xref: /dpdk/drivers/net/e1000/em_ethdev.c (revision ba9de463)
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_log.h>
15 #include <rte_debug.h>
16 #include <rte_pci.h>
17 #include <rte_bus_pci.h>
18 #include <rte_ether.h>
19 #include <rte_ethdev.h>
20 #include <rte_ethdev_pci.h>
21 #include <rte_memory.h>
22 #include <rte_eal.h>
23 #include <rte_atomic.h>
24 #include <rte_malloc.h>
25 #include <rte_dev.h>
26 
27 #include "e1000_logs.h"
28 #include "base/e1000_api.h"
29 #include "e1000_ethdev.h"
30 
31 #define EM_EIAC			0x000DC
32 
33 #define PMD_ROUNDUP(x,y)	(((x) + (y) - 1)/(y) * (y))
34 
35 
36 static int eth_em_configure(struct rte_eth_dev *dev);
37 static int eth_em_start(struct rte_eth_dev *dev);
38 static void eth_em_stop(struct rte_eth_dev *dev);
39 static void eth_em_close(struct rte_eth_dev *dev);
40 static void eth_em_promiscuous_enable(struct rte_eth_dev *dev);
41 static void eth_em_promiscuous_disable(struct rte_eth_dev *dev);
42 static void eth_em_allmulticast_enable(struct rte_eth_dev *dev);
43 static void eth_em_allmulticast_disable(struct rte_eth_dev *dev);
44 static int eth_em_link_update(struct rte_eth_dev *dev,
45 				int wait_to_complete);
46 static int eth_em_stats_get(struct rte_eth_dev *dev,
47 				struct rte_eth_stats *rte_stats);
48 static void eth_em_stats_reset(struct rte_eth_dev *dev);
49 static void eth_em_infos_get(struct rte_eth_dev *dev,
50 				struct rte_eth_dev_info *dev_info);
51 static int eth_em_flow_ctrl_get(struct rte_eth_dev *dev,
52 				struct rte_eth_fc_conf *fc_conf);
53 static int eth_em_flow_ctrl_set(struct rte_eth_dev *dev,
54 				struct rte_eth_fc_conf *fc_conf);
55 static int eth_em_interrupt_setup(struct rte_eth_dev *dev);
56 static int eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev);
57 static int eth_em_interrupt_get_status(struct rte_eth_dev *dev);
58 static int eth_em_interrupt_action(struct rte_eth_dev *dev,
59 				   struct rte_intr_handle *handle);
60 static void eth_em_interrupt_handler(void *param);
61 
62 static int em_hw_init(struct e1000_hw *hw);
63 static int em_hardware_init(struct e1000_hw *hw);
64 static void em_hw_control_acquire(struct e1000_hw *hw);
65 static void em_hw_control_release(struct e1000_hw *hw);
66 static void em_init_manageability(struct e1000_hw *hw);
67 static void em_release_manageability(struct e1000_hw *hw);
68 
69 static int eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
70 
71 static int eth_em_vlan_filter_set(struct rte_eth_dev *dev,
72 		uint16_t vlan_id, int on);
73 static int eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask);
74 static void em_vlan_hw_filter_enable(struct rte_eth_dev *dev);
75 static void em_vlan_hw_filter_disable(struct rte_eth_dev *dev);
76 static void em_vlan_hw_strip_enable(struct rte_eth_dev *dev);
77 static void em_vlan_hw_strip_disable(struct rte_eth_dev *dev);
78 
79 /*
80 static void eth_em_vlan_filter_set(struct rte_eth_dev *dev,
81 					uint16_t vlan_id, int on);
82 */
83 
84 static int eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
85 static int eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
86 static void em_lsc_intr_disable(struct e1000_hw *hw);
87 static void em_rxq_intr_enable(struct e1000_hw *hw);
88 static void em_rxq_intr_disable(struct e1000_hw *hw);
89 
90 static int eth_em_led_on(struct rte_eth_dev *dev);
91 static int eth_em_led_off(struct rte_eth_dev *dev);
92 
93 static int em_get_rx_buffer_size(struct e1000_hw *hw);
94 static int eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
95 			  uint32_t index, uint32_t pool);
96 static void eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index);
97 
98 static int eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
99 				   struct ether_addr *mc_addr_set,
100 				   uint32_t nb_mc_addr);
101 
102 #define EM_FC_PAUSE_TIME 0x0680
103 #define EM_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
104 #define EM_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
105 
106 static enum e1000_fc_mode em_fc_setting = e1000_fc_full;
107 
108 int e1000_logtype_init;
109 int e1000_logtype_driver;
110 
111 /*
112  * The set of PCI devices this driver supports
113  */
114 static const struct rte_pci_id pci_id_em_map[] = {
115 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82540EM) },
116 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_COPPER) },
117 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_FIBER) },
118 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_COPPER) },
119 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_FIBER) },
120 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_QUAD_COPPER) },
121 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_COPPER) },
122 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_FIBER) },
123 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES) },
124 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_DUAL) },
125 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_QUAD) },
126 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER) },
127 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571PT_QUAD_COPPER) },
128 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_FIBER) },
129 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER_LP) },
130 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_COPPER) },
131 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_FIBER) },
132 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_SERDES) },
133 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI) },
134 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82573L) },
135 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574L) },
136 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574LA) },
137 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82583V) },
138 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH2_LV_LM) },
139 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_LM) },
140 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_V) },
141 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_LM) },
142 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_V) },
143 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM2) },
144 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V2) },
145 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM3) },
146 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V3) },
147 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM) },
148 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V) },
149 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM2) },
150 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V2) },
151 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LBG_I219_LM3) },
152 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM4) },
153 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V4) },
154 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM5) },
155 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V5) },
156 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM6) },
157 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V6) },
158 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM7) },
159 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V7) },
160 	{ .vendor_id = 0, /* sentinel */ },
161 };
162 
163 static const struct eth_dev_ops eth_em_ops = {
164 	.dev_configure        = eth_em_configure,
165 	.dev_start            = eth_em_start,
166 	.dev_stop             = eth_em_stop,
167 	.dev_close            = eth_em_close,
168 	.promiscuous_enable   = eth_em_promiscuous_enable,
169 	.promiscuous_disable  = eth_em_promiscuous_disable,
170 	.allmulticast_enable  = eth_em_allmulticast_enable,
171 	.allmulticast_disable = eth_em_allmulticast_disable,
172 	.link_update          = eth_em_link_update,
173 	.stats_get            = eth_em_stats_get,
174 	.stats_reset          = eth_em_stats_reset,
175 	.dev_infos_get        = eth_em_infos_get,
176 	.mtu_set              = eth_em_mtu_set,
177 	.vlan_filter_set      = eth_em_vlan_filter_set,
178 	.vlan_offload_set     = eth_em_vlan_offload_set,
179 	.rx_queue_setup       = eth_em_rx_queue_setup,
180 	.rx_queue_release     = eth_em_rx_queue_release,
181 	.rx_queue_count       = eth_em_rx_queue_count,
182 	.rx_descriptor_done   = eth_em_rx_descriptor_done,
183 	.rx_descriptor_status = eth_em_rx_descriptor_status,
184 	.tx_descriptor_status = eth_em_tx_descriptor_status,
185 	.tx_queue_setup       = eth_em_tx_queue_setup,
186 	.tx_queue_release     = eth_em_tx_queue_release,
187 	.rx_queue_intr_enable = eth_em_rx_queue_intr_enable,
188 	.rx_queue_intr_disable = eth_em_rx_queue_intr_disable,
189 	.dev_led_on           = eth_em_led_on,
190 	.dev_led_off          = eth_em_led_off,
191 	.flow_ctrl_get        = eth_em_flow_ctrl_get,
192 	.flow_ctrl_set        = eth_em_flow_ctrl_set,
193 	.mac_addr_add         = eth_em_rar_set,
194 	.mac_addr_remove      = eth_em_rar_clear,
195 	.set_mc_addr_list     = eth_em_set_mc_addr_list,
196 	.rxq_info_get         = em_rxq_info_get,
197 	.txq_info_get         = em_txq_info_get,
198 };
199 
200 /**
201  * Atomically reads the link status information from global
202  * structure rte_eth_dev.
