xref: /dpdk/drivers/net/ixgbe/ixgbe_ethdev.c (revision c7e9729d)
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 <netinet/in.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_eal.h>
26 #include <rte_alarm.h>
27 #include <rte_ether.h>
28 #include <rte_ethdev_driver.h>
29 #include <rte_ethdev_pci.h>
30 #include <rte_malloc.h>
31 #include <rte_random.h>
32 #include <rte_dev.h>
33 #include <rte_hash_crc.h>
34 #ifdef RTE_LIBRTE_SECURITY
35 #include <rte_security_driver.h>
36 #endif
37 
38 #include "ixgbe_logs.h"
39 #include "base/ixgbe_api.h"
40 #include "base/ixgbe_vf.h"
41 #include "base/ixgbe_common.h"
42 #include "ixgbe_ethdev.h"
43 #include "ixgbe_bypass.h"
44 #include "ixgbe_rxtx.h"
45 #include "base/ixgbe_type.h"
46 #include "base/ixgbe_phy.h"
47 #include "ixgbe_regs.h"
48 
49 /*
50  * High threshold controlling when to start sending XOFF frames. Must be at
51  * least 8 bytes less than receive packet buffer size. This value is in units
52  * of 1024 bytes.
53  */
54 #define IXGBE_FC_HI    0x80
55 
56 /*
57  * Low threshold controlling when to start sending XON frames. This value is
58  * in units of 1024 bytes.
59  */
60 #define IXGBE_FC_LO    0x40
61 
62 /* Default minimum inter-interrupt interval for EITR configuration */
63 #define IXGBE_MIN_INTER_INTERRUPT_INTERVAL_DEFAULT    0x79E
64 
65 /* Timer value included in XOFF frames. */
66 #define IXGBE_FC_PAUSE 0x680
67 
68 /*Default value of Max Rx Queue*/
69 #define IXGBE_MAX_RX_QUEUE_NUM 128
70 
71 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
72 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
73 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
74 
75 #define IXGBE_MMW_SIZE_DEFAULT        0x4
76 #define IXGBE_MMW_SIZE_JUMBO_FRAME    0x14
77 #define IXGBE_MAX_RING_DESC           4096 /* replicate define from rxtx */
78 
79 /*
80  *  Default values for RX/TX configuration
81  */
82 #define IXGBE_DEFAULT_RX_FREE_THRESH  32
83 #define IXGBE_DEFAULT_RX_PTHRESH      8
84 #define IXGBE_DEFAULT_RX_HTHRESH      8
85 #define IXGBE_DEFAULT_RX_WTHRESH      0
86 
87 #define IXGBE_DEFAULT_TX_FREE_THRESH  32
88 #define IXGBE_DEFAULT_TX_PTHRESH      32
89 #define IXGBE_DEFAULT_TX_HTHRESH      0
90 #define IXGBE_DEFAULT_TX_WTHRESH      0
91 #define IXGBE_DEFAULT_TX_RSBIT_THRESH 32
92 
93 /* Bit shift and mask */
94 #define IXGBE_4_BIT_WIDTH  (CHAR_BIT / 2)
95 #define IXGBE_4_BIT_MASK   RTE_LEN2MASK(IXGBE_4_BIT_WIDTH, uint8_t)
96 #define IXGBE_8_BIT_WIDTH  CHAR_BIT
97 #define IXGBE_8_BIT_MASK   UINT8_MAX
98 
99 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
100 
101 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
102 
103 #define IXGBE_HKEY_MAX_INDEX 10
104 
105 /* Additional timesync values. */
106 #define NSEC_PER_SEC             1000000000L
107 #define IXGBE_INCVAL_10GB        0x66666666
108 #define IXGBE_INCVAL_1GB         0x40000000
109 #define IXGBE_INCVAL_100         0x50000000
110 #define IXGBE_INCVAL_SHIFT_10GB  28
111 #define IXGBE_INCVAL_SHIFT_1GB   24
112 #define IXGBE_INCVAL_SHIFT_100   21
113 #define IXGBE_INCVAL_SHIFT_82599 7
114 #define IXGBE_INCPER_SHIFT_82599 24
115 
116 #define IXGBE_CYCLECOUNTER_MASK   0xffffffffffffffffULL
117 
118 #define IXGBE_VT_CTL_POOLING_MODE_MASK         0x00030000
119 #define IXGBE_VT_CTL_POOLING_MODE_ETAG         0x00010000
120 #define DEFAULT_ETAG_ETYPE                     0x893f
121 #define IXGBE_ETAG_ETYPE                       0x00005084
122 #define IXGBE_ETAG_ETYPE_MASK                  0x0000ffff
123 #define IXGBE_ETAG_ETYPE_VALID                 0x80000000
124 #define IXGBE_RAH_ADTYPE                       0x40000000
125 #define IXGBE_RAL_ETAG_FILTER_MASK             0x00003fff
126 #define IXGBE_VMVIR_TAGA_MASK                  0x18000000
127 #define IXGBE_VMVIR_TAGA_ETAG_INSERT           0x08000000
128 #define IXGBE_VMTIR(_i) (0x00017000 + ((_i) * 4)) /* 64 of these (0-63) */
129 #define IXGBE_QDE_STRIP_TAG                    0x00000004
130 #define IXGBE_VTEICR_MASK                      0x07
131 
132 #define IXGBE_EXVET_VET_EXT_SHIFT              16
133 #define IXGBE_DMATXCTL_VT_MASK                 0xFFFF0000
134 
135 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev);
136 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
137 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev);
138 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev);
139 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev);
140 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev);
141 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev);
142 static int  ixgbe_dev_configure(struct rte_eth_dev *dev);
143 static int  ixgbe_dev_start(struct rte_eth_dev *dev);
144 static void ixgbe_dev_stop(struct rte_eth_dev *dev);
145 static int  ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
146 static int  ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
147 static void ixgbe_dev_close(struct rte_eth_dev *dev);
148 static int  ixgbe_dev_reset(struct rte_eth_dev *dev);
149 static void ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
150 static void ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
151 static void ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
152 static void ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
153 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
154 				int wait_to_complete);
155 static int ixgbe_dev_stats_get(struct rte_eth_dev *dev,
156 				struct rte_eth_stats *stats);
157 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
158 				struct rte_eth_xstat *xstats, unsigned n);
159 static int ixgbevf_dev_xstats_get(struct rte_eth_dev *dev,
160 				  struct rte_eth_xstat *xstats, unsigned n);
161 static int
162 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
163 		uint64_t *values, unsigned int n);
164 static void ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
165 static void ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
166 static int ixgbe_dev_xstats_get_names(struct rte_eth_dev *dev,
167 	struct rte_eth_xstat_name *xstats_names,
168 	unsigned int size);
169 static int ixgbevf_dev_xstats_get_names(struct rte_eth_dev *dev,
170 	struct rte_eth_xstat_name *xstats_names, unsigned limit);
171 static int ixgbe_dev_xstats_get_names_by_id(
172 	struct rte_eth_dev *dev,
173 	struct rte_eth_xstat_name *xstats_names,
174 	const uint64_t *ids,
175 	unsigned int limit);
176 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
177 					     uint16_t queue_id,
178 					     uint8_t stat_idx,
179 					     uint8_t is_rx);
180 static int ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
181 				 size_t fw_size);
182 static void ixgbe_dev_info_get(struct rte_eth_dev *dev,
183 			       struct rte_eth_dev_info *dev_info);
184 static const uint32_t *ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev);
185 static void ixgbevf_dev_info_get(struct rte_eth_dev *dev,
186 				 struct rte_eth_dev_info *dev_info);
187 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
188 
189 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
190 		uint16_t vlan_id, int on);
191 static int ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
192 			       enum rte_vlan_type vlan_type,
193 			       uint16_t tpid_id);
194 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
195 		uint16_t queue, bool on);
196 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
197 		int on);
198 static int ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
199 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
200 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
201 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
202 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
203 
204 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
205 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
206 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
207 			       struct rte_eth_fc_conf *fc_conf);
208 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
209 			       struct rte_eth_fc_conf *fc_conf);
210 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
211 		struct rte_eth_pfc_conf *pfc_conf);
212 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
213 			struct rte_eth_rss_reta_entry64 *reta_conf,
214 			uint16_t reta_size);
215 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
216 			struct rte_eth_rss_reta_entry64 *reta_conf,
217 			uint16_t reta_size);
218 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
219 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
220 static int ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev);
221 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
222 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
223 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev,
224 				      struct rte_intr_handle *handle);
225 static void ixgbe_dev_interrupt_handler(void *param);
226 static void ixgbe_dev_interrupt_delayed_handler(void *param);
227 static int ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
228 			 uint32_t index, uint32_t pool);
229 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
230 static void ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
231 					   struct ether_addr *mac_addr);
232 static void ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config);
233 static bool is_device_supported(struct rte_eth_dev *dev,
234 				struct rte_pci_driver *drv);
235 
236 /* For Virtual Function support */
237 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
238 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
239 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
240 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
241 static int ixgbevf_dev_link_update(struct rte_eth_dev *dev,
242 				   int wait_to_complete);
243 static void ixgbevf_dev_stop(struct rte_eth_dev *dev);
244 static void ixgbevf_dev_close(struct rte_eth_dev *dev);
245 static int  ixgbevf_dev_reset(struct rte_eth_dev *dev);
246 static void ixgbevf_intr_disable(struct ixgbe_hw *hw);
247 static void ixgbevf_intr_enable(struct ixgbe_hw *hw);
248 static int ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
249 		struct rte_eth_stats *stats);
250 static void ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
251 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
252 		uint16_t vlan_id, int on);
253 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
254 		uint16_t queue, int on);
255 static int ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
256 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
257 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
258 					    uint16_t queue_id);
259 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
260 					     uint16_t queue_id);
261 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
262 				 uint8_t queue, uint8_t msix_vector);
263 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
264 static void ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev);
265 static void ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev);
266 
267 /* For Eth VMDQ APIs support */
268 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
269 		ether_addr * mac_addr, uint8_t on);
270 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on);
271 static int ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
272 		struct rte_eth_mirror_conf *mirror_conf,
273 		uint8_t rule_id, uint8_t on);
274 static int ixgbe_mirror_rule_reset(struct rte_eth_dev *dev,
275 		uint8_t	rule_id);
276 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
277 					  uint16_t queue_id);
278 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
279 					   uint16_t queue_id);
280 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
281 			       uint8_t queue, uint8_t msix_vector);
282 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
283 
284 static int ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
285 				struct ether_addr *mac_addr,
286 				uint32_t index, uint32_t pool);
287 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
288 static void ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
289 					     struct ether_addr *mac_addr);
290 static int ixgbe_syn_filter_get(struct rte_eth_dev *dev,
291 			struct rte_eth_syn_filter *filter);
292 static int ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
293 			enum rte_filter_op filter_op,
294 			void *arg);
295 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
296 			struct ixgbe_5tuple_filter *filter);
297 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
298 			struct ixgbe_5tuple_filter *filter);
299 static int ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
300 				enum rte_filter_op filter_op,
301 				void *arg);
302 static int ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
303 			struct rte_eth_ntuple_filter *filter);
304 static int ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
305 				enum rte_filter_op filter_op,
306 				void *arg);
307 static int ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
308 			struct rte_eth_ethertype_filter *filter);
309 static int ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
310 		     enum rte_filter_type filter_type,
311 		     enum rte_filter_op filter_op,
312 		     void *arg);
313 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
314 
315 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
316 				      struct ether_addr *mc_addr_set,
317 				      uint32_t nb_mc_addr);
318 static int ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
319 				   struct rte_eth_dcb_info *dcb_info);
320 
321 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
322 static int ixgbe_get_regs(struct rte_eth_dev *dev,
323 			    struct rte_dev_reg_info *regs);
324 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
325 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
326 				struct rte_dev_eeprom_info *eeprom);
327 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
328 				struct rte_dev_eeprom_info *eeprom);
329 
330 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
331 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
332 				struct rte_dev_reg_info *regs);
333 
334 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
335 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
336 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
337 					    struct timespec *timestamp,
338 					    uint32_t flags);
339 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
340 					    struct timespec *timestamp);
341 static int ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
342 static int ixgbe_timesync_read_time(struct rte_eth_dev *dev,
343 				   struct timespec *timestamp);
344 static int ixgbe_timesync_write_time(struct rte_eth_dev *dev,
345 				   const struct timespec *timestamp);
346 static void ixgbevf_dev_interrupt_handler(void *param);
347 
348 static int ixgbe_dev_l2_tunnel_eth_type_conf
349 	(struct rte_eth_dev *dev, struct rte_eth_l2_tunnel_conf *l2_tunnel);
350 static int ixgbe_dev_l2_tunnel_offload_set
351 	(struct rte_eth_dev *dev,
352 	 struct rte_eth_l2_tunnel_conf *l2_tunnel,
353 	 uint32_t mask,
354 	 uint8_t en);
355 static int ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
356 					     enum rte_filter_op filter_op,
357 					     void *arg);
358 
359 static int ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
360 					 struct rte_eth_udp_tunnel *udp_tunnel);
361 static int ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
362 					 struct rte_eth_udp_tunnel *udp_tunnel);
363 static int ixgbe_filter_restore(struct rte_eth_dev *dev);
364 static void ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev);
365 
366 /*
367  * Define VF Stats MACRO for Non "cleared on read" register
368  */
369 #define UPDATE_VF_STAT(reg, last, cur)                          \
370 {                                                               \
371 	uint32_t latest = IXGBE_READ_REG(hw, reg);              \
372 	cur += (latest - last) & UINT_MAX;                      \
373 	last = latest;                                          \
374 }
375 
376 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
377 {                                                                \
378 	u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
379 	u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
380 	u64 latest = ((new_msb << 32) | new_lsb);                \
381 	cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
382 	last = latest;                                           \
383 }
384 
385 #define IXGBE_SET_HWSTRIP(h, q) do {\
386 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
387 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
388 		(h)->bitmap[idx] |= 1 << bit;\
389 	} while (0)
390 
391 #define IXGBE_CLEAR_HWSTRIP(h, q) do {\
392 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
393 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
394 		(h)->bitmap[idx] &= ~(1 << bit);\
395 	} while (0)
396 
397 #define IXGBE_GET_HWSTRIP(h, q, r) do {\
398 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
399 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
400 		(r) = (h)->bitmap[idx] >> bit & 1;\
401 	} while (0)
402 
403 int ixgbe_logtype_init;
404 int ixgbe_logtype_driver;
405 
406 /*
407  * The set of PCI devices this driver supports
408  */
409 static const struct rte_pci_id pci_id_ixgbe_map[] = {
410 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598) },
411 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_BX) },
412 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT) },
413 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT) },
414 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT) },
415 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2) },
416 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM) },
417 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4) },
418 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT) },
419 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT) },
420 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM) },
421 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR) },
422 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4) },
423 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ) },
424 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KR) },
425 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE) },
426 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4) },
427 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP) },
428 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE) },
429 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE) },
430 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_EM) },
431 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2) },
432 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP) },
433 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP) },
434 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP) },
435 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM) },
436 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM) },
437 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_LS) },
438 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T) },
439 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1) },
440 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP) },
441 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T) },
442 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_1G_T) },
443 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T) },
444 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1) },
445 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR) },
446 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR_L) },
447 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP_N) },
448 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII) },
449 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII_L) },
450 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_10G_T) },
451 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP) },
452 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP_N) },
453 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP) },
454 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T) },
455 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T_L) },
456 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4) },
457 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR) },
458 #ifdef RTE_LIBRTE_IXGBE_BYPASS
459 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BYPASS) },
460 #endif
461 	{ .vendor_id = 0, /* sentinel */ },
462 };
463 
464 /*
465  * The set of PCI devices this driver supports (for 82599 VF)
466  */
467 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
468 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF) },
469 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF_HV) },
470 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF) },
471 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF_HV) },
472 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF_HV) },
473 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF) },
474 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF) },
475 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF_HV) },
476 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF) },
477 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF_HV) },
478 	{ .vendor_id = 0, /* sentinel */ },
479 };
480 
481 static const struct rte_eth_desc_lim rx_desc_lim = {
482 	.nb_max = IXGBE_MAX_RING_DESC,
483 	.nb_min = IXGBE_MIN_RING_DESC,
484 	.nb_align = IXGBE_RXD_ALIGN,
485 };
486 
487 static const struct rte_eth_desc_lim tx_desc_lim = {
488 	.nb_max = IXGBE_MAX_RING_DESC,
489 	.nb_min = IXGBE_MIN_RING_DESC,
490 	.nb_align = IXGBE_TXD_ALIGN,
491 	.nb_seg_max = IXGBE_TX_MAX_SEG,
492 	.nb_mtu_seg_max = IXGBE_TX_MAX_SEG,
493 };
494 
495 static const struct eth_dev_ops ixgbe_eth_dev_ops = {
496 	.dev_configure        = ixgbe_dev_configure,
497 	.dev_start            = ixgbe_dev_start,
498 	.dev_stop             = ixgbe_dev_stop,
499 	.dev_set_link_up    = ixgbe_dev_set_link_up,
500 	.dev_set_link_down  = ixgbe_dev_set_link_down,
501 	.dev_close            = ixgbe_dev_close,
502 	.dev_reset	      = ixgbe_dev_reset,
503 	.promiscuous_enable   = ixgbe_dev_promiscuous_enable,
504 	.promiscuous_disable  = ixgbe_dev_promiscuous_disable,
505 	.allmulticast_enable  = ixgbe_dev_allmulticast_enable,
506 	.allmulticast_disable = ixgbe_dev_allmulticast_disable,
507 	.link_update          = ixgbe_dev_link_update,
508 	.stats_get            = ixgbe_dev_stats_get,
509 	.xstats_get           = ixgbe_dev_xstats_get,
510 	.xstats_get_by_id     = ixgbe_dev_xstats_get_by_id,
511 	.stats_reset          = ixgbe_dev_stats_reset,
512 	.xstats_reset         = ixgbe_dev_xstats_reset,
513 	.xstats_get_names     = ixgbe_dev_xstats_get_names,
514 	.xstats_get_names_by_id = ixgbe_dev_xstats_get_names_by_id,
515 	.queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
516 	.fw_version_get       = ixgbe_fw_version_get,
517 	.dev_infos_get        = ixgbe_dev_info_get,
518 	.dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
519 	.mtu_set              = ixgbe_dev_mtu_set,
520 	.vlan_filter_set      = ixgbe_vlan_filter_set,
521 	.vlan_tpid_set        = ixgbe_vlan_tpid_set,
522 	.vlan_offload_set     = ixgbe_vlan_offload_set,
523 	.vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
524 	.rx_queue_start	      = ixgbe_dev_rx_queue_start,
525 	.rx_queue_stop        = ixgbe_dev_rx_queue_stop,
526 	.tx_queue_start	      = ixgbe_dev_tx_queue_start,
527 	.tx_queue_stop        = ixgbe_dev_tx_queue_stop,
528 	.rx_queue_setup       = ixgbe_dev_rx_queue_setup,
529 	.rx_queue_intr_enable = ixgbe_dev_rx_queue_intr_enable,
530 	.rx_queue_intr_disable = ixgbe_dev_rx_queue_intr_disable,
531 	.rx_queue_release     = ixgbe_dev_rx_queue_release,
532 	.rx_queue_count       = ixgbe_dev_rx_queue_count,
533 	.rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
534 	.rx_descriptor_status = ixgbe_dev_rx_descriptor_status,
535 	.tx_descriptor_status = ixgbe_dev_tx_descriptor_status,
536 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
537 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
538 	.dev_led_on           = ixgbe_dev_led_on,
539 	.dev_led_off          = ixgbe_dev_led_off,
540 	.flow_ctrl_get        = ixgbe_flow_ctrl_get,
541 	.flow_ctrl_set        = ixgbe_flow_ctrl_set,
542 	.priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
543 	.mac_addr_add         = ixgbe_add_rar,
544 	.mac_addr_remove      = ixgbe_remove_rar,
545 	.mac_addr_set         = ixgbe_set_default_mac_addr,
546 	.uc_hash_table_set    = ixgbe_uc_hash_table_set,
547 	.uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
548 	.mirror_rule_set      = ixgbe_mirror_rule_set,
549 	.mirror_rule_reset    = ixgbe_mirror_rule_reset,
550 	.set_queue_rate_limit = ixgbe_set_queue_rate_limit,
551 	.reta_update          = ixgbe_dev_rss_reta_update,
552 	.reta_query           = ixgbe_dev_rss_reta_query,
553 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
554 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
555 	.filter_ctrl          = ixgbe_dev_filter_ctrl,
556 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
557 	.rxq_info_get         = ixgbe_rxq_info_get,
558 	.txq_info_get         = ixgbe_txq_info_get,
559 	.timesync_enable      = ixgbe_timesync_enable,
560 	.timesync_disable     = ixgbe_timesync_disable,
561 	.timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
562 	.timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
563 	.get_reg              = ixgbe_get_regs,
564 	.get_eeprom_length    = ixgbe_get_eeprom_length,
565 	.get_eeprom           = ixgbe_get_eeprom,
566 	.set_eeprom           = ixgbe_set_eeprom,
567 	.get_dcb_info         = ixgbe_dev_get_dcb_info,
568 	.timesync_adjust_time = ixgbe_timesync_adjust_time,
569 	.timesync_read_time   = ixgbe_timesync_read_time,
570 	.timesync_write_time  = ixgbe_timesync_write_time,
571 	.l2_tunnel_eth_type_conf = ixgbe_dev_l2_tunnel_eth_type_conf,
572 	.l2_tunnel_offload_set   = ixgbe_dev_l2_tunnel_offload_set,
573 	.udp_tunnel_port_add  = ixgbe_dev_udp_tunnel_port_add,
574 	.udp_tunnel_port_del  = ixgbe_dev_udp_tunnel_port_del,
575 	.tm_ops_get           = ixgbe_tm_ops_get,
576 };
577 
578 /*
579  * dev_ops for virtual function, bare necessities for basic vf
580  * operation have been implemented
581  */
582 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
583 	.dev_configure        = ixgbevf_dev_configure,
584 	.dev_start            = ixgbevf_dev_start,
585 	.dev_stop             = ixgbevf_dev_stop,
586 	.link_update          = ixgbevf_dev_link_update,
587 	.stats_get            = ixgbevf_dev_stats_get,
588 	.xstats_get           = ixgbevf_dev_xstats_get,
589 	.stats_reset          = ixgbevf_dev_stats_reset,
590 	.xstats_reset         = ixgbevf_dev_stats_reset,
591 	.xstats_get_names     = ixgbevf_dev_xstats_get_names,
592 	.dev_close            = ixgbevf_dev_close,
593 	.dev_reset	      = ixgbevf_dev_reset,
594 	.allmulticast_enable  = ixgbevf_dev_allmulticast_enable,
595 	.allmulticast_disable = ixgbevf_dev_allmulticast_disable,
596 	.dev_infos_get        = ixgbevf_dev_info_get,
597 	.dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
598 	.mtu_set              = ixgbevf_dev_set_mtu,
599 	.vlan_filter_set      = ixgbevf_vlan_filter_set,
600 	.vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
601 	.vlan_offload_set     = ixgbevf_vlan_offload_set,
602 	.rx_queue_setup       = ixgbe_dev_rx_queue_setup,
603 	.rx_queue_release     = ixgbe_dev_rx_queue_release,
604 	.rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
605 	.rx_descriptor_status = ixgbe_dev_rx_descriptor_status,
606 	.tx_descriptor_status = ixgbe_dev_tx_descriptor_status,
607 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
608 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
609 	.rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
610 	.rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
611 	.mac_addr_add         = ixgbevf_add_mac_addr,
612 	.mac_addr_remove      = ixgbevf_remove_mac_addr,
613 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
614 	.rxq_info_get         = ixgbe_rxq_info_get,
615 	.txq_info_get         = ixgbe_txq_info_get,
616 	.mac_addr_set         = ixgbevf_set_default_mac_addr,
617 	.get_reg              = ixgbevf_get_regs,
618 	.reta_update          = ixgbe_dev_rss_reta_update,
619 	.reta_query           = ixgbe_dev_rss_reta_query,
620 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
621 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
622 };
623 
624 /* store statistics names and its offset in stats structure */
625 struct rte_ixgbe_xstats_name_off {
626 	char name[RTE_ETH_XSTATS_NAME_SIZE];
627 	unsigned offset;
628 };
629 
630 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
631 	{"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
632 	{"rx_illegal_byte_errors", offsetof(struct ixgbe_hw_stats, illerrc)},
633 	{"rx_error_bytes", offsetof(struct ixgbe_hw_stats, errbc)},
634 	{"mac_local_errors", offsetof(struct ixgbe_hw_stats, mlfc)},
635 	{"mac_remote_errors", offsetof(struct ixgbe_hw_stats, mrfc)},
636 	{"rx_length_errors", offsetof(struct ixgbe_hw_stats, rlec)},
637 	{"tx_xon_packets", offsetof(struct ixgbe_hw_stats, lxontxc)},
638 	{"rx_xon_packets", offsetof(struct ixgbe_hw_stats, lxonrxc)},
639 	{"tx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxofftxc)},
640 	{"rx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
641 	{"rx_size_64_packets", offsetof(struct ixgbe_hw_stats, prc64)},
642 	{"rx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, prc127)},
643 	{"rx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, prc255)},
644 	{"rx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, prc511)},
645 	{"rx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
646 		prc1023)},
647 	{"rx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
648 		prc1522)},
649 	{"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
650 	{"rx_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
651 	{"rx_fragment_errors", offsetof(struct ixgbe_hw_stats, rfc)},
652 	{"rx_undersize_errors", offsetof(struct ixgbe_hw_stats, ruc)},
653 	{"rx_oversize_errors", offsetof(struct ixgbe_hw_stats, roc)},
654 	{"rx_jabber_errors", offsetof(struct ixgbe_hw_stats, rjc)},
655 	{"rx_management_packets", offsetof(struct ixgbe_hw_stats, mngprc)},
656 	{"rx_management_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
657 	{"tx_management_packets", offsetof(struct ixgbe_hw_stats, mngptc)},
658 	{"rx_total_packets", offsetof(struct ixgbe_hw_stats, tpr)},
659 	{"rx_total_bytes", offsetof(struct ixgbe_hw_stats, tor)},
660 	{"tx_total_packets", offsetof(struct ixgbe_hw_stats, tpt)},
661 	{"tx_size_64_packets", offsetof(struct ixgbe_hw_stats, ptc64)},
662 	{"tx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, ptc127)},
663 	{"tx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, ptc255)},
664 	{"tx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, ptc511)},
665 	{"tx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
666 		ptc1023)},
667 	{"tx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
668 		ptc1522)},
669 	{"tx_multicast_packets", offsetof(struct ixgbe_hw_stats, mptc)},
670 	{"tx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bptc)},
671 	{"rx_mac_short_packet_dropped", offsetof(struct ixgbe_hw_stats, mspdc)},
672 	{"rx_l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
673 
674 	{"flow_director_added_filters", offsetof(struct ixgbe_hw_stats,
675 		fdirustat_add)},
676 	{"flow_director_removed_filters", offsetof(struct ixgbe_hw_stats,
677 		fdirustat_remove)},
678 	{"flow_director_filter_add_errors", offsetof(struct ixgbe_hw_stats,
679 		fdirfstat_fadd)},
680 	{"flow_director_filter_remove_errors", offsetof(struct ixgbe_hw_stats,
681 		fdirfstat_fremove)},
682 	{"flow_director_matched_filters", offsetof(struct ixgbe_hw_stats,
683 		fdirmatch)},
684 	{"flow_director_missed_filters", offsetof(struct ixgbe_hw_stats,
685 		fdirmiss)},
686 
687 	{"rx_fcoe_crc_errors", offsetof(struct ixgbe_hw_stats, fccrc)},
688 	{"rx_fcoe_dropped", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
689 	{"rx_fcoe_mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats,
690 		fclast)},
691 	{"rx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeprc)},
692 	{"tx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeptc)},
693 	{"rx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwrc)},
694 	{"tx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwtc)},
695 	{"rx_fcoe_no_direct_data_placement", offsetof(struct ixgbe_hw_stats,
696 		fcoe_noddp)},
697 	{"rx_fcoe_no_direct_data_placement_ext_buff",
698 		offsetof(struct ixgbe_hw_stats, fcoe_noddp_ext_buff)},
699 
700 	{"tx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
701 		lxontxc)},
702 	{"rx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
703 		lxonrxc)},
704 	{"tx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
705 		lxofftxc)},
706 	{"rx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
707 		lxoffrxc)},
708 	{"rx_total_missed_packets", offsetof(struct ixgbe_hw_stats, mpctotal)},
709 };
710 
711 #define IXGBE_NB_HW_STATS (sizeof(rte_ixgbe_stats_strings) / \
712 			   sizeof(rte_ixgbe_stats_strings[0]))
713 
714 /* MACsec statistics */
715 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_macsec_strings[] = {
716 	{"out_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
717 		out_pkts_untagged)},
718 	{"out_pkts_encrypted", offsetof(struct ixgbe_macsec_stats,
719 		out_pkts_encrypted)},
720 	{"out_pkts_protected", offsetof(struct ixgbe_macsec_stats,
721 		out_pkts_protected)},
722 	{"out_octets_encrypted", offsetof(struct ixgbe_macsec_stats,
723 		out_octets_encrypted)},
724 	{"out_octets_protected", offsetof(struct ixgbe_macsec_stats,
725 		out_octets_protected)},
726 	{"in_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
727 		in_pkts_untagged)},
728 	{"in_pkts_badtag", offsetof(struct ixgbe_macsec_stats,
729 		in_pkts_badtag)},
730 	{"in_pkts_nosci", offsetof(struct ixgbe_macsec_stats,
731 		in_pkts_nosci)},
732 	{"in_pkts_unknownsci", offsetof(struct ixgbe_macsec_stats,
733 		in_pkts_unknownsci)},
734 	{"in_octets_decrypted", offsetof(struct ixgbe_macsec_stats,
735 		in_octets_decrypted)},
736 	{"in_octets_validated", offsetof(struct ixgbe_macsec_stats,
737 		in_octets_validated)},
738 	{"in_pkts_unchecked", offsetof(struct ixgbe_macsec_stats,
739 		in_pkts_unchecked)},
740 	{"in_pkts_delayed", offsetof(struct ixgbe_macsec_stats,
741 		in_pkts_delayed)},
742 	{"in_pkts_late", offsetof(struct ixgbe_macsec_stats,
743 		in_pkts_late)},
744 	{"in_pkts_ok", offsetof(struct ixgbe_macsec_stats,
745 		in_pkts_ok)},
746 	{"in_pkts_invalid", offsetof(struct ixgbe_macsec_stats,
747 		in_pkts_invalid)},
748 	{"in_pkts_notvalid", offsetof(struct ixgbe_macsec_stats,
749 		in_pkts_notvalid)},
750 	{"in_pkts_unusedsa", offsetof(struct ixgbe_macsec_stats,
751 		in_pkts_unusedsa)},
752 	{"in_pkts_notusingsa", offsetof(struct ixgbe_macsec_stats,
753 		in_pkts_notusingsa)},
754 };
755 
756 #define IXGBE_NB_MACSEC_STATS (sizeof(rte_ixgbe_macsec_strings) / \
757 			   sizeof(rte_ixgbe_macsec_strings[0]))
758 
759 /* Per-queue statistics */
760 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_rxq_strings[] = {
761 	{"mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats, rnbc)},
762 	{"dropped", offsetof(struct ixgbe_hw_stats, mpc)},
763 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxonrxc)},
764 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxoffrxc)},
765 };
766 
767 #define IXGBE_NB_RXQ_PRIO_STATS (sizeof(rte_ixgbe_rxq_strings) / \
768 			   sizeof(rte_ixgbe_rxq_strings[0]))
769 #define IXGBE_NB_RXQ_PRIO_VALUES 8
770 
771 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_txq_strings[] = {
772 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxontxc)},
773 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxofftxc)},
774 	{"xon_to_xoff_packets", offsetof(struct ixgbe_hw_stats,
775 		pxon2offc)},
776 };
777 
778 #define IXGBE_NB_TXQ_PRIO_STATS (sizeof(rte_ixgbe_txq_strings) / \
779 			   sizeof(rte_ixgbe_txq_strings[0]))
780 #define IXGBE_NB_TXQ_PRIO_VALUES 8
781 
782 static const struct rte_ixgbe_xstats_name_off rte_ixgbevf_stats_strings[] = {
783 	{"rx_multicast_packets", offsetof(struct ixgbevf_hw_stats, vfmprc)},
784 };
785 
786 #define IXGBEVF_NB_XSTATS (sizeof(rte_ixgbevf_stats_strings) /	\
787 		sizeof(rte_ixgbevf_stats_strings[0]))
788 
789 /*
790  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
791  */
792 static inline int
793 ixgbe_is_sfp(struct ixgbe_hw *hw)
794 {
795 	switch (hw->phy.type) {
796 	case ixgbe_phy_sfp_avago:
797 	case ixgbe_phy_sfp_ftl:
798 	case ixgbe_phy_sfp_intel:
799 	case ixgbe_phy_sfp_unknown:
800 	case ixgbe_phy_sfp_passive_tyco:
801 	case ixgbe_phy_sfp_passive_unknown:
802 		return 1;
803 	default:
804 		return 0;
805 	}
806 }
807 
808 static inline int32_t
809 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
810 {
811 	uint32_t ctrl_ext;
812 	int32_t status;
813 
814 	status = ixgbe_reset_hw(hw);
815 
816 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
817 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
818 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
819 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
820 	IXGBE_WRITE_FLUSH(hw);
821 
822 	if (status == IXGBE_ERR_SFP_NOT_PRESENT)
823 		status = IXGBE_SUCCESS;
824 	return status;
825 }
826 
827 static inline void
828 ixgbe_enable_intr(struct rte_eth_dev *dev)
829 {
830 	struct ixgbe_interrupt *intr =
831 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
832 	struct ixgbe_hw *hw =
833 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
834 
835 	IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
836 	IXGBE_WRITE_FLUSH(hw);
837 }
838 
839 /*
840  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
841  */
842 static void
843 ixgbe_disable_intr(struct ixgbe_hw *hw)
844 {
845 	PMD_INIT_FUNC_TRACE();
846 
847 	if (hw->mac.type == ixgbe_mac_82598EB) {
848 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
849 	} else {
850 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
851 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
852 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
853 	}
854 	IXGBE_WRITE_FLUSH(hw);
855 }
856 
857 /*
858  * This function resets queue statistics mapping registers.
