xref: /dpdk/drivers/net/igc/igc_ethdev.c (revision dd4e429c)
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 != ETH_MQ_RX_NONE &&
311 		rx_mq_mode != 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 != 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 & ETH_MQ_RX_RSS_FLAG)
338 		dev->data->dev_conf.rxmode.offloads |= DEV_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 				ETH_LINK_FULL_DUPLEX :
477 				ETH_LINK_HALF_DUPLEX;
478 		link.link_speed = speed;
479 		link.link_status = ETH_LINK_UP;
480 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
481 				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 = ETH_LINK_HALF_DUPLEX;
494 		link.link_status = ETH_LINK_DOWN;
495 		link.link_autoneg = 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 == 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 	if (intr_handle->intr_vec != NULL) {
665 		rte_free(intr_handle->intr_vec);
666 		intr_handle->intr_vec = NULL;
667 	}
668 
669 	return 0;
670 }
671 
672 /*
673  * write interrupt vector allocation register
674  * @hw
675  *  board private structure
676  * @queue_index
677  *  queue index, valid 0,1,2,3
678  * @tx
679  *  tx:1, rx:0
680  * @msix_vector
681  *  msix-vector, valid 0,1,2,3,4
682  */
683 static void
684 igc_write_ivar(struct igc_hw *hw, uint8_t queue_index,
685 		bool tx, uint8_t msix_vector)
686 {
687 	uint8_t offset = 0;
688 	uint8_t reg_index = queue_index >> 1;
689 	uint32_t val;
690 
691 	/*
692 	 * IVAR(0)
693 	 * bit31...24	bit23...16	bit15...8	bit7...0
694 	 * TX1		RX1		TX0		RX0
695 	 *
696 	 * IVAR(1)
697 	 * bit31...24	bit23...16	bit15...8	bit7...0
698 	 * TX3		RX3		TX2		RX2
699 	 */
700 
701 	if (tx)
702 		offset = 8;
703 
704 	if (queue_index & 1)
705 		offset += 16;
706 
707 	val = IGC_READ_REG_ARRAY(hw, IGC_IVAR0, reg_index);
708 
709 	/* clear bits */
710 	val &= ~((uint32_t)0xFF << offset);
711 
712 	/* write vector and valid bit */
713 	val |= (uint32_t)(msix_vector | IGC_IVAR_VALID) << offset;
714 
715 	IGC_WRITE_REG_ARRAY(hw, IGC_IVAR0, reg_index, val);
716 }
717 
718 /* Sets up the hardware to generate MSI-X interrupts properly
719  * @hw
720  *  board private structure
721  */
722 static void
723 igc_configure_msix_intr(struct rte_eth_dev *dev)
724 {
725 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
726 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
727 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
728 
729 	uint32_t intr_mask;
730 	uint32_t vec = IGC_MISC_VEC_ID;
731 	uint32_t base = IGC_MISC_VEC_ID;
732 	uint32_t misc_shift = 0;
733 	int i;
734 
735 	/* won't configure msix register if no mapping is done
736 	 * between intr vector and event fd
737 	 */
738 	if (!rte_intr_dp_is_en(intr_handle))
739 		return;
740 
741 	if (rte_intr_allow_others(intr_handle)) {
742 		base = IGC_RX_VEC_START;
743 		vec = base;
744 		misc_shift = 1;
745 	}
746 
747 	/* turn on MSI-X capability first */
748 	IGC_WRITE_REG(hw, IGC_GPIE, IGC_GPIE_MSIX_MODE |
749 				IGC_GPIE_PBA | IGC_GPIE_EIAME |
750 				IGC_GPIE_NSICR);
751 	intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
752 		misc_shift;
753 
754 	if (dev->data->dev_conf.intr_conf.lsc)
755 		intr_mask |= (1u << IGC_MSIX_OTHER_INTR_VEC);
756 
757 	/* enable msix auto-clear */
758 	igc_read_reg_check_set_bits(hw, IGC_EIAC, intr_mask);
759 
760 	/* set other cause interrupt vector */
761 	igc_read_reg_check_set_bits(hw, IGC_IVAR_MISC,
762 		(uint32_t)(IGC_MSIX_OTHER_INTR_VEC | IGC_IVAR_VALID) << 8);
763 
764 	/* enable auto-mask */
765 	igc_read_reg_check_set_bits(hw, IGC_EIAM, intr_mask);
766 
767 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
768 		igc_write_ivar(hw, i, 0, vec);
769 		intr_handle->intr_vec[i] = vec;
770 		if (vec < base + intr_handle->nb_efd - 1)
771 			vec++;
772 	}
773 
774 	IGC_WRITE_FLUSH(hw);
775 }
776 
777 /**
778  * It enables the interrupt mask and then enable the interrupt.
779  *
780  * @dev
781  *  Pointer to struct rte_eth_dev.
782  * @on
783  *  Enable or Disable
784  */
785 static void
786 igc_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
787 {
788 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
789 
790 	if (on)
791 		intr->mask |= IGC_ICR_LSC;
792 	else
793 		intr->mask &= ~IGC_ICR_LSC;
794 }
795 
796 /*
797  * It enables the interrupt.
798  * It will be called once only during nic initialized.
799  */
800 static void
801 igc_rxq_interrupt_setup(struct rte_eth_dev *dev)
802 {
803 	uint32_t mask;
804 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
805 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
806 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
807 	int misc_shift = rte_intr_allow_others(intr_handle) ? 1 : 0;
808 
809 	/* won't configure msix register if no mapping is done
810 	 * between intr vector and event fd
811 	 */
812 	if (!rte_intr_dp_is_en(intr_handle))
813 		return;
814 
815 	mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) << misc_shift;
816 	IGC_WRITE_REG(hw, IGC_EIMS, mask);
817 }
818 
819 /*
820  *  Get hardware rx-buffer size.
821  */
822 static inline int
823 igc_get_rx_buffer_size(struct igc_hw *hw)
824 {
825 	return (IGC_READ_REG(hw, IGC_RXPBS) & 0x3f) << 10;
826 }
827 
828 /*
829  * igc_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
830  * For ASF and Pass Through versions of f/w this means
831  * that the driver is loaded.
832  */
833 static void
834 igc_hw_control_acquire(struct igc_hw *hw)
835 {
836 	uint32_t ctrl_ext;
837 
838 	/* Let firmware know the driver has taken over */
839 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
840 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext | IGC_CTRL_EXT_DRV_LOAD);
841 }
842 
843 /*
844  * igc_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
845  * For ASF and Pass Through versions of f/w this means that the
846  * driver is no longer loaded.
847  */
848 static void
849 igc_hw_control_release(struct igc_hw *hw)
850 {
851 	uint32_t ctrl_ext;
852 
853 	/* Let firmware taken over control of h/w */
854 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
855 	IGC_WRITE_REG(hw, IGC_CTRL_EXT,
856 			ctrl_ext & ~IGC_CTRL_EXT_DRV_LOAD);
857 }
858 
859 static int
860 igc_hardware_init(struct igc_hw *hw)
861 {
862 	uint32_t rx_buf_size;
863 	int diag;
864 
865 	/* Let the firmware know the OS is in control */
866 	igc_hw_control_acquire(hw);
867 
868 	/* Issue a global reset */
869 	igc_reset_hw(hw);
870 
871 	/* disable all wake up */
872 	IGC_WRITE_REG(hw, IGC_WUC, 0);
873 
874 	/*
875 	 * Hardware flow control
876 	 * - High water mark should allow for at least two standard size (1518)
877 	 *   frames to be received after sending an XOFF.
878 	 * - Low water mark works best when it is very near the high water mark.
879 	 *   This allows the receiver to restart by sending XON when it has
880 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
881 	 *   restart after one full frame is pulled from the buffer. There
882 	 *   could be several smaller frames in the buffer and if so they will
883 	 *   not trigger the XON until their total number reduces the buffer
884 	 *   by 1500.
