xref: /dpdk/drivers/net/ixgbe/ixgbe_ethdev.c (revision ba9de463)
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_atomic.h>
24 #include <rte_branch_prediction.h>
25 #include <rte_memory.h>
26 #include <rte_eal.h>
27 #include <rte_alarm.h>
28 #include <rte_ether.h>
29 #include <rte_ethdev.h>
30 #include <rte_ethdev_pci.h>
31 #include <rte_malloc.h>
32 #include <rte_random.h>
33 #include <rte_dev.h>
34 #include <rte_hash_crc.h>
35 #ifdef RTE_LIBRTE_SECURITY
36 #include <rte_security_driver.h>
37 #endif
38 
39 #include "ixgbe_logs.h"
40 #include "base/ixgbe_api.h"
41 #include "base/ixgbe_vf.h"
42 #include "base/ixgbe_common.h"
43 #include "ixgbe_ethdev.h"
44 #include "ixgbe_bypass.h"
45 #include "ixgbe_rxtx.h"
46 #include "base/ixgbe_type.h"
47 #include "base/ixgbe_phy.h"
48 #include "ixgbe_regs.h"
49 
50 /*
51  * High threshold controlling when to start sending XOFF frames. Must be at
52  * least 8 bytes less than receive packet buffer size. This value is in units
53  * of 1024 bytes.
54  */
55 #define IXGBE_FC_HI    0x80
56 
57 /*
58  * Low threshold controlling when to start sending XON frames. This value is
59  * in units of 1024 bytes.
60  */
61 #define IXGBE_FC_LO    0x40
62 
63 /* Default minimum inter-interrupt interval for EITR configuration */
64 #define IXGBE_MIN_INTER_INTERRUPT_INTERVAL_DEFAULT    0x79E
65 
66 /* Timer value included in XOFF frames. */
67 #define IXGBE_FC_PAUSE 0x680
68 
69 /*Default value of Max Rx Queue*/
70 #define IXGBE_MAX_RX_QUEUE_NUM 128
71 
72 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
73 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
74 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
75 
76 #define IXGBE_MMW_SIZE_DEFAULT        0x4
77 #define IXGBE_MMW_SIZE_JUMBO_FRAME    0x14
78 #define IXGBE_MAX_RING_DESC           4096 /* replicate define from rxtx */
79 
80 /*
81  *  Default values for RX/TX configuration
82  */
83 #define IXGBE_DEFAULT_RX_FREE_THRESH  32
84 #define IXGBE_DEFAULT_RX_PTHRESH      8
85 #define IXGBE_DEFAULT_RX_HTHRESH      8
86 #define IXGBE_DEFAULT_RX_WTHRESH      0
87 
88 #define IXGBE_DEFAULT_TX_FREE_THRESH  32
89 #define IXGBE_DEFAULT_TX_PTHRESH      32
90 #define IXGBE_DEFAULT_TX_HTHRESH      0
91 #define IXGBE_DEFAULT_TX_WTHRESH      0
92 #define IXGBE_DEFAULT_TX_RSBIT_THRESH 32
93 
94 /* Bit shift and mask */
95 #define IXGBE_4_BIT_WIDTH  (CHAR_BIT / 2)
96 #define IXGBE_4_BIT_MASK   RTE_LEN2MASK(IXGBE_4_BIT_WIDTH, uint8_t)
97 #define IXGBE_8_BIT_WIDTH  CHAR_BIT
98 #define IXGBE_8_BIT_MASK   UINT8_MAX
99 
100 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
101 
102 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
103 
104 #define IXGBE_HKEY_MAX_INDEX 10
105 
106 /* Additional timesync values. */
107 #define NSEC_PER_SEC             1000000000L
108 #define IXGBE_INCVAL_10GB        0x66666666
109 #define IXGBE_INCVAL_1GB         0x40000000
110 #define IXGBE_INCVAL_100         0x50000000
111 #define IXGBE_INCVAL_SHIFT_10GB  28
112 #define IXGBE_INCVAL_SHIFT_1GB   24
113 #define IXGBE_INCVAL_SHIFT_100   21
114 #define IXGBE_INCVAL_SHIFT_82599 7
115 #define IXGBE_INCPER_SHIFT_82599 24
116 
117 #define IXGBE_CYCLECOUNTER_MASK   0xffffffffffffffffULL
118 
119 #define IXGBE_VT_CTL_POOLING_MODE_MASK         0x00030000
120 #define IXGBE_VT_CTL_POOLING_MODE_ETAG         0x00010000
121 #define DEFAULT_ETAG_ETYPE                     0x893f
122 #define IXGBE_ETAG_ETYPE                       0x00005084
123 #define IXGBE_ETAG_ETYPE_MASK                  0x0000ffff
124 #define IXGBE_ETAG_ETYPE_VALID                 0x80000000
125 #define IXGBE_RAH_ADTYPE                       0x40000000
126 #define IXGBE_RAL_ETAG_FILTER_MASK             0x00003fff
127 #define IXGBE_VMVIR_TAGA_MASK                  0x18000000
128 #define IXGBE_VMVIR_TAGA_ETAG_INSERT           0x08000000
129 #define IXGBE_VMTIR(_i) (0x00017000 + ((_i) * 4)) /* 64 of these (0-63) */
130 #define IXGBE_QDE_STRIP_TAG                    0x00000004
131 #define IXGBE_VTEICR_MASK                      0x07
132 
133 #define IXGBE_EXVET_VET_EXT_SHIFT              16
134 #define IXGBE_DMATXCTL_VT_MASK                 0xFFFF0000
135 
136 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev);
137 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
138 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev);
139 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev);
140 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev);
141 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev);
142 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev);
143 static int  ixgbe_dev_configure(struct rte_eth_dev *dev);
144 static int  ixgbe_dev_start(struct rte_eth_dev *dev);
145 static void ixgbe_dev_stop(struct rte_eth_dev *dev);
146 static int  ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
147 static int  ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
148 static void ixgbe_dev_close(struct rte_eth_dev *dev);
149 static int  ixgbe_dev_reset(struct rte_eth_dev *dev);
150 static void ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
151 static void ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
152 static void ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
153 static void ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
154 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
155 				int wait_to_complete);
156 static int ixgbe_dev_stats_get(struct rte_eth_dev *dev,
157 				struct rte_eth_stats *stats);
158 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
159 				struct rte_eth_xstat *xstats, unsigned n);
160 static int ixgbevf_dev_xstats_get(struct rte_eth_dev *dev,
161 				  struct rte_eth_xstat *xstats, unsigned n);
162 static int
163 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
164 		uint64_t *values, unsigned int n);
165 static void ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
166 static void ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
167 static int ixgbe_dev_xstats_get_names(struct rte_eth_dev *dev,
168 	struct rte_eth_xstat_name *xstats_names,
169 	unsigned int size);
170 static int ixgbevf_dev_xstats_get_names(struct rte_eth_dev *dev,
171 	struct rte_eth_xstat_name *xstats_names, unsigned limit);
172 static int ixgbe_dev_xstats_get_names_by_id(
173 	struct rte_eth_dev *dev,
174 	struct rte_eth_xstat_name *xstats_names,
175 	const uint64_t *ids,
176 	unsigned int limit);
177 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
178 					     uint16_t queue_id,
179 					     uint8_t stat_idx,
180 					     uint8_t is_rx);
181 static int ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
182 				 size_t fw_size);
183 static void ixgbe_dev_info_get(struct rte_eth_dev *dev,
184 			       struct rte_eth_dev_info *dev_info);
185 static const uint32_t *ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev);
186 static void ixgbevf_dev_info_get(struct rte_eth_dev *dev,
187 				 struct rte_eth_dev_info *dev_info);
188 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
189 
190 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
191 		uint16_t vlan_id, int on);
192 static int ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
193 			       enum rte_vlan_type vlan_type,
194 			       uint16_t tpid_id);
195 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
196 		uint16_t queue, bool on);
197 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
198 		int on);
199 static int ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
200 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
201 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
202 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
203 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
204 
205 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
206 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
207 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
208 			       struct rte_eth_fc_conf *fc_conf);
209 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
210 			       struct rte_eth_fc_conf *fc_conf);
211 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
212 		struct rte_eth_pfc_conf *pfc_conf);
213 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
214 			struct rte_eth_rss_reta_entry64 *reta_conf,
215 			uint16_t reta_size);
216 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
217 			struct rte_eth_rss_reta_entry64 *reta_conf,
218 			uint16_t reta_size);
219 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
220 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
221 static int ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev);
222 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
223 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
224 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev,
225 				      struct rte_intr_handle *handle);
226 static void ixgbe_dev_interrupt_handler(void *param);
227 static void ixgbe_dev_interrupt_delayed_handler(void *param);
228 static int ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
229 			 uint32_t index, uint32_t pool);
230 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
231 static void ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
232 					   struct ether_addr *mac_addr);
233 static void ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config);
234 static bool is_device_supported(struct rte_eth_dev *dev,
235 				struct rte_pci_driver *drv);
236 
237 /* For Virtual Function support */
238 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
239 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
240 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
241 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
242 static int ixgbevf_dev_link_update(struct rte_eth_dev *dev,
243 				   int wait_to_complete);
244 static void ixgbevf_dev_stop(struct rte_eth_dev *dev);
245 static void ixgbevf_dev_close(struct rte_eth_dev *dev);
246 static int  ixgbevf_dev_reset(struct rte_eth_dev *dev);
247 static void ixgbevf_intr_disable(struct ixgbe_hw *hw);
248 static void ixgbevf_intr_enable(struct ixgbe_hw *hw);
249 static int ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
250 		struct rte_eth_stats *stats);
251 static void ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
252 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
253 		uint16_t vlan_id, int on);
254 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
255 		uint16_t queue, int on);
256 static int ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
257 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
258 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
259 					    uint16_t queue_id);
260 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
261 					     uint16_t queue_id);
262 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
263 				 uint8_t queue, uint8_t msix_vector);
264 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
265 static void ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev);
266 static void ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev);
267 
268 /* For Eth VMDQ APIs support */
269 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
270 		ether_addr * mac_addr, uint8_t on);
271 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on);
272 static int ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
273 		struct rte_eth_mirror_conf *mirror_conf,
274 		uint8_t rule_id, uint8_t on);
275 static int ixgbe_mirror_rule_reset(struct rte_eth_dev *dev,
276 		uint8_t	rule_id);
277 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
278 					  uint16_t queue_id);
279 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
280 					   uint16_t queue_id);
281 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
282 			       uint8_t queue, uint8_t msix_vector);
283 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
284 
285 static int ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
286 				struct ether_addr *mac_addr,
287 				uint32_t index, uint32_t pool);
288 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
289 static void ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
290 					     struct ether_addr *mac_addr);
291 static int ixgbe_syn_filter_get(struct rte_eth_dev *dev,
292 			struct rte_eth_syn_filter *filter);
293 static int ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
294 			enum rte_filter_op filter_op,
295 			void *arg);
296 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
297 			struct ixgbe_5tuple_filter *filter);
298 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
299 			struct ixgbe_5tuple_filter *filter);
300 static int ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
301 				enum rte_filter_op filter_op,
302 				void *arg);
303 static int ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
304 			struct rte_eth_ntuple_filter *filter);
305 static int ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
306 				enum rte_filter_op filter_op,
307 				void *arg);
308 static int ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
309 			struct rte_eth_ethertype_filter *filter);
310 static int ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
311 		     enum rte_filter_type filter_type,
312 		     enum rte_filter_op filter_op,
313 		     void *arg);
314 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
315 
316 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
317 				      struct ether_addr *mc_addr_set,
318 				      uint32_t nb_mc_addr);
319 static int ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
320 				   struct rte_eth_dcb_info *dcb_info);
321 
322 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
323 static int ixgbe_get_regs(struct rte_eth_dev *dev,
324 			    struct rte_dev_reg_info *regs);
325 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
326 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
327 				struct rte_dev_eeprom_info *eeprom);
328 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
329 				struct rte_dev_eeprom_info *eeprom);
330 
331 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
332 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
333 				struct rte_dev_reg_info *regs);
334 
335 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
336 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
337 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
338 					    struct timespec *timestamp,
339 					    uint32_t flags);
340 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
341 					    struct timespec *timestamp);
342 static int ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
343 static int ixgbe_timesync_read_time(struct rte_eth_dev *dev,
344 				   struct timespec *timestamp);
345 static int ixgbe_timesync_write_time(struct rte_eth_dev *dev,
346 				   const struct timespec *timestamp);
347 static void ixgbevf_dev_interrupt_handler(void *param);
348 
349 static int ixgbe_dev_l2_tunnel_eth_type_conf
350 	(struct rte_eth_dev *dev, struct rte_eth_l2_tunnel_conf *l2_tunnel);
351 static int ixgbe_dev_l2_tunnel_offload_set
352 	(struct rte_eth_dev *dev,
353 	 struct rte_eth_l2_tunnel_conf *l2_tunnel,
354 	 uint32_t mask,
355 	 uint8_t en);
356 static int ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
357 					     enum rte_filter_op filter_op,
358 					     void *arg);
359 
360 static int ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
361 					 struct rte_eth_udp_tunnel *udp_tunnel);
362 static int ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
363 					 struct rte_eth_udp_tunnel *udp_tunnel);
364 static int ixgbe_filter_restore(struct rte_eth_dev *dev);
365 static void ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev);
366 
367 /*
368  * Define VF Stats MACRO for Non "cleared on read" register
369  */
370 #define UPDATE_VF_STAT(reg, last, cur)                          \
371 {                                                               \
372 	uint32_t latest = IXGBE_READ_REG(hw, reg);              \
373 	cur += (latest - last) & UINT_MAX;                      \
374 	last = latest;                                          \
375 }
376 
377 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
378 {                                                                \
379 	u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
380 	u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
381 	u64 latest = ((new_msb << 32) | new_lsb);                \
382 	cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
383 	last = latest;                                           \
384 }
385 
386 #define IXGBE_SET_HWSTRIP(h, q) do {\
387 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
388 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
389 		(h)->bitmap[idx] |= 1 << bit;\
390 	} while (0)
391 
392 #define IXGBE_CLEAR_HWSTRIP(h, q) do {\
393 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
394 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
395 		(h)->bitmap[idx] &= ~(1 << bit);\
396 	} while (0)
397 
398 #define IXGBE_GET_HWSTRIP(h, q, r) do {\
399 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
400 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
401 		(r) = (h)->bitmap[idx] >> bit & 1;\
402 	} while (0)
403 
404 int ixgbe_logtype_init;
405 int ixgbe_logtype_driver;
406 
407 /*
408  * The set of PCI devices this driver supports
409  */
410 static const struct rte_pci_id pci_id_ixgbe_map[] = {
411 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598) },
412 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_BX) },
413 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT) },
414 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT) },
415 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT) },
416 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2) },
417 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM) },
418 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4) },
419 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT) },
420 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT) },
421 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM) },
422 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR) },
423 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4) },
424 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ) },
425 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KR) },
426 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE) },
427 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4) },
428 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP) },
429 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE) },
430 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE) },
431 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_EM) },
432 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2) },
433 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP) },
434 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP) },
435 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP) },
436 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM) },
437 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM) },
438 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_LS) },
439 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T) },
440 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1) },
441 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP) },
442 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T) },
443 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_1G_T) },
444 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T) },
445 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1) },
446 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR) },
447 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR_L) },
448 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP_N) },
449 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII) },
450 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII_L) },
451 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_10G_T) },
452 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP) },
453 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP_N) },
454 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP) },
455 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T) },
456 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T_L) },
457 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4) },
458 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR) },
459 #ifdef RTE_LIBRTE_IXGBE_BYPASS
460 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BYPASS) },
461 #endif
462 	{ .vendor_id = 0, /* sentinel */ },
463 };
464 
465 /*
466  * The set of PCI devices this driver supports (for 82599 VF)
467  */
468 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
469 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF) },
470 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF_HV) },
471 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF) },
472 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF_HV) },
473 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF_HV) },
474 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF) },
475 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF) },
476 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF_HV) },
477 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF) },
478 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF_HV) },
479 	{ .vendor_id = 0, /* sentinel */ },
480 };
481 
482 static const struct rte_eth_desc_lim rx_desc_lim = {
483 	.nb_max = IXGBE_MAX_RING_DESC,
484 	.nb_min = IXGBE_MIN_RING_DESC,
485 	.nb_align = IXGBE_RXD_ALIGN,
486 };
487 
488 static const struct rte_eth_desc_lim tx_desc_lim = {
489 	.nb_max = IXGBE_MAX_RING_DESC,
490 	.nb_min = IXGBE_MIN_RING_DESC,
491 	.nb_align = IXGBE_TXD_ALIGN,
492 	.nb_seg_max = IXGBE_TX_MAX_SEG,
493 	.nb_mtu_seg_max = IXGBE_TX_MAX_SEG,
494 };
495 
496 static const struct eth_dev_ops ixgbe_eth_dev_ops = {
497 	.dev_configure        = ixgbe_dev_configure,
498 	.dev_start            = ixgbe_dev_start,
499 	.dev_stop             = ixgbe_dev_stop,
500 	.dev_set_link_up    = ixgbe_dev_set_link_up,
501 	.dev_set_link_down  = ixgbe_dev_set_link_down,
502 	.dev_close            = ixgbe_dev_close,
503 	.dev_reset	      = ixgbe_dev_reset,
504 	.promiscuous_enable   = ixgbe_dev_promiscuous_enable,
505 	.promiscuous_disable  = ixgbe_dev_promiscuous_disable,
506 	.allmulticast_enable  = ixgbe_dev_allmulticast_enable,
507 	.allmulticast_disable = ixgbe_dev_allmulticast_disable,
508 	.link_update          = ixgbe_dev_link_update,
509 	.stats_get            = ixgbe_dev_stats_get,
510 	.xstats_get           = ixgbe_dev_xstats_get,
511 	.xstats_get_by_id     = ixgbe_dev_xstats_get_by_id,
512 	.stats_reset          = ixgbe_dev_stats_reset,
513 	.xstats_reset         = ixgbe_dev_xstats_reset,
514 	.xstats_get_names     = ixgbe_dev_xstats_get_names,
515 	.xstats_get_names_by_id = ixgbe_dev_xstats_get_names_by_id,
516 	.queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
517 	.fw_version_get       = ixgbe_fw_version_get,
518 	.dev_infos_get        = ixgbe_dev_info_get,
519 	.dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
520 	.mtu_set              = ixgbe_dev_mtu_set,
521 	.vlan_filter_set      = ixgbe_vlan_filter_set,
522 	.vlan_tpid_set        = ixgbe_vlan_tpid_set,
523 	.vlan_offload_set     = ixgbe_vlan_offload_set,
524 	.vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
525 	.rx_queue_start	      = ixgbe_dev_rx_queue_start,
526 	.rx_queue_stop        = ixgbe_dev_rx_queue_stop,
527 	.tx_queue_start	      = ixgbe_dev_tx_queue_start,
528 	.tx_queue_stop        = ixgbe_dev_tx_queue_stop,
529 	.rx_queue_setup       = ixgbe_dev_rx_queue_setup,
530 	.rx_queue_intr_enable = ixgbe_dev_rx_queue_intr_enable,
531 	.rx_queue_intr_disable = ixgbe_dev_rx_queue_intr_disable,
532 	.rx_queue_release     = ixgbe_dev_rx_queue_release,
533 	.rx_queue_count       = ixgbe_dev_rx_queue_count,
534 	.rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
535 	.rx_descriptor_status = ixgbe_dev_rx_descriptor_status,
536 	.tx_descriptor_status = ixgbe_dev_tx_descriptor_status,
537 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
538 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
539 	.dev_led_on           = ixgbe_dev_led_on,
540 	.dev_led_off          = ixgbe_dev_led_off,
541 	.flow_ctrl_get        = ixgbe_flow_ctrl_get,
542 	.flow_ctrl_set        = ixgbe_flow_ctrl_set,
543 	.priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
544 	.mac_addr_add         = ixgbe_add_rar,
545 	.mac_addr_remove      = ixgbe_remove_rar,
546 	.mac_addr_set         = ixgbe_set_default_mac_addr,
547 	.uc_hash_table_set    = ixgbe_uc_hash_table_set,
548 	.uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
549 	.mirror_rule_set      = ixgbe_mirror_rule_set,
550 	.mirror_rule_reset    = ixgbe_mirror_rule_reset,
551 	.set_queue_rate_limit = ixgbe_set_queue_rate_limit,
552 	.reta_update          = ixgbe_dev_rss_reta_update,
553 	.reta_query           = ixgbe_dev_rss_reta_query,
554 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
555 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
556 	.filter_ctrl          = ixgbe_dev_filter_ctrl,
557 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
558 	.rxq_info_get         = ixgbe_rxq_info_get,
559 	.txq_info_get         = ixgbe_txq_info_get,
560 	.timesync_enable      = ixgbe_timesync_enable,
561 	.timesync_disable     = ixgbe_timesync_disable,
562 	.timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
563 	.timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
564 	.get_reg              = ixgbe_get_regs,
565 	.get_eeprom_length    = ixgbe_get_eeprom_length,
566 	.get_eeprom           = ixgbe_get_eeprom,
567 	.set_eeprom           = ixgbe_set_eeprom,
568 	.get_dcb_info         = ixgbe_dev_get_dcb_info,
569 	.timesync_adjust_time = ixgbe_timesync_adjust_time,
570 	.timesync_read_time   = ixgbe_timesync_read_time,
571 	.timesync_write_time  = ixgbe_timesync_write_time,
572 	.l2_tunnel_eth_type_conf = ixgbe_dev_l2_tunnel_eth_type_conf,
573 	.l2_tunnel_offload_set   = ixgbe_dev_l2_tunnel_offload_set,
574 	.udp_tunnel_port_add  = ixgbe_dev_udp_tunnel_port_add,
575 	.udp_tunnel_port_del  = ixgbe_dev_udp_tunnel_port_del,
576 	.tm_ops_get           = ixgbe_tm_ops_get,
577 };
578 
579 /*
580  * dev_ops for virtual function, bare necessities for basic vf
581  * operation have been implemented
582  */
583 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
584 	.dev_configure        = ixgbevf_dev_configure,
585 	.dev_start            = ixgbevf_dev_start,
586 	.dev_stop             = ixgbevf_dev_stop,
587 	.link_update          = ixgbevf_dev_link_update,
588 	.stats_get            = ixgbevf_dev_stats_get,
589 	.xstats_get           = ixgbevf_dev_xstats_get,
590 	.stats_reset          = ixgbevf_dev_stats_reset,
591 	.xstats_reset         = ixgbevf_dev_stats_reset,
592 	.xstats_get_names     = ixgbevf_dev_xstats_get_names,
593 	.dev_close            = ixgbevf_dev_close,
594 	.dev_reset	      = ixgbevf_dev_reset,
595 	.allmulticast_enable  = ixgbevf_dev_allmulticast_enable,
596 	.allmulticast_disable = ixgbevf_dev_allmulticast_disable,
597 	.dev_infos_get        = ixgbevf_dev_info_get,
598 	.dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
599 	.mtu_set              = ixgbevf_dev_set_mtu,
600 	.vlan_filter_set      = ixgbevf_vlan_filter_set,
601 	.vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
602 	.vlan_offload_set     = ixgbevf_vlan_offload_set,
603 	.rx_queue_setup       = ixgbe_dev_rx_queue_setup,
604 	.rx_queue_release     = ixgbe_dev_rx_queue_release,
605 	.rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
606 	.rx_descriptor_status = ixgbe_dev_rx_descriptor_status,
607 	.tx_descriptor_status = ixgbe_dev_tx_descriptor_status,
608 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
609 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
610 	.rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
611 	.rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
612 	.mac_addr_add         = ixgbevf_add_mac_addr,
613 	.mac_addr_remove      = ixgbevf_remove_mac_addr,
614 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
615 	.rxq_info_get         = ixgbe_rxq_info_get,
616 	.txq_info_get         = ixgbe_txq_info_get,
617 	.mac_addr_set         = ixgbevf_set_default_mac_addr,
618 	.get_reg              = ixgbevf_get_regs,
619 	.reta_update          = ixgbe_dev_rss_reta_update,
620 	.reta_query           = ixgbe_dev_rss_reta_query,
621 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
622 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
623 };
624 
625 /* store statistics names and its offset in stats structure */
626 struct rte_ixgbe_xstats_name_off {
627 	char name[RTE_ETH_XSTATS_NAME_SIZE];
628 	unsigned offset;
629 };
630 
631 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
632 	{"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
633 	{"rx_illegal_byte_errors", offsetof(struct ixgbe_hw_stats, illerrc)},
634 	{"rx_error_bytes", offsetof(struct ixgbe_hw_stats, errbc)},
635 	{"mac_local_errors", offsetof(struct ixgbe_hw_stats, mlfc)},
636 	{"mac_remote_errors", offsetof(struct ixgbe_hw_stats, mrfc)},
637 	{"rx_length_errors", offsetof(struct ixgbe_hw_stats, rlec)},
638 	{"tx_xon_packets", offsetof(struct ixgbe_hw_stats, lxontxc)},
639 	{"rx_xon_packets", offsetof(struct ixgbe_hw_stats, lxonrxc)},
640 	{"tx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxofftxc)},
641 	{"rx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
642 	{"rx_size_64_packets", offsetof(struct ixgbe_hw_stats, prc64)},
643 	{"rx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, prc127)},
644 	{"rx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, prc255)},
645 	{"rx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, prc511)},
646 	{"rx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
647 		prc1023)},
648 	{"rx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
649 		prc1522)},
650 	{"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
651 	{"rx_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
652 	{"rx_fragment_errors", offsetof(struct ixgbe_hw_stats, rfc)},
653 	{"rx_undersize_errors", offsetof(struct ixgbe_hw_stats, ruc)},
654 	{"rx_oversize_errors", offsetof(struct ixgbe_hw_stats, roc)},
655 	{"rx_jabber_errors", offsetof(struct ixgbe_hw_stats, rjc)},
656 	{"rx_management_packets", offsetof(struct ixgbe_hw_stats, mngprc)},
657 	{"rx_management_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
658 	{"tx_management_packets", offsetof(struct ixgbe_hw_stats, mngptc)},
659 	{"rx_total_packets", offsetof(struct ixgbe_hw_stats, tpr)},
660 	{"rx_total_bytes", offsetof(struct ixgbe_hw_stats, tor)},
661 	{"tx_total_packets", offsetof(struct ixgbe_hw_stats, tpt)},
662 	{"tx_size_64_packets", offsetof(struct ixgbe_hw_stats, ptc64)},
663 	{"tx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, ptc127)},
664 	{"tx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, ptc255)},
665 	{"tx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, ptc511)},
666 	{"tx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
667 		ptc1023)},
668 	{"tx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
669 		ptc1522)},
670 	{"tx_multicast_packets", offsetof(struct ixgbe_hw_stats, mptc)},
671 	{"tx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bptc)},
672 	{"rx_mac_short_packet_dropped", offsetof(struct ixgbe_hw_stats, mspdc)},
673 	{"rx_l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
674 
675 	{"flow_director_added_filters", offsetof(struct ixgbe_hw_stats,
676 		fdirustat_add)},
677 	{"flow_director_removed_filters", offsetof(struct ixgbe_hw_stats,
678 		fdirustat_remove)},
679 	{"flow_director_filter_add_errors", offsetof(struct ixgbe_hw_stats,
680 		fdirfstat_fadd)},
681 	{"flow_director_filter_remove_errors", offsetof(struct ixgbe_hw_stats,
682 		fdirfstat_fremove)},
683 	{"flow_director_matched_filters", offsetof(struct ixgbe_hw_stats,
684 		fdirmatch)},
685 	{"flow_director_missed_filters", offsetof(struct ixgbe_hw_stats,
686 		fdirmiss)},
687 
688 	{"rx_fcoe_crc_errors", offsetof(struct ixgbe_hw_stats, fccrc)},
689 	{"rx_fcoe_dropped", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
690 	{"rx_fcoe_mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats,
691 		fclast)},
692 	{"rx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeprc)},
693 	{"tx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeptc)},
694 	{"rx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwrc)},
695 	{"tx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwtc)},
696 	{"rx_fcoe_no_direct_data_placement", offsetof(struct ixgbe_hw_stats,
697 		fcoe_noddp)},
698 	{"rx_fcoe_no_direct_data_placement_ext_buff",
699 		offsetof(struct ixgbe_hw_stats, fcoe_noddp_ext_buff)},
700 
701 	{"tx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
702 		lxontxc)},
703 	{"rx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
704 		lxonrxc)},
705 	{"tx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
706 		lxofftxc)},
707 	{"rx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
708 		lxoffrxc)},
709 	{"rx_total_missed_packets", offsetof(struct ixgbe_hw_stats, mpctotal)},
710 };
711 
712 #define IXGBE_NB_HW_STATS (sizeof(rte_ixgbe_stats_strings) / \
713 			   sizeof(rte_ixgbe_stats_strings[0]))
714 
715 /* MACsec statistics */
716 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_macsec_strings[] = {
717 	{"out_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
718 		out_pkts_untagged)},
719 	{"out_pkts_encrypted", offsetof(struct ixgbe_macsec_stats,
720 		out_pkts_encrypted)},
721 	{"out_pkts_protected", offsetof(struct ixgbe_macsec_stats,
722 		out_pkts_protected)},
723 	{"out_octets_encrypted", offsetof(struct ixgbe_macsec_stats,
724 		out_octets_encrypted)},
725 	{"out_octets_protected", offsetof(struct ixgbe_macsec_stats,
726 		out_octets_protected)},
727 	{"in_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
728 		in_pkts_untagged)},
729 	{"in_pkts_badtag", offsetof(struct ixgbe_macsec_stats,
730 		in_pkts_badtag)},
731 	{"in_pkts_nosci", offsetof(struct ixgbe_macsec_stats,
732 		in_pkts_nosci)},
733 	{"in_pkts_unknownsci", offsetof(struct ixgbe_macsec_stats,
734 		in_pkts_unknownsci)},
735 	{"in_octets_decrypted", offsetof(struct ixgbe_macsec_stats,
736 		in_octets_decrypted)},
737 	{"in_octets_validated", offsetof(struct ixgbe_macsec_stats,
738 		in_octets_validated)},
739 	{"in_pkts_unchecked", offsetof(struct ixgbe_macsec_stats,
740 		in_pkts_unchecked)},
741 	{"in_pkts_delayed", offsetof(struct ixgbe_macsec_stats,
742 		in_pkts_delayed)},
743 	{"in_pkts_late", offsetof(struct ixgbe_macsec_stats,
744 		in_pkts_late)},
745 	{"in_pkts_ok", offsetof(struct ixgbe_macsec_stats,
746 		in_pkts_ok)},
747 	{"in_pkts_invalid", offsetof(struct ixgbe_macsec_stats,
748 		in_pkts_invalid)},
749 	{"in_pkts_notvalid", offsetof(struct ixgbe_macsec_stats,
750 		in_pkts_notvalid)},
751 	{"in_pkts_unusedsa", offsetof(struct ixgbe_macsec_stats,
752 		in_pkts_unusedsa)},
753 	{"in_pkts_notusingsa", offsetof(struct ixgbe_macsec_stats,
754 		in_pkts_notusingsa)},
755 };
756 
757 #define IXGBE_NB_MACSEC_STATS (sizeof(rte_ixgbe_macsec_strings) / \
758 			   sizeof(rte_ixgbe_macsec_strings[0]))
759 
760 /* Per-queue statistics */
761 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_rxq_strings[] = {
762 	{"mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats, rnbc)},
763 	{"dropped", offsetof(struct ixgbe_hw_stats, mpc)},
764 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxonrxc)},
765 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxoffrxc)},
766 };
767 
768 #define IXGBE_NB_RXQ_PRIO_STATS (sizeof(rte_ixgbe_rxq_strings) / \
769 			   sizeof(rte_ixgbe_rxq_strings[0]))
770 #define IXGBE_NB_RXQ_PRIO_VALUES 8
771 
772 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_txq_strings[] = {
773 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxontxc)},
774 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxofftxc)},
775 	{"xon_to_xoff_packets", offsetof(struct ixgbe_hw_stats,
776 		pxon2offc)},
777 };
778 
779 #define IXGBE_NB_TXQ_PRIO_STATS (sizeof(rte_ixgbe_txq_strings) / \
780 			   sizeof(rte_ixgbe_txq_strings[0]))
781 #define IXGBE_NB_TXQ_PRIO_VALUES 8
782 
783 static const struct rte_ixgbe_xstats_name_off rte_ixgbevf_stats_strings[] = {
784 	{"rx_multicast_packets", offsetof(struct ixgbevf_hw_stats, vfmprc)},
785 };
786 
787 #define IXGBEVF_NB_XSTATS (sizeof(rte_ixgbevf_stats_strings) /	\
788 		sizeof(rte_ixgbevf_stats_strings[0]))
789 
790 /**
791  * Atomically reads the link status information from global
792  * structure rte_eth_dev.
793  *
794  * @param dev
795  *   - Pointer to the structure rte_eth_dev to read from.
796  *   - Pointer to the buffer to be saved with the link status.
797  *
798  * @return
799  *   - On success, zero.
800  *   - On failure, negative value.
801  */
802 static inline int
803 rte_ixgbe_dev_atomic_read_link_status(struct rte_eth_dev *dev,
804 				struct rte_eth_link *link)
805 {
806 	struct rte_eth_link *dst = link;
807 	struct rte_eth_link *src = &(dev->data->dev_link);
808 
809 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
810 					*(uint64_t *)src) == 0)
811 		return -1;
812 
813 	return 0;
814 }
815 
816 /**
817  * Atomically writes the link status information into global
818  * structure rte_eth_dev.
819  *
820  * @param dev
821  *   - Pointer to the structure rte_eth_dev to read from.
822  *   - Pointer to the buffer to be saved with the link status.
823  *
824  * @return
825  *   - On success, zero.
826  *   - On failure, negative value.
