xref: /dpdk/drivers/net/ixgbe/ixgbe_ethdev.c (revision 452c1916)
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 		rte_free(eth_dev->data->mac_addrs);
1228 		eth_dev->data->mac_addrs = NULL;
1229 		rte_free(eth_dev->data->hash_mac_addrs);
1230 		eth_dev->data->hash_mac_addrs = NULL;
1231 		return ret;
1232 	}
1233 
1234 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1235 	/* let hardware know driver is loaded */
1236 	ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1237 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1238 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1239 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1240 	IXGBE_WRITE_FLUSH(hw);
1241 
1242 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1243 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1244 			     (int) hw->mac.type, (int) hw->phy.type,
1245 			     (int) hw->phy.sfp_type);
1246 	else
1247 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1248 			     (int) hw->mac.type, (int) hw->phy.type);
1249 
1250 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1251 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1252 		     pci_dev->id.device_id);
1253 
1254 	rte_intr_callback_register(intr_handle,
1255 				   ixgbe_dev_interrupt_handler, eth_dev);
1256 
1257 	/* enable uio/vfio intr/eventfd mapping */
1258 	rte_intr_enable(intr_handle);
1259 
1260 	/* enable support intr */
1261 	ixgbe_enable_intr(eth_dev);
1262 
1263 	/* initialize filter info */
1264 	memset(filter_info, 0,
1265 	       sizeof(struct ixgbe_filter_info));
1266 
1267 	/* initialize 5tuple filter list */
1268 	TAILQ_INIT(&filter_info->fivetuple_list);
1269 
1270 	/* initialize flow director filter list & hash */
1271 	ixgbe_fdir_filter_init(eth_dev);
1272 
1273 	/* initialize l2 tunnel filter list & hash */
1274 	ixgbe_l2_tn_filter_init(eth_dev);
1275 
1276 	/* initialize flow filter lists */
1277 	ixgbe_filterlist_init();
1278 
1279 	/* initialize bandwidth configuration info */
1280 	memset(bw_conf, 0, sizeof(struct ixgbe_bw_conf));
1281 
1282 	/* initialize Traffic Manager configuration */
1283 	ixgbe_tm_conf_init(eth_dev);
1284 
1285 	return 0;
1286 }
1287 
1288 static int
1289 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1290 {
1291 	PMD_INIT_FUNC_TRACE();
1292 
1293 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1294 		return 0;
1295 
1296 	ixgbe_dev_close(eth_dev);
1297 
1298 	return 0;
1299 }
1300 
1301 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
1302 {
1303 	struct ixgbe_filter_info *filter_info =
1304 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1305 	struct ixgbe_5tuple_filter *p_5tuple;
1306 
1307 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
1308 		TAILQ_REMOVE(&filter_info->fivetuple_list,
1309 			     p_5tuple,
1310 			     entries);
1311 		rte_free(p_5tuple);
1312 	}
1313 	memset(filter_info->fivetuple_mask, 0,
1314 	       sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1315 
1316 	return 0;
1317 }
1318 
1319 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev)
1320 {
1321 	struct ixgbe_hw_fdir_info *fdir_info =
1322 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1323 	struct ixgbe_fdir_filter *fdir_filter;
1324 
1325 		if (fdir_info->hash_map)
1326 		rte_free(fdir_info->hash_map);
1327 	if (fdir_info->hash_handle)
1328 		rte_hash_free(fdir_info->hash_handle);
1329 
1330 	while ((fdir_filter = TAILQ_FIRST(&fdir_info->fdir_list))) {
1331 		TAILQ_REMOVE(&fdir_info->fdir_list,
1332 			     fdir_filter,
1333 			     entries);
1334 		rte_free(fdir_filter);
1335 	}
1336 
1337 	return 0;
1338 }
1339 
1340 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev)
1341 {
1342 	struct ixgbe_l2_tn_info *l2_tn_info =
1343 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1344 	struct ixgbe_l2_tn_filter *l2_tn_filter;
1345 
1346 	if (l2_tn_info->hash_map)
1347 		rte_free(l2_tn_info->hash_map);
1348 	if (l2_tn_info->hash_handle)
1349 		rte_hash_free(l2_tn_info->hash_handle);
1350 
1351 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
1352 		TAILQ_REMOVE(&l2_tn_info->l2_tn_list,
1353 			     l2_tn_filter,
1354 			     entries);
1355 		rte_free(l2_tn_filter);
1356 	}
1357 
1358 	return 0;
1359 }
1360 
1361 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev)
1362 {
1363 	struct ixgbe_hw_fdir_info *fdir_info =
1364 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1365 	char fdir_hash_name[RTE_HASH_NAMESIZE];
1366 	struct rte_hash_parameters fdir_hash_params = {
1367 		.name = fdir_hash_name,
1368 		.entries = IXGBE_MAX_FDIR_FILTER_NUM,
1369 		.key_len = sizeof(union ixgbe_atr_input),
1370 		.hash_func = rte_hash_crc,
1371 		.hash_func_init_val = 0,
1372 		.socket_id = rte_socket_id(),
1373 	};
1374 
1375 	TAILQ_INIT(&fdir_info->fdir_list);
1376 	snprintf(fdir_hash_name, RTE_HASH_NAMESIZE,
1377 		 "fdir_%s", eth_dev->device->name);
1378 	fdir_info->hash_handle = rte_hash_create(&fdir_hash_params);
1379 	if (!fdir_info->hash_handle) {
1380 		PMD_INIT_LOG(ERR, "Failed to create fdir hash table!");
1381 		return -EINVAL;
1382 	}
1383 	fdir_info->hash_map = rte_zmalloc("ixgbe",
1384 					  sizeof(struct ixgbe_fdir_filter *) *
1385 					  IXGBE_MAX_FDIR_FILTER_NUM,
1386 					  0);
1387 	if (!fdir_info->hash_map) {
1388 		PMD_INIT_LOG(ERR,
1389 			     "Failed to allocate memory for fdir hash map!");
1390 		rte_hash_free(fdir_info->hash_handle);
1391 		return -ENOMEM;
1392 	}
1393 	fdir_info->mask_added = FALSE;
1394 
1395 	return 0;
1396 }
1397 
1398 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev)
1399 {
1400 	struct ixgbe_l2_tn_info *l2_tn_info =
1401 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1402 	char l2_tn_hash_name[RTE_HASH_NAMESIZE];
1403 	struct rte_hash_parameters l2_tn_hash_params = {
1404 		.name = l2_tn_hash_name,
1405 		.entries = IXGBE_MAX_L2_TN_FILTER_NUM,
1406 		.key_len = sizeof(struct ixgbe_l2_tn_key),
1407 		.hash_func = rte_hash_crc,
1408 		.hash_func_init_val = 0,
1409 		.socket_id = rte_socket_id(),
1410 	};
1411 
1412 	TAILQ_INIT(&l2_tn_info->l2_tn_list);
1413 	snprintf(l2_tn_hash_name, RTE_HASH_NAMESIZE,
1414 		 "l2_tn_%s", eth_dev->device->name);
1415 	l2_tn_info->hash_handle = rte_hash_create(&l2_tn_hash_params);
1416 	if (!l2_tn_info->hash_handle) {
1417 		PMD_INIT_LOG(ERR, "Failed to create L2 TN hash table!");
1418 		return -EINVAL;
1419 	}
1420 	l2_tn_info->hash_map = rte_zmalloc("ixgbe",
1421 				   sizeof(struct ixgbe_l2_tn_filter *) *
1422 				   IXGBE_MAX_L2_TN_FILTER_NUM,
1423 				   0);
1424 	if (!l2_tn_info->hash_map) {
1425 		PMD_INIT_LOG(ERR,
1426 			"Failed to allocate memory for L2 TN hash map!");
1427 		rte_hash_free(l2_tn_info->hash_handle);
1428 		return -ENOMEM;
1429 	}
1430 	l2_tn_info->e_tag_en = FALSE;
1431 	l2_tn_info->e_tag_fwd_en = FALSE;
1432 	l2_tn_info->e_tag_ether_type = RTE_ETHER_TYPE_ETAG;
1433 
1434 	return 0;
1435 }
1436 /*
1437  * Negotiate mailbox API version with the PF.
1438  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1439  * Then we try to negotiate starting with the most recent one.
1440  * If all negotiation attempts fail, then we will proceed with
1441  * the default one (ixgbe_mbox_api_10).
1442  */
1443 static void
1444 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1445 {
1446 	int32_t i;
1447 
1448 	/* start with highest supported, proceed down */
1449 	static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1450 		ixgbe_mbox_api_13,
1451 		ixgbe_mbox_api_12,
1452 		ixgbe_mbox_api_11,
1453 		ixgbe_mbox_api_10,
1454 	};
1455 
1456 	for (i = 0;
1457 			i != RTE_DIM(sup_ver) &&
1458 			ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1459 			i++)
1460 		;
1461 }
1462 
1463 static void
1464 generate_random_mac_addr(struct rte_ether_addr *mac_addr)
1465 {
1466 	uint64_t random;
1467 
1468 	/* Set Organizationally Unique Identifier (OUI) prefix. */
1469 	mac_addr->addr_bytes[0] = 0x00;
1470 	mac_addr->addr_bytes[1] = 0x09;
1471 	mac_addr->addr_bytes[2] = 0xC0;
1472 	/* Force indication of locally assigned MAC address. */
1473 	mac_addr->addr_bytes[0] |= RTE_ETHER_LOCAL_ADMIN_ADDR;
1474 	/* Generate the last 3 bytes of the MAC address with a random number. */
1475 	random = rte_rand();
1476 	memcpy(&mac_addr->addr_bytes[3], &random, 3);
1477 }
1478 
1479 static int
1480 devarg_handle_int(__rte_unused const char *key, const char *value,
1481 		  void *extra_args)
1482 {
1483 	uint16_t *n = extra_args;
1484 
1485 	if (value == NULL || extra_args == NULL)
1486 		return -EINVAL;
1487 
1488 	*n = (uint16_t)strtoul(value, NULL, 0);
1489 	if (*n == USHRT_MAX && errno == ERANGE)
1490 		return -1;
1491 
1492 	return 0;
1493 }
1494 
1495 static void
1496 ixgbevf_parse_devargs(struct ixgbe_adapter *adapter,
1497 		      struct rte_devargs *devargs)
1498 {
1499 	struct rte_kvargs *kvlist;
1500 	uint16_t pflink_fullchk;
1501 
1502 	if (devargs == NULL)
1503 		return;
1504 
1505 	kvlist = rte_kvargs_parse(devargs->args, ixgbevf_valid_arguments);
1506 	if (kvlist == NULL)
1507 		return;
1508 
1509 	if (rte_kvargs_count(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK) == 1 &&
1510 	    rte_kvargs_process(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK,
1511 			       devarg_handle_int, &pflink_fullchk) == 0 &&
1512 	    pflink_fullchk == 1)
1513 		adapter->pflink_fullchk = 1;
1514 
1515 	rte_kvargs_free(kvlist);
1516 }
1517 
1518 /*
1519  * Virtual Function device init
1520  */
1521 static int
1522 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1523 {
1524 	int diag;
1525 	uint32_t tc, tcs;
1526 	struct ixgbe_adapter *ad = eth_dev->data->dev_private;
1527 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1528 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1529 	struct ixgbe_hw *hw =
1530 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1531 	struct ixgbe_vfta *shadow_vfta =
1532 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1533 	struct ixgbe_hwstrip *hwstrip =
1534 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1535 	struct rte_ether_addr *perm_addr =
1536 		(struct rte_ether_addr *)hw->mac.perm_addr;
1537 
1538 	PMD_INIT_FUNC_TRACE();
1539 
1540 	eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1541 	eth_dev->rx_descriptor_status = ixgbe_dev_rx_descriptor_status;
1542 	eth_dev->tx_descriptor_status = ixgbe_dev_tx_descriptor_status;
1543 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1544 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1545 
1546 	/* for secondary processes, we don't initialise any further as primary
1547 	 * has already done this work. Only check we don't need a different
1548 	 * RX function
1549 	 */
1550 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1551 		struct ixgbe_tx_queue *txq;
1552 		/* TX queue function in primary, set by last queue initialized
1553 		 * Tx queue may not initialized by primary process
1554 		 */
1555 		if (eth_dev->data->tx_queues) {
1556 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1557 			ixgbe_set_tx_function(eth_dev, txq);
1558 		} else {
1559 			/* Use default TX function if we get here */
1560 			PMD_INIT_LOG(NOTICE,
1561 				     "No TX queues configured yet. Using default TX function.");
1562 		}
1563 
1564 		ixgbe_set_rx_function(eth_dev);
1565 
1566 		return 0;
1567 	}
1568 
1569 	rte_atomic32_clear(&ad->link_thread_running);
1570 	ixgbevf_parse_devargs(eth_dev->data->dev_private,
1571 			      pci_dev->device.devargs);
1572 
1573 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1574 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1575 
1576 	hw->device_id = pci_dev->id.device_id;
1577 	hw->vendor_id = pci_dev->id.vendor_id;
1578 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1579 
1580 	/* initialize the vfta */
1581 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1582 
1583 	/* initialize the hw strip bitmap*/
1584 	memset(hwstrip, 0, sizeof(*hwstrip));
1585 
1586 	/* Initialize the shared code (base driver) */
1587 	diag = ixgbe_init_shared_code(hw);
1588 	if (diag != IXGBE_SUCCESS) {
1589 		PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1590 		return -EIO;
1591 	}
1592 
1593 	/* init_mailbox_params */
1594 	hw->mbx.ops.init_params(hw);
1595 
1596 	/* Reset the hw statistics */
1597 	ixgbevf_dev_stats_reset(eth_dev);
1598 
1599 	/* Disable the interrupts for VF */
1600 	ixgbevf_intr_disable(eth_dev);
1601 
1602 	hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1603 	diag = hw->mac.ops.reset_hw(hw);
1604 
1605 	/*
1606 	 * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1607 	 * the underlying PF driver has not assigned a MAC address to the VF.
1608 	 * In this case, assign a random MAC address.
1609 	 */
1610 	if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1611 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1612 		/*
1613 		 * This error code will be propagated to the app by
1614 		 * rte_eth_dev_reset, so use a public error code rather than
1615 		 * the internal-only IXGBE_ERR_RESET_FAILED
1616 		 */
1617 		return -EAGAIN;
1618 	}
1619 
1620 	/* negotiate mailbox API version to use with the PF. */
1621 	ixgbevf_negotiate_api(hw);
1622 
1623 	/* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1624 	ixgbevf_get_queues(hw, &tcs, &tc);
1625 
1626 	/* Allocate memory for storing MAC addresses */
1627 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", RTE_ETHER_ADDR_LEN *
1628 					       hw->mac.num_rar_entries, 0);
1629 	if (eth_dev->data->mac_addrs == NULL) {
1630 		PMD_INIT_LOG(ERR,
1631 			     "Failed to allocate %u bytes needed to store "
1632 			     "MAC addresses",
1633 			     RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1634 		return -ENOMEM;
1635 	}
1636 
1637 	/* Generate a random MAC address, if none was assigned by PF. */
1638 	if (rte_is_zero_ether_addr(perm_addr)) {
1639 		generate_random_mac_addr(perm_addr);
1640 		diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1641 		if (diag) {
1642 			rte_free(eth_dev->data->mac_addrs);
1643 			eth_dev->data->mac_addrs = NULL;
1644 			return diag;
1645 		}
1646 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1647 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1648 			     RTE_ETHER_ADDR_PRT_FMT,
1649 				 RTE_ETHER_ADDR_BYTES(perm_addr));
1650 	}
1651 
1652 	/* Copy the permanent MAC address */
1653 	rte_ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1654 
1655 	/* reset the hardware with the new settings */
1656 	diag = hw->mac.ops.start_hw(hw);
1657 	switch (diag) {
1658 	case  0:
1659 		break;
1660 
1661 	default:
1662 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1663 		rte_free(eth_dev->data->mac_addrs);
1664 		eth_dev->data->mac_addrs = NULL;
1665 		return -EIO;
1666 	}
1667 
1668 	rte_intr_callback_register(intr_handle,
1669 				   ixgbevf_dev_interrupt_handler, eth_dev);
1670 	rte_intr_enable(intr_handle);
1671 	ixgbevf_intr_enable(eth_dev);
1672 
1673 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1674 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1675 		     pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1676 
1677 	return 0;
1678 }
1679 
1680 /* Virtual Function device uninit */
1681 
1682 static int
1683 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1684 {
1685 	PMD_INIT_FUNC_TRACE();
1686 
1687 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1688 		return 0;
1689 
1690 	ixgbevf_dev_close(eth_dev);
1691 
1692 	return 0;
1693 }
1694 
1695 static int
1696 eth_ixgbe_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1697 		struct rte_pci_device *pci_dev)
1698 {
1699 	char name[RTE_ETH_NAME_MAX_LEN];
1700 	struct rte_eth_dev *pf_ethdev;
1701 	struct rte_eth_devargs eth_da;
1702 	int i, retval;
1703 
1704 	if (pci_dev->device.devargs) {
1705 		retval = rte_eth_devargs_parse(pci_dev->device.devargs->args,
1706 				&eth_da);
1707 		if (retval)
1708 			return retval;
1709 	} else
1710 		memset(&eth_da, 0, sizeof(eth_da));
1711 
1712 	if (eth_da.nb_representor_ports > 0 &&
1713 	    eth_da.type != RTE_ETH_REPRESENTOR_VF) {
1714 		PMD_DRV_LOG(ERR, "unsupported representor type: %s\n",
1715 			    pci_dev->device.devargs->args);
1716 		return -ENOTSUP;
1717 	}
1718 
1719 	retval = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
1720 		sizeof(struct ixgbe_adapter),
1721 		eth_dev_pci_specific_init, pci_dev,
1722 		eth_ixgbe_dev_init, NULL);
1723 
1724 	if (retval || eth_da.nb_representor_ports < 1)
1725 		return retval;
1726 
1727 	pf_ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1728 	if (pf_ethdev == NULL)
1729 		return -ENODEV;
1730 
1731 	/* probe VF representor ports */
1732 	for (i = 0; i < eth_da.nb_representor_ports; i++) {
1733 		struct ixgbe_vf_info *vfinfo;
1734 		struct ixgbe_vf_representor representor;
1735 
1736 		vfinfo = *IXGBE_DEV_PRIVATE_TO_P_VFDATA(
1737 			pf_ethdev->data->dev_private);
1738 		if (vfinfo == NULL) {
1739 			PMD_DRV_LOG(ERR,
1740 				"no virtual functions supported by PF");
1741 			break;
1742 		}
1743 
1744 		representor.vf_id = eth_da.representor_ports[i];
1745 		representor.switch_domain_id = vfinfo->switch_domain_id;
1746 		representor.pf_ethdev = pf_ethdev;
1747 
1748 		/* representor port net_bdf_port */
1749 		snprintf(name, sizeof(name), "net_%s_representor_%d",
1750 			pci_dev->device.name,
1751 			eth_da.representor_ports[i]);
1752 
1753 		retval = rte_eth_dev_create(&pci_dev->device, name,
1754 			sizeof(struct ixgbe_vf_representor), NULL, NULL,
1755 			ixgbe_vf_representor_init, &representor);
1756 
1757 		if (retval)
1758 			PMD_DRV_LOG(ERR, "failed to create ixgbe vf "
1759 				"representor %s.", name);
1760 	}
1761 
1762 	return 0;
1763 }
1764 
1765 static int eth_ixgbe_pci_remove(struct rte_pci_device *pci_dev)
1766 {
1767 	struct rte_eth_dev *ethdev;
1768 
1769 	ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1770 	if (!ethdev)
1771 		return 0;
1772 
1773 	if (ethdev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
1774 		return rte_eth_dev_pci_generic_remove(pci_dev,
1775 					ixgbe_vf_representor_uninit);
1776 	else
1777 		return rte_eth_dev_pci_generic_remove(pci_dev,
1778 						eth_ixgbe_dev_uninit);
1779 }
1780 
1781 static struct rte_pci_driver rte_ixgbe_pmd = {
1782 	.id_table = pci_id_ixgbe_map,
1783 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1784 	.probe = eth_ixgbe_pci_probe,
1785 	.remove = eth_ixgbe_pci_remove,
1786 };
1787 
1788 static int eth_ixgbevf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1789 	struct rte_pci_device *pci_dev)
1790 {
1791 	return rte_eth_dev_pci_generic_probe(pci_dev,
1792 		sizeof(struct ixgbe_adapter), eth_ixgbevf_dev_init);
1793 }
1794 
1795 static int eth_ixgbevf_pci_remove(struct rte_pci_device *pci_dev)
1796 {
1797 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbevf_dev_uninit);
1798 }
1799 
1800 /*
1801  * virtual function driver struct
1802  */
1803 static struct rte_pci_driver rte_ixgbevf_pmd = {
1804 	.id_table = pci_id_ixgbevf_map,
1805 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1806 	.probe = eth_ixgbevf_pci_probe,
1807 	.remove = eth_ixgbevf_pci_remove,
1808 };
1809 
1810 static int
1811 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1812 {
1813 	struct ixgbe_hw *hw =
1814 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1815 	struct ixgbe_vfta *shadow_vfta =
1816 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1817 	uint32_t vfta;
1818 	uint32_t vid_idx;
1819 	uint32_t vid_bit;
1820 
1821 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1822 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1823 	vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1824 	if (on)
1825 		vfta |= vid_bit;
1826 	else
1827 		vfta &= ~vid_bit;
1828 	IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1829 
1830 	/* update local VFTA copy */
1831 	shadow_vfta->vfta[vid_idx] = vfta;
1832 
1833 	return 0;
1834 }
1835 
1836 static void
1837 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1838 {
1839 	if (on)
1840 		ixgbe_vlan_hw_strip_enable(dev, queue);
1841 	else
1842 		ixgbe_vlan_hw_strip_disable(dev, queue);
1843 }
1844 
1845 static int
1846 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1847 		    enum rte_vlan_type vlan_type,
1848 		    uint16_t tpid)
1849 {
1850 	struct ixgbe_hw *hw =
1851 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1852 	int ret = 0;
1853 	uint32_t reg;
1854 	uint32_t qinq;
1855 
1856 	qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1857 	qinq &= IXGBE_DMATXCTL_GDV;
1858 
1859 	switch (vlan_type) {
1860 	case RTE_ETH_VLAN_TYPE_INNER:
1861 		if (qinq) {
1862 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1863 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1864 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1865 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1866 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1867 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1868 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1869 		} else {
1870 			ret = -ENOTSUP;
1871 			PMD_DRV_LOG(ERR, "Inner type is not supported"
1872 				    " by single VLAN");
1873 		}
1874 		break;
1875 	case RTE_ETH_VLAN_TYPE_OUTER:
1876 		if (qinq) {
1877 			/* Only the high 16-bits is valid */
1878 			IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1879 					IXGBE_EXVET_VET_EXT_SHIFT);
1880 		} else {
1881 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1882 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1883 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1884 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1885 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1886 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1887 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1888 		}
1889 
1890 		break;
1891 	default:
1892 		ret = -EINVAL;
1893 		PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1894 		break;
1895 	}
1896 
1897 	return ret;
1898 }
1899 
1900 void
1901 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1902 {
1903 	struct ixgbe_hw *hw =
1904 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1905 	uint32_t vlnctrl;
1906 
1907 	PMD_INIT_FUNC_TRACE();
1908 
1909 	/* Filter Table Disable */
1910 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1911 	vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1912 
1913 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1914 }
1915 
1916 void
1917 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1918 {
1919 	struct ixgbe_hw *hw =
1920 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1921 	struct ixgbe_vfta *shadow_vfta =
1922 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1923 	uint32_t vlnctrl;
1924 	uint16_t i;
1925 
1926 	PMD_INIT_FUNC_TRACE();
1927 
1928 	/* Filter Table Enable */
1929 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1930 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1931 	vlnctrl |= IXGBE_VLNCTRL_VFE;
1932 
1933 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1934 
1935 	/* write whatever is in local vfta copy */
1936 	for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1937 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1938 }
1939 
1940 static void
1941 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1942 {
1943 	struct ixgbe_hwstrip *hwstrip =
1944 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1945 	struct ixgbe_rx_queue *rxq;
1946 
1947 	if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
1948 		return;
1949 
1950 	if (on)
1951 		IXGBE_SET_HWSTRIP(hwstrip, queue);
1952 	else
1953 		IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1954 
1955 	if (queue >= dev->data->nb_rx_queues)
1956 		return;
1957 
1958 	rxq = dev->data->rx_queues[queue];
1959 
1960 	if (on) {
1961 		rxq->vlan_flags = RTE_MBUF_F_RX_VLAN | RTE_MBUF_F_RX_VLAN_STRIPPED;
1962 		rxq->offloads |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
1963 	} else {
1964 		rxq->vlan_flags = RTE_MBUF_F_RX_VLAN;
1965 		rxq->offloads &= ~RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
1966 	}
1967 }
1968 
1969 static void
1970 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
1971 {
1972 	struct ixgbe_hw *hw =
1973 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1974 	uint32_t ctrl;
1975 
1976 	PMD_INIT_FUNC_TRACE();
1977 
1978 	if (hw->mac.type == ixgbe_mac_82598EB) {
1979 		/* No queue level support */
1980 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1981 		return;
1982 	}
1983 
1984 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1985 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1986 	ctrl &= ~IXGBE_RXDCTL_VME;
1987 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1988 
1989 	/* record those setting for HW strip per queue */
1990 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
1991 }
1992 
1993 static void
1994 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
1995 {
1996 	struct ixgbe_hw *hw =
1997 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1998 	uint32_t ctrl;
1999 
2000 	PMD_INIT_FUNC_TRACE();
2001 
2002 	if (hw->mac.type == ixgbe_mac_82598EB) {
2003 		/* No queue level supported */
2004 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2005 		return;
2006 	}
2007 
2008 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2009 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2010 	ctrl |= IXGBE_RXDCTL_VME;
2011 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2012 
2013 	/* record those setting for HW strip per queue */
2014 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
2015 }
2016 
2017 static void
2018 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2019 {
2020 	struct ixgbe_hw *hw =
2021 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2022 	uint32_t ctrl;
2023 
2024 	PMD_INIT_FUNC_TRACE();
2025 
2026 	/* DMATXCTRL: Geric Double VLAN Disable */
2027 	ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2028 	ctrl &= ~IXGBE_DMATXCTL_GDV;
2029 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2030 
2031 	/* CTRL_EXT: Global Double VLAN Disable */
2032 	ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2033 	ctrl &= ~IXGBE_EXTENDED_VLAN;
2034 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2035 
2036 }
2037 
2038 static void
2039 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2040 {
2041 	struct ixgbe_hw *hw =
2042 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2043 	uint32_t ctrl;
2044 
2045 	PMD_INIT_FUNC_TRACE();
2046 
2047 	/* DMATXCTRL: Geric Double VLAN Enable */
2048 	ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2049 	ctrl |= IXGBE_DMATXCTL_GDV;
2050 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2051 
2052 	/* CTRL_EXT: Global Double VLAN Enable */
2053 	ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2054 	ctrl |= IXGBE_EXTENDED_VLAN;
2055 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2056 
2057 	/* Clear pooling mode of PFVTCTL. It's required by X550. */
2058 	if (hw->mac.type == ixgbe_mac_X550 ||
2059 	    hw->mac.type == ixgbe_mac_X550EM_x ||
2060 	    hw->mac.type == ixgbe_mac_X550EM_a) {
2061 		ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
2062 		ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
