xref: /dpdk/drivers/net/ixgbe/ixgbe_ethdev.c (revision ffe77e91)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2017 Intel Corporation
3  */
4 
5 #include <sys/queue.h>
6 #include <stdio.h>
7 #include <errno.h>
8 #include <stdint.h>
9 #include <string.h>
10 #include <unistd.h>
11 #include <stdarg.h>
12 #include <inttypes.h>
13 #include <rte_string_fns.h>
14 #include <rte_byteorder.h>
15 #include <rte_common.h>
16 #include <rte_cycles.h>
17 
18 #include <rte_interrupts.h>
19 #include <rte_log.h>
20 #include <rte_debug.h>
21 #include <rte_pci.h>
22 #include <rte_bus_pci.h>
23 #include <rte_branch_prediction.h>
24 #include <rte_memory.h>
25 #include <rte_kvargs.h>
26 #include <rte_eal.h>
27 #include <rte_alarm.h>
28 #include <rte_ether.h>
29 #include <ethdev_driver.h>
30 #include <ethdev_pci.h>
31 #include <rte_malloc.h>
32 #include <rte_random.h>
33 #include <rte_dev.h>
34 #include <rte_hash_crc.h>
35 #ifdef RTE_LIB_SECURITY
36 #include <rte_security_driver.h>
37 #endif
38 
39 #include "ixgbe_logs.h"
40 #include "base/ixgbe_api.h"
41 #include "base/ixgbe_vf.h"
42 #include "base/ixgbe_common.h"
43 #include "ixgbe_ethdev.h"
44 #include "ixgbe_bypass.h"
45 #include "ixgbe_rxtx.h"
46 #include "base/ixgbe_type.h"
47 #include "base/ixgbe_phy.h"
48 #include "base/ixgbe_osdep.h"
49 #include "ixgbe_regs.h"
50 
51 /*
52  * High threshold controlling when to start sending XOFF frames. Must be at
53  * least 8 bytes less than receive packet buffer size. This value is in units
54  * of 1024 bytes.
55  */
56 #define IXGBE_FC_HI    0x80
57 
58 /*
59  * Low threshold controlling when to start sending XON frames. This value is
60  * in units of 1024 bytes.
61  */
62 #define IXGBE_FC_LO    0x40
63 
64 /* Timer value included in XOFF frames. */
65 #define IXGBE_FC_PAUSE 0x680
66 
67 /*Default value of Max Rx Queue*/
68 #define IXGBE_MAX_RX_QUEUE_NUM 128
69 
70 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
71 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
72 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
73 
74 #define IXGBE_MMW_SIZE_DEFAULT        0x4
75 #define IXGBE_MMW_SIZE_JUMBO_FRAME    0x14
76 #define IXGBE_MAX_RING_DESC           4096 /* replicate define from rxtx */
77 
78 /*
79  *  Default values for RX/TX configuration
80  */
81 #define IXGBE_DEFAULT_RX_FREE_THRESH  32
82 #define IXGBE_DEFAULT_RX_PTHRESH      8
83 #define IXGBE_DEFAULT_RX_HTHRESH      8
84 #define IXGBE_DEFAULT_RX_WTHRESH      0
85 
86 #define IXGBE_DEFAULT_TX_FREE_THRESH  32
87 #define IXGBE_DEFAULT_TX_PTHRESH      32
88 #define IXGBE_DEFAULT_TX_HTHRESH      0
89 #define IXGBE_DEFAULT_TX_WTHRESH      0
90 #define IXGBE_DEFAULT_TX_RSBIT_THRESH 32
91 
92 /* Bit shift and mask */
93 #define IXGBE_4_BIT_WIDTH  (CHAR_BIT / 2)
94 #define IXGBE_4_BIT_MASK   RTE_LEN2MASK(IXGBE_4_BIT_WIDTH, uint8_t)
95 #define IXGBE_8_BIT_WIDTH  CHAR_BIT
96 #define IXGBE_8_BIT_MASK   UINT8_MAX
97 
98 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
99 
100 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
101 
102 /* Additional timesync values. */
103 #define NSEC_PER_SEC             1000000000L
104 #define IXGBE_INCVAL_10GB        0x66666666
105 #define IXGBE_INCVAL_1GB         0x40000000
106 #define IXGBE_INCVAL_100         0x50000000
107 #define IXGBE_INCVAL_SHIFT_10GB  28
108 #define IXGBE_INCVAL_SHIFT_1GB   24
109 #define IXGBE_INCVAL_SHIFT_100   21
110 #define IXGBE_INCVAL_SHIFT_82599 7
111 #define IXGBE_INCPER_SHIFT_82599 24
112 
113 #define IXGBE_CYCLECOUNTER_MASK   0xffffffffffffffffULL
114 
115 #define IXGBE_VT_CTL_POOLING_MODE_MASK         0x00030000
116 #define IXGBE_VT_CTL_POOLING_MODE_ETAG         0x00010000
117 #define IXGBE_ETAG_ETYPE                       0x00005084
118 #define IXGBE_ETAG_ETYPE_MASK                  0x0000ffff
119 #define IXGBE_ETAG_ETYPE_VALID                 0x80000000
120 #define IXGBE_RAH_ADTYPE                       0x40000000
121 #define IXGBE_RAL_ETAG_FILTER_MASK             0x00003fff
122 #define IXGBE_VMVIR_TAGA_MASK                  0x18000000
123 #define IXGBE_VMVIR_TAGA_ETAG_INSERT           0x08000000
124 #define IXGBE_VMTIR(_i) (0x00017000 + ((_i) * 4)) /* 64 of these (0-63) */
125 #define IXGBE_QDE_STRIP_TAG                    0x00000004
126 #define IXGBE_VTEICR_MASK                      0x07
127 
128 #define IXGBE_EXVET_VET_EXT_SHIFT              16
129 #define IXGBE_DMATXCTL_VT_MASK                 0xFFFF0000
130 
131 #define IXGBEVF_DEVARG_PFLINK_FULLCHK		"pflink_fullchk"
132 
133 static const char * const ixgbevf_valid_arguments[] = {
134 	IXGBEVF_DEVARG_PFLINK_FULLCHK,
135 	NULL
136 };
137 
138 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params);
139 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
140 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev);
141 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev);
142 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev);
143 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev);
144 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev);
145 static int ixgbe_dev_configure(struct rte_eth_dev *dev);
146 static int ixgbe_dev_start(struct rte_eth_dev *dev);
147 static int ixgbe_dev_stop(struct rte_eth_dev *dev);
148 static int ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
149 static int ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
150 static int ixgbe_dev_close(struct rte_eth_dev *dev);
151 static int ixgbe_dev_reset(struct rte_eth_dev *dev);
152 static int ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
153 static int ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
154 static int ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
155 static int ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
156 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
157 				int wait_to_complete);
158 static int ixgbe_dev_stats_get(struct rte_eth_dev *dev,
159 				struct rte_eth_stats *stats);
160 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
161 				struct rte_eth_xstat *xstats, unsigned n);
162 static int ixgbevf_dev_xstats_get(struct rte_eth_dev *dev,
163 				  struct rte_eth_xstat *xstats, unsigned n);
164 static int
165 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
166 		uint64_t *values, unsigned int n);
167 static int ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
168 static int ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
169 static int ixgbe_dev_xstats_get_names(struct rte_eth_dev *dev,
170 	struct rte_eth_xstat_name *xstats_names,
171 	unsigned int size);
172 static int ixgbevf_dev_xstats_get_names(struct rte_eth_dev *dev,
173 	struct rte_eth_xstat_name *xstats_names, unsigned limit);
174 static int ixgbe_dev_xstats_get_names_by_id(
175 	struct rte_eth_dev *dev,
176 	const uint64_t *ids,
177 	struct rte_eth_xstat_name *xstats_names,
178 	unsigned int limit);
179 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
180 					     uint16_t queue_id,
181 					     uint8_t stat_idx,
182 					     uint8_t is_rx);
183 static int ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
184 				 size_t fw_size);
185 static int ixgbe_dev_info_get(struct rte_eth_dev *dev,
186 			      struct rte_eth_dev_info *dev_info);
187 static const uint32_t *ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev);
188 static int ixgbevf_dev_info_get(struct rte_eth_dev *dev,
189 				struct rte_eth_dev_info *dev_info);
190 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
191 
192 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
193 		uint16_t vlan_id, int on);
194 static int ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
195 			       enum rte_vlan_type vlan_type,
196 			       uint16_t tpid_id);
197 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
198 		uint16_t queue, bool on);
199 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
200 		int on);
201 static void ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev,
202 						  int mask);
203 static int ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask);
204 static int ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
205 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
206 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
207 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
208 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
209 
210 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
211 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
212 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
213 			       struct rte_eth_fc_conf *fc_conf);
214 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
215 			       struct rte_eth_fc_conf *fc_conf);
216 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
217 		struct rte_eth_pfc_conf *pfc_conf);
218 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
219 			struct rte_eth_rss_reta_entry64 *reta_conf,
220 			uint16_t reta_size);
221 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
222 			struct rte_eth_rss_reta_entry64 *reta_conf,
223 			uint16_t reta_size);
224 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
225 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
226 static int ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev);
227 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
228 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
229 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev);
230 static void ixgbe_dev_interrupt_handler(void *param);
231 static void ixgbe_dev_interrupt_delayed_handler(void *param);
232 static void *ixgbe_dev_setup_link_thread_handler(void *param);
233 static int ixgbe_dev_wait_setup_link_complete(struct rte_eth_dev *dev,
234 					      uint32_t timeout_ms);
235 
236 static int ixgbe_add_rar(struct rte_eth_dev *dev,
237 			struct rte_ether_addr *mac_addr,
238 			uint32_t index, uint32_t pool);
239 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
240 static int ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
241 					   struct rte_ether_addr *mac_addr);
242 static void ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config);
243 static bool is_device_supported(struct rte_eth_dev *dev,
244 				struct rte_pci_driver *drv);
245 
246 /* For Virtual Function support */
247 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
248 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
249 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
250 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
251 static int ixgbevf_dev_link_update(struct rte_eth_dev *dev,
252 				   int wait_to_complete);
253 static int ixgbevf_dev_stop(struct rte_eth_dev *dev);
254 static int ixgbevf_dev_close(struct rte_eth_dev *dev);
255 static int  ixgbevf_dev_reset(struct rte_eth_dev *dev);
256 static void ixgbevf_intr_disable(struct rte_eth_dev *dev);
257 static void ixgbevf_intr_enable(struct rte_eth_dev *dev);
258 static int ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
259 		struct rte_eth_stats *stats);
260 static int ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
261 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
262 		uint16_t vlan_id, int on);
263 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
264 		uint16_t queue, int on);
265 static int ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask);
266 static int ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
267 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
268 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
269 					    uint16_t queue_id);
270 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
271 					     uint16_t queue_id);
272 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
273 				 uint8_t queue, uint8_t msix_vector);
274 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
275 static int ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev);
276 static int ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev);
277 static int ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev);
278 static int ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev);
279 
280 /* For Eth VMDQ APIs support */
281 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
282 		rte_ether_addr * mac_addr, uint8_t on);
283 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on);
284 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
285 					  uint16_t queue_id);
286 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
287 					   uint16_t queue_id);
288 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
289 			       uint8_t queue, uint8_t msix_vector);
290 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
291 
292 static int ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
293 				struct rte_ether_addr *mac_addr,
294 				uint32_t index, uint32_t pool);
295 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
296 static int ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
297 					     struct rte_ether_addr *mac_addr);
298 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
299 			struct ixgbe_5tuple_filter *filter);
300 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
301 			struct ixgbe_5tuple_filter *filter);
302 static int ixgbe_dev_flow_ops_get(struct rte_eth_dev *dev,
303 				  const struct rte_flow_ops **ops);
304 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
305 
306 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
307 				      struct rte_ether_addr *mc_addr_set,
308 				      uint32_t nb_mc_addr);
309 static int ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
310 				   struct rte_eth_dcb_info *dcb_info);
311 
312 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
313 static int ixgbe_get_regs(struct rte_eth_dev *dev,
314 			    struct rte_dev_reg_info *regs);
315 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
316 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
317 				struct rte_dev_eeprom_info *eeprom);
318 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
319 				struct rte_dev_eeprom_info *eeprom);
320 
321 static int ixgbe_get_module_info(struct rte_eth_dev *dev,
322 				 struct rte_eth_dev_module_info *modinfo);
323 static int ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
324 				   struct rte_dev_eeprom_info *info);
325 
326 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
327 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
328 				struct rte_dev_reg_info *regs);
329 
330 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
331 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
332 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
333 					    struct timespec *timestamp,
334 					    uint32_t flags);
335 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
336 					    struct timespec *timestamp);
337 static int ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
338 static int ixgbe_timesync_read_time(struct rte_eth_dev *dev,
339 				   struct timespec *timestamp);
340 static int ixgbe_timesync_write_time(struct rte_eth_dev *dev,
341 				   const struct timespec *timestamp);
342 static void ixgbevf_dev_interrupt_handler(void *param);
343 
344 static int ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
345 					 struct rte_eth_udp_tunnel *udp_tunnel);
346 static int ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
347 					 struct rte_eth_udp_tunnel *udp_tunnel);
348 static int ixgbe_filter_restore(struct rte_eth_dev *dev);
349 static void ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev);
350 static int ixgbe_wait_for_link_up(struct ixgbe_hw *hw);
351 
352 /*
353  * Define VF Stats MACRO for Non "cleared on read" register
354  */
355 #define UPDATE_VF_STAT(reg, last, cur)                          \
356 {                                                               \
357 	uint32_t latest = IXGBE_READ_REG(hw, reg);              \
358 	cur += (latest - last) & UINT_MAX;                      \
359 	last = latest;                                          \
360 }
361 
362 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
363 {                                                                \
364 	u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
365 	u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
366 	u64 latest = ((new_msb << 32) | new_lsb);                \
367 	cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
368 	last = latest;                                           \
369 }
370 
371 #define IXGBE_SET_HWSTRIP(h, q) do {\
372 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
373 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
374 		(h)->bitmap[idx] |= 1 << bit;\
375 	} while (0)
376 
377 #define IXGBE_CLEAR_HWSTRIP(h, q) do {\
378 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
379 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
380 		(h)->bitmap[idx] &= ~(1 << bit);\
381 	} while (0)
382 
383 #define IXGBE_GET_HWSTRIP(h, q, r) do {\
384 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
385 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
386 		(r) = (h)->bitmap[idx] >> bit & 1;\
387 	} while (0)
388 
389 /*
390  * The set of PCI devices this driver supports
391  */
392 static const struct rte_pci_id pci_id_ixgbe_map[] = {
393 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598) },
394 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_BX) },
395 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT) },
396 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT) },
397 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT) },
398 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2) },
399 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM) },
400 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4) },
401 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT) },
402 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT) },
403 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM) },
404 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR) },
405 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4) },
406 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ) },
407 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KR) },
408 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE) },
409 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4) },
410 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP) },
411 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE) },
412 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE) },
413 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_EM) },
414 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2) },
415 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP) },
416 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP) },
417 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP) },
418 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM) },
419 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM) },
420 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T) },
421 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1) },
422 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP) },
423 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T) },
424 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_1G_T) },
425 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T) },
426 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1) },
427 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR) },
428 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR_L) },
429 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP_N) },
430 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII) },
431 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII_L) },
432 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_10G_T) },
433 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP) },
434 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP_N) },
435 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP) },
436 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T) },
437 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T_L) },
438 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4) },
439 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR) },
440 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_XFI) },
441 #ifdef RTE_LIBRTE_IXGBE_BYPASS
442 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BYPASS) },
443 #endif
444 	{ .vendor_id = 0, /* sentinel */ },
445 };
446 
447 /*
448  * The set of PCI devices this driver supports (for 82599 VF)
449  */
450 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
451 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF) },
452 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF_HV) },
453 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF) },
454 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF_HV) },
455 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF_HV) },
456 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF) },
457 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF) },
458 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF_HV) },
459 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF) },
460 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF_HV) },
461 	{ .vendor_id = 0, /* sentinel */ },
462 };
463 
464 static const struct rte_eth_desc_lim rx_desc_lim = {
465 	.nb_max = IXGBE_MAX_RING_DESC,
466 	.nb_min = IXGBE_MIN_RING_DESC,
467 	.nb_align = IXGBE_RXD_ALIGN,
468 };
469 
470 static const struct rte_eth_desc_lim tx_desc_lim = {
471 	.nb_max = IXGBE_MAX_RING_DESC,
472 	.nb_min = IXGBE_MIN_RING_DESC,
473 	.nb_align = IXGBE_TXD_ALIGN,
474 	.nb_seg_max = IXGBE_TX_MAX_SEG,
475 	.nb_mtu_seg_max = IXGBE_TX_MAX_SEG,
476 };
477 
478 static const struct eth_dev_ops ixgbe_eth_dev_ops = {
479 	.dev_configure        = ixgbe_dev_configure,
480 	.dev_start            = ixgbe_dev_start,
481 	.dev_stop             = ixgbe_dev_stop,
482 	.dev_set_link_up    = ixgbe_dev_set_link_up,
483 	.dev_set_link_down  = ixgbe_dev_set_link_down,
484 	.dev_close            = ixgbe_dev_close,
485 	.dev_reset	      = ixgbe_dev_reset,
486 	.promiscuous_enable   = ixgbe_dev_promiscuous_enable,
487 	.promiscuous_disable  = ixgbe_dev_promiscuous_disable,
488 	.allmulticast_enable  = ixgbe_dev_allmulticast_enable,
489 	.allmulticast_disable = ixgbe_dev_allmulticast_disable,
490 	.link_update          = ixgbe_dev_link_update,
491 	.stats_get            = ixgbe_dev_stats_get,
492 	.xstats_get           = ixgbe_dev_xstats_get,
493 	.xstats_get_by_id     = ixgbe_dev_xstats_get_by_id,
494 	.stats_reset          = ixgbe_dev_stats_reset,
495 	.xstats_reset         = ixgbe_dev_xstats_reset,
496 	.xstats_get_names     = ixgbe_dev_xstats_get_names,
497 	.xstats_get_names_by_id = ixgbe_dev_xstats_get_names_by_id,
498 	.queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
499 	.fw_version_get       = ixgbe_fw_version_get,
500 	.dev_infos_get        = ixgbe_dev_info_get,
501 	.dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
502 	.mtu_set              = ixgbe_dev_mtu_set,
503 	.vlan_filter_set      = ixgbe_vlan_filter_set,
504 	.vlan_tpid_set        = ixgbe_vlan_tpid_set,
505 	.vlan_offload_set     = ixgbe_vlan_offload_set,
506 	.vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
507 	.rx_queue_start	      = ixgbe_dev_rx_queue_start,
508 	.rx_queue_stop        = ixgbe_dev_rx_queue_stop,
509 	.tx_queue_start	      = ixgbe_dev_tx_queue_start,
510 	.tx_queue_stop        = ixgbe_dev_tx_queue_stop,
511 	.rx_queue_setup       = ixgbe_dev_rx_queue_setup,
512 	.rx_queue_intr_enable = ixgbe_dev_rx_queue_intr_enable,
513 	.rx_queue_intr_disable = ixgbe_dev_rx_queue_intr_disable,
514 	.rx_queue_release     = ixgbe_dev_rx_queue_release,
515 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
516 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
517 	.dev_led_on           = ixgbe_dev_led_on,
518 	.dev_led_off          = ixgbe_dev_led_off,
519 	.flow_ctrl_get        = ixgbe_flow_ctrl_get,
520 	.flow_ctrl_set        = ixgbe_flow_ctrl_set,
521 	.priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
522 	.mac_addr_add         = ixgbe_add_rar,
523 	.mac_addr_remove      = ixgbe_remove_rar,
524 	.mac_addr_set         = ixgbe_set_default_mac_addr,
525 	.uc_hash_table_set    = ixgbe_uc_hash_table_set,
526 	.uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
527 	.set_queue_rate_limit = ixgbe_set_queue_rate_limit,
528 	.reta_update          = ixgbe_dev_rss_reta_update,
529 	.reta_query           = ixgbe_dev_rss_reta_query,
530 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
531 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
532 	.flow_ops_get         = ixgbe_dev_flow_ops_get,
533 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
534 	.rxq_info_get         = ixgbe_rxq_info_get,
535 	.txq_info_get         = ixgbe_txq_info_get,
536 	.timesync_enable      = ixgbe_timesync_enable,
537 	.timesync_disable     = ixgbe_timesync_disable,
538 	.timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
539 	.timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
540 	.get_reg              = ixgbe_get_regs,
541 	.get_eeprom_length    = ixgbe_get_eeprom_length,
542 	.get_eeprom           = ixgbe_get_eeprom,
543 	.set_eeprom           = ixgbe_set_eeprom,
544 	.get_module_info      = ixgbe_get_module_info,
545 	.get_module_eeprom    = ixgbe_get_module_eeprom,
546 	.get_dcb_info         = ixgbe_dev_get_dcb_info,
547 	.timesync_adjust_time = ixgbe_timesync_adjust_time,
548 	.timesync_read_time   = ixgbe_timesync_read_time,
549 	.timesync_write_time  = ixgbe_timesync_write_time,
550 	.udp_tunnel_port_add  = ixgbe_dev_udp_tunnel_port_add,
551 	.udp_tunnel_port_del  = ixgbe_dev_udp_tunnel_port_del,
552 	.tm_ops_get           = ixgbe_tm_ops_get,
553 	.tx_done_cleanup      = ixgbe_dev_tx_done_cleanup,
554 	.get_monitor_addr     = ixgbe_get_monitor_addr,
555 };
556 
557 /*
558  * dev_ops for virtual function, bare necessities for basic vf
559  * operation have been implemented
560  */
561 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
562 	.dev_configure        = ixgbevf_dev_configure,
563 	.dev_start            = ixgbevf_dev_start,
564 	.dev_stop             = ixgbevf_dev_stop,
565 	.link_update          = ixgbevf_dev_link_update,
566 	.stats_get            = ixgbevf_dev_stats_get,
567 	.xstats_get           = ixgbevf_dev_xstats_get,
568 	.stats_reset          = ixgbevf_dev_stats_reset,
569 	.xstats_reset         = ixgbevf_dev_stats_reset,
570 	.xstats_get_names     = ixgbevf_dev_xstats_get_names,
571 	.dev_close            = ixgbevf_dev_close,
572 	.dev_reset	      = ixgbevf_dev_reset,
573 	.promiscuous_enable   = ixgbevf_dev_promiscuous_enable,
574 	.promiscuous_disable  = ixgbevf_dev_promiscuous_disable,
575 	.allmulticast_enable  = ixgbevf_dev_allmulticast_enable,
576 	.allmulticast_disable = ixgbevf_dev_allmulticast_disable,
577 	.dev_infos_get        = ixgbevf_dev_info_get,
578 	.dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
579 	.mtu_set              = ixgbevf_dev_set_mtu,
580 	.vlan_filter_set      = ixgbevf_vlan_filter_set,
581 	.vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
582 	.vlan_offload_set     = ixgbevf_vlan_offload_set,
583 	.rx_queue_setup       = ixgbe_dev_rx_queue_setup,
584 	.rx_queue_release     = ixgbe_dev_rx_queue_release,
585 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
586 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
587 	.rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
588 	.rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
589 	.mac_addr_add         = ixgbevf_add_mac_addr,
590 	.mac_addr_remove      = ixgbevf_remove_mac_addr,
591 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
592 	.rxq_info_get         = ixgbe_rxq_info_get,
593 	.txq_info_get         = ixgbe_txq_info_get,
594 	.mac_addr_set         = ixgbevf_set_default_mac_addr,
595 	.get_reg              = ixgbevf_get_regs,
596 	.reta_update          = ixgbe_dev_rss_reta_update,
597 	.reta_query           = ixgbe_dev_rss_reta_query,
598 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
599 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
600 	.tx_done_cleanup      = ixgbe_dev_tx_done_cleanup,
601 	.get_monitor_addr     = ixgbe_get_monitor_addr,
602 };
603 
604 /* store statistics names and its offset in stats structure */
605 struct rte_ixgbe_xstats_name_off {
606 	char name[RTE_ETH_XSTATS_NAME_SIZE];
607 	unsigned offset;
608 };
609 
610 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
611 	{"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
612 	{"rx_illegal_byte_errors", offsetof(struct ixgbe_hw_stats, illerrc)},
613 	{"rx_error_bytes", offsetof(struct ixgbe_hw_stats, errbc)},
614 	{"mac_local_errors", offsetof(struct ixgbe_hw_stats, mlfc)},
615 	{"mac_remote_errors", offsetof(struct ixgbe_hw_stats, mrfc)},
616 	{"rx_length_errors", offsetof(struct ixgbe_hw_stats, rlec)},
617 	{"tx_xon_packets", offsetof(struct ixgbe_hw_stats, lxontxc)},
618 	{"rx_xon_packets", offsetof(struct ixgbe_hw_stats, lxonrxc)},
619 	{"tx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxofftxc)},
620 	{"rx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
621 	{"rx_size_64_packets", offsetof(struct ixgbe_hw_stats, prc64)},
622 	{"rx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, prc127)},
623 	{"rx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, prc255)},
624 	{"rx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, prc511)},
625 	{"rx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
626 		prc1023)},
627 	{"rx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
628 		prc1522)},
629 	{"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
630 	{"rx_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
631 	{"rx_fragment_errors", offsetof(struct ixgbe_hw_stats, rfc)},
632 	{"rx_undersize_errors", offsetof(struct ixgbe_hw_stats, ruc)},
633 	{"rx_oversize_errors", offsetof(struct ixgbe_hw_stats, roc)},
634 	{"rx_jabber_errors", offsetof(struct ixgbe_hw_stats, rjc)},
635 	{"rx_management_packets", offsetof(struct ixgbe_hw_stats, mngprc)},
636 	{"rx_management_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
637 	{"tx_management_packets", offsetof(struct ixgbe_hw_stats, mngptc)},
638 	{"rx_total_packets", offsetof(struct ixgbe_hw_stats, tpr)},
639 	{"rx_total_bytes", offsetof(struct ixgbe_hw_stats, tor)},
640 	{"tx_total_packets", offsetof(struct ixgbe_hw_stats, tpt)},
641 	{"tx_size_64_packets", offsetof(struct ixgbe_hw_stats, ptc64)},
642 	{"tx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, ptc127)},
643 	{"tx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, ptc255)},
644 	{"tx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, ptc511)},
645 	{"tx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
646 		ptc1023)},
647 	{"tx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
648 		ptc1522)},
649 	{"tx_multicast_packets", offsetof(struct ixgbe_hw_stats, mptc)},
650 	{"tx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bptc)},
651 	{"rx_mac_short_packet_dropped", offsetof(struct ixgbe_hw_stats, mspdc)},
652 	{"rx_l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
653 
654 	{"flow_director_added_filters", offsetof(struct ixgbe_hw_stats,
655 		fdirustat_add)},
656 	{"flow_director_removed_filters", offsetof(struct ixgbe_hw_stats,
657 		fdirustat_remove)},
658 	{"flow_director_filter_add_errors", offsetof(struct ixgbe_hw_stats,
659 		fdirfstat_fadd)},
660 	{"flow_director_filter_remove_errors", offsetof(struct ixgbe_hw_stats,
661 		fdirfstat_fremove)},
662 	{"flow_director_matched_filters", offsetof(struct ixgbe_hw_stats,
663 		fdirmatch)},
664 	{"flow_director_missed_filters", offsetof(struct ixgbe_hw_stats,
665 		fdirmiss)},
666 
667 	{"rx_fcoe_crc_errors", offsetof(struct ixgbe_hw_stats, fccrc)},
668 	{"rx_fcoe_dropped", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
669 	{"rx_fcoe_mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats,
670 		fclast)},
671 	{"rx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeprc)},
672 	{"tx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeptc)},
673 	{"rx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwrc)},
674 	{"tx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwtc)},
675 	{"rx_fcoe_no_direct_data_placement", offsetof(struct ixgbe_hw_stats,
676 		fcoe_noddp)},
677 	{"rx_fcoe_no_direct_data_placement_ext_buff",
678 		offsetof(struct ixgbe_hw_stats, fcoe_noddp_ext_buff)},
679 
680 	{"tx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
681 		lxontxc)},
682 	{"rx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
683 		lxonrxc)},
684 	{"tx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
685 		lxofftxc)},
686 	{"rx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
687 		lxoffrxc)},
688 	{"rx_total_missed_packets", offsetof(struct ixgbe_hw_stats, mpctotal)},
689 };
690 
691 #define IXGBE_NB_HW_STATS (sizeof(rte_ixgbe_stats_strings) / \
692 			   sizeof(rte_ixgbe_stats_strings[0]))
693 
694 /* MACsec statistics */
695 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_macsec_strings[] = {
696 	{"out_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
697 		out_pkts_untagged)},
698 	{"out_pkts_encrypted", offsetof(struct ixgbe_macsec_stats,
699 		out_pkts_encrypted)},
700 	{"out_pkts_protected", offsetof(struct ixgbe_macsec_stats,
701 		out_pkts_protected)},
702 	{"out_octets_encrypted", offsetof(struct ixgbe_macsec_stats,
703 		out_octets_encrypted)},
704 	{"out_octets_protected", offsetof(struct ixgbe_macsec_stats,
705 		out_octets_protected)},
706 	{"in_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
707 		in_pkts_untagged)},
708 	{"in_pkts_badtag", offsetof(struct ixgbe_macsec_stats,
709 		in_pkts_badtag)},
710 	{"in_pkts_nosci", offsetof(struct ixgbe_macsec_stats,
711 		in_pkts_nosci)},
712 	{"in_pkts_unknownsci", offsetof(struct ixgbe_macsec_stats,
713 		in_pkts_unknownsci)},
714 	{"in_octets_decrypted", offsetof(struct ixgbe_macsec_stats,
715 		in_octets_decrypted)},
716 	{"in_octets_validated", offsetof(struct ixgbe_macsec_stats,
717 		in_octets_validated)},
718 	{"in_pkts_unchecked", offsetof(struct ixgbe_macsec_stats,
719 		in_pkts_unchecked)},
720 	{"in_pkts_delayed", offsetof(struct ixgbe_macsec_stats,
721 		in_pkts_delayed)},
722 	{"in_pkts_late", offsetof(struct ixgbe_macsec_stats,
723 		in_pkts_late)},
724 	{"in_pkts_ok", offsetof(struct ixgbe_macsec_stats,
725 		in_pkts_ok)},
726 	{"in_pkts_invalid", offsetof(struct ixgbe_macsec_stats,
727 		in_pkts_invalid)},
728 	{"in_pkts_notvalid", offsetof(struct ixgbe_macsec_stats,
729 		in_pkts_notvalid)},
730 	{"in_pkts_unusedsa", offsetof(struct ixgbe_macsec_stats,
731 		in_pkts_unusedsa)},
732 	{"in_pkts_notusingsa", offsetof(struct ixgbe_macsec_stats,
733 		in_pkts_notusingsa)},
734 };
735 
736 #define IXGBE_NB_MACSEC_STATS (sizeof(rte_ixgbe_macsec_strings) / \
737 			   sizeof(rte_ixgbe_macsec_strings[0]))
738 
739 /* Per-queue statistics */
740 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_rxq_strings[] = {
741 	{"mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats, rnbc)},
742 	{"dropped", offsetof(struct ixgbe_hw_stats, mpc)},
743 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxonrxc)},
744 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxoffrxc)},
745 };
746 
747 #define IXGBE_NB_RXQ_PRIO_STATS (sizeof(rte_ixgbe_rxq_strings) / \
748 			   sizeof(rte_ixgbe_rxq_strings[0]))
749 #define IXGBE_NB_RXQ_PRIO_VALUES 8
750 
751 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_txq_strings[] = {
752 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxontxc)},
753 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxofftxc)},
754 	{"xon_to_xoff_packets", offsetof(struct ixgbe_hw_stats,
755 		pxon2offc)},
756 };
757 
758 #define IXGBE_NB_TXQ_PRIO_STATS (sizeof(rte_ixgbe_txq_strings) / \
759 			   sizeof(rte_ixgbe_txq_strings[0]))
760 #define IXGBE_NB_TXQ_PRIO_VALUES 8
761 
762 static const struct rte_ixgbe_xstats_name_off rte_ixgbevf_stats_strings[] = {
763 	{"rx_multicast_packets", offsetof(struct ixgbevf_hw_stats, vfmprc)},
764 };
765 
766 #define IXGBEVF_NB_XSTATS (sizeof(rte_ixgbevf_stats_strings) /	\
767 		sizeof(rte_ixgbevf_stats_strings[0]))
768 
769 /*
770  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
771  */
772 static inline int
773 ixgbe_is_sfp(struct ixgbe_hw *hw)
774 {
775 	switch (hw->phy.type) {
776 	case ixgbe_phy_sfp_avago:
777 	case ixgbe_phy_sfp_ftl:
778 	case ixgbe_phy_sfp_intel:
779 	case ixgbe_phy_sfp_unknown:
780 	case ixgbe_phy_sfp_passive_tyco:
781 	case ixgbe_phy_sfp_passive_unknown:
782 		return 1;
783 	default:
784 		return 0;
785 	}
786 }
787 
788 static inline int32_t
789 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
790 {
791 	uint32_t ctrl_ext;
792 	int32_t status;
793 
794 	status = ixgbe_reset_hw(hw);
795 
796 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
797 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
798 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
799 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
800 	IXGBE_WRITE_FLUSH(hw);
801 
802 	if (status == IXGBE_ERR_SFP_NOT_PRESENT)
803 		status = IXGBE_SUCCESS;
804 	return status;
805 }
806 
807 static inline void
808 ixgbe_enable_intr(struct rte_eth_dev *dev)
809 {
810 	struct ixgbe_interrupt *intr =
811 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
812 	struct ixgbe_hw *hw =
813 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
814 
815 	IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
816 	IXGBE_WRITE_FLUSH(hw);
817 }
818 
819 /*
820  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
821  */
822 static void
823 ixgbe_disable_intr(struct ixgbe_hw *hw)
824 {
825 	PMD_INIT_FUNC_TRACE();
826 
827 	if (hw->mac.type == ixgbe_mac_82598EB) {
828 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
829 	} else {
830 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
831 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
832 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
833 	}
834 	IXGBE_WRITE_FLUSH(hw);
835 }
836 
837 /*
838  * This function resets queue statistics mapping registers.
