xref: /dpdk/drivers/net/igc/igc_ethdev.c (revision 2cdca6ed)
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 #define IGC_FC_PAUSE_TIME		0x0680
24 #define IGC_LINK_UPDATE_CHECK_TIMEOUT	90  /* 9s */
25 #define IGC_LINK_UPDATE_CHECK_INTERVAL	100 /* ms */
26 
27 #define IGC_MISC_VEC_ID			RTE_INTR_VEC_ZERO_OFFSET
28 #define IGC_RX_VEC_START		RTE_INTR_VEC_RXTX_OFFSET
29 #define IGC_MSIX_OTHER_INTR_VEC		0   /* MSI-X other interrupt vector */
30 #define IGC_FLAG_NEED_LINK_UPDATE	(1u << 0)	/* need update link */
31 
32 #define IGC_DEFAULT_RX_FREE_THRESH	32
33 
34 #define IGC_DEFAULT_RX_PTHRESH		8
35 #define IGC_DEFAULT_RX_HTHRESH		8
36 #define IGC_DEFAULT_RX_WTHRESH		4
37 
38 #define IGC_DEFAULT_TX_PTHRESH		8
39 #define IGC_DEFAULT_TX_HTHRESH		1
40 #define IGC_DEFAULT_TX_WTHRESH		16
41 
42 /* MSI-X other interrupt vector */
43 #define IGC_MSIX_OTHER_INTR_VEC		0
44 
45 /* External VLAN Enable bit mask */
46 #define IGC_CTRL_EXT_EXT_VLAN		(1u << 26)
47 
48 /* Speed select */
49 #define IGC_CTRL_SPEED_MASK		(7u << 8)
50 #define IGC_CTRL_SPEED_2500		(6u << 8)
51 
52 /* External VLAN Ether Type bit mask and shift */
53 #define IGC_VET_EXT			0xFFFF0000
54 #define IGC_VET_EXT_SHIFT		16
55 
56 /* Force EEE Auto-negotiation */
57 #define IGC_EEER_EEE_FRC_AN		(1u << 28)
58 
59 /* Per Queue Good Packets Received Count */
60 #define IGC_PQGPRC(idx)		(0x10010 + 0x100 * (idx))
61 /* Per Queue Good Octets Received Count */
62 #define IGC_PQGORC(idx)		(0x10018 + 0x100 * (idx))
63 /* Per Queue Good Octets Transmitted Count */
64 #define IGC_PQGOTC(idx)		(0x10034 + 0x100 * (idx))
65 /* Per Queue Multicast Packets Received Count */
66 #define IGC_PQMPRC(idx)		(0x10038 + 0x100 * (idx))
67 /* Transmit Queue Drop Packet Count */
68 #define IGC_TQDPC(idx)		(0xe030 + 0x40 * (idx))
69 
70 #if RTE_BYTE_ORDER == RTE_LITTLE_ENDIAN
71 #define U32_0_IN_U64		0	/* lower bytes of u64 */
72 #define U32_1_IN_U64		1	/* higher bytes of u64 */
73 #else
74 #define U32_0_IN_U64		1
75 #define U32_1_IN_U64		0
76 #endif
77 
78 #define IGC_ALARM_INTERVAL	8000000u
79 /* us, about 13.6s some per-queue registers will wrap around back to 0. */
80 
81 static const struct rte_eth_desc_lim rx_desc_lim = {
82 	.nb_max = IGC_MAX_RXD,
83 	.nb_min = IGC_MIN_RXD,
84 	.nb_align = IGC_RXD_ALIGN,
85 };
86 
87 static const struct rte_eth_desc_lim tx_desc_lim = {
88 	.nb_max = IGC_MAX_TXD,
89 	.nb_min = IGC_MIN_TXD,
90 	.nb_align = IGC_TXD_ALIGN,
91 	.nb_seg_max = IGC_TX_MAX_SEG,
92 	.nb_mtu_seg_max = IGC_TX_MAX_MTU_SEG,
93 };
94 
95 static const struct rte_pci_id pci_id_igc_map[] = {
96 	{ RTE_PCI_DEVICE(IGC_INTEL_VENDOR_ID, IGC_DEV_ID_I225_LM) },
97 	{ RTE_PCI_DEVICE(IGC_INTEL_VENDOR_ID, IGC_DEV_ID_I225_V)  },
98 	{ RTE_PCI_DEVICE(IGC_INTEL_VENDOR_ID, IGC_DEV_ID_I225_I)  },
99 	{ RTE_PCI_DEVICE(IGC_INTEL_VENDOR_ID, IGC_DEV_ID_I225_K)  },
100 	{ .vendor_id = 0, /* sentinel */ },
101 };
102 
103 /* store statistics names and its offset in stats structure */
104 struct rte_igc_xstats_name_off {
105 	char name[RTE_ETH_XSTATS_NAME_SIZE];
106 	unsigned int offset;
107 };
108 
109 static const struct rte_igc_xstats_name_off rte_igc_stats_strings[] = {
110 	{"rx_crc_errors", offsetof(struct igc_hw_stats, crcerrs)},
111 	{"rx_align_errors", offsetof(struct igc_hw_stats, algnerrc)},
112 	{"rx_errors", offsetof(struct igc_hw_stats, rxerrc)},
113 	{"rx_missed_packets", offsetof(struct igc_hw_stats, mpc)},
114 	{"tx_single_collision_packets", offsetof(struct igc_hw_stats, scc)},
115 	{"tx_multiple_collision_packets", offsetof(struct igc_hw_stats, mcc)},
116 	{"tx_excessive_collision_packets", offsetof(struct igc_hw_stats,
117 		ecol)},
118 	{"tx_late_collisions", offsetof(struct igc_hw_stats, latecol)},
119 	{"tx_total_collisions", offsetof(struct igc_hw_stats, colc)},
120 	{"tx_deferred_packets", offsetof(struct igc_hw_stats, dc)},
121 	{"tx_no_carrier_sense_packets", offsetof(struct igc_hw_stats, tncrs)},
122 	{"tx_discarded_packets", offsetof(struct igc_hw_stats, htdpmc)},
123 	{"rx_length_errors", offsetof(struct igc_hw_stats, rlec)},
124 	{"rx_xon_packets", offsetof(struct igc_hw_stats, xonrxc)},
125 	{"tx_xon_packets", offsetof(struct igc_hw_stats, xontxc)},
126 	{"rx_xoff_packets", offsetof(struct igc_hw_stats, xoffrxc)},
127 	{"tx_xoff_packets", offsetof(struct igc_hw_stats, xofftxc)},
128 	{"rx_flow_control_unsupported_packets", offsetof(struct igc_hw_stats,
129 		fcruc)},
130 	{"rx_size_64_packets", offsetof(struct igc_hw_stats, prc64)},
131 	{"rx_size_65_to_127_packets", offsetof(struct igc_hw_stats, prc127)},
132 	{"rx_size_128_to_255_packets", offsetof(struct igc_hw_stats, prc255)},
133 	{"rx_size_256_to_511_packets", offsetof(struct igc_hw_stats, prc511)},
134 	{"rx_size_512_to_1023_packets", offsetof(struct igc_hw_stats,
135 		prc1023)},
136 	{"rx_size_1024_to_max_packets", offsetof(struct igc_hw_stats,
137 		prc1522)},
138 	{"rx_broadcast_packets", offsetof(struct igc_hw_stats, bprc)},
139 	{"rx_multicast_packets", offsetof(struct igc_hw_stats, mprc)},
140 	{"rx_undersize_errors", offsetof(struct igc_hw_stats, ruc)},
141 	{"rx_fragment_errors", offsetof(struct igc_hw_stats, rfc)},
142 	{"rx_oversize_errors", offsetof(struct igc_hw_stats, roc)},
143 	{"rx_jabber_errors", offsetof(struct igc_hw_stats, rjc)},
144 	{"rx_no_buffers", offsetof(struct igc_hw_stats, rnbc)},
145 	{"rx_management_packets", offsetof(struct igc_hw_stats, mgprc)},
146 	{"rx_management_dropped", offsetof(struct igc_hw_stats, mgpdc)},
147 	{"tx_management_packets", offsetof(struct igc_hw_stats, mgptc)},
148 	{"rx_total_packets", offsetof(struct igc_hw_stats, tpr)},
149 	{"tx_total_packets", offsetof(struct igc_hw_stats, tpt)},
150 	{"rx_total_bytes", offsetof(struct igc_hw_stats, tor)},
151 	{"tx_total_bytes", offsetof(struct igc_hw_stats, tot)},
152 	{"tx_size_64_packets", offsetof(struct igc_hw_stats, ptc64)},
153 	{"tx_size_65_to_127_packets", offsetof(struct igc_hw_stats, ptc127)},
154 	{"tx_size_128_to_255_packets", offsetof(struct igc_hw_stats, ptc255)},
155 	{"tx_size_256_to_511_packets", offsetof(struct igc_hw_stats, ptc511)},
156 	{"tx_size_512_to_1023_packets", offsetof(struct igc_hw_stats,
157 		ptc1023)},
158 	{"tx_size_1023_to_max_packets", offsetof(struct igc_hw_stats,
159 		ptc1522)},
160 	{"tx_multicast_packets", offsetof(struct igc_hw_stats, mptc)},
161 	{"tx_broadcast_packets", offsetof(struct igc_hw_stats, bptc)},
162 	{"tx_tso_packets", offsetof(struct igc_hw_stats, tsctc)},
163 	{"rx_sent_to_host_packets", offsetof(struct igc_hw_stats, rpthc)},
164 	{"tx_sent_by_host_packets", offsetof(struct igc_hw_stats, hgptc)},
165 	{"interrupt_assert_count", offsetof(struct igc_hw_stats, iac)},
166 	{"rx_descriptor_lower_threshold",
167 		offsetof(struct igc_hw_stats, icrxdmtc)},
168 };
169 
170 #define IGC_NB_XSTATS (sizeof(rte_igc_stats_strings) / \
171 		sizeof(rte_igc_stats_strings[0]))
172 
173 static int eth_igc_configure(struct rte_eth_dev *dev);
174 static int eth_igc_link_update(struct rte_eth_dev *dev, int wait_to_complete);
175 static int eth_igc_stop(struct rte_eth_dev *dev);
176 static int eth_igc_start(struct rte_eth_dev *dev);
177 static int eth_igc_set_link_up(struct rte_eth_dev *dev);
178 static int eth_igc_set_link_down(struct rte_eth_dev *dev);
179 static int eth_igc_close(struct rte_eth_dev *dev);
180 static int eth_igc_reset(struct rte_eth_dev *dev);
181 static int eth_igc_promiscuous_enable(struct rte_eth_dev *dev);
182 static int eth_igc_promiscuous_disable(struct rte_eth_dev *dev);
183 static int eth_igc_fw_version_get(struct rte_eth_dev *dev,
184 				char *fw_version, size_t fw_size);
185 static int eth_igc_infos_get(struct rte_eth_dev *dev,
186 			struct rte_eth_dev_info *dev_info);
187 static int eth_igc_led_on(struct rte_eth_dev *dev);
188 static int eth_igc_led_off(struct rte_eth_dev *dev);
189 static const uint32_t *eth_igc_supported_ptypes_get(struct rte_eth_dev *dev);
190 static int eth_igc_rar_set(struct rte_eth_dev *dev,
191 		struct rte_ether_addr *mac_addr, uint32_t index, uint32_t pool);
192 static void eth_igc_rar_clear(struct rte_eth_dev *dev, uint32_t index);
193 static int eth_igc_default_mac_addr_set(struct rte_eth_dev *dev,
194 			struct rte_ether_addr *addr);
195 static int eth_igc_set_mc_addr_list(struct rte_eth_dev *dev,
196 			 struct rte_ether_addr *mc_addr_set,
197 			 uint32_t nb_mc_addr);
198 static int eth_igc_allmulticast_enable(struct rte_eth_dev *dev);
199 static int eth_igc_allmulticast_disable(struct rte_eth_dev *dev);
200 static int eth_igc_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
201 static int eth_igc_stats_get(struct rte_eth_dev *dev,
202 			struct rte_eth_stats *rte_stats);
203 static int eth_igc_xstats_get(struct rte_eth_dev *dev,
204 			struct rte_eth_xstat *xstats, unsigned int n);
205 static int eth_igc_xstats_get_by_id(struct rte_eth_dev *dev,
206 				const uint64_t *ids,
207 				uint64_t *values, unsigned int n);
208 static int eth_igc_xstats_get_names(struct rte_eth_dev *dev,
209 				struct rte_eth_xstat_name *xstats_names,
210 				unsigned int size);
211 static int eth_igc_xstats_get_names_by_id(struct rte_eth_dev *dev,
212 		const uint64_t *ids, struct rte_eth_xstat_name *xstats_names,
213 		unsigned int limit);
214 static int eth_igc_xstats_reset(struct rte_eth_dev *dev);
215 static int
216 eth_igc_queue_stats_mapping_set(struct rte_eth_dev *dev,
217 	uint16_t queue_id, uint8_t stat_idx, uint8_t is_rx);
218 static int
219 eth_igc_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
220 static int
221 eth_igc_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
222 static int
223 eth_igc_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf);
224 static int
225 eth_igc_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf);
226 static int eth_igc_rss_reta_update(struct rte_eth_dev *dev,
227 			struct rte_eth_rss_reta_entry64 *reta_conf,
228 			uint16_t reta_size);
229 static int eth_igc_rss_reta_query(struct rte_eth_dev *dev,
230 		       struct rte_eth_rss_reta_entry64 *reta_conf,
231 		       uint16_t reta_size);
232 static int eth_igc_rss_hash_update(struct rte_eth_dev *dev,
233 			struct rte_eth_rss_conf *rss_conf);
234 static int eth_igc_rss_hash_conf_get(struct rte_eth_dev *dev,