885 	 */
886 	rx_buf_size = igc_get_rx_buffer_size(hw);
887 	hw->fc.high_water = rx_buf_size - (RTE_ETHER_MAX_LEN * 2);
888 	hw->fc.low_water = hw->fc.high_water - 1500;
889 	hw->fc.pause_time = IGC_FC_PAUSE_TIME;
890 	hw->fc.send_xon = 1;
891 	hw->fc.requested_mode = igc_fc_full;
892 
893 	diag = igc_init_hw(hw);
894 	if (diag < 0)
895 		return diag;
896 
897 	igc_get_phy_info(hw);
898 	igc_check_for_link(hw);
899 
900 	return 0;
901 }
902 
903 static int
904 eth_igc_start(struct rte_eth_dev *dev)
905 {
906 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
907 	struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
908 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
909 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
910 	uint32_t *speeds;
911 	int ret;
912 
913 	PMD_INIT_FUNC_TRACE();
914 
915 	/* disable all MSI-X interrupts */
916 	IGC_WRITE_REG(hw, IGC_EIMC, 0x1f);
917 	IGC_WRITE_FLUSH(hw);
918 
919 	/* clear all MSI-X interrupts */
920 	IGC_WRITE_REG(hw, IGC_EICR, 0x1f);
921 
922 	/* disable uio/vfio intr/eventfd mapping */
923 	if (!adapter->stopped)
924 		rte_intr_disable(intr_handle);
925 
926 	/* Power up the phy. Needed to make the link go Up */
927 	eth_igc_set_link_up(dev);
928 
929 	/* Put the address into the Receive Address Array */
930 	igc_rar_set(hw, hw->mac.addr, 0);
931 
932 	/* Initialize the hardware */
933 	if (igc_hardware_init(hw)) {
934 		PMD_DRV_LOG(ERR, "Unable to initialize the hardware");
935 		return -EIO;
936 	}
937 	adapter->stopped = 0;
938 
939 	/* check and configure queue intr-vector mapping */
940 	if (rte_intr_cap_multiple(intr_handle) &&
941 		dev->data->dev_conf.intr_conf.rxq) {
942 		uint32_t intr_vector = dev->data->nb_rx_queues;
943 		if (rte_intr_efd_enable(intr_handle, intr_vector))
944 			return -1;
945 	}
946 
947 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
948 		intr_handle->intr_vec = rte_zmalloc("intr_vec",
949 			dev->data->nb_rx_queues * sizeof(int), 0);
950 		if (intr_handle->intr_vec == NULL) {
951 			PMD_DRV_LOG(ERR,
952 				"Failed to allocate %d rx_queues intr_vec",
953 				dev->data->nb_rx_queues);
954 			return -ENOMEM;
955 		}
956 	}
957 
958 	/* configure msix for rx interrupt */
959 	igc_configure_msix_intr(dev);
960 
961 	igc_tx_init(dev);
962 
963 	/* This can fail when allocating mbufs for descriptor rings */
964 	ret = igc_rx_init(dev);
965 	if (ret) {
966 		PMD_DRV_LOG(ERR, "Unable to initialize RX hardware");
967 		igc_dev_clear_queues(dev);
968 		return ret;
969 	}
970 
971 	igc_clear_hw_cntrs_base_generic(hw);
972 
973 	/* VLAN Offload Settings */
974 	eth_igc_vlan_offload_set(dev,
975 		ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
976 		ETH_VLAN_EXTEND_MASK);
977 
978 	/* Setup link speed and duplex */
979 	speeds = &dev->data->dev_conf.link_speeds;
980 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
981 		hw->phy.autoneg_advertised = IGC_ALL_SPEED_DUPLEX_2500;
982 		hw->mac.autoneg = 1;
983 	} else {
984 		int num_speeds = 0;
985 
986 		if (*speeds & ETH_LINK_SPEED_FIXED) {
987 			PMD_DRV_LOG(ERR,
988 				    "Force speed mode currently not supported");
989 			igc_dev_clear_queues(dev);
990 			return -EINVAL;
991 		}
992 
993 		hw->phy.autoneg_advertised = 0;
994 		hw->mac.autoneg = 1;
995 
996 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
997 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
998 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_2_5G)) {
999 			num_speeds = -1;
1000 			goto error_invalid_config;
1001 		}
1002 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
1003 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1004 			num_speeds++;
1005 		}
1006 		if (*speeds & ETH_LINK_SPEED_10M) {
1007 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1008 			num_speeds++;
1009 		}
1010 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
1011 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1012 			num_speeds++;
1013 		}
1014 		if (*speeds & ETH_LINK_SPEED_100M) {
1015 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1016 			num_speeds++;
1017 		}
1018 		if (*speeds & ETH_LINK_SPEED_1G) {
1019 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1020 			num_speeds++;
1021 		}
1022 		if (*speeds & ETH_LINK_SPEED_2_5G) {
1023 			hw->phy.autoneg_advertised |= ADVERTISE_2500_FULL;
1024 			num_speeds++;
1025 		}
1026 		if (num_speeds == 0)
1027 			goto error_invalid_config;
1028 	}
1029 
1030 	igc_setup_link(hw);
1031 
1032 	if (rte_intr_allow_others(intr_handle)) {
1033 		/* check if lsc interrupt is enabled */
1034 		if (dev->data->dev_conf.intr_conf.lsc)
1035 			igc_lsc_interrupt_setup(dev, 1);
1036 		else
1037 			igc_lsc_interrupt_setup(dev, 0);
1038 	} else {
1039 		rte_intr_callback_unregister(intr_handle,
1040 					     eth_igc_interrupt_handler,
1041 					     (void *)dev);
1042 		if (dev->data->dev_conf.intr_conf.lsc)
1043 			PMD_DRV_LOG(INFO,
1044 				"LSC won't enable because of no intr multiplex");
1045 	}
1046 
1047 	/* enable uio/vfio intr/eventfd mapping */
1048 	rte_intr_enable(intr_handle);
1049 
1050 	rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1051 			igc_update_queue_stats_handler, dev);
1052 
1053 	/* check if rxq interrupt is enabled */
1054 	if (dev->data->dev_conf.intr_conf.rxq &&
1055 			rte_intr_dp_is_en(intr_handle))
1056 		igc_rxq_interrupt_setup(dev);
1057 
1058 	/* resume enabled intr since hw reset */
1059 	igc_intr_other_enable(dev);
1060 
1061 	eth_igc_rxtx_control(dev, true);
1062 	eth_igc_link_update(dev, 0);
1063 
1064 	/* configure MAC-loopback mode */
1065 	if (dev->data->dev_conf.lpbk_mode == 1) {
1066 		uint32_t reg_val;
1067 
1068 		reg_val = IGC_READ_REG(hw, IGC_CTRL);
1069 		reg_val &= ~IGC_CTRL_SPEED_MASK;
1070 		reg_val |= IGC_CTRL_SLU | IGC_CTRL_FRCSPD |
1071 			IGC_CTRL_FRCDPX | IGC_CTRL_FD | IGC_CTRL_SPEED_2500;
1072 		IGC_WRITE_REG(hw, IGC_CTRL, reg_val);
1073 
1074 		igc_read_reg_check_set_bits(hw, IGC_EEER, IGC_EEER_EEE_FRC_AN);
1075 	}
1076 
1077 	return 0;
1078 
1079 error_invalid_config:
1080 	PMD_DRV_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
1081 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
1082 	igc_dev_clear_queues(dev);
1083 	return -EINVAL;
1084 }
1085 
1086 static int
1087 igc_reset_swfw_lock(struct igc_hw *hw)
1088 {
1089 	int ret_val;
1090 
1091 	/*
1092 	 * Do mac ops initialization manually here, since we will need
1093 	 * some function pointers set by this call.
1094 	 */
1095 	ret_val = igc_init_mac_params(hw);
1096 	if (ret_val)
1097 		return ret_val;
1098 
1099 	/*
1100 	 * SMBI lock should not fail in this early stage. If this is the case,
1101 	 * it is due to an improper exit of the application.
1102 	 * So force the release of the faulty lock.
1103 	 */
1104 	if (igc_get_hw_semaphore_generic(hw) < 0)
1105 		PMD_DRV_LOG(DEBUG, "SMBI lock released");
1106 
1107 	igc_put_hw_semaphore_generic(hw);
1108 
1109 	if (hw->mac.ops.acquire_swfw_sync != NULL) {
1110 		uint16_t mask;
1111 
1112 		/*
1113 		 * Phy lock should not fail in this early stage.
1114 		 * If this is the case, it is due to an improper exit of the
1115 		 * application. So force the release of the faulty lock.
1116 		 */
1117 		mask = IGC_SWFW_PHY0_SM;
1118 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
1119 			PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released",
1120 				    hw->bus.func);
1121 		}
1122 		hw->mac.ops.release_swfw_sync(hw, mask);
1123 
1124 		/*
1125 		 * This one is more tricky since it is common to all ports; but
1126 		 * swfw_sync retries last long enough (1s) to be almost sure
1127 		 * that if lock can not be taken it is due to an improper lock
1128 		 * of the semaphore.
1129 		 */
1130 		mask = IGC_SWFW_EEP_SM;
1131 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0)
1132 			PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1133 
1134 		hw->mac.ops.release_swfw_sync(hw, mask);
1135 	}
1136 
1137 	return IGC_SUCCESS;
1138 }
1139 
1140 /*
1141  * free all rx/tx queues.
1142  */
1143 static void
1144 igc_dev_free_queues(struct rte_eth_dev *dev)
1145 {
1146 	uint16_t i;
1147 
1148 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
1149 		eth_igc_rx_queue_release(dev, i);
1150 		dev->data->rx_queues[i] = NULL;
1151 	}
1152 	dev->data->nb_rx_queues = 0;
1153 
1154 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
1155 		eth_igc_tx_queue_release(dev, i);
1156 		dev->data->tx_queues[i] = NULL;
1157 	}
1158 	dev->data->nb_tx_queues = 0;
1159 }
1160 
1161 static int
1162 eth_igc_close(struct rte_eth_dev *dev)
1163 {
1164 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1165 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1166 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1167 	struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
1168 	int retry = 0;
1169 	int ret = 0;
1170 
1171 	PMD_INIT_FUNC_TRACE();
1172 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1173 		return 0;
1174 
1175 	if (!adapter->stopped)
1176 		ret = eth_igc_stop(dev);
1177 
1178 	igc_flow_flush(dev, NULL);
1179 	igc_clear_all_filter(dev);
1180 
1181 	igc_intr_other_disable(dev);
1182 	do {
1183 		int ret = rte_intr_callback_unregister(intr_handle,
1184 				eth_igc_interrupt_handler, dev);
1185 		if (ret >= 0 || ret == -ENOENT || ret == -EINVAL)
1186 			break;
1187 
1188 		PMD_DRV_LOG(ERR, "intr callback unregister failed: %d", ret);
1189 		DELAY(200 * 1000); /* delay 200ms */
1190 	} while (retry++ < 5);
1191 
1192 	igc_phy_hw_reset(hw);
1193 	igc_hw_control_release(hw);
1194 	igc_dev_free_queues(dev);
1195 
1196 	/* Reset any pending lock */
1197 	igc_reset_swfw_lock(hw);
1198 
1199 	return ret;
1200 }
1201 
1202 static void
1203 igc_identify_hardware(struct rte_eth_dev *dev, struct rte_pci_device *pci_dev)
1204 {
1205 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1206 
1207 	hw->vendor_id = pci_dev->id.vendor_id;
1208 	hw->device_id = pci_dev->id.device_id;
1209 	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
1210 	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
1211 }
1212 
1213 static int
1214 eth_igc_dev_init(struct rte_eth_dev *dev)
1215 {
1216 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1217 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
1218 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1219 	int i, error = 0;
1220 
1221 	PMD_INIT_FUNC_TRACE();
1222 	dev->dev_ops = &eth_igc_ops;
1223 	dev->rx_queue_count = eth_igc_rx_queue_count;
1224 	dev->rx_descriptor_status = eth_igc_rx_descriptor_status;
1225 	dev->tx_descriptor_status = eth_igc_tx_descriptor_status;
1226 
1227 	/*
1228 	 * for secondary processes, we don't initialize any further as primary
1229 	 * has already done this work. Only check we don't need a different
1230 	 * RX function.
