xref: /dpdk/drivers/net/igc/igc_ethdev.c (revision ff4e52ef)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2019-2020 Intel Corporation
3  */
4 
5 #include <stdint.h>
6 #include <string.h>
7 
8 #include <rte_string_fns.h>
9 #include <rte_pci.h>
10 #include <rte_bus_pci.h>
11 #include <ethdev_driver.h>
12 #include <ethdev_pci.h>
13 #include <rte_malloc.h>
14 #include <rte_alarm.h>
15 
16 #include "igc_logs.h"
17 #include "igc_txrx.h"
18 #include "igc_filter.h"
19 #include "igc_flow.h"
20 
21 #define IGC_INTEL_VENDOR_ID		0x8086
22 
23 /*
24  * The overhead from MTU to max frame size.
25  * Considering VLAN so tag needs to be counted.
26  */
27 #define IGC_ETH_OVERHEAD		(RTE_ETHER_HDR_LEN + \
28 					RTE_ETHER_CRC_LEN + VLAN_TAG_SIZE)
29 
30 #define IGC_FC_PAUSE_TIME		0x0680
31 #define IGC_LINK_UPDATE_CHECK_TIMEOUT	90  /* 9s */
32 #define IGC_LINK_UPDATE_CHECK_INTERVAL	100 /* ms */
33 
34 #define IGC_MISC_VEC_ID			RTE_INTR_VEC_ZERO_OFFSET
35 #define IGC_RX_VEC_START		RTE_INTR_VEC_RXTX_OFFSET
36 #define IGC_MSIX_OTHER_INTR_VEC		0   /* MSI-X other interrupt vector */
37 #define IGC_FLAG_NEED_LINK_UPDATE	(1u << 0)	/* need update link */
38 
39 #define IGC_DEFAULT_RX_FREE_THRESH	32
40 
41 #define IGC_DEFAULT_RX_PTHRESH		8
42 #define IGC_DEFAULT_RX_HTHRESH		8
43 #define IGC_DEFAULT_RX_WTHRESH		4
44 
45 #define IGC_DEFAULT_TX_PTHRESH		8
46 #define IGC_DEFAULT_TX_HTHRESH		1
47 #define IGC_DEFAULT_TX_WTHRESH		16
48 
49 /* MSI-X other interrupt vector */
50 #define IGC_MSIX_OTHER_INTR_VEC		0
51 
52 /* External VLAN Enable bit mask */
53 #define IGC_CTRL_EXT_EXT_VLAN		(1u << 26)
54 
55 /* Speed select */
56 #define IGC_CTRL_SPEED_MASK		(7u << 8)
57 #define IGC_CTRL_SPEED_2500		(6u << 8)
58 
59 /* External VLAN Ether Type bit mask and shift */
60 #define IGC_VET_EXT			0xFFFF0000
61 #define IGC_VET_EXT_SHIFT		16
62 
63 /* Force EEE Auto-negotiation */
64 #define IGC_EEER_EEE_FRC_AN		(1u << 28)
65 
66 /* Per Queue Good Packets Received Count */
67 #define IGC_PQGPRC(idx)		(0x10010 + 0x100 * (idx))
68 /* Per Queue Good Octets Received Count */
69 #define IGC_PQGORC(idx)		(0x10018 + 0x100 * (idx))
70 /* Per Queue Good Octets Transmitted Count */
71 #define IGC_PQGOTC(idx)		(0x10034 + 0x100 * (idx))
72 /* Per Queue Multicast Packets Received Count */
73 #define IGC_PQMPRC(idx)		(0x10038 + 0x100 * (idx))
74 /* Transmit Queue Drop Packet Count */
75 #define IGC_TQDPC(idx)		(0xe030 + 0x40 * (idx))
76 
77 #if RTE_BYTE_ORDER == RTE_LITTLE_ENDIAN
78 #define U32_0_IN_U64		0	/* lower bytes of u64 */
79 #define U32_1_IN_U64		1	/* higher bytes of u64 */
80 #else
81 #define U32_0_IN_U64		1
82 #define U32_1_IN_U64		0
83 #endif
84 
85 #define IGC_ALARM_INTERVAL	8000000u
86 /* us, about 13.6s some per-queue registers will wrap around back to 0. */
87 
88 static const struct rte_eth_desc_lim rx_desc_lim = {
89 	.nb_max = IGC_MAX_RXD,
90 	.nb_min = IGC_MIN_RXD,
91 	.nb_align = IGC_RXD_ALIGN,
92 };
93 
94 static const struct rte_eth_desc_lim tx_desc_lim = {
95 	.nb_max = IGC_MAX_TXD,
96 	.nb_min = IGC_MIN_TXD,
97 	.nb_align = IGC_TXD_ALIGN,
98 	.nb_seg_max = IGC_TX_MAX_SEG,
99 	.nb_mtu_seg_max = IGC_TX_MAX_MTU_SEG,
100 };
101 
102 static const struct rte_pci_id pci_id_igc_map[] = {
103 	{ RTE_PCI_DEVICE(IGC_INTEL_VENDOR_ID, IGC_DEV_ID_I225_LM) },
104 	{ RTE_PCI_DEVICE(IGC_INTEL_VENDOR_ID, IGC_DEV_ID_I225_V)  },
105 	{ RTE_PCI_DEVICE(IGC_INTEL_VENDOR_ID, IGC_DEV_ID_I225_I)  },
106 	{ RTE_PCI_DEVICE(IGC_INTEL_VENDOR_ID, IGC_DEV_ID_I225_K)  },
107 	{ .vendor_id = 0, /* sentinel */ },
108 };
109 
110 /* store statistics names and its offset in stats structure */
111 struct rte_igc_xstats_name_off {
112 	char name[RTE_ETH_XSTATS_NAME_SIZE];
113 	unsigned int offset;
114 };
115 
116 static const struct rte_igc_xstats_name_off rte_igc_stats_strings[] = {
117 	{"rx_crc_errors", offsetof(struct igc_hw_stats, crcerrs)},
118 	{"rx_align_errors", offsetof(struct igc_hw_stats, algnerrc)},
119 	{"rx_errors", offsetof(struct igc_hw_stats, rxerrc)},
120 	{"rx_missed_packets", offsetof(struct igc_hw_stats, mpc)},
121 	{"tx_single_collision_packets", offsetof(struct igc_hw_stats, scc)},
122 	{"tx_multiple_collision_packets", offsetof(struct igc_hw_stats, mcc)},
123 	{"tx_excessive_collision_packets", offsetof(struct igc_hw_stats,
124 		ecol)},
125 	{"tx_late_collisions", offsetof(struct igc_hw_stats, latecol)},
126 	{"tx_total_collisions", offsetof(struct igc_hw_stats, colc)},
127 	{"tx_deferred_packets", offsetof(struct igc_hw_stats, dc)},
128 	{"tx_no_carrier_sense_packets", offsetof(struct igc_hw_stats, tncrs)},
129 	{"tx_discarded_packets", offsetof(struct igc_hw_stats, htdpmc)},
130 	{"rx_length_errors", offsetof(struct igc_hw_stats, rlec)},
131 	{"rx_xon_packets", offsetof(struct igc_hw_stats, xonrxc)},
132 	{"tx_xon_packets", offsetof(struct igc_hw_stats, xontxc)},
133 	{"rx_xoff_packets", offsetof(struct igc_hw_stats, xoffrxc)},
134 	{"tx_xoff_packets", offsetof(struct igc_hw_stats, xofftxc)},
135 	{"rx_flow_control_unsupported_packets", offsetof(struct igc_hw_stats,
136 		fcruc)},
137 	{"rx_size_64_packets", offsetof(struct igc_hw_stats, prc64)},
138 	{"rx_size_65_to_127_packets", offsetof(struct igc_hw_stats, prc127)},
139 	{"rx_size_128_to_255_packets", offsetof(struct igc_hw_stats, prc255)},
140 	{"rx_size_256_to_511_packets", offsetof(struct igc_hw_stats, prc511)},
141 	{"rx_size_512_to_1023_packets", offsetof(struct igc_hw_stats,
142 		prc1023)},
143 	{"rx_size_1024_to_max_packets", offsetof(struct igc_hw_stats,
144 		prc1522)},
145 	{"rx_broadcast_packets", offsetof(struct igc_hw_stats, bprc)},
146 	{"rx_multicast_packets", offsetof(struct igc_hw_stats, mprc)},
147 	{"rx_undersize_errors", offsetof(struct igc_hw_stats, ruc)},
148 	{"rx_fragment_errors", offsetof(struct igc_hw_stats, rfc)},
149 	{"rx_oversize_errors", offsetof(struct igc_hw_stats, roc)},
150 	{"rx_jabber_errors", offsetof(struct igc_hw_stats, rjc)},
151 	{"rx_no_buffers", offsetof(struct igc_hw_stats, rnbc)},
152 	{"rx_management_packets", offsetof(struct igc_hw_stats, mgprc)},
153 	{"rx_management_dropped", offsetof(struct igc_hw_stats, mgpdc)},
154 	{"tx_management_packets", offsetof(struct igc_hw_stats, mgptc)},
155 	{"rx_total_packets", offsetof(struct igc_hw_stats, tpr)},
156 	{"tx_total_packets", offsetof(struct igc_hw_stats, tpt)},
157 	{"rx_total_bytes", offsetof(struct igc_hw_stats, tor)},
158 	{"tx_total_bytes", offsetof(struct igc_hw_stats, tot)},
159 	{"tx_size_64_packets", offsetof(struct igc_hw_stats, ptc64)},
160 	{"tx_size_65_to_127_packets", offsetof(struct igc_hw_stats, ptc127)},
161 	{"tx_size_128_to_255_packets", offsetof(struct igc_hw_stats, ptc255)},
162 	{"tx_size_256_to_511_packets", offsetof(struct igc_hw_stats, ptc511)},
163 	{"tx_size_512_to_1023_packets", offsetof(struct igc_hw_stats,
164 		ptc1023)},
165 	{"tx_size_1023_to_max_packets", offsetof(struct igc_hw_stats,
166 		ptc1522)},
167 	{"tx_multicast_packets", offsetof(struct igc_hw_stats, mptc)},
168 	{"tx_broadcast_packets", offsetof(struct igc_hw_stats, bptc)},
169 	{"tx_tso_packets", offsetof(struct igc_hw_stats, tsctc)},
170 	{"rx_sent_to_host_packets", offsetof(struct igc_hw_stats, rpthc)},
171 	{"tx_sent_by_host_packets", offsetof(struct igc_hw_stats, hgptc)},
172 	{"interrupt_assert_count", offsetof(struct igc_hw_stats, iac)},
173 	{"rx_descriptor_lower_threshold",
174 		offsetof(struct igc_hw_stats, icrxdmtc)},
175 };
176 
177 #define IGC_NB_XSTATS (sizeof(rte_igc_stats_strings) / \
178 		sizeof(rte_igc_stats_strings[0]))
179 
180 static int eth_igc_configure(struct rte_eth_dev *dev);
181 static int eth_igc_link_update(struct rte_eth_dev *dev, int wait_to_complete);
182 static int eth_igc_stop(struct rte_eth_dev *dev);
183 static int eth_igc_start(struct rte_eth_dev *dev);
184 static int eth_igc_set_link_up(struct rte_eth_dev *dev);
185 static int eth_igc_set_link_down(struct rte_eth_dev *dev);
186 static int eth_igc_close(struct rte_eth_dev *dev);
187 static int eth_igc_reset(struct rte_eth_dev *dev);
188 static int eth_igc_promiscuous_enable(struct rte_eth_dev *dev);
189 static int eth_igc_promiscuous_disable(struct rte_eth_dev *dev);
190 static int eth_igc_fw_version_get(struct rte_eth_dev *dev,
191 				char *fw_version, size_t fw_size);
192 static int eth_igc_infos_get(struct rte_eth_dev *dev,
193 			struct rte_eth_dev_info *dev_info);
194 static int eth_igc_led_on(struct rte_eth_dev *dev);
195 static int eth_igc_led_off(struct rte_eth_dev *dev);
196 static const uint32_t *eth_igc_supported_ptypes_get(struct rte_eth_dev *dev);
197 static int eth_igc_rar_set(struct rte_eth_dev *dev,
198 		struct rte_ether_addr *mac_addr, uint32_t index, uint32_t pool);
199 static void eth_igc_rar_clear(struct rte_eth_dev *dev, uint32_t index);
200 static int eth_igc_default_mac_addr_set(struct rte_eth_dev *dev,
201 			struct rte_ether_addr *addr);
202 static int eth_igc_set_mc_addr_list(struct rte_eth_dev *dev,
203 			 struct rte_ether_addr *mc_addr_set,
204 			 uint32_t nb_mc_addr);
205 static int eth_igc_allmulticast_enable(struct rte_eth_dev *dev);
206 static int eth_igc_allmulticast_disable(struct rte_eth_dev *dev);
207 static int eth_igc_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
208 static int eth_igc_stats_get(struct rte_eth_dev *dev,
209 			struct rte_eth_stats *rte_stats);
210 static int eth_igc_xstats_get(struct rte_eth_dev *dev,
211 			struct rte_eth_xstat *xstats, unsigned int n);
212 static int eth_igc_xstats_get_by_id(struct rte_eth_dev *dev,
213 				const uint64_t *ids,
214 				uint64_t *values, unsigned int n);
215 static int eth_igc_xstats_get_names(struct rte_eth_dev *dev,
216 				struct rte_eth_xstat_name *xstats_names,
217 				unsigned int size);
218 static int eth_igc_xstats_get_names_by_id(struct rte_eth_dev *dev,
219 		const uint64_t *ids, struct rte_eth_xstat_name *xstats_names,
220 		unsigned int limit);
221 static int eth_igc_xstats_reset(struct rte_eth_dev *dev);
222 static int
223 eth_igc_queue_stats_mapping_set(struct rte_eth_dev *dev,
224 	uint16_t queue_id, uint8_t stat_idx, uint8_t is_rx);
225 static int
226 eth_igc_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
227 static int
228 eth_igc_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
229 static int
230 eth_igc_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf);
231 static int
232 eth_igc_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf);
233 static int eth_igc_rss_reta_update(struct rte_eth_dev *dev,
