xref: /dpdk/drivers/net/e1000/em_ethdev.c (revision a3c8a446)
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_done   = eth_em_rx_descriptor_done;
251 	eth_dev->rx_descriptor_status = eth_em_rx_descriptor_status;
252 	eth_dev->tx_descriptor_status = eth_em_tx_descriptor_status;
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 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
269 
270 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
271 	hw->device_id = pci_dev->id.device_id;
272 	adapter->stopped = 0;
273 
274 	/* For ICH8 support we'll need to map the flash memory BAR */
275 	if (eth_em_dev_is_ich8(hw))
276 		hw->flash_address = (void *)pci_dev->mem_resource[1].addr;
277 
278 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS ||
279 			em_hw_init(hw) != 0) {
280 		PMD_INIT_LOG(ERR, "port_id %d vendorID=0x%x deviceID=0x%x: "
281 			"failed to init HW",
282 			eth_dev->data->port_id, pci_dev->id.vendor_id,
283 			pci_dev->id.device_id);
284 		return -ENODEV;
285 	}
286 
287 	/* Allocate memory for storing MAC addresses */
288 	eth_dev->data->mac_addrs = rte_zmalloc("e1000", RTE_ETHER_ADDR_LEN *
289 			hw->mac.rar_entry_count, 0);
290 	if (eth_dev->data->mac_addrs == NULL) {
291 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
292 			"store MAC addresses",
293 			RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
294 		return -ENOMEM;
295 	}
296 
297 	/* Copy the permanent MAC address */
298 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
299 		eth_dev->data->mac_addrs);
300 
301 	/* initialize the vfta */
302 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
303 
304 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
305 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
306 		     pci_dev->id.device_id);
307 
308 	rte_intr_callback_register(intr_handle,
309 				   eth_em_interrupt_handler, eth_dev);
310 
311 	return 0;
312 }
313 
314 static int
315 eth_em_dev_uninit(struct rte_eth_dev *eth_dev)
316 {
317 	PMD_INIT_FUNC_TRACE();
318 
319 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
320 		return 0;
321 
322 	eth_em_close(eth_dev);
323 
324 	return 0;
325 }
326 
327 static int eth_em_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
328 	struct rte_pci_device *pci_dev)
329 {
330 	return rte_eth_dev_pci_generic_probe(pci_dev,
331 		sizeof(struct e1000_adapter), eth_em_dev_init);
332 }
333 
334 static int eth_em_pci_remove(struct rte_pci_device *pci_dev)
335 {
336 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_em_dev_uninit);
337 }
338 
339 static struct rte_pci_driver rte_em_pmd = {
340 	.id_table = pci_id_em_map,
341 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
342 	.probe = eth_em_pci_probe,
343 	.remove = eth_em_pci_remove,
344 };
345 
346 static int
347 em_hw_init(struct e1000_hw *hw)
348 {
349 	int diag;
350 
351 	diag = hw->mac.ops.init_params(hw);
352 	if (diag != 0) {
353 		PMD_INIT_LOG(ERR, "MAC Initialization Error");
354 		return diag;
355 	}
356 	diag = hw->nvm.ops.init_params(hw);
357 	if (diag != 0) {
358 		PMD_INIT_LOG(ERR, "NVM Initialization Error");
359 		return diag;
360 	}
361 	diag = hw->phy.ops.init_params(hw);
362 	if (diag != 0) {
363 		PMD_INIT_LOG(ERR, "PHY Initialization Error");
364 		return diag;
365 	}
366 	(void) e1000_get_bus_info(hw);
367 
368 	hw->mac.autoneg = 1;
369 	hw->phy.autoneg_wait_to_complete = 0;
370 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
371 
372 	e1000_init_script_state_82541(hw, TRUE);
373 	e1000_set_tbi_compatibility_82543(hw, TRUE);
374 
375 	/* Copper options */
376 	if (hw->phy.media_type == e1000_media_type_copper) {
377 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
378 		hw->phy.disable_polarity_correction = 0;
379 		hw->phy.ms_type = e1000_ms_hw_default;
380 	}
381 
382 	/*
383 	 * Start from a known state, this is important in reading the nvm
384 	 * and mac from that.
385 	 */
386 	e1000_reset_hw(hw);
387 
388 	/* Make sure we have a good EEPROM before we read from it */
389 	if (e1000_validate_nvm_checksum(hw) < 0) {
390 		/*
391 		 * Some PCI-E parts fail the first check due to
392 		 * the link being in sleep state, call it again,
393 		 * if it fails a second time its a real issue.
394 		 */
395 		diag = e1000_validate_nvm_checksum(hw);
396 		if (diag < 0) {
397 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
398 			goto error;
399 		}
400 	}
401 
402 	/* Read the permanent MAC address out of the EEPROM */
403 	diag = e1000_read_mac_addr(hw);
404 	if (diag != 0) {
405 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
406 		goto error;
407 	}
408 
409 	/* Now initialize the hardware */
410 	diag = em_hardware_init(hw);
411 	if (diag != 0) {
412 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
413 		goto error;
414 	}
415 
416 	hw->mac.get_link_status = 1;
417 
418 	/* Indicate SOL/IDER usage */
419 	diag = e1000_check_reset_block(hw);
420 	if (diag < 0) {
421 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to "
422 			"SOL/IDER session");
423 	}
424 	return 0;
425 
426 error:
427 	em_hw_control_release(hw);
428 	return diag;
429 }
430 
431 static int
432 eth_em_configure(struct rte_eth_dev *dev)
433 {
434 	struct e1000_interrupt *intr =
435 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
436 
437 	PMD_INIT_FUNC_TRACE();
438 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
439 
440 	PMD_INIT_FUNC_TRACE();
441 
442 	return 0;
443 }
444 
445 static void
446 em_set_pba(struct e1000_hw *hw)
447 {
448 	uint32_t pba;
449 
450 	/*
451 	 * Packet Buffer Allocation (PBA)
452 	 * Writing PBA sets the receive portion of the buffer
453 	 * the remainder is used for the transmit buffer.
