xref: /dpdk/drivers/net/ixgbe/ixgbe_ethdev.c (revision 985e7673)
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 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 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 = PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
1962 		rxq->offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
1963 	} else {
1964 		rxq->vlan_flags = PKT_RX_VLAN;
1965 		rxq->offloads &= ~DEV_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 & DEV_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 & DEV_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 & ETH_VLAN_STRIP_MASK) {
2126 		rxmode = &dev->data->dev_conf.rxmode;
2127 		if (rxmode->offloads & DEV_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 |= DEV_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 &= ~DEV_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 & ETH_VLAN_STRIP_MASK) {
2147 		ixgbe_vlan_hw_strip_config(dev);
2148 	}
2149 
2150 	if (mask & ETH_VLAN_FILTER_MASK) {
2151 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2152 			ixgbe_vlan_hw_filter_enable(dev);
2153 		else
2154 			ixgbe_vlan_hw_filter_disable(dev);
2155 	}
2156 
2157 	if (mask & ETH_VLAN_EXTEND_MASK) {
2158 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2159 			ixgbe_vlan_hw_extend_enable(dev);
2160 		else
2161 			ixgbe_vlan_hw_extend_disable(dev);
2162 	}
2163 
2164 	return 0;
2165 }
2166 
2167 static int
2168 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2169 {
2170 	ixgbe_config_vlan_strip_on_all_queues(dev, mask);
2171 
2172 	ixgbe_vlan_offload_config(dev, mask);
2173 
2174 	return 0;
2175 }
2176 
2177 static void
2178 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2179 {
2180 	struct ixgbe_hw *hw =
2181 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2182 	/* VLNCTRL: enable vlan filtering and allow all vlan tags through */
2183 	uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2184 
2185 	vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
2186 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
2187 }
2188 
2189 static int
2190 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
2191 {
2192 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2193 
2194 	switch (nb_rx_q) {
2195 	case 1:
2196 	case 2:
2197 		RTE_ETH_DEV_SRIOV(dev).active = ETH_64_POOLS;
2198 		break;
2199 	case 4:
2200 		RTE_ETH_DEV_SRIOV(dev).active = ETH_32_POOLS;
2201 		break;
2202 	default:
2203 		return -EINVAL;
2204 	}
2205 
2206 	RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
2207 		IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2208 	RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
2209 		pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2210 	return 0;
2211 }
2212 
2213 static int
2214 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
2215 {
2216 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
2217 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2218 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
2219 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
2220 
2221 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
2222 		/* check multi-queue mode */
2223 		switch (dev_conf->rxmode.mq_mode) {
2224 		case ETH_MQ_RX_VMDQ_DCB:
2225 			PMD_INIT_LOG(INFO, "ETH_MQ_RX_VMDQ_DCB mode supported in SRIOV");
2226 			break;
2227 		case ETH_MQ_RX_VMDQ_DCB_RSS:
2228 			/* DCB/RSS VMDQ in SRIOV mode, not implement yet */
2229 			PMD_INIT_LOG(ERR, "SRIOV active,"
2230 					" unsupported mq_mode rx %d.",
2231 					dev_conf->rxmode.mq_mode);
2232 			return -EINVAL;
2233 		case ETH_MQ_RX_RSS:
2234 		case ETH_MQ_RX_VMDQ_RSS:
2235 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_RSS;
2236 			if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
2237 				if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
2238 					PMD_INIT_LOG(ERR, "SRIOV is active,"
2239 						" invalid queue number"
2240 						" for VMDQ RSS, allowed"
2241 						" value are 1, 2 or 4.");
2242 					return -EINVAL;
2243 				}
2244 			break;
2245 		case ETH_MQ_RX_VMDQ_ONLY:
2246 		case ETH_MQ_RX_NONE:
2247 			/* if nothing mq mode configure, use default scheme */
2248 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
2249 			break;
2250 		default: /* ETH_MQ_RX_DCB, ETH_MQ_RX_DCB_RSS or ETH_MQ_TX_DCB*/
2251 			/* SRIOV only works in VMDq enable mode */
2252 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2253 					" wrong mq_mode rx %d.",
2254 					dev_conf->rxmode.mq_mode);
2255 			return -EINVAL;
2256 		}
2257 
2258 		switch (dev_conf->txmode.mq_mode) {
2259 		case ETH_MQ_TX_VMDQ_DCB:
2260 			PMD_INIT_LOG(INFO, "ETH_MQ_TX_VMDQ_DCB mode supported in SRIOV");
2261 			dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_DCB;
2262 			break;
2263 		default: /* ETH_MQ_TX_VMDQ_ONLY or ETH_MQ_TX_NONE */
2264 			dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_ONLY;
2265 			break;
2266 		}
2267 
2268 		/* check valid queue number */
2269 		if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2270 		    (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2271 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2272 					" nb_rx_q=%d nb_tx_q=%d queue number"
2273 					" must be less than or equal to %d.",
2274 					nb_rx_q, nb_tx_q,
2275 					RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2276 			return -EINVAL;
2277 		}
2278 	} else {
2279 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB_RSS) {
2280 			PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2281 					  " not supported.");
2282 			return -EINVAL;
2283 		}
2284 		/* check configuration for vmdb+dcb mode */
2285 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB) {
2286 			const struct rte_eth_vmdq_dcb_conf *conf;
2287 
2288 			if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2289 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2290 						IXGBE_VMDQ_DCB_NB_QUEUES);
2291 				return -EINVAL;
2292 			}
2293 			conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2294 			if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2295 			       conf->nb_queue_pools == ETH_32_POOLS)) {
2296 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2297 						" nb_queue_pools must be %d or %d.",
2298 						ETH_16_POOLS, ETH_32_POOLS);
2299 				return -EINVAL;
2300 			}
2301 		}
2302 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_VMDQ_DCB) {
2303 			const struct rte_eth_vmdq_dcb_tx_conf *conf;
2304 
2305 			if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2306 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2307 						 IXGBE_VMDQ_DCB_NB_QUEUES);
2308 				return -EINVAL;
2309 			}
2310 			conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2311 			if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2312 			       conf->nb_queue_pools == ETH_32_POOLS)) {
2313 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2314 						" nb_queue_pools != %d and"
2315 						" nb_queue_pools != %d.",
2316 						ETH_16_POOLS, ETH_32_POOLS);
2317 				return -EINVAL;
2318 			}
2319 		}
2320 
2321 		/* For DCB mode check our configuration before we go further */
2322 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB) {
2323 			const struct rte_eth_dcb_rx_conf *conf;
2324 
2325 			conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2326 			if (!(conf->nb_tcs == ETH_4_TCS ||
2327 			       conf->nb_tcs == ETH_8_TCS)) {
2328 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2329 						" and nb_tcs != %d.",
2330 						ETH_4_TCS, ETH_8_TCS);
2331 				return -EINVAL;
2332 			}
2333 		}
2334 
2335 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_DCB) {
2336 			const struct rte_eth_dcb_tx_conf *conf;
2337 
2338 			conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2339 			if (!(conf->nb_tcs == ETH_4_TCS ||
2340 			       conf->nb_tcs == ETH_8_TCS)) {
2341 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2342 						" and nb_tcs != %d.",
2343 						ETH_4_TCS, ETH_8_TCS);
2344 				return -EINVAL;
2345 			}
2346 		}
2347 
2348 		/*
2349 		 * When DCB/VT is off, maximum number of queues changes,
2350 		 * except for 82598EB, which remains constant.
2351 		 */
2352 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
2353 				hw->mac.type != ixgbe_mac_82598EB) {
2354 			if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2355 				PMD_INIT_LOG(ERR,
2356 					     "Neither VT nor DCB are enabled, "
2357 					     "nb_tx_q > %d.",
2358 					     IXGBE_NONE_MODE_TX_NB_QUEUES);
2359 				return -EINVAL;
2360 			}
2361 		}
2362 	}
2363 	return 0;
2364 }
2365 
2366 static int
2367 ixgbe_dev_configure(struct rte_eth_dev *dev)
2368 {
2369 	struct ixgbe_interrupt *intr =
2370 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2371 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2372 	int ret;
2373 
2374 	PMD_INIT_FUNC_TRACE();
2375 
2376 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
2377 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
2378 
2379 	/* multipe queue mode checking */
2380 	ret  = ixgbe_check_mq_mode(dev);
2381 	if (ret != 0) {
2382 		PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2383 			    ret);
2384 		return ret;
2385 	}
2386 
2387 	/* set flag to update link status after init */
2388 	intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2389 
2390 	/*
2391 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2392 	 * allocation or vector Rx preconditions we will reset it.
2393 	 */
2394 	adapter->rx_bulk_alloc_allowed = true;
2395 	adapter->rx_vec_allowed = true;
2396 
2397 	return 0;
2398 }
2399 
2400 static void
2401 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2402 {
2403 	struct ixgbe_hw *hw =
2404 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2405 	struct ixgbe_interrupt *intr =
2406 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2407 	uint32_t gpie;
2408 
2409 	/* only set up it on X550EM_X */
2410 	if (hw->mac.type == ixgbe_mac_X550EM_x) {
2411 		gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2412 		gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2413 		IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2414 		if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2415 			intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2416 	}
2417 }
2418 
2419 int
2420 ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
2421 			uint16_t tx_rate, uint64_t q_msk)
2422 {
2423 	struct ixgbe_hw *hw;
2424 	struct ixgbe_vf_info *vfinfo;
2425 	struct rte_eth_link link;
2426 	uint8_t  nb_q_per_pool;
2427 	uint32_t queue_stride;
2428 	uint32_t queue_idx, idx = 0, vf_idx;
2429 	uint32_t queue_end;
2430 	uint16_t total_rate = 0;
2431 	struct rte_pci_device *pci_dev;
2432 	int ret;
2433 
2434 	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2435 	ret = rte_eth_link_get_nowait(dev->data->port_id, &link);
2436 	if (ret < 0)
2437 		return ret;
2438 
2439 	if (vf >= pci_dev->max_vfs)
2440 		return -EINVAL;
2441 
2442 	if (tx_rate > link.link_speed)
2443 		return -EINVAL;
2444 
2445 	if (q_msk == 0)
2446 		return 0;
2447 
2448 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2449 	vfinfo = *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
2450 	nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2451 	queue_stride = IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2452 	queue_idx = vf * queue_stride;
2453 	queue_end = queue_idx + nb_q_per_pool - 1;
2454 	if (queue_end >= hw->mac.max_tx_queues)
2455 		return -EINVAL;
2456 
2457 	if (vfinfo) {
2458 		for (vf_idx = 0; vf_idx < pci_dev->max_vfs; vf_idx++) {
2459 			if (vf_idx == vf)
2460 				continue;
2461 			for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
2462 				idx++)
2463 				total_rate += vfinfo[vf_idx].tx_rate[idx];
2464 		}
2465 	} else {
2466 		return -EINVAL;
2467 	}
2468 
2469 	/* Store tx_rate for this vf. */
2470 	for (idx = 0; idx < nb_q_per_pool; idx++) {
2471 		if (((uint64_t)0x1 << idx) & q_msk) {
2472 			if (vfinfo[vf].tx_rate[idx] != tx_rate)
2473 				vfinfo[vf].tx_rate[idx] = tx_rate;
2474 			total_rate += tx_rate;
2475 		}
2476 	}
2477 
2478 	if (total_rate > dev->data->dev_link.link_speed) {
2479 		/* Reset stored TX rate of the VF if it causes exceed
2480 		 * link speed.
2481 		 */
2482 		memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
2483 		return -EINVAL;
2484 	}
2485 
2486 	/* Set RTTBCNRC of each queue/pool for vf X  */
2487 	for (; queue_idx <= queue_end; queue_idx++) {
2488 		if (0x1 & q_msk)
2489 			ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
2490 		q_msk = q_msk >> 1;
2491 	}
2492 
2493 	return 0;
2494 }
2495 
2496 static int
2497 ixgbe_flow_ctrl_enable(struct rte_eth_dev *dev, struct ixgbe_hw *hw)
2498 {
2499 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2500 	int err;
2501 	uint32_t mflcn;
2502 
2503 	ixgbe_setup_fc(hw);
2504 
2505 	err = ixgbe_fc_enable(hw);
2506 
2507 	/* Not negotiated is not an error case */
2508 	if (err == IXGBE_SUCCESS || err == IXGBE_ERR_FC_NOT_NEGOTIATED) {
2509 		/*
2510 		 *check if we want to forward MAC frames - driver doesn't
2511 		 *have native capability to do that,
2512 		 *so we'll write the registers ourselves
2513 		 */
2514 
2515 		mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
2516 
2517 		/* set or clear MFLCN.PMCF bit depending on configuration */
2518 		if (adapter->mac_ctrl_frame_fwd != 0)
2519 			mflcn |= IXGBE_MFLCN_PMCF;
2520 		else
2521 			mflcn &= ~IXGBE_MFLCN_PMCF;
2522 
2523 		IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
2524 		IXGBE_WRITE_FLUSH(hw);
2525 
2526 		return 0;
2527 	}
2528 	return err;
2529 }
2530 
2531 /*
2532  * Configure device link speed and setup link.
2533  * It returns 0 on success.
2534  */
2535 static int
2536 ixgbe_dev_start(struct rte_eth_dev *dev)
2537 {
2538 	struct ixgbe_hw *hw =
2539 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2540 	struct ixgbe_vf_info *vfinfo =
2541 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2542 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2543 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2544 	uint32_t intr_vector = 0;
2545 	int err;
2546 	bool link_up = false, negotiate = 0;
2547 	uint32_t speed = 0;
2548 	uint32_t allowed_speeds = 0;
2549 	int mask = 0;
2550 	int status;
2551 	uint16_t vf, idx;
2552 	uint32_t *link_speeds;
2553 	struct ixgbe_tm_conf *tm_conf =
2554 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2555 	struct ixgbe_macsec_setting *macsec_setting =
2556 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
2557 
2558 	PMD_INIT_FUNC_TRACE();
2559 
2560 	/* Stop the link setup handler before resetting the HW. */
2561 	ixgbe_dev_wait_setup_link_complete(dev, 0);
2562 
2563 	/* disable uio/vfio intr/eventfd mapping */
2564 	rte_intr_disable(intr_handle);
2565 
2566 	/* stop adapter */
2567 	hw->adapter_stopped = 0;
2568 	ixgbe_stop_adapter(hw);
2569 
2570 	/* reinitialize adapter
2571 	 * this calls reset and start
2572 	 */
2573 	status = ixgbe_pf_reset_hw(hw);
2574 	if (status != 0)
2575 		return -1;
2576 	hw->mac.ops.start_hw(hw);
2577 	hw->mac.get_link_status = true;
2578 
2579 	/* configure PF module if SRIOV enabled */
2580 	ixgbe_pf_host_configure(dev);
2581 
2582 	ixgbe_dev_phy_intr_setup(dev);
2583 
2584 	/* check and configure queue intr-vector mapping */
2585 	if ((rte_intr_cap_multiple(intr_handle) ||
2586 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
2587 	    dev->data->dev_conf.intr_conf.rxq != 0) {
2588 		intr_vector = dev->data->nb_rx_queues;
2589 		if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2590 			PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2591 					IXGBE_MAX_INTR_QUEUE_NUM);
2592 			return -ENOTSUP;
2593 		}
2594 		if (rte_intr_efd_enable(intr_handle, intr_vector))
2595 			return -1;
2596 	}
2597 
2598 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
2599 		intr_handle->intr_vec =
2600 			rte_zmalloc("intr_vec",
2601 				    dev->data->nb_rx_queues * sizeof(int), 0);
2602 		if (intr_handle->intr_vec == NULL) {
2603 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2604 				     " intr_vec", dev->data->nb_rx_queues);
2605 			return -ENOMEM;
2606 		}
2607 	}
2608 
2609 	/* confiugre msix for sleep until rx interrupt */
2610 	ixgbe_configure_msix(dev);
2611 
2612 	/* initialize transmission unit */
2613 	ixgbe_dev_tx_init(dev);
2614 
2615 	/* This can fail when allocating mbufs for descriptor rings */
2616 	err = ixgbe_dev_rx_init(dev);
2617 	if (err) {
2618 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2619 		goto error;
2620 	}
2621 
2622 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
2623 		ETH_VLAN_EXTEND_MASK;
2624 	err = ixgbe_vlan_offload_config(dev, mask);
2625 	if (err) {
2626 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload");
2627 		goto error;
2628 	}
2629 
2630 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
2631 		/* Enable vlan filtering for VMDq */
2632 		ixgbe_vmdq_vlan_hw_filter_enable(dev);
2633 	}
2634 
2635 	/* Configure DCB hw */
2636 	ixgbe_configure_dcb(dev);
2637 
2638 	if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2639 		err = ixgbe_fdir_configure(dev);
2640 		if (err)
2641 			goto error;
2642 	}
2643 
2644 	/* Restore vf rate limit */
2645 	if (vfinfo != NULL) {
2646 		for (vf = 0; vf < pci_dev->max_vfs; vf++)
2647 			for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2648 				if (vfinfo[vf].tx_rate[idx] != 0)
2649 					ixgbe_set_vf_rate_limit(
2650 						dev, vf,
2651 						vfinfo[vf].tx_rate[idx],
2652 						1 << idx);
2653 	}
2654 
2655 	ixgbe_restore_statistics_mapping(dev);
2656 
2657 	err = ixgbe_flow_ctrl_enable(dev, hw);
2658 	if (err < 0) {
2659 		PMD_INIT_LOG(ERR, "enable flow ctrl err");
2660 		goto error;
2661 	}
2662 
2663 	err = ixgbe_dev_rxtx_start(dev);
2664 	if (err < 0) {
2665 		PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2666 		goto error;
2667 	}
2668 
2669 	/* Skip link setup if loopback mode is enabled. */
2670 	if (dev->data->dev_conf.lpbk_mode != 0) {
2671 		err = ixgbe_check_supported_loopback_mode(dev);
2672 		if (err < 0) {
2673 			PMD_INIT_LOG(ERR, "Unsupported loopback mode");
2674 			goto error;
2675 		} else {
2676 			goto skip_link_setup;
2677 		}
2678 	}
2679 
2680 	if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2681 		err = hw->mac.ops.setup_sfp(hw);
2682 		if (err)
2683 			goto error;
2684 	}
2685 
2686 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2687 		/* Turn on the copper */
2688 		ixgbe_set_phy_power(hw, true);
2689 	} else {
2690 		/* Turn on the laser */
2691 		ixgbe_enable_tx_laser(hw);
2692 	}
2693 
2694 	err = ixgbe_check_link(hw, &speed, &link_up, 0);
2695 	if (err)
2696 		goto error;
2697 	dev->data->dev_link.link_status = link_up;
2698 
2699 	err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2700 	if (err)
2701 		goto error;
2702 
2703 	switch (hw->mac.type) {
2704 	case ixgbe_mac_X550:
2705 	case ixgbe_mac_X550EM_x:
2706 	case ixgbe_mac_X550EM_a:
2707 		allowed_speeds = ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2708 			ETH_LINK_SPEED_2_5G |  ETH_LINK_SPEED_5G |
2709 			ETH_LINK_SPEED_10G;
2710 		if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
2711 				hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
2712 			allowed_speeds = ETH_LINK_SPEED_10M |
2713 				ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G;
2714 		break;
2715 	default:
2716 		allowed_speeds = ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2717 			ETH_LINK_SPEED_10G;
2718 	}
2719 
2720 	link_speeds = &dev->data->dev_conf.link_speeds;
2721 
2722 	/* Ignore autoneg flag bit and check the validity of 
2723 	 * link_speed 
2724 	 */
2725 	if (((*link_speeds) >> 1) & ~(allowed_speeds >> 1)) {
2726 		PMD_INIT_LOG(ERR, "Invalid link setting");
2727 		goto error;
2728 	}
2729 
2730 	speed = 0x0;
2731 	if (*link_speeds == ETH_LINK_SPEED_AUTONEG) {
2732 		switch (hw->mac.type) {
2733 		case ixgbe_mac_82598EB:
2734 			speed = IXGBE_LINK_SPEED_82598_AUTONEG;
2735 			break;
2736 		case ixgbe_mac_82599EB:
2737 		case ixgbe_mac_X540:
2738 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2739 			break;
2740 		case ixgbe_mac_X550:
2741 		case ixgbe_mac_X550EM_x:
2742 		case ixgbe_mac_X550EM_a:
2743 			speed = IXGBE_LINK_SPEED_X550_AUTONEG;
2744 			break;
2745 		default:
2746 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2747 		}
2748 	} else {
2749 		if (*link_speeds & ETH_LINK_SPEED_10G)
2750 			speed |= IXGBE_LINK_SPEED_10GB_FULL;
2751 		if (*link_speeds & ETH_LINK_SPEED_5G)
2752 			speed |= IXGBE_LINK_SPEED_5GB_FULL;
2753 		if (*link_speeds & ETH_LINK_SPEED_2_5G)
2754 			speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
2755 		if (*link_speeds & ETH_LINK_SPEED_1G)
2756 			speed |= IXGBE_LINK_SPEED_1GB_FULL;
2757 		if (*link_speeds & ETH_LINK_SPEED_100M)
2758 			speed |= IXGBE_LINK_SPEED_100_FULL;
2759 		if (*link_speeds & ETH_LINK_SPEED_10M)
2760 			speed |= IXGBE_LINK_SPEED_10_FULL;
2761 	}
2762 
2763 	err = ixgbe_setup_link(hw, speed, link_up);
2764 	if (err)
2765 		goto error;
2766 
2767 skip_link_setup:
2768 
2769 	if (rte_intr_allow_others(intr_handle)) {
2770 		/* check if lsc interrupt is enabled */
2771 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2772 			ixgbe_dev_lsc_interrupt_setup(dev, TRUE);
2773 		else
2774 			ixgbe_dev_lsc_interrupt_setup(dev, FALSE);
2775 		ixgbe_dev_macsec_interrupt_setup(dev);
2776 	} else {
2777 		rte_intr_callback_unregister(intr_handle,
2778 					     ixgbe_dev_interrupt_handler, dev);
2779 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2780 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
2781 				     " no intr multiplex");
2782 	}
2783 
2784 	/* check if rxq interrupt is enabled */
2785 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2786 	    rte_intr_dp_is_en(intr_handle))
2787 		ixgbe_dev_rxq_interrupt_setup(dev);
2788 
2789 	/* enable uio/vfio intr/eventfd mapping */
2790 	rte_intr_enable(intr_handle);
2791 
2792 	/* resume enabled intr since hw reset */
2793 	ixgbe_enable_intr(dev);
2794 	ixgbe_l2_tunnel_conf(dev);
2795 	ixgbe_filter_restore(dev);
2796 
2797 	if (tm_conf->root && !tm_conf->committed)
2798 		PMD_DRV_LOG(WARNING,
2799 			    "please call hierarchy_commit() "
2800 			    "before starting the port");
2801 
2802 	/* wait for the controller to acquire link */
2803 	err = ixgbe_wait_for_link_up(hw);
2804 	if (err)
2805 		goto error;
2806 
2807 	/*
2808 	 * Update link status right before return, because it may
2809 	 * start link configuration process in a separate thread.
2810 	 */
2811 	ixgbe_dev_link_update(dev, 0);
2812 
2813 	/* setup the macsec setting register */
2814 	if (macsec_setting->offload_en)
2815 		ixgbe_dev_macsec_register_enable(dev, macsec_setting);
2816 
2817 	return 0;
2818 
2819 error:
2820 	PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2821 	ixgbe_dev_clear_queues(dev);
2822 	return -EIO;
2823 }
2824 
2825 /*
2826  * Stop device: disable rx and tx functions to allow for reconfiguring.
