xref: /dpdk/drivers/net/e1000/em_ethdev.c (revision c7e9729d)
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_driver.h>
20 #include <rte_ethdev_pci.h>
21 #include <rte_memory.h>
22 #include <rte_eal.h>
23 #include <rte_malloc.h>
24 #include <rte_dev.h>
25 
26 #include "e1000_logs.h"
27 #include "base/e1000_api.h"
28 #include "e1000_ethdev.h"
29 
30 #define EM_EIAC			0x000DC
31 
32 #define PMD_ROUNDUP(x,y)	(((x) + (y) - 1)/(y) * (y))
33 
34 
35 static int eth_em_configure(struct rte_eth_dev *dev);
36 static int eth_em_start(struct rte_eth_dev *dev);
37 static void eth_em_stop(struct rte_eth_dev *dev);
38 static void eth_em_close(struct rte_eth_dev *dev);
39 static void eth_em_promiscuous_enable(struct rte_eth_dev *dev);
40 static void eth_em_promiscuous_disable(struct rte_eth_dev *dev);
41 static void eth_em_allmulticast_enable(struct rte_eth_dev *dev);
42 static void eth_em_allmulticast_disable(struct rte_eth_dev *dev);
43 static int eth_em_link_update(struct rte_eth_dev *dev,
44 				int wait_to_complete);
45 static int eth_em_stats_get(struct rte_eth_dev *dev,
46 				struct rte_eth_stats *rte_stats);
47 static void eth_em_stats_reset(struct rte_eth_dev *dev);
48 static void eth_em_infos_get(struct rte_eth_dev *dev,
49 				struct rte_eth_dev_info *dev_info);
50 static int eth_em_flow_ctrl_get(struct rte_eth_dev *dev,
51 				struct rte_eth_fc_conf *fc_conf);
52 static int eth_em_flow_ctrl_set(struct rte_eth_dev *dev,
53 				struct rte_eth_fc_conf *fc_conf);
54 static int eth_em_interrupt_setup(struct rte_eth_dev *dev);
55 static int eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev);
56 static int eth_em_interrupt_get_status(struct rte_eth_dev *dev);
57 static int eth_em_interrupt_action(struct rte_eth_dev *dev,
58 				   struct rte_intr_handle *handle);
59 static void eth_em_interrupt_handler(void *param);
60 
61 static int em_hw_init(struct e1000_hw *hw);
62 static int em_hardware_init(struct e1000_hw *hw);
63 static void em_hw_control_acquire(struct e1000_hw *hw);
64 static void em_hw_control_release(struct e1000_hw *hw);
65 static void em_init_manageability(struct e1000_hw *hw);
66 static void em_release_manageability(struct e1000_hw *hw);
67 
68 static int eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
69 
70 static int eth_em_vlan_filter_set(struct rte_eth_dev *dev,
71 		uint16_t vlan_id, int on);
72 static int eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask);
73 static void em_vlan_hw_filter_enable(struct rte_eth_dev *dev);
74 static void em_vlan_hw_filter_disable(struct rte_eth_dev *dev);
75 static void em_vlan_hw_strip_enable(struct rte_eth_dev *dev);
76 static void em_vlan_hw_strip_disable(struct rte_eth_dev *dev);
77 
78 /*
79 static void eth_em_vlan_filter_set(struct rte_eth_dev *dev,
80 					uint16_t vlan_id, int on);
81 */
82 
83 static int eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
84 static int eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
85 static void em_lsc_intr_disable(struct e1000_hw *hw);
86 static void em_rxq_intr_enable(struct e1000_hw *hw);
87 static void em_rxq_intr_disable(struct e1000_hw *hw);
88 
89 static int eth_em_led_on(struct rte_eth_dev *dev);
90 static int eth_em_led_off(struct rte_eth_dev *dev);
91 
92 static int em_get_rx_buffer_size(struct e1000_hw *hw);
93 static int eth_em_rar_set(struct rte_eth_dev *dev, struct 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 
97 static int eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
98 				   struct ether_addr *mc_addr_set,
99 				   uint32_t nb_mc_addr);
100 
101 #define EM_FC_PAUSE_TIME 0x0680
102 #define EM_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
103 #define EM_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
104 
105 static enum e1000_fc_mode em_fc_setting = e1000_fc_full;
106 
107 int e1000_logtype_init;
108 int e1000_logtype_driver;
109 
110 /*
111  * The set of PCI devices this driver supports
112  */
113 static const struct rte_pci_id pci_id_em_map[] = {
114 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82540EM) },
115 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_COPPER) },
116 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_FIBER) },
117 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_COPPER) },
118 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_FIBER) },
119 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_QUAD_COPPER) },
120 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_COPPER) },
121 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_FIBER) },
122 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES) },
123 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_DUAL) },
124 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_QUAD) },
125 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER) },
126 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571PT_QUAD_COPPER) },
127 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_FIBER) },
128 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER_LP) },
129 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_COPPER) },
130 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_FIBER) },
131 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_SERDES) },
132 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI) },
133 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82573L) },
134 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574L) },
135 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574LA) },
136 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82583V) },
137 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH2_LV_LM) },
138 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_LM) },
139 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_V) },
140 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_LM) },
141 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_V) },
142 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM2) },
143 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V2) },
144 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM3) },
145 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V3) },
146 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM) },
147 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V) },
148 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM2) },
149 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V2) },
150 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LBG_I219_LM3) },
151 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM4) },
152 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V4) },
153 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM5) },
154 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V5) },
155 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM6) },
156 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V6) },
157 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM7) },
158 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V7) },
159 	{ .vendor_id = 0, /* sentinel */ },
160 };
161 
162 static const struct eth_dev_ops eth_em_ops = {
163 	.dev_configure        = eth_em_configure,
164 	.dev_start            = eth_em_start,
165 	.dev_stop             = eth_em_stop,
166 	.dev_close            = eth_em_close,
167 	.promiscuous_enable   = eth_em_promiscuous_enable,
168 	.promiscuous_disable  = eth_em_promiscuous_disable,
169 	.allmulticast_enable  = eth_em_allmulticast_enable,
170 	.allmulticast_disable = eth_em_allmulticast_disable,
171 	.link_update          = eth_em_link_update,
172 	.stats_get            = eth_em_stats_get,
173 	.stats_reset          = eth_em_stats_reset,
174 	.dev_infos_get        = eth_em_infos_get,
