xref: /dpdk/drivers/net/intel/ixgbe/ixgbe_ethdev.c (revision b87fc2117eb2a35a8c65c9dd74b5aace40fbad95)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2017 Intel Corporation
3  */
4 
5 #include <sys/queue.h>
6 #include <stdio.h>
7 #include <errno.h>
8 #include <stdint.h>
9 #include <string.h>
10 #include <unistd.h>
11 #include <stdarg.h>
12 #include <inttypes.h>
13 #include <rte_string_fns.h>
14 #include <rte_byteorder.h>
15 #include <rte_common.h>
16 #include <rte_cycles.h>
17 
18 #include <rte_interrupts.h>
19 #include <rte_log.h>
20 #include <rte_debug.h>
21 #include <rte_pci.h>
22 #include <bus_pci_driver.h>
23 #include <rte_branch_prediction.h>
24 #include <rte_memory.h>
25 #include <rte_kvargs.h>
26 #include <rte_eal.h>
27 #include <rte_alarm.h>
28 #include <rte_ether.h>
29 #include <ethdev_driver.h>
30 #include <ethdev_pci.h>
31 #include <rte_malloc.h>
32 #include <rte_random.h>
33 #include <dev_driver.h>
34 #include <rte_hash_crc.h>
35 #ifdef RTE_LIB_SECURITY
36 #include <rte_security_driver.h>
37 #endif
38 #include <rte_os_shim.h>
39 
40 #include "ixgbe_logs.h"
41 #include "base/ixgbe_api.h"
42 #include "base/ixgbe_vf.h"
43 #include "base/ixgbe_common.h"
44 #include "ixgbe_ethdev.h"
45 #include "ixgbe_bypass.h"
46 #include "ixgbe_rxtx.h"
47 #include "base/ixgbe_type.h"
48 #include "base/ixgbe_phy.h"
49 #include "base/ixgbe_osdep.h"
50 #include "ixgbe_regs.h"
51 
52 /*
53  * High threshold controlling when to start sending XOFF frames. Must be at
54  * least 8 bytes less than receive packet buffer size. This value is in units
55  * of 1024 bytes.
56  */
57 #define IXGBE_FC_HI    0x80
58 
59 /*
60  * Low threshold controlling when to start sending XON frames. This value is
61  * in units of 1024 bytes.
62  */
63 #define IXGBE_FC_LO    0x40
64 
65 /* Timer value included in XOFF frames. */
66 #define IXGBE_FC_PAUSE 0x680
67 
68 /*Default value of Max Rx Queue*/
69 #define IXGBE_MAX_RX_QUEUE_NUM 128
70 
71 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
72 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
73 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
74 
75 #define IXGBE_MMW_SIZE_DEFAULT        0x4
76 #define IXGBE_MMW_SIZE_JUMBO_FRAME    0x14
77 #define IXGBE_MAX_RING_DESC           8192 /* replicate define from rxtx */
78 
79 /*
80  *  Default values for RX/TX configuration
81  */
82 #define IXGBE_DEFAULT_RX_FREE_THRESH  32
83 #define IXGBE_DEFAULT_RX_PTHRESH      8
84 #define IXGBE_DEFAULT_RX_HTHRESH      8
85 #define IXGBE_DEFAULT_RX_WTHRESH      0
86 
87 #define IXGBE_DEFAULT_TX_FREE_THRESH  32
88 #define IXGBE_DEFAULT_TX_PTHRESH      32
89 #define IXGBE_DEFAULT_TX_HTHRESH      0
90 #define IXGBE_DEFAULT_TX_WTHRESH      0
91 #define IXGBE_DEFAULT_TX_RSBIT_THRESH 32
92 
93 /* Bit shift and mask */
94 #define IXGBE_4_BIT_WIDTH  (CHAR_BIT / 2)
95 #define IXGBE_4_BIT_MASK   RTE_LEN2MASK(IXGBE_4_BIT_WIDTH, uint8_t)
96 #define IXGBE_8_BIT_WIDTH  CHAR_BIT
97 #define IXGBE_8_BIT_MASK   UINT8_MAX
98 
99 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
100 
101 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
102 
103 /* Additional timesync values. */
104 #define NSEC_PER_SEC             1000000000L
105 #define IXGBE_INCVAL_10GB        0x66666666
106 #define IXGBE_INCVAL_1GB         0x40000000
107 #define IXGBE_INCVAL_100         0x50000000
108 #define IXGBE_INCVAL_SHIFT_10GB  28
109 #define IXGBE_INCVAL_SHIFT_1GB   24
110 #define IXGBE_INCVAL_SHIFT_100   21
111 #define IXGBE_INCVAL_SHIFT_82599 7
112 #define IXGBE_INCPER_SHIFT_82599 24
113 
114 #define IXGBE_CYCLECOUNTER_MASK   0xffffffffffffffffULL
115 
116 #define IXGBE_VT_CTL_POOLING_MODE_MASK         0x00030000
117 #define IXGBE_VT_CTL_POOLING_MODE_ETAG         0x00010000
118 #define IXGBE_ETAG_ETYPE                       0x00005084
119 #define IXGBE_ETAG_ETYPE_MASK                  0x0000ffff
120 #define IXGBE_ETAG_ETYPE_VALID                 0x80000000
121 #define IXGBE_RAH_ADTYPE                       0x40000000
122 #define IXGBE_RAL_ETAG_FILTER_MASK             0x00003fff
123 #define IXGBE_VMVIR_TAGA_MASK                  0x18000000
124 #define IXGBE_VMVIR_TAGA_ETAG_INSERT           0x08000000
125 #define IXGBE_VMTIR(_i) (0x00017000 + ((_i) * 4)) /* 64 of these (0-63) */
126 #define IXGBE_QDE_STRIP_TAG                    0x00000004
127 #define IXGBE_VTEICR_MASK                      0x07
128 
129 #define IXGBE_EXVET_VET_EXT_SHIFT              16
130 #define IXGBE_DMATXCTL_VT_MASK                 0xFFFF0000
131 
132 #define IXGBE_DEVARG_FIBER_SDP3_NOT_TX_DISABLE	"fiber_sdp3_no_tx_disable"
133 
134 static const char * const ixgbe_valid_arguments[] = {
135 	IXGBE_DEVARG_FIBER_SDP3_NOT_TX_DISABLE,
136 	NULL
137 };
138 
139 #define IXGBEVF_DEVARG_PFLINK_FULLCHK		"pflink_fullchk"
140 
141 static const char * const ixgbevf_valid_arguments[] = {
142 	IXGBEVF_DEVARG_PFLINK_FULLCHK,
143 	NULL
144 };
145 
146 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params);
147 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
148 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev);
149 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev);
150 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev);
151 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev);
152 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev);
153 static int ixgbe_dev_configure(struct rte_eth_dev *dev);
154 static int ixgbe_dev_start(struct rte_eth_dev *dev);
155 static int ixgbe_dev_stop(struct rte_eth_dev *dev);
156 static int ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
157 static int ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
158 static int ixgbe_dev_close(struct rte_eth_dev *dev);
159 static int ixgbe_dev_reset(struct rte_eth_dev *dev);
160 static int ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
161 static int ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
162 static int ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
163 static int ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
164 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
165 				int wait_to_complete);
166 static int ixgbe_dev_stats_get(struct rte_eth_dev *dev,
167 				struct rte_eth_stats *stats);
168 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
169 				struct rte_eth_xstat *xstats, unsigned n);
170 static int ixgbevf_dev_xstats_get(struct rte_eth_dev *dev,
171 				  struct rte_eth_xstat *xstats, unsigned n);
172 static int
173 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
174 		uint64_t *values, unsigned int n);
175 static int ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
176 static int ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
177 static int ixgbe_dev_xstats_get_names(struct rte_eth_dev *dev,
178 	struct rte_eth_xstat_name *xstats_names,
179 	unsigned int size);
180 static int ixgbevf_dev_xstats_get_names(struct rte_eth_dev *dev,
181 	struct rte_eth_xstat_name *xstats_names, unsigned limit);
182 static int ixgbe_dev_xstats_get_names_by_id(
183 	struct rte_eth_dev *dev,
184 	const uint64_t *ids,
185 	struct rte_eth_xstat_name *xstats_names,
186 	unsigned int limit);
187 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
188 					     uint16_t queue_id,
189 					     uint8_t stat_idx,
190 					     uint8_t is_rx);
191 static int ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
192 				 size_t fw_size);
193 static int ixgbe_dev_info_get(struct rte_eth_dev *dev,
194 			      struct rte_eth_dev_info *dev_info);
195 static const uint32_t *ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev,
196 						      size_t *no_of_elements);
197 static int ixgbevf_dev_info_get(struct rte_eth_dev *dev,
198 				struct rte_eth_dev_info *dev_info);
199 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
200 
201 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
202 		uint16_t vlan_id, int on);
203 static int ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
204 			       enum rte_vlan_type vlan_type,
205 			       uint16_t tpid_id);
206 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
207 		uint16_t queue, bool on);
208 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
209 		int on);
210 static void ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev,
211 						  int mask);
212 static int ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask);
213 static int ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
214 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
215 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
216 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
217 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
218 
219 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
220 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
221 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
222 			       struct rte_eth_fc_conf *fc_conf);
223 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
224 			       struct rte_eth_fc_conf *fc_conf);
225 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
226 		struct rte_eth_pfc_conf *pfc_conf);
227 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
228 			struct rte_eth_rss_reta_entry64 *reta_conf,
229 			uint16_t reta_size);
230 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
231 			struct rte_eth_rss_reta_entry64 *reta_conf,
232 			uint16_t reta_size);
233 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
234 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
235 static int ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev);
236 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
237 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
238 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev);
239 static void ixgbe_dev_interrupt_handler(void *param);
240 static void ixgbe_dev_interrupt_delayed_handler(void *param);
241 static uint32_t ixgbe_dev_setup_link_thread_handler(void *param);
242 static int ixgbe_dev_wait_setup_link_complete(struct rte_eth_dev *dev,
243 					      uint32_t timeout_ms);
244 
245 static int ixgbe_add_rar(struct rte_eth_dev *dev,
246 			struct rte_ether_addr *mac_addr,
247 			uint32_t index, uint32_t pool);
248 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
249 static int ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
250 					   struct rte_ether_addr *mac_addr);
251 static void ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config);
252 static bool is_device_supported(struct rte_eth_dev *dev,
253 				struct rte_pci_driver *drv);
254 
255 /* For Virtual Function support */
256 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
257 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
258 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
259 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
260 static int ixgbevf_dev_link_update(struct rte_eth_dev *dev,
261 				   int wait_to_complete);
262 static int ixgbevf_dev_stop(struct rte_eth_dev *dev);
263 static int ixgbevf_dev_close(struct rte_eth_dev *dev);
264 static int  ixgbevf_dev_reset(struct rte_eth_dev *dev);
265 static void ixgbevf_intr_disable(struct rte_eth_dev *dev);
266 static void ixgbevf_intr_enable(struct rte_eth_dev *dev);
267 static int ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
268 		struct rte_eth_stats *stats);
269 static int ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
270 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
271 		uint16_t vlan_id, int on);
272 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
273 		uint16_t queue, int on);
274 static int ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask);
275 static int ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
276 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
277 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
278 					    uint16_t queue_id);
279 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
280 					     uint16_t queue_id);
281 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
282 				 uint8_t queue, uint8_t msix_vector);
283 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
284 static int ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev);
285 static int ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev);
286 static int ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev);
287 static int ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev);
288 
289 /* For Eth VMDQ APIs support */
290 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
291 		rte_ether_addr * mac_addr, uint8_t on);
292 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on);
293 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
294 					  uint16_t queue_id);
295 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
296 					   uint16_t queue_id);
297 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
298 			       uint8_t queue, uint8_t msix_vector);
299 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
300 
301 static int ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
302 				struct rte_ether_addr *mac_addr,
303 				uint32_t index, uint32_t pool);
304 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
305 static int ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
306 					     struct rte_ether_addr *mac_addr);
307 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
308 			struct ixgbe_5tuple_filter *filter);
309 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
310 			struct ixgbe_5tuple_filter *filter);
311 static int ixgbe_dev_flow_ops_get(struct rte_eth_dev *dev,
312 				  const struct rte_flow_ops **ops);
313 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
314 
315 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
316 				      struct rte_ether_addr *mc_addr_set,
317 				      uint32_t nb_mc_addr);
318 static int ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
319 				   struct rte_eth_dcb_info *dcb_info);
320 
321 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
322 static int ixgbe_get_regs(struct rte_eth_dev *dev,
323 			    struct rte_dev_reg_info *regs);
324 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
325 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
326 				struct rte_dev_eeprom_info *eeprom);
327 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
328 				struct rte_dev_eeprom_info *eeprom);
329 
330 static int ixgbe_get_module_info(struct rte_eth_dev *dev,
331 				 struct rte_eth_dev_module_info *modinfo);
332 static int ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
333 				   struct rte_dev_eeprom_info *info);
334 
335 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
336 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
337 				struct rte_dev_reg_info *regs);
338 
339 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
340 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
341 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
342 					    struct timespec *timestamp,
343 					    uint32_t flags);
344 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
345 					    struct timespec *timestamp);
346 static int ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
347 static int ixgbe_timesync_read_time(struct rte_eth_dev *dev,
348 				   struct timespec *timestamp);
349 static int ixgbe_timesync_write_time(struct rte_eth_dev *dev,
350 				   const struct timespec *timestamp);
351 static void ixgbevf_dev_interrupt_handler(void *param);
352 
353 static int ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
354 					 struct rte_eth_udp_tunnel *udp_tunnel);
355 static int ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
356 					 struct rte_eth_udp_tunnel *udp_tunnel);
357 static int ixgbe_filter_restore(struct rte_eth_dev *dev);
358 static void ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev);
359 static int ixgbe_wait_for_link_up(struct ixgbe_hw *hw);
360 static int devarg_handle_int(__rte_unused const char *key, const char *value,
361 			     void *extra_args);
362 
363 /*
364  * Define VF Stats MACRO for Non "cleared on read" register
365  */
366 #define UPDATE_VF_STAT(reg, last, cur)                          \
367 {                                                               \
368 	uint32_t latest = IXGBE_READ_REG(hw, reg);              \
369 	cur += (latest - last) & UINT_MAX;                      \
370 	last = latest;                                          \
371 }
372 
373 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
374 {                                                                \
375 	u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
376 	u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
377 	u64 latest = ((new_msb << 32) | new_lsb);                \
378 	cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
379 	last = latest;                                           \
380 }
381 
382 #define IXGBE_SET_HWSTRIP(h, q) do {\
383 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
384 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
385 		(h)->bitmap[idx] |= 1 << bit;\
386 	} while (0)
387 
388 #define IXGBE_CLEAR_HWSTRIP(h, q) do {\
389 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
390 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
391 		(h)->bitmap[idx] &= ~(1 << bit);\
392 	} while (0)
393 
394 #define IXGBE_GET_HWSTRIP(h, q, r) do {\
395 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
396 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
397 		(r) = (h)->bitmap[idx] >> bit & 1;\
398 	} while (0)
399 
400 /*
401  * The set of PCI devices this driver supports
402  */
403 static const struct rte_pci_id pci_id_ixgbe_map[] = {
404 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598) },
405 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_BX) },
406 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT) },
407 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT) },
408 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT) },
409 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2) },
410 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM) },
411 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4) },
412 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT) },
413 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT) },
414 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM) },
415 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR) },
416 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4) },
417 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ) },
418 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KR) },
419 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE) },
420 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4) },
421 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP) },
422 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE) },
423 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE) },
424 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_EM) },
425 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2) },
426 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP) },
427 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP) },
428 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP) },
429 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM) },
430 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM) },
431 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_LS) },
432 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T) },
433 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1) },
434 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP) },
435 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T) },
436 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_1G_T) },
437 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T) },
438 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1) },
439 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR) },
440 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR_L) },
441 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP_N) },
442 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII) },
443 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII_L) },
444 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_10G_T) },
445 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP) },
446 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP_N) },
447 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP) },
448 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T) },
449 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T_L) },
450 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4) },
451 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR) },
452 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_XFI) },
453 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_E610_10G_T) },
454 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_E610_2_5G_T) },
455 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_E610_BACKPLANE) },
456 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_E610_SFP) },
457 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_E610_SGMII) },
458 #ifdef RTE_LIBRTE_IXGBE_BYPASS
459 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BYPASS) },
460 #endif
461 	{ .vendor_id = 0, /* sentinel */ },
462 };
463 
464 /*
465  * The set of PCI devices this driver supports (for 82599 VF)
466  */
467 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
468 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF) },
469 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF_HV) },
470 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF) },
471 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF_HV) },
472 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF_HV) },
473 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF) },
474 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF) },
475 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF_HV) },
476 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF) },
477 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF_HV) },
478 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_E610_VF) },
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 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
534 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
535 	.dev_led_on           = ixgbe_dev_led_on,
536 	.dev_led_off          = ixgbe_dev_led_off,
537 	.flow_ctrl_get        = ixgbe_flow_ctrl_get,
538 	.flow_ctrl_set        = ixgbe_flow_ctrl_set,
539 	.priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
540 	.mac_addr_add         = ixgbe_add_rar,
541 	.mac_addr_remove      = ixgbe_remove_rar,
542 	.mac_addr_set         = ixgbe_set_default_mac_addr,
543 	.uc_hash_table_set    = ixgbe_uc_hash_table_set,
544 	.uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
545 	.set_queue_rate_limit = ixgbe_set_queue_rate_limit,
546 	.reta_update          = ixgbe_dev_rss_reta_update,
547 	.reta_query           = ixgbe_dev_rss_reta_query,
548 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
549 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
550 	.flow_ops_get         = ixgbe_dev_flow_ops_get,
551 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
552 	.rxq_info_get         = ixgbe_rxq_info_get,
553 	.txq_info_get         = ixgbe_txq_info_get,
554 	.recycle_rxq_info_get = ixgbe_recycle_rxq_info_get,
555 	.timesync_enable      = ixgbe_timesync_enable,
556 	.timesync_disable     = ixgbe_timesync_disable,
557 	.timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
558 	.timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
559 	.get_reg              = ixgbe_get_regs,
560 	.get_eeprom_length    = ixgbe_get_eeprom_length,
561 	.get_eeprom           = ixgbe_get_eeprom,
562 	.set_eeprom           = ixgbe_set_eeprom,
563 	.get_module_info      = ixgbe_get_module_info,
564 	.get_module_eeprom    = ixgbe_get_module_eeprom,
565 	.get_dcb_info         = ixgbe_dev_get_dcb_info,
566 	.timesync_adjust_time = ixgbe_timesync_adjust_time,
567 	.timesync_read_time   = ixgbe_timesync_read_time,
568 	.timesync_write_time  = ixgbe_timesync_write_time,
569 	.udp_tunnel_port_add  = ixgbe_dev_udp_tunnel_port_add,
570 	.udp_tunnel_port_del  = ixgbe_dev_udp_tunnel_port_del,
571 	.tm_ops_get           = ixgbe_tm_ops_get,
572 	.tx_done_cleanup      = ixgbe_dev_tx_done_cleanup,
573 	.get_monitor_addr     = ixgbe_get_monitor_addr,
574 };
575 
576 /*
577  * dev_ops for virtual function, bare necessities for basic vf
578  * operation have been implemented
579  */
580 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
581 	.dev_configure        = ixgbevf_dev_configure,
582 	.dev_start            = ixgbevf_dev_start,
583 	.dev_stop             = ixgbevf_dev_stop,
584 	.link_update          = ixgbevf_dev_link_update,
585 	.stats_get            = ixgbevf_dev_stats_get,
586 	.xstats_get           = ixgbevf_dev_xstats_get,
587 	.stats_reset          = ixgbevf_dev_stats_reset,
588 	.xstats_reset         = ixgbevf_dev_stats_reset,
589 	.xstats_get_names     = ixgbevf_dev_xstats_get_names,
590 	.dev_close            = ixgbevf_dev_close,
591 	.dev_reset	      = ixgbevf_dev_reset,
592 	.promiscuous_enable   = ixgbevf_dev_promiscuous_enable,
593 	.promiscuous_disable  = ixgbevf_dev_promiscuous_disable,
594 	.allmulticast_enable  = ixgbevf_dev_allmulticast_enable,
595 	.allmulticast_disable = ixgbevf_dev_allmulticast_disable,
596 	.dev_infos_get        = ixgbevf_dev_info_get,
597 	.dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
598 	.mtu_set              = ixgbevf_dev_set_mtu,
599 	.vlan_filter_set      = ixgbevf_vlan_filter_set,
600 	.vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
601 	.vlan_offload_set     = ixgbevf_vlan_offload_set,
602 	.rx_queue_setup       = ixgbe_dev_rx_queue_setup,
603 	.rx_queue_release     = ixgbe_dev_rx_queue_release,
604 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
605 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
606 	.rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
607 	.rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
608 	.mac_addr_add         = ixgbevf_add_mac_addr,
609 	.mac_addr_remove      = ixgbevf_remove_mac_addr,
610 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
611 	.rxq_info_get         = ixgbe_rxq_info_get,
612 	.txq_info_get         = ixgbe_txq_info_get,
613 	.mac_addr_set         = ixgbevf_set_default_mac_addr,
614 	.get_reg              = ixgbevf_get_regs,
615 	.reta_update          = ixgbe_dev_rss_reta_update,
616 	.reta_query           = ixgbe_dev_rss_reta_query,
617 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
618 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
619 	.tx_done_cleanup      = ixgbe_dev_tx_done_cleanup,
620 	.get_monitor_addr     = ixgbe_get_monitor_addr,
621 };
622 
623 /* store statistics names and its offset in stats structure */
624 struct rte_ixgbe_xstats_name_off {
625 	char name[RTE_ETH_XSTATS_NAME_SIZE];
626 	unsigned offset;
627 };
628 
629 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
630 	{"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
631 	{"rx_illegal_byte_errors", offsetof(struct ixgbe_hw_stats, illerrc)},
632 	{"rx_error_bytes", offsetof(struct ixgbe_hw_stats, errbc)},
633 	{"mac_local_errors", offsetof(struct ixgbe_hw_stats, mlfc)},
634 	{"mac_remote_errors", offsetof(struct ixgbe_hw_stats, mrfc)},
635 	{"rx_length_errors", offsetof(struct ixgbe_hw_stats, rlec)},
636 	{"tx_xon_packets", offsetof(struct ixgbe_hw_stats, lxontxc)},
637 	{"rx_xon_packets", offsetof(struct ixgbe_hw_stats, lxonrxc)},
638 	{"tx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxofftxc)},
639 	{"rx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
640 	{"rx_size_64_packets", offsetof(struct ixgbe_hw_stats, prc64)},
641 	{"rx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, prc127)},
642 	{"rx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, prc255)},
643 	{"rx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, prc511)},
644 	{"rx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
645 		prc1023)},
646 	{"rx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
647 		prc1522)},
648 	{"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
649 	{"rx_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
650 	{"rx_fragment_errors", offsetof(struct ixgbe_hw_stats, rfc)},
651 	{"rx_undersize_errors", offsetof(struct ixgbe_hw_stats, ruc)},
652 	{"rx_oversize_errors", offsetof(struct ixgbe_hw_stats, roc)},
653 	{"rx_jabber_errors", offsetof(struct ixgbe_hw_stats, rjc)},
654 	{"rx_management_packets", offsetof(struct ixgbe_hw_stats, mngprc)},
655 	{"rx_management_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
656 	{"tx_management_packets", offsetof(struct ixgbe_hw_stats, mngptc)},
657 	{"rx_total_packets", offsetof(struct ixgbe_hw_stats, tpr)},
658 	{"rx_total_bytes", offsetof(struct ixgbe_hw_stats, tor)},
659 	{"tx_total_packets", offsetof(struct ixgbe_hw_stats, tpt)},
660 	{"tx_size_64_packets", offsetof(struct ixgbe_hw_stats, ptc64)},
661 	{"tx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, ptc127)},
662 	{"tx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, ptc255)},
663 	{"tx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, ptc511)},
664 	{"tx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
665 		ptc1023)},
666 	{"tx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
667 		ptc1522)},
668 	{"tx_multicast_packets", offsetof(struct ixgbe_hw_stats, mptc)},
669 	{"tx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bptc)},
670 	{"rx_mac_short_packet_dropped", offsetof(struct ixgbe_hw_stats, mspdc)},
671 	{"rx_l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
672 
673 	{"flow_director_added_filters", offsetof(struct ixgbe_hw_stats,
674 		fdirustat_add)},
675 	{"flow_director_removed_filters", offsetof(struct ixgbe_hw_stats,
676 		fdirustat_remove)},
677 	{"flow_director_filter_add_errors", offsetof(struct ixgbe_hw_stats,
678 		fdirfstat_fadd)},
679 	{"flow_director_filter_remove_errors", offsetof(struct ixgbe_hw_stats,
680 		fdirfstat_fremove)},
681 	{"flow_director_matched_filters", offsetof(struct ixgbe_hw_stats,
682 		fdirmatch)},
683 	{"flow_director_missed_filters", offsetof(struct ixgbe_hw_stats,
684 		fdirmiss)},
685 
686 	{"rx_fcoe_crc_errors", offsetof(struct ixgbe_hw_stats, fccrc)},
687 	{"rx_fcoe_dropped", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
688 	{"rx_fcoe_mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats,
689 		fclast)},
690 	{"rx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeprc)},
691 	{"tx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeptc)},
692 	{"rx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwrc)},
693 	{"tx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwtc)},
694 	{"rx_fcoe_no_direct_data_placement", offsetof(struct ixgbe_hw_stats,
695 		fcoe_noddp)},
696 	{"rx_fcoe_no_direct_data_placement_ext_buff",
697 		offsetof(struct ixgbe_hw_stats, fcoe_noddp_ext_buff)},
698 
699 	{"tx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
700 		lxontxc)},
701 	{"rx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
702 		lxonrxc)},
703 	{"tx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
704 		lxofftxc)},
705 	{"rx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
706 		lxoffrxc)},
707 	{"rx_total_missed_packets", offsetof(struct ixgbe_hw_stats, mpctotal)},
708 };
709 
710 #define IXGBE_NB_HW_STATS (sizeof(rte_ixgbe_stats_strings) / \
711 			   sizeof(rte_ixgbe_stats_strings[0]))
712 
713 /* MACsec statistics */
714 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_macsec_strings[] = {
715 	{"out_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
716 		out_pkts_untagged)},
717 	{"out_pkts_encrypted", offsetof(struct ixgbe_macsec_stats,
718 		out_pkts_encrypted)},
719 	{"out_pkts_protected", offsetof(struct ixgbe_macsec_stats,
720 		out_pkts_protected)},
721 	{"out_octets_encrypted", offsetof(struct ixgbe_macsec_stats,
722 		out_octets_encrypted)},
723 	{"out_octets_protected", offsetof(struct ixgbe_macsec_stats,
724 		out_octets_protected)},
725 	{"in_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
726 		in_pkts_untagged)},
727 	{"in_pkts_badtag", offsetof(struct ixgbe_macsec_stats,
728 		in_pkts_badtag)},
729 	{"in_pkts_nosci", offsetof(struct ixgbe_macsec_stats,
730 		in_pkts_nosci)},
731 	{"in_pkts_unknownsci", offsetof(struct ixgbe_macsec_stats,
732 		in_pkts_unknownsci)},
733 	{"in_octets_decrypted", offsetof(struct ixgbe_macsec_stats,
734 		in_octets_decrypted)},
735 	{"in_octets_validated", offsetof(struct ixgbe_macsec_stats,
736 		in_octets_validated)},
737 	{"in_pkts_unchecked", offsetof(struct ixgbe_macsec_stats,
738 		in_pkts_unchecked)},
739 	{"in_pkts_delayed", offsetof(struct ixgbe_macsec_stats,
740 		in_pkts_delayed)},
741 	{"in_pkts_late", offsetof(struct ixgbe_macsec_stats,
742 		in_pkts_late)},
743 	{"in_pkts_ok", offsetof(struct ixgbe_macsec_stats,
744 		in_pkts_ok)},
745 	{"in_pkts_invalid", offsetof(struct ixgbe_macsec_stats,
746 		in_pkts_invalid)},
747 	{"in_pkts_notvalid", offsetof(struct ixgbe_macsec_stats,
748 		in_pkts_notvalid)},
749 	{"in_pkts_unusedsa", offsetof(struct ixgbe_macsec_stats,
750 		in_pkts_unusedsa)},
751 	{"in_pkts_notusingsa", offsetof(struct ixgbe_macsec_stats,
752 		in_pkts_notusingsa)},
753 };
754 
755 #define IXGBE_NB_MACSEC_STATS (sizeof(rte_ixgbe_macsec_strings) / \
756 			   sizeof(rte_ixgbe_macsec_strings[0]))
757 
758 /* Per-queue statistics */
759 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_rxq_strings[] = {
760 	{"mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats, rnbc)},
761 	{"dropped", offsetof(struct ixgbe_hw_stats, mpc)},
762 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxonrxc)},
763 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxoffrxc)},
764 };
765 
766 #define IXGBE_NB_RXQ_PRIO_STATS (sizeof(rte_ixgbe_rxq_strings) / \
767 			   sizeof(rte_ixgbe_rxq_strings[0]))
768 #define IXGBE_NB_RXQ_PRIO_VALUES 8
769 
770 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_txq_strings[] = {
771 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxontxc)},
772 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxofftxc)},
773 	{"xon_to_xoff_packets", offsetof(struct ixgbe_hw_stats,
774 		pxon2offc)},
775 };
776 
777 #define IXGBE_NB_TXQ_PRIO_STATS (sizeof(rte_ixgbe_txq_strings) / \
778 			   sizeof(rte_ixgbe_txq_strings[0]))
779 #define IXGBE_NB_TXQ_PRIO_VALUES 8
780 
781 static const struct rte_ixgbe_xstats_name_off rte_ixgbevf_stats_strings[] = {
782 	{"rx_multicast_packets", offsetof(struct ixgbevf_hw_stats, vfmprc)},
783 };
784 
785 #define IXGBEVF_NB_XSTATS (sizeof(rte_ixgbevf_stats_strings) /	\
786 		sizeof(rte_ixgbevf_stats_strings[0]))
787 
788 /*
789  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
790  */
791 static inline int
792 ixgbe_is_sfp(struct ixgbe_hw *hw)
793 {
794 	switch (hw->phy.type) {
795 	case ixgbe_phy_sfp_avago:
796 	case ixgbe_phy_sfp_ftl:
797 	case ixgbe_phy_sfp_intel:
798 	case ixgbe_phy_sfp_unknown:
799 	case ixgbe_phy_sfp_passive_tyco:
800 	case ixgbe_phy_sfp_passive_unknown:
801 		return 1;
802 	default:
803 		/* x550em devices may be SFP, check media type */
804 		switch (hw->mac.type) {
805 		case ixgbe_mac_X550EM_x:
806 		case ixgbe_mac_X550EM_a:
807 			switch (ixgbe_get_media_type(hw)) {
808 			case ixgbe_media_type_fiber:
809 			case ixgbe_media_type_fiber_qsfp:
810 				return 1;
811 			default:
812 				break;
813 			}
814 		default:
815 			break;
816 		}
817 		return 0;
818 	}
819 }
820 
821 static inline int32_t
822 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
823 {
824 	uint32_t ctrl_ext;
825 	int32_t status;
826 
827 	status = ixgbe_reset_hw(hw);
828 
829 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
830 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
831 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
832 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
833 	IXGBE_WRITE_FLUSH(hw);
834 
835 	if (status == IXGBE_ERR_SFP_NOT_PRESENT)
836 		status = IXGBE_SUCCESS;
837 	return status;
838 }
839 
840 static inline void
841 ixgbe_enable_intr(struct rte_eth_dev *dev)
842 {
843 	struct ixgbe_interrupt *intr =
844 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
845 	struct ixgbe_hw *hw =
846 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
847 
848 	IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
849 	IXGBE_WRITE_FLUSH(hw);
850 }
851 
852 /*
853  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
854  */
855 static void
856 ixgbe_disable_intr(struct ixgbe_hw *hw)
857 {
858 	PMD_INIT_FUNC_TRACE();
859 
860 	if (hw->mac.type == ixgbe_mac_82598EB) {
861 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
862 	} else {
863 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
864 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
865 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
866 	}
867 	IXGBE_WRITE_FLUSH(hw);
868 }
869 
870 /*
871  * This function resets queue statistics mapping registers.
872  * From Niantic datasheet, Initialization of Statistics section:
873  * "...if software requires the queue counters, the RQSMR and TQSM registers
874  * must be re-programmed following a device reset.
