xref: /f-stack/dpdk/drivers/net/e1000/em_ethdev.c (revision 2bfe3f2e)
1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
5  *   All rights reserved.
6  *
7  *   Redistribution and use in source and binary forms, with or without
8  *   modification, are permitted provided that the following conditions
9  *   are met:
10  *
11  *     * Redistributions of source code must retain the above copyright
12  *       notice, this list of conditions and the following disclaimer.
13  *     * Redistributions in binary form must reproduce the above copyright
14  *       notice, this list of conditions and the following disclaimer in
15  *       the documentation and/or other materials provided with the
16  *       distribution.
17  *     * Neither the name of Intel Corporation nor the names of its
18  *       contributors may be used to endorse or promote products derived
19  *       from this software without specific prior written permission.
20  *
21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32  */
33 
34 #include <sys/queue.h>
35 #include <stdio.h>
36 #include <errno.h>
37 #include <stdint.h>
38 #include <stdarg.h>
39 
40 #include <rte_common.h>
41 #include <rte_interrupts.h>
42 #include <rte_byteorder.h>
43 #include <rte_log.h>
44 #include <rte_debug.h>
45 #include <rte_pci.h>
46 #include <rte_bus_pci.h>
47 #include <rte_ether.h>
48 #include <rte_ethdev.h>
49 #include <rte_ethdev_pci.h>
50 #include <rte_memory.h>
51 #include <rte_eal.h>
52 #include <rte_atomic.h>
53 #include <rte_malloc.h>
54 #include <rte_dev.h>
55 
56 #include "e1000_logs.h"
57 #include "base/e1000_api.h"
58 #include "e1000_ethdev.h"
59 
60 #define EM_EIAC			0x000DC
61 
62 #define PMD_ROUNDUP(x,y)	(((x) + (y) - 1)/(y) * (y))
63 
64 
65 static int eth_em_configure(struct rte_eth_dev *dev);
66 static int eth_em_start(struct rte_eth_dev *dev);
67 static void eth_em_stop(struct rte_eth_dev *dev);
68 static void eth_em_close(struct rte_eth_dev *dev);
69 static void eth_em_promiscuous_enable(struct rte_eth_dev *dev);
70 static void eth_em_promiscuous_disable(struct rte_eth_dev *dev);
71 static void eth_em_allmulticast_enable(struct rte_eth_dev *dev);
72 static void eth_em_allmulticast_disable(struct rte_eth_dev *dev);
73 static int eth_em_link_update(struct rte_eth_dev *dev,
74 				int wait_to_complete);
75 static int eth_em_stats_get(struct rte_eth_dev *dev,
76 				struct rte_eth_stats *rte_stats);
77 static void eth_em_stats_reset(struct rte_eth_dev *dev);
78 static void eth_em_infos_get(struct rte_eth_dev *dev,
79 				struct rte_eth_dev_info *dev_info);
80 static int eth_em_flow_ctrl_get(struct rte_eth_dev *dev,
81 				struct rte_eth_fc_conf *fc_conf);
82 static int eth_em_flow_ctrl_set(struct rte_eth_dev *dev,
83 				struct rte_eth_fc_conf *fc_conf);
84 static int eth_em_interrupt_setup(struct rte_eth_dev *dev);
85 static int eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev);
86 static int eth_em_interrupt_get_status(struct rte_eth_dev *dev);
87 static int eth_em_interrupt_action(struct rte_eth_dev *dev,
88 				   struct rte_intr_handle *handle);
89 static void eth_em_interrupt_handler(void *param);
90 
91 static int em_hw_init(struct e1000_hw *hw);
92 static int em_hardware_init(struct e1000_hw *hw);
93 static void em_hw_control_acquire(struct e1000_hw *hw);
94 static void em_hw_control_release(struct e1000_hw *hw);
95 static void em_init_manageability(struct e1000_hw *hw);
96 static void em_release_manageability(struct e1000_hw *hw);
97 
98 static int eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
99 
100 static int eth_em_vlan_filter_set(struct rte_eth_dev *dev,
101 		uint16_t vlan_id, int on);
102 static int eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask);
103 static void em_vlan_hw_filter_enable(struct rte_eth_dev *dev);
104 static void em_vlan_hw_filter_disable(struct rte_eth_dev *dev);
105 static void em_vlan_hw_strip_enable(struct rte_eth_dev *dev);
106 static void em_vlan_hw_strip_disable(struct rte_eth_dev *dev);
107 
108 /*
109 static void eth_em_vlan_filter_set(struct rte_eth_dev *dev,
110 					uint16_t vlan_id, int on);
111 */
112 
113 static int eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
114 static int eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
115 static void em_lsc_intr_disable(struct e1000_hw *hw);
116 static void em_rxq_intr_enable(struct e1000_hw *hw);
117 static void em_rxq_intr_disable(struct e1000_hw *hw);
118 
119 static int eth_em_led_on(struct rte_eth_dev *dev);
120 static int eth_em_led_off(struct rte_eth_dev *dev);
121 
122 static int em_get_rx_buffer_size(struct e1000_hw *hw);
123 static int eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
124 			  uint32_t index, uint32_t pool);
125 static void eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index);
126 
127 static int eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
128 				   struct ether_addr *mc_addr_set,
129 				   uint32_t nb_mc_addr);
130 
131 #define EM_FC_PAUSE_TIME 0x0680
132 #define EM_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
133 #define EM_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
134 
135 static enum e1000_fc_mode em_fc_setting = e1000_fc_full;
136 
137 /*
138  * The set of PCI devices this driver supports
139  */
140 static const struct rte_pci_id pci_id_em_map[] = {
141 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82540EM) },
142 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_COPPER) },
143 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_FIBER) },
144 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_COPPER) },
145 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_FIBER) },
146 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_QUAD_COPPER) },
147 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_COPPER) },
148 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_FIBER) },
149 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES) },
150 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_DUAL) },
151 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_QUAD) },
152 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER) },
153 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571PT_QUAD_COPPER) },
154 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_FIBER) },
155 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER_LP) },
156 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_COPPER) },
157 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_FIBER) },
158 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_SERDES) },
159 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI) },
160 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82573L) },
161 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574L) },
162 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574LA) },
163 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82583V) },
164 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_LM) },
165 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_V) },
166 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_LM) },
167 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_V) },
168 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM2) },
169 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V2) },
170 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM3) },
171 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V3) },
172 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM) },
173 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V) },
174 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM2) },
175 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V2) },
176 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LBG_I219_LM3) },
177 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM4) },
178 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V4) },
179 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM5) },
180 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V5) },
181 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM6) },
182 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V6) },
183 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM7) },
184 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V7) },
185 	{ .vendor_id = 0, /* sentinel */ },
186 };
187 
188 static const struct eth_dev_ops eth_em_ops = {
189 	.dev_configure        = eth_em_configure,
190 	.dev_start            = eth_em_start,
191 	.dev_stop             = eth_em_stop,
192 	.dev_close            = eth_em_close,
193 	.promiscuous_enable   = eth_em_promiscuous_enable,
194 	.promiscuous_disable  = eth_em_promiscuous_disable,
195 	.allmulticast_enable  = eth_em_allmulticast_enable,
196 	.allmulticast_disable = eth_em_allmulticast_disable,
197 	.link_update          = eth_em_link_update,
198 	.stats_get            = eth_em_stats_get,
199 	.stats_reset          = eth_em_stats_reset,
200 	.dev_infos_get        = eth_em_infos_get,
201 	.mtu_set              = eth_em_mtu_set,
202 	.vlan_filter_set      = eth_em_vlan_filter_set,
203 	.vlan_offload_set     = eth_em_vlan_offload_set,
204 	.rx_queue_setup       = eth_em_rx_queue_setup,
205 	.rx_queue_release     = eth_em_rx_queue_release,
206 	.rx_queue_count       = eth_em_rx_queue_count,
207 	.rx_descriptor_done   = eth_em_rx_descriptor_done,
208 	.rx_descriptor_status = eth_em_rx_descriptor_status,
209 	.tx_descriptor_status = eth_em_tx_descriptor_status,
210 	.tx_queue_setup       = eth_em_tx_queue_setup,
211 	.tx_queue_release     = eth_em_tx_queue_release,
212 	.rx_queue_intr_enable = eth_em_rx_queue_intr_enable,
213 	.rx_queue_intr_disable = eth_em_rx_queue_intr_disable,
214 	.dev_led_on           = eth_em_led_on,
215 	.dev_led_off          = eth_em_led_off,
216 	.flow_ctrl_get        = eth_em_flow_ctrl_get,
217 	.flow_ctrl_set        = eth_em_flow_ctrl_set,
218 	.mac_addr_add         = eth_em_rar_set,
219 	.mac_addr_remove      = eth_em_rar_clear,
220 	.set_mc_addr_list     = eth_em_set_mc_addr_list,
221 	.rxq_info_get         = em_rxq_info_get,
222 	.txq_info_get         = em_txq_info_get,
223 };
224 
225 /**
226  * Atomically reads the link status information from global
227  * structure rte_eth_dev.