454 	 * Devices before the 82547 had a Packet Buffer of 64K.
455 	 * After the 82547 the buffer was reduced to 40K.
456 	 */
457 	switch (hw->mac.type) {
458 		case e1000_82547:
459 		case e1000_82547_rev_2:
460 		/* 82547: Total Packet Buffer is 40K */
461 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
462 			break;
463 		case e1000_82571:
464 		case e1000_82572:
465 		case e1000_80003es2lan:
466 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
467 			break;
468 		case e1000_82573: /* 82573: Total Packet Buffer is 32K */
469 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
470 			break;
471 		case e1000_82574:
472 		case e1000_82583:
473 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
474 			break;
475 		case e1000_ich8lan:
476 			pba = E1000_PBA_8K;
477 			break;
478 		case e1000_ich9lan:
479 		case e1000_ich10lan:
480 			pba = E1000_PBA_10K;
481 			break;
482 		case e1000_pchlan:
483 		case e1000_pch2lan:
484 		case e1000_pch_lpt:
485 		case e1000_pch_spt:
486 		case e1000_pch_cnp:
487 			pba = E1000_PBA_26K;
488 			break;
489 		default:
490 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
491 	}
492 
493 	E1000_WRITE_REG(hw, E1000_PBA, pba);
494 }
495 
496 static void
497 eth_em_rxtx_control(struct rte_eth_dev *dev,
498 		    bool enable)
499 {
500 	struct e1000_hw *hw =
501 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
502 	uint32_t tctl, rctl;
503 
504 	tctl = E1000_READ_REG(hw, E1000_TCTL);
505 	rctl = E1000_READ_REG(hw, E1000_RCTL);
506 	if (enable) {
507 		/* enable Tx/Rx */
508 		tctl |= E1000_TCTL_EN;
509 		rctl |= E1000_RCTL_EN;
510 	} else {
511 		/* disable Tx/Rx */
512 		tctl &= ~E1000_TCTL_EN;
513 		rctl &= ~E1000_RCTL_EN;
514 	}
515 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
516 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
517 	E1000_WRITE_FLUSH(hw);
518 }
519 
520 static int
521 eth_em_start(struct rte_eth_dev *dev)
522 {
523 	struct e1000_adapter *adapter =
524 		E1000_DEV_PRIVATE(dev->data->dev_private);
525 	struct e1000_hw *hw =
526 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
527 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
528 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
529 	int ret, mask;
530 	uint32_t intr_vector = 0;
531 	uint32_t *speeds;
532 	int num_speeds;
533 	bool autoneg;
534 
535 	PMD_INIT_FUNC_TRACE();
536 
537 	ret = eth_em_stop(dev);
538 	if (ret != 0)
539 		return ret;
540 
541 	e1000_power_up_phy(hw);
542 
543 	/* Set default PBA value */
544 	em_set_pba(hw);
545 
546 	/* Put the address into the Receive Address Array */
547 	e1000_rar_set(hw, hw->mac.addr, 0);
548 
549 	/*
550 	 * With the 82571 adapter, RAR[0] may be overwritten
551 	 * when the other port is reset, we make a duplicate
552 	 * in RAR[14] for that eventuality, this assures
553 	 * the interface continues to function.
554 	 */
555 	if (hw->mac.type == e1000_82571) {
556 		e1000_set_laa_state_82571(hw, TRUE);
557 		e1000_rar_set(hw, hw->mac.addr, E1000_RAR_ENTRIES - 1);
558 	}
559 
560 	/* Initialize the hardware */
561 	if (em_hardware_init(hw)) {
562 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
563 		return -EIO;
564 	}
565 
566 	E1000_WRITE_REG(hw, E1000_VET, RTE_ETHER_TYPE_VLAN);
567 
568 	/* Configure for OS presence */
569 	em_init_manageability(hw);
570 
571 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
572 		intr_vector = dev->data->nb_rx_queues;
573 		if (rte_intr_efd_enable(intr_handle, intr_vector))
574 			return -1;
575 	}
576 
577 	if (rte_intr_dp_is_en(intr_handle)) {
578 		intr_handle->intr_vec =
579 			rte_zmalloc("intr_vec",
580 					dev->data->nb_rx_queues * sizeof(int), 0);
581 		if (intr_handle->intr_vec == NULL) {
582 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
583 						" intr_vec", dev->data->nb_rx_queues);
584 			return -ENOMEM;
585 		}
586 
587 		/* enable rx interrupt */
588 		em_rxq_intr_enable(hw);
589 	}
590 
591 	eth_em_tx_init(dev);
592 
593 	ret = eth_em_rx_init(dev);
594 	if (ret) {
595 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
596 		em_dev_clear_queues(dev);
597 		return ret;
598 	}
599 
600 	e1000_clear_hw_cntrs_base_generic(hw);
601 
602 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
603 			ETH_VLAN_EXTEND_MASK;
604 	ret = eth_em_vlan_offload_set(dev, mask);
605 	if (ret) {
606 		PMD_INIT_LOG(ERR, "Unable to update vlan offload");
607 		em_dev_clear_queues(dev);
608 		return ret;
609 	}
610 
611 	/* Set Interrupt Throttling Rate to maximum allowed value. */
612 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
613 
614 	/* Setup link speed and duplex */
615 	speeds = &dev->data->dev_conf.link_speeds;
616 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
617 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
618 		hw->mac.autoneg = 1;
619 	} else {
620 		num_speeds = 0;
621 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
622 
623 		/* Reset */
624 		hw->phy.autoneg_advertised = 0;
625 
626 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
627 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
628 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
629 			num_speeds = -1;
630 			goto error_invalid_config;
631 		}
632 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
633 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
634 			num_speeds++;
635 		}
636 		if (*speeds & ETH_LINK_SPEED_10M) {
637 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
638 			num_speeds++;
639 		}
640 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
641 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
642 			num_speeds++;
643 		}
644 		if (*speeds & ETH_LINK_SPEED_100M) {
645 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
646 			num_speeds++;
647 		}
648 		if (*speeds & ETH_LINK_SPEED_1G) {
649 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
650 			num_speeds++;
651 		}
652 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
653 			goto error_invalid_config;
654 
655 		/* Set/reset the mac.autoneg based on the link speed,
656 		 * fixed or not
657 		 */
658 		if (!autoneg) {
659 			hw->mac.autoneg = 0;
660 			hw->mac.forced_speed_duplex =
661 					hw->phy.autoneg_advertised;
662 		} else {
663 			hw->mac.autoneg = 1;
664 		}
665 	}
666 
667 	e1000_setup_link(hw);
668 
669 	if (rte_intr_allow_others(intr_handle)) {
670 		/* check if lsc interrupt is enabled */
671 		if (dev->data->dev_conf.intr_conf.lsc != 0) {
672 			ret = eth_em_interrupt_setup(dev);
673 			if (ret) {