2063 		IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
2064 	}
2065 
2066 	/*
2067 	 * VET EXT field in the EXVET register = 0x8100 by default
2068 	 * So no need to change. Same to VT field of DMATXCTL register
2069 	 */
2070 }
2071 
2072 void
2073 ixgbe_vlan_hw_strip_config(struct rte_eth_dev *dev)
2074 {
2075 	struct ixgbe_hw *hw =
2076 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2077 	struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
2078 	uint32_t ctrl;
2079 	uint16_t i;
2080 	struct ixgbe_rx_queue *rxq;
2081 	bool on;
2082 
2083 	PMD_INIT_FUNC_TRACE();
2084 
2085 	if (hw->mac.type == ixgbe_mac_82598EB) {
2086 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP) {
2087 			ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2088 			ctrl |= IXGBE_VLNCTRL_VME;
2089 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2090 		} else {
2091 			ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2092 			ctrl &= ~IXGBE_VLNCTRL_VME;
2093 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2094 		}
2095 	} else {
2096 		/*
2097 		 * Other 10G NIC, the VLAN strip can be setup
2098 		 * per queue in RXDCTL
2099 		 */
2100 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
2101 			rxq = dev->data->rx_queues[i];
2102 			ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2103 			if (rxq->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP) {
2104 				ctrl |= IXGBE_RXDCTL_VME;
2105 				on = TRUE;
2106 			} else {
2107 				ctrl &= ~IXGBE_RXDCTL_VME;
2108 				on = FALSE;
2109 			}
2110 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2111 
2112 			/* record those setting for HW strip per queue */
2113 			ixgbe_vlan_hw_strip_bitmap_set(dev, i, on);
2114 		}
2115 	}
2116 }
2117 
2118 static void
2119 ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev, int mask)
2120 {
2121 	uint16_t i;
2122 	struct rte_eth_rxmode *rxmode;
2123 	struct ixgbe_rx_queue *rxq;
2124 
2125 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
2126 		rxmode = &dev->data->dev_conf.rxmode;
2127 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
2128 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
2129 				rxq = dev->data->rx_queues[i];
2130 				rxq->offloads |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
2131 			}
2132 		else
2133 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
2134 				rxq = dev->data->rx_queues[i];
2135 				rxq->offloads &= ~RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
2136 			}
2137 	}
2138 }
2139 
2140 static int
2141 ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask)
2142 {
2143 	struct rte_eth_rxmode *rxmode;
2144 	rxmode = &dev->data->dev_conf.rxmode;
2145 
2146 	if (mask & RTE_ETH_VLAN_STRIP_MASK)
2147 		ixgbe_vlan_hw_strip_config(dev);
2148 
2149 	if (mask & RTE_ETH_VLAN_FILTER_MASK) {
2150 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
2151 			ixgbe_vlan_hw_filter_enable(dev);
2152 		else
2153 			ixgbe_vlan_hw_filter_disable(dev);
2154 	}
2155 
2156 	if (mask & RTE_ETH_VLAN_EXTEND_MASK) {
2157 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND)
2158 			ixgbe_vlan_hw_extend_enable(dev);
2159 		else
2160 			ixgbe_vlan_hw_extend_disable(dev);
2161 	}
2162 
2163 	return 0;
2164 }
2165 
2166 static int
2167 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2168 {
2169 	ixgbe_config_vlan_strip_on_all_queues(dev, mask);
2170 
2171 	ixgbe_vlan_offload_config(dev, mask);
2172 
2173 	return 0;
2174 }
2175 
2176 static void
2177 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2178 {
2179 	struct ixgbe_hw *hw =
2180 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2181 	/* VLNCTRL: enable vlan filtering and allow all vlan tags through */
2182 	uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2183 
2184 	vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
2185 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
2186 }
2187 
2188 static int
2189 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
2190 {
2191 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2192 
2193 	switch (nb_rx_q) {
2194 	case 1:
2195 	case 2:
2196 		RTE_ETH_DEV_SRIOV(dev).active = RTE_ETH_64_POOLS;
2197 		break;
2198 	case 4:
2199 		RTE_ETH_DEV_SRIOV(dev).active = RTE_ETH_32_POOLS;
2200 		break;
2201 	default:
2202 		return -EINVAL;
2203 	}
2204 
2205 	RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
2206 		IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2207 	RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
2208 		pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2209 	return 0;
2210 }
2211 
2212 static int
2213 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
2214 {
2215 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
2216 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2217 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
2218 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
2219 
2220 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
2221 		/* check multi-queue mode */
2222 		switch (dev_conf->rxmode.mq_mode) {
2223 		case RTE_ETH_MQ_RX_VMDQ_DCB:
2224 			PMD_INIT_LOG(INFO, "RTE_ETH_MQ_RX_VMDQ_DCB mode supported in SRIOV");
2225 			break;
2226 		case RTE_ETH_MQ_RX_VMDQ_DCB_RSS:
2227 			/* DCB/RSS VMDQ in SRIOV mode, not implement yet */
2228 			PMD_INIT_LOG(ERR, "SRIOV active,"
2229 					" unsupported mq_mode rx %d.",
2230 					dev_conf->rxmode.mq_mode);
2231 			return -EINVAL;
2232 		case RTE_ETH_MQ_RX_RSS:
2233 		case RTE_ETH_MQ_RX_VMDQ_RSS:
2234 			dev->data->dev_conf.rxmode.mq_mode = RTE_ETH_MQ_RX_VMDQ_RSS;
2235 			if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
2236 				if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
2237 					PMD_INIT_LOG(ERR, "SRIOV is active,"
2238 						" invalid queue number"
2239 						" for VMDQ RSS, allowed"
2240 						" value are 1, 2 or 4.");
2241 					return -EINVAL;
2242 				}
2243 			break;
2244 		case RTE_ETH_MQ_RX_VMDQ_ONLY:
2245 		case RTE_ETH_MQ_RX_NONE:
2246 			/* if nothing mq mode configure, use default scheme */
2247 			dev->data->dev_conf.rxmode.mq_mode = RTE_ETH_MQ_RX_VMDQ_ONLY;
2248 			break;
2249 		default: /* RTE_ETH_MQ_RX_DCB, RTE_ETH_MQ_RX_DCB_RSS or RTE_ETH_MQ_TX_DCB*/
2250 			/* SRIOV only works in VMDq enable mode */
2251 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2252 					" wrong mq_mode rx %d.",
2253 					dev_conf->rxmode.mq_mode);
2254 			return -EINVAL;
2255 		}
2256 
2257 		switch (dev_conf->txmode.mq_mode) {
2258 		case RTE_ETH_MQ_TX_VMDQ_DCB:
2259 			PMD_INIT_LOG(INFO, "RTE_ETH_MQ_TX_VMDQ_DCB mode supported in SRIOV");
2260 			dev->data->dev_conf.txmode.mq_mode = RTE_ETH_MQ_TX_VMDQ_DCB;
2261 			break;
2262 		default: /* RTE_ETH_MQ_TX_VMDQ_ONLY or RTE_ETH_MQ_TX_NONE */
2263 			dev->data->dev_conf.txmode.mq_mode = RTE_ETH_MQ_TX_VMDQ_ONLY;
2264 			break;
2265 		}
2266 
2267 		/* check valid queue number */
2268 		if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2269 		    (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2270 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2271 					" nb_rx_q=%d nb_tx_q=%d queue number"
2272 					" must be less than or equal to %d.",
2273 					nb_rx_q, nb_tx_q,
2274 					RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2275 			return -EINVAL;
2276 		}
2277 	} else {
2278 		if (dev_conf->rxmode.mq_mode == RTE_ETH_MQ_RX_VMDQ_DCB_RSS) {
2279 			PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2280 					  " not supported.");
2281 			return -EINVAL;
2282 		}
2283 		/* check configuration for vmdb+dcb mode */
2284 		if (dev_conf->rxmode.mq_mode == RTE_ETH_MQ_RX_VMDQ_DCB) {
2285 			const struct rte_eth_vmdq_dcb_conf *conf;
2286 
2287 			if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2288 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2289 						IXGBE_VMDQ_DCB_NB_QUEUES);
2290 				return -EINVAL;
2291 			}
2292 			conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2293 			if (!(conf->nb_queue_pools == RTE_ETH_16_POOLS ||
2294 			       conf->nb_queue_pools == RTE_ETH_32_POOLS)) {
2295 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2296 						" nb_queue_pools must be %d or %d.",
2297 						RTE_ETH_16_POOLS, RTE_ETH_32_POOLS);
2298 				return -EINVAL;
2299 			}
2300 		}
2301 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_VMDQ_DCB) {
2302 			const struct rte_eth_vmdq_dcb_tx_conf *conf;
2303 
2304 			if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2305 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2306 						 IXGBE_VMDQ_DCB_NB_QUEUES);
2307 				return -EINVAL;
2308 			}
2309 			conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2310 			if (!(conf->nb_queue_pools == RTE_ETH_16_POOLS ||
2311 			       conf->nb_queue_pools == RTE_ETH_32_POOLS)) {
2312 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2313 						" nb_queue_pools != %d and"
2314 						" nb_queue_pools != %d.",
2315 						RTE_ETH_16_POOLS, RTE_ETH_32_POOLS);
2316 				return -EINVAL;
2317 			}
2318 		}
2319 
2320 		/* For DCB mode check our configuration before we go further */
2321 		if (dev_conf->rxmode.mq_mode == RTE_ETH_MQ_RX_DCB) {
2322 			const struct rte_eth_dcb_rx_conf *conf;
2323 
2324 			conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2325 			if (!(conf->nb_tcs == RTE_ETH_4_TCS ||
2326 			       conf->nb_tcs == RTE_ETH_8_TCS)) {
2327 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2328 						" and nb_tcs != %d.",
2329 						RTE_ETH_4_TCS, RTE_ETH_8_TCS);
2330 				return -EINVAL;
2331 			}
2332 		}
2333 
2334 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_DCB) {
2335 			const struct rte_eth_dcb_tx_conf *conf;
2336 
2337 			conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2338 			if (!(conf->nb_tcs == RTE_ETH_4_TCS ||
2339 			       conf->nb_tcs == RTE_ETH_8_TCS)) {
2340 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2341 						" and nb_tcs != %d.",
2342 						RTE_ETH_4_TCS, RTE_ETH_8_TCS);
2343 				return -EINVAL;
2344 			}
2345 		}
2346 
2347 		/*
2348 		 * When DCB/VT is off, maximum number of queues changes,
2349 		 * except for 82598EB, which remains constant.
2350 		 */
2351 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_NONE &&
2352 				hw->mac.type != ixgbe_mac_82598EB) {
2353 			if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2354 				PMD_INIT_LOG(ERR,
2355 					     "Neither VT nor DCB are enabled, "
2356 					     "nb_tx_q > %d.",
2357 					     IXGBE_NONE_MODE_TX_NB_QUEUES);
2358 				return -EINVAL;
2359 			}
2360 		}
2361 	}
2362 	return 0;
2363 }
2364 
2365 static int
2366 ixgbe_dev_configure(struct rte_eth_dev *dev)
2367 {
2368 	struct ixgbe_interrupt *intr =
2369 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2370 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2371 	int ret;
2372 
2373 	PMD_INIT_FUNC_TRACE();
2374 
2375 	if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
2376 		dev->data->dev_conf.rxmode.offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
2377 
2378 	/* multipe queue mode checking */
2379 	ret  = ixgbe_check_mq_mode(dev);
2380 	if (ret != 0) {
2381 		PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2382 			    ret);
2383 		return ret;
2384 	}
2385 
2386 	/* set flag to update link status after init */
2387 	intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2388 
2389 	/*
2390 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2391 	 * allocation or vector Rx preconditions we will reset it.
2392 	 */
2393 	adapter->rx_bulk_alloc_allowed = true;
2394 	adapter->rx_vec_allowed = true;
2395 
2396 	return 0;
2397 }
2398 
2399 static void
2400 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2401 {
2402 	struct ixgbe_hw *hw =
2403 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2404 	struct ixgbe_interrupt *intr =
2405 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2406 	uint32_t gpie;
2407 
2408 	/* only set up it on X550EM_X */
2409 	if (hw->mac.type == ixgbe_mac_X550EM_x) {
2410 		gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2411 		gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2412 		IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2413 		if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2414 			intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2415 	}
2416 }
2417 
2418 int
2419 ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
2420 			uint16_t tx_rate, uint64_t q_msk)
2421 {
2422 	struct ixgbe_hw *hw;
2423 	struct ixgbe_vf_info *vfinfo;
2424 	struct rte_eth_link link;
2425 	uint8_t  nb_q_per_pool;
2426 	uint32_t queue_stride;
2427 	uint32_t queue_idx, idx = 0, vf_idx;
2428 	uint32_t queue_end;
2429 	uint16_t total_rate = 0;
2430 	struct rte_pci_device *pci_dev;
2431 	int ret;
2432 
2433 	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2434 	ret = rte_eth_link_get_nowait(dev->data->port_id, &link);
2435 	if (ret < 0)
2436 		return ret;
2437 
2438 	if (vf >= pci_dev->max_vfs)
2439 		return -EINVAL;
2440 
2441 	if (tx_rate > link.link_speed)
2442 		return -EINVAL;
2443 
2444 	if (q_msk == 0)
2445 		return 0;
2446 
2447 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2448 	vfinfo = *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
2449 	nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2450 	queue_stride = IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2451 	queue_idx = vf * queue_stride;
2452 	queue_end = queue_idx + nb_q_per_pool - 1;
2453 	if (queue_end >= hw->mac.max_tx_queues)
2454 		return -EINVAL;
2455 
2456 	if (vfinfo) {
2457 		for (vf_idx = 0; vf_idx < pci_dev->max_vfs; vf_idx++) {
2458 			if (vf_idx == vf)
2459 				continue;
2460 			for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
2461 				idx++)
2462 				total_rate += vfinfo[vf_idx].tx_rate[idx];
2463 		}
2464 	} else {
2465 		return -EINVAL;
2466 	}
2467 
2468 	/* Store tx_rate for this vf. */
2469 	for (idx = 0; idx < nb_q_per_pool; idx++) {
2470 		if (((uint64_t)0x1 << idx) & q_msk) {
2471 			if (vfinfo[vf].tx_rate[idx] != tx_rate)
2472 				vfinfo[vf].tx_rate[idx] = tx_rate;
2473 			total_rate += tx_rate;
2474 		}
2475 	}
2476 
2477 	if (total_rate > dev->data->dev_link.link_speed) {
2478 		/* Reset stored TX rate of the VF if it causes exceed
2479 		 * link speed.
2480 		 */
2481 		memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
2482 		return -EINVAL;
2483 	}
2484 
2485 	/* Set RTTBCNRC of each queue/pool for vf X  */
2486 	for (; queue_idx <= queue_end; queue_idx++) {
2487 		if (0x1 & q_msk)
2488 			ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
2489 		q_msk = q_msk >> 1;
2490 	}
2491 
2492 	return 0;
2493 }
2494 
2495 static int
2496 ixgbe_flow_ctrl_enable(struct rte_eth_dev *dev, struct ixgbe_hw *hw)
2497 {
2498 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2499 	int err;
2500 	uint32_t mflcn;
2501 
2502 	ixgbe_setup_fc(hw);
2503 
2504 	err = ixgbe_fc_enable(hw);
2505 
2506 	/* Not negotiated is not an error case */
2507 	if (err == IXGBE_SUCCESS || err == IXGBE_ERR_FC_NOT_NEGOTIATED) {
2508 		/*
2509 		 *check if we want to forward MAC frames - driver doesn't
2510 		 *have native capability to do that,
2511 		 *so we'll write the registers ourselves
2512 		 */
2513 
2514 		mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
2515 
2516 		/* set or clear MFLCN.PMCF bit depending on configuration */
2517 		if (adapter->mac_ctrl_frame_fwd != 0)
2518 			mflcn |= IXGBE_MFLCN_PMCF;
2519 		else
2520 			mflcn &= ~IXGBE_MFLCN_PMCF;
2521 
2522 		IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
2523 		IXGBE_WRITE_FLUSH(hw);
2524 
2525 		return 0;
2526 	}
2527 	return err;
2528 }
2529 
2530 /*
2531  * Configure device link speed and setup link.
2532  * It returns 0 on success.
2533  */
2534 static int
2535 ixgbe_dev_start(struct rte_eth_dev *dev)
2536 {
2537 	struct ixgbe_hw *hw =
2538 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2539 	struct ixgbe_vf_info *vfinfo =
2540 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2541 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2542 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2543 	uint32_t intr_vector = 0;
2544 	int err;
2545 	bool link_up = false, negotiate = 0;
2546 	uint32_t speed = 0;
2547 	uint32_t allowed_speeds = 0;
2548 	int mask = 0;
2549 	int status;
2550 	uint16_t vf, idx;
2551 	uint32_t *link_speeds;
2552 	struct ixgbe_tm_conf *tm_conf =
2553 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2554 	struct ixgbe_macsec_setting *macsec_setting =
2555 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
2556 
2557 	PMD_INIT_FUNC_TRACE();
2558 
2559 	/* Stop the link setup handler before resetting the HW. */
2560 	ixgbe_dev_wait_setup_link_complete(dev, 0);
2561 
2562 	/* disable uio/vfio intr/eventfd mapping */
2563 	rte_intr_disable(intr_handle);
2564 
2565 	/* stop adapter */
2566 	hw->adapter_stopped = 0;
2567 	ixgbe_stop_adapter(hw);
2568 
2569 	/* reinitialize adapter
2570 	 * this calls reset and start
2571 	 */
2572 	status = ixgbe_pf_reset_hw(hw);
2573 	if (status != 0)
2574 		return -1;
2575 	hw->mac.ops.start_hw(hw);
2576 	hw->mac.get_link_status = true;
2577 
2578 	/* configure PF module if SRIOV enabled */
2579 	ixgbe_pf_host_configure(dev);
2580 
2581 	ixgbe_dev_phy_intr_setup(dev);
2582 
2583 	/* check and configure queue intr-vector mapping */
2584 	if ((rte_intr_cap_multiple(intr_handle) ||
2585 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
2586 	    dev->data->dev_conf.intr_conf.rxq != 0) {
2587 		intr_vector = dev->data->nb_rx_queues;
2588 		if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2589 			PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2590 					IXGBE_MAX_INTR_QUEUE_NUM);
2591 			return -ENOTSUP;
2592 		}
2593 		if (rte_intr_efd_enable(intr_handle, intr_vector))
2594 			return -1;
2595 	}
2596 
2597 	if (rte_intr_dp_is_en(intr_handle)) {
2598 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
2599 						   dev->data->nb_rx_queues)) {
2600 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2601 				     " intr_vec", dev->data->nb_rx_queues);
2602 			return -ENOMEM;
2603 		}
2604 	}
2605 
2606 	/* confiugre msix for sleep until rx interrupt */
2607 	ixgbe_configure_msix(dev);
2608 
2609 	/* initialize transmission unit */
2610 	ixgbe_dev_tx_init(dev);
2611 
2612 	/* This can fail when allocating mbufs for descriptor rings */
2613 	err = ixgbe_dev_rx_init(dev);
2614 	if (err) {
2615 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2616 		goto error;
2617 	}
2618 
2619 	mask = RTE_ETH_VLAN_STRIP_MASK | RTE_ETH_VLAN_FILTER_MASK |
2620 		RTE_ETH_VLAN_EXTEND_MASK;
2621 	err = ixgbe_vlan_offload_config(dev, mask);
2622 	if (err) {
2623 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload");
2624 		goto error;
2625 	}
2626 
2627 	if (dev->data->dev_conf.rxmode.mq_mode == RTE_ETH_MQ_RX_VMDQ_ONLY) {
2628 		/* Enable vlan filtering for VMDq */
2629 		ixgbe_vmdq_vlan_hw_filter_enable(dev);
2630 	}
2631 
2632 	/* Configure DCB hw */
2633 	ixgbe_configure_dcb(dev);
2634 
2635 	if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2636 		err = ixgbe_fdir_configure(dev);
2637 		if (err)
2638 			goto error;
2639 	}
2640 
2641 	/* Restore vf rate limit */
2642 	if (vfinfo != NULL) {
2643 		for (vf = 0; vf < pci_dev->max_vfs; vf++)
2644 			for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2645 				if (vfinfo[vf].tx_rate[idx] != 0)
2646 					ixgbe_set_vf_rate_limit(
2647 						dev, vf,
2648 						vfinfo[vf].tx_rate[idx],
2649 						1 << idx);
2650 	}
2651 
2652 	ixgbe_restore_statistics_mapping(dev);
2653 
2654 	err = ixgbe_flow_ctrl_enable(dev, hw);
2655 	if (err < 0) {
2656 		PMD_INIT_LOG(ERR, "enable flow ctrl err");
2657 		goto error;
2658 	}
2659 
2660 	err = ixgbe_dev_rxtx_start(dev);
2661 	if (err < 0) {
2662 		PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2663 		goto error;
2664 	}
2665 
2666 	/* Skip link setup if loopback mode is enabled. */
2667 	if (dev->data->dev_conf.lpbk_mode != 0) {
2668 		err = ixgbe_check_supported_loopback_mode(dev);
2669 		if (err < 0) {
2670 			PMD_INIT_LOG(ERR, "Unsupported loopback mode");
2671 			goto error;
2672 		} else {
2673 			goto skip_link_setup;
2674 		}
2675 	}
2676 
2677 	if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2678 		err = hw->mac.ops.setup_sfp(hw);
2679 		if (err)
2680 			goto error;
2681 	}
2682 
2683 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2684 		/* Turn on the copper */
2685 		ixgbe_set_phy_power(hw, true);
2686 	} else {
2687 		/* Turn on the laser */
2688 		ixgbe_enable_tx_laser(hw);
2689 	}
2690 
2691 	err = ixgbe_check_link(hw, &speed, &link_up, 0);
2692 	if (err)
2693 		goto error;
2694 	dev->data->dev_link.link_status = link_up;
2695 
2696 	err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2697 	if (err)
2698 		goto error;
2699 
2700 	switch (hw->mac.type) {
2701 	case ixgbe_mac_X550:
2702 	case ixgbe_mac_X550EM_x:
2703 	case ixgbe_mac_X550EM_a:
2704 		allowed_speeds = RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G |
2705 			RTE_ETH_LINK_SPEED_2_5G |  RTE_ETH_LINK_SPEED_5G |
2706 			RTE_ETH_LINK_SPEED_10G;
2707 		if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
2708 				hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
2709 			allowed_speeds = RTE_ETH_LINK_SPEED_10M |
2710 				RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G;
2711 		break;
2712 	default:
2713 		allowed_speeds = RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G |
2714 			RTE_ETH_LINK_SPEED_10G;
2715 	}
2716 
2717 	link_speeds = &dev->data->dev_conf.link_speeds;
2718 
2719 	/* Ignore autoneg flag bit and check the validity of 
2720 	 * link_speed 
2721 	 */
2722 	if (((*link_speeds) >> 1) & ~(allowed_speeds >> 1)) {
2723 		PMD_INIT_LOG(ERR, "Invalid link setting");
2724 		goto error;
2725 	}
2726 
2727 	speed = 0x0;
2728 	if (*link_speeds == RTE_ETH_LINK_SPEED_AUTONEG) {
2729 		switch (hw->mac.type) {
2730 		case ixgbe_mac_82598EB:
2731 			speed = IXGBE_LINK_SPEED_82598_AUTONEG;
2732 			break;
2733 		case ixgbe_mac_82599EB:
2734 		case ixgbe_mac_X540:
2735 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2736 			break;
2737 		case ixgbe_mac_X550:
2738 		case ixgbe_mac_X550EM_x:
2739 		case ixgbe_mac_X550EM_a:
2740 			speed = IXGBE_LINK_SPEED_X550_AUTONEG;
2741 			break;
2742 		default:
2743 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2744 		}
2745 	} else {
2746 		if (*link_speeds & RTE_ETH_LINK_SPEED_10G)
2747 			speed |= IXGBE_LINK_SPEED_10GB_FULL;
2748 		if (*link_speeds & RTE_ETH_LINK_SPEED_5G)
2749 			speed |= IXGBE_LINK_SPEED_5GB_FULL;
2750 		if (*link_speeds & RTE_ETH_LINK_SPEED_2_5G)
2751 			speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
2752 		if (*link_speeds & RTE_ETH_LINK_SPEED_1G)
2753 			speed |= IXGBE_LINK_SPEED_1GB_FULL;
2754 		if (*link_speeds & RTE_ETH_LINK_SPEED_100M)
2755 			speed |= IXGBE_LINK_SPEED_100_FULL;
2756 		if (*link_speeds & RTE_ETH_LINK_SPEED_10M)
2757 			speed |= IXGBE_LINK_SPEED_10_FULL;
2758 	}
2759 
2760 	err = ixgbe_setup_link(hw, speed, link_up);
2761 	if (err)
2762 		goto error;
2763 
2764 skip_link_setup:
2765 
2766 	if (rte_intr_allow_others(intr_handle)) {
2767 		/* check if lsc interrupt is enabled */
2768 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2769 			ixgbe_dev_lsc_interrupt_setup(dev, TRUE);
2770 		else
2771 			ixgbe_dev_lsc_interrupt_setup(dev, FALSE);
2772 		ixgbe_dev_macsec_interrupt_setup(dev);
2773 	} else {
2774 		rte_intr_callback_unregister(intr_handle,
2775 					     ixgbe_dev_interrupt_handler, dev);
2776 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2777 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
2778 				     " no intr multiplex");
2779 	}
2780 
2781 	/* check if rxq interrupt is enabled */
2782 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2783 	    rte_intr_dp_is_en(intr_handle))
2784 		ixgbe_dev_rxq_interrupt_setup(dev);
2785 
2786 	/* enable uio/vfio intr/eventfd mapping */
2787 	rte_intr_enable(intr_handle);
2788 
2789 	/* resume enabled intr since hw reset */
2790 	ixgbe_enable_intr(dev);
2791 	ixgbe_l2_tunnel_conf(dev);
2792 	ixgbe_filter_restore(dev);
2793 
2794 	if (tm_conf->root && !tm_conf->committed)
2795 		PMD_DRV_LOG(WARNING,
2796 			    "please call hierarchy_commit() "
2797 			    "before starting the port");
2798 
2799 	/* wait for the controller to acquire link */
2800 	err = ixgbe_wait_for_link_up(hw);
2801 	if (err)
2802 		goto error;
2803 
2804 	/*
2805 	 * Update link status right before return, because it may
2806 	 * start link configuration process in a separate thread.
2807 	 */
2808 	ixgbe_dev_link_update(dev, 0);
2809 
2810 	/* setup the macsec setting register */
2811 	if (macsec_setting->offload_en)
2812 		ixgbe_dev_macsec_register_enable(dev, macsec_setting);
2813 
2814 	return 0;
2815 
2816 error:
2817 	PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2818 	ixgbe_dev_clear_queues(dev);
2819 	return -EIO;
2820 }
2821 
2822 /*
2823  * Stop device: disable rx and tx functions to allow for reconfiguring.
2824  */
2825 static int
2826 ixgbe_dev_stop(struct rte_eth_dev *dev)
2827 {
2828 	struct rte_eth_link link;
2829 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2830 	struct ixgbe_hw *hw =
2831 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2832 	struct ixgbe_vf_info *vfinfo =
2833 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2834 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2835 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2836 	int vf;
2837 	struct ixgbe_tm_conf *tm_conf =
2838 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2839 
2840 	if (hw->adapter_stopped)
2841 		return 0;
2842 
2843 	PMD_INIT_FUNC_TRACE();
2844 
2845 	ixgbe_dev_wait_setup_link_complete(dev, 0);
2846 
2847 	/* disable interrupts */
2848 	ixgbe_disable_intr(hw);
2849 
2850 	/* reset the NIC */
2851 	ixgbe_pf_reset_hw(hw);
2852 	hw->adapter_stopped = 0;
2853 
2854 	/* stop adapter */
2855 	ixgbe_stop_adapter(hw);
2856 
2857 	for (vf = 0; vfinfo != NULL && vf < pci_dev->max_vfs; vf++)
2858 		vfinfo[vf].clear_to_send = false;
2859 
2860 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2861 		/* Turn off the copper */
2862 		ixgbe_set_phy_power(hw, false);
2863 	} else {
2864 		/* Turn off the laser */
2865 		ixgbe_disable_tx_laser(hw);
2866 	}
2867 
2868 	ixgbe_dev_clear_queues(dev);
2869 
2870 	/* Clear stored conf */
2871 	dev->data->scattered_rx = 0;
2872 	dev->data->lro = 0;
2873 
2874 	/* Clear recorded link status */
2875 	memset(&link, 0, sizeof(link));
2876 	rte_eth_linkstatus_set(dev, &link);
2877 
2878 	if (!rte_intr_allow_others(intr_handle))
2879 		/* resume to the default handler */
2880 		rte_intr_callback_register(intr_handle,
2881 					   ixgbe_dev_interrupt_handler,
2882 					   (void *)dev);
2883 
2884 	/* Clean datapath event and queue/vec mapping */
2885 	rte_intr_efd_disable(intr_handle);
2886 	rte_intr_vec_list_free(intr_handle);
2887 
2888 	/* reset hierarchy commit */
2889 	tm_conf->committed = false;
2890 
2891 	adapter->rss_reta_updated = 0;
2892 
2893 	hw->adapter_stopped = true;
2894 	dev->data->dev_started = 0;
2895 
2896 	return 0;
2897 }
2898 
2899 /*
2900  * Set device link up: enable tx.