228  *
229  * @param dev
230  *   - Pointer to the structure rte_eth_dev to read from.
231  *   - Pointer to the buffer to be saved with the link status.
232  *
233  * @return
234  *   - On success, zero.
235  *   - On failure, negative value.
236  */
237 static inline int
238 rte_em_dev_atomic_read_link_status(struct rte_eth_dev *dev,
239 				struct rte_eth_link *link)
240 {
241 	struct rte_eth_link *dst = link;
242 	struct rte_eth_link *src = &(dev->data->dev_link);
243 
244 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
245 					*(uint64_t *)src) == 0)
246 		return -1;
247 
248 	return 0;
249 }
250 
251 /**
252  * Atomically writes the link status information into global
253  * structure rte_eth_dev.
254  *
255  * @param dev
256  *   - Pointer to the structure rte_eth_dev to read from.
257  *   - Pointer to the buffer to be saved with the link status.
258  *
259  * @return
260  *   - On success, zero.
261  *   - On failure, negative value.
262  */
263 static inline int
264 rte_em_dev_atomic_write_link_status(struct rte_eth_dev *dev,
265 				struct rte_eth_link *link)
266 {
267 	struct rte_eth_link *dst = &(dev->data->dev_link);
268 	struct rte_eth_link *src = link;
269 
270 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
271 					*(uint64_t *)src) == 0)
272 		return -1;
273 
274 	return 0;
275 }
276 
277 /**
278  *  eth_em_dev_is_ich8 - Check for ICH8 device
279  *  @hw: pointer to the HW structure
280  *
281  *  return TRUE for ICH8, otherwise FALSE
282  **/
283 static bool
284 eth_em_dev_is_ich8(struct e1000_hw *hw)
285 {
286 	DEBUGFUNC("eth_em_dev_is_ich8");
287 
288 	switch (hw->device_id) {
289 	case E1000_DEV_ID_PCH_LPT_I217_LM:
290 	case E1000_DEV_ID_PCH_LPT_I217_V:
291 	case E1000_DEV_ID_PCH_LPTLP_I218_LM:
292 	case E1000_DEV_ID_PCH_LPTLP_I218_V:
293 	case E1000_DEV_ID_PCH_I218_V2:
294 	case E1000_DEV_ID_PCH_I218_LM2:
295 	case E1000_DEV_ID_PCH_I218_V3:
296 	case E1000_DEV_ID_PCH_I218_LM3:
297 	case E1000_DEV_ID_PCH_SPT_I219_LM:
298 	case E1000_DEV_ID_PCH_SPT_I219_V:
299 	case E1000_DEV_ID_PCH_SPT_I219_LM2:
300 	case E1000_DEV_ID_PCH_SPT_I219_V2:
301 	case E1000_DEV_ID_PCH_LBG_I219_LM3:
302 	case E1000_DEV_ID_PCH_SPT_I219_LM4:
303 	case E1000_DEV_ID_PCH_SPT_I219_V4:
304 	case E1000_DEV_ID_PCH_SPT_I219_LM5:
305 	case E1000_DEV_ID_PCH_SPT_I219_V5:
306 	case E1000_DEV_ID_PCH_CNP_I219_LM6:
307 	case E1000_DEV_ID_PCH_CNP_I219_V6:
308 	case E1000_DEV_ID_PCH_CNP_I219_LM7:
309 	case E1000_DEV_ID_PCH_CNP_I219_V7:
310 		return 1;
311 	default:
312 		return 0;
313 	}
314 }
315 
316 static int
317 eth_em_dev_init(struct rte_eth_dev *eth_dev)
318 {
319 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
320 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
321 	struct e1000_adapter *adapter =
322 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
323 	struct e1000_hw *hw =
324 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
325 	struct e1000_vfta * shadow_vfta =
326 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
327 
328 	eth_dev->dev_ops = &eth_em_ops;
329 	eth_dev->rx_pkt_burst = (eth_rx_burst_t)&eth_em_recv_pkts;
330 	eth_dev->tx_pkt_burst = (eth_tx_burst_t)&eth_em_xmit_pkts;
331 	eth_dev->tx_pkt_prepare = (eth_tx_prep_t)&eth_em_prep_pkts;
332 
333 	/* for secondary processes, we don't initialise any further as primary
334 	 * has already done this work. Only check we don't need a different
335 	 * RX function */
336 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
337 		if (eth_dev->data->scattered_rx)
338 			eth_dev->rx_pkt_burst =
339 				(eth_rx_burst_t)&eth_em_recv_scattered_pkts;
340 		return 0;
341 	}
342 
343 	rte_eth_copy_pci_info(eth_dev, pci_dev);
344 
345 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
346 	hw->device_id = pci_dev->id.device_id;
347 	adapter->stopped = 0;
348 
349 	/* For ICH8 support we'll need to map the flash memory BAR */
350 	if (eth_em_dev_is_ich8(hw))
351 		hw->flash_address = (void *)pci_dev->mem_resource[1].addr;
352 
353 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS ||
354 			em_hw_init(hw) != 0) {
355 		PMD_INIT_LOG(ERR, "port_id %d vendorID=0x%x deviceID=0x%x: "
356 			"failed to init HW",
357 			eth_dev->data->port_id, pci_dev->id.vendor_id,
358 			pci_dev->id.device_id);
359 		return -ENODEV;
360 	}
361 
362 	/* Allocate memory for storing MAC addresses */
363 	eth_dev->data->mac_addrs = rte_zmalloc("e1000", ETHER_ADDR_LEN *
364 			hw->mac.rar_entry_count, 0);
365 	if (eth_dev->data->mac_addrs == NULL) {
366 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
367 			"store MAC addresses",
368 			ETHER_ADDR_LEN * hw->mac.rar_entry_count);
369 		return -ENOMEM;
370 	}
371 
372 	/* Copy the permanent MAC address */
373 	ether_addr_copy((struct ether_addr *) hw->mac.addr,
374 		eth_dev->data->mac_addrs);
375 
376 	/* initialize the vfta */
377 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
378 
379 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
380 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
381 		     pci_dev->id.device_id);
382 
383 	rte_intr_callback_register(intr_handle,
384 				   eth_em_interrupt_handler, eth_dev);
385 
386 	return 0;
387 }
388 
389 static int
390 eth_em_dev_uninit(struct rte_eth_dev *eth_dev)
391 {
392 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
393 	struct e1000_adapter *adapter =
394 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
395 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
396 
397 	PMD_INIT_FUNC_TRACE();
398 
399 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
400 		return -EPERM;
401 
402 	if (adapter->stopped == 0)
403 		eth_em_close(eth_dev);
404 
405 	eth_dev->dev_ops = NULL;
406 	eth_dev->rx_pkt_burst = NULL;
407 	eth_dev->tx_pkt_burst = NULL;
408 
409 	rte_free(eth_dev->data->mac_addrs);
410 	eth_dev->data->mac_addrs = NULL;
411 
412 	/* disable uio intr before callback unregister */
