xref: /dpdk/drivers/net/igc/igc_ethdev.c (revision 0c2b79e8)
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 void 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 	.rx_queue_count		= eth_igc_rx_queue_count,
276 	.rx_descriptor_done	= eth_igc_rx_descriptor_done,
277 	.rx_descriptor_status	= eth_igc_rx_descriptor_status,
278 	.tx_descriptor_status	= eth_igc_tx_descriptor_status,
279 	.tx_queue_setup		= eth_igc_tx_queue_setup,
280 	.tx_queue_release	= eth_igc_tx_queue_release,
281 	.tx_done_cleanup	= eth_igc_tx_done_cleanup,
282 	.rxq_info_get		= eth_igc_rxq_info_get,
283 	.txq_info_get		= eth_igc_txq_info_get,
284 	.stats_get		= eth_igc_stats_get,
285 	.xstats_get		= eth_igc_xstats_get,
286 	.xstats_get_by_id	= eth_igc_xstats_get_by_id,
287 	.xstats_get_names_by_id	= eth_igc_xstats_get_names_by_id,
288 	.xstats_get_names	= eth_igc_xstats_get_names,
289 	.stats_reset		= eth_igc_xstats_reset,
290 	.xstats_reset		= eth_igc_xstats_reset,
291 	.queue_stats_mapping_set = eth_igc_queue_stats_mapping_set,
292 	.rx_queue_intr_enable	= eth_igc_rx_queue_intr_enable,
293 	.rx_queue_intr_disable	= eth_igc_rx_queue_intr_disable,
294 	.flow_ctrl_get		= eth_igc_flow_ctrl_get,
295 	.flow_ctrl_set		= eth_igc_flow_ctrl_set,
296 	.reta_update		= eth_igc_rss_reta_update,
297 	.reta_query		= eth_igc_rss_reta_query,
298 	.rss_hash_update	= eth_igc_rss_hash_update,
299 	.rss_hash_conf_get	= eth_igc_rss_hash_conf_get,
300 	.vlan_filter_set	= eth_igc_vlan_filter_set,
301 	.vlan_offload_set	= eth_igc_vlan_offload_set,
302 	.vlan_tpid_set		= eth_igc_vlan_tpid_set,
303 	.vlan_strip_queue_set	= eth_igc_vlan_strip_queue_set,
304 	.filter_ctrl		= eth_igc_filter_ctrl,
305 };
306 
307 /*
308  * multiple queue mode checking
309  */
310 static int
311 igc_check_mq_mode(struct rte_eth_dev *dev)
312 {
313 	enum rte_eth_rx_mq_mode rx_mq_mode = dev->data->dev_conf.rxmode.mq_mode;
314 	enum rte_eth_tx_mq_mode tx_mq_mode = dev->data->dev_conf.txmode.mq_mode;
315 
316 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
317 		PMD_INIT_LOG(ERR, "SRIOV is not supported.");
318 		return -EINVAL;
319 	}
320 
321 	if (rx_mq_mode != ETH_MQ_RX_NONE &&
322 		rx_mq_mode != ETH_MQ_RX_RSS) {
323 		/* RSS together with VMDq not supported*/
324 		PMD_INIT_LOG(ERR, "RX mode %d is not supported.",
325 				rx_mq_mode);
326 		return -EINVAL;
327 	}
328 
329 	/* To no break software that set invalid mode, only display
330 	 * warning if invalid mode is used.
331 	 */
332 	if (tx_mq_mode != ETH_MQ_TX_NONE)
333 		PMD_INIT_LOG(WARNING,
334 			"TX mode %d is not supported. Due to meaningless in this driver, just ignore",
335 			tx_mq_mode);
336 
337 	return 0;
338 }
339 
340 static int
341 eth_igc_configure(struct rte_eth_dev *dev)
342 {
343 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
344 	int ret;
345 
346 	PMD_INIT_FUNC_TRACE();
347 
348 	ret  = igc_check_mq_mode(dev);
349 	if (ret != 0)
350 		return ret;
351 
352 	intr->flags |= IGC_FLAG_NEED_LINK_UPDATE;
353 	return 0;
354 }
355 
356 static int
357 eth_igc_set_link_up(struct rte_eth_dev *dev)
358 {
359 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
360 
361 	if (hw->phy.media_type == igc_media_type_copper)
362 		igc_power_up_phy(hw);
363 	else
364 		igc_power_up_fiber_serdes_link(hw);
365 	return 0;
366 }
367 
368 static int
369 eth_igc_set_link_down(struct rte_eth_dev *dev)
370 {
371 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
372 
373 	if (hw->phy.media_type == igc_media_type_copper)
374 		igc_power_down_phy(hw);
375 	else
376 		igc_shutdown_fiber_serdes_link(hw);
377 	return 0;
378 }
379 
380 /*
381  * disable other interrupt
382  */
383 static void
384 igc_intr_other_disable(struct rte_eth_dev *dev)
385 {
386 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
387 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
388 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
389 
390 	if (rte_intr_allow_others(intr_handle) &&
391 		dev->data->dev_conf.intr_conf.lsc) {
392 		IGC_WRITE_REG(hw, IGC_EIMC, 1u << IGC_MSIX_OTHER_INTR_VEC);
393 	}
394 
395 	IGC_WRITE_REG(hw, IGC_IMC, ~0);
396 	IGC_WRITE_FLUSH(hw);
397 }
398 
399 /*
400  * enable other interrupt
401  */
402 static inline void
403 igc_intr_other_enable(struct rte_eth_dev *dev)
404 {
405 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
406 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
407 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
408 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
409 
410 	if (rte_intr_allow_others(intr_handle) &&
411 		dev->data->dev_conf.intr_conf.lsc) {
412 		IGC_WRITE_REG(hw, IGC_EIMS, 1u << IGC_MSIX_OTHER_INTR_VEC);
413 	}
414 
415 	IGC_WRITE_REG(hw, IGC_IMS, intr->mask);
416 	IGC_WRITE_FLUSH(hw);
417 }
418 
419 /*
420  * It reads ICR and gets interrupt causes, check it and set a bit flag
421  * to update link status.
422  */
423 static void
424 eth_igc_interrupt_get_status(struct rte_eth_dev *dev)
425 {
426 	uint32_t icr;
427 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
428 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
429 
430 	/* read-on-clear nic registers here */
431 	icr = IGC_READ_REG(hw, IGC_ICR);
432 
433 	intr->flags = 0;
434 	if (icr & IGC_ICR_LSC)
435 		intr->flags |= IGC_FLAG_NEED_LINK_UPDATE;
436 }
437 
438 /* return 0 means link status changed, -1 means not changed */
439 static int
440 eth_igc_link_update(struct rte_eth_dev *dev, int wait_to_complete)
441 {
442 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
443 	struct rte_eth_link link;
444 	int link_check, count;
445 
446 	link_check = 0;
447 	hw->mac.get_link_status = 1;
448 
449 	/* possible wait-to-complete in up to 9 seconds */
450 	for (count = 0; count < IGC_LINK_UPDATE_CHECK_TIMEOUT; count++) {
451 		/* Read the real link status */
452 		switch (hw->phy.media_type) {
453 		case igc_media_type_copper:
454 			/* Do the work to read phy */
455 			igc_check_for_link(hw);
456 			link_check = !hw->mac.get_link_status;
457 			break;
458 
459 		case igc_media_type_fiber:
460 			igc_check_for_link(hw);
461 			link_check = (IGC_READ_REG(hw, IGC_STATUS) &
462 				      IGC_STATUS_LU);
463 			break;
464 
465 		case igc_media_type_internal_serdes:
466 			igc_check_for_link(hw);
467 			link_check = hw->mac.serdes_has_link;
468 			break;
469 
470 		default:
471 			break;
472 		}
473 		if (link_check || wait_to_complete == 0)
474 			break;
475 		rte_delay_ms(IGC_LINK_UPDATE_CHECK_INTERVAL);
476 	}
477 	memset(&link, 0, sizeof(link));
478 
479 	/* Now we check if a transition has happened */
480 	if (link_check) {
481 		uint16_t duplex, speed;
482 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
483 		link.link_duplex = (duplex == FULL_DUPLEX) ?
484 				ETH_LINK_FULL_DUPLEX :
485 				ETH_LINK_HALF_DUPLEX;
486 		link.link_speed = speed;
487 		link.link_status = ETH_LINK_UP;
488 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
489 				ETH_LINK_SPEED_FIXED);
490 
491 		if (speed == SPEED_2500) {
492 			uint32_t tipg = IGC_READ_REG(hw, IGC_TIPG);
493 			if ((tipg & IGC_TIPG_IPGT_MASK) != 0x0b) {
494 				tipg &= ~IGC_TIPG_IPGT_MASK;
495 				tipg |= 0x0b;
496 				IGC_WRITE_REG(hw, IGC_TIPG, tipg);
497 			}
498 		}
499 	} else {
500 		link.link_speed = 0;
501 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
502 		link.link_status = ETH_LINK_DOWN;
503 		link.link_autoneg = ETH_LINK_FIXED;
504 	}
505 
506 	return rte_eth_linkstatus_set(dev, &link);
507 }
508 
509 /*
510  * It executes link_update after knowing an interrupt is present.
511  */
512 static void
513 eth_igc_interrupt_action(struct rte_eth_dev *dev)
514 {
515 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
516 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
517 	struct rte_eth_link link;
518 	int ret;
519 
520 	if (intr->flags & IGC_FLAG_NEED_LINK_UPDATE) {
521 		intr->flags &= ~IGC_FLAG_NEED_LINK_UPDATE;
522 
523 		/* set get_link_status to check register later */
524 		ret = eth_igc_link_update(dev, 0);
525 
526 		/* check if link has changed */
527 		if (ret < 0)
528 			return;
529 
530 		rte_eth_linkstatus_get(dev, &link);
531 		if (link.link_status)
532 			PMD_DRV_LOG(INFO,
533 				" Port %d: Link Up - speed %u Mbps - %s",
534 				dev->data->port_id,
535 				(unsigned int)link.link_speed,
536 				link.link_duplex == ETH_LINK_FULL_DUPLEX ?
537 				"full-duplex" : "half-duplex");
538 		else
539 			PMD_DRV_LOG(INFO, " Port %d: Link Down",
540 				dev->data->port_id);
541 
542 		PMD_DRV_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
543 				pci_dev->addr.domain,
544 				pci_dev->addr.bus,
545 				pci_dev->addr.devid,
546 				pci_dev->addr.function);
547 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC,
548 				NULL);
549 	}
550 }
551 
552 /*
553  * Interrupt handler which shall be registered at first.
554  *
555  * @handle
556  *  Pointer to interrupt handle.
557  * @param
558  *  The address of parameter (struct rte_eth_dev *) registered before.
559  */
560 static void
561 eth_igc_interrupt_handler(void *param)
562 {
563 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
564 
565 	eth_igc_interrupt_get_status(dev);
566 	eth_igc_interrupt_action(dev);
567 }
568 
569 static void igc_read_queue_stats_register(struct rte_eth_dev *dev);
570 
571 /*
572  * Update the queue status every IGC_ALARM_INTERVAL time.
