xref: /dpdk/drivers/net/ixgbe/ixgbe_ethdev.c (revision fa4dfda5)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2017 Intel Corporation
3  */
4 
5 #include <sys/queue.h>
6 #include <stdio.h>
7 #include <errno.h>
8 #include <stdint.h>
9 #include <string.h>
10 #include <unistd.h>
11 #include <stdarg.h>
12 #include <inttypes.h>
13 #include <rte_string_fns.h>
14 #include <rte_byteorder.h>
15 #include <rte_common.h>
16 #include <rte_cycles.h>
17 
18 #include <rte_interrupts.h>
19 #include <rte_log.h>
20 #include <rte_debug.h>
21 #include <rte_pci.h>
22 #include <rte_bus_pci.h>
23 #include <rte_branch_prediction.h>
24 #include <rte_memory.h>
25 #include <rte_kvargs.h>
26 #include <rte_eal.h>
27 #include <rte_alarm.h>
28 #include <rte_ether.h>
29 #include <ethdev_driver.h>
30 #include <ethdev_pci.h>
31 #include <rte_malloc.h>
32 #include <rte_random.h>
33 #include <rte_dev.h>
34 #include <rte_hash_crc.h>
35 #ifdef RTE_LIB_SECURITY
36 #include <rte_security_driver.h>
37 #endif
38 
39 #include "ixgbe_logs.h"
40 #include "base/ixgbe_api.h"
41 #include "base/ixgbe_vf.h"
42 #include "base/ixgbe_common.h"
43 #include "ixgbe_ethdev.h"
44 #include "ixgbe_bypass.h"
45 #include "ixgbe_rxtx.h"
46 #include "base/ixgbe_type.h"
47 #include "base/ixgbe_phy.h"
48 #include "base/ixgbe_osdep.h"
49 #include "ixgbe_regs.h"
50 
51 /*
52  * High threshold controlling when to start sending XOFF frames. Must be at
53  * least 8 bytes less than receive packet buffer size. This value is in units
54  * of 1024 bytes.
55  */
56 #define IXGBE_FC_HI    0x80
57 
58 /*
59  * Low threshold controlling when to start sending XON frames. This value is
60  * in units of 1024 bytes.
61  */
62 #define IXGBE_FC_LO    0x40
63 
64 /* Timer value included in XOFF frames. */
65 #define IXGBE_FC_PAUSE 0x680
66 
67 /*Default value of Max Rx Queue*/
68 #define IXGBE_MAX_RX_QUEUE_NUM 128
69 
70 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
71 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
72 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
73 
74 #define IXGBE_MMW_SIZE_DEFAULT        0x4
75 #define IXGBE_MMW_SIZE_JUMBO_FRAME    0x14
76 #define IXGBE_MAX_RING_DESC           4096 /* replicate define from rxtx */
77 
78 /*
79  *  Default values for RX/TX configuration
80  */
81 #define IXGBE_DEFAULT_RX_FREE_THRESH  32
82 #define IXGBE_DEFAULT_RX_PTHRESH      8
83 #define IXGBE_DEFAULT_RX_HTHRESH      8
84 #define IXGBE_DEFAULT_RX_WTHRESH      0
85 
86 #define IXGBE_DEFAULT_TX_FREE_THRESH  32
87 #define IXGBE_DEFAULT_TX_PTHRESH      32
88 #define IXGBE_DEFAULT_TX_HTHRESH      0
89 #define IXGBE_DEFAULT_TX_WTHRESH      0
90 #define IXGBE_DEFAULT_TX_RSBIT_THRESH 32
91 
92 /* Bit shift and mask */
93 #define IXGBE_4_BIT_WIDTH  (CHAR_BIT / 2)
94 #define IXGBE_4_BIT_MASK   RTE_LEN2MASK(IXGBE_4_BIT_WIDTH, uint8_t)
95 #define IXGBE_8_BIT_WIDTH  CHAR_BIT
96 #define IXGBE_8_BIT_MASK   UINT8_MAX
97 
98 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
99 
100 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
101 
102 /* Additional timesync values. */
103 #define NSEC_PER_SEC             1000000000L
104 #define IXGBE_INCVAL_10GB        0x66666666
105 #define IXGBE_INCVAL_1GB         0x40000000
106 #define IXGBE_INCVAL_100         0x50000000
107 #define IXGBE_INCVAL_SHIFT_10GB  28
108 #define IXGBE_INCVAL_SHIFT_1GB   24
109 #define IXGBE_INCVAL_SHIFT_100   21
110 #define IXGBE_INCVAL_SHIFT_82599 7
111 #define IXGBE_INCPER_SHIFT_82599 24
112 
113 #define IXGBE_CYCLECOUNTER_MASK   0xffffffffffffffffULL
114 
115 #define IXGBE_VT_CTL_POOLING_MODE_MASK         0x00030000
116 #define IXGBE_VT_CTL_POOLING_MODE_ETAG         0x00010000
117 #define IXGBE_ETAG_ETYPE                       0x00005084
118 #define IXGBE_ETAG_ETYPE_MASK                  0x0000ffff
119 #define IXGBE_ETAG_ETYPE_VALID                 0x80000000
120 #define IXGBE_RAH_ADTYPE                       0x40000000
121 #define IXGBE_RAL_ETAG_FILTER_MASK             0x00003fff
122 #define IXGBE_VMVIR_TAGA_MASK                  0x18000000
123 #define IXGBE_VMVIR_TAGA_ETAG_INSERT           0x08000000
124 #define IXGBE_VMTIR(_i) (0x00017000 + ((_i) * 4)) /* 64 of these (0-63) */
125 #define IXGBE_QDE_STRIP_TAG                    0x00000004
126 #define IXGBE_VTEICR_MASK                      0x07
127 
128 #define IXGBE_EXVET_VET_EXT_SHIFT              16
129 #define IXGBE_DMATXCTL_VT_MASK                 0xFFFF0000
130 
131 #define IXGBEVF_DEVARG_PFLINK_FULLCHK		"pflink_fullchk"
132 
133 static const char * const ixgbevf_valid_arguments[] = {
134 	IXGBEVF_DEVARG_PFLINK_FULLCHK,
135 	NULL
136 };
137 
138 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params);
139 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
140 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev);
141 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev);
142 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev);
143 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev);
144 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev);
145 static int ixgbe_dev_configure(struct rte_eth_dev *dev);
146 static int ixgbe_dev_start(struct rte_eth_dev *dev);
147 static int ixgbe_dev_stop(struct rte_eth_dev *dev);
148 static int ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
149 static int ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
150 static int ixgbe_dev_close(struct rte_eth_dev *dev);
151 static int ixgbe_dev_reset(struct rte_eth_dev *dev);
152 static int ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
153 static int ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
154 static int ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
155 static int ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
156 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
157 				int wait_to_complete);
158 static int ixgbe_dev_stats_get(struct rte_eth_dev *dev,
159 				struct rte_eth_stats *stats);
160 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
161 				struct rte_eth_xstat *xstats, unsigned n);
162 static int ixgbevf_dev_xstats_get(struct rte_eth_dev *dev,
163 				  struct rte_eth_xstat *xstats, unsigned n);
164 static int
165 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
166 		uint64_t *values, unsigned int n);
167 static int ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
168 static int ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
169 static int ixgbe_dev_xstats_get_names(struct rte_eth_dev *dev,
170 	struct rte_eth_xstat_name *xstats_names,
171 	unsigned int size);
172 static int ixgbevf_dev_xstats_get_names(struct rte_eth_dev *dev,
173 	struct rte_eth_xstat_name *xstats_names, unsigned limit);
174 static int ixgbe_dev_xstats_get_names_by_id(
175 	struct rte_eth_dev *dev,
176 	const uint64_t *ids,
177 	struct rte_eth_xstat_name *xstats_names,
178 	unsigned int limit);
179 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
180 					     uint16_t queue_id,
181 					     uint8_t stat_idx,
182 					     uint8_t is_rx);
183 static int ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
184 				 size_t fw_size);
185 static int ixgbe_dev_info_get(struct rte_eth_dev *dev,
186 			      struct rte_eth_dev_info *dev_info);
187 static const uint32_t *ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev);
188 static int ixgbevf_dev_info_get(struct rte_eth_dev *dev,
189 				struct rte_eth_dev_info *dev_info);
190 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
191 
192 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
193 		uint16_t vlan_id, int on);
194 static int ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
195 			       enum rte_vlan_type vlan_type,
196 			       uint16_t tpid_id);
197 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
198 		uint16_t queue, bool on);
199 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
200 		int on);
201 static void ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev,
202 						  int mask);
203 static int ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask);
204 static int ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
205 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
206 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
207 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
208 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
209 
210 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
211 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
212 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
213 			       struct rte_eth_fc_conf *fc_conf);
214 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
215 			       struct rte_eth_fc_conf *fc_conf);
216 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
217 		struct rte_eth_pfc_conf *pfc_conf);
218 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
219 			struct rte_eth_rss_reta_entry64 *reta_conf,
220 			uint16_t reta_size);
221 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
222 			struct rte_eth_rss_reta_entry64 *reta_conf,
223 			uint16_t reta_size);
224 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
225 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
226 static int ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev);
227 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
228 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
229 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev);
230 static void ixgbe_dev_interrupt_handler(void *param);
231 static void ixgbe_dev_interrupt_delayed_handler(void *param);
232 static void *ixgbe_dev_setup_link_thread_handler(void *param);
233 static int ixgbe_dev_wait_setup_link_complete(struct rte_eth_dev *dev,
234 					      uint32_t timeout_ms);
235 
236 static int ixgbe_add_rar(struct rte_eth_dev *dev,
237 			struct rte_ether_addr *mac_addr,
238 			uint32_t index, uint32_t pool);
239 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
240 static int ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
241 					   struct rte_ether_addr *mac_addr);
242 static void ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config);
243 static bool is_device_supported(struct rte_eth_dev *dev,
244 				struct rte_pci_driver *drv);
245 
246 /* For Virtual Function support */
247 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
248 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
249 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
250 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
251 static int ixgbevf_dev_link_update(struct rte_eth_dev *dev,
252 				   int wait_to_complete);
253 static int ixgbevf_dev_stop(struct rte_eth_dev *dev);
254 static int ixgbevf_dev_close(struct rte_eth_dev *dev);
255 static int  ixgbevf_dev_reset(struct rte_eth_dev *dev);
256 static void ixgbevf_intr_disable(struct rte_eth_dev *dev);
257 static void ixgbevf_intr_enable(struct rte_eth_dev *dev);
258 static int ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
259 		struct rte_eth_stats *stats);
260 static int ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
261 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
262 		uint16_t vlan_id, int on);
263 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
264 		uint16_t queue, int on);
265 static int ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask);
266 static int ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
267 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
268 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
269 					    uint16_t queue_id);
270 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
271 					     uint16_t queue_id);
272 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
273 				 uint8_t queue, uint8_t msix_vector);
274 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
275 static int ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev);
276 static int ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev);
277 static int ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev);
278 static int ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev);
279 
280 /* For Eth VMDQ APIs support */
281 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
282 		rte_ether_addr * mac_addr, uint8_t on);
283 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on);
284 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
285 					  uint16_t queue_id);
286 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
287 					   uint16_t queue_id);
288 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
289 			       uint8_t queue, uint8_t msix_vector);
290 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
291 
292 static int ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
293 				struct rte_ether_addr *mac_addr,
294 				uint32_t index, uint32_t pool);
295 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
296 static int ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
297 					     struct rte_ether_addr *mac_addr);
298 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
299 			struct ixgbe_5tuple_filter *filter);
300 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
301 			struct ixgbe_5tuple_filter *filter);
302 static int ixgbe_dev_flow_ops_get(struct rte_eth_dev *dev,
303 				  const struct rte_flow_ops **ops);
304 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
305 
306 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
307 				      struct rte_ether_addr *mc_addr_set,
308 				      uint32_t nb_mc_addr);
309 static int ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
310 				   struct rte_eth_dcb_info *dcb_info);
311 
312 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
313 static int ixgbe_get_regs(struct rte_eth_dev *dev,
314 			    struct rte_dev_reg_info *regs);
315 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
316 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
317 				struct rte_dev_eeprom_info *eeprom);
318 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
319 				struct rte_dev_eeprom_info *eeprom);
320 
321 static int ixgbe_get_module_info(struct rte_eth_dev *dev,
322 				 struct rte_eth_dev_module_info *modinfo);
323 static int ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
324 				   struct rte_dev_eeprom_info *info);
325 
326 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
327 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
328 				struct rte_dev_reg_info *regs);
329 
330 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
331 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
332 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
333 					    struct timespec *timestamp,
334 					    uint32_t flags);
335 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
336 					    struct timespec *timestamp);
337 static int ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
338 static int ixgbe_timesync_read_time(struct rte_eth_dev *dev,
339 				   struct timespec *timestamp);
340 static int ixgbe_timesync_write_time(struct rte_eth_dev *dev,
341 				   const struct timespec *timestamp);
342 static void ixgbevf_dev_interrupt_handler(void *param);
343 
344 static int ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
345 					 struct rte_eth_udp_tunnel *udp_tunnel);
346 static int ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
347 					 struct rte_eth_udp_tunnel *udp_tunnel);
348 static int ixgbe_filter_restore(struct rte_eth_dev *dev);
349 static void ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev);
350 static int ixgbe_wait_for_link_up(struct ixgbe_hw *hw);
351 
352 /*
353  * Define VF Stats MACRO for Non "cleared on read" register
354  */
355 #define UPDATE_VF_STAT(reg, last, cur)                          \
356 {                                                               \
357 	uint32_t latest = IXGBE_READ_REG(hw, reg);              \
358 	cur += (latest - last) & UINT_MAX;                      \
359 	last = latest;                                          \
360 }
361 
362 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
363 {                                                                \
364 	u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
365 	u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
366 	u64 latest = ((new_msb << 32) | new_lsb);                \
367 	cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
368 	last = latest;                                           \
369 }
370 
371 #define IXGBE_SET_HWSTRIP(h, q) do {\
372 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
373 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
374 		(h)->bitmap[idx] |= 1 << bit;\
375 	} while (0)
376 
377 #define IXGBE_CLEAR_HWSTRIP(h, q) do {\
378 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
379 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
380 		(h)->bitmap[idx] &= ~(1 << bit);\
381 	} while (0)
382 
383 #define IXGBE_GET_HWSTRIP(h, q, r) do {\
384 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
385 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
386 		(r) = (h)->bitmap[idx] >> bit & 1;\
387 	} while (0)
388 
389 /*
390  * The set of PCI devices this driver supports
391  */
392 static const struct rte_pci_id pci_id_ixgbe_map[] = {
393 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598) },
394 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_BX) },
395 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT) },
396 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT) },
397 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT) },
398 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2) },
399 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM) },
400 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4) },
401 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT) },
402 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT) },
403 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM) },
404 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR) },
405 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4) },
406 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ) },
407 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KR) },
408 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE) },
409 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4) },
410 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP) },
411 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE) },
412 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE) },
413 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_EM) },
414 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2) },
415 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP) },
416 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP) },
417 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP) },
418 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM) },
419 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM) },
420 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T) },
421 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1) },
422 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP) },
423 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T) },
424 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_1G_T) },
425 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T) },
426 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1) },
427 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR) },
428 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR_L) },
429 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP_N) },
430 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII) },
431 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII_L) },
432 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_10G_T) },
433 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP) },
434 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP_N) },
435 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP) },
436 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T) },
437 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T_L) },
438 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4) },
439 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR) },
440 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_XFI) },
441 #ifdef RTE_LIBRTE_IXGBE_BYPASS
442 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BYPASS) },
443 #endif
444 	{ .vendor_id = 0, /* sentinel */ },
445 };
446 
447 /*
448  * The set of PCI devices this driver supports (for 82599 VF)
449  */
450 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
451 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF) },
452 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF_HV) },
453 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF) },
454 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF_HV) },
455 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF_HV) },
456 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF) },
457 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF) },
458 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF_HV) },
459 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF) },
460 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF_HV) },
461 	{ .vendor_id = 0, /* sentinel */ },
462 };
463 
464 static const struct rte_eth_desc_lim rx_desc_lim = {
465 	.nb_max = IXGBE_MAX_RING_DESC,
466 	.nb_min = IXGBE_MIN_RING_DESC,
467 	.nb_align = IXGBE_RXD_ALIGN,
468 };
469 
470 static const struct rte_eth_desc_lim tx_desc_lim = {
471 	.nb_max = IXGBE_MAX_RING_DESC,
472 	.nb_min = IXGBE_MIN_RING_DESC,
473 	.nb_align = IXGBE_TXD_ALIGN,
474 	.nb_seg_max = IXGBE_TX_MAX_SEG,
475 	.nb_mtu_seg_max = IXGBE_TX_MAX_SEG,
476 };
477 
478 static const struct eth_dev_ops ixgbe_eth_dev_ops = {
479 	.dev_configure        = ixgbe_dev_configure,
480 	.dev_start            = ixgbe_dev_start,
481 	.dev_stop             = ixgbe_dev_stop,
482 	.dev_set_link_up    = ixgbe_dev_set_link_up,
483 	.dev_set_link_down  = ixgbe_dev_set_link_down,
484 	.dev_close            = ixgbe_dev_close,
485 	.dev_reset	      = ixgbe_dev_reset,
486 	.promiscuous_enable   = ixgbe_dev_promiscuous_enable,
487 	.promiscuous_disable  = ixgbe_dev_promiscuous_disable,
488 	.allmulticast_enable  = ixgbe_dev_allmulticast_enable,
489 	.allmulticast_disable = ixgbe_dev_allmulticast_disable,
490 	.link_update          = ixgbe_dev_link_update,
491 	.stats_get            = ixgbe_dev_stats_get,
492 	.xstats_get           = ixgbe_dev_xstats_get,
493 	.xstats_get_by_id     = ixgbe_dev_xstats_get_by_id,
494 	.stats_reset          = ixgbe_dev_stats_reset,
495 	.xstats_reset         = ixgbe_dev_xstats_reset,
496 	.xstats_get_names     = ixgbe_dev_xstats_get_names,
497 	.xstats_get_names_by_id = ixgbe_dev_xstats_get_names_by_id,
498 	.queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
499 	.fw_version_get       = ixgbe_fw_version_get,
500 	.dev_infos_get        = ixgbe_dev_info_get,
501 	.dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
502 	.mtu_set              = ixgbe_dev_mtu_set,
503 	.vlan_filter_set      = ixgbe_vlan_filter_set,
504 	.vlan_tpid_set        = ixgbe_vlan_tpid_set,
505 	.vlan_offload_set     = ixgbe_vlan_offload_set,
506 	.vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
507 	.rx_queue_start	      = ixgbe_dev_rx_queue_start,
508 	.rx_queue_stop        = ixgbe_dev_rx_queue_stop,
509 	.tx_queue_start	      = ixgbe_dev_tx_queue_start,
510 	.tx_queue_stop        = ixgbe_dev_tx_queue_stop,
511 	.rx_queue_setup       = ixgbe_dev_rx_queue_setup,
512 	.rx_queue_intr_enable = ixgbe_dev_rx_queue_intr_enable,
513 	.rx_queue_intr_disable = ixgbe_dev_rx_queue_intr_disable,
514 	.rx_queue_release     = ixgbe_dev_rx_queue_release,
515 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
516 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
517 	.dev_led_on           = ixgbe_dev_led_on,
518 	.dev_led_off          = ixgbe_dev_led_off,
519 	.flow_ctrl_get        = ixgbe_flow_ctrl_get,
520 	.flow_ctrl_set        = ixgbe_flow_ctrl_set,
521 	.priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
522 	.mac_addr_add         = ixgbe_add_rar,
523 	.mac_addr_remove      = ixgbe_remove_rar,
524 	.mac_addr_set         = ixgbe_set_default_mac_addr,
525 	.uc_hash_table_set    = ixgbe_uc_hash_table_set,
526 	.uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
527 	.set_queue_rate_limit = ixgbe_set_queue_rate_limit,
528 	.reta_update          = ixgbe_dev_rss_reta_update,
529 	.reta_query           = ixgbe_dev_rss_reta_query,
530 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
531 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
532 	.flow_ops_get         = ixgbe_dev_flow_ops_get,
533 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
534 	.rxq_info_get         = ixgbe_rxq_info_get,
535 	.txq_info_get         = ixgbe_txq_info_get,
536 	.timesync_enable      = ixgbe_timesync_enable,
537 	.timesync_disable     = ixgbe_timesync_disable,
538 	.timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
539 	.timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
540 	.get_reg              = ixgbe_get_regs,
541 	.get_eeprom_length    = ixgbe_get_eeprom_length,
542 	.get_eeprom           = ixgbe_get_eeprom,
543 	.set_eeprom           = ixgbe_set_eeprom,
544 	.get_module_info      = ixgbe_get_module_info,
545 	.get_module_eeprom    = ixgbe_get_module_eeprom,
546 	.get_dcb_info         = ixgbe_dev_get_dcb_info,
547 	.timesync_adjust_time = ixgbe_timesync_adjust_time,
548 	.timesync_read_time   = ixgbe_timesync_read_time,
549 	.timesync_write_time  = ixgbe_timesync_write_time,
550 	.udp_tunnel_port_add  = ixgbe_dev_udp_tunnel_port_add,
551 	.udp_tunnel_port_del  = ixgbe_dev_udp_tunnel_port_del,
552 	.tm_ops_get           = ixgbe_tm_ops_get,
553 	.tx_done_cleanup      = ixgbe_dev_tx_done_cleanup,
554 	.get_monitor_addr     = ixgbe_get_monitor_addr,
555 };
556 
557 /*
558  * dev_ops for virtual function, bare necessities for basic vf
559  * operation have been implemented
560  */
561 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
562 	.dev_configure        = ixgbevf_dev_configure,
563 	.dev_start            = ixgbevf_dev_start,
564 	.dev_stop             = ixgbevf_dev_stop,
565 	.link_update          = ixgbevf_dev_link_update,
566 	.stats_get            = ixgbevf_dev_stats_get,
567 	.xstats_get           = ixgbevf_dev_xstats_get,
568 	.stats_reset          = ixgbevf_dev_stats_reset,
569 	.xstats_reset         = ixgbevf_dev_stats_reset,
570 	.xstats_get_names     = ixgbevf_dev_xstats_get_names,
571 	.dev_close            = ixgbevf_dev_close,
572 	.dev_reset	      = ixgbevf_dev_reset,
573 	.promiscuous_enable   = ixgbevf_dev_promiscuous_enable,
574 	.promiscuous_disable  = ixgbevf_dev_promiscuous_disable,
575 	.allmulticast_enable  = ixgbevf_dev_allmulticast_enable,
576 	.allmulticast_disable = ixgbevf_dev_allmulticast_disable,
577 	.dev_infos_get        = ixgbevf_dev_info_get,
578 	.dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
579 	.mtu_set              = ixgbevf_dev_set_mtu,
580 	.vlan_filter_set      = ixgbevf_vlan_filter_set,
581 	.vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
582 	.vlan_offload_set     = ixgbevf_vlan_offload_set,
583 	.rx_queue_setup       = ixgbe_dev_rx_queue_setup,
584 	.rx_queue_release     = ixgbe_dev_rx_queue_release,
585 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
586 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
587 	.rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
588 	.rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
589 	.mac_addr_add         = ixgbevf_add_mac_addr,
590 	.mac_addr_remove      = ixgbevf_remove_mac_addr,
591 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
592 	.rxq_info_get         = ixgbe_rxq_info_get,
593 	.txq_info_get         = ixgbe_txq_info_get,
594 	.mac_addr_set         = ixgbevf_set_default_mac_addr,
595 	.get_reg              = ixgbevf_get_regs,
596 	.reta_update          = ixgbe_dev_rss_reta_update,
597 	.reta_query           = ixgbe_dev_rss_reta_query,
598 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
599 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
600 	.tx_done_cleanup      = ixgbe_dev_tx_done_cleanup,
601 	.get_monitor_addr     = ixgbe_get_monitor_addr,
602 };
603 
604 /* store statistics names and its offset in stats structure */
605 struct rte_ixgbe_xstats_name_off {
606 	char name[RTE_ETH_XSTATS_NAME_SIZE];
607 	unsigned offset;
608 };
609 
610 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
611 	{"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
612 	{"rx_illegal_byte_errors", offsetof(struct ixgbe_hw_stats, illerrc)},
613 	{"rx_error_bytes", offsetof(struct ixgbe_hw_stats, errbc)},
614 	{"mac_local_errors", offsetof(struct ixgbe_hw_stats, mlfc)},
615 	{"mac_remote_errors", offsetof(struct ixgbe_hw_stats, mrfc)},
616 	{"rx_length_errors", offsetof(struct ixgbe_hw_stats, rlec)},
617 	{"tx_xon_packets", offsetof(struct ixgbe_hw_stats, lxontxc)},
618 	{"rx_xon_packets", offsetof(struct ixgbe_hw_stats, lxonrxc)},
619 	{"tx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxofftxc)},
620 	{"rx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
621 	{"rx_size_64_packets", offsetof(struct ixgbe_hw_stats, prc64)},
622 	{"rx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, prc127)},
623 	{"rx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, prc255)},
624 	{"rx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, prc511)},
625 	{"rx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
626 		prc1023)},
627 	{"rx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
628 		prc1522)},
629 	{"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
630 	{"rx_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
631 	{"rx_fragment_errors", offsetof(struct ixgbe_hw_stats, rfc)},
632 	{"rx_undersize_errors", offsetof(struct ixgbe_hw_stats, ruc)},
633 	{"rx_oversize_errors", offsetof(struct ixgbe_hw_stats, roc)},
634 	{"rx_jabber_errors", offsetof(struct ixgbe_hw_stats, rjc)},
635 	{"rx_management_packets", offsetof(struct ixgbe_hw_stats, mngprc)},
636 	{"rx_management_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
637 	{"tx_management_packets", offsetof(struct ixgbe_hw_stats, mngptc)},
638 	{"rx_total_packets", offsetof(struct ixgbe_hw_stats, tpr)},
639 	{"rx_total_bytes", offsetof(struct ixgbe_hw_stats, tor)},
640 	{"tx_total_packets", offsetof(struct ixgbe_hw_stats, tpt)},
641 	{"tx_size_64_packets", offsetof(struct ixgbe_hw_stats, ptc64)},
642 	{"tx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, ptc127)},
643 	{"tx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, ptc255)},
644 	{"tx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, ptc511)},
645 	{"tx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
646 		ptc1023)},
647 	{"tx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
648 		ptc1522)},
649 	{"tx_multicast_packets", offsetof(struct ixgbe_hw_stats, mptc)},
650 	{"tx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bptc)},
651 	{"rx_mac_short_packet_dropped", offsetof(struct ixgbe_hw_stats, mspdc)},
652 	{"rx_l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
653 
654 	{"flow_director_added_filters", offsetof(struct ixgbe_hw_stats,
655 		fdirustat_add)},
656 	{"flow_director_removed_filters", offsetof(struct ixgbe_hw_stats,
657 		fdirustat_remove)},
658 	{"flow_director_filter_add_errors", offsetof(struct ixgbe_hw_stats,
659 		fdirfstat_fadd)},
660 	{"flow_director_filter_remove_errors", offsetof(struct ixgbe_hw_stats,
661 		fdirfstat_fremove)},
662 	{"flow_director_matched_filters", offsetof(struct ixgbe_hw_stats,
663 		fdirmatch)},
664 	{"flow_director_missed_filters", offsetof(struct ixgbe_hw_stats,
665 		fdirmiss)},
666 
667 	{"rx_fcoe_crc_errors", offsetof(struct ixgbe_hw_stats, fccrc)},
668 	{"rx_fcoe_dropped", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
669 	{"rx_fcoe_mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats,
670 		fclast)},
671 	{"rx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeprc)},
672 	{"tx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeptc)},
673 	{"rx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwrc)},
674 	{"tx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwtc)},
675 	{"rx_fcoe_no_direct_data_placement", offsetof(struct ixgbe_hw_stats,
676 		fcoe_noddp)},
677 	{"rx_fcoe_no_direct_data_placement_ext_buff",
678 		offsetof(struct ixgbe_hw_stats, fcoe_noddp_ext_buff)},
679 
680 	{"tx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
681 		lxontxc)},
682 	{"rx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
683 		lxonrxc)},
684 	{"tx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
685 		lxofftxc)},
686 	{"rx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
687 		lxoffrxc)},
688 	{"rx_total_missed_packets", offsetof(struct ixgbe_hw_stats, mpctotal)},
689 };
690 
691 #define IXGBE_NB_HW_STATS (sizeof(rte_ixgbe_stats_strings) / \
692 			   sizeof(rte_ixgbe_stats_strings[0]))
693 
694 /* MACsec statistics */
695 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_macsec_strings[] = {
696 	{"out_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
697 		out_pkts_untagged)},
698 	{"out_pkts_encrypted", offsetof(struct ixgbe_macsec_stats,
699 		out_pkts_encrypted)},
700 	{"out_pkts_protected", offsetof(struct ixgbe_macsec_stats,
701 		out_pkts_protected)},
702 	{"out_octets_encrypted", offsetof(struct ixgbe_macsec_stats,
703 		out_octets_encrypted)},
704 	{"out_octets_protected", offsetof(struct ixgbe_macsec_stats,
705 		out_octets_protected)},
706 	{"in_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
707 		in_pkts_untagged)},
708 	{"in_pkts_badtag", offsetof(struct ixgbe_macsec_stats,
709 		in_pkts_badtag)},
710 	{"in_pkts_nosci", offsetof(struct ixgbe_macsec_stats,
711 		in_pkts_nosci)},
712 	{"in_pkts_unknownsci", offsetof(struct ixgbe_macsec_stats,
713 		in_pkts_unknownsci)},
714 	{"in_octets_decrypted", offsetof(struct ixgbe_macsec_stats,
715 		in_octets_decrypted)},
716 	{"in_octets_validated", offsetof(struct ixgbe_macsec_stats,
717 		in_octets_validated)},
718 	{"in_pkts_unchecked", offsetof(struct ixgbe_macsec_stats,
719 		in_pkts_unchecked)},
720 	{"in_pkts_delayed", offsetof(struct ixgbe_macsec_stats,
721 		in_pkts_delayed)},
722 	{"in_pkts_late", offsetof(struct ixgbe_macsec_stats,
723 		in_pkts_late)},
724 	{"in_pkts_ok", offsetof(struct ixgbe_macsec_stats,
725 		in_pkts_ok)},
726 	{"in_pkts_invalid", offsetof(struct ixgbe_macsec_stats,
727 		in_pkts_invalid)},
728 	{"in_pkts_notvalid", offsetof(struct ixgbe_macsec_stats,
729 		in_pkts_notvalid)},
730 	{"in_pkts_unusedsa", offsetof(struct ixgbe_macsec_stats,
731 		in_pkts_unusedsa)},
732 	{"in_pkts_notusingsa", offsetof(struct ixgbe_macsec_stats,
733 		in_pkts_notusingsa)},
734 };
735 
736 #define IXGBE_NB_MACSEC_STATS (sizeof(rte_ixgbe_macsec_strings) / \
737 			   sizeof(rte_ixgbe_macsec_strings[0]))
738 
739 /* Per-queue statistics */
740 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_rxq_strings[] = {
741 	{"mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats, rnbc)},
742 	{"dropped", offsetof(struct ixgbe_hw_stats, mpc)},
743 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxonrxc)},
744 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxoffrxc)},
745 };
746 
747 #define IXGBE_NB_RXQ_PRIO_STATS (sizeof(rte_ixgbe_rxq_strings) / \
748 			   sizeof(rte_ixgbe_rxq_strings[0]))
749 #define IXGBE_NB_RXQ_PRIO_VALUES 8
750 
751 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_txq_strings[] = {
752 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxontxc)},
753 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxofftxc)},
754 	{"xon_to_xoff_packets", offsetof(struct ixgbe_hw_stats,
755 		pxon2offc)},
756 };
757 
758 #define IXGBE_NB_TXQ_PRIO_STATS (sizeof(rte_ixgbe_txq_strings) / \
759 			   sizeof(rte_ixgbe_txq_strings[0]))
760 #define IXGBE_NB_TXQ_PRIO_VALUES 8
761 
762 static const struct rte_ixgbe_xstats_name_off rte_ixgbevf_stats_strings[] = {
763 	{"rx_multicast_packets", offsetof(struct ixgbevf_hw_stats, vfmprc)},
764 };
765 
766 #define IXGBEVF_NB_XSTATS (sizeof(rte_ixgbevf_stats_strings) /	\
767 		sizeof(rte_ixgbevf_stats_strings[0]))
768 
769 /*
770  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
771  */
772 static inline int
773 ixgbe_is_sfp(struct ixgbe_hw *hw)
774 {
775 	switch (hw->phy.type) {
776 	case ixgbe_phy_sfp_avago:
777 	case ixgbe_phy_sfp_ftl:
778 	case ixgbe_phy_sfp_intel:
779 	case ixgbe_phy_sfp_unknown:
780 	case ixgbe_phy_sfp_passive_tyco:
781 	case ixgbe_phy_sfp_passive_unknown:
782 		return 1;
783 	default:
784 		return 0;
785 	}
786 }
787 
788 static inline int32_t
789 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
790 {
791 	uint32_t ctrl_ext;
792 	int32_t status;
793 
794 	status = ixgbe_reset_hw(hw);
795 
796 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
797 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
798 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
799 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
800 	IXGBE_WRITE_FLUSH(hw);
801 
802 	if (status == IXGBE_ERR_SFP_NOT_PRESENT)
803 		status = IXGBE_SUCCESS;
804 	return status;
805 }
806 
807 static inline void
808 ixgbe_enable_intr(struct rte_eth_dev *dev)
809 {
810 	struct ixgbe_interrupt *intr =
811 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
812 	struct ixgbe_hw *hw =
813 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
814 
815 	IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
816 	IXGBE_WRITE_FLUSH(hw);
817 }
818 
819 /*
820  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
821  */
822 static void
823 ixgbe_disable_intr(struct ixgbe_hw *hw)
824 {
825 	PMD_INIT_FUNC_TRACE();
826 
827 	if (hw->mac.type == ixgbe_mac_82598EB) {
828 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
829 	} else {
830 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
831 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
832 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
833 	}
834 	IXGBE_WRITE_FLUSH(hw);
835 }
836 
837 /*
838  * This function resets queue statistics mapping registers.