839  * From Niantic datasheet, Initialization of Statistics section:
840  * "...if software requires the queue counters, the RQSMR and TQSM registers
841  * must be re-programmed following a device reset.
842  */
843 static void
844 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
845 {
846 	uint32_t i;
847 
848 	for (i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
849 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
850 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
851 	}
852 }
853 
854 
855 static int
856 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
857 				  uint16_t queue_id,
858 				  uint8_t stat_idx,
859 				  uint8_t is_rx)
860 {
861 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
862 #define NB_QMAP_FIELDS_PER_QSM_REG 4
863 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
864 
865 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
866 	struct ixgbe_stat_mapping_registers *stat_mappings =
867 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
868 	uint32_t qsmr_mask = 0;
869 	uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
870 	uint32_t q_map;
871 	uint8_t n, offset;
872 
873 	if ((hw->mac.type != ixgbe_mac_82599EB) &&
874 		(hw->mac.type != ixgbe_mac_X540) &&
875 		(hw->mac.type != ixgbe_mac_X550) &&
876 		(hw->mac.type != ixgbe_mac_X550EM_x) &&
877 		(hw->mac.type != ixgbe_mac_X550EM_a))
878 		return -ENOSYS;
879 
880 	PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
881 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
882 		     queue_id, stat_idx);
883 
884 	n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
885 	if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
886 		PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
887 		return -EIO;
888 	}
889 	offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
890 
891 	/* Now clear any previous stat_idx set */
892 	clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
893 	if (!is_rx)
894 		stat_mappings->tqsm[n] &= ~clearing_mask;
895 	else
896 		stat_mappings->rqsmr[n] &= ~clearing_mask;
897 
898 	q_map = (uint32_t)stat_idx;
899 	q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
900 	qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
901 	if (!is_rx)
902 		stat_mappings->tqsm[n] |= qsmr_mask;
903 	else
904 		stat_mappings->rqsmr[n] |= qsmr_mask;
905 
906 	PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
907 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
908 		     queue_id, stat_idx);
909 	PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
910 		     is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
911 
912 	/* Now write the mapping in the appropriate register */
913 	if (is_rx) {
914 		PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
915 			     stat_mappings->rqsmr[n], n);
916 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
917 	} else {
918 		PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
919 			     stat_mappings->tqsm[n], n);
920 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
921 	}
922 	return 0;
923 }
924 
925 static void
926 ixgbe_restore_statistics_mapping(struct rte_eth_dev *dev)
927 {
928 	struct ixgbe_stat_mapping_registers *stat_mappings =
929 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
930 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
931 	int i;
932 
933 	/* write whatever was in stat mapping table to the NIC */
934 	for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
935 		/* rx */
936 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
937 
938 		/* tx */
939 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
940 	}
941 }
942 
943 static void
944 ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config)
945 {
946 	uint8_t i;
947 	struct ixgbe_dcb_tc_config *tc;
948 	uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
949 
950 	dcb_config->num_tcs.pg_tcs = dcb_max_tc;
951 	dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
952 	for (i = 0; i < dcb_max_tc; i++) {
953 		tc = &dcb_config->tc_config[i];
954 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
955 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
956 				 (uint8_t)(100/dcb_max_tc + (i & 1));
957 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
958 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
959 				 (uint8_t)(100/dcb_max_tc + (i & 1));
960 		tc->pfc = ixgbe_dcb_pfc_disabled;
961 	}
962 
963 	/* Initialize default user to priority mapping, UPx->TC0 */
964 	tc = &dcb_config->tc_config[0];
965 	tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
966 	tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
967 	for (i = 0; i < IXGBE_DCB_MAX_BW_GROUP; i++) {
968 		dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
969 		dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
970 	}
971 	dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
972 	dcb_config->pfc_mode_enable = false;
973 	dcb_config->vt_mode = true;
974 	dcb_config->round_robin_enable = false;
975 	/* support all DCB capabilities in 82599 */
976 	dcb_config->support.capabilities = 0xFF;
977 
978 	/*we only support 4 Tcs for X540, X550 */
979 	if (hw->mac.type == ixgbe_mac_X540 ||
980 		hw->mac.type == ixgbe_mac_X550 ||
981 		hw->mac.type == ixgbe_mac_X550EM_x ||
982 		hw->mac.type == ixgbe_mac_X550EM_a) {
983 		dcb_config->num_tcs.pg_tcs = 4;
984 		dcb_config->num_tcs.pfc_tcs = 4;
985 	}
986 }
987 
988 /*
989  * Ensure that all locks are released before first NVM or PHY access
990  */
991 static void
992 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
993 {
994 	uint16_t mask;
995 
996 	/*
997 	 * Phy lock should not fail in this early stage. If this is the case,
998 	 * it is due to an improper exit of the application.
999 	 * So force the release of the faulty lock. Release of common lock
1000 	 * is done automatically by swfw_sync function.
1001 	 */
1002 	mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
1003 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1004 		PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
1005 	}
1006 	ixgbe_release_swfw_semaphore(hw, mask);
1007 
1008 	/*
1009 	 * These ones are more tricky since they are common to all ports; but
1010 	 * swfw_sync retries last long enough (1s) to be almost sure that if
1011 	 * lock can not be taken it is due to an improper lock of the
1012 	 * semaphore.
1013 	 */
1014 	mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
1015 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1016 		PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1017 	}
1018 	ixgbe_release_swfw_semaphore(hw, mask);
1019 }
1020 
1021 /*
1022  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
1023  * It returns 0 on success.
1024  */
1025 static int
1026 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params __rte_unused)
1027 {
1028 	struct ixgbe_adapter *ad = eth_dev->data->dev_private;
1029 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1030 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1031 	struct ixgbe_hw *hw =
1032 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1033 	struct ixgbe_vfta *shadow_vfta =
1034 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1035 	struct ixgbe_hwstrip *hwstrip =
1036 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1037 	struct ixgbe_dcb_config *dcb_config =
1038 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
1039 	struct ixgbe_filter_info *filter_info =
1040 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1041 	struct ixgbe_bw_conf *bw_conf =
1042 		IXGBE_DEV_PRIVATE_TO_BW_CONF(eth_dev->data->dev_private);
1043 	uint32_t ctrl_ext;
1044 	uint16_t csum;
1045 	int diag, i, ret;
1046 
1047 	PMD_INIT_FUNC_TRACE();
1048 
1049 	ixgbe_dev_macsec_setting_reset(eth_dev);
1050 
1051 	eth_dev->dev_ops = &ixgbe_eth_dev_ops;
1052 	eth_dev->rx_queue_count       = ixgbe_dev_rx_queue_count;
1053 	eth_dev->rx_descriptor_status = ixgbe_dev_rx_descriptor_status;
1054 	eth_dev->tx_descriptor_status = ixgbe_dev_tx_descriptor_status;
1055 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1056 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1057 	eth_dev->tx_pkt_prepare = &ixgbe_prep_pkts;
1058 
1059 	/*
1060 	 * For secondary processes, we don't initialise any further as primary
1061 	 * has already done this work. Only check we don't need a different
1062 	 * RX and TX function.
1063 	 */
1064 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1065 		struct ixgbe_tx_queue *txq;
1066 		/* TX queue function in primary, set by last queue initialized
1067 		 * Tx queue may not initialized by primary process
1068 		 */
1069 		if (eth_dev->data->tx_queues) {
1070 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
1071 			ixgbe_set_tx_function(eth_dev, txq);
1072 		} else {
1073 			/* Use default TX function if we get here */
1074 			PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
1075 				     "Using default TX function.");
1076 		}
1077 
1078 		ixgbe_set_rx_function(eth_dev);
1079 
1080 		return 0;
1081 	}
1082 
1083 	rte_atomic32_clear(&ad->link_thread_running);
1084 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1085 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1086 
1087 	/* Vendor and Device ID need to be set before init of shared code */
1088 	hw->device_id = pci_dev->id.device_id;
1089 	hw->vendor_id = pci_dev->id.vendor_id;
1090 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1091 	hw->allow_unsupported_sfp = 1;
1092 
1093 	/* Initialize the shared code (base driver) */
1094 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1095 	diag = ixgbe_bypass_init_shared_code(hw);
1096 #else
1097 	diag = ixgbe_init_shared_code(hw);
1098 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1099 
1100 	if (diag != IXGBE_SUCCESS) {
1101 		PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
1102 		return -EIO;
1103 	}
1104 
1105 	if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
1106 		PMD_INIT_LOG(ERR, "\nERROR: "
1107 			"Firmware recovery mode detected. Limiting functionality.\n"
1108 			"Refer to the Intel(R) Ethernet Adapters and Devices "
1109 			"User Guide for details on firmware recovery mode.");
1110 		return -EIO;
1111 	}
1112 
1113 	/* pick up the PCI bus settings for reporting later */
1114 	ixgbe_get_bus_info(hw);
1115 
1116 	/* Unlock any pending hardware semaphore */
1117 	ixgbe_swfw_lock_reset(hw);
1118 
1119 #ifdef RTE_LIB_SECURITY
1120 	/* Initialize security_ctx only for primary process*/
1121 	if (ixgbe_ipsec_ctx_create(eth_dev))
1122 		return -ENOMEM;
1123 #endif
1124 
1125 	/* Initialize DCB configuration*/
1126 	memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
1127 	ixgbe_dcb_init(hw, dcb_config);
1128 	/* Get Hardware Flow Control setting */
1129 	hw->fc.requested_mode = ixgbe_fc_none;
1130 	hw->fc.current_mode = ixgbe_fc_none;
1131 	hw->fc.pause_time = IXGBE_FC_PAUSE;
1132 	for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
1133 		hw->fc.low_water[i] = IXGBE_FC_LO;
1134 		hw->fc.high_water[i] = IXGBE_FC_HI;
1135 	}
1136 	hw->fc.send_xon = 1;
1137 
1138 	/* Make sure we have a good EEPROM before we read from it */
1139 	diag = ixgbe_validate_eeprom_checksum(hw, &csum);
1140 	if (diag != IXGBE_SUCCESS) {
1141 		PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
1142 		return -EIO;
1143 	}
1144 
1145 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1146 	diag = ixgbe_bypass_init_hw(hw);
1147 #else
1148 	diag = ixgbe_init_hw(hw);
1149 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1150 
1151 	/*
1152 	 * Devices with copper phys will fail to initialise if ixgbe_init_hw()
1153 	 * is called too soon after the kernel driver unbinding/binding occurs.
1154 	 * The failure occurs in ixgbe_identify_phy_generic() for all devices,
1155 	 * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
1156 	 * also called. See ixgbe_identify_phy_82599(). The reason for the
1157 	 * failure is not known, and only occuts when virtualisation features
1158 	 * are disabled in the bios. A delay of 100ms  was found to be enough by
1159 	 * trial-and-error, and is doubled to be safe.
1160 	 */
1161 	if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
1162 		rte_delay_ms(200);
1163 		diag = ixgbe_init_hw(hw);
1164 	}
1165 
1166 	if (diag == IXGBE_ERR_SFP_NOT_PRESENT)
1167 		diag = IXGBE_SUCCESS;
1168 
1169 	if (diag == IXGBE_ERR_EEPROM_VERSION) {
1170 		PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
1171 			     "LOM.  Please be aware there may be issues associated "
1172 			     "with your hardware.");
1173 		PMD_INIT_LOG(ERR, "If you are experiencing problems "
1174 			     "please contact your Intel or hardware representative "
1175 			     "who provided you with this hardware.");
1176 	} else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
1177 		PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
1178 	if (diag) {
1179 		PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
1180 		return -EIO;
1181 	}
1182 
1183 	/* Reset the hw statistics */
1184 	ixgbe_dev_stats_reset(eth_dev);
1185 
1186 	/* disable interrupt */
1187 	ixgbe_disable_intr(hw);
1188 
1189 	/* reset mappings for queue statistics hw counters*/
1190 	ixgbe_reset_qstat_mappings(hw);
1191 
1192 	/* Allocate memory for storing MAC addresses */
1193 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", RTE_ETHER_ADDR_LEN *
1194 					       hw->mac.num_rar_entries, 0);
1195 	if (eth_dev->data->mac_addrs == NULL) {
1196 		PMD_INIT_LOG(ERR,
1197 			     "Failed to allocate %u bytes needed to store "
1198 			     "MAC addresses",
1199 			     RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1200 		return -ENOMEM;
1201 	}
1202 	/* Copy the permanent MAC address */
1203 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.perm_addr,
1204 			&eth_dev->data->mac_addrs[0]);
1205 
1206 	/* Allocate memory for storing hash filter MAC addresses */
1207 	eth_dev->data->hash_mac_addrs = rte_zmalloc(
1208 		"ixgbe", RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC, 0);
1209 	if (eth_dev->data->hash_mac_addrs == NULL) {
1210 		PMD_INIT_LOG(ERR,
1211 			     "Failed to allocate %d bytes needed to store MAC addresses",
1212 			     RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
1213 		rte_free(eth_dev->data->mac_addrs);
1214 		eth_dev->data->mac_addrs = NULL;
1215 		return -ENOMEM;
1216 	}
1217 
1218 	/* initialize the vfta */
1219 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1220 
1221 	/* initialize the hw strip bitmap*/
1222 	memset(hwstrip, 0, sizeof(*hwstrip));
1223 
1224 	/* initialize PF if max_vfs not zero */
1225 	ret = ixgbe_pf_host_init(eth_dev);
1226 	if (ret)
1227 		goto err_pf_host_init;
1228 
1229 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1230 	/* let hardware know driver is loaded */
1231 	ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1232 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1233 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1234 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1235 	IXGBE_WRITE_FLUSH(hw);
1236 
1237 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1238 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1239 			     (int) hw->mac.type, (int) hw->phy.type,
1240 			     (int) hw->phy.sfp_type);
1241 	else
1242 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1243 			     (int) hw->mac.type, (int) hw->phy.type);
1244 
1245 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1246 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1247 		     pci_dev->id.device_id);
1248 
1249 	rte_intr_callback_register(intr_handle,
1250 				   ixgbe_dev_interrupt_handler, eth_dev);
1251 
1252 	/* enable uio/vfio intr/eventfd mapping */
1253 	rte_intr_enable(intr_handle);
1254 
1255 	/* enable support intr */
1256 	ixgbe_enable_intr(eth_dev);
1257 
1258 	/* initialize filter info */
1259 	memset(filter_info, 0,
1260 	       sizeof(struct ixgbe_filter_info));
1261 
1262 	/* initialize 5tuple filter list */
1263 	TAILQ_INIT(&filter_info->fivetuple_list);
1264 
1265 	/* initialize flow director filter list & hash */
1266 	ret = ixgbe_fdir_filter_init(eth_dev);
1267 	if (ret)
1268 		goto err_fdir_filter_init;
1269 
1270 	/* initialize l2 tunnel filter list & hash */
1271 	ret = ixgbe_l2_tn_filter_init(eth_dev);
1272 	if (ret)
1273 		goto err_l2_tn_filter_init;
1274 
1275 	/* initialize flow filter lists */
1276 	ixgbe_filterlist_init();
1277 
1278 	/* initialize bandwidth configuration info */
1279 	memset(bw_conf, 0, sizeof(struct ixgbe_bw_conf));
1280 
1281 	/* initialize Traffic Manager configuration */
1282 	ixgbe_tm_conf_init(eth_dev);
1283 
1284 	return 0;
1285 
1286 err_l2_tn_filter_init:
1287 	ixgbe_fdir_filter_uninit(eth_dev);
1288 err_fdir_filter_init:
1289 	ixgbe_disable_intr(hw);
1290 	rte_intr_disable(intr_handle);
1291 	rte_intr_callback_unregister(intr_handle,
1292 		ixgbe_dev_interrupt_handler, eth_dev);
1293 	ixgbe_pf_host_uninit(eth_dev);
1294 err_pf_host_init:
1295 	rte_free(eth_dev->data->mac_addrs);
1296 	eth_dev->data->mac_addrs = NULL;
1297 	rte_free(eth_dev->data->hash_mac_addrs);
1298 	eth_dev->data->hash_mac_addrs = NULL;
1299 	return ret;
1300 }
1301 
1302 static int
1303 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1304 {
1305 	PMD_INIT_FUNC_TRACE();
1306 
1307 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1308 		return 0;
1309 
1310 	ixgbe_dev_close(eth_dev);
1311 
1312 	return 0;
1313 }
1314 
1315 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
1316 {
1317 	struct ixgbe_filter_info *filter_info =
1318 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1319 	struct ixgbe_5tuple_filter *p_5tuple;
1320 
1321 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
1322 		TAILQ_REMOVE(&filter_info->fivetuple_list,
1323 			     p_5tuple,
1324 			     entries);
1325 		rte_free(p_5tuple);
1326 	}
1327 	memset(filter_info->fivetuple_mask, 0,
1328 	       sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1329 
1330 	return 0;
1331 }
1332 
1333 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev)
1334 {
1335 	struct ixgbe_hw_fdir_info *fdir_info =
1336 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1337 	struct ixgbe_fdir_filter *fdir_filter;
1338 
1339 	rte_free(fdir_info->hash_map);
1340 	rte_hash_free(fdir_info->hash_handle);
1341 
1342 	while ((fdir_filter = TAILQ_FIRST(&fdir_info->fdir_list))) {
1343 		TAILQ_REMOVE(&fdir_info->fdir_list,
1344 			     fdir_filter,
1345 			     entries);
1346 		rte_free(fdir_filter);
1347 	}
1348 
1349 	return 0;
1350 }
1351 
1352 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev)
1353 {
1354 	struct ixgbe_l2_tn_info *l2_tn_info =
1355 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1356 	struct ixgbe_l2_tn_filter *l2_tn_filter;
1357 
1358 	rte_free(l2_tn_info->hash_map);
1359 	rte_hash_free(l2_tn_info->hash_handle);
1360 
1361 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
1362 		TAILQ_REMOVE(&l2_tn_info->l2_tn_list,
1363 			     l2_tn_filter,
1364 			     entries);
1365 		rte_free(l2_tn_filter);
1366 	}
1367 
1368 	return 0;
1369 }
1370 
1371 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev)
1372 {
1373 	struct ixgbe_hw_fdir_info *fdir_info =
1374 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1375 	char fdir_hash_name[RTE_HASH_NAMESIZE];
1376 	struct rte_hash_parameters fdir_hash_params = {
1377 		.name = fdir_hash_name,
1378 		.entries = IXGBE_MAX_FDIR_FILTER_NUM,
1379 		.key_len = sizeof(union ixgbe_atr_input),
1380 		.hash_func = rte_hash_crc,
1381 		.hash_func_init_val = 0,
1382 		.socket_id = rte_socket_id(),
1383 	};
1384 
1385 	TAILQ_INIT(&fdir_info->fdir_list);
1386 	snprintf(fdir_hash_name, RTE_HASH_NAMESIZE,
1387 		 "fdir_%s", eth_dev->device->name);
1388 	fdir_info->hash_handle = rte_hash_create(&fdir_hash_params);
1389 	if (!fdir_info->hash_handle) {
1390 		PMD_INIT_LOG(ERR, "Failed to create fdir hash table!");
1391 		return -EINVAL;
1392 	}
1393 	fdir_info->hash_map = rte_zmalloc("ixgbe",
1394 					  sizeof(struct ixgbe_fdir_filter *) *
1395 					  IXGBE_MAX_FDIR_FILTER_NUM,
1396 					  0);
1397 	if (!fdir_info->hash_map) {
1398 		PMD_INIT_LOG(ERR,
1399 			     "Failed to allocate memory for fdir hash map!");
1400 		rte_hash_free(fdir_info->hash_handle);
1401 		return -ENOMEM;
1402 	}
1403 	fdir_info->mask_added = FALSE;
1404 
1405 	return 0;
1406 }
1407 
1408 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev)
1409 {
1410 	struct ixgbe_l2_tn_info *l2_tn_info =
1411 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1412 	char l2_tn_hash_name[RTE_HASH_NAMESIZE];
1413 	struct rte_hash_parameters l2_tn_hash_params = {
1414 		.name = l2_tn_hash_name,
1415 		.entries = IXGBE_MAX_L2_TN_FILTER_NUM,
1416 		.key_len = sizeof(struct ixgbe_l2_tn_key),
1417 		.hash_func = rte_hash_crc,
1418 		.hash_func_init_val = 0,
1419 		.socket_id = rte_socket_id(),
1420 	};
1421 
1422 	TAILQ_INIT(&l2_tn_info->l2_tn_list);
1423 	snprintf(l2_tn_hash_name, RTE_HASH_NAMESIZE,
1424 		 "l2_tn_%s", eth_dev->device->name);
1425 	l2_tn_info->hash_handle = rte_hash_create(&l2_tn_hash_params);
1426 	if (!l2_tn_info->hash_handle) {
1427 		PMD_INIT_LOG(ERR, "Failed to create L2 TN hash table!");
1428 		return -EINVAL;
1429 	}
1430 	l2_tn_info->hash_map = rte_zmalloc("ixgbe",
1431 				   sizeof(struct ixgbe_l2_tn_filter *) *
1432 				   IXGBE_MAX_L2_TN_FILTER_NUM,
1433 				   0);
1434 	if (!l2_tn_info->hash_map) {
1435 		PMD_INIT_LOG(ERR,
1436 			"Failed to allocate memory for L2 TN hash map!");
1437 		rte_hash_free(l2_tn_info->hash_handle);
1438 		return -ENOMEM;
1439 	}
1440 	l2_tn_info->e_tag_en = FALSE;
1441 	l2_tn_info->e_tag_fwd_en = FALSE;
1442 	l2_tn_info->e_tag_ether_type = RTE_ETHER_TYPE_ETAG;
1443 
1444 	return 0;
1445 }
1446 /*
1447  * Negotiate mailbox API version with the PF.
1448  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1449  * Then we try to negotiate starting with the most recent one.
1450  * If all negotiation attempts fail, then we will proceed with
1451  * the default one (ixgbe_mbox_api_10).
1452  */
1453 static void
1454 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1455 {
1456 	int32_t i;
1457 
1458 	/* start with highest supported, proceed down */
1459 	static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1460 		ixgbe_mbox_api_13,
1461 		ixgbe_mbox_api_12,
1462 		ixgbe_mbox_api_11,
1463 		ixgbe_mbox_api_10,
1464 	};
1465 
1466 	for (i = 0;
1467 			i != RTE_DIM(sup_ver) &&
1468 			ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1469 			i++)
1470 		;
1471 }
1472 
1473 static void
1474 generate_random_mac_addr(struct rte_ether_addr *mac_addr)
1475 {
1476 	uint64_t random;
1477 
1478 	/* Set Organizationally Unique Identifier (OUI) prefix. */
1479 	mac_addr->addr_bytes[0] = 0x00;
1480 	mac_addr->addr_bytes[1] = 0x09;
1481 	mac_addr->addr_bytes[2] = 0xC0;
1482 	/* Force indication of locally assigned MAC address. */
1483 	mac_addr->addr_bytes[0] |= RTE_ETHER_LOCAL_ADMIN_ADDR;
1484 	/* Generate the last 3 bytes of the MAC address with a random number. */
1485 	random = rte_rand();
1486 	memcpy(&mac_addr->addr_bytes[3], &random, 3);
1487 }
1488 
1489 static int
1490 devarg_handle_int(__rte_unused const char *key, const char *value,
1491 		  void *extra_args)
1492 {
1493 	uint16_t *n = extra_args;
1494 
1495 	if (value == NULL || extra_args == NULL)
1496 		return -EINVAL;
1497 
1498 	*n = (uint16_t)strtoul(value, NULL, 0);
1499 	if (*n == USHRT_MAX && errno == ERANGE)
1500 		return -1;
1501 
1502 	return 0;
1503 }
1504 
1505 static void
1506 ixgbevf_parse_devargs(struct ixgbe_adapter *adapter,
1507 		      struct rte_devargs *devargs)
1508 {
1509 	struct rte_kvargs *kvlist;
1510 	uint16_t pflink_fullchk;
1511 
1512 	if (devargs == NULL)
1513 		return;
1514 
1515 	kvlist = rte_kvargs_parse(devargs->args, ixgbevf_valid_arguments);
1516 	if (kvlist == NULL)
1517 		return;
1518 
1519 	if (rte_kvargs_count(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK) == 1 &&
1520 	    rte_kvargs_process(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK,
1521 			       devarg_handle_int, &pflink_fullchk) == 0 &&
1522 	    pflink_fullchk == 1)
1523 		adapter->pflink_fullchk = 1;
1524 
1525 	rte_kvargs_free(kvlist);
1526 }
1527 
1528 /*
1529  * Virtual Function device init
1530  */
1531 static int
1532 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1533 {
1534 	int diag;
1535 	uint32_t tc, tcs;
1536 	struct ixgbe_adapter *ad = eth_dev->data->dev_private;
1537 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1538 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1539 	struct ixgbe_hw *hw =
1540 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1541 	struct ixgbe_vfta *shadow_vfta =
1542 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1543 	struct ixgbe_hwstrip *hwstrip =
1544 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1545 	struct rte_ether_addr *perm_addr =
1546 		(struct rte_ether_addr *)hw->mac.perm_addr;
1547 
1548 	PMD_INIT_FUNC_TRACE();
1549 
1550 	eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1551 	eth_dev->rx_descriptor_status = ixgbe_dev_rx_descriptor_status;
1552 	eth_dev->tx_descriptor_status = ixgbe_dev_tx_descriptor_status;
1553 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1554 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1555 
1556 	/* for secondary processes, we don't initialise any further as primary
1557 	 * has already done this work. Only check we don't need a different
1558 	 * RX function
1559 	 */
1560 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1561 		struct ixgbe_tx_queue *txq;
1562 		/* TX queue function in primary, set by last queue initialized
1563 		 * Tx queue may not initialized by primary process
1564 		 */
1565 		if (eth_dev->data->tx_queues) {
1566 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1567 			ixgbe_set_tx_function(eth_dev, txq);
1568 		} else {
1569 			/* Use default TX function if we get here */
1570 			PMD_INIT_LOG(NOTICE,
1571 				     "No TX queues configured yet. Using default TX function.");
1572 		}
1573 
1574 		ixgbe_set_rx_function(eth_dev);
1575 
1576 		return 0;
1577 	}
1578 
1579 	rte_atomic32_clear(&ad->link_thread_running);
1580 	ixgbevf_parse_devargs(eth_dev->data->dev_private,
1581 			      pci_dev->device.devargs);
1582 
1583 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1584 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1585 
1586 	hw->device_id = pci_dev->id.device_id;
1587 	hw->vendor_id = pci_dev->id.vendor_id;
1588 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1589 
1590 	/* initialize the vfta */
1591 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1592 
1593 	/* initialize the hw strip bitmap*/
1594 	memset(hwstrip, 0, sizeof(*hwstrip));
1595 
1596 	/* Initialize the shared code (base driver) */
1597 	diag = ixgbe_init_shared_code(hw);
1598 	if (diag != IXGBE_SUCCESS) {
1599 		PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1600 		return -EIO;
1601 	}
1602 
1603 	/* init_mailbox_params */
1604 	hw->mbx.ops.init_params(hw);
1605 
1606 	/* Reset the hw statistics */
1607 	ixgbevf_dev_stats_reset(eth_dev);
1608 
1609 	/* Disable the interrupts for VF */
1610 	ixgbevf_intr_disable(eth_dev);
1611 
1612 	hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1613 	diag = hw->mac.ops.reset_hw(hw);
1614 
1615 	/*
1616 	 * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1617 	 * the underlying PF driver has not assigned a MAC address to the VF.
1618 	 * In this case, assign a random MAC address.