875  */
876 static void
877 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
878 {
879 	uint32_t i;
880 
881 	for (i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
882 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
883 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
884 	}
885 }
886 
887 
888 static int
889 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
890 				  uint16_t queue_id,
891 				  uint8_t stat_idx,
892 				  uint8_t is_rx)
893 {
894 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
895 #define NB_QMAP_FIELDS_PER_QSM_REG 4
896 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
897 
898 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
899 	struct ixgbe_stat_mapping_registers *stat_mappings =
900 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
901 	uint32_t qsmr_mask = 0;
902 	uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
903 	uint32_t q_map;
904 	uint8_t n, offset;
905 
906 	if ((hw->mac.type != ixgbe_mac_82599EB) &&
907 		(hw->mac.type != ixgbe_mac_X540) &&
908 		(hw->mac.type != ixgbe_mac_X550) &&
909 		(hw->mac.type != ixgbe_mac_X550EM_x) &&
910 		(hw->mac.type != ixgbe_mac_X550EM_a))
911 		return -ENOSYS;
912 
913 	PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
914 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
915 		     queue_id, stat_idx);
916 
917 	n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
918 	if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
919 		PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
920 		return -EIO;
921 	}
922 	offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
923 
924 	/* Now clear any previous stat_idx set */
925 	clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
926 	if (!is_rx)
927 		stat_mappings->tqsm[n] &= ~clearing_mask;
928 	else
929 		stat_mappings->rqsmr[n] &= ~clearing_mask;
930 
931 	q_map = (uint32_t)stat_idx;
932 	q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
933 	qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
934 	if (!is_rx)
935 		stat_mappings->tqsm[n] |= qsmr_mask;
936 	else
937 		stat_mappings->rqsmr[n] |= qsmr_mask;
938 
939 	PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
940 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
941 		     queue_id, stat_idx);
942 	PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
943 		     is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
944 
945 	/* Now write the mapping in the appropriate register */
946 	if (is_rx) {
947 		PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
948 			     stat_mappings->rqsmr[n], n);
949 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
950 	} else {
951 		PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
952 			     stat_mappings->tqsm[n], n);
953 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
954 	}
955 	return 0;
956 }
957 
958 static void
959 ixgbe_restore_statistics_mapping(struct rte_eth_dev *dev)
960 {
961 	struct ixgbe_stat_mapping_registers *stat_mappings =
962 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
963 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
964 	int i;
965 
966 	/* write whatever was in stat mapping table to the NIC */
967 	for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
968 		/* rx */
969 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
970 
971 		/* tx */
972 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
973 	}
974 }
975 
976 static void
977 ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config)
978 {
979 	uint8_t i;
980 	struct ixgbe_dcb_tc_config *tc;
981 	uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
982 
983 	dcb_config->num_tcs.pg_tcs = dcb_max_tc;
984 	dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
985 	for (i = 0; i < dcb_max_tc; i++) {
986 		tc = &dcb_config->tc_config[i];
987 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
988 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
989 				 (uint8_t)(100/dcb_max_tc + (i & 1));
990 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
991 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
992 				 (uint8_t)(100/dcb_max_tc + (i & 1));
993 		tc->pfc = ixgbe_dcb_pfc_disabled;
994 	}
995 
996 	/* Initialize default user to priority mapping, UPx->TC0 */
997 	tc = &dcb_config->tc_config[0];
998 	tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
999 	tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
1000 	for (i = 0; i < IXGBE_DCB_MAX_BW_GROUP; i++) {
1001 		dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
1002 		dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
1003 	}
1004 	dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
1005 	dcb_config->pfc_mode_enable = false;
1006 	dcb_config->vt_mode = true;
1007 	dcb_config->round_robin_enable = false;
1008 	/* support all DCB capabilities in 82599 */
1009 	dcb_config->support.capabilities = 0xFF;
1010 
1011 	/*we only support 4 Tcs for X540, X550 */
1012 	if (hw->mac.type == ixgbe_mac_X540 ||
1013 		hw->mac.type == ixgbe_mac_X550 ||
1014 		hw->mac.type == ixgbe_mac_X550EM_x ||
1015 		hw->mac.type == ixgbe_mac_X550EM_a) {
1016 		dcb_config->num_tcs.pg_tcs = 4;
1017 		dcb_config->num_tcs.pfc_tcs = 4;
1018 	}
1019 }
1020 
1021 /*
1022  * Ensure that all locks are released before first NVM or PHY access
1023  */
1024 static void
1025 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
1026 {
1027 	uint16_t mask;
1028 
1029 	/*
1030 	 * Phy lock should not fail in this early stage. If this is the case,
1031 	 * it is due to an improper exit of the application.
1032 	 * So force the release of the faulty lock. Release of common lock
1033 	 * is done automatically by swfw_sync function.
1034 	 */
1035 	mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
1036 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1037 		PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
1038 	}
1039 	ixgbe_release_swfw_semaphore(hw, mask);
1040 
1041 	/*
1042 	 * These ones are more tricky since they are common to all ports; but
1043 	 * swfw_sync retries last long enough (1s) to be almost sure that if
1044 	 * lock can not be taken it is due to an improper lock of the
1045 	 * semaphore.
1046 	 */
1047 	mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
1048 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1049 		PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1050 	}
1051 	ixgbe_release_swfw_semaphore(hw, mask);
1052 }
1053 
1054 static void
1055 ixgbe_parse_devargs(struct ixgbe_adapter *adapter,
1056 		      struct rte_devargs *devargs)
1057 {
1058 	struct rte_kvargs *kvlist;
1059 	uint16_t sdp3_no_tx_disable;
1060 
1061 	if (devargs == NULL)
1062 		return;
1063 
1064 	kvlist = rte_kvargs_parse(devargs->args, ixgbe_valid_arguments);
1065 	if (kvlist == NULL)
1066 		return;
1067 
1068 	if (rte_kvargs_count(kvlist, IXGBE_DEVARG_FIBER_SDP3_NOT_TX_DISABLE) == 1 &&
1069 	    rte_kvargs_process(kvlist, IXGBE_DEVARG_FIBER_SDP3_NOT_TX_DISABLE,
1070 			       devarg_handle_int, &sdp3_no_tx_disable) == 0 &&
1071 	    sdp3_no_tx_disable == 1)
1072 		adapter->sdp3_no_tx_disable = 1;
1073 
1074 	rte_kvargs_free(kvlist);
1075 }
1076 
1077 /*
1078  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
1079  * It returns 0 on success.
1080  */
1081 static int
1082 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params __rte_unused)
1083 {
1084 	struct ixgbe_adapter *ad = eth_dev->data->dev_private;
1085 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1086 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1087 	struct ixgbe_hw *hw =
1088 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1089 	struct ixgbe_vfta *shadow_vfta =
1090 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1091 	struct ixgbe_hwstrip *hwstrip =
1092 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1093 	struct ixgbe_dcb_config *dcb_config =
1094 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
1095 	struct ixgbe_filter_info *filter_info =
1096 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1097 	struct ixgbe_bw_conf *bw_conf =
1098 		IXGBE_DEV_PRIVATE_TO_BW_CONF(eth_dev->data->dev_private);
1099 	uint32_t ctrl_ext;
1100 	uint16_t csum;
1101 	int diag, i, ret;
1102 
1103 	PMD_INIT_FUNC_TRACE();
1104 
1105 	ixgbe_dev_macsec_setting_reset(eth_dev);
1106 
1107 	eth_dev->dev_ops = &ixgbe_eth_dev_ops;
1108 	eth_dev->rx_queue_count       = ixgbe_dev_rx_queue_count;
1109 	eth_dev->rx_descriptor_status = ixgbe_dev_rx_descriptor_status;
1110 	eth_dev->tx_descriptor_status = ixgbe_dev_tx_descriptor_status;
1111 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1112 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1113 	eth_dev->tx_pkt_prepare = &ixgbe_prep_pkts;
1114 
1115 	/*
1116 	 * For secondary processes, we don't initialise any further as primary
1117 	 * has already done this work. Only check we don't need a different
1118 	 * RX and TX function.
1119 	 */
1120 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1121 		struct ci_tx_queue *txq;
1122 		/* TX queue function in primary, set by last queue initialized
1123 		 * Tx queue may not initialized by primary process
1124 		 */
1125 		if (eth_dev->data->tx_queues) {
1126 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
1127 			ixgbe_set_tx_function(eth_dev, txq);
1128 		} else {
1129 			/* Use default TX function if we get here */
1130 			PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
1131 				     "Using default TX function.");
1132 		}
1133 
1134 		ixgbe_set_rx_function(eth_dev);
1135 
1136 		return 0;
1137 	}
1138 
1139 	/* NOTE: review for potential ordering optimization */
1140 	rte_atomic_store_explicit(&ad->link_thread_running, 0, rte_memory_order_seq_cst);
1141 	ixgbe_parse_devargs(eth_dev->data->dev_private,
1142 			    pci_dev->device.devargs);
1143 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1144 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1145 
1146 	/* Vendor and Device ID need to be set before init of shared code */
1147 	hw->device_id = pci_dev->id.device_id;
1148 	hw->vendor_id = pci_dev->id.vendor_id;
1149 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1150 	hw->allow_unsupported_sfp = 1;
1151 
1152 	/* Initialize the shared code (base driver) */
1153 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1154 	diag = ixgbe_bypass_init_shared_code(hw);
1155 #else
1156 	diag = ixgbe_init_shared_code(hw);
1157 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1158 
1159 	if (diag != IXGBE_SUCCESS) {
1160 		PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
1161 		return -EIO;
1162 	}
1163 
1164 	if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
1165 		PMD_INIT_LOG(ERR, "ERROR: Firmware recovery mode detected. Limiting functionality.");
1166 		return -EIO;
1167 	}
1168 
1169 	/* pick up the PCI bus settings for reporting later */
1170 	ixgbe_get_bus_info(hw);
1171 
1172 	/* Unlock any pending hardware semaphore */
1173 	ixgbe_swfw_lock_reset(hw);
1174 
1175 #ifdef RTE_LIB_SECURITY
1176 	/* Initialize security_ctx only for primary process*/
1177 	if (ixgbe_ipsec_ctx_create(eth_dev))
1178 		return -ENOMEM;
1179 #endif
1180 
1181 	/* Initialize DCB configuration*/
1182 	memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
1183 	ixgbe_dcb_init(hw, dcb_config);
1184 	/* Get Hardware Flow Control setting */
1185 	hw->fc.requested_mode = ixgbe_fc_none;
1186 	hw->fc.current_mode = ixgbe_fc_none;
1187 	hw->fc.pause_time = IXGBE_FC_PAUSE;
1188 	for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
1189 		hw->fc.low_water[i] = IXGBE_FC_LO;
1190 		hw->fc.high_water[i] = IXGBE_FC_HI;
1191 	}
1192 	hw->fc.send_xon = 1;
1193 
1194 	/* Make sure we have a good EEPROM before we read from it */
1195 	diag = ixgbe_validate_eeprom_checksum(hw, &csum);
1196 	if (diag != IXGBE_SUCCESS) {
1197 		PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
1198 		ret = -EIO;
1199 		goto err_exit;
1200 	}
1201 
1202 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1203 	diag = ixgbe_bypass_init_hw(hw);
1204 #else
1205 	diag = ixgbe_init_hw(hw);
1206 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1207 
1208 	/*
1209 	 * Devices with copper phys will fail to initialise if ixgbe_init_hw()
1210 	 * is called too soon after the kernel driver unbinding/binding occurs.
1211 	 * The failure occurs in ixgbe_identify_phy_generic() for all devices,
1212 	 * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
1213 	 * also called. See ixgbe_identify_phy_82599(). The reason for the
1214 	 * failure is not known, and only occuts when virtualisation features
1215 	 * are disabled in the bios. A delay of 100ms  was found to be enough by
1216 	 * trial-and-error, and is doubled to be safe.
1217 	 */
1218 	if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
1219 		rte_delay_ms(200);
1220 		diag = ixgbe_init_hw(hw);
1221 	}
1222 
1223 	if (diag == IXGBE_ERR_SFP_NOT_PRESENT)
1224 		diag = IXGBE_SUCCESS;
1225 
1226 	if (diag == IXGBE_ERR_EEPROM_VERSION) {
1227 		PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
1228 			     "LOM.  Please be aware there may be issues associated "
1229 			     "with your hardware.");
1230 		PMD_INIT_LOG(ERR, "If you are experiencing problems "
1231 			     "please contact your Intel or hardware representative "
1232 			     "who provided you with this hardware.");
1233 	} else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
1234 		PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
1235 	if (diag) {
1236 		PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
1237 		ret = -EIO;
1238 		goto err_exit;
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", RTE_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 			     RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1258 		ret = -ENOMEM;
1259 		goto err_exit;
1260 	}
1261 	/* Copy the permanent MAC address */
1262 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.perm_addr,
1263 			&eth_dev->data->mac_addrs[0]);
1264 
1265 	/* Allocate memory for storing hash filter MAC addresses */
1266 	eth_dev->data->hash_mac_addrs = rte_zmalloc(
1267 		"ixgbe", RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC, 0);
1268 	if (eth_dev->data->hash_mac_addrs == NULL) {
1269 		PMD_INIT_LOG(ERR,
1270 			     "Failed to allocate %d bytes needed to store MAC addresses",
1271 			     RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
1272 		rte_free(eth_dev->data->mac_addrs);
1273 		eth_dev->data->mac_addrs = NULL;
1274 		ret = -ENOMEM;
1275 		goto err_exit;
1276 	}
1277 
1278 	/* initialize the vfta */
1279 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1280 
1281 	/* initialize the hw strip bitmap*/
1282 	memset(hwstrip, 0, sizeof(*hwstrip));
1283 
1284 	/* initialize PF if max_vfs not zero */
1285 	ret = ixgbe_pf_host_init(eth_dev);
1286 	if (ret)
1287 		goto err_pf_host_init;
1288 
1289 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1290 	/* let hardware know driver is loaded */
1291 	ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1292 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1293 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1294 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1295 	IXGBE_WRITE_FLUSH(hw);
1296 
1297 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1298 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1299 			     (int) hw->mac.type, (int) hw->phy.type,
1300 			     (int) hw->phy.sfp_type);
1301 	else
1302 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1303 			     (int) hw->mac.type, (int) hw->phy.type);
1304 
1305 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1306 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1307 		     pci_dev->id.device_id);
1308 
1309 	rte_intr_callback_register(intr_handle,
1310 				   ixgbe_dev_interrupt_handler, eth_dev);
1311 
1312 	/* enable uio/vfio intr/eventfd mapping */
1313 	rte_intr_enable(intr_handle);
1314 
1315 	/* enable support intr */
1316 	ixgbe_enable_intr(eth_dev);
1317 
1318 	/* initialize filter info */
1319 	memset(filter_info, 0,
1320 	       sizeof(struct ixgbe_filter_info));
1321 
1322 	/* initialize 5tuple filter list */
1323 	TAILQ_INIT(&filter_info->fivetuple_list);
1324 
1325 	/* initialize flow director filter list & hash */
1326 	ret = ixgbe_fdir_filter_init(eth_dev);
1327 	if (ret)
1328 		goto err_fdir_filter_init;
1329 
1330 	/* initialize l2 tunnel filter list & hash */
1331 	ret = ixgbe_l2_tn_filter_init(eth_dev);
1332 	if (ret)
1333 		goto err_l2_tn_filter_init;
1334 
1335 	/* initialize flow filter lists */
1336 	ixgbe_filterlist_init();
1337 
1338 	/* initialize bandwidth configuration info */
1339 	memset(bw_conf, 0, sizeof(struct ixgbe_bw_conf));
1340 
1341 	/* initialize Traffic Manager configuration */
1342 	ixgbe_tm_conf_init(eth_dev);
1343 
1344 	return 0;
1345 
1346 err_l2_tn_filter_init:
1347 	ixgbe_fdir_filter_uninit(eth_dev);
1348 err_fdir_filter_init:
1349 	ixgbe_disable_intr(hw);
1350 	rte_intr_disable(intr_handle);
1351 	rte_intr_callback_unregister(intr_handle,
1352 		ixgbe_dev_interrupt_handler, eth_dev);
1353 	ixgbe_pf_host_uninit(eth_dev);
1354 err_pf_host_init:
1355 	rte_free(eth_dev->data->mac_addrs);
1356 	eth_dev->data->mac_addrs = NULL;
1357 	rte_free(eth_dev->data->hash_mac_addrs);
1358 	eth_dev->data->hash_mac_addrs = NULL;
1359 err_exit:
1360 #ifdef RTE_LIB_SECURITY
1361 	rte_free(eth_dev->security_ctx);
1362 	eth_dev->security_ctx = NULL;
1363 #endif
1364 	return ret;
1365 }
1366 
1367 static int
1368 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1369 {
1370 	PMD_INIT_FUNC_TRACE();
1371 
1372 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1373 		return 0;
1374 
1375 	ixgbe_dev_close(eth_dev);
1376 
1377 	return 0;
1378 }
1379 
1380 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
1381 {
1382 	struct ixgbe_filter_info *filter_info =
1383 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1384 	struct ixgbe_5tuple_filter *p_5tuple;
1385 
1386 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
1387 		TAILQ_REMOVE(&filter_info->fivetuple_list,
1388 			     p_5tuple,
1389 			     entries);
1390 		rte_free(p_5tuple);
1391 	}
1392 	memset(filter_info->fivetuple_mask, 0,
1393 	       sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1394 
1395 	return 0;
1396 }
1397 
1398 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev)
1399 {
1400 	struct ixgbe_hw_fdir_info *fdir_info =
1401 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1402 	struct ixgbe_fdir_filter *fdir_filter;
1403 
1404 	rte_free(fdir_info->hash_map);
1405 	rte_hash_free(fdir_info->hash_handle);
1406 
1407 	while ((fdir_filter = TAILQ_FIRST(&fdir_info->fdir_list))) {
1408 		TAILQ_REMOVE(&fdir_info->fdir_list,
1409 			     fdir_filter,
1410 			     entries);
1411 		rte_free(fdir_filter);
1412 	}
1413 
1414 	return 0;
1415 }
1416 
1417 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev)
1418 {
1419 	struct ixgbe_l2_tn_info *l2_tn_info =
1420 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1421 	struct ixgbe_l2_tn_filter *l2_tn_filter;
1422 
1423 	rte_free(l2_tn_info->hash_map);
1424 	rte_hash_free(l2_tn_info->hash_handle);
1425 
1426 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
1427 		TAILQ_REMOVE(&l2_tn_info->l2_tn_list,
1428 			     l2_tn_filter,
1429 			     entries);
1430 		rte_free(l2_tn_filter);
1431 	}
1432 
1433 	return 0;
1434 }
1435 
1436 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev)
1437 {
1438 	struct ixgbe_hw_fdir_info *fdir_info =
1439 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1440 	char fdir_hash_name[RTE_HASH_NAMESIZE];
1441 	struct rte_hash_parameters fdir_hash_params = {
1442 		.name = fdir_hash_name,
1443 		.entries = IXGBE_MAX_FDIR_FILTER_NUM,
1444 		.key_len = sizeof(union ixgbe_atr_input),
1445 		.hash_func = rte_hash_crc,
1446 		.hash_func_init_val = 0,
1447 		.socket_id = rte_socket_id(),
1448 	};
1449 
1450 	TAILQ_INIT(&fdir_info->fdir_list);
1451 	snprintf(fdir_hash_name, RTE_HASH_NAMESIZE,
1452 		 "fdir_%s", eth_dev->device->name);
1453 	fdir_info->hash_handle = rte_hash_create(&fdir_hash_params);
1454 	if (!fdir_info->hash_handle) {
1455 		PMD_INIT_LOG(ERR, "Failed to create fdir hash table!");
1456 		return -EINVAL;
1457 	}
1458 	fdir_info->hash_map = rte_zmalloc("ixgbe",
1459 					  sizeof(struct ixgbe_fdir_filter *) *
1460 					  IXGBE_MAX_FDIR_FILTER_NUM,
1461 					  0);
1462 	if (!fdir_info->hash_map) {
1463 		PMD_INIT_LOG(ERR,
1464 			     "Failed to allocate memory for fdir hash map!");
1465 		rte_hash_free(fdir_info->hash_handle);
1466 		return -ENOMEM;
1467 	}
1468 	fdir_info->mask_added = FALSE;
1469 
1470 	return 0;
1471 }
1472 
1473 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev)
1474 {
1475 	struct ixgbe_l2_tn_info *l2_tn_info =
1476 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1477 	char l2_tn_hash_name[RTE_HASH_NAMESIZE];
1478 	struct rte_hash_parameters l2_tn_hash_params = {
1479 		.name = l2_tn_hash_name,
1480 		.entries = IXGBE_MAX_L2_TN_FILTER_NUM,
1481 		.key_len = sizeof(struct ixgbe_l2_tn_key),
1482 		.hash_func = rte_hash_crc,
1483 		.hash_func_init_val = 0,
1484 		.socket_id = rte_socket_id(),
1485 	};
1486 
1487 	TAILQ_INIT(&l2_tn_info->l2_tn_list);
1488 	snprintf(l2_tn_hash_name, RTE_HASH_NAMESIZE,
1489 		 "l2_tn_%s", eth_dev->device->name);
1490 	l2_tn_info->hash_handle = rte_hash_create(&l2_tn_hash_params);
1491 	if (!l2_tn_info->hash_handle) {
1492 		PMD_INIT_LOG(ERR, "Failed to create L2 TN hash table!");
1493 		return -EINVAL;
1494 	}
1495 	l2_tn_info->hash_map = rte_zmalloc("ixgbe",
1496 				   sizeof(struct ixgbe_l2_tn_filter *) *
1497 				   IXGBE_MAX_L2_TN_FILTER_NUM,
1498 				   0);
1499 	if (!l2_tn_info->hash_map) {
1500 		PMD_INIT_LOG(ERR,
1501 			"Failed to allocate memory for L2 TN hash map!");
1502 		rte_hash_free(l2_tn_info->hash_handle);
1503 		return -ENOMEM;
1504 	}
1505 	l2_tn_info->e_tag_en = FALSE;
1506 	l2_tn_info->e_tag_fwd_en = FALSE;
1507 	l2_tn_info->e_tag_ether_type = RTE_ETHER_TYPE_ETAG;
1508 
1509 	return 0;
1510 }
1511 /*
1512  * Negotiate mailbox API version with the PF.
1513  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1514  * Then we try to negotiate starting with the most recent one.
1515  * If all negotiation attempts fail, then we will proceed with
1516  * the default one (ixgbe_mbox_api_10).
1517  */
1518 static void
1519 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1520 {
1521 	int32_t i;
1522 
1523 	/* start with highest supported, proceed down */
1524 	static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1525 		ixgbe_mbox_api_13,
1526 		ixgbe_mbox_api_12,
1527 		ixgbe_mbox_api_11,
1528 		ixgbe_mbox_api_10,
1529 	};
1530 
1531 	for (i = 0;
1532 			i != RTE_DIM(sup_ver) &&
1533 			ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1534 			i++)
1535 		;
1536 }
1537 
1538 static void
1539 generate_random_mac_addr(struct rte_ether_addr *mac_addr)
1540 {
1541 	uint64_t random;
1542 
1543 	/* Set Organizationally Unique Identifier (OUI) prefix. */
1544 	mac_addr->addr_bytes[0] = 0x00;
1545 	mac_addr->addr_bytes[1] = 0x09;
1546 	mac_addr->addr_bytes[2] = 0xC0;
1547 	/* Force indication of locally assigned MAC address. */
1548 	mac_addr->addr_bytes[0] |= RTE_ETHER_LOCAL_ADMIN_ADDR;
1549 	/* Generate the last 3 bytes of the MAC address with a random number. */
1550 	random = rte_rand();
1551 	memcpy(&mac_addr->addr_bytes[3], &random, 3);
1552 }
1553 
1554 static int
1555 devarg_handle_int(__rte_unused const char *key, const char *value,
1556 		  void *extra_args)
1557 {
1558 	uint16_t *n = extra_args;
1559 
1560 	if (value == NULL || extra_args == NULL)
1561 		return -EINVAL;
1562 
1563 	*n = (uint16_t)strtoul(value, NULL, 0);
1564 	if (*n == USHRT_MAX && errno == ERANGE)
1565 		return -1;
1566 
1567 	return 0;
1568 }
1569 
1570 static void
1571 ixgbevf_parse_devargs(struct ixgbe_adapter *adapter,
1572 		      struct rte_devargs *devargs)
1573 {
1574 	struct rte_kvargs *kvlist;
1575 	uint16_t pflink_fullchk;
1576 
1577 	if (devargs == NULL)
1578 		return;
1579 
1580 	kvlist = rte_kvargs_parse(devargs->args, ixgbevf_valid_arguments);
1581 	if (kvlist == NULL)
1582 		return;
1583 
1584 	if (rte_kvargs_count(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK) == 1 &&
1585 	    rte_kvargs_process(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK,
1586 			       devarg_handle_int, &pflink_fullchk) == 0 &&
1587 	    pflink_fullchk == 1)
1588 		adapter->pflink_fullchk = 1;
1589 
1590 	rte_kvargs_free(kvlist);
1591 }
1592 
1593 /*
1594  * Virtual Function device init
1595  */
1596 static int
1597 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1598 {
1599 	int diag;
1600 	uint32_t tc, tcs;
1601 	struct ixgbe_adapter *ad = eth_dev->data->dev_private;
1602 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1603 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1604 	struct ixgbe_hw *hw =
1605 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1606 	struct ixgbe_vfta *shadow_vfta =
1607 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1608 	struct ixgbe_hwstrip *hwstrip =
1609 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1610 	struct rte_ether_addr *perm_addr =
1611 		(struct rte_ether_addr *)hw->mac.perm_addr;
1612 
1613 	PMD_INIT_FUNC_TRACE();
1614 
1615 	eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1616 	eth_dev->rx_descriptor_status = ixgbe_dev_rx_descriptor_status;
1617 	eth_dev->tx_descriptor_status = ixgbe_dev_tx_descriptor_status;
1618 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1619 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1620 
1621 	/* for secondary processes, we don't initialise any further as primary
1622 	 * has already done this work. Only check we don't need a different
1623 	 * RX function
1624 	 */
1625 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1626 		struct ci_tx_queue *txq;
1627 		/* TX queue function in primary, set by last queue initialized
1628 		 * Tx queue may not initialized by primary process
1629 		 */
1630 		if (eth_dev->data->tx_queues) {
1631 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1632 			ixgbe_set_tx_function(eth_dev, txq);
1633 		} else {
1634 			/* Use default TX function if we get here */
1635 			PMD_INIT_LOG(NOTICE,
1636 				     "No TX queues configured yet. Using default TX function.");
1637 		}
1638 
1639 		ixgbe_set_rx_function(eth_dev);
1640 
1641 		return 0;
1642 	}
1643 
1644 	/* NOTE: review for potential ordering optimization */
1645 	rte_atomic_store_explicit(&ad->link_thread_running, 0, rte_memory_order_seq_cst);
1646 	ixgbevf_parse_devargs(eth_dev->data->dev_private,
1647 			      pci_dev->device.devargs);
1648 
1649 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1650 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1651 
1652 	hw->device_id = pci_dev->id.device_id;
1653 	hw->vendor_id = pci_dev->id.vendor_id;
1654 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1655 
1656 	/* initialize the vfta */
1657 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1658 
1659 	/* initialize the hw strip bitmap*/
1660 	memset(hwstrip, 0, sizeof(*hwstrip));
1661 
1662 	/* Initialize the shared code (base driver) */
1663 	diag = ixgbe_init_shared_code(hw);
1664 	if (diag != IXGBE_SUCCESS) {
1665 		PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1666 		return -EIO;
1667 	}
1668 
1669 	/* init_mailbox_params */
1670 	hw->mbx.ops[0].init_params(hw);
1671 
1672 	/* Reset the hw statistics */
1673 	ixgbevf_dev_stats_reset(eth_dev);
1674 
1675 	/* Disable the interrupts for VF */
1676 	ixgbevf_intr_disable(eth_dev);
1677 
1678 	hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1679 	diag = hw->mac.ops.reset_hw(hw);
1680 
1681 	/*
1682 	 * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1683 	 * the underlying PF driver has not assigned a MAC address to the VF.
1684 	 * In this case, assign a random MAC address.