235 			struct rte_eth_rss_conf *rss_conf);
236 static int
237 eth_igc_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on);
238 static int eth_igc_vlan_offload_set(struct rte_eth_dev *dev, int mask);
239 static int eth_igc_vlan_tpid_set(struct rte_eth_dev *dev,
240 		      enum rte_vlan_type vlan_type, uint16_t tpid);
241 
242 static const struct eth_dev_ops eth_igc_ops = {
243 	.dev_configure		= eth_igc_configure,
244 	.link_update		= eth_igc_link_update,
245 	.dev_stop		= eth_igc_stop,
246 	.dev_start		= eth_igc_start,
247 	.dev_close		= eth_igc_close,
248 	.dev_reset		= eth_igc_reset,
249 	.dev_set_link_up	= eth_igc_set_link_up,
250 	.dev_set_link_down	= eth_igc_set_link_down,
251 	.promiscuous_enable	= eth_igc_promiscuous_enable,
252 	.promiscuous_disable	= eth_igc_promiscuous_disable,
253 	.allmulticast_enable	= eth_igc_allmulticast_enable,
254 	.allmulticast_disable	= eth_igc_allmulticast_disable,
255 	.fw_version_get		= eth_igc_fw_version_get,
256 	.dev_infos_get		= eth_igc_infos_get,
257 	.dev_led_on		= eth_igc_led_on,
258 	.dev_led_off		= eth_igc_led_off,
259 	.dev_supported_ptypes_get = eth_igc_supported_ptypes_get,
260 	.mtu_set		= eth_igc_mtu_set,
261 	.mac_addr_add		= eth_igc_rar_set,
262 	.mac_addr_remove	= eth_igc_rar_clear,
263 	.mac_addr_set		= eth_igc_default_mac_addr_set,
264 	.set_mc_addr_list	= eth_igc_set_mc_addr_list,
265 
266 	.rx_queue_setup		= eth_igc_rx_queue_setup,
267 	.rx_queue_release	= eth_igc_rx_queue_release,
268 	.tx_queue_setup		= eth_igc_tx_queue_setup,
269 	.tx_queue_release	= eth_igc_tx_queue_release,
270 	.tx_done_cleanup	= eth_igc_tx_done_cleanup,
271 	.rxq_info_get		= eth_igc_rxq_info_get,
272 	.txq_info_get		= eth_igc_txq_info_get,
273 	.stats_get		= eth_igc_stats_get,
274 	.xstats_get		= eth_igc_xstats_get,
275 	.xstats_get_by_id	= eth_igc_xstats_get_by_id,
276 	.xstats_get_names_by_id	= eth_igc_xstats_get_names_by_id,
277 	.xstats_get_names	= eth_igc_xstats_get_names,
278 	.stats_reset		= eth_igc_xstats_reset,
279 	.xstats_reset		= eth_igc_xstats_reset,
280 	.queue_stats_mapping_set = eth_igc_queue_stats_mapping_set,
281 	.rx_queue_intr_enable	= eth_igc_rx_queue_intr_enable,
282 	.rx_queue_intr_disable	= eth_igc_rx_queue_intr_disable,
283 	.flow_ctrl_get		= eth_igc_flow_ctrl_get,
284 	.flow_ctrl_set		= eth_igc_flow_ctrl_set,
285 	.reta_update		= eth_igc_rss_reta_update,
286 	.reta_query		= eth_igc_rss_reta_query,
287 	.rss_hash_update	= eth_igc_rss_hash_update,
288 	.rss_hash_conf_get	= eth_igc_rss_hash_conf_get,
289 	.vlan_filter_set	= eth_igc_vlan_filter_set,
290 	.vlan_offload_set	= eth_igc_vlan_offload_set,
291 	.vlan_tpid_set		= eth_igc_vlan_tpid_set,
292 	.vlan_strip_queue_set	= eth_igc_vlan_strip_queue_set,
293 	.flow_ops_get		= eth_igc_flow_ops_get,
294 };
295 
296 /*
297  * multiple queue mode checking
298  */
299 static int
300 igc_check_mq_mode(struct rte_eth_dev *dev)
301 {
302 	enum rte_eth_rx_mq_mode rx_mq_mode = dev->data->dev_conf.rxmode.mq_mode;
303 	enum rte_eth_tx_mq_mode tx_mq_mode = dev->data->dev_conf.txmode.mq_mode;
304 
305 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
306 		PMD_INIT_LOG(ERR, "SRIOV is not supported.");
307 		return -EINVAL;
308 	}
309 
310 	if (rx_mq_mode != RTE_ETH_MQ_RX_NONE &&
311 		rx_mq_mode != RTE_ETH_MQ_RX_RSS) {
312 		/* RSS together with VMDq not supported*/
313 		PMD_INIT_LOG(ERR, "RX mode %d is not supported.",
314 				rx_mq_mode);
315 		return -EINVAL;
316 	}
317 
318 	/* To no break software that set invalid mode, only display
319 	 * warning if invalid mode is used.
320 	 */
321 	if (tx_mq_mode != RTE_ETH_MQ_TX_NONE)
322 		PMD_INIT_LOG(WARNING,
323 			"TX mode %d is not supported. Due to meaningless in this driver, just ignore",
324 			tx_mq_mode);
325 
326 	return 0;
327 }
328 
329 static int
330 eth_igc_configure(struct rte_eth_dev *dev)
331 {
332 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
333 	int ret;
334 
335 	PMD_INIT_FUNC_TRACE();
336 
337 	if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
338 		dev->data->dev_conf.rxmode.offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
339 
340 	ret  = igc_check_mq_mode(dev);
341 	if (ret != 0)
342 		return ret;
343 
344 	intr->flags |= IGC_FLAG_NEED_LINK_UPDATE;
345 	return 0;
346 }
347 
348 static int
349 eth_igc_set_link_up(struct rte_eth_dev *dev)
350 {
351 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
352 
353 	if (hw->phy.media_type == igc_media_type_copper)
354 		igc_power_up_phy(hw);
355 	else
356 		igc_power_up_fiber_serdes_link(hw);
357 	return 0;
358 }
359 
360 static int
361 eth_igc_set_link_down(struct rte_eth_dev *dev)
362 {
363 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
364 
365 	if (hw->phy.media_type == igc_media_type_copper)
366 		igc_power_down_phy(hw);
367 	else
368 		igc_shutdown_fiber_serdes_link(hw);
369 	return 0;
370 }
371 
372 /*
373  * disable other interrupt
374  */
375 static void
376 igc_intr_other_disable(struct rte_eth_dev *dev)
377 {
378 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
379 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
380 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
381 
382 	if (rte_intr_allow_others(intr_handle) &&
383 		dev->data->dev_conf.intr_conf.lsc) {
384 		IGC_WRITE_REG(hw, IGC_EIMC, 1u << IGC_MSIX_OTHER_INTR_VEC);
385 	}
386 
387 	IGC_WRITE_REG(hw, IGC_IMC, ~0);
388 	IGC_WRITE_FLUSH(hw);
389 }
390 
391 /*
392  * enable other interrupt
393  */
394 static inline void
395 igc_intr_other_enable(struct rte_eth_dev *dev)
396 {
397 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
398 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
399 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
400 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
401 
402 	if (rte_intr_allow_others(intr_handle) &&
403 		dev->data->dev_conf.intr_conf.lsc) {
404 		IGC_WRITE_REG(hw, IGC_EIMS, 1u << IGC_MSIX_OTHER_INTR_VEC);
405 	}
406 
407 	IGC_WRITE_REG(hw, IGC_IMS, intr->mask);
408 	IGC_WRITE_FLUSH(hw);
409 }
410 
411 /*
412  * It reads ICR and gets interrupt causes, check it and set a bit flag
413  * to update link status.
414  */
415 static void
416 eth_igc_interrupt_get_status(struct rte_eth_dev *dev)
417 {
418 	uint32_t icr;
419 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
420 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
421 
422 	/* read-on-clear nic registers here */
423 	icr = IGC_READ_REG(hw, IGC_ICR);
424 
425 	intr->flags = 0;
426 	if (icr & IGC_ICR_LSC)
427 		intr->flags |= IGC_FLAG_NEED_LINK_UPDATE;
428 }
429 
430 /* return 0 means link status changed, -1 means not changed */
431 static int
432 eth_igc_link_update(struct rte_eth_dev *dev, int wait_to_complete)
433 {
434 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
435 	struct rte_eth_link link;
436 	int link_check, count;
437 
438 	link_check = 0;
439 	hw->mac.get_link_status = 1;
440 
441 	/* possible wait-to-complete in up to 9 seconds */
442 	for (count = 0; count < IGC_LINK_UPDATE_CHECK_TIMEOUT; count++) {
443 		/* Read the real link status */
444 		switch (hw->phy.media_type) {
445 		case igc_media_type_copper:
446 			/* Do the work to read phy */
447 			igc_check_for_link(hw);
448 			link_check = !hw->mac.get_link_status;
449 			break;
450 
451 		case igc_media_type_fiber:
452 			igc_check_for_link(hw);
453 			link_check = (IGC_READ_REG(hw, IGC_STATUS) &
454 				      IGC_STATUS_LU);
455 			break;
456 
457 		case igc_media_type_internal_serdes:
458 			igc_check_for_link(hw);
459 			link_check = hw->mac.serdes_has_link;
460 			break;
461 
462 		default:
463 			break;
464 		}
465 		if (link_check || wait_to_complete == 0)
466 			break;
467 		rte_delay_ms(IGC_LINK_UPDATE_CHECK_INTERVAL);
468 	}
469 	memset(&link, 0, sizeof(link));
470 
471 	/* Now we check if a transition has happened */
472 	if (link_check) {
473 		uint16_t duplex, speed;
474 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
475 		link.link_duplex = (duplex == FULL_DUPLEX) ?
476 				RTE_ETH_LINK_FULL_DUPLEX :
477 				RTE_ETH_LINK_HALF_DUPLEX;
478 		link.link_speed = speed;
479 		link.link_status = RTE_ETH_LINK_UP;
480 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
481 				RTE_ETH_LINK_SPEED_FIXED);
482 
483 		if (speed == SPEED_2500) {
484 			uint32_t tipg = IGC_READ_REG(hw, IGC_TIPG);
485 			if ((tipg & IGC_TIPG_IPGT_MASK) != 0x0b) {
486 				tipg &= ~IGC_TIPG_IPGT_MASK;
487 				tipg |= 0x0b;
488 				IGC_WRITE_REG(hw, IGC_TIPG, tipg);
489 			}
490 		}
491 	} else {
492 		link.link_speed = 0;
493 		link.link_duplex = RTE_ETH_LINK_HALF_DUPLEX;
494 		link.link_status = RTE_ETH_LINK_DOWN;
495 		link.link_autoneg = RTE_ETH_LINK_FIXED;
496 	}
497 
498 	return rte_eth_linkstatus_set(dev, &link);
499 }
500 
501 /*
502  * It executes link_update after knowing an interrupt is present.
503  */
504 static void
505 eth_igc_interrupt_action(struct rte_eth_dev *dev)
506 {
507 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
508 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
509 	struct rte_eth_link link;
510 	int ret;
511 
512 	if (intr->flags & IGC_FLAG_NEED_LINK_UPDATE) {
513 		intr->flags &= ~IGC_FLAG_NEED_LINK_UPDATE;
514 
515 		/* set get_link_status to check register later */
516 		ret = eth_igc_link_update(dev, 0);
517 
518 		/* check if link has changed */
519 		if (ret < 0)
520 			return;
521 
522 		rte_eth_linkstatus_get(dev, &link);
523 		if (link.link_status)
524 			PMD_DRV_LOG(INFO,
525 				" Port %d: Link Up - speed %u Mbps - %s",
526 				dev->data->port_id,
527 				(unsigned int)link.link_speed,
528 				link.link_duplex == RTE_ETH_LINK_FULL_DUPLEX ?
529 				"full-duplex" : "half-duplex");
530 		else
531 			PMD_DRV_LOG(INFO, " Port %d: Link Down",
532 				dev->data->port_id);
533 
534 		PMD_DRV_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
535 				pci_dev->addr.domain,
536 				pci_dev->addr.bus,
537 				pci_dev->addr.devid,
538 				pci_dev->addr.function);
539 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
540 	}
541 }
542 
543 /*
544  * Interrupt handler which shall be registered at first.
545  *
546  * @handle
547  *  Pointer to interrupt handle.
548  * @param
549  *  The address of parameter (struct rte_eth_dev *) registered before.