2827  */
2828 static int
2829 ixgbe_dev_stop(struct rte_eth_dev *dev)
2830 {
2831 	struct rte_eth_link link;
2832 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2833 	struct ixgbe_hw *hw =
2834 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2835 	struct ixgbe_vf_info *vfinfo =
2836 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2837 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2838 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2839 	int vf;
2840 	struct ixgbe_tm_conf *tm_conf =
2841 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2842 
2843 	if (hw->adapter_stopped)
2844 		return 0;
2845 
2846 	PMD_INIT_FUNC_TRACE();
2847 
2848 	ixgbe_dev_wait_setup_link_complete(dev, 0);
2849 
2850 	/* disable interrupts */
2851 	ixgbe_disable_intr(hw);
2852 
2853 	/* reset the NIC */
2854 	ixgbe_pf_reset_hw(hw);
2855 	hw->adapter_stopped = 0;
2856 
2857 	/* stop adapter */
2858 	ixgbe_stop_adapter(hw);
2859 
2860 	for (vf = 0; vfinfo != NULL && vf < pci_dev->max_vfs; vf++)
2861 		vfinfo[vf].clear_to_send = false;
2862 
2863 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2864 		/* Turn off the copper */
2865 		ixgbe_set_phy_power(hw, false);
2866 	} else {
2867 		/* Turn off the laser */
2868 		ixgbe_disable_tx_laser(hw);
2869 	}
2870 
2871 	ixgbe_dev_clear_queues(dev);
2872 
2873 	/* Clear stored conf */
2874 	dev->data->scattered_rx = 0;
2875 	dev->data->lro = 0;
2876 
2877 	/* Clear recorded link status */
2878 	memset(&link, 0, sizeof(link));
2879 	rte_eth_linkstatus_set(dev, &link);
2880 
2881 	if (!rte_intr_allow_others(intr_handle))
2882 		/* resume to the default handler */
2883 		rte_intr_callback_register(intr_handle,
2884 					   ixgbe_dev_interrupt_handler,
2885 					   (void *)dev);
2886 
2887 	/* Clean datapath event and queue/vec mapping */
2888 	rte_intr_efd_disable(intr_handle);
2889 	if (intr_handle->intr_vec != NULL) {
2890 		rte_free(intr_handle->intr_vec);
2891 		intr_handle->intr_vec = NULL;
2892 	}
2893 
2894 	/* reset hierarchy commit */
2895 	tm_conf->committed = false;
2896 
2897 	adapter->rss_reta_updated = 0;
2898 
2899 	hw->adapter_stopped = true;
2900 	dev->data->dev_started = 0;
2901 
2902 	return 0;
2903 }
2904 
2905 /*
2906  * Set device link up: enable tx.
2907  */
2908 static int
2909 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2910 {
2911 	struct ixgbe_hw *hw =
2912 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2913 	if (hw->mac.type == ixgbe_mac_82599EB) {
2914 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2915 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2916 			/* Not suported in bypass mode */
2917 			PMD_INIT_LOG(ERR, "Set link up is not supported "
2918 				     "by device id 0x%x", hw->device_id);
2919 			return -ENOTSUP;
2920 		}
2921 #endif
2922 	}
2923 
2924 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2925 		/* Turn on the copper */
2926 		ixgbe_set_phy_power(hw, true);
2927 	} else {
2928 		/* Turn on the laser */
2929 		ixgbe_enable_tx_laser(hw);
2930 		ixgbe_dev_link_update(dev, 0);
2931 	}
2932 
2933 	return 0;
2934 }
2935 
2936 /*
2937  * Set device link down: disable tx.
2938  */
2939 static int
2940 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2941 {
2942 	struct ixgbe_hw *hw =
2943 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2944 	if (hw->mac.type == ixgbe_mac_82599EB) {
2945 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2946 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2947 			/* Not suported in bypass mode */
2948 			PMD_INIT_LOG(ERR, "Set link down is not supported "
2949 				     "by device id 0x%x", hw->device_id);
2950 			return -ENOTSUP;
2951 		}
2952 #endif
2953 	}
2954 
2955 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2956 		/* Turn off the copper */
2957 		ixgbe_set_phy_power(hw, false);
2958 	} else {
2959 		/* Turn off the laser */
2960 		ixgbe_disable_tx_laser(hw);
2961 		ixgbe_dev_link_update(dev, 0);
2962 	}
2963 
2964 	return 0;
2965 }
2966 
2967 /*
2968  * Reset and stop device.
2969  */
2970 static int
2971 ixgbe_dev_close(struct rte_eth_dev *dev)
2972 {
2973 	struct ixgbe_hw *hw =
2974 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2975 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2976 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2977 	int retries = 0;
2978 	int ret;
2979 
2980 	PMD_INIT_FUNC_TRACE();
2981 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2982 		return 0;
2983 
2984 	ixgbe_pf_reset_hw(hw);
2985 
2986 	ret = ixgbe_dev_stop(dev);
2987 
2988 	ixgbe_dev_free_queues(dev);
2989 
2990 	ixgbe_disable_pcie_master(hw);
2991 
2992 	/* reprogram the RAR[0] in case user changed it. */
2993 	ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
2994 
2995 	/* Unlock any pending hardware semaphore */
2996 	ixgbe_swfw_lock_reset(hw);
2997 
2998 	/* disable uio intr before callback unregister */
2999 	rte_intr_disable(intr_handle);
3000 
3001 	do {
3002 		ret = rte_intr_callback_unregister(intr_handle,
3003 				ixgbe_dev_interrupt_handler, dev);
3004 		if (ret >= 0 || ret == -ENOENT) {
3005 			break;
3006 		} else if (ret != -EAGAIN) {
3007 			PMD_INIT_LOG(ERR,
3008 				"intr callback unregister failed: %d",
3009 				ret);
3010 		}
3011 		rte_delay_ms(100);
3012 	} while (retries++ < (10 + IXGBE_LINK_UP_TIME));
3013 
3014 	/* cancel the delay handler before remove dev */
3015 	rte_eal_alarm_cancel(ixgbe_dev_interrupt_delayed_handler, dev);
3016 
3017 	/* uninitialize PF if max_vfs not zero */
3018 	ixgbe_pf_host_uninit(dev);
3019 
3020 	/* remove all the fdir filters & hash */
3021 	ixgbe_fdir_filter_uninit(dev);
3022 
3023 	/* remove all the L2 tunnel filters & hash */
3024 	ixgbe_l2_tn_filter_uninit(dev);
3025 
3026 	/* Remove all ntuple filters of the device */
3027 	ixgbe_ntuple_filter_uninit(dev);
3028 
3029 	/* clear all the filters list */
3030 	ixgbe_filterlist_flush();
3031 
3032 	/* Remove all Traffic Manager configuration */
3033 	ixgbe_tm_conf_uninit(dev);
3034 
3035 #ifdef RTE_LIB_SECURITY
3036 	rte_free(dev->security_ctx);
3037 #endif
3038 
3039 	return ret;
3040 }
3041 
3042 /*
3043  * Reset PF device.
3044  */
3045 static int
3046 ixgbe_dev_reset(struct rte_eth_dev *dev)
3047 {
3048 	int ret;
3049 
3050 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
3051 	 * its VF to make them align with it. The detailed notification
3052 	 * mechanism is PMD specific. As to ixgbe PF, it is rather complex.
3053 	 * To avoid unexpected behavior in VF, currently reset of PF with
3054 	 * SR-IOV activation is not supported. It might be supported later.
3055 	 */
3056 	if (dev->data->sriov.active)
3057 		return -ENOTSUP;
3058 
3059 	ret = eth_ixgbe_dev_uninit(dev);
3060 	if (ret)
3061 		return ret;
3062 
3063 	ret = eth_ixgbe_dev_init(dev, NULL);
3064 
3065 	return ret;
3066 }
3067 
3068 static void
3069 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
3070 			   struct ixgbe_hw_stats *hw_stats,
3071 			   struct ixgbe_macsec_stats *macsec_stats,
3072 			   uint64_t *total_missed_rx, uint64_t *total_qbrc,
3073 			   uint64_t *total_qprc, uint64_t *total_qprdc)
3074 {
3075 	uint32_t bprc, lxon, lxoff, total;
3076 	uint32_t delta_gprc = 0;
3077 	unsigned i;
3078 	/* Workaround for RX byte count not including CRC bytes when CRC
3079 	 * strip is enabled. CRC bytes are removed from counters when crc_strip
3080 	 * is disabled.
3081 	 */
3082 	int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
3083 			IXGBE_HLREG0_RXCRCSTRP);
3084 
3085 	hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
3086 	hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
3087 	hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
3088 	hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
3089 
3090 	for (i = 0; i < 8; i++) {
3091 		uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
3092 
3093 		/* global total per queue */
3094 		hw_stats->mpc[i] += mp;
3095 		/* Running comprehensive total for stats display */
3096 		*total_missed_rx += hw_stats->mpc[i];
3097 		if (hw->mac.type == ixgbe_mac_82598EB) {
3098 			hw_stats->rnbc[i] +=
3099 			    IXGBE_READ_REG(hw, IXGBE_RNBC(i));
3100 			hw_stats->pxonrxc[i] +=
3101 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
3102 			hw_stats->pxoffrxc[i] +=
3103 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
3104 		} else {
3105 			hw_stats->pxonrxc[i] +=
3106 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
3107 			hw_stats->pxoffrxc[i] +=
3108 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
3109 			hw_stats->pxon2offc[i] +=
3110 				IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
3111 		}
3112 		hw_stats->pxontxc[i] +=
3113 		    IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
3114 		hw_stats->pxofftxc[i] +=
3115 		    IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
3116 	}
3117 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3118 		uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
3119 		uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
3120 		uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
3121 
3122 		delta_gprc += delta_qprc;
3123 
3124 		hw_stats->qprc[i] += delta_qprc;
3125 		hw_stats->qptc[i] += delta_qptc;
3126 
3127 		hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
3128 		hw_stats->qbrc[i] +=
3129 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
3130 		if (crc_strip == 0)
3131 			hw_stats->qbrc[i] -= delta_qprc * RTE_ETHER_CRC_LEN;
3132 
3133 		hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
3134 		hw_stats->qbtc[i] +=
3135 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
3136 
3137 		hw_stats->qprdc[i] += delta_qprdc;
3138 		*total_qprdc += hw_stats->qprdc[i];
3139 
3140 		*total_qprc += hw_stats->qprc[i];
3141 		*total_qbrc += hw_stats->qbrc[i];
3142 	}
3143 	hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
3144 	hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
3145 	hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
3146 
3147 	/*
3148 	 * An errata states that gprc actually counts good + missed packets:
3149 	 * Workaround to set gprc to summated queue packet receives
3150 	 */
3151 	hw_stats->gprc = *total_qprc;
3152 
3153 	if (hw->mac.type != ixgbe_mac_82598EB) {
3154 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
3155 		hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
3156 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
3157 		hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
3158 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
3159 		hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
3160 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
3161 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
3162 	} else {
3163 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
3164 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
3165 		/* 82598 only has a counter in the high register */
3166 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
3167 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
3168 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
3169 	}
3170 	uint64_t old_tpr = hw_stats->tpr;
3171 
3172 	hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
3173 	hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
3174 
3175 	if (crc_strip == 0)
3176 		hw_stats->gorc -= delta_gprc * RTE_ETHER_CRC_LEN;
3177 
3178 	uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
3179 	hw_stats->gptc += delta_gptc;
3180 	hw_stats->gotc -= delta_gptc * RTE_ETHER_CRC_LEN;
3181 	hw_stats->tor -= (hw_stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
3182 
3183 	/*
3184 	 * Workaround: mprc hardware is incorrectly counting
3185 	 * broadcasts, so for now we subtract those.
3186 	 */
3187 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
3188 	hw_stats->bprc += bprc;
3189 	hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
3190 	if (hw->mac.type == ixgbe_mac_82598EB)
3191 		hw_stats->mprc -= bprc;
3192 
3193 	hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
3194 	hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
3195 	hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
3196 	hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
3197 	hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
3198 	hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
3199 
3200 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
3201 	hw_stats->lxontxc += lxon;
3202 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
3203 	hw_stats->lxofftxc += lxoff;
3204 	total = lxon + lxoff;
3205 
3206 	hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
3207 	hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
3208 	hw_stats->gptc -= total;
3209 	hw_stats->mptc -= total;
3210 	hw_stats->ptc64 -= total;
3211 	hw_stats->gotc -= total * RTE_ETHER_MIN_LEN;
3212 
3213 	hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3214 	hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3215 	hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3216 	hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
3217 	hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
3218 	hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
3219 	hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
3220 	hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3221 	hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3222 	hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3223 	hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3224 	hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
3225 	hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3226 	hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
3227 	hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
3228 	hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
3229 	/* Only read FCOE on 82599 */
3230 	if (hw->mac.type != ixgbe_mac_82598EB) {
3231 		hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
3232 		hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
3233 		hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
3234 		hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
3235 		hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
3236 	}
3237 
3238 	/* Flow Director Stats registers */
3239 	if (hw->mac.type != ixgbe_mac_82598EB) {
3240 		hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
3241 		hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
3242 		hw_stats->fdirustat_add += IXGBE_READ_REG(hw,
3243 					IXGBE_FDIRUSTAT) & 0xFFFF;
3244 		hw_stats->fdirustat_remove += (IXGBE_READ_REG(hw,
3245 					IXGBE_FDIRUSTAT) >> 16) & 0xFFFF;
3246 		hw_stats->fdirfstat_fadd += IXGBE_READ_REG(hw,
3247 					IXGBE_FDIRFSTAT) & 0xFFFF;
3248 		hw_stats->fdirfstat_fremove += (IXGBE_READ_REG(hw,
3249 					IXGBE_FDIRFSTAT) >> 16) & 0xFFFF;
3250 	}
3251 	/* MACsec Stats registers */
3252 	macsec_stats->out_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECTXUT);
3253 	macsec_stats->out_pkts_encrypted +=
3254 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTE);
3255 	macsec_stats->out_pkts_protected +=
3256 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTP);
3257 	macsec_stats->out_octets_encrypted +=
3258 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTE);
3259 	macsec_stats->out_octets_protected +=
3260 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTP);
3261 	macsec_stats->in_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECRXUT);
3262 	macsec_stats->in_pkts_badtag += IXGBE_READ_REG(hw, IXGBE_LSECRXBAD);
3263 	macsec_stats->in_pkts_nosci += IXGBE_READ_REG(hw, IXGBE_LSECRXNOSCI);
3264 	macsec_stats->in_pkts_unknownsci +=
3265 		IXGBE_READ_REG(hw, IXGBE_LSECRXUNSCI);
3266 	macsec_stats->in_octets_decrypted +=
3267 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTD);
3268 	macsec_stats->in_octets_validated +=
3269 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTV);
3270 	macsec_stats->in_pkts_unchecked += IXGBE_READ_REG(hw, IXGBE_LSECRXUNCH);
3271 	macsec_stats->in_pkts_delayed += IXGBE_READ_REG(hw, IXGBE_LSECRXDELAY);
3272 	macsec_stats->in_pkts_late += IXGBE_READ_REG(hw, IXGBE_LSECRXLATE);
3273 	for (i = 0; i < 2; i++) {
3274 		macsec_stats->in_pkts_ok +=
3275 			IXGBE_READ_REG(hw, IXGBE_LSECRXOK(i));
3276 		macsec_stats->in_pkts_invalid +=
3277 			IXGBE_READ_REG(hw, IXGBE_LSECRXINV(i));
3278 		macsec_stats->in_pkts_notvalid +=
3279 			IXGBE_READ_REG(hw, IXGBE_LSECRXNV(i));
3280 	}
3281 	macsec_stats->in_pkts_unusedsa += IXGBE_READ_REG(hw, IXGBE_LSECRXUNSA);
3282 	macsec_stats->in_pkts_notusingsa +=
3283 		IXGBE_READ_REG(hw, IXGBE_LSECRXNUSA);
3284 }
3285 
3286 /*
3287  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
3288  */
3289 static int
3290 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3291 {
3292 	struct ixgbe_hw *hw =
3293 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3294 	struct ixgbe_hw_stats *hw_stats =
3295 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3296 	struct ixgbe_macsec_stats *macsec_stats =
3297 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3298 				dev->data->dev_private);
3299 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3300 	unsigned i;
3301 
3302 	total_missed_rx = 0;
3303 	total_qbrc = 0;
3304 	total_qprc = 0;
3305 	total_qprdc = 0;
3306 
3307 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3308 			&total_qbrc, &total_qprc, &total_qprdc);
3309 
3310 	if (stats == NULL)
3311 		return -EINVAL;
3312 
3313 	/* Fill out the rte_eth_stats statistics structure */
3314 	stats->ipackets = total_qprc;
3315 	stats->ibytes = total_qbrc;
3316 	stats->opackets = hw_stats->gptc;
3317 	stats->obytes = hw_stats->gotc;
3318 
3319 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3320 		stats->q_ipackets[i] = hw_stats->qprc[i];
3321 		stats->q_opackets[i] = hw_stats->qptc[i];
3322 		stats->q_ibytes[i] = hw_stats->qbrc[i];
3323 		stats->q_obytes[i] = hw_stats->qbtc[i];
3324 		stats->q_errors[i] = hw_stats->qprdc[i];
3325 	}
3326 
3327 	/* Rx Errors */
3328 	stats->imissed  = total_missed_rx;
3329 	stats->ierrors  = hw_stats->crcerrs +
3330 			  hw_stats->mspdc +
3331 			  hw_stats->rlec +
3332 			  hw_stats->ruc +
3333 			  hw_stats->roc +
3334 			  hw_stats->illerrc +
3335 			  hw_stats->errbc +
3336 			  hw_stats->rfc +
3337 			  hw_stats->fccrc +
3338 			  hw_stats->fclast;
3339 
3340 	/*
3341 	 * 82599 errata, UDP frames with a 0 checksum can be marked as checksum
3342 	 * errors.
3343 	 */
3344 	if (hw->mac.type != ixgbe_mac_82599EB)
3345 		stats->ierrors += hw_stats->xec;
3346 
3347 	/* Tx Errors */
3348 	stats->oerrors  = 0;
3349 	return 0;
3350 }
3351 
3352 static int
3353 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
3354 {
3355 	struct ixgbe_hw_stats *stats =
3356 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3357 
3358 	/* HW registers are cleared on read */
3359 	ixgbe_dev_stats_get(dev, NULL);
3360 
3361 	/* Reset software totals */
3362 	memset(stats, 0, sizeof(*stats));
3363 
3364 	return 0;
3365 }
3366 
3367 /* This function calculates the number of xstats based on the current config */
3368 static unsigned
3369 ixgbe_xstats_calc_num(void) {
3370 	return IXGBE_NB_HW_STATS + IXGBE_NB_MACSEC_STATS +
3371 		(IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
3372 		(IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
3373 }
3374 
3375 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3376 	struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned int size)
3377 {
3378 	const unsigned cnt_stats = ixgbe_xstats_calc_num();
3379 	unsigned stat, i, count;
3380 
3381 	if (xstats_names != NULL) {
3382 		count = 0;
3383 
3384 		/* Note: limit >= cnt_stats checked upstream
3385 		 * in rte_eth_xstats_names()
3386 		 */
3387 
3388 		/* Extended stats from ixgbe_hw_stats */
3389 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3390 			strlcpy(xstats_names[count].name,
3391 				rte_ixgbe_stats_strings[i].name,
3392 				sizeof(xstats_names[count].name));
3393 			count++;
3394 		}
3395 
3396 		/* MACsec Stats */
3397 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3398 			strlcpy(xstats_names[count].name,
3399 				rte_ixgbe_macsec_strings[i].name,
3400 				sizeof(xstats_names[count].name));
3401 			count++;
3402 		}
3403 
3404 		/* RX Priority Stats */
3405 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3406 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3407 				snprintf(xstats_names[count].name,
3408 					sizeof(xstats_names[count].name),
3409 					"rx_priority%u_%s", i,
3410 					rte_ixgbe_rxq_strings[stat].name);
3411 				count++;
3412 			}
3413 		}
3414 
3415 		/* TX Priority Stats */
3416 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3417 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3418 				snprintf(xstats_names[count].name,
3419 					sizeof(xstats_names[count].name),
3420 					"tx_priority%u_%s", i,
3421 					rte_ixgbe_txq_strings[stat].name);
3422 				count++;
3423 			}
3424 		}
3425 	}
3426 	return cnt_stats;
3427 }
3428 
3429 static int ixgbe_dev_xstats_get_names_by_id(
3430 	struct rte_eth_dev *dev,
3431 	const uint64_t *ids,
3432 	struct rte_eth_xstat_name *xstats_names,
3433 	unsigned int limit)
3434 {
3435 	if (!ids) {
3436 		const unsigned int cnt_stats = ixgbe_xstats_calc_num();
3437 		unsigned int stat, i, count;
3438 
3439 		if (xstats_names != NULL) {
3440 			count = 0;
3441 
3442 			/* Note: limit >= cnt_stats checked upstream
3443 			 * in rte_eth_xstats_names()
3444 			 */
3445 
3446 			/* Extended stats from ixgbe_hw_stats */
3447 			for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3448 				strlcpy(xstats_names[count].name,
3449 					rte_ixgbe_stats_strings[i].name,
3450 					sizeof(xstats_names[count].name));
3451 				count++;
3452 			}
3453 
3454 			/* MACsec Stats */
3455 			for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3456 				strlcpy(xstats_names[count].name,
3457 					rte_ixgbe_macsec_strings[i].name,
3458 					sizeof(xstats_names[count].name));
3459 				count++;
3460 			}
3461 
3462 			/* RX Priority Stats */
3463 			for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3464 				for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3465 					snprintf(xstats_names[count].name,
3466 					    sizeof(xstats_names[count].name),
3467 					    "rx_priority%u_%s", i,
3468 					    rte_ixgbe_rxq_strings[stat].name);
3469 					count++;
3470 				}
3471 			}
3472 
3473 			/* TX Priority Stats */
3474 			for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3475 				for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3476 					snprintf(xstats_names[count].name,
3477 					    sizeof(xstats_names[count].name),
3478 					    "tx_priority%u_%s", i,
3479 					    rte_ixgbe_txq_strings[stat].name);
3480 					count++;
3481 				}
3482 			}
3483 		}
3484 		return cnt_stats;
3485 	}
3486 
3487 	uint16_t i;
3488 	uint16_t size = ixgbe_xstats_calc_num();
3489 	struct rte_eth_xstat_name xstats_names_copy[size];
3490 
3491 	ixgbe_dev_xstats_get_names_by_id(dev, NULL, xstats_names_copy,
3492 			size);
3493 
3494 	for (i = 0; i < limit; i++) {
3495 		if (ids[i] >= size) {
3496 			PMD_INIT_LOG(ERR, "id value isn't valid");
3497 			return -1;
3498 		}
3499 		strcpy(xstats_names[i].name,
3500 				xstats_names_copy[ids[i]].name);
3501 	}
3502 	return limit;
3503 }
3504 
3505 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3506 	struct rte_eth_xstat_name *xstats_names, unsigned limit)
3507 {
3508 	unsigned i;
3509 
3510 	if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
3511 		return -ENOMEM;
3512 
3513 	if (xstats_names != NULL)
3514 		for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
3515 			strlcpy(xstats_names[i].name,
3516 				rte_ixgbevf_stats_strings[i].name,
3517 				sizeof(xstats_names[i].name));
3518 	return IXGBEVF_NB_XSTATS;
3519 }
3520 
3521 static int
3522 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3523 					 unsigned n)
3524 {
3525 	struct ixgbe_hw *hw =
3526 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3527 	struct ixgbe_hw_stats *hw_stats =
3528 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3529 	struct ixgbe_macsec_stats *macsec_stats =
3530 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3531 				dev->data->dev_private);
3532 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3533 	unsigned i, stat, count = 0;
3534 
3535 	count = ixgbe_xstats_calc_num();
3536 
3537 	if (n < count)
3538 		return count;
3539 
3540 	total_missed_rx = 0;
3541 	total_qbrc = 0;
3542 	total_qprc = 0;
3543 	total_qprdc = 0;
3544 
3545 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3546 			&total_qbrc, &total_qprc, &total_qprdc);
3547 
3548 	/* If this is a reset xstats is NULL, and we have cleared the
3549 	 * registers by reading them.