1685 	 */
1686 	if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1687 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1688 		/*
1689 		 * This error code will be propagated to the app by
1690 		 * rte_eth_dev_reset, so use a public error code rather than
1691 		 * the internal-only IXGBE_ERR_RESET_FAILED
1692 		 */
1693 		return -EAGAIN;
1694 	}
1695 
1696 	/* negotiate mailbox API version to use with the PF. */
1697 	ixgbevf_negotiate_api(hw);
1698 
1699 	/* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1700 	ixgbevf_get_queues(hw, &tcs, &tc);
1701 
1702 	/* Allocate memory for storing MAC addresses */
1703 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", RTE_ETHER_ADDR_LEN *
1704 					       hw->mac.num_rar_entries, 0);
1705 	if (eth_dev->data->mac_addrs == NULL) {
1706 		PMD_INIT_LOG(ERR,
1707 			     "Failed to allocate %u bytes needed to store "
1708 			     "MAC addresses",
1709 			     RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1710 		return -ENOMEM;
1711 	}
1712 
1713 	/* Generate a random MAC address, if none was assigned by PF. */
1714 	if (rte_is_zero_ether_addr(perm_addr)) {
1715 		generate_random_mac_addr(perm_addr);
1716 		diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1717 		if (diag) {
1718 			rte_free(eth_dev->data->mac_addrs);
1719 			eth_dev->data->mac_addrs = NULL;
1720 			return diag;
1721 		}
1722 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1723 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1724 			     RTE_ETHER_ADDR_PRT_FMT,
1725 				 RTE_ETHER_ADDR_BYTES(perm_addr));
1726 	}
1727 
1728 	/* Copy the permanent MAC address */
1729 	rte_ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1730 
1731 	/* reset the hardware with the new settings */
1732 	diag = hw->mac.ops.start_hw(hw);
1733 	switch (diag) {
1734 	case  0:
1735 		break;
1736 
1737 	default:
1738 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1739 		rte_free(eth_dev->data->mac_addrs);
1740 		eth_dev->data->mac_addrs = NULL;
1741 		return -EIO;
1742 	}
1743 
1744 	rte_intr_callback_register(intr_handle,
1745 				   ixgbevf_dev_interrupt_handler, eth_dev);
1746 	rte_intr_enable(intr_handle);
1747 	ixgbevf_intr_enable(eth_dev);
1748 
1749 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1750 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1751 		     pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1752 
1753 	return 0;
1754 }
1755 
1756 /* Virtual Function device uninit */
1757 
1758 static int
1759 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1760 {
1761 	PMD_INIT_FUNC_TRACE();
1762 
1763 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1764 		return 0;
1765 
1766 	ixgbevf_dev_close(eth_dev);
1767 
1768 	return 0;
1769 }
1770 
1771 static int
1772 eth_ixgbe_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1773 		struct rte_pci_device *pci_dev)
1774 {
1775 	char name[RTE_ETH_NAME_MAX_LEN];
1776 	struct rte_eth_dev *pf_ethdev;
1777 	struct rte_eth_devargs eth_da;
1778 	int i, retval;
1779 
1780 	if (pci_dev->device.devargs) {
1781 		retval = rte_eth_devargs_parse(pci_dev->device.devargs->args,
1782 				&eth_da, 1);
1783 		if (retval < 0)
1784 			return retval;
1785 	} else
1786 		memset(&eth_da, 0, sizeof(eth_da));
1787 
1788 	if (eth_da.nb_representor_ports > 0 &&
1789 	    eth_da.type != RTE_ETH_REPRESENTOR_VF) {
1790 		PMD_DRV_LOG(ERR, "unsupported representor type: %s",
1791 			    pci_dev->device.devargs->args);
1792 		return -ENOTSUP;
1793 	}
1794 
1795 	retval = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
1796 		sizeof(struct ixgbe_adapter),
1797 		eth_dev_pci_specific_init, pci_dev,
1798 		eth_ixgbe_dev_init, NULL);
1799 
1800 	if (retval || eth_da.nb_representor_ports < 1)
1801 		return retval;
1802 
1803 	pf_ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1804 	if (pf_ethdev == NULL)
1805 		return -ENODEV;
1806 
1807 	/* probe VF representor ports */
1808 	for (i = 0; i < eth_da.nb_representor_ports; i++) {
1809 		struct ixgbe_vf_info *vfinfo;
1810 		struct ixgbe_vf_representor representor;
1811 
1812 		vfinfo = *IXGBE_DEV_PRIVATE_TO_P_VFDATA(
1813 			pf_ethdev->data->dev_private);
1814 		if (vfinfo == NULL) {
1815 			PMD_DRV_LOG(ERR,
1816 				"no virtual functions supported by PF");
1817 			break;
1818 		}
1819 
1820 		representor.vf_id = eth_da.representor_ports[i];
1821 		representor.switch_domain_id = vfinfo->switch_domain_id;
1822 		representor.pf_ethdev = pf_ethdev;
1823 
1824 		/* representor port net_bdf_port */
1825 		snprintf(name, sizeof(name), "net_%s_representor_%d",
1826 			pci_dev->device.name,
1827 			eth_da.representor_ports[i]);
1828 
1829 		retval = rte_eth_dev_create(&pci_dev->device, name,
1830 			sizeof(struct ixgbe_vf_representor), NULL, NULL,
1831 			ixgbe_vf_representor_init, &representor);
1832 
1833 		if (retval)
1834 			PMD_DRV_LOG(ERR, "failed to create ixgbe vf "
1835 				"representor %s.", name);
1836 	}
1837 
1838 	return 0;
1839 }
1840 
1841 static int eth_ixgbe_pci_remove(struct rte_pci_device *pci_dev)
1842 {
1843 	struct rte_eth_dev *ethdev;
1844 
1845 	ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1846 	if (!ethdev)
1847 		return 0;
1848 
1849 	if (rte_eth_dev_is_repr(ethdev))
1850 		return rte_eth_dev_pci_generic_remove(pci_dev,
1851 					ixgbe_vf_representor_uninit);
1852 	else
1853 		return rte_eth_dev_pci_generic_remove(pci_dev,
1854 						eth_ixgbe_dev_uninit);
1855 }
1856 
1857 static struct rte_pci_driver rte_ixgbe_pmd = {
1858 	.id_table = pci_id_ixgbe_map,
1859 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1860 	.probe = eth_ixgbe_pci_probe,
1861 	.remove = eth_ixgbe_pci_remove,
1862 };
1863 
1864 static int eth_ixgbevf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1865 	struct rte_pci_device *pci_dev)
1866 {
1867 	return rte_eth_dev_pci_generic_probe(pci_dev,
1868 		sizeof(struct ixgbe_adapter), eth_ixgbevf_dev_init);
1869 }
1870 
1871 static int eth_ixgbevf_pci_remove(struct rte_pci_device *pci_dev)
1872 {
1873 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbevf_dev_uninit);
1874 }
1875 
1876 /*
1877  * virtual function driver struct
1878  */
1879 static struct rte_pci_driver rte_ixgbevf_pmd = {
1880 	.id_table = pci_id_ixgbevf_map,
1881 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1882 	.probe = eth_ixgbevf_pci_probe,
1883 	.remove = eth_ixgbevf_pci_remove,
1884 };
1885 
1886 static int
1887 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1888 {
1889 	struct ixgbe_hw *hw =
1890 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1891 	struct ixgbe_vfta *shadow_vfta =
1892 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1893 	uint32_t vfta;
1894 	uint32_t vid_idx;
1895 	uint32_t vid_bit;
1896 
1897 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1898 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1899 	vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1900 	if (on)
1901 		vfta |= vid_bit;
1902 	else
1903 		vfta &= ~vid_bit;
1904 	IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1905 
1906 	/* update local VFTA copy */
1907 	shadow_vfta->vfta[vid_idx] = vfta;
1908 
1909 	return 0;
1910 }
1911 
1912 static void
1913 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1914 {
1915 	if (on)
1916 		ixgbe_vlan_hw_strip_enable(dev, queue);
1917 	else
1918 		ixgbe_vlan_hw_strip_disable(dev, queue);
1919 }
1920 
1921 static int
1922 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1923 		    enum rte_vlan_type vlan_type,
1924 		    uint16_t tpid)
1925 {
1926 	struct ixgbe_hw *hw =
1927 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1928 	int ret = 0;
1929 	uint32_t reg;
1930 	uint32_t qinq;
1931 
1932 	qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1933 	qinq &= IXGBE_DMATXCTL_GDV;
1934 
1935 	switch (vlan_type) {
1936 	case RTE_ETH_VLAN_TYPE_INNER:
1937 		if (qinq) {
1938 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1939 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1940 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1941 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1942 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1943 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1944 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1945 		} else {
1946 			ret = -ENOTSUP;
1947 			PMD_DRV_LOG(ERR, "Inner type is not supported"
1948 				    " by single VLAN");
1949 		}
1950 		break;
1951 	case RTE_ETH_VLAN_TYPE_OUTER:
1952 		if (qinq) {
1953 			/* Only the high 16-bits is valid */
1954 			IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1955 					IXGBE_EXVET_VET_EXT_SHIFT);
1956 		} else {
1957 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1958 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1959 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1960 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1961 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1962 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1963 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1964 		}
1965 
1966 		break;
1967 	default:
1968 		ret = -EINVAL;
1969 		PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1970 		break;
1971 	}
1972 
1973 	return ret;
1974 }
1975 
1976 void
1977 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1978 {
1979 	struct ixgbe_hw *hw =
1980 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1981 	uint32_t vlnctrl;
1982 
1983 	PMD_INIT_FUNC_TRACE();
1984 
1985 	/* Filter Table Disable */
1986 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1987 	vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1988 
1989 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1990 }
1991 
1992 void
1993 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1994 {
1995 	struct ixgbe_hw *hw =
1996 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1997 	struct ixgbe_vfta *shadow_vfta =
1998 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1999 	uint32_t vlnctrl;
2000 	uint16_t i;
2001 
2002 	PMD_INIT_FUNC_TRACE();
2003 
2004 	/* Filter Table Enable */
2005 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2006 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
2007 	vlnctrl |= IXGBE_VLNCTRL_VFE;
2008 
2009 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
2010 
2011 	/* write whatever is in local vfta copy */
2012 	for (i = 0; i < IXGBE_VFTA_SIZE; i++)
2013 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
2014 }
2015 
2016 static void
2017 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
2018 {
2019 	struct ixgbe_hwstrip *hwstrip =
2020 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
2021 	struct ixgbe_rx_queue *rxq;
2022 
2023 	if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
2024 		return;
2025 
2026 	if (on)
2027 		IXGBE_SET_HWSTRIP(hwstrip, queue);
2028 	else
2029 		IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
2030 
2031 	if (queue >= dev->data->nb_rx_queues)
2032 		return;
2033 
2034 	rxq = dev->data->rx_queues[queue];
2035 
2036 	if (on) {
2037 		rxq->vlan_flags = RTE_MBUF_F_RX_VLAN | RTE_MBUF_F_RX_VLAN_STRIPPED;
2038 		rxq->offloads |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
2039 	} else {
2040 		rxq->vlan_flags = RTE_MBUF_F_RX_VLAN;
2041 		rxq->offloads &= ~RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
2042 	}
2043 }
2044 
2045 static void
2046 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
2047 {
2048 	struct ixgbe_hw *hw =
2049 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2050 	uint32_t ctrl;
2051 
2052 	PMD_INIT_FUNC_TRACE();
2053 
2054 	if (hw->mac.type == ixgbe_mac_82598EB) {
2055 		/* No queue level support */
2056 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2057 		return;
2058 	}
2059 
2060 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2061 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2062 	ctrl &= ~IXGBE_RXDCTL_VME;
2063 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2064 
2065 	/* record those setting for HW strip per queue */
2066 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
2067 }
2068 
2069 static void
2070 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
2071 {
2072 	struct ixgbe_hw *hw =
2073 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2074 	uint32_t ctrl;
2075 
2076 	PMD_INIT_FUNC_TRACE();
2077 
2078 	if (hw->mac.type == ixgbe_mac_82598EB) {
2079 		/* No queue level supported */
2080 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2081 		return;
2082 	}
2083 
2084 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2085 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2086 	ctrl |= IXGBE_RXDCTL_VME;
2087 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2088 
2089 	/* record those setting for HW strip per queue */
2090 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
2091 }
2092 
2093 static void
2094 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2095 {
2096 	struct ixgbe_hw *hw =
2097 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2098 	uint32_t ctrl;
2099 
2100 	PMD_INIT_FUNC_TRACE();
2101 
2102 	/* DMATXCTRL: Geric Double VLAN Disable */
2103 	ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2104 	ctrl &= ~IXGBE_DMATXCTL_GDV;
2105 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2106 
2107 	/* CTRL_EXT: Global Double VLAN Disable */
2108 	ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2109 	ctrl &= ~IXGBE_EXTENDED_VLAN;
2110 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2111 
2112 }
2113 
2114 static void
2115 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2116 {
2117 	struct ixgbe_hw *hw =
2118 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2119 	uint32_t ctrl;
2120 
2121 	PMD_INIT_FUNC_TRACE();
2122 
2123 	/* DMATXCTRL: Geric Double VLAN Enable */
2124 	ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2125 	ctrl |= IXGBE_DMATXCTL_GDV;
2126 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2127 
2128 	/* CTRL_EXT: Global Double VLAN Enable */
2129 	ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2130 	ctrl |= IXGBE_EXTENDED_VLAN;
2131 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2132 
2133 	/* Clear pooling mode of PFVTCTL. It's required by X550. */
2134 	if (hw->mac.type == ixgbe_mac_X550 ||
2135 	    hw->mac.type == ixgbe_mac_X550EM_x ||
2136 	    hw->mac.type == ixgbe_mac_X550EM_a) {
2137 		ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
2138 		ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
2139 		IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
2140 	}
2141 
2142 	/*
2143 	 * VET EXT field in the EXVET register = 0x8100 by default
2144 	 * So no need to change. Same to VT field of DMATXCTL register
2145 	 */
2146 }
2147 
2148 void
2149 ixgbe_vlan_hw_strip_config(struct rte_eth_dev *dev)
2150 {
2151 	struct ixgbe_hw *hw =
2152 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2153 	struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
2154 	uint32_t ctrl;
2155 	uint16_t i;
2156 	struct ixgbe_rx_queue *rxq;
2157 	bool on;
2158 
2159 	PMD_INIT_FUNC_TRACE();
2160 
2161 	if (hw->mac.type == ixgbe_mac_82598EB) {
2162 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP) {
2163 			ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2164 			ctrl |= IXGBE_VLNCTRL_VME;
2165 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2166 		} else {
2167 			ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2168 			ctrl &= ~IXGBE_VLNCTRL_VME;
2169 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2170 		}
2171 	} else {
2172 		/*
2173 		 * Other 10G NIC, the VLAN strip can be setup
2174 		 * per queue in RXDCTL
2175 		 */
2176 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
2177 			rxq = dev->data->rx_queues[i];
2178 			ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2179 			if (rxq->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP) {
2180 				ctrl |= IXGBE_RXDCTL_VME;
2181 				on = TRUE;
2182 			} else {
2183 				ctrl &= ~IXGBE_RXDCTL_VME;
2184 				on = FALSE;
2185 			}
2186 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2187 
2188 			/* record those setting for HW strip per queue */
2189 			ixgbe_vlan_hw_strip_bitmap_set(dev, i, on);
2190 		}
2191 	}
2192 }
2193 
2194 static void
2195 ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev, int mask)
2196 {
2197 	uint16_t i;
2198 	struct rte_eth_rxmode *rxmode;
2199 	struct ixgbe_rx_queue *rxq;
2200 
2201 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
2202 		rxmode = &dev->data->dev_conf.rxmode;
2203 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
2204 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
2205 				rxq = dev->data->rx_queues[i];
2206 				rxq->offloads |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
2207 			}
2208 		else
2209 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
2210 				rxq = dev->data->rx_queues[i];
2211 				rxq->offloads &= ~RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
2212 			}
2213 	}
2214 }
2215 
2216 static int
2217 ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask)
2218 {
2219 	struct rte_eth_rxmode *rxmode;
2220 	rxmode = &dev->data->dev_conf.rxmode;
2221 
2222 	if (mask & RTE_ETH_VLAN_STRIP_MASK)
2223 		ixgbe_vlan_hw_strip_config(dev);
2224 
2225 	if (mask & RTE_ETH_VLAN_FILTER_MASK) {
2226 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
2227 			ixgbe_vlan_hw_filter_enable(dev);
2228 		else
2229 			ixgbe_vlan_hw_filter_disable(dev);
2230 	}
2231 
2232 	if (mask & RTE_ETH_VLAN_EXTEND_MASK) {
2233 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND)
2234 			ixgbe_vlan_hw_extend_enable(dev);
2235 		else
2236 			ixgbe_vlan_hw_extend_disable(dev);
2237 	}
2238 
2239 	return 0;
2240 }
2241 
2242 static int
2243 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2244 {
2245 	ixgbe_config_vlan_strip_on_all_queues(dev, mask);
2246 
2247 	ixgbe_vlan_offload_config(dev, mask);
2248 
2249 	return 0;
2250 }
2251 
2252 static void
2253 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2254 {
2255 	struct ixgbe_hw *hw =
2256 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2257 	/* VLNCTRL: enable vlan filtering and allow all vlan tags through */
2258 	uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2259 
2260 	vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
2261 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
2262 }
2263 
2264 static int
2265 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
2266 {
2267 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2268 
2269 	switch (nb_rx_q) {
2270 	case 1:
2271 	case 2:
2272 		RTE_ETH_DEV_SRIOV(dev).active = RTE_ETH_64_POOLS;
2273 		break;
2274 	case 4:
2275 		RTE_ETH_DEV_SRIOV(dev).active = RTE_ETH_32_POOLS;
2276 		break;
2277 	default:
2278 		return -EINVAL;
2279 	}
2280 
2281 	RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
2282 		IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2283 	RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
2284 		pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2285 	return 0;
2286 }
2287 
2288 static int
2289 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
2290 {
2291 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
2292 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2293 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
2294 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
2295 
2296 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
2297 		/* check multi-queue mode */
2298 		switch (dev_conf->rxmode.mq_mode) {
2299 		case RTE_ETH_MQ_RX_VMDQ_DCB:
2300 			PMD_INIT_LOG(INFO, "RTE_ETH_MQ_RX_VMDQ_DCB mode supported in SRIOV");
2301 			break;
2302 		case RTE_ETH_MQ_RX_VMDQ_DCB_RSS:
2303 			/* DCB/RSS VMDQ in SRIOV mode, not implement yet */
2304 			PMD_INIT_LOG(ERR, "SRIOV active,"
2305 					" unsupported mq_mode rx %d.",
2306 					dev_conf->rxmode.mq_mode);
2307 			return -EINVAL;
2308 		case RTE_ETH_MQ_RX_RSS:
2309 		case RTE_ETH_MQ_RX_VMDQ_RSS:
2310 			dev->data->dev_conf.rxmode.mq_mode = RTE_ETH_MQ_RX_VMDQ_RSS;
2311 			if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
2312 				if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
2313 					PMD_INIT_LOG(ERR, "SRIOV is active,"
2314 						" invalid queue number"
2315 						" for VMDQ RSS, allowed"
2316 						" value are 1, 2 or 4.");
2317 					return -EINVAL;
2318 				}
2319 			break;
2320 		case RTE_ETH_MQ_RX_VMDQ_ONLY:
2321 		case RTE_ETH_MQ_RX_NONE:
2322 			/* if nothing mq mode configure, use default scheme */
2323 			dev->data->dev_conf.rxmode.mq_mode = RTE_ETH_MQ_RX_VMDQ_ONLY;
2324 			break;
2325 		default: /* RTE_ETH_MQ_RX_DCB, RTE_ETH_MQ_RX_DCB_RSS or RTE_ETH_MQ_TX_DCB*/
2326 			/* SRIOV only works in VMDq enable mode */
2327 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2328 					" wrong mq_mode rx %d.",
2329 					dev_conf->rxmode.mq_mode);
2330 			return -EINVAL;
2331 		}
2332 
2333 		switch (dev_conf->txmode.mq_mode) {
2334 		case RTE_ETH_MQ_TX_VMDQ_DCB:
2335 			PMD_INIT_LOG(INFO, "RTE_ETH_MQ_TX_VMDQ_DCB mode supported in SRIOV");
2336 			dev->data->dev_conf.txmode.mq_mode = RTE_ETH_MQ_TX_VMDQ_DCB;
2337 			break;
2338 		default: /* RTE_ETH_MQ_TX_VMDQ_ONLY or RTE_ETH_MQ_TX_NONE */
2339 			dev->data->dev_conf.txmode.mq_mode = RTE_ETH_MQ_TX_VMDQ_ONLY;
2340 			break;
2341 		}
2342 
2343 		/* check valid queue number */
2344 		if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2345 		    (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2346 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2347 					" nb_rx_q=%d nb_tx_q=%d queue number"
2348 					" must be less than or equal to %d.",
2349 					nb_rx_q, nb_tx_q,
2350 					RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2351 			return -EINVAL;
2352 		}
2353 	} else {
2354 		if (dev_conf->rxmode.mq_mode == RTE_ETH_MQ_RX_VMDQ_DCB_RSS) {
2355 			PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2356 					  " not supported.");
2357 			return -EINVAL;
2358 		}
2359 		/* check configuration for vmdb+dcb mode */
2360 		if (dev_conf->rxmode.mq_mode == RTE_ETH_MQ_RX_VMDQ_DCB) {
2361 			const struct rte_eth_vmdq_dcb_conf *conf;
2362 
2363 			if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2364 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2365 						IXGBE_VMDQ_DCB_NB_QUEUES);
2366 				return -EINVAL;
2367 			}
2368 			conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2369 			if (!(conf->nb_queue_pools == RTE_ETH_16_POOLS ||
2370 			       conf->nb_queue_pools == RTE_ETH_32_POOLS)) {
2371 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2372 						" nb_queue_pools must be %d or %d.",
2373 						RTE_ETH_16_POOLS, RTE_ETH_32_POOLS);
2374 				return -EINVAL;
2375 			}
2376 		}
2377 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_VMDQ_DCB) {
2378 			const struct rte_eth_vmdq_dcb_tx_conf *conf;
2379 
2380 			if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2381 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2382 						 IXGBE_VMDQ_DCB_NB_QUEUES);
2383 				return -EINVAL;
2384 			}
2385 			conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2386 			if (!(conf->nb_queue_pools == RTE_ETH_16_POOLS ||
2387 			       conf->nb_queue_pools == RTE_ETH_32_POOLS)) {
2388 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2389 						" nb_queue_pools != %d and"
2390 						" nb_queue_pools != %d.",
2391 						RTE_ETH_16_POOLS, RTE_ETH_32_POOLS);
2392 				return -EINVAL;
2393 			}
2394 		}
2395 
2396 		/* For DCB mode check our configuration before we go further */
2397 		if (dev_conf->rxmode.mq_mode == RTE_ETH_MQ_RX_DCB) {
2398 			const struct rte_eth_dcb_rx_conf *conf;
2399 
2400 			conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2401 			if (!(conf->nb_tcs == RTE_ETH_4_TCS ||
2402 			       conf->nb_tcs == RTE_ETH_8_TCS)) {
2403 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2404 						" and nb_tcs != %d.",
2405 						RTE_ETH_4_TCS, RTE_ETH_8_TCS);
2406 				return -EINVAL;
2407 			}
2408 		}
2409 
2410 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_DCB) {
2411 			const struct rte_eth_dcb_tx_conf *conf;
2412 
2413 			conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2414 			if (!(conf->nb_tcs == RTE_ETH_4_TCS ||
2415 			       conf->nb_tcs == RTE_ETH_8_TCS)) {
2416 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2417 						" and nb_tcs != %d.",
2418 						RTE_ETH_4_TCS, RTE_ETH_8_TCS);
2419 				return -EINVAL;
2420 			}
2421 		}
2422 
2423 		/*
2424 		 * When DCB/VT is off, maximum number of queues changes,
2425 		 * except for 82598EB, which remains constant.
2426 		 */
2427 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_NONE &&
2428 				hw->mac.type != ixgbe_mac_82598EB) {
2429 			if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2430 				PMD_INIT_LOG(ERR,
2431 					     "Neither VT nor DCB are enabled, "
2432 					     "nb_tx_q > %d.",
2433 					     IXGBE_NONE_MODE_TX_NB_QUEUES);
2434 				return -EINVAL;
2435 			}
2436 		}
2437 	}
2438 	return 0;
2439 }
2440 
2441 static int
2442 ixgbe_dev_configure(struct rte_eth_dev *dev)
2443 {
2444 	struct ixgbe_interrupt *intr =
2445 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2446 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2447 	int ret;
2448 
2449 	PMD_INIT_FUNC_TRACE();
2450 
2451 	if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
2452 		dev->data->dev_conf.rxmode.offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
2453 
2454 	/* multiple queue mode checking */
2455 	ret  = ixgbe_check_mq_mode(dev);
2456 	if (ret != 0) {
2457 		PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2458 			    ret);
2459 		return ret;
2460 	}
2461 
2462 	/* set flag to update link status after init */
2463 	intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2464 
2465 	/*
2466 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2467 	 * allocation or vector Rx preconditions we will reset it.
2468 	 */
2469 	adapter->rx_bulk_alloc_allowed = true;
2470 	adapter->rx_vec_allowed = true;
2471 
2472 	return 0;
2473 }
2474 
2475 static void
2476 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2477 {
2478 	struct ixgbe_hw *hw =
2479 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2480 	struct ixgbe_interrupt *intr =
2481 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2482 	uint32_t gpie;
2483 
2484 	/* only set up it on X550EM_X */
2485 	if (hw->mac.type == ixgbe_mac_X550EM_x) {
2486 		gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2487 		gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2488 		IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2489 		if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2490 			intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2491 	}
2492 }
2493 
2494 int
2495 ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
2496 			uint32_t tx_rate, uint64_t q_msk)
2497 {
2498 	struct ixgbe_hw *hw;
2499 	struct ixgbe_vf_info *vfinfo;
2500 	struct rte_eth_link link;
2501 	uint8_t  nb_q_per_pool;
2502 	uint32_t queue_stride;
2503 	uint32_t queue_idx, idx = 0, vf_idx;
2504 	uint32_t queue_end;
2505 	uint16_t total_rate = 0;
2506 	struct rte_pci_device *pci_dev;
2507 	int ret;
2508 
2509 	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2510 	ret = rte_eth_link_get_nowait(dev->data->port_id, &link);
2511 	if (ret < 0)
2512 		return ret;
2513 
2514 	if (vf >= pci_dev->max_vfs)
2515 		return -EINVAL;
2516 
2517 	if (tx_rate > link.link_speed)
2518 		return -EINVAL;
2519 
2520 	if (q_msk == 0)
2521 		return 0;
2522 
2523 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2524 	vfinfo = *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
2525 	nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2526 	queue_stride = IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2527 	queue_idx = vf * queue_stride;
2528 	queue_end = queue_idx + nb_q_per_pool - 1;
2529 	if (queue_end >= hw->mac.max_tx_queues)
2530 		return -EINVAL;
2531 
2532 	if (vfinfo) {
2533 		for (vf_idx = 0; vf_idx < pci_dev->max_vfs; vf_idx++) {
2534 			if (vf_idx == vf)
2535 				continue;
2536 			for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
2537 				idx++)
2538 				total_rate += vfinfo[vf_idx].tx_rate[idx];
2539 		}
2540 	} else {
2541 		return -EINVAL;
2542 	}
2543 
2544 	/* Store tx_rate for this vf. */
2545 	for (idx = 0; idx < nb_q_per_pool; idx++) {
2546 		if (((uint64_t)0x1 << idx) & q_msk) {
2547 			if (vfinfo[vf].tx_rate[idx] != tx_rate)
2548 				vfinfo[vf].tx_rate[idx] = tx_rate;
2549 			total_rate += tx_rate;
2550 		}
2551 	}
2552 
2553 	if (total_rate > dev->data->dev_link.link_speed) {
2554 		/* Reset stored TX rate of the VF if it causes exceed
2555 		 * link speed.
2556 		 */
2557 		memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
2558 		return -EINVAL;
2559 	}
2560 
2561 	/* Set RTTBCNRC of each queue/pool for vf X  */
2562 	for (; queue_idx <= queue_end; queue_idx++) {
2563 		if (0x1 & q_msk)
2564 			ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
2565 		q_msk = q_msk >> 1;
2566 	}
2567 
2568 	return 0;
2569 }
2570 
2571 static int
2572 ixgbe_flow_ctrl_enable(struct rte_eth_dev *dev, struct ixgbe_hw *hw)
2573 {
2574 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2575 	int err;
2576 	uint32_t mflcn;
2577 
2578 	ixgbe_setup_fc(hw);
2579 
2580 	err = ixgbe_fc_enable(hw);
2581 
2582 	/* Not negotiated is not an error case */
2583 	if (err == IXGBE_SUCCESS || err == IXGBE_ERR_FC_NOT_NEGOTIATED) {
2584 		/*
2585 		 *check if we want to forward MAC frames - driver doesn't
2586 		 *have native capability to do that,
2587 		 *so we'll write the registers ourselves
2588 		 */
2589 
2590 		mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
2591 
2592 		/* set or clear MFLCN.PMCF bit depending on configuration */
2593 		if (adapter->mac_ctrl_frame_fwd != 0)
2594 			mflcn |= IXGBE_MFLCN_PMCF;
2595 		else
2596 			mflcn &= ~IXGBE_MFLCN_PMCF;
2597 
2598 		IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
2599 		IXGBE_WRITE_FLUSH(hw);
2600 
2601 		return 0;
2602 	}
2603 	return err;
2604 }
2605 
2606 /*
2607  * Configure device link speed and setup link.
2608  * It returns 0 on success.
2609  */
2610 static int
2611 ixgbe_dev_start(struct rte_eth_dev *dev)
2612 {
2613 	struct ixgbe_hw *hw =
2614 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2615 	struct ixgbe_vf_info *vfinfo =
2616 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2617 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2618 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2619 	uint32_t intr_vector = 0;
2620 	int err;
2621 	bool link_up = false, negotiate = 0;
2622 	uint32_t speed = 0;
2623 	uint32_t allowed_speeds = 0;
2624 	int mask = 0;
2625 	int status;
2626 	uint16_t vf, idx;
2627 	uint32_t *link_speeds;
2628 	struct ixgbe_tm_conf *tm_conf =
2629 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2630 	struct ixgbe_macsec_setting *macsec_setting =
2631 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
2632 
2633 	PMD_INIT_FUNC_TRACE();
2634 
2635 	/* Stop the link setup handler before resetting the HW. */
2636 	ixgbe_dev_wait_setup_link_complete(dev, 0);
2637 
2638 	/* disable uio/vfio intr/eventfd mapping */
2639 	rte_intr_disable(intr_handle);
2640 
2641 	/* stop adapter */
2642 	hw->adapter_stopped = 0;
2643 	ixgbe_stop_adapter(hw);
2644 
2645 	/* reinitialize adapter
2646 	 * this calls reset and start
2647 	 */
2648 	status = ixgbe_pf_reset_hw(hw);
2649 	if (status != 0)
2650 		return -1;
2651 	hw->mac.ops.start_hw(hw);
2652 	hw->mac.get_link_status = true;
2653 
2654 	/* configure PF module if SRIOV enabled */
2655 	ixgbe_pf_host_configure(dev);
2656 
2657 	ixgbe_dev_phy_intr_setup(dev);
2658 
2659 	/* check and configure queue intr-vector mapping */
2660 	if ((rte_intr_cap_multiple(intr_handle) ||
2661 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
2662 	    dev->data->dev_conf.intr_conf.rxq != 0) {
2663 		intr_vector = dev->data->nb_rx_queues;
2664 		if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2665 			PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2666 					IXGBE_MAX_INTR_QUEUE_NUM);
2667 			return -ENOTSUP;
2668 		}
2669 		if (rte_intr_efd_enable(intr_handle, intr_vector))
2670 			return -1;
2671 	}
2672 
2673 	if (rte_intr_dp_is_en(intr_handle)) {
2674 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
2675 						   dev->data->nb_rx_queues)) {
2676 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2677 				     " intr_vec", dev->data->nb_rx_queues);
2678 			return -ENOMEM;
2679 		}
2680 	}
2681 
2682 	/* configure MSI-X for sleep until Rx interrupt */
2683 	ixgbe_configure_msix(dev);
2684 
2685 	/* initialize transmission unit */
2686 	ixgbe_dev_tx_init(dev);
2687 
2688 	/* This can fail when allocating mbufs for descriptor rings */
2689 	err = ixgbe_dev_rx_init(dev);
2690 	if (err) {
2691 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2692 		goto error;
2693 	}
2694 
2695 	mask = RTE_ETH_VLAN_STRIP_MASK | RTE_ETH_VLAN_FILTER_MASK |
2696 		RTE_ETH_VLAN_EXTEND_MASK;
2697 	err = ixgbe_vlan_offload_config(dev, mask);
2698 	if (err) {
2699 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload");
2700 		goto error;
2701 	}
2702 
2703 	if (dev->data->dev_conf.rxmode.mq_mode == RTE_ETH_MQ_RX_VMDQ_ONLY) {
2704 		/* Enable vlan filtering for VMDq */
2705 		ixgbe_vmdq_vlan_hw_filter_enable(dev);
2706 	}
2707 
2708 	/* Configure DCB hw */
2709 	ixgbe_configure_dcb(dev);
2710 
2711 	if (IXGBE_DEV_FDIR_CONF(dev)->mode != RTE_FDIR_MODE_NONE) {
2712 		err = ixgbe_fdir_configure(dev);
2713 		if (err)
2714 			goto error;
2715 	}
2716 
2717 	/* Restore vf rate limit */
2718 	if (vfinfo != NULL) {
2719 		for (vf = 0; vf < pci_dev->max_vfs; vf++)
2720 			for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2721 				if (vfinfo[vf].tx_rate[idx] != 0)
2722 					ixgbe_set_vf_rate_limit(
2723 						dev, vf,
2724 						vfinfo[vf].tx_rate[idx],
2725 						1 << idx);
2726 	}
2727 
2728 	ixgbe_restore_statistics_mapping(dev);
2729 
2730 	err = ixgbe_flow_ctrl_enable(dev, hw);
2731 	if (err < 0) {
2732 		PMD_INIT_LOG(ERR, "enable flow ctrl err");
2733 		goto error;
2734 	}
2735 
2736 	err = ixgbe_dev_rxtx_start(dev);
2737 	if (err < 0) {
2738 		PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2739 		goto error;
2740 	}
2741 
2742 	/* Skip link setup if loopback mode is enabled. */
2743 	if (dev->data->dev_conf.lpbk_mode != 0) {
2744 		err = ixgbe_check_supported_loopback_mode(dev);
2745 		if (err < 0) {
2746 			PMD_INIT_LOG(ERR, "Unsupported loopback mode");
2747 			goto error;
2748 		} else {
2749 			goto skip_link_setup;
2750 		}
2751 	}
2752 
2753 	if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2754 		err = hw->mac.ops.setup_sfp(hw);
2755 		if (err)
2756 			goto error;
2757 	}
2758 
2759 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2760 		/* Turn on the copper */
2761 		ixgbe_set_phy_power(hw, true);
2762 	} else {
2763 		/* Turn on the laser */
2764 		ixgbe_enable_tx_laser(hw);
2765 	}
2766 
2767 	err = ixgbe_check_link(hw, &speed, &link_up, 0);
2768 	if (err)
2769 		goto error;
2770 	dev->data->dev_link.link_status = link_up;
2771 
2772 	err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2773 	if (err)
2774 		goto error;
2775 
2776 	switch (hw->mac.type) {
2777 	case ixgbe_mac_X550:
2778 	case ixgbe_mac_X550EM_x:
2779 	case ixgbe_mac_X550EM_a:
2780 		allowed_speeds = RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G |
2781 			RTE_ETH_LINK_SPEED_2_5G |  RTE_ETH_LINK_SPEED_5G |
2782 			RTE_ETH_LINK_SPEED_10G;
2783 		if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
2784 				hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
2785 			allowed_speeds = RTE_ETH_LINK_SPEED_10M |
2786 				RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G;
2787 		break;
2788 	default:
2789 		allowed_speeds = RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G |
2790 			RTE_ETH_LINK_SPEED_10G;
2791 	}
2792 
2793 	link_speeds = &dev->data->dev_conf.link_speeds;
2794 
2795 	/* Ignore autoneg flag bit and check the validity of 
2796 	 * link_speed 
2797 	 */
2798 	if (((*link_speeds) >> 1) & ~(allowed_speeds >> 1)) {
2799 		PMD_INIT_LOG(ERR, "Invalid link setting");
2800 		goto error;
2801 	}
2802 
2803 	speed = 0x0;
2804 	if (*link_speeds == RTE_ETH_LINK_SPEED_AUTONEG) {
2805 		switch (hw->mac.type) {
2806 		case ixgbe_mac_82598EB:
2807 			speed = IXGBE_LINK_SPEED_82598_AUTONEG;
2808 			break;
2809 		case ixgbe_mac_82599EB:
2810 		case ixgbe_mac_X540:
2811 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2812 			break;
2813 		case ixgbe_mac_X550:
2814 		case ixgbe_mac_X550EM_x:
2815 		case ixgbe_mac_X550EM_a:
2816 			speed = IXGBE_LINK_SPEED_X550_AUTONEG;
2817 			break;
2818 		default:
2819 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2820 		}
2821 	} else {
2822 		if (*link_speeds & RTE_ETH_LINK_SPEED_10G)
2823 			speed |= IXGBE_LINK_SPEED_10GB_FULL;
2824 		if (*link_speeds & RTE_ETH_LINK_SPEED_5G)
2825 			speed |= IXGBE_LINK_SPEED_5GB_FULL;
2826 		if (*link_speeds & RTE_ETH_LINK_SPEED_2_5G)
2827 			speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
2828 		if (*link_speeds & RTE_ETH_LINK_SPEED_1G)
2829 			speed |= IXGBE_LINK_SPEED_1GB_FULL;
2830 		if (*link_speeds & RTE_ETH_LINK_SPEED_100M)
2831 			speed |= IXGBE_LINK_SPEED_100_FULL;
2832 		if (*link_speeds & RTE_ETH_LINK_SPEED_10M)
2833 			speed |= IXGBE_LINK_SPEED_10_FULL;
2834 	}
2835 
2836 	err = ixgbe_setup_link(hw, speed, link_up);
2837 	if (err)
2838 		goto error;
2839 
2840 skip_link_setup:
2841 
2842 	if (rte_intr_allow_others(intr_handle)) {
2843 		/* check if lsc interrupt is enabled */
2844 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2845 			ixgbe_dev_lsc_interrupt_setup(dev, TRUE);
2846 		else
2847 			ixgbe_dev_lsc_interrupt_setup(dev, FALSE);
2848 		ixgbe_dev_macsec_interrupt_setup(dev);
2849 	} else {
2850 		rte_intr_callback_unregister(intr_handle,
2851 					     ixgbe_dev_interrupt_handler, dev);
2852 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2853 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
2854 				     " no intr multiplex");
2855 	}
2856 
2857 	/* check if rxq interrupt is enabled */
2858 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2859 	    rte_intr_dp_is_en(intr_handle))
2860 		ixgbe_dev_rxq_interrupt_setup(dev);
2861 
2862 	/* enable uio/vfio intr/eventfd mapping */
2863 	rte_intr_enable(intr_handle);
2864 
2865 	/* resume enabled intr since hw reset */
2866 	ixgbe_enable_intr(dev);
2867 	ixgbe_l2_tunnel_conf(dev);
2868 	ixgbe_filter_restore(dev);
2869 
2870 	if (tm_conf->root && !tm_conf->committed)
2871 		PMD_DRV_LOG(WARNING,
2872 			    "please call hierarchy_commit() "
2873 			    "before starting the port");
2874 
2875 	/* wait for the controller to acquire link */
2876 	err = ixgbe_wait_for_link_up(hw);
2877 	if (err)
2878 		goto error;
2879 
2880 	/*
2881 	 * Update link status right before return, because it may
2882 	 * start link configuration process in a separate thread.
2883 	 */
2884 	ixgbe_dev_link_update(dev, 0);
2885 
2886 	/* setup the macsec setting register */
2887 	if (macsec_setting->offload_en)
2888 		ixgbe_dev_macsec_register_enable(dev, macsec_setting);
2889 
2890 	return 0;
2891 
2892 error:
2893 	PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2894 	ixgbe_dev_clear_queues(dev);
2895 	return -EIO;
2896 }
2897 
2898 /*
2899  * Stop device: disable rx and tx functions to allow for reconfiguring.
2900  */
2901 static int
2902 ixgbe_dev_stop(struct rte_eth_dev *dev)
2903 {
2904 	struct rte_eth_link link;
2905 	struct ixgbe_adapter *adapter = dev->data->dev_private;
2906 	struct ixgbe_hw *hw =
2907 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2908 	struct ixgbe_vf_info *vfinfo =
2909 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2910 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2911 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2912 	int vf;
2913 	struct ixgbe_tm_conf *tm_conf =
2914 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2915 
2916 	if (hw->adapter_stopped)
2917 		return 0;
2918 
2919 	PMD_INIT_FUNC_TRACE();
2920 
2921 	ixgbe_dev_wait_setup_link_complete(dev, 0);
2922 
2923 	/* disable interrupts */
2924 	ixgbe_disable_intr(hw);
2925 
2926 	/* reset the NIC */
2927 	ixgbe_pf_reset_hw(hw);
2928 	hw->adapter_stopped = 0;
2929 
2930 	/* stop adapter */
2931 	ixgbe_stop_adapter(hw);
2932 
2933 	for (vf = 0; vfinfo != NULL && vf < pci_dev->max_vfs; vf++)
2934 		vfinfo[vf].clear_to_send = false;
2935 
2936 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2937 		/* Turn off the copper */
2938 		ixgbe_set_phy_power(hw, false);
2939 	} else {
2940 		/* Turn off the laser */
2941 		ixgbe_disable_tx_laser(hw);
2942 	}
2943 
2944 	ixgbe_dev_clear_queues(dev);
2945 
2946 	/* Clear stored conf */
2947 	dev->data->scattered_rx = 0;
2948 	dev->data->lro = 0;
2949 
2950 	/* Clear recorded link status */
2951 	memset(&link, 0, sizeof(link));
2952 	rte_eth_linkstatus_set(dev, &link);
2953 
2954 	if (!rte_intr_allow_others(intr_handle))
2955 		/* resume to the default handler */
2956 		rte_intr_callback_register(intr_handle,
2957 					   ixgbe_dev_interrupt_handler,
2958 					   (void *)dev);
2959 
2960 	/* Clean datapath event and queue/vec mapping */
2961 	rte_intr_efd_disable(intr_handle);
2962 	rte_intr_vec_list_free(intr_handle);
2963 
2964 	/* reset hierarchy commit */
2965 	tm_conf->committed = false;
2966 
2967 	adapter->rss_reta_updated = 0;
2968 
2969 	hw->adapter_stopped = true;
2970 	dev->data->dev_started = 0;
2971 
2972 	return 0;
2973 }
2974 
2975 /*
2976  * Set device link up: enable tx.
