xref: /f-stack/dpdk/drivers/net/ixgbe/ixgbe_ethdev.c (revision e2d9b9ca)
1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright(c) 2010-2017 Intel Corporation. All rights reserved.
5  *   All rights reserved.
6  *
7  *   Redistribution and use in source and binary forms, with or without
8  *   modification, are permitted provided that the following conditions
9  *   are met:
10  *
11  *     * Redistributions of source code must retain the above copyright
12  *       notice, this list of conditions and the following disclaimer.
13  *     * Redistributions in binary form must reproduce the above copyright
14  *       notice, this list of conditions and the following disclaimer in
15  *       the documentation and/or other materials provided with the
16  *       distribution.
17  *     * Neither the name of Intel Corporation nor the names of its
18  *       contributors may be used to endorse or promote products derived
19  *       from this software without specific prior written permission.
20  *
21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32  */
33 
34 #include <sys/queue.h>
35 #include <stdio.h>
36 #include <errno.h>
37 #include <stdint.h>
38 #include <string.h>
39 #include <unistd.h>
40 #include <stdarg.h>
41 #include <inttypes.h>
42 #include <netinet/in.h>
43 #include <rte_byteorder.h>
44 #include <rte_common.h>
45 #include <rte_cycles.h>
46 
47 #include <rte_interrupts.h>
48 #include <rte_log.h>
49 #include <rte_debug.h>
50 #include <rte_pci.h>
51 #include <rte_bus_pci.h>
52 #include <rte_atomic.h>
53 #include <rte_branch_prediction.h>
54 #include <rte_memory.h>
55 #include <rte_eal.h>
56 #include <rte_alarm.h>
57 #include <rte_ether.h>
58 #include <rte_ethdev.h>
59 #include <rte_ethdev_pci.h>
60 #include <rte_malloc.h>
61 #include <rte_random.h>
62 #include <rte_dev.h>
63 #include <rte_hash_crc.h>
64 #ifdef RTE_LIBRTE_SECURITY
65 #include <rte_security_driver.h>
66 #endif
67 
68 #include "ixgbe_logs.h"
69 #include "base/ixgbe_api.h"
70 #include "base/ixgbe_vf.h"
71 #include "base/ixgbe_common.h"
72 #include "ixgbe_ethdev.h"
73 #include "ixgbe_bypass.h"
74 #include "ixgbe_rxtx.h"
75 #include "base/ixgbe_type.h"
76 #include "base/ixgbe_phy.h"
77 #include "ixgbe_regs.h"
78 
79 /*
80  * High threshold controlling when to start sending XOFF frames. Must be at
81  * least 8 bytes less than receive packet buffer size. This value is in units
82  * of 1024 bytes.
83  */
84 #define IXGBE_FC_HI    0x80
85 
86 /*
87  * Low threshold controlling when to start sending XON frames. This value is
88  * in units of 1024 bytes.
89  */
90 #define IXGBE_FC_LO    0x40
91 
92 /* Default minimum inter-interrupt interval for EITR configuration */
93 #define IXGBE_MIN_INTER_INTERRUPT_INTERVAL_DEFAULT    0x79E
94 
95 /* Timer value included in XOFF frames. */
96 #define IXGBE_FC_PAUSE 0x680
97 
98 /*Default value of Max Rx Queue*/
99 #define IXGBE_MAX_RX_QUEUE_NUM 128
100 
101 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
102 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
103 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
104 
105 #define IXGBE_MMW_SIZE_DEFAULT        0x4
106 #define IXGBE_MMW_SIZE_JUMBO_FRAME    0x14
107 #define IXGBE_MAX_RING_DESC           4096 /* replicate define from rxtx */
108 
109 /*
110  *  Default values for RX/TX configuration
111  */
112 #define IXGBE_DEFAULT_RX_FREE_THRESH  32
113 #define IXGBE_DEFAULT_RX_PTHRESH      8
114 #define IXGBE_DEFAULT_RX_HTHRESH      8
115 #define IXGBE_DEFAULT_RX_WTHRESH      0
116 
117 #define IXGBE_DEFAULT_TX_FREE_THRESH  32
118 #define IXGBE_DEFAULT_TX_PTHRESH      32
119 #define IXGBE_DEFAULT_TX_HTHRESH      0
120 #define IXGBE_DEFAULT_TX_WTHRESH      0
121 #define IXGBE_DEFAULT_TX_RSBIT_THRESH 32
122 
123 /* Bit shift and mask */
124 #define IXGBE_4_BIT_WIDTH  (CHAR_BIT / 2)
125 #define IXGBE_4_BIT_MASK   RTE_LEN2MASK(IXGBE_4_BIT_WIDTH, uint8_t)
126 #define IXGBE_8_BIT_WIDTH  CHAR_BIT
127 #define IXGBE_8_BIT_MASK   UINT8_MAX
128 
129 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
130 
131 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
132 
133 #define IXGBE_HKEY_MAX_INDEX 10
134 
135 /* Additional timesync values. */
136 #define NSEC_PER_SEC             1000000000L
137 #define IXGBE_INCVAL_10GB        0x66666666
138 #define IXGBE_INCVAL_1GB         0x40000000
139 #define IXGBE_INCVAL_100         0x50000000
140 #define IXGBE_INCVAL_SHIFT_10GB  28
141 #define IXGBE_INCVAL_SHIFT_1GB   24
142 #define IXGBE_INCVAL_SHIFT_100   21
143 #define IXGBE_INCVAL_SHIFT_82599 7
144 #define IXGBE_INCPER_SHIFT_82599 24
145 
146 #define IXGBE_CYCLECOUNTER_MASK   0xffffffffffffffffULL
147 
148 #define IXGBE_VT_CTL_POOLING_MODE_MASK         0x00030000
149 #define IXGBE_VT_CTL_POOLING_MODE_ETAG         0x00010000
150 #define DEFAULT_ETAG_ETYPE                     0x893f
151 #define IXGBE_ETAG_ETYPE                       0x00005084
152 #define IXGBE_ETAG_ETYPE_MASK                  0x0000ffff
153 #define IXGBE_ETAG_ETYPE_VALID                 0x80000000
154 #define IXGBE_RAH_ADTYPE                       0x40000000
155 #define IXGBE_RAL_ETAG_FILTER_MASK             0x00003fff
156 #define IXGBE_VMVIR_TAGA_MASK                  0x18000000
157 #define IXGBE_VMVIR_TAGA_ETAG_INSERT           0x08000000
158 #define IXGBE_VMTIR(_i) (0x00017000 + ((_i) * 4)) /* 64 of these (0-63) */
159 #define IXGBE_QDE_STRIP_TAG                    0x00000004
160 #define IXGBE_VTEICR_MASK                      0x07
161 
162 #define IXGBE_EXVET_VET_EXT_SHIFT              16
163 #define IXGBE_DMATXCTL_VT_MASK                 0xFFFF0000
164 
165 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev);
166 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
167 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev);
168 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev);
169 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev);
170 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev);
171 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev);
172 static int  ixgbe_dev_configure(struct rte_eth_dev *dev);
173 static int  ixgbe_dev_start(struct rte_eth_dev *dev);
174 static void ixgbe_dev_stop(struct rte_eth_dev *dev);
175 static int  ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
176 static int  ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
177 static void ixgbe_dev_close(struct rte_eth_dev *dev);
178 static int  ixgbe_dev_reset(struct rte_eth_dev *dev);
179 static void ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
180 static void ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
181 static void ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
182 static void ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
183 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
184 				int wait_to_complete);
185 static int ixgbe_dev_stats_get(struct rte_eth_dev *dev,
186 				struct rte_eth_stats *stats);
187 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
188 				struct rte_eth_xstat *xstats, unsigned n);
189 static int ixgbevf_dev_xstats_get(struct rte_eth_dev *dev,
190 				  struct rte_eth_xstat *xstats, unsigned n);
191 static int
192 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
193 		uint64_t *values, unsigned int n);
194 static void ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
195 static void ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
196 static int ixgbe_dev_xstats_get_names(struct rte_eth_dev *dev,
197 	struct rte_eth_xstat_name *xstats_names,
198 	unsigned int size);
199 static int ixgbevf_dev_xstats_get_names(struct rte_eth_dev *dev,
200 	struct rte_eth_xstat_name *xstats_names, unsigned limit);
201 static int ixgbe_dev_xstats_get_names_by_id(
202 	struct rte_eth_dev *dev,
203 	struct rte_eth_xstat_name *xstats_names,
204 	const uint64_t *ids,
205 	unsigned int limit);
206 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
207 					     uint16_t queue_id,
208 					     uint8_t stat_idx,
209 					     uint8_t is_rx);
210 static int ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
211 				 size_t fw_size);
212 static void ixgbe_dev_info_get(struct rte_eth_dev *dev,
213 			       struct rte_eth_dev_info *dev_info);
214 static const uint32_t *ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev);
215 static void ixgbevf_dev_info_get(struct rte_eth_dev *dev,
216 				 struct rte_eth_dev_info *dev_info);
217 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
218 
219 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
220 		uint16_t vlan_id, int on);
221 static int ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
222 			       enum rte_vlan_type vlan_type,
223 			       uint16_t tpid_id);
224 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
225 		uint16_t queue, bool on);
226 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
227 		int on);
228 static int ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
229 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
230 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
231 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
232 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
233 
234 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
235 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
236 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
237 			       struct rte_eth_fc_conf *fc_conf);
238 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
239 			       struct rte_eth_fc_conf *fc_conf);
240 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
241 		struct rte_eth_pfc_conf *pfc_conf);
242 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
243 			struct rte_eth_rss_reta_entry64 *reta_conf,
244 			uint16_t reta_size);
245 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
246 			struct rte_eth_rss_reta_entry64 *reta_conf,
247 			uint16_t reta_size);
248 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
249 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
250 static int ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev);
251 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
252 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
253 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev,
254 				      struct rte_intr_handle *handle);
255 static void ixgbe_dev_interrupt_handler(void *param);
256 static void ixgbe_dev_interrupt_delayed_handler(void *param);
257 static int ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
258 			 uint32_t index, uint32_t pool);
259 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
260 static void ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
261 					   struct ether_addr *mac_addr);
262 static void ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config);
263 static bool is_device_supported(struct rte_eth_dev *dev,
264 				struct rte_pci_driver *drv);
265 
266 /* For Virtual Function support */
267 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
268 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
269 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
270 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
271 static int ixgbevf_dev_link_update(struct rte_eth_dev *dev,
272 				   int wait_to_complete);
273 static void ixgbevf_dev_stop(struct rte_eth_dev *dev);
274 static void ixgbevf_dev_close(struct rte_eth_dev *dev);
275 static int  ixgbevf_dev_reset(struct rte_eth_dev *dev);
276 static void ixgbevf_intr_disable(struct ixgbe_hw *hw);
277 static void ixgbevf_intr_enable(struct ixgbe_hw *hw);
278 static int ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
279 		struct rte_eth_stats *stats);
280 static void ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
281 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
282 		uint16_t vlan_id, int on);
283 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
284 		uint16_t queue, int on);
285 static int ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
286 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
287 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
288 					    uint16_t queue_id);
289 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
290 					     uint16_t queue_id);
291 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
292 				 uint8_t queue, uint8_t msix_vector);
293 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
294 static void ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev);
295 static void ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev);
296 
297 /* For Eth VMDQ APIs support */
298 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
299 		ether_addr * mac_addr, uint8_t on);
300 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on);
301 static int ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
302 		struct rte_eth_mirror_conf *mirror_conf,
303 		uint8_t rule_id, uint8_t on);
304 static int ixgbe_mirror_rule_reset(struct rte_eth_dev *dev,
305 		uint8_t	rule_id);
306 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
307 					  uint16_t queue_id);
308 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
309 					   uint16_t queue_id);
310 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
311 			       uint8_t queue, uint8_t msix_vector);
312 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
313 
314 static int ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
315 				struct ether_addr *mac_addr,
316 				uint32_t index, uint32_t pool);
317 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
318 static void ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
319 					     struct ether_addr *mac_addr);
320 static int ixgbe_syn_filter_get(struct rte_eth_dev *dev,
321 			struct rte_eth_syn_filter *filter);
322 static int ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
323 			enum rte_filter_op filter_op,
324 			void *arg);
325 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
326 			struct ixgbe_5tuple_filter *filter);
327 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
328 			struct ixgbe_5tuple_filter *filter);
329 static int ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
330 				enum rte_filter_op filter_op,
331 				void *arg);
332 static int ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
333 			struct rte_eth_ntuple_filter *filter);
334 static int ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
335 				enum rte_filter_op filter_op,
336 				void *arg);
337 static int ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
338 			struct rte_eth_ethertype_filter *filter);
339 static int ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
340 		     enum rte_filter_type filter_type,
341 		     enum rte_filter_op filter_op,
342 		     void *arg);
343 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
344 
345 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
346 				      struct ether_addr *mc_addr_set,
347 				      uint32_t nb_mc_addr);
348 static int ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
349 				   struct rte_eth_dcb_info *dcb_info);
350 
351 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
352 static int ixgbe_get_regs(struct rte_eth_dev *dev,
353 			    struct rte_dev_reg_info *regs);
354 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
355 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
356 				struct rte_dev_eeprom_info *eeprom);
357 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
358 				struct rte_dev_eeprom_info *eeprom);
359 
360 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
361 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
362 				struct rte_dev_reg_info *regs);
363 
364 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
365 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
366 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
367 					    struct timespec *timestamp,
368 					    uint32_t flags);
369 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
370 					    struct timespec *timestamp);
371 static int ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
372 static int ixgbe_timesync_read_time(struct rte_eth_dev *dev,
373 				   struct timespec *timestamp);
374 static int ixgbe_timesync_write_time(struct rte_eth_dev *dev,
375 				   const struct timespec *timestamp);
376 static void ixgbevf_dev_interrupt_handler(void *param);
377 
378 static int ixgbe_dev_l2_tunnel_eth_type_conf
379 	(struct rte_eth_dev *dev, struct rte_eth_l2_tunnel_conf *l2_tunnel);
380 static int ixgbe_dev_l2_tunnel_offload_set
381 	(struct rte_eth_dev *dev,
382 	 struct rte_eth_l2_tunnel_conf *l2_tunnel,
383 	 uint32_t mask,
384 	 uint8_t en);
385 static int ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
386 					     enum rte_filter_op filter_op,
387 					     void *arg);
388 
389 static int ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
390 					 struct rte_eth_udp_tunnel *udp_tunnel);
391 static int ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
392 					 struct rte_eth_udp_tunnel *udp_tunnel);
393 static int ixgbe_filter_restore(struct rte_eth_dev *dev);
394 static void ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev);
395 
396 /*
397  * Define VF Stats MACRO for Non "cleared on read" register
398  */
399 #define UPDATE_VF_STAT(reg, last, cur)                          \
400 {                                                               \
401 	uint32_t latest = IXGBE_READ_REG(hw, reg);              \
402 	cur += (latest - last) & UINT_MAX;                      \
403 	last = latest;                                          \
404 }
405 
406 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
407 {                                                                \
408 	u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
409 	u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
410 	u64 latest = ((new_msb << 32) | new_lsb);                \
411 	cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
412 	last = latest;                                           \
413 }
414 
415 #define IXGBE_SET_HWSTRIP(h, q) do {\
416 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
417 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
418 		(h)->bitmap[idx] |= 1 << bit;\
419 	} while (0)
420 
421 #define IXGBE_CLEAR_HWSTRIP(h, q) do {\
422 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
423 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
424 		(h)->bitmap[idx] &= ~(1 << bit);\
425 	} while (0)
426 
427 #define IXGBE_GET_HWSTRIP(h, q, r) do {\
428 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
429 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
430 		(r) = (h)->bitmap[idx] >> bit & 1;\
431 	} while (0)
432 
433 /*
434  * The set of PCI devices this driver supports
435  */
436 static const struct rte_pci_id pci_id_ixgbe_map[] = {
437 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598) },
438 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_BX) },
439 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT) },
440 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT) },
441 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT) },
442 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2) },
443 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM) },
444 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4) },
445 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT) },
446 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT) },
447 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM) },
448 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR) },
449 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4) },
450 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ) },
451 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KR) },
452 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE) },
453 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4) },
454 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP) },
455 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE) },
456 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE) },
457 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_EM) },
458 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2) },
459 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP) },
460 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP) },
461 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP) },
462 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM) },
463 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM) },
464 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_LS) },
465 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T) },
466 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1) },
467 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP) },
468 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T) },
469 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_1G_T) },
470 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T) },
471 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1) },
472 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR) },
473 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR_L) },
474 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP_N) },
475 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII) },
476 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII_L) },
477 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_10G_T) },
478 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP) },
479 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP_N) },
480 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP) },
481 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T) },
482 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T_L) },
483 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4) },
484 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR) },
485 #ifdef RTE_LIBRTE_IXGBE_BYPASS
486 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BYPASS) },
487 #endif
488 	{ .vendor_id = 0, /* sentinel */ },
489 };
490 
491 /*
492  * The set of PCI devices this driver supports (for 82599 VF)
493  */
494 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
495 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF) },
496 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF_HV) },
497 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF) },
498 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF_HV) },
499 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF_HV) },
500 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF) },
501 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF) },
502 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF_HV) },
503 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF) },
504 	{ RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF_HV) },
505 	{ .vendor_id = 0, /* sentinel */ },
506 };
507 
508 static const struct rte_eth_desc_lim rx_desc_lim = {
509 	.nb_max = IXGBE_MAX_RING_DESC,
510 	.nb_min = IXGBE_MIN_RING_DESC,
511 	.nb_align = IXGBE_RXD_ALIGN,
512 };
513 
514 static const struct rte_eth_desc_lim tx_desc_lim = {
515 	.nb_max = IXGBE_MAX_RING_DESC,
516 	.nb_min = IXGBE_MIN_RING_DESC,
517 	.nb_align = IXGBE_TXD_ALIGN,
518 	.nb_seg_max = IXGBE_TX_MAX_SEG,
519 	.nb_mtu_seg_max = IXGBE_TX_MAX_SEG,
520 };
521 
522 static const struct eth_dev_ops ixgbe_eth_dev_ops = {
523 	.dev_configure        = ixgbe_dev_configure,
524 	.dev_start            = ixgbe_dev_start,
525 	.dev_stop             = ixgbe_dev_stop,
526 	.dev_set_link_up    = ixgbe_dev_set_link_up,
527 	.dev_set_link_down  = ixgbe_dev_set_link_down,
528 	.dev_close            = ixgbe_dev_close,
529 	.dev_reset	      = ixgbe_dev_reset,
530 	.promiscuous_enable   = ixgbe_dev_promiscuous_enable,
531 	.promiscuous_disable  = ixgbe_dev_promiscuous_disable,
532 	.allmulticast_enable  = ixgbe_dev_allmulticast_enable,
533 	.allmulticast_disable = ixgbe_dev_allmulticast_disable,
534 	.link_update          = ixgbe_dev_link_update,
535 	.stats_get            = ixgbe_dev_stats_get,
536 	.xstats_get           = ixgbe_dev_xstats_get,
537 	.xstats_get_by_id     = ixgbe_dev_xstats_get_by_id,
538 	.stats_reset          = ixgbe_dev_stats_reset,
539 	.xstats_reset         = ixgbe_dev_xstats_reset,
540 	.xstats_get_names     = ixgbe_dev_xstats_get_names,
541 	.xstats_get_names_by_id = ixgbe_dev_xstats_get_names_by_id,
542 	.queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
543 	.fw_version_get       = ixgbe_fw_version_get,
544 	.dev_infos_get        = ixgbe_dev_info_get,
545 	.dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
546 	.mtu_set              = ixgbe_dev_mtu_set,
547 	.vlan_filter_set      = ixgbe_vlan_filter_set,
548 	.vlan_tpid_set        = ixgbe_vlan_tpid_set,
549 	.vlan_offload_set     = ixgbe_vlan_offload_set,