674 				PMD_INIT_LOG(ERR, "Unable to setup interrupts");
675 				em_dev_clear_queues(dev);
676 				return ret;
677 			}
678 		}
679 	} else {
680 		rte_intr_callback_unregister(intr_handle,
681 						eth_em_interrupt_handler,
682 						(void *)dev);
683 		if (dev->data->dev_conf.intr_conf.lsc != 0)
684 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
685 				     " no intr multiplexn");
686 	}
687 	/* check if rxq interrupt is enabled */
688 	if (dev->data->dev_conf.intr_conf.rxq != 0)
689 		eth_em_rxq_interrupt_setup(dev);
690 
691 	rte_intr_enable(intr_handle);
692 
693 	adapter->stopped = 0;
694 
695 	eth_em_rxtx_control(dev, true);
696 	eth_em_link_update(dev, 0);
697 
698 	PMD_INIT_LOG(DEBUG, "<<");
699 
700 	return 0;
701 
702 error_invalid_config:
703 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
704 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
705 	em_dev_clear_queues(dev);
706 	return -EINVAL;
707 }
708 
709 /*********************************************************************
710  *
711  *  This routine disables all traffic on the adapter by issuing a
712  *  global reset on the MAC.
713  *
714  **********************************************************************/
715 static int
716 eth_em_stop(struct rte_eth_dev *dev)
717 {
718 	struct rte_eth_link link;
719 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
720 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
721 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
722 
723 	dev->data->dev_started = 0;
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 	return 0;
761 }
762 
763 static int
764 eth_em_close(struct rte_eth_dev *dev)
765 {
766 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
767 	struct e1000_adapter *adapter =
768 		E1000_DEV_PRIVATE(dev->data->dev_private);
769 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
770 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
771 	int ret;
772 
773 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
774 		return 0;
775 
776 	ret = eth_em_stop(dev);
777 	adapter->stopped = 1;
778 	em_dev_free_queues(dev);
779 	e1000_phy_hw_reset(hw);
780 	em_release_manageability(hw);
781 	em_hw_control_release(hw);
782 
783 	/* disable uio intr before callback unregister */
784 	rte_intr_disable(intr_handle);
785 	rte_intr_callback_unregister(intr_handle,
786 				     eth_em_interrupt_handler, dev);
787 
788 	return ret;
789 }
790 
791 static int
792 em_get_rx_buffer_size(struct e1000_hw *hw)
793 {
794 	uint32_t rx_buf_size;
795 
796 	rx_buf_size = ((E1000_READ_REG(hw, E1000_PBA) & UINT16_MAX) << 10);
797 	return rx_buf_size;
798 }
799 
800 /*********************************************************************
801  *
802  *  Initialize the hardware
803  *
804  **********************************************************************/
805 static int
806 em_hardware_init(struct e1000_hw *hw)
807 {
808 	uint32_t rx_buf_size;
809 	int diag;
810 
811 	/* Issue a global reset */
812 	e1000_reset_hw(hw);
813 
814 	/* Let the firmware know the OS is in control */
815 	em_hw_control_acquire(hw);
816 
817 	/*
818 	 * These parameters control the automatic generation (Tx) and
819 	 * response (Rx) to Ethernet PAUSE frames.
820 	 * - High water mark should allow for at least two standard size (1518)
821 	 *   frames to be received after sending an XOFF.
822 	 * - Low water mark works best when it is very near the high water mark.
823 	 *   This allows the receiver to restart by sending XON when it has
824 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
825 	 *   restart after one full frame is pulled from the buffer. There
826 	 *   could be several smaller frames in the buffer and if so they will
827 	 *   not trigger the XON until their total number reduces the buffer
828 	 *   by 1500.
829 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
830 	 */
831 	rx_buf_size = em_get_rx_buffer_size(hw);
832 
833 	hw->fc.high_water = rx_buf_size -
834 		PMD_ROUNDUP(RTE_ETHER_MAX_LEN * 2, 1024);
835 	hw->fc.low_water = hw->fc.high_water - 1500;
836 
837 	if (hw->mac.type == e1000_80003es2lan)
838 		hw->fc.pause_time = UINT16_MAX;
839 	else
840 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
841 
842 	hw->fc.send_xon = 1;
843 
844 	/* Set Flow control, use the tunable location if sane */
845 	if (em_fc_setting <= e1000_fc_full)
846 		hw->fc.requested_mode = em_fc_setting;
847 	else
848 		hw->fc.requested_mode = e1000_fc_none;
849 
850 	/* Workaround: no TX flow ctrl for PCH */
851 	if (hw->mac.type == e1000_pchlan)
852 		hw->fc.requested_mode = e1000_fc_rx_pause;
853 
854 	/* Override - settings for PCH2LAN, ya its magic :) */
855 	if (hw->mac.type == e1000_pch2lan) {
856 		hw->fc.high_water = 0x5C20;
857 		hw->fc.low_water = 0x5048;
858 		hw->fc.pause_time = 0x0650;
859 		hw->fc.refresh_time = 0x0400;
860 	} else if (hw->mac.type == e1000_pch_lpt ||
861 		   hw->mac.type == e1000_pch_spt ||
862 		   hw->mac.type == e1000_pch_cnp) {
863 		hw->fc.requested_mode = e1000_fc_full;
864 	}
865 
866 	diag = e1000_init_hw(hw);
867 	if (diag < 0)
868 		return diag;
869 	e1000_check_for_link(hw);
870 	return 0;
871 }
872 
873 /* This function is based on em_update_stats_counters() in e1000/if_em.c */
874 static int
875 eth_em_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
876 {
877 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
878 	struct e1000_hw_stats *stats =
879 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
880 	int pause_frames;
881 
882 	if(hw->phy.media_type == e1000_media_type_copper ||
883 			(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
884 		stats->symerrs += E1000_READ_REG(hw,E1000_SYMERRS);
885 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
886 	}
887 
888 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
889 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
890 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
891 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
892 
893 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
894 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
895 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
896 	stats->dc += E1000_READ_REG(hw, E1000_DC);
897 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
898 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
899 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
900 
901 	/*
902 	 * For watchdog management we need to know if we have been
903 	 * paused during the last interval, so capture that here.