2977  */
2978 static int
2979 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2980 {
2981 	struct ixgbe_hw *hw =
2982 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2983 	if (hw->mac.type == ixgbe_mac_82599EB) {
2984 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2985 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2986 			/* Not supported in bypass mode */
2987 			PMD_INIT_LOG(ERR, "Set link up is not supported "
2988 				     "by device id 0x%x", hw->device_id);
2989 			return -ENOTSUP;
2990 		}
2991 #endif
2992 	}
2993 
2994 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2995 		/* Turn on the copper */
2996 		ixgbe_set_phy_power(hw, true);
2997 	} else {
2998 		/* Turn on the laser */
2999 		ixgbe_enable_tx_laser(hw);
3000 		ixgbe_dev_link_update(dev, 0);
3001 	}
3002 
3003 	return 0;
3004 }
3005 
3006 /*
3007  * Set device link down: disable tx.
3008  */
3009 static int
3010 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
3011 {
3012 	struct ixgbe_hw *hw =
3013 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3014 	if (hw->mac.type == ixgbe_mac_82599EB) {
3015 #ifdef RTE_LIBRTE_IXGBE_BYPASS
3016 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
3017 			/* Not supported in bypass mode */
3018 			PMD_INIT_LOG(ERR, "Set link down is not supported "
3019 				     "by device id 0x%x", hw->device_id);
3020 			return -ENOTSUP;
3021 		}
3022 #endif
3023 	}
3024 
3025 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
3026 		/* Turn off the copper */
3027 		ixgbe_set_phy_power(hw, false);
3028 	} else {
3029 		/* Turn off the laser */
3030 		ixgbe_disable_tx_laser(hw);
3031 		ixgbe_dev_link_update(dev, 0);
3032 	}
3033 
3034 	return 0;
3035 }
3036 
3037 /*
3038  * Reset and stop device.
3039  */
3040 static int
3041 ixgbe_dev_close(struct rte_eth_dev *dev)
3042 {
3043 	struct ixgbe_hw *hw =
3044 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3045 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3046 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
3047 	int retries = 0;
3048 	int ret;
3049 
3050 	PMD_INIT_FUNC_TRACE();
3051 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
3052 		return 0;
3053 
3054 	ixgbe_pf_reset_hw(hw);
3055 
3056 	ret = ixgbe_dev_stop(dev);
3057 
3058 	ixgbe_dev_free_queues(dev);
3059 
3060 	ixgbe_disable_pcie_primary(hw);
3061 
3062 	/* reprogram the RAR[0] in case user changed it. */
3063 	ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
3064 
3065 	/* Unlock any pending hardware semaphore */
3066 	ixgbe_swfw_lock_reset(hw);
3067 
3068 	/* disable uio intr before callback unregister */
3069 	rte_intr_disable(intr_handle);
3070 
3071 	do {
3072 		ret = rte_intr_callback_unregister(intr_handle,
3073 				ixgbe_dev_interrupt_handler, dev);
3074 		if (ret >= 0 || ret == -ENOENT) {
3075 			break;
3076 		} else if (ret != -EAGAIN) {
3077 			PMD_INIT_LOG(ERR,
3078 				"intr callback unregister failed: %d",
3079 				ret);
3080 		}
3081 		rte_delay_ms(100);
3082 	} while (retries++ < (10 + IXGBE_LINK_UP_TIME));
3083 
3084 	/* cancel the delay handler before remove dev */
3085 	rte_eal_alarm_cancel(ixgbe_dev_interrupt_delayed_handler, dev);
3086 
3087 	/* uninitialize PF if max_vfs not zero */
3088 	ixgbe_pf_host_uninit(dev);
3089 
3090 	/* remove all the fdir filters & hash */
3091 	ixgbe_fdir_filter_uninit(dev);
3092 
3093 	/* remove all the L2 tunnel filters & hash */
3094 	ixgbe_l2_tn_filter_uninit(dev);
3095 
3096 	/* Remove all ntuple filters of the device */
3097 	ixgbe_ntuple_filter_uninit(dev);
3098 
3099 	/* clear all the filters list */
3100 	ixgbe_filterlist_flush();
3101 
3102 	/* Remove all Traffic Manager configuration */
3103 	ixgbe_tm_conf_uninit(dev);
3104 
3105 #ifdef RTE_LIB_SECURITY
3106 	rte_free(dev->security_ctx);
3107 	dev->security_ctx = NULL;
3108 #endif
3109 
3110 	return ret;
3111 }
3112 
3113 /*
3114  * Reset PF device.
3115  */
3116 static int
3117 ixgbe_dev_reset(struct rte_eth_dev *dev)
3118 {
3119 	int ret;
3120 
3121 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
3122 	 * its VF to make them align with it. The detailed notification
3123 	 * mechanism is PMD specific. As to ixgbe PF, it is rather complex.
3124 	 * To avoid unexpected behavior in VF, currently reset of PF with
3125 	 * SR-IOV activation is not supported. It might be supported later.
3126 	 */
3127 	if (dev->data->sriov.active)
3128 		return -ENOTSUP;
3129 
3130 	ret = eth_ixgbe_dev_uninit(dev);
3131 	if (ret)
3132 		return ret;
3133 
3134 	ret = eth_ixgbe_dev_init(dev, NULL);
3135 
3136 	return ret;
3137 }
3138 
3139 static void
3140 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
3141 			   struct ixgbe_hw_stats *hw_stats,
3142 			   struct ixgbe_macsec_stats *macsec_stats,
3143 			   uint64_t *total_missed_rx, uint64_t *total_qbrc,
3144 			   uint64_t *total_qprc, uint64_t *total_qprdc)
3145 {
3146 	uint32_t bprc, lxon, lxoff, total;
3147 	uint32_t delta_gprc = 0;
3148 	unsigned i;
3149 	/* Workaround for RX byte count not including CRC bytes when CRC
3150 	 * strip is enabled. CRC bytes are removed from counters when crc_strip
3151 	 * is disabled.
3152 	 */
3153 	int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
3154 			IXGBE_HLREG0_RXCRCSTRP);
3155 
3156 	hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
3157 	hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
3158 	hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
3159 	hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
3160 
3161 	for (i = 0; i < 8; i++) {
3162 		uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
3163 
3164 		/* global total per queue */
3165 		hw_stats->mpc[i] += mp;
3166 		/* Running comprehensive total for stats display */
3167 		*total_missed_rx += hw_stats->mpc[i];
3168 		if (hw->mac.type == ixgbe_mac_82598EB) {
3169 			hw_stats->rnbc[i] +=
3170 			    IXGBE_READ_REG(hw, IXGBE_RNBC(i));
3171 			hw_stats->pxonrxc[i] +=
3172 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
3173 			hw_stats->pxoffrxc[i] +=
3174 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
3175 		} else {
3176 			hw_stats->pxonrxc[i] +=
3177 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
3178 			hw_stats->pxoffrxc[i] +=
3179 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
3180 			hw_stats->pxon2offc[i] +=
3181 				IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
3182 		}
3183 		hw_stats->pxontxc[i] +=
3184 		    IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
3185 		hw_stats->pxofftxc[i] +=
3186 		    IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
3187 	}
3188 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3189 		uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
3190 		uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
3191 		uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
3192 
3193 		delta_gprc += delta_qprc;
3194 
3195 		hw_stats->qprc[i] += delta_qprc;
3196 		hw_stats->qptc[i] += delta_qptc;
3197 
3198 		hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
3199 		hw_stats->qbrc[i] +=
3200 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
3201 		if (crc_strip == 0)
3202 			hw_stats->qbrc[i] -= delta_qprc * RTE_ETHER_CRC_LEN;
3203 
3204 		hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
3205 		hw_stats->qbtc[i] +=
3206 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
3207 
3208 		hw_stats->qprdc[i] += delta_qprdc;
3209 		*total_qprdc += hw_stats->qprdc[i];
3210 
3211 		*total_qprc += hw_stats->qprc[i];
3212 		*total_qbrc += hw_stats->qbrc[i];
3213 	}
3214 	hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
3215 	hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
3216 	hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
3217 
3218 	/*
3219 	 * An errata states that gprc actually counts good + missed packets:
3220 	 * Workaround to set gprc to summated queue packet receives
3221 	 */
3222 	hw_stats->gprc = *total_qprc;
3223 
3224 	if (hw->mac.type != ixgbe_mac_82598EB) {
3225 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
3226 		hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
3227 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
3228 		hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
3229 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
3230 		hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
3231 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
3232 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
3233 	} else {
3234 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
3235 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
3236 		/* 82598 only has a counter in the high register */
3237 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
3238 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
3239 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
3240 	}
3241 	uint64_t old_tpr = hw_stats->tpr;
3242 
3243 	hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
3244 	hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
3245 
3246 	if (crc_strip == 0)
3247 		hw_stats->gorc -= delta_gprc * RTE_ETHER_CRC_LEN;
3248 
3249 	uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
3250 	hw_stats->gptc += delta_gptc;
3251 	hw_stats->gotc -= delta_gptc * RTE_ETHER_CRC_LEN;
3252 	hw_stats->tor -= (hw_stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
3253 
3254 	/*
3255 	 * Workaround: mprc hardware is incorrectly counting
3256 	 * broadcasts, so for now we subtract those.
3257 	 */
3258 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
3259 	hw_stats->bprc += bprc;
3260 	hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
3261 	if (hw->mac.type == ixgbe_mac_82598EB)
3262 		hw_stats->mprc -= bprc;
3263 
3264 	hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
3265 	hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
3266 	hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
3267 	hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
3268 	hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
3269 	hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
3270 
3271 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
3272 	hw_stats->lxontxc += lxon;
3273 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
3274 	hw_stats->lxofftxc += lxoff;
3275 	total = lxon + lxoff;
3276 
3277 	hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
3278 	hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
3279 	hw_stats->gptc -= total;
3280 	hw_stats->mptc -= total;
3281 	hw_stats->ptc64 -= total;
3282 	hw_stats->gotc -= total * RTE_ETHER_MIN_LEN;
3283 
3284 	hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3285 	hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3286 	hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3287 	hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
3288 	hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
3289 	hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
3290 	hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
3291 	hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3292 	hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3293 	hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3294 	hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3295 	hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
3296 	hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3297 	hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
3298 	hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
3299 	hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
3300 	/* Only read FCOE on 82599 */
3301 	if (hw->mac.type != ixgbe_mac_82598EB) {
3302 		hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
3303 		hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
3304 		hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
3305 		hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
3306 		hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
3307 	}
3308 
3309 	/* Flow Director Stats registers */
3310 	if (hw->mac.type != ixgbe_mac_82598EB) {
3311 		hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
3312 		hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
3313 		hw_stats->fdirustat_add += IXGBE_READ_REG(hw,
3314 					IXGBE_FDIRUSTAT) & 0xFFFF;
3315 		hw_stats->fdirustat_remove += (IXGBE_READ_REG(hw,
3316 					IXGBE_FDIRUSTAT) >> 16) & 0xFFFF;
3317 		hw_stats->fdirfstat_fadd += IXGBE_READ_REG(hw,
3318 					IXGBE_FDIRFSTAT) & 0xFFFF;
3319 		hw_stats->fdirfstat_fremove += (IXGBE_READ_REG(hw,
3320 					IXGBE_FDIRFSTAT) >> 16) & 0xFFFF;
3321 	}
3322 	/* MACsec Stats registers */
3323 	macsec_stats->out_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECTXUT);
3324 	macsec_stats->out_pkts_encrypted +=
3325 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTE);
3326 	macsec_stats->out_pkts_protected +=
3327 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTP);
3328 	macsec_stats->out_octets_encrypted +=
3329 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTE);
3330 	macsec_stats->out_octets_protected +=
3331 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTP);
3332 	macsec_stats->in_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECRXUT);
3333 	macsec_stats->in_pkts_badtag += IXGBE_READ_REG(hw, IXGBE_LSECRXBAD);
3334 	macsec_stats->in_pkts_nosci += IXGBE_READ_REG(hw, IXGBE_LSECRXNOSCI);
3335 	macsec_stats->in_pkts_unknownsci +=
3336 		IXGBE_READ_REG(hw, IXGBE_LSECRXUNSCI);
3337 	macsec_stats->in_octets_decrypted +=
3338 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTD);
3339 	macsec_stats->in_octets_validated +=
3340 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTV);
3341 	macsec_stats->in_pkts_unchecked += IXGBE_READ_REG(hw, IXGBE_LSECRXUNCH);
3342 	macsec_stats->in_pkts_delayed += IXGBE_READ_REG(hw, IXGBE_LSECRXDELAY);
3343 	macsec_stats->in_pkts_late += IXGBE_READ_REG(hw, IXGBE_LSECRXLATE);
3344 	for (i = 0; i < 2; i++) {
3345 		macsec_stats->in_pkts_ok +=
3346 			IXGBE_READ_REG(hw, IXGBE_LSECRXOK(i));
3347 		macsec_stats->in_pkts_invalid +=
3348 			IXGBE_READ_REG(hw, IXGBE_LSECRXINV(i));
3349 		macsec_stats->in_pkts_notvalid +=
3350 			IXGBE_READ_REG(hw, IXGBE_LSECRXNV(i));
3351 	}
3352 	macsec_stats->in_pkts_unusedsa += IXGBE_READ_REG(hw, IXGBE_LSECRXUNSA);
3353 	macsec_stats->in_pkts_notusingsa +=
3354 		IXGBE_READ_REG(hw, IXGBE_LSECRXNUSA);
3355 }
3356 
3357 /*
3358  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
3359  */
3360 static int
3361 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3362 {
3363 	struct ixgbe_hw *hw =
3364 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3365 	struct ixgbe_hw_stats *hw_stats =
3366 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3367 	struct ixgbe_macsec_stats *macsec_stats =
3368 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3369 				dev->data->dev_private);
3370 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3371 	unsigned i;
3372 
3373 	total_missed_rx = 0;
3374 	total_qbrc = 0;
3375 	total_qprc = 0;
3376 	total_qprdc = 0;
3377 
3378 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3379 			&total_qbrc, &total_qprc, &total_qprdc);
3380 
3381 	if (stats == NULL)
3382 		return -EINVAL;
3383 
3384 	/* Fill out the rte_eth_stats statistics structure */
3385 	stats->ipackets = total_qprc;
3386 	stats->ibytes = total_qbrc;
3387 	stats->opackets = hw_stats->gptc;
3388 	stats->obytes = hw_stats->gotc;
3389 
3390 	for (i = 0; i < RTE_MIN_T(IXGBE_QUEUE_STAT_COUNTERS,
3391 			RTE_ETHDEV_QUEUE_STAT_CNTRS, typeof(i)); i++) {
3392 		stats->q_ipackets[i] = hw_stats->qprc[i];
3393 		stats->q_opackets[i] = hw_stats->qptc[i];
3394 		stats->q_ibytes[i] = hw_stats->qbrc[i];
3395 		stats->q_obytes[i] = hw_stats->qbtc[i];
3396 		stats->q_errors[i] = hw_stats->qprdc[i];
3397 	}
3398 
3399 	/* Rx Errors */
3400 	stats->imissed  = total_missed_rx;
3401 	stats->ierrors  = hw_stats->crcerrs +
3402 			  hw_stats->mspdc +
3403 			  hw_stats->rlec +
3404 			  hw_stats->ruc +
3405 			  hw_stats->roc +
3406 			  hw_stats->illerrc +
3407 			  hw_stats->errbc +
3408 			  hw_stats->rfc +
3409 			  hw_stats->fccrc +
3410 			  hw_stats->fclast;
3411 
3412 	/*
3413 	 * 82599 errata, UDP frames with a 0 checksum can be marked as checksum
3414 	 * errors.
3415 	 */
3416 	if (hw->mac.type != ixgbe_mac_82599EB)
3417 		stats->ierrors += hw_stats->xec;
3418 
3419 	/* Tx Errors */
3420 	stats->oerrors  = 0;
3421 	return 0;
3422 }
3423 
3424 static int
3425 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
3426 {
3427 	struct ixgbe_hw_stats *stats =
3428 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3429 
3430 	/* HW registers are cleared on read */
3431 	ixgbe_dev_stats_get(dev, NULL);
3432 
3433 	/* Reset software totals */
3434 	memset(stats, 0, sizeof(*stats));
3435 
3436 	return 0;
3437 }
3438 
3439 /* This function calculates the number of xstats based on the current config */
3440 static unsigned
3441 ixgbe_xstats_calc_num(void) {
3442 	return IXGBE_NB_HW_STATS + IXGBE_NB_MACSEC_STATS +
3443 		(IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
3444 		(IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
3445 }
3446 
3447 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3448 	struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned int size)
3449 {
3450 	const unsigned cnt_stats = ixgbe_xstats_calc_num();
3451 	unsigned stat, i, count;
3452 
3453 	if (xstats_names != NULL) {
3454 		count = 0;
3455 
3456 		/* Note: limit >= cnt_stats checked upstream
3457 		 * in rte_eth_xstats_names()
3458 		 */
3459 
3460 		/* Extended stats from ixgbe_hw_stats */
3461 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3462 			strlcpy(xstats_names[count].name,
3463 				rte_ixgbe_stats_strings[i].name,
3464 				sizeof(xstats_names[count].name));
3465 			count++;
3466 		}
3467 
3468 		/* MACsec Stats */
3469 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3470 			strlcpy(xstats_names[count].name,
3471 				rte_ixgbe_macsec_strings[i].name,
3472 				sizeof(xstats_names[count].name));
3473 			count++;
3474 		}
3475 
3476 		/* RX Priority Stats */
3477 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3478 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3479 				snprintf(xstats_names[count].name,
3480 					sizeof(xstats_names[count].name),
3481 					"rx_priority%u_%s", i,
3482 					rte_ixgbe_rxq_strings[stat].name);
3483 				count++;
3484 			}
3485 		}
3486 
3487 		/* TX Priority Stats */
3488 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3489 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3490 				snprintf(xstats_names[count].name,
3491 					sizeof(xstats_names[count].name),
3492 					"tx_priority%u_%s", i,
3493 					rte_ixgbe_txq_strings[stat].name);
3494 				count++;
3495 			}
3496 		}
3497 	}
3498 	return cnt_stats;
3499 }
3500 
3501 static int ixgbe_dev_xstats_get_names_by_id(
3502 	struct rte_eth_dev *dev,
3503 	const uint64_t *ids,
3504 	struct rte_eth_xstat_name *xstats_names,
3505 	unsigned int limit)
3506 {
3507 	if (!ids) {
3508 		const unsigned int cnt_stats = ixgbe_xstats_calc_num();
3509 		unsigned int stat, i, count;
3510 
3511 		if (xstats_names != NULL) {
3512 			count = 0;
3513 
3514 			/* Note: limit >= cnt_stats checked upstream
3515 			 * in rte_eth_xstats_names()
3516 			 */
3517 
3518 			/* Extended stats from ixgbe_hw_stats */
3519 			for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3520 				strlcpy(xstats_names[count].name,
3521 					rte_ixgbe_stats_strings[i].name,
3522 					sizeof(xstats_names[count].name));
3523 				count++;
3524 			}
3525 
3526 			/* MACsec Stats */
3527 			for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3528 				strlcpy(xstats_names[count].name,
3529 					rte_ixgbe_macsec_strings[i].name,
3530 					sizeof(xstats_names[count].name));
3531 				count++;
3532 			}
3533 
3534 			/* RX Priority Stats */
3535 			for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3536 				for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3537 					snprintf(xstats_names[count].name,
3538 					    sizeof(xstats_names[count].name),
3539 					    "rx_priority%u_%s", i,
3540 					    rte_ixgbe_rxq_strings[stat].name);
3541 					count++;
3542 				}
3543 			}
3544 
3545 			/* TX Priority Stats */
3546 			for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3547 				for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3548 					snprintf(xstats_names[count].name,
3549 					    sizeof(xstats_names[count].name),
3550 					    "tx_priority%u_%s", i,
3551 					    rte_ixgbe_txq_strings[stat].name);
3552 					count++;
3553 				}
3554 			}
3555 		}
3556 		return cnt_stats;
3557 	}
3558 
3559 	uint16_t i;
3560 	uint16_t size = ixgbe_xstats_calc_num();
3561 	struct rte_eth_xstat_name xstats_names_copy[size];
3562 
3563 	ixgbe_dev_xstats_get_names_by_id(dev, NULL, xstats_names_copy,
3564 			size);
3565 
3566 	for (i = 0; i < limit; i++) {
3567 		if (ids[i] >= size) {
3568 			PMD_INIT_LOG(ERR, "id value isn't valid");
3569 			return -1;
3570 		}
3571 		strcpy(xstats_names[i].name,
3572 				xstats_names_copy[ids[i]].name);
3573 	}
3574 	return limit;
3575 }
3576 
3577 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3578 	struct rte_eth_xstat_name *xstats_names, unsigned limit)
3579 {
3580 	unsigned i;
3581 
3582 	if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
3583 		return -ENOMEM;
3584 
3585 	if (xstats_names != NULL)
3586 		for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
3587 			strlcpy(xstats_names[i].name,
3588 				rte_ixgbevf_stats_strings[i].name,
3589 				sizeof(xstats_names[i].name));
3590 	return IXGBEVF_NB_XSTATS;
3591 }
3592 
3593 static int
3594 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3595 					 unsigned n)
3596 {
3597 	struct ixgbe_hw *hw =
3598 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3599 	struct ixgbe_hw_stats *hw_stats =
3600 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3601 	struct ixgbe_macsec_stats *macsec_stats =
3602 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3603 				dev->data->dev_private);
3604 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3605 	unsigned i, stat, count = 0;
3606 
3607 	count = ixgbe_xstats_calc_num();
3608 
3609 	if (n < count)
3610 		return count;
3611 
3612 	total_missed_rx = 0;
3613 	total_qbrc = 0;
3614 	total_qprc = 0;
3615 	total_qprdc = 0;
3616 
3617 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3618 			&total_qbrc, &total_qprc, &total_qprdc);
3619 
3620 	/* If this is a reset xstats is NULL, and we have cleared the
3621 	 * registers by reading them.
3622 	 */
3623 	if (!xstats)
3624 		return 0;
3625 
3626 	/* Extended stats from ixgbe_hw_stats */
3627 	count = 0;
3628 	for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3629 		xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3630 				rte_ixgbe_stats_strings[i].offset);
3631 		xstats[count].id = count;
3632 		count++;
3633 	}
3634 
3635 	/* MACsec Stats */
3636 	for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3637 		xstats[count].value = *(uint64_t *)(((char *)macsec_stats) +
3638 				rte_ixgbe_macsec_strings[i].offset);
3639 		xstats[count].id = count;
3640 		count++;
3641 	}
3642 
3643 	/* RX Priority Stats */
3644 	for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3645 		for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3646 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3647 					rte_ixgbe_rxq_strings[stat].offset +
3648 					(sizeof(uint64_t) * i));
3649 			xstats[count].id = count;
3650 			count++;
3651 		}
3652 	}
3653 
3654 	/* TX Priority Stats */
3655 	for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3656 		for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3657 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3658 					rte_ixgbe_txq_strings[stat].offset +
3659 					(sizeof(uint64_t) * i));
3660 			xstats[count].id = count;
3661 			count++;
3662 		}
3663 	}
3664 	return count;
3665 }
3666 
3667 static int
3668 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
3669 		uint64_t *values, unsigned int n)
3670 {
3671 	if (!ids) {
3672 		struct ixgbe_hw *hw =
3673 				IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3674 		struct ixgbe_hw_stats *hw_stats =
3675 				IXGBE_DEV_PRIVATE_TO_STATS(
3676 						dev->data->dev_private);
3677 		struct ixgbe_macsec_stats *macsec_stats =
3678 				IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3679 					dev->data->dev_private);
3680 		uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3681 		unsigned int i, stat, count = 0;
3682 
3683 		count = ixgbe_xstats_calc_num();
3684 
3685 		if (!ids && n < count)
3686 			return count;
3687 
3688 		total_missed_rx = 0;
3689 		total_qbrc = 0;
3690 		total_qprc = 0;
3691 		total_qprdc = 0;
3692 
3693 		ixgbe_read_stats_registers(hw, hw_stats, macsec_stats,
3694 				&total_missed_rx, &total_qbrc, &total_qprc,
3695 				&total_qprdc);
3696 
3697 		/* If this is a reset xstats is NULL, and we have cleared the
3698 		 * registers by reading them.
3699 		 */
3700 		if (!ids && !values)
3701 			return 0;
3702 
3703 		/* Extended stats from ixgbe_hw_stats */
3704 		count = 0;
3705 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3706 			values[count] = *(uint64_t *)(((char *)hw_stats) +
3707 					rte_ixgbe_stats_strings[i].offset);
3708 			count++;
3709 		}
3710 
3711 		/* MACsec Stats */
3712 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3713 			values[count] = *(uint64_t *)(((char *)macsec_stats) +
3714 					rte_ixgbe_macsec_strings[i].offset);
3715 			count++;
3716 		}
3717 
3718 		/* RX Priority Stats */
3719 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3720 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3721 				values[count] =
3722 					*(uint64_t *)(((char *)hw_stats) +
3723 					rte_ixgbe_rxq_strings[stat].offset +
3724 					(sizeof(uint64_t) * i));
3725 				count++;
3726 			}
3727 		}
3728 
3729 		/* TX Priority Stats */
3730 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3731 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3732 				values[count] =
3733 					*(uint64_t *)(((char *)hw_stats) +
3734 					rte_ixgbe_txq_strings[stat].offset +
3735 					(sizeof(uint64_t) * i));
3736 				count++;
3737 			}
3738 		}
3739 		return count;
3740 	}
3741 
3742 	uint16_t i;
3743 	uint16_t size = ixgbe_xstats_calc_num();
3744 	uint64_t values_copy[size];
3745 
3746 	ixgbe_dev_xstats_get_by_id(dev, NULL, values_copy, size);
3747 
3748 	for (i = 0; i < n; i++) {
3749 		if (ids[i] >= size) {
3750 			PMD_INIT_LOG(ERR, "id value isn't valid");
3751 			return -1;
3752 		}
3753 		values[i] = values_copy[ids[i]];
3754 	}
3755 	return n;
3756 }
3757 
3758 static int
3759 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
3760 {
3761 	struct ixgbe_hw_stats *stats =
3762 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3763 	struct ixgbe_macsec_stats *macsec_stats =
3764 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3765 				dev->data->dev_private);
3766 
3767 	unsigned count = ixgbe_xstats_calc_num();
3768 
3769 	/* HW registers are cleared on read */
3770 	ixgbe_dev_xstats_get(dev, NULL, count);
3771 
3772 	/* Reset software totals */
3773 	memset(stats, 0, sizeof(*stats));
3774 	memset(macsec_stats, 0, sizeof(*macsec_stats));
3775 
3776 	return 0;
3777 }
3778 
3779 static void
3780 ixgbevf_update_stats(struct rte_eth_dev *dev)
3781 {
3782 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3783 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3784 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3785 
3786 	/* Good Rx packet, include VF loopback */
3787 	UPDATE_VF_STAT(IXGBE_VFGPRC,
3788 	    hw_stats->last_vfgprc, hw_stats->vfgprc);
3789 
3790 	/* Good Rx octets, include VF loopback */
3791 	UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
3792 	    hw_stats->last_vfgorc, hw_stats->vfgorc);
3793 
3794 	/* Good Tx packet, include VF loopback */
3795 	UPDATE_VF_STAT(IXGBE_VFGPTC,
3796 	    hw_stats->last_vfgptc, hw_stats->vfgptc);
3797 
3798 	/* Good Tx octets, include VF loopback */
3799 	UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
3800 	    hw_stats->last_vfgotc, hw_stats->vfgotc);
3801 
3802 	/* Rx Multicst Packet */
3803 	UPDATE_VF_STAT(IXGBE_VFMPRC,
3804 	    hw_stats->last_vfmprc, hw_stats->vfmprc);
3805 }
3806 
3807 static int
3808 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3809 		       unsigned n)
3810 {
3811 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3812 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3813 	unsigned i;
3814 
3815 	if (n < IXGBEVF_NB_XSTATS)
3816 		return IXGBEVF_NB_XSTATS;
3817 
3818 	ixgbevf_update_stats(dev);
3819 
3820 	if (!xstats)
3821 		return 0;
3822 
3823 	/* Extended stats */
3824 	for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
3825 		xstats[i].id = i;
3826 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
3827 			rte_ixgbevf_stats_strings[i].offset);
3828 	}
3829 
3830 	return IXGBEVF_NB_XSTATS;
3831 }
3832 
3833 static int
3834 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3835 {
3836 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3837 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3838 
3839 	ixgbevf_update_stats(dev);
3840 
3841 	if (stats == NULL)
3842 		return -EINVAL;
3843 
3844 	stats->ipackets = hw_stats->vfgprc;
3845 	stats->ibytes = hw_stats->vfgorc;
3846 	stats->opackets = hw_stats->vfgptc;
3847 	stats->obytes = hw_stats->vfgotc;
3848 	return 0;
3849 }
3850 
3851 static int
3852 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3853 {
3854 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3855 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3856 
3857 	/* Sync HW register to the last stats */
3858 	ixgbevf_dev_stats_get(dev, NULL);
3859 
3860 	/* reset HW current stats*/
3861 	hw_stats->vfgprc = 0;
3862 	hw_stats->vfgorc = 0;
3863 	hw_stats->vfgptc = 0;
3864 	hw_stats->vfgotc = 0;
3865 	hw_stats->vfmprc = 0;
3866 
3867 	return 0;
3868 }
3869 
3870 static int
3871 ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
3872 {
3873 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3874 	struct ixgbe_nvm_version nvm_ver;
3875 	int ret;
3876 
3877 	ixgbe_get_oem_prod_version(hw, &nvm_ver);
3878 	if (nvm_ver.oem_valid) {
3879 		snprintf(fw_version, fw_size, "%x.%x.%x",
3880 			 nvm_ver.oem_major, nvm_ver.oem_minor,
3881 			 nvm_ver.oem_release);
3882 		return 0;
3883 	}
3884 
3885 	ixgbe_get_etk_id(hw, &nvm_ver);
3886 	ixgbe_get_orom_version(hw, &nvm_ver);
3887 
3888 	if (nvm_ver.or_valid) {
3889 		snprintf(fw_version, fw_size, "0x%08x, %d.%d.%d",
3890 			 nvm_ver.etk_id, nvm_ver.or_major,
3891 			 nvm_ver.or_build, nvm_ver.or_patch);
3892 		return 0;
3893 	}
3894 
3895 	ret = snprintf(fw_version, fw_size, "0x%08x", nvm_ver.etk_id);
3896 	if (ret < 0)
3897 		return -EINVAL;
3898 
3899 	return (fw_size < (size_t)ret++) ? ret : 0;
3900 }
3901 
3902 static int
3903 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3904 {
3905 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3906 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3907 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3908 
3909 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3910 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3911 	if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3912 		/*
3913 		 * When DCB/VT is off, maximum number of queues changes,
3914 		 * except for 82598EB, which remains constant.
3915 		 */
3916 		if (dev_conf->txmode.mq_mode == RTE_ETH_MQ_TX_NONE &&
3917 				hw->mac.type != ixgbe_mac_82598EB)
3918 			dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3919 	}
3920 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3921 	dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3922 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3923 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3924 	dev_info->max_vfs = pci_dev->max_vfs;
3925 	if (hw->mac.type == ixgbe_mac_82598EB)
3926 		dev_info->max_vmdq_pools = RTE_ETH_16_POOLS;
3927 	else
3928 		dev_info->max_vmdq_pools = RTE_ETH_64_POOLS;
3929 	dev_info->max_mtu =  dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3930 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
3931 	dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3932 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3933 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3934 				     dev_info->rx_queue_offload_capa);
3935 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3936 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3937 
3938 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3939 		.rx_thresh = {
3940 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3941 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3942 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3943 		},
3944 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3945 		.rx_drop_en = 0,
3946 		.offloads = 0,
3947 	};
3948 
3949 	dev_info->default_txconf = (struct rte_eth_txconf) {
3950 		.tx_thresh = {
3951 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3952 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3953 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3954 		},
3955 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3956 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3957 		.offloads = 0,
3958 	};
3959 
3960 	dev_info->rx_desc_lim = rx_desc_lim;
3961 	dev_info->tx_desc_lim = tx_desc_lim;
3962 
3963 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3964 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3965 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3966 
3967 	dev_info->speed_capa = RTE_ETH_LINK_SPEED_1G | RTE_ETH_LINK_SPEED_10G;
3968 	if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
3969 			hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
3970 		dev_info->speed_capa = RTE_ETH_LINK_SPEED_10M |
3971 			RTE_ETH_LINK_SPEED_100M | RTE_ETH_LINK_SPEED_1G;
3972 
3973 	if (hw->mac.type == ixgbe_mac_X540 ||
3974 	    hw->mac.type == ixgbe_mac_X540_vf ||
3975 	    hw->mac.type == ixgbe_mac_X550 ||
3976 	    hw->mac.type == ixgbe_mac_X550_vf) {
3977 		dev_info->speed_capa |= RTE_ETH_LINK_SPEED_100M;
3978 	}
3979 	if (hw->mac.type == ixgbe_mac_X550) {
3980 		dev_info->speed_capa |= RTE_ETH_LINK_SPEED_2_5G;
3981 		dev_info->speed_capa |= RTE_ETH_LINK_SPEED_5G;
3982 	}
3983 
3984 	/* Driver-preferred Rx/Tx parameters */
3985 	dev_info->default_rxportconf.burst_size = 32;
3986 	dev_info->default_txportconf.burst_size = 32;
3987 	dev_info->default_rxportconf.nb_queues = 1;
3988 	dev_info->default_txportconf.nb_queues = 1;
3989 	dev_info->default_rxportconf.ring_size = 256;
3990 	dev_info->default_txportconf.ring_size = 256;
3991 
3992 	return 0;
3993 }
3994 
3995 static const uint32_t *
3996 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev, size_t *no_of_elements)
3997 {
3998 	static const uint32_t ptypes[] = {
3999 		/* For non-vec functions,
4000 		 * refers to ixgbe_rxd_pkt_info_to_pkt_type();
4001 		 * for vec functions,
4002 		 * refers to _recv_raw_pkts_vec().