2901  */
2902 static int
2903 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2904 {
2905 	struct ixgbe_hw *hw =
2906 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2907 	if (hw->mac.type == ixgbe_mac_82599EB) {
2908 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2909 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2910 			/* Not suported in bypass mode */
2911 			PMD_INIT_LOG(ERR, "Set link up is not supported "
2912 				     "by device id 0x%x", hw->device_id);
2913 			return -ENOTSUP;
2914 		}
2915 #endif
2916 	}
2917 
2918 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2919 		/* Turn on the copper */
2920 		ixgbe_set_phy_power(hw, true);
2921 	} else {
2922 		/* Turn on the laser */
2923 		ixgbe_enable_tx_laser(hw);
2924 		ixgbe_dev_link_update(dev, 0);
2925 	}
2926 
2927 	return 0;
2928 }
2929 
2930 /*
2931  * Set device link down: disable tx.
2932  */
2933 static int
2934 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2935 {
2936 	struct ixgbe_hw *hw =
2937 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2938 	if (hw->mac.type == ixgbe_mac_82599EB) {
2939 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2940 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2941 			/* Not suported in bypass mode */
2942 			PMD_INIT_LOG(ERR, "Set link down is not supported "
2943 				     "by device id 0x%x", hw->device_id);
2944 			return -ENOTSUP;
2945 		}
2946 #endif
2947 	}
2948 
2949 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2950 		/* Turn off the copper */
2951 		ixgbe_set_phy_power(hw, false);
2952 	} else {
2953 		/* Turn off the laser */
2954 		ixgbe_disable_tx_laser(hw);
2955 		ixgbe_dev_link_update(dev, 0);
2956 	}
2957 
2958 	return 0;
2959 }
2960 
2961 /*
2962  * Reset and stop device.
2963  */
2964 static int
2965 ixgbe_dev_close(struct rte_eth_dev *dev)
2966 {
2967 	struct ixgbe_hw *hw =
2968 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2969 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2970 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2971 	int retries = 0;
2972 	int ret;
2973 
2974 	PMD_INIT_FUNC_TRACE();
2975 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2976 		return 0;
2977 
2978 	ixgbe_pf_reset_hw(hw);
2979 
2980 	ret = ixgbe_dev_stop(dev);
2981 
2982 	ixgbe_dev_free_queues(dev);
2983 
2984 	ixgbe_disable_pcie_master(hw);
2985 
2986 	/* reprogram the RAR[0] in case user changed it. */
2987 	ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
2988 
2989 	/* Unlock any pending hardware semaphore */
2990 	ixgbe_swfw_lock_reset(hw);
2991 
2992 	/* disable uio intr before callback unregister */
2993 	rte_intr_disable(intr_handle);
2994 
2995 	do {
2996 		ret = rte_intr_callback_unregister(intr_handle,
2997 				ixgbe_dev_interrupt_handler, dev);
2998 		if (ret >= 0 || ret == -ENOENT) {
2999 			break;
3000 		} else if (ret != -EAGAIN) {
3001 			PMD_INIT_LOG(ERR,
3002 				"intr callback unregister failed: %d",
3003 				ret);
3004 		}
3005 		rte_delay_ms(100);
3006 	} while (retries++ < (10 + IXGBE_LINK_UP_TIME));
3007 
3008 	/* cancel the delay handler before remove dev */
3009 	rte_eal_alarm_cancel(ixgbe_dev_interrupt_delayed_handler, dev);
3010 
3011 	/* uninitialize PF if max_vfs not zero */
3012 	ixgbe_pf_host_uninit(dev);
3013 
3014 	/* remove all the fdir filters & hash */
3015 	ixgbe_fdir_filter_uninit(dev);
3016 
3017 	/* remove all the L2 tunnel filters & hash */
3018 	ixgbe_l2_tn_filter_uninit(dev);
3019 
3020 	/* Remove all ntuple filters of the device */
3021 	ixgbe_ntuple_filter_uninit(dev);
3022 
3023 	/* clear all the filters list */
3024 	ixgbe_filterlist_flush();
3025 
3026 	/* Remove all Traffic Manager configuration */
3027 	ixgbe_tm_conf_uninit(dev);
3028 
3029 #ifdef RTE_LIB_SECURITY
3030 	rte_free(dev->security_ctx);
3031 #endif
3032 
3033 	return ret;
3034 }
3035 
3036 /*
3037  * Reset PF device.
3038  */
3039 static int
3040 ixgbe_dev_reset(struct rte_eth_dev *dev)
3041 {
3042 	int ret;
3043 
3044 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
3045 	 * its VF to make them align with it. The detailed notification
3046 	 * mechanism is PMD specific. As to ixgbe PF, it is rather complex.
3047 	 * To avoid unexpected behavior in VF, currently reset of PF with
3048 	 * SR-IOV activation is not supported. It might be supported later.
3049 	 */
3050 	if (dev->data->sriov.active)
3051 		return -ENOTSUP;
3052 
3053 	ret = eth_ixgbe_dev_uninit(dev);
3054 	if (ret)
3055 		return ret;
3056 
3057 	ret = eth_ixgbe_dev_init(dev, NULL);
3058 
3059 	return ret;
3060 }
3061 
3062 static void
3063 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
3064 			   struct ixgbe_hw_stats *hw_stats,
3065 			   struct ixgbe_macsec_stats *macsec_stats,
3066 			   uint64_t *total_missed_rx, uint64_t *total_qbrc,
3067 			   uint64_t *total_qprc, uint64_t *total_qprdc)
3068 {
3069 	uint32_t bprc, lxon, lxoff, total;
3070 	uint32_t delta_gprc = 0;
3071 	unsigned i;
3072 	/* Workaround for RX byte count not including CRC bytes when CRC
3073 	 * strip is enabled. CRC bytes are removed from counters when crc_strip
3074 	 * is disabled.
3075 	 */
3076 	int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
3077 			IXGBE_HLREG0_RXCRCSTRP);
3078 
3079 	hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
3080 	hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
3081 	hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
3082 	hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
3083 
3084 	for (i = 0; i < 8; i++) {
3085 		uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
3086 
3087 		/* global total per queue */
3088 		hw_stats->mpc[i] += mp;
3089 		/* Running comprehensive total for stats display */
3090 		*total_missed_rx += hw_stats->mpc[i];
3091 		if (hw->mac.type == ixgbe_mac_82598EB) {
3092 			hw_stats->rnbc[i] +=
3093 			    IXGBE_READ_REG(hw, IXGBE_RNBC(i));
3094 			hw_stats->pxonrxc[i] +=
3095 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
3096 			hw_stats->pxoffrxc[i] +=
3097 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
3098 		} else {
3099 			hw_stats->pxonrxc[i] +=
3100 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
3101 			hw_stats->pxoffrxc[i] +=
3102 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
3103 			hw_stats->pxon2offc[i] +=
3104 				IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
3105 		}
3106 		hw_stats->pxontxc[i] +=
3107 		    IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
3108 		hw_stats->pxofftxc[i] +=
3109 		    IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
3110 	}
3111 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3112 		uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
3113 		uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
3114 		uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
3115 
3116 		delta_gprc += delta_qprc;
3117 
3118 		hw_stats->qprc[i] += delta_qprc;
3119 		hw_stats->qptc[i] += delta_qptc;
3120 
3121 		hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
3122 		hw_stats->qbrc[i] +=
3123 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
3124 		if (crc_strip == 0)
3125 			hw_stats->qbrc[i] -= delta_qprc * RTE_ETHER_CRC_LEN;
3126 
3127 		hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
3128 		hw_stats->qbtc[i] +=
3129 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
3130 
3131 		hw_stats->qprdc[i] += delta_qprdc;
3132 		*total_qprdc += hw_stats->qprdc[i];
3133 
3134 		*total_qprc += hw_stats->qprc[i];
3135 		*total_qbrc += hw_stats->qbrc[i];
3136 	}
3137 	hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
3138 	hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
3139 	hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
3140 
3141 	/*
3142 	 * An errata states that gprc actually counts good + missed packets:
3143 	 * Workaround to set gprc to summated queue packet receives
3144 	 */
3145 	hw_stats->gprc = *total_qprc;
3146 
3147 	if (hw->mac.type != ixgbe_mac_82598EB) {
3148 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
3149 		hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
3150 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
3151 		hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
3152 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
3153 		hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
3154 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
3155 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
3156 	} else {
3157 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
3158 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
3159 		/* 82598 only has a counter in the high register */
3160 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
3161 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
3162 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
3163 	}
3164 	uint64_t old_tpr = hw_stats->tpr;
3165 
3166 	hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
3167 	hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
3168 
3169 	if (crc_strip == 0)
3170 		hw_stats->gorc -= delta_gprc * RTE_ETHER_CRC_LEN;
3171 
3172 	uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
3173 	hw_stats->gptc += delta_gptc;
3174 	hw_stats->gotc -= delta_gptc * RTE_ETHER_CRC_LEN;
3175 	hw_stats->tor -= (hw_stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
3176 
3177 	/*
3178 	 * Workaround: mprc hardware is incorrectly counting
3179 	 * broadcasts, so for now we subtract those.
3180 	 */
3181 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
3182 	hw_stats->bprc += bprc;
3183 	hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
3184 	if (hw->mac.type == ixgbe_mac_82598EB)
3185 		hw_stats->mprc -= bprc;
3186 
3187 	hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
3188 	hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
3189 	hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
3190 	hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
3191 	hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
3192 	hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
3193 
3194 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
3195 	hw_stats->lxontxc += lxon;
3196 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
3197 	hw_stats->lxofftxc += lxoff;
3198 	total = lxon + lxoff;
3199 
3200 	hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
3201 	hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
3202 	hw_stats->gptc -= total;
3203 	hw_stats->mptc -= total;
3204 	hw_stats->ptc64 -= total;
3205 	hw_stats->gotc -= total * RTE_ETHER_MIN_LEN;
3206 
3207 	hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3208 	hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3209 	hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3210 	hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
3211 	hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
3212 	hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
3213 	hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
3214 	hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3215 	hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3216 	hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3217 	hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3218 	hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
3219 	hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3220 	hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
3221 	hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
3222 	hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
3223 	/* Only read FCOE on 82599 */
3224 	if (hw->mac.type != ixgbe_mac_82598EB) {
3225 		hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
3226 		hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
3227 		hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
3228 		hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
3229 		hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
3230 	}
3231 
3232 	/* Flow Director Stats registers */
3233 	if (hw->mac.type != ixgbe_mac_82598EB) {
3234 		hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
3235 		hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
3236 		hw_stats->fdirustat_add += IXGBE_READ_REG(hw,
3237 					IXGBE_FDIRUSTAT) & 0xFFFF;
3238 		hw_stats->fdirustat_remove += (IXGBE_READ_REG(hw,
3239 					IXGBE_FDIRUSTAT) >> 16) & 0xFFFF;
3240 		hw_stats->fdirfstat_fadd += IXGBE_READ_REG(hw,
3241 					IXGBE_FDIRFSTAT) & 0xFFFF;
3242 		hw_stats->fdirfstat_fremove += (IXGBE_READ_REG(hw,
3243 					IXGBE_FDIRFSTAT) >> 16) & 0xFFFF;
3244 	}
3245 	/* MACsec Stats registers */
3246 	macsec_stats->out_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECTXUT);
3247 	macsec_stats->out_pkts_encrypted +=
3248 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTE);
3249 	macsec_stats->out_pkts_protected +=
3250 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTP);
3251 	macsec_stats->out_octets_encrypted +=
3252 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTE);
3253 	macsec_stats->out_octets_protected +=
3254 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTP);
3255 	macsec_stats->in_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECRXUT);
3256 	macsec_stats->in_pkts_badtag += IXGBE_READ_REG(hw, IXGBE_LSECRXBAD);
3257 	macsec_stats->in_pkts_nosci += IXGBE_READ_REG(hw, IXGBE_LSECRXNOSCI);
3258 	macsec_stats->in_pkts_unknownsci +=
3259 		IXGBE_READ_REG(hw, IXGBE_LSECRXUNSCI);
3260 	macsec_stats->in_octets_decrypted +=
3261 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTD);
3262 	macsec_stats->in_octets_validated +=
3263 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTV);
3264 	macsec_stats->in_pkts_unchecked += IXGBE_READ_REG(hw, IXGBE_LSECRXUNCH);
3265 	macsec_stats->in_pkts_delayed += IXGBE_READ_REG(hw, IXGBE_LSECRXDELAY);
3266 	macsec_stats->in_pkts_late += IXGBE_READ_REG(hw, IXGBE_LSECRXLATE);
3267 	for (i = 0; i < 2; i++) {
3268 		macsec_stats->in_pkts_ok +=
3269 			IXGBE_READ_REG(hw, IXGBE_LSECRXOK(i));
3270 		macsec_stats->in_pkts_invalid +=
3271 			IXGBE_READ_REG(hw, IXGBE_LSECRXINV(i));
3272 		macsec_stats->in_pkts_notvalid +=
3273 			IXGBE_READ_REG(hw, IXGBE_LSECRXNV(i));
3274 	}
3275 	macsec_stats->in_pkts_unusedsa += IXGBE_READ_REG(hw, IXGBE_LSECRXUNSA);
3276 	macsec_stats->in_pkts_notusingsa +=
3277 		IXGBE_READ_REG(hw, IXGBE_LSECRXNUSA);
3278 }
3279 
3280 /*
3281  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
3282  */
3283 static int
3284 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3285 {
3286 	struct ixgbe_hw *hw =
3287 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3288 	struct ixgbe_hw_stats *hw_stats =
3289 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3290 	struct ixgbe_macsec_stats *macsec_stats =
3291 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3292 				dev->data->dev_private);
3293 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3294 	unsigned i;
3295 
3296 	total_missed_rx = 0;
3297 	total_qbrc = 0;
3298 	total_qprc = 0;
3299 	total_qprdc = 0;
3300 
3301 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3302 			&total_qbrc, &total_qprc, &total_qprdc);
3303 
3304 	if (stats == NULL)
3305 		return -EINVAL;
3306 
3307 	/* Fill out the rte_eth_stats statistics structure */
3308 	stats->ipackets = total_qprc;
3309 	stats->ibytes = total_qbrc;
3310 	stats->opackets = hw_stats->gptc;
3311 	stats->obytes = hw_stats->gotc;
3312 
3313 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3314 		stats->q_ipackets[i] = hw_stats->qprc[i];
3315 		stats->q_opackets[i] = hw_stats->qptc[i];
3316 		stats->q_ibytes[i] = hw_stats->qbrc[i];
3317 		stats->q_obytes[i] = hw_stats->qbtc[i];
3318 		stats->q_errors[i] = hw_stats->qprdc[i];
3319 	}
3320 
3321 	/* Rx Errors */
3322 	stats->imissed  = total_missed_rx;
3323 	stats->ierrors  = hw_stats->crcerrs +
3324 			  hw_stats->mspdc +
3325 			  hw_stats->rlec +
3326 			  hw_stats->ruc +
3327 			  hw_stats->roc +
3328 			  hw_stats->illerrc +
3329 			  hw_stats->errbc +
3330 			  hw_stats->rfc +
3331 			  hw_stats->fccrc +
3332 			  hw_stats->fclast;
3333 
3334 	/*
3335 	 * 82599 errata, UDP frames with a 0 checksum can be marked as checksum
3336 	 * errors.
3337 	 */
3338 	if (hw->mac.type != ixgbe_mac_82599EB)
3339 		stats->ierrors += hw_stats->xec;
3340 
3341 	/* Tx Errors */
3342 	stats->oerrors  = 0;
3343 	return 0;
3344 }
3345 
3346 static int
3347 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
3348 {
3349 	struct ixgbe_hw_stats *stats =
3350 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3351 
3352 	/* HW registers are cleared on read */
3353 	ixgbe_dev_stats_get(dev, NULL);
3354 
3355 	/* Reset software totals */
3356 	memset(stats, 0, sizeof(*stats));
3357 
3358 	return 0;
3359 }
3360 
3361 /* This function calculates the number of xstats based on the current config */
3362 static unsigned
3363 ixgbe_xstats_calc_num(void) {
3364 	return IXGBE_NB_HW_STATS + IXGBE_NB_MACSEC_STATS +
3365 		(IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
3366 		(IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
3367 }
3368 
3369 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3370 	struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned int size)
3371 {
3372 	const unsigned cnt_stats = ixgbe_xstats_calc_num();
3373 	unsigned stat, i, count;
3374 
3375 	if (xstats_names != NULL) {
3376 		count = 0;
3377 
3378 		/* Note: limit >= cnt_stats checked upstream
3379 		 * in rte_eth_xstats_names()
3380 		 */
3381 
3382 		/* Extended stats from ixgbe_hw_stats */
3383 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3384 			strlcpy(xstats_names[count].name,
3385 				rte_ixgbe_stats_strings[i].name,
3386 				sizeof(xstats_names[count].name));
3387 			count++;
3388 		}
3389 
3390 		/* MACsec Stats */
3391 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3392 			strlcpy(xstats_names[count].name,
3393 				rte_ixgbe_macsec_strings[i].name,
3394 				sizeof(xstats_names[count].name));
3395 			count++;
3396 		}
3397 
3398 		/* RX Priority Stats */
3399 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3400 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3401 				snprintf(xstats_names[count].name,
3402 					sizeof(xstats_names[count].name),
3403 					"rx_priority%u_%s", i,
3404 					rte_ixgbe_rxq_strings[stat].name);
3405 				count++;
3406 			}
3407 		}
3408 
3409 		/* TX Priority Stats */
3410 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3411 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3412 				snprintf(xstats_names[count].name,
3413 					sizeof(xstats_names[count].name),
3414 					"tx_priority%u_%s", i,
3415 					rte_ixgbe_txq_strings[stat].name);
3416 				count++;
3417 			}
3418 		}
3419 	}
3420 	return cnt_stats;
3421 }
3422 
3423 static int ixgbe_dev_xstats_get_names_by_id(
3424 	struct rte_eth_dev *dev,
3425 	const uint64_t *ids,
3426 	struct rte_eth_xstat_name *xstats_names,
3427 	unsigned int limit)
3428 {
3429 	if (!ids) {
3430 		const unsigned int cnt_stats = ixgbe_xstats_calc_num();
3431 		unsigned int stat, i, count;
3432 
3433 		if (xstats_names != NULL) {
3434 			count = 0;
3435 
3436 			/* Note: limit >= cnt_stats checked upstream
3437 			 * in rte_eth_xstats_names()
3438 			 */
3439 
3440 			/* Extended stats from ixgbe_hw_stats */
3441 			for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3442 				strlcpy(xstats_names[count].name,
3443 					rte_ixgbe_stats_strings[i].name,
3444 					sizeof(xstats_names[count].name));
3445 				count++;
3446 			}
3447 
3448 			/* MACsec Stats */
3449 			for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3450 				strlcpy(xstats_names[count].name,
3451 					rte_ixgbe_macsec_strings[i].name,
3452 					sizeof(xstats_names[count].name));
3453 				count++;
3454 			}
3455 
3456 			/* RX Priority Stats */
3457 			for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3458 				for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3459 					snprintf(xstats_names[count].name,
3460 					    sizeof(xstats_names[count].name),
3461 					    "rx_priority%u_%s", i,
3462 					    rte_ixgbe_rxq_strings[stat].name);
3463 					count++;
3464 				}
3465 			}
3466 
3467 			/* TX Priority Stats */
3468 			for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3469 				for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3470 					snprintf(xstats_names[count].name,
3471 					    sizeof(xstats_names[count].name),
3472 					    "tx_priority%u_%s", i,
3473 					    rte_ixgbe_txq_strings[stat].name);
3474 					count++;
3475 				}
3476 			}
3477 		}
3478 		return cnt_stats;
3479 	}
3480 
3481 	uint16_t i;
3482 	uint16_t size = ixgbe_xstats_calc_num();
3483 	struct rte_eth_xstat_name xstats_names_copy[size];
3484 
3485 	ixgbe_dev_xstats_get_names_by_id(dev, NULL, xstats_names_copy,
3486 			size);
3487 
3488 	for (i = 0; i < limit; i++) {
3489 		if (ids[i] >= size) {
3490 			PMD_INIT_LOG(ERR, "id value isn't valid");
3491 			return -1;
3492 		}
3493 		strcpy(xstats_names[i].name,
3494 				xstats_names_copy[ids[i]].name);
3495 	}
3496 	return limit;
3497 }
3498 
3499 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3500 	struct rte_eth_xstat_name *xstats_names, unsigned limit)
3501 {
3502 	unsigned i;
3503 
3504 	if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
3505 		return -ENOMEM;
3506 
3507 	if (xstats_names != NULL)
3508 		for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
3509 			strlcpy(xstats_names[i].name,
3510 				rte_ixgbevf_stats_strings[i].name,
3511 				sizeof(xstats_names[i].name));
3512 	return IXGBEVF_NB_XSTATS;
3513 }
3514 
3515 static int
3516 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3517 					 unsigned n)
3518 {
3519 	struct ixgbe_hw *hw =
3520 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3521 	struct ixgbe_hw_stats *hw_stats =
3522 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3523 	struct ixgbe_macsec_stats *macsec_stats =
3524 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3525 				dev->data->dev_private);
3526 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3527 	unsigned i, stat, count = 0;
3528 
3529 	count = ixgbe_xstats_calc_num();
3530 
3531 	if (n < count)
3532 		return count;
3533 
3534 	total_missed_rx = 0;
3535 	total_qbrc = 0;
3536 	total_qprc = 0;
3537 	total_qprdc = 0;
3538 
3539 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3540 			&total_qbrc, &total_qprc, &total_qprdc);
3541 
3542 	/* If this is a reset xstats is NULL, and we have cleared the
3543 	 * registers by reading them.
3544 	 */
3545 	if (!xstats)
3546 		return 0;
3547 
3548 	/* Extended stats from ixgbe_hw_stats */
3549 	count = 0;
3550 	for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3551 		xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3552 				rte_ixgbe_stats_strings[i].offset);
3553 		xstats[count].id = count;
3554 		count++;
3555 	}
3556 
3557 	/* MACsec Stats */
3558 	for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3559 		xstats[count].value = *(uint64_t *)(((char *)macsec_stats) +
3560 				rte_ixgbe_macsec_strings[i].offset);
3561 		xstats[count].id = count;
3562 		count++;
3563 	}
3564 
3565 	/* RX Priority Stats */
3566 	for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3567 		for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3568 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3569 					rte_ixgbe_rxq_strings[stat].offset +
3570 					(sizeof(uint64_t) * i));
3571 			xstats[count].id = count;
3572 			count++;
3573 		}
3574 	}
3575 
3576 	/* TX Priority Stats */
3577 	for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3578 		for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3579 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3580 					rte_ixgbe_txq_strings[stat].offset +
3581 					(sizeof(uint64_t) * i));
3582 			xstats[count].id = count;
3583 			count++;
3584 		}
3585 	}
3586 	return count;
3587 }
3588 
3589 static int
3590 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
3591 		uint64_t *values, unsigned int n)
3592 {
3593 	if (!ids) {
3594 		struct ixgbe_hw *hw =
3595 				IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3596 		struct ixgbe_hw_stats *hw_stats =
3597 				IXGBE_DEV_PRIVATE_TO_STATS(
3598 						dev->data->dev_private);
3599 		struct ixgbe_macsec_stats *macsec_stats =
3600 				IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3601 					dev->data->dev_private);
3602 		uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3603 		unsigned int i, stat, count = 0;
3604 
3605 		count = ixgbe_xstats_calc_num();
3606 
3607 		if (!ids && n < count)
3608 			return count;
3609 
3610 		total_missed_rx = 0;
3611 		total_qbrc = 0;
3612 		total_qprc = 0;
3613 		total_qprdc = 0;
3614 
3615 		ixgbe_read_stats_registers(hw, hw_stats, macsec_stats,
3616 				&total_missed_rx, &total_qbrc, &total_qprc,
3617 				&total_qprdc);
3618 
3619 		/* If this is a reset xstats is NULL, and we have cleared the
3620 		 * registers by reading them.
3621 		 */
3622 		if (!ids && !values)
3623 			return 0;
3624 
3625 		/* Extended stats from ixgbe_hw_stats */
3626 		count = 0;
3627 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3628 			values[count] = *(uint64_t *)(((char *)hw_stats) +
3629 					rte_ixgbe_stats_strings[i].offset);
3630 			count++;
3631 		}
3632 
3633 		/* MACsec Stats */
3634 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3635 			values[count] = *(uint64_t *)(((char *)macsec_stats) +
3636 					rte_ixgbe_macsec_strings[i].offset);
3637 			count++;
3638 		}
3639 
3640 		/* RX Priority Stats */
3641 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3642 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3643 				values[count] =
3644 					*(uint64_t *)(((char *)hw_stats) +
3645 					rte_ixgbe_rxq_strings[stat].offset +
3646 					(sizeof(uint64_t) * i));
3647 				count++;
3648 			}
3649 		}
3650 
3651 		/* TX Priority Stats */
3652 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3653 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3654 				values[count] =
3655 					*(uint64_t *)(((char *)hw_stats) +
3656 					rte_ixgbe_txq_strings[stat].offset +
3657 					(sizeof(uint64_t) * i));
3658 				count++;
3659 			}
3660 		}
3661 		return count;
3662 	}
3663 
3664 	uint16_t i;
3665 	uint16_t size = ixgbe_xstats_calc_num();
3666 	uint64_t values_copy[size];
3667 
3668 	ixgbe_dev_xstats_get_by_id(dev, NULL, values_copy, size);
3669 
3670 	for (i = 0; i < n; i++) {
3671 		if (ids[i] >= size) {
3672 			PMD_INIT_LOG(ERR, "id value isn't valid");
3673 			return -1;
3674 		}
3675 		values[i] = values_copy[ids[i]];
3676 	}
3677 	return n;
3678 }
3679 
3680 static int
3681 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
3682 {
3683 	struct ixgbe_hw_stats *stats =
3684 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3685 	struct ixgbe_macsec_stats *macsec_stats =
3686 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3687 				dev->data->dev_private);
3688 
3689 	unsigned count = ixgbe_xstats_calc_num();
3690 
3691 	/* HW registers are cleared on read */
3692 	ixgbe_dev_xstats_get(dev, NULL, count);
3693 
3694 	/* Reset software totals */
3695 	memset(stats, 0, sizeof(*stats));
3696 	memset(macsec_stats, 0, sizeof(*macsec_stats));
3697 
3698 	return 0;
3699 }
3700 
3701 static void
3702 ixgbevf_update_stats(struct rte_eth_dev *dev)
3703 {
3704 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3705 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3706 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3707 
3708 	/* Good Rx packet, include VF loopback */
3709 	UPDATE_VF_STAT(IXGBE_VFGPRC,
3710 	    hw_stats->last_vfgprc, hw_stats->vfgprc);
3711 
3712 	/* Good Rx octets, include VF loopback */
3713 	UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
3714 	    hw_stats->last_vfgorc, hw_stats->vfgorc);
3715 
3716 	/* Good Tx packet, include VF loopback */
3717 	UPDATE_VF_STAT(IXGBE_VFGPTC,
3718 	    hw_stats->last_vfgptc, hw_stats->vfgptc);
3719 
3720 	/* Good Tx octets, include VF loopback */
3721 	UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
3722 	    hw_stats->last_vfgotc, hw_stats->vfgotc);
3723 
3724 	/* Rx Multicst Packet */
3725 	UPDATE_VF_STAT(IXGBE_VFMPRC,
3726 	    hw_stats->last_vfmprc, hw_stats->vfmprc);
3727 }
3728 
3729 static int
3730 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3731 		       unsigned n)
3732 {
3733 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3734 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3735 	unsigned i;
3736 
3737 	if (n < IXGBEVF_NB_XSTATS)
3738 		return IXGBEVF_NB_XSTATS;
3739 
3740 	ixgbevf_update_stats(dev);
3741 
3742 	if (!xstats)
3743 		return 0;
3744 
3745 	/* Extended stats */
3746 	for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
3747 		xstats[i].id = i;
3748 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
3749 			rte_ixgbevf_stats_strings[i].offset);
3750 	}
3751 
3752 	return IXGBEVF_NB_XSTATS;
3753 }
3754 
3755 static int
3756 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3757 {
3758 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3759 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3760 
3761 	ixgbevf_update_stats(dev);
3762 
3763 	if (stats == NULL)
3764 		return -EINVAL;
3765 
3766 	stats->ipackets = hw_stats->vfgprc;
3767 	stats->ibytes = hw_stats->vfgorc;
3768 	stats->opackets = hw_stats->vfgptc;
3769 	stats->obytes = hw_stats->vfgotc;
3770 	return 0;
3771 }
3772 
3773 static int
3774 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3775 {
3776 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3777 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3778 
3779 	/* Sync HW register to the last stats */
3780 	ixgbevf_dev_stats_get(dev, NULL);
3781 
3782 	/* reset HW current stats*/
3783 	hw_stats->vfgprc = 0;
3784 	hw_stats->vfgorc = 0;
3785 	hw_stats->vfgptc = 0;
3786 	hw_stats->vfgotc = 0;
3787 	hw_stats->vfmprc = 0;
3788 
3789 	return 0;
3790 }
3791 
3792 static int
3793 ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
3794 {
3795 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3796 	u16 eeprom_verh, eeprom_verl;
3797 	u32 etrack_id;
3798 	int ret;
3799 
3800 	ixgbe_read_eeprom(hw, 0x2e, &eeprom_verh);
3801 	ixgbe_read_eeprom(hw, 0x2d, &eeprom_verl);
3802 
3803 	etrack_id = (eeprom_verh << 16) | eeprom_verl;
3804 	ret = snprintf(fw_version, fw_size, "0x%08x", etrack_id);
3805 	if (ret < 0)
3806 		return -EINVAL;
3807 
3808 	ret += 1; /* add the size of '\0' */
3809 	if (fw_size < (size_t)ret)
3810 		return ret;
3811 	else
3812 		return 0;
3813 }
3814 
3815 static int
3816 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3817 {
3818 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3819 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3820 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3821 
3822 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3823 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3824 	if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3825 		/*
3826 		 * When DCB/VT is off, maximum number of queues changes,
3827 		 * except for 82598EB, which remains constant.