234 			struct rte_eth_rss_reta_entry64 *reta_conf,
235 			uint16_t reta_size);
236 static int eth_igc_rss_reta_query(struct rte_eth_dev *dev,
237 		       struct rte_eth_rss_reta_entry64 *reta_conf,
238 		       uint16_t reta_size);
239 static int eth_igc_rss_hash_update(struct rte_eth_dev *dev,
240 			struct rte_eth_rss_conf *rss_conf);
241 static int eth_igc_rss_hash_conf_get(struct rte_eth_dev *dev,
242 			struct rte_eth_rss_conf *rss_conf);
243 static int
244 eth_igc_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on);
245 static int eth_igc_vlan_offload_set(struct rte_eth_dev *dev, int mask);
246 static int eth_igc_vlan_tpid_set(struct rte_eth_dev *dev,
247 		      enum rte_vlan_type vlan_type, uint16_t tpid);
248 
249 static const struct eth_dev_ops eth_igc_ops = {
250 	.dev_configure		= eth_igc_configure,
251 	.link_update		= eth_igc_link_update,
252 	.dev_stop		= eth_igc_stop,
253 	.dev_start		= eth_igc_start,
254 	.dev_close		= eth_igc_close,
255 	.dev_reset		= eth_igc_reset,
256 	.dev_set_link_up	= eth_igc_set_link_up,
257 	.dev_set_link_down	= eth_igc_set_link_down,
258 	.promiscuous_enable	= eth_igc_promiscuous_enable,
259 	.promiscuous_disable	= eth_igc_promiscuous_disable,
260 	.allmulticast_enable	= eth_igc_allmulticast_enable,
261 	.allmulticast_disable	= eth_igc_allmulticast_disable,
262 	.fw_version_get		= eth_igc_fw_version_get,
263 	.dev_infos_get		= eth_igc_infos_get,
264 	.dev_led_on		= eth_igc_led_on,
265 	.dev_led_off		= eth_igc_led_off,
266 	.dev_supported_ptypes_get = eth_igc_supported_ptypes_get,
267 	.mtu_set		= eth_igc_mtu_set,
268 	.mac_addr_add		= eth_igc_rar_set,
269 	.mac_addr_remove	= eth_igc_rar_clear,
270 	.mac_addr_set		= eth_igc_default_mac_addr_set,
271 	.set_mc_addr_list	= eth_igc_set_mc_addr_list,
272 
273 	.rx_queue_setup		= eth_igc_rx_queue_setup,
274 	.rx_queue_release	= eth_igc_rx_queue_release,
275 	.tx_queue_setup		= eth_igc_tx_queue_setup,
276 	.tx_queue_release	= eth_igc_tx_queue_release,
277 	.tx_done_cleanup	= eth_igc_tx_done_cleanup,
278 	.rxq_info_get		= eth_igc_rxq_info_get,
279 	.txq_info_get		= eth_igc_txq_info_get,
280 	.stats_get		= eth_igc_stats_get,
281 	.xstats_get		= eth_igc_xstats_get,
282 	.xstats_get_by_id	= eth_igc_xstats_get_by_id,
283 	.xstats_get_names_by_id	= eth_igc_xstats_get_names_by_id,
284 	.xstats_get_names	= eth_igc_xstats_get_names,
285 	.stats_reset		= eth_igc_xstats_reset,
286 	.xstats_reset		= eth_igc_xstats_reset,
287 	.queue_stats_mapping_set = eth_igc_queue_stats_mapping_set,
288 	.rx_queue_intr_enable	= eth_igc_rx_queue_intr_enable,
289 	.rx_queue_intr_disable	= eth_igc_rx_queue_intr_disable,
290 	.flow_ctrl_get		= eth_igc_flow_ctrl_get,
291 	.flow_ctrl_set		= eth_igc_flow_ctrl_set,
292 	.reta_update		= eth_igc_rss_reta_update,
293 	.reta_query		= eth_igc_rss_reta_query,
294 	.rss_hash_update	= eth_igc_rss_hash_update,
295 	.rss_hash_conf_get	= eth_igc_rss_hash_conf_get,
296 	.vlan_filter_set	= eth_igc_vlan_filter_set,
297 	.vlan_offload_set	= eth_igc_vlan_offload_set,
298 	.vlan_tpid_set		= eth_igc_vlan_tpid_set,
299 	.vlan_strip_queue_set	= eth_igc_vlan_strip_queue_set,
300 	.flow_ops_get		= eth_igc_flow_ops_get,
301 };
302 
303 /*
304  * multiple queue mode checking
305  */
306 static int
307 igc_check_mq_mode(struct rte_eth_dev *dev)
308 {
309 	enum rte_eth_rx_mq_mode rx_mq_mode = dev->data->dev_conf.rxmode.mq_mode;
310 	enum rte_eth_tx_mq_mode tx_mq_mode = dev->data->dev_conf.txmode.mq_mode;
311 
312 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
313 		PMD_INIT_LOG(ERR, "SRIOV is not supported.");
314 		return -EINVAL;
315 	}
316 
317 	if (rx_mq_mode != ETH_MQ_RX_NONE &&
318 		rx_mq_mode != ETH_MQ_RX_RSS) {
319 		/* RSS together with VMDq not supported*/
320 		PMD_INIT_LOG(ERR, "RX mode %d is not supported.",
321 				rx_mq_mode);
322 		return -EINVAL;
323 	}
324 
325 	/* To no break software that set invalid mode, only display
326 	 * warning if invalid mode is used.
327 	 */
328 	if (tx_mq_mode != ETH_MQ_TX_NONE)
329 		PMD_INIT_LOG(WARNING,
330 			"TX mode %d is not supported. Due to meaningless in this driver, just ignore",
331 			tx_mq_mode);
332 
333 	return 0;
334 }
335 
336 static int
337 eth_igc_configure(struct rte_eth_dev *dev)
338 {
339 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
340 	int ret;
341 
342 	PMD_INIT_FUNC_TRACE();
343 
344 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
345 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
346 
347 	ret  = igc_check_mq_mode(dev);
348 	if (ret != 0)
349 		return ret;
350 
351 	intr->flags |= IGC_FLAG_NEED_LINK_UPDATE;
352 	return 0;
353 }
354 
355 static int
356 eth_igc_set_link_up(struct rte_eth_dev *dev)
357 {
358 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
359 
360 	if (hw->phy.media_type == igc_media_type_copper)
361 		igc_power_up_phy(hw);
362 	else
363 		igc_power_up_fiber_serdes_link(hw);
364 	return 0;
365 }
366 
367 static int
368 eth_igc_set_link_down(struct rte_eth_dev *dev)
369 {
370 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
371 
372 	if (hw->phy.media_type == igc_media_type_copper)
373 		igc_power_down_phy(hw);
374 	else
375 		igc_shutdown_fiber_serdes_link(hw);
376 	return 0;
377 }
378 
379 /*
380  * disable other interrupt
381  */
382 static void
383 igc_intr_other_disable(struct rte_eth_dev *dev)
384 {
385 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
386 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
387 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
388 
389 	if (rte_intr_allow_others(intr_handle) &&
390 		dev->data->dev_conf.intr_conf.lsc) {
391 		IGC_WRITE_REG(hw, IGC_EIMC, 1u << IGC_MSIX_OTHER_INTR_VEC);
392 	}
393 
394 	IGC_WRITE_REG(hw, IGC_IMC, ~0);
395 	IGC_WRITE_FLUSH(hw);
396 }
397 
398 /*
399  * enable other interrupt
400  */
401 static inline void
402 igc_intr_other_enable(struct rte_eth_dev *dev)
403 {
404 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
405 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
406 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
407 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
408 
409 	if (rte_intr_allow_others(intr_handle) &&
410 		dev->data->dev_conf.intr_conf.lsc) {
411 		IGC_WRITE_REG(hw, IGC_EIMS, 1u << IGC_MSIX_OTHER_INTR_VEC);
412 	}
413 
414 	IGC_WRITE_REG(hw, IGC_IMS, intr->mask);
415 	IGC_WRITE_FLUSH(hw);
416 }
417 
418 /*
419  * It reads ICR and gets interrupt causes, check it and set a bit flag
420  * to update link status.
421  */
422 static void
423 eth_igc_interrupt_get_status(struct rte_eth_dev *dev)
424 {
425 	uint32_t icr;
426 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
427 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
428 
429 	/* read-on-clear nic registers here */
430 	icr = IGC_READ_REG(hw, IGC_ICR);
431 
432 	intr->flags = 0;
433 	if (icr & IGC_ICR_LSC)
434 		intr->flags |= IGC_FLAG_NEED_LINK_UPDATE;
435 }
436 
437 /* return 0 means link status changed, -1 means not changed */
438 static int
439 eth_igc_link_update(struct rte_eth_dev *dev, int wait_to_complete)
440 {
441 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
442 	struct rte_eth_link link;
443 	int link_check, count;
444 
445 	link_check = 0;
446 	hw->mac.get_link_status = 1;
447 
448 	/* possible wait-to-complete in up to 9 seconds */
449 	for (count = 0; count < IGC_LINK_UPDATE_CHECK_TIMEOUT; count++) {
450 		/* Read the real link status */
451 		switch (hw->phy.media_type) {
452 		case igc_media_type_copper:
453 			/* Do the work to read phy */
454 			igc_check_for_link(hw);
455 			link_check = !hw->mac.get_link_status;
456 			break;
457 
458 		case igc_media_type_fiber:
459 			igc_check_for_link(hw);
460 			link_check = (IGC_READ_REG(hw, IGC_STATUS) &
461 				      IGC_STATUS_LU);
462 			break;
463 
464 		case igc_media_type_internal_serdes:
465 			igc_check_for_link(hw);
466 			link_check = hw->mac.serdes_has_link;
467 			break;
468 
469 		default:
470 			break;
471 		}
472 		if (link_check || wait_to_complete == 0)
473 			break;
474 		rte_delay_ms(IGC_LINK_UPDATE_CHECK_INTERVAL);
475 	}
476 	memset(&link, 0, sizeof(link));
477 
478 	/* Now we check if a transition has happened */
479 	if (link_check) {
480 		uint16_t duplex, speed;
481 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
482 		link.link_duplex = (duplex == FULL_DUPLEX) ?
483 				ETH_LINK_FULL_DUPLEX :
484 				ETH_LINK_HALF_DUPLEX;
485 		link.link_speed = speed;
486 		link.link_status = ETH_LINK_UP;
487 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
488 				ETH_LINK_SPEED_FIXED);
489 
490 		if (speed == SPEED_2500) {
491 			uint32_t tipg = IGC_READ_REG(hw, IGC_TIPG);
492 			if ((tipg & IGC_TIPG_IPGT_MASK) != 0x0b) {
493 				tipg &= ~IGC_TIPG_IPGT_MASK;
494 				tipg |= 0x0b;
495 				IGC_WRITE_REG(hw, IGC_TIPG, tipg);
496 			}
497 		}
498 	} else {
499 		link.link_speed = 0;
500 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
501 		link.link_status = ETH_LINK_DOWN;
502 		link.link_autoneg = ETH_LINK_FIXED;
503 	}
504 
505 	return rte_eth_linkstatus_set(dev, &link);
506 }
507 
508 /*
509  * It executes link_update after knowing an interrupt is present.
510  */
511 static void
512 eth_igc_interrupt_action(struct rte_eth_dev *dev)
513 {
514 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
515 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
516 	struct rte_eth_link link;
517 	int ret;
518 
519 	if (intr->flags & IGC_FLAG_NEED_LINK_UPDATE) {
520 		intr->flags &= ~IGC_FLAG_NEED_LINK_UPDATE;
521 
522 		/* set get_link_status to check register later */
523 		ret = eth_igc_link_update(dev, 0);
524 
525 		/* check if link has changed */
526 		if (ret < 0)
527 			return;
528 
529 		rte_eth_linkstatus_get(dev, &link);
530 		if (link.link_status)
531 			PMD_DRV_LOG(INFO,
532 				" Port %d: Link Up - speed %u Mbps - %s",
533 				dev->data->port_id,
534 				(unsigned int)link.link_speed,
535 				link.link_duplex == ETH_LINK_FULL_DUPLEX ?
536 				"full-duplex" : "half-duplex");
537 		else
538 			PMD_DRV_LOG(INFO, " Port %d: Link Down",
539 				dev->data->port_id);
540 
541 		PMD_DRV_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
542 				pci_dev->addr.domain,
543 				pci_dev->addr.bus,
544 				pci_dev->addr.devid,
545 				pci_dev->addr.function);
546 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
547 	}
548 }
549 
550 /*
551  * Interrupt handler which shall be registered at first.