413 	rte_intr_disable(intr_handle);
414 	rte_intr_callback_unregister(intr_handle,
415 				     eth_em_interrupt_handler, eth_dev);
416 
417 	return 0;
418 }
419 
420 static int eth_em_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
421 	struct rte_pci_device *pci_dev)
422 {
423 	return rte_eth_dev_pci_generic_probe(pci_dev,
424 		sizeof(struct e1000_adapter), eth_em_dev_init);
425 }
426 
427 static int eth_em_pci_remove(struct rte_pci_device *pci_dev)
428 {
429 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_em_dev_uninit);
430 }
431 
432 static struct rte_pci_driver rte_em_pmd = {
433 	.id_table = pci_id_em_map,
434 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
435 		     RTE_PCI_DRV_IOVA_AS_VA,
436 	.probe = eth_em_pci_probe,
437 	.remove = eth_em_pci_remove,
438 };
439 
440 static int
441 em_hw_init(struct e1000_hw *hw)
442 {
443 	int diag;
444 
445 	diag = hw->mac.ops.init_params(hw);
446 	if (diag != 0) {
447 		PMD_INIT_LOG(ERR, "MAC Initialization Error");
448 		return diag;
449 	}
450 	diag = hw->nvm.ops.init_params(hw);
451 	if (diag != 0) {
452 		PMD_INIT_LOG(ERR, "NVM Initialization Error");
453 		return diag;
454 	}
455 	diag = hw->phy.ops.init_params(hw);
456 	if (diag != 0) {
457 		PMD_INIT_LOG(ERR, "PHY Initialization Error");
458 		return diag;
459 	}
460 	(void) e1000_get_bus_info(hw);
461 
462 	hw->mac.autoneg = 1;
463 	hw->phy.autoneg_wait_to_complete = 0;
464 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
465 
466 	e1000_init_script_state_82541(hw, TRUE);
467 	e1000_set_tbi_compatibility_82543(hw, TRUE);
468 
469 	/* Copper options */
470 	if (hw->phy.media_type == e1000_media_type_copper) {
471 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
472 		hw->phy.disable_polarity_correction = 0;
473 		hw->phy.ms_type = e1000_ms_hw_default;
474 	}
475 
476 	/*
477 	 * Start from a known state, this is important in reading the nvm
478 	 * and mac from that.
479 	 */
480 	e1000_reset_hw(hw);
481 
482 	/* Make sure we have a good EEPROM before we read from it */
483 	if (e1000_validate_nvm_checksum(hw) < 0) {
484 		/*
485 		 * Some PCI-E parts fail the first check due to
486 		 * the link being in sleep state, call it again,
487 		 * if it fails a second time its a real issue.
488 		 */
489 		diag = e1000_validate_nvm_checksum(hw);
490 		if (diag < 0) {
491 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
492 			goto error;
493 		}
494 	}
495 
496 	/* Read the permanent MAC address out of the EEPROM */
497 	diag = e1000_read_mac_addr(hw);
498 	if (diag != 0) {
499 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
500 		goto error;
501 	}
502 
503 	/* Now initialize the hardware */
504 	diag = em_hardware_init(hw);
505 	if (diag != 0) {
506 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
507 		goto error;
508 	}
509 
510 	hw->mac.get_link_status = 1;
511 
512 	/* Indicate SOL/IDER usage */
513 	diag = e1000_check_reset_block(hw);
514 	if (diag < 0) {
515 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to "
516 			"SOL/IDER session");
517 	}
518 	return 0;
519 
520 error:
521 	em_hw_control_release(hw);
522 	return diag;
523 }
524 
525 static int
526 eth_em_configure(struct rte_eth_dev *dev)
527 {
528 	struct e1000_interrupt *intr =
529 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
530 
531 	PMD_INIT_FUNC_TRACE();
532 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
533 	PMD_INIT_FUNC_TRACE();
534 
535 	return 0;
536 }
537 
538 static void
539 em_set_pba(struct e1000_hw *hw)
540 {
541 	uint32_t pba;
542 
543 	/*
544 	 * Packet Buffer Allocation (PBA)
545 	 * Writing PBA sets the receive portion of the buffer
546 	 * the remainder is used for the transmit buffer.
547 	 * Devices before the 82547 had a Packet Buffer of 64K.
548 	 * After the 82547 the buffer was reduced to 40K.
549 	 */
550 	switch (hw->mac.type) {
551 		case e1000_82547:
552 		case e1000_82547_rev_2:
553 		/* 82547: Total Packet Buffer is 40K */
554 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
555 			break;
556 		case e1000_82571:
557 		case e1000_82572:
558 		case e1000_80003es2lan:
559 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
560 			break;
561 		case e1000_82573: /* 82573: Total Packet Buffer is 32K */
562 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
563 			break;
564 		case e1000_82574:
565 		case e1000_82583:
566 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
567 			break;
568 		case e1000_ich8lan:
569 			pba = E1000_PBA_8K;
570 			break;
571 		case e1000_ich9lan:
572 		case e1000_ich10lan:
573 			pba = E1000_PBA_10K;
574 			break;
575 		case e1000_pchlan:
576 		case e1000_pch2lan:
577 		case e1000_pch_lpt:
578 		case e1000_pch_spt:
579 		case e1000_pch_cnp:
580 			pba = E1000_PBA_26K;
581 			break;
582 		default:
583 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
584 	}
585 
586 	E1000_WRITE_REG(hw, E1000_PBA, pba);
587 }
588 
589 static int
590 eth_em_start(struct rte_eth_dev *dev)
591 {
592 	struct e1000_adapter *adapter =
593 		E1000_DEV_PRIVATE(dev->data->dev_private);
594 	struct e1000_hw *hw =
595 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
596 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
597 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
598 	int ret, mask;
599 	uint32_t intr_vector = 0;
600 	uint32_t *speeds;
601 	int num_speeds;
602 	bool autoneg;
603 
604 	PMD_INIT_FUNC_TRACE();
605 
606 	eth_em_stop(dev);
607 
608 	e1000_power_up_phy(hw);
609 
610 	/* Set default PBA value */
611 	em_set_pba(hw);
612 
613 	/* Put the address into the Receive Address Array */
614 	e1000_rar_set(hw, hw->mac.addr, 0);
615 
616 	/*
617 	 * With the 82571 adapter, RAR[0] may be overwritten
618 	 * when the other port is reset, we make a duplicate
619 	 * in RAR[14] for that eventuality, this assures
620 	 * the interface continues to function.
