xref: /dpdk/drivers/net/igc/igc_ethdev.c (revision 7a292619)
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 <rte_ethdev_driver.h>
12 #include <rte_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 void 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 		struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
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 	.filter_ctrl		= eth_igc_filter_ctrl,
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 	ret  = igc_check_mq_mode(dev);
345 	if (ret != 0)
346 		return ret;
347 
348 	intr->flags |= IGC_FLAG_NEED_LINK_UPDATE;
349 	return 0;
350 }
351 
352 static int
353 eth_igc_set_link_up(struct rte_eth_dev *dev)
354 {
355 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
356 
357 	if (hw->phy.media_type == igc_media_type_copper)
358 		igc_power_up_phy(hw);
359 	else
360 		igc_power_up_fiber_serdes_link(hw);
361 	return 0;
362 }
363 
364 static int
365 eth_igc_set_link_down(struct rte_eth_dev *dev)
366 {
367 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
368 
369 	if (hw->phy.media_type == igc_media_type_copper)
370 		igc_power_down_phy(hw);
371 	else
372 		igc_shutdown_fiber_serdes_link(hw);
373 	return 0;
374 }
375 
376 /*
377  * disable other interrupt
378  */
379 static void
380 igc_intr_other_disable(struct rte_eth_dev *dev)
381 {
382 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
383 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
384 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
385 
386 	if (rte_intr_allow_others(intr_handle) &&
387 		dev->data->dev_conf.intr_conf.lsc) {
388 		IGC_WRITE_REG(hw, IGC_EIMC, 1u << IGC_MSIX_OTHER_INTR_VEC);
389 	}
390 
391 	IGC_WRITE_REG(hw, IGC_IMC, ~0);
392 	IGC_WRITE_FLUSH(hw);
393 }
394 
395 /*
396  * enable other interrupt
397  */
398 static inline void
399 igc_intr_other_enable(struct rte_eth_dev *dev)
400 {
401 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
402 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
403 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
404 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
405 
406 	if (rte_intr_allow_others(intr_handle) &&
407 		dev->data->dev_conf.intr_conf.lsc) {
408 		IGC_WRITE_REG(hw, IGC_EIMS, 1u << IGC_MSIX_OTHER_INTR_VEC);
409 	}
410 
411 	IGC_WRITE_REG(hw, IGC_IMS, intr->mask);
412 	IGC_WRITE_FLUSH(hw);
413 }
414 
415 /*
416  * It reads ICR and gets interrupt causes, check it and set a bit flag
417  * to update link status.
418  */
419 static void
420 eth_igc_interrupt_get_status(struct rte_eth_dev *dev)
421 {
422 	uint32_t icr;
423 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
424 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
425 
426 	/* read-on-clear nic registers here */
427 	icr = IGC_READ_REG(hw, IGC_ICR);
428 
429 	intr->flags = 0;
430 	if (icr & IGC_ICR_LSC)
431 		intr->flags |= IGC_FLAG_NEED_LINK_UPDATE;
432 }
433 
434 /* return 0 means link status changed, -1 means not changed */
435 static int
436 eth_igc_link_update(struct rte_eth_dev *dev, int wait_to_complete)
437 {
438 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
439 	struct rte_eth_link link;
440 	int link_check, count;
441 
442 	link_check = 0;
443 	hw->mac.get_link_status = 1;
444 
445 	/* possible wait-to-complete in up to 9 seconds */
446 	for (count = 0; count < IGC_LINK_UPDATE_CHECK_TIMEOUT; count++) {
447 		/* Read the real link status */
448 		switch (hw->phy.media_type) {
449 		case igc_media_type_copper:
450 			/* Do the work to read phy */
451 			igc_check_for_link(hw);
452 			link_check = !hw->mac.get_link_status;
453 			break;
454 
455 		case igc_media_type_fiber:
456 			igc_check_for_link(hw);
457 			link_check = (IGC_READ_REG(hw, IGC_STATUS) &
458 				      IGC_STATUS_LU);
459 			break;
460 
461 		case igc_media_type_internal_serdes:
462 			igc_check_for_link(hw);
463 			link_check = hw->mac.serdes_has_link;
464 			break;
465 
466 		default:
467 			break;
468 		}
469 		if (link_check || wait_to_complete == 0)
470 			break;
471 		rte_delay_ms(IGC_LINK_UPDATE_CHECK_INTERVAL);
472 	}
473 	memset(&link, 0, sizeof(link));
474 
475 	/* Now we check if a transition has happened */
476 	if (link_check) {
477 		uint16_t duplex, speed;
478 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
479 		link.link_duplex = (duplex == FULL_DUPLEX) ?
480 				ETH_LINK_FULL_DUPLEX :
481 				ETH_LINK_HALF_DUPLEX;
482 		link.link_speed = speed;
483 		link.link_status = ETH_LINK_UP;
484 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
485 				ETH_LINK_SPEED_FIXED);
486 
487 		if (speed == SPEED_2500) {
488 			uint32_t tipg = IGC_READ_REG(hw, IGC_TIPG);
489 			if ((tipg & IGC_TIPG_IPGT_MASK) != 0x0b) {
490 				tipg &= ~IGC_TIPG_IPGT_MASK;
491 				tipg |= 0x0b;
492 				IGC_WRITE_REG(hw, IGC_TIPG, tipg);
493 			}
494 		}
495 	} else {
496 		link.link_speed = 0;
497 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
498 		link.link_status = ETH_LINK_DOWN;
499 		link.link_autoneg = ETH_LINK_FIXED;
500 	}
501 
502 	return rte_eth_linkstatus_set(dev, &link);
503 }
504 
505 /*
506  * It executes link_update after knowing an interrupt is present.
507  */
508 static void
509 eth_igc_interrupt_action(struct rte_eth_dev *dev)
510 {
511 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
512 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
513 	struct rte_eth_link link;
514 	int ret;
515 
516 	if (intr->flags & IGC_FLAG_NEED_LINK_UPDATE) {
517 		intr->flags &= ~IGC_FLAG_NEED_LINK_UPDATE;
518 
519 		/* set get_link_status to check register later */
520 		ret = eth_igc_link_update(dev, 0);
521 
522 		/* check if link has changed */
523 		if (ret < 0)
524 			return;
525 
526 		rte_eth_linkstatus_get(dev, &link);
527 		if (link.link_status)
528 			PMD_DRV_LOG(INFO,
529 				" Port %d: Link Up - speed %u Mbps - %s",
530 				dev->data->port_id,
531 				(unsigned int)link.link_speed,
532 				link.link_duplex == ETH_LINK_FULL_DUPLEX ?
533 				"full-duplex" : "half-duplex");
534 		else
535 			PMD_DRV_LOG(INFO, " Port %d: Link Down",
536 				dev->data->port_id);
537 
538 		PMD_DRV_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
539 				pci_dev->addr.domain,
540 				pci_dev->addr.bus,
541 				pci_dev->addr.devid,
542 				pci_dev->addr.function);
543 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
544 	}
545 }
546 
547 /*
548  * Interrupt handler which shall be registered at first.
549  *
550  * @handle
551  *  Pointer to interrupt handle.
552  * @param
553  *  The address of parameter (struct rte_eth_dev *) registered before.
554  */
555 static void
556 eth_igc_interrupt_handler(void *param)
557 {
558 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
559 
560 	eth_igc_interrupt_get_status(dev);
561 	eth_igc_interrupt_action(dev);
562 }
563 
564 static void igc_read_queue_stats_register(struct rte_eth_dev *dev);
565 
566 /*
567  * Update the queue status every IGC_ALARM_INTERVAL time.
568  * @param
569  *  The address of parameter (struct rte_eth_dev *) registered before.
570  */
571 static void
572 igc_update_queue_stats_handler(void *param)
573 {
574 	struct rte_eth_dev *dev = param;
575 	igc_read_queue_stats_register(dev);
576 	rte_eal_alarm_set(IGC_ALARM_INTERVAL,
577 			igc_update_queue_stats_handler, dev);
578 }
579 
580 /*
581  * rx,tx enable/disable
582  */
583 static void
584 eth_igc_rxtx_control(struct rte_eth_dev *dev, bool enable)
585 {
586 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
587 	uint32_t tctl, rctl;
588 
589 	tctl = IGC_READ_REG(hw, IGC_TCTL);
590 	rctl = IGC_READ_REG(hw, IGC_RCTL);
591 
592 	if (enable) {
593 		/* enable Tx/Rx */
594 		tctl |= IGC_TCTL_EN;
595 		rctl |= IGC_RCTL_EN;
596 	} else {
597 		/* disable Tx/Rx */
598 		tctl &= ~IGC_TCTL_EN;
599 		rctl &= ~IGC_RCTL_EN;
600 	}
601 	IGC_WRITE_REG(hw, IGC_TCTL, tctl);
602 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
603 	IGC_WRITE_FLUSH(hw);
604 }
605 
606 /*
607  *  This routine disables all traffic on the adapter by issuing a
608  *  global reset on the MAC.
