xref: /dpdk/drivers/net/e1000/em_ethdev.c (revision 985e7673)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2016 Intel Corporation
3  */
4 
5 #include <sys/queue.h>
6 #include <stdio.h>
7 #include <errno.h>
8 #include <stdint.h>
9 #include <stdarg.h>
10 
11 #include <rte_common.h>
12 #include <rte_interrupts.h>
13 #include <rte_byteorder.h>
14 #include <rte_debug.h>
15 #include <rte_pci.h>
16 #include <rte_bus_pci.h>
17 #include <rte_ether.h>
18 #include <ethdev_driver.h>
19 #include <ethdev_pci.h>
20 #include <rte_memory.h>
21 #include <rte_eal.h>
22 #include <rte_malloc.h>
23 #include <rte_dev.h>
24 
25 #include "e1000_logs.h"
26 #include "base/e1000_api.h"
27 #include "e1000_ethdev.h"
28 
29 #define EM_EIAC			0x000DC
30 
31 #define PMD_ROUNDUP(x,y)	(((x) + (y) - 1)/(y) * (y))
32 
33 
34 static int eth_em_configure(struct rte_eth_dev *dev);
35 static int eth_em_start(struct rte_eth_dev *dev);
36 static int eth_em_stop(struct rte_eth_dev *dev);
37 static int eth_em_close(struct rte_eth_dev *dev);
38 static int eth_em_promiscuous_enable(struct rte_eth_dev *dev);
39 static int eth_em_promiscuous_disable(struct rte_eth_dev *dev);
40 static int eth_em_allmulticast_enable(struct rte_eth_dev *dev);
41 static int eth_em_allmulticast_disable(struct rte_eth_dev *dev);
42 static int eth_em_link_update(struct rte_eth_dev *dev,
43 				int wait_to_complete);
44 static int eth_em_stats_get(struct rte_eth_dev *dev,
45 				struct rte_eth_stats *rte_stats);
46 static int eth_em_stats_reset(struct rte_eth_dev *dev);
47 static int eth_em_infos_get(struct rte_eth_dev *dev,
48 				struct rte_eth_dev_info *dev_info);
49 static int eth_em_flow_ctrl_get(struct rte_eth_dev *dev,
50 				struct rte_eth_fc_conf *fc_conf);
51 static int eth_em_flow_ctrl_set(struct rte_eth_dev *dev,
52 				struct rte_eth_fc_conf *fc_conf);
53 static int eth_em_interrupt_setup(struct rte_eth_dev *dev);
54 static int eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev);
55 static int eth_em_interrupt_get_status(struct rte_eth_dev *dev);
56 static int eth_em_interrupt_action(struct rte_eth_dev *dev,
57 				   struct rte_intr_handle *handle);
58 static void eth_em_interrupt_handler(void *param);
59 
60 static int em_hw_init(struct e1000_hw *hw);
61 static int em_hardware_init(struct e1000_hw *hw);
62 static void em_hw_control_acquire(struct e1000_hw *hw);
63 static void em_hw_control_release(struct e1000_hw *hw);
64 static void em_init_manageability(struct e1000_hw *hw);
65 static void em_release_manageability(struct e1000_hw *hw);
66 
67 static int eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
68 
69 static int eth_em_vlan_filter_set(struct rte_eth_dev *dev,
70 		uint16_t vlan_id, int on);
71 static int eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask);
72 static void em_vlan_hw_filter_enable(struct rte_eth_dev *dev);
73 static void em_vlan_hw_filter_disable(struct rte_eth_dev *dev);
74 static void em_vlan_hw_strip_enable(struct rte_eth_dev *dev);
75 static void em_vlan_hw_strip_disable(struct rte_eth_dev *dev);
76 
77 /*
78 static void eth_em_vlan_filter_set(struct rte_eth_dev *dev,
79 					uint16_t vlan_id, int on);
80 */
81 
82 static int eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
83 static int eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
84 static void em_lsc_intr_disable(struct e1000_hw *hw);
85 static void em_rxq_intr_enable(struct e1000_hw *hw);
86 static void em_rxq_intr_disable(struct e1000_hw *hw);
87 
88 static int eth_em_led_on(struct rte_eth_dev *dev);
89 static int eth_em_led_off(struct rte_eth_dev *dev);
90 
91 static int em_get_rx_buffer_size(struct e1000_hw *hw);
92 static int eth_em_rar_set(struct rte_eth_dev *dev,
93 			struct rte_ether_addr *mac_addr,
94 			uint32_t index, uint32_t pool);
95 static void eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index);
96 static int eth_em_default_mac_addr_set(struct rte_eth_dev *dev,
97 					 struct rte_ether_addr *addr);
98 
99 static int eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
100 				   struct rte_ether_addr *mc_addr_set,
101 				   uint32_t nb_mc_addr);
102 
103 #define EM_FC_PAUSE_TIME 0x0680
104 #define EM_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
105 #define EM_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
106 
107 static enum e1000_fc_mode em_fc_setting = e1000_fc_full;
108 
109 /*
110  * The set of PCI devices this driver supports
111  */
112 static const struct rte_pci_id pci_id_em_map[] = {
113 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82540EM) },
114 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_COPPER) },
115 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_FIBER) },
116 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_COPPER) },
117 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_FIBER) },
118 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_QUAD_COPPER) },
119 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_COPPER) },
120 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_FIBER) },
121 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES) },
122 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_DUAL) },
123 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_QUAD) },
124 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER) },
125 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571PT_QUAD_COPPER) },
126 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_FIBER) },
127 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER_LP) },
128 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_COPPER) },
129 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_FIBER) },
130 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_SERDES) },
131 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI) },
132 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82573L) },
133 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574L) },
134 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574LA) },
135 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82583V) },
136 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH2_LV_LM) },
137 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_LM) },
138 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_V) },
139 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_LM) },
140 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_V) },
141 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM2) },
142 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V2) },
143 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM3) },
144 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V3) },
145 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM) },
146 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V) },
147 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM2) },
148 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V2) },
149 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LBG_I219_LM3) },
150 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM4) },
151 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V4) },
152 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM5) },
153 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V5) },
154 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM6) },
155 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V6) },
156 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM7) },
157 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V7) },
158 	{ .vendor_id = 0, /* sentinel */ },
159 };
160 
161 static const struct eth_dev_ops eth_em_ops = {