550 	.vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
551 	.rx_queue_start	      = ixgbe_dev_rx_queue_start,
552 	.rx_queue_stop        = ixgbe_dev_rx_queue_stop,
553 	.tx_queue_start	      = ixgbe_dev_tx_queue_start,
554 	.tx_queue_stop        = ixgbe_dev_tx_queue_stop,
555 	.rx_queue_setup       = ixgbe_dev_rx_queue_setup,
556 	.rx_queue_intr_enable = ixgbe_dev_rx_queue_intr_enable,
557 	.rx_queue_intr_disable = ixgbe_dev_rx_queue_intr_disable,
558 	.rx_queue_release     = ixgbe_dev_rx_queue_release,
559 	.rx_queue_count       = ixgbe_dev_rx_queue_count,
560 	.rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
561 	.rx_descriptor_status = ixgbe_dev_rx_descriptor_status,
562 	.tx_descriptor_status = ixgbe_dev_tx_descriptor_status,
563 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
564 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
565 	.dev_led_on           = ixgbe_dev_led_on,
566 	.dev_led_off          = ixgbe_dev_led_off,
567 	.flow_ctrl_get        = ixgbe_flow_ctrl_get,
568 	.flow_ctrl_set        = ixgbe_flow_ctrl_set,
569 	.priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
570 	.mac_addr_add         = ixgbe_add_rar,
571 	.mac_addr_remove      = ixgbe_remove_rar,
572 	.mac_addr_set         = ixgbe_set_default_mac_addr,
573 	.uc_hash_table_set    = ixgbe_uc_hash_table_set,
574 	.uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
575 	.mirror_rule_set      = ixgbe_mirror_rule_set,
576 	.mirror_rule_reset    = ixgbe_mirror_rule_reset,
577 	.set_queue_rate_limit = ixgbe_set_queue_rate_limit,
578 	.reta_update          = ixgbe_dev_rss_reta_update,
579 	.reta_query           = ixgbe_dev_rss_reta_query,
580 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
581 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
582 	.filter_ctrl          = ixgbe_dev_filter_ctrl,
583 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
584 	.rxq_info_get         = ixgbe_rxq_info_get,
585 	.txq_info_get         = ixgbe_txq_info_get,
586 	.timesync_enable      = ixgbe_timesync_enable,
587 	.timesync_disable     = ixgbe_timesync_disable,
588 	.timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
589 	.timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
590 	.get_reg              = ixgbe_get_regs,
591 	.get_eeprom_length    = ixgbe_get_eeprom_length,
592 	.get_eeprom           = ixgbe_get_eeprom,
593 	.set_eeprom           = ixgbe_set_eeprom,
594 	.get_dcb_info         = ixgbe_dev_get_dcb_info,
595 	.timesync_adjust_time = ixgbe_timesync_adjust_time,
596 	.timesync_read_time   = ixgbe_timesync_read_time,
597 	.timesync_write_time  = ixgbe_timesync_write_time,
598 	.l2_tunnel_eth_type_conf = ixgbe_dev_l2_tunnel_eth_type_conf,
599 	.l2_tunnel_offload_set   = ixgbe_dev_l2_tunnel_offload_set,
600 	.udp_tunnel_port_add  = ixgbe_dev_udp_tunnel_port_add,
601 	.udp_tunnel_port_del  = ixgbe_dev_udp_tunnel_port_del,
602 	.tm_ops_get           = ixgbe_tm_ops_get,
603 };
604 
605 /*
606  * dev_ops for virtual function, bare necessities for basic vf
607  * operation have been implemented
608  */
609 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
610 	.dev_configure        = ixgbevf_dev_configure,
611 	.dev_start            = ixgbevf_dev_start,
612 	.dev_stop             = ixgbevf_dev_stop,
613 	.link_update          = ixgbevf_dev_link_update,
614 	.stats_get            = ixgbevf_dev_stats_get,
615 	.xstats_get           = ixgbevf_dev_xstats_get,
616 	.stats_reset          = ixgbevf_dev_stats_reset,
617 	.xstats_reset         = ixgbevf_dev_stats_reset,
618 	.xstats_get_names     = ixgbevf_dev_xstats_get_names,
619 	.dev_close            = ixgbevf_dev_close,
620 	.dev_reset	      = ixgbevf_dev_reset,
621 	.allmulticast_enable  = ixgbevf_dev_allmulticast_enable,
622 	.allmulticast_disable = ixgbevf_dev_allmulticast_disable,
623 	.dev_infos_get        = ixgbevf_dev_info_get,
624 	.dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
625 	.mtu_set              = ixgbevf_dev_set_mtu,
626 	.vlan_filter_set      = ixgbevf_vlan_filter_set,
627 	.vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
628 	.vlan_offload_set     = ixgbevf_vlan_offload_set,
629 	.rx_queue_setup       = ixgbe_dev_rx_queue_setup,
630 	.rx_queue_release     = ixgbe_dev_rx_queue_release,
631 	.rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
632 	.rx_descriptor_status = ixgbe_dev_rx_descriptor_status,
633 	.tx_descriptor_status = ixgbe_dev_tx_descriptor_status,
634 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
635 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
636 	.rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
637 	.rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
638 	.mac_addr_add         = ixgbevf_add_mac_addr,
639 	.mac_addr_remove      = ixgbevf_remove_mac_addr,
640 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
641 	.rxq_info_get         = ixgbe_rxq_info_get,
642 	.txq_info_get         = ixgbe_txq_info_get,
643 	.mac_addr_set         = ixgbevf_set_default_mac_addr,
644 	.get_reg              = ixgbevf_get_regs,
645 	.reta_update          = ixgbe_dev_rss_reta_update,
646 	.reta_query           = ixgbe_dev_rss_reta_query,
647 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
648 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
649 };
650 
651 /* store statistics names and its offset in stats structure */
652 struct rte_ixgbe_xstats_name_off {
653 	char name[RTE_ETH_XSTATS_NAME_SIZE];
654 	unsigned offset;
655 };
656 
657 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
658 	{"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
659 	{"rx_illegal_byte_errors", offsetof(struct ixgbe_hw_stats, illerrc)},
660 	{"rx_error_bytes", offsetof(struct ixgbe_hw_stats, errbc)},
661 	{"mac_local_errors", offsetof(struct ixgbe_hw_stats, mlfc)},
662 	{"mac_remote_errors", offsetof(struct ixgbe_hw_stats, mrfc)},
663 	{"rx_length_errors", offsetof(struct ixgbe_hw_stats, rlec)},
664 	{"tx_xon_packets", offsetof(struct ixgbe_hw_stats, lxontxc)},
665 	{"rx_xon_packets", offsetof(struct ixgbe_hw_stats, lxonrxc)},
666 	{"tx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxofftxc)},
667 	{"rx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
668 	{"rx_size_64_packets", offsetof(struct ixgbe_hw_stats, prc64)},
669 	{"rx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, prc127)},
670 	{"rx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, prc255)},
671 	{"rx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, prc511)},
672 	{"rx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
673 		prc1023)},
674 	{"rx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
675 		prc1522)},
676 	{"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
677 	{"rx_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
678 	{"rx_fragment_errors", offsetof(struct ixgbe_hw_stats, rfc)},
679 	{"rx_undersize_errors", offsetof(struct ixgbe_hw_stats, ruc)},
680 	{"rx_oversize_errors", offsetof(struct ixgbe_hw_stats, roc)},
681 	{"rx_jabber_errors", offsetof(struct ixgbe_hw_stats, rjc)},
682 	{"rx_management_packets", offsetof(struct ixgbe_hw_stats, mngprc)},
683 	{"rx_management_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
684 	{"tx_management_packets", offsetof(struct ixgbe_hw_stats, mngptc)},
685 	{"rx_total_packets", offsetof(struct ixgbe_hw_stats, tpr)},
686 	{"rx_total_bytes", offsetof(struct ixgbe_hw_stats, tor)},
687 	{"tx_total_packets", offsetof(struct ixgbe_hw_stats, tpt)},
688 	{"tx_size_64_packets", offsetof(struct ixgbe_hw_stats, ptc64)},
689 	{"tx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, ptc127)},
690 	{"tx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, ptc255)},
691 	{"tx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, ptc511)},
692 	{"tx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
693 		ptc1023)},
694 	{"tx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
695 		ptc1522)},
696 	{"tx_multicast_packets", offsetof(struct ixgbe_hw_stats, mptc)},
697 	{"tx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bptc)},
698 	{"rx_mac_short_packet_dropped", offsetof(struct ixgbe_hw_stats, mspdc)},
699 	{"rx_l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
700 
701 	{"flow_director_added_filters", offsetof(struct ixgbe_hw_stats,
702 		fdirustat_add)},
703 	{"flow_director_removed_filters", offsetof(struct ixgbe_hw_stats,
704 		fdirustat_remove)},
705 	{"flow_director_filter_add_errors", offsetof(struct ixgbe_hw_stats,
706 		fdirfstat_fadd)},
707 	{"flow_director_filter_remove_errors", offsetof(struct ixgbe_hw_stats,
708 		fdirfstat_fremove)},
709 	{"flow_director_matched_filters", offsetof(struct ixgbe_hw_stats,
710 		fdirmatch)},
711 	{"flow_director_missed_filters", offsetof(struct ixgbe_hw_stats,
712 		fdirmiss)},
713 
714 	{"rx_fcoe_crc_errors", offsetof(struct ixgbe_hw_stats, fccrc)},
715 	{"rx_fcoe_dropped", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
716 	{"rx_fcoe_mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats,
717 		fclast)},
718 	{"rx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeprc)},
719 	{"tx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeptc)},
720 	{"rx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwrc)},
721 	{"tx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwtc)},
722 	{"rx_fcoe_no_direct_data_placement", offsetof(struct ixgbe_hw_stats,
723 		fcoe_noddp)},
724 	{"rx_fcoe_no_direct_data_placement_ext_buff",
725 		offsetof(struct ixgbe_hw_stats, fcoe_noddp_ext_buff)},
726 
727 	{"tx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
728 		lxontxc)},
729 	{"rx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
730 		lxonrxc)},
731 	{"tx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
732 		lxofftxc)},
733 	{"rx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
734 		lxoffrxc)},
735 	{"rx_total_missed_packets", offsetof(struct ixgbe_hw_stats, mpctotal)},
736 };
737 
738 #define IXGBE_NB_HW_STATS (sizeof(rte_ixgbe_stats_strings) / \
739 			   sizeof(rte_ixgbe_stats_strings[0]))
740 
741 /* MACsec statistics */
742 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_macsec_strings[] = {
743 	{"out_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
744 		out_pkts_untagged)},
745 	{"out_pkts_encrypted", offsetof(struct ixgbe_macsec_stats,
746 		out_pkts_encrypted)},
747 	{"out_pkts_protected", offsetof(struct ixgbe_macsec_stats,
748 		out_pkts_protected)},
749 	{"out_octets_encrypted", offsetof(struct ixgbe_macsec_stats,
750 		out_octets_encrypted)},
751 	{"out_octets_protected", offsetof(struct ixgbe_macsec_stats,
752 		out_octets_protected)},
753 	{"in_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
754 		in_pkts_untagged)},
755 	{"in_pkts_badtag", offsetof(struct ixgbe_macsec_stats,
756 		in_pkts_badtag)},
757 	{"in_pkts_nosci", offsetof(struct ixgbe_macsec_stats,
758 		in_pkts_nosci)},
759 	{"in_pkts_unknownsci", offsetof(struct ixgbe_macsec_stats,
760 		in_pkts_unknownsci)},
761 	{"in_octets_decrypted", offsetof(struct ixgbe_macsec_stats,
762 		in_octets_decrypted)},
763 	{"in_octets_validated", offsetof(struct ixgbe_macsec_stats,
764 		in_octets_validated)},
765 	{"in_pkts_unchecked", offsetof(struct ixgbe_macsec_stats,
766 		in_pkts_unchecked)},
767 	{"in_pkts_delayed", offsetof(struct ixgbe_macsec_stats,
768 		in_pkts_delayed)},
769 	{"in_pkts_late", offsetof(struct ixgbe_macsec_stats,
770 		in_pkts_late)},
771 	{"in_pkts_ok", offsetof(struct ixgbe_macsec_stats,
772 		in_pkts_ok)},
773 	{"in_pkts_invalid", offsetof(struct ixgbe_macsec_stats,
774 		in_pkts_invalid)},
775 	{"in_pkts_notvalid", offsetof(struct ixgbe_macsec_stats,
776 		in_pkts_notvalid)},
777 	{"in_pkts_unusedsa", offsetof(struct ixgbe_macsec_stats,
778 		in_pkts_unusedsa)},
779 	{"in_pkts_notusingsa", offsetof(struct ixgbe_macsec_stats,
780 		in_pkts_notusingsa)},
781 };
782 
783 #define IXGBE_NB_MACSEC_STATS (sizeof(rte_ixgbe_macsec_strings) / \
784 			   sizeof(rte_ixgbe_macsec_strings[0]))
785 
786 /* Per-queue statistics */
787 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_rxq_strings[] = {
788 	{"mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats, rnbc)},
789 	{"dropped", offsetof(struct ixgbe_hw_stats, mpc)},
790 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxonrxc)},
791 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxoffrxc)},
792 };
793 
794 #define IXGBE_NB_RXQ_PRIO_STATS (sizeof(rte_ixgbe_rxq_strings) / \
795 			   sizeof(rte_ixgbe_rxq_strings[0]))
796 #define IXGBE_NB_RXQ_PRIO_VALUES 8
797 
798 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_txq_strings[] = {
799 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxontxc)},
800 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxofftxc)},
801 	{"xon_to_xoff_packets", offsetof(struct ixgbe_hw_stats,
802 		pxon2offc)},
803 };
804 
805 #define IXGBE_NB_TXQ_PRIO_STATS (sizeof(rte_ixgbe_txq_strings) / \
806 			   sizeof(rte_ixgbe_txq_strings[0]))
807 #define IXGBE_NB_TXQ_PRIO_VALUES 8
808 
809 static const struct rte_ixgbe_xstats_name_off rte_ixgbevf_stats_strings[] = {
810 	{"rx_multicast_packets", offsetof(struct ixgbevf_hw_stats, vfmprc)},
811 };
812 
813 #define IXGBEVF_NB_XSTATS (sizeof(rte_ixgbevf_stats_strings) /	\
814 		sizeof(rte_ixgbevf_stats_strings[0]))
815 
816 /**
817  * Atomically reads the link status information from global
818  * structure rte_eth_dev.
819  *
820  * @param dev
821  *   - Pointer to the structure rte_eth_dev to read from.
822  *   - Pointer to the buffer to be saved with the link status.
823  *
824  * @return
825  *   - On success, zero.
826  *   - On failure, negative value.
827  */
828 static inline int
829 rte_ixgbe_dev_atomic_read_link_status(struct rte_eth_dev *dev,
830 				struct rte_eth_link *link)
831 {
832 	struct rte_eth_link *dst = link;
833 	struct rte_eth_link *src = &(dev->data->dev_link);
834 
835 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
836 					*(uint64_t *)src) == 0)
837 		return -1;
838 
839 	return 0;
840 }
841 
842 /**
843  * Atomically writes the link status information into global
844  * structure rte_eth_dev.
845  *
846  * @param dev
847  *   - Pointer to the structure rte_eth_dev to read from.
848  *   - Pointer to the buffer to be saved with the link status.
849  *
850  * @return
851  *   - On success, zero.
852  *   - On failure, negative value.
853  */
854 static inline int
855 rte_ixgbe_dev_atomic_write_link_status(struct rte_eth_dev *dev,
856 				struct rte_eth_link *link)
857 {
858 	struct rte_eth_link *dst = &(dev->data->dev_link);
859 	struct rte_eth_link *src = link;
860 
861 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
862 					*(uint64_t *)src) == 0)
863 		return -1;
864 
865 	return 0;
866 }
867 
868 /*
869  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
870  */
871 static inline int
872 ixgbe_is_sfp(struct ixgbe_hw *hw)
873 {
874 	switch (hw->phy.type) {
875 	case ixgbe_phy_sfp_avago:
876 	case ixgbe_phy_sfp_ftl:
877 	case ixgbe_phy_sfp_intel:
878 	case ixgbe_phy_sfp_unknown:
879 	case ixgbe_phy_sfp_passive_tyco:
880 	case ixgbe_phy_sfp_passive_unknown:
881 		return 1;
882 	default:
883 		return 0;
884 	}
885 }
886 
887 static inline int32_t
888 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
889 {
890 	uint32_t ctrl_ext;
891 	int32_t status;
892 
893 	status = ixgbe_reset_hw(hw);
894 
895 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
896 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
897 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
898 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
899 	IXGBE_WRITE_FLUSH(hw);
900 
901 	if (status == IXGBE_ERR_SFP_NOT_PRESENT)
902 		status = IXGBE_SUCCESS;
903 	return status;
904 }
905 
906 static inline void
907 ixgbe_enable_intr(struct rte_eth_dev *dev)
908 {
909 	struct ixgbe_interrupt *intr =
910 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
911 	struct ixgbe_hw *hw =
912 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
913 
914 	IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
915 	IXGBE_WRITE_FLUSH(hw);
916 }
917 
918 /*
919  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
920  */
921 static void
922 ixgbe_disable_intr(struct ixgbe_hw *hw)
923 {
924 	PMD_INIT_FUNC_TRACE();
925 
926 	if (hw->mac.type == ixgbe_mac_82598EB) {
927 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
928 	} else {
929 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
930 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
931 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
932 	}
933 	IXGBE_WRITE_FLUSH(hw);
934 }
935 
936 /*
937  * This function resets queue statistics mapping registers.
938  * From Niantic datasheet, Initialization of Statistics section:
939  * "...if software requires the queue counters, the RQSMR and TQSM registers
940  * must be re-programmed following a device reset.
941  */
942 static void
943 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
944 {
945 	uint32_t i;
946 
947 	for (i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
948 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
949 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
950 	}
951 }
952 
953 
954 static int
955 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
956 				  uint16_t queue_id,
957 				  uint8_t stat_idx,
958 				  uint8_t is_rx)
959 {
960 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
961 #define NB_QMAP_FIELDS_PER_QSM_REG 4
962 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
963 
964 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
965 	struct ixgbe_stat_mapping_registers *stat_mappings =
966 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
967 	uint32_t qsmr_mask = 0;
968 	uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
969 	uint32_t q_map;
970 	uint8_t n, offset;
971 
972 	if ((hw->mac.type != ixgbe_mac_82599EB) &&
973 		(hw->mac.type != ixgbe_mac_X540) &&
974 		(hw->mac.type != ixgbe_mac_X550) &&
975 		(hw->mac.type != ixgbe_mac_X550EM_x) &&
976 		(hw->mac.type != ixgbe_mac_X550EM_a))
977 		return -ENOSYS;
978 
979 	PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
980 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
981 		     queue_id, stat_idx);
982 
983 	n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
984 	if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
985 		PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
986 		return -EIO;
987 	}
988 	offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
989 
990 	/* Now clear any previous stat_idx set */
991 	clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
992 	if (!is_rx)
993 		stat_mappings->tqsm[n] &= ~clearing_mask;
994 	else
995 		stat_mappings->rqsmr[n] &= ~clearing_mask;
996 
997 	q_map = (uint32_t)stat_idx;
998 	q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
999 	qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
1000 	if (!is_rx)
1001 		stat_mappings->tqsm[n] |= qsmr_mask;
1002 	else
1003 		stat_mappings->rqsmr[n] |= qsmr_mask;
1004 
1005 	PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
1006 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
1007 		     queue_id, stat_idx);
1008 	PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
1009 		     is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
1010 
1011 	/* Now write the mapping in the appropriate register */
1012 	if (is_rx) {
1013 		PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
1014 			     stat_mappings->rqsmr[n], n);
1015 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
1016 	} else {
1017 		PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
1018 			     stat_mappings->tqsm[n], n);
1019 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
1020 	}
1021 	return 0;
1022 }
1023 
1024 static void
1025 ixgbe_restore_statistics_mapping(struct rte_eth_dev *dev)
1026 {
1027 	struct ixgbe_stat_mapping_registers *stat_mappings =
1028 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
1029 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1030 	int i;
1031 
1032 	/* write whatever was in stat mapping table to the NIC */
1033 	for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
1034 		/* rx */
1035 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
1036 
1037 		/* tx */
1038 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
1039 	}
1040 }
1041 
1042 static void
1043 ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config)
1044 {
1045 	uint8_t i;
1046 	struct ixgbe_dcb_tc_config *tc;
1047 	uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
1048 
1049 	dcb_config->num_tcs.pg_tcs = dcb_max_tc;
1050 	dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
1051 	for (i = 0; i < dcb_max_tc; i++) {
1052 		tc = &dcb_config->tc_config[i];
1053 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
1054 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
1055 				 (uint8_t)(100/dcb_max_tc + (i & 1));
1056 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
1057 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
1058 				 (uint8_t)(100/dcb_max_tc + (i & 1));
1059 		tc->pfc = ixgbe_dcb_pfc_disabled;
1060 	}
1061 
1062 	/* Initialize default user to priority mapping, UPx->TC0 */
1063 	tc = &dcb_config->tc_config[0];
1064 	tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
1065 	tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
1066 	for (i = 0; i < IXGBE_DCB_MAX_BW_GROUP; i++) {
1067 		dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
1068 		dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
1069 	}
1070 	dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
1071 	dcb_config->pfc_mode_enable = false;
1072 	dcb_config->vt_mode = true;
1073 	dcb_config->round_robin_enable = false;
1074 	/* support all DCB capabilities in 82599 */
1075 	dcb_config->support.capabilities = 0xFF;
1076 
1077 	/*we only support 4 Tcs for X540, X550 */
1078 	if (hw->mac.type == ixgbe_mac_X540 ||
1079 		hw->mac.type == ixgbe_mac_X550 ||
1080 		hw->mac.type == ixgbe_mac_X550EM_x ||
1081 		hw->mac.type == ixgbe_mac_X550EM_a) {
1082 		dcb_config->num_tcs.pg_tcs = 4;
1083 		dcb_config->num_tcs.pfc_tcs = 4;
1084 	}
1085 }
1086 
1087 /*
1088  * Ensure that all locks are released before first NVM or PHY access
1089  */
1090 static void
1091 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
1092 {
1093 	uint16_t mask;
1094 
1095 	/*
1096 	 * Phy lock should not fail in this early stage. If this is the case,
1097 	 * it is due to an improper exit of the application.
1098 	 * So force the release of the faulty lock. Release of common lock
1099 	 * is done automatically by swfw_sync function.
1100 	 */
1101 	mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
1102 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1103 		PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
1104 	}
1105 	ixgbe_release_swfw_semaphore(hw, mask);
1106 
1107 	/*
1108 	 * These ones are more tricky since they are common to all ports; but
1109 	 * swfw_sync retries last long enough (1s) to be almost sure that if
1110 	 * lock can not be taken it is due to an improper lock of the
1111 	 * semaphore.
1112 	 */
1113 	mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
1114 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1115 		PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1116 	}
1117 	ixgbe_release_swfw_semaphore(hw, mask);
1118 }
1119 
1120 /*
1121  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
1122  * It returns 0 on success.
1123  */
1124 static int
1125 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev)
1126 {
1127 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1128 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1129 	struct ixgbe_hw *hw =
1130 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1131 	struct ixgbe_vfta *shadow_vfta =
1132 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1133 	struct ixgbe_hwstrip *hwstrip =
1134 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1135 	struct ixgbe_dcb_config *dcb_config =
1136 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
1137 	struct ixgbe_filter_info *filter_info =
1138 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1139 	struct ixgbe_bw_conf *bw_conf =
1140 		IXGBE_DEV_PRIVATE_TO_BW_CONF(eth_dev->data->dev_private);
1141 	uint32_t ctrl_ext;
1142 	uint16_t csum;
1143 	int diag, i;
1144 
1145 	PMD_INIT_FUNC_TRACE();
1146 
1147 	eth_dev->dev_ops = &ixgbe_eth_dev_ops;
1148 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1149 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1150 	eth_dev->tx_pkt_prepare = &ixgbe_prep_pkts;
1151 
1152 	/*
1153 	 * For secondary processes, we don't initialise any further as primary
1154 	 * has already done this work. Only check we don't need a different
1155 	 * RX and TX function.