609  */
610 static void
611 eth_igc_stop(struct rte_eth_dev *dev)
612 {
613 	struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
614 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
615 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
616 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
617 	struct rte_eth_link link;
618 
619 	adapter->stopped = 1;
620 
621 	/* disable receive and transmit */
622 	eth_igc_rxtx_control(dev, false);
623 
624 	/* disable all MSI-X interrupts */
625 	IGC_WRITE_REG(hw, IGC_EIMC, 0x1f);
626 	IGC_WRITE_FLUSH(hw);
627 
628 	/* clear all MSI-X interrupts */
629 	IGC_WRITE_REG(hw, IGC_EICR, 0x1f);
630 
631 	igc_intr_other_disable(dev);
632 
633 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
634 
635 	/* disable intr eventfd mapping */
636 	rte_intr_disable(intr_handle);
637 
638 	igc_reset_hw(hw);
639 
640 	/* disable all wake up */
641 	IGC_WRITE_REG(hw, IGC_WUC, 0);
642 
643 	/* disable checking EEE operation in MAC loopback mode */
644 	igc_read_reg_check_clear_bits(hw, IGC_EEER, IGC_EEER_EEE_FRC_AN);
645 
646 	/* Set bit for Go Link disconnect */
647 	igc_read_reg_check_set_bits(hw, IGC_82580_PHY_POWER_MGMT,
648 			IGC_82580_PM_GO_LINKD);
649 
650 	/* Power down the phy. Needed to make the link go Down */
651 	eth_igc_set_link_down(dev);
652 
653 	igc_dev_clear_queues(dev);
654 
655 	/* clear the recorded link status */
656 	memset(&link, 0, sizeof(link));
657 	rte_eth_linkstatus_set(dev, &link);
658 
659 	if (!rte_intr_allow_others(intr_handle))
660 		/* resume to the default handler */
661 		rte_intr_callback_register(intr_handle,
662 					   eth_igc_interrupt_handler,
663 					   (void *)dev);
664 
665 	/* Clean datapath event and queue/vec mapping */
666 	rte_intr_efd_disable(intr_handle);
667 	if (intr_handle->intr_vec != NULL) {
668 		rte_free(intr_handle->intr_vec);
669 		intr_handle->intr_vec = NULL;
670 	}
671 }
672 
673 /*
674  * write interrupt vector allocation register
675  * @hw
676  *  board private structure
677  * @queue_index
678  *  queue index, valid 0,1,2,3
679  * @tx
680  *  tx:1, rx:0
681  * @msix_vector
682  *  msix-vector, valid 0,1,2,3,4
683  */
684 static void
685 igc_write_ivar(struct igc_hw *hw, uint8_t queue_index,
686 		bool tx, uint8_t msix_vector)
687 {
688 	uint8_t offset = 0;
689 	uint8_t reg_index = queue_index >> 1;
690 	uint32_t val;
691 
692 	/*
693 	 * IVAR(0)
694 	 * bit31...24	bit23...16	bit15...8	bit7...0
695 	 * TX1		RX1		TX0		RX0
696 	 *
697 	 * IVAR(1)
698 	 * bit31...24	bit23...16	bit15...8	bit7...0
699 	 * TX3		RX3		TX2		RX2
700 	 */
701 
702 	if (tx)
703 		offset = 8;
704 
705 	if (queue_index & 1)
706 		offset += 16;
707 
708 	val = IGC_READ_REG_ARRAY(hw, IGC_IVAR0, reg_index);
709 
710 	/* clear bits */
711 	val &= ~((uint32_t)0xFF << offset);
712 
713 	/* write vector and valid bit */
714 	val |= (uint32_t)(msix_vector | IGC_IVAR_VALID) << offset;
715 
716 	IGC_WRITE_REG_ARRAY(hw, IGC_IVAR0, reg_index, val);
717 }
718 
719 /* Sets up the hardware to generate MSI-X interrupts properly
720  * @hw
721  *  board private structure
722  */
723 static void
724 igc_configure_msix_intr(struct rte_eth_dev *dev)
725 {
726 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
727 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
728 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
729 
730 	uint32_t intr_mask;
731 	uint32_t vec = IGC_MISC_VEC_ID;
732 	uint32_t base = IGC_MISC_VEC_ID;
733 	uint32_t misc_shift = 0;
734 	int i;
735 
736 	/* won't configure msix register if no mapping is done
737 	 * between intr vector and event fd
738 	 */
739 	if (!rte_intr_dp_is_en(intr_handle))
740 		return;
741 
742 	if (rte_intr_allow_others(intr_handle)) {
743 		base = IGC_RX_VEC_START;
744 		vec = base;
745 		misc_shift = 1;
746 	}
747 
748 	/* turn on MSI-X capability first */
749 	IGC_WRITE_REG(hw, IGC_GPIE, IGC_GPIE_MSIX_MODE |
750 				IGC_GPIE_PBA | IGC_GPIE_EIAME |
751 				IGC_GPIE_NSICR);
752 	intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
753 		misc_shift;
754 
755 	if (dev->data->dev_conf.intr_conf.lsc)
756 		intr_mask |= (1u << IGC_MSIX_OTHER_INTR_VEC);
757 
758 	/* enable msix auto-clear */
759 	igc_read_reg_check_set_bits(hw, IGC_EIAC, intr_mask);
760 
761 	/* set other cause interrupt vector */
762 	igc_read_reg_check_set_bits(hw, IGC_IVAR_MISC,
763 		(uint32_t)(IGC_MSIX_OTHER_INTR_VEC | IGC_IVAR_VALID) << 8);
764 
765 	/* enable auto-mask */
766 	igc_read_reg_check_set_bits(hw, IGC_EIAM, intr_mask);
767 
768 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
769 		igc_write_ivar(hw, i, 0, vec);
770 		intr_handle->intr_vec[i] = vec;
771 		if (vec < base + intr_handle->nb_efd - 1)
772 			vec++;
773 	}
774 
775 	IGC_WRITE_FLUSH(hw);
776 }
777 
778 /**
779  * It enables the interrupt mask and then enable the interrupt.
780  *
781  * @dev
782  *  Pointer to struct rte_eth_dev.
783  * @on
784  *  Enable or Disable
785  */
786 static void
787 igc_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
788 {
789 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
790 
791 	if (on)
792 		intr->mask |= IGC_ICR_LSC;
793 	else
794 		intr->mask &= ~IGC_ICR_LSC;
795 }
796 
797 /*
798  * It enables the interrupt.
799  * It will be called once only during nic initialized.
800  */
801 static void
802 igc_rxq_interrupt_setup(struct rte_eth_dev *dev)
803 {
804 	uint32_t mask;
805 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
806 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
807 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
808 	int misc_shift = rte_intr_allow_others(intr_handle) ? 1 : 0;
809 
810 	/* won't configure msix register if no mapping is done
811 	 * between intr vector and event fd
812 	 */
813 	if (!rte_intr_dp_is_en(intr_handle))
814 		return;
815 
816 	mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) << misc_shift;
817 	IGC_WRITE_REG(hw, IGC_EIMS, mask);
818 }
819 
820 /*
821  *  Get hardware rx-buffer size.
822  */
823 static inline int
824 igc_get_rx_buffer_size(struct igc_hw *hw)
825 {
826 	return (IGC_READ_REG(hw, IGC_RXPBS) & 0x3f) << 10;
827 }
828 
829 /*
830  * igc_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
831  * For ASF and Pass Through versions of f/w this means
832  * that the driver is loaded.
833  */
834 static void
835 igc_hw_control_acquire(struct igc_hw *hw)
836 {
837 	uint32_t ctrl_ext;
838 
839 	/* Let firmware know the driver has taken over */
840 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
841 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext | IGC_CTRL_EXT_DRV_LOAD);
842 }
843 
844 /*
845  * igc_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
846  * For ASF and Pass Through versions of f/w this means that the
847  * driver is no longer loaded.
848  */
849 static void
850 igc_hw_control_release(struct igc_hw *hw)
851 {
852 	uint32_t ctrl_ext;
853 
854 	/* Let firmware taken over control of h/w */
855 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
856 	IGC_WRITE_REG(hw, IGC_CTRL_EXT,
857 			ctrl_ext & ~IGC_CTRL_EXT_DRV_LOAD);
858 }
859 
860 static int
861 igc_hardware_init(struct igc_hw *hw)
862 {
863 	uint32_t rx_buf_size;
864 	int diag;
865 
866 	/* Let the firmware know the OS is in control */
867 	igc_hw_control_acquire(hw);
868 
869 	/* Issue a global reset */
870 	igc_reset_hw(hw);
871 
872 	/* disable all wake up */
873 	IGC_WRITE_REG(hw, IGC_WUC, 0);
874 
875 	/*
876 	 * Hardware flow control
877 	 * - High water mark should allow for at least two standard size (1518)
878 	 *   frames to be received after sending an XOFF.
879 	 * - Low water mark works best when it is very near the high water mark.
880 	 *   This allows the receiver to restart by sending XON when it has
881 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
882 	 *   restart after one full frame is pulled from the buffer. There
883 	 *   could be several smaller frames in the buffer and if so they will
884 	 *   not trigger the XON until their total number reduces the buffer
885 	 *   by 1500.