3828 		 */
3829 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_NONE &&
3830 				hw->mac.type != ixgbe_mac_82598EB)
3831 			dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3832 	}
3833 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3834 	dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3835 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3836 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3837 	dev_info->max_vfs = pci_dev->max_vfs;
3838 	if (hw->mac.type == ixgbe_mac_82598EB)
3839 		dev_info->max_vmdq_pools = RTE_ETH_16_POOLS;
3840 	else
3841 		dev_info->max_vmdq_pools = RTE_ETH_64_POOLS;
3842 	dev_info->max_mtu =  dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3843 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
3844 	dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3845 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3846 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3847 				     dev_info->rx_queue_offload_capa);
3848 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3849 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3850 
3851 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3852 		.rx_thresh = {
3853 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3854 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3855 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3856 		},
3857 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3858 		.rx_drop_en = 0,
3859 		.offloads = 0,
3860 	};
3861 
3862 	dev_info->default_txconf = (struct rte_eth_txconf) {
3863 		.tx_thresh = {
3864 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3865 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3866 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3867 		},
3868 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3869 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3870 		.offloads = 0,
3871 	};
3872 
3873 	dev_info->rx_desc_lim = rx_desc_lim;
3874 	dev_info->tx_desc_lim = tx_desc_lim;
3875 
3876 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3877 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3878 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3879 
3880 	dev_info->speed_capa = RTE_ETH_LINK_SPEED_1G | RTE_ETH_LINK_SPEED_10G;
3881 	if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
3882 			hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
3883 		dev_info->speed_capa = RTE_ETH_LINK_SPEED_10M |
3884 			RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G;
3885 
3886 	if (hw->mac.type == ixgbe_mac_X540 ||
3887 	    hw->mac.type == ixgbe_mac_X540_vf ||
3888 	    hw->mac.type == ixgbe_mac_X550 ||
3889 	    hw->mac.type == ixgbe_mac_X550_vf) {
3890 		dev_info->speed_capa |= RTE_ETH_LINK_SPEED_100M;
3891 	}
3892 	if (hw->mac.type == ixgbe_mac_X550) {
3893 		dev_info->speed_capa |= RTE_ETH_LINK_SPEED_2_5G;
3894 		dev_info->speed_capa |= RTE_ETH_LINK_SPEED_5G;
3895 	}
3896 
3897 	/* Driver-preferred Rx/Tx parameters */
3898 	dev_info->default_rxportconf.burst_size = 32;
3899 	dev_info->default_txportconf.burst_size = 32;
3900 	dev_info->default_rxportconf.nb_queues = 1;
3901 	dev_info->default_txportconf.nb_queues = 1;
3902 	dev_info->default_rxportconf.ring_size = 256;
3903 	dev_info->default_txportconf.ring_size = 256;
3904 
3905 	return 0;
3906 }
3907 
3908 static const uint32_t *
3909 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3910 {
3911 	static const uint32_t ptypes[] = {
3912 		/* For non-vec functions,
3913 		 * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3914 		 * for vec functions,
3915 		 * refers to _recv_raw_pkts_vec().
3916 		 */
3917 		RTE_PTYPE_L2_ETHER,
3918 		RTE_PTYPE_L3_IPV4,
3919 		RTE_PTYPE_L3_IPV4_EXT,
3920 		RTE_PTYPE_L3_IPV6,
3921 		RTE_PTYPE_L3_IPV6_EXT,
3922 		RTE_PTYPE_L4_SCTP,
3923 		RTE_PTYPE_L4_TCP,
3924 		RTE_PTYPE_L4_UDP,
3925 		RTE_PTYPE_TUNNEL_IP,
3926 		RTE_PTYPE_INNER_L3_IPV6,
3927 		RTE_PTYPE_INNER_L3_IPV6_EXT,
3928 		RTE_PTYPE_INNER_L4_TCP,
3929 		RTE_PTYPE_INNER_L4_UDP,
3930 		RTE_PTYPE_UNKNOWN
3931 	};
3932 
3933 	if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3934 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3935 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3936 	    dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3937 		return ptypes;
3938 
3939 #if defined(RTE_ARCH_X86) || defined(__ARM_NEON)
3940 	if (dev->rx_pkt_burst == ixgbe_recv_pkts_vec ||
3941 	    dev->rx_pkt_burst == ixgbe_recv_scattered_pkts_vec)
3942 		return ptypes;
3943 #endif
3944 	return NULL;
3945 }
3946 
3947 static int
3948 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3949 		     struct rte_eth_dev_info *dev_info)
3950 {
3951 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3952 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3953 
3954 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3955 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3956 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3957 	dev_info->max_rx_pktlen = 9728; /* includes CRC, cf MAXFRS reg */
3958 	dev_info->max_mtu = dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3959 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3960 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3961 	dev_info->max_vfs = pci_dev->max_vfs;
3962 	if (hw->mac.type == ixgbe_mac_82598EB)
3963 		dev_info->max_vmdq_pools = RTE_ETH_16_POOLS;
3964 	else
3965 		dev_info->max_vmdq_pools = RTE_ETH_64_POOLS;
3966 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3967 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3968 				     dev_info->rx_queue_offload_capa);
3969 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3970 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3971 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3972 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3973 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3974 
3975 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3976 		.rx_thresh = {
3977 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3978 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3979 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3980 		},
3981 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3982 		.rx_drop_en = 0,
3983 		.offloads = 0,
3984 	};
3985 
3986 	dev_info->default_txconf = (struct rte_eth_txconf) {
3987 		.tx_thresh = {
3988 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3989 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3990 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3991 		},
3992 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3993 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3994 		.offloads = 0,
3995 	};
3996 
3997 	dev_info->rx_desc_lim = rx_desc_lim;
3998 	dev_info->tx_desc_lim = tx_desc_lim;
3999 
4000 	return 0;
4001 }
4002 
4003 static int
4004 ixgbevf_check_link(struct ixgbe_hw *hw, ixgbe_link_speed *speed,
4005 		   bool *link_up, int wait_to_complete)
4006 {
4007 	struct ixgbe_adapter *adapter = container_of(hw,
4008 						     struct ixgbe_adapter, hw);
4009 	struct ixgbe_mbx_info *mbx = &hw->mbx;
4010 	struct ixgbe_mac_info *mac = &hw->mac;
4011 	uint32_t links_reg, in_msg;
4012 	int ret_val = 0;
4013 
4014 	/* If we were hit with a reset drop the link */
4015 	if (!mbx->ops.check_for_rst(hw, 0) || !mbx->timeout)
4016 		mac->get_link_status = true;
4017 
4018 	if (!mac->get_link_status)
4019 		goto out;
4020 
4021 	/* if link status is down no point in checking to see if pf is up */
4022 	links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4023 	if (!(links_reg & IXGBE_LINKS_UP))
4024 		goto out;
4025 
4026 	/* for SFP+ modules and DA cables on 82599 it can take up to 500usecs
4027 	 * before the link status is correct
4028 	 */
4029 	if (mac->type == ixgbe_mac_82599_vf && wait_to_complete) {
4030 		int i;
4031 
4032 		for (i = 0; i < 5; i++) {
4033 			rte_delay_us(100);
4034 			links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4035 
4036 			if (!(links_reg & IXGBE_LINKS_UP))
4037 				goto out;
4038 		}
4039 	}
4040 
4041 	switch (links_reg & IXGBE_LINKS_SPEED_82599) {
4042 	case IXGBE_LINKS_SPEED_10G_82599:
4043 		*speed = IXGBE_LINK_SPEED_10GB_FULL;
4044 		if (hw->mac.type >= ixgbe_mac_X550) {
4045 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4046 				*speed = IXGBE_LINK_SPEED_2_5GB_FULL;
4047 		}
4048 		break;
4049 	case IXGBE_LINKS_SPEED_1G_82599:
4050 		*speed = IXGBE_LINK_SPEED_1GB_FULL;
4051 		break;
4052 	case IXGBE_LINKS_SPEED_100_82599:
4053 		*speed = IXGBE_LINK_SPEED_100_FULL;
4054 		if (hw->mac.type == ixgbe_mac_X550) {
4055 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4056 				*speed = IXGBE_LINK_SPEED_5GB_FULL;
4057 		}
4058 		break;
4059 	case IXGBE_LINKS_SPEED_10_X550EM_A:
4060 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
4061 		/* Since Reserved in older MAC's */
4062 		if (hw->mac.type >= ixgbe_mac_X550)
4063 			*speed = IXGBE_LINK_SPEED_10_FULL;
4064 		break;
4065 	default:
4066 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
4067 	}
4068 
4069 	if (wait_to_complete == 0 && adapter->pflink_fullchk == 0) {
4070 		if (*speed == IXGBE_LINK_SPEED_UNKNOWN)
4071 			mac->get_link_status = true;
4072 		else
4073 			mac->get_link_status = false;
4074 
4075 		goto out;
4076 	}
4077 
4078 	/* if the read failed it could just be a mailbox collision, best wait
4079 	 * until we are called again and don't report an error
4080 	 */
4081 	if (mbx->ops.read(hw, &in_msg, 1, 0))
4082 		goto out;
4083 
4084 	if (!(in_msg & IXGBE_VT_MSGTYPE_CTS)) {
4085 		/* msg is not CTS and is NACK we must have lost CTS status */
4086 		if (in_msg & IXGBE_VT_MSGTYPE_NACK)
4087 			mac->get_link_status = false;
4088 		goto out;
4089 	}
4090 
4091 	/* the pf is talking, if we timed out in the past we reinit */
4092 	if (!mbx->timeout) {
4093 		ret_val = -1;
4094 		goto out;
4095 	}
4096 
4097 	/* if we passed all the tests above then the link is up and we no
4098 	 * longer need to check for link
4099 	 */
4100 	mac->get_link_status = false;
4101 
4102 out:
4103 	*link_up = !mac->get_link_status;
4104 	return ret_val;
4105 }
4106 
4107 /*
4108  * If @timeout_ms was 0, it means that it will not return until link complete.
4109  * It returns 1 on complete, return 0 on timeout.
4110  */
4111 static int
4112 ixgbe_dev_wait_setup_link_complete(struct rte_eth_dev *dev, uint32_t timeout_ms)
4113 {
4114 #define WARNING_TIMEOUT    9000 /* 9s  in total */
4115 	struct ixgbe_adapter *ad = dev->data->dev_private;
4116 	uint32_t timeout = timeout_ms ? timeout_ms : WARNING_TIMEOUT;
4117 
4118 	while (rte_atomic32_read(&ad->link_thread_running)) {
4119 		msec_delay(1);
4120 		timeout--;
4121 
4122 		if (timeout_ms) {
4123 			if (!timeout)
4124 				return 0;
4125 		} else if (!timeout) {
4126 			/* It will not return until link complete */
4127 			timeout = WARNING_TIMEOUT;
4128 			PMD_DRV_LOG(ERR, "IXGBE link thread not complete too long time!");
4129 		}
4130 	}
4131 
4132 	return 1;
4133 }
4134 
4135 static void *
4136 ixgbe_dev_setup_link_thread_handler(void *param)
4137 {
4138 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4139 	struct ixgbe_adapter *ad = dev->data->dev_private;
4140 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4141 	struct ixgbe_interrupt *intr =
4142 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4143 	u32 speed;
4144 	bool autoneg = false;
4145 
4146 	pthread_detach(pthread_self());
4147 	speed = hw->phy.autoneg_advertised;
4148 	if (!speed)
4149 		ixgbe_get_link_capabilities(hw, &speed, &autoneg);
4150 
4151 	ixgbe_setup_link(hw, speed, true);
4152 
4153 	intr->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4154 	rte_atomic32_clear(&ad->link_thread_running);
4155 	return NULL;
4156 }
4157 
4158 /*
4159  * In freebsd environment, nic_uio drivers do not support interrupts,
4160  * rte_intr_callback_register() will fail to register interrupts.
4161  * We can not make link status to change from down to up by interrupt
4162  * callback. So we need to wait for the controller to acquire link
4163  * when ports start.
4164  * It returns 0 on link up.
4165  */
4166 static int
4167 ixgbe_wait_for_link_up(struct ixgbe_hw *hw)
4168 {
4169 #ifdef RTE_EXEC_ENV_FREEBSD
4170 	int err, i;
4171 	bool link_up = false;
4172 	uint32_t speed = 0;
4173 	const int nb_iter = 25;
4174 
4175 	for (i = 0; i < nb_iter; i++) {
4176 		err = ixgbe_check_link(hw, &speed, &link_up, 0);
4177 		if (err)
4178 			return err;
4179 		if (link_up)
4180 			return 0;
4181 		msec_delay(200);
4182 	}
4183 
4184 	return 0;
4185 #else
4186 	RTE_SET_USED(hw);
4187 	return 0;
4188 #endif
4189 }
4190 
4191 /* return 0 means link status changed, -1 means not changed */
4192 int
4193 ixgbe_dev_link_update_share(struct rte_eth_dev *dev,
4194 			    int wait_to_complete, int vf)
4195 {
4196 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4197 	struct ixgbe_adapter *ad = dev->data->dev_private;
4198 	struct rte_eth_link link;
4199 	ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
4200 	struct ixgbe_interrupt *intr =
4201 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4202 	bool link_up;
4203 	int diag;
4204 	int wait = 1;
4205 	u32 esdp_reg;
4206 
4207 	memset(&link, 0, sizeof(link));
4208 	link.link_status = RTE_ETH_LINK_DOWN;
4209 	link.link_speed = RTE_ETH_SPEED_NUM_NONE;
4210 	link.link_duplex = RTE_ETH_LINK_HALF_DUPLEX;
4211 	link.link_autoneg = !(dev->data->dev_conf.link_speeds &
4212 			RTE_ETH_LINK_SPEED_FIXED);
4213 
4214 	hw->mac.get_link_status = true;
4215 
4216 	if (intr->flags & IXGBE_FLAG_NEED_LINK_CONFIG)
4217 		return rte_eth_linkstatus_set(dev, &link);
4218 
4219 	/* check if it needs to wait to complete, if lsc interrupt is enabled */
4220 	if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
4221 		wait = 0;
4222 
4223 /* BSD has no interrupt mechanism, so force NIC status synchronization. */
4224 #ifdef RTE_EXEC_ENV_FREEBSD
4225 	wait = 1;
4226 #endif
4227 
4228 	if (vf)
4229 		diag = ixgbevf_check_link(hw, &link_speed, &link_up, wait);
4230 	else
4231 		diag = ixgbe_check_link(hw, &link_speed, &link_up, wait);
4232 
4233 	if (diag != 0) {
4234 		link.link_speed = RTE_ETH_SPEED_NUM_100M;
4235 		link.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
4236 		return rte_eth_linkstatus_set(dev, &link);
4237 	}
4238 
4239 	if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4240 		esdp_reg = IXGBE_READ_REG(hw, IXGBE_ESDP);
4241 		if ((esdp_reg & IXGBE_ESDP_SDP3))
4242 			link_up = 0;
4243 	}
4244 
4245 	if (link_up == 0) {
4246 		if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4247 			ixgbe_dev_wait_setup_link_complete(dev, 0);
4248 			if (rte_atomic32_test_and_set(&ad->link_thread_running)) {
4249 				/* To avoid race condition between threads, set
4250 				 * the IXGBE_FLAG_NEED_LINK_CONFIG flag only
4251 				 * when there is no link thread running.
4252 				 */
4253 				intr->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
4254 				if (rte_ctrl_thread_create(&ad->link_thread_tid,
4255 					"ixgbe-link-handler",
4256 					NULL,
4257 					ixgbe_dev_setup_link_thread_handler,
4258 					dev) < 0) {
4259 					PMD_DRV_LOG(ERR,
4260 						"Create link thread failed!");
4261 					rte_atomic32_clear(&ad->link_thread_running);
4262 				}
4263 			} else {
4264 				PMD_DRV_LOG(ERR,
4265 					"Other link thread is running now!");
4266 			}
4267 		}
4268 		return rte_eth_linkstatus_set(dev, &link);
4269 	}
4270 
4271 	link.link_status = RTE_ETH_LINK_UP;
4272 	link.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
4273 
4274 	switch (link_speed) {
4275 	default:
4276 	case IXGBE_LINK_SPEED_UNKNOWN:
4277 		link.link_speed = RTE_ETH_SPEED_NUM_UNKNOWN;
4278 		break;
4279 
4280 	case IXGBE_LINK_SPEED_10_FULL:
4281 		link.link_speed = RTE_ETH_SPEED_NUM_10M;
4282 		break;
4283 
4284 	case IXGBE_LINK_SPEED_100_FULL:
4285 		link.link_speed = RTE_ETH_SPEED_NUM_100M;
4286 		break;
4287 
4288 	case IXGBE_LINK_SPEED_1GB_FULL:
4289 		link.link_speed = RTE_ETH_SPEED_NUM_1G;
4290 		break;
4291 
4292 	case IXGBE_LINK_SPEED_2_5GB_FULL:
4293 		link.link_speed = RTE_ETH_SPEED_NUM_2_5G;
4294 		break;
4295 
4296 	case IXGBE_LINK_SPEED_5GB_FULL:
4297 		link.link_speed = RTE_ETH_SPEED_NUM_5G;
4298 		break;
4299 
4300 	case IXGBE_LINK_SPEED_10GB_FULL:
4301 		link.link_speed = RTE_ETH_SPEED_NUM_10G;
4302 		break;
4303 	}
4304 
4305 	return rte_eth_linkstatus_set(dev, &link);
4306 }
4307 
4308 static int
4309 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4310 {
4311 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 0);
4312 }
4313 
4314 static int
4315 ixgbevf_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4316 {
4317 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 1);
4318 }
4319 
4320 static int
4321 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
4322 {
4323 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4324 	uint32_t fctrl;
4325 
4326 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4327 	fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4328 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4329 
4330 	return 0;
4331 }
4332 
4333 static int
4334 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
4335 {
4336 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4337 	uint32_t fctrl;
4338 
4339 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4340 	fctrl &= (~IXGBE_FCTRL_UPE);
4341 	if (dev->data->all_multicast == 1)
4342 		fctrl |= IXGBE_FCTRL_MPE;
4343 	else
4344 		fctrl &= (~IXGBE_FCTRL_MPE);
4345 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4346 
4347 	return 0;
4348 }
4349 
4350 static int
4351 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
4352 {
4353 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4354 	uint32_t fctrl;
4355 
4356 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4357 	fctrl |= IXGBE_FCTRL_MPE;
4358 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4359 
4360 	return 0;
4361 }
4362 
4363 static int
4364 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
4365 {
4366 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4367 	uint32_t fctrl;
4368 
4369 	if (dev->data->promiscuous == 1)
4370 		return 0; /* must remain in all_multicast mode */
4371 
4372 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4373 	fctrl &= (~IXGBE_FCTRL_MPE);
4374 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4375 
4376 	return 0;
4377 }
4378 
4379 /**
4380  * It clears the interrupt causes and enables the interrupt.
4381  * It will be called once only during nic initialized.
4382  *
4383  * @param dev
4384  *  Pointer to struct rte_eth_dev.
4385  * @param on
4386  *  Enable or Disable.
4387  *
4388  * @return
4389  *  - On success, zero.
4390  *  - On failure, a negative value.
4391  */
4392 static int
4393 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
4394 {
4395 	struct ixgbe_interrupt *intr =
4396 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4397 
4398 	ixgbe_dev_link_status_print(dev);
4399 	if (on)
4400 		intr->mask |= IXGBE_EICR_LSC;
4401 	else
4402 		intr->mask &= ~IXGBE_EICR_LSC;
4403 
4404 	return 0;
4405 }
4406 
4407 /**
4408  * It clears the interrupt causes and enables the interrupt.
4409  * It will be called once only during nic initialized.
4410  *
4411  * @param dev
4412  *  Pointer to struct rte_eth_dev.
4413  *
4414  * @return
4415  *  - On success, zero.
4416  *  - On failure, a negative value.
4417  */
4418 static int
4419 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
4420 {
4421 	struct ixgbe_interrupt *intr =
4422 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4423 
4424 	intr->mask |= IXGBE_EICR_RTX_QUEUE;
4425 
4426 	return 0;
4427 }
4428 
4429 /**
4430  * It clears the interrupt causes and enables the interrupt.
4431  * It will be called once only during nic initialized.
4432  *
4433  * @param dev
4434  *  Pointer to struct rte_eth_dev.
4435  *
4436  * @return
4437  *  - On success, zero.
4438  *  - On failure, a negative value.
4439  */
4440 static int
4441 ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev)
4442 {
4443 	struct ixgbe_interrupt *intr =
4444 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4445 
4446 	intr->mask |= IXGBE_EICR_LINKSEC;
4447 
4448 	return 0;
4449 }
4450 
4451 /*
4452  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
4453  *
4454  * @param dev
4455  *  Pointer to struct rte_eth_dev.
4456  *
4457  * @return
4458  *  - On success, zero.
4459  *  - On failure, a negative value.
4460  */
4461 static int
4462 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
4463 {
4464 	uint32_t eicr;
4465 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4466 	struct ixgbe_interrupt *intr =
4467 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4468 
4469 	/* clear all cause mask */
4470 	ixgbe_disable_intr(hw);
4471 
4472 	/* read-on-clear nic registers here */
4473 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4474 	PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
4475 
4476 	intr->flags = 0;
4477 
4478 	/* set flag for async link update */
4479 	if (eicr & IXGBE_EICR_LSC)
4480 		intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4481 
4482 	if (eicr & IXGBE_EICR_MAILBOX)
4483 		intr->flags |= IXGBE_FLAG_MAILBOX;
4484 
4485 	if (eicr & IXGBE_EICR_LINKSEC)
4486 		intr->flags |= IXGBE_FLAG_MACSEC;
4487 
4488 	if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
4489 	    hw->phy.type == ixgbe_phy_x550em_ext_t &&
4490 	    (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
4491 		intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
4492 
4493 	return 0;
4494 }
4495 
4496 /**
4497  * It gets and then prints the link status.
4498  *
4499  * @param dev
4500  *  Pointer to struct rte_eth_dev.
4501  *
4502  * @return
4503  *  - On success, zero.
4504  *  - On failure, a negative value.
4505  */
4506 static void
4507 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
4508 {
4509 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4510 	struct rte_eth_link link;
4511 
4512 	rte_eth_linkstatus_get(dev, &link);
4513 
4514 	if (link.link_status) {
4515 		PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
4516 					(int)(dev->data->port_id),
4517 					(unsigned)link.link_speed,
4518 			link.link_duplex == RTE_ETH_LINK_FULL_DUPLEX ?
4519 					"full-duplex" : "half-duplex");
4520 	} else {
4521 		PMD_INIT_LOG(INFO, " Port %d: Link Down",
4522 				(int)(dev->data->port_id));
4523 	}
4524 	PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
4525 				pci_dev->addr.domain,
4526 				pci_dev->addr.bus,
4527 				pci_dev->addr.devid,
4528 				pci_dev->addr.function);
4529 }
4530 
4531 /*
4532  * It executes link_update after knowing an interrupt occurred.
4533  *
4534  * @param dev
4535  *  Pointer to struct rte_eth_dev.
4536  *
4537  * @return
4538  *  - On success, zero.
4539  *  - On failure, a negative value.
4540  */
4541 static int
4542 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
4543 {
4544 	struct ixgbe_interrupt *intr =
4545 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4546 	int64_t timeout;
4547 	struct ixgbe_hw *hw =
4548 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4549 
4550 	PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
4551 
4552 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
4553 		ixgbe_pf_mbx_process(dev);
4554 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
4555 	}
4556 
4557 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4558 		ixgbe_handle_lasi(hw);
4559 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4560 	}
4561 
4562 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4563 		struct rte_eth_link link;
4564 
4565 		/* get the link status before link update, for predicting later */
4566 		rte_eth_linkstatus_get(dev, &link);
4567 
4568 		ixgbe_dev_link_update(dev, 0);
4569 
4570 		/* likely to up */
4571 		if (!link.link_status)
4572 			/* handle it 1 sec later, wait it being stable */
4573 			timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
4574 		/* likely to down */
4575 		else
4576 			/* handle it 4 sec later, wait it being stable */
4577 			timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
4578 
4579 		ixgbe_dev_link_status_print(dev);
4580 		if (rte_eal_alarm_set(timeout * 1000,
4581 				      ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
4582 			PMD_DRV_LOG(ERR, "Error setting alarm");
4583 		else {
4584 			/* remember original mask */
4585 			intr->mask_original = intr->mask;
4586 			/* only disable lsc interrupt */
4587 			intr->mask &= ~IXGBE_EIMS_LSC;
4588 		}
4589 	}
4590 
4591 	PMD_DRV_LOG(DEBUG, "enable intr immediately");
4592 	ixgbe_enable_intr(dev);
4593 
4594 	return 0;
4595 }
4596 
4597 /**
4598  * Interrupt handler which shall be registered for alarm callback for delayed
4599  * handling specific interrupt to wait for the stable nic state. As the
4600  * NIC interrupt state is not stable for ixgbe after link is just down,
4601  * it needs to wait 4 seconds to get the stable status.