175 	.mtu_set              = eth_em_mtu_set,
176 	.vlan_filter_set      = eth_em_vlan_filter_set,
177 	.vlan_offload_set     = eth_em_vlan_offload_set,
178 	.rx_queue_setup       = eth_em_rx_queue_setup,
179 	.rx_queue_release     = eth_em_rx_queue_release,
180 	.rx_queue_count       = eth_em_rx_queue_count,
181 	.rx_descriptor_done   = eth_em_rx_descriptor_done,
182 	.rx_descriptor_status = eth_em_rx_descriptor_status,
183 	.tx_descriptor_status = eth_em_tx_descriptor_status,
184 	.tx_queue_setup       = eth_em_tx_queue_setup,
185 	.tx_queue_release     = eth_em_tx_queue_release,
186 	.rx_queue_intr_enable = eth_em_rx_queue_intr_enable,
187 	.rx_queue_intr_disable = eth_em_rx_queue_intr_disable,
188 	.dev_led_on           = eth_em_led_on,
189 	.dev_led_off          = eth_em_led_off,
190 	.flow_ctrl_get        = eth_em_flow_ctrl_get,
191 	.flow_ctrl_set        = eth_em_flow_ctrl_set,
192 	.mac_addr_add         = eth_em_rar_set,
193 	.mac_addr_remove      = eth_em_rar_clear,
194 	.set_mc_addr_list     = eth_em_set_mc_addr_list,
195 	.rxq_info_get         = em_rxq_info_get,
196 	.txq_info_get         = em_txq_info_get,
197 };
198 
199 
200 /**
201  *  eth_em_dev_is_ich8 - Check for ICH8 device
202  *  @hw: pointer to the HW structure
203  *
204  *  return TRUE for ICH8, otherwise FALSE
205  **/
206 static bool
207 eth_em_dev_is_ich8(struct e1000_hw *hw)
208 {
209 	DEBUGFUNC("eth_em_dev_is_ich8");
210 
211 	switch (hw->device_id) {
212 	case E1000_DEV_ID_PCH2_LV_LM:
213 	case E1000_DEV_ID_PCH_LPT_I217_LM:
214 	case E1000_DEV_ID_PCH_LPT_I217_V:
215 	case E1000_DEV_ID_PCH_LPTLP_I218_LM:
216 	case E1000_DEV_ID_PCH_LPTLP_I218_V:
217 	case E1000_DEV_ID_PCH_I218_V2:
218 	case E1000_DEV_ID_PCH_I218_LM2:
219 	case E1000_DEV_ID_PCH_I218_V3:
220 	case E1000_DEV_ID_PCH_I218_LM3:
221 	case E1000_DEV_ID_PCH_SPT_I219_LM:
222 	case E1000_DEV_ID_PCH_SPT_I219_V:
223 	case E1000_DEV_ID_PCH_SPT_I219_LM2:
224 	case E1000_DEV_ID_PCH_SPT_I219_V2:
225 	case E1000_DEV_ID_PCH_LBG_I219_LM3:
226 	case E1000_DEV_ID_PCH_SPT_I219_LM4:
227 	case E1000_DEV_ID_PCH_SPT_I219_V4:
228 	case E1000_DEV_ID_PCH_SPT_I219_LM5:
229 	case E1000_DEV_ID_PCH_SPT_I219_V5:
230 	case E1000_DEV_ID_PCH_CNP_I219_LM6:
231 	case E1000_DEV_ID_PCH_CNP_I219_V6:
232 	case E1000_DEV_ID_PCH_CNP_I219_LM7:
233 	case E1000_DEV_ID_PCH_CNP_I219_V7:
234 		return 1;
235 	default:
236 		return 0;
237 	}
238 }
239 
240 static int
241 eth_em_dev_init(struct rte_eth_dev *eth_dev)
242 {
243 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
244 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
245 	struct e1000_adapter *adapter =
246 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
247 	struct e1000_hw *hw =
248 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
249 	struct e1000_vfta * shadow_vfta =
250 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
251 
252 	eth_dev->dev_ops = &eth_em_ops;
253 	eth_dev->rx_pkt_burst = (eth_rx_burst_t)&eth_em_recv_pkts;
254 	eth_dev->tx_pkt_burst = (eth_tx_burst_t)&eth_em_xmit_pkts;
255 	eth_dev->tx_pkt_prepare = (eth_tx_prep_t)&eth_em_prep_pkts;
256 
257 	/* for secondary processes, we don't initialise any further as primary
258 	 * has already done this work. Only check we don't need a different
259 	 * RX function */
260 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
261 		if (eth_dev->data->scattered_rx)
262 			eth_dev->rx_pkt_burst =
263 				(eth_rx_burst_t)&eth_em_recv_scattered_pkts;
264 		return 0;
265 	}
266 
267 	rte_eth_copy_pci_info(eth_dev, pci_dev);
268 
269 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
270 	hw->device_id = pci_dev->id.device_id;
271 	adapter->stopped = 0;
272 
273 	/* For ICH8 support we'll need to map the flash memory BAR */
274 	if (eth_em_dev_is_ich8(hw))
275 		hw->flash_address = (void *)pci_dev->mem_resource[1].addr;
276 
277 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS ||
278 			em_hw_init(hw) != 0) {
279 		PMD_INIT_LOG(ERR, "port_id %d vendorID=0x%x deviceID=0x%x: "
280 			"failed to init HW",
281 			eth_dev->data->port_id, pci_dev->id.vendor_id,
282 			pci_dev->id.device_id);
283 		return -ENODEV;
284 	}
285 
286 	/* Allocate memory for storing MAC addresses */
287 	eth_dev->data->mac_addrs = rte_zmalloc("e1000", ETHER_ADDR_LEN *
288 			hw->mac.rar_entry_count, 0);
289 	if (eth_dev->data->mac_addrs == NULL) {
290 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
291 			"store MAC addresses",
292 			ETHER_ADDR_LEN * hw->mac.rar_entry_count);
293 		return -ENOMEM;
294 	}
295 
296 	/* Copy the permanent MAC address */
297 	ether_addr_copy((struct ether_addr *) hw->mac.addr,
298 		eth_dev->data->mac_addrs);
299 
300 	/* initialize the vfta */
301 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
302 
303 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
304 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
305 		     pci_dev->id.device_id);
306 
307 	rte_intr_callback_register(intr_handle,
308 				   eth_em_interrupt_handler, eth_dev);
309 
310 	return 0;
311 }
312 
313 static int
314 eth_em_dev_uninit(struct rte_eth_dev *eth_dev)
315 {
316 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
317 	struct e1000_adapter *adapter =
318 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
319 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
320 
321 	PMD_INIT_FUNC_TRACE();
322 
323 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
324 		return -EPERM;
325 
326 	if (adapter->stopped == 0)
327 		eth_em_close(eth_dev);
328 
329 	eth_dev->dev_ops = NULL;
330 	eth_dev->rx_pkt_burst = NULL;
331 	eth_dev->tx_pkt_burst = NULL;
332 
333 	rte_free(eth_dev->data->mac_addrs);
334 	eth_dev->data->mac_addrs = NULL;
335 
336 	/* disable uio intr before callback unregister */
337 	rte_intr_disable(intr_handle);
338 	rte_intr_callback_unregister(intr_handle,
339 				     eth_em_interrupt_handler, eth_dev);
340 
341 	return 0;
342 }
343 
344 static int eth_em_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
345 	struct rte_pci_device *pci_dev)
346 {
347 	return rte_eth_dev_pci_generic_probe(pci_dev,
348 		sizeof(struct e1000_adapter), eth_em_dev_init);
349 }
350 
351 static int eth_em_pci_remove(struct rte_pci_device *pci_dev)
352 {
353 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_em_dev_uninit);
354 }
355 
356 static struct rte_pci_driver rte_em_pmd = {
357 	.id_table = pci_id_em_map,
358 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
359 		     RTE_PCI_DRV_IOVA_AS_VA,
360 	.probe = eth_em_pci_probe,
361 	.remove = eth_em_pci_remove,
362 };
363 
364 static int
365 em_hw_init(struct e1000_hw *hw)
366 {
367 	int diag;
368 
369 	diag = hw->mac.ops.init_params(hw);
370 	if (diag != 0) {
371 		PMD_INIT_LOG(ERR, "MAC Initialization Error");
372 		return diag;
373 	}
374 	diag = hw->nvm.ops.init_params(hw);
375 	if (diag != 0) {
376 		PMD_INIT_LOG(ERR, "NVM Initialization Error");
377 		return diag;
378 	}
379 	diag = hw->phy.ops.init_params(hw);
380 	if (diag != 0) {
381 		PMD_INIT_LOG(ERR, "PHY Initialization Error");
382 		return diag;
383 	}
384 	(void) e1000_get_bus_info(hw);
385 
386 	hw->mac.autoneg = 1;
387 	hw->phy.autoneg_wait_to_complete = 0;
388 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
389 
390 	e1000_init_script_state_82541(hw, TRUE);
391 	e1000_set_tbi_compatibility_82543(hw, TRUE);
392 
393 	/* Copper options */
394 	if (hw->phy.media_type == e1000_media_type_copper) {
395 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
396 		hw->phy.disable_polarity_correction = 0;
397 		hw->phy.ms_type = e1000_ms_hw_default;
398 	}
399 
400 	/*
401 	 * Start from a known state, this is important in reading the nvm
402 	 * and mac from that.