859  * From Niantic datasheet, Initialization of Statistics section:
860  * "...if software requires the queue counters, the RQSMR and TQSM registers
861  * must be re-programmed following a device reset.
862  */
863 static void
864 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
865 {
866 	uint32_t i;
867 
868 	for (i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
869 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
870 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
871 	}
872 }
873 
874 
875 static int
876 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
877 				  uint16_t queue_id,
878 				  uint8_t stat_idx,
879 				  uint8_t is_rx)
880 {
881 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
882 #define NB_QMAP_FIELDS_PER_QSM_REG 4
883 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
884 
885 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
886 	struct ixgbe_stat_mapping_registers *stat_mappings =
887 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
888 	uint32_t qsmr_mask = 0;
889 	uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
890 	uint32_t q_map;
891 	uint8_t n, offset;
892 
893 	if ((hw->mac.type != ixgbe_mac_82599EB) &&
894 		(hw->mac.type != ixgbe_mac_X540) &&
895 		(hw->mac.type != ixgbe_mac_X550) &&
896 		(hw->mac.type != ixgbe_mac_X550EM_x) &&
897 		(hw->mac.type != ixgbe_mac_X550EM_a))
898 		return -ENOSYS;
899 
900 	PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
901 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
902 		     queue_id, stat_idx);
903 
904 	n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
905 	if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
906 		PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
907 		return -EIO;
908 	}
909 	offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
910 
911 	/* Now clear any previous stat_idx set */
912 	clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
913 	if (!is_rx)
914 		stat_mappings->tqsm[n] &= ~clearing_mask;
915 	else
916 		stat_mappings->rqsmr[n] &= ~clearing_mask;
917 
918 	q_map = (uint32_t)stat_idx;
919 	q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
920 	qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
921 	if (!is_rx)
922 		stat_mappings->tqsm[n] |= qsmr_mask;
923 	else
924 		stat_mappings->rqsmr[n] |= qsmr_mask;
925 
926 	PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
927 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
928 		     queue_id, stat_idx);
929 	PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
930 		     is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
931 
932 	/* Now write the mapping in the appropriate register */
933 	if (is_rx) {
934 		PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
935 			     stat_mappings->rqsmr[n], n);
936 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
937 	} else {
938 		PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
939 			     stat_mappings->tqsm[n], n);
940 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
941 	}
942 	return 0;
943 }
944 
945 static void
946 ixgbe_restore_statistics_mapping(struct rte_eth_dev *dev)
947 {
948 	struct ixgbe_stat_mapping_registers *stat_mappings =
949 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
950 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
951 	int i;
952 
953 	/* write whatever was in stat mapping table to the NIC */
954 	for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
955 		/* rx */
956 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
957 
958 		/* tx */
959 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
960 	}
961 }
962 
963 static void
964 ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config)
965 {
966 	uint8_t i;
967 	struct ixgbe_dcb_tc_config *tc;
968 	uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
969 
970 	dcb_config->num_tcs.pg_tcs = dcb_max_tc;
971 	dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
972 	for (i = 0; i < dcb_max_tc; i++) {
973 		tc = &dcb_config->tc_config[i];
974 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
975 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
976 				 (uint8_t)(100/dcb_max_tc + (i & 1));
977 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
978 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
979 				 (uint8_t)(100/dcb_max_tc + (i & 1));
980 		tc->pfc = ixgbe_dcb_pfc_disabled;
981 	}
982 
983 	/* Initialize default user to priority mapping, UPx->TC0 */
984 	tc = &dcb_config->tc_config[0];
985 	tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
986 	tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
987 	for (i = 0; i < IXGBE_DCB_MAX_BW_GROUP; i++) {
988 		dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
989 		dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
990 	}
991 	dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
992 	dcb_config->pfc_mode_enable = false;
993 	dcb_config->vt_mode = true;
994 	dcb_config->round_robin_enable = false;
995 	/* support all DCB capabilities in 82599 */
996 	dcb_config->support.capabilities = 0xFF;
997 
998 	/*we only support 4 Tcs for X540, X550 */
999 	if (hw->mac.type == ixgbe_mac_X540 ||
1000 		hw->mac.type == ixgbe_mac_X550 ||
1001 		hw->mac.type == ixgbe_mac_X550EM_x ||
1002 		hw->mac.type == ixgbe_mac_X550EM_a) {
1003 		dcb_config->num_tcs.pg_tcs = 4;
1004 		dcb_config->num_tcs.pfc_tcs = 4;
1005 	}
1006 }
1007 
1008 /*
1009  * Ensure that all locks are released before first NVM or PHY access
1010  */
1011 static void
1012 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
1013 {
1014 	uint16_t mask;
1015 
1016 	/*
1017 	 * Phy lock should not fail in this early stage. If this is the case,
1018 	 * it is due to an improper exit of the application.
1019 	 * So force the release of the faulty lock. Release of common lock
1020 	 * is done automatically by swfw_sync function.
1021 	 */
1022 	mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
1023 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1024 		PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
1025 	}
1026 	ixgbe_release_swfw_semaphore(hw, mask);
1027 
1028 	/*
1029 	 * These ones are more tricky since they are common to all ports; but
1030 	 * swfw_sync retries last long enough (1s) to be almost sure that if
1031 	 * lock can not be taken it is due to an improper lock of the
1032 	 * semaphore.
1033 	 */
1034 	mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
1035 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1036 		PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1037 	}
1038 	ixgbe_release_swfw_semaphore(hw, mask);
1039 }
1040 
1041 /*
1042  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
1043  * It returns 0 on success.
1044  */
1045 static int
1046 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev)
1047 {
1048 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1049 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1050 	struct ixgbe_hw *hw =
1051 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1052 	struct ixgbe_vfta *shadow_vfta =
1053 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1054 	struct ixgbe_hwstrip *hwstrip =
1055 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1056 	struct ixgbe_dcb_config *dcb_config =
1057 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
1058 	struct ixgbe_filter_info *filter_info =
1059 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1060 	struct ixgbe_bw_conf *bw_conf =
1061 		IXGBE_DEV_PRIVATE_TO_BW_CONF(eth_dev->data->dev_private);
1062 	uint32_t ctrl_ext;
1063 	uint16_t csum;
1064 	int diag, i;
1065 
1066 	PMD_INIT_FUNC_TRACE();
1067 
1068 	eth_dev->dev_ops = &ixgbe_eth_dev_ops;
1069 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1070 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1071 	eth_dev->tx_pkt_prepare = &ixgbe_prep_pkts;
1072 
1073 	/*
1074 	 * For secondary processes, we don't initialise any further as primary
1075 	 * has already done this work. Only check we don't need a different
1076 	 * RX and TX function.
1077 	 */
1078 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1079 		struct ixgbe_tx_queue *txq;
1080 		/* TX queue function in primary, set by last queue initialized
1081 		 * Tx queue may not initialized by primary process
1082 		 */
1083 		if (eth_dev->data->tx_queues) {
1084 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
1085 			ixgbe_set_tx_function(eth_dev, txq);
1086 		} else {
1087 			/* Use default TX function if we get here */
1088 			PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
1089 				     "Using default TX function.");
1090 		}
1091 
1092 		ixgbe_set_rx_function(eth_dev);
1093 
1094 		return 0;
1095 	}
1096 
1097 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1098 
1099 	/* Vendor and Device ID need to be set before init of shared code */
1100 	hw->device_id = pci_dev->id.device_id;
1101 	hw->vendor_id = pci_dev->id.vendor_id;
1102 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1103 	hw->allow_unsupported_sfp = 1;
1104 
1105 	/* Initialize the shared code (base driver) */
1106 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1107 	diag = ixgbe_bypass_init_shared_code(hw);
1108 #else
1109 	diag = ixgbe_init_shared_code(hw);
1110 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1111 
1112 	if (diag != IXGBE_SUCCESS) {
1113 		PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
1114 		return -EIO;
1115 	}
1116 
1117 	/* pick up the PCI bus settings for reporting later */
1118 	ixgbe_get_bus_info(hw);
1119 
1120 	/* Unlock any pending hardware semaphore */
1121 	ixgbe_swfw_lock_reset(hw);
1122 
1123 #ifdef RTE_LIBRTE_SECURITY
1124 	/* Initialize security_ctx only for primary process*/
1125 	if (ixgbe_ipsec_ctx_create(eth_dev))
1126 		return -ENOMEM;
1127 #endif
1128 
1129 	/* Initialize DCB configuration*/
1130 	memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
1131 	ixgbe_dcb_init(hw, dcb_config);
1132 	/* Get Hardware Flow Control setting */
1133 	hw->fc.requested_mode = ixgbe_fc_full;
1134 	hw->fc.current_mode = ixgbe_fc_full;
1135 	hw->fc.pause_time = IXGBE_FC_PAUSE;
1136 	for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
1137 		hw->fc.low_water[i] = IXGBE_FC_LO;
1138 		hw->fc.high_water[i] = IXGBE_FC_HI;
1139 	}
1140 	hw->fc.send_xon = 1;
1141 
1142 	/* Make sure we have a good EEPROM before we read from it */
1143 	diag = ixgbe_validate_eeprom_checksum(hw, &csum);
1144 	if (diag != IXGBE_SUCCESS) {
1145 		PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
1146 		return -EIO;
1147 	}
1148 
1149 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1150 	diag = ixgbe_bypass_init_hw(hw);
1151 #else
1152 	diag = ixgbe_init_hw(hw);
1153 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1154 
1155 	/*
1156 	 * Devices with copper phys will fail to initialise if ixgbe_init_hw()
1157 	 * is called too soon after the kernel driver unbinding/binding occurs.
1158 	 * The failure occurs in ixgbe_identify_phy_generic() for all devices,
1159 	 * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
1160 	 * also called. See ixgbe_identify_phy_82599(). The reason for the
1161 	 * failure is not known, and only occuts when virtualisation features
1162 	 * are disabled in the bios. A delay of 100ms  was found to be enough by
1163 	 * trial-and-error, and is doubled to be safe.
1164 	 */
1165 	if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
1166 		rte_delay_ms(200);
1167 		diag = ixgbe_init_hw(hw);
1168 	}
1169 
1170 	if (diag == IXGBE_ERR_SFP_NOT_PRESENT)
1171 		diag = IXGBE_SUCCESS;
1172 
1173 	if (diag == IXGBE_ERR_EEPROM_VERSION) {
1174 		PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
1175 			     "LOM.  Please be aware there may be issues associated "
1176 			     "with your hardware.");
1177 		PMD_INIT_LOG(ERR, "If you are experiencing problems "
1178 			     "please contact your Intel or hardware representative "
1179 			     "who provided you with this hardware.");
1180 	} else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
1181 		PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
1182 	if (diag) {
1183 		PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
1184 		return -EIO;
1185 	}
1186 
1187 	/* Reset the hw statistics */
1188 	ixgbe_dev_stats_reset(eth_dev);
1189 
1190 	/* disable interrupt */
1191 	ixgbe_disable_intr(hw);
1192 
1193 	/* reset mappings for queue statistics hw counters*/
1194 	ixgbe_reset_qstat_mappings(hw);
1195 
1196 	/* Allocate memory for storing MAC addresses */
1197 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
1198 					       hw->mac.num_rar_entries, 0);
1199 	if (eth_dev->data->mac_addrs == NULL) {
1200 		PMD_INIT_LOG(ERR,
1201 			     "Failed to allocate %u bytes needed to store "
1202 			     "MAC addresses",
1203 			     ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1204 		return -ENOMEM;
1205 	}
1206 	/* Copy the permanent MAC address */
1207 	ether_addr_copy((struct ether_addr *) hw->mac.perm_addr,
1208 			&eth_dev->data->mac_addrs[0]);
1209 
1210 	/* Allocate memory for storing hash filter MAC addresses */
1211 	eth_dev->data->hash_mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
1212 						    IXGBE_VMDQ_NUM_UC_MAC, 0);
1213 	if (eth_dev->data->hash_mac_addrs == NULL) {
1214 		PMD_INIT_LOG(ERR,
1215 			     "Failed to allocate %d bytes needed to store MAC addresses",
1216 			     ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
1217 		return -ENOMEM;
1218 	}
1219 
1220 	/* initialize the vfta */
1221 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1222 
1223 	/* initialize the hw strip bitmap*/
1224 	memset(hwstrip, 0, sizeof(*hwstrip));
1225 
1226 	/* initialize PF if max_vfs not zero */
1227 	ixgbe_pf_host_init(eth_dev);
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 	ixgbe_fdir_filter_init(eth_dev);
1267 
1268 	/* initialize l2 tunnel filter list & hash */
1269 	ixgbe_l2_tn_filter_init(eth_dev);
1270 
1271 	/* initialize flow filter lists */
1272 	ixgbe_filterlist_init();
1273 
1274 	/* initialize bandwidth configuration info */
1275 	memset(bw_conf, 0, sizeof(struct ixgbe_bw_conf));
1276 
1277 	/* initialize Traffic Manager configuration */
1278 	ixgbe_tm_conf_init(eth_dev);
1279 
1280 	return 0;
1281 }
1282 
1283 static int
1284 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1285 {
1286 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1287 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1288 	struct ixgbe_hw *hw;
1289 	int retries = 0;
1290 	int ret;
1291 
1292 	PMD_INIT_FUNC_TRACE();
1293 
1294 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1295 		return -EPERM;
1296 
1297 	hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1298 
1299 	if (hw->adapter_stopped == 0)
1300 		ixgbe_dev_close(eth_dev);
1301 
1302 	eth_dev->dev_ops = NULL;
1303 	eth_dev->rx_pkt_burst = NULL;
1304 	eth_dev->tx_pkt_burst = NULL;
1305 
1306 	/* Unlock any pending hardware semaphore */
1307 	ixgbe_swfw_lock_reset(hw);
1308 
1309 	/* disable uio intr before callback unregister */
1310 	rte_intr_disable(intr_handle);
1311 
1312 	do {
1313 		ret = rte_intr_callback_unregister(intr_handle,
1314 				ixgbe_dev_interrupt_handler, eth_dev);
1315 		if (ret >= 0) {
1316 			break;
1317 		} else if (ret != -EAGAIN) {
1318 			PMD_INIT_LOG(ERR,
1319 				"intr callback unregister failed: %d",
1320 				ret);
1321 			return ret;
1322 		}
1323 		rte_delay_ms(100);
1324 	} while (retries++ < (10 + IXGBE_LINK_UP_TIME));
1325 
1326 	/* uninitialize PF if max_vfs not zero */
1327 	ixgbe_pf_host_uninit(eth_dev);
1328 
1329 	rte_free(eth_dev->data->mac_addrs);
1330 	eth_dev->data->mac_addrs = NULL;
1331 
1332 	rte_free(eth_dev->data->hash_mac_addrs);
1333 	eth_dev->data->hash_mac_addrs = NULL;
1334 
1335 	/* remove all the fdir filters & hash */
1336 	ixgbe_fdir_filter_uninit(eth_dev);
1337 
1338 	/* remove all the L2 tunnel filters & hash */
1339 	ixgbe_l2_tn_filter_uninit(eth_dev);
1340 
1341 	/* Remove all ntuple filters of the device */
1342 	ixgbe_ntuple_filter_uninit(eth_dev);
1343 
1344 	/* clear all the filters list */
1345 	ixgbe_filterlist_flush();
1346 
1347 	/* Remove all Traffic Manager configuration */
1348 	ixgbe_tm_conf_uninit(eth_dev);
1349 
1350 #ifdef RTE_LIBRTE_SECURITY
1351 	rte_free(eth_dev->security_ctx);
1352 #endif
1353 
1354 	return 0;
1355 }
1356 
1357 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
1358 {
1359 	struct ixgbe_filter_info *filter_info =
1360 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1361 	struct ixgbe_5tuple_filter *p_5tuple;
1362 
1363 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
1364 		TAILQ_REMOVE(&filter_info->fivetuple_list,
1365 			     p_5tuple,
1366 			     entries);
1367 		rte_free(p_5tuple);
1368 	}
1369 	memset(filter_info->fivetuple_mask, 0,
1370 	       sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1371 
1372 	return 0;
1373 }
1374 
1375 static int ixgbe_fdir_filter_uninit(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 	struct ixgbe_fdir_filter *fdir_filter;
1380 
1381 		if (fdir_info->hash_map)
1382 		rte_free(fdir_info->hash_map);
1383 	if (fdir_info->hash_handle)
1384 		rte_hash_free(fdir_info->hash_handle);
1385 
1386 	while ((fdir_filter = TAILQ_FIRST(&fdir_info->fdir_list))) {
1387 		TAILQ_REMOVE(&fdir_info->fdir_list,
1388 			     fdir_filter,
1389 			     entries);
1390 		rte_free(fdir_filter);
1391 	}
1392 
1393 	return 0;
1394 }
1395 
1396 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev)
1397 {
1398 	struct ixgbe_l2_tn_info *l2_tn_info =
1399 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1400 	struct ixgbe_l2_tn_filter *l2_tn_filter;
1401 
1402 	if (l2_tn_info->hash_map)
1403 		rte_free(l2_tn_info->hash_map);
1404 	if (l2_tn_info->hash_handle)
1405 		rte_hash_free(l2_tn_info->hash_handle);
1406 
1407 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
1408 		TAILQ_REMOVE(&l2_tn_info->l2_tn_list,
1409 			     l2_tn_filter,
1410 			     entries);
1411 		rte_free(l2_tn_filter);
1412 	}
1413 
1414 	return 0;
1415 }
1416 
1417 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev)
1418 {
1419 	struct ixgbe_hw_fdir_info *fdir_info =
1420 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1421 	char fdir_hash_name[RTE_HASH_NAMESIZE];
1422 	struct rte_hash_parameters fdir_hash_params = {
1423 		.name = fdir_hash_name,
1424 		.entries = IXGBE_MAX_FDIR_FILTER_NUM,
1425 		.key_len = sizeof(union ixgbe_atr_input),
1426 		.hash_func = rte_hash_crc,
1427 		.hash_func_init_val = 0,
1428 		.socket_id = rte_socket_id(),
1429 	};
1430 
1431 	TAILQ_INIT(&fdir_info->fdir_list);
1432 	snprintf(fdir_hash_name, RTE_HASH_NAMESIZE,
1433 		 "fdir_%s", eth_dev->device->name);
1434 	fdir_info->hash_handle = rte_hash_create(&fdir_hash_params);
1435 	if (!fdir_info->hash_handle) {
1436 		PMD_INIT_LOG(ERR, "Failed to create fdir hash table!");
1437 		return -EINVAL;
1438 	}
1439 	fdir_info->hash_map = rte_zmalloc("ixgbe",
1440 					  sizeof(struct ixgbe_fdir_filter *) *
1441 					  IXGBE_MAX_FDIR_FILTER_NUM,
1442 					  0);
1443 	if (!fdir_info->hash_map) {
1444 		PMD_INIT_LOG(ERR,
1445 			     "Failed to allocate memory for fdir hash map!");
1446 		return -ENOMEM;
1447 	}
1448 	fdir_info->mask_added = FALSE;
1449 
1450 	return 0;
1451 }
1452 
1453 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev)
1454 {
1455 	struct ixgbe_l2_tn_info *l2_tn_info =
1456 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1457 	char l2_tn_hash_name[RTE_HASH_NAMESIZE];
1458 	struct rte_hash_parameters l2_tn_hash_params = {
1459 		.name = l2_tn_hash_name,
1460 		.entries = IXGBE_MAX_L2_TN_FILTER_NUM,
1461 		.key_len = sizeof(struct ixgbe_l2_tn_key),
1462 		.hash_func = rte_hash_crc,
1463 		.hash_func_init_val = 0,
1464 		.socket_id = rte_socket_id(),
1465 	};
1466 
1467 	TAILQ_INIT(&l2_tn_info->l2_tn_list);
1468 	snprintf(l2_tn_hash_name, RTE_HASH_NAMESIZE,
1469 		 "l2_tn_%s", eth_dev->device->name);
1470 	l2_tn_info->hash_handle = rte_hash_create(&l2_tn_hash_params);
1471 	if (!l2_tn_info->hash_handle) {
1472 		PMD_INIT_LOG(ERR, "Failed to create L2 TN hash table!");
1473 		return -EINVAL;
1474 	}
1475 	l2_tn_info->hash_map = rte_zmalloc("ixgbe",
1476 				   sizeof(struct ixgbe_l2_tn_filter *) *
1477 				   IXGBE_MAX_L2_TN_FILTER_NUM,
1478 				   0);
1479 	if (!l2_tn_info->hash_map) {
1480 		PMD_INIT_LOG(ERR,
1481 			"Failed to allocate memory for L2 TN hash map!");
1482 		return -ENOMEM;
1483 	}
1484 	l2_tn_info->e_tag_en = FALSE;
1485 	l2_tn_info->e_tag_fwd_en = FALSE;
1486 	l2_tn_info->e_tag_ether_type = DEFAULT_ETAG_ETYPE;
1487 
1488 	return 0;
1489 }
1490 /*
1491  * Negotiate mailbox API version with the PF.
1492  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1493  * Then we try to negotiate starting with the most recent one.
1494  * If all negotiation attempts fail, then we will proceed with
1495  * the default one (ixgbe_mbox_api_10).
1496  */
1497 static void
1498 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1499 {
1500 	int32_t i;
1501 
1502 	/* start with highest supported, proceed down */
1503 	static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1504 		ixgbe_mbox_api_12,
1505 		ixgbe_mbox_api_11,
1506 		ixgbe_mbox_api_10,
1507 	};
1508 
1509 	for (i = 0;
1510 			i != RTE_DIM(sup_ver) &&
1511 			ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1512 			i++)
1513 		;
1514 }
1515 
1516 static void
1517 generate_random_mac_addr(struct ether_addr *mac_addr)
1518 {
1519 	uint64_t random;
1520 
1521 	/* Set Organizationally Unique Identifier (OUI) prefix. */
1522 	mac_addr->addr_bytes[0] = 0x00;
1523 	mac_addr->addr_bytes[1] = 0x09;
1524 	mac_addr->addr_bytes[2] = 0xC0;
1525 	/* Force indication of locally assigned MAC address. */
1526 	mac_addr->addr_bytes[0] |= ETHER_LOCAL_ADMIN_ADDR;
1527 	/* Generate the last 3 bytes of the MAC address with a random number. */
1528 	random = rte_rand();
1529 	memcpy(&mac_addr->addr_bytes[3], &random, 3);
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 rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1541 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1542 	struct ixgbe_hw *hw =
1543 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1544 	struct ixgbe_vfta *shadow_vfta =
1545 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1546 	struct ixgbe_hwstrip *hwstrip =
1547 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1548 	struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
1549 
1550 	PMD_INIT_FUNC_TRACE();
1551 
1552 	eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1553 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1554 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1555 
1556 	/* for secondary processes, we don't initialise any further as primary
1557 	 * has already done this work. Only check we don't need a different
1558 	 * RX function
1559 	 */
1560 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1561 		struct ixgbe_tx_queue *txq;
1562 		/* TX queue function in primary, set by last queue initialized
1563 		 * Tx queue may not initialized by primary process
1564 		 */
1565 		if (eth_dev->data->tx_queues) {
1566 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1567 			ixgbe_set_tx_function(eth_dev, txq);
1568 		} else {
1569 			/* Use default TX function if we get here */
1570 			PMD_INIT_LOG(NOTICE,
1571 				     "No TX queues configured yet. Using default TX function.");
1572 		}
1573 
1574 		ixgbe_set_rx_function(eth_dev);
1575 
1576 		return 0;
1577 	}
1578 
1579 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1580 
1581 	hw->device_id = pci_dev->id.device_id;
1582 	hw->vendor_id = pci_dev->id.vendor_id;
1583 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1584 
1585 	/* initialize the vfta */
1586 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1587 
1588 	/* initialize the hw strip bitmap*/
1589 	memset(hwstrip, 0, sizeof(*hwstrip));
1590 
1591 	/* Initialize the shared code (base driver) */
1592 	diag = ixgbe_init_shared_code(hw);
1593 	if (diag != IXGBE_SUCCESS) {
1594 		PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1595 		return -EIO;
1596 	}
1597 
1598 	/* init_mailbox_params */
1599 	hw->mbx.ops.init_params(hw);
1600 
1601 	/* Reset the hw statistics */
1602 	ixgbevf_dev_stats_reset(eth_dev);
1603 
1604 	/* Disable the interrupts for VF */
1605 	ixgbevf_intr_disable(hw);
1606 
1607 	hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1608 	diag = hw->mac.ops.reset_hw(hw);
1609 
1610 	/*
1611 	 * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1612 	 * the underlying PF driver has not assigned a MAC address to the VF.
1613 	 * In this case, assign a random MAC address.