1231 	 */
1232 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1233 		return 0;
1234 
1235 	rte_eth_copy_pci_info(dev, pci_dev);
1236 	dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1237 
1238 	hw->back = pci_dev;
1239 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1240 
1241 	igc_identify_hardware(dev, pci_dev);
1242 	if (igc_setup_init_funcs(hw, false) != IGC_SUCCESS) {
1243 		error = -EIO;
1244 		goto err_late;
1245 	}
1246 
1247 	igc_get_bus_info(hw);
1248 
1249 	/* Reset any pending lock */
1250 	if (igc_reset_swfw_lock(hw) != IGC_SUCCESS) {
1251 		error = -EIO;
1252 		goto err_late;
1253 	}
1254 
1255 	/* Finish initialization */
1256 	if (igc_setup_init_funcs(hw, true) != IGC_SUCCESS) {
1257 		error = -EIO;
1258 		goto err_late;
1259 	}
1260 
1261 	hw->mac.autoneg = 1;
1262 	hw->phy.autoneg_wait_to_complete = 0;
1263 	hw->phy.autoneg_advertised = IGC_ALL_SPEED_DUPLEX_2500;
1264 
1265 	/* Copper options */
1266 	if (hw->phy.media_type == igc_media_type_copper) {
1267 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
1268 		hw->phy.disable_polarity_correction = 0;
1269 		hw->phy.ms_type = igc_ms_hw_default;
1270 	}
1271 
1272 	/*
1273 	 * Start from a known state, this is important in reading the nvm
1274 	 * and mac from that.
1275 	 */
1276 	igc_reset_hw(hw);
1277 
1278 	/* Make sure we have a good EEPROM before we read from it */
1279 	if (igc_validate_nvm_checksum(hw) < 0) {
1280 		/*
1281 		 * Some PCI-E parts fail the first check due to
1282 		 * the link being in sleep state, call it again,
1283 		 * if it fails a second time its a real issue.
1284 		 */
1285 		if (igc_validate_nvm_checksum(hw) < 0) {
1286 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
1287 			error = -EIO;
1288 			goto err_late;
1289 		}
1290 	}
1291 
1292 	/* Read the permanent MAC address out of the EEPROM */
1293 	if (igc_read_mac_addr(hw) != 0) {
1294 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
1295 		error = -EIO;
1296 		goto err_late;
1297 	}
1298 
1299 	/* Allocate memory for storing MAC addresses */
1300 	dev->data->mac_addrs = rte_zmalloc("igc",
1301 		RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count, 0);
1302 	if (dev->data->mac_addrs == NULL) {
1303 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes for storing MAC",
1304 				RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
1305 		error = -ENOMEM;
1306 		goto err_late;
1307 	}
1308 
1309 	/* Copy the permanent MAC address */
1310 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
1311 			&dev->data->mac_addrs[0]);
1312 
1313 	/* Now initialize the hardware */
1314 	if (igc_hardware_init(hw) != 0) {
1315 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
1316 		rte_free(dev->data->mac_addrs);
1317 		dev->data->mac_addrs = NULL;
1318 		error = -ENODEV;
1319 		goto err_late;
1320 	}
1321 
1322 	hw->mac.get_link_status = 1;
1323 	igc->stopped = 0;
1324 
1325 	/* Indicate SOL/IDER usage */
1326 	if (igc_check_reset_block(hw) < 0)
1327 		PMD_INIT_LOG(ERR,
1328 			"PHY reset is blocked due to SOL/IDER session.");
1329 
1330 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
1331 			dev->data->port_id, pci_dev->id.vendor_id,
1332 			pci_dev->id.device_id);
1333 
1334 	rte_intr_callback_register(&pci_dev->intr_handle,
1335 			eth_igc_interrupt_handler, (void *)dev);
1336 
1337 	/* enable uio/vfio intr/eventfd mapping */
1338 	rte_intr_enable(&pci_dev->intr_handle);
1339 
1340 	/* enable support intr */
1341 	igc_intr_other_enable(dev);
1342 
1343 	/* initiate queue status */
1344 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1345 		igc->txq_stats_map[i] = -1;
1346 		igc->rxq_stats_map[i] = -1;
1347 	}
1348 
1349 	igc_flow_init(dev);
1350 	igc_clear_all_filter(dev);
1351 	return 0;
1352 
1353 err_late:
1354 	igc_hw_control_release(hw);
1355 	return error;
1356 }
1357 
1358 static int
1359 eth_igc_dev_uninit(__rte_unused struct rte_eth_dev *eth_dev)
1360 {
1361 	PMD_INIT_FUNC_TRACE();
1362 	eth_igc_close(eth_dev);
1363 	return 0;
1364 }
1365 
1366 static int
1367 eth_igc_reset(struct rte_eth_dev *dev)
1368 {
1369 	int ret;
1370 
1371 	PMD_INIT_FUNC_TRACE();
1372 
1373 	ret = eth_igc_dev_uninit(dev);
1374 	if (ret)
1375 		return ret;
1376 
1377 	return eth_igc_dev_init(dev);
1378 }
1379 
1380 static int
1381 eth_igc_promiscuous_enable(struct rte_eth_dev *dev)
1382 {
1383 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1384 	uint32_t rctl;
1385 
1386 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1387 	rctl |= (IGC_RCTL_UPE | IGC_RCTL_MPE);
1388 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1389 	return 0;
1390 }
1391 
1392 static int
1393 eth_igc_promiscuous_disable(struct rte_eth_dev *dev)
1394 {
1395 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1396 	uint32_t rctl;
1397 
1398 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1399 	rctl &= (~IGC_RCTL_UPE);
1400 	if (dev->data->all_multicast == 1)
1401 		rctl |= IGC_RCTL_MPE;
1402 	else
1403 		rctl &= (~IGC_RCTL_MPE);
1404 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1405 	return 0;
1406 }
1407 
1408 static int
1409 eth_igc_allmulticast_enable(struct rte_eth_dev *dev)
1410 {
1411 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1412 	uint32_t rctl;
1413 
1414 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1415 	rctl |= IGC_RCTL_MPE;
1416 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1417 	return 0;
1418 }
1419 
1420 static int
1421 eth_igc_allmulticast_disable(struct rte_eth_dev *dev)
1422 {
1423 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1424 	uint32_t rctl;
1425 
1426 	if (dev->data->promiscuous == 1)
1427 		return 0;	/* must remain in all_multicast mode */
1428 
1429 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1430 	rctl &= (~IGC_RCTL_MPE);
1431 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1432 	return 0;
1433 }
1434 
1435 static int
1436 eth_igc_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
1437 		       size_t fw_size)
1438 {
1439 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1440 	struct igc_fw_version fw;
1441 	int ret;
1442 
1443 	igc_get_fw_version(hw, &fw);
1444 
1445 	/* if option rom is valid, display its version too */
1446 	if (fw.or_valid) {
1447 		ret = snprintf(fw_version, fw_size,
1448 			 "%d.%d, 0x%08x, %d.%d.%d",
1449 			 fw.eep_major, fw.eep_minor, fw.etrack_id,
1450 			 fw.or_major, fw.or_build, fw.or_patch);
1451 	/* no option rom */
1452 	} else {
1453 		if (fw.etrack_id != 0X0000) {
1454 			ret = snprintf(fw_version, fw_size,
1455 				 "%d.%d, 0x%08x",
1456 				 fw.eep_major, fw.eep_minor,
1457 				 fw.etrack_id);
1458 		} else {
1459 			ret = snprintf(fw_version, fw_size,
1460 				 "%d.%d.%d",
1461 				 fw.eep_major, fw.eep_minor,
1462 				 fw.eep_build);
1463 		}
1464 	}
1465 	if (ret < 0)