162 	.dev_configure        = eth_em_configure,
163 	.dev_start            = eth_em_start,
164 	.dev_stop             = eth_em_stop,
165 	.dev_close            = eth_em_close,
166 	.promiscuous_enable   = eth_em_promiscuous_enable,
167 	.promiscuous_disable  = eth_em_promiscuous_disable,
168 	.allmulticast_enable  = eth_em_allmulticast_enable,
169 	.allmulticast_disable = eth_em_allmulticast_disable,
170 	.link_update          = eth_em_link_update,
171 	.stats_get            = eth_em_stats_get,
172 	.stats_reset          = eth_em_stats_reset,
173 	.dev_infos_get        = eth_em_infos_get,
174 	.mtu_set              = eth_em_mtu_set,
175 	.vlan_filter_set      = eth_em_vlan_filter_set,
176 	.vlan_offload_set     = eth_em_vlan_offload_set,
177 	.rx_queue_setup       = eth_em_rx_queue_setup,
178 	.rx_queue_release     = eth_em_rx_queue_release,
179 	.tx_queue_setup       = eth_em_tx_queue_setup,
180 	.tx_queue_release     = eth_em_tx_queue_release,
181 	.rx_queue_intr_enable = eth_em_rx_queue_intr_enable,
182 	.rx_queue_intr_disable = eth_em_rx_queue_intr_disable,
183 	.dev_led_on           = eth_em_led_on,
184 	.dev_led_off          = eth_em_led_off,
185 	.flow_ctrl_get        = eth_em_flow_ctrl_get,
186 	.flow_ctrl_set        = eth_em_flow_ctrl_set,
187 	.mac_addr_set         = eth_em_default_mac_addr_set,
188 	.mac_addr_add         = eth_em_rar_set,
189 	.mac_addr_remove      = eth_em_rar_clear,
190 	.set_mc_addr_list     = eth_em_set_mc_addr_list,
191 	.rxq_info_get         = em_rxq_info_get,
192 	.txq_info_get         = em_txq_info_get,
193 };
194 
195 
196 /**
197  *  eth_em_dev_is_ich8 - Check for ICH8 device
198  *  @hw: pointer to the HW structure
199  *
200  *  return TRUE for ICH8, otherwise FALSE
201  **/
202 static bool
203 eth_em_dev_is_ich8(struct e1000_hw *hw)
204 {
205 	DEBUGFUNC("eth_em_dev_is_ich8");
206 
207 	switch (hw->device_id) {
208 	case E1000_DEV_ID_PCH2_LV_LM:
209 	case E1000_DEV_ID_PCH_LPT_I217_LM:
210 	case E1000_DEV_ID_PCH_LPT_I217_V:
211 	case E1000_DEV_ID_PCH_LPTLP_I218_LM:
212 	case E1000_DEV_ID_PCH_LPTLP_I218_V:
213 	case E1000_DEV_ID_PCH_I218_V2:
214 	case E1000_DEV_ID_PCH_I218_LM2:
215 	case E1000_DEV_ID_PCH_I218_V3:
216 	case E1000_DEV_ID_PCH_I218_LM3:
217 	case E1000_DEV_ID_PCH_SPT_I219_LM:
218 	case E1000_DEV_ID_PCH_SPT_I219_V:
219 	case E1000_DEV_ID_PCH_SPT_I219_LM2:
220 	case E1000_DEV_ID_PCH_SPT_I219_V2:
221 	case E1000_DEV_ID_PCH_LBG_I219_LM3:
222 	case E1000_DEV_ID_PCH_SPT_I219_LM4:
223 	case E1000_DEV_ID_PCH_SPT_I219_V4:
224 	case E1000_DEV_ID_PCH_SPT_I219_LM5:
225 	case E1000_DEV_ID_PCH_SPT_I219_V5:
226 	case E1000_DEV_ID_PCH_CNP_I219_LM6:
227 	case E1000_DEV_ID_PCH_CNP_I219_V6:
228 	case E1000_DEV_ID_PCH_CNP_I219_LM7:
229 	case E1000_DEV_ID_PCH_CNP_I219_V7:
230 		return 1;
231 	default:
232 		return 0;
233 	}
234 }
235 
236 static int
237 eth_em_dev_init(struct rte_eth_dev *eth_dev)
238 {
239 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
240 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
241 	struct e1000_adapter *adapter =
242 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
243 	struct e1000_hw *hw =
244 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
245 	struct e1000_vfta * shadow_vfta =
246 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
247 
248 	eth_dev->dev_ops = &eth_em_ops;
249 	eth_dev->rx_queue_count = eth_em_rx_queue_count;
250 	eth_dev->rx_descriptor_status = eth_em_rx_descriptor_status;
251 	eth_dev->tx_descriptor_status = eth_em_tx_descriptor_status;
252 	eth_dev->rx_pkt_burst = (eth_rx_burst_t)&eth_em_recv_pkts;
253 	eth_dev->tx_pkt_burst = (eth_tx_burst_t)&eth_em_xmit_pkts;
254 	eth_dev->tx_pkt_prepare = (eth_tx_prep_t)&eth_em_prep_pkts;
255 
256 	/* for secondary processes, we don't initialise any further as primary
257 	 * has already done this work. Only check we don't need a different
258 	 * RX function */
259 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
260 		if (eth_dev->data->scattered_rx)
261 			eth_dev->rx_pkt_burst =
262 				(eth_rx_burst_t)&eth_em_recv_scattered_pkts;
263 		return 0;
264 	}
265 
266 	rte_eth_copy_pci_info(eth_dev, pci_dev);
267 
268 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
269 	hw->device_id = pci_dev->id.device_id;
270 	adapter->stopped = 0;
271 
272 	/* For ICH8 support we'll need to map the flash memory BAR */
273 	if (eth_em_dev_is_ich8(hw))
274 		hw->flash_address = (void *)pci_dev->mem_resource[1].addr;
275 
276 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS ||
277 			em_hw_init(hw) != 0) {
278 		PMD_INIT_LOG(ERR, "port_id %d vendorID=0x%x deviceID=0x%x: "
279 			"failed to init HW",
280 			eth_dev->data->port_id, pci_dev->id.vendor_id,
281 			pci_dev->id.device_id);
282 		return -ENODEV;
283 	}
284 
285 	/* Allocate memory for storing MAC addresses */
286 	eth_dev->data->mac_addrs = rte_zmalloc("e1000", RTE_ETHER_ADDR_LEN *
287 			hw->mac.rar_entry_count, 0);
288 	if (eth_dev->data->mac_addrs == NULL) {
289 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
290 			"store MAC addresses",
291 			RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
292 		return -ENOMEM;
293 	}
294 
295 	/* Copy the permanent MAC address */
296 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
297 		eth_dev->data->mac_addrs);
298 
299 	/* initialize the vfta */
300 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
301 
302 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
303 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
304 		     pci_dev->id.device_id);
305 
306 	rte_intr_callback_register(intr_handle,
307 				   eth_em_interrupt_handler, eth_dev);
308 
309 	return 0;
310 }
311 
312 static int
313 eth_em_dev_uninit(struct rte_eth_dev *eth_dev)
314 {
315 	PMD_INIT_FUNC_TRACE();
316 
317 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
318 		return 0;
319 
320 	eth_em_close(eth_dev);
321 
322 	return 0;
323 }
324 
325 static int eth_em_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
326 	struct rte_pci_device *pci_dev)
327 {
328 	return rte_eth_dev_pci_generic_probe(pci_dev,
329 		sizeof(struct e1000_adapter), eth_em_dev_init);
330 }
331 
332 static int eth_em_pci_remove(struct rte_pci_device *pci_dev)
333 {
334 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_em_dev_uninit);
335 }
336 
337 static struct rte_pci_driver rte_em_pmd = {
338 	.id_table = pci_id_em_map,
339 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
340 	.probe = eth_em_pci_probe,
341 	.remove = eth_em_pci_remove,
342 };
343 
344 static int
345 em_hw_init(struct e1000_hw *hw)
346 {
347 	int diag;
348 
349 	diag = hw->mac.ops.init_params(hw);
350 	if (diag != 0) {
351 		PMD_INIT_LOG(ERR, "MAC Initialization Error");
352 		return diag;
353 	}
354 	diag = hw->nvm.ops.init_params(hw);
355 	if (diag != 0) {
356 		PMD_INIT_LOG(ERR, "NVM Initialization Error");
357 		return diag;
358 	}
359 	diag = hw->phy.ops.init_params(hw);
360 	if (diag != 0) {
361 		PMD_INIT_LOG(ERR, "PHY Initialization Error");
362 		return diag;
363 	}
364 	(void) e1000_get_bus_info(hw);
365 
366 	hw->mac.autoneg = 1;
367 	hw->phy.autoneg_wait_to_complete = 0;
368 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
369 
370 	e1000_init_script_state_82541(hw, TRUE);
371 	e1000_set_tbi_compatibility_82543(hw, TRUE);
372 
373 	/* Copper options */
374 	if (hw->phy.media_type == e1000_media_type_copper) {
375 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
376 		hw->phy.disable_polarity_correction = 0;
377 		hw->phy.ms_type = e1000_ms_hw_default;
378 	}
379 
380 	/*
381 	 * Start from a known state, this is important in reading the nvm
382 	 * and mac from that.