1614 	 */
1615 	if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1616 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1617 		return diag;
1618 	}
1619 
1620 	/* negotiate mailbox API version to use with the PF. */
1621 	ixgbevf_negotiate_api(hw);
1622 
1623 	/* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1624 	ixgbevf_get_queues(hw, &tcs, &tc);
1625 
1626 	/* Allocate memory for storing MAC addresses */
1627 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", ETHER_ADDR_LEN *
1628 					       hw->mac.num_rar_entries, 0);
1629 	if (eth_dev->data->mac_addrs == NULL) {
1630 		PMD_INIT_LOG(ERR,
1631 			     "Failed to allocate %u bytes needed to store "
1632 			     "MAC addresses",
1633 			     ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1634 		return -ENOMEM;
1635 	}
1636 
1637 	/* Generate a random MAC address, if none was assigned by PF. */
1638 	if (is_zero_ether_addr(perm_addr)) {
1639 		generate_random_mac_addr(perm_addr);
1640 		diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1641 		if (diag) {
1642 			rte_free(eth_dev->data->mac_addrs);
1643 			eth_dev->data->mac_addrs = NULL;
1644 			return diag;
1645 		}
1646 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1647 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1648 			     "%02x:%02x:%02x:%02x:%02x:%02x",
1649 			     perm_addr->addr_bytes[0],
1650 			     perm_addr->addr_bytes[1],
1651 			     perm_addr->addr_bytes[2],
1652 			     perm_addr->addr_bytes[3],
1653 			     perm_addr->addr_bytes[4],
1654 			     perm_addr->addr_bytes[5]);
1655 	}
1656 
1657 	/* Copy the permanent MAC address */
1658 	ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1659 
1660 	/* reset the hardware with the new settings */
1661 	diag = hw->mac.ops.start_hw(hw);
1662 	switch (diag) {
1663 	case  0:
1664 		break;
1665 
1666 	default:
1667 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1668 		return -EIO;
1669 	}
1670 
1671 	rte_intr_callback_register(intr_handle,
1672 				   ixgbevf_dev_interrupt_handler, eth_dev);
1673 	rte_intr_enable(intr_handle);
1674 	ixgbevf_intr_enable(hw);
1675 
1676 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1677 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1678 		     pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1679 
1680 	return 0;
1681 }
1682 
1683 /* Virtual Function device uninit */
1684 
1685 static int
1686 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1687 {
1688 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1689 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1690 	struct ixgbe_hw *hw;
1691 
1692 	PMD_INIT_FUNC_TRACE();
1693 
1694 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1695 		return -EPERM;
1696 
1697 	hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1698 
1699 	if (hw->adapter_stopped == 0)
1700 		ixgbevf_dev_close(eth_dev);
1701 
1702 	eth_dev->dev_ops = NULL;
1703 	eth_dev->rx_pkt_burst = NULL;
1704 	eth_dev->tx_pkt_burst = NULL;
1705 
1706 	/* Disable the interrupts for VF */
1707 	ixgbevf_intr_disable(hw);
1708 
1709 	rte_free(eth_dev->data->mac_addrs);
1710 	eth_dev->data->mac_addrs = NULL;
1711 
1712 	rte_intr_disable(intr_handle);
1713 	rte_intr_callback_unregister(intr_handle,
1714 				     ixgbevf_dev_interrupt_handler, eth_dev);
1715 
1716 	return 0;
1717 }
1718 
1719 static int eth_ixgbe_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1720 	struct rte_pci_device *pci_dev)
1721 {
1722 	return rte_eth_dev_pci_generic_probe(pci_dev,
1723 		sizeof(struct ixgbe_adapter), eth_ixgbe_dev_init);
1724 }
1725 
1726 static int eth_ixgbe_pci_remove(struct rte_pci_device *pci_dev)
1727 {
1728 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbe_dev_uninit);
1729 }
1730 
1731 static struct rte_pci_driver rte_ixgbe_pmd = {
1732 	.id_table = pci_id_ixgbe_map,
1733 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
1734 		     RTE_PCI_DRV_IOVA_AS_VA,
1735 	.probe = eth_ixgbe_pci_probe,
1736 	.remove = eth_ixgbe_pci_remove,
1737 };
1738 
1739 static int eth_ixgbevf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1740 	struct rte_pci_device *pci_dev)
1741 {
1742 	return rte_eth_dev_pci_generic_probe(pci_dev,
1743 		sizeof(struct ixgbe_adapter), eth_ixgbevf_dev_init);
1744 }
1745 
1746 static int eth_ixgbevf_pci_remove(struct rte_pci_device *pci_dev)
1747 {
1748 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbevf_dev_uninit);
1749 }
1750 
1751 /*
1752  * virtual function driver struct
1753  */
1754 static struct rte_pci_driver rte_ixgbevf_pmd = {
1755 	.id_table = pci_id_ixgbevf_map,
1756 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_IOVA_AS_VA,
1757 	.probe = eth_ixgbevf_pci_probe,
1758 	.remove = eth_ixgbevf_pci_remove,
1759 };
1760 
1761 static int
1762 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1763 {
1764 	struct ixgbe_hw *hw =
1765 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1766 	struct ixgbe_vfta *shadow_vfta =
1767 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1768 	uint32_t vfta;
1769 	uint32_t vid_idx;
1770 	uint32_t vid_bit;
1771 
1772 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1773 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1774 	vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1775 	if (on)
1776 		vfta |= vid_bit;
1777 	else
1778 		vfta &= ~vid_bit;
1779 	IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1780 
1781 	/* update local VFTA copy */
1782 	shadow_vfta->vfta[vid_idx] = vfta;
1783 
1784 	return 0;
1785 }
1786 
1787 static void
1788 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1789 {
1790 	if (on)
1791 		ixgbe_vlan_hw_strip_enable(dev, queue);
1792 	else
1793 		ixgbe_vlan_hw_strip_disable(dev, queue);
1794 }
1795 
1796 static int
1797 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1798 		    enum rte_vlan_type vlan_type,
1799 		    uint16_t tpid)
1800 {
1801 	struct ixgbe_hw *hw =
1802 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1803 	int ret = 0;
1804 	uint32_t reg;
1805 	uint32_t qinq;
1806 
1807 	qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1808 	qinq &= IXGBE_DMATXCTL_GDV;
1809 
1810 	switch (vlan_type) {
1811 	case ETH_VLAN_TYPE_INNER:
1812 		if (qinq) {
1813 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1814 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1815 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1816 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1817 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1818 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1819 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1820 		} else {
1821 			ret = -ENOTSUP;
1822 			PMD_DRV_LOG(ERR, "Inner type is not supported"
1823 				    " by single VLAN");
1824 		}
1825 		break;
1826 	case ETH_VLAN_TYPE_OUTER:
1827 		if (qinq) {
1828 			/* Only the high 16-bits is valid */
1829 			IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1830 					IXGBE_EXVET_VET_EXT_SHIFT);
1831 		} else {
1832 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1833 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1834 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1835 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1836 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1837 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1838 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1839 		}
1840 
1841 		break;
1842 	default:
1843 		ret = -EINVAL;
1844 		PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1845 		break;
1846 	}
1847 
1848 	return ret;
1849 }
1850 
1851 void
1852 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1853 {
1854 	struct ixgbe_hw *hw =
1855 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1856 	uint32_t vlnctrl;
1857 
1858 	PMD_INIT_FUNC_TRACE();
1859 
1860 	/* Filter Table Disable */
1861 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1862 	vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1863 
1864 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1865 }
1866 
1867 void
1868 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1869 {
1870 	struct ixgbe_hw *hw =
1871 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1872 	struct ixgbe_vfta *shadow_vfta =
1873 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1874 	uint32_t vlnctrl;
1875 	uint16_t i;
1876 
1877 	PMD_INIT_FUNC_TRACE();
1878 
1879 	/* Filter Table Enable */
1880 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1881 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1882 	vlnctrl |= IXGBE_VLNCTRL_VFE;
1883 
1884 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1885 
1886 	/* write whatever is in local vfta copy */
1887 	for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1888 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1889 }
1890 
1891 static void
1892 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1893 {
1894 	struct ixgbe_hwstrip *hwstrip =
1895 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1896 	struct ixgbe_rx_queue *rxq;
1897 
1898 	if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
1899 		return;
1900 
1901 	if (on)
1902 		IXGBE_SET_HWSTRIP(hwstrip, queue);
1903 	else
1904 		IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1905 
1906 	if (queue >= dev->data->nb_rx_queues)
1907 		return;
1908 
1909 	rxq = dev->data->rx_queues[queue];
1910 
1911 	if (on)
1912 		rxq->vlan_flags = PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
1913 	else
1914 		rxq->vlan_flags = PKT_RX_VLAN;
1915 }
1916 
1917 static void
1918 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
1919 {
1920 	struct ixgbe_hw *hw =
1921 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1922 	uint32_t ctrl;
1923 
1924 	PMD_INIT_FUNC_TRACE();
1925 
1926 	if (hw->mac.type == ixgbe_mac_82598EB) {
1927 		/* No queue level support */
1928 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1929 		return;
1930 	}
1931 
1932 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1933 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1934 	ctrl &= ~IXGBE_RXDCTL_VME;
1935 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1936 
1937 	/* record those setting for HW strip per queue */
1938 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
1939 }
1940 
1941 static void
1942 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
1943 {
1944 	struct ixgbe_hw *hw =
1945 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1946 	uint32_t ctrl;
1947 
1948 	PMD_INIT_FUNC_TRACE();
1949 
1950 	if (hw->mac.type == ixgbe_mac_82598EB) {
1951 		/* No queue level supported */
1952 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1953 		return;
1954 	}
1955 
1956 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1957 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1958 	ctrl |= IXGBE_RXDCTL_VME;
1959 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1960 
1961 	/* record those setting for HW strip per queue */
1962 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
1963 }
1964 
1965 static void
1966 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
1967 {
1968 	struct ixgbe_hw *hw =
1969 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1970 	uint32_t ctrl;
1971 
1972 	PMD_INIT_FUNC_TRACE();
1973 
1974 	/* DMATXCTRL: Geric Double VLAN Disable */
1975 	ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1976 	ctrl &= ~IXGBE_DMATXCTL_GDV;
1977 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
1978 
1979 	/* CTRL_EXT: Global Double VLAN Disable */
1980 	ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1981 	ctrl &= ~IXGBE_EXTENDED_VLAN;
1982 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
1983 
1984 }
1985 
1986 static void
1987 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
1988 {
1989 	struct ixgbe_hw *hw =
1990 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1991 	uint32_t ctrl;
1992 
1993 	PMD_INIT_FUNC_TRACE();
1994 
1995 	/* DMATXCTRL: Geric Double VLAN Enable */
1996 	ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1997 	ctrl |= IXGBE_DMATXCTL_GDV;
1998 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
1999 
2000 	/* CTRL_EXT: Global Double VLAN Enable */
2001 	ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2002 	ctrl |= IXGBE_EXTENDED_VLAN;
2003 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2004 
2005 	/* Clear pooling mode of PFVTCTL. It's required by X550. */
2006 	if (hw->mac.type == ixgbe_mac_X550 ||
2007 	    hw->mac.type == ixgbe_mac_X550EM_x ||
2008 	    hw->mac.type == ixgbe_mac_X550EM_a) {
2009 		ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
2010 		ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
2011 		IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
2012 	}
2013 
2014 	/*
2015 	 * VET EXT field in the EXVET register = 0x8100 by default
2016 	 * So no need to change. Same to VT field of DMATXCTL register
2017 	 */
2018 }
2019 
2020 void
2021 ixgbe_vlan_hw_strip_config(struct rte_eth_dev *dev)
2022 {
2023 	struct ixgbe_hw *hw =
2024 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2025 	struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
2026 	uint32_t ctrl;
2027 	uint16_t i;
2028 	struct ixgbe_rx_queue *rxq;
2029 	bool on;
2030 
2031 	PMD_INIT_FUNC_TRACE();
2032 
2033 	if (hw->mac.type == ixgbe_mac_82598EB) {
2034 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP) {
2035 			ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2036 			ctrl |= IXGBE_VLNCTRL_VME;
2037 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2038 		} else {
2039 			ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2040 			ctrl &= ~IXGBE_VLNCTRL_VME;
2041 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2042 		}
2043 	} else {
2044 		/*
2045 		 * Other 10G NIC, the VLAN strip can be setup
2046 		 * per queue in RXDCTL
2047 		 */
2048 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
2049 			rxq = dev->data->rx_queues[i];
2050 			ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2051 			if (rxq->offloads & DEV_RX_OFFLOAD_VLAN_STRIP) {
2052 				ctrl |= IXGBE_RXDCTL_VME;
2053 				on = TRUE;
2054 			} else {
2055 				ctrl &= ~IXGBE_RXDCTL_VME;
2056 				on = FALSE;
2057 			}
2058 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2059 
2060 			/* record those setting for HW strip per queue */
2061 			ixgbe_vlan_hw_strip_bitmap_set(dev, i, on);
2062 		}
2063 	}
2064 }
2065 
2066 static int
2067 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2068 {
2069 	struct rte_eth_rxmode *rxmode;
2070 	rxmode = &dev->data->dev_conf.rxmode;
2071 
2072 	if (mask & ETH_VLAN_STRIP_MASK) {
2073 		ixgbe_vlan_hw_strip_config(dev);
2074 	}
2075 
2076 	if (mask & ETH_VLAN_FILTER_MASK) {
2077 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2078 			ixgbe_vlan_hw_filter_enable(dev);
2079 		else
2080 			ixgbe_vlan_hw_filter_disable(dev);
2081 	}
2082 
2083 	if (mask & ETH_VLAN_EXTEND_MASK) {
2084 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2085 			ixgbe_vlan_hw_extend_enable(dev);
2086 		else
2087 			ixgbe_vlan_hw_extend_disable(dev);
2088 	}
2089 
2090 	return 0;
2091 }
2092 
2093 static void
2094 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2095 {
2096 	struct ixgbe_hw *hw =
2097 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2098 	/* VLNCTRL: enable vlan filtering and allow all vlan tags through */
2099 	uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2100 
2101 	vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
2102 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
2103 }
2104 
2105 static int
2106 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
2107 {
2108 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2109 
2110 	switch (nb_rx_q) {
2111 	case 1:
2112 	case 2:
2113 		RTE_ETH_DEV_SRIOV(dev).active = ETH_64_POOLS;
2114 		break;
2115 	case 4:
2116 		RTE_ETH_DEV_SRIOV(dev).active = ETH_32_POOLS;
2117 		break;
2118 	default:
2119 		return -EINVAL;
2120 	}
2121 
2122 	RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
2123 		IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2124 	RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
2125 		pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2126 	return 0;
2127 }
2128 
2129 static int
2130 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
2131 {
2132 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
2133 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2134 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
2135 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
2136 
2137 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
2138 		/* check multi-queue mode */
2139 		switch (dev_conf->rxmode.mq_mode) {
2140 		case ETH_MQ_RX_VMDQ_DCB:
2141 			PMD_INIT_LOG(INFO, "ETH_MQ_RX_VMDQ_DCB mode supported in SRIOV");
2142 			break;
2143 		case ETH_MQ_RX_VMDQ_DCB_RSS:
2144 			/* DCB/RSS VMDQ in SRIOV mode, not implement yet */
2145 			PMD_INIT_LOG(ERR, "SRIOV active,"
2146 					" unsupported mq_mode rx %d.",
2147 					dev_conf->rxmode.mq_mode);
2148 			return -EINVAL;
2149 		case ETH_MQ_RX_RSS:
2150 		case ETH_MQ_RX_VMDQ_RSS:
2151 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_RSS;
2152 			if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
2153 				if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
2154 					PMD_INIT_LOG(ERR, "SRIOV is active,"
2155 						" invalid queue number"
2156 						" for VMDQ RSS, allowed"
2157 						" value are 1, 2 or 4.");
2158 					return -EINVAL;
2159 				}
2160 			break;
2161 		case ETH_MQ_RX_VMDQ_ONLY:
2162 		case ETH_MQ_RX_NONE:
2163 			/* if nothing mq mode configure, use default scheme */
2164 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
2165 			break;
2166 		default: /* ETH_MQ_RX_DCB, ETH_MQ_RX_DCB_RSS or ETH_MQ_TX_DCB*/
2167 			/* SRIOV only works in VMDq enable mode */
2168 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2169 					" wrong mq_mode rx %d.",
2170 					dev_conf->rxmode.mq_mode);
2171 			return -EINVAL;
2172 		}
2173 
2174 		switch (dev_conf->txmode.mq_mode) {
2175 		case ETH_MQ_TX_VMDQ_DCB:
2176 			PMD_INIT_LOG(INFO, "ETH_MQ_TX_VMDQ_DCB mode supported in SRIOV");
2177 			dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_DCB;
2178 			break;
2179 		default: /* ETH_MQ_TX_VMDQ_ONLY or ETH_MQ_TX_NONE */
2180 			dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_ONLY;
2181 			break;
2182 		}
2183 
2184 		/* check valid queue number */
2185 		if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2186 		    (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2187 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2188 					" nb_rx_q=%d nb_tx_q=%d queue number"
2189 					" must be less than or equal to %d.",
2190 					nb_rx_q, nb_tx_q,
2191 					RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2192 			return -EINVAL;
2193 		}
2194 	} else {
2195 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB_RSS) {
2196 			PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2197 					  " not supported.");
2198 			return -EINVAL;
2199 		}
2200 		/* check configuration for vmdb+dcb mode */
2201 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB) {
2202 			const struct rte_eth_vmdq_dcb_conf *conf;
2203 
2204 			if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2205 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2206 						IXGBE_VMDQ_DCB_NB_QUEUES);
2207 				return -EINVAL;
2208 			}
2209 			conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2210 			if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2211 			       conf->nb_queue_pools == ETH_32_POOLS)) {
2212 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2213 						" nb_queue_pools must be %d or %d.",
2214 						ETH_16_POOLS, ETH_32_POOLS);
2215 				return -EINVAL;
2216 			}
2217 		}
2218 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_VMDQ_DCB) {
2219 			const struct rte_eth_vmdq_dcb_tx_conf *conf;
2220 
2221 			if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2222 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2223 						 IXGBE_VMDQ_DCB_NB_QUEUES);
2224 				return -EINVAL;
2225 			}
2226 			conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2227 			if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2228 			       conf->nb_queue_pools == ETH_32_POOLS)) {
2229 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2230 						" nb_queue_pools != %d and"
2231 						" nb_queue_pools != %d.",
2232 						ETH_16_POOLS, ETH_32_POOLS);
2233 				return -EINVAL;
2234 			}
2235 		}
2236 
2237 		/* For DCB mode check our configuration before we go further */
2238 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB) {
2239 			const struct rte_eth_dcb_rx_conf *conf;
2240 
2241 			if (nb_rx_q != IXGBE_DCB_NB_QUEUES) {
2242 				PMD_INIT_LOG(ERR, "DCB selected, nb_rx_q != %d.",
2243 						 IXGBE_DCB_NB_QUEUES);
2244 				return -EINVAL;
2245 			}
2246 			conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2247 			if (!(conf->nb_tcs == ETH_4_TCS ||
2248 			       conf->nb_tcs == ETH_8_TCS)) {
2249 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2250 						" and nb_tcs != %d.",
2251 						ETH_4_TCS, ETH_8_TCS);
2252 				return -EINVAL;
2253 			}
2254 		}
2255 
2256 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_DCB) {
2257 			const struct rte_eth_dcb_tx_conf *conf;
2258 
2259 			if (nb_tx_q != IXGBE_DCB_NB_QUEUES) {
2260 				PMD_INIT_LOG(ERR, "DCB, nb_tx_q != %d.",
2261 						 IXGBE_DCB_NB_QUEUES);
2262 				return -EINVAL;
2263 			}
2264 			conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2265 			if (!(conf->nb_tcs == ETH_4_TCS ||
2266 			       conf->nb_tcs == ETH_8_TCS)) {
2267 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2268 						" and nb_tcs != %d.",
2269 						ETH_4_TCS, ETH_8_TCS);
2270 				return -EINVAL;
2271 			}
2272 		}
2273 
2274 		/*
2275 		 * When DCB/VT is off, maximum number of queues changes,
2276 		 * except for 82598EB, which remains constant.
2277 		 */
2278 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
2279 				hw->mac.type != ixgbe_mac_82598EB) {
2280 			if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2281 				PMD_INIT_LOG(ERR,
2282 					     "Neither VT nor DCB are enabled, "
2283 					     "nb_tx_q > %d.",
2284 					     IXGBE_NONE_MODE_TX_NB_QUEUES);
2285 				return -EINVAL;
2286 			}
2287 		}
2288 	}
2289 	return 0;
2290 }
2291 
2292 static int
2293 ixgbe_dev_configure(struct rte_eth_dev *dev)
2294 {
2295 	struct ixgbe_interrupt *intr =
2296 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2297 	struct ixgbe_adapter *adapter =
2298 		(struct ixgbe_adapter *)dev->data->dev_private;
2299 	struct rte_eth_dev_info dev_info;
2300 	uint64_t rx_offloads;
2301 	uint64_t tx_offloads;
2302 	int ret;
2303 
2304 	PMD_INIT_FUNC_TRACE();
2305 	/* multipe queue mode checking */
2306 	ret  = ixgbe_check_mq_mode(dev);
2307 	if (ret != 0) {
2308 		PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2309 			    ret);
2310 		return ret;
2311 	}
2312 
2313 	ixgbe_dev_info_get(dev, &dev_info);
2314 	rx_offloads = dev->data->dev_conf.rxmode.offloads;
2315 	if ((rx_offloads & dev_info.rx_offload_capa) != rx_offloads) {
2316 		PMD_DRV_LOG(ERR, "Some Rx offloads are not supported "
2317 			    "requested 0x%" PRIx64 " supported 0x%" PRIx64,
2318 			    rx_offloads, dev_info.rx_offload_capa);
2319 		return -ENOTSUP;
2320 	}
2321 	tx_offloads = dev->data->dev_conf.txmode.offloads;
2322 	if ((tx_offloads & dev_info.tx_offload_capa) != tx_offloads) {
2323 		PMD_DRV_LOG(ERR, "Some Tx offloads are not supported "
2324 			    "requested 0x%" PRIx64 " supported 0x%" PRIx64,
2325 			    tx_offloads, dev_info.tx_offload_capa);
2326 		return -ENOTSUP;
2327 	}
2328 
2329 	/* set flag to update link status after init */
2330 	intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2331 
2332 	/*
2333 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2334 	 * allocation or vector Rx preconditions we will reset it.
2335 	 */
2336 	adapter->rx_bulk_alloc_allowed = true;
2337 	adapter->rx_vec_allowed = true;
2338 
2339 	return 0;
2340 }
2341 
2342 static void
2343 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2344 {
2345 	struct ixgbe_hw *hw =
2346 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2347 	struct ixgbe_interrupt *intr =
2348 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2349 	uint32_t gpie;
2350 
2351 	/* only set up it on X550EM_X */
2352 	if (hw->mac.type == ixgbe_mac_X550EM_x) {
2353 		gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2354 		gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2355 		IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2356 		if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2357 			intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2358 	}
2359 }
2360 
2361 int
2362 ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
2363 			uint16_t tx_rate, uint64_t q_msk)
2364 {
2365 	struct ixgbe_hw *hw;
2366 	struct ixgbe_vf_info *vfinfo;
2367 	struct rte_eth_link link;
2368 	uint8_t  nb_q_per_pool;
2369 	uint32_t queue_stride;
2370 	uint32_t queue_idx, idx = 0, vf_idx;
2371 	uint32_t queue_end;
2372 	uint16_t total_rate = 0;
2373 	struct rte_pci_device *pci_dev;
2374 
2375 	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2376 	rte_eth_link_get_nowait(dev->data->port_id, &link);
2377 
2378 	if (vf >= pci_dev->max_vfs)
2379 		return -EINVAL;
2380 
2381 	if (tx_rate > link.link_speed)
2382 		return -EINVAL;
2383 
2384 	if (q_msk == 0)
2385 		return 0;
2386 
2387 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2388 	vfinfo = *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
2389 	nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2390 	queue_stride = IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2391 	queue_idx = vf * queue_stride;
2392 	queue_end = queue_idx + nb_q_per_pool - 1;
2393 	if (queue_end >= hw->mac.max_tx_queues)
2394 		return -EINVAL;
2395 
2396 	if (vfinfo) {
2397 		for (vf_idx = 0; vf_idx < pci_dev->max_vfs; vf_idx++) {
2398 			if (vf_idx == vf)
2399 				continue;
2400 			for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
2401 				idx++)
2402 				total_rate += vfinfo[vf_idx].tx_rate[idx];
2403 		}
2404 	} else {
2405 		return -EINVAL;
2406 	}
2407 
2408 	/* Store tx_rate for this vf. */
2409 	for (idx = 0; idx < nb_q_per_pool; idx++) {
2410 		if (((uint64_t)0x1 << idx) & q_msk) {
2411 			if (vfinfo[vf].tx_rate[idx] != tx_rate)
2412 				vfinfo[vf].tx_rate[idx] = tx_rate;
2413 			total_rate += tx_rate;
2414 		}
2415 	}
2416 
2417 	if (total_rate > dev->data->dev_link.link_speed) {
2418 		/* Reset stored TX rate of the VF if it causes exceed
2419 		 * link speed.
2420 		 */
2421 		memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
2422 		return -EINVAL;
2423 	}
2424 
2425 	/* Set RTTBCNRC of each queue/pool for vf X  */
2426 	for (; queue_idx <= queue_end; queue_idx++) {
2427 		if (0x1 & q_msk)
2428 			ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
2429 		q_msk = q_msk >> 1;
2430 	}
2431 
2432 	return 0;
2433 }
2434 
2435 /*
2436  * Configure device link speed and setup link.
2437  * It returns 0 on success.
2438  */
2439 static int
2440 ixgbe_dev_start(struct rte_eth_dev *dev)
2441 {
2442 	struct ixgbe_hw *hw =
2443 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2444 	struct ixgbe_vf_info *vfinfo =
2445 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2446 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2447 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2448 	uint32_t intr_vector = 0;
2449 	int err, link_up = 0, negotiate = 0;
2450 	uint32_t speed = 0;
2451 	int mask = 0;
2452 	int status;
2453 	uint16_t vf, idx;
2454 	uint32_t *link_speeds;
2455 	struct ixgbe_tm_conf *tm_conf =
2456 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2457 
2458 	PMD_INIT_FUNC_TRACE();
2459 
2460 	/* IXGBE devices don't support:
2461 	*    - half duplex (checked afterwards for valid speeds)
2462 	*    - fixed speed: TODO implement
2463 	*/
2464 	if (dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED) {
2465 		PMD_INIT_LOG(ERR,
2466 		"Invalid link_speeds for port %u, fix speed not supported",
2467 				dev->data->port_id);
2468 		return -EINVAL;
2469 	}
2470 
2471 	/* disable uio/vfio intr/eventfd mapping */
2472 	rte_intr_disable(intr_handle);
2473 
2474 	/* stop adapter */
2475 	hw->adapter_stopped = 0;
2476 	ixgbe_stop_adapter(hw);
2477 
2478 	/* reinitialize adapter
2479 	 * this calls reset and start
2480 	 */
2481 	status = ixgbe_pf_reset_hw(hw);
2482 	if (status != 0)
2483 		return -1;
2484 	hw->mac.ops.start_hw(hw);
2485 	hw->mac.get_link_status = true;
2486 
2487 	/* configure PF module if SRIOV enabled */
2488 	ixgbe_pf_host_configure(dev);
2489 
2490 	ixgbe_dev_phy_intr_setup(dev);
2491 
2492 	/* check and configure queue intr-vector mapping */
2493 	if ((rte_intr_cap_multiple(intr_handle) ||
2494 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
2495 	    dev->data->dev_conf.intr_conf.rxq != 0) {
2496 		intr_vector = dev->data->nb_rx_queues;
2497 		if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2498 			PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2499 					IXGBE_MAX_INTR_QUEUE_NUM);
2500 			return -ENOTSUP;
2501 		}
2502 		if (rte_intr_efd_enable(intr_handle, intr_vector))
2503 			return -1;
2504 	}
2505 
2506 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
2507 		intr_handle->intr_vec =
2508 			rte_zmalloc("intr_vec",
2509 				    dev->data->nb_rx_queues * sizeof(int), 0);
2510 		if (intr_handle->intr_vec == NULL) {
2511 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2512 				     " intr_vec", dev->data->nb_rx_queues);
2513 			return -ENOMEM;
2514 		}
2515 	}
2516 
2517 	/* confiugre msix for sleep until rx interrupt */
2518 	ixgbe_configure_msix(dev);
2519 
2520 	/* initialize transmission unit */
2521 	ixgbe_dev_tx_init(dev);
2522 
2523 	/* This can fail when allocating mbufs for descriptor rings */
2524 	err = ixgbe_dev_rx_init(dev);
2525 	if (err) {
2526 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2527 		goto error;
2528 	}
2529 
2530 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
2531 		ETH_VLAN_EXTEND_MASK;
2532 	err = ixgbe_vlan_offload_set(dev, mask);
2533 	if (err) {
2534 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload");
2535 		goto error;
2536 	}
2537 
2538 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
2539 		/* Enable vlan filtering for VMDq */
2540 		ixgbe_vmdq_vlan_hw_filter_enable(dev);
2541 	}
2542 
2543 	/* Configure DCB hw */
2544 	ixgbe_configure_dcb(dev);
2545 
2546 	if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2547 		err = ixgbe_fdir_configure(dev);
2548 		if (err)
2549 			goto error;
2550 	}
2551 
2552 	/* Restore vf rate limit */
2553 	if (vfinfo != NULL) {
2554 		for (vf = 0; vf < pci_dev->max_vfs; vf++)
2555 			for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2556 				if (vfinfo[vf].tx_rate[idx] != 0)
2557 					ixgbe_set_vf_rate_limit(
2558 						dev, vf,
2559 						vfinfo[vf].tx_rate[idx],
2560 						1 << idx);
2561 	}
2562 
2563 	ixgbe_restore_statistics_mapping(dev);
2564 
2565 	err = ixgbe_dev_rxtx_start(dev);
2566 	if (err < 0) {
2567 		PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2568 		goto error;
2569 	}
2570 
2571 	/* Skip link setup if loopback mode is enabled for 82599. */
2572 	if (hw->mac.type == ixgbe_mac_82599EB &&
2573 			dev->data->dev_conf.lpbk_mode == IXGBE_LPBK_82599_TX_RX)
2574 		goto skip_link_setup;
2575 
2576 	if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2577 		err = hw->mac.ops.setup_sfp(hw);
2578 		if (err)
2579 			goto error;
2580 	}
2581 
2582 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2583 		/* Turn on the copper */
2584 		ixgbe_set_phy_power(hw, true);
2585 	} else {
2586 		/* Turn on the laser */
2587 		ixgbe_enable_tx_laser(hw);
2588 	}
2589 
2590 	err = ixgbe_check_link(hw, &speed, &link_up, 0);
2591 	if (err)
2592 		goto error;
2593 	dev->data->dev_link.link_status = link_up;
2594 
2595 	err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2596 	if (err)
2597 		goto error;
2598 
2599 	link_speeds = &dev->data->dev_conf.link_speeds;
2600 	if (*link_speeds & ~(ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2601 			ETH_LINK_SPEED_10G)) {
2602 		PMD_INIT_LOG(ERR, "Invalid link setting");
2603 		goto error;
2604 	}
2605 
2606 	speed = 0x0;
2607 	if (*link_speeds == ETH_LINK_SPEED_AUTONEG) {
2608 		switch (hw->mac.type) {
2609 		case ixgbe_mac_82598EB:
2610 			speed = IXGBE_LINK_SPEED_82598_AUTONEG;
2611 			break;
2612 		case ixgbe_mac_82599EB:
2613 		case ixgbe_mac_X540:
2614 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2615 			break;
2616 		case ixgbe_mac_X550:
2617 		case ixgbe_mac_X550EM_x:
2618 		case ixgbe_mac_X550EM_a:
2619 			speed = IXGBE_LINK_SPEED_X550_AUTONEG;
2620 			break;
2621 		default:
2622 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2623 		}
2624 	} else {
2625 		if (*link_speeds & ETH_LINK_SPEED_10G)
2626 			speed |= IXGBE_LINK_SPEED_10GB_FULL;
2627 		if (*link_speeds & ETH_LINK_SPEED_1G)
2628 			speed |= IXGBE_LINK_SPEED_1GB_FULL;
2629 		if (*link_speeds & ETH_LINK_SPEED_100M)
2630 			speed |= IXGBE_LINK_SPEED_100_FULL;
2631 	}
2632 
2633 	err = ixgbe_setup_link(hw, speed, link_up);
2634 	if (err)
2635 		goto error;
2636 
2637 skip_link_setup:
2638 
2639 	if (rte_intr_allow_others(intr_handle)) {
2640 		/* check if lsc interrupt is enabled */
2641 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2642 			ixgbe_dev_lsc_interrupt_setup(dev, TRUE);
2643 		else
2644 			ixgbe_dev_lsc_interrupt_setup(dev, FALSE);
2645 		ixgbe_dev_macsec_interrupt_setup(dev);
2646 	} else {
2647 		rte_intr_callback_unregister(intr_handle,
2648 					     ixgbe_dev_interrupt_handler, dev);
2649 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2650 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
2651 				     " no intr multiplex");
2652 	}
2653 
2654 	/* check if rxq interrupt is enabled */
2655 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2656 	    rte_intr_dp_is_en(intr_handle))
2657 		ixgbe_dev_rxq_interrupt_setup(dev);
2658 
2659 	/* enable uio/vfio intr/eventfd mapping */
2660 	rte_intr_enable(intr_handle);
2661 
2662 	/* resume enabled intr since hw reset */
2663 	ixgbe_enable_intr(dev);
2664 	ixgbe_l2_tunnel_conf(dev);
2665 	ixgbe_filter_restore(dev);
2666 
2667 	if (tm_conf->root && !tm_conf->committed)
2668 		PMD_DRV_LOG(WARNING,
2669 			    "please call hierarchy_commit() "
2670 			    "before starting the port");
2671 
2672 	return 0;
2673 
2674 error:
2675 	PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2676 	ixgbe_dev_clear_queues(dev);
2677 	return -EIO;
2678 }
2679 
2680 /*
2681  * Stop device: disable rx and tx functions to allow for reconfiguring.
2682  */
2683 static void
2684 ixgbe_dev_stop(struct rte_eth_dev *dev)
2685 {
2686 	struct rte_eth_link link;
2687 	struct ixgbe_hw *hw =
2688 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2689 	struct ixgbe_vf_info *vfinfo =
2690 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2691 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2692 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2693 	int vf;
2694 	struct ixgbe_tm_conf *tm_conf =
2695 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2696 
2697 	PMD_INIT_FUNC_TRACE();
2698 
2699 	/* disable interrupts */
2700 	ixgbe_disable_intr(hw);
2701 
2702 	/* reset the NIC */
2703 	ixgbe_pf_reset_hw(hw);
2704 	hw->adapter_stopped = 0;
2705 
2706 	/* stop adapter */
2707 	ixgbe_stop_adapter(hw);
2708 
2709 	for (vf = 0; vfinfo != NULL && vf < pci_dev->max_vfs; vf++)
2710 		vfinfo[vf].clear_to_send = false;
2711 
2712 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2713 		/* Turn off the copper */
2714 		ixgbe_set_phy_power(hw, false);
2715 	} else {
2716 		/* Turn off the laser */
2717 		ixgbe_disable_tx_laser(hw);
2718 	}
2719 
2720 	ixgbe_dev_clear_queues(dev);
2721 
2722 	/* Clear stored conf */
2723 	dev->data->scattered_rx = 0;
2724 	dev->data->lro = 0;
2725 
2726 	/* Clear recorded link status */
2727 	memset(&link, 0, sizeof(link));
2728 	rte_eth_linkstatus_set(dev, &link);
2729 
2730 	if (!rte_intr_allow_others(intr_handle))
2731 		/* resume to the default handler */
2732 		rte_intr_callback_register(intr_handle,
2733 					   ixgbe_dev_interrupt_handler,
2734 					   (void *)dev);
2735 
2736 	/* Clean datapath event and queue/vec mapping */
2737 	rte_intr_efd_disable(intr_handle);
2738 	if (intr_handle->intr_vec != NULL) {
2739 		rte_free(intr_handle->intr_vec);
2740 		intr_handle->intr_vec = NULL;
2741 	}
2742 
2743 	/* reset hierarchy commit */
2744 	tm_conf->committed = false;
2745 }
2746 
2747 /*
2748  * Set device link up: enable tx.
2749  */
2750 static int
2751 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2752 {
2753 	struct ixgbe_hw *hw =
2754 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2755 	if (hw->mac.type == ixgbe_mac_82599EB) {
2756 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2757 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2758 			/* Not suported in bypass mode */
2759 			PMD_INIT_LOG(ERR, "Set link up is not supported "
2760 				     "by device id 0x%x", hw->device_id);
2761 			return -ENOTSUP;
2762 		}
2763 #endif
2764 	}
2765 
2766 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2767 		/* Turn on the copper */
2768 		ixgbe_set_phy_power(hw, true);
2769 	} else {
2770 		/* Turn on the laser */
2771 		ixgbe_enable_tx_laser(hw);
2772 	}
2773 
2774 	return 0;
2775 }
2776 
2777 /*
2778  * Set device link down: disable tx.
2779  */
2780 static int
2781 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2782 {
2783 	struct ixgbe_hw *hw =
2784 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2785 	if (hw->mac.type == ixgbe_mac_82599EB) {
2786 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2787 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2788 			/* Not suported in bypass mode */
2789 			PMD_INIT_LOG(ERR, "Set link down is not supported "
2790 				     "by device id 0x%x", hw->device_id);
2791 			return -ENOTSUP;
2792 		}
2793 #endif
2794 	}
2795 
2796 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2797 		/* Turn off the copper */
2798 		ixgbe_set_phy_power(hw, false);
2799 	} else {
2800 		/* Turn off the laser */
2801 		ixgbe_disable_tx_laser(hw);
2802 	}
2803 
2804 	return 0;
2805 }
2806 
2807 /*
2808  * Reset and stop device.
2809  */
2810 static void
2811 ixgbe_dev_close(struct rte_eth_dev *dev)
2812 {
2813 	struct ixgbe_hw *hw =
2814 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2815 
2816 	PMD_INIT_FUNC_TRACE();
2817 
2818 	ixgbe_pf_reset_hw(hw);
2819 
2820 	ixgbe_dev_stop(dev);
2821 	hw->adapter_stopped = 1;
2822 
2823 	ixgbe_dev_free_queues(dev);
2824 
2825 	ixgbe_disable_pcie_master(hw);
2826 
2827 	/* reprogram the RAR[0] in case user changed it. */
2828 	ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
2829 }
2830 
2831 /*
2832  * Reset PF device.
2833  */
2834 static int
2835 ixgbe_dev_reset(struct rte_eth_dev *dev)
2836 {
2837 	int ret;
2838 
2839 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
2840 	 * its VF to make them align with it. The detailed notification
2841 	 * mechanism is PMD specific. As to ixgbe PF, it is rather complex.
2842 	 * To avoid unexpected behavior in VF, currently reset of PF with
2843 	 * SR-IOV activation is not supported. It might be supported later.
2844 	 */
2845 	if (dev->data->sriov.active)
2846 		return -ENOTSUP;
2847 
2848 	ret = eth_ixgbe_dev_uninit(dev);
2849 	if (ret)
2850 		return ret;
2851 
2852 	ret = eth_ixgbe_dev_init(dev);
2853 
2854 	return ret;
2855 }
2856 
2857 static void
2858 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
2859 			   struct ixgbe_hw_stats *hw_stats,
2860 			   struct ixgbe_macsec_stats *macsec_stats,
2861 			   uint64_t *total_missed_rx, uint64_t *total_qbrc,
2862 			   uint64_t *total_qprc, uint64_t *total_qprdc)
2863 {
2864 	uint32_t bprc, lxon, lxoff, total;
2865 	uint32_t delta_gprc = 0;
2866 	unsigned i;
2867 	/* Workaround for RX byte count not including CRC bytes when CRC
2868 	 * strip is enabled. CRC bytes are removed from counters when crc_strip
2869 	 * is disabled.
2870 	 */
2871 	int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
2872 			IXGBE_HLREG0_RXCRCSTRP);
2873 
2874 	hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
2875 	hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
2876 	hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
2877 	hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
2878 
2879 	for (i = 0; i < 8; i++) {
2880 		uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
2881 
2882 		/* global total per queue */
2883 		hw_stats->mpc[i] += mp;
2884 		/* Running comprehensive total for stats display */
2885 		*total_missed_rx += hw_stats->mpc[i];
2886 		if (hw->mac.type == ixgbe_mac_82598EB) {
2887 			hw_stats->rnbc[i] +=
2888 			    IXGBE_READ_REG(hw, IXGBE_RNBC(i));
2889 			hw_stats->pxonrxc[i] +=
2890 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
2891 			hw_stats->pxoffrxc[i] +=
2892 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
2893 		} else {
2894 			hw_stats->pxonrxc[i] +=
2895 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
2896 			hw_stats->pxoffrxc[i] +=
2897 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
2898 			hw_stats->pxon2offc[i] +=
2899 				IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
2900 		}
2901 		hw_stats->pxontxc[i] +=
2902 		    IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
2903 		hw_stats->pxofftxc[i] +=
2904 		    IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
2905 	}
2906 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
2907 		uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
2908 		uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
2909 		uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
2910 
2911 		delta_gprc += delta_qprc;
2912 
2913 		hw_stats->qprc[i] += delta_qprc;
2914 		hw_stats->qptc[i] += delta_qptc;
2915 
2916 		hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
2917 		hw_stats->qbrc[i] +=
2918 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
2919 		if (crc_strip == 0)
2920 			hw_stats->qbrc[i] -= delta_qprc * ETHER_CRC_LEN;
2921 
2922 		hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
2923 		hw_stats->qbtc[i] +=
2924 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
2925 
2926 		hw_stats->qprdc[i] += delta_qprdc;
2927 		*total_qprdc += hw_stats->qprdc[i];
2928 
2929 		*total_qprc += hw_stats->qprc[i];
2930 		*total_qbrc += hw_stats->qbrc[i];
2931 	}
2932 	hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
2933 	hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
2934 	hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
2935 
2936 	/*
2937 	 * An errata states that gprc actually counts good + missed packets:
2938 	 * Workaround to set gprc to summated queue packet receives
2939 	 */
2940 	hw_stats->gprc = *total_qprc;
2941 
2942 	if (hw->mac.type != ixgbe_mac_82598EB) {
2943 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
2944 		hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
2945 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
2946 		hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
2947 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
2948 		hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
2949 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
2950 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
2951 	} else {
2952 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
2953 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
2954 		/* 82598 only has a counter in the high register */
2955 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
2956 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
2957 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
2958 	}
2959 	uint64_t old_tpr = hw_stats->tpr;
2960 
2961 	hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
2962 	hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
2963 
2964 	if (crc_strip == 0)
2965 		hw_stats->gorc -= delta_gprc * ETHER_CRC_LEN;
2966 
2967 	uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
2968 	hw_stats->gptc += delta_gptc;
2969 	hw_stats->gotc -= delta_gptc * ETHER_CRC_LEN;
2970 	hw_stats->tor -= (hw_stats->tpr - old_tpr) * ETHER_CRC_LEN;
2971 
2972 	/*
2973 	 * Workaround: mprc hardware is incorrectly counting
2974 	 * broadcasts, so for now we subtract those.