550  */
551 static void
552 eth_igc_interrupt_handler(void *param)
553 {
554 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
555 
556 	eth_igc_interrupt_get_status(dev);
557 	eth_igc_interrupt_action(dev);
558 }
559 
560 static void igc_read_queue_stats_register(struct rte_eth_dev *dev);
561 
562 /*
563  * Update the queue status every IGC_ALARM_INTERVAL time.
564  * @param
565  *  The address of parameter (struct rte_eth_dev *) registered before.
566  */
567 static void
568 igc_update_queue_stats_handler(void *param)
569 {
570 	struct rte_eth_dev *dev = param;
571 	igc_read_queue_stats_register(dev);
572 	rte_eal_alarm_set(IGC_ALARM_INTERVAL,
573 			igc_update_queue_stats_handler, dev);
574 }
575 
576 /*
577  * rx,tx enable/disable
578  */
579 static void
580 eth_igc_rxtx_control(struct rte_eth_dev *dev, bool enable)
581 {
582 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
583 	uint32_t tctl, rctl;
584 
585 	tctl = IGC_READ_REG(hw, IGC_TCTL);
586 	rctl = IGC_READ_REG(hw, IGC_RCTL);
587 
588 	if (enable) {
589 		/* enable Tx/Rx */
590 		tctl |= IGC_TCTL_EN;
591 		rctl |= IGC_RCTL_EN;
592 	} else {
593 		/* disable Tx/Rx */
594 		tctl &= ~IGC_TCTL_EN;
595 		rctl &= ~IGC_RCTL_EN;
596 	}
597 	IGC_WRITE_REG(hw, IGC_TCTL, tctl);
598 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
599 	IGC_WRITE_FLUSH(hw);
600 }
601 
602 /*
603  *  This routine disables all traffic on the adapter by issuing a
604  *  global reset on the MAC.
605  */
606 static int
607 eth_igc_stop(struct rte_eth_dev *dev)
608 {
609 	struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
610 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
611 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
612 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
613 	struct rte_eth_link link;
614 
615 	dev->data->dev_started = 0;
616 	adapter->stopped = 1;
617 
618 	/* disable receive and transmit */
619 	eth_igc_rxtx_control(dev, false);
620 
621 	/* disable all MSI-X interrupts */
622 	IGC_WRITE_REG(hw, IGC_EIMC, 0x1f);
623 	IGC_WRITE_FLUSH(hw);
624 
625 	/* clear all MSI-X interrupts */
626 	IGC_WRITE_REG(hw, IGC_EICR, 0x1f);
627 
628 	igc_intr_other_disable(dev);
629 
630 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
631 
632 	/* disable intr eventfd mapping */
633 	rte_intr_disable(intr_handle);
634 
635 	igc_reset_hw(hw);
636 
637 	/* disable all wake up */
638 	IGC_WRITE_REG(hw, IGC_WUC, 0);
639 
640 	/* disable checking EEE operation in MAC loopback mode */
641 	igc_read_reg_check_clear_bits(hw, IGC_EEER, IGC_EEER_EEE_FRC_AN);
642 
643 	/* Set bit for Go Link disconnect */
644 	igc_read_reg_check_set_bits(hw, IGC_82580_PHY_POWER_MGMT,
645 			IGC_82580_PM_GO_LINKD);
646 
647 	/* Power down the phy. Needed to make the link go Down */
648 	eth_igc_set_link_down(dev);
649 
650 	igc_dev_clear_queues(dev);
651 
652 	/* clear the recorded link status */
653 	memset(&link, 0, sizeof(link));
654 	rte_eth_linkstatus_set(dev, &link);
655 
656 	if (!rte_intr_allow_others(intr_handle))
657 		/* resume to the default handler */
658 		rte_intr_callback_register(intr_handle,
659 					   eth_igc_interrupt_handler,
660 					   (void *)dev);
661 
662 	/* Clean datapath event and queue/vec mapping */
663 	rte_intr_efd_disable(intr_handle);
664 	rte_intr_vec_list_free(intr_handle);
665 
666 	return 0;
667 }
668 
669 /*
670  * write interrupt vector allocation register
671  * @hw
672  *  board private structure
673  * @queue_index
674  *  queue index, valid 0,1,2,3
675  * @tx
676  *  tx:1, rx:0
677  * @msix_vector
678  *  msix-vector, valid 0,1,2,3,4
679  */
680 static void
681 igc_write_ivar(struct igc_hw *hw, uint8_t queue_index,
682 		bool tx, uint8_t msix_vector)
683 {
684 	uint8_t offset = 0;
685 	uint8_t reg_index = queue_index >> 1;
686 	uint32_t val;
687 
688 	/*
689 	 * IVAR(0)
690 	 * bit31...24	bit23...16	bit15...8	bit7...0
691 	 * TX1		RX1		TX0		RX0
692 	 *
693 	 * IVAR(1)
694 	 * bit31...24	bit23...16	bit15...8	bit7...0
695 	 * TX3		RX3		TX2		RX2
696 	 */
697 
698 	if (tx)
699 		offset = 8;
700 
701 	if (queue_index & 1)
702 		offset += 16;
703 
704 	val = IGC_READ_REG_ARRAY(hw, IGC_IVAR0, reg_index);
705 
706 	/* clear bits */
707 	val &= ~((uint32_t)0xFF << offset);
708 
709 	/* write vector and valid bit */
710 	val |= (uint32_t)(msix_vector | IGC_IVAR_VALID) << offset;
711 
712 	IGC_WRITE_REG_ARRAY(hw, IGC_IVAR0, reg_index, val);
713 }
714 
715 /* Sets up the hardware to generate MSI-X interrupts properly
716  * @hw
717  *  board private structure
718  */
719 static void
720 igc_configure_msix_intr(struct rte_eth_dev *dev)
721 {
722 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
723 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
724 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
725 
726 	uint32_t intr_mask;
727 	uint32_t vec = IGC_MISC_VEC_ID;
728 	uint32_t base = IGC_MISC_VEC_ID;
729 	uint32_t misc_shift = 0;
730 	int i;
731 
732 	/* won't configure msix register if no mapping is done
733 	 * between intr vector and event fd
734 	 */
735 	if (!rte_intr_dp_is_en(intr_handle))
736 		return;
737 
738 	if (rte_intr_allow_others(intr_handle)) {
739 		base = IGC_RX_VEC_START;
740 		vec = base;
741 		misc_shift = 1;
742 	}
743 
744 	/* turn on MSI-X capability first */
745 	IGC_WRITE_REG(hw, IGC_GPIE, IGC_GPIE_MSIX_MODE |
746 				IGC_GPIE_PBA | IGC_GPIE_EIAME |
747 				IGC_GPIE_NSICR);
748 	intr_mask = RTE_LEN2MASK(rte_intr_nb_efd_get(intr_handle),
749 				 uint32_t) << misc_shift;
750 
751 	if (dev->data->dev_conf.intr_conf.lsc)
752 		intr_mask |= (1u << IGC_MSIX_OTHER_INTR_VEC);
753 
754 	/* enable msix auto-clear */
755 	igc_read_reg_check_set_bits(hw, IGC_EIAC, intr_mask);
756 
757 	/* set other cause interrupt vector */
758 	igc_read_reg_check_set_bits(hw, IGC_IVAR_MISC,
759 		(uint32_t)(IGC_MSIX_OTHER_INTR_VEC | IGC_IVAR_VALID) << 8);
760 
761 	/* enable auto-mask */
762 	igc_read_reg_check_set_bits(hw, IGC_EIAM, intr_mask);
763 
764 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
765 		igc_write_ivar(hw, i, 0, vec);
766 		rte_intr_vec_list_index_set(intr_handle, i, vec);
767 		if (vec < base + rte_intr_nb_efd_get(intr_handle) - 1)
768 			vec++;
769 	}
770 
771 	IGC_WRITE_FLUSH(hw);
772 }
773 
774 /**
775  * It enables the interrupt mask and then enable the interrupt.
776  *
777  * @dev
778  *  Pointer to struct rte_eth_dev.
779  * @on
780  *  Enable or Disable
781  */
782 static void
783 igc_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
784 {
785 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
786 
787 	if (on)
788 		intr->mask |= IGC_ICR_LSC;
789 	else
790 		intr->mask &= ~IGC_ICR_LSC;
791 }
792 
793 /*
794  * It enables the interrupt.
795  * It will be called once only during nic initialized.
796  */
797 static void
798 igc_rxq_interrupt_setup(struct rte_eth_dev *dev)
799 {
800 	uint32_t mask;
801 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
802 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
803 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
804 	int misc_shift = rte_intr_allow_others(intr_handle) ? 1 : 0;
805 
806 	/* won't configure msix register if no mapping is done
807 	 * between intr vector and event fd
808 	 */
809 	if (!rte_intr_dp_is_en(intr_handle))
810 		return;
811 
812 	mask = RTE_LEN2MASK(rte_intr_nb_efd_get(intr_handle), uint32_t)
813 		<< misc_shift;
814 	IGC_WRITE_REG(hw, IGC_EIMS, mask);
815 }
816 
817 /*
818  *  Get hardware rx-buffer size.
819  */
820 static inline int
821 igc_get_rx_buffer_size(struct igc_hw *hw)
822 {
823 	return (IGC_READ_REG(hw, IGC_RXPBS) & 0x3f) << 10;
824 }
825 
826 /*
827  * igc_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
828  * For ASF and Pass Through versions of f/w this means
829  * that the driver is loaded.
830  */
831 static void
832 igc_hw_control_acquire(struct igc_hw *hw)
833 {
834 	uint32_t ctrl_ext;
835 
836 	/* Let firmware know the driver has taken over */
837 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
838 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext | IGC_CTRL_EXT_DRV_LOAD);
839 }
840 
841 /*
842  * igc_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
843  * For ASF and Pass Through versions of f/w this means that the
844  * driver is no longer loaded.
845  */
846 static void
847 igc_hw_control_release(struct igc_hw *hw)
848 {
849 	uint32_t ctrl_ext;
850 
851 	/* Let firmware taken over control of h/w */
852 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
853 	IGC_WRITE_REG(hw, IGC_CTRL_EXT,
854 			ctrl_ext & ~IGC_CTRL_EXT_DRV_LOAD);
855 }
856 
857 static int
858 igc_hardware_init(struct igc_hw *hw)
859 {
860 	uint32_t rx_buf_size;
861 	int diag;
862 
863 	/* Let the firmware know the OS is in control */
864 	igc_hw_control_acquire(hw);
865 
866 	/* Issue a global reset */
867 	igc_reset_hw(hw);
868 
869 	/* disable all wake up */
870 	IGC_WRITE_REG(hw, IGC_WUC, 0);
871 
872 	/*
873 	 * Hardware flow control
874 	 * - High water mark should allow for at least two standard size (1518)
875 	 *   frames to be received after sending an XOFF.
876 	 * - Low water mark works best when it is very near the high water mark.
877 	 *   This allows the receiver to restart by sending XON when it has
878 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
879 	 *   restart after one full frame is pulled from the buffer. There
880 	 *   could be several smaller frames in the buffer and if so they will
881 	 *   not trigger the XON until their total number reduces the buffer
882 	 *   by 1500.