1619 	 */
1620 	if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1621 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1622 		/*
1623 		 * This error code will be propagated to the app by
1624 		 * rte_eth_dev_reset, so use a public error code rather than
1625 		 * the internal-only IXGBE_ERR_RESET_FAILED
1626 		 */
1627 		return -EAGAIN;
1628 	}
1629 
1630 	/* negotiate mailbox API version to use with the PF. */
1631 	ixgbevf_negotiate_api(hw);
1632 
1633 	/* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1634 	ixgbevf_get_queues(hw, &tcs, &tc);
1635 
1636 	/* Allocate memory for storing MAC addresses */
1637 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", RTE_ETHER_ADDR_LEN *
1638 					       hw->mac.num_rar_entries, 0);
1639 	if (eth_dev->data->mac_addrs == NULL) {
1640 		PMD_INIT_LOG(ERR,
1641 			     "Failed to allocate %u bytes needed to store "
1642 			     "MAC addresses",
1643 			     RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1644 		return -ENOMEM;
1645 	}
1646 
1647 	/* Generate a random MAC address, if none was assigned by PF. */
1648 	if (rte_is_zero_ether_addr(perm_addr)) {
1649 		generate_random_mac_addr(perm_addr);
1650 		diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1651 		if (diag) {
1652 			rte_free(eth_dev->data->mac_addrs);
1653 			eth_dev->data->mac_addrs = NULL;
1654 			return diag;
1655 		}
1656 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1657 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1658 			     RTE_ETHER_ADDR_PRT_FMT,
1659 				 RTE_ETHER_ADDR_BYTES(perm_addr));
1660 	}
1661 
1662 	/* Copy the permanent MAC address */
1663 	rte_ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1664 
1665 	/* reset the hardware with the new settings */
1666 	diag = hw->mac.ops.start_hw(hw);
1667 	switch (diag) {
1668 	case  0:
1669 		break;
1670 
1671 	default:
1672 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1673 		rte_free(eth_dev->data->mac_addrs);
1674 		eth_dev->data->mac_addrs = NULL;
1675 		return -EIO;
1676 	}
1677 
1678 	rte_intr_callback_register(intr_handle,
1679 				   ixgbevf_dev_interrupt_handler, eth_dev);
1680 	rte_intr_enable(intr_handle);
1681 	ixgbevf_intr_enable(eth_dev);
1682 
1683 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1684 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1685 		     pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1686 
1687 	return 0;
1688 }
1689 
1690 /* Virtual Function device uninit */
1691 
1692 static int
1693 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1694 {
1695 	PMD_INIT_FUNC_TRACE();
1696 
1697 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1698 		return 0;
1699 
1700 	ixgbevf_dev_close(eth_dev);
1701 
1702 	return 0;
1703 }
1704 
1705 static int
1706 eth_ixgbe_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1707 		struct rte_pci_device *pci_dev)
1708 {
1709 	char name[RTE_ETH_NAME_MAX_LEN];
1710 	struct rte_eth_dev *pf_ethdev;
1711 	struct rte_eth_devargs eth_da;
1712 	int i, retval;
1713 
1714 	if (pci_dev->device.devargs) {
1715 		retval = rte_eth_devargs_parse(pci_dev->device.devargs->args,
1716 				&eth_da);
1717 		if (retval)
1718 			return retval;
1719 	} else
1720 		memset(&eth_da, 0, sizeof(eth_da));
1721 
1722 	if (eth_da.nb_representor_ports > 0 &&
1723 	    eth_da.type != RTE_ETH_REPRESENTOR_VF) {
1724 		PMD_DRV_LOG(ERR, "unsupported representor type: %s\n",
1725 			    pci_dev->device.devargs->args);
1726 		return -ENOTSUP;
1727 	}
1728 
1729 	retval = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
1730 		sizeof(struct ixgbe_adapter),
1731 		eth_dev_pci_specific_init, pci_dev,
1732 		eth_ixgbe_dev_init, NULL);
1733 
1734 	if (retval || eth_da.nb_representor_ports < 1)
1735 		return retval;
1736 
1737 	pf_ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1738 	if (pf_ethdev == NULL)
1739 		return -ENODEV;
1740 
1741 	/* probe VF representor ports */
1742 	for (i = 0; i < eth_da.nb_representor_ports; i++) {
1743 		struct ixgbe_vf_info *vfinfo;
1744 		struct ixgbe_vf_representor representor;
1745 
1746 		vfinfo = *IXGBE_DEV_PRIVATE_TO_P_VFDATA(
1747 			pf_ethdev->data->dev_private);
1748 		if (vfinfo == NULL) {
1749 			PMD_DRV_LOG(ERR,
1750 				"no virtual functions supported by PF");
1751 			break;
1752 		}
1753 
1754 		representor.vf_id = eth_da.representor_ports[i];
1755 		representor.switch_domain_id = vfinfo->switch_domain_id;
1756 		representor.pf_ethdev = pf_ethdev;
1757 
1758 		/* representor port net_bdf_port */
1759 		snprintf(name, sizeof(name), "net_%s_representor_%d",
1760 			pci_dev->device.name,
1761 			eth_da.representor_ports[i]);
1762 
1763 		retval = rte_eth_dev_create(&pci_dev->device, name,
1764 			sizeof(struct ixgbe_vf_representor), NULL, NULL,
1765 			ixgbe_vf_representor_init, &representor);
1766 
1767 		if (retval)
1768 			PMD_DRV_LOG(ERR, "failed to create ixgbe vf "
1769 				"representor %s.", name);
1770 	}
1771 
1772 	return 0;
1773 }
1774 
1775 static int eth_ixgbe_pci_remove(struct rte_pci_device *pci_dev)
1776 {
1777 	struct rte_eth_dev *ethdev;
1778 
1779 	ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1780 	if (!ethdev)
1781 		return 0;
1782 
1783 	if (ethdev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
1784 		return rte_eth_dev_pci_generic_remove(pci_dev,
1785 					ixgbe_vf_representor_uninit);
1786 	else
1787 		return rte_eth_dev_pci_generic_remove(pci_dev,
1788 						eth_ixgbe_dev_uninit);
1789 }
1790 
1791 static struct rte_pci_driver rte_ixgbe_pmd = {
1792 	.id_table = pci_id_ixgbe_map,
1793 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1794 	.probe = eth_ixgbe_pci_probe,
1795 	.remove = eth_ixgbe_pci_remove,
1796 };
1797 
1798 static int eth_ixgbevf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1799 	struct rte_pci_device *pci_dev)
1800 {
1801 	return rte_eth_dev_pci_generic_probe(pci_dev,
1802 		sizeof(struct ixgbe_adapter), eth_ixgbevf_dev_init);
1803 }
1804 
1805 static int eth_ixgbevf_pci_remove(struct rte_pci_device *pci_dev)
1806 {
1807 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbevf_dev_uninit);
1808 }
1809 
1810 /*
1811  * virtual function driver struct
1812  */
1813 static struct rte_pci_driver rte_ixgbevf_pmd = {
1814 	.id_table = pci_id_ixgbevf_map,
1815 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1816 	.probe = eth_ixgbevf_pci_probe,
1817 	.remove = eth_ixgbevf_pci_remove,
1818 };
1819 
1820 static int
1821 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1822 {
1823 	struct ixgbe_hw *hw =
1824 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1825 	struct ixgbe_vfta *shadow_vfta =
1826 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1827 	uint32_t vfta;
1828 	uint32_t vid_idx;
1829 	uint32_t vid_bit;
1830 
1831 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1832 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1833 	vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1834 	if (on)
1835 		vfta |= vid_bit;
1836 	else
1837 		vfta &= ~vid_bit;
1838 	IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1839 
1840 	/* update local VFTA copy */
1841 	shadow_vfta->vfta[vid_idx] = vfta;
1842 
1843 	return 0;
1844 }
1845 
1846 static void
1847 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1848 {
1849 	if (on)
1850 		ixgbe_vlan_hw_strip_enable(dev, queue);
1851 	else
1852 		ixgbe_vlan_hw_strip_disable(dev, queue);
1853 }
1854 
1855 static int
1856 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1857 		    enum rte_vlan_type vlan_type,
1858 		    uint16_t tpid)
1859 {
1860 	struct ixgbe_hw *hw =
1861 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1862 	int ret = 0;
1863 	uint32_t reg;
1864 	uint32_t qinq;
1865 
1866 	qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1867 	qinq &= IXGBE_DMATXCTL_GDV;
1868 
1869 	switch (vlan_type) {
1870 	case RTE_ETH_VLAN_TYPE_INNER:
1871 		if (qinq) {
1872 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1873 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1874 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1875 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1876 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1877 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1878 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1879 		} else {
1880 			ret = -ENOTSUP;
1881 			PMD_DRV_LOG(ERR, "Inner type is not supported"
1882 				    " by single VLAN");
1883 		}
1884 		break;
1885 	case RTE_ETH_VLAN_TYPE_OUTER:
1886 		if (qinq) {
1887 			/* Only the high 16-bits is valid */
1888 			IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1889 					IXGBE_EXVET_VET_EXT_SHIFT);
1890 		} else {
1891 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1892 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1893 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1894 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1895 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1896 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1897 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1898 		}
1899 
1900 		break;
1901 	default:
1902 		ret = -EINVAL;
1903 		PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1904 		break;
1905 	}
1906 
1907 	return ret;
1908 }
1909 
1910 void
1911 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1912 {
1913 	struct ixgbe_hw *hw =
1914 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1915 	uint32_t vlnctrl;
1916 
1917 	PMD_INIT_FUNC_TRACE();
1918 
1919 	/* Filter Table Disable */
1920 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1921 	vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1922 
1923 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1924 }
1925 
1926 void
1927 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1928 {
1929 	struct ixgbe_hw *hw =
1930 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1931 	struct ixgbe_vfta *shadow_vfta =
1932 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1933 	uint32_t vlnctrl;
1934 	uint16_t i;
1935 
1936 	PMD_INIT_FUNC_TRACE();
1937 
1938 	/* Filter Table Enable */
1939 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1940 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1941 	vlnctrl |= IXGBE_VLNCTRL_VFE;
1942 
1943 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1944 
1945 	/* write whatever is in local vfta copy */
1946 	for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1947 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1948 }
1949 
1950 static void
1951 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1952 {
1953 	struct ixgbe_hwstrip *hwstrip =
1954 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1955 	struct ixgbe_rx_queue *rxq;
1956 
1957 	if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
1958 		return;
1959 
1960 	if (on)
1961 		IXGBE_SET_HWSTRIP(hwstrip, queue);
1962 	else
1963 		IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1964 
1965 	if (queue >= dev->data->nb_rx_queues)
1966 		return;
1967 
1968 	rxq = dev->data->rx_queues[queue];
1969 
1970 	if (on) {
1971 		rxq->vlan_flags = RTE_MBUF_F_RX_VLAN | RTE_MBUF_F_RX_VLAN_STRIPPED;
1972 		rxq->offloads |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
1973 	} else {
1974 		rxq->vlan_flags = RTE_MBUF_F_RX_VLAN;
1975 		rxq->offloads &= ~RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
1976 	}
1977 }
1978 
1979 static void
1980 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
1981 {
1982 	struct ixgbe_hw *hw =
1983 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1984 	uint32_t ctrl;
1985 
1986 	PMD_INIT_FUNC_TRACE();
1987 
1988 	if (hw->mac.type == ixgbe_mac_82598EB) {
1989 		/* No queue level support */
1990 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1991 		return;
1992 	}
1993 
1994 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1995 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1996 	ctrl &= ~IXGBE_RXDCTL_VME;
1997 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1998 
1999 	/* record those setting for HW strip per queue */
2000 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
2001 }
2002 
2003 static void
2004 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
2005 {
2006 	struct ixgbe_hw *hw =
2007 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2008 	uint32_t ctrl;
2009 
2010 	PMD_INIT_FUNC_TRACE();
2011 
2012 	if (hw->mac.type == ixgbe_mac_82598EB) {
2013 		/* No queue level supported */
2014 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2015 		return;
2016 	}
2017 
2018 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2019 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2020 	ctrl |= IXGBE_RXDCTL_VME;
2021 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2022 
2023 	/* record those setting for HW strip per queue */
2024 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
2025 }
2026 
2027 static void
2028 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2029 {
2030 	struct ixgbe_hw *hw =
2031 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2032 	uint32_t ctrl;
2033 
2034 	PMD_INIT_FUNC_TRACE();
2035 
2036 	/* DMATXCTRL: Geric Double VLAN Disable */
2037 	ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2038 	ctrl &= ~IXGBE_DMATXCTL_GDV;
2039 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2040 
2041 	/* CTRL_EXT: Global Double VLAN Disable */
2042 	ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2043 	ctrl &= ~IXGBE_EXTENDED_VLAN;
2044 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2045 
2046 }
2047 
2048 static void
2049 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2050 {
2051 	struct ixgbe_hw *hw =
2052 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2053 	uint32_t ctrl;
2054 
2055 	PMD_INIT_FUNC_TRACE();
2056 
2057 	/* DMATXCTRL: Geric Double VLAN Enable */
2058 	ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2059 	ctrl |= IXGBE_DMATXCTL_GDV;
2060 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2061 
2062 	/* CTRL_EXT: Global Double VLAN Enable */
2063 	ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2064 	ctrl |= IXGBE_EXTENDED_VLAN;
2065 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2066 
2067 	/* Clear pooling mode of PFVTCTL. It's required by X550. */
2068 	if (hw->mac.type == ixgbe_mac_X550 ||
2069 	    hw->mac.type == ixgbe_mac_X550EM_x ||
2070 	    hw->mac.type == ixgbe_mac_X550EM_a) {
2071 		ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
2072 		ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
2073 		IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
2074 	}
2075 
2076 	/*
2077 	 * VET EXT field in the EXVET register = 0x8100 by default
2078 	 * So no need to change. Same to VT field of DMATXCTL register
2079 	 */
2080 }
2081 
2082 void
2083 ixgbe_vlan_hw_strip_config(struct rte_eth_dev *dev)
2084 {
2085 	struct ixgbe_hw *hw =
2086 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2087 	struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
2088 	uint32_t ctrl;
2089 	uint16_t i;
2090 	struct ixgbe_rx_queue *rxq;
2091 	bool on;
2092 
2093 	PMD_INIT_FUNC_TRACE();
2094 
2095 	if (hw->mac.type == ixgbe_mac_82598EB) {
2096 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP) {
2097 			ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2098 			ctrl |= IXGBE_VLNCTRL_VME;
2099 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2100 		} else {
2101 			ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2102 			ctrl &= ~IXGBE_VLNCTRL_VME;
2103 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2104 		}
2105 	} else {
2106 		/*
2107 		 * Other 10G NIC, the VLAN strip can be setup
2108 		 * per queue in RXDCTL
2109 		 */
2110 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
2111 			rxq = dev->data->rx_queues[i];
2112 			ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2113 			if (rxq->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP) {
2114 				ctrl |= IXGBE_RXDCTL_VME;
2115 				on = TRUE;
2116 			} else {
2117 				ctrl &= ~IXGBE_RXDCTL_VME;
2118 				on = FALSE;
2119 			}
2120 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2121 
2122 			/* record those setting for HW strip per queue */
2123 			ixgbe_vlan_hw_strip_bitmap_set(dev, i, on);
2124 		}
2125 	}
2126 }
2127 
2128 static void
2129 ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev, int mask)
2130 {
2131 	uint16_t i;
2132 	struct rte_eth_rxmode *rxmode;
2133 	struct ixgbe_rx_queue *rxq;
2134 
2135 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
2136 		rxmode = &dev->data->dev_conf.rxmode;
2137 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
2138 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
2139 				rxq = dev->data->rx_queues[i];
2140 				rxq->offloads |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
2141 			}
2142 		else
2143 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
2144 				rxq = dev->data->rx_queues[i];
2145 				rxq->offloads &= ~RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
2146 			}
2147 	}
2148 }
2149 
2150 static int
2151 ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask)
2152 {
2153 	struct rte_eth_rxmode *rxmode;
2154 	rxmode = &dev->data->dev_conf.rxmode;
2155 
2156 	if (mask & RTE_ETH_VLAN_STRIP_MASK)
2157 		ixgbe_vlan_hw_strip_config(dev);
2158 
2159 	if (mask & RTE_ETH_VLAN_FILTER_MASK) {
2160 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
2161 			ixgbe_vlan_hw_filter_enable(dev);
2162 		else
2163 			ixgbe_vlan_hw_filter_disable(dev);
2164 	}
2165 
2166 	if (mask & RTE_ETH_VLAN_EXTEND_MASK) {
2167 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND)
2168 			ixgbe_vlan_hw_extend_enable(dev);
2169 		else
2170 			ixgbe_vlan_hw_extend_disable(dev);
2171 	}
2172 
2173 	return 0;
2174 }
2175 
2176 static int
2177 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2178 {
2179 	ixgbe_config_vlan_strip_on_all_queues(dev, mask);
2180 
2181 	ixgbe_vlan_offload_config(dev, mask);
2182 
2183 	return 0;
2184 }
2185 
2186 static void
2187 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2188 {
2189 	struct ixgbe_hw *hw =
2190 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2191 	/* VLNCTRL: enable vlan filtering and allow all vlan tags through */
2192 	uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2193 
2194 	vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
2195 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
2196 }
2197 
2198 static int
2199 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
2200 {
2201 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2202 
2203 	switch (nb_rx_q) {
2204 	case 1:
2205 	case 2:
2206 		RTE_ETH_DEV_SRIOV(dev).active = RTE_ETH_64_POOLS;
2207 		break;
2208 	case 4:
2209 		RTE_ETH_DEV_SRIOV(dev).active = RTE_ETH_32_POOLS;
2210 		break;
2211 	default:
2212 		return -EINVAL;
2213 	}
2214 
2215 	RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
2216 		IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2217 	RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
2218 		pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2219 	return 0;
2220 }
2221 
2222 static int
2223 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
2224 {
2225 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
2226 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2227 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
2228 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
2229 
2230 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
2231 		/* check multi-queue mode */
2232 		switch (dev_conf->rxmode.mq_mode) {
2233 		case RTE_ETH_MQ_RX_VMDQ_DCB:
2234 			PMD_INIT_LOG(INFO, "RTE_ETH_MQ_RX_VMDQ_DCB mode supported in SRIOV");
2235 			break;
2236 		case RTE_ETH_MQ_RX_VMDQ_DCB_RSS:
2237 			/* DCB/RSS VMDQ in SRIOV mode, not implement yet */
2238 			PMD_INIT_LOG(ERR, "SRIOV active,"
2239 					" unsupported mq_mode rx %d.",
2240 					dev_conf->rxmode.mq_mode);
2241 			return -EINVAL;
2242 		case RTE_ETH_MQ_RX_RSS:
2243 		case RTE_ETH_MQ_RX_VMDQ_RSS:
2244 			dev->data->dev_conf.rxmode.mq_mode = RTE_ETH_MQ_RX_VMDQ_RSS;
2245 			if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
2246 				if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
2247 					PMD_INIT_LOG(ERR, "SRIOV is active,"
2248 						" invalid queue number"
2249 						" for VMDQ RSS, allowed"
2250 						" value are 1, 2 or 4.");
2251 					return -EINVAL;
2252 				}
2253 			break;
2254 		case RTE_ETH_MQ_RX_VMDQ_ONLY:
2255 		case RTE_ETH_MQ_RX_NONE:
2256 			/* if nothing mq mode configure, use default scheme */
2257 			dev->data->dev_conf.rxmode.mq_mode = RTE_ETH_MQ_RX_VMDQ_ONLY;
2258 			break;
2259 		default: /* RTE_ETH_MQ_RX_DCB, RTE_ETH_MQ_RX_DCB_RSS or RTE_ETH_MQ_TX_DCB*/
2260 			/* SRIOV only works in VMDq enable mode */
2261 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2262 					" wrong mq_mode rx %d.",
2263 					dev_conf->rxmode.mq_mode);
2264 			return -EINVAL;
2265 		}
2266 
2267 		switch (dev_conf->txmode.mq_mode) {
2268 		case RTE_ETH_MQ_TX_VMDQ_DCB:
2269 			PMD_INIT_LOG(INFO, "RTE_ETH_MQ_TX_VMDQ_DCB mode supported in SRIOV");
2270 			dev->data->dev_conf.txmode.mq_mode = RTE_ETH_MQ_TX_VMDQ_DCB;
2271 			break;
2272 		default: /* RTE_ETH_MQ_TX_VMDQ_ONLY or RTE_ETH_MQ_TX_NONE */
2273 			dev->data->dev_conf.txmode.mq_mode = RTE_ETH_MQ_TX_VMDQ_ONLY;
2274 			break;
2275 		}
2276 
2277 		/* check valid queue number */
2278 		if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2279 		    (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2280 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2281 					" nb_rx_q=%d nb_tx_q=%d queue number"
2282 					" must be less than or equal to %d.",
2283 					nb_rx_q, nb_tx_q,
2284 					RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2285 			return -EINVAL;
2286 		}
2287 	} else {
2288 		if (dev_conf->rxmode.mq_mode == RTE_ETH_MQ_RX_VMDQ_DCB_RSS) {
2289 			PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2290 					  " not supported.");
2291 			return -EINVAL;
2292 		}
2293 		/* check configuration for vmdb+dcb mode */
2294 		if (dev_conf->rxmode.mq_mode == RTE_ETH_MQ_RX_VMDQ_DCB) {
2295 			const struct rte_eth_vmdq_dcb_conf *conf;
2296 
2297 			if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2298 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2299 						IXGBE_VMDQ_DCB_NB_QUEUES);
2300 				return -EINVAL;
2301 			}
2302 			conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2303 			if (!(conf->nb_queue_pools == RTE_ETH_16_POOLS ||
2304 			       conf->nb_queue_pools == RTE_ETH_32_POOLS)) {
2305 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2306 						" nb_queue_pools must be %d or %d.",
2307 						RTE_ETH_16_POOLS, RTE_ETH_32_POOLS);
2308 				return -EINVAL;
2309 			}
2310 		}
2311 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_VMDQ_DCB) {
2312 			const struct rte_eth_vmdq_dcb_tx_conf *conf;
2313 
2314 			if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2315 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2316 						 IXGBE_VMDQ_DCB_NB_QUEUES);
2317 				return -EINVAL;
2318 			}
2319 			conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2320 			if (!(conf->nb_queue_pools == RTE_ETH_16_POOLS ||
2321 			       conf->nb_queue_pools == RTE_ETH_32_POOLS)) {
2322 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2323 						" nb_queue_pools != %d and"
2324 						" nb_queue_pools != %d.",
2325 						RTE_ETH_16_POOLS, RTE_ETH_32_POOLS);
2326 				return -EINVAL;
2327 			}
2328 		}
2329 
2330 		/* For DCB mode check our configuration before we go further */
2331 		if (dev_conf->rxmode.mq_mode == RTE_ETH_MQ_RX_DCB) {
2332 			const struct rte_eth_dcb_rx_conf *conf;
2333 
2334 			conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2335 			if (!(conf->nb_tcs == RTE_ETH_4_TCS ||
2336 			       conf->nb_tcs == RTE_ETH_8_TCS)) {
2337 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2338 						" and nb_tcs != %d.",
2339 						RTE_ETH_4_TCS, RTE_ETH_8_TCS);
2340 				return -EINVAL;
2341 			}
2342 		}
2343 
2344 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_DCB) {
2345 			const struct rte_eth_dcb_tx_conf *conf;
2346 
2347 			conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2348 			if (!(conf->nb_tcs == RTE_ETH_4_TCS ||
2349 			       conf->nb_tcs == RTE_ETH_8_TCS)) {
2350 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2351 						" and nb_tcs != %d.",
2352 						RTE_ETH_4_TCS, RTE_ETH_8_TCS);
2353 				return -EINVAL;
2354 			}
2355 		}
2356 
2357 		/*
2358 		 * When DCB/VT is off, maximum number of queues changes,
2359 		 * except for 82598EB, which remains constant.
2360 		 */
2361 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_NONE &&
2362 				hw->mac.type != ixgbe_mac_82598EB) {
2363 			if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2364 				PMD_INIT_LOG(ERR,
2365 					     "Neither VT nor DCB are enabled, "
2366 					     "nb_tx_q > %d.",
2367 					     IXGBE_NONE_MODE_TX_NB_QUEUES);
2368 				return -EINVAL;
2369 			}
2370 		}
2371 	}
2372 	return 0;
2373 }
2374 
2375 static int
2376 ixgbe_dev_configure(struct rte_eth_dev *dev)
2377 {
2378 	struct ixgbe_interrupt *intr =
2379 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2380 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2381 	int ret;
2382 
2383 	PMD_INIT_FUNC_TRACE();
2384 
2385 	if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
2386 		dev->data->dev_conf.rxmode.offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
2387 
2388 	/* multiple queue mode checking */
2389 	ret  = ixgbe_check_mq_mode(dev);
2390 	if (ret != 0) {
2391 		PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2392 			    ret);
2393 		return ret;
2394 	}
2395 
2396 	/* set flag to update link status after init */
2397 	intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2398 
2399 	/*
2400 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2401 	 * allocation or vector Rx preconditions we will reset it.
2402 	 */
2403 	adapter->rx_bulk_alloc_allowed = true;
2404 	adapter->rx_vec_allowed = true;
2405 
2406 	return 0;
2407 }
2408 
2409 static void
2410 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2411 {
2412 	struct ixgbe_hw *hw =
2413 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2414 	struct ixgbe_interrupt *intr =
2415 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2416 	uint32_t gpie;
2417 
2418 	/* only set up it on X550EM_X */
2419 	if (hw->mac.type == ixgbe_mac_X550EM_x) {
2420 		gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2421 		gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2422 		IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2423 		if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2424 			intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2425 	}
2426 }
2427 
2428 int
2429 ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
2430 			uint16_t tx_rate, uint64_t q_msk)
2431 {
2432 	struct ixgbe_hw *hw;
2433 	struct ixgbe_vf_info *vfinfo;
2434 	struct rte_eth_link link;
2435 	uint8_t  nb_q_per_pool;
2436 	uint32_t queue_stride;
2437 	uint32_t queue_idx, idx = 0, vf_idx;
2438 	uint32_t queue_end;
2439 	uint16_t total_rate = 0;
2440 	struct rte_pci_device *pci_dev;
2441 	int ret;
2442 
2443 	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2444 	ret = rte_eth_link_get_nowait(dev->data->port_id, &link);
2445 	if (ret < 0)
2446 		return ret;
2447 
2448 	if (vf >= pci_dev->max_vfs)
2449 		return -EINVAL;
2450 
2451 	if (tx_rate > link.link_speed)
2452 		return -EINVAL;
2453 
2454 	if (q_msk == 0)
2455 		return 0;
2456 
2457 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2458 	vfinfo = *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
2459 	nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2460 	queue_stride = IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2461 	queue_idx = vf * queue_stride;
2462 	queue_end = queue_idx + nb_q_per_pool - 1;
2463 	if (queue_end >= hw->mac.max_tx_queues)
2464 		return -EINVAL;
2465 
2466 	if (vfinfo) {
2467 		for (vf_idx = 0; vf_idx < pci_dev->max_vfs; vf_idx++) {
2468 			if (vf_idx == vf)
2469 				continue;
2470 			for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
2471 				idx++)
2472 				total_rate += vfinfo[vf_idx].tx_rate[idx];
2473 		}
2474 	} else {
2475 		return -EINVAL;
2476 	}
2477 
2478 	/* Store tx_rate for this vf. */
2479 	for (idx = 0; idx < nb_q_per_pool; idx++) {
2480 		if (((uint64_t)0x1 << idx) & q_msk) {
2481 			if (vfinfo[vf].tx_rate[idx] != tx_rate)
2482 				vfinfo[vf].tx_rate[idx] = tx_rate;
2483 			total_rate += tx_rate;
2484 		}
2485 	}
2486 
2487 	if (total_rate > dev->data->dev_link.link_speed) {
2488 		/* Reset stored TX rate of the VF if it causes exceed
2489 		 * link speed.
2490 		 */
2491 		memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
2492 		return -EINVAL;
2493 	}
2494 
2495 	/* Set RTTBCNRC of each queue/pool for vf X  */
2496 	for (; queue_idx <= queue_end; queue_idx++) {
2497 		if (0x1 & q_msk)
2498 			ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
2499 		q_msk = q_msk >> 1;
2500 	}
2501 
2502 	return 0;
2503 }
2504 
2505 static int
2506 ixgbe_flow_ctrl_enable(struct rte_eth_dev *dev, struct ixgbe_hw *hw)
2507 {
2508 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2509 	int err;
2510 	uint32_t mflcn;
2511 
2512 	ixgbe_setup_fc(hw);
2513 
2514 	err = ixgbe_fc_enable(hw);
2515 
2516 	/* Not negotiated is not an error case */
2517 	if (err == IXGBE_SUCCESS || err == IXGBE_ERR_FC_NOT_NEGOTIATED) {
2518 		/*
2519 		 *check if we want to forward MAC frames - driver doesn't
2520 		 *have native capability to do that,
2521 		 *so we'll write the registers ourselves
2522 		 */
2523 
2524 		mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
2525 
2526 		/* set or clear MFLCN.PMCF bit depending on configuration */
2527 		if (adapter->mac_ctrl_frame_fwd != 0)
2528 			mflcn |= IXGBE_MFLCN_PMCF;
2529 		else
2530 			mflcn &= ~IXGBE_MFLCN_PMCF;
2531 
2532 		IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
2533 		IXGBE_WRITE_FLUSH(hw);
2534 
2535 		return 0;
2536 	}
2537 	return err;
2538 }
2539 
2540 /*
2541  * Configure device link speed and setup link.
2542  * It returns 0 on success.
2543  */
2544 static int
2545 ixgbe_dev_start(struct rte_eth_dev *dev)
2546 {
2547 	struct ixgbe_hw *hw =
2548 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2549 	struct ixgbe_vf_info *vfinfo =
2550 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2551 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2552 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2553 	uint32_t intr_vector = 0;
2554 	int err;
2555 	bool link_up = false, negotiate = 0;
2556 	uint32_t speed = 0;
2557 	uint32_t allowed_speeds = 0;
2558 	int mask = 0;
2559 	int status;
2560 	uint16_t vf, idx;
2561 	uint32_t *link_speeds;
2562 	struct ixgbe_tm_conf *tm_conf =
2563 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2564 	struct ixgbe_macsec_setting *macsec_setting =
2565 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
2566 
2567 	PMD_INIT_FUNC_TRACE();
2568 
2569 	/* Stop the link setup handler before resetting the HW. */
2570 	ixgbe_dev_wait_setup_link_complete(dev, 0);
2571 
2572 	/* disable uio/vfio intr/eventfd mapping */
2573 	rte_intr_disable(intr_handle);
2574 
2575 	/* stop adapter */
2576 	hw->adapter_stopped = 0;
2577 	ixgbe_stop_adapter(hw);
2578 
2579 	/* reinitialize adapter
2580 	 * this calls reset and start
2581 	 */
2582 	status = ixgbe_pf_reset_hw(hw);
2583 	if (status != 0)
2584 		return -1;
2585 	hw->mac.ops.start_hw(hw);
2586 	hw->mac.get_link_status = true;
2587 
2588 	/* configure PF module if SRIOV enabled */
2589 	ixgbe_pf_host_configure(dev);
2590 
2591 	ixgbe_dev_phy_intr_setup(dev);
2592 
2593 	/* check and configure queue intr-vector mapping */
2594 	if ((rte_intr_cap_multiple(intr_handle) ||
2595 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
2596 	    dev->data->dev_conf.intr_conf.rxq != 0) {
2597 		intr_vector = dev->data->nb_rx_queues;
2598 		if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2599 			PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2600 					IXGBE_MAX_INTR_QUEUE_NUM);
2601 			return -ENOTSUP;
2602 		}
2603 		if (rte_intr_efd_enable(intr_handle, intr_vector))
2604 			return -1;
2605 	}
2606 
2607 	if (rte_intr_dp_is_en(intr_handle)) {
2608 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
2609 						   dev->data->nb_rx_queues)) {
2610 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2611 				     " intr_vec", dev->data->nb_rx_queues);
2612 			return -ENOMEM;
2613 		}
2614 	}
2615 
2616 	/* configure MSI-X for sleep until Rx interrupt */
2617 	ixgbe_configure_msix(dev);
2618 
2619 	/* initialize transmission unit */
2620 	ixgbe_dev_tx_init(dev);
2621 
2622 	/* This can fail when allocating mbufs for descriptor rings */
2623 	err = ixgbe_dev_rx_init(dev);
2624 	if (err) {
2625 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2626 		goto error;
2627 	}
2628 
2629 	mask = RTE_ETH_VLAN_STRIP_MASK | RTE_ETH_VLAN_FILTER_MASK |
2630 		RTE_ETH_VLAN_EXTEND_MASK;
2631 	err = ixgbe_vlan_offload_config(dev, mask);
2632 	if (err) {
2633 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload");
2634 		goto error;
2635 	}
2636 
2637 	if (dev->data->dev_conf.rxmode.mq_mode == RTE_ETH_MQ_RX_VMDQ_ONLY) {
2638 		/* Enable vlan filtering for VMDq */
2639 		ixgbe_vmdq_vlan_hw_filter_enable(dev);
2640 	}
2641 
2642 	/* Configure DCB hw */
2643 	ixgbe_configure_dcb(dev);
2644 
2645 	if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2646 		err = ixgbe_fdir_configure(dev);
2647 		if (err)
2648 			goto error;
2649 	}
2650 
2651 	/* Restore vf rate limit */
2652 	if (vfinfo != NULL) {
2653 		for (vf = 0; vf < pci_dev->max_vfs; vf++)
2654 			for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2655 				if (vfinfo[vf].tx_rate[idx] != 0)
2656 					ixgbe_set_vf_rate_limit(
2657 						dev, vf,
2658 						vfinfo[vf].tx_rate[idx],
2659 						1 << idx);
2660 	}
2661 
2662 	ixgbe_restore_statistics_mapping(dev);
2663 
2664 	err = ixgbe_flow_ctrl_enable(dev, hw);
2665 	if (err < 0) {
2666 		PMD_INIT_LOG(ERR, "enable flow ctrl err");
2667 		goto error;
2668 	}
2669 
2670 	err = ixgbe_dev_rxtx_start(dev);
2671 	if (err < 0) {
2672 		PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2673 		goto error;
2674 	}
2675 
2676 	/* Skip link setup if loopback mode is enabled. */
2677 	if (dev->data->dev_conf.lpbk_mode != 0) {
2678 		err = ixgbe_check_supported_loopback_mode(dev);
2679 		if (err < 0) {
2680 			PMD_INIT_LOG(ERR, "Unsupported loopback mode");
2681 			goto error;
2682 		} else {
2683 			goto skip_link_setup;
2684 		}
2685 	}
2686 
2687 	if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2688 		err = hw->mac.ops.setup_sfp(hw);
2689 		if (err)
2690 			goto error;
2691 	}
2692 
2693 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2694 		/* Turn on the copper */
2695 		ixgbe_set_phy_power(hw, true);
2696 	} else {
2697 		/* Turn on the laser */
2698 		ixgbe_enable_tx_laser(hw);
2699 	}
2700 
2701 	err = ixgbe_check_link(hw, &speed, &link_up, 0);
2702 	if (err)
2703 		goto error;
2704 	dev->data->dev_link.link_status = link_up;
2705 
2706 	err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2707 	if (err)
2708 		goto error;
2709 
2710 	switch (hw->mac.type) {
2711 	case ixgbe_mac_X550:
2712 	case ixgbe_mac_X550EM_x:
2713 	case ixgbe_mac_X550EM_a:
2714 		allowed_speeds = RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G |
2715 			RTE_ETH_LINK_SPEED_2_5G |  RTE_ETH_LINK_SPEED_5G |
2716 			RTE_ETH_LINK_SPEED_10G;
2717 		if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
2718 				hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
2719 			allowed_speeds = RTE_ETH_LINK_SPEED_10M |
2720 				RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G;
2721 		break;
2722 	default:
2723 		allowed_speeds = RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G |
2724 			RTE_ETH_LINK_SPEED_10G;
2725 	}
2726 
2727 	link_speeds = &dev->data->dev_conf.link_speeds;
2728 
2729 	/* Ignore autoneg flag bit and check the validity of 
2730 	 * link_speed 
2731 	 */
2732 	if (((*link_speeds) >> 1) & ~(allowed_speeds >> 1)) {
2733 		PMD_INIT_LOG(ERR, "Invalid link setting");
2734 		goto error;
2735 	}
2736 
2737 	speed = 0x0;
2738 	if (*link_speeds == RTE_ETH_LINK_SPEED_AUTONEG) {
2739 		switch (hw->mac.type) {
2740 		case ixgbe_mac_82598EB:
2741 			speed = IXGBE_LINK_SPEED_82598_AUTONEG;
2742 			break;
2743 		case ixgbe_mac_82599EB:
2744 		case ixgbe_mac_X540:
2745 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2746 			break;
2747 		case ixgbe_mac_X550:
2748 		case ixgbe_mac_X550EM_x:
2749 		case ixgbe_mac_X550EM_a:
2750 			speed = IXGBE_LINK_SPEED_X550_AUTONEG;
2751 			break;
2752 		default:
2753 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2754 		}
2755 	} else {
2756 		if (*link_speeds & RTE_ETH_LINK_SPEED_10G)
2757 			speed |= IXGBE_LINK_SPEED_10GB_FULL;
2758 		if (*link_speeds & RTE_ETH_LINK_SPEED_5G)
2759 			speed |= IXGBE_LINK_SPEED_5GB_FULL;
2760 		if (*link_speeds & RTE_ETH_LINK_SPEED_2_5G)
2761 			speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
2762 		if (*link_speeds & RTE_ETH_LINK_SPEED_1G)
2763 			speed |= IXGBE_LINK_SPEED_1GB_FULL;
2764 		if (*link_speeds & RTE_ETH_LINK_SPEED_100M)
2765 			speed |= IXGBE_LINK_SPEED_100_FULL;
2766 		if (*link_speeds & RTE_ETH_LINK_SPEED_10M)
2767 			speed |= IXGBE_LINK_SPEED_10_FULL;
2768 	}
2769 
2770 	err = ixgbe_setup_link(hw, speed, link_up);
2771 	if (err)
2772 		goto error;
2773 
2774 skip_link_setup:
2775 
2776 	if (rte_intr_allow_others(intr_handle)) {
2777 		/* check if lsc interrupt is enabled */
2778 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2779 			ixgbe_dev_lsc_interrupt_setup(dev, TRUE);
2780 		else
2781 			ixgbe_dev_lsc_interrupt_setup(dev, FALSE);
2782 		ixgbe_dev_macsec_interrupt_setup(dev);
2783 	} else {
2784 		rte_intr_callback_unregister(intr_handle,
2785 					     ixgbe_dev_interrupt_handler, dev);
2786 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2787 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
2788 				     " no intr multiplex");
2789 	}
2790 
2791 	/* check if rxq interrupt is enabled */
2792 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2793 	    rte_intr_dp_is_en(intr_handle))
2794 		ixgbe_dev_rxq_interrupt_setup(dev);
2795 
2796 	/* enable uio/vfio intr/eventfd mapping */
2797 	rte_intr_enable(intr_handle);
2798 
2799 	/* resume enabled intr since hw reset */
2800 	ixgbe_enable_intr(dev);
2801 	ixgbe_l2_tunnel_conf(dev);
2802 	ixgbe_filter_restore(dev);
2803 
2804 	if (tm_conf->root && !tm_conf->committed)
2805 		PMD_DRV_LOG(WARNING,
2806 			    "please call hierarchy_commit() "
2807 			    "before starting the port");
2808 
2809 	/* wait for the controller to acquire link */
2810 	err = ixgbe_wait_for_link_up(hw);
2811 	if (err)
2812 		goto error;
2813 
2814 	/*
2815 	 * Update link status right before return, because it may
2816 	 * start link configuration process in a separate thread.
