xref: /dpdk/drivers/net/ixgbe/ixgbe_ethdev.c (revision e0ae3db0)
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_done   = ixgbe_dev_rx_descriptor_done;
1054 	eth_dev->rx_descriptor_status = ixgbe_dev_rx_descriptor_status;
1055 	eth_dev->tx_descriptor_status = ixgbe_dev_tx_descriptor_status;
1056 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1057 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1058 	eth_dev->tx_pkt_prepare = &ixgbe_prep_pkts;
1059 
1060 	/*
1061 	 * For secondary processes, we don't initialise any further as primary
1062 	 * has already done this work. Only check we don't need a different
1063 	 * RX and TX function.
1064 	 */
1065 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1066 		struct ixgbe_tx_queue *txq;
1067 		/* TX queue function in primary, set by last queue initialized
1068 		 * Tx queue may not initialized by primary process
1069 		 */
1070 		if (eth_dev->data->tx_queues) {
1071 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
1072 			ixgbe_set_tx_function(eth_dev, txq);
1073 		} else {
1074 			/* Use default TX function if we get here */
1075 			PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
1076 				     "Using default TX function.");
1077 		}
1078 
1079 		ixgbe_set_rx_function(eth_dev);
1080 
1081 		return 0;
1082 	}
1083 
1084 	rte_atomic32_clear(&ad->link_thread_running);
1085 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1086 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1087 
1088 	/* Vendor and Device ID need to be set before init of shared code */
1089 	hw->device_id = pci_dev->id.device_id;
1090 	hw->vendor_id = pci_dev->id.vendor_id;
1091 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1092 	hw->allow_unsupported_sfp = 1;
1093 
1094 	/* Initialize the shared code (base driver) */
1095 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1096 	diag = ixgbe_bypass_init_shared_code(hw);
1097 #else
1098 	diag = ixgbe_init_shared_code(hw);
1099 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1100 
1101 	if (diag != IXGBE_SUCCESS) {
1102 		PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
1103 		return -EIO;
1104 	}
1105 
1106 	if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
1107 		PMD_INIT_LOG(ERR, "\nERROR: "
1108 			"Firmware recovery mode detected. Limiting functionality.\n"
1109 			"Refer to the Intel(R) Ethernet Adapters and Devices "
1110 			"User Guide for details on firmware recovery mode.");
1111 		return -EIO;
1112 	}
1113 
1114 	/* pick up the PCI bus settings for reporting later */
1115 	ixgbe_get_bus_info(hw);
1116 
1117 	/* Unlock any pending hardware semaphore */
1118 	ixgbe_swfw_lock_reset(hw);
1119 
1120 #ifdef RTE_LIB_SECURITY
1121 	/* Initialize security_ctx only for primary process*/
1122 	if (ixgbe_ipsec_ctx_create(eth_dev))
1123 		return -ENOMEM;
1124 #endif
1125 
1126 	/* Initialize DCB configuration*/
1127 	memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
1128 	ixgbe_dcb_init(hw, dcb_config);
1129 	/* Get Hardware Flow Control setting */
1130 	hw->fc.requested_mode = ixgbe_fc_none;
1131 	hw->fc.current_mode = ixgbe_fc_none;
1132 	hw->fc.pause_time = IXGBE_FC_PAUSE;
1133 	for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
1134 		hw->fc.low_water[i] = IXGBE_FC_LO;
1135 		hw->fc.high_water[i] = IXGBE_FC_HI;
1136 	}
1137 	hw->fc.send_xon = 1;
1138 
1139 	/* Make sure we have a good EEPROM before we read from it */
1140 	diag = ixgbe_validate_eeprom_checksum(hw, &csum);
1141 	if (diag != IXGBE_SUCCESS) {
1142 		PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
1143 		return -EIO;
1144 	}
1145 
1146 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1147 	diag = ixgbe_bypass_init_hw(hw);
1148 #else
1149 	diag = ixgbe_init_hw(hw);
1150 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1151 
1152 	/*
1153 	 * Devices with copper phys will fail to initialise if ixgbe_init_hw()
1154 	 * is called too soon after the kernel driver unbinding/binding occurs.
1155 	 * The failure occurs in ixgbe_identify_phy_generic() for all devices,
1156 	 * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
1157 	 * also called. See ixgbe_identify_phy_82599(). The reason for the
1158 	 * failure is not known, and only occuts when virtualisation features
1159 	 * are disabled in the bios. A delay of 100ms  was found to be enough by
1160 	 * trial-and-error, and is doubled to be safe.
1161 	 */
1162 	if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
1163 		rte_delay_ms(200);
1164 		diag = ixgbe_init_hw(hw);
1165 	}
1166 
1167 	if (diag == IXGBE_ERR_SFP_NOT_PRESENT)
1168 		diag = IXGBE_SUCCESS;
1169 
1170 	if (diag == IXGBE_ERR_EEPROM_VERSION) {
1171 		PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
1172 			     "LOM.  Please be aware there may be issues associated "
1173 			     "with your hardware.");
1174 		PMD_INIT_LOG(ERR, "If you are experiencing problems "
1175 			     "please contact your Intel or hardware representative "
1176 			     "who provided you with this hardware.");
1177 	} else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
1178 		PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
1179 	if (diag) {
1180 		PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
1181 		return -EIO;
1182 	}
1183 
1184 	/* Reset the hw statistics */
1185 	ixgbe_dev_stats_reset(eth_dev);
1186 
1187 	/* disable interrupt */
1188 	ixgbe_disable_intr(hw);
1189 
1190 	/* reset mappings for queue statistics hw counters*/
1191 	ixgbe_reset_qstat_mappings(hw);
1192 
1193 	/* Allocate memory for storing MAC addresses */
1194 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", RTE_ETHER_ADDR_LEN *
1195 					       hw->mac.num_rar_entries, 0);
1196 	if (eth_dev->data->mac_addrs == NULL) {
1197 		PMD_INIT_LOG(ERR,
1198 			     "Failed to allocate %u bytes needed to store "
1199 			     "MAC addresses",
1200 			     RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1201 		return -ENOMEM;
1202 	}
1203 	/* Copy the permanent MAC address */
1204 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.perm_addr,
1205 			&eth_dev->data->mac_addrs[0]);
1206 
1207 	/* Allocate memory for storing hash filter MAC addresses */
1208 	eth_dev->data->hash_mac_addrs = rte_zmalloc(
1209 		"ixgbe", RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC, 0);
1210 	if (eth_dev->data->hash_mac_addrs == NULL) {
1211 		PMD_INIT_LOG(ERR,
1212 			     "Failed to allocate %d bytes needed to store MAC addresses",
1213 			     RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
1214 		rte_free(eth_dev->data->mac_addrs);
1215 		eth_dev->data->mac_addrs = NULL;
1216 		return -ENOMEM;
1217 	}
1218 
1219 	/* initialize the vfta */
1220 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1221 
1222 	/* initialize the hw strip bitmap*/
1223 	memset(hwstrip, 0, sizeof(*hwstrip));
1224 
1225 	/* initialize PF if max_vfs not zero */
1226 	ret = ixgbe_pf_host_init(eth_dev);
1227 	if (ret) {
1228 		rte_free(eth_dev->data->mac_addrs);
1229 		eth_dev->data->mac_addrs = NULL;
1230 		rte_free(eth_dev->data->hash_mac_addrs);
1231 		eth_dev->data->hash_mac_addrs = NULL;
1232 		return ret;
1233 	}
1234 
1235 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1236 	/* let hardware know driver is loaded */
1237 	ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1238 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1239 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1240 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1241 	IXGBE_WRITE_FLUSH(hw);
1242 
1243 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1244 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1245 			     (int) hw->mac.type, (int) hw->phy.type,
1246 			     (int) hw->phy.sfp_type);
1247 	else
1248 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1249 			     (int) hw->mac.type, (int) hw->phy.type);
1250 
1251 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1252 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1253 		     pci_dev->id.device_id);
1254 
1255 	rte_intr_callback_register(intr_handle,
1256 				   ixgbe_dev_interrupt_handler, eth_dev);
1257 
1258 	/* enable uio/vfio intr/eventfd mapping */
1259 	rte_intr_enable(intr_handle);
1260 
1261 	/* enable support intr */
1262 	ixgbe_enable_intr(eth_dev);
1263 
1264 	/* initialize filter info */
1265 	memset(filter_info, 0,
1266 	       sizeof(struct ixgbe_filter_info));
1267 
1268 	/* initialize 5tuple filter list */
1269 	TAILQ_INIT(&filter_info->fivetuple_list);
1270 
1271 	/* initialize flow director filter list & hash */
1272 	ixgbe_fdir_filter_init(eth_dev);
1273 
1274 	/* initialize l2 tunnel filter list & hash */
1275 	ixgbe_l2_tn_filter_init(eth_dev);
1276 
1277 	/* initialize flow filter lists */
1278 	ixgbe_filterlist_init();
1279 
1280 	/* initialize bandwidth configuration info */
1281 	memset(bw_conf, 0, sizeof(struct ixgbe_bw_conf));
1282 
1283 	/* initialize Traffic Manager configuration */
1284 	ixgbe_tm_conf_init(eth_dev);
1285 
1286 	return 0;
1287 }
1288 
1289 static int
1290 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1291 {
1292 	PMD_INIT_FUNC_TRACE();
1293 
1294 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1295 		return 0;
1296 
1297 	ixgbe_dev_close(eth_dev);
1298 
1299 	return 0;
1300 }
1301 
1302 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
1303 {
1304 	struct ixgbe_filter_info *filter_info =
1305 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1306 	struct ixgbe_5tuple_filter *p_5tuple;
1307 
1308 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
1309 		TAILQ_REMOVE(&filter_info->fivetuple_list,
1310 			     p_5tuple,
1311 			     entries);
1312 		rte_free(p_5tuple);
1313 	}
1314 	memset(filter_info->fivetuple_mask, 0,
1315 	       sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1316 
1317 	return 0;
1318 }
1319 
1320 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev)
1321 {
1322 	struct ixgbe_hw_fdir_info *fdir_info =
1323 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1324 	struct ixgbe_fdir_filter *fdir_filter;
1325 
1326 		if (fdir_info->hash_map)
1327 		rte_free(fdir_info->hash_map);
1328 	if (fdir_info->hash_handle)
1329 		rte_hash_free(fdir_info->hash_handle);
1330 
1331 	while ((fdir_filter = TAILQ_FIRST(&fdir_info->fdir_list))) {
1332 		TAILQ_REMOVE(&fdir_info->fdir_list,
1333 			     fdir_filter,
1334 			     entries);
1335 		rte_free(fdir_filter);
1336 	}
1337 
1338 	return 0;
1339 }
1340 
1341 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev)
1342 {
1343 	struct ixgbe_l2_tn_info *l2_tn_info =
1344 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1345 	struct ixgbe_l2_tn_filter *l2_tn_filter;
1346 
1347 	if (l2_tn_info->hash_map)
1348 		rte_free(l2_tn_info->hash_map);
1349 	if (l2_tn_info->hash_handle)
1350 		rte_hash_free(l2_tn_info->hash_handle);
1351 
1352 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
1353 		TAILQ_REMOVE(&l2_tn_info->l2_tn_list,
1354 			     l2_tn_filter,
1355 			     entries);
1356 		rte_free(l2_tn_filter);
1357 	}
1358 
1359 	return 0;
1360 }
1361 
1362 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev)
1363 {
1364 	struct ixgbe_hw_fdir_info *fdir_info =
1365 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1366 	char fdir_hash_name[RTE_HASH_NAMESIZE];
1367 	struct rte_hash_parameters fdir_hash_params = {
1368 		.name = fdir_hash_name,
1369 		.entries = IXGBE_MAX_FDIR_FILTER_NUM,
1370 		.key_len = sizeof(union ixgbe_atr_input),
1371 		.hash_func = rte_hash_crc,
1372 		.hash_func_init_val = 0,
1373 		.socket_id = rte_socket_id(),
1374 	};
1375 
1376 	TAILQ_INIT(&fdir_info->fdir_list);
1377 	snprintf(fdir_hash_name, RTE_HASH_NAMESIZE,
1378 		 "fdir_%s", eth_dev->device->name);
1379 	fdir_info->hash_handle = rte_hash_create(&fdir_hash_params);
1380 	if (!fdir_info->hash_handle) {
1381 		PMD_INIT_LOG(ERR, "Failed to create fdir hash table!");
1382 		return -EINVAL;
1383 	}
1384 	fdir_info->hash_map = rte_zmalloc("ixgbe",
1385 					  sizeof(struct ixgbe_fdir_filter *) *
1386 					  IXGBE_MAX_FDIR_FILTER_NUM,
1387 					  0);
1388 	if (!fdir_info->hash_map) {
1389 		PMD_INIT_LOG(ERR,
1390 			     "Failed to allocate memory for fdir hash map!");
1391 		rte_hash_free(fdir_info->hash_handle);
1392 		return -ENOMEM;
1393 	}
1394 	fdir_info->mask_added = FALSE;
1395 
1396 	return 0;
1397 }
1398 
1399 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev)
1400 {
1401 	struct ixgbe_l2_tn_info *l2_tn_info =
1402 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1403 	char l2_tn_hash_name[RTE_HASH_NAMESIZE];
1404 	struct rte_hash_parameters l2_tn_hash_params = {
1405 		.name = l2_tn_hash_name,
1406 		.entries = IXGBE_MAX_L2_TN_FILTER_NUM,
1407 		.key_len = sizeof(struct ixgbe_l2_tn_key),
1408 		.hash_func = rte_hash_crc,
1409 		.hash_func_init_val = 0,
1410 		.socket_id = rte_socket_id(),
1411 	};
1412 
1413 	TAILQ_INIT(&l2_tn_info->l2_tn_list);
1414 	snprintf(l2_tn_hash_name, RTE_HASH_NAMESIZE,
1415 		 "l2_tn_%s", eth_dev->device->name);
1416 	l2_tn_info->hash_handle = rte_hash_create(&l2_tn_hash_params);
1417 	if (!l2_tn_info->hash_handle) {
1418 		PMD_INIT_LOG(ERR, "Failed to create L2 TN hash table!");
1419 		return -EINVAL;
1420 	}
1421 	l2_tn_info->hash_map = rte_zmalloc("ixgbe",
1422 				   sizeof(struct ixgbe_l2_tn_filter *) *
1423 				   IXGBE_MAX_L2_TN_FILTER_NUM,
1424 				   0);
1425 	if (!l2_tn_info->hash_map) {
1426 		PMD_INIT_LOG(ERR,
1427 			"Failed to allocate memory for L2 TN hash map!");
1428 		rte_hash_free(l2_tn_info->hash_handle);
1429 		return -ENOMEM;
1430 	}
1431 	l2_tn_info->e_tag_en = FALSE;
1432 	l2_tn_info->e_tag_fwd_en = FALSE;
1433 	l2_tn_info->e_tag_ether_type = RTE_ETHER_TYPE_ETAG;
1434 
1435 	return 0;
1436 }
1437 /*
1438  * Negotiate mailbox API version with the PF.
1439  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1440  * Then we try to negotiate starting with the most recent one.
1441  * If all negotiation attempts fail, then we will proceed with
1442  * the default one (ixgbe_mbox_api_10).
1443  */
1444 static void
1445 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1446 {
1447 	int32_t i;
1448 
1449 	/* start with highest supported, proceed down */
1450 	static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1451 		ixgbe_mbox_api_13,
1452 		ixgbe_mbox_api_12,
1453 		ixgbe_mbox_api_11,
1454 		ixgbe_mbox_api_10,
1455 	};
1456 
1457 	for (i = 0;
1458 			i != RTE_DIM(sup_ver) &&
1459 			ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1460 			i++)
1461 		;
1462 }
1463 
1464 static void
1465 generate_random_mac_addr(struct rte_ether_addr *mac_addr)
1466 {
1467 	uint64_t random;
1468 
1469 	/* Set Organizationally Unique Identifier (OUI) prefix. */
1470 	mac_addr->addr_bytes[0] = 0x00;
1471 	mac_addr->addr_bytes[1] = 0x09;
1472 	mac_addr->addr_bytes[2] = 0xC0;
1473 	/* Force indication of locally assigned MAC address. */
1474 	mac_addr->addr_bytes[0] |= RTE_ETHER_LOCAL_ADMIN_ADDR;
1475 	/* Generate the last 3 bytes of the MAC address with a random number. */
1476 	random = rte_rand();
1477 	memcpy(&mac_addr->addr_bytes[3], &random, 3);
1478 }
1479 
1480 static int
1481 devarg_handle_int(__rte_unused const char *key, const char *value,
1482 		  void *extra_args)
1483 {
1484 	uint16_t *n = extra_args;
1485 
1486 	if (value == NULL || extra_args == NULL)
1487 		return -EINVAL;
1488 
1489 	*n = (uint16_t)strtoul(value, NULL, 0);
1490 	if (*n == USHRT_MAX && errno == ERANGE)
1491 		return -1;
1492 
1493 	return 0;
1494 }
1495 
1496 static void
1497 ixgbevf_parse_devargs(struct ixgbe_adapter *adapter,
1498 		      struct rte_devargs *devargs)
1499 {
1500 	struct rte_kvargs *kvlist;
1501 	uint16_t pflink_fullchk;
1502 
1503 	if (devargs == NULL)
1504 		return;
1505 
1506 	kvlist = rte_kvargs_parse(devargs->args, ixgbevf_valid_arguments);
1507 	if (kvlist == NULL)
1508 		return;
1509 
1510 	if (rte_kvargs_count(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK) == 1 &&
1511 	    rte_kvargs_process(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK,
1512 			       devarg_handle_int, &pflink_fullchk) == 0 &&
1513 	    pflink_fullchk == 1)
1514 		adapter->pflink_fullchk = 1;
1515 
1516 	rte_kvargs_free(kvlist);
1517 }
1518 
1519 /*
1520  * Virtual Function device init
1521  */
1522 static int
1523 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1524 {
1525 	int diag;
1526 	uint32_t tc, tcs;
1527 	struct ixgbe_adapter *ad = eth_dev->data->dev_private;
1528 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1529 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1530 	struct ixgbe_hw *hw =
1531 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1532 	struct ixgbe_vfta *shadow_vfta =
1533 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1534 	struct ixgbe_hwstrip *hwstrip =
1535 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1536 	struct rte_ether_addr *perm_addr =
1537 		(struct rte_ether_addr *)hw->mac.perm_addr;
1538 
1539 	PMD_INIT_FUNC_TRACE();
1540 
1541 	eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1542 	eth_dev->rx_descriptor_done   = ixgbe_dev_rx_descriptor_done;
1543 	eth_dev->rx_descriptor_status = ixgbe_dev_rx_descriptor_status;
1544 	eth_dev->tx_descriptor_status = ixgbe_dev_tx_descriptor_status;
1545 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1546 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1547 
1548 	/* for secondary processes, we don't initialise any further as primary
1549 	 * has already done this work. Only check we don't need a different
1550 	 * RX function
1551 	 */
1552 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1553 		struct ixgbe_tx_queue *txq;
1554 		/* TX queue function in primary, set by last queue initialized
1555 		 * Tx queue may not initialized by primary process
1556 		 */
1557 		if (eth_dev->data->tx_queues) {
1558 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1559 			ixgbe_set_tx_function(eth_dev, txq);
1560 		} else {
1561 			/* Use default TX function if we get here */
1562 			PMD_INIT_LOG(NOTICE,
1563 				     "No TX queues configured yet. Using default TX function.");
1564 		}
1565 
1566 		ixgbe_set_rx_function(eth_dev);
1567 
1568 		return 0;
1569 	}
1570 
1571 	rte_atomic32_clear(&ad->link_thread_running);
1572 	ixgbevf_parse_devargs(eth_dev->data->dev_private,
1573 			      pci_dev->device.devargs);
1574 
1575 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1576 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1577 
1578 	hw->device_id = pci_dev->id.device_id;
1579 	hw->vendor_id = pci_dev->id.vendor_id;
1580 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1581 
1582 	/* initialize the vfta */
1583 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1584 
1585 	/* initialize the hw strip bitmap*/
1586 	memset(hwstrip, 0, sizeof(*hwstrip));
1587 
1588 	/* Initialize the shared code (base driver) */
1589 	diag = ixgbe_init_shared_code(hw);
1590 	if (diag != IXGBE_SUCCESS) {
1591 		PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1592 		return -EIO;
1593 	}
1594 
1595 	/* init_mailbox_params */
1596 	hw->mbx.ops.init_params(hw);
1597 
1598 	/* Reset the hw statistics */
1599 	ixgbevf_dev_stats_reset(eth_dev);
1600 
1601 	/* Disable the interrupts for VF */
1602 	ixgbevf_intr_disable(eth_dev);
1603 
1604 	hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1605 	diag = hw->mac.ops.reset_hw(hw);
1606 
1607 	/*
1608 	 * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1609 	 * the underlying PF driver has not assigned a MAC address to the VF.
1610 	 * In this case, assign a random MAC address.