403 	 */
404 	e1000_reset_hw(hw);
405 
406 	/* Make sure we have a good EEPROM before we read from it */
407 	if (e1000_validate_nvm_checksum(hw) < 0) {
408 		/*
409 		 * Some PCI-E parts fail the first check due to
410 		 * the link being in sleep state, call it again,
411 		 * if it fails a second time its a real issue.
412 		 */
413 		diag = e1000_validate_nvm_checksum(hw);
414 		if (diag < 0) {
415 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
416 			goto error;
417 		}
418 	}
419 
420 	/* Read the permanent MAC address out of the EEPROM */
421 	diag = e1000_read_mac_addr(hw);
422 	if (diag != 0) {
423 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
424 		goto error;
425 	}
426 
427 	/* Now initialize the hardware */
428 	diag = em_hardware_init(hw);
429 	if (diag != 0) {
430 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
431 		goto error;
432 	}
433 
434 	hw->mac.get_link_status = 1;
435 
436 	/* Indicate SOL/IDER usage */
437 	diag = e1000_check_reset_block(hw);
438 	if (diag < 0) {
439 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to "
440 			"SOL/IDER session");
441 	}
442 	return 0;
443 
444 error:
445 	em_hw_control_release(hw);
446 	return diag;
447 }
448 
449 static int
450 eth_em_configure(struct rte_eth_dev *dev)
451 {
452 	struct e1000_interrupt *intr =
453 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
454 	struct rte_eth_dev_info dev_info;
455 	uint64_t rx_offloads;
456 	uint64_t tx_offloads;
457 
458 	PMD_INIT_FUNC_TRACE();
459 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
460 
461 	eth_em_infos_get(dev, &dev_info);
462 	rx_offloads = dev->data->dev_conf.rxmode.offloads;
463 	if ((rx_offloads & dev_info.rx_offload_capa) != rx_offloads) {
464 		PMD_DRV_LOG(ERR, "Some Rx offloads are not supported "
465 			    "requested 0x%" PRIx64 " supported 0x%" PRIx64,
466 			    rx_offloads, dev_info.rx_offload_capa);
467 		return -ENOTSUP;
468 	}
469 	tx_offloads = dev->data->dev_conf.txmode.offloads;
470 	if ((tx_offloads & dev_info.tx_offload_capa) != tx_offloads) {
471 		PMD_DRV_LOG(ERR, "Some Tx offloads are not supported "
472 			    "requested 0x%" PRIx64 " supported 0x%" PRIx64,
473 			    tx_offloads, dev_info.tx_offload_capa);
474 		return -ENOTSUP;
475 	}
476 
477 	PMD_INIT_FUNC_TRACE();
478 
479 	return 0;
480 }
481 
482 static void
483 em_set_pba(struct e1000_hw *hw)
484 {
485 	uint32_t pba;
486 
487 	/*
488 	 * Packet Buffer Allocation (PBA)
489 	 * Writing PBA sets the receive portion of the buffer
490 	 * the remainder is used for the transmit buffer.
491 	 * Devices before the 82547 had a Packet Buffer of 64K.
492 	 * After the 82547 the buffer was reduced to 40K.
493 	 */
494 	switch (hw->mac.type) {
495 		case e1000_82547:
496 		case e1000_82547_rev_2:
497 		/* 82547: Total Packet Buffer is 40K */
498 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
499 			break;
500 		case e1000_82571:
501 		case e1000_82572:
502 		case e1000_80003es2lan:
503 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
504 			break;
505 		case e1000_82573: /* 82573: Total Packet Buffer is 32K */
506 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
507 			break;
508 		case e1000_82574:
509 		case e1000_82583:
510 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
511 			break;
512 		case e1000_ich8lan:
513 			pba = E1000_PBA_8K;
514 			break;
515 		case e1000_ich9lan:
516 		case e1000_ich10lan:
517 			pba = E1000_PBA_10K;
518 			break;
519 		case e1000_pchlan:
520 		case e1000_pch2lan:
521 		case e1000_pch_lpt:
522 		case e1000_pch_spt:
523 		case e1000_pch_cnp:
524 			pba = E1000_PBA_26K;
525 			break;
526 		default:
527 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
528 	}
529 
530 	E1000_WRITE_REG(hw, E1000_PBA, pba);
531 }
532 
533 static void
534 eth_em_rxtx_control(struct rte_eth_dev *dev,
535 		    bool enable)
536 {
537 	struct e1000_hw *hw =
538 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
539 	uint32_t tctl, rctl;
540 
541 	tctl = E1000_READ_REG(hw, E1000_TCTL);
542 	rctl = E1000_READ_REG(hw, E1000_RCTL);
543 	if (enable) {
544 		/* enable Tx/Rx */
545 		tctl |= E1000_TCTL_EN;
546 		rctl |= E1000_RCTL_EN;
547 	} else {
548 		/* disable Tx/Rx */
549 		tctl &= ~E1000_TCTL_EN;
550 		rctl &= ~E1000_RCTL_EN;
551 	}
552 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
553 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
554 	E1000_WRITE_FLUSH(hw);
555 }
556 
557 static int
558 eth_em_start(struct rte_eth_dev *dev)
559 {
560 	struct e1000_adapter *adapter =
561 		E1000_DEV_PRIVATE(dev->data->dev_private);
562 	struct e1000_hw *hw =
563 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
564 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
565 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
566 	int ret, mask;
567 	uint32_t intr_vector = 0;
568 	uint32_t *speeds;
569 	int num_speeds;
570 	bool autoneg;
571 
572 	PMD_INIT_FUNC_TRACE();
573 
574 	eth_em_stop(dev);
575 
576 	e1000_power_up_phy(hw);
577 
578 	/* Set default PBA value */
579 	em_set_pba(hw);
580 
581 	/* Put the address into the Receive Address Array */
582 	e1000_rar_set(hw, hw->mac.addr, 0);
583 
584 	/*
585 	 * With the 82571 adapter, RAR[0] may be overwritten
586 	 * when the other port is reset, we make a duplicate
587 	 * in RAR[14] for that eventuality, this assures
588 	 * the interface continues to function.