1466 		return -EINVAL;
1467 
1468 	ret += 1; /* add the size of '\0' */
1469 	if (fw_size < (size_t)ret)
1470 		return ret;
1471 	else
1472 		return 0;
1473 }
1474 
1475 static int
1476 eth_igc_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1477 {
1478 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1479 
1480 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1481 	dev_info->max_rx_pktlen = MAX_RX_JUMBO_FRAME_SIZE;
1482 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1483 	dev_info->rx_offload_capa = IGC_RX_OFFLOAD_ALL;
1484 	dev_info->tx_offload_capa = IGC_TX_OFFLOAD_ALL;
1485 	dev_info->rx_queue_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP;
1486 
1487 	dev_info->max_rx_queues = IGC_QUEUE_PAIRS_NUM;
1488 	dev_info->max_tx_queues = IGC_QUEUE_PAIRS_NUM;
1489 	dev_info->max_vmdq_pools = 0;
1490 
1491 	dev_info->hash_key_size = IGC_HKEY_MAX_INDEX * sizeof(uint32_t);
1492 	dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
1493 	dev_info->flow_type_rss_offloads = IGC_RSS_OFFLOAD_ALL;
1494 
1495 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
1496 		.rx_thresh = {
1497 			.pthresh = IGC_DEFAULT_RX_PTHRESH,
1498 			.hthresh = IGC_DEFAULT_RX_HTHRESH,
1499 			.wthresh = IGC_DEFAULT_RX_WTHRESH,
1500 		},
1501 		.rx_free_thresh = IGC_DEFAULT_RX_FREE_THRESH,
1502 		.rx_drop_en = 0,
1503 		.offloads = 0,
1504 	};
1505 
1506 	dev_info->default_txconf = (struct rte_eth_txconf) {
1507 		.tx_thresh = {
1508 			.pthresh = IGC_DEFAULT_TX_PTHRESH,
1509 			.hthresh = IGC_DEFAULT_TX_HTHRESH,
1510 			.wthresh = IGC_DEFAULT_TX_WTHRESH,
1511 		},
1512 		.offloads = 0,
1513 	};
1514 
1515 	dev_info->rx_desc_lim = rx_desc_lim;
1516 	dev_info->tx_desc_lim = tx_desc_lim;
1517 
1518 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1519 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1520 			ETH_LINK_SPEED_1G | ETH_LINK_SPEED_2_5G;
1521 
1522 	dev_info->max_mtu = dev_info->max_rx_pktlen - IGC_ETH_OVERHEAD;
1523 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
1524 	return 0;
1525 }
1526 
1527 static int
1528 eth_igc_led_on(struct rte_eth_dev *dev)
1529 {
1530 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1531 
1532 	return igc_led_on(hw) == IGC_SUCCESS ? 0 : -ENOTSUP;
1533 }
1534 
1535 static int
1536 eth_igc_led_off(struct rte_eth_dev *dev)
1537 {
1538 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1539 
1540 	return igc_led_off(hw) == IGC_SUCCESS ? 0 : -ENOTSUP;
1541 }
1542 
1543 static const uint32_t *
1544 eth_igc_supported_ptypes_get(__rte_unused struct rte_eth_dev *dev)
1545 {
1546 	static const uint32_t ptypes[] = {
1547 		/* refers to rx_desc_pkt_info_to_pkt_type() */
1548 		RTE_PTYPE_L2_ETHER,
1549 		RTE_PTYPE_L3_IPV4,
1550 		RTE_PTYPE_L3_IPV4_EXT,
1551 		RTE_PTYPE_L3_IPV6,
1552 		RTE_PTYPE_L3_IPV6_EXT,
1553 		RTE_PTYPE_L4_TCP,
1554 		RTE_PTYPE_L4_UDP,
1555 		RTE_PTYPE_L4_SCTP,
1556 		RTE_PTYPE_TUNNEL_IP,
1557 		RTE_PTYPE_INNER_L3_IPV6,
1558 		RTE_PTYPE_INNER_L3_IPV6_EXT,
1559 		RTE_PTYPE_INNER_L4_TCP,
1560 		RTE_PTYPE_INNER_L4_UDP,
1561 		RTE_PTYPE_UNKNOWN
1562 	};
1563 
1564 	return ptypes;
1565 }
1566 
1567 static int
1568 eth_igc_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1569 {
1570 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1571 	uint32_t frame_size = mtu + IGC_ETH_OVERHEAD;
1572 	uint32_t rctl;
1573 
1574 	/* if extend vlan has been enabled */
1575 	if (IGC_READ_REG(hw, IGC_CTRL_EXT) & IGC_CTRL_EXT_EXT_VLAN)
1576 		frame_size += VLAN_TAG_SIZE;
1577 
1578 	/* check that mtu is within the allowed range */
1579 	if (mtu < RTE_ETHER_MIN_MTU ||
1580 		frame_size > MAX_RX_JUMBO_FRAME_SIZE)
1581 		return -EINVAL;
1582 
1583 	/*
1584 	 * If device is started, refuse mtu that requires the support of
1585 	 * scattered packets when this feature has not been enabled before.
1586 	 */
1587 	if (dev->data->dev_started && !dev->data->scattered_rx &&
1588 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
1589 		PMD_INIT_LOG(ERR, "Stop port first.");
1590 		return -EINVAL;
1591 	}
1592 
1593 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1594 	if (mtu > RTE_ETHER_MTU)
1595 		rctl |= IGC_RCTL_LPE;
1596 	else
1597 		rctl &= ~IGC_RCTL_LPE;
1598 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1599 
1600 	IGC_WRITE_REG(hw, IGC_RLPML, frame_size);
1601 
1602 	return 0;
1603 }
1604 
1605 static int
1606 eth_igc_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
1607 		uint32_t index, uint32_t pool)
1608 {
1609 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1610 
1611 	igc_rar_set(hw, mac_addr->addr_bytes, index);
1612 	RTE_SET_USED(pool);
1613 	return 0;
1614 }
1615 
1616 static void
1617 eth_igc_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1618 {
1619 	uint8_t addr[RTE_ETHER_ADDR_LEN];
1620 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1621 
1622 	memset(addr, 0, sizeof(addr));
1623 	igc_rar_set(hw, addr, index);
1624 }
1625 
1626 static int
1627 eth_igc_default_mac_addr_set(struct rte_eth_dev *dev,
1628 			struct rte_ether_addr *addr)
1629 {
1630 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1631 	igc_rar_set(hw, addr->addr_bytes, 0);
1632 	return 0;
1633 }
1634 
1635 static int
1636 eth_igc_set_mc_addr_list(struct rte_eth_dev *dev,
1637 			 struct rte_ether_addr *mc_addr_set,
1638 			 uint32_t nb_mc_addr)
1639 {
1640 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1641 	igc_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1642 	return 0;
1643 }
1644 
1645 /*
1646  * Read hardware registers
1647  */
1648 static void
1649 igc_read_stats_registers(struct igc_hw *hw, struct igc_hw_stats *stats)
1650 {
1651 	int pause_frames;
1652 
1653 	uint64_t old_gprc  = stats->gprc;
1654 	uint64_t old_gptc  = stats->gptc;
1655 	uint64_t old_tpr   = stats->tpr;
1656 	uint64_t old_tpt   = stats->tpt;
1657 	uint64_t old_rpthc = stats->rpthc;
1658 	uint64_t old_hgptc = stats->hgptc;
1659 
1660 	stats->crcerrs += IGC_READ_REG(hw, IGC_CRCERRS);
1661 	stats->algnerrc += IGC_READ_REG(hw, IGC_ALGNERRC);
1662 	stats->rxerrc += IGC_READ_REG(hw, IGC_RXERRC);
1663 	stats->mpc += IGC_READ_REG(hw, IGC_MPC);
1664 	stats->scc += IGC_READ_REG(hw, IGC_SCC);
1665 	stats->ecol += IGC_READ_REG(hw, IGC_ECOL);
1666 
1667 	stats->mcc += IGC_READ_REG(hw, IGC_MCC);
1668 	stats->latecol += IGC_READ_REG(hw, IGC_LATECOL);
1669 	stats->colc += IGC_READ_REG(hw, IGC_COLC);
1670 
1671 	stats->dc += IGC_READ_REG(hw, IGC_DC);
1672 	stats->tncrs += IGC_READ_REG(hw, IGC_TNCRS);
1673 	stats->htdpmc += IGC_READ_REG(hw, IGC_HTDPMC);
1674 	stats->rlec += IGC_READ_REG(hw, IGC_RLEC);
1675 	stats->xonrxc += IGC_READ_REG(hw, IGC_XONRXC);
1676 	stats->xontxc += IGC_READ_REG(hw, IGC_XONTXC);
1677 
1678 	/*
1679 	 * For watchdog management we need to know if we have been
1680 	 * paused during the last interval, so capture that here.