573  * @param
574  *  The address of parameter (struct rte_eth_dev *) registered before.
575  */
576 static void
577 igc_update_queue_stats_handler(void *param)
578 {
579 	struct rte_eth_dev *dev = param;
580 	igc_read_queue_stats_register(dev);
581 	rte_eal_alarm_set(IGC_ALARM_INTERVAL,
582 			igc_update_queue_stats_handler, dev);
583 }
584 
585 /*
586  * rx,tx enable/disable
587  */
588 static void
589 eth_igc_rxtx_control(struct rte_eth_dev *dev, bool enable)
590 {
591 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
592 	uint32_t tctl, rctl;
593 
594 	tctl = IGC_READ_REG(hw, IGC_TCTL);
595 	rctl = IGC_READ_REG(hw, IGC_RCTL);
596 
597 	if (enable) {
598 		/* enable Tx/Rx */
599 		tctl |= IGC_TCTL_EN;
600 		rctl |= IGC_RCTL_EN;
601 	} else {
602 		/* disable Tx/Rx */
603 		tctl &= ~IGC_TCTL_EN;
604 		rctl &= ~IGC_RCTL_EN;
605 	}
606 	IGC_WRITE_REG(hw, IGC_TCTL, tctl);
607 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
608 	IGC_WRITE_FLUSH(hw);
609 }
610 
611 /*
612  *  This routine disables all traffic on the adapter by issuing a
613  *  global reset on the MAC.
614  */
615 static void
616 eth_igc_stop(struct rte_eth_dev *dev)
617 {
618 	struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
619 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
620 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
621 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
622 	struct rte_eth_link link;
623 
624 	adapter->stopped = 1;
625 
626 	/* disable receive and transmit */
627 	eth_igc_rxtx_control(dev, false);
628 
629 	/* disable all MSI-X interrupts */
630 	IGC_WRITE_REG(hw, IGC_EIMC, 0x1f);
631 	IGC_WRITE_FLUSH(hw);
632 
633 	/* clear all MSI-X interrupts */
634 	IGC_WRITE_REG(hw, IGC_EICR, 0x1f);
635 
636 	igc_intr_other_disable(dev);
637 
638 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
639 
640 	/* disable intr eventfd mapping */
641 	rte_intr_disable(intr_handle);
642 
643 	igc_reset_hw(hw);
644 
645 	/* disable all wake up */
646 	IGC_WRITE_REG(hw, IGC_WUC, 0);
647 
648 	/* disable checking EEE operation in MAC loopback mode */
649 	igc_read_reg_check_clear_bits(hw, IGC_EEER, IGC_EEER_EEE_FRC_AN);
650 
651 	/* Set bit for Go Link disconnect */
652 	igc_read_reg_check_set_bits(hw, IGC_82580_PHY_POWER_MGMT,
653 			IGC_82580_PM_GO_LINKD);
654 
655 	/* Power down the phy. Needed to make the link go Down */
656 	eth_igc_set_link_down(dev);
657 
658 	igc_dev_clear_queues(dev);
659 
660 	/* clear the recorded link status */
661 	memset(&link, 0, sizeof(link));
662 	rte_eth_linkstatus_set(dev, &link);
663 
664 	if (!rte_intr_allow_others(intr_handle))
665 		/* resume to the default handler */
666 		rte_intr_callback_register(intr_handle,
667 					   eth_igc_interrupt_handler,
668 					   (void *)dev);
669 
670 	/* Clean datapath event and queue/vec mapping */
671 	rte_intr_efd_disable(intr_handle);
672 	if (intr_handle->intr_vec != NULL) {
673 		rte_free(intr_handle->intr_vec);
674 		intr_handle->intr_vec = NULL;
675 	}
676 }
677 
678 /*
679  * write interrupt vector allocation register
680  * @hw
681  *  board private structure
682  * @queue_index
683  *  queue index, valid 0,1,2,3
684  * @tx
685  *  tx:1, rx:0
686  * @msix_vector
687  *  msix-vector, valid 0,1,2,3,4
688  */
689 static void
690 igc_write_ivar(struct igc_hw *hw, uint8_t queue_index,
691 		bool tx, uint8_t msix_vector)
692 {
693 	uint8_t offset = 0;
694 	uint8_t reg_index = queue_index >> 1;
695 	uint32_t val;
696 
697 	/*
698 	 * IVAR(0)
699 	 * bit31...24	bit23...16	bit15...8	bit7...0
700 	 * TX1		RX1		TX0		RX0
701 	 *
702 	 * IVAR(1)
703 	 * bit31...24	bit23...16	bit15...8	bit7...0
704 	 * TX3		RX3		TX2		RX2
705 	 */
706 
707 	if (tx)
708 		offset = 8;
709 
710 	if (queue_index & 1)
711 		offset += 16;
712 
713 	val = IGC_READ_REG_ARRAY(hw, IGC_IVAR0, reg_index);
714 
715 	/* clear bits */
716 	val &= ~((uint32_t)0xFF << offset);
717 
718 	/* write vector and valid bit */
719 	val |= (uint32_t)(msix_vector | IGC_IVAR_VALID) << offset;
720 
721 	IGC_WRITE_REG_ARRAY(hw, IGC_IVAR0, reg_index, val);
722 }
723 
724 /* Sets up the hardware to generate MSI-X interrupts properly
725  * @hw
726  *  board private structure
727  */
728 static void
729 igc_configure_msix_intr(struct rte_eth_dev *dev)
730 {
731 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
732 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
733 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
734 
735 	uint32_t intr_mask;
736 	uint32_t vec = IGC_MISC_VEC_ID;
737 	uint32_t base = IGC_MISC_VEC_ID;
738 	uint32_t misc_shift = 0;
739 	int i;
740 
741 	/* won't configure msix register if no mapping is done
742 	 * between intr vector and event fd
743 	 */
744 	if (!rte_intr_dp_is_en(intr_handle))
745 		return;
746 
747 	if (rte_intr_allow_others(intr_handle)) {
748 		base = IGC_RX_VEC_START;
749 		vec = base;
750 		misc_shift = 1;
751 	}
752 
753 	/* turn on MSI-X capability first */
754 	IGC_WRITE_REG(hw, IGC_GPIE, IGC_GPIE_MSIX_MODE |
755 				IGC_GPIE_PBA | IGC_GPIE_EIAME |
756 				IGC_GPIE_NSICR);
757 	intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
758 		misc_shift;
759 
760 	if (dev->data->dev_conf.intr_conf.lsc)
761 		intr_mask |= (1u << IGC_MSIX_OTHER_INTR_VEC);
762 
763 	/* enable msix auto-clear */
764 	igc_read_reg_check_set_bits(hw, IGC_EIAC, intr_mask);
765 
766 	/* set other cause interrupt vector */
767 	igc_read_reg_check_set_bits(hw, IGC_IVAR_MISC,
768 		(uint32_t)(IGC_MSIX_OTHER_INTR_VEC | IGC_IVAR_VALID) << 8);
769 
770 	/* enable auto-mask */
771 	igc_read_reg_check_set_bits(hw, IGC_EIAM, intr_mask);
772 
773 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
774 		igc_write_ivar(hw, i, 0, vec);
775 		intr_handle->intr_vec[i] = vec;
776 		if (vec < base + intr_handle->nb_efd - 1)
777 			vec++;
778 	}
779 
780 	IGC_WRITE_FLUSH(hw);
781 }
782 
783 /**
784  * It enables the interrupt mask and then enable the interrupt.
785  *
786  * @dev
787  *  Pointer to struct rte_eth_dev.
788  * @on
789  *  Enable or Disable
790  */
791 static void
792 igc_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
793 {
794 	struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
795 
796 	if (on)
797 		intr->mask |= IGC_ICR_LSC;
798 	else
799 		intr->mask &= ~IGC_ICR_LSC;
800 }
801 
802 /*
803  * It enables the interrupt.
804  * It will be called once only during nic initialized.
805  */
806 static void
807 igc_rxq_interrupt_setup(struct rte_eth_dev *dev)
808 {
809 	uint32_t mask;
810 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
811 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
812 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
813 	int misc_shift = rte_intr_allow_others(intr_handle) ? 1 : 0;
814 
815 	/* won't configure msix register if no mapping is done
816 	 * between intr vector and event fd
817 	 */
818 	if (!rte_intr_dp_is_en(intr_handle))
819 		return;
820 
821 	mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) << misc_shift;
822 	IGC_WRITE_REG(hw, IGC_EIMS, mask);
823 }
824 
825 /*
826  *  Get hardware rx-buffer size.
827  */
828 static inline int
829 igc_get_rx_buffer_size(struct igc_hw *hw)
830 {
831 	return (IGC_READ_REG(hw, IGC_RXPBS) & 0x3f) << 10;
832 }
833 
834 /*
835  * igc_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
836  * For ASF and Pass Through versions of f/w this means
837  * that the driver is loaded.
838  */
839 static void
840 igc_hw_control_acquire(struct igc_hw *hw)
841 {
842 	uint32_t ctrl_ext;
843 
844 	/* Let firmware know the driver has taken over */
845 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
846 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext | IGC_CTRL_EXT_DRV_LOAD);
847 }
848 
849 /*
850  * igc_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
851  * For ASF and Pass Through versions of f/w this means that the
852  * driver is no longer loaded.
853  */
854 static void
855 igc_hw_control_release(struct igc_hw *hw)
856 {
857 	uint32_t ctrl_ext;
858 
859 	/* Let firmware taken over control of h/w */
860 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
861 	IGC_WRITE_REG(hw, IGC_CTRL_EXT,
862 			ctrl_ext & ~IGC_CTRL_EXT_DRV_LOAD);
863 }
864 
865 static int
866 igc_hardware_init(struct igc_hw *hw)
867 {
868 	uint32_t rx_buf_size;
869 	int diag;
870 
871 	/* Let the firmware know the OS is in control */
872 	igc_hw_control_acquire(hw);
873 
874 	/* Issue a global reset */
875 	igc_reset_hw(hw);
876 
877 	/* disable all wake up */
878 	IGC_WRITE_REG(hw, IGC_WUC, 0);
879 
880 	/*
881 	 * Hardware flow control
882 	 * - High water mark should allow for at least two standard size (1518)
883 	 *   frames to be received after sending an XOFF.
884 	 * - Low water mark works best when it is very near the high water mark.
885 	 *   This allows the receiver to restart by sending XON when it has
886 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
887 	 *   restart after one full frame is pulled from the buffer. There
888 	 *   could be several smaller frames in the buffer and if so they will
889 	 *   not trigger the XON until their total number reduces the buffer
890 	 *   by 1500.