4003 		 */
4004 		RTE_PTYPE_L2_ETHER,
4005 		RTE_PTYPE_L3_IPV4,
4006 		RTE_PTYPE_L3_IPV4_EXT,
4007 		RTE_PTYPE_L3_IPV6,
4008 		RTE_PTYPE_L3_IPV6_EXT,
4009 		RTE_PTYPE_L4_SCTP,
4010 		RTE_PTYPE_L4_TCP,
4011 		RTE_PTYPE_L4_UDP,
4012 		RTE_PTYPE_TUNNEL_IP,
4013 		RTE_PTYPE_INNER_L3_IPV6,
4014 		RTE_PTYPE_INNER_L3_IPV6_EXT,
4015 		RTE_PTYPE_INNER_L4_TCP,
4016 		RTE_PTYPE_INNER_L4_UDP,
4017 	};
4018 
4019 	if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
4020 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
4021 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
4022 	    dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc) {
4023 		*no_of_elements = RTE_DIM(ptypes);
4024 		return ptypes;
4025 	}
4026 
4027 #if defined(RTE_ARCH_X86) || defined(__ARM_NEON)
4028 	if (dev->rx_pkt_burst == ixgbe_recv_pkts_vec ||
4029 	    dev->rx_pkt_burst == ixgbe_recv_scattered_pkts_vec) {
4030 		*no_of_elements = RTE_DIM(ptypes);
4031 		return ptypes;
4032 	}
4033 #endif
4034 	return NULL;
4035 }
4036 
4037 static int
4038 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
4039 		     struct rte_eth_dev_info *dev_info)
4040 {
4041 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4042 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4043 
4044 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
4045 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
4046 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
4047 	dev_info->max_rx_pktlen = 9728; /* includes CRC, cf MAXFRS reg */
4048 	dev_info->max_mtu = dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
4049 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
4050 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
4051 	dev_info->max_vfs = pci_dev->max_vfs;
4052 	if (hw->mac.type == ixgbe_mac_82598EB)
4053 		dev_info->max_vmdq_pools = RTE_ETH_16_POOLS;
4054 	else
4055 		dev_info->max_vmdq_pools = RTE_ETH_64_POOLS;
4056 	dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
4057 	dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
4058 				     dev_info->rx_queue_offload_capa);
4059 	dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
4060 	dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
4061 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
4062 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
4063 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
4064 
4065 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
4066 		.rx_thresh = {
4067 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
4068 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
4069 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
4070 		},
4071 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
4072 		.rx_drop_en = 0,
4073 		.offloads = 0,
4074 	};
4075 
4076 	dev_info->default_txconf = (struct rte_eth_txconf) {
4077 		.tx_thresh = {
4078 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
4079 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
4080 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
4081 		},
4082 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
4083 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
4084 		.offloads = 0,
4085 	};
4086 
4087 	dev_info->rx_desc_lim = rx_desc_lim;
4088 	dev_info->tx_desc_lim = tx_desc_lim;
4089 
4090 	dev_info->err_handle_mode = RTE_ETH_ERROR_HANDLE_MODE_PASSIVE;
4091 
4092 	return 0;
4093 }
4094 
4095 static int
4096 ixgbevf_check_link(struct ixgbe_hw *hw, ixgbe_link_speed *speed,
4097 		   bool *link_up, int wait_to_complete)
4098 {
4099 	struct ixgbe_adapter *adapter = container_of(hw,
4100 						     struct ixgbe_adapter, hw);
4101 	struct ixgbe_mbx_info *mbx = &hw->mbx;
4102 	struct ixgbe_mac_info *mac = &hw->mac;
4103 	uint32_t links_reg, in_msg;
4104 	int ret_val = 0;
4105 
4106 	/* If we were hit with a reset drop the link */
4107 	if (!mbx->ops[0].check_for_rst(hw, 0) || !mbx->timeout)
4108 		mac->get_link_status = true;
4109 
4110 	if (!mac->get_link_status)
4111 		goto out;
4112 
4113 	/* if link status is down no point in checking to see if pf is up */
4114 	links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4115 	if (!(links_reg & IXGBE_LINKS_UP))
4116 		goto out;
4117 
4118 	/* for SFP+ modules and DA cables on 82599 it can take up to 500usecs
4119 	 * before the link status is correct
4120 	 */
4121 	if (mac->type == ixgbe_mac_82599_vf && wait_to_complete) {
4122 		int i;
4123 
4124 		for (i = 0; i < 5; i++) {
4125 			rte_delay_us(100);
4126 			links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4127 
4128 			if (!(links_reg & IXGBE_LINKS_UP))
4129 				goto out;
4130 		}
4131 	}
4132 
4133 	switch (links_reg & IXGBE_LINKS_SPEED_82599) {
4134 	case IXGBE_LINKS_SPEED_10G_82599:
4135 		*speed = IXGBE_LINK_SPEED_10GB_FULL;
4136 		if (hw->mac.type >= ixgbe_mac_X550) {
4137 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4138 				*speed = IXGBE_LINK_SPEED_2_5GB_FULL;
4139 		}
4140 		break;
4141 	case IXGBE_LINKS_SPEED_1G_82599:
4142 		*speed = IXGBE_LINK_SPEED_1GB_FULL;
4143 		break;
4144 	case IXGBE_LINKS_SPEED_100_82599:
4145 		*speed = IXGBE_LINK_SPEED_100_FULL;
4146 		if (hw->mac.type == ixgbe_mac_X550) {
4147 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4148 				*speed = IXGBE_LINK_SPEED_5GB_FULL;
4149 		}
4150 		break;
4151 	case IXGBE_LINKS_SPEED_10_X550EM_A:
4152 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
4153 		/* Since Reserved in older MAC's */
4154 		if (hw->mac.type >= ixgbe_mac_X550)
4155 			*speed = IXGBE_LINK_SPEED_10_FULL;
4156 		break;
4157 	default:
4158 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
4159 	}
4160 
4161 	if (wait_to_complete == 0 && adapter->pflink_fullchk == 0) {
4162 		if (*speed == IXGBE_LINK_SPEED_UNKNOWN)
4163 			mac->get_link_status = true;
4164 		else
4165 			mac->get_link_status = false;
4166 
4167 		goto out;
4168 	}
4169 
4170 	/* if the read failed it could just be a mailbox collision, best wait
4171 	 * until we are called again and don't report an error
4172 	 */
4173 
4174 	/*
4175 	 * on MinGW, the read op call is interpreted as call into read() macro,
4176 	 * so avoid calling it directly.
4177 	 */
4178 	if ((mbx->ops[0].read)(hw, &in_msg, 1, 0))
4179 		goto out;
4180 
4181 	if (!(in_msg & IXGBE_VT_MSGTYPE_CTS)) {
4182 		/* msg is not CTS and is FAILURE we must have lost CTS status */
4183 		if (in_msg & IXGBE_VT_MSGTYPE_FAILURE)
4184 			mac->get_link_status = false;
4185 		goto out;
4186 	}
4187 
4188 	/* the pf is talking, if we timed out in the past we reinit */
4189 	if (!mbx->timeout) {
4190 		ret_val = -1;
4191 		goto out;
4192 	}
4193 
4194 	/* if we passed all the tests above then the link is up and we no
4195 	 * longer need to check for link
4196 	 */
4197 	mac->get_link_status = false;
4198 
4199 out:
4200 	*link_up = !mac->get_link_status;
4201 	return ret_val;
4202 }
4203 
4204 /*
4205  * If @timeout_ms was 0, it means that it will not return until link complete.
4206  * It returns 1 on complete, return 0 on timeout.
4207  */
4208 static int
4209 ixgbe_dev_wait_setup_link_complete(struct rte_eth_dev *dev, uint32_t timeout_ms)
4210 {
4211 #define WARNING_TIMEOUT    9000 /* 9s  in total */
4212 	struct ixgbe_adapter *ad = dev->data->dev_private;
4213 	uint32_t timeout = timeout_ms ? timeout_ms : WARNING_TIMEOUT;
4214 
4215 	/* NOTE: review for potential ordering optimization */
4216 	while (rte_atomic_load_explicit(&ad->link_thread_running, rte_memory_order_seq_cst)) {
4217 		msec_delay(1);
4218 		timeout--;
4219 
4220 		if (timeout_ms) {
4221 			if (!timeout)
4222 				return 0;
4223 		} else if (!timeout) {
4224 			/* It will not return until link complete */
4225 			timeout = WARNING_TIMEOUT;
4226 			PMD_DRV_LOG(ERR, "IXGBE link thread not complete too long time!");
4227 		}
4228 	}
4229 
4230 	return 1;
4231 }
4232 
4233 static uint32_t
4234 ixgbe_dev_setup_link_thread_handler(void *param)
4235 {
4236 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4237 	struct ixgbe_adapter *ad = dev->data->dev_private;
4238 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4239 	struct ixgbe_interrupt *intr =
4240 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4241 	u32 speed;
4242 	bool autoneg = false;
4243 
4244 	rte_thread_detach(rte_thread_self());
4245 	speed = hw->phy.autoneg_advertised;
4246 	if (!speed)
4247 		ixgbe_get_link_capabilities(hw, &speed, &autoneg);
4248 
4249 	ixgbe_setup_link(hw, speed, true);
4250 
4251 	intr->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4252 	/* NOTE: review for potential ordering optimization */
4253 	rte_atomic_store_explicit(&ad->link_thread_running, 0, rte_memory_order_seq_cst);
4254 	return 0;
4255 }
4256 
4257 /*
4258  * In freebsd environment, nic_uio drivers do not support interrupts,
4259  * rte_intr_callback_register() will fail to register interrupts.
4260  * We can not make link status to change from down to up by interrupt
4261  * callback. So we need to wait for the controller to acquire link
4262  * when ports start.
4263  * It returns 0 on link up.
4264  */
4265 static int
4266 ixgbe_wait_for_link_up(struct ixgbe_hw *hw)
4267 {
4268 #ifdef RTE_EXEC_ENV_FREEBSD
4269 	int err, i;
4270 	bool link_up = false;
4271 	uint32_t speed = 0;
4272 	const int nb_iter = 25;
4273 
4274 	for (i = 0; i < nb_iter; i++) {
4275 		err = ixgbe_check_link(hw, &speed, &link_up, 0);
4276 		if (err)
4277 			return err;
4278 		if (link_up)
4279 			return 0;
4280 		msec_delay(200);
4281 	}
4282 
4283 	return 0;
4284 #else
4285 	RTE_SET_USED(hw);
4286 	return 0;
4287 #endif
4288 }
4289 
4290 /* return 0 means link status changed, -1 means not changed */
4291 int
4292 ixgbe_dev_link_update_share(struct rte_eth_dev *dev,
4293 			    int wait_to_complete, int vf)
4294 {
4295 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4296 	struct ixgbe_adapter *ad = dev->data->dev_private;
4297 	struct rte_eth_link link;
4298 	ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
4299 	struct ixgbe_interrupt *intr =
4300 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4301 	bool link_up;
4302 	int diag;
4303 	int wait = 1;
4304 	u32 esdp_reg;
4305 
4306 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
4307 		return -1;
4308 
4309 	memset(&link, 0, sizeof(link));
4310 	link.link_status = RTE_ETH_LINK_DOWN;
4311 	link.link_speed = RTE_ETH_SPEED_NUM_NONE;
4312 	link.link_duplex = RTE_ETH_LINK_HALF_DUPLEX;
4313 	link.link_autoneg = !(dev->data->dev_conf.link_speeds &
4314 			RTE_ETH_LINK_SPEED_FIXED);
4315 
4316 	hw->mac.get_link_status = true;
4317 
4318 	if (intr->flags & IXGBE_FLAG_NEED_LINK_CONFIG)
4319 		return rte_eth_linkstatus_set(dev, &link);
4320 
4321 	/* check if it needs to wait to complete, if lsc interrupt is enabled */
4322 	if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
4323 		wait = 0;
4324 
4325 	if (vf)
4326 		diag = ixgbevf_check_link(hw, &link_speed, &link_up, wait);
4327 	else
4328 		diag = ixgbe_check_link(hw, &link_speed, &link_up, wait);
4329 
4330 	if (diag != 0) {
4331 		link.link_speed = RTE_ETH_SPEED_NUM_100M;
4332 		link.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
4333 		return rte_eth_linkstatus_set(dev, &link);
4334 	}
4335 
4336 	if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber &&
4337 	    !ad->sdp3_no_tx_disable) {
4338 		esdp_reg = IXGBE_READ_REG(hw, IXGBE_ESDP);
4339 		if ((esdp_reg & IXGBE_ESDP_SDP3))
4340 			link_up = 0;
4341 	}
4342 
4343 	if (link_up == 0) {
4344 		if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4345 			ixgbe_dev_wait_setup_link_complete(dev, 0);
4346 			/* NOTE: review for potential ordering optimization */
4347 			if (!rte_atomic_exchange_explicit(&ad->link_thread_running, 1,
4348 					rte_memory_order_seq_cst)) {
4349 				/* To avoid race condition between threads, set
4350 				 * the IXGBE_FLAG_NEED_LINK_CONFIG flag only
4351 				 * when there is no link thread running.
4352 				 */
4353 				intr->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
4354 				if (rte_thread_create_internal_control(&ad->link_thread_tid,
4355 						"ixgbe-link",
4356 						ixgbe_dev_setup_link_thread_handler, dev) < 0) {
4357 					PMD_DRV_LOG(ERR,
4358 						"Create link thread failed!");
4359 					/* NOTE: review for potential ordering optimization */
4360 					rte_atomic_store_explicit(&ad->link_thread_running, 0,
4361 							rte_memory_order_seq_cst);
4362 				}
4363 			} else {
4364 				PMD_DRV_LOG(ERR,
4365 					"Other link thread is running now!");
4366 			}
4367 		}
4368 		return rte_eth_linkstatus_set(dev, &link);
4369 	}
4370 
4371 	link.link_status = RTE_ETH_LINK_UP;
4372 	link.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
4373 
4374 	switch (link_speed) {
4375 	default:
4376 	case IXGBE_LINK_SPEED_UNKNOWN:
4377 		link.link_speed = RTE_ETH_SPEED_NUM_UNKNOWN;
4378 		break;
4379 
4380 	case IXGBE_LINK_SPEED_10_FULL:
4381 		link.link_speed = RTE_ETH_SPEED_NUM_10M;
4382 		break;
4383 
4384 	case IXGBE_LINK_SPEED_100_FULL:
4385 		link.link_speed = RTE_ETH_SPEED_NUM_100M;
4386 		break;
4387 
4388 	case IXGBE_LINK_SPEED_1GB_FULL:
4389 		link.link_speed = RTE_ETH_SPEED_NUM_1G;
4390 		break;
4391 
4392 	case IXGBE_LINK_SPEED_2_5GB_FULL:
4393 		link.link_speed = RTE_ETH_SPEED_NUM_2_5G;
4394 		break;
4395 
4396 	case IXGBE_LINK_SPEED_5GB_FULL:
4397 		link.link_speed = RTE_ETH_SPEED_NUM_5G;
4398 		break;
4399 
4400 	case IXGBE_LINK_SPEED_10GB_FULL:
4401 		link.link_speed = RTE_ETH_SPEED_NUM_10G;
4402 		break;
4403 	}
4404 
4405 	return rte_eth_linkstatus_set(dev, &link);
4406 }
4407 
4408 static int
4409 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4410 {
4411 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 0);
4412 }
4413 
4414 static int
4415 ixgbevf_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4416 {
4417 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 1);
4418 }
4419 
4420 static int
4421 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
4422 {
4423 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4424 	uint32_t fctrl;
4425 
4426 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4427 	fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4428 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4429 
4430 	return 0;
4431 }
4432 
4433 static int
4434 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
4435 {
4436 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4437 	uint32_t fctrl;
4438 
4439 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4440 	fctrl &= (~IXGBE_FCTRL_UPE);
4441 	if (dev->data->all_multicast == 1)
4442 		fctrl |= IXGBE_FCTRL_MPE;
4443 	else
4444 		fctrl &= (~IXGBE_FCTRL_MPE);
4445 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4446 
4447 	return 0;
4448 }
4449 
4450 static int
4451 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
4452 {
4453 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4454 	uint32_t fctrl;
4455 
4456 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4457 	fctrl |= IXGBE_FCTRL_MPE;
4458 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4459 
4460 	return 0;
4461 }
4462 
4463 static int
4464 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
4465 {
4466 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4467 	uint32_t fctrl;
4468 
4469 	if (dev->data->promiscuous == 1)
4470 		return 0; /* must remain in all_multicast mode */
4471 
4472 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4473 	fctrl &= (~IXGBE_FCTRL_MPE);
4474 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4475 
4476 	return 0;
4477 }
4478 
4479 /**
4480  * It clears the interrupt causes and enables the interrupt.
4481  * It will be called once only during nic initialized.
4482  *
4483  * @param dev
4484  *  Pointer to struct rte_eth_dev.
4485  * @param on
4486  *  Enable or Disable.
4487  *
4488  * @return
4489  *  - On success, zero.
4490  *  - On failure, a negative value.
4491  */
4492 static int
4493 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
4494 {
4495 	struct ixgbe_interrupt *intr =
4496 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4497 
4498 	ixgbe_dev_link_status_print(dev);
4499 	if (on)
4500 		intr->mask |= IXGBE_EICR_LSC;
4501 	else
4502 		intr->mask &= ~IXGBE_EICR_LSC;
4503 
4504 	return 0;
4505 }
4506 
4507 /**
4508  * It clears the interrupt causes and enables the interrupt.
4509  * It will be called once only during nic initialized.
4510  *
4511  * @param dev
4512  *  Pointer to struct rte_eth_dev.
4513  *
4514  * @return
4515  *  - On success, zero.
4516  *  - On failure, a negative value.
4517  */
4518 static int
4519 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
4520 {
4521 	struct ixgbe_interrupt *intr =
4522 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4523 
4524 	intr->mask |= IXGBE_EICR_RTX_QUEUE;
4525 
4526 	return 0;
4527 }
4528 
4529 /**
4530  * It clears the interrupt causes and enables the interrupt.
4531  * It will be called once only during nic initialized.
4532  *
4533  * @param dev
4534  *  Pointer to struct rte_eth_dev.
4535  *
4536  * @return
4537  *  - On success, zero.
4538  *  - On failure, a negative value.
4539  */
4540 static int
4541 ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev)
4542 {
4543 	struct ixgbe_interrupt *intr =
4544 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4545 
4546 	intr->mask |= IXGBE_EICR_LINKSEC;
4547 
4548 	return 0;
4549 }
4550 
4551 /*
4552  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
4553  *
4554  * @param dev
4555  *  Pointer to struct rte_eth_dev.
4556  *
4557  * @return
4558  *  - On success, zero.
4559  *  - On failure, a negative value.
4560  */
4561 static int
4562 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
4563 {
4564 	uint32_t eicr;
4565 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4566 	struct ixgbe_interrupt *intr =
4567 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4568 
4569 	/* clear all cause mask */
4570 	ixgbe_disable_intr(hw);
4571 
4572 	/* read-on-clear nic registers here */
4573 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4574 	PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
4575 
4576 	intr->flags = 0;
4577 
4578 	/* set flag for async link update */
4579 	if (eicr & IXGBE_EICR_LSC)
4580 		intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4581 
4582 	if (eicr & IXGBE_EICR_MAILBOX)
4583 		intr->flags |= IXGBE_FLAG_MAILBOX;
4584 
4585 	if (eicr & IXGBE_EICR_LINKSEC)
4586 		intr->flags |= IXGBE_FLAG_MACSEC;
4587 
4588 	if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
4589 	    hw->phy.type == ixgbe_phy_x550em_ext_t &&
4590 	    (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
4591 		intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
4592 
4593 	return 0;
4594 }
4595 
4596 /**
4597  * It gets and then prints the link status.
4598  *
4599  * @param dev
4600  *  Pointer to struct rte_eth_dev.
4601  *
4602  * @return
4603  *  - On success, zero.
4604  *  - On failure, a negative value.
4605  */
4606 static void
4607 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
4608 {
4609 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4610 	struct rte_eth_link link;
4611 
4612 	rte_eth_linkstatus_get(dev, &link);
4613 
4614 	if (link.link_status) {
4615 		PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
4616 					(int)(dev->data->port_id),
4617 					(unsigned)link.link_speed,
4618 			link.link_duplex == RTE_ETH_LINK_FULL_DUPLEX ?
4619 					"full-duplex" : "half-duplex");
4620 	} else {
4621 		PMD_INIT_LOG(INFO, " Port %d: Link Down",
4622 				(int)(dev->data->port_id));
4623 	}
4624 	PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
4625 				pci_dev->addr.domain,
4626 				pci_dev->addr.bus,
4627 				pci_dev->addr.devid,
4628 				pci_dev->addr.function);
4629 }
4630 
4631 /*
4632  * It executes link_update after knowing an interrupt occurred.
4633  *
4634  * @param dev
4635  *  Pointer to struct rte_eth_dev.
4636  *
4637  * @return
4638  *  - On success, zero.
4639  *  - On failure, a negative value.
4640  */
4641 static int
4642 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
4643 {
4644 	struct ixgbe_interrupt *intr =
4645 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4646 	int64_t timeout;
4647 	struct ixgbe_hw *hw =
4648 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4649 
4650 	PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
4651 
4652 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
4653 		ixgbe_pf_mbx_process(dev);
4654 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
4655 	}
4656 
4657 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4658 		ixgbe_handle_lasi(hw);
4659 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4660 	}
4661 
4662 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4663 		struct rte_eth_link link;
4664 
4665 		/* get the link status before link update, for predicting later */
4666 		rte_eth_linkstatus_get(dev, &link);
4667 
4668 		ixgbe_dev_link_update(dev, 0);
4669 
4670 		/* likely to up */
4671 		if (!link.link_status)
4672 			/* handle it 1 sec later, wait it being stable */
4673 			timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
4674 		/* likely to down */
4675 		else
4676 			/* handle it 4 sec later, wait it being stable */
4677 			timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
4678 
4679 		ixgbe_dev_link_status_print(dev);
4680 
4681 		/* Don't program delayed handler if LSC interrupt is disabled.
4682 		 * It means one is already programmed.
4683 		 */
4684 		if (intr->mask & IXGBE_EIMS_LSC) {
4685 			if (rte_eal_alarm_set(timeout * 1000,
4686 					      ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
4687 				PMD_DRV_LOG(ERR, "Error setting alarm");
4688 			else {
4689 				/* remember original mask */
4690 				intr->mask_original = intr->mask;
4691 				/* only disable lsc interrupt */
4692 				intr->mask &= ~IXGBE_EIMS_LSC;
4693 			}
4694 		}
4695 	}
4696 
4697 	PMD_DRV_LOG(DEBUG, "enable intr immediately");
4698 	ixgbe_enable_intr(dev);
4699 
4700 	return 0;
4701 }
4702 
4703 /**
4704  * Interrupt handler which shall be registered for alarm callback for delayed
4705  * handling specific interrupt to wait for the stable nic state. As the
4706  * NIC interrupt state is not stable for ixgbe after link is just down,
4707  * it needs to wait 4 seconds to get the stable status.
4708  *
4709  * @param handle
4710  *  Pointer to interrupt handle.
4711  * @param param
4712  *  The address of parameter (struct rte_eth_dev *) registered before.
4713  *
4714  * @return
4715  *  void
4716  */
4717 static void
4718 ixgbe_dev_interrupt_delayed_handler(void *param)
4719 {
4720 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4721 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4722 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
4723 	struct ixgbe_interrupt *intr =
4724 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4725 	struct ixgbe_hw *hw =
4726 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4727 	uint32_t eicr;
4728 
4729 	ixgbe_disable_intr(hw);
4730 
4731 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4732 	if (eicr & IXGBE_EICR_MAILBOX)
4733 		ixgbe_pf_mbx_process(dev);
4734 
4735 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4736 		ixgbe_handle_lasi(hw);
4737 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4738 	}
4739 
4740 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4741 		ixgbe_dev_link_update(dev, 0);
4742 		intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4743 		ixgbe_dev_link_status_print(dev);
4744 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
4745 	}
4746 
4747 	if (intr->flags & IXGBE_FLAG_MACSEC) {
4748 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_MACSEC, NULL);
4749 		intr->flags &= ~IXGBE_FLAG_MACSEC;
4750 	}
4751 
4752 	/* restore original mask */
4753 	intr->mask = intr->mask_original;
4754 	intr->mask_original = 0;
4755 
4756 	PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
4757 	ixgbe_enable_intr(dev);
4758 	rte_intr_ack(intr_handle);
4759 }
4760 
4761 /**
4762  * Interrupt handler triggered by NIC  for handling
4763  * specific interrupt.
4764  *
4765  * @param handle
4766  *  Pointer to interrupt handle.
4767  * @param param
4768  *  The address of parameter (struct rte_eth_dev *) registered before.
4769  *
4770  * @return
4771  *  void
4772  */
4773 static void
4774 ixgbe_dev_interrupt_handler(void *param)
4775 {
4776 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4777 
4778 	ixgbe_dev_interrupt_get_status(dev);
4779 	ixgbe_dev_interrupt_action(dev);
4780 }
4781 
4782 static int
4783 ixgbe_dev_led_on(struct rte_eth_dev *dev)
4784 {
4785 	struct ixgbe_hw *hw;
4786 
4787 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4788 	return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4789 }
4790 
4791 static int
4792 ixgbe_dev_led_off(struct rte_eth_dev *dev)
4793 {
4794 	struct ixgbe_hw *hw;
4795 
4796 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4797 	return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4798 }
4799 
4800 static int
4801 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4802 {
4803 	struct ixgbe_hw *hw;
4804 	uint32_t mflcn_reg;
4805 	uint32_t fccfg_reg;
4806 	int rx_pause;
4807 	int tx_pause;
4808 
4809 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4810 
4811 	fc_conf->pause_time = hw->fc.pause_time;
4812 	fc_conf->high_water = hw->fc.high_water[0];
4813 	fc_conf->low_water = hw->fc.low_water[0];
4814 	fc_conf->send_xon = hw->fc.send_xon;
4815 	fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
4816 
4817 	/*
4818 	 * Return rx_pause status according to actual setting of
4819 	 * MFLCN register.
4820 	 */
4821 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4822 	if (mflcn_reg & IXGBE_MFLCN_PMCF)
4823 		fc_conf->mac_ctrl_frame_fwd = 1;
4824 	else
4825 		fc_conf->mac_ctrl_frame_fwd = 0;
4826 
4827 	if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
4828 		rx_pause = 1;
4829 	else
4830 		rx_pause = 0;
4831 
4832 	/*
4833 	 * Return tx_pause status according to actual setting of
4834 	 * FCCFG register.
4835 	 */
4836 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4837 	if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
4838 		tx_pause = 1;
4839 	else
4840 		tx_pause = 0;
4841 
4842 	if (rx_pause && tx_pause)
4843 		fc_conf->mode = RTE_ETH_FC_FULL;
4844 	else if (rx_pause)
4845 		fc_conf->mode = RTE_ETH_FC_RX_PAUSE;
4846 	else if (tx_pause)
4847 		fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
4848 	else
4849 		fc_conf->mode = RTE_ETH_FC_NONE;
4850 
4851 	return 0;
4852 }
4853 
4854 static int
4855 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4856 {
4857 	struct ixgbe_hw *hw;
4858 	struct ixgbe_adapter *adapter = dev->data->dev_private;
4859 	int err;
4860 	uint32_t rx_buf_size;
4861 	uint32_t max_high_water;
4862 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4863 		ixgbe_fc_none,
4864 		ixgbe_fc_rx_pause,
4865 		ixgbe_fc_tx_pause,
4866 		ixgbe_fc_full
4867 	};
4868 
4869 	PMD_INIT_FUNC_TRACE();
4870 
4871 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4872 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
4873 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4874 
4875 	/*
4876 	 * At least reserve one Ethernet frame for watermark
4877 	 * high_water/low_water in kilo bytes for ixgbe
4878 	 */
4879 	max_high_water = (rx_buf_size -
4880 			RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4881 	if ((fc_conf->high_water > max_high_water) ||
4882 		(fc_conf->high_water < fc_conf->low_water)) {
4883 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4884 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4885 		return -EINVAL;
4886 	}
4887 
4888 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
4889 	hw->fc.pause_time     = fc_conf->pause_time;
4890 	hw->fc.high_water[0]  = fc_conf->high_water;
4891 	hw->fc.low_water[0]   = fc_conf->low_water;
4892 	hw->fc.send_xon       = fc_conf->send_xon;
4893 	hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
4894 	adapter->mac_ctrl_frame_fwd = fc_conf->mac_ctrl_frame_fwd;
4895 
4896 	err = ixgbe_flow_ctrl_enable(dev, hw);
4897 	if (err < 0) {
4898 		PMD_INIT_LOG(ERR, "ixgbe_flow_ctrl_enable = 0x%x", err);
4899 		return -EIO;
4900 	}
4901 	return err;
4902 }
4903 
4904 /**
4905  *  ixgbe_pfc_enable_generic - Enable flow control
4906  *  @hw: pointer to hardware structure
4907  *  @tc_num: traffic class number
4908  *  Enable flow control according to the current settings.
4909  */
4910 static int
4911 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
4912 {
4913 	int ret_val = 0;
4914 	uint32_t mflcn_reg, fccfg_reg;
4915 	uint32_t reg;
4916 	uint32_t fcrtl, fcrth;
4917 	uint8_t i;
4918 	uint8_t nb_rx_en;
4919 
4920 	/* Validate the water mark configuration */
4921 	if (!hw->fc.pause_time) {
4922 		ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4923 		goto out;
4924 	}
4925 
4926 	/* Low water mark of zero causes XOFF floods */
4927 	if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
4928 		 /* High/Low water can not be 0 */
4929 		if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_water[tc_num])) {
4930 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4931 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4932 			goto out;
4933 		}
4934 
4935 		if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
4936 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4937 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4938 			goto out;
4939 		}
4940 	}
4941 	/* Negotiate the fc mode to use */
4942 	ixgbe_fc_autoneg(hw);
4943 
4944 	/* Disable any previous flow control settings */
4945 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4946 	mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
4947 
4948 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4949 	fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
4950 
4951 	switch (hw->fc.current_mode) {
4952 	case ixgbe_fc_none:
4953 		/*
4954 		 * If the count of enabled RX Priority Flow control >1,
4955 		 * and the TX pause can not be disabled
4956 		 */
4957 		nb_rx_en = 0;
4958 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4959 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4960 			if (reg & IXGBE_FCRTH_FCEN)
4961 				nb_rx_en++;
4962 		}
4963 		if (nb_rx_en > 1)
4964 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4965 		break;
4966 	case ixgbe_fc_rx_pause:
4967 		/*
4968 		 * Rx Flow control is enabled and Tx Flow control is
4969 		 * disabled by software override. Since there really
4970 		 * isn't a way to advertise that we are capable of RX
4971 		 * Pause ONLY, we will advertise that we support both
4972 		 * symmetric and asymmetric Rx PAUSE.  Later, we will
4973 		 * disable the adapter's ability to send PAUSE frames.
4974 		 */
4975 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4976 		/*
4977 		 * If the count of enabled RX Priority Flow control >1,
4978 		 * and the TX pause can not be disabled
4979 		 */
4980 		nb_rx_en = 0;
4981 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4982 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4983 			if (reg & IXGBE_FCRTH_FCEN)
4984 				nb_rx_en++;
4985 		}
4986 		if (nb_rx_en > 1)
4987 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4988 		break;
4989 	case ixgbe_fc_tx_pause:
4990 		/*
4991 		 * Tx Flow control is enabled, and Rx Flow control is
4992 		 * disabled by software override.
4993 		 */
4994 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4995 		break;
4996 	case ixgbe_fc_full:
4997 		/* Flow control (both Rx and Tx) is enabled by SW override. */
4998 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4999 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
5000 		break;
5001 	default:
5002 		PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
5003 		ret_val = IXGBE_ERR_CONFIG;
5004 		goto out;
5005 	}
5006 
5007 	/* Set 802.3x based flow control settings. */
5008 	mflcn_reg |= IXGBE_MFLCN_DPF;
5009 	IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
5010 	IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
5011 
5012 	/* Set up and enable Rx high/low water mark thresholds, enable XON. */
5013 	if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
5014 		hw->fc.high_water[tc_num]) {
5015 		fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
5016 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
5017 		fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
5018 	} else {
5019 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
5020 		/*
5021 		 * In order to prevent Tx hangs when the internal Tx
5022 		 * switch is enabled we must set the high water mark
5023 		 * to the maximum FCRTH value.  This allows the Tx
5024 		 * switch to function even under heavy Rx workloads.