589 	 */
590 	if (hw->mac.type == e1000_82571) {
591 		e1000_set_laa_state_82571(hw, TRUE);
592 		e1000_rar_set(hw, hw->mac.addr, E1000_RAR_ENTRIES - 1);
593 	}
594 
595 	/* Initialize the hardware */
596 	if (em_hardware_init(hw)) {
597 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
598 		return -EIO;
599 	}
600 
601 	E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN);
602 
603 	/* Configure for OS presence */
604 	em_init_manageability(hw);
605 
606 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
607 		intr_vector = dev->data->nb_rx_queues;
608 		if (rte_intr_efd_enable(intr_handle, intr_vector))
609 			return -1;
610 	}
611 
612 	if (rte_intr_dp_is_en(intr_handle)) {
613 		intr_handle->intr_vec =
614 			rte_zmalloc("intr_vec",
615 					dev->data->nb_rx_queues * sizeof(int), 0);
616 		if (intr_handle->intr_vec == NULL) {
617 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
618 						" intr_vec", dev->data->nb_rx_queues);
619 			return -ENOMEM;
620 		}
621 
622 		/* enable rx interrupt */
623 		em_rxq_intr_enable(hw);
624 	}
625 
626 	eth_em_tx_init(dev);
627 
628 	ret = eth_em_rx_init(dev);
629 	if (ret) {
630 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
631 		em_dev_clear_queues(dev);
632 		return ret;
633 	}
634 
635 	e1000_clear_hw_cntrs_base_generic(hw);
636 
637 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
638 			ETH_VLAN_EXTEND_MASK;
639 	ret = eth_em_vlan_offload_set(dev, mask);
640 	if (ret) {
641 		PMD_INIT_LOG(ERR, "Unable to update vlan offload");
642 		em_dev_clear_queues(dev);
643 		return ret;
644 	}
645 
646 	/* Set Interrupt Throttling Rate to maximum allowed value. */
647 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
648 
649 	/* Setup link speed and duplex */
650 	speeds = &dev->data->dev_conf.link_speeds;
651 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
652 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
653 		hw->mac.autoneg = 1;
654 	} else {
655 		num_speeds = 0;
656 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
657 
658 		/* Reset */
659 		hw->phy.autoneg_advertised = 0;
660 
661 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
662 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
663 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
664 			num_speeds = -1;
665 			goto error_invalid_config;
666 		}
667 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
668 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
669 			num_speeds++;
670 		}
671 		if (*speeds & ETH_LINK_SPEED_10M) {
672 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
673 			num_speeds++;
674 		}
675 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
676 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
677 			num_speeds++;
678 		}
679 		if (*speeds & ETH_LINK_SPEED_100M) {
680 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
681 			num_speeds++;
682 		}
683 		if (*speeds & ETH_LINK_SPEED_1G) {
684 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
685 			num_speeds++;
686 		}
687 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
688 			goto error_invalid_config;
689 
690 		/* Set/reset the mac.autoneg based on the link speed,
691 		 * fixed or not
692 		 */
693 		if (!autoneg) {
694 			hw->mac.autoneg = 0;
695 			hw->mac.forced_speed_duplex =
696 					hw->phy.autoneg_advertised;
697 		} else {
698 			hw->mac.autoneg = 1;
699 		}
700 	}
701 
702 	e1000_setup_link(hw);
703 
704 	if (rte_intr_allow_others(intr_handle)) {
705 		/* check if lsc interrupt is enabled */
706 		if (dev->data->dev_conf.intr_conf.lsc != 0) {
707 			ret = eth_em_interrupt_setup(dev);
708 			if (ret) {
709 				PMD_INIT_LOG(ERR, "Unable to setup interrupts");
710 				em_dev_clear_queues(dev);
711 				return ret;
712 			}
713 		}
714 	} else {
715 		rte_intr_callback_unregister(intr_handle,
716 						eth_em_interrupt_handler,
717 						(void *)dev);
718 		if (dev->data->dev_conf.intr_conf.lsc != 0)
719 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
720 				     " no intr multiplexn");
721 	}
722 	/* check if rxq interrupt is enabled */
723 	if (dev->data->dev_conf.intr_conf.rxq != 0)
724 		eth_em_rxq_interrupt_setup(dev);
725 
726 	rte_intr_enable(intr_handle);
727 
728 	adapter->stopped = 0;
729 
730 	eth_em_rxtx_control(dev, true);
731 	eth_em_link_update(dev, 0);
732 
733 	PMD_INIT_LOG(DEBUG, "<<");
734 
735 	return 0;
736 
737 error_invalid_config:
738 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
739 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
740 	em_dev_clear_queues(dev);
741 	return -EINVAL;
742 }
743 
744 /*********************************************************************
745  *
746  *  This routine disables all traffic on the adapter by issuing a
747  *  global reset on the MAC.
748  *
749  **********************************************************************/
750 static void
751 eth_em_stop(struct rte_eth_dev *dev)
752 {
753 	struct rte_eth_link link;
754 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
755 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
756 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
757 
758 	eth_em_rxtx_control(dev, false);
759 	em_rxq_intr_disable(hw);
760 	em_lsc_intr_disable(hw);
761 
762 	e1000_reset_hw(hw);
763 	if (hw->mac.type >= e1000_82544)
764 		E1000_WRITE_REG(hw, E1000_WUC, 0);
765 
766 	/* Power down the phy. Needed to make the link go down */
767 	e1000_power_down_phy(hw);
768 
769 	em_dev_clear_queues(dev);
770 
771 	/* clear the recorded link status */
772 	memset(&link, 0, sizeof(link));
773 	rte_eth_linkstatus_set(dev, &link);
774 
775 	if (!rte_intr_allow_others(intr_handle))
776 		/* resume to the default handler */
777 		rte_intr_callback_register(intr_handle,
778 					   eth_em_interrupt_handler,
779 					   (void *)dev);
780 
781 	/* Clean datapath event and queue/vec mapping */
782 	rte_intr_efd_disable(intr_handle);
783 	if (intr_handle->intr_vec != NULL) {
784 		rte_free(intr_handle->intr_vec);
785 		intr_handle->intr_vec = NULL;
786 	}
787 }
788 
789 static void
790 eth_em_close(struct rte_eth_dev *dev)
791 {
792 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
793 	struct e1000_adapter *adapter =
794 		E1000_DEV_PRIVATE(dev->data->dev_private);
795 
796 	eth_em_stop(dev);
797 	adapter->stopped = 1;
798 	em_dev_free_queues(dev);
799 	e1000_phy_hw_reset(hw);
800 	em_release_manageability(hw);
801 	em_hw_control_release(hw);
802 }
803 
804 static int
805 em_get_rx_buffer_size(struct e1000_hw *hw)
806 {
807 	uint32_t rx_buf_size;
808 
809 	rx_buf_size = ((E1000_READ_REG(hw, E1000_PBA) & UINT16_MAX) << 10);
810 	return rx_buf_size;
811 }
812 
813 /*********************************************************************
814  *
815  *  Initialize the hardware
816  *
817  **********************************************************************/
818 static int
819 em_hardware_init(struct e1000_hw *hw)
820 {
821 	uint32_t rx_buf_size;
822 	int diag;
823 
824 	/* Issue a global reset */
825 	e1000_reset_hw(hw);
826 
827 	/* Let the firmware know the OS is in control */
828 	em_hw_control_acquire(hw);
829 
830 	/*
831 	 * These parameters control the automatic generation (Tx) and
832 	 * response (Rx) to Ethernet PAUSE frames.
833 	 * - High water mark should allow for at least two standard size (1518)
834 	 *   frames to be received after sending an XOFF.
835 	 * - Low water mark works best when it is very near the high water mark.
836 	 *   This allows the receiver to restart by sending XON when it has
837 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
838 	 *   restart after one full frame is pulled from the buffer. There
839 	 *   could be several smaller frames in the buffer and if so they will
840 	 *   not trigger the XON until their total number reduces the buffer
841 	 *   by 1500.