1611 	 */
1612 	if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1613 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1614 		/*
1615 		 * This error code will be propagated to the app by
1616 		 * rte_eth_dev_reset, so use a public error code rather than
1617 		 * the internal-only IXGBE_ERR_RESET_FAILED
1618 		 */
1619 		return -EAGAIN;
1620 	}
1621 
1622 	/* negotiate mailbox API version to use with the PF. */
1623 	ixgbevf_negotiate_api(hw);
1624 
1625 	/* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1626 	ixgbevf_get_queues(hw, &tcs, &tc);
1627 
1628 	/* Allocate memory for storing MAC addresses */
1629 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", RTE_ETHER_ADDR_LEN *
1630 					       hw->mac.num_rar_entries, 0);
1631 	if (eth_dev->data->mac_addrs == NULL) {
1632 		PMD_INIT_LOG(ERR,
1633 			     "Failed to allocate %u bytes needed to store "
1634 			     "MAC addresses",
1635 			     RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1636 		return -ENOMEM;
1637 	}
1638 
1639 	/* Generate a random MAC address, if none was assigned by PF. */
1640 	if (rte_is_zero_ether_addr(perm_addr)) {
1641 		generate_random_mac_addr(perm_addr);
1642 		diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1643 		if (diag) {
1644 			rte_free(eth_dev->data->mac_addrs);
1645 			eth_dev->data->mac_addrs = NULL;
1646 			return diag;
1647 		}
1648 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1649 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1650 			     RTE_ETHER_ADDR_PRT_FMT,
1651 				 RTE_ETHER_ADDR_BYTES(perm_addr));
1652 	}
1653 
1654 	/* Copy the permanent MAC address */
1655 	rte_ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1656 
1657 	/* reset the hardware with the new settings */
1658 	diag = hw->mac.ops.start_hw(hw);
1659 	switch (diag) {
1660 	case  0:
1661 		break;
1662 
1663 	default:
1664 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1665 		rte_free(eth_dev->data->mac_addrs);
1666 		eth_dev->data->mac_addrs = NULL;
1667 		return -EIO;
1668 	}
1669 
1670 	rte_intr_callback_register(intr_handle,
1671 				   ixgbevf_dev_interrupt_handler, eth_dev);
1672 	rte_intr_enable(intr_handle);
1673 	ixgbevf_intr_enable(eth_dev);
1674 
1675 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1676 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1677 		     pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1678 
1679 	return 0;
1680 }
1681 
1682 /* Virtual Function device uninit */
1683 
1684 static int
1685 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1686 {
1687 	PMD_INIT_FUNC_TRACE();
1688 
1689 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1690 		return 0;
1691 
1692 	ixgbevf_dev_close(eth_dev);
1693 
1694 	return 0;
1695 }
1696 
1697 static int
1698 eth_ixgbe_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1699 		struct rte_pci_device *pci_dev)
1700 {
1701 	char name[RTE_ETH_NAME_MAX_LEN];
1702 	struct rte_eth_dev *pf_ethdev;
1703 	struct rte_eth_devargs eth_da;
1704 	int i, retval;
1705 
1706 	if (pci_dev->device.devargs) {
1707 		retval = rte_eth_devargs_parse(pci_dev->device.devargs->args,
1708 				&eth_da);
1709 		if (retval)
1710 			return retval;
1711 	} else
1712 		memset(&eth_da, 0, sizeof(eth_da));
1713 
1714 	if (eth_da.nb_representor_ports > 0 &&
1715 	    eth_da.type != RTE_ETH_REPRESENTOR_VF) {
1716 		PMD_DRV_LOG(ERR, "unsupported representor type: %s\n",
1717 			    pci_dev->device.devargs->args);
1718 		return -ENOTSUP;
1719 	}
1720 
1721 	retval = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
1722 		sizeof(struct ixgbe_adapter),
1723 		eth_dev_pci_specific_init, pci_dev,
1724 		eth_ixgbe_dev_init, NULL);
1725 
1726 	if (retval || eth_da.nb_representor_ports < 1)
1727 		return retval;
1728 
1729 	pf_ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1730 	if (pf_ethdev == NULL)
1731 		return -ENODEV;
1732 
1733 	/* probe VF representor ports */
1734 	for (i = 0; i < eth_da.nb_representor_ports; i++) {
1735 		struct ixgbe_vf_info *vfinfo;
1736 		struct ixgbe_vf_representor representor;
1737 
1738 		vfinfo = *IXGBE_DEV_PRIVATE_TO_P_VFDATA(
1739 			pf_ethdev->data->dev_private);
1740 		if (vfinfo == NULL) {
1741 			PMD_DRV_LOG(ERR,
1742 				"no virtual functions supported by PF");
1743 			break;
1744 		}
1745 
1746 		representor.vf_id = eth_da.representor_ports[i];
1747 		representor.switch_domain_id = vfinfo->switch_domain_id;
1748 		representor.pf_ethdev = pf_ethdev;
1749 
1750 		/* representor port net_bdf_port */
1751 		snprintf(name, sizeof(name), "net_%s_representor_%d",
1752 			pci_dev->device.name,
1753 			eth_da.representor_ports[i]);
1754 
1755 		retval = rte_eth_dev_create(&pci_dev->device, name,
1756 			sizeof(struct ixgbe_vf_representor), NULL, NULL,
1757 			ixgbe_vf_representor_init, &representor);
1758 
1759 		if (retval)
1760 			PMD_DRV_LOG(ERR, "failed to create ixgbe vf "
1761 				"representor %s.", name);
1762 	}
1763 
1764 	return 0;
1765 }
1766 
1767 static int eth_ixgbe_pci_remove(struct rte_pci_device *pci_dev)
1768 {
1769 	struct rte_eth_dev *ethdev;
1770 
1771 	ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1772 	if (!ethdev)
1773 		return 0;
1774 
1775 	if (ethdev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
1776 		return rte_eth_dev_pci_generic_remove(pci_dev,
1777 					ixgbe_vf_representor_uninit);
1778 	else
1779 		return rte_eth_dev_pci_generic_remove(pci_dev,
1780 						eth_ixgbe_dev_uninit);
1781 }
1782 
1783 static struct rte_pci_driver rte_ixgbe_pmd = {
1784 	.id_table = pci_id_ixgbe_map,
1785 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1786 	.probe = eth_ixgbe_pci_probe,
1787 	.remove = eth_ixgbe_pci_remove,
1788 };
1789 
1790 static int eth_ixgbevf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1791 	struct rte_pci_device *pci_dev)
1792 {
1793 	return rte_eth_dev_pci_generic_probe(pci_dev,
1794 		sizeof(struct ixgbe_adapter), eth_ixgbevf_dev_init);
1795 }
1796 
1797 static int eth_ixgbevf_pci_remove(struct rte_pci_device *pci_dev)
1798 {
1799 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbevf_dev_uninit);
1800 }
1801 
1802 /*
1803  * virtual function driver struct
1804  */
1805 static struct rte_pci_driver rte_ixgbevf_pmd = {
1806 	.id_table = pci_id_ixgbevf_map,
1807 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1808 	.probe = eth_ixgbevf_pci_probe,
1809 	.remove = eth_ixgbevf_pci_remove,
1810 };
1811 
1812 static int
1813 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1814 {
1815 	struct ixgbe_hw *hw =
1816 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1817 	struct ixgbe_vfta *shadow_vfta =
1818 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1819 	uint32_t vfta;
1820 	uint32_t vid_idx;
1821 	uint32_t vid_bit;
1822 
1823 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1824 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1825 	vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1826 	if (on)
1827 		vfta |= vid_bit;
1828 	else
1829 		vfta &= ~vid_bit;
1830 	IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1831 
1832 	/* update local VFTA copy */
1833 	shadow_vfta->vfta[vid_idx] = vfta;
1834 
1835 	return 0;
1836 }
1837 
1838 static void
1839 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1840 {
1841 	if (on)
1842 		ixgbe_vlan_hw_strip_enable(dev, queue);
1843 	else
1844 		ixgbe_vlan_hw_strip_disable(dev, queue);
1845 }
1846 
1847 static int
1848 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1849 		    enum rte_vlan_type vlan_type,
1850 		    uint16_t tpid)
1851 {
1852 	struct ixgbe_hw *hw =
1853 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1854 	int ret = 0;
1855 	uint32_t reg;
1856 	uint32_t qinq;
1857 
1858 	qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1859 	qinq &= IXGBE_DMATXCTL_GDV;
1860 
1861 	switch (vlan_type) {
1862 	case ETH_VLAN_TYPE_INNER:
1863 		if (qinq) {
1864 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1865 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1866 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1867 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1868 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1869 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1870 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1871 		} else {
1872 			ret = -ENOTSUP;
1873 			PMD_DRV_LOG(ERR, "Inner type is not supported"
1874 				    " by single VLAN");
1875 		}
1876 		break;
1877 	case ETH_VLAN_TYPE_OUTER:
1878 		if (qinq) {
1879 			/* Only the high 16-bits is valid */
1880 			IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1881 					IXGBE_EXVET_VET_EXT_SHIFT);
1882 		} else {
1883 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1884 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1885 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1886 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1887 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1888 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1889 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1890 		}
1891 
1892 		break;
1893 	default:
1894 		ret = -EINVAL;
1895 		PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1896 		break;
1897 	}
1898 
1899 	return ret;
1900 }
1901 
1902 void
1903 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1904 {
1905 	struct ixgbe_hw *hw =
1906 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1907 	uint32_t vlnctrl;
1908 
1909 	PMD_INIT_FUNC_TRACE();
1910 
1911 	/* Filter Table Disable */
1912 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1913 	vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1914 
1915 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1916 }
1917 
1918 void
1919 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1920 {
1921 	struct ixgbe_hw *hw =
1922 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1923 	struct ixgbe_vfta *shadow_vfta =
1924 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1925 	uint32_t vlnctrl;
1926 	uint16_t i;
1927 
1928 	PMD_INIT_FUNC_TRACE();
1929 
1930 	/* Filter Table Enable */
1931 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1932 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1933 	vlnctrl |= IXGBE_VLNCTRL_VFE;
1934 
1935 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1936 
1937 	/* write whatever is in local vfta copy */
1938 	for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1939 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1940 }
1941 
1942 static void
1943 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1944 {
1945 	struct ixgbe_hwstrip *hwstrip =
1946 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1947 	struct ixgbe_rx_queue *rxq;
1948 
1949 	if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
1950 		return;
1951 
1952 	if (on)
1953 		IXGBE_SET_HWSTRIP(hwstrip, queue);
1954 	else
1955 		IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1956 
1957 	if (queue >= dev->data->nb_rx_queues)
1958 		return;
1959 
1960 	rxq = dev->data->rx_queues[queue];
1961 
1962 	if (on) {
1963 		rxq->vlan_flags = PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
1964 		rxq->offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
1965 	} else {
1966 		rxq->vlan_flags = PKT_RX_VLAN;
1967 		rxq->offloads &= ~DEV_RX_OFFLOAD_VLAN_STRIP;
1968 	}
1969 }
1970 
1971 static void
1972 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
1973 {
1974 	struct ixgbe_hw *hw =
1975 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1976 	uint32_t ctrl;
1977 
1978 	PMD_INIT_FUNC_TRACE();
1979 
1980 	if (hw->mac.type == ixgbe_mac_82598EB) {
1981 		/* No queue level support */
1982 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1983 		return;
1984 	}
1985 
1986 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1987 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1988 	ctrl &= ~IXGBE_RXDCTL_VME;
1989 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1990 
1991 	/* record those setting for HW strip per queue */
1992 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
1993 }
1994 
1995 static void
1996 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
1997 {
1998 	struct ixgbe_hw *hw =
1999 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2000 	uint32_t ctrl;
2001 
2002 	PMD_INIT_FUNC_TRACE();
2003 
2004 	if (hw->mac.type == ixgbe_mac_82598EB) {
2005 		/* No queue level supported */
2006 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2007 		return;
2008 	}
2009 
2010 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2011 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2012 	ctrl |= IXGBE_RXDCTL_VME;
2013 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2014 
2015 	/* record those setting for HW strip per queue */
2016 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
2017 }
2018 
2019 static void
2020 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2021 {
2022 	struct ixgbe_hw *hw =
2023 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2024 	uint32_t ctrl;
2025 
2026 	PMD_INIT_FUNC_TRACE();
2027 
2028 	/* DMATXCTRL: Geric Double VLAN Disable */
2029 	ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2030 	ctrl &= ~IXGBE_DMATXCTL_GDV;
2031 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2032 
2033 	/* CTRL_EXT: Global Double VLAN Disable */
2034 	ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2035 	ctrl &= ~IXGBE_EXTENDED_VLAN;
2036 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2037 
2038 }
2039 
2040 static void
2041 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2042 {
2043 	struct ixgbe_hw *hw =
2044 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2045 	uint32_t ctrl;
2046 
2047 	PMD_INIT_FUNC_TRACE();
2048 
2049 	/* DMATXCTRL: Geric Double VLAN Enable */
2050 	ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2051 	ctrl |= IXGBE_DMATXCTL_GDV;
2052 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2053 
2054 	/* CTRL_EXT: Global Double VLAN Enable */
2055 	ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2056 	ctrl |= IXGBE_EXTENDED_VLAN;
2057 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2058 
2059 	/* Clear pooling mode of PFVTCTL. It's required by X550. */
2060 	if (hw->mac.type == ixgbe_mac_X550 ||
2061 	    hw->mac.type == ixgbe_mac_X550EM_x ||
2062 	    hw->mac.type == ixgbe_mac_X550EM_a) {
2063 		ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
2064 		ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
2065 		IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
2066 	}
2067 
2068 	/*
2069 	 * VET EXT field in the EXVET register = 0x8100 by default
2070 	 * So no need to change. Same to VT field of DMATXCTL register
2071 	 */
2072 }
2073 
2074 void
2075 ixgbe_vlan_hw_strip_config(struct rte_eth_dev *dev)
2076 {
2077 	struct ixgbe_hw *hw =
2078 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2079 	struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
2080 	uint32_t ctrl;
2081 	uint16_t i;
2082 	struct ixgbe_rx_queue *rxq;
2083 	bool on;
2084 
2085 	PMD_INIT_FUNC_TRACE();
2086 
2087 	if (hw->mac.type == ixgbe_mac_82598EB) {
2088 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP) {
2089 			ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2090 			ctrl |= IXGBE_VLNCTRL_VME;
2091 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2092 		} else {
2093 			ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2094 			ctrl &= ~IXGBE_VLNCTRL_VME;
2095 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2096 		}
2097 	} else {
2098 		/*
2099 		 * Other 10G NIC, the VLAN strip can be setup
2100 		 * per queue in RXDCTL
2101 		 */
2102 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
2103 			rxq = dev->data->rx_queues[i];
2104 			ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2105 			if (rxq->offloads & DEV_RX_OFFLOAD_VLAN_STRIP) {
2106 				ctrl |= IXGBE_RXDCTL_VME;
2107 				on = TRUE;
2108 			} else {
2109 				ctrl &= ~IXGBE_RXDCTL_VME;
2110 				on = FALSE;
2111 			}
2112 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2113 
2114 			/* record those setting for HW strip per queue */
2115 			ixgbe_vlan_hw_strip_bitmap_set(dev, i, on);
2116 		}
2117 	}
2118 }
2119 
2120 static void
2121 ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev, int mask)
2122 {
2123 	uint16_t i;
2124 	struct rte_eth_rxmode *rxmode;
2125 	struct ixgbe_rx_queue *rxq;
2126 
2127 	if (mask & ETH_VLAN_STRIP_MASK) {
2128 		rxmode = &dev->data->dev_conf.rxmode;
2129 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2130 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
2131 				rxq = dev->data->rx_queues[i];
2132 				rxq->offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
2133 			}
2134 		else
2135 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
2136 				rxq = dev->data->rx_queues[i];
2137 				rxq->offloads &= ~DEV_RX_OFFLOAD_VLAN_STRIP;
2138 			}
2139 	}
2140 }
2141 
2142 static int
2143 ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask)
2144 {
2145 	struct rte_eth_rxmode *rxmode;
2146 	rxmode = &dev->data->dev_conf.rxmode;
2147 
2148 	if (mask & ETH_VLAN_STRIP_MASK) {
2149 		ixgbe_vlan_hw_strip_config(dev);
2150 	}
2151 
2152 	if (mask & ETH_VLAN_FILTER_MASK) {
2153 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2154 			ixgbe_vlan_hw_filter_enable(dev);
2155 		else
2156 			ixgbe_vlan_hw_filter_disable(dev);
2157 	}
2158 
2159 	if (mask & ETH_VLAN_EXTEND_MASK) {
2160 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2161 			ixgbe_vlan_hw_extend_enable(dev);
2162 		else
2163 			ixgbe_vlan_hw_extend_disable(dev);
2164 	}
2165 
2166 	return 0;
2167 }
2168 
2169 static int
2170 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2171 {
2172 	ixgbe_config_vlan_strip_on_all_queues(dev, mask);
2173 
2174 	ixgbe_vlan_offload_config(dev, mask);
2175 
2176 	return 0;
2177 }
2178 
2179 static void
2180 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2181 {
2182 	struct ixgbe_hw *hw =
2183 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2184 	/* VLNCTRL: enable vlan filtering and allow all vlan tags through */
2185 	uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2186 
2187 	vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
2188 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
2189 }
2190 
2191 static int
2192 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
2193 {
2194 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2195 
2196 	switch (nb_rx_q) {
2197 	case 1:
2198 	case 2:
2199 		RTE_ETH_DEV_SRIOV(dev).active = ETH_64_POOLS;
2200 		break;
2201 	case 4:
2202 		RTE_ETH_DEV_SRIOV(dev).active = ETH_32_POOLS;
2203 		break;
2204 	default:
2205 		return -EINVAL;
2206 	}
2207 
2208 	RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
2209 		IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2210 	RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
2211 		pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2212 	return 0;
2213 }
2214 
2215 static int
2216 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
2217 {
2218 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
2219 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2220 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
2221 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
2222 
2223 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
2224 		/* check multi-queue mode */
2225 		switch (dev_conf->rxmode.mq_mode) {
2226 		case ETH_MQ_RX_VMDQ_DCB:
2227 			PMD_INIT_LOG(INFO, "ETH_MQ_RX_VMDQ_DCB mode supported in SRIOV");
2228 			break;
2229 		case ETH_MQ_RX_VMDQ_DCB_RSS:
2230 			/* DCB/RSS VMDQ in SRIOV mode, not implement yet */
2231 			PMD_INIT_LOG(ERR, "SRIOV active,"
2232 					" unsupported mq_mode rx %d.",
2233 					dev_conf->rxmode.mq_mode);
2234 			return -EINVAL;
2235 		case ETH_MQ_RX_RSS:
2236 		case ETH_MQ_RX_VMDQ_RSS:
2237 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_RSS;
2238 			if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
2239 				if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
2240 					PMD_INIT_LOG(ERR, "SRIOV is active,"
2241 						" invalid queue number"
2242 						" for VMDQ RSS, allowed"
2243 						" value are 1, 2 or 4.");
2244 					return -EINVAL;
2245 				}
2246 			break;
2247 		case ETH_MQ_RX_VMDQ_ONLY:
2248 		case ETH_MQ_RX_NONE:
2249 			/* if nothing mq mode configure, use default scheme */
2250 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
2251 			break;
2252 		default: /* ETH_MQ_RX_DCB, ETH_MQ_RX_DCB_RSS or ETH_MQ_TX_DCB*/
2253 			/* SRIOV only works in VMDq enable mode */
2254 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2255 					" wrong mq_mode rx %d.",
2256 					dev_conf->rxmode.mq_mode);
2257 			return -EINVAL;
2258 		}
2259 
2260 		switch (dev_conf->txmode.mq_mode) {
2261 		case ETH_MQ_TX_VMDQ_DCB:
2262 			PMD_INIT_LOG(INFO, "ETH_MQ_TX_VMDQ_DCB mode supported in SRIOV");
2263 			dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_DCB;
2264 			break;
2265 		default: /* ETH_MQ_TX_VMDQ_ONLY or ETH_MQ_TX_NONE */
2266 			dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_ONLY;
2267 			break;
2268 		}
2269 
2270 		/* check valid queue number */
2271 		if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2272 		    (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2273 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2274 					" nb_rx_q=%d nb_tx_q=%d queue number"
2275 					" must be less than or equal to %d.",
2276 					nb_rx_q, nb_tx_q,
2277 					RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2278 			return -EINVAL;
2279 		}
2280 	} else {
2281 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB_RSS) {
2282 			PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2283 					  " not supported.");
2284 			return -EINVAL;
2285 		}
2286 		/* check configuration for vmdb+dcb mode */
2287 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB) {
2288 			const struct rte_eth_vmdq_dcb_conf *conf;
2289 
2290 			if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2291 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2292 						IXGBE_VMDQ_DCB_NB_QUEUES);
2293 				return -EINVAL;
2294 			}
2295 			conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2296 			if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2297 			       conf->nb_queue_pools == ETH_32_POOLS)) {
2298 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2299 						" nb_queue_pools must be %d or %d.",
2300 						ETH_16_POOLS, ETH_32_POOLS);
2301 				return -EINVAL;
2302 			}
2303 		}
2304 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_VMDQ_DCB) {
2305 			const struct rte_eth_vmdq_dcb_tx_conf *conf;
2306 
2307 			if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2308 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2309 						 IXGBE_VMDQ_DCB_NB_QUEUES);
2310 				return -EINVAL;
2311 			}
2312 			conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2313 			if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2314 			       conf->nb_queue_pools == ETH_32_POOLS)) {
2315 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2316 						" nb_queue_pools != %d and"
2317 						" nb_queue_pools != %d.",
2318 						ETH_16_POOLS, ETH_32_POOLS);
2319 				return -EINVAL;
2320 			}
2321 		}
2322 
2323 		/* For DCB mode check our configuration before we go further */
2324 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB) {
2325 			const struct rte_eth_dcb_rx_conf *conf;
2326 
2327 			conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2328 			if (!(conf->nb_tcs == ETH_4_TCS ||
2329 			       conf->nb_tcs == ETH_8_TCS)) {
2330 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2331 						" and nb_tcs != %d.",
2332 						ETH_4_TCS, ETH_8_TCS);
2333 				return -EINVAL;
2334 			}
2335 		}
2336 
2337 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_DCB) {
2338 			const struct rte_eth_dcb_tx_conf *conf;
2339 
2340 			conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2341 			if (!(conf->nb_tcs == ETH_4_TCS ||
2342 			       conf->nb_tcs == ETH_8_TCS)) {
2343 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2344 						" and nb_tcs != %d.",
2345 						ETH_4_TCS, ETH_8_TCS);
2346 				return -EINVAL;
2347 			}
2348 		}
2349 
2350 		/*
2351 		 * When DCB/VT is off, maximum number of queues changes,
2352 		 * except for 82598EB, which remains constant.
2353 		 */
2354 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
2355 				hw->mac.type != ixgbe_mac_82598EB) {
2356 			if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2357 				PMD_INIT_LOG(ERR,
2358 					     "Neither VT nor DCB are enabled, "
2359 					     "nb_tx_q > %d.",
2360 					     IXGBE_NONE_MODE_TX_NB_QUEUES);
2361 				return -EINVAL;
2362 			}
2363 		}
2364 	}
2365 	return 0;
2366 }
2367 
2368 static int
2369 ixgbe_dev_configure(struct rte_eth_dev *dev)
2370 {
2371 	struct ixgbe_interrupt *intr =
2372 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2373 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2374 	int ret;
2375 
2376 	PMD_INIT_FUNC_TRACE();
2377 
2378 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
2379 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
2380 
2381 	/* multipe queue mode checking */
2382 	ret  = ixgbe_check_mq_mode(dev);
2383 	if (ret != 0) {
2384 		PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2385 			    ret);
2386 		return ret;
2387 	}
2388 
2389 	/* set flag to update link status after init */
2390 	intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2391 
2392 	/*
2393 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2394 	 * allocation or vector Rx preconditions we will reset it.
2395 	 */
2396 	adapter->rx_bulk_alloc_allowed = true;
2397 	adapter->rx_vec_allowed = true;
2398 
2399 	return 0;
2400 }
2401 
2402 static void
2403 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2404 {
2405 	struct ixgbe_hw *hw =
2406 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2407 	struct ixgbe_interrupt *intr =
2408 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2409 	uint32_t gpie;
2410 
2411 	/* only set up it on X550EM_X */
2412 	if (hw->mac.type == ixgbe_mac_X550EM_x) {
2413 		gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2414 		gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2415 		IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2416 		if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2417 			intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2418 	}
2419 }
2420 
2421 int
2422 ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
2423 			uint16_t tx_rate, uint64_t q_msk)
2424 {
2425 	struct ixgbe_hw *hw;
2426 	struct ixgbe_vf_info *vfinfo;
2427 	struct rte_eth_link link;
2428 	uint8_t  nb_q_per_pool;
2429 	uint32_t queue_stride;
2430 	uint32_t queue_idx, idx = 0, vf_idx;
2431 	uint32_t queue_end;
2432 	uint16_t total_rate = 0;
2433 	struct rte_pci_device *pci_dev;
2434 	int ret;
2435 
2436 	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2437 	ret = rte_eth_link_get_nowait(dev->data->port_id, &link);
2438 	if (ret < 0)
2439 		return ret;
2440 
2441 	if (vf >= pci_dev->max_vfs)
2442 		return -EINVAL;
2443 
2444 	if (tx_rate > link.link_speed)
2445 		return -EINVAL;
2446 
2447 	if (q_msk == 0)
2448 		return 0;
2449 
2450 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2451 	vfinfo = *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
2452 	nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2453 	queue_stride = IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2454 	queue_idx = vf * queue_stride;
2455 	queue_end = queue_idx + nb_q_per_pool - 1;
2456 	if (queue_end >= hw->mac.max_tx_queues)
2457 		return -EINVAL;
2458 
2459 	if (vfinfo) {
2460 		for (vf_idx = 0; vf_idx < pci_dev->max_vfs; vf_idx++) {
2461 			if (vf_idx == vf)
2462 				continue;
2463 			for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
2464 				idx++)
2465 				total_rate += vfinfo[vf_idx].tx_rate[idx];
2466 		}
2467 	} else {
2468 		return -EINVAL;
2469 	}
2470 
2471 	/* Store tx_rate for this vf. */
2472 	for (idx = 0; idx < nb_q_per_pool; idx++) {
2473 		if (((uint64_t)0x1 << idx) & q_msk) {
2474 			if (vfinfo[vf].tx_rate[idx] != tx_rate)
2475 				vfinfo[vf].tx_rate[idx] = tx_rate;
2476 			total_rate += tx_rate;
2477 		}
2478 	}
2479 
2480 	if (total_rate > dev->data->dev_link.link_speed) {
2481 		/* Reset stored TX rate of the VF if it causes exceed
2482 		 * link speed.
2483 		 */
2484 		memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
2485 		return -EINVAL;
2486 	}
2487 
2488 	/* Set RTTBCNRC of each queue/pool for vf X  */
2489 	for (; queue_idx <= queue_end; queue_idx++) {
2490 		if (0x1 & q_msk)
2491 			ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
2492 		q_msk = q_msk >> 1;
2493 	}
2494 
2495 	return 0;
2496 }
2497 
2498 static int
2499 ixgbe_flow_ctrl_enable(struct rte_eth_dev *dev, struct ixgbe_hw *hw)
2500 {
2501 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2502 	int err;
2503 	uint32_t mflcn;
2504 
2505 	ixgbe_setup_fc(hw);
2506 
2507 	err = ixgbe_fc_enable(hw);
2508 
2509 	/* Not negotiated is not an error case */
2510 	if (err == IXGBE_SUCCESS || err == IXGBE_ERR_FC_NOT_NEGOTIATED) {
2511 		/*
2512 		 *check if we want to forward MAC frames - driver doesn't
2513 		 *have native capability to do that,
2514 		 *so we'll write the registers ourselves
2515 		 */
2516 
2517 		mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
2518 
2519 		/* set or clear MFLCN.PMCF bit depending on configuration */
2520 		if (adapter->mac_ctrl_frame_fwd != 0)
2521 			mflcn |= IXGBE_MFLCN_PMCF;
2522 		else
2523 			mflcn &= ~IXGBE_MFLCN_PMCF;
2524 
2525 		IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
2526 		IXGBE_WRITE_FLUSH(hw);
2527 
2528 		return 0;
2529 	}
2530 	return err;
2531 }
2532 
2533 /*
2534  * Configure device link speed and setup link.
2535  * It returns 0 on success.
2536  */
2537 static int
2538 ixgbe_dev_start(struct rte_eth_dev *dev)
2539 {
2540 	struct ixgbe_hw *hw =
2541 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2542 	struct ixgbe_vf_info *vfinfo =
2543 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2544 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2545 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2546 	uint32_t intr_vector = 0;
2547 	int err;
2548 	bool link_up = false, negotiate = 0;
2549 	uint32_t speed = 0;
2550 	uint32_t allowed_speeds = 0;
2551 	int mask = 0;
2552 	int status;
2553 	uint16_t vf, idx;
2554 	uint32_t *link_speeds;
2555 	struct ixgbe_tm_conf *tm_conf =
2556 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2557 	struct ixgbe_macsec_setting *macsec_setting =
2558 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
2559 
2560 	PMD_INIT_FUNC_TRACE();
2561 
2562 	/* Stop the link setup handler before resetting the HW. */
2563 	ixgbe_dev_wait_setup_link_complete(dev, 0);
2564 
2565 	/* disable uio/vfio intr/eventfd mapping */
2566 	rte_intr_disable(intr_handle);
2567 
2568 	/* stop adapter */
2569 	hw->adapter_stopped = 0;
2570 	ixgbe_stop_adapter(hw);
2571 
2572 	/* reinitialize adapter
2573 	 * this calls reset and start
2574 	 */
2575 	status = ixgbe_pf_reset_hw(hw);
2576 	if (status != 0)
2577 		return -1;
2578 	hw->mac.ops.start_hw(hw);
2579 	hw->mac.get_link_status = true;
2580 
2581 	/* configure PF module if SRIOV enabled */
2582 	ixgbe_pf_host_configure(dev);
2583 
2584 	ixgbe_dev_phy_intr_setup(dev);
2585 
2586 	/* check and configure queue intr-vector mapping */
2587 	if ((rte_intr_cap_multiple(intr_handle) ||
2588 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
2589 	    dev->data->dev_conf.intr_conf.rxq != 0) {
2590 		intr_vector = dev->data->nb_rx_queues;
2591 		if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2592 			PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2593 					IXGBE_MAX_INTR_QUEUE_NUM);
2594 			return -ENOTSUP;
2595 		}
2596 		if (rte_intr_efd_enable(intr_handle, intr_vector))
2597 			return -1;
2598 	}
2599 
2600 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
2601 		intr_handle->intr_vec =
2602 			rte_zmalloc("intr_vec",
2603 				    dev->data->nb_rx_queues * sizeof(int), 0);
2604 		if (intr_handle->intr_vec == NULL) {
2605 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2606 				     " intr_vec", dev->data->nb_rx_queues);
2607 			return -ENOMEM;
2608 		}
2609 	}
2610 
2611 	/* confiugre msix for sleep until rx interrupt */
2612 	ixgbe_configure_msix(dev);
2613 
2614 	/* initialize transmission unit */
2615 	ixgbe_dev_tx_init(dev);
2616 
2617 	/* This can fail when allocating mbufs for descriptor rings */
2618 	err = ixgbe_dev_rx_init(dev);
2619 	if (err) {
2620 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2621 		goto error;
2622 	}
2623 
2624 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
2625 		ETH_VLAN_EXTEND_MASK;
2626 	err = ixgbe_vlan_offload_config(dev, mask);
2627 	if (err) {
2628 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload");
2629 		goto error;
2630 	}
2631 
2632 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
2633 		/* Enable vlan filtering for VMDq */
2634 		ixgbe_vmdq_vlan_hw_filter_enable(dev);
2635 	}
2636 
2637 	/* Configure DCB hw */
2638 	ixgbe_configure_dcb(dev);
2639 
2640 	if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2641 		err = ixgbe_fdir_configure(dev);
2642 		if (err)
2643 			goto error;
2644 	}
2645 
2646 	/* Restore vf rate limit */
2647 	if (vfinfo != NULL) {
2648 		for (vf = 0; vf < pci_dev->max_vfs; vf++)
2649 			for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2650 				if (vfinfo[vf].tx_rate[idx] != 0)
2651 					ixgbe_set_vf_rate_limit(
2652 						dev, vf,
2653 						vfinfo[vf].tx_rate[idx],
2654 						1 << idx);
2655 	}
2656 
2657 	ixgbe_restore_statistics_mapping(dev);
2658 
2659 	err = ixgbe_flow_ctrl_enable(dev, hw);
2660 	if (err < 0) {
2661 		PMD_INIT_LOG(ERR, "enable flow ctrl err");
2662 		goto error;
2663 	}
2664 
2665 	err = ixgbe_dev_rxtx_start(dev);
2666 	if (err < 0) {
2667 		PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2668 		goto error;
2669 	}
2670 
2671 	/* Skip link setup if loopback mode is enabled. */
2672 	if (dev->data->dev_conf.lpbk_mode != 0) {
2673 		err = ixgbe_check_supported_loopback_mode(dev);
2674 		if (err < 0) {
2675 			PMD_INIT_LOG(ERR, "Unsupported loopback mode");
2676 			goto error;
2677 		} else {
2678 			goto skip_link_setup;
2679 		}
2680 	}
2681 
2682 	if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2683 		err = hw->mac.ops.setup_sfp(hw);
2684 		if (err)
2685 			goto error;
2686 	}
2687 
2688 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2689 		/* Turn on the copper */
2690 		ixgbe_set_phy_power(hw, true);
2691 	} else {
2692 		/* Turn on the laser */
2693 		ixgbe_enable_tx_laser(hw);
2694 	}
2695 
2696 	err = ixgbe_check_link(hw, &speed, &link_up, 0);
2697 	if (err)
2698 		goto error;
2699 	dev->data->dev_link.link_status = link_up;
2700 
2701 	err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2702 	if (err)
2703 		goto error;
2704 
2705 	switch (hw->mac.type) {
2706 	case ixgbe_mac_X550:
2707 	case ixgbe_mac_X550EM_x:
2708 	case ixgbe_mac_X550EM_a:
2709 		allowed_speeds = ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2710 			ETH_LINK_SPEED_2_5G |  ETH_LINK_SPEED_5G |
2711 			ETH_LINK_SPEED_10G;
2712 		if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
2713 				hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
2714 			allowed_speeds = ETH_LINK_SPEED_10M |
2715 				ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G;
2716 		break;
2717 	default:
2718 		allowed_speeds = ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2719 			ETH_LINK_SPEED_10G;
2720 	}
2721 
2722 	link_speeds = &dev->data->dev_conf.link_speeds;
2723 
2724 	/* Ignore autoneg flag bit and check the validity of 
2725 	 * link_speed 
2726 	 */
2727 	if (((*link_speeds) >> 1) & ~(allowed_speeds >> 1)) {
2728 		PMD_INIT_LOG(ERR, "Invalid link setting");
2729 		goto error;
2730 	}
2731 
2732 	speed = 0x0;
2733 	if (*link_speeds == ETH_LINK_SPEED_AUTONEG) {
2734 		switch (hw->mac.type) {
2735 		case ixgbe_mac_82598EB:
2736 			speed = IXGBE_LINK_SPEED_82598_AUTONEG;
2737 			break;
2738 		case ixgbe_mac_82599EB:
2739 		case ixgbe_mac_X540:
2740 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2741 			break;
2742 		case ixgbe_mac_X550:
2743 		case ixgbe_mac_X550EM_x:
2744 		case ixgbe_mac_X550EM_a:
2745 			speed = IXGBE_LINK_SPEED_X550_AUTONEG;
2746 			break;
2747 		default:
2748 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2749 		}
2750 	} else {
2751 		if (*link_speeds & ETH_LINK_SPEED_10G)
2752 			speed |= IXGBE_LINK_SPEED_10GB_FULL;
2753 		if (*link_speeds & ETH_LINK_SPEED_5G)
2754 			speed |= IXGBE_LINK_SPEED_5GB_FULL;
2755 		if (*link_speeds & ETH_LINK_SPEED_2_5G)
2756 			speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
2757 		if (*link_speeds & ETH_LINK_SPEED_1G)
2758 			speed |= IXGBE_LINK_SPEED_1GB_FULL;
2759 		if (*link_speeds & ETH_LINK_SPEED_100M)
2760 			speed |= IXGBE_LINK_SPEED_100_FULL;
2761 		if (*link_speeds & ETH_LINK_SPEED_10M)
2762 			speed |= IXGBE_LINK_SPEED_10_FULL;
2763 	}
2764 
2765 	err = ixgbe_setup_link(hw, speed, link_up);
2766 	if (err)
2767 		goto error;
2768 
2769 skip_link_setup:
2770 
2771 	if (rte_intr_allow_others(intr_handle)) {
2772 		/* check if lsc interrupt is enabled */
2773 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2774 			ixgbe_dev_lsc_interrupt_setup(dev, TRUE);
2775 		else
2776 			ixgbe_dev_lsc_interrupt_setup(dev, FALSE);
2777 		ixgbe_dev_macsec_interrupt_setup(dev);
2778 	} else {
2779 		rte_intr_callback_unregister(intr_handle,
2780 					     ixgbe_dev_interrupt_handler, dev);
2781 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2782 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
2783 				     " no intr multiplex");
2784 	}
2785 
2786 	/* check if rxq interrupt is enabled */
2787 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2788 	    rte_intr_dp_is_en(intr_handle))
2789 		ixgbe_dev_rxq_interrupt_setup(dev);
2790 
2791 	/* enable uio/vfio intr/eventfd mapping */
2792 	rte_intr_enable(intr_handle);
2793 
2794 	/* resume enabled intr since hw reset */
2795 	ixgbe_enable_intr(dev);
2796 	ixgbe_l2_tunnel_conf(dev);
2797 	ixgbe_filter_restore(dev);
2798 
2799 	if (tm_conf->root && !tm_conf->committed)
2800 		PMD_DRV_LOG(WARNING,
2801 			    "please call hierarchy_commit() "
2802 			    "before starting the port");
2803 
2804 	/* wait for the controller to acquire link */
2805 	err = ixgbe_wait_for_link_up(hw);
2806 	if (err)
2807 		goto error;
2808 
2809 	/*
2810 	 * Update link status right before return, because it may
2811 	 * start link configuration process in a separate thread.