1681 	 */
1682 	pause_frames = IGC_READ_REG(hw, IGC_XOFFRXC);
1683 	stats->xoffrxc += pause_frames;
1684 	stats->xofftxc += IGC_READ_REG(hw, IGC_XOFFTXC);
1685 	stats->fcruc += IGC_READ_REG(hw, IGC_FCRUC);
1686 	stats->prc64 += IGC_READ_REG(hw, IGC_PRC64);
1687 	stats->prc127 += IGC_READ_REG(hw, IGC_PRC127);
1688 	stats->prc255 += IGC_READ_REG(hw, IGC_PRC255);
1689 	stats->prc511 += IGC_READ_REG(hw, IGC_PRC511);
1690 	stats->prc1023 += IGC_READ_REG(hw, IGC_PRC1023);
1691 	stats->prc1522 += IGC_READ_REG(hw, IGC_PRC1522);
1692 	stats->gprc += IGC_READ_REG(hw, IGC_GPRC);
1693 	stats->bprc += IGC_READ_REG(hw, IGC_BPRC);
1694 	stats->mprc += IGC_READ_REG(hw, IGC_MPRC);
1695 	stats->gptc += IGC_READ_REG(hw, IGC_GPTC);
1696 
1697 	/* For the 64-bit byte counters the low dword must be read first. */
1698 	/* Both registers clear on the read of the high dword */
1699 
1700 	/* Workaround CRC bytes included in size, take away 4 bytes/packet */
1701 	stats->gorc += IGC_READ_REG(hw, IGC_GORCL);
1702 	stats->gorc += ((uint64_t)IGC_READ_REG(hw, IGC_GORCH) << 32);
1703 	stats->gorc -= (stats->gprc - old_gprc) * RTE_ETHER_CRC_LEN;
1704 	stats->gotc += IGC_READ_REG(hw, IGC_GOTCL);
1705 	stats->gotc += ((uint64_t)IGC_READ_REG(hw, IGC_GOTCH) << 32);
1706 	stats->gotc -= (stats->gptc - old_gptc) * RTE_ETHER_CRC_LEN;
1707 
1708 	stats->rnbc += IGC_READ_REG(hw, IGC_RNBC);
1709 	stats->ruc += IGC_READ_REG(hw, IGC_RUC);
1710 	stats->rfc += IGC_READ_REG(hw, IGC_RFC);
1711 	stats->roc += IGC_READ_REG(hw, IGC_ROC);
1712 	stats->rjc += IGC_READ_REG(hw, IGC_RJC);
1713 
1714 	stats->mgprc += IGC_READ_REG(hw, IGC_MGTPRC);
1715 	stats->mgpdc += IGC_READ_REG(hw, IGC_MGTPDC);
1716 	stats->mgptc += IGC_READ_REG(hw, IGC_MGTPTC);
1717 	stats->b2ospc += IGC_READ_REG(hw, IGC_B2OSPC);
1718 	stats->b2ogprc += IGC_READ_REG(hw, IGC_B2OGPRC);
1719 	stats->o2bgptc += IGC_READ_REG(hw, IGC_O2BGPTC);
1720 	stats->o2bspc += IGC_READ_REG(hw, IGC_O2BSPC);
1721 
1722 	stats->tpr += IGC_READ_REG(hw, IGC_TPR);
1723 	stats->tpt += IGC_READ_REG(hw, IGC_TPT);
1724 
1725 	stats->tor += IGC_READ_REG(hw, IGC_TORL);
1726 	stats->tor += ((uint64_t)IGC_READ_REG(hw, IGC_TORH) << 32);
1727 	stats->tor -= (stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
1728 	stats->tot += IGC_READ_REG(hw, IGC_TOTL);
1729 	stats->tot += ((uint64_t)IGC_READ_REG(hw, IGC_TOTH) << 32);
1730 	stats->tot -= (stats->tpt - old_tpt) * RTE_ETHER_CRC_LEN;
1731 
1732 	stats->ptc64 += IGC_READ_REG(hw, IGC_PTC64);
1733 	stats->ptc127 += IGC_READ_REG(hw, IGC_PTC127);
1734 	stats->ptc255 += IGC_READ_REG(hw, IGC_PTC255);
1735 	stats->ptc511 += IGC_READ_REG(hw, IGC_PTC511);
1736 	stats->ptc1023 += IGC_READ_REG(hw, IGC_PTC1023);
1737 	stats->ptc1522 += IGC_READ_REG(hw, IGC_PTC1522);
1738 	stats->mptc += IGC_READ_REG(hw, IGC_MPTC);
1739 	stats->bptc += IGC_READ_REG(hw, IGC_BPTC);
1740 	stats->tsctc += IGC_READ_REG(hw, IGC_TSCTC);
1741 
1742 	stats->iac += IGC_READ_REG(hw, IGC_IAC);
1743 	stats->rpthc += IGC_READ_REG(hw, IGC_RPTHC);
1744 	stats->hgptc += IGC_READ_REG(hw, IGC_HGPTC);
1745 	stats->icrxdmtc += IGC_READ_REG(hw, IGC_ICRXDMTC);
1746 
1747 	/* Host to Card Statistics */
1748 	stats->hgorc += IGC_READ_REG(hw, IGC_HGORCL);
1749 	stats->hgorc += ((uint64_t)IGC_READ_REG(hw, IGC_HGORCH) << 32);
1750 	stats->hgorc -= (stats->rpthc - old_rpthc) * RTE_ETHER_CRC_LEN;
1751 	stats->hgotc += IGC_READ_REG(hw, IGC_HGOTCL);
1752 	stats->hgotc += ((uint64_t)IGC_READ_REG(hw, IGC_HGOTCH) << 32);
1753 	stats->hgotc -= (stats->hgptc - old_hgptc) * RTE_ETHER_CRC_LEN;
1754 	stats->lenerrs += IGC_READ_REG(hw, IGC_LENERRS);
1755 }
1756 
1757 /*
1758  * Write 0 to all queue status registers
1759  */
1760 static void
1761 igc_reset_queue_stats_register(struct igc_hw *hw)
1762 {
1763 	int i;
1764 
1765 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1766 		IGC_WRITE_REG(hw, IGC_PQGPRC(i), 0);
1767 		IGC_WRITE_REG(hw, IGC_PQGPTC(i), 0);
1768 		IGC_WRITE_REG(hw, IGC_PQGORC(i), 0);
1769 		IGC_WRITE_REG(hw, IGC_PQGOTC(i), 0);
1770 		IGC_WRITE_REG(hw, IGC_PQMPRC(i), 0);
1771 		IGC_WRITE_REG(hw, IGC_RQDPC(i), 0);
1772 		IGC_WRITE_REG(hw, IGC_TQDPC(i), 0);
1773 	}
1774 }
1775 
1776 /*
1777  * Read all hardware queue status registers
1778  */
1779 static void
1780 igc_read_queue_stats_register(struct rte_eth_dev *dev)
1781 {
1782 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1783 	struct igc_hw_queue_stats *queue_stats =
1784 				IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1785 	int i;
1786 
1787 	/*
1788 	 * This register is not cleared on read. Furthermore, the register wraps
1789 	 * around back to 0x00000000 on the next increment when reaching a value
1790 	 * of 0xFFFFFFFF and then continues normal count operation.
1791 	 */
1792 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1793 		union {
1794 			u64 ddword;
1795 			u32 dword[2];
1796 		} value;
1797 		u32 tmp;
1798 
1799 		/*
1800 		 * Read the register first, if the value is smaller than that
1801 		 * previous read, that mean the register has been overflowed,
1802 		 * then we add the high 4 bytes by 1 and replace the low 4
1803 		 * bytes by the new value.
1804 		 */
1805 		tmp = IGC_READ_REG(hw, IGC_PQGPRC(i));
1806 		value.ddword = queue_stats->pqgprc[i];
1807 		if (value.dword[U32_0_IN_U64] > tmp)
1808 			value.dword[U32_1_IN_U64]++;
1809 		value.dword[U32_0_IN_U64] = tmp;
1810 		queue_stats->pqgprc[i] = value.ddword;
1811 
1812 		tmp = IGC_READ_REG(hw, IGC_PQGPTC(i));
1813 		value.ddword = queue_stats->pqgptc[i];
1814 		if (value.dword[U32_0_IN_U64] > tmp)
1815 			value.dword[U32_1_IN_U64]++;
1816 		value.dword[U32_0_IN_U64] = tmp;
1817 		queue_stats->pqgptc[i] = value.ddword;
1818 
1819 		tmp = IGC_READ_REG(hw, IGC_PQGORC(i));
1820 		value.ddword = queue_stats->pqgorc[i];
1821 		if (value.dword[U32_0_IN_U64] > tmp)
1822 			value.dword[U32_1_IN_U64]++;
1823 		value.dword[U32_0_IN_U64] = tmp;
1824 		queue_stats->pqgorc[i] = value.ddword;
1825 
1826 		tmp = IGC_READ_REG(hw, IGC_PQGOTC(i));
1827 		value.ddword = queue_stats->pqgotc[i];
1828 		if (value.dword[U32_0_IN_U64] > tmp)
1829 			value.dword[U32_1_IN_U64]++;
1830 		value.dword[U32_0_IN_U64] = tmp;
1831 		queue_stats->pqgotc[i] = value.ddword;
1832 
1833 		tmp = IGC_READ_REG(hw, IGC_PQMPRC(i));
1834 		value.ddword = queue_stats->pqmprc[i];
1835 		if (value.dword[U32_0_IN_U64] > tmp)
1836 			value.dword[U32_1_IN_U64]++;
1837 		value.dword[U32_0_IN_U64] = tmp;
1838 		queue_stats->pqmprc[i] = value.ddword;
1839 
1840 		tmp = IGC_READ_REG(hw, IGC_RQDPC(i));
1841 		value.ddword = queue_stats->rqdpc[i];
1842 		if (value.dword[U32_0_IN_U64] > tmp)
1843 			value.dword[U32_1_IN_U64]++;
1844 		value.dword[U32_0_IN_U64] = tmp;
1845 		queue_stats->rqdpc[i] = value.ddword;
1846 
1847 		tmp = IGC_READ_REG(hw, IGC_TQDPC(i));
1848 		value.ddword = queue_stats->tqdpc[i];
1849 		if (value.dword[U32_0_IN_U64] > tmp)
1850 			value.dword[U32_1_IN_U64]++;
1851 		value.dword[U32_0_IN_U64] = tmp;
1852 		queue_stats->tqdpc[i] = value.ddword;
1853 	}
1854 }
1855 
1856 static int
1857 eth_igc_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1858 {
1859 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
1860 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1861 	struct igc_hw_stats *stats = IGC_DEV_PRIVATE_STATS(dev);
1862 	struct igc_hw_queue_stats *queue_stats =
1863 			IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1864 	int i;
1865 
1866 	/*
1867 	 * Cancel status handler since it will read the queue status registers
1868 	 */
1869 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
1870 
1871 	/* Read status register */
1872 	igc_read_queue_stats_register(dev);
1873 	igc_read_stats_registers(hw, stats);
1874 
1875 	if (rte_stats == NULL) {
1876 		/* Restart queue status handler */
1877 		rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1878 				igc_update_queue_stats_handler, dev);
1879 		return -EINVAL;
1880 	}
1881 
1882 	/* Rx Errors */
1883 	rte_stats->imissed = stats->mpc;
1884 	rte_stats->ierrors = stats->crcerrs + stats->rlec +
1885 			stats->rxerrc + stats->algnerrc;
1886 
1887 	/* Tx Errors */
1888 	rte_stats->oerrors = stats->ecol + stats->latecol;
1889 
1890 	rte_stats->ipackets = stats->gprc;
1891 	rte_stats->opackets = stats->gptc;
1892 	rte_stats->ibytes   = stats->gorc;
1893 	rte_stats->obytes   = stats->gotc;
1894 
1895 	/* Get per-queue statuses */
1896 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1897 		/* GET TX queue statuses */
1898 		int map_id = igc->txq_stats_map[i];
1899 		if (map_id >= 0) {
1900 			rte_stats->q_opackets[map_id] += queue_stats->pqgptc[i];
1901 			rte_stats->q_obytes[map_id] += queue_stats->pqgotc[i];
1902 		}
1903 		/* Get RX queue statuses */
1904 		map_id = igc->rxq_stats_map[i];
1905 		if (map_id >= 0) {
1906 			rte_stats->q_ipackets[map_id] += queue_stats->pqgprc[i];
1907 			rte_stats->q_ibytes[map_id] += queue_stats->pqgorc[i];
1908 			rte_stats->q_errors[map_id] += queue_stats->rqdpc[i];
1909 		}
1910 	}
1911 
1912 	/* Restart queue status handler */
1913 	rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1914 			igc_update_queue_stats_handler, dev);
1915 	return 0;
1916 }
1917 
1918 static int
1919 eth_igc_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1920 		   unsigned int n)
1921 {
1922 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1923 	struct igc_hw_stats *hw_stats =
1924 			IGC_DEV_PRIVATE_STATS(dev);
1925 	unsigned int i;
1926 
1927 	igc_read_stats_registers(hw, hw_stats);
1928 
1929 	if (n < IGC_NB_XSTATS)
1930 		return IGC_NB_XSTATS;
1931 
1932 	/* If this is a reset xstats is NULL, and we have cleared the
1933 	 * registers by reading them.