2975 	 */
2976 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
2977 	hw_stats->bprc += bprc;
2978 	hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
2979 	if (hw->mac.type == ixgbe_mac_82598EB)
2980 		hw_stats->mprc -= bprc;
2981 
2982 	hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
2983 	hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
2984 	hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
2985 	hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
2986 	hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
2987 	hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
2988 
2989 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
2990 	hw_stats->lxontxc += lxon;
2991 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
2992 	hw_stats->lxofftxc += lxoff;
2993 	total = lxon + lxoff;
2994 
2995 	hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
2996 	hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
2997 	hw_stats->gptc -= total;
2998 	hw_stats->mptc -= total;
2999 	hw_stats->ptc64 -= total;
3000 	hw_stats->gotc -= total * ETHER_MIN_LEN;
3001 
3002 	hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3003 	hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3004 	hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3005 	hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
3006 	hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
3007 	hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
3008 	hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
3009 	hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3010 	hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3011 	hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3012 	hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3013 	hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
3014 	hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3015 	hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
3016 	hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
3017 	hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
3018 	/* Only read FCOE on 82599 */
3019 	if (hw->mac.type != ixgbe_mac_82598EB) {
3020 		hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
3021 		hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
3022 		hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
3023 		hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
3024 		hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
3025 	}
3026 
3027 	/* Flow Director Stats registers */
3028 	hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
3029 	hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
3030 
3031 	/* MACsec Stats registers */
3032 	macsec_stats->out_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECTXUT);
3033 	macsec_stats->out_pkts_encrypted +=
3034 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTE);
3035 	macsec_stats->out_pkts_protected +=
3036 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTP);
3037 	macsec_stats->out_octets_encrypted +=
3038 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTE);
3039 	macsec_stats->out_octets_protected +=
3040 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTP);
3041 	macsec_stats->in_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECRXUT);
3042 	macsec_stats->in_pkts_badtag += IXGBE_READ_REG(hw, IXGBE_LSECRXBAD);
3043 	macsec_stats->in_pkts_nosci += IXGBE_READ_REG(hw, IXGBE_LSECRXNOSCI);
3044 	macsec_stats->in_pkts_unknownsci +=
3045 		IXGBE_READ_REG(hw, IXGBE_LSECRXUNSCI);
3046 	macsec_stats->in_octets_decrypted +=
3047 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTD);
3048 	macsec_stats->in_octets_validated +=
3049 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTV);
3050 	macsec_stats->in_pkts_unchecked += IXGBE_READ_REG(hw, IXGBE_LSECRXUNCH);
3051 	macsec_stats->in_pkts_delayed += IXGBE_READ_REG(hw, IXGBE_LSECRXDELAY);
3052 	macsec_stats->in_pkts_late += IXGBE_READ_REG(hw, IXGBE_LSECRXLATE);
3053 	for (i = 0; i < 2; i++) {
3054 		macsec_stats->in_pkts_ok +=
3055 			IXGBE_READ_REG(hw, IXGBE_LSECRXOK(i));
3056 		macsec_stats->in_pkts_invalid +=
3057 			IXGBE_READ_REG(hw, IXGBE_LSECRXINV(i));
3058 		macsec_stats->in_pkts_notvalid +=
3059 			IXGBE_READ_REG(hw, IXGBE_LSECRXNV(i));
3060 	}
3061 	macsec_stats->in_pkts_unusedsa += IXGBE_READ_REG(hw, IXGBE_LSECRXUNSA);
3062 	macsec_stats->in_pkts_notusingsa +=
3063 		IXGBE_READ_REG(hw, IXGBE_LSECRXNUSA);
3064 }
3065 
3066 /*
3067  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
3068  */
3069 static int
3070 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3071 {
3072 	struct ixgbe_hw *hw =
3073 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3074 	struct ixgbe_hw_stats *hw_stats =
3075 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3076 	struct ixgbe_macsec_stats *macsec_stats =
3077 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3078 				dev->data->dev_private);
3079 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3080 	unsigned i;
3081 
3082 	total_missed_rx = 0;
3083 	total_qbrc = 0;
3084 	total_qprc = 0;
3085 	total_qprdc = 0;
3086 
3087 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3088 			&total_qbrc, &total_qprc, &total_qprdc);
3089 
3090 	if (stats == NULL)
3091 		return -EINVAL;
3092 
3093 	/* Fill out the rte_eth_stats statistics structure */
3094 	stats->ipackets = total_qprc;
3095 	stats->ibytes = total_qbrc;
3096 	stats->opackets = hw_stats->gptc;
3097 	stats->obytes = hw_stats->gotc;
3098 
3099 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3100 		stats->q_ipackets[i] = hw_stats->qprc[i];
3101 		stats->q_opackets[i] = hw_stats->qptc[i];
3102 		stats->q_ibytes[i] = hw_stats->qbrc[i];
3103 		stats->q_obytes[i] = hw_stats->qbtc[i];
3104 		stats->q_errors[i] = hw_stats->qprdc[i];
3105 	}
3106 
3107 	/* Rx Errors */
3108 	stats->imissed  = total_missed_rx;
3109 	stats->ierrors  = hw_stats->crcerrs +
3110 			  hw_stats->mspdc +
3111 			  hw_stats->rlec +
3112 			  hw_stats->ruc +
3113 			  hw_stats->roc +
3114 			  hw_stats->illerrc +
3115 			  hw_stats->errbc +
3116 			  hw_stats->rfc +
3117 			  hw_stats->fccrc +
3118 			  hw_stats->fclast;
3119 
3120 	/* Tx Errors */
3121 	stats->oerrors  = 0;
3122 	return 0;
3123 }
3124 
3125 static void
3126 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
3127 {
3128 	struct ixgbe_hw_stats *stats =
3129 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3130 
3131 	/* HW registers are cleared on read */
3132 	ixgbe_dev_stats_get(dev, NULL);
3133 
3134 	/* Reset software totals */
3135 	memset(stats, 0, sizeof(*stats));
3136 }
3137 
3138 /* This function calculates the number of xstats based on the current config */
3139 static unsigned
3140 ixgbe_xstats_calc_num(void) {
3141 	return IXGBE_NB_HW_STATS + IXGBE_NB_MACSEC_STATS +
3142 		(IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
3143 		(IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
3144 }
3145 
3146 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3147 	struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned int size)
3148 {
3149 	const unsigned cnt_stats = ixgbe_xstats_calc_num();
3150 	unsigned stat, i, count;
3151 
3152 	if (xstats_names != NULL) {
3153 		count = 0;
3154 
3155 		/* Note: limit >= cnt_stats checked upstream
3156 		 * in rte_eth_xstats_names()
3157 		 */
3158 
3159 		/* Extended stats from ixgbe_hw_stats */
3160 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3161 			snprintf(xstats_names[count].name,
3162 				sizeof(xstats_names[count].name),
3163 				"%s",
3164 				rte_ixgbe_stats_strings[i].name);
3165 			count++;
3166 		}
3167 
3168 		/* MACsec Stats */
3169 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3170 			snprintf(xstats_names[count].name,
3171 				sizeof(xstats_names[count].name),
3172 				"%s",
3173 				rte_ixgbe_macsec_strings[i].name);
3174 			count++;
3175 		}
3176 
3177 		/* RX Priority Stats */
3178 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3179 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3180 				snprintf(xstats_names[count].name,
3181 					sizeof(xstats_names[count].name),
3182 					"rx_priority%u_%s", i,
3183 					rte_ixgbe_rxq_strings[stat].name);
3184 				count++;
3185 			}
3186 		}
3187 
3188 		/* TX Priority Stats */
3189 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3190 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3191 				snprintf(xstats_names[count].name,
3192 					sizeof(xstats_names[count].name),
3193 					"tx_priority%u_%s", i,
3194 					rte_ixgbe_txq_strings[stat].name);
3195 				count++;
3196 			}
3197 		}
3198 	}
3199 	return cnt_stats;
3200 }
3201 
3202 static int ixgbe_dev_xstats_get_names_by_id(
3203 	struct rte_eth_dev *dev,
3204 	struct rte_eth_xstat_name *xstats_names,
3205 	const uint64_t *ids,
3206 	unsigned int limit)
3207 {
3208 	if (!ids) {
3209 		const unsigned int cnt_stats = ixgbe_xstats_calc_num();
3210 		unsigned int stat, i, count;
3211 
3212 		if (xstats_names != NULL) {
3213 			count = 0;
3214 
3215 			/* Note: limit >= cnt_stats checked upstream
3216 			 * in rte_eth_xstats_names()
3217 			 */
3218 
3219 			/* Extended stats from ixgbe_hw_stats */
3220 			for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3221 				snprintf(xstats_names[count].name,
3222 					sizeof(xstats_names[count].name),
3223 					"%s",
3224 					rte_ixgbe_stats_strings[i].name);
3225 				count++;
3226 			}
3227 
3228 			/* MACsec Stats */
3229 			for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3230 				snprintf(xstats_names[count].name,
3231 					sizeof(xstats_names[count].name),
3232 					"%s",
3233 					rte_ixgbe_macsec_strings[i].name);
3234 				count++;
3235 			}
3236 
3237 			/* RX Priority Stats */
3238 			for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3239 				for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3240 					snprintf(xstats_names[count].name,
3241 					    sizeof(xstats_names[count].name),
3242 					    "rx_priority%u_%s", i,
3243 					    rte_ixgbe_rxq_strings[stat].name);
3244 					count++;
3245 				}
3246 			}
3247 
3248 			/* TX Priority Stats */
3249 			for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3250 				for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3251 					snprintf(xstats_names[count].name,
3252 					    sizeof(xstats_names[count].name),
3253 					    "tx_priority%u_%s", i,
3254 					    rte_ixgbe_txq_strings[stat].name);
3255 					count++;
3256 				}
3257 			}
3258 		}
3259 		return cnt_stats;
3260 	}
3261 
3262 	uint16_t i;
3263 	uint16_t size = ixgbe_xstats_calc_num();
3264 	struct rte_eth_xstat_name xstats_names_copy[size];
3265 
3266 	ixgbe_dev_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
3267 			size);
3268 
3269 	for (i = 0; i < limit; i++) {
3270 		if (ids[i] >= size) {
3271 			PMD_INIT_LOG(ERR, "id value isn't valid");
3272 			return -1;
3273 		}
3274 		strcpy(xstats_names[i].name,
3275 				xstats_names_copy[ids[i]].name);
3276 	}
3277 	return limit;
3278 }
3279 
3280 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3281 	struct rte_eth_xstat_name *xstats_names, unsigned limit)
3282 {
3283 	unsigned i;
3284 
3285 	if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
3286 		return -ENOMEM;
3287 
3288 	if (xstats_names != NULL)
3289 		for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
3290 			snprintf(xstats_names[i].name,
3291 				sizeof(xstats_names[i].name),
3292 				"%s", rte_ixgbevf_stats_strings[i].name);
3293 	return IXGBEVF_NB_XSTATS;
3294 }
3295 
3296 static int
3297 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3298 					 unsigned n)
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, stat, count = 0;
3309 
3310 	count = ixgbe_xstats_calc_num();
3311 
3312 	if (n < count)
3313 		return count;
3314 
3315 	total_missed_rx = 0;
3316 	total_qbrc = 0;
3317 	total_qprc = 0;
3318 	total_qprdc = 0;
3319 
3320 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3321 			&total_qbrc, &total_qprc, &total_qprdc);
3322 
3323 	/* If this is a reset xstats is NULL, and we have cleared the
3324 	 * registers by reading them.
3325 	 */
3326 	if (!xstats)
3327 		return 0;
3328 
3329 	/* Extended stats from ixgbe_hw_stats */
3330 	count = 0;
3331 	for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3332 		xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3333 				rte_ixgbe_stats_strings[i].offset);
3334 		xstats[count].id = count;
3335 		count++;
3336 	}
3337 
3338 	/* MACsec Stats */
3339 	for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3340 		xstats[count].value = *(uint64_t *)(((char *)macsec_stats) +
3341 				rte_ixgbe_macsec_strings[i].offset);
3342 		xstats[count].id = count;
3343 		count++;
3344 	}
3345 
3346 	/* RX Priority Stats */
3347 	for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3348 		for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3349 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3350 					rte_ixgbe_rxq_strings[stat].offset +
3351 					(sizeof(uint64_t) * i));
3352 			xstats[count].id = count;
3353 			count++;
3354 		}
3355 	}
3356 
3357 	/* TX Priority Stats */
3358 	for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3359 		for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3360 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3361 					rte_ixgbe_txq_strings[stat].offset +
3362 					(sizeof(uint64_t) * i));
3363 			xstats[count].id = count;
3364 			count++;
3365 		}
3366 	}
3367 	return count;
3368 }
3369 
3370 static int
3371 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
3372 		uint64_t *values, unsigned int n)
3373 {
3374 	if (!ids) {
3375 		struct ixgbe_hw *hw =
3376 				IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3377 		struct ixgbe_hw_stats *hw_stats =
3378 				IXGBE_DEV_PRIVATE_TO_STATS(
3379 						dev->data->dev_private);
3380 		struct ixgbe_macsec_stats *macsec_stats =
3381 				IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3382 					dev->data->dev_private);
3383 		uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3384 		unsigned int i, stat, count = 0;
3385 
3386 		count = ixgbe_xstats_calc_num();
3387 
3388 		if (!ids && n < count)
3389 			return count;
3390 
3391 		total_missed_rx = 0;
3392 		total_qbrc = 0;
3393 		total_qprc = 0;
3394 		total_qprdc = 0;
3395 
3396 		ixgbe_read_stats_registers(hw, hw_stats, macsec_stats,
3397 				&total_missed_rx, &total_qbrc, &total_qprc,
3398 				&total_qprdc);
3399 
3400 		/* If this is a reset xstats is NULL, and we have cleared the
3401 		 * registers by reading them.
3402 		 */
3403 		if (!ids && !values)
3404 			return 0;
3405 
3406 		/* Extended stats from ixgbe_hw_stats */
3407 		count = 0;
3408 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3409 			values[count] = *(uint64_t *)(((char *)hw_stats) +
3410 					rte_ixgbe_stats_strings[i].offset);
3411 			count++;
3412 		}
3413 
3414 		/* MACsec Stats */
3415 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3416 			values[count] = *(uint64_t *)(((char *)macsec_stats) +
3417 					rte_ixgbe_macsec_strings[i].offset);
3418 			count++;
3419 		}
3420 
3421 		/* RX Priority Stats */
3422 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3423 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3424 				values[count] =
3425 					*(uint64_t *)(((char *)hw_stats) +
3426 					rte_ixgbe_rxq_strings[stat].offset +
3427 					(sizeof(uint64_t) * i));
3428 				count++;
3429 			}
3430 		}
3431 
3432 		/* TX Priority Stats */
3433 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3434 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3435 				values[count] =
3436 					*(uint64_t *)(((char *)hw_stats) +
3437 					rte_ixgbe_txq_strings[stat].offset +
3438 					(sizeof(uint64_t) * i));
3439 				count++;
3440 			}
3441 		}
3442 		return count;
3443 	}
3444 
3445 	uint16_t i;
3446 	uint16_t size = ixgbe_xstats_calc_num();
3447 	uint64_t values_copy[size];
3448 
3449 	ixgbe_dev_xstats_get_by_id(dev, NULL, values_copy, size);
3450 
3451 	for (i = 0; i < n; i++) {
3452 		if (ids[i] >= size) {
3453 			PMD_INIT_LOG(ERR, "id value isn't valid");
3454 			return -1;
3455 		}
3456 		values[i] = values_copy[ids[i]];
3457 	}
3458 	return n;
3459 }
3460 
3461 static void
3462 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
3463 {
3464 	struct ixgbe_hw_stats *stats =
3465 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3466 	struct ixgbe_macsec_stats *macsec_stats =
3467 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3468 				dev->data->dev_private);
3469 
3470 	unsigned count = ixgbe_xstats_calc_num();
3471 
3472 	/* HW registers are cleared on read */
3473 	ixgbe_dev_xstats_get(dev, NULL, count);
3474 
3475 	/* Reset software totals */
3476 	memset(stats, 0, sizeof(*stats));
3477 	memset(macsec_stats, 0, sizeof(*macsec_stats));
3478 }
3479 
3480 static void
3481 ixgbevf_update_stats(struct rte_eth_dev *dev)
3482 {
3483 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3484 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3485 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3486 
3487 	/* Good Rx packet, include VF loopback */
3488 	UPDATE_VF_STAT(IXGBE_VFGPRC,
3489 	    hw_stats->last_vfgprc, hw_stats->vfgprc);
3490 
3491 	/* Good Rx octets, include VF loopback */
3492 	UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
3493 	    hw_stats->last_vfgorc, hw_stats->vfgorc);
3494 
3495 	/* Good Tx packet, include VF loopback */
3496 	UPDATE_VF_STAT(IXGBE_VFGPTC,
3497 	    hw_stats->last_vfgptc, hw_stats->vfgptc);
3498 
3499 	/* Good Tx octets, include VF loopback */
3500 	UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
3501 	    hw_stats->last_vfgotc, hw_stats->vfgotc);
3502 
3503 	/* Rx Multicst Packet */
3504 	UPDATE_VF_STAT(IXGBE_VFMPRC,
3505 	    hw_stats->last_vfmprc, hw_stats->vfmprc);
3506 }
3507 
3508 static int
3509 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3510 		       unsigned n)
3511 {
3512 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3513 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3514 	unsigned i;
3515 
3516 	if (n < IXGBEVF_NB_XSTATS)
3517 		return IXGBEVF_NB_XSTATS;
3518 
3519 	ixgbevf_update_stats(dev);
3520 
3521 	if (!xstats)
3522 		return 0;
3523 
3524 	/* Extended stats */
3525 	for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
3526 		xstats[i].id = i;
3527 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
3528 			rte_ixgbevf_stats_strings[i].offset);
3529 	}
3530 
3531 	return IXGBEVF_NB_XSTATS;
3532 }
3533 
3534 static int
3535 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3536 {
3537 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3538 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3539 
3540 	ixgbevf_update_stats(dev);
3541 
3542 	if (stats == NULL)
3543 		return -EINVAL;
3544 
3545 	stats->ipackets = hw_stats->vfgprc;
3546 	stats->ibytes = hw_stats->vfgorc;
3547 	stats->opackets = hw_stats->vfgptc;
3548 	stats->obytes = hw_stats->vfgotc;
3549 	return 0;
3550 }
3551 
3552 static void
3553 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3554 {
3555 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3556 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3557 
3558 	/* Sync HW register to the last stats */
3559 	ixgbevf_dev_stats_get(dev, NULL);
3560 
3561 	/* reset HW current stats*/
3562 	hw_stats->vfgprc = 0;
3563 	hw_stats->vfgorc = 0;
3564 	hw_stats->vfgptc = 0;
3565 	hw_stats->vfgotc = 0;
3566 }
3567 
3568 static int
3569 ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
3570 {
3571 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3572 	u16 eeprom_verh, eeprom_verl;
3573 	u32 etrack_id;
3574 	int ret;
3575 
3576 	ixgbe_read_eeprom(hw, 0x2e, &eeprom_verh);
3577 	ixgbe_read_eeprom(hw, 0x2d, &eeprom_verl);
3578 
3579 	etrack_id = (eeprom_verh << 16) | eeprom_verl;
3580 	ret = snprintf(fw_version, fw_size, "0x%08x", etrack_id);
3581 
3582 	ret += 1; /* add the size of '\0' */
3583 	if (fw_size < (u32)ret)
3584 		return ret;
3585 	else
3586 		return 0;
3587 }
3588 
3589 static void
3590 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3591 {
3592 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3593 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3594 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3595 
3596 	dev_info->pci_dev = pci_dev;
3597 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3598 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3599 	if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3600 		/*
3601 		 * When DCB/VT is off, maximum number of queues changes,
3602 		 * except for 82598EB, which remains constant.
3603 		 */
3604 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
3605 				hw->mac.type != ixgbe_mac_82598EB)
3606 			dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3607 	}
3608 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3609 	dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3610 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3611 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3612 	dev_info->max_vfs = pci_dev->max_vfs;
3613 	if (hw->mac.type == ixgbe_mac_82598EB)
3614 		dev_info->max_vmdq_pools = ETH_16_POOLS;
3615 	else
3616 		dev_info->max_vmdq_pools = ETH_64_POOLS;
3617 	dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3618 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3619 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3620 				     dev_info->rx_queue_offload_capa);
3621 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3622 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3623 
3624 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3625 		.rx_thresh = {
3626 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3627 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3628 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3629 		},
3630 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3631 		.rx_drop_en = 0,
3632 		.offloads = 0,
3633 	};
3634 
3635 	dev_info->default_txconf = (struct rte_eth_txconf) {
3636 		.tx_thresh = {
3637 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3638 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3639 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3640 		},
3641 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3642 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3643 		.txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
3644 			     ETH_TXQ_FLAGS_NOOFFLOADS |
3645 			     ETH_TXQ_FLAGS_IGNORE,
3646 		.offloads = 0,
3647 	};
3648 
3649 	dev_info->rx_desc_lim = rx_desc_lim;
3650 	dev_info->tx_desc_lim = tx_desc_lim;
3651 
3652 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3653 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3654 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3655 
3656 	dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G;
3657 	if (hw->mac.type == ixgbe_mac_X540 ||
3658 	    hw->mac.type == ixgbe_mac_X540_vf ||
3659 	    hw->mac.type == ixgbe_mac_X550 ||
3660 	    hw->mac.type == ixgbe_mac_X550_vf) {
3661 		dev_info->speed_capa |= ETH_LINK_SPEED_100M;
3662 	}
3663 	if (hw->mac.type == ixgbe_mac_X550) {
3664 		dev_info->speed_capa |= ETH_LINK_SPEED_2_5G;
3665 		dev_info->speed_capa |= ETH_LINK_SPEED_5G;
3666 	}
3667 }
3668 
3669 static const uint32_t *
3670 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3671 {
3672 	static const uint32_t ptypes[] = {
3673 		/* For non-vec functions,
3674 		 * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3675 		 * for vec functions,
3676 		 * refers to _recv_raw_pkts_vec().
3677 		 */
3678 		RTE_PTYPE_L2_ETHER,
3679 		RTE_PTYPE_L3_IPV4,
3680 		RTE_PTYPE_L3_IPV4_EXT,
3681 		RTE_PTYPE_L3_IPV6,
3682 		RTE_PTYPE_L3_IPV6_EXT,
3683 		RTE_PTYPE_L4_SCTP,
3684 		RTE_PTYPE_L4_TCP,
3685 		RTE_PTYPE_L4_UDP,
3686 		RTE_PTYPE_TUNNEL_IP,
3687 		RTE_PTYPE_INNER_L3_IPV6,
3688 		RTE_PTYPE_INNER_L3_IPV6_EXT,
3689 		RTE_PTYPE_INNER_L4_TCP,
3690 		RTE_PTYPE_INNER_L4_UDP,
3691 		RTE_PTYPE_UNKNOWN
3692 	};
3693 
3694 	if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3695 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3696 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3697 	    dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3698 		return ptypes;
3699 
3700 #if defined(RTE_ARCH_X86)
3701 	if (dev->rx_pkt_burst == ixgbe_recv_pkts_vec ||
3702 	    dev->rx_pkt_burst == ixgbe_recv_scattered_pkts_vec)
3703 		return ptypes;
3704 #endif
3705 	return NULL;
3706 }
3707 
3708 static void
3709 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3710 		     struct rte_eth_dev_info *dev_info)
3711 {
3712 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3713 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3714 
3715 	dev_info->pci_dev = pci_dev;
3716 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3717 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3718 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3719 	dev_info->max_rx_pktlen = 9728; /* includes CRC, cf MAXFRS reg */
3720 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3721 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3722 	dev_info->max_vfs = pci_dev->max_vfs;
3723 	if (hw->mac.type == ixgbe_mac_82598EB)
3724 		dev_info->max_vmdq_pools = ETH_16_POOLS;
3725 	else
3726 		dev_info->max_vmdq_pools = ETH_64_POOLS;
3727 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3728 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3729 				     dev_info->rx_queue_offload_capa);
3730 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3731 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3732 
3733 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3734 		.rx_thresh = {
3735 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3736 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3737 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3738 		},
3739 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3740 		.rx_drop_en = 0,
3741 		.offloads = 0,
3742 	};
3743 
3744 	dev_info->default_txconf = (struct rte_eth_txconf) {
3745 		.tx_thresh = {
3746 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3747 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3748 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3749 		},
3750 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3751 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3752 		.txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
3753 			     ETH_TXQ_FLAGS_NOOFFLOADS |
3754 			     ETH_TXQ_FLAGS_IGNORE,
3755 		.offloads = 0,
3756 	};
3757 
3758 	dev_info->rx_desc_lim = rx_desc_lim;
3759 	dev_info->tx_desc_lim = tx_desc_lim;
3760 }
3761 
3762 static int
3763 ixgbevf_check_link(struct ixgbe_hw *hw, ixgbe_link_speed *speed,
3764 		   int *link_up, int wait_to_complete)
3765 {
3766 	/**
3767 	 * for a quick link status checking, wait_to_compelet == 0,
3768 	 * skip PF link status checking
3769 	 */
3770 	bool no_pflink_check = wait_to_complete == 0;
3771 	struct ixgbe_mbx_info *mbx = &hw->mbx;
3772 	struct ixgbe_mac_info *mac = &hw->mac;
3773 	uint32_t links_reg, in_msg;
3774 	int ret_val = 0;
3775 
3776 	/* If we were hit with a reset drop the link */
3777 	if (!mbx->ops.check_for_rst(hw, 0) || !mbx->timeout)
3778 		mac->get_link_status = true;
3779 
3780 	if (!mac->get_link_status)
3781 		goto out;
3782 
3783 	/* if link status is down no point in checking to see if pf is up */
3784 	links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
3785 	if (!(links_reg & IXGBE_LINKS_UP))
3786 		goto out;
3787 
3788 	/* for SFP+ modules and DA cables on 82599 it can take up to 500usecs
3789 	 * before the link status is correct
3790 	 */
3791 	if (mac->type == ixgbe_mac_82599_vf) {
3792 		int i;
3793 
3794 		for (i = 0; i < 5; i++) {
3795 			rte_delay_us(100);
3796 			links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
3797 
3798 			if (!(links_reg & IXGBE_LINKS_UP))
3799 				goto out;
3800 		}
3801 	}
3802 
3803 	switch (links_reg & IXGBE_LINKS_SPEED_82599) {
3804 	case IXGBE_LINKS_SPEED_10G_82599:
3805 		*speed = IXGBE_LINK_SPEED_10GB_FULL;
3806 		if (hw->mac.type >= ixgbe_mac_X550) {
3807 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
3808 				*speed = IXGBE_LINK_SPEED_2_5GB_FULL;
3809 		}
3810 		break;
3811 	case IXGBE_LINKS_SPEED_1G_82599:
3812 		*speed = IXGBE_LINK_SPEED_1GB_FULL;
3813 		break;
3814 	case IXGBE_LINKS_SPEED_100_82599:
3815 		*speed = IXGBE_LINK_SPEED_100_FULL;
3816 		if (hw->mac.type == ixgbe_mac_X550) {
3817 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
3818 				*speed = IXGBE_LINK_SPEED_5GB_FULL;
3819 		}
3820 		break;
3821 	case IXGBE_LINKS_SPEED_10_X550EM_A:
3822 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
3823 		/* Since Reserved in older MAC's */
3824 		if (hw->mac.type >= ixgbe_mac_X550)
3825 			*speed = IXGBE_LINK_SPEED_10_FULL;
3826 		break;
3827 	default:
3828 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
3829 	}
3830 
3831 	if (no_pflink_check) {
3832 		if (*speed == IXGBE_LINK_SPEED_UNKNOWN)
3833 			mac->get_link_status = true;
3834 		else
3835 			mac->get_link_status = false;
3836 
3837 		goto out;
3838 	}
3839 	/* if the read failed it could just be a mailbox collision, best wait
3840 	 * until we are called again and don't report an error
3841 	 */
3842 	if (mbx->ops.read(hw, &in_msg, 1, 0))
3843 		goto out;
3844 
3845 	if (!(in_msg & IXGBE_VT_MSGTYPE_CTS)) {
3846 		/* msg is not CTS and is NACK we must have lost CTS status */
3847 		if (in_msg & IXGBE_VT_MSGTYPE_NACK)
3848 			ret_val = -1;
3849 		goto out;
3850 	}
3851 
3852 	/* the pf is talking, if we timed out in the past we reinit */
3853 	if (!mbx->timeout) {
3854 		ret_val = -1;
3855 		goto out;
3856 	}
3857 
3858 	/* if we passed all the tests above then the link is up and we no
3859 	 * longer need to check for link
3860 	 */
3861 	mac->get_link_status = false;
3862 
3863 out:
3864 	*link_up = !mac->get_link_status;
3865 	return ret_val;
3866 }
3867 
3868 /* return 0 means link status changed, -1 means not changed */
3869 static int
3870 ixgbe_dev_link_update_share(struct rte_eth_dev *dev,
3871 			    int wait_to_complete, int vf)
3872 {
3873 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3874 	struct rte_eth_link link;
3875 	ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
3876 	struct ixgbe_interrupt *intr =
3877 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3878 	int link_up;
3879 	int diag;
3880 	u32 speed = 0;
3881 	int wait = 1;
3882 	bool autoneg = false;
3883 
3884 	memset(&link, 0, sizeof(link));
3885 	link.link_status = ETH_LINK_DOWN;
3886 	link.link_speed = 0;
3887 	link.link_duplex = ETH_LINK_HALF_DUPLEX;
3888 	link.link_autoneg = ETH_LINK_AUTONEG;
3889 
3890 	hw->mac.get_link_status = true;
3891 
3892 	if ((intr->flags & IXGBE_FLAG_NEED_LINK_CONFIG) &&
3893 		ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
3894 		speed = hw->phy.autoneg_advertised;
3895 		if (!speed)
3896 			ixgbe_get_link_capabilities(hw, &speed, &autoneg);
3897 		ixgbe_setup_link(hw, speed, true);
3898 	}
3899 
3900 	/* check if it needs to wait to complete, if lsc interrupt is enabled */
3901 	if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
3902 		wait = 0;
3903 
3904 	if (vf)
3905 		diag = ixgbevf_check_link(hw, &link_speed, &link_up, wait);
3906 	else
3907 		diag = ixgbe_check_link(hw, &link_speed, &link_up, wait);
3908 
3909 	if (diag != 0) {
3910 		link.link_speed = ETH_SPEED_NUM_100M;
3911 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
3912 		return rte_eth_linkstatus_set(dev, &link);
3913 	}
3914 
3915 	if (link_up == 0) {
3916 		intr->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
3917 		return rte_eth_linkstatus_set(dev, &link);
3918 	}
3919 
3920 	intr->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
3921 	link.link_status = ETH_LINK_UP;
3922 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
3923 
3924 	switch (link_speed) {
3925 	default:
3926 	case IXGBE_LINK_SPEED_UNKNOWN:
3927 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
3928 		link.link_speed = ETH_SPEED_NUM_100M;
3929 		break;
3930 
3931 	case IXGBE_LINK_SPEED_100_FULL:
3932 		link.link_speed = ETH_SPEED_NUM_100M;
3933 		break;
3934 
3935 	case IXGBE_LINK_SPEED_1GB_FULL:
3936 		link.link_speed = ETH_SPEED_NUM_1G;
3937 		break;
3938 
3939 	case IXGBE_LINK_SPEED_2_5GB_FULL:
3940 		link.link_speed = ETH_SPEED_NUM_2_5G;
3941 		break;
3942 
3943 	case IXGBE_LINK_SPEED_5GB_FULL:
3944 		link.link_speed = ETH_SPEED_NUM_5G;
3945 		break;
3946 
3947 	case IXGBE_LINK_SPEED_10GB_FULL:
3948 		link.link_speed = ETH_SPEED_NUM_10G;
3949 		break;
3950 	}
3951 
3952 	return rte_eth_linkstatus_set(dev, &link);
3953 }
3954 
3955 static int
3956 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
3957 {
3958 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 0);
3959 }
3960 
3961 static int
3962 ixgbevf_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
3963 {
3964 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 1);
3965 }
3966 
3967 static void
3968 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
3969 {
3970 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3971 	uint32_t fctrl;
3972 
3973 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3974 	fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3975 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3976 }
3977 
3978 static void
3979 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
3980 {
3981 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3982 	uint32_t fctrl;
3983 
3984 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3985 	fctrl &= (~IXGBE_FCTRL_UPE);
3986 	if (dev->data->all_multicast == 1)
3987 		fctrl |= IXGBE_FCTRL_MPE;
3988 	else
3989 		fctrl &= (~IXGBE_FCTRL_MPE);
3990 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3991 }
3992 
3993 static void
3994 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
3995 {
3996 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3997 	uint32_t fctrl;
3998 
3999 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4000 	fctrl |= IXGBE_FCTRL_MPE;
4001 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4002 }
4003 
4004 static void
4005 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
4006 {
4007 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4008 	uint32_t fctrl;
4009 
4010 	if (dev->data->promiscuous == 1)
4011 		return; /* must remain in all_multicast mode */
4012 
4013 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4014 	fctrl &= (~IXGBE_FCTRL_MPE);
4015 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4016 }
4017 
4018 /**
4019  * It clears the interrupt causes and enables the interrupt.
4020  * It will be called once only during nic initialized.
4021  *
4022  * @param dev
4023  *  Pointer to struct rte_eth_dev.
4024  * @param on
4025  *  Enable or Disable.