383 	 */
384 	e1000_reset_hw(hw);
385 
386 	/* Make sure we have a good EEPROM before we read from it */
387 	if (e1000_validate_nvm_checksum(hw) < 0) {
388 		/*
389 		 * Some PCI-E parts fail the first check due to
390 		 * the link being in sleep state, call it again,
391 		 * if it fails a second time its a real issue.
392 		 */
393 		diag = e1000_validate_nvm_checksum(hw);
394 		if (diag < 0) {
395 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
396 			goto error;
397 		}
398 	}
399 
400 	/* Read the permanent MAC address out of the EEPROM */
401 	diag = e1000_read_mac_addr(hw);
402 	if (diag != 0) {
403 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
404 		goto error;
405 	}
406 
407 	/* Now initialize the hardware */
408 	diag = em_hardware_init(hw);
409 	if (diag != 0) {
410 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
411 		goto error;
412 	}
413 
414 	hw->mac.get_link_status = 1;
415 
416 	/* Indicate SOL/IDER usage */
417 	diag = e1000_check_reset_block(hw);
418 	if (diag < 0) {
419 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to "
420 			"SOL/IDER session");
421 	}
422 	return 0;
423 
424 error:
425 	em_hw_control_release(hw);
426 	return diag;
427 }
428 
429 static int
430 eth_em_configure(struct rte_eth_dev *dev)
431 {
432 	struct e1000_interrupt *intr =
433 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
434 
435 	PMD_INIT_FUNC_TRACE();
436 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
437 
438 	PMD_INIT_FUNC_TRACE();
439 
440 	return 0;
441 }
442 
443 static void
444 em_set_pba(struct e1000_hw *hw)
445 {
446 	uint32_t pba;
447 
448 	/*
449 	 * Packet Buffer Allocation (PBA)
450 	 * Writing PBA sets the receive portion of the buffer
451 	 * the remainder is used for the transmit buffer.
452 	 * Devices before the 82547 had a Packet Buffer of 64K.
453 	 * After the 82547 the buffer was reduced to 40K.
454 	 */
455 	switch (hw->mac.type) {
456 		case e1000_82547:
457 		case e1000_82547_rev_2:
458 		/* 82547: Total Packet Buffer is 40K */
459 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
460 			break;
461 		case e1000_82571:
462 		case e1000_82572:
463 		case e1000_80003es2lan:
464 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
465 			break;
466 		case e1000_82573: /* 82573: Total Packet Buffer is 32K */
467 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
468 			break;
469 		case e1000_82574:
470 		case e1000_82583:
471 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
472 			break;
473 		case e1000_ich8lan:
474 			pba = E1000_PBA_8K;
475 			break;
476 		case e1000_ich9lan:
477 		case e1000_ich10lan:
478 			pba = E1000_PBA_10K;
479 			break;
480 		case e1000_pchlan:
481 		case e1000_pch2lan:
482 		case e1000_pch_lpt:
483 		case e1000_pch_spt:
484 		case e1000_pch_cnp:
485 			pba = E1000_PBA_26K;
486 			break;
487 		default:
488 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
489 	}
490 
491 	E1000_WRITE_REG(hw, E1000_PBA, pba);
492 }
493 
494 static void
495 eth_em_rxtx_control(struct rte_eth_dev *dev,
496 		    bool enable)
497 {
498 	struct e1000_hw *hw =
499 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
500 	uint32_t tctl, rctl;
501 
502 	tctl = E1000_READ_REG(hw, E1000_TCTL);
503 	rctl = E1000_READ_REG(hw, E1000_RCTL);
504 	if (enable) {
505 		/* enable Tx/Rx */
506 		tctl |= E1000_TCTL_EN;
507 		rctl |= E1000_RCTL_EN;
508 	} else {
509 		/* disable Tx/Rx */
510 		tctl &= ~E1000_TCTL_EN;
511 		rctl &= ~E1000_RCTL_EN;
512 	}
513 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
514 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
515 	E1000_WRITE_FLUSH(hw);
516 }
517 
518 static int
519 eth_em_start(struct rte_eth_dev *dev)
520 {
521 	struct e1000_adapter *adapter =
522 		E1000_DEV_PRIVATE(dev->data->dev_private);
523 	struct e1000_hw *hw =
524 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
525 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
526 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
527 	int ret, mask;
528 	uint32_t intr_vector = 0;
529 	uint32_t *speeds;
530 	int num_speeds;
531 	bool autoneg;
532 
533 	PMD_INIT_FUNC_TRACE();
534 
535 	ret = eth_em_stop(dev);
536 	if (ret != 0)
537 		return ret;
538 
539 	e1000_power_up_phy(hw);
540 
541 	/* Set default PBA value */
542 	em_set_pba(hw);
543 
544 	/* Put the address into the Receive Address Array */
545 	e1000_rar_set(hw, hw->mac.addr, 0);
546 
547 	/*
548 	 * With the 82571 adapter, RAR[0] may be overwritten
549 	 * when the other port is reset, we make a duplicate
550 	 * in RAR[14] for that eventuality, this assures
551 	 * the interface continues to function.