886 	 */
887 	rx_buf_size = igc_get_rx_buffer_size(hw);
888 	hw->fc.high_water = rx_buf_size - (RTE_ETHER_MAX_LEN * 2);
889 	hw->fc.low_water = hw->fc.high_water - 1500;
890 	hw->fc.pause_time = IGC_FC_PAUSE_TIME;
891 	hw->fc.send_xon = 1;
892 	hw->fc.requested_mode = igc_fc_full;
893 
894 	diag = igc_init_hw(hw);
895 	if (diag < 0)
896 		return diag;
897 
898 	igc_get_phy_info(hw);
899 	igc_check_for_link(hw);
900 
901 	return 0;
902 }
903 
904 static int
905 eth_igc_start(struct rte_eth_dev *dev)
906 {
907 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
908 	struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
909 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
910 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
911 	uint32_t *speeds;
912 	int ret;
913 
914 	PMD_INIT_FUNC_TRACE();
915 
916 	/* disable all MSI-X interrupts */
917 	IGC_WRITE_REG(hw, IGC_EIMC, 0x1f);
918 	IGC_WRITE_FLUSH(hw);
919 
920 	/* clear all MSI-X interrupts */
921 	IGC_WRITE_REG(hw, IGC_EICR, 0x1f);
922 
923 	/* disable uio/vfio intr/eventfd mapping */
924 	if (!adapter->stopped)
925 		rte_intr_disable(intr_handle);
926 
927 	/* Power up the phy. Needed to make the link go Up */
928 	eth_igc_set_link_up(dev);
929 
930 	/* Put the address into the Receive Address Array */
931 	igc_rar_set(hw, hw->mac.addr, 0);
932 
933 	/* Initialize the hardware */
934 	if (igc_hardware_init(hw)) {
935 		PMD_DRV_LOG(ERR, "Unable to initialize the hardware");
936 		return -EIO;
937 	}
938 	adapter->stopped = 0;
939 
940 	/* check and configure queue intr-vector mapping */
941 	if (rte_intr_cap_multiple(intr_handle) &&
942 		dev->data->dev_conf.intr_conf.rxq) {
943 		uint32_t intr_vector = dev->data->nb_rx_queues;
944 		if (rte_intr_efd_enable(intr_handle, intr_vector))
945 			return -1;
946 	}
947 
948 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
949 		intr_handle->intr_vec = rte_zmalloc("intr_vec",
950 			dev->data->nb_rx_queues * sizeof(int), 0);
951 		if (intr_handle->intr_vec == NULL) {
952 			PMD_DRV_LOG(ERR,
953 				"Failed to allocate %d rx_queues intr_vec",
954 				dev->data->nb_rx_queues);
955 			return -ENOMEM;
956 		}
957 	}
958 
959 	/* configure msix for rx interrupt */
960 	igc_configure_msix_intr(dev);
961 
962 	igc_tx_init(dev);
963 
964 	/* This can fail when allocating mbufs for descriptor rings */
965 	ret = igc_rx_init(dev);
966 	if (ret) {
967 		PMD_DRV_LOG(ERR, "Unable to initialize RX hardware");
968 		igc_dev_clear_queues(dev);
969 		return ret;
970 	}
971 
972 	igc_clear_hw_cntrs_base_generic(hw);
973 
974 	/* VLAN Offload Settings */
975 	eth_igc_vlan_offload_set(dev,
976 		ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
977 		ETH_VLAN_EXTEND_MASK);
978 
979 	/* Setup link speed and duplex */
980 	speeds = &dev->data->dev_conf.link_speeds;
981 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
982 		hw->phy.autoneg_advertised = IGC_ALL_SPEED_DUPLEX_2500;
983 		hw->mac.autoneg = 1;
984 	} else {
985 		int num_speeds = 0;
986 		bool autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
987 
988 		/* Reset */
989 		hw->phy.autoneg_advertised = 0;
990 
991 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
992 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
993 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_2_5G |
994 				ETH_LINK_SPEED_FIXED)) {
995 			num_speeds = -1;
996 			goto error_invalid_config;
997 		}
998 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
999 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1000 			num_speeds++;
1001 		}
1002 		if (*speeds & ETH_LINK_SPEED_10M) {
1003 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1004 			num_speeds++;
1005 		}
1006 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
1007 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1008 			num_speeds++;
1009 		}
1010 		if (*speeds & ETH_LINK_SPEED_100M) {
1011 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1012 			num_speeds++;
1013 		}
1014 		if (*speeds & ETH_LINK_SPEED_1G) {
1015 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1016 			num_speeds++;
1017 		}
1018 		if (*speeds & ETH_LINK_SPEED_2_5G) {
1019 			hw->phy.autoneg_advertised |= ADVERTISE_2500_FULL;
1020 			num_speeds++;
1021 		}
1022 		if (num_speeds == 0 || (!autoneg && num_speeds > 1))
1023 			goto error_invalid_config;
1024 
1025 		/* Set/reset the mac.autoneg based on the link speed,
1026 		 * fixed or not
1027 		 */
1028 		if (!autoneg) {
1029 			hw->mac.autoneg = 0;
1030 			hw->mac.forced_speed_duplex =
1031 					hw->phy.autoneg_advertised;
1032 		} else {
1033 			hw->mac.autoneg = 1;
1034 		}
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->data->rx_queues[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->data->tx_queues[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 
1177 	PMD_INIT_FUNC_TRACE();
1178 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1179 		return 0;
1180 
1181 	if (!adapter->stopped)
1182 		eth_igc_stop(dev);
1183 
1184 	igc_flow_flush(dev, NULL);
1185 	igc_clear_all_filter(dev);
1186 
1187 	igc_intr_other_disable(dev);
1188 	do {
1189 		int ret = rte_intr_callback_unregister(intr_handle,
1190 				eth_igc_interrupt_handler, dev);
1191 		if (ret >= 0 || ret == -ENOENT || ret == -EINVAL)
1192 			break;
1193 
1194 		PMD_DRV_LOG(ERR, "intr callback unregister failed: %d", ret);
1195 		DELAY(200 * 1000); /* delay 200ms */
1196 	} while (retry++ < 5);
1197 
1198 	igc_phy_hw_reset(hw);
1199 	igc_hw_control_release(hw);
1200 	igc_dev_free_queues(dev);
1201 
1202 	/* Reset any pending lock */
1203 	igc_reset_swfw_lock(hw);
1204 
1205 	return 0;
1206 }
1207 
1208 static void
1209 igc_identify_hardware(struct rte_eth_dev *dev, struct rte_pci_device *pci_dev)
1210 {
1211 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1212 
1213 	hw->vendor_id = pci_dev->id.vendor_id;
1214 	hw->device_id = pci_dev->id.device_id;
1215 	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
1216 	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
1217 }
1218 
1219 static int
1220 eth_igc_dev_init(struct rte_eth_dev *dev)
1221 {
1222 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1223 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
1224 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1225 	int i, error = 0;
1226 
1227 	PMD_INIT_FUNC_TRACE();
1228 	dev->dev_ops = &eth_igc_ops;
1229 	dev->rx_descriptor_done	= eth_igc_rx_descriptor_done;
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 
1244 	hw->back = pci_dev;
1245 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1246 
1247 	igc_identify_hardware(dev, pci_dev);
1248 	if (igc_setup_init_funcs(hw, false) != IGC_SUCCESS) {
1249 		error = -EIO;
1250 		goto err_late;
1251 	}
1252 
1253 	igc_get_bus_info(hw);
1254 
1255 	/* Reset any pending lock */
1256 	if (igc_reset_swfw_lock(hw) != IGC_SUCCESS) {
1257 		error = -EIO;
1258 		goto err_late;
1259 	}
1260 
1261 	/* Finish initialization */
1262 	if (igc_setup_init_funcs(hw, true) != IGC_SUCCESS) {
1263 		error = -EIO;
1264 		goto err_late;
1265 	}
1266 
1267 	hw->mac.autoneg = 1;
1268 	hw->phy.autoneg_wait_to_complete = 0;
1269 	hw->phy.autoneg_advertised = IGC_ALL_SPEED_DUPLEX_2500;
1270 
1271 	/* Copper options */
1272 	if (hw->phy.media_type == igc_media_type_copper) {
1273 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
1274 		hw->phy.disable_polarity_correction = 0;
1275 		hw->phy.ms_type = igc_ms_hw_default;
1276 	}
1277 
1278 	/*
1279 	 * Start from a known state, this is important in reading the nvm
1280 	 * and mac from that.
1281 	 */
1282 	igc_reset_hw(hw);
1283 
1284 	/* Make sure we have a good EEPROM before we read from it */
1285 	if (igc_validate_nvm_checksum(hw) < 0) {
1286 		/*
1287 		 * Some PCI-E parts fail the first check due to
1288 		 * the link being in sleep state, call it again,
1289 		 * if it fails a second time its a real issue.
1290 		 */
1291 		if (igc_validate_nvm_checksum(hw) < 0) {
1292 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
1293 			error = -EIO;
1294 			goto err_late;
1295 		}
1296 	}
1297 
1298 	/* Read the permanent MAC address out of the EEPROM */
1299 	if (igc_read_mac_addr(hw) != 0) {
1300 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
1301 		error = -EIO;
1302 		goto err_late;
1303 	}
1304 
1305 	/* Allocate memory for storing MAC addresses */
1306 	dev->data->mac_addrs = rte_zmalloc("igc",
1307 		RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count, 0);
1308 	if (dev->data->mac_addrs == NULL) {
1309 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes for storing MAC",
1310 				RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
1311 		error = -ENOMEM;
1312 		goto err_late;
1313 	}
1314 
1315 	/* Copy the permanent MAC address */
1316 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
1317 			&dev->data->mac_addrs[0]);
1318 
1319 	/* Now initialize the hardware */
1320 	if (igc_hardware_init(hw) != 0) {
1321 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
1322 		rte_free(dev->data->mac_addrs);
1323 		dev->data->mac_addrs = NULL;
1324 		error = -ENODEV;
1325 		goto err_late;
1326 	}
1327 
1328 	hw->mac.get_link_status = 1;
1329 	igc->stopped = 0;
1330 
1331 	/* Indicate SOL/IDER usage */
1332 	if (igc_check_reset_block(hw) < 0)
1333 		PMD_INIT_LOG(ERR,
1334 			"PHY reset is blocked due to SOL/IDER session.");
1335 
1336 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
1337 			dev->data->port_id, pci_dev->id.vendor_id,
1338 			pci_dev->id.device_id);
1339 
1340 	rte_intr_callback_register(&pci_dev->intr_handle,
1341 			eth_igc_interrupt_handler, (void *)dev);
1342 
1343 	/* enable uio/vfio intr/eventfd mapping */
1344 	rte_intr_enable(&pci_dev->intr_handle);
1345 
1346 	/* enable support intr */
1347 	igc_intr_other_enable(dev);
1348 
1349 	/* initiate queue status */
1350 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1351 		igc->txq_stats_map[i] = -1;
1352 		igc->rxq_stats_map[i] = -1;
1353 	}
1354 
1355 	igc_flow_init(dev);
1356 	igc_clear_all_filter(dev);
1357 	return 0;
1358 
1359 err_late:
1360 	igc_hw_control_release(hw);
1361 	return error;
1362 }
1363 
1364 static int
1365 eth_igc_dev_uninit(__rte_unused struct rte_eth_dev *eth_dev)
1366 {
1367 	PMD_INIT_FUNC_TRACE();
1368 	eth_igc_close(eth_dev);
1369 	return 0;
1370 }
1371 
1372 static int
1373 eth_igc_reset(struct rte_eth_dev *dev)
1374 {
1375 	int ret;
1376 
1377 	PMD_INIT_FUNC_TRACE();
1378 
1379 	ret = eth_igc_dev_uninit(dev);
1380 	if (ret)
1381 		return ret;
1382 
1383 	return eth_igc_dev_init(dev);
1384 }
1385 
1386 static int
1387 eth_igc_promiscuous_enable(struct rte_eth_dev *dev)
1388 {
1389 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1390 	uint32_t rctl;
1391 
1392 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1393 	rctl |= (IGC_RCTL_UPE | IGC_RCTL_MPE);
1394 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1395 	return 0;
1396 }
1397 
1398 static int
1399 eth_igc_promiscuous_disable(struct rte_eth_dev *dev)
1400 {
1401 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1402 	uint32_t rctl;
1403 
1404 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1405 	rctl &= (~IGC_RCTL_UPE);
1406 	if (dev->data->all_multicast == 1)
1407 		rctl |= IGC_RCTL_MPE;
1408 	else
1409 		rctl &= (~IGC_RCTL_MPE);
1410 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1411 	return 0;
1412 }
1413 
1414 static int
1415 eth_igc_allmulticast_enable(struct rte_eth_dev *dev)
1416 {
1417 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1418 	uint32_t rctl;
1419 
1420 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1421 	rctl |= IGC_RCTL_MPE;
1422 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1423 	return 0;
1424 }
1425 
1426 static int
1427 eth_igc_allmulticast_disable(struct rte_eth_dev *dev)
1428 {
1429 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1430 	uint32_t rctl;
1431 
1432 	if (dev->data->promiscuous == 1)
1433 		return 0;	/* must remain in all_multicast mode */
1434 
1435 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1436 	rctl &= (~IGC_RCTL_MPE);
1437 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1438 	return 0;
1439 }
1440 
1441 static int
1442 eth_igc_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
1443 		       size_t fw_size)
1444 {
1445 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1446 	struct igc_fw_version fw;
1447 	int ret;
1448 
1449 	igc_get_fw_version(hw, &fw);
1450 
1451 	/* if option rom is valid, display its version too */
1452 	if (fw.or_valid) {
1453 		ret = snprintf(fw_version, fw_size,
1454 			 "%d.%d, 0x%08x, %d.%d.%d",
1455 			 fw.eep_major, fw.eep_minor, fw.etrack_id,
1456 			 fw.or_major, fw.or_build, fw.or_patch);
1457 	/* no option rom */
1458 	} else {
1459 		if (fw.etrack_id != 0X0000) {
1460 			ret = snprintf(fw_version, fw_size,
1461 				 "%d.%d, 0x%08x",
1462 				 fw.eep_major, fw.eep_minor,
1463 				 fw.etrack_id);
1464 		} else {
1465 			ret = snprintf(fw_version, fw_size,
1466 				 "%d.%d.%d",
1467 				 fw.eep_major, fw.eep_minor,
1468 				 fw.eep_build);
1469 		}
1470 	}
1471 
1472 	ret += 1; /* add the size of '\0' */
1473 	if (fw_size < (u32)ret)
1474 		return ret;
1475 	else
1476 		return 0;
1477 }
1478 
1479 static int
1480 eth_igc_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1481 {
1482 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1483 
1484 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1485 	dev_info->max_rx_pktlen = MAX_RX_JUMBO_FRAME_SIZE;
1486 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1487 	dev_info->rx_offload_capa = IGC_RX_OFFLOAD_ALL;
1488 	dev_info->tx_offload_capa = IGC_TX_OFFLOAD_ALL;
1489 	dev_info->rx_queue_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP;
1490 
1491 	dev_info->max_rx_queues = IGC_QUEUE_PAIRS_NUM;
1492 	dev_info->max_tx_queues = IGC_QUEUE_PAIRS_NUM;
1493 	dev_info->max_vmdq_pools = 0;
1494 
1495 	dev_info->hash_key_size = IGC_HKEY_MAX_INDEX * sizeof(uint32_t);
1496 	dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
1497 	dev_info->flow_type_rss_offloads = IGC_RSS_OFFLOAD_ALL;
1498 
1499 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
1500 		.rx_thresh = {
1501 			.pthresh = IGC_DEFAULT_RX_PTHRESH,
1502 			.hthresh = IGC_DEFAULT_RX_HTHRESH,
1503 			.wthresh = IGC_DEFAULT_RX_WTHRESH,
1504 		},
1505 		.rx_free_thresh = IGC_DEFAULT_RX_FREE_THRESH,
1506 		.rx_drop_en = 0,
1507 		.offloads = 0,
1508 	};
1509 
1510 	dev_info->default_txconf = (struct rte_eth_txconf) {
1511 		.tx_thresh = {
1512 			.pthresh = IGC_DEFAULT_TX_PTHRESH,
1513 			.hthresh = IGC_DEFAULT_TX_HTHRESH,
1514 			.wthresh = IGC_DEFAULT_TX_WTHRESH,
1515 		},
1516 		.offloads = 0,
1517 	};
1518 
1519 	dev_info->rx_desc_lim = rx_desc_lim;
1520 	dev_info->tx_desc_lim = tx_desc_lim;
1521 
1522 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1523 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1524 			ETH_LINK_SPEED_1G | ETH_LINK_SPEED_2_5G;
1525 
1526 	dev_info->max_mtu = dev_info->max_rx_pktlen - IGC_ETH_OVERHEAD;
1527 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
1528 	return 0;
1529 }
1530 
1531 static int
1532 eth_igc_led_on(struct rte_eth_dev *dev)
1533 {
1534 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1535 
1536 	return igc_led_on(hw) == IGC_SUCCESS ? 0 : -ENOTSUP;
1537 }
1538 
1539 static int
1540 eth_igc_led_off(struct rte_eth_dev *dev)
1541 {
1542 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1543 
1544 	return igc_led_off(hw) == IGC_SUCCESS ? 0 : -ENOTSUP;
1545 }
1546 
1547 static const uint32_t *
1548 eth_igc_supported_ptypes_get(__rte_unused struct rte_eth_dev *dev)
1549 {
1550 	static const uint32_t ptypes[] = {
1551 		/* refers to rx_desc_pkt_info_to_pkt_type() */
1552 		RTE_PTYPE_L2_ETHER,
1553 		RTE_PTYPE_L3_IPV4,
1554 		RTE_PTYPE_L3_IPV4_EXT,
1555 		RTE_PTYPE_L3_IPV6,
1556 		RTE_PTYPE_L3_IPV6_EXT,
1557 		RTE_PTYPE_L4_TCP,
1558 		RTE_PTYPE_L4_UDP,
1559 		RTE_PTYPE_L4_SCTP,
1560 		RTE_PTYPE_TUNNEL_IP,
1561 		RTE_PTYPE_INNER_L3_IPV6,
1562 		RTE_PTYPE_INNER_L3_IPV6_EXT,
1563 		RTE_PTYPE_INNER_L4_TCP,
1564 		RTE_PTYPE_INNER_L4_UDP,
1565 		RTE_PTYPE_UNKNOWN
1566 	};
1567 
1568 	return ptypes;
1569 }
1570 
1571 static int
1572 eth_igc_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1573 {
1574 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1575 	uint32_t frame_size = mtu + IGC_ETH_OVERHEAD;
1576 	uint32_t rctl;
1577 
1578 	/* if extend vlan has been enabled */
1579 	if (IGC_READ_REG(hw, IGC_CTRL_EXT) & IGC_CTRL_EXT_EXT_VLAN)
1580 		frame_size += VLAN_TAG_SIZE;
1581 
1582 	/* check that mtu is within the allowed range */
1583 	if (mtu < RTE_ETHER_MIN_MTU ||
1584 		frame_size > MAX_RX_JUMBO_FRAME_SIZE)
1585 		return -EINVAL;
1586 
1587 	/*
1588 	 * refuse mtu that requires the support of scattered packets when
1589 	 * this feature has not been enabled before.
1590 	 */
1591 	if (!dev->data->scattered_rx &&
1592 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
1593 		return -EINVAL;
1594 
1595 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1596 
1597 	/* switch to jumbo mode if needed */
1598 	if (mtu > RTE_ETHER_MTU) {
1599 		dev->data->dev_conf.rxmode.offloads |=
1600 			DEV_RX_OFFLOAD_JUMBO_FRAME;
1601 		rctl |= IGC_RCTL_LPE;
1602 	} else {
1603 		dev->data->dev_conf.rxmode.offloads &=
1604 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
1605 		rctl &= ~IGC_RCTL_LPE;
1606 	}
1607 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1608 
1609 	/* update max frame size */
1610 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1611 
1612 	IGC_WRITE_REG(hw, IGC_RLPML,
1613 			dev->data->dev_conf.rxmode.max_rx_pkt_len);
1614 
1615 	return 0;
1616 }
1617 
1618 static int
1619 eth_igc_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
1620 		uint32_t index, uint32_t pool)
1621 {
1622 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1623 
1624 	igc_rar_set(hw, mac_addr->addr_bytes, index);
1625 	RTE_SET_USED(pool);
1626 	return 0;
1627 }
1628 
1629 static void
1630 eth_igc_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1631 {
1632 	uint8_t addr[RTE_ETHER_ADDR_LEN];
1633 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1634 
1635 	memset(addr, 0, sizeof(addr));
1636 	igc_rar_set(hw, addr, index);
1637 }
1638 
1639 static int
1640 eth_igc_default_mac_addr_set(struct rte_eth_dev *dev,
1641 			struct rte_ether_addr *addr)
1642 {
1643 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1644 	igc_rar_set(hw, addr->addr_bytes, 0);
1645 	return 0;
1646 }
1647 
1648 static int
1649 eth_igc_set_mc_addr_list(struct rte_eth_dev *dev,
1650 			 struct rte_ether_addr *mc_addr_set,
1651 			 uint32_t nb_mc_addr)
1652 {
1653 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1654 	igc_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1655 	return 0;
1656 }
1657 
1658 /*
1659  * Read hardware registers
1660  */
1661 static void
1662 igc_read_stats_registers(struct igc_hw *hw, struct igc_hw_stats *stats)
1663 {
1664 	int pause_frames;
1665 
1666 	uint64_t old_gprc  = stats->gprc;
1667 	uint64_t old_gptc  = stats->gptc;
1668 	uint64_t old_tpr   = stats->tpr;
1669 	uint64_t old_tpt   = stats->tpt;
1670 	uint64_t old_rpthc = stats->rpthc;
1671 	uint64_t old_hgptc = stats->hgptc;
1672 
1673 	stats->crcerrs += IGC_READ_REG(hw, IGC_CRCERRS);
1674 	stats->algnerrc += IGC_READ_REG(hw, IGC_ALGNERRC);
1675 	stats->rxerrc += IGC_READ_REG(hw, IGC_RXERRC);
1676 	stats->mpc += IGC_READ_REG(hw, IGC_MPC);
1677 	stats->scc += IGC_READ_REG(hw, IGC_SCC);
1678 	stats->ecol += IGC_READ_REG(hw, IGC_ECOL);
1679 
1680 	stats->mcc += IGC_READ_REG(hw, IGC_MCC);
1681 	stats->latecol += IGC_READ_REG(hw, IGC_LATECOL);
1682 	stats->colc += IGC_READ_REG(hw, IGC_COLC);
1683 
1684 	stats->dc += IGC_READ_REG(hw, IGC_DC);
1685 	stats->tncrs += IGC_READ_REG(hw, IGC_TNCRS);
1686 	stats->htdpmc += IGC_READ_REG(hw, IGC_HTDPMC);
1687 	stats->rlec += IGC_READ_REG(hw, IGC_RLEC);
1688 	stats->xonrxc += IGC_READ_REG(hw, IGC_XONRXC);
1689 	stats->xontxc += IGC_READ_REG(hw, IGC_XONTXC);
1690 
1691 	/*
1692 	 * For watchdog management we need to know if we have been
1693 	 * paused during the last interval, so capture that here.