621 	 */
622 	if (hw->mac.type == e1000_82571) {
623 		e1000_set_laa_state_82571(hw, TRUE);
624 		e1000_rar_set(hw, hw->mac.addr, E1000_RAR_ENTRIES - 1);
625 	}
626 
627 	/* Initialize the hardware */
628 	if (em_hardware_init(hw)) {
629 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
630 		return -EIO;
631 	}
632 
633 	E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN);
634 
635 	/* Configure for OS presence */
636 	em_init_manageability(hw);
637 
638 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
639 		intr_vector = dev->data->nb_rx_queues;
640 		if (rte_intr_efd_enable(intr_handle, intr_vector))
641 			return -1;
642 	}
643 
644 	if (rte_intr_dp_is_en(intr_handle)) {
645 		intr_handle->intr_vec =
646 			rte_zmalloc("intr_vec",
647 					dev->data->nb_rx_queues * sizeof(int), 0);
648 		if (intr_handle->intr_vec == NULL) {
649 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
650 						" intr_vec", dev->data->nb_rx_queues);
651 			return -ENOMEM;
652 		}
653 
654 		/* enable rx interrupt */
655 		em_rxq_intr_enable(hw);
656 	}
657 
658 	eth_em_tx_init(dev);
659 
660 	ret = eth_em_rx_init(dev);
661 	if (ret) {
662 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
663 		em_dev_clear_queues(dev);
664 		return ret;
665 	}
666 
667 	e1000_clear_hw_cntrs_base_generic(hw);
668 
669 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
670 			ETH_VLAN_EXTEND_MASK;
671 	ret = eth_em_vlan_offload_set(dev, mask);
672 	if (ret) {
673 		PMD_INIT_LOG(ERR, "Unable to update vlan offload");
674 		em_dev_clear_queues(dev);
675 		return ret;
676 	}
677 
678 	/* Set Interrupt Throttling Rate to maximum allowed value. */
679 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
680 
681 	/* Setup link speed and duplex */
682 	speeds = &dev->data->dev_conf.link_speeds;
683 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
684 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
685 		hw->mac.autoneg = 1;
686 	} else {
687 		num_speeds = 0;
688 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
689 
690 		/* Reset */
691 		hw->phy.autoneg_advertised = 0;
692 
693 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
694 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
695 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
696 			num_speeds = -1;
697 			goto error_invalid_config;
698 		}
699 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
700 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
701 			num_speeds++;
702 		}
703 		if (*speeds & ETH_LINK_SPEED_10M) {
704 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
705 			num_speeds++;
706 		}
707 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
708 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
709 			num_speeds++;
710 		}
711 		if (*speeds & ETH_LINK_SPEED_100M) {
712 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
713 			num_speeds++;
714 		}
715 		if (*speeds & ETH_LINK_SPEED_1G) {
716 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
717 			num_speeds++;
718 		}
719 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
720 			goto error_invalid_config;
721 
722 		/* Set/reset the mac.autoneg based on the link speed,
723 		 * fixed or not
724 		 */
725 		if (!autoneg) {
726 			hw->mac.autoneg = 0;
727 			hw->mac.forced_speed_duplex =
728 					hw->phy.autoneg_advertised;
729 		} else {
730 			hw->mac.autoneg = 1;
731 		}
732 	}
733 
734 	e1000_setup_link(hw);
735 
736 	if (rte_intr_allow_others(intr_handle)) {
737 		/* check if lsc interrupt is enabled */
738 		if (dev->data->dev_conf.intr_conf.lsc != 0) {
739 			ret = eth_em_interrupt_setup(dev);
740 			if (ret) {
741 				PMD_INIT_LOG(ERR, "Unable to setup interrupts");
742 				em_dev_clear_queues(dev);
743 				return ret;
744 			}
745 		}
746 	} else {
747 		rte_intr_callback_unregister(intr_handle,
748 						eth_em_interrupt_handler,
749 						(void *)dev);
750 		if (dev->data->dev_conf.intr_conf.lsc != 0)
751 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
752 				     " no intr multiplexn");
753 	}
754 	/* check if rxq interrupt is enabled */
755 	if (dev->data->dev_conf.intr_conf.rxq != 0)
756 		eth_em_rxq_interrupt_setup(dev);
757 
758 	rte_intr_enable(intr_handle);
759 
760 	adapter->stopped = 0;
761 
762 	PMD_INIT_LOG(DEBUG, "<<");
763 
764 	return 0;
765 
766 error_invalid_config:
767 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
768 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
769 	em_dev_clear_queues(dev);
770 	return -EINVAL;
771 }
772 
773 /*********************************************************************
774  *
775  *  This routine disables all traffic on the adapter by issuing a
776  *  global reset on the MAC.
777  *
778  **********************************************************************/
779 static void
780 eth_em_stop(struct rte_eth_dev *dev)
781 {
782 	struct rte_eth_link link;
783 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
784 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
785 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
786 
787 	em_rxq_intr_disable(hw);
788 	em_lsc_intr_disable(hw);
789 
790 	e1000_reset_hw(hw);
791 	if (hw->mac.type >= e1000_82544)
792 		E1000_WRITE_REG(hw, E1000_WUC, 0);
793 
794 	/* Power down the phy. Needed to make the link go down */
795 	e1000_power_down_phy(hw);
796 
797 	em_dev_clear_queues(dev);
798 
799 	/* clear the recorded link status */
800 	memset(&link, 0, sizeof(link));
801 	rte_em_dev_atomic_write_link_status(dev, &link);
802 
803 	if (!rte_intr_allow_others(intr_handle))
804 		/* resume to the default handler */
805 		rte_intr_callback_register(intr_handle,
806 					   eth_em_interrupt_handler,
807 					   (void *)dev);
808 
809 	/* Clean datapath event and queue/vec mapping */
810 	rte_intr_efd_disable(intr_handle);
811 	if (intr_handle->intr_vec != NULL) {
812 		rte_free(intr_handle->intr_vec);
813 		intr_handle->intr_vec = NULL;
814 	}
815 }
816 
817 static void
818 eth_em_close(struct rte_eth_dev *dev)
819 {
820 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
821 	struct e1000_adapter *adapter =
822 		E1000_DEV_PRIVATE(dev->data->dev_private);
823 
824 	eth_em_stop(dev);
825 	adapter->stopped = 1;
826 	em_dev_free_queues(dev);
827 	e1000_phy_hw_reset(hw);
828 	em_release_manageability(hw);
829 	em_hw_control_release(hw);
830 }
831 
832 static int
833 em_get_rx_buffer_size(struct e1000_hw *hw)
834 {
835 	uint32_t rx_buf_size;
836 
837 	rx_buf_size = ((E1000_READ_REG(hw, E1000_PBA) & UINT16_MAX) << 10);
838 	return rx_buf_size;
839 }
840 
841 /*********************************************************************
842  *
843  *  Initialize the hardware
844  *
845  **********************************************************************/
846 static int
847 em_hardware_init(struct e1000_hw *hw)
848 {
849 	uint32_t rx_buf_size;
850 	int diag;
851 
852 	/* Issue a global reset */
853 	e1000_reset_hw(hw);
854 
855 	/* Let the firmware know the OS is in control */
856 	em_hw_control_acquire(hw);
857 
858 	/*
859 	 * These parameters control the automatic generation (Tx) and
860 	 * response (Rx) to Ethernet PAUSE frames.
861 	 * - High water mark should allow for at least two standard size (1518)
862 	 *   frames to be received after sending an XOFF.
863 	 * - Low water mark works best when it is very near the high water mark.
864 	 *   This allows the receiver to restart by sending XON when it has
865 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
866 	 *   restart after one full frame is pulled from the buffer. There
867 	 *   could be several smaller frames in the buffer and if so they will
868 	 *   not trigger the XON until their total number reduces the buffer
869 	 *   by 1500.