904 	 */
905 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
906 	stats->xoffrxc += pause_frames;
907 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
908 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
909 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
910 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
911 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
912 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
913 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
914 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
915 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
916 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
917 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
918 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
919 
920 	/*
921 	 * For the 64-bit byte counters the low dword must be read first.
922 	 * Both registers clear on the read of the high dword.
923 	 */
924 
925 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
926 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
927 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
928 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
929 
930 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
931 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
932 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
933 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
934 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
935 
936 	stats->tor += E1000_READ_REG(hw, E1000_TORH);
937 	stats->tot += E1000_READ_REG(hw, E1000_TOTH);
938 
939 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
940 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
941 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
942 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
943 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
944 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
945 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
946 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
947 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
948 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
949 
950 	/* Interrupt Counts */
951 
952 	if (hw->mac.type >= e1000_82571) {
953 		stats->iac += E1000_READ_REG(hw, E1000_IAC);
954 		stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
955 		stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
956 		stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
957 		stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
958 		stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
959 		stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
960 		stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
961 		stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
962 	}
963 
964 	if (hw->mac.type >= e1000_82543) {
965 		stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
966 		stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
967 		stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
968 		stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
969 		stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
970 		stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
971 	}
972 
973 	if (rte_stats == NULL)
974 		return -EINVAL;
975 
976 	/* Rx Errors */
977 	rte_stats->imissed = stats->mpc;
978 	rte_stats->ierrors = stats->crcerrs + stats->rlec +
979 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
980 
981 	/* Tx Errors */
982 	rte_stats->oerrors = stats->ecol + stats->latecol;
983 
984 	rte_stats->ipackets = stats->gprc;
985 	rte_stats->opackets = stats->gptc;
986 	rte_stats->ibytes   = stats->gorc;
987 	rte_stats->obytes   = stats->gotc;
988 	return 0;
989 }
990 
991 static int
992 eth_em_stats_reset(struct rte_eth_dev *dev)
993 {
994 	struct e1000_hw_stats *hw_stats =
995 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
996 
997 	/* HW registers are cleared on read */
998 	eth_em_stats_get(dev, NULL);
999 
1000 	/* Reset software totals */
1001 	memset(hw_stats, 0, sizeof(*hw_stats));
1002 
1003 	return 0;
1004 }
1005 
1006 static int
1007 eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1008 {
1009 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1010 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1011 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1012 
1013 	em_rxq_intr_enable(hw);
1014 	rte_intr_ack(intr_handle);
1015 
1016 	return 0;
1017 }
1018 
1019 static int
1020 eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1021 {
1022 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1023 
1024 	em_rxq_intr_disable(hw);
1025 
1026 	return 0;
1027 }
1028 
1029 uint32_t
1030 em_get_max_pktlen(struct rte_eth_dev *dev)
1031 {
1032 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1033 
1034 	switch (hw->mac.type) {
1035 	case e1000_82571:
1036 	case e1000_82572:
1037 	case e1000_ich9lan:
1038 	case e1000_ich10lan:
1039 	case e1000_pch2lan:
1040 	case e1000_pch_lpt:
1041 	case e1000_pch_spt:
1042 	case e1000_pch_cnp:
1043 	case e1000_82574:
1044 	case e1000_80003es2lan: /* 9K Jumbo Frame size */
1045 	case e1000_82583:
1046 		return 0x2412;
1047 	case e1000_pchlan:
1048 		return 0x1000;
1049 	/* Adapters that do not support jumbo frames */
1050 	case e1000_ich8lan:
1051 		return RTE_ETHER_MAX_LEN;
1052 	default:
1053 		return MAX_JUMBO_FRAME_SIZE;
1054 	}
1055 }
1056 
1057 static int
1058 eth_em_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1059 {
1060 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1061 
1062 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1063 	dev_info->max_rx_pktlen = em_get_max_pktlen(dev);
1064 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1065 
1066 	/*
1067 	 * Starting with 631xESB hw supports 2 TX/RX queues per port.
1068 	 * Unfortunatelly, all these nics have just one TX context.
1069 	 * So we have few choises for TX:
1070 	 * - Use just one TX queue.
1071 	 * - Allow cksum offload only for one TX queue.
1072 	 * - Don't allow TX cksum offload at all.
1073 	 * For now, option #1 was chosen.
1074 	 * To use second RX queue we have to use extended RX descriptor
1075 	 * (Multiple Receive Queues are mutually exclusive with UDP
1076 	 * fragmentation and are not supported when a legacy receive
1077 	 * descriptor format is used).
1078 	 * Which means separate RX routinies - as legacy nics (82540, 82545)
1079 	 * don't support extended RXD.
1080 	 * To avoid it we support just one RX queue for now (no RSS).