827  */
828 static inline int
829 rte_ixgbe_dev_atomic_write_link_status(struct rte_eth_dev *dev,
830 				struct rte_eth_link *link)
831 {
832 	struct rte_eth_link *dst = &(dev->data->dev_link);
833 	struct rte_eth_link *src = link;
834 
835 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
836 					*(uint64_t *)src) == 0)
837 		return -1;
838 
839 	return 0;
840 }
841 
842 /*
843  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
844  */
845 static inline int
846 ixgbe_is_sfp(struct ixgbe_hw *hw)
847 {
848 	switch (hw->phy.type) {
849 	case ixgbe_phy_sfp_avago:
850 	case ixgbe_phy_sfp_ftl:
851 	case ixgbe_phy_sfp_intel:
852 	case ixgbe_phy_sfp_unknown:
853 	case ixgbe_phy_sfp_passive_tyco:
854 	case ixgbe_phy_sfp_passive_unknown:
855 		return 1;
856 	default:
857 		return 0;
858 	}
859 }
860 
861 static inline int32_t
862 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
863 {
864 	uint32_t ctrl_ext;
865 	int32_t status;
866 
867 	status = ixgbe_reset_hw(hw);
868 
869 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
870 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
871 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
872 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
873 	IXGBE_WRITE_FLUSH(hw);
874 
875 	if (status == IXGBE_ERR_SFP_NOT_PRESENT)
876 		status = IXGBE_SUCCESS;
877 	return status;
878 }
879 
880 static inline void
881 ixgbe_enable_intr(struct rte_eth_dev *dev)
882 {
883 	struct ixgbe_interrupt *intr =
884 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
885 	struct ixgbe_hw *hw =
886 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
887 
888 	IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
889 	IXGBE_WRITE_FLUSH(hw);
890 }
891 
892 /*
893  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
894  */
895 static void
896 ixgbe_disable_intr(struct ixgbe_hw *hw)
897 {
898 	PMD_INIT_FUNC_TRACE();
899 
900 	if (hw->mac.type == ixgbe_mac_82598EB) {
901 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
902 	} else {
903 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
904 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
905 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
906 	}
907 	IXGBE_WRITE_FLUSH(hw);
908 }
909 
910 /*
911  * This function resets queue statistics mapping registers.
912  * From Niantic datasheet, Initialization of Statistics section:
913  * "...if software requires the queue counters, the RQSMR and TQSM registers
914  * must be re-programmed following a device reset.
915  */
916 static void
917 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
918 {
919 	uint32_t i;
920 
921 	for (i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
922 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
923 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
924 	}
925 }
926 
927 
928 static int
929 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
930 				  uint16_t queue_id,
931 				  uint8_t stat_idx,
932 				  uint8_t is_rx)
933 {
934 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
935 #define NB_QMAP_FIELDS_PER_QSM_REG 4
936 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
937 
938 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
939 	struct ixgbe_stat_mapping_registers *stat_mappings =
940 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
941 	uint32_t qsmr_mask = 0;
942 	uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
943 	uint32_t q_map;
944 	uint8_t n, offset;
945 
946 	if ((hw->mac.type != ixgbe_mac_82599EB) &&
947 		(hw->mac.type != ixgbe_mac_X540) &&
948 		(hw->mac.type != ixgbe_mac_X550) &&
949 		(hw->mac.type != ixgbe_mac_X550EM_x) &&
950 		(hw->mac.type != ixgbe_mac_X550EM_a))
951 		return -ENOSYS;
952 
953 	PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
954 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
955 		     queue_id, stat_idx);
956 
957 	n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
958 	if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
959 		PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
960 		return -EIO;
961 	}
962 	offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
963 
964 	/* Now clear any previous stat_idx set */
965 	clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
966 	if (!is_rx)
967 		stat_mappings->tqsm[n] &= ~clearing_mask;
968 	else
969 		stat_mappings->rqsmr[n] &= ~clearing_mask;
970 
971 	q_map = (uint32_t)stat_idx;
972 	q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
973 	qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
974 	if (!is_rx)
975 		stat_mappings->tqsm[n] |= qsmr_mask;
976 	else
977 		stat_mappings->rqsmr[n] |= qsmr_mask;
978 
979 	PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
980 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
981 		     queue_id, stat_idx);
982 	PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
983 		     is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
984 
985 	/* Now write the mapping in the appropriate register */
986 	if (is_rx) {
987 		PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
988 			     stat_mappings->rqsmr[n], n);
989 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
990 	} else {
991 		PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
992 			     stat_mappings->tqsm[n], n);
993 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
994 	}
995 	return 0;
996 }
997 
998 static void
999 ixgbe_restore_statistics_mapping(struct rte_eth_dev *dev)
1000 {
1001 	struct ixgbe_stat_mapping_registers *stat_mappings =
1002 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
1003 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1004 	int i;
1005 
1006 	/* write whatever was in stat mapping table to the NIC */
1007 	for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
1008 		/* rx */
1009 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
1010 
1011 		/* tx */
1012 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
1013 	}
1014 }
1015 
1016 static void
1017 ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config)
1018 {
1019 	uint8_t i;
1020 	struct ixgbe_dcb_tc_config *tc;
1021 	uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
1022 
1023 	dcb_config->num_tcs.pg_tcs = dcb_max_tc;
1024 	dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
1025 	for (i = 0; i < dcb_max_tc; i++) {
1026 		tc = &dcb_config->tc_config[i];
1027 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
1028 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
1029 				 (uint8_t)(100/dcb_max_tc + (i & 1));
1030 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
1031 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
1032 				 (uint8_t)(100/dcb_max_tc + (i & 1));
1033 		tc->pfc = ixgbe_dcb_pfc_disabled;
1034 	}
1035 
1036 	/* Initialize default user to priority mapping, UPx->TC0 */
1037 	tc = &dcb_config->tc_config[0];
1038 	tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
1039 	tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
1040 	for (i = 0; i < IXGBE_DCB_MAX_BW_GROUP; i++) {
1041 		dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
1042 		dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
1043 	}
1044 	dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
1045 	dcb_config->pfc_mode_enable = false;
1046 	dcb_config->vt_mode = true;
1047 	dcb_config->round_robin_enable = false;
1048 	/* support all DCB capabilities in 82599 */
1049 	dcb_config->support.capabilities = 0xFF;
1050 
1051 	/*we only support 4 Tcs for X540, X550 */
1052 	if (hw->mac.type == ixgbe_mac_X540 ||
1053 		hw->mac.type == ixgbe_mac_X550 ||
1054 		hw->mac.type == ixgbe_mac_X550EM_x ||
1055 		hw->mac.type == ixgbe_mac_X550EM_a) {
1056 		dcb_config->num_tcs.pg_tcs = 4;
1057 		dcb_config->num_tcs.pfc_tcs = 4;
1058 	}
1059 }
1060 
1061 /*
1062  * Ensure that all locks are released before first NVM or PHY access
1063  */
1064 static void
1065 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
1066 {
1067 	uint16_t mask;
1068 
1069 	/*
1070 	 * Phy lock should not fail in this early stage. If this is the case,
1071 	 * it is due to an improper exit of the application.
1072 	 * So force the release of the faulty lock. Release of common lock
1073 	 * is done automatically by swfw_sync function.
1074 	 */
1075 	mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
1076 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1077 		PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
1078 	}
1079 	ixgbe_release_swfw_semaphore(hw, mask);
1080 
1081 	/*
1082 	 * These ones are more tricky since they are common to all ports; but
1083 	 * swfw_sync retries last long enough (1s) to be almost sure that if
1084 	 * lock can not be taken it is due to an improper lock of the
1085 	 * semaphore.
1086 	 */
1087 	mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
1088 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1089 		PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1090 	}
1091 	ixgbe_release_swfw_semaphore(hw, mask);
1092 }
1093 
1094 /*
1095  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
1096  * It returns 0 on success.
1097  */
1098 static int
1099 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev)
1100 {
1101 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1102 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1103 	struct ixgbe_hw *hw =
1104 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1105 	struct ixgbe_vfta *shadow_vfta =
1106 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1107 	struct ixgbe_hwstrip *hwstrip =
1108 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1109 	struct ixgbe_dcb_config *dcb_config =
1110 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
1111 	struct ixgbe_filter_info *filter_info =
1112 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1113 	struct ixgbe_bw_conf *bw_conf =
1114 		IXGBE_DEV_PRIVATE_TO_BW_CONF(eth_dev->data->dev_private);
1115 	uint32_t ctrl_ext;
1116 	uint16_t csum;
1117 	int diag, i;
1118 
1119 	PMD_INIT_FUNC_TRACE();
1120 
1121 	eth_dev->dev_ops = &ixgbe_eth_dev_ops;
1122 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1123 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1124 	eth_dev->tx_pkt_prepare = &ixgbe_prep_pkts;
1125 
1126 	/*
1127 	 * For secondary processes, we don't initialise any further as primary
1128 	 * has already done this work. Only check we don't need a different
1129 	 * RX and TX function.
1130 	 */
1131 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1132 		struct ixgbe_tx_queue *txq;
1133 		/* TX queue function in primary, set by last queue initialized
1134 		 * Tx queue may not initialized by primary process
1135 		 */
1136 		if (eth_dev->data->tx_queues) {
1137 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
1138 			ixgbe_set_tx_function(eth_dev, txq);
1139 		} else {
1140 			/* Use default TX function if we get here */
1141 			PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
1142 				     "Using default TX function.");
1143 		}
1144 
1145 		ixgbe_set_rx_function(eth_dev);
1146 
1147 		return 0;
1148 	}
1149 
1150 #ifdef RTE_LIBRTE_SECURITY
1151 	/* Initialize security_ctx only for primary process*/
1152 	eth_dev->security_ctx = ixgbe_ipsec_ctx_create(eth_dev);
1153 	if (eth_dev->security_ctx == NULL)
1154 		return -ENOMEM;
1155 #endif
1156 
1157 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1158 
1159 	/* Vendor and Device ID need to be set before init of shared code */
1160 	hw->device_id = pci_dev->id.device_id;
1161 	hw->vendor_id = pci_dev->id.vendor_id;
1162 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1163 	hw->allow_unsupported_sfp = 1;
1164 
1165 	/* Initialize the shared code (base driver) */
1166 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1167 	diag = ixgbe_bypass_init_shared_code(hw);
1168 #else
1169 	diag = ixgbe_init_shared_code(hw);
1170 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1171 
1172 	if (diag != IXGBE_SUCCESS) {
1173 		PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
1174 		return -EIO;
1175 	}
1176 
1177 	/* pick up the PCI bus settings for reporting later */
1178 	ixgbe_get_bus_info(hw);
1179 
1180 	/* Unlock any pending hardware semaphore */
1181 	ixgbe_swfw_lock_reset(hw);
1182 
1183 	/* Initialize DCB configuration*/
1184 	memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
1185 	ixgbe_dcb_init(hw, dcb_config);
1186 	/* Get Hardware Flow Control setting */
1187 	hw->fc.requested_mode = ixgbe_fc_full;
1188 	hw->fc.current_mode = ixgbe_fc_full;
1189 	hw->fc.pause_time = IXGBE_FC_PAUSE;
1190 	for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
1191 		hw->fc.low_water[i] = IXGBE_FC_LO;
1192 		hw->fc.high_water[i] = IXGBE_FC_HI;
1193 	}
1194 	hw->fc.send_xon = 1;
1195 
1196 	/* Make sure we have a good EEPROM before we read from it */
1197 	diag = ixgbe_validate_eeprom_checksum(hw, &csum);
1198 	if (diag != IXGBE_SUCCESS) {
1199 		PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
1200 		return -EIO;
1201 	}
1202 
1203 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1204 	diag = ixgbe_bypass_init_hw(hw);
1205 #else
1206 	diag = ixgbe_init_hw(hw);
1207 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1208 
1209 	/*
1210 	 * Devices with copper phys will fail to initialise if ixgbe_init_hw()
1211 	 * is called too soon after the kernel driver unbinding/binding occurs.
1212 	 * The failure occurs in ixgbe_identify_phy_generic() for all devices,
1213 	 * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
1214 	 * also called. See ixgbe_identify_phy_82599(). The reason for the
1215 	 * failure is not known, and only occuts when virtualisation features
1216 	 * are disabled in the bios. A delay of 100ms  was found to be enough by
1217 	 * trial-and-error, and is doubled to be safe.
1218 	 */
1219 	if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
1220 		rte_delay_ms(200);
1221 		diag = ixgbe_init_hw(hw);
1222 	}
1223 
1224 	if (diag == IXGBE_ERR_SFP_NOT_PRESENT)
1225 		diag = IXGBE_SUCCESS;
1226 
1227 	if (diag == IXGBE_ERR_EEPROM_VERSION) {
1228 		PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
1229 			     "LOM.  Please be aware there may be issues associated "
1230 			     "with your hardware.");
1231 		PMD_INIT_LOG(ERR, "If you are experiencing problems "
1232 			     "please contact your Intel or hardware representative "
1233 			     "who provided you with this hardware.");
1234 	} else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
1235 		PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
1236 	if (diag) {
1237 		PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
1238 		return -EIO;
1239 	}
1240 
1241 	/* Reset the hw statistics */
1242 	ixgbe_dev_stats_reset(eth_dev);
1243 
1244 	/* disable interrupt */
1245 	ixgbe_disable_intr(hw);
1246 
1247 	/* reset mappings for queue statistics hw counters*/
1248 	ixgbe_reset_qstat_mappings(hw);
1249 
1250 	/* Allocate memory for storing MAC addresses */
1251 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
1252 					       hw->mac.num_rar_entries, 0);
1253 	if (eth_dev->data->mac_addrs == NULL) {
1254 		PMD_INIT_LOG(ERR,
1255 			     "Failed to allocate %u bytes needed to store "
1256 			     "MAC addresses",
1257 			     ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1258 		return -ENOMEM;
1259 	}
1260 	/* Copy the permanent MAC address */
1261 	ether_addr_copy((struct ether_addr *) hw->mac.perm_addr,
1262 			&eth_dev->data->mac_addrs[0]);
1263 
1264 	/* Allocate memory for storing hash filter MAC addresses */
1265 	eth_dev->data->hash_mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
1266 						    IXGBE_VMDQ_NUM_UC_MAC, 0);
1267 	if (eth_dev->data->hash_mac_addrs == NULL) {
1268 		PMD_INIT_LOG(ERR,
1269 			     "Failed to allocate %d bytes needed to store MAC addresses",
1270 			     ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
1271 		return -ENOMEM;
1272 	}
1273 
1274 	/* initialize the vfta */
1275 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1276 
1277 	/* initialize the hw strip bitmap*/
1278 	memset(hwstrip, 0, sizeof(*hwstrip));
1279 
1280 	/* initialize PF if max_vfs not zero */
1281 	ixgbe_pf_host_init(eth_dev);
1282 
1283 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1284 	/* let hardware know driver is loaded */
1285 	ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1286 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1287 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1288 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1289 	IXGBE_WRITE_FLUSH(hw);
1290 
1291 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1292 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1293 			     (int) hw->mac.type, (int) hw->phy.type,
1294 			     (int) hw->phy.sfp_type);
1295 	else
1296 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1297 			     (int) hw->mac.type, (int) hw->phy.type);
1298 
1299 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1300 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1301 		     pci_dev->id.device_id);
1302 
1303 	rte_intr_callback_register(intr_handle,
1304 				   ixgbe_dev_interrupt_handler, eth_dev);
1305 
1306 	/* enable uio/vfio intr/eventfd mapping */
1307 	rte_intr_enable(intr_handle);
1308 
1309 	/* enable support intr */
1310 	ixgbe_enable_intr(eth_dev);
1311 
1312 	/* initialize filter info */
1313 	memset(filter_info, 0,
1314 	       sizeof(struct ixgbe_filter_info));
1315 
1316 	/* initialize 5tuple filter list */
1317 	TAILQ_INIT(&filter_info->fivetuple_list);
1318 
1319 	/* initialize flow director filter list & hash */
1320 	ixgbe_fdir_filter_init(eth_dev);
1321 
1322 	/* initialize l2 tunnel filter list & hash */
1323 	ixgbe_l2_tn_filter_init(eth_dev);
1324 
1325 	/* initialize flow filter lists */
1326 	ixgbe_filterlist_init();
1327 
1328 	/* initialize bandwidth configuration info */
1329 	memset(bw_conf, 0, sizeof(struct ixgbe_bw_conf));
1330 
1331 	/* initialize Traffic Manager configuration */
1332 	ixgbe_tm_conf_init(eth_dev);
1333 
1334 	return 0;
1335 }
1336 
1337 static int
1338 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1339 {
1340 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1341 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1342 	struct ixgbe_hw *hw;
1343 
1344 	PMD_INIT_FUNC_TRACE();
1345 
1346 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1347 		return -EPERM;
1348 
1349 	hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1350 
1351 	if (hw->adapter_stopped == 0)
1352 		ixgbe_dev_close(eth_dev);
1353 
1354 	eth_dev->dev_ops = NULL;
1355 	eth_dev->rx_pkt_burst = NULL;
1356 	eth_dev->tx_pkt_burst = NULL;
1357 
1358 	/* Unlock any pending hardware semaphore */
1359 	ixgbe_swfw_lock_reset(hw);
1360 
1361 	/* disable uio intr before callback unregister */
1362 	rte_intr_disable(intr_handle);
1363 	rte_intr_callback_unregister(intr_handle,
1364 				     ixgbe_dev_interrupt_handler, eth_dev);
1365 
1366 	/* uninitialize PF if max_vfs not zero */
1367 	ixgbe_pf_host_uninit(eth_dev);
1368 
1369 	rte_free(eth_dev->data->mac_addrs);
1370 	eth_dev->data->mac_addrs = NULL;
1371 
1372 	rte_free(eth_dev->data->hash_mac_addrs);
1373 	eth_dev->data->hash_mac_addrs = NULL;
1374 
1375 	/* remove all the fdir filters & hash */
1376 	ixgbe_fdir_filter_uninit(eth_dev);
1377 
1378 	/* remove all the L2 tunnel filters & hash */
1379 	ixgbe_l2_tn_filter_uninit(eth_dev);
1380 
1381 	/* Remove all ntuple filters of the device */
1382 	ixgbe_ntuple_filter_uninit(eth_dev);
1383 
1384 	/* clear all the filters list */
1385 	ixgbe_filterlist_flush();
1386 
1387 	/* Remove all Traffic Manager configuration */
1388 	ixgbe_tm_conf_uninit(eth_dev);
1389 
1390 #ifdef RTE_LIBRTE_SECURITY
1391 	rte_free(eth_dev->security_ctx);
1392 #endif
1393 
1394 	return 0;
1395 }
1396 
1397 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
1398 {
1399 	struct ixgbe_filter_info *filter_info =
1400 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1401 	struct ixgbe_5tuple_filter *p_5tuple;
1402 
1403 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
1404 		TAILQ_REMOVE(&filter_info->fivetuple_list,
1405 			     p_5tuple,
1406 			     entries);
1407 		rte_free(p_5tuple);
1408 	}
1409 	memset(filter_info->fivetuple_mask, 0,
1410 	       sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1411 
1412 	return 0;
1413 }
1414 
1415 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev)
1416 {
1417 	struct ixgbe_hw_fdir_info *fdir_info =
1418 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1419 	struct ixgbe_fdir_filter *fdir_filter;
1420 
1421 		if (fdir_info->hash_map)
1422 		rte_free(fdir_info->hash_map);
1423 	if (fdir_info->hash_handle)
1424 		rte_hash_free(fdir_info->hash_handle);
1425 
1426 	while ((fdir_filter = TAILQ_FIRST(&fdir_info->fdir_list))) {
1427 		TAILQ_REMOVE(&fdir_info->fdir_list,
1428 			     fdir_filter,
1429 			     entries);
1430 		rte_free(fdir_filter);
1431 	}
1432 
1433 	return 0;
1434 }
1435 
1436 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev)
1437 {
1438 	struct ixgbe_l2_tn_info *l2_tn_info =
1439 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1440 	struct ixgbe_l2_tn_filter *l2_tn_filter;
1441 
1442 	if (l2_tn_info->hash_map)
1443 		rte_free(l2_tn_info->hash_map);
1444 	if (l2_tn_info->hash_handle)
1445 		rte_hash_free(l2_tn_info->hash_handle);
1446 
1447 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
1448 		TAILQ_REMOVE(&l2_tn_info->l2_tn_list,
1449 			     l2_tn_filter,
1450 			     entries);
1451 		rte_free(l2_tn_filter);
1452 	}
1453 
1454 	return 0;
1455 }
1456 
1457 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev)
1458 {
1459 	struct ixgbe_hw_fdir_info *fdir_info =
1460 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1461 	char fdir_hash_name[RTE_HASH_NAMESIZE];
1462 	struct rte_hash_parameters fdir_hash_params = {
1463 		.name = fdir_hash_name,
1464 		.entries = IXGBE_MAX_FDIR_FILTER_NUM,
1465 		.key_len = sizeof(union ixgbe_atr_input),
1466 		.hash_func = rte_hash_crc,
1467 		.hash_func_init_val = 0,
1468 		.socket_id = rte_socket_id(),
1469 	};
1470 
1471 	TAILQ_INIT(&fdir_info->fdir_list);
1472 	snprintf(fdir_hash_name, RTE_HASH_NAMESIZE,
1473 		 "fdir_%s", eth_dev->device->name);
1474 	fdir_info->hash_handle = rte_hash_create(&fdir_hash_params);
1475 	if (!fdir_info->hash_handle) {
1476 		PMD_INIT_LOG(ERR, "Failed to create fdir hash table!");
1477 		return -EINVAL;
1478 	}
1479 	fdir_info->hash_map = rte_zmalloc("ixgbe",
1480 					  sizeof(struct ixgbe_fdir_filter *) *
1481 					  IXGBE_MAX_FDIR_FILTER_NUM,
1482 					  0);
1483 	if (!fdir_info->hash_map) {
1484 		PMD_INIT_LOG(ERR,
1485 			     "Failed to allocate memory for fdir hash map!");
1486 		return -ENOMEM;
1487 	}
1488 	fdir_info->mask_added = FALSE;
1489 
1490 	return 0;
1491 }
1492 
1493 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev)
1494 {
1495 	struct ixgbe_l2_tn_info *l2_tn_info =
1496 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1497 	char l2_tn_hash_name[RTE_HASH_NAMESIZE];
1498 	struct rte_hash_parameters l2_tn_hash_params = {
1499 		.name = l2_tn_hash_name,
1500 		.entries = IXGBE_MAX_L2_TN_FILTER_NUM,
1501 		.key_len = sizeof(struct ixgbe_l2_tn_key),
1502 		.hash_func = rte_hash_crc,
1503 		.hash_func_init_val = 0,
1504 		.socket_id = rte_socket_id(),
1505 	};
1506 
1507 	TAILQ_INIT(&l2_tn_info->l2_tn_list);
1508 	snprintf(l2_tn_hash_name, RTE_HASH_NAMESIZE,
1509 		 "l2_tn_%s", eth_dev->device->name);
1510 	l2_tn_info->hash_handle = rte_hash_create(&l2_tn_hash_params);
1511 	if (!l2_tn_info->hash_handle) {
1512 		PMD_INIT_LOG(ERR, "Failed to create L2 TN hash table!");
1513 		return -EINVAL;
1514 	}
1515 	l2_tn_info->hash_map = rte_zmalloc("ixgbe",
1516 				   sizeof(struct ixgbe_l2_tn_filter *) *
1517 				   IXGBE_MAX_L2_TN_FILTER_NUM,
1518 				   0);
1519 	if (!l2_tn_info->hash_map) {
1520 		PMD_INIT_LOG(ERR,
1521 			"Failed to allocate memory for L2 TN hash map!");
1522 		return -ENOMEM;
1523 	}
1524 	l2_tn_info->e_tag_en = FALSE;
1525 	l2_tn_info->e_tag_fwd_en = FALSE;
1526 	l2_tn_info->e_tag_ether_type = DEFAULT_ETAG_ETYPE;
1527 
1528 	return 0;
1529 }
1530 /*
1531  * Negotiate mailbox API version with the PF.
1532  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1533  * Then we try to negotiate starting with the most recent one.
1534  * If all negotiation attempts fail, then we will proceed with
1535  * the default one (ixgbe_mbox_api_10).
1536  */
1537 static void
1538 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1539 {
1540 	int32_t i;
1541 
1542 	/* start with highest supported, proceed down */
1543 	static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1544 		ixgbe_mbox_api_12,
1545 		ixgbe_mbox_api_11,
1546 		ixgbe_mbox_api_10,
1547 	};
1548 
1549 	for (i = 0;
1550 			i != RTE_DIM(sup_ver) &&
1551 			ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1552 			i++)
1553 		;
1554 }
1555 
1556 static void
1557 generate_random_mac_addr(struct ether_addr *mac_addr)
1558 {
1559 	uint64_t random;
1560 
1561 	/* Set Organizationally Unique Identifier (OUI) prefix. */
1562 	mac_addr->addr_bytes[0] = 0x00;
1563 	mac_addr->addr_bytes[1] = 0x09;
1564 	mac_addr->addr_bytes[2] = 0xC0;
1565 	/* Force indication of locally assigned MAC address. */
1566 	mac_addr->addr_bytes[0] |= ETHER_LOCAL_ADMIN_ADDR;
1567 	/* Generate the last 3 bytes of the MAC address with a random number. */
1568 	random = rte_rand();
1569 	memcpy(&mac_addr->addr_bytes[3], &random, 3);
1570 }
1571 
1572 /*
1573  * Virtual Function device init
1574  */
1575 static int
1576 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1577 {
1578 	int diag;
1579 	uint32_t tc, tcs;
1580 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1581 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1582 	struct ixgbe_hw *hw =
1583 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1584 	struct ixgbe_vfta *shadow_vfta =
1585 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1586 	struct ixgbe_hwstrip *hwstrip =
1587 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1588 	struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
1589 
1590 	PMD_INIT_FUNC_TRACE();
1591 
1592 	eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1593 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1594 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1595 
1596 	/* for secondary processes, we don't initialise any further as primary
1597 	 * has already done this work. Only check we don't need a different
1598 	 * RX function
1599 	 */
1600 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1601 		struct ixgbe_tx_queue *txq;
1602 		/* TX queue function in primary, set by last queue initialized
1603 		 * Tx queue may not initialized by primary process
1604 		 */
1605 		if (eth_dev->data->tx_queues) {
1606 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1607 			ixgbe_set_tx_function(eth_dev, txq);
1608 		} else {
1609 			/* Use default TX function if we get here */
1610 			PMD_INIT_LOG(NOTICE,
1611 				     "No TX queues configured yet. Using default TX function.");
1612 		}
1613 
1614 		ixgbe_set_rx_function(eth_dev);
1615 
1616 		return 0;
1617 	}
1618 
1619 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1620 
1621 	hw->device_id = pci_dev->id.device_id;
1622 	hw->vendor_id = pci_dev->id.vendor_id;
1623 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1624 
1625 	/* initialize the vfta */
1626 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1627 
1628 	/* initialize the hw strip bitmap*/
1629 	memset(hwstrip, 0, sizeof(*hwstrip));
1630 
1631 	/* Initialize the shared code (base driver) */
1632 	diag = ixgbe_init_shared_code(hw);
1633 	if (diag != IXGBE_SUCCESS) {
1634 		PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1635 		return -EIO;
1636 	}
1637 
1638 	/* init_mailbox_params */
1639 	hw->mbx.ops.init_params(hw);
1640 
1641 	/* Reset the hw statistics */
1642 	ixgbevf_dev_stats_reset(eth_dev);
1643 
1644 	/* Disable the interrupts for VF */
1645 	ixgbevf_intr_disable(hw);
1646 
1647 	hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1648 	diag = hw->mac.ops.reset_hw(hw);
1649 
1650 	/*
1651 	 * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1652 	 * the underlying PF driver has not assigned a MAC address to the VF.
1653 	 * In this case, assign a random MAC address.