552 	 */
553 	if (hw->mac.type == e1000_82571) {
554 		e1000_set_laa_state_82571(hw, TRUE);
555 		e1000_rar_set(hw, hw->mac.addr, E1000_RAR_ENTRIES - 1);
556 	}
557 
558 	/* Initialize the hardware */
559 	if (em_hardware_init(hw)) {
560 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
561 		return -EIO;
562 	}
563 
564 	E1000_WRITE_REG(hw, E1000_VET, RTE_ETHER_TYPE_VLAN);
565 
566 	/* Configure for OS presence */
567 	em_init_manageability(hw);
568 
569 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
570 		intr_vector = dev->data->nb_rx_queues;
571 		if (rte_intr_efd_enable(intr_handle, intr_vector))
572 			return -1;
573 	}
574 
575 	if (rte_intr_dp_is_en(intr_handle)) {
576 		intr_handle->intr_vec =
577 			rte_zmalloc("intr_vec",
578 					dev->data->nb_rx_queues * sizeof(int), 0);
579 		if (intr_handle->intr_vec == NULL) {
580 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
581 						" intr_vec", dev->data->nb_rx_queues);
582 			return -ENOMEM;
583 		}
584 
585 		/* enable rx interrupt */
586 		em_rxq_intr_enable(hw);
587 	}
588 
589 	eth_em_tx_init(dev);
590 
591 	ret = eth_em_rx_init(dev);
592 	if (ret) {
593 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
594 		em_dev_clear_queues(dev);
595 		return ret;
596 	}
597 
598 	e1000_clear_hw_cntrs_base_generic(hw);
599 
600 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
601 			ETH_VLAN_EXTEND_MASK;
602 	ret = eth_em_vlan_offload_set(dev, mask);
603 	if (ret) {
604 		PMD_INIT_LOG(ERR, "Unable to update vlan offload");
605 		em_dev_clear_queues(dev);
606 		return ret;
607 	}
608 
609 	/* Set Interrupt Throttling Rate to maximum allowed value. */
610 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
611 
612 	/* Setup link speed and duplex */
613 	speeds = &dev->data->dev_conf.link_speeds;
614 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
615 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
616 		hw->mac.autoneg = 1;
617 	} else {
618 		num_speeds = 0;
619 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
620 
621 		/* Reset */
622 		hw->phy.autoneg_advertised = 0;
623 
624 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
625 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
626 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
627 			num_speeds = -1;
628 			goto error_invalid_config;
629 		}
630 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
631 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
632 			num_speeds++;
633 		}
634 		if (*speeds & ETH_LINK_SPEED_10M) {
635 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
636 			num_speeds++;
637 		}
638 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
639 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
640 			num_speeds++;
641 		}
642 		if (*speeds & ETH_LINK_SPEED_100M) {
643 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
644 			num_speeds++;
645 		}
646 		if (*speeds & ETH_LINK_SPEED_1G) {
647 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
648 			num_speeds++;
649 		}
650 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
651 			goto error_invalid_config;
652 
653 		/* Set/reset the mac.autoneg based on the link speed,
654 		 * fixed or not
655 		 */
656 		if (!autoneg) {
657 			hw->mac.autoneg = 0;
658 			hw->mac.forced_speed_duplex =
659 					hw->phy.autoneg_advertised;
660 		} else {
661 			hw->mac.autoneg = 1;
662 		}
663 	}
664 
665 	e1000_setup_link(hw);
666 
667 	if (rte_intr_allow_others(intr_handle)) {
668 		/* check if lsc interrupt is enabled */
669 		if (dev->data->dev_conf.intr_conf.lsc != 0) {
670 			ret = eth_em_interrupt_setup(dev);
671 			if (ret) {
672 				PMD_INIT_LOG(ERR, "Unable to setup interrupts");
673 				em_dev_clear_queues(dev);
674 				return ret;
675 			}
676 		}
677 	} else {
678 		rte_intr_callback_unregister(intr_handle,
679 						eth_em_interrupt_handler,
680 						(void *)dev);
681 		if (dev->data->dev_conf.intr_conf.lsc != 0)
682 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
683 				     " no intr multiplexn");
684 	}
685 	/* check if rxq interrupt is enabled */
686 	if (dev->data->dev_conf.intr_conf.rxq != 0)
687 		eth_em_rxq_interrupt_setup(dev);
688 
689 	rte_intr_enable(intr_handle);
690 
691 	adapter->stopped = 0;
692 
693 	eth_em_rxtx_control(dev, true);
694 	eth_em_link_update(dev, 0);
695 
696 	PMD_INIT_LOG(DEBUG, "<<");
697 
698 	return 0;
699 
700 error_invalid_config:
701 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
702 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
703 	em_dev_clear_queues(dev);
704 	return -EINVAL;
705 }
706 
707 /*********************************************************************
708  *
709  *  This routine disables all traffic on the adapter by issuing a
710  *  global reset on the MAC.
711  *
712  **********************************************************************/
713 static int
714 eth_em_stop(struct rte_eth_dev *dev)
715 {
716 	struct rte_eth_link link;
717 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
718 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
719 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
720 
721 	dev->data->dev_started = 0;
722 
723 	eth_em_rxtx_control(dev, false);
724 	em_rxq_intr_disable(hw);
725 	em_lsc_intr_disable(hw);
726 
727 	e1000_reset_hw(hw);
728 
729 	/* Flush desc rings for i219 */
730 	if (hw->mac.type == e1000_pch_spt || hw->mac.type == e1000_pch_cnp)
731 		em_flush_desc_rings(dev);
732 
733 	if (hw->mac.type >= e1000_82544)
734 		E1000_WRITE_REG(hw, E1000_WUC, 0);
735 
736 	/* Power down the phy. Needed to make the link go down */
737 	e1000_power_down_phy(hw);
738 
739 	em_dev_clear_queues(dev);
740 
741 	/* clear the recorded link status */
742 	memset(&link, 0, sizeof(link));
743 	rte_eth_linkstatus_set(dev, &link);
744 
745 	if (!rte_intr_allow_others(intr_handle))
746 		/* resume to the default handler */
747 		rte_intr_callback_register(intr_handle,
748 					   eth_em_interrupt_handler,
749 					   (void *)dev);
750 
751 	/* Clean datapath event and queue/vec mapping */
752 	rte_intr_efd_disable(intr_handle);
753 	if (intr_handle->intr_vec != NULL) {
754 		rte_free(intr_handle->intr_vec);
755 		intr_handle->intr_vec = NULL;
756 	}
757 
758 	return 0;
759 }
760 
761 static int
762 eth_em_close(struct rte_eth_dev *dev)
763 {
764 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
765 	struct e1000_adapter *adapter =
766 		E1000_DEV_PRIVATE(dev->data->dev_private);
767 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
768 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
769 	int ret;
770 
771 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
772 		return 0;
773 
774 	ret = eth_em_stop(dev);
775 	adapter->stopped = 1;
776 	em_dev_free_queues(dev);
777 	e1000_phy_hw_reset(hw);
778 	em_release_manageability(hw);
779 	em_hw_control_release(hw);
780 
781 	/* disable uio intr before callback unregister */
782 	rte_intr_disable(intr_handle);
783 	rte_intr_callback_unregister(intr_handle,
784 				     eth_em_interrupt_handler, dev);
785 
786 	return ret;
787 }
788 
789 static int
790 em_get_rx_buffer_size(struct e1000_hw *hw)
791 {
792 	uint32_t rx_buf_size;
793 
794 	rx_buf_size = ((E1000_READ_REG(hw, E1000_PBA) & UINT16_MAX) << 10);
795 	return rx_buf_size;
796 }
797 
798 /*********************************************************************
799  *
800  *  Initialize the hardware
801  *
802  **********************************************************************/
803 static int
804 em_hardware_init(struct e1000_hw *hw)
805 {
806 	uint32_t rx_buf_size;
807 	int diag;
808 
809 	/* Issue a global reset */
810 	e1000_reset_hw(hw);
811 
812 	/* Let the firmware know the OS is in control */
813 	em_hw_control_acquire(hw);
814 
815 	/*
816 	 * These parameters control the automatic generation (Tx) and
817 	 * response (Rx) to Ethernet PAUSE frames.
818 	 * - High water mark should allow for at least two standard size (1518)
819 	 *   frames to be received after sending an XOFF.
820 	 * - Low water mark works best when it is very near the high water mark.
821 	 *   This allows the receiver to restart by sending XON when it has
822 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
823 	 *   restart after one full frame is pulled from the buffer. There
824 	 *   could be several smaller frames in the buffer and if so they will
825 	 *   not trigger the XON until their total number reduces the buffer
826 	 *   by 1500.