1694 	 */
1695 	pause_frames = IGC_READ_REG(hw, IGC_XOFFRXC);
1696 	stats->xoffrxc += pause_frames;
1697 	stats->xofftxc += IGC_READ_REG(hw, IGC_XOFFTXC);
1698 	stats->fcruc += IGC_READ_REG(hw, IGC_FCRUC);
1699 	stats->prc64 += IGC_READ_REG(hw, IGC_PRC64);
1700 	stats->prc127 += IGC_READ_REG(hw, IGC_PRC127);
1701 	stats->prc255 += IGC_READ_REG(hw, IGC_PRC255);
1702 	stats->prc511 += IGC_READ_REG(hw, IGC_PRC511);
1703 	stats->prc1023 += IGC_READ_REG(hw, IGC_PRC1023);
1704 	stats->prc1522 += IGC_READ_REG(hw, IGC_PRC1522);
1705 	stats->gprc += IGC_READ_REG(hw, IGC_GPRC);
1706 	stats->bprc += IGC_READ_REG(hw, IGC_BPRC);
1707 	stats->mprc += IGC_READ_REG(hw, IGC_MPRC);
1708 	stats->gptc += IGC_READ_REG(hw, IGC_GPTC);
1709 
1710 	/* For the 64-bit byte counters the low dword must be read first. */
1711 	/* Both registers clear on the read of the high dword */
1712 
1713 	/* Workaround CRC bytes included in size, take away 4 bytes/packet */
1714 	stats->gorc += IGC_READ_REG(hw, IGC_GORCL);
1715 	stats->gorc += ((uint64_t)IGC_READ_REG(hw, IGC_GORCH) << 32);
1716 	stats->gorc -= (stats->gprc - old_gprc) * RTE_ETHER_CRC_LEN;
1717 	stats->gotc += IGC_READ_REG(hw, IGC_GOTCL);
1718 	stats->gotc += ((uint64_t)IGC_READ_REG(hw, IGC_GOTCH) << 32);
1719 	stats->gotc -= (stats->gptc - old_gptc) * RTE_ETHER_CRC_LEN;
1720 
1721 	stats->rnbc += IGC_READ_REG(hw, IGC_RNBC);
1722 	stats->ruc += IGC_READ_REG(hw, IGC_RUC);
1723 	stats->rfc += IGC_READ_REG(hw, IGC_RFC);
1724 	stats->roc += IGC_READ_REG(hw, IGC_ROC);
1725 	stats->rjc += IGC_READ_REG(hw, IGC_RJC);
1726 
1727 	stats->mgprc += IGC_READ_REG(hw, IGC_MGTPRC);
1728 	stats->mgpdc += IGC_READ_REG(hw, IGC_MGTPDC);
1729 	stats->mgptc += IGC_READ_REG(hw, IGC_MGTPTC);
1730 	stats->b2ospc += IGC_READ_REG(hw, IGC_B2OSPC);
1731 	stats->b2ogprc += IGC_READ_REG(hw, IGC_B2OGPRC);
1732 	stats->o2bgptc += IGC_READ_REG(hw, IGC_O2BGPTC);
1733 	stats->o2bspc += IGC_READ_REG(hw, IGC_O2BSPC);
1734 
1735 	stats->tpr += IGC_READ_REG(hw, IGC_TPR);
1736 	stats->tpt += IGC_READ_REG(hw, IGC_TPT);
1737 
1738 	stats->tor += IGC_READ_REG(hw, IGC_TORL);
1739 	stats->tor += ((uint64_t)IGC_READ_REG(hw, IGC_TORH) << 32);
1740 	stats->tor -= (stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
1741 	stats->tot += IGC_READ_REG(hw, IGC_TOTL);
1742 	stats->tot += ((uint64_t)IGC_READ_REG(hw, IGC_TOTH) << 32);
1743 	stats->tot -= (stats->tpt - old_tpt) * RTE_ETHER_CRC_LEN;
1744 
1745 	stats->ptc64 += IGC_READ_REG(hw, IGC_PTC64);
1746 	stats->ptc127 += IGC_READ_REG(hw, IGC_PTC127);
1747 	stats->ptc255 += IGC_READ_REG(hw, IGC_PTC255);
1748 	stats->ptc511 += IGC_READ_REG(hw, IGC_PTC511);
1749 	stats->ptc1023 += IGC_READ_REG(hw, IGC_PTC1023);
1750 	stats->ptc1522 += IGC_READ_REG(hw, IGC_PTC1522);
1751 	stats->mptc += IGC_READ_REG(hw, IGC_MPTC);
1752 	stats->bptc += IGC_READ_REG(hw, IGC_BPTC);
1753 	stats->tsctc += IGC_READ_REG(hw, IGC_TSCTC);
1754 
1755 	stats->iac += IGC_READ_REG(hw, IGC_IAC);
1756 	stats->rpthc += IGC_READ_REG(hw, IGC_RPTHC);
1757 	stats->hgptc += IGC_READ_REG(hw, IGC_HGPTC);
1758 	stats->icrxdmtc += IGC_READ_REG(hw, IGC_ICRXDMTC);
1759 
1760 	/* Host to Card Statistics */
1761 	stats->hgorc += IGC_READ_REG(hw, IGC_HGORCL);
1762 	stats->hgorc += ((uint64_t)IGC_READ_REG(hw, IGC_HGORCH) << 32);
1763 	stats->hgorc -= (stats->rpthc - old_rpthc) * RTE_ETHER_CRC_LEN;
1764 	stats->hgotc += IGC_READ_REG(hw, IGC_HGOTCL);
1765 	stats->hgotc += ((uint64_t)IGC_READ_REG(hw, IGC_HGOTCH) << 32);
1766 	stats->hgotc -= (stats->hgptc - old_hgptc) * RTE_ETHER_CRC_LEN;
1767 	stats->lenerrs += IGC_READ_REG(hw, IGC_LENERRS);
1768 }
1769 
1770 /*
1771  * Write 0 to all queue status registers
1772  */
1773 static void
1774 igc_reset_queue_stats_register(struct igc_hw *hw)
1775 {
1776 	int i;
1777 
1778 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1779 		IGC_WRITE_REG(hw, IGC_PQGPRC(i), 0);
1780 		IGC_WRITE_REG(hw, IGC_PQGPTC(i), 0);
1781 		IGC_WRITE_REG(hw, IGC_PQGORC(i), 0);
1782 		IGC_WRITE_REG(hw, IGC_PQGOTC(i), 0);
1783 		IGC_WRITE_REG(hw, IGC_PQMPRC(i), 0);
1784 		IGC_WRITE_REG(hw, IGC_RQDPC(i), 0);
1785 		IGC_WRITE_REG(hw, IGC_TQDPC(i), 0);
1786 	}
1787 }
1788 
1789 /*
1790  * Read all hardware queue status registers
1791  */
1792 static void
1793 igc_read_queue_stats_register(struct rte_eth_dev *dev)
1794 {
1795 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1796 	struct igc_hw_queue_stats *queue_stats =
1797 				IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1798 	int i;
1799 
1800 	/*
1801 	 * This register is not cleared on read. Furthermore, the register wraps
1802 	 * around back to 0x00000000 on the next increment when reaching a value
1803 	 * of 0xFFFFFFFF and then continues normal count operation.
1804 	 */
1805 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1806 		union {
1807 			u64 ddword;
1808 			u32 dword[2];
1809 		} value;
1810 		u32 tmp;
1811 
1812 		/*
1813 		 * Read the register first, if the value is smaller than that
1814 		 * previous read, that mean the register has been overflowed,
1815 		 * then we add the high 4 bytes by 1 and replace the low 4
1816 		 * bytes by the new value.