552  *
553  * @handle
554  *  Pointer to interrupt handle.
555  * @param
556  *  The address of parameter (struct rte_eth_dev *) registered before.
557  */
558 static void
559 eth_igc_interrupt_handler(void *param)
560 {
561 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
562 
563 	eth_igc_interrupt_get_status(dev);
564 	eth_igc_interrupt_action(dev);
565 }
566 
567 static void igc_read_queue_stats_register(struct rte_eth_dev *dev);
568 
569 /*
570  * Update the queue status every IGC_ALARM_INTERVAL time.
571  * @param
572  *  The address of parameter (struct rte_eth_dev *) registered before.
573  */
574 static void
575 igc_update_queue_stats_handler(void *param)
576 {
577 	struct rte_eth_dev *dev = param;
578 	igc_read_queue_stats_register(dev);
579 	rte_eal_alarm_set(IGC_ALARM_INTERVAL,
580 			igc_update_queue_stats_handler, dev);
581 }
582 
583 /*
584  * rx,tx enable/disable
585  */
586 static void
587 eth_igc_rxtx_control(struct rte_eth_dev *dev, bool enable)
588 {
589 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
590 	uint32_t tctl, rctl;
591 
592 	tctl = IGC_READ_REG(hw, IGC_TCTL);
593 	rctl = IGC_READ_REG(hw, IGC_RCTL);
594 
595 	if (enable) {
596 		/* enable Tx/Rx */
597 		tctl |= IGC_TCTL_EN;
598 		rctl |= IGC_RCTL_EN;
599 	} else {
600 		/* disable Tx/Rx */
601 		tctl &= ~IGC_TCTL_EN;
602 		rctl &= ~IGC_RCTL_EN;
603 	}
604 	IGC_WRITE_REG(hw, IGC_TCTL, tctl);
605 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
606 	IGC_WRITE_FLUSH(hw);
607 }
608 
609 /*
610  *  This routine disables all traffic on the adapter by issuing a
611  *  global reset on the MAC.
612  */
613 static int
614 eth_igc_stop(struct rte_eth_dev *dev)
615 {
616 	struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
617 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
618 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
619 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
620 	struct rte_eth_link link;
621 
622 	dev->data->dev_started = 0;
623 	adapter->stopped = 1;
624 
625 	/* disable receive and transmit */
626 	eth_igc_rxtx_control(dev, false);
627 
628 	/* disable all MSI-X interrupts */
629 	IGC_WRITE_REG(hw, IGC_EIMC, 0x1f);
630 	IGC_WRITE_FLUSH(hw);
631 
632 	/* clear all MSI-X interrupts */
633 	IGC_WRITE_REG(hw, IGC_EICR, 0x1f);
634 
635 	igc_intr_other_disable(dev);
636 
637 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
638 
639 	/* disable intr eventfd mapping */
640 	rte_intr_disable(intr_handle);
641 
642 	igc_reset_hw(hw);
643 
644 	/* disable all wake up */
645 	IGC_WRITE_REG(hw, IGC_WUC, 0);
646 
647 	/* disable checking EEE operation in MAC loopback mode */
648 	igc_read_reg_check_clear_bits(hw, IGC_EEER, IGC_EEER_EEE_FRC_AN);
649 
650 	/* Set bit for Go Link disconnect */
651 	igc_read_reg_check_set_bits(hw, IGC_82580_PHY_POWER_MGMT,
652 			IGC_82580_PM_GO_LINKD);
653 
654 	/* Power down the phy. Needed to make the link go Down */
655 	eth_igc_set_link_down(dev);
656 
657 	igc_dev_clear_queues(dev);
658 
659 	/* clear the recorded link status */
660 	memset(&link, 0, sizeof(link));
661 	rte_eth_linkstatus_set(dev, &link);
662 
663 	if (!rte_intr_allow_others(intr_handle))
664 		/* resume to the default handler */
665 		rte_intr_callback_register(intr_handle,
666 					   eth_igc_interrupt_handler,
667 					   (void *)dev);
668 
669 	/* Clean datapath event and queue/vec mapping */
670 	rte_intr_efd_disable(intr_handle);
671 	if (intr_handle->intr_vec != NULL) {
672 		rte_free(intr_handle->intr_vec);
673 		intr_handle->intr_vec = NULL;
674 	}
675 
676 	return 0;
677 }
678 
679 /*
680  * write interrupt vector allocation register
681  * @hw
682  *  board private structure
683  * @queue_index
684  *  queue index, valid 0,1,2,3
685  * @tx
686  *  tx:1, rx:0
687  * @msix_vector
688  *  msix-vector, valid 0,1,2,3,4
689  */
690 static void
691 igc_write_ivar(struct igc_hw *hw, uint8_t queue_index,
692 		bool tx, uint8_t msix_vector)
693 {
694 	uint8_t offset = 0;
695 	uint8_t reg_index = queue_index >> 1;
696 	uint32_t val;
697 
698 	/*
699 	 * IVAR(0)
700 	 * bit31...24	bit23...16	bit15...8	bit7...0
701 	 * TX1		RX1		TX0		RX0
702 	 *
703 	 * IVAR(1)
704 	 * bit31...24	bit23...16	bit15...8	bit7...0
705 	 * TX3		RX3		TX2		RX2
706 	 */
707 
708 	if (tx)
709 		offset = 8;
710 
711 	if (queue_index & 1)
712 		offset += 16;
713 
714 	val = IGC_READ_REG_ARRAY(hw, IGC_IVAR0, reg_index);
715 
716 	/* clear bits */
717 	val &= ~((uint32_t)0xFF << offset);
718 
719 	/* write vector and valid bit */
720 	val |= (uint32_t)(msix_vector | IGC_IVAR_VALID) << offset;
721 
722 	IGC_WRITE_REG_ARRAY(hw, IGC_IVAR0, reg_index, val);
723 }
724 
725 /* Sets up the hardware to generate MSI-X interrupts properly
726  * @hw
727  *  board private structure
728  */
729 static void
730 igc_configure_msix_intr(struct rte_eth_dev *dev)
731 {
732 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
733 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
734 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
735 
736 	uint32_t intr_mask;
737 	uint32_t vec = IGC_MISC_VEC_ID;
738 	uint32_t base = IGC_MISC_VEC_ID;
739 	uint32_t misc_shift = 0;
740 	int i;
741 
742 	/* won't configure msix register if no mapping is done
743 	 * between intr vector and event fd
744 	 */
745 	if (!rte_intr_dp_is_en(intr_handle))
746 		return;
747 
748 	if (rte_intr_allow_others(intr_handle)) {
749 		base = IGC_RX_VEC_START;
750 		vec = base;
751 		misc_shift = 1;
752 	}
753 
754 	/* turn on MSI-X capability first */
755 	IGC_WRITE_REG(hw, IGC_GPIE, IGC_GPIE_MSIX_MODE |
756 				IGC_GPIE_PBA | IGC_GPIE_EIAME |
757 				IGC_GPIE_NSICR);
758 	intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
759 		misc_shift;
760 
761 	if (dev->data->dev_conf.intr_conf.lsc)
762 		intr_mask |= (1u << IGC_MSIX_OTHER_INTR_VEC);
763 
764 	/* enable msix auto-clear */
765 	igc_read_reg_check_set_bits(hw, IGC_EIAC, intr_mask);
766 
767 	/* set other cause interrupt vector */
768 	igc_read_reg_check_set_bits(hw, IGC_IVAR_MISC,
769 		(uint32_t)(IGC_MSIX_OTHER_INTR_VEC | IGC_IVAR_VALID) << 8);
770 
771 	/* enable auto-mask */
772 	igc_read_reg_check_set_bits(hw, IGC_EIAM, intr_mask);
773 
774 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
775 		igc_write_ivar(hw, i, 0, vec);
776 		intr_handle->intr_vec[i] = vec;
777 		if (vec < base + intr_handle->nb_efd - 1)
778 			vec++;
779 	}
780 
781 	IGC_WRITE_FLUSH(hw);
782 }
783 
784 /**
785  * It enables the interrupt mask and then enable the interrupt.
786  *
787  * @dev
788  *  Pointer to struct rte_eth_dev.
789  * @on
790  *  Enable or Disable
791  */
792 static void
793 igc_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
794 {
795 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
796 
797 	if (on)
798 		intr->mask |= IGC_ICR_LSC;
799 	else
800 		intr->mask &= ~IGC_ICR_LSC;
801 }
802 
803 /*
804  * It enables the interrupt.
805  * It will be called once only during nic initialized.
806  */
807 static void
808 igc_rxq_interrupt_setup(struct rte_eth_dev *dev)
809 {
810 	uint32_t mask;
811 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
812 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
813 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
814 	int misc_shift = rte_intr_allow_others(intr_handle) ? 1 : 0;
815 
816 	/* won't configure msix register if no mapping is done
817 	 * between intr vector and event fd
818 	 */
819 	if (!rte_intr_dp_is_en(intr_handle))
820 		return;
821 
822 	mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) << misc_shift;
823 	IGC_WRITE_REG(hw, IGC_EIMS, mask);
824 }
825 
826 /*
827  *  Get hardware rx-buffer size.
828  */
829 static inline int
830 igc_get_rx_buffer_size(struct igc_hw *hw)
831 {
832 	return (IGC_READ_REG(hw, IGC_RXPBS) & 0x3f) << 10;
833 }
834 
835 /*
836  * igc_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
837  * For ASF and Pass Through versions of f/w this means
838  * that the driver is loaded.
839  */
840 static void
841 igc_hw_control_acquire(struct igc_hw *hw)
842 {
843 	uint32_t ctrl_ext;
844 
845 	/* Let firmware know the driver has taken over */
846 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
847 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext | IGC_CTRL_EXT_DRV_LOAD);
848 }
849 
850 /*
851  * igc_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
852  * For ASF and Pass Through versions of f/w this means that the
853  * driver is no longer loaded.
854  */
855 static void
856 igc_hw_control_release(struct igc_hw *hw)
857 {
858 	uint32_t ctrl_ext;
859 
860 	/* Let firmware taken over control of h/w */
861 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
862 	IGC_WRITE_REG(hw, IGC_CTRL_EXT,
863 			ctrl_ext & ~IGC_CTRL_EXT_DRV_LOAD);
864 }
865 
866 static int
867 igc_hardware_init(struct igc_hw *hw)
868 {
869 	uint32_t rx_buf_size;
870 	int diag;
871 
872 	/* Let the firmware know the OS is in control */
873 	igc_hw_control_acquire(hw);
874 
875 	/* Issue a global reset */
876 	igc_reset_hw(hw);
877 
878 	/* disable all wake up */
879 	IGC_WRITE_REG(hw, IGC_WUC, 0);
880 
881 	/*
882 	 * Hardware flow control
883 	 * - High water mark should allow for at least two standard size (1518)
884 	 *   frames to be received after sending an XOFF.
885 	 * - Low water mark works best when it is very near the high water mark.
886 	 *   This allows the receiver to restart by sending XON when it has
887 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
888 	 *   restart after one full frame is pulled from the buffer. There
889 	 *   could be several smaller frames in the buffer and if so they will
890 	 *   not trigger the XON until their total number reduces the buffer
891 	 *   by 1500.