842 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
843 	 */
844 	rx_buf_size = em_get_rx_buffer_size(hw);
845 
846 	hw->fc.high_water = rx_buf_size - PMD_ROUNDUP(ETHER_MAX_LEN * 2, 1024);
847 	hw->fc.low_water = hw->fc.high_water - 1500;
848 
849 	if (hw->mac.type == e1000_80003es2lan)
850 		hw->fc.pause_time = UINT16_MAX;
851 	else
852 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
853 
854 	hw->fc.send_xon = 1;
855 
856 	/* Set Flow control, use the tunable location if sane */
857 	if (em_fc_setting <= e1000_fc_full)
858 		hw->fc.requested_mode = em_fc_setting;
859 	else
860 		hw->fc.requested_mode = e1000_fc_none;
861 
862 	/* Workaround: no TX flow ctrl for PCH */
863 	if (hw->mac.type == e1000_pchlan)
864 		hw->fc.requested_mode = e1000_fc_rx_pause;
865 
866 	/* Override - settings for PCH2LAN, ya its magic :) */
867 	if (hw->mac.type == e1000_pch2lan) {
868 		hw->fc.high_water = 0x5C20;
869 		hw->fc.low_water = 0x5048;
870 		hw->fc.pause_time = 0x0650;
871 		hw->fc.refresh_time = 0x0400;
872 	} else if (hw->mac.type == e1000_pch_lpt ||
873 		   hw->mac.type == e1000_pch_spt ||
874 		   hw->mac.type == e1000_pch_cnp) {
875 		hw->fc.requested_mode = e1000_fc_full;
876 	}
877 
878 	diag = e1000_init_hw(hw);
879 	if (diag < 0)
880 		return diag;
881 	e1000_check_for_link(hw);
882 	return 0;
883 }
884 
885 /* This function is based on em_update_stats_counters() in e1000/if_em.c */
886 static int
887 eth_em_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
888 {
889 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
890 	struct e1000_hw_stats *stats =
891 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
892 	int pause_frames;
893 
894 	if(hw->phy.media_type == e1000_media_type_copper ||
895 			(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
896 		stats->symerrs += E1000_READ_REG(hw,E1000_SYMERRS);
897 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
898 	}
899 
900 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
901 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
902 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
903 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
904 
905 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
906 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
907 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
908 	stats->dc += E1000_READ_REG(hw, E1000_DC);
909 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
910 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
911 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
912 
913 	/*
914 	 * For watchdog management we need to know if we have been
915 	 * paused during the last interval, so capture that here.
916 	 */
917 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
918 	stats->xoffrxc += pause_frames;
919 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
920 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
921 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
922 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
923 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
924 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
925 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
926 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
927 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
928 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
929 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
930 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
931 
932 	/*
933 	 * For the 64-bit byte counters the low dword must be read first.
934 	 * Both registers clear on the read of the high dword.
935 	 */
936 
937 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
938 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
939 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
940 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
941 
942 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
943 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
944 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
945 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
946 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
947 
948 	stats->tor += E1000_READ_REG(hw, E1000_TORH);
949 	stats->tot += E1000_READ_REG(hw, E1000_TOTH);
950 
951 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
952 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
953 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
954 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
955 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
956 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
957 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
958 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
959 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
960 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
961 
962 	/* Interrupt Counts */
963 
964 	if (hw->mac.type >= e1000_82571) {
965 		stats->iac += E1000_READ_REG(hw, E1000_IAC);
966 		stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
967 		stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
968 		stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
969 		stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
970 		stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
971 		stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
972 		stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
973 		stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
974 	}
975 
976 	if (hw->mac.type >= e1000_82543) {
977 		stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
978 		stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
979 		stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
980 		stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
981 		stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
982 		stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
983 	}
984 
985 	if (rte_stats == NULL)
986 		return -EINVAL;
987 
988 	/* Rx Errors */
989 	rte_stats->imissed = stats->mpc;
990 	rte_stats->ierrors = stats->crcerrs +
991 	                     stats->rlec + stats->ruc + stats->roc +
992 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
993 
994 	/* Tx Errors */
995 	rte_stats->oerrors = stats->ecol + stats->latecol;
996 
997 	rte_stats->ipackets = stats->gprc;
998 	rte_stats->opackets = stats->gptc;
999 	rte_stats->ibytes   = stats->gorc;
1000 	rte_stats->obytes   = stats->gotc;
1001 	return 0;
1002 }
1003 
1004 static void
1005 eth_em_stats_reset(struct rte_eth_dev *dev)
1006 {
1007 	struct e1000_hw_stats *hw_stats =
1008 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1009 
1010 	/* HW registers are cleared on read */
1011 	eth_em_stats_get(dev, NULL);
1012 
1013 	/* Reset software totals */
1014 	memset(hw_stats, 0, sizeof(*hw_stats));
1015 }
1016 
1017 static int
1018 eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1019 {
1020 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1021 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1022 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1023 
1024 	em_rxq_intr_enable(hw);
1025 	rte_intr_enable(intr_handle);
1026 
1027 	return 0;
1028 }
1029 
1030 static int
1031 eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1032 {
1033 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1034 
1035 	em_rxq_intr_disable(hw);
1036 
1037 	return 0;
1038 }
1039 
1040 uint32_t
1041 em_get_max_pktlen(struct rte_eth_dev *dev)
1042 {
1043 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1044 
1045 	switch (hw->mac.type) {
1046 	case e1000_82571:
1047 	case e1000_82572:
1048 	case e1000_ich9lan:
1049 	case e1000_ich10lan:
1050 	case e1000_pch2lan:
1051 	case e1000_pch_lpt:
1052 	case e1000_pch_spt:
1053 	case e1000_pch_cnp:
1054 	case e1000_82574:
1055 	case e1000_80003es2lan: /* 9K Jumbo Frame size */
1056 	case e1000_82583:
1057 		return 0x2412;
1058 	case e1000_pchlan:
1059 		return 0x1000;
1060 	/* Adapters that do not support jumbo frames */
1061 	case e1000_ich8lan:
1062 		return ETHER_MAX_LEN;
1063 	default:
1064 		return MAX_JUMBO_FRAME_SIZE;
1065 	}
1066 }
1067 
1068 static void
1069 eth_em_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1070 {
1071 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1072 
1073 	dev_info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1074 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1075 	dev_info->max_rx_pktlen = em_get_max_pktlen(dev);
1076 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1077 
1078 	/*
1079 	 * Starting with 631xESB hw supports 2 TX/RX queues per port.
1080 	 * Unfortunatelly, all these nics have just one TX context.
1081 	 * So we have few choises for TX:
1082 	 * - Use just one TX queue.
1083 	 * - Allow cksum offload only for one TX queue.
1084 	 * - Don't allow TX cksum offload at all.
1085 	 * For now, option #1 was chosen.
1086 	 * To use second RX queue we have to use extended RX descriptor
1087 	 * (Multiple Receive Queues are mutually exclusive with UDP
1088 	 * fragmentation and are not supported when a legacy receive
1089 	 * descriptor format is used).
1090 	 * Which means separate RX routinies - as legacy nics (82540, 82545)
1091 	 * don't support extended RXD.
1092 	 * To avoid it we support just one RX queue for now (no RSS).