4602  *
4603  * @param handle
4604  *  Pointer to interrupt handle.
4605  * @param param
4606  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4607  *
4608  * @return
4609  *  void
4610  */
4611 static void
4612 ixgbe_dev_interrupt_delayed_handler(void *param)
4613 {
4614 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4615 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4616 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
4617 	struct ixgbe_interrupt *intr =
4618 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4619 	struct ixgbe_hw *hw =
4620 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4621 	uint32_t eicr;
4622 
4623 	ixgbe_disable_intr(hw);
4624 
4625 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4626 	if (eicr & IXGBE_EICR_MAILBOX)
4627 		ixgbe_pf_mbx_process(dev);
4628 
4629 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4630 		ixgbe_handle_lasi(hw);
4631 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4632 	}
4633 
4634 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4635 		ixgbe_dev_link_update(dev, 0);
4636 		intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4637 		ixgbe_dev_link_status_print(dev);
4638 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
4639 	}
4640 
4641 	if (intr->flags & IXGBE_FLAG_MACSEC) {
4642 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_MACSEC, NULL);
4643 		intr->flags &= ~IXGBE_FLAG_MACSEC;
4644 	}
4645 
4646 	/* restore original mask */
4647 	intr->mask = intr->mask_original;
4648 	intr->mask_original = 0;
4649 
4650 	PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
4651 	ixgbe_enable_intr(dev);
4652 	rte_intr_ack(intr_handle);
4653 }
4654 
4655 /**
4656  * Interrupt handler triggered by NIC  for handling
4657  * specific interrupt.
4658  *
4659  * @param handle
4660  *  Pointer to interrupt handle.
4661  * @param param
4662  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4663  *
4664  * @return
4665  *  void
4666  */
4667 static void
4668 ixgbe_dev_interrupt_handler(void *param)
4669 {
4670 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4671 
4672 	ixgbe_dev_interrupt_get_status(dev);
4673 	ixgbe_dev_interrupt_action(dev);
4674 }
4675 
4676 static int
4677 ixgbe_dev_led_on(struct rte_eth_dev *dev)
4678 {
4679 	struct ixgbe_hw *hw;
4680 
4681 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4682 	return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4683 }
4684 
4685 static int
4686 ixgbe_dev_led_off(struct rte_eth_dev *dev)
4687 {
4688 	struct ixgbe_hw *hw;
4689 
4690 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4691 	return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4692 }
4693 
4694 static int
4695 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4696 {
4697 	struct ixgbe_hw *hw;
4698 	uint32_t mflcn_reg;
4699 	uint32_t fccfg_reg;
4700 	int rx_pause;
4701 	int tx_pause;
4702 
4703 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4704 
4705 	fc_conf->pause_time = hw->fc.pause_time;
4706 	fc_conf->high_water = hw->fc.high_water[0];
4707 	fc_conf->low_water = hw->fc.low_water[0];
4708 	fc_conf->send_xon = hw->fc.send_xon;
4709 	fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
4710 
4711 	/*
4712 	 * Return rx_pause status according to actual setting of
4713 	 * MFLCN register.
4714 	 */
4715 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4716 	if (mflcn_reg & IXGBE_MFLCN_PMCF)
4717 		fc_conf->mac_ctrl_frame_fwd = 1;
4718 	else
4719 		fc_conf->mac_ctrl_frame_fwd = 0;
4720 
4721 	if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
4722 		rx_pause = 1;
4723 	else
4724 		rx_pause = 0;
4725 
4726 	/*
4727 	 * Return tx_pause status according to actual setting of
4728 	 * FCCFG register.
4729 	 */
4730 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4731 	if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
4732 		tx_pause = 1;
4733 	else
4734 		tx_pause = 0;
4735 
4736 	if (rx_pause && tx_pause)
4737 		fc_conf->mode = RTE_ETH_FC_FULL;
4738 	else if (rx_pause)
4739 		fc_conf->mode = RTE_ETH_FC_RX_PAUSE;
4740 	else if (tx_pause)
4741 		fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
4742 	else
4743 		fc_conf->mode = RTE_ETH_FC_NONE;
4744 
4745 	return 0;
4746 }
4747 
4748 static int
4749 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4750 {
4751 	struct ixgbe_hw *hw;
4752 	struct ixgbe_adapter *adapter = dev->data->dev_private;
4753 	int err;
4754 	uint32_t rx_buf_size;
4755 	uint32_t max_high_water;
4756 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4757 		ixgbe_fc_none,
4758 		ixgbe_fc_rx_pause,
4759 		ixgbe_fc_tx_pause,
4760 		ixgbe_fc_full
4761 	};
4762 
4763 	PMD_INIT_FUNC_TRACE();
4764 
4765 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4766 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
4767 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4768 
4769 	/*
4770 	 * At least reserve one Ethernet frame for watermark
4771 	 * high_water/low_water in kilo bytes for ixgbe
4772 	 */
4773 	max_high_water = (rx_buf_size -
4774 			RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4775 	if ((fc_conf->high_water > max_high_water) ||
4776 		(fc_conf->high_water < fc_conf->low_water)) {
4777 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4778 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4779 		return -EINVAL;
4780 	}
4781 
4782 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
4783 	hw->fc.pause_time     = fc_conf->pause_time;
4784 	hw->fc.high_water[0]  = fc_conf->high_water;
4785 	hw->fc.low_water[0]   = fc_conf->low_water;
4786 	hw->fc.send_xon       = fc_conf->send_xon;
4787 	hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
4788 	adapter->mac_ctrl_frame_fwd = fc_conf->mac_ctrl_frame_fwd;
4789 
4790 	err = ixgbe_flow_ctrl_enable(dev, hw);
4791 	if (err < 0) {
4792 		PMD_INIT_LOG(ERR, "ixgbe_flow_ctrl_enable = 0x%x", err);
4793 		return -EIO;
4794 	}
4795 	return err;
4796 }
4797 
4798 /**
4799  *  ixgbe_pfc_enable_generic - Enable flow control
4800  *  @hw: pointer to hardware structure
4801  *  @tc_num: traffic class number
4802  *  Enable flow control according to the current settings.
4803  */
4804 static int
4805 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
4806 {
4807 	int ret_val = 0;
4808 	uint32_t mflcn_reg, fccfg_reg;
4809 	uint32_t reg;
4810 	uint32_t fcrtl, fcrth;
4811 	uint8_t i;
4812 	uint8_t nb_rx_en;
4813 
4814 	/* Validate the water mark configuration */
4815 	if (!hw->fc.pause_time) {
4816 		ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4817 		goto out;
4818 	}
4819 
4820 	/* Low water mark of zero causes XOFF floods */
4821 	if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
4822 		 /* High/Low water can not be 0 */
4823 		if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_water[tc_num])) {
4824 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4825 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4826 			goto out;
4827 		}
4828 
4829 		if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
4830 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4831 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4832 			goto out;
4833 		}
4834 	}
4835 	/* Negotiate the fc mode to use */
4836 	ixgbe_fc_autoneg(hw);
4837 
4838 	/* Disable any previous flow control settings */
4839 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4840 	mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
4841 
4842 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4843 	fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
4844 
4845 	switch (hw->fc.current_mode) {
4846 	case ixgbe_fc_none:
4847 		/*
4848 		 * If the count of enabled RX Priority Flow control >1,
4849 		 * and the TX pause can not be disabled
4850 		 */
4851 		nb_rx_en = 0;
4852 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4853 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4854 			if (reg & IXGBE_FCRTH_FCEN)
4855 				nb_rx_en++;
4856 		}
4857 		if (nb_rx_en > 1)
4858 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4859 		break;
4860 	case ixgbe_fc_rx_pause:
4861 		/*
4862 		 * Rx Flow control is enabled and Tx Flow control is
4863 		 * disabled by software override. Since there really
4864 		 * isn't a way to advertise that we are capable of RX
4865 		 * Pause ONLY, we will advertise that we support both
4866 		 * symmetric and asymmetric Rx PAUSE.  Later, we will
4867 		 * disable the adapter's ability to send PAUSE frames.
4868 		 */
4869 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4870 		/*
4871 		 * If the count of enabled RX Priority Flow control >1,
4872 		 * and the TX pause can not be disabled
4873 		 */
4874 		nb_rx_en = 0;
4875 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4876 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4877 			if (reg & IXGBE_FCRTH_FCEN)
4878 				nb_rx_en++;
4879 		}
4880 		if (nb_rx_en > 1)
4881 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4882 		break;
4883 	case ixgbe_fc_tx_pause:
4884 		/*
4885 		 * Tx Flow control is enabled, and Rx Flow control is
4886 		 * disabled by software override.
4887 		 */
4888 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4889 		break;
4890 	case ixgbe_fc_full:
4891 		/* Flow control (both Rx and Tx) is enabled by SW override. */
4892 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4893 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4894 		break;
4895 	default:
4896 		PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
4897 		ret_val = IXGBE_ERR_CONFIG;
4898 		goto out;
4899 	}
4900 
4901 	/* Set 802.3x based flow control settings. */
4902 	mflcn_reg |= IXGBE_MFLCN_DPF;
4903 	IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
4904 	IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
4905 
4906 	/* Set up and enable Rx high/low water mark thresholds, enable XON. */
4907 	if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
4908 		hw->fc.high_water[tc_num]) {
4909 		fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
4910 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
4911 		fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
4912 	} else {
4913 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
4914 		/*
4915 		 * In order to prevent Tx hangs when the internal Tx
4916 		 * switch is enabled we must set the high water mark
4917 		 * to the maximum FCRTH value.  This allows the Tx
4918 		 * switch to function even under heavy Rx workloads.
4919 		 */
4920 		fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
4921 	}
4922 	IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
4923 
4924 	/* Configure pause time (2 TCs per register) */
4925 	reg = hw->fc.pause_time * 0x00010001;
4926 	for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
4927 		IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
4928 
4929 	/* Configure flow control refresh threshold value */
4930 	IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
4931 
4932 out:
4933 	return ret_val;
4934 }
4935 
4936 static int
4937 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
4938 {
4939 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4940 	int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
4941 
4942 	if (hw->mac.type != ixgbe_mac_82598EB) {
4943 		ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
4944 	}
4945 	return ret_val;
4946 }
4947 
4948 static int
4949 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
4950 {
4951 	int err;
4952 	uint32_t rx_buf_size;
4953 	uint32_t max_high_water;
4954 	uint8_t tc_num;
4955 	uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
4956 	struct ixgbe_hw *hw =
4957 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4958 	struct ixgbe_dcb_config *dcb_config =
4959 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
4960 
4961 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4962 		ixgbe_fc_none,
4963 		ixgbe_fc_rx_pause,
4964 		ixgbe_fc_tx_pause,
4965 		ixgbe_fc_full
4966 	};
4967 
4968 	PMD_INIT_FUNC_TRACE();
4969 
4970 	ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
4971 	tc_num = map[pfc_conf->priority];
4972 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
4973 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4974 	/*
4975 	 * At least reserve one Ethernet frame for watermark
4976 	 * high_water/low_water in kilo bytes for ixgbe
4977 	 */
4978 	max_high_water = (rx_buf_size -
4979 			RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4980 	if ((pfc_conf->fc.high_water > max_high_water) ||
4981 	    (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
4982 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4983 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4984 		return -EINVAL;
4985 	}
4986 
4987 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
4988 	hw->fc.pause_time = pfc_conf->fc.pause_time;
4989 	hw->fc.send_xon = pfc_conf->fc.send_xon;
4990 	hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
4991 	hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
4992 
4993 	err = ixgbe_dcb_pfc_enable(dev, tc_num);
4994 
4995 	/* Not negotiated is not an error case */
4996 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
4997 		return 0;
4998 
4999 	PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
5000 	return -EIO;
5001 }
5002 
5003 static int
5004 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
5005 			  struct rte_eth_rss_reta_entry64 *reta_conf,
5006 			  uint16_t reta_size)
5007 {
5008 	uint16_t i, sp_reta_size;
5009 	uint8_t j, mask;
5010 	uint32_t reta, r;
5011 	uint16_t idx, shift;
5012 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5013 	struct rte_eth_dev_data *dev_data = dev->data;
5014 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5015 	uint32_t reta_reg;
5016 
5017 	PMD_INIT_FUNC_TRACE();
5018 
5019 	if (!dev_data->dev_started) {
5020 		PMD_DRV_LOG(ERR,
5021 			"port %d must be started before rss reta update",
5022 			 dev_data->port_id);
5023 		return -EIO;
5024 	}
5025 
5026 	if (!ixgbe_rss_update_sp(hw->mac.type)) {
5027 		PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
5028 			"NIC.");
5029 		return -ENOTSUP;
5030 	}
5031 
5032 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5033 	if (reta_size != sp_reta_size) {
5034 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5035 			"(%d) doesn't match the number hardware can supported "
5036 			"(%d)", reta_size, sp_reta_size);
5037 		return -EINVAL;
5038 	}
5039 
5040 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5041 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
5042 		shift = i % RTE_ETH_RETA_GROUP_SIZE;
5043 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5044 						IXGBE_4_BIT_MASK);
5045 		if (!mask)
5046 			continue;
5047 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5048 		if (mask == IXGBE_4_BIT_MASK)
5049 			r = 0;
5050 		else
5051 			r = IXGBE_READ_REG(hw, reta_reg);
5052 		for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5053 			if (mask & (0x1 << j))
5054 				reta |= reta_conf[idx].reta[shift + j] <<
5055 							(CHAR_BIT * j);
5056 			else
5057 				reta |= r & (IXGBE_8_BIT_MASK <<
5058 						(CHAR_BIT * j));
5059 		}
5060 		IXGBE_WRITE_REG(hw, reta_reg, reta);
5061 	}
5062 	adapter->rss_reta_updated = 1;
5063 
5064 	return 0;
5065 }
5066 
5067 static int
5068 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
5069 			 struct rte_eth_rss_reta_entry64 *reta_conf,
5070 			 uint16_t reta_size)
5071 {
5072 	uint16_t i, sp_reta_size;
5073 	uint8_t j, mask;
5074 	uint32_t reta;
5075 	uint16_t idx, shift;
5076 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5077 	uint32_t reta_reg;
5078 
5079 	PMD_INIT_FUNC_TRACE();
5080 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5081 	if (reta_size != sp_reta_size) {
5082 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5083 			"(%d) doesn't match the number hardware can supported "
5084 			"(%d)", reta_size, sp_reta_size);
5085 		return -EINVAL;
5086 	}
5087 
5088 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5089 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
5090 		shift = i % RTE_ETH_RETA_GROUP_SIZE;
5091 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5092 						IXGBE_4_BIT_MASK);
5093 		if (!mask)
5094 			continue;
5095 
5096 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5097 		reta = IXGBE_READ_REG(hw, reta_reg);
5098 		for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5099 			if (mask & (0x1 << j))
5100 				reta_conf[idx].reta[shift + j] =
5101 					((reta >> (CHAR_BIT * j)) &
5102 						IXGBE_8_BIT_MASK);
5103 		}
5104 	}
5105 
5106 	return 0;
5107 }
5108 
5109 static int
5110 ixgbe_add_rar(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
5111 				uint32_t index, uint32_t pool)
5112 {
5113 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5114 	uint32_t enable_addr = 1;
5115 
5116 	return ixgbe_set_rar(hw, index, mac_addr->addr_bytes,
5117 			     pool, enable_addr);
5118 }
5119 
5120 static void
5121 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
5122 {
5123 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5124 
5125 	ixgbe_clear_rar(hw, index);
5126 }
5127 
5128 static int
5129 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
5130 {
5131 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5132 
5133 	ixgbe_remove_rar(dev, 0);
5134 	ixgbe_add_rar(dev, addr, 0, pci_dev->max_vfs);
5135 
5136 	return 0;
5137 }
5138 
5139 static bool
5140 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
5141 {
5142 	if (strcmp(dev->device->driver->name, drv->driver.name))
5143 		return false;
5144 
5145 	return true;
5146 }
5147 
5148 bool
5149 is_ixgbe_supported(struct rte_eth_dev *dev)
5150 {
5151 	return is_device_supported(dev, &rte_ixgbe_pmd);
5152 }
5153 
5154 static int
5155 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
5156 {
5157 	uint32_t hlreg0;
5158 	uint32_t maxfrs;
5159 	struct ixgbe_hw *hw;
5160 	struct rte_eth_dev_info dev_info;
5161 	uint32_t frame_size = mtu + IXGBE_ETH_OVERHEAD;
5162 	int ret;
5163 
5164 	ret = ixgbe_dev_info_get(dev, &dev_info);
5165 	if (ret != 0)
5166 		return ret;
5167 
5168 	/* check that mtu is within the allowed range */
5169 	if (mtu < RTE_ETHER_MIN_MTU || frame_size > dev_info.max_rx_pktlen)
5170 		return -EINVAL;
5171 
5172 	/* If device is started, refuse mtu that requires the support of
5173 	 * scattered packets when this feature has not been enabled before.
5174 	 */
5175 	if (dev->data->dev_started && !dev->data->scattered_rx &&
5176 	    frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
5177 			dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
5178 		PMD_INIT_LOG(ERR, "Stop port first.");
5179 		return -EINVAL;
5180 	}
5181 
5182 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5183 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
5184 
5185 	/* switch to jumbo mode if needed */
5186 	if (mtu > RTE_ETHER_MTU)
5187 		hlreg0 |= IXGBE_HLREG0_JUMBOEN;
5188 	else
5189 		hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
5190 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
5191 
5192 	maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
5193 	maxfrs &= 0x0000FFFF;
5194 	maxfrs |= (frame_size << 16);
5195 	IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
5196 
5197 	return 0;
5198 }
5199 
5200 /*
5201  * Virtual Function operations
5202  */
5203 static void
5204 ixgbevf_intr_disable(struct rte_eth_dev *dev)
5205 {
5206 	struct ixgbe_interrupt *intr =
5207 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5208 	struct ixgbe_hw *hw =
5209 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5210 
5211 	PMD_INIT_FUNC_TRACE();
5212 
5213 	/* Clear interrupt mask to stop from interrupts being generated */
5214 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
5215 
5216 	IXGBE_WRITE_FLUSH(hw);
5217 
5218 	/* Clear mask value. */
5219 	intr->mask = 0;
5220 }
5221 
5222 static void
5223 ixgbevf_intr_enable(struct rte_eth_dev *dev)
5224 {
5225 	struct ixgbe_interrupt *intr =
5226 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5227 	struct ixgbe_hw *hw =
5228 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5229 
5230 	PMD_INIT_FUNC_TRACE();
5231 
5232 	/* VF enable interrupt autoclean */
5233 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
5234 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
5235 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
5236 
5237 	IXGBE_WRITE_FLUSH(hw);
5238 
5239 	/* Save IXGBE_VTEIMS value to mask. */
5240 	intr->mask = IXGBE_VF_IRQ_ENABLE_MASK;
5241 }
5242 
5243 static int
5244 ixgbevf_dev_configure(struct rte_eth_dev *dev)
5245 {
5246 	struct rte_eth_conf *conf = &dev->data->dev_conf;
5247 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5248 
5249 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
5250 		     dev->data->port_id);
5251 
5252 	if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
5253 		dev->data->dev_conf.rxmode.offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
5254 
5255 	/*
5256 	 * VF has no ability to enable/disable HW CRC
5257 	 * Keep the persistent behavior the same as Host PF
5258 	 */
5259 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
5260 	if (conf->rxmode.offloads & RTE_ETH_RX_OFFLOAD_KEEP_CRC) {
5261 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
5262 		conf->rxmode.offloads &= ~RTE_ETH_RX_OFFLOAD_KEEP_CRC;
5263 	}
5264 #else
5265 	if (!(conf->rxmode.offloads & RTE_ETH_RX_OFFLOAD_KEEP_CRC)) {
5266 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
5267 		conf->rxmode.offloads |= RTE_ETH_RX_OFFLOAD_KEEP_CRC;
5268 	}
5269 #endif
5270 
5271 	/*
5272 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
5273 	 * allocation or vector Rx preconditions we will reset it.
5274 	 */
5275 	adapter->rx_bulk_alloc_allowed = true;
5276 	adapter->rx_vec_allowed = true;
5277 
5278 	return 0;
5279 }
5280 
5281 static int
5282 ixgbevf_dev_start(struct rte_eth_dev *dev)
5283 {
5284 	struct ixgbe_hw *hw =
5285 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5286 	uint32_t intr_vector = 0;
5287 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5288 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5289 
5290 	int err, mask = 0;
5291 
5292 	PMD_INIT_FUNC_TRACE();
5293 
5294 	/* Stop the link setup handler before resetting the HW. */
5295 	ixgbe_dev_wait_setup_link_complete(dev, 0);
5296 
5297 	err = hw->mac.ops.reset_hw(hw);
5298 
5299 	/**
5300 	 * In this case, reuses the MAC address assigned by VF
5301 	 * initialization.
5302 	 */
5303 	if (err != IXGBE_SUCCESS && err != IXGBE_ERR_INVALID_MAC_ADDR) {
5304 		PMD_INIT_LOG(ERR, "Unable to reset vf hardware (%d)", err);
5305 		return err;
5306 	}
5307 
5308 	hw->mac.get_link_status = true;
5309 
5310 	/* negotiate mailbox API version to use with the PF. */
5311 	ixgbevf_negotiate_api(hw);
5312 
5313 	ixgbevf_dev_tx_init(dev);
5314 
5315 	/* This can fail when allocating mbufs for descriptor rings */
5316 	err = ixgbevf_dev_rx_init(dev);
5317 	if (err) {
5318 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
5319 		ixgbe_dev_clear_queues(dev);
5320 		return err;
5321 	}
5322 
5323 	/* Set vfta */
5324 	ixgbevf_set_vfta_all(dev, 1);
5325 
5326 	/* Set HW strip */
5327 	mask = RTE_ETH_VLAN_STRIP_MASK | RTE_ETH_VLAN_FILTER_MASK |
5328 		RTE_ETH_VLAN_EXTEND_MASK;
5329 	err = ixgbevf_vlan_offload_config(dev, mask);
5330 	if (err) {
5331 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
5332 		ixgbe_dev_clear_queues(dev);
5333 		return err;
5334 	}
5335 
5336 	ixgbevf_dev_rxtx_start(dev);
5337 
5338 	/* check and configure queue intr-vector mapping */
5339 	if (rte_intr_cap_multiple(intr_handle) &&
5340 	    dev->data->dev_conf.intr_conf.rxq) {
5341 		/* According to datasheet, only vector 0/1/2 can be used,
5342 		 * now only one vector is used for Rx queue
5343 		 */
5344 		intr_vector = 1;
5345 		if (rte_intr_efd_enable(intr_handle, intr_vector)) {
5346 			ixgbe_dev_clear_queues(dev);
5347 			return -1;
5348 		}
5349 	}
5350 
5351 	if (rte_intr_dp_is_en(intr_handle)) {
5352 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
5353 						   dev->data->nb_rx_queues)) {
5354 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
5355 				     " intr_vec", dev->data->nb_rx_queues);
5356 			ixgbe_dev_clear_queues(dev);
5357 			return -ENOMEM;
5358 		}
5359 	}
5360 	ixgbevf_configure_msix(dev);
5361 
5362 	/* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
5363 	 * is mapped to VFIO vector 0 in eth_ixgbevf_dev_init( ).
5364 	 * If previous VFIO interrupt mapping setting in eth_ixgbevf_dev_init( )
5365 	 * is not cleared, it will fail when following rte_intr_enable( ) tries
5366 	 * to map Rx queue interrupt to other VFIO vectors.
5367 	 * So clear uio/vfio intr/evevnfd first to avoid failure.
5368 	 */
5369 	rte_intr_disable(intr_handle);
5370 
5371 	rte_intr_enable(intr_handle);
5372 
5373 	/* Re-enable interrupt for VF */
5374 	ixgbevf_intr_enable(dev);
5375 
5376 	/*
5377 	 * Update link status right before return, because it may
5378 	 * start link configuration process in a separate thread.