5025 		 */
5026 		fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
5027 	}
5028 	IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
5029 
5030 	/* Configure pause time (2 TCs per register) */
5031 	reg = hw->fc.pause_time * 0x00010001;
5032 	for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
5033 		IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
5034 
5035 	/* Configure flow control refresh threshold value */
5036 	IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
5037 
5038 out:
5039 	return ret_val;
5040 }
5041 
5042 static int
5043 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
5044 {
5045 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5046 	int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
5047 
5048 	if (hw->mac.type != ixgbe_mac_82598EB) {
5049 		ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
5050 	}
5051 	return ret_val;
5052 }
5053 
5054 static int
5055 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
5056 {
5057 	int err;
5058 	uint32_t rx_buf_size;
5059 	uint32_t max_high_water;
5060 	uint8_t tc_num;
5061 	uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
5062 	struct ixgbe_hw *hw =
5063 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5064 	struct ixgbe_dcb_config *dcb_config =
5065 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
5066 
5067 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
5068 		ixgbe_fc_none,
5069 		ixgbe_fc_rx_pause,
5070 		ixgbe_fc_tx_pause,
5071 		ixgbe_fc_full
5072 	};
5073 
5074 	PMD_INIT_FUNC_TRACE();
5075 
5076 	ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
5077 	tc_num = map[pfc_conf->priority];
5078 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
5079 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
5080 	/*
5081 	 * At least reserve one Ethernet frame for watermark
5082 	 * high_water/low_water in kilo bytes for ixgbe
5083 	 */
5084 	max_high_water = (rx_buf_size -
5085 			RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
5086 	if ((pfc_conf->fc.high_water > max_high_water) ||
5087 	    (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
5088 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
5089 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
5090 		return -EINVAL;
5091 	}
5092 
5093 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
5094 	hw->fc.pause_time = pfc_conf->fc.pause_time;
5095 	hw->fc.send_xon = pfc_conf->fc.send_xon;
5096 	hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
5097 	hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
5098 
5099 	err = ixgbe_dcb_pfc_enable(dev, tc_num);
5100 
5101 	/* Not negotiated is not an error case */
5102 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
5103 		return 0;
5104 
5105 	PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
5106 	return -EIO;
5107 }
5108 
5109 static int
5110 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
5111 			  struct rte_eth_rss_reta_entry64 *reta_conf,
5112 			  uint16_t reta_size)
5113 {
5114 	uint16_t i, sp_reta_size;
5115 	uint8_t j, mask;
5116 	uint32_t reta, r;
5117 	uint16_t idx, shift;
5118 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5119 	struct rte_eth_dev_data *dev_data = dev->data;
5120 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5121 	uint32_t reta_reg;
5122 
5123 	PMD_INIT_FUNC_TRACE();
5124 
5125 	if (!dev_data->dev_started) {
5126 		PMD_DRV_LOG(ERR,
5127 			"port %d must be started before rss reta update",
5128 			 dev_data->port_id);
5129 		return -EIO;
5130 	}
5131 
5132 	if (!ixgbe_rss_update_sp(hw->mac.type)) {
5133 		PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
5134 			"NIC.");
5135 		return -ENOTSUP;
5136 	}
5137 
5138 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5139 	if (reta_size != sp_reta_size) {
5140 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5141 			"(%d) doesn't match the number hardware can supported "
5142 			"(%d)", reta_size, sp_reta_size);
5143 		return -EINVAL;
5144 	}
5145 
5146 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5147 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
5148 		shift = i % RTE_ETH_RETA_GROUP_SIZE;
5149 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5150 						IXGBE_4_BIT_MASK);
5151 		if (!mask)
5152 			continue;
5153 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5154 		if (mask == IXGBE_4_BIT_MASK)
5155 			r = 0;
5156 		else
5157 			r = IXGBE_READ_REG(hw, reta_reg);
5158 		for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5159 			if (mask & (0x1 << j))
5160 				reta |= reta_conf[idx].reta[shift + j] <<
5161 							(CHAR_BIT * j);
5162 			else
5163 				reta |= r & (IXGBE_8_BIT_MASK <<
5164 						(CHAR_BIT * j));
5165 		}
5166 		IXGBE_WRITE_REG(hw, reta_reg, reta);
5167 	}
5168 	adapter->rss_reta_updated = 1;
5169 
5170 	return 0;
5171 }
5172 
5173 static int
5174 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
5175 			 struct rte_eth_rss_reta_entry64 *reta_conf,
5176 			 uint16_t reta_size)
5177 {
5178 	uint16_t i, sp_reta_size;
5179 	uint8_t j, mask;
5180 	uint32_t reta;
5181 	uint16_t idx, shift;
5182 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5183 	uint32_t reta_reg;
5184 
5185 	PMD_INIT_FUNC_TRACE();
5186 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5187 	if (reta_size != sp_reta_size) {
5188 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5189 			"(%d) doesn't match the number hardware can supported "
5190 			"(%d)", reta_size, sp_reta_size);
5191 		return -EINVAL;
5192 	}
5193 
5194 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5195 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
5196 		shift = i % RTE_ETH_RETA_GROUP_SIZE;
5197 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5198 						IXGBE_4_BIT_MASK);
5199 		if (!mask)
5200 			continue;
5201 
5202 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5203 		reta = IXGBE_READ_REG(hw, reta_reg);
5204 		for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5205 			if (mask & (0x1 << j))
5206 				reta_conf[idx].reta[shift + j] =
5207 					((reta >> (CHAR_BIT * j)) &
5208 						IXGBE_8_BIT_MASK);
5209 		}
5210 	}
5211 
5212 	return 0;
5213 }
5214 
5215 static int
5216 ixgbe_add_rar(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
5217 				uint32_t index, uint32_t pool)
5218 {
5219 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5220 	uint32_t enable_addr = 1;
5221 
5222 	return ixgbe_set_rar(hw, index, mac_addr->addr_bytes,
5223 			     pool, enable_addr);
5224 }
5225 
5226 static void
5227 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
5228 {
5229 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5230 
5231 	ixgbe_clear_rar(hw, index);
5232 }
5233 
5234 static int
5235 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
5236 {
5237 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5238 
5239 	ixgbe_remove_rar(dev, 0);
5240 	ixgbe_add_rar(dev, addr, 0, pci_dev->max_vfs);
5241 
5242 	return 0;
5243 }
5244 
5245 static bool
5246 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
5247 {
5248 	if (strcmp(dev->device->driver->name, drv->driver.name))
5249 		return false;
5250 
5251 	return true;
5252 }
5253 
5254 bool
5255 is_ixgbe_supported(struct rte_eth_dev *dev)
5256 {
5257 	return is_device_supported(dev, &rte_ixgbe_pmd);
5258 }
5259 
5260 static int
5261 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
5262 {
5263 	uint32_t hlreg0;
5264 	uint32_t maxfrs;
5265 	struct ixgbe_hw *hw;
5266 	struct rte_eth_dev_info dev_info;
5267 	uint32_t frame_size = mtu + IXGBE_ETH_OVERHEAD;
5268 	int ret;
5269 
5270 	ret = ixgbe_dev_info_get(dev, &dev_info);
5271 	if (ret != 0)
5272 		return ret;
5273 
5274 	/* check that mtu is within the allowed range */
5275 	if (mtu < RTE_ETHER_MIN_MTU || frame_size > dev_info.max_rx_pktlen)
5276 		return -EINVAL;
5277 
5278 	/* If device is started, refuse mtu that requires the support of
5279 	 * scattered packets when this feature has not been enabled before.
5280 	 */
5281 	if (dev->data->dev_started && !dev->data->scattered_rx &&
5282 	    frame_size + 2 * RTE_VLAN_HLEN >
5283 			dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
5284 		PMD_INIT_LOG(ERR, "Stop port first.");
5285 		return -EINVAL;
5286 	}
5287 
5288 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5289 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
5290 
5291 	/* switch to jumbo mode if needed */
5292 	if (mtu > RTE_ETHER_MTU)
5293 		hlreg0 |= IXGBE_HLREG0_JUMBOEN;
5294 	else
5295 		hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
5296 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
5297 
5298 	maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
5299 	maxfrs &= 0x0000FFFF;
5300 	maxfrs |= (frame_size << 16);
5301 	IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
5302 
5303 	return 0;
5304 }
5305 
5306 /*
5307  * Virtual Function operations
5308  */
5309 static void
5310 ixgbevf_intr_disable(struct rte_eth_dev *dev)
5311 {
5312 	struct ixgbe_interrupt *intr =
5313 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5314 	struct ixgbe_hw *hw =
5315 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5316 
5317 	PMD_INIT_FUNC_TRACE();
5318 
5319 	/* Clear interrupt mask to stop from interrupts being generated */
5320 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
5321 
5322 	IXGBE_WRITE_FLUSH(hw);
5323 
5324 	/* Clear mask value. */
5325 	intr->mask = 0;
5326 }
5327 
5328 static void
5329 ixgbevf_intr_enable(struct rte_eth_dev *dev)
5330 {
5331 	struct ixgbe_interrupt *intr =
5332 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5333 	struct ixgbe_hw *hw =
5334 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5335 
5336 	PMD_INIT_FUNC_TRACE();
5337 
5338 	/* VF enable interrupt autoclean */
5339 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
5340 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
5341 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
5342 
5343 	IXGBE_WRITE_FLUSH(hw);
5344 
5345 	/* Save IXGBE_VTEIMS value to mask. */
5346 	intr->mask = IXGBE_VF_IRQ_ENABLE_MASK;
5347 }
5348 
5349 static int
5350 ixgbevf_dev_configure(struct rte_eth_dev *dev)
5351 {
5352 	struct rte_eth_conf *conf = &dev->data->dev_conf;
5353 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5354 
5355 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
5356 		     dev->data->port_id);
5357 
5358 	if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
5359 		dev->data->dev_conf.rxmode.offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
5360 
5361 	/*
5362 	 * VF has no ability to enable/disable HW CRC
5363 	 * Keep the persistent behavior the same as Host PF
5364 	 */
5365 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
5366 	if (conf->rxmode.offloads & RTE_ETH_RX_OFFLOAD_KEEP_CRC) {
5367 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
5368 		conf->rxmode.offloads &= ~RTE_ETH_RX_OFFLOAD_KEEP_CRC;
5369 	}
5370 #else
5371 	if (!(conf->rxmode.offloads & RTE_ETH_RX_OFFLOAD_KEEP_CRC)) {
5372 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
5373 		conf->rxmode.offloads |= RTE_ETH_RX_OFFLOAD_KEEP_CRC;
5374 	}
5375 #endif
5376 
5377 	/*
5378 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
5379 	 * allocation or vector Rx preconditions we will reset it.
5380 	 */
5381 	adapter->rx_bulk_alloc_allowed = true;
5382 	adapter->rx_vec_allowed = true;
5383 
5384 	return 0;
5385 }
5386 
5387 static int
5388 ixgbevf_dev_start(struct rte_eth_dev *dev)
5389 {
5390 	struct ixgbe_hw *hw =
5391 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5392 	uint32_t intr_vector = 0;
5393 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5394 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5395 
5396 	int err, mask = 0;
5397 
5398 	PMD_INIT_FUNC_TRACE();
5399 
5400 	/* Stop the link setup handler before resetting the HW. */
5401 	ixgbe_dev_wait_setup_link_complete(dev, 0);
5402 
5403 	err = hw->mac.ops.reset_hw(hw);
5404 
5405 	/**
5406 	 * In this case, reuses the MAC address assigned by VF
5407 	 * initialization.
5408 	 */
5409 	if (err != IXGBE_SUCCESS && err != IXGBE_ERR_INVALID_MAC_ADDR) {
5410 		PMD_INIT_LOG(ERR, "Unable to reset vf hardware (%d)", err);
5411 		return err;
5412 	}
5413 
5414 	hw->mac.get_link_status = true;
5415 
5416 	/* negotiate mailbox API version to use with the PF. */
5417 	ixgbevf_negotiate_api(hw);
5418 
5419 	ixgbevf_dev_tx_init(dev);
5420 
5421 	/* This can fail when allocating mbufs for descriptor rings */
5422 	err = ixgbevf_dev_rx_init(dev);
5423 	if (err) {
5424 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
5425 		ixgbe_dev_clear_queues(dev);
5426 		return err;
5427 	}
5428 
5429 	/* Set vfta */
5430 	ixgbevf_set_vfta_all(dev, 1);
5431 
5432 	/* Set HW strip */
5433 	mask = RTE_ETH_VLAN_STRIP_MASK | RTE_ETH_VLAN_FILTER_MASK |
5434 		RTE_ETH_VLAN_EXTEND_MASK;
5435 	err = ixgbevf_vlan_offload_config(dev, mask);
5436 	if (err) {
5437 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
5438 		ixgbe_dev_clear_queues(dev);
5439 		return err;
5440 	}
5441 
5442 	ixgbevf_dev_rxtx_start(dev);
5443 
5444 	/* check and configure queue intr-vector mapping */
5445 	if (rte_intr_cap_multiple(intr_handle) &&
5446 	    dev->data->dev_conf.intr_conf.rxq) {
5447 		/* According to datasheet, only vector 0/1/2 can be used,
5448 		 * now only one vector is used for Rx queue
5449 		 */
5450 		intr_vector = 1;
5451 		if (rte_intr_efd_enable(intr_handle, intr_vector)) {
5452 			ixgbe_dev_clear_queues(dev);
5453 			return -1;
5454 		}
5455 	}
5456 
5457 	if (rte_intr_dp_is_en(intr_handle)) {
5458 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
5459 						   dev->data->nb_rx_queues)) {
5460 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
5461 				     " intr_vec", dev->data->nb_rx_queues);
5462 			ixgbe_dev_clear_queues(dev);
5463 			return -ENOMEM;
5464 		}
5465 	}
5466 	ixgbevf_configure_msix(dev);
5467 
5468 	/* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
5469 	 * is mapped to VFIO vector 0 in eth_ixgbevf_dev_init( ).
5470 	 * If previous VFIO interrupt mapping setting in eth_ixgbevf_dev_init( )
5471 	 * is not cleared, it will fail when following rte_intr_enable( ) tries
5472 	 * to map Rx queue interrupt to other VFIO vectors.
5473 	 * So clear uio/vfio intr/evevnfd first to avoid failure.
5474 	 */
5475 	rte_intr_disable(intr_handle);
5476 
5477 	rte_intr_enable(intr_handle);
5478 
5479 	/* Re-enable interrupt for VF */
5480 	ixgbevf_intr_enable(dev);
5481 
5482 	/*
5483 	 * Update link status right before return, because it may
5484 	 * start link configuration process in a separate thread.
5485 	 */
5486 	ixgbevf_dev_link_update(dev, 0);
5487 
5488 	hw->adapter_stopped = false;
5489 
5490 	return 0;
5491 }
5492 
5493 static int
5494 ixgbevf_dev_stop(struct rte_eth_dev *dev)
5495 {
5496 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5497 	struct ixgbe_adapter *adapter = dev->data->dev_private;
5498 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5499 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5500 
5501 	if (hw->adapter_stopped)
5502 		return 0;
5503 
5504 	PMD_INIT_FUNC_TRACE();
5505 
5506 	ixgbe_dev_wait_setup_link_complete(dev, 0);
5507 
5508 	ixgbevf_intr_disable(dev);
5509 
5510 	dev->data->dev_started = 0;
5511 	hw->adapter_stopped = 1;
5512 	ixgbe_stop_adapter(hw);
5513 
5514 	/*
5515 	  * Clear what we set, but we still keep shadow_vfta to
5516 	  * restore after device starts
5517 	  */
5518 	ixgbevf_set_vfta_all(dev, 0);
5519 
5520 	/* Clear stored conf */
5521 	dev->data->scattered_rx = 0;
5522 
5523 	ixgbe_dev_clear_queues(dev);
5524 
5525 	/* Clean datapath event and queue/vec mapping */
5526 	rte_intr_efd_disable(intr_handle);
5527 	rte_intr_vec_list_free(intr_handle);
5528 
5529 	adapter->rss_reta_updated = 0;
5530 
5531 	return 0;
5532 }
5533 
5534 static int
5535 ixgbevf_dev_close(struct rte_eth_dev *dev)
5536 {
5537 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5538 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5539 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5540 	int ret;
5541 
5542 	PMD_INIT_FUNC_TRACE();
5543 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5544 		return 0;
5545 
5546 	ixgbe_reset_hw(hw);
5547 
5548 	ret = ixgbevf_dev_stop(dev);
5549 
5550 	ixgbe_dev_free_queues(dev);
5551 
5552 	/**
5553 	 * Remove the VF MAC address ro ensure
5554 	 * that the VF traffic goes to the PF
5555 	 * after stop, close and detach of the VF
5556 	 **/
5557 	ixgbevf_remove_mac_addr(dev, 0);
5558 
5559 	rte_intr_disable(intr_handle);
5560 	rte_intr_callback_unregister(intr_handle,
5561 				     ixgbevf_dev_interrupt_handler, dev);
5562 
5563 	return ret;
5564 }
5565 
5566 /*
5567  * Reset VF device
5568  */
5569 static int
5570 ixgbevf_dev_reset(struct rte_eth_dev *dev)
5571 {
5572 	int ret;
5573 
5574 	ret = eth_ixgbevf_dev_uninit(dev);
5575 	if (ret)
5576 		return ret;
5577 
5578 	ret = eth_ixgbevf_dev_init(dev);
5579 
5580 	return ret;
5581 }
5582 
5583 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
5584 {
5585 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5586 	struct ixgbe_vfta *shadow_vfta =
5587 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5588 	int i = 0, j = 0, vfta = 0, mask = 1;
5589 
5590 	for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
5591 		vfta = shadow_vfta->vfta[i];
5592 		if (vfta) {
5593 			mask = 1;
5594 			for (j = 0; j < 32; j++) {
5595 				if (vfta & mask)
5596 					ixgbe_set_vfta(hw, (i<<5)+j, 0,
5597 						       on, false);
5598 				mask <<= 1;
5599 			}
5600 		}
5601 	}
5602 
5603 }
5604 
5605 static int
5606 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
5607 {
5608 	struct ixgbe_hw *hw =
5609 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5610 	struct ixgbe_vfta *shadow_vfta =
5611 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5612 	uint32_t vid_idx = 0;
5613 	uint32_t vid_bit = 0;
5614 	int ret = 0;
5615 
5616 	PMD_INIT_FUNC_TRACE();
5617 
5618 	/* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
5619 	ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
5620 	if (ret) {
5621 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
5622 		return ret;
5623 	}
5624 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
5625 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
5626 
5627 	/* Save what we set and retore it after device reset */
5628 	if (on)
5629 		shadow_vfta->vfta[vid_idx] |= vid_bit;
5630 	else
5631 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
5632 
5633 	return 0;
5634 }
5635 
5636 static void
5637 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
5638 {
5639 	struct ixgbe_hw *hw =
5640 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5641 	uint32_t ctrl;
5642 
5643 	PMD_INIT_FUNC_TRACE();
5644 
5645 	if (queue >= hw->mac.max_rx_queues)
5646 		return;
5647 
5648 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
5649 	if (on)
5650 		ctrl |= IXGBE_RXDCTL_VME;
5651 	else
5652 		ctrl &= ~IXGBE_RXDCTL_VME;
5653 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
5654 
5655 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
5656 }
5657 
5658 static int
5659 ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask)
5660 {
5661 	struct ixgbe_rx_queue *rxq;
5662 	uint16_t i;
5663 	int on = 0;
5664 
5665 	/* VF function only support hw strip feature, others are not support */
5666 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
5667 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
5668 			rxq = dev->data->rx_queues[i];
5669 			on = !!(rxq->offloads &	RTE_ETH_RX_OFFLOAD_VLAN_STRIP);
5670 			ixgbevf_vlan_strip_queue_set(dev, i, on);
5671 		}
5672 	}
5673 
5674 	return 0;
5675 }
5676 
5677 static int
5678 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
5679 {
5680 	ixgbe_config_vlan_strip_on_all_queues(dev, mask);
5681 
5682 	ixgbevf_vlan_offload_config(dev, mask);
5683 
5684 	return 0;
5685 }
5686 
5687 int
5688 ixgbe_vt_check(struct ixgbe_hw *hw)
5689 {
5690 	uint32_t reg_val;
5691 
5692 	/* if Virtualization Technology is enabled */
5693 	reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
5694 	if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
5695 		PMD_INIT_LOG(ERR, "VT must be enabled for this setting");
5696 		return -1;
5697 	}
5698 
5699 	return 0;
5700 }
5701 
5702 static uint32_t
5703 ixgbe_uta_vector(struct ixgbe_hw *hw, struct rte_ether_addr *uc_addr)
5704 {
5705 	uint32_t vector = 0;
5706 
5707 	switch (hw->mac.mc_filter_type) {
5708 	case 0:   /* use bits [47:36] of the address */
5709 		vector = ((uc_addr->addr_bytes[4] >> 4) |
5710 			(((uint16_t)uc_addr->addr_bytes[5]) << 4));
5711 		break;
5712 	case 1:   /* use bits [46:35] of the address */
5713 		vector = ((uc_addr->addr_bytes[4] >> 3) |
5714 			(((uint16_t)uc_addr->addr_bytes[5]) << 5));
5715 		break;
5716 	case 2:   /* use bits [45:34] of the address */
5717 		vector = ((uc_addr->addr_bytes[4] >> 2) |
5718 			(((uint16_t)uc_addr->addr_bytes[5]) << 6));
5719 		break;
5720 	case 3:   /* use bits [43:32] of the address */
5721 		vector = ((uc_addr->addr_bytes[4]) |
5722 			(((uint16_t)uc_addr->addr_bytes[5]) << 8));
5723 		break;
5724 	default:  /* Invalid mc_filter_type */
5725 		break;
5726 	}
5727 
5728 	/* vector can only be 12-bits or boundary will be exceeded */
5729 	vector &= 0xFFF;
5730 	return vector;
5731 }
5732 
5733 static int
5734 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev,
5735 			struct rte_ether_addr *mac_addr, uint8_t on)
5736 {
5737 	uint32_t vector;
5738 	uint32_t uta_idx;
5739 	uint32_t reg_val;
5740 	uint32_t uta_shift;
5741 	uint32_t rc;
5742 	const uint32_t ixgbe_uta_idx_mask = 0x7F;
5743 	const uint32_t ixgbe_uta_bit_shift = 5;
5744 	const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
5745 	const uint32_t bit1 = 0x1;
5746 
5747 	struct ixgbe_hw *hw =
5748 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5749 	struct ixgbe_uta_info *uta_info =
5750 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5751 
5752 	/* The UTA table only exists on 82599 hardware and newer */
5753 	if (hw->mac.type < ixgbe_mac_82599EB)
5754 		return -ENOTSUP;
5755 
5756 	vector = ixgbe_uta_vector(hw, mac_addr);
5757 	uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
5758 	uta_shift = vector & ixgbe_uta_bit_mask;
5759 
5760 	rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
5761 	if (rc == on)
5762 		return 0;
5763 
5764 	reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
5765 	if (on) {
5766 		uta_info->uta_in_use++;
5767 		reg_val |= (bit1 << uta_shift);
5768 		uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
5769 	} else {
5770 		uta_info->uta_in_use--;
5771 		reg_val &= ~(bit1 << uta_shift);
5772 		uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
5773 	}
5774 
5775 	IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
5776 
5777 	if (uta_info->uta_in_use > 0)
5778 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
5779 				IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
5780 	else
5781 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
5782 
5783 	return 0;
5784 }
5785 
5786 static int
5787 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
5788 {
5789 	int i;
5790 	struct ixgbe_hw *hw =
5791 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5792 	struct ixgbe_uta_info *uta_info =
5793 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5794 
5795 	/* The UTA table only exists on 82599 hardware and newer */
5796 	if (hw->mac.type < ixgbe_mac_82599EB)
5797 		return -ENOTSUP;
5798 
5799 	if (on) {
5800 		for (i = 0; i < RTE_ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5801 			uta_info->uta_shadow[i] = ~0;
5802 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
5803 		}
5804 	} else {
5805 		for (i = 0; i < RTE_ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5806 			uta_info->uta_shadow[i] = 0;
5807 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
5808 		}
5809 	}
5810 	return 0;
5811 
5812 }
5813 
5814 uint32_t
5815 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
5816 {
5817 	uint32_t new_val = orig_val;
5818 
5819 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_UNTAG)
5820 		new_val |= IXGBE_VMOLR_AUPE;
5821 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_HASH_MC)
5822 		new_val |= IXGBE_VMOLR_ROMPE;
5823 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_HASH_UC)
5824 		new_val |= IXGBE_VMOLR_ROPE;
5825 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_BROADCAST)
5826 		new_val |= IXGBE_VMOLR_BAM;
5827 	if (rx_mask & RTE_ETH_VMDQ_ACCEPT_MULTICAST)
5828 		new_val |= IXGBE_VMOLR_MPE;
5829 
5830 	return new_val;
5831 }
5832 
5833 static int
5834 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5835 {
5836 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5837 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5838 	struct ixgbe_interrupt *intr =
5839 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5840 	struct ixgbe_hw *hw =
5841 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5842 	uint32_t vec = IXGBE_MISC_VEC_ID;
5843 
5844 	if (rte_intr_allow_others(intr_handle))
5845 		vec = IXGBE_RX_VEC_START;
5846 	intr->mask |= (1 << vec);
5847 	RTE_SET_USED(queue_id);
5848 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5849 
5850 	rte_intr_ack(intr_handle);
5851 
5852 	return 0;
5853 }
5854 
5855 static int
5856 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5857 {
5858 	struct ixgbe_interrupt *intr =
5859 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5860 	struct ixgbe_hw *hw =
5861 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5862 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5863 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5864 	uint32_t vec = IXGBE_MISC_VEC_ID;
5865 
5866 	if (rte_intr_allow_others(intr_handle))
5867 		vec = IXGBE_RX_VEC_START;
5868 	intr->mask &= ~(1 << vec);
5869 	RTE_SET_USED(queue_id);
5870 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5871 
5872 	return 0;
5873 }
5874 
5875 static int
5876 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5877 {
5878 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5879 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
5880 	uint32_t mask;
5881 	struct ixgbe_hw *hw =
5882 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5883 	struct ixgbe_interrupt *intr =
5884 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5885 
5886 	if (queue_id < 16) {
5887 		ixgbe_disable_intr(hw);
5888 		intr->mask |= (1 << queue_id);
5889 		ixgbe_enable_intr(dev);
5890 	} else if (queue_id < 32) {
5891 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5892 		mask &= (1 << queue_id);
5893 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5894 	} else if (queue_id < 64) {
5895 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5896 		mask &= (1 << (queue_id - 32));
5897 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5898 	}
5899 	rte_intr_ack(intr_handle);
5900 
5901 	return 0;
5902 }
5903 
5904 static int
5905 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5906 {
5907 	uint32_t mask;
5908 	struct ixgbe_hw *hw =
5909 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5910 	struct ixgbe_interrupt *intr =
5911 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5912 
5913 	if (queue_id < 16) {
5914 		ixgbe_disable_intr(hw);
5915 		intr->mask &= ~(1 << queue_id);
5916 		ixgbe_enable_intr(dev);
5917 	} else if (queue_id < 32) {
5918 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5919 		mask &= ~(1 << queue_id);
5920 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5921 	} else if (queue_id < 64) {
5922 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5923 		mask &= ~(1 << (queue_id - 32));
5924 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5925 	}
5926 
5927 	return 0;
5928 }
5929 
5930 static void
5931 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5932 		     uint8_t queue, uint8_t msix_vector)
5933 {
5934 	uint32_t tmp, idx;
5935 
5936 	if (direction == -1) {
5937 		/* other causes */
5938 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5939 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
5940 		tmp &= ~0xFF;
5941 		tmp |= msix_vector;
5942 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
5943 	} else {
5944 		/* rx or tx cause */
5945 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5946 		idx = ((16 * (queue & 1)) + (8 * direction));
5947 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
5948 		tmp &= ~(0xFF << idx);
5949 		tmp |= (msix_vector << idx);
5950 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
5951 	}
5952 }
5953 
5954 /**
5955  * set the IVAR registers, mapping interrupt causes to vectors
5956  * @param hw
5957  *  pointer to ixgbe_hw struct
5958  * @direction
5959  *  0 for Rx, 1 for Tx, -1 for other causes
5960  * @queue
5961  *  queue to map the corresponding interrupt to
5962  * @msix_vector
5963  *  the vector to map to the corresponding queue
5964  */
5965 static void
5966 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5967 		   uint8_t queue, uint8_t msix_vector)
5968 {
5969 	uint32_t tmp, idx;
5970 
5971 	msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5972 	if (hw->mac.type == ixgbe_mac_82598EB) {
5973 		if (direction == -1)
5974 			direction = 0;
5975 		idx = (((direction * 64) + queue) >> 2) & 0x1F;
5976 		tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
5977 		tmp &= ~(0xFF << (8 * (queue & 0x3)));
5978 		tmp |= (msix_vector << (8 * (queue & 0x3)));
5979 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
5980 	} else if ((hw->mac.type == ixgbe_mac_82599EB) ||
5981 			(hw->mac.type == ixgbe_mac_X540) ||
5982 			(hw->mac.type == ixgbe_mac_X550) ||
5983 			(hw->mac.type == ixgbe_mac_X550EM_a) ||
5984 			(hw->mac.type == ixgbe_mac_X550EM_x) ||
5985 			(hw->mac.type == ixgbe_mac_E610)) {
5986 		if (direction == -1) {
5987 			/* other causes */
5988 			idx = ((queue & 1) * 8);
5989 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
5990 			tmp &= ~(0xFF << idx);
5991 			tmp |= (msix_vector << idx);
5992 			IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
5993 		} else {
5994 			/* rx or tx causes */
5995 			idx = ((16 * (queue & 1)) + (8 * direction));
5996 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
5997 			tmp &= ~(0xFF << idx);
5998 			tmp |= (msix_vector << idx);
5999 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
6000 		}
6001 	}
6002 }
6003 
6004 static void
6005 ixgbevf_configure_msix(struct rte_eth_dev *dev)
6006 {
6007 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
6008 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
6009 	struct ixgbe_hw *hw =
6010 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6011 	uint32_t q_idx;
6012 	uint32_t vector_idx = IXGBE_MISC_VEC_ID;
6013 	uint32_t base = IXGBE_MISC_VEC_ID;
6014 
6015 	/* Configure VF other cause ivar */
6016 	ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
6017 
6018 	/* won't configure msix register if no mapping is done
6019 	 * between intr vector and event fd.
6020 	 */
6021 	if (!rte_intr_dp_is_en(intr_handle))
6022 		return;
6023 
6024 	if (rte_intr_allow_others(intr_handle)) {
6025 		base = IXGBE_RX_VEC_START;
6026 		vector_idx = IXGBE_RX_VEC_START;
6027 	}
6028 
6029 	/* Configure all RX queues of VF */
6030 	for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
6031 		/* Force all queue use vector 0,
6032 		 * as IXGBE_VF_MAXMSIVECTOR = 1
6033 		 */
6034 		ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
6035 		rte_intr_vec_list_index_set(intr_handle, q_idx,
6036 						   vector_idx);
6037 		if (vector_idx < base + rte_intr_nb_efd_get(intr_handle)
6038 		    - 1)
6039 			vector_idx++;
6040 	}
6041 
6042 	/* As RX queue setting above show, all queues use the vector 0.
6043 	 * Set only the ITR value of IXGBE_MISC_VEC_ID.
6044 	 */
6045 	IXGBE_WRITE_REG(hw, IXGBE_VTEITR(IXGBE_MISC_VEC_ID),
6046 			IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
6047 			| IXGBE_EITR_CNT_WDIS);
6048 }
6049 
6050 /**
6051  * Sets up the hardware to properly generate MSI-X interrupts
6052  * @hw
6053  *  board private structure
6054  */
6055 static void
6056 ixgbe_configure_msix(struct rte_eth_dev *dev)
6057 {
6058 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
6059 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
6060 	struct ixgbe_hw *hw =
6061 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6062 	uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
6063 	uint32_t vec = IXGBE_MISC_VEC_ID;
6064 	uint32_t mask;
6065 	uint32_t gpie;
6066 
6067 	/* won't configure msix register if no mapping is done
6068 	 * between intr vector and event fd
6069 	 * but if misx has been enabled already, need to configure
6070 	 * auto clean, auto mask and throttling.
6071 	 */
6072 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
6073 	if (!rte_intr_dp_is_en(intr_handle) &&
6074 	    !(gpie & (IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT)))
6075 		return;
6076 
6077 	if (rte_intr_allow_others(intr_handle))
6078 		vec = base = IXGBE_RX_VEC_START;
6079 
6080 	/* setup GPIE for MSI-x mode */
6081 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
6082 	gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
6083 		IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
6084 	/* auto clearing and auto setting corresponding bits in EIMS
6085 	 * when MSI-X interrupt is triggered
6086 	 */
6087 	if (hw->mac.type == ixgbe_mac_82598EB) {
6088 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
6089 	} else {
6090 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
6091 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
6092 	}
6093 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
6094 
6095 	/* Populate the IVAR table and set the ITR values to the
6096 	 * corresponding register.
6097 	 */
6098 	if (rte_intr_dp_is_en(intr_handle)) {
6099 		for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
6100 			queue_id++) {
6101 			/* by default, 1:1 mapping */
6102 			ixgbe_set_ivar_map(hw, 0, queue_id, vec);
6103 			rte_intr_vec_list_index_set(intr_handle,
6104 							   queue_id, vec);
6105 			if (vec < base + rte_intr_nb_efd_get(intr_handle)
6106 			    - 1)
6107 				vec++;
6108 		}
6109 
6110 		switch (hw->mac.type) {
6111 		case ixgbe_mac_82598EB:
6112 			ixgbe_set_ivar_map(hw, -1,
6113 					   IXGBE_IVAR_OTHER_CAUSES_INDEX,
6114 					   IXGBE_MISC_VEC_ID);
6115 			break;
6116 		case ixgbe_mac_82599EB:
6117 		case ixgbe_mac_X540:
6118 		case ixgbe_mac_X550:
6119 		case ixgbe_mac_X550EM_x:
6120 		case ixgbe_mac_E610:
6121 			ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
6122 			break;
6123 		default:
6124 			break;
6125 		}
6126 	}
6127 	IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
6128 			IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
6129 			| IXGBE_EITR_CNT_WDIS);
6130 
6131 	/* set up to autoclear timer, and the vectors */
6132 	mask = IXGBE_EIMS_ENABLE_MASK;
6133 	mask &= ~(IXGBE_EIMS_OTHER |
6134 		  IXGBE_EIMS_MAILBOX |
6135 		  IXGBE_EIMS_LSC);
6136 
6137 	IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
6138 }
6139 
6140 int
6141 ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
6142 			   uint16_t queue_idx, uint32_t tx_rate)
6143 {
6144 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6145 	uint32_t rf_dec, rf_int;
6146 	uint32_t bcnrc_val;
6147 	uint16_t link_speed = dev->data->dev_link.link_speed;
6148 
6149 	if (queue_idx >= hw->mac.max_tx_queues)
6150 		return -EINVAL;
6151 
6152 	if (tx_rate != 0) {
6153 		/* Calculate the rate factor values to set */
6154 		rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
6155 		rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
6156 		rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
6157 
6158 		bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
6159 		bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
6160 				IXGBE_RTTBCNRC_RF_INT_MASK_M);
6161 		bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
6162 	} else {
6163 		bcnrc_val = 0;
6164 	}
6165 
6166 	/*
6167 	 * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
6168 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
6169 	 * set as 0x4.