839  * From Niantic datasheet, Initialization of Statistics section:
840  * "...if software requires the queue counters, the RQSMR and TQSM registers
841  * must be re-programmed following a device reset.
842  */
843 static void
844 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
845 {
846 	uint32_t i;
847 
848 	for (i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
849 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
850 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
851 	}
852 }
853 
854 
855 static int
856 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
857 				  uint16_t queue_id,
858 				  uint8_t stat_idx,
859 				  uint8_t is_rx)
860 {
861 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
862 #define NB_QMAP_FIELDS_PER_QSM_REG 4
863 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
864 
865 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
866 	struct ixgbe_stat_mapping_registers *stat_mappings =
867 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
868 	uint32_t qsmr_mask = 0;
869 	uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
870 	uint32_t q_map;
871 	uint8_t n, offset;
872 
873 	if ((hw->mac.type != ixgbe_mac_82599EB) &&
874 		(hw->mac.type != ixgbe_mac_X540) &&
875 		(hw->mac.type != ixgbe_mac_X550) &&
876 		(hw->mac.type != ixgbe_mac_X550EM_x) &&
877 		(hw->mac.type != ixgbe_mac_X550EM_a))
878 		return -ENOSYS;
879 
880 	PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
881 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
882 		     queue_id, stat_idx);
883 
884 	n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
885 	if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
886 		PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
887 		return -EIO;
888 	}
889 	offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
890 
891 	/* Now clear any previous stat_idx set */
892 	clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
893 	if (!is_rx)
894 		stat_mappings->tqsm[n] &= ~clearing_mask;
895 	else
896 		stat_mappings->rqsmr[n] &= ~clearing_mask;
897 
898 	q_map = (uint32_t)stat_idx;
899 	q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
900 	qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
901 	if (!is_rx)
902 		stat_mappings->tqsm[n] |= qsmr_mask;
903 	else
904 		stat_mappings->rqsmr[n] |= qsmr_mask;
905 
906 	PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
907 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
908 		     queue_id, stat_idx);
909 	PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
910 		     is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
911 
912 	/* Now write the mapping in the appropriate register */
913 	if (is_rx) {
914 		PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
915 			     stat_mappings->rqsmr[n], n);
916 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
917 	} else {
918 		PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
919 			     stat_mappings->tqsm[n], n);
920 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
921 	}
922 	return 0;
923 }
924 
925 static void
926 ixgbe_restore_statistics_mapping(struct rte_eth_dev *dev)
927 {
928 	struct ixgbe_stat_mapping_registers *stat_mappings =
929 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
930 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
931 	int i;
932 
933 	/* write whatever was in stat mapping table to the NIC */
934 	for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
935 		/* rx */
936 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
937 
938 		/* tx */
939 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
940 	}
941 }
942 
943 static void
944 ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config)
945 {
946 	uint8_t i;
947 	struct ixgbe_dcb_tc_config *tc;
948 	uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
949 
950 	dcb_config->num_tcs.pg_tcs = dcb_max_tc;
951 	dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
952 	for (i = 0; i < dcb_max_tc; i++) {
953 		tc = &dcb_config->tc_config[i];
954 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
955 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
956 				 (uint8_t)(100/dcb_max_tc + (i & 1));
957 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
958 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
959 				 (uint8_t)(100/dcb_max_tc + (i & 1));
960 		tc->pfc = ixgbe_dcb_pfc_disabled;
961 	}
962 
963 	/* Initialize default user to priority mapping, UPx->TC0 */
964 	tc = &dcb_config->tc_config[0];
965 	tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
966 	tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
967 	for (i = 0; i < IXGBE_DCB_MAX_BW_GROUP; i++) {
968 		dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
969 		dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
970 	}
971 	dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
972 	dcb_config->pfc_mode_enable = false;
973 	dcb_config->vt_mode = true;
974 	dcb_config->round_robin_enable = false;
975 	/* support all DCB capabilities in 82599 */
976 	dcb_config->support.capabilities = 0xFF;
977 
978 	/*we only support 4 Tcs for X540, X550 */
979 	if (hw->mac.type == ixgbe_mac_X540 ||
980 		hw->mac.type == ixgbe_mac_X550 ||
981 		hw->mac.type == ixgbe_mac_X550EM_x ||
982 		hw->mac.type == ixgbe_mac_X550EM_a) {
983 		dcb_config->num_tcs.pg_tcs = 4;
984 		dcb_config->num_tcs.pfc_tcs = 4;
985 	}
986 }
987 
988 /*
989  * Ensure that all locks are released before first NVM or PHY access
990  */
991 static void
992 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
993 {
994 	uint16_t mask;
995 
996 	/*
997 	 * Phy lock should not fail in this early stage. If this is the case,
998 	 * it is due to an improper exit of the application.
999 	 * So force the release of the faulty lock. Release of common lock
1000 	 * is done automatically by swfw_sync function.
1001 	 */
1002 	mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
1003 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1004 		PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
1005 	}
1006 	ixgbe_release_swfw_semaphore(hw, mask);
1007 
1008 	/*
1009 	 * These ones are more tricky since they are common to all ports; but
1010 	 * swfw_sync retries last long enough (1s) to be almost sure that if
1011 	 * lock can not be taken it is due to an improper lock of the
1012 	 * semaphore.
1013 	 */
1014 	mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
1015 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1016 		PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1017 	}
1018 	ixgbe_release_swfw_semaphore(hw, mask);
1019 }
1020 
1021 /*
1022  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
1023  * It returns 0 on success.
1024  */
1025 static int
1026 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params __rte_unused)
1027 {
1028 	struct ixgbe_adapter *ad = eth_dev->data->dev_private;
1029 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1030 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1031 	struct ixgbe_hw *hw =
1032 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1033 	struct ixgbe_vfta *shadow_vfta =
1034 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1035 	struct ixgbe_hwstrip *hwstrip =
1036 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1037 	struct ixgbe_dcb_config *dcb_config =
1038 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
1039 	struct ixgbe_filter_info *filter_info =
1040 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1041 	struct ixgbe_bw_conf *bw_conf =
1042 		IXGBE_DEV_PRIVATE_TO_BW_CONF(eth_dev->data->dev_private);
1043 	uint32_t ctrl_ext;
1044 	uint16_t csum;
1045 	int diag, i, ret;
1046 
1047 	PMD_INIT_FUNC_TRACE();
1048 
1049 	ixgbe_dev_macsec_setting_reset(eth_dev);
1050 
1051 	eth_dev->dev_ops = &ixgbe_eth_dev_ops;
1052 	eth_dev->rx_queue_count       = ixgbe_dev_rx_queue_count;
1053 	eth_dev->rx_descriptor_status = ixgbe_dev_rx_descriptor_status;
1054 	eth_dev->tx_descriptor_status = ixgbe_dev_tx_descriptor_status;
1055 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1056 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1057 	eth_dev->tx_pkt_prepare = &ixgbe_prep_pkts;
1058 
1059 	/*
1060 	 * For secondary processes, we don't initialise any further as primary
1061 	 * has already done this work. Only check we don't need a different
1062 	 * RX and TX function.
1063 	 */
1064 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1065 		struct ixgbe_tx_queue *txq;
1066 		/* TX queue function in primary, set by last queue initialized
1067 		 * Tx queue may not initialized by primary process
1068 		 */
1069 		if (eth_dev->data->tx_queues) {
1070 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
1071 			ixgbe_set_tx_function(eth_dev, txq);
1072 		} else {
1073 			/* Use default TX function if we get here */
1074 			PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
1075 				     "Using default TX function.");
1076 		}
1077 
1078 		ixgbe_set_rx_function(eth_dev);
1079 
1080 		return 0;
1081 	}
1082 
1083 	rte_atomic32_clear(&ad->link_thread_running);
1084 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1085 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1086 
1087 	/* Vendor and Device ID need to be set before init of shared code */
1088 	hw->device_id = pci_dev->id.device_id;
1089 	hw->vendor_id = pci_dev->id.vendor_id;
1090 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1091 	hw->allow_unsupported_sfp = 1;
1092 
1093 	/* Initialize the shared code (base driver) */
1094 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1095 	diag = ixgbe_bypass_init_shared_code(hw);
1096 #else
1097 	diag = ixgbe_init_shared_code(hw);
1098 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1099 
1100 	if (diag != IXGBE_SUCCESS) {
1101 		PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
1102 		return -EIO;
1103 	}
1104 
1105 	if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
1106 		PMD_INIT_LOG(ERR, "\nERROR: "
1107 			"Firmware recovery mode detected. Limiting functionality.\n"
1108 			"Refer to the Intel(R) Ethernet Adapters and Devices "
1109 			"User Guide for details on firmware recovery mode.");
1110 		return -EIO;
1111 	}
1112 
1113 	/* pick up the PCI bus settings for reporting later */
1114 	ixgbe_get_bus_info(hw);
1115 
1116 	/* Unlock any pending hardware semaphore */
1117 	ixgbe_swfw_lock_reset(hw);
1118 
1119 #ifdef RTE_LIB_SECURITY
1120 	/* Initialize security_ctx only for primary process*/
1121 	if (ixgbe_ipsec_ctx_create(eth_dev))
1122 		return -ENOMEM;
1123 #endif
1124 
1125 	/* Initialize DCB configuration*/
1126 	memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
1127 	ixgbe_dcb_init(hw, dcb_config);
1128 	/* Get Hardware Flow Control setting */
1129 	hw->fc.requested_mode = ixgbe_fc_none;
1130 	hw->fc.current_mode = ixgbe_fc_none;
1131 	hw->fc.pause_time = IXGBE_FC_PAUSE;
1132 	for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
1133 		hw->fc.low_water[i] = IXGBE_FC_LO;
1134 		hw->fc.high_water[i] = IXGBE_FC_HI;
1135 	}
1136 	hw->fc.send_xon = 1;
1137 
1138 	/* Make sure we have a good EEPROM before we read from it */
1139 	diag = ixgbe_validate_eeprom_checksum(hw, &csum);
1140 	if (diag != IXGBE_SUCCESS) {
1141 		PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
1142 		return -EIO;
1143 	}
1144 
1145 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1146 	diag = ixgbe_bypass_init_hw(hw);
1147 #else
1148 	diag = ixgbe_init_hw(hw);
1149 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1150 
1151 	/*
1152 	 * Devices with copper phys will fail to initialise if ixgbe_init_hw()
1153 	 * is called too soon after the kernel driver unbinding/binding occurs.
1154 	 * The failure occurs in ixgbe_identify_phy_generic() for all devices,
1155 	 * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
1156 	 * also called. See ixgbe_identify_phy_82599(). The reason for the
1157 	 * failure is not known, and only occuts when virtualisation features
1158 	 * are disabled in the bios. A delay of 100ms  was found to be enough by
1159 	 * trial-and-error, and is doubled to be safe.
1160 	 */
1161 	if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
1162 		rte_delay_ms(200);
1163 		diag = ixgbe_init_hw(hw);
1164 	}
1165 
1166 	if (diag == IXGBE_ERR_SFP_NOT_PRESENT)
1167 		diag = IXGBE_SUCCESS;
1168 
1169 	if (diag == IXGBE_ERR_EEPROM_VERSION) {
1170 		PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
1171 			     "LOM.  Please be aware there may be issues associated "
1172 			     "with your hardware.");
1173 		PMD_INIT_LOG(ERR, "If you are experiencing problems "
1174 			     "please contact your Intel or hardware representative "
1175 			     "who provided you with this hardware.");
1176 	} else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
1177 		PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
1178 	if (diag) {
1179 		PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
1180 		return -EIO;
1181 	}
1182 
1183 	/* Reset the hw statistics */
1184 	ixgbe_dev_stats_reset(eth_dev);
1185 
1186 	/* disable interrupt */
1187 	ixgbe_disable_intr(hw);
1188 
1189 	/* reset mappings for queue statistics hw counters*/
1190 	ixgbe_reset_qstat_mappings(hw);
1191 
1192 	/* Allocate memory for storing MAC addresses */
1193 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", RTE_ETHER_ADDR_LEN *
1194 					       hw->mac.num_rar_entries, 0);
1195 	if (eth_dev->data->mac_addrs == NULL) {
1196 		PMD_INIT_LOG(ERR,
1197 			     "Failed to allocate %u bytes needed to store "
1198 			     "MAC addresses",
1199 			     RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1200 		return -ENOMEM;
1201 	}
1202 	/* Copy the permanent MAC address */
1203 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.perm_addr,
1204 			&eth_dev->data->mac_addrs[0]);
1205 
1206 	/* Allocate memory for storing hash filter MAC addresses */
1207 	eth_dev->data->hash_mac_addrs = rte_zmalloc(
1208 		"ixgbe", RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC, 0);
1209 	if (eth_dev->data->hash_mac_addrs == NULL) {
1210 		PMD_INIT_LOG(ERR,
1211 			     "Failed to allocate %d bytes needed to store MAC addresses",
1212 			     RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
1213 		rte_free(eth_dev->data->mac_addrs);
1214 		eth_dev->data->mac_addrs = NULL;
1215 		return -ENOMEM;
1216 	}
1217 
1218 	/* initialize the vfta */
1219 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1220 
1221 	/* initialize the hw strip bitmap*/
1222 	memset(hwstrip, 0, sizeof(*hwstrip));
1223 
1224 	/* initialize PF if max_vfs not zero */
1225 	ret = ixgbe_pf_host_init(eth_dev);
1226 	if (ret)
1227 		goto err_pf_host_init;
1228 
1229 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1230 	/* let hardware know driver is loaded */
1231 	ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1232 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1233 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1234 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1235 	IXGBE_WRITE_FLUSH(hw);
1236 
1237 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1238 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1239 			     (int) hw->mac.type, (int) hw->phy.type,
1240 			     (int) hw->phy.sfp_type);
1241 	else
1242 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1243 			     (int) hw->mac.type, (int) hw->phy.type);
1244 
1245 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1246 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1247 		     pci_dev->id.device_id);
1248 
1249 	rte_intr_callback_register(intr_handle,
1250 				   ixgbe_dev_interrupt_handler, eth_dev);
1251 
1252 	/* enable uio/vfio intr/eventfd mapping */
1253 	rte_intr_enable(intr_handle);
1254 
1255 	/* enable support intr */
1256 	ixgbe_enable_intr(eth_dev);
1257 
1258 	/* initialize filter info */
1259 	memset(filter_info, 0,
1260 	       sizeof(struct ixgbe_filter_info));
1261 
1262 	/* initialize 5tuple filter list */
1263 	TAILQ_INIT(&filter_info->fivetuple_list);
1264 
1265 	/* initialize flow director filter list & hash */
1266 	ret = ixgbe_fdir_filter_init(eth_dev);
1267 	if (ret)
1268 		goto err_fdir_filter_init;
1269 
1270 	/* initialize l2 tunnel filter list & hash */
1271 	ret = ixgbe_l2_tn_filter_init(eth_dev);
1272 	if (ret)
1273 		goto err_l2_tn_filter_init;
1274 
1275 	/* initialize flow filter lists */
1276 	ixgbe_filterlist_init();
1277 
1278 	/* initialize bandwidth configuration info */
1279 	memset(bw_conf, 0, sizeof(struct ixgbe_bw_conf));
1280 
1281 	/* initialize Traffic Manager configuration */
1282 	ixgbe_tm_conf_init(eth_dev);
1283 
1284 	return 0;
1285 
1286 err_l2_tn_filter_init:
1287 	ixgbe_fdir_filter_uninit(eth_dev);
1288 err_fdir_filter_init:
1289 	ixgbe_disable_intr(hw);
1290 	rte_intr_disable(intr_handle);
1291 	rte_intr_callback_unregister(intr_handle,
1292 		ixgbe_dev_interrupt_handler, eth_dev);
1293 	ixgbe_pf_host_uninit(eth_dev);
1294 err_pf_host_init:
1295 	rte_free(eth_dev->data->mac_addrs);
1296 	eth_dev->data->mac_addrs = NULL;
1297 	rte_free(eth_dev->data->hash_mac_addrs);
1298 	eth_dev->data->hash_mac_addrs = NULL;
1299 	return ret;
1300 }
1301 
1302 static int
1303 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1304 {
1305 	PMD_INIT_FUNC_TRACE();
1306 
1307 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1308 		return 0;
1309 
1310 	ixgbe_dev_close(eth_dev);
1311 
1312 	return 0;
1313 }
1314 
1315 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
1316 {
1317 	struct ixgbe_filter_info *filter_info =
1318 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1319 	struct ixgbe_5tuple_filter *p_5tuple;
1320 
1321 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
1322 		TAILQ_REMOVE(&filter_info->fivetuple_list,
1323 			     p_5tuple,
1324 			     entries);
1325 		rte_free(p_5tuple);
1326 	}
1327 	memset(filter_info->fivetuple_mask, 0,
1328 	       sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1329 
1330 	return 0;
1331 }
1332 
1333 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev)
1334 {
1335 	struct ixgbe_hw_fdir_info *fdir_info =
1336 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1337 	struct ixgbe_fdir_filter *fdir_filter;
1338 
1339 		if (fdir_info->hash_map)
1340 		rte_free(fdir_info->hash_map);
1341 	if (fdir_info->hash_handle)
1342 		rte_hash_free(fdir_info->hash_handle);
1343 
1344 	while ((fdir_filter = TAILQ_FIRST(&fdir_info->fdir_list))) {
1345 		TAILQ_REMOVE(&fdir_info->fdir_list,
1346 			     fdir_filter,
1347 			     entries);
1348 		rte_free(fdir_filter);
1349 	}
1350 
1351 	return 0;
1352 }
1353 
1354 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev)
1355 {
1356 	struct ixgbe_l2_tn_info *l2_tn_info =
1357 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1358 	struct ixgbe_l2_tn_filter *l2_tn_filter;
1359 
1360 	if (l2_tn_info->hash_map)
1361 		rte_free(l2_tn_info->hash_map);
1362 	if (l2_tn_info->hash_handle)
1363 		rte_hash_free(l2_tn_info->hash_handle);
1364 
1365 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
1366 		TAILQ_REMOVE(&l2_tn_info->l2_tn_list,
1367 			     l2_tn_filter,
1368 			     entries);
1369 		rte_free(l2_tn_filter);
1370 	}
1371 
1372 	return 0;
1373 }
1374 
1375 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev)
1376 {
1377 	struct ixgbe_hw_fdir_info *fdir_info =
1378 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1379 	char fdir_hash_name[RTE_HASH_NAMESIZE];
1380 	struct rte_hash_parameters fdir_hash_params = {
1381 		.name = fdir_hash_name,
1382 		.entries = IXGBE_MAX_FDIR_FILTER_NUM,
1383 		.key_len = sizeof(union ixgbe_atr_input),
1384 		.hash_func = rte_hash_crc,
1385 		.hash_func_init_val = 0,
1386 		.socket_id = rte_socket_id(),
1387 	};
1388 
1389 	TAILQ_INIT(&fdir_info->fdir_list);
1390 	snprintf(fdir_hash_name, RTE_HASH_NAMESIZE,
1391 		 "fdir_%s", eth_dev->device->name);
1392 	fdir_info->hash_handle = rte_hash_create(&fdir_hash_params);
1393 	if (!fdir_info->hash_handle) {
1394 		PMD_INIT_LOG(ERR, "Failed to create fdir hash table!");
1395 		return -EINVAL;
1396 	}
1397 	fdir_info->hash_map = rte_zmalloc("ixgbe",
1398 					  sizeof(struct ixgbe_fdir_filter *) *
1399 					  IXGBE_MAX_FDIR_FILTER_NUM,
1400 					  0);
1401 	if (!fdir_info->hash_map) {
1402 		PMD_INIT_LOG(ERR,
1403 			     "Failed to allocate memory for fdir hash map!");
1404 		rte_hash_free(fdir_info->hash_handle);
1405 		return -ENOMEM;
1406 	}
1407 	fdir_info->mask_added = FALSE;
1408 
1409 	return 0;
1410 }
1411 
1412 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev)
1413 {
1414 	struct ixgbe_l2_tn_info *l2_tn_info =
1415 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1416 	char l2_tn_hash_name[RTE_HASH_NAMESIZE];
1417 	struct rte_hash_parameters l2_tn_hash_params = {
1418 		.name = l2_tn_hash_name,
1419 		.entries = IXGBE_MAX_L2_TN_FILTER_NUM,
1420 		.key_len = sizeof(struct ixgbe_l2_tn_key),
1421 		.hash_func = rte_hash_crc,
1422 		.hash_func_init_val = 0,
1423 		.socket_id = rte_socket_id(),
1424 	};
1425 
1426 	TAILQ_INIT(&l2_tn_info->l2_tn_list);
1427 	snprintf(l2_tn_hash_name, RTE_HASH_NAMESIZE,
1428 		 "l2_tn_%s", eth_dev->device->name);
1429 	l2_tn_info->hash_handle = rte_hash_create(&l2_tn_hash_params);
1430 	if (!l2_tn_info->hash_handle) {
1431 		PMD_INIT_LOG(ERR, "Failed to create L2 TN hash table!");
1432 		return -EINVAL;
1433 	}
1434 	l2_tn_info->hash_map = rte_zmalloc("ixgbe",
1435 				   sizeof(struct ixgbe_l2_tn_filter *) *
1436 				   IXGBE_MAX_L2_TN_FILTER_NUM,
1437 				   0);
1438 	if (!l2_tn_info->hash_map) {
1439 		PMD_INIT_LOG(ERR,
1440 			"Failed to allocate memory for L2 TN hash map!");
1441 		rte_hash_free(l2_tn_info->hash_handle);
1442 		return -ENOMEM;
1443 	}
1444 	l2_tn_info->e_tag_en = FALSE;
1445 	l2_tn_info->e_tag_fwd_en = FALSE;
1446 	l2_tn_info->e_tag_ether_type = RTE_ETHER_TYPE_ETAG;
1447 
1448 	return 0;
1449 }
1450 /*
1451  * Negotiate mailbox API version with the PF.
1452  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1453  * Then we try to negotiate starting with the most recent one.
1454  * If all negotiation attempts fail, then we will proceed with
1455  * the default one (ixgbe_mbox_api_10).
1456  */
1457 static void
1458 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1459 {
1460 	int32_t i;
1461 
1462 	/* start with highest supported, proceed down */
1463 	static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1464 		ixgbe_mbox_api_13,
1465 		ixgbe_mbox_api_12,
1466 		ixgbe_mbox_api_11,
1467 		ixgbe_mbox_api_10,
1468 	};
1469 
1470 	for (i = 0;
1471 			i != RTE_DIM(sup_ver) &&
1472 			ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1473 			i++)
1474 		;
1475 }
1476 
1477 static void
1478 generate_random_mac_addr(struct rte_ether_addr *mac_addr)
1479 {
1480 	uint64_t random;
1481 
1482 	/* Set Organizationally Unique Identifier (OUI) prefix. */
1483 	mac_addr->addr_bytes[0] = 0x00;
1484 	mac_addr->addr_bytes[1] = 0x09;
1485 	mac_addr->addr_bytes[2] = 0xC0;
1486 	/* Force indication of locally assigned MAC address. */
1487 	mac_addr->addr_bytes[0] |= RTE_ETHER_LOCAL_ADMIN_ADDR;
1488 	/* Generate the last 3 bytes of the MAC address with a random number. */
1489 	random = rte_rand();
1490 	memcpy(&mac_addr->addr_bytes[3], &random, 3);
1491 }
1492 
1493 static int
1494 devarg_handle_int(__rte_unused const char *key, const char *value,
1495 		  void *extra_args)
1496 {
1497 	uint16_t *n = extra_args;
1498 
1499 	if (value == NULL || extra_args == NULL)
1500 		return -EINVAL;
1501 
1502 	*n = (uint16_t)strtoul(value, NULL, 0);
1503 	if (*n == USHRT_MAX && errno == ERANGE)
1504 		return -1;
1505 
1506 	return 0;
1507 }
1508 
1509 static void
1510 ixgbevf_parse_devargs(struct ixgbe_adapter *adapter,
1511 		      struct rte_devargs *devargs)
1512 {
1513 	struct rte_kvargs *kvlist;
1514 	uint16_t pflink_fullchk;
1515 
1516 	if (devargs == NULL)
1517 		return;
1518 
1519 	kvlist = rte_kvargs_parse(devargs->args, ixgbevf_valid_arguments);
1520 	if (kvlist == NULL)
1521 		return;
1522 
1523 	if (rte_kvargs_count(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK) == 1 &&
1524 	    rte_kvargs_process(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK,
1525 			       devarg_handle_int, &pflink_fullchk) == 0 &&
1526 	    pflink_fullchk == 1)
1527 		adapter->pflink_fullchk = 1;
1528 
1529 	rte_kvargs_free(kvlist);
1530 }
1531 
1532 /*
1533  * Virtual Function device init
1534  */
1535 static int
1536 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1537 {
1538 	int diag;
1539 	uint32_t tc, tcs;
1540 	struct ixgbe_adapter *ad = eth_dev->data->dev_private;
1541 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1542 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1543 	struct ixgbe_hw *hw =
1544 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1545 	struct ixgbe_vfta *shadow_vfta =
1546 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1547 	struct ixgbe_hwstrip *hwstrip =
1548 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1549 	struct rte_ether_addr *perm_addr =
1550 		(struct rte_ether_addr *)hw->mac.perm_addr;
1551 
1552 	PMD_INIT_FUNC_TRACE();
1553 
1554 	eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1555 	eth_dev->rx_descriptor_status = ixgbe_dev_rx_descriptor_status;
1556 	eth_dev->tx_descriptor_status = ixgbe_dev_tx_descriptor_status;
1557 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1558 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1559 
1560 	/* for secondary processes, we don't initialise any further as primary
1561 	 * has already done this work. Only check we don't need a different
1562 	 * RX function
1563 	 */
1564 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1565 		struct ixgbe_tx_queue *txq;
1566 		/* TX queue function in primary, set by last queue initialized
1567 		 * Tx queue may not initialized by primary process
1568 		 */
1569 		if (eth_dev->data->tx_queues) {
1570 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1571 			ixgbe_set_tx_function(eth_dev, txq);
1572 		} else {
1573 			/* Use default TX function if we get here */
1574 			PMD_INIT_LOG(NOTICE,
1575 				     "No TX queues configured yet. Using default TX function.");
1576 		}
1577 
1578 		ixgbe_set_rx_function(eth_dev);
1579 
1580 		return 0;
1581 	}
1582 
1583 	rte_atomic32_clear(&ad->link_thread_running);
1584 	ixgbevf_parse_devargs(eth_dev->data->dev_private,
1585 			      pci_dev->device.devargs);
1586 
1587 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1588 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1589 
1590 	hw->device_id = pci_dev->id.device_id;
1591 	hw->vendor_id = pci_dev->id.vendor_id;
1592 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1593 
1594 	/* initialize the vfta */
1595 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1596 
1597 	/* initialize the hw strip bitmap*/
1598 	memset(hwstrip, 0, sizeof(*hwstrip));
1599 
1600 	/* Initialize the shared code (base driver) */
1601 	diag = ixgbe_init_shared_code(hw);
1602 	if (diag != IXGBE_SUCCESS) {
1603 		PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1604 		return -EIO;
1605 	}
1606 
1607 	/* init_mailbox_params */
1608 	hw->mbx.ops.init_params(hw);
1609 
1610 	/* Reset the hw statistics */
1611 	ixgbevf_dev_stats_reset(eth_dev);
1612 
1613 	/* Disable the interrupts for VF */
1614 	ixgbevf_intr_disable(eth_dev);
1615 
1616 	hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1617 	diag = hw->mac.ops.reset_hw(hw);
1618 
1619 	/*
1620 	 * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1621 	 * the underlying PF driver has not assigned a MAC address to the VF.
1622 	 * In this case, assign a random MAC address.