891 	 */
892 	rx_buf_size = igc_get_rx_buffer_size(hw);
893 	hw->fc.high_water = rx_buf_size - (RTE_ETHER_MAX_LEN * 2);
894 	hw->fc.low_water = hw->fc.high_water - 1500;
895 	hw->fc.pause_time = IGC_FC_PAUSE_TIME;
896 	hw->fc.send_xon = 1;
897 	hw->fc.requested_mode = igc_fc_full;
898 
899 	diag = igc_init_hw(hw);
900 	if (diag < 0)
901 		return diag;
902 
903 	igc_get_phy_info(hw);
904 	igc_check_for_link(hw);
905 
906 	return 0;
907 }
908 
909 static int
910 eth_igc_start(struct rte_eth_dev *dev)
911 {
912 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
913 	struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
914 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
915 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
916 	uint32_t *speeds;
917 	int ret;
918 
919 	PMD_INIT_FUNC_TRACE();
920 
921 	/* disable all MSI-X interrupts */
922 	IGC_WRITE_REG(hw, IGC_EIMC, 0x1f);
923 	IGC_WRITE_FLUSH(hw);
924 
925 	/* clear all MSI-X interrupts */
926 	IGC_WRITE_REG(hw, IGC_EICR, 0x1f);
927 
928 	/* disable uio/vfio intr/eventfd mapping */
929 	if (!adapter->stopped)
930 		rte_intr_disable(intr_handle);
931 
932 	/* Power up the phy. Needed to make the link go Up */
933 	eth_igc_set_link_up(dev);
934 
935 	/* Put the address into the Receive Address Array */
936 	igc_rar_set(hw, hw->mac.addr, 0);
937 
938 	/* Initialize the hardware */
939 	if (igc_hardware_init(hw)) {
940 		PMD_DRV_LOG(ERR, "Unable to initialize the hardware");
941 		return -EIO;
942 	}
943 	adapter->stopped = 0;
944 
945 	/* check and configure queue intr-vector mapping */
946 	if (rte_intr_cap_multiple(intr_handle) &&
947 		dev->data->dev_conf.intr_conf.rxq) {
948 		uint32_t intr_vector = dev->data->nb_rx_queues;
949 		if (rte_intr_efd_enable(intr_handle, intr_vector))
950 			return -1;
951 	}
952 
953 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
954 		intr_handle->intr_vec = rte_zmalloc("intr_vec",
955 			dev->data->nb_rx_queues * sizeof(int), 0);
956 		if (intr_handle->intr_vec == NULL) {
957 			PMD_DRV_LOG(ERR,
958 				"Failed to allocate %d rx_queues intr_vec",
959 				dev->data->nb_rx_queues);
960 			return -ENOMEM;
961 		}
962 	}
963 
964 	/* configure msix for rx interrupt */
965 	igc_configure_msix_intr(dev);
966 
967 	igc_tx_init(dev);
968 
969 	/* This can fail when allocating mbufs for descriptor rings */
970 	ret = igc_rx_init(dev);
971 	if (ret) {
972 		PMD_DRV_LOG(ERR, "Unable to initialize RX hardware");
973 		igc_dev_clear_queues(dev);
974 		return ret;
975 	}
976 
977 	igc_clear_hw_cntrs_base_generic(hw);
978 
979 	/* VLAN Offload Settings */
980 	eth_igc_vlan_offload_set(dev,
981 		ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
982 		ETH_VLAN_EXTEND_MASK);
983 
984 	/* Setup link speed and duplex */
985 	speeds = &dev->data->dev_conf.link_speeds;
986 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
987 		hw->phy.autoneg_advertised = IGC_ALL_SPEED_DUPLEX_2500;
988 		hw->mac.autoneg = 1;
989 	} else {
990 		int num_speeds = 0;
991 		bool autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
992 
993 		/* Reset */
994 		hw->phy.autoneg_advertised = 0;
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 				ETH_LINK_SPEED_FIXED)) {
1000 			num_speeds = -1;
1001 			goto error_invalid_config;
1002 		}
1003 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
1004 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1005 			num_speeds++;
1006 		}
1007 		if (*speeds & ETH_LINK_SPEED_10M) {
1008 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1009 			num_speeds++;
1010 		}
1011 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
1012 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1013 			num_speeds++;
1014 		}
1015 		if (*speeds & ETH_LINK_SPEED_100M) {
1016 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1017 			num_speeds++;
1018 		}
1019 		if (*speeds & ETH_LINK_SPEED_1G) {
1020 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1021 			num_speeds++;
1022 		}
1023 		if (*speeds & ETH_LINK_SPEED_2_5G) {
1024 			hw->phy.autoneg_advertised |= ADVERTISE_2500_FULL;
1025 			num_speeds++;
1026 		}
1027 		if (num_speeds == 0 || (!autoneg && num_speeds > 1))
1028 			goto error_invalid_config;
1029 
1030 		/* Set/reset the mac.autoneg based on the link speed,
1031 		 * fixed or not
1032 		 */
1033 		if (!autoneg) {
1034 			hw->mac.autoneg = 0;
1035 			hw->mac.forced_speed_duplex =
1036 					hw->phy.autoneg_advertised;
1037 		} else {
1038 			hw->mac.autoneg = 1;
1039 		}
1040 	}
1041 
1042 	igc_setup_link(hw);
1043 
1044 	if (rte_intr_allow_others(intr_handle)) {
1045 		/* check if lsc interrupt is enabled */
1046 		if (dev->data->dev_conf.intr_conf.lsc)
1047 			igc_lsc_interrupt_setup(dev, 1);
1048 		else
1049 			igc_lsc_interrupt_setup(dev, 0);
1050 	} else {
1051 		rte_intr_callback_unregister(intr_handle,
1052 					     eth_igc_interrupt_handler,
1053 					     (void *)dev);
1054 		if (dev->data->dev_conf.intr_conf.lsc)
1055 			PMD_DRV_LOG(INFO,
1056 				"LSC won't enable because of no intr multiplex");
1057 	}
1058 
1059 	/* enable uio/vfio intr/eventfd mapping */
1060 	rte_intr_enable(intr_handle);
1061 
1062 	rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1063 			igc_update_queue_stats_handler, dev);
1064 
1065 	/* check if rxq interrupt is enabled */
1066 	if (dev->data->dev_conf.intr_conf.rxq &&
1067 			rte_intr_dp_is_en(intr_handle))
1068 		igc_rxq_interrupt_setup(dev);
1069 
1070 	/* resume enabled intr since hw reset */
1071 	igc_intr_other_enable(dev);
1072 
1073 	eth_igc_rxtx_control(dev, true);
1074 	eth_igc_link_update(dev, 0);
1075 
1076 	/* configure MAC-loopback mode */
1077 	if (dev->data->dev_conf.lpbk_mode == 1) {
1078 		uint32_t reg_val;
1079 
1080 		reg_val = IGC_READ_REG(hw, IGC_CTRL);
1081 		reg_val &= ~IGC_CTRL_SPEED_MASK;
1082 		reg_val |= IGC_CTRL_SLU | IGC_CTRL_FRCSPD |
1083 			IGC_CTRL_FRCDPX | IGC_CTRL_FD | IGC_CTRL_SPEED_2500;
1084 		IGC_WRITE_REG(hw, IGC_CTRL, reg_val);
1085 
1086 		igc_read_reg_check_set_bits(hw, IGC_EEER, IGC_EEER_EEE_FRC_AN);
1087 	}
1088 
1089 	return 0;
1090 
1091 error_invalid_config:
1092 	PMD_DRV_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
1093 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
1094 	igc_dev_clear_queues(dev);
1095 	return -EINVAL;
1096 }
1097 
1098 static int
1099 igc_reset_swfw_lock(struct igc_hw *hw)
1100 {
1101 	int ret_val;
1102 
1103 	/*
1104 	 * Do mac ops initialization manually here, since we will need
1105 	 * some function pointers set by this call.
1106 	 */
1107 	ret_val = igc_init_mac_params(hw);
1108 	if (ret_val)
1109 		return ret_val;
1110 
1111 	/*
1112 	 * SMBI lock should not fail in this early stage. If this is the case,
1113 	 * it is due to an improper exit of the application.
1114 	 * So force the release of the faulty lock.
1115 	 */
1116 	if (igc_get_hw_semaphore_generic(hw) < 0)
1117 		PMD_DRV_LOG(DEBUG, "SMBI lock released");
1118 
1119 	igc_put_hw_semaphore_generic(hw);
1120 
1121 	if (hw->mac.ops.acquire_swfw_sync != NULL) {
1122 		uint16_t mask;
1123 
1124 		/*
1125 		 * Phy lock should not fail in this early stage.
1126 		 * If this is the case, it is due to an improper exit of the
1127 		 * application. So force the release of the faulty lock.
1128 		 */
1129 		mask = IGC_SWFW_PHY0_SM;
1130 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
1131 			PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released",
1132 				    hw->bus.func);
1133 		}
1134 		hw->mac.ops.release_swfw_sync(hw, mask);
1135 
1136 		/*
1137 		 * This one is more tricky since it is common to all ports; but
1138 		 * swfw_sync retries last long enough (1s) to be almost sure
1139 		 * that if lock can not be taken it is due to an improper lock
1140 		 * of the semaphore.
1141 		 */
1142 		mask = IGC_SWFW_EEP_SM;
1143 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0)
1144 			PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1145 
1146 		hw->mac.ops.release_swfw_sync(hw, mask);
1147 	}
1148 
1149 	return IGC_SUCCESS;
1150 }
1151 
1152 /*
1153  * free all rx/tx queues.
1154  */
1155 static void
1156 igc_dev_free_queues(struct rte_eth_dev *dev)
1157 {
1158 	uint16_t i;
1159 
1160 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
1161 		eth_igc_rx_queue_release(dev->data->rx_queues[i]);
1162 		dev->data->rx_queues[i] = NULL;
1163 	}
1164 	dev->data->nb_rx_queues = 0;
1165 
1166 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
1167 		eth_igc_tx_queue_release(dev->data->tx_queues[i]);
1168 		dev->data->tx_queues[i] = NULL;
1169 	}
1170 	dev->data->nb_tx_queues = 0;
1171 }
1172 
1173 static void
1174 eth_igc_close(struct rte_eth_dev *dev)
1175 {
1176 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1177 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1178 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1179 	struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
1180 	int retry = 0;
1181 
1182 	PMD_INIT_FUNC_TRACE();
1183 
1184 	if (!adapter->stopped)
1185 		eth_igc_stop(dev);
1186 
1187 	igc_flow_flush(dev, NULL);
1188 	igc_clear_all_filter(dev);
1189 
1190 	igc_intr_other_disable(dev);
1191 	do {
1192 		int ret = rte_intr_callback_unregister(intr_handle,
1193 				eth_igc_interrupt_handler, dev);
1194 		if (ret >= 0 || ret == -ENOENT || ret == -EINVAL)
1195 			break;
1196 
1197 		PMD_DRV_LOG(ERR, "intr callback unregister failed: %d", ret);
1198 		DELAY(200 * 1000); /* delay 200ms */
1199 	} while (retry++ < 5);
1200 
1201 	igc_phy_hw_reset(hw);
1202 	igc_hw_control_release(hw);
1203 	igc_dev_free_queues(dev);
1204 
1205 	/* Reset any pending lock */
1206 	igc_reset_swfw_lock(hw);
1207 }
1208 
1209 static void
1210 igc_identify_hardware(struct rte_eth_dev *dev, struct rte_pci_device *pci_dev)
1211 {
1212 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1213 
1214 	hw->vendor_id = pci_dev->id.vendor_id;
1215 	hw->device_id = pci_dev->id.device_id;
1216 	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
1217 	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
1218 }
1219 
1220 static int
1221 eth_igc_dev_init(struct rte_eth_dev *dev)
1222 {
1223 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1224 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
1225 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1226 	int i, error = 0;
1227 
1228 	PMD_INIT_FUNC_TRACE();
1229 	dev->dev_ops = &eth_igc_ops;
1230 
1231 	/*
1232 	 * for secondary processes, we don't initialize any further as primary
1233 	 * has already done this work. Only check we don't need a different
1234 	 * RX function.