883 	 */
884 	rx_buf_size = igc_get_rx_buffer_size(hw);
885 	hw->fc.high_water = rx_buf_size - (RTE_ETHER_MAX_LEN * 2);
886 	hw->fc.low_water = hw->fc.high_water - 1500;
887 	hw->fc.pause_time = IGC_FC_PAUSE_TIME;
888 	hw->fc.send_xon = 1;
889 	hw->fc.requested_mode = igc_fc_full;
890 
891 	diag = igc_init_hw(hw);
892 	if (diag < 0)
893 		return diag;
894 
895 	igc_get_phy_info(hw);
896 	igc_check_for_link(hw);
897 
898 	return 0;
899 }
900 
901 static int
902 eth_igc_start(struct rte_eth_dev *dev)
903 {
904 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
905 	struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
906 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
907 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
908 	uint32_t *speeds;
909 	int ret;
910 
911 	PMD_INIT_FUNC_TRACE();
912 
913 	/* disable all MSI-X interrupts */
914 	IGC_WRITE_REG(hw, IGC_EIMC, 0x1f);
915 	IGC_WRITE_FLUSH(hw);
916 
917 	/* clear all MSI-X interrupts */
918 	IGC_WRITE_REG(hw, IGC_EICR, 0x1f);
919 
920 	/* disable uio/vfio intr/eventfd mapping */
921 	if (!adapter->stopped)
922 		rte_intr_disable(intr_handle);
923 
924 	/* Power up the phy. Needed to make the link go Up */
925 	eth_igc_set_link_up(dev);
926 
927 	/* Put the address into the Receive Address Array */
928 	igc_rar_set(hw, hw->mac.addr, 0);
929 
930 	/* Initialize the hardware */
931 	if (igc_hardware_init(hw)) {
932 		PMD_DRV_LOG(ERR, "Unable to initialize the hardware");
933 		return -EIO;
934 	}
935 	adapter->stopped = 0;
936 
937 	/* check and configure queue intr-vector mapping */
938 	if (rte_intr_cap_multiple(intr_handle) &&
939 		dev->data->dev_conf.intr_conf.rxq) {
940 		uint32_t intr_vector = dev->data->nb_rx_queues;
941 		if (rte_intr_efd_enable(intr_handle, intr_vector))
942 			return -1;
943 	}
944 
945 	if (rte_intr_dp_is_en(intr_handle)) {
946 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
947 						  dev->data->nb_rx_queues)) {
948 			PMD_DRV_LOG(ERR,
949 				"Failed to allocate %d rx_queues intr_vec",
950 				dev->data->nb_rx_queues);
951 			return -ENOMEM;
952 		}
953 	}
954 
955 	/* configure msix for rx interrupt */
956 	igc_configure_msix_intr(dev);
957 
958 	igc_tx_init(dev);
959 
960 	/* This can fail when allocating mbufs for descriptor rings */
961 	ret = igc_rx_init(dev);
962 	if (ret) {
963 		PMD_DRV_LOG(ERR, "Unable to initialize RX hardware");
964 		igc_dev_clear_queues(dev);
965 		return ret;
966 	}
967 
968 	igc_clear_hw_cntrs_base_generic(hw);
969 
970 	/* VLAN Offload Settings */
971 	eth_igc_vlan_offload_set(dev,
972 		RTE_ETH_VLAN_STRIP_MASK | RTE_ETH_VLAN_FILTER_MASK |
973 		RTE_ETH_VLAN_EXTEND_MASK);
974 
975 	/* Setup link speed and duplex */
976 	speeds = &dev->data->dev_conf.link_speeds;
977 	if (*speeds == RTE_ETH_LINK_SPEED_AUTONEG) {
978 		hw->phy.autoneg_advertised = IGC_ALL_SPEED_DUPLEX_2500;
979 		hw->mac.autoneg = 1;
980 	} else {
981 		int num_speeds = 0;
982 
983 		if (*speeds & RTE_ETH_LINK_SPEED_FIXED) {
984 			PMD_DRV_LOG(ERR,
985 				    "Force speed mode currently not supported");
986 			igc_dev_clear_queues(dev);
987 			return -EINVAL;
988 		}
989 
990 		hw->phy.autoneg_advertised = 0;
991 		hw->mac.autoneg = 1;
992 
993 		if (*speeds & ~(RTE_ETH_LINK_SPEED_10M_HD | RTE_ETH_LINK_SPEED_10M |
994 				RTE_ETH_LINK_SPEED_100M_HD | RTE_ETH_LINK_SPEED_100M |
995 				RTE_ETH_LINK_SPEED_1G | RTE_ETH_LINK_SPEED_2_5G)) {
996 			num_speeds = -1;
997 			goto error_invalid_config;
998 		}
999 		if (*speeds & RTE_ETH_LINK_SPEED_10M_HD) {
1000 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1001 			num_speeds++;
1002 		}
1003 		if (*speeds & RTE_ETH_LINK_SPEED_10M) {
1004 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1005 			num_speeds++;
1006 		}
1007 		if (*speeds & RTE_ETH_LINK_SPEED_100M_HD) {
1008 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1009 			num_speeds++;
1010 		}
1011 		if (*speeds & RTE_ETH_LINK_SPEED_100M) {
1012 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1013 			num_speeds++;
1014 		}
1015 		if (*speeds & RTE_ETH_LINK_SPEED_1G) {
1016 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1017 			num_speeds++;
1018 		}
1019 		if (*speeds & RTE_ETH_LINK_SPEED_2_5G) {
1020 			hw->phy.autoneg_advertised |= ADVERTISE_2500_FULL;
1021 			num_speeds++;
1022 		}
1023 		if (num_speeds == 0)
1024 			goto error_invalid_config;
1025 	}
1026 
1027 	igc_setup_link(hw);
1028 
1029 	if (rte_intr_allow_others(intr_handle)) {
1030 		/* check if lsc interrupt is enabled */
1031 		if (dev->data->dev_conf.intr_conf.lsc)
1032 			igc_lsc_interrupt_setup(dev, 1);
1033 		else
1034 			igc_lsc_interrupt_setup(dev, 0);
1035 	} else {
1036 		rte_intr_callback_unregister(intr_handle,
1037 					     eth_igc_interrupt_handler,
1038 					     (void *)dev);
1039 		if (dev->data->dev_conf.intr_conf.lsc)
1040 			PMD_DRV_LOG(INFO,
1041 				"LSC won't enable because of no intr multiplex");
1042 	}
1043 
1044 	/* enable uio/vfio intr/eventfd mapping */
1045 	rte_intr_enable(intr_handle);
1046 
1047 	rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1048 			igc_update_queue_stats_handler, dev);
1049 
1050 	/* check if rxq interrupt is enabled */
1051 	if (dev->data->dev_conf.intr_conf.rxq &&
1052 			rte_intr_dp_is_en(intr_handle))
1053 		igc_rxq_interrupt_setup(dev);
1054 
1055 	/* resume enabled intr since hw reset */
1056 	igc_intr_other_enable(dev);
1057 
1058 	eth_igc_rxtx_control(dev, true);
1059 	eth_igc_link_update(dev, 0);
1060 
1061 	/* configure MAC-loopback mode */
1062 	if (dev->data->dev_conf.lpbk_mode == 1) {
1063 		uint32_t reg_val;
1064 
1065 		reg_val = IGC_READ_REG(hw, IGC_CTRL);
1066 		reg_val &= ~IGC_CTRL_SPEED_MASK;
1067 		reg_val |= IGC_CTRL_SLU | IGC_CTRL_FRCSPD |
1068 			IGC_CTRL_FRCDPX | IGC_CTRL_FD | IGC_CTRL_SPEED_2500;
1069 		IGC_WRITE_REG(hw, IGC_CTRL, reg_val);
1070 
1071 		igc_read_reg_check_set_bits(hw, IGC_EEER, IGC_EEER_EEE_FRC_AN);
1072 	}
1073 
1074 	return 0;
1075 
1076 error_invalid_config:
1077 	PMD_DRV_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
1078 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
1079 	igc_dev_clear_queues(dev);
1080 	return -EINVAL;
1081 }
1082 
1083 static int
1084 igc_reset_swfw_lock(struct igc_hw *hw)
1085 {
1086 	int ret_val;
1087 
1088 	/*
1089 	 * Do mac ops initialization manually here, since we will need
1090 	 * some function pointers set by this call.
1091 	 */
1092 	ret_val = igc_init_mac_params(hw);
1093 	if (ret_val)
1094 		return ret_val;
1095 
1096 	/*
1097 	 * SMBI lock should not fail in this early stage. If this is the case,
1098 	 * it is due to an improper exit of the application.
1099 	 * So force the release of the faulty lock.
1100 	 */
1101 	if (igc_get_hw_semaphore_generic(hw) < 0)
1102 		PMD_DRV_LOG(DEBUG, "SMBI lock released");
1103 
1104 	igc_put_hw_semaphore_generic(hw);
1105 
1106 	if (hw->mac.ops.acquire_swfw_sync != NULL) {
1107 		uint16_t mask;
1108 
1109 		/*
1110 		 * Phy lock should not fail in this early stage.
1111 		 * If this is the case, it is due to an improper exit of the
1112 		 * application. So force the release of the faulty lock.
1113 		 */
1114 		mask = IGC_SWFW_PHY0_SM;
1115 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
1116 			PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released",
1117 				    hw->bus.func);
1118 		}
1119 		hw->mac.ops.release_swfw_sync(hw, mask);
1120 
1121 		/*
1122 		 * This one is more tricky since it is common to all ports; but
1123 		 * swfw_sync retries last long enough (1s) to be almost sure
1124 		 * that if lock can not be taken it is due to an improper lock
1125 		 * of the semaphore.
1126 		 */
1127 		mask = IGC_SWFW_EEP_SM;
1128 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0)
1129 			PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1130 
1131 		hw->mac.ops.release_swfw_sync(hw, mask);
1132 	}
1133 
1134 	return IGC_SUCCESS;
1135 }
1136 
1137 /*
1138  * free all rx/tx queues.
1139  */
1140 static void
1141 igc_dev_free_queues(struct rte_eth_dev *dev)
1142 {
1143 	uint16_t i;
1144 
1145 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
1146 		eth_igc_rx_queue_release(dev, i);
1147 		dev->data->rx_queues[i] = NULL;
1148 	}
1149 	dev->data->nb_rx_queues = 0;
1150 
1151 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
1152 		eth_igc_tx_queue_release(dev, i);
1153 		dev->data->tx_queues[i] = NULL;
1154 	}
1155 	dev->data->nb_tx_queues = 0;
1156 }
1157 
1158 static int
1159 eth_igc_close(struct rte_eth_dev *dev)
1160 {
1161 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1162 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1163 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1164 	struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
1165 	int retry = 0;
1166 	int ret = 0;
1167 
1168 	PMD_INIT_FUNC_TRACE();
1169 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1170 		return 0;
1171 
1172 	if (!adapter->stopped)
1173 		ret = eth_igc_stop(dev);
1174 
1175 	igc_flow_flush(dev, NULL);
1176 	igc_clear_all_filter(dev);
1177 
1178 	igc_intr_other_disable(dev);
1179 	do {
1180 		int ret = rte_intr_callback_unregister(intr_handle,
1181 				eth_igc_interrupt_handler, dev);
1182 		if (ret >= 0 || ret == -ENOENT || ret == -EINVAL)
1183 			break;
1184 
1185 		PMD_DRV_LOG(ERR, "intr callback unregister failed: %d", ret);
1186 		DELAY(200 * 1000); /* delay 200ms */
1187 	} while (retry++ < 5);
1188 
1189 	igc_phy_hw_reset(hw);
1190 	igc_hw_control_release(hw);
1191 	igc_dev_free_queues(dev);
1192 
1193 	/* Reset any pending lock */
1194 	igc_reset_swfw_lock(hw);
1195 
1196 	return ret;
1197 }
1198 
1199 static void
1200 igc_identify_hardware(struct rte_eth_dev *dev, struct rte_pci_device *pci_dev)
1201 {
1202 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1203 
1204 	hw->vendor_id = pci_dev->id.vendor_id;
1205 	hw->device_id = pci_dev->id.device_id;
1206 	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
1207 	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
1208 }
1209 
1210 static int
1211 eth_igc_dev_init(struct rte_eth_dev *dev)
1212 {
1213 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1214 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
1215 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1216 	int i, error = 0;
1217 
1218 	PMD_INIT_FUNC_TRACE();
1219 	dev->dev_ops = &eth_igc_ops;
1220 	dev->rx_queue_count = eth_igc_rx_queue_count;
1221 	dev->rx_descriptor_status = eth_igc_rx_descriptor_status;
1222 	dev->tx_descriptor_status = eth_igc_tx_descriptor_status;
1223 
1224 	/*
1225 	 * for secondary processes, we don't initialize any further as primary
1226 	 * has already done this work. Only check we don't need a different
1227 	 * RX function.
1228 	 */
1229 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1230 		return 0;
1231 
1232 	rte_eth_copy_pci_info(dev, pci_dev);
1233 	dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1234 
1235 	hw->back = pci_dev;
1236 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1237 
1238 	igc_identify_hardware(dev, pci_dev);
1239 	if (igc_setup_init_funcs(hw, false) != IGC_SUCCESS) {
1240 		error = -EIO;
1241 		goto err_late;
1242 	}
1243 
1244 	igc_get_bus_info(hw);
1245 
1246 	/* Reset any pending lock */
1247 	if (igc_reset_swfw_lock(hw) != IGC_SUCCESS) {
1248 		error = -EIO;
1249 		goto err_late;
1250 	}
1251 
1252 	/* Finish initialization */
1253 	if (igc_setup_init_funcs(hw, true) != IGC_SUCCESS) {
1254 		error = -EIO;
1255 		goto err_late;
1256 	}
1257 
1258 	hw->mac.autoneg = 1;
1259 	hw->phy.autoneg_wait_to_complete = 0;
1260 	hw->phy.autoneg_advertised = IGC_ALL_SPEED_DUPLEX_2500;
1261 
1262 	/* Copper options */
1263 	if (hw->phy.media_type == igc_media_type_copper) {
1264 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
1265 		hw->phy.disable_polarity_correction = 0;
1266 		hw->phy.ms_type = igc_ms_hw_default;
1267 	}
1268 
1269 	/*
1270 	 * Start from a known state, this is important in reading the nvm
1271 	 * and mac from that.
1272 	 */
1273 	igc_reset_hw(hw);
1274 
1275 	/* Make sure we have a good EEPROM before we read from it */
1276 	if (igc_validate_nvm_checksum(hw) < 0) {
1277 		/*
1278 		 * Some PCI-E parts fail the first check due to
1279 		 * the link being in sleep state, call it again,
1280 		 * if it fails a second time its a real issue.