827 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
828 	 */
829 	rx_buf_size = em_get_rx_buffer_size(hw);
830 
831 	hw->fc.high_water = rx_buf_size -
832 		PMD_ROUNDUP(RTE_ETHER_MAX_LEN * 2, 1024);
833 	hw->fc.low_water = hw->fc.high_water - 1500;
834 
835 	if (hw->mac.type == e1000_80003es2lan)
836 		hw->fc.pause_time = UINT16_MAX;
837 	else
838 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
839 
840 	hw->fc.send_xon = 1;
841 
842 	/* Set Flow control, use the tunable location if sane */
843 	if (em_fc_setting <= e1000_fc_full)
844 		hw->fc.requested_mode = em_fc_setting;
845 	else
846 		hw->fc.requested_mode = e1000_fc_none;
847 
848 	/* Workaround: no TX flow ctrl for PCH */
849 	if (hw->mac.type == e1000_pchlan)
850 		hw->fc.requested_mode = e1000_fc_rx_pause;
851 
852 	/* Override - settings for PCH2LAN, ya its magic :) */
853 	if (hw->mac.type == e1000_pch2lan) {
854 		hw->fc.high_water = 0x5C20;
855 		hw->fc.low_water = 0x5048;
856 		hw->fc.pause_time = 0x0650;
857 		hw->fc.refresh_time = 0x0400;
858 	} else if (hw->mac.type == e1000_pch_lpt ||
859 		   hw->mac.type == e1000_pch_spt ||
860 		   hw->mac.type == e1000_pch_cnp) {
861 		hw->fc.requested_mode = e1000_fc_full;
862 	}
863 
864 	diag = e1000_init_hw(hw);
865 	if (diag < 0)
866 		return diag;
867 	e1000_check_for_link(hw);
868 	return 0;
869 }
870 
871 /* This function is based on em_update_stats_counters() in e1000/if_em.c */
872 static int
873 eth_em_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
874 {
875 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
876 	struct e1000_hw_stats *stats =
877 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
878 	int pause_frames;
879 
880 	if(hw->phy.media_type == e1000_media_type_copper ||
881 			(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
882 		stats->symerrs += E1000_READ_REG(hw,E1000_SYMERRS);
883 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
884 	}
885 
886 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
887 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
888 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
889 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
890 
891 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
892 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
893 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
894 	stats->dc += E1000_READ_REG(hw, E1000_DC);
895 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
896 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
897 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
898 
899 	/*
900 	 * For watchdog management we need to know if we have been
901 	 * paused during the last interval, so capture that here.
902 	 */
903 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
904 	stats->xoffrxc += pause_frames;
905 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
906 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
907 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
908 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
909 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
910 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
911 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
912 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
913 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
914 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
915 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
916 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
917 
918 	/*
919 	 * For the 64-bit byte counters the low dword must be read first.
920 	 * Both registers clear on the read of the high dword.
921 	 */
922 
923 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
924 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
925 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
926 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
927 
928 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
929 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
930 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
931 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
932 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
933 
934 	stats->tor += E1000_READ_REG(hw, E1000_TORH);
935 	stats->tot += E1000_READ_REG(hw, E1000_TOTH);
936 
937 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
938 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
939 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
940 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
941 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
942 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
943 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
944 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
945 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
946 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
947 
948 	/* Interrupt Counts */
949 
950 	if (hw->mac.type >= e1000_82571) {
951 		stats->iac += E1000_READ_REG(hw, E1000_IAC);
952 		stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
953 		stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
954 		stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
955 		stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
956 		stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
957 		stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
958 		stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
959 		stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
960 	}
961 
962 	if (hw->mac.type >= e1000_82543) {
963 		stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
964 		stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
965 		stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
966 		stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
967 		stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
968 		stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
969 	}
970 
971 	if (rte_stats == NULL)
972 		return -EINVAL;
973 
974 	/* Rx Errors */
975 	rte_stats->imissed = stats->mpc;
976 	rte_stats->ierrors = stats->crcerrs + stats->rlec +
977 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
978 
979 	/* Tx Errors */
980 	rte_stats->oerrors = stats->ecol + stats->latecol;
981 
982 	rte_stats->ipackets = stats->gprc;
983 	rte_stats->opackets = stats->gptc;
984 	rte_stats->ibytes   = stats->gorc;
985 	rte_stats->obytes   = stats->gotc;
986 	return 0;
987 }
988 
989 static int
990 eth_em_stats_reset(struct rte_eth_dev *dev)
991 {
992 	struct e1000_hw_stats *hw_stats =
993 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
994 
995 	/* HW registers are cleared on read */
996 	eth_em_stats_get(dev, NULL);
997 
998 	/* Reset software totals */
999 	memset(hw_stats, 0, sizeof(*hw_stats));
1000 
1001 	return 0;
1002 }
1003 
1004 static int
1005 eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1006 {
1007 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1008 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1009 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1010 
1011 	em_rxq_intr_enable(hw);
1012 	rte_intr_ack(intr_handle);
1013 
1014 	return 0;
1015 }
1016 
1017 static int
1018 eth_em_rx_queue_intr_disable(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 
1022 	em_rxq_intr_disable(hw);
1023 
1024 	return 0;
1025 }
1026 
1027 uint32_t
1028 em_get_max_pktlen(struct rte_eth_dev *dev)
1029 {
1030 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1031 
1032 	switch (hw->mac.type) {
1033 	case e1000_82571:
1034 	case e1000_82572:
1035 	case e1000_ich9lan:
1036 	case e1000_ich10lan:
1037 	case e1000_pch2lan:
1038 	case e1000_pch_lpt:
1039 	case e1000_pch_spt:
1040 	case e1000_pch_cnp:
1041 	case e1000_82574:
1042 	case e1000_80003es2lan: /* 9K Jumbo Frame size */
1043 	case e1000_82583:
1044 		return 0x2412;
1045 	case e1000_pchlan:
1046 		return 0x1000;
1047 	/* Adapters that do not support jumbo frames */
1048 	case e1000_ich8lan:
1049 		return RTE_ETHER_MAX_LEN;
1050 	default:
1051 		return MAX_JUMBO_FRAME_SIZE;
1052 	}
1053 }
1054 
1055 static int
1056 eth_em_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1057 {
1058 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1059 
1060 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1061 	dev_info->max_rx_pktlen = em_get_max_pktlen(dev);
1062 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1063 
1064 	/*
1065 	 * Starting with 631xESB hw supports 2 TX/RX queues per port.
1066 	 * Unfortunatelly, all these nics have just one TX context.
1067 	 * So we have few choises for TX:
1068 	 * - Use just one TX queue.
1069 	 * - Allow cksum offload only for one TX queue.
1070 	 * - Don't allow TX cksum offload at all.
1071 	 * For now, option #1 was chosen.
1072 	 * To use second RX queue we have to use extended RX descriptor
1073 	 * (Multiple Receive Queues are mutually exclusive with UDP
1074 	 * fragmentation and are not supported when a legacy receive
1075 	 * descriptor format is used).
1076 	 * Which means separate RX routinies - as legacy nics (82540, 82545)
1077 	 * don't support extended RXD.
1078 	 * To avoid it we support just one RX queue for now (no RSS).