5379 	 */
5380 	ixgbevf_dev_link_update(dev, 0);
5381 
5382 	hw->adapter_stopped = false;
5383 
5384 	return 0;
5385 }
5386 
5387 static int
5388 ixgbevf_dev_stop(struct rte_eth_dev *dev)
5389 {
5390 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5391 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5392 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5393 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5394 
5395 	if (hw->adapter_stopped)
5396 		return 0;
5397 
5398 	PMD_INIT_FUNC_TRACE();
5399 
5400 	ixgbe_dev_wait_setup_link_complete(dev, 0);
5401 
5402 	ixgbevf_intr_disable(dev);
5403 
5404 	dev->data->dev_started = 0;
5405 	hw->adapter_stopped = 1;
5406 	ixgbe_stop_adapter(hw);
5407 
5408 	/*
5409 	  * Clear what we set, but we still keep shadow_vfta to
5410 	  * restore after device starts
5411 	  */
5412 	ixgbevf_set_vfta_all(dev, 0);
5413 
5414 	/* Clear stored conf */
5415 	dev->data->scattered_rx = 0;
5416 
5417 	ixgbe_dev_clear_queues(dev);
5418 
5419 	/* Clean datapath event and queue/vec mapping */
5420 	rte_intr_efd_disable(intr_handle);
5421 	rte_intr_vec_list_free(intr_handle);
5422 
5423 	adapter->rss_reta_updated = 0;
5424 
5425 	return 0;
5426 }
5427 
5428 static int
5429 ixgbevf_dev_close(struct rte_eth_dev *dev)
5430 {
5431 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5432 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5433 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5434 	int ret;
5435 
5436 	PMD_INIT_FUNC_TRACE();
5437 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5438 		return 0;
5439 
5440 	ixgbe_reset_hw(hw);
5441 
5442 	ret = ixgbevf_dev_stop(dev);
5443 
5444 	ixgbe_dev_free_queues(dev);
5445 
5446 	/**
5447 	 * Remove the VF MAC address ro ensure
5448 	 * that the VF traffic goes to the PF
5449 	 * after stop, close and detach of the VF
5450 	 **/
5451 	ixgbevf_remove_mac_addr(dev, 0);
5452 
5453 	rte_intr_disable(intr_handle);
5454 	rte_intr_callback_unregister(intr_handle,
5455 				     ixgbevf_dev_interrupt_handler, dev);
5456 
5457 	return ret;
5458 }
5459 
5460 /*
5461  * Reset VF device
5462  */
5463 static int
5464 ixgbevf_dev_reset(struct rte_eth_dev *dev)
5465 {
5466 	int ret;
5467 
5468 	ret = eth_ixgbevf_dev_uninit(dev);
5469 	if (ret)
5470 		return ret;
5471 
5472 	ret = eth_ixgbevf_dev_init(dev);
5473 
5474 	return ret;
5475 }
5476 
5477 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
5478 {
5479 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5480 	struct ixgbe_vfta *shadow_vfta =
5481 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5482 	int i = 0, j = 0, vfta = 0, mask = 1;
5483 
5484 	for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
5485 		vfta = shadow_vfta->vfta[i];
5486 		if (vfta) {
5487 			mask = 1;
5488 			for (j = 0; j < 32; j++) {
5489 				if (vfta & mask)
5490 					ixgbe_set_vfta(hw, (i<<5)+j, 0,
5491 						       on, false);
5492 				mask <<= 1;
5493 			}
5494 		}
5495 	}
5496 
5497 }
5498 
5499 static int
5500 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
5501 {
5502 	struct ixgbe_hw *hw =
5503 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5504 	struct ixgbe_vfta *shadow_vfta =
5505 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5506 	uint32_t vid_idx = 0;
5507 	uint32_t vid_bit = 0;
5508 	int ret = 0;
5509 
5510 	PMD_INIT_FUNC_TRACE();
5511 
5512 	/* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
5513 	ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
5514 	if (ret) {
5515 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
5516 		return ret;
5517 	}
5518 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
5519 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
5520 
5521 	/* Save what we set and retore it after device reset */
5522 	if (on)
5523 		shadow_vfta->vfta[vid_idx] |= vid_bit;
5524 	else
5525 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
5526 
5527 	return 0;
5528 }
5529 
5530 static void
5531 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
5532 {
5533 	struct ixgbe_hw *hw =
5534 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5535 	uint32_t ctrl;
5536 
5537 	PMD_INIT_FUNC_TRACE();
5538 
5539 	if (queue >= hw->mac.max_rx_queues)
5540 		return;
5541 
5542 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
5543 	if (on)
5544 		ctrl |= IXGBE_RXDCTL_VME;
5545 	else
5546 		ctrl &= ~IXGBE_RXDCTL_VME;
5547 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
5548 
5549 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
5550 }
5551 
5552 static int
5553 ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask)
5554 {
5555 	struct ixgbe_rx_queue *rxq;
5556 	uint16_t i;
5557 	int on = 0;
5558 
5559 	/* VF function only support hw strip feature, others are not support */
5560 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
5561 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
5562 			rxq = dev->data->rx_queues[i];
5563 			on = !!(rxq->offloads &	RTE_ETH_RX_OFFLOAD_VLAN_STRIP);
5564 			ixgbevf_vlan_strip_queue_set(dev, i, on);
5565 		}
5566 	}
5567 
5568 	return 0;
5569 }
5570 
5571 static int
5572 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
5573 {
5574 	ixgbe_config_vlan_strip_on_all_queues(dev, mask);
5575 
5576 	ixgbevf_vlan_offload_config(dev, mask);
5577 
5578 	return 0;
5579 }
5580 
5581 int
5582 ixgbe_vt_check(struct ixgbe_hw *hw)
5583 {
5584 	uint32_t reg_val;
5585 
5586 	/* if Virtualization Technology is enabled */
5587 	reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
5588 	if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
5589 		PMD_INIT_LOG(ERR, "VT must be enabled for this setting");
5590 		return -1;
5591 	}
5592 
5593 	return 0;
5594 }
5595 
5596 static uint32_t
5597 ixgbe_uta_vector(struct ixgbe_hw *hw, struct rte_ether_addr *uc_addr)
5598 {
5599 	uint32_t vector = 0;
5600 
5601 	switch (hw->mac.mc_filter_type) {
5602 	case 0:   /* use bits [47:36] of the address */
5603 		vector = ((uc_addr->addr_bytes[4] >> 4) |
5604 			(((uint16_t)uc_addr->addr_bytes[5]) << 4));
5605 		break;
5606 	case 1:   /* use bits [46:35] of the address */
5607 		vector = ((uc_addr->addr_bytes[4] >> 3) |
5608 			(((uint16_t)uc_addr->addr_bytes[5]) << 5));
5609 		break;
5610 	case 2:   /* use bits [45:34] of the address */
5611 		vector = ((uc_addr->addr_bytes[4] >> 2) |
5612 			(((uint16_t)uc_addr->addr_bytes[5]) << 6));
5613 		break;
5614 	case 3:   /* use bits [43:32] of the address */
5615 		vector = ((uc_addr->addr_bytes[4]) |
5616 			(((uint16_t)uc_addr->addr_bytes[5]) << 8));
5617 		break;
5618 	default:  /* Invalid mc_filter_type */
5619 		break;
5620 	}
5621 
5622 	/* vector can only be 12-bits or boundary will be exceeded */
5623 	vector &= 0xFFF;
5624 	return vector;
5625 }
5626 
5627 static int
5628 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev,
5629 			struct rte_ether_addr *mac_addr, uint8_t on)
5630 {
5631 	uint32_t vector;
5632 	uint32_t uta_idx;
5633 	uint32_t reg_val;
5634 	uint32_t uta_shift;
5635 	uint32_t rc;
5636 	const uint32_t ixgbe_uta_idx_mask = 0x7F;
5637 	const uint32_t ixgbe_uta_bit_shift = 5;
5638 	const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
5639 	const uint32_t bit1 = 0x1;
5640 
5641 	struct ixgbe_hw *hw =
5642 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5643 	struct ixgbe_uta_info *uta_info =
5644 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5645 
5646 	/* The UTA table only exists on 82599 hardware and newer */
5647 	if (hw->mac.type < ixgbe_mac_82599EB)
5648 		return -ENOTSUP;
5649 
5650 	vector = ixgbe_uta_vector(hw, mac_addr);
5651 	uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
5652 	uta_shift = vector & ixgbe_uta_bit_mask;
5653 
5654 	rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
5655 	if (rc == on)
5656 		return 0;
5657 
5658 	reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
5659 	if (on) {
5660 		uta_info->uta_in_use++;
5661 		reg_val |= (bit1 << uta_shift);
5662 		uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
5663 	} else {
5664 		uta_info->uta_in_use--;
5665 		reg_val &= ~(bit1 << uta_shift);
5666 		uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
5667 	}
5668 
5669 	IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
5670 
5671 	if (uta_info->uta_in_use > 0)
5672 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
5673 				IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
5674 	else
5675 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
5676 
5677 	return 0;
5678 }
5679 
5680 static int
5681 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
5682 {
5683 	int i;
5684 	struct ixgbe_hw *hw =
5685 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5686 	struct ixgbe_uta_info *uta_info =
5687 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5688 
5689 	/* The UTA table only exists on 82599 hardware and newer */
5690 	if (hw->mac.type < ixgbe_mac_82599EB)
5691 		return -ENOTSUP;
5692 
5693 	if (on) {
5694 		for (i = 0; i < RTE_ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5695 			uta_info->uta_shadow[i] = ~0;
5696 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
5697 		}
5698 	} else {
5699 		for (i = 0; i < RTE_ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5700 			uta_info->uta_shadow[i] = 0;
5701 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
5702 		}
5703 	}
5704 	return 0;
5705 
5706 }
5707 
5708 uint32_t
5709 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
5710 {
5711 	uint32_t new_val = orig_val;
5712 
5713 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_UNTAG)
5714 		new_val |= IXGBE_VMOLR_AUPE;
5715 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_HASH_MC)
5716 		new_val |= IXGBE_VMOLR_ROMPE;
5717 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_HASH_UC)
5718 		new_val |= IXGBE_VMOLR_ROPE;
5719 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_BROADCAST)
5720 		new_val |= IXGBE_VMOLR_BAM;
5721 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_MULTICAST)
5722 		new_val |= IXGBE_VMOLR_MPE;
5723 
5724 	return new_val;
5725 }
5726 
5727 static int
5728 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5729 {
5730 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5731 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5732 	struct ixgbe_interrupt *intr =
5733 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5734 	struct ixgbe_hw *hw =
5735 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5736 	uint32_t vec = IXGBE_MISC_VEC_ID;
5737 
5738 	if (rte_intr_allow_others(intr_handle))
5739 		vec = IXGBE_RX_VEC_START;
5740 	intr->mask |= (1 << vec);
5741 	RTE_SET_USED(queue_id);
5742 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5743 
5744 	rte_intr_ack(intr_handle);
5745 
5746 	return 0;
5747 }
5748 
5749 static int
5750 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5751 {
5752 	struct ixgbe_interrupt *intr =
5753 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5754 	struct ixgbe_hw *hw =
5755 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5756 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5757 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5758 	uint32_t vec = IXGBE_MISC_VEC_ID;
5759 
5760 	if (rte_intr_allow_others(intr_handle))
5761 		vec = IXGBE_RX_VEC_START;
5762 	intr->mask &= ~(1 << vec);
5763 	RTE_SET_USED(queue_id);
5764 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5765 
5766 	return 0;
5767 }
5768 
5769 static int
5770 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5771 {
5772 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5773 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5774 	uint32_t mask;
5775 	struct ixgbe_hw *hw =
5776 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5777 	struct ixgbe_interrupt *intr =
5778 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5779 
5780 	if (queue_id < 16) {
5781 		ixgbe_disable_intr(hw);
5782 		intr->mask |= (1 << queue_id);
5783 		ixgbe_enable_intr(dev);
5784 	} else if (queue_id < 32) {
5785 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5786 		mask &= (1 << queue_id);
5787 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5788 	} else if (queue_id < 64) {
5789 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5790 		mask &= (1 << (queue_id - 32));
5791 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5792 	}
5793 	rte_intr_ack(intr_handle);
5794 
5795 	return 0;
5796 }
5797 
5798 static int
5799 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5800 {
5801 	uint32_t mask;
5802 	struct ixgbe_hw *hw =
5803 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5804 	struct ixgbe_interrupt *intr =
5805 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5806 
5807 	if (queue_id < 16) {
5808 		ixgbe_disable_intr(hw);
5809 		intr->mask &= ~(1 << queue_id);
5810 		ixgbe_enable_intr(dev);
5811 	} else if (queue_id < 32) {
5812 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5813 		mask &= ~(1 << queue_id);
5814 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5815 	} else if (queue_id < 64) {
5816 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5817 		mask &= ~(1 << (queue_id - 32));
5818 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5819 	}
5820 
5821 	return 0;
5822 }
5823 
5824 static void
5825 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5826 		     uint8_t queue, uint8_t msix_vector)
5827 {
5828 	uint32_t tmp, idx;
5829 
5830 	if (direction == -1) {
5831 		/* other causes */
5832 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5833 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
5834 		tmp &= ~0xFF;
5835 		tmp |= msix_vector;
5836 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
5837 	} else {
5838 		/* rx or tx cause */
5839 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5840 		idx = ((16 * (queue & 1)) + (8 * direction));
5841 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
5842 		tmp &= ~(0xFF << idx);
5843 		tmp |= (msix_vector << idx);
5844 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
5845 	}
5846 }
5847 
5848 /**
5849  * set the IVAR registers, mapping interrupt causes to vectors
5850  * @param hw
5851  *  pointer to ixgbe_hw struct
5852  * @direction
5853  *  0 for Rx, 1 for Tx, -1 for other causes
5854  * @queue
5855  *  queue to map the corresponding interrupt to
5856  * @msix_vector
5857  *  the vector to map to the corresponding queue
5858  */
5859 static void
5860 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5861 		   uint8_t queue, uint8_t msix_vector)
5862 {
5863 	uint32_t tmp, idx;
5864 
5865 	msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5866 	if (hw->mac.type == ixgbe_mac_82598EB) {
5867 		if (direction == -1)
5868 			direction = 0;
5869 		idx = (((direction * 64) + queue) >> 2) & 0x1F;
5870 		tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
5871 		tmp &= ~(0xFF << (8 * (queue & 0x3)));
5872 		tmp |= (msix_vector << (8 * (queue & 0x3)));
5873 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
5874 	} else if ((hw->mac.type == ixgbe_mac_82599EB) ||
5875 			(hw->mac.type == ixgbe_mac_X540) ||
5876 			(hw->mac.type == ixgbe_mac_X550) ||
5877 			(hw->mac.type == ixgbe_mac_X550EM_x)) {
5878 		if (direction == -1) {
5879 			/* other causes */
5880 			idx = ((queue & 1) * 8);
5881 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
5882 			tmp &= ~(0xFF << idx);
5883 			tmp |= (msix_vector << idx);
5884 			IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
5885 		} else {
5886 			/* rx or tx causes */
5887 			idx = ((16 * (queue & 1)) + (8 * direction));
5888 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
5889 			tmp &= ~(0xFF << idx);
5890 			tmp |= (msix_vector << idx);
5891 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
5892 		}
5893 	}
5894 }
5895 
5896 static void
5897 ixgbevf_configure_msix(struct rte_eth_dev *dev)
5898 {
5899 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5900 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5901 	struct ixgbe_hw *hw =
5902 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5903 	uint32_t q_idx;
5904 	uint32_t vector_idx = IXGBE_MISC_VEC_ID;
5905 	uint32_t base = IXGBE_MISC_VEC_ID;
5906 
5907 	/* Configure VF other cause ivar */
5908 	ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
5909 
5910 	/* won't configure msix register if no mapping is done
5911 	 * between intr vector and event fd.
5912 	 */
5913 	if (!rte_intr_dp_is_en(intr_handle))
5914 		return;
5915 
5916 	if (rte_intr_allow_others(intr_handle)) {
5917 		base = IXGBE_RX_VEC_START;
5918 		vector_idx = IXGBE_RX_VEC_START;
5919 	}
5920 
5921 	/* Configure all RX queues of VF */
5922 	for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
5923 		/* Force all queue use vector 0,
5924 		 * as IXGBE_VF_MAXMSIVECOTR = 1
5925 		 */
5926 		ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
5927 		rte_intr_vec_list_index_set(intr_handle, q_idx,
5928 						   vector_idx);
5929 		if (vector_idx < base + rte_intr_nb_efd_get(intr_handle)
5930 		    - 1)
5931 			vector_idx++;
5932 	}
5933 
5934 	/* As RX queue setting above show, all queues use the vector 0.
5935 	 * Set only the ITR value of IXGBE_MISC_VEC_ID.
5936 	 */
5937 	IXGBE_WRITE_REG(hw, IXGBE_VTEITR(IXGBE_MISC_VEC_ID),
5938 			IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
5939 			| IXGBE_EITR_CNT_WDIS);
5940 }
5941 
5942 /**
5943  * Sets up the hardware to properly generate MSI-X interrupts
5944  * @hw
5945  *  board private structure
5946  */
5947 static void
5948 ixgbe_configure_msix(struct rte_eth_dev *dev)
5949 {
5950 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5951 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5952 	struct ixgbe_hw *hw =
5953 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5954 	uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
5955 	uint32_t vec = IXGBE_MISC_VEC_ID;
5956 	uint32_t mask;
5957 	uint32_t gpie;
5958 
5959 	/* won't configure msix register if no mapping is done
5960 	 * between intr vector and event fd
5961 	 * but if misx has been enabled already, need to configure
5962 	 * auto clean, auto mask and throttling.
5963 	 */
5964 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5965 	if (!rte_intr_dp_is_en(intr_handle) &&
5966 	    !(gpie & (IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT)))
5967 		return;
5968 
5969 	if (rte_intr_allow_others(intr_handle))
5970 		vec = base = IXGBE_RX_VEC_START;
5971 
5972 	/* setup GPIE for MSI-x mode */
5973 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5974 	gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5975 		IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
5976 	/* auto clearing and auto setting corresponding bits in EIMS
5977 	 * when MSI-X interrupt is triggered
5978 	 */
5979 	if (hw->mac.type == ixgbe_mac_82598EB) {
5980 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5981 	} else {
5982 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5983 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5984 	}
5985 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5986 
5987 	/* Populate the IVAR table and set the ITR values to the
5988 	 * corresponding register.
5989 	 */
5990 	if (rte_intr_dp_is_en(intr_handle)) {
5991 		for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
5992 			queue_id++) {
5993 			/* by default, 1:1 mapping */
5994 			ixgbe_set_ivar_map(hw, 0, queue_id, vec);
5995 			rte_intr_vec_list_index_set(intr_handle,
5996 							   queue_id, vec);
5997 			if (vec < base + rte_intr_nb_efd_get(intr_handle)
5998 			    - 1)
5999 				vec++;
6000 		}
6001 
6002 		switch (hw->mac.type) {
6003 		case ixgbe_mac_82598EB:
6004 			ixgbe_set_ivar_map(hw, -1,
6005 					   IXGBE_IVAR_OTHER_CAUSES_INDEX,
6006 					   IXGBE_MISC_VEC_ID);
6007 			break;
6008 		case ixgbe_mac_82599EB:
6009 		case ixgbe_mac_X540:
6010 		case ixgbe_mac_X550:
6011 		case ixgbe_mac_X550EM_x:
6012 			ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
6013 			break;
6014 		default:
6015 			break;
6016 		}
6017 	}
6018 	IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
6019 			IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
6020 			| IXGBE_EITR_CNT_WDIS);
6021 
6022 	/* set up to autoclear timer, and the vectors */
6023 	mask = IXGBE_EIMS_ENABLE_MASK;
6024 	mask &= ~(IXGBE_EIMS_OTHER |
6025 		  IXGBE_EIMS_MAILBOX |
6026 		  IXGBE_EIMS_LSC);
6027 
6028 	IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
6029 }
6030 
6031 int
6032 ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
6033 			   uint16_t queue_idx, uint16_t tx_rate)
6034 {
6035 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6036 	uint32_t rf_dec, rf_int;
6037 	uint32_t bcnrc_val;
6038 	uint16_t link_speed = dev->data->dev_link.link_speed;
6039 
6040 	if (queue_idx >= hw->mac.max_tx_queues)
6041 		return -EINVAL;
6042 
6043 	if (tx_rate != 0) {
6044 		/* Calculate the rate factor values to set */
6045 		rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
6046 		rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
6047 		rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
6048 
6049 		bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
6050 		bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
6051 				IXGBE_RTTBCNRC_RF_INT_MASK_M);
6052 		bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
6053 	} else {
6054 		bcnrc_val = 0;
6055 	}
6056 
6057 	/*
6058 	 * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
6059 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
6060 	 * set as 0x4.
6061 	 */
6062 	if (dev->data->mtu + IXGBE_ETH_OVERHEAD >= IXGBE_MAX_JUMBO_FRAME_SIZE)
6063 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM, IXGBE_MMW_SIZE_JUMBO_FRAME);
6064 	else
6065 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM, IXGBE_MMW_SIZE_DEFAULT);
6066 
6067 	/* Set RTTBCNRC of queue X */
6068 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
6069 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
6070 	IXGBE_WRITE_FLUSH(hw);
6071 
6072 	return 0;
6073 }
6074 
6075 static int
6076 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
6077 		     __rte_unused uint32_t index,
6078 		     __rte_unused uint32_t pool)
6079 {
6080 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6081 	int diag;
6082 
6083 	/*
6084 	 * On a 82599 VF, adding again the same MAC addr is not an idempotent
6085 	 * operation. Trap this case to avoid exhausting the [very limited]
6086 	 * set of PF resources used to store VF MAC addresses.
6087 	 */
6088 	if (memcmp(hw->mac.perm_addr, mac_addr,
6089 			sizeof(struct rte_ether_addr)) == 0)
6090 		return -1;
6091 	diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6092 	if (diag != 0)
6093 		PMD_DRV_LOG(ERR, "Unable to add MAC address "
6094 			    RTE_ETHER_ADDR_PRT_FMT " - diag=%d",
6095 			    RTE_ETHER_ADDR_BYTES(mac_addr), diag);
6096 	return diag;
6097 }
6098 
6099 static void
6100 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
6101 {
6102 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6103 	struct rte_ether_addr *perm_addr =
6104 		(struct rte_ether_addr *)hw->mac.perm_addr;
6105 	struct rte_ether_addr *mac_addr;
6106 	uint32_t i;
6107 	int diag;
6108 
6109 	/*
6110 	 * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
6111 	 * not support the deletion of a given MAC address.
6112 	 * Instead, it imposes to delete all MAC addresses, then to add again
6113 	 * all MAC addresses with the exception of the one to be deleted.
6114 	 */
6115 	(void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
6116 
6117 	/*
6118 	 * Add again all MAC addresses, with the exception of the deleted one
6119 	 * and of the permanent MAC address.
6120 	 */
6121 	for (i = 0, mac_addr = dev->data->mac_addrs;
6122 	     i < hw->mac.num_rar_entries; i++, mac_addr++) {
6123 		/* Skip the deleted MAC address */
6124 		if (i == index)
6125 			continue;
6126 		/* Skip NULL MAC addresses */
6127 		if (rte_is_zero_ether_addr(mac_addr))
6128 			continue;
6129 		/* Skip the permanent MAC address */
6130 		if (memcmp(perm_addr, mac_addr,
6131 				sizeof(struct rte_ether_addr)) == 0)
6132 			continue;
6133 		diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6134 		if (diag != 0)
6135 			PMD_DRV_LOG(ERR,
6136 				    "Adding again MAC address "
6137 				     RTE_ETHER_ADDR_PRT_FMT " failed "
6138 				    "diag=%d", RTE_ETHER_ADDR_BYTES(mac_addr),
6139 				    diag);
6140 	}
6141 }
6142 
6143 static int
6144 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
6145 			struct rte_ether_addr *addr)
6146 {
6147 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6148 
6149 	hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
6150 
6151 	return 0;
6152 }
6153 
6154 int
6155 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
6156 			struct rte_eth_syn_filter *filter,
6157 			bool add)
6158 {
6159 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6160 	struct ixgbe_filter_info *filter_info =
6161 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6162 	uint32_t syn_info;
6163 	uint32_t synqf;
6164 
6165 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6166 		return -EINVAL;
6167 
6168 	syn_info = filter_info->syn_info;
6169 
6170 	if (add) {
6171 		if (syn_info & IXGBE_SYN_FILTER_ENABLE)
6172 			return -EINVAL;
6173 		synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
6174 			IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
6175 
6176 		if (filter->hig_pri)
6177 			synqf |= IXGBE_SYN_FILTER_SYNQFP;
6178 		else
6179 			synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
6180 	} else {
6181 		synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6182 		if (!(syn_info & IXGBE_SYN_FILTER_ENABLE))
6183 			return -ENOENT;
6184 		synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
6185 	}
6186 
6187 	filter_info->syn_info = synqf;
6188 	IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
6189 	IXGBE_WRITE_FLUSH(hw);
6190 	return 0;
6191 }
6192 
6193 
6194 static inline enum ixgbe_5tuple_protocol
6195 convert_protocol_type(uint8_t protocol_value)
6196 {
6197 	if (protocol_value == IPPROTO_TCP)
6198 		return IXGBE_FILTER_PROTOCOL_TCP;
6199 	else if (protocol_value == IPPROTO_UDP)
6200 		return IXGBE_FILTER_PROTOCOL_UDP;
6201 	else if (protocol_value == IPPROTO_SCTP)
6202 		return IXGBE_FILTER_PROTOCOL_SCTP;
6203 	else
6204 		return IXGBE_FILTER_PROTOCOL_NONE;
6205 }
6206 
6207 /* inject a 5-tuple filter to HW */
6208 static inline void
6209 ixgbe_inject_5tuple_filter(struct rte_eth_dev *dev,
6210 			   struct ixgbe_5tuple_filter *filter)
6211 {
6212 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6213 	int i;
6214 	uint32_t ftqf, sdpqf;
6215 	uint32_t l34timir = 0;
6216 	uint8_t mask = 0xff;
6217 
6218 	i = filter->index;
6219 
6220 	sdpqf = (uint32_t)(filter->filter_info.dst_port <<
6221 				IXGBE_SDPQF_DSTPORT_SHIFT);
6222 	sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
6223 
6224 	ftqf = (uint32_t)(filter->filter_info.proto &
6225 		IXGBE_FTQF_PROTOCOL_MASK);
6226 	ftqf |= (uint32_t)((filter->filter_info.priority &
6227 		IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
6228 	if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
6229 		mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
6230 	if (filter->filter_info.dst_ip_mask == 0)
6231 		mask &= IXGBE_FTQF_DEST_ADDR_MASK;
6232 	if (filter->filter_info.src_port_mask == 0)
6233 		mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
6234 	if (filter->filter_info.dst_port_mask == 0)
6235 		mask &= IXGBE_FTQF_DEST_PORT_MASK;
6236 	if (filter->filter_info.proto_mask == 0)
6237 		mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
6238 	ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
6239 	ftqf |= IXGBE_FTQF_POOL_MASK_EN;
6240 	ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
6241 
6242 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
6243 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
6244 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
6245 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
6246 
6247 	l34timir |= IXGBE_L34T_IMIR_RESERVE;
6248 	l34timir |= (uint32_t)(filter->queue <<
6249 				IXGBE_L34T_IMIR_QUEUE_SHIFT);
6250 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
6251 }
6252 
6253 /*
6254  * add a 5tuple filter
6255  *
6256  * @param
6257  * dev: Pointer to struct rte_eth_dev.
6258  * index: the index the filter allocates.
6259  * filter: ponter to the filter that will be added.
6260  * rx_queue: the queue id the filter assigned to.
6261  *
6262  * @return
6263  *    - On success, zero.
6264  *    - On failure, a negative value.
6265  */
6266 static int
6267 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
6268 			struct ixgbe_5tuple_filter *filter)
6269 {
6270 	struct ixgbe_filter_info *filter_info =
6271 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6272 	int i, idx, shift;
6273 
6274 	/*
6275 	 * look for an unused 5tuple filter index,
6276 	 * and insert the filter to list.
6277 	 */
6278 	for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
6279 		idx = i / (sizeof(uint32_t) * NBBY);
6280 		shift = i % (sizeof(uint32_t) * NBBY);
6281 		if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
6282 			filter_info->fivetuple_mask[idx] |= 1 << shift;
6283 			filter->index = i;
6284 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
6285 					  filter,
6286 					  entries);
6287 			break;
6288 		}
6289 	}
6290 	if (i >= IXGBE_MAX_FTQF_FILTERS) {
6291 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
6292 		return -ENOSYS;
6293 	}
6294 
6295 	ixgbe_inject_5tuple_filter(dev, filter);
6296 
6297 	return 0;
6298 }
6299 
6300 /*
6301  * remove a 5tuple filter
6302  *
6303  * @param
6304  * dev: Pointer to struct rte_eth_dev.
6305  * filter: the pointer of the filter will be removed.
6306  */
6307 static void
6308 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
6309 			struct ixgbe_5tuple_filter *filter)
6310 {
6311 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6312 	struct ixgbe_filter_info *filter_info =
6313 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6314 	uint16_t index = filter->index;
6315 
6316 	filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
6317 				~(1 << (index % (sizeof(uint32_t) * NBBY)));
6318 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
6319 	rte_free(filter);
6320 
6321 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
6322 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
6323 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
6324 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
6325 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
6326 }
6327 
6328 static int
6329 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
6330 {
6331 	struct ixgbe_hw *hw;
6332 	uint32_t max_frame = mtu + IXGBE_ETH_OVERHEAD;
6333 	struct rte_eth_dev_data *dev_data = dev->data;
6334 
6335 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6336 
6337 	if (mtu < RTE_ETHER_MIN_MTU || max_frame > RTE_ETHER_MAX_JUMBO_FRAME_LEN)
6338 		return -EINVAL;
6339 
6340 	/* If device is started, refuse mtu that requires the support of
6341 	 * scattered packets when this feature has not been enabled before.
6342 	 */
6343 	if (dev_data->dev_started && !dev_data->scattered_rx &&
6344 	    (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
6345 			dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
6346 		PMD_INIT_LOG(ERR, "Stop port first.");
6347 		return -EINVAL;
6348 	}
6349 
6350 	/*
6351 	 * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
6352 	 * request of the version 2.0 of the mailbox API.
6353 	 * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
6354 	 * of the mailbox API.
6355 	 * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
6356 	 * prior to 3.11.33 which contains the following change:
6357 	 * "ixgbe: Enable jumbo frames support w/ SR-IOV"
6358 	 */
6359 	if (ixgbevf_rlpml_set_vf(hw, max_frame))
6360 		return -EINVAL;
6361 
6362 	return 0;
6363 }
6364 
6365 static inline struct ixgbe_5tuple_filter *
6366 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
6367 			struct ixgbe_5tuple_filter_info *key)
6368 {
6369 	struct ixgbe_5tuple_filter *it;
6370 
6371 	TAILQ_FOREACH(it, filter_list, entries) {
6372 		if (memcmp(key, &it->filter_info,
6373 			sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
6374 			return it;
6375 		}
6376 	}
6377 	return NULL;
6378 }
6379 
6380 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
6381 static inline int
6382 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
6383 			struct ixgbe_5tuple_filter_info *filter_info)
6384 {
6385 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
6386 		filter->priority > IXGBE_5TUPLE_MAX_PRI ||
6387 		filter->priority < IXGBE_5TUPLE_MIN_PRI)
6388 		return -EINVAL;
6389 
6390 	switch (filter->dst_ip_mask) {
6391 	case UINT32_MAX:
6392 		filter_info->dst_ip_mask = 0;
6393 		filter_info->dst_ip = filter->dst_ip;
6394 		break;
6395 	case 0:
6396 		filter_info->dst_ip_mask = 1;
6397 		break;
6398 	default:
6399 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
6400 		return -EINVAL;
6401 	}
6402 
6403 	switch (filter->src_ip_mask) {
6404 	case UINT32_MAX:
6405 		filter_info->src_ip_mask = 0;
6406 		filter_info->src_ip = filter->src_ip;
6407 		break;
6408 	case 0:
6409 		filter_info->src_ip_mask = 1;
6410 		break;
6411 	default:
6412 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
6413 		return -EINVAL;
6414 	}
6415 
6416 	switch (filter->dst_port_mask) {
6417 	case UINT16_MAX:
6418 		filter_info->dst_port_mask = 0;
6419 		filter_info->dst_port = filter->dst_port;
6420 		break;
6421 	case 0:
6422 		filter_info->dst_port_mask = 1;
6423 		break;
6424 	default:
6425 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
6426 		return -EINVAL;
6427 	}
6428 
6429 	switch (filter->src_port_mask) {
6430 	case UINT16_MAX:
6431 		filter_info->src_port_mask = 0;
6432 		filter_info->src_port = filter->src_port;
6433 		break;
6434 	case 0:
6435 		filter_info->src_port_mask = 1;
6436 		break;
6437 	default:
6438 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
6439 		return -EINVAL;
6440 	}
6441 
6442 	switch (filter->proto_mask) {
6443 	case UINT8_MAX:
6444 		filter_info->proto_mask = 0;
6445 		filter_info->proto =
6446 			convert_protocol_type(filter->proto);
6447 		break;
6448 	case 0:
6449 		filter_info->proto_mask = 1;
6450 		break;
6451 	default:
6452 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
6453 		return -EINVAL;
6454 	}
6455 
6456 	filter_info->priority = (uint8_t)filter->priority;
6457 	return 0;
6458 }
6459 
6460 /*
6461  * add or delete a ntuple filter
6462  *
6463  * @param
6464  * dev: Pointer to struct rte_eth_dev.