892 	 */
893 	rx_buf_size = igc_get_rx_buffer_size(hw);
894 	hw->fc.high_water = rx_buf_size - (RTE_ETHER_MAX_LEN * 2);
895 	hw->fc.low_water = hw->fc.high_water - 1500;
896 	hw->fc.pause_time = IGC_FC_PAUSE_TIME;
897 	hw->fc.send_xon = 1;
898 	hw->fc.requested_mode = igc_fc_full;
899 
900 	diag = igc_init_hw(hw);
901 	if (diag < 0)
902 		return diag;
903 
904 	igc_get_phy_info(hw);
905 	igc_check_for_link(hw);
906 
907 	return 0;
908 }
909 
910 static int
911 eth_igc_start(struct rte_eth_dev *dev)
912 {
913 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
914 	struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
915 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
916 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
917 	uint32_t *speeds;
918 	int ret;
919 
920 	PMD_INIT_FUNC_TRACE();
921 
922 	/* disable all MSI-X interrupts */
923 	IGC_WRITE_REG(hw, IGC_EIMC, 0x1f);
924 	IGC_WRITE_FLUSH(hw);
925 
926 	/* clear all MSI-X interrupts */
927 	IGC_WRITE_REG(hw, IGC_EICR, 0x1f);
928 
929 	/* disable uio/vfio intr/eventfd mapping */
930 	if (!adapter->stopped)
931 		rte_intr_disable(intr_handle);
932 
933 	/* Power up the phy. Needed to make the link go Up */
934 	eth_igc_set_link_up(dev);
935 
936 	/* Put the address into the Receive Address Array */
937 	igc_rar_set(hw, hw->mac.addr, 0);
938 
939 	/* Initialize the hardware */
940 	if (igc_hardware_init(hw)) {
941 		PMD_DRV_LOG(ERR, "Unable to initialize the hardware");
942 		return -EIO;
943 	}
944 	adapter->stopped = 0;
945 
946 	/* check and configure queue intr-vector mapping */
947 	if (rte_intr_cap_multiple(intr_handle) &&
948 		dev->data->dev_conf.intr_conf.rxq) {
949 		uint32_t intr_vector = dev->data->nb_rx_queues;
950 		if (rte_intr_efd_enable(intr_handle, intr_vector))
951 			return -1;
952 	}
953 
954 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
955 		intr_handle->intr_vec = rte_zmalloc("intr_vec",
956 			dev->data->nb_rx_queues * sizeof(int), 0);
957 		if (intr_handle->intr_vec == NULL) {
958 			PMD_DRV_LOG(ERR,
959 				"Failed to allocate %d rx_queues intr_vec",
960 				dev->data->nb_rx_queues);
961 			return -ENOMEM;
962 		}
963 	}
964 
965 	/* configure msix for rx interrupt */
966 	igc_configure_msix_intr(dev);
967 
968 	igc_tx_init(dev);
969 
970 	/* This can fail when allocating mbufs for descriptor rings */
971 	ret = igc_rx_init(dev);
972 	if (ret) {
973 		PMD_DRV_LOG(ERR, "Unable to initialize RX hardware");
974 		igc_dev_clear_queues(dev);
975 		return ret;
976 	}
977 
978 	igc_clear_hw_cntrs_base_generic(hw);
979 
980 	/* VLAN Offload Settings */
981 	eth_igc_vlan_offload_set(dev,
982 		ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
983 		ETH_VLAN_EXTEND_MASK);
984 
985 	/* Setup link speed and duplex */
986 	speeds = &dev->data->dev_conf.link_speeds;
987 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
988 		hw->phy.autoneg_advertised = IGC_ALL_SPEED_DUPLEX_2500;
989 		hw->mac.autoneg = 1;
990 	} else {
991 		int num_speeds = 0;
992 
993 		if (*speeds & ETH_LINK_SPEED_FIXED) {
994 			PMD_DRV_LOG(ERR,
995 				    "Force speed mode currently not supported");
996 			igc_dev_clear_queues(dev);
997 			return -EINVAL;
998 		}
999 
1000 		hw->phy.autoneg_advertised = 0;
1001 		hw->mac.autoneg = 1;
1002 
1003 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1004 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1005 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_2_5G)) {
1006 			num_speeds = -1;
1007 			goto error_invalid_config;
1008 		}
1009 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
1010 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1011 			num_speeds++;
1012 		}
1013 		if (*speeds & ETH_LINK_SPEED_10M) {
1014 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1015 			num_speeds++;
1016 		}
1017 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
1018 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1019 			num_speeds++;
1020 		}
1021 		if (*speeds & ETH_LINK_SPEED_100M) {
1022 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1023 			num_speeds++;
1024 		}
1025 		if (*speeds & ETH_LINK_SPEED_1G) {
1026 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1027 			num_speeds++;
1028 		}
1029 		if (*speeds & ETH_LINK_SPEED_2_5G) {
1030 			hw->phy.autoneg_advertised |= ADVERTISE_2500_FULL;
1031 			num_speeds++;
1032 		}
1033 		if (num_speeds == 0)
1034 			goto error_invalid_config;
1035 	}
1036 
1037 	igc_setup_link(hw);
1038 
1039 	if (rte_intr_allow_others(intr_handle)) {
1040 		/* check if lsc interrupt is enabled */
1041 		if (dev->data->dev_conf.intr_conf.lsc)
1042 			igc_lsc_interrupt_setup(dev, 1);
1043 		else
1044 			igc_lsc_interrupt_setup(dev, 0);
1045 	} else {
1046 		rte_intr_callback_unregister(intr_handle,
1047 					     eth_igc_interrupt_handler,
1048 					     (void *)dev);
1049 		if (dev->data->dev_conf.intr_conf.lsc)
1050 			PMD_DRV_LOG(INFO,
1051 				"LSC won't enable because of no intr multiplex");
1052 	}
1053 
1054 	/* enable uio/vfio intr/eventfd mapping */
1055 	rte_intr_enable(intr_handle);
1056 
1057 	rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1058 			igc_update_queue_stats_handler, dev);
1059 
1060 	/* check if rxq interrupt is enabled */
1061 	if (dev->data->dev_conf.intr_conf.rxq &&
1062 			rte_intr_dp_is_en(intr_handle))
1063 		igc_rxq_interrupt_setup(dev);
1064 
1065 	/* resume enabled intr since hw reset */
1066 	igc_intr_other_enable(dev);
1067 
1068 	eth_igc_rxtx_control(dev, true);
1069 	eth_igc_link_update(dev, 0);
1070 
1071 	/* configure MAC-loopback mode */
1072 	if (dev->data->dev_conf.lpbk_mode == 1) {
1073 		uint32_t reg_val;
1074 
1075 		reg_val = IGC_READ_REG(hw, IGC_CTRL);
1076 		reg_val &= ~IGC_CTRL_SPEED_MASK;
1077 		reg_val |= IGC_CTRL_SLU | IGC_CTRL_FRCSPD |
1078 			IGC_CTRL_FRCDPX | IGC_CTRL_FD | IGC_CTRL_SPEED_2500;
1079 		IGC_WRITE_REG(hw, IGC_CTRL, reg_val);
1080 
1081 		igc_read_reg_check_set_bits(hw, IGC_EEER, IGC_EEER_EEE_FRC_AN);
1082 	}
1083 
1084 	return 0;
1085 
1086 error_invalid_config:
1087 	PMD_DRV_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
1088 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
1089 	igc_dev_clear_queues(dev);
1090 	return -EINVAL;
1091 }
1092 
1093 static int
1094 igc_reset_swfw_lock(struct igc_hw *hw)
1095 {
1096 	int ret_val;
1097 
1098 	/*
1099 	 * Do mac ops initialization manually here, since we will need
1100 	 * some function pointers set by this call.
1101 	 */
1102 	ret_val = igc_init_mac_params(hw);
1103 	if (ret_val)
1104 		return ret_val;
1105 
1106 	/*
1107 	 * SMBI lock should not fail in this early stage. If this is the case,
1108 	 * it is due to an improper exit of the application.
1109 	 * So force the release of the faulty lock.
1110 	 */
1111 	if (igc_get_hw_semaphore_generic(hw) < 0)
1112 		PMD_DRV_LOG(DEBUG, "SMBI lock released");
1113 
1114 	igc_put_hw_semaphore_generic(hw);
1115 
1116 	if (hw->mac.ops.acquire_swfw_sync != NULL) {
1117 		uint16_t mask;
1118 
1119 		/*
1120 		 * Phy lock should not fail in this early stage.
1121 		 * If this is the case, it is due to an improper exit of the
1122 		 * application. So force the release of the faulty lock.
1123 		 */
1124 		mask = IGC_SWFW_PHY0_SM;
1125 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
1126 			PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released",
1127 				    hw->bus.func);
1128 		}
1129 		hw->mac.ops.release_swfw_sync(hw, mask);
1130 
1131 		/*
1132 		 * This one is more tricky since it is common to all ports; but
1133 		 * swfw_sync retries last long enough (1s) to be almost sure
1134 		 * that if lock can not be taken it is due to an improper lock
1135 		 * of the semaphore.
1136 		 */
1137 		mask = IGC_SWFW_EEP_SM;
1138 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0)
1139 			PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1140 
1141 		hw->mac.ops.release_swfw_sync(hw, mask);
1142 	}
1143 
1144 	return IGC_SUCCESS;
1145 }
1146 
1147 /*
1148  * free all rx/tx queues.
1149  */
1150 static void
1151 igc_dev_free_queues(struct rte_eth_dev *dev)
1152 {
1153 	uint16_t i;
1154 
1155 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
1156 		eth_igc_rx_queue_release(dev, i);
1157 		dev->data->rx_queues[i] = NULL;
1158 	}
1159 	dev->data->nb_rx_queues = 0;
1160 
1161 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
1162 		eth_igc_tx_queue_release(dev, i);
1163 		dev->data->tx_queues[i] = NULL;
1164 	}
1165 	dev->data->nb_tx_queues = 0;
1166 }
1167 
1168 static int
1169 eth_igc_close(struct rte_eth_dev *dev)
1170 {
1171 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1172 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1173 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1174 	struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
1175 	int retry = 0;
1176 	int ret = 0;
1177 
1178 	PMD_INIT_FUNC_TRACE();
1179 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1180 		return 0;
1181 
1182 	if (!adapter->stopped)
1183 		ret = eth_igc_stop(dev);
1184 
1185 	igc_flow_flush(dev, NULL);
1186 	igc_clear_all_filter(dev);
1187 
1188 	igc_intr_other_disable(dev);
1189 	do {
1190 		int ret = rte_intr_callback_unregister(intr_handle,
1191 				eth_igc_interrupt_handler, dev);
1192 		if (ret >= 0 || ret == -ENOENT || ret == -EINVAL)
1193 			break;
1194 
1195 		PMD_DRV_LOG(ERR, "intr callback unregister failed: %d", ret);
1196 		DELAY(200 * 1000); /* delay 200ms */
1197 	} while (retry++ < 5);
1198 
1199 	igc_phy_hw_reset(hw);
1200 	igc_hw_control_release(hw);
1201 	igc_dev_free_queues(dev);
1202 
1203 	/* Reset any pending lock */
1204 	igc_reset_swfw_lock(hw);
1205 
1206 	return ret;
1207 }
1208 
1209 static void
1210 igc_identify_hardware(struct rte_eth_dev *dev, struct rte_pci_device *pci_dev)
1211 {
1212 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1213 
1214 	hw->vendor_id = pci_dev->id.vendor_id;
1215 	hw->device_id = pci_dev->id.device_id;
1216 	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
1217 	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
1218 }
1219 
1220 static int
1221 eth_igc_dev_init(struct rte_eth_dev *dev)
1222 {
1223 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1224 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
1225 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1226 	int i, error = 0;
1227 
1228 	PMD_INIT_FUNC_TRACE();
1229 	dev->dev_ops = &eth_igc_ops;
1230 	dev->rx_queue_count = eth_igc_rx_queue_count;
1231 	dev->rx_descriptor_status = eth_igc_rx_descriptor_status;
1232 	dev->tx_descriptor_status = eth_igc_tx_descriptor_status;
1233 
1234 	/*
1235 	 * for secondary processes, we don't initialize any further as primary
1236 	 * has already done this work. Only check we don't need a different
1237 	 * RX function.
1238 	 */
1239 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1240 		return 0;
1241 
1242 	rte_eth_copy_pci_info(dev, pci_dev);
1243 	dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1244 
1245 	hw->back = pci_dev;
1246 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1247 
1248 	igc_identify_hardware(dev, pci_dev);
1249 	if (igc_setup_init_funcs(hw, false) != IGC_SUCCESS) {
1250 		error = -EIO;
1251 		goto err_late;
1252 	}
1253 
1254 	igc_get_bus_info(hw);
1255 
1256 	/* Reset any pending lock */
1257 	if (igc_reset_swfw_lock(hw) != IGC_SUCCESS) {
1258 		error = -EIO;
1259 		goto err_late;
1260 	}
1261 
1262 	/* Finish initialization */
1263 	if (igc_setup_init_funcs(hw, true) != IGC_SUCCESS) {
1264 		error = -EIO;
1265 		goto err_late;
1266 	}
1267 
1268 	hw->mac.autoneg = 1;
1269 	hw->phy.autoneg_wait_to_complete = 0;
1270 	hw->phy.autoneg_advertised = IGC_ALL_SPEED_DUPLEX_2500;
1271 
1272 	/* Copper options */
1273 	if (hw->phy.media_type == igc_media_type_copper) {
1274 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
1275 		hw->phy.disable_polarity_correction = 0;
1276 		hw->phy.ms_type = igc_ms_hw_default;
1277 	}
1278 
1279 	/*
1280 	 * Start from a known state, this is important in reading the nvm
1281 	 * and mac from that.
1282 	 */
1283 	igc_reset_hw(hw);
1284 
1285 	/* Make sure we have a good EEPROM before we read from it */
1286 	if (igc_validate_nvm_checksum(hw) < 0) {
1287 		/*
1288 		 * Some PCI-E parts fail the first check due to
1289 		 * the link being in sleep state, call it again,
1290 		 * if it fails a second time its a real issue.