1654 	 */
1655 	if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1656 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1657 		return diag;
1658 	}
1659 
1660 	/* negotiate mailbox API version to use with the PF. */
1661 	ixgbevf_negotiate_api(hw);
1662 
1663 	/* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1664 	ixgbevf_get_queues(hw, &tcs, &tc);
1665 
1666 	/* Allocate memory for storing MAC addresses */
1667 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", ETHER_ADDR_LEN *
1668 					       hw->mac.num_rar_entries, 0);
1669 	if (eth_dev->data->mac_addrs == NULL) {
1670 		PMD_INIT_LOG(ERR,
1671 			     "Failed to allocate %u bytes needed to store "
1672 			     "MAC addresses",
1673 			     ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1674 		return -ENOMEM;
1675 	}
1676 
1677 	/* Generate a random MAC address, if none was assigned by PF. */
1678 	if (is_zero_ether_addr(perm_addr)) {
1679 		generate_random_mac_addr(perm_addr);
1680 		diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1681 		if (diag) {
1682 			rte_free(eth_dev->data->mac_addrs);
1683 			eth_dev->data->mac_addrs = NULL;
1684 			return diag;
1685 		}
1686 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1687 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1688 			     "%02x:%02x:%02x:%02x:%02x:%02x",
1689 			     perm_addr->addr_bytes[0],
1690 			     perm_addr->addr_bytes[1],
1691 			     perm_addr->addr_bytes[2],
1692 			     perm_addr->addr_bytes[3],
1693 			     perm_addr->addr_bytes[4],
1694 			     perm_addr->addr_bytes[5]);
1695 	}
1696 
1697 	/* Copy the permanent MAC address */
1698 	ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1699 
1700 	/* reset the hardware with the new settings */
1701 	diag = hw->mac.ops.start_hw(hw);
1702 	switch (diag) {
1703 	case  0:
1704 		break;
1705 
1706 	default:
1707 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1708 		return -EIO;
1709 	}
1710 
1711 	rte_intr_callback_register(intr_handle,
1712 				   ixgbevf_dev_interrupt_handler, eth_dev);
1713 	rte_intr_enable(intr_handle);
1714 	ixgbevf_intr_enable(hw);
1715 
1716 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1717 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1718 		     pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1719 
1720 	return 0;
1721 }
1722 
1723 /* Virtual Function device uninit */
1724 
1725 static int
1726 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1727 {
1728 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1729 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1730 	struct ixgbe_hw *hw;
1731 
1732 	PMD_INIT_FUNC_TRACE();
1733 
1734 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1735 		return -EPERM;
1736 
1737 	hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1738 
1739 	if (hw->adapter_stopped == 0)
1740 		ixgbevf_dev_close(eth_dev);
1741 
1742 	eth_dev->dev_ops = NULL;
1743 	eth_dev->rx_pkt_burst = NULL;
1744 	eth_dev->tx_pkt_burst = NULL;
1745 
1746 	/* Disable the interrupts for VF */
1747 	ixgbevf_intr_disable(hw);
1748 
1749 	rte_free(eth_dev->data->mac_addrs);
1750 	eth_dev->data->mac_addrs = NULL;
1751 
1752 	rte_intr_disable(intr_handle);
1753 	rte_intr_callback_unregister(intr_handle,
1754 				     ixgbevf_dev_interrupt_handler, eth_dev);
1755 
1756 	return 0;
1757 }
1758 
1759 static int eth_ixgbe_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1760 	struct rte_pci_device *pci_dev)
1761 {
1762 	return rte_eth_dev_pci_generic_probe(pci_dev,
1763 		sizeof(struct ixgbe_adapter), eth_ixgbe_dev_init);
1764 }
1765 
1766 static int eth_ixgbe_pci_remove(struct rte_pci_device *pci_dev)
1767 {
1768 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbe_dev_uninit);
1769 }
1770 
1771 static struct rte_pci_driver rte_ixgbe_pmd = {
1772 	.id_table = pci_id_ixgbe_map,
1773 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
1774 		     RTE_PCI_DRV_IOVA_AS_VA,
1775 	.probe = eth_ixgbe_pci_probe,
1776 	.remove = eth_ixgbe_pci_remove,
1777 };
1778 
1779 static int eth_ixgbevf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1780 	struct rte_pci_device *pci_dev)
1781 {
1782 	return rte_eth_dev_pci_generic_probe(pci_dev,
1783 		sizeof(struct ixgbe_adapter), eth_ixgbevf_dev_init);
1784 }
1785 
1786 static int eth_ixgbevf_pci_remove(struct rte_pci_device *pci_dev)
1787 {
1788 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbevf_dev_uninit);
1789 }
1790 
1791 /*
1792  * virtual function driver struct
1793  */
1794 static struct rte_pci_driver rte_ixgbevf_pmd = {
1795 	.id_table = pci_id_ixgbevf_map,
1796 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_IOVA_AS_VA,
1797 	.probe = eth_ixgbevf_pci_probe,
1798 	.remove = eth_ixgbevf_pci_remove,
1799 };
1800 
1801 static int
1802 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1803 {
1804 	struct ixgbe_hw *hw =
1805 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1806 	struct ixgbe_vfta *shadow_vfta =
1807 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1808 	uint32_t vfta;
1809 	uint32_t vid_idx;
1810 	uint32_t vid_bit;
1811 
1812 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1813 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1814 	vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1815 	if (on)
1816 		vfta |= vid_bit;
1817 	else
1818 		vfta &= ~vid_bit;
1819 	IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1820 
1821 	/* update local VFTA copy */
1822 	shadow_vfta->vfta[vid_idx] = vfta;
1823 
1824 	return 0;
1825 }
1826 
1827 static void
1828 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1829 {
1830 	if (on)
1831 		ixgbe_vlan_hw_strip_enable(dev, queue);
1832 	else
1833 		ixgbe_vlan_hw_strip_disable(dev, queue);
1834 }
1835 
1836 static int
1837 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1838 		    enum rte_vlan_type vlan_type,
1839 		    uint16_t tpid)
1840 {
1841 	struct ixgbe_hw *hw =
1842 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1843 	int ret = 0;
1844 	uint32_t reg;
1845 	uint32_t qinq;
1846 
1847 	qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1848 	qinq &= IXGBE_DMATXCTL_GDV;
1849 
1850 	switch (vlan_type) {
1851 	case ETH_VLAN_TYPE_INNER:
1852 		if (qinq) {
1853 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1854 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1855 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1856 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1857 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1858 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1859 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1860 		} else {
1861 			ret = -ENOTSUP;
1862 			PMD_DRV_LOG(ERR, "Inner type is not supported"
1863 				    " by single VLAN");
1864 		}
1865 		break;
1866 	case ETH_VLAN_TYPE_OUTER:
1867 		if (qinq) {
1868 			/* Only the high 16-bits is valid */
1869 			IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1870 					IXGBE_EXVET_VET_EXT_SHIFT);
1871 		} else {
1872 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1873 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1874 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1875 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1876 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1877 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1878 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1879 		}
1880 
1881 		break;
1882 	default:
1883 		ret = -EINVAL;
1884 		PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1885 		break;
1886 	}
1887 
1888 	return ret;
1889 }
1890 
1891 void
1892 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1893 {
1894 	struct ixgbe_hw *hw =
1895 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1896 	uint32_t vlnctrl;
1897 
1898 	PMD_INIT_FUNC_TRACE();
1899 
1900 	/* Filter Table Disable */
1901 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1902 	vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1903 
1904 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1905 }
1906 
1907 void
1908 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1909 {
1910 	struct ixgbe_hw *hw =
1911 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1912 	struct ixgbe_vfta *shadow_vfta =
1913 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1914 	uint32_t vlnctrl;
1915 	uint16_t i;
1916 
1917 	PMD_INIT_FUNC_TRACE();
1918 
1919 	/* Filter Table Enable */
1920 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1921 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1922 	vlnctrl |= IXGBE_VLNCTRL_VFE;
1923 
1924 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1925 
1926 	/* write whatever is in local vfta copy */
1927 	for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1928 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1929 }
1930 
1931 static void
1932 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1933 {
1934 	struct ixgbe_hwstrip *hwstrip =
1935 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1936 	struct ixgbe_rx_queue *rxq;
1937 
1938 	if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
1939 		return;
1940 
1941 	if (on)
1942 		IXGBE_SET_HWSTRIP(hwstrip, queue);
1943 	else
1944 		IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1945 
1946 	if (queue >= dev->data->nb_rx_queues)
1947 		return;
1948 
1949 	rxq = dev->data->rx_queues[queue];
1950 
1951 	if (on)
1952 		rxq->vlan_flags = PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
1953 	else
1954 		rxq->vlan_flags = PKT_RX_VLAN;
1955 }
1956 
1957 static void
1958 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
1959 {
1960 	struct ixgbe_hw *hw =
1961 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1962 	uint32_t ctrl;
1963 
1964 	PMD_INIT_FUNC_TRACE();
1965 
1966 	if (hw->mac.type == ixgbe_mac_82598EB) {
1967 		/* No queue level support */
1968 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1969 		return;
1970 	}
1971 
1972 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1973 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1974 	ctrl &= ~IXGBE_RXDCTL_VME;
1975 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1976 
1977 	/* record those setting for HW strip per queue */
1978 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
1979 }
1980 
1981 static void
1982 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
1983 {
1984 	struct ixgbe_hw *hw =
1985 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1986 	uint32_t ctrl;
1987 
1988 	PMD_INIT_FUNC_TRACE();
1989 
1990 	if (hw->mac.type == ixgbe_mac_82598EB) {
1991 		/* No queue level supported */
1992 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1993 		return;
1994 	}
1995 
1996 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1997 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1998 	ctrl |= IXGBE_RXDCTL_VME;
1999 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2000 
2001 	/* record those setting for HW strip per queue */
2002 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
2003 }
2004 
2005 void
2006 ixgbe_vlan_hw_strip_disable_all(struct rte_eth_dev *dev)
2007 {
2008 	struct ixgbe_hw *hw =
2009 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2010 	uint32_t ctrl;
2011 	uint16_t i;
2012 	struct ixgbe_rx_queue *rxq;
2013 
2014 	PMD_INIT_FUNC_TRACE();
2015 
2016 	if (hw->mac.type == ixgbe_mac_82598EB) {
2017 		ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2018 		ctrl &= ~IXGBE_VLNCTRL_VME;
2019 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2020 	} else {
2021 		/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2022 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
2023 			rxq = dev->data->rx_queues[i];
2024 			ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2025 			ctrl &= ~IXGBE_RXDCTL_VME;
2026 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2027 
2028 			/* record those setting for HW strip per queue */
2029 			ixgbe_vlan_hw_strip_bitmap_set(dev, i, 0);
2030 		}
2031 	}
2032 }
2033 
2034 void
2035 ixgbe_vlan_hw_strip_enable_all(struct rte_eth_dev *dev)
2036 {
2037 	struct ixgbe_hw *hw =
2038 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2039 	uint32_t ctrl;
2040 	uint16_t i;
2041 	struct ixgbe_rx_queue *rxq;
2042 
2043 	PMD_INIT_FUNC_TRACE();
2044 
2045 	if (hw->mac.type == ixgbe_mac_82598EB) {
2046 		ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2047 		ctrl |= IXGBE_VLNCTRL_VME;
2048 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2049 	} else {
2050 		/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2051 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
2052 			rxq = dev->data->rx_queues[i];
2053 			ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2054 			ctrl |= IXGBE_RXDCTL_VME;
2055 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2056 
2057 			/* record those setting for HW strip per queue */
2058 			ixgbe_vlan_hw_strip_bitmap_set(dev, i, 1);
2059 		}
2060 	}
2061 }
2062 
2063 static void
2064 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2065 {
2066 	struct ixgbe_hw *hw =
2067 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2068 	uint32_t ctrl;
2069 
2070 	PMD_INIT_FUNC_TRACE();
2071 
2072 	/* DMATXCTRL: Geric Double VLAN Disable */
2073 	ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2074 	ctrl &= ~IXGBE_DMATXCTL_GDV;
2075 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2076 
2077 	/* CTRL_EXT: Global Double VLAN Disable */
2078 	ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2079 	ctrl &= ~IXGBE_EXTENDED_VLAN;
2080 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2081 
2082 }
2083 
2084 static void
2085 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2086 {
2087 	struct ixgbe_hw *hw =
2088 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2089 	uint32_t ctrl;
2090 
2091 	PMD_INIT_FUNC_TRACE();
2092 
2093 	/* DMATXCTRL: Geric Double VLAN Enable */
2094 	ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2095 	ctrl |= IXGBE_DMATXCTL_GDV;
2096 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2097 
2098 	/* CTRL_EXT: Global Double VLAN Enable */
2099 	ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2100 	ctrl |= IXGBE_EXTENDED_VLAN;
2101 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2102 
2103 	/* Clear pooling mode of PFVTCTL. It's required by X550. */
2104 	if (hw->mac.type == ixgbe_mac_X550 ||
2105 	    hw->mac.type == ixgbe_mac_X550EM_x ||
2106 	    hw->mac.type == ixgbe_mac_X550EM_a) {
2107 		ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
2108 		ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
2109 		IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
2110 	}
2111 
2112 	/*
2113 	 * VET EXT field in the EXVET register = 0x8100 by default
2114 	 * So no need to change. Same to VT field of DMATXCTL register
2115 	 */
2116 }
2117 
2118 static int
2119 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2120 {
2121 	if (mask & ETH_VLAN_STRIP_MASK) {
2122 		if (dev->data->dev_conf.rxmode.hw_vlan_strip)
2123 			ixgbe_vlan_hw_strip_enable_all(dev);
2124 		else
2125 			ixgbe_vlan_hw_strip_disable_all(dev);
2126 	}
2127 
2128 	if (mask & ETH_VLAN_FILTER_MASK) {
2129 		if (dev->data->dev_conf.rxmode.hw_vlan_filter)
2130 			ixgbe_vlan_hw_filter_enable(dev);
2131 		else
2132 			ixgbe_vlan_hw_filter_disable(dev);
2133 	}
2134 
2135 	if (mask & ETH_VLAN_EXTEND_MASK) {
2136 		if (dev->data->dev_conf.rxmode.hw_vlan_extend)
2137 			ixgbe_vlan_hw_extend_enable(dev);
2138 		else
2139 			ixgbe_vlan_hw_extend_disable(dev);
2140 	}
2141 
2142 	return 0;
2143 }
2144 
2145 static void
2146 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2147 {
2148 	struct ixgbe_hw *hw =
2149 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2150 	/* VLNCTRL: enable vlan filtering and allow all vlan tags through */
2151 	uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2152 
2153 	vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
2154 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
2155 }
2156 
2157 static int
2158 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
2159 {
2160 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2161 
2162 	switch (nb_rx_q) {
2163 	case 1:
2164 	case 2:
2165 		RTE_ETH_DEV_SRIOV(dev).active = ETH_64_POOLS;
2166 		break;
2167 	case 4:
2168 		RTE_ETH_DEV_SRIOV(dev).active = ETH_32_POOLS;
2169 		break;
2170 	default:
2171 		return -EINVAL;
2172 	}
2173 
2174 	RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
2175 		IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2176 	RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
2177 		pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2178 	return 0;
2179 }
2180 
2181 static int
2182 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
2183 {
2184 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
2185 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2186 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
2187 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
2188 
2189 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
2190 		/* check multi-queue mode */
2191 		switch (dev_conf->rxmode.mq_mode) {
2192 		case ETH_MQ_RX_VMDQ_DCB:
2193 			PMD_INIT_LOG(INFO, "ETH_MQ_RX_VMDQ_DCB mode supported in SRIOV");
2194 			break;
2195 		case ETH_MQ_RX_VMDQ_DCB_RSS:
2196 			/* DCB/RSS VMDQ in SRIOV mode, not implement yet */
2197 			PMD_INIT_LOG(ERR, "SRIOV active,"
2198 					" unsupported mq_mode rx %d.",
2199 					dev_conf->rxmode.mq_mode);
2200 			return -EINVAL;
2201 		case ETH_MQ_RX_RSS:
2202 		case ETH_MQ_RX_VMDQ_RSS:
2203 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_RSS;
2204 			if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
2205 				if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
2206 					PMD_INIT_LOG(ERR, "SRIOV is active,"
2207 						" invalid queue number"
2208 						" for VMDQ RSS, allowed"
2209 						" value are 1, 2 or 4.");
2210 					return -EINVAL;
2211 				}
2212 			break;
2213 		case ETH_MQ_RX_VMDQ_ONLY:
2214 		case ETH_MQ_RX_NONE:
2215 			/* if nothing mq mode configure, use default scheme */
2216 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
2217 			break;
2218 		default: /* ETH_MQ_RX_DCB, ETH_MQ_RX_DCB_RSS or ETH_MQ_TX_DCB*/
2219 			/* SRIOV only works in VMDq enable mode */
2220 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2221 					" wrong mq_mode rx %d.",
2222 					dev_conf->rxmode.mq_mode);
2223 			return -EINVAL;
2224 		}
2225 
2226 		switch (dev_conf->txmode.mq_mode) {
2227 		case ETH_MQ_TX_VMDQ_DCB:
2228 			PMD_INIT_LOG(INFO, "ETH_MQ_TX_VMDQ_DCB mode supported in SRIOV");
2229 			dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_DCB;
2230 			break;
2231 		default: /* ETH_MQ_TX_VMDQ_ONLY or ETH_MQ_TX_NONE */
2232 			dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_ONLY;
2233 			break;
2234 		}
2235 
2236 		/* check valid queue number */
2237 		if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2238 		    (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2239 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2240 					" nb_rx_q=%d nb_tx_q=%d queue number"
2241 					" must be less than or equal to %d.",
2242 					nb_rx_q, nb_tx_q,
2243 					RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2244 			return -EINVAL;
2245 		}
2246 	} else {
2247 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB_RSS) {
2248 			PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2249 					  " not supported.");
2250 			return -EINVAL;
2251 		}
2252 		/* check configuration for vmdb+dcb mode */
2253 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB) {
2254 			const struct rte_eth_vmdq_dcb_conf *conf;
2255 
2256 			if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2257 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2258 						IXGBE_VMDQ_DCB_NB_QUEUES);
2259 				return -EINVAL;
2260 			}
2261 			conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2262 			if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2263 			       conf->nb_queue_pools == ETH_32_POOLS)) {
2264 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2265 						" nb_queue_pools must be %d or %d.",
2266 						ETH_16_POOLS, ETH_32_POOLS);
2267 				return -EINVAL;
2268 			}
2269 		}
2270 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_VMDQ_DCB) {
2271 			const struct rte_eth_vmdq_dcb_tx_conf *conf;
2272 
2273 			if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2274 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2275 						 IXGBE_VMDQ_DCB_NB_QUEUES);
2276 				return -EINVAL;
2277 			}
2278 			conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2279 			if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2280 			       conf->nb_queue_pools == ETH_32_POOLS)) {
2281 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2282 						" nb_queue_pools != %d and"
2283 						" nb_queue_pools != %d.",
2284 						ETH_16_POOLS, ETH_32_POOLS);
2285 				return -EINVAL;
2286 			}
2287 		}
2288 
2289 		/* For DCB mode check our configuration before we go further */
2290 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB) {
2291 			const struct rte_eth_dcb_rx_conf *conf;
2292 
2293 			if (nb_rx_q != IXGBE_DCB_NB_QUEUES) {
2294 				PMD_INIT_LOG(ERR, "DCB selected, nb_rx_q != %d.",
2295 						 IXGBE_DCB_NB_QUEUES);
2296 				return -EINVAL;
2297 			}
2298 			conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2299 			if (!(conf->nb_tcs == ETH_4_TCS ||
2300 			       conf->nb_tcs == ETH_8_TCS)) {
2301 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2302 						" and nb_tcs != %d.",
2303 						ETH_4_TCS, ETH_8_TCS);
2304 				return -EINVAL;
2305 			}
2306 		}
2307 
2308 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_DCB) {
2309 			const struct rte_eth_dcb_tx_conf *conf;
2310 
2311 			if (nb_tx_q != IXGBE_DCB_NB_QUEUES) {
2312 				PMD_INIT_LOG(ERR, "DCB, nb_tx_q != %d.",
2313 						 IXGBE_DCB_NB_QUEUES);
2314 				return -EINVAL;
2315 			}
2316 			conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2317 			if (!(conf->nb_tcs == ETH_4_TCS ||
2318 			       conf->nb_tcs == ETH_8_TCS)) {
2319 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2320 						" and nb_tcs != %d.",
2321 						ETH_4_TCS, ETH_8_TCS);
2322 				return -EINVAL;
2323 			}
2324 		}
2325 
2326 		/*
2327 		 * When DCB/VT is off, maximum number of queues changes,
2328 		 * except for 82598EB, which remains constant.
2329 		 */
2330 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
2331 				hw->mac.type != ixgbe_mac_82598EB) {
2332 			if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2333 				PMD_INIT_LOG(ERR,
2334 					     "Neither VT nor DCB are enabled, "
2335 					     "nb_tx_q > %d.",
2336 					     IXGBE_NONE_MODE_TX_NB_QUEUES);
2337 				return -EINVAL;
2338 			}
2339 		}
2340 	}
2341 	return 0;
2342 }
2343 
2344 static int
2345 ixgbe_dev_configure(struct rte_eth_dev *dev)
2346 {
2347 	struct ixgbe_interrupt *intr =
2348 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2349 	struct ixgbe_adapter *adapter =
2350 		(struct ixgbe_adapter *)dev->data->dev_private;
2351 	int ret;
2352 
2353 	PMD_INIT_FUNC_TRACE();
2354 	/* multipe queue mode checking */
2355 	ret  = ixgbe_check_mq_mode(dev);
2356 	if (ret != 0) {
2357 		PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2358 			    ret);
2359 		return ret;
2360 	}
2361 
2362 	/* set flag to update link status after init */
2363 	intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2364 
2365 	/*
2366 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2367 	 * allocation or vector Rx preconditions we will reset it.
2368 	 */
2369 	adapter->rx_bulk_alloc_allowed = true;
2370 	adapter->rx_vec_allowed = true;
2371 
2372 	return 0;
2373 }
2374 
2375 static void
2376 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2377 {
2378 	struct ixgbe_hw *hw =
2379 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2380 	struct ixgbe_interrupt *intr =
2381 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2382 	uint32_t gpie;
2383 
2384 	/* only set up it on X550EM_X */
2385 	if (hw->mac.type == ixgbe_mac_X550EM_x) {
2386 		gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2387 		gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2388 		IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2389 		if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2390 			intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2391 	}
2392 }
2393 
2394 int
2395 ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
2396 			uint16_t tx_rate, uint64_t q_msk)
2397 {
2398 	struct ixgbe_hw *hw;
2399 	struct ixgbe_vf_info *vfinfo;
2400 	struct rte_eth_link link;
2401 	uint8_t  nb_q_per_pool;
2402 	uint32_t queue_stride;
2403 	uint32_t queue_idx, idx = 0, vf_idx;
2404 	uint32_t queue_end;
2405 	uint16_t total_rate = 0;
2406 	struct rte_pci_device *pci_dev;
2407 
2408 	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2409 	rte_eth_link_get_nowait(dev->data->port_id, &link);
2410 
2411 	if (vf >= pci_dev->max_vfs)
2412 		return -EINVAL;
2413 
2414 	if (tx_rate > link.link_speed)
2415 		return -EINVAL;
2416 
2417 	if (q_msk == 0)
2418 		return 0;
2419 
2420 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2421 	vfinfo = *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
2422 	nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2423 	queue_stride = IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2424 	queue_idx = vf * queue_stride;
2425 	queue_end = queue_idx + nb_q_per_pool - 1;
2426 	if (queue_end >= hw->mac.max_tx_queues)
2427 		return -EINVAL;
2428 
2429 	if (vfinfo) {
2430 		for (vf_idx = 0; vf_idx < pci_dev->max_vfs; vf_idx++) {
2431 			if (vf_idx == vf)
2432 				continue;
2433 			for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
2434 				idx++)
2435 				total_rate += vfinfo[vf_idx].tx_rate[idx];
2436 		}
2437 	} else {
2438 		return -EINVAL;
2439 	}
2440 
2441 	/* Store tx_rate for this vf. */
2442 	for (idx = 0; idx < nb_q_per_pool; idx++) {
2443 		if (((uint64_t)0x1 << idx) & q_msk) {
2444 			if (vfinfo[vf].tx_rate[idx] != tx_rate)
2445 				vfinfo[vf].tx_rate[idx] = tx_rate;
2446 			total_rate += tx_rate;
2447 		}
2448 	}
2449 
2450 	if (total_rate > dev->data->dev_link.link_speed) {
2451 		/* Reset stored TX rate of the VF if it causes exceed
2452 		 * link speed.
2453 		 */
2454 		memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
2455 		return -EINVAL;
2456 	}
2457 
2458 	/* Set RTTBCNRC of each queue/pool for vf X  */
2459 	for (; queue_idx <= queue_end; queue_idx++) {
2460 		if (0x1 & q_msk)
2461 			ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
2462 		q_msk = q_msk >> 1;
2463 	}
2464 
2465 	return 0;
2466 }
2467 
2468 /*
2469  * Configure device link speed and setup link.
2470  * It returns 0 on success.
2471  */
2472 static int
2473 ixgbe_dev_start(struct rte_eth_dev *dev)
2474 {
2475 	struct ixgbe_hw *hw =
2476 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2477 	struct ixgbe_vf_info *vfinfo =
2478 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2479 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2480 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2481 	uint32_t intr_vector = 0;
2482 	int err, link_up = 0, negotiate = 0;
2483 	uint32_t speed = 0;
2484 	int mask = 0;
2485 	int status;
2486 	uint16_t vf, idx;
2487 	uint32_t *link_speeds;
2488 	struct ixgbe_tm_conf *tm_conf =
2489 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2490 
2491 	PMD_INIT_FUNC_TRACE();
2492 
2493 	/* IXGBE devices don't support:
2494 	*    - half duplex (checked afterwards for valid speeds)
2495 	*    - fixed speed: TODO implement
2496 	*/
2497 	if (dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED) {
2498 		PMD_INIT_LOG(ERR,
2499 		"Invalid link_speeds for port %u, fix speed not supported",
2500 				dev->data->port_id);
2501 		return -EINVAL;
2502 	}
2503 
2504 	/* disable uio/vfio intr/eventfd mapping */
2505 	rte_intr_disable(intr_handle);
2506 
2507 	/* stop adapter */
2508 	hw->adapter_stopped = 0;
2509 	ixgbe_stop_adapter(hw);
2510 
2511 	/* reinitialize adapter
2512 	 * this calls reset and start
2513 	 */
2514 	status = ixgbe_pf_reset_hw(hw);
2515 	if (status != 0)
2516 		return -1;
2517 	hw->mac.ops.start_hw(hw);
2518 	hw->mac.get_link_status = true;
2519 
2520 	/* configure PF module if SRIOV enabled */
2521 	ixgbe_pf_host_configure(dev);
2522 
2523 	ixgbe_dev_phy_intr_setup(dev);
2524 
2525 	/* check and configure queue intr-vector mapping */
2526 	if ((rte_intr_cap_multiple(intr_handle) ||
2527 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
2528 	    dev->data->dev_conf.intr_conf.rxq != 0) {
2529 		intr_vector = dev->data->nb_rx_queues;
2530 		if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2531 			PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2532 					IXGBE_MAX_INTR_QUEUE_NUM);
2533 			return -ENOTSUP;
2534 		}
2535 		if (rte_intr_efd_enable(intr_handle, intr_vector))
2536 			return -1;
2537 	}
2538 
2539 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
2540 		intr_handle->intr_vec =
2541 			rte_zmalloc("intr_vec",
2542 				    dev->data->nb_rx_queues * sizeof(int), 0);
2543 		if (intr_handle->intr_vec == NULL) {
2544 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2545 				     " intr_vec", dev->data->nb_rx_queues);
2546 			return -ENOMEM;
2547 		}
2548 	}
2549 
2550 	/* confiugre msix for sleep until rx interrupt */
2551 	ixgbe_configure_msix(dev);
2552 
2553 	/* initialize transmission unit */
2554 	ixgbe_dev_tx_init(dev);
2555 
2556 	/* This can fail when allocating mbufs for descriptor rings */
2557 	err = ixgbe_dev_rx_init(dev);
2558 	if (err) {
2559 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2560 		goto error;
2561 	}
2562 
2563 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
2564 		ETH_VLAN_EXTEND_MASK;
2565 	err = ixgbe_vlan_offload_set(dev, mask);
2566 	if (err) {
2567 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload");
2568 		goto error;
2569 	}
2570 
2571 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
2572 		/* Enable vlan filtering for VMDq */
2573 		ixgbe_vmdq_vlan_hw_filter_enable(dev);
2574 	}
2575 
2576 	/* Configure DCB hw */
2577 	ixgbe_configure_dcb(dev);
2578 
2579 	if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2580 		err = ixgbe_fdir_configure(dev);
2581 		if (err)
2582 			goto error;
2583 	}
2584 
2585 	/* Restore vf rate limit */
2586 	if (vfinfo != NULL) {
2587 		for (vf = 0; vf < pci_dev->max_vfs; vf++)
2588 			for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2589 				if (vfinfo[vf].tx_rate[idx] != 0)
2590 					ixgbe_set_vf_rate_limit(
2591 						dev, vf,
2592 						vfinfo[vf].tx_rate[idx],
2593 						1 << idx);
2594 	}
2595 
2596 	ixgbe_restore_statistics_mapping(dev);
2597 
2598 	err = ixgbe_dev_rxtx_start(dev);
2599 	if (err < 0) {
2600 		PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2601 		goto error;
2602 	}
2603 
2604 	/* Skip link setup if loopback mode is enabled for 82599. */
2605 	if (hw->mac.type == ixgbe_mac_82599EB &&
2606 			dev->data->dev_conf.lpbk_mode == IXGBE_LPBK_82599_TX_RX)
2607 		goto skip_link_setup;
2608 
2609 	if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2610 		err = hw->mac.ops.setup_sfp(hw);
2611 		if (err)
2612 			goto error;
2613 	}
2614 
2615 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2616 		/* Turn on the copper */
2617 		ixgbe_set_phy_power(hw, true);
2618 	} else {
2619 		/* Turn on the laser */
2620 		ixgbe_enable_tx_laser(hw);
2621 	}
2622 
2623 	err = ixgbe_check_link(hw, &speed, &link_up, 0);
2624 	if (err)
2625 		goto error;
2626 	dev->data->dev_link.link_status = link_up;
2627 
2628 	err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2629 	if (err)
2630 		goto error;
2631 
2632 	link_speeds = &dev->data->dev_conf.link_speeds;
2633 	if (*link_speeds & ~(ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2634 			ETH_LINK_SPEED_10G)) {
2635 		PMD_INIT_LOG(ERR, "Invalid link setting");
2636 		goto error;
2637 	}
2638 
2639 	speed = 0x0;
2640 	if (*link_speeds == ETH_LINK_SPEED_AUTONEG) {
2641 		switch (hw->mac.type) {
2642 		case ixgbe_mac_82598EB:
2643 			speed = IXGBE_LINK_SPEED_82598_AUTONEG;
2644 			break;
2645 		case ixgbe_mac_82599EB:
2646 		case ixgbe_mac_X540:
2647 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2648 			break;
2649 		case ixgbe_mac_X550:
2650 		case ixgbe_mac_X550EM_x:
2651 		case ixgbe_mac_X550EM_a:
2652 			speed = IXGBE_LINK_SPEED_X550_AUTONEG;
2653 			break;
2654 		default:
2655 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2656 		}
2657 	} else {
2658 		if (*link_speeds & ETH_LINK_SPEED_10G)
2659 			speed |= IXGBE_LINK_SPEED_10GB_FULL;
2660 		if (*link_speeds & ETH_LINK_SPEED_1G)
2661 			speed |= IXGBE_LINK_SPEED_1GB_FULL;
2662 		if (*link_speeds & ETH_LINK_SPEED_100M)
2663 			speed |= IXGBE_LINK_SPEED_100_FULL;
2664 	}
2665 
2666 	err = ixgbe_setup_link(hw, speed, link_up);
2667 	if (err)
2668 		goto error;
2669 
2670 skip_link_setup:
2671 
2672 	if (rte_intr_allow_others(intr_handle)) {
2673 		/* check if lsc interrupt is enabled */
2674 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2675 			ixgbe_dev_lsc_interrupt_setup(dev, TRUE);
2676 		else
2677 			ixgbe_dev_lsc_interrupt_setup(dev, FALSE);
2678 		ixgbe_dev_macsec_interrupt_setup(dev);
2679 	} else {
2680 		rte_intr_callback_unregister(intr_handle,
2681 					     ixgbe_dev_interrupt_handler, dev);
2682 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2683 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
2684 				     " no intr multiplex");
2685 	}
2686 
2687 	/* check if rxq interrupt is enabled */
2688 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2689 	    rte_intr_dp_is_en(intr_handle))
2690 		ixgbe_dev_rxq_interrupt_setup(dev);
2691 
2692 	/* enable uio/vfio intr/eventfd mapping */
2693 	rte_intr_enable(intr_handle);
2694 
2695 	/* resume enabled intr since hw reset */
2696 	ixgbe_enable_intr(dev);
2697 	ixgbe_l2_tunnel_conf(dev);
2698 	ixgbe_filter_restore(dev);
2699 
2700 	if (tm_conf->root && !tm_conf->committed)
2701 		PMD_DRV_LOG(WARNING,
2702 			    "please call hierarchy_commit() "
2703 			    "before starting the port");
2704 
2705 	return 0;
2706 
2707 error:
2708 	PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2709 	ixgbe_dev_clear_queues(dev);
2710 	return -EIO;
2711 }
2712 
2713 /*
2714  * Stop device: disable rx and tx functions to allow for reconfiguring.
2715  */
2716 static void
2717 ixgbe_dev_stop(struct rte_eth_dev *dev)
2718 {
2719 	struct rte_eth_link link;
2720 	struct ixgbe_hw *hw =
2721 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2722 	struct ixgbe_vf_info *vfinfo =
2723 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2724 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2725 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2726 	int vf;
2727 	struct ixgbe_tm_conf *tm_conf =
2728 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2729 
2730 	PMD_INIT_FUNC_TRACE();
2731 
2732 	/* disable interrupts */
2733 	ixgbe_disable_intr(hw);
2734 
2735 	/* reset the NIC */
2736 	ixgbe_pf_reset_hw(hw);
2737 	hw->adapter_stopped = 0;
2738 
2739 	/* stop adapter */
2740 	ixgbe_stop_adapter(hw);
2741 
2742 	for (vf = 0; vfinfo != NULL && vf < pci_dev->max_vfs; vf++)
2743 		vfinfo[vf].clear_to_send = false;
2744 
2745 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2746 		/* Turn off the copper */
2747 		ixgbe_set_phy_power(hw, false);
2748 	} else {
2749 		/* Turn off the laser */
2750 		ixgbe_disable_tx_laser(hw);
2751 	}
2752 
2753 	ixgbe_dev_clear_queues(dev);
2754 
2755 	/* Clear stored conf */
2756 	dev->data->scattered_rx = 0;
2757 	dev->data->lro = 0;
2758 
2759 	/* Clear recorded link status */
2760 	memset(&link, 0, sizeof(link));
2761 	rte_ixgbe_dev_atomic_write_link_status(dev, &link);
2762 
2763 	if (!rte_intr_allow_others(intr_handle))
2764 		/* resume to the default handler */
2765 		rte_intr_callback_register(intr_handle,
2766 					   ixgbe_dev_interrupt_handler,
2767 					   (void *)dev);
2768 
2769 	/* Clean datapath event and queue/vec mapping */
2770 	rte_intr_efd_disable(intr_handle);
2771 	if (intr_handle->intr_vec != NULL) {
2772 		rte_free(intr_handle->intr_vec);
2773 		intr_handle->intr_vec = NULL;
2774 	}
2775 
2776 	/* reset hierarchy commit */
2777 	tm_conf->committed = false;
2778 }
2779 
2780 /*
2781  * Set device link up: enable tx.
2782  */
2783 static int
2784 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2785 {
2786 	struct ixgbe_hw *hw =
2787 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2788 	if (hw->mac.type == ixgbe_mac_82599EB) {
2789 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2790 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2791 			/* Not suported in bypass mode */
2792 			PMD_INIT_LOG(ERR, "Set link up is not supported "
2793 				     "by device id 0x%x", hw->device_id);
2794 			return -ENOTSUP;
2795 		}
2796 #endif
2797 	}
2798 
2799 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2800 		/* Turn on the copper */
2801 		ixgbe_set_phy_power(hw, true);
2802 	} else {
2803 		/* Turn on the laser */
2804 		ixgbe_enable_tx_laser(hw);
2805 	}
2806 
2807 	return 0;
2808 }
2809 
2810 /*
2811  * Set device link down: disable tx.
2812  */
2813 static int
2814 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2815 {
2816 	struct ixgbe_hw *hw =
2817 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2818 	if (hw->mac.type == ixgbe_mac_82599EB) {
2819 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2820 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2821 			/* Not suported in bypass mode */
2822 			PMD_INIT_LOG(ERR, "Set link down is not supported "
2823 				     "by device id 0x%x", hw->device_id);
2824 			return -ENOTSUP;
2825 		}
2826 #endif
2827 	}
2828 
2829 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2830 		/* Turn off the copper */
2831 		ixgbe_set_phy_power(hw, false);
2832 	} else {
2833 		/* Turn off the laser */
2834 		ixgbe_disable_tx_laser(hw);
2835 	}
2836 
2837 	return 0;
2838 }
2839 
2840 /*
2841  * Reset and stop device.
2842  */
2843 static void
2844 ixgbe_dev_close(struct rte_eth_dev *dev)
2845 {
2846 	struct ixgbe_hw *hw =
2847 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2848 
2849 	PMD_INIT_FUNC_TRACE();
2850 
2851 	ixgbe_pf_reset_hw(hw);
2852 
2853 	ixgbe_dev_stop(dev);
2854 	hw->adapter_stopped = 1;
2855 
2856 	ixgbe_dev_free_queues(dev);
2857 
2858 	ixgbe_disable_pcie_master(hw);
2859 
2860 	/* reprogram the RAR[0] in case user changed it. */
2861 	ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
2862 }
2863 
2864 /*
2865  * Reset PF device.
2866  */
2867 static int
2868 ixgbe_dev_reset(struct rte_eth_dev *dev)
2869 {
2870 	int ret;
2871 
2872 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
2873 	 * its VF to make them align with it. The detailed notification
2874 	 * mechanism is PMD specific. As to ixgbe PF, it is rather complex.
2875 	 * To avoid unexpected behavior in VF, currently reset of PF with
2876 	 * SR-IOV activation is not supported. It might be supported later.
2877 	 */
2878 	if (dev->data->sriov.active)
2879 		return -ENOTSUP;
2880 
2881 	ret = eth_ixgbe_dev_uninit(dev);
2882 	if (ret)
2883 		return ret;
2884 
2885 	ret = eth_ixgbe_dev_init(dev);
2886 
2887 	return ret;
2888 }
2889 
2890 static void
2891 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
2892 			   struct ixgbe_hw_stats *hw_stats,
2893 			   struct ixgbe_macsec_stats *macsec_stats,
2894 			   uint64_t *total_missed_rx, uint64_t *total_qbrc,
2895 			   uint64_t *total_qprc, uint64_t *total_qprdc)
2896 {
2897 	uint32_t bprc, lxon, lxoff, total;
2898 	uint32_t delta_gprc = 0;
2899 	unsigned i;
2900 	/* Workaround for RX byte count not including CRC bytes when CRC
2901 	 * strip is enabled. CRC bytes are removed from counters when crc_strip
2902 	 * is disabled.
2903 	 */
2904 	int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
2905 			IXGBE_HLREG0_RXCRCSTRP);
2906 
2907 	hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
2908 	hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
2909 	hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
2910 	hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
2911 
2912 	for (i = 0; i < 8; i++) {
2913 		uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
2914 
2915 		/* global total per queue */
2916 		hw_stats->mpc[i] += mp;
2917 		/* Running comprehensive total for stats display */
2918 		*total_missed_rx += hw_stats->mpc[i];
2919 		if (hw->mac.type == ixgbe_mac_82598EB) {
2920 			hw_stats->rnbc[i] +=
2921 			    IXGBE_READ_REG(hw, IXGBE_RNBC(i));
2922 			hw_stats->pxonrxc[i] +=
2923 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
2924 			hw_stats->pxoffrxc[i] +=
2925 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
2926 		} else {
2927 			hw_stats->pxonrxc[i] +=
2928 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
2929 			hw_stats->pxoffrxc[i] +=
2930 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
2931 			hw_stats->pxon2offc[i] +=
2932 				IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
2933 		}
2934 		hw_stats->pxontxc[i] +=
2935 		    IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
2936 		hw_stats->pxofftxc[i] +=
2937 		    IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
2938 	}
2939 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
2940 		uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
2941 		uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
2942 		uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
2943 
2944 		delta_gprc += delta_qprc;
2945 
2946 		hw_stats->qprc[i] += delta_qprc;
2947 		hw_stats->qptc[i] += delta_qptc;
2948 
2949 		hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
2950 		hw_stats->qbrc[i] +=
2951 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
2952 		if (crc_strip == 0)
2953 			hw_stats->qbrc[i] -= delta_qprc * ETHER_CRC_LEN;
2954 
2955 		hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
2956 		hw_stats->qbtc[i] +=
2957 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
2958 
2959 		hw_stats->qprdc[i] += delta_qprdc;
2960 		*total_qprdc += hw_stats->qprdc[i];
2961 
2962 		*total_qprc += hw_stats->qprc[i];
2963 		*total_qbrc += hw_stats->qbrc[i];
2964 	}
2965 	hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
2966 	hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
2967 	hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
2968 
2969 	/*
2970 	 * An errata states that gprc actually counts good + missed packets:
2971 	 * Workaround to set gprc to summated queue packet receives
2972 	 */
2973 	hw_stats->gprc = *total_qprc;
2974 
2975 	if (hw->mac.type != ixgbe_mac_82598EB) {
2976 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
2977 		hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
2978 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
2979 		hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
2980 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
2981 		hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
2982 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
2983 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
2984 	} else {
2985 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
2986 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
2987 		/* 82598 only has a counter in the high register */
2988 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
2989 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
2990 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
2991 	}
2992 	uint64_t old_tpr = hw_stats->tpr;
2993 
2994 	hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
2995 	hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
2996 
2997 	if (crc_strip == 0)
2998 		hw_stats->gorc -= delta_gprc * ETHER_CRC_LEN;
2999 
3000 	uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
3001 	hw_stats->gptc += delta_gptc;
3002 	hw_stats->gotc -= delta_gptc * ETHER_CRC_LEN;
3003 	hw_stats->tor -= (hw_stats->tpr - old_tpr) * ETHER_CRC_LEN;
3004 
3005 	/*
3006 	 * Workaround: mprc hardware is incorrectly counting
3007 	 * broadcasts, so for now we subtract those.