1817 		 */
1818 		tmp = IGC_READ_REG(hw, IGC_PQGPRC(i));
1819 		value.ddword = queue_stats->pqgprc[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->pqgprc[i] = value.ddword;
1824 
1825 		tmp = IGC_READ_REG(hw, IGC_PQGPTC(i));
1826 		value.ddword = queue_stats->pqgptc[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->pqgptc[i] = value.ddword;
1831 
1832 		tmp = IGC_READ_REG(hw, IGC_PQGORC(i));
1833 		value.ddword = queue_stats->pqgorc[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->pqgorc[i] = value.ddword;
1838 
1839 		tmp = IGC_READ_REG(hw, IGC_PQGOTC(i));
1840 		value.ddword = queue_stats->pqgotc[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->pqgotc[i] = value.ddword;
1845 
1846 		tmp = IGC_READ_REG(hw, IGC_PQMPRC(i));
1847 		value.ddword = queue_stats->pqmprc[i];
1848 		if (value.dword[U32_0_IN_U64] > tmp)
1849 			value.dword[U32_1_IN_U64]++;
1850 		value.dword[U32_0_IN_U64] = tmp;
1851 		queue_stats->pqmprc[i] = value.ddword;
1852 
1853 		tmp = IGC_READ_REG(hw, IGC_RQDPC(i));
1854 		value.ddword = queue_stats->rqdpc[i];
1855 		if (value.dword[U32_0_IN_U64] > tmp)
1856 			value.dword[U32_1_IN_U64]++;
1857 		value.dword[U32_0_IN_U64] = tmp;
1858 		queue_stats->rqdpc[i] = value.ddword;
1859 
1860 		tmp = IGC_READ_REG(hw, IGC_TQDPC(i));
1861 		value.ddword = queue_stats->tqdpc[i];
1862 		if (value.dword[U32_0_IN_U64] > tmp)
1863 			value.dword[U32_1_IN_U64]++;
1864 		value.dword[U32_0_IN_U64] = tmp;
1865 		queue_stats->tqdpc[i] = value.ddword;
1866 	}
1867 }
1868 
1869 static int
1870 eth_igc_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1871 {
1872 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
1873 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1874 	struct igc_hw_stats *stats = IGC_DEV_PRIVATE_STATS(dev);
1875 	struct igc_hw_queue_stats *queue_stats =
1876 			IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1877 	int i;
1878 
1879 	/*
1880 	 * Cancel status handler since it will read the queue status registers
1881 	 */
1882 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
1883 
1884 	/* Read status register */
1885 	igc_read_queue_stats_register(dev);
1886 	igc_read_stats_registers(hw, stats);
1887 
1888 	if (rte_stats == NULL) {
1889 		/* Restart queue status handler */
1890 		rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1891 				igc_update_queue_stats_handler, dev);
1892 		return -EINVAL;
1893 	}
1894 
1895 	/* Rx Errors */
1896 	rte_stats->imissed = stats->mpc;
1897 	rte_stats->ierrors = stats->crcerrs +
1898 			stats->rlec + stats->ruc + stats->roc +
1899 			stats->rxerrc + stats->algnerrc;
1900 
1901 	/* Tx Errors */
1902 	rte_stats->oerrors = stats->ecol + stats->latecol;
1903 
1904 	rte_stats->ipackets = stats->gprc;
1905 	rte_stats->opackets = stats->gptc;
1906 	rte_stats->ibytes   = stats->gorc;
1907 	rte_stats->obytes   = stats->gotc;
1908 
1909 	/* Get per-queue statuses */
1910 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1911 		/* GET TX queue statuses */
1912 		int map_id = igc->txq_stats_map[i];
1913 		if (map_id >= 0) {
1914 			rte_stats->q_opackets[map_id] += queue_stats->pqgptc[i];
1915 			rte_stats->q_obytes[map_id] += queue_stats->pqgotc[i];
1916 		}
1917 		/* Get RX queue statuses */
1918 		map_id = igc->rxq_stats_map[i];
1919 		if (map_id >= 0) {
1920 			rte_stats->q_ipackets[map_id] += queue_stats->pqgprc[i];
1921 			rte_stats->q_ibytes[map_id] += queue_stats->pqgorc[i];
1922 			rte_stats->q_errors[map_id] += queue_stats->rqdpc[i];
1923 		}
1924 	}
1925 
1926 	/* Restart queue status handler */
1927 	rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1928 			igc_update_queue_stats_handler, dev);
1929 	return 0;
1930 }
1931 
1932 static int
1933 eth_igc_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1934 		   unsigned int n)
1935 {
1936 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1937 	struct igc_hw_stats *hw_stats =
1938 			IGC_DEV_PRIVATE_STATS(dev);
1939 	unsigned int i;
1940 
1941 	igc_read_stats_registers(hw, hw_stats);
1942 
1943 	if (n < IGC_NB_XSTATS)
1944 		return IGC_NB_XSTATS;
1945 
1946 	/* If this is a reset xstats is NULL, and we have cleared the
1947 	 * registers by reading them.
1948 	 */
1949 	if (!xstats)
1950 		return 0;
1951 
1952 	/* Extended stats */
1953 	for (i = 0; i < IGC_NB_XSTATS; i++) {
1954 		xstats[i].id = i;
1955 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1956 			rte_igc_stats_strings[i].offset);
1957 	}
1958 
1959 	return IGC_NB_XSTATS;
1960 }
1961 
1962 static int
1963 eth_igc_xstats_reset(struct rte_eth_dev *dev)
1964 {
1965 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1966 	struct igc_hw_stats *hw_stats = IGC_DEV_PRIVATE_STATS(dev);
1967 	struct igc_hw_queue_stats *queue_stats =
1968 			IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1969 
1970 	/* Cancel queue status handler for avoid conflict */
1971 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
1972 
1973 	/* HW registers are cleared on read */
1974 	igc_reset_queue_stats_register(hw);
1975 	igc_read_stats_registers(hw, hw_stats);
1976 
1977 	/* Reset software totals */
1978 	memset(hw_stats, 0, sizeof(*hw_stats));
1979 	memset(queue_stats, 0, sizeof(*queue_stats));
1980 
1981 	/* Restart the queue status handler */
1982 	rte_eal_alarm_set(IGC_ALARM_INTERVAL, igc_update_queue_stats_handler,
1983 			dev);
1984 
1985 	return 0;
1986 }
1987 
1988 static int
1989 eth_igc_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1990 	struct rte_eth_xstat_name *xstats_names, unsigned int size)
1991 {
1992 	unsigned int i;
1993 
1994 	if (xstats_names == NULL)
1995 		return IGC_NB_XSTATS;
1996 
1997 	if (size < IGC_NB_XSTATS) {
1998 		PMD_DRV_LOG(ERR, "not enough buffers!");
1999 		return IGC_NB_XSTATS;
2000 	}
2001 
2002 	for (i = 0; i < IGC_NB_XSTATS; i++)
2003 		strlcpy(xstats_names[i].name, rte_igc_stats_strings[i].name,
2004 			sizeof(xstats_names[i].name));
2005 
2006 	return IGC_NB_XSTATS;
2007 }
2008 
2009 static int
2010 eth_igc_xstats_get_names_by_id(struct rte_eth_dev *dev,
2011 		struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
2012 		unsigned int limit)
2013 {
2014 	unsigned int i;
2015 
2016 	if (!ids)
2017 		return eth_igc_xstats_get_names(dev, xstats_names, limit);
2018 
2019 	for (i = 0; i < limit; i++) {
2020 		if (ids[i] >= IGC_NB_XSTATS) {
2021 			PMD_DRV_LOG(ERR, "id value isn't valid");
2022 			return -EINVAL;
2023 		}
2024 		strlcpy(xstats_names[i].name,
2025 			rte_igc_stats_strings[ids[i]].name,
2026 			sizeof(xstats_names[i].name));
2027 	}
2028 	return limit;
2029 }
2030 
2031 static int
2032 eth_igc_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
2033 		uint64_t *values, unsigned int n)
2034 {
2035 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2036 	struct igc_hw_stats *hw_stats = IGC_DEV_PRIVATE_STATS(dev);
2037 	unsigned int i;
2038 
2039 	igc_read_stats_registers(hw, hw_stats);
2040 
2041 	if (!ids) {
2042 		if (n < IGC_NB_XSTATS)
2043 			return IGC_NB_XSTATS;
2044 
2045 		/* If this is a reset xstats is NULL, and we have cleared the
2046 		 * registers by reading them.
2047 		 */
2048 		if (!values)
2049 			return 0;
2050 
2051 		/* Extended stats */
2052 		for (i = 0; i < IGC_NB_XSTATS; i++)
2053 			values[i] = *(uint64_t *)(((char *)hw_stats) +
2054 					rte_igc_stats_strings[i].offset);
2055 
2056 		return IGC_NB_XSTATS;
2057 
2058 	} else {
2059 		for (i = 0; i < n; i++) {
2060 			if (ids[i] >= IGC_NB_XSTATS) {
2061 				PMD_DRV_LOG(ERR, "id value isn't valid");
2062 				return -EINVAL;
2063 			}
2064 			values[i] = *(uint64_t *)(((char *)hw_stats) +
2065 					rte_igc_stats_strings[ids[i]].offset);
2066 		}
2067 		return n;
2068 	}
2069 }
2070 
2071 static int
2072 eth_igc_queue_stats_mapping_set(struct rte_eth_dev *dev,
2073 		uint16_t queue_id, uint8_t stat_idx, uint8_t is_rx)
2074 {
2075 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
2076 
2077 	/* check queue id is valid */
2078 	if (queue_id >= IGC_QUEUE_PAIRS_NUM) {
2079 		PMD_DRV_LOG(ERR, "queue id(%u) error, max is %u",
2080 			queue_id, IGC_QUEUE_PAIRS_NUM - 1);
2081 		return -EINVAL;
2082 	}
2083 
2084 	/* store the mapping status id */
2085 	if (is_rx)
2086 		igc->rxq_stats_map[queue_id] = stat_idx;
2087 	else
2088 		igc->txq_stats_map[queue_id] = stat_idx;
2089 
2090 	return 0;
2091 }
2092 
2093 static int
2094 eth_igc_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
2095 {
2096 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2097 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2098 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2099 	uint32_t vec = IGC_MISC_VEC_ID;
2100 
2101 	if (rte_intr_allow_others(intr_handle))
2102 		vec = IGC_RX_VEC_START;
2103 
2104 	uint32_t mask = 1u << (queue_id + vec);
2105 
2106 	IGC_WRITE_REG(hw, IGC_EIMC, mask);
2107 	IGC_WRITE_FLUSH(hw);
2108 
2109 	return 0;
2110 }
2111 
2112 static int
2113 eth_igc_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
2114 {
2115 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2116 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2117 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2118 	uint32_t vec = IGC_MISC_VEC_ID;
2119 
2120 	if (rte_intr_allow_others(intr_handle))
2121 		vec = IGC_RX_VEC_START;
2122 
2123 	uint32_t mask = 1u << (queue_id + vec);
2124 
2125 	IGC_WRITE_REG(hw, IGC_EIMS, mask);
2126 	IGC_WRITE_FLUSH(hw);
2127 
2128 	rte_intr_enable(intr_handle);
2129 
2130 	return 0;
2131 }
2132 
2133 static int
2134 eth_igc_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2135 {
2136 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2137 	uint32_t ctrl;
2138 	int tx_pause;
2139 	int rx_pause;
2140 
2141 	fc_conf->pause_time = hw->fc.pause_time;
2142 	fc_conf->high_water = hw->fc.high_water;
2143 	fc_conf->low_water = hw->fc.low_water;
2144 	fc_conf->send_xon = hw->fc.send_xon;
2145 	fc_conf->autoneg = hw->mac.autoneg;
2146 
2147 	/*
2148 	 * Return rx_pause and tx_pause status according to actual setting of
2149 	 * the TFCE and RFCE bits in the CTRL register.