2817 	 */
2818 	ixgbe_dev_link_update(dev, 0);
2819 
2820 	/* setup the macsec setting register */
2821 	if (macsec_setting->offload_en)
2822 		ixgbe_dev_macsec_register_enable(dev, macsec_setting);
2823 
2824 	return 0;
2825 
2826 error:
2827 	PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2828 	ixgbe_dev_clear_queues(dev);
2829 	return -EIO;
2830 }
2831 
2832 /*
2833  * Stop device: disable rx and tx functions to allow for reconfiguring.
2834  */
2835 static int
2836 ixgbe_dev_stop(struct rte_eth_dev *dev)
2837 {
2838 	struct rte_eth_link link;
2839 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2840 	struct ixgbe_hw *hw =
2841 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2842 	struct ixgbe_vf_info *vfinfo =
2843 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2844 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2845 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2846 	int vf;
2847 	struct ixgbe_tm_conf *tm_conf =
2848 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2849 
2850 	if (hw->adapter_stopped)
2851 		return 0;
2852 
2853 	PMD_INIT_FUNC_TRACE();
2854 
2855 	ixgbe_dev_wait_setup_link_complete(dev, 0);
2856 
2857 	/* disable interrupts */
2858 	ixgbe_disable_intr(hw);
2859 
2860 	/* reset the NIC */
2861 	ixgbe_pf_reset_hw(hw);
2862 	hw->adapter_stopped = 0;
2863 
2864 	/* stop adapter */
2865 	ixgbe_stop_adapter(hw);
2866 
2867 	for (vf = 0; vfinfo != NULL && vf < pci_dev->max_vfs; vf++)
2868 		vfinfo[vf].clear_to_send = false;
2869 
2870 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2871 		/* Turn off the copper */
2872 		ixgbe_set_phy_power(hw, false);
2873 	} else {
2874 		/* Turn off the laser */
2875 		ixgbe_disable_tx_laser(hw);
2876 	}
2877 
2878 	ixgbe_dev_clear_queues(dev);
2879 
2880 	/* Clear stored conf */
2881 	dev->data->scattered_rx = 0;
2882 	dev->data->lro = 0;
2883 
2884 	/* Clear recorded link status */
2885 	memset(&link, 0, sizeof(link));
2886 	rte_eth_linkstatus_set(dev, &link);
2887 
2888 	if (!rte_intr_allow_others(intr_handle))
2889 		/* resume to the default handler */
2890 		rte_intr_callback_register(intr_handle,
2891 					   ixgbe_dev_interrupt_handler,
2892 					   (void *)dev);
2893 
2894 	/* Clean datapath event and queue/vec mapping */
2895 	rte_intr_efd_disable(intr_handle);
2896 	rte_intr_vec_list_free(intr_handle);
2897 
2898 	/* reset hierarchy commit */
2899 	tm_conf->committed = false;
2900 
2901 	adapter->rss_reta_updated = 0;
2902 
2903 	hw->adapter_stopped = true;
2904 	dev->data->dev_started = 0;
2905 
2906 	return 0;
2907 }
2908 
2909 /*
2910  * Set device link up: enable tx.
2911  */
2912 static int
2913 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2914 {
2915 	struct ixgbe_hw *hw =
2916 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2917 	if (hw->mac.type == ixgbe_mac_82599EB) {
2918 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2919 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2920 			/* Not supported in bypass mode */
2921 			PMD_INIT_LOG(ERR, "Set link up is not supported "
2922 				     "by device id 0x%x", hw->device_id);
2923 			return -ENOTSUP;
2924 		}
2925 #endif
2926 	}
2927 
2928 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2929 		/* Turn on the copper */
2930 		ixgbe_set_phy_power(hw, true);
2931 	} else {
2932 		/* Turn on the laser */
2933 		ixgbe_enable_tx_laser(hw);
2934 		ixgbe_dev_link_update(dev, 0);
2935 	}
2936 
2937 	return 0;
2938 }
2939 
2940 /*
2941  * Set device link down: disable tx.
2942  */
2943 static int
2944 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2945 {
2946 	struct ixgbe_hw *hw =
2947 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2948 	if (hw->mac.type == ixgbe_mac_82599EB) {
2949 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2950 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2951 			/* Not supported in bypass mode */
2952 			PMD_INIT_LOG(ERR, "Set link down is not supported "
2953 				     "by device id 0x%x", hw->device_id);
2954 			return -ENOTSUP;
2955 		}
2956 #endif
2957 	}
2958 
2959 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2960 		/* Turn off the copper */
2961 		ixgbe_set_phy_power(hw, false);
2962 	} else {
2963 		/* Turn off the laser */
2964 		ixgbe_disable_tx_laser(hw);
2965 		ixgbe_dev_link_update(dev, 0);
2966 	}
2967 
2968 	return 0;
2969 }
2970 
2971 /*
2972  * Reset and stop device.
2973  */
2974 static int
2975 ixgbe_dev_close(struct rte_eth_dev *dev)
2976 {
2977 	struct ixgbe_hw *hw =
2978 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2979 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2980 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2981 	int retries = 0;
2982 	int ret;
2983 
2984 	PMD_INIT_FUNC_TRACE();
2985 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2986 		return 0;
2987 
2988 	ixgbe_pf_reset_hw(hw);
2989 
2990 	ret = ixgbe_dev_stop(dev);
2991 
2992 	ixgbe_dev_free_queues(dev);
2993 
2994 	ixgbe_disable_pcie_master(hw);
2995 
2996 	/* reprogram the RAR[0] in case user changed it. */
2997 	ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
2998 
2999 	/* Unlock any pending hardware semaphore */
3000 	ixgbe_swfw_lock_reset(hw);
3001 
3002 	/* disable uio intr before callback unregister */
3003 	rte_intr_disable(intr_handle);
3004 
3005 	do {
3006 		ret = rte_intr_callback_unregister(intr_handle,
3007 				ixgbe_dev_interrupt_handler, dev);
3008 		if (ret >= 0 || ret == -ENOENT) {
3009 			break;
3010 		} else if (ret != -EAGAIN) {
3011 			PMD_INIT_LOG(ERR,
3012 				"intr callback unregister failed: %d",
3013 				ret);
3014 		}
3015 		rte_delay_ms(100);
3016 	} while (retries++ < (10 + IXGBE_LINK_UP_TIME));
3017 
3018 	/* cancel the delay handler before remove dev */
3019 	rte_eal_alarm_cancel(ixgbe_dev_interrupt_delayed_handler, dev);
3020 
3021 	/* uninitialize PF if max_vfs not zero */
3022 	ixgbe_pf_host_uninit(dev);
3023 
3024 	/* remove all the fdir filters & hash */
3025 	ixgbe_fdir_filter_uninit(dev);
3026 
3027 	/* remove all the L2 tunnel filters & hash */
3028 	ixgbe_l2_tn_filter_uninit(dev);
3029 
3030 	/* Remove all ntuple filters of the device */
3031 	ixgbe_ntuple_filter_uninit(dev);
3032 
3033 	/* clear all the filters list */
3034 	ixgbe_filterlist_flush();
3035 
3036 	/* Remove all Traffic Manager configuration */
3037 	ixgbe_tm_conf_uninit(dev);
3038 
3039 #ifdef RTE_LIB_SECURITY
3040 	rte_free(dev->security_ctx);
3041 	dev->security_ctx = NULL;
3042 #endif
3043 
3044 	return ret;
3045 }
3046 
3047 /*
3048  * Reset PF device.
3049  */
3050 static int
3051 ixgbe_dev_reset(struct rte_eth_dev *dev)
3052 {
3053 	int ret;
3054 
3055 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
3056 	 * its VF to make them align with it. The detailed notification
3057 	 * mechanism is PMD specific. As to ixgbe PF, it is rather complex.
3058 	 * To avoid unexpected behavior in VF, currently reset of PF with
3059 	 * SR-IOV activation is not supported. It might be supported later.
3060 	 */
3061 	if (dev->data->sriov.active)
3062 		return -ENOTSUP;
3063 
3064 	ret = eth_ixgbe_dev_uninit(dev);
3065 	if (ret)
3066 		return ret;
3067 
3068 	ret = eth_ixgbe_dev_init(dev, NULL);
3069 
3070 	return ret;
3071 }
3072 
3073 static void
3074 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
3075 			   struct ixgbe_hw_stats *hw_stats,
3076 			   struct ixgbe_macsec_stats *macsec_stats,
3077 			   uint64_t *total_missed_rx, uint64_t *total_qbrc,
3078 			   uint64_t *total_qprc, uint64_t *total_qprdc)
3079 {
3080 	uint32_t bprc, lxon, lxoff, total;
3081 	uint32_t delta_gprc = 0;
3082 	unsigned i;
3083 	/* Workaround for RX byte count not including CRC bytes when CRC
3084 	 * strip is enabled. CRC bytes are removed from counters when crc_strip
3085 	 * is disabled.
3086 	 */
3087 	int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
3088 			IXGBE_HLREG0_RXCRCSTRP);
3089 
3090 	hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
3091 	hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
3092 	hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
3093 	hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
3094 
3095 	for (i = 0; i < 8; i++) {
3096 		uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
3097 
3098 		/* global total per queue */
3099 		hw_stats->mpc[i] += mp;
3100 		/* Running comprehensive total for stats display */
3101 		*total_missed_rx += hw_stats->mpc[i];
3102 		if (hw->mac.type == ixgbe_mac_82598EB) {
3103 			hw_stats->rnbc[i] +=
3104 			    IXGBE_READ_REG(hw, IXGBE_RNBC(i));
3105 			hw_stats->pxonrxc[i] +=
3106 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
3107 			hw_stats->pxoffrxc[i] +=
3108 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
3109 		} else {
3110 			hw_stats->pxonrxc[i] +=
3111 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
3112 			hw_stats->pxoffrxc[i] +=
3113 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
3114 			hw_stats->pxon2offc[i] +=
3115 				IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
3116 		}
3117 		hw_stats->pxontxc[i] +=
3118 		    IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
3119 		hw_stats->pxofftxc[i] +=
3120 		    IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
3121 	}
3122 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3123 		uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
3124 		uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
3125 		uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
3126 
3127 		delta_gprc += delta_qprc;
3128 
3129 		hw_stats->qprc[i] += delta_qprc;
3130 		hw_stats->qptc[i] += delta_qptc;
3131 
3132 		hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
3133 		hw_stats->qbrc[i] +=
3134 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
3135 		if (crc_strip == 0)
3136 			hw_stats->qbrc[i] -= delta_qprc * RTE_ETHER_CRC_LEN;
3137 
3138 		hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
3139 		hw_stats->qbtc[i] +=
3140 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
3141 
3142 		hw_stats->qprdc[i] += delta_qprdc;
3143 		*total_qprdc += hw_stats->qprdc[i];
3144 
3145 		*total_qprc += hw_stats->qprc[i];
3146 		*total_qbrc += hw_stats->qbrc[i];
3147 	}
3148 	hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
3149 	hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
3150 	hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
3151 
3152 	/*
3153 	 * An errata states that gprc actually counts good + missed packets:
3154 	 * Workaround to set gprc to summated queue packet receives
3155 	 */
3156 	hw_stats->gprc = *total_qprc;
3157 
3158 	if (hw->mac.type != ixgbe_mac_82598EB) {
3159 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
3160 		hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
3161 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
3162 		hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
3163 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
3164 		hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
3165 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
3166 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
3167 	} else {
3168 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
3169 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
3170 		/* 82598 only has a counter in the high register */
3171 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
3172 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
3173 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
3174 	}
3175 	uint64_t old_tpr = hw_stats->tpr;
3176 
3177 	hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
3178 	hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
3179 
3180 	if (crc_strip == 0)
3181 		hw_stats->gorc -= delta_gprc * RTE_ETHER_CRC_LEN;
3182 
3183 	uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
3184 	hw_stats->gptc += delta_gptc;
3185 	hw_stats->gotc -= delta_gptc * RTE_ETHER_CRC_LEN;
3186 	hw_stats->tor -= (hw_stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
3187 
3188 	/*
3189 	 * Workaround: mprc hardware is incorrectly counting
3190 	 * broadcasts, so for now we subtract those.
3191 	 */
3192 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
3193 	hw_stats->bprc += bprc;
3194 	hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
3195 	if (hw->mac.type == ixgbe_mac_82598EB)
3196 		hw_stats->mprc -= bprc;
3197 
3198 	hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
3199 	hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
3200 	hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
3201 	hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
3202 	hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
3203 	hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
3204 
3205 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
3206 	hw_stats->lxontxc += lxon;
3207 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
3208 	hw_stats->lxofftxc += lxoff;
3209 	total = lxon + lxoff;
3210 
3211 	hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
3212 	hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
3213 	hw_stats->gptc -= total;
3214 	hw_stats->mptc -= total;
3215 	hw_stats->ptc64 -= total;
3216 	hw_stats->gotc -= total * RTE_ETHER_MIN_LEN;
3217 
3218 	hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3219 	hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3220 	hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3221 	hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
3222 	hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
3223 	hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
3224 	hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
3225 	hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3226 	hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3227 	hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3228 	hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3229 	hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
3230 	hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3231 	hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
3232 	hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
3233 	hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
3234 	/* Only read FCOE on 82599 */
3235 	if (hw->mac.type != ixgbe_mac_82598EB) {
3236 		hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
3237 		hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
3238 		hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
3239 		hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
3240 		hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
3241 	}
3242 
3243 	/* Flow Director Stats registers */
3244 	if (hw->mac.type != ixgbe_mac_82598EB) {
3245 		hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
3246 		hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
3247 		hw_stats->fdirustat_add += IXGBE_READ_REG(hw,
3248 					IXGBE_FDIRUSTAT) & 0xFFFF;
3249 		hw_stats->fdirustat_remove += (IXGBE_READ_REG(hw,
3250 					IXGBE_FDIRUSTAT) >> 16) & 0xFFFF;
3251 		hw_stats->fdirfstat_fadd += IXGBE_READ_REG(hw,
3252 					IXGBE_FDIRFSTAT) & 0xFFFF;
3253 		hw_stats->fdirfstat_fremove += (IXGBE_READ_REG(hw,
3254 					IXGBE_FDIRFSTAT) >> 16) & 0xFFFF;
3255 	}
3256 	/* MACsec Stats registers */
3257 	macsec_stats->out_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECTXUT);
3258 	macsec_stats->out_pkts_encrypted +=
3259 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTE);
3260 	macsec_stats->out_pkts_protected +=
3261 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTP);
3262 	macsec_stats->out_octets_encrypted +=
3263 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTE);
3264 	macsec_stats->out_octets_protected +=
3265 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTP);
3266 	macsec_stats->in_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECRXUT);
3267 	macsec_stats->in_pkts_badtag += IXGBE_READ_REG(hw, IXGBE_LSECRXBAD);
3268 	macsec_stats->in_pkts_nosci += IXGBE_READ_REG(hw, IXGBE_LSECRXNOSCI);
3269 	macsec_stats->in_pkts_unknownsci +=
3270 		IXGBE_READ_REG(hw, IXGBE_LSECRXUNSCI);
3271 	macsec_stats->in_octets_decrypted +=
3272 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTD);
3273 	macsec_stats->in_octets_validated +=
3274 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTV);
3275 	macsec_stats->in_pkts_unchecked += IXGBE_READ_REG(hw, IXGBE_LSECRXUNCH);
3276 	macsec_stats->in_pkts_delayed += IXGBE_READ_REG(hw, IXGBE_LSECRXDELAY);
3277 	macsec_stats->in_pkts_late += IXGBE_READ_REG(hw, IXGBE_LSECRXLATE);
3278 	for (i = 0; i < 2; i++) {
3279 		macsec_stats->in_pkts_ok +=
3280 			IXGBE_READ_REG(hw, IXGBE_LSECRXOK(i));
3281 		macsec_stats->in_pkts_invalid +=
3282 			IXGBE_READ_REG(hw, IXGBE_LSECRXINV(i));
3283 		macsec_stats->in_pkts_notvalid +=
3284 			IXGBE_READ_REG(hw, IXGBE_LSECRXNV(i));
3285 	}
3286 	macsec_stats->in_pkts_unusedsa += IXGBE_READ_REG(hw, IXGBE_LSECRXUNSA);
3287 	macsec_stats->in_pkts_notusingsa +=
3288 		IXGBE_READ_REG(hw, IXGBE_LSECRXNUSA);
3289 }
3290 
3291 /*
3292  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
3293  */
3294 static int
3295 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3296 {
3297 	struct ixgbe_hw *hw =
3298 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3299 	struct ixgbe_hw_stats *hw_stats =
3300 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3301 	struct ixgbe_macsec_stats *macsec_stats =
3302 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3303 				dev->data->dev_private);
3304 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3305 	unsigned i;
3306 
3307 	total_missed_rx = 0;
3308 	total_qbrc = 0;
3309 	total_qprc = 0;
3310 	total_qprdc = 0;
3311 
3312 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3313 			&total_qbrc, &total_qprc, &total_qprdc);
3314 
3315 	if (stats == NULL)
3316 		return -EINVAL;
3317 
3318 	/* Fill out the rte_eth_stats statistics structure */
3319 	stats->ipackets = total_qprc;
3320 	stats->ibytes = total_qbrc;
3321 	stats->opackets = hw_stats->gptc;
3322 	stats->obytes = hw_stats->gotc;
3323 
3324 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3325 		stats->q_ipackets[i] = hw_stats->qprc[i];
3326 		stats->q_opackets[i] = hw_stats->qptc[i];
3327 		stats->q_ibytes[i] = hw_stats->qbrc[i];
3328 		stats->q_obytes[i] = hw_stats->qbtc[i];
3329 		stats->q_errors[i] = hw_stats->qprdc[i];
3330 	}
3331 
3332 	/* Rx Errors */
3333 	stats->imissed  = total_missed_rx;
3334 	stats->ierrors  = hw_stats->crcerrs +
3335 			  hw_stats->mspdc +
3336 			  hw_stats->rlec +
3337 			  hw_stats->ruc +
3338 			  hw_stats->roc +
3339 			  hw_stats->illerrc +
3340 			  hw_stats->errbc +
3341 			  hw_stats->rfc +
3342 			  hw_stats->fccrc +
3343 			  hw_stats->fclast;
3344 
3345 	/*
3346 	 * 82599 errata, UDP frames with a 0 checksum can be marked as checksum
3347 	 * errors.
3348 	 */
3349 	if (hw->mac.type != ixgbe_mac_82599EB)
3350 		stats->ierrors += hw_stats->xec;
3351 
3352 	/* Tx Errors */
3353 	stats->oerrors  = 0;
3354 	return 0;
3355 }
3356 
3357 static int
3358 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
3359 {
3360 	struct ixgbe_hw_stats *stats =
3361 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3362 
3363 	/* HW registers are cleared on read */
3364 	ixgbe_dev_stats_get(dev, NULL);
3365 
3366 	/* Reset software totals */
3367 	memset(stats, 0, sizeof(*stats));
3368 
3369 	return 0;
3370 }
3371 
3372 /* This function calculates the number of xstats based on the current config */
3373 static unsigned
3374 ixgbe_xstats_calc_num(void) {
3375 	return IXGBE_NB_HW_STATS + IXGBE_NB_MACSEC_STATS +
3376 		(IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
3377 		(IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
3378 }
3379 
3380 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3381 	struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned int size)
3382 {
3383 	const unsigned cnt_stats = ixgbe_xstats_calc_num();
3384 	unsigned stat, i, count;
3385 
3386 	if (xstats_names != NULL) {
3387 		count = 0;
3388 
3389 		/* Note: limit >= cnt_stats checked upstream
3390 		 * in rte_eth_xstats_names()
3391 		 */
3392 
3393 		/* Extended stats from ixgbe_hw_stats */
3394 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3395 			strlcpy(xstats_names[count].name,
3396 				rte_ixgbe_stats_strings[i].name,
3397 				sizeof(xstats_names[count].name));
3398 			count++;
3399 		}
3400 
3401 		/* MACsec Stats */
3402 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3403 			strlcpy(xstats_names[count].name,
3404 				rte_ixgbe_macsec_strings[i].name,
3405 				sizeof(xstats_names[count].name));
3406 			count++;
3407 		}
3408 
3409 		/* RX Priority Stats */
3410 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3411 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3412 				snprintf(xstats_names[count].name,
3413 					sizeof(xstats_names[count].name),
3414 					"rx_priority%u_%s", i,
3415 					rte_ixgbe_rxq_strings[stat].name);
3416 				count++;
3417 			}
3418 		}
3419 
3420 		/* TX Priority Stats */
3421 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3422 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3423 				snprintf(xstats_names[count].name,
3424 					sizeof(xstats_names[count].name),
3425 					"tx_priority%u_%s", i,
3426 					rte_ixgbe_txq_strings[stat].name);
3427 				count++;
3428 			}
3429 		}
3430 	}
3431 	return cnt_stats;
3432 }
3433 
3434 static int ixgbe_dev_xstats_get_names_by_id(
3435 	struct rte_eth_dev *dev,
3436 	const uint64_t *ids,
3437 	struct rte_eth_xstat_name *xstats_names,
3438 	unsigned int limit)
3439 {
3440 	if (!ids) {
3441 		const unsigned int cnt_stats = ixgbe_xstats_calc_num();
3442 		unsigned int stat, i, count;
3443 
3444 		if (xstats_names != NULL) {
3445 			count = 0;
3446 
3447 			/* Note: limit >= cnt_stats checked upstream
3448 			 * in rte_eth_xstats_names()
3449 			 */
3450 
3451 			/* Extended stats from ixgbe_hw_stats */
3452 			for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3453 				strlcpy(xstats_names[count].name,
3454 					rte_ixgbe_stats_strings[i].name,
3455 					sizeof(xstats_names[count].name));
3456 				count++;
3457 			}
3458 
3459 			/* MACsec Stats */
3460 			for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3461 				strlcpy(xstats_names[count].name,
3462 					rte_ixgbe_macsec_strings[i].name,
3463 					sizeof(xstats_names[count].name));
3464 				count++;
3465 			}
3466 
3467 			/* RX Priority Stats */
3468 			for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3469 				for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3470 					snprintf(xstats_names[count].name,
3471 					    sizeof(xstats_names[count].name),
3472 					    "rx_priority%u_%s", i,
3473 					    rte_ixgbe_rxq_strings[stat].name);
3474 					count++;
3475 				}
3476 			}
3477 
3478 			/* TX Priority Stats */
3479 			for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3480 				for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3481 					snprintf(xstats_names[count].name,
3482 					    sizeof(xstats_names[count].name),
3483 					    "tx_priority%u_%s", i,
3484 					    rte_ixgbe_txq_strings[stat].name);
3485 					count++;
3486 				}
3487 			}
3488 		}
3489 		return cnt_stats;
3490 	}
3491 
3492 	uint16_t i;
3493 	uint16_t size = ixgbe_xstats_calc_num();
3494 	struct rte_eth_xstat_name xstats_names_copy[size];
3495 
3496 	ixgbe_dev_xstats_get_names_by_id(dev, NULL, xstats_names_copy,
3497 			size);
3498 
3499 	for (i = 0; i < limit; i++) {
3500 		if (ids[i] >= size) {
3501 			PMD_INIT_LOG(ERR, "id value isn't valid");
3502 			return -1;
3503 		}
3504 		strcpy(xstats_names[i].name,
3505 				xstats_names_copy[ids[i]].name);
3506 	}
3507 	return limit;
3508 }
3509 
3510 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3511 	struct rte_eth_xstat_name *xstats_names, unsigned limit)
3512 {
3513 	unsigned i;
3514 
3515 	if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
3516 		return -ENOMEM;
3517 
3518 	if (xstats_names != NULL)
3519 		for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
3520 			strlcpy(xstats_names[i].name,
3521 				rte_ixgbevf_stats_strings[i].name,
3522 				sizeof(xstats_names[i].name));
3523 	return IXGBEVF_NB_XSTATS;
3524 }
3525 
3526 static int
3527 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3528 					 unsigned n)
3529 {
3530 	struct ixgbe_hw *hw =
3531 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3532 	struct ixgbe_hw_stats *hw_stats =
3533 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3534 	struct ixgbe_macsec_stats *macsec_stats =
3535 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3536 				dev->data->dev_private);
3537 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3538 	unsigned i, stat, count = 0;
3539 
3540 	count = ixgbe_xstats_calc_num();
3541 
3542 	if (n < count)
3543 		return count;
3544 
3545 	total_missed_rx = 0;
3546 	total_qbrc = 0;
3547 	total_qprc = 0;
3548 	total_qprdc = 0;
3549 
3550 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3551 			&total_qbrc, &total_qprc, &total_qprdc);
3552 
3553 	/* If this is a reset xstats is NULL, and we have cleared the
3554 	 * registers by reading them.
3555 	 */
3556 	if (!xstats)
3557 		return 0;
3558 
3559 	/* Extended stats from ixgbe_hw_stats */
3560 	count = 0;
3561 	for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3562 		xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3563 				rte_ixgbe_stats_strings[i].offset);
3564 		xstats[count].id = count;
3565 		count++;
3566 	}
3567 
3568 	/* MACsec Stats */
3569 	for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3570 		xstats[count].value = *(uint64_t *)(((char *)macsec_stats) +
3571 				rte_ixgbe_macsec_strings[i].offset);
3572 		xstats[count].id = count;
3573 		count++;
3574 	}
3575 
3576 	/* RX Priority Stats */
3577 	for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3578 		for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3579 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3580 					rte_ixgbe_rxq_strings[stat].offset +
3581 					(sizeof(uint64_t) * i));
3582 			xstats[count].id = count;
3583 			count++;
3584 		}
3585 	}
3586 
3587 	/* TX Priority Stats */
3588 	for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3589 		for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3590 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3591 					rte_ixgbe_txq_strings[stat].offset +
3592 					(sizeof(uint64_t) * i));
3593 			xstats[count].id = count;
3594 			count++;
3595 		}
3596 	}
3597 	return count;
3598 }
3599 
3600 static int
3601 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
3602 		uint64_t *values, unsigned int n)
3603 {
3604 	if (!ids) {
3605 		struct ixgbe_hw *hw =
3606 				IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3607 		struct ixgbe_hw_stats *hw_stats =
3608 				IXGBE_DEV_PRIVATE_TO_STATS(
3609 						dev->data->dev_private);
3610 		struct ixgbe_macsec_stats *macsec_stats =
3611 				IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3612 					dev->data->dev_private);
3613 		uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3614 		unsigned int i, stat, count = 0;
3615 
3616 		count = ixgbe_xstats_calc_num();
3617 
3618 		if (!ids && n < count)
3619 			return count;
3620 
3621 		total_missed_rx = 0;
3622 		total_qbrc = 0;
3623 		total_qprc = 0;
3624 		total_qprdc = 0;
3625 
3626 		ixgbe_read_stats_registers(hw, hw_stats, macsec_stats,
3627 				&total_missed_rx, &total_qbrc, &total_qprc,
3628 				&total_qprdc);
3629 
3630 		/* If this is a reset xstats is NULL, and we have cleared the
3631 		 * registers by reading them.
3632 		 */
3633 		if (!ids && !values)
3634 			return 0;
3635 
3636 		/* Extended stats from ixgbe_hw_stats */
3637 		count = 0;
3638 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3639 			values[count] = *(uint64_t *)(((char *)hw_stats) +
3640 					rte_ixgbe_stats_strings[i].offset);
3641 			count++;
3642 		}
3643 
3644 		/* MACsec Stats */
3645 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3646 			values[count] = *(uint64_t *)(((char *)macsec_stats) +
3647 					rte_ixgbe_macsec_strings[i].offset);
3648 			count++;
3649 		}
3650 
3651 		/* RX Priority Stats */
3652 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3653 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3654 				values[count] =
3655 					*(uint64_t *)(((char *)hw_stats) +
3656 					rte_ixgbe_rxq_strings[stat].offset +
3657 					(sizeof(uint64_t) * i));
3658 				count++;
3659 			}
3660 		}
3661 
3662 		/* TX Priority Stats */
3663 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3664 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3665 				values[count] =
3666 					*(uint64_t *)(((char *)hw_stats) +
3667 					rte_ixgbe_txq_strings[stat].offset +
3668 					(sizeof(uint64_t) * i));
3669 				count++;
3670 			}
3671 		}
3672 		return count;
3673 	}
3674 
3675 	uint16_t i;
3676 	uint16_t size = ixgbe_xstats_calc_num();
3677 	uint64_t values_copy[size];
3678 
3679 	ixgbe_dev_xstats_get_by_id(dev, NULL, values_copy, size);
3680 
3681 	for (i = 0; i < n; i++) {
3682 		if (ids[i] >= size) {
3683 			PMD_INIT_LOG(ERR, "id value isn't valid");
3684 			return -1;
3685 		}
3686 		values[i] = values_copy[ids[i]];
3687 	}
3688 	return n;
3689 }
3690 
3691 static int
3692 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
3693 {
3694 	struct ixgbe_hw_stats *stats =
3695 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3696 	struct ixgbe_macsec_stats *macsec_stats =
3697 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3698 				dev->data->dev_private);
3699 
3700 	unsigned count = ixgbe_xstats_calc_num();
3701 
3702 	/* HW registers are cleared on read */
3703 	ixgbe_dev_xstats_get(dev, NULL, count);
3704 
3705 	/* Reset software totals */
3706 	memset(stats, 0, sizeof(*stats));
3707 	memset(macsec_stats, 0, sizeof(*macsec_stats));
3708 
3709 	return 0;
3710 }
3711 
3712 static void
3713 ixgbevf_update_stats(struct rte_eth_dev *dev)
3714 {
3715 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3716 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3717 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3718 
3719 	/* Good Rx packet, include VF loopback */
3720 	UPDATE_VF_STAT(IXGBE_VFGPRC,
3721 	    hw_stats->last_vfgprc, hw_stats->vfgprc);
3722 
3723 	/* Good Rx octets, include VF loopback */
3724 	UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
3725 	    hw_stats->last_vfgorc, hw_stats->vfgorc);
3726 
3727 	/* Good Tx packet, include VF loopback */
3728 	UPDATE_VF_STAT(IXGBE_VFGPTC,
3729 	    hw_stats->last_vfgptc, hw_stats->vfgptc);
3730 
3731 	/* Good Tx octets, include VF loopback */
3732 	UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
3733 	    hw_stats->last_vfgotc, hw_stats->vfgotc);
3734 
3735 	/* Rx Multicst Packet */
3736 	UPDATE_VF_STAT(IXGBE_VFMPRC,
3737 	    hw_stats->last_vfmprc, hw_stats->vfmprc);
3738 }
3739 
3740 static int
3741 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3742 		       unsigned n)
3743 {
3744 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3745 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3746 	unsigned i;
3747 
3748 	if (n < IXGBEVF_NB_XSTATS)
3749 		return IXGBEVF_NB_XSTATS;
3750 
3751 	ixgbevf_update_stats(dev);
3752 
3753 	if (!xstats)
3754 		return 0;
3755 
3756 	/* Extended stats */
3757 	for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
3758 		xstats[i].id = i;
3759 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
3760 			rte_ixgbevf_stats_strings[i].offset);
3761 	}
3762 
3763 	return IXGBEVF_NB_XSTATS;
3764 }
3765 
3766 static int
3767 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3768 {
3769 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3770 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3771 
3772 	ixgbevf_update_stats(dev);
3773 
3774 	if (stats == NULL)
3775 		return -EINVAL;
3776 
3777 	stats->ipackets = hw_stats->vfgprc;
3778 	stats->ibytes = hw_stats->vfgorc;
3779 	stats->opackets = hw_stats->vfgptc;
3780 	stats->obytes = hw_stats->vfgotc;
3781 	return 0;
3782 }
3783 
3784 static int
3785 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3786 {
3787 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3788 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3789 
3790 	/* Sync HW register to the last stats */
3791 	ixgbevf_dev_stats_get(dev, NULL);
3792 
3793 	/* reset HW current stats*/
3794 	hw_stats->vfgprc = 0;
3795 	hw_stats->vfgorc = 0;
3796 	hw_stats->vfgptc = 0;
3797 	hw_stats->vfgotc = 0;
3798 	hw_stats->vfmprc = 0;
3799 
3800 	return 0;
3801 }
3802 
3803 static int
3804 ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
3805 {
3806 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3807 	u16 eeprom_verh, eeprom_verl;
3808 	u32 etrack_id;
3809 	int ret;
3810 
3811 	ixgbe_read_eeprom(hw, 0x2e, &eeprom_verh);
3812 	ixgbe_read_eeprom(hw, 0x2d, &eeprom_verl);
3813 
3814 	etrack_id = (eeprom_verh << 16) | eeprom_verl;
3815 	ret = snprintf(fw_version, fw_size, "0x%08x", etrack_id);
3816 	if (ret < 0)
3817 		return -EINVAL;
3818 
3819 	ret += 1; /* add the size of '\0' */
3820 	if (fw_size < (size_t)ret)
3821 		return ret;
3822 	else
3823 		return 0;
3824 }
3825 
3826 static int
3827 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3828 {
3829 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3830 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3831 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3832 
3833 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3834 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3835 	if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3836 		/*
3837 		 * When DCB/VT is off, maximum number of queues changes,
3838 		 * except for 82598EB, which remains constant.