203  *
204  * @param dev
205  *   - Pointer to the structure rte_eth_dev to read from.
206  *   - Pointer to the buffer to be saved with the link status.
207  *
208  * @return
209  *   - On success, zero.
210  *   - On failure, negative value.
211  */
212 static inline int
213 rte_em_dev_atomic_read_link_status(struct rte_eth_dev *dev,
214 				struct rte_eth_link *link)
215 {
216 	struct rte_eth_link *dst = link;
217 	struct rte_eth_link *src = &(dev->data->dev_link);
218 
219 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
220 					*(uint64_t *)src) == 0)
221 		return -1;
222 
223 	return 0;
224 }
225 
226 /**
227  * Atomically writes the link status information into global
228  * structure rte_eth_dev.
229  *
230  * @param dev
231  *   - Pointer to the structure rte_eth_dev to read from.
232  *   - Pointer to the buffer to be saved with the link status.
233  *
234  * @return
235  *   - On success, zero.
236  *   - On failure, negative value.
237  */
238 static inline int
239 rte_em_dev_atomic_write_link_status(struct rte_eth_dev *dev,
240 				struct rte_eth_link *link)
241 {
242 	struct rte_eth_link *dst = &(dev->data->dev_link);
243 	struct rte_eth_link *src = link;
244 
245 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
246 					*(uint64_t *)src) == 0)
247 		return -1;
248 
249 	return 0;
250 }
251 
252 /**
253  *  eth_em_dev_is_ich8 - Check for ICH8 device
254  *  @hw: pointer to the HW structure
255  *
256  *  return TRUE for ICH8, otherwise FALSE
257  **/
258 static bool
259 eth_em_dev_is_ich8(struct e1000_hw *hw)
260 {
261 	DEBUGFUNC("eth_em_dev_is_ich8");
262 
263 	switch (hw->device_id) {
264 	case E1000_DEV_ID_PCH2_LV_LM:
265 	case E1000_DEV_ID_PCH_LPT_I217_LM:
266 	case E1000_DEV_ID_PCH_LPT_I217_V:
267 	case E1000_DEV_ID_PCH_LPTLP_I218_LM:
268 	case E1000_DEV_ID_PCH_LPTLP_I218_V:
269 	case E1000_DEV_ID_PCH_I218_V2:
270 	case E1000_DEV_ID_PCH_I218_LM2:
271 	case E1000_DEV_ID_PCH_I218_V3:
272 	case E1000_DEV_ID_PCH_I218_LM3:
273 	case E1000_DEV_ID_PCH_SPT_I219_LM:
274 	case E1000_DEV_ID_PCH_SPT_I219_V:
275 	case E1000_DEV_ID_PCH_SPT_I219_LM2:
276 	case E1000_DEV_ID_PCH_SPT_I219_V2:
277 	case E1000_DEV_ID_PCH_LBG_I219_LM3:
278 	case E1000_DEV_ID_PCH_SPT_I219_LM4:
279 	case E1000_DEV_ID_PCH_SPT_I219_V4:
280 	case E1000_DEV_ID_PCH_SPT_I219_LM5:
281 	case E1000_DEV_ID_PCH_SPT_I219_V5:
282 	case E1000_DEV_ID_PCH_CNP_I219_LM6:
283 	case E1000_DEV_ID_PCH_CNP_I219_V6:
284 	case E1000_DEV_ID_PCH_CNP_I219_LM7:
285 	case E1000_DEV_ID_PCH_CNP_I219_V7:
286 		return 1;
287 	default:
288 		return 0;
289 	}
290 }
291 
292 static int
293 eth_em_dev_init(struct rte_eth_dev *eth_dev)
294 {
295 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
296 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
297 	struct e1000_adapter *adapter =
298 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
299 	struct e1000_hw *hw =
300 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
301 	struct e1000_vfta * shadow_vfta =
302 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
303 
304 	eth_dev->dev_ops = &eth_em_ops;
305 	eth_dev->rx_pkt_burst = (eth_rx_burst_t)&eth_em_recv_pkts;
306 	eth_dev->tx_pkt_burst = (eth_tx_burst_t)&eth_em_xmit_pkts;
307 	eth_dev->tx_pkt_prepare = (eth_tx_prep_t)&eth_em_prep_pkts;
308 
309 	/* for secondary processes, we don't initialise any further as primary
310 	 * has already done this work. Only check we don't need a different
311 	 * RX function */
312 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
313 		if (eth_dev->data->scattered_rx)
314 			eth_dev->rx_pkt_burst =
315 				(eth_rx_burst_t)&eth_em_recv_scattered_pkts;
316 		return 0;
317 	}
318 
319 	rte_eth_copy_pci_info(eth_dev, pci_dev);
320 
321 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
322 	hw->device_id = pci_dev->id.device_id;
323 	adapter->stopped = 0;
324 
325 	/* For ICH8 support we'll need to map the flash memory BAR */
326 	if (eth_em_dev_is_ich8(hw))
327 		hw->flash_address = (void *)pci_dev->mem_resource[1].addr;
328 
329 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS ||
330 			em_hw_init(hw) != 0) {
331 		PMD_INIT_LOG(ERR, "port_id %d vendorID=0x%x deviceID=0x%x: "
332 			"failed to init HW",
333 			eth_dev->data->port_id, pci_dev->id.vendor_id,
334 			pci_dev->id.device_id);
335 		return -ENODEV;
336 	}
337 
338 	/* Allocate memory for storing MAC addresses */
339 	eth_dev->data->mac_addrs = rte_zmalloc("e1000", ETHER_ADDR_LEN *
340 			hw->mac.rar_entry_count, 0);
341 	if (eth_dev->data->mac_addrs == NULL) {
342 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
343 			"store MAC addresses",
344 			ETHER_ADDR_LEN * hw->mac.rar_entry_count);
345 		return -ENOMEM;
346 	}
347 
348 	/* Copy the permanent MAC address */
349 	ether_addr_copy((struct ether_addr *) hw->mac.addr,
350 		eth_dev->data->mac_addrs);
351 
352 	/* initialize the vfta */
353 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
354 
355 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
356 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
357 		     pci_dev->id.device_id);
358 
359 	rte_intr_callback_register(intr_handle,
360 				   eth_em_interrupt_handler, eth_dev);
361 
362 	return 0;
363 }
364 
365 static int
366 eth_em_dev_uninit(struct rte_eth_dev *eth_dev)
367 {
368 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
369 	struct e1000_adapter *adapter =
370 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
371 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
372 
373 	PMD_INIT_FUNC_TRACE();
374 
375 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
376 		return -EPERM;
377 
378 	if (adapter->stopped == 0)
379 		eth_em_close(eth_dev);
380 
381 	eth_dev->dev_ops = NULL;
382 	eth_dev->rx_pkt_burst = NULL;
383 	eth_dev->tx_pkt_burst = NULL;
384 
385 	rte_free(eth_dev->data->mac_addrs);
386 	eth_dev->data->mac_addrs = NULL;
387 
388 	/* disable uio intr before callback unregister */
389 	rte_intr_disable(intr_handle);
390 	rte_intr_callback_unregister(intr_handle,
391 				     eth_em_interrupt_handler, eth_dev);
392 
393 	return 0;
394 }
395 
396 static int eth_em_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
397 	struct rte_pci_device *pci_dev)
398 {
399 	return rte_eth_dev_pci_generic_probe(pci_dev,
400 		sizeof(struct e1000_adapter), eth_em_dev_init);
401 }
402 
403 static int eth_em_pci_remove(struct rte_pci_device *pci_dev)
404 {
405 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_em_dev_uninit);
406 }
407 
408 static struct rte_pci_driver rte_em_pmd = {
409 	.id_table = pci_id_em_map,
410 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
411 		     RTE_PCI_DRV_IOVA_AS_VA,
412 	.probe = eth_em_pci_probe,
413 	.remove = eth_em_pci_remove,
414 };
415 
416 static int
417 em_hw_init(struct e1000_hw *hw)
418 {
419 	int diag;
420 
421 	diag = hw->mac.ops.init_params(hw);
422 	if (diag != 0) {
423 		PMD_INIT_LOG(ERR, "MAC Initialization Error");
424 		return diag;
425 	}
426 	diag = hw->nvm.ops.init_params(hw);
427 	if (diag != 0) {
428 		PMD_INIT_LOG(ERR, "NVM Initialization Error");
429 		return diag;
430 	}
431 	diag = hw->phy.ops.init_params(hw);
432 	if (diag != 0) {
433 		PMD_INIT_LOG(ERR, "PHY Initialization Error");
434 		return diag;
435 	}
436 	(void) e1000_get_bus_info(hw);
437 
438 	hw->mac.autoneg = 1;
439 	hw->phy.autoneg_wait_to_complete = 0;
440 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
441 
442 	e1000_init_script_state_82541(hw, TRUE);
443 	e1000_set_tbi_compatibility_82543(hw, TRUE);
444 
445 	/* Copper options */
446 	if (hw->phy.media_type == e1000_media_type_copper) {
447 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
448 		hw->phy.disable_polarity_correction = 0;
449 		hw->phy.ms_type = e1000_ms_hw_default;
450 	}
451 
452 	/*
453 	 * Start from a known state, this is important in reading the nvm
454 	 * and mac from that.