1081 	 */
1082 
1083 	dev_info->max_rx_queues = 1;
1084 	dev_info->max_tx_queues = 1;
1085 
1086 	dev_info->rx_queue_offload_capa = em_get_rx_queue_offloads_capa(dev);
1087 	dev_info->rx_offload_capa = em_get_rx_port_offloads_capa(dev) |
1088 				    dev_info->rx_queue_offload_capa;
1089 	dev_info->tx_queue_offload_capa = em_get_tx_queue_offloads_capa(dev);
1090 	dev_info->tx_offload_capa = em_get_tx_port_offloads_capa(dev) |
1091 				    dev_info->tx_queue_offload_capa;
1092 
1093 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1094 		.nb_max = E1000_MAX_RING_DESC,
1095 		.nb_min = E1000_MIN_RING_DESC,
1096 		.nb_align = EM_RXD_ALIGN,
1097 	};
1098 
1099 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1100 		.nb_max = E1000_MAX_RING_DESC,
1101 		.nb_min = E1000_MIN_RING_DESC,
1102 		.nb_align = EM_TXD_ALIGN,
1103 		.nb_seg_max = EM_TX_MAX_SEG,
1104 		.nb_mtu_seg_max = EM_TX_MAX_MTU_SEG,
1105 	};
1106 
1107 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1108 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1109 			ETH_LINK_SPEED_1G;
1110 
1111 	/* Preferred queue parameters */
1112 	dev_info->default_rxportconf.nb_queues = 1;
1113 	dev_info->default_txportconf.nb_queues = 1;
1114 	dev_info->default_txportconf.ring_size = 256;
1115 	dev_info->default_rxportconf.ring_size = 256;
1116 
1117 	return 0;
1118 }
1119 
1120 /* return 0 means link status changed, -1 means not changed */
1121 static int
1122 eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1123 {
1124 	struct e1000_hw *hw =
1125 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1126 	struct rte_eth_link link;
1127 	int link_up, count;
1128 
1129 	link_up = 0;
1130 	hw->mac.get_link_status = 1;
1131 
1132 	/* possible wait-to-complete in up to 9 seconds */
1133 	for (count = 0; count < EM_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
1134 		/* Read the real link status */
1135 		switch (hw->phy.media_type) {
1136 		case e1000_media_type_copper:
1137 			/* Do the work to read phy */
1138 			e1000_check_for_link(hw);
1139 			link_up = !hw->mac.get_link_status;
1140 			break;
1141 
1142 		case e1000_media_type_fiber:
1143 			e1000_check_for_link(hw);
1144 			link_up = (E1000_READ_REG(hw, E1000_STATUS) &
1145 					E1000_STATUS_LU);
1146 			break;
1147 
1148 		case e1000_media_type_internal_serdes:
1149 			e1000_check_for_link(hw);
1150 			link_up = hw->mac.serdes_has_link;
1151 			break;
1152 
1153 		default:
1154 			break;
1155 		}
1156 		if (link_up || wait_to_complete == 0)
1157 			break;
1158 		rte_delay_ms(EM_LINK_UPDATE_CHECK_INTERVAL);
1159 	}
1160 	memset(&link, 0, sizeof(link));
1161 
1162 	/* Now we check if a transition has happened */
1163 	if (link_up) {
1164 		uint16_t duplex, speed;
1165 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
1166 		link.link_duplex = (duplex == FULL_DUPLEX) ?
1167 				ETH_LINK_FULL_DUPLEX :
1168 				ETH_LINK_HALF_DUPLEX;
1169 		link.link_speed = speed;
1170 		link.link_status = ETH_LINK_UP;
1171 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1172 				ETH_LINK_SPEED_FIXED);
1173 	} else {
1174 		link.link_speed = ETH_SPEED_NUM_NONE;
1175 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
1176 		link.link_status = ETH_LINK_DOWN;
1177 		link.link_autoneg = ETH_LINK_FIXED;
1178 	}
1179 
1180 	return rte_eth_linkstatus_set(dev, &link);
1181 }
1182 
1183 /*
1184  * em_hw_control_acquire sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
1185  * For ASF and Pass Through versions of f/w this means
1186  * that the driver is loaded. For AMT version type f/w
1187  * this means that the network i/f is open.
1188  */
1189 static void
1190 em_hw_control_acquire(struct e1000_hw *hw)
1191 {
1192 	uint32_t ctrl_ext, swsm;
1193 
1194 	/* Let firmware know the driver has taken over */
1195 	if (hw->mac.type == e1000_82573) {
1196 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1197 		E1000_WRITE_REG(hw, E1000_SWSM, swsm | E1000_SWSM_DRV_LOAD);
1198 
1199 	} else {
1200 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1201 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1202 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1203 	}
1204 }
1205 
1206 /*
1207  * em_hw_control_release resets {CTRL_EXTT|FWSM}:DRV_LOAD bit.
1208  * For ASF and Pass Through versions of f/w this means that the
1209  * driver is no longer loaded. For AMT versions of the
1210  * f/w this means that the network i/f is closed.
1211  */
1212 static void
1213 em_hw_control_release(struct e1000_hw *hw)
1214 {
1215 	uint32_t ctrl_ext, swsm;
1216 
1217 	/* Let firmware taken over control of h/w */
1218 	if (hw->mac.type == e1000_82573) {
1219 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1220 		E1000_WRITE_REG(hw, E1000_SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
1221 	} else {
1222 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1223 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1224 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1225 	}
1226 }
1227 
1228 /*
1229  * Bit of a misnomer, what this really means is
1230  * to enable OS management of the system... aka
1231  * to disable special hardware management features.
1232  */
1233 static void
1234 em_init_manageability(struct e1000_hw *hw)
1235 {
1236 	if (e1000_enable_mng_pass_thru(hw)) {
1237 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
1238 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
1239 
1240 		/* disable hardware interception of ARP */
1241 		manc &= ~(E1000_MANC_ARP_EN);
1242 
1243 		/* enable receiving management packets to the host */
1244 		manc |= E1000_MANC_EN_MNG2HOST;
1245 		manc2h |= 1 << 5;  /* Mng Port 623 */
1246 		manc2h |= 1 << 6;  /* Mng Port 664 */
1247 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
1248 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1249 	}
1250 }
1251 
1252 /*
1253  * Give control back to hardware management
1254  * controller if there is one.