1281 		 */
1282 		if (igc_validate_nvm_checksum(hw) < 0) {
1283 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
1284 			error = -EIO;
1285 			goto err_late;
1286 		}
1287 	}
1288 
1289 	/* Read the permanent MAC address out of the EEPROM */
1290 	if (igc_read_mac_addr(hw) != 0) {
1291 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
1292 		error = -EIO;
1293 		goto err_late;
1294 	}
1295 
1296 	/* Allocate memory for storing MAC addresses */
1297 	dev->data->mac_addrs = rte_zmalloc("igc",
1298 		RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count, 0);
1299 	if (dev->data->mac_addrs == NULL) {
1300 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes for storing MAC",
1301 				RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
1302 		error = -ENOMEM;
1303 		goto err_late;
1304 	}
1305 
1306 	/* Copy the permanent MAC address */
1307 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
1308 			&dev->data->mac_addrs[0]);
1309 
1310 	/* Now initialize the hardware */
1311 	if (igc_hardware_init(hw) != 0) {
1312 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
1313 		rte_free(dev->data->mac_addrs);
1314 		dev->data->mac_addrs = NULL;
1315 		error = -ENODEV;
1316 		goto err_late;
1317 	}
1318 
1319 	hw->mac.get_link_status = 1;
1320 	igc->stopped = 0;
1321 
1322 	/* Indicate SOL/IDER usage */
1323 	if (igc_check_reset_block(hw) < 0)
1324 		PMD_INIT_LOG(ERR,
1325 			"PHY reset is blocked due to SOL/IDER session.");
1326 
1327 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
1328 			dev->data->port_id, pci_dev->id.vendor_id,
1329 			pci_dev->id.device_id);
1330 
1331 	rte_intr_callback_register(pci_dev->intr_handle,
1332 			eth_igc_interrupt_handler, (void *)dev);
1333 
1334 	/* enable uio/vfio intr/eventfd mapping */
1335 	rte_intr_enable(pci_dev->intr_handle);
1336 
1337 	/* enable support intr */
1338 	igc_intr_other_enable(dev);
1339 
1340 	/* initiate queue status */
1341 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1342 		igc->txq_stats_map[i] = -1;
1343 		igc->rxq_stats_map[i] = -1;
1344 	}
1345 
1346 	igc_flow_init(dev);
1347 	igc_clear_all_filter(dev);
1348 	return 0;
1349 
1350 err_late:
1351 	igc_hw_control_release(hw);
1352 	return error;
1353 }
1354 
1355 static int
1356 eth_igc_dev_uninit(__rte_unused struct rte_eth_dev *eth_dev)
1357 {
1358 	PMD_INIT_FUNC_TRACE();
1359 	eth_igc_close(eth_dev);
1360 	return 0;
1361 }
1362 
1363 static int
1364 eth_igc_reset(struct rte_eth_dev *dev)
1365 {
1366 	int ret;
1367 
1368 	PMD_INIT_FUNC_TRACE();
1369 
1370 	ret = eth_igc_dev_uninit(dev);
1371 	if (ret)
1372 		return ret;
1373 
1374 	return eth_igc_dev_init(dev);
1375 }
1376 
1377 static int
1378 eth_igc_promiscuous_enable(struct rte_eth_dev *dev)
1379 {
1380 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1381 	uint32_t rctl;
1382 
1383 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1384 	rctl |= (IGC_RCTL_UPE | IGC_RCTL_MPE);
1385 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1386 	return 0;
1387 }
1388 
1389 static int
1390 eth_igc_promiscuous_disable(struct rte_eth_dev *dev)
1391 {
1392 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1393 	uint32_t rctl;
1394 
1395 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1396 	rctl &= (~IGC_RCTL_UPE);
1397 	if (dev->data->all_multicast == 1)
1398 		rctl |= IGC_RCTL_MPE;
1399 	else
1400 		rctl &= (~IGC_RCTL_MPE);
1401 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1402 	return 0;
1403 }
1404 
1405 static int
1406 eth_igc_allmulticast_enable(struct rte_eth_dev *dev)
1407 {
1408 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1409 	uint32_t rctl;
1410 
1411 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1412 	rctl |= IGC_RCTL_MPE;
1413 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1414 	return 0;
1415 }
1416 
1417 static int
1418 eth_igc_allmulticast_disable(struct rte_eth_dev *dev)
1419 {
1420 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1421 	uint32_t rctl;
1422 
1423 	if (dev->data->promiscuous == 1)
1424 		return 0;	/* must remain in all_multicast mode */
1425 
1426 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1427 	rctl &= (~IGC_RCTL_MPE);
1428 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1429 	return 0;
1430 }
1431 
1432 static int
1433 eth_igc_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
1434 		       size_t fw_size)
1435 {
1436 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1437 	struct igc_fw_version fw;
1438 	int ret;
1439 
1440 	igc_get_fw_version(hw, &fw);
1441 
1442 	/* if option rom is valid, display its version too */
1443 	if (fw.or_valid) {
1444 		ret = snprintf(fw_version, fw_size,
1445 			 "%d.%d, 0x%08x, %d.%d.%d",
1446 			 fw.eep_major, fw.eep_minor, fw.etrack_id,
1447 			 fw.or_major, fw.or_build, fw.or_patch);
1448 	/* no option rom */
1449 	} else {
1450 		if (fw.etrack_id != 0X0000) {
1451 			ret = snprintf(fw_version, fw_size,
1452 				 "%d.%d, 0x%08x",
1453 				 fw.eep_major, fw.eep_minor,
1454 				 fw.etrack_id);
1455 		} else {
1456 			ret = snprintf(fw_version, fw_size,
1457 				 "%d.%d.%d",
1458 				 fw.eep_major, fw.eep_minor,
1459 				 fw.eep_build);
1460 		}
1461 	}
1462 	if (ret < 0)
1463 		return -EINVAL;
1464 
1465 	ret += 1; /* add the size of '\0' */
1466 	if (fw_size < (size_t)ret)
1467 		return ret;
1468 	else
1469 		return 0;
1470 }
1471 
1472 static int
1473 eth_igc_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1474 {
1475 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1476 
1477 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1478 	dev_info->max_rx_pktlen = MAX_RX_JUMBO_FRAME_SIZE;
1479 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1480 	dev_info->rx_offload_capa = IGC_RX_OFFLOAD_ALL;
1481 	dev_info->tx_offload_capa = IGC_TX_OFFLOAD_ALL;
1482 	dev_info->rx_queue_offload_capa = RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
1483 
1484 	dev_info->max_rx_queues = IGC_QUEUE_PAIRS_NUM;
1485 	dev_info->max_tx_queues = IGC_QUEUE_PAIRS_NUM;
1486 	dev_info->max_vmdq_pools = 0;
1487 
1488 	dev_info->hash_key_size = IGC_HKEY_MAX_INDEX * sizeof(uint32_t);
1489 	dev_info->reta_size = RTE_ETH_RSS_RETA_SIZE_128;
1490 	dev_info->flow_type_rss_offloads = IGC_RSS_OFFLOAD_ALL;
1491 
1492 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
1493 		.rx_thresh = {
1494 			.pthresh = IGC_DEFAULT_RX_PTHRESH,
1495 			.hthresh = IGC_DEFAULT_RX_HTHRESH,
1496 			.wthresh = IGC_DEFAULT_RX_WTHRESH,
1497 		},
1498 		.rx_free_thresh = IGC_DEFAULT_RX_FREE_THRESH,
1499 		.rx_drop_en = 0,
1500 		.offloads = 0,
1501 	};
1502 
1503 	dev_info->default_txconf = (struct rte_eth_txconf) {
1504 		.tx_thresh = {
1505 			.pthresh = IGC_DEFAULT_TX_PTHRESH,
1506 			.hthresh = IGC_DEFAULT_TX_HTHRESH,
1507 			.wthresh = IGC_DEFAULT_TX_WTHRESH,
1508 		},
1509 		.offloads = 0,
1510 	};
1511 
1512 	dev_info->rx_desc_lim = rx_desc_lim;
1513 	dev_info->tx_desc_lim = tx_desc_lim;
1514 
1515 	dev_info->speed_capa = RTE_ETH_LINK_SPEED_10M_HD | RTE_ETH_LINK_SPEED_10M |
1516 			RTE_ETH_LINK_SPEED_100M_HD | RTE_ETH_LINK_SPEED_100M |
1517 			RTE_ETH_LINK_SPEED_1G | RTE_ETH_LINK_SPEED_2_5G;
1518 
1519 	dev_info->max_mtu = dev_info->max_rx_pktlen - IGC_ETH_OVERHEAD;
1520 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
1521 	return 0;
1522 }
1523 
1524 static int
1525 eth_igc_led_on(struct rte_eth_dev *dev)
1526 {
1527 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1528 
1529 	return igc_led_on(hw) == IGC_SUCCESS ? 0 : -ENOTSUP;
1530 }
1531 
1532 static int
1533 eth_igc_led_off(struct rte_eth_dev *dev)
1534 {
1535 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1536 
1537 	return igc_led_off(hw) == IGC_SUCCESS ? 0 : -ENOTSUP;
1538 }
1539 
1540 static const uint32_t *
1541 eth_igc_supported_ptypes_get(__rte_unused struct rte_eth_dev *dev)
1542 {
1543 	static const uint32_t ptypes[] = {
1544 		/* refers to rx_desc_pkt_info_to_pkt_type() */
1545 		RTE_PTYPE_L2_ETHER,
1546 		RTE_PTYPE_L3_IPV4,
1547 		RTE_PTYPE_L3_IPV4_EXT,
1548 		RTE_PTYPE_L3_IPV6,
1549 		RTE_PTYPE_L3_IPV6_EXT,
1550 		RTE_PTYPE_L4_TCP,
1551 		RTE_PTYPE_L4_UDP,
1552 		RTE_PTYPE_L4_SCTP,
1553 		RTE_PTYPE_TUNNEL_IP,
1554 		RTE_PTYPE_INNER_L3_IPV6,
1555 		RTE_PTYPE_INNER_L3_IPV6_EXT,
1556 		RTE_PTYPE_INNER_L4_TCP,
1557 		RTE_PTYPE_INNER_L4_UDP,
1558 		RTE_PTYPE_UNKNOWN
1559 	};
1560 
1561 	return ptypes;
1562 }
1563 
1564 static int
1565 eth_igc_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1566 {
1567 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1568 	uint32_t frame_size = mtu + IGC_ETH_OVERHEAD;
1569 	uint32_t rctl;
1570 
1571 	/* if extend vlan has been enabled */
1572 	if (IGC_READ_REG(hw, IGC_CTRL_EXT) & IGC_CTRL_EXT_EXT_VLAN)
1573 		frame_size += VLAN_TAG_SIZE;
1574 
1575 	/*
1576 	 * If device is started, refuse mtu that requires the support of
1577 	 * scattered packets when this feature has not been enabled before.
1578 	 */
1579 	if (dev->data->dev_started && !dev->data->scattered_rx &&
1580 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
1581 		PMD_INIT_LOG(ERR, "Stop port first.");
1582 		return -EINVAL;
1583 	}
1584 
1585 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1586 	if (mtu > RTE_ETHER_MTU)
1587 		rctl |= IGC_RCTL_LPE;
1588 	else
1589 		rctl &= ~IGC_RCTL_LPE;
1590 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1591 
1592 	IGC_WRITE_REG(hw, IGC_RLPML, frame_size);
1593 
1594 	return 0;
1595 }
1596 
1597 static int
1598 eth_igc_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
1599 		uint32_t index, uint32_t pool)
1600 {
1601 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1602 
1603 	igc_rar_set(hw, mac_addr->addr_bytes, index);
1604 	RTE_SET_USED(pool);
1605 	return 0;
1606 }
1607 
1608 static void
1609 eth_igc_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1610 {
1611 	uint8_t addr[RTE_ETHER_ADDR_LEN];
1612 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1613 
1614 	memset(addr, 0, sizeof(addr));
1615 	igc_rar_set(hw, addr, index);
1616 }
1617 
1618 static int
1619 eth_igc_default_mac_addr_set(struct rte_eth_dev *dev,
1620 			struct rte_ether_addr *addr)
1621 {
1622 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1623 	igc_rar_set(hw, addr->addr_bytes, 0);
1624 	return 0;
1625 }
1626 
1627 static int
1628 eth_igc_set_mc_addr_list(struct rte_eth_dev *dev,
1629 			 struct rte_ether_addr *mc_addr_set,
1630 			 uint32_t nb_mc_addr)
1631 {
1632 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1633 	igc_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1634 	return 0;
1635 }
1636 
1637 /*
1638  * Read hardware registers
1639  */
1640 static void
1641 igc_read_stats_registers(struct igc_hw *hw, struct igc_hw_stats *stats)
1642 {
1643 	int pause_frames;
1644 
1645 	uint64_t old_gprc  = stats->gprc;
1646 	uint64_t old_gptc  = stats->gptc;
1647 	uint64_t old_tpr   = stats->tpr;
1648 	uint64_t old_tpt   = stats->tpt;
1649 	uint64_t old_rpthc = stats->rpthc;
1650 	uint64_t old_hgptc = stats->hgptc;
1651 
1652 	stats->crcerrs += IGC_READ_REG(hw, IGC_CRCERRS);
1653 	stats->algnerrc += IGC_READ_REG(hw, IGC_ALGNERRC);
1654 	stats->rxerrc += IGC_READ_REG(hw, IGC_RXERRC);
1655 	stats->mpc += IGC_READ_REG(hw, IGC_MPC);
1656 	stats->scc += IGC_READ_REG(hw, IGC_SCC);
1657 	stats->ecol += IGC_READ_REG(hw, IGC_ECOL);
1658 
1659 	stats->mcc += IGC_READ_REG(hw, IGC_MCC);
1660 	stats->latecol += IGC_READ_REG(hw, IGC_LATECOL);
1661 	stats->colc += IGC_READ_REG(hw, IGC_COLC);
1662 
1663 	stats->dc += IGC_READ_REG(hw, IGC_DC);
1664 	stats->tncrs += IGC_READ_REG(hw, IGC_TNCRS);
1665 	stats->htdpmc += IGC_READ_REG(hw, IGC_HTDPMC);
1666 	stats->rlec += IGC_READ_REG(hw, IGC_RLEC);
1667 	stats->xonrxc += IGC_READ_REG(hw, IGC_XONRXC);
1668 	stats->xontxc += IGC_READ_REG(hw, IGC_XONTXC);
1669 
1670 	/*
1671 	 * For watchdog management we need to know if we have been
1672 	 * paused during the last interval, so capture that here.