1235 	 */
1236 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1237 		return 0;
1238 
1239 	rte_eth_copy_pci_info(dev, pci_dev);
1240 
1241 	hw->back = pci_dev;
1242 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1243 
1244 	igc_identify_hardware(dev, pci_dev);
1245 	if (igc_setup_init_funcs(hw, false) != IGC_SUCCESS) {
1246 		error = -EIO;
1247 		goto err_late;
1248 	}
1249 
1250 	igc_get_bus_info(hw);
1251 
1252 	/* Reset any pending lock */
1253 	if (igc_reset_swfw_lock(hw) != IGC_SUCCESS) {
1254 		error = -EIO;
1255 		goto err_late;
1256 	}
1257 
1258 	/* Finish initialization */
1259 	if (igc_setup_init_funcs(hw, true) != IGC_SUCCESS) {
1260 		error = -EIO;
1261 		goto err_late;
1262 	}
1263 
1264 	hw->mac.autoneg = 1;
1265 	hw->phy.autoneg_wait_to_complete = 0;
1266 	hw->phy.autoneg_advertised = IGC_ALL_SPEED_DUPLEX_2500;
1267 
1268 	/* Copper options */
1269 	if (hw->phy.media_type == igc_media_type_copper) {
1270 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
1271 		hw->phy.disable_polarity_correction = 0;
1272 		hw->phy.ms_type = igc_ms_hw_default;
1273 	}
1274 
1275 	/*
1276 	 * Start from a known state, this is important in reading the nvm
1277 	 * and mac from that.
1278 	 */
1279 	igc_reset_hw(hw);
1280 
1281 	/* Make sure we have a good EEPROM before we read from it */
1282 	if (igc_validate_nvm_checksum(hw) < 0) {
1283 		/*
1284 		 * Some PCI-E parts fail the first check due to
1285 		 * the link being in sleep state, call it again,
1286 		 * if it fails a second time its a real issue.
1287 		 */
1288 		if (igc_validate_nvm_checksum(hw) < 0) {
1289 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
1290 			error = -EIO;
1291 			goto err_late;
1292 		}
1293 	}
1294 
1295 	/* Read the permanent MAC address out of the EEPROM */
1296 	if (igc_read_mac_addr(hw) != 0) {
1297 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
1298 		error = -EIO;
1299 		goto err_late;
1300 	}
1301 
1302 	/* Allocate memory for storing MAC addresses */
1303 	dev->data->mac_addrs = rte_zmalloc("igc",
1304 		RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count, 0);
1305 	if (dev->data->mac_addrs == NULL) {
1306 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes for storing MAC",
1307 				RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
1308 		error = -ENOMEM;
1309 		goto err_late;
1310 	}
1311 
1312 	/* Copy the permanent MAC address */
1313 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
1314 			&dev->data->mac_addrs[0]);
1315 
1316 	/* Now initialize the hardware */
1317 	if (igc_hardware_init(hw) != 0) {
1318 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
1319 		rte_free(dev->data->mac_addrs);
1320 		dev->data->mac_addrs = NULL;
1321 		error = -ENODEV;
1322 		goto err_late;
1323 	}
1324 
1325 	/* Pass the information to the rte_eth_dev_close() that it should also
1326 	 * release the private port resources.
1327 	 */
1328 	dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
1329 
1330 	hw->mac.get_link_status = 1;
1331 	igc->stopped = 0;
1332 
1333 	/* Indicate SOL/IDER usage */
1334 	if (igc_check_reset_block(hw) < 0)
1335 		PMD_INIT_LOG(ERR,
1336 			"PHY reset is blocked due to SOL/IDER session.");
1337 
1338 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
1339 			dev->data->port_id, pci_dev->id.vendor_id,
1340 			pci_dev->id.device_id);
1341 
1342 	rte_intr_callback_register(&pci_dev->intr_handle,
1343 			eth_igc_interrupt_handler, (void *)dev);
1344 
1345 	/* enable uio/vfio intr/eventfd mapping */
1346 	rte_intr_enable(&pci_dev->intr_handle);
1347 
1348 	/* enable support intr */
1349 	igc_intr_other_enable(dev);
1350 
1351 	/* initiate queue status */
1352 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1353 		igc->txq_stats_map[i] = -1;
1354 		igc->rxq_stats_map[i] = -1;
1355 	}
1356 
1357 	igc_flow_init(dev);
1358 	igc_clear_all_filter(dev);
1359 	return 0;
1360 
1361 err_late:
1362 	igc_hw_control_release(hw);
1363 	return error;
1364 }
1365 
1366 static int
1367 eth_igc_dev_uninit(__rte_unused struct rte_eth_dev *eth_dev)
1368 {
1369 	PMD_INIT_FUNC_TRACE();
1370 
1371 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1372 		return 0;
1373 
1374 	eth_igc_close(eth_dev);
1375 	return 0;
1376 }
1377 
1378 static int
1379 eth_igc_reset(struct rte_eth_dev *dev)
1380 {
1381 	int ret;
1382 
1383 	PMD_INIT_FUNC_TRACE();
1384 
1385 	ret = eth_igc_dev_uninit(dev);
1386 	if (ret)
1387 		return ret;
1388 
1389 	return eth_igc_dev_init(dev);
1390 }
1391 
1392 static int
1393 eth_igc_promiscuous_enable(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 | IGC_RCTL_MPE);
1400 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1401 	return 0;
1402 }
1403 
1404 static int
1405 eth_igc_promiscuous_disable(struct rte_eth_dev *dev)
1406 {
1407 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1408 	uint32_t rctl;
1409 
1410 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1411 	rctl &= (~IGC_RCTL_UPE);
1412 	if (dev->data->all_multicast == 1)
1413 		rctl |= IGC_RCTL_MPE;
1414 	else
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_enable(struct rte_eth_dev *dev)
1422 {
1423 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1424 	uint32_t rctl;
1425 
1426 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1427 	rctl |= IGC_RCTL_MPE;
1428 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1429 	return 0;
1430 }
1431 
1432 static int
1433 eth_igc_allmulticast_disable(struct rte_eth_dev *dev)
1434 {
1435 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1436 	uint32_t rctl;
1437 
1438 	if (dev->data->promiscuous == 1)
1439 		return 0;	/* must remain in all_multicast mode */
1440 
1441 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1442 	rctl &= (~IGC_RCTL_MPE);
1443 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1444 	return 0;
1445 }
1446 
1447 static int
1448 eth_igc_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
1449 		       size_t fw_size)
1450 {
1451 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1452 	struct igc_fw_version fw;
1453 	int ret;
1454 
1455 	igc_get_fw_version(hw, &fw);
1456 
1457 	/* if option rom is valid, display its version too */
1458 	if (fw.or_valid) {
1459 		ret = snprintf(fw_version, fw_size,
1460 			 "%d.%d, 0x%08x, %d.%d.%d",
1461 			 fw.eep_major, fw.eep_minor, fw.etrack_id,
1462 			 fw.or_major, fw.or_build, fw.or_patch);
1463 	/* no option rom */
1464 	} else {
1465 		if (fw.etrack_id != 0X0000) {
1466 			ret = snprintf(fw_version, fw_size,
1467 				 "%d.%d, 0x%08x",
1468 				 fw.eep_major, fw.eep_minor,
1469 				 fw.etrack_id);
1470 		} else {
1471 			ret = snprintf(fw_version, fw_size,
1472 				 "%d.%d.%d",
1473 				 fw.eep_major, fw.eep_minor,
1474 				 fw.eep_build);
1475 		}
1476 	}
1477 
1478 	ret += 1; /* add the size of '\0' */
1479 	if (fw_size < (u32)ret)
1480 		return ret;
1481 	else
1482 		return 0;
1483 }
1484 
1485 static int
1486 eth_igc_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1487 {
1488 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1489 
1490 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1491 	dev_info->max_rx_pktlen = MAX_RX_JUMBO_FRAME_SIZE;
1492 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1493 	dev_info->rx_offload_capa = IGC_RX_OFFLOAD_ALL;
1494 	dev_info->tx_offload_capa = IGC_TX_OFFLOAD_ALL;
1495 	dev_info->rx_queue_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP;
1496 
1497 	dev_info->max_rx_queues = IGC_QUEUE_PAIRS_NUM;
1498 	dev_info->max_tx_queues = IGC_QUEUE_PAIRS_NUM;
1499 	dev_info->max_vmdq_pools = 0;
1500 
1501 	dev_info->hash_key_size = IGC_HKEY_MAX_INDEX * sizeof(uint32_t);
1502 	dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
1503 	dev_info->flow_type_rss_offloads = IGC_RSS_OFFLOAD_ALL;
1504 
1505 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
1506 		.rx_thresh = {
1507 			.pthresh = IGC_DEFAULT_RX_PTHRESH,
1508 			.hthresh = IGC_DEFAULT_RX_HTHRESH,
1509 			.wthresh = IGC_DEFAULT_RX_WTHRESH,
1510 		},
1511 		.rx_free_thresh = IGC_DEFAULT_RX_FREE_THRESH,
1512 		.rx_drop_en = 0,
1513 		.offloads = 0,
1514 	};
1515 
1516 	dev_info->default_txconf = (struct rte_eth_txconf) {
1517 		.tx_thresh = {
1518 			.pthresh = IGC_DEFAULT_TX_PTHRESH,
1519 			.hthresh = IGC_DEFAULT_TX_HTHRESH,
1520 			.wthresh = IGC_DEFAULT_TX_WTHRESH,
1521 		},
1522 		.offloads = 0,
1523 	};
1524 
1525 	dev_info->rx_desc_lim = rx_desc_lim;
1526 	dev_info->tx_desc_lim = tx_desc_lim;
1527 
1528 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1529 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1530 			ETH_LINK_SPEED_1G | ETH_LINK_SPEED_2_5G;
1531 
1532 	dev_info->max_mtu = dev_info->max_rx_pktlen - IGC_ETH_OVERHEAD;
1533 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
1534 	return 0;
1535 }
1536 
1537 static int
1538 eth_igc_led_on(struct rte_eth_dev *dev)
1539 {
1540 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1541 
1542 	return igc_led_on(hw) == IGC_SUCCESS ? 0 : -ENOTSUP;
1543 }
1544 
1545 static int
1546 eth_igc_led_off(struct rte_eth_dev *dev)
1547 {
1548 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1549 
1550 	return igc_led_off(hw) == IGC_SUCCESS ? 0 : -ENOTSUP;
1551 }
1552 
1553 static const uint32_t *
1554 eth_igc_supported_ptypes_get(__rte_unused struct rte_eth_dev *dev)
1555 {
1556 	static const uint32_t ptypes[] = {
1557 		/* refers to rx_desc_pkt_info_to_pkt_type() */
1558 		RTE_PTYPE_L2_ETHER,
1559 		RTE_PTYPE_L3_IPV4,
1560 		RTE_PTYPE_L3_IPV4_EXT,
1561 		RTE_PTYPE_L3_IPV6,
1562 		RTE_PTYPE_L3_IPV6_EXT,
1563 		RTE_PTYPE_L4_TCP,
1564 		RTE_PTYPE_L4_UDP,
1565 		RTE_PTYPE_L4_SCTP,
1566 		RTE_PTYPE_TUNNEL_IP,
1567 		RTE_PTYPE_INNER_L3_IPV6,
1568 		RTE_PTYPE_INNER_L3_IPV6_EXT,
1569 		RTE_PTYPE_INNER_L4_TCP,
1570 		RTE_PTYPE_INNER_L4_UDP,
1571 		RTE_PTYPE_UNKNOWN
1572 	};
1573 
1574 	return ptypes;
1575 }
1576 
1577 static int
1578 eth_igc_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1579 {
1580 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1581 	uint32_t frame_size = mtu + IGC_ETH_OVERHEAD;
1582 	uint32_t rctl;
1583 
1584 	/* if extend vlan has been enabled */
1585 	if (IGC_READ_REG(hw, IGC_CTRL_EXT) & IGC_CTRL_EXT_EXT_VLAN)
1586 		frame_size += VLAN_TAG_SIZE;
1587 
1588 	/* check that mtu is within the allowed range */
1589 	if (mtu < RTE_ETHER_MIN_MTU ||
1590 		frame_size > MAX_RX_JUMBO_FRAME_SIZE)
1591 		return -EINVAL;
1592 
1593 	/*
1594 	 * refuse mtu that requires the support of scattered packets when
1595 	 * this feature has not been enabled before.