3008 	 */
3009 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
3010 	hw_stats->bprc += bprc;
3011 	hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
3012 	if (hw->mac.type == ixgbe_mac_82598EB)
3013 		hw_stats->mprc -= bprc;
3014 
3015 	hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
3016 	hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
3017 	hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
3018 	hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
3019 	hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
3020 	hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
3021 
3022 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
3023 	hw_stats->lxontxc += lxon;
3024 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
3025 	hw_stats->lxofftxc += lxoff;
3026 	total = lxon + lxoff;
3027 
3028 	hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
3029 	hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
3030 	hw_stats->gptc -= total;
3031 	hw_stats->mptc -= total;
3032 	hw_stats->ptc64 -= total;
3033 	hw_stats->gotc -= total * ETHER_MIN_LEN;
3034 
3035 	hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3036 	hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3037 	hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3038 	hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
3039 	hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
3040 	hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
3041 	hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
3042 	hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3043 	hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3044 	hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3045 	hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3046 	hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
3047 	hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3048 	hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
3049 	hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
3050 	hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
3051 	/* Only read FCOE on 82599 */
3052 	if (hw->mac.type != ixgbe_mac_82598EB) {
3053 		hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
3054 		hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
3055 		hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
3056 		hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
3057 		hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
3058 	}
3059 
3060 	/* Flow Director Stats registers */
3061 	hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
3062 	hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
3063 
3064 	/* MACsec Stats registers */
3065 	macsec_stats->out_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECTXUT);
3066 	macsec_stats->out_pkts_encrypted +=
3067 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTE);
3068 	macsec_stats->out_pkts_protected +=
3069 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTP);
3070 	macsec_stats->out_octets_encrypted +=
3071 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTE);
3072 	macsec_stats->out_octets_protected +=
3073 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTP);
3074 	macsec_stats->in_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECRXUT);
3075 	macsec_stats->in_pkts_badtag += IXGBE_READ_REG(hw, IXGBE_LSECRXBAD);
3076 	macsec_stats->in_pkts_nosci += IXGBE_READ_REG(hw, IXGBE_LSECRXNOSCI);
3077 	macsec_stats->in_pkts_unknownsci +=
3078 		IXGBE_READ_REG(hw, IXGBE_LSECRXUNSCI);
3079 	macsec_stats->in_octets_decrypted +=
3080 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTD);
3081 	macsec_stats->in_octets_validated +=
3082 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTV);
3083 	macsec_stats->in_pkts_unchecked += IXGBE_READ_REG(hw, IXGBE_LSECRXUNCH);
3084 	macsec_stats->in_pkts_delayed += IXGBE_READ_REG(hw, IXGBE_LSECRXDELAY);
3085 	macsec_stats->in_pkts_late += IXGBE_READ_REG(hw, IXGBE_LSECRXLATE);
3086 	for (i = 0; i < 2; i++) {
3087 		macsec_stats->in_pkts_ok +=
3088 			IXGBE_READ_REG(hw, IXGBE_LSECRXOK(i));
3089 		macsec_stats->in_pkts_invalid +=
3090 			IXGBE_READ_REG(hw, IXGBE_LSECRXINV(i));
3091 		macsec_stats->in_pkts_notvalid +=
3092 			IXGBE_READ_REG(hw, IXGBE_LSECRXNV(i));
3093 	}
3094 	macsec_stats->in_pkts_unusedsa += IXGBE_READ_REG(hw, IXGBE_LSECRXUNSA);
3095 	macsec_stats->in_pkts_notusingsa +=
3096 		IXGBE_READ_REG(hw, IXGBE_LSECRXNUSA);
3097 }
3098 
3099 /*
3100  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
3101  */
3102 static int
3103 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3104 {
3105 	struct ixgbe_hw *hw =
3106 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3107 	struct ixgbe_hw_stats *hw_stats =
3108 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3109 	struct ixgbe_macsec_stats *macsec_stats =
3110 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3111 				dev->data->dev_private);
3112 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3113 	unsigned i;
3114 
3115 	total_missed_rx = 0;
3116 	total_qbrc = 0;
3117 	total_qprc = 0;
3118 	total_qprdc = 0;
3119 
3120 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3121 			&total_qbrc, &total_qprc, &total_qprdc);
3122 
3123 	if (stats == NULL)
3124 		return -EINVAL;
3125 
3126 	/* Fill out the rte_eth_stats statistics structure */
3127 	stats->ipackets = total_qprc;
3128 	stats->ibytes = total_qbrc;
3129 	stats->opackets = hw_stats->gptc;
3130 	stats->obytes = hw_stats->gotc;
3131 
3132 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3133 		stats->q_ipackets[i] = hw_stats->qprc[i];
3134 		stats->q_opackets[i] = hw_stats->qptc[i];
3135 		stats->q_ibytes[i] = hw_stats->qbrc[i];
3136 		stats->q_obytes[i] = hw_stats->qbtc[i];
3137 		stats->q_errors[i] = hw_stats->qprdc[i];
3138 	}
3139 
3140 	/* Rx Errors */
3141 	stats->imissed  = total_missed_rx;
3142 	stats->ierrors  = hw_stats->crcerrs +
3143 			  hw_stats->mspdc +
3144 			  hw_stats->rlec +
3145 			  hw_stats->ruc +
3146 			  hw_stats->roc +
3147 			  hw_stats->illerrc +
3148 			  hw_stats->errbc +
3149 			  hw_stats->rfc +
3150 			  hw_stats->fccrc +
3151 			  hw_stats->fclast;
3152 
3153 	/* Tx Errors */
3154 	stats->oerrors  = 0;
3155 	return 0;
3156 }
3157 
3158 static void
3159 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
3160 {
3161 	struct ixgbe_hw_stats *stats =
3162 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3163 
3164 	/* HW registers are cleared on read */
3165 	ixgbe_dev_stats_get(dev, NULL);
3166 
3167 	/* Reset software totals */
3168 	memset(stats, 0, sizeof(*stats));
3169 }
3170 
3171 /* This function calculates the number of xstats based on the current config */
3172 static unsigned
3173 ixgbe_xstats_calc_num(void) {
3174 	return IXGBE_NB_HW_STATS + IXGBE_NB_MACSEC_STATS +
3175 		(IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
3176 		(IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
3177 }
3178 
3179 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3180 	struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned int size)
3181 {
3182 	const unsigned cnt_stats = ixgbe_xstats_calc_num();
3183 	unsigned stat, i, count;
3184 
3185 	if (xstats_names != NULL) {
3186 		count = 0;
3187 
3188 		/* Note: limit >= cnt_stats checked upstream
3189 		 * in rte_eth_xstats_names()
3190 		 */
3191 
3192 		/* Extended stats from ixgbe_hw_stats */
3193 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3194 			snprintf(xstats_names[count].name,
3195 				sizeof(xstats_names[count].name),
3196 				"%s",
3197 				rte_ixgbe_stats_strings[i].name);
3198 			count++;
3199 		}
3200 
3201 		/* MACsec Stats */
3202 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3203 			snprintf(xstats_names[count].name,
3204 				sizeof(xstats_names[count].name),
3205 				"%s",
3206 				rte_ixgbe_macsec_strings[i].name);
3207 			count++;
3208 		}
3209 
3210 		/* RX Priority Stats */
3211 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3212 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3213 				snprintf(xstats_names[count].name,
3214 					sizeof(xstats_names[count].name),
3215 					"rx_priority%u_%s", i,
3216 					rte_ixgbe_rxq_strings[stat].name);
3217 				count++;
3218 			}
3219 		}
3220 
3221 		/* TX Priority Stats */
3222 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3223 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3224 				snprintf(xstats_names[count].name,
3225 					sizeof(xstats_names[count].name),
3226 					"tx_priority%u_%s", i,
3227 					rte_ixgbe_txq_strings[stat].name);
3228 				count++;
3229 			}
3230 		}
3231 	}
3232 	return cnt_stats;
3233 }
3234 
3235 static int ixgbe_dev_xstats_get_names_by_id(
3236 	struct rte_eth_dev *dev,
3237 	struct rte_eth_xstat_name *xstats_names,
3238 	const uint64_t *ids,
3239 	unsigned int limit)
3240 {
3241 	if (!ids) {
3242 		const unsigned int cnt_stats = ixgbe_xstats_calc_num();
3243 		unsigned int stat, i, count;
3244 
3245 		if (xstats_names != NULL) {
3246 			count = 0;
3247 
3248 			/* Note: limit >= cnt_stats checked upstream
3249 			 * in rte_eth_xstats_names()
3250 			 */
3251 
3252 			/* Extended stats from ixgbe_hw_stats */
3253 			for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3254 				snprintf(xstats_names[count].name,
3255 					sizeof(xstats_names[count].name),
3256 					"%s",
3257 					rte_ixgbe_stats_strings[i].name);
3258 				count++;
3259 			}
3260 
3261 			/* MACsec Stats */
3262 			for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3263 				snprintf(xstats_names[count].name,
3264 					sizeof(xstats_names[count].name),
3265 					"%s",
3266 					rte_ixgbe_macsec_strings[i].name);
3267 				count++;
3268 			}
3269 
3270 			/* RX Priority Stats */
3271 			for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3272 				for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3273 					snprintf(xstats_names[count].name,
3274 					    sizeof(xstats_names[count].name),
3275 					    "rx_priority%u_%s", i,
3276 					    rte_ixgbe_rxq_strings[stat].name);
3277 					count++;
3278 				}
3279 			}
3280 
3281 			/* TX Priority Stats */
3282 			for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3283 				for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3284 					snprintf(xstats_names[count].name,
3285 					    sizeof(xstats_names[count].name),
3286 					    "tx_priority%u_%s", i,
3287 					    rte_ixgbe_txq_strings[stat].name);
3288 					count++;
3289 				}
3290 			}
3291 		}
3292 		return cnt_stats;
3293 	}
3294 
3295 	uint16_t i;
3296 	uint16_t size = ixgbe_xstats_calc_num();
3297 	struct rte_eth_xstat_name xstats_names_copy[size];
3298 
3299 	ixgbe_dev_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
3300 			size);
3301 
3302 	for (i = 0; i < limit; i++) {
3303 		if (ids[i] >= size) {
3304 			PMD_INIT_LOG(ERR, "id value isn't valid");
3305 			return -1;
3306 		}
3307 		strcpy(xstats_names[i].name,
3308 				xstats_names_copy[ids[i]].name);
3309 	}
3310 	return limit;
3311 }
3312 
3313 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3314 	struct rte_eth_xstat_name *xstats_names, unsigned limit)
3315 {
3316 	unsigned i;
3317 
3318 	if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
3319 		return -ENOMEM;
3320 
3321 	if (xstats_names != NULL)
3322 		for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
3323 			snprintf(xstats_names[i].name,
3324 				sizeof(xstats_names[i].name),
3325 				"%s", rte_ixgbevf_stats_strings[i].name);
3326 	return IXGBEVF_NB_XSTATS;
3327 }
3328 
3329 static int
3330 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3331 					 unsigned n)
3332 {
3333 	struct ixgbe_hw *hw =
3334 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3335 	struct ixgbe_hw_stats *hw_stats =
3336 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3337 	struct ixgbe_macsec_stats *macsec_stats =
3338 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3339 				dev->data->dev_private);
3340 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3341 	unsigned i, stat, count = 0;
3342 
3343 	count = ixgbe_xstats_calc_num();
3344 
3345 	if (n < count)
3346 		return count;
3347 
3348 	total_missed_rx = 0;
3349 	total_qbrc = 0;
3350 	total_qprc = 0;
3351 	total_qprdc = 0;
3352 
3353 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3354 			&total_qbrc, &total_qprc, &total_qprdc);
3355 
3356 	/* If this is a reset xstats is NULL, and we have cleared the
3357 	 * registers by reading them.
3358 	 */
3359 	if (!xstats)
3360 		return 0;
3361 
3362 	/* Extended stats from ixgbe_hw_stats */
3363 	count = 0;
3364 	for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3365 		xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3366 				rte_ixgbe_stats_strings[i].offset);
3367 		xstats[count].id = count;
3368 		count++;
3369 	}
3370 
3371 	/* MACsec Stats */
3372 	for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3373 		xstats[count].value = *(uint64_t *)(((char *)macsec_stats) +
3374 				rte_ixgbe_macsec_strings[i].offset);
3375 		xstats[count].id = count;
3376 		count++;
3377 	}
3378 
3379 	/* RX Priority Stats */
3380 	for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3381 		for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3382 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3383 					rte_ixgbe_rxq_strings[stat].offset +
3384 					(sizeof(uint64_t) * i));
3385 			xstats[count].id = count;
3386 			count++;
3387 		}
3388 	}
3389 
3390 	/* TX Priority Stats */
3391 	for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3392 		for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3393 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3394 					rte_ixgbe_txq_strings[stat].offset +
3395 					(sizeof(uint64_t) * i));
3396 			xstats[count].id = count;
3397 			count++;
3398 		}
3399 	}
3400 	return count;
3401 }
3402 
3403 static int
3404 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
3405 		uint64_t *values, unsigned int n)
3406 {
3407 	if (!ids) {
3408 		struct ixgbe_hw *hw =
3409 				IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3410 		struct ixgbe_hw_stats *hw_stats =
3411 				IXGBE_DEV_PRIVATE_TO_STATS(
3412 						dev->data->dev_private);
3413 		struct ixgbe_macsec_stats *macsec_stats =
3414 				IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3415 					dev->data->dev_private);
3416 		uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3417 		unsigned int i, stat, count = 0;
3418 
3419 		count = ixgbe_xstats_calc_num();
3420 
3421 		if (!ids && n < count)
3422 			return count;
3423 
3424 		total_missed_rx = 0;
3425 		total_qbrc = 0;
3426 		total_qprc = 0;
3427 		total_qprdc = 0;
3428 
3429 		ixgbe_read_stats_registers(hw, hw_stats, macsec_stats,
3430 				&total_missed_rx, &total_qbrc, &total_qprc,
3431 				&total_qprdc);
3432 
3433 		/* If this is a reset xstats is NULL, and we have cleared the
3434 		 * registers by reading them.
3435 		 */
3436 		if (!ids && !values)
3437 			return 0;
3438 
3439 		/* Extended stats from ixgbe_hw_stats */
3440 		count = 0;
3441 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3442 			values[count] = *(uint64_t *)(((char *)hw_stats) +
3443 					rte_ixgbe_stats_strings[i].offset);
3444 			count++;
3445 		}
3446 
3447 		/* MACsec Stats */
3448 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3449 			values[count] = *(uint64_t *)(((char *)macsec_stats) +
3450 					rte_ixgbe_macsec_strings[i].offset);
3451 			count++;
3452 		}
3453 
3454 		/* RX Priority Stats */
3455 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3456 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3457 				values[count] =
3458 					*(uint64_t *)(((char *)hw_stats) +
3459 					rte_ixgbe_rxq_strings[stat].offset +
3460 					(sizeof(uint64_t) * i));
3461 				count++;
3462 			}
3463 		}
3464 
3465 		/* TX Priority Stats */
3466 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3467 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3468 				values[count] =
3469 					*(uint64_t *)(((char *)hw_stats) +
3470 					rte_ixgbe_txq_strings[stat].offset +
3471 					(sizeof(uint64_t) * i));
3472 				count++;
3473 			}
3474 		}
3475 		return count;
3476 	}
3477 
3478 	uint16_t i;
3479 	uint16_t size = ixgbe_xstats_calc_num();
3480 	uint64_t values_copy[size];
3481 
3482 	ixgbe_dev_xstats_get_by_id(dev, NULL, values_copy, size);
3483 
3484 	for (i = 0; i < n; i++) {
3485 		if (ids[i] >= size) {
3486 			PMD_INIT_LOG(ERR, "id value isn't valid");
3487 			return -1;
3488 		}
3489 		values[i] = values_copy[ids[i]];
3490 	}
3491 	return n;
3492 }
3493 
3494 static void
3495 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
3496 {
3497 	struct ixgbe_hw_stats *stats =
3498 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3499 	struct ixgbe_macsec_stats *macsec_stats =
3500 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3501 				dev->data->dev_private);
3502 
3503 	unsigned count = ixgbe_xstats_calc_num();
3504 
3505 	/* HW registers are cleared on read */
3506 	ixgbe_dev_xstats_get(dev, NULL, count);
3507 
3508 	/* Reset software totals */
3509 	memset(stats, 0, sizeof(*stats));
3510 	memset(macsec_stats, 0, sizeof(*macsec_stats));
3511 }
3512 
3513 static void
3514 ixgbevf_update_stats(struct rte_eth_dev *dev)
3515 {
3516 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3517 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3518 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3519 
3520 	/* Good Rx packet, include VF loopback */
3521 	UPDATE_VF_STAT(IXGBE_VFGPRC,
3522 	    hw_stats->last_vfgprc, hw_stats->vfgprc);
3523 
3524 	/* Good Rx octets, include VF loopback */
3525 	UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
3526 	    hw_stats->last_vfgorc, hw_stats->vfgorc);
3527 
3528 	/* Good Tx packet, include VF loopback */
3529 	UPDATE_VF_STAT(IXGBE_VFGPTC,
3530 	    hw_stats->last_vfgptc, hw_stats->vfgptc);
3531 
3532 	/* Good Tx octets, include VF loopback */
3533 	UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
3534 	    hw_stats->last_vfgotc, hw_stats->vfgotc);
3535 
3536 	/* Rx Multicst Packet */
3537 	UPDATE_VF_STAT(IXGBE_VFMPRC,
3538 	    hw_stats->last_vfmprc, hw_stats->vfmprc);
3539 }
3540 
3541 static int
3542 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3543 		       unsigned n)
3544 {
3545 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3546 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3547 	unsigned i;
3548 
3549 	if (n < IXGBEVF_NB_XSTATS)
3550 		return IXGBEVF_NB_XSTATS;
3551 
3552 	ixgbevf_update_stats(dev);
3553 
3554 	if (!xstats)
3555 		return 0;
3556 
3557 	/* Extended stats */
3558 	for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
3559 		xstats[i].id = i;
3560 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
3561 			rte_ixgbevf_stats_strings[i].offset);
3562 	}
3563 
3564 	return IXGBEVF_NB_XSTATS;
3565 }
3566 
3567 static int
3568 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3569 {
3570 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3571 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3572 
3573 	ixgbevf_update_stats(dev);
3574 
3575 	if (stats == NULL)
3576 		return -EINVAL;
3577 
3578 	stats->ipackets = hw_stats->vfgprc;
3579 	stats->ibytes = hw_stats->vfgorc;
3580 	stats->opackets = hw_stats->vfgptc;
3581 	stats->obytes = hw_stats->vfgotc;
3582 	return 0;
3583 }
3584 
3585 static void
3586 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3587 {
3588 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3589 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3590 
3591 	/* Sync HW register to the last stats */
3592 	ixgbevf_dev_stats_get(dev, NULL);
3593 
3594 	/* reset HW current stats*/
3595 	hw_stats->vfgprc = 0;
3596 	hw_stats->vfgorc = 0;
3597 	hw_stats->vfgptc = 0;
3598 	hw_stats->vfgotc = 0;
3599 }
3600 
3601 static int
3602 ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
3603 {
3604 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3605 	u16 eeprom_verh, eeprom_verl;
3606 	u32 etrack_id;
3607 	int ret;
3608 
3609 	ixgbe_read_eeprom(hw, 0x2e, &eeprom_verh);
3610 	ixgbe_read_eeprom(hw, 0x2d, &eeprom_verl);
3611 
3612 	etrack_id = (eeprom_verh << 16) | eeprom_verl;
3613 	ret = snprintf(fw_version, fw_size, "0x%08x", etrack_id);
3614 
3615 	ret += 1; /* add the size of '\0' */
3616 	if (fw_size < (u32)ret)
3617 		return ret;
3618 	else
3619 		return 0;
3620 }
3621 
3622 static void
3623 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3624 {
3625 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3626 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3627 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3628 
3629 	dev_info->pci_dev = pci_dev;
3630 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3631 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3632 	if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3633 		/*
3634 		 * When DCB/VT is off, maximum number of queues changes,
3635 		 * except for 82598EB, which remains constant.
3636 		 */
3637 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
3638 				hw->mac.type != ixgbe_mac_82598EB)
3639 			dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3640 	}
3641 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3642 	dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3643 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3644 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3645 	dev_info->max_vfs = pci_dev->max_vfs;
3646 	if (hw->mac.type == ixgbe_mac_82598EB)
3647 		dev_info->max_vmdq_pools = ETH_16_POOLS;
3648 	else
3649 		dev_info->max_vmdq_pools = ETH_64_POOLS;
3650 	dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3651 	dev_info->rx_offload_capa =
3652 		DEV_RX_OFFLOAD_VLAN_STRIP |
3653 		DEV_RX_OFFLOAD_IPV4_CKSUM |
3654 		DEV_RX_OFFLOAD_UDP_CKSUM  |
3655 		DEV_RX_OFFLOAD_TCP_CKSUM;
3656 
3657 	/*
3658 	 * RSC is only supported by 82599 and x540 PF devices in a non-SR-IOV
3659 	 * mode.
3660 	 */
3661 	if ((hw->mac.type == ixgbe_mac_82599EB ||
3662 	     hw->mac.type == ixgbe_mac_X540) &&
3663 	    !RTE_ETH_DEV_SRIOV(dev).active)
3664 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TCP_LRO;
3665 
3666 	if (hw->mac.type == ixgbe_mac_82599EB ||
3667 	    hw->mac.type == ixgbe_mac_X540)
3668 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_MACSEC_STRIP;
3669 
3670 	if (hw->mac.type == ixgbe_mac_X550 ||
3671 	    hw->mac.type == ixgbe_mac_X550EM_x ||
3672 	    hw->mac.type == ixgbe_mac_X550EM_a)
3673 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM;
3674 
3675 	dev_info->tx_offload_capa =
3676 		DEV_TX_OFFLOAD_VLAN_INSERT |
3677 		DEV_TX_OFFLOAD_IPV4_CKSUM  |
3678 		DEV_TX_OFFLOAD_UDP_CKSUM   |
3679 		DEV_TX_OFFLOAD_TCP_CKSUM   |
3680 		DEV_TX_OFFLOAD_SCTP_CKSUM  |
3681 		DEV_TX_OFFLOAD_TCP_TSO;
3682 
3683 	if (hw->mac.type == ixgbe_mac_82599EB ||
3684 	    hw->mac.type == ixgbe_mac_X540)
3685 		dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_MACSEC_INSERT;
3686 
3687 	if (hw->mac.type == ixgbe_mac_X550 ||
3688 	    hw->mac.type == ixgbe_mac_X550EM_x ||
3689 	    hw->mac.type == ixgbe_mac_X550EM_a)
3690 		dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM;
3691 
3692 #ifdef RTE_LIBRTE_SECURITY
3693 	dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_SECURITY;
3694 	dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_SECURITY;
3695 #endif
3696 
3697 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3698 		.rx_thresh = {
3699 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3700 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3701 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3702 		},
3703 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3704 		.rx_drop_en = 0,
3705 	};
3706 
3707 	dev_info->default_txconf = (struct rte_eth_txconf) {
3708 		.tx_thresh = {
3709 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3710 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3711 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3712 		},
3713 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3714 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3715 		.txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
3716 				ETH_TXQ_FLAGS_NOOFFLOADS,
3717 	};
3718 
3719 	dev_info->rx_desc_lim = rx_desc_lim;
3720 	dev_info->tx_desc_lim = tx_desc_lim;
3721 
3722 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3723 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3724 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3725 
3726 	dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G;
3727 	if (hw->mac.type == ixgbe_mac_X540 ||
3728 	    hw->mac.type == ixgbe_mac_X540_vf ||
3729 	    hw->mac.type == ixgbe_mac_X550 ||
3730 	    hw->mac.type == ixgbe_mac_X550_vf) {
3731 		dev_info->speed_capa |= ETH_LINK_SPEED_100M;
3732 	}
3733 	if (hw->mac.type == ixgbe_mac_X550) {
3734 		dev_info->speed_capa |= ETH_LINK_SPEED_2_5G;
3735 		dev_info->speed_capa |= ETH_LINK_SPEED_5G;
3736 	}
3737 }
3738 
3739 static const uint32_t *
3740 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3741 {
3742 	static const uint32_t ptypes[] = {
3743 		/* For non-vec functions,
3744 		 * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3745 		 * for vec functions,
3746 		 * refers to _recv_raw_pkts_vec().
3747 		 */
3748 		RTE_PTYPE_L2_ETHER,
3749 		RTE_PTYPE_L3_IPV4,
3750 		RTE_PTYPE_L3_IPV4_EXT,
3751 		RTE_PTYPE_L3_IPV6,
3752 		RTE_PTYPE_L3_IPV6_EXT,
3753 		RTE_PTYPE_L4_SCTP,
3754 		RTE_PTYPE_L4_TCP,
3755 		RTE_PTYPE_L4_UDP,
3756 		RTE_PTYPE_TUNNEL_IP,
3757 		RTE_PTYPE_INNER_L3_IPV6,
3758 		RTE_PTYPE_INNER_L3_IPV6_EXT,
3759 		RTE_PTYPE_INNER_L4_TCP,
3760 		RTE_PTYPE_INNER_L4_UDP,
3761 		RTE_PTYPE_UNKNOWN
3762 	};
3763 
3764 	if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3765 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3766 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3767 	    dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3768 		return ptypes;
3769 
3770 #if defined(RTE_ARCH_X86)
3771 	if (dev->rx_pkt_burst == ixgbe_recv_pkts_vec ||
3772 	    dev->rx_pkt_burst == ixgbe_recv_scattered_pkts_vec)
3773 		return ptypes;
3774 #endif
3775 	return NULL;
3776 }
3777 
3778 static void
3779 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3780 		     struct rte_eth_dev_info *dev_info)
3781 {
3782 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3783 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3784 
3785 	dev_info->pci_dev = pci_dev;
3786 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3787 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3788 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3789 	dev_info->max_rx_pktlen = 9728; /* includes CRC, cf MAXFRS reg */
3790 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3791 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3792 	dev_info->max_vfs = pci_dev->max_vfs;
3793 	if (hw->mac.type == ixgbe_mac_82598EB)
3794 		dev_info->max_vmdq_pools = ETH_16_POOLS;
3795 	else
3796 		dev_info->max_vmdq_pools = ETH_64_POOLS;
3797 	dev_info->rx_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP |
3798 				DEV_RX_OFFLOAD_IPV4_CKSUM |
3799 				DEV_RX_OFFLOAD_UDP_CKSUM  |
3800 				DEV_RX_OFFLOAD_TCP_CKSUM;
3801 	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
3802 				DEV_TX_OFFLOAD_IPV4_CKSUM  |
3803 				DEV_TX_OFFLOAD_UDP_CKSUM   |
3804 				DEV_TX_OFFLOAD_TCP_CKSUM   |
3805 				DEV_TX_OFFLOAD_SCTP_CKSUM  |
3806 				DEV_TX_OFFLOAD_TCP_TSO;
3807 
3808 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3809 		.rx_thresh = {
3810 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3811 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3812 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3813 		},
3814 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3815 		.rx_drop_en = 0,
3816 	};
3817 
3818 	dev_info->default_txconf = (struct rte_eth_txconf) {
3819 		.tx_thresh = {
3820 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3821 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3822 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3823 		},
3824 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3825 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3826 		.txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
3827 				ETH_TXQ_FLAGS_NOOFFLOADS,
3828 	};
3829 
3830 	dev_info->rx_desc_lim = rx_desc_lim;
3831 	dev_info->tx_desc_lim = tx_desc_lim;
3832 }
3833 
3834 static int
3835 ixgbevf_check_link(struct ixgbe_hw *hw, ixgbe_link_speed *speed,
3836 		   int *link_up, int wait_to_complete)
3837 {
3838 	/**
3839 	 * for a quick link status checking, wait_to_compelet == 0,
3840 	 * skip PF link status checking
3841 	 */
3842 	bool no_pflink_check = wait_to_complete == 0;
3843 	struct ixgbe_mbx_info *mbx = &hw->mbx;
3844 	struct ixgbe_mac_info *mac = &hw->mac;
3845 	uint32_t links_reg, in_msg;
3846 	int ret_val = 0;
3847 
3848 	/* If we were hit with a reset drop the link */
3849 	if (!mbx->ops.check_for_rst(hw, 0) || !mbx->timeout)
3850 		mac->get_link_status = true;
3851 
3852 	if (!mac->get_link_status)
3853 		goto out;
3854 
3855 	/* if link status is down no point in checking to see if pf is up */
3856 	links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
3857 	if (!(links_reg & IXGBE_LINKS_UP))
3858 		goto out;
3859 
3860 	/* for SFP+ modules and DA cables on 82599 it can take up to 500usecs
3861 	 * before the link status is correct
3862 	 */
3863 	if (mac->type == ixgbe_mac_82599_vf) {
3864 		int i;
3865 
3866 		for (i = 0; i < 5; i++) {
3867 			rte_delay_us(100);
3868 			links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
3869 
3870 			if (!(links_reg & IXGBE_LINKS_UP))
3871 				goto out;
3872 		}
3873 	}
3874 
3875 	switch (links_reg & IXGBE_LINKS_SPEED_82599) {
3876 	case IXGBE_LINKS_SPEED_10G_82599:
3877 		*speed = IXGBE_LINK_SPEED_10GB_FULL;
3878 		if (hw->mac.type >= ixgbe_mac_X550) {
3879 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
3880 				*speed = IXGBE_LINK_SPEED_2_5GB_FULL;
3881 		}
3882 		break;
3883 	case IXGBE_LINKS_SPEED_1G_82599:
3884 		*speed = IXGBE_LINK_SPEED_1GB_FULL;
3885 		break;
3886 	case IXGBE_LINKS_SPEED_100_82599:
3887 		*speed = IXGBE_LINK_SPEED_100_FULL;
3888 		if (hw->mac.type == ixgbe_mac_X550) {
3889 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
3890 				*speed = IXGBE_LINK_SPEED_5GB_FULL;
3891 		}
3892 		break;
3893 	case IXGBE_LINKS_SPEED_10_X550EM_A:
3894 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
3895 		/* Since Reserved in older MAC's */
3896 		if (hw->mac.type >= ixgbe_mac_X550)
3897 			*speed = IXGBE_LINK_SPEED_10_FULL;
3898 		break;
3899 	default:
3900 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
3901 	}
3902 
3903 	if (no_pflink_check) {
3904 		if (*speed == IXGBE_LINK_SPEED_UNKNOWN)
3905 			mac->get_link_status = true;
3906 		else
3907 			mac->get_link_status = false;
3908 
3909 		goto out;
3910 	}
3911 	/* if the read failed it could just be a mailbox collision, best wait
3912 	 * until we are called again and don't report an error
3913 	 */
3914 	if (mbx->ops.read(hw, &in_msg, 1, 0))
3915 		goto out;
3916 
3917 	if (!(in_msg & IXGBE_VT_MSGTYPE_CTS)) {
3918 		/* msg is not CTS and is NACK we must have lost CTS status */
3919 		if (in_msg & IXGBE_VT_MSGTYPE_NACK)
3920 			ret_val = -1;
3921 		goto out;
3922 	}
3923 
3924 	/* the pf is talking, if we timed out in the past we reinit */
3925 	if (!mbx->timeout) {
3926 		ret_val = -1;
3927 		goto out;
3928 	}
3929 
3930 	/* if we passed all the tests above then the link is up and we no
3931 	 * longer need to check for link
3932 	 */
3933 	mac->get_link_status = false;
3934 
3935 out:
3936 	*link_up = !mac->get_link_status;
3937 	return ret_val;
3938 }
3939 
3940 /* return 0 means link status changed, -1 means not changed */
3941 static int
3942 ixgbe_dev_link_update_share(struct rte_eth_dev *dev,
3943 			    int wait_to_complete, int vf)
3944 {
3945 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3946 	struct rte_eth_link link, old;
3947 	ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
3948 	struct ixgbe_interrupt *intr =
3949 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3950 	int link_up;
3951 	int diag;
3952 	u32 speed = 0;
3953 	int wait = 1;
3954 	bool autoneg = false;
3955 
3956 	link.link_status = ETH_LINK_DOWN;
3957 	link.link_speed = 0;
3958 	link.link_duplex = ETH_LINK_HALF_DUPLEX;
3959 	link.link_autoneg = ETH_LINK_AUTONEG;
3960 	memset(&old, 0, sizeof(old));
3961 	rte_ixgbe_dev_atomic_read_link_status(dev, &old);
3962 
3963 	hw->mac.get_link_status = true;
3964 
3965 	if ((intr->flags & IXGBE_FLAG_NEED_LINK_CONFIG) &&
3966 		ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
3967 		speed = hw->phy.autoneg_advertised;
3968 		if (!speed)
3969 			ixgbe_get_link_capabilities(hw, &speed, &autoneg);
3970 		ixgbe_setup_link(hw, speed, true);
3971 	}
3972 
3973 	/* check if it needs to wait to complete, if lsc interrupt is enabled */
3974 	if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
3975 		wait = 0;
3976 
3977 	if (vf)
3978 		diag = ixgbevf_check_link(hw, &link_speed, &link_up, wait);
3979 	else
3980 		diag = ixgbe_check_link(hw, &link_speed, &link_up, wait);
3981 
3982 	if (diag != 0) {
3983 		link.link_speed = ETH_SPEED_NUM_100M;
3984 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
3985 		rte_ixgbe_dev_atomic_write_link_status(dev, &link);
3986 		if (link.link_status == old.link_status)
3987 			return -1;
3988 		return 0;
3989 	}
3990 
3991 	if (link_up == 0) {
3992 		rte_ixgbe_dev_atomic_write_link_status(dev, &link);
3993 		intr->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
3994 		if (link.link_status == old.link_status)
3995 			return -1;
3996 		return 0;
3997 	}
3998 	intr->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
3999 	link.link_status = ETH_LINK_UP;
4000 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
4001 
4002 	switch (link_speed) {
4003 	default:
4004 	case IXGBE_LINK_SPEED_UNKNOWN:
4005 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
4006 		link.link_speed = ETH_SPEED_NUM_100M;
4007 		break;
4008 
4009 	case IXGBE_LINK_SPEED_100_FULL:
4010 		link.link_speed = ETH_SPEED_NUM_100M;
4011 		break;
4012 
4013 	case IXGBE_LINK_SPEED_1GB_FULL:
4014 		link.link_speed = ETH_SPEED_NUM_1G;
4015 		break;
4016 
4017 	case IXGBE_LINK_SPEED_2_5GB_FULL:
4018 		link.link_speed = ETH_SPEED_NUM_2_5G;
4019 		break;
4020 
4021 	case IXGBE_LINK_SPEED_5GB_FULL:
4022 		link.link_speed = ETH_SPEED_NUM_5G;
4023 		break;
4024 
4025 	case IXGBE_LINK_SPEED_10GB_FULL:
4026 		link.link_speed = ETH_SPEED_NUM_10G;
4027 		break;
4028 	}
4029 	rte_ixgbe_dev_atomic_write_link_status(dev, &link);
4030 
4031 	if (link.link_status == old.link_status)
4032 		return -1;
4033 
4034 	return 0;
4035 }
4036 
4037 static int
4038 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4039 {
4040 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 0);
4041 }
4042 
4043 static int
4044 ixgbevf_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4045 {
4046 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 1);
4047 }
4048 
4049 static void
4050 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
4051 {
4052 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4053 	uint32_t fctrl;
4054 
4055 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4056 	fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4057 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4058 }
4059 
4060 static void
4061 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
4062 {
4063 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4064 	uint32_t fctrl;
4065 
4066 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4067 	fctrl &= (~IXGBE_FCTRL_UPE);
4068 	if (dev->data->all_multicast == 1)
4069 		fctrl |= IXGBE_FCTRL_MPE;
4070 	else
4071 		fctrl &= (~IXGBE_FCTRL_MPE);
4072 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4073 }
4074 
4075 static void
4076 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
4077 {
4078 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4079 	uint32_t fctrl;
4080 
4081 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4082 	fctrl |= IXGBE_FCTRL_MPE;
4083 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4084 }
4085 
4086 static void
4087 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
4088 {
4089 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4090 	uint32_t fctrl;
4091 
4092 	if (dev->data->promiscuous == 1)
4093 		return; /* must remain in all_multicast mode */
4094 
4095 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4096 	fctrl &= (~IXGBE_FCTRL_MPE);
4097 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4098 }
4099 
4100 /**
4101  * It clears the interrupt causes and enables the interrupt.