6170 	 */
6171 	if (dev->data->mtu + IXGBE_ETH_OVERHEAD >= IXGBE_MAX_JUMBO_FRAME_SIZE)
6172 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM, IXGBE_MMW_SIZE_JUMBO_FRAME);
6173 	else
6174 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM, IXGBE_MMW_SIZE_DEFAULT);
6175 
6176 	/* Set RTTBCNRC of queue X */
6177 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
6178 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
6179 	IXGBE_WRITE_FLUSH(hw);
6180 
6181 	return 0;
6182 }
6183 
6184 static int
6185 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
6186 		     __rte_unused uint32_t index,
6187 		     __rte_unused uint32_t pool)
6188 {
6189 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6190 	int diag;
6191 
6192 	/*
6193 	 * On a 82599 VF, adding again the same MAC addr is not an idempotent
6194 	 * operation. Trap this case to avoid exhausting the [very limited]
6195 	 * set of PF resources used to store VF MAC addresses.
6196 	 */
6197 	if (memcmp(hw->mac.perm_addr, mac_addr,
6198 			sizeof(struct rte_ether_addr)) == 0)
6199 		return -1;
6200 	diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6201 	if (diag != 0)
6202 		PMD_DRV_LOG(ERR, "Unable to add MAC address "
6203 			    RTE_ETHER_ADDR_PRT_FMT " - diag=%d",
6204 			    RTE_ETHER_ADDR_BYTES(mac_addr), diag);
6205 	return diag;
6206 }
6207 
6208 static void
6209 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
6210 {
6211 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6212 	struct rte_ether_addr *perm_addr =
6213 		(struct rte_ether_addr *)hw->mac.perm_addr;
6214 	struct rte_ether_addr *mac_addr;
6215 	uint32_t i;
6216 	int diag;
6217 
6218 	/*
6219 	 * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
6220 	 * not support the deletion of a given MAC address.
6221 	 * Instead, it imposes to delete all MAC addresses, then to add again
6222 	 * all MAC addresses with the exception of the one to be deleted.
6223 	 */
6224 	(void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
6225 
6226 	/*
6227 	 * Add again all MAC addresses, with the exception of the deleted one
6228 	 * and of the permanent MAC address.
6229 	 */
6230 	for (i = 0, mac_addr = dev->data->mac_addrs;
6231 	     i < hw->mac.num_rar_entries; i++, mac_addr++) {
6232 		/* Skip the deleted MAC address */
6233 		if (i == index)
6234 			continue;
6235 		/* Skip NULL MAC addresses */
6236 		if (rte_is_zero_ether_addr(mac_addr))
6237 			continue;
6238 		/* Skip the permanent MAC address */
6239 		if (memcmp(perm_addr, mac_addr,
6240 				sizeof(struct rte_ether_addr)) == 0)
6241 			continue;
6242 		diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6243 		if (diag != 0)
6244 			PMD_DRV_LOG(ERR,
6245 				    "Adding again MAC address "
6246 				     RTE_ETHER_ADDR_PRT_FMT " failed "
6247 				    "diag=%d", RTE_ETHER_ADDR_BYTES(mac_addr),
6248 				    diag);
6249 	}
6250 }
6251 
6252 static int
6253 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
6254 			struct rte_ether_addr *addr)
6255 {
6256 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6257 
6258 	hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
6259 
6260 	return 0;
6261 }
6262 
6263 int
6264 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
6265 			struct rte_eth_syn_filter *filter,
6266 			bool add)
6267 {
6268 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6269 	struct ixgbe_filter_info *filter_info =
6270 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6271 	uint32_t syn_info;
6272 	uint32_t synqf;
6273 
6274 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6275 		return -EINVAL;
6276 
6277 	syn_info = filter_info->syn_info;
6278 
6279 	if (add) {
6280 		if (syn_info & IXGBE_SYN_FILTER_ENABLE)
6281 			return -EINVAL;
6282 		synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
6283 			IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
6284 
6285 		if (filter->hig_pri)
6286 			synqf |= IXGBE_SYN_FILTER_SYNQFP;
6287 		else
6288 			synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
6289 	} else {
6290 		synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6291 		if (!(syn_info & IXGBE_SYN_FILTER_ENABLE))
6292 			return -ENOENT;
6293 		synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
6294 	}
6295 
6296 	filter_info->syn_info = synqf;
6297 	IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
6298 	IXGBE_WRITE_FLUSH(hw);
6299 	return 0;
6300 }
6301 
6302 
6303 static inline enum ixgbe_5tuple_protocol
6304 convert_protocol_type(uint8_t protocol_value)
6305 {
6306 	if (protocol_value == IPPROTO_TCP)
6307 		return IXGBE_FILTER_PROTOCOL_TCP;
6308 	else if (protocol_value == IPPROTO_UDP)
6309 		return IXGBE_FILTER_PROTOCOL_UDP;
6310 	else if (protocol_value == IPPROTO_SCTP)
6311 		return IXGBE_FILTER_PROTOCOL_SCTP;
6312 	else
6313 		return IXGBE_FILTER_PROTOCOL_NONE;
6314 }
6315 
6316 /* inject a 5-tuple filter to HW */
6317 static inline void
6318 ixgbe_inject_5tuple_filter(struct rte_eth_dev *dev,
6319 			   struct ixgbe_5tuple_filter *filter)
6320 {
6321 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6322 	int i;
6323 	uint32_t ftqf, sdpqf;
6324 	uint32_t l34timir = 0;
6325 	uint8_t mask = 0xff;
6326 
6327 	i = filter->index;
6328 
6329 	sdpqf = (uint32_t)(filter->filter_info.dst_port <<
6330 				IXGBE_SDPQF_DSTPORT_SHIFT);
6331 	sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
6332 
6333 	ftqf = (uint32_t)(filter->filter_info.proto &
6334 		IXGBE_FTQF_PROTOCOL_MASK);
6335 	ftqf |= (uint32_t)((filter->filter_info.priority &
6336 		IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
6337 	if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
6338 		mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
6339 	if (filter->filter_info.dst_ip_mask == 0)
6340 		mask &= IXGBE_FTQF_DEST_ADDR_MASK;
6341 	if (filter->filter_info.src_port_mask == 0)
6342 		mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
6343 	if (filter->filter_info.dst_port_mask == 0)
6344 		mask &= IXGBE_FTQF_DEST_PORT_MASK;
6345 	if (filter->filter_info.proto_mask == 0)
6346 		mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
6347 	ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
6348 	ftqf |= IXGBE_FTQF_POOL_MASK_EN;
6349 	ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
6350 
6351 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
6352 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
6353 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
6354 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
6355 
6356 	l34timir |= IXGBE_L34T_IMIR_RESERVE;
6357 	l34timir |= (uint32_t)(filter->queue <<
6358 				IXGBE_L34T_IMIR_QUEUE_SHIFT);
6359 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
6360 }
6361 
6362 /*
6363  * add a 5tuple filter
6364  *
6365  * @param
6366  * dev: Pointer to struct rte_eth_dev.
6367  * index: the index the filter allocates.
6368  * filter: pointer to the filter that will be added.
6369  * rx_queue: the queue id the filter assigned to.
6370  *
6371  * @return
6372  *    - On success, zero.
6373  *    - On failure, a negative value.
6374  */
6375 static int
6376 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
6377 			struct ixgbe_5tuple_filter *filter)
6378 {
6379 	struct ixgbe_filter_info *filter_info =
6380 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6381 	int i, idx, shift;
6382 
6383 	/*
6384 	 * look for an unused 5tuple filter index,
6385 	 * and insert the filter to list.
6386 	 */
6387 	for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
6388 		idx = i / (sizeof(uint32_t) * NBBY);
6389 		shift = i % (sizeof(uint32_t) * NBBY);
6390 		if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
6391 			filter_info->fivetuple_mask[idx] |= 1 << shift;
6392 			filter->index = i;
6393 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
6394 					  filter,
6395 					  entries);
6396 			break;
6397 		}
6398 	}
6399 	if (i >= IXGBE_MAX_FTQF_FILTERS) {
6400 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
6401 		return -ENOSYS;
6402 	}
6403 
6404 	ixgbe_inject_5tuple_filter(dev, filter);
6405 
6406 	return 0;
6407 }
6408 
6409 /*
6410  * remove a 5tuple filter
6411  *
6412  * @param
6413  * dev: Pointer to struct rte_eth_dev.
6414  * filter: the pointer of the filter will be removed.
6415  */
6416 static void
6417 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
6418 			struct ixgbe_5tuple_filter *filter)
6419 {
6420 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6421 	struct ixgbe_filter_info *filter_info =
6422 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6423 	uint16_t index = filter->index;
6424 
6425 	filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
6426 				~(1 << (index % (sizeof(uint32_t) * NBBY)));
6427 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
6428 	rte_free(filter);
6429 
6430 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
6431 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
6432 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
6433 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
6434 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
6435 }
6436 
6437 static int
6438 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
6439 {
6440 	struct ixgbe_hw *hw;
6441 	uint32_t max_frame = mtu + IXGBE_ETH_OVERHEAD;
6442 	struct rte_eth_dev_data *dev_data = dev->data;
6443 
6444 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6445 
6446 	if (mtu < RTE_ETHER_MIN_MTU || max_frame > RTE_ETHER_MAX_JUMBO_FRAME_LEN)
6447 		return -EINVAL;
6448 
6449 	/* If device is started, refuse mtu that requires the support of
6450 	 * scattered packets when this feature has not been enabled before.
6451 	 */
6452 	if (dev_data->dev_started && !dev_data->scattered_rx &&
6453 	    (max_frame + 2 * RTE_VLAN_HLEN >
6454 			dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
6455 		PMD_INIT_LOG(ERR, "Stop port first.");
6456 		return -EINVAL;
6457 	}
6458 
6459 	/*
6460 	 * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
6461 	 * request of the version 2.0 of the mailbox API.
6462 	 * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
6463 	 * of the mailbox API.
6464 	 * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
6465 	 * prior to 3.11.33 which contains the following change:
6466 	 * "ixgbe: Enable jumbo frames support w/ SR-IOV"
6467 	 */
6468 	if (ixgbevf_rlpml_set_vf(hw, max_frame))
6469 		return -EINVAL;
6470 
6471 	return 0;
6472 }
6473 
6474 static inline struct ixgbe_5tuple_filter *
6475 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
6476 			struct ixgbe_5tuple_filter_info *key)
6477 {
6478 	struct ixgbe_5tuple_filter *it;
6479 
6480 	TAILQ_FOREACH(it, filter_list, entries) {
6481 		if (memcmp(key, &it->filter_info,
6482 			sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
6483 			return it;
6484 		}
6485 	}
6486 	return NULL;
6487 }
6488 
6489 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
6490 static inline int
6491 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
6492 			struct ixgbe_5tuple_filter_info *filter_info)
6493 {
6494 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
6495 		filter->priority > IXGBE_5TUPLE_MAX_PRI ||
6496 		filter->priority < IXGBE_5TUPLE_MIN_PRI)
6497 		return -EINVAL;
6498 
6499 	switch (filter->dst_ip_mask) {
6500 	case UINT32_MAX:
6501 		filter_info->dst_ip_mask = 0;
6502 		filter_info->dst_ip = filter->dst_ip;
6503 		break;
6504 	case 0:
6505 		filter_info->dst_ip_mask = 1;
6506 		break;
6507 	default:
6508 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
6509 		return -EINVAL;
6510 	}
6511 
6512 	switch (filter->src_ip_mask) {
6513 	case UINT32_MAX:
6514 		filter_info->src_ip_mask = 0;
6515 		filter_info->src_ip = filter->src_ip;
6516 		break;
6517 	case 0:
6518 		filter_info->src_ip_mask = 1;
6519 		break;
6520 	default:
6521 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
6522 		return -EINVAL;
6523 	}
6524 
6525 	switch (filter->dst_port_mask) {
6526 	case UINT16_MAX:
6527 		filter_info->dst_port_mask = 0;
6528 		filter_info->dst_port = filter->dst_port;
6529 		break;
6530 	case 0:
6531 		filter_info->dst_port_mask = 1;
6532 		break;
6533 	default:
6534 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
6535 		return -EINVAL;
6536 	}
6537 
6538 	switch (filter->src_port_mask) {
6539 	case UINT16_MAX:
6540 		filter_info->src_port_mask = 0;
6541 		filter_info->src_port = filter->src_port;
6542 		break;
6543 	case 0:
6544 		filter_info->src_port_mask = 1;
6545 		break;
6546 	default:
6547 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
6548 		return -EINVAL;
6549 	}
6550 
6551 	switch (filter->proto_mask) {
6552 	case UINT8_MAX:
6553 		filter_info->proto_mask = 0;
6554 		filter_info->proto =
6555 			convert_protocol_type(filter->proto);
6556 		break;
6557 	case 0:
6558 		filter_info->proto_mask = 1;
6559 		break;
6560 	default:
6561 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
6562 		return -EINVAL;
6563 	}
6564 
6565 	filter_info->priority = (uint8_t)filter->priority;
6566 	return 0;
6567 }
6568 
6569 /*
6570  * add or delete a ntuple filter
6571  *
6572  * @param
6573  * dev: Pointer to struct rte_eth_dev.
6574  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6575  * add: if true, add filter, if false, remove filter
6576  *
6577  * @return
6578  *    - On success, zero.
6579  *    - On failure, a negative value.
6580  */
6581 int
6582 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
6583 			struct rte_eth_ntuple_filter *ntuple_filter,
6584 			bool add)
6585 {
6586 	struct ixgbe_filter_info *filter_info =
6587 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6588 	struct ixgbe_5tuple_filter_info filter_5tuple;
6589 	struct ixgbe_5tuple_filter *filter;
6590 	int ret;
6591 
6592 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6593 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6594 		return -EINVAL;
6595 	}
6596 
6597 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6598 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6599 	if (ret < 0)
6600 		return ret;
6601 
6602 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6603 					 &filter_5tuple);
6604 	if (filter != NULL && add) {
6605 		PMD_DRV_LOG(ERR, "filter exists.");
6606 		return -EEXIST;
6607 	}
6608 	if (filter == NULL && !add) {
6609 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
6610 		return -ENOENT;
6611 	}
6612 
6613 	if (add) {
6614 		filter = rte_zmalloc("ixgbe_5tuple_filter",
6615 				sizeof(struct ixgbe_5tuple_filter), 0);
6616 		if (filter == NULL)
6617 			return -ENOMEM;
6618 		rte_memcpy(&filter->filter_info,
6619 				 &filter_5tuple,
6620 				 sizeof(struct ixgbe_5tuple_filter_info));
6621 		filter->queue = ntuple_filter->queue;
6622 		ret = ixgbe_add_5tuple_filter(dev, filter);
6623 		if (ret < 0) {
6624 			rte_free(filter);
6625 			return ret;
6626 		}
6627 	} else
6628 		ixgbe_remove_5tuple_filter(dev, filter);
6629 
6630 	return 0;
6631 }
6632 
6633 int
6634 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
6635 			struct rte_eth_ethertype_filter *filter,
6636 			bool add)
6637 {
6638 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6639 	struct ixgbe_filter_info *filter_info =
6640 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6641 	uint32_t etqf = 0;
6642 	uint32_t etqs = 0;
6643 	int ret;
6644 	struct ixgbe_ethertype_filter ethertype_filter;
6645 
6646 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6647 		return -EINVAL;
6648 
6649 	if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
6650 		filter->ether_type == RTE_ETHER_TYPE_IPV6) {
6651 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
6652 			" ethertype filter.", filter->ether_type);
6653 		return -EINVAL;
6654 	}
6655 
6656 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
6657 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
6658 		return -EINVAL;
6659 	}
6660 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
6661 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
6662 		return -EINVAL;
6663 	}
6664 
6665 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6666 	if (ret >= 0 && add) {
6667 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
6668 			    filter->ether_type);
6669 		return -EEXIST;
6670 	}
6671 	if (ret < 0 && !add) {
6672 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6673 			    filter->ether_type);
6674 		return -ENOENT;
6675 	}
6676 
6677 	if (add) {
6678 		etqf = IXGBE_ETQF_FILTER_EN;
6679 		etqf |= (uint32_t)filter->ether_type;
6680 		etqs |= (uint32_t)((filter->queue <<
6681 				    IXGBE_ETQS_RX_QUEUE_SHIFT) &
6682 				    IXGBE_ETQS_RX_QUEUE);
6683 		etqs |= IXGBE_ETQS_QUEUE_EN;
6684 
6685 		ethertype_filter.ethertype = filter->ether_type;
6686 		ethertype_filter.etqf = etqf;
6687 		ethertype_filter.etqs = etqs;
6688 		ethertype_filter.conf = FALSE;
6689 		ret = ixgbe_ethertype_filter_insert(filter_info,
6690 						    &ethertype_filter);
6691 		if (ret < 0) {
6692 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
6693 			return -ENOSPC;
6694 		}
6695 	} else {
6696 		ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
6697 		if (ret < 0)
6698 			return -ENOSYS;
6699 	}
6700 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
6701 	IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
6702 	IXGBE_WRITE_FLUSH(hw);
6703 
6704 	return 0;
6705 }
6706 
6707 static int
6708 ixgbe_dev_flow_ops_get(__rte_unused struct rte_eth_dev *dev,
6709 		       const struct rte_flow_ops **ops)
6710 {
6711 	*ops = &ixgbe_flow_ops;
6712 	return 0;
6713 }
6714 
6715 static u8 *
6716 ixgbe_dev_addr_list_itr(__rte_unused struct ixgbe_hw *hw,
6717 			u8 **mc_addr_ptr, u32 *vmdq)
6718 {
6719 	u8 *mc_addr;
6720 
6721 	*vmdq = 0;
6722 	mc_addr = *mc_addr_ptr;
6723 	*mc_addr_ptr = (mc_addr + sizeof(struct rte_ether_addr));
6724 	return mc_addr;
6725 }
6726 
6727 static int
6728 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
6729 			  struct rte_ether_addr *mc_addr_set,
6730 			  uint32_t nb_mc_addr)
6731 {
6732 	struct ixgbe_hw *hw;
6733 	u8 *mc_addr_list;
6734 
6735 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6736 	mc_addr_list = (u8 *)mc_addr_set;
6737 	return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
6738 					 ixgbe_dev_addr_list_itr, TRUE);
6739 }
6740 
6741 static uint64_t
6742 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
6743 {
6744 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6745 	uint64_t systime_cycles;
6746 
6747 	switch (hw->mac.type) {
6748 	case ixgbe_mac_X550:
6749 	case ixgbe_mac_X550EM_x:
6750 	case ixgbe_mac_X550EM_a:
6751 	case ixgbe_mac_E610:
6752 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
6753 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6754 		systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6755 				* NSEC_PER_SEC;
6756 		break;
6757 	default:
6758 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6759 		systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6760 				<< 32;
6761 	}
6762 
6763 	return systime_cycles;
6764 }
6765 
6766 static uint64_t
6767 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6768 {
6769 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6770 	uint64_t rx_tstamp_cycles;
6771 
6772 	switch (hw->mac.type) {
6773 	case ixgbe_mac_X550:
6774 	case ixgbe_mac_X550EM_x:
6775 	case ixgbe_mac_X550EM_a:
6776 	case ixgbe_mac_E610:
6777 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6778 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6779 		rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6780 				* NSEC_PER_SEC;
6781 		break;
6782 	default:
6783 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6784 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6785 		rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6786 				<< 32;
6787 	}
6788 
6789 	return rx_tstamp_cycles;
6790 }
6791 
6792 static uint64_t
6793 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6794 {
6795 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6796 	uint64_t tx_tstamp_cycles;
6797 
6798 	switch (hw->mac.type) {
6799 	case ixgbe_mac_X550:
6800 	case ixgbe_mac_X550EM_x:
6801 	case ixgbe_mac_X550EM_a:
6802 	case ixgbe_mac_E610:
6803 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6804 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6805 		tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6806 				* NSEC_PER_SEC;
6807 		break;
6808 	default:
6809 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6810 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6811 		tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6812 				<< 32;
6813 	}
6814 
6815 	return tx_tstamp_cycles;
6816 }
6817 
6818 static void
6819 ixgbe_start_timecounters(struct rte_eth_dev *dev)
6820 {
6821 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6822 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6823 	struct rte_eth_link link;
6824 	uint32_t incval = 0;
6825 	uint32_t shift = 0;
6826 
6827 	/* Get current link speed. */
6828 	ixgbe_dev_link_update(dev, 1);
6829 	rte_eth_linkstatus_get(dev, &link);
6830 
6831 	switch (link.link_speed) {
6832 	case RTE_ETH_SPEED_NUM_100M:
6833 		incval = IXGBE_INCVAL_100;
6834 		shift = IXGBE_INCVAL_SHIFT_100;
6835 		break;
6836 	case RTE_ETH_SPEED_NUM_1G:
6837 		incval = IXGBE_INCVAL_1GB;
6838 		shift = IXGBE_INCVAL_SHIFT_1GB;
6839 		break;
6840 	case RTE_ETH_SPEED_NUM_10G:
6841 	default:
6842 		incval = IXGBE_INCVAL_10GB;
6843 		shift = IXGBE_INCVAL_SHIFT_10GB;
6844 		break;
6845 	}
6846 
6847 	switch (hw->mac.type) {
6848 	case ixgbe_mac_X550:
6849 	case ixgbe_mac_X550EM_x:
6850 	case ixgbe_mac_X550EM_a:
6851 	case ixgbe_mac_E610:
6852 		/* Independent of link speed. */
6853 		incval = 1;
6854 		/* Cycles read will be interpreted as ns. */
6855 		shift = 0;
6856 		/* Fall-through */
6857 	case ixgbe_mac_X540:
6858 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
6859 		break;
6860 	case ixgbe_mac_82599EB:
6861 		incval >>= IXGBE_INCVAL_SHIFT_82599;
6862 		shift -= IXGBE_INCVAL_SHIFT_82599;
6863 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
6864 				(1 << IXGBE_INCPER_SHIFT_82599) | incval);
6865 		break;
6866 	default:
6867 		/* Not supported. */
6868 		return;
6869 	}
6870 
6871 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
6872 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6873 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6874 
6875 	adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6876 	adapter->systime_tc.cc_shift = shift;
6877 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
6878 
6879 	adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6880 	adapter->rx_tstamp_tc.cc_shift = shift;
6881 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6882 
6883 	adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6884 	adapter->tx_tstamp_tc.cc_shift = shift;
6885 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6886 }
6887 
6888 static int
6889 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
6890 {
6891 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6892 
6893 	adapter->systime_tc.nsec += delta;
6894 	adapter->rx_tstamp_tc.nsec += delta;
6895 	adapter->tx_tstamp_tc.nsec += delta;
6896 
6897 	return 0;
6898 }
6899 
6900 static int
6901 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
6902 {
6903 	uint64_t ns;
6904 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6905 
6906 	ns = rte_timespec_to_ns(ts);
6907 	/* Set the timecounters to a new value. */
6908 	adapter->systime_tc.nsec = ns;
6909 	adapter->rx_tstamp_tc.nsec = ns;
6910 	adapter->tx_tstamp_tc.nsec = ns;
6911 
6912 	return 0;
6913 }
6914 
6915 static int
6916 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
6917 {
6918 	uint64_t ns, systime_cycles;
6919 	struct ixgbe_adapter *adapter = dev->data->dev_private;
6920 
6921 	systime_cycles = ixgbe_read_systime_cyclecounter(dev);
6922 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
6923 	*ts = rte_ns_to_timespec(ns);
6924 
6925 	return 0;
6926 }
6927 
6928 static int
6929 ixgbe_timesync_enable(struct rte_eth_dev *dev)
6930 {
6931 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6932 	uint32_t tsync_ctl;
6933 	uint32_t tsauxc;
6934 	struct timespec ts;
6935 
6936 	memset(&ts, 0, sizeof(struct timespec));
6937 
6938 	/* get current system time */
6939 	clock_gettime(CLOCK_REALTIME, &ts);
6940 
6941 	/* Stop the timesync system time. */
6942 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
6943 	/* Reset the timesync system time value. */
6944 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
6945 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
6946 
6947 	/* Enable system time for platforms where it isn't on by default. */
6948 	tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
6949 	tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
6950 	IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
6951 
6952 	ixgbe_start_timecounters(dev);
6953 
6954 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6955 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
6956 			(RTE_ETHER_TYPE_1588 |
6957 			 IXGBE_ETQF_FILTER_EN |
6958 			 IXGBE_ETQF_1588));
6959 
6960 	/* Enable timestamping of received PTP packets. */
6961 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6962 	tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
6963 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6964 
6965 	/* Enable timestamping of transmitted PTP packets. */
6966 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6967 	tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
6968 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6969 
6970 	IXGBE_WRITE_FLUSH(hw);
6971 
6972 	/* ixgbe uses zero-based timestamping so only adjust timecounter */
6973 	ixgbe_timesync_write_time(dev, &ts);
6974 
6975 	return 0;
6976 }
6977 
6978 static int
6979 ixgbe_timesync_disable(struct rte_eth_dev *dev)
6980 {
6981 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6982 	uint32_t tsync_ctl;
6983 
6984 	/* Disable timestamping of transmitted PTP packets. */
6985 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6986 	tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
6987 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6988 
6989 	/* Disable timestamping of received PTP packets. */
6990 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6991 	tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
6992 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6993 
6994 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6995 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
6996 
6997 	/* Stop incrementing the System Time registers. */
6998 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
6999 
7000 	return 0;
7001 }
7002 
7003 static int
7004 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
7005 				 struct timespec *timestamp,
7006 				 uint32_t flags __rte_unused)
7007 {
7008 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7009 	struct ixgbe_adapter *adapter = dev->data->dev_private;
7010 	uint32_t tsync_rxctl;
7011 	uint64_t rx_tstamp_cycles;
7012 	uint64_t ns;
7013 
7014 	tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7015 	if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
7016 		return -EINVAL;
7017 
7018 	rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
7019 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
7020 	*timestamp = rte_ns_to_timespec(ns);
7021 
7022 	return  0;
7023 }
7024 
7025 static int
7026 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
7027 				 struct timespec *timestamp)
7028 {
7029 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7030 	struct ixgbe_adapter *adapter = dev->data->dev_private;
7031 	uint32_t tsync_txctl;
7032 	uint64_t tx_tstamp_cycles;
7033 	uint64_t ns;
7034 
7035 	tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7036 	if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
7037 		return -EINVAL;
7038 
7039 	tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
7040 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
7041 	*timestamp = rte_ns_to_timespec(ns);
7042 
7043 	return 0;
7044 }
7045 
7046 static int
7047 ixgbe_get_reg_length(struct rte_eth_dev *dev)
7048 {
7049 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7050 	int count = 0;
7051 	int g_ind = 0;
7052 	const struct reg_info *reg_group;
7053 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7054 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7055 
7056 	while ((reg_group = reg_set[g_ind++]))
7057 		count += ixgbe_regs_group_count(reg_group);
7058 
7059 	return count;
7060 }
7061 
7062 static int
7063 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
7064 {
7065 	int count = 0;
7066 	int g_ind = 0;
7067 	const struct reg_info *reg_group;
7068 
7069 	while ((reg_group = ixgbevf_regs[g_ind++]))
7070 		count += ixgbe_regs_group_count(reg_group);
7071 
7072 	return count;
7073 }
7074 
7075 static int
7076 ixgbe_get_regs(struct rte_eth_dev *dev,
7077 	      struct rte_dev_reg_info *regs)
7078 {
7079 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7080 	uint32_t *data = regs->data;
7081 	int g_ind = 0;
7082 	int count = 0;
7083 	const struct reg_info *reg_group;
7084 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7085 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7086 
7087 	if (data == NULL) {
7088 		regs->length = ixgbe_get_reg_length(dev);
7089 		regs->width = sizeof(uint32_t);
7090 		return 0;
7091 	}
7092 
7093 	/* Support only full register dump */
7094 	if ((regs->length == 0) ||
7095 	    (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
7096 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7097 			hw->device_id;
7098 		while ((reg_group = reg_set[g_ind++]))
7099 			count += ixgbe_read_regs_group(dev, &data[count],
7100 				reg_group);
7101 		return 0;
7102 	}
7103 
7104 	return -ENOTSUP;
7105 }
7106 
7107 static int
7108 ixgbevf_get_regs(struct rte_eth_dev *dev,
7109 		struct rte_dev_reg_info *regs)
7110 {
7111 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7112 	uint32_t *data = regs->data;
7113 	int g_ind = 0;
7114 	int count = 0;
7115 	const struct reg_info *reg_group;
7116 
7117 	if (data == NULL) {
7118 		regs->length = ixgbevf_get_reg_length(dev);
7119 		regs->width = sizeof(uint32_t);
7120 		return 0;
7121 	}
7122 
7123 	/* Support only full register dump */
7124 	if ((regs->length == 0) ||
7125 	    (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
7126 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7127 			hw->device_id;
7128 		while ((reg_group = ixgbevf_regs[g_ind++]))
7129 			count += ixgbe_read_regs_group(dev, &data[count],
7130 						      reg_group);
7131 		return 0;
7132 	}
7133 
7134 	return -ENOTSUP;
7135 }
7136 
7137 static int
7138 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
7139 {
7140 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7141 
7142 	/* Return unit is byte count */
7143 	return hw->eeprom.word_size * 2;
7144 }
7145 
7146 static int
7147 ixgbe_get_eeprom(struct rte_eth_dev *dev,
7148 		struct rte_dev_eeprom_info *in_eeprom)
7149 {
7150 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7151 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7152 	uint16_t *data = in_eeprom->data;
7153 	int first, length;
7154 
7155 	first = in_eeprom->offset >> 1;
7156 	length = in_eeprom->length >> 1;
7157 	if ((first > hw->eeprom.word_size) ||
7158 	    ((first + length) > hw->eeprom.word_size))
7159 		return -EINVAL;
7160 
7161 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7162 
7163 	return eeprom->ops.read_buffer(hw, first, length, data);
7164 }
7165 
7166 static int
7167 ixgbe_set_eeprom(struct rte_eth_dev *dev,
7168 		struct rte_dev_eeprom_info *in_eeprom)
7169 {
7170 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7171 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7172 	uint16_t *data = in_eeprom->data;
7173 	int first, length;
7174 
7175 	first = in_eeprom->offset >> 1;
7176 	length = in_eeprom->length >> 1;
7177 	if ((first > hw->eeprom.word_size) ||
7178 	    ((first + length) > hw->eeprom.word_size))
7179 		return -EINVAL;
7180 
7181 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7182 
7183 	return eeprom->ops.write_buffer(hw,  first, length, data);
7184 }
7185 
7186 static int
7187 ixgbe_get_module_info(struct rte_eth_dev *dev,
7188 		      struct rte_eth_dev_module_info *modinfo)
7189 {
7190 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7191 	uint32_t status;
7192 	uint8_t sff8472_rev, addr_mode;
7193 	bool page_swap = false;
7194 
7195 	/* Check whether we support SFF-8472 or not */
7196 	status = hw->phy.ops.read_i2c_eeprom(hw,
7197 					     IXGBE_SFF_SFF_8472_COMP,
7198 					     &sff8472_rev);
7199 	if (status != 0)
7200 		return -EIO;
7201 
7202 	/* addressing mode is not supported */
7203 	status = hw->phy.ops.read_i2c_eeprom(hw,
7204 					     IXGBE_SFF_SFF_8472_SWAP,
7205 					     &addr_mode);
7206 	if (status != 0)
7207 		return -EIO;
7208 
7209 	if (addr_mode & IXGBE_SFF_ADDRESSING_MODE) {
7210 		PMD_DRV_LOG(ERR,
7211 			    "Address change required to access page 0xA2, "
7212 			    "but not supported. Please report the module "
7213 			    "type to the driver maintainers.");
7214 		page_swap = true;
7215 	}
7216 
7217 	if (sff8472_rev == IXGBE_SFF_SFF_8472_UNSUP || page_swap) {
7218 		/* We have a SFP, but it does not support SFF-8472 */
7219 		modinfo->type = RTE_ETH_MODULE_SFF_8079;
7220 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
7221 	} else {
7222 		/* We have a SFP which supports a revision of SFF-8472. */
7223 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
7224 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
7225 	}
7226 
7227 	return 0;
7228 }
7229 
7230 static int
7231 ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
7232 			struct rte_dev_eeprom_info *info)
7233 {
7234 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7235 	uint32_t status = IXGBE_ERR_PHY_ADDR_INVALID;
7236 	uint8_t databyte = 0xFF;
7237 	uint8_t *data = info->data;
7238 	uint32_t i = 0;
7239 
7240 	for (i = info->offset; i < info->offset + info->length; i++) {
7241 		if (i < RTE_ETH_MODULE_SFF_8079_LEN)
7242 			status = hw->phy.ops.read_i2c_eeprom(hw, i, &databyte);
7243 		else
7244 			status = hw->phy.ops.read_i2c_sff8472(hw, i, &databyte);
7245 
7246 		if (status != 0)
7247 			return -EIO;
7248 
7249 		data[i - info->offset] = databyte;
7250 	}
7251 
7252 	return 0;
7253 }
7254 
7255 uint16_t
7256 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
7257 	switch (mac_type) {
7258 	case ixgbe_mac_X550:
7259 	case ixgbe_mac_X550EM_x:
7260 	case ixgbe_mac_X550EM_a:
7261 	case ixgbe_mac_E610:
7262 		return RTE_ETH_RSS_RETA_SIZE_512;
7263 	case ixgbe_mac_X550_vf:
7264 	case ixgbe_mac_X550EM_x_vf:
7265 	case ixgbe_mac_X550EM_a_vf:
7266 	case ixgbe_mac_E610_vf:
7267 		return RTE_ETH_RSS_RETA_SIZE_64;
7268 	case ixgbe_mac_X540_vf:
7269 	case ixgbe_mac_82599_vf:
7270 		return 0;
7271 	default:
7272 		return RTE_ETH_RSS_RETA_SIZE_128;