1673 	 */
1674 	pause_frames = IGC_READ_REG(hw, IGC_XOFFRXC);
1675 	stats->xoffrxc += pause_frames;
1676 	stats->xofftxc += IGC_READ_REG(hw, IGC_XOFFTXC);
1677 	stats->fcruc += IGC_READ_REG(hw, IGC_FCRUC);
1678 	stats->prc64 += IGC_READ_REG(hw, IGC_PRC64);
1679 	stats->prc127 += IGC_READ_REG(hw, IGC_PRC127);
1680 	stats->prc255 += IGC_READ_REG(hw, IGC_PRC255);
1681 	stats->prc511 += IGC_READ_REG(hw, IGC_PRC511);
1682 	stats->prc1023 += IGC_READ_REG(hw, IGC_PRC1023);
1683 	stats->prc1522 += IGC_READ_REG(hw, IGC_PRC1522);
1684 	stats->gprc += IGC_READ_REG(hw, IGC_GPRC);
1685 	stats->bprc += IGC_READ_REG(hw, IGC_BPRC);
1686 	stats->mprc += IGC_READ_REG(hw, IGC_MPRC);
1687 	stats->gptc += IGC_READ_REG(hw, IGC_GPTC);
1688 
1689 	/* For the 64-bit byte counters the low dword must be read first. */
1690 	/* Both registers clear on the read of the high dword */
1691 
1692 	/* Workaround CRC bytes included in size, take away 4 bytes/packet */
1693 	stats->gorc += IGC_READ_REG(hw, IGC_GORCL);
1694 	stats->gorc += ((uint64_t)IGC_READ_REG(hw, IGC_GORCH) << 32);
1695 	stats->gorc -= (stats->gprc - old_gprc) * RTE_ETHER_CRC_LEN;
1696 	stats->gotc += IGC_READ_REG(hw, IGC_GOTCL);
1697 	stats->gotc += ((uint64_t)IGC_READ_REG(hw, IGC_GOTCH) << 32);
1698 	stats->gotc -= (stats->gptc - old_gptc) * RTE_ETHER_CRC_LEN;
1699 
1700 	stats->rnbc += IGC_READ_REG(hw, IGC_RNBC);
1701 	stats->ruc += IGC_READ_REG(hw, IGC_RUC);
1702 	stats->rfc += IGC_READ_REG(hw, IGC_RFC);
1703 	stats->roc += IGC_READ_REG(hw, IGC_ROC);
1704 	stats->rjc += IGC_READ_REG(hw, IGC_RJC);
1705 
1706 	stats->mgprc += IGC_READ_REG(hw, IGC_MGTPRC);
1707 	stats->mgpdc += IGC_READ_REG(hw, IGC_MGTPDC);
1708 	stats->mgptc += IGC_READ_REG(hw, IGC_MGTPTC);
1709 	stats->b2ospc += IGC_READ_REG(hw, IGC_B2OSPC);
1710 	stats->b2ogprc += IGC_READ_REG(hw, IGC_B2OGPRC);
1711 	stats->o2bgptc += IGC_READ_REG(hw, IGC_O2BGPTC);
1712 	stats->o2bspc += IGC_READ_REG(hw, IGC_O2BSPC);
1713 
1714 	stats->tpr += IGC_READ_REG(hw, IGC_TPR);
1715 	stats->tpt += IGC_READ_REG(hw, IGC_TPT);
1716 
1717 	stats->tor += IGC_READ_REG(hw, IGC_TORL);
1718 	stats->tor += ((uint64_t)IGC_READ_REG(hw, IGC_TORH) << 32);
1719 	stats->tor -= (stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
1720 	stats->tot += IGC_READ_REG(hw, IGC_TOTL);
1721 	stats->tot += ((uint64_t)IGC_READ_REG(hw, IGC_TOTH) << 32);
1722 	stats->tot -= (stats->tpt - old_tpt) * RTE_ETHER_CRC_LEN;
1723 
1724 	stats->ptc64 += IGC_READ_REG(hw, IGC_PTC64);
1725 	stats->ptc127 += IGC_READ_REG(hw, IGC_PTC127);
1726 	stats->ptc255 += IGC_READ_REG(hw, IGC_PTC255);
1727 	stats->ptc511 += IGC_READ_REG(hw, IGC_PTC511);
1728 	stats->ptc1023 += IGC_READ_REG(hw, IGC_PTC1023);
1729 	stats->ptc1522 += IGC_READ_REG(hw, IGC_PTC1522);
1730 	stats->mptc += IGC_READ_REG(hw, IGC_MPTC);
1731 	stats->bptc += IGC_READ_REG(hw, IGC_BPTC);
1732 	stats->tsctc += IGC_READ_REG(hw, IGC_TSCTC);
1733 
1734 	stats->iac += IGC_READ_REG(hw, IGC_IAC);
1735 	stats->rpthc += IGC_READ_REG(hw, IGC_RPTHC);
1736 	stats->hgptc += IGC_READ_REG(hw, IGC_HGPTC);
1737 	stats->icrxdmtc += IGC_READ_REG(hw, IGC_ICRXDMTC);
1738 
1739 	/* Host to Card Statistics */
1740 	stats->hgorc += IGC_READ_REG(hw, IGC_HGORCL);
1741 	stats->hgorc += ((uint64_t)IGC_READ_REG(hw, IGC_HGORCH) << 32);
1742 	stats->hgorc -= (stats->rpthc - old_rpthc) * RTE_ETHER_CRC_LEN;
1743 	stats->hgotc += IGC_READ_REG(hw, IGC_HGOTCL);
1744 	stats->hgotc += ((uint64_t)IGC_READ_REG(hw, IGC_HGOTCH) << 32);
1745 	stats->hgotc -= (stats->hgptc - old_hgptc) * RTE_ETHER_CRC_LEN;
1746 	stats->lenerrs += IGC_READ_REG(hw, IGC_LENERRS);
1747 }
1748 
1749 /*
1750  * Write 0 to all queue status registers
1751  */
1752 static void
1753 igc_reset_queue_stats_register(struct igc_hw *hw)
1754 {
1755 	int i;
1756 
1757 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1758 		IGC_WRITE_REG(hw, IGC_PQGPRC(i), 0);
1759 		IGC_WRITE_REG(hw, IGC_PQGPTC(i), 0);
1760 		IGC_WRITE_REG(hw, IGC_PQGORC(i), 0);
1761 		IGC_WRITE_REG(hw, IGC_PQGOTC(i), 0);
1762 		IGC_WRITE_REG(hw, IGC_PQMPRC(i), 0);
1763 		IGC_WRITE_REG(hw, IGC_RQDPC(i), 0);
1764 		IGC_WRITE_REG(hw, IGC_TQDPC(i), 0);
1765 	}
1766 }
1767 
1768 /*
1769  * Read all hardware queue status registers
1770  */
1771 static void
1772 igc_read_queue_stats_register(struct rte_eth_dev *dev)
1773 {
1774 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1775 	struct igc_hw_queue_stats *queue_stats =
1776 				IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1777 	int i;
1778 
1779 	/*
1780 	 * This register is not cleared on read. Furthermore, the register wraps
1781 	 * around back to 0x00000000 on the next increment when reaching a value
1782 	 * of 0xFFFFFFFF and then continues normal count operation.
1783 	 */
1784 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1785 		union {
1786 			u64 ddword;
1787 			u32 dword[2];
1788 		} value;
1789 		u32 tmp;
1790 
1791 		/*
1792 		 * Read the register first, if the value is smaller than that
1793 		 * previous read, that mean the register has been overflowed,
1794 		 * then we add the high 4 bytes by 1 and replace the low 4
1795 		 * bytes by the new value.
1796 		 */
1797 		tmp = IGC_READ_REG(hw, IGC_PQGPRC(i));
1798 		value.ddword = queue_stats->pqgprc[i];
1799 		if (value.dword[U32_0_IN_U64] > tmp)
1800 			value.dword[U32_1_IN_U64]++;
1801 		value.dword[U32_0_IN_U64] = tmp;
1802 		queue_stats->pqgprc[i] = value.ddword;
1803 
1804 		tmp = IGC_READ_REG(hw, IGC_PQGPTC(i));
1805 		value.ddword = queue_stats->pqgptc[i];
1806 		if (value.dword[U32_0_IN_U64] > tmp)
1807 			value.dword[U32_1_IN_U64]++;
1808 		value.dword[U32_0_IN_U64] = tmp;
1809 		queue_stats->pqgptc[i] = value.ddword;
1810 
1811 		tmp = IGC_READ_REG(hw, IGC_PQGORC(i));
1812 		value.ddword = queue_stats->pqgorc[i];
1813 		if (value.dword[U32_0_IN_U64] > tmp)
1814 			value.dword[U32_1_IN_U64]++;
1815 		value.dword[U32_0_IN_U64] = tmp;
1816 		queue_stats->pqgorc[i] = value.ddword;
1817 
1818 		tmp = IGC_READ_REG(hw, IGC_PQGOTC(i));
1819 		value.ddword = queue_stats->pqgotc[i];
1820 		if (value.dword[U32_0_IN_U64] > tmp)
1821 			value.dword[U32_1_IN_U64]++;
1822 		value.dword[U32_0_IN_U64] = tmp;
1823 		queue_stats->pqgotc[i] = value.ddword;
1824 
1825 		tmp = IGC_READ_REG(hw, IGC_PQMPRC(i));
1826 		value.ddword = queue_stats->pqmprc[i];
1827 		if (value.dword[U32_0_IN_U64] > tmp)
1828 			value.dword[U32_1_IN_U64]++;
1829 		value.dword[U32_0_IN_U64] = tmp;
1830 		queue_stats->pqmprc[i] = value.ddword;
1831 
1832 		tmp = IGC_READ_REG(hw, IGC_RQDPC(i));
1833 		value.ddword = queue_stats->rqdpc[i];
1834 		if (value.dword[U32_0_IN_U64] > tmp)
1835 			value.dword[U32_1_IN_U64]++;
1836 		value.dword[U32_0_IN_U64] = tmp;
1837 		queue_stats->rqdpc[i] = value.ddword;
1838 
1839 		tmp = IGC_READ_REG(hw, IGC_TQDPC(i));
1840 		value.ddword = queue_stats->tqdpc[i];
1841 		if (value.dword[U32_0_IN_U64] > tmp)
1842 			value.dword[U32_1_IN_U64]++;
1843 		value.dword[U32_0_IN_U64] = tmp;
1844 		queue_stats->tqdpc[i] = value.ddword;
1845 	}
1846 }
1847 
1848 static int
1849 eth_igc_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1850 {
1851 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
1852 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1853 	struct igc_hw_stats *stats = IGC_DEV_PRIVATE_STATS(dev);
1854 	struct igc_hw_queue_stats *queue_stats =
1855 			IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1856 	int i;
1857 
1858 	/*
1859 	 * Cancel status handler since it will read the queue status registers
1860 	 */
1861 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
1862 
1863 	/* Read status register */
1864 	igc_read_queue_stats_register(dev);
1865 	igc_read_stats_registers(hw, stats);
1866 
1867 	if (rte_stats == NULL) {
1868 		/* Restart queue status handler */
1869 		rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1870 				igc_update_queue_stats_handler, dev);
1871 		return -EINVAL;
1872 	}
1873 
1874 	/* Rx Errors */
1875 	rte_stats->imissed = stats->mpc;
1876 	rte_stats->ierrors = stats->crcerrs + stats->rlec +
1877 			stats->rxerrc + stats->algnerrc;
1878 
1879 	/* Tx Errors */
1880 	rte_stats->oerrors = stats->ecol + stats->latecol;
1881 
1882 	rte_stats->ipackets = stats->gprc;
1883 	rte_stats->opackets = stats->gptc;
1884 	rte_stats->ibytes   = stats->gorc;
1885 	rte_stats->obytes   = stats->gotc;
1886 
1887 	/* Get per-queue statuses */
1888 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1889 		/* GET TX queue statuses */
1890 		int map_id = igc->txq_stats_map[i];
1891 		if (map_id >= 0) {
1892 			rte_stats->q_opackets[map_id] += queue_stats->pqgptc[i];
1893 			rte_stats->q_obytes[map_id] += queue_stats->pqgotc[i];
1894 		}
1895 		/* Get RX queue statuses */
1896 		map_id = igc->rxq_stats_map[i];
1897 		if (map_id >= 0) {
1898 			rte_stats->q_ipackets[map_id] += queue_stats->pqgprc[i];
1899 			rte_stats->q_ibytes[map_id] += queue_stats->pqgorc[i];
1900 			rte_stats->q_errors[map_id] += queue_stats->rqdpc[i];
1901 		}
1902 	}
1903 
1904 	/* Restart queue status handler */
1905 	rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1906 			igc_update_queue_stats_handler, dev);
1907 	return 0;
1908 }
1909 
1910 static int
1911 eth_igc_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1912 		   unsigned int n)
1913 {
1914 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1915 	struct igc_hw_stats *hw_stats =
1916 			IGC_DEV_PRIVATE_STATS(dev);
1917 	unsigned int i;
1918 
1919 	igc_read_stats_registers(hw, hw_stats);
1920 
1921 	if (n < IGC_NB_XSTATS)
1922 		return IGC_NB_XSTATS;
1923 
1924 	/* If this is a reset xstats is NULL, and we have cleared the
1925 	 * registers by reading them.