4026  *
4027  * @return
4028  *  - On success, zero.
4029  *  - On failure, a negative value.
4030  */
4031 static int
4032 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
4033 {
4034 	struct ixgbe_interrupt *intr =
4035 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4036 
4037 	ixgbe_dev_link_status_print(dev);
4038 	if (on)
4039 		intr->mask |= IXGBE_EICR_LSC;
4040 	else
4041 		intr->mask &= ~IXGBE_EICR_LSC;
4042 
4043 	return 0;
4044 }
4045 
4046 /**
4047  * It clears the interrupt causes and enables the interrupt.
4048  * It will be called once only during nic initialized.
4049  *
4050  * @param dev
4051  *  Pointer to struct rte_eth_dev.
4052  *
4053  * @return
4054  *  - On success, zero.
4055  *  - On failure, a negative value.
4056  */
4057 static int
4058 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
4059 {
4060 	struct ixgbe_interrupt *intr =
4061 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4062 
4063 	intr->mask |= IXGBE_EICR_RTX_QUEUE;
4064 
4065 	return 0;
4066 }
4067 
4068 /**
4069  * It clears the interrupt causes and enables the interrupt.
4070  * It will be called once only during nic initialized.
4071  *
4072  * @param dev
4073  *  Pointer to struct rte_eth_dev.
4074  *
4075  * @return
4076  *  - On success, zero.
4077  *  - On failure, a negative value.
4078  */
4079 static int
4080 ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev)
4081 {
4082 	struct ixgbe_interrupt *intr =
4083 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4084 
4085 	intr->mask |= IXGBE_EICR_LINKSEC;
4086 
4087 	return 0;
4088 }
4089 
4090 /*
4091  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
4092  *
4093  * @param dev
4094  *  Pointer to struct rte_eth_dev.
4095  *
4096  * @return
4097  *  - On success, zero.
4098  *  - On failure, a negative value.
4099  */
4100 static int
4101 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
4102 {
4103 	uint32_t eicr;
4104 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4105 	struct ixgbe_interrupt *intr =
4106 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4107 
4108 	/* clear all cause mask */
4109 	ixgbe_disable_intr(hw);
4110 
4111 	/* read-on-clear nic registers here */
4112 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4113 	PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
4114 
4115 	intr->flags = 0;
4116 
4117 	/* set flag for async link update */
4118 	if (eicr & IXGBE_EICR_LSC)
4119 		intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4120 
4121 	if (eicr & IXGBE_EICR_MAILBOX)
4122 		intr->flags |= IXGBE_FLAG_MAILBOX;
4123 
4124 	if (eicr & IXGBE_EICR_LINKSEC)
4125 		intr->flags |= IXGBE_FLAG_MACSEC;
4126 
4127 	if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
4128 	    hw->phy.type == ixgbe_phy_x550em_ext_t &&
4129 	    (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
4130 		intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
4131 
4132 	return 0;
4133 }
4134 
4135 /**
4136  * It gets and then prints the link status.
4137  *
4138  * @param dev
4139  *  Pointer to struct rte_eth_dev.
4140  *
4141  * @return
4142  *  - On success, zero.
4143  *  - On failure, a negative value.
4144  */
4145 static void
4146 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
4147 {
4148 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4149 	struct rte_eth_link link;
4150 
4151 	rte_eth_linkstatus_get(dev, &link);
4152 
4153 	if (link.link_status) {
4154 		PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
4155 					(int)(dev->data->port_id),
4156 					(unsigned)link.link_speed,
4157 			link.link_duplex == ETH_LINK_FULL_DUPLEX ?
4158 					"full-duplex" : "half-duplex");
4159 	} else {
4160 		PMD_INIT_LOG(INFO, " Port %d: Link Down",
4161 				(int)(dev->data->port_id));
4162 	}
4163 	PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
4164 				pci_dev->addr.domain,
4165 				pci_dev->addr.bus,
4166 				pci_dev->addr.devid,
4167 				pci_dev->addr.function);
4168 }
4169 
4170 /*
4171  * It executes link_update after knowing an interrupt occurred.
4172  *
4173  * @param dev
4174  *  Pointer to struct rte_eth_dev.
4175  *
4176  * @return
4177  *  - On success, zero.
4178  *  - On failure, a negative value.
4179  */
4180 static int
4181 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev,
4182 			   struct rte_intr_handle *intr_handle)
4183 {
4184 	struct ixgbe_interrupt *intr =
4185 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4186 	int64_t timeout;
4187 	struct ixgbe_hw *hw =
4188 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4189 
4190 	PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
4191 
4192 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
4193 		ixgbe_pf_mbx_process(dev);
4194 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
4195 	}
4196 
4197 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4198 		ixgbe_handle_lasi(hw);
4199 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4200 	}
4201 
4202 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4203 		struct rte_eth_link link;
4204 
4205 		/* get the link status before link update, for predicting later */
4206 		rte_eth_linkstatus_get(dev, &link);
4207 
4208 		ixgbe_dev_link_update(dev, 0);
4209 
4210 		/* likely to up */
4211 		if (!link.link_status)
4212 			/* handle it 1 sec later, wait it being stable */
4213 			timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
4214 		/* likely to down */
4215 		else
4216 			/* handle it 4 sec later, wait it being stable */
4217 			timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
4218 
4219 		ixgbe_dev_link_status_print(dev);
4220 		if (rte_eal_alarm_set(timeout * 1000,
4221 				      ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
4222 			PMD_DRV_LOG(ERR, "Error setting alarm");
4223 		else {
4224 			/* remember original mask */
4225 			intr->mask_original = intr->mask;
4226 			/* only disable lsc interrupt */
4227 			intr->mask &= ~IXGBE_EIMS_LSC;
4228 		}
4229 	}
4230 
4231 	PMD_DRV_LOG(DEBUG, "enable intr immediately");
4232 	ixgbe_enable_intr(dev);
4233 	rte_intr_enable(intr_handle);
4234 
4235 	return 0;
4236 }
4237 
4238 /**
4239  * Interrupt handler which shall be registered for alarm callback for delayed
4240  * handling specific interrupt to wait for the stable nic state. As the
4241  * NIC interrupt state is not stable for ixgbe after link is just down,
4242  * it needs to wait 4 seconds to get the stable status.
4243  *
4244  * @param handle
4245  *  Pointer to interrupt handle.
4246  * @param param
4247  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4248  *
4249  * @return
4250  *  void
4251  */
4252 static void
4253 ixgbe_dev_interrupt_delayed_handler(void *param)
4254 {
4255 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4256 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4257 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
4258 	struct ixgbe_interrupt *intr =
4259 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4260 	struct ixgbe_hw *hw =
4261 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4262 	uint32_t eicr;
4263 
4264 	ixgbe_disable_intr(hw);
4265 
4266 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4267 	if (eicr & IXGBE_EICR_MAILBOX)
4268 		ixgbe_pf_mbx_process(dev);
4269 
4270 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4271 		ixgbe_handle_lasi(hw);
4272 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4273 	}
4274 
4275 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4276 		ixgbe_dev_link_update(dev, 0);
4277 		intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4278 		ixgbe_dev_link_status_print(dev);
4279 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC,
4280 					      NULL);
4281 	}
4282 
4283 	if (intr->flags & IXGBE_FLAG_MACSEC) {
4284 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_MACSEC,
4285 					      NULL);
4286 		intr->flags &= ~IXGBE_FLAG_MACSEC;
4287 	}
4288 
4289 	/* restore original mask */
4290 	intr->mask = intr->mask_original;
4291 	intr->mask_original = 0;
4292 
4293 	PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
4294 	ixgbe_enable_intr(dev);
4295 	rte_intr_enable(intr_handle);
4296 }
4297 
4298 /**
4299  * Interrupt handler triggered by NIC  for handling
4300  * specific interrupt.
4301  *
4302  * @param handle
4303  *  Pointer to interrupt handle.
4304  * @param param
4305  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4306  *
4307  * @return
4308  *  void
4309  */
4310 static void
4311 ixgbe_dev_interrupt_handler(void *param)
4312 {
4313 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4314 
4315 	ixgbe_dev_interrupt_get_status(dev);
4316 	ixgbe_dev_interrupt_action(dev, dev->intr_handle);
4317 }
4318 
4319 static int
4320 ixgbe_dev_led_on(struct rte_eth_dev *dev)
4321 {
4322 	struct ixgbe_hw *hw;
4323 
4324 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4325 	return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4326 }
4327 
4328 static int
4329 ixgbe_dev_led_off(struct rte_eth_dev *dev)
4330 {
4331 	struct ixgbe_hw *hw;
4332 
4333 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4334 	return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4335 }
4336 
4337 static int
4338 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4339 {
4340 	struct ixgbe_hw *hw;
4341 	uint32_t mflcn_reg;
4342 	uint32_t fccfg_reg;
4343 	int rx_pause;
4344 	int tx_pause;
4345 
4346 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4347 
4348 	fc_conf->pause_time = hw->fc.pause_time;
4349 	fc_conf->high_water = hw->fc.high_water[0];
4350 	fc_conf->low_water = hw->fc.low_water[0];
4351 	fc_conf->send_xon = hw->fc.send_xon;
4352 	fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
4353 
4354 	/*
4355 	 * Return rx_pause status according to actual setting of
4356 	 * MFLCN register.
4357 	 */
4358 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4359 	if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
4360 		rx_pause = 1;
4361 	else
4362 		rx_pause = 0;
4363 
4364 	/*
4365 	 * Return tx_pause status according to actual setting of
4366 	 * FCCFG register.
4367 	 */
4368 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4369 	if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
4370 		tx_pause = 1;
4371 	else
4372 		tx_pause = 0;
4373 
4374 	if (rx_pause && tx_pause)
4375 		fc_conf->mode = RTE_FC_FULL;
4376 	else if (rx_pause)
4377 		fc_conf->mode = RTE_FC_RX_PAUSE;
4378 	else if (tx_pause)
4379 		fc_conf->mode = RTE_FC_TX_PAUSE;
4380 	else
4381 		fc_conf->mode = RTE_FC_NONE;
4382 
4383 	return 0;
4384 }
4385 
4386 static int
4387 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4388 {
4389 	struct ixgbe_hw *hw;
4390 	int err;
4391 	uint32_t rx_buf_size;
4392 	uint32_t max_high_water;
4393 	uint32_t mflcn;
4394 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4395 		ixgbe_fc_none,
4396 		ixgbe_fc_rx_pause,
4397 		ixgbe_fc_tx_pause,
4398 		ixgbe_fc_full
4399 	};
4400 
4401 	PMD_INIT_FUNC_TRACE();
4402 
4403 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4404 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
4405 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4406 
4407 	/*
4408 	 * At least reserve one Ethernet frame for watermark
4409 	 * high_water/low_water in kilo bytes for ixgbe
4410 	 */
4411 	max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4412 	if ((fc_conf->high_water > max_high_water) ||
4413 		(fc_conf->high_water < fc_conf->low_water)) {
4414 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4415 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4416 		return -EINVAL;
4417 	}
4418 
4419 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
4420 	hw->fc.pause_time     = fc_conf->pause_time;
4421 	hw->fc.high_water[0]  = fc_conf->high_water;
4422 	hw->fc.low_water[0]   = fc_conf->low_water;
4423 	hw->fc.send_xon       = fc_conf->send_xon;
4424 	hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
4425 
4426 	err = ixgbe_fc_enable(hw);
4427 
4428 	/* Not negotiated is not an error case */
4429 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED)) {
4430 
4431 		/* check if we want to forward MAC frames - driver doesn't have native
4432 		 * capability to do that, so we'll write the registers ourselves */
4433 
4434 		mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4435 
4436 		/* set or clear MFLCN.PMCF bit depending on configuration */
4437 		if (fc_conf->mac_ctrl_frame_fwd != 0)
4438 			mflcn |= IXGBE_MFLCN_PMCF;
4439 		else
4440 			mflcn &= ~IXGBE_MFLCN_PMCF;
4441 
4442 		IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
4443 		IXGBE_WRITE_FLUSH(hw);
4444 
4445 		return 0;
4446 	}
4447 
4448 	PMD_INIT_LOG(ERR, "ixgbe_fc_enable = 0x%x", err);
4449 	return -EIO;
4450 }
4451 
4452 /**
4453  *  ixgbe_pfc_enable_generic - Enable flow control
4454  *  @hw: pointer to hardware structure
4455  *  @tc_num: traffic class number
4456  *  Enable flow control according to the current settings.
4457  */
4458 static int
4459 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
4460 {
4461 	int ret_val = 0;
4462 	uint32_t mflcn_reg, fccfg_reg;
4463 	uint32_t reg;
4464 	uint32_t fcrtl, fcrth;
4465 	uint8_t i;
4466 	uint8_t nb_rx_en;
4467 
4468 	/* Validate the water mark configuration */
4469 	if (!hw->fc.pause_time) {
4470 		ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4471 		goto out;
4472 	}
4473 
4474 	/* Low water mark of zero causes XOFF floods */
4475 	if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
4476 		 /* High/Low water can not be 0 */
4477 		if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_water[tc_num])) {
4478 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4479 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4480 			goto out;
4481 		}
4482 
4483 		if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
4484 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4485 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4486 			goto out;
4487 		}
4488 	}
4489 	/* Negotiate the fc mode to use */
4490 	ixgbe_fc_autoneg(hw);
4491 
4492 	/* Disable any previous flow control settings */
4493 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4494 	mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
4495 
4496 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4497 	fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
4498 
4499 	switch (hw->fc.current_mode) {
4500 	case ixgbe_fc_none:
4501 		/*
4502 		 * If the count of enabled RX Priority Flow control >1,
4503 		 * and the TX pause can not be disabled
4504 		 */
4505 		nb_rx_en = 0;
4506 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4507 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4508 			if (reg & IXGBE_FCRTH_FCEN)
4509 				nb_rx_en++;
4510 		}
4511 		if (nb_rx_en > 1)
4512 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4513 		break;
4514 	case ixgbe_fc_rx_pause:
4515 		/*
4516 		 * Rx Flow control is enabled and Tx Flow control is
4517 		 * disabled by software override. Since there really
4518 		 * isn't a way to advertise that we are capable of RX
4519 		 * Pause ONLY, we will advertise that we support both
4520 		 * symmetric and asymmetric Rx PAUSE.  Later, we will
4521 		 * disable the adapter's ability to send PAUSE frames.
4522 		 */
4523 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4524 		/*
4525 		 * If the count of enabled RX Priority Flow control >1,
4526 		 * and the TX pause can not be disabled
4527 		 */
4528 		nb_rx_en = 0;
4529 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4530 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4531 			if (reg & IXGBE_FCRTH_FCEN)
4532 				nb_rx_en++;
4533 		}
4534 		if (nb_rx_en > 1)
4535 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4536 		break;
4537 	case ixgbe_fc_tx_pause:
4538 		/*
4539 		 * Tx Flow control is enabled, and Rx Flow control is
4540 		 * disabled by software override.
4541 		 */
4542 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4543 		break;
4544 	case ixgbe_fc_full:
4545 		/* Flow control (both Rx and Tx) is enabled by SW override. */
4546 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4547 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4548 		break;
4549 	default:
4550 		PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
4551 		ret_val = IXGBE_ERR_CONFIG;
4552 		goto out;
4553 	}
4554 
4555 	/* Set 802.3x based flow control settings. */
4556 	mflcn_reg |= IXGBE_MFLCN_DPF;
4557 	IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
4558 	IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
4559 
4560 	/* Set up and enable Rx high/low water mark thresholds, enable XON. */
4561 	if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
4562 		hw->fc.high_water[tc_num]) {
4563 		fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
4564 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
4565 		fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
4566 	} else {
4567 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
4568 		/*
4569 		 * In order to prevent Tx hangs when the internal Tx
4570 		 * switch is enabled we must set the high water mark
4571 		 * to the maximum FCRTH value.  This allows the Tx
4572 		 * switch to function even under heavy Rx workloads.
4573 		 */
4574 		fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
4575 	}
4576 	IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
4577 
4578 	/* Configure pause time (2 TCs per register) */
4579 	reg = hw->fc.pause_time * 0x00010001;
4580 	for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
4581 		IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
4582 
4583 	/* Configure flow control refresh threshold value */
4584 	IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
4585 
4586 out:
4587 	return ret_val;
4588 }
4589 
4590 static int
4591 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
4592 {
4593 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4594 	int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
4595 
4596 	if (hw->mac.type != ixgbe_mac_82598EB) {
4597 		ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
4598 	}
4599 	return ret_val;
4600 }
4601 
4602 static int
4603 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
4604 {
4605 	int err;
4606 	uint32_t rx_buf_size;
4607 	uint32_t max_high_water;
4608 	uint8_t tc_num;
4609 	uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
4610 	struct ixgbe_hw *hw =
4611 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4612 	struct ixgbe_dcb_config *dcb_config =
4613 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
4614 
4615 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4616 		ixgbe_fc_none,
4617 		ixgbe_fc_rx_pause,
4618 		ixgbe_fc_tx_pause,
4619 		ixgbe_fc_full
4620 	};
4621 
4622 	PMD_INIT_FUNC_TRACE();
4623 
4624 	ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
4625 	tc_num = map[pfc_conf->priority];
4626 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
4627 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4628 	/*
4629 	 * At least reserve one Ethernet frame for watermark
4630 	 * high_water/low_water in kilo bytes for ixgbe
4631 	 */
4632 	max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4633 	if ((pfc_conf->fc.high_water > max_high_water) ||
4634 	    (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
4635 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4636 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4637 		return -EINVAL;
4638 	}
4639 
4640 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
4641 	hw->fc.pause_time = pfc_conf->fc.pause_time;
4642 	hw->fc.send_xon = pfc_conf->fc.send_xon;
4643 	hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
4644 	hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
4645 
4646 	err = ixgbe_dcb_pfc_enable(dev, tc_num);
4647 
4648 	/* Not negotiated is not an error case */
4649 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
4650 		return 0;
4651 
4652 	PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
4653 	return -EIO;
4654 }
4655 
4656 static int
4657 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
4658 			  struct rte_eth_rss_reta_entry64 *reta_conf,
4659 			  uint16_t reta_size)
4660 {
4661 	uint16_t i, sp_reta_size;
4662 	uint8_t j, mask;
4663 	uint32_t reta, r;
4664 	uint16_t idx, shift;
4665 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4666 	uint32_t reta_reg;
4667 
4668 	PMD_INIT_FUNC_TRACE();
4669 
4670 	if (!ixgbe_rss_update_sp(hw->mac.type)) {
4671 		PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
4672 			"NIC.");
4673 		return -ENOTSUP;
4674 	}
4675 
4676 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
4677 	if (reta_size != sp_reta_size) {
4678 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
4679 			"(%d) doesn't match the number hardware can supported "
4680 			"(%d)", reta_size, sp_reta_size);
4681 		return -EINVAL;
4682 	}
4683 
4684 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
4685 		idx = i / RTE_RETA_GROUP_SIZE;
4686 		shift = i % RTE_RETA_GROUP_SIZE;
4687 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
4688 						IXGBE_4_BIT_MASK);
4689 		if (!mask)
4690 			continue;
4691 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
4692 		if (mask == IXGBE_4_BIT_MASK)
4693 			r = 0;
4694 		else
4695 			r = IXGBE_READ_REG(hw, reta_reg);
4696 		for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
4697 			if (mask & (0x1 << j))
4698 				reta |= reta_conf[idx].reta[shift + j] <<
4699 							(CHAR_BIT * j);
4700 			else
4701 				reta |= r & (IXGBE_8_BIT_MASK <<
4702 						(CHAR_BIT * j));
4703 		}
4704 		IXGBE_WRITE_REG(hw, reta_reg, reta);
4705 	}
4706 
4707 	return 0;
4708 }
4709 
4710 static int
4711 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
4712 			 struct rte_eth_rss_reta_entry64 *reta_conf,
4713 			 uint16_t reta_size)
4714 {
4715 	uint16_t i, sp_reta_size;
4716 	uint8_t j, mask;
4717 	uint32_t reta;
4718 	uint16_t idx, shift;
4719 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4720 	uint32_t reta_reg;
4721 
4722 	PMD_INIT_FUNC_TRACE();
4723 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
4724 	if (reta_size != sp_reta_size) {
4725 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
4726 			"(%d) doesn't match the number hardware can supported "
4727 			"(%d)", reta_size, sp_reta_size);
4728 		return -EINVAL;
4729 	}
4730 
4731 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
4732 		idx = i / RTE_RETA_GROUP_SIZE;
4733 		shift = i % RTE_RETA_GROUP_SIZE;
4734 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
4735 						IXGBE_4_BIT_MASK);
4736 		if (!mask)
4737 			continue;
4738 
4739 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
4740 		reta = IXGBE_READ_REG(hw, reta_reg);
4741 		for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
4742 			if (mask & (0x1 << j))
4743 				reta_conf[idx].reta[shift + j] =
4744 					((reta >> (CHAR_BIT * j)) &
4745 						IXGBE_8_BIT_MASK);
4746 		}
4747 	}
4748 
4749 	return 0;
4750 }
4751 
4752 static int
4753 ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
4754 				uint32_t index, uint32_t pool)
4755 {
4756 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4757 	uint32_t enable_addr = 1;
4758 
4759 	return ixgbe_set_rar(hw, index, mac_addr->addr_bytes,
4760 			     pool, enable_addr);
4761 }
4762 
4763 static void
4764 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
4765 {
4766 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4767 
4768 	ixgbe_clear_rar(hw, index);
4769 }
4770 
4771 static void
4772 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
4773 {
4774 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4775 
4776 	ixgbe_remove_rar(dev, 0);
4777 
4778 	ixgbe_add_rar(dev, addr, 0, pci_dev->max_vfs);
4779 }
4780 
4781 static bool
4782 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
4783 {
4784 	if (strcmp(dev->device->driver->name, drv->driver.name))
4785 		return false;
4786 
4787 	return true;
4788 }
4789 
4790 bool
4791 is_ixgbe_supported(struct rte_eth_dev *dev)
4792 {
4793 	return is_device_supported(dev, &rte_ixgbe_pmd);
4794 }
4795 
4796 static int
4797 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4798 {
4799 	uint32_t hlreg0;
4800 	uint32_t maxfrs;
4801 	struct ixgbe_hw *hw;
4802 	struct rte_eth_dev_info dev_info;
4803 	uint32_t frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
4804 	struct rte_eth_dev_data *dev_data = dev->data;
4805 
4806 	ixgbe_dev_info_get(dev, &dev_info);
4807 
4808 	/* check that mtu is within the allowed range */
4809 	if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
4810 		return -EINVAL;
4811 
4812 	/* If device is started, refuse mtu that requires the support of
4813 	 * scattered packets when this feature has not been enabled before.
4814 	 */
4815 	if (dev_data->dev_started && !dev_data->scattered_rx &&
4816 	    (frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
4817 	     dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
4818 		PMD_INIT_LOG(ERR, "Stop port first.");
4819 		return -EINVAL;
4820 	}
4821 
4822 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4823 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4824 
4825 	/* switch to jumbo mode if needed */
4826 	if (frame_size > ETHER_MAX_LEN) {
4827 		dev->data->dev_conf.rxmode.offloads |=
4828 			DEV_RX_OFFLOAD_JUMBO_FRAME;
4829 		hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4830 	} else {
4831 		dev->data->dev_conf.rxmode.offloads &=
4832 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
4833 		hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
4834 	}
4835 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4836 
4837 	/* update max frame size */
4838 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4839 
4840 	maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
4841 	maxfrs &= 0x0000FFFF;
4842 	maxfrs |= (dev->data->dev_conf.rxmode.max_rx_pkt_len << 16);
4843 	IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
4844 
4845 	return 0;
4846 }
4847 
4848 /*
4849  * Virtual Function operations
4850  */
4851 static void
4852 ixgbevf_intr_disable(struct ixgbe_hw *hw)
4853 {
4854 	PMD_INIT_FUNC_TRACE();
4855 
4856 	/* Clear interrupt mask to stop from interrupts being generated */
4857 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
4858 
4859 	IXGBE_WRITE_FLUSH(hw);
4860 }
4861 
4862 static void
4863 ixgbevf_intr_enable(struct ixgbe_hw *hw)
4864 {
4865 	PMD_INIT_FUNC_TRACE();
4866 
4867 	/* VF enable interrupt autoclean */
4868 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
4869 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
4870 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
4871 
4872 	IXGBE_WRITE_FLUSH(hw);
4873 }
4874 
4875 static int
4876 ixgbevf_dev_configure(struct rte_eth_dev *dev)
4877 {
4878 	struct rte_eth_conf *conf = &dev->data->dev_conf;
4879 	struct ixgbe_adapter *adapter =
4880 			(struct ixgbe_adapter *)dev->data->dev_private;
4881 	struct rte_eth_dev_info dev_info;
4882 	uint64_t rx_offloads;
4883 	uint64_t tx_offloads;
4884 
4885 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
4886 		     dev->data->port_id);
4887 
4888 	ixgbevf_dev_info_get(dev, &dev_info);
4889 	rx_offloads = dev->data->dev_conf.rxmode.offloads;
4890 	if ((rx_offloads & dev_info.rx_offload_capa) != rx_offloads) {
4891 		PMD_DRV_LOG(ERR, "Some Rx offloads are not supported "
4892 			    "requested 0x%" PRIx64 " supported 0x%" PRIx64,
4893 			    rx_offloads, dev_info.rx_offload_capa);
4894 		return -ENOTSUP;
4895 	}
4896 	tx_offloads = dev->data->dev_conf.txmode.offloads;
4897 	if ((tx_offloads & dev_info.tx_offload_capa) != tx_offloads) {
4898 		PMD_DRV_LOG(ERR, "Some Tx offloads are not supported "
4899 			    "requested 0x%" PRIx64 " supported 0x%" PRIx64,
4900 			    tx_offloads, dev_info.tx_offload_capa);
4901 		return -ENOTSUP;
4902 	}
4903 
4904 	/*
4905 	 * VF has no ability to enable/disable HW CRC
4906 	 * Keep the persistent behavior the same as Host PF
4907 	 */
4908 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
4909 	if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_CRC_STRIP)) {
4910 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
4911 		conf->rxmode.offloads |= DEV_RX_OFFLOAD_CRC_STRIP;
4912 	}
4913 #else
4914 	if (conf->rxmode.offloads & DEV_RX_OFFLOAD_CRC_STRIP) {
4915 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
4916 		conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_CRC_STRIP;
4917 	}
4918 #endif
4919 
4920 	/*
4921 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
4922 	 * allocation or vector Rx preconditions we will reset it.
4923 	 */
4924 	adapter->rx_bulk_alloc_allowed = true;
4925 	adapter->rx_vec_allowed = true;
4926 
4927 	return 0;
4928 }
4929 
4930 static int
4931 ixgbevf_dev_start(struct rte_eth_dev *dev)
4932 {
4933 	struct ixgbe_hw *hw =
4934 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4935 	uint32_t intr_vector = 0;
4936 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4937 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
4938 
4939 	int err, mask = 0;
4940 
4941 	PMD_INIT_FUNC_TRACE();
4942 
4943 	err = hw->mac.ops.reset_hw(hw);
4944 	if (err) {
4945 		PMD_INIT_LOG(ERR, "Unable to reset vf hardware (%d)", err);
4946 		return err;
4947 	}
4948 	hw->mac.get_link_status = true;
4949 
4950 	/* negotiate mailbox API version to use with the PF. */
4951 	ixgbevf_negotiate_api(hw);
4952 
4953 	ixgbevf_dev_tx_init(dev);
4954 
4955 	/* This can fail when allocating mbufs for descriptor rings */
4956 	err = ixgbevf_dev_rx_init(dev);
4957 	if (err) {
4958 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
4959 		ixgbe_dev_clear_queues(dev);
4960 		return err;
4961 	}
4962 
4963 	/* Set vfta */
4964 	ixgbevf_set_vfta_all(dev, 1);
4965 
4966 	/* Set HW strip */
4967 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
4968 		ETH_VLAN_EXTEND_MASK;
4969 	err = ixgbevf_vlan_offload_set(dev, mask);
4970 	if (err) {
4971 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
4972 		ixgbe_dev_clear_queues(dev);
4973 		return err;
4974 	}
4975 
4976 	ixgbevf_dev_rxtx_start(dev);
4977 
4978 	/* check and configure queue intr-vector mapping */
4979 	if (rte_intr_cap_multiple(intr_handle) &&
4980 	    dev->data->dev_conf.intr_conf.rxq) {
4981 		/* According to datasheet, only vector 0/1/2 can be used,
4982 		 * now only one vector is used for Rx queue
4983 		 */
4984 		intr_vector = 1;
4985 		if (rte_intr_efd_enable(intr_handle, intr_vector))
4986 			return -1;
4987 	}
4988 
4989 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
4990 		intr_handle->intr_vec =
4991 			rte_zmalloc("intr_vec",
4992 				    dev->data->nb_rx_queues * sizeof(int), 0);
4993 		if (intr_handle->intr_vec == NULL) {
4994 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
4995 				     " intr_vec", dev->data->nb_rx_queues);
4996 			return -ENOMEM;
4997 		}
4998 	}
4999 	ixgbevf_configure_msix(dev);
5000 
5001 	/* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
5002 	 * is mapped to VFIO vector 0 in eth_ixgbevf_dev_init( ).
5003 	 * If previous VFIO interrupt mapping setting in eth_ixgbevf_dev_init( )
5004 	 * is not cleared, it will fail when following rte_intr_enable( ) tries
5005 	 * to map Rx queue interrupt to other VFIO vectors.
5006 	 * So clear uio/vfio intr/evevnfd first to avoid failure.