1079 	 */
1080 
1081 	dev_info->max_rx_queues = 1;
1082 	dev_info->max_tx_queues = 1;
1083 
1084 	dev_info->rx_queue_offload_capa = em_get_rx_queue_offloads_capa();
1085 	dev_info->rx_offload_capa = em_get_rx_port_offloads_capa() |
1086 				    dev_info->rx_queue_offload_capa;
1087 	dev_info->tx_queue_offload_capa = em_get_tx_queue_offloads_capa(dev);
1088 	dev_info->tx_offload_capa = em_get_tx_port_offloads_capa(dev) |
1089 				    dev_info->tx_queue_offload_capa;
1090 
1091 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1092 		.nb_max = E1000_MAX_RING_DESC,
1093 		.nb_min = E1000_MIN_RING_DESC,
1094 		.nb_align = EM_RXD_ALIGN,
1095 	};
1096 
1097 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1098 		.nb_max = E1000_MAX_RING_DESC,
1099 		.nb_min = E1000_MIN_RING_DESC,
1100 		.nb_align = EM_TXD_ALIGN,
1101 		.nb_seg_max = EM_TX_MAX_SEG,
1102 		.nb_mtu_seg_max = EM_TX_MAX_MTU_SEG,
1103 	};
1104 
1105 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1106 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1107 			ETH_LINK_SPEED_1G;
1108 
1109 	/* Preferred queue parameters */
1110 	dev_info->default_rxportconf.nb_queues = 1;
1111 	dev_info->default_txportconf.nb_queues = 1;
1112 	dev_info->default_txportconf.ring_size = 256;
1113 	dev_info->default_rxportconf.ring_size = 256;
1114 
1115 	return 0;
1116 }
1117 
1118 /* return 0 means link status changed, -1 means not changed */
1119 static int
1120 eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1121 {
1122 	struct e1000_hw *hw =
1123 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1124 	struct rte_eth_link link;
1125 	int link_up, count;
1126 
1127 	link_up = 0;
1128 	hw->mac.get_link_status = 1;
1129 
1130 	/* possible wait-to-complete in up to 9 seconds */
1131 	for (count = 0; count < EM_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
1132 		/* Read the real link status */
1133 		switch (hw->phy.media_type) {
1134 		case e1000_media_type_copper:
1135 			/* Do the work to read phy */
1136 			e1000_check_for_link(hw);
1137 			link_up = !hw->mac.get_link_status;
1138 			break;
1139 
1140 		case e1000_media_type_fiber:
1141 			e1000_check_for_link(hw);
1142 			link_up = (E1000_READ_REG(hw, E1000_STATUS) &
1143 					E1000_STATUS_LU);
1144 			break;
1145 
1146 		case e1000_media_type_internal_serdes:
1147 			e1000_check_for_link(hw);
1148 			link_up = hw->mac.serdes_has_link;
1149 			break;
1150 
1151 		default:
1152 			break;
1153 		}
1154 		if (link_up || wait_to_complete == 0)
1155 			break;
1156 		rte_delay_ms(EM_LINK_UPDATE_CHECK_INTERVAL);
1157 	}
1158 	memset(&link, 0, sizeof(link));
1159 
1160 	/* Now we check if a transition has happened */
1161 	if (link_up) {
1162 		uint16_t duplex, speed;
1163 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
1164 		link.link_duplex = (duplex == FULL_DUPLEX) ?
1165 				ETH_LINK_FULL_DUPLEX :
1166 				ETH_LINK_HALF_DUPLEX;
1167 		link.link_speed = speed;
1168 		link.link_status = ETH_LINK_UP;
1169 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1170 				ETH_LINK_SPEED_FIXED);
1171 	} else {
1172 		link.link_speed = ETH_SPEED_NUM_NONE;
1173 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
1174 		link.link_status = ETH_LINK_DOWN;
1175 		link.link_autoneg = ETH_LINK_FIXED;
1176 	}
1177 
1178 	return rte_eth_linkstatus_set(dev, &link);
1179 }
1180 
1181 /*
1182  * em_hw_control_acquire sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
1183  * For ASF and Pass Through versions of f/w this means
1184  * that the driver is loaded. For AMT version type f/w
1185  * this means that the network i/f is open.
1186  */
1187 static void
1188 em_hw_control_acquire(struct e1000_hw *hw)
1189 {
1190 	uint32_t ctrl_ext, swsm;
1191 
1192 	/* Let firmware know the driver has taken over */
1193 	if (hw->mac.type == e1000_82573) {
1194 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1195 		E1000_WRITE_REG(hw, E1000_SWSM, swsm | E1000_SWSM_DRV_LOAD);
1196 
1197 	} else {
1198 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1199 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1200 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1201 	}
1202 }
1203 
1204 /*
1205  * em_hw_control_release resets {CTRL_EXTT|FWSM}:DRV_LOAD bit.
1206  * For ASF and Pass Through versions of f/w this means that the
1207  * driver is no longer loaded. For AMT versions of the
1208  * f/w this means that the network i/f is closed.
1209  */
1210 static void
1211 em_hw_control_release(struct e1000_hw *hw)
1212 {
1213 	uint32_t ctrl_ext, swsm;
1214 
1215 	/* Let firmware taken over control of h/w */
1216 	if (hw->mac.type == e1000_82573) {
1217 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1218 		E1000_WRITE_REG(hw, E1000_SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
1219 	} else {
1220 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1221 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1222 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1223 	}
1224 }
1225 
1226 /*
1227  * Bit of a misnomer, what this really means is
1228  * to enable OS management of the system... aka
1229  * to disable special hardware management features.
1230  */
1231 static void
1232 em_init_manageability(struct e1000_hw *hw)
1233 {
1234 	if (e1000_enable_mng_pass_thru(hw)) {
1235 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
1236 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
1237 
1238 		/* disable hardware interception of ARP */
1239 		manc &= ~(E1000_MANC_ARP_EN);
1240 
1241 		/* enable receiving management packets to the host */
1242 		manc |= E1000_MANC_EN_MNG2HOST;
1243 		manc2h |= 1 << 5;  /* Mng Port 623 */
1244 		manc2h |= 1 << 6;  /* Mng Port 664 */
1245 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
1246 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1247 	}
1248 }
1249 
1250 /*
1251  * Give control back to hardware management
1252  * controller if there is one.