1291 		 */
1292 		if (igc_validate_nvm_checksum(hw) < 0) {
1293 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
1294 			error = -EIO;
1295 			goto err_late;
1296 		}
1297 	}
1298 
1299 	/* Read the permanent MAC address out of the EEPROM */
1300 	if (igc_read_mac_addr(hw) != 0) {
1301 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
1302 		error = -EIO;
1303 		goto err_late;
1304 	}
1305 
1306 	/* Allocate memory for storing MAC addresses */
1307 	dev->data->mac_addrs = rte_zmalloc("igc",
1308 		RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count, 0);
1309 	if (dev->data->mac_addrs == NULL) {
1310 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes for storing MAC",
1311 				RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
1312 		error = -ENOMEM;
1313 		goto err_late;
1314 	}
1315 
1316 	/* Copy the permanent MAC address */
1317 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
1318 			&dev->data->mac_addrs[0]);
1319 
1320 	/* Now initialize the hardware */
1321 	if (igc_hardware_init(hw) != 0) {
1322 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
1323 		rte_free(dev->data->mac_addrs);
1324 		dev->data->mac_addrs = NULL;
1325 		error = -ENODEV;
1326 		goto err_late;
1327 	}
1328 
1329 	hw->mac.get_link_status = 1;
1330 	igc->stopped = 0;
1331 
1332 	/* Indicate SOL/IDER usage */
1333 	if (igc_check_reset_block(hw) < 0)
1334 		PMD_INIT_LOG(ERR,
1335 			"PHY reset is blocked due to SOL/IDER session.");
1336 
1337 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
1338 			dev->data->port_id, pci_dev->id.vendor_id,
1339 			pci_dev->id.device_id);
1340 
1341 	rte_intr_callback_register(&pci_dev->intr_handle,
1342 			eth_igc_interrupt_handler, (void *)dev);
1343 
1344 	/* enable uio/vfio intr/eventfd mapping */
1345 	rte_intr_enable(&pci_dev->intr_handle);
1346 
1347 	/* enable support intr */
1348 	igc_intr_other_enable(dev);
1349 
1350 	/* initiate queue status */
1351 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1352 		igc->txq_stats_map[i] = -1;
1353 		igc->rxq_stats_map[i] = -1;
1354 	}
1355 
1356 	igc_flow_init(dev);
1357 	igc_clear_all_filter(dev);
1358 	return 0;
1359 
1360 err_late:
1361 	igc_hw_control_release(hw);
1362 	return error;
1363 }
1364 
1365 static int
1366 eth_igc_dev_uninit(__rte_unused struct rte_eth_dev *eth_dev)
1367 {
1368 	PMD_INIT_FUNC_TRACE();
1369 	eth_igc_close(eth_dev);
1370 	return 0;
1371 }
1372 
1373 static int
1374 eth_igc_reset(struct rte_eth_dev *dev)
1375 {
1376 	int ret;
1377 
1378 	PMD_INIT_FUNC_TRACE();
1379 
1380 	ret = eth_igc_dev_uninit(dev);
1381 	if (ret)
1382 		return ret;
1383 
1384 	return eth_igc_dev_init(dev);
1385 }
1386 
1387 static int
1388 eth_igc_promiscuous_enable(struct rte_eth_dev *dev)
1389 {
1390 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1391 	uint32_t rctl;
1392 
1393 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1394 	rctl |= (IGC_RCTL_UPE | IGC_RCTL_MPE);
1395 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1396 	return 0;
1397 }
1398 
1399 static int
1400 eth_igc_promiscuous_disable(struct rte_eth_dev *dev)
1401 {
1402 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1403 	uint32_t rctl;
1404 
1405 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1406 	rctl &= (~IGC_RCTL_UPE);
1407 	if (dev->data->all_multicast == 1)
1408 		rctl |= IGC_RCTL_MPE;
1409 	else
1410 		rctl &= (~IGC_RCTL_MPE);
1411 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1412 	return 0;
1413 }
1414 
1415 static int
1416 eth_igc_allmulticast_enable(struct rte_eth_dev *dev)
1417 {
1418 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1419 	uint32_t rctl;
1420 
1421 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1422 	rctl |= IGC_RCTL_MPE;
1423 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1424 	return 0;
1425 }
1426 
1427 static int
1428 eth_igc_allmulticast_disable(struct rte_eth_dev *dev)
1429 {
1430 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1431 	uint32_t rctl;
1432 
1433 	if (dev->data->promiscuous == 1)
1434 		return 0;	/* must remain in all_multicast mode */
1435 
1436 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1437 	rctl &= (~IGC_RCTL_MPE);
1438 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1439 	return 0;
1440 }
1441 
1442 static int
1443 eth_igc_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
1444 		       size_t fw_size)
1445 {
1446 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1447 	struct igc_fw_version fw;
1448 	int ret;
1449 
1450 	igc_get_fw_version(hw, &fw);
1451 
1452 	/* if option rom is valid, display its version too */
1453 	if (fw.or_valid) {
1454 		ret = snprintf(fw_version, fw_size,
1455 			 "%d.%d, 0x%08x, %d.%d.%d",
1456 			 fw.eep_major, fw.eep_minor, fw.etrack_id,
1457 			 fw.or_major, fw.or_build, fw.or_patch);
1458 	/* no option rom */
1459 	} else {
1460 		if (fw.etrack_id != 0X0000) {
1461 			ret = snprintf(fw_version, fw_size,
1462 				 "%d.%d, 0x%08x",
1463 				 fw.eep_major, fw.eep_minor,
1464 				 fw.etrack_id);
1465 		} else {
1466 			ret = snprintf(fw_version, fw_size,
1467 				 "%d.%d.%d",
1468 				 fw.eep_major, fw.eep_minor,
1469 				 fw.eep_build);
1470 		}
1471 	}
1472 	if (ret < 0)
1473 		return -EINVAL;
1474 
1475 	ret += 1; /* add the size of '\0' */
1476 	if (fw_size < (size_t)ret)
1477 		return ret;
1478 	else
1479 		return 0;
1480 }
1481 
1482 static int
1483 eth_igc_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1484 {
1485 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1486 
1487 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1488 	dev_info->max_rx_pktlen = MAX_RX_JUMBO_FRAME_SIZE;
1489 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1490 	dev_info->rx_offload_capa = IGC_RX_OFFLOAD_ALL;
1491 	dev_info->tx_offload_capa = IGC_TX_OFFLOAD_ALL;
1492 	dev_info->rx_queue_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP;
1493 
1494 	dev_info->max_rx_queues = IGC_QUEUE_PAIRS_NUM;
1495 	dev_info->max_tx_queues = IGC_QUEUE_PAIRS_NUM;
1496 	dev_info->max_vmdq_pools = 0;
1497 
1498 	dev_info->hash_key_size = IGC_HKEY_MAX_INDEX * sizeof(uint32_t);
1499 	dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
1500 	dev_info->flow_type_rss_offloads = IGC_RSS_OFFLOAD_ALL;
1501 
1502 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
1503 		.rx_thresh = {
1504 			.pthresh = IGC_DEFAULT_RX_PTHRESH,
1505 			.hthresh = IGC_DEFAULT_RX_HTHRESH,
1506 			.wthresh = IGC_DEFAULT_RX_WTHRESH,
1507 		},
1508 		.rx_free_thresh = IGC_DEFAULT_RX_FREE_THRESH,
1509 		.rx_drop_en = 0,
1510 		.offloads = 0,
1511 	};
1512 
1513 	dev_info->default_txconf = (struct rte_eth_txconf) {
1514 		.tx_thresh = {
1515 			.pthresh = IGC_DEFAULT_TX_PTHRESH,
1516 			.hthresh = IGC_DEFAULT_TX_HTHRESH,
1517 			.wthresh = IGC_DEFAULT_TX_WTHRESH,
1518 		},
1519 		.offloads = 0,
1520 	};
1521 
1522 	dev_info->rx_desc_lim = rx_desc_lim;
1523 	dev_info->tx_desc_lim = tx_desc_lim;
1524 
1525 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1526 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1527 			ETH_LINK_SPEED_1G | ETH_LINK_SPEED_2_5G;
1528 
1529 	dev_info->max_mtu = dev_info->max_rx_pktlen - IGC_ETH_OVERHEAD;
1530 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
1531 	return 0;
1532 }
1533 
1534 static int
1535 eth_igc_led_on(struct rte_eth_dev *dev)
1536 {
1537 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1538 
1539 	return igc_led_on(hw) == IGC_SUCCESS ? 0 : -ENOTSUP;
1540 }
1541 
1542 static int
1543 eth_igc_led_off(struct rte_eth_dev *dev)
1544 {
1545 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1546 
1547 	return igc_led_off(hw) == IGC_SUCCESS ? 0 : -ENOTSUP;
1548 }
1549 
1550 static const uint32_t *
1551 eth_igc_supported_ptypes_get(__rte_unused struct rte_eth_dev *dev)
1552 {
1553 	static const uint32_t ptypes[] = {
1554 		/* refers to rx_desc_pkt_info_to_pkt_type() */
1555 		RTE_PTYPE_L2_ETHER,
1556 		RTE_PTYPE_L3_IPV4,
1557 		RTE_PTYPE_L3_IPV4_EXT,
1558 		RTE_PTYPE_L3_IPV6,
1559 		RTE_PTYPE_L3_IPV6_EXT,
1560 		RTE_PTYPE_L4_TCP,
1561 		RTE_PTYPE_L4_UDP,
1562 		RTE_PTYPE_L4_SCTP,
1563 		RTE_PTYPE_TUNNEL_IP,
1564 		RTE_PTYPE_INNER_L3_IPV6,
1565 		RTE_PTYPE_INNER_L3_IPV6_EXT,
1566 		RTE_PTYPE_INNER_L4_TCP,
1567 		RTE_PTYPE_INNER_L4_UDP,
1568 		RTE_PTYPE_UNKNOWN
1569 	};
1570 
1571 	return ptypes;
1572 }
1573 
1574 static int
1575 eth_igc_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1576 {
1577 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1578 	uint32_t frame_size = mtu + IGC_ETH_OVERHEAD;
1579 	uint32_t rctl;
1580 
1581 	/* if extend vlan has been enabled */
1582 	if (IGC_READ_REG(hw, IGC_CTRL_EXT) & IGC_CTRL_EXT_EXT_VLAN)
1583 		frame_size += VLAN_TAG_SIZE;
1584 
1585 	/* check that mtu is within the allowed range */
1586 	if (mtu < RTE_ETHER_MIN_MTU ||
1587 		frame_size > MAX_RX_JUMBO_FRAME_SIZE)
1588 		return -EINVAL;
1589 
1590 	/*
1591 	 * If device is started, refuse mtu that requires the support of
1592 	 * scattered packets when this feature has not been enabled before.
1593 	 */
1594 	if (dev->data->dev_started && !dev->data->scattered_rx &&
1595 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
1596 		PMD_INIT_LOG(ERR, "Stop port first.");
1597 		return -EINVAL;
1598 	}
1599 
1600 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1601 
1602 	/* switch to jumbo mode if needed */
1603 	if (mtu > RTE_ETHER_MTU) {
1604 		dev->data->dev_conf.rxmode.offloads |=
1605 			DEV_RX_OFFLOAD_JUMBO_FRAME;
1606 		rctl |= IGC_RCTL_LPE;
1607 	} else {
1608 		dev->data->dev_conf.rxmode.offloads &=
1609 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
1610 		rctl &= ~IGC_RCTL_LPE;
1611 	}
1612 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1613 
1614 	/* update max frame size */
1615 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1616 
1617 	IGC_WRITE_REG(hw, IGC_RLPML,
1618 			dev->data->dev_conf.rxmode.max_rx_pkt_len);
1619 
1620 	return 0;
1621 }
1622 
1623 static int
1624 eth_igc_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
1625 		uint32_t index, uint32_t pool)
1626 {
1627 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1628 
1629 	igc_rar_set(hw, mac_addr->addr_bytes, index);
1630 	RTE_SET_USED(pool);
1631 	return 0;
1632 }
1633 
1634 static void
1635 eth_igc_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1636 {
1637 	uint8_t addr[RTE_ETHER_ADDR_LEN];
1638 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1639 
1640 	memset(addr, 0, sizeof(addr));
1641 	igc_rar_set(hw, addr, index);
1642 }
1643 
1644 static int
1645 eth_igc_default_mac_addr_set(struct rte_eth_dev *dev,
1646 			struct rte_ether_addr *addr)
1647 {
1648 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1649 	igc_rar_set(hw, addr->addr_bytes, 0);
1650 	return 0;
1651 }
1652 
1653 static int
1654 eth_igc_set_mc_addr_list(struct rte_eth_dev *dev,
1655 			 struct rte_ether_addr *mc_addr_set,
1656 			 uint32_t nb_mc_addr)
1657 {
1658 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1659 	igc_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1660 	return 0;
1661 }
1662 
1663 /*
1664  * Read hardware registers
1665  */
1666 static void
1667 igc_read_stats_registers(struct igc_hw *hw, struct igc_hw_stats *stats)
1668 {
1669 	int pause_frames;
1670 
1671 	uint64_t old_gprc  = stats->gprc;
1672 	uint64_t old_gptc  = stats->gptc;
1673 	uint64_t old_tpr   = stats->tpr;
1674 	uint64_t old_tpt   = stats->tpt;
1675 	uint64_t old_rpthc = stats->rpthc;
1676 	uint64_t old_hgptc = stats->hgptc;
1677 
1678 	stats->crcerrs += IGC_READ_REG(hw, IGC_CRCERRS);
1679 	stats->algnerrc += IGC_READ_REG(hw, IGC_ALGNERRC);
1680 	stats->rxerrc += IGC_READ_REG(hw, IGC_RXERRC);
1681 	stats->mpc += IGC_READ_REG(hw, IGC_MPC);
1682 	stats->scc += IGC_READ_REG(hw, IGC_SCC);
1683 	stats->ecol += IGC_READ_REG(hw, IGC_ECOL);
1684 
1685 	stats->mcc += IGC_READ_REG(hw, IGC_MCC);
1686 	stats->latecol += IGC_READ_REG(hw, IGC_LATECOL);
1687 	stats->colc += IGC_READ_REG(hw, IGC_COLC);
1688 
1689 	stats->dc += IGC_READ_REG(hw, IGC_DC);
1690 	stats->tncrs += IGC_READ_REG(hw, IGC_TNCRS);
1691 	stats->htdpmc += IGC_READ_REG(hw, IGC_HTDPMC);
1692 	stats->rlec += IGC_READ_REG(hw, IGC_RLEC);
1693 	stats->xonrxc += IGC_READ_REG(hw, IGC_XONRXC);
1694 	stats->xontxc += IGC_READ_REG(hw, IGC_XONTXC);
1695 
1696 	/*
1697 	 * For watchdog management we need to know if we have been
1698 	 * paused during the last interval, so capture that here.