1926 	 */
1927 	if (!xstats)
1928 		return 0;
1929 
1930 	/* Extended stats */
1931 	for (i = 0; i < IGC_NB_XSTATS; i++) {
1932 		xstats[i].id = i;
1933 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1934 			rte_igc_stats_strings[i].offset);
1935 	}
1936 
1937 	return IGC_NB_XSTATS;
1938 }
1939 
1940 static int
1941 eth_igc_xstats_reset(struct rte_eth_dev *dev)
1942 {
1943 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1944 	struct igc_hw_stats *hw_stats = IGC_DEV_PRIVATE_STATS(dev);
1945 	struct igc_hw_queue_stats *queue_stats =
1946 			IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1947 
1948 	/* Cancel queue status handler for avoid conflict */
1949 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
1950 
1951 	/* HW registers are cleared on read */
1952 	igc_reset_queue_stats_register(hw);
1953 	igc_read_stats_registers(hw, hw_stats);
1954 
1955 	/* Reset software totals */
1956 	memset(hw_stats, 0, sizeof(*hw_stats));
1957 	memset(queue_stats, 0, sizeof(*queue_stats));
1958 
1959 	/* Restart the queue status handler */
1960 	rte_eal_alarm_set(IGC_ALARM_INTERVAL, igc_update_queue_stats_handler,
1961 			dev);
1962 
1963 	return 0;
1964 }
1965 
1966 static int
1967 eth_igc_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1968 	struct rte_eth_xstat_name *xstats_names, unsigned int size)
1969 {
1970 	unsigned int i;
1971 
1972 	if (xstats_names == NULL)
1973 		return IGC_NB_XSTATS;
1974 
1975 	if (size < IGC_NB_XSTATS) {
1976 		PMD_DRV_LOG(ERR, "not enough buffers!");
1977 		return IGC_NB_XSTATS;
1978 	}
1979 
1980 	for (i = 0; i < IGC_NB_XSTATS; i++)
1981 		strlcpy(xstats_names[i].name, rte_igc_stats_strings[i].name,
1982 			sizeof(xstats_names[i].name));
1983 
1984 	return IGC_NB_XSTATS;
1985 }
1986 
1987 static int
1988 eth_igc_xstats_get_names_by_id(struct rte_eth_dev *dev,
1989 		const uint64_t *ids, struct rte_eth_xstat_name *xstats_names,
1990 		unsigned int limit)
1991 {
1992 	unsigned int i;
1993 
1994 	if (!ids)
1995 		return eth_igc_xstats_get_names(dev, xstats_names, limit);
1996 
1997 	for (i = 0; i < limit; i++) {
1998 		if (ids[i] >= IGC_NB_XSTATS) {
1999 			PMD_DRV_LOG(ERR, "id value isn't valid");
2000 			return -EINVAL;
2001 		}
2002 		strlcpy(xstats_names[i].name,
2003 			rte_igc_stats_strings[ids[i]].name,
2004 			sizeof(xstats_names[i].name));
2005 	}
2006 	return limit;
2007 }
2008 
2009 static int
2010 eth_igc_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
2011 		uint64_t *values, unsigned int n)
2012 {
2013 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2014 	struct igc_hw_stats *hw_stats = IGC_DEV_PRIVATE_STATS(dev);
2015 	unsigned int i;
2016 
2017 	igc_read_stats_registers(hw, hw_stats);
2018 
2019 	if (!ids) {
2020 		if (n < IGC_NB_XSTATS)
2021 			return IGC_NB_XSTATS;
2022 
2023 		/* If this is a reset xstats is NULL, and we have cleared the
2024 		 * registers by reading them.
2025 		 */
2026 		if (!values)
2027 			return 0;
2028 
2029 		/* Extended stats */
2030 		for (i = 0; i < IGC_NB_XSTATS; i++)
2031 			values[i] = *(uint64_t *)(((char *)hw_stats) +
2032 					rte_igc_stats_strings[i].offset);
2033 
2034 		return IGC_NB_XSTATS;
2035 
2036 	} else {
2037 		for (i = 0; i < n; i++) {
2038 			if (ids[i] >= IGC_NB_XSTATS) {
2039 				PMD_DRV_LOG(ERR, "id value isn't valid");
2040 				return -EINVAL;
2041 			}
2042 			values[i] = *(uint64_t *)(((char *)hw_stats) +
2043 					rte_igc_stats_strings[ids[i]].offset);
2044 		}
2045 		return n;
2046 	}
2047 }
2048 
2049 static int
2050 eth_igc_queue_stats_mapping_set(struct rte_eth_dev *dev,
2051 		uint16_t queue_id, uint8_t stat_idx, uint8_t is_rx)
2052 {
2053 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
2054 
2055 	/* check queue id is valid */
2056 	if (queue_id >= IGC_QUEUE_PAIRS_NUM) {
2057 		PMD_DRV_LOG(ERR, "queue id(%u) error, max is %u",
2058 			queue_id, IGC_QUEUE_PAIRS_NUM - 1);
2059 		return -EINVAL;
2060 	}
2061 
2062 	/* store the mapping status id */
2063 	if (is_rx)
2064 		igc->rxq_stats_map[queue_id] = stat_idx;
2065 	else
2066 		igc->txq_stats_map[queue_id] = stat_idx;
2067 
2068 	return 0;
2069 }
2070 
2071 static int
2072 eth_igc_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
2073 {
2074 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2075 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2076 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2077 	uint32_t vec = IGC_MISC_VEC_ID;
2078 
2079 	if (rte_intr_allow_others(intr_handle))
2080 		vec = IGC_RX_VEC_START;
2081 
2082 	uint32_t mask = 1u << (queue_id + vec);
2083 
2084 	IGC_WRITE_REG(hw, IGC_EIMC, mask);
2085 	IGC_WRITE_FLUSH(hw);
2086 
2087 	return 0;
2088 }
2089 
2090 static int
2091 eth_igc_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
2092 {
2093 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2094 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2095 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2096 	uint32_t vec = IGC_MISC_VEC_ID;
2097 
2098 	if (rte_intr_allow_others(intr_handle))
2099 		vec = IGC_RX_VEC_START;
2100 
2101 	uint32_t mask = 1u << (queue_id + vec);
2102 
2103 	IGC_WRITE_REG(hw, IGC_EIMS, mask);
2104 	IGC_WRITE_FLUSH(hw);
2105 
2106 	rte_intr_enable(intr_handle);
2107 
2108 	return 0;
2109 }
2110 
2111 static int
2112 eth_igc_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2113 {
2114 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2115 	uint32_t ctrl;
2116 	int tx_pause;
2117 	int rx_pause;
2118 
2119 	fc_conf->pause_time = hw->fc.pause_time;
2120 	fc_conf->high_water = hw->fc.high_water;
2121 	fc_conf->low_water = hw->fc.low_water;
2122 	fc_conf->send_xon = hw->fc.send_xon;
2123 	fc_conf->autoneg = hw->mac.autoneg;
2124 
2125 	/*
2126 	 * Return rx_pause and tx_pause status according to actual setting of
2127 	 * the TFCE and RFCE bits in the CTRL register.
2128 	 */
2129 	ctrl = IGC_READ_REG(hw, IGC_CTRL);
2130 	if (ctrl & IGC_CTRL_TFCE)
2131 		tx_pause = 1;
2132 	else
2133 		tx_pause = 0;
2134 
2135 	if (ctrl & IGC_CTRL_RFCE)
2136 		rx_pause = 1;
2137 	else
2138 		rx_pause = 0;
2139 
2140 	if (rx_pause && tx_pause)
2141 		fc_conf->mode = RTE_ETH_FC_FULL;
2142 	else if (rx_pause)
2143 		fc_conf->mode = RTE_ETH_FC_RX_PAUSE;
2144 	else if (tx_pause)
2145 		fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
2146 	else
2147 		fc_conf->mode = RTE_ETH_FC_NONE;
2148 
2149 	return 0;
2150 }
2151 
2152 static int
2153 eth_igc_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2154 {
2155 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2156 	uint32_t rx_buf_size;
2157 	uint32_t max_high_water;
2158 	uint32_t rctl;
2159 	int err;
2160 
2161 	if (fc_conf->autoneg != hw->mac.autoneg)
2162 		return -ENOTSUP;
2163 
2164 	rx_buf_size = igc_get_rx_buffer_size(hw);
2165 	PMD_DRV_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
2166 
2167 	/* At least reserve one Ethernet frame for watermark */
2168 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
2169 	if (fc_conf->high_water > max_high_water ||
2170 		fc_conf->high_water < fc_conf->low_water) {
2171 		PMD_DRV_LOG(ERR,
2172 			"Incorrect high(%u)/low(%u) water value, max is %u",
2173 			fc_conf->high_water, fc_conf->low_water,
2174 			max_high_water);
2175 		return -EINVAL;
2176 	}
2177 
2178 	switch (fc_conf->mode) {
2179 	case RTE_ETH_FC_NONE:
2180 		hw->fc.requested_mode = igc_fc_none;
2181 		break;
2182 	case RTE_ETH_FC_RX_PAUSE:
2183 		hw->fc.requested_mode = igc_fc_rx_pause;
2184 		break;
2185 	case RTE_ETH_FC_TX_PAUSE:
2186 		hw->fc.requested_mode = igc_fc_tx_pause;
2187 		break;
2188 	case RTE_ETH_FC_FULL:
2189 		hw->fc.requested_mode = igc_fc_full;
2190 		break;
2191 	default:
2192 		PMD_DRV_LOG(ERR, "unsupported fc mode: %u", fc_conf->mode);
2193 		return -EINVAL;
2194 	}
2195 
2196 	hw->fc.pause_time     = fc_conf->pause_time;
2197 	hw->fc.high_water     = fc_conf->high_water;
2198 	hw->fc.low_water      = fc_conf->low_water;
2199 	hw->fc.send_xon	      = fc_conf->send_xon;
2200 
2201 	err = igc_setup_link_generic(hw);
2202 	if (err == IGC_SUCCESS) {
2203 		/**
2204 		 * check if we want to forward MAC frames - driver doesn't have
2205 		 * native capability to do that, so we'll write the registers
2206 		 * ourselves
2207 		 **/
2208 		rctl = IGC_READ_REG(hw, IGC_RCTL);
2209 
2210 		/* set or clear MFLCN.PMCF bit depending on configuration */
2211 		if (fc_conf->mac_ctrl_frame_fwd != 0)
2212 			rctl |= IGC_RCTL_PMCF;
2213 		else
2214 			rctl &= ~IGC_RCTL_PMCF;
2215 
2216 		IGC_WRITE_REG(hw, IGC_RCTL, rctl);
2217 		IGC_WRITE_FLUSH(hw);
2218 
2219 		return 0;
2220 	}
2221 
2222 	PMD_DRV_LOG(ERR, "igc_setup_link_generic = 0x%x", err);
2223 	return -EIO;
2224 }
2225 
2226 static int
2227 eth_igc_rss_reta_update(struct rte_eth_dev *dev,
2228 			struct rte_eth_rss_reta_entry64 *reta_conf,
2229 			uint16_t reta_size)
2230 {
2231 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2232 	uint16_t i;
2233 
2234 	if (reta_size != RTE_ETH_RSS_RETA_SIZE_128) {
2235 		PMD_DRV_LOG(ERR,
2236 			"The size of RSS redirection table configured(%d) doesn't match the number hardware can supported(%d)",
2237 			reta_size, RTE_ETH_RSS_RETA_SIZE_128);
2238 		return -EINVAL;
2239 	}
2240 
2241 	RTE_BUILD_BUG_ON(RTE_ETH_RSS_RETA_SIZE_128 % IGC_RSS_RDT_REG_SIZE);
2242 
2243 	/* set redirection table */
2244 	for (i = 0; i < RTE_ETH_RSS_RETA_SIZE_128; i += IGC_RSS_RDT_REG_SIZE) {
2245 		union igc_rss_reta_reg reta, reg;
2246 		uint16_t idx, shift;
2247 		uint8_t j, mask;
2248 
2249 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2250 		shift = i % RTE_ETH_RETA_GROUP_SIZE;
2251 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
2252 				IGC_RSS_RDT_REG_SIZE_MASK);
2253 
2254 		/* if no need to update the register */
2255 		if (!mask ||
2256 		    shift > (RTE_ETH_RETA_GROUP_SIZE - IGC_RSS_RDT_REG_SIZE))
2257 			continue;
2258 
2259 		/* check mask whether need to read the register value first */
2260 		if (mask == IGC_RSS_RDT_REG_SIZE_MASK)
2261 			reg.dword = 0;
2262 		else
2263 			reg.dword = IGC_READ_REG_LE_VALUE(hw,
2264 					IGC_RETA(i / IGC_RSS_RDT_REG_SIZE));
2265 
2266 		/* update the register */
2267 		RTE_BUILD_BUG_ON(sizeof(reta.bytes) != IGC_RSS_RDT_REG_SIZE);
2268 		for (j = 0; j < IGC_RSS_RDT_REG_SIZE; j++) {
2269 			if (mask & (1u << j))
2270 				reta.bytes[j] =
2271 					(uint8_t)reta_conf[idx].reta[shift + j];