1623 	 */
1624 	if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1625 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1626 		/*
1627 		 * This error code will be propagated to the app by
1628 		 * rte_eth_dev_reset, so use a public error code rather than
1629 		 * the internal-only IXGBE_ERR_RESET_FAILED
1630 		 */
1631 		return -EAGAIN;
1632 	}
1633 
1634 	/* negotiate mailbox API version to use with the PF. */
1635 	ixgbevf_negotiate_api(hw);
1636 
1637 	/* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1638 	ixgbevf_get_queues(hw, &tcs, &tc);
1639 
1640 	/* Allocate memory for storing MAC addresses */
1641 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", RTE_ETHER_ADDR_LEN *
1642 					       hw->mac.num_rar_entries, 0);
1643 	if (eth_dev->data->mac_addrs == NULL) {
1644 		PMD_INIT_LOG(ERR,
1645 			     "Failed to allocate %u bytes needed to store "
1646 			     "MAC addresses",
1647 			     RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1648 		return -ENOMEM;
1649 	}
1650 
1651 	/* Generate a random MAC address, if none was assigned by PF. */
1652 	if (rte_is_zero_ether_addr(perm_addr)) {
1653 		generate_random_mac_addr(perm_addr);
1654 		diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1655 		if (diag) {
1656 			rte_free(eth_dev->data->mac_addrs);
1657 			eth_dev->data->mac_addrs = NULL;
1658 			return diag;
1659 		}
1660 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1661 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1662 			     RTE_ETHER_ADDR_PRT_FMT,
1663 				 RTE_ETHER_ADDR_BYTES(perm_addr));
1664 	}
1665 
1666 	/* Copy the permanent MAC address */
1667 	rte_ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1668 
1669 	/* reset the hardware with the new settings */
1670 	diag = hw->mac.ops.start_hw(hw);
1671 	switch (diag) {
1672 	case  0:
1673 		break;
1674 
1675 	default:
1676 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1677 		rte_free(eth_dev->data->mac_addrs);
1678 		eth_dev->data->mac_addrs = NULL;
1679 		return -EIO;
1680 	}
1681 
1682 	rte_intr_callback_register(intr_handle,
1683 				   ixgbevf_dev_interrupt_handler, eth_dev);
1684 	rte_intr_enable(intr_handle);
1685 	ixgbevf_intr_enable(eth_dev);
1686 
1687 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1688 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1689 		     pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1690 
1691 	return 0;
1692 }
1693 
1694 /* Virtual Function device uninit */
1695 
1696 static int
1697 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1698 {
1699 	PMD_INIT_FUNC_TRACE();
1700 
1701 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1702 		return 0;
1703 
1704 	ixgbevf_dev_close(eth_dev);
1705 
1706 	return 0;
1707 }
1708 
1709 static int
1710 eth_ixgbe_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1711 		struct rte_pci_device *pci_dev)
1712 {
1713 	char name[RTE_ETH_NAME_MAX_LEN];
1714 	struct rte_eth_dev *pf_ethdev;
1715 	struct rte_eth_devargs eth_da;
1716 	int i, retval;
1717 
1718 	if (pci_dev->device.devargs) {
1719 		retval = rte_eth_devargs_parse(pci_dev->device.devargs->args,
1720 				&eth_da);
1721 		if (retval)
1722 			return retval;
1723 	} else
1724 		memset(&eth_da, 0, sizeof(eth_da));
1725 
1726 	if (eth_da.nb_representor_ports > 0 &&
1727 	    eth_da.type != RTE_ETH_REPRESENTOR_VF) {
1728 		PMD_DRV_LOG(ERR, "unsupported representor type: %s\n",
1729 			    pci_dev->device.devargs->args);
1730 		return -ENOTSUP;
1731 	}
1732 
1733 	retval = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
1734 		sizeof(struct ixgbe_adapter),
1735 		eth_dev_pci_specific_init, pci_dev,
1736 		eth_ixgbe_dev_init, NULL);
1737 
1738 	if (retval || eth_da.nb_representor_ports < 1)
1739 		return retval;
1740 
1741 	pf_ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1742 	if (pf_ethdev == NULL)
1743 		return -ENODEV;
1744 
1745 	/* probe VF representor ports */
1746 	for (i = 0; i < eth_da.nb_representor_ports; i++) {
1747 		struct ixgbe_vf_info *vfinfo;
1748 		struct ixgbe_vf_representor representor;
1749 
1750 		vfinfo = *IXGBE_DEV_PRIVATE_TO_P_VFDATA(
1751 			pf_ethdev->data->dev_private);
1752 		if (vfinfo == NULL) {
1753 			PMD_DRV_LOG(ERR,
1754 				"no virtual functions supported by PF");
1755 			break;
1756 		}
1757 
1758 		representor.vf_id = eth_da.representor_ports[i];
1759 		representor.switch_domain_id = vfinfo->switch_domain_id;
1760 		representor.pf_ethdev = pf_ethdev;
1761 
1762 		/* representor port net_bdf_port */
1763 		snprintf(name, sizeof(name), "net_%s_representor_%d",
1764 			pci_dev->device.name,
1765 			eth_da.representor_ports[i]);
1766 
1767 		retval = rte_eth_dev_create(&pci_dev->device, name,
1768 			sizeof(struct ixgbe_vf_representor), NULL, NULL,
1769 			ixgbe_vf_representor_init, &representor);
1770 
1771 		if (retval)
1772 			PMD_DRV_LOG(ERR, "failed to create ixgbe vf "
1773 				"representor %s.", name);
1774 	}
1775 
1776 	return 0;
1777 }
1778 
1779 static int eth_ixgbe_pci_remove(struct rte_pci_device *pci_dev)
1780 {
1781 	struct rte_eth_dev *ethdev;
1782 
1783 	ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1784 	if (!ethdev)
1785 		return 0;
1786 
1787 	if (ethdev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
1788 		return rte_eth_dev_pci_generic_remove(pci_dev,
1789 					ixgbe_vf_representor_uninit);
1790 	else
1791 		return rte_eth_dev_pci_generic_remove(pci_dev,
1792 						eth_ixgbe_dev_uninit);
1793 }
1794 
1795 static struct rte_pci_driver rte_ixgbe_pmd = {
1796 	.id_table = pci_id_ixgbe_map,
1797 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1798 	.probe = eth_ixgbe_pci_probe,
1799 	.remove = eth_ixgbe_pci_remove,
1800 };
1801 
1802 static int eth_ixgbevf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1803 	struct rte_pci_device *pci_dev)
1804 {
1805 	return rte_eth_dev_pci_generic_probe(pci_dev,
1806 		sizeof(struct ixgbe_adapter), eth_ixgbevf_dev_init);
1807 }
1808 
1809 static int eth_ixgbevf_pci_remove(struct rte_pci_device *pci_dev)
1810 {
1811 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbevf_dev_uninit);
1812 }
1813 
1814 /*
1815  * virtual function driver struct
1816  */
1817 static struct rte_pci_driver rte_ixgbevf_pmd = {
1818 	.id_table = pci_id_ixgbevf_map,
1819 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1820 	.probe = eth_ixgbevf_pci_probe,
1821 	.remove = eth_ixgbevf_pci_remove,
1822 };
1823 
1824 static int
1825 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1826 {
1827 	struct ixgbe_hw *hw =
1828 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1829 	struct ixgbe_vfta *shadow_vfta =
1830 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1831 	uint32_t vfta;
1832 	uint32_t vid_idx;
1833 	uint32_t vid_bit;
1834 
1835 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1836 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1837 	vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1838 	if (on)
1839 		vfta |= vid_bit;
1840 	else
1841 		vfta &= ~vid_bit;
1842 	IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1843 
1844 	/* update local VFTA copy */
1845 	shadow_vfta->vfta[vid_idx] = vfta;
1846 
1847 	return 0;
1848 }
1849 
1850 static void
1851 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1852 {
1853 	if (on)
1854 		ixgbe_vlan_hw_strip_enable(dev, queue);
1855 	else
1856 		ixgbe_vlan_hw_strip_disable(dev, queue);
1857 }
1858 
1859 static int
1860 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1861 		    enum rte_vlan_type vlan_type,
1862 		    uint16_t tpid)
1863 {
1864 	struct ixgbe_hw *hw =
1865 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1866 	int ret = 0;
1867 	uint32_t reg;
1868 	uint32_t qinq;
1869 
1870 	qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1871 	qinq &= IXGBE_DMATXCTL_GDV;
1872 
1873 	switch (vlan_type) {
1874 	case RTE_ETH_VLAN_TYPE_INNER:
1875 		if (qinq) {
1876 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1877 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1878 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1879 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1880 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1881 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1882 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1883 		} else {
1884 			ret = -ENOTSUP;
1885 			PMD_DRV_LOG(ERR, "Inner type is not supported"
1886 				    " by single VLAN");
1887 		}
1888 		break;
1889 	case RTE_ETH_VLAN_TYPE_OUTER:
1890 		if (qinq) {
1891 			/* Only the high 16-bits is valid */
1892 			IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1893 					IXGBE_EXVET_VET_EXT_SHIFT);
1894 		} else {
1895 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1896 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1897 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1898 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1899 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1900 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1901 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1902 		}
1903 
1904 		break;
1905 	default:
1906 		ret = -EINVAL;
1907 		PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1908 		break;
1909 	}
1910 
1911 	return ret;
1912 }
1913 
1914 void
1915 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1916 {
1917 	struct ixgbe_hw *hw =
1918 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1919 	uint32_t vlnctrl;
1920 
1921 	PMD_INIT_FUNC_TRACE();
1922 
1923 	/* Filter Table Disable */
1924 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1925 	vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1926 
1927 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1928 }
1929 
1930 void
1931 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1932 {
1933 	struct ixgbe_hw *hw =
1934 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1935 	struct ixgbe_vfta *shadow_vfta =
1936 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1937 	uint32_t vlnctrl;
1938 	uint16_t i;
1939 
1940 	PMD_INIT_FUNC_TRACE();
1941 
1942 	/* Filter Table Enable */
1943 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1944 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1945 	vlnctrl |= IXGBE_VLNCTRL_VFE;
1946 
1947 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1948 
1949 	/* write whatever is in local vfta copy */
1950 	for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1951 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1952 }
1953 
1954 static void
1955 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1956 {
1957 	struct ixgbe_hwstrip *hwstrip =
1958 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1959 	struct ixgbe_rx_queue *rxq;
1960 
1961 	if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
1962 		return;
1963 
1964 	if (on)
1965 		IXGBE_SET_HWSTRIP(hwstrip, queue);
1966 	else
1967 		IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1968 
1969 	if (queue >= dev->data->nb_rx_queues)
1970 		return;
1971 
1972 	rxq = dev->data->rx_queues[queue];
1973 
1974 	if (on) {
1975 		rxq->vlan_flags = RTE_MBUF_F_RX_VLAN | RTE_MBUF_F_RX_VLAN_STRIPPED;
1976 		rxq->offloads |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
1977 	} else {
1978 		rxq->vlan_flags = RTE_MBUF_F_RX_VLAN;
1979 		rxq->offloads &= ~RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
1980 	}
1981 }
1982 
1983 static void
1984 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
1985 {
1986 	struct ixgbe_hw *hw =
1987 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1988 	uint32_t ctrl;
1989 
1990 	PMD_INIT_FUNC_TRACE();
1991 
1992 	if (hw->mac.type == ixgbe_mac_82598EB) {
1993 		/* No queue level support */
1994 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1995 		return;
1996 	}
1997 
1998 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1999 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2000 	ctrl &= ~IXGBE_RXDCTL_VME;
2001 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2002 
2003 	/* record those setting for HW strip per queue */
2004 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
2005 }
2006 
2007 static void
2008 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
2009 {
2010 	struct ixgbe_hw *hw =
2011 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2012 	uint32_t ctrl;
2013 
2014 	PMD_INIT_FUNC_TRACE();
2015 
2016 	if (hw->mac.type == ixgbe_mac_82598EB) {
2017 		/* No queue level supported */
2018 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2019 		return;
2020 	}
2021 
2022 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2023 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2024 	ctrl |= IXGBE_RXDCTL_VME;
2025 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2026 
2027 	/* record those setting for HW strip per queue */
2028 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
2029 }
2030 
2031 static void
2032 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2033 {
2034 	struct ixgbe_hw *hw =
2035 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2036 	uint32_t ctrl;
2037 
2038 	PMD_INIT_FUNC_TRACE();
2039 
2040 	/* DMATXCTRL: Geric Double VLAN Disable */
2041 	ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2042 	ctrl &= ~IXGBE_DMATXCTL_GDV;
2043 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2044 
2045 	/* CTRL_EXT: Global Double VLAN Disable */
2046 	ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2047 	ctrl &= ~IXGBE_EXTENDED_VLAN;
2048 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2049 
2050 }
2051 
2052 static void
2053 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2054 {
2055 	struct ixgbe_hw *hw =
2056 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2057 	uint32_t ctrl;
2058 
2059 	PMD_INIT_FUNC_TRACE();
2060 
2061 	/* DMATXCTRL: Geric Double VLAN Enable */
2062 	ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2063 	ctrl |= IXGBE_DMATXCTL_GDV;
2064 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2065 
2066 	/* CTRL_EXT: Global Double VLAN Enable */
2067 	ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2068 	ctrl |= IXGBE_EXTENDED_VLAN;
2069 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2070 
2071 	/* Clear pooling mode of PFVTCTL. It's required by X550. */
2072 	if (hw->mac.type == ixgbe_mac_X550 ||
2073 	    hw->mac.type == ixgbe_mac_X550EM_x ||
2074 	    hw->mac.type == ixgbe_mac_X550EM_a) {
2075 		ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
2076 		ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
2077 		IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
2078 	}
2079 
2080 	/*
2081 	 * VET EXT field in the EXVET register = 0x8100 by default
2082 	 * So no need to change. Same to VT field of DMATXCTL register
2083 	 */
2084 }
2085 
2086 void
2087 ixgbe_vlan_hw_strip_config(struct rte_eth_dev *dev)
2088 {
2089 	struct ixgbe_hw *hw =
2090 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2091 	struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
2092 	uint32_t ctrl;
2093 	uint16_t i;
2094 	struct ixgbe_rx_queue *rxq;
2095 	bool on;
2096 
2097 	PMD_INIT_FUNC_TRACE();
2098 
2099 	if (hw->mac.type == ixgbe_mac_82598EB) {
2100 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP) {
2101 			ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2102 			ctrl |= IXGBE_VLNCTRL_VME;
2103 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2104 		} else {
2105 			ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2106 			ctrl &= ~IXGBE_VLNCTRL_VME;
2107 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2108 		}
2109 	} else {
2110 		/*
2111 		 * Other 10G NIC, the VLAN strip can be setup
2112 		 * per queue in RXDCTL
2113 		 */
2114 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
2115 			rxq = dev->data->rx_queues[i];
2116 			ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2117 			if (rxq->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP) {
2118 				ctrl |= IXGBE_RXDCTL_VME;
2119 				on = TRUE;
2120 			} else {
2121 				ctrl &= ~IXGBE_RXDCTL_VME;
2122 				on = FALSE;
2123 			}
2124 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2125 
2126 			/* record those setting for HW strip per queue */
2127 			ixgbe_vlan_hw_strip_bitmap_set(dev, i, on);
2128 		}
2129 	}
2130 }
2131 
2132 static void
2133 ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev, int mask)
2134 {
2135 	uint16_t i;
2136 	struct rte_eth_rxmode *rxmode;
2137 	struct ixgbe_rx_queue *rxq;
2138 
2139 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
2140 		rxmode = &dev->data->dev_conf.rxmode;
2141 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
2142 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
2143 				rxq = dev->data->rx_queues[i];
2144 				rxq->offloads |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
2145 			}
2146 		else
2147 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
2148 				rxq = dev->data->rx_queues[i];
2149 				rxq->offloads &= ~RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
2150 			}
2151 	}
2152 }
2153 
2154 static int
2155 ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask)
2156 {
2157 	struct rte_eth_rxmode *rxmode;
2158 	rxmode = &dev->data->dev_conf.rxmode;
2159 
2160 	if (mask & RTE_ETH_VLAN_STRIP_MASK)
2161 		ixgbe_vlan_hw_strip_config(dev);
2162 
2163 	if (mask & RTE_ETH_VLAN_FILTER_MASK) {
2164 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
2165 			ixgbe_vlan_hw_filter_enable(dev);
2166 		else
2167 			ixgbe_vlan_hw_filter_disable(dev);
2168 	}
2169 
2170 	if (mask & RTE_ETH_VLAN_EXTEND_MASK) {
2171 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND)
2172 			ixgbe_vlan_hw_extend_enable(dev);
2173 		else
2174 			ixgbe_vlan_hw_extend_disable(dev);
2175 	}
2176 
2177 	return 0;
2178 }
2179 
2180 static int
2181 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2182 {
2183 	ixgbe_config_vlan_strip_on_all_queues(dev, mask);
2184 
2185 	ixgbe_vlan_offload_config(dev, mask);
2186 
2187 	return 0;
2188 }
2189 
2190 static void
2191 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2192 {
2193 	struct ixgbe_hw *hw =
2194 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2195 	/* VLNCTRL: enable vlan filtering and allow all vlan tags through */
2196 	uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2197 
2198 	vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
2199 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
2200 }
2201 
2202 static int
2203 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
2204 {
2205 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2206 
2207 	switch (nb_rx_q) {
2208 	case 1:
2209 	case 2:
2210 		RTE_ETH_DEV_SRIOV(dev).active = RTE_ETH_64_POOLS;
2211 		break;
2212 	case 4:
2213 		RTE_ETH_DEV_SRIOV(dev).active = RTE_ETH_32_POOLS;
2214 		break;
2215 	default:
2216 		return -EINVAL;
2217 	}
2218 
2219 	RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
2220 		IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2221 	RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
2222 		pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2223 	return 0;
2224 }
2225 
2226 static int
2227 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
2228 {
2229 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
2230 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2231 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
2232 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
2233 
2234 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
2235 		/* check multi-queue mode */
2236 		switch (dev_conf->rxmode.mq_mode) {
2237 		case RTE_ETH_MQ_RX_VMDQ_DCB:
2238 			PMD_INIT_LOG(INFO, "RTE_ETH_MQ_RX_VMDQ_DCB mode supported in SRIOV");
2239 			break;
2240 		case RTE_ETH_MQ_RX_VMDQ_DCB_RSS:
2241 			/* DCB/RSS VMDQ in SRIOV mode, not implement yet */
2242 			PMD_INIT_LOG(ERR, "SRIOV active,"
2243 					" unsupported mq_mode rx %d.",
2244 					dev_conf->rxmode.mq_mode);
2245 			return -EINVAL;
2246 		case RTE_ETH_MQ_RX_RSS:
2247 		case RTE_ETH_MQ_RX_VMDQ_RSS:
2248 			dev->data->dev_conf.rxmode.mq_mode = RTE_ETH_MQ_RX_VMDQ_RSS;
2249 			if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
2250 				if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
2251 					PMD_INIT_LOG(ERR, "SRIOV is active,"
2252 						" invalid queue number"
2253 						" for VMDQ RSS, allowed"
2254 						" value are 1, 2 or 4.");
2255 					return -EINVAL;
2256 				}
2257 			break;
2258 		case RTE_ETH_MQ_RX_VMDQ_ONLY:
2259 		case RTE_ETH_MQ_RX_NONE:
2260 			/* if nothing mq mode configure, use default scheme */
2261 			dev->data->dev_conf.rxmode.mq_mode = RTE_ETH_MQ_RX_VMDQ_ONLY;
2262 			break;
2263 		default: /* RTE_ETH_MQ_RX_DCB, RTE_ETH_MQ_RX_DCB_RSS or RTE_ETH_MQ_TX_DCB*/
2264 			/* SRIOV only works in VMDq enable mode */
2265 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2266 					" wrong mq_mode rx %d.",
2267 					dev_conf->rxmode.mq_mode);
2268 			return -EINVAL;
2269 		}
2270 
2271 		switch (dev_conf->txmode.mq_mode) {
2272 		case RTE_ETH_MQ_TX_VMDQ_DCB:
2273 			PMD_INIT_LOG(INFO, "RTE_ETH_MQ_TX_VMDQ_DCB mode supported in SRIOV");
2274 			dev->data->dev_conf.txmode.mq_mode = RTE_ETH_MQ_TX_VMDQ_DCB;
2275 			break;
2276 		default: /* RTE_ETH_MQ_TX_VMDQ_ONLY or RTE_ETH_MQ_TX_NONE */
2277 			dev->data->dev_conf.txmode.mq_mode = RTE_ETH_MQ_TX_VMDQ_ONLY;
2278 			break;
2279 		}
2280 
2281 		/* check valid queue number */
2282 		if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2283 		    (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2284 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2285 					" nb_rx_q=%d nb_tx_q=%d queue number"
2286 					" must be less than or equal to %d.",
2287 					nb_rx_q, nb_tx_q,
2288 					RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2289 			return -EINVAL;
2290 		}
2291 	} else {
2292 		if (dev_conf->rxmode.mq_mode == RTE_ETH_MQ_RX_VMDQ_DCB_RSS) {
2293 			PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2294 					  " not supported.");
2295 			return -EINVAL;
2296 		}
2297 		/* check configuration for vmdb+dcb mode */
2298 		if (dev_conf->rxmode.mq_mode == RTE_ETH_MQ_RX_VMDQ_DCB) {
2299 			const struct rte_eth_vmdq_dcb_conf *conf;
2300 
2301 			if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2302 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2303 						IXGBE_VMDQ_DCB_NB_QUEUES);
2304 				return -EINVAL;
2305 			}
2306 			conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2307 			if (!(conf->nb_queue_pools == RTE_ETH_16_POOLS ||
2308 			       conf->nb_queue_pools == RTE_ETH_32_POOLS)) {
2309 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2310 						" nb_queue_pools must be %d or %d.",
2311 						RTE_ETH_16_POOLS, RTE_ETH_32_POOLS);
2312 				return -EINVAL;
2313 			}
2314 		}
2315 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_VMDQ_DCB) {
2316 			const struct rte_eth_vmdq_dcb_tx_conf *conf;
2317 
2318 			if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2319 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2320 						 IXGBE_VMDQ_DCB_NB_QUEUES);
2321 				return -EINVAL;
2322 			}
2323 			conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2324 			if (!(conf->nb_queue_pools == RTE_ETH_16_POOLS ||
2325 			       conf->nb_queue_pools == RTE_ETH_32_POOLS)) {
2326 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2327 						" nb_queue_pools != %d and"
2328 						" nb_queue_pools != %d.",
2329 						RTE_ETH_16_POOLS, RTE_ETH_32_POOLS);
2330 				return -EINVAL;
2331 			}
2332 		}
2333 
2334 		/* For DCB mode check our configuration before we go further */
2335 		if (dev_conf->rxmode.mq_mode == RTE_ETH_MQ_RX_DCB) {
2336 			const struct rte_eth_dcb_rx_conf *conf;
2337 
2338 			conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2339 			if (!(conf->nb_tcs == RTE_ETH_4_TCS ||
2340 			       conf->nb_tcs == RTE_ETH_8_TCS)) {
2341 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2342 						" and nb_tcs != %d.",
2343 						RTE_ETH_4_TCS, RTE_ETH_8_TCS);
2344 				return -EINVAL;
2345 			}
2346 		}
2347 
2348 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_DCB) {
2349 			const struct rte_eth_dcb_tx_conf *conf;
2350 
2351 			conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2352 			if (!(conf->nb_tcs == RTE_ETH_4_TCS ||
2353 			       conf->nb_tcs == RTE_ETH_8_TCS)) {
2354 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2355 						" and nb_tcs != %d.",
2356 						RTE_ETH_4_TCS, RTE_ETH_8_TCS);
2357 				return -EINVAL;
2358 			}
2359 		}
2360 
2361 		/*
2362 		 * When DCB/VT is off, maximum number of queues changes,
2363 		 * except for 82598EB, which remains constant.
2364 		 */
2365 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_NONE &&
2366 				hw->mac.type != ixgbe_mac_82598EB) {
2367 			if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2368 				PMD_INIT_LOG(ERR,
2369 					     "Neither VT nor DCB are enabled, "
2370 					     "nb_tx_q > %d.",
2371 					     IXGBE_NONE_MODE_TX_NB_QUEUES);
2372 				return -EINVAL;
2373 			}
2374 		}
2375 	}
2376 	return 0;
2377 }
2378 
2379 static int
2380 ixgbe_dev_configure(struct rte_eth_dev *dev)
2381 {
2382 	struct ixgbe_interrupt *intr =
2383 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2384 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2385 	int ret;
2386 
2387 	PMD_INIT_FUNC_TRACE();
2388 
2389 	if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
2390 		dev->data->dev_conf.rxmode.offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
2391 
2392 	/* multiple queue mode checking */
2393 	ret  = ixgbe_check_mq_mode(dev);
2394 	if (ret != 0) {
2395 		PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2396 			    ret);
2397 		return ret;
2398 	}
2399 
2400 	/* set flag to update link status after init */
2401 	intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2402 
2403 	/*
2404 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2405 	 * allocation or vector Rx preconditions we will reset it.
2406 	 */
2407 	adapter->rx_bulk_alloc_allowed = true;
2408 	adapter->rx_vec_allowed = true;
2409 
2410 	return 0;
2411 }
2412 
2413 static void
2414 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2415 {
2416 	struct ixgbe_hw *hw =
2417 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2418 	struct ixgbe_interrupt *intr =
2419 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2420 	uint32_t gpie;
2421 
2422 	/* only set up it on X550EM_X */
2423 	if (hw->mac.type == ixgbe_mac_X550EM_x) {
2424 		gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2425 		gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2426 		IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2427 		if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2428 			intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2429 	}
2430 }
2431 
2432 int
2433 ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
2434 			uint16_t tx_rate, uint64_t q_msk)
2435 {
2436 	struct ixgbe_hw *hw;
2437 	struct ixgbe_vf_info *vfinfo;
2438 	struct rte_eth_link link;
2439 	uint8_t  nb_q_per_pool;
2440 	uint32_t queue_stride;
2441 	uint32_t queue_idx, idx = 0, vf_idx;
2442 	uint32_t queue_end;
2443 	uint16_t total_rate = 0;
2444 	struct rte_pci_device *pci_dev;
2445 	int ret;
2446 
2447 	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2448 	ret = rte_eth_link_get_nowait(dev->data->port_id, &link);
2449 	if (ret < 0)
2450 		return ret;
2451 
2452 	if (vf >= pci_dev->max_vfs)
2453 		return -EINVAL;
2454 
2455 	if (tx_rate > link.link_speed)
2456 		return -EINVAL;
2457 
2458 	if (q_msk == 0)
2459 		return 0;
2460 
2461 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2462 	vfinfo = *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
2463 	nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2464 	queue_stride = IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2465 	queue_idx = vf * queue_stride;
2466 	queue_end = queue_idx + nb_q_per_pool - 1;
2467 	if (queue_end >= hw->mac.max_tx_queues)
2468 		return -EINVAL;
2469 
2470 	if (vfinfo) {
2471 		for (vf_idx = 0; vf_idx < pci_dev->max_vfs; vf_idx++) {
2472 			if (vf_idx == vf)
2473 				continue;
2474 			for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
2475 				idx++)
2476 				total_rate += vfinfo[vf_idx].tx_rate[idx];
2477 		}
2478 	} else {
2479 		return -EINVAL;
2480 	}
2481 
2482 	/* Store tx_rate for this vf. */
2483 	for (idx = 0; idx < nb_q_per_pool; idx++) {
2484 		if (((uint64_t)0x1 << idx) & q_msk) {
2485 			if (vfinfo[vf].tx_rate[idx] != tx_rate)
2486 				vfinfo[vf].tx_rate[idx] = tx_rate;
2487 			total_rate += tx_rate;
2488 		}
2489 	}
2490 
2491 	if (total_rate > dev->data->dev_link.link_speed) {
2492 		/* Reset stored TX rate of the VF if it causes exceed
2493 		 * link speed.
2494 		 */
2495 		memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
2496 		return -EINVAL;
2497 	}
2498 
2499 	/* Set RTTBCNRC of each queue/pool for vf X  */
2500 	for (; queue_idx <= queue_end; queue_idx++) {
2501 		if (0x1 & q_msk)
2502 			ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
2503 		q_msk = q_msk >> 1;
2504 	}
2505 
2506 	return 0;
2507 }
2508 
2509 static int
2510 ixgbe_flow_ctrl_enable(struct rte_eth_dev *dev, struct ixgbe_hw *hw)
2511 {
2512 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2513 	int err;
2514 	uint32_t mflcn;
2515 
2516 	ixgbe_setup_fc(hw);
2517 
2518 	err = ixgbe_fc_enable(hw);
2519 
2520 	/* Not negotiated is not an error case */
2521 	if (err == IXGBE_SUCCESS || err == IXGBE_ERR_FC_NOT_NEGOTIATED) {
2522 		/*
2523 		 *check if we want to forward MAC frames - driver doesn't
2524 		 *have native capability to do that,
2525 		 *so we'll write the registers ourselves
2526 		 */
2527 
2528 		mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
2529 
2530 		/* set or clear MFLCN.PMCF bit depending on configuration */
2531 		if (adapter->mac_ctrl_frame_fwd != 0)
2532 			mflcn |= IXGBE_MFLCN_PMCF;
2533 		else
2534 			mflcn &= ~IXGBE_MFLCN_PMCF;
2535 
2536 		IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
2537 		IXGBE_WRITE_FLUSH(hw);
2538 
2539 		return 0;
2540 	}
2541 	return err;
2542 }
2543 
2544 /*
2545  * Configure device link speed and setup link.
2546  * It returns 0 on success.
2547  */
2548 static int
2549 ixgbe_dev_start(struct rte_eth_dev *dev)
2550 {
2551 	struct ixgbe_hw *hw =
2552 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2553 	struct ixgbe_vf_info *vfinfo =
2554 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2555 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2556 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2557 	uint32_t intr_vector = 0;
2558 	int err;
2559 	bool link_up = false, negotiate = 0;
2560 	uint32_t speed = 0;
2561 	uint32_t allowed_speeds = 0;
2562 	int mask = 0;
2563 	int status;
2564 	uint16_t vf, idx;
2565 	uint32_t *link_speeds;
2566 	struct ixgbe_tm_conf *tm_conf =
2567 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2568 	struct ixgbe_macsec_setting *macsec_setting =
2569 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
2570 
2571 	PMD_INIT_FUNC_TRACE();
2572 
2573 	/* Stop the link setup handler before resetting the HW. */
2574 	ixgbe_dev_wait_setup_link_complete(dev, 0);
2575 
2576 	/* disable uio/vfio intr/eventfd mapping */
2577 	rte_intr_disable(intr_handle);
2578 
2579 	/* stop adapter */
2580 	hw->adapter_stopped = 0;
2581 	ixgbe_stop_adapter(hw);
2582 
2583 	/* reinitialize adapter
2584 	 * this calls reset and start
2585 	 */
2586 	status = ixgbe_pf_reset_hw(hw);
2587 	if (status != 0)
2588 		return -1;
2589 	hw->mac.ops.start_hw(hw);
2590 	hw->mac.get_link_status = true;
2591 
2592 	/* configure PF module if SRIOV enabled */
2593 	ixgbe_pf_host_configure(dev);
2594 
2595 	ixgbe_dev_phy_intr_setup(dev);
2596 
2597 	/* check and configure queue intr-vector mapping */
2598 	if ((rte_intr_cap_multiple(intr_handle) ||
2599 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
2600 	    dev->data->dev_conf.intr_conf.rxq != 0) {
2601 		intr_vector = dev->data->nb_rx_queues;
2602 		if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2603 			PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2604 					IXGBE_MAX_INTR_QUEUE_NUM);
2605 			return -ENOTSUP;
2606 		}
2607 		if (rte_intr_efd_enable(intr_handle, intr_vector))
2608 			return -1;
2609 	}
2610 
2611 	if (rte_intr_dp_is_en(intr_handle)) {
2612 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
2613 						   dev->data->nb_rx_queues)) {
2614 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2615 				     " intr_vec", dev->data->nb_rx_queues);
2616 			return -ENOMEM;
2617 		}
2618 	}
2619 
2620 	/* configure MSI-X for sleep until Rx interrupt */
2621 	ixgbe_configure_msix(dev);
2622 
2623 	/* initialize transmission unit */
2624 	ixgbe_dev_tx_init(dev);
2625 
2626 	/* This can fail when allocating mbufs for descriptor rings */
2627 	err = ixgbe_dev_rx_init(dev);
2628 	if (err) {
2629 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2630 		goto error;
2631 	}
2632 
2633 	mask = RTE_ETH_VLAN_STRIP_MASK | RTE_ETH_VLAN_FILTER_MASK |
2634 		RTE_ETH_VLAN_EXTEND_MASK;
2635 	err = ixgbe_vlan_offload_config(dev, mask);
2636 	if (err) {
2637 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload");
2638 		goto error;
2639 	}
2640 
2641 	if (dev->data->dev_conf.rxmode.mq_mode == RTE_ETH_MQ_RX_VMDQ_ONLY) {
2642 		/* Enable vlan filtering for VMDq */
2643 		ixgbe_vmdq_vlan_hw_filter_enable(dev);
2644 	}
2645 
2646 	/* Configure DCB hw */
2647 	ixgbe_configure_dcb(dev);
2648 
2649 	if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2650 		err = ixgbe_fdir_configure(dev);
2651 		if (err)
2652 			goto error;
2653 	}
2654 
2655 	/* Restore vf rate limit */
2656 	if (vfinfo != NULL) {
2657 		for (vf = 0; vf < pci_dev->max_vfs; vf++)
2658 			for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2659 				if (vfinfo[vf].tx_rate[idx] != 0)
2660 					ixgbe_set_vf_rate_limit(
2661 						dev, vf,
2662 						vfinfo[vf].tx_rate[idx],
2663 						1 << idx);
2664 	}
2665 
2666 	ixgbe_restore_statistics_mapping(dev);
2667 
2668 	err = ixgbe_flow_ctrl_enable(dev, hw);
2669 	if (err < 0) {
2670 		PMD_INIT_LOG(ERR, "enable flow ctrl err");
2671 		goto error;
2672 	}
2673 
2674 	err = ixgbe_dev_rxtx_start(dev);
2675 	if (err < 0) {
2676 		PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2677 		goto error;
2678 	}
2679 
2680 	/* Skip link setup if loopback mode is enabled. */
2681 	if (dev->data->dev_conf.lpbk_mode != 0) {
2682 		err = ixgbe_check_supported_loopback_mode(dev);
2683 		if (err < 0) {
2684 			PMD_INIT_LOG(ERR, "Unsupported loopback mode");
2685 			goto error;
2686 		} else {
2687 			goto skip_link_setup;
2688 		}
2689 	}
2690 
2691 	if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2692 		err = hw->mac.ops.setup_sfp(hw);
2693 		if (err)
2694 			goto error;
2695 	}
2696 
2697 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2698 		/* Turn on the copper */
2699 		ixgbe_set_phy_power(hw, true);
2700 	} else {
2701 		/* Turn on the laser */
2702 		ixgbe_enable_tx_laser(hw);
2703 	}
2704 
2705 	err = ixgbe_check_link(hw, &speed, &link_up, 0);
2706 	if (err)
2707 		goto error;
2708 	dev->data->dev_link.link_status = link_up;
2709 
2710 	err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2711 	if (err)
2712 		goto error;
2713 
2714 	switch (hw->mac.type) {
2715 	case ixgbe_mac_X550:
2716 	case ixgbe_mac_X550EM_x:
2717 	case ixgbe_mac_X550EM_a:
2718 		allowed_speeds = RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G |
2719 			RTE_ETH_LINK_SPEED_2_5G |  RTE_ETH_LINK_SPEED_5G |
2720 			RTE_ETH_LINK_SPEED_10G;
2721 		if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
2722 				hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
2723 			allowed_speeds = RTE_ETH_LINK_SPEED_10M |
2724 				RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G;
2725 		break;
2726 	default:
2727 		allowed_speeds = RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G |
2728 			RTE_ETH_LINK_SPEED_10G;
2729 	}
2730 
2731 	link_speeds = &dev->data->dev_conf.link_speeds;
2732 
2733 	/* Ignore autoneg flag bit and check the validity of 
2734 	 * link_speed 
2735 	 */
2736 	if (((*link_speeds) >> 1) & ~(allowed_speeds >> 1)) {
2737 		PMD_INIT_LOG(ERR, "Invalid link setting");
2738 		goto error;
2739 	}
2740 
2741 	speed = 0x0;
2742 	if (*link_speeds == RTE_ETH_LINK_SPEED_AUTONEG) {
2743 		switch (hw->mac.type) {
2744 		case ixgbe_mac_82598EB:
2745 			speed = IXGBE_LINK_SPEED_82598_AUTONEG;
2746 			break;
2747 		case ixgbe_mac_82599EB:
2748 		case ixgbe_mac_X540:
2749 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2750 			break;
2751 		case ixgbe_mac_X550:
2752 		case ixgbe_mac_X550EM_x:
2753 		case ixgbe_mac_X550EM_a:
2754 			speed = IXGBE_LINK_SPEED_X550_AUTONEG;
2755 			break;
2756 		default:
2757 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2758 		}
2759 	} else {
2760 		if (*link_speeds & RTE_ETH_LINK_SPEED_10G)
2761 			speed |= IXGBE_LINK_SPEED_10GB_FULL;
2762 		if (*link_speeds & RTE_ETH_LINK_SPEED_5G)
2763 			speed |= IXGBE_LINK_SPEED_5GB_FULL;
2764 		if (*link_speeds & RTE_ETH_LINK_SPEED_2_5G)
2765 			speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
2766 		if (*link_speeds & RTE_ETH_LINK_SPEED_1G)
2767 			speed |= IXGBE_LINK_SPEED_1GB_FULL;
2768 		if (*link_speeds & RTE_ETH_LINK_SPEED_100M)
2769 			speed |= IXGBE_LINK_SPEED_100_FULL;
2770 		if (*link_speeds & RTE_ETH_LINK_SPEED_10M)
2771 			speed |= IXGBE_LINK_SPEED_10_FULL;
2772 	}
2773 
2774 	err = ixgbe_setup_link(hw, speed, link_up);
2775 	if (err)
2776 		goto error;
2777 
2778 skip_link_setup:
2779 
2780 	if (rte_intr_allow_others(intr_handle)) {
2781 		/* check if lsc interrupt is enabled */
2782 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2783 			ixgbe_dev_lsc_interrupt_setup(dev, TRUE);
2784 		else
2785 			ixgbe_dev_lsc_interrupt_setup(dev, FALSE);
2786 		ixgbe_dev_macsec_interrupt_setup(dev);
2787 	} else {
2788 		rte_intr_callback_unregister(intr_handle,
2789 					     ixgbe_dev_interrupt_handler, dev);
2790 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2791 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
2792 				     " no intr multiplex");
2793 	}
2794 
2795 	/* check if rxq interrupt is enabled */
2796 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2797 	    rte_intr_dp_is_en(intr_handle))
2798 		ixgbe_dev_rxq_interrupt_setup(dev);
2799 
2800 	/* enable uio/vfio intr/eventfd mapping */
2801 	rte_intr_enable(intr_handle);
2802 
2803 	/* resume enabled intr since hw reset */
2804 	ixgbe_enable_intr(dev);
2805 	ixgbe_l2_tunnel_conf(dev);
2806 	ixgbe_filter_restore(dev);
2807 
2808 	if (tm_conf->root && !tm_conf->committed)
2809 		PMD_DRV_LOG(WARNING,
2810 			    "please call hierarchy_commit() "
2811 			    "before starting the port");
2812 
2813 	/* wait for the controller to acquire link */
2814 	err = ixgbe_wait_for_link_up(hw);
2815 	if (err)
2816 		goto error;
2817 
2818 	/*
2819 	 * Update link status right before return, because it may
2820 	 * start link configuration process in a separate thread.