1699 	 */
1700 	pause_frames = IGC_READ_REG(hw, IGC_XOFFRXC);
1701 	stats->xoffrxc += pause_frames;
1702 	stats->xofftxc += IGC_READ_REG(hw, IGC_XOFFTXC);
1703 	stats->fcruc += IGC_READ_REG(hw, IGC_FCRUC);
1704 	stats->prc64 += IGC_READ_REG(hw, IGC_PRC64);
1705 	stats->prc127 += IGC_READ_REG(hw, IGC_PRC127);
1706 	stats->prc255 += IGC_READ_REG(hw, IGC_PRC255);
1707 	stats->prc511 += IGC_READ_REG(hw, IGC_PRC511);
1708 	stats->prc1023 += IGC_READ_REG(hw, IGC_PRC1023);
1709 	stats->prc1522 += IGC_READ_REG(hw, IGC_PRC1522);
1710 	stats->gprc += IGC_READ_REG(hw, IGC_GPRC);
1711 	stats->bprc += IGC_READ_REG(hw, IGC_BPRC);
1712 	stats->mprc += IGC_READ_REG(hw, IGC_MPRC);
1713 	stats->gptc += IGC_READ_REG(hw, IGC_GPTC);
1714 
1715 	/* For the 64-bit byte counters the low dword must be read first. */
1716 	/* Both registers clear on the read of the high dword */
1717 
1718 	/* Workaround CRC bytes included in size, take away 4 bytes/packet */
1719 	stats->gorc += IGC_READ_REG(hw, IGC_GORCL);
1720 	stats->gorc += ((uint64_t)IGC_READ_REG(hw, IGC_GORCH) << 32);
1721 	stats->gorc -= (stats->gprc - old_gprc) * RTE_ETHER_CRC_LEN;
1722 	stats->gotc += IGC_READ_REG(hw, IGC_GOTCL);
1723 	stats->gotc += ((uint64_t)IGC_READ_REG(hw, IGC_GOTCH) << 32);
1724 	stats->gotc -= (stats->gptc - old_gptc) * RTE_ETHER_CRC_LEN;
1725 
1726 	stats->rnbc += IGC_READ_REG(hw, IGC_RNBC);
1727 	stats->ruc += IGC_READ_REG(hw, IGC_RUC);
1728 	stats->rfc += IGC_READ_REG(hw, IGC_RFC);
1729 	stats->roc += IGC_READ_REG(hw, IGC_ROC);
1730 	stats->rjc += IGC_READ_REG(hw, IGC_RJC);
1731 
1732 	stats->mgprc += IGC_READ_REG(hw, IGC_MGTPRC);
1733 	stats->mgpdc += IGC_READ_REG(hw, IGC_MGTPDC);
1734 	stats->mgptc += IGC_READ_REG(hw, IGC_MGTPTC);
1735 	stats->b2ospc += IGC_READ_REG(hw, IGC_B2OSPC);
1736 	stats->b2ogprc += IGC_READ_REG(hw, IGC_B2OGPRC);
1737 	stats->o2bgptc += IGC_READ_REG(hw, IGC_O2BGPTC);
1738 	stats->o2bspc += IGC_READ_REG(hw, IGC_O2BSPC);
1739 
1740 	stats->tpr += IGC_READ_REG(hw, IGC_TPR);
1741 	stats->tpt += IGC_READ_REG(hw, IGC_TPT);
1742 
1743 	stats->tor += IGC_READ_REG(hw, IGC_TORL);
1744 	stats->tor += ((uint64_t)IGC_READ_REG(hw, IGC_TORH) << 32);
1745 	stats->tor -= (stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
1746 	stats->tot += IGC_READ_REG(hw, IGC_TOTL);
1747 	stats->tot += ((uint64_t)IGC_READ_REG(hw, IGC_TOTH) << 32);
1748 	stats->tot -= (stats->tpt - old_tpt) * RTE_ETHER_CRC_LEN;
1749 
1750 	stats->ptc64 += IGC_READ_REG(hw, IGC_PTC64);
1751 	stats->ptc127 += IGC_READ_REG(hw, IGC_PTC127);
1752 	stats->ptc255 += IGC_READ_REG(hw, IGC_PTC255);
1753 	stats->ptc511 += IGC_READ_REG(hw, IGC_PTC511);
1754 	stats->ptc1023 += IGC_READ_REG(hw, IGC_PTC1023);
1755 	stats->ptc1522 += IGC_READ_REG(hw, IGC_PTC1522);
1756 	stats->mptc += IGC_READ_REG(hw, IGC_MPTC);
1757 	stats->bptc += IGC_READ_REG(hw, IGC_BPTC);
1758 	stats->tsctc += IGC_READ_REG(hw, IGC_TSCTC);
1759 
1760 	stats->iac += IGC_READ_REG(hw, IGC_IAC);
1761 	stats->rpthc += IGC_READ_REG(hw, IGC_RPTHC);
1762 	stats->hgptc += IGC_READ_REG(hw, IGC_HGPTC);
1763 	stats->icrxdmtc += IGC_READ_REG(hw, IGC_ICRXDMTC);
1764 
1765 	/* Host to Card Statistics */
1766 	stats->hgorc += IGC_READ_REG(hw, IGC_HGORCL);
1767 	stats->hgorc += ((uint64_t)IGC_READ_REG(hw, IGC_HGORCH) << 32);
1768 	stats->hgorc -= (stats->rpthc - old_rpthc) * RTE_ETHER_CRC_LEN;
1769 	stats->hgotc += IGC_READ_REG(hw, IGC_HGOTCL);
1770 	stats->hgotc += ((uint64_t)IGC_READ_REG(hw, IGC_HGOTCH) << 32);
1771 	stats->hgotc -= (stats->hgptc - old_hgptc) * RTE_ETHER_CRC_LEN;
1772 	stats->lenerrs += IGC_READ_REG(hw, IGC_LENERRS);
1773 }
1774 
1775 /*
1776  * Write 0 to all queue status registers
1777  */
1778 static void
1779 igc_reset_queue_stats_register(struct igc_hw *hw)
1780 {
1781 	int i;
1782 
1783 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1784 		IGC_WRITE_REG(hw, IGC_PQGPRC(i), 0);
1785 		IGC_WRITE_REG(hw, IGC_PQGPTC(i), 0);
1786 		IGC_WRITE_REG(hw, IGC_PQGORC(i), 0);
1787 		IGC_WRITE_REG(hw, IGC_PQGOTC(i), 0);
1788 		IGC_WRITE_REG(hw, IGC_PQMPRC(i), 0);
1789 		IGC_WRITE_REG(hw, IGC_RQDPC(i), 0);
1790 		IGC_WRITE_REG(hw, IGC_TQDPC(i), 0);
1791 	}
1792 }
1793 
1794 /*
1795  * Read all hardware queue status registers
1796  */
1797 static void
1798 igc_read_queue_stats_register(struct rte_eth_dev *dev)
1799 {
1800 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1801 	struct igc_hw_queue_stats *queue_stats =
1802 				IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1803 	int i;
1804 
1805 	/*
1806 	 * This register is not cleared on read. Furthermore, the register wraps
1807 	 * around back to 0x00000000 on the next increment when reaching a value
1808 	 * of 0xFFFFFFFF and then continues normal count operation.
1809 	 */
1810 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1811 		union {
1812 			u64 ddword;
1813 			u32 dword[2];
1814 		} value;
1815 		u32 tmp;
1816 
1817 		/*
1818 		 * Read the register first, if the value is smaller than that
1819 		 * previous read, that mean the register has been overflowed,
1820 		 * then we add the high 4 bytes by 1 and replace the low 4
1821 		 * bytes by the new value.
1822 		 */
1823 		tmp = IGC_READ_REG(hw, IGC_PQGPRC(i));
1824 		value.ddword = queue_stats->pqgprc[i];
1825 		if (value.dword[U32_0_IN_U64] > tmp)
1826 			value.dword[U32_1_IN_U64]++;
1827 		value.dword[U32_0_IN_U64] = tmp;
1828 		queue_stats->pqgprc[i] = value.ddword;
1829 
1830 		tmp = IGC_READ_REG(hw, IGC_PQGPTC(i));
1831 		value.ddword = queue_stats->pqgptc[i];
1832 		if (value.dword[U32_0_IN_U64] > tmp)
1833 			value.dword[U32_1_IN_U64]++;
1834 		value.dword[U32_0_IN_U64] = tmp;
1835 		queue_stats->pqgptc[i] = value.ddword;
1836 
1837 		tmp = IGC_READ_REG(hw, IGC_PQGORC(i));
1838 		value.ddword = queue_stats->pqgorc[i];
1839 		if (value.dword[U32_0_IN_U64] > tmp)
1840 			value.dword[U32_1_IN_U64]++;
1841 		value.dword[U32_0_IN_U64] = tmp;
1842 		queue_stats->pqgorc[i] = value.ddword;
1843 
1844 		tmp = IGC_READ_REG(hw, IGC_PQGOTC(i));
1845 		value.ddword = queue_stats->pqgotc[i];
1846 		if (value.dword[U32_0_IN_U64] > tmp)
1847 			value.dword[U32_1_IN_U64]++;
1848 		value.dword[U32_0_IN_U64] = tmp;
1849 		queue_stats->pqgotc[i] = value.ddword;
1850 
1851 		tmp = IGC_READ_REG(hw, IGC_PQMPRC(i));
1852 		value.ddword = queue_stats->pqmprc[i];
1853 		if (value.dword[U32_0_IN_U64] > tmp)
1854 			value.dword[U32_1_IN_U64]++;
1855 		value.dword[U32_0_IN_U64] = tmp;
1856 		queue_stats->pqmprc[i] = value.ddword;
1857 
1858 		tmp = IGC_READ_REG(hw, IGC_RQDPC(i));
1859 		value.ddword = queue_stats->rqdpc[i];
1860 		if (value.dword[U32_0_IN_U64] > tmp)
1861 			value.dword[U32_1_IN_U64]++;
1862 		value.dword[U32_0_IN_U64] = tmp;
1863 		queue_stats->rqdpc[i] = value.ddword;
1864 
1865 		tmp = IGC_READ_REG(hw, IGC_TQDPC(i));
1866 		value.ddword = queue_stats->tqdpc[i];
1867 		if (value.dword[U32_0_IN_U64] > tmp)
1868 			value.dword[U32_1_IN_U64]++;
1869 		value.dword[U32_0_IN_U64] = tmp;
1870 		queue_stats->tqdpc[i] = value.ddword;
1871 	}
1872 }
1873 
1874 static int
1875 eth_igc_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1876 {
1877 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
1878 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1879 	struct igc_hw_stats *stats = IGC_DEV_PRIVATE_STATS(dev);
1880 	struct igc_hw_queue_stats *queue_stats =
1881 			IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1882 	int i;
1883 
1884 	/*
1885 	 * Cancel status handler since it will read the queue status registers
1886 	 */
1887 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
1888 
1889 	/* Read status register */
1890 	igc_read_queue_stats_register(dev);
1891 	igc_read_stats_registers(hw, stats);
1892 
1893 	if (rte_stats == NULL) {
1894 		/* Restart queue status handler */
1895 		rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1896 				igc_update_queue_stats_handler, dev);
1897 		return -EINVAL;
1898 	}
1899 
1900 	/* Rx Errors */
1901 	rte_stats->imissed = stats->mpc;
1902 	rte_stats->ierrors = stats->crcerrs + stats->rlec +
1903 			stats->rxerrc + stats->algnerrc;
1904 
1905 	/* Tx Errors */
1906 	rte_stats->oerrors = stats->ecol + stats->latecol;
1907 
1908 	rte_stats->ipackets = stats->gprc;
1909 	rte_stats->opackets = stats->gptc;
1910 	rte_stats->ibytes   = stats->gorc;
1911 	rte_stats->obytes   = stats->gotc;
1912 
1913 	/* Get per-queue statuses */
1914 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1915 		/* GET TX queue statuses */
1916 		int map_id = igc->txq_stats_map[i];
1917 		if (map_id >= 0) {
1918 			rte_stats->q_opackets[map_id] += queue_stats->pqgptc[i];
1919 			rte_stats->q_obytes[map_id] += queue_stats->pqgotc[i];
1920 		}
1921 		/* Get RX queue statuses */
1922 		map_id = igc->rxq_stats_map[i];
1923 		if (map_id >= 0) {
1924 			rte_stats->q_ipackets[map_id] += queue_stats->pqgprc[i];
1925 			rte_stats->q_ibytes[map_id] += queue_stats->pqgorc[i];
1926 			rte_stats->q_errors[map_id] += queue_stats->rqdpc[i];
1927 		}
1928 	}
1929 
1930 	/* Restart queue status handler */
1931 	rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1932 			igc_update_queue_stats_handler, dev);
1933 	return 0;
1934 }
1935 
1936 static int
1937 eth_igc_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1938 		   unsigned int n)
1939 {
1940 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1941 	struct igc_hw_stats *hw_stats =
1942 			IGC_DEV_PRIVATE_STATS(dev);
1943 	unsigned int i;
1944 
1945 	igc_read_stats_registers(hw, hw_stats);
1946 
1947 	if (n < IGC_NB_XSTATS)
1948 		return IGC_NB_XSTATS;
1949 
1950 	/* If this is a reset xstats is NULL, and we have cleared the
1951 	 * registers by reading them.