1596 	 */
1597 	if (!dev->data->scattered_rx &&
1598 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
1599 		return -EINVAL;
1600 
1601 	rctl = IGC_READ_REG(hw, IGC_RCTL);
1602 
1603 	/* switch to jumbo mode if needed */
1604 	if (mtu > RTE_ETHER_MTU) {
1605 		dev->data->dev_conf.rxmode.offloads |=
1606 			DEV_RX_OFFLOAD_JUMBO_FRAME;
1607 		rctl |= IGC_RCTL_LPE;
1608 	} else {
1609 		dev->data->dev_conf.rxmode.offloads &=
1610 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
1611 		rctl &= ~IGC_RCTL_LPE;
1612 	}
1613 	IGC_WRITE_REG(hw, IGC_RCTL, rctl);
1614 
1615 	/* update max frame size */
1616 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1617 
1618 	IGC_WRITE_REG(hw, IGC_RLPML,
1619 			dev->data->dev_conf.rxmode.max_rx_pkt_len);
1620 
1621 	return 0;
1622 }
1623 
1624 static int
1625 eth_igc_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
1626 		uint32_t index, uint32_t pool)
1627 {
1628 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1629 
1630 	igc_rar_set(hw, mac_addr->addr_bytes, index);
1631 	RTE_SET_USED(pool);
1632 	return 0;
1633 }
1634 
1635 static void
1636 eth_igc_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1637 {
1638 	uint8_t addr[RTE_ETHER_ADDR_LEN];
1639 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1640 
1641 	memset(addr, 0, sizeof(addr));
1642 	igc_rar_set(hw, addr, index);
1643 }
1644 
1645 static int
1646 eth_igc_default_mac_addr_set(struct rte_eth_dev *dev,
1647 			struct rte_ether_addr *addr)
1648 {
1649 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1650 	igc_rar_set(hw, addr->addr_bytes, 0);
1651 	return 0;
1652 }
1653 
1654 static int
1655 eth_igc_set_mc_addr_list(struct rte_eth_dev *dev,
1656 			 struct rte_ether_addr *mc_addr_set,
1657 			 uint32_t nb_mc_addr)
1658 {
1659 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1660 	igc_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1661 	return 0;
1662 }
1663 
1664 /*
1665  * Read hardware registers
1666  */
1667 static void
1668 igc_read_stats_registers(struct igc_hw *hw, struct igc_hw_stats *stats)
1669 {
1670 	int pause_frames;
1671 
1672 	uint64_t old_gprc  = stats->gprc;
1673 	uint64_t old_gptc  = stats->gptc;
1674 	uint64_t old_tpr   = stats->tpr;
1675 	uint64_t old_tpt   = stats->tpt;
1676 	uint64_t old_rpthc = stats->rpthc;
1677 	uint64_t old_hgptc = stats->hgptc;
1678 
1679 	stats->crcerrs += IGC_READ_REG(hw, IGC_CRCERRS);
1680 	stats->algnerrc += IGC_READ_REG(hw, IGC_ALGNERRC);
1681 	stats->rxerrc += IGC_READ_REG(hw, IGC_RXERRC);
1682 	stats->mpc += IGC_READ_REG(hw, IGC_MPC);
1683 	stats->scc += IGC_READ_REG(hw, IGC_SCC);
1684 	stats->ecol += IGC_READ_REG(hw, IGC_ECOL);
1685 
1686 	stats->mcc += IGC_READ_REG(hw, IGC_MCC);
1687 	stats->latecol += IGC_READ_REG(hw, IGC_LATECOL);
1688 	stats->colc += IGC_READ_REG(hw, IGC_COLC);
1689 
1690 	stats->dc += IGC_READ_REG(hw, IGC_DC);
1691 	stats->tncrs += IGC_READ_REG(hw, IGC_TNCRS);
1692 	stats->htdpmc += IGC_READ_REG(hw, IGC_HTDPMC);
1693 	stats->rlec += IGC_READ_REG(hw, IGC_RLEC);
1694 	stats->xonrxc += IGC_READ_REG(hw, IGC_XONRXC);
1695 	stats->xontxc += IGC_READ_REG(hw, IGC_XONTXC);
1696 
1697 	/*
1698 	 * For watchdog management we need to know if we have been
1699 	 * paused during the last interval, so capture that here.
1700 	 */
1701 	pause_frames = IGC_READ_REG(hw, IGC_XOFFRXC);
1702 	stats->xoffrxc += pause_frames;
1703 	stats->xofftxc += IGC_READ_REG(hw, IGC_XOFFTXC);
1704 	stats->fcruc += IGC_READ_REG(hw, IGC_FCRUC);
1705 	stats->prc64 += IGC_READ_REG(hw, IGC_PRC64);
1706 	stats->prc127 += IGC_READ_REG(hw, IGC_PRC127);
1707 	stats->prc255 += IGC_READ_REG(hw, IGC_PRC255);
1708 	stats->prc511 += IGC_READ_REG(hw, IGC_PRC511);
1709 	stats->prc1023 += IGC_READ_REG(hw, IGC_PRC1023);
1710 	stats->prc1522 += IGC_READ_REG(hw, IGC_PRC1522);
1711 	stats->gprc += IGC_READ_REG(hw, IGC_GPRC);
1712 	stats->bprc += IGC_READ_REG(hw, IGC_BPRC);
1713 	stats->mprc += IGC_READ_REG(hw, IGC_MPRC);
1714 	stats->gptc += IGC_READ_REG(hw, IGC_GPTC);
1715 
1716 	/* For the 64-bit byte counters the low dword must be read first. */
1717 	/* Both registers clear on the read of the high dword */
1718 
1719 	/* Workaround CRC bytes included in size, take away 4 bytes/packet */
1720 	stats->gorc += IGC_READ_REG(hw, IGC_GORCL);
1721 	stats->gorc += ((uint64_t)IGC_READ_REG(hw, IGC_GORCH) << 32);
1722 	stats->gorc -= (stats->gprc - old_gprc) * RTE_ETHER_CRC_LEN;
1723 	stats->gotc += IGC_READ_REG(hw, IGC_GOTCL);
1724 	stats->gotc += ((uint64_t)IGC_READ_REG(hw, IGC_GOTCH) << 32);
1725 	stats->gotc -= (stats->gptc - old_gptc) * RTE_ETHER_CRC_LEN;
1726 
1727 	stats->rnbc += IGC_READ_REG(hw, IGC_RNBC);
1728 	stats->ruc += IGC_READ_REG(hw, IGC_RUC);
1729 	stats->rfc += IGC_READ_REG(hw, IGC_RFC);
1730 	stats->roc += IGC_READ_REG(hw, IGC_ROC);
1731 	stats->rjc += IGC_READ_REG(hw, IGC_RJC);
1732 
1733 	stats->mgprc += IGC_READ_REG(hw, IGC_MGTPRC);
1734 	stats->mgpdc += IGC_READ_REG(hw, IGC_MGTPDC);
1735 	stats->mgptc += IGC_READ_REG(hw, IGC_MGTPTC);
1736 	stats->b2ospc += IGC_READ_REG(hw, IGC_B2OSPC);
1737 	stats->b2ogprc += IGC_READ_REG(hw, IGC_B2OGPRC);
1738 	stats->o2bgptc += IGC_READ_REG(hw, IGC_O2BGPTC);
1739 	stats->o2bspc += IGC_READ_REG(hw, IGC_O2BSPC);
1740 
1741 	stats->tpr += IGC_READ_REG(hw, IGC_TPR);
1742 	stats->tpt += IGC_READ_REG(hw, IGC_TPT);
1743 
1744 	stats->tor += IGC_READ_REG(hw, IGC_TORL);
1745 	stats->tor += ((uint64_t)IGC_READ_REG(hw, IGC_TORH) << 32);
1746 	stats->tor -= (stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
1747 	stats->tot += IGC_READ_REG(hw, IGC_TOTL);
1748 	stats->tot += ((uint64_t)IGC_READ_REG(hw, IGC_TOTH) << 32);
1749 	stats->tot -= (stats->tpt - old_tpt) * RTE_ETHER_CRC_LEN;
1750 
1751 	stats->ptc64 += IGC_READ_REG(hw, IGC_PTC64);
1752 	stats->ptc127 += IGC_READ_REG(hw, IGC_PTC127);
1753 	stats->ptc255 += IGC_READ_REG(hw, IGC_PTC255);
1754 	stats->ptc511 += IGC_READ_REG(hw, IGC_PTC511);
1755 	stats->ptc1023 += IGC_READ_REG(hw, IGC_PTC1023);
1756 	stats->ptc1522 += IGC_READ_REG(hw, IGC_PTC1522);
1757 	stats->mptc += IGC_READ_REG(hw, IGC_MPTC);
1758 	stats->bptc += IGC_READ_REG(hw, IGC_BPTC);
1759 	stats->tsctc += IGC_READ_REG(hw, IGC_TSCTC);
1760 
1761 	stats->iac += IGC_READ_REG(hw, IGC_IAC);
1762 	stats->rpthc += IGC_READ_REG(hw, IGC_RPTHC);
1763 	stats->hgptc += IGC_READ_REG(hw, IGC_HGPTC);
1764 	stats->icrxdmtc += IGC_READ_REG(hw, IGC_ICRXDMTC);
1765 
1766 	/* Host to Card Statistics */
1767 	stats->hgorc += IGC_READ_REG(hw, IGC_HGORCL);
1768 	stats->hgorc += ((uint64_t)IGC_READ_REG(hw, IGC_HGORCH) << 32);
1769 	stats->hgorc -= (stats->rpthc - old_rpthc) * RTE_ETHER_CRC_LEN;
1770 	stats->hgotc += IGC_READ_REG(hw, IGC_HGOTCL);
1771 	stats->hgotc += ((uint64_t)IGC_READ_REG(hw, IGC_HGOTCH) << 32);
1772 	stats->hgotc -= (stats->hgptc - old_hgptc) * RTE_ETHER_CRC_LEN;
1773 	stats->lenerrs += IGC_READ_REG(hw, IGC_LENERRS);
1774 }
1775 
1776 /*
1777  * Write 0 to all queue status registers
1778  */
1779 static void
1780 igc_reset_queue_stats_register(struct igc_hw *hw)
1781 {
1782 	int i;
1783 
1784 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1785 		IGC_WRITE_REG(hw, IGC_PQGPRC(i), 0);
1786 		IGC_WRITE_REG(hw, IGC_PQGPTC(i), 0);
1787 		IGC_WRITE_REG(hw, IGC_PQGORC(i), 0);
1788 		IGC_WRITE_REG(hw, IGC_PQGOTC(i), 0);
1789 		IGC_WRITE_REG(hw, IGC_PQMPRC(i), 0);
1790 		IGC_WRITE_REG(hw, IGC_RQDPC(i), 0);
1791 		IGC_WRITE_REG(hw, IGC_TQDPC(i), 0);
1792 	}
1793 }
1794 
1795 /*
1796  * Read all hardware queue status registers
1797  */
1798 static void
1799 igc_read_queue_stats_register(struct rte_eth_dev *dev)
1800 {
1801 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1802 	struct igc_hw_queue_stats *queue_stats =
1803 				IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1804 	int i;
1805 
1806 	/*
1807 	 * This register is not cleared on read. Furthermore, the register wraps
1808 	 * around back to 0x00000000 on the next increment when reaching a value
1809 	 * of 0xFFFFFFFF and then continues normal count operation.