4102  * It will be called once only during nic initialized.
4103  *
4104  * @param dev
4105  *  Pointer to struct rte_eth_dev.
4106  * @param on
4107  *  Enable or Disable.
4108  *
4109  * @return
4110  *  - On success, zero.
4111  *  - On failure, a negative value.
4112  */
4113 static int
4114 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
4115 {
4116 	struct ixgbe_interrupt *intr =
4117 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4118 
4119 	ixgbe_dev_link_status_print(dev);
4120 	if (on)
4121 		intr->mask |= IXGBE_EICR_LSC;
4122 	else
4123 		intr->mask &= ~IXGBE_EICR_LSC;
4124 
4125 	return 0;
4126 }
4127 
4128 /**
4129  * It clears the interrupt causes and enables the interrupt.
4130  * It will be called once only during nic initialized.
4131  *
4132  * @param dev
4133  *  Pointer to struct rte_eth_dev.
4134  *
4135  * @return
4136  *  - On success, zero.
4137  *  - On failure, a negative value.
4138  */
4139 static int
4140 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
4141 {
4142 	struct ixgbe_interrupt *intr =
4143 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4144 
4145 	intr->mask |= IXGBE_EICR_RTX_QUEUE;
4146 
4147 	return 0;
4148 }
4149 
4150 /**
4151  * It clears the interrupt causes and enables the interrupt.
4152  * It will be called once only during nic initialized.
4153  *
4154  * @param dev
4155  *  Pointer to struct rte_eth_dev.
4156  *
4157  * @return
4158  *  - On success, zero.
4159  *  - On failure, a negative value.
4160  */
4161 static int
4162 ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev)
4163 {
4164 	struct ixgbe_interrupt *intr =
4165 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4166 
4167 	intr->mask |= IXGBE_EICR_LINKSEC;
4168 
4169 	return 0;
4170 }
4171 
4172 /*
4173  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
4174  *
4175  * @param dev
4176  *  Pointer to struct rte_eth_dev.
4177  *
4178  * @return
4179  *  - On success, zero.
4180  *  - On failure, a negative value.
4181  */
4182 static int
4183 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
4184 {
4185 	uint32_t eicr;
4186 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4187 	struct ixgbe_interrupt *intr =
4188 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4189 
4190 	/* clear all cause mask */
4191 	ixgbe_disable_intr(hw);
4192 
4193 	/* read-on-clear nic registers here */
4194 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4195 	PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
4196 
4197 	intr->flags = 0;
4198 
4199 	/* set flag for async link update */
4200 	if (eicr & IXGBE_EICR_LSC)
4201 		intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4202 
4203 	if (eicr & IXGBE_EICR_MAILBOX)
4204 		intr->flags |= IXGBE_FLAG_MAILBOX;
4205 
4206 	if (eicr & IXGBE_EICR_LINKSEC)
4207 		intr->flags |= IXGBE_FLAG_MACSEC;
4208 
4209 	if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
4210 	    hw->phy.type == ixgbe_phy_x550em_ext_t &&
4211 	    (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
4212 		intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
4213 
4214 	return 0;
4215 }
4216 
4217 /**
4218  * It gets and then prints the link status.
4219  *
4220  * @param dev
4221  *  Pointer to struct rte_eth_dev.
4222  *
4223  * @return
4224  *  - On success, zero.
4225  *  - On failure, a negative value.
4226  */
4227 static void
4228 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
4229 {
4230 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4231 	struct rte_eth_link link;
4232 
4233 	memset(&link, 0, sizeof(link));
4234 	rte_ixgbe_dev_atomic_read_link_status(dev, &link);
4235 	if (link.link_status) {
4236 		PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
4237 					(int)(dev->data->port_id),
4238 					(unsigned)link.link_speed,
4239 			link.link_duplex == ETH_LINK_FULL_DUPLEX ?
4240 					"full-duplex" : "half-duplex");
4241 	} else {
4242 		PMD_INIT_LOG(INFO, " Port %d: Link Down",
4243 				(int)(dev->data->port_id));
4244 	}
4245 	PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
4246 				pci_dev->addr.domain,
4247 				pci_dev->addr.bus,
4248 				pci_dev->addr.devid,
4249 				pci_dev->addr.function);
4250 }
4251 
4252 /*
4253  * It executes link_update after knowing an interrupt occurred.
4254  *
4255  * @param dev
4256  *  Pointer to struct rte_eth_dev.
4257  *
4258  * @return
4259  *  - On success, zero.
4260  *  - On failure, a negative value.
4261  */
4262 static int
4263 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev,
4264 			   struct rte_intr_handle *intr_handle)
4265 {
4266 	struct ixgbe_interrupt *intr =
4267 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4268 	int64_t timeout;
4269 	struct rte_eth_link link;
4270 	struct ixgbe_hw *hw =
4271 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4272 
4273 	PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
4274 
4275 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
4276 		ixgbe_pf_mbx_process(dev);
4277 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
4278 	}
4279 
4280 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4281 		ixgbe_handle_lasi(hw);
4282 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4283 	}
4284 
4285 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4286 		/* get the link status before link update, for predicting later */
4287 		memset(&link, 0, sizeof(link));
4288 		rte_ixgbe_dev_atomic_read_link_status(dev, &link);
4289 
4290 		ixgbe_dev_link_update(dev, 0);
4291 
4292 		/* likely to up */
4293 		if (!link.link_status)
4294 			/* handle it 1 sec later, wait it being stable */
4295 			timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
4296 		/* likely to down */
4297 		else
4298 			/* handle it 4 sec later, wait it being stable */
4299 			timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
4300 
4301 		ixgbe_dev_link_status_print(dev);
4302 		if (rte_eal_alarm_set(timeout * 1000,
4303 				      ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
4304 			PMD_DRV_LOG(ERR, "Error setting alarm");
4305 		else {
4306 			/* remember original mask */
4307 			intr->mask_original = intr->mask;
4308 			/* only disable lsc interrupt */
4309 			intr->mask &= ~IXGBE_EIMS_LSC;
4310 		}
4311 	}
4312 
4313 	PMD_DRV_LOG(DEBUG, "enable intr immediately");
4314 	ixgbe_enable_intr(dev);
4315 	rte_intr_enable(intr_handle);
4316 
4317 	return 0;
4318 }
4319 
4320 /**
4321  * Interrupt handler which shall be registered for alarm callback for delayed
4322  * handling specific interrupt to wait for the stable nic state. As the
4323  * NIC interrupt state is not stable for ixgbe after link is just down,
4324  * it needs to wait 4 seconds to get the stable status.
4325  *
4326  * @param handle
4327  *  Pointer to interrupt handle.
4328  * @param param
4329  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4330  *
4331  * @return
4332  *  void
4333  */
4334 static void
4335 ixgbe_dev_interrupt_delayed_handler(void *param)
4336 {
4337 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4338 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4339 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
4340 	struct ixgbe_interrupt *intr =
4341 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4342 	struct ixgbe_hw *hw =
4343 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4344 	uint32_t eicr;
4345 
4346 	ixgbe_disable_intr(hw);
4347 
4348 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4349 	if (eicr & IXGBE_EICR_MAILBOX)
4350 		ixgbe_pf_mbx_process(dev);
4351 
4352 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4353 		ixgbe_handle_lasi(hw);
4354 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4355 	}
4356 
4357 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4358 		ixgbe_dev_link_update(dev, 0);
4359 		intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4360 		ixgbe_dev_link_status_print(dev);
4361 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC,
4362 					      NULL);
4363 	}
4364 
4365 	if (intr->flags & IXGBE_FLAG_MACSEC) {
4366 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_MACSEC,
4367 					      NULL);
4368 		intr->flags &= ~IXGBE_FLAG_MACSEC;
4369 	}
4370 
4371 	/* restore original mask */
4372 	intr->mask = intr->mask_original;
4373 	intr->mask_original = 0;
4374 
4375 	PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
4376 	ixgbe_enable_intr(dev);
4377 	rte_intr_enable(intr_handle);
4378 }
4379 
4380 /**
4381  * Interrupt handler triggered by NIC  for handling
4382  * specific interrupt.
4383  *
4384  * @param handle
4385  *  Pointer to interrupt handle.
4386  * @param param
4387  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4388  *
4389  * @return
4390  *  void
4391  */
4392 static void
4393 ixgbe_dev_interrupt_handler(void *param)
4394 {
4395 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4396 
4397 	ixgbe_dev_interrupt_get_status(dev);
4398 	ixgbe_dev_interrupt_action(dev, dev->intr_handle);
4399 }
4400 
4401 static int
4402 ixgbe_dev_led_on(struct rte_eth_dev *dev)
4403 {
4404 	struct ixgbe_hw *hw;
4405 
4406 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4407 	return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4408 }
4409 
4410 static int
4411 ixgbe_dev_led_off(struct rte_eth_dev *dev)
4412 {
4413 	struct ixgbe_hw *hw;
4414 
4415 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4416 	return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4417 }
4418 
4419 static int
4420 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4421 {
4422 	struct ixgbe_hw *hw;
4423 	uint32_t mflcn_reg;
4424 	uint32_t fccfg_reg;
4425 	int rx_pause;
4426 	int tx_pause;
4427 
4428 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4429 
4430 	fc_conf->pause_time = hw->fc.pause_time;
4431 	fc_conf->high_water = hw->fc.high_water[0];
4432 	fc_conf->low_water = hw->fc.low_water[0];
4433 	fc_conf->send_xon = hw->fc.send_xon;
4434 	fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
4435 
4436 	/*
4437 	 * Return rx_pause status according to actual setting of
4438 	 * MFLCN register.
4439 	 */
4440 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4441 	if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
4442 		rx_pause = 1;
4443 	else
4444 		rx_pause = 0;
4445 
4446 	/*
4447 	 * Return tx_pause status according to actual setting of
4448 	 * FCCFG register.
4449 	 */
4450 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4451 	if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
4452 		tx_pause = 1;
4453 	else
4454 		tx_pause = 0;
4455 
4456 	if (rx_pause && tx_pause)
4457 		fc_conf->mode = RTE_FC_FULL;
4458 	else if (rx_pause)
4459 		fc_conf->mode = RTE_FC_RX_PAUSE;
4460 	else if (tx_pause)
4461 		fc_conf->mode = RTE_FC_TX_PAUSE;
4462 	else
4463 		fc_conf->mode = RTE_FC_NONE;
4464 
4465 	return 0;
4466 }
4467 
4468 static int
4469 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4470 {
4471 	struct ixgbe_hw *hw;
4472 	int err;
4473 	uint32_t rx_buf_size;
4474 	uint32_t max_high_water;
4475 	uint32_t mflcn;
4476 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4477 		ixgbe_fc_none,
4478 		ixgbe_fc_rx_pause,
4479 		ixgbe_fc_tx_pause,
4480 		ixgbe_fc_full
4481 	};
4482 
4483 	PMD_INIT_FUNC_TRACE();
4484 
4485 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4486 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
4487 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4488 
4489 	/*
4490 	 * At least reserve one Ethernet frame for watermark
4491 	 * high_water/low_water in kilo bytes for ixgbe
4492 	 */
4493 	max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4494 	if ((fc_conf->high_water > max_high_water) ||
4495 		(fc_conf->high_water < fc_conf->low_water)) {
4496 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4497 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4498 		return -EINVAL;
4499 	}
4500 
4501 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
4502 	hw->fc.pause_time     = fc_conf->pause_time;
4503 	hw->fc.high_water[0]  = fc_conf->high_water;
4504 	hw->fc.low_water[0]   = fc_conf->low_water;
4505 	hw->fc.send_xon       = fc_conf->send_xon;
4506 	hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
4507 
4508 	err = ixgbe_fc_enable(hw);
4509 
4510 	/* Not negotiated is not an error case */
4511 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED)) {
4512 
4513 		/* check if we want to forward MAC frames - driver doesn't have native
4514 		 * capability to do that, so we'll write the registers ourselves */
4515 
4516 		mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4517 
4518 		/* set or clear MFLCN.PMCF bit depending on configuration */
4519 		if (fc_conf->mac_ctrl_frame_fwd != 0)
4520 			mflcn |= IXGBE_MFLCN_PMCF;
4521 		else
4522 			mflcn &= ~IXGBE_MFLCN_PMCF;
4523 
4524 		IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
4525 		IXGBE_WRITE_FLUSH(hw);
4526 
4527 		return 0;
4528 	}
4529 
4530 	PMD_INIT_LOG(ERR, "ixgbe_fc_enable = 0x%x", err);
4531 	return -EIO;
4532 }
4533 
4534 /**
4535  *  ixgbe_pfc_enable_generic - Enable flow control
4536  *  @hw: pointer to hardware structure
4537  *  @tc_num: traffic class number
4538  *  Enable flow control according to the current settings.
4539  */
4540 static int
4541 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
4542 {
4543 	int ret_val = 0;
4544 	uint32_t mflcn_reg, fccfg_reg;
4545 	uint32_t reg;
4546 	uint32_t fcrtl, fcrth;
4547 	uint8_t i;
4548 	uint8_t nb_rx_en;
4549 
4550 	/* Validate the water mark configuration */
4551 	if (!hw->fc.pause_time) {
4552 		ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4553 		goto out;
4554 	}
4555 
4556 	/* Low water mark of zero causes XOFF floods */
4557 	if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
4558 		 /* High/Low water can not be 0 */
4559 		if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_water[tc_num])) {
4560 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4561 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4562 			goto out;
4563 		}
4564 
4565 		if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
4566 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4567 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4568 			goto out;
4569 		}
4570 	}
4571 	/* Negotiate the fc mode to use */
4572 	ixgbe_fc_autoneg(hw);
4573 
4574 	/* Disable any previous flow control settings */
4575 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4576 	mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
4577 
4578 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4579 	fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
4580 
4581 	switch (hw->fc.current_mode) {
4582 	case ixgbe_fc_none:
4583 		/*
4584 		 * If the count of enabled RX Priority Flow control >1,
4585 		 * and the TX pause can not be disabled
4586 		 */
4587 		nb_rx_en = 0;
4588 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4589 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4590 			if (reg & IXGBE_FCRTH_FCEN)
4591 				nb_rx_en++;
4592 		}
4593 		if (nb_rx_en > 1)
4594 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4595 		break;
4596 	case ixgbe_fc_rx_pause:
4597 		/*
4598 		 * Rx Flow control is enabled and Tx Flow control is
4599 		 * disabled by software override. Since there really
4600 		 * isn't a way to advertise that we are capable of RX
4601 		 * Pause ONLY, we will advertise that we support both
4602 		 * symmetric and asymmetric Rx PAUSE.  Later, we will
4603 		 * disable the adapter's ability to send PAUSE frames.
4604 		 */
4605 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4606 		/*
4607 		 * If the count of enabled RX Priority Flow control >1,
4608 		 * and the TX pause can not be disabled
4609 		 */
4610 		nb_rx_en = 0;
4611 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4612 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4613 			if (reg & IXGBE_FCRTH_FCEN)
4614 				nb_rx_en++;
4615 		}
4616 		if (nb_rx_en > 1)
4617 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4618 		break;
4619 	case ixgbe_fc_tx_pause:
4620 		/*
4621 		 * Tx Flow control is enabled, and Rx Flow control is
4622 		 * disabled by software override.
4623 		 */
4624 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4625 		break;
4626 	case ixgbe_fc_full:
4627 		/* Flow control (both Rx and Tx) is enabled by SW override. */
4628 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4629 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4630 		break;
4631 	default:
4632 		PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
4633 		ret_val = IXGBE_ERR_CONFIG;
4634 		goto out;
4635 	}
4636 
4637 	/* Set 802.3x based flow control settings. */
4638 	mflcn_reg |= IXGBE_MFLCN_DPF;
4639 	IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
4640 	IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
4641 
4642 	/* Set up and enable Rx high/low water mark thresholds, enable XON. */
4643 	if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
4644 		hw->fc.high_water[tc_num]) {
4645 		fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
4646 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
4647 		fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
4648 	} else {
4649 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
4650 		/*
4651 		 * In order to prevent Tx hangs when the internal Tx
4652 		 * switch is enabled we must set the high water mark
4653 		 * to the maximum FCRTH value.  This allows the Tx
4654 		 * switch to function even under heavy Rx workloads.
4655 		 */
4656 		fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
4657 	}
4658 	IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
4659 
4660 	/* Configure pause time (2 TCs per register) */
4661 	reg = hw->fc.pause_time * 0x00010001;
4662 	for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
4663 		IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
4664 
4665 	/* Configure flow control refresh threshold value */
4666 	IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
4667 
4668 out:
4669 	return ret_val;
4670 }
4671 
4672 static int
4673 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
4674 {
4675 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4676 	int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
4677 
4678 	if (hw->mac.type != ixgbe_mac_82598EB) {
4679 		ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
4680 	}
4681 	return ret_val;
4682 }
4683 
4684 static int
4685 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
4686 {
4687 	int err;
4688 	uint32_t rx_buf_size;
4689 	uint32_t max_high_water;
4690 	uint8_t tc_num;
4691 	uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
4692 	struct ixgbe_hw *hw =
4693 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4694 	struct ixgbe_dcb_config *dcb_config =
4695 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
4696 
4697 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4698 		ixgbe_fc_none,
4699 		ixgbe_fc_rx_pause,
4700 		ixgbe_fc_tx_pause,
4701 		ixgbe_fc_full
4702 	};
4703 
4704 	PMD_INIT_FUNC_TRACE();
4705 
4706 	ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
4707 	tc_num = map[pfc_conf->priority];
4708 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
4709 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4710 	/*
4711 	 * At least reserve one Ethernet frame for watermark
4712 	 * high_water/low_water in kilo bytes for ixgbe
4713 	 */
4714 	max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4715 	if ((pfc_conf->fc.high_water > max_high_water) ||
4716 	    (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
4717 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4718 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4719 		return -EINVAL;
4720 	}
4721 
4722 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
4723 	hw->fc.pause_time = pfc_conf->fc.pause_time;
4724 	hw->fc.send_xon = pfc_conf->fc.send_xon;
4725 	hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
4726 	hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
4727 
4728 	err = ixgbe_dcb_pfc_enable(dev, tc_num);
4729 
4730 	/* Not negotiated is not an error case */
4731 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
4732 		return 0;
4733 
4734 	PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
4735 	return -EIO;
4736 }
4737 
4738 static int
4739 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
4740 			  struct rte_eth_rss_reta_entry64 *reta_conf,
4741 			  uint16_t reta_size)
4742 {
4743 	uint16_t i, sp_reta_size;
4744 	uint8_t j, mask;
4745 	uint32_t reta, r;
4746 	uint16_t idx, shift;
4747 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4748 	uint32_t reta_reg;
4749 
4750 	PMD_INIT_FUNC_TRACE();
4751 
4752 	if (!ixgbe_rss_update_sp(hw->mac.type)) {
4753 		PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
4754 			"NIC.");
4755 		return -ENOTSUP;
4756 	}
4757 
4758 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
4759 	if (reta_size != sp_reta_size) {
4760 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
4761 			"(%d) doesn't match the number hardware can supported "
4762 			"(%d)", reta_size, sp_reta_size);
4763 		return -EINVAL;
4764 	}
4765 
4766 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
4767 		idx = i / RTE_RETA_GROUP_SIZE;
4768 		shift = i % RTE_RETA_GROUP_SIZE;
4769 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
4770 						IXGBE_4_BIT_MASK);
4771 		if (!mask)
4772 			continue;
4773 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
4774 		if (mask == IXGBE_4_BIT_MASK)
4775 			r = 0;
4776 		else
4777 			r = IXGBE_READ_REG(hw, reta_reg);
4778 		for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
4779 			if (mask & (0x1 << j))
4780 				reta |= reta_conf[idx].reta[shift + j] <<
4781 							(CHAR_BIT * j);
4782 			else
4783 				reta |= r & (IXGBE_8_BIT_MASK <<
4784 						(CHAR_BIT * j));
4785 		}
4786 		IXGBE_WRITE_REG(hw, reta_reg, reta);
4787 	}
4788 
4789 	return 0;
4790 }
4791 
4792 static int
4793 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
4794 			 struct rte_eth_rss_reta_entry64 *reta_conf,
4795 			 uint16_t reta_size)
4796 {
4797 	uint16_t i, sp_reta_size;
4798 	uint8_t j, mask;
4799 	uint32_t reta;
4800 	uint16_t idx, shift;
4801 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4802 	uint32_t reta_reg;
4803 
4804 	PMD_INIT_FUNC_TRACE();
4805 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
4806 	if (reta_size != sp_reta_size) {
4807 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
4808 			"(%d) doesn't match the number hardware can supported "
4809 			"(%d)", reta_size, sp_reta_size);
4810 		return -EINVAL;
4811 	}
4812 
4813 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
4814 		idx = i / RTE_RETA_GROUP_SIZE;
4815 		shift = i % RTE_RETA_GROUP_SIZE;
4816 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
4817 						IXGBE_4_BIT_MASK);
4818 		if (!mask)
4819 			continue;
4820 
4821 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
4822 		reta = IXGBE_READ_REG(hw, reta_reg);
4823 		for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
4824 			if (mask & (0x1 << j))
4825 				reta_conf[idx].reta[shift + j] =
4826 					((reta >> (CHAR_BIT * j)) &
4827 						IXGBE_8_BIT_MASK);
4828 		}
4829 	}
4830 
4831 	return 0;
4832 }
4833 
4834 static int
4835 ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
4836 				uint32_t index, uint32_t pool)
4837 {
4838 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4839 	uint32_t enable_addr = 1;
4840 
4841 	return ixgbe_set_rar(hw, index, mac_addr->addr_bytes,
4842 			     pool, enable_addr);
4843 }
4844 
4845 static void
4846 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
4847 {
4848 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4849 
4850 	ixgbe_clear_rar(hw, index);
4851 }
4852 
4853 static void
4854 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
4855 {
4856 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4857 
4858 	ixgbe_remove_rar(dev, 0);
4859 
4860 	ixgbe_add_rar(dev, addr, 0, pci_dev->max_vfs);
4861 }
4862 
4863 static bool
4864 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
4865 {
4866 	if (strcmp(dev->device->driver->name, drv->driver.name))
4867 		return false;
4868 
4869 	return true;
4870 }
4871 
4872 bool
4873 is_ixgbe_supported(struct rte_eth_dev *dev)
4874 {
4875 	return is_device_supported(dev, &rte_ixgbe_pmd);
4876 }
4877 
4878 static int
4879 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4880 {
4881 	uint32_t hlreg0;
4882 	uint32_t maxfrs;
4883 	struct ixgbe_hw *hw;
4884 	struct rte_eth_dev_info dev_info;
4885 	uint32_t frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
4886 	struct rte_eth_dev_data *dev_data = dev->data;
4887 
4888 	ixgbe_dev_info_get(dev, &dev_info);
4889 
4890 	/* check that mtu is within the allowed range */
4891 	if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
4892 		return -EINVAL;
4893 
4894 	/* If device is started, refuse mtu that requires the support of
4895 	 * scattered packets when this feature has not been enabled before.
4896 	 */
4897 	if (dev_data->dev_started && !dev_data->scattered_rx &&
4898 	    (frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
4899 	     dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
4900 		PMD_INIT_LOG(ERR, "Stop port first.");
4901 		return -EINVAL;
4902 	}
4903 
4904 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4905 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4906 
4907 	/* switch to jumbo mode if needed */
4908 	if (frame_size > ETHER_MAX_LEN) {
4909 		dev->data->dev_conf.rxmode.jumbo_frame = 1;
4910 		hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4911 	} else {
4912 		dev->data->dev_conf.rxmode.jumbo_frame = 0;
4913 		hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
4914 	}
4915 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4916 
4917 	/* update max frame size */
4918 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4919 
4920 	maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
4921 	maxfrs &= 0x0000FFFF;
4922 	maxfrs |= (dev->data->dev_conf.rxmode.max_rx_pkt_len << 16);
4923 	IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
4924 
4925 	return 0;
4926 }
4927 
4928 /*
4929  * Virtual Function operations
4930  */
4931 static void
4932 ixgbevf_intr_disable(struct ixgbe_hw *hw)
4933 {
4934 	PMD_INIT_FUNC_TRACE();
4935 
4936 	/* Clear interrupt mask to stop from interrupts being generated */
4937 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
4938 
4939 	IXGBE_WRITE_FLUSH(hw);
4940 }
4941 
4942 static void
4943 ixgbevf_intr_enable(struct ixgbe_hw *hw)
4944 {
4945 	PMD_INIT_FUNC_TRACE();
4946 
4947 	/* VF enable interrupt autoclean */
4948 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
4949 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
4950 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
4951 
4952 	IXGBE_WRITE_FLUSH(hw);
4953 }
4954 
4955 static int
4956 ixgbevf_dev_configure(struct rte_eth_dev *dev)
4957 {
4958 	struct rte_eth_conf *conf = &dev->data->dev_conf;
4959 	struct ixgbe_adapter *adapter =
4960 			(struct ixgbe_adapter *)dev->data->dev_private;
4961 
4962 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
4963 		     dev->data->port_id);
4964 
4965 	/*
4966 	 * VF has no ability to enable/disable HW CRC
4967 	 * Keep the persistent behavior the same as Host PF
4968 	 */
4969 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
4970 	if (!conf->rxmode.hw_strip_crc) {
4971 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
4972 		conf->rxmode.hw_strip_crc = 1;
4973 	}
4974 #else
4975 	if (conf->rxmode.hw_strip_crc) {
4976 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
4977 		conf->rxmode.hw_strip_crc = 0;
4978 	}
4979 #endif
4980 
4981 	/*
4982 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
4983 	 * allocation or vector Rx preconditions we will reset it.
4984 	 */
4985 	adapter->rx_bulk_alloc_allowed = true;
4986 	adapter->rx_vec_allowed = true;
4987 
4988 	return 0;
4989 }
4990 
4991 static int
4992 ixgbevf_dev_start(struct rte_eth_dev *dev)
4993 {
4994 	struct ixgbe_hw *hw =
4995 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4996 	uint32_t intr_vector = 0;
4997 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4998 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
4999 
5000 	int err, mask = 0;
5001 
5002 	PMD_INIT_FUNC_TRACE();
5003 
5004 	hw->mac.ops.reset_hw(hw);
5005 	hw->mac.get_link_status = true;
5006 
5007 	/* negotiate mailbox API version to use with the PF. */
5008 	ixgbevf_negotiate_api(hw);
5009 
5010 	ixgbevf_dev_tx_init(dev);
5011 
5012 	/* This can fail when allocating mbufs for descriptor rings */
5013 	err = ixgbevf_dev_rx_init(dev);
5014 	if (err) {
5015 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
5016 		ixgbe_dev_clear_queues(dev);
5017 		return err;
5018 	}
5019 
5020 	/* Set vfta */
5021 	ixgbevf_set_vfta_all(dev, 1);
5022 
5023 	/* Set HW strip */
5024 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
5025 		ETH_VLAN_EXTEND_MASK;
5026 	err = ixgbevf_vlan_offload_set(dev, mask);
5027 	if (err) {
5028 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
5029 		ixgbe_dev_clear_queues(dev);
5030 		return err;
5031 	}
5032 
5033 	ixgbevf_dev_rxtx_start(dev);
5034 
5035 	/* check and configure queue intr-vector mapping */
5036 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
5037 		/* According to datasheet, only vector 0/1/2 can be used,
5038 		 * now only one vector is used for Rx queue
5039 		 */
5040 		intr_vector = 1;
5041 		if (rte_intr_efd_enable(intr_handle, intr_vector))
5042 			return -1;
5043 	}
5044 
5045 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
5046 		intr_handle->intr_vec =
5047 			rte_zmalloc("intr_vec",
5048 				    dev->data->nb_rx_queues * sizeof(int), 0);
5049 		if (intr_handle->intr_vec == NULL) {
5050 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
5051 				     " intr_vec", dev->data->nb_rx_queues);
5052 			return -ENOMEM;
5053 		}
5054 	}
5055 	ixgbevf_configure_msix(dev);
5056 
5057 	/* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
5058 	 * is mapped to VFIO vector 0 in eth_ixgbevf_dev_init( ).
5059 	 * If previous VFIO interrupt mapping setting in eth_ixgbevf_dev_init( )
5060 	 * is not cleared, it will fail when following rte_intr_enable( ) tries
5061 	 * to map Rx queue interrupt to other VFIO vectors.
5062 	 * So clear uio/vfio intr/evevnfd first to avoid failure.