870 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
871 	 */
872 	rx_buf_size = em_get_rx_buffer_size(hw);
873 
874 	hw->fc.high_water = rx_buf_size - PMD_ROUNDUP(ETHER_MAX_LEN * 2, 1024);
875 	hw->fc.low_water = hw->fc.high_water - 1500;
876 
877 	if (hw->mac.type == e1000_80003es2lan)
878 		hw->fc.pause_time = UINT16_MAX;
879 	else
880 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
881 
882 	hw->fc.send_xon = 1;
883 
884 	/* Set Flow control, use the tunable location if sane */
885 	if (em_fc_setting <= e1000_fc_full)
886 		hw->fc.requested_mode = em_fc_setting;
887 	else
888 		hw->fc.requested_mode = e1000_fc_none;
889 
890 	/* Workaround: no TX flow ctrl for PCH */
891 	if (hw->mac.type == e1000_pchlan)
892 		hw->fc.requested_mode = e1000_fc_rx_pause;
893 
894 	/* Override - settings for PCH2LAN, ya its magic :) */
895 	if (hw->mac.type == e1000_pch2lan) {
896 		hw->fc.high_water = 0x5C20;
897 		hw->fc.low_water = 0x5048;
898 		hw->fc.pause_time = 0x0650;
899 		hw->fc.refresh_time = 0x0400;
900 	} else if (hw->mac.type == e1000_pch_lpt ||
901 		   hw->mac.type == e1000_pch_spt ||
902 		   hw->mac.type == e1000_pch_cnp) {
903 		hw->fc.requested_mode = e1000_fc_full;
904 	}
905 
906 	diag = e1000_init_hw(hw);
907 	if (diag < 0)
908 		return diag;
909 	e1000_check_for_link(hw);
910 	return 0;
911 }
912 
913 /* This function is based on em_update_stats_counters() in e1000/if_em.c */
914 static int
915 eth_em_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
916 {
917 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
918 	struct e1000_hw_stats *stats =
919 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
920 	int pause_frames;
921 
922 	if(hw->phy.media_type == e1000_media_type_copper ||
923 			(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
924 		stats->symerrs += E1000_READ_REG(hw,E1000_SYMERRS);
925 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
926 	}
927 
928 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
929 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
930 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
931 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
932 
933 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
934 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
935 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
936 	stats->dc += E1000_READ_REG(hw, E1000_DC);
937 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
938 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
939 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
940 
941 	/*
942 	 * For watchdog management we need to know if we have been
943 	 * paused during the last interval, so capture that here.
944 	 */
945 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
946 	stats->xoffrxc += pause_frames;
947 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
948 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
949 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
950 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
951 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
952 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
953 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
954 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
955 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
956 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
957 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
958 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
959 
960 	/*
961 	 * For the 64-bit byte counters the low dword must be read first.
962 	 * Both registers clear on the read of the high dword.
963 	 */
964 
965 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
966 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
967 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
968 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
969 
970 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
971 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
972 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
973 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
974 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
975 
976 	stats->tor += E1000_READ_REG(hw, E1000_TORH);
977 	stats->tot += E1000_READ_REG(hw, E1000_TOTH);
978 
979 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
980 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
981 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
982 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
983 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
984 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
985 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
986 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
987 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
988 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
989 
990 	/* Interrupt Counts */
991 
992 	if (hw->mac.type >= e1000_82571) {
993 		stats->iac += E1000_READ_REG(hw, E1000_IAC);
994 		stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
995 		stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
996 		stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
997 		stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
998 		stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
999 		stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
1000 		stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
1001 		stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
1002 	}
1003 
1004 	if (hw->mac.type >= e1000_82543) {
1005 		stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
1006 		stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
1007 		stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
1008 		stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
1009 		stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
1010 		stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
1011 	}
1012 
1013 	if (rte_stats == NULL)
1014 		return -EINVAL;
1015 
1016 	/* Rx Errors */
1017 	rte_stats->imissed = stats->mpc;
1018 	rte_stats->ierrors = stats->crcerrs +
1019 	                     stats->rlec + stats->ruc + stats->roc +
1020 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
1021 
1022 	/* Tx Errors */
1023 	rte_stats->oerrors = stats->ecol + stats->latecol;
1024 
1025 	rte_stats->ipackets = stats->gprc;
1026 	rte_stats->opackets = stats->gptc;
1027 	rte_stats->ibytes   = stats->gorc;
1028 	rte_stats->obytes   = stats->gotc;
1029 	return 0;
1030 }
1031 
1032 static void
1033 eth_em_stats_reset(struct rte_eth_dev *dev)
1034 {
1035 	struct e1000_hw_stats *hw_stats =
1036 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1037 
1038 	/* HW registers are cleared on read */
1039 	eth_em_stats_get(dev, NULL);
1040 
1041 	/* Reset software totals */
1042 	memset(hw_stats, 0, sizeof(*hw_stats));
1043 }
1044 
1045 static int
1046 eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1047 {
1048 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1049 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1050 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1051 
1052 	em_rxq_intr_enable(hw);
1053 	rte_intr_enable(intr_handle);
1054 
1055 	return 0;
1056 }
1057 
1058 static int
1059 eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1060 {
1061 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1062 
1063 	em_rxq_intr_disable(hw);
1064 
1065 	return 0;
1066 }
1067 
1068 static uint32_t
1069 em_get_max_pktlen(const struct e1000_hw *hw)
1070 {
1071 	switch (hw->mac.type) {
1072 	case e1000_82571:
1073 	case e1000_82572:
1074 	case e1000_ich9lan:
1075 	case e1000_ich10lan:
1076 	case e1000_pch2lan:
1077 	case e1000_pch_lpt:
1078 	case e1000_pch_spt:
1079 	case e1000_pch_cnp:
1080 	case e1000_82574:
1081 	case e1000_80003es2lan: /* 9K Jumbo Frame size */
1082 	case e1000_82583:
1083 		return 0x2412;
1084 	case e1000_pchlan:
1085 		return 0x1000;
1086 	/* Adapters that do not support jumbo frames */
1087 	case e1000_ich8lan:
1088 		return ETHER_MAX_LEN;
1089 	default:
1090 		return MAX_JUMBO_FRAME_SIZE;
1091 	}
1092 }
1093 
1094 static void
1095 eth_em_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1096 {
1097 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1098 
1099 	dev_info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1100 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1101 	dev_info->max_rx_pktlen = em_get_max_pktlen(hw);
1102 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1103 	dev_info->rx_offload_capa =
1104 		DEV_RX_OFFLOAD_VLAN_STRIP |
1105 		DEV_RX_OFFLOAD_IPV4_CKSUM |
1106 		DEV_RX_OFFLOAD_UDP_CKSUM  |
1107 		DEV_RX_OFFLOAD_TCP_CKSUM;
1108 	dev_info->tx_offload_capa =
1109 		DEV_TX_OFFLOAD_VLAN_INSERT |
1110 		DEV_TX_OFFLOAD_IPV4_CKSUM  |
1111 		DEV_TX_OFFLOAD_UDP_CKSUM   |
1112 		DEV_TX_OFFLOAD_TCP_CKSUM;
1113 
1114 	/*
1115 	 * Starting with 631xESB hw supports 2 TX/RX queues per port.
1116 	 * Unfortunatelly, all these nics have just one TX context.
1117 	 * So we have few choises for TX:
1118 	 * - Use just one TX queue.
1119 	 * - Allow cksum offload only for one TX queue.
1120 	 * - Don't allow TX cksum offload at all.
1121 	 * For now, option #1 was chosen.
1122 	 * To use second RX queue we have to use extended RX descriptor
1123 	 * (Multiple Receive Queues are mutually exclusive with UDP
1124 	 * fragmentation and are not supported when a legacy receive
1125 	 * descriptor format is used).
1126 	 * Which means separate RX routinies - as legacy nics (82540, 82545)
1127 	 * don't support extended RXD.
1128 	 * To avoid it we support just one RX queue for now (no RSS).