2272 			else
2273 				reta.bytes[j] = reg.bytes[j];
2274 		}
2275 		IGC_WRITE_REG_LE_VALUE(hw,
2276 			IGC_RETA(i / IGC_RSS_RDT_REG_SIZE), reta.dword);
2277 	}
2278 
2279 	return 0;
2280 }
2281 
2282 static int
2283 eth_igc_rss_reta_query(struct rte_eth_dev *dev,
2284 		       struct rte_eth_rss_reta_entry64 *reta_conf,
2285 		       uint16_t reta_size)
2286 {
2287 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2288 	uint16_t i;
2289 
2290 	if (reta_size != RTE_ETH_RSS_RETA_SIZE_128) {
2291 		PMD_DRV_LOG(ERR,
2292 			"The size of RSS redirection table configured(%d) doesn't match the number hardware can supported(%d)",
2293 			reta_size, RTE_ETH_RSS_RETA_SIZE_128);
2294 		return -EINVAL;
2295 	}
2296 
2297 	RTE_BUILD_BUG_ON(RTE_ETH_RSS_RETA_SIZE_128 % IGC_RSS_RDT_REG_SIZE);
2298 
2299 	/* read redirection table */
2300 	for (i = 0; i < RTE_ETH_RSS_RETA_SIZE_128; i += IGC_RSS_RDT_REG_SIZE) {
2301 		union igc_rss_reta_reg reta;
2302 		uint16_t idx, shift;
2303 		uint8_t j, mask;
2304 
2305 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2306 		shift = i % RTE_ETH_RETA_GROUP_SIZE;
2307 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
2308 				IGC_RSS_RDT_REG_SIZE_MASK);
2309 
2310 		/* if no need to read register */
2311 		if (!mask ||
2312 		    shift > (RTE_ETH_RETA_GROUP_SIZE - IGC_RSS_RDT_REG_SIZE))
2313 			continue;
2314 
2315 		/* read register and get the queue index */
2316 		RTE_BUILD_BUG_ON(sizeof(reta.bytes) != IGC_RSS_RDT_REG_SIZE);
2317 		reta.dword = IGC_READ_REG_LE_VALUE(hw,
2318 				IGC_RETA(i / IGC_RSS_RDT_REG_SIZE));
2319 		for (j = 0; j < IGC_RSS_RDT_REG_SIZE; j++) {
2320 			if (mask & (1u << j))
2321 				reta_conf[idx].reta[shift + j] = reta.bytes[j];
2322 		}
2323 	}
2324 
2325 	return 0;
2326 }
2327 
2328 static int
2329 eth_igc_rss_hash_update(struct rte_eth_dev *dev,
2330 			struct rte_eth_rss_conf *rss_conf)
2331 {
2332 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2333 	igc_hw_rss_hash_set(hw, rss_conf);
2334 	return 0;
2335 }
2336 
2337 static int
2338 eth_igc_rss_hash_conf_get(struct rte_eth_dev *dev,
2339 			struct rte_eth_rss_conf *rss_conf)
2340 {
2341 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2342 	uint32_t *hash_key = (uint32_t *)rss_conf->rss_key;
2343 	uint32_t mrqc;
2344 	uint64_t rss_hf;
2345 
2346 	if (hash_key != NULL) {
2347 		int i;
2348 
2349 		/* if not enough space for store hash key */
2350 		if (rss_conf->rss_key_len != IGC_HKEY_SIZE) {
2351 			PMD_DRV_LOG(ERR,
2352 				"RSS hash key size %u in parameter doesn't match the hardware hash key size %u",
2353 				rss_conf->rss_key_len, IGC_HKEY_SIZE);
2354 			return -EINVAL;
2355 		}
2356 
2357 		/* read RSS key from register */
2358 		for (i = 0; i < IGC_HKEY_MAX_INDEX; i++)
2359 			hash_key[i] = IGC_READ_REG_LE_VALUE(hw, IGC_RSSRK(i));
2360 	}
2361 
2362 	/* get RSS functions configured in MRQC register */
2363 	mrqc = IGC_READ_REG(hw, IGC_MRQC);
2364 	if ((mrqc & IGC_MRQC_ENABLE_RSS_4Q) == 0)
2365 		return 0;
2366 
2367 	rss_hf = 0;
2368 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4)
2369 		rss_hf |= RTE_ETH_RSS_IPV4;
2370 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4_TCP)
2371 		rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_TCP;
2372 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6)
2373 		rss_hf |= RTE_ETH_RSS_IPV6;
2374 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_EX)
2375 		rss_hf |= RTE_ETH_RSS_IPV6_EX;
2376 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_TCP)
2377 		rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_TCP;
2378 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_TCP_EX)
2379 		rss_hf |= RTE_ETH_RSS_IPV6_TCP_EX;
2380 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4_UDP)
2381 		rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_UDP;
2382 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_UDP)
2383 		rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_UDP;
2384 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_UDP_EX)
2385 		rss_hf |= RTE_ETH_RSS_IPV6_UDP_EX;
2386 
2387 	rss_conf->rss_hf |= rss_hf;
2388 	return 0;
2389 }
2390 
2391 static int
2392 eth_igc_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2393 {
2394 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2395 	struct igc_vfta *shadow_vfta = IGC_DEV_PRIVATE_VFTA(dev);
2396 	uint32_t vfta;
2397 	uint32_t vid_idx;
2398 	uint32_t vid_bit;
2399 
2400 	vid_idx = (vlan_id >> IGC_VFTA_ENTRY_SHIFT) & IGC_VFTA_ENTRY_MASK;
2401 	vid_bit = 1u << (vlan_id & IGC_VFTA_ENTRY_BIT_SHIFT_MASK);
2402 	vfta = shadow_vfta->vfta[vid_idx];
2403 	if (on)
2404 		vfta |= vid_bit;
2405 	else
2406 		vfta &= ~vid_bit;
2407 	IGC_WRITE_REG_ARRAY(hw, IGC_VFTA, vid_idx, vfta);
2408 
2409 	/* update local VFTA copy */
2410 	shadow_vfta->vfta[vid_idx] = vfta;
2411 
2412 	return 0;
2413 }
2414 
2415 static void
2416 igc_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2417 {
2418 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2419 	igc_read_reg_check_clear_bits(hw, IGC_RCTL,
2420 			IGC_RCTL_CFIEN | IGC_RCTL_VFE);
2421 }
2422 
2423 static void
2424 igc_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2425 {
2426 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2427 	struct igc_vfta *shadow_vfta = IGC_DEV_PRIVATE_VFTA(dev);
2428 	uint32_t reg_val;
2429 	int i;
2430 
2431 	/* Filter Table Enable, CFI not used for packet acceptance */
2432 	reg_val = IGC_READ_REG(hw, IGC_RCTL);
2433 	reg_val &= ~IGC_RCTL_CFIEN;
2434 	reg_val |= IGC_RCTL_VFE;
2435 	IGC_WRITE_REG(hw, IGC_RCTL, reg_val);
2436 
2437 	/* restore VFTA table */
2438 	for (i = 0; i < IGC_VFTA_SIZE; i++)
2439 		IGC_WRITE_REG_ARRAY(hw, IGC_VFTA, i, shadow_vfta->vfta[i]);
2440 }
2441 
2442 static void
2443 igc_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2444 {
2445 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2446 
2447 	igc_read_reg_check_clear_bits(hw, IGC_CTRL, IGC_CTRL_VME);
2448 }
2449 
2450 static void
2451 igc_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2452 {
2453 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2454 
2455 	igc_read_reg_check_set_bits(hw, IGC_CTRL, IGC_CTRL_VME);
2456 }
2457 
2458 static int
2459 igc_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2460 {
2461 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2462 	uint32_t frame_size = dev->data->mtu + IGC_ETH_OVERHEAD;
2463 	uint32_t ctrl_ext;
2464 
2465 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
2466 
2467 	/* if extend vlan hasn't been enabled */
2468 	if ((ctrl_ext & IGC_CTRL_EXT_EXT_VLAN) == 0)
2469 		return 0;
2470 
2471 	/* Update maximum packet length */
2472 	if (frame_size < RTE_ETHER_MIN_MTU + VLAN_TAG_SIZE) {
2473 		PMD_DRV_LOG(ERR, "Maximum packet length %u error, min is %u",
2474 			frame_size, VLAN_TAG_SIZE + RTE_ETHER_MIN_MTU);
2475 		return -EINVAL;
2476 	}
2477 	IGC_WRITE_REG(hw, IGC_RLPML, frame_size - VLAN_TAG_SIZE);
2478 
2479 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext & ~IGC_CTRL_EXT_EXT_VLAN);
2480 	return 0;
2481 }
2482 
2483 static int
2484 igc_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2485 {
2486 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2487 	uint32_t frame_size = dev->data->mtu + IGC_ETH_OVERHEAD;
2488 	uint32_t ctrl_ext;
2489 
2490 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
2491 
2492 	/* if extend vlan has been enabled */
2493 	if (ctrl_ext & IGC_CTRL_EXT_EXT_VLAN)
2494 		return 0;
2495 
2496 	/* Update maximum packet length */
2497 	if (frame_size > MAX_RX_JUMBO_FRAME_SIZE) {
2498 		PMD_DRV_LOG(ERR, "Maximum packet length %u error, max is %u",
2499 			frame_size, MAX_RX_JUMBO_FRAME_SIZE);
2500 		return -EINVAL;
2501 	}
2502 	IGC_WRITE_REG(hw, IGC_RLPML, frame_size);
2503 
2504 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext | IGC_CTRL_EXT_EXT_VLAN);
2505 	return 0;
2506 }
2507 
2508 static int
2509 eth_igc_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2510 {
2511 	struct rte_eth_rxmode *rxmode;
2512 
2513 	rxmode = &dev->data->dev_conf.rxmode;
2514 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
2515 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
2516 			igc_vlan_hw_strip_enable(dev);
2517 		else
2518 			igc_vlan_hw_strip_disable(dev);
2519 	}
2520 
2521 	if (mask & RTE_ETH_VLAN_FILTER_MASK) {
2522 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
2523 			igc_vlan_hw_filter_enable(dev);
2524 		else
2525 			igc_vlan_hw_filter_disable(dev);
2526 	}
2527 
2528 	if (mask & RTE_ETH_VLAN_EXTEND_MASK) {
2529 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND)
2530 			return igc_vlan_hw_extend_enable(dev);
2531 		else
2532 			return igc_vlan_hw_extend_disable(dev);
2533 	}
2534 
2535 	return 0;
2536 }
2537 
2538 static int
2539 eth_igc_vlan_tpid_set(struct rte_eth_dev *dev,
2540 		      enum rte_vlan_type vlan_type,
2541 		      uint16_t tpid)
2542 {
2543 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2544 	uint32_t reg_val;
2545 
2546 	/* only outer TPID of double VLAN can be configured*/
2547 	if (vlan_type == RTE_ETH_VLAN_TYPE_OUTER) {
2548 		reg_val = IGC_READ_REG(hw, IGC_VET);
2549 		reg_val = (reg_val & (~IGC_VET_EXT)) |
2550 			((uint32_t)tpid << IGC_VET_EXT_SHIFT);
2551 		IGC_WRITE_REG(hw, IGC_VET, reg_val);
2552 
2553 		return 0;
2554 	}
2555 
2556 	/* all other TPID values are read-only*/
2557 	PMD_DRV_LOG(ERR, "Not supported");
2558 	return -ENOTSUP;
2559 }
2560 
2561 static int
2562 eth_igc_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
2563 	struct rte_pci_device *pci_dev)
2564 {
2565 	PMD_INIT_FUNC_TRACE();
2566 	return rte_eth_dev_pci_generic_probe(pci_dev,
2567 		sizeof(struct igc_adapter), eth_igc_dev_init);
2568 }
2569 
2570 static int
2571 eth_igc_pci_remove(struct rte_pci_device *pci_dev)
2572 {
2573 	PMD_INIT_FUNC_TRACE();
2574 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igc_dev_uninit);
2575 }
2576 
2577 static struct rte_pci_driver rte_igc_pmd = {
2578 	.id_table = pci_id_igc_map,
2579 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
2580 	.probe = eth_igc_pci_probe,
2581 	.remove = eth_igc_pci_remove,
2582 };
2583 
2584 RTE_PMD_REGISTER_PCI(net_igc, rte_igc_pmd);
2585 RTE_PMD_REGISTER_PCI_TABLE(net_igc, pci_id_igc_map);
2586 RTE_PMD_REGISTER_KMOD_DEP(net_igc, "* igb_uio | uio_pci_generic | vfio-pci");
2587