2812 	 */
2813 	ixgbe_dev_link_update(dev, 0);
2814 
2815 	/* setup the macsec setting register */
2816 	if (macsec_setting->offload_en)
2817 		ixgbe_dev_macsec_register_enable(dev, macsec_setting);
2818 
2819 	return 0;
2820 
2821 error:
2822 	PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2823 	ixgbe_dev_clear_queues(dev);
2824 	return -EIO;
2825 }
2826 
2827 /*
2828  * Stop device: disable rx and tx functions to allow for reconfiguring.
2829  */
2830 static int
2831 ixgbe_dev_stop(struct rte_eth_dev *dev)
2832 {
2833 	struct rte_eth_link link;
2834 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2835 	struct ixgbe_hw *hw =
2836 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2837 	struct ixgbe_vf_info *vfinfo =
2838 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2839 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2840 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2841 	int vf;
2842 	struct ixgbe_tm_conf *tm_conf =
2843 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2844 
2845 	if (hw->adapter_stopped)
2846 		return 0;
2847 
2848 	PMD_INIT_FUNC_TRACE();
2849 
2850 	ixgbe_dev_wait_setup_link_complete(dev, 0);
2851 
2852 	/* disable interrupts */
2853 	ixgbe_disable_intr(hw);
2854 
2855 	/* reset the NIC */
2856 	ixgbe_pf_reset_hw(hw);
2857 	hw->adapter_stopped = 0;
2858 
2859 	/* stop adapter */
2860 	ixgbe_stop_adapter(hw);
2861 
2862 	for (vf = 0; vfinfo != NULL && vf < pci_dev->max_vfs; vf++)
2863 		vfinfo[vf].clear_to_send = false;
2864 
2865 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2866 		/* Turn off the copper */
2867 		ixgbe_set_phy_power(hw, false);
2868 	} else {
2869 		/* Turn off the laser */
2870 		ixgbe_disable_tx_laser(hw);
2871 	}
2872 
2873 	ixgbe_dev_clear_queues(dev);
2874 
2875 	/* Clear stored conf */
2876 	dev->data->scattered_rx = 0;
2877 	dev->data->lro = 0;
2878 
2879 	/* Clear recorded link status */
2880 	memset(&link, 0, sizeof(link));
2881 	rte_eth_linkstatus_set(dev, &link);
2882 
2883 	if (!rte_intr_allow_others(intr_handle))
2884 		/* resume to the default handler */
2885 		rte_intr_callback_register(intr_handle,
2886 					   ixgbe_dev_interrupt_handler,
2887 					   (void *)dev);
2888 
2889 	/* Clean datapath event and queue/vec mapping */
2890 	rte_intr_efd_disable(intr_handle);
2891 	if (intr_handle->intr_vec != NULL) {
2892 		rte_free(intr_handle->intr_vec);
2893 		intr_handle->intr_vec = NULL;
2894 	}
2895 
2896 	/* reset hierarchy commit */
2897 	tm_conf->committed = false;
2898 
2899 	adapter->rss_reta_updated = 0;
2900 
2901 	hw->adapter_stopped = true;
2902 	dev->data->dev_started = 0;
2903 
2904 	return 0;
2905 }
2906 
2907 /*
2908  * Set device link up: enable tx.
2909  */
2910 static int
2911 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2912 {
2913 	struct ixgbe_hw *hw =
2914 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2915 	if (hw->mac.type == ixgbe_mac_82599EB) {
2916 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2917 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2918 			/* Not suported in bypass mode */
2919 			PMD_INIT_LOG(ERR, "Set link up is not supported "
2920 				     "by device id 0x%x", hw->device_id);
2921 			return -ENOTSUP;
2922 		}
2923 #endif
2924 	}
2925 
2926 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2927 		/* Turn on the copper */
2928 		ixgbe_set_phy_power(hw, true);
2929 	} else {
2930 		/* Turn on the laser */
2931 		ixgbe_enable_tx_laser(hw);
2932 		ixgbe_dev_link_update(dev, 0);
2933 	}
2934 
2935 	return 0;
2936 }
2937 
2938 /*
2939  * Set device link down: disable tx.
2940  */
2941 static int
2942 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2943 {
2944 	struct ixgbe_hw *hw =
2945 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2946 	if (hw->mac.type == ixgbe_mac_82599EB) {
2947 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2948 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2949 			/* Not suported in bypass mode */
2950 			PMD_INIT_LOG(ERR, "Set link down is not supported "
2951 				     "by device id 0x%x", hw->device_id);
2952 			return -ENOTSUP;
2953 		}
2954 #endif
2955 	}
2956 
2957 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2958 		/* Turn off the copper */
2959 		ixgbe_set_phy_power(hw, false);
2960 	} else {
2961 		/* Turn off the laser */
2962 		ixgbe_disable_tx_laser(hw);
2963 		ixgbe_dev_link_update(dev, 0);
2964 	}
2965 
2966 	return 0;
2967 }
2968 
2969 /*
2970  * Reset and stop device.
2971  */
2972 static int
2973 ixgbe_dev_close(struct rte_eth_dev *dev)
2974 {
2975 	struct ixgbe_hw *hw =
2976 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2977 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2978 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2979 	int retries = 0;
2980 	int ret;
2981 
2982 	PMD_INIT_FUNC_TRACE();
2983 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2984 		return 0;
2985 
2986 	ixgbe_pf_reset_hw(hw);
2987 
2988 	ret = ixgbe_dev_stop(dev);
2989 
2990 	ixgbe_dev_free_queues(dev);
2991 
2992 	ixgbe_disable_pcie_master(hw);
2993 
2994 	/* reprogram the RAR[0] in case user changed it. */
2995 	ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
2996 
2997 	/* Unlock any pending hardware semaphore */
2998 	ixgbe_swfw_lock_reset(hw);
2999 
3000 	/* disable uio intr before callback unregister */
3001 	rte_intr_disable(intr_handle);
3002 
3003 	do {
3004 		ret = rte_intr_callback_unregister(intr_handle,
3005 				ixgbe_dev_interrupt_handler, dev);
3006 		if (ret >= 0 || ret == -ENOENT) {
3007 			break;
3008 		} else if (ret != -EAGAIN) {
3009 			PMD_INIT_LOG(ERR,
3010 				"intr callback unregister failed: %d",
3011 				ret);
3012 		}
3013 		rte_delay_ms(100);
3014 	} while (retries++ < (10 + IXGBE_LINK_UP_TIME));
3015 
3016 	/* cancel the delay handler before remove dev */
3017 	rte_eal_alarm_cancel(ixgbe_dev_interrupt_delayed_handler, dev);
3018 
3019 	/* uninitialize PF if max_vfs not zero */
3020 	ixgbe_pf_host_uninit(dev);
3021 
3022 	/* remove all the fdir filters & hash */
3023 	ixgbe_fdir_filter_uninit(dev);
3024 
3025 	/* remove all the L2 tunnel filters & hash */
3026 	ixgbe_l2_tn_filter_uninit(dev);
3027 
3028 	/* Remove all ntuple filters of the device */
3029 	ixgbe_ntuple_filter_uninit(dev);
3030 
3031 	/* clear all the filters list */
3032 	ixgbe_filterlist_flush();
3033 
3034 	/* Remove all Traffic Manager configuration */
3035 	ixgbe_tm_conf_uninit(dev);
3036 
3037 #ifdef RTE_LIB_SECURITY
3038 	rte_free(dev->security_ctx);
3039 #endif
3040 
3041 	return ret;
3042 }
3043 
3044 /*
3045  * Reset PF device.
3046  */
3047 static int
3048 ixgbe_dev_reset(struct rte_eth_dev *dev)
3049 {
3050 	int ret;
3051 
3052 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
3053 	 * its VF to make them align with it. The detailed notification
3054 	 * mechanism is PMD specific. As to ixgbe PF, it is rather complex.
3055 	 * To avoid unexpected behavior in VF, currently reset of PF with
3056 	 * SR-IOV activation is not supported. It might be supported later.
3057 	 */
3058 	if (dev->data->sriov.active)
3059 		return -ENOTSUP;
3060 
3061 	ret = eth_ixgbe_dev_uninit(dev);
3062 	if (ret)
3063 		return ret;
3064 
3065 	ret = eth_ixgbe_dev_init(dev, NULL);
3066 
3067 	return ret;
3068 }
3069 
3070 static void
3071 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
3072 			   struct ixgbe_hw_stats *hw_stats,
3073 			   struct ixgbe_macsec_stats *macsec_stats,
3074 			   uint64_t *total_missed_rx, uint64_t *total_qbrc,
3075 			   uint64_t *total_qprc, uint64_t *total_qprdc)
3076 {
3077 	uint32_t bprc, lxon, lxoff, total;
3078 	uint32_t delta_gprc = 0;
3079 	unsigned i;
3080 	/* Workaround for RX byte count not including CRC bytes when CRC
3081 	 * strip is enabled. CRC bytes are removed from counters when crc_strip
3082 	 * is disabled.
3083 	 */
3084 	int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
3085 			IXGBE_HLREG0_RXCRCSTRP);
3086 
3087 	hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
3088 	hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
3089 	hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
3090 	hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
3091 
3092 	for (i = 0; i < 8; i++) {
3093 		uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
3094 
3095 		/* global total per queue */
3096 		hw_stats->mpc[i] += mp;
3097 		/* Running comprehensive total for stats display */
3098 		*total_missed_rx += hw_stats->mpc[i];
3099 		if (hw->mac.type == ixgbe_mac_82598EB) {
3100 			hw_stats->rnbc[i] +=
3101 			    IXGBE_READ_REG(hw, IXGBE_RNBC(i));
3102 			hw_stats->pxonrxc[i] +=
3103 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
3104 			hw_stats->pxoffrxc[i] +=
3105 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
3106 		} else {
3107 			hw_stats->pxonrxc[i] +=
3108 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
3109 			hw_stats->pxoffrxc[i] +=
3110 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
3111 			hw_stats->pxon2offc[i] +=
3112 				IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
3113 		}
3114 		hw_stats->pxontxc[i] +=
3115 		    IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
3116 		hw_stats->pxofftxc[i] +=
3117 		    IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
3118 	}
3119 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3120 		uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
3121 		uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
3122 		uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
3123 
3124 		delta_gprc += delta_qprc;
3125 
3126 		hw_stats->qprc[i] += delta_qprc;
3127 		hw_stats->qptc[i] += delta_qptc;
3128 
3129 		hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
3130 		hw_stats->qbrc[i] +=
3131 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
3132 		if (crc_strip == 0)
3133 			hw_stats->qbrc[i] -= delta_qprc * RTE_ETHER_CRC_LEN;
3134 
3135 		hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
3136 		hw_stats->qbtc[i] +=
3137 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
3138 
3139 		hw_stats->qprdc[i] += delta_qprdc;
3140 		*total_qprdc += hw_stats->qprdc[i];
3141 
3142 		*total_qprc += hw_stats->qprc[i];
3143 		*total_qbrc += hw_stats->qbrc[i];
3144 	}
3145 	hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
3146 	hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
3147 	hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
3148 
3149 	/*
3150 	 * An errata states that gprc actually counts good + missed packets:
3151 	 * Workaround to set gprc to summated queue packet receives
3152 	 */
3153 	hw_stats->gprc = *total_qprc;
3154 
3155 	if (hw->mac.type != ixgbe_mac_82598EB) {
3156 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
3157 		hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
3158 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
3159 		hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
3160 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
3161 		hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
3162 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
3163 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
3164 	} else {
3165 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
3166 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
3167 		/* 82598 only has a counter in the high register */
3168 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
3169 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
3170 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
3171 	}
3172 	uint64_t old_tpr = hw_stats->tpr;
3173 
3174 	hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
3175 	hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
3176 
3177 	if (crc_strip == 0)
3178 		hw_stats->gorc -= delta_gprc * RTE_ETHER_CRC_LEN;
3179 
3180 	uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
3181 	hw_stats->gptc += delta_gptc;
3182 	hw_stats->gotc -= delta_gptc * RTE_ETHER_CRC_LEN;
3183 	hw_stats->tor -= (hw_stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
3184 
3185 	/*
3186 	 * Workaround: mprc hardware is incorrectly counting
3187 	 * broadcasts, so for now we subtract those.
3188 	 */
3189 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
3190 	hw_stats->bprc += bprc;
3191 	hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
3192 	if (hw->mac.type == ixgbe_mac_82598EB)
3193 		hw_stats->mprc -= bprc;
3194 
3195 	hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
3196 	hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
3197 	hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
3198 	hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
3199 	hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
3200 	hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
3201 
3202 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
3203 	hw_stats->lxontxc += lxon;
3204 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
3205 	hw_stats->lxofftxc += lxoff;
3206 	total = lxon + lxoff;
3207 
3208 	hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
3209 	hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
3210 	hw_stats->gptc -= total;
3211 	hw_stats->mptc -= total;
3212 	hw_stats->ptc64 -= total;
3213 	hw_stats->gotc -= total * RTE_ETHER_MIN_LEN;
3214 
3215 	hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3216 	hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3217 	hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3218 	hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
3219 	hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
3220 	hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
3221 	hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
3222 	hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3223 	hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3224 	hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3225 	hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3226 	hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
3227 	hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3228 	hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
3229 	hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
3230 	hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
3231 	/* Only read FCOE on 82599 */
3232 	if (hw->mac.type != ixgbe_mac_82598EB) {
3233 		hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
3234 		hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
3235 		hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
3236 		hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
3237 		hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
3238 	}
3239 
3240 	/* Flow Director Stats registers */
3241 	if (hw->mac.type != ixgbe_mac_82598EB) {
3242 		hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
3243 		hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
3244 		hw_stats->fdirustat_add += IXGBE_READ_REG(hw,
3245 					IXGBE_FDIRUSTAT) & 0xFFFF;
3246 		hw_stats->fdirustat_remove += (IXGBE_READ_REG(hw,
3247 					IXGBE_FDIRUSTAT) >> 16) & 0xFFFF;
3248 		hw_stats->fdirfstat_fadd += IXGBE_READ_REG(hw,
3249 					IXGBE_FDIRFSTAT) & 0xFFFF;
3250 		hw_stats->fdirfstat_fremove += (IXGBE_READ_REG(hw,
3251 					IXGBE_FDIRFSTAT) >> 16) & 0xFFFF;
3252 	}
3253 	/* MACsec Stats registers */
3254 	macsec_stats->out_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECTXUT);
3255 	macsec_stats->out_pkts_encrypted +=
3256 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTE);
3257 	macsec_stats->out_pkts_protected +=
3258 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTP);
3259 	macsec_stats->out_octets_encrypted +=
3260 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTE);
3261 	macsec_stats->out_octets_protected +=
3262 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTP);
3263 	macsec_stats->in_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECRXUT);
3264 	macsec_stats->in_pkts_badtag += IXGBE_READ_REG(hw, IXGBE_LSECRXBAD);
3265 	macsec_stats->in_pkts_nosci += IXGBE_READ_REG(hw, IXGBE_LSECRXNOSCI);
3266 	macsec_stats->in_pkts_unknownsci +=
3267 		IXGBE_READ_REG(hw, IXGBE_LSECRXUNSCI);
3268 	macsec_stats->in_octets_decrypted +=
3269 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTD);
3270 	macsec_stats->in_octets_validated +=
3271 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTV);
3272 	macsec_stats->in_pkts_unchecked += IXGBE_READ_REG(hw, IXGBE_LSECRXUNCH);
3273 	macsec_stats->in_pkts_delayed += IXGBE_READ_REG(hw, IXGBE_LSECRXDELAY);
3274 	macsec_stats->in_pkts_late += IXGBE_READ_REG(hw, IXGBE_LSECRXLATE);
3275 	for (i = 0; i < 2; i++) {
3276 		macsec_stats->in_pkts_ok +=
3277 			IXGBE_READ_REG(hw, IXGBE_LSECRXOK(i));
3278 		macsec_stats->in_pkts_invalid +=
3279 			IXGBE_READ_REG(hw, IXGBE_LSECRXINV(i));
3280 		macsec_stats->in_pkts_notvalid +=
3281 			IXGBE_READ_REG(hw, IXGBE_LSECRXNV(i));
3282 	}
3283 	macsec_stats->in_pkts_unusedsa += IXGBE_READ_REG(hw, IXGBE_LSECRXUNSA);
3284 	macsec_stats->in_pkts_notusingsa +=
3285 		IXGBE_READ_REG(hw, IXGBE_LSECRXNUSA);
3286 }
3287 
3288 /*
3289  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
3290  */
3291 static int
3292 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3293 {
3294 	struct ixgbe_hw *hw =
3295 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3296 	struct ixgbe_hw_stats *hw_stats =
3297 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3298 	struct ixgbe_macsec_stats *macsec_stats =
3299 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3300 				dev->data->dev_private);
3301 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3302 	unsigned i;
3303 
3304 	total_missed_rx = 0;
3305 	total_qbrc = 0;
3306 	total_qprc = 0;
3307 	total_qprdc = 0;
3308 
3309 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3310 			&total_qbrc, &total_qprc, &total_qprdc);
3311 
3312 	if (stats == NULL)
3313 		return -EINVAL;
3314 
3315 	/* Fill out the rte_eth_stats statistics structure */
3316 	stats->ipackets = total_qprc;
3317 	stats->ibytes = total_qbrc;
3318 	stats->opackets = hw_stats->gptc;
3319 	stats->obytes = hw_stats->gotc;
3320 
3321 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3322 		stats->q_ipackets[i] = hw_stats->qprc[i];
3323 		stats->q_opackets[i] = hw_stats->qptc[i];
3324 		stats->q_ibytes[i] = hw_stats->qbrc[i];
3325 		stats->q_obytes[i] = hw_stats->qbtc[i];
3326 		stats->q_errors[i] = hw_stats->qprdc[i];
3327 	}
3328 
3329 	/* Rx Errors */
3330 	stats->imissed  = total_missed_rx;
3331 	stats->ierrors  = hw_stats->crcerrs +
3332 			  hw_stats->mspdc +
3333 			  hw_stats->rlec +
3334 			  hw_stats->ruc +
3335 			  hw_stats->roc +
3336 			  hw_stats->illerrc +
3337 			  hw_stats->errbc +
3338 			  hw_stats->rfc +
3339 			  hw_stats->fccrc +
3340 			  hw_stats->fclast;
3341 
3342 	/*
3343 	 * 82599 errata, UDP frames with a 0 checksum can be marked as checksum
3344 	 * errors.
3345 	 */
3346 	if (hw->mac.type != ixgbe_mac_82599EB)
3347 		stats->ierrors += hw_stats->xec;
3348 
3349 	/* Tx Errors */
3350 	stats->oerrors  = 0;
3351 	return 0;
3352 }
3353 
3354 static int
3355 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
3356 {
3357 	struct ixgbe_hw_stats *stats =
3358 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3359 
3360 	/* HW registers are cleared on read */
3361 	ixgbe_dev_stats_get(dev, NULL);
3362 
3363 	/* Reset software totals */
3364 	memset(stats, 0, sizeof(*stats));
3365 
3366 	return 0;
3367 }
3368 
3369 /* This function calculates the number of xstats based on the current config */
3370 static unsigned
3371 ixgbe_xstats_calc_num(void) {
3372 	return IXGBE_NB_HW_STATS + IXGBE_NB_MACSEC_STATS +
3373 		(IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
3374 		(IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
3375 }
3376 
3377 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3378 	struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned int size)
3379 {
3380 	const unsigned cnt_stats = ixgbe_xstats_calc_num();
3381 	unsigned stat, i, count;
3382 
3383 	if (xstats_names != NULL) {
3384 		count = 0;
3385 
3386 		/* Note: limit >= cnt_stats checked upstream
3387 		 * in rte_eth_xstats_names()
3388 		 */
3389 
3390 		/* Extended stats from ixgbe_hw_stats */
3391 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3392 			strlcpy(xstats_names[count].name,
3393 				rte_ixgbe_stats_strings[i].name,
3394 				sizeof(xstats_names[count].name));
3395 			count++;
3396 		}
3397 
3398 		/* MACsec Stats */
3399 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3400 			strlcpy(xstats_names[count].name,
3401 				rte_ixgbe_macsec_strings[i].name,
3402 				sizeof(xstats_names[count].name));
3403 			count++;
3404 		}
3405 
3406 		/* RX Priority Stats */
3407 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3408 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3409 				snprintf(xstats_names[count].name,
3410 					sizeof(xstats_names[count].name),
3411 					"rx_priority%u_%s", i,
3412 					rte_ixgbe_rxq_strings[stat].name);
3413 				count++;
3414 			}
3415 		}
3416 
3417 		/* TX Priority Stats */
3418 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3419 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3420 				snprintf(xstats_names[count].name,
3421 					sizeof(xstats_names[count].name),
3422 					"tx_priority%u_%s", i,
3423 					rte_ixgbe_txq_strings[stat].name);
3424 				count++;
3425 			}
3426 		}
3427 	}
3428 	return cnt_stats;
3429 }
3430 
3431 static int ixgbe_dev_xstats_get_names_by_id(
3432 	struct rte_eth_dev *dev,
3433 	const uint64_t *ids,
3434 	struct rte_eth_xstat_name *xstats_names,
3435 	unsigned int limit)
3436 {
3437 	if (!ids) {
3438 		const unsigned int cnt_stats = ixgbe_xstats_calc_num();
3439 		unsigned int stat, i, count;
3440 
3441 		if (xstats_names != NULL) {
3442 			count = 0;
3443 
3444 			/* Note: limit >= cnt_stats checked upstream
3445 			 * in rte_eth_xstats_names()
3446 			 */
3447 
3448 			/* Extended stats from ixgbe_hw_stats */
3449 			for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3450 				strlcpy(xstats_names[count].name,
3451 					rte_ixgbe_stats_strings[i].name,
3452 					sizeof(xstats_names[count].name));
3453 				count++;
3454 			}
3455 
3456 			/* MACsec Stats */
3457 			for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3458 				strlcpy(xstats_names[count].name,
3459 					rte_ixgbe_macsec_strings[i].name,
3460 					sizeof(xstats_names[count].name));
3461 				count++;
3462 			}
3463 
3464 			/* RX Priority Stats */
3465 			for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3466 				for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3467 					snprintf(xstats_names[count].name,
3468 					    sizeof(xstats_names[count].name),
3469 					    "rx_priority%u_%s", i,
3470 					    rte_ixgbe_rxq_strings[stat].name);
3471 					count++;
3472 				}
3473 			}
3474 
3475 			/* TX Priority Stats */
3476 			for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3477 				for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3478 					snprintf(xstats_names[count].name,
3479 					    sizeof(xstats_names[count].name),
3480 					    "tx_priority%u_%s", i,
3481 					    rte_ixgbe_txq_strings[stat].name);
3482 					count++;
3483 				}
3484 			}
3485 		}
3486 		return cnt_stats;
3487 	}
3488 
3489 	uint16_t i;
3490 	uint16_t size = ixgbe_xstats_calc_num();
3491 	struct rte_eth_xstat_name xstats_names_copy[size];
3492 
3493 	ixgbe_dev_xstats_get_names_by_id(dev, NULL, xstats_names_copy,
3494 			size);
3495 
3496 	for (i = 0; i < limit; i++) {
3497 		if (ids[i] >= size) {
3498 			PMD_INIT_LOG(ERR, "id value isn't valid");
3499 			return -1;
3500 		}
3501 		strcpy(xstats_names[i].name,
3502 				xstats_names_copy[ids[i]].name);
3503 	}
3504 	return limit;
3505 }
3506 
3507 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3508 	struct rte_eth_xstat_name *xstats_names, unsigned limit)
3509 {
3510 	unsigned i;
3511 
3512 	if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
3513 		return -ENOMEM;
3514 
3515 	if (xstats_names != NULL)
3516 		for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
3517 			strlcpy(xstats_names[i].name,
3518 				rte_ixgbevf_stats_strings[i].name,
3519 				sizeof(xstats_names[i].name));
3520 	return IXGBEVF_NB_XSTATS;
3521 }
3522 
3523 static int
3524 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3525 					 unsigned n)
3526 {
3527 	struct ixgbe_hw *hw =
3528 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3529 	struct ixgbe_hw_stats *hw_stats =
3530 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3531 	struct ixgbe_macsec_stats *macsec_stats =
3532 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3533 				dev->data->dev_private);
3534 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3535 	unsigned i, stat, count = 0;
3536 
3537 	count = ixgbe_xstats_calc_num();
3538 
3539 	if (n < count)
3540 		return count;
3541 
3542 	total_missed_rx = 0;
3543 	total_qbrc = 0;
3544 	total_qprc = 0;
3545 	total_qprdc = 0;
3546 
3547 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3548 			&total_qbrc, &total_qprc, &total_qprdc);
3549 
3550 	/* If this is a reset xstats is NULL, and we have cleared the
3551 	 * registers by reading them.
3552 	 */
3553 	if (!xstats)
3554 		return 0;
3555 
3556 	/* Extended stats from ixgbe_hw_stats */
3557 	count = 0;
3558 	for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3559 		xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3560 				rte_ixgbe_stats_strings[i].offset);
3561 		xstats[count].id = count;
3562 		count++;
3563 	}
3564 
3565 	/* MACsec Stats */
3566 	for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3567 		xstats[count].value = *(uint64_t *)(((char *)macsec_stats) +
3568 				rte_ixgbe_macsec_strings[i].offset);
3569 		xstats[count].id = count;
3570 		count++;
3571 	}
3572 
3573 	/* RX Priority Stats */
3574 	for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3575 		for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3576 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3577 					rte_ixgbe_rxq_strings[stat].offset +
3578 					(sizeof(uint64_t) * i));
3579 			xstats[count].id = count;
3580 			count++;
3581 		}
3582 	}
3583 
3584 	/* TX Priority Stats */
3585 	for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3586 		for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3587 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3588 					rte_ixgbe_txq_strings[stat].offset +
3589 					(sizeof(uint64_t) * i));
3590 			xstats[count].id = count;
3591 			count++;
3592 		}
3593 	}
3594 	return count;
3595 }
3596 
3597 static int
3598 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
3599 		uint64_t *values, unsigned int n)
3600 {
3601 	if (!ids) {
3602 		struct ixgbe_hw *hw =
3603 				IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3604 		struct ixgbe_hw_stats *hw_stats =
3605 				IXGBE_DEV_PRIVATE_TO_STATS(
3606 						dev->data->dev_private);
3607 		struct ixgbe_macsec_stats *macsec_stats =
3608 				IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3609 					dev->data->dev_private);
3610 		uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3611 		unsigned int i, stat, count = 0;
3612 
3613 		count = ixgbe_xstats_calc_num();
3614 
3615 		if (!ids && n < count)
3616 			return count;
3617 
3618 		total_missed_rx = 0;
3619 		total_qbrc = 0;
3620 		total_qprc = 0;
3621 		total_qprdc = 0;
3622 
3623 		ixgbe_read_stats_registers(hw, hw_stats, macsec_stats,
3624 				&total_missed_rx, &total_qbrc, &total_qprc,
3625 				&total_qprdc);
3626 
3627 		/* If this is a reset xstats is NULL, and we have cleared the
3628 		 * registers by reading them.