3550 	 */
3551 	if (!xstats)
3552 		return 0;
3553 
3554 	/* Extended stats from ixgbe_hw_stats */
3555 	count = 0;
3556 	for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3557 		xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3558 				rte_ixgbe_stats_strings[i].offset);
3559 		xstats[count].id = count;
3560 		count++;
3561 	}
3562 
3563 	/* MACsec Stats */
3564 	for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3565 		xstats[count].value = *(uint64_t *)(((char *)macsec_stats) +
3566 				rte_ixgbe_macsec_strings[i].offset);
3567 		xstats[count].id = count;
3568 		count++;
3569 	}
3570 
3571 	/* RX Priority Stats */
3572 	for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3573 		for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3574 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3575 					rte_ixgbe_rxq_strings[stat].offset +
3576 					(sizeof(uint64_t) * i));
3577 			xstats[count].id = count;
3578 			count++;
3579 		}
3580 	}
3581 
3582 	/* TX Priority Stats */
3583 	for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3584 		for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3585 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3586 					rte_ixgbe_txq_strings[stat].offset +
3587 					(sizeof(uint64_t) * i));
3588 			xstats[count].id = count;
3589 			count++;
3590 		}
3591 	}
3592 	return count;
3593 }
3594 
3595 static int
3596 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
3597 		uint64_t *values, unsigned int n)
3598 {
3599 	if (!ids) {
3600 		struct ixgbe_hw *hw =
3601 				IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3602 		struct ixgbe_hw_stats *hw_stats =
3603 				IXGBE_DEV_PRIVATE_TO_STATS(
3604 						dev->data->dev_private);
3605 		struct ixgbe_macsec_stats *macsec_stats =
3606 				IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3607 					dev->data->dev_private);
3608 		uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3609 		unsigned int i, stat, count = 0;
3610 
3611 		count = ixgbe_xstats_calc_num();
3612 
3613 		if (!ids && n < count)
3614 			return count;
3615 
3616 		total_missed_rx = 0;
3617 		total_qbrc = 0;
3618 		total_qprc = 0;
3619 		total_qprdc = 0;
3620 
3621 		ixgbe_read_stats_registers(hw, hw_stats, macsec_stats,
3622 				&total_missed_rx, &total_qbrc, &total_qprc,
3623 				&total_qprdc);
3624 
3625 		/* If this is a reset xstats is NULL, and we have cleared the
3626 		 * registers by reading them.
3627 		 */
3628 		if (!ids && !values)
3629 			return 0;
3630 
3631 		/* Extended stats from ixgbe_hw_stats */
3632 		count = 0;
3633 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3634 			values[count] = *(uint64_t *)(((char *)hw_stats) +
3635 					rte_ixgbe_stats_strings[i].offset);
3636 			count++;
3637 		}
3638 
3639 		/* MACsec Stats */
3640 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3641 			values[count] = *(uint64_t *)(((char *)macsec_stats) +
3642 					rte_ixgbe_macsec_strings[i].offset);
3643 			count++;
3644 		}
3645 
3646 		/* RX Priority Stats */
3647 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3648 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3649 				values[count] =
3650 					*(uint64_t *)(((char *)hw_stats) +
3651 					rte_ixgbe_rxq_strings[stat].offset +
3652 					(sizeof(uint64_t) * i));
3653 				count++;
3654 			}
3655 		}
3656 
3657 		/* TX Priority Stats */
3658 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3659 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3660 				values[count] =
3661 					*(uint64_t *)(((char *)hw_stats) +
3662 					rte_ixgbe_txq_strings[stat].offset +
3663 					(sizeof(uint64_t) * i));
3664 				count++;
3665 			}
3666 		}
3667 		return count;
3668 	}
3669 
3670 	uint16_t i;
3671 	uint16_t size = ixgbe_xstats_calc_num();
3672 	uint64_t values_copy[size];
3673 
3674 	ixgbe_dev_xstats_get_by_id(dev, NULL, values_copy, size);
3675 
3676 	for (i = 0; i < n; i++) {
3677 		if (ids[i] >= size) {
3678 			PMD_INIT_LOG(ERR, "id value isn't valid");
3679 			return -1;
3680 		}
3681 		values[i] = values_copy[ids[i]];
3682 	}
3683 	return n;
3684 }
3685 
3686 static int
3687 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
3688 {
3689 	struct ixgbe_hw_stats *stats =
3690 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3691 	struct ixgbe_macsec_stats *macsec_stats =
3692 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3693 				dev->data->dev_private);
3694 
3695 	unsigned count = ixgbe_xstats_calc_num();
3696 
3697 	/* HW registers are cleared on read */
3698 	ixgbe_dev_xstats_get(dev, NULL, count);
3699 
3700 	/* Reset software totals */
3701 	memset(stats, 0, sizeof(*stats));
3702 	memset(macsec_stats, 0, sizeof(*macsec_stats));
3703 
3704 	return 0;
3705 }
3706 
3707 static void
3708 ixgbevf_update_stats(struct rte_eth_dev *dev)
3709 {
3710 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3711 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3712 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3713 
3714 	/* Good Rx packet, include VF loopback */
3715 	UPDATE_VF_STAT(IXGBE_VFGPRC,
3716 	    hw_stats->last_vfgprc, hw_stats->vfgprc);
3717 
3718 	/* Good Rx octets, include VF loopback */
3719 	UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
3720 	    hw_stats->last_vfgorc, hw_stats->vfgorc);
3721 
3722 	/* Good Tx packet, include VF loopback */
3723 	UPDATE_VF_STAT(IXGBE_VFGPTC,
3724 	    hw_stats->last_vfgptc, hw_stats->vfgptc);
3725 
3726 	/* Good Tx octets, include VF loopback */
3727 	UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
3728 	    hw_stats->last_vfgotc, hw_stats->vfgotc);
3729 
3730 	/* Rx Multicst Packet */
3731 	UPDATE_VF_STAT(IXGBE_VFMPRC,
3732 	    hw_stats->last_vfmprc, hw_stats->vfmprc);
3733 }
3734 
3735 static int
3736 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3737 		       unsigned n)
3738 {
3739 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3740 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3741 	unsigned i;
3742 
3743 	if (n < IXGBEVF_NB_XSTATS)
3744 		return IXGBEVF_NB_XSTATS;
3745 
3746 	ixgbevf_update_stats(dev);
3747 
3748 	if (!xstats)
3749 		return 0;
3750 
3751 	/* Extended stats */
3752 	for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
3753 		xstats[i].id = i;
3754 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
3755 			rte_ixgbevf_stats_strings[i].offset);
3756 	}
3757 
3758 	return IXGBEVF_NB_XSTATS;
3759 }
3760 
3761 static int
3762 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3763 {
3764 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3765 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3766 
3767 	ixgbevf_update_stats(dev);
3768 
3769 	if (stats == NULL)
3770 		return -EINVAL;
3771 
3772 	stats->ipackets = hw_stats->vfgprc;
3773 	stats->ibytes = hw_stats->vfgorc;
3774 	stats->opackets = hw_stats->vfgptc;
3775 	stats->obytes = hw_stats->vfgotc;
3776 	return 0;
3777 }
3778 
3779 static int
3780 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3781 {
3782 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3783 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3784 
3785 	/* Sync HW register to the last stats */
3786 	ixgbevf_dev_stats_get(dev, NULL);
3787 
3788 	/* reset HW current stats*/
3789 	hw_stats->vfgprc = 0;
3790 	hw_stats->vfgorc = 0;
3791 	hw_stats->vfgptc = 0;
3792 	hw_stats->vfgotc = 0;
3793 	hw_stats->vfmprc = 0;
3794 
3795 	return 0;
3796 }
3797 
3798 static int
3799 ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
3800 {
3801 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3802 	u16 eeprom_verh, eeprom_verl;
3803 	u32 etrack_id;
3804 	int ret;
3805 
3806 	ixgbe_read_eeprom(hw, 0x2e, &eeprom_verh);
3807 	ixgbe_read_eeprom(hw, 0x2d, &eeprom_verl);
3808 
3809 	etrack_id = (eeprom_verh << 16) | eeprom_verl;
3810 	ret = snprintf(fw_version, fw_size, "0x%08x", etrack_id);
3811 	if (ret < 0)
3812 		return -EINVAL;
3813 
3814 	ret += 1; /* add the size of '\0' */
3815 	if (fw_size < (size_t)ret)
3816 		return ret;
3817 	else
3818 		return 0;
3819 }
3820 
3821 static int
3822 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3823 {
3824 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3825 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3826 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3827 
3828 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3829 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3830 	if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3831 		/*
3832 		 * When DCB/VT is off, maximum number of queues changes,
3833 		 * except for 82598EB, which remains constant.
3834 		 */
3835 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
3836 				hw->mac.type != ixgbe_mac_82598EB)
3837 			dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3838 	}
3839 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3840 	dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3841 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3842 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3843 	dev_info->max_vfs = pci_dev->max_vfs;
3844 	if (hw->mac.type == ixgbe_mac_82598EB)
3845 		dev_info->max_vmdq_pools = ETH_16_POOLS;
3846 	else
3847 		dev_info->max_vmdq_pools = ETH_64_POOLS;
3848 	dev_info->max_mtu =  dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3849 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
3850 	dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3851 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3852 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3853 				     dev_info->rx_queue_offload_capa);
3854 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3855 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3856 
3857 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3858 		.rx_thresh = {
3859 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3860 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3861 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3862 		},
3863 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3864 		.rx_drop_en = 0,
3865 		.offloads = 0,
3866 	};
3867 
3868 	dev_info->default_txconf = (struct rte_eth_txconf) {
3869 		.tx_thresh = {
3870 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3871 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3872 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3873 		},
3874 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3875 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3876 		.offloads = 0,
3877 	};
3878 
3879 	dev_info->rx_desc_lim = rx_desc_lim;
3880 	dev_info->tx_desc_lim = tx_desc_lim;
3881 
3882 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3883 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3884 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3885 
3886 	dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G;
3887 	if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
3888 			hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
3889 		dev_info->speed_capa = ETH_LINK_SPEED_10M |
3890 			ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G;
3891 
3892 	if (hw->mac.type == ixgbe_mac_X540 ||
3893 	    hw->mac.type == ixgbe_mac_X540_vf ||
3894 	    hw->mac.type == ixgbe_mac_X550 ||
3895 	    hw->mac.type == ixgbe_mac_X550_vf) {
3896 		dev_info->speed_capa |= ETH_LINK_SPEED_100M;
3897 	}
3898 	if (hw->mac.type == ixgbe_mac_X550) {
3899 		dev_info->speed_capa |= ETH_LINK_SPEED_2_5G;
3900 		dev_info->speed_capa |= ETH_LINK_SPEED_5G;
3901 	}
3902 
3903 	/* Driver-preferred Rx/Tx parameters */
3904 	dev_info->default_rxportconf.burst_size = 32;
3905 	dev_info->default_txportconf.burst_size = 32;
3906 	dev_info->default_rxportconf.nb_queues = 1;
3907 	dev_info->default_txportconf.nb_queues = 1;
3908 	dev_info->default_rxportconf.ring_size = 256;
3909 	dev_info->default_txportconf.ring_size = 256;
3910 
3911 	return 0;
3912 }
3913 
3914 static const uint32_t *
3915 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3916 {
3917 	static const uint32_t ptypes[] = {
3918 		/* For non-vec functions,
3919 		 * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3920 		 * for vec functions,
3921 		 * refers to _recv_raw_pkts_vec().
3922 		 */
3923 		RTE_PTYPE_L2_ETHER,
3924 		RTE_PTYPE_L3_IPV4,
3925 		RTE_PTYPE_L3_IPV4_EXT,
3926 		RTE_PTYPE_L3_IPV6,
3927 		RTE_PTYPE_L3_IPV6_EXT,
3928 		RTE_PTYPE_L4_SCTP,
3929 		RTE_PTYPE_L4_TCP,
3930 		RTE_PTYPE_L4_UDP,
3931 		RTE_PTYPE_TUNNEL_IP,
3932 		RTE_PTYPE_INNER_L3_IPV6,
3933 		RTE_PTYPE_INNER_L3_IPV6_EXT,
3934 		RTE_PTYPE_INNER_L4_TCP,
3935 		RTE_PTYPE_INNER_L4_UDP,
3936 		RTE_PTYPE_UNKNOWN
3937 	};
3938 
3939 	if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3940 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3941 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3942 	    dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3943 		return ptypes;
3944 
3945 #if defined(RTE_ARCH_X86) || defined(__ARM_NEON)
3946 	if (dev->rx_pkt_burst == ixgbe_recv_pkts_vec ||
3947 	    dev->rx_pkt_burst == ixgbe_recv_scattered_pkts_vec)
3948 		return ptypes;
3949 #endif
3950 	return NULL;
3951 }
3952 
3953 static int
3954 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3955 		     struct rte_eth_dev_info *dev_info)
3956 {
3957 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3958 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3959 
3960 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3961 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3962 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3963 	dev_info->max_rx_pktlen = 9728; /* includes CRC, cf MAXFRS reg */
3964 	dev_info->max_mtu = dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3965 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3966 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3967 	dev_info->max_vfs = pci_dev->max_vfs;
3968 	if (hw->mac.type == ixgbe_mac_82598EB)
3969 		dev_info->max_vmdq_pools = ETH_16_POOLS;
3970 	else
3971 		dev_info->max_vmdq_pools = ETH_64_POOLS;
3972 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3973 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3974 				     dev_info->rx_queue_offload_capa);
3975 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3976 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3977 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3978 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3979 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3980 
3981 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3982 		.rx_thresh = {
3983 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3984 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3985 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3986 		},
3987 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3988 		.rx_drop_en = 0,
3989 		.offloads = 0,
3990 	};
3991 
3992 	dev_info->default_txconf = (struct rte_eth_txconf) {
3993 		.tx_thresh = {
3994 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3995 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3996 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3997 		},
3998 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3999 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
4000 		.offloads = 0,
4001 	};
4002 
4003 	dev_info->rx_desc_lim = rx_desc_lim;
4004 	dev_info->tx_desc_lim = tx_desc_lim;
4005 
4006 	return 0;
4007 }
4008 
4009 static int
4010 ixgbevf_check_link(struct ixgbe_hw *hw, ixgbe_link_speed *speed,
4011 		   bool *link_up, int wait_to_complete)
4012 {
4013 	struct ixgbe_adapter *adapter = container_of(hw,
4014 						     struct ixgbe_adapter, hw);
4015 	struct ixgbe_mbx_info *mbx = &hw->mbx;
4016 	struct ixgbe_mac_info *mac = &hw->mac;
4017 	uint32_t links_reg, in_msg;
4018 	int ret_val = 0;
4019 
4020 	/* If we were hit with a reset drop the link */
4021 	if (!mbx->ops.check_for_rst(hw, 0) || !mbx->timeout)
4022 		mac->get_link_status = true;
4023 
4024 	if (!mac->get_link_status)
4025 		goto out;
4026 
4027 	/* if link status is down no point in checking to see if pf is up */
4028 	links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4029 	if (!(links_reg & IXGBE_LINKS_UP))
4030 		goto out;
4031 
4032 	/* for SFP+ modules and DA cables on 82599 it can take up to 500usecs
4033 	 * before the link status is correct
4034 	 */
4035 	if (mac->type == ixgbe_mac_82599_vf && wait_to_complete) {
4036 		int i;
4037 
4038 		for (i = 0; i < 5; i++) {
4039 			rte_delay_us(100);
4040 			links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4041 
4042 			if (!(links_reg & IXGBE_LINKS_UP))
4043 				goto out;
4044 		}
4045 	}
4046 
4047 	switch (links_reg & IXGBE_LINKS_SPEED_82599) {
4048 	case IXGBE_LINKS_SPEED_10G_82599:
4049 		*speed = IXGBE_LINK_SPEED_10GB_FULL;
4050 		if (hw->mac.type >= ixgbe_mac_X550) {
4051 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4052 				*speed = IXGBE_LINK_SPEED_2_5GB_FULL;
4053 		}
4054 		break;
4055 	case IXGBE_LINKS_SPEED_1G_82599:
4056 		*speed = IXGBE_LINK_SPEED_1GB_FULL;
4057 		break;
4058 	case IXGBE_LINKS_SPEED_100_82599:
4059 		*speed = IXGBE_LINK_SPEED_100_FULL;
4060 		if (hw->mac.type == ixgbe_mac_X550) {
4061 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4062 				*speed = IXGBE_LINK_SPEED_5GB_FULL;
4063 		}
4064 		break;
4065 	case IXGBE_LINKS_SPEED_10_X550EM_A:
4066 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
4067 		/* Since Reserved in older MAC's */
4068 		if (hw->mac.type >= ixgbe_mac_X550)
4069 			*speed = IXGBE_LINK_SPEED_10_FULL;
4070 		break;
4071 	default:
4072 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
4073 	}
4074 
4075 	if (wait_to_complete == 0 && adapter->pflink_fullchk == 0) {
4076 		if (*speed == IXGBE_LINK_SPEED_UNKNOWN)
4077 			mac->get_link_status = true;
4078 		else
4079 			mac->get_link_status = false;
4080 
4081 		goto out;
4082 	}
4083 
4084 	/* if the read failed it could just be a mailbox collision, best wait
4085 	 * until we are called again and don't report an error
4086 	 */
4087 	if (mbx->ops.read(hw, &in_msg, 1, 0))
4088 		goto out;
4089 
4090 	if (!(in_msg & IXGBE_VT_MSGTYPE_CTS)) {
4091 		/* msg is not CTS and is NACK we must have lost CTS status */
4092 		if (in_msg & IXGBE_VT_MSGTYPE_NACK)
4093 			mac->get_link_status = false;
4094 		goto out;
4095 	}
4096 
4097 	/* the pf is talking, if we timed out in the past we reinit */
4098 	if (!mbx->timeout) {
4099 		ret_val = -1;
4100 		goto out;
4101 	}
4102 
4103 	/* if we passed all the tests above then the link is up and we no
4104 	 * longer need to check for link
4105 	 */
4106 	mac->get_link_status = false;
4107 
4108 out:
4109 	*link_up = !mac->get_link_status;
4110 	return ret_val;
4111 }
4112 
4113 /*
4114  * If @timeout_ms was 0, it means that it will not return until link complete.
4115  * It returns 1 on complete, return 0 on timeout.
4116  */
4117 static int
4118 ixgbe_dev_wait_setup_link_complete(struct rte_eth_dev *dev, uint32_t timeout_ms)
4119 {
4120 #define WARNING_TIMEOUT    9000 /* 9s  in total */
4121 	struct ixgbe_adapter *ad = dev->data->dev_private;
4122 	uint32_t timeout = timeout_ms ? timeout_ms : WARNING_TIMEOUT;
4123 
4124 	while (rte_atomic32_read(&ad->link_thread_running)) {
4125 		msec_delay(1);
4126 		timeout--;
4127 
4128 		if (timeout_ms) {
4129 			if (!timeout)
4130 				return 0;
4131 		} else if (!timeout) {
4132 			/* It will not return until link complete */
4133 			timeout = WARNING_TIMEOUT;
4134 			PMD_DRV_LOG(ERR, "IXGBE link thread not complete too long time!");
4135 		}
4136 	}
4137 
4138 	return 1;
4139 }
4140 
4141 static void *
4142 ixgbe_dev_setup_link_thread_handler(void *param)
4143 {
4144 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4145 	struct ixgbe_adapter *ad = dev->data->dev_private;
4146 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4147 	struct ixgbe_interrupt *intr =
4148 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4149 	u32 speed;
4150 	bool autoneg = false;
4151 
4152 	pthread_detach(pthread_self());
4153 	speed = hw->phy.autoneg_advertised;
4154 	if (!speed)
4155 		ixgbe_get_link_capabilities(hw, &speed, &autoneg);
4156 
4157 	ixgbe_setup_link(hw, speed, true);
4158 
4159 	intr->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4160 	rte_atomic32_clear(&ad->link_thread_running);
4161 	return NULL;
4162 }
4163 
4164 /*
4165  * In freebsd environment, nic_uio drivers do not support interrupts,
4166  * rte_intr_callback_register() will fail to register interrupts.
4167  * We can not make link status to change from down to up by interrupt
4168  * callback. So we need to wait for the controller to acquire link
4169  * when ports start.
4170  * It returns 0 on link up.
4171  */
4172 static int
4173 ixgbe_wait_for_link_up(struct ixgbe_hw *hw)
4174 {
4175 #ifdef RTE_EXEC_ENV_FREEBSD
4176 	int err, i;
4177 	bool link_up = false;
4178 	uint32_t speed = 0;
4179 	const int nb_iter = 25;
4180 
4181 	for (i = 0; i < nb_iter; i++) {
4182 		err = ixgbe_check_link(hw, &speed, &link_up, 0);
4183 		if (err)
4184 			return err;
4185 		if (link_up)
4186 			return 0;
4187 		msec_delay(200);
4188 	}
4189 
4190 	return 0;
4191 #else
4192 	RTE_SET_USED(hw);
4193 	return 0;
4194 #endif
4195 }
4196 
4197 /* return 0 means link status changed, -1 means not changed */
4198 int
4199 ixgbe_dev_link_update_share(struct rte_eth_dev *dev,
4200 			    int wait_to_complete, int vf)
4201 {
4202 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4203 	struct ixgbe_adapter *ad = dev->data->dev_private;
4204 	struct rte_eth_link link;
4205 	ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
4206 	struct ixgbe_interrupt *intr =
4207 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4208 	bool link_up;
4209 	int diag;
4210 	int wait = 1;
4211 	u32 esdp_reg;
4212 
4213 	memset(&link, 0, sizeof(link));
4214 	link.link_status = ETH_LINK_DOWN;
4215 	link.link_speed = ETH_SPEED_NUM_NONE;
4216 	link.link_duplex = ETH_LINK_HALF_DUPLEX;
4217 	link.link_autoneg = !(dev->data->dev_conf.link_speeds &
4218 			ETH_LINK_SPEED_FIXED);
4219 
4220 	hw->mac.get_link_status = true;
4221 
4222 	if (intr->flags & IXGBE_FLAG_NEED_LINK_CONFIG)
4223 		return rte_eth_linkstatus_set(dev, &link);
4224 
4225 	/* check if it needs to wait to complete, if lsc interrupt is enabled */
4226 	if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
4227 		wait = 0;
4228 
4229 /* BSD has no interrupt mechanism, so force NIC status synchronization. */
4230 #ifdef RTE_EXEC_ENV_FREEBSD
4231 	wait = 1;
4232 #endif
4233 
4234 	if (vf)
4235 		diag = ixgbevf_check_link(hw, &link_speed, &link_up, wait);
4236 	else
4237 		diag = ixgbe_check_link(hw, &link_speed, &link_up, wait);
4238 
4239 	if (diag != 0) {
4240 		link.link_speed = ETH_SPEED_NUM_100M;
4241 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
4242 		return rte_eth_linkstatus_set(dev, &link);
4243 	}
4244 
4245 	if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4246 		esdp_reg = IXGBE_READ_REG(hw, IXGBE_ESDP);
4247 		if ((esdp_reg & IXGBE_ESDP_SDP3))
4248 			link_up = 0;
4249 	}
4250 
4251 	if (link_up == 0) {
4252 		if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4253 			ixgbe_dev_wait_setup_link_complete(dev, 0);
4254 			if (rte_atomic32_test_and_set(&ad->link_thread_running)) {
4255 				/* To avoid race condition between threads, set
4256 				 * the IXGBE_FLAG_NEED_LINK_CONFIG flag only
4257 				 * when there is no link thread running.