2821 	 */
2822 	ixgbe_dev_link_update(dev, 0);
2823 
2824 	/* setup the macsec setting register */
2825 	if (macsec_setting->offload_en)
2826 		ixgbe_dev_macsec_register_enable(dev, macsec_setting);
2827 
2828 	return 0;
2829 
2830 error:
2831 	PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2832 	ixgbe_dev_clear_queues(dev);
2833 	return -EIO;
2834 }
2835 
2836 /*
2837  * Stop device: disable rx and tx functions to allow for reconfiguring.
2838  */
2839 static int
2840 ixgbe_dev_stop(struct rte_eth_dev *dev)
2841 {
2842 	struct rte_eth_link link;
2843 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2844 	struct ixgbe_hw *hw =
2845 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2846 	struct ixgbe_vf_info *vfinfo =
2847 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2848 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2849 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2850 	int vf;
2851 	struct ixgbe_tm_conf *tm_conf =
2852 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2853 
2854 	if (hw->adapter_stopped)
2855 		return 0;
2856 
2857 	PMD_INIT_FUNC_TRACE();
2858 
2859 	ixgbe_dev_wait_setup_link_complete(dev, 0);
2860 
2861 	/* disable interrupts */
2862 	ixgbe_disable_intr(hw);
2863 
2864 	/* reset the NIC */
2865 	ixgbe_pf_reset_hw(hw);
2866 	hw->adapter_stopped = 0;
2867 
2868 	/* stop adapter */
2869 	ixgbe_stop_adapter(hw);
2870 
2871 	for (vf = 0; vfinfo != NULL && vf < pci_dev->max_vfs; vf++)
2872 		vfinfo[vf].clear_to_send = false;
2873 
2874 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2875 		/* Turn off the copper */
2876 		ixgbe_set_phy_power(hw, false);
2877 	} else {
2878 		/* Turn off the laser */
2879 		ixgbe_disable_tx_laser(hw);
2880 	}
2881 
2882 	ixgbe_dev_clear_queues(dev);
2883 
2884 	/* Clear stored conf */
2885 	dev->data->scattered_rx = 0;
2886 	dev->data->lro = 0;
2887 
2888 	/* Clear recorded link status */
2889 	memset(&link, 0, sizeof(link));
2890 	rte_eth_linkstatus_set(dev, &link);
2891 
2892 	if (!rte_intr_allow_others(intr_handle))
2893 		/* resume to the default handler */
2894 		rte_intr_callback_register(intr_handle,
2895 					   ixgbe_dev_interrupt_handler,
2896 					   (void *)dev);
2897 
2898 	/* Clean datapath event and queue/vec mapping */
2899 	rte_intr_efd_disable(intr_handle);
2900 	rte_intr_vec_list_free(intr_handle);
2901 
2902 	/* reset hierarchy commit */
2903 	tm_conf->committed = false;
2904 
2905 	adapter->rss_reta_updated = 0;
2906 
2907 	hw->adapter_stopped = true;
2908 	dev->data->dev_started = 0;
2909 
2910 	return 0;
2911 }
2912 
2913 /*
2914  * Set device link up: enable tx.
2915  */
2916 static int
2917 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2918 {
2919 	struct ixgbe_hw *hw =
2920 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2921 	if (hw->mac.type == ixgbe_mac_82599EB) {
2922 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2923 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2924 			/* Not supported in bypass mode */
2925 			PMD_INIT_LOG(ERR, "Set link up is not supported "
2926 				     "by device id 0x%x", hw->device_id);
2927 			return -ENOTSUP;
2928 		}
2929 #endif
2930 	}
2931 
2932 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2933 		/* Turn on the copper */
2934 		ixgbe_set_phy_power(hw, true);
2935 	} else {
2936 		/* Turn on the laser */
2937 		ixgbe_enable_tx_laser(hw);
2938 		ixgbe_dev_link_update(dev, 0);
2939 	}
2940 
2941 	return 0;
2942 }
2943 
2944 /*
2945  * Set device link down: disable tx.
2946  */
2947 static int
2948 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2949 {
2950 	struct ixgbe_hw *hw =
2951 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2952 	if (hw->mac.type == ixgbe_mac_82599EB) {
2953 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2954 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2955 			/* Not supported in bypass mode */
2956 			PMD_INIT_LOG(ERR, "Set link down is not supported "
2957 				     "by device id 0x%x", hw->device_id);
2958 			return -ENOTSUP;
2959 		}
2960 #endif
2961 	}
2962 
2963 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2964 		/* Turn off the copper */
2965 		ixgbe_set_phy_power(hw, false);
2966 	} else {
2967 		/* Turn off the laser */
2968 		ixgbe_disable_tx_laser(hw);
2969 		ixgbe_dev_link_update(dev, 0);
2970 	}
2971 
2972 	return 0;
2973 }
2974 
2975 /*
2976  * Reset and stop device.
2977  */
2978 static int
2979 ixgbe_dev_close(struct rte_eth_dev *dev)
2980 {
2981 	struct ixgbe_hw *hw =
2982 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2983 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2984 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2985 	int retries = 0;
2986 	int ret;
2987 
2988 	PMD_INIT_FUNC_TRACE();
2989 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2990 		return 0;
2991 
2992 	ixgbe_pf_reset_hw(hw);
2993 
2994 	ret = ixgbe_dev_stop(dev);
2995 
2996 	ixgbe_dev_free_queues(dev);
2997 
2998 	ixgbe_disable_pcie_master(hw);
2999 
3000 	/* reprogram the RAR[0] in case user changed it. */
3001 	ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
3002 
3003 	/* Unlock any pending hardware semaphore */
3004 	ixgbe_swfw_lock_reset(hw);
3005 
3006 	/* disable uio intr before callback unregister */
3007 	rte_intr_disable(intr_handle);
3008 
3009 	do {
3010 		ret = rte_intr_callback_unregister(intr_handle,
3011 				ixgbe_dev_interrupt_handler, dev);
3012 		if (ret >= 0 || ret == -ENOENT) {
3013 			break;
3014 		} else if (ret != -EAGAIN) {
3015 			PMD_INIT_LOG(ERR,
3016 				"intr callback unregister failed: %d",
3017 				ret);
3018 		}
3019 		rte_delay_ms(100);
3020 	} while (retries++ < (10 + IXGBE_LINK_UP_TIME));
3021 
3022 	/* cancel the delay handler before remove dev */
3023 	rte_eal_alarm_cancel(ixgbe_dev_interrupt_delayed_handler, dev);
3024 
3025 	/* uninitialize PF if max_vfs not zero */
3026 	ixgbe_pf_host_uninit(dev);
3027 
3028 	/* remove all the fdir filters & hash */
3029 	ixgbe_fdir_filter_uninit(dev);
3030 
3031 	/* remove all the L2 tunnel filters & hash */
3032 	ixgbe_l2_tn_filter_uninit(dev);
3033 
3034 	/* Remove all ntuple filters of the device */
3035 	ixgbe_ntuple_filter_uninit(dev);
3036 
3037 	/* clear all the filters list */
3038 	ixgbe_filterlist_flush();
3039 
3040 	/* Remove all Traffic Manager configuration */
3041 	ixgbe_tm_conf_uninit(dev);
3042 
3043 #ifdef RTE_LIB_SECURITY
3044 	rte_free(dev->security_ctx);
3045 #endif
3046 
3047 	return ret;
3048 }
3049 
3050 /*
3051  * Reset PF device.
3052  */
3053 static int
3054 ixgbe_dev_reset(struct rte_eth_dev *dev)
3055 {
3056 	int ret;
3057 
3058 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
3059 	 * its VF to make them align with it. The detailed notification
3060 	 * mechanism is PMD specific. As to ixgbe PF, it is rather complex.
3061 	 * To avoid unexpected behavior in VF, currently reset of PF with
3062 	 * SR-IOV activation is not supported. It might be supported later.
3063 	 */
3064 	if (dev->data->sriov.active)
3065 		return -ENOTSUP;
3066 
3067 	ret = eth_ixgbe_dev_uninit(dev);
3068 	if (ret)
3069 		return ret;
3070 
3071 	ret = eth_ixgbe_dev_init(dev, NULL);
3072 
3073 	return ret;
3074 }
3075 
3076 static void
3077 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
3078 			   struct ixgbe_hw_stats *hw_stats,
3079 			   struct ixgbe_macsec_stats *macsec_stats,
3080 			   uint64_t *total_missed_rx, uint64_t *total_qbrc,
3081 			   uint64_t *total_qprc, uint64_t *total_qprdc)
3082 {
3083 	uint32_t bprc, lxon, lxoff, total;
3084 	uint32_t delta_gprc = 0;
3085 	unsigned i;
3086 	/* Workaround for RX byte count not including CRC bytes when CRC
3087 	 * strip is enabled. CRC bytes are removed from counters when crc_strip
3088 	 * is disabled.
3089 	 */
3090 	int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
3091 			IXGBE_HLREG0_RXCRCSTRP);
3092 
3093 	hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
3094 	hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
3095 	hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
3096 	hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
3097 
3098 	for (i = 0; i < 8; i++) {
3099 		uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
3100 
3101 		/* global total per queue */
3102 		hw_stats->mpc[i] += mp;
3103 		/* Running comprehensive total for stats display */
3104 		*total_missed_rx += hw_stats->mpc[i];
3105 		if (hw->mac.type == ixgbe_mac_82598EB) {
3106 			hw_stats->rnbc[i] +=
3107 			    IXGBE_READ_REG(hw, IXGBE_RNBC(i));
3108 			hw_stats->pxonrxc[i] +=
3109 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
3110 			hw_stats->pxoffrxc[i] +=
3111 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
3112 		} else {
3113 			hw_stats->pxonrxc[i] +=
3114 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
3115 			hw_stats->pxoffrxc[i] +=
3116 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
3117 			hw_stats->pxon2offc[i] +=
3118 				IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
3119 		}
3120 		hw_stats->pxontxc[i] +=
3121 		    IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
3122 		hw_stats->pxofftxc[i] +=
3123 		    IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
3124 	}
3125 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3126 		uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
3127 		uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
3128 		uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
3129 
3130 		delta_gprc += delta_qprc;
3131 
3132 		hw_stats->qprc[i] += delta_qprc;
3133 		hw_stats->qptc[i] += delta_qptc;
3134 
3135 		hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
3136 		hw_stats->qbrc[i] +=
3137 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
3138 		if (crc_strip == 0)
3139 			hw_stats->qbrc[i] -= delta_qprc * RTE_ETHER_CRC_LEN;
3140 
3141 		hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
3142 		hw_stats->qbtc[i] +=
3143 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
3144 
3145 		hw_stats->qprdc[i] += delta_qprdc;
3146 		*total_qprdc += hw_stats->qprdc[i];
3147 
3148 		*total_qprc += hw_stats->qprc[i];
3149 		*total_qbrc += hw_stats->qbrc[i];
3150 	}
3151 	hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
3152 	hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
3153 	hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
3154 
3155 	/*
3156 	 * An errata states that gprc actually counts good + missed packets:
3157 	 * Workaround to set gprc to summated queue packet receives
3158 	 */
3159 	hw_stats->gprc = *total_qprc;
3160 
3161 	if (hw->mac.type != ixgbe_mac_82598EB) {
3162 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
3163 		hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
3164 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
3165 		hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
3166 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
3167 		hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
3168 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
3169 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
3170 	} else {
3171 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
3172 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
3173 		/* 82598 only has a counter in the high register */
3174 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
3175 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
3176 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
3177 	}
3178 	uint64_t old_tpr = hw_stats->tpr;
3179 
3180 	hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
3181 	hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
3182 
3183 	if (crc_strip == 0)
3184 		hw_stats->gorc -= delta_gprc * RTE_ETHER_CRC_LEN;
3185 
3186 	uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
3187 	hw_stats->gptc += delta_gptc;
3188 	hw_stats->gotc -= delta_gptc * RTE_ETHER_CRC_LEN;
3189 	hw_stats->tor -= (hw_stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
3190 
3191 	/*
3192 	 * Workaround: mprc hardware is incorrectly counting
3193 	 * broadcasts, so for now we subtract those.
3194 	 */
3195 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
3196 	hw_stats->bprc += bprc;
3197 	hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
3198 	if (hw->mac.type == ixgbe_mac_82598EB)
3199 		hw_stats->mprc -= bprc;
3200 
3201 	hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
3202 	hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
3203 	hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
3204 	hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
3205 	hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
3206 	hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
3207 
3208 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
3209 	hw_stats->lxontxc += lxon;
3210 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
3211 	hw_stats->lxofftxc += lxoff;
3212 	total = lxon + lxoff;
3213 
3214 	hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
3215 	hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
3216 	hw_stats->gptc -= total;
3217 	hw_stats->mptc -= total;
3218 	hw_stats->ptc64 -= total;
3219 	hw_stats->gotc -= total * RTE_ETHER_MIN_LEN;
3220 
3221 	hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3222 	hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3223 	hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3224 	hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
3225 	hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
3226 	hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
3227 	hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
3228 	hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3229 	hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3230 	hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3231 	hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3232 	hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
3233 	hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3234 	hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
3235 	hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
3236 	hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
3237 	/* Only read FCOE on 82599 */
3238 	if (hw->mac.type != ixgbe_mac_82598EB) {
3239 		hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
3240 		hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
3241 		hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
3242 		hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
3243 		hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
3244 	}
3245 
3246 	/* Flow Director Stats registers */
3247 	if (hw->mac.type != ixgbe_mac_82598EB) {
3248 		hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
3249 		hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
3250 		hw_stats->fdirustat_add += IXGBE_READ_REG(hw,
3251 					IXGBE_FDIRUSTAT) & 0xFFFF;
3252 		hw_stats->fdirustat_remove += (IXGBE_READ_REG(hw,
3253 					IXGBE_FDIRUSTAT) >> 16) & 0xFFFF;
3254 		hw_stats->fdirfstat_fadd += IXGBE_READ_REG(hw,
3255 					IXGBE_FDIRFSTAT) & 0xFFFF;
3256 		hw_stats->fdirfstat_fremove += (IXGBE_READ_REG(hw,
3257 					IXGBE_FDIRFSTAT) >> 16) & 0xFFFF;
3258 	}
3259 	/* MACsec Stats registers */
3260 	macsec_stats->out_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECTXUT);
3261 	macsec_stats->out_pkts_encrypted +=
3262 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTE);
3263 	macsec_stats->out_pkts_protected +=
3264 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTP);
3265 	macsec_stats->out_octets_encrypted +=
3266 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTE);
3267 	macsec_stats->out_octets_protected +=
3268 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTP);
3269 	macsec_stats->in_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECRXUT);
3270 	macsec_stats->in_pkts_badtag += IXGBE_READ_REG(hw, IXGBE_LSECRXBAD);
3271 	macsec_stats->in_pkts_nosci += IXGBE_READ_REG(hw, IXGBE_LSECRXNOSCI);
3272 	macsec_stats->in_pkts_unknownsci +=
3273 		IXGBE_READ_REG(hw, IXGBE_LSECRXUNSCI);
3274 	macsec_stats->in_octets_decrypted +=
3275 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTD);
3276 	macsec_stats->in_octets_validated +=
3277 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTV);
3278 	macsec_stats->in_pkts_unchecked += IXGBE_READ_REG(hw, IXGBE_LSECRXUNCH);
3279 	macsec_stats->in_pkts_delayed += IXGBE_READ_REG(hw, IXGBE_LSECRXDELAY);
3280 	macsec_stats->in_pkts_late += IXGBE_READ_REG(hw, IXGBE_LSECRXLATE);
3281 	for (i = 0; i < 2; i++) {
3282 		macsec_stats->in_pkts_ok +=
3283 			IXGBE_READ_REG(hw, IXGBE_LSECRXOK(i));
3284 		macsec_stats->in_pkts_invalid +=
3285 			IXGBE_READ_REG(hw, IXGBE_LSECRXINV(i));
3286 		macsec_stats->in_pkts_notvalid +=
3287 			IXGBE_READ_REG(hw, IXGBE_LSECRXNV(i));
3288 	}
3289 	macsec_stats->in_pkts_unusedsa += IXGBE_READ_REG(hw, IXGBE_LSECRXUNSA);
3290 	macsec_stats->in_pkts_notusingsa +=
3291 		IXGBE_READ_REG(hw, IXGBE_LSECRXNUSA);
3292 }
3293 
3294 /*
3295  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
3296  */
3297 static int
3298 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3299 {
3300 	struct ixgbe_hw *hw =
3301 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3302 	struct ixgbe_hw_stats *hw_stats =
3303 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3304 	struct ixgbe_macsec_stats *macsec_stats =
3305 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3306 				dev->data->dev_private);
3307 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3308 	unsigned i;
3309 
3310 	total_missed_rx = 0;
3311 	total_qbrc = 0;
3312 	total_qprc = 0;
3313 	total_qprdc = 0;
3314 
3315 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3316 			&total_qbrc, &total_qprc, &total_qprdc);
3317 
3318 	if (stats == NULL)
3319 		return -EINVAL;
3320 
3321 	/* Fill out the rte_eth_stats statistics structure */
3322 	stats->ipackets = total_qprc;
3323 	stats->ibytes = total_qbrc;
3324 	stats->opackets = hw_stats->gptc;
3325 	stats->obytes = hw_stats->gotc;
3326 
3327 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3328 		stats->q_ipackets[i] = hw_stats->qprc[i];
3329 		stats->q_opackets[i] = hw_stats->qptc[i];
3330 		stats->q_ibytes[i] = hw_stats->qbrc[i];
3331 		stats->q_obytes[i] = hw_stats->qbtc[i];
3332 		stats->q_errors[i] = hw_stats->qprdc[i];
3333 	}
3334 
3335 	/* Rx Errors */
3336 	stats->imissed  = total_missed_rx;
3337 	stats->ierrors  = hw_stats->crcerrs +
3338 			  hw_stats->mspdc +
3339 			  hw_stats->rlec +
3340 			  hw_stats->ruc +
3341 			  hw_stats->roc +
3342 			  hw_stats->illerrc +
3343 			  hw_stats->errbc +
3344 			  hw_stats->rfc +
3345 			  hw_stats->fccrc +
3346 			  hw_stats->fclast;
3347 
3348 	/*
3349 	 * 82599 errata, UDP frames with a 0 checksum can be marked as checksum
3350 	 * errors.
3351 	 */
3352 	if (hw->mac.type != ixgbe_mac_82599EB)
3353 		stats->ierrors += hw_stats->xec;
3354 
3355 	/* Tx Errors */
3356 	stats->oerrors  = 0;
3357 	return 0;
3358 }
3359 
3360 static int
3361 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
3362 {
3363 	struct ixgbe_hw_stats *stats =
3364 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3365 
3366 	/* HW registers are cleared on read */
3367 	ixgbe_dev_stats_get(dev, NULL);
3368 
3369 	/* Reset software totals */
3370 	memset(stats, 0, sizeof(*stats));
3371 
3372 	return 0;
3373 }
3374 
3375 /* This function calculates the number of xstats based on the current config */
3376 static unsigned
3377 ixgbe_xstats_calc_num(void) {
3378 	return IXGBE_NB_HW_STATS + IXGBE_NB_MACSEC_STATS +
3379 		(IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
3380 		(IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
3381 }
3382 
3383 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3384 	struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned int size)
3385 {
3386 	const unsigned cnt_stats = ixgbe_xstats_calc_num();
3387 	unsigned stat, i, count;
3388 
3389 	if (xstats_names != NULL) {
3390 		count = 0;
3391 
3392 		/* Note: limit >= cnt_stats checked upstream
3393 		 * in rte_eth_xstats_names()
3394 		 */
3395 
3396 		/* Extended stats from ixgbe_hw_stats */
3397 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3398 			strlcpy(xstats_names[count].name,
3399 				rte_ixgbe_stats_strings[i].name,
3400 				sizeof(xstats_names[count].name));
3401 			count++;
3402 		}
3403 
3404 		/* MACsec Stats */
3405 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3406 			strlcpy(xstats_names[count].name,
3407 				rte_ixgbe_macsec_strings[i].name,
3408 				sizeof(xstats_names[count].name));
3409 			count++;
3410 		}
3411 
3412 		/* RX Priority Stats */
3413 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3414 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3415 				snprintf(xstats_names[count].name,
3416 					sizeof(xstats_names[count].name),
3417 					"rx_priority%u_%s", i,
3418 					rte_ixgbe_rxq_strings[stat].name);
3419 				count++;
3420 			}
3421 		}
3422 
3423 		/* TX Priority Stats */
3424 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3425 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3426 				snprintf(xstats_names[count].name,
3427 					sizeof(xstats_names[count].name),
3428 					"tx_priority%u_%s", i,
3429 					rte_ixgbe_txq_strings[stat].name);
3430 				count++;
3431 			}
3432 		}
3433 	}
3434 	return cnt_stats;
3435 }
3436 
3437 static int ixgbe_dev_xstats_get_names_by_id(
3438 	struct rte_eth_dev *dev,
3439 	const uint64_t *ids,
3440 	struct rte_eth_xstat_name *xstats_names,
3441 	unsigned int limit)
3442 {
3443 	if (!ids) {
3444 		const unsigned int cnt_stats = ixgbe_xstats_calc_num();
3445 		unsigned int stat, i, count;
3446 
3447 		if (xstats_names != NULL) {
3448 			count = 0;
3449 
3450 			/* Note: limit >= cnt_stats checked upstream
3451 			 * in rte_eth_xstats_names()
3452 			 */
3453 
3454 			/* Extended stats from ixgbe_hw_stats */
3455 			for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3456 				strlcpy(xstats_names[count].name,
3457 					rte_ixgbe_stats_strings[i].name,
3458 					sizeof(xstats_names[count].name));
3459 				count++;
3460 			}
3461 
3462 			/* MACsec Stats */
3463 			for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3464 				strlcpy(xstats_names[count].name,
3465 					rte_ixgbe_macsec_strings[i].name,
3466 					sizeof(xstats_names[count].name));
3467 				count++;
3468 			}
3469 
3470 			/* RX Priority Stats */
3471 			for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3472 				for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3473 					snprintf(xstats_names[count].name,
3474 					    sizeof(xstats_names[count].name),
3475 					    "rx_priority%u_%s", i,
3476 					    rte_ixgbe_rxq_strings[stat].name);
3477 					count++;
3478 				}
3479 			}
3480 
3481 			/* TX Priority Stats */
3482 			for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3483 				for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3484 					snprintf(xstats_names[count].name,
3485 					    sizeof(xstats_names[count].name),
3486 					    "tx_priority%u_%s", i,
3487 					    rte_ixgbe_txq_strings[stat].name);
3488 					count++;
3489 				}
3490 			}
3491 		}
3492 		return cnt_stats;
3493 	}
3494 
3495 	uint16_t i;
3496 	uint16_t size = ixgbe_xstats_calc_num();
3497 	struct rte_eth_xstat_name xstats_names_copy[size];
3498 
3499 	ixgbe_dev_xstats_get_names_by_id(dev, NULL, xstats_names_copy,
3500 			size);
3501 
3502 	for (i = 0; i < limit; i++) {
3503 		if (ids[i] >= size) {
3504 			PMD_INIT_LOG(ERR, "id value isn't valid");
3505 			return -1;
3506 		}
3507 		strcpy(xstats_names[i].name,
3508 				xstats_names_copy[ids[i]].name);
3509 	}
3510 	return limit;
3511 }
3512 
3513 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3514 	struct rte_eth_xstat_name *xstats_names, unsigned limit)
3515 {
3516 	unsigned i;
3517 
3518 	if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
3519 		return -ENOMEM;
3520 
3521 	if (xstats_names != NULL)
3522 		for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
3523 			strlcpy(xstats_names[i].name,
3524 				rte_ixgbevf_stats_strings[i].name,
3525 				sizeof(xstats_names[i].name));
3526 	return IXGBEVF_NB_XSTATS;
3527 }
3528 
3529 static int
3530 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3531 					 unsigned n)
3532 {
3533 	struct ixgbe_hw *hw =
3534 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3535 	struct ixgbe_hw_stats *hw_stats =
3536 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3537 	struct ixgbe_macsec_stats *macsec_stats =
3538 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3539 				dev->data->dev_private);
3540 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3541 	unsigned i, stat, count = 0;
3542 
3543 	count = ixgbe_xstats_calc_num();
3544 
3545 	if (n < count)
3546 		return count;
3547 
3548 	total_missed_rx = 0;
3549 	total_qbrc = 0;
3550 	total_qprc = 0;
3551 	total_qprdc = 0;
3552 
3553 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3554 			&total_qbrc, &total_qprc, &total_qprdc);
3555 
3556 	/* If this is a reset xstats is NULL, and we have cleared the
3557 	 * registers by reading them.
3558 	 */
3559 	if (!xstats)
3560 		return 0;
3561 
3562 	/* Extended stats from ixgbe_hw_stats */
3563 	count = 0;
3564 	for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3565 		xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3566 				rte_ixgbe_stats_strings[i].offset);
3567 		xstats[count].id = count;
3568 		count++;
3569 	}
3570 
3571 	/* MACsec Stats */
3572 	for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3573 		xstats[count].value = *(uint64_t *)(((char *)macsec_stats) +
3574 				rte_ixgbe_macsec_strings[i].offset);
3575 		xstats[count].id = count;
3576 		count++;
3577 	}
3578 
3579 	/* RX Priority Stats */
3580 	for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3581 		for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3582 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3583 					rte_ixgbe_rxq_strings[stat].offset +
3584 					(sizeof(uint64_t) * i));
3585 			xstats[count].id = count;
3586 			count++;
3587 		}
3588 	}
3589 
3590 	/* TX Priority Stats */
3591 	for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3592 		for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3593 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3594 					rte_ixgbe_txq_strings[stat].offset +
3595 					(sizeof(uint64_t) * i));
3596 			xstats[count].id = count;
3597 			count++;
3598 		}
3599 	}
3600 	return count;
3601 }
3602 
3603 static int
3604 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
3605 		uint64_t *values, unsigned int n)
3606 {
3607 	if (!ids) {
3608 		struct ixgbe_hw *hw =
3609 				IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3610 		struct ixgbe_hw_stats *hw_stats =
3611 				IXGBE_DEV_PRIVATE_TO_STATS(
3612 						dev->data->dev_private);
3613 		struct ixgbe_macsec_stats *macsec_stats =
3614 				IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3615 					dev->data->dev_private);
3616 		uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3617 		unsigned int i, stat, count = 0;
3618 
3619 		count = ixgbe_xstats_calc_num();
3620 
3621 		if (!ids && n < count)
3622 			return count;
3623 
3624 		total_missed_rx = 0;
3625 		total_qbrc = 0;
3626 		total_qprc = 0;
3627 		total_qprdc = 0;
3628 
3629 		ixgbe_read_stats_registers(hw, hw_stats, macsec_stats,
3630 				&total_missed_rx, &total_qbrc, &total_qprc,
3631 				&total_qprdc);
3632 
3633 		/* If this is a reset xstats is NULL, and we have cleared the
3634 		 * registers by reading them.
3635 		 */
3636 		if (!ids && !values)
3637 			return 0;
3638 
3639 		/* Extended stats from ixgbe_hw_stats */
3640 		count = 0;
3641 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3642 			values[count] = *(uint64_t *)(((char *)hw_stats) +
3643 					rte_ixgbe_stats_strings[i].offset);
3644 			count++;
3645 		}
3646 
3647 		/* MACsec Stats */
3648 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3649 			values[count] = *(uint64_t *)(((char *)macsec_stats) +
3650 					rte_ixgbe_macsec_strings[i].offset);
3651 			count++;
3652 		}
3653 
3654 		/* RX Priority Stats */
3655 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3656 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3657 				values[count] =
3658 					*(uint64_t *)(((char *)hw_stats) +
3659 					rte_ixgbe_rxq_strings[stat].offset +
3660 					(sizeof(uint64_t) * i));
3661 				count++;
3662 			}
3663 		}
3664 
3665 		/* TX Priority Stats */
3666 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3667 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3668 				values[count] =
3669 					*(uint64_t *)(((char *)hw_stats) +
3670 					rte_ixgbe_txq_strings[stat].offset +
3671 					(sizeof(uint64_t) * i));
3672 				count++;
3673 			}
3674 		}
3675 		return count;
3676 	}
3677 
3678 	uint16_t i;
3679 	uint16_t size = ixgbe_xstats_calc_num();
3680 	uint64_t values_copy[size];
3681 
3682 	ixgbe_dev_xstats_get_by_id(dev, NULL, values_copy, size);
3683 
3684 	for (i = 0; i < n; i++) {
3685 		if (ids[i] >= size) {
3686 			PMD_INIT_LOG(ERR, "id value isn't valid");
3687 			return -1;
3688 		}
3689 		values[i] = values_copy[ids[i]];
3690 	}
3691 	return n;
3692 }
3693 
3694 static int
3695 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
3696 {
3697 	struct ixgbe_hw_stats *stats =
3698 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3699 	struct ixgbe_macsec_stats *macsec_stats =
3700 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3701 				dev->data->dev_private);
3702 
3703 	unsigned count = ixgbe_xstats_calc_num();
3704 
3705 	/* HW registers are cleared on read */
3706 	ixgbe_dev_xstats_get(dev, NULL, count);
3707 
3708 	/* Reset software totals */
3709 	memset(stats, 0, sizeof(*stats));
3710 	memset(macsec_stats, 0, sizeof(*macsec_stats));
3711 
3712 	return 0;
3713 }
3714 
3715 static void
3716 ixgbevf_update_stats(struct rte_eth_dev *dev)
3717 {
3718 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3719 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3720 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3721 
3722 	/* Good Rx packet, include VF loopback */
3723 	UPDATE_VF_STAT(IXGBE_VFGPRC,
3724 	    hw_stats->last_vfgprc, hw_stats->vfgprc);
3725 
3726 	/* Good Rx octets, include VF loopback */
3727 	UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
3728 	    hw_stats->last_vfgorc, hw_stats->vfgorc);
3729 
3730 	/* Good Tx packet, include VF loopback */
3731 	UPDATE_VF_STAT(IXGBE_VFGPTC,
3732 	    hw_stats->last_vfgptc, hw_stats->vfgptc);
3733 
3734 	/* Good Tx octets, include VF loopback */
3735 	UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
3736 	    hw_stats->last_vfgotc, hw_stats->vfgotc);
3737 
3738 	/* Rx Multicst Packet */
3739 	UPDATE_VF_STAT(IXGBE_VFMPRC,
3740 	    hw_stats->last_vfmprc, hw_stats->vfmprc);
3741 }
3742 
3743 static int
3744 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3745 		       unsigned n)
3746 {
3747 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3748 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3749 	unsigned i;
3750 
3751 	if (n < IXGBEVF_NB_XSTATS)
3752 		return IXGBEVF_NB_XSTATS;
3753 
3754 	ixgbevf_update_stats(dev);
3755 
3756 	if (!xstats)
3757 		return 0;
3758 
3759 	/* Extended stats */
3760 	for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
3761 		xstats[i].id = i;
3762 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
3763 			rte_ixgbevf_stats_strings[i].offset);
3764 	}
3765 
3766 	return IXGBEVF_NB_XSTATS;
3767 }
3768 
3769 static int
3770 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3771 {
3772 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3773 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3774 
3775 	ixgbevf_update_stats(dev);
3776 
3777 	if (stats == NULL)
3778 		return -EINVAL;
3779 
3780 	stats->ipackets = hw_stats->vfgprc;
3781 	stats->ibytes = hw_stats->vfgorc;
3782 	stats->opackets = hw_stats->vfgptc;
3783 	stats->obytes = hw_stats->vfgotc;
3784 	return 0;
3785 }
3786 
3787 static int
3788 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3789 {
3790 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3791 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3792 
3793 	/* Sync HW register to the last stats */
3794 	ixgbevf_dev_stats_get(dev, NULL);
3795 
3796 	/* reset HW current stats*/
3797 	hw_stats->vfgprc = 0;
3798 	hw_stats->vfgorc = 0;
3799 	hw_stats->vfgptc = 0;
3800 	hw_stats->vfgotc = 0;
3801 	hw_stats->vfmprc = 0;
3802 
3803 	return 0;
3804 }
3805 
3806 static int
3807 ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
3808 {
3809 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3810 	u16 eeprom_verh, eeprom_verl;
3811 	u32 etrack_id;
3812 	int ret;
3813 
3814 	ixgbe_read_eeprom(hw, 0x2e, &eeprom_verh);
3815 	ixgbe_read_eeprom(hw, 0x2d, &eeprom_verl);
3816 
3817 	etrack_id = (eeprom_verh << 16) | eeprom_verl;
3818 	ret = snprintf(fw_version, fw_size, "0x%08x", etrack_id);
3819 	if (ret < 0)
3820 		return -EINVAL;
3821 
3822 	ret += 1; /* add the size of '\0' */
3823 	if (fw_size < (size_t)ret)
3824 		return ret;
3825 	else
3826 		return 0;
3827 }
3828 
3829 static int
3830 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3831 {
3832 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3833 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3834 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3835 
3836 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3837 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3838 	if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3839 		/*
3840 		 * When DCB/VT is off, maximum number of queues changes,
3841 		 * except for 82598EB, which remains constant.