3629 		 */
3630 		if (!ids && !values)
3631 			return 0;
3632 
3633 		/* Extended stats from ixgbe_hw_stats */
3634 		count = 0;
3635 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3636 			values[count] = *(uint64_t *)(((char *)hw_stats) +
3637 					rte_ixgbe_stats_strings[i].offset);
3638 			count++;
3639 		}
3640 
3641 		/* MACsec Stats */
3642 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3643 			values[count] = *(uint64_t *)(((char *)macsec_stats) +
3644 					rte_ixgbe_macsec_strings[i].offset);
3645 			count++;
3646 		}
3647 
3648 		/* RX Priority Stats */
3649 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3650 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3651 				values[count] =
3652 					*(uint64_t *)(((char *)hw_stats) +
3653 					rte_ixgbe_rxq_strings[stat].offset +
3654 					(sizeof(uint64_t) * i));
3655 				count++;
3656 			}
3657 		}
3658 
3659 		/* TX Priority Stats */
3660 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3661 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3662 				values[count] =
3663 					*(uint64_t *)(((char *)hw_stats) +
3664 					rte_ixgbe_txq_strings[stat].offset +
3665 					(sizeof(uint64_t) * i));
3666 				count++;
3667 			}
3668 		}
3669 		return count;
3670 	}
3671 
3672 	uint16_t i;
3673 	uint16_t size = ixgbe_xstats_calc_num();
3674 	uint64_t values_copy[size];
3675 
3676 	ixgbe_dev_xstats_get_by_id(dev, NULL, values_copy, size);
3677 
3678 	for (i = 0; i < n; i++) {
3679 		if (ids[i] >= size) {
3680 			PMD_INIT_LOG(ERR, "id value isn't valid");
3681 			return -1;
3682 		}
3683 		values[i] = values_copy[ids[i]];
3684 	}
3685 	return n;
3686 }
3687 
3688 static int
3689 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
3690 {
3691 	struct ixgbe_hw_stats *stats =
3692 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3693 	struct ixgbe_macsec_stats *macsec_stats =
3694 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3695 				dev->data->dev_private);
3696 
3697 	unsigned count = ixgbe_xstats_calc_num();
3698 
3699 	/* HW registers are cleared on read */
3700 	ixgbe_dev_xstats_get(dev, NULL, count);
3701 
3702 	/* Reset software totals */
3703 	memset(stats, 0, sizeof(*stats));
3704 	memset(macsec_stats, 0, sizeof(*macsec_stats));
3705 
3706 	return 0;
3707 }
3708 
3709 static void
3710 ixgbevf_update_stats(struct rte_eth_dev *dev)
3711 {
3712 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3713 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3714 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3715 
3716 	/* Good Rx packet, include VF loopback */
3717 	UPDATE_VF_STAT(IXGBE_VFGPRC,
3718 	    hw_stats->last_vfgprc, hw_stats->vfgprc);
3719 
3720 	/* Good Rx octets, include VF loopback */
3721 	UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
3722 	    hw_stats->last_vfgorc, hw_stats->vfgorc);
3723 
3724 	/* Good Tx packet, include VF loopback */
3725 	UPDATE_VF_STAT(IXGBE_VFGPTC,
3726 	    hw_stats->last_vfgptc, hw_stats->vfgptc);
3727 
3728 	/* Good Tx octets, include VF loopback */
3729 	UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
3730 	    hw_stats->last_vfgotc, hw_stats->vfgotc);
3731 
3732 	/* Rx Multicst Packet */
3733 	UPDATE_VF_STAT(IXGBE_VFMPRC,
3734 	    hw_stats->last_vfmprc, hw_stats->vfmprc);
3735 }
3736 
3737 static int
3738 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3739 		       unsigned n)
3740 {
3741 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3742 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3743 	unsigned i;
3744 
3745 	if (n < IXGBEVF_NB_XSTATS)
3746 		return IXGBEVF_NB_XSTATS;
3747 
3748 	ixgbevf_update_stats(dev);
3749 
3750 	if (!xstats)
3751 		return 0;
3752 
3753 	/* Extended stats */
3754 	for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
3755 		xstats[i].id = i;
3756 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
3757 			rte_ixgbevf_stats_strings[i].offset);
3758 	}
3759 
3760 	return IXGBEVF_NB_XSTATS;
3761 }
3762 
3763 static int
3764 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3765 {
3766 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3767 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3768 
3769 	ixgbevf_update_stats(dev);
3770 
3771 	if (stats == NULL)
3772 		return -EINVAL;
3773 
3774 	stats->ipackets = hw_stats->vfgprc;
3775 	stats->ibytes = hw_stats->vfgorc;
3776 	stats->opackets = hw_stats->vfgptc;
3777 	stats->obytes = hw_stats->vfgotc;
3778 	return 0;
3779 }
3780 
3781 static int
3782 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3783 {
3784 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3785 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3786 
3787 	/* Sync HW register to the last stats */
3788 	ixgbevf_dev_stats_get(dev, NULL);
3789 
3790 	/* reset HW current stats*/
3791 	hw_stats->vfgprc = 0;
3792 	hw_stats->vfgorc = 0;
3793 	hw_stats->vfgptc = 0;
3794 	hw_stats->vfgotc = 0;
3795 	hw_stats->vfmprc = 0;
3796 
3797 	return 0;
3798 }
3799 
3800 static int
3801 ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
3802 {
3803 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3804 	u16 eeprom_verh, eeprom_verl;
3805 	u32 etrack_id;
3806 	int ret;
3807 
3808 	ixgbe_read_eeprom(hw, 0x2e, &eeprom_verh);
3809 	ixgbe_read_eeprom(hw, 0x2d, &eeprom_verl);
3810 
3811 	etrack_id = (eeprom_verh << 16) | eeprom_verl;
3812 	ret = snprintf(fw_version, fw_size, "0x%08x", etrack_id);
3813 	if (ret < 0)
3814 		return -EINVAL;
3815 
3816 	ret += 1; /* add the size of '\0' */
3817 	if (fw_size < (size_t)ret)
3818 		return ret;
3819 	else
3820 		return 0;
3821 }
3822 
3823 static int
3824 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3825 {
3826 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3827 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3828 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3829 
3830 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3831 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3832 	if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3833 		/*
3834 		 * When DCB/VT is off, maximum number of queues changes,
3835 		 * except for 82598EB, which remains constant.
3836 		 */
3837 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
3838 				hw->mac.type != ixgbe_mac_82598EB)
3839 			dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3840 	}
3841 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3842 	dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3843 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3844 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3845 	dev_info->max_vfs = pci_dev->max_vfs;
3846 	if (hw->mac.type == ixgbe_mac_82598EB)
3847 		dev_info->max_vmdq_pools = ETH_16_POOLS;
3848 	else
3849 		dev_info->max_vmdq_pools = ETH_64_POOLS;
3850 	dev_info->max_mtu =  dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3851 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
3852 	dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3853 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3854 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3855 				     dev_info->rx_queue_offload_capa);
3856 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3857 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3858 
3859 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3860 		.rx_thresh = {
3861 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3862 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3863 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3864 		},
3865 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3866 		.rx_drop_en = 0,
3867 		.offloads = 0,
3868 	};
3869 
3870 	dev_info->default_txconf = (struct rte_eth_txconf) {
3871 		.tx_thresh = {
3872 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3873 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3874 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3875 		},
3876 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3877 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3878 		.offloads = 0,
3879 	};
3880 
3881 	dev_info->rx_desc_lim = rx_desc_lim;
3882 	dev_info->tx_desc_lim = tx_desc_lim;
3883 
3884 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3885 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3886 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3887 
3888 	dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G;
3889 	if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
3890 			hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
3891 		dev_info->speed_capa = ETH_LINK_SPEED_10M |
3892 			ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G;
3893 
3894 	if (hw->mac.type == ixgbe_mac_X540 ||
3895 	    hw->mac.type == ixgbe_mac_X540_vf ||
3896 	    hw->mac.type == ixgbe_mac_X550 ||
3897 	    hw->mac.type == ixgbe_mac_X550_vf) {
3898 		dev_info->speed_capa |= ETH_LINK_SPEED_100M;
3899 	}
3900 	if (hw->mac.type == ixgbe_mac_X550) {
3901 		dev_info->speed_capa |= ETH_LINK_SPEED_2_5G;
3902 		dev_info->speed_capa |= ETH_LINK_SPEED_5G;
3903 	}
3904 
3905 	/* Driver-preferred Rx/Tx parameters */
3906 	dev_info->default_rxportconf.burst_size = 32;
3907 	dev_info->default_txportconf.burst_size = 32;
3908 	dev_info->default_rxportconf.nb_queues = 1;
3909 	dev_info->default_txportconf.nb_queues = 1;
3910 	dev_info->default_rxportconf.ring_size = 256;
3911 	dev_info->default_txportconf.ring_size = 256;
3912 
3913 	return 0;
3914 }
3915 
3916 static const uint32_t *
3917 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3918 {
3919 	static const uint32_t ptypes[] = {
3920 		/* For non-vec functions,
3921 		 * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3922 		 * for vec functions,
3923 		 * refers to _recv_raw_pkts_vec().
3924 		 */
3925 		RTE_PTYPE_L2_ETHER,
3926 		RTE_PTYPE_L3_IPV4,
3927 		RTE_PTYPE_L3_IPV4_EXT,
3928 		RTE_PTYPE_L3_IPV6,
3929 		RTE_PTYPE_L3_IPV6_EXT,
3930 		RTE_PTYPE_L4_SCTP,
3931 		RTE_PTYPE_L4_TCP,
3932 		RTE_PTYPE_L4_UDP,
3933 		RTE_PTYPE_TUNNEL_IP,
3934 		RTE_PTYPE_INNER_L3_IPV6,
3935 		RTE_PTYPE_INNER_L3_IPV6_EXT,
3936 		RTE_PTYPE_INNER_L4_TCP,
3937 		RTE_PTYPE_INNER_L4_UDP,
3938 		RTE_PTYPE_UNKNOWN
3939 	};
3940 
3941 	if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3942 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3943 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3944 	    dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3945 		return ptypes;
3946 
3947 #if defined(RTE_ARCH_X86) || defined(__ARM_NEON)
3948 	if (dev->rx_pkt_burst == ixgbe_recv_pkts_vec ||
3949 	    dev->rx_pkt_burst == ixgbe_recv_scattered_pkts_vec)
3950 		return ptypes;
3951 #endif
3952 	return NULL;
3953 }
3954 
3955 static int
3956 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3957 		     struct rte_eth_dev_info *dev_info)
3958 {
3959 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3960 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3961 
3962 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3963 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3964 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3965 	dev_info->max_rx_pktlen = 9728; /* includes CRC, cf MAXFRS reg */
3966 	dev_info->max_mtu = dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3967 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3968 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3969 	dev_info->max_vfs = pci_dev->max_vfs;
3970 	if (hw->mac.type == ixgbe_mac_82598EB)
3971 		dev_info->max_vmdq_pools = ETH_16_POOLS;
3972 	else
3973 		dev_info->max_vmdq_pools = ETH_64_POOLS;
3974 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3975 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3976 				     dev_info->rx_queue_offload_capa);
3977 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3978 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3979 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3980 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3981 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3982 
3983 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3984 		.rx_thresh = {
3985 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3986 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3987 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3988 		},
3989 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3990 		.rx_drop_en = 0,
3991 		.offloads = 0,
3992 	};
3993 
3994 	dev_info->default_txconf = (struct rte_eth_txconf) {
3995 		.tx_thresh = {
3996 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3997 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3998 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3999 		},
4000 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
4001 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
4002 		.offloads = 0,
4003 	};
4004 
4005 	dev_info->rx_desc_lim = rx_desc_lim;
4006 	dev_info->tx_desc_lim = tx_desc_lim;
4007 
4008 	return 0;
4009 }
4010 
4011 static int
4012 ixgbevf_check_link(struct ixgbe_hw *hw, ixgbe_link_speed *speed,
4013 		   bool *link_up, int wait_to_complete)
4014 {
4015 	struct ixgbe_adapter *adapter = container_of(hw,
4016 						     struct ixgbe_adapter, hw);
4017 	struct ixgbe_mbx_info *mbx = &hw->mbx;
4018 	struct ixgbe_mac_info *mac = &hw->mac;
4019 	uint32_t links_reg, in_msg;
4020 	int ret_val = 0;
4021 
4022 	/* If we were hit with a reset drop the link */
4023 	if (!mbx->ops.check_for_rst(hw, 0) || !mbx->timeout)
4024 		mac->get_link_status = true;
4025 
4026 	if (!mac->get_link_status)
4027 		goto out;
4028 
4029 	/* if link status is down no point in checking to see if pf is up */
4030 	links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4031 	if (!(links_reg & IXGBE_LINKS_UP))
4032 		goto out;
4033 
4034 	/* for SFP+ modules and DA cables on 82599 it can take up to 500usecs
4035 	 * before the link status is correct
4036 	 */
4037 	if (mac->type == ixgbe_mac_82599_vf && wait_to_complete) {
4038 		int i;
4039 
4040 		for (i = 0; i < 5; i++) {
4041 			rte_delay_us(100);
4042 			links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4043 
4044 			if (!(links_reg & IXGBE_LINKS_UP))
4045 				goto out;
4046 		}
4047 	}
4048 
4049 	switch (links_reg & IXGBE_LINKS_SPEED_82599) {
4050 	case IXGBE_LINKS_SPEED_10G_82599:
4051 		*speed = IXGBE_LINK_SPEED_10GB_FULL;
4052 		if (hw->mac.type >= ixgbe_mac_X550) {
4053 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4054 				*speed = IXGBE_LINK_SPEED_2_5GB_FULL;
4055 		}
4056 		break;
4057 	case IXGBE_LINKS_SPEED_1G_82599:
4058 		*speed = IXGBE_LINK_SPEED_1GB_FULL;
4059 		break;
4060 	case IXGBE_LINKS_SPEED_100_82599:
4061 		*speed = IXGBE_LINK_SPEED_100_FULL;
4062 		if (hw->mac.type == ixgbe_mac_X550) {
4063 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4064 				*speed = IXGBE_LINK_SPEED_5GB_FULL;
4065 		}
4066 		break;
4067 	case IXGBE_LINKS_SPEED_10_X550EM_A:
4068 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
4069 		/* Since Reserved in older MAC's */
4070 		if (hw->mac.type >= ixgbe_mac_X550)
4071 			*speed = IXGBE_LINK_SPEED_10_FULL;
4072 		break;
4073 	default:
4074 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
4075 	}
4076 
4077 	if (wait_to_complete == 0 && adapter->pflink_fullchk == 0) {
4078 		if (*speed == IXGBE_LINK_SPEED_UNKNOWN)
4079 			mac->get_link_status = true;
4080 		else
4081 			mac->get_link_status = false;
4082 
4083 		goto out;
4084 	}
4085 
4086 	/* if the read failed it could just be a mailbox collision, best wait
4087 	 * until we are called again and don't report an error
4088 	 */
4089 	if (mbx->ops.read(hw, &in_msg, 1, 0))
4090 		goto out;
4091 
4092 	if (!(in_msg & IXGBE_VT_MSGTYPE_CTS)) {
4093 		/* msg is not CTS and is NACK we must have lost CTS status */
4094 		if (in_msg & IXGBE_VT_MSGTYPE_NACK)
4095 			mac->get_link_status = false;
4096 		goto out;
4097 	}
4098 
4099 	/* the pf is talking, if we timed out in the past we reinit */
4100 	if (!mbx->timeout) {
4101 		ret_val = -1;
4102 		goto out;
4103 	}
4104 
4105 	/* if we passed all the tests above then the link is up and we no
4106 	 * longer need to check for link
4107 	 */
4108 	mac->get_link_status = false;
4109 
4110 out:
4111 	*link_up = !mac->get_link_status;
4112 	return ret_val;
4113 }
4114 
4115 /*
4116  * If @timeout_ms was 0, it means that it will not return until link complete.
4117  * It returns 1 on complete, return 0 on timeout.
4118  */
4119 static int
4120 ixgbe_dev_wait_setup_link_complete(struct rte_eth_dev *dev, uint32_t timeout_ms)
4121 {
4122 #define WARNING_TIMEOUT    9000 /* 9s  in total */
4123 	struct ixgbe_adapter *ad = dev->data->dev_private;
4124 	uint32_t timeout = timeout_ms ? timeout_ms : WARNING_TIMEOUT;
4125 
4126 	while (rte_atomic32_read(&ad->link_thread_running)) {
4127 		msec_delay(1);
4128 		timeout--;
4129 
4130 		if (timeout_ms) {
4131 			if (!timeout)
4132 				return 0;
4133 		} else if (!timeout) {
4134 			/* It will not return until link complete */
4135 			timeout = WARNING_TIMEOUT;
4136 			PMD_DRV_LOG(ERR, "IXGBE link thread not complete too long time!");
4137 		}
4138 	}
4139 
4140 	return 1;
4141 }
4142 
4143 static void *
4144 ixgbe_dev_setup_link_thread_handler(void *param)
4145 {
4146 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4147 	struct ixgbe_adapter *ad = dev->data->dev_private;
4148 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4149 	struct ixgbe_interrupt *intr =
4150 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4151 	u32 speed;
4152 	bool autoneg = false;
4153 
4154 	pthread_detach(pthread_self());
4155 	speed = hw->phy.autoneg_advertised;
4156 	if (!speed)
4157 		ixgbe_get_link_capabilities(hw, &speed, &autoneg);
4158 
4159 	ixgbe_setup_link(hw, speed, true);
4160 
4161 	intr->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4162 	rte_atomic32_clear(&ad->link_thread_running);
4163 	return NULL;
4164 }
4165 
4166 /*
4167  * In freebsd environment, nic_uio drivers do not support interrupts,
4168  * rte_intr_callback_register() will fail to register interrupts.
4169  * We can not make link status to change from down to up by interrupt
4170  * callback. So we need to wait for the controller to acquire link
4171  * when ports start.
4172  * It returns 0 on link up.
4173  */
4174 static int
4175 ixgbe_wait_for_link_up(struct ixgbe_hw *hw)
4176 {
4177 #ifdef RTE_EXEC_ENV_FREEBSD
4178 	int err, i;
4179 	bool link_up = false;
4180 	uint32_t speed = 0;
4181 	const int nb_iter = 25;
4182 
4183 	for (i = 0; i < nb_iter; i++) {
4184 		err = ixgbe_check_link(hw, &speed, &link_up, 0);
4185 		if (err)
4186 			return err;
4187 		if (link_up)
4188 			return 0;
4189 		msec_delay(200);
4190 	}
4191 
4192 	return 0;
4193 #else
4194 	RTE_SET_USED(hw);
4195 	return 0;
4196 #endif
4197 }
4198 
4199 /* return 0 means link status changed, -1 means not changed */
4200 int
4201 ixgbe_dev_link_update_share(struct rte_eth_dev *dev,
4202 			    int wait_to_complete, int vf)
4203 {
4204 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4205 	struct ixgbe_adapter *ad = dev->data->dev_private;
4206 	struct rte_eth_link link;
4207 	ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
4208 	struct ixgbe_interrupt *intr =
4209 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4210 	bool link_up;
4211 	int diag;
4212 	int wait = 1;
4213 	u32 esdp_reg;
4214 
4215 	memset(&link, 0, sizeof(link));
4216 	link.link_status = ETH_LINK_DOWN;
4217 	link.link_speed = ETH_SPEED_NUM_NONE;
4218 	link.link_duplex = ETH_LINK_HALF_DUPLEX;
4219 	link.link_autoneg = !(dev->data->dev_conf.link_speeds &
4220 			ETH_LINK_SPEED_FIXED);
4221 
4222 	hw->mac.get_link_status = true;
4223 
4224 	if (intr->flags & IXGBE_FLAG_NEED_LINK_CONFIG)
4225 		return rte_eth_linkstatus_set(dev, &link);
4226 
4227 	/* check if it needs to wait to complete, if lsc interrupt is enabled */
4228 	if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
4229 		wait = 0;
4230 
4231 /* BSD has no interrupt mechanism, so force NIC status synchronization. */
4232 #ifdef RTE_EXEC_ENV_FREEBSD
4233 	wait = 1;
4234 #endif
4235 
4236 	if (vf)
4237 		diag = ixgbevf_check_link(hw, &link_speed, &link_up, wait);
4238 	else
4239 		diag = ixgbe_check_link(hw, &link_speed, &link_up, wait);
4240 
4241 	if (diag != 0) {
4242 		link.link_speed = ETH_SPEED_NUM_100M;
4243 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
4244 		return rte_eth_linkstatus_set(dev, &link);
4245 	}
4246 
4247 	if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4248 		esdp_reg = IXGBE_READ_REG(hw, IXGBE_ESDP);
4249 		if ((esdp_reg & IXGBE_ESDP_SDP3))
4250 			link_up = 0;
4251 	}
4252 
4253 	if (link_up == 0) {
4254 		if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4255 			ixgbe_dev_wait_setup_link_complete(dev, 0);
4256 			if (rte_atomic32_test_and_set(&ad->link_thread_running)) {
4257 				/* To avoid race condition between threads, set
4258 				 * the IXGBE_FLAG_NEED_LINK_CONFIG flag only
4259 				 * when there is no link thread running.
4260 				 */
4261 				intr->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
4262 				if (rte_ctrl_thread_create(&ad->link_thread_tid,
4263 					"ixgbe-link-handler",
4264 					NULL,
4265 					ixgbe_dev_setup_link_thread_handler,
4266 					dev) < 0) {
4267 					PMD_DRV_LOG(ERR,
4268 						"Create link thread failed!");
4269 					rte_atomic32_clear(&ad->link_thread_running);
4270 				}
4271 			} else {
4272 				PMD_DRV_LOG(ERR,
4273 					"Other link thread is running now!");
4274 			}
4275 		}
4276 		return rte_eth_linkstatus_set(dev, &link);
4277 	}
4278 
4279 	link.link_status = ETH_LINK_UP;
4280 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
4281 
4282 	switch (link_speed) {
4283 	default:
4284 	case IXGBE_LINK_SPEED_UNKNOWN:
4285 		link.link_speed = ETH_SPEED_NUM_UNKNOWN;
4286 		break;
4287 
4288 	case IXGBE_LINK_SPEED_10_FULL:
4289 		link.link_speed = ETH_SPEED_NUM_10M;
4290 		break;
4291 
4292 	case IXGBE_LINK_SPEED_100_FULL:
4293 		link.link_speed = ETH_SPEED_NUM_100M;
4294 		break;
4295 
4296 	case IXGBE_LINK_SPEED_1GB_FULL:
4297 		link.link_speed = ETH_SPEED_NUM_1G;
4298 		break;
4299 
4300 	case IXGBE_LINK_SPEED_2_5GB_FULL:
4301 		link.link_speed = ETH_SPEED_NUM_2_5G;
4302 		break;
4303 
4304 	case IXGBE_LINK_SPEED_5GB_FULL:
4305 		link.link_speed = ETH_SPEED_NUM_5G;
4306 		break;
4307 
4308 	case IXGBE_LINK_SPEED_10GB_FULL:
4309 		link.link_speed = ETH_SPEED_NUM_10G;
4310 		break;
4311 	}
4312 
4313 	return rte_eth_linkstatus_set(dev, &link);
4314 }
4315 
4316 static int
4317 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4318 {
4319 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 0);
4320 }
4321 
4322 static int
4323 ixgbevf_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4324 {
4325 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 1);
4326 }
4327 
4328 static int
4329 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
4330 {
4331 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4332 	uint32_t fctrl;
4333 
4334 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4335 	fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4336 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4337 
4338 	return 0;
4339 }
4340 
4341 static int
4342 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
4343 {
4344 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4345 	uint32_t fctrl;
4346 
4347 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4348 	fctrl &= (~IXGBE_FCTRL_UPE);
4349 	if (dev->data->all_multicast == 1)
4350 		fctrl |= IXGBE_FCTRL_MPE;
4351 	else
4352 		fctrl &= (~IXGBE_FCTRL_MPE);
4353 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4354 
4355 	return 0;
4356 }
4357 
4358 static int
4359 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
4360 {
4361 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4362 	uint32_t fctrl;
4363 
4364 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4365 	fctrl |= IXGBE_FCTRL_MPE;
4366 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4367 
4368 	return 0;
4369 }
4370 
4371 static int
4372 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
4373 {
4374 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4375 	uint32_t fctrl;
4376 
4377 	if (dev->data->promiscuous == 1)
4378 		return 0; /* must remain in all_multicast mode */
4379 
4380 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4381 	fctrl &= (~IXGBE_FCTRL_MPE);
4382 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4383 
4384 	return 0;
4385 }
4386 
4387 /**
4388  * It clears the interrupt causes and enables the interrupt.
4389  * It will be called once only during nic initialized.
4390  *
4391  * @param dev
4392  *  Pointer to struct rte_eth_dev.
4393  * @param on
4394  *  Enable or Disable.
4395  *
4396  * @return
4397  *  - On success, zero.
4398  *  - On failure, a negative value.
4399  */
4400 static int
4401 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
4402 {
4403 	struct ixgbe_interrupt *intr =
4404 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4405 
4406 	ixgbe_dev_link_status_print(dev);
4407 	if (on)
4408 		intr->mask |= IXGBE_EICR_LSC;
4409 	else
4410 		intr->mask &= ~IXGBE_EICR_LSC;
4411 
4412 	return 0;
4413 }
4414 
4415 /**
4416  * It clears the interrupt causes and enables the interrupt.
4417  * It will be called once only during nic initialized.
4418  *
4419  * @param dev
4420  *  Pointer to struct rte_eth_dev.
4421  *
4422  * @return
4423  *  - On success, zero.
4424  *  - On failure, a negative value.
4425  */
4426 static int
4427 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
4428 {
4429 	struct ixgbe_interrupt *intr =
4430 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4431 
4432 	intr->mask |= IXGBE_EICR_RTX_QUEUE;
4433 
4434 	return 0;
4435 }
4436 
4437 /**
4438  * It clears the interrupt causes and enables the interrupt.
4439  * It will be called once only during nic initialized.
4440  *
4441  * @param dev
4442  *  Pointer to struct rte_eth_dev.
4443  *
4444  * @return
4445  *  - On success, zero.
4446  *  - On failure, a negative value.
4447  */
4448 static int
4449 ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev)
4450 {
4451 	struct ixgbe_interrupt *intr =
4452 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4453 
4454 	intr->mask |= IXGBE_EICR_LINKSEC;
4455 
4456 	return 0;
4457 }
4458 
4459 /*
4460  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
4461  *
4462  * @param dev
4463  *  Pointer to struct rte_eth_dev.
4464  *
4465  * @return
4466  *  - On success, zero.
4467  *  - On failure, a negative value.
4468  */
4469 static int
4470 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
4471 {
4472 	uint32_t eicr;
4473 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4474 	struct ixgbe_interrupt *intr =
4475 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4476 
4477 	/* clear all cause mask */
4478 	ixgbe_disable_intr(hw);
4479 
4480 	/* read-on-clear nic registers here */
4481 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4482 	PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
4483 
4484 	intr->flags = 0;
4485 
4486 	/* set flag for async link update */
4487 	if (eicr & IXGBE_EICR_LSC)
4488 		intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4489 
4490 	if (eicr & IXGBE_EICR_MAILBOX)
4491 		intr->flags |= IXGBE_FLAG_MAILBOX;
4492 
4493 	if (eicr & IXGBE_EICR_LINKSEC)
4494 		intr->flags |= IXGBE_FLAG_MACSEC;
4495 
4496 	if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
4497 	    hw->phy.type == ixgbe_phy_x550em_ext_t &&
4498 	    (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
4499 		intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
4500 
4501 	return 0;
4502 }
4503 
4504 /**
4505  * It gets and then prints the link status.
4506  *
4507  * @param dev
4508  *  Pointer to struct rte_eth_dev.
4509  *
4510  * @return
4511  *  - On success, zero.
4512  *  - On failure, a negative value.
4513  */
4514 static void
4515 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
4516 {
4517 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4518 	struct rte_eth_link link;
4519 
4520 	rte_eth_linkstatus_get(dev, &link);
4521 
4522 	if (link.link_status) {
4523 		PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
4524 					(int)(dev->data->port_id),
4525 					(unsigned)link.link_speed,
4526 			link.link_duplex == ETH_LINK_FULL_DUPLEX ?