4258 				 */
4259 				intr->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
4260 				if (rte_ctrl_thread_create(&ad->link_thread_tid,
4261 					"ixgbe-link-handler",
4262 					NULL,
4263 					ixgbe_dev_setup_link_thread_handler,
4264 					dev) < 0) {
4265 					PMD_DRV_LOG(ERR,
4266 						"Create link thread failed!");
4267 					rte_atomic32_clear(&ad->link_thread_running);
4268 				}
4269 			} else {
4270 				PMD_DRV_LOG(ERR,
4271 					"Other link thread is running now!");
4272 			}
4273 		}
4274 		return rte_eth_linkstatus_set(dev, &link);
4275 	}
4276 
4277 	link.link_status = ETH_LINK_UP;
4278 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
4279 
4280 	switch (link_speed) {
4281 	default:
4282 	case IXGBE_LINK_SPEED_UNKNOWN:
4283 		link.link_speed = ETH_SPEED_NUM_UNKNOWN;
4284 		break;
4285 
4286 	case IXGBE_LINK_SPEED_10_FULL:
4287 		link.link_speed = ETH_SPEED_NUM_10M;
4288 		break;
4289 
4290 	case IXGBE_LINK_SPEED_100_FULL:
4291 		link.link_speed = ETH_SPEED_NUM_100M;
4292 		break;
4293 
4294 	case IXGBE_LINK_SPEED_1GB_FULL:
4295 		link.link_speed = ETH_SPEED_NUM_1G;
4296 		break;
4297 
4298 	case IXGBE_LINK_SPEED_2_5GB_FULL:
4299 		link.link_speed = ETH_SPEED_NUM_2_5G;
4300 		break;
4301 
4302 	case IXGBE_LINK_SPEED_5GB_FULL:
4303 		link.link_speed = ETH_SPEED_NUM_5G;
4304 		break;
4305 
4306 	case IXGBE_LINK_SPEED_10GB_FULL:
4307 		link.link_speed = ETH_SPEED_NUM_10G;
4308 		break;
4309 	}
4310 
4311 	return rte_eth_linkstatus_set(dev, &link);
4312 }
4313 
4314 static int
4315 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4316 {
4317 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 0);
4318 }
4319 
4320 static int
4321 ixgbevf_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4322 {
4323 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 1);
4324 }
4325 
4326 static int
4327 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
4328 {
4329 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4330 	uint32_t fctrl;
4331 
4332 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4333 	fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4334 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4335 
4336 	return 0;
4337 }
4338 
4339 static int
4340 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
4341 {
4342 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4343 	uint32_t fctrl;
4344 
4345 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4346 	fctrl &= (~IXGBE_FCTRL_UPE);
4347 	if (dev->data->all_multicast == 1)
4348 		fctrl |= IXGBE_FCTRL_MPE;
4349 	else
4350 		fctrl &= (~IXGBE_FCTRL_MPE);
4351 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4352 
4353 	return 0;
4354 }
4355 
4356 static int
4357 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
4358 {
4359 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4360 	uint32_t fctrl;
4361 
4362 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4363 	fctrl |= IXGBE_FCTRL_MPE;
4364 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4365 
4366 	return 0;
4367 }
4368 
4369 static int
4370 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
4371 {
4372 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4373 	uint32_t fctrl;
4374 
4375 	if (dev->data->promiscuous == 1)
4376 		return 0; /* must remain in all_multicast mode */
4377 
4378 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4379 	fctrl &= (~IXGBE_FCTRL_MPE);
4380 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4381 
4382 	return 0;
4383 }
4384 
4385 /**
4386  * It clears the interrupt causes and enables the interrupt.
4387  * It will be called once only during nic initialized.
4388  *
4389  * @param dev
4390  *  Pointer to struct rte_eth_dev.
4391  * @param on
4392  *  Enable or Disable.
4393  *
4394  * @return
4395  *  - On success, zero.
4396  *  - On failure, a negative value.
4397  */
4398 static int
4399 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
4400 {
4401 	struct ixgbe_interrupt *intr =
4402 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4403 
4404 	ixgbe_dev_link_status_print(dev);
4405 	if (on)
4406 		intr->mask |= IXGBE_EICR_LSC;
4407 	else
4408 		intr->mask &= ~IXGBE_EICR_LSC;
4409 
4410 	return 0;
4411 }
4412 
4413 /**
4414  * It clears the interrupt causes and enables the interrupt.
4415  * It will be called once only during nic initialized.
4416  *
4417  * @param dev
4418  *  Pointer to struct rte_eth_dev.
4419  *
4420  * @return
4421  *  - On success, zero.
4422  *  - On failure, a negative value.
4423  */
4424 static int
4425 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
4426 {
4427 	struct ixgbe_interrupt *intr =
4428 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4429 
4430 	intr->mask |= IXGBE_EICR_RTX_QUEUE;
4431 
4432 	return 0;
4433 }
4434 
4435 /**
4436  * It clears the interrupt causes and enables the interrupt.
4437  * It will be called once only during nic initialized.
4438  *
4439  * @param dev
4440  *  Pointer to struct rte_eth_dev.
4441  *
4442  * @return
4443  *  - On success, zero.
4444  *  - On failure, a negative value.
4445  */
4446 static int
4447 ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev)
4448 {
4449 	struct ixgbe_interrupt *intr =
4450 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4451 
4452 	intr->mask |= IXGBE_EICR_LINKSEC;
4453 
4454 	return 0;
4455 }
4456 
4457 /*
4458  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
4459  *
4460  * @param dev
4461  *  Pointer to struct rte_eth_dev.
4462  *
4463  * @return
4464  *  - On success, zero.
4465  *  - On failure, a negative value.
4466  */
4467 static int
4468 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
4469 {
4470 	uint32_t eicr;
4471 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4472 	struct ixgbe_interrupt *intr =
4473 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4474 
4475 	/* clear all cause mask */
4476 	ixgbe_disable_intr(hw);
4477 
4478 	/* read-on-clear nic registers here */
4479 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4480 	PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
4481 
4482 	intr->flags = 0;
4483 
4484 	/* set flag for async link update */
4485 	if (eicr & IXGBE_EICR_LSC)
4486 		intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4487 
4488 	if (eicr & IXGBE_EICR_MAILBOX)
4489 		intr->flags |= IXGBE_FLAG_MAILBOX;
4490 
4491 	if (eicr & IXGBE_EICR_LINKSEC)
4492 		intr->flags |= IXGBE_FLAG_MACSEC;
4493 
4494 	if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
4495 	    hw->phy.type == ixgbe_phy_x550em_ext_t &&
4496 	    (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
4497 		intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
4498 
4499 	return 0;
4500 }
4501 
4502 /**
4503  * It gets and then prints the link status.
4504  *
4505  * @param dev
4506  *  Pointer to struct rte_eth_dev.
4507  *
4508  * @return
4509  *  - On success, zero.
4510  *  - On failure, a negative value.
4511  */
4512 static void
4513 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
4514 {
4515 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4516 	struct rte_eth_link link;
4517 
4518 	rte_eth_linkstatus_get(dev, &link);
4519 
4520 	if (link.link_status) {
4521 		PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
4522 					(int)(dev->data->port_id),
4523 					(unsigned)link.link_speed,
4524 			link.link_duplex == ETH_LINK_FULL_DUPLEX ?
4525 					"full-duplex" : "half-duplex");
4526 	} else {
4527 		PMD_INIT_LOG(INFO, " Port %d: Link Down",
4528 				(int)(dev->data->port_id));
4529 	}
4530 	PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
4531 				pci_dev->addr.domain,
4532 				pci_dev->addr.bus,
4533 				pci_dev->addr.devid,
4534 				pci_dev->addr.function);
4535 }
4536 
4537 /*
4538  * It executes link_update after knowing an interrupt occurred.
4539  *
4540  * @param dev
4541  *  Pointer to struct rte_eth_dev.
4542  *
4543  * @return
4544  *  - On success, zero.
4545  *  - On failure, a negative value.
4546  */
4547 static int
4548 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
4549 {
4550 	struct ixgbe_interrupt *intr =
4551 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4552 	int64_t timeout;
4553 	struct ixgbe_hw *hw =
4554 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4555 
4556 	PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
4557 
4558 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
4559 		ixgbe_pf_mbx_process(dev);
4560 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
4561 	}
4562 
4563 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4564 		ixgbe_handle_lasi(hw);
4565 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4566 	}
4567 
4568 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4569 		struct rte_eth_link link;
4570 
4571 		/* get the link status before link update, for predicting later */
4572 		rte_eth_linkstatus_get(dev, &link);
4573 
4574 		ixgbe_dev_link_update(dev, 0);
4575 
4576 		/* likely to up */
4577 		if (!link.link_status)
4578 			/* handle it 1 sec later, wait it being stable */
4579 			timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
4580 		/* likely to down */
4581 		else
4582 			/* handle it 4 sec later, wait it being stable */
4583 			timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
4584 
4585 		ixgbe_dev_link_status_print(dev);
4586 		if (rte_eal_alarm_set(timeout * 1000,
4587 				      ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
4588 			PMD_DRV_LOG(ERR, "Error setting alarm");
4589 		else {
4590 			/* remember original mask */
4591 			intr->mask_original = intr->mask;
4592 			/* only disable lsc interrupt */
4593 			intr->mask &= ~IXGBE_EIMS_LSC;
4594 		}
4595 	}
4596 
4597 	PMD_DRV_LOG(DEBUG, "enable intr immediately");
4598 	ixgbe_enable_intr(dev);
4599 
4600 	return 0;
4601 }
4602 
4603 /**
4604  * Interrupt handler which shall be registered for alarm callback for delayed
4605  * handling specific interrupt to wait for the stable nic state. As the
4606  * NIC interrupt state is not stable for ixgbe after link is just down,
4607  * it needs to wait 4 seconds to get the stable status.
4608  *
4609  * @param handle
4610  *  Pointer to interrupt handle.
4611  * @param param
4612  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4613  *
4614  * @return
4615  *  void
4616  */
4617 static void
4618 ixgbe_dev_interrupt_delayed_handler(void *param)
4619 {
4620 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4621 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4622 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
4623 	struct ixgbe_interrupt *intr =
4624 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4625 	struct ixgbe_hw *hw =
4626 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4627 	uint32_t eicr;
4628 
4629 	ixgbe_disable_intr(hw);
4630 
4631 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4632 	if (eicr & IXGBE_EICR_MAILBOX)
4633 		ixgbe_pf_mbx_process(dev);
4634 
4635 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4636 		ixgbe_handle_lasi(hw);
4637 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4638 	}
4639 
4640 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4641 		ixgbe_dev_link_update(dev, 0);
4642 		intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4643 		ixgbe_dev_link_status_print(dev);
4644 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
4645 	}
4646 
4647 	if (intr->flags & IXGBE_FLAG_MACSEC) {
4648 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_MACSEC, NULL);
4649 		intr->flags &= ~IXGBE_FLAG_MACSEC;
4650 	}
4651 
4652 	/* restore original mask */
4653 	intr->mask = intr->mask_original;
4654 	intr->mask_original = 0;
4655 
4656 	PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
4657 	ixgbe_enable_intr(dev);
4658 	rte_intr_ack(intr_handle);
4659 }
4660 
4661 /**
4662  * Interrupt handler triggered by NIC  for handling
4663  * specific interrupt.
4664  *
4665  * @param handle
4666  *  Pointer to interrupt handle.
4667  * @param param
4668  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4669  *
4670  * @return
4671  *  void
4672  */
4673 static void
4674 ixgbe_dev_interrupt_handler(void *param)
4675 {
4676 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4677 
4678 	ixgbe_dev_interrupt_get_status(dev);
4679 	ixgbe_dev_interrupt_action(dev);
4680 }
4681 
4682 static int
4683 ixgbe_dev_led_on(struct rte_eth_dev *dev)
4684 {
4685 	struct ixgbe_hw *hw;
4686 
4687 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4688 	return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4689 }
4690 
4691 static int
4692 ixgbe_dev_led_off(struct rte_eth_dev *dev)
4693 {
4694 	struct ixgbe_hw *hw;
4695 
4696 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4697 	return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4698 }
4699 
4700 static int
4701 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4702 {
4703 	struct ixgbe_hw *hw;
4704 	uint32_t mflcn_reg;
4705 	uint32_t fccfg_reg;
4706 	int rx_pause;
4707 	int tx_pause;
4708 
4709 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4710 
4711 	fc_conf->pause_time = hw->fc.pause_time;
4712 	fc_conf->high_water = hw->fc.high_water[0];
4713 	fc_conf->low_water = hw->fc.low_water[0];
4714 	fc_conf->send_xon = hw->fc.send_xon;
4715 	fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
4716 
4717 	/*
4718 	 * Return rx_pause status according to actual setting of
4719 	 * MFLCN register.
4720 	 */
4721 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4722 	if (mflcn_reg & IXGBE_MFLCN_PMCF)
4723 		fc_conf->mac_ctrl_frame_fwd = 1;
4724 	else
4725 		fc_conf->mac_ctrl_frame_fwd = 0;
4726 
4727 	if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
4728 		rx_pause = 1;
4729 	else
4730 		rx_pause = 0;
4731 
4732 	/*
4733 	 * Return tx_pause status according to actual setting of
4734 	 * FCCFG register.
4735 	 */
4736 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4737 	if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
4738 		tx_pause = 1;
4739 	else
4740 		tx_pause = 0;
4741 
4742 	if (rx_pause && tx_pause)
4743 		fc_conf->mode = RTE_FC_FULL;
4744 	else if (rx_pause)
4745 		fc_conf->mode = RTE_FC_RX_PAUSE;
4746 	else if (tx_pause)
4747 		fc_conf->mode = RTE_FC_TX_PAUSE;
4748 	else
4749 		fc_conf->mode = RTE_FC_NONE;
4750 
4751 	return 0;
4752 }
4753 
4754 static int
4755 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4756 {
4757 	struct ixgbe_hw *hw;
4758 	struct ixgbe_adapter *adapter = dev->data->dev_private;
4759 	int err;
4760 	uint32_t rx_buf_size;
4761 	uint32_t max_high_water;
4762 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4763 		ixgbe_fc_none,
4764 		ixgbe_fc_rx_pause,
4765 		ixgbe_fc_tx_pause,
4766 		ixgbe_fc_full
4767 	};
4768 
4769 	PMD_INIT_FUNC_TRACE();
4770 
4771 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4772 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
4773 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4774 
4775 	/*
4776 	 * At least reserve one Ethernet frame for watermark
4777 	 * high_water/low_water in kilo bytes for ixgbe
4778 	 */
4779 	max_high_water = (rx_buf_size -
4780 			RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4781 	if ((fc_conf->high_water > max_high_water) ||
4782 		(fc_conf->high_water < fc_conf->low_water)) {
4783 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4784 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4785 		return -EINVAL;
4786 	}
4787 
4788 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
4789 	hw->fc.pause_time     = fc_conf->pause_time;
4790 	hw->fc.high_water[0]  = fc_conf->high_water;
4791 	hw->fc.low_water[0]   = fc_conf->low_water;
4792 	hw->fc.send_xon       = fc_conf->send_xon;
4793 	hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
4794 	adapter->mac_ctrl_frame_fwd = fc_conf->mac_ctrl_frame_fwd;
4795 
4796 	err = ixgbe_flow_ctrl_enable(dev, hw);
4797 	if (err < 0) {
4798 		PMD_INIT_LOG(ERR, "ixgbe_flow_ctrl_enable = 0x%x", err);
4799 		return -EIO;
4800 	}
4801 	return err;
4802 }
4803 
4804 /**
4805  *  ixgbe_pfc_enable_generic - Enable flow control
4806  *  @hw: pointer to hardware structure
4807  *  @tc_num: traffic class number
4808  *  Enable flow control according to the current settings.
4809  */
4810 static int
4811 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
4812 {
4813 	int ret_val = 0;
4814 	uint32_t mflcn_reg, fccfg_reg;
4815 	uint32_t reg;
4816 	uint32_t fcrtl, fcrth;
4817 	uint8_t i;
4818 	uint8_t nb_rx_en;
4819 
4820 	/* Validate the water mark configuration */
4821 	if (!hw->fc.pause_time) {
4822 		ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4823 		goto out;
4824 	}
4825 
4826 	/* Low water mark of zero causes XOFF floods */
4827 	if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
4828 		 /* High/Low water can not be 0 */
4829 		if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_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 		if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
4836 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4837 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4838 			goto out;
4839 		}
4840 	}
4841 	/* Negotiate the fc mode to use */
4842 	ixgbe_fc_autoneg(hw);
4843 
4844 	/* Disable any previous flow control settings */
4845 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4846 	mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
4847 
4848 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4849 	fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
4850 
4851 	switch (hw->fc.current_mode) {
4852 	case ixgbe_fc_none:
4853 		/*
4854 		 * If the count of enabled RX Priority Flow control >1,
4855 		 * and the TX pause can not be disabled
4856 		 */
4857 		nb_rx_en = 0;
4858 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4859 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4860 			if (reg & IXGBE_FCRTH_FCEN)
4861 				nb_rx_en++;
4862 		}
4863 		if (nb_rx_en > 1)
4864 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4865 		break;
4866 	case ixgbe_fc_rx_pause:
4867 		/*
4868 		 * Rx Flow control is enabled and Tx Flow control is
4869 		 * disabled by software override. Since there really
4870 		 * isn't a way to advertise that we are capable of RX
4871 		 * Pause ONLY, we will advertise that we support both
4872 		 * symmetric and asymmetric Rx PAUSE.  Later, we will
4873 		 * disable the adapter's ability to send PAUSE frames.
4874 		 */
4875 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4876 		/*
4877 		 * If the count of enabled RX Priority Flow control >1,
4878 		 * and the TX pause can not be disabled
4879 		 */
4880 		nb_rx_en = 0;
4881 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4882 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4883 			if (reg & IXGBE_FCRTH_FCEN)
4884 				nb_rx_en++;
4885 		}
4886 		if (nb_rx_en > 1)
4887 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4888 		break;
4889 	case ixgbe_fc_tx_pause:
4890 		/*
4891 		 * Tx Flow control is enabled, and Rx Flow control is
4892 		 * disabled by software override.
4893 		 */
4894 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4895 		break;
4896 	case ixgbe_fc_full:
4897 		/* Flow control (both Rx and Tx) is enabled by SW override. */
4898 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4899 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4900 		break;
4901 	default:
4902 		PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
4903 		ret_val = IXGBE_ERR_CONFIG;
4904 		goto out;
4905 	}
4906 
4907 	/* Set 802.3x based flow control settings. */
4908 	mflcn_reg |= IXGBE_MFLCN_DPF;
4909 	IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
4910 	IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
4911 
4912 	/* Set up and enable Rx high/low water mark thresholds, enable XON. */
4913 	if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
4914 		hw->fc.high_water[tc_num]) {
4915 		fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
4916 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
4917 		fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
4918 	} else {
4919 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
4920 		/*
4921 		 * In order to prevent Tx hangs when the internal Tx
4922 		 * switch is enabled we must set the high water mark
4923 		 * to the maximum FCRTH value.  This allows the Tx
4924 		 * switch to function even under heavy Rx workloads.
4925 		 */
4926 		fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
4927 	}
4928 	IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
4929 
4930 	/* Configure pause time (2 TCs per register) */
4931 	reg = hw->fc.pause_time * 0x00010001;
4932 	for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
4933 		IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
4934 
4935 	/* Configure flow control refresh threshold value */
4936 	IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
4937 
4938 out:
4939 	return ret_val;
4940 }
4941 
4942 static int
4943 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
4944 {
4945 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4946 	int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
4947 
4948 	if (hw->mac.type != ixgbe_mac_82598EB) {
4949 		ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
4950 	}
4951 	return ret_val;
4952 }
4953 
4954 static int
4955 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
4956 {
4957 	int err;
4958 	uint32_t rx_buf_size;
4959 	uint32_t max_high_water;
4960 	uint8_t tc_num;
4961 	uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
4962 	struct ixgbe_hw *hw =
4963 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4964 	struct ixgbe_dcb_config *dcb_config =
4965 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
4966 
4967 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4968 		ixgbe_fc_none,
4969 		ixgbe_fc_rx_pause,
4970 		ixgbe_fc_tx_pause,
4971 		ixgbe_fc_full
4972 	};
4973 
4974 	PMD_INIT_FUNC_TRACE();
4975 
4976 	ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
4977 	tc_num = map[pfc_conf->priority];
4978 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
4979 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4980 	/*
4981 	 * At least reserve one Ethernet frame for watermark
4982 	 * high_water/low_water in kilo bytes for ixgbe
4983 	 */
4984 	max_high_water = (rx_buf_size -
4985 			RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4986 	if ((pfc_conf->fc.high_water > max_high_water) ||
4987 	    (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
4988 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4989 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4990 		return -EINVAL;
4991 	}
4992 
4993 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
4994 	hw->fc.pause_time = pfc_conf->fc.pause_time;
4995 	hw->fc.send_xon = pfc_conf->fc.send_xon;
4996 	hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
4997 	hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
4998 
4999 	err = ixgbe_dcb_pfc_enable(dev, tc_num);
5000 
5001 	/* Not negotiated is not an error case */
5002 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
5003 		return 0;
5004 
5005 	PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
5006 	return -EIO;
5007 }
5008 
5009 static int
5010 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
5011 			  struct rte_eth_rss_reta_entry64 *reta_conf,
5012 			  uint16_t reta_size)
5013 {
5014 	uint16_t i, sp_reta_size;
5015 	uint8_t j, mask;
5016 	uint32_t reta, r;
5017 	uint16_t idx, shift;
5018 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5019 	struct rte_eth_dev_data *dev_data = dev->data;
5020 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5021 	uint32_t reta_reg;
5022 
5023 	PMD_INIT_FUNC_TRACE();
5024 
5025 	if (!dev_data->dev_started) {
5026 		PMD_DRV_LOG(ERR,
5027 			"port %d must be started before rss reta update",
5028 			 dev_data->port_id);
5029 		return -EIO;
5030 	}
5031 
5032 	if (!ixgbe_rss_update_sp(hw->mac.type)) {
5033 		PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
5034 			"NIC.");
5035 		return -ENOTSUP;
5036 	}
5037 
5038 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5039 	if (reta_size != sp_reta_size) {
5040 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5041 			"(%d) doesn't match the number hardware can supported "
5042 			"(%d)", reta_size, sp_reta_size);
5043 		return -EINVAL;
5044 	}
5045 
5046 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5047 		idx = i / RTE_RETA_GROUP_SIZE;
5048 		shift = i % RTE_RETA_GROUP_SIZE;
5049 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5050 						IXGBE_4_BIT_MASK);
5051 		if (!mask)
5052 			continue;
5053 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5054 		if (mask == IXGBE_4_BIT_MASK)
5055 			r = 0;
5056 		else
5057 			r = IXGBE_READ_REG(hw, reta_reg);
5058 		for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5059 			if (mask & (0x1 << j))
5060 				reta |= reta_conf[idx].reta[shift + j] <<
5061 							(CHAR_BIT * j);
5062 			else
5063 				reta |= r & (IXGBE_8_BIT_MASK <<
5064 						(CHAR_BIT * j));
5065 		}
5066 		IXGBE_WRITE_REG(hw, reta_reg, reta);
5067 	}
5068 	adapter->rss_reta_updated = 1;
5069 
5070 	return 0;
5071 }
5072 
5073 static int
5074 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
5075 			 struct rte_eth_rss_reta_entry64 *reta_conf,
5076 			 uint16_t reta_size)
5077 {
5078 	uint16_t i, sp_reta_size;
5079 	uint8_t j, mask;
5080 	uint32_t reta;
5081 	uint16_t idx, shift;
5082 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5083 	uint32_t reta_reg;
5084 
5085 	PMD_INIT_FUNC_TRACE();
5086 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5087 	if (reta_size != sp_reta_size) {
5088 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5089 			"(%d) doesn't match the number hardware can supported "
5090 			"(%d)", reta_size, sp_reta_size);
5091 		return -EINVAL;
5092 	}
5093 
5094 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5095 		idx = i / RTE_RETA_GROUP_SIZE;
5096 		shift = i % RTE_RETA_GROUP_SIZE;
5097 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5098 						IXGBE_4_BIT_MASK);
5099 		if (!mask)
5100 			continue;
5101 
5102 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5103 		reta = IXGBE_READ_REG(hw, reta_reg);
5104 		for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5105 			if (mask & (0x1 << j))
5106 				reta_conf[idx].reta[shift + j] =
5107 					((reta >> (CHAR_BIT * j)) &
5108 						IXGBE_8_BIT_MASK);
5109 		}
5110 	}
5111 
5112 	return 0;
5113 }
5114 
5115 static int
5116 ixgbe_add_rar(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
5117 				uint32_t index, uint32_t pool)
5118 {
5119 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5120 	uint32_t enable_addr = 1;
5121 
5122 	return ixgbe_set_rar(hw, index, mac_addr->addr_bytes,
5123 			     pool, enable_addr);
5124 }
5125 
5126 static void
5127 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
5128 {
5129 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5130 
5131 	ixgbe_clear_rar(hw, index);
5132 }
5133 
5134 static int
5135 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
5136 {
5137 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5138 
5139 	ixgbe_remove_rar(dev, 0);
5140 	ixgbe_add_rar(dev, addr, 0, pci_dev->max_vfs);
5141 
5142 	return 0;
5143 }
5144 
5145 static bool
5146 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
5147 {
5148 	if (strcmp(dev->device->driver->name, drv->driver.name))
5149 		return false;
5150 
5151 	return true;
5152 }
5153 
5154 bool
5155 is_ixgbe_supported(struct rte_eth_dev *dev)
5156 {
5157 	return is_device_supported(dev, &rte_ixgbe_pmd);
5158 }
5159 
5160 static int
5161 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
5162 {
5163 	uint32_t hlreg0;
5164 	uint32_t maxfrs;
5165 	struct ixgbe_hw *hw;
5166 	struct rte_eth_dev_info dev_info;
5167 	uint32_t frame_size = mtu + IXGBE_ETH_OVERHEAD;
5168 	int ret;
5169 
5170 	ret = ixgbe_dev_info_get(dev, &dev_info);
5171 	if (ret != 0)
5172 		return ret;
5173 
5174 	/* check that mtu is within the allowed range */
5175 	if (mtu < RTE_ETHER_MIN_MTU || frame_size > dev_info.max_rx_pktlen)
5176 		return -EINVAL;
5177 
5178 	/* If device is started, refuse mtu that requires the support of
5179 	 * scattered packets when this feature has not been enabled before.