1156 	 */
1157 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1158 		struct ixgbe_tx_queue *txq;
1159 		/* TX queue function in primary, set by last queue initialized
1160 		 * Tx queue may not initialized by primary process
1161 		 */
1162 		if (eth_dev->data->tx_queues) {
1163 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
1164 			ixgbe_set_tx_function(eth_dev, txq);
1165 		} else {
1166 			/* Use default TX function if we get here */
1167 			PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
1168 				     "Using default TX function.");
1169 		}
1170 
1171 		ixgbe_set_rx_function(eth_dev);
1172 
1173 		return 0;
1174 	}
1175 
1176 #ifdef RTE_LIBRTE_SECURITY
1177 	/* Initialize security_ctx only for primary process*/
1178 	eth_dev->security_ctx = ixgbe_ipsec_ctx_create(eth_dev);
1179 	if (eth_dev->security_ctx == NULL)
1180 		return -ENOMEM;
1181 #endif
1182 
1183 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1184 
1185 	/* Vendor and Device ID need to be set before init of shared code */
1186 	hw->device_id = pci_dev->id.device_id;
1187 	hw->vendor_id = pci_dev->id.vendor_id;
1188 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1189 	hw->allow_unsupported_sfp = 1;
1190 
1191 	/* Initialize the shared code (base driver) */
1192 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1193 	diag = ixgbe_bypass_init_shared_code(hw);
1194 #else
1195 	diag = ixgbe_init_shared_code(hw);
1196 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1197 
1198 	if (diag != IXGBE_SUCCESS) {
1199 		PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
1200 		return -EIO;
1201 	}
1202 
1203 	/* pick up the PCI bus settings for reporting later */
1204 	ixgbe_get_bus_info(hw);
1205 
1206 	/* Unlock any pending hardware semaphore */
1207 	ixgbe_swfw_lock_reset(hw);
1208 
1209 	/* Initialize DCB configuration*/
1210 	memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
1211 	ixgbe_dcb_init(hw, dcb_config);
1212 	/* Get Hardware Flow Control setting */
1213 	hw->fc.requested_mode = ixgbe_fc_full;
1214 	hw->fc.current_mode = ixgbe_fc_full;
1215 	hw->fc.pause_time = IXGBE_FC_PAUSE;
1216 	for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
1217 		hw->fc.low_water[i] = IXGBE_FC_LO;
1218 		hw->fc.high_water[i] = IXGBE_FC_HI;
1219 	}
1220 	hw->fc.send_xon = 1;
1221 
1222 	/* Make sure we have a good EEPROM before we read from it */
1223 	diag = ixgbe_validate_eeprom_checksum(hw, &csum);
1224 	if (diag != IXGBE_SUCCESS) {
1225 		PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
1226 		return -EIO;
1227 	}
1228 
1229 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1230 	diag = ixgbe_bypass_init_hw(hw);
1231 #else
1232 	diag = ixgbe_init_hw(hw);
1233 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1234 
1235 	/*
1236 	 * Devices with copper phys will fail to initialise if ixgbe_init_hw()
1237 	 * is called too soon after the kernel driver unbinding/binding occurs.
1238 	 * The failure occurs in ixgbe_identify_phy_generic() for all devices,
1239 	 * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
1240 	 * also called. See ixgbe_identify_phy_82599(). The reason for the
1241 	 * failure is not known, and only occuts when virtualisation features
1242 	 * are disabled in the bios. A delay of 100ms  was found to be enough by
1243 	 * trial-and-error, and is doubled to be safe.
1244 	 */
1245 	if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
1246 		rte_delay_ms(200);
1247 		diag = ixgbe_init_hw(hw);
1248 	}
1249 
1250 	if (diag == IXGBE_ERR_SFP_NOT_PRESENT)
1251 		diag = IXGBE_SUCCESS;
1252 
1253 	if (diag == IXGBE_ERR_EEPROM_VERSION) {
1254 		PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
1255 			     "LOM.  Please be aware there may be issues associated "
1256 			     "with your hardware.");
1257 		PMD_INIT_LOG(ERR, "If you are experiencing problems "
1258 			     "please contact your Intel or hardware representative "
1259 			     "who provided you with this hardware.");
1260 	} else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
1261 		PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
1262 	if (diag) {
1263 		PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
1264 		return -EIO;
1265 	}
1266 
1267 	/* Reset the hw statistics */
1268 	ixgbe_dev_stats_reset(eth_dev);
1269 
1270 	/* disable interrupt */
1271 	ixgbe_disable_intr(hw);
1272 
1273 	/* reset mappings for queue statistics hw counters*/
1274 	ixgbe_reset_qstat_mappings(hw);
1275 
1276 	/* Allocate memory for storing MAC addresses */
1277 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
1278 					       hw->mac.num_rar_entries, 0);
1279 	if (eth_dev->data->mac_addrs == NULL) {
1280 		PMD_INIT_LOG(ERR,
1281 			     "Failed to allocate %u bytes needed to store "
1282 			     "MAC addresses",
1283 			     ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1284 		return -ENOMEM;
1285 	}
1286 	/* Copy the permanent MAC address */
1287 	ether_addr_copy((struct ether_addr *) hw->mac.perm_addr,
1288 			&eth_dev->data->mac_addrs[0]);
1289 
1290 	/* Allocate memory for storing hash filter MAC addresses */
1291 	eth_dev->data->hash_mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
1292 						    IXGBE_VMDQ_NUM_UC_MAC, 0);
1293 	if (eth_dev->data->hash_mac_addrs == NULL) {
1294 		PMD_INIT_LOG(ERR,
1295 			     "Failed to allocate %d bytes needed to store MAC addresses",
1296 			     ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
1297 		return -ENOMEM;
1298 	}
1299 
1300 	/* initialize the vfta */
1301 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1302 
1303 	/* initialize the hw strip bitmap*/
1304 	memset(hwstrip, 0, sizeof(*hwstrip));
1305 
1306 	/* initialize PF if max_vfs not zero */
1307 	ixgbe_pf_host_init(eth_dev);
1308 
1309 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1310 	/* let hardware know driver is loaded */
1311 	ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1312 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1313 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1314 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1315 	IXGBE_WRITE_FLUSH(hw);
1316 
1317 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1318 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1319 			     (int) hw->mac.type, (int) hw->phy.type,
1320 			     (int) hw->phy.sfp_type);
1321 	else
1322 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1323 			     (int) hw->mac.type, (int) hw->phy.type);
1324 
1325 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1326 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1327 		     pci_dev->id.device_id);
1328 
1329 	rte_intr_callback_register(intr_handle,
1330 				   ixgbe_dev_interrupt_handler, eth_dev);
1331 
1332 	/* enable uio/vfio intr/eventfd mapping */
1333 	rte_intr_enable(intr_handle);
1334 
1335 	/* enable support intr */
1336 	ixgbe_enable_intr(eth_dev);
1337 
1338 	/* initialize filter info */
1339 	memset(filter_info, 0,
1340 	       sizeof(struct ixgbe_filter_info));
1341 
1342 	/* initialize 5tuple filter list */
1343 	TAILQ_INIT(&filter_info->fivetuple_list);
1344 
1345 	/* initialize flow director filter list & hash */
1346 	ixgbe_fdir_filter_init(eth_dev);
1347 
1348 	/* initialize l2 tunnel filter list & hash */
1349 	ixgbe_l2_tn_filter_init(eth_dev);
1350 
1351 	/* initialize flow filter lists */
1352 	ixgbe_filterlist_init();
1353 
1354 	/* initialize bandwidth configuration info */
1355 	memset(bw_conf, 0, sizeof(struct ixgbe_bw_conf));
1356 
1357 	/* initialize Traffic Manager configuration */
1358 	ixgbe_tm_conf_init(eth_dev);
1359 
1360 	return 0;
1361 }
1362 
1363 static int
1364 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1365 {
1366 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1367 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1368 	struct ixgbe_hw *hw;
1369 	int retries = 0;
1370 	int ret;
1371 
1372 	PMD_INIT_FUNC_TRACE();
1373 
1374 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1375 		return -EPERM;
1376 
1377 	hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1378 
1379 	if (hw->adapter_stopped == 0)
1380 		ixgbe_dev_close(eth_dev);
1381 
1382 	eth_dev->dev_ops = NULL;
1383 	eth_dev->rx_pkt_burst = NULL;
1384 	eth_dev->tx_pkt_burst = NULL;
1385 
1386 	/* Unlock any pending hardware semaphore */
1387 	ixgbe_swfw_lock_reset(hw);
1388 
1389 	/* disable uio intr before callback unregister */
1390 	rte_intr_disable(intr_handle);
1391 
1392 	do {
1393 		ret = rte_intr_callback_unregister(intr_handle,
1394 				ixgbe_dev_interrupt_handler, eth_dev);
1395 		if (ret >= 0) {
1396 			break;
1397 		} else if (ret != -EAGAIN) {
1398 			PMD_INIT_LOG(ERR,
1399 				"intr callback unregister failed: %d",
1400 				ret);
1401 			return ret;
1402 		}
1403 		rte_delay_ms(100);
1404 	} while (retries++ < (10 + IXGBE_LINK_UP_TIME));
1405 
1406 	/* uninitialize PF if max_vfs not zero */
1407 	ixgbe_pf_host_uninit(eth_dev);
1408 
1409 	rte_free(eth_dev->data->mac_addrs);
1410 	eth_dev->data->mac_addrs = NULL;
1411 
1412 	rte_free(eth_dev->data->hash_mac_addrs);
1413 	eth_dev->data->hash_mac_addrs = NULL;
1414 
1415 	/* remove all the fdir filters & hash */
1416 	ixgbe_fdir_filter_uninit(eth_dev);
1417 
1418 	/* remove all the L2 tunnel filters & hash */
1419 	ixgbe_l2_tn_filter_uninit(eth_dev);
1420 
1421 	/* Remove all ntuple filters of the device */
1422 	ixgbe_ntuple_filter_uninit(eth_dev);
1423 
1424 	/* clear all the filters list */
1425 	ixgbe_filterlist_flush();
1426 
1427 	/* Remove all Traffic Manager configuration */
1428 	ixgbe_tm_conf_uninit(eth_dev);
1429 
1430 #ifdef RTE_LIBRTE_SECURITY
1431 	rte_free(eth_dev->security_ctx);
1432 #endif
1433 
1434 	return 0;
1435 }
1436 
1437 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
1438 {
1439 	struct ixgbe_filter_info *filter_info =
1440 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1441 	struct ixgbe_5tuple_filter *p_5tuple;
1442 
1443 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
1444 		TAILQ_REMOVE(&filter_info->fivetuple_list,
1445 			     p_5tuple,
1446 			     entries);
1447 		rte_free(p_5tuple);
1448 	}
1449 	memset(filter_info->fivetuple_mask, 0,
1450 	       sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1451 
1452 	return 0;
1453 }
1454 
1455 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev)
1456 {
1457 	struct ixgbe_hw_fdir_info *fdir_info =
1458 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1459 	struct ixgbe_fdir_filter *fdir_filter;
1460 
1461 		if (fdir_info->hash_map)
1462 		rte_free(fdir_info->hash_map);
1463 	if (fdir_info->hash_handle)
1464 		rte_hash_free(fdir_info->hash_handle);
1465 
1466 	while ((fdir_filter = TAILQ_FIRST(&fdir_info->fdir_list))) {
1467 		TAILQ_REMOVE(&fdir_info->fdir_list,
1468 			     fdir_filter,
1469 			     entries);
1470 		rte_free(fdir_filter);
1471 	}
1472 
1473 	return 0;
1474 }
1475 
1476 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev)
1477 {
1478 	struct ixgbe_l2_tn_info *l2_tn_info =
1479 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1480 	struct ixgbe_l2_tn_filter *l2_tn_filter;
1481 
1482 	if (l2_tn_info->hash_map)
1483 		rte_free(l2_tn_info->hash_map);
1484 	if (l2_tn_info->hash_handle)
1485 		rte_hash_free(l2_tn_info->hash_handle);
1486 
1487 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
1488 		TAILQ_REMOVE(&l2_tn_info->l2_tn_list,
1489 			     l2_tn_filter,
1490 			     entries);
1491 		rte_free(l2_tn_filter);
1492 	}
1493 
1494 	return 0;
1495 }
1496 
1497 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev)
1498 {
1499 	struct ixgbe_hw_fdir_info *fdir_info =
1500 		IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1501 	char fdir_hash_name[RTE_HASH_NAMESIZE];
1502 	struct rte_hash_parameters fdir_hash_params = {
1503 		.name = fdir_hash_name,
1504 		.entries = IXGBE_MAX_FDIR_FILTER_NUM,
1505 		.key_len = sizeof(union ixgbe_atr_input),
1506 		.hash_func = rte_hash_crc,
1507 		.hash_func_init_val = 0,
1508 		.socket_id = rte_socket_id(),
1509 	};
1510 
1511 	TAILQ_INIT(&fdir_info->fdir_list);
1512 	snprintf(fdir_hash_name, RTE_HASH_NAMESIZE,
1513 		 "fdir_%s", eth_dev->device->name);
1514 	fdir_info->hash_handle = rte_hash_create(&fdir_hash_params);
1515 	if (!fdir_info->hash_handle) {
1516 		PMD_INIT_LOG(ERR, "Failed to create fdir hash table!");
1517 		return -EINVAL;
1518 	}
1519 	fdir_info->hash_map = rte_zmalloc("ixgbe",
1520 					  sizeof(struct ixgbe_fdir_filter *) *
1521 					  IXGBE_MAX_FDIR_FILTER_NUM,
1522 					  0);
1523 	if (!fdir_info->hash_map) {
1524 		PMD_INIT_LOG(ERR,
1525 			     "Failed to allocate memory for fdir hash map!");
1526 		return -ENOMEM;
1527 	}
1528 	fdir_info->mask_added = FALSE;
1529 
1530 	return 0;
1531 }
1532 
1533 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev)
1534 {
1535 	struct ixgbe_l2_tn_info *l2_tn_info =
1536 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1537 	char l2_tn_hash_name[RTE_HASH_NAMESIZE];
1538 	struct rte_hash_parameters l2_tn_hash_params = {
1539 		.name = l2_tn_hash_name,
1540 		.entries = IXGBE_MAX_L2_TN_FILTER_NUM,
1541 		.key_len = sizeof(struct ixgbe_l2_tn_key),
1542 		.hash_func = rte_hash_crc,
1543 		.hash_func_init_val = 0,
1544 		.socket_id = rte_socket_id(),
1545 	};
1546 
1547 	TAILQ_INIT(&l2_tn_info->l2_tn_list);
1548 	snprintf(l2_tn_hash_name, RTE_HASH_NAMESIZE,
1549 		 "l2_tn_%s", eth_dev->device->name);
1550 	l2_tn_info->hash_handle = rte_hash_create(&l2_tn_hash_params);
1551 	if (!l2_tn_info->hash_handle) {
1552 		PMD_INIT_LOG(ERR, "Failed to create L2 TN hash table!");
1553 		return -EINVAL;
1554 	}
1555 	l2_tn_info->hash_map = rte_zmalloc("ixgbe",
1556 				   sizeof(struct ixgbe_l2_tn_filter *) *
1557 				   IXGBE_MAX_L2_TN_FILTER_NUM,
1558 				   0);
1559 	if (!l2_tn_info->hash_map) {
1560 		PMD_INIT_LOG(ERR,
1561 			"Failed to allocate memory for L2 TN hash map!");
1562 		return -ENOMEM;
1563 	}
1564 	l2_tn_info->e_tag_en = FALSE;
1565 	l2_tn_info->e_tag_fwd_en = FALSE;
1566 	l2_tn_info->e_tag_ether_type = DEFAULT_ETAG_ETYPE;
1567 
1568 	return 0;
1569 }
1570 /*
1571  * Negotiate mailbox API version with the PF.
1572  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1573  * Then we try to negotiate starting with the most recent one.
1574  * If all negotiation attempts fail, then we will proceed with
1575  * the default one (ixgbe_mbox_api_10).
1576  */
1577 static void
1578 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1579 {
1580 	int32_t i;
1581 
1582 	/* start with highest supported, proceed down */
1583 	static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1584 		ixgbe_mbox_api_12,
1585 		ixgbe_mbox_api_11,
1586 		ixgbe_mbox_api_10,
1587 	};
1588 
1589 	for (i = 0;
1590 			i != RTE_DIM(sup_ver) &&
1591 			ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1592 			i++)
1593 		;
1594 }
1595 
1596 static void
1597 generate_random_mac_addr(struct ether_addr *mac_addr)
1598 {
1599 	uint64_t random;
1600 
1601 	/* Set Organizationally Unique Identifier (OUI) prefix. */
1602 	mac_addr->addr_bytes[0] = 0x00;
1603 	mac_addr->addr_bytes[1] = 0x09;
1604 	mac_addr->addr_bytes[2] = 0xC0;
1605 	/* Force indication of locally assigned MAC address. */
1606 	mac_addr->addr_bytes[0] |= ETHER_LOCAL_ADMIN_ADDR;
1607 	/* Generate the last 3 bytes of the MAC address with a random number. */
1608 	random = rte_rand();
1609 	memcpy(&mac_addr->addr_bytes[3], &random, 3);
1610 }
1611 
1612 /*
1613  * Virtual Function device init
1614  */
1615 static int
1616 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1617 {
1618 	int diag;
1619 	uint32_t tc, tcs;
1620 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1621 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1622 	struct ixgbe_hw *hw =
1623 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1624 	struct ixgbe_vfta *shadow_vfta =
1625 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1626 	struct ixgbe_hwstrip *hwstrip =
1627 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1628 	struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
1629 
1630 	PMD_INIT_FUNC_TRACE();
1631 
1632 	eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1633 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1634 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1635 
1636 	/* for secondary processes, we don't initialise any further as primary
1637 	 * has already done this work. Only check we don't need a different
1638 	 * RX function
1639 	 */
1640 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1641 		struct ixgbe_tx_queue *txq;
1642 		/* TX queue function in primary, set by last queue initialized
1643 		 * Tx queue may not initialized by primary process
1644 		 */
1645 		if (eth_dev->data->tx_queues) {
1646 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1647 			ixgbe_set_tx_function(eth_dev, txq);
1648 		} else {
1649 			/* Use default TX function if we get here */
1650 			PMD_INIT_LOG(NOTICE,
1651 				     "No TX queues configured yet. Using default TX function.");
1652 		}
1653 
1654 		ixgbe_set_rx_function(eth_dev);
1655 
1656 		return 0;
1657 	}
1658 
1659 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1660 
1661 	hw->device_id = pci_dev->id.device_id;
1662 	hw->vendor_id = pci_dev->id.vendor_id;
1663 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1664 
1665 	/* initialize the vfta */
1666 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1667 
1668 	/* initialize the hw strip bitmap*/
1669 	memset(hwstrip, 0, sizeof(*hwstrip));
1670 
1671 	/* Initialize the shared code (base driver) */
1672 	diag = ixgbe_init_shared_code(hw);
1673 	if (diag != IXGBE_SUCCESS) {
1674 		PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1675 		return -EIO;
1676 	}
1677 
1678 	/* init_mailbox_params */
1679 	hw->mbx.ops.init_params(hw);
1680 
1681 	/* Reset the hw statistics */
1682 	ixgbevf_dev_stats_reset(eth_dev);
1683 
1684 	/* Disable the interrupts for VF */
1685 	ixgbevf_intr_disable(hw);
1686 
1687 	hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1688 	diag = hw->mac.ops.reset_hw(hw);
1689 
1690 	/*
1691 	 * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1692 	 * the underlying PF driver has not assigned a MAC address to the VF.
1693 	 * In this case, assign a random MAC address.