4527 					"full-duplex" : "half-duplex");
4528 	} else {
4529 		PMD_INIT_LOG(INFO, " Port %d: Link Down",
4530 				(int)(dev->data->port_id));
4531 	}
4532 	PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
4533 				pci_dev->addr.domain,
4534 				pci_dev->addr.bus,
4535 				pci_dev->addr.devid,
4536 				pci_dev->addr.function);
4537 }
4538 
4539 /*
4540  * It executes link_update after knowing an interrupt occurred.
4541  *
4542  * @param dev
4543  *  Pointer to struct rte_eth_dev.
4544  *
4545  * @return
4546  *  - On success, zero.
4547  *  - On failure, a negative value.
4548  */
4549 static int
4550 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
4551 {
4552 	struct ixgbe_interrupt *intr =
4553 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4554 	int64_t timeout;
4555 	struct ixgbe_hw *hw =
4556 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4557 
4558 	PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
4559 
4560 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
4561 		ixgbe_pf_mbx_process(dev);
4562 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
4563 	}
4564 
4565 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4566 		ixgbe_handle_lasi(hw);
4567 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4568 	}
4569 
4570 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4571 		struct rte_eth_link link;
4572 
4573 		/* get the link status before link update, for predicting later */
4574 		rte_eth_linkstatus_get(dev, &link);
4575 
4576 		ixgbe_dev_link_update(dev, 0);
4577 
4578 		/* likely to up */
4579 		if (!link.link_status)
4580 			/* handle it 1 sec later, wait it being stable */
4581 			timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
4582 		/* likely to down */
4583 		else
4584 			/* handle it 4 sec later, wait it being stable */
4585 			timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
4586 
4587 		ixgbe_dev_link_status_print(dev);
4588 		if (rte_eal_alarm_set(timeout * 1000,
4589 				      ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
4590 			PMD_DRV_LOG(ERR, "Error setting alarm");
4591 		else {
4592 			/* remember original mask */
4593 			intr->mask_original = intr->mask;
4594 			/* only disable lsc interrupt */
4595 			intr->mask &= ~IXGBE_EIMS_LSC;
4596 		}
4597 	}
4598 
4599 	PMD_DRV_LOG(DEBUG, "enable intr immediately");
4600 	ixgbe_enable_intr(dev);
4601 
4602 	return 0;
4603 }
4604 
4605 /**
4606  * Interrupt handler which shall be registered for alarm callback for delayed
4607  * handling specific interrupt to wait for the stable nic state. As the
4608  * NIC interrupt state is not stable for ixgbe after link is just down,
4609  * it needs to wait 4 seconds to get the stable status.
4610  *
4611  * @param handle
4612  *  Pointer to interrupt handle.
4613  * @param param
4614  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4615  *
4616  * @return
4617  *  void
4618  */
4619 static void
4620 ixgbe_dev_interrupt_delayed_handler(void *param)
4621 {
4622 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4623 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4624 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
4625 	struct ixgbe_interrupt *intr =
4626 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4627 	struct ixgbe_hw *hw =
4628 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4629 	uint32_t eicr;
4630 
4631 	ixgbe_disable_intr(hw);
4632 
4633 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4634 	if (eicr & IXGBE_EICR_MAILBOX)
4635 		ixgbe_pf_mbx_process(dev);
4636 
4637 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4638 		ixgbe_handle_lasi(hw);
4639 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4640 	}
4641 
4642 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4643 		ixgbe_dev_link_update(dev, 0);
4644 		intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4645 		ixgbe_dev_link_status_print(dev);
4646 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
4647 	}
4648 
4649 	if (intr->flags & IXGBE_FLAG_MACSEC) {
4650 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_MACSEC, NULL);
4651 		intr->flags &= ~IXGBE_FLAG_MACSEC;
4652 	}
4653 
4654 	/* restore original mask */
4655 	intr->mask = intr->mask_original;
4656 	intr->mask_original = 0;
4657 
4658 	PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
4659 	ixgbe_enable_intr(dev);
4660 	rte_intr_ack(intr_handle);
4661 }
4662 
4663 /**
4664  * Interrupt handler triggered by NIC  for handling
4665  * specific interrupt.
4666  *
4667  * @param handle
4668  *  Pointer to interrupt handle.
4669  * @param param
4670  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4671  *
4672  * @return
4673  *  void
4674  */
4675 static void
4676 ixgbe_dev_interrupt_handler(void *param)
4677 {
4678 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4679 
4680 	ixgbe_dev_interrupt_get_status(dev);
4681 	ixgbe_dev_interrupt_action(dev);
4682 }
4683 
4684 static int
4685 ixgbe_dev_led_on(struct rte_eth_dev *dev)
4686 {
4687 	struct ixgbe_hw *hw;
4688 
4689 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4690 	return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4691 }
4692 
4693 static int
4694 ixgbe_dev_led_off(struct rte_eth_dev *dev)
4695 {
4696 	struct ixgbe_hw *hw;
4697 
4698 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4699 	return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4700 }
4701 
4702 static int
4703 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4704 {
4705 	struct ixgbe_hw *hw;
4706 	uint32_t mflcn_reg;
4707 	uint32_t fccfg_reg;
4708 	int rx_pause;
4709 	int tx_pause;
4710 
4711 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4712 
4713 	fc_conf->pause_time = hw->fc.pause_time;
4714 	fc_conf->high_water = hw->fc.high_water[0];
4715 	fc_conf->low_water = hw->fc.low_water[0];
4716 	fc_conf->send_xon = hw->fc.send_xon;
4717 	fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
4718 
4719 	/*
4720 	 * Return rx_pause status according to actual setting of
4721 	 * MFLCN register.
4722 	 */
4723 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4724 	if (mflcn_reg & IXGBE_MFLCN_PMCF)
4725 		fc_conf->mac_ctrl_frame_fwd = 1;
4726 	else
4727 		fc_conf->mac_ctrl_frame_fwd = 0;
4728 
4729 	if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
4730 		rx_pause = 1;
4731 	else
4732 		rx_pause = 0;
4733 
4734 	/*
4735 	 * Return tx_pause status according to actual setting of
4736 	 * FCCFG register.
4737 	 */
4738 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4739 	if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
4740 		tx_pause = 1;
4741 	else
4742 		tx_pause = 0;
4743 
4744 	if (rx_pause && tx_pause)
4745 		fc_conf->mode = RTE_FC_FULL;
4746 	else if (rx_pause)
4747 		fc_conf->mode = RTE_FC_RX_PAUSE;
4748 	else if (tx_pause)
4749 		fc_conf->mode = RTE_FC_TX_PAUSE;
4750 	else
4751 		fc_conf->mode = RTE_FC_NONE;
4752 
4753 	return 0;
4754 }
4755 
4756 static int
4757 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4758 {
4759 	struct ixgbe_hw *hw;
4760 	struct ixgbe_adapter *adapter = dev->data->dev_private;
4761 	int err;
4762 	uint32_t rx_buf_size;
4763 	uint32_t max_high_water;
4764 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4765 		ixgbe_fc_none,
4766 		ixgbe_fc_rx_pause,
4767 		ixgbe_fc_tx_pause,
4768 		ixgbe_fc_full
4769 	};
4770 
4771 	PMD_INIT_FUNC_TRACE();
4772 
4773 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4774 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
4775 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4776 
4777 	/*
4778 	 * At least reserve one Ethernet frame for watermark
4779 	 * high_water/low_water in kilo bytes for ixgbe
4780 	 */
4781 	max_high_water = (rx_buf_size -
4782 			RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4783 	if ((fc_conf->high_water > max_high_water) ||
4784 		(fc_conf->high_water < fc_conf->low_water)) {
4785 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4786 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4787 		return -EINVAL;
4788 	}
4789 
4790 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
4791 	hw->fc.pause_time     = fc_conf->pause_time;
4792 	hw->fc.high_water[0]  = fc_conf->high_water;
4793 	hw->fc.low_water[0]   = fc_conf->low_water;
4794 	hw->fc.send_xon       = fc_conf->send_xon;
4795 	hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
4796 	adapter->mac_ctrl_frame_fwd = fc_conf->mac_ctrl_frame_fwd;
4797 
4798 	err = ixgbe_flow_ctrl_enable(dev, hw);
4799 	if (err < 0) {
4800 		PMD_INIT_LOG(ERR, "ixgbe_flow_ctrl_enable = 0x%x", err);
4801 		return -EIO;
4802 	}
4803 	return err;
4804 }
4805 
4806 /**
4807  *  ixgbe_pfc_enable_generic - Enable flow control
4808  *  @hw: pointer to hardware structure
4809  *  @tc_num: traffic class number
4810  *  Enable flow control according to the current settings.
4811  */
4812 static int
4813 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
4814 {
4815 	int ret_val = 0;
4816 	uint32_t mflcn_reg, fccfg_reg;
4817 	uint32_t reg;
4818 	uint32_t fcrtl, fcrth;
4819 	uint8_t i;
4820 	uint8_t nb_rx_en;
4821 
4822 	/* Validate the water mark configuration */
4823 	if (!hw->fc.pause_time) {
4824 		ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4825 		goto out;
4826 	}
4827 
4828 	/* Low water mark of zero causes XOFF floods */
4829 	if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
4830 		 /* High/Low water can not be 0 */
4831 		if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_water[tc_num])) {
4832 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4833 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4834 			goto out;
4835 		}
4836 
4837 		if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
4838 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4839 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4840 			goto out;
4841 		}
4842 	}
4843 	/* Negotiate the fc mode to use */
4844 	ixgbe_fc_autoneg(hw);
4845 
4846 	/* Disable any previous flow control settings */
4847 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4848 	mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
4849 
4850 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4851 	fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
4852 
4853 	switch (hw->fc.current_mode) {
4854 	case ixgbe_fc_none:
4855 		/*
4856 		 * If the count of enabled RX Priority Flow control >1,
4857 		 * and the TX pause can not be disabled
4858 		 */
4859 		nb_rx_en = 0;
4860 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4861 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4862 			if (reg & IXGBE_FCRTH_FCEN)
4863 				nb_rx_en++;
4864 		}
4865 		if (nb_rx_en > 1)
4866 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4867 		break;
4868 	case ixgbe_fc_rx_pause:
4869 		/*
4870 		 * Rx Flow control is enabled and Tx Flow control is
4871 		 * disabled by software override. Since there really
4872 		 * isn't a way to advertise that we are capable of RX
4873 		 * Pause ONLY, we will advertise that we support both
4874 		 * symmetric and asymmetric Rx PAUSE.  Later, we will
4875 		 * disable the adapter's ability to send PAUSE frames.
4876 		 */
4877 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4878 		/*
4879 		 * If the count of enabled RX Priority Flow control >1,
4880 		 * and the TX pause can not be disabled
4881 		 */
4882 		nb_rx_en = 0;
4883 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4884 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4885 			if (reg & IXGBE_FCRTH_FCEN)
4886 				nb_rx_en++;
4887 		}
4888 		if (nb_rx_en > 1)
4889 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4890 		break;
4891 	case ixgbe_fc_tx_pause:
4892 		/*
4893 		 * Tx Flow control is enabled, and Rx Flow control is
4894 		 * disabled by software override.
4895 		 */
4896 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4897 		break;
4898 	case ixgbe_fc_full:
4899 		/* Flow control (both Rx and Tx) is enabled by SW override. */
4900 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4901 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4902 		break;
4903 	default:
4904 		PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
4905 		ret_val = IXGBE_ERR_CONFIG;
4906 		goto out;
4907 	}
4908 
4909 	/* Set 802.3x based flow control settings. */
4910 	mflcn_reg |= IXGBE_MFLCN_DPF;
4911 	IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
4912 	IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
4913 
4914 	/* Set up and enable Rx high/low water mark thresholds, enable XON. */
4915 	if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
4916 		hw->fc.high_water[tc_num]) {
4917 		fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
4918 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
4919 		fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
4920 	} else {
4921 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
4922 		/*
4923 		 * In order to prevent Tx hangs when the internal Tx
4924 		 * switch is enabled we must set the high water mark
4925 		 * to the maximum FCRTH value.  This allows the Tx
4926 		 * switch to function even under heavy Rx workloads.
4927 		 */
4928 		fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
4929 	}
4930 	IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
4931 
4932 	/* Configure pause time (2 TCs per register) */
4933 	reg = hw->fc.pause_time * 0x00010001;
4934 	for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
4935 		IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
4936 
4937 	/* Configure flow control refresh threshold value */
4938 	IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
4939 
4940 out:
4941 	return ret_val;
4942 }
4943 
4944 static int
4945 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
4946 {
4947 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4948 	int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
4949 
4950 	if (hw->mac.type != ixgbe_mac_82598EB) {
4951 		ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
4952 	}
4953 	return ret_val;
4954 }
4955 
4956 static int
4957 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
4958 {
4959 	int err;
4960 	uint32_t rx_buf_size;
4961 	uint32_t max_high_water;
4962 	uint8_t tc_num;
4963 	uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
4964 	struct ixgbe_hw *hw =
4965 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4966 	struct ixgbe_dcb_config *dcb_config =
4967 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
4968 
4969 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4970 		ixgbe_fc_none,
4971 		ixgbe_fc_rx_pause,
4972 		ixgbe_fc_tx_pause,
4973 		ixgbe_fc_full
4974 	};
4975 
4976 	PMD_INIT_FUNC_TRACE();
4977 
4978 	ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
4979 	tc_num = map[pfc_conf->priority];
4980 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
4981 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4982 	/*
4983 	 * At least reserve one Ethernet frame for watermark
4984 	 * high_water/low_water in kilo bytes for ixgbe
4985 	 */
4986 	max_high_water = (rx_buf_size -
4987 			RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4988 	if ((pfc_conf->fc.high_water > max_high_water) ||
4989 	    (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
4990 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4991 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4992 		return -EINVAL;
4993 	}
4994 
4995 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
4996 	hw->fc.pause_time = pfc_conf->fc.pause_time;
4997 	hw->fc.send_xon = pfc_conf->fc.send_xon;
4998 	hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
4999 	hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
5000 
5001 	err = ixgbe_dcb_pfc_enable(dev, tc_num);
5002 
5003 	/* Not negotiated is not an error case */
5004 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
5005 		return 0;
5006 
5007 	PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
5008 	return -EIO;
5009 }
5010 
5011 static int
5012 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
5013 			  struct rte_eth_rss_reta_entry64 *reta_conf,
5014 			  uint16_t reta_size)
5015 {
5016 	uint16_t i, sp_reta_size;
5017 	uint8_t j, mask;
5018 	uint32_t reta, r;
5019 	uint16_t idx, shift;
5020 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5021 	struct rte_eth_dev_data *dev_data = dev->data;
5022 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5023 	uint32_t reta_reg;
5024 
5025 	PMD_INIT_FUNC_TRACE();
5026 
5027 	if (!dev_data->dev_started) {
5028 		PMD_DRV_LOG(ERR,
5029 			"port %d must be started before rss reta update",
5030 			 dev_data->port_id);
5031 		return -EIO;
5032 	}
5033 
5034 	if (!ixgbe_rss_update_sp(hw->mac.type)) {
5035 		PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
5036 			"NIC.");
5037 		return -ENOTSUP;
5038 	}
5039 
5040 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5041 	if (reta_size != sp_reta_size) {
5042 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5043 			"(%d) doesn't match the number hardware can supported "
5044 			"(%d)", reta_size, sp_reta_size);
5045 		return -EINVAL;
5046 	}
5047 
5048 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5049 		idx = i / RTE_RETA_GROUP_SIZE;
5050 		shift = i % RTE_RETA_GROUP_SIZE;
5051 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5052 						IXGBE_4_BIT_MASK);
5053 		if (!mask)
5054 			continue;
5055 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5056 		if (mask == IXGBE_4_BIT_MASK)
5057 			r = 0;
5058 		else
5059 			r = IXGBE_READ_REG(hw, reta_reg);
5060 		for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5061 			if (mask & (0x1 << j))
5062 				reta |= reta_conf[idx].reta[shift + j] <<
5063 							(CHAR_BIT * j);
5064 			else
5065 				reta |= r & (IXGBE_8_BIT_MASK <<
5066 						(CHAR_BIT * j));
5067 		}
5068 		IXGBE_WRITE_REG(hw, reta_reg, reta);
5069 	}
5070 	adapter->rss_reta_updated = 1;
5071 
5072 	return 0;
5073 }
5074 
5075 static int
5076 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
5077 			 struct rte_eth_rss_reta_entry64 *reta_conf,
5078 			 uint16_t reta_size)
5079 {
5080 	uint16_t i, sp_reta_size;
5081 	uint8_t j, mask;
5082 	uint32_t reta;
5083 	uint16_t idx, shift;
5084 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5085 	uint32_t reta_reg;
5086 
5087 	PMD_INIT_FUNC_TRACE();
5088 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5089 	if (reta_size != sp_reta_size) {
5090 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5091 			"(%d) doesn't match the number hardware can supported "
5092 			"(%d)", reta_size, sp_reta_size);
5093 		return -EINVAL;
5094 	}
5095 
5096 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5097 		idx = i / RTE_RETA_GROUP_SIZE;
5098 		shift = i % RTE_RETA_GROUP_SIZE;
5099 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5100 						IXGBE_4_BIT_MASK);
5101 		if (!mask)
5102 			continue;
5103 
5104 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5105 		reta = IXGBE_READ_REG(hw, reta_reg);
5106 		for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5107 			if (mask & (0x1 << j))
5108 				reta_conf[idx].reta[shift + j] =
5109 					((reta >> (CHAR_BIT * j)) &
5110 						IXGBE_8_BIT_MASK);
5111 		}
5112 	}
5113 
5114 	return 0;
5115 }
5116 
5117 static int
5118 ixgbe_add_rar(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
5119 				uint32_t index, uint32_t pool)
5120 {
5121 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5122 	uint32_t enable_addr = 1;
5123 
5124 	return ixgbe_set_rar(hw, index, mac_addr->addr_bytes,
5125 			     pool, enable_addr);
5126 }
5127 
5128 static void
5129 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
5130 {
5131 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5132 
5133 	ixgbe_clear_rar(hw, index);
5134 }
5135 
5136 static int
5137 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
5138 {
5139 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5140 
5141 	ixgbe_remove_rar(dev, 0);
5142 	ixgbe_add_rar(dev, addr, 0, pci_dev->max_vfs);
5143 
5144 	return 0;
5145 }
5146 
5147 static bool
5148 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
5149 {
5150 	if (strcmp(dev->device->driver->name, drv->driver.name))
5151 		return false;
5152 
5153 	return true;
5154 }
5155 
5156 bool
5157 is_ixgbe_supported(struct rte_eth_dev *dev)
5158 {
5159 	return is_device_supported(dev, &rte_ixgbe_pmd);
5160 }
5161 
5162 static int
5163 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
5164 {
5165 	uint32_t hlreg0;
5166 	uint32_t maxfrs;
5167 	struct ixgbe_hw *hw;
5168 	struct rte_eth_dev_info dev_info;
5169 	uint32_t frame_size = mtu + IXGBE_ETH_OVERHEAD;
5170 	struct rte_eth_dev_data *dev_data = dev->data;
5171 	int ret;
5172 
5173 	ret = ixgbe_dev_info_get(dev, &dev_info);
5174 	if (ret != 0)
5175 		return ret;
5176 
5177 	/* check that mtu is within the allowed range */
5178 	if (mtu < RTE_ETHER_MIN_MTU || frame_size > dev_info.max_rx_pktlen)
5179 		return -EINVAL;
5180 
5181 	/* If device is started, refuse mtu that requires the support of
5182 	 * scattered packets when this feature has not been enabled before.
5183 	 */
5184 	if (dev_data->dev_started && !dev_data->scattered_rx &&
5185 	    (frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
5186 	     dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
5187 		PMD_INIT_LOG(ERR, "Stop port first.");
5188 		return -EINVAL;
5189 	}
5190 
5191 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5192 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
5193 
5194 	/* switch to jumbo mode if needed */
5195 	if (frame_size > IXGBE_ETH_MAX_LEN) {
5196 		dev->data->dev_conf.rxmode.offloads |=
5197 			DEV_RX_OFFLOAD_JUMBO_FRAME;
5198 		hlreg0 |= IXGBE_HLREG0_JUMBOEN;
5199 	} else {
5200 		dev->data->dev_conf.rxmode.offloads &=
5201 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
5202 		hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
5203 	}
5204 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
5205 
5206 	/* update max frame size */
5207 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
5208 
5209 	maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
5210 	maxfrs &= 0x0000FFFF;
5211 	maxfrs |= (dev->data->dev_conf.rxmode.max_rx_pkt_len << 16);
5212 	IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
5213 
5214 	return 0;
5215 }
5216 
5217 /*
5218  * Virtual Function operations
5219  */
5220 static void
5221 ixgbevf_intr_disable(struct rte_eth_dev *dev)
5222 {
5223 	struct ixgbe_interrupt *intr =
5224 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5225 	struct ixgbe_hw *hw =
5226 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5227 
5228 	PMD_INIT_FUNC_TRACE();
5229 
5230 	/* Clear interrupt mask to stop from interrupts being generated */
5231 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
5232 
5233 	IXGBE_WRITE_FLUSH(hw);
5234 
5235 	/* Clear mask value. */
5236 	intr->mask = 0;
5237 }
5238 
5239 static void
5240 ixgbevf_intr_enable(struct rte_eth_dev *dev)
5241 {
5242 	struct ixgbe_interrupt *intr =
5243 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5244 	struct ixgbe_hw *hw =
5245 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5246 
5247 	PMD_INIT_FUNC_TRACE();
5248 
5249 	/* VF enable interrupt autoclean */
5250 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
5251 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
5252 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
5253 
5254 	IXGBE_WRITE_FLUSH(hw);
5255 
5256 	/* Save IXGBE_VTEIMS value to mask. */
5257 	intr->mask = IXGBE_VF_IRQ_ENABLE_MASK;
5258 }
5259 
5260 static int
5261 ixgbevf_dev_configure(struct rte_eth_dev *dev)
5262 {
5263 	struct rte_eth_conf *conf = &dev->data->dev_conf;
5264 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5265 
5266 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
5267 		     dev->data->port_id);
5268 
5269 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
5270 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
5271 
5272 	/*
5273 	 * VF has no ability to enable/disable HW CRC
5274 	 * Keep the persistent behavior the same as Host PF
5275 	 */
5276 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
5277 	if (conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
5278 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
5279 		conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_KEEP_CRC;
5280 	}
5281 #else
5282 	if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC)) {
5283 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
5284 		conf->rxmode.offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
5285 	}
5286 #endif
5287 
5288 	/*
5289 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
5290 	 * allocation or vector Rx preconditions we will reset it.
5291 	 */
5292 	adapter->rx_bulk_alloc_allowed = true;
5293 	adapter->rx_vec_allowed = true;
5294 
5295 	return 0;
5296 }
5297 
5298 static int
5299 ixgbevf_dev_start(struct rte_eth_dev *dev)
5300 {
5301 	struct ixgbe_hw *hw =
5302 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5303 	uint32_t intr_vector = 0;
5304 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5305 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5306 
5307 	int err, mask = 0;
5308 
5309 	PMD_INIT_FUNC_TRACE();
5310 
5311 	/* Stop the link setup handler before resetting the HW. */
5312 	ixgbe_dev_wait_setup_link_complete(dev, 0);
5313 
5314 	err = hw->mac.ops.reset_hw(hw);
5315 
5316 	/**
5317 	 * In this case, reuses the MAC address assigned by VF
5318 	 * initialization.
5319 	 */
5320 	if (err != IXGBE_SUCCESS && err != IXGBE_ERR_INVALID_MAC_ADDR) {
5321 		PMD_INIT_LOG(ERR, "Unable to reset vf hardware (%d)", err);
5322 		return err;
5323 	}
5324 
5325 	hw->mac.get_link_status = true;
5326 
5327 	/* negotiate mailbox API version to use with the PF. */
5328 	ixgbevf_negotiate_api(hw);
5329 
5330 	ixgbevf_dev_tx_init(dev);
5331 
5332 	/* This can fail when allocating mbufs for descriptor rings */
5333 	err = ixgbevf_dev_rx_init(dev);
5334 	if (err) {
5335 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
5336 		ixgbe_dev_clear_queues(dev);
5337 		return err;
5338 	}
5339 
5340 	/* Set vfta */
5341 	ixgbevf_set_vfta_all(dev, 1);
5342 
5343 	/* Set HW strip */
5344 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
5345 		ETH_VLAN_EXTEND_MASK;
5346 	err = ixgbevf_vlan_offload_config(dev, mask);
5347 	if (err) {
5348 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
5349 		ixgbe_dev_clear_queues(dev);
5350 		return err;
5351 	}
5352 
5353 	ixgbevf_dev_rxtx_start(dev);
5354 
5355 	/* check and configure queue intr-vector mapping */
5356 	if (rte_intr_cap_multiple(intr_handle) &&
5357 	    dev->data->dev_conf.intr_conf.rxq) {
5358 		/* According to datasheet, only vector 0/1/2 can be used,
5359 		 * now only one vector is used for Rx queue
5360 		 */
5361 		intr_vector = 1;
5362 		if (rte_intr_efd_enable(intr_handle, intr_vector)) {
5363 			ixgbe_dev_clear_queues(dev);
5364 			return -1;
5365 		}
5366 	}
5367 
5368 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
5369 		intr_handle->intr_vec =
5370 			rte_zmalloc("intr_vec",
5371 				    dev->data->nb_rx_queues * sizeof(int), 0);
5372 		if (intr_handle->intr_vec == NULL) {
5373 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
5374 				     " intr_vec", dev->data->nb_rx_queues);
5375 			ixgbe_dev_clear_queues(dev);
5376 			return -ENOMEM;
5377 		}
5378 	}
5379 	ixgbevf_configure_msix(dev);
5380 
5381 	/* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
5382 	 * is mapped to VFIO vector 0 in eth_ixgbevf_dev_init( ).
5383 	 * If previous VFIO interrupt mapping setting in eth_ixgbevf_dev_init( )
5384 	 * is not cleared, it will fail when following rte_intr_enable( ) tries
5385 	 * to map Rx queue interrupt to other VFIO vectors.
5386 	 * So clear uio/vfio intr/evevnfd first to avoid failure.
5387 	 */
5388 	rte_intr_disable(intr_handle);
5389 
5390 	rte_intr_enable(intr_handle);
5391 
5392 	/* Re-enable interrupt for VF */
5393 	ixgbevf_intr_enable(dev);
5394 
5395 	/*
5396 	 * Update link status right before return, because it may
5397 	 * start link configuration process in a separate thread.