1129 	 */
1130 
1131 	dev_info->max_rx_queues = 1;
1132 	dev_info->max_tx_queues = 1;
1133 
1134 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1135 		.nb_max = E1000_MAX_RING_DESC,
1136 		.nb_min = E1000_MIN_RING_DESC,
1137 		.nb_align = EM_RXD_ALIGN,
1138 	};
1139 
1140 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1141 		.nb_max = E1000_MAX_RING_DESC,
1142 		.nb_min = E1000_MIN_RING_DESC,
1143 		.nb_align = EM_TXD_ALIGN,
1144 		.nb_seg_max = EM_TX_MAX_SEG,
1145 		.nb_mtu_seg_max = EM_TX_MAX_MTU_SEG,
1146 	};
1147 
1148 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1149 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1150 			ETH_LINK_SPEED_1G;
1151 }
1152 
1153 /* return 0 means link status changed, -1 means not changed */
1154 static int
1155 eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1156 {
1157 	struct e1000_hw *hw =
1158 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1159 	struct rte_eth_link link, old;
1160 	int link_check, count;
1161 
1162 	link_check = 0;
1163 	hw->mac.get_link_status = 1;
1164 
1165 	/* possible wait-to-complete in up to 9 seconds */
1166 	for (count = 0; count < EM_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
1167 		/* Read the real link status */
1168 		switch (hw->phy.media_type) {
1169 		case e1000_media_type_copper:
1170 			/* Do the work to read phy */
1171 			e1000_check_for_link(hw);
1172 			link_check = !hw->mac.get_link_status;
1173 			break;
1174 
1175 		case e1000_media_type_fiber:
1176 			e1000_check_for_link(hw);
1177 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1178 					E1000_STATUS_LU);
1179 			break;
1180 
1181 		case e1000_media_type_internal_serdes:
1182 			e1000_check_for_link(hw);
1183 			link_check = hw->mac.serdes_has_link;
1184 			break;
1185 
1186 		default:
1187 			break;
1188 		}
1189 		if (link_check || wait_to_complete == 0)
1190 			break;
1191 		rte_delay_ms(EM_LINK_UPDATE_CHECK_INTERVAL);
1192 	}
1193 	memset(&link, 0, sizeof(link));
1194 	rte_em_dev_atomic_read_link_status(dev, &link);
1195 	old = link;
1196 
1197 	/* Now we check if a transition has happened */
1198 	if (link_check && (link.link_status == ETH_LINK_DOWN)) {
1199 		uint16_t duplex, speed;
1200 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
1201 		link.link_duplex = (duplex == FULL_DUPLEX) ?
1202 				ETH_LINK_FULL_DUPLEX :
1203 				ETH_LINK_HALF_DUPLEX;
1204 		link.link_speed = speed;
1205 		link.link_status = ETH_LINK_UP;
1206 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1207 				ETH_LINK_SPEED_FIXED);
1208 	} else if (!link_check && (link.link_status == ETH_LINK_UP)) {
1209 		link.link_speed = 0;
1210 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
1211 		link.link_status = ETH_LINK_DOWN;
1212 		link.link_autoneg = ETH_LINK_FIXED;
1213 	}
1214 	rte_em_dev_atomic_write_link_status(dev, &link);
1215 
1216 	/* not changed */
1217 	if (old.link_status == link.link_status)
1218 		return -1;
1219 
1220 	/* changed */
1221 	return 0;
1222 }
1223 
1224 /*
1225  * em_hw_control_acquire sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
1226  * For ASF and Pass Through versions of f/w this means
1227  * that the driver is loaded. For AMT version type f/w
1228  * this means that the network i/f is open.
1229  */
1230 static void
1231 em_hw_control_acquire(struct e1000_hw *hw)
1232 {
1233 	uint32_t ctrl_ext, swsm;
1234 
1235 	/* Let firmware know the driver has taken over */
1236 	if (hw->mac.type == e1000_82573) {
1237 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1238 		E1000_WRITE_REG(hw, E1000_SWSM, swsm | E1000_SWSM_DRV_LOAD);
1239 
1240 	} else {
1241 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1242 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1243 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1244 	}
1245 }
1246 
1247 /*
1248  * em_hw_control_release resets {CTRL_EXTT|FWSM}:DRV_LOAD bit.
1249  * For ASF and Pass Through versions of f/w this means that the
1250  * driver is no longer loaded. For AMT versions of the
1251  * f/w this means that the network i/f is closed.
1252  */
1253 static void
1254 em_hw_control_release(struct e1000_hw *hw)
1255 {
1256 	uint32_t ctrl_ext, swsm;
1257 
1258 	/* Let firmware taken over control of h/w */
1259 	if (hw->mac.type == e1000_82573) {
1260 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1261 		E1000_WRITE_REG(hw, E1000_SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
1262 	} else {
1263 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1264 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1265 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1266 	}
1267 }
1268 
1269 /*
1270  * Bit of a misnomer, what this really means is
1271  * to enable OS management of the system... aka
1272  * to disable special hardware management features.
1273  */
1274 static void
1275 em_init_manageability(struct e1000_hw *hw)
1276 {
1277 	if (e1000_enable_mng_pass_thru(hw)) {
1278 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
1279 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
1280 
1281 		/* disable hardware interception of ARP */
1282 		manc &= ~(E1000_MANC_ARP_EN);
1283 
1284 		/* enable receiving management packets to the host */
1285 		manc |= E1000_MANC_EN_MNG2HOST;
1286 		manc2h |= 1 << 5;  /* Mng Port 623 */
1287 		manc2h |= 1 << 6;  /* Mng Port 664 */
1288 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
1289 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1290 	}
1291 }
1292 
1293 /*
1294  * Give control back to hardware management
1295  * controller if there is one.
1296  */
1297 static void
1298 em_release_manageability(struct e1000_hw *hw)
1299 {
1300 	uint32_t manc;
1301 
1302 	if (e1000_enable_mng_pass_thru(hw)) {
1303 		manc = E1000_READ_REG(hw, E1000_MANC);
1304 
1305 		/* re-enable hardware interception of ARP */
1306 		manc |= E1000_MANC_ARP_EN;
1307 		manc &= ~E1000_MANC_EN_MNG2HOST;
1308 
1309 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1310 	}
1311 }
1312 
1313 static void
1314 eth_em_promiscuous_enable(struct rte_eth_dev *dev)
1315 {
1316 	struct e1000_hw *hw =
1317 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1318 	uint32_t rctl;
1319 
1320 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1321 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1322 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1323 }
1324 
1325 static void
1326 eth_em_promiscuous_disable(struct rte_eth_dev *dev)
1327 {
1328 	struct e1000_hw *hw =
1329 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1330 	uint32_t rctl;
1331 
1332 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1333 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_SBP);
1334 	if (dev->data->all_multicast == 1)
1335 		rctl |= E1000_RCTL_MPE;
1336 	else
1337 		rctl &= (~E1000_RCTL_MPE);
1338 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1339 }
1340 
1341 static void
1342 eth_em_allmulticast_enable(struct rte_eth_dev *dev)
1343 {
1344 	struct e1000_hw *hw =
1345 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1346 	uint32_t rctl;
1347 
1348 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1349 	rctl |= E1000_RCTL_MPE;
1350 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1351 }
1352 
1353 static void
1354 eth_em_allmulticast_disable(struct rte_eth_dev *dev)
1355 {
1356 	struct e1000_hw *hw =
1357 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1358 	uint32_t rctl;
1359 
1360 	if (dev->data->promiscuous == 1)
1361 		return; /* must remain in all_multicast mode */
1362 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1363 	rctl &= (~E1000_RCTL_MPE);
1364 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1365 }
1366 
1367 static int
1368 eth_em_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1369 {
1370 	struct e1000_hw *hw =
1371 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1372 	struct e1000_vfta * shadow_vfta =
1373 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1374 	uint32_t vfta;
1375 	uint32_t vid_idx;
1376 	uint32_t vid_bit;