3839 		 */
3840 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_NONE &&
3841 				hw->mac.type != ixgbe_mac_82598EB)
3842 			dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3843 	}
3844 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3845 	dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3846 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3847 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3848 	dev_info->max_vfs = pci_dev->max_vfs;
3849 	if (hw->mac.type == ixgbe_mac_82598EB)
3850 		dev_info->max_vmdq_pools = RTE_ETH_16_POOLS;
3851 	else
3852 		dev_info->max_vmdq_pools = RTE_ETH_64_POOLS;
3853 	dev_info->max_mtu =  dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3854 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
3855 	dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3856 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3857 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3858 				     dev_info->rx_queue_offload_capa);
3859 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3860 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3861 
3862 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3863 		.rx_thresh = {
3864 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3865 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3866 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3867 		},
3868 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3869 		.rx_drop_en = 0,
3870 		.offloads = 0,
3871 	};
3872 
3873 	dev_info->default_txconf = (struct rte_eth_txconf) {
3874 		.tx_thresh = {
3875 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3876 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3877 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3878 		},
3879 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3880 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3881 		.offloads = 0,
3882 	};
3883 
3884 	dev_info->rx_desc_lim = rx_desc_lim;
3885 	dev_info->tx_desc_lim = tx_desc_lim;
3886 
3887 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3888 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3889 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3890 
3891 	dev_info->speed_capa = RTE_ETH_LINK_SPEED_1G | RTE_ETH_LINK_SPEED_10G;
3892 	if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
3893 			hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
3894 		dev_info->speed_capa = RTE_ETH_LINK_SPEED_10M |
3895 			RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G;
3896 
3897 	if (hw->mac.type == ixgbe_mac_X540 ||
3898 	    hw->mac.type == ixgbe_mac_X540_vf ||
3899 	    hw->mac.type == ixgbe_mac_X550 ||
3900 	    hw->mac.type == ixgbe_mac_X550_vf) {
3901 		dev_info->speed_capa |= RTE_ETH_LINK_SPEED_100M;
3902 	}
3903 	if (hw->mac.type == ixgbe_mac_X550) {
3904 		dev_info->speed_capa |= RTE_ETH_LINK_SPEED_2_5G;
3905 		dev_info->speed_capa |= RTE_ETH_LINK_SPEED_5G;
3906 	}
3907 
3908 	/* Driver-preferred Rx/Tx parameters */
3909 	dev_info->default_rxportconf.burst_size = 32;
3910 	dev_info->default_txportconf.burst_size = 32;
3911 	dev_info->default_rxportconf.nb_queues = 1;
3912 	dev_info->default_txportconf.nb_queues = 1;
3913 	dev_info->default_rxportconf.ring_size = 256;
3914 	dev_info->default_txportconf.ring_size = 256;
3915 
3916 	return 0;
3917 }
3918 
3919 static const uint32_t *
3920 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3921 {
3922 	static const uint32_t ptypes[] = {
3923 		/* For non-vec functions,
3924 		 * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3925 		 * for vec functions,
3926 		 * refers to _recv_raw_pkts_vec().
3927 		 */
3928 		RTE_PTYPE_L2_ETHER,
3929 		RTE_PTYPE_L3_IPV4,
3930 		RTE_PTYPE_L3_IPV4_EXT,
3931 		RTE_PTYPE_L3_IPV6,
3932 		RTE_PTYPE_L3_IPV6_EXT,
3933 		RTE_PTYPE_L4_SCTP,
3934 		RTE_PTYPE_L4_TCP,
3935 		RTE_PTYPE_L4_UDP,
3936 		RTE_PTYPE_TUNNEL_IP,
3937 		RTE_PTYPE_INNER_L3_IPV6,
3938 		RTE_PTYPE_INNER_L3_IPV6_EXT,
3939 		RTE_PTYPE_INNER_L4_TCP,
3940 		RTE_PTYPE_INNER_L4_UDP,
3941 		RTE_PTYPE_UNKNOWN
3942 	};
3943 
3944 	if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3945 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3946 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3947 	    dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3948 		return ptypes;
3949 
3950 #if defined(RTE_ARCH_X86) || defined(__ARM_NEON)
3951 	if (dev->rx_pkt_burst == ixgbe_recv_pkts_vec ||
3952 	    dev->rx_pkt_burst == ixgbe_recv_scattered_pkts_vec)
3953 		return ptypes;
3954 #endif
3955 	return NULL;
3956 }
3957 
3958 static int
3959 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3960 		     struct rte_eth_dev_info *dev_info)
3961 {
3962 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3963 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3964 
3965 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3966 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3967 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3968 	dev_info->max_rx_pktlen = 9728; /* includes CRC, cf MAXFRS reg */
3969 	dev_info->max_mtu = dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3970 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3971 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3972 	dev_info->max_vfs = pci_dev->max_vfs;
3973 	if (hw->mac.type == ixgbe_mac_82598EB)
3974 		dev_info->max_vmdq_pools = RTE_ETH_16_POOLS;
3975 	else
3976 		dev_info->max_vmdq_pools = RTE_ETH_64_POOLS;
3977 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3978 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3979 				     dev_info->rx_queue_offload_capa);
3980 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3981 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3982 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3983 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3984 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3985 
3986 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3987 		.rx_thresh = {
3988 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3989 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3990 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3991 		},
3992 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3993 		.rx_drop_en = 0,
3994 		.offloads = 0,
3995 	};
3996 
3997 	dev_info->default_txconf = (struct rte_eth_txconf) {
3998 		.tx_thresh = {
3999 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
4000 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
4001 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
4002 		},
4003 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
4004 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
4005 		.offloads = 0,
4006 	};
4007 
4008 	dev_info->rx_desc_lim = rx_desc_lim;
4009 	dev_info->tx_desc_lim = tx_desc_lim;
4010 
4011 	return 0;
4012 }
4013 
4014 static int
4015 ixgbevf_check_link(struct ixgbe_hw *hw, ixgbe_link_speed *speed,
4016 		   bool *link_up, int wait_to_complete)
4017 {
4018 	struct ixgbe_adapter *adapter = container_of(hw,
4019 						     struct ixgbe_adapter, hw);
4020 	struct ixgbe_mbx_info *mbx = &hw->mbx;
4021 	struct ixgbe_mac_info *mac = &hw->mac;
4022 	uint32_t links_reg, in_msg;
4023 	int ret_val = 0;
4024 
4025 	/* If we were hit with a reset drop the link */
4026 	if (!mbx->ops.check_for_rst(hw, 0) || !mbx->timeout)
4027 		mac->get_link_status = true;
4028 
4029 	if (!mac->get_link_status)
4030 		goto out;
4031 
4032 	/* if link status is down no point in checking to see if pf is up */
4033 	links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4034 	if (!(links_reg & IXGBE_LINKS_UP))
4035 		goto out;
4036 
4037 	/* for SFP+ modules and DA cables on 82599 it can take up to 500usecs
4038 	 * before the link status is correct
4039 	 */
4040 	if (mac->type == ixgbe_mac_82599_vf && wait_to_complete) {
4041 		int i;
4042 
4043 		for (i = 0; i < 5; i++) {
4044 			rte_delay_us(100);
4045 			links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4046 
4047 			if (!(links_reg & IXGBE_LINKS_UP))
4048 				goto out;
4049 		}
4050 	}
4051 
4052 	switch (links_reg & IXGBE_LINKS_SPEED_82599) {
4053 	case IXGBE_LINKS_SPEED_10G_82599:
4054 		*speed = IXGBE_LINK_SPEED_10GB_FULL;
4055 		if (hw->mac.type >= ixgbe_mac_X550) {
4056 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4057 				*speed = IXGBE_LINK_SPEED_2_5GB_FULL;
4058 		}
4059 		break;
4060 	case IXGBE_LINKS_SPEED_1G_82599:
4061 		*speed = IXGBE_LINK_SPEED_1GB_FULL;
4062 		break;
4063 	case IXGBE_LINKS_SPEED_100_82599:
4064 		*speed = IXGBE_LINK_SPEED_100_FULL;
4065 		if (hw->mac.type == ixgbe_mac_X550) {
4066 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4067 				*speed = IXGBE_LINK_SPEED_5GB_FULL;
4068 		}
4069 		break;
4070 	case IXGBE_LINKS_SPEED_10_X550EM_A:
4071 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
4072 		/* Since Reserved in older MAC's */
4073 		if (hw->mac.type >= ixgbe_mac_X550)
4074 			*speed = IXGBE_LINK_SPEED_10_FULL;
4075 		break;
4076 	default:
4077 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
4078 	}
4079 
4080 	if (wait_to_complete == 0 && adapter->pflink_fullchk == 0) {
4081 		if (*speed == IXGBE_LINK_SPEED_UNKNOWN)
4082 			mac->get_link_status = true;
4083 		else
4084 			mac->get_link_status = false;
4085 
4086 		goto out;
4087 	}
4088 
4089 	/* if the read failed it could just be a mailbox collision, best wait
4090 	 * until we are called again and don't report an error
4091 	 */
4092 	if (mbx->ops.read(hw, &in_msg, 1, 0))
4093 		goto out;
4094 
4095 	if (!(in_msg & IXGBE_VT_MSGTYPE_CTS)) {
4096 		/* msg is not CTS and is NACK we must have lost CTS status */
4097 		if (in_msg & IXGBE_VT_MSGTYPE_NACK)
4098 			mac->get_link_status = false;
4099 		goto out;
4100 	}
4101 
4102 	/* the pf is talking, if we timed out in the past we reinit */
4103 	if (!mbx->timeout) {
4104 		ret_val = -1;
4105 		goto out;
4106 	}
4107 
4108 	/* if we passed all the tests above then the link is up and we no
4109 	 * longer need to check for link
4110 	 */
4111 	mac->get_link_status = false;
4112 
4113 out:
4114 	*link_up = !mac->get_link_status;
4115 	return ret_val;
4116 }
4117 
4118 /*
4119  * If @timeout_ms was 0, it means that it will not return until link complete.
4120  * It returns 1 on complete, return 0 on timeout.
4121  */
4122 static int
4123 ixgbe_dev_wait_setup_link_complete(struct rte_eth_dev *dev, uint32_t timeout_ms)
4124 {
4125 #define WARNING_TIMEOUT    9000 /* 9s  in total */
4126 	struct ixgbe_adapter *ad = dev->data->dev_private;
4127 	uint32_t timeout = timeout_ms ? timeout_ms : WARNING_TIMEOUT;
4128 
4129 	while (rte_atomic32_read(&ad->link_thread_running)) {
4130 		msec_delay(1);
4131 		timeout--;
4132 
4133 		if (timeout_ms) {
4134 			if (!timeout)
4135 				return 0;
4136 		} else if (!timeout) {
4137 			/* It will not return until link complete */
4138 			timeout = WARNING_TIMEOUT;
4139 			PMD_DRV_LOG(ERR, "IXGBE link thread not complete too long time!");
4140 		}
4141 	}
4142 
4143 	return 1;
4144 }
4145 
4146 static void *
4147 ixgbe_dev_setup_link_thread_handler(void *param)
4148 {
4149 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4150 	struct ixgbe_adapter *ad = dev->data->dev_private;
4151 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4152 	struct ixgbe_interrupt *intr =
4153 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4154 	u32 speed;
4155 	bool autoneg = false;
4156 
4157 	pthread_detach(pthread_self());
4158 	speed = hw->phy.autoneg_advertised;
4159 	if (!speed)
4160 		ixgbe_get_link_capabilities(hw, &speed, &autoneg);
4161 
4162 	ixgbe_setup_link(hw, speed, true);
4163 
4164 	intr->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4165 	rte_atomic32_clear(&ad->link_thread_running);
4166 	return NULL;
4167 }
4168 
4169 /*
4170  * In freebsd environment, nic_uio drivers do not support interrupts,
4171  * rte_intr_callback_register() will fail to register interrupts.
4172  * We can not make link status to change from down to up by interrupt
4173  * callback. So we need to wait for the controller to acquire link
4174  * when ports start.
4175  * It returns 0 on link up.
4176  */
4177 static int
4178 ixgbe_wait_for_link_up(struct ixgbe_hw *hw)
4179 {
4180 #ifdef RTE_EXEC_ENV_FREEBSD
4181 	int err, i;
4182 	bool link_up = false;
4183 	uint32_t speed = 0;
4184 	const int nb_iter = 25;
4185 
4186 	for (i = 0; i < nb_iter; i++) {
4187 		err = ixgbe_check_link(hw, &speed, &link_up, 0);
4188 		if (err)
4189 			return err;
4190 		if (link_up)
4191 			return 0;
4192 		msec_delay(200);
4193 	}
4194 
4195 	return 0;
4196 #else
4197 	RTE_SET_USED(hw);
4198 	return 0;
4199 #endif
4200 }
4201 
4202 /* return 0 means link status changed, -1 means not changed */
4203 int
4204 ixgbe_dev_link_update_share(struct rte_eth_dev *dev,
4205 			    int wait_to_complete, int vf)
4206 {
4207 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4208 	struct ixgbe_adapter *ad = dev->data->dev_private;
4209 	struct rte_eth_link link;
4210 	ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
4211 	struct ixgbe_interrupt *intr =
4212 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4213 	bool link_up;
4214 	int diag;
4215 	int wait = 1;
4216 	u32 esdp_reg;
4217 
4218 	memset(&link, 0, sizeof(link));
4219 	link.link_status = RTE_ETH_LINK_DOWN;
4220 	link.link_speed = RTE_ETH_SPEED_NUM_NONE;
4221 	link.link_duplex = RTE_ETH_LINK_HALF_DUPLEX;
4222 	link.link_autoneg = !(dev->data->dev_conf.link_speeds &
4223 			RTE_ETH_LINK_SPEED_FIXED);
4224 
4225 	hw->mac.get_link_status = true;
4226 
4227 	if (intr->flags & IXGBE_FLAG_NEED_LINK_CONFIG)
4228 		return rte_eth_linkstatus_set(dev, &link);
4229 
4230 	/* check if it needs to wait to complete, if lsc interrupt is enabled */
4231 	if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
4232 		wait = 0;
4233 
4234 /* BSD has no interrupt mechanism, so force NIC status synchronization. */
4235 #ifdef RTE_EXEC_ENV_FREEBSD
4236 	wait = 1;
4237 #endif
4238 
4239 	if (vf)
4240 		diag = ixgbevf_check_link(hw, &link_speed, &link_up, wait);
4241 	else
4242 		diag = ixgbe_check_link(hw, &link_speed, &link_up, wait);
4243 
4244 	if (diag != 0) {
4245 		link.link_speed = RTE_ETH_SPEED_NUM_100M;
4246 		link.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
4247 		return rte_eth_linkstatus_set(dev, &link);
4248 	}
4249 
4250 	if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4251 		esdp_reg = IXGBE_READ_REG(hw, IXGBE_ESDP);
4252 		if ((esdp_reg & IXGBE_ESDP_SDP3))
4253 			link_up = 0;
4254 	}
4255 
4256 	if (link_up == 0) {
4257 		if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4258 			ixgbe_dev_wait_setup_link_complete(dev, 0);
4259 			if (rte_atomic32_test_and_set(&ad->link_thread_running)) {
4260 				/* To avoid race condition between threads, set
4261 				 * the IXGBE_FLAG_NEED_LINK_CONFIG flag only
4262 				 * when there is no link thread running.
4263 				 */
4264 				intr->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
4265 				if (rte_ctrl_thread_create(&ad->link_thread_tid,
4266 					"ixgbe-link-handler",
4267 					NULL,
4268 					ixgbe_dev_setup_link_thread_handler,
4269 					dev) < 0) {
4270 					PMD_DRV_LOG(ERR,
4271 						"Create link thread failed!");
4272 					rte_atomic32_clear(&ad->link_thread_running);
4273 				}
4274 			} else {
4275 				PMD_DRV_LOG(ERR,
4276 					"Other link thread is running now!");
4277 			}
4278 		}
4279 		return rte_eth_linkstatus_set(dev, &link);
4280 	}
4281 
4282 	link.link_status = RTE_ETH_LINK_UP;
4283 	link.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
4284 
4285 	switch (link_speed) {
4286 	default:
4287 	case IXGBE_LINK_SPEED_UNKNOWN:
4288 		link.link_speed = RTE_ETH_SPEED_NUM_UNKNOWN;
4289 		break;
4290 
4291 	case IXGBE_LINK_SPEED_10_FULL:
4292 		link.link_speed = RTE_ETH_SPEED_NUM_10M;
4293 		break;
4294 
4295 	case IXGBE_LINK_SPEED_100_FULL:
4296 		link.link_speed = RTE_ETH_SPEED_NUM_100M;
4297 		break;
4298 
4299 	case IXGBE_LINK_SPEED_1GB_FULL:
4300 		link.link_speed = RTE_ETH_SPEED_NUM_1G;
4301 		break;
4302 
4303 	case IXGBE_LINK_SPEED_2_5GB_FULL:
4304 		link.link_speed = RTE_ETH_SPEED_NUM_2_5G;
4305 		break;
4306 
4307 	case IXGBE_LINK_SPEED_5GB_FULL:
4308 		link.link_speed = RTE_ETH_SPEED_NUM_5G;
4309 		break;
4310 
4311 	case IXGBE_LINK_SPEED_10GB_FULL:
4312 		link.link_speed = RTE_ETH_SPEED_NUM_10G;
4313 		break;
4314 	}
4315 
4316 	return rte_eth_linkstatus_set(dev, &link);
4317 }
4318 
4319 static int
4320 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4321 {
4322 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 0);
4323 }
4324 
4325 static int
4326 ixgbevf_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4327 {
4328 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 1);
4329 }
4330 
4331 static int
4332 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
4333 {
4334 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4335 	uint32_t fctrl;
4336 
4337 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4338 	fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4339 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4340 
4341 	return 0;
4342 }
4343 
4344 static int
4345 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
4346 {
4347 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4348 	uint32_t fctrl;
4349 
4350 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4351 	fctrl &= (~IXGBE_FCTRL_UPE);
4352 	if (dev->data->all_multicast == 1)
4353 		fctrl |= IXGBE_FCTRL_MPE;
4354 	else
4355 		fctrl &= (~IXGBE_FCTRL_MPE);
4356 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4357 
4358 	return 0;
4359 }
4360 
4361 static int
4362 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
4363 {
4364 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4365 	uint32_t fctrl;
4366 
4367 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4368 	fctrl |= IXGBE_FCTRL_MPE;
4369 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4370 
4371 	return 0;
4372 }
4373 
4374 static int
4375 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
4376 {
4377 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4378 	uint32_t fctrl;
4379 
4380 	if (dev->data->promiscuous == 1)
4381 		return 0; /* must remain in all_multicast mode */
4382 
4383 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4384 	fctrl &= (~IXGBE_FCTRL_MPE);
4385 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4386 
4387 	return 0;
4388 }
4389 
4390 /**
4391  * It clears the interrupt causes and enables the interrupt.
4392  * It will be called once only during nic initialized.
4393  *
4394  * @param dev
4395  *  Pointer to struct rte_eth_dev.
4396  * @param on
4397  *  Enable or Disable.
4398  *
4399  * @return
4400  *  - On success, zero.
4401  *  - On failure, a negative value.
4402  */
4403 static int
4404 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
4405 {
4406 	struct ixgbe_interrupt *intr =
4407 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4408 
4409 	ixgbe_dev_link_status_print(dev);
4410 	if (on)
4411 		intr->mask |= IXGBE_EICR_LSC;
4412 	else
4413 		intr->mask &= ~IXGBE_EICR_LSC;
4414 
4415 	return 0;
4416 }
4417 
4418 /**
4419  * It clears the interrupt causes and enables the interrupt.
4420  * It will be called once only during nic initialized.
4421  *
4422  * @param dev
4423  *  Pointer to struct rte_eth_dev.
4424  *
4425  * @return
4426  *  - On success, zero.
4427  *  - On failure, a negative value.
4428  */
4429 static int
4430 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
4431 {
4432 	struct ixgbe_interrupt *intr =
4433 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4434 
4435 	intr->mask |= IXGBE_EICR_RTX_QUEUE;
4436 
4437 	return 0;
4438 }
4439 
4440 /**
4441  * It clears the interrupt causes and enables the interrupt.
4442  * It will be called once only during nic initialized.
4443  *
4444  * @param dev
4445  *  Pointer to struct rte_eth_dev.
4446  *
4447  * @return
4448  *  - On success, zero.
4449  *  - On failure, a negative value.
4450  */
4451 static int
4452 ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev)
4453 {
4454 	struct ixgbe_interrupt *intr =
4455 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4456 
4457 	intr->mask |= IXGBE_EICR_LINKSEC;
4458 
4459 	return 0;
4460 }
4461 
4462 /*
4463  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
4464  *
4465  * @param dev
4466  *  Pointer to struct rte_eth_dev.
4467  *
4468  * @return
4469  *  - On success, zero.
4470  *  - On failure, a negative value.
4471  */
4472 static int
4473 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
4474 {
4475 	uint32_t eicr;
4476 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4477 	struct ixgbe_interrupt *intr =
4478 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4479 
4480 	/* clear all cause mask */
4481 	ixgbe_disable_intr(hw);
4482 
4483 	/* read-on-clear nic registers here */
4484 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4485 	PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
4486 
4487 	intr->flags = 0;
4488 
4489 	/* set flag for async link update */
4490 	if (eicr & IXGBE_EICR_LSC)
4491 		intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4492 
4493 	if (eicr & IXGBE_EICR_MAILBOX)
4494 		intr->flags |= IXGBE_FLAG_MAILBOX;
4495 
4496 	if (eicr & IXGBE_EICR_LINKSEC)
4497 		intr->flags |= IXGBE_FLAG_MACSEC;
4498 
4499 	if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
4500 	    hw->phy.type == ixgbe_phy_x550em_ext_t &&
4501 	    (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
4502 		intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
4503 
4504 	return 0;
4505 }
4506 
4507 /**
4508  * It gets and then prints the link status.
4509  *
4510  * @param dev
4511  *  Pointer to struct rte_eth_dev.
4512  *
4513  * @return
4514  *  - On success, zero.
4515  *  - On failure, a negative value.
4516  */
4517 static void
4518 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
4519 {
4520 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4521 	struct rte_eth_link link;
4522 
4523 	rte_eth_linkstatus_get(dev, &link);
4524 
4525 	if (link.link_status) {
4526 		PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
4527 					(int)(dev->data->port_id),
4528 					(unsigned)link.link_speed,
4529 			link.link_duplex == RTE_ETH_LINK_FULL_DUPLEX ?
4530 					"full-duplex" : "half-duplex");
4531 	} else {
4532 		PMD_INIT_LOG(INFO, " Port %d: Link Down",
4533 				(int)(dev->data->port_id));
4534 	}
4535 	PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
4536 				pci_dev->addr.domain,
4537 				pci_dev->addr.bus,
4538 				pci_dev->addr.devid,
4539 				pci_dev->addr.function);
4540 }
4541 
4542 /*
4543  * It executes link_update after knowing an interrupt occurred.
4544  *
4545  * @param dev
4546  *  Pointer to struct rte_eth_dev.
4547  *
4548  * @return
4549  *  - On success, zero.
4550  *  - On failure, a negative value.
4551  */
4552 static int
4553 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
4554 {
4555 	struct ixgbe_interrupt *intr =
4556 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4557 	int64_t timeout;
4558 	struct ixgbe_hw *hw =
4559 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4560 
4561 	PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
4562 
4563 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
4564 		ixgbe_pf_mbx_process(dev);
4565 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
4566 	}
4567 
4568 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4569 		ixgbe_handle_lasi(hw);
4570 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4571 	}
4572 
4573 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4574 		struct rte_eth_link link;
4575 
4576 		/* get the link status before link update, for predicting later */
4577 		rte_eth_linkstatus_get(dev, &link);
4578 
4579 		ixgbe_dev_link_update(dev, 0);
4580 
4581 		/* likely to up */
4582 		if (!link.link_status)
4583 			/* handle it 1 sec later, wait it being stable */
4584 			timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
4585 		/* likely to down */
4586 		else
4587 			/* handle it 4 sec later, wait it being stable */
4588 			timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
4589 
4590 		ixgbe_dev_link_status_print(dev);
4591 		if (rte_eal_alarm_set(timeout * 1000,
4592 				      ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
4593 			PMD_DRV_LOG(ERR, "Error setting alarm");
4594 		else {
4595 			/* remember original mask */
4596 			intr->mask_original = intr->mask;
4597 			/* only disable lsc interrupt */
4598 			intr->mask &= ~IXGBE_EIMS_LSC;
4599 		}
4600 	}
4601 
4602 	PMD_DRV_LOG(DEBUG, "enable intr immediately");
4603 	ixgbe_enable_intr(dev);
4604 
4605 	return 0;
4606 }
4607 
4608 /**
4609  * Interrupt handler which shall be registered for alarm callback for delayed
4610  * handling specific interrupt to wait for the stable nic state. As the
4611  * NIC interrupt state is not stable for ixgbe after link is just down,
4612  * it needs to wait 4 seconds to get the stable status.
4613  *
4614  * @param handle
4615  *  Pointer to interrupt handle.
4616  * @param param
4617  *  The address of parameter (struct rte_eth_dev *) registered before.
4618  *
4619  * @return
4620  *  void
4621  */
4622 static void
4623 ixgbe_dev_interrupt_delayed_handler(void *param)
4624 {
4625 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4626 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4627 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
4628 	struct ixgbe_interrupt *intr =
4629 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4630 	struct ixgbe_hw *hw =
4631 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4632 	uint32_t eicr;
4633 
4634 	ixgbe_disable_intr(hw);
4635 
4636 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4637 	if (eicr & IXGBE_EICR_MAILBOX)
4638 		ixgbe_pf_mbx_process(dev);
4639 
4640 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4641 		ixgbe_handle_lasi(hw);
4642 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4643 	}
4644 
4645 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4646 		ixgbe_dev_link_update(dev, 0);
4647 		intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4648 		ixgbe_dev_link_status_print(dev);
4649 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
4650 	}
4651 
4652 	if (intr->flags & IXGBE_FLAG_MACSEC) {
4653 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_MACSEC, NULL);
4654 		intr->flags &= ~IXGBE_FLAG_MACSEC;
4655 	}
4656 
4657 	/* restore original mask */
4658 	intr->mask = intr->mask_original;
4659 	intr->mask_original = 0;
4660 
4661 	PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
4662 	ixgbe_enable_intr(dev);
4663 	rte_intr_ack(intr_handle);
4664 }
4665 
4666 /**
4667  * Interrupt handler triggered by NIC  for handling
4668  * specific interrupt.
4669  *
4670  * @param handle
4671  *  Pointer to interrupt handle.
4672  * @param param
4673  *  The address of parameter (struct rte_eth_dev *) registered before.
4674  *
4675  * @return
4676  *  void
4677  */
4678 static void
4679 ixgbe_dev_interrupt_handler(void *param)
4680 {
4681 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4682 
4683 	ixgbe_dev_interrupt_get_status(dev);
4684 	ixgbe_dev_interrupt_action(dev);
4685 }
4686 
4687 static int
4688 ixgbe_dev_led_on(struct rte_eth_dev *dev)
4689 {
4690 	struct ixgbe_hw *hw;
4691 
4692 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4693 	return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4694 }
4695 
4696 static int
4697 ixgbe_dev_led_off(struct rte_eth_dev *dev)
4698 {
4699 	struct ixgbe_hw *hw;
4700 
4701 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4702 	return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4703 }
4704 
4705 static int
4706 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4707 {
4708 	struct ixgbe_hw *hw;
4709 	uint32_t mflcn_reg;
4710 	uint32_t fccfg_reg;
4711 	int rx_pause;
4712 	int tx_pause;
4713 
4714 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4715 
4716 	fc_conf->pause_time = hw->fc.pause_time;
4717 	fc_conf->high_water = hw->fc.high_water[0];
4718 	fc_conf->low_water = hw->fc.low_water[0];
4719 	fc_conf->send_xon = hw->fc.send_xon;
4720 	fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
4721 
4722 	/*
4723 	 * Return rx_pause status according to actual setting of
4724 	 * MFLCN register.
4725 	 */
4726 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4727 	if (mflcn_reg & IXGBE_MFLCN_PMCF)
4728 		fc_conf->mac_ctrl_frame_fwd = 1;
4729 	else
4730 		fc_conf->mac_ctrl_frame_fwd = 0;
4731 
4732 	if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
4733 		rx_pause = 1;
4734 	else
4735 		rx_pause = 0;
4736 
4737 	/*
4738 	 * Return tx_pause status according to actual setting of
4739 	 * FCCFG register.
4740 	 */
4741 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4742 	if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
4743 		tx_pause = 1;
4744 	else
4745 		tx_pause = 0;
4746 
4747 	if (rx_pause && tx_pause)
4748 		fc_conf->mode = RTE_ETH_FC_FULL;
4749 	else if (rx_pause)
4750 		fc_conf->mode = RTE_ETH_FC_RX_PAUSE;
4751 	else if (tx_pause)
4752 		fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
4753 	else
4754 		fc_conf->mode = RTE_ETH_FC_NONE;
4755 
4756 	return 0;
4757 }
4758 
4759 static int
4760 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4761 {
4762 	struct ixgbe_hw *hw;
4763 	struct ixgbe_adapter *adapter = dev->data->dev_private;
4764 	int err;
4765 	uint32_t rx_buf_size;
4766 	uint32_t max_high_water;
4767 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4768 		ixgbe_fc_none,
4769 		ixgbe_fc_rx_pause,
4770 		ixgbe_fc_tx_pause,
4771 		ixgbe_fc_full
4772 	};
4773 
4774 	PMD_INIT_FUNC_TRACE();
4775 
4776 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4777 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
4778 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4779 
4780 	/*
4781 	 * At least reserve one Ethernet frame for watermark
4782 	 * high_water/low_water in kilo bytes for ixgbe
4783 	 */
4784 	max_high_water = (rx_buf_size -
4785 			RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4786 	if ((fc_conf->high_water > max_high_water) ||
4787 		(fc_conf->high_water < fc_conf->low_water)) {
4788 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4789 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4790 		return -EINVAL;
4791 	}
4792 
4793 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
4794 	hw->fc.pause_time     = fc_conf->pause_time;
4795 	hw->fc.high_water[0]  = fc_conf->high_water;
4796 	hw->fc.low_water[0]   = fc_conf->low_water;
4797 	hw->fc.send_xon       = fc_conf->send_xon;
4798 	hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
4799 	adapter->mac_ctrl_frame_fwd = fc_conf->mac_ctrl_frame_fwd;
4800 
4801 	err = ixgbe_flow_ctrl_enable(dev, hw);
4802 	if (err < 0) {
4803 		PMD_INIT_LOG(ERR, "ixgbe_flow_ctrl_enable = 0x%x", err);
4804 		return -EIO;
4805 	}
4806 	return err;
4807 }
4808 
4809 /**
4810  *  ixgbe_pfc_enable_generic - Enable flow control
4811  *  @hw: pointer to hardware structure
4812  *  @tc_num: traffic class number
4813  *  Enable flow control according to the current settings.
4814  */
4815 static int
4816 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
4817 {
4818 	int ret_val = 0;
4819 	uint32_t mflcn_reg, fccfg_reg;
4820 	uint32_t reg;
4821 	uint32_t fcrtl, fcrth;
4822 	uint8_t i;
4823 	uint8_t nb_rx_en;
4824 
4825 	/* Validate the water mark configuration */
4826 	if (!hw->fc.pause_time) {
4827 		ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4828 		goto out;
4829 	}
4830 
4831 	/* Low water mark of zero causes XOFF floods */
4832 	if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
4833 		 /* High/Low water can not be 0 */
4834 		if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_water[tc_num])) {
4835 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4836 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4837 			goto out;
4838 		}
4839 
4840 		if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
4841 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4842 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4843 			goto out;
4844 		}
4845 	}
4846 	/* Negotiate the fc mode to use */
4847 	ixgbe_fc_autoneg(hw);
4848 
4849 	/* Disable any previous flow control settings */
4850 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4851 	mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
4852 
4853 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4854 	fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
4855 
4856 	switch (hw->fc.current_mode) {
4857 	case ixgbe_fc_none:
4858 		/*
4859 		 * If the count of enabled RX Priority Flow control >1,
4860 		 * and the TX pause can not be disabled
4861 		 */
4862 		nb_rx_en = 0;
4863 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4864 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4865 			if (reg & IXGBE_FCRTH_FCEN)
4866 				nb_rx_en++;
4867 		}
4868 		if (nb_rx_en > 1)
4869 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4870 		break;
4871 	case ixgbe_fc_rx_pause:
4872 		/*
4873 		 * Rx Flow control is enabled and Tx Flow control is
4874 		 * disabled by software override. Since there really
4875 		 * isn't a way to advertise that we are capable of RX
4876 		 * Pause ONLY, we will advertise that we support both
4877 		 * symmetric and asymmetric Rx PAUSE.  Later, we will
4878 		 * disable the adapter's ability to send PAUSE frames.
4879 		 */
4880 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4881 		/*
4882 		 * If the count of enabled RX Priority Flow control >1,
4883 		 * and the TX pause can not be disabled
4884 		 */
4885 		nb_rx_en = 0;
4886 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4887 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4888 			if (reg & IXGBE_FCRTH_FCEN)
4889 				nb_rx_en++;
4890 		}
4891 		if (nb_rx_en > 1)
4892 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4893 		break;
4894 	case ixgbe_fc_tx_pause:
4895 		/*
4896 		 * Tx Flow control is enabled, and Rx Flow control is
4897 		 * disabled by software override.
4898 		 */
4899 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4900 		break;
4901 	case ixgbe_fc_full:
4902 		/* Flow control (both Rx and Tx) is enabled by SW override. */
4903 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4904 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4905 		break;
4906 	default:
4907 		PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
4908 		ret_val = IXGBE_ERR_CONFIG;
4909 		goto out;
4910 	}
4911 
4912 	/* Set 802.3x based flow control settings. */
4913 	mflcn_reg |= IXGBE_MFLCN_DPF;
4914 	IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
4915 	IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
4916 
4917 	/* Set up and enable Rx high/low water mark thresholds, enable XON. */
4918 	if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
4919 		hw->fc.high_water[tc_num]) {
4920 		fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
4921 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
4922 		fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
4923 	} else {
4924 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
4925 		/*
4926 		 * In order to prevent Tx hangs when the internal Tx
4927 		 * switch is enabled we must set the high water mark
4928 		 * to the maximum FCRTH value.  This allows the Tx
4929 		 * switch to function even under heavy Rx workloads.