5180 	 */
5181 	if (dev->data->dev_started && !dev->data->scattered_rx &&
5182 	    frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
5183 			dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
5184 		PMD_INIT_LOG(ERR, "Stop port first.");
5185 		return -EINVAL;
5186 	}
5187 
5188 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5189 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
5190 
5191 	/* switch to jumbo mode if needed */
5192 	if (mtu > RTE_ETHER_MTU)
5193 		hlreg0 |= IXGBE_HLREG0_JUMBOEN;
5194 	else
5195 		hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
5196 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
5197 
5198 	maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
5199 	maxfrs &= 0x0000FFFF;
5200 	maxfrs |= (frame_size << 16);
5201 	IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
5202 
5203 	return 0;
5204 }
5205 
5206 /*
5207  * Virtual Function operations
5208  */
5209 static void
5210 ixgbevf_intr_disable(struct rte_eth_dev *dev)
5211 {
5212 	struct ixgbe_interrupt *intr =
5213 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5214 	struct ixgbe_hw *hw =
5215 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5216 
5217 	PMD_INIT_FUNC_TRACE();
5218 
5219 	/* Clear interrupt mask to stop from interrupts being generated */
5220 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
5221 
5222 	IXGBE_WRITE_FLUSH(hw);
5223 
5224 	/* Clear mask value. */
5225 	intr->mask = 0;
5226 }
5227 
5228 static void
5229 ixgbevf_intr_enable(struct rte_eth_dev *dev)
5230 {
5231 	struct ixgbe_interrupt *intr =
5232 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5233 	struct ixgbe_hw *hw =
5234 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5235 
5236 	PMD_INIT_FUNC_TRACE();
5237 
5238 	/* VF enable interrupt autoclean */
5239 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
5240 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
5241 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
5242 
5243 	IXGBE_WRITE_FLUSH(hw);
5244 
5245 	/* Save IXGBE_VTEIMS value to mask. */
5246 	intr->mask = IXGBE_VF_IRQ_ENABLE_MASK;
5247 }
5248 
5249 static int
5250 ixgbevf_dev_configure(struct rte_eth_dev *dev)
5251 {
5252 	struct rte_eth_conf *conf = &dev->data->dev_conf;
5253 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5254 
5255 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
5256 		     dev->data->port_id);
5257 
5258 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
5259 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
5260 
5261 	/*
5262 	 * VF has no ability to enable/disable HW CRC
5263 	 * Keep the persistent behavior the same as Host PF
5264 	 */
5265 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
5266 	if (conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
5267 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
5268 		conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_KEEP_CRC;
5269 	}
5270 #else
5271 	if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC)) {
5272 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
5273 		conf->rxmode.offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
5274 	}
5275 #endif
5276 
5277 	/*
5278 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
5279 	 * allocation or vector Rx preconditions we will reset it.
5280 	 */
5281 	adapter->rx_bulk_alloc_allowed = true;
5282 	adapter->rx_vec_allowed = true;
5283 
5284 	return 0;
5285 }
5286 
5287 static int
5288 ixgbevf_dev_start(struct rte_eth_dev *dev)
5289 {
5290 	struct ixgbe_hw *hw =
5291 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5292 	uint32_t intr_vector = 0;
5293 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5294 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5295 
5296 	int err, mask = 0;
5297 
5298 	PMD_INIT_FUNC_TRACE();
5299 
5300 	/* Stop the link setup handler before resetting the HW. */
5301 	ixgbe_dev_wait_setup_link_complete(dev, 0);
5302 
5303 	err = hw->mac.ops.reset_hw(hw);
5304 
5305 	/**
5306 	 * In this case, reuses the MAC address assigned by VF
5307 	 * initialization.
5308 	 */
5309 	if (err != IXGBE_SUCCESS && err != IXGBE_ERR_INVALID_MAC_ADDR) {
5310 		PMD_INIT_LOG(ERR, "Unable to reset vf hardware (%d)", err);
5311 		return err;
5312 	}
5313 
5314 	hw->mac.get_link_status = true;
5315 
5316 	/* negotiate mailbox API version to use with the PF. */
5317 	ixgbevf_negotiate_api(hw);
5318 
5319 	ixgbevf_dev_tx_init(dev);
5320 
5321 	/* This can fail when allocating mbufs for descriptor rings */
5322 	err = ixgbevf_dev_rx_init(dev);
5323 	if (err) {
5324 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
5325 		ixgbe_dev_clear_queues(dev);
5326 		return err;
5327 	}
5328 
5329 	/* Set vfta */
5330 	ixgbevf_set_vfta_all(dev, 1);
5331 
5332 	/* Set HW strip */
5333 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
5334 		ETH_VLAN_EXTEND_MASK;
5335 	err = ixgbevf_vlan_offload_config(dev, mask);
5336 	if (err) {
5337 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
5338 		ixgbe_dev_clear_queues(dev);
5339 		return err;
5340 	}
5341 
5342 	ixgbevf_dev_rxtx_start(dev);
5343 
5344 	/* check and configure queue intr-vector mapping */
5345 	if (rte_intr_cap_multiple(intr_handle) &&
5346 	    dev->data->dev_conf.intr_conf.rxq) {
5347 		/* According to datasheet, only vector 0/1/2 can be used,
5348 		 * now only one vector is used for Rx queue
5349 		 */
5350 		intr_vector = 1;
5351 		if (rte_intr_efd_enable(intr_handle, intr_vector)) {
5352 			ixgbe_dev_clear_queues(dev);
5353 			return -1;
5354 		}
5355 	}
5356 
5357 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
5358 		intr_handle->intr_vec =
5359 			rte_zmalloc("intr_vec",
5360 				    dev->data->nb_rx_queues * sizeof(int), 0);
5361 		if (intr_handle->intr_vec == NULL) {
5362 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
5363 				     " intr_vec", dev->data->nb_rx_queues);
5364 			ixgbe_dev_clear_queues(dev);
5365 			return -ENOMEM;
5366 		}
5367 	}
5368 	ixgbevf_configure_msix(dev);
5369 
5370 	/* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
5371 	 * is mapped to VFIO vector 0 in eth_ixgbevf_dev_init( ).
5372 	 * If previous VFIO interrupt mapping setting in eth_ixgbevf_dev_init( )
5373 	 * is not cleared, it will fail when following rte_intr_enable( ) tries
5374 	 * to map Rx queue interrupt to other VFIO vectors.
5375 	 * So clear uio/vfio intr/evevnfd first to avoid failure.
5376 	 */
5377 	rte_intr_disable(intr_handle);
5378 
5379 	rte_intr_enable(intr_handle);
5380 
5381 	/* Re-enable interrupt for VF */
5382 	ixgbevf_intr_enable(dev);
5383 
5384 	/*
5385 	 * Update link status right before return, because it may
5386 	 * start link configuration process in a separate thread.
5387 	 */
5388 	ixgbevf_dev_link_update(dev, 0);
5389 
5390 	hw->adapter_stopped = false;
5391 
5392 	return 0;
5393 }
5394 
5395 static int
5396 ixgbevf_dev_stop(struct rte_eth_dev *dev)
5397 {
5398 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5399 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5400 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5401 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5402 
5403 	if (hw->adapter_stopped)
5404 		return 0;
5405 
5406 	PMD_INIT_FUNC_TRACE();
5407 
5408 	ixgbe_dev_wait_setup_link_complete(dev, 0);
5409 
5410 	ixgbevf_intr_disable(dev);
5411 
5412 	dev->data->dev_started = 0;
5413 	hw->adapter_stopped = 1;
5414 	ixgbe_stop_adapter(hw);
5415 
5416 	/*
5417 	  * Clear what we set, but we still keep shadow_vfta to
5418 	  * restore after device starts
5419 	  */
5420 	ixgbevf_set_vfta_all(dev, 0);
5421 
5422 	/* Clear stored conf */
5423 	dev->data->scattered_rx = 0;
5424 
5425 	ixgbe_dev_clear_queues(dev);
5426 
5427 	/* Clean datapath event and queue/vec mapping */
5428 	rte_intr_efd_disable(intr_handle);
5429 	if (intr_handle->intr_vec != NULL) {
5430 		rte_free(intr_handle->intr_vec);
5431 		intr_handle->intr_vec = NULL;
5432 	}
5433 
5434 	adapter->rss_reta_updated = 0;
5435 
5436 	return 0;
5437 }
5438 
5439 static int
5440 ixgbevf_dev_close(struct rte_eth_dev *dev)
5441 {
5442 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5443 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5444 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5445 	int ret;
5446 
5447 	PMD_INIT_FUNC_TRACE();
5448 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5449 		return 0;
5450 
5451 	ixgbe_reset_hw(hw);
5452 
5453 	ret = ixgbevf_dev_stop(dev);
5454 
5455 	ixgbe_dev_free_queues(dev);
5456 
5457 	/**
5458 	 * Remove the VF MAC address ro ensure
5459 	 * that the VF traffic goes to the PF
5460 	 * after stop, close and detach of the VF
5461 	 **/
5462 	ixgbevf_remove_mac_addr(dev, 0);
5463 
5464 	rte_intr_disable(intr_handle);
5465 	rte_intr_callback_unregister(intr_handle,
5466 				     ixgbevf_dev_interrupt_handler, dev);
5467 
5468 	return ret;
5469 }
5470 
5471 /*
5472  * Reset VF device
5473  */
5474 static int
5475 ixgbevf_dev_reset(struct rte_eth_dev *dev)
5476 {
5477 	int ret;
5478 
5479 	ret = eth_ixgbevf_dev_uninit(dev);
5480 	if (ret)
5481 		return ret;
5482 
5483 	ret = eth_ixgbevf_dev_init(dev);
5484 
5485 	return ret;
5486 }
5487 
5488 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
5489 {
5490 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5491 	struct ixgbe_vfta *shadow_vfta =
5492 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5493 	int i = 0, j = 0, vfta = 0, mask = 1;
5494 
5495 	for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
5496 		vfta = shadow_vfta->vfta[i];
5497 		if (vfta) {
5498 			mask = 1;
5499 			for (j = 0; j < 32; j++) {
5500 				if (vfta & mask)
5501 					ixgbe_set_vfta(hw, (i<<5)+j, 0,
5502 						       on, false);
5503 				mask <<= 1;
5504 			}
5505 		}
5506 	}
5507 
5508 }
5509 
5510 static int
5511 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
5512 {
5513 	struct ixgbe_hw *hw =
5514 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5515 	struct ixgbe_vfta *shadow_vfta =
5516 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5517 	uint32_t vid_idx = 0;
5518 	uint32_t vid_bit = 0;
5519 	int ret = 0;
5520 
5521 	PMD_INIT_FUNC_TRACE();
5522 
5523 	/* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
5524 	ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
5525 	if (ret) {
5526 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
5527 		return ret;
5528 	}
5529 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
5530 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
5531 
5532 	/* Save what we set and retore it after device reset */
5533 	if (on)
5534 		shadow_vfta->vfta[vid_idx] |= vid_bit;
5535 	else
5536 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
5537 
5538 	return 0;
5539 }
5540 
5541 static void
5542 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
5543 {
5544 	struct ixgbe_hw *hw =
5545 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5546 	uint32_t ctrl;
5547 
5548 	PMD_INIT_FUNC_TRACE();
5549 
5550 	if (queue >= hw->mac.max_rx_queues)
5551 		return;
5552 
5553 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
5554 	if (on)
5555 		ctrl |= IXGBE_RXDCTL_VME;
5556 	else
5557 		ctrl &= ~IXGBE_RXDCTL_VME;
5558 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
5559 
5560 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
5561 }
5562 
5563 static int
5564 ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask)
5565 {
5566 	struct ixgbe_rx_queue *rxq;
5567 	uint16_t i;
5568 	int on = 0;
5569 
5570 	/* VF function only support hw strip feature, others are not support */
5571 	if (mask & ETH_VLAN_STRIP_MASK) {
5572 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
5573 			rxq = dev->data->rx_queues[i];
5574 			on = !!(rxq->offloads &	DEV_RX_OFFLOAD_VLAN_STRIP);
5575 			ixgbevf_vlan_strip_queue_set(dev, i, on);
5576 		}
5577 	}
5578 
5579 	return 0;
5580 }
5581 
5582 static int
5583 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
5584 {
5585 	ixgbe_config_vlan_strip_on_all_queues(dev, mask);
5586 
5587 	ixgbevf_vlan_offload_config(dev, mask);
5588 
5589 	return 0;
5590 }
5591 
5592 int
5593 ixgbe_vt_check(struct ixgbe_hw *hw)
5594 {
5595 	uint32_t reg_val;
5596 
5597 	/* if Virtualization Technology is enabled */
5598 	reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
5599 	if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
5600 		PMD_INIT_LOG(ERR, "VT must be enabled for this setting");
5601 		return -1;
5602 	}
5603 
5604 	return 0;
5605 }
5606 
5607 static uint32_t
5608 ixgbe_uta_vector(struct ixgbe_hw *hw, struct rte_ether_addr *uc_addr)
5609 {
5610 	uint32_t vector = 0;
5611 
5612 	switch (hw->mac.mc_filter_type) {
5613 	case 0:   /* use bits [47:36] of the address */
5614 		vector = ((uc_addr->addr_bytes[4] >> 4) |
5615 			(((uint16_t)uc_addr->addr_bytes[5]) << 4));
5616 		break;
5617 	case 1:   /* use bits [46:35] of the address */
5618 		vector = ((uc_addr->addr_bytes[4] >> 3) |
5619 			(((uint16_t)uc_addr->addr_bytes[5]) << 5));
5620 		break;
5621 	case 2:   /* use bits [45:34] of the address */
5622 		vector = ((uc_addr->addr_bytes[4] >> 2) |
5623 			(((uint16_t)uc_addr->addr_bytes[5]) << 6));
5624 		break;
5625 	case 3:   /* use bits [43:32] of the address */
5626 		vector = ((uc_addr->addr_bytes[4]) |
5627 			(((uint16_t)uc_addr->addr_bytes[5]) << 8));
5628 		break;
5629 	default:  /* Invalid mc_filter_type */
5630 		break;
5631 	}
5632 
5633 	/* vector can only be 12-bits or boundary will be exceeded */
5634 	vector &= 0xFFF;
5635 	return vector;
5636 }
5637 
5638 static int
5639 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev,
5640 			struct rte_ether_addr *mac_addr, uint8_t on)
5641 {
5642 	uint32_t vector;
5643 	uint32_t uta_idx;
5644 	uint32_t reg_val;
5645 	uint32_t uta_shift;
5646 	uint32_t rc;
5647 	const uint32_t ixgbe_uta_idx_mask = 0x7F;
5648 	const uint32_t ixgbe_uta_bit_shift = 5;
5649 	const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
5650 	const uint32_t bit1 = 0x1;
5651 
5652 	struct ixgbe_hw *hw =
5653 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5654 	struct ixgbe_uta_info *uta_info =
5655 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5656 
5657 	/* The UTA table only exists on 82599 hardware and newer */
5658 	if (hw->mac.type < ixgbe_mac_82599EB)
5659 		return -ENOTSUP;
5660 
5661 	vector = ixgbe_uta_vector(hw, mac_addr);
5662 	uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
5663 	uta_shift = vector & ixgbe_uta_bit_mask;
5664 
5665 	rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
5666 	if (rc == on)
5667 		return 0;
5668 
5669 	reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
5670 	if (on) {
5671 		uta_info->uta_in_use++;
5672 		reg_val |= (bit1 << uta_shift);
5673 		uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
5674 	} else {
5675 		uta_info->uta_in_use--;
5676 		reg_val &= ~(bit1 << uta_shift);
5677 		uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
5678 	}
5679 
5680 	IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
5681 
5682 	if (uta_info->uta_in_use > 0)
5683 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
5684 				IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
5685 	else
5686 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
5687 
5688 	return 0;
5689 }
5690 
5691 static int
5692 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
5693 {
5694 	int i;
5695 	struct ixgbe_hw *hw =
5696 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5697 	struct ixgbe_uta_info *uta_info =
5698 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5699 
5700 	/* The UTA table only exists on 82599 hardware and newer */
5701 	if (hw->mac.type < ixgbe_mac_82599EB)
5702 		return -ENOTSUP;
5703 
5704 	if (on) {
5705 		for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5706 			uta_info->uta_shadow[i] = ~0;
5707 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
5708 		}
5709 	} else {
5710 		for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5711 			uta_info->uta_shadow[i] = 0;
5712 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
5713 		}
5714 	}
5715 	return 0;
5716 
5717 }
5718 
5719 uint32_t
5720 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
5721 {
5722 	uint32_t new_val = orig_val;
5723 
5724 	if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG)
5725 		new_val |= IXGBE_VMOLR_AUPE;
5726 	if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC)
5727 		new_val |= IXGBE_VMOLR_ROMPE;
5728 	if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
5729 		new_val |= IXGBE_VMOLR_ROPE;
5730 	if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
5731 		new_val |= IXGBE_VMOLR_BAM;
5732 	if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
5733 		new_val |= IXGBE_VMOLR_MPE;
5734 
5735 	return new_val;
5736 }
5737 
5738 static int
5739 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5740 {
5741 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5742 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5743 	struct ixgbe_interrupt *intr =
5744 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5745 	struct ixgbe_hw *hw =
5746 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5747 	uint32_t vec = IXGBE_MISC_VEC_ID;
5748 
5749 	if (rte_intr_allow_others(intr_handle))
5750 		vec = IXGBE_RX_VEC_START;
5751 	intr->mask |= (1 << vec);
5752 	RTE_SET_USED(queue_id);
5753 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5754 
5755 	rte_intr_ack(intr_handle);
5756 
5757 	return 0;
5758 }
5759 
5760 static int
5761 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5762 {
5763 	struct ixgbe_interrupt *intr =
5764 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5765 	struct ixgbe_hw *hw =
5766 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5767 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5768 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5769 	uint32_t vec = IXGBE_MISC_VEC_ID;
5770 
5771 	if (rte_intr_allow_others(intr_handle))
5772 		vec = IXGBE_RX_VEC_START;
5773 	intr->mask &= ~(1 << vec);
5774 	RTE_SET_USED(queue_id);
5775 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5776 
5777 	return 0;
5778 }
5779 
5780 static int
5781 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5782 {
5783 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5784 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5785 	uint32_t mask;
5786 	struct ixgbe_hw *hw =
5787 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5788 	struct ixgbe_interrupt *intr =
5789 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5790 
5791 	if (queue_id < 16) {
5792 		ixgbe_disable_intr(hw);
5793 		intr->mask |= (1 << queue_id);
5794 		ixgbe_enable_intr(dev);
5795 	} else if (queue_id < 32) {
5796 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5797 		mask &= (1 << queue_id);
5798 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5799 	} else if (queue_id < 64) {
5800 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5801 		mask &= (1 << (queue_id - 32));
5802 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5803 	}
5804 	rte_intr_ack(intr_handle);
5805 
5806 	return 0;
5807 }
5808 
5809 static int
5810 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5811 {
5812 	uint32_t mask;
5813 	struct ixgbe_hw *hw =
5814 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5815 	struct ixgbe_interrupt *intr =
5816 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5817 
5818 	if (queue_id < 16) {
5819 		ixgbe_disable_intr(hw);
5820 		intr->mask &= ~(1 << queue_id);
5821 		ixgbe_enable_intr(dev);
5822 	} else if (queue_id < 32) {
5823 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5824 		mask &= ~(1 << queue_id);
5825 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5826 	} else if (queue_id < 64) {
5827 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5828 		mask &= ~(1 << (queue_id - 32));
5829 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5830 	}
5831 
5832 	return 0;
5833 }
5834 
5835 static void
5836 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5837 		     uint8_t queue, uint8_t msix_vector)
5838 {
5839 	uint32_t tmp, idx;
5840 
5841 	if (direction == -1) {
5842 		/* other causes */
5843 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5844 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
5845 		tmp &= ~0xFF;
5846 		tmp |= msix_vector;
5847 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
5848 	} else {
5849 		/* rx or tx cause */
5850 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5851 		idx = ((16 * (queue & 1)) + (8 * direction));
5852 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
5853 		tmp &= ~(0xFF << idx);
5854 		tmp |= (msix_vector << idx);
5855 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
5856 	}
5857 }
5858 
5859 /**
5860  * set the IVAR registers, mapping interrupt causes to vectors
5861  * @param hw
5862  *  pointer to ixgbe_hw struct
5863  * @direction
5864  *  0 for Rx, 1 for Tx, -1 for other causes
5865  * @queue
5866  *  queue to map the corresponding interrupt to
5867  * @msix_vector
5868  *  the vector to map to the corresponding queue
5869  */
5870 static void
5871 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5872 		   uint8_t queue, uint8_t msix_vector)
5873 {
5874 	uint32_t tmp, idx;
5875 
5876 	msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5877 	if (hw->mac.type == ixgbe_mac_82598EB) {
5878 		if (direction == -1)
5879 			direction = 0;
5880 		idx = (((direction * 64) + queue) >> 2) & 0x1F;
5881 		tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
5882 		tmp &= ~(0xFF << (8 * (queue & 0x3)));
5883 		tmp |= (msix_vector << (8 * (queue & 0x3)));
5884 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
5885 	} else if ((hw->mac.type == ixgbe_mac_82599EB) ||
5886 			(hw->mac.type == ixgbe_mac_X540) ||
5887 			(hw->mac.type == ixgbe_mac_X550) ||
5888 			(hw->mac.type == ixgbe_mac_X550EM_x)) {
5889 		if (direction == -1) {
5890 			/* other causes */
5891 			idx = ((queue & 1) * 8);
5892 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
5893 			tmp &= ~(0xFF << idx);
5894 			tmp |= (msix_vector << idx);
5895 			IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
5896 		} else {
5897 			/* rx or tx causes */
5898 			idx = ((16 * (queue & 1)) + (8 * direction));
5899 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
5900 			tmp &= ~(0xFF << idx);
5901 			tmp |= (msix_vector << idx);
5902 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
5903 		}
5904 	}
5905 }
5906 
5907 static void
5908 ixgbevf_configure_msix(struct rte_eth_dev *dev)
5909 {
5910 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5911 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5912 	struct ixgbe_hw *hw =
5913 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5914 	uint32_t q_idx;
5915 	uint32_t vector_idx = IXGBE_MISC_VEC_ID;
5916 	uint32_t base = IXGBE_MISC_VEC_ID;
5917 
5918 	/* Configure VF other cause ivar */
5919 	ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
5920 
5921 	/* won't configure msix register if no mapping is done
5922 	 * between intr vector and event fd.