5398 	 */
5399 	ixgbevf_dev_link_update(dev, 0);
5400 
5401 	hw->adapter_stopped = false;
5402 
5403 	return 0;
5404 }
5405 
5406 static int
5407 ixgbevf_dev_stop(struct rte_eth_dev *dev)
5408 {
5409 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5410 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5411 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5412 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5413 
5414 	if (hw->adapter_stopped)
5415 		return 0;
5416 
5417 	PMD_INIT_FUNC_TRACE();
5418 
5419 	ixgbe_dev_wait_setup_link_complete(dev, 0);
5420 
5421 	ixgbevf_intr_disable(dev);
5422 
5423 	dev->data->dev_started = 0;
5424 	hw->adapter_stopped = 1;
5425 	ixgbe_stop_adapter(hw);
5426 
5427 	/*
5428 	  * Clear what we set, but we still keep shadow_vfta to
5429 	  * restore after device starts
5430 	  */
5431 	ixgbevf_set_vfta_all(dev, 0);
5432 
5433 	/* Clear stored conf */
5434 	dev->data->scattered_rx = 0;
5435 
5436 	ixgbe_dev_clear_queues(dev);
5437 
5438 	/* Clean datapath event and queue/vec mapping */
5439 	rte_intr_efd_disable(intr_handle);
5440 	if (intr_handle->intr_vec != NULL) {
5441 		rte_free(intr_handle->intr_vec);
5442 		intr_handle->intr_vec = NULL;
5443 	}
5444 
5445 	adapter->rss_reta_updated = 0;
5446 
5447 	return 0;
5448 }
5449 
5450 static int
5451 ixgbevf_dev_close(struct rte_eth_dev *dev)
5452 {
5453 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5454 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5455 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5456 	int ret;
5457 
5458 	PMD_INIT_FUNC_TRACE();
5459 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5460 		return 0;
5461 
5462 	ixgbe_reset_hw(hw);
5463 
5464 	ret = ixgbevf_dev_stop(dev);
5465 
5466 	ixgbe_dev_free_queues(dev);
5467 
5468 	/**
5469 	 * Remove the VF MAC address ro ensure
5470 	 * that the VF traffic goes to the PF
5471 	 * after stop, close and detach of the VF
5472 	 **/
5473 	ixgbevf_remove_mac_addr(dev, 0);
5474 
5475 	rte_intr_disable(intr_handle);
5476 	rte_intr_callback_unregister(intr_handle,
5477 				     ixgbevf_dev_interrupt_handler, dev);
5478 
5479 	return ret;
5480 }
5481 
5482 /*
5483  * Reset VF device
5484  */
5485 static int
5486 ixgbevf_dev_reset(struct rte_eth_dev *dev)
5487 {
5488 	int ret;
5489 
5490 	ret = eth_ixgbevf_dev_uninit(dev);
5491 	if (ret)
5492 		return ret;
5493 
5494 	ret = eth_ixgbevf_dev_init(dev);
5495 
5496 	return ret;
5497 }
5498 
5499 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
5500 {
5501 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5502 	struct ixgbe_vfta *shadow_vfta =
5503 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5504 	int i = 0, j = 0, vfta = 0, mask = 1;
5505 
5506 	for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
5507 		vfta = shadow_vfta->vfta[i];
5508 		if (vfta) {
5509 			mask = 1;
5510 			for (j = 0; j < 32; j++) {
5511 				if (vfta & mask)
5512 					ixgbe_set_vfta(hw, (i<<5)+j, 0,
5513 						       on, false);
5514 				mask <<= 1;
5515 			}
5516 		}
5517 	}
5518 
5519 }
5520 
5521 static int
5522 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
5523 {
5524 	struct ixgbe_hw *hw =
5525 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5526 	struct ixgbe_vfta *shadow_vfta =
5527 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5528 	uint32_t vid_idx = 0;
5529 	uint32_t vid_bit = 0;
5530 	int ret = 0;
5531 
5532 	PMD_INIT_FUNC_TRACE();
5533 
5534 	/* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
5535 	ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
5536 	if (ret) {
5537 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
5538 		return ret;
5539 	}
5540 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
5541 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
5542 
5543 	/* Save what we set and retore it after device reset */
5544 	if (on)
5545 		shadow_vfta->vfta[vid_idx] |= vid_bit;
5546 	else
5547 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
5548 
5549 	return 0;
5550 }
5551 
5552 static void
5553 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
5554 {
5555 	struct ixgbe_hw *hw =
5556 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5557 	uint32_t ctrl;
5558 
5559 	PMD_INIT_FUNC_TRACE();
5560 
5561 	if (queue >= hw->mac.max_rx_queues)
5562 		return;
5563 
5564 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
5565 	if (on)
5566 		ctrl |= IXGBE_RXDCTL_VME;
5567 	else
5568 		ctrl &= ~IXGBE_RXDCTL_VME;
5569 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
5570 
5571 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
5572 }
5573 
5574 static int
5575 ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask)
5576 {
5577 	struct ixgbe_rx_queue *rxq;
5578 	uint16_t i;
5579 	int on = 0;
5580 
5581 	/* VF function only support hw strip feature, others are not support */
5582 	if (mask & ETH_VLAN_STRIP_MASK) {
5583 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
5584 			rxq = dev->data->rx_queues[i];
5585 			on = !!(rxq->offloads &	DEV_RX_OFFLOAD_VLAN_STRIP);
5586 			ixgbevf_vlan_strip_queue_set(dev, i, on);
5587 		}
5588 	}
5589 
5590 	return 0;
5591 }
5592 
5593 static int
5594 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
5595 {
5596 	ixgbe_config_vlan_strip_on_all_queues(dev, mask);
5597 
5598 	ixgbevf_vlan_offload_config(dev, mask);
5599 
5600 	return 0;
5601 }
5602 
5603 int
5604 ixgbe_vt_check(struct ixgbe_hw *hw)
5605 {
5606 	uint32_t reg_val;
5607 
5608 	/* if Virtualization Technology is enabled */
5609 	reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
5610 	if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
5611 		PMD_INIT_LOG(ERR, "VT must be enabled for this setting");
5612 		return -1;
5613 	}
5614 
5615 	return 0;
5616 }
5617 
5618 static uint32_t
5619 ixgbe_uta_vector(struct ixgbe_hw *hw, struct rte_ether_addr *uc_addr)
5620 {
5621 	uint32_t vector = 0;
5622 
5623 	switch (hw->mac.mc_filter_type) {
5624 	case 0:   /* use bits [47:36] of the address */
5625 		vector = ((uc_addr->addr_bytes[4] >> 4) |
5626 			(((uint16_t)uc_addr->addr_bytes[5]) << 4));
5627 		break;
5628 	case 1:   /* use bits [46:35] of the address */
5629 		vector = ((uc_addr->addr_bytes[4] >> 3) |
5630 			(((uint16_t)uc_addr->addr_bytes[5]) << 5));
5631 		break;
5632 	case 2:   /* use bits [45:34] of the address */
5633 		vector = ((uc_addr->addr_bytes[4] >> 2) |
5634 			(((uint16_t)uc_addr->addr_bytes[5]) << 6));
5635 		break;
5636 	case 3:   /* use bits [43:32] of the address */
5637 		vector = ((uc_addr->addr_bytes[4]) |
5638 			(((uint16_t)uc_addr->addr_bytes[5]) << 8));
5639 		break;
5640 	default:  /* Invalid mc_filter_type */
5641 		break;
5642 	}
5643 
5644 	/* vector can only be 12-bits or boundary will be exceeded */
5645 	vector &= 0xFFF;
5646 	return vector;
5647 }
5648 
5649 static int
5650 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev,
5651 			struct rte_ether_addr *mac_addr, uint8_t on)
5652 {
5653 	uint32_t vector;
5654 	uint32_t uta_idx;
5655 	uint32_t reg_val;
5656 	uint32_t uta_shift;
5657 	uint32_t rc;
5658 	const uint32_t ixgbe_uta_idx_mask = 0x7F;
5659 	const uint32_t ixgbe_uta_bit_shift = 5;
5660 	const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
5661 	const uint32_t bit1 = 0x1;
5662 
5663 	struct ixgbe_hw *hw =
5664 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5665 	struct ixgbe_uta_info *uta_info =
5666 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5667 
5668 	/* The UTA table only exists on 82599 hardware and newer */
5669 	if (hw->mac.type < ixgbe_mac_82599EB)
5670 		return -ENOTSUP;
5671 
5672 	vector = ixgbe_uta_vector(hw, mac_addr);
5673 	uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
5674 	uta_shift = vector & ixgbe_uta_bit_mask;
5675 
5676 	rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
5677 	if (rc == on)
5678 		return 0;
5679 
5680 	reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
5681 	if (on) {
5682 		uta_info->uta_in_use++;
5683 		reg_val |= (bit1 << uta_shift);
5684 		uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
5685 	} else {
5686 		uta_info->uta_in_use--;
5687 		reg_val &= ~(bit1 << uta_shift);
5688 		uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
5689 	}
5690 
5691 	IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
5692 
5693 	if (uta_info->uta_in_use > 0)
5694 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
5695 				IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
5696 	else
5697 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
5698 
5699 	return 0;
5700 }
5701 
5702 static int
5703 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
5704 {
5705 	int i;
5706 	struct ixgbe_hw *hw =
5707 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5708 	struct ixgbe_uta_info *uta_info =
5709 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5710 
5711 	/* The UTA table only exists on 82599 hardware and newer */
5712 	if (hw->mac.type < ixgbe_mac_82599EB)
5713 		return -ENOTSUP;
5714 
5715 	if (on) {
5716 		for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5717 			uta_info->uta_shadow[i] = ~0;
5718 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
5719 		}
5720 	} else {
5721 		for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5722 			uta_info->uta_shadow[i] = 0;
5723 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
5724 		}
5725 	}
5726 	return 0;
5727 
5728 }
5729 
5730 uint32_t
5731 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
5732 {
5733 	uint32_t new_val = orig_val;
5734 
5735 	if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG)
5736 		new_val |= IXGBE_VMOLR_AUPE;
5737 	if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC)
5738 		new_val |= IXGBE_VMOLR_ROMPE;
5739 	if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
5740 		new_val |= IXGBE_VMOLR_ROPE;
5741 	if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
5742 		new_val |= IXGBE_VMOLR_BAM;
5743 	if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
5744 		new_val |= IXGBE_VMOLR_MPE;
5745 
5746 	return new_val;
5747 }
5748 
5749 static int
5750 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5751 {
5752 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5753 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5754 	struct ixgbe_interrupt *intr =
5755 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5756 	struct ixgbe_hw *hw =
5757 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5758 	uint32_t vec = IXGBE_MISC_VEC_ID;
5759 
5760 	if (rte_intr_allow_others(intr_handle))
5761 		vec = IXGBE_RX_VEC_START;
5762 	intr->mask |= (1 << vec);
5763 	RTE_SET_USED(queue_id);
5764 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5765 
5766 	rte_intr_ack(intr_handle);
5767 
5768 	return 0;
5769 }
5770 
5771 static int
5772 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5773 {
5774 	struct ixgbe_interrupt *intr =
5775 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5776 	struct ixgbe_hw *hw =
5777 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5778 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5779 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5780 	uint32_t vec = IXGBE_MISC_VEC_ID;
5781 
5782 	if (rte_intr_allow_others(intr_handle))
5783 		vec = IXGBE_RX_VEC_START;
5784 	intr->mask &= ~(1 << vec);
5785 	RTE_SET_USED(queue_id);
5786 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5787 
5788 	return 0;
5789 }
5790 
5791 static int
5792 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5793 {
5794 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5795 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5796 	uint32_t mask;
5797 	struct ixgbe_hw *hw =
5798 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5799 	struct ixgbe_interrupt *intr =
5800 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5801 
5802 	if (queue_id < 16) {
5803 		ixgbe_disable_intr(hw);
5804 		intr->mask |= (1 << queue_id);
5805 		ixgbe_enable_intr(dev);
5806 	} else if (queue_id < 32) {
5807 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5808 		mask &= (1 << queue_id);
5809 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5810 	} else if (queue_id < 64) {
5811 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5812 		mask &= (1 << (queue_id - 32));
5813 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5814 	}
5815 	rte_intr_ack(intr_handle);
5816 
5817 	return 0;
5818 }
5819 
5820 static int
5821 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5822 {
5823 	uint32_t mask;
5824 	struct ixgbe_hw *hw =
5825 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5826 	struct ixgbe_interrupt *intr =
5827 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5828 
5829 	if (queue_id < 16) {
5830 		ixgbe_disable_intr(hw);
5831 		intr->mask &= ~(1 << queue_id);
5832 		ixgbe_enable_intr(dev);
5833 	} else if (queue_id < 32) {
5834 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5835 		mask &= ~(1 << queue_id);
5836 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5837 	} else if (queue_id < 64) {
5838 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5839 		mask &= ~(1 << (queue_id - 32));
5840 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5841 	}
5842 
5843 	return 0;
5844 }
5845 
5846 static void
5847 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5848 		     uint8_t queue, uint8_t msix_vector)
5849 {
5850 	uint32_t tmp, idx;
5851 
5852 	if (direction == -1) {
5853 		/* other causes */
5854 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5855 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
5856 		tmp &= ~0xFF;
5857 		tmp |= msix_vector;
5858 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
5859 	} else {
5860 		/* rx or tx cause */
5861 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5862 		idx = ((16 * (queue & 1)) + (8 * direction));
5863 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
5864 		tmp &= ~(0xFF << idx);
5865 		tmp |= (msix_vector << idx);
5866 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
5867 	}
5868 }
5869 
5870 /**
5871  * set the IVAR registers, mapping interrupt causes to vectors
5872  * @param hw
5873  *  pointer to ixgbe_hw struct
5874  * @direction
5875  *  0 for Rx, 1 for Tx, -1 for other causes
5876  * @queue
5877  *  queue to map the corresponding interrupt to
5878  * @msix_vector
5879  *  the vector to map to the corresponding queue
5880  */
5881 static void
5882 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5883 		   uint8_t queue, uint8_t msix_vector)
5884 {
5885 	uint32_t tmp, idx;
5886 
5887 	msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5888 	if (hw->mac.type == ixgbe_mac_82598EB) {
5889 		if (direction == -1)
5890 			direction = 0;
5891 		idx = (((direction * 64) + queue) >> 2) & 0x1F;
5892 		tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
5893 		tmp &= ~(0xFF << (8 * (queue & 0x3)));
5894 		tmp |= (msix_vector << (8 * (queue & 0x3)));
5895 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
5896 	} else if ((hw->mac.type == ixgbe_mac_82599EB) ||
5897 			(hw->mac.type == ixgbe_mac_X540) ||
5898 			(hw->mac.type == ixgbe_mac_X550) ||
5899 			(hw->mac.type == ixgbe_mac_X550EM_x)) {
5900 		if (direction == -1) {
5901 			/* other causes */
5902 			idx = ((queue & 1) * 8);
5903 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
5904 			tmp &= ~(0xFF << idx);
5905 			tmp |= (msix_vector << idx);
5906 			IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
5907 		} else {
5908 			/* rx or tx causes */
5909 			idx = ((16 * (queue & 1)) + (8 * direction));
5910 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
5911 			tmp &= ~(0xFF << idx);
5912 			tmp |= (msix_vector << idx);
5913 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
5914 		}
5915 	}
5916 }
5917 
5918 static void
5919 ixgbevf_configure_msix(struct rte_eth_dev *dev)
5920 {
5921 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5922 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5923 	struct ixgbe_hw *hw =
5924 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5925 	uint32_t q_idx;
5926 	uint32_t vector_idx = IXGBE_MISC_VEC_ID;
5927 	uint32_t base = IXGBE_MISC_VEC_ID;
5928 
5929 	/* Configure VF other cause ivar */
5930 	ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
5931 
5932 	/* won't configure msix register if no mapping is done
5933 	 * between intr vector and event fd.
5934 	 */
5935 	if (!rte_intr_dp_is_en(intr_handle))
5936 		return;
5937 
5938 	if (rte_intr_allow_others(intr_handle)) {
5939 		base = IXGBE_RX_VEC_START;
5940 		vector_idx = IXGBE_RX_VEC_START;
5941 	}
5942 
5943 	/* Configure all RX queues of VF */
5944 	for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
5945 		/* Force all queue use vector 0,
5946 		 * as IXGBE_VF_MAXMSIVECOTR = 1
5947 		 */
5948 		ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
5949 		intr_handle->intr_vec[q_idx] = vector_idx;
5950 		if (vector_idx < base + intr_handle->nb_efd - 1)
5951 			vector_idx++;
5952 	}
5953 
5954 	/* As RX queue setting above show, all queues use the vector 0.
5955 	 * Set only the ITR value of IXGBE_MISC_VEC_ID.
5956 	 */
5957 	IXGBE_WRITE_REG(hw, IXGBE_VTEITR(IXGBE_MISC_VEC_ID),
5958 			IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
5959 			| IXGBE_EITR_CNT_WDIS);
5960 }
5961 
5962 /**
5963  * Sets up the hardware to properly generate MSI-X interrupts
5964  * @hw
5965  *  board private structure
5966  */
5967 static void
5968 ixgbe_configure_msix(struct rte_eth_dev *dev)
5969 {
5970 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5971 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5972 	struct ixgbe_hw *hw =
5973 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5974 	uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
5975 	uint32_t vec = IXGBE_MISC_VEC_ID;
5976 	uint32_t mask;
5977 	uint32_t gpie;
5978 
5979 	/* won't configure msix register if no mapping is done
5980 	 * between intr vector and event fd
5981 	 * but if misx has been enabled already, need to configure
5982 	 * auto clean, auto mask and throttling.
5983 	 */
5984 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5985 	if (!rte_intr_dp_is_en(intr_handle) &&
5986 	    !(gpie & (IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT)))
5987 		return;
5988 
5989 	if (rte_intr_allow_others(intr_handle))
5990 		vec = base = IXGBE_RX_VEC_START;
5991 
5992 	/* setup GPIE for MSI-x mode */
5993 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5994 	gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5995 		IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
5996 	/* auto clearing and auto setting corresponding bits in EIMS
5997 	 * when MSI-X interrupt is triggered
5998 	 */
5999 	if (hw->mac.type == ixgbe_mac_82598EB) {
6000 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
6001 	} else {
6002 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
6003 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
6004 	}
6005 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
6006 
6007 	/* Populate the IVAR table and set the ITR values to the
6008 	 * corresponding register.
6009 	 */
6010 	if (rte_intr_dp_is_en(intr_handle)) {
6011 		for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
6012 			queue_id++) {
6013 			/* by default, 1:1 mapping */
6014 			ixgbe_set_ivar_map(hw, 0, queue_id, vec);
6015 			intr_handle->intr_vec[queue_id] = vec;
6016 			if (vec < base + intr_handle->nb_efd - 1)
6017 				vec++;
6018 		}
6019 
6020 		switch (hw->mac.type) {
6021 		case ixgbe_mac_82598EB:
6022 			ixgbe_set_ivar_map(hw, -1,
6023 					   IXGBE_IVAR_OTHER_CAUSES_INDEX,
6024 					   IXGBE_MISC_VEC_ID);
6025 			break;
6026 		case ixgbe_mac_82599EB:
6027 		case ixgbe_mac_X540:
6028 		case ixgbe_mac_X550:
6029 		case ixgbe_mac_X550EM_x:
6030 			ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
6031 			break;
6032 		default:
6033 			break;
6034 		}
6035 	}
6036 	IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
6037 			IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
6038 			| IXGBE_EITR_CNT_WDIS);
6039 
6040 	/* set up to autoclear timer, and the vectors */
6041 	mask = IXGBE_EIMS_ENABLE_MASK;
6042 	mask &= ~(IXGBE_EIMS_OTHER |
6043 		  IXGBE_EIMS_MAILBOX |
6044 		  IXGBE_EIMS_LSC);
6045 
6046 	IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
6047 }
6048 
6049 int
6050 ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
6051 			   uint16_t queue_idx, uint16_t tx_rate)
6052 {
6053 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6054 	struct rte_eth_rxmode *rxmode;
6055 	uint32_t rf_dec, rf_int;
6056 	uint32_t bcnrc_val;
6057 	uint16_t link_speed = dev->data->dev_link.link_speed;
6058 
6059 	if (queue_idx >= hw->mac.max_tx_queues)
6060 		return -EINVAL;
6061 
6062 	if (tx_rate != 0) {
6063 		/* Calculate the rate factor values to set */
6064 		rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
6065 		rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
6066 		rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
6067 
6068 		bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
6069 		bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
6070 				IXGBE_RTTBCNRC_RF_INT_MASK_M);
6071 		bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
6072 	} else {
6073 		bcnrc_val = 0;
6074 	}
6075 
6076 	rxmode = &dev->data->dev_conf.rxmode;
6077 	/*
6078 	 * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
6079 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
6080 	 * set as 0x4.
6081 	 */
6082 	if ((rxmode->offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) &&
6083 	    (rxmode->max_rx_pkt_len >= IXGBE_MAX_JUMBO_FRAME_SIZE))
6084 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
6085 			IXGBE_MMW_SIZE_JUMBO_FRAME);
6086 	else
6087 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
6088 			IXGBE_MMW_SIZE_DEFAULT);
6089 
6090 	/* Set RTTBCNRC of queue X */
6091 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
6092 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
6093 	IXGBE_WRITE_FLUSH(hw);
6094 
6095 	return 0;
6096 }
6097 
6098 static int
6099 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
6100 		     __rte_unused uint32_t index,
6101 		     __rte_unused uint32_t pool)
6102 {
6103 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6104 	int diag;
6105 
6106 	/*
6107 	 * On a 82599 VF, adding again the same MAC addr is not an idempotent
6108 	 * operation. Trap this case to avoid exhausting the [very limited]
6109 	 * set of PF resources used to store VF MAC addresses.
6110 	 */
6111 	if (memcmp(hw->mac.perm_addr, mac_addr,
6112 			sizeof(struct rte_ether_addr)) == 0)
6113 		return -1;
6114 	diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6115 	if (diag != 0)
6116 		PMD_DRV_LOG(ERR, "Unable to add MAC address "
6117 			    RTE_ETHER_ADDR_PRT_FMT " - diag=%d",
6118 			    RTE_ETHER_ADDR_BYTES(mac_addr), diag);
6119 	return diag;
6120 }
6121 
6122 static void
6123 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
6124 {
6125 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6126 	struct rte_ether_addr *perm_addr =
6127 		(struct rte_ether_addr *)hw->mac.perm_addr;
6128 	struct rte_ether_addr *mac_addr;
6129 	uint32_t i;
6130 	int diag;
6131 
6132 	/*
6133 	 * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
6134 	 * not support the deletion of a given MAC address.
6135 	 * Instead, it imposes to delete all MAC addresses, then to add again
6136 	 * all MAC addresses with the exception of the one to be deleted.
6137 	 */
6138 	(void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
6139 
6140 	/*
6141 	 * Add again all MAC addresses, with the exception of the deleted one
6142 	 * and of the permanent MAC address.
6143 	 */
6144 	for (i = 0, mac_addr = dev->data->mac_addrs;
6145 	     i < hw->mac.num_rar_entries; i++, mac_addr++) {
6146 		/* Skip the deleted MAC address */
6147 		if (i == index)
6148 			continue;
6149 		/* Skip NULL MAC addresses */
6150 		if (rte_is_zero_ether_addr(mac_addr))
6151 			continue;
6152 		/* Skip the permanent MAC address */
6153 		if (memcmp(perm_addr, mac_addr,
6154 				sizeof(struct rte_ether_addr)) == 0)
6155 			continue;
6156 		diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6157 		if (diag != 0)
6158 			PMD_DRV_LOG(ERR,
6159 				    "Adding again MAC address "
6160 				     RTE_ETHER_ADDR_PRT_FMT " failed "
6161 				    "diag=%d", RTE_ETHER_ADDR_BYTES(mac_addr),
6162 				    diag);
6163 	}
6164 }
6165 
6166 static int
6167 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
6168 			struct rte_ether_addr *addr)
6169 {
6170 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6171 
6172 	hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
6173 
6174 	return 0;
6175 }
6176 
6177 int
6178 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
6179 			struct rte_eth_syn_filter *filter,
6180 			bool add)
6181 {
6182 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6183 	struct ixgbe_filter_info *filter_info =
6184 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6185 	uint32_t syn_info;
6186 	uint32_t synqf;
6187 
6188 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6189 		return -EINVAL;
6190 
6191 	syn_info = filter_info->syn_info;
6192 
6193 	if (add) {
6194 		if (syn_info & IXGBE_SYN_FILTER_ENABLE)
6195 			return -EINVAL;
6196 		synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
6197 			IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
6198 
6199 		if (filter->hig_pri)
6200 			synqf |= IXGBE_SYN_FILTER_SYNQFP;
6201 		else
6202 			synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
6203 	} else {
6204 		synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6205 		if (!(syn_info & IXGBE_SYN_FILTER_ENABLE))
6206 			return -ENOENT;
6207 		synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
6208 	}
6209 
6210 	filter_info->syn_info = synqf;
6211 	IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
6212 	IXGBE_WRITE_FLUSH(hw);
6213 	return 0;
6214 }
6215 
6216 
6217 static inline enum ixgbe_5tuple_protocol
6218 convert_protocol_type(uint8_t protocol_value)
6219 {
6220 	if (protocol_value == IPPROTO_TCP)
6221 		return IXGBE_FILTER_PROTOCOL_TCP;
6222 	else if (protocol_value == IPPROTO_UDP)
6223 		return IXGBE_FILTER_PROTOCOL_UDP;
6224 	else if (protocol_value == IPPROTO_SCTP)
6225 		return IXGBE_FILTER_PROTOCOL_SCTP;
6226 	else
6227 		return IXGBE_FILTER_PROTOCOL_NONE;
6228 }
6229 
6230 /* inject a 5-tuple filter to HW */
6231 static inline void
6232 ixgbe_inject_5tuple_filter(struct rte_eth_dev *dev,
6233 			   struct ixgbe_5tuple_filter *filter)
6234 {
6235 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6236 	int i;
6237 	uint32_t ftqf, sdpqf;
6238 	uint32_t l34timir = 0;
6239 	uint8_t mask = 0xff;
6240 
6241 	i = filter->index;
6242 
6243 	sdpqf = (uint32_t)(filter->filter_info.dst_port <<
6244 				IXGBE_SDPQF_DSTPORT_SHIFT);
6245 	sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
6246 
6247 	ftqf = (uint32_t)(filter->filter_info.proto &
6248 		IXGBE_FTQF_PROTOCOL_MASK);
6249 	ftqf |= (uint32_t)((filter->filter_info.priority &
6250 		IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
6251 	if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
6252 		mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
6253 	if (filter->filter_info.dst_ip_mask == 0)
6254 		mask &= IXGBE_FTQF_DEST_ADDR_MASK;
6255 	if (filter->filter_info.src_port_mask == 0)
6256 		mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
6257 	if (filter->filter_info.dst_port_mask == 0)
6258 		mask &= IXGBE_FTQF_DEST_PORT_MASK;
6259 	if (filter->filter_info.proto_mask == 0)
6260 		mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
6261 	ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
6262 	ftqf |= IXGBE_FTQF_POOL_MASK_EN;
6263 	ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
6264 
6265 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
6266 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
6267 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
6268 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
6269 
6270 	l34timir |= IXGBE_L34T_IMIR_RESERVE;
6271 	l34timir |= (uint32_t)(filter->queue <<
6272 				IXGBE_L34T_IMIR_QUEUE_SHIFT);
6273 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
6274 }
6275 
6276 /*
6277  * add a 5tuple filter
6278  *
6279  * @param
6280  * dev: Pointer to struct rte_eth_dev.
6281  * index: the index the filter allocates.
6282  * filter: ponter to the filter that will be added.
6283  * rx_queue: the queue id the filter assigned to.
6284  *
6285  * @return
6286  *    - On success, zero.
6287  *    - On failure, a negative value.
6288  */
6289 static int
6290 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
6291 			struct ixgbe_5tuple_filter *filter)
6292 {
6293 	struct ixgbe_filter_info *filter_info =
6294 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6295 	int i, idx, shift;
6296 
6297 	/*
6298 	 * look for an unused 5tuple filter index,
6299 	 * and insert the filter to list.
6300 	 */
6301 	for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
6302 		idx = i / (sizeof(uint32_t) * NBBY);
6303 		shift = i % (sizeof(uint32_t) * NBBY);
6304 		if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
6305 			filter_info->fivetuple_mask[idx] |= 1 << shift;
6306 			filter->index = i;
6307 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
6308 					  filter,
6309 					  entries);
6310 			break;
6311 		}
6312 	}
6313 	if (i >= IXGBE_MAX_FTQF_FILTERS) {
6314 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
6315 		return -ENOSYS;
6316 	}
6317 
6318 	ixgbe_inject_5tuple_filter(dev, filter);
6319 
6320 	return 0;
6321 }
6322 
6323 /*
6324  * remove a 5tuple filter
6325  *
6326  * @param
6327  * dev: Pointer to struct rte_eth_dev.
6328  * filter: the pointer of the filter will be removed.
6329  */
6330 static void
6331 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
6332 			struct ixgbe_5tuple_filter *filter)
6333 {
6334 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6335 	struct ixgbe_filter_info *filter_info =
6336 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6337 	uint16_t index = filter->index;
6338 
6339 	filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
6340 				~(1 << (index % (sizeof(uint32_t) * NBBY)));
6341 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
6342 	rte_free(filter);
6343 
6344 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
6345 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
6346 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
6347 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
6348 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
6349 }
6350 
6351 static int
6352 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
6353 {
6354 	struct ixgbe_hw *hw;
6355 	uint32_t max_frame = mtu + IXGBE_ETH_OVERHEAD;
6356 	struct rte_eth_dev_data *dev_data = dev->data;
6357 
6358 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6359 
6360 	if (mtu < RTE_ETHER_MIN_MTU ||
6361 			max_frame > RTE_ETHER_MAX_JUMBO_FRAME_LEN)
6362 		return -EINVAL;
6363 
6364 	/* If device is started, refuse mtu that requires the support of
6365 	 * scattered packets when this feature has not been enabled before.
6366 	 */
6367 	if (dev_data->dev_started && !dev_data->scattered_rx &&
6368 	    (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
6369 	     dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
6370 		PMD_INIT_LOG(ERR, "Stop port first.");
6371 		return -EINVAL;
6372 	}
6373 
6374 	/*
6375 	 * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
6376 	 * request of the version 2.0 of the mailbox API.
6377 	 * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
6378 	 * of the mailbox API.