4930 		 */
4931 		fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
4932 	}
4933 	IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
4934 
4935 	/* Configure pause time (2 TCs per register) */
4936 	reg = hw->fc.pause_time * 0x00010001;
4937 	for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
4938 		IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
4939 
4940 	/* Configure flow control refresh threshold value */
4941 	IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
4942 
4943 out:
4944 	return ret_val;
4945 }
4946 
4947 static int
4948 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
4949 {
4950 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4951 	int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
4952 
4953 	if (hw->mac.type != ixgbe_mac_82598EB) {
4954 		ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
4955 	}
4956 	return ret_val;
4957 }
4958 
4959 static int
4960 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
4961 {
4962 	int err;
4963 	uint32_t rx_buf_size;
4964 	uint32_t max_high_water;
4965 	uint8_t tc_num;
4966 	uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
4967 	struct ixgbe_hw *hw =
4968 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4969 	struct ixgbe_dcb_config *dcb_config =
4970 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
4971 
4972 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4973 		ixgbe_fc_none,
4974 		ixgbe_fc_rx_pause,
4975 		ixgbe_fc_tx_pause,
4976 		ixgbe_fc_full
4977 	};
4978 
4979 	PMD_INIT_FUNC_TRACE();
4980 
4981 	ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
4982 	tc_num = map[pfc_conf->priority];
4983 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
4984 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4985 	/*
4986 	 * At least reserve one Ethernet frame for watermark
4987 	 * high_water/low_water in kilo bytes for ixgbe
4988 	 */
4989 	max_high_water = (rx_buf_size -
4990 			RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4991 	if ((pfc_conf->fc.high_water > max_high_water) ||
4992 	    (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
4993 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4994 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4995 		return -EINVAL;
4996 	}
4997 
4998 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
4999 	hw->fc.pause_time = pfc_conf->fc.pause_time;
5000 	hw->fc.send_xon = pfc_conf->fc.send_xon;
5001 	hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
5002 	hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
5003 
5004 	err = ixgbe_dcb_pfc_enable(dev, tc_num);
5005 
5006 	/* Not negotiated is not an error case */
5007 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
5008 		return 0;
5009 
5010 	PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
5011 	return -EIO;
5012 }
5013 
5014 static int
5015 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
5016 			  struct rte_eth_rss_reta_entry64 *reta_conf,
5017 			  uint16_t reta_size)
5018 {
5019 	uint16_t i, sp_reta_size;
5020 	uint8_t j, mask;
5021 	uint32_t reta, r;
5022 	uint16_t idx, shift;
5023 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5024 	struct rte_eth_dev_data *dev_data = dev->data;
5025 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5026 	uint32_t reta_reg;
5027 
5028 	PMD_INIT_FUNC_TRACE();
5029 
5030 	if (!dev_data->dev_started) {
5031 		PMD_DRV_LOG(ERR,
5032 			"port %d must be started before rss reta update",
5033 			 dev_data->port_id);
5034 		return -EIO;
5035 	}
5036 
5037 	if (!ixgbe_rss_update_sp(hw->mac.type)) {
5038 		PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
5039 			"NIC.");
5040 		return -ENOTSUP;
5041 	}
5042 
5043 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5044 	if (reta_size != sp_reta_size) {
5045 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5046 			"(%d) doesn't match the number hardware can supported "
5047 			"(%d)", reta_size, sp_reta_size);
5048 		return -EINVAL;
5049 	}
5050 
5051 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5052 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
5053 		shift = i % RTE_ETH_RETA_GROUP_SIZE;
5054 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5055 						IXGBE_4_BIT_MASK);
5056 		if (!mask)
5057 			continue;
5058 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5059 		if (mask == IXGBE_4_BIT_MASK)
5060 			r = 0;
5061 		else
5062 			r = IXGBE_READ_REG(hw, reta_reg);
5063 		for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5064 			if (mask & (0x1 << j))
5065 				reta |= reta_conf[idx].reta[shift + j] <<
5066 							(CHAR_BIT * j);
5067 			else
5068 				reta |= r & (IXGBE_8_BIT_MASK <<
5069 						(CHAR_BIT * j));
5070 		}
5071 		IXGBE_WRITE_REG(hw, reta_reg, reta);
5072 	}
5073 	adapter->rss_reta_updated = 1;
5074 
5075 	return 0;
5076 }
5077 
5078 static int
5079 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
5080 			 struct rte_eth_rss_reta_entry64 *reta_conf,
5081 			 uint16_t reta_size)
5082 {
5083 	uint16_t i, sp_reta_size;
5084 	uint8_t j, mask;
5085 	uint32_t reta;
5086 	uint16_t idx, shift;
5087 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5088 	uint32_t reta_reg;
5089 
5090 	PMD_INIT_FUNC_TRACE();
5091 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5092 	if (reta_size != sp_reta_size) {
5093 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5094 			"(%d) doesn't match the number hardware can supported "
5095 			"(%d)", reta_size, sp_reta_size);
5096 		return -EINVAL;
5097 	}
5098 
5099 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5100 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
5101 		shift = i % RTE_ETH_RETA_GROUP_SIZE;
5102 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5103 						IXGBE_4_BIT_MASK);
5104 		if (!mask)
5105 			continue;
5106 
5107 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5108 		reta = IXGBE_READ_REG(hw, reta_reg);
5109 		for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5110 			if (mask & (0x1 << j))
5111 				reta_conf[idx].reta[shift + j] =
5112 					((reta >> (CHAR_BIT * j)) &
5113 						IXGBE_8_BIT_MASK);
5114 		}
5115 	}
5116 
5117 	return 0;
5118 }
5119 
5120 static int
5121 ixgbe_add_rar(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
5122 				uint32_t index, uint32_t pool)
5123 {
5124 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5125 	uint32_t enable_addr = 1;
5126 
5127 	return ixgbe_set_rar(hw, index, mac_addr->addr_bytes,
5128 			     pool, enable_addr);
5129 }
5130 
5131 static void
5132 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
5133 {
5134 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5135 
5136 	ixgbe_clear_rar(hw, index);
5137 }
5138 
5139 static int
5140 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
5141 {
5142 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5143 
5144 	ixgbe_remove_rar(dev, 0);
5145 	ixgbe_add_rar(dev, addr, 0, pci_dev->max_vfs);
5146 
5147 	return 0;
5148 }
5149 
5150 static bool
5151 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
5152 {
5153 	if (strcmp(dev->device->driver->name, drv->driver.name))
5154 		return false;
5155 
5156 	return true;
5157 }
5158 
5159 bool
5160 is_ixgbe_supported(struct rte_eth_dev *dev)
5161 {
5162 	return is_device_supported(dev, &rte_ixgbe_pmd);
5163 }
5164 
5165 static int
5166 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
5167 {
5168 	uint32_t hlreg0;
5169 	uint32_t maxfrs;
5170 	struct ixgbe_hw *hw;
5171 	struct rte_eth_dev_info dev_info;
5172 	uint32_t frame_size = mtu + IXGBE_ETH_OVERHEAD;
5173 	int ret;
5174 
5175 	ret = ixgbe_dev_info_get(dev, &dev_info);
5176 	if (ret != 0)
5177 		return ret;
5178 
5179 	/* check that mtu is within the allowed range */
5180 	if (mtu < RTE_ETHER_MIN_MTU || frame_size > dev_info.max_rx_pktlen)
5181 		return -EINVAL;
5182 
5183 	/* If device is started, refuse mtu that requires the support of
5184 	 * scattered packets when this feature has not been enabled before.
5185 	 */
5186 	if (dev->data->dev_started && !dev->data->scattered_rx &&
5187 	    frame_size + 2 * RTE_VLAN_HLEN >
5188 			dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
5189 		PMD_INIT_LOG(ERR, "Stop port first.");
5190 		return -EINVAL;
5191 	}
5192 
5193 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5194 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
5195 
5196 	/* switch to jumbo mode if needed */
5197 	if (mtu > RTE_ETHER_MTU)
5198 		hlreg0 |= IXGBE_HLREG0_JUMBOEN;
5199 	else
5200 		hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
5201 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
5202 
5203 	maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
5204 	maxfrs &= 0x0000FFFF;
5205 	maxfrs |= (frame_size << 16);
5206 	IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
5207 
5208 	return 0;
5209 }
5210 
5211 /*
5212  * Virtual Function operations
5213  */
5214 static void
5215 ixgbevf_intr_disable(struct rte_eth_dev *dev)
5216 {
5217 	struct ixgbe_interrupt *intr =
5218 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5219 	struct ixgbe_hw *hw =
5220 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5221 
5222 	PMD_INIT_FUNC_TRACE();
5223 
5224 	/* Clear interrupt mask to stop from interrupts being generated */
5225 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
5226 
5227 	IXGBE_WRITE_FLUSH(hw);
5228 
5229 	/* Clear mask value. */
5230 	intr->mask = 0;
5231 }
5232 
5233 static void
5234 ixgbevf_intr_enable(struct rte_eth_dev *dev)
5235 {
5236 	struct ixgbe_interrupt *intr =
5237 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5238 	struct ixgbe_hw *hw =
5239 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5240 
5241 	PMD_INIT_FUNC_TRACE();
5242 
5243 	/* VF enable interrupt autoclean */
5244 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
5245 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
5246 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
5247 
5248 	IXGBE_WRITE_FLUSH(hw);
5249 
5250 	/* Save IXGBE_VTEIMS value to mask. */
5251 	intr->mask = IXGBE_VF_IRQ_ENABLE_MASK;
5252 }
5253 
5254 static int
5255 ixgbevf_dev_configure(struct rte_eth_dev *dev)
5256 {
5257 	struct rte_eth_conf *conf = &dev->data->dev_conf;
5258 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5259 
5260 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
5261 		     dev->data->port_id);
5262 
5263 	if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
5264 		dev->data->dev_conf.rxmode.offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
5265 
5266 	/*
5267 	 * VF has no ability to enable/disable HW CRC
5268 	 * Keep the persistent behavior the same as Host PF
5269 	 */
5270 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
5271 	if (conf->rxmode.offloads & RTE_ETH_RX_OFFLOAD_KEEP_CRC) {
5272 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
5273 		conf->rxmode.offloads &= ~RTE_ETH_RX_OFFLOAD_KEEP_CRC;
5274 	}
5275 #else
5276 	if (!(conf->rxmode.offloads & RTE_ETH_RX_OFFLOAD_KEEP_CRC)) {
5277 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
5278 		conf->rxmode.offloads |= RTE_ETH_RX_OFFLOAD_KEEP_CRC;
5279 	}
5280 #endif
5281 
5282 	/*
5283 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
5284 	 * allocation or vector Rx preconditions we will reset it.
5285 	 */
5286 	adapter->rx_bulk_alloc_allowed = true;
5287 	adapter->rx_vec_allowed = true;
5288 
5289 	return 0;
5290 }
5291 
5292 static int
5293 ixgbevf_dev_start(struct rte_eth_dev *dev)
5294 {
5295 	struct ixgbe_hw *hw =
5296 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5297 	uint32_t intr_vector = 0;
5298 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5299 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5300 
5301 	int err, mask = 0;
5302 
5303 	PMD_INIT_FUNC_TRACE();
5304 
5305 	/* Stop the link setup handler before resetting the HW. */
5306 	ixgbe_dev_wait_setup_link_complete(dev, 0);
5307 
5308 	err = hw->mac.ops.reset_hw(hw);
5309 
5310 	/**
5311 	 * In this case, reuses the MAC address assigned by VF
5312 	 * initialization.
5313 	 */
5314 	if (err != IXGBE_SUCCESS && err != IXGBE_ERR_INVALID_MAC_ADDR) {
5315 		PMD_INIT_LOG(ERR, "Unable to reset vf hardware (%d)", err);
5316 		return err;
5317 	}
5318 
5319 	hw->mac.get_link_status = true;
5320 
5321 	/* negotiate mailbox API version to use with the PF. */
5322 	ixgbevf_negotiate_api(hw);
5323 
5324 	ixgbevf_dev_tx_init(dev);
5325 
5326 	/* This can fail when allocating mbufs for descriptor rings */
5327 	err = ixgbevf_dev_rx_init(dev);
5328 	if (err) {
5329 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
5330 		ixgbe_dev_clear_queues(dev);
5331 		return err;
5332 	}
5333 
5334 	/* Set vfta */
5335 	ixgbevf_set_vfta_all(dev, 1);
5336 
5337 	/* Set HW strip */
5338 	mask = RTE_ETH_VLAN_STRIP_MASK | RTE_ETH_VLAN_FILTER_MASK |
5339 		RTE_ETH_VLAN_EXTEND_MASK;
5340 	err = ixgbevf_vlan_offload_config(dev, mask);
5341 	if (err) {
5342 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
5343 		ixgbe_dev_clear_queues(dev);
5344 		return err;
5345 	}
5346 
5347 	ixgbevf_dev_rxtx_start(dev);
5348 
5349 	/* check and configure queue intr-vector mapping */
5350 	if (rte_intr_cap_multiple(intr_handle) &&
5351 	    dev->data->dev_conf.intr_conf.rxq) {
5352 		/* According to datasheet, only vector 0/1/2 can be used,
5353 		 * now only one vector is used for Rx queue
5354 		 */
5355 		intr_vector = 1;
5356 		if (rte_intr_efd_enable(intr_handle, intr_vector)) {
5357 			ixgbe_dev_clear_queues(dev);
5358 			return -1;
5359 		}
5360 	}
5361 
5362 	if (rte_intr_dp_is_en(intr_handle)) {
5363 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
5364 						   dev->data->nb_rx_queues)) {
5365 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
5366 				     " intr_vec", dev->data->nb_rx_queues);
5367 			ixgbe_dev_clear_queues(dev);
5368 			return -ENOMEM;
5369 		}
5370 	}
5371 	ixgbevf_configure_msix(dev);
5372 
5373 	/* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
5374 	 * is mapped to VFIO vector 0 in eth_ixgbevf_dev_init( ).
5375 	 * If previous VFIO interrupt mapping setting in eth_ixgbevf_dev_init( )
5376 	 * is not cleared, it will fail when following rte_intr_enable( ) tries
5377 	 * to map Rx queue interrupt to other VFIO vectors.
5378 	 * So clear uio/vfio intr/evevnfd first to avoid failure.
5379 	 */
5380 	rte_intr_disable(intr_handle);
5381 
5382 	rte_intr_enable(intr_handle);
5383 
5384 	/* Re-enable interrupt for VF */
5385 	ixgbevf_intr_enable(dev);
5386 
5387 	/*
5388 	 * Update link status right before return, because it may
5389 	 * start link configuration process in a separate thread.
5390 	 */
5391 	ixgbevf_dev_link_update(dev, 0);
5392 
5393 	hw->adapter_stopped = false;
5394 
5395 	return 0;
5396 }
5397 
5398 static int
5399 ixgbevf_dev_stop(struct rte_eth_dev *dev)
5400 {
5401 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5402 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5403 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5404 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5405 
5406 	if (hw->adapter_stopped)
5407 		return 0;
5408 
5409 	PMD_INIT_FUNC_TRACE();
5410 
5411 	ixgbe_dev_wait_setup_link_complete(dev, 0);
5412 
5413 	ixgbevf_intr_disable(dev);
5414 
5415 	dev->data->dev_started = 0;
5416 	hw->adapter_stopped = 1;
5417 	ixgbe_stop_adapter(hw);
5418 
5419 	/*
5420 	  * Clear what we set, but we still keep shadow_vfta to
5421 	  * restore after device starts
5422 	  */
5423 	ixgbevf_set_vfta_all(dev, 0);
5424 
5425 	/* Clear stored conf */
5426 	dev->data->scattered_rx = 0;
5427 
5428 	ixgbe_dev_clear_queues(dev);
5429 
5430 	/* Clean datapath event and queue/vec mapping */
5431 	rte_intr_efd_disable(intr_handle);
5432 	rte_intr_vec_list_free(intr_handle);
5433 
5434 	adapter->rss_reta_updated = 0;
5435 
5436 	return 0;
5437 }
5438 
5439 static int
5440 ixgbevf_dev_close(struct rte_eth_dev *dev)
5441 {
5442 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5443 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5444 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5445 	int ret;
5446 
5447 	PMD_INIT_FUNC_TRACE();
5448 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5449 		return 0;
5450 
5451 	ixgbe_reset_hw(hw);
5452 
5453 	ret = ixgbevf_dev_stop(dev);
5454 
5455 	ixgbe_dev_free_queues(dev);
5456 
5457 	/**
5458 	 * Remove the VF MAC address ro ensure
5459 	 * that the VF traffic goes to the PF
5460 	 * after stop, close and detach of the VF
5461 	 **/
5462 	ixgbevf_remove_mac_addr(dev, 0);
5463 
5464 	rte_intr_disable(intr_handle);
5465 	rte_intr_callback_unregister(intr_handle,
5466 				     ixgbevf_dev_interrupt_handler, dev);
5467 
5468 	return ret;
5469 }
5470 
5471 /*
5472  * Reset VF device
5473  */
5474 static int
5475 ixgbevf_dev_reset(struct rte_eth_dev *dev)
5476 {
5477 	int ret;
5478 
5479 	ret = eth_ixgbevf_dev_uninit(dev);
5480 	if (ret)
5481 		return ret;
5482 
5483 	ret = eth_ixgbevf_dev_init(dev);
5484 
5485 	return ret;
5486 }
5487 
5488 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
5489 {
5490 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5491 	struct ixgbe_vfta *shadow_vfta =
5492 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5493 	int i = 0, j = 0, vfta = 0, mask = 1;
5494 
5495 	for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
5496 		vfta = shadow_vfta->vfta[i];
5497 		if (vfta) {
5498 			mask = 1;
5499 			for (j = 0; j < 32; j++) {
5500 				if (vfta & mask)
5501 					ixgbe_set_vfta(hw, (i<<5)+j, 0,
5502 						       on, false);
5503 				mask <<= 1;
5504 			}
5505 		}
5506 	}
5507 
5508 }
5509 
5510 static int
5511 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
5512 {
5513 	struct ixgbe_hw *hw =
5514 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5515 	struct ixgbe_vfta *shadow_vfta =
5516 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5517 	uint32_t vid_idx = 0;
5518 	uint32_t vid_bit = 0;
5519 	int ret = 0;
5520 
5521 	PMD_INIT_FUNC_TRACE();
5522 
5523 	/* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
5524 	ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
5525 	if (ret) {
5526 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
5527 		return ret;
5528 	}
5529 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
5530 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
5531 
5532 	/* Save what we set and retore it after device reset */
5533 	if (on)
5534 		shadow_vfta->vfta[vid_idx] |= vid_bit;
5535 	else
5536 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
5537 
5538 	return 0;
5539 }
5540 
5541 static void
5542 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
5543 {
5544 	struct ixgbe_hw *hw =
5545 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5546 	uint32_t ctrl;
5547 
5548 	PMD_INIT_FUNC_TRACE();
5549 
5550 	if (queue >= hw->mac.max_rx_queues)
5551 		return;
5552 
5553 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
5554 	if (on)
5555 		ctrl |= IXGBE_RXDCTL_VME;
5556 	else
5557 		ctrl &= ~IXGBE_RXDCTL_VME;
5558 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
5559 
5560 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
5561 }
5562 
5563 static int
5564 ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask)
5565 {
5566 	struct ixgbe_rx_queue *rxq;
5567 	uint16_t i;
5568 	int on = 0;
5569 
5570 	/* VF function only support hw strip feature, others are not support */
5571 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
5572 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
5573 			rxq = dev->data->rx_queues[i];
5574 			on = !!(rxq->offloads &	RTE_ETH_RX_OFFLOAD_VLAN_STRIP);
5575 			ixgbevf_vlan_strip_queue_set(dev, i, on);
5576 		}
5577 	}
5578 
5579 	return 0;
5580 }
5581 
5582 static int
5583 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
5584 {
5585 	ixgbe_config_vlan_strip_on_all_queues(dev, mask);
5586 
5587 	ixgbevf_vlan_offload_config(dev, mask);
5588 
5589 	return 0;
5590 }
5591 
5592 int
5593 ixgbe_vt_check(struct ixgbe_hw *hw)
5594 {
5595 	uint32_t reg_val;
5596 
5597 	/* if Virtualization Technology is enabled */
5598 	reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
5599 	if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
5600 		PMD_INIT_LOG(ERR, "VT must be enabled for this setting");
5601 		return -1;
5602 	}
5603 
5604 	return 0;
5605 }
5606 
5607 static uint32_t
5608 ixgbe_uta_vector(struct ixgbe_hw *hw, struct rte_ether_addr *uc_addr)
5609 {
5610 	uint32_t vector = 0;
5611 
5612 	switch (hw->mac.mc_filter_type) {
5613 	case 0:   /* use bits [47:36] of the address */
5614 		vector = ((uc_addr->addr_bytes[4] >> 4) |
5615 			(((uint16_t)uc_addr->addr_bytes[5]) << 4));
5616 		break;
5617 	case 1:   /* use bits [46:35] of the address */
5618 		vector = ((uc_addr->addr_bytes[4] >> 3) |
5619 			(((uint16_t)uc_addr->addr_bytes[5]) << 5));
5620 		break;
5621 	case 2:   /* use bits [45:34] of the address */
5622 		vector = ((uc_addr->addr_bytes[4] >> 2) |
5623 			(((uint16_t)uc_addr->addr_bytes[5]) << 6));
5624 		break;
5625 	case 3:   /* use bits [43:32] of the address */
5626 		vector = ((uc_addr->addr_bytes[4]) |
5627 			(((uint16_t)uc_addr->addr_bytes[5]) << 8));
5628 		break;
5629 	default:  /* Invalid mc_filter_type */
5630 		break;
5631 	}
5632 
5633 	/* vector can only be 12-bits or boundary will be exceeded */
5634 	vector &= 0xFFF;
5635 	return vector;
5636 }
5637 
5638 static int
5639 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev,
5640 			struct rte_ether_addr *mac_addr, uint8_t on)
5641 {
5642 	uint32_t vector;
5643 	uint32_t uta_idx;
5644 	uint32_t reg_val;
5645 	uint32_t uta_shift;
5646 	uint32_t rc;
5647 	const uint32_t ixgbe_uta_idx_mask = 0x7F;
5648 	const uint32_t ixgbe_uta_bit_shift = 5;
5649 	const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
5650 	const uint32_t bit1 = 0x1;
5651 
5652 	struct ixgbe_hw *hw =
5653 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5654 	struct ixgbe_uta_info *uta_info =
5655 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5656 
5657 	/* The UTA table only exists on 82599 hardware and newer */
5658 	if (hw->mac.type < ixgbe_mac_82599EB)
5659 		return -ENOTSUP;
5660 
5661 	vector = ixgbe_uta_vector(hw, mac_addr);
5662 	uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
5663 	uta_shift = vector & ixgbe_uta_bit_mask;
5664 
5665 	rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
5666 	if (rc == on)
5667 		return 0;
5668 
5669 	reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
5670 	if (on) {
5671 		uta_info->uta_in_use++;
5672 		reg_val |= (bit1 << uta_shift);
5673 		uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
5674 	} else {
5675 		uta_info->uta_in_use--;
5676 		reg_val &= ~(bit1 << uta_shift);
5677 		uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
5678 	}
5679 
5680 	IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
5681 
5682 	if (uta_info->uta_in_use > 0)
5683 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
5684 				IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
5685 	else
5686 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
5687 
5688 	return 0;
5689 }
5690 
5691 static int
5692 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
5693 {
5694 	int i;
5695 	struct ixgbe_hw *hw =
5696 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5697 	struct ixgbe_uta_info *uta_info =
5698 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5699 
5700 	/* The UTA table only exists on 82599 hardware and newer */
5701 	if (hw->mac.type < ixgbe_mac_82599EB)
5702 		return -ENOTSUP;
5703 
5704 	if (on) {
5705 		for (i = 0; i < RTE_ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5706 			uta_info->uta_shadow[i] = ~0;
5707 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
5708 		}
5709 	} else {
5710 		for (i = 0; i < RTE_ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5711 			uta_info->uta_shadow[i] = 0;
5712 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
5713 		}
5714 	}
5715 	return 0;
5716 
5717 }
5718 
5719 uint32_t
5720 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
5721 {
5722 	uint32_t new_val = orig_val;
5723 
5724 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_UNTAG)
5725 		new_val |= IXGBE_VMOLR_AUPE;
5726 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_HASH_MC)
5727 		new_val |= IXGBE_VMOLR_ROMPE;
5728 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_HASH_UC)
5729 		new_val |= IXGBE_VMOLR_ROPE;
5730 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_BROADCAST)
5731 		new_val |= IXGBE_VMOLR_BAM;
5732 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_MULTICAST)
5733 		new_val |= IXGBE_VMOLR_MPE;
5734 
5735 	return new_val;
5736 }
5737 
5738 static int
5739 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5740 {
5741 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5742 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5743 	struct ixgbe_interrupt *intr =
5744 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5745 	struct ixgbe_hw *hw =
5746 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5747 	uint32_t vec = IXGBE_MISC_VEC_ID;
5748 
5749 	if (rte_intr_allow_others(intr_handle))
5750 		vec = IXGBE_RX_VEC_START;
5751 	intr->mask |= (1 << vec);
5752 	RTE_SET_USED(queue_id);
5753 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5754 
5755 	rte_intr_ack(intr_handle);
5756 
5757 	return 0;
5758 }
5759 
5760 static int
5761 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5762 {
5763 	struct ixgbe_interrupt *intr =
5764 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5765 	struct ixgbe_hw *hw =
5766 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5767 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5768 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5769 	uint32_t vec = IXGBE_MISC_VEC_ID;
5770 
5771 	if (rte_intr_allow_others(intr_handle))
5772 		vec = IXGBE_RX_VEC_START;
5773 	intr->mask &= ~(1 << vec);
5774 	RTE_SET_USED(queue_id);
5775 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5776 
5777 	return 0;
5778 }
5779 
5780 static int
5781 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5782 {
5783 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5784 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5785 	uint32_t mask;
5786 	struct ixgbe_hw *hw =
5787 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5788 	struct ixgbe_interrupt *intr =
5789 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5790 
5791 	if (queue_id < 16) {
5792 		ixgbe_disable_intr(hw);
5793 		intr->mask |= (1 << queue_id);
5794 		ixgbe_enable_intr(dev);
5795 	} else if (queue_id < 32) {
5796 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5797 		mask &= (1 << queue_id);
5798 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5799 	} else if (queue_id < 64) {
5800 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5801 		mask &= (1 << (queue_id - 32));
5802 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5803 	}
5804 	rte_intr_ack(intr_handle);
5805 
5806 	return 0;
5807 }
5808 
5809 static int
5810 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5811 {
5812 	uint32_t mask;
5813 	struct ixgbe_hw *hw =
5814 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5815 	struct ixgbe_interrupt *intr =
5816 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5817 
5818 	if (queue_id < 16) {
5819 		ixgbe_disable_intr(hw);
5820 		intr->mask &= ~(1 << queue_id);
5821 		ixgbe_enable_intr(dev);
5822 	} else if (queue_id < 32) {
5823 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5824 		mask &= ~(1 << queue_id);
5825 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5826 	} else if (queue_id < 64) {
5827 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5828 		mask &= ~(1 << (queue_id - 32));
5829 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5830 	}
5831 
5832 	return 0;
5833 }
5834 
5835 static void
5836 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5837 		     uint8_t queue, uint8_t msix_vector)
5838 {
5839 	uint32_t tmp, idx;
5840 
5841 	if (direction == -1) {
5842 		/* other causes */
5843 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5844 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
5845 		tmp &= ~0xFF;
5846 		tmp |= msix_vector;
5847 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
5848 	} else {
5849 		/* rx or tx cause */
5850 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5851 		idx = ((16 * (queue & 1)) + (8 * direction));
5852 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
5853 		tmp &= ~(0xFF << idx);
5854 		tmp |= (msix_vector << idx);
5855 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
5856 	}
5857 }
5858 
5859 /**
5860  * set the IVAR registers, mapping interrupt causes to vectors
5861  * @param hw
5862  *  pointer to ixgbe_hw struct
5863  * @direction
5864  *  0 for Rx, 1 for Tx, -1 for other causes
5865  * @queue
5866  *  queue to map the corresponding interrupt to
5867  * @msix_vector
5868  *  the vector to map to the corresponding queue
5869  */
5870 static void
5871 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5872 		   uint8_t queue, uint8_t msix_vector)
5873 {
5874 	uint32_t tmp, idx;
5875 
5876 	msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5877 	if (hw->mac.type == ixgbe_mac_82598EB) {
5878 		if (direction == -1)
5879 			direction = 0;
5880 		idx = (((direction * 64) + queue) >> 2) & 0x1F;
5881 		tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
5882 		tmp &= ~(0xFF << (8 * (queue & 0x3)));
5883 		tmp |= (msix_vector << (8 * (queue & 0x3)));
5884 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
5885 	} else if ((hw->mac.type == ixgbe_mac_82599EB) ||
5886 			(hw->mac.type == ixgbe_mac_X540) ||
5887 			(hw->mac.type == ixgbe_mac_X550) ||
5888 			(hw->mac.type == ixgbe_mac_X550EM_x)) {
5889 		if (direction == -1) {
5890 			/* other causes */
5891 			idx = ((queue & 1) * 8);
5892 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
5893 			tmp &= ~(0xFF << idx);
5894 			tmp |= (msix_vector << idx);
5895 			IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
5896 		} else {
5897 			/* rx or tx causes */
5898 			idx = ((16 * (queue & 1)) + (8 * direction));
5899 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
5900 			tmp &= ~(0xFF << idx);
5901 			tmp |= (msix_vector << idx);
5902 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
5903 		}
5904 	}
5905 }
5906 
5907 static void
5908 ixgbevf_configure_msix(struct rte_eth_dev *dev)
5909 {
5910 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5911 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5912 	struct ixgbe_hw *hw =
5913 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5914 	uint32_t q_idx;
5915 	uint32_t vector_idx = IXGBE_MISC_VEC_ID;
5916 	uint32_t base = IXGBE_MISC_VEC_ID;
5917 
5918 	/* Configure VF other cause ivar */
5919 	ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
5920 
5921 	/* won't configure msix register if no mapping is done
5922 	 * between intr vector and event fd.
5923 	 */
5924 	if (!rte_intr_dp_is_en(intr_handle))
5925 		return;
5926 
5927 	if (rte_intr_allow_others(intr_handle)) {
5928 		base = IXGBE_RX_VEC_START;
5929 		vector_idx = IXGBE_RX_VEC_START;
5930 	}
5931 
5932 	/* Configure all RX queues of VF */
5933 	for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
5934 		/* Force all queue use vector 0,
5935 		 * as IXGBE_VF_MAXMSIVECTOR = 1
5936 		 */
5937 		ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
5938 		rte_intr_vec_list_index_set(intr_handle, q_idx,
5939 						   vector_idx);
5940 		if (vector_idx < base + rte_intr_nb_efd_get(intr_handle)
5941 		    - 1)
5942 			vector_idx++;
5943 	}
5944 
5945 	/* As RX queue setting above show, all queues use the vector 0.
5946 	 * Set only the ITR value of IXGBE_MISC_VEC_ID.
5947 	 */
5948 	IXGBE_WRITE_REG(hw, IXGBE_VTEITR(IXGBE_MISC_VEC_ID),
5949 			IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
5950 			| IXGBE_EITR_CNT_WDIS);
5951 }
5952 
5953 /**
5954  * Sets up the hardware to properly generate MSI-X interrupts
5955  * @hw
5956  *  board private structure
5957  */
5958 static void
5959 ixgbe_configure_msix(struct rte_eth_dev *dev)
5960 {
5961 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5962 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5963 	struct ixgbe_hw *hw =
5964 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5965 	uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
5966 	uint32_t vec = IXGBE_MISC_VEC_ID;
5967 	uint32_t mask;
5968 	uint32_t gpie;
5969 
5970 	/* won't configure msix register if no mapping is done
5971 	 * between intr vector and event fd
5972 	 * but if misx has been enabled already, need to configure
5973 	 * auto clean, auto mask and throttling.
5974 	 */
5975 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5976 	if (!rte_intr_dp_is_en(intr_handle) &&
5977 	    !(gpie & (IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT)))
5978 		return;
5979 
5980 	if (rte_intr_allow_others(intr_handle))
5981 		vec = base = IXGBE_RX_VEC_START;
5982 
5983 	/* setup GPIE for MSI-x mode */
5984 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5985 	gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5986 		IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
5987 	/* auto clearing and auto setting corresponding bits in EIMS
5988 	 * when MSI-X interrupt is triggered
5989 	 */
5990 	if (hw->mac.type == ixgbe_mac_82598EB) {
5991 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5992 	} else {
5993 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5994 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5995 	}
5996 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5997 
5998 	/* Populate the IVAR table and set the ITR values to the
5999 	 * corresponding register.
6000 	 */
6001 	if (rte_intr_dp_is_en(intr_handle)) {
6002 		for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
6003 			queue_id++) {
6004 			/* by default, 1:1 mapping */
6005 			ixgbe_set_ivar_map(hw, 0, queue_id, vec);
6006 			rte_intr_vec_list_index_set(intr_handle,
6007 							   queue_id, vec);
6008 			if (vec < base + rte_intr_nb_efd_get(intr_handle)
6009 			    - 1)
6010 				vec++;
6011 		}
6012 
6013 		switch (hw->mac.type) {
6014 		case ixgbe_mac_82598EB:
6015 			ixgbe_set_ivar_map(hw, -1,
6016 					   IXGBE_IVAR_OTHER_CAUSES_INDEX,
6017 					   IXGBE_MISC_VEC_ID);
6018 			break;
6019 		case ixgbe_mac_82599EB:
6020 		case ixgbe_mac_X540:
6021 		case ixgbe_mac_X550:
6022 		case ixgbe_mac_X550EM_x:
6023 			ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
6024 			break;
6025 		default:
6026 			break;
6027 		}
6028 	}
6029 	IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
6030 			IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
6031 			| IXGBE_EITR_CNT_WDIS);
6032 
6033 	/* set up to autoclear timer, and the vectors */
6034 	mask = IXGBE_EIMS_ENABLE_MASK;
6035 	mask &= ~(IXGBE_EIMS_OTHER |
6036 		  IXGBE_EIMS_MAILBOX |
6037 		  IXGBE_EIMS_LSC);
6038 
6039 	IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
6040 }
6041 
6042 int
6043 ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
6044 			   uint16_t queue_idx, uint16_t tx_rate)
6045 {
6046 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6047 	uint32_t rf_dec, rf_int;
6048 	uint32_t bcnrc_val;
6049 	uint16_t link_speed = dev->data->dev_link.link_speed;
6050 
6051 	if (queue_idx >= hw->mac.max_tx_queues)
6052 		return -EINVAL;
6053 
6054 	if (tx_rate != 0) {
6055 		/* Calculate the rate factor values to set */
6056 		rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
6057 		rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
6058 		rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
6059 
6060 		bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
6061 		bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
6062 				IXGBE_RTTBCNRC_RF_INT_MASK_M);
6063 		bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
6064 	} else {
6065 		bcnrc_val = 0;
6066 	}
6067 
6068 	/*
6069 	 * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
6070 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
6071 	 * set as 0x4.