3842 		 */
3843 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_NONE &&
3844 				hw->mac.type != ixgbe_mac_82598EB)
3845 			dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3846 	}
3847 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3848 	dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3849 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3850 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3851 	dev_info->max_vfs = pci_dev->max_vfs;
3852 	if (hw->mac.type == ixgbe_mac_82598EB)
3853 		dev_info->max_vmdq_pools = RTE_ETH_16_POOLS;
3854 	else
3855 		dev_info->max_vmdq_pools = RTE_ETH_64_POOLS;
3856 	dev_info->max_mtu =  dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3857 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
3858 	dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3859 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3860 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3861 				     dev_info->rx_queue_offload_capa);
3862 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3863 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3864 
3865 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3866 		.rx_thresh = {
3867 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3868 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3869 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3870 		},
3871 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3872 		.rx_drop_en = 0,
3873 		.offloads = 0,
3874 	};
3875 
3876 	dev_info->default_txconf = (struct rte_eth_txconf) {
3877 		.tx_thresh = {
3878 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3879 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3880 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3881 		},
3882 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3883 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3884 		.offloads = 0,
3885 	};
3886 
3887 	dev_info->rx_desc_lim = rx_desc_lim;
3888 	dev_info->tx_desc_lim = tx_desc_lim;
3889 
3890 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3891 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3892 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3893 
3894 	dev_info->speed_capa = RTE_ETH_LINK_SPEED_1G | RTE_ETH_LINK_SPEED_10G;
3895 	if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
3896 			hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
3897 		dev_info->speed_capa = RTE_ETH_LINK_SPEED_10M |
3898 			RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G;
3899 
3900 	if (hw->mac.type == ixgbe_mac_X540 ||
3901 	    hw->mac.type == ixgbe_mac_X540_vf ||
3902 	    hw->mac.type == ixgbe_mac_X550 ||
3903 	    hw->mac.type == ixgbe_mac_X550_vf) {
3904 		dev_info->speed_capa |= RTE_ETH_LINK_SPEED_100M;
3905 	}
3906 	if (hw->mac.type == ixgbe_mac_X550) {
3907 		dev_info->speed_capa |= RTE_ETH_LINK_SPEED_2_5G;
3908 		dev_info->speed_capa |= RTE_ETH_LINK_SPEED_5G;
3909 	}
3910 
3911 	/* Driver-preferred Rx/Tx parameters */
3912 	dev_info->default_rxportconf.burst_size = 32;
3913 	dev_info->default_txportconf.burst_size = 32;
3914 	dev_info->default_rxportconf.nb_queues = 1;
3915 	dev_info->default_txportconf.nb_queues = 1;
3916 	dev_info->default_rxportconf.ring_size = 256;
3917 	dev_info->default_txportconf.ring_size = 256;
3918 
3919 	return 0;
3920 }
3921 
3922 static const uint32_t *
3923 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3924 {
3925 	static const uint32_t ptypes[] = {
3926 		/* For non-vec functions,
3927 		 * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3928 		 * for vec functions,
3929 		 * refers to _recv_raw_pkts_vec().
3930 		 */
3931 		RTE_PTYPE_L2_ETHER,
3932 		RTE_PTYPE_L3_IPV4,
3933 		RTE_PTYPE_L3_IPV4_EXT,
3934 		RTE_PTYPE_L3_IPV6,
3935 		RTE_PTYPE_L3_IPV6_EXT,
3936 		RTE_PTYPE_L4_SCTP,
3937 		RTE_PTYPE_L4_TCP,
3938 		RTE_PTYPE_L4_UDP,
3939 		RTE_PTYPE_TUNNEL_IP,
3940 		RTE_PTYPE_INNER_L3_IPV6,
3941 		RTE_PTYPE_INNER_L3_IPV6_EXT,
3942 		RTE_PTYPE_INNER_L4_TCP,
3943 		RTE_PTYPE_INNER_L4_UDP,
3944 		RTE_PTYPE_UNKNOWN
3945 	};
3946 
3947 	if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3948 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3949 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3950 	    dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3951 		return ptypes;
3952 
3953 #if defined(RTE_ARCH_X86) || defined(__ARM_NEON)
3954 	if (dev->rx_pkt_burst == ixgbe_recv_pkts_vec ||
3955 	    dev->rx_pkt_burst == ixgbe_recv_scattered_pkts_vec)
3956 		return ptypes;
3957 #endif
3958 	return NULL;
3959 }
3960 
3961 static int
3962 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3963 		     struct rte_eth_dev_info *dev_info)
3964 {
3965 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3966 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3967 
3968 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3969 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3970 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3971 	dev_info->max_rx_pktlen = 9728; /* includes CRC, cf MAXFRS reg */
3972 	dev_info->max_mtu = dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3973 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3974 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3975 	dev_info->max_vfs = pci_dev->max_vfs;
3976 	if (hw->mac.type == ixgbe_mac_82598EB)
3977 		dev_info->max_vmdq_pools = RTE_ETH_16_POOLS;
3978 	else
3979 		dev_info->max_vmdq_pools = RTE_ETH_64_POOLS;
3980 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3981 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3982 				     dev_info->rx_queue_offload_capa);
3983 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3984 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3985 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3986 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3987 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3988 
3989 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3990 		.rx_thresh = {
3991 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3992 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3993 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3994 		},
3995 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3996 		.rx_drop_en = 0,
3997 		.offloads = 0,
3998 	};
3999 
4000 	dev_info->default_txconf = (struct rte_eth_txconf) {
4001 		.tx_thresh = {
4002 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
4003 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
4004 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
4005 		},
4006 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
4007 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
4008 		.offloads = 0,
4009 	};
4010 
4011 	dev_info->rx_desc_lim = rx_desc_lim;
4012 	dev_info->tx_desc_lim = tx_desc_lim;
4013 
4014 	return 0;
4015 }
4016 
4017 static int
4018 ixgbevf_check_link(struct ixgbe_hw *hw, ixgbe_link_speed *speed,
4019 		   bool *link_up, int wait_to_complete)
4020 {
4021 	struct ixgbe_adapter *adapter = container_of(hw,
4022 						     struct ixgbe_adapter, hw);
4023 	struct ixgbe_mbx_info *mbx = &hw->mbx;
4024 	struct ixgbe_mac_info *mac = &hw->mac;
4025 	uint32_t links_reg, in_msg;
4026 	int ret_val = 0;
4027 
4028 	/* If we were hit with a reset drop the link */
4029 	if (!mbx->ops.check_for_rst(hw, 0) || !mbx->timeout)
4030 		mac->get_link_status = true;
4031 
4032 	if (!mac->get_link_status)
4033 		goto out;
4034 
4035 	/* if link status is down no point in checking to see if pf is up */
4036 	links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4037 	if (!(links_reg & IXGBE_LINKS_UP))
4038 		goto out;
4039 
4040 	/* for SFP+ modules and DA cables on 82599 it can take up to 500usecs
4041 	 * before the link status is correct
4042 	 */
4043 	if (mac->type == ixgbe_mac_82599_vf && wait_to_complete) {
4044 		int i;
4045 
4046 		for (i = 0; i < 5; i++) {
4047 			rte_delay_us(100);
4048 			links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4049 
4050 			if (!(links_reg & IXGBE_LINKS_UP))
4051 				goto out;
4052 		}
4053 	}
4054 
4055 	switch (links_reg & IXGBE_LINKS_SPEED_82599) {
4056 	case IXGBE_LINKS_SPEED_10G_82599:
4057 		*speed = IXGBE_LINK_SPEED_10GB_FULL;
4058 		if (hw->mac.type >= ixgbe_mac_X550) {
4059 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4060 				*speed = IXGBE_LINK_SPEED_2_5GB_FULL;
4061 		}
4062 		break;
4063 	case IXGBE_LINKS_SPEED_1G_82599:
4064 		*speed = IXGBE_LINK_SPEED_1GB_FULL;
4065 		break;
4066 	case IXGBE_LINKS_SPEED_100_82599:
4067 		*speed = IXGBE_LINK_SPEED_100_FULL;
4068 		if (hw->mac.type == ixgbe_mac_X550) {
4069 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4070 				*speed = IXGBE_LINK_SPEED_5GB_FULL;
4071 		}
4072 		break;
4073 	case IXGBE_LINKS_SPEED_10_X550EM_A:
4074 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
4075 		/* Since Reserved in older MAC's */
4076 		if (hw->mac.type >= ixgbe_mac_X550)
4077 			*speed = IXGBE_LINK_SPEED_10_FULL;
4078 		break;
4079 	default:
4080 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
4081 	}
4082 
4083 	if (wait_to_complete == 0 && adapter->pflink_fullchk == 0) {
4084 		if (*speed == IXGBE_LINK_SPEED_UNKNOWN)
4085 			mac->get_link_status = true;
4086 		else
4087 			mac->get_link_status = false;
4088 
4089 		goto out;
4090 	}
4091 
4092 	/* if the read failed it could just be a mailbox collision, best wait
4093 	 * until we are called again and don't report an error
4094 	 */
4095 	if (mbx->ops.read(hw, &in_msg, 1, 0))
4096 		goto out;
4097 
4098 	if (!(in_msg & IXGBE_VT_MSGTYPE_CTS)) {
4099 		/* msg is not CTS and is NACK we must have lost CTS status */
4100 		if (in_msg & IXGBE_VT_MSGTYPE_NACK)
4101 			mac->get_link_status = false;
4102 		goto out;
4103 	}
4104 
4105 	/* the pf is talking, if we timed out in the past we reinit */
4106 	if (!mbx->timeout) {
4107 		ret_val = -1;
4108 		goto out;
4109 	}
4110 
4111 	/* if we passed all the tests above then the link is up and we no
4112 	 * longer need to check for link
4113 	 */
4114 	mac->get_link_status = false;
4115 
4116 out:
4117 	*link_up = !mac->get_link_status;
4118 	return ret_val;
4119 }
4120 
4121 /*
4122  * If @timeout_ms was 0, it means that it will not return until link complete.
4123  * It returns 1 on complete, return 0 on timeout.
4124  */
4125 static int
4126 ixgbe_dev_wait_setup_link_complete(struct rte_eth_dev *dev, uint32_t timeout_ms)
4127 {
4128 #define WARNING_TIMEOUT    9000 /* 9s  in total */
4129 	struct ixgbe_adapter *ad = dev->data->dev_private;
4130 	uint32_t timeout = timeout_ms ? timeout_ms : WARNING_TIMEOUT;
4131 
4132 	while (rte_atomic32_read(&ad->link_thread_running)) {
4133 		msec_delay(1);
4134 		timeout--;
4135 
4136 		if (timeout_ms) {
4137 			if (!timeout)
4138 				return 0;
4139 		} else if (!timeout) {
4140 			/* It will not return until link complete */
4141 			timeout = WARNING_TIMEOUT;
4142 			PMD_DRV_LOG(ERR, "IXGBE link thread not complete too long time!");
4143 		}
4144 	}
4145 
4146 	return 1;
4147 }
4148 
4149 static void *
4150 ixgbe_dev_setup_link_thread_handler(void *param)
4151 {
4152 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4153 	struct ixgbe_adapter *ad = dev->data->dev_private;
4154 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4155 	struct ixgbe_interrupt *intr =
4156 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4157 	u32 speed;
4158 	bool autoneg = false;
4159 
4160 	pthread_detach(pthread_self());
4161 	speed = hw->phy.autoneg_advertised;
4162 	if (!speed)
4163 		ixgbe_get_link_capabilities(hw, &speed, &autoneg);
4164 
4165 	ixgbe_setup_link(hw, speed, true);
4166 
4167 	intr->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4168 	rte_atomic32_clear(&ad->link_thread_running);
4169 	return NULL;
4170 }
4171 
4172 /*
4173  * In freebsd environment, nic_uio drivers do not support interrupts,
4174  * rte_intr_callback_register() will fail to register interrupts.
4175  * We can not make link status to change from down to up by interrupt
4176  * callback. So we need to wait for the controller to acquire link
4177  * when ports start.
4178  * It returns 0 on link up.
4179  */
4180 static int
4181 ixgbe_wait_for_link_up(struct ixgbe_hw *hw)
4182 {
4183 #ifdef RTE_EXEC_ENV_FREEBSD
4184 	int err, i;
4185 	bool link_up = false;
4186 	uint32_t speed = 0;
4187 	const int nb_iter = 25;
4188 
4189 	for (i = 0; i < nb_iter; i++) {
4190 		err = ixgbe_check_link(hw, &speed, &link_up, 0);
4191 		if (err)
4192 			return err;
4193 		if (link_up)
4194 			return 0;
4195 		msec_delay(200);
4196 	}
4197 
4198 	return 0;
4199 #else
4200 	RTE_SET_USED(hw);
4201 	return 0;
4202 #endif
4203 }
4204 
4205 /* return 0 means link status changed, -1 means not changed */
4206 int
4207 ixgbe_dev_link_update_share(struct rte_eth_dev *dev,
4208 			    int wait_to_complete, int vf)
4209 {
4210 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4211 	struct ixgbe_adapter *ad = dev->data->dev_private;
4212 	struct rte_eth_link link;
4213 	ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
4214 	struct ixgbe_interrupt *intr =
4215 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4216 	bool link_up;
4217 	int diag;
4218 	int wait = 1;
4219 	u32 esdp_reg;
4220 
4221 	memset(&link, 0, sizeof(link));
4222 	link.link_status = RTE_ETH_LINK_DOWN;
4223 	link.link_speed = RTE_ETH_SPEED_NUM_NONE;
4224 	link.link_duplex = RTE_ETH_LINK_HALF_DUPLEX;
4225 	link.link_autoneg = !(dev->data->dev_conf.link_speeds &
4226 			RTE_ETH_LINK_SPEED_FIXED);
4227 
4228 	hw->mac.get_link_status = true;
4229 
4230 	if (intr->flags & IXGBE_FLAG_NEED_LINK_CONFIG)
4231 		return rte_eth_linkstatus_set(dev, &link);
4232 
4233 	/* check if it needs to wait to complete, if lsc interrupt is enabled */
4234 	if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
4235 		wait = 0;
4236 
4237 /* BSD has no interrupt mechanism, so force NIC status synchronization. */
4238 #ifdef RTE_EXEC_ENV_FREEBSD
4239 	wait = 1;
4240 #endif
4241 
4242 	if (vf)
4243 		diag = ixgbevf_check_link(hw, &link_speed, &link_up, wait);
4244 	else
4245 		diag = ixgbe_check_link(hw, &link_speed, &link_up, wait);
4246 
4247 	if (diag != 0) {
4248 		link.link_speed = RTE_ETH_SPEED_NUM_100M;
4249 		link.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
4250 		return rte_eth_linkstatus_set(dev, &link);
4251 	}
4252 
4253 	if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4254 		esdp_reg = IXGBE_READ_REG(hw, IXGBE_ESDP);
4255 		if ((esdp_reg & IXGBE_ESDP_SDP3))
4256 			link_up = 0;
4257 	}
4258 
4259 	if (link_up == 0) {
4260 		if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4261 			ixgbe_dev_wait_setup_link_complete(dev, 0);
4262 			if (rte_atomic32_test_and_set(&ad->link_thread_running)) {
4263 				/* To avoid race condition between threads, set
4264 				 * the IXGBE_FLAG_NEED_LINK_CONFIG flag only
4265 				 * when there is no link thread running.
4266 				 */
4267 				intr->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
4268 				if (rte_ctrl_thread_create(&ad->link_thread_tid,
4269 					"ixgbe-link-handler",
4270 					NULL,
4271 					ixgbe_dev_setup_link_thread_handler,
4272 					dev) < 0) {
4273 					PMD_DRV_LOG(ERR,
4274 						"Create link thread failed!");
4275 					rte_atomic32_clear(&ad->link_thread_running);
4276 				}
4277 			} else {
4278 				PMD_DRV_LOG(ERR,
4279 					"Other link thread is running now!");
4280 			}
4281 		}
4282 		return rte_eth_linkstatus_set(dev, &link);
4283 	}
4284 
4285 	link.link_status = RTE_ETH_LINK_UP;
4286 	link.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
4287 
4288 	switch (link_speed) {
4289 	default:
4290 	case IXGBE_LINK_SPEED_UNKNOWN:
4291 		link.link_speed = RTE_ETH_SPEED_NUM_UNKNOWN;
4292 		break;
4293 
4294 	case IXGBE_LINK_SPEED_10_FULL:
4295 		link.link_speed = RTE_ETH_SPEED_NUM_10M;
4296 		break;
4297 
4298 	case IXGBE_LINK_SPEED_100_FULL:
4299 		link.link_speed = RTE_ETH_SPEED_NUM_100M;
4300 		break;
4301 
4302 	case IXGBE_LINK_SPEED_1GB_FULL:
4303 		link.link_speed = RTE_ETH_SPEED_NUM_1G;
4304 		break;
4305 
4306 	case IXGBE_LINK_SPEED_2_5GB_FULL:
4307 		link.link_speed = RTE_ETH_SPEED_NUM_2_5G;
4308 		break;
4309 
4310 	case IXGBE_LINK_SPEED_5GB_FULL:
4311 		link.link_speed = RTE_ETH_SPEED_NUM_5G;
4312 		break;
4313 
4314 	case IXGBE_LINK_SPEED_10GB_FULL:
4315 		link.link_speed = RTE_ETH_SPEED_NUM_10G;
4316 		break;
4317 	}
4318 
4319 	return rte_eth_linkstatus_set(dev, &link);
4320 }
4321 
4322 static int
4323 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4324 {
4325 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 0);
4326 }
4327 
4328 static int
4329 ixgbevf_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4330 {
4331 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 1);
4332 }
4333 
4334 static int
4335 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
4336 {
4337 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4338 	uint32_t fctrl;
4339 
4340 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4341 	fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4342 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4343 
4344 	return 0;
4345 }
4346 
4347 static int
4348 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
4349 {
4350 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4351 	uint32_t fctrl;
4352 
4353 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4354 	fctrl &= (~IXGBE_FCTRL_UPE);
4355 	if (dev->data->all_multicast == 1)
4356 		fctrl |= IXGBE_FCTRL_MPE;
4357 	else
4358 		fctrl &= (~IXGBE_FCTRL_MPE);
4359 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4360 
4361 	return 0;
4362 }
4363 
4364 static int
4365 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
4366 {
4367 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4368 	uint32_t fctrl;
4369 
4370 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4371 	fctrl |= IXGBE_FCTRL_MPE;
4372 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4373 
4374 	return 0;
4375 }
4376 
4377 static int
4378 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
4379 {
4380 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4381 	uint32_t fctrl;
4382 
4383 	if (dev->data->promiscuous == 1)
4384 		return 0; /* must remain in all_multicast mode */
4385 
4386 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4387 	fctrl &= (~IXGBE_FCTRL_MPE);
4388 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4389 
4390 	return 0;
4391 }
4392 
4393 /**
4394  * It clears the interrupt causes and enables the interrupt.
4395  * It will be called once only during nic initialized.
4396  *
4397  * @param dev
4398  *  Pointer to struct rte_eth_dev.
4399  * @param on
4400  *  Enable or Disable.
4401  *
4402  * @return
4403  *  - On success, zero.
4404  *  - On failure, a negative value.
4405  */
4406 static int
4407 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
4408 {
4409 	struct ixgbe_interrupt *intr =
4410 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4411 
4412 	ixgbe_dev_link_status_print(dev);
4413 	if (on)
4414 		intr->mask |= IXGBE_EICR_LSC;
4415 	else
4416 		intr->mask &= ~IXGBE_EICR_LSC;
4417 
4418 	return 0;
4419 }
4420 
4421 /**
4422  * It clears the interrupt causes and enables the interrupt.
4423  * It will be called once only during nic initialized.
4424  *
4425  * @param dev
4426  *  Pointer to struct rte_eth_dev.
4427  *
4428  * @return
4429  *  - On success, zero.
4430  *  - On failure, a negative value.
4431  */
4432 static int
4433 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
4434 {
4435 	struct ixgbe_interrupt *intr =
4436 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4437 
4438 	intr->mask |= IXGBE_EICR_RTX_QUEUE;
4439 
4440 	return 0;
4441 }
4442 
4443 /**
4444  * It clears the interrupt causes and enables the interrupt.
4445  * It will be called once only during nic initialized.
4446  *
4447  * @param dev
4448  *  Pointer to struct rte_eth_dev.
4449  *
4450  * @return
4451  *  - On success, zero.
4452  *  - On failure, a negative value.
4453  */
4454 static int
4455 ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev)
4456 {
4457 	struct ixgbe_interrupt *intr =
4458 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4459 
4460 	intr->mask |= IXGBE_EICR_LINKSEC;
4461 
4462 	return 0;
4463 }
4464 
4465 /*
4466  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
4467  *
4468  * @param dev
4469  *  Pointer to struct rte_eth_dev.
4470  *
4471  * @return
4472  *  - On success, zero.
4473  *  - On failure, a negative value.
4474  */
4475 static int
4476 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
4477 {
4478 	uint32_t eicr;
4479 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4480 	struct ixgbe_interrupt *intr =
4481 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4482 
4483 	/* clear all cause mask */
4484 	ixgbe_disable_intr(hw);
4485 
4486 	/* read-on-clear nic registers here */
4487 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4488 	PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
4489 
4490 	intr->flags = 0;
4491 
4492 	/* set flag for async link update */
4493 	if (eicr & IXGBE_EICR_LSC)
4494 		intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4495 
4496 	if (eicr & IXGBE_EICR_MAILBOX)
4497 		intr->flags |= IXGBE_FLAG_MAILBOX;
4498 
4499 	if (eicr & IXGBE_EICR_LINKSEC)
4500 		intr->flags |= IXGBE_FLAG_MACSEC;
4501 
4502 	if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
4503 	    hw->phy.type == ixgbe_phy_x550em_ext_t &&
4504 	    (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
4505 		intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
4506 
4507 	return 0;
4508 }
4509 
4510 /**
4511  * It gets and then prints the link status.
4512  *
4513  * @param dev
4514  *  Pointer to struct rte_eth_dev.
4515  *
4516  * @return
4517  *  - On success, zero.
4518  *  - On failure, a negative value.
4519  */
4520 static void
4521 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
4522 {
4523 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4524 	struct rte_eth_link link;
4525 
4526 	rte_eth_linkstatus_get(dev, &link);
4527 
4528 	if (link.link_status) {
4529 		PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
4530 					(int)(dev->data->port_id),
4531 					(unsigned)link.link_speed,
4532 			link.link_duplex == RTE_ETH_LINK_FULL_DUPLEX ?
4533 					"full-duplex" : "half-duplex");
4534 	} else {
4535 		PMD_INIT_LOG(INFO, " Port %d: Link Down",
4536 				(int)(dev->data->port_id));
4537 	}
4538 	PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
4539 				pci_dev->addr.domain,
4540 				pci_dev->addr.bus,
4541 				pci_dev->addr.devid,
4542 				pci_dev->addr.function);
4543 }
4544 
4545 /*
4546  * It executes link_update after knowing an interrupt occurred.
4547  *
4548  * @param dev
4549  *  Pointer to struct rte_eth_dev.
4550  *
4551  * @return
4552  *  - On success, zero.
4553  *  - On failure, a negative value.
4554  */
4555 static int
4556 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
4557 {
4558 	struct ixgbe_interrupt *intr =
4559 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4560 	int64_t timeout;
4561 	struct ixgbe_hw *hw =
4562 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4563 
4564 	PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
4565 
4566 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
4567 		ixgbe_pf_mbx_process(dev);
4568 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
4569 	}
4570 
4571 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4572 		ixgbe_handle_lasi(hw);
4573 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4574 	}
4575 
4576 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4577 		struct rte_eth_link link;
4578 
4579 		/* get the link status before link update, for predicting later */
4580 		rte_eth_linkstatus_get(dev, &link);
4581 
4582 		ixgbe_dev_link_update(dev, 0);
4583 
4584 		/* likely to up */
4585 		if (!link.link_status)
4586 			/* handle it 1 sec later, wait it being stable */
4587 			timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
4588 		/* likely to down */
4589 		else
4590 			/* handle it 4 sec later, wait it being stable */
4591 			timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
4592 
4593 		ixgbe_dev_link_status_print(dev);
4594 		if (rte_eal_alarm_set(timeout * 1000,
4595 				      ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
4596 			PMD_DRV_LOG(ERR, "Error setting alarm");
4597 		else {
4598 			/* remember original mask */
4599 			intr->mask_original = intr->mask;
4600 			/* only disable lsc interrupt */
4601 			intr->mask &= ~IXGBE_EIMS_LSC;
4602 		}
4603 	}
4604 
4605 	PMD_DRV_LOG(DEBUG, "enable intr immediately");
4606 	ixgbe_enable_intr(dev);
4607 
4608 	return 0;
4609 }
4610 
4611 /**
4612  * Interrupt handler which shall be registered for alarm callback for delayed
4613  * handling specific interrupt to wait for the stable nic state. As the
4614  * NIC interrupt state is not stable for ixgbe after link is just down,
4615  * it needs to wait 4 seconds to get the stable status.
4616  *
4617  * @param handle
4618  *  Pointer to interrupt handle.
4619  * @param param
4620  *  The address of parameter (struct rte_eth_dev *) registered before.
4621  *
4622  * @return
4623  *  void
4624  */
4625 static void
4626 ixgbe_dev_interrupt_delayed_handler(void *param)
4627 {
4628 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4629 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4630 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
4631 	struct ixgbe_interrupt *intr =
4632 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4633 	struct ixgbe_hw *hw =
4634 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4635 	uint32_t eicr;
4636 
4637 	ixgbe_disable_intr(hw);
4638 
4639 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4640 	if (eicr & IXGBE_EICR_MAILBOX)
4641 		ixgbe_pf_mbx_process(dev);
4642 
4643 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4644 		ixgbe_handle_lasi(hw);
4645 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4646 	}
4647 
4648 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4649 		ixgbe_dev_link_update(dev, 0);
4650 		intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4651 		ixgbe_dev_link_status_print(dev);
4652 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
4653 	}
4654 
4655 	if (intr->flags & IXGBE_FLAG_MACSEC) {
4656 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_MACSEC, NULL);
4657 		intr->flags &= ~IXGBE_FLAG_MACSEC;
4658 	}
4659 
4660 	/* restore original mask */
4661 	intr->mask = intr->mask_original;
4662 	intr->mask_original = 0;
4663 
4664 	PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
4665 	ixgbe_enable_intr(dev);
4666 	rte_intr_ack(intr_handle);
4667 }
4668 
4669 /**
4670  * Interrupt handler triggered by NIC  for handling
4671  * specific interrupt.
4672  *
4673  * @param handle
4674  *  Pointer to interrupt handle.
4675  * @param param
4676  *  The address of parameter (struct rte_eth_dev *) registered before.
4677  *
4678  * @return
4679  *  void
4680  */
4681 static void
4682 ixgbe_dev_interrupt_handler(void *param)
4683 {
4684 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4685 
4686 	ixgbe_dev_interrupt_get_status(dev);
4687 	ixgbe_dev_interrupt_action(dev);
4688 }
4689 
4690 static int
4691 ixgbe_dev_led_on(struct rte_eth_dev *dev)
4692 {
4693 	struct ixgbe_hw *hw;
4694 
4695 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4696 	return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4697 }
4698 
4699 static int
4700 ixgbe_dev_led_off(struct rte_eth_dev *dev)
4701 {
4702 	struct ixgbe_hw *hw;
4703 
4704 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4705 	return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4706 }
4707 
4708 static int
4709 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4710 {
4711 	struct ixgbe_hw *hw;
4712 	uint32_t mflcn_reg;
4713 	uint32_t fccfg_reg;
4714 	int rx_pause;
4715 	int tx_pause;
4716 
4717 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4718 
4719 	fc_conf->pause_time = hw->fc.pause_time;
4720 	fc_conf->high_water = hw->fc.high_water[0];
4721 	fc_conf->low_water = hw->fc.low_water[0];
4722 	fc_conf->send_xon = hw->fc.send_xon;
4723 	fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
4724 
4725 	/*
4726 	 * Return rx_pause status according to actual setting of
4727 	 * MFLCN register.
4728 	 */
4729 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4730 	if (mflcn_reg & IXGBE_MFLCN_PMCF)
4731 		fc_conf->mac_ctrl_frame_fwd = 1;
4732 	else
4733 		fc_conf->mac_ctrl_frame_fwd = 0;
4734 
4735 	if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
4736 		rx_pause = 1;
4737 	else
4738 		rx_pause = 0;
4739 
4740 	/*
4741 	 * Return tx_pause status according to actual setting of
4742 	 * FCCFG register.
4743 	 */
4744 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4745 	if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
4746 		tx_pause = 1;
4747 	else
4748 		tx_pause = 0;
4749 
4750 	if (rx_pause && tx_pause)
4751 		fc_conf->mode = RTE_ETH_FC_FULL;
4752 	else if (rx_pause)
4753 		fc_conf->mode = RTE_ETH_FC_RX_PAUSE;
4754 	else if (tx_pause)
4755 		fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
4756 	else
4757 		fc_conf->mode = RTE_ETH_FC_NONE;
4758 
4759 	return 0;
4760 }
4761 
4762 static int
4763 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4764 {
4765 	struct ixgbe_hw *hw;
4766 	struct ixgbe_adapter *adapter = dev->data->dev_private;
4767 	int err;
4768 	uint32_t rx_buf_size;
4769 	uint32_t max_high_water;
4770 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4771 		ixgbe_fc_none,
4772 		ixgbe_fc_rx_pause,
4773 		ixgbe_fc_tx_pause,
4774 		ixgbe_fc_full
4775 	};
4776 
4777 	PMD_INIT_FUNC_TRACE();
4778 
4779 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4780 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
4781 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4782 
4783 	/*
4784 	 * At least reserve one Ethernet frame for watermark
4785 	 * high_water/low_water in kilo bytes for ixgbe
4786 	 */
4787 	max_high_water = (rx_buf_size -
4788 			RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4789 	if ((fc_conf->high_water > max_high_water) ||
4790 		(fc_conf->high_water < fc_conf->low_water)) {
4791 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4792 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4793 		return -EINVAL;
4794 	}
4795 
4796 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
4797 	hw->fc.pause_time     = fc_conf->pause_time;
4798 	hw->fc.high_water[0]  = fc_conf->high_water;
4799 	hw->fc.low_water[0]   = fc_conf->low_water;
4800 	hw->fc.send_xon       = fc_conf->send_xon;
4801 	hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
4802 	adapter->mac_ctrl_frame_fwd = fc_conf->mac_ctrl_frame_fwd;
4803 
4804 	err = ixgbe_flow_ctrl_enable(dev, hw);
4805 	if (err < 0) {
4806 		PMD_INIT_LOG(ERR, "ixgbe_flow_ctrl_enable = 0x%x", err);
4807 		return -EIO;
4808 	}
4809 	return err;
4810 }
4811 
4812 /**
4813  *  ixgbe_pfc_enable_generic - Enable flow control
4814  *  @hw: pointer to hardware structure
4815  *  @tc_num: traffic class number
4816  *  Enable flow control according to the current settings.
4817  */
4818 static int
4819 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
4820 {
4821 	int ret_val = 0;
4822 	uint32_t mflcn_reg, fccfg_reg;
4823 	uint32_t reg;
4824 	uint32_t fcrtl, fcrth;
4825 	uint8_t i;
4826 	uint8_t nb_rx_en;
4827 
4828 	/* Validate the water mark configuration */
4829 	if (!hw->fc.pause_time) {
4830 		ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4831 		goto out;
4832 	}
4833 
4834 	/* Low water mark of zero causes XOFF floods */
4835 	if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
4836 		 /* High/Low water can not be 0 */
4837 		if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_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 		if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
4844 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4845 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4846 			goto out;
4847 		}
4848 	}
4849 	/* Negotiate the fc mode to use */
4850 	ixgbe_fc_autoneg(hw);
4851 
4852 	/* Disable any previous flow control settings */
4853 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4854 	mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
4855 
4856 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4857 	fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
4858 
4859 	switch (hw->fc.current_mode) {
4860 	case ixgbe_fc_none:
4861 		/*
4862 		 * If the count of enabled RX Priority Flow control >1,
4863 		 * and the TX pause can not be disabled
4864 		 */
4865 		nb_rx_en = 0;
4866 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4867 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4868 			if (reg & IXGBE_FCRTH_FCEN)
4869 				nb_rx_en++;
4870 		}
4871 		if (nb_rx_en > 1)
4872 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4873 		break;
4874 	case ixgbe_fc_rx_pause:
4875 		/*
4876 		 * Rx Flow control is enabled and Tx Flow control is
4877 		 * disabled by software override. Since there really
4878 		 * isn't a way to advertise that we are capable of RX
4879 		 * Pause ONLY, we will advertise that we support both
4880 		 * symmetric and asymmetric Rx PAUSE.  Later, we will
4881 		 * disable the adapter's ability to send PAUSE frames.
4882 		 */
4883 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4884 		/*
4885 		 * If the count of enabled RX Priority Flow control >1,
4886 		 * and the TX pause can not be disabled
4887 		 */
4888 		nb_rx_en = 0;
4889 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4890 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4891 			if (reg & IXGBE_FCRTH_FCEN)
4892 				nb_rx_en++;
4893 		}
4894 		if (nb_rx_en > 1)
4895 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4896 		break;
4897 	case ixgbe_fc_tx_pause:
4898 		/*
4899 		 * Tx Flow control is enabled, and Rx Flow control is
4900 		 * disabled by software override.
4901 		 */
4902 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4903 		break;
4904 	case ixgbe_fc_full:
4905 		/* Flow control (both Rx and Tx) is enabled by SW override. */
4906 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4907 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4908 		break;
4909 	default:
4910 		PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
4911 		ret_val = IXGBE_ERR_CONFIG;
4912 		goto out;
4913 	}
4914 
4915 	/* Set 802.3x based flow control settings. */
4916 	mflcn_reg |= IXGBE_MFLCN_DPF;
4917 	IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
4918 	IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
4919 
4920 	/* Set up and enable Rx high/low water mark thresholds, enable XON. */
4921 	if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
4922 		hw->fc.high_water[tc_num]) {
4923 		fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
4924 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
4925 		fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
4926 	} else {
4927 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
4928 		/*
4929 		 * In order to prevent Tx hangs when the internal Tx
4930 		 * switch is enabled we must set the high water mark
4931 		 * to the maximum FCRTH value.  This allows the Tx
4932 		 * switch to function even under heavy Rx workloads.