5923 	 */
5924 	if (!rte_intr_dp_is_en(intr_handle))
5925 		return;
5926 
5927 	if (rte_intr_allow_others(intr_handle)) {
5928 		base = IXGBE_RX_VEC_START;
5929 		vector_idx = IXGBE_RX_VEC_START;
5930 	}
5931 
5932 	/* Configure all RX queues of VF */
5933 	for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
5934 		/* Force all queue use vector 0,
5935 		 * as IXGBE_VF_MAXMSIVECOTR = 1
5936 		 */
5937 		ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
5938 		intr_handle->intr_vec[q_idx] = vector_idx;
5939 		if (vector_idx < base + intr_handle->nb_efd - 1)
5940 			vector_idx++;
5941 	}
5942 
5943 	/* As RX queue setting above show, all queues use the vector 0.
5944 	 * Set only the ITR value of IXGBE_MISC_VEC_ID.
5945 	 */
5946 	IXGBE_WRITE_REG(hw, IXGBE_VTEITR(IXGBE_MISC_VEC_ID),
5947 			IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
5948 			| IXGBE_EITR_CNT_WDIS);
5949 }
5950 
5951 /**
5952  * Sets up the hardware to properly generate MSI-X interrupts
5953  * @hw
5954  *  board private structure
5955  */
5956 static void
5957 ixgbe_configure_msix(struct rte_eth_dev *dev)
5958 {
5959 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5960 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5961 	struct ixgbe_hw *hw =
5962 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5963 	uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
5964 	uint32_t vec = IXGBE_MISC_VEC_ID;
5965 	uint32_t mask;
5966 	uint32_t gpie;
5967 
5968 	/* won't configure msix register if no mapping is done
5969 	 * between intr vector and event fd
5970 	 * but if misx has been enabled already, need to configure
5971 	 * auto clean, auto mask and throttling.
5972 	 */
5973 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5974 	if (!rte_intr_dp_is_en(intr_handle) &&
5975 	    !(gpie & (IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT)))
5976 		return;
5977 
5978 	if (rte_intr_allow_others(intr_handle))
5979 		vec = base = IXGBE_RX_VEC_START;
5980 
5981 	/* setup GPIE for MSI-x mode */
5982 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5983 	gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5984 		IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
5985 	/* auto clearing and auto setting corresponding bits in EIMS
5986 	 * when MSI-X interrupt is triggered
5987 	 */
5988 	if (hw->mac.type == ixgbe_mac_82598EB) {
5989 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5990 	} else {
5991 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5992 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5993 	}
5994 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5995 
5996 	/* Populate the IVAR table and set the ITR values to the
5997 	 * corresponding register.
5998 	 */
5999 	if (rte_intr_dp_is_en(intr_handle)) {
6000 		for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
6001 			queue_id++) {
6002 			/* by default, 1:1 mapping */
6003 			ixgbe_set_ivar_map(hw, 0, queue_id, vec);
6004 			intr_handle->intr_vec[queue_id] = vec;
6005 			if (vec < base + intr_handle->nb_efd - 1)
6006 				vec++;
6007 		}
6008 
6009 		switch (hw->mac.type) {
6010 		case ixgbe_mac_82598EB:
6011 			ixgbe_set_ivar_map(hw, -1,
6012 					   IXGBE_IVAR_OTHER_CAUSES_INDEX,
6013 					   IXGBE_MISC_VEC_ID);
6014 			break;
6015 		case ixgbe_mac_82599EB:
6016 		case ixgbe_mac_X540:
6017 		case ixgbe_mac_X550:
6018 		case ixgbe_mac_X550EM_x:
6019 			ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
6020 			break;
6021 		default:
6022 			break;
6023 		}
6024 	}
6025 	IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
6026 			IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
6027 			| IXGBE_EITR_CNT_WDIS);
6028 
6029 	/* set up to autoclear timer, and the vectors */
6030 	mask = IXGBE_EIMS_ENABLE_MASK;
6031 	mask &= ~(IXGBE_EIMS_OTHER |
6032 		  IXGBE_EIMS_MAILBOX |
6033 		  IXGBE_EIMS_LSC);
6034 
6035 	IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
6036 }
6037 
6038 int
6039 ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
6040 			   uint16_t queue_idx, uint16_t tx_rate)
6041 {
6042 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6043 	uint32_t rf_dec, rf_int;
6044 	uint32_t bcnrc_val;
6045 	uint16_t link_speed = dev->data->dev_link.link_speed;
6046 
6047 	if (queue_idx >= hw->mac.max_tx_queues)
6048 		return -EINVAL;
6049 
6050 	if (tx_rate != 0) {
6051 		/* Calculate the rate factor values to set */
6052 		rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
6053 		rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
6054 		rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
6055 
6056 		bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
6057 		bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
6058 				IXGBE_RTTBCNRC_RF_INT_MASK_M);
6059 		bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
6060 	} else {
6061 		bcnrc_val = 0;
6062 	}
6063 
6064 	/*
6065 	 * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
6066 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
6067 	 * set as 0x4.
6068 	 */
6069 	if (dev->data->mtu + IXGBE_ETH_OVERHEAD >= IXGBE_MAX_JUMBO_FRAME_SIZE)
6070 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM, IXGBE_MMW_SIZE_JUMBO_FRAME);
6071 	else
6072 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM, IXGBE_MMW_SIZE_DEFAULT);
6073 
6074 	/* Set RTTBCNRC of queue X */
6075 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
6076 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
6077 	IXGBE_WRITE_FLUSH(hw);
6078 
6079 	return 0;
6080 }
6081 
6082 static int
6083 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
6084 		     __rte_unused uint32_t index,
6085 		     __rte_unused uint32_t pool)
6086 {
6087 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6088 	int diag;
6089 
6090 	/*
6091 	 * On a 82599 VF, adding again the same MAC addr is not an idempotent
6092 	 * operation. Trap this case to avoid exhausting the [very limited]
6093 	 * set of PF resources used to store VF MAC addresses.
6094 	 */
6095 	if (memcmp(hw->mac.perm_addr, mac_addr,
6096 			sizeof(struct rte_ether_addr)) == 0)
6097 		return -1;
6098 	diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6099 	if (diag != 0)
6100 		PMD_DRV_LOG(ERR, "Unable to add MAC address "
6101 			    RTE_ETHER_ADDR_PRT_FMT " - diag=%d",
6102 			    RTE_ETHER_ADDR_BYTES(mac_addr), diag);
6103 	return diag;
6104 }
6105 
6106 static void
6107 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
6108 {
6109 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6110 	struct rte_ether_addr *perm_addr =
6111 		(struct rte_ether_addr *)hw->mac.perm_addr;
6112 	struct rte_ether_addr *mac_addr;
6113 	uint32_t i;
6114 	int diag;
6115 
6116 	/*
6117 	 * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
6118 	 * not support the deletion of a given MAC address.
6119 	 * Instead, it imposes to delete all MAC addresses, then to add again
6120 	 * all MAC addresses with the exception of the one to be deleted.
6121 	 */
6122 	(void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
6123 
6124 	/*
6125 	 * Add again all MAC addresses, with the exception of the deleted one
6126 	 * and of the permanent MAC address.
6127 	 */
6128 	for (i = 0, mac_addr = dev->data->mac_addrs;
6129 	     i < hw->mac.num_rar_entries; i++, mac_addr++) {
6130 		/* Skip the deleted MAC address */
6131 		if (i == index)
6132 			continue;
6133 		/* Skip NULL MAC addresses */
6134 		if (rte_is_zero_ether_addr(mac_addr))
6135 			continue;
6136 		/* Skip the permanent MAC address */
6137 		if (memcmp(perm_addr, mac_addr,
6138 				sizeof(struct rte_ether_addr)) == 0)
6139 			continue;
6140 		diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6141 		if (diag != 0)
6142 			PMD_DRV_LOG(ERR,
6143 				    "Adding again MAC address "
6144 				     RTE_ETHER_ADDR_PRT_FMT " failed "
6145 				    "diag=%d", RTE_ETHER_ADDR_BYTES(mac_addr),
6146 				    diag);
6147 	}
6148 }
6149 
6150 static int
6151 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
6152 			struct rte_ether_addr *addr)
6153 {
6154 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6155 
6156 	hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
6157 
6158 	return 0;
6159 }
6160 
6161 int
6162 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
6163 			struct rte_eth_syn_filter *filter,
6164 			bool add)
6165 {
6166 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6167 	struct ixgbe_filter_info *filter_info =
6168 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6169 	uint32_t syn_info;
6170 	uint32_t synqf;
6171 
6172 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6173 		return -EINVAL;
6174 
6175 	syn_info = filter_info->syn_info;
6176 
6177 	if (add) {
6178 		if (syn_info & IXGBE_SYN_FILTER_ENABLE)
6179 			return -EINVAL;
6180 		synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
6181 			IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
6182 
6183 		if (filter->hig_pri)
6184 			synqf |= IXGBE_SYN_FILTER_SYNQFP;
6185 		else
6186 			synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
6187 	} else {
6188 		synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6189 		if (!(syn_info & IXGBE_SYN_FILTER_ENABLE))
6190 			return -ENOENT;
6191 		synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
6192 	}
6193 
6194 	filter_info->syn_info = synqf;
6195 	IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
6196 	IXGBE_WRITE_FLUSH(hw);
6197 	return 0;
6198 }
6199 
6200 
6201 static inline enum ixgbe_5tuple_protocol
6202 convert_protocol_type(uint8_t protocol_value)
6203 {
6204 	if (protocol_value == IPPROTO_TCP)
6205 		return IXGBE_FILTER_PROTOCOL_TCP;
6206 	else if (protocol_value == IPPROTO_UDP)
6207 		return IXGBE_FILTER_PROTOCOL_UDP;
6208 	else if (protocol_value == IPPROTO_SCTP)
6209 		return IXGBE_FILTER_PROTOCOL_SCTP;
6210 	else
6211 		return IXGBE_FILTER_PROTOCOL_NONE;
6212 }
6213 
6214 /* inject a 5-tuple filter to HW */
6215 static inline void
6216 ixgbe_inject_5tuple_filter(struct rte_eth_dev *dev,
6217 			   struct ixgbe_5tuple_filter *filter)
6218 {
6219 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6220 	int i;
6221 	uint32_t ftqf, sdpqf;
6222 	uint32_t l34timir = 0;
6223 	uint8_t mask = 0xff;
6224 
6225 	i = filter->index;
6226 
6227 	sdpqf = (uint32_t)(filter->filter_info.dst_port <<
6228 				IXGBE_SDPQF_DSTPORT_SHIFT);
6229 	sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
6230 
6231 	ftqf = (uint32_t)(filter->filter_info.proto &
6232 		IXGBE_FTQF_PROTOCOL_MASK);
6233 	ftqf |= (uint32_t)((filter->filter_info.priority &
6234 		IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
6235 	if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
6236 		mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
6237 	if (filter->filter_info.dst_ip_mask == 0)
6238 		mask &= IXGBE_FTQF_DEST_ADDR_MASK;
6239 	if (filter->filter_info.src_port_mask == 0)
6240 		mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
6241 	if (filter->filter_info.dst_port_mask == 0)
6242 		mask &= IXGBE_FTQF_DEST_PORT_MASK;
6243 	if (filter->filter_info.proto_mask == 0)
6244 		mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
6245 	ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
6246 	ftqf |= IXGBE_FTQF_POOL_MASK_EN;
6247 	ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
6248 
6249 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
6250 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
6251 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
6252 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
6253 
6254 	l34timir |= IXGBE_L34T_IMIR_RESERVE;
6255 	l34timir |= (uint32_t)(filter->queue <<
6256 				IXGBE_L34T_IMIR_QUEUE_SHIFT);
6257 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
6258 }
6259 
6260 /*
6261  * add a 5tuple filter
6262  *
6263  * @param
6264  * dev: Pointer to struct rte_eth_dev.
6265  * index: the index the filter allocates.
6266  * filter: ponter to the filter that will be added.
6267  * rx_queue: the queue id the filter assigned to.
6268  *
6269  * @return
6270  *    - On success, zero.
6271  *    - On failure, a negative value.
6272  */
6273 static int
6274 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
6275 			struct ixgbe_5tuple_filter *filter)
6276 {
6277 	struct ixgbe_filter_info *filter_info =
6278 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6279 	int i, idx, shift;
6280 
6281 	/*
6282 	 * look for an unused 5tuple filter index,
6283 	 * and insert the filter to list.
6284 	 */
6285 	for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
6286 		idx = i / (sizeof(uint32_t) * NBBY);
6287 		shift = i % (sizeof(uint32_t) * NBBY);
6288 		if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
6289 			filter_info->fivetuple_mask[idx] |= 1 << shift;
6290 			filter->index = i;
6291 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
6292 					  filter,
6293 					  entries);
6294 			break;
6295 		}
6296 	}
6297 	if (i >= IXGBE_MAX_FTQF_FILTERS) {
6298 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
6299 		return -ENOSYS;
6300 	}
6301 
6302 	ixgbe_inject_5tuple_filter(dev, filter);
6303 
6304 	return 0;
6305 }
6306 
6307 /*
6308  * remove a 5tuple filter
6309  *
6310  * @param
6311  * dev: Pointer to struct rte_eth_dev.
6312  * filter: the pointer of the filter will be removed.
6313  */
6314 static void
6315 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
6316 			struct ixgbe_5tuple_filter *filter)
6317 {
6318 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6319 	struct ixgbe_filter_info *filter_info =
6320 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6321 	uint16_t index = filter->index;
6322 
6323 	filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
6324 				~(1 << (index % (sizeof(uint32_t) * NBBY)));
6325 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
6326 	rte_free(filter);
6327 
6328 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
6329 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
6330 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
6331 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
6332 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
6333 }
6334 
6335 static int
6336 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
6337 {
6338 	struct ixgbe_hw *hw;
6339 	uint32_t max_frame = mtu + IXGBE_ETH_OVERHEAD;
6340 	struct rte_eth_dev_data *dev_data = dev->data;
6341 
6342 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6343 
6344 	if (mtu < RTE_ETHER_MIN_MTU || max_frame > RTE_ETHER_MAX_JUMBO_FRAME_LEN)
6345 		return -EINVAL;
6346 
6347 	/* If device is started, refuse mtu that requires the support of
6348 	 * scattered packets when this feature has not been enabled before.
6349 	 */
6350 	if (dev_data->dev_started && !dev_data->scattered_rx &&
6351 	    (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
6352 			dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
6353 		PMD_INIT_LOG(ERR, "Stop port first.");
6354 		return -EINVAL;
6355 	}
6356 
6357 	/*
6358 	 * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
6359 	 * request of the version 2.0 of the mailbox API.
6360 	 * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
6361 	 * of the mailbox API.
6362 	 * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
6363 	 * prior to 3.11.33 which contains the following change:
6364 	 * "ixgbe: Enable jumbo frames support w/ SR-IOV"
6365 	 */
6366 	if (ixgbevf_rlpml_set_vf(hw, max_frame))
6367 		return -EINVAL;
6368 
6369 	return 0;
6370 }
6371 
6372 static inline struct ixgbe_5tuple_filter *
6373 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
6374 			struct ixgbe_5tuple_filter_info *key)
6375 {
6376 	struct ixgbe_5tuple_filter *it;
6377 
6378 	TAILQ_FOREACH(it, filter_list, entries) {
6379 		if (memcmp(key, &it->filter_info,
6380 			sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
6381 			return it;
6382 		}
6383 	}
6384 	return NULL;
6385 }
6386 
6387 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
6388 static inline int
6389 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
6390 			struct ixgbe_5tuple_filter_info *filter_info)
6391 {
6392 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
6393 		filter->priority > IXGBE_5TUPLE_MAX_PRI ||
6394 		filter->priority < IXGBE_5TUPLE_MIN_PRI)
6395 		return -EINVAL;
6396 
6397 	switch (filter->dst_ip_mask) {
6398 	case UINT32_MAX:
6399 		filter_info->dst_ip_mask = 0;
6400 		filter_info->dst_ip = filter->dst_ip;
6401 		break;
6402 	case 0:
6403 		filter_info->dst_ip_mask = 1;
6404 		break;
6405 	default:
6406 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
6407 		return -EINVAL;
6408 	}
6409 
6410 	switch (filter->src_ip_mask) {
6411 	case UINT32_MAX:
6412 		filter_info->src_ip_mask = 0;
6413 		filter_info->src_ip = filter->src_ip;
6414 		break;
6415 	case 0:
6416 		filter_info->src_ip_mask = 1;
6417 		break;
6418 	default:
6419 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
6420 		return -EINVAL;
6421 	}
6422 
6423 	switch (filter->dst_port_mask) {
6424 	case UINT16_MAX:
6425 		filter_info->dst_port_mask = 0;
6426 		filter_info->dst_port = filter->dst_port;
6427 		break;
6428 	case 0:
6429 		filter_info->dst_port_mask = 1;
6430 		break;
6431 	default:
6432 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
6433 		return -EINVAL;
6434 	}
6435 
6436 	switch (filter->src_port_mask) {
6437 	case UINT16_MAX:
6438 		filter_info->src_port_mask = 0;
6439 		filter_info->src_port = filter->src_port;
6440 		break;
6441 	case 0:
6442 		filter_info->src_port_mask = 1;
6443 		break;
6444 	default:
6445 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
6446 		return -EINVAL;
6447 	}
6448 
6449 	switch (filter->proto_mask) {
6450 	case UINT8_MAX:
6451 		filter_info->proto_mask = 0;
6452 		filter_info->proto =
6453 			convert_protocol_type(filter->proto);
6454 		break;
6455 	case 0:
6456 		filter_info->proto_mask = 1;
6457 		break;
6458 	default:
6459 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
6460 		return -EINVAL;
6461 	}
6462 
6463 	filter_info->priority = (uint8_t)filter->priority;
6464 	return 0;
6465 }
6466 
6467 /*
6468  * add or delete a ntuple filter
6469  *
6470  * @param
6471  * dev: Pointer to struct rte_eth_dev.
6472  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6473  * add: if true, add filter, if false, remove filter
6474  *
6475  * @return
6476  *    - On success, zero.
6477  *    - On failure, a negative value.