6072 	 */
6073 	if (dev->data->mtu + IXGBE_ETH_OVERHEAD >= IXGBE_MAX_JUMBO_FRAME_SIZE)
6074 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM, IXGBE_MMW_SIZE_JUMBO_FRAME);
6075 	else
6076 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM, IXGBE_MMW_SIZE_DEFAULT);
6077 
6078 	/* Set RTTBCNRC of queue X */
6079 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
6080 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
6081 	IXGBE_WRITE_FLUSH(hw);
6082 
6083 	return 0;
6084 }
6085 
6086 static int
6087 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
6088 		     __rte_unused uint32_t index,
6089 		     __rte_unused uint32_t pool)
6090 {
6091 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6092 	int diag;
6093 
6094 	/*
6095 	 * On a 82599 VF, adding again the same MAC addr is not an idempotent
6096 	 * operation. Trap this case to avoid exhausting the [very limited]
6097 	 * set of PF resources used to store VF MAC addresses.
6098 	 */
6099 	if (memcmp(hw->mac.perm_addr, mac_addr,
6100 			sizeof(struct rte_ether_addr)) == 0)
6101 		return -1;
6102 	diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6103 	if (diag != 0)
6104 		PMD_DRV_LOG(ERR, "Unable to add MAC address "
6105 			    RTE_ETHER_ADDR_PRT_FMT " - diag=%d",
6106 			    RTE_ETHER_ADDR_BYTES(mac_addr), diag);
6107 	return diag;
6108 }
6109 
6110 static void
6111 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
6112 {
6113 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6114 	struct rte_ether_addr *perm_addr =
6115 		(struct rte_ether_addr *)hw->mac.perm_addr;
6116 	struct rte_ether_addr *mac_addr;
6117 	uint32_t i;
6118 	int diag;
6119 
6120 	/*
6121 	 * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
6122 	 * not support the deletion of a given MAC address.
6123 	 * Instead, it imposes to delete all MAC addresses, then to add again
6124 	 * all MAC addresses with the exception of the one to be deleted.
6125 	 */
6126 	(void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
6127 
6128 	/*
6129 	 * Add again all MAC addresses, with the exception of the deleted one
6130 	 * and of the permanent MAC address.
6131 	 */
6132 	for (i = 0, mac_addr = dev->data->mac_addrs;
6133 	     i < hw->mac.num_rar_entries; i++, mac_addr++) {
6134 		/* Skip the deleted MAC address */
6135 		if (i == index)
6136 			continue;
6137 		/* Skip NULL MAC addresses */
6138 		if (rte_is_zero_ether_addr(mac_addr))
6139 			continue;
6140 		/* Skip the permanent MAC address */
6141 		if (memcmp(perm_addr, mac_addr,
6142 				sizeof(struct rte_ether_addr)) == 0)
6143 			continue;
6144 		diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6145 		if (diag != 0)
6146 			PMD_DRV_LOG(ERR,
6147 				    "Adding again MAC address "
6148 				     RTE_ETHER_ADDR_PRT_FMT " failed "
6149 				    "diag=%d", RTE_ETHER_ADDR_BYTES(mac_addr),
6150 				    diag);
6151 	}
6152 }
6153 
6154 static int
6155 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
6156 			struct rte_ether_addr *addr)
6157 {
6158 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6159 
6160 	hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
6161 
6162 	return 0;
6163 }
6164 
6165 int
6166 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
6167 			struct rte_eth_syn_filter *filter,
6168 			bool add)
6169 {
6170 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6171 	struct ixgbe_filter_info *filter_info =
6172 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6173 	uint32_t syn_info;
6174 	uint32_t synqf;
6175 
6176 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6177 		return -EINVAL;
6178 
6179 	syn_info = filter_info->syn_info;
6180 
6181 	if (add) {
6182 		if (syn_info & IXGBE_SYN_FILTER_ENABLE)
6183 			return -EINVAL;
6184 		synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
6185 			IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
6186 
6187 		if (filter->hig_pri)
6188 			synqf |= IXGBE_SYN_FILTER_SYNQFP;
6189 		else
6190 			synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
6191 	} else {
6192 		synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6193 		if (!(syn_info & IXGBE_SYN_FILTER_ENABLE))
6194 			return -ENOENT;
6195 		synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
6196 	}
6197 
6198 	filter_info->syn_info = synqf;
6199 	IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
6200 	IXGBE_WRITE_FLUSH(hw);
6201 	return 0;
6202 }
6203 
6204 
6205 static inline enum ixgbe_5tuple_protocol
6206 convert_protocol_type(uint8_t protocol_value)
6207 {
6208 	if (protocol_value == IPPROTO_TCP)
6209 		return IXGBE_FILTER_PROTOCOL_TCP;
6210 	else if (protocol_value == IPPROTO_UDP)
6211 		return IXGBE_FILTER_PROTOCOL_UDP;
6212 	else if (protocol_value == IPPROTO_SCTP)
6213 		return IXGBE_FILTER_PROTOCOL_SCTP;
6214 	else
6215 		return IXGBE_FILTER_PROTOCOL_NONE;
6216 }
6217 
6218 /* inject a 5-tuple filter to HW */
6219 static inline void
6220 ixgbe_inject_5tuple_filter(struct rte_eth_dev *dev,
6221 			   struct ixgbe_5tuple_filter *filter)
6222 {
6223 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6224 	int i;
6225 	uint32_t ftqf, sdpqf;
6226 	uint32_t l34timir = 0;
6227 	uint8_t mask = 0xff;
6228 
6229 	i = filter->index;
6230 
6231 	sdpqf = (uint32_t)(filter->filter_info.dst_port <<
6232 				IXGBE_SDPQF_DSTPORT_SHIFT);
6233 	sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
6234 
6235 	ftqf = (uint32_t)(filter->filter_info.proto &
6236 		IXGBE_FTQF_PROTOCOL_MASK);
6237 	ftqf |= (uint32_t)((filter->filter_info.priority &
6238 		IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
6239 	if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
6240 		mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
6241 	if (filter->filter_info.dst_ip_mask == 0)
6242 		mask &= IXGBE_FTQF_DEST_ADDR_MASK;
6243 	if (filter->filter_info.src_port_mask == 0)
6244 		mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
6245 	if (filter->filter_info.dst_port_mask == 0)
6246 		mask &= IXGBE_FTQF_DEST_PORT_MASK;
6247 	if (filter->filter_info.proto_mask == 0)
6248 		mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
6249 	ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
6250 	ftqf |= IXGBE_FTQF_POOL_MASK_EN;
6251 	ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
6252 
6253 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
6254 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
6255 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
6256 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
6257 
6258 	l34timir |= IXGBE_L34T_IMIR_RESERVE;
6259 	l34timir |= (uint32_t)(filter->queue <<
6260 				IXGBE_L34T_IMIR_QUEUE_SHIFT);
6261 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
6262 }
6263 
6264 /*
6265  * add a 5tuple filter
6266  *
6267  * @param
6268  * dev: Pointer to struct rte_eth_dev.
6269  * index: the index the filter allocates.
6270  * filter: pointer to the filter that will be added.
6271  * rx_queue: the queue id the filter assigned to.
6272  *
6273  * @return
6274  *    - On success, zero.
6275  *    - On failure, a negative value.
6276  */
6277 static int
6278 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
6279 			struct ixgbe_5tuple_filter *filter)
6280 {
6281 	struct ixgbe_filter_info *filter_info =
6282 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6283 	int i, idx, shift;
6284 
6285 	/*
6286 	 * look for an unused 5tuple filter index,
6287 	 * and insert the filter to list.
6288 	 */
6289 	for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
6290 		idx = i / (sizeof(uint32_t) * NBBY);
6291 		shift = i % (sizeof(uint32_t) * NBBY);
6292 		if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
6293 			filter_info->fivetuple_mask[idx] |= 1 << shift;
6294 			filter->index = i;
6295 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
6296 					  filter,
6297 					  entries);
6298 			break;
6299 		}
6300 	}
6301 	if (i >= IXGBE_MAX_FTQF_FILTERS) {
6302 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
6303 		return -ENOSYS;
6304 	}
6305 
6306 	ixgbe_inject_5tuple_filter(dev, filter);
6307 
6308 	return 0;
6309 }
6310 
6311 /*
6312  * remove a 5tuple filter
6313  *
6314  * @param
6315  * dev: Pointer to struct rte_eth_dev.
6316  * filter: the pointer of the filter will be removed.
6317  */
6318 static void
6319 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
6320 			struct ixgbe_5tuple_filter *filter)
6321 {
6322 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6323 	struct ixgbe_filter_info *filter_info =
6324 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6325 	uint16_t index = filter->index;
6326 
6327 	filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
6328 				~(1 << (index % (sizeof(uint32_t) * NBBY)));
6329 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
6330 	rte_free(filter);
6331 
6332 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
6333 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
6334 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
6335 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
6336 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
6337 }
6338 
6339 static int
6340 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
6341 {
6342 	struct ixgbe_hw *hw;
6343 	uint32_t max_frame = mtu + IXGBE_ETH_OVERHEAD;
6344 	struct rte_eth_dev_data *dev_data = dev->data;
6345 
6346 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6347 
6348 	if (mtu < RTE_ETHER_MIN_MTU || max_frame > RTE_ETHER_MAX_JUMBO_FRAME_LEN)
6349 		return -EINVAL;
6350 
6351 	/* If device is started, refuse mtu that requires the support of
6352 	 * scattered packets when this feature has not been enabled before.
6353 	 */
6354 	if (dev_data->dev_started && !dev_data->scattered_rx &&
6355 	    (max_frame + 2 * RTE_VLAN_HLEN >
6356 			dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
6357 		PMD_INIT_LOG(ERR, "Stop port first.");
6358 		return -EINVAL;
6359 	}
6360 
6361 	/*
6362 	 * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
6363 	 * request of the version 2.0 of the mailbox API.
6364 	 * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
6365 	 * of the mailbox API.
6366 	 * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
6367 	 * prior to 3.11.33 which contains the following change:
6368 	 * "ixgbe: Enable jumbo frames support w/ SR-IOV"
6369 	 */
6370 	if (ixgbevf_rlpml_set_vf(hw, max_frame))
6371 		return -EINVAL;
6372 
6373 	return 0;
6374 }
6375 
6376 static inline struct ixgbe_5tuple_filter *
6377 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
6378 			struct ixgbe_5tuple_filter_info *key)
6379 {
6380 	struct ixgbe_5tuple_filter *it;
6381 
6382 	TAILQ_FOREACH(it, filter_list, entries) {
6383 		if (memcmp(key, &it->filter_info,
6384 			sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
6385 			return it;
6386 		}
6387 	}
6388 	return NULL;
6389 }
6390 
6391 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
6392 static inline int
6393 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
6394 			struct ixgbe_5tuple_filter_info *filter_info)
6395 {
6396 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
6397 		filter->priority > IXGBE_5TUPLE_MAX_PRI ||
6398 		filter->priority < IXGBE_5TUPLE_MIN_PRI)
6399 		return -EINVAL;
6400 
6401 	switch (filter->dst_ip_mask) {
6402 	case UINT32_MAX:
6403 		filter_info->dst_ip_mask = 0;
6404 		filter_info->dst_ip = filter->dst_ip;
6405 		break;
6406 	case 0:
6407 		filter_info->dst_ip_mask = 1;
6408 		break;
6409 	default:
6410 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
6411 		return -EINVAL;
6412 	}
6413 
6414 	switch (filter->src_ip_mask) {
6415 	case UINT32_MAX:
6416 		filter_info->src_ip_mask = 0;
6417 		filter_info->src_ip = filter->src_ip;
6418 		break;
6419 	case 0:
6420 		filter_info->src_ip_mask = 1;
6421 		break;
6422 	default:
6423 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
6424 		return -EINVAL;
6425 	}
6426 
6427 	switch (filter->dst_port_mask) {
6428 	case UINT16_MAX:
6429 		filter_info->dst_port_mask = 0;
6430 		filter_info->dst_port = filter->dst_port;
6431 		break;
6432 	case 0:
6433 		filter_info->dst_port_mask = 1;
6434 		break;
6435 	default:
6436 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
6437 		return -EINVAL;
6438 	}
6439 
6440 	switch (filter->src_port_mask) {
6441 	case UINT16_MAX:
6442 		filter_info->src_port_mask = 0;
6443 		filter_info->src_port = filter->src_port;
6444 		break;
6445 	case 0:
6446 		filter_info->src_port_mask = 1;
6447 		break;
6448 	default:
6449 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
6450 		return -EINVAL;
6451 	}
6452 
6453 	switch (filter->proto_mask) {
6454 	case UINT8_MAX:
6455 		filter_info->proto_mask = 0;
6456 		filter_info->proto =
6457 			convert_protocol_type(filter->proto);
6458 		break;
6459 	case 0:
6460 		filter_info->proto_mask = 1;
6461 		break;
6462 	default:
6463 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
6464 		return -EINVAL;
6465 	}
6466 
6467 	filter_info->priority = (uint8_t)filter->priority;
6468 	return 0;
6469 }
6470 
6471 /*
6472  * add or delete a ntuple filter
6473  *
6474  * @param
6475  * dev: Pointer to struct rte_eth_dev.
6476  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6477  * add: if true, add filter, if false, remove filter
6478  *
6479  * @return
6480  *    - On success, zero.
6481  *    - On failure, a negative value.
6482  */
6483 int
6484 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
6485 			struct rte_eth_ntuple_filter *ntuple_filter,
6486 			bool add)
6487 {
6488 	struct ixgbe_filter_info *filter_info =
6489 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6490 	struct ixgbe_5tuple_filter_info filter_5tuple;
6491 	struct ixgbe_5tuple_filter *filter;
6492 	int ret;
6493 
6494 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6495 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6496 		return -EINVAL;
6497 	}
6498 
6499 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6500 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6501 	if (ret < 0)
6502 		return ret;
6503 
6504 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6505 					 &filter_5tuple);
6506 	if (filter != NULL && add) {
6507 		PMD_DRV_LOG(ERR, "filter exists.");
6508 		return -EEXIST;
6509 	}
6510 	if (filter == NULL && !add) {
6511 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
6512 		return -ENOENT;
6513 	}
6514 
6515 	if (add) {
6516 		filter = rte_zmalloc("ixgbe_5tuple_filter",
6517 				sizeof(struct ixgbe_5tuple_filter), 0);
6518 		if (filter == NULL)
6519 			return -ENOMEM;
6520 		rte_memcpy(&filter->filter_info,
6521 				 &filter_5tuple,
6522 				 sizeof(struct ixgbe_5tuple_filter_info));
6523 		filter->queue = ntuple_filter->queue;
6524 		ret = ixgbe_add_5tuple_filter(dev, filter);
6525 		if (ret < 0) {
6526 			rte_free(filter);
6527 			return ret;
6528 		}
6529 	} else
6530 		ixgbe_remove_5tuple_filter(dev, filter);
6531 
6532 	return 0;
6533 }
6534 
6535 int
6536 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
6537 			struct rte_eth_ethertype_filter *filter,
6538 			bool add)
6539 {
6540 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6541 	struct ixgbe_filter_info *filter_info =
6542 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6543 	uint32_t etqf = 0;
6544 	uint32_t etqs = 0;
6545 	int ret;
6546 	struct ixgbe_ethertype_filter ethertype_filter;
6547 
6548 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6549 		return -EINVAL;
6550 
6551 	if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
6552 		filter->ether_type == RTE_ETHER_TYPE_IPV6) {
6553 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
6554 			" ethertype filter.", filter->ether_type);
6555 		return -EINVAL;
6556 	}
6557 
6558 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
6559 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
6560 		return -EINVAL;
6561 	}
6562 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
6563 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
6564 		return -EINVAL;
6565 	}
6566 
6567 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6568 	if (ret >= 0 && add) {
6569 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
6570 			    filter->ether_type);
6571 		return -EEXIST;
6572 	}
6573 	if (ret < 0 && !add) {
6574 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6575 			    filter->ether_type);
6576 		return -ENOENT;
6577 	}
6578 
6579 	if (add) {
6580 		etqf = IXGBE_ETQF_FILTER_EN;
6581 		etqf |= (uint32_t)filter->ether_type;
6582 		etqs |= (uint32_t)((filter->queue <<
6583 				    IXGBE_ETQS_RX_QUEUE_SHIFT) &
6584 				    IXGBE_ETQS_RX_QUEUE);
6585 		etqs |= IXGBE_ETQS_QUEUE_EN;
6586 
6587 		ethertype_filter.ethertype = filter->ether_type;
6588 		ethertype_filter.etqf = etqf;
6589 		ethertype_filter.etqs = etqs;
6590 		ethertype_filter.conf = FALSE;
6591 		ret = ixgbe_ethertype_filter_insert(filter_info,
6592 						    &ethertype_filter);
6593 		if (ret < 0) {
6594 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
6595 			return -ENOSPC;
6596 		}
6597 	} else {
6598 		ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
6599 		if (ret < 0)
6600 			return -ENOSYS;
6601 	}
6602 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
6603 	IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
6604 	IXGBE_WRITE_FLUSH(hw);
6605 
6606 	return 0;
6607 }
6608 
6609 static int
6610 ixgbe_dev_flow_ops_get(__rte_unused struct rte_eth_dev *dev,
6611 		       const struct rte_flow_ops **ops)
6612 {
6613 	*ops = &ixgbe_flow_ops;
6614 	return 0;
6615 }
6616 
6617 static u8 *
6618 ixgbe_dev_addr_list_itr(__rte_unused struct ixgbe_hw *hw,
6619 			u8 **mc_addr_ptr, u32 *vmdq)
6620 {
6621 	u8 *mc_addr;
6622 
6623 	*vmdq = 0;
6624 	mc_addr = *mc_addr_ptr;
6625 	*mc_addr_ptr = (mc_addr + sizeof(struct rte_ether_addr));
6626 	return mc_addr;
6627 }
6628 
6629 static int
6630 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
6631 			  struct rte_ether_addr *mc_addr_set,
6632 			  uint32_t nb_mc_addr)
6633 {
6634 	struct ixgbe_hw *hw;
6635 	u8 *mc_addr_list;
6636 
6637 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6638 	mc_addr_list = (u8 *)mc_addr_set;
6639 	return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
6640 					 ixgbe_dev_addr_list_itr, TRUE);
6641 }
6642 
6643 static uint64_t
6644 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
6645 {
6646 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6647 	uint64_t systime_cycles;
6648 
6649 	switch (hw->mac.type) {
6650 	case ixgbe_mac_X550:
6651 	case ixgbe_mac_X550EM_x:
6652 	case ixgbe_mac_X550EM_a:
6653 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
6654 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6655 		systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6656 				* NSEC_PER_SEC;
6657 		break;
6658 	default:
6659 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6660 		systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6661 				<< 32;
6662 	}
6663 
6664 	return systime_cycles;
6665 }
6666 
6667 static uint64_t
6668 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6669 {
6670 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6671 	uint64_t rx_tstamp_cycles;
6672 
6673 	switch (hw->mac.type) {
6674 	case ixgbe_mac_X550:
6675 	case ixgbe_mac_X550EM_x:
6676 	case ixgbe_mac_X550EM_a:
6677 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6678 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6679 		rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6680 				* NSEC_PER_SEC;
6681 		break;
6682 	default:
6683 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6684 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6685 		rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6686 				<< 32;
6687 	}
6688 
6689 	return rx_tstamp_cycles;
6690 }
6691 
6692 static uint64_t
6693 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6694 {
6695 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6696 	uint64_t tx_tstamp_cycles;
6697 
6698 	switch (hw->mac.type) {
6699 	case ixgbe_mac_X550:
6700 	case ixgbe_mac_X550EM_x:
6701 	case ixgbe_mac_X550EM_a:
6702 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6703 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6704 		tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6705 				* NSEC_PER_SEC;
6706 		break;
6707 	default:
6708 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6709 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6710 		tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6711 				<< 32;
6712 	}
6713 
6714 	return tx_tstamp_cycles;
6715 }
6716 
6717 static void
6718 ixgbe_start_timecounters(struct rte_eth_dev *dev)
6719 {
6720 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6721 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6722 	struct rte_eth_link link;
6723 	uint32_t incval = 0;
6724 	uint32_t shift = 0;
6725 
6726 	/* Get current link speed. */
6727 	ixgbe_dev_link_update(dev, 1);
6728 	rte_eth_linkstatus_get(dev, &link);
6729 
6730 	switch (link.link_speed) {
6731 	case RTE_ETH_SPEED_NUM_100M:
6732 		incval = IXGBE_INCVAL_100;
6733 		shift = IXGBE_INCVAL_SHIFT_100;
6734 		break;
6735 	case RTE_ETH_SPEED_NUM_1G:
6736 		incval = IXGBE_INCVAL_1GB;
6737 		shift = IXGBE_INCVAL_SHIFT_1GB;
6738 		break;
6739 	case RTE_ETH_SPEED_NUM_10G:
6740 	default:
6741 		incval = IXGBE_INCVAL_10GB;
6742 		shift = IXGBE_INCVAL_SHIFT_10GB;
6743 		break;
6744 	}
6745 
6746 	switch (hw->mac.type) {
6747 	case ixgbe_mac_X550:
6748 	case ixgbe_mac_X550EM_x:
6749 	case ixgbe_mac_X550EM_a:
6750 		/* Independent of link speed. */
6751 		incval = 1;
6752 		/* Cycles read will be interpreted as ns. */
6753 		shift = 0;
6754 		/* Fall-through */
6755 	case ixgbe_mac_X540:
6756 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
6757 		break;
6758 	case ixgbe_mac_82599EB:
6759 		incval >>= IXGBE_INCVAL_SHIFT_82599;
6760 		shift -= IXGBE_INCVAL_SHIFT_82599;
6761 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
6762 				(1 << IXGBE_INCPER_SHIFT_82599) | incval);
6763 		break;
6764 	default:
6765 		/* Not supported. */
6766 		return;
6767 	}
6768 
6769 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
6770 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6771 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6772 
6773 	adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6774 	adapter->systime_tc.cc_shift = shift;
6775 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
6776 
6777 	adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6778 	adapter->rx_tstamp_tc.cc_shift = shift;
6779 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6780 
6781 	adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6782 	adapter->tx_tstamp_tc.cc_shift = shift;
6783 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6784 }
6785 
6786 static int
6787 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
6788 {
6789 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6790 
6791 	adapter->systime_tc.nsec += delta;
6792 	adapter->rx_tstamp_tc.nsec += delta;
6793 	adapter->tx_tstamp_tc.nsec += delta;
6794 
6795 	return 0;
6796 }
6797 
6798 static int
6799 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
6800 {
6801 	uint64_t ns;
6802 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6803 
6804 	ns = rte_timespec_to_ns(ts);
6805 	/* Set the timecounters to a new value. */
6806 	adapter->systime_tc.nsec = ns;
6807 	adapter->rx_tstamp_tc.nsec = ns;
6808 	adapter->tx_tstamp_tc.nsec = ns;
6809 
6810 	return 0;
6811 }
6812 
6813 static int
6814 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
6815 {
6816 	uint64_t ns, systime_cycles;
6817 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6818 
6819 	systime_cycles = ixgbe_read_systime_cyclecounter(dev);
6820 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
6821 	*ts = rte_ns_to_timespec(ns);
6822 
6823 	return 0;
6824 }
6825 
6826 static int
6827 ixgbe_timesync_enable(struct rte_eth_dev *dev)
6828 {
6829 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6830 	uint32_t tsync_ctl;
6831 	uint32_t tsauxc;
6832 
6833 	/* Stop the timesync system time. */
6834 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
6835 	/* Reset the timesync system time value. */
6836 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
6837 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
6838 
6839 	/* Enable system time for platforms where it isn't on by default. */
6840 	tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
6841 	tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
6842 	IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
6843 
6844 	ixgbe_start_timecounters(dev);
6845 
6846 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6847 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
6848 			(RTE_ETHER_TYPE_1588 |
6849 			 IXGBE_ETQF_FILTER_EN |
6850 			 IXGBE_ETQF_1588));
6851 
6852 	/* Enable timestamping of received PTP packets. */
6853 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6854 	tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
6855 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6856 
6857 	/* Enable timestamping of transmitted PTP packets. */
6858 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6859 	tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
6860 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6861 
6862 	IXGBE_WRITE_FLUSH(hw);
6863 
6864 	return 0;
6865 }
6866 
6867 static int
6868 ixgbe_timesync_disable(struct rte_eth_dev *dev)
6869 {
6870 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6871 	uint32_t tsync_ctl;
6872 
6873 	/* Disable timestamping of transmitted PTP packets. */
6874 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6875 	tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
6876 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6877 
6878 	/* Disable timestamping of received PTP packets. */
6879 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6880 	tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
6881 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6882 
6883 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6884 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
6885 
6886 	/* Stop incrementing the System Time registers. */
6887 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
6888 
6889 	return 0;
6890 }
6891 
6892 static int
6893 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
6894 				 struct timespec *timestamp,
6895 				 uint32_t flags __rte_unused)
6896 {
6897 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6898 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6899 	uint32_t tsync_rxctl;
6900 	uint64_t rx_tstamp_cycles;
6901 	uint64_t ns;
6902 
6903 	tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6904 	if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
6905 		return -EINVAL;
6906 
6907 	rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
6908 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
6909 	*timestamp = rte_ns_to_timespec(ns);
6910 
6911 	return  0;
6912 }
6913 
6914 static int
6915 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
6916 				 struct timespec *timestamp)
6917 {
6918 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6919 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6920 	uint32_t tsync_txctl;
6921 	uint64_t tx_tstamp_cycles;
6922 	uint64_t ns;
6923 
6924 	tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6925 	if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
6926 		return -EINVAL;
6927 
6928 	tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
6929 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
6930 	*timestamp = rte_ns_to_timespec(ns);
6931 
6932 	return 0;
6933 }
6934 
6935 static int
6936 ixgbe_get_reg_length(struct rte_eth_dev *dev)
6937 {
6938 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6939 	int count = 0;
6940 	int g_ind = 0;
6941 	const struct reg_info *reg_group;
6942 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6943 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6944 
6945 	while ((reg_group = reg_set[g_ind++]))
6946 		count += ixgbe_regs_group_count(reg_group);
6947 
6948 	return count;
6949 }
6950 
6951 static int
6952 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
6953 {
6954 	int count = 0;
6955 	int g_ind = 0;
6956 	const struct reg_info *reg_group;
6957 
6958 	while ((reg_group = ixgbevf_regs[g_ind++]))
6959 		count += ixgbe_regs_group_count(reg_group);
6960 
6961 	return count;
6962 }
6963 
6964 static int
6965 ixgbe_get_regs(struct rte_eth_dev *dev,
6966 	      struct rte_dev_reg_info *regs)
6967 {
6968 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6969 	uint32_t *data = regs->data;
6970 	int g_ind = 0;
6971 	int count = 0;
6972 	const struct reg_info *reg_group;
6973 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6974 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6975 
6976 	if (data == NULL) {
6977 		regs->length = ixgbe_get_reg_length(dev);
6978 		regs->width = sizeof(uint32_t);
6979 		return 0;
6980 	}
6981 
6982 	/* Support only full register dump */
6983 	if ((regs->length == 0) ||
6984 	    (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
6985 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
6986 			hw->device_id;
6987 		while ((reg_group = reg_set[g_ind++]))
6988 			count += ixgbe_read_regs_group(dev, &data[count],
6989 				reg_group);
6990 		return 0;
6991 	}
6992 
6993 	return -ENOTSUP;
6994 }
6995 
6996 static int
6997 ixgbevf_get_regs(struct rte_eth_dev *dev,
6998 		struct rte_dev_reg_info *regs)
6999 {
7000 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7001 	uint32_t *data = regs->data;
7002 	int g_ind = 0;
7003 	int count = 0;
7004 	const struct reg_info *reg_group;
7005 
7006 	if (data == NULL) {
7007 		regs->length = ixgbevf_get_reg_length(dev);
7008 		regs->width = sizeof(uint32_t);
7009 		return 0;
7010 	}
7011 
7012 	/* Support only full register dump */
7013 	if ((regs->length == 0) ||
7014 	    (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
7015 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7016 			hw->device_id;
7017 		while ((reg_group = ixgbevf_regs[g_ind++]))
7018 			count += ixgbe_read_regs_group(dev, &data[count],
7019 						      reg_group);
7020 		return 0;
7021 	}
7022 
7023 	return -ENOTSUP;
7024 }
7025 
7026 static int
7027 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
7028 {
7029 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7030 
7031 	/* Return unit is byte count */
7032 	return hw->eeprom.word_size * 2;
7033 }
7034 
7035 static int
7036 ixgbe_get_eeprom(struct rte_eth_dev *dev,
7037 		struct rte_dev_eeprom_info *in_eeprom)
7038 {
7039 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7040 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7041 	uint16_t *data = in_eeprom->data;
7042 	int first, length;
7043 
7044 	first = in_eeprom->offset >> 1;
7045 	length = in_eeprom->length >> 1;
7046 	if ((first > hw->eeprom.word_size) ||
7047 	    ((first + length) > hw->eeprom.word_size))
7048 		return -EINVAL;
7049 
7050 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7051 
7052 	return eeprom->ops.read_buffer(hw, first, length, data);
7053 }
7054 
7055 static int
7056 ixgbe_set_eeprom(struct rte_eth_dev *dev,
7057 		struct rte_dev_eeprom_info *in_eeprom)
7058 {
7059 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7060 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7061 	uint16_t *data = in_eeprom->data;
7062 	int first, length;
7063 
7064 	first = in_eeprom->offset >> 1;
7065 	length = in_eeprom->length >> 1;
7066 	if ((first > hw->eeprom.word_size) ||
7067 	    ((first + length) > hw->eeprom.word_size))
7068 		return -EINVAL;
7069 
7070 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7071 
7072 	return eeprom->ops.write_buffer(hw,  first, length, data);
7073 }
7074 
7075 static int
7076 ixgbe_get_module_info(struct rte_eth_dev *dev,
7077 		      struct rte_eth_dev_module_info *modinfo)
7078 {
7079 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7080 	uint32_t status;
7081 	uint8_t sff8472_rev, addr_mode;
7082 	bool page_swap = false;
7083 
7084 	/* Check whether we support SFF-8472 or not */
7085 	status = hw->phy.ops.read_i2c_eeprom(hw,
7086 					     IXGBE_SFF_SFF_8472_COMP,
7087 					     &sff8472_rev);
7088 	if (status != 0)
7089 		return -EIO;
7090 
7091 	/* addressing mode is not supported */
7092 	status = hw->phy.ops.read_i2c_eeprom(hw,
7093 					     IXGBE_SFF_SFF_8472_SWAP,
7094 					     &addr_mode);
7095 	if (status != 0)
7096 		return -EIO;
7097 
7098 	if (addr_mode & IXGBE_SFF_ADDRESSING_MODE) {
7099 		PMD_DRV_LOG(ERR,
7100 			    "Address change required to access page 0xA2, "
7101 			    "but not supported. Please report the module "
7102 			    "type to the driver maintainers.");
7103 		page_swap = true;
7104 	}
7105 
7106 	if (sff8472_rev == IXGBE_SFF_SFF_8472_UNSUP || page_swap) {
7107 		/* We have a SFP, but it does not support SFF-8472 */
7108 		modinfo->type = RTE_ETH_MODULE_SFF_8079;
7109 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
7110 	} else {
7111 		/* We have a SFP which supports a revision of SFF-8472. */
7112 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
7113 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
7114 	}
7115 
7116 	return 0;
7117 }
7118 
7119 static int
7120 ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
7121 			struct rte_dev_eeprom_info *info)
7122 {
7123 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7124 	uint32_t status = IXGBE_ERR_PHY_ADDR_INVALID;
7125 	uint8_t databyte = 0xFF;
7126 	uint8_t *data = info->data;
7127 	uint32_t i = 0;
7128 
7129 	for (i = info->offset; i < info->offset + info->length; i++) {
7130 		if (i < RTE_ETH_MODULE_SFF_8079_LEN)
7131 			status = hw->phy.ops.read_i2c_eeprom(hw, i, &databyte);
7132 		else
7133 			status = hw->phy.ops.read_i2c_sff8472(hw, i, &databyte);
7134 
7135 		if (status != 0)
7136 			return -EIO;
7137 
7138 		data[i - info->offset] = databyte;
7139 	}
7140 
7141 	return 0;
7142 }
7143 
7144 uint16_t
7145 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
7146 	switch (mac_type) {
7147 	case ixgbe_mac_X550:
7148 	case ixgbe_mac_X550EM_x:
7149 	case ixgbe_mac_X550EM_a:
7150 		return RTE_ETH_RSS_RETA_SIZE_512;
7151 	case ixgbe_mac_X550_vf:
7152 	case ixgbe_mac_X550EM_x_vf:
7153 	case ixgbe_mac_X550EM_a_vf:
7154 		return RTE_ETH_RSS_RETA_SIZE_64;
7155 	case ixgbe_mac_X540_vf:
7156 	case ixgbe_mac_82599_vf:
7157 		return 0;
7158 	default:
7159 		return RTE_ETH_RSS_RETA_SIZE_128;
7160 	}
7161 }
7162 
7163 uint32_t