1255  */
1256 static void
1257 em_release_manageability(struct e1000_hw *hw)
1258 {
1259 	uint32_t manc;
1260 
1261 	if (e1000_enable_mng_pass_thru(hw)) {
1262 		manc = E1000_READ_REG(hw, E1000_MANC);
1263 
1264 		/* re-enable hardware interception of ARP */
1265 		manc |= E1000_MANC_ARP_EN;
1266 		manc &= ~E1000_MANC_EN_MNG2HOST;
1267 
1268 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1269 	}
1270 }
1271 
1272 static int
1273 eth_em_promiscuous_enable(struct rte_eth_dev *dev)
1274 {
1275 	struct e1000_hw *hw =
1276 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1277 	uint32_t rctl;
1278 
1279 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1280 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1281 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1282 
1283 	return 0;
1284 }
1285 
1286 static int
1287 eth_em_promiscuous_disable(struct rte_eth_dev *dev)
1288 {
1289 	struct e1000_hw *hw =
1290 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1291 	uint32_t rctl;
1292 
1293 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1294 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_SBP);
1295 	if (dev->data->all_multicast == 1)
1296 		rctl |= E1000_RCTL_MPE;
1297 	else
1298 		rctl &= (~E1000_RCTL_MPE);
1299 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1300 
1301 	return 0;
1302 }
1303 
1304 static int
1305 eth_em_allmulticast_enable(struct rte_eth_dev *dev)
1306 {
1307 	struct e1000_hw *hw =
1308 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1309 	uint32_t rctl;
1310 
1311 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1312 	rctl |= E1000_RCTL_MPE;
1313 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1314 
1315 	return 0;
1316 }
1317 
1318 static int
1319 eth_em_allmulticast_disable(struct rte_eth_dev *dev)
1320 {
1321 	struct e1000_hw *hw =
1322 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1323 	uint32_t rctl;
1324 
1325 	if (dev->data->promiscuous == 1)
1326 		return 0; /* must remain in all_multicast mode */
1327 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1328 	rctl &= (~E1000_RCTL_MPE);
1329 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1330 
1331 	return 0;
1332 }
1333 
1334 static int
1335 eth_em_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1336 {
1337 	struct e1000_hw *hw =
1338 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1339 	struct e1000_vfta * shadow_vfta =
1340 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1341 	uint32_t vfta;
1342 	uint32_t vid_idx;
1343 	uint32_t vid_bit;
1344 
1345 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
1346 			      E1000_VFTA_ENTRY_MASK);
1347 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
1348 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
1349 	if (on)
1350 		vfta |= vid_bit;
1351 	else
1352 		vfta &= ~vid_bit;
1353 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
1354 
1355 	/* update local VFTA copy */
1356 	shadow_vfta->vfta[vid_idx] = vfta;
1357 
1358 	return 0;
1359 }
1360 
1361 static void
1362 em_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1363 {
1364 	struct e1000_hw *hw =
1365 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1366 	uint32_t reg;
1367 
1368 	/* Filter Table Disable */
1369 	reg = E1000_READ_REG(hw, E1000_RCTL);
1370 	reg &= ~E1000_RCTL_CFIEN;
1371 	reg &= ~E1000_RCTL_VFE;
1372 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1373 }
1374 
1375 static void
1376 em_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1377 {
1378 	struct e1000_hw *hw =
1379 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1380 	struct e1000_vfta * shadow_vfta =
1381 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1382 	uint32_t reg;
1383 	int i;
1384 
1385 	/* Filter Table Enable, CFI not used for packet acceptance */
1386 	reg = E1000_READ_REG(hw, E1000_RCTL);
1387 	reg &= ~E1000_RCTL_CFIEN;
1388 	reg |= E1000_RCTL_VFE;
1389 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1390 
1391 	/* restore vfta from local copy */
1392 	for (i = 0; i < IGB_VFTA_SIZE; i++)
1393 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
1394 }
1395 
1396 static void
1397 em_vlan_hw_strip_disable(struct rte_eth_dev *dev)
1398 {
1399 	struct e1000_hw *hw =
1400 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1401 	uint32_t reg;
1402 
1403 	/* VLAN Mode Disable */
1404 	reg = E1000_READ_REG(hw, E1000_CTRL);
1405 	reg &= ~E1000_CTRL_VME;
1406 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1407 
1408 }
1409 
1410 static void
1411 em_vlan_hw_strip_enable(struct rte_eth_dev *dev)
1412 {
1413 	struct e1000_hw *hw =
1414 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1415 	uint32_t reg;
1416 
1417 	/* VLAN Mode Enable */
1418 	reg = E1000_READ_REG(hw, E1000_CTRL);
1419 	reg |= E1000_CTRL_VME;
1420 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1421 }
1422 
1423 static int
1424 eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1425 {
1426 	struct rte_eth_rxmode *rxmode;
1427 
1428 	rxmode = &dev->data->dev_conf.rxmode;
1429 	if(mask & ETH_VLAN_STRIP_MASK){
1430 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1431 			em_vlan_hw_strip_enable(dev);
1432 		else
1433 			em_vlan_hw_strip_disable(dev);
1434 	}
1435 
1436 	if(mask & ETH_VLAN_FILTER_MASK){
1437 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1438 			em_vlan_hw_filter_enable(dev);
1439 		else
1440 			em_vlan_hw_filter_disable(dev);
1441 	}
1442 
1443 	return 0;
1444 }
1445 
1446 /*
1447  * It enables the interrupt mask and then enable the interrupt.
1448  *
1449  * @param dev
1450  *  Pointer to struct rte_eth_dev.
1451  *
1452  * @return
1453  *  - On success, zero.
1454  *  - On failure, a negative value.
1455  */
1456 static int
1457 eth_em_interrupt_setup(struct rte_eth_dev *dev)
1458 {
1459 	uint32_t regval;
1460 	struct e1000_hw *hw =
1461 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1462 
1463 	/* clear interrupt */
1464 	E1000_READ_REG(hw, E1000_ICR);
1465 	regval = E1000_READ_REG(hw, E1000_IMS);
1466 	E1000_WRITE_REG(hw, E1000_IMS,
1467 			regval | E1000_ICR_LSC | E1000_ICR_OTHER);
1468 	return 0;
1469 }
1470 
1471 /*
1472  * It clears the interrupt causes and enables the interrupt.