5007 	 */
5008 	rte_intr_disable(intr_handle);
5009 
5010 	rte_intr_enable(intr_handle);
5011 
5012 	/* Re-enable interrupt for VF */
5013 	ixgbevf_intr_enable(hw);
5014 
5015 	return 0;
5016 }
5017 
5018 static void
5019 ixgbevf_dev_stop(struct rte_eth_dev *dev)
5020 {
5021 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5022 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5023 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5024 
5025 	PMD_INIT_FUNC_TRACE();
5026 
5027 	ixgbevf_intr_disable(hw);
5028 
5029 	hw->adapter_stopped = 1;
5030 	ixgbe_stop_adapter(hw);
5031 
5032 	/*
5033 	  * Clear what we set, but we still keep shadow_vfta to
5034 	  * restore after device starts
5035 	  */
5036 	ixgbevf_set_vfta_all(dev, 0);
5037 
5038 	/* Clear stored conf */
5039 	dev->data->scattered_rx = 0;
5040 
5041 	ixgbe_dev_clear_queues(dev);
5042 
5043 	/* Clean datapath event and queue/vec mapping */
5044 	rte_intr_efd_disable(intr_handle);
5045 	if (intr_handle->intr_vec != NULL) {
5046 		rte_free(intr_handle->intr_vec);
5047 		intr_handle->intr_vec = NULL;
5048 	}
5049 }
5050 
5051 static void
5052 ixgbevf_dev_close(struct rte_eth_dev *dev)
5053 {
5054 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5055 
5056 	PMD_INIT_FUNC_TRACE();
5057 
5058 	ixgbe_reset_hw(hw);
5059 
5060 	ixgbevf_dev_stop(dev);
5061 
5062 	ixgbe_dev_free_queues(dev);
5063 
5064 	/**
5065 	 * Remove the VF MAC address ro ensure
5066 	 * that the VF traffic goes to the PF
5067 	 * after stop, close and detach of the VF
5068 	 **/
5069 	ixgbevf_remove_mac_addr(dev, 0);
5070 }
5071 
5072 /*
5073  * Reset VF device
5074  */
5075 static int
5076 ixgbevf_dev_reset(struct rte_eth_dev *dev)
5077 {
5078 	int ret;
5079 
5080 	ret = eth_ixgbevf_dev_uninit(dev);
5081 	if (ret)
5082 		return ret;
5083 
5084 	ret = eth_ixgbevf_dev_init(dev);
5085 
5086 	return ret;
5087 }
5088 
5089 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
5090 {
5091 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5092 	struct ixgbe_vfta *shadow_vfta =
5093 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5094 	int i = 0, j = 0, vfta = 0, mask = 1;
5095 
5096 	for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
5097 		vfta = shadow_vfta->vfta[i];
5098 		if (vfta) {
5099 			mask = 1;
5100 			for (j = 0; j < 32; j++) {
5101 				if (vfta & mask)
5102 					ixgbe_set_vfta(hw, (i<<5)+j, 0,
5103 						       on, false);
5104 				mask <<= 1;
5105 			}
5106 		}
5107 	}
5108 
5109 }
5110 
5111 static int
5112 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
5113 {
5114 	struct ixgbe_hw *hw =
5115 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5116 	struct ixgbe_vfta *shadow_vfta =
5117 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5118 	uint32_t vid_idx = 0;
5119 	uint32_t vid_bit = 0;
5120 	int ret = 0;
5121 
5122 	PMD_INIT_FUNC_TRACE();
5123 
5124 	/* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
5125 	ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
5126 	if (ret) {
5127 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
5128 		return ret;
5129 	}
5130 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
5131 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
5132 
5133 	/* Save what we set and retore it after device reset */
5134 	if (on)
5135 		shadow_vfta->vfta[vid_idx] |= vid_bit;
5136 	else
5137 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
5138 
5139 	return 0;
5140 }
5141 
5142 static void
5143 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
5144 {
5145 	struct ixgbe_hw *hw =
5146 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5147 	uint32_t ctrl;
5148 
5149 	PMD_INIT_FUNC_TRACE();
5150 
5151 	if (queue >= hw->mac.max_rx_queues)
5152 		return;
5153 
5154 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
5155 	if (on)
5156 		ctrl |= IXGBE_RXDCTL_VME;
5157 	else
5158 		ctrl &= ~IXGBE_RXDCTL_VME;
5159 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
5160 
5161 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
5162 }
5163 
5164 static int
5165 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
5166 {
5167 	struct ixgbe_hw *hw =
5168 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5169 	struct ixgbe_rx_queue *rxq;
5170 	uint16_t i;
5171 	int on = 0;
5172 
5173 	/* VF function only support hw strip feature, others are not support */
5174 	if (mask & ETH_VLAN_STRIP_MASK) {
5175 		for (i = 0; i < hw->mac.max_rx_queues; i++) {
5176 			rxq = dev->data->rx_queues[i];
5177 			on = !!(rxq->offloads &	DEV_RX_OFFLOAD_VLAN_STRIP);
5178 			ixgbevf_vlan_strip_queue_set(dev, i, on);
5179 		}
5180 	}
5181 
5182 	return 0;
5183 }
5184 
5185 int
5186 ixgbe_vt_check(struct ixgbe_hw *hw)
5187 {
5188 	uint32_t reg_val;
5189 
5190 	/* if Virtualization Technology is enabled */
5191 	reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
5192 	if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
5193 		PMD_INIT_LOG(ERR, "VT must be enabled for this setting");
5194 		return -1;
5195 	}
5196 
5197 	return 0;
5198 }
5199 
5200 static uint32_t
5201 ixgbe_uta_vector(struct ixgbe_hw *hw, struct ether_addr *uc_addr)
5202 {
5203 	uint32_t vector = 0;
5204 
5205 	switch (hw->mac.mc_filter_type) {
5206 	case 0:   /* use bits [47:36] of the address */
5207 		vector = ((uc_addr->addr_bytes[4] >> 4) |
5208 			(((uint16_t)uc_addr->addr_bytes[5]) << 4));
5209 		break;
5210 	case 1:   /* use bits [46:35] of the address */
5211 		vector = ((uc_addr->addr_bytes[4] >> 3) |
5212 			(((uint16_t)uc_addr->addr_bytes[5]) << 5));
5213 		break;
5214 	case 2:   /* use bits [45:34] of the address */
5215 		vector = ((uc_addr->addr_bytes[4] >> 2) |
5216 			(((uint16_t)uc_addr->addr_bytes[5]) << 6));
5217 		break;
5218 	case 3:   /* use bits [43:32] of the address */
5219 		vector = ((uc_addr->addr_bytes[4]) |
5220 			(((uint16_t)uc_addr->addr_bytes[5]) << 8));
5221 		break;
5222 	default:  /* Invalid mc_filter_type */
5223 		break;
5224 	}
5225 
5226 	/* vector can only be 12-bits or boundary will be exceeded */
5227 	vector &= 0xFFF;
5228 	return vector;
5229 }
5230 
5231 static int
5232 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
5233 			uint8_t on)
5234 {
5235 	uint32_t vector;
5236 	uint32_t uta_idx;
5237 	uint32_t reg_val;
5238 	uint32_t uta_shift;
5239 	uint32_t rc;
5240 	const uint32_t ixgbe_uta_idx_mask = 0x7F;
5241 	const uint32_t ixgbe_uta_bit_shift = 5;
5242 	const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
5243 	const uint32_t bit1 = 0x1;
5244 
5245 	struct ixgbe_hw *hw =
5246 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5247 	struct ixgbe_uta_info *uta_info =
5248 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5249 
5250 	/* The UTA table only exists on 82599 hardware and newer */
5251 	if (hw->mac.type < ixgbe_mac_82599EB)
5252 		return -ENOTSUP;
5253 
5254 	vector = ixgbe_uta_vector(hw, mac_addr);
5255 	uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
5256 	uta_shift = vector & ixgbe_uta_bit_mask;
5257 
5258 	rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
5259 	if (rc == on)
5260 		return 0;
5261 
5262 	reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
5263 	if (on) {
5264 		uta_info->uta_in_use++;
5265 		reg_val |= (bit1 << uta_shift);
5266 		uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
5267 	} else {
5268 		uta_info->uta_in_use--;
5269 		reg_val &= ~(bit1 << uta_shift);
5270 		uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
5271 	}
5272 
5273 	IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
5274 
5275 	if (uta_info->uta_in_use > 0)
5276 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
5277 				IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
5278 	else
5279 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
5280 
5281 	return 0;
5282 }
5283 
5284 static int
5285 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
5286 {
5287 	int i;
5288 	struct ixgbe_hw *hw =
5289 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5290 	struct ixgbe_uta_info *uta_info =
5291 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5292 
5293 	/* The UTA table only exists on 82599 hardware and newer */
5294 	if (hw->mac.type < ixgbe_mac_82599EB)
5295 		return -ENOTSUP;
5296 
5297 	if (on) {
5298 		for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5299 			uta_info->uta_shadow[i] = ~0;
5300 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
5301 		}
5302 	} else {
5303 		for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5304 			uta_info->uta_shadow[i] = 0;
5305 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
5306 		}
5307 	}
5308 	return 0;
5309 
5310 }
5311 
5312 uint32_t
5313 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
5314 {
5315 	uint32_t new_val = orig_val;
5316 
5317 	if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG)
5318 		new_val |= IXGBE_VMOLR_AUPE;
5319 	if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC)
5320 		new_val |= IXGBE_VMOLR_ROMPE;
5321 	if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
5322 		new_val |= IXGBE_VMOLR_ROPE;
5323 	if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
5324 		new_val |= IXGBE_VMOLR_BAM;
5325 	if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
5326 		new_val |= IXGBE_VMOLR_MPE;
5327 
5328 	return new_val;
5329 }
5330 
5331 #define IXGBE_MRCTL_VPME  0x01 /* Virtual Pool Mirroring. */
5332 #define IXGBE_MRCTL_UPME  0x02 /* Uplink Port Mirroring. */
5333 #define IXGBE_MRCTL_DPME  0x04 /* Downlink Port Mirroring. */
5334 #define IXGBE_MRCTL_VLME  0x08 /* VLAN Mirroring. */
5335 #define IXGBE_INVALID_MIRROR_TYPE(mirror_type) \
5336 	((mirror_type) & ~(uint8_t)(ETH_MIRROR_VIRTUAL_POOL_UP | \
5337 	ETH_MIRROR_UPLINK_PORT | ETH_MIRROR_DOWNLINK_PORT | ETH_MIRROR_VLAN))
5338 
5339 static int
5340 ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
5341 		      struct rte_eth_mirror_conf *mirror_conf,
5342 		      uint8_t rule_id, uint8_t on)
5343 {
5344 	uint32_t mr_ctl, vlvf;
5345 	uint32_t mp_lsb = 0;
5346 	uint32_t mv_msb = 0;
5347 	uint32_t mv_lsb = 0;
5348 	uint32_t mp_msb = 0;
5349 	uint8_t i = 0;
5350 	int reg_index = 0;
5351 	uint64_t vlan_mask = 0;
5352 
5353 	const uint8_t pool_mask_offset = 32;
5354 	const uint8_t vlan_mask_offset = 32;
5355 	const uint8_t dst_pool_offset = 8;
5356 	const uint8_t rule_mr_offset  = 4;
5357 	const uint8_t mirror_rule_mask = 0x0F;
5358 
5359 	struct ixgbe_mirror_info *mr_info =
5360 			(IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5361 	struct ixgbe_hw *hw =
5362 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5363 	uint8_t mirror_type = 0;
5364 
5365 	if (ixgbe_vt_check(hw) < 0)
5366 		return -ENOTSUP;
5367 
5368 	if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5369 		return -EINVAL;
5370 
5371 	if (IXGBE_INVALID_MIRROR_TYPE(mirror_conf->rule_type)) {
5372 		PMD_DRV_LOG(ERR, "unsupported mirror type 0x%x.",
5373 			    mirror_conf->rule_type);
5374 		return -EINVAL;
5375 	}
5376 
5377 	if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5378 		mirror_type |= IXGBE_MRCTL_VLME;
5379 		/* Check if vlan id is valid and find conresponding VLAN ID
5380 		 * index in VLVF
5381 		 */
5382 		for (i = 0; i < IXGBE_VLVF_ENTRIES; i++) {
5383 			if (mirror_conf->vlan.vlan_mask & (1ULL << i)) {
5384 				/* search vlan id related pool vlan filter
5385 				 * index
5386 				 */
5387 				reg_index = ixgbe_find_vlvf_slot(
5388 						hw,
5389 						mirror_conf->vlan.vlan_id[i],
5390 						false);
5391 				if (reg_index < 0)
5392 					return -EINVAL;
5393 				vlvf = IXGBE_READ_REG(hw,
5394 						      IXGBE_VLVF(reg_index));
5395 				if ((vlvf & IXGBE_VLVF_VIEN) &&
5396 				    ((vlvf & IXGBE_VLVF_VLANID_MASK) ==
5397 				      mirror_conf->vlan.vlan_id[i]))
5398 					vlan_mask |= (1ULL << reg_index);
5399 				else
5400 					return -EINVAL;
5401 			}
5402 		}
5403 
5404 		if (on) {
5405 			mv_lsb = vlan_mask & 0xFFFFFFFF;
5406 			mv_msb = vlan_mask >> vlan_mask_offset;
5407 
5408 			mr_info->mr_conf[rule_id].vlan.vlan_mask =
5409 						mirror_conf->vlan.vlan_mask;
5410 			for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++) {
5411 				if (mirror_conf->vlan.vlan_mask & (1ULL << i))
5412 					mr_info->mr_conf[rule_id].vlan.vlan_id[i] =
5413 						mirror_conf->vlan.vlan_id[i];
5414 			}
5415 		} else {
5416 			mv_lsb = 0;
5417 			mv_msb = 0;
5418 			mr_info->mr_conf[rule_id].vlan.vlan_mask = 0;
5419 			for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++)
5420 				mr_info->mr_conf[rule_id].vlan.vlan_id[i] = 0;
5421 		}
5422 	}
5423 
5424 	/**
5425 	 * if enable pool mirror, write related pool mask register,if disable
5426 	 * pool mirror, clear PFMRVM register
5427 	 */
5428 	if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5429 		mirror_type |= IXGBE_MRCTL_VPME;
5430 		if (on) {
5431 			mp_lsb = mirror_conf->pool_mask & 0xFFFFFFFF;
5432 			mp_msb = mirror_conf->pool_mask >> pool_mask_offset;
5433 			mr_info->mr_conf[rule_id].pool_mask =
5434 					mirror_conf->pool_mask;
5435 
5436 		} else {
5437 			mp_lsb = 0;
5438 			mp_msb = 0;
5439 			mr_info->mr_conf[rule_id].pool_mask = 0;
5440 		}
5441 	}
5442 	if (mirror_conf->rule_type & ETH_MIRROR_UPLINK_PORT)
5443 		mirror_type |= IXGBE_MRCTL_UPME;
5444 	if (mirror_conf->rule_type & ETH_MIRROR_DOWNLINK_PORT)
5445 		mirror_type |= IXGBE_MRCTL_DPME;
5446 
5447 	/* read  mirror control register and recalculate it */
5448 	mr_ctl = IXGBE_READ_REG(hw, IXGBE_MRCTL(rule_id));
5449 
5450 	if (on) {
5451 		mr_ctl |= mirror_type;
5452 		mr_ctl &= mirror_rule_mask;
5453 		mr_ctl |= mirror_conf->dst_pool << dst_pool_offset;
5454 	} else {
5455 		mr_ctl &= ~(mirror_conf->rule_type & mirror_rule_mask);
5456 	}
5457 
5458 	mr_info->mr_conf[rule_id].rule_type = mirror_conf->rule_type;
5459 	mr_info->mr_conf[rule_id].dst_pool = mirror_conf->dst_pool;
5460 
5461 	/* write mirrror control  register */
5462 	IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5463 
5464 	/* write pool mirrror control  register */
5465 	if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5466 		IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), mp_lsb);
5467 		IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset),
5468 				mp_msb);
5469 	}
5470 	/* write VLAN mirrror control  register */
5471 	if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5472 		IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), mv_lsb);
5473 		IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset),
5474 				mv_msb);
5475 	}
5476 
5477 	return 0;
5478 }
5479 
5480 static int
5481 ixgbe_mirror_rule_reset(struct rte_eth_dev *dev, uint8_t rule_id)
5482 {
5483 	int mr_ctl = 0;
5484 	uint32_t lsb_val = 0;
5485 	uint32_t msb_val = 0;
5486 	const uint8_t rule_mr_offset = 4;
5487 
5488 	struct ixgbe_hw *hw =
5489 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5490 	struct ixgbe_mirror_info *mr_info =
5491 		(IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5492 
5493 	if (ixgbe_vt_check(hw) < 0)
5494 		return -ENOTSUP;
5495 
5496 	if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5497 		return -EINVAL;
5498 
5499 	memset(&mr_info->mr_conf[rule_id], 0,
5500 	       sizeof(struct rte_eth_mirror_conf));
5501 
5502 	/* clear PFVMCTL register */
5503 	IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5504 
5505 	/* clear pool mask register */
5506 	IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), lsb_val);
5507 	IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset), msb_val);
5508 
5509 	/* clear vlan mask register */
5510 	IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), lsb_val);
5511 	IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset), msb_val);
5512 
5513 	return 0;
5514 }
5515 
5516 static int
5517 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5518 {
5519 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5520 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5521 	uint32_t mask;
5522 	struct ixgbe_hw *hw =
5523 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5524 	uint32_t vec = IXGBE_MISC_VEC_ID;
5525 
5526 	mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
5527 	if (rte_intr_allow_others(intr_handle))
5528 		vec = IXGBE_RX_VEC_START;
5529 	mask |= (1 << vec);
5530 	RTE_SET_USED(queue_id);
5531 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
5532 
5533 	rte_intr_enable(intr_handle);
5534 
5535 	return 0;
5536 }
5537 
5538 static int
5539 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5540 {
5541 	uint32_t mask;
5542 	struct ixgbe_hw *hw =
5543 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5544 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5545 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5546 	uint32_t vec = IXGBE_MISC_VEC_ID;
5547 
5548 	mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
5549 	if (rte_intr_allow_others(intr_handle))
5550 		vec = IXGBE_RX_VEC_START;
5551 	mask &= ~(1 << vec);
5552 	RTE_SET_USED(queue_id);
5553 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
5554 
5555 	return 0;
5556 }
5557 
5558 static int
5559 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5560 {
5561 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5562 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5563 	uint32_t mask;
5564 	struct ixgbe_hw *hw =
5565 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5566 	struct ixgbe_interrupt *intr =
5567 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5568 
5569 	if (queue_id < 16) {
5570 		ixgbe_disable_intr(hw);
5571 		intr->mask |= (1 << queue_id);
5572 		ixgbe_enable_intr(dev);
5573 	} else if (queue_id < 32) {
5574 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5575 		mask &= (1 << queue_id);
5576 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5577 	} else if (queue_id < 64) {
5578 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5579 		mask &= (1 << (queue_id - 32));
5580 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5581 	}
5582 	rte_intr_enable(intr_handle);
5583 
5584 	return 0;
5585 }
5586 
5587 static int
5588 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5589 {
5590 	uint32_t mask;
5591 	struct ixgbe_hw *hw =
5592 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5593 	struct ixgbe_interrupt *intr =
5594 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5595 
5596 	if (queue_id < 16) {
5597 		ixgbe_disable_intr(hw);
5598 		intr->mask &= ~(1 << queue_id);
5599 		ixgbe_enable_intr(dev);
5600 	} else if (queue_id < 32) {
5601 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5602 		mask &= ~(1 << queue_id);
5603 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5604 	} else if (queue_id < 64) {
5605 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5606 		mask &= ~(1 << (queue_id - 32));
5607 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5608 	}
5609 
5610 	return 0;
5611 }
5612 
5613 static void
5614 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5615 		     uint8_t queue, uint8_t msix_vector)
5616 {
5617 	uint32_t tmp, idx;
5618 
5619 	if (direction == -1) {
5620 		/* other causes */
5621 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5622 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
5623 		tmp &= ~0xFF;
5624 		tmp |= msix_vector;
5625 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
5626 	} else {
5627 		/* rx or tx cause */
5628 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5629 		idx = ((16 * (queue & 1)) + (8 * direction));
5630 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
5631 		tmp &= ~(0xFF << idx);
5632 		tmp |= (msix_vector << idx);
5633 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
5634 	}
5635 }
5636 
5637 /**
5638  * set the IVAR registers, mapping interrupt causes to vectors
5639  * @param hw
5640  *  pointer to ixgbe_hw struct
5641  * @direction
5642  *  0 for Rx, 1 for Tx, -1 for other causes
5643  * @queue
5644  *  queue to map the corresponding interrupt to
5645  * @msix_vector
5646  *  the vector to map to the corresponding queue
5647  */
5648 static void
5649 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5650 		   uint8_t queue, uint8_t msix_vector)
5651 {
5652 	uint32_t tmp, idx;
5653 
5654 	msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5655 	if (hw->mac.type == ixgbe_mac_82598EB) {
5656 		if (direction == -1)
5657 			direction = 0;
5658 		idx = (((direction * 64) + queue) >> 2) & 0x1F;
5659 		tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
5660 		tmp &= ~(0xFF << (8 * (queue & 0x3)));
5661 		tmp |= (msix_vector << (8 * (queue & 0x3)));
5662 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
5663 	} else if ((hw->mac.type == ixgbe_mac_82599EB) ||
5664 			(hw->mac.type == ixgbe_mac_X540) ||
5665 			(hw->mac.type == ixgbe_mac_X550)) {
5666 		if (direction == -1) {
5667 			/* other causes */
5668 			idx = ((queue & 1) * 8);
5669 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
5670 			tmp &= ~(0xFF << idx);
5671 			tmp |= (msix_vector << idx);
5672 			IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
5673 		} else {
5674 			/* rx or tx causes */
5675 			idx = ((16 * (queue & 1)) + (8 * direction));
5676 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
5677 			tmp &= ~(0xFF << idx);
5678 			tmp |= (msix_vector << idx);
5679 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
5680 		}
5681 	}
5682 }
5683 
5684 static void
5685 ixgbevf_configure_msix(struct rte_eth_dev *dev)
5686 {
5687 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5688 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5689 	struct ixgbe_hw *hw =
5690 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5691 	uint32_t q_idx;
5692 	uint32_t vector_idx = IXGBE_MISC_VEC_ID;
5693 	uint32_t base = IXGBE_MISC_VEC_ID;
5694 
5695 	/* Configure VF other cause ivar */
5696 	ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
5697 
5698 	/* won't configure msix register if no mapping is done
5699 	 * between intr vector and event fd.
5700 	 */
5701 	if (!rte_intr_dp_is_en(intr_handle))
5702 		return;
5703 
5704 	if (rte_intr_allow_others(intr_handle)) {
5705 		base = IXGBE_RX_VEC_START;
5706 		vector_idx = IXGBE_RX_VEC_START;
5707 	}
5708 
5709 	/* Configure all RX queues of VF */
5710 	for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
5711 		/* Force all queue use vector 0,
5712 		 * as IXGBE_VF_MAXMSIVECOTR = 1
5713 		 */
5714 		ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
5715 		intr_handle->intr_vec[q_idx] = vector_idx;
5716 		if (vector_idx < base + intr_handle->nb_efd - 1)
5717 			vector_idx++;
5718 	}
5719 }
5720 
5721 /**
5722  * Sets up the hardware to properly generate MSI-X interrupts
5723  * @hw
5724  *  board private structure
5725  */
5726 static void
5727 ixgbe_configure_msix(struct rte_eth_dev *dev)
5728 {
5729 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5730 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5731 	struct ixgbe_hw *hw =
5732 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5733 	uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
5734 	uint32_t vec = IXGBE_MISC_VEC_ID;
5735 	uint32_t mask;
5736 	uint32_t gpie;
5737 
5738 	/* won't configure msix register if no mapping is done
5739 	 * between intr vector and event fd
5740 	 */
5741 	if (!rte_intr_dp_is_en(intr_handle))
5742 		return;
5743 
5744 	if (rte_intr_allow_others(intr_handle))
5745 		vec = base = IXGBE_RX_VEC_START;
5746 
5747 	/* setup GPIE for MSI-x mode */
5748 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5749 	gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5750 		IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
5751 	/* auto clearing and auto setting corresponding bits in EIMS
5752 	 * when MSI-X interrupt is triggered
5753 	 */
5754 	if (hw->mac.type == ixgbe_mac_82598EB) {
5755 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5756 	} else {
5757 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5758 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5759 	}
5760 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5761 
5762 	/* Populate the IVAR table and set the ITR values to the
5763 	 * corresponding register.
5764 	 */
5765 	for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
5766 	     queue_id++) {
5767 		/* by default, 1:1 mapping */
5768 		ixgbe_set_ivar_map(hw, 0, queue_id, vec);
5769 		intr_handle->intr_vec[queue_id] = vec;
5770 		if (vec < base + intr_handle->nb_efd - 1)
5771 			vec++;
5772 	}
5773 
5774 	switch (hw->mac.type) {
5775 	case ixgbe_mac_82598EB:
5776 		ixgbe_set_ivar_map(hw, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
5777 				   IXGBE_MISC_VEC_ID);
5778 		break;
5779 	case ixgbe_mac_82599EB:
5780 	case ixgbe_mac_X540:
5781 	case ixgbe_mac_X550:
5782 		ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
5783 		break;
5784 	default:
5785 		break;
5786 	}
5787 	IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
5788 			IXGBE_MIN_INTER_INTERRUPT_INTERVAL_DEFAULT & 0xFFF);
5789 
5790 	/* set up to autoclear timer, and the vectors */
5791 	mask = IXGBE_EIMS_ENABLE_MASK;
5792 	mask &= ~(IXGBE_EIMS_OTHER |
5793 		  IXGBE_EIMS_MAILBOX |
5794 		  IXGBE_EIMS_LSC);
5795 
5796 	IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
5797 }
5798 
5799 int
5800 ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
5801 			   uint16_t queue_idx, uint16_t tx_rate)
5802 {
5803 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5804 	struct rte_eth_rxmode *rxmode;
5805 	uint32_t rf_dec, rf_int;
5806 	uint32_t bcnrc_val;
5807 	uint16_t link_speed = dev->data->dev_link.link_speed;
5808 
5809 	if (queue_idx >= hw->mac.max_tx_queues)
5810 		return -EINVAL;
5811 
5812 	if (tx_rate != 0) {
5813 		/* Calculate the rate factor values to set */
5814 		rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
5815 		rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
5816 		rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
5817 
5818 		bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
5819 		bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
5820 				IXGBE_RTTBCNRC_RF_INT_MASK_M);
5821 		bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
5822 	} else {
5823 		bcnrc_val = 0;
5824 	}
5825 
5826 	rxmode = &dev->data->dev_conf.rxmode;
5827 	/*
5828 	 * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
5829 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
5830 	 * set as 0x4.
5831 	 */
5832 	if ((rxmode->offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) &&
5833 	    (rxmode->max_rx_pkt_len >= IXGBE_MAX_JUMBO_FRAME_SIZE))
5834 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
5835 			IXGBE_MMW_SIZE_JUMBO_FRAME);
5836 	else
5837 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
5838 			IXGBE_MMW_SIZE_DEFAULT);
5839 
5840 	/* Set RTTBCNRC of queue X */
5841 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
5842 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
5843 	IXGBE_WRITE_FLUSH(hw);
5844 
5845 	return 0;
5846 }
5847 
5848 static int
5849 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
5850 		     __attribute__((unused)) uint32_t index,
5851 		     __attribute__((unused)) uint32_t pool)
5852 {
5853 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5854 	int diag;
5855 
5856 	/*
5857 	 * On a 82599 VF, adding again the same MAC addr is not an idempotent
5858 	 * operation. Trap this case to avoid exhausting the [very limited]
5859 	 * set of PF resources used to store VF MAC addresses.
5860 	 */
5861 	if (memcmp(hw->mac.perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
5862 		return -1;
5863 	diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
5864 	if (diag != 0)
5865 		PMD_DRV_LOG(ERR, "Unable to add MAC address "
5866 			    "%02x:%02x:%02x:%02x:%02x:%02x - diag=%d",
5867 			    mac_addr->addr_bytes[0],
5868 			    mac_addr->addr_bytes[1],
5869 			    mac_addr->addr_bytes[2],
5870 			    mac_addr->addr_bytes[3],
5871 			    mac_addr->addr_bytes[4],
5872 			    mac_addr->addr_bytes[5],
5873 			    diag);
5874 	return diag;
5875 }
5876 
5877 static void
5878 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
5879 {
5880 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5881 	struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
5882 	struct ether_addr *mac_addr;
5883 	uint32_t i;
5884 	int diag;
5885 
5886 	/*
5887 	 * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
5888 	 * not support the deletion of a given MAC address.
5889 	 * Instead, it imposes to delete all MAC addresses, then to add again
5890 	 * all MAC addresses with the exception of the one to be deleted.
5891 	 */
5892 	(void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
5893 
5894 	/*
5895 	 * Add again all MAC addresses, with the exception of the deleted one
5896 	 * and of the permanent MAC address.
5897 	 */
5898 	for (i = 0, mac_addr = dev->data->mac_addrs;
5899 	     i < hw->mac.num_rar_entries; i++, mac_addr++) {
5900 		/* Skip the deleted MAC address */
5901 		if (i == index)
5902 			continue;
5903 		/* Skip NULL MAC addresses */
5904 		if (is_zero_ether_addr(mac_addr))
5905 			continue;
5906 		/* Skip the permanent MAC address */
5907 		if (memcmp(perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
5908 			continue;
5909 		diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
5910 		if (diag != 0)
5911 			PMD_DRV_LOG(ERR,
5912 				    "Adding again MAC address "
5913 				    "%02x:%02x:%02x:%02x:%02x:%02x failed "
5914 				    "diag=%d",
5915 				    mac_addr->addr_bytes[0],
5916 				    mac_addr->addr_bytes[1],
5917 				    mac_addr->addr_bytes[2],
5918 				    mac_addr->addr_bytes[3],
5919 				    mac_addr->addr_bytes[4],
5920 				    mac_addr->addr_bytes[5],
5921 				    diag);
5922 	}
5923 }
5924 
5925 static void
5926 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
5927 {
5928 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5929 
5930 	hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
5931 }
5932 
5933 int
5934 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
5935 			struct rte_eth_syn_filter *filter,
5936 			bool add)
5937 {
5938 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5939 	struct ixgbe_filter_info *filter_info =
5940 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5941 	uint32_t syn_info;
5942 	uint32_t synqf;
5943 
5944 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
5945 		return -EINVAL;
5946 
5947 	syn_info = filter_info->syn_info;
5948 
5949 	if (add) {
5950 		if (syn_info & IXGBE_SYN_FILTER_ENABLE)
5951 			return -EINVAL;
5952 		synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
5953 			IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
5954 
5955 		if (filter->hig_pri)
5956 			synqf |= IXGBE_SYN_FILTER_SYNQFP;
5957 		else
5958 			synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
5959 	} else {
5960 		synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
5961 		if (!(syn_info & IXGBE_SYN_FILTER_ENABLE))
5962 			return -ENOENT;
5963 		synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
5964 	}
5965 
5966 	filter_info->syn_info = synqf;
5967 	IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
5968 	IXGBE_WRITE_FLUSH(hw);
5969 	return 0;
5970 }
5971 
5972 static int
5973 ixgbe_syn_filter_get(struct rte_eth_dev *dev,
5974 			struct rte_eth_syn_filter *filter)
5975 {
5976 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5977 	uint32_t synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
5978 
5979 	if (synqf & IXGBE_SYN_FILTER_ENABLE) {
5980 		filter->hig_pri = (synqf & IXGBE_SYN_FILTER_SYNQFP) ? 1 : 0;
5981 		filter->queue = (uint16_t)((synqf & IXGBE_SYN_FILTER_QUEUE) >> 1);
5982 		return 0;
5983 	}
5984 	return -ENOENT;
5985 }
5986 
5987 static int
5988 ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
5989 			enum rte_filter_op filter_op,
5990 			void *arg)
5991 {
5992 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5993 	int ret;
5994 
5995 	MAC_TYPE_FILTER_SUP(hw->mac.type);
5996 
5997 	if (filter_op == RTE_ETH_FILTER_NOP)
5998 		return 0;
5999 
6000 	if (arg == NULL) {
6001 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
6002 			    filter_op);
6003 		return -EINVAL;
6004 	}
6005 
6006 	switch (filter_op) {
6007 	case RTE_ETH_FILTER_ADD:
6008 		ret = ixgbe_syn_filter_set(dev,
6009 				(struct rte_eth_syn_filter *)arg,
6010 				TRUE);
6011 		break;
6012 	case RTE_ETH_FILTER_DELETE:
6013 		ret = ixgbe_syn_filter_set(dev,
6014 				(struct rte_eth_syn_filter *)arg,
6015 				FALSE);
6016 		break;
6017 	case RTE_ETH_FILTER_GET:
6018 		ret = ixgbe_syn_filter_get(dev,
6019 				(struct rte_eth_syn_filter *)arg);
6020 		break;
6021 	default:
6022 		PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
6023 		ret = -EINVAL;
6024 		break;
6025 	}
6026 
6027 	return ret;
6028 }
6029 
6030 
6031 static inline enum ixgbe_5tuple_protocol
6032 convert_protocol_type(uint8_t protocol_value)
6033 {
6034 	if (protocol_value == IPPROTO_TCP)
6035 		return IXGBE_FILTER_PROTOCOL_TCP;
6036 	else if (protocol_value == IPPROTO_UDP)
6037 		return IXGBE_FILTER_PROTOCOL_UDP;
6038 	else if (protocol_value == IPPROTO_SCTP)
6039 		return IXGBE_FILTER_PROTOCOL_SCTP;
6040 	else
6041 		return IXGBE_FILTER_PROTOCOL_NONE;
6042 }
6043 
6044 /* inject a 5-tuple filter to HW */
6045 static inline void
6046 ixgbe_inject_5tuple_filter(struct rte_eth_dev *dev,
6047 			   struct ixgbe_5tuple_filter *filter)
6048 {
6049 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6050 	int i;
6051 	uint32_t ftqf, sdpqf;
6052 	uint32_t l34timir = 0;
6053 	uint8_t mask = 0xff;
6054 
6055 	i = filter->index;
6056 
6057 	sdpqf = (uint32_t)(filter->filter_info.dst_port <<
6058 				IXGBE_SDPQF_DSTPORT_SHIFT);
6059 	sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
6060 
6061 	ftqf = (uint32_t)(filter->filter_info.proto &
6062 		IXGBE_FTQF_PROTOCOL_MASK);
6063 	ftqf |= (uint32_t)((filter->filter_info.priority &
6064 		IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
6065 	if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
6066 		mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
6067 	if (filter->filter_info.dst_ip_mask == 0)
6068 		mask &= IXGBE_FTQF_DEST_ADDR_MASK;
6069 	if (filter->filter_info.src_port_mask == 0)
6070 		mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
6071 	if (filter->filter_info.dst_port_mask == 0)
6072 		mask &= IXGBE_FTQF_DEST_PORT_MASK;
6073 	if (filter->filter_info.proto_mask == 0)
6074 		mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
6075 	ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
6076 	ftqf |= IXGBE_FTQF_POOL_MASK_EN;
6077 	ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
6078 
6079 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
6080 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
6081 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
6082 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
6083 
6084 	l34timir |= IXGBE_L34T_IMIR_RESERVE;
6085 	l34timir |= (uint32_t)(filter->queue <<
6086 				IXGBE_L34T_IMIR_QUEUE_SHIFT);
6087 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
6088 }
6089 
6090 /*
6091  * add a 5tuple filter
6092  *
6093  * @param
6094  * dev: Pointer to struct rte_eth_dev.
6095  * index: the index the filter allocates.
6096  * filter: ponter to the filter that will be added.
6097  * rx_queue: the queue id the filter assigned to.
6098  *
6099  * @return
6100  *    - On success, zero.
6101  *    - On failure, a negative value.
6102  */
6103 static int
6104 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
6105 			struct ixgbe_5tuple_filter *filter)
6106 {
6107 	struct ixgbe_filter_info *filter_info =
6108 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6109 	int i, idx, shift;
6110 
6111 	/*
6112 	 * look for an unused 5tuple filter index,
6113 	 * and insert the filter to list.
6114 	 */
6115 	for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
6116 		idx = i / (sizeof(uint32_t) * NBBY);
6117 		shift = i % (sizeof(uint32_t) * NBBY);
6118 		if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
6119 			filter_info->fivetuple_mask[idx] |= 1 << shift;
6120 			filter->index = i;
6121 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
6122 					  filter,
6123 					  entries);
6124 			break;
6125 		}
6126 	}
6127 	if (i >= IXGBE_MAX_FTQF_FILTERS) {
6128 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
6129 		return -ENOSYS;
6130 	}
6131 
6132 	ixgbe_inject_5tuple_filter(dev, filter);
6133 
6134 	return 0;
6135 }
6136 
6137 /*
6138  * remove a 5tuple filter
6139  *
6140  * @param
6141  * dev: Pointer to struct rte_eth_dev.
6142  * filter: the pointer of the filter will be removed.
6143  */
6144 static void
6145 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
6146 			struct ixgbe_5tuple_filter *filter)
6147 {
6148 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6149 	struct ixgbe_filter_info *filter_info =
6150 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6151 	uint16_t index = filter->index;
6152 
6153 	filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
6154 				~(1 << (index % (sizeof(uint32_t) * NBBY)));
6155 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
6156 	rte_free(filter);
6157 
6158 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
6159 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
6160 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
6161 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
6162 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
6163 }
6164 
6165 static int
6166 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
6167 {
6168 	struct ixgbe_hw *hw;
6169 	uint32_t max_frame = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
6170 	struct rte_eth_rxmode *rx_conf = &dev->data->dev_conf.rxmode;
6171 
6172 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6173 
6174 	if ((mtu < ETHER_MIN_MTU) || (max_frame > ETHER_MAX_JUMBO_FRAME_LEN))
6175 		return -EINVAL;
6176 
6177 	/* refuse mtu that requires the support of scattered packets when this
6178 	 * feature has not been enabled before.
6179 	 */
6180 	if (!(rx_conf->offloads & DEV_RX_OFFLOAD_SCATTER) &&
6181 	    (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
6182 	     dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM))
6183 		return -EINVAL;
6184 
6185 	/*
6186 	 * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
6187 	 * request of the version 2.0 of the mailbox API.