455 	 */
456 	e1000_reset_hw(hw);
457 
458 	/* Make sure we have a good EEPROM before we read from it */
459 	if (e1000_validate_nvm_checksum(hw) < 0) {
460 		/*
461 		 * Some PCI-E parts fail the first check due to
462 		 * the link being in sleep state, call it again,
463 		 * if it fails a second time its a real issue.
464 		 */
465 		diag = e1000_validate_nvm_checksum(hw);
466 		if (diag < 0) {
467 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
468 			goto error;
469 		}
470 	}
471 
472 	/* Read the permanent MAC address out of the EEPROM */
473 	diag = e1000_read_mac_addr(hw);
474 	if (diag != 0) {
475 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
476 		goto error;
477 	}
478 
479 	/* Now initialize the hardware */
480 	diag = em_hardware_init(hw);
481 	if (diag != 0) {
482 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
483 		goto error;
484 	}
485 
486 	hw->mac.get_link_status = 1;
487 
488 	/* Indicate SOL/IDER usage */
489 	diag = e1000_check_reset_block(hw);
490 	if (diag < 0) {
491 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to "
492 			"SOL/IDER session");
493 	}
494 	return 0;
495 
496 error:
497 	em_hw_control_release(hw);
498 	return diag;
499 }
500 
501 static int
502 eth_em_configure(struct rte_eth_dev *dev)
503 {
504 	struct e1000_interrupt *intr =
505 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
506 
507 	PMD_INIT_FUNC_TRACE();
508 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
509 	PMD_INIT_FUNC_TRACE();
510 
511 	return 0;
512 }
513 
514 static void
515 em_set_pba(struct e1000_hw *hw)
516 {
517 	uint32_t pba;
518 
519 	/*
520 	 * Packet Buffer Allocation (PBA)
521 	 * Writing PBA sets the receive portion of the buffer
522 	 * the remainder is used for the transmit buffer.
523 	 * Devices before the 82547 had a Packet Buffer of 64K.
524 	 * After the 82547 the buffer was reduced to 40K.
525 	 */
526 	switch (hw->mac.type) {
527 		case e1000_82547:
528 		case e1000_82547_rev_2:
529 		/* 82547: Total Packet Buffer is 40K */
530 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
531 			break;
532 		case e1000_82571:
533 		case e1000_82572:
534 		case e1000_80003es2lan:
535 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
536 			break;
537 		case e1000_82573: /* 82573: Total Packet Buffer is 32K */
538 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
539 			break;
540 		case e1000_82574:
541 		case e1000_82583:
542 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
543 			break;
544 		case e1000_ich8lan:
545 			pba = E1000_PBA_8K;
546 			break;
547 		case e1000_ich9lan:
548 		case e1000_ich10lan:
549 			pba = E1000_PBA_10K;
550 			break;
551 		case e1000_pchlan:
552 		case e1000_pch2lan:
553 		case e1000_pch_lpt:
554 		case e1000_pch_spt:
555 		case e1000_pch_cnp:
556 			pba = E1000_PBA_26K;
557 			break;
558 		default:
559 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
560 	}
561 
562 	E1000_WRITE_REG(hw, E1000_PBA, pba);
563 }
564 
565 static void
566 eth_em_rxtx_control(struct rte_eth_dev *dev,
567 		    bool enable)
568 {
569 	struct e1000_hw *hw =
570 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
571 	uint32_t tctl, rctl;
572 
573 	tctl = E1000_READ_REG(hw, E1000_TCTL);
574 	rctl = E1000_READ_REG(hw, E1000_RCTL);
575 	if (enable) {
576 		/* enable Tx/Rx */
577 		tctl |= E1000_TCTL_EN;
578 		rctl |= E1000_RCTL_EN;
579 	} else {
580 		/* disable Tx/Rx */
581 		tctl &= ~E1000_TCTL_EN;
582 		rctl &= ~E1000_RCTL_EN;
583 	}
584 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
585 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
586 	E1000_WRITE_FLUSH(hw);
587 }
588 
589 static int
590 eth_em_start(struct rte_eth_dev *dev)
591 {
592 	struct e1000_adapter *adapter =
593 		E1000_DEV_PRIVATE(dev->data->dev_private);
594 	struct e1000_hw *hw =
595 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
596 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
597 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
598 	int ret, mask;
599 	uint32_t intr_vector = 0;
600 	uint32_t *speeds;
601 	int num_speeds;
602 	bool autoneg;
603 
604 	PMD_INIT_FUNC_TRACE();
605 
606 	eth_em_stop(dev);
607 
608 	e1000_power_up_phy(hw);
609 
610 	/* Set default PBA value */
611 	em_set_pba(hw);
612 
613 	/* Put the address into the Receive Address Array */
614 	e1000_rar_set(hw, hw->mac.addr, 0);
615 
616 	/*
617 	 * With the 82571 adapter, RAR[0] may be overwritten
618 	 * when the other port is reset, we make a duplicate
619 	 * in RAR[14] for that eventuality, this assures
620 	 * the interface continues to function.