1473  * It will be called once only during nic initialized.
1474  *
1475  * @param dev
1476  *  Pointer to struct rte_eth_dev.
1477  *
1478  * @return
1479  *  - On success, zero.
1480  *  - On failure, a negative value.
1481  */
1482 static int
1483 eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev)
1484 {
1485 	struct e1000_hw *hw =
1486 	E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1487 
1488 	E1000_READ_REG(hw, E1000_ICR);
1489 	em_rxq_intr_enable(hw);
1490 	return 0;
1491 }
1492 
1493 /*
1494  * It enable receive packet interrupt.
1495  * @param hw
1496  * Pointer to struct e1000_hw
1497  *
1498  * @return
1499  */
1500 static void
1501 em_rxq_intr_enable(struct e1000_hw *hw)
1502 {
1503 	E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_RXT0);
1504 	E1000_WRITE_FLUSH(hw);
1505 }
1506 
1507 /*
1508  * It disabled lsc interrupt.
1509  * @param hw
1510  * Pointer to struct e1000_hw
1511  *
1512  * @return
1513  */
1514 static void
1515 em_lsc_intr_disable(struct e1000_hw *hw)
1516 {
1517 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC | E1000_IMS_OTHER);
1518 	E1000_WRITE_FLUSH(hw);
1519 }
1520 
1521 /*
1522  * It disabled receive packet interrupt.
1523  * @param hw
1524  * Pointer to struct e1000_hw
1525  *
1526  * @return
1527  */
1528 static void
1529 em_rxq_intr_disable(struct e1000_hw *hw)
1530 {
1531 	E1000_READ_REG(hw, E1000_ICR);
1532 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_RXT0);
1533 	E1000_WRITE_FLUSH(hw);
1534 }
1535 
1536 /*
1537  * It reads ICR and gets interrupt causes, check it and set a bit flag
1538  * to update link status.
1539  *
1540  * @param dev
1541  *  Pointer to struct rte_eth_dev.
1542  *
1543  * @return
1544  *  - On success, zero.
1545  *  - On failure, a negative value.
1546  */
1547 static int
1548 eth_em_interrupt_get_status(struct rte_eth_dev *dev)
1549 {
1550 	uint32_t icr;
1551 	struct e1000_hw *hw =
1552 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1553 	struct e1000_interrupt *intr =
1554 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1555 
1556 	/* read-on-clear nic registers here */
1557 	icr = E1000_READ_REG(hw, E1000_ICR);
1558 	if (icr & E1000_ICR_LSC) {
1559 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1560 	}
1561 
1562 	return 0;
1563 }
1564 
1565 /*
1566  * It executes link_update after knowing an interrupt is prsent.
1567  *
1568  * @param dev
1569  *  Pointer to struct rte_eth_dev.
1570  *
1571  * @return
1572  *  - On success, zero.
1573  *  - On failure, a negative value.
1574  */
1575 static int
1576 eth_em_interrupt_action(struct rte_eth_dev *dev,
1577 			struct rte_intr_handle *intr_handle)
1578 {
1579 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1580 	struct e1000_hw *hw =
1581 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1582 	struct e1000_interrupt *intr =
1583 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1584 	struct rte_eth_link link;
1585 	int ret;
1586 
1587 	if (!(intr->flags & E1000_FLAG_NEED_LINK_UPDATE))
1588 		return -1;
1589 
1590 	intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
1591 	rte_intr_ack(intr_handle);
1592 
1593 	/* set get_link_status to check register later */
1594 	hw->mac.get_link_status = 1;
1595 	ret = eth_em_link_update(dev, 0);
1596 
1597 	/* check if link has changed */
1598 	if (ret < 0)
1599 		return 0;
1600 
1601 	rte_eth_linkstatus_get(dev, &link);
1602 
1603 	if (link.link_status) {
1604 		PMD_INIT_LOG(INFO, " Port %d: Link Up - speed %u Mbps - %s",
1605 			     dev->data->port_id, link.link_speed,
1606 			     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
1607 			     "full-duplex" : "half-duplex");
1608 	} else {
1609 		PMD_INIT_LOG(INFO, " Port %d: Link Down", dev->data->port_id);
1610 	}
1611 	PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
1612 		     pci_dev->addr.domain, pci_dev->addr.bus,
1613 		     pci_dev->addr.devid, pci_dev->addr.function);
1614 
1615 	return 0;
1616 }
1617 
1618 /**
1619  * Interrupt handler which shall be registered at first.
1620  *
1621  * @param handle
1622  *  Pointer to interrupt handle.
1623  * @param param
1624  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1625  *
1626  * @return
1627  *  void
1628  */
1629 static void
1630 eth_em_interrupt_handler(void *param)
1631 {
1632 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1633 
1634 	eth_em_interrupt_get_status(dev);
1635 	eth_em_interrupt_action(dev, dev->intr_handle);
1636 	rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
1637 }
1638 
1639 static int
1640 eth_em_led_on(struct rte_eth_dev *dev)
1641 {
1642 	struct e1000_hw *hw;
1643 
1644 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1645 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1646 }
1647 
1648 static int
1649 eth_em_led_off(struct rte_eth_dev *dev)
1650 {
1651 	struct e1000_hw *hw;
1652 
1653 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1654 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1655 }
1656 
1657 static int
1658 eth_em_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1659 {
1660 	struct e1000_hw *hw;
1661 	uint32_t ctrl;
1662 	int tx_pause;
1663 	int rx_pause;
1664 
1665 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1666 	fc_conf->pause_time = hw->fc.pause_time;
1667 	fc_conf->high_water = hw->fc.high_water;
1668 	fc_conf->low_water = hw->fc.low_water;
1669 	fc_conf->send_xon = hw->fc.send_xon;
1670 	fc_conf->autoneg = hw->mac.autoneg;
1671 
1672 	/*
1673 	 * Return rx_pause and tx_pause status according to actual setting of
1674 	 * the TFCE and RFCE bits in the CTRL register.