6465  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6466  * add: if true, add filter, if false, remove filter
6467  *
6468  * @return
6469  *    - On success, zero.
6470  *    - On failure, a negative value.
6471  */
6472 int
6473 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
6474 			struct rte_eth_ntuple_filter *ntuple_filter,
6475 			bool add)
6476 {
6477 	struct ixgbe_filter_info *filter_info =
6478 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6479 	struct ixgbe_5tuple_filter_info filter_5tuple;
6480 	struct ixgbe_5tuple_filter *filter;
6481 	int ret;
6482 
6483 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6484 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6485 		return -EINVAL;
6486 	}
6487 
6488 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6489 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6490 	if (ret < 0)
6491 		return ret;
6492 
6493 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6494 					 &filter_5tuple);
6495 	if (filter != NULL && add) {
6496 		PMD_DRV_LOG(ERR, "filter exists.");
6497 		return -EEXIST;
6498 	}
6499 	if (filter == NULL && !add) {
6500 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
6501 		return -ENOENT;
6502 	}
6503 
6504 	if (add) {
6505 		filter = rte_zmalloc("ixgbe_5tuple_filter",
6506 				sizeof(struct ixgbe_5tuple_filter), 0);
6507 		if (filter == NULL)
6508 			return -ENOMEM;
6509 		rte_memcpy(&filter->filter_info,
6510 				 &filter_5tuple,
6511 				 sizeof(struct ixgbe_5tuple_filter_info));
6512 		filter->queue = ntuple_filter->queue;
6513 		ret = ixgbe_add_5tuple_filter(dev, filter);
6514 		if (ret < 0) {
6515 			rte_free(filter);
6516 			return ret;
6517 		}
6518 	} else
6519 		ixgbe_remove_5tuple_filter(dev, filter);
6520 
6521 	return 0;
6522 }
6523 
6524 int
6525 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
6526 			struct rte_eth_ethertype_filter *filter,
6527 			bool add)
6528 {
6529 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6530 	struct ixgbe_filter_info *filter_info =
6531 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6532 	uint32_t etqf = 0;
6533 	uint32_t etqs = 0;
6534 	int ret;
6535 	struct ixgbe_ethertype_filter ethertype_filter;
6536 
6537 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6538 		return -EINVAL;
6539 
6540 	if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
6541 		filter->ether_type == RTE_ETHER_TYPE_IPV6) {
6542 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
6543 			" ethertype filter.", filter->ether_type);
6544 		return -EINVAL;
6545 	}
6546 
6547 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
6548 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
6549 		return -EINVAL;
6550 	}
6551 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
6552 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
6553 		return -EINVAL;
6554 	}
6555 
6556 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6557 	if (ret >= 0 && add) {
6558 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
6559 			    filter->ether_type);
6560 		return -EEXIST;
6561 	}
6562 	if (ret < 0 && !add) {
6563 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6564 			    filter->ether_type);
6565 		return -ENOENT;
6566 	}
6567 
6568 	if (add) {
6569 		etqf = IXGBE_ETQF_FILTER_EN;
6570 		etqf |= (uint32_t)filter->ether_type;
6571 		etqs |= (uint32_t)((filter->queue <<
6572 				    IXGBE_ETQS_RX_QUEUE_SHIFT) &
6573 				    IXGBE_ETQS_RX_QUEUE);
6574 		etqs |= IXGBE_ETQS_QUEUE_EN;
6575 
6576 		ethertype_filter.ethertype = filter->ether_type;
6577 		ethertype_filter.etqf = etqf;
6578 		ethertype_filter.etqs = etqs;
6579 		ethertype_filter.conf = FALSE;
6580 		ret = ixgbe_ethertype_filter_insert(filter_info,
6581 						    &ethertype_filter);
6582 		if (ret < 0) {
6583 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
6584 			return -ENOSPC;
6585 		}
6586 	} else {
6587 		ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
6588 		if (ret < 0)
6589 			return -ENOSYS;
6590 	}
6591 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
6592 	IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
6593 	IXGBE_WRITE_FLUSH(hw);
6594 
6595 	return 0;
6596 }
6597 
6598 static int
6599 ixgbe_dev_flow_ops_get(__rte_unused struct rte_eth_dev *dev,
6600 		       const struct rte_flow_ops **ops)
6601 {
6602 	*ops = &ixgbe_flow_ops;
6603 	return 0;
6604 }
6605 
6606 static u8 *
6607 ixgbe_dev_addr_list_itr(__rte_unused struct ixgbe_hw *hw,
6608 			u8 **mc_addr_ptr, u32 *vmdq)
6609 {
6610 	u8 *mc_addr;
6611 
6612 	*vmdq = 0;
6613 	mc_addr = *mc_addr_ptr;
6614 	*mc_addr_ptr = (mc_addr + sizeof(struct rte_ether_addr));
6615 	return mc_addr;
6616 }
6617 
6618 static int
6619 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
6620 			  struct rte_ether_addr *mc_addr_set,
6621 			  uint32_t nb_mc_addr)
6622 {
6623 	struct ixgbe_hw *hw;
6624 	u8 *mc_addr_list;
6625 
6626 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6627 	mc_addr_list = (u8 *)mc_addr_set;
6628 	return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
6629 					 ixgbe_dev_addr_list_itr, TRUE);
6630 }
6631 
6632 static uint64_t
6633 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
6634 {
6635 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6636 	uint64_t systime_cycles;
6637 
6638 	switch (hw->mac.type) {
6639 	case ixgbe_mac_X550:
6640 	case ixgbe_mac_X550EM_x:
6641 	case ixgbe_mac_X550EM_a:
6642 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
6643 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6644 		systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6645 				* NSEC_PER_SEC;
6646 		break;
6647 	default:
6648 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6649 		systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6650 				<< 32;
6651 	}
6652 
6653 	return systime_cycles;
6654 }
6655 
6656 static uint64_t
6657 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6658 {
6659 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6660 	uint64_t rx_tstamp_cycles;
6661 
6662 	switch (hw->mac.type) {
6663 	case ixgbe_mac_X550:
6664 	case ixgbe_mac_X550EM_x:
6665 	case ixgbe_mac_X550EM_a:
6666 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6667 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6668 		rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6669 				* NSEC_PER_SEC;
6670 		break;
6671 	default:
6672 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6673 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6674 		rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6675 				<< 32;
6676 	}
6677 
6678 	return rx_tstamp_cycles;
6679 }
6680 
6681 static uint64_t
6682 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6683 {
6684 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6685 	uint64_t tx_tstamp_cycles;
6686 
6687 	switch (hw->mac.type) {
6688 	case ixgbe_mac_X550:
6689 	case ixgbe_mac_X550EM_x:
6690 	case ixgbe_mac_X550EM_a:
6691 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6692 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6693 		tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6694 				* NSEC_PER_SEC;
6695 		break;
6696 	default:
6697 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6698 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6699 		tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6700 				<< 32;
6701 	}
6702 
6703 	return tx_tstamp_cycles;
6704 }
6705 
6706 static void
6707 ixgbe_start_timecounters(struct rte_eth_dev *dev)
6708 {
6709 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6710 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6711 	struct rte_eth_link link;
6712 	uint32_t incval = 0;
6713 	uint32_t shift = 0;
6714 
6715 	/* Get current link speed. */
6716 	ixgbe_dev_link_update(dev, 1);
6717 	rte_eth_linkstatus_get(dev, &link);
6718 
6719 	switch (link.link_speed) {
6720 	case RTE_ETH_SPEED_NUM_100M:
6721 		incval = IXGBE_INCVAL_100;
6722 		shift = IXGBE_INCVAL_SHIFT_100;
6723 		break;
6724 	case RTE_ETH_SPEED_NUM_1G:
6725 		incval = IXGBE_INCVAL_1GB;
6726 		shift = IXGBE_INCVAL_SHIFT_1GB;
6727 		break;
6728 	case RTE_ETH_SPEED_NUM_10G:
6729 	default:
6730 		incval = IXGBE_INCVAL_10GB;
6731 		shift = IXGBE_INCVAL_SHIFT_10GB;
6732 		break;
6733 	}
6734 
6735 	switch (hw->mac.type) {
6736 	case ixgbe_mac_X550:
6737 	case ixgbe_mac_X550EM_x:
6738 	case ixgbe_mac_X550EM_a:
6739 		/* Independent of link speed. */
6740 		incval = 1;
6741 		/* Cycles read will be interpreted as ns. */
6742 		shift = 0;
6743 		/* Fall-through */
6744 	case ixgbe_mac_X540:
6745 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
6746 		break;
6747 	case ixgbe_mac_82599EB:
6748 		incval >>= IXGBE_INCVAL_SHIFT_82599;
6749 		shift -= IXGBE_INCVAL_SHIFT_82599;
6750 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
6751 				(1 << IXGBE_INCPER_SHIFT_82599) | incval);
6752 		break;
6753 	default:
6754 		/* Not supported. */
6755 		return;
6756 	}
6757 
6758 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
6759 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6760 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6761 
6762 	adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6763 	adapter->systime_tc.cc_shift = shift;
6764 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
6765 
6766 	adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6767 	adapter->rx_tstamp_tc.cc_shift = shift;
6768 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6769 
6770 	adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6771 	adapter->tx_tstamp_tc.cc_shift = shift;
6772 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6773 }
6774 
6775 static int
6776 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
6777 {
6778 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6779 
6780 	adapter->systime_tc.nsec += delta;
6781 	adapter->rx_tstamp_tc.nsec += delta;
6782 	adapter->tx_tstamp_tc.nsec += delta;
6783 
6784 	return 0;
6785 }
6786 
6787 static int
6788 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
6789 {
6790 	uint64_t ns;
6791 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6792 
6793 	ns = rte_timespec_to_ns(ts);
6794 	/* Set the timecounters to a new value. */
6795 	adapter->systime_tc.nsec = ns;
6796 	adapter->rx_tstamp_tc.nsec = ns;
6797 	adapter->tx_tstamp_tc.nsec = ns;
6798 
6799 	return 0;
6800 }
6801 
6802 static int
6803 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
6804 {
6805 	uint64_t ns, systime_cycles;
6806 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6807 
6808 	systime_cycles = ixgbe_read_systime_cyclecounter(dev);
6809 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
6810 	*ts = rte_ns_to_timespec(ns);
6811 
6812 	return 0;
6813 }
6814 
6815 static int
6816 ixgbe_timesync_enable(struct rte_eth_dev *dev)
6817 {
6818 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6819 	uint32_t tsync_ctl;
6820 	uint32_t tsauxc;
6821 
6822 	/* Stop the timesync system time. */
6823 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
6824 	/* Reset the timesync system time value. */
6825 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
6826 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
6827 
6828 	/* Enable system time for platforms where it isn't on by default. */
6829 	tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
6830 	tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
6831 	IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
6832 
6833 	ixgbe_start_timecounters(dev);
6834 
6835 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6836 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
6837 			(RTE_ETHER_TYPE_1588 |
6838 			 IXGBE_ETQF_FILTER_EN |
6839 			 IXGBE_ETQF_1588));
6840 
6841 	/* Enable timestamping of received PTP packets. */
6842 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6843 	tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
6844 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6845 
6846 	/* Enable timestamping of transmitted PTP packets. */
6847 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6848 	tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
6849 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6850 
6851 	IXGBE_WRITE_FLUSH(hw);
6852 
6853 	return 0;
6854 }
6855 
6856 static int
6857 ixgbe_timesync_disable(struct rte_eth_dev *dev)
6858 {
6859 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6860 	uint32_t tsync_ctl;
6861 
6862 	/* Disable timestamping of transmitted PTP packets. */
6863 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6864 	tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
6865 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6866 
6867 	/* Disable timestamping of received PTP packets. */
6868 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6869 	tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
6870 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6871 
6872 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6873 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
6874 
6875 	/* Stop incrementating the System Time registers. */
6876 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
6877 
6878 	return 0;
6879 }
6880 
6881 static int
6882 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
6883 				 struct timespec *timestamp,
6884 				 uint32_t flags __rte_unused)
6885 {
6886 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6887 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6888 	uint32_t tsync_rxctl;
6889 	uint64_t rx_tstamp_cycles;
6890 	uint64_t ns;
6891 
6892 	tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6893 	if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
6894 		return -EINVAL;
6895 
6896 	rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
6897 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
6898 	*timestamp = rte_ns_to_timespec(ns);
6899 
6900 	return  0;
6901 }
6902 
6903 static int
6904 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
6905 				 struct timespec *timestamp)
6906 {
6907 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6908 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6909 	uint32_t tsync_txctl;
6910 	uint64_t tx_tstamp_cycles;
6911 	uint64_t ns;
6912 
6913 	tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6914 	if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
6915 		return -EINVAL;
6916 
6917 	tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
6918 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
6919 	*timestamp = rte_ns_to_timespec(ns);
6920 
6921 	return 0;
6922 }
6923 
6924 static int
6925 ixgbe_get_reg_length(struct rte_eth_dev *dev)
6926 {
6927 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6928 	int count = 0;
6929 	int g_ind = 0;
6930 	const struct reg_info *reg_group;
6931 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6932 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6933 
6934 	while ((reg_group = reg_set[g_ind++]))
6935 		count += ixgbe_regs_group_count(reg_group);
6936 
6937 	return count;
6938 }
6939 
6940 static int
6941 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
6942 {
6943 	int count = 0;
6944 	int g_ind = 0;
6945 	const struct reg_info *reg_group;
6946 
6947 	while ((reg_group = ixgbevf_regs[g_ind++]))
6948 		count += ixgbe_regs_group_count(reg_group);
6949 
6950 	return count;
6951 }
6952 
6953 static int
6954 ixgbe_get_regs(struct rte_eth_dev *dev,
6955 	      struct rte_dev_reg_info *regs)
6956 {
6957 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6958 	uint32_t *data = regs->data;
6959 	int g_ind = 0;
6960 	int count = 0;
6961 	const struct reg_info *reg_group;
6962 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6963 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6964 
6965 	if (data == NULL) {
6966 		regs->length = ixgbe_get_reg_length(dev);
6967 		regs->width = sizeof(uint32_t);
6968 		return 0;
6969 	}
6970 
6971 	/* Support only full register dump */
6972 	if ((regs->length == 0) ||
6973 	    (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
6974 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
6975 			hw->device_id;
6976 		while ((reg_group = reg_set[g_ind++]))
6977 			count += ixgbe_read_regs_group(dev, &data[count],
6978 				reg_group);
6979 		return 0;
6980 	}
6981 
6982 	return -ENOTSUP;
6983 }
6984 
6985 static int
6986 ixgbevf_get_regs(struct rte_eth_dev *dev,
6987 		struct rte_dev_reg_info *regs)
6988 {
6989 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6990 	uint32_t *data = regs->data;
6991 	int g_ind = 0;
6992 	int count = 0;
6993 	const struct reg_info *reg_group;
6994 
6995 	if (data == NULL) {
6996 		regs->length = ixgbevf_get_reg_length(dev);
6997 		regs->width = sizeof(uint32_t);
6998 		return 0;
6999 	}
7000 
7001 	/* Support only full register dump */
7002 	if ((regs->length == 0) ||
7003 	    (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
7004 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7005 			hw->device_id;
7006 		while ((reg_group = ixgbevf_regs[g_ind++]))
7007 			count += ixgbe_read_regs_group(dev, &data[count],
7008 						      reg_group);
7009 		return 0;
7010 	}
7011 
7012 	return -ENOTSUP;
7013 }
7014 
7015 static int
7016 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
7017 {
7018 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7019 
7020 	/* Return unit is byte count */
7021 	return hw->eeprom.word_size * 2;
7022 }
7023 
7024 static int
7025 ixgbe_get_eeprom(struct rte_eth_dev *dev,
7026 		struct rte_dev_eeprom_info *in_eeprom)
7027 {
7028 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7029 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7030 	uint16_t *data = in_eeprom->data;
7031 	int first, length;
7032 
7033 	first = in_eeprom->offset >> 1;
7034 	length = in_eeprom->length >> 1;
7035 	if ((first > hw->eeprom.word_size) ||
7036 	    ((first + length) > hw->eeprom.word_size))
7037 		return -EINVAL;
7038 
7039 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7040 
7041 	return eeprom->ops.read_buffer(hw, first, length, data);
7042 }
7043 
7044 static int
7045 ixgbe_set_eeprom(struct rte_eth_dev *dev,
7046 		struct rte_dev_eeprom_info *in_eeprom)
7047 {
7048 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7049 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7050 	uint16_t *data = in_eeprom->data;
7051 	int first, length;
7052 
7053 	first = in_eeprom->offset >> 1;
7054 	length = in_eeprom->length >> 1;
7055 	if ((first > hw->eeprom.word_size) ||
7056 	    ((first + length) > hw->eeprom.word_size))
7057 		return -EINVAL;
7058 
7059 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7060 
7061 	return eeprom->ops.write_buffer(hw,  first, length, data);
7062 }
7063 
7064 static int
7065 ixgbe_get_module_info(struct rte_eth_dev *dev,
7066 		      struct rte_eth_dev_module_info *modinfo)
7067 {
7068 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7069 	uint32_t status;
7070 	uint8_t sff8472_rev, addr_mode;
7071 	bool page_swap = false;
7072 
7073 	/* Check whether we support SFF-8472 or not */
7074 	status = hw->phy.ops.read_i2c_eeprom(hw,
7075 					     IXGBE_SFF_SFF_8472_COMP,
7076 					     &sff8472_rev);
7077 	if (status != 0)
7078 		return -EIO;
7079 
7080 	/* addressing mode is not supported */
7081 	status = hw->phy.ops.read_i2c_eeprom(hw,
7082 					     IXGBE_SFF_SFF_8472_SWAP,
7083 					     &addr_mode);
7084 	if (status != 0)
7085 		return -EIO;
7086 
7087 	if (addr_mode & IXGBE_SFF_ADDRESSING_MODE) {
7088 		PMD_DRV_LOG(ERR,
7089 			    "Address change required to access page 0xA2, "
7090 			    "but not supported. Please report the module "
7091 			    "type to the driver maintainers.");
7092 		page_swap = true;
7093 	}
7094 
7095 	if (sff8472_rev == IXGBE_SFF_SFF_8472_UNSUP || page_swap) {
7096 		/* We have a SFP, but it does not support SFF-8472 */
7097 		modinfo->type = RTE_ETH_MODULE_SFF_8079;
7098 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
7099 	} else {
7100 		/* We have a SFP which supports a revision of SFF-8472. */
7101 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
7102 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
7103 	}
7104 
7105 	return 0;
7106 }
7107 
7108 static int
7109 ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
7110 			struct rte_dev_eeprom_info *info)
7111 {
7112 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7113 	uint32_t status = IXGBE_ERR_PHY_ADDR_INVALID;
7114 	uint8_t databyte = 0xFF;
7115 	uint8_t *data = info->data;
7116 	uint32_t i = 0;
7117 
7118 	for (i = info->offset; i < info->offset + info->length; i++) {
7119 		if (i < RTE_ETH_MODULE_SFF_8079_LEN)
7120 			status = hw->phy.ops.read_i2c_eeprom(hw, i, &databyte);
7121 		else
7122 			status = hw->phy.ops.read_i2c_sff8472(hw, i, &databyte);
7123 
7124 		if (status != 0)
7125 			return -EIO;
7126 
7127 		data[i - info->offset] = databyte;
7128 	}
7129 
7130 	return 0;
7131 }
7132 
7133 uint16_t
7134 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
7135 	switch (mac_type) {
7136 	case ixgbe_mac_X550:
7137 	case ixgbe_mac_X550EM_x:
7138 	case ixgbe_mac_X550EM_a:
7139 		return RTE_ETH_RSS_RETA_SIZE_512;
7140 	case ixgbe_mac_X550_vf:
7141 	case ixgbe_mac_X550EM_x_vf:
7142 	case ixgbe_mac_X550EM_a_vf:
7143 		return RTE_ETH_RSS_RETA_SIZE_64;
7144 	case ixgbe_mac_X540_vf:
7145 	case ixgbe_mac_82599_vf:
7146 		return 0;
7147 	default:
7148 		return RTE_ETH_RSS_RETA_SIZE_128;
7149 	}
7150 }
7151 
7152 uint32_t
7153 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
7154 	switch (mac_type) {
7155 	case ixgbe_mac_X550:
7156 	case ixgbe_mac_X550EM_x:
7157 	case ixgbe_mac_X550EM_a:
7158 		if (reta_idx < RTE_ETH_RSS_RETA_SIZE_128)
7159 			return IXGBE_RETA(reta_idx >> 2);
7160 		else
7161 			return IXGBE_ERETA((reta_idx - RTE_ETH_RSS_RETA_SIZE_128) >> 2);
7162 	case ixgbe_mac_X550_vf:
7163 	case ixgbe_mac_X550EM_x_vf:
7164 	case ixgbe_mac_X550EM_a_vf:
7165 		return IXGBE_VFRETA(reta_idx >> 2);
7166 	default:
7167 		return IXGBE_RETA(reta_idx >> 2);
7168 	}
7169 }
7170 
7171 uint32_t
7172 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
7173 	switch (mac_type) {
7174 	case ixgbe_mac_X550_vf:
7175 	case ixgbe_mac_X550EM_x_vf:
7176 	case ixgbe_mac_X550EM_a_vf:
7177 		return IXGBE_VFMRQC;
7178 	default:
7179 		return IXGBE_MRQC;
7180 	}
7181 }
7182 
7183 uint32_t
7184 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
7185 	switch (mac_type) {
7186 	case ixgbe_mac_X550_vf:
7187 	case ixgbe_mac_X550EM_x_vf:
7188 	case ixgbe_mac_X550EM_a_vf:
7189 		return IXGBE_VFRSSRK(i);
7190 	default:
7191 		return IXGBE_RSSRK(i);
7192 	}
7193 }
7194 
7195 bool
7196 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
7197 	switch (mac_type) {
7198 	case ixgbe_mac_82599_vf:
7199 	case ixgbe_mac_X540_vf:
7200 		return 0;
7201 	default:
7202 		return 1;
7203 	}
7204 }
7205 
7206 static int
7207 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
7208 			struct rte_eth_dcb_info *dcb_info)
7209 {
7210 	struct ixgbe_dcb_config *dcb_config =
7211 			IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
7212 	struct ixgbe_dcb_tc_config *tc;
7213 	struct rte_eth_dcb_tc_queue_mapping *tc_queue;
7214 	uint8_t nb_tcs;
7215 	uint8_t i, j;
7216 
7217 	if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_DCB_FLAG)
7218 		dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
7219 	else
7220 		dcb_info->nb_tcs = 1;
7221 
7222 	tc_queue = &dcb_info->tc_queue;
7223 	nb_tcs = dcb_info->nb_tcs;
7224 
7225 	if (dcb_config->vt_mode) { /* vt is enabled*/
7226 		struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
7227 				&dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
7228 		for (i = 0; i < RTE_ETH_DCB_NUM_USER_PRIORITIES; i++)
7229 			dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
7230 		if (RTE_ETH_DEV_SRIOV(dev).active > 0) {
7231 			for (j = 0; j < nb_tcs; j++) {
7232 				tc_queue->tc_rxq[0][j].base = j;
7233 				tc_queue->tc_rxq[0][j].nb_queue = 1;
7234 				tc_queue->tc_txq[0][j].base = j;
7235 				tc_queue->tc_txq[0][j].nb_queue = 1;
7236 			}
7237 		} else {
7238 			for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
7239 				for (j = 0; j < nb_tcs; j++) {
7240 					tc_queue->tc_rxq[i][j].base =
7241 						i * nb_tcs + j;
7242 					tc_queue->tc_rxq[i][j].nb_queue = 1;
7243 					tc_queue->tc_txq[i][j].base =
7244 						i * nb_tcs + j;
7245 					tc_queue->tc_txq[i][j].nb_queue = 1;
7246 				}
7247 			}
7248 		}
7249 	} else { /* vt is disabled*/
7250 		struct rte_eth_dcb_rx_conf *rx_conf =
7251 				&dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
7252 		for (i = 0; i < RTE_ETH_DCB_NUM_USER_PRIORITIES; i++)
7253 			dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
7254 		if (dcb_info->nb_tcs == RTE_ETH_4_TCS) {
7255 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7256 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
7257 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7258 			}
7259 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7260 			dcb_info->tc_queue.tc_txq[0][1].base = 64;
7261 			dcb_info->tc_queue.tc_txq[0][2].base = 96;
7262 			dcb_info->tc_queue.tc_txq[0][3].base = 112;
7263 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
7264 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7265 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7266 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7267 		} else if (dcb_info->nb_tcs == RTE_ETH_8_TCS) {
7268 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7269 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
7270 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7271 			}
7272 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7273 			dcb_info->tc_queue.tc_txq[0][1].base = 32;
7274 			dcb_info->tc_queue.tc_txq[0][2].base = 64;
7275 			dcb_info->tc_queue.tc_txq[0][3].base = 80;
7276 			dcb_info->tc_queue.tc_txq[0][4].base = 96;
7277 			dcb_info->tc_queue.tc_txq[0][5].base = 104;
7278 			dcb_info->tc_queue.tc_txq[0][6].base = 112;
7279 			dcb_info->tc_queue.tc_txq[0][7].base = 120;
7280 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
7281 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7282 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7283 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7284 			dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
7285 			dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
7286 			dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
7287 			dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
7288 		}
7289 	}
7290 	for (i = 0; i < dcb_info->nb_tcs; i++) {
7291 		tc = &dcb_config->tc_config[i];
7292 		dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
7293 	}
7294 	return 0;
7295 }
7296 
7297 /* Update e-tag ether type */
7298 static int
7299 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
7300 			    uint16_t ether_type)
7301 {
7302 	uint32_t etag_etype;
7303 
7304 	if (hw->mac.type != ixgbe_mac_X550 &&
7305 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7306 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7307 		return -ENOTSUP;
7308 	}
7309 
7310 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7311 	etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
7312 	etag_etype |= ether_type;
7313 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7314 	IXGBE_WRITE_FLUSH(hw);
7315 
7316 	return 0;
7317 }
7318 
7319 /* Enable e-tag tunnel */
7320 static int
7321 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
7322 {
7323 	uint32_t etag_etype;
7324 
7325 	if (hw->mac.type != ixgbe_mac_X550 &&
7326 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7327 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7328 		return -ENOTSUP;
7329 	}
7330 
7331 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7332 	etag_etype |= IXGBE_ETAG_ETYPE_VALID;
7333 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7334 	IXGBE_WRITE_FLUSH(hw);
7335 
7336 	return 0;
7337 }
7338 
7339 static int
7340 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
7341 		       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7342 {
7343 	int ret = 0;
7344 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7345 	uint32_t i, rar_entries;
7346 	uint32_t rar_low, rar_high;
7347 
7348 	if (hw->mac.type != ixgbe_mac_X550 &&
7349 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7350 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7351 		return -ENOTSUP;
7352 	}
7353 
7354 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7355 
7356 	for (i = 1; i < rar_entries; i++) {
7357 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7358 		rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
7359 		if ((rar_high & IXGBE_RAH_AV) &&
7360 		    (rar_high & IXGBE_RAH_ADTYPE) &&
7361 		    ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
7362 		     l2_tunnel->tunnel_id)) {
7363 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
7364 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
7365 
7366 			ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
7367 
7368 			return ret;
7369 		}
7370 	}
7371 
7372 	return ret;
7373 }
7374 
7375 static int
7376 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
7377 		       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7378 {
7379 	int ret = 0;
7380 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7381 	uint32_t i, rar_entries;
7382 	uint32_t rar_low, rar_high;
7383 
7384 	if (hw->mac.type != ixgbe_mac_X550 &&
7385 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7386 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7387 		return -ENOTSUP;
7388 	}
7389 
7390 	/* One entry for one tunnel. Try to remove potential existing entry. */
7391 	ixgbe_e_tag_filter_del(dev, l2_tunnel);
7392 
7393 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7394 
7395 	for (i = 1; i < rar_entries; i++) {
7396 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7397 		if (rar_high & IXGBE_RAH_AV) {
7398 			continue;
7399 		} else {
7400 			ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
7401 			rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
7402 			rar_low = l2_tunnel->tunnel_id;
7403 
7404 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
7405 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
7406 
7407 			return ret;
7408 		}
7409 	}
7410 
7411 	PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
7412 		     " Please remove a rule before adding a new one.");
7413 	return -EINVAL;
7414 }
7415 
7416 static inline struct ixgbe_l2_tn_filter *
7417 ixgbe_l2_tn_filter_lookup(struct ixgbe_l2_tn_info *l2_tn_info,
7418 			  struct ixgbe_l2_tn_key *key)
7419 {
7420 	int ret;
7421 
7422 	ret = rte_hash_lookup(l2_tn_info->hash_handle, (const void *)key);
7423 	if (ret < 0)
7424 		return NULL;
7425 
7426 	return l2_tn_info->hash_map[ret];
7427 }
7428 
7429 static inline int
7430 ixgbe_insert_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7431 			  struct ixgbe_l2_tn_filter *l2_tn_filter)
7432 {
7433 	int ret;
7434 
7435 	ret = rte_hash_add_key(l2_tn_info->hash_handle,
7436 			       &l2_tn_filter->key);
7437 
7438 	if (ret < 0) {
7439 		PMD_DRV_LOG(ERR,
7440 			    "Failed to insert L2 tunnel filter"
7441 			    " to hash table %d!",
7442 			    ret);
7443 		return ret;
7444 	}
7445 
7446 	l2_tn_info->hash_map[ret] = l2_tn_filter;
7447 
7448 	TAILQ_INSERT_TAIL(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7449 
7450 	return 0;
7451 }
7452 
7453 static inline int
7454 ixgbe_remove_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7455 			  struct ixgbe_l2_tn_key *key)
7456 {
7457 	int ret;
7458 	struct ixgbe_l2_tn_filter *l2_tn_filter;
7459 
7460 	ret = rte_hash_del_key(l2_tn_info->hash_handle, key);
7461 
7462 	if (ret < 0) {
7463 		PMD_DRV_LOG(ERR,
7464 			    "No such L2 tunnel filter to delete %d!",
7465 			    ret);
7466 		return ret;
7467 	}
7468 
7469 	l2_tn_filter = l2_tn_info->hash_map[ret];
7470 	l2_tn_info->hash_map[ret] = NULL;
7471 
7472 	TAILQ_REMOVE(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7473 	rte_free(l2_tn_filter);
7474 
7475 	return 0;
7476 }
7477 
7478 /* Add l2 tunnel filter */
7479 int
7480 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
7481 			       struct ixgbe_l2_tunnel_conf *l2_tunnel,
7482 			       bool restore)
7483 {
7484 	int ret;
7485 	struct ixgbe_l2_tn_info *l2_tn_info =
7486 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7487 	struct ixgbe_l2_tn_key key;
7488 	struct ixgbe_l2_tn_filter *node;
7489 
7490 	if (!restore) {
7491 		key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7492 		key.tn_id = l2_tunnel->tunnel_id;
7493 
7494 		node = ixgbe_l2_tn_filter_lookup(l2_tn_info, &key);
7495 
7496 		if (node) {
7497 			PMD_DRV_LOG(ERR,
7498 				    "The L2 tunnel filter already exists!");
7499 			return -EINVAL;
7500 		}
7501 
7502 		node = rte_zmalloc("ixgbe_l2_tn",
7503 				   sizeof(struct ixgbe_l2_tn_filter),
7504 				   0);
7505 		if (!node)
7506 			return -ENOMEM;
7507 
7508 		rte_memcpy(&node->key,
7509 				 &key,
7510 				 sizeof(struct ixgbe_l2_tn_key));
7511 		node->pool = l2_tunnel->pool;
7512 		ret = ixgbe_insert_l2_tn_filter(l2_tn_info, node);
7513 		if (ret < 0) {
7514 			rte_free(node);
7515 			return ret;
7516 		}
7517 	}
7518 
7519 	switch (l2_tunnel->l2_tunnel_type) {
7520 	case RTE_ETH_L2_TUNNEL_TYPE_E_TAG:
7521 		ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
7522 		break;
7523 	default:
7524 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7525 		ret = -EINVAL;
7526 		break;
7527 	}
7528 
7529 	if ((!restore) && (ret < 0))
7530 		(void)ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7531 
7532 	return ret;
7533 }
7534 
7535 /* Delete l2 tunnel filter */
7536 int
7537 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
7538 			       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7539 {
7540 	int ret;
7541 	struct ixgbe_l2_tn_info *l2_tn_info =
7542 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7543 	struct ixgbe_l2_tn_key key;
7544 
7545 	key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7546 	key.tn_id = l2_tunnel->tunnel_id;
7547 	ret = ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7548 	if (ret < 0)
7549 		return ret;
7550 
7551 	switch (l2_tunnel->l2_tunnel_type) {
7552 	case RTE_ETH_L2_TUNNEL_TYPE_E_TAG:
7553 		ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
7554 		break;
7555 	default:
7556 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7557 		ret = -EINVAL;
7558 		break;
7559 	}
7560 
7561 	return ret;
7562 }
7563 
7564 static int
7565 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
7566 {
7567 	int ret = 0;
7568 	uint32_t ctrl;
7569 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7570 
7571 	if (hw->mac.type != ixgbe_mac_X550 &&
7572 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7573 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7574 		return -ENOTSUP;
7575 	}
7576 
7577 	ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
7578 	ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
7579 	if (en)
7580 		ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
7581 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
7582 
7583 	return ret;
7584 }
7585 
7586 static int
7587 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
7588 			uint16_t port)
7589 {
7590 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
7591 	IXGBE_WRITE_FLUSH(hw);
7592 
7593 	return 0;
7594 }
7595 
7596 /* There's only one register for VxLAN UDP port.