1810 	 */
1811 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1812 		union {
1813 			u64 ddword;
1814 			u32 dword[2];
1815 		} value;
1816 		u32 tmp;
1817 
1818 		/*
1819 		 * Read the register first, if the value is smaller than that
1820 		 * previous read, that mean the register has been overflowed,
1821 		 * then we add the high 4 bytes by 1 and replace the low 4
1822 		 * bytes by the new value.
1823 		 */
1824 		tmp = IGC_READ_REG(hw, IGC_PQGPRC(i));
1825 		value.ddword = queue_stats->pqgprc[i];
1826 		if (value.dword[U32_0_IN_U64] > tmp)
1827 			value.dword[U32_1_IN_U64]++;
1828 		value.dword[U32_0_IN_U64] = tmp;
1829 		queue_stats->pqgprc[i] = value.ddword;
1830 
1831 		tmp = IGC_READ_REG(hw, IGC_PQGPTC(i));
1832 		value.ddword = queue_stats->pqgptc[i];
1833 		if (value.dword[U32_0_IN_U64] > tmp)
1834 			value.dword[U32_1_IN_U64]++;
1835 		value.dword[U32_0_IN_U64] = tmp;
1836 		queue_stats->pqgptc[i] = value.ddword;
1837 
1838 		tmp = IGC_READ_REG(hw, IGC_PQGORC(i));
1839 		value.ddword = queue_stats->pqgorc[i];
1840 		if (value.dword[U32_0_IN_U64] > tmp)
1841 			value.dword[U32_1_IN_U64]++;
1842 		value.dword[U32_0_IN_U64] = tmp;
1843 		queue_stats->pqgorc[i] = value.ddword;
1844 
1845 		tmp = IGC_READ_REG(hw, IGC_PQGOTC(i));
1846 		value.ddword = queue_stats->pqgotc[i];
1847 		if (value.dword[U32_0_IN_U64] > tmp)
1848 			value.dword[U32_1_IN_U64]++;
1849 		value.dword[U32_0_IN_U64] = tmp;
1850 		queue_stats->pqgotc[i] = value.ddword;
1851 
1852 		tmp = IGC_READ_REG(hw, IGC_PQMPRC(i));
1853 		value.ddword = queue_stats->pqmprc[i];
1854 		if (value.dword[U32_0_IN_U64] > tmp)
1855 			value.dword[U32_1_IN_U64]++;
1856 		value.dword[U32_0_IN_U64] = tmp;
1857 		queue_stats->pqmprc[i] = value.ddword;
1858 
1859 		tmp = IGC_READ_REG(hw, IGC_RQDPC(i));
1860 		value.ddword = queue_stats->rqdpc[i];
1861 		if (value.dword[U32_0_IN_U64] > tmp)
1862 			value.dword[U32_1_IN_U64]++;
1863 		value.dword[U32_0_IN_U64] = tmp;
1864 		queue_stats->rqdpc[i] = value.ddword;
1865 
1866 		tmp = IGC_READ_REG(hw, IGC_TQDPC(i));
1867 		value.ddword = queue_stats->tqdpc[i];
1868 		if (value.dword[U32_0_IN_U64] > tmp)
1869 			value.dword[U32_1_IN_U64]++;
1870 		value.dword[U32_0_IN_U64] = tmp;
1871 		queue_stats->tqdpc[i] = value.ddword;
1872 	}
1873 }
1874 
1875 static int
1876 eth_igc_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1877 {
1878 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
1879 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1880 	struct igc_hw_stats *stats = IGC_DEV_PRIVATE_STATS(dev);
1881 	struct igc_hw_queue_stats *queue_stats =
1882 			IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1883 	int i;
1884 
1885 	/*
1886 	 * Cancel status handler since it will read the queue status registers
1887 	 */
1888 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
1889 
1890 	/* Read status register */
1891 	igc_read_queue_stats_register(dev);
1892 	igc_read_stats_registers(hw, stats);
1893 
1894 	if (rte_stats == NULL) {
1895 		/* Restart queue status handler */
1896 		rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1897 				igc_update_queue_stats_handler, dev);
1898 		return -EINVAL;
1899 	}
1900 
1901 	/* Rx Errors */
1902 	rte_stats->imissed = stats->mpc;
1903 	rte_stats->ierrors = stats->crcerrs +
1904 			stats->rlec + stats->ruc + stats->roc +
1905 			stats->rxerrc + stats->algnerrc;
1906 
1907 	/* Tx Errors */
1908 	rte_stats->oerrors = stats->ecol + stats->latecol;
1909 
1910 	rte_stats->ipackets = stats->gprc;
1911 	rte_stats->opackets = stats->gptc;
1912 	rte_stats->ibytes   = stats->gorc;
1913 	rte_stats->obytes   = stats->gotc;
1914 
1915 	/* Get per-queue statuses */
1916 	for (i = 0; i < IGC_QUEUE_PAIRS_NUM; i++) {
1917 		/* GET TX queue statuses */
1918 		int map_id = igc->txq_stats_map[i];
1919 		if (map_id >= 0) {
1920 			rte_stats->q_opackets[map_id] += queue_stats->pqgptc[i];
1921 			rte_stats->q_obytes[map_id] += queue_stats->pqgotc[i];
1922 		}
1923 		/* Get RX queue statuses */
1924 		map_id = igc->rxq_stats_map[i];
1925 		if (map_id >= 0) {
1926 			rte_stats->q_ipackets[map_id] += queue_stats->pqgprc[i];
1927 			rte_stats->q_ibytes[map_id] += queue_stats->pqgorc[i];
1928 			rte_stats->q_errors[map_id] += queue_stats->rqdpc[i];
1929 		}
1930 	}
1931 
1932 	/* Restart queue status handler */
1933 	rte_eal_alarm_set(IGC_ALARM_INTERVAL,
1934 			igc_update_queue_stats_handler, dev);
1935 	return 0;
1936 }
1937 
1938 static int
1939 eth_igc_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1940 		   unsigned int n)
1941 {
1942 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1943 	struct igc_hw_stats *hw_stats =
1944 			IGC_DEV_PRIVATE_STATS(dev);
1945 	unsigned int i;
1946 
1947 	igc_read_stats_registers(hw, hw_stats);
1948 
1949 	if (n < IGC_NB_XSTATS)
1950 		return IGC_NB_XSTATS;
1951 
1952 	/* If this is a reset xstats is NULL, and we have cleared the
1953 	 * registers by reading them.
1954 	 */
1955 	if (!xstats)
1956 		return 0;
1957 
1958 	/* Extended stats */
1959 	for (i = 0; i < IGC_NB_XSTATS; i++) {
1960 		xstats[i].id = i;
1961 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1962 			rte_igc_stats_strings[i].offset);
1963 	}
1964 
1965 	return IGC_NB_XSTATS;
1966 }
1967 
1968 static int
1969 eth_igc_xstats_reset(struct rte_eth_dev *dev)
1970 {
1971 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
1972 	struct igc_hw_stats *hw_stats = IGC_DEV_PRIVATE_STATS(dev);
1973 	struct igc_hw_queue_stats *queue_stats =
1974 			IGC_DEV_PRIVATE_QUEUE_STATS(dev);
1975 
1976 	/* Cancel queue status handler for avoid conflict */
1977 	rte_eal_alarm_cancel(igc_update_queue_stats_handler, dev);
1978 
1979 	/* HW registers are cleared on read */
1980 	igc_reset_queue_stats_register(hw);
1981 	igc_read_stats_registers(hw, hw_stats);
1982 
1983 	/* Reset software totals */
1984 	memset(hw_stats, 0, sizeof(*hw_stats));
1985 	memset(queue_stats, 0, sizeof(*queue_stats));
1986 
1987 	/* Restart the queue status handler */
1988 	rte_eal_alarm_set(IGC_ALARM_INTERVAL, igc_update_queue_stats_handler,
1989 			dev);
1990 
1991 	return 0;
1992 }
1993 
1994 static int
1995 eth_igc_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1996 	struct rte_eth_xstat_name *xstats_names, unsigned int size)
1997 {
1998 	unsigned int i;
1999 
2000 	if (xstats_names == NULL)
2001 		return IGC_NB_XSTATS;
2002 
2003 	if (size < IGC_NB_XSTATS) {
2004 		PMD_DRV_LOG(ERR, "not enough buffers!");
2005 		return IGC_NB_XSTATS;
2006 	}
2007 
2008 	for (i = 0; i < IGC_NB_XSTATS; i++)
2009 		strlcpy(xstats_names[i].name, rte_igc_stats_strings[i].name,
2010 			sizeof(xstats_names[i].name));
2011 
2012 	return IGC_NB_XSTATS;
2013 }
2014 
2015 static int
2016 eth_igc_xstats_get_names_by_id(struct rte_eth_dev *dev,
2017 		struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
2018 		unsigned int limit)
2019 {
2020 	unsigned int i;
2021 
2022 	if (!ids)
2023 		return eth_igc_xstats_get_names(dev, xstats_names, limit);
2024 
2025 	for (i = 0; i < limit; i++) {
2026 		if (ids[i] >= IGC_NB_XSTATS) {
2027 			PMD_DRV_LOG(ERR, "id value isn't valid");
2028 			return -EINVAL;
2029 		}
2030 		strlcpy(xstats_names[i].name,
2031 			rte_igc_stats_strings[ids[i]].name,
2032 			sizeof(xstats_names[i].name));
2033 	}
2034 	return limit;
2035 }
2036 
2037 static int
2038 eth_igc_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
2039 		uint64_t *values, unsigned int n)
2040 {
2041 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2042 	struct igc_hw_stats *hw_stats = IGC_DEV_PRIVATE_STATS(dev);
2043 	unsigned int i;
2044 
2045 	igc_read_stats_registers(hw, hw_stats);
2046 
2047 	if (!ids) {
2048 		if (n < IGC_NB_XSTATS)
2049 			return IGC_NB_XSTATS;
2050 
2051 		/* If this is a reset xstats is NULL, and we have cleared the
2052 		 * registers by reading them.