1952 	 */
1953 	if (!xstats)
1954 		return 0;
1955 
1956 	/* Extended stats */
1957 	for (i = 0; i < IGC_NB_XSTATS; i++) {
1958 		xstats[i].id = i;
1959 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1960 			rte_igc_stats_strings[i].offset);
1961 	}
1962 
1963 	return IGC_NB_XSTATS;
1964 }
1965 
1966 static int
1967 eth_igc_xstats_reset(struct rte_eth_dev *dev)
1968 {
1969 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1970 	struct igc_hw_stats *hw_stats = IGC_DEV_PRIVATE_STATS(dev);
1971 	struct igc_hw_queue_stats *queue_stats =
1972 			IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1973 
1974 	/* Cancel queue status handler for avoid conflict */
1975 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
1976 
1977 	/* HW registers are cleared on read */
1978 	igc_reset_queue_stats_register(hw);
1979 	igc_read_stats_registers(hw, hw_stats);
1980 
1981 	/* Reset software totals */
1982 	memset(hw_stats, 0, sizeof(*hw_stats));
1983 	memset(queue_stats, 0, sizeof(*queue_stats));
1984 
1985 	/* Restart the queue status handler */
1986 	rte_eal_alarm_set(IGC_ALARM_INTERVAL, igc_update_queue_stats_handler,
1987 			dev);
1988 
1989 	return 0;
1990 }
1991 
1992 static int
1993 eth_igc_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1994 	struct rte_eth_xstat_name *xstats_names, unsigned int size)
1995 {
1996 	unsigned int i;
1997 
1998 	if (xstats_names == NULL)
1999 		return IGC_NB_XSTATS;
2000 
2001 	if (size < IGC_NB_XSTATS) {
2002 		PMD_DRV_LOG(ERR, "not enough buffers!");
2003 		return IGC_NB_XSTATS;
2004 	}
2005 
2006 	for (i = 0; i < IGC_NB_XSTATS; i++)
2007 		strlcpy(xstats_names[i].name, rte_igc_stats_strings[i].name,
2008 			sizeof(xstats_names[i].name));
2009 
2010 	return IGC_NB_XSTATS;
2011 }
2012 
2013 static int
2014 eth_igc_xstats_get_names_by_id(struct rte_eth_dev *dev,
2015 		const uint64_t *ids, struct rte_eth_xstat_name *xstats_names,
2016 		unsigned int limit)
2017 {
2018 	unsigned int i;
2019 
2020 	if (!ids)
2021 		return eth_igc_xstats_get_names(dev, xstats_names, limit);
2022 
2023 	for (i = 0; i < limit; i++) {
2024 		if (ids[i] >= IGC_NB_XSTATS) {
2025 			PMD_DRV_LOG(ERR, "id value isn't valid");
2026 			return -EINVAL;
2027 		}
2028 		strlcpy(xstats_names[i].name,
2029 			rte_igc_stats_strings[ids[i]].name,
2030 			sizeof(xstats_names[i].name));
2031 	}
2032 	return limit;
2033 }
2034 
2035 static int
2036 eth_igc_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
2037 		uint64_t *values, unsigned int n)
2038 {
2039 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2040 	struct igc_hw_stats *hw_stats = IGC_DEV_PRIVATE_STATS(dev);
2041 	unsigned int i;
2042 
2043 	igc_read_stats_registers(hw, hw_stats);
2044 
2045 	if (!ids) {
2046 		if (n < IGC_NB_XSTATS)
2047 			return IGC_NB_XSTATS;
2048 
2049 		/* If this is a reset xstats is NULL, and we have cleared the
2050 		 * registers by reading them.
2051 		 */
2052 		if (!values)
2053 			return 0;
2054 
2055 		/* Extended stats */
2056 		for (i = 0; i < IGC_NB_XSTATS; i++)
2057 			values[i] = *(uint64_t *)(((char *)hw_stats) +
2058 					rte_igc_stats_strings[i].offset);
2059 
2060 		return IGC_NB_XSTATS;
2061 
2062 	} else {
2063 		for (i = 0; i < n; i++) {
2064 			if (ids[i] >= IGC_NB_XSTATS) {
2065 				PMD_DRV_LOG(ERR, "id value isn't valid");
2066 				return -EINVAL;
2067 			}
2068 			values[i] = *(uint64_t *)(((char *)hw_stats) +
2069 					rte_igc_stats_strings[ids[i]].offset);
2070 		}
2071 		return n;
2072 	}
2073 }
2074 
2075 static int
2076 eth_igc_queue_stats_mapping_set(struct rte_eth_dev *dev,
2077 		uint16_t queue_id, uint8_t stat_idx, uint8_t is_rx)
2078 {
2079 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
2080 
2081 	/* check queue id is valid */
2082 	if (queue_id >= IGC_QUEUE_PAIRS_NUM) {
2083 		PMD_DRV_LOG(ERR, "queue id(%u) error, max is %u",
2084 			queue_id, IGC_QUEUE_PAIRS_NUM - 1);
2085 		return -EINVAL;
2086 	}
2087 
2088 	/* store the mapping status id */
2089 	if (is_rx)
2090 		igc->rxq_stats_map[queue_id] = stat_idx;
2091 	else
2092 		igc->txq_stats_map[queue_id] = stat_idx;
2093 
2094 	return 0;
2095 }
2096 
2097 static int
2098 eth_igc_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
2099 {
2100 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2101 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2102 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2103 	uint32_t vec = IGC_MISC_VEC_ID;
2104 
2105 	if (rte_intr_allow_others(intr_handle))
2106 		vec = IGC_RX_VEC_START;
2107 
2108 	uint32_t mask = 1u << (queue_id + vec);
2109 
2110 	IGC_WRITE_REG(hw, IGC_EIMC, mask);
2111 	IGC_WRITE_FLUSH(hw);
2112 
2113 	return 0;
2114 }
2115 
2116 static int
2117 eth_igc_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
2118 {
2119 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2120 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2121 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2122 	uint32_t vec = IGC_MISC_VEC_ID;
2123 
2124 	if (rte_intr_allow_others(intr_handle))
2125 		vec = IGC_RX_VEC_START;
2126 
2127 	uint32_t mask = 1u << (queue_id + vec);
2128 
2129 	IGC_WRITE_REG(hw, IGC_EIMS, mask);
2130 	IGC_WRITE_FLUSH(hw);
2131 
2132 	rte_intr_enable(intr_handle);
2133 
2134 	return 0;
2135 }
2136 
2137 static int
2138 eth_igc_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2139 {
2140 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2141 	uint32_t ctrl;
2142 	int tx_pause;
2143 	int rx_pause;
2144 
2145 	fc_conf->pause_time = hw->fc.pause_time;
2146 	fc_conf->high_water = hw->fc.high_water;
2147 	fc_conf->low_water = hw->fc.low_water;
2148 	fc_conf->send_xon = hw->fc.send_xon;
2149 	fc_conf->autoneg = hw->mac.autoneg;
2150 
2151 	/*
2152 	 * Return rx_pause and tx_pause status according to actual setting of
2153 	 * the TFCE and RFCE bits in the CTRL register.
2154 	 */
2155 	ctrl = IGC_READ_REG(hw, IGC_CTRL);
2156 	if (ctrl & IGC_CTRL_TFCE)
2157 		tx_pause = 1;
2158 	else
2159 		tx_pause = 0;
2160 
2161 	if (ctrl & IGC_CTRL_RFCE)
2162 		rx_pause = 1;
2163 	else
2164 		rx_pause = 0;
2165 
2166 	if (rx_pause && tx_pause)
2167 		fc_conf->mode = RTE_FC_FULL;
2168 	else if (rx_pause)
2169 		fc_conf->mode = RTE_FC_RX_PAUSE;
2170 	else if (tx_pause)
2171 		fc_conf->mode = RTE_FC_TX_PAUSE;
2172 	else
2173 		fc_conf->mode = RTE_FC_NONE;
2174 
2175 	return 0;
2176 }
2177 
2178 static int
2179 eth_igc_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2180 {
2181 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2182 	uint32_t rx_buf_size;
2183 	uint32_t max_high_water;
2184 	uint32_t rctl;
2185 	int err;
2186 
2187 	if (fc_conf->autoneg != hw->mac.autoneg)
2188 		return -ENOTSUP;
2189 
2190 	rx_buf_size = igc_get_rx_buffer_size(hw);
2191 	PMD_DRV_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
2192 
2193 	/* At least reserve one Ethernet frame for watermark */
2194 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
2195 	if (fc_conf->high_water > max_high_water ||
2196 		fc_conf->high_water < fc_conf->low_water) {
2197 		PMD_DRV_LOG(ERR,
2198 			"Incorrect high(%u)/low(%u) water value, max is %u",
2199 			fc_conf->high_water, fc_conf->low_water,
2200 			max_high_water);
2201 		return -EINVAL;
2202 	}
2203 
2204 	switch (fc_conf->mode) {
2205 	case RTE_FC_NONE:
2206 		hw->fc.requested_mode = igc_fc_none;
2207 		break;
2208 	case RTE_FC_RX_PAUSE:
2209 		hw->fc.requested_mode = igc_fc_rx_pause;
2210 		break;
2211 	case RTE_FC_TX_PAUSE:
2212 		hw->fc.requested_mode = igc_fc_tx_pause;
2213 		break;
2214 	case RTE_FC_FULL:
2215 		hw->fc.requested_mode = igc_fc_full;
2216 		break;
2217 	default:
2218 		PMD_DRV_LOG(ERR, "unsupported fc mode: %u", fc_conf->mode);
2219 		return -EINVAL;
2220 	}
2221 
2222 	hw->fc.pause_time     = fc_conf->pause_time;
2223 	hw->fc.high_water     = fc_conf->high_water;
2224 	hw->fc.low_water      = fc_conf->low_water;
2225 	hw->fc.send_xon	      = fc_conf->send_xon;
2226 
2227 	err = igc_setup_link_generic(hw);
2228 	if (err == IGC_SUCCESS) {
2229 		/**
2230 		 * check if we want to forward MAC frames - driver doesn't have
2231 		 * native capability to do that, so we'll write the registers
2232 		 * ourselves
2233 		 **/
2234 		rctl = IGC_READ_REG(hw, IGC_RCTL);
2235 
2236 		/* set or clear MFLCN.PMCF bit depending on configuration */
2237 		if (fc_conf->mac_ctrl_frame_fwd != 0)
2238 			rctl |= IGC_RCTL_PMCF;
2239 		else
2240 			rctl &= ~IGC_RCTL_PMCF;
2241 
2242 		IGC_WRITE_REG(hw, IGC_RCTL, rctl);
2243 		IGC_WRITE_FLUSH(hw);
2244 
2245 		return 0;
2246 	}
2247 
2248 	PMD_DRV_LOG(ERR, "igc_setup_link_generic = 0x%x", err);
2249 	return -EIO;
2250 }
2251 
2252 static int
2253 eth_igc_rss_reta_update(struct rte_eth_dev *dev,
2254 			struct rte_eth_rss_reta_entry64 *reta_conf,
2255 			uint16_t reta_size)
2256 {
2257 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2258 	uint16_t i;
2259 
2260 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
2261 		PMD_DRV_LOG(ERR,
2262 			"The size of RSS redirection table configured(%d) doesn't match the number hardware can supported(%d)",
2263 			reta_size, ETH_RSS_RETA_SIZE_128);
2264 		return -EINVAL;
2265 	}
2266 
2267 	RTE_BUILD_BUG_ON(ETH_RSS_RETA_SIZE_128 % IGC_RSS_RDT_REG_SIZE);
2268 
2269 	/* set redirection table */
2270 	for (i = 0; i < ETH_RSS_RETA_SIZE_128; i += IGC_RSS_RDT_REG_SIZE) {
2271 		union igc_rss_reta_reg reta, reg;
2272 		uint16_t idx, shift;
2273 		uint8_t j, mask;
2274 
2275 		idx = i / RTE_RETA_GROUP_SIZE;
2276 		shift = i % RTE_RETA_GROUP_SIZE;
2277 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
2278 				IGC_RSS_RDT_REG_SIZE_MASK);
2279 
2280 		/* if no need to update the register */
2281 		if (!mask ||
2282 		    shift > (RTE_RETA_GROUP_SIZE - IGC_RSS_RDT_REG_SIZE))
2283 			continue;
2284 
2285 		/* check mask whether need to read the register value first */
2286 		if (mask == IGC_RSS_RDT_REG_SIZE_MASK)
2287 			reg.dword = 0;
2288 		else
2289 			reg.dword = IGC_READ_REG_LE_VALUE(hw,
2290 					IGC_RETA(i / IGC_RSS_RDT_REG_SIZE));
2291 
2292 		/* update the register */
2293 		RTE_BUILD_BUG_ON(sizeof(reta.bytes) != IGC_RSS_RDT_REG_SIZE);
2294 		for (j = 0; j < IGC_RSS_RDT_REG_SIZE; j++) {
2295 			if (mask & (1u << j))
2296 				reta.bytes[j] =
2297 					(uint8_t)reta_conf[idx].reta[shift + j];
2298 			else
2299 				reta.bytes[j] = reg.bytes[j];
2300 		}
2301 		IGC_WRITE_REG_LE_VALUE(hw,
2302 			IGC_RETA(i / IGC_RSS_RDT_REG_SIZE), reta.dword);
2303 	}
2304 
2305 	return 0;
2306 }
2307 
2308 static int
2309 eth_igc_rss_reta_query(struct rte_eth_dev *dev,