1694 	 */
1695 	if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1696 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1697 		return diag;
1698 	}
1699 
1700 	/* negotiate mailbox API version to use with the PF. */
1701 	ixgbevf_negotiate_api(hw);
1702 
1703 	/* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1704 	ixgbevf_get_queues(hw, &tcs, &tc);
1705 
1706 	/* Allocate memory for storing MAC addresses */
1707 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", ETHER_ADDR_LEN *
1708 					       hw->mac.num_rar_entries, 0);
1709 	if (eth_dev->data->mac_addrs == NULL) {
1710 		PMD_INIT_LOG(ERR,
1711 			     "Failed to allocate %u bytes needed to store "
1712 			     "MAC addresses",
1713 			     ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1714 		return -ENOMEM;
1715 	}
1716 
1717 	/* Generate a random MAC address, if none was assigned by PF. */
1718 	if (is_zero_ether_addr(perm_addr)) {
1719 		generate_random_mac_addr(perm_addr);
1720 		diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1721 		if (diag) {
1722 			rte_free(eth_dev->data->mac_addrs);
1723 			eth_dev->data->mac_addrs = NULL;
1724 			return diag;
1725 		}
1726 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1727 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1728 			     "%02x:%02x:%02x:%02x:%02x:%02x",
1729 			     perm_addr->addr_bytes[0],
1730 			     perm_addr->addr_bytes[1],
1731 			     perm_addr->addr_bytes[2],
1732 			     perm_addr->addr_bytes[3],
1733 			     perm_addr->addr_bytes[4],
1734 			     perm_addr->addr_bytes[5]);
1735 	}
1736 
1737 	/* Copy the permanent MAC address */
1738 	ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1739 
1740 	/* reset the hardware with the new settings */
1741 	diag = hw->mac.ops.start_hw(hw);
1742 	switch (diag) {
1743 	case  0:
1744 		break;
1745 
1746 	default:
1747 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1748 		return -EIO;
1749 	}
1750 
1751 	rte_intr_callback_register(intr_handle,
1752 				   ixgbevf_dev_interrupt_handler, eth_dev);
1753 	rte_intr_enable(intr_handle);
1754 	ixgbevf_intr_enable(hw);
1755 
1756 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1757 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1758 		     pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1759 
1760 	return 0;
1761 }
1762 
1763 /* Virtual Function device uninit */
1764 
1765 static int
1766 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1767 {
1768 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1769 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1770 	struct ixgbe_hw *hw;
1771 
1772 	PMD_INIT_FUNC_TRACE();
1773 
1774 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1775 		return -EPERM;
1776 
1777 	hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1778 
1779 	if (hw->adapter_stopped == 0)
1780 		ixgbevf_dev_close(eth_dev);
1781 
1782 	eth_dev->dev_ops = NULL;
1783 	eth_dev->rx_pkt_burst = NULL;
1784 	eth_dev->tx_pkt_burst = NULL;
1785 
1786 	/* Disable the interrupts for VF */
1787 	ixgbevf_intr_disable(hw);
1788 
1789 	rte_free(eth_dev->data->mac_addrs);
1790 	eth_dev->data->mac_addrs = NULL;
1791 
1792 	rte_intr_disable(intr_handle);
1793 	rte_intr_callback_unregister(intr_handle,
1794 				     ixgbevf_dev_interrupt_handler, eth_dev);
1795 
1796 	return 0;
1797 }
1798 
1799 static int eth_ixgbe_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1800 	struct rte_pci_device *pci_dev)
1801 {
1802 	return rte_eth_dev_pci_generic_probe(pci_dev,
1803 		sizeof(struct ixgbe_adapter), eth_ixgbe_dev_init);
1804 }
1805 
1806 static int eth_ixgbe_pci_remove(struct rte_pci_device *pci_dev)
1807 {
1808 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbe_dev_uninit);
1809 }
1810 
1811 static struct rte_pci_driver rte_ixgbe_pmd = {
1812 	.id_table = pci_id_ixgbe_map,
1813 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
1814 		     RTE_PCI_DRV_IOVA_AS_VA,
1815 	.probe = eth_ixgbe_pci_probe,
1816 	.remove = eth_ixgbe_pci_remove,
1817 };
1818 
1819 static int eth_ixgbevf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1820 	struct rte_pci_device *pci_dev)
1821 {
1822 	return rte_eth_dev_pci_generic_probe(pci_dev,
1823 		sizeof(struct ixgbe_adapter), eth_ixgbevf_dev_init);
1824 }
1825 
1826 static int eth_ixgbevf_pci_remove(struct rte_pci_device *pci_dev)
1827 {
1828 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbevf_dev_uninit);
1829 }
1830 
1831 /*
1832  * virtual function driver struct
1833  */
1834 static struct rte_pci_driver rte_ixgbevf_pmd = {
1835 	.id_table = pci_id_ixgbevf_map,
1836 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_IOVA_AS_VA,
1837 	.probe = eth_ixgbevf_pci_probe,
1838 	.remove = eth_ixgbevf_pci_remove,
1839 };
1840 
1841 static int
1842 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1843 {
1844 	struct ixgbe_hw *hw =
1845 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1846 	struct ixgbe_vfta *shadow_vfta =
1847 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1848 	uint32_t vfta;
1849 	uint32_t vid_idx;
1850 	uint32_t vid_bit;
1851 
1852 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1853 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1854 	vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1855 	if (on)
1856 		vfta |= vid_bit;
1857 	else
1858 		vfta &= ~vid_bit;
1859 	IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1860 
1861 	/* update local VFTA copy */
1862 	shadow_vfta->vfta[vid_idx] = vfta;
1863 
1864 	return 0;
1865 }
1866 
1867 static void
1868 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1869 {
1870 	if (on)
1871 		ixgbe_vlan_hw_strip_enable(dev, queue);
1872 	else
1873 		ixgbe_vlan_hw_strip_disable(dev, queue);
1874 }
1875 
1876 static int
1877 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1878 		    enum rte_vlan_type vlan_type,
1879 		    uint16_t tpid)
1880 {
1881 	struct ixgbe_hw *hw =
1882 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1883 	int ret = 0;
1884 	uint32_t reg;
1885 	uint32_t qinq;
1886 
1887 	qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1888 	qinq &= IXGBE_DMATXCTL_GDV;
1889 
1890 	switch (vlan_type) {
1891 	case ETH_VLAN_TYPE_INNER:
1892 		if (qinq) {
1893 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1894 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1895 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1896 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1897 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1898 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1899 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1900 		} else {
1901 			ret = -ENOTSUP;
1902 			PMD_DRV_LOG(ERR, "Inner type is not supported"
1903 				    " by single VLAN");
1904 		}
1905 		break;
1906 	case ETH_VLAN_TYPE_OUTER:
1907 		if (qinq) {
1908 			/* Only the high 16-bits is valid */
1909 			IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1910 					IXGBE_EXVET_VET_EXT_SHIFT);
1911 		} else {
1912 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1913 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1914 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1915 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1916 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1917 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1918 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1919 		}
1920 
1921 		break;
1922 	default:
1923 		ret = -EINVAL;
1924 		PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1925 		break;
1926 	}
1927 
1928 	return ret;
1929 }
1930 
1931 void
1932 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1933 {
1934 	struct ixgbe_hw *hw =
1935 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1936 	uint32_t vlnctrl;
1937 
1938 	PMD_INIT_FUNC_TRACE();
1939 
1940 	/* Filter Table Disable */
1941 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1942 	vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1943 
1944 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1945 }
1946 
1947 void
1948 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1949 {
1950 	struct ixgbe_hw *hw =
1951 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1952 	struct ixgbe_vfta *shadow_vfta =
1953 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1954 	uint32_t vlnctrl;
1955 	uint16_t i;
1956 
1957 	PMD_INIT_FUNC_TRACE();
1958 
1959 	/* Filter Table Enable */
1960 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1961 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1962 	vlnctrl |= IXGBE_VLNCTRL_VFE;
1963 
1964 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1965 
1966 	/* write whatever is in local vfta copy */
1967 	for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1968 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1969 }
1970 
1971 static void
1972 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1973 {
1974 	struct ixgbe_hwstrip *hwstrip =
1975 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1976 	struct ixgbe_rx_queue *rxq;
1977 
1978 	if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
1979 		return;
1980 
1981 	if (on)
1982 		IXGBE_SET_HWSTRIP(hwstrip, queue);
1983 	else
1984 		IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1985 
1986 	if (queue >= dev->data->nb_rx_queues)
1987 		return;
1988 
1989 	rxq = dev->data->rx_queues[queue];
1990 
1991 	if (on)
1992 		rxq->vlan_flags = PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
1993 	else
1994 		rxq->vlan_flags = PKT_RX_VLAN;
1995 }
1996 
1997 static void
1998 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
1999 {
2000 	struct ixgbe_hw *hw =
2001 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2002 	uint32_t ctrl;
2003 
2004 	PMD_INIT_FUNC_TRACE();
2005 
2006 	if (hw->mac.type == ixgbe_mac_82598EB) {
2007 		/* No queue level support */
2008 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2009 		return;
2010 	}
2011 
2012 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2013 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2014 	ctrl &= ~IXGBE_RXDCTL_VME;
2015 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2016 
2017 	/* record those setting for HW strip per queue */
2018 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
2019 }
2020 
2021 static void
2022 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
2023 {
2024 	struct ixgbe_hw *hw =
2025 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2026 	uint32_t ctrl;
2027 
2028 	PMD_INIT_FUNC_TRACE();
2029 
2030 	if (hw->mac.type == ixgbe_mac_82598EB) {
2031 		/* No queue level supported */
2032 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2033 		return;
2034 	}
2035 
2036 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2037 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2038 	ctrl |= IXGBE_RXDCTL_VME;
2039 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2040 
2041 	/* record those setting for HW strip per queue */
2042 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
2043 }
2044 
2045 void
2046 ixgbe_vlan_hw_strip_disable_all(struct rte_eth_dev *dev)
2047 {
2048 	struct ixgbe_hw *hw =
2049 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2050 	uint32_t ctrl;
2051 	uint16_t i;
2052 	struct ixgbe_rx_queue *rxq;
2053 
2054 	PMD_INIT_FUNC_TRACE();
2055 
2056 	if (hw->mac.type == ixgbe_mac_82598EB) {
2057 		ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2058 		ctrl &= ~IXGBE_VLNCTRL_VME;
2059 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2060 	} else {
2061 		/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2062 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
2063 			rxq = dev->data->rx_queues[i];
2064 			ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2065 			ctrl &= ~IXGBE_RXDCTL_VME;
2066 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2067 
2068 			/* record those setting for HW strip per queue */
2069 			ixgbe_vlan_hw_strip_bitmap_set(dev, i, 0);
2070 		}
2071 	}
2072 }
2073 
2074 void
2075 ixgbe_vlan_hw_strip_enable_all(struct rte_eth_dev *dev)
2076 {
2077 	struct ixgbe_hw *hw =
2078 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2079 	uint32_t ctrl;
2080 	uint16_t i;
2081 	struct ixgbe_rx_queue *rxq;
2082 
2083 	PMD_INIT_FUNC_TRACE();
2084 
2085 	if (hw->mac.type == ixgbe_mac_82598EB) {
2086 		ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2087 		ctrl |= IXGBE_VLNCTRL_VME;
2088 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2089 	} else {
2090 		/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2091 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
2092 			rxq = dev->data->rx_queues[i];
2093 			ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2094 			ctrl |= IXGBE_RXDCTL_VME;
2095 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2096 
2097 			/* record those setting for HW strip per queue */
2098 			ixgbe_vlan_hw_strip_bitmap_set(dev, i, 1);
2099 		}
2100 	}
2101 }
2102 
2103 static void
2104 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2105 {
2106 	struct ixgbe_hw *hw =
2107 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2108 	uint32_t ctrl;
2109 
2110 	PMD_INIT_FUNC_TRACE();
2111 
2112 	/* DMATXCTRL: Geric Double VLAN Disable */
2113 	ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2114 	ctrl &= ~IXGBE_DMATXCTL_GDV;
2115 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2116 
2117 	/* CTRL_EXT: Global Double VLAN Disable */
2118 	ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2119 	ctrl &= ~IXGBE_EXTENDED_VLAN;
2120 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2121 
2122 }
2123 
2124 static void
2125 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2126 {
2127 	struct ixgbe_hw *hw =
2128 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2129 	uint32_t ctrl;
2130 
2131 	PMD_INIT_FUNC_TRACE();
2132 
2133 	/* DMATXCTRL: Geric Double VLAN Enable */
2134 	ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2135 	ctrl |= IXGBE_DMATXCTL_GDV;
2136 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2137 
2138 	/* CTRL_EXT: Global Double VLAN Enable */
2139 	ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2140 	ctrl |= IXGBE_EXTENDED_VLAN;
2141 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2142 
2143 	/* Clear pooling mode of PFVTCTL. It's required by X550. */
2144 	if (hw->mac.type == ixgbe_mac_X550 ||
2145 	    hw->mac.type == ixgbe_mac_X550EM_x ||
2146 	    hw->mac.type == ixgbe_mac_X550EM_a) {
2147 		ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
2148 		ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
2149 		IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
2150 	}
2151 
2152 	/*
2153 	 * VET EXT field in the EXVET register = 0x8100 by default
2154 	 * So no need to change. Same to VT field of DMATXCTL register
2155 	 */
2156 }
2157 
2158 static int
2159 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2160 {
2161 	if (mask & ETH_VLAN_STRIP_MASK) {
2162 		if (dev->data->dev_conf.rxmode.hw_vlan_strip)
2163 			ixgbe_vlan_hw_strip_enable_all(dev);
2164 		else
2165 			ixgbe_vlan_hw_strip_disable_all(dev);
2166 	}
2167 
2168 	if (mask & ETH_VLAN_FILTER_MASK) {
2169 		if (dev->data->dev_conf.rxmode.hw_vlan_filter)
2170 			ixgbe_vlan_hw_filter_enable(dev);
2171 		else
2172 			ixgbe_vlan_hw_filter_disable(dev);
2173 	}
2174 
2175 	if (mask & ETH_VLAN_EXTEND_MASK) {
2176 		if (dev->data->dev_conf.rxmode.hw_vlan_extend)
2177 			ixgbe_vlan_hw_extend_enable(dev);
2178 		else
2179 			ixgbe_vlan_hw_extend_disable(dev);
2180 	}
2181 
2182 	return 0;
2183 }
2184 
2185 static void
2186 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2187 {
2188 	struct ixgbe_hw *hw =
2189 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2190 	/* VLNCTRL: enable vlan filtering and allow all vlan tags through */
2191 	uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2192 
2193 	vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
2194 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
2195 }
2196 
2197 static int
2198 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
2199 {
2200 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2201 
2202 	switch (nb_rx_q) {
2203 	case 1:
2204 	case 2:
2205 		RTE_ETH_DEV_SRIOV(dev).active = ETH_64_POOLS;
2206 		break;
2207 	case 4:
2208 		RTE_ETH_DEV_SRIOV(dev).active = ETH_32_POOLS;
2209 		break;
2210 	default:
2211 		return -EINVAL;
2212 	}
2213 
2214 	RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
2215 		IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2216 	RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
2217 		pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2218 	return 0;
2219 }
2220 
2221 static int
2222 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
2223 {
2224 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
2225 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2226 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
2227 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
2228 
2229 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
2230 		/* check multi-queue mode */
2231 		switch (dev_conf->rxmode.mq_mode) {
2232 		case ETH_MQ_RX_VMDQ_DCB:
2233 			PMD_INIT_LOG(INFO, "ETH_MQ_RX_VMDQ_DCB mode supported in SRIOV");
2234 			break;
2235 		case ETH_MQ_RX_VMDQ_DCB_RSS:
2236 			/* DCB/RSS VMDQ in SRIOV mode, not implement yet */
2237 			PMD_INIT_LOG(ERR, "SRIOV active,"
2238 					" unsupported mq_mode rx %d.",
2239 					dev_conf->rxmode.mq_mode);
2240 			return -EINVAL;
2241 		case ETH_MQ_RX_RSS:
2242 		case ETH_MQ_RX_VMDQ_RSS:
2243 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_RSS;
2244 			if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
2245 				if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
2246 					PMD_INIT_LOG(ERR, "SRIOV is active,"
2247 						" invalid queue number"
2248 						" for VMDQ RSS, allowed"
2249 						" value are 1, 2 or 4.");
2250 					return -EINVAL;
2251 				}
2252 			break;
2253 		case ETH_MQ_RX_VMDQ_ONLY:
2254 		case ETH_MQ_RX_NONE:
2255 			/* if nothing mq mode configure, use default scheme */
2256 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
2257 			break;
2258 		default: /* ETH_MQ_RX_DCB, ETH_MQ_RX_DCB_RSS or ETH_MQ_TX_DCB*/
2259 			/* SRIOV only works in VMDq enable mode */
2260 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2261 					" wrong mq_mode rx %d.",
2262 					dev_conf->rxmode.mq_mode);
2263 			return -EINVAL;
2264 		}
2265 
2266 		switch (dev_conf->txmode.mq_mode) {
2267 		case ETH_MQ_TX_VMDQ_DCB:
2268 			PMD_INIT_LOG(INFO, "ETH_MQ_TX_VMDQ_DCB mode supported in SRIOV");
2269 			dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_DCB;
2270 			break;
2271 		default: /* ETH_MQ_TX_VMDQ_ONLY or ETH_MQ_TX_NONE */
2272 			dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_ONLY;
2273 			break;
2274 		}
2275 
2276 		/* check valid queue number */
2277 		if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2278 		    (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2279 			PMD_INIT_LOG(ERR, "SRIOV is active,"
2280 					" nb_rx_q=%d nb_tx_q=%d queue number"
2281 					" must be less than or equal to %d.",
2282 					nb_rx_q, nb_tx_q,
2283 					RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2284 			return -EINVAL;
2285 		}
2286 	} else {
2287 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB_RSS) {
2288 			PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2289 					  " not supported.");
2290 			return -EINVAL;
2291 		}
2292 		/* check configuration for vmdb+dcb mode */
2293 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB) {
2294 			const struct rte_eth_vmdq_dcb_conf *conf;
2295 
2296 			if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2297 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2298 						IXGBE_VMDQ_DCB_NB_QUEUES);
2299 				return -EINVAL;
2300 			}
2301 			conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2302 			if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2303 			       conf->nb_queue_pools == ETH_32_POOLS)) {
2304 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2305 						" nb_queue_pools must be %d or %d.",
2306 						ETH_16_POOLS, ETH_32_POOLS);
2307 				return -EINVAL;
2308 			}
2309 		}
2310 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_VMDQ_DCB) {
2311 			const struct rte_eth_vmdq_dcb_tx_conf *conf;
2312 
2313 			if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2314 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2315 						 IXGBE_VMDQ_DCB_NB_QUEUES);
2316 				return -EINVAL;
2317 			}
2318 			conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2319 			if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2320 			       conf->nb_queue_pools == ETH_32_POOLS)) {
2321 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2322 						" nb_queue_pools != %d and"
2323 						" nb_queue_pools != %d.",
2324 						ETH_16_POOLS, ETH_32_POOLS);
2325 				return -EINVAL;
2326 			}
2327 		}
2328 
2329 		/* For DCB mode check our configuration before we go further */
2330 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB) {
2331 			const struct rte_eth_dcb_rx_conf *conf;
2332 
2333 			conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2334 			if (!(conf->nb_tcs == ETH_4_TCS ||
2335 			       conf->nb_tcs == ETH_8_TCS)) {
2336 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2337 						" and nb_tcs != %d.",
2338 						ETH_4_TCS, ETH_8_TCS);
2339 				return -EINVAL;
2340 			}
2341 		}
2342 
2343 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_DCB) {
2344 			const struct rte_eth_dcb_tx_conf *conf;
2345 
2346 			conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2347 			if (!(conf->nb_tcs == ETH_4_TCS ||
2348 			       conf->nb_tcs == ETH_8_TCS)) {
2349 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2350 						" and nb_tcs != %d.",
2351 						ETH_4_TCS, ETH_8_TCS);
2352 				return -EINVAL;
2353 			}
2354 		}
2355 
2356 		/*
2357 		 * When DCB/VT is off, maximum number of queues changes,
2358 		 * except for 82598EB, which remains constant.
2359 		 */
2360 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
2361 				hw->mac.type != ixgbe_mac_82598EB) {
2362 			if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2363 				PMD_INIT_LOG(ERR,
2364 					     "Neither VT nor DCB are enabled, "
2365 					     "nb_tx_q > %d.",
2366 					     IXGBE_NONE_MODE_TX_NB_QUEUES);
2367 				return -EINVAL;
2368 			}
2369 		}
2370 	}
2371 	return 0;
2372 }
2373 
2374 static int
2375 ixgbe_dev_configure(struct rte_eth_dev *dev)
2376 {
2377 	struct ixgbe_interrupt *intr =
2378 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2379 	struct ixgbe_adapter *adapter =
2380 		(struct ixgbe_adapter *)dev->data->dev_private;
2381 	int ret;
2382 
2383 	PMD_INIT_FUNC_TRACE();
2384 	/* multipe queue mode checking */
2385 	ret  = ixgbe_check_mq_mode(dev);
2386 	if (ret != 0) {
2387 		PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2388 			    ret);
2389 		return ret;
2390 	}
2391 
2392 	/* set flag to update link status after init */
2393 	intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2394 
2395 	/*
2396 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2397 	 * allocation or vector Rx preconditions we will reset it.
2398 	 */
2399 	adapter->rx_bulk_alloc_allowed = true;
2400 	adapter->rx_vec_allowed = true;
2401 
2402 	return 0;
2403 }
2404 
2405 static void
2406 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2407 {
2408 	struct ixgbe_hw *hw =
2409 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2410 	struct ixgbe_interrupt *intr =
2411 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2412 	uint32_t gpie;
2413 
2414 	/* only set up it on X550EM_X */
2415 	if (hw->mac.type == ixgbe_mac_X550EM_x) {
2416 		gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2417 		gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2418 		IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2419 		if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2420 			intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2421 	}
2422 }
2423 
2424 int
2425 ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
2426 			uint16_t tx_rate, uint64_t q_msk)
2427 {
2428 	struct ixgbe_hw *hw;
2429 	struct ixgbe_vf_info *vfinfo;
2430 	struct rte_eth_link link;
2431 	uint8_t  nb_q_per_pool;
2432 	uint32_t queue_stride;
2433 	uint32_t queue_idx, idx = 0, vf_idx;
2434 	uint32_t queue_end;
2435 	uint16_t total_rate = 0;
2436 	struct rte_pci_device *pci_dev;
2437 
2438 	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2439 	rte_eth_link_get_nowait(dev->data->port_id, &link);
2440 
2441 	if (vf >= pci_dev->max_vfs)
2442 		return -EINVAL;
2443 
2444 	if (tx_rate > link.link_speed)
2445 		return -EINVAL;
2446 
2447 	if (q_msk == 0)
2448 		return 0;
2449 
2450 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2451 	vfinfo = *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
2452 	nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2453 	queue_stride = IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2454 	queue_idx = vf * queue_stride;
2455 	queue_end = queue_idx + nb_q_per_pool - 1;
2456 	if (queue_end >= hw->mac.max_tx_queues)
2457 		return -EINVAL;
2458 
2459 	if (vfinfo) {
2460 		for (vf_idx = 0; vf_idx < pci_dev->max_vfs; vf_idx++) {
2461 			if (vf_idx == vf)
2462 				continue;
2463 			for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
2464 				idx++)
2465 				total_rate += vfinfo[vf_idx].tx_rate[idx];
2466 		}
2467 	} else {
2468 		return -EINVAL;
2469 	}
2470 
2471 	/* Store tx_rate for this vf. */
2472 	for (idx = 0; idx < nb_q_per_pool; idx++) {
2473 		if (((uint64_t)0x1 << idx) & q_msk) {
2474 			if (vfinfo[vf].tx_rate[idx] != tx_rate)
2475 				vfinfo[vf].tx_rate[idx] = tx_rate;
2476 			total_rate += tx_rate;
2477 		}
2478 	}
2479 
2480 	if (total_rate > dev->data->dev_link.link_speed) {
2481 		/* Reset stored TX rate of the VF if it causes exceed
2482 		 * link speed.