1934 	 */
1935 	if (!xstats)
1936 		return 0;
1937 
1938 	/* Extended stats */
1939 	for (i = 0; i < IGC_NB_XSTATS; i++) {
1940 		xstats[i].id = i;
1941 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1942 			rte_igc_stats_strings[i].offset);
1943 	}
1944 
1945 	return IGC_NB_XSTATS;
1946 }
1947 
1948 static int
1949 eth_igc_xstats_reset(struct rte_eth_dev *dev)
1950 {
1951 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1952 	struct igc_hw_stats *hw_stats = IGC_DEV_PRIVATE_STATS(dev);
1953 	struct igc_hw_queue_stats *queue_stats =
1954 			IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1955 
1956 	/* Cancel queue status handler for avoid conflict */
1957 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
1958 
1959 	/* HW registers are cleared on read */
1960 	igc_reset_queue_stats_register(hw);
1961 	igc_read_stats_registers(hw, hw_stats);
1962 
1963 	/* Reset software totals */
1964 	memset(hw_stats, 0, sizeof(*hw_stats));
1965 	memset(queue_stats, 0, sizeof(*queue_stats));
1966 
1967 	/* Restart the queue status handler */
1968 	rte_eal_alarm_set(IGC_ALARM_INTERVAL, igc_update_queue_stats_handler,
1969 			dev);
1970 
1971 	return 0;
1972 }
1973 
1974 static int
1975 eth_igc_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1976 	struct rte_eth_xstat_name *xstats_names, unsigned int size)
1977 {
1978 	unsigned int i;
1979 
1980 	if (xstats_names == NULL)
1981 		return IGC_NB_XSTATS;
1982 
1983 	if (size < IGC_NB_XSTATS) {
1984 		PMD_DRV_LOG(ERR, "not enough buffers!");
1985 		return IGC_NB_XSTATS;
1986 	}
1987 
1988 	for (i = 0; i < IGC_NB_XSTATS; i++)
1989 		strlcpy(xstats_names[i].name, rte_igc_stats_strings[i].name,
1990 			sizeof(xstats_names[i].name));
1991 
1992 	return IGC_NB_XSTATS;
1993 }
1994 
1995 static int
1996 eth_igc_xstats_get_names_by_id(struct rte_eth_dev *dev,
1997 		const uint64_t *ids, struct rte_eth_xstat_name *xstats_names,
1998 		unsigned int limit)
1999 {
2000 	unsigned int i;
2001 
2002 	if (!ids)
2003 		return eth_igc_xstats_get_names(dev, xstats_names, limit);
2004 
2005 	for (i = 0; i < limit; i++) {
2006 		if (ids[i] >= IGC_NB_XSTATS) {
2007 			PMD_DRV_LOG(ERR, "id value isn't valid");
2008 			return -EINVAL;
2009 		}
2010 		strlcpy(xstats_names[i].name,
2011 			rte_igc_stats_strings[ids[i]].name,
2012 			sizeof(xstats_names[i].name));
2013 	}
2014 	return limit;
2015 }
2016 
2017 static int
2018 eth_igc_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
2019 		uint64_t *values, unsigned int n)
2020 {
2021 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2022 	struct igc_hw_stats *hw_stats = IGC_DEV_PRIVATE_STATS(dev);
2023 	unsigned int i;
2024 
2025 	igc_read_stats_registers(hw, hw_stats);
2026 
2027 	if (!ids) {
2028 		if (n < IGC_NB_XSTATS)
2029 			return IGC_NB_XSTATS;
2030 
2031 		/* If this is a reset xstats is NULL, and we have cleared the
2032 		 * registers by reading them.
2033 		 */
2034 		if (!values)
2035 			return 0;
2036 
2037 		/* Extended stats */
2038 		for (i = 0; i < IGC_NB_XSTATS; i++)
2039 			values[i] = *(uint64_t *)(((char *)hw_stats) +
2040 					rte_igc_stats_strings[i].offset);
2041 
2042 		return IGC_NB_XSTATS;
2043 
2044 	} else {
2045 		for (i = 0; i < n; i++) {
2046 			if (ids[i] >= IGC_NB_XSTATS) {
2047 				PMD_DRV_LOG(ERR, "id value isn't valid");
2048 				return -EINVAL;
2049 			}
2050 			values[i] = *(uint64_t *)(((char *)hw_stats) +
2051 					rte_igc_stats_strings[ids[i]].offset);
2052 		}
2053 		return n;
2054 	}
2055 }
2056 
2057 static int
2058 eth_igc_queue_stats_mapping_set(struct rte_eth_dev *dev,
2059 		uint16_t queue_id, uint8_t stat_idx, uint8_t is_rx)
2060 {
2061 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
2062 
2063 	/* check queue id is valid */
2064 	if (queue_id >= IGC_QUEUE_PAIRS_NUM) {
2065 		PMD_DRV_LOG(ERR, "queue id(%u) error, max is %u",
2066 			queue_id, IGC_QUEUE_PAIRS_NUM - 1);
2067 		return -EINVAL;
2068 	}
2069 
2070 	/* store the mapping status id */
2071 	if (is_rx)
2072 		igc->rxq_stats_map[queue_id] = stat_idx;
2073 	else
2074 		igc->txq_stats_map[queue_id] = stat_idx;
2075 
2076 	return 0;
2077 }
2078 
2079 static int
2080 eth_igc_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
2081 {
2082 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2083 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2084 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2085 	uint32_t vec = IGC_MISC_VEC_ID;
2086 
2087 	if (rte_intr_allow_others(intr_handle))
2088 		vec = IGC_RX_VEC_START;
2089 
2090 	uint32_t mask = 1u << (queue_id + vec);
2091 
2092 	IGC_WRITE_REG(hw, IGC_EIMC, mask);
2093 	IGC_WRITE_FLUSH(hw);
2094 
2095 	return 0;
2096 }
2097 
2098 static int
2099 eth_igc_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
2100 {
2101 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2102 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2103 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2104 	uint32_t vec = IGC_MISC_VEC_ID;
2105 
2106 	if (rte_intr_allow_others(intr_handle))
2107 		vec = IGC_RX_VEC_START;
2108 
2109 	uint32_t mask = 1u << (queue_id + vec);
2110 
2111 	IGC_WRITE_REG(hw, IGC_EIMS, mask);
2112 	IGC_WRITE_FLUSH(hw);
2113 
2114 	rte_intr_enable(intr_handle);
2115 
2116 	return 0;
2117 }
2118 
2119 static int
2120 eth_igc_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2121 {
2122 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2123 	uint32_t ctrl;
2124 	int tx_pause;
2125 	int rx_pause;
2126 
2127 	fc_conf->pause_time = hw->fc.pause_time;
2128 	fc_conf->high_water = hw->fc.high_water;
2129 	fc_conf->low_water = hw->fc.low_water;
2130 	fc_conf->send_xon = hw->fc.send_xon;
2131 	fc_conf->autoneg = hw->mac.autoneg;
2132 
2133 	/*
2134 	 * Return rx_pause and tx_pause status according to actual setting of
2135 	 * the TFCE and RFCE bits in the CTRL register.