2150 	 */
2151 	ctrl = IGC_READ_REG(hw, IGC_CTRL);
2152 	if (ctrl & IGC_CTRL_TFCE)
2153 		tx_pause = 1;
2154 	else
2155 		tx_pause = 0;
2156 
2157 	if (ctrl & IGC_CTRL_RFCE)
2158 		rx_pause = 1;
2159 	else
2160 		rx_pause = 0;
2161 
2162 	if (rx_pause && tx_pause)
2163 		fc_conf->mode = RTE_FC_FULL;
2164 	else if (rx_pause)
2165 		fc_conf->mode = RTE_FC_RX_PAUSE;
2166 	else if (tx_pause)
2167 		fc_conf->mode = RTE_FC_TX_PAUSE;
2168 	else
2169 		fc_conf->mode = RTE_FC_NONE;
2170 
2171 	return 0;
2172 }
2173 
2174 static int
2175 eth_igc_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2176 {
2177 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2178 	uint32_t rx_buf_size;
2179 	uint32_t max_high_water;
2180 	uint32_t rctl;
2181 	int err;
2182 
2183 	if (fc_conf->autoneg != hw->mac.autoneg)
2184 		return -ENOTSUP;
2185 
2186 	rx_buf_size = igc_get_rx_buffer_size(hw);
2187 	PMD_DRV_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
2188 
2189 	/* At least reserve one Ethernet frame for watermark */
2190 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
2191 	if (fc_conf->high_water > max_high_water ||
2192 		fc_conf->high_water < fc_conf->low_water) {
2193 		PMD_DRV_LOG(ERR,
2194 			"Incorrect high(%u)/low(%u) water value, max is %u",
2195 			fc_conf->high_water, fc_conf->low_water,
2196 			max_high_water);
2197 		return -EINVAL;
2198 	}
2199 
2200 	switch (fc_conf->mode) {
2201 	case RTE_FC_NONE:
2202 		hw->fc.requested_mode = igc_fc_none;
2203 		break;
2204 	case RTE_FC_RX_PAUSE:
2205 		hw->fc.requested_mode = igc_fc_rx_pause;
2206 		break;
2207 	case RTE_FC_TX_PAUSE:
2208 		hw->fc.requested_mode = igc_fc_tx_pause;
2209 		break;
2210 	case RTE_FC_FULL:
2211 		hw->fc.requested_mode = igc_fc_full;
2212 		break;
2213 	default:
2214 		PMD_DRV_LOG(ERR, "unsupported fc mode: %u", fc_conf->mode);
2215 		return -EINVAL;
2216 	}
2217 
2218 	hw->fc.pause_time     = fc_conf->pause_time;
2219 	hw->fc.high_water     = fc_conf->high_water;
2220 	hw->fc.low_water      = fc_conf->low_water;
2221 	hw->fc.send_xon	      = fc_conf->send_xon;
2222 
2223 	err = igc_setup_link_generic(hw);
2224 	if (err == IGC_SUCCESS) {
2225 		/**
2226 		 * check if we want to forward MAC frames - driver doesn't have
2227 		 * native capability to do that, so we'll write the registers
2228 		 * ourselves
2229 		 **/
2230 		rctl = IGC_READ_REG(hw, IGC_RCTL);
2231 
2232 		/* set or clear MFLCN.PMCF bit depending on configuration */
2233 		if (fc_conf->mac_ctrl_frame_fwd != 0)
2234 			rctl |= IGC_RCTL_PMCF;
2235 		else
2236 			rctl &= ~IGC_RCTL_PMCF;
2237 
2238 		IGC_WRITE_REG(hw, IGC_RCTL, rctl);
2239 		IGC_WRITE_FLUSH(hw);
2240 
2241 		return 0;
2242 	}
2243 
2244 	PMD_DRV_LOG(ERR, "igc_setup_link_generic = 0x%x", err);
2245 	return -EIO;
2246 }
2247 
2248 static int
2249 eth_igc_rss_reta_update(struct rte_eth_dev *dev,
2250 			struct rte_eth_rss_reta_entry64 *reta_conf,
2251 			uint16_t reta_size)
2252 {
2253 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2254 	uint16_t i;
2255 
2256 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
2257 		PMD_DRV_LOG(ERR,
2258 			"The size of RSS redirection table configured(%d) doesn't match the number hardware can supported(%d)",
2259 			reta_size, ETH_RSS_RETA_SIZE_128);
2260 		return -EINVAL;
2261 	}
2262 
2263 	RTE_BUILD_BUG_ON(ETH_RSS_RETA_SIZE_128 % IGC_RSS_RDT_REG_SIZE);
2264 
2265 	/* set redirection table */
2266 	for (i = 0; i < ETH_RSS_RETA_SIZE_128; i += IGC_RSS_RDT_REG_SIZE) {
2267 		union igc_rss_reta_reg reta, reg;
2268 		uint16_t idx, shift;
2269 		uint8_t j, mask;
2270 
2271 		idx = i / RTE_RETA_GROUP_SIZE;
2272 		shift = i % RTE_RETA_GROUP_SIZE;
2273 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
2274 				IGC_RSS_RDT_REG_SIZE_MASK);
2275 
2276 		/* if no need to update the register */
2277 		if (!mask ||
2278 		    shift > (RTE_RETA_GROUP_SIZE - IGC_RSS_RDT_REG_SIZE))
2279 			continue;
2280 
2281 		/* check mask whether need to read the register value first */
2282 		if (mask == IGC_RSS_RDT_REG_SIZE_MASK)
2283 			reg.dword = 0;
2284 		else
2285 			reg.dword = IGC_READ_REG_LE_VALUE(hw,
2286 					IGC_RETA(i / IGC_RSS_RDT_REG_SIZE));
2287 
2288 		/* update the register */
2289 		RTE_BUILD_BUG_ON(sizeof(reta.bytes) != IGC_RSS_RDT_REG_SIZE);
2290 		for (j = 0; j < IGC_RSS_RDT_REG_SIZE; j++) {
2291 			if (mask & (1u << j))
2292 				reta.bytes[j] =
2293 					(uint8_t)reta_conf[idx].reta[shift + j];
2294 			else
2295 				reta.bytes[j] = reg.bytes[j];
2296 		}
2297 		IGC_WRITE_REG_LE_VALUE(hw,
2298 			IGC_RETA(i / IGC_RSS_RDT_REG_SIZE), reta.dword);
2299 	}
2300 
2301 	return 0;
2302 }
2303 
2304 static int
2305 eth_igc_rss_reta_query(struct rte_eth_dev *dev,
2306 		       struct rte_eth_rss_reta_entry64 *reta_conf,
2307 		       uint16_t reta_size)
2308 {
2309 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2310 	uint16_t i;
2311 
2312 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
2313 		PMD_DRV_LOG(ERR,
2314 			"The size of RSS redirection table configured(%d) doesn't match the number hardware can supported(%d)",
2315 			reta_size, ETH_RSS_RETA_SIZE_128);
2316 		return -EINVAL;
2317 	}
2318 
2319 	RTE_BUILD_BUG_ON(ETH_RSS_RETA_SIZE_128 % IGC_RSS_RDT_REG_SIZE);
2320 
2321 	/* read redirection table */
2322 	for (i = 0; i < ETH_RSS_RETA_SIZE_128; i += IGC_RSS_RDT_REG_SIZE) {
2323 		union igc_rss_reta_reg reta;
2324 		uint16_t idx, shift;
2325 		uint8_t j, mask;
2326 
2327 		idx = i / RTE_RETA_GROUP_SIZE;
2328 		shift = i % RTE_RETA_GROUP_SIZE;
2329 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
2330 				IGC_RSS_RDT_REG_SIZE_MASK);
2331 
2332 		/* if no need to read register */
2333 		if (!mask ||
2334 		    shift > (RTE_RETA_GROUP_SIZE - IGC_RSS_RDT_REG_SIZE))
2335 			continue;
2336 
2337 		/* read register and get the queue index */
2338 		RTE_BUILD_BUG_ON(sizeof(reta.bytes) != IGC_RSS_RDT_REG_SIZE);
2339 		reta.dword = IGC_READ_REG_LE_VALUE(hw,
2340 				IGC_RETA(i / IGC_RSS_RDT_REG_SIZE));
2341 		for (j = 0; j < IGC_RSS_RDT_REG_SIZE; j++) {
2342 			if (mask & (1u << j))
2343 				reta_conf[idx].reta[shift + j] = reta.bytes[j];
2344 		}
2345 	}
2346 
2347 	return 0;
2348 }
2349 
2350 static int
2351 eth_igc_rss_hash_update(struct rte_eth_dev *dev,
2352 			struct rte_eth_rss_conf *rss_conf)
2353 {
2354 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2355 	igc_hw_rss_hash_set(hw, rss_conf);
2356 	return 0;
2357 }
2358 
2359 static int
2360 eth_igc_rss_hash_conf_get(struct rte_eth_dev *dev,
2361 			struct rte_eth_rss_conf *rss_conf)
2362 {
2363 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2364 	uint32_t *hash_key = (uint32_t *)rss_conf->rss_key;
2365 	uint32_t mrqc;
2366 	uint64_t rss_hf;
2367 
2368 	if (hash_key != NULL) {
2369 		int i;
2370 
2371 		/* if not enough space for store hash key */
2372 		if (rss_conf->rss_key_len != IGC_HKEY_SIZE) {
2373 			PMD_DRV_LOG(ERR,
2374 				"RSS hash key size %u in parameter doesn't match the hardware hash key size %u",
2375 				rss_conf->rss_key_len, IGC_HKEY_SIZE);
2376 			return -EINVAL;
2377 		}
2378 
2379 		/* read RSS key from register */
2380 		for (i = 0; i < IGC_HKEY_MAX_INDEX; i++)
2381 			hash_key[i] = IGC_READ_REG_LE_VALUE(hw, IGC_RSSRK(i));
2382 	}
2383 
2384 	/* get RSS functions configured in MRQC register */
2385 	mrqc = IGC_READ_REG(hw, IGC_MRQC);
2386 	if ((mrqc & IGC_MRQC_ENABLE_RSS_4Q) == 0)
2387 		return 0;
2388 
2389 	rss_hf = 0;
2390 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4)
2391 		rss_hf |= ETH_RSS_IPV4;
2392 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4_TCP)