2483 		 */
2484 		memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
2485 		return -EINVAL;
2486 	}
2487 
2488 	/* Set RTTBCNRC of each queue/pool for vf X  */
2489 	for (; queue_idx <= queue_end; queue_idx++) {
2490 		if (0x1 & q_msk)
2491 			ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
2492 		q_msk = q_msk >> 1;
2493 	}
2494 
2495 	return 0;
2496 }
2497 
2498 /*
2499  * Configure device link speed and setup link.
2500  * It returns 0 on success.
2501  */
2502 static int
2503 ixgbe_dev_start(struct rte_eth_dev *dev)
2504 {
2505 	struct ixgbe_hw *hw =
2506 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2507 	struct ixgbe_vf_info *vfinfo =
2508 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2509 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2510 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2511 	uint32_t intr_vector = 0;
2512 	int err, link_up = 0, negotiate = 0;
2513 	uint32_t speed = 0;
2514 	int mask = 0;
2515 	int status;
2516 	uint16_t vf, idx;
2517 	uint32_t *link_speeds;
2518 	struct ixgbe_tm_conf *tm_conf =
2519 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2520 
2521 	PMD_INIT_FUNC_TRACE();
2522 
2523 	/* IXGBE devices don't support:
2524 	*    - half duplex (checked afterwards for valid speeds)
2525 	*    - fixed speed: TODO implement
2526 	*/
2527 	if (dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED) {
2528 		PMD_INIT_LOG(ERR,
2529 		"Invalid link_speeds for port %u, fix speed not supported",
2530 				dev->data->port_id);
2531 		return -EINVAL;
2532 	}
2533 
2534 	/* disable uio/vfio intr/eventfd mapping */
2535 	rte_intr_disable(intr_handle);
2536 
2537 	/* stop adapter */
2538 	hw->adapter_stopped = 0;
2539 	ixgbe_stop_adapter(hw);
2540 
2541 	/* reinitialize adapter
2542 	 * this calls reset and start
2543 	 */
2544 	status = ixgbe_pf_reset_hw(hw);
2545 	if (status != 0)
2546 		return -1;
2547 	hw->mac.ops.start_hw(hw);
2548 	hw->mac.get_link_status = true;
2549 
2550 	/* configure PF module if SRIOV enabled */
2551 	ixgbe_pf_host_configure(dev);
2552 
2553 	ixgbe_dev_phy_intr_setup(dev);
2554 
2555 	/* check and configure queue intr-vector mapping */
2556 	if ((rte_intr_cap_multiple(intr_handle) ||
2557 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
2558 	    dev->data->dev_conf.intr_conf.rxq != 0) {
2559 		intr_vector = dev->data->nb_rx_queues;
2560 		if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2561 			PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2562 					IXGBE_MAX_INTR_QUEUE_NUM);
2563 			return -ENOTSUP;
2564 		}
2565 		if (rte_intr_efd_enable(intr_handle, intr_vector))
2566 			return -1;
2567 	}
2568 
2569 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
2570 		intr_handle->intr_vec =
2571 			rte_zmalloc("intr_vec",
2572 				    dev->data->nb_rx_queues * sizeof(int), 0);
2573 		if (intr_handle->intr_vec == NULL) {
2574 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2575 				     " intr_vec", dev->data->nb_rx_queues);
2576 			return -ENOMEM;
2577 		}
2578 	}
2579 
2580 	/* confiugre msix for sleep until rx interrupt */
2581 	ixgbe_configure_msix(dev);
2582 
2583 	/* initialize transmission unit */
2584 	ixgbe_dev_tx_init(dev);
2585 
2586 	/* This can fail when allocating mbufs for descriptor rings */
2587 	err = ixgbe_dev_rx_init(dev);
2588 	if (err) {
2589 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2590 		goto error;
2591 	}
2592 
2593 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
2594 		ETH_VLAN_EXTEND_MASK;
2595 	err = ixgbe_vlan_offload_set(dev, mask);
2596 	if (err) {
2597 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload");
2598 		goto error;
2599 	}
2600 
2601 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
2602 		/* Enable vlan filtering for VMDq */
2603 		ixgbe_vmdq_vlan_hw_filter_enable(dev);
2604 	}
2605 
2606 	/* Configure DCB hw */
2607 	ixgbe_configure_dcb(dev);
2608 
2609 	if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2610 		err = ixgbe_fdir_configure(dev);
2611 		if (err)
2612 			goto error;
2613 	}
2614 
2615 	/* Restore vf rate limit */
2616 	if (vfinfo != NULL) {
2617 		for (vf = 0; vf < pci_dev->max_vfs; vf++)
2618 			for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2619 				if (vfinfo[vf].tx_rate[idx] != 0)
2620 					ixgbe_set_vf_rate_limit(
2621 						dev, vf,
2622 						vfinfo[vf].tx_rate[idx],
2623 						1 << idx);
2624 	}
2625 
2626 	ixgbe_restore_statistics_mapping(dev);
2627 
2628 	err = ixgbe_dev_rxtx_start(dev);
2629 	if (err < 0) {
2630 		PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2631 		goto error;
2632 	}
2633 
2634 	/* Skip link setup if loopback mode is enabled for 82599. */
2635 	if (hw->mac.type == ixgbe_mac_82599EB &&
2636 			dev->data->dev_conf.lpbk_mode == IXGBE_LPBK_82599_TX_RX)
2637 		goto skip_link_setup;
2638 
2639 	if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2640 		err = hw->mac.ops.setup_sfp(hw);
2641 		if (err)
2642 			goto error;
2643 	}
2644 
2645 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2646 		/* Turn on the copper */
2647 		ixgbe_set_phy_power(hw, true);
2648 	} else {
2649 		/* Turn on the laser */
2650 		ixgbe_enable_tx_laser(hw);
2651 	}
2652 
2653 	err = ixgbe_check_link(hw, &speed, &link_up, 0);
2654 	if (err)
2655 		goto error;
2656 	dev->data->dev_link.link_status = link_up;
2657 
2658 	err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2659 	if (err)
2660 		goto error;
2661 
2662 	link_speeds = &dev->data->dev_conf.link_speeds;
2663 	if (*link_speeds & ~(ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2664 			ETH_LINK_SPEED_10G)) {
2665 		PMD_INIT_LOG(ERR, "Invalid link setting");
2666 		goto error;
2667 	}
2668 
2669 	speed = 0x0;
2670 	if (*link_speeds == ETH_LINK_SPEED_AUTONEG) {
2671 		switch (hw->mac.type) {
2672 		case ixgbe_mac_82598EB:
2673 			speed = IXGBE_LINK_SPEED_82598_AUTONEG;
2674 			break;
2675 		case ixgbe_mac_82599EB:
2676 		case ixgbe_mac_X540:
2677 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2678 			break;
2679 		case ixgbe_mac_X550:
2680 		case ixgbe_mac_X550EM_x:
2681 		case ixgbe_mac_X550EM_a:
2682 			speed = IXGBE_LINK_SPEED_X550_AUTONEG;
2683 			break;
2684 		default:
2685 			speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2686 		}
2687 	} else {
2688 		if (*link_speeds & ETH_LINK_SPEED_10G)
2689 			speed |= IXGBE_LINK_SPEED_10GB_FULL;
2690 		if (*link_speeds & ETH_LINK_SPEED_1G)
2691 			speed |= IXGBE_LINK_SPEED_1GB_FULL;
2692 		if (*link_speeds & ETH_LINK_SPEED_100M)
2693 			speed |= IXGBE_LINK_SPEED_100_FULL;
2694 	}
2695 
2696 	err = ixgbe_setup_link(hw, speed, link_up);
2697 	if (err)
2698 		goto error;
2699 
2700 skip_link_setup:
2701 
2702 	if (rte_intr_allow_others(intr_handle)) {
2703 		/* check if lsc interrupt is enabled */
2704 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2705 			ixgbe_dev_lsc_interrupt_setup(dev, TRUE);
2706 		else
2707 			ixgbe_dev_lsc_interrupt_setup(dev, FALSE);
2708 		ixgbe_dev_macsec_interrupt_setup(dev);
2709 	} else {
2710 		rte_intr_callback_unregister(intr_handle,
2711 					     ixgbe_dev_interrupt_handler, dev);
2712 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2713 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
2714 				     " no intr multiplex");
2715 	}
2716 
2717 	/* check if rxq interrupt is enabled */
2718 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2719 	    rte_intr_dp_is_en(intr_handle))
2720 		ixgbe_dev_rxq_interrupt_setup(dev);
2721 
2722 	/* enable uio/vfio intr/eventfd mapping */
2723 	rte_intr_enable(intr_handle);
2724 
2725 	/* resume enabled intr since hw reset */
2726 	ixgbe_enable_intr(dev);
2727 	ixgbe_l2_tunnel_conf(dev);
2728 	ixgbe_filter_restore(dev);
2729 
2730 	if (tm_conf->root && !tm_conf->committed)
2731 		PMD_DRV_LOG(WARNING,
2732 			    "please call hierarchy_commit() "
2733 			    "before starting the port");
2734 
2735 	return 0;
2736 
2737 error:
2738 	PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2739 	ixgbe_dev_clear_queues(dev);
2740 	return -EIO;
2741 }
2742 
2743 /*
2744  * Stop device: disable rx and tx functions to allow for reconfiguring.
2745  */
2746 static void
2747 ixgbe_dev_stop(struct rte_eth_dev *dev)
2748 {
2749 	struct rte_eth_link link;
2750 	struct ixgbe_hw *hw =
2751 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2752 	struct ixgbe_vf_info *vfinfo =
2753 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2754 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2755 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2756 	int vf;
2757 	struct ixgbe_tm_conf *tm_conf =
2758 		IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2759 
2760 	PMD_INIT_FUNC_TRACE();
2761 
2762 	/* disable interrupts */
2763 	ixgbe_disable_intr(hw);
2764 
2765 	/* reset the NIC */
2766 	ixgbe_pf_reset_hw(hw);
2767 	hw->adapter_stopped = 0;
2768 
2769 	/* stop adapter */
2770 	ixgbe_stop_adapter(hw);
2771 
2772 	for (vf = 0; vfinfo != NULL && vf < pci_dev->max_vfs; vf++)
2773 		vfinfo[vf].clear_to_send = false;
2774 
2775 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2776 		/* Turn off the copper */
2777 		ixgbe_set_phy_power(hw, false);
2778 	} else {
2779 		/* Turn off the laser */
2780 		ixgbe_disable_tx_laser(hw);
2781 	}
2782 
2783 	ixgbe_dev_clear_queues(dev);
2784 
2785 	/* Clear stored conf */
2786 	dev->data->scattered_rx = 0;
2787 	dev->data->lro = 0;
2788 
2789 	/* Clear recorded link status */
2790 	memset(&link, 0, sizeof(link));
2791 	rte_ixgbe_dev_atomic_write_link_status(dev, &link);
2792 
2793 	if (!rte_intr_allow_others(intr_handle))
2794 		/* resume to the default handler */
2795 		rte_intr_callback_register(intr_handle,
2796 					   ixgbe_dev_interrupt_handler,
2797 					   (void *)dev);
2798 
2799 	/* Clean datapath event and queue/vec mapping */
2800 	rte_intr_efd_disable(intr_handle);
2801 	if (intr_handle->intr_vec != NULL) {
2802 		rte_free(intr_handle->intr_vec);
2803 		intr_handle->intr_vec = NULL;
2804 	}
2805 
2806 	/* reset hierarchy commit */
2807 	tm_conf->committed = false;
2808 }
2809 
2810 /*
2811  * Set device link up: enable tx.
2812  */
2813 static int
2814 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2815 {
2816 	struct ixgbe_hw *hw =
2817 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2818 	if (hw->mac.type == ixgbe_mac_82599EB) {
2819 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2820 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2821 			/* Not suported in bypass mode */
2822 			PMD_INIT_LOG(ERR, "Set link up is not supported "
2823 				     "by device id 0x%x", hw->device_id);
2824 			return -ENOTSUP;
2825 		}
2826 #endif
2827 	}
2828 
2829 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2830 		/* Turn on the copper */
2831 		ixgbe_set_phy_power(hw, true);
2832 	} else {
2833 		/* Turn on the laser */
2834 		ixgbe_enable_tx_laser(hw);
2835 	}
2836 
2837 	return 0;
2838 }
2839 
2840 /*
2841  * Set device link down: disable tx.
2842  */
2843 static int
2844 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2845 {
2846 	struct ixgbe_hw *hw =
2847 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2848 	if (hw->mac.type == ixgbe_mac_82599EB) {
2849 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2850 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2851 			/* Not suported in bypass mode */
2852 			PMD_INIT_LOG(ERR, "Set link down is not supported "
2853 				     "by device id 0x%x", hw->device_id);
2854 			return -ENOTSUP;
2855 		}
2856 #endif
2857 	}
2858 
2859 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2860 		/* Turn off the copper */
2861 		ixgbe_set_phy_power(hw, false);
2862 	} else {
2863 		/* Turn off the laser */
2864 		ixgbe_disable_tx_laser(hw);
2865 	}
2866 
2867 	return 0;
2868 }
2869 
2870 /*
2871  * Reset and stop device.
2872  */
2873 static void
2874 ixgbe_dev_close(struct rte_eth_dev *dev)
2875 {
2876 	struct ixgbe_hw *hw =
2877 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2878 
2879 	PMD_INIT_FUNC_TRACE();
2880 
2881 	ixgbe_pf_reset_hw(hw);
2882 
2883 	ixgbe_dev_stop(dev);
2884 	hw->adapter_stopped = 1;
2885 
2886 	ixgbe_dev_free_queues(dev);
2887 
2888 	ixgbe_disable_pcie_master(hw);
2889 
2890 	/* reprogram the RAR[0] in case user changed it. */
2891 	ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
2892 }
2893 
2894 /*
2895  * Reset PF device.
2896  */
2897 static int
2898 ixgbe_dev_reset(struct rte_eth_dev *dev)
2899 {
2900 	int ret;
2901 
2902 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
2903 	 * its VF to make them align with it. The detailed notification
2904 	 * mechanism is PMD specific. As to ixgbe PF, it is rather complex.
2905 	 * To avoid unexpected behavior in VF, currently reset of PF with
2906 	 * SR-IOV activation is not supported. It might be supported later.
2907 	 */
2908 	if (dev->data->sriov.active)
2909 		return -ENOTSUP;
2910 
2911 	ret = eth_ixgbe_dev_uninit(dev);
2912 	if (ret)
2913 		return ret;
2914 
2915 	ret = eth_ixgbe_dev_init(dev);
2916 
2917 	return ret;
2918 }
2919 
2920 static void
2921 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
2922 			   struct ixgbe_hw_stats *hw_stats,
2923 			   struct ixgbe_macsec_stats *macsec_stats,
2924 			   uint64_t *total_missed_rx, uint64_t *total_qbrc,
2925 			   uint64_t *total_qprc, uint64_t *total_qprdc)
2926 {
2927 	uint32_t bprc, lxon, lxoff, total;
2928 	uint32_t delta_gprc = 0;
2929 	unsigned i;
2930 	/* Workaround for RX byte count not including CRC bytes when CRC
2931 	 * strip is enabled. CRC bytes are removed from counters when crc_strip
2932 	 * is disabled.
2933 	 */
2934 	int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
2935 			IXGBE_HLREG0_RXCRCSTRP);
2936 
2937 	hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
2938 	hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
2939 	hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
2940 	hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
2941 
2942 	for (i = 0; i < 8; i++) {
2943 		uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
2944 
2945 		/* global total per queue */
2946 		hw_stats->mpc[i] += mp;
2947 		/* Running comprehensive total for stats display */
2948 		*total_missed_rx += hw_stats->mpc[i];
2949 		if (hw->mac.type == ixgbe_mac_82598EB) {
2950 			hw_stats->rnbc[i] +=
2951 			    IXGBE_READ_REG(hw, IXGBE_RNBC(i));
2952 			hw_stats->pxonrxc[i] +=
2953 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
2954 			hw_stats->pxoffrxc[i] +=
2955 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
2956 		} else {
2957 			hw_stats->pxonrxc[i] +=
2958 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
2959 			hw_stats->pxoffrxc[i] +=
2960 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
2961 			hw_stats->pxon2offc[i] +=
2962 				IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
2963 		}
2964 		hw_stats->pxontxc[i] +=
2965 		    IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
2966 		hw_stats->pxofftxc[i] +=
2967 		    IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
2968 	}
2969 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
2970 		uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
2971 		uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
2972 		uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
2973 
2974 		delta_gprc += delta_qprc;
2975 
2976 		hw_stats->qprc[i] += delta_qprc;
2977 		hw_stats->qptc[i] += delta_qptc;
2978 
2979 		hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
2980 		hw_stats->qbrc[i] +=
2981 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
2982 		if (crc_strip == 0)
2983 			hw_stats->qbrc[i] -= delta_qprc * ETHER_CRC_LEN;
2984 
2985 		hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
2986 		hw_stats->qbtc[i] +=
2987 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
2988 
2989 		hw_stats->qprdc[i] += delta_qprdc;
2990 		*total_qprdc += hw_stats->qprdc[i];
2991 
2992 		*total_qprc += hw_stats->qprc[i];
2993 		*total_qbrc += hw_stats->qbrc[i];
2994 	}
2995 	hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
2996 	hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
2997 	hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
2998 
2999 	/*
3000 	 * An errata states that gprc actually counts good + missed packets:
3001 	 * Workaround to set gprc to summated queue packet receives
3002 	 */
3003 	hw_stats->gprc = *total_qprc;
3004 
3005 	if (hw->mac.type != ixgbe_mac_82598EB) {
3006 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
3007 		hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
3008 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
3009 		hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
3010 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
3011 		hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
3012 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
3013 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
3014 	} else {
3015 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
3016 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
3017 		/* 82598 only has a counter in the high register */
3018 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
3019 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
3020 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
3021 	}
3022 	uint64_t old_tpr = hw_stats->tpr;
3023 
3024 	hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
3025 	hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
3026 
3027 	if (crc_strip == 0)
3028 		hw_stats->gorc -= delta_gprc * ETHER_CRC_LEN;
3029 
3030 	uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
3031 	hw_stats->gptc += delta_gptc;
3032 	hw_stats->gotc -= delta_gptc * ETHER_CRC_LEN;
3033 	hw_stats->tor -= (hw_stats->tpr - old_tpr) * ETHER_CRC_LEN;
3034 
3035 	/*
3036 	 * Workaround: mprc hardware is incorrectly counting
3037 	 * broadcasts, so for now we subtract those.