4933 		 */
4934 		fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
4935 	}
4936 	IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
4937 
4938 	/* Configure pause time (2 TCs per register) */
4939 	reg = hw->fc.pause_time * 0x00010001;
4940 	for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
4941 		IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
4942 
4943 	/* Configure flow control refresh threshold value */
4944 	IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
4945 
4946 out:
4947 	return ret_val;
4948 }
4949 
4950 static int
4951 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
4952 {
4953 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4954 	int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
4955 
4956 	if (hw->mac.type != ixgbe_mac_82598EB) {
4957 		ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
4958 	}
4959 	return ret_val;
4960 }
4961 
4962 static int
4963 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
4964 {
4965 	int err;
4966 	uint32_t rx_buf_size;
4967 	uint32_t max_high_water;
4968 	uint8_t tc_num;
4969 	uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
4970 	struct ixgbe_hw *hw =
4971 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4972 	struct ixgbe_dcb_config *dcb_config =
4973 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
4974 
4975 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4976 		ixgbe_fc_none,
4977 		ixgbe_fc_rx_pause,
4978 		ixgbe_fc_tx_pause,
4979 		ixgbe_fc_full
4980 	};
4981 
4982 	PMD_INIT_FUNC_TRACE();
4983 
4984 	ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
4985 	tc_num = map[pfc_conf->priority];
4986 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
4987 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4988 	/*
4989 	 * At least reserve one Ethernet frame for watermark
4990 	 * high_water/low_water in kilo bytes for ixgbe
4991 	 */
4992 	max_high_water = (rx_buf_size -
4993 			RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4994 	if ((pfc_conf->fc.high_water > max_high_water) ||
4995 	    (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
4996 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4997 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4998 		return -EINVAL;
4999 	}
5000 
5001 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
5002 	hw->fc.pause_time = pfc_conf->fc.pause_time;
5003 	hw->fc.send_xon = pfc_conf->fc.send_xon;
5004 	hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
5005 	hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
5006 
5007 	err = ixgbe_dcb_pfc_enable(dev, tc_num);
5008 
5009 	/* Not negotiated is not an error case */
5010 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
5011 		return 0;
5012 
5013 	PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
5014 	return -EIO;
5015 }
5016 
5017 static int
5018 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
5019 			  struct rte_eth_rss_reta_entry64 *reta_conf,
5020 			  uint16_t reta_size)
5021 {
5022 	uint16_t i, sp_reta_size;
5023 	uint8_t j, mask;
5024 	uint32_t reta, r;
5025 	uint16_t idx, shift;
5026 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5027 	struct rte_eth_dev_data *dev_data = dev->data;
5028 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5029 	uint32_t reta_reg;
5030 
5031 	PMD_INIT_FUNC_TRACE();
5032 
5033 	if (!dev_data->dev_started) {
5034 		PMD_DRV_LOG(ERR,
5035 			"port %d must be started before rss reta update",
5036 			 dev_data->port_id);
5037 		return -EIO;
5038 	}
5039 
5040 	if (!ixgbe_rss_update_sp(hw->mac.type)) {
5041 		PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
5042 			"NIC.");
5043 		return -ENOTSUP;
5044 	}
5045 
5046 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5047 	if (reta_size != sp_reta_size) {
5048 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5049 			"(%d) doesn't match the number hardware can supported "
5050 			"(%d)", reta_size, sp_reta_size);
5051 		return -EINVAL;
5052 	}
5053 
5054 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5055 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
5056 		shift = i % RTE_ETH_RETA_GROUP_SIZE;
5057 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5058 						IXGBE_4_BIT_MASK);
5059 		if (!mask)
5060 			continue;
5061 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5062 		if (mask == IXGBE_4_BIT_MASK)
5063 			r = 0;
5064 		else
5065 			r = IXGBE_READ_REG(hw, reta_reg);
5066 		for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5067 			if (mask & (0x1 << j))
5068 				reta |= reta_conf[idx].reta[shift + j] <<
5069 							(CHAR_BIT * j);
5070 			else
5071 				reta |= r & (IXGBE_8_BIT_MASK <<
5072 						(CHAR_BIT * j));
5073 		}
5074 		IXGBE_WRITE_REG(hw, reta_reg, reta);
5075 	}
5076 	adapter->rss_reta_updated = 1;
5077 
5078 	return 0;
5079 }
5080 
5081 static int
5082 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
5083 			 struct rte_eth_rss_reta_entry64 *reta_conf,
5084 			 uint16_t reta_size)
5085 {
5086 	uint16_t i, sp_reta_size;
5087 	uint8_t j, mask;
5088 	uint32_t reta;
5089 	uint16_t idx, shift;
5090 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5091 	uint32_t reta_reg;
5092 
5093 	PMD_INIT_FUNC_TRACE();
5094 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5095 	if (reta_size != sp_reta_size) {
5096 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5097 			"(%d) doesn't match the number hardware can supported "
5098 			"(%d)", reta_size, sp_reta_size);
5099 		return -EINVAL;
5100 	}
5101 
5102 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5103 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
5104 		shift = i % RTE_ETH_RETA_GROUP_SIZE;
5105 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5106 						IXGBE_4_BIT_MASK);
5107 		if (!mask)
5108 			continue;
5109 
5110 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5111 		reta = IXGBE_READ_REG(hw, reta_reg);
5112 		for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5113 			if (mask & (0x1 << j))
5114 				reta_conf[idx].reta[shift + j] =
5115 					((reta >> (CHAR_BIT * j)) &
5116 						IXGBE_8_BIT_MASK);
5117 		}
5118 	}
5119 
5120 	return 0;
5121 }
5122 
5123 static int
5124 ixgbe_add_rar(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
5125 				uint32_t index, uint32_t pool)
5126 {
5127 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5128 	uint32_t enable_addr = 1;
5129 
5130 	return ixgbe_set_rar(hw, index, mac_addr->addr_bytes,
5131 			     pool, enable_addr);
5132 }
5133 
5134 static void
5135 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
5136 {
5137 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5138 
5139 	ixgbe_clear_rar(hw, index);
5140 }
5141 
5142 static int
5143 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
5144 {
5145 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5146 
5147 	ixgbe_remove_rar(dev, 0);
5148 	ixgbe_add_rar(dev, addr, 0, pci_dev->max_vfs);
5149 
5150 	return 0;
5151 }
5152 
5153 static bool
5154 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
5155 {
5156 	if (strcmp(dev->device->driver->name, drv->driver.name))
5157 		return false;
5158 
5159 	return true;
5160 }
5161 
5162 bool
5163 is_ixgbe_supported(struct rte_eth_dev *dev)
5164 {
5165 	return is_device_supported(dev, &rte_ixgbe_pmd);
5166 }
5167 
5168 static int
5169 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
5170 {
5171 	uint32_t hlreg0;
5172 	uint32_t maxfrs;
5173 	struct ixgbe_hw *hw;
5174 	struct rte_eth_dev_info dev_info;
5175 	uint32_t frame_size = mtu + IXGBE_ETH_OVERHEAD;
5176 	int ret;
5177 
5178 	ret = ixgbe_dev_info_get(dev, &dev_info);
5179 	if (ret != 0)
5180 		return ret;
5181 
5182 	/* check that mtu is within the allowed range */
5183 	if (mtu < RTE_ETHER_MIN_MTU || frame_size > dev_info.max_rx_pktlen)
5184 		return -EINVAL;
5185 
5186 	/* If device is started, refuse mtu that requires the support of
5187 	 * scattered packets when this feature has not been enabled before.
5188 	 */
5189 	if (dev->data->dev_started && !dev->data->scattered_rx &&
5190 	    frame_size + 2 * RTE_VLAN_HLEN >
5191 			dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
5192 		PMD_INIT_LOG(ERR, "Stop port first.");
5193 		return -EINVAL;
5194 	}
5195 
5196 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5197 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
5198 
5199 	/* switch to jumbo mode if needed */
5200 	if (mtu > RTE_ETHER_MTU)
5201 		hlreg0 |= IXGBE_HLREG0_JUMBOEN;
5202 	else
5203 		hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
5204 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
5205 
5206 	maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
5207 	maxfrs &= 0x0000FFFF;
5208 	maxfrs |= (frame_size << 16);
5209 	IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
5210 
5211 	return 0;
5212 }
5213 
5214 /*
5215  * Virtual Function operations
5216  */
5217 static void
5218 ixgbevf_intr_disable(struct rte_eth_dev *dev)
5219 {
5220 	struct ixgbe_interrupt *intr =
5221 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5222 	struct ixgbe_hw *hw =
5223 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5224 
5225 	PMD_INIT_FUNC_TRACE();
5226 
5227 	/* Clear interrupt mask to stop from interrupts being generated */
5228 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
5229 
5230 	IXGBE_WRITE_FLUSH(hw);
5231 
5232 	/* Clear mask value. */
5233 	intr->mask = 0;
5234 }
5235 
5236 static void
5237 ixgbevf_intr_enable(struct rte_eth_dev *dev)
5238 {
5239 	struct ixgbe_interrupt *intr =
5240 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5241 	struct ixgbe_hw *hw =
5242 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5243 
5244 	PMD_INIT_FUNC_TRACE();
5245 
5246 	/* VF enable interrupt autoclean */
5247 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
5248 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
5249 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
5250 
5251 	IXGBE_WRITE_FLUSH(hw);
5252 
5253 	/* Save IXGBE_VTEIMS value to mask. */
5254 	intr->mask = IXGBE_VF_IRQ_ENABLE_MASK;
5255 }
5256 
5257 static int
5258 ixgbevf_dev_configure(struct rte_eth_dev *dev)
5259 {
5260 	struct rte_eth_conf *conf = &dev->data->dev_conf;
5261 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5262 
5263 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
5264 		     dev->data->port_id);
5265 
5266 	if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
5267 		dev->data->dev_conf.rxmode.offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
5268 
5269 	/*
5270 	 * VF has no ability to enable/disable HW CRC
5271 	 * Keep the persistent behavior the same as Host PF
5272 	 */
5273 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
5274 	if (conf->rxmode.offloads & RTE_ETH_RX_OFFLOAD_KEEP_CRC) {
5275 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
5276 		conf->rxmode.offloads &= ~RTE_ETH_RX_OFFLOAD_KEEP_CRC;
5277 	}
5278 #else
5279 	if (!(conf->rxmode.offloads & RTE_ETH_RX_OFFLOAD_KEEP_CRC)) {
5280 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
5281 		conf->rxmode.offloads |= RTE_ETH_RX_OFFLOAD_KEEP_CRC;
5282 	}
5283 #endif
5284 
5285 	/*
5286 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
5287 	 * allocation or vector Rx preconditions we will reset it.
5288 	 */
5289 	adapter->rx_bulk_alloc_allowed = true;
5290 	adapter->rx_vec_allowed = true;
5291 
5292 	return 0;
5293 }
5294 
5295 static int
5296 ixgbevf_dev_start(struct rte_eth_dev *dev)
5297 {
5298 	struct ixgbe_hw *hw =
5299 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5300 	uint32_t intr_vector = 0;
5301 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5302 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5303 
5304 	int err, mask = 0;
5305 
5306 	PMD_INIT_FUNC_TRACE();
5307 
5308 	/* Stop the link setup handler before resetting the HW. */
5309 	ixgbe_dev_wait_setup_link_complete(dev, 0);
5310 
5311 	err = hw->mac.ops.reset_hw(hw);
5312 
5313 	/**
5314 	 * In this case, reuses the MAC address assigned by VF
5315 	 * initialization.
5316 	 */
5317 	if (err != IXGBE_SUCCESS && err != IXGBE_ERR_INVALID_MAC_ADDR) {
5318 		PMD_INIT_LOG(ERR, "Unable to reset vf hardware (%d)", err);
5319 		return err;
5320 	}
5321 
5322 	hw->mac.get_link_status = true;
5323 
5324 	/* negotiate mailbox API version to use with the PF. */
5325 	ixgbevf_negotiate_api(hw);
5326 
5327 	ixgbevf_dev_tx_init(dev);
5328 
5329 	/* This can fail when allocating mbufs for descriptor rings */
5330 	err = ixgbevf_dev_rx_init(dev);
5331 	if (err) {
5332 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
5333 		ixgbe_dev_clear_queues(dev);
5334 		return err;
5335 	}
5336 
5337 	/* Set vfta */
5338 	ixgbevf_set_vfta_all(dev, 1);
5339 
5340 	/* Set HW strip */
5341 	mask = RTE_ETH_VLAN_STRIP_MASK | RTE_ETH_VLAN_FILTER_MASK |
5342 		RTE_ETH_VLAN_EXTEND_MASK;
5343 	err = ixgbevf_vlan_offload_config(dev, mask);
5344 	if (err) {
5345 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
5346 		ixgbe_dev_clear_queues(dev);
5347 		return err;
5348 	}
5349 
5350 	ixgbevf_dev_rxtx_start(dev);
5351 
5352 	/* check and configure queue intr-vector mapping */
5353 	if (rte_intr_cap_multiple(intr_handle) &&
5354 	    dev->data->dev_conf.intr_conf.rxq) {
5355 		/* According to datasheet, only vector 0/1/2 can be used,
5356 		 * now only one vector is used for Rx queue
5357 		 */
5358 		intr_vector = 1;
5359 		if (rte_intr_efd_enable(intr_handle, intr_vector)) {
5360 			ixgbe_dev_clear_queues(dev);
5361 			return -1;
5362 		}
5363 	}
5364 
5365 	if (rte_intr_dp_is_en(intr_handle)) {
5366 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
5367 						   dev->data->nb_rx_queues)) {
5368 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
5369 				     " intr_vec", dev->data->nb_rx_queues);
5370 			ixgbe_dev_clear_queues(dev);
5371 			return -ENOMEM;
5372 		}
5373 	}
5374 	ixgbevf_configure_msix(dev);
5375 
5376 	/* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
5377 	 * is mapped to VFIO vector 0 in eth_ixgbevf_dev_init( ).
5378 	 * If previous VFIO interrupt mapping setting in eth_ixgbevf_dev_init( )
5379 	 * is not cleared, it will fail when following rte_intr_enable( ) tries
5380 	 * to map Rx queue interrupt to other VFIO vectors.
5381 	 * So clear uio/vfio intr/evevnfd first to avoid failure.
5382 	 */
5383 	rte_intr_disable(intr_handle);
5384 
5385 	rte_intr_enable(intr_handle);
5386 
5387 	/* Re-enable interrupt for VF */
5388 	ixgbevf_intr_enable(dev);
5389 
5390 	/*
5391 	 * Update link status right before return, because it may
5392 	 * start link configuration process in a separate thread.
5393 	 */
5394 	ixgbevf_dev_link_update(dev, 0);
5395 
5396 	hw->adapter_stopped = false;
5397 
5398 	return 0;
5399 }
5400 
5401 static int
5402 ixgbevf_dev_stop(struct rte_eth_dev *dev)
5403 {
5404 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5405 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5406 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5407 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5408 
5409 	if (hw->adapter_stopped)
5410 		return 0;
5411 
5412 	PMD_INIT_FUNC_TRACE();
5413 
5414 	ixgbe_dev_wait_setup_link_complete(dev, 0);
5415 
5416 	ixgbevf_intr_disable(dev);
5417 
5418 	dev->data->dev_started = 0;
5419 	hw->adapter_stopped = 1;
5420 	ixgbe_stop_adapter(hw);
5421 
5422 	/*
5423 	  * Clear what we set, but we still keep shadow_vfta to
5424 	  * restore after device starts
5425 	  */
5426 	ixgbevf_set_vfta_all(dev, 0);
5427 
5428 	/* Clear stored conf */
5429 	dev->data->scattered_rx = 0;
5430 
5431 	ixgbe_dev_clear_queues(dev);
5432 
5433 	/* Clean datapath event and queue/vec mapping */
5434 	rte_intr_efd_disable(intr_handle);
5435 	rte_intr_vec_list_free(intr_handle);
5436 
5437 	adapter->rss_reta_updated = 0;
5438 
5439 	return 0;
5440 }
5441 
5442 static int
5443 ixgbevf_dev_close(struct rte_eth_dev *dev)
5444 {
5445 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5446 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5447 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5448 	int ret;
5449 
5450 	PMD_INIT_FUNC_TRACE();
5451 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5452 		return 0;
5453 
5454 	ixgbe_reset_hw(hw);
5455 
5456 	ret = ixgbevf_dev_stop(dev);
5457 
5458 	ixgbe_dev_free_queues(dev);
5459 
5460 	/**
5461 	 * Remove the VF MAC address ro ensure
5462 	 * that the VF traffic goes to the PF
5463 	 * after stop, close and detach of the VF
5464 	 **/
5465 	ixgbevf_remove_mac_addr(dev, 0);
5466 
5467 	rte_intr_disable(intr_handle);
5468 	rte_intr_callback_unregister(intr_handle,
5469 				     ixgbevf_dev_interrupt_handler, dev);
5470 
5471 	return ret;
5472 }
5473 
5474 /*
5475  * Reset VF device
5476  */
5477 static int
5478 ixgbevf_dev_reset(struct rte_eth_dev *dev)
5479 {
5480 	int ret;
5481 
5482 	ret = eth_ixgbevf_dev_uninit(dev);
5483 	if (ret)
5484 		return ret;
5485 
5486 	ret = eth_ixgbevf_dev_init(dev);
5487 
5488 	return ret;
5489 }
5490 
5491 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
5492 {
5493 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5494 	struct ixgbe_vfta *shadow_vfta =
5495 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5496 	int i = 0, j = 0, vfta = 0, mask = 1;
5497 
5498 	for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
5499 		vfta = shadow_vfta->vfta[i];
5500 		if (vfta) {
5501 			mask = 1;
5502 			for (j = 0; j < 32; j++) {
5503 				if (vfta & mask)
5504 					ixgbe_set_vfta(hw, (i<<5)+j, 0,
5505 						       on, false);
5506 				mask <<= 1;
5507 			}
5508 		}
5509 	}
5510 
5511 }
5512 
5513 static int
5514 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
5515 {
5516 	struct ixgbe_hw *hw =
5517 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5518 	struct ixgbe_vfta *shadow_vfta =
5519 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5520 	uint32_t vid_idx = 0;
5521 	uint32_t vid_bit = 0;
5522 	int ret = 0;
5523 
5524 	PMD_INIT_FUNC_TRACE();
5525 
5526 	/* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
5527 	ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
5528 	if (ret) {
5529 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
5530 		return ret;
5531 	}
5532 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
5533 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
5534 
5535 	/* Save what we set and retore it after device reset */
5536 	if (on)
5537 		shadow_vfta->vfta[vid_idx] |= vid_bit;
5538 	else
5539 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
5540 
5541 	return 0;
5542 }
5543 
5544 static void
5545 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
5546 {
5547 	struct ixgbe_hw *hw =
5548 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5549 	uint32_t ctrl;
5550 
5551 	PMD_INIT_FUNC_TRACE();
5552 
5553 	if (queue >= hw->mac.max_rx_queues)
5554 		return;
5555 
5556 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
5557 	if (on)
5558 		ctrl |= IXGBE_RXDCTL_VME;
5559 	else
5560 		ctrl &= ~IXGBE_RXDCTL_VME;
5561 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
5562 
5563 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
5564 }
5565 
5566 static int
5567 ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask)
5568 {
5569 	struct ixgbe_rx_queue *rxq;
5570 	uint16_t i;
5571 	int on = 0;
5572 
5573 	/* VF function only support hw strip feature, others are not support */
5574 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
5575 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
5576 			rxq = dev->data->rx_queues[i];
5577 			on = !!(rxq->offloads &	RTE_ETH_RX_OFFLOAD_VLAN_STRIP);
5578 			ixgbevf_vlan_strip_queue_set(dev, i, on);
5579 		}
5580 	}
5581 
5582 	return 0;
5583 }
5584 
5585 static int
5586 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
5587 {
5588 	ixgbe_config_vlan_strip_on_all_queues(dev, mask);
5589 
5590 	ixgbevf_vlan_offload_config(dev, mask);
5591 
5592 	return 0;
5593 }
5594 
5595 int
5596 ixgbe_vt_check(struct ixgbe_hw *hw)
5597 {
5598 	uint32_t reg_val;
5599 
5600 	/* if Virtualization Technology is enabled */
5601 	reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
5602 	if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
5603 		PMD_INIT_LOG(ERR, "VT must be enabled for this setting");
5604 		return -1;
5605 	}
5606 
5607 	return 0;
5608 }
5609 
5610 static uint32_t
5611 ixgbe_uta_vector(struct ixgbe_hw *hw, struct rte_ether_addr *uc_addr)
5612 {
5613 	uint32_t vector = 0;
5614 
5615 	switch (hw->mac.mc_filter_type) {
5616 	case 0:   /* use bits [47:36] of the address */
5617 		vector = ((uc_addr->addr_bytes[4] >> 4) |
5618 			(((uint16_t)uc_addr->addr_bytes[5]) << 4));
5619 		break;
5620 	case 1:   /* use bits [46:35] of the address */
5621 		vector = ((uc_addr->addr_bytes[4] >> 3) |
5622 			(((uint16_t)uc_addr->addr_bytes[5]) << 5));
5623 		break;
5624 	case 2:   /* use bits [45:34] of the address */
5625 		vector = ((uc_addr->addr_bytes[4] >> 2) |
5626 			(((uint16_t)uc_addr->addr_bytes[5]) << 6));
5627 		break;
5628 	case 3:   /* use bits [43:32] of the address */
5629 		vector = ((uc_addr->addr_bytes[4]) |
5630 			(((uint16_t)uc_addr->addr_bytes[5]) << 8));
5631 		break;
5632 	default:  /* Invalid mc_filter_type */
5633 		break;
5634 	}
5635 
5636 	/* vector can only be 12-bits or boundary will be exceeded */
5637 	vector &= 0xFFF;
5638 	return vector;
5639 }
5640 
5641 static int
5642 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev,
5643 			struct rte_ether_addr *mac_addr, uint8_t on)
5644 {
5645 	uint32_t vector;
5646 	uint32_t uta_idx;
5647 	uint32_t reg_val;
5648 	uint32_t uta_shift;
5649 	uint32_t rc;
5650 	const uint32_t ixgbe_uta_idx_mask = 0x7F;
5651 	const uint32_t ixgbe_uta_bit_shift = 5;
5652 	const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
5653 	const uint32_t bit1 = 0x1;
5654 
5655 	struct ixgbe_hw *hw =
5656 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5657 	struct ixgbe_uta_info *uta_info =
5658 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5659 
5660 	/* The UTA table only exists on 82599 hardware and newer */
5661 	if (hw->mac.type < ixgbe_mac_82599EB)
5662 		return -ENOTSUP;
5663 
5664 	vector = ixgbe_uta_vector(hw, mac_addr);
5665 	uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
5666 	uta_shift = vector & ixgbe_uta_bit_mask;
5667 
5668 	rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
5669 	if (rc == on)
5670 		return 0;
5671 
5672 	reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
5673 	if (on) {
5674 		uta_info->uta_in_use++;
5675 		reg_val |= (bit1 << uta_shift);
5676 		uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
5677 	} else {
5678 		uta_info->uta_in_use--;
5679 		reg_val &= ~(bit1 << uta_shift);
5680 		uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
5681 	}
5682 
5683 	IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
5684 
5685 	if (uta_info->uta_in_use > 0)
5686 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
5687 				IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
5688 	else
5689 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
5690 
5691 	return 0;
5692 }
5693 
5694 static int
5695 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
5696 {
5697 	int i;
5698 	struct ixgbe_hw *hw =
5699 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5700 	struct ixgbe_uta_info *uta_info =
5701 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5702 
5703 	/* The UTA table only exists on 82599 hardware and newer */
5704 	if (hw->mac.type < ixgbe_mac_82599EB)
5705 		return -ENOTSUP;
5706 
5707 	if (on) {
5708 		for (i = 0; i < RTE_ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5709 			uta_info->uta_shadow[i] = ~0;
5710 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
5711 		}
5712 	} else {
5713 		for (i = 0; i < RTE_ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5714 			uta_info->uta_shadow[i] = 0;
5715 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
5716 		}
5717 	}
5718 	return 0;
5719 
5720 }
5721 
5722 uint32_t
5723 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
5724 {
5725 	uint32_t new_val = orig_val;
5726 
5727 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_UNTAG)
5728 		new_val |= IXGBE_VMOLR_AUPE;
5729 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_HASH_MC)
5730 		new_val |= IXGBE_VMOLR_ROMPE;
5731 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_HASH_UC)
5732 		new_val |= IXGBE_VMOLR_ROPE;
5733 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_BROADCAST)
5734 		new_val |= IXGBE_VMOLR_BAM;
5735 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_MULTICAST)
5736 		new_val |= IXGBE_VMOLR_MPE;
5737 
5738 	return new_val;
5739 }
5740 
5741 static int
5742 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5743 {
5744 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5745 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5746 	struct ixgbe_interrupt *intr =
5747 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5748 	struct ixgbe_hw *hw =
5749 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5750 	uint32_t vec = IXGBE_MISC_VEC_ID;
5751 
5752 	if (rte_intr_allow_others(intr_handle))
5753 		vec = IXGBE_RX_VEC_START;
5754 	intr->mask |= (1 << vec);
5755 	RTE_SET_USED(queue_id);
5756 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5757 
5758 	rte_intr_ack(intr_handle);
5759 
5760 	return 0;
5761 }
5762 
5763 static int
5764 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5765 {
5766 	struct ixgbe_interrupt *intr =
5767 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5768 	struct ixgbe_hw *hw =
5769 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5770 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5771 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5772 	uint32_t vec = IXGBE_MISC_VEC_ID;
5773 
5774 	if (rte_intr_allow_others(intr_handle))
5775 		vec = IXGBE_RX_VEC_START;
5776 	intr->mask &= ~(1 << vec);
5777 	RTE_SET_USED(queue_id);
5778 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5779 
5780 	return 0;
5781 }
5782 
5783 static int
5784 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5785 {
5786 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5787 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5788 	uint32_t mask;
5789 	struct ixgbe_hw *hw =
5790 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5791 	struct ixgbe_interrupt *intr =
5792 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5793 
5794 	if (queue_id < 16) {
5795 		ixgbe_disable_intr(hw);
5796 		intr->mask |= (1 << queue_id);
5797 		ixgbe_enable_intr(dev);
5798 	} else if (queue_id < 32) {
5799 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5800 		mask &= (1 << queue_id);
5801 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5802 	} else if (queue_id < 64) {
5803 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5804 		mask &= (1 << (queue_id - 32));
5805 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5806 	}
5807 	rte_intr_ack(intr_handle);
5808 
5809 	return 0;
5810 }
5811 
5812 static int
5813 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5814 {
5815 	uint32_t mask;
5816 	struct ixgbe_hw *hw =
5817 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5818 	struct ixgbe_interrupt *intr =
5819 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5820 
5821 	if (queue_id < 16) {
5822 		ixgbe_disable_intr(hw);
5823 		intr->mask &= ~(1 << queue_id);
5824 		ixgbe_enable_intr(dev);
5825 	} else if (queue_id < 32) {
5826 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5827 		mask &= ~(1 << queue_id);
5828 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5829 	} else if (queue_id < 64) {
5830 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5831 		mask &= ~(1 << (queue_id - 32));
5832 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5833 	}
5834 
5835 	return 0;
5836 }
5837 
5838 static void
5839 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5840 		     uint8_t queue, uint8_t msix_vector)
5841 {
5842 	uint32_t tmp, idx;
5843 
5844 	if (direction == -1) {
5845 		/* other causes */
5846 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5847 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
5848 		tmp &= ~0xFF;
5849 		tmp |= msix_vector;
5850 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
5851 	} else {
5852 		/* rx or tx cause */
5853 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5854 		idx = ((16 * (queue & 1)) + (8 * direction));
5855 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
5856 		tmp &= ~(0xFF << idx);
5857 		tmp |= (msix_vector << idx);
5858 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
5859 	}
5860 }
5861 
5862 /**
5863  * set the IVAR registers, mapping interrupt causes to vectors
5864  * @param hw
5865  *  pointer to ixgbe_hw struct
5866  * @direction
5867  *  0 for Rx, 1 for Tx, -1 for other causes
5868  * @queue
5869  *  queue to map the corresponding interrupt to
5870  * @msix_vector
5871  *  the vector to map to the corresponding queue
5872  */
5873 static void
5874 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5875 		   uint8_t queue, uint8_t msix_vector)
5876 {
5877 	uint32_t tmp, idx;
5878 
5879 	msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5880 	if (hw->mac.type == ixgbe_mac_82598EB) {
5881 		if (direction == -1)
5882 			direction = 0;
5883 		idx = (((direction * 64) + queue) >> 2) & 0x1F;
5884 		tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
5885 		tmp &= ~(0xFF << (8 * (queue & 0x3)));
5886 		tmp |= (msix_vector << (8 * (queue & 0x3)));
5887 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
5888 	} else if ((hw->mac.type == ixgbe_mac_82599EB) ||
5889 			(hw->mac.type == ixgbe_mac_X540) ||
5890 			(hw->mac.type == ixgbe_mac_X550) ||
5891 			(hw->mac.type == ixgbe_mac_X550EM_x)) {
5892 		if (direction == -1) {
5893 			/* other causes */
5894 			idx = ((queue & 1) * 8);
5895 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
5896 			tmp &= ~(0xFF << idx);
5897 			tmp |= (msix_vector << idx);
5898 			IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
5899 		} else {
5900 			/* rx or tx causes */
5901 			idx = ((16 * (queue & 1)) + (8 * direction));
5902 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
5903 			tmp &= ~(0xFF << idx);
5904 			tmp |= (msix_vector << idx);
5905 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
5906 		}
5907 	}
5908 }
5909 
5910 static void
5911 ixgbevf_configure_msix(struct rte_eth_dev *dev)
5912 {
5913 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5914 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5915 	struct ixgbe_hw *hw =
5916 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5917 	uint32_t q_idx;
5918 	uint32_t vector_idx = IXGBE_MISC_VEC_ID;
5919 	uint32_t base = IXGBE_MISC_VEC_ID;
5920 
5921 	/* Configure VF other cause ivar */
5922 	ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
5923 
5924 	/* won't configure msix register if no mapping is done
5925 	 * between intr vector and event fd.
5926 	 */
5927 	if (!rte_intr_dp_is_en(intr_handle))
5928 		return;
5929 
5930 	if (rte_intr_allow_others(intr_handle)) {
5931 		base = IXGBE_RX_VEC_START;
5932 		vector_idx = IXGBE_RX_VEC_START;
5933 	}
5934 
5935 	/* Configure all RX queues of VF */
5936 	for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
5937 		/* Force all queue use vector 0,
5938 		 * as IXGBE_VF_MAXMSIVECTOR = 1
5939 		 */
5940 		ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
5941 		rte_intr_vec_list_index_set(intr_handle, q_idx,
5942 						   vector_idx);
5943 		if (vector_idx < base + rte_intr_nb_efd_get(intr_handle)
5944 		    - 1)
5945 			vector_idx++;
5946 	}
5947 
5948 	/* As RX queue setting above show, all queues use the vector 0.
5949 	 * Set only the ITR value of IXGBE_MISC_VEC_ID.
5950 	 */
5951 	IXGBE_WRITE_REG(hw, IXGBE_VTEITR(IXGBE_MISC_VEC_ID),
5952 			IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
5953 			| IXGBE_EITR_CNT_WDIS);
5954 }
5955 
5956 /**
5957  * Sets up the hardware to properly generate MSI-X interrupts
5958  * @hw
5959  *  board private structure
5960  */
5961 static void
5962 ixgbe_configure_msix(struct rte_eth_dev *dev)
5963 {
5964 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5965 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5966 	struct ixgbe_hw *hw =
5967 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5968 	uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
5969 	uint32_t vec = IXGBE_MISC_VEC_ID;
5970 	uint32_t mask;
5971 	uint32_t gpie;
5972 
5973 	/* won't configure msix register if no mapping is done
5974 	 * between intr vector and event fd
5975 	 * but if misx has been enabled already, need to configure
5976 	 * auto clean, auto mask and throttling.
5977 	 */
5978 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5979 	if (!rte_intr_dp_is_en(intr_handle) &&
5980 	    !(gpie & (IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT)))
5981 		return;
5982 
5983 	if (rte_intr_allow_others(intr_handle))
5984 		vec = base = IXGBE_RX_VEC_START;
5985 
5986 	/* setup GPIE for MSI-x mode */
5987 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5988 	gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5989 		IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
5990 	/* auto clearing and auto setting corresponding bits in EIMS
5991 	 * when MSI-X interrupt is triggered
5992 	 */
5993 	if (hw->mac.type == ixgbe_mac_82598EB) {
5994 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5995 	} else {
5996 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5997 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5998 	}
5999 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
6000 
6001 	/* Populate the IVAR table and set the ITR values to the
6002 	 * corresponding register.
6003 	 */
6004 	if (rte_intr_dp_is_en(intr_handle)) {
6005 		for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
6006 			queue_id++) {
6007 			/* by default, 1:1 mapping */
6008 			ixgbe_set_ivar_map(hw, 0, queue_id, vec);
6009 			rte_intr_vec_list_index_set(intr_handle,
6010 							   queue_id, vec);
6011 			if (vec < base + rte_intr_nb_efd_get(intr_handle)
6012 			    - 1)
6013 				vec++;
6014 		}
6015 
6016 		switch (hw->mac.type) {
6017 		case ixgbe_mac_82598EB:
6018 			ixgbe_set_ivar_map(hw, -1,
6019 					   IXGBE_IVAR_OTHER_CAUSES_INDEX,
6020 					   IXGBE_MISC_VEC_ID);
6021 			break;
6022 		case ixgbe_mac_82599EB:
6023 		case ixgbe_mac_X540:
6024 		case ixgbe_mac_X550:
6025 		case ixgbe_mac_X550EM_x:
6026 			ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
6027 			break;
6028 		default:
6029 			break;
6030 		}
6031 	}
6032 	IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
6033 			IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
6034 			| IXGBE_EITR_CNT_WDIS);
6035 
6036 	/* set up to autoclear timer, and the vectors */
6037 	mask = IXGBE_EIMS_ENABLE_MASK;
6038 	mask &= ~(IXGBE_EIMS_OTHER |
6039 		  IXGBE_EIMS_MAILBOX |
6040 		  IXGBE_EIMS_LSC);
6041 
6042 	IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
6043 }
6044 
6045 int
6046 ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
6047 			   uint16_t queue_idx, uint16_t tx_rate)
6048 {
6049 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6050 	uint32_t rf_dec, rf_int;
6051 	uint32_t bcnrc_val;
6052 	uint16_t link_speed = dev->data->dev_link.link_speed;
6053 
6054 	if (queue_idx >= hw->mac.max_tx_queues)
6055 		return -EINVAL;
6056 
6057 	if (tx_rate != 0) {
6058 		/* Calculate the rate factor values to set */
6059 		rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
6060 		rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
6061 		rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
6062 
6063 		bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
6064 		bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
6065 				IXGBE_RTTBCNRC_RF_INT_MASK_M);
6066 		bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
6067 	} else {
6068 		bcnrc_val = 0;
6069 	}
6070 
6071 	/*
6072 	 * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
6073 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
6074 	 * set as 0x4.