621 	 */
622 	if (hw->mac.type == e1000_82571) {
623 		e1000_set_laa_state_82571(hw, TRUE);
624 		e1000_rar_set(hw, hw->mac.addr, E1000_RAR_ENTRIES - 1);
625 	}
626 
627 	/* Initialize the hardware */
628 	if (em_hardware_init(hw)) {
629 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
630 		return -EIO;
631 	}
632 
633 	E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN);
634 
635 	/* Configure for OS presence */
636 	em_init_manageability(hw);
637 
638 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
639 		intr_vector = dev->data->nb_rx_queues;
640 		if (rte_intr_efd_enable(intr_handle, intr_vector))
641 			return -1;
642 	}
643 
644 	if (rte_intr_dp_is_en(intr_handle)) {
645 		intr_handle->intr_vec =
646 			rte_zmalloc("intr_vec",
647 					dev->data->nb_rx_queues * sizeof(int), 0);
648 		if (intr_handle->intr_vec == NULL) {
649 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
650 						" intr_vec", dev->data->nb_rx_queues);
651 			return -ENOMEM;
652 		}
653 
654 		/* enable rx interrupt */
655 		em_rxq_intr_enable(hw);
656 	}
657 
658 	eth_em_tx_init(dev);
659 
660 	ret = eth_em_rx_init(dev);
661 	if (ret) {
662 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
663 		em_dev_clear_queues(dev);
664 		return ret;
665 	}
666 
667 	e1000_clear_hw_cntrs_base_generic(hw);
668 
669 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
670 			ETH_VLAN_EXTEND_MASK;
671 	ret = eth_em_vlan_offload_set(dev, mask);
672 	if (ret) {
673 		PMD_INIT_LOG(ERR, "Unable to update vlan offload");
674 		em_dev_clear_queues(dev);
675 		return ret;
676 	}
677 
678 	/* Set Interrupt Throttling Rate to maximum allowed value. */
679 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
680 
681 	/* Setup link speed and duplex */
682 	speeds = &dev->data->dev_conf.link_speeds;
683 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
684 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
685 		hw->mac.autoneg = 1;
686 	} else {
687 		num_speeds = 0;
688 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
689 
690 		/* Reset */
691 		hw->phy.autoneg_advertised = 0;
692 
693 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
694 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
695 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
696 			num_speeds = -1;
697 			goto error_invalid_config;
698 		}
699 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
700 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
701 			num_speeds++;
702 		}
703 		if (*speeds & ETH_LINK_SPEED_10M) {
704 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
705 			num_speeds++;
706 		}
707 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
708 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
709 			num_speeds++;
710 		}
711 		if (*speeds & ETH_LINK_SPEED_100M) {
712 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
713 			num_speeds++;
714 		}
715 		if (*speeds & ETH_LINK_SPEED_1G) {
716 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
717 			num_speeds++;
718 		}
719 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
720 			goto error_invalid_config;
721 
722 		/* Set/reset the mac.autoneg based on the link speed,
723 		 * fixed or not
724 		 */
725 		if (!autoneg) {
726 			hw->mac.autoneg = 0;
727 			hw->mac.forced_speed_duplex =
728 					hw->phy.autoneg_advertised;
729 		} else {
730 			hw->mac.autoneg = 1;
731 		}
732 	}
733 
734 	e1000_setup_link(hw);
735 
736 	if (rte_intr_allow_others(intr_handle)) {
737 		/* check if lsc interrupt is enabled */
738 		if (dev->data->dev_conf.intr_conf.lsc != 0) {
739 			ret = eth_em_interrupt_setup(dev);
740 			if (ret) {
741 				PMD_INIT_LOG(ERR, "Unable to setup interrupts");
742 				em_dev_clear_queues(dev);
743 				return ret;
744 			}
745 		}
746 	} else {
747 		rte_intr_callback_unregister(intr_handle,
748 						eth_em_interrupt_handler,
749 						(void *)dev);
750 		if (dev->data->dev_conf.intr_conf.lsc != 0)
751 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
752 				     " no intr multiplexn");
753 	}
754 	/* check if rxq interrupt is enabled */
755 	if (dev->data->dev_conf.intr_conf.rxq != 0)
756 		eth_em_rxq_interrupt_setup(dev);
757 
758 	rte_intr_enable(intr_handle);
759 
760 	adapter->stopped = 0;
761 
762 	eth_em_rxtx_control(dev, true);
763 	eth_em_link_update(dev, 0);
764 
765 	PMD_INIT_LOG(DEBUG, "<<");
766 
767 	return 0;
768 
769 error_invalid_config:
770 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
771 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
772 	em_dev_clear_queues(dev);
773 	return -EINVAL;
774 }
775 
776 /*********************************************************************
777  *
778  *  This routine disables all traffic on the adapter by issuing a
779  *  global reset on the MAC.
780  *
781  **********************************************************************/
782 static void
783 eth_em_stop(struct rte_eth_dev *dev)
784 {
785 	struct rte_eth_link link;
786 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
787 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
788 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
789 
790 	eth_em_rxtx_control(dev, false);
791 	em_rxq_intr_disable(hw);
792 	em_lsc_intr_disable(hw);
793 
794 	e1000_reset_hw(hw);
795 	if (hw->mac.type >= e1000_82544)
796 		E1000_WRITE_REG(hw, E1000_WUC, 0);
797 
798 	/* Power down the phy. Needed to make the link go down */
799 	e1000_power_down_phy(hw);
800 
801 	em_dev_clear_queues(dev);
802 
803 	/* clear the recorded link status */
804 	memset(&link, 0, sizeof(link));
805 	rte_em_dev_atomic_write_link_status(dev, &link);
806 
807 	if (!rte_intr_allow_others(intr_handle))
808 		/* resume to the default handler */
809 		rte_intr_callback_register(intr_handle,
810 					   eth_em_interrupt_handler,
811 					   (void *)dev);
812 
813 	/* Clean datapath event and queue/vec mapping */
814 	rte_intr_efd_disable(intr_handle);
815 	if (intr_handle->intr_vec != NULL) {
816 		rte_free(intr_handle->intr_vec);
817 		intr_handle->intr_vec = NULL;
818 	}
819 }
820 
821 static void
822 eth_em_close(struct rte_eth_dev *dev)
823 {
824 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
825 	struct e1000_adapter *adapter =
826 		E1000_DEV_PRIVATE(dev->data->dev_private);
827 
828 	eth_em_stop(dev);
829 	adapter->stopped = 1;
830 	em_dev_free_queues(dev);
831 	e1000_phy_hw_reset(hw);
832 	em_release_manageability(hw);
833 	em_hw_control_release(hw);
834 }
835 
836 static int
837 em_get_rx_buffer_size(struct e1000_hw *hw)
838 {
839 	uint32_t rx_buf_size;
840 
841 	rx_buf_size = ((E1000_READ_REG(hw, E1000_PBA) & UINT16_MAX) << 10);
842 	return rx_buf_size;
843 }
844 
845 /*********************************************************************
846  *
847  *  Initialize the hardware
848  *
849  **********************************************************************/
850 static int
851 em_hardware_init(struct e1000_hw *hw)
852 {
853 	uint32_t rx_buf_size;
854 	int diag;
855 
856 	/* Issue a global reset */
857 	e1000_reset_hw(hw);
858 
859 	/* Let the firmware know the OS is in control */
860 	em_hw_control_acquire(hw);
861 
862 	/*
863 	 * These parameters control the automatic generation (Tx) and
864 	 * response (Rx) to Ethernet PAUSE frames.
865 	 * - High water mark should allow for at least two standard size (1518)
866 	 *   frames to be received after sending an XOFF.
867 	 * - Low water mark works best when it is very near the high water mark.
868 	 *   This allows the receiver to restart by sending XON when it has
869 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
870 	 *   restart after one full frame is pulled from the buffer. There
871 	 *   could be several smaller frames in the buffer and if so they will
872 	 *   not trigger the XON until their total number reduces the buffer
873 	 *   by 1500.