2053 		 */
2054 		if (!values)
2055 			return 0;
2056 
2057 		/* Extended stats */
2058 		for (i = 0; i < IGC_NB_XSTATS; i++)
2059 			values[i] = *(uint64_t *)(((char *)hw_stats) +
2060 					rte_igc_stats_strings[i].offset);
2061 
2062 		return IGC_NB_XSTATS;
2063 
2064 	} else {
2065 		for (i = 0; i < n; i++) {
2066 			if (ids[i] >= IGC_NB_XSTATS) {
2067 				PMD_DRV_LOG(ERR, "id value isn't valid");
2068 				return -EINVAL;
2069 			}
2070 			values[i] = *(uint64_t *)(((char *)hw_stats) +
2071 					rte_igc_stats_strings[ids[i]].offset);
2072 		}
2073 		return n;
2074 	}
2075 }
2076 
2077 static int
2078 eth_igc_queue_stats_mapping_set(struct rte_eth_dev *dev,
2079 		uint16_t queue_id, uint8_t stat_idx, uint8_t is_rx)
2080 {
2081 	struct igc_adapter *igc = IGC_DEV_PRIVATE(dev);
2082 
2083 	/* check queue id is valid */
2084 	if (queue_id >= IGC_QUEUE_PAIRS_NUM) {
2085 		PMD_DRV_LOG(ERR, "queue id(%u) error, max is %u",
2086 			queue_id, IGC_QUEUE_PAIRS_NUM - 1);
2087 		return -EINVAL;
2088 	}
2089 
2090 	/* store the mapping status id */
2091 	if (is_rx)
2092 		igc->rxq_stats_map[queue_id] = stat_idx;
2093 	else
2094 		igc->txq_stats_map[queue_id] = stat_idx;
2095 
2096 	return 0;
2097 }
2098 
2099 static int
2100 eth_igc_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
2101 {
2102 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2103 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2104 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2105 	uint32_t vec = IGC_MISC_VEC_ID;
2106 
2107 	if (rte_intr_allow_others(intr_handle))
2108 		vec = IGC_RX_VEC_START;
2109 
2110 	uint32_t mask = 1u << (queue_id + vec);
2111 
2112 	IGC_WRITE_REG(hw, IGC_EIMC, mask);
2113 	IGC_WRITE_FLUSH(hw);
2114 
2115 	return 0;
2116 }
2117 
2118 static int
2119 eth_igc_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
2120 {
2121 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2122 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2123 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2124 	uint32_t vec = IGC_MISC_VEC_ID;
2125 
2126 	if (rte_intr_allow_others(intr_handle))
2127 		vec = IGC_RX_VEC_START;
2128 
2129 	uint32_t mask = 1u << (queue_id + vec);
2130 
2131 	IGC_WRITE_REG(hw, IGC_EIMS, mask);
2132 	IGC_WRITE_FLUSH(hw);
2133 
2134 	rte_intr_enable(intr_handle);
2135 
2136 	return 0;
2137 }
2138 
2139 static int
2140 eth_igc_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2141 {
2142 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2143 	uint32_t ctrl;
2144 	int tx_pause;
2145 	int rx_pause;
2146 
2147 	fc_conf->pause_time = hw->fc.pause_time;
2148 	fc_conf->high_water = hw->fc.high_water;
2149 	fc_conf->low_water = hw->fc.low_water;
2150 	fc_conf->send_xon = hw->fc.send_xon;
2151 	fc_conf->autoneg = hw->mac.autoneg;
2152 
2153 	/*
2154 	 * Return rx_pause and tx_pause status according to actual setting of
2155 	 * the TFCE and RFCE bits in the CTRL register.
2156 	 */
2157 	ctrl = IGC_READ_REG(hw, IGC_CTRL);
2158 	if (ctrl & IGC_CTRL_TFCE)
2159 		tx_pause = 1;
2160 	else
2161 		tx_pause = 0;
2162 
2163 	if (ctrl & IGC_CTRL_RFCE)
2164 		rx_pause = 1;
2165 	else
2166 		rx_pause = 0;
2167 
2168 	if (rx_pause && tx_pause)
2169 		fc_conf->mode = RTE_FC_FULL;
2170 	else if (rx_pause)
2171 		fc_conf->mode = RTE_FC_RX_PAUSE;
2172 	else if (tx_pause)
2173 		fc_conf->mode = RTE_FC_TX_PAUSE;
2174 	else
2175 		fc_conf->mode = RTE_FC_NONE;
2176 
2177 	return 0;
2178 }
2179 
2180 static int
2181 eth_igc_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2182 {
2183 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2184 	uint32_t rx_buf_size;
2185 	uint32_t max_high_water;
2186 	uint32_t rctl;
2187 	int err;
2188 
2189 	if (fc_conf->autoneg != hw->mac.autoneg)
2190 		return -ENOTSUP;
2191 
2192 	rx_buf_size = igc_get_rx_buffer_size(hw);
2193 	PMD_DRV_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
2194 
2195 	/* At least reserve one Ethernet frame for watermark */
2196 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
2197 	if (fc_conf->high_water > max_high_water ||
2198 		fc_conf->high_water < fc_conf->low_water) {
2199 		PMD_DRV_LOG(ERR,
2200 			"Incorrect high(%u)/low(%u) water value, max is %u",
2201 			fc_conf->high_water, fc_conf->low_water,
2202 			max_high_water);
2203 		return -EINVAL;
2204 	}
2205 
2206 	switch (fc_conf->mode) {
2207 	case RTE_FC_NONE:
2208 		hw->fc.requested_mode = igc_fc_none;
2209 		break;
2210 	case RTE_FC_RX_PAUSE:
2211 		hw->fc.requested_mode = igc_fc_rx_pause;
2212 		break;
2213 	case RTE_FC_TX_PAUSE:
2214 		hw->fc.requested_mode = igc_fc_tx_pause;
2215 		break;
2216 	case RTE_FC_FULL:
2217 		hw->fc.requested_mode = igc_fc_full;
2218 		break;
2219 	default:
2220 		PMD_DRV_LOG(ERR, "unsupported fc mode: %u", fc_conf->mode);
2221 		return -EINVAL;
2222 	}
2223 
2224 	hw->fc.pause_time     = fc_conf->pause_time;
2225 	hw->fc.high_water     = fc_conf->high_water;
2226 	hw->fc.low_water      = fc_conf->low_water;
2227 	hw->fc.send_xon	      = fc_conf->send_xon;
2228 
2229 	err = igc_setup_link_generic(hw);
2230 	if (err == IGC_SUCCESS) {
2231 		/**
2232 		 * check if we want to forward MAC frames - driver doesn't have
2233 		 * native capability to do that, so we'll write the registers
2234 		 * ourselves
2235 		 **/
2236 		rctl = IGC_READ_REG(hw, IGC_RCTL);
2237 
2238 		/* set or clear MFLCN.PMCF bit depending on configuration */
2239 		if (fc_conf->mac_ctrl_frame_fwd != 0)
2240 			rctl |= IGC_RCTL_PMCF;
2241 		else
2242 			rctl &= ~IGC_RCTL_PMCF;
2243 
2244 		IGC_WRITE_REG(hw, IGC_RCTL, rctl);
2245 		IGC_WRITE_FLUSH(hw);
2246 
2247 		return 0;
2248 	}
2249 
2250 	PMD_DRV_LOG(ERR, "igc_setup_link_generic = 0x%x", err);
2251 	return -EIO;
2252 }
2253 
2254 static int
2255 eth_igc_rss_reta_update(struct rte_eth_dev *dev,
2256 			struct rte_eth_rss_reta_entry64 *reta_conf,
2257 			uint16_t reta_size)
2258 {
2259 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2260 	uint16_t i;
2261 
2262 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
2263 		PMD_DRV_LOG(ERR,
2264 			"The size of RSS redirection table configured(%d) doesn't match the number hardware can supported(%d)",
2265 			reta_size, ETH_RSS_RETA_SIZE_128);
2266 		return -EINVAL;
2267 	}
2268 
2269 	/* set redirection table */
2270 	for (i = 0; i < ETH_RSS_RETA_SIZE_128; i += IGC_RSS_RDT_REG_SIZE) {
2271 		union igc_rss_reta_reg reta, reg;
2272 		uint16_t idx, shift;
2273 		uint8_t j, mask;
2274 
2275 		idx = i / RTE_RETA_GROUP_SIZE;
2276 		shift = i % RTE_RETA_GROUP_SIZE;
2277 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
2278 				IGC_RSS_RDT_REG_SIZE_MASK);
2279 
2280 		/* if no need to update the register */
2281 		if (!mask)
2282 			continue;
2283 
2284 		/* check mask whether need to read the register value first */
2285 		if (mask == IGC_RSS_RDT_REG_SIZE_MASK)
2286 			reg.dword = 0;
2287 		else
2288 			reg.dword = IGC_READ_REG_LE_VALUE(hw,
2289 					IGC_RETA(i / IGC_RSS_RDT_REG_SIZE));
2290 
2291 		/* update the register */
2292 		for (j = 0; j < IGC_RSS_RDT_REG_SIZE; j++) {
2293 			if (mask & (1u << j))
2294 				reta.bytes[j] =
2295 					(uint8_t)reta_conf[idx].reta[shift + j];
2296 			else
2297 				reta.bytes[j] = reg.bytes[j];
2298 		}
2299 		IGC_WRITE_REG_LE_VALUE(hw,
2300 			IGC_RETA(i / IGC_RSS_RDT_REG_SIZE), reta.dword);
2301 	}
2302 
2303 	return 0;
2304 }
2305 
2306 static int
2307 eth_igc_rss_reta_query(struct rte_eth_dev *dev,
2308 		       struct rte_eth_rss_reta_entry64 *reta_conf,
2309 		       uint16_t reta_size)
2310 {
2311 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2312 	uint16_t i;
2313 
2314 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
2315 		PMD_DRV_LOG(ERR,
2316 			"The size of RSS redirection table configured(%d) doesn't match the number hardware can supported(%d)",
2317 			reta_size, ETH_RSS_RETA_SIZE_128);
2318 		return -EINVAL;
2319 	}
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 			continue;
2335 
2336 		/* read register and get the queue index */
2337 		reta.dword = IGC_READ_REG_LE_VALUE(hw,
2338 				IGC_RETA(i / IGC_RSS_RDT_REG_SIZE));
2339 		for (j = 0; j < IGC_RSS_RDT_REG_SIZE; j++) {
2340 			if (mask & (1u << j))
2341 				reta_conf[idx].reta[shift + j] = reta.bytes[j];
2342 		}
2343 	}
2344 
2345 	return 0;
2346 }
2347 
2348 static int
2349 eth_igc_rss_hash_update(struct rte_eth_dev *dev,
2350 			struct rte_eth_rss_conf *rss_conf)
2351 {
2352 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2353 	igc_hw_rss_hash_set(hw, rss_conf);
2354 	return 0;
2355 }
2356 
2357 static int
2358 eth_igc_rss_hash_conf_get(struct rte_eth_dev *dev,
2359 			struct rte_eth_rss_conf *rss_conf)
2360 {