7597  * So, we cannot add several ports. Will update it.
7598  */
7599 static int
7600 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
7601 		     uint16_t port)
7602 {
7603 	if (port == 0) {
7604 		PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
7605 		return -EINVAL;
7606 	}
7607 
7608 	return ixgbe_update_vxlan_port(hw, port);
7609 }
7610 
7611 /* We cannot delete the VxLAN port. For there's a register for VxLAN
7612  * UDP port, it must have a value.
7613  * So, will reset it to the original value 0.
7614  */
7615 static int
7616 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
7617 		     uint16_t port)
7618 {
7619 	uint16_t cur_port;
7620 
7621 	cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
7622 
7623 	if (cur_port != port) {
7624 		PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
7625 		return -EINVAL;
7626 	}
7627 
7628 	return ixgbe_update_vxlan_port(hw, 0);
7629 }
7630 
7631 /* Add UDP tunneling port */
7632 static int
7633 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
7634 			      struct rte_eth_udp_tunnel *udp_tunnel)
7635 {
7636 	int ret = 0;
7637 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7638 
7639 	if (hw->mac.type != ixgbe_mac_X550 &&
7640 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7641 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7642 		return -ENOTSUP;
7643 	}
7644 
7645 	if (udp_tunnel == NULL)
7646 		return -EINVAL;
7647 
7648 	switch (udp_tunnel->prot_type) {
7649 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
7650 		ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
7651 		break;
7652 
7653 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
7654 	case RTE_ETH_TUNNEL_TYPE_TEREDO:
7655 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7656 		ret = -EINVAL;
7657 		break;
7658 
7659 	default:
7660 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7661 		ret = -EINVAL;
7662 		break;
7663 	}
7664 
7665 	return ret;
7666 }
7667 
7668 /* Remove UDP tunneling port */
7669 static int
7670 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
7671 			      struct rte_eth_udp_tunnel *udp_tunnel)
7672 {
7673 	int ret = 0;
7674 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7675 
7676 	if (hw->mac.type != ixgbe_mac_X550 &&
7677 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7678 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7679 		return -ENOTSUP;
7680 	}
7681 
7682 	if (udp_tunnel == NULL)
7683 		return -EINVAL;
7684 
7685 	switch (udp_tunnel->prot_type) {
7686 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
7687 		ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
7688 		break;
7689 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
7690 	case RTE_ETH_TUNNEL_TYPE_TEREDO:
7691 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7692 		ret = -EINVAL;
7693 		break;
7694 	default:
7695 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7696 		ret = -EINVAL;
7697 		break;
7698 	}
7699 
7700 	return ret;
7701 }
7702 
7703 static int
7704 ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev)
7705 {
7706 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7707 	int ret;
7708 
7709 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_PROMISC)) {
7710 	case IXGBE_SUCCESS:
7711 		ret = 0;
7712 		break;
7713 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7714 		ret = -ENOTSUP;
7715 		break;
7716 	default:
7717 		ret = -EAGAIN;
7718 		break;
7719 	}
7720 
7721 	return ret;
7722 }
7723 
7724 static int
7725 ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev)
7726 {
7727 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7728 	int ret;
7729 
7730 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_NONE)) {
7731 	case IXGBE_SUCCESS:
7732 		ret = 0;
7733 		break;
7734 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7735 		ret = -ENOTSUP;
7736 		break;
7737 	default:
7738 		ret = -EAGAIN;
7739 		break;
7740 	}
7741 
7742 	return ret;
7743 }
7744 
7745 static int
7746 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
7747 {
7748 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7749 	int ret;
7750 	int mode = IXGBEVF_XCAST_MODE_ALLMULTI;
7751 
7752 	switch (hw->mac.ops.update_xcast_mode(hw, mode)) {
7753 	case IXGBE_SUCCESS:
7754 		ret = 0;
7755 		break;
7756 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7757 		ret = -ENOTSUP;
7758 		break;
7759 	default:
7760 		ret = -EAGAIN;
7761 		break;
7762 	}
7763 
7764 	return ret;
7765 }
7766 
7767 static int
7768 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
7769 {
7770 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7771 	int ret;
7772 
7773 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_MULTI)) {
7774 	case IXGBE_SUCCESS:
7775 		ret = 0;
7776 		break;
7777 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7778 		ret = -ENOTSUP;
7779 		break;
7780 	default:
7781 		ret = -EAGAIN;
7782 		break;
7783 	}
7784 
7785 	return ret;
7786 }
7787 
7788 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
7789 {
7790 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7791 	u32 in_msg = 0;
7792 
7793 	/* peek the message first */
7794 	in_msg = IXGBE_READ_REG(hw, IXGBE_VFMBMEM);
7795 
7796 	/* PF reset VF event */
7797 	if (in_msg == IXGBE_PF_CONTROL_MSG) {
7798 		/* dummy mbx read to ack pf */
7799 		if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
7800 			return;
7801 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
7802 					     NULL);
7803 	}
7804 }
7805 
7806 static int
7807 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
7808 {
7809 	uint32_t eicr;
7810 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7811 	struct ixgbe_interrupt *intr =
7812 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7813 	ixgbevf_intr_disable(dev);
7814 
7815 	/* read-on-clear nic registers here */
7816 	eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
7817 	intr->flags = 0;
7818 
7819 	/* only one misc vector supported - mailbox */
7820 	eicr &= IXGBE_VTEICR_MASK;
7821 	if (eicr == IXGBE_MISC_VEC_ID)
7822 		intr->flags |= IXGBE_FLAG_MAILBOX;
7823 
7824 	return 0;
7825 }
7826 
7827 static int
7828 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
7829 {
7830 	struct ixgbe_interrupt *intr =
7831 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7832 
7833 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
7834 		ixgbevf_mbx_process(dev);
7835 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
7836 	}
7837 
7838 	ixgbevf_intr_enable(dev);
7839 
7840 	return 0;
7841 }
7842 
7843 static void
7844 ixgbevf_dev_interrupt_handler(void *param)
7845 {
7846 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
7847 
7848 	ixgbevf_dev_interrupt_get_status(dev);
7849 	ixgbevf_dev_interrupt_action(dev);
7850 }
7851 
7852 /**
7853  *  ixgbe_disable_sec_tx_path_generic - Stops the transmit data path
7854  *  @hw: pointer to hardware structure
7855  *
7856  *  Stops the transmit data path and waits for the HW to internally empty
7857  *  the Tx security block
7858  **/
7859 int ixgbe_disable_sec_tx_path_generic(struct ixgbe_hw *hw)
7860 {
7861 #define IXGBE_MAX_SECTX_POLL 40
7862 
7863 	int i;
7864 	int sectxreg;
7865 
7866 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
7867 	sectxreg |= IXGBE_SECTXCTRL_TX_DIS;
7868 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
7869 	for (i = 0; i < IXGBE_MAX_SECTX_POLL; i++) {
7870 		sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXSTAT);
7871 		if (sectxreg & IXGBE_SECTXSTAT_SECTX_RDY)
7872 			break;
7873 		/* Use interrupt-safe sleep just in case */
7874 		usec_delay(1000);
7875 	}
7876 
7877 	/* For informational purposes only */
7878 	if (i >= IXGBE_MAX_SECTX_POLL)
7879 		PMD_DRV_LOG(DEBUG, "Tx unit being enabled before security "
7880 			 "path fully disabled.  Continuing with init.");
7881 
7882 	return IXGBE_SUCCESS;
7883 }
7884 
7885 /**
7886  *  ixgbe_enable_sec_tx_path_generic - Enables the transmit data path
7887  *  @hw: pointer to hardware structure
7888  *
7889  *  Enables the transmit data path.
7890  **/
7891 int ixgbe_enable_sec_tx_path_generic(struct ixgbe_hw *hw)
7892 {
7893 	uint32_t sectxreg;
7894 
7895 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
7896 	sectxreg &= ~IXGBE_SECTXCTRL_TX_DIS;
7897 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
7898 	IXGBE_WRITE_FLUSH(hw);
7899 
7900 	return IXGBE_SUCCESS;
7901 }
7902 
7903 /* restore n-tuple filter */
7904 static inline void
7905 ixgbe_ntuple_filter_restore(struct rte_eth_dev *dev)
7906 {
7907 	struct ixgbe_filter_info *filter_info =
7908 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7909 	struct ixgbe_5tuple_filter *node;
7910 
7911 	TAILQ_FOREACH(node, &filter_info->fivetuple_list, entries) {
7912 		ixgbe_inject_5tuple_filter(dev, node);
7913 	}
7914 }
7915 
7916 /* restore ethernet type filter */
7917 static inline void
7918 ixgbe_ethertype_filter_restore(struct rte_eth_dev *dev)
7919 {
7920 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7921 	struct ixgbe_filter_info *filter_info =
7922 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7923 	int i;
7924 
7925 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
7926 		if (filter_info->ethertype_mask & (1 << i)) {
7927 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i),
7928 					filter_info->ethertype_filters[i].etqf);
7929 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i),
7930 					filter_info->ethertype_filters[i].etqs);
7931 			IXGBE_WRITE_FLUSH(hw);
7932 		}
7933 	}
7934 }
7935 
7936 /* restore SYN filter */
7937 static inline void
7938 ixgbe_syn_filter_restore(struct rte_eth_dev *dev)
7939 {
7940 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7941 	struct ixgbe_filter_info *filter_info =
7942 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7943 	uint32_t synqf;
7944 
7945 	synqf = filter_info->syn_info;
7946 
7947 	if (synqf & IXGBE_SYN_FILTER_ENABLE) {
7948 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
7949 		IXGBE_WRITE_FLUSH(hw);
7950 	}
7951 }
7952 
7953 /* restore L2 tunnel filter */
7954 static inline void
7955 ixgbe_l2_tn_filter_restore(struct rte_eth_dev *dev)
7956 {
7957 	struct ixgbe_l2_tn_info *l2_tn_info =
7958 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7959 	struct ixgbe_l2_tn_filter *node;
7960 	struct ixgbe_l2_tunnel_conf l2_tn_conf;
7961 
7962 	TAILQ_FOREACH(node, &l2_tn_info->l2_tn_list, entries) {
7963 		l2_tn_conf.l2_tunnel_type = node->key.l2_tn_type;
7964 		l2_tn_conf.tunnel_id      = node->key.tn_id;
7965 		l2_tn_conf.pool           = node->pool;
7966 		(void)ixgbe_dev_l2_tunnel_filter_add(dev, &l2_tn_conf, TRUE);
7967 	}
7968 }
7969 
7970 /* restore rss filter */
7971 static inline void
7972 ixgbe_rss_filter_restore(struct rte_eth_dev *dev)
7973 {
7974 	struct ixgbe_filter_info *filter_info =
7975 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7976 
7977 	if (filter_info->rss_info.conf.queue_num)
7978 		ixgbe_config_rss_filter(dev,
7979 			&filter_info->rss_info, TRUE);
7980 }
7981 
7982 static int
7983 ixgbe_filter_restore(struct rte_eth_dev *dev)
7984 {
7985 	ixgbe_ntuple_filter_restore(dev);
7986 	ixgbe_ethertype_filter_restore(dev);
7987 	ixgbe_syn_filter_restore(dev);
7988 	ixgbe_fdir_filter_restore(dev);
7989 	ixgbe_l2_tn_filter_restore(dev);
7990 	ixgbe_rss_filter_restore(dev);
7991 
7992 	return 0;
7993 }
7994 
7995 static void
7996 ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev)
7997 {
7998 	struct ixgbe_l2_tn_info *l2_tn_info =
7999 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8000 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8001 
8002 	if (l2_tn_info->e_tag_en)
8003 		(void)ixgbe_e_tag_enable(hw);
8004 
8005 	if (l2_tn_info->e_tag_fwd_en)
8006 		(void)ixgbe_e_tag_forwarding_en_dis(dev, 1);
8007 
8008 	(void)ixgbe_update_e_tag_eth_type(hw, l2_tn_info->e_tag_ether_type);
8009 }
8010 
8011 /* remove all the n-tuple filters */
8012 void
8013 ixgbe_clear_all_ntuple_filter(struct rte_eth_dev *dev)
8014 {
8015 	struct ixgbe_filter_info *filter_info =
8016 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8017 	struct ixgbe_5tuple_filter *p_5tuple;
8018 
8019 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
8020 		ixgbe_remove_5tuple_filter(dev, p_5tuple);
8021 }
8022 
8023 /* remove all the ether type filters */
8024 void
8025 ixgbe_clear_all_ethertype_filter(struct rte_eth_dev *dev)
8026 {
8027 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8028 	struct ixgbe_filter_info *filter_info =
8029 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8030 	int i;
8031 
8032 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8033 		if (filter_info->ethertype_mask & (1 << i) &&
8034 		    !filter_info->ethertype_filters[i].conf) {
8035 			(void)ixgbe_ethertype_filter_remove(filter_info,
8036 							    (uint8_t)i);
8037 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i), 0);
8038 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i), 0);
8039 			IXGBE_WRITE_FLUSH(hw);
8040 		}
8041 	}
8042 }
8043 
8044 /* remove the SYN filter */
8045 void
8046 ixgbe_clear_syn_filter(struct rte_eth_dev *dev)
8047 {
8048 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8049 	struct ixgbe_filter_info *filter_info =
8050 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8051 
8052 	if (filter_info->syn_info & IXGBE_SYN_FILTER_ENABLE) {
8053 		filter_info->syn_info = 0;
8054 
8055 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, 0);
8056 		IXGBE_WRITE_FLUSH(hw);
8057 	}
8058 }
8059 
8060 /* remove all the L2 tunnel filters */
8061 int
8062 ixgbe_clear_all_l2_tn_filter(struct rte_eth_dev *dev)
8063 {
8064 	struct ixgbe_l2_tn_info *l2_tn_info =
8065 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8066 	struct ixgbe_l2_tn_filter *l2_tn_filter;
8067 	struct ixgbe_l2_tunnel_conf l2_tn_conf;
8068 	int ret = 0;
8069 
8070 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
8071 		l2_tn_conf.l2_tunnel_type = l2_tn_filter->key.l2_tn_type;
8072 		l2_tn_conf.tunnel_id      = l2_tn_filter->key.tn_id;
8073 		l2_tn_conf.pool           = l2_tn_filter->pool;
8074 		ret = ixgbe_dev_l2_tunnel_filter_del(dev, &l2_tn_conf);
8075 		if (ret < 0)
8076 			return ret;
8077 	}
8078 
8079 	return 0;
8080 }
8081 
8082 void
8083 ixgbe_dev_macsec_setting_save(struct rte_eth_dev *dev,
8084 				struct ixgbe_macsec_setting *macsec_setting)
8085 {
8086 	struct ixgbe_macsec_setting *macsec =
8087 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8088 
8089 	macsec->offload_en = macsec_setting->offload_en;
8090 	macsec->encrypt_en = macsec_setting->encrypt_en;
8091 	macsec->replayprotect_en = macsec_setting->replayprotect_en;
8092 }
8093 
8094 void
8095 ixgbe_dev_macsec_setting_reset(struct rte_eth_dev *dev)
8096 {
8097 	struct ixgbe_macsec_setting *macsec =
8098 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8099 
8100 	macsec->offload_en = 0;
8101 	macsec->encrypt_en = 0;
8102 	macsec->replayprotect_en = 0;
8103 }
8104 
8105 void
8106 ixgbe_dev_macsec_register_enable(struct rte_eth_dev *dev,
8107 				struct ixgbe_macsec_setting *macsec_setting)
8108 {
8109 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8110 	uint32_t ctrl;
8111 	uint8_t en = macsec_setting->encrypt_en;
8112 	uint8_t rp = macsec_setting->replayprotect_en;
8113 
8114 	/**
8115 	 * Workaround:
8116 	 * As no ixgbe_disable_sec_rx_path equivalent is
8117 	 * implemented for tx in the base code, and we are
8118 	 * not allowed to modify the base code in DPDK, so
8119 	 * just call the hand-written one directly for now.
8120 	 * The hardware support has been checked by
8121 	 * ixgbe_disable_sec_rx_path().
8122 	 */
8123 	ixgbe_disable_sec_tx_path_generic(hw);
8124 
8125 	/* Enable Ethernet CRC (required by MACsec offload) */
8126 	ctrl = IXGBE_READ_REG(hw, IXGBE_HLREG0);
8127 	ctrl |= IXGBE_HLREG0_TXCRCEN | IXGBE_HLREG0_RXCRCSTRP;
8128 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, ctrl);
8129 
8130 	/* Enable the TX and RX crypto engines */
8131 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8132 	ctrl &= ~IXGBE_SECTXCTRL_SECTX_DIS;
8133 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8134 
8135 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8136 	ctrl &= ~IXGBE_SECRXCTRL_SECRX_DIS;
8137 	IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8138 
8139 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
8140 	ctrl &= ~IXGBE_SECTX_MINSECIFG_MASK;
8141 	ctrl |= 0x3;
8142 	IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, ctrl);
8143 
8144 	/* Enable SA lookup */
8145 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8146 	ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8147 	ctrl |= en ? IXGBE_LSECTXCTRL_AUTH_ENCRYPT :
8148 		     IXGBE_LSECTXCTRL_AUTH;
8149 	ctrl |= IXGBE_LSECTXCTRL_AISCI;
8150 	ctrl &= ~IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8151 	ctrl |= IXGBE_MACSEC_PNTHRSH & IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8152 	IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8153 
8154 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8155 	ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8156 	ctrl |= IXGBE_LSECRXCTRL_STRICT << IXGBE_LSECRXCTRL_EN_SHIFT;
8157 	ctrl &= ~IXGBE_LSECRXCTRL_PLSH;
8158 	if (rp)
8159 		ctrl |= IXGBE_LSECRXCTRL_RP;
8160 	else
8161 		ctrl &= ~IXGBE_LSECRXCTRL_RP;
8162 	IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8163 
8164 	/* Start the data paths */
8165 	ixgbe_enable_sec_rx_path(hw);
8166 	/**
8167 	 * Workaround:
8168 	 * As no ixgbe_enable_sec_rx_path equivalent is
8169 	 * implemented for tx in the base code, and we are
8170 	 * not allowed to modify the base code in DPDK, so
8171 	 * just call the hand-written one directly for now.
8172 	 */
8173 	ixgbe_enable_sec_tx_path_generic(hw);
8174 }
8175 
8176 void
8177 ixgbe_dev_macsec_register_disable(struct rte_eth_dev *dev)
8178 {
8179 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8180 	uint32_t ctrl;
8181 
8182 	/**
8183 	 * Workaround:
8184 	 * As no ixgbe_disable_sec_rx_path equivalent is
8185 	 * implemented for tx in the base code, and we are
8186 	 * not allowed to modify the base code in DPDK, so
8187 	 * just call the hand-written one directly for now.
8188 	 * The hardware support has been checked by
8189 	 * ixgbe_disable_sec_rx_path().
8190 	 */
8191 	ixgbe_disable_sec_tx_path_generic(hw);
8192 
8193 	/* Disable the TX and RX crypto engines */
8194 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8195 	ctrl |= IXGBE_SECTXCTRL_SECTX_DIS;
8196 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8197 
8198 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8199 	ctrl |= IXGBE_SECRXCTRL_SECRX_DIS;
8200 	IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8201 
8202 	/* Disable SA lookup */
8203 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8204 	ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8205 	ctrl |= IXGBE_LSECTXCTRL_DISABLE;
8206 	IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8207 
8208 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8209 	ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8210 	ctrl |= IXGBE_LSECRXCTRL_DISABLE << IXGBE_LSECRXCTRL_EN_SHIFT;
8211 	IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8212 
8213 	/* Start the data paths */
8214 	ixgbe_enable_sec_rx_path(hw);
8215 	/**
8216 	 * Workaround:
8217 	 * As no ixgbe_enable_sec_rx_path equivalent is
8218 	 * implemented for tx in the base code, and we are
8219 	 * not allowed to modify the base code in DPDK, so
8220 	 * just call the hand-written one directly for now.
8221 	 */
8222 	ixgbe_enable_sec_tx_path_generic(hw);
8223 }
8224 
8225 RTE_PMD_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd);
8226 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
8227 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe, "* igb_uio | uio_pci_generic | vfio-pci");
8228 RTE_PMD_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd);
8229 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);
8230 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe_vf, "* igb_uio | vfio-pci");
8231 RTE_PMD_REGISTER_PARAM_STRING(net_ixgbe_vf,
8232 			      IXGBEVF_DEVARG_PFLINK_FULLCHK "=<0|1>");
8233 
8234 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_init, init, NOTICE);
8235 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_driver, driver, NOTICE);
8236 
8237 #ifdef RTE_ETHDEV_DEBUG_RX
8238 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_rx, rx, DEBUG);
8239 #endif
8240 #ifdef RTE_ETHDEV_DEBUG_TX
8241 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_tx, tx, DEBUG);
8242 #endif
8243