874 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
875 	 */
876 	rx_buf_size = em_get_rx_buffer_size(hw);
877 
878 	hw->fc.high_water = rx_buf_size - PMD_ROUNDUP(ETHER_MAX_LEN * 2, 1024);
879 	hw->fc.low_water = hw->fc.high_water - 1500;
880 
881 	if (hw->mac.type == e1000_80003es2lan)
882 		hw->fc.pause_time = UINT16_MAX;
883 	else
884 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
885 
886 	hw->fc.send_xon = 1;
887 
888 	/* Set Flow control, use the tunable location if sane */
889 	if (em_fc_setting <= e1000_fc_full)
890 		hw->fc.requested_mode = em_fc_setting;
891 	else
892 		hw->fc.requested_mode = e1000_fc_none;
893 
894 	/* Workaround: no TX flow ctrl for PCH */
895 	if (hw->mac.type == e1000_pchlan)
896 		hw->fc.requested_mode = e1000_fc_rx_pause;
897 
898 	/* Override - settings for PCH2LAN, ya its magic :) */
899 	if (hw->mac.type == e1000_pch2lan) {
900 		hw->fc.high_water = 0x5C20;
901 		hw->fc.low_water = 0x5048;
902 		hw->fc.pause_time = 0x0650;
903 		hw->fc.refresh_time = 0x0400;
904 	} else if (hw->mac.type == e1000_pch_lpt ||
905 		   hw->mac.type == e1000_pch_spt ||
906 		   hw->mac.type == e1000_pch_cnp) {
907 		hw->fc.requested_mode = e1000_fc_full;
908 	}
909 
910 	diag = e1000_init_hw(hw);
911 	if (diag < 0)
912 		return diag;
913 	e1000_check_for_link(hw);
914 	return 0;
915 }
916 
917 /* This function is based on em_update_stats_counters() in e1000/if_em.c */
918 static int
919 eth_em_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
920 {
921 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
922 	struct e1000_hw_stats *stats =
923 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
924 	int pause_frames;
925 
926 	if(hw->phy.media_type == e1000_media_type_copper ||
927 			(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
928 		stats->symerrs += E1000_READ_REG(hw,E1000_SYMERRS);
929 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
930 	}
931 
932 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
933 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
934 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
935 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
936 
937 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
938 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
939 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
940 	stats->dc += E1000_READ_REG(hw, E1000_DC);
941 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
942 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
943 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
944 
945 	/*
946 	 * For watchdog management we need to know if we have been
947 	 * paused during the last interval, so capture that here.
948 	 */
949 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
950 	stats->xoffrxc += pause_frames;
951 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
952 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
953 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
954 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
955 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
956 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
957 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
958 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
959 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
960 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
961 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
962 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
963 
964 	/*
965 	 * For the 64-bit byte counters the low dword must be read first.
966 	 * Both registers clear on the read of the high dword.
967 	 */
968 
969 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
970 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
971 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
972 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
973 
974 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
975 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
976 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
977 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
978 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
979 
980 	stats->tor += E1000_READ_REG(hw, E1000_TORH);
981 	stats->tot += E1000_READ_REG(hw, E1000_TOTH);
982 
983 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
984 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
985 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
986 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
987 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
988 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
989 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
990 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
991 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
992 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
993 
994 	/* Interrupt Counts */
995 
996 	if (hw->mac.type >= e1000_82571) {
997 		stats->iac += E1000_READ_REG(hw, E1000_IAC);
998 		stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
999 		stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
1000 		stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
1001 		stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
1002 		stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
1003 		stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
1004 		stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
1005 		stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
1006 	}
1007 
1008 	if (hw->mac.type >= e1000_82543) {
1009 		stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
1010 		stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
1011 		stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
1012 		stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
1013 		stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
1014 		stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
1015 	}
1016 
1017 	if (rte_stats == NULL)
1018 		return -EINVAL;
1019 
1020 	/* Rx Errors */
1021 	rte_stats->imissed = stats->mpc;
1022 	rte_stats->ierrors = stats->crcerrs +
1023 	                     stats->rlec + stats->ruc + stats->roc +
1024 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
1025 
1026 	/* Tx Errors */
1027 	rte_stats->oerrors = stats->ecol + stats->latecol;
1028 
1029 	rte_stats->ipackets = stats->gprc;
1030 	rte_stats->opackets = stats->gptc;
1031 	rte_stats->ibytes   = stats->gorc;
1032 	rte_stats->obytes   = stats->gotc;
1033 	return 0;
1034 }
1035 
1036 static void
1037 eth_em_stats_reset(struct rte_eth_dev *dev)
1038 {
1039 	struct e1000_hw_stats *hw_stats =
1040 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1041 
1042 	/* HW registers are cleared on read */
1043 	eth_em_stats_get(dev, NULL);
1044 
1045 	/* Reset software totals */
1046 	memset(hw_stats, 0, sizeof(*hw_stats));
1047 }
1048 
1049 static int
1050 eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1051 {
1052 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1053 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1054 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1055 
1056 	em_rxq_intr_enable(hw);
1057 	rte_intr_enable(intr_handle);
1058 
1059 	return 0;
1060 }
1061 
1062 static int
1063 eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1064 {
1065 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1066 
1067 	em_rxq_intr_disable(hw);
1068 
1069 	return 0;
1070 }
1071 
1072 static uint32_t
1073 em_get_max_pktlen(const struct e1000_hw *hw)
1074 {
1075 	switch (hw->mac.type) {
1076 	case e1000_82571:
1077 	case e1000_82572:
1078 	case e1000_ich9lan:
1079 	case e1000_ich10lan:
1080 	case e1000_pch2lan:
1081 	case e1000_pch_lpt:
1082 	case e1000_pch_spt:
1083 	case e1000_pch_cnp:
1084 	case e1000_82574:
1085 	case e1000_80003es2lan: /* 9K Jumbo Frame size */
1086 	case e1000_82583:
1087 		return 0x2412;
1088 	case e1000_pchlan:
1089 		return 0x1000;
1090 	/* Adapters that do not support jumbo frames */
1091 	case e1000_ich8lan:
1092 		return ETHER_MAX_LEN;
1093 	default:
1094 		return MAX_JUMBO_FRAME_SIZE;
1095 	}
1096 }
1097 
1098 static void
1099 eth_em_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1100 {
1101 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1102 
1103 	dev_info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1104 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1105 	dev_info->max_rx_pktlen = em_get_max_pktlen(hw);
1106 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1107 	dev_info->rx_offload_capa =
1108 		DEV_RX_OFFLOAD_VLAN_STRIP |
1109 		DEV_RX_OFFLOAD_IPV4_CKSUM |
1110 		DEV_RX_OFFLOAD_UDP_CKSUM  |
1111 		DEV_RX_OFFLOAD_TCP_CKSUM;
1112 	dev_info->tx_offload_capa =
1113 		DEV_TX_OFFLOAD_VLAN_INSERT |
1114 		DEV_TX_OFFLOAD_IPV4_CKSUM  |
1115 		DEV_TX_OFFLOAD_UDP_CKSUM   |
1116 		DEV_TX_OFFLOAD_TCP_CKSUM;
1117 
1118 	/*
1119 	 * Starting with 631xESB hw supports 2 TX/RX queues per port.
1120 	 * Unfortunatelly, all these nics have just one TX context.
1121 	 * So we have few choises for TX:
1122 	 * - Use just one TX queue.
1123 	 * - Allow cksum offload only for one TX queue.
1124 	 * - Don't allow TX cksum offload at all.
1125 	 * For now, option #1 was chosen.
1126 	 * To use second RX queue we have to use extended RX descriptor
1127 	 * (Multiple Receive Queues are mutually exclusive with UDP
1128 	 * fragmentation and are not supported when a legacy receive
1129 	 * descriptor format is used).
1130 	 * Which means separate RX routinies - as legacy nics (82540, 82545)
1131 	 * don't support extended RXD.
1132 	 * To avoid it we support just one RX queue for now (no RSS).