2310 		       struct rte_eth_rss_reta_entry64 *reta_conf,
2311 		       uint16_t reta_size)
2312 {
2313 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2314 	uint16_t i;
2315 
2316 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
2317 		PMD_DRV_LOG(ERR,
2318 			"The size of RSS redirection table configured(%d) doesn't match the number hardware can supported(%d)",
2319 			reta_size, ETH_RSS_RETA_SIZE_128);
2320 		return -EINVAL;
2321 	}
2322 
2323 	RTE_BUILD_BUG_ON(ETH_RSS_RETA_SIZE_128 % IGC_RSS_RDT_REG_SIZE);
2324 
2325 	/* read redirection table */
2326 	for (i = 0; i < ETH_RSS_RETA_SIZE_128; i += IGC_RSS_RDT_REG_SIZE) {
2327 		union igc_rss_reta_reg reta;
2328 		uint16_t idx, shift;
2329 		uint8_t j, mask;
2330 
2331 		idx = i / RTE_RETA_GROUP_SIZE;
2332 		shift = i % RTE_RETA_GROUP_SIZE;
2333 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
2334 				IGC_RSS_RDT_REG_SIZE_MASK);
2335 
2336 		/* if no need to read register */
2337 		if (!mask ||
2338 		    shift > (RTE_RETA_GROUP_SIZE - IGC_RSS_RDT_REG_SIZE))
2339 			continue;
2340 
2341 		/* read register and get the queue index */
2342 		RTE_BUILD_BUG_ON(sizeof(reta.bytes) != IGC_RSS_RDT_REG_SIZE);
2343 		reta.dword = IGC_READ_REG_LE_VALUE(hw,
2344 				IGC_RETA(i / IGC_RSS_RDT_REG_SIZE));
2345 		for (j = 0; j < IGC_RSS_RDT_REG_SIZE; j++) {
2346 			if (mask & (1u << j))
2347 				reta_conf[idx].reta[shift + j] = reta.bytes[j];
2348 		}
2349 	}
2350 
2351 	return 0;
2352 }
2353 
2354 static int
2355 eth_igc_rss_hash_update(struct rte_eth_dev *dev,
2356 			struct rte_eth_rss_conf *rss_conf)
2357 {
2358 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2359 	igc_hw_rss_hash_set(hw, rss_conf);
2360 	return 0;
2361 }
2362 
2363 static int
2364 eth_igc_rss_hash_conf_get(struct rte_eth_dev *dev,
2365 			struct rte_eth_rss_conf *rss_conf)
2366 {
2367 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2368 	uint32_t *hash_key = (uint32_t *)rss_conf->rss_key;
2369 	uint32_t mrqc;
2370 	uint64_t rss_hf;
2371 
2372 	if (hash_key != NULL) {
2373 		int i;
2374 
2375 		/* if not enough space for store hash key */
2376 		if (rss_conf->rss_key_len != IGC_HKEY_SIZE) {
2377 			PMD_DRV_LOG(ERR,
2378 				"RSS hash key size %u in parameter doesn't match the hardware hash key size %u",
2379 				rss_conf->rss_key_len, IGC_HKEY_SIZE);
2380 			return -EINVAL;
2381 		}
2382 
2383 		/* read RSS key from register */
2384 		for (i = 0; i < IGC_HKEY_MAX_INDEX; i++)
2385 			hash_key[i] = IGC_READ_REG_LE_VALUE(hw, IGC_RSSRK(i));
2386 	}
2387 
2388 	/* get RSS functions configured in MRQC register */
2389 	mrqc = IGC_READ_REG(hw, IGC_MRQC);
2390 	if ((mrqc & IGC_MRQC_ENABLE_RSS_4Q) == 0)
2391 		return 0;
2392 
2393 	rss_hf = 0;
2394 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4)
2395 		rss_hf |= ETH_RSS_IPV4;
2396 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4_TCP)
2397 		rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2398 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6)
2399 		rss_hf |= ETH_RSS_IPV6;
2400 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_EX)
2401 		rss_hf |= ETH_RSS_IPV6_EX;
2402 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_TCP)
2403 		rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2404 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_TCP_EX)
2405 		rss_hf |= ETH_RSS_IPV6_TCP_EX;
2406 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4_UDP)
2407 		rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2408 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_UDP)
2409 		rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2410 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_UDP_EX)
2411 		rss_hf |= ETH_RSS_IPV6_UDP_EX;
2412 
2413 	rss_conf->rss_hf |= rss_hf;
2414 	return 0;
2415 }
2416 
2417 static int
2418 eth_igc_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2419 {
2420 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2421 	struct igc_vfta *shadow_vfta = IGC_DEV_PRIVATE_VFTA(dev);
2422 	uint32_t vfta;
2423 	uint32_t vid_idx;
2424 	uint32_t vid_bit;
2425 
2426 	vid_idx = (vlan_id >> IGC_VFTA_ENTRY_SHIFT) & IGC_VFTA_ENTRY_MASK;
2427 	vid_bit = 1u << (vlan_id & IGC_VFTA_ENTRY_BIT_SHIFT_MASK);
2428 	vfta = shadow_vfta->vfta[vid_idx];
2429 	if (on)
2430 		vfta |= vid_bit;
2431 	else
2432 		vfta &= ~vid_bit;
2433 	IGC_WRITE_REG_ARRAY(hw, IGC_VFTA, vid_idx, vfta);
2434 
2435 	/* update local VFTA copy */
2436 	shadow_vfta->vfta[vid_idx] = vfta;
2437 
2438 	return 0;
2439 }
2440 
2441 static void
2442 igc_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2443 {
2444 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2445 	igc_read_reg_check_clear_bits(hw, IGC_RCTL,
2446 			IGC_RCTL_CFIEN | IGC_RCTL_VFE);
2447 }
2448 
2449 static void
2450 igc_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2451 {
2452 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2453 	struct igc_vfta *shadow_vfta = IGC_DEV_PRIVATE_VFTA(dev);
2454 	uint32_t reg_val;
2455 	int i;
2456 
2457 	/* Filter Table Enable, CFI not used for packet acceptance */
2458 	reg_val = IGC_READ_REG(hw, IGC_RCTL);
2459 	reg_val &= ~IGC_RCTL_CFIEN;
2460 	reg_val |= IGC_RCTL_VFE;
2461 	IGC_WRITE_REG(hw, IGC_RCTL, reg_val);
2462 
2463 	/* restore VFTA table */
2464 	for (i = 0; i < IGC_VFTA_SIZE; i++)
2465 		IGC_WRITE_REG_ARRAY(hw, IGC_VFTA, i, shadow_vfta->vfta[i]);
2466 }
2467 
2468 static void
2469 igc_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2470 {
2471 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2472 
2473 	igc_read_reg_check_clear_bits(hw, IGC_CTRL, IGC_CTRL_VME);
2474 }
2475 
2476 static void
2477 igc_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2478 {
2479 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2480 
2481 	igc_read_reg_check_set_bits(hw, IGC_CTRL, IGC_CTRL_VME);
2482 }
2483 
2484 static int
2485 igc_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2486 {
2487 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2488 	uint32_t ctrl_ext;
2489 
2490 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
2491 
2492 	/* if extend vlan hasn't been enabled */
2493 	if ((ctrl_ext & IGC_CTRL_EXT_EXT_VLAN) == 0)
2494 		return 0;
2495 
2496 	if ((dev->data->dev_conf.rxmode.offloads &
2497 			DEV_RX_OFFLOAD_JUMBO_FRAME) == 0)
2498 		goto write_ext_vlan;
2499 
2500 	/* Update maximum packet length */
2501 	if (dev->data->dev_conf.rxmode.max_rx_pkt_len <
2502 		RTE_ETHER_MIN_MTU + VLAN_TAG_SIZE) {
2503 		PMD_DRV_LOG(ERR, "Maximum packet length %u error, min is %u",
2504 			dev->data->dev_conf.rxmode.max_rx_pkt_len,
2505 			VLAN_TAG_SIZE + RTE_ETHER_MIN_MTU);
2506 		return -EINVAL;
2507 	}
2508 	dev->data->dev_conf.rxmode.max_rx_pkt_len -= VLAN_TAG_SIZE;
2509 	IGC_WRITE_REG(hw, IGC_RLPML,
2510 		dev->data->dev_conf.rxmode.max_rx_pkt_len);
2511 
2512 write_ext_vlan:
2513 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext & ~IGC_CTRL_EXT_EXT_VLAN);
2514 	return 0;
2515 }
2516 
2517 static int
2518 igc_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2519 {
2520 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2521 	uint32_t ctrl_ext;
2522 
2523 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
2524 
2525 	/* if extend vlan has been enabled */
2526 	if (ctrl_ext & IGC_CTRL_EXT_EXT_VLAN)
2527 		return 0;
2528 
2529 	if ((dev->data->dev_conf.rxmode.offloads &
2530 			DEV_RX_OFFLOAD_JUMBO_FRAME) == 0)
2531 		goto write_ext_vlan;
2532 
2533 	/* Update maximum packet length */
2534 	if (dev->data->dev_conf.rxmode.max_rx_pkt_len >
2535 		MAX_RX_JUMBO_FRAME_SIZE - VLAN_TAG_SIZE) {
2536 		PMD_DRV_LOG(ERR, "Maximum packet length %u error, max is %u",
2537 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2538 			VLAN_TAG_SIZE, MAX_RX_JUMBO_FRAME_SIZE);
2539 		return -EINVAL;
2540 	}
2541 	dev->data->dev_conf.rxmode.max_rx_pkt_len += VLAN_TAG_SIZE;
2542 	IGC_WRITE_REG(hw, IGC_RLPML,
2543 		dev->data->dev_conf.rxmode.max_rx_pkt_len);
2544 
2545 write_ext_vlan:
2546 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext | IGC_CTRL_EXT_EXT_VLAN);
2547 	return 0;
2548 }
2549 
2550 static int
2551 eth_igc_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2552 {
2553 	struct rte_eth_rxmode *rxmode;
2554 
2555 	rxmode = &dev->data->dev_conf.rxmode;
2556 	if (mask & ETH_VLAN_STRIP_MASK) {
2557 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2558 			igc_vlan_hw_strip_enable(dev);
2559 		else
2560 			igc_vlan_hw_strip_disable(dev);
2561 	}
2562 
2563 	if (mask & ETH_VLAN_FILTER_MASK) {
2564 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2565 			igc_vlan_hw_filter_enable(dev);
2566 		else
2567 			igc_vlan_hw_filter_disable(dev);
2568 	}
2569 
2570 	if (mask & ETH_VLAN_EXTEND_MASK) {
2571 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2572 			return igc_vlan_hw_extend_enable(dev);
2573 		else
2574 			return igc_vlan_hw_extend_disable(dev);
2575 	}
2576 
2577 	return 0;
2578 }
2579 
2580 static int
2581 eth_igc_vlan_tpid_set(struct rte_eth_dev *dev,
2582 		      enum rte_vlan_type vlan_type,
2583 		      uint16_t tpid)
2584 {
2585 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2586 	uint32_t reg_val;
2587 
2588 	/* only outer TPID of double VLAN can be configured*/
2589 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2590 		reg_val = IGC_READ_REG(hw, IGC_VET);
2591 		reg_val = (reg_val & (~IGC_VET_EXT)) |
2592 			((uint32_t)tpid << IGC_VET_EXT_SHIFT);
2593 		IGC_WRITE_REG(hw, IGC_VET, reg_val);
2594 
2595 		return 0;
2596 	}
2597 
2598 	/* all other TPID values are read-only*/
2599 	PMD_DRV_LOG(ERR, "Not supported");
2600 	return -ENOTSUP;
2601 }
2602 
2603 static int
2604 eth_igc_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
2605 	struct rte_pci_device *pci_dev)
2606 {
2607 	PMD_INIT_FUNC_TRACE();
2608 	return rte_eth_dev_pci_generic_probe(pci_dev,
2609 		sizeof(struct igc_adapter), eth_igc_dev_init);
2610 }
2611 
2612 static int
2613 eth_igc_pci_remove(struct rte_pci_device *pci_dev)
2614 {
2615 	PMD_INIT_FUNC_TRACE();
2616 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igc_dev_uninit);
2617 }
2618 
2619 static struct rte_pci_driver rte_igc_pmd = {
2620 	.id_table = pci_id_igc_map,
2621 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
2622 	.probe = eth_igc_pci_probe,
2623 	.remove = eth_igc_pci_remove,
2624 };
2625 
2626 RTE_PMD_REGISTER_PCI(net_igc, rte_igc_pmd);
2627 RTE_PMD_REGISTER_PCI_TABLE(net_igc, pci_id_igc_map);
2628 RTE_PMD_REGISTER_KMOD_DEP(net_igc, "* igb_uio | uio_pci_generic | vfio-pci");
2629