6379 	 * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
6380 	 * prior to 3.11.33 which contains the following change:
6381 	 * "ixgbe: Enable jumbo frames support w/ SR-IOV"
6382 	 */
6383 	if (ixgbevf_rlpml_set_vf(hw, max_frame))
6384 		return -EINVAL;
6385 
6386 	/* update max frame size */
6387 	dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame;
6388 	return 0;
6389 }
6390 
6391 static inline struct ixgbe_5tuple_filter *
6392 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
6393 			struct ixgbe_5tuple_filter_info *key)
6394 {
6395 	struct ixgbe_5tuple_filter *it;
6396 
6397 	TAILQ_FOREACH(it, filter_list, entries) {
6398 		if (memcmp(key, &it->filter_info,
6399 			sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
6400 			return it;
6401 		}
6402 	}
6403 	return NULL;
6404 }
6405 
6406 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
6407 static inline int
6408 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
6409 			struct ixgbe_5tuple_filter_info *filter_info)
6410 {
6411 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
6412 		filter->priority > IXGBE_5TUPLE_MAX_PRI ||
6413 		filter->priority < IXGBE_5TUPLE_MIN_PRI)
6414 		return -EINVAL;
6415 
6416 	switch (filter->dst_ip_mask) {
6417 	case UINT32_MAX:
6418 		filter_info->dst_ip_mask = 0;
6419 		filter_info->dst_ip = filter->dst_ip;
6420 		break;
6421 	case 0:
6422 		filter_info->dst_ip_mask = 1;
6423 		break;
6424 	default:
6425 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
6426 		return -EINVAL;
6427 	}
6428 
6429 	switch (filter->src_ip_mask) {
6430 	case UINT32_MAX:
6431 		filter_info->src_ip_mask = 0;
6432 		filter_info->src_ip = filter->src_ip;
6433 		break;
6434 	case 0:
6435 		filter_info->src_ip_mask = 1;
6436 		break;
6437 	default:
6438 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
6439 		return -EINVAL;
6440 	}
6441 
6442 	switch (filter->dst_port_mask) {
6443 	case UINT16_MAX:
6444 		filter_info->dst_port_mask = 0;
6445 		filter_info->dst_port = filter->dst_port;
6446 		break;
6447 	case 0:
6448 		filter_info->dst_port_mask = 1;
6449 		break;
6450 	default:
6451 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
6452 		return -EINVAL;
6453 	}
6454 
6455 	switch (filter->src_port_mask) {
6456 	case UINT16_MAX:
6457 		filter_info->src_port_mask = 0;
6458 		filter_info->src_port = filter->src_port;
6459 		break;
6460 	case 0:
6461 		filter_info->src_port_mask = 1;
6462 		break;
6463 	default:
6464 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
6465 		return -EINVAL;
6466 	}
6467 
6468 	switch (filter->proto_mask) {
6469 	case UINT8_MAX:
6470 		filter_info->proto_mask = 0;
6471 		filter_info->proto =
6472 			convert_protocol_type(filter->proto);
6473 		break;
6474 	case 0:
6475 		filter_info->proto_mask = 1;
6476 		break;
6477 	default:
6478 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
6479 		return -EINVAL;
6480 	}
6481 
6482 	filter_info->priority = (uint8_t)filter->priority;
6483 	return 0;
6484 }
6485 
6486 /*
6487  * add or delete a ntuple filter
6488  *
6489  * @param
6490  * dev: Pointer to struct rte_eth_dev.
6491  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6492  * add: if true, add filter, if false, remove filter
6493  *
6494  * @return
6495  *    - On success, zero.
6496  *    - On failure, a negative value.
6497  */
6498 int
6499 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
6500 			struct rte_eth_ntuple_filter *ntuple_filter,
6501 			bool add)
6502 {
6503 	struct ixgbe_filter_info *filter_info =
6504 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6505 	struct ixgbe_5tuple_filter_info filter_5tuple;
6506 	struct ixgbe_5tuple_filter *filter;
6507 	int ret;
6508 
6509 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6510 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6511 		return -EINVAL;
6512 	}
6513 
6514 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6515 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6516 	if (ret < 0)
6517 		return ret;
6518 
6519 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6520 					 &filter_5tuple);
6521 	if (filter != NULL && add) {
6522 		PMD_DRV_LOG(ERR, "filter exists.");
6523 		return -EEXIST;
6524 	}
6525 	if (filter == NULL && !add) {
6526 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
6527 		return -ENOENT;
6528 	}
6529 
6530 	if (add) {
6531 		filter = rte_zmalloc("ixgbe_5tuple_filter",
6532 				sizeof(struct ixgbe_5tuple_filter), 0);
6533 		if (filter == NULL)
6534 			return -ENOMEM;
6535 		rte_memcpy(&filter->filter_info,
6536 				 &filter_5tuple,
6537 				 sizeof(struct ixgbe_5tuple_filter_info));
6538 		filter->queue = ntuple_filter->queue;
6539 		ret = ixgbe_add_5tuple_filter(dev, filter);
6540 		if (ret < 0) {
6541 			rte_free(filter);
6542 			return ret;
6543 		}
6544 	} else
6545 		ixgbe_remove_5tuple_filter(dev, filter);
6546 
6547 	return 0;
6548 }
6549 
6550 int
6551 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
6552 			struct rte_eth_ethertype_filter *filter,
6553 			bool add)
6554 {
6555 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6556 	struct ixgbe_filter_info *filter_info =
6557 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6558 	uint32_t etqf = 0;
6559 	uint32_t etqs = 0;
6560 	int ret;
6561 	struct ixgbe_ethertype_filter ethertype_filter;
6562 
6563 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6564 		return -EINVAL;
6565 
6566 	if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
6567 		filter->ether_type == RTE_ETHER_TYPE_IPV6) {
6568 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
6569 			" ethertype filter.", filter->ether_type);
6570 		return -EINVAL;
6571 	}
6572 
6573 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
6574 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
6575 		return -EINVAL;
6576 	}
6577 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
6578 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
6579 		return -EINVAL;
6580 	}
6581 
6582 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6583 	if (ret >= 0 && add) {
6584 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
6585 			    filter->ether_type);
6586 		return -EEXIST;
6587 	}
6588 	if (ret < 0 && !add) {
6589 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6590 			    filter->ether_type);
6591 		return -ENOENT;
6592 	}
6593 
6594 	if (add) {
6595 		etqf = IXGBE_ETQF_FILTER_EN;
6596 		etqf |= (uint32_t)filter->ether_type;
6597 		etqs |= (uint32_t)((filter->queue <<
6598 				    IXGBE_ETQS_RX_QUEUE_SHIFT) &
6599 				    IXGBE_ETQS_RX_QUEUE);
6600 		etqs |= IXGBE_ETQS_QUEUE_EN;
6601 
6602 		ethertype_filter.ethertype = filter->ether_type;
6603 		ethertype_filter.etqf = etqf;
6604 		ethertype_filter.etqs = etqs;
6605 		ethertype_filter.conf = FALSE;
6606 		ret = ixgbe_ethertype_filter_insert(filter_info,
6607 						    &ethertype_filter);
6608 		if (ret < 0) {
6609 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
6610 			return -ENOSPC;
6611 		}
6612 	} else {
6613 		ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
6614 		if (ret < 0)
6615 			return -ENOSYS;
6616 	}
6617 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
6618 	IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
6619 	IXGBE_WRITE_FLUSH(hw);
6620 
6621 	return 0;
6622 }
6623 
6624 static int
6625 ixgbe_dev_flow_ops_get(__rte_unused struct rte_eth_dev *dev,
6626 		       const struct rte_flow_ops **ops)
6627 {
6628 	*ops = &ixgbe_flow_ops;
6629 	return 0;
6630 }
6631 
6632 static u8 *
6633 ixgbe_dev_addr_list_itr(__rte_unused struct ixgbe_hw *hw,
6634 			u8 **mc_addr_ptr, u32 *vmdq)
6635 {
6636 	u8 *mc_addr;
6637 
6638 	*vmdq = 0;
6639 	mc_addr = *mc_addr_ptr;
6640 	*mc_addr_ptr = (mc_addr + sizeof(struct rte_ether_addr));
6641 	return mc_addr;
6642 }
6643 
6644 static int
6645 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
6646 			  struct rte_ether_addr *mc_addr_set,
6647 			  uint32_t nb_mc_addr)
6648 {
6649 	struct ixgbe_hw *hw;
6650 	u8 *mc_addr_list;
6651 
6652 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6653 	mc_addr_list = (u8 *)mc_addr_set;
6654 	return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
6655 					 ixgbe_dev_addr_list_itr, TRUE);
6656 }
6657 
6658 static uint64_t
6659 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
6660 {
6661 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6662 	uint64_t systime_cycles;
6663 
6664 	switch (hw->mac.type) {
6665 	case ixgbe_mac_X550:
6666 	case ixgbe_mac_X550EM_x:
6667 	case ixgbe_mac_X550EM_a:
6668 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
6669 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6670 		systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6671 				* NSEC_PER_SEC;
6672 		break;
6673 	default:
6674 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6675 		systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6676 				<< 32;
6677 	}
6678 
6679 	return systime_cycles;
6680 }
6681 
6682 static uint64_t
6683 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6684 {
6685 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6686 	uint64_t rx_tstamp_cycles;
6687 
6688 	switch (hw->mac.type) {
6689 	case ixgbe_mac_X550:
6690 	case ixgbe_mac_X550EM_x:
6691 	case ixgbe_mac_X550EM_a:
6692 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6693 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6694 		rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6695 				* NSEC_PER_SEC;
6696 		break;
6697 	default:
6698 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6699 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6700 		rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6701 				<< 32;
6702 	}
6703 
6704 	return rx_tstamp_cycles;
6705 }
6706 
6707 static uint64_t
6708 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6709 {
6710 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6711 	uint64_t tx_tstamp_cycles;
6712 
6713 	switch (hw->mac.type) {
6714 	case ixgbe_mac_X550:
6715 	case ixgbe_mac_X550EM_x:
6716 	case ixgbe_mac_X550EM_a:
6717 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6718 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6719 		tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6720 				* NSEC_PER_SEC;
6721 		break;
6722 	default:
6723 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6724 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6725 		tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6726 				<< 32;
6727 	}
6728 
6729 	return tx_tstamp_cycles;
6730 }
6731 
6732 static void
6733 ixgbe_start_timecounters(struct rte_eth_dev *dev)
6734 {
6735 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6736 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6737 	struct rte_eth_link link;
6738 	uint32_t incval = 0;
6739 	uint32_t shift = 0;
6740 
6741 	/* Get current link speed. */
6742 	ixgbe_dev_link_update(dev, 1);
6743 	rte_eth_linkstatus_get(dev, &link);
6744 
6745 	switch (link.link_speed) {
6746 	case ETH_SPEED_NUM_100M:
6747 		incval = IXGBE_INCVAL_100;
6748 		shift = IXGBE_INCVAL_SHIFT_100;
6749 		break;
6750 	case ETH_SPEED_NUM_1G:
6751 		incval = IXGBE_INCVAL_1GB;
6752 		shift = IXGBE_INCVAL_SHIFT_1GB;
6753 		break;
6754 	case ETH_SPEED_NUM_10G:
6755 	default:
6756 		incval = IXGBE_INCVAL_10GB;
6757 		shift = IXGBE_INCVAL_SHIFT_10GB;
6758 		break;
6759 	}
6760 
6761 	switch (hw->mac.type) {
6762 	case ixgbe_mac_X550:
6763 	case ixgbe_mac_X550EM_x:
6764 	case ixgbe_mac_X550EM_a:
6765 		/* Independent of link speed. */
6766 		incval = 1;
6767 		/* Cycles read will be interpreted as ns. */
6768 		shift = 0;
6769 		/* Fall-through */
6770 	case ixgbe_mac_X540:
6771 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
6772 		break;
6773 	case ixgbe_mac_82599EB:
6774 		incval >>= IXGBE_INCVAL_SHIFT_82599;
6775 		shift -= IXGBE_INCVAL_SHIFT_82599;
6776 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
6777 				(1 << IXGBE_INCPER_SHIFT_82599) | incval);
6778 		break;
6779 	default:
6780 		/* Not supported. */
6781 		return;
6782 	}
6783 
6784 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
6785 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6786 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6787 
6788 	adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6789 	adapter->systime_tc.cc_shift = shift;
6790 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
6791 
6792 	adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6793 	adapter->rx_tstamp_tc.cc_shift = shift;
6794 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6795 
6796 	adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6797 	adapter->tx_tstamp_tc.cc_shift = shift;
6798 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6799 }
6800 
6801 static int
6802 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
6803 {
6804 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6805 
6806 	adapter->systime_tc.nsec += delta;
6807 	adapter->rx_tstamp_tc.nsec += delta;
6808 	adapter->tx_tstamp_tc.nsec += delta;
6809 
6810 	return 0;
6811 }
6812 
6813 static int
6814 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
6815 {
6816 	uint64_t ns;
6817 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6818 
6819 	ns = rte_timespec_to_ns(ts);
6820 	/* Set the timecounters to a new value. */
6821 	adapter->systime_tc.nsec = ns;
6822 	adapter->rx_tstamp_tc.nsec = ns;
6823 	adapter->tx_tstamp_tc.nsec = ns;
6824 
6825 	return 0;
6826 }
6827 
6828 static int
6829 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
6830 {
6831 	uint64_t ns, systime_cycles;
6832 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6833 
6834 	systime_cycles = ixgbe_read_systime_cyclecounter(dev);
6835 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
6836 	*ts = rte_ns_to_timespec(ns);
6837 
6838 	return 0;
6839 }
6840 
6841 static int
6842 ixgbe_timesync_enable(struct rte_eth_dev *dev)
6843 {
6844 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6845 	uint32_t tsync_ctl;
6846 	uint32_t tsauxc;
6847 
6848 	/* Stop the timesync system time. */
6849 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
6850 	/* Reset the timesync system time value. */
6851 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
6852 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
6853 
6854 	/* Enable system time for platforms where it isn't on by default. */
6855 	tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
6856 	tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
6857 	IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
6858 
6859 	ixgbe_start_timecounters(dev);
6860 
6861 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6862 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
6863 			(RTE_ETHER_TYPE_1588 |
6864 			 IXGBE_ETQF_FILTER_EN |
6865 			 IXGBE_ETQF_1588));
6866 
6867 	/* Enable timestamping of received PTP packets. */
6868 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6869 	tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
6870 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6871 
6872 	/* Enable timestamping of transmitted PTP packets. */
6873 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6874 	tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
6875 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6876 
6877 	IXGBE_WRITE_FLUSH(hw);
6878 
6879 	return 0;
6880 }
6881 
6882 static int
6883 ixgbe_timesync_disable(struct rte_eth_dev *dev)
6884 {
6885 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6886 	uint32_t tsync_ctl;
6887 
6888 	/* Disable timestamping of transmitted PTP packets. */
6889 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6890 	tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
6891 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6892 
6893 	/* Disable timestamping of received PTP packets. */
6894 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6895 	tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
6896 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6897 
6898 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6899 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
6900 
6901 	/* Stop incrementating the System Time registers. */
6902 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
6903 
6904 	return 0;
6905 }
6906 
6907 static int
6908 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
6909 				 struct timespec *timestamp,
6910 				 uint32_t flags __rte_unused)
6911 {
6912 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6913 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6914 	uint32_t tsync_rxctl;
6915 	uint64_t rx_tstamp_cycles;
6916 	uint64_t ns;
6917 
6918 	tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6919 	if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
6920 		return -EINVAL;
6921 
6922 	rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
6923 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
6924 	*timestamp = rte_ns_to_timespec(ns);
6925 
6926 	return  0;
6927 }
6928 
6929 static int
6930 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
6931 				 struct timespec *timestamp)
6932 {
6933 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6934 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6935 	uint32_t tsync_txctl;
6936 	uint64_t tx_tstamp_cycles;
6937 	uint64_t ns;
6938 
6939 	tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6940 	if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
6941 		return -EINVAL;
6942 
6943 	tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
6944 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
6945 	*timestamp = rte_ns_to_timespec(ns);
6946 
6947 	return 0;
6948 }
6949 
6950 static int
6951 ixgbe_get_reg_length(struct rte_eth_dev *dev)
6952 {
6953 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6954 	int count = 0;
6955 	int g_ind = 0;
6956 	const struct reg_info *reg_group;
6957 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6958 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6959 
6960 	while ((reg_group = reg_set[g_ind++]))
6961 		count += ixgbe_regs_group_count(reg_group);
6962 
6963 	return count;
6964 }
6965 
6966 static int
6967 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
6968 {
6969 	int count = 0;
6970 	int g_ind = 0;
6971 	const struct reg_info *reg_group;
6972 
6973 	while ((reg_group = ixgbevf_regs[g_ind++]))
6974 		count += ixgbe_regs_group_count(reg_group);
6975 
6976 	return count;
6977 }
6978 
6979 static int
6980 ixgbe_get_regs(struct rte_eth_dev *dev,
6981 	      struct rte_dev_reg_info *regs)
6982 {
6983 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6984 	uint32_t *data = regs->data;
6985 	int g_ind = 0;
6986 	int count = 0;
6987 	const struct reg_info *reg_group;
6988 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6989 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6990 
6991 	if (data == NULL) {
6992 		regs->length = ixgbe_get_reg_length(dev);
6993 		regs->width = sizeof(uint32_t);
6994 		return 0;
6995 	}
6996 
6997 	/* Support only full register dump */
6998 	if ((regs->length == 0) ||
6999 	    (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
7000 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7001 			hw->device_id;
7002 		while ((reg_group = reg_set[g_ind++]))
7003 			count += ixgbe_read_regs_group(dev, &data[count],
7004 				reg_group);
7005 		return 0;
7006 	}
7007 
7008 	return -ENOTSUP;
7009 }
7010 
7011 static int
7012 ixgbevf_get_regs(struct rte_eth_dev *dev,
7013 		struct rte_dev_reg_info *regs)
7014 {
7015 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7016 	uint32_t *data = regs->data;
7017 	int g_ind = 0;
7018 	int count = 0;
7019 	const struct reg_info *reg_group;
7020 
7021 	if (data == NULL) {
7022 		regs->length = ixgbevf_get_reg_length(dev);
7023 		regs->width = sizeof(uint32_t);
7024 		return 0;
7025 	}
7026 
7027 	/* Support only full register dump */
7028 	if ((regs->length == 0) ||
7029 	    (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
7030 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7031 			hw->device_id;
7032 		while ((reg_group = ixgbevf_regs[g_ind++]))
7033 			count += ixgbe_read_regs_group(dev, &data[count],
7034 						      reg_group);
7035 		return 0;
7036 	}
7037 
7038 	return -ENOTSUP;
7039 }
7040 
7041 static int
7042 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
7043 {
7044 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7045 
7046 	/* Return unit is byte count */
7047 	return hw->eeprom.word_size * 2;
7048 }
7049 
7050 static int
7051 ixgbe_get_eeprom(struct rte_eth_dev *dev,
7052 		struct rte_dev_eeprom_info *in_eeprom)
7053 {
7054 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7055 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7056 	uint16_t *data = in_eeprom->data;
7057 	int first, length;
7058 
7059 	first = in_eeprom->offset >> 1;
7060 	length = in_eeprom->length >> 1;
7061 	if ((first > hw->eeprom.word_size) ||
7062 	    ((first + length) > hw->eeprom.word_size))
7063 		return -EINVAL;
7064 
7065 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7066 
7067 	return eeprom->ops.read_buffer(hw, first, length, data);
7068 }
7069 
7070 static int
7071 ixgbe_set_eeprom(struct rte_eth_dev *dev,
7072 		struct rte_dev_eeprom_info *in_eeprom)
7073 {
7074 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7075 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7076 	uint16_t *data = in_eeprom->data;
7077 	int first, length;
7078 
7079 	first = in_eeprom->offset >> 1;
7080 	length = in_eeprom->length >> 1;
7081 	if ((first > hw->eeprom.word_size) ||
7082 	    ((first + length) > hw->eeprom.word_size))
7083 		return -EINVAL;
7084 
7085 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7086 
7087 	return eeprom->ops.write_buffer(hw,  first, length, data);
7088 }
7089 
7090 static int
7091 ixgbe_get_module_info(struct rte_eth_dev *dev,
7092 		      struct rte_eth_dev_module_info *modinfo)
7093 {
7094 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7095 	uint32_t status;
7096 	uint8_t sff8472_rev, addr_mode;
7097 	bool page_swap = false;
7098 
7099 	/* Check whether we support SFF-8472 or not */
7100 	status = hw->phy.ops.read_i2c_eeprom(hw,
7101 					     IXGBE_SFF_SFF_8472_COMP,
7102 					     &sff8472_rev);
7103 	if (status != 0)
7104 		return -EIO;
7105 
7106 	/* addressing mode is not supported */
7107 	status = hw->phy.ops.read_i2c_eeprom(hw,
7108 					     IXGBE_SFF_SFF_8472_SWAP,
7109 					     &addr_mode);
7110 	if (status != 0)
7111 		return -EIO;
7112 
7113 	if (addr_mode & IXGBE_SFF_ADDRESSING_MODE) {
7114 		PMD_DRV_LOG(ERR,
7115 			    "Address change required to access page 0xA2, "
7116 			    "but not supported. Please report the module "
7117 			    "type to the driver maintainers.");
7118 		page_swap = true;
7119 	}
7120 
7121 	if (sff8472_rev == IXGBE_SFF_SFF_8472_UNSUP || page_swap) {
7122 		/* We have a SFP, but it does not support SFF-8472 */
7123 		modinfo->type = RTE_ETH_MODULE_SFF_8079;
7124 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
7125 	} else {
7126 		/* We have a SFP which supports a revision of SFF-8472. */
7127 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
7128 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
7129 	}
7130 
7131 	return 0;
7132 }
7133 
7134 static int
7135 ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
7136 			struct rte_dev_eeprom_info *info)
7137 {
7138 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7139 	uint32_t status = IXGBE_ERR_PHY_ADDR_INVALID;
7140 	uint8_t databyte = 0xFF;
7141 	uint8_t *data = info->data;
7142 	uint32_t i = 0;
7143 
7144 	for (i = info->offset; i < info->offset + info->length; i++) {
7145 		if (i < RTE_ETH_MODULE_SFF_8079_LEN)
7146 			status = hw->phy.ops.read_i2c_eeprom(hw, i, &databyte);
7147 		else
7148 			status = hw->phy.ops.read_i2c_sff8472(hw, i, &databyte);
7149 
7150 		if (status != 0)
7151 			return -EIO;
7152 
7153 		data[i - info->offset] = databyte;
7154 	}
7155 
7156 	return 0;
7157 }
7158 
7159 uint16_t
7160 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
7161 	switch (mac_type) {
7162 	case ixgbe_mac_X550:
7163 	case ixgbe_mac_X550EM_x:
7164 	case ixgbe_mac_X550EM_a:
7165 		return ETH_RSS_RETA_SIZE_512;
7166 	case ixgbe_mac_X550_vf:
7167 	case ixgbe_mac_X550EM_x_vf:
7168 	case ixgbe_mac_X550EM_a_vf:
7169 		return ETH_RSS_RETA_SIZE_64;
7170 	case ixgbe_mac_X540_vf:
7171 	case ixgbe_mac_82599_vf:
7172 		return 0;
7173 	default:
7174 		return ETH_RSS_RETA_SIZE_128;
7175 	}
7176 }
7177 
7178 uint32_t
7179 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
7180 	switch (mac_type) {
7181 	case ixgbe_mac_X550:
7182 	case ixgbe_mac_X550EM_x:
7183 	case ixgbe_mac_X550EM_a:
7184 		if (reta_idx < ETH_RSS_RETA_SIZE_128)
7185 			return IXGBE_RETA(reta_idx >> 2);
7186 		else
7187 			return IXGBE_ERETA((reta_idx - ETH_RSS_RETA_SIZE_128) >> 2);
7188 	case ixgbe_mac_X550_vf:
7189 	case ixgbe_mac_X550EM_x_vf:
7190 	case ixgbe_mac_X550EM_a_vf:
7191 		return IXGBE_VFRETA(reta_idx >> 2);
7192 	default:
7193 		return IXGBE_RETA(reta_idx >> 2);
7194 	}
7195 }
7196 
7197 uint32_t
7198 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
7199 	switch (mac_type) {
7200 	case ixgbe_mac_X550_vf:
7201 	case ixgbe_mac_X550EM_x_vf:
7202 	case ixgbe_mac_X550EM_a_vf:
7203 		return IXGBE_VFMRQC;
7204 	default:
7205 		return IXGBE_MRQC;
7206 	}
7207 }
7208 
7209 uint32_t
7210 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
7211 	switch (mac_type) {
7212 	case ixgbe_mac_X550_vf:
7213 	case ixgbe_mac_X550EM_x_vf:
7214 	case ixgbe_mac_X550EM_a_vf:
7215 		return IXGBE_VFRSSRK(i);
7216 	default:
7217 		return IXGBE_RSSRK(i);
7218 	}
7219 }
7220 
7221 bool
7222 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
7223 	switch (mac_type) {
7224 	case ixgbe_mac_82599_vf:
7225 	case ixgbe_mac_X540_vf:
7226 		return 0;
7227 	default:
7228 		return 1;
7229 	}
7230 }
7231 
7232 static int
7233 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
7234 			struct rte_eth_dcb_info *dcb_info)
7235 {
7236 	struct ixgbe_dcb_config *dcb_config =
7237 			IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
7238 	struct ixgbe_dcb_tc_config *tc;
7239 	struct rte_eth_dcb_tc_queue_mapping *tc_queue;
7240 	uint8_t nb_tcs;
7241 	uint8_t i, j;
7242 
7243 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_DCB_FLAG)
7244 		dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
7245 	else
7246 		dcb_info->nb_tcs = 1;
7247 
7248 	tc_queue = &dcb_info->tc_queue;
7249 	nb_tcs = dcb_info->nb_tcs;
7250 
7251 	if (dcb_config->vt_mode) { /* vt is enabled*/
7252 		struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
7253 				&dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
7254 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7255 			dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
7256 		if (RTE_ETH_DEV_SRIOV(dev).active > 0) {
7257 			for (j = 0; j < nb_tcs; j++) {
7258 				tc_queue->tc_rxq[0][j].base = j;
7259 				tc_queue->tc_rxq[0][j].nb_queue = 1;
7260 				tc_queue->tc_txq[0][j].base = j;
7261 				tc_queue->tc_txq[0][j].nb_queue = 1;
7262 			}
7263 		} else {
7264 			for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
7265 				for (j = 0; j < nb_tcs; j++) {
7266 					tc_queue->tc_rxq[i][j].base =
7267 						i * nb_tcs + j;
7268 					tc_queue->tc_rxq[i][j].nb_queue = 1;
7269 					tc_queue->tc_txq[i][j].base =
7270 						i * nb_tcs + j;
7271 					tc_queue->tc_txq[i][j].nb_queue = 1;
7272 				}
7273 			}
7274 		}
7275 	} else { /* vt is disabled*/
7276 		struct rte_eth_dcb_rx_conf *rx_conf =
7277 				&dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
7278 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7279 			dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
7280 		if (dcb_info->nb_tcs == ETH_4_TCS) {
7281 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7282 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
7283 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7284 			}
7285 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7286 			dcb_info->tc_queue.tc_txq[0][1].base = 64;
7287 			dcb_info->tc_queue.tc_txq[0][2].base = 96;
7288 			dcb_info->tc_queue.tc_txq[0][3].base = 112;
7289 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
7290 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7291 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7292 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7293 		} else if (dcb_info->nb_tcs == ETH_8_TCS) {
7294 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7295 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
7296 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7297 			}
7298 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7299 			dcb_info->tc_queue.tc_txq[0][1].base = 32;
7300 			dcb_info->tc_queue.tc_txq[0][2].base = 64;
7301 			dcb_info->tc_queue.tc_txq[0][3].base = 80;
7302 			dcb_info->tc_queue.tc_txq[0][4].base = 96;
7303 			dcb_info->tc_queue.tc_txq[0][5].base = 104;
7304 			dcb_info->tc_queue.tc_txq[0][6].base = 112;
7305 			dcb_info->tc_queue.tc_txq[0][7].base = 120;
7306 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
7307 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7308 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7309 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7310 			dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
7311 			dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
7312 			dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
7313 			dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
7314 		}
7315 	}
7316 	for (i = 0; i < dcb_info->nb_tcs; i++) {
7317 		tc = &dcb_config->tc_config[i];
7318 		dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
7319 	}
7320 	return 0;
7321 }
7322 
7323 /* Update e-tag ether type */
7324 static int
7325 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
7326 			    uint16_t ether_type)
7327 {
7328 	uint32_t etag_etype;
7329 
7330 	if (hw->mac.type != ixgbe_mac_X550 &&
7331 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7332 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7333 		return -ENOTSUP;
7334 	}
7335 
7336 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7337 	etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
7338 	etag_etype |= ether_type;
7339 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7340 	IXGBE_WRITE_FLUSH(hw);
7341 
7342 	return 0;
7343 }
7344 
7345 /* Enable e-tag tunnel */
7346 static int
7347 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
7348 {
7349 	uint32_t etag_etype;
7350 
7351 	if (hw->mac.type != ixgbe_mac_X550 &&
7352 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7353 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7354 		return -ENOTSUP;
7355 	}
7356 
7357 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7358 	etag_etype |= IXGBE_ETAG_ETYPE_VALID;
7359 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7360 	IXGBE_WRITE_FLUSH(hw);
7361 
7362 	return 0;
7363 }
7364 
7365 static int
7366 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
7367 		       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7368 {
7369 	int ret = 0;
7370 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7371 	uint32_t i, rar_entries;
7372 	uint32_t rar_low, rar_high;
7373 
7374 	if (hw->mac.type != ixgbe_mac_X550 &&
7375 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7376 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7377 		return -ENOTSUP;
7378 	}
7379 
7380 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7381 
7382 	for (i = 1; i < rar_entries; i++) {
7383 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7384 		rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
7385 		if ((rar_high & IXGBE_RAH_AV) &&
7386 		    (rar_high & IXGBE_RAH_ADTYPE) &&
7387 		    ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
7388 		     l2_tunnel->tunnel_id)) {
7389 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
7390 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
7391 
7392 			ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
7393 
7394 			return ret;
7395 		}
7396 	}
7397 
7398 	return ret;
7399 }
7400 
7401 static int
7402 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
7403 		       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7404 {
7405 	int ret = 0;
7406 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7407 	uint32_t i, rar_entries;
7408 	uint32_t rar_low, rar_high;
7409 
7410 	if (hw->mac.type != ixgbe_mac_X550 &&
7411 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7412 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7413 		return -ENOTSUP;
7414 	}
7415 
7416 	/* One entry for one tunnel. Try to remove potential existing entry. */
7417 	ixgbe_e_tag_filter_del(dev, l2_tunnel);
7418 
7419 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7420 
7421 	for (i = 1; i < rar_entries; i++) {
7422 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7423 		if (rar_high & IXGBE_RAH_AV) {
7424 			continue;
7425 		} else {
7426 			ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
7427 			rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
7428 			rar_low = l2_tunnel->tunnel_id;
7429 
7430 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
7431 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
7432 
7433 			return ret;
7434 		}
7435 	}
7436 
7437 	PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
7438 		     " Please remove a rule before adding a new one.");
7439 	return -EINVAL;
7440 }
7441 
7442 static inline struct ixgbe_l2_tn_filter *
7443 ixgbe_l2_tn_filter_lookup(struct ixgbe_l2_tn_info *l2_tn_info,
7444 			  struct ixgbe_l2_tn_key *key)
7445 {
7446 	int ret;
7447 
7448 	ret = rte_hash_lookup(l2_tn_info->hash_handle, (const void *)key);
7449 	if (ret < 0)
7450 		return NULL;
7451 
7452 	return l2_tn_info->hash_map[ret];
7453 }
7454 
7455 static inline int
7456 ixgbe_insert_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7457 			  struct ixgbe_l2_tn_filter *l2_tn_filter)
7458 {
7459 	int ret;
7460 
7461 	ret = rte_hash_add_key(l2_tn_info->hash_handle,
7462 			       &l2_tn_filter->key);
7463 
7464 	if (ret < 0) {
7465 		PMD_DRV_LOG(ERR,
7466 			    "Failed to insert L2 tunnel filter"
7467 			    " to hash table %d!",
7468 			    ret);
7469 		return ret;
7470 	}
7471 
7472 	l2_tn_info->hash_map[ret] = l2_tn_filter;
7473 
7474 	TAILQ_INSERT_TAIL(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7475 
7476 	return 0;
7477 }
7478 
7479 static inline int
7480 ixgbe_remove_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7481 			  struct ixgbe_l2_tn_key *key)
7482 {
7483 	int ret;
7484 	struct ixgbe_l2_tn_filter *l2_tn_filter;
7485 
7486 	ret = rte_hash_del_key(l2_tn_info->hash_handle, key);
7487 
7488 	if (ret < 0) {
7489 		PMD_DRV_LOG(ERR,
7490 			    "No such L2 tunnel filter to delete %d!",
7491 			    ret);
7492 		return ret;
7493 	}
7494 
7495 	l2_tn_filter = l2_tn_info->hash_map[ret];
7496 	l2_tn_info->hash_map[ret] = NULL;
7497 
7498 	TAILQ_REMOVE(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7499 	rte_free(l2_tn_filter);
7500 
7501 	return 0;
7502 }
7503 
7504 /* Add l2 tunnel filter */
7505 int
7506 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
7507 			       struct ixgbe_l2_tunnel_conf *l2_tunnel,
7508 			       bool restore)
7509 {
7510 	int ret;
7511 	struct ixgbe_l2_tn_info *l2_tn_info =
7512 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7513 	struct ixgbe_l2_tn_key key;
7514 	struct ixgbe_l2_tn_filter *node;
7515 
7516 	if (!restore) {
7517 		key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7518 		key.tn_id = l2_tunnel->tunnel_id;
7519 
7520 		node = ixgbe_l2_tn_filter_lookup(l2_tn_info, &key);
7521 
7522 		if (node) {
7523 			PMD_DRV_LOG(ERR,
7524 				    "The L2 tunnel filter already exists!");
7525 			return -EINVAL;
7526 		}
7527 
7528 		node = rte_zmalloc("ixgbe_l2_tn",
7529 				   sizeof(struct ixgbe_l2_tn_filter),
7530 				   0);
7531 		if (!node)
7532 			return -ENOMEM;
7533 
7534 		rte_memcpy(&node->key,
7535 				 &key,
7536 				 sizeof(struct ixgbe_l2_tn_key));
7537 		node->pool = l2_tunnel->pool;
7538 		ret = ixgbe_insert_l2_tn_filter(l2_tn_info, node);
7539 		if (ret < 0) {
7540 			rte_free(node);
7541 			return ret;
7542 		}
7543 	}
7544 
7545 	switch (l2_tunnel->l2_tunnel_type) {
7546 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7547 		ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
7548 		break;
7549 	default:
7550 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7551 		ret = -EINVAL;
7552 		break;
7553 	}
7554 
7555 	if ((!restore) && (ret < 0))
7556 		(void)ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7557 
7558 	return ret;
7559 }
7560 
7561 /* Delete l2 tunnel filter */
7562 int
7563 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
7564 			       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7565 {
7566 	int ret;
7567 	struct ixgbe_l2_tn_info *l2_tn_info =
7568 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7569 	struct ixgbe_l2_tn_key key;
7570 
7571 	key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7572 	key.tn_id = l2_tunnel->tunnel_id;
7573 	ret = ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7574 	if (ret < 0)
7575 		return ret;
7576 
7577 	switch (l2_tunnel->l2_tunnel_type) {
7578 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7579 		ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
7580 		break;
7581 	default:
7582 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7583 		ret = -EINVAL;
7584 		break;
7585 	}
7586 
7587 	return ret;
7588 }
7589 
7590 static int
7591 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
7592 {
7593 	int ret = 0;
7594 	uint32_t ctrl;
7595 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7596 
7597 	if (hw->mac.type != ixgbe_mac_X550 &&
7598 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7599 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7600 		return -ENOTSUP;
7601 	}
7602 
7603 	ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
7604 	ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
7605 	if (en)
7606 		ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
7607 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
7608 
7609 	return ret;
7610 }
7611 
7612 static int
7613 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
7614 			uint16_t port)
7615 {
7616 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
7617 	IXGBE_WRITE_FLUSH(hw);
7618 
7619 	return 0;
7620 }
7621 
7622 /* There's only one register for VxLAN UDP port.