6188 	 * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
6189 	 * of the mailbox API.
6190 	 * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
6191 	 * prior to 3.11.33 which contains the following change:
6192 	 * "ixgbe: Enable jumbo frames support w/ SR-IOV"
6193 	 */
6194 	ixgbevf_rlpml_set_vf(hw, max_frame);
6195 
6196 	/* update max frame size */
6197 	dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame;
6198 	return 0;
6199 }
6200 
6201 static inline struct ixgbe_5tuple_filter *
6202 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
6203 			struct ixgbe_5tuple_filter_info *key)
6204 {
6205 	struct ixgbe_5tuple_filter *it;
6206 
6207 	TAILQ_FOREACH(it, filter_list, entries) {
6208 		if (memcmp(key, &it->filter_info,
6209 			sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
6210 			return it;
6211 		}
6212 	}
6213 	return NULL;
6214 }
6215 
6216 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
6217 static inline int
6218 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
6219 			struct ixgbe_5tuple_filter_info *filter_info)
6220 {
6221 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
6222 		filter->priority > IXGBE_5TUPLE_MAX_PRI ||
6223 		filter->priority < IXGBE_5TUPLE_MIN_PRI)
6224 		return -EINVAL;
6225 
6226 	switch (filter->dst_ip_mask) {
6227 	case UINT32_MAX:
6228 		filter_info->dst_ip_mask = 0;
6229 		filter_info->dst_ip = filter->dst_ip;
6230 		break;
6231 	case 0:
6232 		filter_info->dst_ip_mask = 1;
6233 		break;
6234 	default:
6235 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
6236 		return -EINVAL;
6237 	}
6238 
6239 	switch (filter->src_ip_mask) {
6240 	case UINT32_MAX:
6241 		filter_info->src_ip_mask = 0;
6242 		filter_info->src_ip = filter->src_ip;
6243 		break;
6244 	case 0:
6245 		filter_info->src_ip_mask = 1;
6246 		break;
6247 	default:
6248 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
6249 		return -EINVAL;
6250 	}
6251 
6252 	switch (filter->dst_port_mask) {
6253 	case UINT16_MAX:
6254 		filter_info->dst_port_mask = 0;
6255 		filter_info->dst_port = filter->dst_port;
6256 		break;
6257 	case 0:
6258 		filter_info->dst_port_mask = 1;
6259 		break;
6260 	default:
6261 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
6262 		return -EINVAL;
6263 	}
6264 
6265 	switch (filter->src_port_mask) {
6266 	case UINT16_MAX:
6267 		filter_info->src_port_mask = 0;
6268 		filter_info->src_port = filter->src_port;
6269 		break;
6270 	case 0:
6271 		filter_info->src_port_mask = 1;
6272 		break;
6273 	default:
6274 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
6275 		return -EINVAL;
6276 	}
6277 
6278 	switch (filter->proto_mask) {
6279 	case UINT8_MAX:
6280 		filter_info->proto_mask = 0;
6281 		filter_info->proto =
6282 			convert_protocol_type(filter->proto);
6283 		break;
6284 	case 0:
6285 		filter_info->proto_mask = 1;
6286 		break;
6287 	default:
6288 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
6289 		return -EINVAL;
6290 	}
6291 
6292 	filter_info->priority = (uint8_t)filter->priority;
6293 	return 0;
6294 }
6295 
6296 /*
6297  * add or delete a ntuple filter
6298  *
6299  * @param
6300  * dev: Pointer to struct rte_eth_dev.
6301  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6302  * add: if true, add filter, if false, remove filter
6303  *
6304  * @return
6305  *    - On success, zero.
6306  *    - On failure, a negative value.
6307  */
6308 int
6309 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
6310 			struct rte_eth_ntuple_filter *ntuple_filter,
6311 			bool add)
6312 {
6313 	struct ixgbe_filter_info *filter_info =
6314 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6315 	struct ixgbe_5tuple_filter_info filter_5tuple;
6316 	struct ixgbe_5tuple_filter *filter;
6317 	int ret;
6318 
6319 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6320 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6321 		return -EINVAL;
6322 	}
6323 
6324 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6325 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6326 	if (ret < 0)
6327 		return ret;
6328 
6329 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6330 					 &filter_5tuple);
6331 	if (filter != NULL && add) {
6332 		PMD_DRV_LOG(ERR, "filter exists.");
6333 		return -EEXIST;
6334 	}
6335 	if (filter == NULL && !add) {
6336 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
6337 		return -ENOENT;
6338 	}
6339 
6340 	if (add) {
6341 		filter = rte_zmalloc("ixgbe_5tuple_filter",
6342 				sizeof(struct ixgbe_5tuple_filter), 0);
6343 		if (filter == NULL)
6344 			return -ENOMEM;
6345 		rte_memcpy(&filter->filter_info,
6346 				 &filter_5tuple,
6347 				 sizeof(struct ixgbe_5tuple_filter_info));
6348 		filter->queue = ntuple_filter->queue;
6349 		ret = ixgbe_add_5tuple_filter(dev, filter);
6350 		if (ret < 0) {
6351 			rte_free(filter);
6352 			return ret;
6353 		}
6354 	} else
6355 		ixgbe_remove_5tuple_filter(dev, filter);
6356 
6357 	return 0;
6358 }
6359 
6360 /*
6361  * get a ntuple filter
6362  *
6363  * @param
6364  * dev: Pointer to struct rte_eth_dev.
6365  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6366  *
6367  * @return
6368  *    - On success, zero.
6369  *    - On failure, a negative value.
6370  */
6371 static int
6372 ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
6373 			struct rte_eth_ntuple_filter *ntuple_filter)
6374 {
6375 	struct ixgbe_filter_info *filter_info =
6376 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6377 	struct ixgbe_5tuple_filter_info filter_5tuple;
6378 	struct ixgbe_5tuple_filter *filter;
6379 	int ret;
6380 
6381 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6382 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6383 		return -EINVAL;
6384 	}
6385 
6386 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6387 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6388 	if (ret < 0)
6389 		return ret;
6390 
6391 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6392 					 &filter_5tuple);
6393 	if (filter == NULL) {
6394 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
6395 		return -ENOENT;
6396 	}
6397 	ntuple_filter->queue = filter->queue;
6398 	return 0;
6399 }
6400 
6401 /*
6402  * ixgbe_ntuple_filter_handle - Handle operations for ntuple filter.
6403  * @dev: pointer to rte_eth_dev structure
6404  * @filter_op:operation will be taken.
6405  * @arg: a pointer to specific structure corresponding to the filter_op
6406  *
6407  * @return
6408  *    - On success, zero.
6409  *    - On failure, a negative value.
6410  */
6411 static int
6412 ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
6413 				enum rte_filter_op filter_op,
6414 				void *arg)
6415 {
6416 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6417 	int ret;
6418 
6419 	MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
6420 
6421 	if (filter_op == RTE_ETH_FILTER_NOP)
6422 		return 0;
6423 
6424 	if (arg == NULL) {
6425 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
6426 			    filter_op);
6427 		return -EINVAL;
6428 	}
6429 
6430 	switch (filter_op) {
6431 	case RTE_ETH_FILTER_ADD:
6432 		ret = ixgbe_add_del_ntuple_filter(dev,
6433 			(struct rte_eth_ntuple_filter *)arg,
6434 			TRUE);
6435 		break;
6436 	case RTE_ETH_FILTER_DELETE:
6437 		ret = ixgbe_add_del_ntuple_filter(dev,
6438 			(struct rte_eth_ntuple_filter *)arg,
6439 			FALSE);
6440 		break;
6441 	case RTE_ETH_FILTER_GET:
6442 		ret = ixgbe_get_ntuple_filter(dev,
6443 			(struct rte_eth_ntuple_filter *)arg);
6444 		break;
6445 	default:
6446 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
6447 		ret = -EINVAL;
6448 		break;
6449 	}
6450 	return ret;
6451 }
6452 
6453 int
6454 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
6455 			struct rte_eth_ethertype_filter *filter,
6456 			bool add)
6457 {
6458 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6459 	struct ixgbe_filter_info *filter_info =
6460 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6461 	uint32_t etqf = 0;
6462 	uint32_t etqs = 0;
6463 	int ret;
6464 	struct ixgbe_ethertype_filter ethertype_filter;
6465 
6466 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6467 		return -EINVAL;
6468 
6469 	if (filter->ether_type == ETHER_TYPE_IPv4 ||
6470 		filter->ether_type == ETHER_TYPE_IPv6) {
6471 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
6472 			" ethertype filter.", filter->ether_type);
6473 		return -EINVAL;
6474 	}
6475 
6476 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
6477 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
6478 		return -EINVAL;
6479 	}
6480 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
6481 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
6482 		return -EINVAL;
6483 	}
6484 
6485 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6486 	if (ret >= 0 && add) {
6487 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
6488 			    filter->ether_type);
6489 		return -EEXIST;
6490 	}
6491 	if (ret < 0 && !add) {
6492 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6493 			    filter->ether_type);
6494 		return -ENOENT;
6495 	}
6496 
6497 	if (add) {
6498 		etqf = IXGBE_ETQF_FILTER_EN;
6499 		etqf |= (uint32_t)filter->ether_type;
6500 		etqs |= (uint32_t)((filter->queue <<
6501 				    IXGBE_ETQS_RX_QUEUE_SHIFT) &
6502 				    IXGBE_ETQS_RX_QUEUE);
6503 		etqs |= IXGBE_ETQS_QUEUE_EN;
6504 
6505 		ethertype_filter.ethertype = filter->ether_type;
6506 		ethertype_filter.etqf = etqf;
6507 		ethertype_filter.etqs = etqs;
6508 		ethertype_filter.conf = FALSE;
6509 		ret = ixgbe_ethertype_filter_insert(filter_info,
6510 						    &ethertype_filter);
6511 		if (ret < 0) {
6512 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
6513 			return -ENOSPC;
6514 		}
6515 	} else {
6516 		ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
6517 		if (ret < 0)
6518 			return -ENOSYS;
6519 	}
6520 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
6521 	IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
6522 	IXGBE_WRITE_FLUSH(hw);
6523 
6524 	return 0;
6525 }
6526 
6527 static int
6528 ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
6529 			struct rte_eth_ethertype_filter *filter)
6530 {
6531 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6532 	struct ixgbe_filter_info *filter_info =
6533 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6534 	uint32_t etqf, etqs;
6535 	int ret;
6536 
6537 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6538 	if (ret < 0) {
6539 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6540 			    filter->ether_type);
6541 		return -ENOENT;
6542 	}
6543 
6544 	etqf = IXGBE_READ_REG(hw, IXGBE_ETQF(ret));
6545 	if (etqf & IXGBE_ETQF_FILTER_EN) {
6546 		etqs = IXGBE_READ_REG(hw, IXGBE_ETQS(ret));
6547 		filter->ether_type = etqf & IXGBE_ETQF_ETHERTYPE;
6548 		filter->flags = 0;
6549 		filter->queue = (etqs & IXGBE_ETQS_RX_QUEUE) >>
6550 			       IXGBE_ETQS_RX_QUEUE_SHIFT;
6551 		return 0;
6552 	}
6553 	return -ENOENT;
6554 }
6555 
6556 /*
6557  * ixgbe_ethertype_filter_handle - Handle operations for ethertype filter.
6558  * @dev: pointer to rte_eth_dev structure
6559  * @filter_op:operation will be taken.
6560  * @arg: a pointer to specific structure corresponding to the filter_op
6561  */
6562 static int
6563 ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
6564 				enum rte_filter_op filter_op,
6565 				void *arg)
6566 {
6567 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6568 	int ret;
6569 
6570 	MAC_TYPE_FILTER_SUP(hw->mac.type);
6571 
6572 	if (filter_op == RTE_ETH_FILTER_NOP)
6573 		return 0;
6574 
6575 	if (arg == NULL) {
6576 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
6577 			    filter_op);
6578 		return -EINVAL;
6579 	}
6580 
6581 	switch (filter_op) {
6582 	case RTE_ETH_FILTER_ADD:
6583 		ret = ixgbe_add_del_ethertype_filter(dev,
6584 			(struct rte_eth_ethertype_filter *)arg,
6585 			TRUE);
6586 		break;
6587 	case RTE_ETH_FILTER_DELETE:
6588 		ret = ixgbe_add_del_ethertype_filter(dev,
6589 			(struct rte_eth_ethertype_filter *)arg,
6590 			FALSE);
6591 		break;
6592 	case RTE_ETH_FILTER_GET:
6593 		ret = ixgbe_get_ethertype_filter(dev,
6594 			(struct rte_eth_ethertype_filter *)arg);
6595 		break;
6596 	default:
6597 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
6598 		ret = -EINVAL;
6599 		break;
6600 	}
6601 	return ret;
6602 }
6603 
6604 static int
6605 ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
6606 		     enum rte_filter_type filter_type,
6607 		     enum rte_filter_op filter_op,
6608 		     void *arg)
6609 {
6610 	int ret = 0;
6611 
6612 	switch (filter_type) {
6613 	case RTE_ETH_FILTER_NTUPLE:
6614 		ret = ixgbe_ntuple_filter_handle(dev, filter_op, arg);
6615 		break;
6616 	case RTE_ETH_FILTER_ETHERTYPE:
6617 		ret = ixgbe_ethertype_filter_handle(dev, filter_op, arg);
6618 		break;
6619 	case RTE_ETH_FILTER_SYN:
6620 		ret = ixgbe_syn_filter_handle(dev, filter_op, arg);
6621 		break;
6622 	case RTE_ETH_FILTER_FDIR:
6623 		ret = ixgbe_fdir_ctrl_func(dev, filter_op, arg);
6624 		break;
6625 	case RTE_ETH_FILTER_L2_TUNNEL:
6626 		ret = ixgbe_dev_l2_tunnel_filter_handle(dev, filter_op, arg);
6627 		break;
6628 	case RTE_ETH_FILTER_GENERIC:
6629 		if (filter_op != RTE_ETH_FILTER_GET)
6630 			return -EINVAL;
6631 		*(const void **)arg = &ixgbe_flow_ops;
6632 		break;
6633 	default:
6634 		PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
6635 							filter_type);
6636 		ret = -EINVAL;
6637 		break;
6638 	}
6639 
6640 	return ret;
6641 }
6642 
6643 static u8 *
6644 ixgbe_dev_addr_list_itr(__attribute__((unused)) struct ixgbe_hw *hw,
6645 			u8 **mc_addr_ptr, u32 *vmdq)
6646 {
6647 	u8 *mc_addr;
6648 
6649 	*vmdq = 0;
6650 	mc_addr = *mc_addr_ptr;
6651 	*mc_addr_ptr = (mc_addr + sizeof(struct ether_addr));
6652 	return mc_addr;
6653 }
6654 
6655 static int
6656 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
6657 			  struct ether_addr *mc_addr_set,
6658 			  uint32_t nb_mc_addr)
6659 {
6660 	struct ixgbe_hw *hw;
6661 	u8 *mc_addr_list;
6662 
6663 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6664 	mc_addr_list = (u8 *)mc_addr_set;
6665 	return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
6666 					 ixgbe_dev_addr_list_itr, TRUE);
6667 }
6668 
6669 static uint64_t
6670 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
6671 {
6672 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6673 	uint64_t systime_cycles;
6674 
6675 	switch (hw->mac.type) {
6676 	case ixgbe_mac_X550:
6677 	case ixgbe_mac_X550EM_x:
6678 	case ixgbe_mac_X550EM_a:
6679 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
6680 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6681 		systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6682 				* NSEC_PER_SEC;
6683 		break;
6684 	default:
6685 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6686 		systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6687 				<< 32;
6688 	}
6689 
6690 	return systime_cycles;
6691 }
6692 
6693 static uint64_t
6694 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6695 {
6696 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6697 	uint64_t rx_tstamp_cycles;
6698 
6699 	switch (hw->mac.type) {
6700 	case ixgbe_mac_X550:
6701 	case ixgbe_mac_X550EM_x:
6702 	case ixgbe_mac_X550EM_a:
6703 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6704 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6705 		rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6706 				* NSEC_PER_SEC;
6707 		break;
6708 	default:
6709 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6710 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6711 		rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6712 				<< 32;
6713 	}
6714 
6715 	return rx_tstamp_cycles;
6716 }
6717 
6718 static uint64_t
6719 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6720 {
6721 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6722 	uint64_t tx_tstamp_cycles;
6723 
6724 	switch (hw->mac.type) {
6725 	case ixgbe_mac_X550:
6726 	case ixgbe_mac_X550EM_x:
6727 	case ixgbe_mac_X550EM_a:
6728 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6729 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6730 		tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6731 				* NSEC_PER_SEC;
6732 		break;
6733 	default:
6734 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6735 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6736 		tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6737 				<< 32;
6738 	}
6739 
6740 	return tx_tstamp_cycles;
6741 }
6742 
6743 static void
6744 ixgbe_start_timecounters(struct rte_eth_dev *dev)
6745 {
6746 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6747 	struct ixgbe_adapter *adapter =
6748 		(struct ixgbe_adapter *)dev->data->dev_private;
6749 	struct rte_eth_link link;
6750 	uint32_t incval = 0;
6751 	uint32_t shift = 0;
6752 
6753 	/* Get current link speed. */
6754 	ixgbe_dev_link_update(dev, 1);
6755 	rte_eth_linkstatus_get(dev, &link);
6756 
6757 	switch (link.link_speed) {
6758 	case ETH_SPEED_NUM_100M:
6759 		incval = IXGBE_INCVAL_100;
6760 		shift = IXGBE_INCVAL_SHIFT_100;
6761 		break;
6762 	case ETH_SPEED_NUM_1G:
6763 		incval = IXGBE_INCVAL_1GB;
6764 		shift = IXGBE_INCVAL_SHIFT_1GB;
6765 		break;
6766 	case ETH_SPEED_NUM_10G:
6767 	default:
6768 		incval = IXGBE_INCVAL_10GB;
6769 		shift = IXGBE_INCVAL_SHIFT_10GB;
6770 		break;
6771 	}
6772 
6773 	switch (hw->mac.type) {
6774 	case ixgbe_mac_X550:
6775 	case ixgbe_mac_X550EM_x:
6776 	case ixgbe_mac_X550EM_a:
6777 		/* Independent of link speed. */
6778 		incval = 1;
6779 		/* Cycles read will be interpreted as ns. */
6780 		shift = 0;
6781 		/* Fall-through */
6782 	case ixgbe_mac_X540:
6783 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
6784 		break;
6785 	case ixgbe_mac_82599EB:
6786 		incval >>= IXGBE_INCVAL_SHIFT_82599;
6787 		shift -= IXGBE_INCVAL_SHIFT_82599;
6788 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
6789 				(1 << IXGBE_INCPER_SHIFT_82599) | incval);
6790 		break;
6791 	default:
6792 		/* Not supported. */
6793 		return;
6794 	}
6795 
6796 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
6797 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6798 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6799 
6800 	adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6801 	adapter->systime_tc.cc_shift = shift;
6802 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
6803 
6804 	adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6805 	adapter->rx_tstamp_tc.cc_shift = shift;
6806 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6807 
6808 	adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6809 	adapter->tx_tstamp_tc.cc_shift = shift;
6810 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6811 }
6812 
6813 static int
6814 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
6815 {
6816 	struct ixgbe_adapter *adapter =
6817 			(struct ixgbe_adapter *)dev->data->dev_private;
6818 
6819 	adapter->systime_tc.nsec += delta;
6820 	adapter->rx_tstamp_tc.nsec += delta;
6821 	adapter->tx_tstamp_tc.nsec += delta;
6822 
6823 	return 0;
6824 }
6825 
6826 static int
6827 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
6828 {
6829 	uint64_t ns;
6830 	struct ixgbe_adapter *adapter =
6831 			(struct ixgbe_adapter *)dev->data->dev_private;
6832 
6833 	ns = rte_timespec_to_ns(ts);
6834 	/* Set the timecounters to a new value. */
6835 	adapter->systime_tc.nsec = ns;
6836 	adapter->rx_tstamp_tc.nsec = ns;
6837 	adapter->tx_tstamp_tc.nsec = ns;
6838 
6839 	return 0;
6840 }
6841 
6842 static int
6843 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
6844 {
6845 	uint64_t ns, systime_cycles;
6846 	struct ixgbe_adapter *adapter =
6847 			(struct ixgbe_adapter *)dev->data->dev_private;
6848 
6849 	systime_cycles = ixgbe_read_systime_cyclecounter(dev);
6850 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
6851 	*ts = rte_ns_to_timespec(ns);
6852 
6853 	return 0;
6854 }
6855 
6856 static int
6857 ixgbe_timesync_enable(struct rte_eth_dev *dev)
6858 {
6859 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6860 	uint32_t tsync_ctl;
6861 	uint32_t tsauxc;
6862 
6863 	/* Stop the timesync system time. */
6864 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
6865 	/* Reset the timesync system time value. */
6866 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
6867 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
6868 
6869 	/* Enable system time for platforms where it isn't on by default. */
6870 	tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
6871 	tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
6872 	IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
6873 
6874 	ixgbe_start_timecounters(dev);
6875 
6876 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6877 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
6878 			(ETHER_TYPE_1588 |
6879 			 IXGBE_ETQF_FILTER_EN |
6880 			 IXGBE_ETQF_1588));
6881 
6882 	/* Enable timestamping of received PTP packets. */
6883 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6884 	tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
6885 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6886 
6887 	/* Enable timestamping of transmitted PTP packets. */
6888 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6889 	tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
6890 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6891 
6892 	IXGBE_WRITE_FLUSH(hw);
6893 
6894 	return 0;
6895 }
6896 
6897 static int
6898 ixgbe_timesync_disable(struct rte_eth_dev *dev)
6899 {
6900 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6901 	uint32_t tsync_ctl;
6902 
6903 	/* Disable timestamping of transmitted PTP packets. */
6904 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6905 	tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
6906 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6907 
6908 	/* Disable timestamping of received PTP packets. */
6909 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6910 	tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
6911 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6912 
6913 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6914 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
6915 
6916 	/* Stop incrementating the System Time registers. */
6917 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
6918 
6919 	return 0;
6920 }
6921 
6922 static int
6923 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
6924 				 struct timespec *timestamp,
6925 				 uint32_t flags __rte_unused)
6926 {
6927 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6928 	struct ixgbe_adapter *adapter =
6929 		(struct ixgbe_adapter *)dev->data->dev_private;
6930 	uint32_t tsync_rxctl;
6931 	uint64_t rx_tstamp_cycles;
6932 	uint64_t ns;
6933 
6934 	tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6935 	if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
6936 		return -EINVAL;
6937 
6938 	rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
6939 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
6940 	*timestamp = rte_ns_to_timespec(ns);
6941 
6942 	return  0;
6943 }
6944 
6945 static int
6946 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
6947 				 struct timespec *timestamp)
6948 {
6949 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6950 	struct ixgbe_adapter *adapter =
6951 		(struct ixgbe_adapter *)dev->data->dev_private;
6952 	uint32_t tsync_txctl;
6953 	uint64_t tx_tstamp_cycles;
6954 	uint64_t ns;
6955 
6956 	tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6957 	if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
6958 		return -EINVAL;
6959 
6960 	tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
6961 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
6962 	*timestamp = rte_ns_to_timespec(ns);
6963 
6964 	return 0;
6965 }
6966 
6967 static int
6968 ixgbe_get_reg_length(struct rte_eth_dev *dev)
6969 {
6970 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6971 	int count = 0;
6972 	int g_ind = 0;
6973 	const struct reg_info *reg_group;
6974 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6975 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6976 
6977 	while ((reg_group = reg_set[g_ind++]))
6978 		count += ixgbe_regs_group_count(reg_group);
6979 
6980 	return count;
6981 }
6982 
6983 static int
6984 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
6985 {
6986 	int count = 0;
6987 	int g_ind = 0;
6988 	const struct reg_info *reg_group;
6989 
6990 	while ((reg_group = ixgbevf_regs[g_ind++]))
6991 		count += ixgbe_regs_group_count(reg_group);
6992 
6993 	return count;
6994 }
6995 
6996 static int
6997 ixgbe_get_regs(struct rte_eth_dev *dev,
6998 	      struct rte_dev_reg_info *regs)
6999 {
7000 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7001 	uint32_t *data = regs->data;
7002 	int g_ind = 0;
7003 	int count = 0;
7004 	const struct reg_info *reg_group;
7005 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7006 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7007 
7008 	if (data == NULL) {
7009 		regs->length = ixgbe_get_reg_length(dev);
7010 		regs->width = sizeof(uint32_t);
7011 		return 0;
7012 	}
7013 
7014 	/* Support only full register dump */
7015 	if ((regs->length == 0) ||
7016 	    (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
7017 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7018 			hw->device_id;
7019 		while ((reg_group = reg_set[g_ind++]))
7020 			count += ixgbe_read_regs_group(dev, &data[count],
7021 				reg_group);
7022 		return 0;
7023 	}
7024 
7025 	return -ENOTSUP;
7026 }
7027 
7028 static int
7029 ixgbevf_get_regs(struct rte_eth_dev *dev,
7030 		struct rte_dev_reg_info *regs)
7031 {
7032 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7033 	uint32_t *data = regs->data;
7034 	int g_ind = 0;
7035 	int count = 0;
7036 	const struct reg_info *reg_group;
7037 
7038 	if (data == NULL) {
7039 		regs->length = ixgbevf_get_reg_length(dev);
7040 		regs->width = sizeof(uint32_t);
7041 		return 0;
7042 	}
7043 
7044 	/* Support only full register dump */
7045 	if ((regs->length == 0) ||
7046 	    (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
7047 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7048 			hw->device_id;
7049 		while ((reg_group = ixgbevf_regs[g_ind++]))
7050 			count += ixgbe_read_regs_group(dev, &data[count],
7051 						      reg_group);
7052 		return 0;
7053 	}
7054 
7055 	return -ENOTSUP;
7056 }
7057 
7058 static int
7059 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
7060 {
7061 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7062 
7063 	/* Return unit is byte count */
7064 	return hw->eeprom.word_size * 2;
7065 }
7066 
7067 static int
7068 ixgbe_get_eeprom(struct rte_eth_dev *dev,
7069 		struct rte_dev_eeprom_info *in_eeprom)
7070 {
7071 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7072 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7073 	uint16_t *data = in_eeprom->data;
7074 	int first, length;
7075 
7076 	first = in_eeprom->offset >> 1;
7077 	length = in_eeprom->length >> 1;
7078 	if ((first > hw->eeprom.word_size) ||
7079 	    ((first + length) > hw->eeprom.word_size))
7080 		return -EINVAL;
7081 
7082 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7083 
7084 	return eeprom->ops.read_buffer(hw, first, length, data);
7085 }
7086 
7087 static int
7088 ixgbe_set_eeprom(struct rte_eth_dev *dev,
7089 		struct rte_dev_eeprom_info *in_eeprom)
7090 {
7091 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7092 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7093 	uint16_t *data = in_eeprom->data;
7094 	int first, length;
7095 
7096 	first = in_eeprom->offset >> 1;
7097 	length = in_eeprom->length >> 1;
7098 	if ((first > hw->eeprom.word_size) ||
7099 	    ((first + length) > hw->eeprom.word_size))
7100 		return -EINVAL;
7101 
7102 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7103 
7104 	return eeprom->ops.write_buffer(hw,  first, length, data);
7105 }
7106 
7107 uint16_t
7108 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
7109 	switch (mac_type) {
7110 	case ixgbe_mac_X550:
7111 	case ixgbe_mac_X550EM_x:
7112 	case ixgbe_mac_X550EM_a:
7113 		return ETH_RSS_RETA_SIZE_512;
7114 	case ixgbe_mac_X550_vf:
7115 	case ixgbe_mac_X550EM_x_vf:
7116 	case ixgbe_mac_X550EM_a_vf:
7117 		return ETH_RSS_RETA_SIZE_64;
7118 	default:
7119 		return ETH_RSS_RETA_SIZE_128;
7120 	}
7121 }
7122 
7123 uint32_t
7124 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
7125 	switch (mac_type) {
7126 	case ixgbe_mac_X550:
7127 	case ixgbe_mac_X550EM_x:
7128 	case ixgbe_mac_X550EM_a:
7129 		if (reta_idx < ETH_RSS_RETA_SIZE_128)
7130 			return IXGBE_RETA(reta_idx >> 2);
7131 		else
7132 			return IXGBE_ERETA((reta_idx - ETH_RSS_RETA_SIZE_128) >> 2);
7133 	case ixgbe_mac_X550_vf:
7134 	case ixgbe_mac_X550EM_x_vf:
7135 	case ixgbe_mac_X550EM_a_vf:
7136 		return IXGBE_VFRETA(reta_idx >> 2);
7137 	default:
7138 		return IXGBE_RETA(reta_idx >> 2);
7139 	}
7140 }
7141 
7142 uint32_t
7143 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
7144 	switch (mac_type) {
7145 	case ixgbe_mac_X550_vf:
7146 	case ixgbe_mac_X550EM_x_vf:
7147 	case ixgbe_mac_X550EM_a_vf:
7148 		return IXGBE_VFMRQC;
7149 	default:
7150 		return IXGBE_MRQC;
7151 	}
7152 }
7153 
7154 uint32_t
7155 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
7156 	switch (mac_type) {
7157 	case ixgbe_mac_X550_vf:
7158 	case ixgbe_mac_X550EM_x_vf:
7159 	case ixgbe_mac_X550EM_a_vf:
7160 		return IXGBE_VFRSSRK(i);
7161 	default:
7162 		return IXGBE_RSSRK(i);
7163 	}
7164 }
7165 
7166 bool
7167 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
7168 	switch (mac_type) {
7169 	case ixgbe_mac_82599_vf:
7170 	case ixgbe_mac_X540_vf:
7171 		return 0;
7172 	default:
7173 		return 1;
7174 	}
7175 }
7176 
7177 static int
7178 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
7179 			struct rte_eth_dcb_info *dcb_info)
7180 {
7181 	struct ixgbe_dcb_config *dcb_config =
7182 			IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
7183 	struct ixgbe_dcb_tc_config *tc;
7184 	struct rte_eth_dcb_tc_queue_mapping *tc_queue;
7185 	uint8_t nb_tcs;
7186 	uint8_t i, j;
7187 
7188 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_DCB_FLAG)
7189 		dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
7190 	else
7191 		dcb_info->nb_tcs = 1;
7192 
7193 	tc_queue = &dcb_info->tc_queue;
7194 	nb_tcs = dcb_info->nb_tcs;
7195 
7196 	if (dcb_config->vt_mode) { /* vt is enabled*/
7197 		struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
7198 				&dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
7199 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7200 			dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
7201 		if (RTE_ETH_DEV_SRIOV(dev).active > 0) {
7202 			for (j = 0; j < nb_tcs; j++) {
7203 				tc_queue->tc_rxq[0][j].base = j;
7204 				tc_queue->tc_rxq[0][j].nb_queue = 1;
7205 				tc_queue->tc_txq[0][j].base = j;
7206 				tc_queue->tc_txq[0][j].nb_queue = 1;
7207 			}
7208 		} else {
7209 			for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
7210 				for (j = 0; j < nb_tcs; j++) {
7211 					tc_queue->tc_rxq[i][j].base =
7212 						i * nb_tcs + j;
7213 					tc_queue->tc_rxq[i][j].nb_queue = 1;
7214 					tc_queue->tc_txq[i][j].base =
7215 						i * nb_tcs + j;
7216 					tc_queue->tc_txq[i][j].nb_queue = 1;
7217 				}
7218 			}
7219 		}
7220 	} else { /* vt is disabled*/
7221 		struct rte_eth_dcb_rx_conf *rx_conf =
7222 				&dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
7223 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7224 			dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
7225 		if (dcb_info->nb_tcs == ETH_4_TCS) {
7226 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7227 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
7228 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7229 			}
7230 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7231 			dcb_info->tc_queue.tc_txq[0][1].base = 64;
7232 			dcb_info->tc_queue.tc_txq[0][2].base = 96;
7233 			dcb_info->tc_queue.tc_txq[0][3].base = 112;
7234 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
7235 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7236 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7237 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7238 		} else if (dcb_info->nb_tcs == ETH_8_TCS) {
7239 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7240 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
7241 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7242 			}
7243 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7244 			dcb_info->tc_queue.tc_txq[0][1].base = 32;
7245 			dcb_info->tc_queue.tc_txq[0][2].base = 64;
7246 			dcb_info->tc_queue.tc_txq[0][3].base = 80;
7247 			dcb_info->tc_queue.tc_txq[0][4].base = 96;
7248 			dcb_info->tc_queue.tc_txq[0][5].base = 104;
7249 			dcb_info->tc_queue.tc_txq[0][6].base = 112;
7250 			dcb_info->tc_queue.tc_txq[0][7].base = 120;
7251 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
7252 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7253 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7254 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7255 			dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
7256 			dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
7257 			dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
7258 			dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
7259 		}
7260 	}
7261 	for (i = 0; i < dcb_info->nb_tcs; i++) {
7262 		tc = &dcb_config->tc_config[i];
7263 		dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
7264 	}
7265 	return 0;
7266 }
7267 
7268 /* Update e-tag ether type */
7269 static int
7270 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
7271 			    uint16_t ether_type)
7272 {
7273 	uint32_t etag_etype;
7274 
7275 	if (hw->mac.type != ixgbe_mac_X550 &&
7276 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7277 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7278 		return -ENOTSUP;
7279 	}
7280 
7281 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7282 	etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
7283 	etag_etype |= ether_type;
7284 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7285 	IXGBE_WRITE_FLUSH(hw);
7286 
7287 	return 0;
7288 }
7289 
7290 /* Config l2 tunnel ether type */
7291 static int
7292 ixgbe_dev_l2_tunnel_eth_type_conf(struct rte_eth_dev *dev,
7293 				  struct rte_eth_l2_tunnel_conf *l2_tunnel)
7294 {
7295 	int ret = 0;
7296 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7297 	struct ixgbe_l2_tn_info *l2_tn_info =
7298 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7299 
7300 	if (l2_tunnel == NULL)
7301 		return -EINVAL;
7302 
7303 	switch (l2_tunnel->l2_tunnel_type) {
7304 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7305 		l2_tn_info->e_tag_ether_type = l2_tunnel->ether_type;
7306 		ret = ixgbe_update_e_tag_eth_type(hw, l2_tunnel->ether_type);
7307 		break;
7308 	default:
7309 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7310 		ret = -EINVAL;
7311 		break;
7312 	}
7313 
7314 	return ret;
7315 }
7316 
7317 /* Enable e-tag tunnel */
7318 static int
7319 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
7320 {
7321 	uint32_t etag_etype;
7322 
7323 	if (hw->mac.type != ixgbe_mac_X550 &&
7324 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7325 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7326 		return -ENOTSUP;
7327 	}
7328 
7329 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7330 	etag_etype |= IXGBE_ETAG_ETYPE_VALID;
7331 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7332 	IXGBE_WRITE_FLUSH(hw);
7333 
7334 	return 0;
7335 }
7336 
7337 /* Enable l2 tunnel */
7338 static int
7339 ixgbe_dev_l2_tunnel_enable(struct rte_eth_dev *dev,
7340 			   enum rte_eth_tunnel_type l2_tunnel_type)
7341 {
7342 	int ret = 0;
7343 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7344 	struct ixgbe_l2_tn_info *l2_tn_info =
7345 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7346 
7347 	switch (l2_tunnel_type) {
7348 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7349 		l2_tn_info->e_tag_en = TRUE;
7350 		ret = ixgbe_e_tag_enable(hw);
7351 		break;
7352 	default:
7353 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7354 		ret = -EINVAL;
7355 		break;
7356 	}
7357 
7358 	return ret;
7359 }
7360 
7361 /* Disable e-tag tunnel */
7362 static int
7363 ixgbe_e_tag_disable(struct ixgbe_hw *hw)
7364 {
7365 	uint32_t etag_etype;
7366 
7367 	if (hw->mac.type != ixgbe_mac_X550 &&
7368 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7369 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7370 		return -ENOTSUP;
7371 	}
7372 
7373 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7374 	etag_etype &= ~IXGBE_ETAG_ETYPE_VALID;
7375 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7376 	IXGBE_WRITE_FLUSH(hw);
7377 
7378 	return 0;
7379 }
7380 
7381 /* Disable l2 tunnel */
7382 static int
7383 ixgbe_dev_l2_tunnel_disable(struct rte_eth_dev *dev,
7384 			    enum rte_eth_tunnel_type l2_tunnel_type)
7385 {
7386 	int ret = 0;
7387 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7388 	struct ixgbe_l2_tn_info *l2_tn_info =
7389 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7390 
7391 	switch (l2_tunnel_type) {
7392 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7393 		l2_tn_info->e_tag_en = FALSE;
7394 		ret = ixgbe_e_tag_disable(hw);
7395 		break;
7396 	default:
7397 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7398 		ret = -EINVAL;
7399 		break;
7400 	}
7401 
7402 	return ret;
7403 }
7404 
7405 static int
7406 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
7407 		       struct rte_eth_l2_tunnel_conf *l2_tunnel)
7408 {
7409 	int ret = 0;
7410 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7411 	uint32_t i, rar_entries;
7412 	uint32_t rar_low, rar_high;
7413 
7414 	if (hw->mac.type != ixgbe_mac_X550 &&
7415 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7416 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7417 		return -ENOTSUP;
7418 	}
7419 
7420 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7421 
7422 	for (i = 1; i < rar_entries; i++) {
7423 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7424 		rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
7425 		if ((rar_high & IXGBE_RAH_AV) &&
7426 		    (rar_high & IXGBE_RAH_ADTYPE) &&
7427 		    ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
7428 		     l2_tunnel->tunnel_id)) {
7429 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
7430 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
7431 
7432 			ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
7433 
7434 			return ret;
7435 		}
7436 	}
7437 
7438 	return ret;
7439 }
7440 
7441 static int
7442 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
7443 		       struct rte_eth_l2_tunnel_conf *l2_tunnel)
7444 {
7445 	int ret = 0;
7446 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7447 	uint32_t i, rar_entries;
7448 	uint32_t rar_low, rar_high;
7449 
7450 	if (hw->mac.type != ixgbe_mac_X550 &&
7451 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7452 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7453 		return -ENOTSUP;
7454 	}
7455 
7456 	/* One entry for one tunnel. Try to remove potential existing entry. */
7457 	ixgbe_e_tag_filter_del(dev, l2_tunnel);
7458 
7459 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7460 
7461 	for (i = 1; i < rar_entries; i++) {
7462 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7463 		if (rar_high & IXGBE_RAH_AV) {
7464 			continue;
7465 		} else {
7466 			ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
7467 			rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
7468 			rar_low = l2_tunnel->tunnel_id;
7469 
7470 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
7471 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
7472 
7473 			return ret;
7474 		}
7475 	}
7476 
7477 	PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
7478 		     " Please remove a rule before adding a new one.");
7479 	return -EINVAL;
7480 }
7481 
7482 static inline struct ixgbe_l2_tn_filter *
7483 ixgbe_l2_tn_filter_lookup(struct ixgbe_l2_tn_info *l2_tn_info,
7484 			  struct ixgbe_l2_tn_key *key)
7485 {
7486 	int ret;
7487 
7488 	ret = rte_hash_lookup(l2_tn_info->hash_handle, (const void *)key);
7489 	if (ret < 0)
7490 		return NULL;
7491 
7492 	return l2_tn_info->hash_map[ret];
7493 }
7494 
7495 static inline int
7496 ixgbe_insert_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7497 			  struct ixgbe_l2_tn_filter *l2_tn_filter)
7498 {
7499 	int ret;
7500 
7501 	ret = rte_hash_add_key(l2_tn_info->hash_handle,
7502 			       &l2_tn_filter->key);
7503 
7504 	if (ret < 0) {
7505 		PMD_DRV_LOG(ERR,
7506 			    "Failed to insert L2 tunnel filter"
7507 			    " to hash table %d!",
7508 			    ret);
7509 		return ret;
7510 	}
7511 
7512 	l2_tn_info->hash_map[ret] = l2_tn_filter;
7513 
7514 	TAILQ_INSERT_TAIL(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7515 
7516 	return 0;
7517 }
7518 
7519 static inline int
7520 ixgbe_remove_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7521 			  struct ixgbe_l2_tn_key *key)
7522 {
7523 	int ret;
7524 	struct ixgbe_l2_tn_filter *l2_tn_filter;
7525 
7526 	ret = rte_hash_del_key(l2_tn_info->hash_handle, key);
7527 
7528 	if (ret < 0) {
7529 		PMD_DRV_LOG(ERR,
7530 			    "No such L2 tunnel filter to delete %d!",
7531 			    ret);
7532 		return ret;
7533 	}
7534 
7535 	l2_tn_filter = l2_tn_info->hash_map[ret];
7536 	l2_tn_info->hash_map[ret] = NULL;
7537 
7538 	TAILQ_REMOVE(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7539 	rte_free(l2_tn_filter);
7540 
7541 	return 0;
7542 }
7543 
7544 /* Add l2 tunnel filter */
7545 int
7546 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
7547 			       struct rte_eth_l2_tunnel_conf *l2_tunnel,
7548 			       bool restore)
7549 {
7550 	int ret;
7551 	struct ixgbe_l2_tn_info *l2_tn_info =
7552 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7553 	struct ixgbe_l2_tn_key key;
7554 	struct ixgbe_l2_tn_filter *node;
7555 
7556 	if (!restore) {
7557 		key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7558 		key.tn_id = l2_tunnel->tunnel_id;
7559 
7560 		node = ixgbe_l2_tn_filter_lookup(l2_tn_info, &key);
7561 
7562 		if (node) {
7563 			PMD_DRV_LOG(ERR,
7564 				    "The L2 tunnel filter already exists!");
7565 			return -EINVAL;
7566 		}
7567 
7568 		node = rte_zmalloc("ixgbe_l2_tn",
7569 				   sizeof(struct ixgbe_l2_tn_filter),
7570 				   0);
7571 		if (!node)
7572 			return -ENOMEM;
7573 
7574 		rte_memcpy(&node->key,
7575 				 &key,
7576 				 sizeof(struct ixgbe_l2_tn_key));
7577 		node->pool = l2_tunnel->pool;
7578 		ret = ixgbe_insert_l2_tn_filter(l2_tn_info, node);
7579 		if (ret < 0) {
7580 			rte_free(node);
7581 			return ret;
7582 		}
7583 	}
7584 
7585 	switch (l2_tunnel->l2_tunnel_type) {
7586 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7587 		ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
7588 		break;
7589 	default:
7590 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7591 		ret = -EINVAL;
7592 		break;
7593 	}
7594 
7595 	if ((!restore) && (ret < 0))
7596 		(void)ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7597 
7598 	return ret;
7599 }
7600 
7601 /* Delete l2 tunnel filter */
7602 int
7603 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
7604 			       struct rte_eth_l2_tunnel_conf *l2_tunnel)
7605 {
7606 	int ret;
7607 	struct ixgbe_l2_tn_info *l2_tn_info =
7608 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7609 	struct ixgbe_l2_tn_key key;
7610 
7611 	key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7612 	key.tn_id = l2_tunnel->tunnel_id;
7613 	ret = ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7614 	if (ret < 0)
7615 		return ret;
7616 
7617 	switch (l2_tunnel->l2_tunnel_type) {
7618 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7619 		ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
7620 		break;
7621 	default:
7622 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7623 		ret = -EINVAL;
7624 		break;
7625 	}
7626 
7627 	return ret;
7628 }
7629 
7630 /**
7631  * ixgbe_dev_l2_tunnel_filter_handle - Handle operations for l2 tunnel filter.