2136 	 */
2137 	ctrl = IGC_READ_REG(hw, IGC_CTRL);
2138 	if (ctrl & IGC_CTRL_TFCE)
2139 		tx_pause = 1;
2140 	else
2141 		tx_pause = 0;
2142 
2143 	if (ctrl & IGC_CTRL_RFCE)
2144 		rx_pause = 1;
2145 	else
2146 		rx_pause = 0;
2147 
2148 	if (rx_pause && tx_pause)
2149 		fc_conf->mode = RTE_FC_FULL;
2150 	else if (rx_pause)
2151 		fc_conf->mode = RTE_FC_RX_PAUSE;
2152 	else if (tx_pause)
2153 		fc_conf->mode = RTE_FC_TX_PAUSE;
2154 	else
2155 		fc_conf->mode = RTE_FC_NONE;
2156 
2157 	return 0;
2158 }
2159 
2160 static int
2161 eth_igc_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2162 {
2163 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2164 	uint32_t rx_buf_size;
2165 	uint32_t max_high_water;
2166 	uint32_t rctl;
2167 	int err;
2168 
2169 	if (fc_conf->autoneg != hw->mac.autoneg)
2170 		return -ENOTSUP;
2171 
2172 	rx_buf_size = igc_get_rx_buffer_size(hw);
2173 	PMD_DRV_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
2174 
2175 	/* At least reserve one Ethernet frame for watermark */
2176 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
2177 	if (fc_conf->high_water > max_high_water ||
2178 		fc_conf->high_water < fc_conf->low_water) {
2179 		PMD_DRV_LOG(ERR,
2180 			"Incorrect high(%u)/low(%u) water value, max is %u",
2181 			fc_conf->high_water, fc_conf->low_water,
2182 			max_high_water);
2183 		return -EINVAL;
2184 	}
2185 
2186 	switch (fc_conf->mode) {
2187 	case RTE_FC_NONE:
2188 		hw->fc.requested_mode = igc_fc_none;
2189 		break;
2190 	case RTE_FC_RX_PAUSE:
2191 		hw->fc.requested_mode = igc_fc_rx_pause;
2192 		break;
2193 	case RTE_FC_TX_PAUSE:
2194 		hw->fc.requested_mode = igc_fc_tx_pause;
2195 		break;
2196 	case RTE_FC_FULL:
2197 		hw->fc.requested_mode = igc_fc_full;
2198 		break;
2199 	default:
2200 		PMD_DRV_LOG(ERR, "unsupported fc mode: %u", fc_conf->mode);
2201 		return -EINVAL;
2202 	}
2203 
2204 	hw->fc.pause_time     = fc_conf->pause_time;
2205 	hw->fc.high_water     = fc_conf->high_water;
2206 	hw->fc.low_water      = fc_conf->low_water;
2207 	hw->fc.send_xon	      = fc_conf->send_xon;
2208 
2209 	err = igc_setup_link_generic(hw);
2210 	if (err == IGC_SUCCESS) {
2211 		/**
2212 		 * check if we want to forward MAC frames - driver doesn't have
2213 		 * native capability to do that, so we'll write the registers
2214 		 * ourselves
2215 		 **/
2216 		rctl = IGC_READ_REG(hw, IGC_RCTL);
2217 
2218 		/* set or clear MFLCN.PMCF bit depending on configuration */
2219 		if (fc_conf->mac_ctrl_frame_fwd != 0)
2220 			rctl |= IGC_RCTL_PMCF;
2221 		else
2222 			rctl &= ~IGC_RCTL_PMCF;
2223 
2224 		IGC_WRITE_REG(hw, IGC_RCTL, rctl);
2225 		IGC_WRITE_FLUSH(hw);
2226 
2227 		return 0;
2228 	}
2229 
2230 	PMD_DRV_LOG(ERR, "igc_setup_link_generic = 0x%x", err);
2231 	return -EIO;
2232 }
2233 
2234 static int
2235 eth_igc_rss_reta_update(struct rte_eth_dev *dev,
2236 			struct rte_eth_rss_reta_entry64 *reta_conf,
2237 			uint16_t reta_size)
2238 {
2239 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2240 	uint16_t i;
2241 
2242 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
2243 		PMD_DRV_LOG(ERR,
2244 			"The size of RSS redirection table configured(%d) doesn't match the number hardware can supported(%d)",
2245 			reta_size, ETH_RSS_RETA_SIZE_128);
2246 		return -EINVAL;
2247 	}
2248 
2249 	RTE_BUILD_BUG_ON(ETH_RSS_RETA_SIZE_128 % IGC_RSS_RDT_REG_SIZE);
2250 
2251 	/* set redirection table */
2252 	for (i = 0; i < ETH_RSS_RETA_SIZE_128; i += IGC_RSS_RDT_REG_SIZE) {
2253 		union igc_rss_reta_reg reta, reg;
2254 		uint16_t idx, shift;
2255 		uint8_t j, mask;
2256 
2257 		idx = i / RTE_RETA_GROUP_SIZE;
2258 		shift = i % RTE_RETA_GROUP_SIZE;
2259 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
2260 				IGC_RSS_RDT_REG_SIZE_MASK);
2261 
2262 		/* if no need to update the register */
2263 		if (!mask ||
2264 		    shift > (RTE_RETA_GROUP_SIZE - IGC_RSS_RDT_REG_SIZE))
2265 			continue;
2266 
2267 		/* check mask whether need to read the register value first */
2268 		if (mask == IGC_RSS_RDT_REG_SIZE_MASK)
2269 			reg.dword = 0;
2270 		else
2271 			reg.dword = IGC_READ_REG_LE_VALUE(hw,
2272 					IGC_RETA(i / IGC_RSS_RDT_REG_SIZE));
2273 
2274 		/* update the register */
2275 		RTE_BUILD_BUG_ON(sizeof(reta.bytes) != IGC_RSS_RDT_REG_SIZE);
2276 		for (j = 0; j < IGC_RSS_RDT_REG_SIZE; j++) {
2277 			if (mask & (1u << j))
2278 				reta.bytes[j] =
2279 					(uint8_t)reta_conf[idx].reta[shift + j];
2280 			else
2281 				reta.bytes[j] = reg.bytes[j];
2282 		}
2283 		IGC_WRITE_REG_LE_VALUE(hw,
2284 			IGC_RETA(i / IGC_RSS_RDT_REG_SIZE), reta.dword);
2285 	}
2286 
2287 	return 0;
2288 }
2289 
2290 static int
2291 eth_igc_rss_reta_query(struct rte_eth_dev *dev,
2292 		       struct rte_eth_rss_reta_entry64 *reta_conf,
2293 		       uint16_t reta_size)
2294 {
2295 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2296 	uint16_t i;
2297 
2298 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
2299 		PMD_DRV_LOG(ERR,
2300 			"The size of RSS redirection table configured(%d) doesn't match the number hardware can supported(%d)",
2301 			reta_size, ETH_RSS_RETA_SIZE_128);
2302 		return -EINVAL;
2303 	}
2304 
2305 	RTE_BUILD_BUG_ON(ETH_RSS_RETA_SIZE_128 % IGC_RSS_RDT_REG_SIZE);
2306 
2307 	/* read redirection table */
2308 	for (i = 0; i < ETH_RSS_RETA_SIZE_128; i += IGC_RSS_RDT_REG_SIZE) {
2309 		union igc_rss_reta_reg reta;
2310 		uint16_t idx, shift;
2311 		uint8_t j, mask;
2312 
2313 		idx = i / RTE_RETA_GROUP_SIZE;
2314 		shift = i % RTE_RETA_GROUP_SIZE;
2315 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
2316 				IGC_RSS_RDT_REG_SIZE_MASK);
2317 
2318 		/* if no need to read register */
2319 		if (!mask ||
2320 		    shift > (RTE_RETA_GROUP_SIZE - IGC_RSS_RDT_REG_SIZE))
2321 			continue;
2322 
2323 		/* read register and get the queue index */
2324 		RTE_BUILD_BUG_ON(sizeof(reta.bytes) != IGC_RSS_RDT_REG_SIZE);
2325 		reta.dword = IGC_READ_REG_LE_VALUE(hw,
2326 				IGC_RETA(i / IGC_RSS_RDT_REG_SIZE));
2327 		for (j = 0; j < IGC_RSS_RDT_REG_SIZE; j++) {
2328 			if (mask & (1u << j))
2329 				reta_conf[idx].reta[shift + j] = reta.bytes[j];
2330 		}
2331 	}
2332 
2333 	return 0;
2334 }
2335 
2336 static int
2337 eth_igc_rss_hash_update(struct rte_eth_dev *dev,
2338 			struct rte_eth_rss_conf *rss_conf)
2339 {
2340 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2341 	igc_hw_rss_hash_set(hw, rss_conf);
2342 	return 0;
2343 }
2344 
2345 static int
2346 eth_igc_rss_hash_conf_get(struct rte_eth_dev *dev,
2347 			struct rte_eth_rss_conf *rss_conf)
2348 {
2349 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2350 	uint32_t *hash_key = (uint32_t *)rss_conf->rss_key;
2351 	uint32_t mrqc;
2352 	uint64_t rss_hf;
2353 
2354 	if (hash_key != NULL) {
2355 		int i;
2356 
2357 		/* if not enough space for store hash key */
2358 		if (rss_conf->rss_key_len != IGC_HKEY_SIZE) {
2359 			PMD_DRV_LOG(ERR,
2360 				"RSS hash key size %u in parameter doesn't match the hardware hash key size %u",
2361 				rss_conf->rss_key_len, IGC_HKEY_SIZE);
2362 			return -EINVAL;
2363 		}
2364 
2365 		/* read RSS key from register */
2366 		for (i = 0; i < IGC_HKEY_MAX_INDEX; i++)
2367 			hash_key[i] = IGC_READ_REG_LE_VALUE(hw, IGC_RSSRK(i));
2368 	}
2369 
2370 	/* get RSS functions configured in MRQC register */
2371 	mrqc = IGC_READ_REG(hw, IGC_MRQC);