3038 	 */
3039 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
3040 	hw_stats->bprc += bprc;
3041 	hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
3042 	if (hw->mac.type == ixgbe_mac_82598EB)
3043 		hw_stats->mprc -= bprc;
3044 
3045 	hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
3046 	hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
3047 	hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
3048 	hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
3049 	hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
3050 	hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
3051 
3052 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
3053 	hw_stats->lxontxc += lxon;
3054 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
3055 	hw_stats->lxofftxc += lxoff;
3056 	total = lxon + lxoff;
3057 
3058 	hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
3059 	hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
3060 	hw_stats->gptc -= total;
3061 	hw_stats->mptc -= total;
3062 	hw_stats->ptc64 -= total;
3063 	hw_stats->gotc -= total * ETHER_MIN_LEN;
3064 
3065 	hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3066 	hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3067 	hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3068 	hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
3069 	hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
3070 	hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
3071 	hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
3072 	hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3073 	hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3074 	hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3075 	hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3076 	hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
3077 	hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3078 	hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
3079 	hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
3080 	hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
3081 	/* Only read FCOE on 82599 */
3082 	if (hw->mac.type != ixgbe_mac_82598EB) {
3083 		hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
3084 		hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
3085 		hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
3086 		hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
3087 		hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
3088 	}
3089 
3090 	/* Flow Director Stats registers */
3091 	hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
3092 	hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
3093 
3094 	/* MACsec Stats registers */
3095 	macsec_stats->out_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECTXUT);
3096 	macsec_stats->out_pkts_encrypted +=
3097 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTE);
3098 	macsec_stats->out_pkts_protected +=
3099 		IXGBE_READ_REG(hw, IXGBE_LSECTXPKTP);
3100 	macsec_stats->out_octets_encrypted +=
3101 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTE);
3102 	macsec_stats->out_octets_protected +=
3103 		IXGBE_READ_REG(hw, IXGBE_LSECTXOCTP);
3104 	macsec_stats->in_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECRXUT);
3105 	macsec_stats->in_pkts_badtag += IXGBE_READ_REG(hw, IXGBE_LSECRXBAD);
3106 	macsec_stats->in_pkts_nosci += IXGBE_READ_REG(hw, IXGBE_LSECRXNOSCI);
3107 	macsec_stats->in_pkts_unknownsci +=
3108 		IXGBE_READ_REG(hw, IXGBE_LSECRXUNSCI);
3109 	macsec_stats->in_octets_decrypted +=
3110 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTD);
3111 	macsec_stats->in_octets_validated +=
3112 		IXGBE_READ_REG(hw, IXGBE_LSECRXOCTV);
3113 	macsec_stats->in_pkts_unchecked += IXGBE_READ_REG(hw, IXGBE_LSECRXUNCH);
3114 	macsec_stats->in_pkts_delayed += IXGBE_READ_REG(hw, IXGBE_LSECRXDELAY);
3115 	macsec_stats->in_pkts_late += IXGBE_READ_REG(hw, IXGBE_LSECRXLATE);
3116 	for (i = 0; i < 2; i++) {
3117 		macsec_stats->in_pkts_ok +=
3118 			IXGBE_READ_REG(hw, IXGBE_LSECRXOK(i));
3119 		macsec_stats->in_pkts_invalid +=
3120 			IXGBE_READ_REG(hw, IXGBE_LSECRXINV(i));
3121 		macsec_stats->in_pkts_notvalid +=
3122 			IXGBE_READ_REG(hw, IXGBE_LSECRXNV(i));
3123 	}
3124 	macsec_stats->in_pkts_unusedsa += IXGBE_READ_REG(hw, IXGBE_LSECRXUNSA);
3125 	macsec_stats->in_pkts_notusingsa +=
3126 		IXGBE_READ_REG(hw, IXGBE_LSECRXNUSA);
3127 }
3128 
3129 /*
3130  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
3131  */
3132 static int
3133 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3134 {
3135 	struct ixgbe_hw *hw =
3136 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3137 	struct ixgbe_hw_stats *hw_stats =
3138 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3139 	struct ixgbe_macsec_stats *macsec_stats =
3140 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3141 				dev->data->dev_private);
3142 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3143 	unsigned i;
3144 
3145 	total_missed_rx = 0;
3146 	total_qbrc = 0;
3147 	total_qprc = 0;
3148 	total_qprdc = 0;
3149 
3150 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3151 			&total_qbrc, &total_qprc, &total_qprdc);
3152 
3153 	if (stats == NULL)
3154 		return -EINVAL;
3155 
3156 	/* Fill out the rte_eth_stats statistics structure */
3157 	stats->ipackets = total_qprc;
3158 	stats->ibytes = total_qbrc;
3159 	stats->opackets = hw_stats->gptc;
3160 	stats->obytes = hw_stats->gotc;
3161 
3162 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3163 		stats->q_ipackets[i] = hw_stats->qprc[i];
3164 		stats->q_opackets[i] = hw_stats->qptc[i];
3165 		stats->q_ibytes[i] = hw_stats->qbrc[i];
3166 		stats->q_obytes[i] = hw_stats->qbtc[i];
3167 		stats->q_errors[i] = hw_stats->qprdc[i];
3168 	}
3169 
3170 	/* Rx Errors */
3171 	stats->imissed  = total_missed_rx;
3172 	stats->ierrors  = hw_stats->crcerrs +
3173 			  hw_stats->mspdc +
3174 			  hw_stats->rlec +
3175 			  hw_stats->ruc +
3176 			  hw_stats->roc +
3177 			  hw_stats->illerrc +
3178 			  hw_stats->errbc +
3179 			  hw_stats->rfc +
3180 			  hw_stats->fccrc +
3181 			  hw_stats->fclast;
3182 
3183 	/* Tx Errors */
3184 	stats->oerrors  = 0;
3185 	return 0;
3186 }
3187 
3188 static void
3189 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
3190 {
3191 	struct ixgbe_hw_stats *stats =
3192 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3193 
3194 	/* HW registers are cleared on read */
3195 	ixgbe_dev_stats_get(dev, NULL);
3196 
3197 	/* Reset software totals */
3198 	memset(stats, 0, sizeof(*stats));
3199 }
3200 
3201 /* This function calculates the number of xstats based on the current config */
3202 static unsigned
3203 ixgbe_xstats_calc_num(void) {
3204 	return IXGBE_NB_HW_STATS + IXGBE_NB_MACSEC_STATS +
3205 		(IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
3206 		(IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
3207 }
3208 
3209 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3210 	struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned int size)
3211 {
3212 	const unsigned cnt_stats = ixgbe_xstats_calc_num();
3213 	unsigned stat, i, count;
3214 
3215 	if (xstats_names != NULL) {
3216 		count = 0;
3217 
3218 		/* Note: limit >= cnt_stats checked upstream
3219 		 * in rte_eth_xstats_names()
3220 		 */
3221 
3222 		/* Extended stats from ixgbe_hw_stats */
3223 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3224 			snprintf(xstats_names[count].name,
3225 				sizeof(xstats_names[count].name),
3226 				"%s",
3227 				rte_ixgbe_stats_strings[i].name);
3228 			count++;
3229 		}
3230 
3231 		/* MACsec Stats */
3232 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3233 			snprintf(xstats_names[count].name,
3234 				sizeof(xstats_names[count].name),
3235 				"%s",
3236 				rte_ixgbe_macsec_strings[i].name);
3237 			count++;
3238 		}
3239 
3240 		/* RX Priority Stats */
3241 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3242 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3243 				snprintf(xstats_names[count].name,
3244 					sizeof(xstats_names[count].name),
3245 					"rx_priority%u_%s", i,
3246 					rte_ixgbe_rxq_strings[stat].name);
3247 				count++;
3248 			}
3249 		}
3250 
3251 		/* TX Priority Stats */
3252 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3253 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3254 				snprintf(xstats_names[count].name,
3255 					sizeof(xstats_names[count].name),
3256 					"tx_priority%u_%s", i,
3257 					rte_ixgbe_txq_strings[stat].name);
3258 				count++;
3259 			}
3260 		}
3261 	}
3262 	return cnt_stats;
3263 }
3264 
3265 static int ixgbe_dev_xstats_get_names_by_id(
3266 	struct rte_eth_dev *dev,
3267 	struct rte_eth_xstat_name *xstats_names,
3268 	const uint64_t *ids,
3269 	unsigned int limit)
3270 {
3271 	if (!ids) {
3272 		const unsigned int cnt_stats = ixgbe_xstats_calc_num();
3273 		unsigned int stat, i, count;
3274 
3275 		if (xstats_names != NULL) {
3276 			count = 0;
3277 
3278 			/* Note: limit >= cnt_stats checked upstream
3279 			 * in rte_eth_xstats_names()
3280 			 */
3281 
3282 			/* Extended stats from ixgbe_hw_stats */
3283 			for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3284 				snprintf(xstats_names[count].name,
3285 					sizeof(xstats_names[count].name),
3286 					"%s",
3287 					rte_ixgbe_stats_strings[i].name);
3288 				count++;
3289 			}
3290 
3291 			/* MACsec Stats */
3292 			for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3293 				snprintf(xstats_names[count].name,
3294 					sizeof(xstats_names[count].name),
3295 					"%s",
3296 					rte_ixgbe_macsec_strings[i].name);
3297 				count++;
3298 			}
3299 
3300 			/* RX Priority Stats */
3301 			for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3302 				for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3303 					snprintf(xstats_names[count].name,
3304 					    sizeof(xstats_names[count].name),
3305 					    "rx_priority%u_%s", i,
3306 					    rte_ixgbe_rxq_strings[stat].name);
3307 					count++;
3308 				}
3309 			}
3310 
3311 			/* TX Priority Stats */
3312 			for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3313 				for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3314 					snprintf(xstats_names[count].name,
3315 					    sizeof(xstats_names[count].name),
3316 					    "tx_priority%u_%s", i,
3317 					    rte_ixgbe_txq_strings[stat].name);
3318 					count++;
3319 				}
3320 			}
3321 		}
3322 		return cnt_stats;
3323 	}
3324 
3325 	uint16_t i;
3326 	uint16_t size = ixgbe_xstats_calc_num();
3327 	struct rte_eth_xstat_name xstats_names_copy[size];
3328 
3329 	ixgbe_dev_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
3330 			size);
3331 
3332 	for (i = 0; i < limit; i++) {
3333 		if (ids[i] >= size) {
3334 			PMD_INIT_LOG(ERR, "id value isn't valid");
3335 			return -1;
3336 		}
3337 		strcpy(xstats_names[i].name,
3338 				xstats_names_copy[ids[i]].name);
3339 	}
3340 	return limit;
3341 }
3342 
3343 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3344 	struct rte_eth_xstat_name *xstats_names, unsigned limit)
3345 {
3346 	unsigned i;
3347 
3348 	if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
3349 		return -ENOMEM;
3350 
3351 	if (xstats_names != NULL)
3352 		for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
3353 			snprintf(xstats_names[i].name,
3354 				sizeof(xstats_names[i].name),
3355 				"%s", rte_ixgbevf_stats_strings[i].name);
3356 	return IXGBEVF_NB_XSTATS;
3357 }
3358 
3359 static int
3360 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3361 					 unsigned n)
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, stat, count = 0;
3372 
3373 	count = ixgbe_xstats_calc_num();
3374 
3375 	if (n < count)
3376 		return count;
3377 
3378 	total_missed_rx = 0;
3379 	total_qbrc = 0;
3380 	total_qprc = 0;
3381 	total_qprdc = 0;
3382 
3383 	ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3384 			&total_qbrc, &total_qprc, &total_qprdc);
3385 
3386 	/* If this is a reset xstats is NULL, and we have cleared the
3387 	 * registers by reading them.
3388 	 */
3389 	if (!xstats)
3390 		return 0;
3391 
3392 	/* Extended stats from ixgbe_hw_stats */
3393 	count = 0;
3394 	for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3395 		xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3396 				rte_ixgbe_stats_strings[i].offset);
3397 		xstats[count].id = count;
3398 		count++;
3399 	}
3400 
3401 	/* MACsec Stats */
3402 	for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3403 		xstats[count].value = *(uint64_t *)(((char *)macsec_stats) +
3404 				rte_ixgbe_macsec_strings[i].offset);
3405 		xstats[count].id = count;
3406 		count++;
3407 	}
3408 
3409 	/* RX Priority Stats */
3410 	for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3411 		for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3412 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3413 					rte_ixgbe_rxq_strings[stat].offset +
3414 					(sizeof(uint64_t) * i));
3415 			xstats[count].id = count;
3416 			count++;
3417 		}
3418 	}
3419 
3420 	/* TX Priority Stats */
3421 	for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3422 		for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3423 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3424 					rte_ixgbe_txq_strings[stat].offset +
3425 					(sizeof(uint64_t) * i));
3426 			xstats[count].id = count;
3427 			count++;
3428 		}
3429 	}
3430 	return count;
3431 }
3432 
3433 static int
3434 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
3435 		uint64_t *values, unsigned int n)
3436 {
3437 	if (!ids) {
3438 		struct ixgbe_hw *hw =
3439 				IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3440 		struct ixgbe_hw_stats *hw_stats =
3441 				IXGBE_DEV_PRIVATE_TO_STATS(
3442 						dev->data->dev_private);
3443 		struct ixgbe_macsec_stats *macsec_stats =
3444 				IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3445 					dev->data->dev_private);
3446 		uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3447 		unsigned int i, stat, count = 0;
3448 
3449 		count = ixgbe_xstats_calc_num();
3450 
3451 		if (!ids && n < count)
3452 			return count;
3453 
3454 		total_missed_rx = 0;
3455 		total_qbrc = 0;
3456 		total_qprc = 0;
3457 		total_qprdc = 0;
3458 
3459 		ixgbe_read_stats_registers(hw, hw_stats, macsec_stats,
3460 				&total_missed_rx, &total_qbrc, &total_qprc,
3461 				&total_qprdc);
3462 
3463 		/* If this is a reset xstats is NULL, and we have cleared the
3464 		 * registers by reading them.
3465 		 */
3466 		if (!ids && !values)
3467 			return 0;
3468 
3469 		/* Extended stats from ixgbe_hw_stats */
3470 		count = 0;
3471 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3472 			values[count] = *(uint64_t *)(((char *)hw_stats) +
3473 					rte_ixgbe_stats_strings[i].offset);
3474 			count++;
3475 		}
3476 
3477 		/* MACsec Stats */
3478 		for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3479 			values[count] = *(uint64_t *)(((char *)macsec_stats) +
3480 					rte_ixgbe_macsec_strings[i].offset);
3481 			count++;
3482 		}
3483 
3484 		/* RX Priority Stats */
3485 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3486 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3487 				values[count] =
3488 					*(uint64_t *)(((char *)hw_stats) +
3489 					rte_ixgbe_rxq_strings[stat].offset +
3490 					(sizeof(uint64_t) * i));
3491 				count++;
3492 			}
3493 		}
3494 
3495 		/* TX Priority Stats */
3496 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3497 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3498 				values[count] =
3499 					*(uint64_t *)(((char *)hw_stats) +
3500 					rte_ixgbe_txq_strings[stat].offset +
3501 					(sizeof(uint64_t) * i));
3502 				count++;
3503 			}
3504 		}
3505 		return count;
3506 	}
3507 
3508 	uint16_t i;
3509 	uint16_t size = ixgbe_xstats_calc_num();
3510 	uint64_t values_copy[size];
3511 
3512 	ixgbe_dev_xstats_get_by_id(dev, NULL, values_copy, size);
3513 
3514 	for (i = 0; i < n; i++) {
3515 		if (ids[i] >= size) {
3516 			PMD_INIT_LOG(ERR, "id value isn't valid");
3517 			return -1;
3518 		}
3519 		values[i] = values_copy[ids[i]];
3520 	}
3521 	return n;
3522 }
3523 
3524 static void
3525 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
3526 {
3527 	struct ixgbe_hw_stats *stats =
3528 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3529 	struct ixgbe_macsec_stats *macsec_stats =
3530 			IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3531 				dev->data->dev_private);
3532 
3533 	unsigned count = ixgbe_xstats_calc_num();
3534 
3535 	/* HW registers are cleared on read */
3536 	ixgbe_dev_xstats_get(dev, NULL, count);
3537 
3538 	/* Reset software totals */
3539 	memset(stats, 0, sizeof(*stats));
3540 	memset(macsec_stats, 0, sizeof(*macsec_stats));
3541 }
3542 
3543 static void
3544 ixgbevf_update_stats(struct rte_eth_dev *dev)
3545 {
3546 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3547 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3548 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3549 
3550 	/* Good Rx packet, include VF loopback */
3551 	UPDATE_VF_STAT(IXGBE_VFGPRC,
3552 	    hw_stats->last_vfgprc, hw_stats->vfgprc);
3553 
3554 	/* Good Rx octets, include VF loopback */
3555 	UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
3556 	    hw_stats->last_vfgorc, hw_stats->vfgorc);
3557 
3558 	/* Good Tx packet, include VF loopback */
3559 	UPDATE_VF_STAT(IXGBE_VFGPTC,
3560 	    hw_stats->last_vfgptc, hw_stats->vfgptc);
3561 
3562 	/* Good Tx octets, include VF loopback */
3563 	UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
3564 	    hw_stats->last_vfgotc, hw_stats->vfgotc);
3565 
3566 	/* Rx Multicst Packet */
3567 	UPDATE_VF_STAT(IXGBE_VFMPRC,
3568 	    hw_stats->last_vfmprc, hw_stats->vfmprc);
3569 }
3570 
3571 static int
3572 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3573 		       unsigned n)
3574 {
3575 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3576 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3577 	unsigned i;
3578 
3579 	if (n < IXGBEVF_NB_XSTATS)
3580 		return IXGBEVF_NB_XSTATS;
3581 
3582 	ixgbevf_update_stats(dev);
3583 
3584 	if (!xstats)
3585 		return 0;
3586 
3587 	/* Extended stats */
3588 	for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
3589 		xstats[i].id = i;
3590 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
3591 			rte_ixgbevf_stats_strings[i].offset);
3592 	}
3593 
3594 	return IXGBEVF_NB_XSTATS;
3595 }
3596 
3597 static int
3598 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3599 {
3600 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3601 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3602 
3603 	ixgbevf_update_stats(dev);
3604 
3605 	if (stats == NULL)
3606 		return -EINVAL;
3607 
3608 	stats->ipackets = hw_stats->vfgprc;
3609 	stats->ibytes = hw_stats->vfgorc;
3610 	stats->opackets = hw_stats->vfgptc;
3611 	stats->obytes = hw_stats->vfgotc;
3612 	return 0;
3613 }
3614 
3615 static void
3616 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3617 {
3618 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3619 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3620 
3621 	/* Sync HW register to the last stats */
3622 	ixgbevf_dev_stats_get(dev, NULL);
3623 
3624 	/* reset HW current stats*/
3625 	hw_stats->vfgprc = 0;
3626 	hw_stats->vfgorc = 0;
3627 	hw_stats->vfgptc = 0;
3628 	hw_stats->vfgotc = 0;
3629 }
3630 
3631 static int
3632 ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
3633 {
3634 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3635 	u16 eeprom_verh, eeprom_verl;
3636 	u32 etrack_id;
3637 	int ret;
3638 
3639 	ixgbe_read_eeprom(hw, 0x2e, &eeprom_verh);
3640 	ixgbe_read_eeprom(hw, 0x2d, &eeprom_verl);
3641 
3642 	etrack_id = (eeprom_verh << 16) | eeprom_verl;
3643 	ret = snprintf(fw_version, fw_size, "0x%08x", etrack_id);
3644 
3645 	ret += 1; /* add the size of '\0' */
3646 	if (fw_size < (u32)ret)
3647 		return ret;
3648 	else
3649 		return 0;
3650 }
3651 
3652 static void
3653 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3654 {
3655 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3656 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3657 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3658 
3659 	dev_info->pci_dev = pci_dev;
3660 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3661 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3662 	if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3663 		/*
3664 		 * When DCB/VT is off, maximum number of queues changes,
3665 		 * except for 82598EB, which remains constant.
3666 		 */
3667 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
3668 				hw->mac.type != ixgbe_mac_82598EB)
3669 			dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3670 	}
3671 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3672 	dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3673 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3674 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3675 	dev_info->max_vfs = pci_dev->max_vfs;
3676 	if (hw->mac.type == ixgbe_mac_82598EB)
3677 		dev_info->max_vmdq_pools = ETH_16_POOLS;
3678 	else
3679 		dev_info->max_vmdq_pools = ETH_64_POOLS;
3680 	dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3681 	dev_info->rx_offload_capa =
3682 		DEV_RX_OFFLOAD_VLAN_STRIP |
3683 		DEV_RX_OFFLOAD_IPV4_CKSUM |
3684 		DEV_RX_OFFLOAD_UDP_CKSUM  |
3685 		DEV_RX_OFFLOAD_TCP_CKSUM;
3686 
3687 	/*
3688 	 * RSC is only supported by 82599 and x540 PF devices in a non-SR-IOV
3689 	 * mode.
3690 	 */
3691 	if ((hw->mac.type == ixgbe_mac_82599EB ||
3692 	     hw->mac.type == ixgbe_mac_X540) &&
3693 	    !RTE_ETH_DEV_SRIOV(dev).active)
3694 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TCP_LRO;
3695 
3696 	if (hw->mac.type == ixgbe_mac_82599EB ||
3697 	    hw->mac.type == ixgbe_mac_X540)
3698 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_MACSEC_STRIP;
3699 
3700 	if (hw->mac.type == ixgbe_mac_X550 ||
3701 	    hw->mac.type == ixgbe_mac_X550EM_x ||
3702 	    hw->mac.type == ixgbe_mac_X550EM_a)
3703 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM;
3704 
3705 	dev_info->tx_offload_capa =
3706 		DEV_TX_OFFLOAD_VLAN_INSERT |
3707 		DEV_TX_OFFLOAD_IPV4_CKSUM  |
3708 		DEV_TX_OFFLOAD_UDP_CKSUM   |
3709 		DEV_TX_OFFLOAD_TCP_CKSUM   |
3710 		DEV_TX_OFFLOAD_SCTP_CKSUM  |
3711 		DEV_TX_OFFLOAD_TCP_TSO;
3712 
3713 	if (hw->mac.type == ixgbe_mac_82599EB ||
3714 	    hw->mac.type == ixgbe_mac_X540)
3715 		dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_MACSEC_INSERT;
3716 
3717 	if (hw->mac.type == ixgbe_mac_X550 ||
3718 	    hw->mac.type == ixgbe_mac_X550EM_x ||
3719 	    hw->mac.type == ixgbe_mac_X550EM_a)
3720 		dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM;
3721 
3722 #ifdef RTE_LIBRTE_SECURITY
3723 	dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_SECURITY;
3724 	dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_SECURITY;
3725 #endif
3726 
3727 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3728 		.rx_thresh = {
3729 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3730 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3731 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3732 		},
3733 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3734 		.rx_drop_en = 0,
3735 	};
3736 
3737 	dev_info->default_txconf = (struct rte_eth_txconf) {
3738 		.tx_thresh = {
3739 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3740 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3741 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3742 		},
3743 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3744 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3745 		.txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
3746 				ETH_TXQ_FLAGS_NOOFFLOADS,
3747 	};
3748 
3749 	dev_info->rx_desc_lim = rx_desc_lim;
3750 	dev_info->tx_desc_lim = tx_desc_lim;
3751 
3752 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3753 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3754 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3755 
3756 	dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G;
3757 	if (hw->mac.type == ixgbe_mac_X540 ||
3758 	    hw->mac.type == ixgbe_mac_X540_vf ||
3759 	    hw->mac.type == ixgbe_mac_X550 ||
3760 	    hw->mac.type == ixgbe_mac_X550_vf) {
3761 		dev_info->speed_capa |= ETH_LINK_SPEED_100M;
3762 	}
3763 	if (hw->mac.type == ixgbe_mac_X550) {
3764 		dev_info->speed_capa |= ETH_LINK_SPEED_2_5G;
3765 		dev_info->speed_capa |= ETH_LINK_SPEED_5G;
3766 	}
3767 }
3768 
3769 static const uint32_t *
3770 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3771 {
3772 	static const uint32_t ptypes[] = {
3773 		/* For non-vec functions,
3774 		 * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3775 		 * for vec functions,
3776 		 * refers to _recv_raw_pkts_vec().