1377 
1378 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
1379 			      E1000_VFTA_ENTRY_MASK);
1380 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
1381 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
1382 	if (on)
1383 		vfta |= vid_bit;
1384 	else
1385 		vfta &= ~vid_bit;
1386 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
1387 
1388 	/* update local VFTA copy */
1389 	shadow_vfta->vfta[vid_idx] = vfta;
1390 
1391 	return 0;
1392 }
1393 
1394 static void
1395 em_vlan_hw_filter_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 	/* Filter Table Disable */
1402 	reg = E1000_READ_REG(hw, E1000_RCTL);
1403 	reg &= ~E1000_RCTL_CFIEN;
1404 	reg &= ~E1000_RCTL_VFE;
1405 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1406 }
1407 
1408 static void
1409 em_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1410 {
1411 	struct e1000_hw *hw =
1412 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1413 	struct e1000_vfta * shadow_vfta =
1414 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1415 	uint32_t reg;
1416 	int i;
1417 
1418 	/* Filter Table Enable, CFI not used for packet acceptance */
1419 	reg = E1000_READ_REG(hw, E1000_RCTL);
1420 	reg &= ~E1000_RCTL_CFIEN;
1421 	reg |= E1000_RCTL_VFE;
1422 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1423 
1424 	/* restore vfta from local copy */
1425 	for (i = 0; i < IGB_VFTA_SIZE; i++)
1426 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
1427 }
1428 
1429 static void
1430 em_vlan_hw_strip_disable(struct rte_eth_dev *dev)
1431 {
1432 	struct e1000_hw *hw =
1433 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1434 	uint32_t reg;
1435 
1436 	/* VLAN Mode Disable */
1437 	reg = E1000_READ_REG(hw, E1000_CTRL);
1438 	reg &= ~E1000_CTRL_VME;
1439 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1440 
1441 }
1442 
1443 static void
1444 em_vlan_hw_strip_enable(struct rte_eth_dev *dev)
1445 {
1446 	struct e1000_hw *hw =
1447 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1448 	uint32_t reg;
1449 
1450 	/* VLAN Mode Enable */
1451 	reg = E1000_READ_REG(hw, E1000_CTRL);
1452 	reg |= E1000_CTRL_VME;
1453 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1454 }
1455 
1456 static int
1457 eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1458 {
1459 	if(mask & ETH_VLAN_STRIP_MASK){
1460 		if (dev->data->dev_conf.rxmode.hw_vlan_strip)
1461 			em_vlan_hw_strip_enable(dev);
1462 		else
1463 			em_vlan_hw_strip_disable(dev);
1464 	}
1465 
1466 	if(mask & ETH_VLAN_FILTER_MASK){
1467 		if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1468 			em_vlan_hw_filter_enable(dev);
1469 		else
1470 			em_vlan_hw_filter_disable(dev);
1471 	}
1472 
1473 	return 0;
1474 }
1475 
1476 /*
1477  * It enables the interrupt mask and then enable the interrupt.
1478  *
1479  * @param dev
1480  *  Pointer to struct rte_eth_dev.
1481  *
1482  * @return
1483  *  - On success, zero.
1484  *  - On failure, a negative value.
1485  */
1486 static int
1487 eth_em_interrupt_setup(struct rte_eth_dev *dev)
1488 {
1489 	uint32_t regval;
1490 	struct e1000_hw *hw =
1491 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1492 
1493 	/* clear interrupt */
1494 	E1000_READ_REG(hw, E1000_ICR);
1495 	regval = E1000_READ_REG(hw, E1000_IMS);
1496 	E1000_WRITE_REG(hw, E1000_IMS, regval | E1000_ICR_LSC);
1497 	return 0;
1498 }
1499 
1500 /*
1501  * It clears the interrupt causes and enables the interrupt.
1502  * It will be called once only during nic initialized.
1503  *
1504  * @param dev
1505  *  Pointer to struct rte_eth_dev.
1506  *
1507  * @return
1508  *  - On success, zero.
1509  *  - On failure, a negative value.
1510  */
1511 static int
1512 eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev)
1513 {
1514 	struct e1000_hw *hw =
1515 	E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1516 
1517 	E1000_READ_REG(hw, E1000_ICR);
1518 	em_rxq_intr_enable(hw);
1519 	return 0;
1520 }
1521 
1522 /*
1523  * It enable receive packet interrupt.
1524  * @param hw
1525  * Pointer to struct e1000_hw
1526  *
1527  * @return
1528  */
1529 static void
1530 em_rxq_intr_enable(struct e1000_hw *hw)
1531 {
1532 	E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_RXT0);
1533 	E1000_WRITE_FLUSH(hw);
1534 }
1535 
1536 /*
1537  * It disabled lsc interrupt.
1538  * @param hw
1539  * Pointer to struct e1000_hw
1540  *
1541  * @return
1542  */
1543 static void
1544 em_lsc_intr_disable(struct e1000_hw *hw)
1545 {
1546 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC);
1547 	E1000_WRITE_FLUSH(hw);
1548 }
1549 
1550 /*
1551  * It disabled receive packet interrupt.
1552  * @param hw
1553  * Pointer to struct e1000_hw
1554  *
1555  * @return
1556  */
1557 static void
1558 em_rxq_intr_disable(struct e1000_hw *hw)
1559 {
1560 	E1000_READ_REG(hw, E1000_ICR);
1561 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_RXT0);
1562 	E1000_WRITE_FLUSH(hw);
1563 }
1564 
1565 /*
1566  * It reads ICR and gets interrupt causes, check it and set a bit flag
1567  * to update link status.
1568  *
1569  * @param dev
1570  *  Pointer to struct rte_eth_dev.
1571  *
1572  * @return
1573  *  - On success, zero.
1574  *  - On failure, a negative value.
1575  */
1576 static int
1577 eth_em_interrupt_get_status(struct rte_eth_dev *dev)
1578 {
1579 	uint32_t icr;
1580 	struct e1000_hw *hw =
1581 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1582 	struct e1000_interrupt *intr =
1583 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1584 
1585 	/* read-on-clear nic registers here */
1586 	icr = E1000_READ_REG(hw, E1000_ICR);
1587 	if (icr & E1000_ICR_LSC) {
1588 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1589 	}
1590 
1591 	return 0;
1592 }
1593 
1594 /*
1595  * It executes link_update after knowing an interrupt is prsent.
1596  *
1597  * @param dev
1598  *  Pointer to struct rte_eth_dev.
1599  *
1600  * @return
1601  *  - On success, zero.
1602  *  - On failure, a negative value.
1603  */
1604 static int
1605 eth_em_interrupt_action(struct rte_eth_dev *dev,
1606 			struct rte_intr_handle *intr_handle)
1607 {
1608 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1609 	struct e1000_hw *hw =
1610 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1611 	struct e1000_interrupt *intr =
1612 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1613 	uint32_t tctl, rctl;
1614 	struct rte_eth_link link;
1615 	int ret;
1616 
1617 	if (!(intr->flags & E1000_FLAG_NEED_LINK_UPDATE))
1618 		return -1;
1619 
1620 	intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
1621 	rte_intr_enable(intr_handle);
1622 
1623 	/* set get_link_status to check register later */
1624 	hw->mac.get_link_status = 1;
1625 	ret = eth_em_link_update(dev, 0);
1626 
1627 	/* check if link has changed */
1628 	if (ret < 0)
1629 		return 0;
1630 
1631 	memset(&link, 0, sizeof(link));
1632 	rte_em_dev_atomic_read_link_status(dev, &link);
1633 	if (link.link_status) {
1634 		PMD_INIT_LOG(INFO, " Port %d: Link Up - speed %u Mbps - %s",
1635 			     dev->data->port_id, link.link_speed,
1636 			     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
1637 			     "full-duplex" : "half-duplex");
1638 	} else {
1639 		PMD_INIT_LOG(INFO, " Port %d: Link Down", dev->data->port_id);
1640 	}
1641 	PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
1642 		     pci_dev->addr.domain, pci_dev->addr.bus,
1643 		     pci_dev->addr.devid, pci_dev->addr.function);
1644 
1645 	tctl = E1000_READ_REG(hw, E1000_TCTL);
1646 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1647 	if (link.link_status) {
1648 		/* enable Tx/Rx */
1649 		tctl |= E1000_TCTL_EN;
1650 		rctl |= E1000_RCTL_EN;
1651 	} else {
1652 		/* disable Tx/Rx */
1653 		tctl &= ~E1000_TCTL_EN;
1654 		rctl &= ~E1000_RCTL_EN;
1655 	}
1656 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
1657 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1658 	E1000_WRITE_FLUSH(hw);
1659 
1660 	return 0;
1661 }
1662 
1663 /**
1664  * Interrupt handler which shall be registered at first.