7164 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
7165 	switch (mac_type) {
7166 	case ixgbe_mac_X550:
7167 	case ixgbe_mac_X550EM_x:
7168 	case ixgbe_mac_X550EM_a:
7169 		if (reta_idx < RTE_ETH_RSS_RETA_SIZE_128)
7170 			return IXGBE_RETA(reta_idx >> 2);
7171 		else
7172 			return IXGBE_ERETA((reta_idx - RTE_ETH_RSS_RETA_SIZE_128) >> 2);
7173 	case ixgbe_mac_X550_vf:
7174 	case ixgbe_mac_X550EM_x_vf:
7175 	case ixgbe_mac_X550EM_a_vf:
7176 		return IXGBE_VFRETA(reta_idx >> 2);
7177 	default:
7178 		return IXGBE_RETA(reta_idx >> 2);
7179 	}
7180 }
7181 
7182 uint32_t
7183 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
7184 	switch (mac_type) {
7185 	case ixgbe_mac_X550_vf:
7186 	case ixgbe_mac_X550EM_x_vf:
7187 	case ixgbe_mac_X550EM_a_vf:
7188 		return IXGBE_VFMRQC;
7189 	default:
7190 		return IXGBE_MRQC;
7191 	}
7192 }
7193 
7194 uint32_t
7195 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
7196 	switch (mac_type) {
7197 	case ixgbe_mac_X550_vf:
7198 	case ixgbe_mac_X550EM_x_vf:
7199 	case ixgbe_mac_X550EM_a_vf:
7200 		return IXGBE_VFRSSRK(i);
7201 	default:
7202 		return IXGBE_RSSRK(i);
7203 	}
7204 }
7205 
7206 bool
7207 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
7208 	switch (mac_type) {
7209 	case ixgbe_mac_82599_vf:
7210 	case ixgbe_mac_X540_vf:
7211 		return 0;
7212 	default:
7213 		return 1;
7214 	}
7215 }
7216 
7217 static int
7218 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
7219 			struct rte_eth_dcb_info *dcb_info)
7220 {
7221 	struct ixgbe_dcb_config *dcb_config =
7222 			IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
7223 	struct ixgbe_dcb_tc_config *tc;
7224 	struct rte_eth_dcb_tc_queue_mapping *tc_queue;
7225 	uint8_t nb_tcs;
7226 	uint8_t i, j;
7227 
7228 	if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_DCB_FLAG)
7229 		dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
7230 	else
7231 		dcb_info->nb_tcs = 1;
7232 
7233 	tc_queue = &dcb_info->tc_queue;
7234 	nb_tcs = dcb_info->nb_tcs;
7235 
7236 	if (dcb_config->vt_mode) { /* vt is enabled*/
7237 		struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
7238 				&dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
7239 		for (i = 0; i < RTE_ETH_DCB_NUM_USER_PRIORITIES; i++)
7240 			dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
7241 		if (RTE_ETH_DEV_SRIOV(dev).active > 0) {
7242 			for (j = 0; j < nb_tcs; j++) {
7243 				tc_queue->tc_rxq[0][j].base = j;
7244 				tc_queue->tc_rxq[0][j].nb_queue = 1;
7245 				tc_queue->tc_txq[0][j].base = j;
7246 				tc_queue->tc_txq[0][j].nb_queue = 1;
7247 			}
7248 		} else {
7249 			for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
7250 				for (j = 0; j < nb_tcs; j++) {
7251 					tc_queue->tc_rxq[i][j].base =
7252 						i * nb_tcs + j;
7253 					tc_queue->tc_rxq[i][j].nb_queue = 1;
7254 					tc_queue->tc_txq[i][j].base =
7255 						i * nb_tcs + j;
7256 					tc_queue->tc_txq[i][j].nb_queue = 1;
7257 				}
7258 			}
7259 		}
7260 	} else { /* vt is disabled*/
7261 		struct rte_eth_dcb_rx_conf *rx_conf =
7262 				&dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
7263 		for (i = 0; i < RTE_ETH_DCB_NUM_USER_PRIORITIES; i++)
7264 			dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
7265 		if (dcb_info->nb_tcs == RTE_ETH_4_TCS) {
7266 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7267 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
7268 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7269 			}
7270 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7271 			dcb_info->tc_queue.tc_txq[0][1].base = 64;
7272 			dcb_info->tc_queue.tc_txq[0][2].base = 96;
7273 			dcb_info->tc_queue.tc_txq[0][3].base = 112;
7274 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
7275 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7276 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7277 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7278 		} else if (dcb_info->nb_tcs == RTE_ETH_8_TCS) {
7279 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7280 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
7281 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7282 			}
7283 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7284 			dcb_info->tc_queue.tc_txq[0][1].base = 32;
7285 			dcb_info->tc_queue.tc_txq[0][2].base = 64;
7286 			dcb_info->tc_queue.tc_txq[0][3].base = 80;
7287 			dcb_info->tc_queue.tc_txq[0][4].base = 96;
7288 			dcb_info->tc_queue.tc_txq[0][5].base = 104;
7289 			dcb_info->tc_queue.tc_txq[0][6].base = 112;
7290 			dcb_info->tc_queue.tc_txq[0][7].base = 120;
7291 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
7292 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7293 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7294 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7295 			dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
7296 			dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
7297 			dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
7298 			dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
7299 		}
7300 	}
7301 	for (i = 0; i < dcb_info->nb_tcs; i++) {
7302 		tc = &dcb_config->tc_config[i];
7303 		dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
7304 	}
7305 	return 0;
7306 }
7307 
7308 /* Update e-tag ether type */
7309 static int
7310 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
7311 			    uint16_t ether_type)
7312 {
7313 	uint32_t etag_etype;
7314 
7315 	if (hw->mac.type != ixgbe_mac_X550 &&
7316 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7317 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7318 		return -ENOTSUP;
7319 	}
7320 
7321 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7322 	etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
7323 	etag_etype |= ether_type;
7324 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7325 	IXGBE_WRITE_FLUSH(hw);
7326 
7327 	return 0;
7328 }
7329 
7330 /* Enable e-tag tunnel */
7331 static int
7332 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
7333 {
7334 	uint32_t etag_etype;
7335 
7336 	if (hw->mac.type != ixgbe_mac_X550 &&
7337 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7338 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7339 		return -ENOTSUP;
7340 	}
7341 
7342 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7343 	etag_etype |= IXGBE_ETAG_ETYPE_VALID;
7344 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7345 	IXGBE_WRITE_FLUSH(hw);
7346 
7347 	return 0;
7348 }
7349 
7350 static int
7351 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
7352 		       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7353 {
7354 	int ret = 0;
7355 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7356 	uint32_t i, rar_entries;
7357 	uint32_t rar_low, rar_high;
7358 
7359 	if (hw->mac.type != ixgbe_mac_X550 &&
7360 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7361 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7362 		return -ENOTSUP;
7363 	}
7364 
7365 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7366 
7367 	for (i = 1; i < rar_entries; i++) {
7368 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7369 		rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
7370 		if ((rar_high & IXGBE_RAH_AV) &&
7371 		    (rar_high & IXGBE_RAH_ADTYPE) &&
7372 		    ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
7373 		     l2_tunnel->tunnel_id)) {
7374 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
7375 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
7376 
7377 			ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
7378 
7379 			return ret;
7380 		}
7381 	}
7382 
7383 	return ret;
7384 }
7385 
7386 static int
7387 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
7388 		       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7389 {
7390 	int ret = 0;
7391 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7392 	uint32_t i, rar_entries;
7393 	uint32_t rar_low, rar_high;
7394 
7395 	if (hw->mac.type != ixgbe_mac_X550 &&
7396 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7397 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7398 		return -ENOTSUP;
7399 	}
7400 
7401 	/* One entry for one tunnel. Try to remove potential existing entry. */
7402 	ixgbe_e_tag_filter_del(dev, l2_tunnel);
7403 
7404 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7405 
7406 	for (i = 1; i < rar_entries; i++) {
7407 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7408 		if (rar_high & IXGBE_RAH_AV) {
7409 			continue;
7410 		} else {
7411 			ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
7412 			rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
7413 			rar_low = l2_tunnel->tunnel_id;
7414 
7415 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
7416 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
7417 
7418 			return ret;
7419 		}
7420 	}
7421 
7422 	PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
7423 		     " Please remove a rule before adding a new one.");
7424 	return -EINVAL;
7425 }
7426 
7427 static inline struct ixgbe_l2_tn_filter *
7428 ixgbe_l2_tn_filter_lookup(struct ixgbe_l2_tn_info *l2_tn_info,
7429 			  struct ixgbe_l2_tn_key *key)
7430 {
7431 	int ret;
7432 
7433 	ret = rte_hash_lookup(l2_tn_info->hash_handle, (const void *)key);
7434 	if (ret < 0)
7435 		return NULL;
7436 
7437 	return l2_tn_info->hash_map[ret];
7438 }
7439 
7440 static inline int
7441 ixgbe_insert_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7442 			  struct ixgbe_l2_tn_filter *l2_tn_filter)
7443 {
7444 	int ret;
7445 
7446 	ret = rte_hash_add_key(l2_tn_info->hash_handle,
7447 			       &l2_tn_filter->key);
7448 
7449 	if (ret < 0) {
7450 		PMD_DRV_LOG(ERR,
7451 			    "Failed to insert L2 tunnel filter"
7452 			    " to hash table %d!",
7453 			    ret);
7454 		return ret;
7455 	}
7456 
7457 	l2_tn_info->hash_map[ret] = l2_tn_filter;
7458 
7459 	TAILQ_INSERT_TAIL(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7460 
7461 	return 0;
7462 }
7463 
7464 static inline int
7465 ixgbe_remove_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7466 			  struct ixgbe_l2_tn_key *key)
7467 {
7468 	int ret;
7469 	struct ixgbe_l2_tn_filter *l2_tn_filter;
7470 
7471 	ret = rte_hash_del_key(l2_tn_info->hash_handle, key);
7472 
7473 	if (ret < 0) {
7474 		PMD_DRV_LOG(ERR,
7475 			    "No such L2 tunnel filter to delete %d!",
7476 			    ret);
7477 		return ret;
7478 	}
7479 
7480 	l2_tn_filter = l2_tn_info->hash_map[ret];
7481 	l2_tn_info->hash_map[ret] = NULL;
7482 
7483 	TAILQ_REMOVE(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7484 	rte_free(l2_tn_filter);
7485 
7486 	return 0;
7487 }
7488 
7489 /* Add l2 tunnel filter */
7490 int
7491 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
7492 			       struct ixgbe_l2_tunnel_conf *l2_tunnel,
7493 			       bool restore)
7494 {
7495 	int ret;
7496 	struct ixgbe_l2_tn_info *l2_tn_info =
7497 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7498 	struct ixgbe_l2_tn_key key;
7499 	struct ixgbe_l2_tn_filter *node;
7500 
7501 	if (!restore) {
7502 		key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7503 		key.tn_id = l2_tunnel->tunnel_id;
7504 
7505 		node = ixgbe_l2_tn_filter_lookup(l2_tn_info, &key);
7506 
7507 		if (node) {
7508 			PMD_DRV_LOG(ERR,
7509 				    "The L2 tunnel filter already exists!");
7510 			return -EINVAL;
7511 		}
7512 
7513 		node = rte_zmalloc("ixgbe_l2_tn",
7514 				   sizeof(struct ixgbe_l2_tn_filter),
7515 				   0);
7516 		if (!node)
7517 			return -ENOMEM;
7518 
7519 		rte_memcpy(&node->key,
7520 				 &key,
7521 				 sizeof(struct ixgbe_l2_tn_key));
7522 		node->pool = l2_tunnel->pool;
7523 		ret = ixgbe_insert_l2_tn_filter(l2_tn_info, node);
7524 		if (ret < 0) {
7525 			rte_free(node);
7526 			return ret;
7527 		}
7528 	}
7529 
7530 	switch (l2_tunnel->l2_tunnel_type) {
7531 	case RTE_ETH_L2_TUNNEL_TYPE_E_TAG:
7532 		ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
7533 		break;
7534 	default:
7535 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7536 		ret = -EINVAL;
7537 		break;
7538 	}
7539 
7540 	if ((!restore) && (ret < 0))
7541 		(void)ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7542 
7543 	return ret;
7544 }
7545 
7546 /* Delete l2 tunnel filter */
7547 int
7548 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
7549 			       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7550 {
7551 	int ret;
7552 	struct ixgbe_l2_tn_info *l2_tn_info =
7553 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7554 	struct ixgbe_l2_tn_key key;
7555 
7556 	key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7557 	key.tn_id = l2_tunnel->tunnel_id;
7558 	ret = ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7559 	if (ret < 0)
7560 		return ret;
7561 
7562 	switch (l2_tunnel->l2_tunnel_type) {
7563 	case RTE_ETH_L2_TUNNEL_TYPE_E_TAG:
7564 		ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
7565 		break;
7566 	default:
7567 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7568 		ret = -EINVAL;
7569 		break;
7570 	}
7571 
7572 	return ret;
7573 }
7574 
7575 static int
7576 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
7577 {
7578 	int ret = 0;
7579 	uint32_t ctrl;
7580 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7581 
7582 	if (hw->mac.type != ixgbe_mac_X550 &&
7583 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7584 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7585 		return -ENOTSUP;
7586 	}
7587 
7588 	ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
7589 	ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
7590 	if (en)
7591 		ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
7592 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
7593 
7594 	return ret;
7595 }
7596 
7597 static int
7598 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
7599 			uint16_t port)
7600 {
7601 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
7602 	IXGBE_WRITE_FLUSH(hw);
7603 
7604 	return 0;
7605 }
7606 
7607 /* There's only one register for VxLAN UDP port.
7608  * So, we cannot add several ports. Will update it.
7609  */
7610 static int
7611 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
7612 		     uint16_t port)
7613 {
7614 	if (port == 0) {
7615 		PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
7616 		return -EINVAL;
7617 	}
7618 
7619 	return ixgbe_update_vxlan_port(hw, port);
7620 }
7621 
7622 /* We cannot delete the VxLAN port. For there's a register for VxLAN
7623  * UDP port, it must have a value.
7624  * So, will reset it to the original value 0.
7625  */
7626 static int
7627 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
7628 		     uint16_t port)
7629 {
7630 	uint16_t cur_port;
7631 
7632 	cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
7633 
7634 	if (cur_port != port) {
7635 		PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
7636 		return -EINVAL;
7637 	}
7638 
7639 	return ixgbe_update_vxlan_port(hw, 0);
7640 }
7641 
7642 /* Add UDP tunneling port */
7643 static int
7644 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
7645 			      struct rte_eth_udp_tunnel *udp_tunnel)
7646 {
7647 	int ret = 0;
7648 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7649 
7650 	if (hw->mac.type != ixgbe_mac_X550 &&
7651 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7652 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7653 		return -ENOTSUP;
7654 	}
7655 
7656 	if (udp_tunnel == NULL)
7657 		return -EINVAL;
7658 
7659 	switch (udp_tunnel->prot_type) {
7660 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
7661 		ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
7662 		break;
7663 
7664 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
7665 	case RTE_ETH_TUNNEL_TYPE_TEREDO:
7666 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7667 		ret = -EINVAL;
7668 		break;
7669 
7670 	default:
7671 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7672 		ret = -EINVAL;
7673 		break;
7674 	}
7675 
7676 	return ret;
7677 }
7678 
7679 /* Remove UDP tunneling port */
7680 static int
7681 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
7682 			      struct rte_eth_udp_tunnel *udp_tunnel)
7683 {
7684 	int ret = 0;
7685 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7686 
7687 	if (hw->mac.type != ixgbe_mac_X550 &&
7688 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7689 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7690 		return -ENOTSUP;
7691 	}
7692 
7693 	if (udp_tunnel == NULL)
7694 		return -EINVAL;
7695 
7696 	switch (udp_tunnel->prot_type) {
7697 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
7698 		ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
7699 		break;
7700 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
7701 	case RTE_ETH_TUNNEL_TYPE_TEREDO:
7702 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7703 		ret = -EINVAL;
7704 		break;
7705 	default:
7706 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7707 		ret = -EINVAL;
7708 		break;
7709 	}
7710 
7711 	return ret;
7712 }
7713 
7714 static int
7715 ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev)
7716 {
7717 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7718 	int ret;
7719 
7720 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_PROMISC)) {
7721 	case IXGBE_SUCCESS:
7722 		ret = 0;
7723 		break;
7724 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7725 		ret = -ENOTSUP;
7726 		break;
7727 	default:
7728 		ret = -EAGAIN;
7729 		break;
7730 	}
7731 
7732 	return ret;
7733 }
7734 
7735 static int
7736 ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev)
7737 {
7738 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7739 	int ret;
7740 
7741 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_NONE)) {
7742 	case IXGBE_SUCCESS:
7743 		ret = 0;
7744 		break;
7745 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7746 		ret = -ENOTSUP;
7747 		break;
7748 	default:
7749 		ret = -EAGAIN;
7750 		break;
7751 	}
7752 
7753 	return ret;
7754 }
7755 
7756 static int
7757 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
7758 {
7759 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7760 	int ret;
7761 	int mode = IXGBEVF_XCAST_MODE_ALLMULTI;
7762 
7763 	switch (hw->mac.ops.update_xcast_mode(hw, mode)) {
7764 	case IXGBE_SUCCESS:
7765 		ret = 0;
7766 		break;
7767 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7768 		ret = -ENOTSUP;
7769 		break;
7770 	default:
7771 		ret = -EAGAIN;
7772 		break;
7773 	}
7774 
7775 	return ret;
7776 }
7777 
7778 static int
7779 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
7780 {
7781 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7782 	int ret;
7783 
7784 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_MULTI)) {
7785 	case IXGBE_SUCCESS:
7786 		ret = 0;
7787 		break;
7788 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7789 		ret = -ENOTSUP;
7790 		break;
7791 	default:
7792 		ret = -EAGAIN;
7793 		break;
7794 	}
7795 
7796 	return ret;
7797 }
7798 
7799 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
7800 {
7801 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7802 	u32 in_msg = 0;
7803 
7804 	/* peek the message first */
7805 	in_msg = IXGBE_READ_REG(hw, IXGBE_VFMBMEM);
7806 
7807 	/* PF reset VF event */
7808 	if (in_msg == IXGBE_PF_CONTROL_MSG) {
7809 		/* dummy mbx read to ack pf */
7810 		if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
7811 			return;
7812 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
7813 					     NULL);
7814 	}
7815 }
7816 
7817 static int
7818 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
7819 {
7820 	uint32_t eicr;
7821 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7822 	struct ixgbe_interrupt *intr =
7823 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7824 	ixgbevf_intr_disable(dev);
7825 
7826 	/* read-on-clear nic registers here */
7827 	eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
7828 	intr->flags = 0;
7829 
7830 	/* only one misc vector supported - mailbox */
7831 	eicr &= IXGBE_VTEICR_MASK;
7832 	if (eicr == IXGBE_MISC_VEC_ID)
7833 		intr->flags |= IXGBE_FLAG_MAILBOX;
7834 
7835 	return 0;
7836 }
7837 
7838 static int
7839 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
7840 {
7841 	struct ixgbe_interrupt *intr =
7842 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7843 
7844 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
7845 		ixgbevf_mbx_process(dev);
7846 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
7847 	}
7848 
7849 	ixgbevf_intr_enable(dev);
7850 
7851 	return 0;
7852 }
7853 
7854 static void
7855 ixgbevf_dev_interrupt_handler(void *param)
7856 {
7857 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
7858 
7859 	ixgbevf_dev_interrupt_get_status(dev);
7860 	ixgbevf_dev_interrupt_action(dev);
7861 }
7862 
7863 /**
7864  *  ixgbe_disable_sec_tx_path_generic - Stops the transmit data path
7865  *  @hw: pointer to hardware structure
7866  *
7867  *  Stops the transmit data path and waits for the HW to internally empty
7868  *  the Tx security block
7869  **/
7870 int ixgbe_disable_sec_tx_path_generic(struct ixgbe_hw *hw)
7871 {
7872 #define IXGBE_MAX_SECTX_POLL 40
7873 
7874 	int i;
7875 	int sectxreg;
7876 
7877 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
7878 	sectxreg |= IXGBE_SECTXCTRL_TX_DIS;
7879 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
7880 	for (i = 0; i < IXGBE_MAX_SECTX_POLL; i++) {
7881 		sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXSTAT);
7882 		if (sectxreg & IXGBE_SECTXSTAT_SECTX_RDY)
7883 			break;
7884 		/* Use interrupt-safe sleep just in case */
7885 		usec_delay(1000);
7886 	}
7887 
7888 	/* For informational purposes only */
7889 	if (i >= IXGBE_MAX_SECTX_POLL)
7890 		PMD_DRV_LOG(DEBUG, "Tx unit being enabled before security "
7891 			 "path fully disabled.  Continuing with init.");
7892 
7893 	return IXGBE_SUCCESS;
7894 }
7895 
7896 /**
7897  *  ixgbe_enable_sec_tx_path_generic - Enables the transmit data path
7898  *  @hw: pointer to hardware structure
7899  *
7900  *  Enables the transmit data path.
7901  **/
7902 int ixgbe_enable_sec_tx_path_generic(struct ixgbe_hw *hw)
7903 {
7904 	uint32_t sectxreg;
7905 
7906 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
7907 	sectxreg &= ~IXGBE_SECTXCTRL_TX_DIS;
7908 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
7909 	IXGBE_WRITE_FLUSH(hw);
7910 
7911 	return IXGBE_SUCCESS;
7912 }
7913 
7914 /* restore n-tuple filter */
7915 static inline void
7916 ixgbe_ntuple_filter_restore(struct rte_eth_dev *dev)
7917 {
7918 	struct ixgbe_filter_info *filter_info =
7919 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7920 	struct ixgbe_5tuple_filter *node;
7921 
7922 	TAILQ_FOREACH(node, &filter_info->fivetuple_list, entries) {
7923 		ixgbe_inject_5tuple_filter(dev, node);
7924 	}
7925 }
7926 
7927 /* restore ethernet type filter */
7928 static inline void
7929 ixgbe_ethertype_filter_restore(struct rte_eth_dev *dev)
7930 {
7931 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7932 	struct ixgbe_filter_info *filter_info =
7933 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7934 	int i;
7935 
7936 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
7937 		if (filter_info->ethertype_mask & (1 << i)) {
7938 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i),
7939 					filter_info->ethertype_filters[i].etqf);
7940 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i),
7941 					filter_info->ethertype_filters[i].etqs);
7942 			IXGBE_WRITE_FLUSH(hw);
7943 		}
7944 	}
7945 }
7946 
7947 /* restore SYN filter */
7948 static inline void
7949 ixgbe_syn_filter_restore(struct rte_eth_dev *dev)
7950 {
7951 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7952 	struct ixgbe_filter_info *filter_info =
7953 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7954 	uint32_t synqf;
7955 
7956 	synqf = filter_info->syn_info;
7957 
7958 	if (synqf & IXGBE_SYN_FILTER_ENABLE) {
7959 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
7960 		IXGBE_WRITE_FLUSH(hw);
7961 	}
7962 }
7963 
7964 /* restore L2 tunnel filter */
7965 static inline void
7966 ixgbe_l2_tn_filter_restore(struct rte_eth_dev *dev)
7967 {
7968 	struct ixgbe_l2_tn_info *l2_tn_info =
7969 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7970 	struct ixgbe_l2_tn_filter *node;
7971 	struct ixgbe_l2_tunnel_conf l2_tn_conf;
7972 
7973 	TAILQ_FOREACH(node, &l2_tn_info->l2_tn_list, entries) {
7974 		l2_tn_conf.l2_tunnel_type = node->key.l2_tn_type;
7975 		l2_tn_conf.tunnel_id      = node->key.tn_id;
7976 		l2_tn_conf.pool           = node->pool;
7977 		(void)ixgbe_dev_l2_tunnel_filter_add(dev, &l2_tn_conf, TRUE);
7978 	}
7979 }
7980 
7981 /* restore rss filter */
7982 static inline void
7983 ixgbe_rss_filter_restore(struct rte_eth_dev *dev)
7984 {
7985 	struct ixgbe_filter_info *filter_info =
7986 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7987 
7988 	if (filter_info->rss_info.conf.queue_num)
7989 		ixgbe_config_rss_filter(dev,
7990 			&filter_info->rss_info, TRUE);
7991 }
7992 
7993 static int
7994 ixgbe_filter_restore(struct rte_eth_dev *dev)
7995 {
7996 	ixgbe_ntuple_filter_restore(dev);
7997 	ixgbe_ethertype_filter_restore(dev);
7998 	ixgbe_syn_filter_restore(dev);
7999 	ixgbe_fdir_filter_restore(dev);
8000 	ixgbe_l2_tn_filter_restore(dev);
8001 	ixgbe_rss_filter_restore(dev);
8002 
8003 	return 0;
8004 }
8005 
8006 static void
8007 ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev)
8008 {
8009 	struct ixgbe_l2_tn_info *l2_tn_info =
8010 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8011 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8012 
8013 	if (l2_tn_info->e_tag_en)
8014 		(void)ixgbe_e_tag_enable(hw);
8015 
8016 	if (l2_tn_info->e_tag_fwd_en)
8017 		(void)ixgbe_e_tag_forwarding_en_dis(dev, 1);
8018 
8019 	(void)ixgbe_update_e_tag_eth_type(hw, l2_tn_info->e_tag_ether_type);
8020 }
8021 
8022 /* remove all the n-tuple filters */
8023 void
8024 ixgbe_clear_all_ntuple_filter(struct rte_eth_dev *dev)
8025 {
8026 	struct ixgbe_filter_info *filter_info =
8027 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8028 	struct ixgbe_5tuple_filter *p_5tuple;
8029 
8030 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
8031 		ixgbe_remove_5tuple_filter(dev, p_5tuple);
8032 }
8033 
8034 /* remove all the ether type filters */
8035 void
8036 ixgbe_clear_all_ethertype_filter(struct rte_eth_dev *dev)
8037 {
8038 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8039 	struct ixgbe_filter_info *filter_info =
8040 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8041 	int i;
8042 
8043 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8044 		if (filter_info->ethertype_mask & (1 << i) &&
8045 		    !filter_info->ethertype_filters[i].conf) {
8046 			(void)ixgbe_ethertype_filter_remove(filter_info,
8047 							    (uint8_t)i);
8048 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i), 0);
8049 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i), 0);
8050 			IXGBE_WRITE_FLUSH(hw);
8051 		}
8052 	}
8053 }
8054 
8055 /* remove the SYN filter */
8056 void
8057 ixgbe_clear_syn_filter(struct rte_eth_dev *dev)
8058 {
8059 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8060 	struct ixgbe_filter_info *filter_info =
8061 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8062 
8063 	if (filter_info->syn_info & IXGBE_SYN_FILTER_ENABLE) {
8064 		filter_info->syn_info = 0;
8065 
8066 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, 0);
8067 		IXGBE_WRITE_FLUSH(hw);
8068 	}
8069 }
8070 
8071 /* remove all the L2 tunnel filters */
8072 int
8073 ixgbe_clear_all_l2_tn_filter(struct rte_eth_dev *dev)
8074 {
8075 	struct ixgbe_l2_tn_info *l2_tn_info =
8076 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8077 	struct ixgbe_l2_tn_filter *l2_tn_filter;
8078 	struct ixgbe_l2_tunnel_conf l2_tn_conf;
8079 	int ret = 0;
8080 
8081 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
8082 		l2_tn_conf.l2_tunnel_type = l2_tn_filter->key.l2_tn_type;
8083 		l2_tn_conf.tunnel_id      = l2_tn_filter->key.tn_id;
8084 		l2_tn_conf.pool           = l2_tn_filter->pool;
8085 		ret = ixgbe_dev_l2_tunnel_filter_del(dev, &l2_tn_conf);
8086 		if (ret < 0)
8087 			return ret;
8088 	}
8089 
8090 	return 0;
8091 }
8092 
8093 void
8094 ixgbe_dev_macsec_setting_save(struct rte_eth_dev *dev,
8095 				struct ixgbe_macsec_setting *macsec_setting)
8096 {
8097 	struct ixgbe_macsec_setting *macsec =
8098 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8099 
8100 	macsec->offload_en = macsec_setting->offload_en;
8101 	macsec->encrypt_en = macsec_setting->encrypt_en;
8102 	macsec->replayprotect_en = macsec_setting->replayprotect_en;
8103 }
8104 
8105 void
8106 ixgbe_dev_macsec_setting_reset(struct rte_eth_dev *dev)
8107 {
8108 	struct ixgbe_macsec_setting *macsec =
8109 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8110 
8111 	macsec->offload_en = 0;
8112 	macsec->encrypt_en = 0;
8113 	macsec->replayprotect_en = 0;
8114 }
8115 
8116 void
8117 ixgbe_dev_macsec_register_enable(struct rte_eth_dev *dev,
8118 				struct ixgbe_macsec_setting *macsec_setting)
8119 {
8120 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8121 	uint32_t ctrl;
8122 	uint8_t en = macsec_setting->encrypt_en;
8123 	uint8_t rp = macsec_setting->replayprotect_en;
8124 
8125 	/**
8126 	 * Workaround:
8127 	 * As no ixgbe_disable_sec_rx_path equivalent is
8128 	 * implemented for tx in the base code, and we are
8129 	 * not allowed to modify the base code in DPDK, so
8130 	 * just call the hand-written one directly for now.
8131 	 * The hardware support has been checked by
8132 	 * ixgbe_disable_sec_rx_path().
8133 	 */
8134 	ixgbe_disable_sec_tx_path_generic(hw);
8135 
8136 	/* Enable Ethernet CRC (required by MACsec offload) */
8137 	ctrl = IXGBE_READ_REG(hw, IXGBE_HLREG0);
8138 	ctrl |= IXGBE_HLREG0_TXCRCEN | IXGBE_HLREG0_RXCRCSTRP;
8139 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, ctrl);
8140 
8141 	/* Enable the TX and RX crypto engines */
8142 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8143 	ctrl &= ~IXGBE_SECTXCTRL_SECTX_DIS;
8144 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8145 
8146 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8147 	ctrl &= ~IXGBE_SECRXCTRL_SECRX_DIS;
8148 	IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8149 
8150 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
8151 	ctrl &= ~IXGBE_SECTX_MINSECIFG_MASK;
8152 	ctrl |= 0x3;
8153 	IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, ctrl);
8154 
8155 	/* Enable SA lookup */
8156 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8157 	ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8158 	ctrl |= en ? IXGBE_LSECTXCTRL_AUTH_ENCRYPT :
8159 		     IXGBE_LSECTXCTRL_AUTH;
8160 	ctrl |= IXGBE_LSECTXCTRL_AISCI;
8161 	ctrl &= ~IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8162 	ctrl |= IXGBE_MACSEC_PNTHRSH & IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8163 	IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8164 
8165 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8166 	ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8167 	ctrl |= IXGBE_LSECRXCTRL_STRICT << IXGBE_LSECRXCTRL_EN_SHIFT;
8168 	ctrl &= ~IXGBE_LSECRXCTRL_PLSH;
8169 	if (rp)
8170 		ctrl |= IXGBE_LSECRXCTRL_RP;
8171 	else
8172 		ctrl &= ~IXGBE_LSECRXCTRL_RP;
8173 	IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8174 
8175 	/* Start the data paths */
8176 	ixgbe_enable_sec_rx_path(hw);
8177 	/**
8178 	 * Workaround:
8179 	 * As no ixgbe_enable_sec_rx_path equivalent is
8180 	 * implemented for tx in the base code, and we are
8181 	 * not allowed to modify the base code in DPDK, so
8182 	 * just call the hand-written one directly for now.
8183 	 */
8184 	ixgbe_enable_sec_tx_path_generic(hw);
8185 }
8186 
8187 void
8188 ixgbe_dev_macsec_register_disable(struct rte_eth_dev *dev)
8189 {
8190 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8191 	uint32_t ctrl;
8192 
8193 	/**
8194 	 * Workaround:
8195 	 * As no ixgbe_disable_sec_rx_path equivalent is
8196 	 * implemented for tx in the base code, and we are
8197 	 * not allowed to modify the base code in DPDK, so
8198 	 * just call the hand-written one directly for now.
8199 	 * The hardware support has been checked by
8200 	 * ixgbe_disable_sec_rx_path().
8201 	 */
8202 	ixgbe_disable_sec_tx_path_generic(hw);
8203 
8204 	/* Disable the TX and RX crypto engines */
8205 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8206 	ctrl |= IXGBE_SECTXCTRL_SECTX_DIS;
8207 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8208 
8209 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8210 	ctrl |= IXGBE_SECRXCTRL_SECRX_DIS;
8211 	IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8212 
8213 	/* Disable SA lookup */
8214 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8215 	ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8216 	ctrl |= IXGBE_LSECTXCTRL_DISABLE;
8217 	IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8218 
8219 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8220 	ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8221 	ctrl |= IXGBE_LSECRXCTRL_DISABLE << IXGBE_LSECRXCTRL_EN_SHIFT;
8222 	IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8223 
8224 	/* Start the data paths */
8225 	ixgbe_enable_sec_rx_path(hw);
8226 	/**
8227 	 * Workaround:
8228 	 * As no ixgbe_enable_sec_rx_path equivalent is
8229 	 * implemented for tx in the base code, and we are
8230 	 * not allowed to modify the base code in DPDK, so
8231 	 * just call the hand-written one directly for now.
8232 	 */
8233 	ixgbe_enable_sec_tx_path_generic(hw);
8234 }
8235 
8236 RTE_PMD_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd);
8237 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
8238 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe, "* igb_uio | uio_pci_generic | vfio-pci");
8239 RTE_PMD_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd);
8240 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);
8241 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe_vf, "* igb_uio | vfio-pci");
8242 RTE_PMD_REGISTER_PARAM_STRING(net_ixgbe_vf,
8243 			      IXGBEVF_DEVARG_PFLINK_FULLCHK "=<0|1>");
8244 
8245 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_init, init, NOTICE);
8246 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_driver, driver, NOTICE);
8247 
8248 #ifdef RTE_ETHDEV_DEBUG_RX
8249 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_rx, rx, DEBUG);
8250 #endif
8251 #ifdef RTE_ETHDEV_DEBUG_TX
8252 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_tx, tx, DEBUG);
8253 #endif
8254