7632  * @dev: pointer to rte_eth_dev structure
7633  * @filter_op:operation will be taken.
7634  * @arg: a pointer to specific structure corresponding to the filter_op
7635  */
7636 static int
7637 ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
7638 				  enum rte_filter_op filter_op,
7639 				  void *arg)
7640 {
7641 	int ret;
7642 
7643 	if (filter_op == RTE_ETH_FILTER_NOP)
7644 		return 0;
7645 
7646 	if (arg == NULL) {
7647 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
7648 			    filter_op);
7649 		return -EINVAL;
7650 	}
7651 
7652 	switch (filter_op) {
7653 	case RTE_ETH_FILTER_ADD:
7654 		ret = ixgbe_dev_l2_tunnel_filter_add
7655 			(dev,
7656 			 (struct rte_eth_l2_tunnel_conf *)arg,
7657 			 FALSE);
7658 		break;
7659 	case RTE_ETH_FILTER_DELETE:
7660 		ret = ixgbe_dev_l2_tunnel_filter_del
7661 			(dev,
7662 			 (struct rte_eth_l2_tunnel_conf *)arg);
7663 		break;
7664 	default:
7665 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
7666 		ret = -EINVAL;
7667 		break;
7668 	}
7669 	return ret;
7670 }
7671 
7672 static int
7673 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
7674 {
7675 	int ret = 0;
7676 	uint32_t ctrl;
7677 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7678 
7679 	if (hw->mac.type != ixgbe_mac_X550 &&
7680 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7681 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7682 		return -ENOTSUP;
7683 	}
7684 
7685 	ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
7686 	ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
7687 	if (en)
7688 		ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
7689 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
7690 
7691 	return ret;
7692 }
7693 
7694 /* Enable l2 tunnel forwarding */
7695 static int
7696 ixgbe_dev_l2_tunnel_forwarding_enable
7697 	(struct rte_eth_dev *dev,
7698 	 enum rte_eth_tunnel_type l2_tunnel_type)
7699 {
7700 	struct ixgbe_l2_tn_info *l2_tn_info =
7701 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7702 	int ret = 0;
7703 
7704 	switch (l2_tunnel_type) {
7705 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7706 		l2_tn_info->e_tag_fwd_en = TRUE;
7707 		ret = ixgbe_e_tag_forwarding_en_dis(dev, 1);
7708 		break;
7709 	default:
7710 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7711 		ret = -EINVAL;
7712 		break;
7713 	}
7714 
7715 	return ret;
7716 }
7717 
7718 /* Disable l2 tunnel forwarding */
7719 static int
7720 ixgbe_dev_l2_tunnel_forwarding_disable
7721 	(struct rte_eth_dev *dev,
7722 	 enum rte_eth_tunnel_type l2_tunnel_type)
7723 {
7724 	struct ixgbe_l2_tn_info *l2_tn_info =
7725 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7726 	int ret = 0;
7727 
7728 	switch (l2_tunnel_type) {
7729 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7730 		l2_tn_info->e_tag_fwd_en = FALSE;
7731 		ret = ixgbe_e_tag_forwarding_en_dis(dev, 0);
7732 		break;
7733 	default:
7734 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7735 		ret = -EINVAL;
7736 		break;
7737 	}
7738 
7739 	return ret;
7740 }
7741 
7742 static int
7743 ixgbe_e_tag_insertion_en_dis(struct rte_eth_dev *dev,
7744 			     struct rte_eth_l2_tunnel_conf *l2_tunnel,
7745 			     bool en)
7746 {
7747 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
7748 	int ret = 0;
7749 	uint32_t vmtir, vmvir;
7750 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7751 
7752 	if (l2_tunnel->vf_id >= pci_dev->max_vfs) {
7753 		PMD_DRV_LOG(ERR,
7754 			    "VF id %u should be less than %u",
7755 			    l2_tunnel->vf_id,
7756 			    pci_dev->max_vfs);
7757 		return -EINVAL;
7758 	}
7759 
7760 	if (hw->mac.type != ixgbe_mac_X550 &&
7761 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7762 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7763 		return -ENOTSUP;
7764 	}
7765 
7766 	if (en)
7767 		vmtir = l2_tunnel->tunnel_id;
7768 	else
7769 		vmtir = 0;
7770 
7771 	IXGBE_WRITE_REG(hw, IXGBE_VMTIR(l2_tunnel->vf_id), vmtir);
7772 
7773 	vmvir = IXGBE_READ_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id));
7774 	vmvir &= ~IXGBE_VMVIR_TAGA_MASK;
7775 	if (en)
7776 		vmvir |= IXGBE_VMVIR_TAGA_ETAG_INSERT;
7777 	IXGBE_WRITE_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id), vmvir);
7778 
7779 	return ret;
7780 }
7781 
7782 /* Enable l2 tunnel tag insertion */
7783 static int
7784 ixgbe_dev_l2_tunnel_insertion_enable(struct rte_eth_dev *dev,
7785 				     struct rte_eth_l2_tunnel_conf *l2_tunnel)
7786 {
7787 	int ret = 0;
7788 
7789 	switch (l2_tunnel->l2_tunnel_type) {
7790 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7791 		ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 1);
7792 		break;
7793 	default:
7794 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7795 		ret = -EINVAL;
7796 		break;
7797 	}
7798 
7799 	return ret;
7800 }
7801 
7802 /* Disable l2 tunnel tag insertion */
7803 static int
7804 ixgbe_dev_l2_tunnel_insertion_disable
7805 	(struct rte_eth_dev *dev,
7806 	 struct rte_eth_l2_tunnel_conf *l2_tunnel)
7807 {
7808 	int ret = 0;
7809 
7810 	switch (l2_tunnel->l2_tunnel_type) {
7811 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7812 		ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 0);
7813 		break;
7814 	default:
7815 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7816 		ret = -EINVAL;
7817 		break;
7818 	}
7819 
7820 	return ret;
7821 }
7822 
7823 static int
7824 ixgbe_e_tag_stripping_en_dis(struct rte_eth_dev *dev,
7825 			     bool en)
7826 {
7827 	int ret = 0;
7828 	uint32_t qde;
7829 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7830 
7831 	if (hw->mac.type != ixgbe_mac_X550 &&
7832 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7833 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7834 		return -ENOTSUP;
7835 	}
7836 
7837 	qde = IXGBE_READ_REG(hw, IXGBE_QDE);
7838 	if (en)
7839 		qde |= IXGBE_QDE_STRIP_TAG;
7840 	else
7841 		qde &= ~IXGBE_QDE_STRIP_TAG;
7842 	qde &= ~IXGBE_QDE_READ;
7843 	qde |= IXGBE_QDE_WRITE;
7844 	IXGBE_WRITE_REG(hw, IXGBE_QDE, qde);
7845 
7846 	return ret;
7847 }
7848 
7849 /* Enable l2 tunnel tag stripping */
7850 static int
7851 ixgbe_dev_l2_tunnel_stripping_enable
7852 	(struct rte_eth_dev *dev,
7853 	 enum rte_eth_tunnel_type l2_tunnel_type)
7854 {
7855 	int ret = 0;
7856 
7857 	switch (l2_tunnel_type) {
7858 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7859 		ret = ixgbe_e_tag_stripping_en_dis(dev, 1);
7860 		break;
7861 	default:
7862 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7863 		ret = -EINVAL;
7864 		break;
7865 	}
7866 
7867 	return ret;
7868 }
7869 
7870 /* Disable l2 tunnel tag stripping */
7871 static int
7872 ixgbe_dev_l2_tunnel_stripping_disable
7873 	(struct rte_eth_dev *dev,
7874 	 enum rte_eth_tunnel_type l2_tunnel_type)
7875 {
7876 	int ret = 0;
7877 
7878 	switch (l2_tunnel_type) {
7879 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7880 		ret = ixgbe_e_tag_stripping_en_dis(dev, 0);
7881 		break;
7882 	default:
7883 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7884 		ret = -EINVAL;
7885 		break;
7886 	}
7887 
7888 	return ret;
7889 }
7890 
7891 /* Enable/disable l2 tunnel offload functions */
7892 static int
7893 ixgbe_dev_l2_tunnel_offload_set
7894 	(struct rte_eth_dev *dev,
7895 	 struct rte_eth_l2_tunnel_conf *l2_tunnel,
7896 	 uint32_t mask,
7897 	 uint8_t en)
7898 {
7899 	int ret = 0;
7900 
7901 	if (l2_tunnel == NULL)
7902 		return -EINVAL;
7903 
7904 	ret = -EINVAL;
7905 	if (mask & ETH_L2_TUNNEL_ENABLE_MASK) {
7906 		if (en)
7907 			ret = ixgbe_dev_l2_tunnel_enable(
7908 				dev,
7909 				l2_tunnel->l2_tunnel_type);
7910 		else
7911 			ret = ixgbe_dev_l2_tunnel_disable(
7912 				dev,
7913 				l2_tunnel->l2_tunnel_type);
7914 	}
7915 
7916 	if (mask & ETH_L2_TUNNEL_INSERTION_MASK) {
7917 		if (en)
7918 			ret = ixgbe_dev_l2_tunnel_insertion_enable(
7919 				dev,
7920 				l2_tunnel);
7921 		else
7922 			ret = ixgbe_dev_l2_tunnel_insertion_disable(
7923 				dev,
7924 				l2_tunnel);
7925 	}
7926 
7927 	if (mask & ETH_L2_TUNNEL_STRIPPING_MASK) {
7928 		if (en)
7929 			ret = ixgbe_dev_l2_tunnel_stripping_enable(
7930 				dev,
7931 				l2_tunnel->l2_tunnel_type);
7932 		else
7933 			ret = ixgbe_dev_l2_tunnel_stripping_disable(
7934 				dev,
7935 				l2_tunnel->l2_tunnel_type);
7936 	}
7937 
7938 	if (mask & ETH_L2_TUNNEL_FORWARDING_MASK) {
7939 		if (en)
7940 			ret = ixgbe_dev_l2_tunnel_forwarding_enable(
7941 				dev,
7942 				l2_tunnel->l2_tunnel_type);
7943 		else
7944 			ret = ixgbe_dev_l2_tunnel_forwarding_disable(
7945 				dev,
7946 				l2_tunnel->l2_tunnel_type);
7947 	}
7948 
7949 	return ret;
7950 }
7951 
7952 static int
7953 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
7954 			uint16_t port)
7955 {
7956 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
7957 	IXGBE_WRITE_FLUSH(hw);
7958 
7959 	return 0;
7960 }
7961 
7962 /* There's only one register for VxLAN UDP port.
7963  * So, we cannot add several ports. Will update it.
7964  */
7965 static int
7966 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
7967 		     uint16_t port)
7968 {
7969 	if (port == 0) {
7970 		PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
7971 		return -EINVAL;
7972 	}
7973 
7974 	return ixgbe_update_vxlan_port(hw, port);
7975 }
7976 
7977 /* We cannot delete the VxLAN port. For there's a register for VxLAN
7978  * UDP port, it must have a value.
7979  * So, will reset it to the original value 0.
7980  */
7981 static int
7982 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
7983 		     uint16_t port)
7984 {
7985 	uint16_t cur_port;
7986 
7987 	cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
7988 
7989 	if (cur_port != port) {
7990 		PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
7991 		return -EINVAL;
7992 	}
7993 
7994 	return ixgbe_update_vxlan_port(hw, 0);
7995 }
7996 
7997 /* Add UDP tunneling port */
7998 static int
7999 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
8000 			      struct rte_eth_udp_tunnel *udp_tunnel)
8001 {
8002 	int ret = 0;
8003 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8004 
8005 	if (hw->mac.type != ixgbe_mac_X550 &&
8006 	    hw->mac.type != ixgbe_mac_X550EM_x &&
8007 	    hw->mac.type != ixgbe_mac_X550EM_a) {
8008 		return -ENOTSUP;
8009 	}
8010 
8011 	if (udp_tunnel == NULL)
8012 		return -EINVAL;
8013 
8014 	switch (udp_tunnel->prot_type) {
8015 	case RTE_TUNNEL_TYPE_VXLAN:
8016 		ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
8017 		break;
8018 
8019 	case RTE_TUNNEL_TYPE_GENEVE:
8020 	case RTE_TUNNEL_TYPE_TEREDO:
8021 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
8022 		ret = -EINVAL;
8023 		break;
8024 
8025 	default:
8026 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
8027 		ret = -EINVAL;
8028 		break;
8029 	}
8030 
8031 	return ret;
8032 }
8033 
8034 /* Remove UDP tunneling port */
8035 static int
8036 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
8037 			      struct rte_eth_udp_tunnel *udp_tunnel)
8038 {
8039 	int ret = 0;
8040 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8041 
8042 	if (hw->mac.type != ixgbe_mac_X550 &&
8043 	    hw->mac.type != ixgbe_mac_X550EM_x &&
8044 	    hw->mac.type != ixgbe_mac_X550EM_a) {
8045 		return -ENOTSUP;
8046 	}
8047 
8048 	if (udp_tunnel == NULL)
8049 		return -EINVAL;
8050 
8051 	switch (udp_tunnel->prot_type) {
8052 	case RTE_TUNNEL_TYPE_VXLAN:
8053 		ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
8054 		break;
8055 	case RTE_TUNNEL_TYPE_GENEVE:
8056 	case RTE_TUNNEL_TYPE_TEREDO:
8057 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
8058 		ret = -EINVAL;
8059 		break;
8060 	default:
8061 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
8062 		ret = -EINVAL;
8063 		break;
8064 	}
8065 
8066 	return ret;
8067 }
8068 
8069 static void
8070 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
8071 {
8072 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8073 
8074 	hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_ALLMULTI);
8075 }
8076 
8077 static void
8078 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
8079 {
8080 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8081 
8082 	hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_MULTI);
8083 }
8084 
8085 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
8086 {
8087 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8088 	u32 in_msg = 0;
8089 
8090 	/* peek the message first */
8091 	in_msg = IXGBE_READ_REG(hw, IXGBE_VFMBMEM);
8092 
8093 	/* PF reset VF event */
8094 	if (in_msg == IXGBE_PF_CONTROL_MSG) {
8095 		/* dummy mbx read to ack pf */
8096 		if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
8097 			return;
8098 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
8099 					      NULL);
8100 	}
8101 }
8102 
8103 static int
8104 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
8105 {
8106 	uint32_t eicr;
8107 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8108 	struct ixgbe_interrupt *intr =
8109 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8110 	ixgbevf_intr_disable(hw);
8111 
8112 	/* read-on-clear nic registers here */
8113 	eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
8114 	intr->flags = 0;
8115 
8116 	/* only one misc vector supported - mailbox */
8117 	eicr &= IXGBE_VTEICR_MASK;
8118 	if (eicr == IXGBE_MISC_VEC_ID)
8119 		intr->flags |= IXGBE_FLAG_MAILBOX;
8120 
8121 	return 0;
8122 }
8123 
8124 static int
8125 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
8126 {
8127 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8128 	struct ixgbe_interrupt *intr =
8129 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8130 
8131 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
8132 		ixgbevf_mbx_process(dev);
8133 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
8134 	}
8135 
8136 	ixgbevf_intr_enable(hw);
8137 
8138 	return 0;
8139 }
8140 
8141 static void
8142 ixgbevf_dev_interrupt_handler(void *param)
8143 {
8144 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
8145 
8146 	ixgbevf_dev_interrupt_get_status(dev);
8147 	ixgbevf_dev_interrupt_action(dev);
8148 }
8149 
8150 /**
8151  *  ixgbe_disable_sec_tx_path_generic - Stops the transmit data path
8152  *  @hw: pointer to hardware structure
8153  *
8154  *  Stops the transmit data path and waits for the HW to internally empty
8155  *  the Tx security block
8156  **/
8157 int ixgbe_disable_sec_tx_path_generic(struct ixgbe_hw *hw)
8158 {
8159 #define IXGBE_MAX_SECTX_POLL 40
8160 
8161 	int i;
8162 	int sectxreg;
8163 
8164 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8165 	sectxreg |= IXGBE_SECTXCTRL_TX_DIS;
8166 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8167 	for (i = 0; i < IXGBE_MAX_SECTX_POLL; i++) {
8168 		sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXSTAT);
8169 		if (sectxreg & IXGBE_SECTXSTAT_SECTX_RDY)
8170 			break;
8171 		/* Use interrupt-safe sleep just in case */
8172 		usec_delay(1000);
8173 	}
8174 
8175 	/* For informational purposes only */
8176 	if (i >= IXGBE_MAX_SECTX_POLL)
8177 		PMD_DRV_LOG(DEBUG, "Tx unit being enabled before security "
8178 			 "path fully disabled.  Continuing with init.");
8179 
8180 	return IXGBE_SUCCESS;
8181 }
8182 
8183 /**
8184  *  ixgbe_enable_sec_tx_path_generic - Enables the transmit data path
8185  *  @hw: pointer to hardware structure
8186  *
8187  *  Enables the transmit data path.
8188  **/
8189 int ixgbe_enable_sec_tx_path_generic(struct ixgbe_hw *hw)
8190 {
8191 	uint32_t sectxreg;
8192 
8193 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8194 	sectxreg &= ~IXGBE_SECTXCTRL_TX_DIS;
8195 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8196 	IXGBE_WRITE_FLUSH(hw);
8197 
8198 	return IXGBE_SUCCESS;
8199 }
8200 
8201 /* restore n-tuple filter */
8202 static inline void
8203 ixgbe_ntuple_filter_restore(struct rte_eth_dev *dev)
8204 {
8205 	struct ixgbe_filter_info *filter_info =
8206 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8207 	struct ixgbe_5tuple_filter *node;
8208 
8209 	TAILQ_FOREACH(node, &filter_info->fivetuple_list, entries) {
8210 		ixgbe_inject_5tuple_filter(dev, node);
8211 	}
8212 }
8213 
8214 /* restore ethernet type filter */
8215 static inline void
8216 ixgbe_ethertype_filter_restore(struct rte_eth_dev *dev)
8217 {
8218 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8219 	struct ixgbe_filter_info *filter_info =
8220 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8221 	int i;
8222 
8223 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8224 		if (filter_info->ethertype_mask & (1 << i)) {
8225 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i),
8226 					filter_info->ethertype_filters[i].etqf);
8227 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i),
8228 					filter_info->ethertype_filters[i].etqs);
8229 			IXGBE_WRITE_FLUSH(hw);
8230 		}
8231 	}
8232 }
8233 
8234 /* restore SYN filter */
8235 static inline void
8236 ixgbe_syn_filter_restore(struct rte_eth_dev *dev)
8237 {
8238 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8239 	struct ixgbe_filter_info *filter_info =
8240 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8241 	uint32_t synqf;
8242 
8243 	synqf = filter_info->syn_info;
8244 
8245 	if (synqf & IXGBE_SYN_FILTER_ENABLE) {
8246 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
8247 		IXGBE_WRITE_FLUSH(hw);
8248 	}
8249 }
8250 
8251 /* restore L2 tunnel filter */
8252 static inline void
8253 ixgbe_l2_tn_filter_restore(struct rte_eth_dev *dev)
8254 {
8255 	struct ixgbe_l2_tn_info *l2_tn_info =
8256 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8257 	struct ixgbe_l2_tn_filter *node;
8258 	struct rte_eth_l2_tunnel_conf l2_tn_conf;
8259 
8260 	TAILQ_FOREACH(node, &l2_tn_info->l2_tn_list, entries) {
8261 		l2_tn_conf.l2_tunnel_type = node->key.l2_tn_type;
8262 		l2_tn_conf.tunnel_id      = node->key.tn_id;
8263 		l2_tn_conf.pool           = node->pool;
8264 		(void)ixgbe_dev_l2_tunnel_filter_add(dev, &l2_tn_conf, TRUE);
8265 	}
8266 }
8267 
8268 /* restore rss filter */
8269 static inline void
8270 ixgbe_rss_filter_restore(struct rte_eth_dev *dev)
8271 {
8272 	struct ixgbe_filter_info *filter_info =
8273 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8274 
8275 	if (filter_info->rss_info.num)
8276 		ixgbe_config_rss_filter(dev,
8277 			&filter_info->rss_info, TRUE);
8278 }
8279 
8280 static int
8281 ixgbe_filter_restore(struct rte_eth_dev *dev)
8282 {
8283 	ixgbe_ntuple_filter_restore(dev);
8284 	ixgbe_ethertype_filter_restore(dev);
8285 	ixgbe_syn_filter_restore(dev);
8286 	ixgbe_fdir_filter_restore(dev);
8287 	ixgbe_l2_tn_filter_restore(dev);
8288 	ixgbe_rss_filter_restore(dev);
8289 
8290 	return 0;
8291 }
8292 
8293 static void
8294 ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev)
8295 {
8296 	struct ixgbe_l2_tn_info *l2_tn_info =
8297 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8298 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8299 
8300 	if (l2_tn_info->e_tag_en)
8301 		(void)ixgbe_e_tag_enable(hw);
8302 
8303 	if (l2_tn_info->e_tag_fwd_en)
8304 		(void)ixgbe_e_tag_forwarding_en_dis(dev, 1);
8305 
8306 	(void)ixgbe_update_e_tag_eth_type(hw, l2_tn_info->e_tag_ether_type);
8307 }
8308 
8309 /* remove all the n-tuple filters */
8310 void
8311 ixgbe_clear_all_ntuple_filter(struct rte_eth_dev *dev)
8312 {
8313 	struct ixgbe_filter_info *filter_info =
8314 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8315 	struct ixgbe_5tuple_filter *p_5tuple;
8316 
8317 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
8318 		ixgbe_remove_5tuple_filter(dev, p_5tuple);
8319 }
8320 
8321 /* remove all the ether type filters */
8322 void
8323 ixgbe_clear_all_ethertype_filter(struct rte_eth_dev *dev)
8324 {
8325 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8326 	struct ixgbe_filter_info *filter_info =
8327 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8328 	int i;
8329 
8330 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8331 		if (filter_info->ethertype_mask & (1 << i) &&
8332 		    !filter_info->ethertype_filters[i].conf) {
8333 			(void)ixgbe_ethertype_filter_remove(filter_info,
8334 							    (uint8_t)i);
8335 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i), 0);
8336 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i), 0);
8337 			IXGBE_WRITE_FLUSH(hw);
8338 		}
8339 	}
8340 }
8341 
8342 /* remove the SYN filter */
8343 void
8344 ixgbe_clear_syn_filter(struct rte_eth_dev *dev)
8345 {
8346 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8347 	struct ixgbe_filter_info *filter_info =
8348 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8349 
8350 	if (filter_info->syn_info & IXGBE_SYN_FILTER_ENABLE) {
8351 		filter_info->syn_info = 0;
8352 
8353 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, 0);
8354 		IXGBE_WRITE_FLUSH(hw);
8355 	}
8356 }
8357 
8358 /* remove all the L2 tunnel filters */
8359 int
8360 ixgbe_clear_all_l2_tn_filter(struct rte_eth_dev *dev)
8361 {
8362 	struct ixgbe_l2_tn_info *l2_tn_info =
8363 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8364 	struct ixgbe_l2_tn_filter *l2_tn_filter;
8365 	struct rte_eth_l2_tunnel_conf l2_tn_conf;
8366 	int ret = 0;
8367 
8368 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
8369 		l2_tn_conf.l2_tunnel_type = l2_tn_filter->key.l2_tn_type;
8370 		l2_tn_conf.tunnel_id      = l2_tn_filter->key.tn_id;
8371 		l2_tn_conf.pool           = l2_tn_filter->pool;
8372 		ret = ixgbe_dev_l2_tunnel_filter_del(dev, &l2_tn_conf);
8373 		if (ret < 0)
8374 			return ret;
8375 	}
8376 
8377 	return 0;
8378 }
8379 
8380 RTE_PMD_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd);
8381 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
8382 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe, "* igb_uio | uio_pci_generic | vfio-pci");
8383 RTE_PMD_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd);
8384 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);
8385 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe_vf, "* igb_uio | vfio-pci");
8386 
8387 RTE_INIT(ixgbe_init_log);
8388 static void
8389 ixgbe_init_log(void)
8390 {
8391 	ixgbe_logtype_init = rte_log_register("pmd.net.ixgbe.init");
8392 	if (ixgbe_logtype_init >= 0)
8393 		rte_log_set_level(ixgbe_logtype_init, RTE_LOG_NOTICE);
8394 	ixgbe_logtype_driver = rte_log_register("pmd.net.ixgbe.driver");
8395 	if (ixgbe_logtype_driver >= 0)
8396 		rte_log_set_level(ixgbe_logtype_driver, RTE_LOG_NOTICE);
8397 }
8398