1665  *
1666  * @param handle
1667  *  Pointer to interrupt handle.
1668  * @param param
1669  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1670  *
1671  * @return
1672  *  void
1673  */
1674 static void
1675 eth_em_interrupt_handler(void *param)
1676 {
1677 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1678 
1679 	eth_em_interrupt_get_status(dev);
1680 	eth_em_interrupt_action(dev, dev->intr_handle);
1681 	_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL, NULL);
1682 }
1683 
1684 static int
1685 eth_em_led_on(struct rte_eth_dev *dev)
1686 {
1687 	struct e1000_hw *hw;
1688 
1689 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1690 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1691 }
1692 
1693 static int
1694 eth_em_led_off(struct rte_eth_dev *dev)
1695 {
1696 	struct e1000_hw *hw;
1697 
1698 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1699 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1700 }
1701 
1702 static int
1703 eth_em_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1704 {
1705 	struct e1000_hw *hw;
1706 	uint32_t ctrl;
1707 	int tx_pause;
1708 	int rx_pause;
1709 
1710 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1711 	fc_conf->pause_time = hw->fc.pause_time;
1712 	fc_conf->high_water = hw->fc.high_water;
1713 	fc_conf->low_water = hw->fc.low_water;
1714 	fc_conf->send_xon = hw->fc.send_xon;
1715 	fc_conf->autoneg = hw->mac.autoneg;
1716 
1717 	/*
1718 	 * Return rx_pause and tx_pause status according to actual setting of
1719 	 * the TFCE and RFCE bits in the CTRL register.
1720 	 */
1721 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
1722 	if (ctrl & E1000_CTRL_TFCE)
1723 		tx_pause = 1;
1724 	else
1725 		tx_pause = 0;
1726 
1727 	if (ctrl & E1000_CTRL_RFCE)
1728 		rx_pause = 1;
1729 	else
1730 		rx_pause = 0;
1731 
1732 	if (rx_pause && tx_pause)
1733 		fc_conf->mode = RTE_FC_FULL;
1734 	else if (rx_pause)
1735 		fc_conf->mode = RTE_FC_RX_PAUSE;
1736 	else if (tx_pause)
1737 		fc_conf->mode = RTE_FC_TX_PAUSE;
1738 	else
1739 		fc_conf->mode = RTE_FC_NONE;
1740 
1741 	return 0;
1742 }
1743 
1744 static int
1745 eth_em_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1746 {
1747 	struct e1000_hw *hw;
1748 	int err;
1749 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
1750 		e1000_fc_none,
1751 		e1000_fc_rx_pause,
1752 		e1000_fc_tx_pause,
1753 		e1000_fc_full
1754 	};
1755 	uint32_t rx_buf_size;
1756 	uint32_t max_high_water;
1757 	uint32_t rctl;
1758 
1759 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1760 	if (fc_conf->autoneg != hw->mac.autoneg)
1761 		return -ENOTSUP;
1762 	rx_buf_size = em_get_rx_buffer_size(hw);
1763 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
1764 
1765 	/* At least reserve one Ethernet frame for watermark */
1766 	max_high_water = rx_buf_size - ETHER_MAX_LEN;
1767 	if ((fc_conf->high_water > max_high_water) ||
1768 	    (fc_conf->high_water < fc_conf->low_water)) {
1769 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
1770 		PMD_INIT_LOG(ERR, "high water must <= 0x%x", max_high_water);
1771 		return -EINVAL;
1772 	}
1773 
1774 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
1775 	hw->fc.pause_time     = fc_conf->pause_time;
1776 	hw->fc.high_water     = fc_conf->high_water;
1777 	hw->fc.low_water      = fc_conf->low_water;
1778 	hw->fc.send_xon	      = fc_conf->send_xon;
1779 
1780 	err = e1000_setup_link_generic(hw);
1781 	if (err == E1000_SUCCESS) {
1782 
1783 		/* check if we want to forward MAC frames - driver doesn't have native
1784 		 * capability to do that, so we'll write the registers ourselves */
1785 
1786 		rctl = E1000_READ_REG(hw, E1000_RCTL);
1787 
1788 		/* set or clear MFLCN.PMCF bit depending on configuration */
1789 		if (fc_conf->mac_ctrl_frame_fwd != 0)
1790 			rctl |= E1000_RCTL_PMCF;
1791 		else
1792 			rctl &= ~E1000_RCTL_PMCF;
1793 
1794 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1795 		E1000_WRITE_FLUSH(hw);
1796 
1797 		return 0;
1798 	}
1799 
1800 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
1801 	return -EIO;
1802 }
1803 
1804 static int
1805 eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
1806 		uint32_t index, __rte_unused uint32_t pool)
1807 {
1808 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1809 
1810 	return e1000_rar_set(hw, mac_addr->addr_bytes, index);
1811 }
1812 
1813 static void
1814 eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1815 {
1816 	uint8_t addr[ETHER_ADDR_LEN];
1817 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1818 
1819 	memset(addr, 0, sizeof(addr));
1820 
1821 	e1000_rar_set(hw, addr, index);
1822 }
1823 
1824 static int
1825 eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1826 {
1827 	struct rte_eth_dev_info dev_info;
1828 	struct e1000_hw *hw;
1829 	uint32_t frame_size;
1830 	uint32_t rctl;
1831 
1832 	eth_em_infos_get(dev, &dev_info);
1833 	frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + VLAN_TAG_SIZE;
1834 
1835 	/* check that mtu is within the allowed range */
1836 	if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
1837 		return -EINVAL;
1838 
1839 	/* refuse mtu that requires the support of scattered packets when this
1840 	 * feature has not been enabled before. */
1841 	if (!dev->data->scattered_rx &&
1842 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
1843 		return -EINVAL;
1844 
1845 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1846 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1847 
1848 	/* switch to jumbo mode if needed */
1849 	if (frame_size > ETHER_MAX_LEN) {
1850 		dev->data->dev_conf.rxmode.jumbo_frame = 1;
1851 		rctl |= E1000_RCTL_LPE;
1852 	} else {
1853 		dev->data->dev_conf.rxmode.jumbo_frame = 0;
1854 		rctl &= ~E1000_RCTL_LPE;
1855 	}
1856 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1857 
1858 	/* update max frame size */
1859 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1860 	return 0;
1861 }
1862 
1863 static int
1864 eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
1865 			struct ether_addr *mc_addr_set,
1866 			uint32_t nb_mc_addr)
1867 {
1868 	struct e1000_hw *hw;
1869 
1870 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1871 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1872 	return 0;
1873 }
1874 
1875 RTE_PMD_REGISTER_PCI(net_e1000_em, rte_em_pmd);
1876 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_em, pci_id_em_map);
1877 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_em, "* igb_uio | uio_pci_generic | vfio-pci");
1878