6478  */
6479 int
6480 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
6481 			struct rte_eth_ntuple_filter *ntuple_filter,
6482 			bool add)
6483 {
6484 	struct ixgbe_filter_info *filter_info =
6485 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6486 	struct ixgbe_5tuple_filter_info filter_5tuple;
6487 	struct ixgbe_5tuple_filter *filter;
6488 	int ret;
6489 
6490 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6491 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6492 		return -EINVAL;
6493 	}
6494 
6495 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6496 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6497 	if (ret < 0)
6498 		return ret;
6499 
6500 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6501 					 &filter_5tuple);
6502 	if (filter != NULL && add) {
6503 		PMD_DRV_LOG(ERR, "filter exists.");
6504 		return -EEXIST;
6505 	}
6506 	if (filter == NULL && !add) {
6507 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
6508 		return -ENOENT;
6509 	}
6510 
6511 	if (add) {
6512 		filter = rte_zmalloc("ixgbe_5tuple_filter",
6513 				sizeof(struct ixgbe_5tuple_filter), 0);
6514 		if (filter == NULL)
6515 			return -ENOMEM;
6516 		rte_memcpy(&filter->filter_info,
6517 				 &filter_5tuple,
6518 				 sizeof(struct ixgbe_5tuple_filter_info));
6519 		filter->queue = ntuple_filter->queue;
6520 		ret = ixgbe_add_5tuple_filter(dev, filter);
6521 		if (ret < 0) {
6522 			rte_free(filter);
6523 			return ret;
6524 		}
6525 	} else
6526 		ixgbe_remove_5tuple_filter(dev, filter);
6527 
6528 	return 0;
6529 }
6530 
6531 int
6532 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
6533 			struct rte_eth_ethertype_filter *filter,
6534 			bool add)
6535 {
6536 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6537 	struct ixgbe_filter_info *filter_info =
6538 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6539 	uint32_t etqf = 0;
6540 	uint32_t etqs = 0;
6541 	int ret;
6542 	struct ixgbe_ethertype_filter ethertype_filter;
6543 
6544 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6545 		return -EINVAL;
6546 
6547 	if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
6548 		filter->ether_type == RTE_ETHER_TYPE_IPV6) {
6549 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
6550 			" ethertype filter.", filter->ether_type);
6551 		return -EINVAL;
6552 	}
6553 
6554 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
6555 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
6556 		return -EINVAL;
6557 	}
6558 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
6559 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
6560 		return -EINVAL;
6561 	}
6562 
6563 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6564 	if (ret >= 0 && add) {
6565 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
6566 			    filter->ether_type);
6567 		return -EEXIST;
6568 	}
6569 	if (ret < 0 && !add) {
6570 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6571 			    filter->ether_type);
6572 		return -ENOENT;
6573 	}
6574 
6575 	if (add) {
6576 		etqf = IXGBE_ETQF_FILTER_EN;
6577 		etqf |= (uint32_t)filter->ether_type;
6578 		etqs |= (uint32_t)((filter->queue <<
6579 				    IXGBE_ETQS_RX_QUEUE_SHIFT) &
6580 				    IXGBE_ETQS_RX_QUEUE);
6581 		etqs |= IXGBE_ETQS_QUEUE_EN;
6582 
6583 		ethertype_filter.ethertype = filter->ether_type;
6584 		ethertype_filter.etqf = etqf;
6585 		ethertype_filter.etqs = etqs;
6586 		ethertype_filter.conf = FALSE;
6587 		ret = ixgbe_ethertype_filter_insert(filter_info,
6588 						    &ethertype_filter);
6589 		if (ret < 0) {
6590 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
6591 			return -ENOSPC;
6592 		}
6593 	} else {
6594 		ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
6595 		if (ret < 0)
6596 			return -ENOSYS;
6597 	}
6598 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
6599 	IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
6600 	IXGBE_WRITE_FLUSH(hw);
6601 
6602 	return 0;
6603 }
6604 
6605 static int
6606 ixgbe_dev_flow_ops_get(__rte_unused struct rte_eth_dev *dev,
6607 		       const struct rte_flow_ops **ops)
6608 {
6609 	*ops = &ixgbe_flow_ops;
6610 	return 0;
6611 }
6612 
6613 static u8 *
6614 ixgbe_dev_addr_list_itr(__rte_unused struct ixgbe_hw *hw,
6615 			u8 **mc_addr_ptr, u32 *vmdq)
6616 {
6617 	u8 *mc_addr;
6618 
6619 	*vmdq = 0;
6620 	mc_addr = *mc_addr_ptr;
6621 	*mc_addr_ptr = (mc_addr + sizeof(struct rte_ether_addr));
6622 	return mc_addr;
6623 }
6624 
6625 static int
6626 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
6627 			  struct rte_ether_addr *mc_addr_set,
6628 			  uint32_t nb_mc_addr)
6629 {
6630 	struct ixgbe_hw *hw;
6631 	u8 *mc_addr_list;
6632 
6633 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6634 	mc_addr_list = (u8 *)mc_addr_set;
6635 	return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
6636 					 ixgbe_dev_addr_list_itr, TRUE);
6637 }
6638 
6639 static uint64_t
6640 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
6641 {
6642 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6643 	uint64_t systime_cycles;
6644 
6645 	switch (hw->mac.type) {
6646 	case ixgbe_mac_X550:
6647 	case ixgbe_mac_X550EM_x:
6648 	case ixgbe_mac_X550EM_a:
6649 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
6650 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6651 		systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6652 				* NSEC_PER_SEC;
6653 		break;
6654 	default:
6655 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6656 		systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6657 				<< 32;
6658 	}
6659 
6660 	return systime_cycles;
6661 }
6662 
6663 static uint64_t
6664 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6665 {
6666 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6667 	uint64_t rx_tstamp_cycles;
6668 
6669 	switch (hw->mac.type) {
6670 	case ixgbe_mac_X550:
6671 	case ixgbe_mac_X550EM_x:
6672 	case ixgbe_mac_X550EM_a:
6673 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6674 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6675 		rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6676 				* NSEC_PER_SEC;
6677 		break;
6678 	default:
6679 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6680 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6681 		rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6682 				<< 32;
6683 	}
6684 
6685 	return rx_tstamp_cycles;
6686 }
6687 
6688 static uint64_t
6689 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6690 {
6691 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6692 	uint64_t tx_tstamp_cycles;
6693 
6694 	switch (hw->mac.type) {
6695 	case ixgbe_mac_X550:
6696 	case ixgbe_mac_X550EM_x:
6697 	case ixgbe_mac_X550EM_a:
6698 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6699 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6700 		tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6701 				* NSEC_PER_SEC;
6702 		break;
6703 	default:
6704 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6705 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6706 		tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6707 				<< 32;
6708 	}
6709 
6710 	return tx_tstamp_cycles;
6711 }
6712 
6713 static void
6714 ixgbe_start_timecounters(struct rte_eth_dev *dev)
6715 {
6716 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6717 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6718 	struct rte_eth_link link;
6719 	uint32_t incval = 0;
6720 	uint32_t shift = 0;
6721 
6722 	/* Get current link speed. */
6723 	ixgbe_dev_link_update(dev, 1);
6724 	rte_eth_linkstatus_get(dev, &link);
6725 
6726 	switch (link.link_speed) {
6727 	case ETH_SPEED_NUM_100M:
6728 		incval = IXGBE_INCVAL_100;
6729 		shift = IXGBE_INCVAL_SHIFT_100;
6730 		break;
6731 	case ETH_SPEED_NUM_1G:
6732 		incval = IXGBE_INCVAL_1GB;
6733 		shift = IXGBE_INCVAL_SHIFT_1GB;
6734 		break;
6735 	case ETH_SPEED_NUM_10G:
6736 	default:
6737 		incval = IXGBE_INCVAL_10GB;
6738 		shift = IXGBE_INCVAL_SHIFT_10GB;
6739 		break;
6740 	}
6741 
6742 	switch (hw->mac.type) {
6743 	case ixgbe_mac_X550:
6744 	case ixgbe_mac_X550EM_x:
6745 	case ixgbe_mac_X550EM_a:
6746 		/* Independent of link speed. */
6747 		incval = 1;
6748 		/* Cycles read will be interpreted as ns. */
6749 		shift = 0;
6750 		/* Fall-through */
6751 	case ixgbe_mac_X540:
6752 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
6753 		break;
6754 	case ixgbe_mac_82599EB:
6755 		incval >>= IXGBE_INCVAL_SHIFT_82599;
6756 		shift -= IXGBE_INCVAL_SHIFT_82599;
6757 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
6758 				(1 << IXGBE_INCPER_SHIFT_82599) | incval);
6759 		break;
6760 	default:
6761 		/* Not supported. */
6762 		return;
6763 	}
6764 
6765 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
6766 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6767 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6768 
6769 	adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6770 	adapter->systime_tc.cc_shift = shift;
6771 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
6772 
6773 	adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6774 	adapter->rx_tstamp_tc.cc_shift = shift;
6775 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6776 
6777 	adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6778 	adapter->tx_tstamp_tc.cc_shift = shift;
6779 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6780 }
6781 
6782 static int
6783 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
6784 {
6785 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6786 
6787 	adapter->systime_tc.nsec += delta;
6788 	adapter->rx_tstamp_tc.nsec += delta;
6789 	adapter->tx_tstamp_tc.nsec += delta;
6790 
6791 	return 0;
6792 }
6793 
6794 static int
6795 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
6796 {
6797 	uint64_t ns;
6798 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6799 
6800 	ns = rte_timespec_to_ns(ts);
6801 	/* Set the timecounters to a new value. */
6802 	adapter->systime_tc.nsec = ns;
6803 	adapter->rx_tstamp_tc.nsec = ns;
6804 	adapter->tx_tstamp_tc.nsec = ns;
6805 
6806 	return 0;
6807 }
6808 
6809 static int
6810 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
6811 {
6812 	uint64_t ns, systime_cycles;
6813 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6814 
6815 	systime_cycles = ixgbe_read_systime_cyclecounter(dev);
6816 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
6817 	*ts = rte_ns_to_timespec(ns);
6818 
6819 	return 0;
6820 }
6821 
6822 static int
6823 ixgbe_timesync_enable(struct rte_eth_dev *dev)
6824 {
6825 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6826 	uint32_t tsync_ctl;
6827 	uint32_t tsauxc;
6828 
6829 	/* Stop the timesync system time. */
6830 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
6831 	/* Reset the timesync system time value. */
6832 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
6833 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
6834 
6835 	/* Enable system time for platforms where it isn't on by default. */
6836 	tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
6837 	tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
6838 	IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
6839 
6840 	ixgbe_start_timecounters(dev);
6841 
6842 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6843 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
6844 			(RTE_ETHER_TYPE_1588 |
6845 			 IXGBE_ETQF_FILTER_EN |
6846 			 IXGBE_ETQF_1588));
6847 
6848 	/* Enable timestamping of received PTP packets. */
6849 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6850 	tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
6851 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6852 
6853 	/* Enable timestamping of transmitted PTP packets. */
6854 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6855 	tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
6856 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6857 
6858 	IXGBE_WRITE_FLUSH(hw);
6859 
6860 	return 0;
6861 }
6862 
6863 static int
6864 ixgbe_timesync_disable(struct rte_eth_dev *dev)
6865 {
6866 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6867 	uint32_t tsync_ctl;
6868 
6869 	/* Disable timestamping of transmitted PTP packets. */
6870 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6871 	tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
6872 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6873 
6874 	/* Disable timestamping of received PTP packets. */
6875 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6876 	tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
6877 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6878 
6879 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6880 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
6881 
6882 	/* Stop incrementating the System Time registers. */
6883 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
6884 
6885 	return 0;
6886 }
6887 
6888 static int
6889 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
6890 				 struct timespec *timestamp,
6891 				 uint32_t flags __rte_unused)
6892 {
6893 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6894 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6895 	uint32_t tsync_rxctl;
6896 	uint64_t rx_tstamp_cycles;
6897 	uint64_t ns;
6898 
6899 	tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6900 	if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
6901 		return -EINVAL;
6902 
6903 	rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
6904 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
6905 	*timestamp = rte_ns_to_timespec(ns);
6906 
6907 	return  0;
6908 }
6909 
6910 static int
6911 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
6912 				 struct timespec *timestamp)
6913 {
6914 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6915 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6916 	uint32_t tsync_txctl;
6917 	uint64_t tx_tstamp_cycles;
6918 	uint64_t ns;
6919 
6920 	tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6921 	if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
6922 		return -EINVAL;
6923 
6924 	tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
6925 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
6926 	*timestamp = rte_ns_to_timespec(ns);
6927 
6928 	return 0;
6929 }
6930 
6931 static int
6932 ixgbe_get_reg_length(struct rte_eth_dev *dev)
6933 {
6934 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6935 	int count = 0;
6936 	int g_ind = 0;
6937 	const struct reg_info *reg_group;
6938 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6939 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6940 
6941 	while ((reg_group = reg_set[g_ind++]))
6942 		count += ixgbe_regs_group_count(reg_group);
6943 
6944 	return count;
6945 }
6946 
6947 static int
6948 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
6949 {
6950 	int count = 0;
6951 	int g_ind = 0;
6952 	const struct reg_info *reg_group;
6953 
6954 	while ((reg_group = ixgbevf_regs[g_ind++]))
6955 		count += ixgbe_regs_group_count(reg_group);
6956 
6957 	return count;
6958 }
6959 
6960 static int
6961 ixgbe_get_regs(struct rte_eth_dev *dev,
6962 	      struct rte_dev_reg_info *regs)
6963 {
6964 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6965 	uint32_t *data = regs->data;
6966 	int g_ind = 0;
6967 	int count = 0;
6968 	const struct reg_info *reg_group;
6969 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6970 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6971 
6972 	if (data == NULL) {
6973 		regs->length = ixgbe_get_reg_length(dev);
6974 		regs->width = sizeof(uint32_t);
6975 		return 0;
6976 	}
6977 
6978 	/* Support only full register dump */
6979 	if ((regs->length == 0) ||
6980 	    (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
6981 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
6982 			hw->device_id;
6983 		while ((reg_group = reg_set[g_ind++]))
6984 			count += ixgbe_read_regs_group(dev, &data[count],
6985 				reg_group);
6986 		return 0;
6987 	}
6988 
6989 	return -ENOTSUP;
6990 }
6991 
6992 static int
6993 ixgbevf_get_regs(struct rte_eth_dev *dev,
6994 		struct rte_dev_reg_info *regs)
6995 {
6996 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6997 	uint32_t *data = regs->data;
6998 	int g_ind = 0;
6999 	int count = 0;
7000 	const struct reg_info *reg_group;
7001 
7002 	if (data == NULL) {
7003 		regs->length = ixgbevf_get_reg_length(dev);
7004 		regs->width = sizeof(uint32_t);
7005 		return 0;
7006 	}
7007 
7008 	/* Support only full register dump */
7009 	if ((regs->length == 0) ||
7010 	    (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
7011 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7012 			hw->device_id;
7013 		while ((reg_group = ixgbevf_regs[g_ind++]))
7014 			count += ixgbe_read_regs_group(dev, &data[count],
7015 						      reg_group);
7016 		return 0;
7017 	}
7018 
7019 	return -ENOTSUP;
7020 }
7021 
7022 static int
7023 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
7024 {
7025 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7026 
7027 	/* Return unit is byte count */
7028 	return hw->eeprom.word_size * 2;
7029 }
7030 
7031 static int
7032 ixgbe_get_eeprom(struct rte_eth_dev *dev,
7033 		struct rte_dev_eeprom_info *in_eeprom)
7034 {
7035 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7036 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7037 	uint16_t *data = in_eeprom->data;
7038 	int first, length;
7039 
7040 	first = in_eeprom->offset >> 1;
7041 	length = in_eeprom->length >> 1;
7042 	if ((first > hw->eeprom.word_size) ||
7043 	    ((first + length) > hw->eeprom.word_size))
7044 		return -EINVAL;
7045 
7046 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7047 
7048 	return eeprom->ops.read_buffer(hw, first, length, data);
7049 }
7050 
7051 static int
7052 ixgbe_set_eeprom(struct rte_eth_dev *dev,
7053 		struct rte_dev_eeprom_info *in_eeprom)
7054 {
7055 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7056 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7057 	uint16_t *data = in_eeprom->data;
7058 	int first, length;
7059 
7060 	first = in_eeprom->offset >> 1;
7061 	length = in_eeprom->length >> 1;
7062 	if ((first > hw->eeprom.word_size) ||
7063 	    ((first + length) > hw->eeprom.word_size))
7064 		return -EINVAL;
7065 
7066 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7067 
7068 	return eeprom->ops.write_buffer(hw,  first, length, data);
7069 }
7070 
7071 static int
7072 ixgbe_get_module_info(struct rte_eth_dev *dev,
7073 		      struct rte_eth_dev_module_info *modinfo)
7074 {
7075 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7076 	uint32_t status;
7077 	uint8_t sff8472_rev, addr_mode;
7078 	bool page_swap = false;
7079 
7080 	/* Check whether we support SFF-8472 or not */
7081 	status = hw->phy.ops.read_i2c_eeprom(hw,
7082 					     IXGBE_SFF_SFF_8472_COMP,
7083 					     &sff8472_rev);
7084 	if (status != 0)
7085 		return -EIO;
7086 
7087 	/* addressing mode is not supported */
7088 	status = hw->phy.ops.read_i2c_eeprom(hw,
7089 					     IXGBE_SFF_SFF_8472_SWAP,
7090 					     &addr_mode);
7091 	if (status != 0)
7092 		return -EIO;
7093 
7094 	if (addr_mode & IXGBE_SFF_ADDRESSING_MODE) {
7095 		PMD_DRV_LOG(ERR,
7096 			    "Address change required to access page 0xA2, "
7097 			    "but not supported. Please report the module "
7098 			    "type to the driver maintainers.");
7099 		page_swap = true;
7100 	}
7101 
7102 	if (sff8472_rev == IXGBE_SFF_SFF_8472_UNSUP || page_swap) {
7103 		/* We have a SFP, but it does not support SFF-8472 */
7104 		modinfo->type = RTE_ETH_MODULE_SFF_8079;
7105 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
7106 	} else {
7107 		/* We have a SFP which supports a revision of SFF-8472. */
7108 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
7109 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
7110 	}
7111 
7112 	return 0;
7113 }
7114 
7115 static int
7116 ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
7117 			struct rte_dev_eeprom_info *info)
7118 {
7119 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7120 	uint32_t status = IXGBE_ERR_PHY_ADDR_INVALID;
7121 	uint8_t databyte = 0xFF;
7122 	uint8_t *data = info->data;
7123 	uint32_t i = 0;
7124 
7125 	for (i = info->offset; i < info->offset + info->length; i++) {
7126 		if (i < RTE_ETH_MODULE_SFF_8079_LEN)
7127 			status = hw->phy.ops.read_i2c_eeprom(hw, i, &databyte);
7128 		else
7129 			status = hw->phy.ops.read_i2c_sff8472(hw, i, &databyte);
7130 
7131 		if (status != 0)
7132 			return -EIO;
7133 
7134 		data[i - info->offset] = databyte;
7135 	}
7136 
7137 	return 0;
7138 }
7139 
7140 uint16_t
7141 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
7142 	switch (mac_type) {
7143 	case ixgbe_mac_X550:
7144 	case ixgbe_mac_X550EM_x:
7145 	case ixgbe_mac_X550EM_a:
7146 		return ETH_RSS_RETA_SIZE_512;
7147 	case ixgbe_mac_X550_vf:
7148 	case ixgbe_mac_X550EM_x_vf:
7149 	case ixgbe_mac_X550EM_a_vf:
7150 		return ETH_RSS_RETA_SIZE_64;
7151 	case ixgbe_mac_X540_vf:
7152 	case ixgbe_mac_82599_vf:
7153 		return 0;
7154 	default:
7155 		return ETH_RSS_RETA_SIZE_128;
7156 	}
7157 }
7158 
7159 uint32_t
7160 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
7161 	switch (mac_type) {
7162 	case ixgbe_mac_X550:
7163 	case ixgbe_mac_X550EM_x:
7164 	case ixgbe_mac_X550EM_a:
7165 		if (reta_idx < ETH_RSS_RETA_SIZE_128)
7166 			return IXGBE_RETA(reta_idx >> 2);
7167 		else
7168 			return IXGBE_ERETA((reta_idx - ETH_RSS_RETA_SIZE_128) >> 2);
7169 	case ixgbe_mac_X550_vf:
7170 	case ixgbe_mac_X550EM_x_vf:
7171 	case ixgbe_mac_X550EM_a_vf:
7172 		return IXGBE_VFRETA(reta_idx >> 2);
7173 	default:
7174 		return IXGBE_RETA(reta_idx >> 2);
7175 	}
7176 }
7177 
7178 uint32_t
7179 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
7180 	switch (mac_type) {
7181 	case ixgbe_mac_X550_vf:
7182 	case ixgbe_mac_X550EM_x_vf:
7183 	case ixgbe_mac_X550EM_a_vf:
7184 		return IXGBE_VFMRQC;
7185 	default:
7186 		return IXGBE_MRQC;
7187 	}
7188 }
7189 
7190 uint32_t
7191 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
7192 	switch (mac_type) {
7193 	case ixgbe_mac_X550_vf:
7194 	case ixgbe_mac_X550EM_x_vf:
7195 	case ixgbe_mac_X550EM_a_vf:
7196 		return IXGBE_VFRSSRK(i);
7197 	default:
7198 		return IXGBE_RSSRK(i);
7199 	}
7200 }
7201 
7202 bool
7203 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
7204 	switch (mac_type) {
7205 	case ixgbe_mac_82599_vf:
7206 	case ixgbe_mac_X540_vf:
7207 		return 0;
7208 	default:
7209 		return 1;
7210 	}
7211 }
7212 
7213 static int
7214 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
7215 			struct rte_eth_dcb_info *dcb_info)
7216 {
7217 	struct ixgbe_dcb_config *dcb_config =
7218 			IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
7219 	struct ixgbe_dcb_tc_config *tc;
7220 	struct rte_eth_dcb_tc_queue_mapping *tc_queue;
7221 	uint8_t nb_tcs;
7222 	uint8_t i, j;
7223 
7224 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_DCB_FLAG)
7225 		dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
7226 	else
7227 		dcb_info->nb_tcs = 1;
7228 
7229 	tc_queue = &dcb_info->tc_queue;
7230 	nb_tcs = dcb_info->nb_tcs;
7231 
7232 	if (dcb_config->vt_mode) { /* vt is enabled*/
7233 		struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
7234 				&dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
7235 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7236 			dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
7237 		if (RTE_ETH_DEV_SRIOV(dev).active > 0) {
7238 			for (j = 0; j < nb_tcs; j++) {
7239 				tc_queue->tc_rxq[0][j].base = j;
7240 				tc_queue->tc_rxq[0][j].nb_queue = 1;
7241 				tc_queue->tc_txq[0][j].base = j;
7242 				tc_queue->tc_txq[0][j].nb_queue = 1;
7243 			}
7244 		} else {
7245 			for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
7246 				for (j = 0; j < nb_tcs; j++) {
7247 					tc_queue->tc_rxq[i][j].base =
7248 						i * nb_tcs + j;
7249 					tc_queue->tc_rxq[i][j].nb_queue = 1;
7250 					tc_queue->tc_txq[i][j].base =
7251 						i * nb_tcs + j;
7252 					tc_queue->tc_txq[i][j].nb_queue = 1;
7253 				}
7254 			}
7255 		}
7256 	} else { /* vt is disabled*/
7257 		struct rte_eth_dcb_rx_conf *rx_conf =
7258 				&dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
7259 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7260 			dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
7261 		if (dcb_info->nb_tcs == ETH_4_TCS) {
7262 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7263 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
7264 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7265 			}
7266 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7267 			dcb_info->tc_queue.tc_txq[0][1].base = 64;
7268 			dcb_info->tc_queue.tc_txq[0][2].base = 96;
7269 			dcb_info->tc_queue.tc_txq[0][3].base = 112;
7270 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
7271 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7272 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7273 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7274 		} else if (dcb_info->nb_tcs == ETH_8_TCS) {
7275 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7276 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
7277 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7278 			}
7279 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7280 			dcb_info->tc_queue.tc_txq[0][1].base = 32;
7281 			dcb_info->tc_queue.tc_txq[0][2].base = 64;
7282 			dcb_info->tc_queue.tc_txq[0][3].base = 80;
7283 			dcb_info->tc_queue.tc_txq[0][4].base = 96;
7284 			dcb_info->tc_queue.tc_txq[0][5].base = 104;
7285 			dcb_info->tc_queue.tc_txq[0][6].base = 112;
7286 			dcb_info->tc_queue.tc_txq[0][7].base = 120;
7287 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
7288 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7289 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7290 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7291 			dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
7292 			dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
7293 			dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
7294 			dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
7295 		}
7296 	}
7297 	for (i = 0; i < dcb_info->nb_tcs; i++) {
7298 		tc = &dcb_config->tc_config[i];
7299 		dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
7300 	}
7301 	return 0;
7302 }
7303 
7304 /* Update e-tag ether type */
7305 static int
7306 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
7307 			    uint16_t ether_type)
7308 {
7309 	uint32_t etag_etype;
7310 
7311 	if (hw->mac.type != ixgbe_mac_X550 &&
7312 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7313 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7314 		return -ENOTSUP;
7315 	}
7316 
7317 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7318 	etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
7319 	etag_etype |= ether_type;
7320 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7321 	IXGBE_WRITE_FLUSH(hw);
7322 
7323 	return 0;
7324 }
7325 
7326 /* Enable e-tag tunnel */
7327 static int
7328 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
7329 {
7330 	uint32_t etag_etype;
7331 
7332 	if (hw->mac.type != ixgbe_mac_X550 &&
7333 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7334 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7335 		return -ENOTSUP;
7336 	}
7337 
7338 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7339 	etag_etype |= IXGBE_ETAG_ETYPE_VALID;
7340 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7341 	IXGBE_WRITE_FLUSH(hw);
7342 
7343 	return 0;
7344 }
7345 
7346 static int
7347 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
7348 		       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7349 {
7350 	int ret = 0;
7351 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7352 	uint32_t i, rar_entries;
7353 	uint32_t rar_low, rar_high;
7354 
7355 	if (hw->mac.type != ixgbe_mac_X550 &&
7356 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7357 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7358 		return -ENOTSUP;
7359 	}
7360 
7361 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7362 
7363 	for (i = 1; i < rar_entries; i++) {
7364 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7365 		rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
7366 		if ((rar_high & IXGBE_RAH_AV) &&
7367 		    (rar_high & IXGBE_RAH_ADTYPE) &&
7368 		    ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
7369 		     l2_tunnel->tunnel_id)) {
7370 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
7371 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
7372 
7373 			ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
7374 
7375 			return ret;
7376 		}
7377 	}
7378 
7379 	return ret;
7380 }
7381 
7382 static int
7383 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
7384 		       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7385 {
7386 	int ret = 0;
7387 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7388 	uint32_t i, rar_entries;
7389 	uint32_t rar_low, rar_high;
7390 
7391 	if (hw->mac.type != ixgbe_mac_X550 &&
7392 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7393 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7394 		return -ENOTSUP;
7395 	}
7396 
7397 	/* One entry for one tunnel. Try to remove potential existing entry. */
7398 	ixgbe_e_tag_filter_del(dev, l2_tunnel);
7399 
7400 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7401 
7402 	for (i = 1; i < rar_entries; i++) {
7403 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7404 		if (rar_high & IXGBE_RAH_AV) {
7405 			continue;
7406 		} else {
7407 			ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
7408 			rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
7409 			rar_low = l2_tunnel->tunnel_id;
7410 
7411 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
7412 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
7413 
7414 			return ret;
7415 		}
7416 	}
7417 
7418 	PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
7419 		     " Please remove a rule before adding a new one.");
7420 	return -EINVAL;
7421 }
7422 
7423 static inline struct ixgbe_l2_tn_filter *
7424 ixgbe_l2_tn_filter_lookup(struct ixgbe_l2_tn_info *l2_tn_info,
7425 			  struct ixgbe_l2_tn_key *key)
7426 {
7427 	int ret;
7428 
7429 	ret = rte_hash_lookup(l2_tn_info->hash_handle, (const void *)key);
7430 	if (ret < 0)
7431 		return NULL;
7432 
7433 	return l2_tn_info->hash_map[ret];
7434 }
7435 
7436 static inline int
7437 ixgbe_insert_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7438 			  struct ixgbe_l2_tn_filter *l2_tn_filter)
7439 {
7440 	int ret;
7441 
7442 	ret = rte_hash_add_key(l2_tn_info->hash_handle,
7443 			       &l2_tn_filter->key);
7444 
7445 	if (ret < 0) {
7446 		PMD_DRV_LOG(ERR,
7447 			    "Failed to insert L2 tunnel filter"
7448 			    " to hash table %d!",
7449 			    ret);
7450 		return ret;
7451 	}
7452 
7453 	l2_tn_info->hash_map[ret] = l2_tn_filter;
7454 
7455 	TAILQ_INSERT_TAIL(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7456 
7457 	return 0;
7458 }
7459 
7460 static inline int
7461 ixgbe_remove_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7462 			  struct ixgbe_l2_tn_key *key)
7463 {
7464 	int ret;
7465 	struct ixgbe_l2_tn_filter *l2_tn_filter;
7466 
7467 	ret = rte_hash_del_key(l2_tn_info->hash_handle, key);
7468 
7469 	if (ret < 0) {
7470 		PMD_DRV_LOG(ERR,
7471 			    "No such L2 tunnel filter to delete %d!",
7472 			    ret);
7473 		return ret;
7474 	}
7475 
7476 	l2_tn_filter = l2_tn_info->hash_map[ret];
7477 	l2_tn_info->hash_map[ret] = NULL;
7478 
7479 	TAILQ_REMOVE(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7480 	rte_free(l2_tn_filter);
7481 
7482 	return 0;
7483 }
7484 
7485 /* Add l2 tunnel filter */
7486 int
7487 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
7488 			       struct ixgbe_l2_tunnel_conf *l2_tunnel,
7489 			       bool restore)
7490 {
7491 	int ret;
7492 	struct ixgbe_l2_tn_info *l2_tn_info =
7493 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7494 	struct ixgbe_l2_tn_key key;
7495 	struct ixgbe_l2_tn_filter *node;
7496 
7497 	if (!restore) {
7498 		key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7499 		key.tn_id = l2_tunnel->tunnel_id;
7500 
7501 		node = ixgbe_l2_tn_filter_lookup(l2_tn_info, &key);
7502 
7503 		if (node) {
7504 			PMD_DRV_LOG(ERR,
7505 				    "The L2 tunnel filter already exists!");
7506 			return -EINVAL;
7507 		}
7508 
7509 		node = rte_zmalloc("ixgbe_l2_tn",
7510 				   sizeof(struct ixgbe_l2_tn_filter),
7511 				   0);
7512 		if (!node)
7513 			return -ENOMEM;
7514 
7515 		rte_memcpy(&node->key,
7516 				 &key,
7517 				 sizeof(struct ixgbe_l2_tn_key));
7518 		node->pool = l2_tunnel->pool;
7519 		ret = ixgbe_insert_l2_tn_filter(l2_tn_info, node);
7520 		if (ret < 0) {
7521 			rte_free(node);
7522 			return ret;
7523 		}
7524 	}
7525 
7526 	switch (l2_tunnel->l2_tunnel_type) {
7527 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7528 		ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
7529 		break;
7530 	default:
7531 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7532 		ret = -EINVAL;
7533 		break;
7534 	}
7535 
7536 	if ((!restore) && (ret < 0))
7537 		(void)ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7538 
7539 	return ret;
7540 }
7541 
7542 /* Delete l2 tunnel filter */
7543 int
7544 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
7545 			       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7546 {
7547 	int ret;
7548 	struct ixgbe_l2_tn_info *l2_tn_info =
7549 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7550 	struct ixgbe_l2_tn_key key;
7551 
7552 	key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7553 	key.tn_id = l2_tunnel->tunnel_id;
7554 	ret = ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7555 	if (ret < 0)
7556 		return ret;
7557 
7558 	switch (l2_tunnel->l2_tunnel_type) {
7559 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7560 		ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
7561 		break;
7562 	default:
7563 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7564 		ret = -EINVAL;
7565 		break;
7566 	}
7567 
7568 	return ret;
7569 }
7570 
7571 static int
7572 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
7573 {
7574 	int ret = 0;
7575 	uint32_t ctrl;
7576 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7577 
7578 	if (hw->mac.type != ixgbe_mac_X550 &&
7579 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7580 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7581 		return -ENOTSUP;
7582 	}
7583 
7584 	ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
7585 	ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
7586 	if (en)
7587 		ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
7588 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
7589 
7590 	return ret;
7591 }
7592 
7593 static int
7594 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
7595 			uint16_t port)
7596 {
7597 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
7598 	IXGBE_WRITE_FLUSH(hw);
7599 
7600 	return 0;
7601 }
7602 
7603 /* There's only one register for VxLAN UDP port.