1253  */
1254 static void
1255 em_release_manageability(struct e1000_hw *hw)
1256 {
1257 	uint32_t manc;
1258 
1259 	if (e1000_enable_mng_pass_thru(hw)) {
1260 		manc = E1000_READ_REG(hw, E1000_MANC);
1261 
1262 		/* re-enable hardware interception of ARP */
1263 		manc |= E1000_MANC_ARP_EN;
1264 		manc &= ~E1000_MANC_EN_MNG2HOST;
1265 
1266 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1267 	}
1268 }
1269 
1270 static int
1271 eth_em_promiscuous_enable(struct rte_eth_dev *dev)
1272 {
1273 	struct e1000_hw *hw =
1274 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1275 	uint32_t rctl;
1276 
1277 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1278 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1279 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1280 
1281 	return 0;
1282 }
1283 
1284 static int
1285 eth_em_promiscuous_disable(struct rte_eth_dev *dev)
1286 {
1287 	struct e1000_hw *hw =
1288 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1289 	uint32_t rctl;
1290 
1291 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1292 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_SBP);
1293 	if (dev->data->all_multicast == 1)
1294 		rctl |= E1000_RCTL_MPE;
1295 	else
1296 		rctl &= (~E1000_RCTL_MPE);
1297 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1298 
1299 	return 0;
1300 }
1301 
1302 static int
1303 eth_em_allmulticast_enable(struct rte_eth_dev *dev)
1304 {
1305 	struct e1000_hw *hw =
1306 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1307 	uint32_t rctl;
1308 
1309 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1310 	rctl |= E1000_RCTL_MPE;
1311 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1312 
1313 	return 0;
1314 }
1315 
1316 static int
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 0; /* 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 	return 0;
1330 }
1331 
1332 static int
1333 eth_em_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1334 {
1335 	struct e1000_hw *hw =
1336 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1337 	struct e1000_vfta * shadow_vfta =
1338 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1339 	uint32_t vfta;
1340 	uint32_t vid_idx;
1341 	uint32_t vid_bit;
1342 
1343 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
1344 			      E1000_VFTA_ENTRY_MASK);
1345 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
1346 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
1347 	if (on)
1348 		vfta |= vid_bit;
1349 	else
1350 		vfta &= ~vid_bit;
1351 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
1352 
1353 	/* update local VFTA copy */
1354 	shadow_vfta->vfta[vid_idx] = vfta;
1355 
1356 	return 0;
1357 }
1358 
1359 static void
1360 em_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1361 {
1362 	struct e1000_hw *hw =
1363 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1364 	uint32_t reg;
1365 
1366 	/* Filter Table Disable */
1367 	reg = E1000_READ_REG(hw, E1000_RCTL);
1368 	reg &= ~E1000_RCTL_CFIEN;
1369 	reg &= ~E1000_RCTL_VFE;
1370 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1371 }
1372 
1373 static void
1374 em_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1375 {
1376 	struct e1000_hw *hw =
1377 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1378 	struct e1000_vfta * shadow_vfta =
1379 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1380 	uint32_t reg;
1381 	int i;
1382 
1383 	/* Filter Table Enable, CFI not used for packet acceptance */
1384 	reg = E1000_READ_REG(hw, E1000_RCTL);
1385 	reg &= ~E1000_RCTL_CFIEN;
1386 	reg |= E1000_RCTL_VFE;
1387 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1388 
1389 	/* restore vfta from local copy */
1390 	for (i = 0; i < IGB_VFTA_SIZE; i++)
1391 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
1392 }
1393 
1394 static void
1395 em_vlan_hw_strip_disable(struct rte_eth_dev *dev)
1396 {
1397 	struct e1000_hw *hw =
1398 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1399 	uint32_t reg;
1400 
1401 	/* VLAN Mode Disable */
1402 	reg = E1000_READ_REG(hw, E1000_CTRL);
1403 	reg &= ~E1000_CTRL_VME;
1404 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1405 
1406 }
1407 
1408 static void
1409 em_vlan_hw_strip_enable(struct rte_eth_dev *dev)
1410 {
1411 	struct e1000_hw *hw =
1412 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1413 	uint32_t reg;
1414 
1415 	/* VLAN Mode Enable */
1416 	reg = E1000_READ_REG(hw, E1000_CTRL);
1417 	reg |= E1000_CTRL_VME;
1418 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1419 }
1420 
1421 static int
1422 eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1423 {
1424 	struct rte_eth_rxmode *rxmode;
1425 
1426 	rxmode = &dev->data->dev_conf.rxmode;
1427 	if(mask & ETH_VLAN_STRIP_MASK){
1428 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1429 			em_vlan_hw_strip_enable(dev);
1430 		else
1431 			em_vlan_hw_strip_disable(dev);
1432 	}
1433 
1434 	if(mask & ETH_VLAN_FILTER_MASK){
1435 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1436 			em_vlan_hw_filter_enable(dev);
1437 		else
1438 			em_vlan_hw_filter_disable(dev);
1439 	}
1440 
1441 	return 0;
1442 }
1443 
1444 /*
1445  * It enables the interrupt mask and then enable the interrupt.
1446  *
1447  * @param dev
1448  *  Pointer to struct rte_eth_dev.
1449  *
1450  * @return
1451  *  - On success, zero.
1452  *  - On failure, a negative value.
1453  */
1454 static int
1455 eth_em_interrupt_setup(struct rte_eth_dev *dev)
1456 {
1457 	uint32_t regval;
1458 	struct e1000_hw *hw =
1459 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1460 
1461 	/* clear interrupt */
1462 	E1000_READ_REG(hw, E1000_ICR);
1463 	regval = E1000_READ_REG(hw, E1000_IMS);
1464 	E1000_WRITE_REG(hw, E1000_IMS,
1465 			regval | E1000_ICR_LSC | E1000_ICR_OTHER);
1466 	return 0;
1467 }
1468 
1469 /*
1470  * It clears the interrupt causes and enables the interrupt.
1471  * It will be called once only during nic initialized.
1472  *
1473  * @param dev
1474  *  Pointer to struct rte_eth_dev.
1475  *
1476  * @return
1477  *  - On success, zero.
1478  *  - On failure, a negative value.
1479  */
1480 static int
1481 eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev)
1482 {
1483 	struct e1000_hw *hw =
1484 	E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1485 
1486 	E1000_READ_REG(hw, E1000_ICR);
1487 	em_rxq_intr_enable(hw);
1488 	return 0;
1489 }
1490 
1491 /*
1492  * It enable receive packet interrupt.
1493  * @param hw
1494  * Pointer to struct e1000_hw
1495  *
1496  * @return
1497  */
1498 static void
1499 em_rxq_intr_enable(struct e1000_hw *hw)
1500 {
1501 	E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_RXT0);
1502 	E1000_WRITE_FLUSH(hw);
1503 }
1504 
1505 /*
1506  * It disabled lsc interrupt.
1507  * @param hw
1508  * Pointer to struct e1000_hw
1509  *
1510  * @return
1511  */
1512 static void
1513 em_lsc_intr_disable(struct e1000_hw *hw)
1514 {
1515 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC | E1000_IMS_OTHER);
1516 	E1000_WRITE_FLUSH(hw);
1517 }
1518 
1519 /*
1520  * It disabled receive packet interrupt.
1521  * @param hw
1522  * Pointer to struct e1000_hw
1523  *
1524  * @return
1525  */
1526 static void
1527 em_rxq_intr_disable(struct e1000_hw *hw)
1528 {
1529 	E1000_READ_REG(hw, E1000_ICR);
1530 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_RXT0);
1531 	E1000_WRITE_FLUSH(hw);
1532 }
1533 
1534 /*
1535  * It reads ICR and gets interrupt causes, check it and set a bit flag
1536  * to update link status.
1537  *
1538  * @param dev
1539  *  Pointer to struct rte_eth_dev.
1540  *
1541  * @return
1542  *  - On success, zero.
1543  *  - On failure, a negative value.
1544  */
1545 static int
1546 eth_em_interrupt_get_status(struct rte_eth_dev *dev)
1547 {
1548 	uint32_t icr;
1549 	struct e1000_hw *hw =
1550 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1551 	struct e1000_interrupt *intr =
1552 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1553 
1554 	/* read-on-clear nic registers here */
1555 	icr = E1000_READ_REG(hw, E1000_ICR);
1556 	if (icr & E1000_ICR_LSC) {
1557 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1558 	}
1559 
1560 	return 0;
1561 }
1562 
1563 /*
1564  * It executes link_update after knowing an interrupt is prsent.
1565  *
1566  * @param dev
1567  *  Pointer to struct rte_eth_dev.
1568  *
1569  * @return
1570  *  - On success, zero.
1571  *  - On failure, a negative value.
1572  */
1573 static int
1574 eth_em_interrupt_action(struct rte_eth_dev *dev,
1575 			struct rte_intr_handle *intr_handle)
1576 {
1577 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1578 	struct e1000_hw *hw =
1579 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1580 	struct e1000_interrupt *intr =
1581 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1582 	struct rte_eth_link link;
1583 	int ret;
1584 
1585 	if (!(intr->flags & E1000_FLAG_NEED_LINK_UPDATE))
1586 		return -1;
1587 
1588 	intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
1589 	rte_intr_ack(intr_handle);
1590 
1591 	/* set get_link_status to check register later */
1592 	hw->mac.get_link_status = 1;
1593 	ret = eth_em_link_update(dev, 0);
1594 
1595 	/* check if link has changed */
1596 	if (ret < 0)
1597 		return 0;
1598 
1599 	rte_eth_linkstatus_get(dev, &link);
1600 
1601 	if (link.link_status) {
1602 		PMD_INIT_LOG(INFO, " Port %d: Link Up - speed %u Mbps - %s",
1603 			     dev->data->port_id, link.link_speed,
1604 			     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
1605 			     "full-duplex" : "half-duplex");
1606 	} else {
1607 		PMD_INIT_LOG(INFO, " Port %d: Link Down", dev->data->port_id);
1608 	}
1609 	PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
1610 		     pci_dev->addr.domain, pci_dev->addr.bus,
1611 		     pci_dev->addr.devid, pci_dev->addr.function);
1612 
1613 	return 0;
1614 }
1615 
1616 /**
1617  * Interrupt handler which shall be registered at first.