6075 	 */
6076 	if (dev->data->mtu + IXGBE_ETH_OVERHEAD >= IXGBE_MAX_JUMBO_FRAME_SIZE)
6077 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM, IXGBE_MMW_SIZE_JUMBO_FRAME);
6078 	else
6079 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM, IXGBE_MMW_SIZE_DEFAULT);
6080 
6081 	/* Set RTTBCNRC of queue X */
6082 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
6083 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
6084 	IXGBE_WRITE_FLUSH(hw);
6085 
6086 	return 0;
6087 }
6088 
6089 static int
6090 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
6091 		     __rte_unused uint32_t index,
6092 		     __rte_unused uint32_t pool)
6093 {
6094 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6095 	int diag;
6096 
6097 	/*
6098 	 * On a 82599 VF, adding again the same MAC addr is not an idempotent
6099 	 * operation. Trap this case to avoid exhausting the [very limited]
6100 	 * set of PF resources used to store VF MAC addresses.
6101 	 */
6102 	if (memcmp(hw->mac.perm_addr, mac_addr,
6103 			sizeof(struct rte_ether_addr)) == 0)
6104 		return -1;
6105 	diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6106 	if (diag != 0)
6107 		PMD_DRV_LOG(ERR, "Unable to add MAC address "
6108 			    RTE_ETHER_ADDR_PRT_FMT " - diag=%d",
6109 			    RTE_ETHER_ADDR_BYTES(mac_addr), diag);
6110 	return diag;
6111 }
6112 
6113 static void
6114 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
6115 {
6116 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6117 	struct rte_ether_addr *perm_addr =
6118 		(struct rte_ether_addr *)hw->mac.perm_addr;
6119 	struct rte_ether_addr *mac_addr;
6120 	uint32_t i;
6121 	int diag;
6122 
6123 	/*
6124 	 * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
6125 	 * not support the deletion of a given MAC address.
6126 	 * Instead, it imposes to delete all MAC addresses, then to add again
6127 	 * all MAC addresses with the exception of the one to be deleted.
6128 	 */
6129 	(void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
6130 
6131 	/*
6132 	 * Add again all MAC addresses, with the exception of the deleted one
6133 	 * and of the permanent MAC address.
6134 	 */
6135 	for (i = 0, mac_addr = dev->data->mac_addrs;
6136 	     i < hw->mac.num_rar_entries; i++, mac_addr++) {
6137 		/* Skip the deleted MAC address */
6138 		if (i == index)
6139 			continue;
6140 		/* Skip NULL MAC addresses */
6141 		if (rte_is_zero_ether_addr(mac_addr))
6142 			continue;
6143 		/* Skip the permanent MAC address */
6144 		if (memcmp(perm_addr, mac_addr,
6145 				sizeof(struct rte_ether_addr)) == 0)
6146 			continue;
6147 		diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6148 		if (diag != 0)
6149 			PMD_DRV_LOG(ERR,
6150 				    "Adding again MAC address "
6151 				     RTE_ETHER_ADDR_PRT_FMT " failed "
6152 				    "diag=%d", RTE_ETHER_ADDR_BYTES(mac_addr),
6153 				    diag);
6154 	}
6155 }
6156 
6157 static int
6158 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
6159 			struct rte_ether_addr *addr)
6160 {
6161 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6162 
6163 	hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
6164 
6165 	return 0;
6166 }
6167 
6168 int
6169 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
6170 			struct rte_eth_syn_filter *filter,
6171 			bool add)
6172 {
6173 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6174 	struct ixgbe_filter_info *filter_info =
6175 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6176 	uint32_t syn_info;
6177 	uint32_t synqf;
6178 
6179 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6180 		return -EINVAL;
6181 
6182 	syn_info = filter_info->syn_info;
6183 
6184 	if (add) {
6185 		if (syn_info & IXGBE_SYN_FILTER_ENABLE)
6186 			return -EINVAL;
6187 		synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
6188 			IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
6189 
6190 		if (filter->hig_pri)
6191 			synqf |= IXGBE_SYN_FILTER_SYNQFP;
6192 		else
6193 			synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
6194 	} else {
6195 		synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6196 		if (!(syn_info & IXGBE_SYN_FILTER_ENABLE))
6197 			return -ENOENT;
6198 		synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
6199 	}
6200 
6201 	filter_info->syn_info = synqf;
6202 	IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
6203 	IXGBE_WRITE_FLUSH(hw);
6204 	return 0;
6205 }
6206 
6207 
6208 static inline enum ixgbe_5tuple_protocol
6209 convert_protocol_type(uint8_t protocol_value)
6210 {
6211 	if (protocol_value == IPPROTO_TCP)
6212 		return IXGBE_FILTER_PROTOCOL_TCP;
6213 	else if (protocol_value == IPPROTO_UDP)
6214 		return IXGBE_FILTER_PROTOCOL_UDP;
6215 	else if (protocol_value == IPPROTO_SCTP)
6216 		return IXGBE_FILTER_PROTOCOL_SCTP;
6217 	else
6218 		return IXGBE_FILTER_PROTOCOL_NONE;
6219 }
6220 
6221 /* inject a 5-tuple filter to HW */
6222 static inline void
6223 ixgbe_inject_5tuple_filter(struct rte_eth_dev *dev,
6224 			   struct ixgbe_5tuple_filter *filter)
6225 {
6226 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6227 	int i;
6228 	uint32_t ftqf, sdpqf;
6229 	uint32_t l34timir = 0;
6230 	uint8_t mask = 0xff;
6231 
6232 	i = filter->index;
6233 
6234 	sdpqf = (uint32_t)(filter->filter_info.dst_port <<
6235 				IXGBE_SDPQF_DSTPORT_SHIFT);
6236 	sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
6237 
6238 	ftqf = (uint32_t)(filter->filter_info.proto &
6239 		IXGBE_FTQF_PROTOCOL_MASK);
6240 	ftqf |= (uint32_t)((filter->filter_info.priority &
6241 		IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
6242 	if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
6243 		mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
6244 	if (filter->filter_info.dst_ip_mask == 0)
6245 		mask &= IXGBE_FTQF_DEST_ADDR_MASK;
6246 	if (filter->filter_info.src_port_mask == 0)
6247 		mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
6248 	if (filter->filter_info.dst_port_mask == 0)
6249 		mask &= IXGBE_FTQF_DEST_PORT_MASK;
6250 	if (filter->filter_info.proto_mask == 0)
6251 		mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
6252 	ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
6253 	ftqf |= IXGBE_FTQF_POOL_MASK_EN;
6254 	ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
6255 
6256 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
6257 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
6258 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
6259 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
6260 
6261 	l34timir |= IXGBE_L34T_IMIR_RESERVE;
6262 	l34timir |= (uint32_t)(filter->queue <<
6263 				IXGBE_L34T_IMIR_QUEUE_SHIFT);
6264 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
6265 }
6266 
6267 /*
6268  * add a 5tuple filter
6269  *
6270  * @param
6271  * dev: Pointer to struct rte_eth_dev.
6272  * index: the index the filter allocates.
6273  * filter: pointer to the filter that will be added.
6274  * rx_queue: the queue id the filter assigned to.
6275  *
6276  * @return
6277  *    - On success, zero.
6278  *    - On failure, a negative value.
6279  */
6280 static int
6281 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
6282 			struct ixgbe_5tuple_filter *filter)
6283 {
6284 	struct ixgbe_filter_info *filter_info =
6285 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6286 	int i, idx, shift;
6287 
6288 	/*
6289 	 * look for an unused 5tuple filter index,
6290 	 * and insert the filter to list.
6291 	 */
6292 	for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
6293 		idx = i / (sizeof(uint32_t) * NBBY);
6294 		shift = i % (sizeof(uint32_t) * NBBY);
6295 		if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
6296 			filter_info->fivetuple_mask[idx] |= 1 << shift;
6297 			filter->index = i;
6298 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
6299 					  filter,
6300 					  entries);
6301 			break;
6302 		}
6303 	}
6304 	if (i >= IXGBE_MAX_FTQF_FILTERS) {
6305 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
6306 		return -ENOSYS;
6307 	}
6308 
6309 	ixgbe_inject_5tuple_filter(dev, filter);
6310 
6311 	return 0;
6312 }
6313 
6314 /*
6315  * remove a 5tuple filter
6316  *
6317  * @param
6318  * dev: Pointer to struct rte_eth_dev.
6319  * filter: the pointer of the filter will be removed.
6320  */
6321 static void
6322 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
6323 			struct ixgbe_5tuple_filter *filter)
6324 {
6325 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6326 	struct ixgbe_filter_info *filter_info =
6327 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6328 	uint16_t index = filter->index;
6329 
6330 	filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
6331 				~(1 << (index % (sizeof(uint32_t) * NBBY)));
6332 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
6333 	rte_free(filter);
6334 
6335 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
6336 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
6337 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
6338 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
6339 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
6340 }
6341 
6342 static int
6343 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
6344 {
6345 	struct ixgbe_hw *hw;
6346 	uint32_t max_frame = mtu + IXGBE_ETH_OVERHEAD;
6347 	struct rte_eth_dev_data *dev_data = dev->data;
6348 
6349 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6350 
6351 	if (mtu < RTE_ETHER_MIN_MTU || max_frame > RTE_ETHER_MAX_JUMBO_FRAME_LEN)
6352 		return -EINVAL;
6353 
6354 	/* If device is started, refuse mtu that requires the support of
6355 	 * scattered packets when this feature has not been enabled before.
6356 	 */
6357 	if (dev_data->dev_started && !dev_data->scattered_rx &&
6358 	    (max_frame + 2 * RTE_VLAN_HLEN >
6359 			dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
6360 		PMD_INIT_LOG(ERR, "Stop port first.");
6361 		return -EINVAL;
6362 	}
6363 
6364 	/*
6365 	 * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
6366 	 * request of the version 2.0 of the mailbox API.
6367 	 * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
6368 	 * of the mailbox API.
6369 	 * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
6370 	 * prior to 3.11.33 which contains the following change:
6371 	 * "ixgbe: Enable jumbo frames support w/ SR-IOV"
6372 	 */
6373 	if (ixgbevf_rlpml_set_vf(hw, max_frame))
6374 		return -EINVAL;
6375 
6376 	return 0;
6377 }
6378 
6379 static inline struct ixgbe_5tuple_filter *
6380 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
6381 			struct ixgbe_5tuple_filter_info *key)
6382 {
6383 	struct ixgbe_5tuple_filter *it;
6384 
6385 	TAILQ_FOREACH(it, filter_list, entries) {
6386 		if (memcmp(key, &it->filter_info,
6387 			sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
6388 			return it;
6389 		}
6390 	}
6391 	return NULL;
6392 }
6393 
6394 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
6395 static inline int
6396 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
6397 			struct ixgbe_5tuple_filter_info *filter_info)
6398 {
6399 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
6400 		filter->priority > IXGBE_5TUPLE_MAX_PRI ||
6401 		filter->priority < IXGBE_5TUPLE_MIN_PRI)
6402 		return -EINVAL;
6403 
6404 	switch (filter->dst_ip_mask) {
6405 	case UINT32_MAX:
6406 		filter_info->dst_ip_mask = 0;
6407 		filter_info->dst_ip = filter->dst_ip;
6408 		break;
6409 	case 0:
6410 		filter_info->dst_ip_mask = 1;
6411 		break;
6412 	default:
6413 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
6414 		return -EINVAL;
6415 	}
6416 
6417 	switch (filter->src_ip_mask) {
6418 	case UINT32_MAX:
6419 		filter_info->src_ip_mask = 0;
6420 		filter_info->src_ip = filter->src_ip;
6421 		break;
6422 	case 0:
6423 		filter_info->src_ip_mask = 1;
6424 		break;
6425 	default:
6426 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
6427 		return -EINVAL;
6428 	}
6429 
6430 	switch (filter->dst_port_mask) {
6431 	case UINT16_MAX:
6432 		filter_info->dst_port_mask = 0;
6433 		filter_info->dst_port = filter->dst_port;
6434 		break;
6435 	case 0:
6436 		filter_info->dst_port_mask = 1;
6437 		break;
6438 	default:
6439 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
6440 		return -EINVAL;
6441 	}
6442 
6443 	switch (filter->src_port_mask) {
6444 	case UINT16_MAX:
6445 		filter_info->src_port_mask = 0;
6446 		filter_info->src_port = filter->src_port;
6447 		break;
6448 	case 0:
6449 		filter_info->src_port_mask = 1;
6450 		break;
6451 	default:
6452 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
6453 		return -EINVAL;
6454 	}
6455 
6456 	switch (filter->proto_mask) {
6457 	case UINT8_MAX:
6458 		filter_info->proto_mask = 0;
6459 		filter_info->proto =
6460 			convert_protocol_type(filter->proto);
6461 		break;
6462 	case 0:
6463 		filter_info->proto_mask = 1;
6464 		break;
6465 	default:
6466 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
6467 		return -EINVAL;
6468 	}
6469 
6470 	filter_info->priority = (uint8_t)filter->priority;
6471 	return 0;
6472 }
6473 
6474 /*
6475  * add or delete a ntuple filter
6476  *
6477  * @param
6478  * dev: Pointer to struct rte_eth_dev.
6479  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6480  * add: if true, add filter, if false, remove filter
6481  *
6482  * @return
6483  *    - On success, zero.
6484  *    - On failure, a negative value.
6485  */
6486 int
6487 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
6488 			struct rte_eth_ntuple_filter *ntuple_filter,
6489 			bool add)
6490 {
6491 	struct ixgbe_filter_info *filter_info =
6492 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6493 	struct ixgbe_5tuple_filter_info filter_5tuple;
6494 	struct ixgbe_5tuple_filter *filter;
6495 	int ret;
6496 
6497 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6498 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6499 		return -EINVAL;
6500 	}
6501 
6502 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6503 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6504 	if (ret < 0)
6505 		return ret;
6506 
6507 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6508 					 &filter_5tuple);
6509 	if (filter != NULL && add) {
6510 		PMD_DRV_LOG(ERR, "filter exists.");
6511 		return -EEXIST;
6512 	}
6513 	if (filter == NULL && !add) {
6514 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
6515 		return -ENOENT;
6516 	}
6517 
6518 	if (add) {
6519 		filter = rte_zmalloc("ixgbe_5tuple_filter",
6520 				sizeof(struct ixgbe_5tuple_filter), 0);
6521 		if (filter == NULL)
6522 			return -ENOMEM;
6523 		rte_memcpy(&filter->filter_info,
6524 				 &filter_5tuple,
6525 				 sizeof(struct ixgbe_5tuple_filter_info));
6526 		filter->queue = ntuple_filter->queue;
6527 		ret = ixgbe_add_5tuple_filter(dev, filter);
6528 		if (ret < 0) {
6529 			rte_free(filter);
6530 			return ret;
6531 		}
6532 	} else
6533 		ixgbe_remove_5tuple_filter(dev, filter);
6534 
6535 	return 0;
6536 }
6537 
6538 int
6539 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
6540 			struct rte_eth_ethertype_filter *filter,
6541 			bool add)
6542 {
6543 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6544 	struct ixgbe_filter_info *filter_info =
6545 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6546 	uint32_t etqf = 0;
6547 	uint32_t etqs = 0;
6548 	int ret;
6549 	struct ixgbe_ethertype_filter ethertype_filter;
6550 
6551 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6552 		return -EINVAL;
6553 
6554 	if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
6555 		filter->ether_type == RTE_ETHER_TYPE_IPV6) {
6556 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
6557 			" ethertype filter.", filter->ether_type);
6558 		return -EINVAL;
6559 	}
6560 
6561 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
6562 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
6563 		return -EINVAL;
6564 	}
6565 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
6566 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
6567 		return -EINVAL;
6568 	}
6569 
6570 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6571 	if (ret >= 0 && add) {
6572 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
6573 			    filter->ether_type);
6574 		return -EEXIST;
6575 	}
6576 	if (ret < 0 && !add) {
6577 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6578 			    filter->ether_type);
6579 		return -ENOENT;
6580 	}
6581 
6582 	if (add) {
6583 		etqf = IXGBE_ETQF_FILTER_EN;
6584 		etqf |= (uint32_t)filter->ether_type;
6585 		etqs |= (uint32_t)((filter->queue <<
6586 				    IXGBE_ETQS_RX_QUEUE_SHIFT) &
6587 				    IXGBE_ETQS_RX_QUEUE);
6588 		etqs |= IXGBE_ETQS_QUEUE_EN;
6589 
6590 		ethertype_filter.ethertype = filter->ether_type;
6591 		ethertype_filter.etqf = etqf;
6592 		ethertype_filter.etqs = etqs;
6593 		ethertype_filter.conf = FALSE;
6594 		ret = ixgbe_ethertype_filter_insert(filter_info,
6595 						    &ethertype_filter);
6596 		if (ret < 0) {
6597 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
6598 			return -ENOSPC;
6599 		}
6600 	} else {
6601 		ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
6602 		if (ret < 0)
6603 			return -ENOSYS;
6604 	}
6605 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
6606 	IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
6607 	IXGBE_WRITE_FLUSH(hw);
6608 
6609 	return 0;
6610 }
6611 
6612 static int
6613 ixgbe_dev_flow_ops_get(__rte_unused struct rte_eth_dev *dev,
6614 		       const struct rte_flow_ops **ops)
6615 {
6616 	*ops = &ixgbe_flow_ops;
6617 	return 0;
6618 }
6619 
6620 static u8 *
6621 ixgbe_dev_addr_list_itr(__rte_unused struct ixgbe_hw *hw,
6622 			u8 **mc_addr_ptr, u32 *vmdq)
6623 {
6624 	u8 *mc_addr;
6625 
6626 	*vmdq = 0;
6627 	mc_addr = *mc_addr_ptr;
6628 	*mc_addr_ptr = (mc_addr + sizeof(struct rte_ether_addr));
6629 	return mc_addr;
6630 }
6631 
6632 static int
6633 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
6634 			  struct rte_ether_addr *mc_addr_set,
6635 			  uint32_t nb_mc_addr)
6636 {
6637 	struct ixgbe_hw *hw;
6638 	u8 *mc_addr_list;
6639 
6640 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6641 	mc_addr_list = (u8 *)mc_addr_set;
6642 	return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
6643 					 ixgbe_dev_addr_list_itr, TRUE);
6644 }
6645 
6646 static uint64_t
6647 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
6648 {
6649 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6650 	uint64_t systime_cycles;
6651 
6652 	switch (hw->mac.type) {
6653 	case ixgbe_mac_X550:
6654 	case ixgbe_mac_X550EM_x:
6655 	case ixgbe_mac_X550EM_a:
6656 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
6657 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6658 		systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6659 				* NSEC_PER_SEC;
6660 		break;
6661 	default:
6662 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6663 		systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6664 				<< 32;
6665 	}
6666 
6667 	return systime_cycles;
6668 }
6669 
6670 static uint64_t
6671 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6672 {
6673 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6674 	uint64_t rx_tstamp_cycles;
6675 
6676 	switch (hw->mac.type) {
6677 	case ixgbe_mac_X550:
6678 	case ixgbe_mac_X550EM_x:
6679 	case ixgbe_mac_X550EM_a:
6680 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6681 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6682 		rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6683 				* NSEC_PER_SEC;
6684 		break;
6685 	default:
6686 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6687 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6688 		rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6689 				<< 32;
6690 	}
6691 
6692 	return rx_tstamp_cycles;
6693 }
6694 
6695 static uint64_t
6696 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6697 {
6698 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6699 	uint64_t tx_tstamp_cycles;
6700 
6701 	switch (hw->mac.type) {
6702 	case ixgbe_mac_X550:
6703 	case ixgbe_mac_X550EM_x:
6704 	case ixgbe_mac_X550EM_a:
6705 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6706 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6707 		tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6708 				* NSEC_PER_SEC;
6709 		break;
6710 	default:
6711 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6712 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6713 		tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6714 				<< 32;
6715 	}
6716 
6717 	return tx_tstamp_cycles;
6718 }
6719 
6720 static void
6721 ixgbe_start_timecounters(struct rte_eth_dev *dev)
6722 {
6723 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6724 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6725 	struct rte_eth_link link;
6726 	uint32_t incval = 0;
6727 	uint32_t shift = 0;
6728 
6729 	/* Get current link speed. */
6730 	ixgbe_dev_link_update(dev, 1);
6731 	rte_eth_linkstatus_get(dev, &link);
6732 
6733 	switch (link.link_speed) {
6734 	case RTE_ETH_SPEED_NUM_100M:
6735 		incval = IXGBE_INCVAL_100;
6736 		shift = IXGBE_INCVAL_SHIFT_100;
6737 		break;
6738 	case RTE_ETH_SPEED_NUM_1G:
6739 		incval = IXGBE_INCVAL_1GB;
6740 		shift = IXGBE_INCVAL_SHIFT_1GB;
6741 		break;
6742 	case RTE_ETH_SPEED_NUM_10G:
6743 	default:
6744 		incval = IXGBE_INCVAL_10GB;
6745 		shift = IXGBE_INCVAL_SHIFT_10GB;
6746 		break;
6747 	}
6748 
6749 	switch (hw->mac.type) {
6750 	case ixgbe_mac_X550:
6751 	case ixgbe_mac_X550EM_x:
6752 	case ixgbe_mac_X550EM_a:
6753 		/* Independent of link speed. */
6754 		incval = 1;
6755 		/* Cycles read will be interpreted as ns. */
6756 		shift = 0;
6757 		/* Fall-through */
6758 	case ixgbe_mac_X540:
6759 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
6760 		break;
6761 	case ixgbe_mac_82599EB:
6762 		incval >>= IXGBE_INCVAL_SHIFT_82599;
6763 		shift -= IXGBE_INCVAL_SHIFT_82599;
6764 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
6765 				(1 << IXGBE_INCPER_SHIFT_82599) | incval);
6766 		break;
6767 	default:
6768 		/* Not supported. */
6769 		return;
6770 	}
6771 
6772 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
6773 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6774 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6775 
6776 	adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6777 	adapter->systime_tc.cc_shift = shift;
6778 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
6779 
6780 	adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6781 	adapter->rx_tstamp_tc.cc_shift = shift;
6782 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6783 
6784 	adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6785 	adapter->tx_tstamp_tc.cc_shift = shift;
6786 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6787 }
6788 
6789 static int
6790 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
6791 {
6792 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6793 
6794 	adapter->systime_tc.nsec += delta;
6795 	adapter->rx_tstamp_tc.nsec += delta;
6796 	adapter->tx_tstamp_tc.nsec += delta;
6797 
6798 	return 0;
6799 }
6800 
6801 static int
6802 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
6803 {
6804 	uint64_t ns;
6805 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6806 
6807 	ns = rte_timespec_to_ns(ts);
6808 	/* Set the timecounters to a new value. */
6809 	adapter->systime_tc.nsec = ns;
6810 	adapter->rx_tstamp_tc.nsec = ns;
6811 	adapter->tx_tstamp_tc.nsec = ns;
6812 
6813 	return 0;
6814 }
6815 
6816 static int
6817 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
6818 {
6819 	uint64_t ns, systime_cycles;
6820 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6821 
6822 	systime_cycles = ixgbe_read_systime_cyclecounter(dev);
6823 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
6824 	*ts = rte_ns_to_timespec(ns);
6825 
6826 	return 0;
6827 }
6828 
6829 static int
6830 ixgbe_timesync_enable(struct rte_eth_dev *dev)
6831 {
6832 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6833 	uint32_t tsync_ctl;
6834 	uint32_t tsauxc;
6835 
6836 	/* Stop the timesync system time. */
6837 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
6838 	/* Reset the timesync system time value. */
6839 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
6840 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
6841 
6842 	/* Enable system time for platforms where it isn't on by default. */
6843 	tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
6844 	tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
6845 	IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
6846 
6847 	ixgbe_start_timecounters(dev);
6848 
6849 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6850 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
6851 			(RTE_ETHER_TYPE_1588 |
6852 			 IXGBE_ETQF_FILTER_EN |
6853 			 IXGBE_ETQF_1588));
6854 
6855 	/* Enable timestamping of received PTP packets. */
6856 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6857 	tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
6858 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6859 
6860 	/* Enable timestamping of transmitted PTP packets. */
6861 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6862 	tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
6863 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6864 
6865 	IXGBE_WRITE_FLUSH(hw);
6866 
6867 	return 0;
6868 }
6869 
6870 static int
6871 ixgbe_timesync_disable(struct rte_eth_dev *dev)
6872 {
6873 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6874 	uint32_t tsync_ctl;
6875 
6876 	/* Disable timestamping of transmitted PTP packets. */
6877 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6878 	tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
6879 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6880 
6881 	/* Disable timestamping of received PTP packets. */
6882 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6883 	tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
6884 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6885 
6886 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6887 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
6888 
6889 	/* Stop incrementing the System Time registers. */
6890 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
6891 
6892 	return 0;
6893 }
6894 
6895 static int
6896 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
6897 				 struct timespec *timestamp,
6898 				 uint32_t flags __rte_unused)
6899 {
6900 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6901 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6902 	uint32_t tsync_rxctl;
6903 	uint64_t rx_tstamp_cycles;
6904 	uint64_t ns;
6905 
6906 	tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6907 	if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
6908 		return -EINVAL;
6909 
6910 	rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
6911 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
6912 	*timestamp = rte_ns_to_timespec(ns);
6913 
6914 	return  0;
6915 }
6916 
6917 static int
6918 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
6919 				 struct timespec *timestamp)
6920 {
6921 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6922 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6923 	uint32_t tsync_txctl;
6924 	uint64_t tx_tstamp_cycles;
6925 	uint64_t ns;
6926 
6927 	tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6928 	if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
6929 		return -EINVAL;
6930 
6931 	tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
6932 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
6933 	*timestamp = rte_ns_to_timespec(ns);
6934 
6935 	return 0;
6936 }
6937 
6938 static int
6939 ixgbe_get_reg_length(struct rte_eth_dev *dev)
6940 {
6941 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6942 	int count = 0;
6943 	int g_ind = 0;
6944 	const struct reg_info *reg_group;
6945 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6946 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6947 
6948 	while ((reg_group = reg_set[g_ind++]))
6949 		count += ixgbe_regs_group_count(reg_group);
6950 
6951 	return count;
6952 }
6953 
6954 static int
6955 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
6956 {
6957 	int count = 0;
6958 	int g_ind = 0;
6959 	const struct reg_info *reg_group;
6960 
6961 	while ((reg_group = ixgbevf_regs[g_ind++]))
6962 		count += ixgbe_regs_group_count(reg_group);
6963 
6964 	return count;
6965 }
6966 
6967 static int
6968 ixgbe_get_regs(struct rte_eth_dev *dev,
6969 	      struct rte_dev_reg_info *regs)
6970 {
6971 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6972 	uint32_t *data = regs->data;
6973 	int g_ind = 0;
6974 	int count = 0;
6975 	const struct reg_info *reg_group;
6976 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6977 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6978 
6979 	if (data == NULL) {
6980 		regs->length = ixgbe_get_reg_length(dev);
6981 		regs->width = sizeof(uint32_t);
6982 		return 0;
6983 	}
6984 
6985 	/* Support only full register dump */
6986 	if ((regs->length == 0) ||
6987 	    (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
6988 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
6989 			hw->device_id;
6990 		while ((reg_group = reg_set[g_ind++]))
6991 			count += ixgbe_read_regs_group(dev, &data[count],
6992 				reg_group);
6993 		return 0;
6994 	}
6995 
6996 	return -ENOTSUP;
6997 }
6998 
6999 static int
7000 ixgbevf_get_regs(struct rte_eth_dev *dev,
7001 		struct rte_dev_reg_info *regs)
7002 {
7003 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7004 	uint32_t *data = regs->data;
7005 	int g_ind = 0;
7006 	int count = 0;
7007 	const struct reg_info *reg_group;
7008 
7009 	if (data == NULL) {
7010 		regs->length = ixgbevf_get_reg_length(dev);
7011 		regs->width = sizeof(uint32_t);
7012 		return 0;
7013 	}
7014 
7015 	/* Support only full register dump */
7016 	if ((regs->length == 0) ||
7017 	    (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
7018 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7019 			hw->device_id;
7020 		while ((reg_group = ixgbevf_regs[g_ind++]))
7021 			count += ixgbe_read_regs_group(dev, &data[count],
7022 						      reg_group);
7023 		return 0;
7024 	}
7025 
7026 	return -ENOTSUP;
7027 }
7028 
7029 static int
7030 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
7031 {
7032 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7033 
7034 	/* Return unit is byte count */
7035 	return hw->eeprom.word_size * 2;
7036 }
7037 
7038 static int
7039 ixgbe_get_eeprom(struct rte_eth_dev *dev,
7040 		struct rte_dev_eeprom_info *in_eeprom)
7041 {
7042 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7043 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7044 	uint16_t *data = in_eeprom->data;
7045 	int first, length;
7046 
7047 	first = in_eeprom->offset >> 1;
7048 	length = in_eeprom->length >> 1;
7049 	if ((first > hw->eeprom.word_size) ||
7050 	    ((first + length) > hw->eeprom.word_size))
7051 		return -EINVAL;
7052 
7053 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7054 
7055 	return eeprom->ops.read_buffer(hw, first, length, data);
7056 }
7057 
7058 static int
7059 ixgbe_set_eeprom(struct rte_eth_dev *dev,
7060 		struct rte_dev_eeprom_info *in_eeprom)
7061 {
7062 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7063 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7064 	uint16_t *data = in_eeprom->data;
7065 	int first, length;
7066 
7067 	first = in_eeprom->offset >> 1;
7068 	length = in_eeprom->length >> 1;
7069 	if ((first > hw->eeprom.word_size) ||
7070 	    ((first + length) > hw->eeprom.word_size))
7071 		return -EINVAL;
7072 
7073 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7074 
7075 	return eeprom->ops.write_buffer(hw,  first, length, data);
7076 }
7077 
7078 static int
7079 ixgbe_get_module_info(struct rte_eth_dev *dev,
7080 		      struct rte_eth_dev_module_info *modinfo)
7081 {
7082 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7083 	uint32_t status;
7084 	uint8_t sff8472_rev, addr_mode;
7085 	bool page_swap = false;
7086 
7087 	/* Check whether we support SFF-8472 or not */
7088 	status = hw->phy.ops.read_i2c_eeprom(hw,
7089 					     IXGBE_SFF_SFF_8472_COMP,
7090 					     &sff8472_rev);
7091 	if (status != 0)
7092 		return -EIO;
7093 
7094 	/* addressing mode is not supported */
7095 	status = hw->phy.ops.read_i2c_eeprom(hw,
7096 					     IXGBE_SFF_SFF_8472_SWAP,
7097 					     &addr_mode);
7098 	if (status != 0)
7099 		return -EIO;
7100 
7101 	if (addr_mode & IXGBE_SFF_ADDRESSING_MODE) {
7102 		PMD_DRV_LOG(ERR,
7103 			    "Address change required to access page 0xA2, "
7104 			    "but not supported. Please report the module "
7105 			    "type to the driver maintainers.");
7106 		page_swap = true;
7107 	}
7108 
7109 	if (sff8472_rev == IXGBE_SFF_SFF_8472_UNSUP || page_swap) {
7110 		/* We have a SFP, but it does not support SFF-8472 */
7111 		modinfo->type = RTE_ETH_MODULE_SFF_8079;
7112 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
7113 	} else {
7114 		/* We have a SFP which supports a revision of SFF-8472. */
7115 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
7116 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
7117 	}
7118 
7119 	return 0;
7120 }
7121 
7122 static int
7123 ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
7124 			struct rte_dev_eeprom_info *info)
7125 {
7126 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7127 	uint32_t status = IXGBE_ERR_PHY_ADDR_INVALID;
7128 	uint8_t databyte = 0xFF;
7129 	uint8_t *data = info->data;
7130 	uint32_t i = 0;
7131 
7132 	for (i = info->offset; i < info->offset + info->length; i++) {
7133 		if (i < RTE_ETH_MODULE_SFF_8079_LEN)
7134 			status = hw->phy.ops.read_i2c_eeprom(hw, i, &databyte);
7135 		else
7136 			status = hw->phy.ops.read_i2c_sff8472(hw, i, &databyte);
7137 
7138 		if (status != 0)
7139 			return -EIO;
7140 
7141 		data[i - info->offset] = databyte;
7142 	}
7143 
7144 	return 0;
7145 }
7146 
7147 uint16_t
7148 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
7149 	switch (mac_type) {
7150 	case ixgbe_mac_X550:
7151 	case ixgbe_mac_X550EM_x:
7152 	case ixgbe_mac_X550EM_a:
7153 		return RTE_ETH_RSS_RETA_SIZE_512;
7154 	case ixgbe_mac_X550_vf:
7155 	case ixgbe_mac_X550EM_x_vf:
7156 	case ixgbe_mac_X550EM_a_vf:
7157 		return RTE_ETH_RSS_RETA_SIZE_64;
7158 	case ixgbe_mac_X540_vf:
7159 	case ixgbe_mac_82599_vf:
7160 		return 0;
7161 	default:
7162 		return RTE_ETH_RSS_RETA_SIZE_128;