5063 	 */
5064 	rte_intr_disable(intr_handle);
5065 
5066 	rte_intr_enable(intr_handle);
5067 
5068 	/* Re-enable interrupt for VF */
5069 	ixgbevf_intr_enable(hw);
5070 
5071 	return 0;
5072 }
5073 
5074 static void
5075 ixgbevf_dev_stop(struct rte_eth_dev *dev)
5076 {
5077 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5078 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5079 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5080 
5081 	PMD_INIT_FUNC_TRACE();
5082 
5083 	ixgbevf_intr_disable(hw);
5084 
5085 	hw->adapter_stopped = 1;
5086 	ixgbe_stop_adapter(hw);
5087 
5088 	/*
5089 	  * Clear what we set, but we still keep shadow_vfta to
5090 	  * restore after device starts
5091 	  */
5092 	ixgbevf_set_vfta_all(dev, 0);
5093 
5094 	/* Clear stored conf */
5095 	dev->data->scattered_rx = 0;
5096 
5097 	ixgbe_dev_clear_queues(dev);
5098 
5099 	/* Clean datapath event and queue/vec mapping */
5100 	rte_intr_efd_disable(intr_handle);
5101 	if (intr_handle->intr_vec != NULL) {
5102 		rte_free(intr_handle->intr_vec);
5103 		intr_handle->intr_vec = NULL;
5104 	}
5105 }
5106 
5107 static void
5108 ixgbevf_dev_close(struct rte_eth_dev *dev)
5109 {
5110 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5111 
5112 	PMD_INIT_FUNC_TRACE();
5113 
5114 	ixgbe_reset_hw(hw);
5115 
5116 	ixgbevf_dev_stop(dev);
5117 
5118 	ixgbe_dev_free_queues(dev);
5119 
5120 	/**
5121 	 * Remove the VF MAC address ro ensure
5122 	 * that the VF traffic goes to the PF
5123 	 * after stop, close and detach of the VF
5124 	 **/
5125 	ixgbevf_remove_mac_addr(dev, 0);
5126 }
5127 
5128 /*
5129  * Reset VF device
5130  */
5131 static int
5132 ixgbevf_dev_reset(struct rte_eth_dev *dev)
5133 {
5134 	int ret;
5135 
5136 	ret = eth_ixgbevf_dev_uninit(dev);
5137 	if (ret)
5138 		return ret;
5139 
5140 	ret = eth_ixgbevf_dev_init(dev);
5141 
5142 	return ret;
5143 }
5144 
5145 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
5146 {
5147 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5148 	struct ixgbe_vfta *shadow_vfta =
5149 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5150 	int i = 0, j = 0, vfta = 0, mask = 1;
5151 
5152 	for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
5153 		vfta = shadow_vfta->vfta[i];
5154 		if (vfta) {
5155 			mask = 1;
5156 			for (j = 0; j < 32; j++) {
5157 				if (vfta & mask)
5158 					ixgbe_set_vfta(hw, (i<<5)+j, 0,
5159 						       on, false);
5160 				mask <<= 1;
5161 			}
5162 		}
5163 	}
5164 
5165 }
5166 
5167 static int
5168 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
5169 {
5170 	struct ixgbe_hw *hw =
5171 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5172 	struct ixgbe_vfta *shadow_vfta =
5173 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5174 	uint32_t vid_idx = 0;
5175 	uint32_t vid_bit = 0;
5176 	int ret = 0;
5177 
5178 	PMD_INIT_FUNC_TRACE();
5179 
5180 	/* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
5181 	ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
5182 	if (ret) {
5183 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
5184 		return ret;
5185 	}
5186 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
5187 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
5188 
5189 	/* Save what we set and retore it after device reset */
5190 	if (on)
5191 		shadow_vfta->vfta[vid_idx] |= vid_bit;
5192 	else
5193 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
5194 
5195 	return 0;
5196 }
5197 
5198 static void
5199 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
5200 {
5201 	struct ixgbe_hw *hw =
5202 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5203 	uint32_t ctrl;
5204 
5205 	PMD_INIT_FUNC_TRACE();
5206 
5207 	if (queue >= hw->mac.max_rx_queues)
5208 		return;
5209 
5210 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
5211 	if (on)
5212 		ctrl |= IXGBE_RXDCTL_VME;
5213 	else
5214 		ctrl &= ~IXGBE_RXDCTL_VME;
5215 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
5216 
5217 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
5218 }
5219 
5220 static int
5221 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
5222 {
5223 	struct ixgbe_hw *hw =
5224 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5225 	uint16_t i;
5226 	int on = 0;
5227 
5228 	/* VF function only support hw strip feature, others are not support */
5229 	if (mask & ETH_VLAN_STRIP_MASK) {
5230 		on = !!(dev->data->dev_conf.rxmode.hw_vlan_strip);
5231 
5232 		for (i = 0; i < hw->mac.max_rx_queues; i++)
5233 			ixgbevf_vlan_strip_queue_set(dev, i, on);
5234 	}
5235 
5236 	return 0;
5237 }
5238 
5239 int
5240 ixgbe_vt_check(struct ixgbe_hw *hw)
5241 {
5242 	uint32_t reg_val;
5243 
5244 	/* if Virtualization Technology is enabled */
5245 	reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
5246 	if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
5247 		PMD_INIT_LOG(ERR, "VT must be enabled for this setting");
5248 		return -1;
5249 	}
5250 
5251 	return 0;
5252 }
5253 
5254 static uint32_t
5255 ixgbe_uta_vector(struct ixgbe_hw *hw, struct ether_addr *uc_addr)
5256 {
5257 	uint32_t vector = 0;
5258 
5259 	switch (hw->mac.mc_filter_type) {
5260 	case 0:   /* use bits [47:36] of the address */
5261 		vector = ((uc_addr->addr_bytes[4] >> 4) |
5262 			(((uint16_t)uc_addr->addr_bytes[5]) << 4));
5263 		break;
5264 	case 1:   /* use bits [46:35] of the address */
5265 		vector = ((uc_addr->addr_bytes[4] >> 3) |
5266 			(((uint16_t)uc_addr->addr_bytes[5]) << 5));
5267 		break;
5268 	case 2:   /* use bits [45:34] of the address */
5269 		vector = ((uc_addr->addr_bytes[4] >> 2) |
5270 			(((uint16_t)uc_addr->addr_bytes[5]) << 6));
5271 		break;
5272 	case 3:   /* use bits [43:32] of the address */
5273 		vector = ((uc_addr->addr_bytes[4]) |
5274 			(((uint16_t)uc_addr->addr_bytes[5]) << 8));
5275 		break;
5276 	default:  /* Invalid mc_filter_type */
5277 		break;
5278 	}
5279 
5280 	/* vector can only be 12-bits or boundary will be exceeded */
5281 	vector &= 0xFFF;
5282 	return vector;
5283 }
5284 
5285 static int
5286 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
5287 			uint8_t on)
5288 {
5289 	uint32_t vector;
5290 	uint32_t uta_idx;
5291 	uint32_t reg_val;
5292 	uint32_t uta_shift;
5293 	uint32_t rc;
5294 	const uint32_t ixgbe_uta_idx_mask = 0x7F;
5295 	const uint32_t ixgbe_uta_bit_shift = 5;
5296 	const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
5297 	const uint32_t bit1 = 0x1;
5298 
5299 	struct ixgbe_hw *hw =
5300 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5301 	struct ixgbe_uta_info *uta_info =
5302 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5303 
5304 	/* The UTA table only exists on 82599 hardware and newer */
5305 	if (hw->mac.type < ixgbe_mac_82599EB)
5306 		return -ENOTSUP;
5307 
5308 	vector = ixgbe_uta_vector(hw, mac_addr);
5309 	uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
5310 	uta_shift = vector & ixgbe_uta_bit_mask;
5311 
5312 	rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
5313 	if (rc == on)
5314 		return 0;
5315 
5316 	reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
5317 	if (on) {
5318 		uta_info->uta_in_use++;
5319 		reg_val |= (bit1 << uta_shift);
5320 		uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
5321 	} else {
5322 		uta_info->uta_in_use--;
5323 		reg_val &= ~(bit1 << uta_shift);
5324 		uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
5325 	}
5326 
5327 	IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
5328 
5329 	if (uta_info->uta_in_use > 0)
5330 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
5331 				IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
5332 	else
5333 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
5334 
5335 	return 0;
5336 }
5337 
5338 static int
5339 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
5340 {
5341 	int i;
5342 	struct ixgbe_hw *hw =
5343 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5344 	struct ixgbe_uta_info *uta_info =
5345 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5346 
5347 	/* The UTA table only exists on 82599 hardware and newer */
5348 	if (hw->mac.type < ixgbe_mac_82599EB)
5349 		return -ENOTSUP;
5350 
5351 	if (on) {
5352 		for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5353 			uta_info->uta_shadow[i] = ~0;
5354 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
5355 		}
5356 	} else {
5357 		for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5358 			uta_info->uta_shadow[i] = 0;
5359 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
5360 		}
5361 	}
5362 	return 0;
5363 
5364 }
5365 
5366 uint32_t
5367 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
5368 {
5369 	uint32_t new_val = orig_val;
5370 
5371 	if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG)
5372 		new_val |= IXGBE_VMOLR_AUPE;
5373 	if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC)
5374 		new_val |= IXGBE_VMOLR_ROMPE;
5375 	if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
5376 		new_val |= IXGBE_VMOLR_ROPE;
5377 	if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
5378 		new_val |= IXGBE_VMOLR_BAM;
5379 	if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
5380 		new_val |= IXGBE_VMOLR_MPE;
5381 
5382 	return new_val;
5383 }
5384 
5385 #define IXGBE_MRCTL_VPME  0x01 /* Virtual Pool Mirroring. */
5386 #define IXGBE_MRCTL_UPME  0x02 /* Uplink Port Mirroring. */
5387 #define IXGBE_MRCTL_DPME  0x04 /* Downlink Port Mirroring. */
5388 #define IXGBE_MRCTL_VLME  0x08 /* VLAN Mirroring. */
5389 #define IXGBE_INVALID_MIRROR_TYPE(mirror_type) \
5390 	((mirror_type) & ~(uint8_t)(ETH_MIRROR_VIRTUAL_POOL_UP | \
5391 	ETH_MIRROR_UPLINK_PORT | ETH_MIRROR_DOWNLINK_PORT | ETH_MIRROR_VLAN))
5392 
5393 static int
5394 ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
5395 		      struct rte_eth_mirror_conf *mirror_conf,
5396 		      uint8_t rule_id, uint8_t on)
5397 {
5398 	uint32_t mr_ctl, vlvf;
5399 	uint32_t mp_lsb = 0;
5400 	uint32_t mv_msb = 0;
5401 	uint32_t mv_lsb = 0;
5402 	uint32_t mp_msb = 0;
5403 	uint8_t i = 0;
5404 	int reg_index = 0;
5405 	uint64_t vlan_mask = 0;
5406 
5407 	const uint8_t pool_mask_offset = 32;
5408 	const uint8_t vlan_mask_offset = 32;
5409 	const uint8_t dst_pool_offset = 8;
5410 	const uint8_t rule_mr_offset  = 4;
5411 	const uint8_t mirror_rule_mask = 0x0F;
5412 
5413 	struct ixgbe_mirror_info *mr_info =
5414 			(IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5415 	struct ixgbe_hw *hw =
5416 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5417 	uint8_t mirror_type = 0;
5418 
5419 	if (ixgbe_vt_check(hw) < 0)
5420 		return -ENOTSUP;
5421 
5422 	if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5423 		return -EINVAL;
5424 
5425 	if (IXGBE_INVALID_MIRROR_TYPE(mirror_conf->rule_type)) {
5426 		PMD_DRV_LOG(ERR, "unsupported mirror type 0x%x.",
5427 			    mirror_conf->rule_type);
5428 		return -EINVAL;
5429 	}
5430 
5431 	if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5432 		mirror_type |= IXGBE_MRCTL_VLME;
5433 		/* Check if vlan id is valid and find conresponding VLAN ID
5434 		 * index in VLVF
5435 		 */
5436 		for (i = 0; i < IXGBE_VLVF_ENTRIES; i++) {
5437 			if (mirror_conf->vlan.vlan_mask & (1ULL << i)) {
5438 				/* search vlan id related pool vlan filter
5439 				 * index
5440 				 */
5441 				reg_index = ixgbe_find_vlvf_slot(
5442 						hw,
5443 						mirror_conf->vlan.vlan_id[i],
5444 						false);
5445 				if (reg_index < 0)
5446 					return -EINVAL;
5447 				vlvf = IXGBE_READ_REG(hw,
5448 						      IXGBE_VLVF(reg_index));
5449 				if ((vlvf & IXGBE_VLVF_VIEN) &&
5450 				    ((vlvf & IXGBE_VLVF_VLANID_MASK) ==
5451 				      mirror_conf->vlan.vlan_id[i]))
5452 					vlan_mask |= (1ULL << reg_index);
5453 				else
5454 					return -EINVAL;
5455 			}
5456 		}
5457 
5458 		if (on) {
5459 			mv_lsb = vlan_mask & 0xFFFFFFFF;
5460 			mv_msb = vlan_mask >> vlan_mask_offset;
5461 
5462 			mr_info->mr_conf[rule_id].vlan.vlan_mask =
5463 						mirror_conf->vlan.vlan_mask;
5464 			for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++) {
5465 				if (mirror_conf->vlan.vlan_mask & (1ULL << i))
5466 					mr_info->mr_conf[rule_id].vlan.vlan_id[i] =
5467 						mirror_conf->vlan.vlan_id[i];
5468 			}
5469 		} else {
5470 			mv_lsb = 0;
5471 			mv_msb = 0;
5472 			mr_info->mr_conf[rule_id].vlan.vlan_mask = 0;
5473 			for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++)
5474 				mr_info->mr_conf[rule_id].vlan.vlan_id[i] = 0;
5475 		}
5476 	}
5477 
5478 	/**
5479 	 * if enable pool mirror, write related pool mask register,if disable
5480 	 * pool mirror, clear PFMRVM register
5481 	 */
5482 	if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5483 		mirror_type |= IXGBE_MRCTL_VPME;
5484 		if (on) {
5485 			mp_lsb = mirror_conf->pool_mask & 0xFFFFFFFF;
5486 			mp_msb = mirror_conf->pool_mask >> pool_mask_offset;
5487 			mr_info->mr_conf[rule_id].pool_mask =
5488 					mirror_conf->pool_mask;
5489 
5490 		} else {
5491 			mp_lsb = 0;
5492 			mp_msb = 0;
5493 			mr_info->mr_conf[rule_id].pool_mask = 0;
5494 		}
5495 	}
5496 	if (mirror_conf->rule_type & ETH_MIRROR_UPLINK_PORT)
5497 		mirror_type |= IXGBE_MRCTL_UPME;
5498 	if (mirror_conf->rule_type & ETH_MIRROR_DOWNLINK_PORT)
5499 		mirror_type |= IXGBE_MRCTL_DPME;
5500 
5501 	/* read  mirror control register and recalculate it */
5502 	mr_ctl = IXGBE_READ_REG(hw, IXGBE_MRCTL(rule_id));
5503 
5504 	if (on) {
5505 		mr_ctl |= mirror_type;
5506 		mr_ctl &= mirror_rule_mask;
5507 		mr_ctl |= mirror_conf->dst_pool << dst_pool_offset;
5508 	} else {
5509 		mr_ctl &= ~(mirror_conf->rule_type & mirror_rule_mask);
5510 	}
5511 
5512 	mr_info->mr_conf[rule_id].rule_type = mirror_conf->rule_type;
5513 	mr_info->mr_conf[rule_id].dst_pool = mirror_conf->dst_pool;
5514 
5515 	/* write mirrror control  register */
5516 	IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5517 
5518 	/* write pool mirrror control  register */
5519 	if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5520 		IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), mp_lsb);
5521 		IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset),
5522 				mp_msb);
5523 	}
5524 	/* write VLAN mirrror control  register */
5525 	if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5526 		IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), mv_lsb);
5527 		IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset),
5528 				mv_msb);
5529 	}
5530 
5531 	return 0;
5532 }
5533 
5534 static int
5535 ixgbe_mirror_rule_reset(struct rte_eth_dev *dev, uint8_t rule_id)
5536 {
5537 	int mr_ctl = 0;
5538 	uint32_t lsb_val = 0;
5539 	uint32_t msb_val = 0;
5540 	const uint8_t rule_mr_offset = 4;
5541 
5542 	struct ixgbe_hw *hw =
5543 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5544 	struct ixgbe_mirror_info *mr_info =
5545 		(IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5546 
5547 	if (ixgbe_vt_check(hw) < 0)
5548 		return -ENOTSUP;
5549 
5550 	if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5551 		return -EINVAL;
5552 
5553 	memset(&mr_info->mr_conf[rule_id], 0,
5554 	       sizeof(struct rte_eth_mirror_conf));
5555 
5556 	/* clear PFVMCTL register */
5557 	IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5558 
5559 	/* clear pool mask register */
5560 	IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), lsb_val);
5561 	IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset), msb_val);
5562 
5563 	/* clear vlan mask register */
5564 	IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), lsb_val);
5565 	IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset), msb_val);
5566 
5567 	return 0;
5568 }
5569 
5570 static int
5571 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5572 {
5573 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5574 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5575 	uint32_t mask;
5576 	struct ixgbe_hw *hw =
5577 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5578 	uint32_t vec = IXGBE_MISC_VEC_ID;
5579 
5580 	mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
5581 	if (rte_intr_allow_others(intr_handle))
5582 		vec = IXGBE_RX_VEC_START;
5583 	mask |= (1 << vec);
5584 	RTE_SET_USED(queue_id);
5585 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
5586 
5587 	rte_intr_enable(intr_handle);
5588 
5589 	return 0;
5590 }
5591 
5592 static int
5593 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5594 {
5595 	uint32_t mask;
5596 	struct ixgbe_hw *hw =
5597 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5598 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5599 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5600 	uint32_t vec = IXGBE_MISC_VEC_ID;
5601 
5602 	mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
5603 	if (rte_intr_allow_others(intr_handle))
5604 		vec = IXGBE_RX_VEC_START;
5605 	mask &= ~(1 << vec);
5606 	RTE_SET_USED(queue_id);
5607 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
5608 
5609 	return 0;
5610 }
5611 
5612 static int
5613 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5614 {
5615 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5616 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5617 	uint32_t mask;
5618 	struct ixgbe_hw *hw =
5619 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5620 	struct ixgbe_interrupt *intr =
5621 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5622 
5623 	if (queue_id < 16) {
5624 		ixgbe_disable_intr(hw);
5625 		intr->mask |= (1 << queue_id);
5626 		ixgbe_enable_intr(dev);
5627 	} else if (queue_id < 32) {
5628 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5629 		mask &= (1 << queue_id);
5630 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5631 	} else if (queue_id < 64) {
5632 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5633 		mask &= (1 << (queue_id - 32));
5634 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5635 	}
5636 	rte_intr_enable(intr_handle);
5637 
5638 	return 0;
5639 }
5640 
5641 static int
5642 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5643 {
5644 	uint32_t mask;
5645 	struct ixgbe_hw *hw =
5646 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5647 	struct ixgbe_interrupt *intr =
5648 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5649 
5650 	if (queue_id < 16) {
5651 		ixgbe_disable_intr(hw);
5652 		intr->mask &= ~(1 << queue_id);
5653 		ixgbe_enable_intr(dev);
5654 	} else if (queue_id < 32) {
5655 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5656 		mask &= ~(1 << queue_id);
5657 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5658 	} else if (queue_id < 64) {
5659 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5660 		mask &= ~(1 << (queue_id - 32));
5661 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5662 	}
5663 
5664 	return 0;
5665 }
5666 
5667 static void
5668 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5669 		     uint8_t queue, uint8_t msix_vector)
5670 {
5671 	uint32_t tmp, idx;
5672 
5673 	if (direction == -1) {
5674 		/* other causes */
5675 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5676 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
5677 		tmp &= ~0xFF;
5678 		tmp |= msix_vector;
5679 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
5680 	} else {
5681 		/* rx or tx cause */
5682 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5683 		idx = ((16 * (queue & 1)) + (8 * direction));
5684 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
5685 		tmp &= ~(0xFF << idx);
5686 		tmp |= (msix_vector << idx);
5687 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
5688 	}
5689 }
5690 
5691 /**
5692  * set the IVAR registers, mapping interrupt causes to vectors
5693  * @param hw
5694  *  pointer to ixgbe_hw struct
5695  * @direction
5696  *  0 for Rx, 1 for Tx, -1 for other causes
5697  * @queue
5698  *  queue to map the corresponding interrupt to
5699  * @msix_vector
5700  *  the vector to map to the corresponding queue
5701  */
5702 static void
5703 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5704 		   uint8_t queue, uint8_t msix_vector)
5705 {
5706 	uint32_t tmp, idx;
5707 
5708 	msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5709 	if (hw->mac.type == ixgbe_mac_82598EB) {
5710 		if (direction == -1)
5711 			direction = 0;
5712 		idx = (((direction * 64) + queue) >> 2) & 0x1F;
5713 		tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
5714 		tmp &= ~(0xFF << (8 * (queue & 0x3)));
5715 		tmp |= (msix_vector << (8 * (queue & 0x3)));
5716 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
5717 	} else if ((hw->mac.type == ixgbe_mac_82599EB) ||
5718 			(hw->mac.type == ixgbe_mac_X540) ||
5719 			(hw->mac.type == ixgbe_mac_X550)) {
5720 		if (direction == -1) {
5721 			/* other causes */
5722 			idx = ((queue & 1) * 8);
5723 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
5724 			tmp &= ~(0xFF << idx);
5725 			tmp |= (msix_vector << idx);
5726 			IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
5727 		} else {
5728 			/* rx or tx causes */
5729 			idx = ((16 * (queue & 1)) + (8 * direction));
5730 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
5731 			tmp &= ~(0xFF << idx);
5732 			tmp |= (msix_vector << idx);
5733 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
5734 		}
5735 	}
5736 }
5737 
5738 static void
5739 ixgbevf_configure_msix(struct rte_eth_dev *dev)
5740 {
5741 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5742 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5743 	struct ixgbe_hw *hw =
5744 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5745 	uint32_t q_idx;
5746 	uint32_t vector_idx = IXGBE_MISC_VEC_ID;
5747 	uint32_t base = IXGBE_MISC_VEC_ID;
5748 
5749 	/* Configure VF other cause ivar */
5750 	ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
5751 
5752 	/* won't configure msix register if no mapping is done
5753 	 * between intr vector and event fd.
5754 	 */
5755 	if (!rte_intr_dp_is_en(intr_handle))
5756 		return;
5757 
5758 	if (rte_intr_allow_others(intr_handle)) {
5759 		base = IXGBE_RX_VEC_START;
5760 		vector_idx = IXGBE_RX_VEC_START;
5761 	}
5762 
5763 	/* Configure all RX queues of VF */
5764 	for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
5765 		/* Force all queue use vector 0,
5766 		 * as IXGBE_VF_MAXMSIVECOTR = 1
5767 		 */
5768 		ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
5769 		intr_handle->intr_vec[q_idx] = vector_idx;
5770 		if (vector_idx < base + intr_handle->nb_efd - 1)
5771 			vector_idx++;
5772 	}
5773 }
5774 
5775 /**
5776  * Sets up the hardware to properly generate MSI-X interrupts
5777  * @hw
5778  *  board private structure
5779  */
5780 static void
5781 ixgbe_configure_msix(struct rte_eth_dev *dev)
5782 {
5783 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5784 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5785 	struct ixgbe_hw *hw =
5786 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5787 	uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
5788 	uint32_t vec = IXGBE_MISC_VEC_ID;
5789 	uint32_t mask;
5790 	uint32_t gpie;
5791 
5792 	/* won't configure msix register if no mapping is done
5793 	 * between intr vector and event fd
5794 	 */
5795 	if (!rte_intr_dp_is_en(intr_handle))
5796 		return;
5797 
5798 	if (rte_intr_allow_others(intr_handle))
5799 		vec = base = IXGBE_RX_VEC_START;
5800 
5801 	/* setup GPIE for MSI-x mode */
5802 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5803 	gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5804 		IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
5805 	/* auto clearing and auto setting corresponding bits in EIMS
5806 	 * when MSI-X interrupt is triggered
5807 	 */
5808 	if (hw->mac.type == ixgbe_mac_82598EB) {
5809 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5810 	} else {
5811 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5812 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5813 	}
5814 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5815 
5816 	/* Populate the IVAR table and set the ITR values to the
5817 	 * corresponding register.
5818 	 */
5819 	for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
5820 	     queue_id++) {
5821 		/* by default, 1:1 mapping */
5822 		ixgbe_set_ivar_map(hw, 0, queue_id, vec);
5823 		intr_handle->intr_vec[queue_id] = vec;
5824 		if (vec < base + intr_handle->nb_efd - 1)
5825 			vec++;
5826 	}
5827 
5828 	switch (hw->mac.type) {
5829 	case ixgbe_mac_82598EB:
5830 		ixgbe_set_ivar_map(hw, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
5831 				   IXGBE_MISC_VEC_ID);
5832 		break;
5833 	case ixgbe_mac_82599EB:
5834 	case ixgbe_mac_X540:
5835 	case ixgbe_mac_X550:
5836 		ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
5837 		break;
5838 	default:
5839 		break;
5840 	}
5841 	IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
5842 			IXGBE_MIN_INTER_INTERRUPT_INTERVAL_DEFAULT & 0xFFF);
5843 
5844 	/* set up to autoclear timer, and the vectors */
5845 	mask = IXGBE_EIMS_ENABLE_MASK;
5846 	mask &= ~(IXGBE_EIMS_OTHER |
5847 		  IXGBE_EIMS_MAILBOX |
5848 		  IXGBE_EIMS_LSC);
5849 
5850 	IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
5851 }
5852 
5853 int
5854 ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
5855 			   uint16_t queue_idx, uint16_t tx_rate)
5856 {
5857 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5858 	uint32_t rf_dec, rf_int;
5859 	uint32_t bcnrc_val;
5860 	uint16_t link_speed = dev->data->dev_link.link_speed;
5861 
5862 	if (queue_idx >= hw->mac.max_tx_queues)
5863 		return -EINVAL;
5864 
5865 	if (tx_rate != 0) {
5866 		/* Calculate the rate factor values to set */
5867 		rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
5868 		rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
5869 		rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
5870 
5871 		bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
5872 		bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
5873 				IXGBE_RTTBCNRC_RF_INT_MASK_M);
5874 		bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
5875 	} else {
5876 		bcnrc_val = 0;
5877 	}
5878 
5879 	/*
5880 	 * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
5881 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
5882 	 * set as 0x4.
5883 	 */
5884 	if ((dev->data->dev_conf.rxmode.jumbo_frame == 1) &&
5885 		(dev->data->dev_conf.rxmode.max_rx_pkt_len >=
5886 				IXGBE_MAX_JUMBO_FRAME_SIZE))
5887 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
5888 			IXGBE_MMW_SIZE_JUMBO_FRAME);
5889 	else
5890 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
5891 			IXGBE_MMW_SIZE_DEFAULT);
5892 
5893 	/* Set RTTBCNRC of queue X */
5894 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
5895 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
5896 	IXGBE_WRITE_FLUSH(hw);
5897 
5898 	return 0;
5899 }
5900 
5901 static int
5902 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
5903 		     __attribute__((unused)) uint32_t index,
5904 		     __attribute__((unused)) uint32_t pool)
5905 {
5906 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5907 	int diag;
5908 
5909 	/*
5910 	 * On a 82599 VF, adding again the same MAC addr is not an idempotent
5911 	 * operation. Trap this case to avoid exhausting the [very limited]
5912 	 * set of PF resources used to store VF MAC addresses.
5913 	 */
5914 	if (memcmp(hw->mac.perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
5915 		return -1;
5916 	diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
5917 	if (diag != 0)
5918 		PMD_DRV_LOG(ERR, "Unable to add MAC address "
5919 			    "%02x:%02x:%02x:%02x:%02x:%02x - diag=%d",
5920 			    mac_addr->addr_bytes[0],
5921 			    mac_addr->addr_bytes[1],
5922 			    mac_addr->addr_bytes[2],
5923 			    mac_addr->addr_bytes[3],
5924 			    mac_addr->addr_bytes[4],
5925 			    mac_addr->addr_bytes[5],
5926 			    diag);
5927 	return diag;
5928 }
5929 
5930 static void
5931 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
5932 {
5933 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5934 	struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
5935 	struct ether_addr *mac_addr;
5936 	uint32_t i;
5937 	int diag;
5938 
5939 	/*
5940 	 * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
5941 	 * not support the deletion of a given MAC address.
5942 	 * Instead, it imposes to delete all MAC addresses, then to add again
5943 	 * all MAC addresses with the exception of the one to be deleted.
5944 	 */
5945 	(void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
5946 
5947 	/*
5948 	 * Add again all MAC addresses, with the exception of the deleted one
5949 	 * and of the permanent MAC address.
5950 	 */
5951 	for (i = 0, mac_addr = dev->data->mac_addrs;
5952 	     i < hw->mac.num_rar_entries; i++, mac_addr++) {
5953 		/* Skip the deleted MAC address */
5954 		if (i == index)
5955 			continue;
5956 		/* Skip NULL MAC addresses */
5957 		if (is_zero_ether_addr(mac_addr))
5958 			continue;
5959 		/* Skip the permanent MAC address */
5960 		if (memcmp(perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
5961 			continue;
5962 		diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
5963 		if (diag != 0)
5964 			PMD_DRV_LOG(ERR,
5965 				    "Adding again MAC address "
5966 				    "%02x:%02x:%02x:%02x:%02x:%02x failed "
5967 				    "diag=%d",
5968 				    mac_addr->addr_bytes[0],
5969 				    mac_addr->addr_bytes[1],
5970 				    mac_addr->addr_bytes[2],
5971 				    mac_addr->addr_bytes[3],
5972 				    mac_addr->addr_bytes[4],
5973 				    mac_addr->addr_bytes[5],
5974 				    diag);
5975 	}
5976 }
5977 
5978 static void
5979 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
5980 {
5981 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5982 
5983 	hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
5984 }
5985 
5986 int
5987 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
5988 			struct rte_eth_syn_filter *filter,
5989 			bool add)
5990 {
5991 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5992 	struct ixgbe_filter_info *filter_info =
5993 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5994 	uint32_t syn_info;
5995 	uint32_t synqf;
5996 
5997 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
5998 		return -EINVAL;
5999 
6000 	syn_info = filter_info->syn_info;
6001 
6002 	if (add) {
6003 		if (syn_info & IXGBE_SYN_FILTER_ENABLE)
6004 			return -EINVAL;
6005 		synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
6006 			IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
6007 
6008 		if (filter->hig_pri)
6009 			synqf |= IXGBE_SYN_FILTER_SYNQFP;
6010 		else
6011 			synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
6012 	} else {
6013 		synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6014 		if (!(syn_info & IXGBE_SYN_FILTER_ENABLE))
6015 			return -ENOENT;
6016 		synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
6017 	}
6018 
6019 	filter_info->syn_info = synqf;
6020 	IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
6021 	IXGBE_WRITE_FLUSH(hw);
6022 	return 0;
6023 }
6024 
6025 static int
6026 ixgbe_syn_filter_get(struct rte_eth_dev *dev,
6027 			struct rte_eth_syn_filter *filter)
6028 {
6029 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6030 	uint32_t synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6031 
6032 	if (synqf & IXGBE_SYN_FILTER_ENABLE) {
6033 		filter->hig_pri = (synqf & IXGBE_SYN_FILTER_SYNQFP) ? 1 : 0;
6034 		filter->queue = (uint16_t)((synqf & IXGBE_SYN_FILTER_QUEUE) >> 1);
6035 		return 0;
6036 	}
6037 	return -ENOENT;
6038 }
6039 
6040 static int
6041 ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
6042 			enum rte_filter_op filter_op,
6043 			void *arg)
6044 {
6045 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6046 	int ret;
6047 
6048 	MAC_TYPE_FILTER_SUP(hw->mac.type);
6049 
6050 	if (filter_op == RTE_ETH_FILTER_NOP)
6051 		return 0;
6052 
6053 	if (arg == NULL) {
6054 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
6055 			    filter_op);
6056 		return -EINVAL;
6057 	}
6058 
6059 	switch (filter_op) {
6060 	case RTE_ETH_FILTER_ADD:
6061 		ret = ixgbe_syn_filter_set(dev,
6062 				(struct rte_eth_syn_filter *)arg,
6063 				TRUE);
6064 		break;
6065 	case RTE_ETH_FILTER_DELETE:
6066 		ret = ixgbe_syn_filter_set(dev,
6067 				(struct rte_eth_syn_filter *)arg,
6068 				FALSE);
6069 		break;
6070 	case RTE_ETH_FILTER_GET:
6071 		ret = ixgbe_syn_filter_get(dev,
6072 				(struct rte_eth_syn_filter *)arg);
6073 		break;
6074 	default:
6075 		PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
6076 		ret = -EINVAL;
6077 		break;
6078 	}
6079 
6080 	return ret;
6081 }
6082 
6083 
6084 static inline enum ixgbe_5tuple_protocol
6085 convert_protocol_type(uint8_t protocol_value)
6086 {
6087 	if (protocol_value == IPPROTO_TCP)
6088 		return IXGBE_FILTER_PROTOCOL_TCP;
6089 	else if (protocol_value == IPPROTO_UDP)
6090 		return IXGBE_FILTER_PROTOCOL_UDP;
6091 	else if (protocol_value == IPPROTO_SCTP)
6092 		return IXGBE_FILTER_PROTOCOL_SCTP;
6093 	else
6094 		return IXGBE_FILTER_PROTOCOL_NONE;
6095 }
6096 
6097 /* inject a 5-tuple filter to HW */
6098 static inline void
6099 ixgbe_inject_5tuple_filter(struct rte_eth_dev *dev,
6100 			   struct ixgbe_5tuple_filter *filter)
6101 {
6102 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6103 	int i;
6104 	uint32_t ftqf, sdpqf;
6105 	uint32_t l34timir = 0;
6106 	uint8_t mask = 0xff;
6107 
6108 	i = filter->index;
6109 
6110 	sdpqf = (uint32_t)(filter->filter_info.dst_port <<
6111 				IXGBE_SDPQF_DSTPORT_SHIFT);
6112 	sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
6113 
6114 	ftqf = (uint32_t)(filter->filter_info.proto &
6115 		IXGBE_FTQF_PROTOCOL_MASK);
6116 	ftqf |= (uint32_t)((filter->filter_info.priority &
6117 		IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
6118 	if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
6119 		mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
6120 	if (filter->filter_info.dst_ip_mask == 0)
6121 		mask &= IXGBE_FTQF_DEST_ADDR_MASK;
6122 	if (filter->filter_info.src_port_mask == 0)
6123 		mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
6124 	if (filter->filter_info.dst_port_mask == 0)
6125 		mask &= IXGBE_FTQF_DEST_PORT_MASK;
6126 	if (filter->filter_info.proto_mask == 0)
6127 		mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
6128 	ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
6129 	ftqf |= IXGBE_FTQF_POOL_MASK_EN;
6130 	ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
6131 
6132 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
6133 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
6134 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
6135 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
6136 
6137 	l34timir |= IXGBE_L34T_IMIR_RESERVE;
6138 	l34timir |= (uint32_t)(filter->queue <<
6139 				IXGBE_L34T_IMIR_QUEUE_SHIFT);
6140 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
6141 }
6142 
6143 /*
6144  * add a 5tuple filter
6145  *
6146  * @param
6147  * dev: Pointer to struct rte_eth_dev.
6148  * index: the index the filter allocates.
6149  * filter: ponter to the filter that will be added.
6150  * rx_queue: the queue id the filter assigned to.
6151  *
6152  * @return
6153  *    - On success, zero.
6154  *    - On failure, a negative value.
6155  */
6156 static int
6157 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
6158 			struct ixgbe_5tuple_filter *filter)
6159 {
6160 	struct ixgbe_filter_info *filter_info =
6161 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6162 	int i, idx, shift;
6163 
6164 	/*
6165 	 * look for an unused 5tuple filter index,
6166 	 * and insert the filter to list.
6167 	 */
6168 	for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
6169 		idx = i / (sizeof(uint32_t) * NBBY);
6170 		shift = i % (sizeof(uint32_t) * NBBY);
6171 		if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
6172 			filter_info->fivetuple_mask[idx] |= 1 << shift;
6173 			filter->index = i;
6174 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
6175 					  filter,
6176 					  entries);
6177 			break;
6178 		}
6179 	}
6180 	if (i >= IXGBE_MAX_FTQF_FILTERS) {
6181 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
6182 		return -ENOSYS;
6183 	}
6184 
6185 	ixgbe_inject_5tuple_filter(dev, filter);
6186 
6187 	return 0;
6188 }
6189 
6190 /*
6191  * remove a 5tuple filter
6192  *
6193  * @param
6194  * dev: Pointer to struct rte_eth_dev.
6195  * filter: the pointer of the filter will be removed.