2361 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2362 	uint32_t *hash_key = (uint32_t *)rss_conf->rss_key;
2363 	uint32_t mrqc;
2364 	uint64_t rss_hf;
2365 
2366 	if (hash_key != NULL) {
2367 		int i;
2368 
2369 		/* if not enough space for store hash key */
2370 		if (rss_conf->rss_key_len != IGC_HKEY_SIZE) {
2371 			PMD_DRV_LOG(ERR,
2372 				"RSS hash key size %u in parameter doesn't match the hardware hash key size %u",
2373 				rss_conf->rss_key_len, IGC_HKEY_SIZE);
2374 			return -EINVAL;
2375 		}
2376 
2377 		/* read RSS key from register */
2378 		for (i = 0; i < IGC_HKEY_MAX_INDEX; i++)
2379 			hash_key[i] = IGC_READ_REG_LE_VALUE(hw, IGC_RSSRK(i));
2380 	}
2381 
2382 	/* get RSS functions configured in MRQC register */
2383 	mrqc = IGC_READ_REG(hw, IGC_MRQC);
2384 	if ((mrqc & IGC_MRQC_ENABLE_RSS_4Q) == 0)
2385 		return 0;
2386 
2387 	rss_hf = 0;
2388 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4)
2389 		rss_hf |= ETH_RSS_IPV4;
2390 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4_TCP)
2391 		rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2392 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6)
2393 		rss_hf |= ETH_RSS_IPV6;
2394 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_EX)
2395 		rss_hf |= ETH_RSS_IPV6_EX;
2396 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_TCP)
2397 		rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2398 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_TCP_EX)
2399 		rss_hf |= ETH_RSS_IPV6_TCP_EX;
2400 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV4_UDP)
2401 		rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2402 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_UDP)
2403 		rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2404 	if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_UDP_EX)
2405 		rss_hf |= ETH_RSS_IPV6_UDP_EX;
2406 
2407 	rss_conf->rss_hf |= rss_hf;
2408 	return 0;
2409 }
2410 
2411 static int
2412 eth_igc_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2413 {
2414 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2415 	struct igc_vfta *shadow_vfta = IGC_DEV_PRIVATE_VFTA(dev);
2416 	uint32_t vfta;
2417 	uint32_t vid_idx;
2418 	uint32_t vid_bit;
2419 
2420 	vid_idx = (vlan_id >> IGC_VFTA_ENTRY_SHIFT) & IGC_VFTA_ENTRY_MASK;
2421 	vid_bit = 1u << (vlan_id & IGC_VFTA_ENTRY_BIT_SHIFT_MASK);
2422 	vfta = shadow_vfta->vfta[vid_idx];
2423 	if (on)
2424 		vfta |= vid_bit;
2425 	else
2426 		vfta &= ~vid_bit;
2427 	IGC_WRITE_REG_ARRAY(hw, IGC_VFTA, vid_idx, vfta);
2428 
2429 	/* update local VFTA copy */
2430 	shadow_vfta->vfta[vid_idx] = vfta;
2431 
2432 	return 0;
2433 }
2434 
2435 static void
2436 igc_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2437 {
2438 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2439 	igc_read_reg_check_clear_bits(hw, IGC_RCTL,
2440 			IGC_RCTL_CFIEN | IGC_RCTL_VFE);
2441 }
2442 
2443 static void
2444 igc_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2445 {
2446 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2447 	struct igc_vfta *shadow_vfta = IGC_DEV_PRIVATE_VFTA(dev);
2448 	uint32_t reg_val;
2449 	int i;
2450 
2451 	/* Filter Table Enable, CFI not used for packet acceptance */
2452 	reg_val = IGC_READ_REG(hw, IGC_RCTL);
2453 	reg_val &= ~IGC_RCTL_CFIEN;
2454 	reg_val |= IGC_RCTL_VFE;
2455 	IGC_WRITE_REG(hw, IGC_RCTL, reg_val);
2456 
2457 	/* restore VFTA table */
2458 	for (i = 0; i < IGC_VFTA_SIZE; i++)
2459 		IGC_WRITE_REG_ARRAY(hw, IGC_VFTA, i, shadow_vfta->vfta[i]);
2460 }
2461 
2462 static void
2463 igc_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2464 {
2465 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2466 
2467 	igc_read_reg_check_clear_bits(hw, IGC_CTRL, IGC_CTRL_VME);
2468 }
2469 
2470 static void
2471 igc_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2472 {
2473 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2474 
2475 	igc_read_reg_check_set_bits(hw, IGC_CTRL, IGC_CTRL_VME);
2476 }
2477 
2478 static int
2479 igc_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2480 {
2481 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2482 	uint32_t ctrl_ext;
2483 
2484 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
2485 
2486 	/* if extend vlan hasn't been enabled */
2487 	if ((ctrl_ext & IGC_CTRL_EXT_EXT_VLAN) == 0)
2488 		return 0;
2489 
2490 	if ((dev->data->dev_conf.rxmode.offloads &
2491 			DEV_RX_OFFLOAD_JUMBO_FRAME) == 0)
2492 		goto write_ext_vlan;
2493 
2494 	/* Update maximum packet length */
2495 	if (dev->data->dev_conf.rxmode.max_rx_pkt_len <
2496 		RTE_ETHER_MIN_MTU + VLAN_TAG_SIZE) {
2497 		PMD_DRV_LOG(ERR, "Maximum packet length %u error, min is %u",
2498 			dev->data->dev_conf.rxmode.max_rx_pkt_len,
2499 			VLAN_TAG_SIZE + RTE_ETHER_MIN_MTU);
2500 		return -EINVAL;
2501 	}
2502 	dev->data->dev_conf.rxmode.max_rx_pkt_len -= VLAN_TAG_SIZE;
2503 	IGC_WRITE_REG(hw, IGC_RLPML,
2504 		dev->data->dev_conf.rxmode.max_rx_pkt_len);
2505 
2506 write_ext_vlan:
2507 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext & ~IGC_CTRL_EXT_EXT_VLAN);
2508 	return 0;
2509 }
2510 
2511 static int
2512 igc_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2513 {
2514 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2515 	uint32_t ctrl_ext;
2516 
2517 	ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
2518 
2519 	/* if extend vlan has been enabled */
2520 	if (ctrl_ext & IGC_CTRL_EXT_EXT_VLAN)
2521 		return 0;
2522 
2523 	if ((dev->data->dev_conf.rxmode.offloads &
2524 			DEV_RX_OFFLOAD_JUMBO_FRAME) == 0)
2525 		goto write_ext_vlan;
2526 
2527 	/* Update maximum packet length */
2528 	if (dev->data->dev_conf.rxmode.max_rx_pkt_len >
2529 		MAX_RX_JUMBO_FRAME_SIZE - VLAN_TAG_SIZE) {
2530 		PMD_DRV_LOG(ERR, "Maximum packet length %u error, max is %u",
2531 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2532 			VLAN_TAG_SIZE, MAX_RX_JUMBO_FRAME_SIZE);
2533 		return -EINVAL;
2534 	}
2535 	dev->data->dev_conf.rxmode.max_rx_pkt_len += VLAN_TAG_SIZE;
2536 	IGC_WRITE_REG(hw, IGC_RLPML,
2537 		dev->data->dev_conf.rxmode.max_rx_pkt_len);
2538 
2539 write_ext_vlan:
2540 	IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext | IGC_CTRL_EXT_EXT_VLAN);
2541 	return 0;
2542 }
2543 
2544 static int
2545 eth_igc_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2546 {
2547 	struct rte_eth_rxmode *rxmode;
2548 
2549 	rxmode = &dev->data->dev_conf.rxmode;
2550 	if (mask & ETH_VLAN_STRIP_MASK) {
2551 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2552 			igc_vlan_hw_strip_enable(dev);
2553 		else
2554 			igc_vlan_hw_strip_disable(dev);
2555 	}
2556 
2557 	if (mask & ETH_VLAN_FILTER_MASK) {
2558 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2559 			igc_vlan_hw_filter_enable(dev);
2560 		else
2561 			igc_vlan_hw_filter_disable(dev);
2562 	}
2563 
2564 	if (mask & ETH_VLAN_EXTEND_MASK) {
2565 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2566 			return igc_vlan_hw_extend_enable(dev);
2567 		else
2568 			return igc_vlan_hw_extend_disable(dev);
2569 	}
2570 
2571 	return 0;
2572 }
2573 
2574 static int
2575 eth_igc_vlan_tpid_set(struct rte_eth_dev *dev,
2576 		      enum rte_vlan_type vlan_type,
2577 		      uint16_t tpid)
2578 {
2579 	struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
2580 	uint32_t reg_val;
2581 
2582 	/* only outer TPID of double VLAN can be configured*/
2583 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2584 		reg_val = IGC_READ_REG(hw, IGC_VET);
2585 		reg_val = (reg_val & (~IGC_VET_EXT)) |
2586 			((uint32_t)tpid << IGC_VET_EXT_SHIFT);
2587 		IGC_WRITE_REG(hw, IGC_VET, reg_val);
2588 
2589 		return 0;
2590 	}
2591 
2592 	/* all other TPID values are read-only*/
2593 	PMD_DRV_LOG(ERR, "Not supported");
2594 	return -ENOTSUP;
2595 }
2596 
2597 static int
2598 eth_igc_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
2599 	struct rte_pci_device *pci_dev)
2600 {
2601 	PMD_INIT_FUNC_TRACE();
2602 	return rte_eth_dev_pci_generic_probe(pci_dev,
2603 		sizeof(struct igc_adapter), eth_igc_dev_init);
2604 }
2605 
2606 static int
2607 eth_igc_pci_remove(struct rte_pci_device *pci_dev)
2608 {
2609 	PMD_INIT_FUNC_TRACE();
2610 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igc_dev_uninit);
2611 }
2612 
2613 static struct rte_pci_driver rte_igc_pmd = {
2614 	.id_table = pci_id_igc_map,
2615 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
2616 	.probe = eth_igc_pci_probe,
2617 	.remove = eth_igc_pci_remove,
2618 };
2619 
2620 RTE_PMD_REGISTER_PCI(net_igc, rte_igc_pmd);
2621 RTE_PMD_REGISTER_PCI_TABLE(net_igc, pci_id_igc_map);
2622 RTE_PMD_REGISTER_KMOD_DEP(net_igc, "* igb_uio | uio_pci_generic | vfio-pci");
2623