1093 	 */
1094 
1095 	dev_info->max_rx_queues = 1;
1096 	dev_info->max_tx_queues = 1;
1097 
1098 	dev_info->rx_queue_offload_capa = em_get_rx_queue_offloads_capa(dev);
1099 	dev_info->rx_offload_capa = em_get_rx_port_offloads_capa(dev) |
1100 				    dev_info->rx_queue_offload_capa;
1101 	dev_info->tx_queue_offload_capa = em_get_tx_queue_offloads_capa(dev);
1102 	dev_info->tx_offload_capa = em_get_tx_port_offloads_capa(dev) |
1103 				    dev_info->tx_queue_offload_capa;
1104 
1105 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1106 		.nb_max = E1000_MAX_RING_DESC,
1107 		.nb_min = E1000_MIN_RING_DESC,
1108 		.nb_align = EM_RXD_ALIGN,
1109 	};
1110 
1111 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1112 		.nb_max = E1000_MAX_RING_DESC,
1113 		.nb_min = E1000_MIN_RING_DESC,
1114 		.nb_align = EM_TXD_ALIGN,
1115 		.nb_seg_max = EM_TX_MAX_SEG,
1116 		.nb_mtu_seg_max = EM_TX_MAX_MTU_SEG,
1117 	};
1118 
1119 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1120 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1121 			ETH_LINK_SPEED_1G;
1122 }
1123 
1124 /* return 0 means link status changed, -1 means not changed */
1125 static int
1126 eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1127 {
1128 	struct e1000_hw *hw =
1129 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1130 	struct rte_eth_link link;
1131 	int link_check, count;
1132 
1133 	link_check = 0;
1134 	hw->mac.get_link_status = 1;
1135 
1136 	/* possible wait-to-complete in up to 9 seconds */
1137 	for (count = 0; count < EM_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
1138 		/* Read the real link status */
1139 		switch (hw->phy.media_type) {
1140 		case e1000_media_type_copper:
1141 			/* Do the work to read phy */
1142 			e1000_check_for_link(hw);
1143 			link_check = !hw->mac.get_link_status;
1144 			break;
1145 
1146 		case e1000_media_type_fiber:
1147 			e1000_check_for_link(hw);
1148 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1149 					E1000_STATUS_LU);
1150 			break;
1151 
1152 		case e1000_media_type_internal_serdes:
1153 			e1000_check_for_link(hw);
1154 			link_check = hw->mac.serdes_has_link;
1155 			break;
1156 
1157 		default:
1158 			break;
1159 		}
1160 		if (link_check || wait_to_complete == 0)
1161 			break;
1162 		rte_delay_ms(EM_LINK_UPDATE_CHECK_INTERVAL);
1163 	}
1164 	memset(&link, 0, sizeof(link));
1165 
1166 	/* Now we check if a transition has happened */
1167 	if (link_check && (link.link_status == ETH_LINK_DOWN)) {
1168 		uint16_t duplex, speed;
1169 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
1170 		link.link_duplex = (duplex == FULL_DUPLEX) ?
1171 				ETH_LINK_FULL_DUPLEX :
1172 				ETH_LINK_HALF_DUPLEX;
1173 		link.link_speed = speed;
1174 		link.link_status = ETH_LINK_UP;
1175 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1176 				ETH_LINK_SPEED_FIXED);
1177 	} else if (!link_check && (link.link_status == ETH_LINK_UP)) {
1178 		link.link_speed = 0;
1179 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
1180 		link.link_status = ETH_LINK_DOWN;
1181 		link.link_autoneg = ETH_LINK_FIXED;
1182 	}
1183 
1184 	return rte_eth_linkstatus_set(dev, &link);
1185 }
1186 
1187 /*
1188  * em_hw_control_acquire sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
1189  * For ASF and Pass Through versions of f/w this means
1190  * that the driver is loaded. For AMT version type f/w
1191  * this means that the network i/f is open.
1192  */
1193 static void
1194 em_hw_control_acquire(struct e1000_hw *hw)
1195 {
1196 	uint32_t ctrl_ext, swsm;
1197 
1198 	/* Let firmware know the driver has taken over */
1199 	if (hw->mac.type == e1000_82573) {
1200 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1201 		E1000_WRITE_REG(hw, E1000_SWSM, swsm | E1000_SWSM_DRV_LOAD);
1202 
1203 	} else {
1204 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1205 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1206 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1207 	}
1208 }
1209 
1210 /*
1211  * em_hw_control_release resets {CTRL_EXTT|FWSM}:DRV_LOAD bit.
1212  * For ASF and Pass Through versions of f/w this means that the
1213  * driver is no longer loaded. For AMT versions of the
1214  * f/w this means that the network i/f is closed.
1215  */
1216 static void
1217 em_hw_control_release(struct e1000_hw *hw)
1218 {
1219 	uint32_t ctrl_ext, swsm;
1220 
1221 	/* Let firmware taken over control of h/w */
1222 	if (hw->mac.type == e1000_82573) {
1223 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1224 		E1000_WRITE_REG(hw, E1000_SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
1225 	} else {
1226 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1227 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1228 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1229 	}
1230 }
1231 
1232 /*
1233  * Bit of a misnomer, what this really means is
1234  * to enable OS management of the system... aka
1235  * to disable special hardware management features.
1236  */
1237 static void
1238 em_init_manageability(struct e1000_hw *hw)
1239 {
1240 	if (e1000_enable_mng_pass_thru(hw)) {
1241 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
1242 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
1243 
1244 		/* disable hardware interception of ARP */
1245 		manc &= ~(E1000_MANC_ARP_EN);
1246 
1247 		/* enable receiving management packets to the host */
1248 		manc |= E1000_MANC_EN_MNG2HOST;
1249 		manc2h |= 1 << 5;  /* Mng Port 623 */
1250 		manc2h |= 1 << 6;  /* Mng Port 664 */
1251 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
1252 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1253 	}
1254 }
1255 
1256 /*
1257  * Give control back to hardware management
1258  * controller if there is one.
1259  */
1260 static void
1261 em_release_manageability(struct e1000_hw *hw)
1262 {
1263 	uint32_t manc;
1264 
1265 	if (e1000_enable_mng_pass_thru(hw)) {
1266 		manc = E1000_READ_REG(hw, E1000_MANC);
1267 
1268 		/* re-enable hardware interception of ARP */
1269 		manc |= E1000_MANC_ARP_EN;
1270 		manc &= ~E1000_MANC_EN_MNG2HOST;
1271 
1272 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1273 	}
1274 }
1275 
1276 static void
1277 eth_em_promiscuous_enable(struct rte_eth_dev *dev)
1278 {
1279 	struct e1000_hw *hw =
1280 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1281 	uint32_t rctl;
1282 
1283 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1284 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1285 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1286 }
1287 
1288 static void
1289 eth_em_promiscuous_disable(struct rte_eth_dev *dev)
1290 {
1291 	struct e1000_hw *hw =
1292 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1293 	uint32_t rctl;
1294 
1295 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1296 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_SBP);
1297 	if (dev->data->all_multicast == 1)
1298 		rctl |= E1000_RCTL_MPE;
1299 	else
1300 		rctl &= (~E1000_RCTL_MPE);
1301 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1302 }
1303 
1304 static void
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 
1316 static void
1317 eth_em_allmulticast_disable(struct rte_eth_dev *dev)
1318 {
1319 	struct e1000_hw *hw =
1320 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1321 	uint32_t rctl;
1322 
1323 	if (dev->data->promiscuous == 1)
1324 		return; /* must remain in all_multicast mode */
1325 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1326 	rctl &= (~E1000_RCTL_MPE);
1327 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1328 }
1329 
1330 static int
1331 eth_em_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1332 {