1618  *
1619  * @param handle
1620  *  Pointer to interrupt handle.
1621  * @param param
1622  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1623  *
1624  * @return
1625  *  void
1626  */
1627 static void
1628 eth_em_interrupt_handler(void *param)
1629 {
1630 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1631 
1632 	eth_em_interrupt_get_status(dev);
1633 	eth_em_interrupt_action(dev, dev->intr_handle);
1634 	rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
1635 }
1636 
1637 static int
1638 eth_em_led_on(struct rte_eth_dev *dev)
1639 {
1640 	struct e1000_hw *hw;
1641 
1642 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1643 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1644 }
1645 
1646 static int
1647 eth_em_led_off(struct rte_eth_dev *dev)
1648 {
1649 	struct e1000_hw *hw;
1650 
1651 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1652 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1653 }
1654 
1655 static int
1656 eth_em_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1657 {
1658 	struct e1000_hw *hw;
1659 	uint32_t ctrl;
1660 	int tx_pause;
1661 	int rx_pause;
1662 
1663 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1664 	fc_conf->pause_time = hw->fc.pause_time;
1665 	fc_conf->high_water = hw->fc.high_water;
1666 	fc_conf->low_water = hw->fc.low_water;
1667 	fc_conf->send_xon = hw->fc.send_xon;
1668 	fc_conf->autoneg = hw->mac.autoneg;
1669 
1670 	/*
1671 	 * Return rx_pause and tx_pause status according to actual setting of
1672 	 * the TFCE and RFCE bits in the CTRL register.
1673 	 */
1674 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
1675 	if (ctrl & E1000_CTRL_TFCE)
1676 		tx_pause = 1;
1677 	else
1678 		tx_pause = 0;
1679 
1680 	if (ctrl & E1000_CTRL_RFCE)
1681 		rx_pause = 1;
1682 	else
1683 		rx_pause = 0;
1684 
1685 	if (rx_pause && tx_pause)
1686 		fc_conf->mode = RTE_FC_FULL;
1687 	else if (rx_pause)
1688 		fc_conf->mode = RTE_FC_RX_PAUSE;
1689 	else if (tx_pause)
1690 		fc_conf->mode = RTE_FC_TX_PAUSE;
1691 	else
1692 		fc_conf->mode = RTE_FC_NONE;
1693 
1694 	return 0;
1695 }
1696 
1697 static int
1698 eth_em_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1699 {
1700 	struct e1000_hw *hw;
1701 	int err;
1702 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
1703 		e1000_fc_none,
1704 		e1000_fc_rx_pause,
1705 		e1000_fc_tx_pause,
1706 		e1000_fc_full
1707 	};
1708 	uint32_t rx_buf_size;
1709 	uint32_t max_high_water;
1710 	uint32_t rctl;
1711 
1712 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1713 	if (fc_conf->autoneg != hw->mac.autoneg)
1714 		return -ENOTSUP;
1715 	rx_buf_size = em_get_rx_buffer_size(hw);
1716 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
1717 
1718 	/* At least reserve one Ethernet frame for watermark */
1719 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
1720 	if ((fc_conf->high_water > max_high_water) ||
1721 	    (fc_conf->high_water < fc_conf->low_water)) {
1722 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
1723 		PMD_INIT_LOG(ERR, "high water must <= 0x%x", max_high_water);
1724 		return -EINVAL;
1725 	}
1726 
1727 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
1728 	hw->fc.pause_time     = fc_conf->pause_time;
1729 	hw->fc.high_water     = fc_conf->high_water;
1730 	hw->fc.low_water      = fc_conf->low_water;
1731 	hw->fc.send_xon	      = fc_conf->send_xon;
1732 
1733 	err = e1000_setup_link_generic(hw);
1734 	if (err == E1000_SUCCESS) {
1735 
1736 		/* check if we want to forward MAC frames - driver doesn't have native
1737 		 * capability to do that, so we'll write the registers ourselves */
1738 
1739 		rctl = E1000_READ_REG(hw, E1000_RCTL);
1740 
1741 		/* set or clear MFLCN.PMCF bit depending on configuration */
1742 		if (fc_conf->mac_ctrl_frame_fwd != 0)
1743 			rctl |= E1000_RCTL_PMCF;
1744 		else
1745 			rctl &= ~E1000_RCTL_PMCF;
1746 
1747 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1748 		E1000_WRITE_FLUSH(hw);
1749 
1750 		return 0;
1751 	}
1752 
1753 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
1754 	return -EIO;
1755 }
1756 
1757 static int
1758 eth_em_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
1759 		uint32_t index, __rte_unused uint32_t pool)
1760 {
1761 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1762 
1763 	return e1000_rar_set(hw, mac_addr->addr_bytes, index);
1764 }
1765 
1766 static void
1767 eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1768 {
1769 	uint8_t addr[RTE_ETHER_ADDR_LEN];
1770 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1771 
1772 	memset(addr, 0, sizeof(addr));
1773 
1774 	e1000_rar_set(hw, addr, index);
1775 }
1776 
1777 static int
1778 eth_em_default_mac_addr_set(struct rte_eth_dev *dev,
1779 			    struct rte_ether_addr *addr)
1780 {
1781 	eth_em_rar_clear(dev, 0);
1782 
1783 	return eth_em_rar_set(dev, (void *)addr, 0, 0);
1784 }
1785 
1786 static int
1787 eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1788 {
1789 	struct e1000_hw *hw;
1790 	uint32_t frame_size;
1791 	uint32_t rctl;
1792 
1793 	frame_size = mtu + E1000_ETH_OVERHEAD;
1794 
1795 	/*
1796 	 * If device is started, refuse mtu that requires the support of
1797 	 * scattered packets when this feature has not been enabled before.
1798 	 */
1799 	if (dev->data->dev_started && !dev->data->scattered_rx &&
1800 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
1801 		PMD_INIT_LOG(ERR, "Stop port first.");
1802 		return -EINVAL;
1803 	}
1804 
1805 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1806 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1807 
1808 	/* switch to jumbo mode if needed */
1809 	if (mtu > RTE_ETHER_MTU)
1810 		rctl |= E1000_RCTL_LPE;
1811 	else
1812 		rctl &= ~E1000_RCTL_LPE;
1813 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1814 
1815 	return 0;
1816 }
1817 
1818 static int
1819 eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
1820 			struct rte_ether_addr *mc_addr_set,
1821 			uint32_t nb_mc_addr)
1822 {
1823 	struct e1000_hw *hw;
1824 
1825 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1826 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1827 	return 0;
1828 }
1829 
1830 RTE_PMD_REGISTER_PCI(net_e1000_em, rte_em_pmd);
1831 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_em, pci_id_em_map);
1832 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_em, "* igb_uio | uio_pci_generic | vfio-pci");
1833