1133 	 */
1134 
1135 	dev_info->max_rx_queues = 1;
1136 	dev_info->max_tx_queues = 1;
1137 
1138 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1139 		.nb_max = E1000_MAX_RING_DESC,
1140 		.nb_min = E1000_MIN_RING_DESC,
1141 		.nb_align = EM_RXD_ALIGN,
1142 	};
1143 
1144 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1145 		.nb_max = E1000_MAX_RING_DESC,
1146 		.nb_min = E1000_MIN_RING_DESC,
1147 		.nb_align = EM_TXD_ALIGN,
1148 		.nb_seg_max = EM_TX_MAX_SEG,
1149 		.nb_mtu_seg_max = EM_TX_MAX_MTU_SEG,
1150 	};
1151 
1152 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1153 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1154 			ETH_LINK_SPEED_1G;
1155 }
1156 
1157 /* return 0 means link status changed, -1 means not changed */
1158 static int
1159 eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1160 {
1161 	struct e1000_hw *hw =
1162 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1163 	struct rte_eth_link link, old;
1164 	int link_check, count;
1165 
1166 	link_check = 0;
1167 	hw->mac.get_link_status = 1;
1168 
1169 	/* possible wait-to-complete in up to 9 seconds */
1170 	for (count = 0; count < EM_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
1171 		/* Read the real link status */
1172 		switch (hw->phy.media_type) {
1173 		case e1000_media_type_copper:
1174 			/* Do the work to read phy */
1175 			e1000_check_for_link(hw);
1176 			link_check = !hw->mac.get_link_status;
1177 			break;
1178 
1179 		case e1000_media_type_fiber:
1180 			e1000_check_for_link(hw);
1181 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1182 					E1000_STATUS_LU);
1183 			break;
1184 
1185 		case e1000_media_type_internal_serdes:
1186 			e1000_check_for_link(hw);
1187 			link_check = hw->mac.serdes_has_link;
1188 			break;
1189 
1190 		default:
1191 			break;
1192 		}
1193 		if (link_check || wait_to_complete == 0)
1194 			break;
1195 		rte_delay_ms(EM_LINK_UPDATE_CHECK_INTERVAL);
1196 	}
1197 	memset(&link, 0, sizeof(link));
1198 	rte_em_dev_atomic_read_link_status(dev, &link);
1199 	old = link;
1200 
1201 	/* Now we check if a transition has happened */
1202 	if (link_check && (link.link_status == ETH_LINK_DOWN)) {
1203 		uint16_t duplex, speed;
1204 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
1205 		link.link_duplex = (duplex == FULL_DUPLEX) ?
1206 				ETH_LINK_FULL_DUPLEX :
1207 				ETH_LINK_HALF_DUPLEX;
1208 		link.link_speed = speed;
1209 		link.link_status = ETH_LINK_UP;
1210 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1211 				ETH_LINK_SPEED_FIXED);
1212 	} else if (!link_check && (link.link_status == ETH_LINK_UP)) {
1213 		link.link_speed = 0;
1214 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
1215 		link.link_status = ETH_LINK_DOWN;
1216 		link.link_autoneg = ETH_LINK_FIXED;
1217 	}
1218 	rte_em_dev_atomic_write_link_status(dev, &link);
1219 
1220 	/* not changed */
1221 	if (old.link_status == link.link_status)
1222 		return -1;
1223 
1224 	/* changed */
1225 	return 0;
1226 }
1227 
1228 /*
1229  * em_hw_control_acquire sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
1230  * For ASF and Pass Through versions of f/w this means
1231  * that the driver is loaded. For AMT version type f/w
1232  * this means that the network i/f is open.
1233  */
1234 static void
1235 em_hw_control_acquire(struct e1000_hw *hw)
1236 {
1237 	uint32_t ctrl_ext, swsm;
1238 
1239 	/* Let firmware know the driver has taken over */
1240 	if (hw->mac.type == e1000_82573) {
1241 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1242 		E1000_WRITE_REG(hw, E1000_SWSM, swsm | E1000_SWSM_DRV_LOAD);
1243 
1244 	} else {
1245 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1246 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1247 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1248 	}
1249 }
1250 
1251 /*
1252  * em_hw_control_release resets {CTRL_EXTT|FWSM}:DRV_LOAD bit.
1253  * For ASF and Pass Through versions of f/w this means that the
1254  * driver is no longer loaded. For AMT versions of the
1255  * f/w this means that the network i/f is closed.
1256  */
1257 static void
1258 em_hw_control_release(struct e1000_hw *hw)
1259 {
1260 	uint32_t ctrl_ext, swsm;
1261 
1262 	/* Let firmware taken over control of h/w */
1263 	if (hw->mac.type == e1000_82573) {
1264 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1265 		E1000_WRITE_REG(hw, E1000_SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
1266 	} else {
1267 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1268 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1269 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1270 	}
1271 }
1272 
1273 /*
1274  * Bit of a misnomer, what this really means is
1275  * to enable OS management of the system... aka
1276  * to disable special hardware management features.
1277  */
1278 static void
1279 em_init_manageability(struct e1000_hw *hw)
1280 {
1281 	if (e1000_enable_mng_pass_thru(hw)) {
1282 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
1283 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
1284 
1285 		/* disable hardware interception of ARP */
1286 		manc &= ~(E1000_MANC_ARP_EN);
1287 
1288 		/* enable receiving management packets to the host */
1289 		manc |= E1000_MANC_EN_MNG2HOST;
1290 		manc2h |= 1 << 5;  /* Mng Port 623 */
1291 		manc2h |= 1 << 6;  /* Mng Port 664 */
1292 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
1293 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1294 	}
1295 }
1296 
1297 /*
1298  * Give control back to hardware management
1299  * controller if there is one.
1300  */
1301 static void
1302 em_release_manageability(struct e1000_hw *hw)
1303 {
1304 	uint32_t manc;
1305 
1306 	if (e1000_enable_mng_pass_thru(hw)) {
1307 		manc = E1000_READ_REG(hw, E1000_MANC);
1308 
1309 		/* re-enable hardware interception of ARP */
1310 		manc |= E1000_MANC_ARP_EN;
1311 		manc &= ~E1000_MANC_EN_MNG2HOST;
1312 
1313 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1314 	}
1315 }
1316 
1317 static void
1318 eth_em_promiscuous_enable(struct rte_eth_dev *dev)
1319 {
1320 	struct e1000_hw *hw =
1321 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1322 	uint32_t rctl;
1323 
1324 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1325 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1326 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1327 }
1328 
1329 static void
1330 eth_em_promiscuous_disable(struct rte_eth_dev *dev)
1331 {
1332 	struct e1000_hw *hw =
1333 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1334 	uint32_t rctl;
1335 
1336 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1337 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_SBP);
1338 	if (dev->data->all_multicast == 1)
1339 		rctl |= E1000_RCTL_MPE;
1340 	else
1341 		rctl &= (~E1000_RCTL_MPE);
1342 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1343 }
1344 
1345 static void
1346 eth_em_allmulticast_enable(struct rte_eth_dev *dev)
1347 {
1348 	struct e1000_hw *hw =
1349 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1350 	uint32_t rctl;
1351 
1352 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1353 	rctl |= E1000_RCTL_MPE;
1354 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1355 }
1356 
1357 static void
1358 eth_em_allmulticast_disable(struct rte_eth_dev *dev)
1359 {
1360 	struct e1000_hw *hw =
1361 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1362 	uint32_t rctl;
1363 
1364 	if (dev->data->promiscuous == 1)
1365 		return; /* must remain in all_multicast mode */
1366 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1367 	rctl &= (~E1000_RCTL_MPE);
1368 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1369 }
1370 
1371 static int
1372 eth_em_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1373 {
1374 	struct e1000_hw *hw =