6196  */
6197 static void
6198 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
6199 			struct ixgbe_5tuple_filter *filter)
6200 {
6201 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6202 	struct ixgbe_filter_info *filter_info =
6203 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6204 	uint16_t index = filter->index;
6205 
6206 	filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
6207 				~(1 << (index % (sizeof(uint32_t) * NBBY)));
6208 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
6209 	rte_free(filter);
6210 
6211 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
6212 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
6213 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
6214 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
6215 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
6216 }
6217 
6218 static int
6219 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
6220 {
6221 	struct ixgbe_hw *hw;
6222 	uint32_t max_frame = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
6223 	struct rte_eth_rxmode *rx_conf = &dev->data->dev_conf.rxmode;
6224 
6225 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6226 
6227 	if ((mtu < ETHER_MIN_MTU) || (max_frame > ETHER_MAX_JUMBO_FRAME_LEN))
6228 		return -EINVAL;
6229 
6230 	/* refuse mtu that requires the support of scattered packets when this
6231 	 * feature has not been enabled before.
6232 	 */
6233 	if (!rx_conf->enable_scatter &&
6234 	    (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
6235 	     dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM))
6236 		return -EINVAL;
6237 
6238 	/*
6239 	 * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
6240 	 * request of the version 2.0 of the mailbox API.
6241 	 * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
6242 	 * of the mailbox API.
6243 	 * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
6244 	 * prior to 3.11.33 which contains the following change:
6245 	 * "ixgbe: Enable jumbo frames support w/ SR-IOV"
6246 	 */
6247 	ixgbevf_rlpml_set_vf(hw, max_frame);
6248 
6249 	/* update max frame size */
6250 	dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame;
6251 	return 0;
6252 }
6253 
6254 static inline struct ixgbe_5tuple_filter *
6255 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
6256 			struct ixgbe_5tuple_filter_info *key)
6257 {
6258 	struct ixgbe_5tuple_filter *it;
6259 
6260 	TAILQ_FOREACH(it, filter_list, entries) {
6261 		if (memcmp(key, &it->filter_info,
6262 			sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
6263 			return it;
6264 		}
6265 	}
6266 	return NULL;
6267 }
6268 
6269 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
6270 static inline int
6271 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
6272 			struct ixgbe_5tuple_filter_info *filter_info)
6273 {
6274 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
6275 		filter->priority > IXGBE_5TUPLE_MAX_PRI ||
6276 		filter->priority < IXGBE_5TUPLE_MIN_PRI)
6277 		return -EINVAL;
6278 
6279 	switch (filter->dst_ip_mask) {
6280 	case UINT32_MAX:
6281 		filter_info->dst_ip_mask = 0;
6282 		filter_info->dst_ip = filter->dst_ip;
6283 		break;
6284 	case 0:
6285 		filter_info->dst_ip_mask = 1;
6286 		break;
6287 	default:
6288 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
6289 		return -EINVAL;
6290 	}
6291 
6292 	switch (filter->src_ip_mask) {
6293 	case UINT32_MAX:
6294 		filter_info->src_ip_mask = 0;
6295 		filter_info->src_ip = filter->src_ip;
6296 		break;
6297 	case 0:
6298 		filter_info->src_ip_mask = 1;
6299 		break;
6300 	default:
6301 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
6302 		return -EINVAL;
6303 	}
6304 
6305 	switch (filter->dst_port_mask) {
6306 	case UINT16_MAX:
6307 		filter_info->dst_port_mask = 0;
6308 		filter_info->dst_port = filter->dst_port;
6309 		break;
6310 	case 0:
6311 		filter_info->dst_port_mask = 1;
6312 		break;
6313 	default:
6314 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
6315 		return -EINVAL;
6316 	}
6317 
6318 	switch (filter->src_port_mask) {
6319 	case UINT16_MAX:
6320 		filter_info->src_port_mask = 0;
6321 		filter_info->src_port = filter->src_port;
6322 		break;
6323 	case 0:
6324 		filter_info->src_port_mask = 1;
6325 		break;
6326 	default:
6327 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
6328 		return -EINVAL;
6329 	}
6330 
6331 	switch (filter->proto_mask) {
6332 	case UINT8_MAX:
6333 		filter_info->proto_mask = 0;
6334 		filter_info->proto =
6335 			convert_protocol_type(filter->proto);
6336 		break;
6337 	case 0:
6338 		filter_info->proto_mask = 1;
6339 		break;
6340 	default:
6341 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
6342 		return -EINVAL;
6343 	}
6344 
6345 	filter_info->priority = (uint8_t)filter->priority;
6346 	return 0;
6347 }
6348 
6349 /*
6350  * add or delete a ntuple filter
6351  *
6352  * @param
6353  * dev: Pointer to struct rte_eth_dev.
6354  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6355  * add: if true, add filter, if false, remove filter
6356  *
6357  * @return
6358  *    - On success, zero.
6359  *    - On failure, a negative value.
6360  */
6361 int
6362 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
6363 			struct rte_eth_ntuple_filter *ntuple_filter,
6364 			bool add)
6365 {
6366 	struct ixgbe_filter_info *filter_info =
6367 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6368 	struct ixgbe_5tuple_filter_info filter_5tuple;
6369 	struct ixgbe_5tuple_filter *filter;
6370 	int ret;
6371 
6372 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6373 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6374 		return -EINVAL;
6375 	}
6376 
6377 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6378 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6379 	if (ret < 0)
6380 		return ret;
6381 
6382 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6383 					 &filter_5tuple);
6384 	if (filter != NULL && add) {
6385 		PMD_DRV_LOG(ERR, "filter exists.");
6386 		return -EEXIST;
6387 	}
6388 	if (filter == NULL && !add) {
6389 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
6390 		return -ENOENT;
6391 	}
6392 
6393 	if (add) {
6394 		filter = rte_zmalloc("ixgbe_5tuple_filter",
6395 				sizeof(struct ixgbe_5tuple_filter), 0);
6396 		if (filter == NULL)
6397 			return -ENOMEM;
6398 		rte_memcpy(&filter->filter_info,
6399 				 &filter_5tuple,
6400 				 sizeof(struct ixgbe_5tuple_filter_info));
6401 		filter->queue = ntuple_filter->queue;
6402 		ret = ixgbe_add_5tuple_filter(dev, filter);
6403 		if (ret < 0) {
6404 			rte_free(filter);
6405 			return ret;
6406 		}
6407 	} else
6408 		ixgbe_remove_5tuple_filter(dev, filter);
6409 
6410 	return 0;
6411 }
6412 
6413 /*
6414  * get a ntuple filter
6415  *
6416  * @param
6417  * dev: Pointer to struct rte_eth_dev.
6418  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6419  *
6420  * @return
6421  *    - On success, zero.
6422  *    - On failure, a negative value.
6423  */
6424 static int
6425 ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
6426 			struct rte_eth_ntuple_filter *ntuple_filter)
6427 {
6428 	struct ixgbe_filter_info *filter_info =
6429 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6430 	struct ixgbe_5tuple_filter_info filter_5tuple;
6431 	struct ixgbe_5tuple_filter *filter;
6432 	int ret;
6433 
6434 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6435 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6436 		return -EINVAL;
6437 	}
6438 
6439 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6440 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6441 	if (ret < 0)
6442 		return ret;
6443 
6444 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6445 					 &filter_5tuple);
6446 	if (filter == NULL) {
6447 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
6448 		return -ENOENT;
6449 	}
6450 	ntuple_filter->queue = filter->queue;
6451 	return 0;
6452 }
6453 
6454 /*
6455  * ixgbe_ntuple_filter_handle - Handle operations for ntuple filter.
6456  * @dev: pointer to rte_eth_dev structure
6457  * @filter_op:operation will be taken.
6458  * @arg: a pointer to specific structure corresponding to the filter_op
6459  *
6460  * @return
6461  *    - On success, zero.
6462  *    - On failure, a negative value.
6463  */
6464 static int
6465 ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
6466 				enum rte_filter_op filter_op,
6467 				void *arg)
6468 {
6469 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6470 	int ret;
6471 
6472 	MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
6473 
6474 	if (filter_op == RTE_ETH_FILTER_NOP)
6475 		return 0;
6476 
6477 	if (arg == NULL) {
6478 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
6479 			    filter_op);
6480 		return -EINVAL;
6481 	}
6482 
6483 	switch (filter_op) {
6484 	case RTE_ETH_FILTER_ADD:
6485 		ret = ixgbe_add_del_ntuple_filter(dev,
6486 			(struct rte_eth_ntuple_filter *)arg,
6487 			TRUE);
6488 		break;
6489 	case RTE_ETH_FILTER_DELETE:
6490 		ret = ixgbe_add_del_ntuple_filter(dev,
6491 			(struct rte_eth_ntuple_filter *)arg,
6492 			FALSE);
6493 		break;
6494 	case RTE_ETH_FILTER_GET:
6495 		ret = ixgbe_get_ntuple_filter(dev,
6496 			(struct rte_eth_ntuple_filter *)arg);
6497 		break;
6498 	default:
6499 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
6500 		ret = -EINVAL;
6501 		break;
6502 	}
6503 	return ret;
6504 }
6505 
6506 int
6507 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
6508 			struct rte_eth_ethertype_filter *filter,
6509 			bool add)
6510 {
6511 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6512 	struct ixgbe_filter_info *filter_info =
6513 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6514 	uint32_t etqf = 0;
6515 	uint32_t etqs = 0;
6516 	int ret;
6517 	struct ixgbe_ethertype_filter ethertype_filter;
6518 
6519 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6520 		return -EINVAL;
6521 
6522 	if (filter->ether_type == ETHER_TYPE_IPv4 ||
6523 		filter->ether_type == ETHER_TYPE_IPv6) {
6524 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
6525 			" ethertype filter.", filter->ether_type);
6526 		return -EINVAL;
6527 	}
6528 
6529 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
6530 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
6531 		return -EINVAL;
6532 	}
6533 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
6534 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
6535 		return -EINVAL;
6536 	}
6537 
6538 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6539 	if (ret >= 0 && add) {
6540 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
6541 			    filter->ether_type);
6542 		return -EEXIST;
6543 	}
6544 	if (ret < 0 && !add) {
6545 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6546 			    filter->ether_type);
6547 		return -ENOENT;
6548 	}
6549 
6550 	if (add) {
6551 		etqf = IXGBE_ETQF_FILTER_EN;
6552 		etqf |= (uint32_t)filter->ether_type;
6553 		etqs |= (uint32_t)((filter->queue <<
6554 				    IXGBE_ETQS_RX_QUEUE_SHIFT) &
6555 				    IXGBE_ETQS_RX_QUEUE);
6556 		etqs |= IXGBE_ETQS_QUEUE_EN;
6557 
6558 		ethertype_filter.ethertype = filter->ether_type;
6559 		ethertype_filter.etqf = etqf;
6560 		ethertype_filter.etqs = etqs;
6561 		ethertype_filter.conf = FALSE;
6562 		ret = ixgbe_ethertype_filter_insert(filter_info,
6563 						    &ethertype_filter);
6564 		if (ret < 0) {
6565 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
6566 			return -ENOSPC;
6567 		}
6568 	} else {
6569 		ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
6570 		if (ret < 0)
6571 			return -ENOSYS;
6572 	}
6573 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
6574 	IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
6575 	IXGBE_WRITE_FLUSH(hw);
6576 
6577 	return 0;
6578 }
6579 
6580 static int
6581 ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
6582 			struct rte_eth_ethertype_filter *filter)
6583 {
6584 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6585 	struct ixgbe_filter_info *filter_info =
6586 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6587 	uint32_t etqf, etqs;
6588 	int ret;
6589 
6590 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6591 	if (ret < 0) {
6592 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6593 			    filter->ether_type);
6594 		return -ENOENT;
6595 	}
6596 
6597 	etqf = IXGBE_READ_REG(hw, IXGBE_ETQF(ret));
6598 	if (etqf & IXGBE_ETQF_FILTER_EN) {
6599 		etqs = IXGBE_READ_REG(hw, IXGBE_ETQS(ret));
6600 		filter->ether_type = etqf & IXGBE_ETQF_ETHERTYPE;
6601 		filter->flags = 0;
6602 		filter->queue = (etqs & IXGBE_ETQS_RX_QUEUE) >>
6603 			       IXGBE_ETQS_RX_QUEUE_SHIFT;
6604 		return 0;
6605 	}
6606 	return -ENOENT;
6607 }
6608 
6609 /*
6610  * ixgbe_ethertype_filter_handle - Handle operations for ethertype filter.
6611  * @dev: pointer to rte_eth_dev structure
6612  * @filter_op:operation will be taken.
6613  * @arg: a pointer to specific structure corresponding to the filter_op
6614  */
6615 static int
6616 ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
6617 				enum rte_filter_op filter_op,
6618 				void *arg)
6619 {
6620 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6621 	int ret;
6622 
6623 	MAC_TYPE_FILTER_SUP(hw->mac.type);
6624 
6625 	if (filter_op == RTE_ETH_FILTER_NOP)
6626 		return 0;
6627 
6628 	if (arg == NULL) {
6629 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
6630 			    filter_op);
6631 		return -EINVAL;
6632 	}
6633 
6634 	switch (filter_op) {
6635 	case RTE_ETH_FILTER_ADD:
6636 		ret = ixgbe_add_del_ethertype_filter(dev,
6637 			(struct rte_eth_ethertype_filter *)arg,
6638 			TRUE);
6639 		break;
6640 	case RTE_ETH_FILTER_DELETE:
6641 		ret = ixgbe_add_del_ethertype_filter(dev,
6642 			(struct rte_eth_ethertype_filter *)arg,
6643 			FALSE);
6644 		break;
6645 	case RTE_ETH_FILTER_GET:
6646 		ret = ixgbe_get_ethertype_filter(dev,
6647 			(struct rte_eth_ethertype_filter *)arg);
6648 		break;
6649 	default:
6650 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
6651 		ret = -EINVAL;
6652 		break;
6653 	}
6654 	return ret;
6655 }
6656 
6657 static int
6658 ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
6659 		     enum rte_filter_type filter_type,
6660 		     enum rte_filter_op filter_op,
6661 		     void *arg)
6662 {
6663 	int ret = 0;
6664 
6665 	switch (filter_type) {
6666 	case RTE_ETH_FILTER_NTUPLE:
6667 		ret = ixgbe_ntuple_filter_handle(dev, filter_op, arg);
6668 		break;
6669 	case RTE_ETH_FILTER_ETHERTYPE:
6670 		ret = ixgbe_ethertype_filter_handle(dev, filter_op, arg);
6671 		break;
6672 	case RTE_ETH_FILTER_SYN:
6673 		ret = ixgbe_syn_filter_handle(dev, filter_op, arg);
6674 		break;
6675 	case RTE_ETH_FILTER_FDIR:
6676 		ret = ixgbe_fdir_ctrl_func(dev, filter_op, arg);
6677 		break;
6678 	case RTE_ETH_FILTER_L2_TUNNEL:
6679 		ret = ixgbe_dev_l2_tunnel_filter_handle(dev, filter_op, arg);
6680 		break;
6681 	case RTE_ETH_FILTER_GENERIC:
6682 		if (filter_op != RTE_ETH_FILTER_GET)
6683 			return -EINVAL;
6684 		*(const void **)arg = &ixgbe_flow_ops;
6685 		break;
6686 	default:
6687 		PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
6688 							filter_type);
6689 		ret = -EINVAL;
6690 		break;
6691 	}
6692 
6693 	return ret;
6694 }
6695 
6696 static u8 *
6697 ixgbe_dev_addr_list_itr(__attribute__((unused)) struct ixgbe_hw *hw,
6698 			u8 **mc_addr_ptr, u32 *vmdq)
6699 {
6700 	u8 *mc_addr;
6701 
6702 	*vmdq = 0;
6703 	mc_addr = *mc_addr_ptr;
6704 	*mc_addr_ptr = (mc_addr + sizeof(struct ether_addr));
6705 	return mc_addr;
6706 }
6707 
6708 static int
6709 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
6710 			  struct ether_addr *mc_addr_set,
6711 			  uint32_t nb_mc_addr)
6712 {
6713 	struct ixgbe_hw *hw;
6714 	u8 *mc_addr_list;
6715 
6716 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6717 	mc_addr_list = (u8 *)mc_addr_set;
6718 	return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
6719 					 ixgbe_dev_addr_list_itr, TRUE);
6720 }
6721 
6722 static uint64_t
6723 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
6724 {
6725 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6726 	uint64_t systime_cycles;
6727 
6728 	switch (hw->mac.type) {
6729 	case ixgbe_mac_X550:
6730 	case ixgbe_mac_X550EM_x:
6731 	case ixgbe_mac_X550EM_a:
6732 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
6733 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6734 		systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6735 				* NSEC_PER_SEC;
6736 		break;
6737 	default:
6738 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6739 		systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6740 				<< 32;
6741 	}
6742 
6743 	return systime_cycles;
6744 }
6745 
6746 static uint64_t
6747 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6748 {
6749 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6750 	uint64_t rx_tstamp_cycles;
6751 
6752 	switch (hw->mac.type) {
6753 	case ixgbe_mac_X550:
6754 	case ixgbe_mac_X550EM_x:
6755 	case ixgbe_mac_X550EM_a:
6756 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6757 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6758 		rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6759 				* NSEC_PER_SEC;
6760 		break;
6761 	default:
6762 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6763 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6764 		rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6765 				<< 32;
6766 	}
6767 
6768 	return rx_tstamp_cycles;
6769 }
6770 
6771 static uint64_t
6772 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6773 {
6774 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6775 	uint64_t tx_tstamp_cycles;
6776 
6777 	switch (hw->mac.type) {
6778 	case ixgbe_mac_X550:
6779 	case ixgbe_mac_X550EM_x:
6780 	case ixgbe_mac_X550EM_a:
6781 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6782 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6783 		tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6784 				* NSEC_PER_SEC;
6785 		break;
6786 	default:
6787 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6788 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6789 		tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6790 				<< 32;
6791 	}
6792 
6793 	return tx_tstamp_cycles;
6794 }
6795 
6796 static void
6797 ixgbe_start_timecounters(struct rte_eth_dev *dev)
6798 {
6799 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6800 	struct ixgbe_adapter *adapter =
6801 		(struct ixgbe_adapter *)dev->data->dev_private;
6802 	struct rte_eth_link link;
6803 	uint32_t incval = 0;
6804 	uint32_t shift = 0;
6805 
6806 	/* Get current link speed. */
6807 	memset(&link, 0, sizeof(link));
6808 	ixgbe_dev_link_update(dev, 1);
6809 	rte_ixgbe_dev_atomic_read_link_status(dev, &link);
6810 
6811 	switch (link.link_speed) {
6812 	case ETH_SPEED_NUM_100M:
6813 		incval = IXGBE_INCVAL_100;
6814 		shift = IXGBE_INCVAL_SHIFT_100;
6815 		break;
6816 	case ETH_SPEED_NUM_1G:
6817 		incval = IXGBE_INCVAL_1GB;
6818 		shift = IXGBE_INCVAL_SHIFT_1GB;
6819 		break;
6820 	case ETH_SPEED_NUM_10G:
6821 	default:
6822 		incval = IXGBE_INCVAL_10GB;
6823 		shift = IXGBE_INCVAL_SHIFT_10GB;
6824 		break;
6825 	}
6826 
6827 	switch (hw->mac.type) {
6828 	case ixgbe_mac_X550:
6829 	case ixgbe_mac_X550EM_x:
6830 	case ixgbe_mac_X550EM_a:
6831 		/* Independent of link speed. */
6832 		incval = 1;
6833 		/* Cycles read will be interpreted as ns. */
6834 		shift = 0;
6835 		/* Fall-through */
6836 	case ixgbe_mac_X540:
6837 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
6838 		break;
6839 	case ixgbe_mac_82599EB:
6840 		incval >>= IXGBE_INCVAL_SHIFT_82599;
6841 		shift -= IXGBE_INCVAL_SHIFT_82599;
6842 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
6843 				(1 << IXGBE_INCPER_SHIFT_82599) | incval);
6844 		break;
6845 	default:
6846 		/* Not supported. */
6847 		return;
6848 	}
6849 
6850 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
6851 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6852 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6853 
6854 	adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6855 	adapter->systime_tc.cc_shift = shift;
6856 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
6857 
6858 	adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6859 	adapter->rx_tstamp_tc.cc_shift = shift;
6860 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6861 
6862 	adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6863 	adapter->tx_tstamp_tc.cc_shift = shift;
6864 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6865 }
6866 
6867 static int
6868 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
6869 {
6870 	struct ixgbe_adapter *adapter =
6871 			(struct ixgbe_adapter *)dev->data->dev_private;
6872 
6873 	adapter->systime_tc.nsec += delta;
6874 	adapter->rx_tstamp_tc.nsec += delta;
6875 	adapter->tx_tstamp_tc.nsec += delta;
6876 
6877 	return 0;
6878 }
6879 
6880 static int
6881 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
6882 {
6883 	uint64_t ns;
6884 	struct ixgbe_adapter *adapter =
6885 			(struct ixgbe_adapter *)dev->data->dev_private;
6886 
6887 	ns = rte_timespec_to_ns(ts);
6888 	/* Set the timecounters to a new value. */
6889 	adapter->systime_tc.nsec = ns;
6890 	adapter->rx_tstamp_tc.nsec = ns;
6891 	adapter->tx_tstamp_tc.nsec = ns;
6892 
6893 	return 0;
6894 }
6895 
6896 static int
6897 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
6898 {
6899 	uint64_t ns, systime_cycles;
6900 	struct ixgbe_adapter *adapter =
6901 			(struct ixgbe_adapter *)dev->data->dev_private;
6902 
6903 	systime_cycles = ixgbe_read_systime_cyclecounter(dev);
6904 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
6905 	*ts = rte_ns_to_timespec(ns);
6906 
6907 	return 0;
6908 }
6909 
6910 static int
6911 ixgbe_timesync_enable(struct rte_eth_dev *dev)
6912 {
6913 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6914 	uint32_t tsync_ctl;
6915 	uint32_t tsauxc;
6916 
6917 	/* Stop the timesync system time. */
6918 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
6919 	/* Reset the timesync system time value. */
6920 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
6921 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
6922 
6923 	/* Enable system time for platforms where it isn't on by default. */
6924 	tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
6925 	tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
6926 	IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
6927 
6928 	ixgbe_start_timecounters(dev);
6929 
6930 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6931 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
6932 			(ETHER_TYPE_1588 |
6933 			 IXGBE_ETQF_FILTER_EN |
6934 			 IXGBE_ETQF_1588));
6935 
6936 	/* Enable timestamping of received PTP packets. */
6937 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6938 	tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
6939 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6940 
6941 	/* Enable timestamping of transmitted PTP packets. */
6942 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6943 	tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
6944 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6945 
6946 	IXGBE_WRITE_FLUSH(hw);
6947 
6948 	return 0;
6949 }
6950 
6951 static int
6952 ixgbe_timesync_disable(struct rte_eth_dev *dev)
6953 {
6954 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6955 	uint32_t tsync_ctl;
6956 
6957 	/* Disable timestamping of transmitted PTP packets. */
6958 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6959 	tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
6960 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6961 
6962 	/* Disable timestamping of received PTP packets. */
6963 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6964 	tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
6965 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6966 
6967 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6968 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
6969 
6970 	/* Stop incrementating the System Time registers. */
6971 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
6972 
6973 	return 0;
6974 }
6975 
6976 static int
6977 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
6978 				 struct timespec *timestamp,
6979 				 uint32_t flags __rte_unused)
6980 {
6981 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6982 	struct ixgbe_adapter *adapter =
6983 		(struct ixgbe_adapter *)dev->data->dev_private;
6984 	uint32_t tsync_rxctl;
6985 	uint64_t rx_tstamp_cycles;
6986 	uint64_t ns;
6987 
6988 	tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6989 	if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
6990 		return -EINVAL;
6991 
6992 	rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
6993 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
6994 	*timestamp = rte_ns_to_timespec(ns);
6995 
6996 	return  0;
6997 }
6998 
6999 static int
7000 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
7001 				 struct timespec *timestamp)
7002 {
7003 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7004 	struct ixgbe_adapter *adapter =
7005 		(struct ixgbe_adapter *)dev->data->dev_private;
7006 	uint32_t tsync_txctl;
7007 	uint64_t tx_tstamp_cycles;
7008 	uint64_t ns;
7009 
7010 	tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7011 	if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
7012 		return -EINVAL;
7013 
7014 	tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
7015 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
7016 	*timestamp = rte_ns_to_timespec(ns);
7017 
7018 	return 0;
7019 }
7020 
7021 static int
7022 ixgbe_get_reg_length(struct rte_eth_dev *dev)
7023 {
7024 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7025 	int count = 0;
7026 	int g_ind = 0;
7027 	const struct reg_info *reg_group;
7028 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7029 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7030 
7031 	while ((reg_group = reg_set[g_ind++]))
7032 		count += ixgbe_regs_group_count(reg_group);
7033 
7034 	return count;
7035 }
7036 
7037 static int
7038 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
7039 {
7040 	int count = 0;
7041 	int g_ind = 0;
7042 	const struct reg_info *reg_group;
7043 
7044 	while ((reg_group = ixgbevf_regs[g_ind++]))
7045 		count += ixgbe_regs_group_count(reg_group);
7046 
7047 	return count;
7048 }
7049 
7050 static int
7051 ixgbe_get_regs(struct rte_eth_dev *dev,
7052 	      struct rte_dev_reg_info *regs)
7053 {
7054 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7055 	uint32_t *data = regs->data;
7056 	int g_ind = 0;
7057 	int count = 0;
7058 	const struct reg_info *reg_group;
7059 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7060 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7061 
7062 	if (data == NULL) {
7063 		regs->length = ixgbe_get_reg_length(dev);
7064 		regs->width = sizeof(uint32_t);
7065 		return 0;
7066 	}
7067 
7068 	/* Support only full register dump */
7069 	if ((regs->length == 0) ||
7070 	    (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
7071 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7072 			hw->device_id;
7073 		while ((reg_group = reg_set[g_ind++]))
7074 			count += ixgbe_read_regs_group(dev, &data[count],
7075 				reg_group);
7076 		return 0;
7077 	}
7078 
7079 	return -ENOTSUP;
7080 }
7081 
7082 static int
7083 ixgbevf_get_regs(struct rte_eth_dev *dev,
7084 		struct rte_dev_reg_info *regs)
7085 {
7086 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7087 	uint32_t *data = regs->data;
7088 	int g_ind = 0;
7089 	int count = 0;
7090 	const struct reg_info *reg_group;
7091 
7092 	if (data == NULL) {
7093 		regs->length = ixgbevf_get_reg_length(dev);
7094 		regs->width = sizeof(uint32_t);
7095 		return 0;
7096 	}
7097 
7098 	/* Support only full register dump */
7099 	if ((regs->length == 0) ||
7100 	    (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
7101 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7102 			hw->device_id;
7103 		while ((reg_group = ixgbevf_regs[g_ind++]))
7104 			count += ixgbe_read_regs_group(dev, &data[count],
7105 						      reg_group);
7106 		return 0;
7107 	}
7108 
7109 	return -ENOTSUP;
7110 }
7111 
7112 static int
7113 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
7114 {
7115 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7116 
7117 	/* Return unit is byte count */
7118 	return hw->eeprom.word_size * 2;
7119 }
7120 
7121 static int
7122 ixgbe_get_eeprom(struct rte_eth_dev *dev,
7123 		struct rte_dev_eeprom_info *in_eeprom)
7124 {
7125 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7126 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7127 	uint16_t *data = in_eeprom->data;
7128 	int first, length;
7129 
7130 	first = in_eeprom->offset >> 1;
7131 	length = in_eeprom->length >> 1;
7132 	if ((first > hw->eeprom.word_size) ||
7133 	    ((first + length) > hw->eeprom.word_size))
7134 		return -EINVAL;
7135 
7136 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7137 
7138 	return eeprom->ops.read_buffer(hw, first, length, data);
7139 }
7140 
7141 static int
7142 ixgbe_set_eeprom(struct rte_eth_dev *dev,
7143 		struct rte_dev_eeprom_info *in_eeprom)
7144 {
7145 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7146 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7147 	uint16_t *data = in_eeprom->data;
7148 	int first, length;
7149 
7150 	first = in_eeprom->offset >> 1;
7151 	length = in_eeprom->length >> 1;
7152 	if ((first > hw->eeprom.word_size) ||
7153 	    ((first + length) > hw->eeprom.word_size))
7154 		return -EINVAL;
7155 
7156 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7157 
7158 	return eeprom->ops.write_buffer(hw,  first, length, data);
7159 }
7160 
7161 uint16_t
7162 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
7163 	switch (mac_type) {
7164 	case ixgbe_mac_X550:
7165 	case ixgbe_mac_X550EM_x:
7166 	case ixgbe_mac_X550EM_a:
7167 		return ETH_RSS_RETA_SIZE_512;
7168 	case ixgbe_mac_X550_vf:
7169 	case ixgbe_mac_X550EM_x_vf:
7170 	case ixgbe_mac_X550EM_a_vf:
7171 		return ETH_RSS_RETA_SIZE_64;
7172 	default:
7173 		return ETH_RSS_RETA_SIZE_128;
7174 	}
7175 }
7176 
7177 uint32_t
7178 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
7179 	switch (mac_type) {
7180 	case ixgbe_mac_X550:
7181 	case ixgbe_mac_X550EM_x:
7182 	case ixgbe_mac_X550EM_a:
7183 		if (reta_idx < ETH_RSS_RETA_SIZE_128)
7184 			return IXGBE_RETA(reta_idx >> 2);
7185 		else
7186 			return IXGBE_ERETA((reta_idx - ETH_RSS_RETA_SIZE_128) >> 2);
7187 	case ixgbe_mac_X550_vf:
7188 	case ixgbe_mac_X550EM_x_vf:
7189 	case ixgbe_mac_X550EM_a_vf:
7190 		return IXGBE_VFRETA(reta_idx >> 2);
7191 	default:
7192 		return IXGBE_RETA(reta_idx >> 2);
7193 	}
7194 }
7195 
7196 uint32_t
7197 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
7198 	switch (mac_type) {
7199 	case ixgbe_mac_X550_vf:
7200 	case ixgbe_mac_X550EM_x_vf:
7201 	case ixgbe_mac_X550EM_a_vf:
7202 		return IXGBE_VFMRQC;
7203 	default:
7204 		return IXGBE_MRQC;
7205 	}
7206 }
7207 
7208 uint32_t
7209 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
7210 	switch (mac_type) {
7211 	case ixgbe_mac_X550_vf:
7212 	case ixgbe_mac_X550EM_x_vf:
7213 	case ixgbe_mac_X550EM_a_vf:
7214 		return IXGBE_VFRSSRK(i);
7215 	default:
7216 		return IXGBE_RSSRK(i);
7217 	}
7218 }
7219 
7220 bool
7221 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
7222 	switch (mac_type) {
7223 	case ixgbe_mac_82599_vf:
7224 	case ixgbe_mac_X540_vf:
7225 		return 0;
7226 	default:
7227 		return 1;
7228 	}
7229 }
7230 
7231 static int
7232 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
7233 			struct rte_eth_dcb_info *dcb_info)
7234 {
7235 	struct ixgbe_dcb_config *dcb_config =
7236 			IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
7237 	struct ixgbe_dcb_tc_config *tc;
7238 	struct rte_eth_dcb_tc_queue_mapping *tc_queue;
7239 	uint8_t nb_tcs;
7240 	uint8_t i, j;
7241 
7242 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_DCB_FLAG)
7243 		dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
7244 	else
7245 		dcb_info->nb_tcs = 1;
7246 
7247 	tc_queue = &dcb_info->tc_queue;
7248 	nb_tcs = dcb_info->nb_tcs;
7249 
7250 	if (dcb_config->vt_mode) { /* vt is enabled*/
7251 		struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
7252 				&dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
7253 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7254 			dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
7255 		if (RTE_ETH_DEV_SRIOV(dev).active > 0) {
7256 			for (j = 0; j < nb_tcs; j++) {
7257 				tc_queue->tc_rxq[0][j].base = j;
7258 				tc_queue->tc_rxq[0][j].nb_queue = 1;
7259 				tc_queue->tc_txq[0][j].base = j;
7260 				tc_queue->tc_txq[0][j].nb_queue = 1;
7261 			}
7262 		} else {
7263 			for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
7264 				for (j = 0; j < nb_tcs; j++) {
7265 					tc_queue->tc_rxq[i][j].base =
7266 						i * nb_tcs + j;
7267 					tc_queue->tc_rxq[i][j].nb_queue = 1;
7268 					tc_queue->tc_txq[i][j].base =
7269 						i * nb_tcs + j;
7270 					tc_queue->tc_txq[i][j].nb_queue = 1;
7271 				}
7272 			}
7273 		}
7274 	} else { /* vt is disabled*/
7275 		struct rte_eth_dcb_rx_conf *rx_conf =
7276 				&dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
7277 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7278 			dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
7279 		if (dcb_info->nb_tcs == ETH_4_TCS) {
7280 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7281 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
7282 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7283 			}
7284 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7285 			dcb_info->tc_queue.tc_txq[0][1].base = 64;
7286 			dcb_info->tc_queue.tc_txq[0][2].base = 96;
7287 			dcb_info->tc_queue.tc_txq[0][3].base = 112;
7288 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
7289 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7290 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7291 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7292 		} else if (dcb_info->nb_tcs == ETH_8_TCS) {
7293 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7294 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
7295 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7296 			}
7297 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7298 			dcb_info->tc_queue.tc_txq[0][1].base = 32;
7299 			dcb_info->tc_queue.tc_txq[0][2].base = 64;
7300 			dcb_info->tc_queue.tc_txq[0][3].base = 80;
7301 			dcb_info->tc_queue.tc_txq[0][4].base = 96;
7302 			dcb_info->tc_queue.tc_txq[0][5].base = 104;
7303 			dcb_info->tc_queue.tc_txq[0][6].base = 112;
7304 			dcb_info->tc_queue.tc_txq[0][7].base = 120;
7305 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
7306 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7307 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7308 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7309 			dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
7310 			dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
7311 			dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
7312 			dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
7313 		}
7314 	}
7315 	for (i = 0; i < dcb_info->nb_tcs; i++) {
7316 		tc = &dcb_config->tc_config[i];
7317 		dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
7318 	}
7319 	return 0;
7320 }
7321 
7322 /* Update e-tag ether type */
7323 static int
7324 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
7325 			    uint16_t ether_type)
7326 {
7327 	uint32_t etag_etype;
7328 
7329 	if (hw->mac.type != ixgbe_mac_X550 &&
7330 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7331 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7332 		return -ENOTSUP;
7333 	}
7334 
7335 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7336 	etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
7337 	etag_etype |= ether_type;
7338 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7339 	IXGBE_WRITE_FLUSH(hw);
7340 
7341 	return 0;
7342 }
7343 
7344 /* Config l2 tunnel ether type */
7345 static int
7346 ixgbe_dev_l2_tunnel_eth_type_conf(struct rte_eth_dev *dev,
7347 				  struct rte_eth_l2_tunnel_conf *l2_tunnel)
7348 {
7349 	int ret = 0;
7350 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7351 	struct ixgbe_l2_tn_info *l2_tn_info =
7352 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7353 
7354 	if (l2_tunnel == NULL)
7355 		return -EINVAL;
7356 
7357 	switch (l2_tunnel->l2_tunnel_type) {
7358 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7359 		l2_tn_info->e_tag_ether_type = l2_tunnel->ether_type;
7360 		ret = ixgbe_update_e_tag_eth_type(hw, l2_tunnel->ether_type);
7361 		break;
7362 	default:
7363 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7364 		ret = -EINVAL;
7365 		break;
7366 	}
7367 
7368 	return ret;
7369 }
7370 
7371 /* Enable e-tag tunnel */
7372 static int
7373 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
7374 {
7375 	uint32_t etag_etype;
7376 
7377 	if (hw->mac.type != ixgbe_mac_X550 &&
7378 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7379 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7380 		return -ENOTSUP;
7381 	}
7382 
7383 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7384 	etag_etype |= IXGBE_ETAG_ETYPE_VALID;
7385 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7386 	IXGBE_WRITE_FLUSH(hw);
7387 
7388 	return 0;
7389 }
7390 
7391 /* Enable l2 tunnel */
7392 static int
7393 ixgbe_dev_l2_tunnel_enable(struct rte_eth_dev *dev,
7394 			   enum rte_eth_tunnel_type l2_tunnel_type)
7395 {
7396 	int ret = 0;
7397 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7398 	struct ixgbe_l2_tn_info *l2_tn_info =
7399 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7400 
7401 	switch (l2_tunnel_type) {
7402 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7403 		l2_tn_info->e_tag_en = TRUE;
7404 		ret = ixgbe_e_tag_enable(hw);
7405 		break;
7406 	default:
7407 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7408 		ret = -EINVAL;
7409 		break;
7410 	}
7411 
7412 	return ret;
7413 }
7414 
7415 /* Disable e-tag tunnel */
7416 static int
7417 ixgbe_e_tag_disable(struct ixgbe_hw *hw)
7418 {
7419 	uint32_t etag_etype;
7420 
7421 	if (hw->mac.type != ixgbe_mac_X550 &&
7422 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7423 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7424 		return -ENOTSUP;
7425 	}
7426 
7427 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7428 	etag_etype &= ~IXGBE_ETAG_ETYPE_VALID;
7429 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7430 	IXGBE_WRITE_FLUSH(hw);
7431 
7432 	return 0;
7433 }
7434 
7435 /* Disable l2 tunnel */
7436 static int
7437 ixgbe_dev_l2_tunnel_disable(struct rte_eth_dev *dev,
7438 			    enum rte_eth_tunnel_type l2_tunnel_type)
7439 {
7440 	int ret = 0;
7441 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7442 	struct ixgbe_l2_tn_info *l2_tn_info =
7443 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7444 
7445 	switch (l2_tunnel_type) {
7446 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7447 		l2_tn_info->e_tag_en = FALSE;
7448 		ret = ixgbe_e_tag_disable(hw);
7449 		break;
7450 	default:
7451 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7452 		ret = -EINVAL;
7453 		break;
7454 	}
7455 
7456 	return ret;
7457 }
7458 
7459 static int
7460 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
7461 		       struct rte_eth_l2_tunnel_conf *l2_tunnel)
7462 {
7463 	int ret = 0;
7464 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7465 	uint32_t i, rar_entries;
7466 	uint32_t rar_low, rar_high;
7467 
7468 	if (hw->mac.type != ixgbe_mac_X550 &&
7469 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7470 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7471 		return -ENOTSUP;
7472 	}
7473 
7474 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7475 
7476 	for (i = 1; i < rar_entries; i++) {
7477 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7478 		rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
7479 		if ((rar_high & IXGBE_RAH_AV) &&
7480 		    (rar_high & IXGBE_RAH_ADTYPE) &&
7481 		    ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
7482 		     l2_tunnel->tunnel_id)) {
7483 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
7484 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
7485 
7486 			ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
7487 
7488 			return ret;
7489 		}
7490 	}
7491 
7492 	return ret;
7493 }
7494 
7495 static int
7496 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
7497 		       struct rte_eth_l2_tunnel_conf *l2_tunnel)
7498 {
7499 	int ret = 0;
7500 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7501 	uint32_t i, rar_entries;
7502 	uint32_t rar_low, rar_high;
7503 
7504 	if (hw->mac.type != ixgbe_mac_X550 &&
7505 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7506 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7507 		return -ENOTSUP;
7508 	}
7509 
7510 	/* One entry for one tunnel. Try to remove potential existing entry. */
7511 	ixgbe_e_tag_filter_del(dev, l2_tunnel);
7512 
7513 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7514 
7515 	for (i = 1; i < rar_entries; i++) {
7516 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7517 		if (rar_high & IXGBE_RAH_AV) {
7518 			continue;
7519 		} else {
7520 			ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
7521 			rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
7522 			rar_low = l2_tunnel->tunnel_id;
7523 
7524 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
7525 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
7526 
7527 			return ret;
7528 		}
7529 	}
7530 
7531 	PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
7532 		     " Please remove a rule before adding a new one.");
7533 	return -EINVAL;
7534 }
7535 
7536 static inline struct ixgbe_l2_tn_filter *
7537 ixgbe_l2_tn_filter_lookup(struct ixgbe_l2_tn_info *l2_tn_info,
7538 			  struct ixgbe_l2_tn_key *key)
7539 {
7540 	int ret;
7541 
7542 	ret = rte_hash_lookup(l2_tn_info->hash_handle, (const void *)key);
7543 	if (ret < 0)
7544 		return NULL;
7545 
7546 	return l2_tn_info->hash_map[ret];
7547 }
7548 
7549 static inline int
7550 ixgbe_insert_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7551 			  struct ixgbe_l2_tn_filter *l2_tn_filter)
7552 {
7553 	int ret;
7554 
7555 	ret = rte_hash_add_key(l2_tn_info->hash_handle,
7556 			       &l2_tn_filter->key);
7557 
7558 	if (ret < 0) {
7559 		PMD_DRV_LOG(ERR,
7560 			    "Failed to insert L2 tunnel filter"
7561 			    " to hash table %d!",
7562 			    ret);
7563 		return ret;
7564 	}
7565 
7566 	l2_tn_info->hash_map[ret] = l2_tn_filter;
7567 
7568 	TAILQ_INSERT_TAIL(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7569 
7570 	return 0;
7571 }
7572 
7573 static inline int
7574 ixgbe_remove_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7575 			  struct ixgbe_l2_tn_key *key)
7576 {
7577 	int ret;
7578 	struct ixgbe_l2_tn_filter *l2_tn_filter;
7579 
7580 	ret = rte_hash_del_key(l2_tn_info->hash_handle, key);
7581 
7582 	if (ret < 0) {
7583 		PMD_DRV_LOG(ERR,
7584 			    "No such L2 tunnel filter to delete %d!",
7585 			    ret);
7586 		return ret;
7587 	}
7588 
7589 	l2_tn_filter = l2_tn_info->hash_map[ret];
7590 	l2_tn_info->hash_map[ret] = NULL;
7591 
7592 	TAILQ_REMOVE(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7593 	rte_free(l2_tn_filter);
7594 
7595 	return 0;
7596 }
7597 
7598 /* Add l2 tunnel filter */
7599 int
7600 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
7601 			       struct rte_eth_l2_tunnel_conf *l2_tunnel,
7602 			       bool restore)
7603 {
7604 	int ret;
7605 	struct ixgbe_l2_tn_info *l2_tn_info =
7606 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7607 	struct ixgbe_l2_tn_key key;
7608 	struct ixgbe_l2_tn_filter *node;
7609 
7610 	if (!restore) {
7611 		key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7612 		key.tn_id = l2_tunnel->tunnel_id;
7613 
7614 		node = ixgbe_l2_tn_filter_lookup(l2_tn_info, &key);
7615 
7616 		if (node) {
7617 			PMD_DRV_LOG(ERR,
7618 				    "The L2 tunnel filter already exists!");
7619 			return -EINVAL;
7620 		}
7621 
7622 		node = rte_zmalloc("ixgbe_l2_tn",
7623 				   sizeof(struct ixgbe_l2_tn_filter),
7624 				   0);
7625 		if (!node)
7626 			return -ENOMEM;
7627 
7628 		rte_memcpy(&node->key,
7629 				 &key,
7630 				 sizeof(struct ixgbe_l2_tn_key));
7631 		node->pool = l2_tunnel->pool;
7632 		ret = ixgbe_insert_l2_tn_filter(l2_tn_info, node);
7633 		if (ret < 0) {
7634 			rte_free(node);
7635 			return ret;
7636 		}
7637 	}
7638 
7639 	switch (l2_tunnel->l2_tunnel_type) {
7640 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7641 		ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
7642 		break;
7643 	default:
7644 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7645 		ret = -EINVAL;
7646 		break;
7647 	}
7648 
7649 	if ((!restore) && (ret < 0))
7650 		(void)ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7651 
7652 	return ret;
7653 }
7654 
7655 /* Delete l2 tunnel filter */
7656 int
7657 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
7658 			       struct rte_eth_l2_tunnel_conf *l2_tunnel)
7659 {
7660 	int ret;
7661 	struct ixgbe_l2_tn_info *l2_tn_info =
7662 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7663 	struct ixgbe_l2_tn_key key;
7664 
7665 	key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7666 	key.tn_id = l2_tunnel->tunnel_id;
7667 	ret = ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7668 	if (ret < 0)
7669 		return ret;
7670 
7671 	switch (l2_tunnel->l2_tunnel_type) {
7672 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7673 		ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
7674 		break;
7675 	default:
7676 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7677 		ret = -EINVAL;
7678 		break;
7679 	}
7680 
7681 	return ret;
7682 }
7683 
7684 /**
7685  * ixgbe_dev_l2_tunnel_filter_handle - Handle operations for l2 tunnel filter.