3777 		 */
3778 		RTE_PTYPE_L2_ETHER,
3779 		RTE_PTYPE_L3_IPV4,
3780 		RTE_PTYPE_L3_IPV4_EXT,
3781 		RTE_PTYPE_L3_IPV6,
3782 		RTE_PTYPE_L3_IPV6_EXT,
3783 		RTE_PTYPE_L4_SCTP,
3784 		RTE_PTYPE_L4_TCP,
3785 		RTE_PTYPE_L4_UDP,
3786 		RTE_PTYPE_TUNNEL_IP,
3787 		RTE_PTYPE_INNER_L3_IPV6,
3788 		RTE_PTYPE_INNER_L3_IPV6_EXT,
3789 		RTE_PTYPE_INNER_L4_TCP,
3790 		RTE_PTYPE_INNER_L4_UDP,
3791 		RTE_PTYPE_UNKNOWN
3792 	};
3793 
3794 	if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3795 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3796 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3797 	    dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3798 		return ptypes;
3799 
3800 #if defined(RTE_ARCH_X86)
3801 	if (dev->rx_pkt_burst == ixgbe_recv_pkts_vec ||
3802 	    dev->rx_pkt_burst == ixgbe_recv_scattered_pkts_vec)
3803 		return ptypes;
3804 #endif
3805 	return NULL;
3806 }
3807 
3808 static void
3809 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3810 		     struct rte_eth_dev_info *dev_info)
3811 {
3812 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3813 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3814 
3815 	dev_info->pci_dev = pci_dev;
3816 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3817 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3818 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3819 	dev_info->max_rx_pktlen = 9728; /* includes CRC, cf MAXFRS reg */
3820 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3821 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3822 	dev_info->max_vfs = pci_dev->max_vfs;
3823 	if (hw->mac.type == ixgbe_mac_82598EB)
3824 		dev_info->max_vmdq_pools = ETH_16_POOLS;
3825 	else
3826 		dev_info->max_vmdq_pools = ETH_64_POOLS;
3827 	dev_info->rx_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP |
3828 				DEV_RX_OFFLOAD_IPV4_CKSUM |
3829 				DEV_RX_OFFLOAD_UDP_CKSUM  |
3830 				DEV_RX_OFFLOAD_TCP_CKSUM;
3831 	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
3832 				DEV_TX_OFFLOAD_IPV4_CKSUM  |
3833 				DEV_TX_OFFLOAD_UDP_CKSUM   |
3834 				DEV_TX_OFFLOAD_TCP_CKSUM   |
3835 				DEV_TX_OFFLOAD_SCTP_CKSUM  |
3836 				DEV_TX_OFFLOAD_TCP_TSO;
3837 
3838 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3839 		.rx_thresh = {
3840 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3841 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3842 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3843 		},
3844 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3845 		.rx_drop_en = 0,
3846 	};
3847 
3848 	dev_info->default_txconf = (struct rte_eth_txconf) {
3849 		.tx_thresh = {
3850 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3851 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3852 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3853 		},
3854 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3855 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3856 		.txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
3857 				ETH_TXQ_FLAGS_NOOFFLOADS,
3858 	};
3859 
3860 	dev_info->rx_desc_lim = rx_desc_lim;
3861 	dev_info->tx_desc_lim = tx_desc_lim;
3862 }
3863 
3864 static int
3865 ixgbevf_check_link(struct ixgbe_hw *hw, ixgbe_link_speed *speed,
3866 		   int *link_up, int wait_to_complete)
3867 {
3868 	/**
3869 	 * for a quick link status checking, wait_to_compelet == 0,
3870 	 * skip PF link status checking
3871 	 */
3872 	bool no_pflink_check = wait_to_complete == 0;
3873 	struct ixgbe_mbx_info *mbx = &hw->mbx;
3874 	struct ixgbe_mac_info *mac = &hw->mac;
3875 	uint32_t links_reg, in_msg;
3876 	int ret_val = 0;
3877 
3878 	/* If we were hit with a reset drop the link */
3879 	if (!mbx->ops.check_for_rst(hw, 0) || !mbx->timeout)
3880 		mac->get_link_status = true;
3881 
3882 	if (!mac->get_link_status)
3883 		goto out;
3884 
3885 	/* if link status is down no point in checking to see if pf is up */
3886 	links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
3887 	if (!(links_reg & IXGBE_LINKS_UP))
3888 		goto out;
3889 
3890 	/* for SFP+ modules and DA cables on 82599 it can take up to 500usecs
3891 	 * before the link status is correct
3892 	 */
3893 	if (mac->type == ixgbe_mac_82599_vf && wait_to_complete) {
3894 		int i;
3895 
3896 		for (i = 0; i < 5; i++) {
3897 			rte_delay_us(100);
3898 			links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
3899 
3900 			if (!(links_reg & IXGBE_LINKS_UP))
3901 				goto out;
3902 		}
3903 	}
3904 
3905 	switch (links_reg & IXGBE_LINKS_SPEED_82599) {
3906 	case IXGBE_LINKS_SPEED_10G_82599:
3907 		*speed = IXGBE_LINK_SPEED_10GB_FULL;
3908 		if (hw->mac.type >= ixgbe_mac_X550) {
3909 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
3910 				*speed = IXGBE_LINK_SPEED_2_5GB_FULL;
3911 		}
3912 		break;
3913 	case IXGBE_LINKS_SPEED_1G_82599:
3914 		*speed = IXGBE_LINK_SPEED_1GB_FULL;
3915 		break;
3916 	case IXGBE_LINKS_SPEED_100_82599:
3917 		*speed = IXGBE_LINK_SPEED_100_FULL;
3918 		if (hw->mac.type == ixgbe_mac_X550) {
3919 			if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
3920 				*speed = IXGBE_LINK_SPEED_5GB_FULL;
3921 		}
3922 		break;
3923 	case IXGBE_LINKS_SPEED_10_X550EM_A:
3924 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
3925 		/* Since Reserved in older MAC's */
3926 		if (hw->mac.type >= ixgbe_mac_X550)
3927 			*speed = IXGBE_LINK_SPEED_10_FULL;
3928 		break;
3929 	default:
3930 		*speed = IXGBE_LINK_SPEED_UNKNOWN;
3931 	}
3932 
3933 	if (no_pflink_check) {
3934 		if (*speed == IXGBE_LINK_SPEED_UNKNOWN)
3935 			mac->get_link_status = true;
3936 		else
3937 			mac->get_link_status = false;
3938 
3939 		goto out;
3940 	}
3941 	/* if the read failed it could just be a mailbox collision, best wait
3942 	 * until we are called again and don't report an error
3943 	 */
3944 	if (mbx->ops.read(hw, &in_msg, 1, 0))
3945 		goto out;
3946 
3947 	if (!(in_msg & IXGBE_VT_MSGTYPE_CTS)) {
3948 		/* msg is not CTS and is NACK we must have lost CTS status */
3949 		if (in_msg & IXGBE_VT_MSGTYPE_NACK)
3950 			ret_val = -1;
3951 		goto out;
3952 	}
3953 
3954 	/* the pf is talking, if we timed out in the past we reinit */
3955 	if (!mbx->timeout) {
3956 		ret_val = -1;
3957 		goto out;
3958 	}
3959 
3960 	/* if we passed all the tests above then the link is up and we no
3961 	 * longer need to check for link
3962 	 */
3963 	mac->get_link_status = false;
3964 
3965 out:
3966 	*link_up = !mac->get_link_status;
3967 	return ret_val;
3968 }
3969 
3970 /* return 0 means link status changed, -1 means not changed */
3971 static int
3972 ixgbe_dev_link_update_share(struct rte_eth_dev *dev,
3973 			    int wait_to_complete, int vf)
3974 {
3975 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3976 	struct rte_eth_link link, old;
3977 	ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
3978 	struct ixgbe_interrupt *intr =
3979 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3980 	int link_up;
3981 	int diag;
3982 	u32 speed = 0;
3983 	int wait = 1;
3984 	bool autoneg = false;
3985 
3986 	link.link_status = ETH_LINK_DOWN;
3987 	link.link_speed = 0;
3988 	link.link_duplex = ETH_LINK_HALF_DUPLEX;
3989 	link.link_autoneg = ETH_LINK_AUTONEG;
3990 	memset(&old, 0, sizeof(old));
3991 	rte_ixgbe_dev_atomic_read_link_status(dev, &old);
3992 
3993 	hw->mac.get_link_status = true;
3994 
3995 	if ((intr->flags & IXGBE_FLAG_NEED_LINK_CONFIG) &&
3996 		ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
3997 		speed = hw->phy.autoneg_advertised;
3998 		if (!speed)
3999 			ixgbe_get_link_capabilities(hw, &speed, &autoneg);
4000 		ixgbe_setup_link(hw, speed, true);
4001 	}
4002 
4003 	/* check if it needs to wait to complete, if lsc interrupt is enabled */
4004 	if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
4005 		wait = 0;
4006 
4007 	if (vf)
4008 		diag = ixgbevf_check_link(hw, &link_speed, &link_up, wait);
4009 	else
4010 		diag = ixgbe_check_link(hw, &link_speed, &link_up, wait);
4011 
4012 	if (diag != 0) {
4013 		link.link_speed = ETH_SPEED_NUM_100M;
4014 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
4015 		rte_ixgbe_dev_atomic_write_link_status(dev, &link);
4016 		if (link.link_status == old.link_status)
4017 			return -1;
4018 		return 0;
4019 	}
4020 
4021 	if (link_up == 0) {
4022 		rte_ixgbe_dev_atomic_write_link_status(dev, &link);
4023 		intr->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
4024 		if (link.link_status == old.link_status)
4025 			return -1;
4026 		return 0;
4027 	}
4028 	intr->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4029 	link.link_status = ETH_LINK_UP;
4030 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
4031 
4032 	switch (link_speed) {
4033 	default:
4034 	case IXGBE_LINK_SPEED_UNKNOWN:
4035 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
4036 		link.link_speed = ETH_SPEED_NUM_100M;
4037 		break;
4038 
4039 	case IXGBE_LINK_SPEED_100_FULL:
4040 		link.link_speed = ETH_SPEED_NUM_100M;
4041 		break;
4042 
4043 	case IXGBE_LINK_SPEED_1GB_FULL:
4044 		link.link_speed = ETH_SPEED_NUM_1G;
4045 		break;
4046 
4047 	case IXGBE_LINK_SPEED_2_5GB_FULL:
4048 		link.link_speed = ETH_SPEED_NUM_2_5G;
4049 		break;
4050 
4051 	case IXGBE_LINK_SPEED_5GB_FULL:
4052 		link.link_speed = ETH_SPEED_NUM_5G;
4053 		break;
4054 
4055 	case IXGBE_LINK_SPEED_10GB_FULL:
4056 		link.link_speed = ETH_SPEED_NUM_10G;
4057 		break;
4058 	}
4059 	rte_ixgbe_dev_atomic_write_link_status(dev, &link);
4060 
4061 	if (link.link_status == old.link_status)
4062 		return -1;
4063 
4064 	return 0;
4065 }
4066 
4067 static int
4068 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4069 {
4070 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 0);
4071 }
4072 
4073 static int
4074 ixgbevf_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4075 {
4076 	return ixgbe_dev_link_update_share(dev, wait_to_complete, 1);
4077 }
4078 
4079 static void
4080 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
4081 {
4082 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4083 	uint32_t fctrl;
4084 
4085 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4086 	fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4087 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4088 }
4089 
4090 static void
4091 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
4092 {
4093 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4094 	uint32_t fctrl;
4095 
4096 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4097 	fctrl &= (~IXGBE_FCTRL_UPE);
4098 	if (dev->data->all_multicast == 1)
4099 		fctrl |= IXGBE_FCTRL_MPE;
4100 	else
4101 		fctrl &= (~IXGBE_FCTRL_MPE);
4102 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4103 }
4104 
4105 static void
4106 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
4107 {
4108 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4109 	uint32_t fctrl;
4110 
4111 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4112 	fctrl |= IXGBE_FCTRL_MPE;
4113 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4114 }
4115 
4116 static void
4117 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
4118 {
4119 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4120 	uint32_t fctrl;
4121 
4122 	if (dev->data->promiscuous == 1)
4123 		return; /* must remain in all_multicast mode */
4124 
4125 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4126 	fctrl &= (~IXGBE_FCTRL_MPE);
4127 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4128 }
4129 
4130 /**
4131  * It clears the interrupt causes and enables the interrupt.
4132  * It will be called once only during nic initialized.
4133  *
4134  * @param dev
4135  *  Pointer to struct rte_eth_dev.
4136  * @param on
4137  *  Enable or Disable.
4138  *
4139  * @return
4140  *  - On success, zero.
4141  *  - On failure, a negative value.
4142  */
4143 static int
4144 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
4145 {
4146 	struct ixgbe_interrupt *intr =
4147 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4148 
4149 	ixgbe_dev_link_status_print(dev);
4150 	if (on)
4151 		intr->mask |= IXGBE_EICR_LSC;
4152 	else
4153 		intr->mask &= ~IXGBE_EICR_LSC;
4154 
4155 	return 0;
4156 }
4157 
4158 /**
4159  * It clears the interrupt causes and enables the interrupt.
4160  * It will be called once only during nic initialized.
4161  *
4162  * @param dev
4163  *  Pointer to struct rte_eth_dev.
4164  *
4165  * @return
4166  *  - On success, zero.
4167  *  - On failure, a negative value.
4168  */
4169 static int
4170 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
4171 {
4172 	struct ixgbe_interrupt *intr =
4173 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4174 
4175 	intr->mask |= IXGBE_EICR_RTX_QUEUE;
4176 
4177 	return 0;
4178 }
4179 
4180 /**
4181  * It clears the interrupt causes and enables the interrupt.
4182  * It will be called once only during nic initialized.
4183  *
4184  * @param dev
4185  *  Pointer to struct rte_eth_dev.
4186  *
4187  * @return
4188  *  - On success, zero.
4189  *  - On failure, a negative value.
4190  */
4191 static int
4192 ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev)
4193 {
4194 	struct ixgbe_interrupt *intr =
4195 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4196 
4197 	intr->mask |= IXGBE_EICR_LINKSEC;
4198 
4199 	return 0;
4200 }
4201 
4202 /*
4203  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
4204  *
4205  * @param dev
4206  *  Pointer to struct rte_eth_dev.
4207  *
4208  * @return
4209  *  - On success, zero.
4210  *  - On failure, a negative value.
4211  */
4212 static int
4213 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
4214 {
4215 	uint32_t eicr;
4216 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4217 	struct ixgbe_interrupt *intr =
4218 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4219 
4220 	/* clear all cause mask */
4221 	ixgbe_disable_intr(hw);
4222 
4223 	/* read-on-clear nic registers here */
4224 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4225 	PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
4226 
4227 	intr->flags = 0;
4228 
4229 	/* set flag for async link update */
4230 	if (eicr & IXGBE_EICR_LSC)
4231 		intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4232 
4233 	if (eicr & IXGBE_EICR_MAILBOX)
4234 		intr->flags |= IXGBE_FLAG_MAILBOX;
4235 
4236 	if (eicr & IXGBE_EICR_LINKSEC)
4237 		intr->flags |= IXGBE_FLAG_MACSEC;
4238 
4239 	if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
4240 	    hw->phy.type == ixgbe_phy_x550em_ext_t &&
4241 	    (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
4242 		intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
4243 
4244 	return 0;
4245 }
4246 
4247 /**
4248  * It gets and then prints the link status.
4249  *
4250  * @param dev
4251  *  Pointer to struct rte_eth_dev.
4252  *
4253  * @return
4254  *  - On success, zero.
4255  *  - On failure, a negative value.
4256  */
4257 static void
4258 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
4259 {
4260 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4261 	struct rte_eth_link link;
4262 
4263 	memset(&link, 0, sizeof(link));
4264 	rte_ixgbe_dev_atomic_read_link_status(dev, &link);
4265 	if (link.link_status) {
4266 		PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
4267 					(int)(dev->data->port_id),
4268 					(unsigned)link.link_speed,
4269 			link.link_duplex == ETH_LINK_FULL_DUPLEX ?
4270 					"full-duplex" : "half-duplex");
4271 	} else {
4272 		PMD_INIT_LOG(INFO, " Port %d: Link Down",
4273 				(int)(dev->data->port_id));
4274 	}
4275 	PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
4276 				pci_dev->addr.domain,
4277 				pci_dev->addr.bus,
4278 				pci_dev->addr.devid,
4279 				pci_dev->addr.function);
4280 }
4281 
4282 /*
4283  * It executes link_update after knowing an interrupt occurred.
4284  *
4285  * @param dev
4286  *  Pointer to struct rte_eth_dev.
4287  *
4288  * @return
4289  *  - On success, zero.
4290  *  - On failure, a negative value.
4291  */
4292 static int
4293 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev,
4294 			   struct rte_intr_handle *intr_handle)
4295 {
4296 	struct ixgbe_interrupt *intr =
4297 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4298 	int64_t timeout;
4299 	struct rte_eth_link link;
4300 	struct ixgbe_hw *hw =
4301 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4302 
4303 	PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
4304 
4305 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
4306 		ixgbe_pf_mbx_process(dev);
4307 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
4308 	}
4309 
4310 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4311 		ixgbe_handle_lasi(hw);
4312 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4313 	}
4314 
4315 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4316 		/* get the link status before link update, for predicting later */
4317 		memset(&link, 0, sizeof(link));
4318 		rte_ixgbe_dev_atomic_read_link_status(dev, &link);
4319 
4320 		ixgbe_dev_link_update(dev, 0);
4321 
4322 		/* likely to up */
4323 		if (!link.link_status)
4324 			/* handle it 1 sec later, wait it being stable */
4325 			timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
4326 		/* likely to down */
4327 		else
4328 			/* handle it 4 sec later, wait it being stable */
4329 			timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
4330 
4331 		ixgbe_dev_link_status_print(dev);
4332 		if (rte_eal_alarm_set(timeout * 1000,
4333 				      ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
4334 			PMD_DRV_LOG(ERR, "Error setting alarm");
4335 		else {
4336 			/* remember original mask */
4337 			intr->mask_original = intr->mask;
4338 			/* only disable lsc interrupt */
4339 			intr->mask &= ~IXGBE_EIMS_LSC;
4340 		}
4341 	}
4342 
4343 	PMD_DRV_LOG(DEBUG, "enable intr immediately");
4344 	ixgbe_enable_intr(dev);
4345 	rte_intr_enable(intr_handle);
4346 
4347 	return 0;
4348 }
4349 
4350 /**
4351  * Interrupt handler which shall be registered for alarm callback for delayed
4352  * handling specific interrupt to wait for the stable nic state. As the
4353  * NIC interrupt state is not stable for ixgbe after link is just down,
4354  * it needs to wait 4 seconds to get the stable status.
4355  *
4356  * @param handle
4357  *  Pointer to interrupt handle.
4358  * @param param
4359  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4360  *
4361  * @return
4362  *  void
4363  */
4364 static void
4365 ixgbe_dev_interrupt_delayed_handler(void *param)
4366 {
4367 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4368 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4369 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
4370 	struct ixgbe_interrupt *intr =
4371 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4372 	struct ixgbe_hw *hw =
4373 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4374 	uint32_t eicr;
4375 
4376 	ixgbe_disable_intr(hw);
4377 
4378 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4379 	if (eicr & IXGBE_EICR_MAILBOX)
4380 		ixgbe_pf_mbx_process(dev);
4381 
4382 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4383 		ixgbe_handle_lasi(hw);
4384 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4385 	}
4386 
4387 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4388 		ixgbe_dev_link_update(dev, 0);
4389 		intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4390 		ixgbe_dev_link_status_print(dev);
4391 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC,
4392 					      NULL, NULL);
4393 	}
4394 
4395 	if (intr->flags & IXGBE_FLAG_MACSEC) {
4396 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_MACSEC,
4397 					      NULL, NULL);
4398 		intr->flags &= ~IXGBE_FLAG_MACSEC;
4399 	}
4400 
4401 	/* restore original mask */
4402 	intr->mask = intr->mask_original;
4403 	intr->mask_original = 0;
4404 
4405 	PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
4406 	ixgbe_enable_intr(dev);
4407 	rte_intr_enable(intr_handle);
4408 }
4409 
4410 /**
4411  * Interrupt handler triggered by NIC  for handling
4412  * specific interrupt.
4413  *
4414  * @param handle
4415  *  Pointer to interrupt handle.
4416  * @param param
4417  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4418  *
4419  * @return
4420  *  void
4421  */
4422 static void
4423 ixgbe_dev_interrupt_handler(void *param)
4424 {
4425 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4426 
4427 	ixgbe_dev_interrupt_get_status(dev);
4428 	ixgbe_dev_interrupt_action(dev, dev->intr_handle);
4429 }
4430 
4431 static int
4432 ixgbe_dev_led_on(struct rte_eth_dev *dev)
4433 {
4434 	struct ixgbe_hw *hw;
4435 
4436 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4437 	return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4438 }
4439 
4440 static int
4441 ixgbe_dev_led_off(struct rte_eth_dev *dev)
4442 {
4443 	struct ixgbe_hw *hw;
4444 
4445 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4446 	return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4447 }
4448 
4449 static int
4450 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4451 {
4452 	struct ixgbe_hw *hw;
4453 	uint32_t mflcn_reg;
4454 	uint32_t fccfg_reg;
4455 	int rx_pause;
4456 	int tx_pause;
4457 
4458 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4459 
4460 	fc_conf->pause_time = hw->fc.pause_time;
4461 	fc_conf->high_water = hw->fc.high_water[0];
4462 	fc_conf->low_water = hw->fc.low_water[0];
4463 	fc_conf->send_xon = hw->fc.send_xon;
4464 	fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
4465 
4466 	/*
4467 	 * Return rx_pause status according to actual setting of
4468 	 * MFLCN register.