1333 	struct e1000_hw *hw =
1334 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1335 	struct e1000_vfta * shadow_vfta =
1336 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1337 	uint32_t vfta;
1338 	uint32_t vid_idx;
1339 	uint32_t vid_bit;
1340 
1341 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
1342 			      E1000_VFTA_ENTRY_MASK);
1343 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
1344 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
1345 	if (on)
1346 		vfta |= vid_bit;
1347 	else
1348 		vfta &= ~vid_bit;
1349 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
1350 
1351 	/* update local VFTA copy */
1352 	shadow_vfta->vfta[vid_idx] = vfta;
1353 
1354 	return 0;
1355 }
1356 
1357 static void
1358 em_vlan_hw_filter_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 reg;
1363 
1364 	/* Filter Table Disable */
1365 	reg = E1000_READ_REG(hw, E1000_RCTL);
1366 	reg &= ~E1000_RCTL_CFIEN;
1367 	reg &= ~E1000_RCTL_VFE;
1368 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1369 }
1370 
1371 static void
1372 em_vlan_hw_filter_enable(struct rte_eth_dev *dev)
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 reg;
1379 	int i;
1380 
1381 	/* Filter Table Enable, CFI not used for packet acceptance */
1382 	reg = E1000_READ_REG(hw, E1000_RCTL);
1383 	reg &= ~E1000_RCTL_CFIEN;
1384 	reg |= E1000_RCTL_VFE;
1385 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1386 
1387 	/* restore vfta from local copy */
1388 	for (i = 0; i < IGB_VFTA_SIZE; i++)
1389 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
1390 }
1391 
1392 static void
1393 em_vlan_hw_strip_disable(struct rte_eth_dev *dev)
1394 {
1395 	struct e1000_hw *hw =
1396 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1397 	uint32_t reg;
1398 
1399 	/* VLAN Mode Disable */
1400 	reg = E1000_READ_REG(hw, E1000_CTRL);
1401 	reg &= ~E1000_CTRL_VME;
1402 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1403 
1404 }
1405 
1406 static void
1407 em_vlan_hw_strip_enable(struct rte_eth_dev *dev)
1408 {
1409 	struct e1000_hw *hw =
1410 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1411 	uint32_t reg;
1412 
1413 	/* VLAN Mode Enable */
1414 	reg = E1000_READ_REG(hw, E1000_CTRL);
1415 	reg |= E1000_CTRL_VME;
1416 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1417 }
1418 
1419 static int
1420 eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1421 {
1422 	struct rte_eth_rxmode *rxmode;
1423 
1424 	rxmode = &dev->data->dev_conf.rxmode;
1425 	if(mask & ETH_VLAN_STRIP_MASK){
1426 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1427 			em_vlan_hw_strip_enable(dev);
1428 		else
1429 			em_vlan_hw_strip_disable(dev);
1430 	}
1431 
1432 	if(mask & ETH_VLAN_FILTER_MASK){
1433 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1434 			em_vlan_hw_filter_enable(dev);
1435 		else
1436 			em_vlan_hw_filter_disable(dev);
1437 	}
1438 
1439 	return 0;
1440 }
1441 
1442 /*
1443  * It enables the interrupt mask and then enable the interrupt.
1444  *
1445  * @param dev
1446  *  Pointer to struct rte_eth_dev.
1447  *
1448  * @return
1449  *  - On success, zero.
1450  *  - On failure, a negative value.
1451  */
1452 static int
1453 eth_em_interrupt_setup(struct rte_eth_dev *dev)
1454 {
1455 	uint32_t regval;
1456 	struct e1000_hw *hw =
1457 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1458 
1459 	/* clear interrupt */
1460 	E1000_READ_REG(hw, E1000_ICR);
1461 	regval = E1000_READ_REG(hw, E1000_IMS);
1462 	E1000_WRITE_REG(hw, E1000_IMS, regval | E1000_ICR_LSC);
1463 	return 0;
1464 }
1465 
1466 /*
1467  * It clears the interrupt causes and enables the interrupt.
1468  * It will be called once only during nic initialized.
1469  *
1470  * @param dev
1471  *  Pointer to struct rte_eth_dev.
1472  *
1473  * @return
1474  *  - On success, zero.
1475  *  - On failure, a negative value.
1476  */
1477 static int
1478 eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev)
1479 {
1480 	struct e1000_hw *hw =
1481 	E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1482 
1483 	E1000_READ_REG(hw, E1000_ICR);
1484 	em_rxq_intr_enable(hw);
1485 	return 0;
1486 }
1487 
1488 /*
1489  * It enable receive packet interrupt.
1490  * @param hw
1491  * Pointer to struct e1000_hw
1492  *
1493  * @return
1494  */
1495 static void
1496 em_rxq_intr_enable(struct e1000_hw *hw)
1497 {
1498 	E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_RXT0);
1499 	E1000_WRITE_FLUSH(hw);
1500 }
1501 
1502 /*
1503  * It disabled lsc interrupt.
1504  * @param hw
1505  * Pointer to struct e1000_hw
1506  *
1507  * @return
1508  */
1509 static void
1510 em_lsc_intr_disable(struct e1000_hw *hw)
1511 {
1512 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC);
1513 	E1000_WRITE_FLUSH(hw);
1514 }
1515 
1516 /*
1517  * It disabled receive packet interrupt.
1518  * @param hw
1519  * Pointer to struct e1000_hw
1520  *
1521  * @return
1522  */
1523 static void
1524 em_rxq_intr_disable(struct e1000_hw *hw)
1525 {
1526 	E1000_READ_REG(hw, E1000_ICR);
1527 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_RXT0);
1528 	E1000_WRITE_FLUSH(hw);
1529 }
1530 
1531 /*
1532  * It reads ICR and gets interrupt causes, check it and set a bit flag
1533  * to update link status.
1534  *
1535  * @param dev
1536  *  Pointer to struct rte_eth_dev.
1537  *
1538  * @return
1539  *  - On success, zero.
1540  *  - On failure, a negative value.
1541  */
1542 static int
1543 eth_em_interrupt_get_status(struct rte_eth_dev *dev)
1544 {
1545 	uint32_t icr;
1546 	struct e1000_hw *hw =
1547 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1548 	struct e1000_interrupt *intr =
1549 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1550 
1551 	/* read-on-clear nic registers here */
1552 	icr = E1000_READ_REG(hw, E1000_ICR);
1553 	if (icr & E1000_ICR_LSC) {
1554 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1555 	}
1556 
1557 	return 0;
1558 }
1559 
1560 /*
1561  * It executes link_update after knowing an interrupt is prsent.
1562  *
1563  * @param dev
1564  *  Pointer to struct rte_eth_dev.
1565  *
1566  * @return
1567  *  - On success, zero.
1568  *  - On failure, a negative value.
1569  */
1570 static int
1571 eth_em_interrupt_action(struct rte_eth_dev *dev,
1572 			struct rte_intr_handle *intr_handle)
1573 {
1574 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1575 	struct e1000_hw *hw =
1576 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1577 	struct e1000_interrupt *intr =
1578 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1579 	struct rte_eth_link link;
1580 	int ret;
1581 
1582 	if (!(intr->flags & E1000_FLAG_NEED_LINK_UPDATE))
1583 		return -1;
1584 
1585 	intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
1586 	rte_intr_enable(intr_handle);
1587 
1588 	/* set get_link_status to check register later */
1589 	hw->mac.get_link_status = 1;
1590 	ret = eth_em_link_update(dev, 0);
1591 
1592 	/* check if link has changed */
1593 	if (ret < 0)
1594 		return 0;
1595 
1596 	rte_eth_linkstatus_get(dev, &link);
1597 
1598 	if (link.link_status) {
1599 		PMD_INIT_LOG(INFO, " Port %d: Link Up - speed %u Mbps - %s",
1600 			     dev->data->port_id, link.link_speed,
1601 			     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
1602 			     "full-duplex" : "half-duplex");
1603 	} else {
1604 		PMD_INIT_LOG(INFO, " Port %d: Link Down", dev->data->port_id);
1605 	}
1606 	PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
1607 		     pci_dev->addr.domain, pci_dev->addr.bus,
1608 		     pci_dev->addr.devid, pci_dev->addr.function);
1609 
1610 	return 0;
1611 }
1612 
1613 /**
1614  * Interrupt handler which shall be registered at first.