1675 	 */
1676 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
1677 	if (ctrl & E1000_CTRL_TFCE)
1678 		tx_pause = 1;
1679 	else
1680 		tx_pause = 0;
1681 
1682 	if (ctrl & E1000_CTRL_RFCE)
1683 		rx_pause = 1;
1684 	else
1685 		rx_pause = 0;
1686 
1687 	if (rx_pause && tx_pause)
1688 		fc_conf->mode = RTE_FC_FULL;
1689 	else if (rx_pause)
1690 		fc_conf->mode = RTE_FC_RX_PAUSE;
1691 	else if (tx_pause)
1692 		fc_conf->mode = RTE_FC_TX_PAUSE;
1693 	else
1694 		fc_conf->mode = RTE_FC_NONE;
1695 
1696 	return 0;
1697 }
1698 
1699 static int
1700 eth_em_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1701 {
1702 	struct e1000_hw *hw;
1703 	int err;
1704 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
1705 		e1000_fc_none,
1706 		e1000_fc_rx_pause,
1707 		e1000_fc_tx_pause,
1708 		e1000_fc_full
1709 	};
1710 	uint32_t rx_buf_size;
1711 	uint32_t max_high_water;
1712 	uint32_t rctl;
1713 
1714 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1715 	if (fc_conf->autoneg != hw->mac.autoneg)
1716 		return -ENOTSUP;
1717 	rx_buf_size = em_get_rx_buffer_size(hw);
1718 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
1719 
1720 	/* At least reserve one Ethernet frame for watermark */
1721 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
1722 	if ((fc_conf->high_water > max_high_water) ||
1723 	    (fc_conf->high_water < fc_conf->low_water)) {
1724 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
1725 		PMD_INIT_LOG(ERR, "high water must <= 0x%x", max_high_water);
1726 		return -EINVAL;
1727 	}
1728 
1729 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
1730 	hw->fc.pause_time     = fc_conf->pause_time;
1731 	hw->fc.high_water     = fc_conf->high_water;
1732 	hw->fc.low_water      = fc_conf->low_water;
1733 	hw->fc.send_xon	      = fc_conf->send_xon;
1734 
1735 	err = e1000_setup_link_generic(hw);
1736 	if (err == E1000_SUCCESS) {
1737 
1738 		/* check if we want to forward MAC frames - driver doesn't have native
1739 		 * capability to do that, so we'll write the registers ourselves */
1740 
1741 		rctl = E1000_READ_REG(hw, E1000_RCTL);
1742 
1743 		/* set or clear MFLCN.PMCF bit depending on configuration */
1744 		if (fc_conf->mac_ctrl_frame_fwd != 0)
1745 			rctl |= E1000_RCTL_PMCF;
1746 		else
1747 			rctl &= ~E1000_RCTL_PMCF;
1748 
1749 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1750 		E1000_WRITE_FLUSH(hw);
1751 
1752 		return 0;
1753 	}
1754 
1755 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
1756 	return -EIO;
1757 }
1758 
1759 static int
1760 eth_em_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
1761 		uint32_t index, __rte_unused uint32_t pool)
1762 {
1763 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1764 
1765 	return e1000_rar_set(hw, mac_addr->addr_bytes, index);
1766 }
1767 
1768 static void
1769 eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1770 {
1771 	uint8_t addr[RTE_ETHER_ADDR_LEN];
1772 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1773 
1774 	memset(addr, 0, sizeof(addr));
1775 
1776 	e1000_rar_set(hw, addr, index);
1777 }
1778 
1779 static int
1780 eth_em_default_mac_addr_set(struct rte_eth_dev *dev,
1781 			    struct rte_ether_addr *addr)
1782 {
1783 	eth_em_rar_clear(dev, 0);
1784 
1785 	return eth_em_rar_set(dev, (void *)addr, 0, 0);
1786 }
1787 
1788 static int
1789 eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1790 {
1791 	struct rte_eth_dev_info dev_info;
1792 	struct e1000_hw *hw;
1793 	uint32_t frame_size;
1794 	uint32_t rctl;
1795 	int ret;
1796 
1797 	ret = eth_em_infos_get(dev, &dev_info);
1798 	if (ret != 0)
1799 		return ret;
1800 
1801 	frame_size = mtu + E1000_ETH_OVERHEAD;
1802 
1803 	/* check that mtu is within the allowed range */
1804 	if (mtu < RTE_ETHER_MIN_MTU || frame_size > dev_info.max_rx_pktlen)
1805 		return -EINVAL;
1806 
1807 	/*
1808 	 * If device is started, refuse mtu that requires the support of
1809 	 * scattered packets when this feature has not been enabled before.
1810 	 */
1811 	if (dev->data->dev_started && !dev->data->scattered_rx &&
1812 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
1813 		PMD_INIT_LOG(ERR, "Stop port first.");
1814 		return -EINVAL;
1815 	}
1816 
1817 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1818 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1819 
1820 	/* switch to jumbo mode if needed */
1821 	if (frame_size > E1000_ETH_MAX_LEN) {
1822 		dev->data->dev_conf.rxmode.offloads |=
1823 			DEV_RX_OFFLOAD_JUMBO_FRAME;
1824 		rctl |= E1000_RCTL_LPE;
1825 	} else {
1826 		dev->data->dev_conf.rxmode.offloads &=
1827 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
1828 		rctl &= ~E1000_RCTL_LPE;
1829 	}
1830 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1831 
1832 	/* update max frame size */
1833 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1834 	return 0;
1835 }
1836 
1837 static int
1838 eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
1839 			struct rte_ether_addr *mc_addr_set,
1840 			uint32_t nb_mc_addr)
1841 {
1842 	struct e1000_hw *hw;
1843 
1844 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1845 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1846 	return 0;
1847 }
1848 
1849 RTE_PMD_REGISTER_PCI(net_e1000_em, rte_em_pmd);
1850 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_em, pci_id_em_map);
1851 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_em, "* igb_uio | uio_pci_generic | vfio-pci");
1852