1375 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1376 	struct e1000_vfta * shadow_vfta =
1377 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1378 	uint32_t vfta;
1379 	uint32_t vid_idx;
1380 	uint32_t vid_bit;
1381 
1382 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
1383 			      E1000_VFTA_ENTRY_MASK);
1384 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
1385 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
1386 	if (on)
1387 		vfta |= vid_bit;
1388 	else
1389 		vfta &= ~vid_bit;
1390 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
1391 
1392 	/* update local VFTA copy */
1393 	shadow_vfta->vfta[vid_idx] = vfta;
1394 
1395 	return 0;
1396 }
1397 
1398 static void
1399 em_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1400 {
1401 	struct e1000_hw *hw =
1402 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1403 	uint32_t reg;
1404 
1405 	/* Filter Table Disable */
1406 	reg = E1000_READ_REG(hw, E1000_RCTL);
1407 	reg &= ~E1000_RCTL_CFIEN;
1408 	reg &= ~E1000_RCTL_VFE;
1409 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1410 }
1411 
1412 static void
1413 em_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1414 {
1415 	struct e1000_hw *hw =
1416 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1417 	struct e1000_vfta * shadow_vfta =
1418 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1419 	uint32_t reg;
1420 	int i;
1421 
1422 	/* Filter Table Enable, CFI not used for packet acceptance */
1423 	reg = E1000_READ_REG(hw, E1000_RCTL);
1424 	reg &= ~E1000_RCTL_CFIEN;
1425 	reg |= E1000_RCTL_VFE;
1426 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1427 
1428 	/* restore vfta from local copy */
1429 	for (i = 0; i < IGB_VFTA_SIZE; i++)
1430 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
1431 }
1432 
1433 static void
1434 em_vlan_hw_strip_disable(struct rte_eth_dev *dev)
1435 {
1436 	struct e1000_hw *hw =
1437 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1438 	uint32_t reg;
1439 
1440 	/* VLAN Mode Disable */
1441 	reg = E1000_READ_REG(hw, E1000_CTRL);
1442 	reg &= ~E1000_CTRL_VME;
1443 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1444 
1445 }
1446 
1447 static void
1448 em_vlan_hw_strip_enable(struct rte_eth_dev *dev)
1449 {
1450 	struct e1000_hw *hw =
1451 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1452 	uint32_t reg;
1453 
1454 	/* VLAN Mode Enable */
1455 	reg = E1000_READ_REG(hw, E1000_CTRL);
1456 	reg |= E1000_CTRL_VME;
1457 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1458 }
1459 
1460 static int
1461 eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1462 {
1463 	if(mask & ETH_VLAN_STRIP_MASK){
1464 		if (dev->data->dev_conf.rxmode.hw_vlan_strip)
1465 			em_vlan_hw_strip_enable(dev);
1466 		else
1467 			em_vlan_hw_strip_disable(dev);
1468 	}
1469 
1470 	if(mask & ETH_VLAN_FILTER_MASK){
1471 		if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1472 			em_vlan_hw_filter_enable(dev);
1473 		else
1474 			em_vlan_hw_filter_disable(dev);
1475 	}
1476 
1477 	return 0;
1478 }
1479 
1480 /*
1481  * It enables the interrupt mask and then enable the interrupt.
1482  *
1483  * @param dev
1484  *  Pointer to struct rte_eth_dev.
1485  *
1486  * @return
1487  *  - On success, zero.
1488  *  - On failure, a negative value.
1489  */
1490 static int
1491 eth_em_interrupt_setup(struct rte_eth_dev *dev)
1492 {
1493 	uint32_t regval;
1494 	struct e1000_hw *hw =
1495 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1496 
1497 	/* clear interrupt */
1498 	E1000_READ_REG(hw, E1000_ICR);
1499 	regval = E1000_READ_REG(hw, E1000_IMS);
1500 	E1000_WRITE_REG(hw, E1000_IMS, regval | E1000_ICR_LSC);
1501 	return 0;
1502 }
1503 
1504 /*
1505  * It clears the interrupt causes and enables the interrupt.
1506  * It will be called once only during nic initialized.
1507  *
1508  * @param dev
1509  *  Pointer to struct rte_eth_dev.
1510  *
1511  * @return
1512  *  - On success, zero.
1513  *  - On failure, a negative value.
1514  */
1515 static int
1516 eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev)
1517 {
1518 	struct e1000_hw *hw =
1519 	E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1520 
1521 	E1000_READ_REG(hw, E1000_ICR);
1522 	em_rxq_intr_enable(hw);
1523 	return 0;
1524 }
1525 
1526 /*
1527  * It enable receive packet interrupt.
1528  * @param hw
1529  * Pointer to struct e1000_hw
1530  *
1531  * @return
1532  */
1533 static void
1534 em_rxq_intr_enable(struct e1000_hw *hw)
1535 {
1536 	E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_RXT0);
1537 	E1000_WRITE_FLUSH(hw);
1538 }
1539 
1540 /*
1541  * It disabled lsc interrupt.
1542  * @param hw
1543  * Pointer to struct e1000_hw
1544  *
1545  * @return
1546  */
1547 static void
1548 em_lsc_intr_disable(struct e1000_hw *hw)
1549 {
1550 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC);
1551 	E1000_WRITE_FLUSH(hw);
1552 }
1553 
1554 /*
1555  * It disabled receive packet interrupt.
1556  * @param hw
1557  * Pointer to struct e1000_hw
1558  *
1559  * @return
1560  */
1561 static void
1562 em_rxq_intr_disable(struct e1000_hw *hw)
1563 {
1564 	E1000_READ_REG(hw, E1000_ICR);
1565 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_RXT0);
1566 	E1000_WRITE_FLUSH(hw);
1567 }
1568 
1569 /*
1570  * It reads ICR and gets interrupt causes, check it and set a bit flag
1571  * to update link status.
1572  *
1573  * @param dev
1574  *  Pointer to struct rte_eth_dev.
1575  *
1576  * @return
1577  *  - On success, zero.
1578  *  - On failure, a negative value.
1579  */
1580 static int
1581 eth_em_interrupt_get_status(struct rte_eth_dev *dev)
1582 {
1583 	uint32_t icr;
1584 	struct e1000_hw *hw =
1585 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1586 	struct e1000_interrupt *intr =
1587 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1588 
1589 	/* read-on-clear nic registers here */
1590 	icr = E1000_READ_REG(hw, E1000_ICR);
1591 	if (icr & E1000_ICR_LSC) {
1592 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1593 	}
1594 
1595 	return 0;
1596 }
1597 
1598 /*
1599  * It executes link_update after knowing an interrupt is prsent.
1600  *
1601  * @param dev
1602  *  Pointer to struct rte_eth_dev.
1603  *
1604  * @return
1605  *  - On success, zero.
1606  *  - On failure, a negative value.
1607  */
1608 static int
1609 eth_em_interrupt_action(struct rte_eth_dev *dev,
1610 			struct rte_intr_handle *intr_handle)
1611 {
1612 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1613 	struct e1000_hw *hw =
1614 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1615 	struct e1000_interrupt *intr =
1616 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1617 	struct rte_eth_link link;
1618 	int ret;
1619 
1620 	if (!(intr->flags & E1000_FLAG_NEED_LINK_UPDATE))
1621 		return -1;
1622 
1623 	intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
1624 	rte_intr_enable(intr_handle);
1625 
1626 	/* set get_link_status to check register later */
1627 	hw->mac.get_link_status = 1;
1628 	ret = eth_em_link_update(dev, 0);
1629 
1630 	/* check if link has changed */
1631 	if (ret < 0)
1632 		return 0;
1633 
1634 	memset(&link, 0, sizeof(link));
1635 	rte_em_dev_atomic_read_link_status(dev, &link);
1636 	if (link.link_status) {
1637 		PMD_INIT_LOG(INFO, " Port %d: Link Up - speed %u Mbps - %s",
1638 			     dev->data->port_id, link.link_speed,
1639 			     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
1640 			     "full-duplex" : "half-duplex");
1641 	} else {
1642 		PMD_INIT_LOG(INFO, " Port %d: Link Down", dev->data->port_id);
1643 	}
1644 	PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
1645 		     pci_dev->addr.domain, pci_dev->addr.bus,
1646 		     pci_dev->addr.devid, pci_dev->addr.function);
1647 
1648 	return 0;
1649 }
1650 
1651 /**
1652  * Interrupt handler which shall be registered at first.