1615  *
1616  * @param handle
1617  *  Pointer to interrupt handle.
1618  * @param param
1619  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1620  *
1621  * @return
1622  *  void
1623  */
1624 static void
1625 eth_em_interrupt_handler(void *param)
1626 {
1627 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1628 
1629 	eth_em_interrupt_get_status(dev);
1630 	eth_em_interrupt_action(dev, dev->intr_handle);
1631 	_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
1632 }
1633 
1634 static int
1635 eth_em_led_on(struct rte_eth_dev *dev)
1636 {
1637 	struct e1000_hw *hw;
1638 
1639 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1640 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1641 }
1642 
1643 static int
1644 eth_em_led_off(struct rte_eth_dev *dev)
1645 {
1646 	struct e1000_hw *hw;
1647 
1648 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1649 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1650 }
1651 
1652 static int
1653 eth_em_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1654 {
1655 	struct e1000_hw *hw;
1656 	uint32_t ctrl;
1657 	int tx_pause;
1658 	int rx_pause;
1659 
1660 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1661 	fc_conf->pause_time = hw->fc.pause_time;
1662 	fc_conf->high_water = hw->fc.high_water;
1663 	fc_conf->low_water = hw->fc.low_water;
1664 	fc_conf->send_xon = hw->fc.send_xon;
1665 	fc_conf->autoneg = hw->mac.autoneg;
1666 
1667 	/*
1668 	 * Return rx_pause and tx_pause status according to actual setting of
1669 	 * the TFCE and RFCE bits in the CTRL register.
1670 	 */
1671 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
1672 	if (ctrl & E1000_CTRL_TFCE)
1673 		tx_pause = 1;
1674 	else
1675 		tx_pause = 0;
1676 
1677 	if (ctrl & E1000_CTRL_RFCE)
1678 		rx_pause = 1;
1679 	else
1680 		rx_pause = 0;
1681 
1682 	if (rx_pause && tx_pause)
1683 		fc_conf->mode = RTE_FC_FULL;
1684 	else if (rx_pause)
1685 		fc_conf->mode = RTE_FC_RX_PAUSE;
1686 	else if (tx_pause)
1687 		fc_conf->mode = RTE_FC_TX_PAUSE;
1688 	else
1689 		fc_conf->mode = RTE_FC_NONE;
1690 
1691 	return 0;
1692 }
1693 
1694 static int
1695 eth_em_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1696 {
1697 	struct e1000_hw *hw;
1698 	int err;
1699 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
1700 		e1000_fc_none,
1701 		e1000_fc_rx_pause,
1702 		e1000_fc_tx_pause,
1703 		e1000_fc_full
1704 	};
1705 	uint32_t rx_buf_size;
1706 	uint32_t max_high_water;
1707 	uint32_t rctl;
1708 
1709 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1710 	if (fc_conf->autoneg != hw->mac.autoneg)
1711 		return -ENOTSUP;
1712 	rx_buf_size = em_get_rx_buffer_size(hw);
1713 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
1714 
1715 	/* At least reserve one Ethernet frame for watermark */
1716 	max_high_water = rx_buf_size - ETHER_MAX_LEN;
1717 	if ((fc_conf->high_water > max_high_water) ||
1718 	    (fc_conf->high_water < fc_conf->low_water)) {
1719 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
1720 		PMD_INIT_LOG(ERR, "high water must <= 0x%x", max_high_water);
1721 		return -EINVAL;
1722 	}
1723 
1724 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
1725 	hw->fc.pause_time     = fc_conf->pause_time;
1726 	hw->fc.high_water     = fc_conf->high_water;
1727 	hw->fc.low_water      = fc_conf->low_water;
1728 	hw->fc.send_xon	      = fc_conf->send_xon;
1729 
1730 	err = e1000_setup_link_generic(hw);
1731 	if (err == E1000_SUCCESS) {
1732 
1733 		/* check if we want to forward MAC frames - driver doesn't have native
1734 		 * capability to do that, so we'll write the registers ourselves */
1735 
1736 		rctl = E1000_READ_REG(hw, E1000_RCTL);
1737 
1738 		/* set or clear MFLCN.PMCF bit depending on configuration */
1739 		if (fc_conf->mac_ctrl_frame_fwd != 0)
1740 			rctl |= E1000_RCTL_PMCF;
1741 		else
1742 			rctl &= ~E1000_RCTL_PMCF;
1743 
1744 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1745 		E1000_WRITE_FLUSH(hw);
1746 
1747 		return 0;
1748 	}
1749 
1750 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
1751 	return -EIO;
1752 }
1753 
1754 static int
1755 eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
1756 		uint32_t index, __rte_unused uint32_t pool)
1757 {
1758 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1759 
1760 	return e1000_rar_set(hw, mac_addr->addr_bytes, index);
1761 }
1762 
1763 static void
1764 eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1765 {
1766 	uint8_t addr[ETHER_ADDR_LEN];
1767 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1768 
1769 	memset(addr, 0, sizeof(addr));
1770 
1771 	e1000_rar_set(hw, addr, index);
1772 }
1773 
1774 static int
1775 eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1776 {
1777 	struct rte_eth_dev_info dev_info;
1778 	struct e1000_hw *hw;
1779 	uint32_t frame_size;
1780 	uint32_t rctl;
1781 
1782 	eth_em_infos_get(dev, &dev_info);
1783 	frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + VLAN_TAG_SIZE;
1784 
1785 	/* check that mtu is within the allowed range */
1786 	if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
1787 		return -EINVAL;
1788 
1789 	/* refuse mtu that requires the support of scattered packets when this
1790 	 * feature has not been enabled before. */
1791 	if (!dev->data->scattered_rx &&
1792 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
1793 		return -EINVAL;
1794 
1795 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1796 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1797 
1798 	/* switch to jumbo mode if needed */
1799 	if (frame_size > ETHER_MAX_LEN) {
1800 		dev->data->dev_conf.rxmode.offloads |=
1801 			DEV_RX_OFFLOAD_JUMBO_FRAME;
1802 		rctl |= E1000_RCTL_LPE;
1803 	} else {
1804 		dev->data->dev_conf.rxmode.offloads &=
1805 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
1806 		rctl &= ~E1000_RCTL_LPE;
1807 	}
1808 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1809 
1810 	/* update max frame size */
1811 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1812 	return 0;
1813 }
1814 
1815 static int
1816 eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
1817 			struct ether_addr *mc_addr_set,
1818 			uint32_t nb_mc_addr)
1819 {
1820 	struct e1000_hw *hw;
1821 
1822 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1823 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1824 	return 0;
1825 }
1826 
1827 RTE_PMD_REGISTER_PCI(net_e1000_em, rte_em_pmd);
1828 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_em, pci_id_em_map);
1829 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_em, "* igb_uio | uio_pci_generic | vfio-pci");
1830 
1831 RTE_INIT(e1000_init_log);
1832 static void
1833 e1000_init_log(void)
1834 {
1835 	e1000_logtype_init = rte_log_register("pmd.net.e1000.init");
1836 	if (e1000_logtype_init >= 0)
1837 		rte_log_set_level(e1000_logtype_init, RTE_LOG_NOTICE);
1838 	e1000_logtype_driver = rte_log_register("pmd.net.e1000.driver");
1839 	if (e1000_logtype_driver >= 0)
1840 		rte_log_set_level(e1000_logtype_driver, RTE_LOG_NOTICE);
1841 }
1842