7623  * So, we cannot add several ports. Will update it.
7624  */
7625 static int
7626 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
7627 		     uint16_t port)
7628 {
7629 	if (port == 0) {
7630 		PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
7631 		return -EINVAL;
7632 	}
7633 
7634 	return ixgbe_update_vxlan_port(hw, port);
7635 }
7636 
7637 /* We cannot delete the VxLAN port. For there's a register for VxLAN
7638  * UDP port, it must have a value.
7639  * So, will reset it to the original value 0.
7640  */
7641 static int
7642 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
7643 		     uint16_t port)
7644 {
7645 	uint16_t cur_port;
7646 
7647 	cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
7648 
7649 	if (cur_port != port) {
7650 		PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
7651 		return -EINVAL;
7652 	}
7653 
7654 	return ixgbe_update_vxlan_port(hw, 0);
7655 }
7656 
7657 /* Add UDP tunneling port */
7658 static int
7659 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
7660 			      struct rte_eth_udp_tunnel *udp_tunnel)
7661 {
7662 	int ret = 0;
7663 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7664 
7665 	if (hw->mac.type != ixgbe_mac_X550 &&
7666 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7667 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7668 		return -ENOTSUP;
7669 	}
7670 
7671 	if (udp_tunnel == NULL)
7672 		return -EINVAL;
7673 
7674 	switch (udp_tunnel->prot_type) {
7675 	case RTE_TUNNEL_TYPE_VXLAN:
7676 		ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
7677 		break;
7678 
7679 	case RTE_TUNNEL_TYPE_GENEVE:
7680 	case RTE_TUNNEL_TYPE_TEREDO:
7681 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7682 		ret = -EINVAL;
7683 		break;
7684 
7685 	default:
7686 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7687 		ret = -EINVAL;
7688 		break;
7689 	}
7690 
7691 	return ret;
7692 }
7693 
7694 /* Remove UDP tunneling port */
7695 static int
7696 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
7697 			      struct rte_eth_udp_tunnel *udp_tunnel)
7698 {
7699 	int ret = 0;
7700 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7701 
7702 	if (hw->mac.type != ixgbe_mac_X550 &&
7703 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7704 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7705 		return -ENOTSUP;
7706 	}
7707 
7708 	if (udp_tunnel == NULL)
7709 		return -EINVAL;
7710 
7711 	switch (udp_tunnel->prot_type) {
7712 	case RTE_TUNNEL_TYPE_VXLAN:
7713 		ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
7714 		break;
7715 	case RTE_TUNNEL_TYPE_GENEVE:
7716 	case RTE_TUNNEL_TYPE_TEREDO:
7717 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7718 		ret = -EINVAL;
7719 		break;
7720 	default:
7721 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7722 		ret = -EINVAL;
7723 		break;
7724 	}
7725 
7726 	return ret;
7727 }
7728 
7729 static int
7730 ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev)
7731 {
7732 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7733 	int ret;
7734 
7735 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_PROMISC)) {
7736 	case IXGBE_SUCCESS:
7737 		ret = 0;
7738 		break;
7739 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7740 		ret = -ENOTSUP;
7741 		break;
7742 	default:
7743 		ret = -EAGAIN;
7744 		break;
7745 	}
7746 
7747 	return ret;
7748 }
7749 
7750 static int
7751 ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev)
7752 {
7753 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7754 	int ret;
7755 
7756 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_NONE)) {
7757 	case IXGBE_SUCCESS:
7758 		ret = 0;
7759 		break;
7760 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7761 		ret = -ENOTSUP;
7762 		break;
7763 	default:
7764 		ret = -EAGAIN;
7765 		break;
7766 	}
7767 
7768 	return ret;
7769 }
7770 
7771 static int
7772 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
7773 {
7774 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7775 	int ret;
7776 	int mode = IXGBEVF_XCAST_MODE_ALLMULTI;
7777 
7778 	switch (hw->mac.ops.update_xcast_mode(hw, mode)) {
7779 	case IXGBE_SUCCESS:
7780 		ret = 0;
7781 		break;
7782 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7783 		ret = -ENOTSUP;
7784 		break;
7785 	default:
7786 		ret = -EAGAIN;
7787 		break;
7788 	}
7789 
7790 	return ret;
7791 }
7792 
7793 static int
7794 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
7795 {
7796 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7797 	int ret;
7798 
7799 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_MULTI)) {
7800 	case IXGBE_SUCCESS:
7801 		ret = 0;
7802 		break;
7803 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7804 		ret = -ENOTSUP;
7805 		break;
7806 	default:
7807 		ret = -EAGAIN;
7808 		break;
7809 	}
7810 
7811 	return ret;
7812 }
7813 
7814 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
7815 {
7816 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7817 	u32 in_msg = 0;
7818 
7819 	/* peek the message first */
7820 	in_msg = IXGBE_READ_REG(hw, IXGBE_VFMBMEM);
7821 
7822 	/* PF reset VF event */
7823 	if (in_msg == IXGBE_PF_CONTROL_MSG) {
7824 		/* dummy mbx read to ack pf */
7825 		if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
7826 			return;
7827 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
7828 					     NULL);
7829 	}
7830 }
7831 
7832 static int
7833 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
7834 {
7835 	uint32_t eicr;
7836 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7837 	struct ixgbe_interrupt *intr =
7838 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7839 	ixgbevf_intr_disable(dev);
7840 
7841 	/* read-on-clear nic registers here */
7842 	eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
7843 	intr->flags = 0;
7844 
7845 	/* only one misc vector supported - mailbox */
7846 	eicr &= IXGBE_VTEICR_MASK;
7847 	if (eicr == IXGBE_MISC_VEC_ID)
7848 		intr->flags |= IXGBE_FLAG_MAILBOX;
7849 
7850 	return 0;
7851 }
7852 
7853 static int
7854 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
7855 {
7856 	struct ixgbe_interrupt *intr =
7857 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7858 
7859 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
7860 		ixgbevf_mbx_process(dev);
7861 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
7862 	}
7863 
7864 	ixgbevf_intr_enable(dev);
7865 
7866 	return 0;
7867 }
7868 
7869 static void
7870 ixgbevf_dev_interrupt_handler(void *param)
7871 {
7872 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
7873 
7874 	ixgbevf_dev_interrupt_get_status(dev);
7875 	ixgbevf_dev_interrupt_action(dev);
7876 }
7877 
7878 /**
7879  *  ixgbe_disable_sec_tx_path_generic - Stops the transmit data path
7880  *  @hw: pointer to hardware structure
7881  *
7882  *  Stops the transmit data path and waits for the HW to internally empty
7883  *  the Tx security block
7884  **/
7885 int ixgbe_disable_sec_tx_path_generic(struct ixgbe_hw *hw)
7886 {
7887 #define IXGBE_MAX_SECTX_POLL 40
7888 
7889 	int i;
7890 	int sectxreg;
7891 
7892 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
7893 	sectxreg |= IXGBE_SECTXCTRL_TX_DIS;
7894 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
7895 	for (i = 0; i < IXGBE_MAX_SECTX_POLL; i++) {
7896 		sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXSTAT);
7897 		if (sectxreg & IXGBE_SECTXSTAT_SECTX_RDY)
7898 			break;
7899 		/* Use interrupt-safe sleep just in case */
7900 		usec_delay(1000);
7901 	}
7902 
7903 	/* For informational purposes only */
7904 	if (i >= IXGBE_MAX_SECTX_POLL)
7905 		PMD_DRV_LOG(DEBUG, "Tx unit being enabled before security "
7906 			 "path fully disabled.  Continuing with init.");
7907 
7908 	return IXGBE_SUCCESS;
7909 }
7910 
7911 /**
7912  *  ixgbe_enable_sec_tx_path_generic - Enables the transmit data path
7913  *  @hw: pointer to hardware structure
7914  *
7915  *  Enables the transmit data path.
7916  **/
7917 int ixgbe_enable_sec_tx_path_generic(struct ixgbe_hw *hw)
7918 {
7919 	uint32_t sectxreg;
7920 
7921 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
7922 	sectxreg &= ~IXGBE_SECTXCTRL_TX_DIS;
7923 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
7924 	IXGBE_WRITE_FLUSH(hw);
7925 
7926 	return IXGBE_SUCCESS;
7927 }
7928 
7929 /* restore n-tuple filter */
7930 static inline void
7931 ixgbe_ntuple_filter_restore(struct rte_eth_dev *dev)
7932 {
7933 	struct ixgbe_filter_info *filter_info =
7934 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7935 	struct ixgbe_5tuple_filter *node;
7936 
7937 	TAILQ_FOREACH(node, &filter_info->fivetuple_list, entries) {
7938 		ixgbe_inject_5tuple_filter(dev, node);
7939 	}
7940 }
7941 
7942 /* restore ethernet type filter */
7943 static inline void
7944 ixgbe_ethertype_filter_restore(struct rte_eth_dev *dev)
7945 {
7946 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7947 	struct ixgbe_filter_info *filter_info =
7948 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7949 	int i;
7950 
7951 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
7952 		if (filter_info->ethertype_mask & (1 << i)) {
7953 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i),
7954 					filter_info->ethertype_filters[i].etqf);
7955 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i),
7956 					filter_info->ethertype_filters[i].etqs);
7957 			IXGBE_WRITE_FLUSH(hw);
7958 		}
7959 	}
7960 }
7961 
7962 /* restore SYN filter */
7963 static inline void
7964 ixgbe_syn_filter_restore(struct rte_eth_dev *dev)
7965 {
7966 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7967 	struct ixgbe_filter_info *filter_info =
7968 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7969 	uint32_t synqf;
7970 
7971 	synqf = filter_info->syn_info;
7972 
7973 	if (synqf & IXGBE_SYN_FILTER_ENABLE) {
7974 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
7975 		IXGBE_WRITE_FLUSH(hw);
7976 	}
7977 }
7978 
7979 /* restore L2 tunnel filter */
7980 static inline void
7981 ixgbe_l2_tn_filter_restore(struct rte_eth_dev *dev)
7982 {
7983 	struct ixgbe_l2_tn_info *l2_tn_info =
7984 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7985 	struct ixgbe_l2_tn_filter *node;
7986 	struct ixgbe_l2_tunnel_conf l2_tn_conf;
7987 
7988 	TAILQ_FOREACH(node, &l2_tn_info->l2_tn_list, entries) {
7989 		l2_tn_conf.l2_tunnel_type = node->key.l2_tn_type;
7990 		l2_tn_conf.tunnel_id      = node->key.tn_id;
7991 		l2_tn_conf.pool           = node->pool;
7992 		(void)ixgbe_dev_l2_tunnel_filter_add(dev, &l2_tn_conf, TRUE);
7993 	}
7994 }
7995 
7996 /* restore rss filter */
7997 static inline void
7998 ixgbe_rss_filter_restore(struct rte_eth_dev *dev)
7999 {
8000 	struct ixgbe_filter_info *filter_info =
8001 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8002 
8003 	if (filter_info->rss_info.conf.queue_num)
8004 		ixgbe_config_rss_filter(dev,
8005 			&filter_info->rss_info, TRUE);
8006 }
8007 
8008 static int
8009 ixgbe_filter_restore(struct rte_eth_dev *dev)
8010 {
8011 	ixgbe_ntuple_filter_restore(dev);
8012 	ixgbe_ethertype_filter_restore(dev);
8013 	ixgbe_syn_filter_restore(dev);
8014 	ixgbe_fdir_filter_restore(dev);
8015 	ixgbe_l2_tn_filter_restore(dev);
8016 	ixgbe_rss_filter_restore(dev);
8017 
8018 	return 0;
8019 }
8020 
8021 static void
8022 ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev)
8023 {
8024 	struct ixgbe_l2_tn_info *l2_tn_info =
8025 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8026 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8027 
8028 	if (l2_tn_info->e_tag_en)
8029 		(void)ixgbe_e_tag_enable(hw);
8030 
8031 	if (l2_tn_info->e_tag_fwd_en)
8032 		(void)ixgbe_e_tag_forwarding_en_dis(dev, 1);
8033 
8034 	(void)ixgbe_update_e_tag_eth_type(hw, l2_tn_info->e_tag_ether_type);
8035 }
8036 
8037 /* remove all the n-tuple filters */
8038 void
8039 ixgbe_clear_all_ntuple_filter(struct rte_eth_dev *dev)
8040 {
8041 	struct ixgbe_filter_info *filter_info =
8042 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8043 	struct ixgbe_5tuple_filter *p_5tuple;
8044 
8045 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
8046 		ixgbe_remove_5tuple_filter(dev, p_5tuple);
8047 }
8048 
8049 /* remove all the ether type filters */
8050 void
8051 ixgbe_clear_all_ethertype_filter(struct rte_eth_dev *dev)
8052 {
8053 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8054 	struct ixgbe_filter_info *filter_info =
8055 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8056 	int i;
8057 
8058 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8059 		if (filter_info->ethertype_mask & (1 << i) &&
8060 		    !filter_info->ethertype_filters[i].conf) {
8061 			(void)ixgbe_ethertype_filter_remove(filter_info,
8062 							    (uint8_t)i);
8063 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i), 0);
8064 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i), 0);
8065 			IXGBE_WRITE_FLUSH(hw);
8066 		}
8067 	}
8068 }
8069 
8070 /* remove the SYN filter */
8071 void
8072 ixgbe_clear_syn_filter(struct rte_eth_dev *dev)
8073 {
8074 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8075 	struct ixgbe_filter_info *filter_info =
8076 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8077 
8078 	if (filter_info->syn_info & IXGBE_SYN_FILTER_ENABLE) {
8079 		filter_info->syn_info = 0;
8080 
8081 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, 0);
8082 		IXGBE_WRITE_FLUSH(hw);
8083 	}
8084 }
8085 
8086 /* remove all the L2 tunnel filters */
8087 int
8088 ixgbe_clear_all_l2_tn_filter(struct rte_eth_dev *dev)
8089 {
8090 	struct ixgbe_l2_tn_info *l2_tn_info =
8091 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8092 	struct ixgbe_l2_tn_filter *l2_tn_filter;
8093 	struct ixgbe_l2_tunnel_conf l2_tn_conf;
8094 	int ret = 0;
8095 
8096 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
8097 		l2_tn_conf.l2_tunnel_type = l2_tn_filter->key.l2_tn_type;
8098 		l2_tn_conf.tunnel_id      = l2_tn_filter->key.tn_id;
8099 		l2_tn_conf.pool           = l2_tn_filter->pool;
8100 		ret = ixgbe_dev_l2_tunnel_filter_del(dev, &l2_tn_conf);
8101 		if (ret < 0)
8102 			return ret;
8103 	}
8104 
8105 	return 0;
8106 }
8107 
8108 void
8109 ixgbe_dev_macsec_setting_save(struct rte_eth_dev *dev,
8110 				struct ixgbe_macsec_setting *macsec_setting)
8111 {
8112 	struct ixgbe_macsec_setting *macsec =
8113 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8114 
8115 	macsec->offload_en = macsec_setting->offload_en;
8116 	macsec->encrypt_en = macsec_setting->encrypt_en;
8117 	macsec->replayprotect_en = macsec_setting->replayprotect_en;
8118 }
8119 
8120 void
8121 ixgbe_dev_macsec_setting_reset(struct rte_eth_dev *dev)
8122 {
8123 	struct ixgbe_macsec_setting *macsec =
8124 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8125 
8126 	macsec->offload_en = 0;
8127 	macsec->encrypt_en = 0;
8128 	macsec->replayprotect_en = 0;
8129 }
8130 
8131 void
8132 ixgbe_dev_macsec_register_enable(struct rte_eth_dev *dev,
8133 				struct ixgbe_macsec_setting *macsec_setting)
8134 {
8135 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8136 	uint32_t ctrl;
8137 	uint8_t en = macsec_setting->encrypt_en;
8138 	uint8_t rp = macsec_setting->replayprotect_en;
8139 
8140 	/**
8141 	 * Workaround:
8142 	 * As no ixgbe_disable_sec_rx_path equivalent is
8143 	 * implemented for tx in the base code, and we are
8144 	 * not allowed to modify the base code in DPDK, so
8145 	 * just call the hand-written one directly for now.
8146 	 * The hardware support has been checked by
8147 	 * ixgbe_disable_sec_rx_path().
8148 	 */
8149 	ixgbe_disable_sec_tx_path_generic(hw);
8150 
8151 	/* Enable Ethernet CRC (required by MACsec offload) */
8152 	ctrl = IXGBE_READ_REG(hw, IXGBE_HLREG0);
8153 	ctrl |= IXGBE_HLREG0_TXCRCEN | IXGBE_HLREG0_RXCRCSTRP;
8154 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, ctrl);
8155 
8156 	/* Enable the TX and RX crypto engines */
8157 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8158 	ctrl &= ~IXGBE_SECTXCTRL_SECTX_DIS;
8159 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8160 
8161 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8162 	ctrl &= ~IXGBE_SECRXCTRL_SECRX_DIS;
8163 	IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8164 
8165 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
8166 	ctrl &= ~IXGBE_SECTX_MINSECIFG_MASK;
8167 	ctrl |= 0x3;
8168 	IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, ctrl);
8169 
8170 	/* Enable SA lookup */
8171 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8172 	ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8173 	ctrl |= en ? IXGBE_LSECTXCTRL_AUTH_ENCRYPT :
8174 		     IXGBE_LSECTXCTRL_AUTH;
8175 	ctrl |= IXGBE_LSECTXCTRL_AISCI;
8176 	ctrl &= ~IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8177 	ctrl |= IXGBE_MACSEC_PNTHRSH & IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8178 	IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8179 
8180 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8181 	ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8182 	ctrl |= IXGBE_LSECRXCTRL_STRICT << IXGBE_LSECRXCTRL_EN_SHIFT;
8183 	ctrl &= ~IXGBE_LSECRXCTRL_PLSH;
8184 	if (rp)
8185 		ctrl |= IXGBE_LSECRXCTRL_RP;
8186 	else
8187 		ctrl &= ~IXGBE_LSECRXCTRL_RP;
8188 	IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8189 
8190 	/* Start the data paths */
8191 	ixgbe_enable_sec_rx_path(hw);
8192 	/**
8193 	 * Workaround:
8194 	 * As no ixgbe_enable_sec_rx_path equivalent is
8195 	 * implemented for tx in the base code, and we are
8196 	 * not allowed to modify the base code in DPDK, so
8197 	 * just call the hand-written one directly for now.
8198 	 */
8199 	ixgbe_enable_sec_tx_path_generic(hw);
8200 }
8201 
8202 void
8203 ixgbe_dev_macsec_register_disable(struct rte_eth_dev *dev)
8204 {
8205 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8206 	uint32_t ctrl;
8207 
8208 	/**
8209 	 * Workaround:
8210 	 * As no ixgbe_disable_sec_rx_path equivalent is
8211 	 * implemented for tx in the base code, and we are
8212 	 * not allowed to modify the base code in DPDK, so
8213 	 * just call the hand-written one directly for now.
8214 	 * The hardware support has been checked by
8215 	 * ixgbe_disable_sec_rx_path().
8216 	 */
8217 	ixgbe_disable_sec_tx_path_generic(hw);
8218 
8219 	/* Disable the TX and RX crypto engines */
8220 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8221 	ctrl |= IXGBE_SECTXCTRL_SECTX_DIS;
8222 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8223 
8224 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8225 	ctrl |= IXGBE_SECRXCTRL_SECRX_DIS;
8226 	IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8227 
8228 	/* Disable SA lookup */
8229 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8230 	ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8231 	ctrl |= IXGBE_LSECTXCTRL_DISABLE;
8232 	IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8233 
8234 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8235 	ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8236 	ctrl |= IXGBE_LSECRXCTRL_DISABLE << IXGBE_LSECRXCTRL_EN_SHIFT;
8237 	IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8238 
8239 	/* Start the data paths */
8240 	ixgbe_enable_sec_rx_path(hw);
8241 	/**
8242 	 * Workaround:
8243 	 * As no ixgbe_enable_sec_rx_path equivalent is
8244 	 * implemented for tx in the base code, and we are
8245 	 * not allowed to modify the base code in DPDK, so
8246 	 * just call the hand-written one directly for now.
8247 	 */
8248 	ixgbe_enable_sec_tx_path_generic(hw);
8249 }
8250 
8251 RTE_PMD_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd);
8252 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
8253 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe, "* igb_uio | uio_pci_generic | vfio-pci");
8254 RTE_PMD_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd);
8255 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);
8256 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe_vf, "* igb_uio | vfio-pci");
8257 RTE_PMD_REGISTER_PARAM_STRING(net_ixgbe_vf,
8258 			      IXGBEVF_DEVARG_PFLINK_FULLCHK "=<0|1>");
8259 
8260 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_init, init, NOTICE);
8261 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_driver, driver, NOTICE);
8262 
8263 #ifdef RTE_ETHDEV_DEBUG_RX
8264 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_rx, rx, DEBUG);
8265 #endif
8266 #ifdef RTE_ETHDEV_DEBUG_TX
8267 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_tx, tx, DEBUG);
8268 #endif
8269