2393 		rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2394 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6)
2395 		rss_hf |= ETH_RSS_IPV6;
2396 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_EX)
2397 		rss_hf |= ETH_RSS_IPV6_EX;
2398 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_TCP)
2399 		rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2400 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_TCP_EX)
2401 		rss_hf |= ETH_RSS_IPV6_TCP_EX;
2402 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4_UDP)
2403 		rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2404 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_UDP)
2405 		rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2406 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_UDP_EX)
2407 		rss_hf |= ETH_RSS_IPV6_UDP_EX;
2408 
2409 	rss_conf->rss_hf |= rss_hf;
2410 	return 0;
2411 }
2412 
2413 static int
2414 eth_igc_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2415 {
2416 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2417 	struct igc_vfta *shadow_vfta = IGC_DEV_PRIVATE_VFTA(dev);
2418 	uint32_t vfta;
2419 	uint32_t vid_idx;
2420 	uint32_t vid_bit;
2421 
2422 	vid_idx = (vlan_id >> IGC_VFTA_ENTRY_SHIFT) & IGC_VFTA_ENTRY_MASK;
2423 	vid_bit = 1u << (vlan_id & IGC_VFTA_ENTRY_BIT_SHIFT_MASK);
2424 	vfta = shadow_vfta->vfta[vid_idx];
2425 	if (on)
2426 		vfta |= vid_bit;
2427 	else
2428 		vfta &= ~vid_bit;
2429 	IGC_WRITE_REG_ARRAY(hw, IGC_VFTA, vid_idx, vfta);
2430 
2431 	/* update local VFTA copy */
2432 	shadow_vfta->vfta[vid_idx] = vfta;
2433 
2434 	return 0;
2435 }
2436 
2437 static void
2438 igc_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2439 {
2440 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2441 	igc_read_reg_check_clear_bits(hw, IGC_RCTL,
2442 			IGC_RCTL_CFIEN | IGC_RCTL_VFE);
2443 }
2444 
2445 static void
2446 igc_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2447 {
2448 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2449 	struct igc_vfta *shadow_vfta = IGC_DEV_PRIVATE_VFTA(dev);
2450 	uint32_t reg_val;
2451 	int i;
2452 
2453 	/* Filter Table Enable, CFI not used for packet acceptance */
2454 	reg_val = IGC_READ_REG(hw, IGC_RCTL);
2455 	reg_val &= ~IGC_RCTL_CFIEN;
2456 	reg_val |= IGC_RCTL_VFE;
2457 	IGC_WRITE_REG(hw, IGC_RCTL, reg_val);
2458 
2459 	/* restore VFTA table */
2460 	for (i = 0; i < IGC_VFTA_SIZE; i++)
2461 		IGC_WRITE_REG_ARRAY(hw, IGC_VFTA, i, shadow_vfta->vfta[i]);
2462 }
2463 
2464 static void
2465 igc_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2466 {
2467 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2468 
2469 	igc_read_reg_check_clear_bits(hw, IGC_CTRL, IGC_CTRL_VME);
2470 }
2471 
2472 static void
2473 igc_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2474 {
2475 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2476 
2477 	igc_read_reg_check_set_bits(hw, IGC_CTRL, IGC_CTRL_VME);
2478 }
2479 
2480 static int
2481 igc_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2482 {
2483 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2484 	uint32_t ctrl_ext;
2485 
2486 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
2487 
2488 	/* if extend vlan hasn't been enabled */
2489 	if ((ctrl_ext & IGC_CTRL_EXT_EXT_VLAN) == 0)
2490 		return 0;
2491 
2492 	if ((dev->data->dev_conf.rxmode.offloads &
2493 			DEV_RX_OFFLOAD_JUMBO_FRAME) == 0)
2494 		goto write_ext_vlan;
2495 
2496 	/* Update maximum packet length */
2497 	if (dev->data->dev_conf.rxmode.max_rx_pkt_len <
2498 		RTE_ETHER_MIN_MTU + VLAN_TAG_SIZE) {
2499 		PMD_DRV_LOG(ERR, "Maximum packet length %u error, min is %u",
2500 			dev->data->dev_conf.rxmode.max_rx_pkt_len,
2501 			VLAN_TAG_SIZE + RTE_ETHER_MIN_MTU);
2502 		return -EINVAL;
2503 	}
2504 	dev->data->dev_conf.rxmode.max_rx_pkt_len -= VLAN_TAG_SIZE;
2505 	IGC_WRITE_REG(hw, IGC_RLPML,
2506 		dev->data->dev_conf.rxmode.max_rx_pkt_len);
2507 
2508 write_ext_vlan:
2509 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext & ~IGC_CTRL_EXT_EXT_VLAN);
2510 	return 0;
2511 }
2512 
2513 static int
2514 igc_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2515 {
2516 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2517 	uint32_t ctrl_ext;
2518 
2519 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
2520 
2521 	/* if extend vlan has been enabled */
2522 	if (ctrl_ext & IGC_CTRL_EXT_EXT_VLAN)
2523 		return 0;
2524 
2525 	if ((dev->data->dev_conf.rxmode.offloads &
2526 			DEV_RX_OFFLOAD_JUMBO_FRAME) == 0)
2527 		goto write_ext_vlan;
2528 
2529 	/* Update maximum packet length */
2530 	if (dev->data->dev_conf.rxmode.max_rx_pkt_len >
2531 		MAX_RX_JUMBO_FRAME_SIZE - VLAN_TAG_SIZE) {
2532 		PMD_DRV_LOG(ERR, "Maximum packet length %u error, max is %u",
2533 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2534 			VLAN_TAG_SIZE, MAX_RX_JUMBO_FRAME_SIZE);
2535 		return -EINVAL;
2536 	}
2537 	dev->data->dev_conf.rxmode.max_rx_pkt_len += VLAN_TAG_SIZE;
2538 	IGC_WRITE_REG(hw, IGC_RLPML,
2539 		dev->data->dev_conf.rxmode.max_rx_pkt_len);
2540 
2541 write_ext_vlan:
2542 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext | IGC_CTRL_EXT_EXT_VLAN);
2543 	return 0;
2544 }
2545 
2546 static int
2547 eth_igc_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2548 {
2549 	struct rte_eth_rxmode *rxmode;
2550 
2551 	rxmode = &dev->data->dev_conf.rxmode;
2552 	if (mask & ETH_VLAN_STRIP_MASK) {
2553 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2554 			igc_vlan_hw_strip_enable(dev);
2555 		else
2556 			igc_vlan_hw_strip_disable(dev);
2557 	}
2558 
2559 	if (mask & ETH_VLAN_FILTER_MASK) {
2560 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2561 			igc_vlan_hw_filter_enable(dev);
2562 		else
2563 			igc_vlan_hw_filter_disable(dev);
2564 	}
2565 
2566 	if (mask & ETH_VLAN_EXTEND_MASK) {
2567 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2568 			return igc_vlan_hw_extend_enable(dev);
2569 		else
2570 			return igc_vlan_hw_extend_disable(dev);
2571 	}
2572 
2573 	return 0;
2574 }
2575 
2576 static int
2577 eth_igc_vlan_tpid_set(struct rte_eth_dev *dev,
2578 		      enum rte_vlan_type vlan_type,
2579 		      uint16_t tpid)
2580 {
2581 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2582 	uint32_t reg_val;
2583 
2584 	/* only outer TPID of double VLAN can be configured*/
2585 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2586 		reg_val = IGC_READ_REG(hw, IGC_VET);
2587 		reg_val = (reg_val & (~IGC_VET_EXT)) |
2588 			((uint32_t)tpid << IGC_VET_EXT_SHIFT);
2589 		IGC_WRITE_REG(hw, IGC_VET, reg_val);
2590 
2591 		return 0;
2592 	}
2593 
2594 	/* all other TPID values are read-only*/
2595 	PMD_DRV_LOG(ERR, "Not supported");
2596 	return -ENOTSUP;
2597 }
2598 
2599 static int
2600 eth_igc_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
2601 	struct rte_pci_device *pci_dev)
2602 {
2603 	PMD_INIT_FUNC_TRACE();
2604 	return rte_eth_dev_pci_generic_probe(pci_dev,
2605 		sizeof(struct igc_adapter), eth_igc_dev_init);
2606 }
2607 
2608 static int
2609 eth_igc_pci_remove(struct rte_pci_device *pci_dev)
2610 {
2611 	PMD_INIT_FUNC_TRACE();
2612 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igc_dev_uninit);
2613 }
2614 
2615 static struct rte_pci_driver rte_igc_pmd = {
2616 	.id_table = pci_id_igc_map,
2617 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
2618 	.probe = eth_igc_pci_probe,
2619 	.remove = eth_igc_pci_remove,
2620 };
2621 
2622 RTE_PMD_REGISTER_PCI(net_igc, rte_igc_pmd);
2623 RTE_PMD_REGISTER_PCI_TABLE(net_igc, pci_id_igc_map);
2624 RTE_PMD_REGISTER_KMOD_DEP(net_igc, "* igb_uio | uio_pci_generic | vfio-pci");
2625