1653  *
1654  * @param handle
1655  *  Pointer to interrupt handle.
1656  * @param param
1657  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1658  *
1659  * @return
1660  *  void
1661  */
1662 static void
1663 eth_em_interrupt_handler(void *param)
1664 {
1665 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1666 
1667 	eth_em_interrupt_get_status(dev);
1668 	eth_em_interrupt_action(dev, dev->intr_handle);
1669 	_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
1670 }
1671 
1672 static int
1673 eth_em_led_on(struct rte_eth_dev *dev)
1674 {
1675 	struct e1000_hw *hw;
1676 
1677 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1678 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1679 }
1680 
1681 static int
1682 eth_em_led_off(struct rte_eth_dev *dev)
1683 {
1684 	struct e1000_hw *hw;
1685 
1686 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1687 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1688 }
1689 
1690 static int
1691 eth_em_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1692 {
1693 	struct e1000_hw *hw;
1694 	uint32_t ctrl;
1695 	int tx_pause;
1696 	int rx_pause;
1697 
1698 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1699 	fc_conf->pause_time = hw->fc.pause_time;
1700 	fc_conf->high_water = hw->fc.high_water;
1701 	fc_conf->low_water = hw->fc.low_water;
1702 	fc_conf->send_xon = hw->fc.send_xon;
1703 	fc_conf->autoneg = hw->mac.autoneg;
1704 
1705 	/*
1706 	 * Return rx_pause and tx_pause status according to actual setting of
1707 	 * the TFCE and RFCE bits in the CTRL register.
1708 	 */
1709 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
1710 	if (ctrl & E1000_CTRL_TFCE)
1711 		tx_pause = 1;
1712 	else
1713 		tx_pause = 0;
1714 
1715 	if (ctrl & E1000_CTRL_RFCE)
1716 		rx_pause = 1;
1717 	else
1718 		rx_pause = 0;
1719 
1720 	if (rx_pause && tx_pause)
1721 		fc_conf->mode = RTE_FC_FULL;
1722 	else if (rx_pause)
1723 		fc_conf->mode = RTE_FC_RX_PAUSE;
1724 	else if (tx_pause)
1725 		fc_conf->mode = RTE_FC_TX_PAUSE;
1726 	else
1727 		fc_conf->mode = RTE_FC_NONE;
1728 
1729 	return 0;
1730 }
1731 
1732 static int
1733 eth_em_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1734 {
1735 	struct e1000_hw *hw;
1736 	int err;
1737 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
1738 		e1000_fc_none,
1739 		e1000_fc_rx_pause,
1740 		e1000_fc_tx_pause,
1741 		e1000_fc_full
1742 	};
1743 	uint32_t rx_buf_size;
1744 	uint32_t max_high_water;
1745 	uint32_t rctl;
1746 
1747 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1748 	if (fc_conf->autoneg != hw->mac.autoneg)
1749 		return -ENOTSUP;
1750 	rx_buf_size = em_get_rx_buffer_size(hw);
1751 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
1752 
1753 	/* At least reserve one Ethernet frame for watermark */
1754 	max_high_water = rx_buf_size - ETHER_MAX_LEN;
1755 	if ((fc_conf->high_water > max_high_water) ||
1756 	    (fc_conf->high_water < fc_conf->low_water)) {
1757 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
1758 		PMD_INIT_LOG(ERR, "high water must <= 0x%x", max_high_water);
1759 		return -EINVAL;
1760 	}
1761 
1762 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
1763 	hw->fc.pause_time     = fc_conf->pause_time;
1764 	hw->fc.high_water     = fc_conf->high_water;
1765 	hw->fc.low_water      = fc_conf->low_water;
1766 	hw->fc.send_xon	      = fc_conf->send_xon;
1767 
1768 	err = e1000_setup_link_generic(hw);
1769 	if (err == E1000_SUCCESS) {
1770 
1771 		/* check if we want to forward MAC frames - driver doesn't have native
1772 		 * capability to do that, so we'll write the registers ourselves */
1773 
1774 		rctl = E1000_READ_REG(hw, E1000_RCTL);
1775 
1776 		/* set or clear MFLCN.PMCF bit depending on configuration */
1777 		if (fc_conf->mac_ctrl_frame_fwd != 0)
1778 			rctl |= E1000_RCTL_PMCF;
1779 		else
1780 			rctl &= ~E1000_RCTL_PMCF;
1781 
1782 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1783 		E1000_WRITE_FLUSH(hw);
1784 
1785 		return 0;
1786 	}
1787 
1788 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
1789 	return -EIO;
1790 }
1791 
1792 static int
1793 eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
1794 		uint32_t index, __rte_unused uint32_t pool)
1795 {
1796 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1797 
1798 	return e1000_rar_set(hw, mac_addr->addr_bytes, index);
1799 }
1800 
1801 static void
1802 eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1803 {
1804 	uint8_t addr[ETHER_ADDR_LEN];
1805 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1806 
1807 	memset(addr, 0, sizeof(addr));
1808 
1809 	e1000_rar_set(hw, addr, index);
1810 }
1811 
1812 static int
1813 eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1814 {
1815 	struct rte_eth_dev_info dev_info;
1816 	struct e1000_hw *hw;
1817 	uint32_t frame_size;
1818 	uint32_t rctl;
1819 
1820 	eth_em_infos_get(dev, &dev_info);
1821 	frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + VLAN_TAG_SIZE;
1822 
1823 	/* check that mtu is within the allowed range */
1824 	if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
1825 		return -EINVAL;
1826 
1827 	/* refuse mtu that requires the support of scattered packets when this
1828 	 * feature has not been enabled before. */
1829 	if (!dev->data->scattered_rx &&
1830 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
1831 		return -EINVAL;
1832 
1833 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1834 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1835 
1836 	/* switch to jumbo mode if needed */
1837 	if (frame_size > ETHER_MAX_LEN) {
1838 		dev->data->dev_conf.rxmode.jumbo_frame = 1;
1839 		rctl |= E1000_RCTL_LPE;
1840 	} else {
1841 		dev->data->dev_conf.rxmode.jumbo_frame = 0;
1842 		rctl &= ~E1000_RCTL_LPE;
1843 	}
1844 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1845 
1846 	/* update max frame size */
1847 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1848 	return 0;
1849 }
1850 
1851 static int
1852 eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
1853 			struct ether_addr *mc_addr_set,
1854 			uint32_t nb_mc_addr)
1855 {
1856 	struct e1000_hw *hw;
1857 
1858 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1859 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1860 	return 0;
1861 }
1862 
1863 RTE_PMD_REGISTER_PCI(net_e1000_em, rte_em_pmd);
1864 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_em, pci_id_em_map);
1865 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_em, "* igb_uio | uio_pci_generic | vfio-pci");
1866 
1867 RTE_INIT(e1000_init_log);
1868 static void
1869 e1000_init_log(void)
1870 {
1871 	e1000_logtype_init = rte_log_register("pmd.e1000.init");
1872 	if (e1000_logtype_init >= 0)
1873 		rte_log_set_level(e1000_logtype_init, RTE_LOG_NOTICE);
1874 	e1000_logtype_driver = rte_log_register("pmd.e1000.driver");
1875 	if (e1000_logtype_driver >= 0)
1876 		rte_log_set_level(e1000_logtype_driver, RTE_LOG_NOTICE);
1877 }
1878