7686  * @dev: pointer to rte_eth_dev structure
7687  * @filter_op:operation will be taken.
7688  * @arg: a pointer to specific structure corresponding to the filter_op
7689  */
7690 static int
7691 ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
7692 				  enum rte_filter_op filter_op,
7693 				  void *arg)
7694 {
7695 	int ret;
7696 
7697 	if (filter_op == RTE_ETH_FILTER_NOP)
7698 		return 0;
7699 
7700 	if (arg == NULL) {
7701 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
7702 			    filter_op);
7703 		return -EINVAL;
7704 	}
7705 
7706 	switch (filter_op) {
7707 	case RTE_ETH_FILTER_ADD:
7708 		ret = ixgbe_dev_l2_tunnel_filter_add
7709 			(dev,
7710 			 (struct rte_eth_l2_tunnel_conf *)arg,
7711 			 FALSE);
7712 		break;
7713 	case RTE_ETH_FILTER_DELETE:
7714 		ret = ixgbe_dev_l2_tunnel_filter_del
7715 			(dev,
7716 			 (struct rte_eth_l2_tunnel_conf *)arg);
7717 		break;
7718 	default:
7719 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
7720 		ret = -EINVAL;
7721 		break;
7722 	}
7723 	return ret;
7724 }
7725 
7726 static int
7727 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
7728 {
7729 	int ret = 0;
7730 	uint32_t ctrl;
7731 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7732 
7733 	if (hw->mac.type != ixgbe_mac_X550 &&
7734 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7735 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7736 		return -ENOTSUP;
7737 	}
7738 
7739 	ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
7740 	ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
7741 	if (en)
7742 		ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
7743 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
7744 
7745 	return ret;
7746 }
7747 
7748 /* Enable l2 tunnel forwarding */
7749 static int
7750 ixgbe_dev_l2_tunnel_forwarding_enable
7751 	(struct rte_eth_dev *dev,
7752 	 enum rte_eth_tunnel_type l2_tunnel_type)
7753 {
7754 	struct ixgbe_l2_tn_info *l2_tn_info =
7755 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7756 	int ret = 0;
7757 
7758 	switch (l2_tunnel_type) {
7759 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7760 		l2_tn_info->e_tag_fwd_en = TRUE;
7761 		ret = ixgbe_e_tag_forwarding_en_dis(dev, 1);
7762 		break;
7763 	default:
7764 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7765 		ret = -EINVAL;
7766 		break;
7767 	}
7768 
7769 	return ret;
7770 }
7771 
7772 /* Disable l2 tunnel forwarding */
7773 static int
7774 ixgbe_dev_l2_tunnel_forwarding_disable
7775 	(struct rte_eth_dev *dev,
7776 	 enum rte_eth_tunnel_type l2_tunnel_type)
7777 {
7778 	struct ixgbe_l2_tn_info *l2_tn_info =
7779 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7780 	int ret = 0;
7781 
7782 	switch (l2_tunnel_type) {
7783 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7784 		l2_tn_info->e_tag_fwd_en = FALSE;
7785 		ret = ixgbe_e_tag_forwarding_en_dis(dev, 0);
7786 		break;
7787 	default:
7788 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7789 		ret = -EINVAL;
7790 		break;
7791 	}
7792 
7793 	return ret;
7794 }
7795 
7796 static int
7797 ixgbe_e_tag_insertion_en_dis(struct rte_eth_dev *dev,
7798 			     struct rte_eth_l2_tunnel_conf *l2_tunnel,
7799 			     bool en)
7800 {
7801 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
7802 	int ret = 0;
7803 	uint32_t vmtir, vmvir;
7804 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7805 
7806 	if (l2_tunnel->vf_id >= pci_dev->max_vfs) {
7807 		PMD_DRV_LOG(ERR,
7808 			    "VF id %u should be less than %u",
7809 			    l2_tunnel->vf_id,
7810 			    pci_dev->max_vfs);
7811 		return -EINVAL;
7812 	}
7813 
7814 	if (hw->mac.type != ixgbe_mac_X550 &&
7815 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7816 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7817 		return -ENOTSUP;
7818 	}
7819 
7820 	if (en)
7821 		vmtir = l2_tunnel->tunnel_id;
7822 	else
7823 		vmtir = 0;
7824 
7825 	IXGBE_WRITE_REG(hw, IXGBE_VMTIR(l2_tunnel->vf_id), vmtir);
7826 
7827 	vmvir = IXGBE_READ_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id));
7828 	vmvir &= ~IXGBE_VMVIR_TAGA_MASK;
7829 	if (en)
7830 		vmvir |= IXGBE_VMVIR_TAGA_ETAG_INSERT;
7831 	IXGBE_WRITE_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id), vmvir);
7832 
7833 	return ret;
7834 }
7835 
7836 /* Enable l2 tunnel tag insertion */
7837 static int
7838 ixgbe_dev_l2_tunnel_insertion_enable(struct rte_eth_dev *dev,
7839 				     struct rte_eth_l2_tunnel_conf *l2_tunnel)
7840 {
7841 	int ret = 0;
7842 
7843 	switch (l2_tunnel->l2_tunnel_type) {
7844 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7845 		ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 1);
7846 		break;
7847 	default:
7848 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7849 		ret = -EINVAL;
7850 		break;
7851 	}
7852 
7853 	return ret;
7854 }
7855 
7856 /* Disable l2 tunnel tag insertion */
7857 static int
7858 ixgbe_dev_l2_tunnel_insertion_disable
7859 	(struct rte_eth_dev *dev,
7860 	 struct rte_eth_l2_tunnel_conf *l2_tunnel)
7861 {
7862 	int ret = 0;
7863 
7864 	switch (l2_tunnel->l2_tunnel_type) {
7865 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7866 		ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 0);
7867 		break;
7868 	default:
7869 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7870 		ret = -EINVAL;
7871 		break;
7872 	}
7873 
7874 	return ret;
7875 }
7876 
7877 static int
7878 ixgbe_e_tag_stripping_en_dis(struct rte_eth_dev *dev,
7879 			     bool en)
7880 {
7881 	int ret = 0;
7882 	uint32_t qde;
7883 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7884 
7885 	if (hw->mac.type != ixgbe_mac_X550 &&
7886 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7887 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7888 		return -ENOTSUP;
7889 	}
7890 
7891 	qde = IXGBE_READ_REG(hw, IXGBE_QDE);
7892 	if (en)
7893 		qde |= IXGBE_QDE_STRIP_TAG;
7894 	else
7895 		qde &= ~IXGBE_QDE_STRIP_TAG;
7896 	qde &= ~IXGBE_QDE_READ;
7897 	qde |= IXGBE_QDE_WRITE;
7898 	IXGBE_WRITE_REG(hw, IXGBE_QDE, qde);
7899 
7900 	return ret;
7901 }
7902 
7903 /* Enable l2 tunnel tag stripping */
7904 static int
7905 ixgbe_dev_l2_tunnel_stripping_enable
7906 	(struct rte_eth_dev *dev,
7907 	 enum rte_eth_tunnel_type l2_tunnel_type)
7908 {
7909 	int ret = 0;
7910 
7911 	switch (l2_tunnel_type) {
7912 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7913 		ret = ixgbe_e_tag_stripping_en_dis(dev, 1);
7914 		break;
7915 	default:
7916 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7917 		ret = -EINVAL;
7918 		break;
7919 	}
7920 
7921 	return ret;
7922 }
7923 
7924 /* Disable l2 tunnel tag stripping */
7925 static int
7926 ixgbe_dev_l2_tunnel_stripping_disable
7927 	(struct rte_eth_dev *dev,
7928 	 enum rte_eth_tunnel_type l2_tunnel_type)
7929 {
7930 	int ret = 0;
7931 
7932 	switch (l2_tunnel_type) {
7933 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7934 		ret = ixgbe_e_tag_stripping_en_dis(dev, 0);
7935 		break;
7936 	default:
7937 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7938 		ret = -EINVAL;
7939 		break;
7940 	}
7941 
7942 	return ret;
7943 }
7944 
7945 /* Enable/disable l2 tunnel offload functions */
7946 static int
7947 ixgbe_dev_l2_tunnel_offload_set
7948 	(struct rte_eth_dev *dev,
7949 	 struct rte_eth_l2_tunnel_conf *l2_tunnel,
7950 	 uint32_t mask,
7951 	 uint8_t en)
7952 {
7953 	int ret = 0;
7954 
7955 	if (l2_tunnel == NULL)
7956 		return -EINVAL;
7957 
7958 	ret = -EINVAL;
7959 	if (mask & ETH_L2_TUNNEL_ENABLE_MASK) {
7960 		if (en)
7961 			ret = ixgbe_dev_l2_tunnel_enable(
7962 				dev,
7963 				l2_tunnel->l2_tunnel_type);
7964 		else
7965 			ret = ixgbe_dev_l2_tunnel_disable(
7966 				dev,
7967 				l2_tunnel->l2_tunnel_type);
7968 	}
7969 
7970 	if (mask & ETH_L2_TUNNEL_INSERTION_MASK) {
7971 		if (en)
7972 			ret = ixgbe_dev_l2_tunnel_insertion_enable(
7973 				dev,
7974 				l2_tunnel);
7975 		else
7976 			ret = ixgbe_dev_l2_tunnel_insertion_disable(
7977 				dev,
7978 				l2_tunnel);
7979 	}
7980 
7981 	if (mask & ETH_L2_TUNNEL_STRIPPING_MASK) {
7982 		if (en)
7983 			ret = ixgbe_dev_l2_tunnel_stripping_enable(
7984 				dev,
7985 				l2_tunnel->l2_tunnel_type);
7986 		else
7987 			ret = ixgbe_dev_l2_tunnel_stripping_disable(
7988 				dev,
7989 				l2_tunnel->l2_tunnel_type);
7990 	}
7991 
7992 	if (mask & ETH_L2_TUNNEL_FORWARDING_MASK) {
7993 		if (en)
7994 			ret = ixgbe_dev_l2_tunnel_forwarding_enable(
7995 				dev,
7996 				l2_tunnel->l2_tunnel_type);
7997 		else
7998 			ret = ixgbe_dev_l2_tunnel_forwarding_disable(
7999 				dev,
8000 				l2_tunnel->l2_tunnel_type);
8001 	}
8002 
8003 	return ret;
8004 }
8005 
8006 static int
8007 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
8008 			uint16_t port)
8009 {
8010 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
8011 	IXGBE_WRITE_FLUSH(hw);
8012 
8013 	return 0;
8014 }
8015 
8016 /* There's only one register for VxLAN UDP port.
8017  * So, we cannot add several ports. Will update it.
8018  */
8019 static int
8020 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
8021 		     uint16_t port)
8022 {
8023 	if (port == 0) {
8024 		PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
8025 		return -EINVAL;
8026 	}
8027 
8028 	return ixgbe_update_vxlan_port(hw, port);
8029 }
8030 
8031 /* We cannot delete the VxLAN port. For there's a register for VxLAN
8032  * UDP port, it must have a value.
8033  * So, will reset it to the original value 0.
8034  */
8035 static int
8036 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
8037 		     uint16_t port)
8038 {
8039 	uint16_t cur_port;
8040 
8041 	cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
8042 
8043 	if (cur_port != port) {
8044 		PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
8045 		return -EINVAL;
8046 	}
8047 
8048 	return ixgbe_update_vxlan_port(hw, 0);
8049 }
8050 
8051 /* Add UDP tunneling port */
8052 static int
8053 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
8054 			      struct rte_eth_udp_tunnel *udp_tunnel)
8055 {
8056 	int ret = 0;
8057 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8058 
8059 	if (hw->mac.type != ixgbe_mac_X550 &&
8060 	    hw->mac.type != ixgbe_mac_X550EM_x &&
8061 	    hw->mac.type != ixgbe_mac_X550EM_a) {
8062 		return -ENOTSUP;
8063 	}
8064 
8065 	if (udp_tunnel == NULL)
8066 		return -EINVAL;
8067 
8068 	switch (udp_tunnel->prot_type) {
8069 	case RTE_TUNNEL_TYPE_VXLAN:
8070 		ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
8071 		break;
8072 
8073 	case RTE_TUNNEL_TYPE_GENEVE:
8074 	case RTE_TUNNEL_TYPE_TEREDO:
8075 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
8076 		ret = -EINVAL;
8077 		break;
8078 
8079 	default:
8080 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
8081 		ret = -EINVAL;
8082 		break;
8083 	}
8084 
8085 	return ret;
8086 }
8087 
8088 /* Remove UDP tunneling port */
8089 static int
8090 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
8091 			      struct rte_eth_udp_tunnel *udp_tunnel)
8092 {
8093 	int ret = 0;
8094 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8095 
8096 	if (hw->mac.type != ixgbe_mac_X550 &&
8097 	    hw->mac.type != ixgbe_mac_X550EM_x &&
8098 	    hw->mac.type != ixgbe_mac_X550EM_a) {
8099 		return -ENOTSUP;
8100 	}
8101 
8102 	if (udp_tunnel == NULL)
8103 		return -EINVAL;
8104 
8105 	switch (udp_tunnel->prot_type) {
8106 	case RTE_TUNNEL_TYPE_VXLAN:
8107 		ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
8108 		break;
8109 	case RTE_TUNNEL_TYPE_GENEVE:
8110 	case RTE_TUNNEL_TYPE_TEREDO:
8111 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
8112 		ret = -EINVAL;
8113 		break;
8114 	default:
8115 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
8116 		ret = -EINVAL;
8117 		break;
8118 	}
8119 
8120 	return ret;
8121 }
8122 
8123 static void
8124 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
8125 {
8126 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8127 
8128 	hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_ALLMULTI);
8129 }
8130 
8131 static void
8132 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
8133 {
8134 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8135 
8136 	hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_MULTI);
8137 }
8138 
8139 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
8140 {
8141 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8142 	u32 in_msg = 0;
8143 
8144 	/* peek the message first */
8145 	in_msg = IXGBE_READ_REG(hw, IXGBE_VFMBMEM);
8146 
8147 	/* PF reset VF event */
8148 	if (in_msg == IXGBE_PF_CONTROL_MSG) {
8149 		/* dummy mbx read to ack pf */
8150 		if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
8151 			return;
8152 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
8153 					      NULL);
8154 	}
8155 }
8156 
8157 static int
8158 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
8159 {
8160 	uint32_t eicr;
8161 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8162 	struct ixgbe_interrupt *intr =
8163 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8164 	ixgbevf_intr_disable(hw);
8165 
8166 	/* read-on-clear nic registers here */
8167 	eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
8168 	intr->flags = 0;
8169 
8170 	/* only one misc vector supported - mailbox */
8171 	eicr &= IXGBE_VTEICR_MASK;
8172 	if (eicr == IXGBE_MISC_VEC_ID)
8173 		intr->flags |= IXGBE_FLAG_MAILBOX;
8174 
8175 	return 0;
8176 }
8177 
8178 static int
8179 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
8180 {
8181 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8182 	struct ixgbe_interrupt *intr =
8183 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8184 
8185 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
8186 		ixgbevf_mbx_process(dev);
8187 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
8188 	}
8189 
8190 	ixgbevf_intr_enable(hw);
8191 
8192 	return 0;
8193 }
8194 
8195 static void
8196 ixgbevf_dev_interrupt_handler(void *param)
8197 {
8198 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
8199 
8200 	ixgbevf_dev_interrupt_get_status(dev);
8201 	ixgbevf_dev_interrupt_action(dev);
8202 }
8203 
8204 /**
8205  *  ixgbe_disable_sec_tx_path_generic - Stops the transmit data path
8206  *  @hw: pointer to hardware structure
8207  *
8208  *  Stops the transmit data path and waits for the HW to internally empty
8209  *  the Tx security block
8210  **/
8211 int ixgbe_disable_sec_tx_path_generic(struct ixgbe_hw *hw)
8212 {
8213 #define IXGBE_MAX_SECTX_POLL 40
8214 
8215 	int i;
8216 	int sectxreg;
8217 
8218 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8219 	sectxreg |= IXGBE_SECTXCTRL_TX_DIS;
8220 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8221 	for (i = 0; i < IXGBE_MAX_SECTX_POLL; i++) {
8222 		sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXSTAT);
8223 		if (sectxreg & IXGBE_SECTXSTAT_SECTX_RDY)
8224 			break;
8225 		/* Use interrupt-safe sleep just in case */
8226 		usec_delay(1000);
8227 	}
8228 
8229 	/* For informational purposes only */
8230 	if (i >= IXGBE_MAX_SECTX_POLL)
8231 		PMD_DRV_LOG(DEBUG, "Tx unit being enabled before security "
8232 			 "path fully disabled.  Continuing with init.");
8233 
8234 	return IXGBE_SUCCESS;
8235 }
8236 
8237 /**
8238  *  ixgbe_enable_sec_tx_path_generic - Enables the transmit data path
8239  *  @hw: pointer to hardware structure
8240  *
8241  *  Enables the transmit data path.
8242  **/
8243 int ixgbe_enable_sec_tx_path_generic(struct ixgbe_hw *hw)
8244 {
8245 	uint32_t sectxreg;
8246 
8247 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8248 	sectxreg &= ~IXGBE_SECTXCTRL_TX_DIS;
8249 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8250 	IXGBE_WRITE_FLUSH(hw);
8251 
8252 	return IXGBE_SUCCESS;
8253 }
8254 
8255 /* restore n-tuple filter */
8256 static inline void
8257 ixgbe_ntuple_filter_restore(struct rte_eth_dev *dev)
8258 {
8259 	struct ixgbe_filter_info *filter_info =
8260 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8261 	struct ixgbe_5tuple_filter *node;
8262 
8263 	TAILQ_FOREACH(node, &filter_info->fivetuple_list, entries) {
8264 		ixgbe_inject_5tuple_filter(dev, node);
8265 	}
8266 }
8267 
8268 /* restore ethernet type filter */
8269 static inline void
8270 ixgbe_ethertype_filter_restore(struct rte_eth_dev *dev)
8271 {
8272 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8273 	struct ixgbe_filter_info *filter_info =
8274 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8275 	int i;
8276 
8277 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8278 		if (filter_info->ethertype_mask & (1 << i)) {
8279 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i),
8280 					filter_info->ethertype_filters[i].etqf);
8281 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i),
8282 					filter_info->ethertype_filters[i].etqs);
8283 			IXGBE_WRITE_FLUSH(hw);
8284 		}
8285 	}
8286 }
8287 
8288 /* restore SYN filter */
8289 static inline void
8290 ixgbe_syn_filter_restore(struct rte_eth_dev *dev)
8291 {
8292 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8293 	struct ixgbe_filter_info *filter_info =
8294 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8295 	uint32_t synqf;
8296 
8297 	synqf = filter_info->syn_info;
8298 
8299 	if (synqf & IXGBE_SYN_FILTER_ENABLE) {
8300 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
8301 		IXGBE_WRITE_FLUSH(hw);
8302 	}
8303 }
8304 
8305 /* restore L2 tunnel filter */
8306 static inline void
8307 ixgbe_l2_tn_filter_restore(struct rte_eth_dev *dev)
8308 {
8309 	struct ixgbe_l2_tn_info *l2_tn_info =
8310 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8311 	struct ixgbe_l2_tn_filter *node;
8312 	struct rte_eth_l2_tunnel_conf l2_tn_conf;
8313 
8314 	TAILQ_FOREACH(node, &l2_tn_info->l2_tn_list, entries) {
8315 		l2_tn_conf.l2_tunnel_type = node->key.l2_tn_type;
8316 		l2_tn_conf.tunnel_id      = node->key.tn_id;
8317 		l2_tn_conf.pool           = node->pool;
8318 		(void)ixgbe_dev_l2_tunnel_filter_add(dev, &l2_tn_conf, TRUE);
8319 	}
8320 }
8321 
8322 /* restore rss filter */
8323 static inline void
8324 ixgbe_rss_filter_restore(struct rte_eth_dev *dev)
8325 {
8326 	struct ixgbe_filter_info *filter_info =
8327 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8328 
8329 	if (filter_info->rss_info.num)
8330 		ixgbe_config_rss_filter(dev,
8331 			&filter_info->rss_info, TRUE);
8332 }
8333 
8334 static int
8335 ixgbe_filter_restore(struct rte_eth_dev *dev)
8336 {
8337 	ixgbe_ntuple_filter_restore(dev);
8338 	ixgbe_ethertype_filter_restore(dev);
8339 	ixgbe_syn_filter_restore(dev);
8340 	ixgbe_fdir_filter_restore(dev);
8341 	ixgbe_l2_tn_filter_restore(dev);
8342 	ixgbe_rss_filter_restore(dev);
8343 
8344 	return 0;
8345 }
8346 
8347 static void
8348 ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev)
8349 {
8350 	struct ixgbe_l2_tn_info *l2_tn_info =
8351 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8352 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8353 
8354 	if (l2_tn_info->e_tag_en)
8355 		(void)ixgbe_e_tag_enable(hw);
8356 
8357 	if (l2_tn_info->e_tag_fwd_en)
8358 		(void)ixgbe_e_tag_forwarding_en_dis(dev, 1);
8359 
8360 	(void)ixgbe_update_e_tag_eth_type(hw, l2_tn_info->e_tag_ether_type);
8361 }
8362 
8363 /* remove all the n-tuple filters */
8364 void
8365 ixgbe_clear_all_ntuple_filter(struct rte_eth_dev *dev)
8366 {
8367 	struct ixgbe_filter_info *filter_info =
8368 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8369 	struct ixgbe_5tuple_filter *p_5tuple;
8370 
8371 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
8372 		ixgbe_remove_5tuple_filter(dev, p_5tuple);
8373 }
8374 
8375 /* remove all the ether type filters */
8376 void
8377 ixgbe_clear_all_ethertype_filter(struct rte_eth_dev *dev)
8378 {
8379 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8380 	struct ixgbe_filter_info *filter_info =
8381 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8382 	int i;
8383 
8384 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8385 		if (filter_info->ethertype_mask & (1 << i) &&
8386 		    !filter_info->ethertype_filters[i].conf) {
8387 			(void)ixgbe_ethertype_filter_remove(filter_info,
8388 							    (uint8_t)i);
8389 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i), 0);
8390 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i), 0);
8391 			IXGBE_WRITE_FLUSH(hw);
8392 		}
8393 	}
8394 }
8395 
8396 /* remove the SYN filter */
8397 void
8398 ixgbe_clear_syn_filter(struct rte_eth_dev *dev)
8399 {
8400 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8401 	struct ixgbe_filter_info *filter_info =
8402 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8403 
8404 	if (filter_info->syn_info & IXGBE_SYN_FILTER_ENABLE) {
8405 		filter_info->syn_info = 0;
8406 
8407 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, 0);
8408 		IXGBE_WRITE_FLUSH(hw);
8409 	}
8410 }
8411 
8412 /* remove all the L2 tunnel filters */
8413 int
8414 ixgbe_clear_all_l2_tn_filter(struct rte_eth_dev *dev)
8415 {
8416 	struct ixgbe_l2_tn_info *l2_tn_info =
8417 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8418 	struct ixgbe_l2_tn_filter *l2_tn_filter;
8419 	struct rte_eth_l2_tunnel_conf l2_tn_conf;
8420 	int ret = 0;
8421 
8422 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
8423 		l2_tn_conf.l2_tunnel_type = l2_tn_filter->key.l2_tn_type;
8424 		l2_tn_conf.tunnel_id      = l2_tn_filter->key.tn_id;
8425 		l2_tn_conf.pool           = l2_tn_filter->pool;
8426 		ret = ixgbe_dev_l2_tunnel_filter_del(dev, &l2_tn_conf);
8427 		if (ret < 0)
8428 			return ret;
8429 	}
8430 
8431 	return 0;
8432 }
8433 
8434 RTE_PMD_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd);
8435 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
8436 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe, "* igb_uio | uio_pci_generic | vfio-pci");
8437 RTE_PMD_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd);
8438 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);
8439 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe_vf, "* igb_uio | vfio-pci");
8440 
8441 RTE_INIT(ixgbe_init_log);
8442 static void
8443 ixgbe_init_log(void)
8444 {
8445 	ixgbe_logtype_init = rte_log_register("pmd.ixgbe.init");
8446 	if (ixgbe_logtype_init >= 0)
8447 		rte_log_set_level(ixgbe_logtype_init, RTE_LOG_NOTICE);
8448 	ixgbe_logtype_driver = rte_log_register("pmd.ixgbe.driver");
8449 	if (ixgbe_logtype_driver >= 0)
8450 		rte_log_set_level(ixgbe_logtype_driver, RTE_LOG_NOTICE);
8451 }
8452