2372 	if ((mrqc & IGC_MRQC_ENABLE_RSS_4Q) == 0)
2373 		return 0;
2374 
2375 	rss_hf = 0;
2376 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4)
2377 		rss_hf |= ETH_RSS_IPV4;
2378 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4_TCP)
2379 		rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2380 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6)
2381 		rss_hf |= ETH_RSS_IPV6;
2382 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_EX)
2383 		rss_hf |= ETH_RSS_IPV6_EX;
2384 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_TCP)
2385 		rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2386 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_TCP_EX)
2387 		rss_hf |= ETH_RSS_IPV6_TCP_EX;
2388 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4_UDP)
2389 		rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2390 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_UDP)
2391 		rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2392 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_UDP_EX)
2393 		rss_hf |= ETH_RSS_IPV6_UDP_EX;
2394 
2395 	rss_conf->rss_hf |= rss_hf;
2396 	return 0;
2397 }
2398 
2399 static int
2400 eth_igc_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2401 {
2402 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2403 	struct igc_vfta *shadow_vfta = IGC_DEV_PRIVATE_VFTA(dev);
2404 	uint32_t vfta;
2405 	uint32_t vid_idx;
2406 	uint32_t vid_bit;
2407 
2408 	vid_idx = (vlan_id >> IGC_VFTA_ENTRY_SHIFT) & IGC_VFTA_ENTRY_MASK;
2409 	vid_bit = 1u << (vlan_id & IGC_VFTA_ENTRY_BIT_SHIFT_MASK);
2410 	vfta = shadow_vfta->vfta[vid_idx];
2411 	if (on)
2412 		vfta |= vid_bit;
2413 	else
2414 		vfta &= ~vid_bit;
2415 	IGC_WRITE_REG_ARRAY(hw, IGC_VFTA, vid_idx, vfta);
2416 
2417 	/* update local VFTA copy */
2418 	shadow_vfta->vfta[vid_idx] = vfta;
2419 
2420 	return 0;
2421 }
2422 
2423 static void
2424 igc_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2425 {
2426 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2427 	igc_read_reg_check_clear_bits(hw, IGC_RCTL,
2428 			IGC_RCTL_CFIEN | IGC_RCTL_VFE);
2429 }
2430 
2431 static void
2432 igc_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2433 {
2434 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2435 	struct igc_vfta *shadow_vfta = IGC_DEV_PRIVATE_VFTA(dev);
2436 	uint32_t reg_val;
2437 	int i;
2438 
2439 	/* Filter Table Enable, CFI not used for packet acceptance */
2440 	reg_val = IGC_READ_REG(hw, IGC_RCTL);
2441 	reg_val &= ~IGC_RCTL_CFIEN;
2442 	reg_val |= IGC_RCTL_VFE;
2443 	IGC_WRITE_REG(hw, IGC_RCTL, reg_val);
2444 
2445 	/* restore VFTA table */
2446 	for (i = 0; i < IGC_VFTA_SIZE; i++)
2447 		IGC_WRITE_REG_ARRAY(hw, IGC_VFTA, i, shadow_vfta->vfta[i]);
2448 }
2449 
2450 static void
2451 igc_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2452 {
2453 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2454 
2455 	igc_read_reg_check_clear_bits(hw, IGC_CTRL, IGC_CTRL_VME);
2456 }
2457 
2458 static void
2459 igc_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2460 {
2461 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2462 
2463 	igc_read_reg_check_set_bits(hw, IGC_CTRL, IGC_CTRL_VME);
2464 }
2465 
2466 static int
2467 igc_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2468 {
2469 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2470 	uint32_t frame_size = dev->data->mtu + IGC_ETH_OVERHEAD;
2471 	uint32_t ctrl_ext;
2472 
2473 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
2474 
2475 	/* if extend vlan hasn't been enabled */
2476 	if ((ctrl_ext & IGC_CTRL_EXT_EXT_VLAN) == 0)
2477 		return 0;
2478 
2479 	/* Update maximum packet length */
2480 	if (frame_size < RTE_ETHER_MIN_MTU + VLAN_TAG_SIZE) {
2481 		PMD_DRV_LOG(ERR, "Maximum packet length %u error, min is %u",
2482 			frame_size, VLAN_TAG_SIZE + RTE_ETHER_MIN_MTU);
2483 		return -EINVAL;
2484 	}
2485 	IGC_WRITE_REG(hw, IGC_RLPML, frame_size - VLAN_TAG_SIZE);
2486 
2487 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext & ~IGC_CTRL_EXT_EXT_VLAN);
2488 	return 0;
2489 }
2490 
2491 static int
2492 igc_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2493 {
2494 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2495 	uint32_t frame_size = dev->data->mtu + IGC_ETH_OVERHEAD;
2496 	uint32_t ctrl_ext;
2497 
2498 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
2499 
2500 	/* if extend vlan has been enabled */
2501 	if (ctrl_ext & IGC_CTRL_EXT_EXT_VLAN)
2502 		return 0;
2503 
2504 	/* Update maximum packet length */
2505 	if (frame_size > MAX_RX_JUMBO_FRAME_SIZE) {
2506 		PMD_DRV_LOG(ERR, "Maximum packet length %u error, max is %u",
2507 			frame_size, MAX_RX_JUMBO_FRAME_SIZE);
2508 		return -EINVAL;
2509 	}
2510 	IGC_WRITE_REG(hw, IGC_RLPML, frame_size);
2511 
2512 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext | IGC_CTRL_EXT_EXT_VLAN);
2513 	return 0;
2514 }
2515 
2516 static int
2517 eth_igc_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2518 {
2519 	struct rte_eth_rxmode *rxmode;
2520 
2521 	rxmode = &dev->data->dev_conf.rxmode;
2522 	if (mask & ETH_VLAN_STRIP_MASK) {
2523 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2524 			igc_vlan_hw_strip_enable(dev);
2525 		else
2526 			igc_vlan_hw_strip_disable(dev);
2527 	}
2528 
2529 	if (mask & ETH_VLAN_FILTER_MASK) {
2530 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2531 			igc_vlan_hw_filter_enable(dev);
2532 		else
2533 			igc_vlan_hw_filter_disable(dev);
2534 	}
2535 
2536 	if (mask & ETH_VLAN_EXTEND_MASK) {
2537 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2538 			return igc_vlan_hw_extend_enable(dev);
2539 		else
2540 			return igc_vlan_hw_extend_disable(dev);
2541 	}
2542 
2543 	return 0;
2544 }
2545 
2546 static int
2547 eth_igc_vlan_tpid_set(struct rte_eth_dev *dev,
2548 		      enum rte_vlan_type vlan_type,
2549 		      uint16_t tpid)
2550 {
2551 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2552 	uint32_t reg_val;
2553 
2554 	/* only outer TPID of double VLAN can be configured*/
2555 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2556 		reg_val = IGC_READ_REG(hw, IGC_VET);
2557 		reg_val = (reg_val & (~IGC_VET_EXT)) |
2558 			((uint32_t)tpid << IGC_VET_EXT_SHIFT);
2559 		IGC_WRITE_REG(hw, IGC_VET, reg_val);
2560 
2561 		return 0;
2562 	}
2563 
2564 	/* all other TPID values are read-only*/
2565 	PMD_DRV_LOG(ERR, "Not supported");
2566 	return -ENOTSUP;
2567 }
2568 
2569 static int
2570 eth_igc_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
2571 	struct rte_pci_device *pci_dev)
2572 {
2573 	PMD_INIT_FUNC_TRACE();
2574 	return rte_eth_dev_pci_generic_probe(pci_dev,
2575 		sizeof(struct igc_adapter), eth_igc_dev_init);
2576 }
2577 
2578 static int
2579 eth_igc_pci_remove(struct rte_pci_device *pci_dev)
2580 {
2581 	PMD_INIT_FUNC_TRACE();
2582 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igc_dev_uninit);
2583 }
2584 
2585 static struct rte_pci_driver rte_igc_pmd = {
2586 	.id_table = pci_id_igc_map,
2587 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
2588 	.probe = eth_igc_pci_probe,
2589 	.remove = eth_igc_pci_remove,
2590 };
2591 
2592 RTE_PMD_REGISTER_PCI(net_igc, rte_igc_pmd);
2593 RTE_PMD_REGISTER_PCI_TABLE(net_igc, pci_id_igc_map);
2594 RTE_PMD_REGISTER_KMOD_DEP(net_igc, "* igb_uio | uio_pci_generic | vfio-pci");
2595