7163 	}
7164 }
7165 
7166 uint32_t
7167 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
7168 	switch (mac_type) {
7169 	case ixgbe_mac_X550:
7170 	case ixgbe_mac_X550EM_x:
7171 	case ixgbe_mac_X550EM_a:
7172 		if (reta_idx < RTE_ETH_RSS_RETA_SIZE_128)
7173 			return IXGBE_RETA(reta_idx >> 2);
7174 		else
7175 			return IXGBE_ERETA((reta_idx - RTE_ETH_RSS_RETA_SIZE_128) >> 2);
7176 	case ixgbe_mac_X550_vf:
7177 	case ixgbe_mac_X550EM_x_vf:
7178 	case ixgbe_mac_X550EM_a_vf:
7179 		return IXGBE_VFRETA(reta_idx >> 2);
7180 	default:
7181 		return IXGBE_RETA(reta_idx >> 2);
7182 	}
7183 }
7184 
7185 uint32_t
7186 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
7187 	switch (mac_type) {
7188 	case ixgbe_mac_X550_vf:
7189 	case ixgbe_mac_X550EM_x_vf:
7190 	case ixgbe_mac_X550EM_a_vf:
7191 		return IXGBE_VFMRQC;
7192 	default:
7193 		return IXGBE_MRQC;
7194 	}
7195 }
7196 
7197 uint32_t
7198 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
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_VFRSSRK(i);
7204 	default:
7205 		return IXGBE_RSSRK(i);
7206 	}
7207 }
7208 
7209 bool
7210 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
7211 	switch (mac_type) {
7212 	case ixgbe_mac_82599_vf:
7213 	case ixgbe_mac_X540_vf:
7214 		return 0;
7215 	default:
7216 		return 1;
7217 	}
7218 }
7219 
7220 static int
7221 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
7222 			struct rte_eth_dcb_info *dcb_info)
7223 {
7224 	struct ixgbe_dcb_config *dcb_config =
7225 			IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
7226 	struct ixgbe_dcb_tc_config *tc;
7227 	struct rte_eth_dcb_tc_queue_mapping *tc_queue;
7228 	uint8_t nb_tcs;
7229 	uint8_t i, j;
7230 
7231 	if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_DCB_FLAG)
7232 		dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
7233 	else
7234 		dcb_info->nb_tcs = 1;
7235 
7236 	tc_queue = &dcb_info->tc_queue;
7237 	nb_tcs = dcb_info->nb_tcs;
7238 
7239 	if (dcb_config->vt_mode) { /* vt is enabled*/
7240 		struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
7241 				&dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
7242 		for (i = 0; i < RTE_ETH_DCB_NUM_USER_PRIORITIES; i++)
7243 			dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
7244 		if (RTE_ETH_DEV_SRIOV(dev).active > 0) {
7245 			for (j = 0; j < nb_tcs; j++) {
7246 				tc_queue->tc_rxq[0][j].base = j;
7247 				tc_queue->tc_rxq[0][j].nb_queue = 1;
7248 				tc_queue->tc_txq[0][j].base = j;
7249 				tc_queue->tc_txq[0][j].nb_queue = 1;
7250 			}
7251 		} else {
7252 			for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
7253 				for (j = 0; j < nb_tcs; j++) {
7254 					tc_queue->tc_rxq[i][j].base =
7255 						i * nb_tcs + j;
7256 					tc_queue->tc_rxq[i][j].nb_queue = 1;
7257 					tc_queue->tc_txq[i][j].base =
7258 						i * nb_tcs + j;
7259 					tc_queue->tc_txq[i][j].nb_queue = 1;
7260 				}
7261 			}
7262 		}
7263 	} else { /* vt is disabled*/
7264 		struct rte_eth_dcb_rx_conf *rx_conf =
7265 				&dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
7266 		for (i = 0; i < RTE_ETH_DCB_NUM_USER_PRIORITIES; i++)
7267 			dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
7268 		if (dcb_info->nb_tcs == RTE_ETH_4_TCS) {
7269 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7270 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
7271 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7272 			}
7273 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7274 			dcb_info->tc_queue.tc_txq[0][1].base = 64;
7275 			dcb_info->tc_queue.tc_txq[0][2].base = 96;
7276 			dcb_info->tc_queue.tc_txq[0][3].base = 112;
7277 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
7278 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7279 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7280 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7281 		} else if (dcb_info->nb_tcs == RTE_ETH_8_TCS) {
7282 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7283 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
7284 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7285 			}
7286 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7287 			dcb_info->tc_queue.tc_txq[0][1].base = 32;
7288 			dcb_info->tc_queue.tc_txq[0][2].base = 64;
7289 			dcb_info->tc_queue.tc_txq[0][3].base = 80;
7290 			dcb_info->tc_queue.tc_txq[0][4].base = 96;
7291 			dcb_info->tc_queue.tc_txq[0][5].base = 104;
7292 			dcb_info->tc_queue.tc_txq[0][6].base = 112;
7293 			dcb_info->tc_queue.tc_txq[0][7].base = 120;
7294 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
7295 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7296 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7297 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7298 			dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
7299 			dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
7300 			dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
7301 			dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
7302 		}
7303 	}
7304 	for (i = 0; i < dcb_info->nb_tcs; i++) {
7305 		tc = &dcb_config->tc_config[i];
7306 		dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
7307 	}
7308 	return 0;
7309 }
7310 
7311 /* Update e-tag ether type */
7312 static int
7313 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
7314 			    uint16_t ether_type)
7315 {
7316 	uint32_t etag_etype;
7317 
7318 	if (hw->mac.type != ixgbe_mac_X550 &&
7319 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7320 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7321 		return -ENOTSUP;
7322 	}
7323 
7324 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7325 	etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
7326 	etag_etype |= ether_type;
7327 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7328 	IXGBE_WRITE_FLUSH(hw);
7329 
7330 	return 0;
7331 }
7332 
7333 /* Enable e-tag tunnel */
7334 static int
7335 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
7336 {
7337 	uint32_t etag_etype;
7338 
7339 	if (hw->mac.type != ixgbe_mac_X550 &&
7340 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7341 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7342 		return -ENOTSUP;
7343 	}
7344 
7345 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7346 	etag_etype |= IXGBE_ETAG_ETYPE_VALID;
7347 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7348 	IXGBE_WRITE_FLUSH(hw);
7349 
7350 	return 0;
7351 }
7352 
7353 static int
7354 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
7355 		       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7356 {
7357 	int ret = 0;
7358 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7359 	uint32_t i, rar_entries;
7360 	uint32_t rar_low, rar_high;
7361 
7362 	if (hw->mac.type != ixgbe_mac_X550 &&
7363 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7364 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7365 		return -ENOTSUP;
7366 	}
7367 
7368 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7369 
7370 	for (i = 1; i < rar_entries; i++) {
7371 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7372 		rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
7373 		if ((rar_high & IXGBE_RAH_AV) &&
7374 		    (rar_high & IXGBE_RAH_ADTYPE) &&
7375 		    ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
7376 		     l2_tunnel->tunnel_id)) {
7377 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
7378 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
7379 
7380 			ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
7381 
7382 			return ret;
7383 		}
7384 	}
7385 
7386 	return ret;
7387 }
7388 
7389 static int
7390 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
7391 		       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7392 {
7393 	int ret = 0;
7394 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7395 	uint32_t i, rar_entries;
7396 	uint32_t rar_low, rar_high;
7397 
7398 	if (hw->mac.type != ixgbe_mac_X550 &&
7399 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7400 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7401 		return -ENOTSUP;
7402 	}
7403 
7404 	/* One entry for one tunnel. Try to remove potential existing entry. */
7405 	ixgbe_e_tag_filter_del(dev, l2_tunnel);
7406 
7407 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7408 
7409 	for (i = 1; i < rar_entries; i++) {
7410 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7411 		if (rar_high & IXGBE_RAH_AV) {
7412 			continue;
7413 		} else {
7414 			ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
7415 			rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
7416 			rar_low = l2_tunnel->tunnel_id;
7417 
7418 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
7419 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
7420 
7421 			return ret;
7422 		}
7423 	}
7424 
7425 	PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
7426 		     " Please remove a rule before adding a new one.");
7427 	return -EINVAL;
7428 }
7429 
7430 static inline struct ixgbe_l2_tn_filter *
7431 ixgbe_l2_tn_filter_lookup(struct ixgbe_l2_tn_info *l2_tn_info,
7432 			  struct ixgbe_l2_tn_key *key)
7433 {
7434 	int ret;
7435 
7436 	ret = rte_hash_lookup(l2_tn_info->hash_handle, (const void *)key);
7437 	if (ret < 0)
7438 		return NULL;
7439 
7440 	return l2_tn_info->hash_map[ret];
7441 }
7442 
7443 static inline int
7444 ixgbe_insert_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7445 			  struct ixgbe_l2_tn_filter *l2_tn_filter)
7446 {
7447 	int ret;
7448 
7449 	ret = rte_hash_add_key(l2_tn_info->hash_handle,
7450 			       &l2_tn_filter->key);
7451 
7452 	if (ret < 0) {
7453 		PMD_DRV_LOG(ERR,
7454 			    "Failed to insert L2 tunnel filter"
7455 			    " to hash table %d!",
7456 			    ret);
7457 		return ret;
7458 	}
7459 
7460 	l2_tn_info->hash_map[ret] = l2_tn_filter;
7461 
7462 	TAILQ_INSERT_TAIL(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7463 
7464 	return 0;
7465 }
7466 
7467 static inline int
7468 ixgbe_remove_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7469 			  struct ixgbe_l2_tn_key *key)
7470 {
7471 	int ret;
7472 	struct ixgbe_l2_tn_filter *l2_tn_filter;
7473 
7474 	ret = rte_hash_del_key(l2_tn_info->hash_handle, key);
7475 
7476 	if (ret < 0) {
7477 		PMD_DRV_LOG(ERR,
7478 			    "No such L2 tunnel filter to delete %d!",
7479 			    ret);
7480 		return ret;
7481 	}
7482 
7483 	l2_tn_filter = l2_tn_info->hash_map[ret];
7484 	l2_tn_info->hash_map[ret] = NULL;
7485 
7486 	TAILQ_REMOVE(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7487 	rte_free(l2_tn_filter);
7488 
7489 	return 0;
7490 }
7491 
7492 /* Add l2 tunnel filter */
7493 int
7494 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
7495 			       struct ixgbe_l2_tunnel_conf *l2_tunnel,
7496 			       bool restore)
7497 {
7498 	int ret;
7499 	struct ixgbe_l2_tn_info *l2_tn_info =
7500 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7501 	struct ixgbe_l2_tn_key key;
7502 	struct ixgbe_l2_tn_filter *node;
7503 
7504 	if (!restore) {
7505 		key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7506 		key.tn_id = l2_tunnel->tunnel_id;
7507 
7508 		node = ixgbe_l2_tn_filter_lookup(l2_tn_info, &key);
7509 
7510 		if (node) {
7511 			PMD_DRV_LOG(ERR,
7512 				    "The L2 tunnel filter already exists!");
7513 			return -EINVAL;
7514 		}
7515 
7516 		node = rte_zmalloc("ixgbe_l2_tn",
7517 				   sizeof(struct ixgbe_l2_tn_filter),
7518 				   0);
7519 		if (!node)
7520 			return -ENOMEM;
7521 
7522 		rte_memcpy(&node->key,
7523 				 &key,
7524 				 sizeof(struct ixgbe_l2_tn_key));
7525 		node->pool = l2_tunnel->pool;
7526 		ret = ixgbe_insert_l2_tn_filter(l2_tn_info, node);
7527 		if (ret < 0) {
7528 			rte_free(node);
7529 			return ret;
7530 		}
7531 	}
7532 
7533 	switch (l2_tunnel->l2_tunnel_type) {
7534 	case RTE_ETH_L2_TUNNEL_TYPE_E_TAG:
7535 		ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
7536 		break;
7537 	default:
7538 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7539 		ret = -EINVAL;
7540 		break;
7541 	}
7542 
7543 	if ((!restore) && (ret < 0))
7544 		(void)ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7545 
7546 	return ret;
7547 }
7548 
7549 /* Delete l2 tunnel filter */
7550 int
7551 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
7552 			       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7553 {
7554 	int ret;
7555 	struct ixgbe_l2_tn_info *l2_tn_info =
7556 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7557 	struct ixgbe_l2_tn_key key;
7558 
7559 	key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7560 	key.tn_id = l2_tunnel->tunnel_id;
7561 	ret = ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7562 	if (ret < 0)
7563 		return ret;
7564 
7565 	switch (l2_tunnel->l2_tunnel_type) {
7566 	case RTE_ETH_L2_TUNNEL_TYPE_E_TAG:
7567 		ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
7568 		break;
7569 	default:
7570 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7571 		ret = -EINVAL;
7572 		break;
7573 	}
7574 
7575 	return ret;
7576 }
7577 
7578 static int
7579 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
7580 {
7581 	int ret = 0;
7582 	uint32_t ctrl;
7583 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7584 
7585 	if (hw->mac.type != ixgbe_mac_X550 &&
7586 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7587 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7588 		return -ENOTSUP;
7589 	}
7590 
7591 	ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
7592 	ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
7593 	if (en)
7594 		ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
7595 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
7596 
7597 	return ret;
7598 }
7599 
7600 static int
7601 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
7602 			uint16_t port)
7603 {
7604 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
7605 	IXGBE_WRITE_FLUSH(hw);
7606 
7607 	return 0;
7608 }
7609 
7610 /* There's only one register for VxLAN UDP port.
7611  * So, we cannot add several ports. Will update it.
7612  */
7613 static int
7614 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
7615 		     uint16_t port)
7616 {
7617 	if (port == 0) {
7618 		PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
7619 		return -EINVAL;
7620 	}
7621 
7622 	return ixgbe_update_vxlan_port(hw, port);
7623 }
7624 
7625 /* We cannot delete the VxLAN port. For there's a register for VxLAN
7626  * UDP port, it must have a value.
7627  * So, will reset it to the original value 0.
7628  */
7629 static int
7630 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
7631 		     uint16_t port)
7632 {
7633 	uint16_t cur_port;
7634 
7635 	cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
7636 
7637 	if (cur_port != port) {
7638 		PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
7639 		return -EINVAL;
7640 	}
7641 
7642 	return ixgbe_update_vxlan_port(hw, 0);
7643 }
7644 
7645 /* Add UDP tunneling port */
7646 static int
7647 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
7648 			      struct rte_eth_udp_tunnel *udp_tunnel)
7649 {
7650 	int ret = 0;
7651 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7652 
7653 	if (hw->mac.type != ixgbe_mac_X550 &&
7654 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7655 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7656 		return -ENOTSUP;
7657 	}
7658 
7659 	if (udp_tunnel == NULL)
7660 		return -EINVAL;
7661 
7662 	switch (udp_tunnel->prot_type) {
7663 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
7664 		ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
7665 		break;
7666 
7667 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
7668 	case RTE_ETH_TUNNEL_TYPE_TEREDO:
7669 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7670 		ret = -EINVAL;
7671 		break;
7672 
7673 	default:
7674 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7675 		ret = -EINVAL;
7676 		break;
7677 	}
7678 
7679 	return ret;
7680 }
7681 
7682 /* Remove UDP tunneling port */
7683 static int
7684 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
7685 			      struct rte_eth_udp_tunnel *udp_tunnel)
7686 {
7687 	int ret = 0;
7688 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7689 
7690 	if (hw->mac.type != ixgbe_mac_X550 &&
7691 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7692 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7693 		return -ENOTSUP;
7694 	}
7695 
7696 	if (udp_tunnel == NULL)
7697 		return -EINVAL;
7698 
7699 	switch (udp_tunnel->prot_type) {
7700 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
7701 		ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
7702 		break;
7703 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
7704 	case RTE_ETH_TUNNEL_TYPE_TEREDO:
7705 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7706 		ret = -EINVAL;
7707 		break;
7708 	default:
7709 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7710 		ret = -EINVAL;
7711 		break;
7712 	}
7713 
7714 	return ret;
7715 }
7716 
7717 static int
7718 ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev)
7719 {
7720 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7721 	int ret;
7722 
7723 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_PROMISC)) {
7724 	case IXGBE_SUCCESS:
7725 		ret = 0;
7726 		break;
7727 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7728 		ret = -ENOTSUP;
7729 		break;
7730 	default:
7731 		ret = -EAGAIN;
7732 		break;
7733 	}
7734 
7735 	return ret;
7736 }
7737 
7738 static int
7739 ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev)
7740 {
7741 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7742 	int ret;
7743 
7744 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_NONE)) {
7745 	case IXGBE_SUCCESS:
7746 		ret = 0;
7747 		break;
7748 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7749 		ret = -ENOTSUP;
7750 		break;
7751 	default:
7752 		ret = -EAGAIN;
7753 		break;
7754 	}
7755 
7756 	return ret;
7757 }
7758 
7759 static int
7760 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
7761 {
7762 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7763 	int ret;
7764 	int mode = IXGBEVF_XCAST_MODE_ALLMULTI;
7765 
7766 	switch (hw->mac.ops.update_xcast_mode(hw, mode)) {
7767 	case IXGBE_SUCCESS:
7768 		ret = 0;
7769 		break;
7770 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7771 		ret = -ENOTSUP;
7772 		break;
7773 	default:
7774 		ret = -EAGAIN;
7775 		break;
7776 	}
7777 
7778 	return ret;
7779 }
7780 
7781 static int
7782 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
7783 {
7784 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7785 	int ret;
7786 
7787 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_MULTI)) {
7788 	case IXGBE_SUCCESS:
7789 		ret = 0;
7790 		break;
7791 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7792 		ret = -ENOTSUP;
7793 		break;
7794 	default:
7795 		ret = -EAGAIN;
7796 		break;
7797 	}
7798 
7799 	return ret;
7800 }
7801 
7802 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
7803 {
7804 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7805 	u32 in_msg = 0;
7806 
7807 	/* peek the message first */
7808 	in_msg = IXGBE_READ_REG(hw, IXGBE_VFMBMEM);
7809 
7810 	/* PF reset VF event */
7811 	if (in_msg == IXGBE_PF_CONTROL_MSG) {
7812 		/* dummy mbx read to ack pf */
7813 		if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
7814 			return;
7815 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
7816 					     NULL);
7817 	}
7818 }
7819 
7820 static int
7821 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
7822 {
7823 	uint32_t eicr;
7824 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7825 	struct ixgbe_interrupt *intr =
7826 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7827 	ixgbevf_intr_disable(dev);
7828 
7829 	/* read-on-clear nic registers here */
7830 	eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
7831 	intr->flags = 0;
7832 
7833 	/* only one misc vector supported - mailbox */
7834 	eicr &= IXGBE_VTEICR_MASK;
7835 	if (eicr == IXGBE_MISC_VEC_ID)
7836 		intr->flags |= IXGBE_FLAG_MAILBOX;
7837 
7838 	return 0;
7839 }
7840 
7841 static int
7842 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
7843 {
7844 	struct ixgbe_interrupt *intr =
7845 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7846 
7847 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
7848 		ixgbevf_mbx_process(dev);
7849 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
7850 	}
7851 
7852 	ixgbevf_intr_enable(dev);
7853 
7854 	return 0;
7855 }
7856 
7857 static void
7858 ixgbevf_dev_interrupt_handler(void *param)
7859 {
7860 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
7861 
7862 	ixgbevf_dev_interrupt_get_status(dev);
7863 	ixgbevf_dev_interrupt_action(dev);
7864 }
7865 
7866 /**
7867  *  ixgbe_disable_sec_tx_path_generic - Stops the transmit data path
7868  *  @hw: pointer to hardware structure
7869  *
7870  *  Stops the transmit data path and waits for the HW to internally empty
7871  *  the Tx security block
7872  **/
7873 int ixgbe_disable_sec_tx_path_generic(struct ixgbe_hw *hw)
7874 {
7875 #define IXGBE_MAX_SECTX_POLL 40
7876 
7877 	int i;
7878 	int sectxreg;
7879 
7880 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
7881 	sectxreg |= IXGBE_SECTXCTRL_TX_DIS;
7882 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
7883 	for (i = 0; i < IXGBE_MAX_SECTX_POLL; i++) {
7884 		sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXSTAT);
7885 		if (sectxreg & IXGBE_SECTXSTAT_SECTX_RDY)
7886 			break;
7887 		/* Use interrupt-safe sleep just in case */
7888 		usec_delay(1000);
7889 	}
7890 
7891 	/* For informational purposes only */
7892 	if (i >= IXGBE_MAX_SECTX_POLL)
7893 		PMD_DRV_LOG(DEBUG, "Tx unit being enabled before security "
7894 			 "path fully disabled.  Continuing with init.");
7895 
7896 	return IXGBE_SUCCESS;
7897 }
7898 
7899 /**
7900  *  ixgbe_enable_sec_tx_path_generic - Enables the transmit data path
7901  *  @hw: pointer to hardware structure
7902  *
7903  *  Enables the transmit data path.
7904  **/
7905 int ixgbe_enable_sec_tx_path_generic(struct ixgbe_hw *hw)
7906 {
7907 	uint32_t sectxreg;
7908 
7909 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
7910 	sectxreg &= ~IXGBE_SECTXCTRL_TX_DIS;
7911 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
7912 	IXGBE_WRITE_FLUSH(hw);
7913 
7914 	return IXGBE_SUCCESS;
7915 }
7916 
7917 /* restore n-tuple filter */
7918 static inline void
7919 ixgbe_ntuple_filter_restore(struct rte_eth_dev *dev)
7920 {
7921 	struct ixgbe_filter_info *filter_info =
7922 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7923 	struct ixgbe_5tuple_filter *node;
7924 
7925 	TAILQ_FOREACH(node, &filter_info->fivetuple_list, entries) {
7926 		ixgbe_inject_5tuple_filter(dev, node);
7927 	}
7928 }
7929 
7930 /* restore ethernet type filter */
7931 static inline void
7932 ixgbe_ethertype_filter_restore(struct rte_eth_dev *dev)
7933 {
7934 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7935 	struct ixgbe_filter_info *filter_info =
7936 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7937 	int i;
7938 
7939 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
7940 		if (filter_info->ethertype_mask & (1 << i)) {
7941 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i),
7942 					filter_info->ethertype_filters[i].etqf);
7943 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i),
7944 					filter_info->ethertype_filters[i].etqs);
7945 			IXGBE_WRITE_FLUSH(hw);
7946 		}
7947 	}
7948 }
7949 
7950 /* restore SYN filter */
7951 static inline void
7952 ixgbe_syn_filter_restore(struct rte_eth_dev *dev)
7953 {
7954 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7955 	struct ixgbe_filter_info *filter_info =
7956 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7957 	uint32_t synqf;
7958 
7959 	synqf = filter_info->syn_info;
7960 
7961 	if (synqf & IXGBE_SYN_FILTER_ENABLE) {
7962 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
7963 		IXGBE_WRITE_FLUSH(hw);
7964 	}
7965 }
7966 
7967 /* restore L2 tunnel filter */
7968 static inline void
7969 ixgbe_l2_tn_filter_restore(struct rte_eth_dev *dev)
7970 {
7971 	struct ixgbe_l2_tn_info *l2_tn_info =
7972 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7973 	struct ixgbe_l2_tn_filter *node;
7974 	struct ixgbe_l2_tunnel_conf l2_tn_conf;
7975 
7976 	TAILQ_FOREACH(node, &l2_tn_info->l2_tn_list, entries) {
7977 		l2_tn_conf.l2_tunnel_type = node->key.l2_tn_type;
7978 		l2_tn_conf.tunnel_id      = node->key.tn_id;
7979 		l2_tn_conf.pool           = node->pool;
7980 		(void)ixgbe_dev_l2_tunnel_filter_add(dev, &l2_tn_conf, TRUE);
7981 	}
7982 }
7983 
7984 /* restore rss filter */
7985 static inline void
7986 ixgbe_rss_filter_restore(struct rte_eth_dev *dev)
7987 {
7988 	struct ixgbe_filter_info *filter_info =
7989 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
7990 
7991 	if (filter_info->rss_info.conf.queue_num)
7992 		ixgbe_config_rss_filter(dev,
7993 			&filter_info->rss_info, TRUE);
7994 }
7995 
7996 static int
7997 ixgbe_filter_restore(struct rte_eth_dev *dev)
7998 {
7999 	ixgbe_ntuple_filter_restore(dev);
8000 	ixgbe_ethertype_filter_restore(dev);
8001 	ixgbe_syn_filter_restore(dev);
8002 	ixgbe_fdir_filter_restore(dev);
8003 	ixgbe_l2_tn_filter_restore(dev);
8004 	ixgbe_rss_filter_restore(dev);
8005 
8006 	return 0;
8007 }
8008 
8009 static void
8010 ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev)
8011 {
8012 	struct ixgbe_l2_tn_info *l2_tn_info =
8013 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8014 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8015 
8016 	if (l2_tn_info->e_tag_en)
8017 		(void)ixgbe_e_tag_enable(hw);
8018 
8019 	if (l2_tn_info->e_tag_fwd_en)
8020 		(void)ixgbe_e_tag_forwarding_en_dis(dev, 1);
8021 
8022 	(void)ixgbe_update_e_tag_eth_type(hw, l2_tn_info->e_tag_ether_type);
8023 }
8024 
8025 /* remove all the n-tuple filters */
8026 void
8027 ixgbe_clear_all_ntuple_filter(struct rte_eth_dev *dev)
8028 {
8029 	struct ixgbe_filter_info *filter_info =
8030 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8031 	struct ixgbe_5tuple_filter *p_5tuple;
8032 
8033 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
8034 		ixgbe_remove_5tuple_filter(dev, p_5tuple);
8035 }
8036 
8037 /* remove all the ether type filters */
8038 void
8039 ixgbe_clear_all_ethertype_filter(struct rte_eth_dev *dev)
8040 {
8041 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8042 	struct ixgbe_filter_info *filter_info =
8043 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8044 	int i;
8045 
8046 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8047 		if (filter_info->ethertype_mask & (1 << i) &&
8048 		    !filter_info->ethertype_filters[i].conf) {
8049 			(void)ixgbe_ethertype_filter_remove(filter_info,
8050 							    (uint8_t)i);
8051 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i), 0);
8052 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i), 0);
8053 			IXGBE_WRITE_FLUSH(hw);
8054 		}
8055 	}
8056 }
8057 
8058 /* remove the SYN filter */
8059 void
8060 ixgbe_clear_syn_filter(struct rte_eth_dev *dev)
8061 {
8062 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8063 	struct ixgbe_filter_info *filter_info =
8064 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8065 
8066 	if (filter_info->syn_info & IXGBE_SYN_FILTER_ENABLE) {
8067 		filter_info->syn_info = 0;
8068 
8069 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, 0);
8070 		IXGBE_WRITE_FLUSH(hw);
8071 	}
8072 }
8073 
8074 /* remove all the L2 tunnel filters */
8075 int
8076 ixgbe_clear_all_l2_tn_filter(struct rte_eth_dev *dev)
8077 {
8078 	struct ixgbe_l2_tn_info *l2_tn_info =
8079 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8080 	struct ixgbe_l2_tn_filter *l2_tn_filter;
8081 	struct ixgbe_l2_tunnel_conf l2_tn_conf;
8082 	int ret = 0;
8083 
8084 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
8085 		l2_tn_conf.l2_tunnel_type = l2_tn_filter->key.l2_tn_type;
8086 		l2_tn_conf.tunnel_id      = l2_tn_filter->key.tn_id;
8087 		l2_tn_conf.pool           = l2_tn_filter->pool;
8088 		ret = ixgbe_dev_l2_tunnel_filter_del(dev, &l2_tn_conf);
8089 		if (ret < 0)
8090 			return ret;
8091 	}
8092 
8093 	return 0;
8094 }
8095 
8096 void
8097 ixgbe_dev_macsec_setting_save(struct rte_eth_dev *dev,
8098 				struct ixgbe_macsec_setting *macsec_setting)
8099 {
8100 	struct ixgbe_macsec_setting *macsec =
8101 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8102 
8103 	macsec->offload_en = macsec_setting->offload_en;
8104 	macsec->encrypt_en = macsec_setting->encrypt_en;
8105 	macsec->replayprotect_en = macsec_setting->replayprotect_en;
8106 }
8107 
8108 void
8109 ixgbe_dev_macsec_setting_reset(struct rte_eth_dev *dev)
8110 {
8111 	struct ixgbe_macsec_setting *macsec =
8112 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8113 
8114 	macsec->offload_en = 0;
8115 	macsec->encrypt_en = 0;
8116 	macsec->replayprotect_en = 0;
8117 }
8118 
8119 void
8120 ixgbe_dev_macsec_register_enable(struct rte_eth_dev *dev,
8121 				struct ixgbe_macsec_setting *macsec_setting)
8122 {
8123 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8124 	uint32_t ctrl;
8125 	uint8_t en = macsec_setting->encrypt_en;
8126 	uint8_t rp = macsec_setting->replayprotect_en;
8127 
8128 	/**
8129 	 * Workaround:
8130 	 * As no ixgbe_disable_sec_rx_path equivalent is
8131 	 * implemented for tx in the base code, and we are
8132 	 * not allowed to modify the base code in DPDK, so
8133 	 * just call the hand-written one directly for now.
8134 	 * The hardware support has been checked by
8135 	 * ixgbe_disable_sec_rx_path().
8136 	 */
8137 	ixgbe_disable_sec_tx_path_generic(hw);
8138 
8139 	/* Enable Ethernet CRC (required by MACsec offload) */
8140 	ctrl = IXGBE_READ_REG(hw, IXGBE_HLREG0);
8141 	ctrl |= IXGBE_HLREG0_TXCRCEN | IXGBE_HLREG0_RXCRCSTRP;
8142 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, ctrl);
8143 
8144 	/* Enable the TX and RX crypto engines */
8145 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8146 	ctrl &= ~IXGBE_SECTXCTRL_SECTX_DIS;
8147 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8148 
8149 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8150 	ctrl &= ~IXGBE_SECRXCTRL_SECRX_DIS;
8151 	IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8152 
8153 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
8154 	ctrl &= ~IXGBE_SECTX_MINSECIFG_MASK;
8155 	ctrl |= 0x3;
8156 	IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, ctrl);
8157 
8158 	/* Enable SA lookup */
8159 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8160 	ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8161 	ctrl |= en ? IXGBE_LSECTXCTRL_AUTH_ENCRYPT :
8162 		     IXGBE_LSECTXCTRL_AUTH;
8163 	ctrl |= IXGBE_LSECTXCTRL_AISCI;
8164 	ctrl &= ~IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8165 	ctrl |= IXGBE_MACSEC_PNTHRSH & IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8166 	IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8167 
8168 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8169 	ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8170 	ctrl |= IXGBE_LSECRXCTRL_STRICT << IXGBE_LSECRXCTRL_EN_SHIFT;
8171 	ctrl &= ~IXGBE_LSECRXCTRL_PLSH;
8172 	if (rp)
8173 		ctrl |= IXGBE_LSECRXCTRL_RP;
8174 	else
8175 		ctrl &= ~IXGBE_LSECRXCTRL_RP;
8176 	IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8177 
8178 	/* Start the data paths */
8179 	ixgbe_enable_sec_rx_path(hw);
8180 	/**
8181 	 * Workaround:
8182 	 * As no ixgbe_enable_sec_rx_path equivalent is
8183 	 * implemented for tx in the base code, and we are
8184 	 * not allowed to modify the base code in DPDK, so
8185 	 * just call the hand-written one directly for now.
8186 	 */
8187 	ixgbe_enable_sec_tx_path_generic(hw);
8188 }
8189 
8190 void
8191 ixgbe_dev_macsec_register_disable(struct rte_eth_dev *dev)
8192 {
8193 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8194 	uint32_t ctrl;
8195 
8196 	/**
8197 	 * Workaround:
8198 	 * As no ixgbe_disable_sec_rx_path equivalent is
8199 	 * implemented for tx in the base code, and we are
8200 	 * not allowed to modify the base code in DPDK, so
8201 	 * just call the hand-written one directly for now.
8202 	 * The hardware support has been checked by
8203 	 * ixgbe_disable_sec_rx_path().
8204 	 */
8205 	ixgbe_disable_sec_tx_path_generic(hw);
8206 
8207 	/* Disable the TX and RX crypto engines */
8208 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8209 	ctrl |= IXGBE_SECTXCTRL_SECTX_DIS;
8210 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8211 
8212 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8213 	ctrl |= IXGBE_SECRXCTRL_SECRX_DIS;
8214 	IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8215 
8216 	/* Disable SA lookup */
8217 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8218 	ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8219 	ctrl |= IXGBE_LSECTXCTRL_DISABLE;
8220 	IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8221 
8222 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8223 	ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8224 	ctrl |= IXGBE_LSECRXCTRL_DISABLE << IXGBE_LSECRXCTRL_EN_SHIFT;
8225 	IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8226 
8227 	/* Start the data paths */
8228 	ixgbe_enable_sec_rx_path(hw);
8229 	/**
8230 	 * Workaround:
8231 	 * As no ixgbe_enable_sec_rx_path equivalent is
8232 	 * implemented for tx in the base code, and we are
8233 	 * not allowed to modify the base code in DPDK, so
8234 	 * just call the hand-written one directly for now.
8235 	 */
8236 	ixgbe_enable_sec_tx_path_generic(hw);
8237 }
8238 
8239 RTE_PMD_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd);
8240 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
8241 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe, "* igb_uio | uio_pci_generic | vfio-pci");
8242 RTE_PMD_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd);
8243 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);
8244 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe_vf, "* igb_uio | vfio-pci");
8245 RTE_PMD_REGISTER_PARAM_STRING(net_ixgbe_vf,
8246 			      IXGBEVF_DEVARG_PFLINK_FULLCHK "=<0|1>");
8247 
8248 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_init, init, NOTICE);
8249 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_driver, driver, NOTICE);
8250 
8251 #ifdef RTE_ETHDEV_DEBUG_RX
8252 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_rx, rx, DEBUG);
8253 #endif
8254 #ifdef RTE_ETHDEV_DEBUG_TX
8255 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_tx, tx, DEBUG);
8256 #endif
8257