7604  * So, we cannot add several ports. Will update it.
7605  */
7606 static int
7607 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
7608 		     uint16_t port)
7609 {
7610 	if (port == 0) {
7611 		PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
7612 		return -EINVAL;
7613 	}
7614 
7615 	return ixgbe_update_vxlan_port(hw, port);
7616 }
7617 
7618 /* We cannot delete the VxLAN port. For there's a register for VxLAN
7619  * UDP port, it must have a value.
7620  * So, will reset it to the original value 0.
7621  */
7622 static int
7623 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
7624 		     uint16_t port)
7625 {
7626 	uint16_t cur_port;
7627 
7628 	cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
7629 
7630 	if (cur_port != port) {
7631 		PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
7632 		return -EINVAL;
7633 	}
7634 
7635 	return ixgbe_update_vxlan_port(hw, 0);
7636 }
7637 
7638 /* Add UDP tunneling port */
7639 static int
7640 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
7641 			      struct rte_eth_udp_tunnel *udp_tunnel)
7642 {
7643 	int ret = 0;
7644 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7645 
7646 	if (hw->mac.type != ixgbe_mac_X550 &&
7647 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7648 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7649 		return -ENOTSUP;
7650 	}
7651 
7652 	if (udp_tunnel == NULL)
7653 		return -EINVAL;
7654 
7655 	switch (udp_tunnel->prot_type) {
7656 	case RTE_TUNNEL_TYPE_VXLAN:
7657 		ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
7658 		break;
7659 
7660 	case RTE_TUNNEL_TYPE_GENEVE:
7661 	case RTE_TUNNEL_TYPE_TEREDO:
7662 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7663 		ret = -EINVAL;
7664 		break;
7665 
7666 	default:
7667 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7668 		ret = -EINVAL;
7669 		break;
7670 	}
7671 
7672 	return ret;
7673 }
7674 
7675 /* Remove UDP tunneling port */
7676 static int
7677 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
7678 			      struct rte_eth_udp_tunnel *udp_tunnel)
7679 {
7680 	int ret = 0;
7681 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7682 
7683 	if (hw->mac.type != ixgbe_mac_X550 &&
7684 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7685 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7686 		return -ENOTSUP;
7687 	}
7688 
7689 	if (udp_tunnel == NULL)
7690 		return -EINVAL;
7691 
7692 	switch (udp_tunnel->prot_type) {
7693 	case RTE_TUNNEL_TYPE_VXLAN:
7694 		ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
7695 		break;
7696 	case RTE_TUNNEL_TYPE_GENEVE:
7697 	case RTE_TUNNEL_TYPE_TEREDO:
7698 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7699 		ret = -EINVAL;
7700 		break;
7701 	default:
7702 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7703 		ret = -EINVAL;
7704 		break;
7705 	}
7706 
7707 	return ret;
7708 }
7709 
7710 static int
7711 ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev)
7712 {
7713 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7714 	int ret;
7715 
7716 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_PROMISC)) {
7717 	case IXGBE_SUCCESS:
7718 		ret = 0;
7719 		break;
7720 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7721 		ret = -ENOTSUP;
7722 		break;
7723 	default:
7724 		ret = -EAGAIN;
7725 		break;
7726 	}
7727 
7728 	return ret;
7729 }
7730 
7731 static int
7732 ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev)
7733 {
7734 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7735 	int ret;
7736 
7737 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_NONE)) {
7738 	case IXGBE_SUCCESS:
7739 		ret = 0;
7740 		break;
7741 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7742 		ret = -ENOTSUP;
7743 		break;
7744 	default:
7745 		ret = -EAGAIN;
7746 		break;
7747 	}
7748 
7749 	return ret;
7750 }
7751 
7752 static int
7753 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
7754 {
7755 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7756 	int ret;
7757 	int mode = IXGBEVF_XCAST_MODE_ALLMULTI;
7758 
7759 	switch (hw->mac.ops.update_xcast_mode(hw, mode)) {
7760 	case IXGBE_SUCCESS:
7761 		ret = 0;
7762 		break;
7763 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7764 		ret = -ENOTSUP;
7765 		break;
7766 	default:
7767 		ret = -EAGAIN;
7768 		break;
7769 	}
7770 
7771 	return ret;
7772 }
7773 
7774 static int
7775 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
7776 {
7777 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7778 	int ret;
7779 
7780 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_MULTI)) {
7781 	case IXGBE_SUCCESS:
7782 		ret = 0;
7783 		break;
7784 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7785 		ret = -ENOTSUP;
7786 		break;
7787 	default:
7788 		ret = -EAGAIN;
7789 		break;
7790 	}
7791 
7792 	return ret;
7793 }
7794 
7795 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
7796 {
7797 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7798 	u32 in_msg = 0;
7799 
7800 	/* peek the message first */
7801 	in_msg = IXGBE_READ_REG(hw, IXGBE_VFMBMEM);
7802 
7803 	/* PF reset VF event */
7804 	if (in_msg == IXGBE_PF_CONTROL_MSG) {
7805 		/* dummy mbx read to ack pf */
7806 		if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
7807 			return;
7808 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
7809 					     NULL);
7810 	}
7811 }
7812 
7813 static int
7814 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
7815 {
7816 	uint32_t eicr;
7817 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7818 	struct ixgbe_interrupt *intr =
7819 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7820 	ixgbevf_intr_disable(dev);
7821 
7822 	/* read-on-clear nic registers here */
7823 	eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
7824 	intr->flags = 0;
7825 
7826 	/* only one misc vector supported - mailbox */
7827 	eicr &= IXGBE_VTEICR_MASK;
7828 	if (eicr == IXGBE_MISC_VEC_ID)
7829 		intr->flags |= IXGBE_FLAG_MAILBOX;
7830 
7831 	return 0;
7832 }
7833 
7834 static int
7835 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
7836 {
7837 	struct ixgbe_interrupt *intr =
7838 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7839 
7840 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
7841 		ixgbevf_mbx_process(dev);
7842 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
7843 	}
7844 
7845 	ixgbevf_intr_enable(dev);
7846 
7847 	return 0;
7848 }
7849 
7850 static void
7851 ixgbevf_dev_interrupt_handler(void *param)
7852 {
7853 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
7854 
7855 	ixgbevf_dev_interrupt_get_status(dev);
7856 	ixgbevf_dev_interrupt_action(dev);
7857 }
7858 
7859 /**
7860  *  ixgbe_disable_sec_tx_path_generic - Stops the transmit data path
7861  *  @hw: pointer to hardware structure
7862  *
7863  *  Stops the transmit data path and waits for the HW to internally empty
7864  *  the Tx security block
7865  **/
7866 int ixgbe_disable_sec_tx_path_generic(struct ixgbe_hw *hw)
7867 {
7868 #define IXGBE_MAX_SECTX_POLL 40
7869 
7870 	int i;
7871 	int sectxreg;
7872 
7873 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
7874 	sectxreg |= IXGBE_SECTXCTRL_TX_DIS;
7875 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
7876 	for (i = 0; i < IXGBE_MAX_SECTX_POLL; i++) {
7877 		sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXSTAT);
7878 		if (sectxreg & IXGBE_SECTXSTAT_SECTX_RDY)
7879 			break;
7880 		/* Use interrupt-safe sleep just in case */
7881 		usec_delay(1000);
7882 	}
7883 
7884 	/* For informational purposes only */
7885 	if (i >= IXGBE_MAX_SECTX_POLL)
7886 		PMD_DRV_LOG(DEBUG, "Tx unit being enabled before security "
7887 			 "path fully disabled.  Continuing with init.");
7888 
7889 	return IXGBE_SUCCESS;
7890 }
7891 
7892 /**
7893  *  ixgbe_enable_sec_tx_path_generic - Enables the transmit data path
7894  *  @hw: pointer to hardware structure
7895  *
7896  *  Enables the transmit data path.
7897  **/
7898 int ixgbe_enable_sec_tx_path_generic(struct ixgbe_hw *hw)
7899 {
7900 	uint32_t sectxreg;
7901 
7902 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
7903 	sectxreg &= ~IXGBE_SECTXCTRL_TX_DIS;
7904 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
7905 	IXGBE_WRITE_FLUSH(hw);
7906 
7907 	return IXGBE_SUCCESS;
7908 }
7909 
7910 /* restore n-tuple filter */
7911 static inline void
7912 ixgbe_ntuple_filter_restore(struct rte_eth_dev *dev)
7913 {
7914 	struct ixgbe_filter_info *filter_info =
7915 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7916 	struct ixgbe_5tuple_filter *node;
7917 
7918 	TAILQ_FOREACH(node, &filter_info->fivetuple_list, entries) {
7919 		ixgbe_inject_5tuple_filter(dev, node);
7920 	}
7921 }
7922 
7923 /* restore ethernet type filter */
7924 static inline void
7925 ixgbe_ethertype_filter_restore(struct rte_eth_dev *dev)
7926 {
7927 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7928 	struct ixgbe_filter_info *filter_info =
7929 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7930 	int i;
7931 
7932 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
7933 		if (filter_info->ethertype_mask & (1 << i)) {
7934 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i),
7935 					filter_info->ethertype_filters[i].etqf);
7936 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i),
7937 					filter_info->ethertype_filters[i].etqs);
7938 			IXGBE_WRITE_FLUSH(hw);
7939 		}
7940 	}
7941 }
7942 
7943 /* restore SYN filter */
7944 static inline void
7945 ixgbe_syn_filter_restore(struct rte_eth_dev *dev)
7946 {
7947 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7948 	struct ixgbe_filter_info *filter_info =
7949 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7950 	uint32_t synqf;
7951 
7952 	synqf = filter_info->syn_info;
7953 
7954 	if (synqf & IXGBE_SYN_FILTER_ENABLE) {
7955 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
7956 		IXGBE_WRITE_FLUSH(hw);
7957 	}
7958 }
7959 
7960 /* restore L2 tunnel filter */
7961 static inline void
7962 ixgbe_l2_tn_filter_restore(struct rte_eth_dev *dev)
7963 {
7964 	struct ixgbe_l2_tn_info *l2_tn_info =
7965 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7966 	struct ixgbe_l2_tn_filter *node;
7967 	struct ixgbe_l2_tunnel_conf l2_tn_conf;
7968 
7969 	TAILQ_FOREACH(node, &l2_tn_info->l2_tn_list, entries) {
7970 		l2_tn_conf.l2_tunnel_type = node->key.l2_tn_type;
7971 		l2_tn_conf.tunnel_id      = node->key.tn_id;
7972 		l2_tn_conf.pool           = node->pool;
7973 		(void)ixgbe_dev_l2_tunnel_filter_add(dev, &l2_tn_conf, TRUE);
7974 	}
7975 }
7976 
7977 /* restore rss filter */
7978 static inline void
7979 ixgbe_rss_filter_restore(struct rte_eth_dev *dev)
7980 {
7981 	struct ixgbe_filter_info *filter_info =
7982 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7983 
7984 	if (filter_info->rss_info.conf.queue_num)
7985 		ixgbe_config_rss_filter(dev,
7986 			&filter_info->rss_info, TRUE);
7987 }
7988 
7989 static int
7990 ixgbe_filter_restore(struct rte_eth_dev *dev)
7991 {
7992 	ixgbe_ntuple_filter_restore(dev);
7993 	ixgbe_ethertype_filter_restore(dev);
7994 	ixgbe_syn_filter_restore(dev);
7995 	ixgbe_fdir_filter_restore(dev);
7996 	ixgbe_l2_tn_filter_restore(dev);
7997 	ixgbe_rss_filter_restore(dev);
7998 
7999 	return 0;
8000 }
8001 
8002 static void
8003 ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev)
8004 {
8005 	struct ixgbe_l2_tn_info *l2_tn_info =
8006 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8007 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8008 
8009 	if (l2_tn_info->e_tag_en)
8010 		(void)ixgbe_e_tag_enable(hw);
8011 
8012 	if (l2_tn_info->e_tag_fwd_en)
8013 		(void)ixgbe_e_tag_forwarding_en_dis(dev, 1);
8014 
8015 	(void)ixgbe_update_e_tag_eth_type(hw, l2_tn_info->e_tag_ether_type);
8016 }
8017 
8018 /* remove all the n-tuple filters */
8019 void
8020 ixgbe_clear_all_ntuple_filter(struct rte_eth_dev *dev)
8021 {
8022 	struct ixgbe_filter_info *filter_info =
8023 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8024 	struct ixgbe_5tuple_filter *p_5tuple;
8025 
8026 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
8027 		ixgbe_remove_5tuple_filter(dev, p_5tuple);
8028 }
8029 
8030 /* remove all the ether type filters */
8031 void
8032 ixgbe_clear_all_ethertype_filter(struct rte_eth_dev *dev)
8033 {
8034 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8035 	struct ixgbe_filter_info *filter_info =
8036 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8037 	int i;
8038 
8039 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8040 		if (filter_info->ethertype_mask & (1 << i) &&
8041 		    !filter_info->ethertype_filters[i].conf) {
8042 			(void)ixgbe_ethertype_filter_remove(filter_info,
8043 							    (uint8_t)i);
8044 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i), 0);
8045 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i), 0);
8046 			IXGBE_WRITE_FLUSH(hw);
8047 		}
8048 	}
8049 }
8050 
8051 /* remove the SYN filter */
8052 void
8053 ixgbe_clear_syn_filter(struct rte_eth_dev *dev)
8054 {
8055 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8056 	struct ixgbe_filter_info *filter_info =
8057 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8058 
8059 	if (filter_info->syn_info & IXGBE_SYN_FILTER_ENABLE) {
8060 		filter_info->syn_info = 0;
8061 
8062 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, 0);
8063 		IXGBE_WRITE_FLUSH(hw);
8064 	}
8065 }
8066 
8067 /* remove all the L2 tunnel filters */
8068 int
8069 ixgbe_clear_all_l2_tn_filter(struct rte_eth_dev *dev)
8070 {
8071 	struct ixgbe_l2_tn_info *l2_tn_info =
8072 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8073 	struct ixgbe_l2_tn_filter *l2_tn_filter;
8074 	struct ixgbe_l2_tunnel_conf l2_tn_conf;
8075 	int ret = 0;
8076 
8077 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
8078 		l2_tn_conf.l2_tunnel_type = l2_tn_filter->key.l2_tn_type;
8079 		l2_tn_conf.tunnel_id      = l2_tn_filter->key.tn_id;
8080 		l2_tn_conf.pool           = l2_tn_filter->pool;
8081 		ret = ixgbe_dev_l2_tunnel_filter_del(dev, &l2_tn_conf);
8082 		if (ret < 0)
8083 			return ret;
8084 	}
8085 
8086 	return 0;
8087 }
8088 
8089 void
8090 ixgbe_dev_macsec_setting_save(struct rte_eth_dev *dev,
8091 				struct ixgbe_macsec_setting *macsec_setting)
8092 {
8093 	struct ixgbe_macsec_setting *macsec =
8094 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8095 
8096 	macsec->offload_en = macsec_setting->offload_en;
8097 	macsec->encrypt_en = macsec_setting->encrypt_en;
8098 	macsec->replayprotect_en = macsec_setting->replayprotect_en;
8099 }
8100 
8101 void
8102 ixgbe_dev_macsec_setting_reset(struct rte_eth_dev *dev)
8103 {
8104 	struct ixgbe_macsec_setting *macsec =
8105 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8106 
8107 	macsec->offload_en = 0;
8108 	macsec->encrypt_en = 0;
8109 	macsec->replayprotect_en = 0;
8110 }
8111 
8112 void
8113 ixgbe_dev_macsec_register_enable(struct rte_eth_dev *dev,
8114 				struct ixgbe_macsec_setting *macsec_setting)
8115 {
8116 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8117 	uint32_t ctrl;
8118 	uint8_t en = macsec_setting->encrypt_en;
8119 	uint8_t rp = macsec_setting->replayprotect_en;
8120 
8121 	/**
8122 	 * Workaround:
8123 	 * As no ixgbe_disable_sec_rx_path equivalent is
8124 	 * implemented for tx in the base code, and we are
8125 	 * not allowed to modify the base code in DPDK, so
8126 	 * just call the hand-written one directly for now.
8127 	 * The hardware support has been checked by
8128 	 * ixgbe_disable_sec_rx_path().
8129 	 */
8130 	ixgbe_disable_sec_tx_path_generic(hw);
8131 
8132 	/* Enable Ethernet CRC (required by MACsec offload) */
8133 	ctrl = IXGBE_READ_REG(hw, IXGBE_HLREG0);
8134 	ctrl |= IXGBE_HLREG0_TXCRCEN | IXGBE_HLREG0_RXCRCSTRP;
8135 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, ctrl);
8136 
8137 	/* Enable the TX and RX crypto engines */
8138 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8139 	ctrl &= ~IXGBE_SECTXCTRL_SECTX_DIS;
8140 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8141 
8142 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8143 	ctrl &= ~IXGBE_SECRXCTRL_SECRX_DIS;
8144 	IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8145 
8146 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
8147 	ctrl &= ~IXGBE_SECTX_MINSECIFG_MASK;
8148 	ctrl |= 0x3;
8149 	IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, ctrl);
8150 
8151 	/* Enable SA lookup */
8152 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8153 	ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8154 	ctrl |= en ? IXGBE_LSECTXCTRL_AUTH_ENCRYPT :
8155 		     IXGBE_LSECTXCTRL_AUTH;
8156 	ctrl |= IXGBE_LSECTXCTRL_AISCI;
8157 	ctrl &= ~IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8158 	ctrl |= IXGBE_MACSEC_PNTHRSH & IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8159 	IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8160 
8161 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8162 	ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8163 	ctrl |= IXGBE_LSECRXCTRL_STRICT << IXGBE_LSECRXCTRL_EN_SHIFT;
8164 	ctrl &= ~IXGBE_LSECRXCTRL_PLSH;
8165 	if (rp)
8166 		ctrl |= IXGBE_LSECRXCTRL_RP;
8167 	else
8168 		ctrl &= ~IXGBE_LSECRXCTRL_RP;
8169 	IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8170 
8171 	/* Start the data paths */
8172 	ixgbe_enable_sec_rx_path(hw);
8173 	/**
8174 	 * Workaround:
8175 	 * As no ixgbe_enable_sec_rx_path equivalent is
8176 	 * implemented for tx in the base code, and we are
8177 	 * not allowed to modify the base code in DPDK, so
8178 	 * just call the hand-written one directly for now.
8179 	 */
8180 	ixgbe_enable_sec_tx_path_generic(hw);
8181 }
8182 
8183 void
8184 ixgbe_dev_macsec_register_disable(struct rte_eth_dev *dev)
8185 {
8186 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8187 	uint32_t ctrl;
8188 
8189 	/**
8190 	 * Workaround:
8191 	 * As no ixgbe_disable_sec_rx_path equivalent is
8192 	 * implemented for tx in the base code, and we are
8193 	 * not allowed to modify the base code in DPDK, so
8194 	 * just call the hand-written one directly for now.
8195 	 * The hardware support has been checked by
8196 	 * ixgbe_disable_sec_rx_path().
8197 	 */
8198 	ixgbe_disable_sec_tx_path_generic(hw);
8199 
8200 	/* Disable the TX and RX crypto engines */
8201 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8202 	ctrl |= IXGBE_SECTXCTRL_SECTX_DIS;
8203 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8204 
8205 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8206 	ctrl |= IXGBE_SECRXCTRL_SECRX_DIS;
8207 	IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8208 
8209 	/* Disable SA lookup */
8210 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8211 	ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8212 	ctrl |= IXGBE_LSECTXCTRL_DISABLE;
8213 	IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8214 
8215 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8216 	ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8217 	ctrl |= IXGBE_LSECRXCTRL_DISABLE << IXGBE_LSECRXCTRL_EN_SHIFT;
8218 	IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8219 
8220 	/* Start the data paths */
8221 	ixgbe_enable_sec_rx_path(hw);
8222 	/**
8223 	 * Workaround:
8224 	 * As no ixgbe_enable_sec_rx_path equivalent is
8225 	 * implemented for tx in the base code, and we are
8226 	 * not allowed to modify the base code in DPDK, so
8227 	 * just call the hand-written one directly for now.
8228 	 */
8229 	ixgbe_enable_sec_tx_path_generic(hw);
8230 }
8231 
8232 RTE_PMD_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd);
8233 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
8234 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe, "* igb_uio | uio_pci_generic | vfio-pci");
8235 RTE_PMD_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd);
8236 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);
8237 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe_vf, "* igb_uio | vfio-pci");
8238 RTE_PMD_REGISTER_PARAM_STRING(net_ixgbe_vf,
8239 			      IXGBEVF_DEVARG_PFLINK_FULLCHK "=<0|1>");
8240 
8241 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_init, init, NOTICE);
8242 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_driver, driver, NOTICE);
8243 
8244 #ifdef RTE_ETHDEV_DEBUG_RX
8245 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_rx, rx, DEBUG);
8246 #endif
8247 #ifdef RTE_ETHDEV_DEBUG_TX
8248 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_tx, tx, DEBUG);
8249 #endif
8250