7273 	}
7274 }
7275 
7276 uint32_t
7277 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
7278 	switch (mac_type) {
7279 	case ixgbe_mac_X550:
7280 	case ixgbe_mac_X550EM_x:
7281 	case ixgbe_mac_X550EM_a:
7282 	case ixgbe_mac_E610:
7283 		if (reta_idx < RTE_ETH_RSS_RETA_SIZE_128)
7284 			return IXGBE_RETA(reta_idx >> 2);
7285 		else
7286 			return IXGBE_ERETA((reta_idx - RTE_ETH_RSS_RETA_SIZE_128) >> 2);
7287 	case ixgbe_mac_X550_vf:
7288 	case ixgbe_mac_X550EM_x_vf:
7289 	case ixgbe_mac_X550EM_a_vf:
7290 	case ixgbe_mac_E610_vf:
7291 		return IXGBE_VFRETA(reta_idx >> 2);
7292 	default:
7293 		return IXGBE_RETA(reta_idx >> 2);
7294 	}
7295 }
7296 
7297 uint32_t
7298 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
7299 	switch (mac_type) {
7300 	case ixgbe_mac_X550_vf:
7301 	case ixgbe_mac_X550EM_x_vf:
7302 	case ixgbe_mac_X550EM_a_vf:
7303 		return IXGBE_VFMRQC;
7304 	default:
7305 		return IXGBE_MRQC;
7306 	}
7307 }
7308 
7309 uint32_t
7310 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
7311 	switch (mac_type) {
7312 	case ixgbe_mac_X550_vf:
7313 	case ixgbe_mac_X550EM_x_vf:
7314 	case ixgbe_mac_X550EM_a_vf:
7315 		return IXGBE_VFRSSRK(i);
7316 	default:
7317 		return IXGBE_RSSRK(i);
7318 	}
7319 }
7320 
7321 bool
7322 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
7323 	switch (mac_type) {
7324 	case ixgbe_mac_82599_vf:
7325 	case ixgbe_mac_X540_vf:
7326 		return 0;
7327 	default:
7328 		return 1;
7329 	}
7330 }
7331 
7332 static int
7333 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
7334 			struct rte_eth_dcb_info *dcb_info)
7335 {
7336 	struct ixgbe_dcb_config *dcb_config =
7337 			IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
7338 	struct ixgbe_dcb_tc_config *tc;
7339 	struct rte_eth_dcb_tc_queue_mapping *tc_queue;
7340 	uint8_t nb_tcs;
7341 	uint8_t i, j;
7342 
7343 	if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_DCB_FLAG)
7344 		dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
7345 	else
7346 		dcb_info->nb_tcs = 1;
7347 
7348 	tc_queue = &dcb_info->tc_queue;
7349 	nb_tcs = dcb_info->nb_tcs;
7350 
7351 	if (dcb_config->vt_mode) { /* vt is enabled*/
7352 		struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
7353 				&dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
7354 		for (i = 0; i < RTE_ETH_DCB_NUM_USER_PRIORITIES; i++)
7355 			dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
7356 		if (RTE_ETH_DEV_SRIOV(dev).active > 0) {
7357 			for (j = 0; j < nb_tcs; j++) {
7358 				tc_queue->tc_rxq[0][j].base = j;
7359 				tc_queue->tc_rxq[0][j].nb_queue = 1;
7360 				tc_queue->tc_txq[0][j].base = j;
7361 				tc_queue->tc_txq[0][j].nb_queue = 1;
7362 			}
7363 		} else {
7364 			for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
7365 				for (j = 0; j < nb_tcs; j++) {
7366 					tc_queue->tc_rxq[i][j].base =
7367 						i * nb_tcs + j;
7368 					tc_queue->tc_rxq[i][j].nb_queue = 1;
7369 					tc_queue->tc_txq[i][j].base =
7370 						i * nb_tcs + j;
7371 					tc_queue->tc_txq[i][j].nb_queue = 1;
7372 				}
7373 			}
7374 		}
7375 	} else { /* vt is disabled*/
7376 		struct rte_eth_dcb_rx_conf *rx_conf =
7377 				&dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
7378 		for (i = 0; i < RTE_ETH_DCB_NUM_USER_PRIORITIES; i++)
7379 			dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
7380 		if (dcb_info->nb_tcs == RTE_ETH_4_TCS) {
7381 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7382 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
7383 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7384 			}
7385 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7386 			dcb_info->tc_queue.tc_txq[0][1].base = 64;
7387 			dcb_info->tc_queue.tc_txq[0][2].base = 96;
7388 			dcb_info->tc_queue.tc_txq[0][3].base = 112;
7389 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
7390 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7391 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7392 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7393 		} else if (dcb_info->nb_tcs == RTE_ETH_8_TCS) {
7394 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7395 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
7396 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7397 			}
7398 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7399 			dcb_info->tc_queue.tc_txq[0][1].base = 32;
7400 			dcb_info->tc_queue.tc_txq[0][2].base = 64;
7401 			dcb_info->tc_queue.tc_txq[0][3].base = 80;
7402 			dcb_info->tc_queue.tc_txq[0][4].base = 96;
7403 			dcb_info->tc_queue.tc_txq[0][5].base = 104;
7404 			dcb_info->tc_queue.tc_txq[0][6].base = 112;
7405 			dcb_info->tc_queue.tc_txq[0][7].base = 120;
7406 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
7407 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7408 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7409 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7410 			dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
7411 			dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
7412 			dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
7413 			dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
7414 		}
7415 	}
7416 	for (i = 0; i < dcb_info->nb_tcs; i++) {
7417 		tc = &dcb_config->tc_config[i];
7418 		dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
7419 	}
7420 	return 0;
7421 }
7422 
7423 /* Update e-tag ether type */
7424 static int
7425 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
7426 			    uint16_t ether_type)
7427 {
7428 	uint32_t etag_etype;
7429 
7430 	if (hw->mac.type != ixgbe_mac_X550 &&
7431 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7432 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7433 		return -ENOTSUP;
7434 	}
7435 
7436 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7437 	etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
7438 	etag_etype |= ether_type;
7439 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7440 	IXGBE_WRITE_FLUSH(hw);
7441 
7442 	return 0;
7443 }
7444 
7445 /* Enable e-tag tunnel */
7446 static int
7447 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
7448 {
7449 	uint32_t etag_etype;
7450 
7451 	if (hw->mac.type != ixgbe_mac_X550 &&
7452 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7453 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7454 		return -ENOTSUP;
7455 	}
7456 
7457 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7458 	etag_etype |= IXGBE_ETAG_ETYPE_VALID;
7459 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7460 	IXGBE_WRITE_FLUSH(hw);
7461 
7462 	return 0;
7463 }
7464 
7465 static int
7466 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
7467 		       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7468 {
7469 	int ret = 0;
7470 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7471 	uint32_t i, rar_entries;
7472 	uint32_t rar_low, rar_high;
7473 
7474 	if (hw->mac.type != ixgbe_mac_X550 &&
7475 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7476 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7477 		return -ENOTSUP;
7478 	}
7479 
7480 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7481 
7482 	for (i = 1; i < rar_entries; i++) {
7483 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7484 		rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
7485 		if ((rar_high & IXGBE_RAH_AV) &&
7486 		    (rar_high & IXGBE_RAH_ADTYPE) &&
7487 		    ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
7488 		     l2_tunnel->tunnel_id)) {
7489 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
7490 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
7491 
7492 			ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
7493 
7494 			return ret;
7495 		}
7496 	}
7497 
7498 	return ret;
7499 }
7500 
7501 static int
7502 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
7503 		       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7504 {
7505 	int ret = 0;
7506 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7507 	uint32_t i, rar_entries;
7508 	uint32_t rar_low, rar_high;
7509 
7510 	if (hw->mac.type != ixgbe_mac_X550 &&
7511 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7512 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7513 		return -ENOTSUP;
7514 	}
7515 
7516 	/* One entry for one tunnel. Try to remove potential existing entry. */
7517 	ixgbe_e_tag_filter_del(dev, l2_tunnel);
7518 
7519 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7520 
7521 	for (i = 1; i < rar_entries; i++) {
7522 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7523 		if (rar_high & IXGBE_RAH_AV) {
7524 			continue;
7525 		} else {
7526 			ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
7527 			rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
7528 			rar_low = l2_tunnel->tunnel_id;
7529 
7530 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
7531 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
7532 
7533 			return ret;
7534 		}
7535 	}
7536 
7537 	PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
7538 		     " Please remove a rule before adding a new one.");
7539 	return -EINVAL;
7540 }
7541 
7542 static inline struct ixgbe_l2_tn_filter *
7543 ixgbe_l2_tn_filter_lookup(struct ixgbe_l2_tn_info *l2_tn_info,
7544 			  struct ixgbe_l2_tn_key *key)
7545 {
7546 	int ret;
7547 
7548 	ret = rte_hash_lookup(l2_tn_info->hash_handle, (const void *)key);
7549 	if (ret < 0)
7550 		return NULL;
7551 
7552 	return l2_tn_info->hash_map[ret];
7553 }
7554 
7555 static inline int
7556 ixgbe_insert_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7557 			  struct ixgbe_l2_tn_filter *l2_tn_filter)
7558 {
7559 	int ret;
7560 
7561 	ret = rte_hash_add_key(l2_tn_info->hash_handle,
7562 			       &l2_tn_filter->key);
7563 
7564 	if (ret < 0) {
7565 		PMD_DRV_LOG(ERR,
7566 			    "Failed to insert L2 tunnel filter"
7567 			    " to hash table %d!",
7568 			    ret);
7569 		return ret;
7570 	}
7571 
7572 	l2_tn_info->hash_map[ret] = l2_tn_filter;
7573 
7574 	TAILQ_INSERT_TAIL(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7575 
7576 	return 0;
7577 }
7578 
7579 static inline int
7580 ixgbe_remove_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7581 			  struct ixgbe_l2_tn_key *key)
7582 {
7583 	int ret;
7584 	struct ixgbe_l2_tn_filter *l2_tn_filter;
7585 
7586 	ret = rte_hash_del_key(l2_tn_info->hash_handle, key);
7587 
7588 	if (ret < 0) {
7589 		PMD_DRV_LOG(ERR,
7590 			    "No such L2 tunnel filter to delete %d!",
7591 			    ret);
7592 		return ret;
7593 	}
7594 
7595 	l2_tn_filter = l2_tn_info->hash_map[ret];
7596 	l2_tn_info->hash_map[ret] = NULL;
7597 
7598 	TAILQ_REMOVE(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7599 	rte_free(l2_tn_filter);
7600 
7601 	return 0;
7602 }
7603 
7604 /* Add l2 tunnel filter */
7605 int
7606 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
7607 			       struct ixgbe_l2_tunnel_conf *l2_tunnel,
7608 			       bool restore)
7609 {
7610 	int ret;
7611 	struct ixgbe_l2_tn_info *l2_tn_info =
7612 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7613 	struct ixgbe_l2_tn_key key;
7614 	struct ixgbe_l2_tn_filter *node;
7615 
7616 	if (!restore) {
7617 		key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7618 		key.tn_id = l2_tunnel->tunnel_id;
7619 
7620 		node = ixgbe_l2_tn_filter_lookup(l2_tn_info, &key);
7621 
7622 		if (node) {
7623 			PMD_DRV_LOG(ERR,
7624 				    "The L2 tunnel filter already exists!");
7625 			return -EINVAL;
7626 		}
7627 
7628 		node = rte_zmalloc("ixgbe_l2_tn",
7629 				   sizeof(struct ixgbe_l2_tn_filter),
7630 				   0);
7631 		if (!node)
7632 			return -ENOMEM;
7633 
7634 		rte_memcpy(&node->key,
7635 				 &key,
7636 				 sizeof(struct ixgbe_l2_tn_key));
7637 		node->pool = l2_tunnel->pool;
7638 		ret = ixgbe_insert_l2_tn_filter(l2_tn_info, node);
7639 		if (ret < 0) {
7640 			rte_free(node);
7641 			return ret;
7642 		}
7643 	}
7644 
7645 	switch (l2_tunnel->l2_tunnel_type) {
7646 	case RTE_ETH_L2_TUNNEL_TYPE_E_TAG:
7647 		ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
7648 		break;
7649 	default:
7650 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7651 		ret = -EINVAL;
7652 		break;
7653 	}
7654 
7655 	if ((!restore) && (ret < 0))
7656 		(void)ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7657 
7658 	return ret;
7659 }
7660 
7661 /* Delete l2 tunnel filter */
7662 int
7663 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
7664 			       struct ixgbe_l2_tunnel_conf *l2_tunnel)
7665 {
7666 	int ret;
7667 	struct ixgbe_l2_tn_info *l2_tn_info =
7668 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7669 	struct ixgbe_l2_tn_key key;
7670 
7671 	key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7672 	key.tn_id = l2_tunnel->tunnel_id;
7673 	ret = ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7674 	if (ret < 0)
7675 		return ret;
7676 
7677 	switch (l2_tunnel->l2_tunnel_type) {
7678 	case RTE_ETH_L2_TUNNEL_TYPE_E_TAG:
7679 		ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
7680 		break;
7681 	default:
7682 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7683 		ret = -EINVAL;
7684 		break;
7685 	}
7686 
7687 	return ret;
7688 }
7689 
7690 static int
7691 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
7692 {
7693 	int ret = 0;
7694 	uint32_t ctrl;
7695 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7696 
7697 	if (hw->mac.type != ixgbe_mac_X550 &&
7698 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7699 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7700 		return -ENOTSUP;
7701 	}
7702 
7703 	ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
7704 	ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
7705 	if (en)
7706 		ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
7707 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
7708 
7709 	return ret;
7710 }
7711 
7712 static int
7713 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
7714 			uint16_t port)
7715 {
7716 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
7717 	IXGBE_WRITE_FLUSH(hw);
7718 
7719 	return 0;
7720 }
7721 
7722 /* There's only one register for VxLAN UDP port.
7723  * So, we cannot add several ports. Will update it.
7724  */
7725 static int
7726 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
7727 		     uint16_t port)
7728 {
7729 	if (port == 0) {
7730 		PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
7731 		return -EINVAL;
7732 	}
7733 
7734 	return ixgbe_update_vxlan_port(hw, port);
7735 }
7736 
7737 /* We cannot delete the VxLAN port. For there's a register for VxLAN
7738  * UDP port, it must have a value.
7739  * So, will reset it to the original value 0.
7740  */
7741 static int
7742 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
7743 		     uint16_t port)
7744 {
7745 	uint16_t cur_port;
7746 
7747 	cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
7748 
7749 	if (cur_port != port) {
7750 		PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
7751 		return -EINVAL;
7752 	}
7753 
7754 	return ixgbe_update_vxlan_port(hw, 0);
7755 }
7756 
7757 /* Add UDP tunneling port */
7758 static int
7759 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
7760 			      struct rte_eth_udp_tunnel *udp_tunnel)
7761 {
7762 	int ret = 0;
7763 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7764 
7765 	if (hw->mac.type != ixgbe_mac_X550 &&
7766 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7767 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7768 		return -ENOTSUP;
7769 	}
7770 
7771 	if (udp_tunnel == NULL)
7772 		return -EINVAL;
7773 
7774 	switch (udp_tunnel->prot_type) {
7775 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
7776 		ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
7777 		break;
7778 
7779 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
7780 	case RTE_ETH_TUNNEL_TYPE_TEREDO:
7781 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7782 		ret = -EINVAL;
7783 		break;
7784 
7785 	default:
7786 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7787 		ret = -EINVAL;
7788 		break;
7789 	}
7790 
7791 	return ret;
7792 }
7793 
7794 /* Remove UDP tunneling port */
7795 static int
7796 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
7797 			      struct rte_eth_udp_tunnel *udp_tunnel)
7798 {
7799 	int ret = 0;
7800 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7801 
7802 	if (hw->mac.type != ixgbe_mac_X550 &&
7803 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7804 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7805 		return -ENOTSUP;
7806 	}
7807 
7808 	if (udp_tunnel == NULL)
7809 		return -EINVAL;
7810 
7811 	switch (udp_tunnel->prot_type) {
7812 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
7813 		ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
7814 		break;
7815 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
7816 	case RTE_ETH_TUNNEL_TYPE_TEREDO:
7817 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7818 		ret = -EINVAL;
7819 		break;
7820 	default:
7821 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7822 		ret = -EINVAL;
7823 		break;
7824 	}
7825 
7826 	return ret;
7827 }
7828 
7829 static int
7830 ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev)
7831 {
7832 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7833 	int ret;
7834 
7835 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_PROMISC)) {
7836 	case IXGBE_SUCCESS:
7837 		ret = 0;
7838 		break;
7839 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7840 		ret = -ENOTSUP;
7841 		break;
7842 	default:
7843 		ret = -EAGAIN;
7844 		break;
7845 	}
7846 
7847 	return ret;
7848 }
7849 
7850 static int
7851 ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev)
7852 {
7853 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7854 	int mode = IXGBEVF_XCAST_MODE_NONE;
7855 	int ret;
7856 
7857 	if (dev->data->all_multicast)
7858 		mode = IXGBEVF_XCAST_MODE_ALLMULTI;
7859 
7860 	switch (hw->mac.ops.update_xcast_mode(hw, mode)) {
7861 	case IXGBE_SUCCESS:
7862 		ret = 0;
7863 		break;
7864 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7865 		ret = -ENOTSUP;
7866 		break;
7867 	default:
7868 		ret = -EAGAIN;
7869 		break;
7870 	}
7871 
7872 	return ret;
7873 }
7874 
7875 static int
7876 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
7877 {
7878 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7879 	int ret;
7880 	int mode = IXGBEVF_XCAST_MODE_ALLMULTI;
7881 
7882 	if (dev->data->promiscuous)
7883 		return 0;
7884 
7885 	switch (hw->mac.ops.update_xcast_mode(hw, mode)) {
7886 	case IXGBE_SUCCESS:
7887 		ret = 0;
7888 		break;
7889 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7890 		ret = -ENOTSUP;
7891 		break;
7892 	default:
7893 		ret = -EAGAIN;
7894 		break;
7895 	}
7896 
7897 	return ret;
7898 }
7899 
7900 static int
7901 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
7902 {
7903 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7904 	int ret;
7905 
7906 	if (dev->data->promiscuous)
7907 		return 0;
7908 
7909 	switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_MULTI)) {
7910 	case IXGBE_SUCCESS:
7911 		ret = 0;
7912 		break;
7913 	case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7914 		ret = -ENOTSUP;
7915 		break;
7916 	default:
7917 		ret = -EAGAIN;
7918 		break;
7919 	}
7920 
7921 	return ret;
7922 }
7923 
7924 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
7925 {
7926 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7927 	u32 in_msg = 0;
7928 
7929 	/* peek the message first */
7930 	in_msg = IXGBE_READ_REG(hw, IXGBE_VFMBMEM);
7931 
7932 	/* PF reset VF event */
7933 	if (in_msg == IXGBE_PF_CONTROL_MSG) {
7934 		/* dummy mbx read to ack pf */
7935 		if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
7936 			return;
7937 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
7938 					     NULL);
7939 	}
7940 }
7941 
7942 static int
7943 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
7944 {
7945 	uint32_t eicr;
7946 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7947 	struct ixgbe_interrupt *intr =
7948 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7949 	ixgbevf_intr_disable(dev);
7950 
7951 	/* read-on-clear nic registers here */
7952 	eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
7953 	intr->flags = 0;
7954 
7955 	/* only one misc vector supported - mailbox */
7956 	eicr &= IXGBE_VTEICR_MASK;
7957 	if (eicr == IXGBE_MISC_VEC_ID)
7958 		intr->flags |= IXGBE_FLAG_MAILBOX;
7959 
7960 	return 0;
7961 }
7962 
7963 static int
7964 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
7965 {
7966 	struct ixgbe_interrupt *intr =
7967 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7968 
7969 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
7970 		ixgbevf_mbx_process(dev);
7971 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
7972 	}
7973 
7974 	ixgbevf_intr_enable(dev);
7975 
7976 	return 0;
7977 }
7978 
7979 static void
7980 ixgbevf_dev_interrupt_handler(void *param)
7981 {
7982 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
7983 
7984 	ixgbevf_dev_interrupt_get_status(dev);
7985 	ixgbevf_dev_interrupt_action(dev);
7986 }
7987 
7988 /**
7989  *  ixgbe_disable_sec_tx_path_generic - Stops the transmit data path
7990  *  @hw: pointer to hardware structure
7991  *
7992  *  Stops the transmit data path and waits for the HW to internally empty
7993  *  the Tx security block
7994  **/
7995 int ixgbe_disable_sec_tx_path_generic(struct ixgbe_hw *hw)
7996 {
7997 #define IXGBE_MAX_SECTX_POLL 40
7998 
7999 	int i;
8000 	int sectxreg;
8001 
8002 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8003 	sectxreg |= IXGBE_SECTXCTRL_TX_DIS;
8004 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8005 	for (i = 0; i < IXGBE_MAX_SECTX_POLL; i++) {
8006 		sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXSTAT);
8007 		if (sectxreg & IXGBE_SECTXSTAT_SECTX_RDY)
8008 			break;
8009 		/* Use interrupt-safe sleep just in case */
8010 		usec_delay(1000);
8011 	}
8012 
8013 	/* For informational purposes only */
8014 	if (i >= IXGBE_MAX_SECTX_POLL)
8015 		PMD_DRV_LOG(DEBUG, "Tx unit being enabled before security "
8016 			 "path fully disabled.  Continuing with init.");
8017 
8018 	return IXGBE_SUCCESS;
8019 }
8020 
8021 /**
8022  *  ixgbe_enable_sec_tx_path_generic - Enables the transmit data path
8023  *  @hw: pointer to hardware structure
8024  *
8025  *  Enables the transmit data path.
8026  **/
8027 int ixgbe_enable_sec_tx_path_generic(struct ixgbe_hw *hw)
8028 {
8029 	uint32_t sectxreg;
8030 
8031 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8032 	sectxreg &= ~IXGBE_SECTXCTRL_TX_DIS;
8033 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8034 	IXGBE_WRITE_FLUSH(hw);
8035 
8036 	return IXGBE_SUCCESS;
8037 }
8038 
8039 /* restore n-tuple filter */
8040 static inline void
8041 ixgbe_ntuple_filter_restore(struct rte_eth_dev *dev)
8042 {
8043 	struct ixgbe_filter_info *filter_info =
8044 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8045 	struct ixgbe_5tuple_filter *node;
8046 
8047 	TAILQ_FOREACH(node, &filter_info->fivetuple_list, entries) {
8048 		ixgbe_inject_5tuple_filter(dev, node);
8049 	}
8050 }
8051 
8052 /* restore ethernet type filter */
8053 static inline void
8054 ixgbe_ethertype_filter_restore(struct rte_eth_dev *dev)
8055 {
8056 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8057 	struct ixgbe_filter_info *filter_info =
8058 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8059 	int i;
8060 
8061 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8062 		if (filter_info->ethertype_mask & (1 << i)) {
8063 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i),
8064 					filter_info->ethertype_filters[i].etqf);
8065 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i),
8066 					filter_info->ethertype_filters[i].etqs);
8067 			IXGBE_WRITE_FLUSH(hw);
8068 		}
8069 	}
8070 }
8071 
8072 /* restore SYN filter */
8073 static inline void
8074 ixgbe_syn_filter_restore(struct rte_eth_dev *dev)
8075 {
8076 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8077 	struct ixgbe_filter_info *filter_info =
8078 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8079 	uint32_t synqf;
8080 
8081 	synqf = filter_info->syn_info;
8082 
8083 	if (synqf & IXGBE_SYN_FILTER_ENABLE) {
8084 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
8085 		IXGBE_WRITE_FLUSH(hw);
8086 	}
8087 }
8088 
8089 /* restore L2 tunnel filter */
8090 static inline void
8091 ixgbe_l2_tn_filter_restore(struct rte_eth_dev *dev)
8092 {
8093 	struct ixgbe_l2_tn_info *l2_tn_info =
8094 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8095 	struct ixgbe_l2_tn_filter *node;
8096 	struct ixgbe_l2_tunnel_conf l2_tn_conf;
8097 
8098 	TAILQ_FOREACH(node, &l2_tn_info->l2_tn_list, entries) {
8099 		l2_tn_conf.l2_tunnel_type = node->key.l2_tn_type;
8100 		l2_tn_conf.tunnel_id      = node->key.tn_id;
8101 		l2_tn_conf.pool           = node->pool;
8102 		(void)ixgbe_dev_l2_tunnel_filter_add(dev, &l2_tn_conf, TRUE);
8103 	}
8104 }
8105 
8106 /* restore rss filter */
8107 static inline void
8108 ixgbe_rss_filter_restore(struct rte_eth_dev *dev)
8109 {
8110 	struct ixgbe_filter_info *filter_info =
8111 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8112 
8113 	if (filter_info->rss_info.conf.queue_num)
8114 		ixgbe_config_rss_filter(dev,
8115 			&filter_info->rss_info, TRUE);
8116 }
8117 
8118 static int
8119 ixgbe_filter_restore(struct rte_eth_dev *dev)
8120 {
8121 	ixgbe_ntuple_filter_restore(dev);
8122 	ixgbe_ethertype_filter_restore(dev);
8123 	ixgbe_syn_filter_restore(dev);
8124 	ixgbe_fdir_filter_restore(dev);
8125 	ixgbe_l2_tn_filter_restore(dev);
8126 	ixgbe_rss_filter_restore(dev);
8127 
8128 	return 0;
8129 }
8130 
8131 static void
8132 ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev)
8133 {
8134 	struct ixgbe_l2_tn_info *l2_tn_info =
8135 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8136 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8137 
8138 	if (l2_tn_info->e_tag_en)
8139 		(void)ixgbe_e_tag_enable(hw);
8140 
8141 	if (l2_tn_info->e_tag_fwd_en)
8142 		(void)ixgbe_e_tag_forwarding_en_dis(dev, 1);
8143 
8144 	(void)ixgbe_update_e_tag_eth_type(hw, l2_tn_info->e_tag_ether_type);
8145 }
8146 
8147 /* remove all the n-tuple filters */
8148 void
8149 ixgbe_clear_all_ntuple_filter(struct rte_eth_dev *dev)
8150 {
8151 	struct ixgbe_filter_info *filter_info =
8152 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8153 	struct ixgbe_5tuple_filter *p_5tuple;
8154 
8155 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
8156 		ixgbe_remove_5tuple_filter(dev, p_5tuple);
8157 }
8158 
8159 /* remove all the ether type filters */
8160 void
8161 ixgbe_clear_all_ethertype_filter(struct rte_eth_dev *dev)
8162 {
8163 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8164 	struct ixgbe_filter_info *filter_info =
8165 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8166 	int i;
8167 
8168 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8169 		if (filter_info->ethertype_mask & (1 << i) &&
8170 		    !filter_info->ethertype_filters[i].conf) {
8171 			(void)ixgbe_ethertype_filter_remove(filter_info,
8172 							    (uint8_t)i);
8173 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i), 0);
8174 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i), 0);
8175 			IXGBE_WRITE_FLUSH(hw);
8176 		}
8177 	}
8178 }
8179 
8180 /* remove the SYN filter */
8181 void
8182 ixgbe_clear_syn_filter(struct rte_eth_dev *dev)
8183 {
8184 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8185 	struct ixgbe_filter_info *filter_info =
8186 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8187 
8188 	if (filter_info->syn_info & IXGBE_SYN_FILTER_ENABLE) {
8189 		filter_info->syn_info = 0;
8190 
8191 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, 0);
8192 		IXGBE_WRITE_FLUSH(hw);
8193 	}
8194 }
8195 
8196 /* remove all the L2 tunnel filters */
8197 int
8198 ixgbe_clear_all_l2_tn_filter(struct rte_eth_dev *dev)
8199 {
8200 	struct ixgbe_l2_tn_info *l2_tn_info =
8201 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8202 	struct ixgbe_l2_tn_filter *l2_tn_filter;
8203 	struct ixgbe_l2_tunnel_conf l2_tn_conf;
8204 	int ret = 0;
8205 
8206 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
8207 		l2_tn_conf.l2_tunnel_type = l2_tn_filter->key.l2_tn_type;
8208 		l2_tn_conf.tunnel_id      = l2_tn_filter->key.tn_id;
8209 		l2_tn_conf.pool           = l2_tn_filter->pool;
8210 		ret = ixgbe_dev_l2_tunnel_filter_del(dev, &l2_tn_conf);
8211 		if (ret < 0)
8212 			return ret;
8213 	}
8214 
8215 	return 0;
8216 }
8217 
8218 void
8219 ixgbe_dev_macsec_setting_save(struct rte_eth_dev *dev,
8220 				struct ixgbe_macsec_setting *macsec_setting)
8221 {
8222 	struct ixgbe_macsec_setting *macsec =
8223 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8224 
8225 	macsec->offload_en = macsec_setting->offload_en;
8226 	macsec->encrypt_en = macsec_setting->encrypt_en;
8227 	macsec->replayprotect_en = macsec_setting->replayprotect_en;
8228 }
8229 
8230 void
8231 ixgbe_dev_macsec_setting_reset(struct rte_eth_dev *dev)
8232 {
8233 	struct ixgbe_macsec_setting *macsec =
8234 		IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8235 
8236 	macsec->offload_en = 0;
8237 	macsec->encrypt_en = 0;
8238 	macsec->replayprotect_en = 0;
8239 }
8240 
8241 void
8242 ixgbe_dev_macsec_register_enable(struct rte_eth_dev *dev,
8243 				struct ixgbe_macsec_setting *macsec_setting)
8244 {
8245 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8246 	uint32_t ctrl;
8247 	uint8_t en = macsec_setting->encrypt_en;
8248 	uint8_t rp = macsec_setting->replayprotect_en;
8249 
8250 	/**
8251 	 * Workaround:
8252 	 * As no ixgbe_disable_sec_rx_path equivalent is
8253 	 * implemented for tx in the base code, and we are
8254 	 * not allowed to modify the base code in DPDK, so
8255 	 * just call the hand-written one directly for now.
8256 	 * The hardware support has been checked by
8257 	 * ixgbe_disable_sec_rx_path().
8258 	 */
8259 	ixgbe_disable_sec_tx_path_generic(hw);
8260 
8261 	/* Enable Ethernet CRC (required by MACsec offload) */
8262 	ctrl = IXGBE_READ_REG(hw, IXGBE_HLREG0);
8263 	ctrl |= IXGBE_HLREG0_TXCRCEN | IXGBE_HLREG0_RXCRCSTRP;
8264 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, ctrl);
8265 
8266 	/* Enable the TX and RX crypto engines */
8267 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8268 	ctrl &= ~IXGBE_SECTXCTRL_SECTX_DIS;
8269 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8270 
8271 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8272 	ctrl &= ~IXGBE_SECRXCTRL_SECRX_DIS;
8273 	IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8274 
8275 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
8276 	ctrl &= ~IXGBE_SECTX_MINSECIFG_MASK;
8277 	ctrl |= 0x3;
8278 	IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, ctrl);
8279 
8280 	/* Enable SA lookup */
8281 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8282 	ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8283 	ctrl |= en ? IXGBE_LSECTXCTRL_AUTH_ENCRYPT :
8284 		     IXGBE_LSECTXCTRL_AUTH;
8285 	ctrl |= IXGBE_LSECTXCTRL_AISCI;
8286 	ctrl &= ~IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8287 	ctrl |= IXGBE_MACSEC_PNTHRSH & IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8288 	IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8289 
8290 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8291 	ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8292 	ctrl |= IXGBE_LSECRXCTRL_STRICT << IXGBE_LSECRXCTRL_EN_SHIFT;
8293 	ctrl &= ~IXGBE_LSECRXCTRL_PLSH;
8294 	if (rp)
8295 		ctrl |= IXGBE_LSECRXCTRL_RP;
8296 	else
8297 		ctrl &= ~IXGBE_LSECRXCTRL_RP;
8298 	IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8299 
8300 	/* Start the data paths */
8301 	ixgbe_enable_sec_rx_path(hw);
8302 	/**
8303 	 * Workaround:
8304 	 * As no ixgbe_enable_sec_rx_path equivalent is
8305 	 * implemented for tx in the base code, and we are
8306 	 * not allowed to modify the base code in DPDK, so
8307 	 * just call the hand-written one directly for now.
8308 	 */
8309 	ixgbe_enable_sec_tx_path_generic(hw);
8310 }
8311 
8312 void
8313 ixgbe_dev_macsec_register_disable(struct rte_eth_dev *dev)
8314 {
8315 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8316 	uint32_t ctrl;
8317 
8318 	/**
8319 	 * Workaround:
8320 	 * As no ixgbe_disable_sec_rx_path equivalent is
8321 	 * implemented for tx in the base code, and we are
8322 	 * not allowed to modify the base code in DPDK, so
8323 	 * just call the hand-written one directly for now.
8324 	 * The hardware support has been checked by
8325 	 * ixgbe_disable_sec_rx_path().
8326 	 */
8327 	ixgbe_disable_sec_tx_path_generic(hw);
8328 
8329 	/* Disable the TX and RX crypto engines */
8330 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8331 	ctrl |= IXGBE_SECTXCTRL_SECTX_DIS;
8332 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8333 
8334 	ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8335 	ctrl |= IXGBE_SECRXCTRL_SECRX_DIS;
8336 	IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8337 
8338 	/* Disable SA lookup */
8339 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8340 	ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8341 	ctrl |= IXGBE_LSECTXCTRL_DISABLE;
8342 	IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8343 
8344 	ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8345 	ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8346 	ctrl |= IXGBE_LSECRXCTRL_DISABLE << IXGBE_LSECRXCTRL_EN_SHIFT;
8347 	IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8348 
8349 	/* Start the data paths */
8350 	ixgbe_enable_sec_rx_path(hw);
8351 	/**
8352 	 * Workaround:
8353 	 * As no ixgbe_enable_sec_rx_path equivalent is
8354 	 * implemented for tx in the base code, and we are
8355 	 * not allowed to modify the base code in DPDK, so
8356 	 * just call the hand-written one directly for now.
8357 	 */
8358 	ixgbe_enable_sec_tx_path_generic(hw);
8359 }
8360 
8361 RTE_PMD_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd);
8362 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
8363 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe, "* igb_uio | uio_pci_generic | vfio-pci");
8364 RTE_PMD_REGISTER_PARAM_STRING(net_ixgbe,
8365 			      IXGBE_DEVARG_FIBER_SDP3_NOT_TX_DISABLE "=<0|1>");
8366 RTE_PMD_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd);
8367 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);
8368 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe_vf, "* igb_uio | vfio-pci");
8369 RTE_PMD_REGISTER_PARAM_STRING(net_ixgbe_vf,
8370 			      IXGBEVF_DEVARG_PFLINK_FULLCHK "=<0|1>");
8371 
8372 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_init, init, NOTICE);
8373 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_driver, driver, NOTICE);
8374 
8375 #ifdef RTE_ETHDEV_DEBUG_RX
8376 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_rx, rx, DEBUG);
8377 #endif
8378 #ifdef RTE_ETHDEV_DEBUG_TX
8379 RTE_LOG_REGISTER_SUFFIX(ixgbe_logtype_tx, tx, DEBUG);
8380 #endif
8381