4469 	 */
4470 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4471 	if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
4472 		rx_pause = 1;
4473 	else
4474 		rx_pause = 0;
4475 
4476 	/*
4477 	 * Return tx_pause status according to actual setting of
4478 	 * FCCFG register.
4479 	 */
4480 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4481 	if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
4482 		tx_pause = 1;
4483 	else
4484 		tx_pause = 0;
4485 
4486 	if (rx_pause && tx_pause)
4487 		fc_conf->mode = RTE_FC_FULL;
4488 	else if (rx_pause)
4489 		fc_conf->mode = RTE_FC_RX_PAUSE;
4490 	else if (tx_pause)
4491 		fc_conf->mode = RTE_FC_TX_PAUSE;
4492 	else
4493 		fc_conf->mode = RTE_FC_NONE;
4494 
4495 	return 0;
4496 }
4497 
4498 static int
4499 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4500 {
4501 	struct ixgbe_hw *hw;
4502 	int err;
4503 	uint32_t rx_buf_size;
4504 	uint32_t max_high_water;
4505 	uint32_t mflcn;
4506 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4507 		ixgbe_fc_none,
4508 		ixgbe_fc_rx_pause,
4509 		ixgbe_fc_tx_pause,
4510 		ixgbe_fc_full
4511 	};
4512 
4513 	PMD_INIT_FUNC_TRACE();
4514 
4515 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4516 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
4517 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4518 
4519 	/*
4520 	 * At least reserve one Ethernet frame for watermark
4521 	 * high_water/low_water in kilo bytes for ixgbe
4522 	 */
4523 	max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4524 	if ((fc_conf->high_water > max_high_water) ||
4525 		(fc_conf->high_water < fc_conf->low_water)) {
4526 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4527 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4528 		return -EINVAL;
4529 	}
4530 
4531 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
4532 	hw->fc.pause_time     = fc_conf->pause_time;
4533 	hw->fc.high_water[0]  = fc_conf->high_water;
4534 	hw->fc.low_water[0]   = fc_conf->low_water;
4535 	hw->fc.send_xon       = fc_conf->send_xon;
4536 	hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
4537 
4538 	err = ixgbe_fc_enable(hw);
4539 
4540 	/* Not negotiated is not an error case */
4541 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED)) {
4542 
4543 		/* check if we want to forward MAC frames - driver doesn't have native
4544 		 * capability to do that, so we'll write the registers ourselves */
4545 
4546 		mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4547 
4548 		/* set or clear MFLCN.PMCF bit depending on configuration */
4549 		if (fc_conf->mac_ctrl_frame_fwd != 0)
4550 			mflcn |= IXGBE_MFLCN_PMCF;
4551 		else
4552 			mflcn &= ~IXGBE_MFLCN_PMCF;
4553 
4554 		IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
4555 		IXGBE_WRITE_FLUSH(hw);
4556 
4557 		return 0;
4558 	}
4559 
4560 	PMD_INIT_LOG(ERR, "ixgbe_fc_enable = 0x%x", err);
4561 	return -EIO;
4562 }
4563 
4564 /**
4565  *  ixgbe_pfc_enable_generic - Enable flow control
4566  *  @hw: pointer to hardware structure
4567  *  @tc_num: traffic class number
4568  *  Enable flow control according to the current settings.
4569  */
4570 static int
4571 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
4572 {
4573 	int ret_val = 0;
4574 	uint32_t mflcn_reg, fccfg_reg;
4575 	uint32_t reg;
4576 	uint32_t fcrtl, fcrth;
4577 	uint8_t i;
4578 	uint8_t nb_rx_en;
4579 
4580 	/* Validate the water mark configuration */
4581 	if (!hw->fc.pause_time) {
4582 		ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4583 		goto out;
4584 	}
4585 
4586 	/* Low water mark of zero causes XOFF floods */
4587 	if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
4588 		 /* High/Low water can not be 0 */
4589 		if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_water[tc_num])) {
4590 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4591 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4592 			goto out;
4593 		}
4594 
4595 		if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
4596 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4597 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4598 			goto out;
4599 		}
4600 	}
4601 	/* Negotiate the fc mode to use */
4602 	ixgbe_fc_autoneg(hw);
4603 
4604 	/* Disable any previous flow control settings */
4605 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4606 	mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
4607 
4608 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4609 	fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
4610 
4611 	switch (hw->fc.current_mode) {
4612 	case ixgbe_fc_none:
4613 		/*
4614 		 * If the count of enabled RX Priority Flow control >1,
4615 		 * and the TX pause can not be disabled
4616 		 */
4617 		nb_rx_en = 0;
4618 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4619 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4620 			if (reg & IXGBE_FCRTH_FCEN)
4621 				nb_rx_en++;
4622 		}
4623 		if (nb_rx_en > 1)
4624 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4625 		break;
4626 	case ixgbe_fc_rx_pause:
4627 		/*
4628 		 * Rx Flow control is enabled and Tx Flow control is
4629 		 * disabled by software override. Since there really
4630 		 * isn't a way to advertise that we are capable of RX
4631 		 * Pause ONLY, we will advertise that we support both
4632 		 * symmetric and asymmetric Rx PAUSE.  Later, we will
4633 		 * disable the adapter's ability to send PAUSE frames.
4634 		 */
4635 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4636 		/*
4637 		 * If the count of enabled RX Priority Flow control >1,
4638 		 * and the TX pause can not be disabled
4639 		 */
4640 		nb_rx_en = 0;
4641 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4642 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4643 			if (reg & IXGBE_FCRTH_FCEN)
4644 				nb_rx_en++;
4645 		}
4646 		if (nb_rx_en > 1)
4647 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4648 		break;
4649 	case ixgbe_fc_tx_pause:
4650 		/*
4651 		 * Tx Flow control is enabled, and Rx Flow control is
4652 		 * disabled by software override.
4653 		 */
4654 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4655 		break;
4656 	case ixgbe_fc_full:
4657 		/* Flow control (both Rx and Tx) is enabled by SW override. */
4658 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
4659 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4660 		break;
4661 	default:
4662 		PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
4663 		ret_val = IXGBE_ERR_CONFIG;
4664 		goto out;
4665 	}
4666 
4667 	/* Set 802.3x based flow control settings. */
4668 	mflcn_reg |= IXGBE_MFLCN_DPF;
4669 	IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
4670 	IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
4671 
4672 	/* Set up and enable Rx high/low water mark thresholds, enable XON. */
4673 	if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
4674 		hw->fc.high_water[tc_num]) {
4675 		fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
4676 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
4677 		fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
4678 	} else {
4679 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
4680 		/*
4681 		 * In order to prevent Tx hangs when the internal Tx
4682 		 * switch is enabled we must set the high water mark
4683 		 * to the maximum FCRTH value.  This allows the Tx
4684 		 * switch to function even under heavy Rx workloads.
4685 		 */
4686 		fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
4687 	}
4688 	IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
4689 
4690 	/* Configure pause time (2 TCs per register) */
4691 	reg = hw->fc.pause_time * 0x00010001;
4692 	for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
4693 		IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
4694 
4695 	/* Configure flow control refresh threshold value */
4696 	IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
4697 
4698 out:
4699 	return ret_val;
4700 }
4701 
4702 static int
4703 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
4704 {
4705 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4706 	int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
4707 
4708 	if (hw->mac.type != ixgbe_mac_82598EB) {
4709 		ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
4710 	}
4711 	return ret_val;
4712 }
4713 
4714 static int
4715 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
4716 {
4717 	int err;
4718 	uint32_t rx_buf_size;
4719 	uint32_t max_high_water;
4720 	uint8_t tc_num;
4721 	uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
4722 	struct ixgbe_hw *hw =
4723 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4724 	struct ixgbe_dcb_config *dcb_config =
4725 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
4726 
4727 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4728 		ixgbe_fc_none,
4729 		ixgbe_fc_rx_pause,
4730 		ixgbe_fc_tx_pause,
4731 		ixgbe_fc_full
4732 	};
4733 
4734 	PMD_INIT_FUNC_TRACE();
4735 
4736 	ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
4737 	tc_num = map[pfc_conf->priority];
4738 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
4739 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4740 	/*
4741 	 * At least reserve one Ethernet frame for watermark
4742 	 * high_water/low_water in kilo bytes for ixgbe
4743 	 */
4744 	max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4745 	if ((pfc_conf->fc.high_water > max_high_water) ||
4746 	    (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
4747 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4748 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4749 		return -EINVAL;
4750 	}
4751 
4752 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
4753 	hw->fc.pause_time = pfc_conf->fc.pause_time;
4754 	hw->fc.send_xon = pfc_conf->fc.send_xon;
4755 	hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
4756 	hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
4757 
4758 	err = ixgbe_dcb_pfc_enable(dev, tc_num);
4759 
4760 	/* Not negotiated is not an error case */
4761 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
4762 		return 0;
4763 
4764 	PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
4765 	return -EIO;
4766 }
4767 
4768 static int
4769 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
4770 			  struct rte_eth_rss_reta_entry64 *reta_conf,
4771 			  uint16_t reta_size)
4772 {
4773 	uint16_t i, sp_reta_size;
4774 	uint8_t j, mask;
4775 	uint32_t reta, r;
4776 	uint16_t idx, shift;
4777 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4778 	uint32_t reta_reg;
4779 
4780 	PMD_INIT_FUNC_TRACE();
4781 
4782 	if (!ixgbe_rss_update_sp(hw->mac.type)) {
4783 		PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
4784 			"NIC.");
4785 		return -ENOTSUP;
4786 	}
4787 
4788 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
4789 	if (reta_size != sp_reta_size) {
4790 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
4791 			"(%d) doesn't match the number hardware can supported "
4792 			"(%d)", reta_size, sp_reta_size);
4793 		return -EINVAL;
4794 	}
4795 
4796 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
4797 		idx = i / RTE_RETA_GROUP_SIZE;
4798 		shift = i % RTE_RETA_GROUP_SIZE;
4799 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
4800 						IXGBE_4_BIT_MASK);
4801 		if (!mask)
4802 			continue;
4803 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
4804 		if (mask == IXGBE_4_BIT_MASK)
4805 			r = 0;
4806 		else
4807 			r = IXGBE_READ_REG(hw, reta_reg);
4808 		for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
4809 			if (mask & (0x1 << j))
4810 				reta |= reta_conf[idx].reta[shift + j] <<
4811 							(CHAR_BIT * j);
4812 			else
4813 				reta |= r & (IXGBE_8_BIT_MASK <<
4814 						(CHAR_BIT * j));
4815 		}
4816 		IXGBE_WRITE_REG(hw, reta_reg, reta);
4817 	}
4818 
4819 	return 0;
4820 }
4821 
4822 static int
4823 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
4824 			 struct rte_eth_rss_reta_entry64 *reta_conf,
4825 			 uint16_t reta_size)
4826 {
4827 	uint16_t i, sp_reta_size;
4828 	uint8_t j, mask;
4829 	uint32_t reta;
4830 	uint16_t idx, shift;
4831 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4832 	uint32_t reta_reg;
4833 
4834 	PMD_INIT_FUNC_TRACE();
4835 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
4836 	if (reta_size != sp_reta_size) {
4837 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
4838 			"(%d) doesn't match the number hardware can supported "
4839 			"(%d)", reta_size, sp_reta_size);
4840 		return -EINVAL;
4841 	}
4842 
4843 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
4844 		idx = i / RTE_RETA_GROUP_SIZE;
4845 		shift = i % RTE_RETA_GROUP_SIZE;
4846 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
4847 						IXGBE_4_BIT_MASK);
4848 		if (!mask)
4849 			continue;
4850 
4851 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
4852 		reta = IXGBE_READ_REG(hw, reta_reg);
4853 		for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
4854 			if (mask & (0x1 << j))
4855 				reta_conf[idx].reta[shift + j] =
4856 					((reta >> (CHAR_BIT * j)) &
4857 						IXGBE_8_BIT_MASK);
4858 		}
4859 	}
4860 
4861 	return 0;
4862 }
4863 
4864 static int
4865 ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
4866 				uint32_t index, uint32_t pool)
4867 {
4868 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4869 	uint32_t enable_addr = 1;
4870 
4871 	return ixgbe_set_rar(hw, index, mac_addr->addr_bytes,
4872 			     pool, enable_addr);
4873 }
4874 
4875 static void
4876 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
4877 {
4878 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4879 
4880 	ixgbe_clear_rar(hw, index);
4881 }
4882 
4883 static void
4884 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
4885 {
4886 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4887 
4888 	ixgbe_remove_rar(dev, 0);
4889 
4890 	ixgbe_add_rar(dev, addr, 0, pci_dev->max_vfs);
4891 }
4892 
4893 static bool
4894 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
4895 {
4896 	if (strcmp(dev->device->driver->name, drv->driver.name))
4897 		return false;
4898 
4899 	return true;
4900 }
4901 
4902 bool
4903 is_ixgbe_supported(struct rte_eth_dev *dev)
4904 {
4905 	return is_device_supported(dev, &rte_ixgbe_pmd);
4906 }
4907 
4908 static int
4909 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4910 {
4911 	uint32_t hlreg0;
4912 	uint32_t maxfrs;
4913 	struct ixgbe_hw *hw;
4914 	struct rte_eth_dev_info dev_info;
4915 	uint32_t frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
4916 	struct rte_eth_dev_data *dev_data = dev->data;
4917 
4918 	ixgbe_dev_info_get(dev, &dev_info);
4919 
4920 	/* check that mtu is within the allowed range */
4921 	if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
4922 		return -EINVAL;
4923 
4924 	/* If device is started, refuse mtu that requires the support of
4925 	 * scattered packets when this feature has not been enabled before.
4926 	 */
4927 	if (dev_data->dev_started && !dev_data->scattered_rx &&
4928 	    (frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
4929 	     dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
4930 		PMD_INIT_LOG(ERR, "Stop port first.");
4931 		return -EINVAL;
4932 	}
4933 
4934 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4935 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4936 
4937 	/* switch to jumbo mode if needed */
4938 	if (frame_size > ETHER_MAX_LEN) {
4939 		dev->data->dev_conf.rxmode.jumbo_frame = 1;
4940 		hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4941 	} else {
4942 		dev->data->dev_conf.rxmode.jumbo_frame = 0;
4943 		hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
4944 	}
4945 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4946 
4947 	/* update max frame size */
4948 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4949 
4950 	maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
4951 	maxfrs &= 0x0000FFFF;
4952 	maxfrs |= (dev->data->dev_conf.rxmode.max_rx_pkt_len << 16);
4953 	IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
4954 
4955 	return 0;
4956 }
4957 
4958 /*
4959  * Virtual Function operations
4960  */
4961 static void
4962 ixgbevf_intr_disable(struct ixgbe_hw *hw)
4963 {
4964 	PMD_INIT_FUNC_TRACE();
4965 
4966 	/* Clear interrupt mask to stop from interrupts being generated */
4967 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
4968 
4969 	IXGBE_WRITE_FLUSH(hw);
4970 }
4971 
4972 static void
4973 ixgbevf_intr_enable(struct ixgbe_hw *hw)
4974 {
4975 	PMD_INIT_FUNC_TRACE();
4976 
4977 	/* VF enable interrupt autoclean */
4978 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
4979 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
4980 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
4981 
4982 	IXGBE_WRITE_FLUSH(hw);
4983 }
4984 
4985 static int
4986 ixgbevf_dev_configure(struct rte_eth_dev *dev)
4987 {
4988 	struct rte_eth_conf *conf = &dev->data->dev_conf;
4989 	struct ixgbe_adapter *adapter =
4990 			(struct ixgbe_adapter *)dev->data->dev_private;
4991 
4992 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
4993 		     dev->data->port_id);
4994 
4995 	/*
4996 	 * VF has no ability to enable/disable HW CRC
4997 	 * Keep the persistent behavior the same as Host PF
4998 	 */
4999 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
5000 	if (!conf->rxmode.hw_strip_crc) {
5001 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
5002 		conf->rxmode.hw_strip_crc = 1;
5003 	}
5004 #else
5005 	if (conf->rxmode.hw_strip_crc) {
5006 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
5007 		conf->rxmode.hw_strip_crc = 0;
5008 	}
5009 #endif
5010 
5011 	/*
5012 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
5013 	 * allocation or vector Rx preconditions we will reset it.
5014 	 */
5015 	adapter->rx_bulk_alloc_allowed = true;
5016 	adapter->rx_vec_allowed = true;
5017 
5018 	return 0;
5019 }
5020 
5021 static int
5022 ixgbevf_dev_start(struct rte_eth_dev *dev)
5023 {
5024 	struct ixgbe_hw *hw =
5025 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5026 	uint32_t intr_vector = 0;
5027 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5028 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5029 
5030 	int err, mask = 0;
5031 
5032 	PMD_INIT_FUNC_TRACE();
5033 
5034 	err = hw->mac.ops.reset_hw(hw);
5035 	if (err) {
5036 		PMD_INIT_LOG(ERR, "Unable to reset vf hardware (%d)", err);
5037 		return err;
5038 	}
5039 	hw->mac.get_link_status = true;
5040 
5041 	/* negotiate mailbox API version to use with the PF. */
5042 	ixgbevf_negotiate_api(hw);
5043 
5044 	ixgbevf_dev_tx_init(dev);
5045 
5046 	/* This can fail when allocating mbufs for descriptor rings */
5047 	err = ixgbevf_dev_rx_init(dev);
5048 	if (err) {
5049 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
5050 		ixgbe_dev_clear_queues(dev);
5051 		return err;
5052 	}
5053 
5054 	/* Set vfta */
5055 	ixgbevf_set_vfta_all(dev, 1);
5056 
5057 	/* Set HW strip */
5058 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
5059 		ETH_VLAN_EXTEND_MASK;
5060 	err = ixgbevf_vlan_offload_set(dev, mask);
5061 	if (err) {
5062 		PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
5063 		ixgbe_dev_clear_queues(dev);
5064 		return err;
5065 	}
5066 
5067 	ixgbevf_dev_rxtx_start(dev);
5068 
5069 	/* check and configure queue intr-vector mapping */
5070 	if (rte_intr_cap_multiple(intr_handle) &&
5071 	    dev->data->dev_conf.intr_conf.rxq) {
5072 		/* According to datasheet, only vector 0/1/2 can be used,
5073 		 * now only one vector is used for Rx queue
5074 		 */
5075 		intr_vector = 1;
5076 		if (rte_intr_efd_enable(intr_handle, intr_vector))
5077 			return -1;
5078 	}
5079 
5080 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
5081 		intr_handle->intr_vec =
5082 			rte_zmalloc("intr_vec",
5083 				    dev->data->nb_rx_queues * sizeof(int), 0);
5084 		if (intr_handle->intr_vec == NULL) {
5085 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
5086 				     " intr_vec", dev->data->nb_rx_queues);
5087 			return -ENOMEM;
5088 		}
5089 	}
5090 	ixgbevf_configure_msix(dev);
5091 
5092 	/* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
5093 	 * is mapped to VFIO vector 0 in eth_ixgbevf_dev_init( ).
5094 	 * If previous VFIO interrupt mapping setting in eth_ixgbevf_dev_init( )
5095 	 * is not cleared, it will fail when following rte_intr_enable( ) tries
5096 	 * to map Rx queue interrupt to other VFIO vectors.
5097 	 * So clear uio/vfio intr/evevnfd first to avoid failure.
5098 	 */
5099 	rte_intr_disable(intr_handle);
5100 
5101 	rte_intr_enable(intr_handle);
5102 
5103 	/* Re-enable interrupt for VF */
5104 	ixgbevf_intr_enable(hw);
5105 
5106 	return 0;
5107 }
5108 
5109 static void
5110 ixgbevf_dev_stop(struct rte_eth_dev *dev)
5111 {
5112 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5113 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5114 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5115 
5116 	PMD_INIT_FUNC_TRACE();
5117 
5118 	ixgbevf_intr_disable(hw);
5119 
5120 	hw->adapter_stopped = 1;
5121 	ixgbe_stop_adapter(hw);
5122 
5123 	/*
5124 	  * Clear what we set, but we still keep shadow_vfta to
5125 	  * restore after device starts
5126 	  */
5127 	ixgbevf_set_vfta_all(dev, 0);
5128 
5129 	/* Clear stored conf */
5130 	dev->data->scattered_rx = 0;
5131 
5132 	ixgbe_dev_clear_queues(dev);
5133 
5134 	/* Clean datapath event and queue/vec mapping */
5135 	rte_intr_efd_disable(intr_handle);
5136 	if (intr_handle->intr_vec != NULL) {
5137 		rte_free(intr_handle->intr_vec);
5138 		intr_handle->intr_vec = NULL;
5139 	}
5140 }
5141 
5142 static void
5143 ixgbevf_dev_close(struct rte_eth_dev *dev)
5144 {
5145 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5146 
5147 	PMD_INIT_FUNC_TRACE();
5148 
5149 	ixgbe_reset_hw(hw);
5150 
5151 	ixgbevf_dev_stop(dev);
5152 
5153 	ixgbe_dev_free_queues(dev);
5154 
5155 	/**
5156 	 * Remove the VF MAC address ro ensure
5157 	 * that the VF traffic goes to the PF
5158 	 * after stop, close and detach of the VF
5159 	 **/
5160 	ixgbevf_remove_mac_addr(dev, 0);
5161 }
5162 
5163 /*
5164  * Reset VF device
5165  */
5166 static int
5167 ixgbevf_dev_reset(struct rte_eth_dev *dev)
5168 {
5169 	int ret;
5170 
5171 	ret = eth_ixgbevf_dev_uninit(dev);
5172 	if (ret)
5173 		return ret;
5174 
5175 	ret = eth_ixgbevf_dev_init(dev);
5176 
5177 	return ret;
5178 }
5179 
5180 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
5181 {
5182 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5183 	struct ixgbe_vfta *shadow_vfta =
5184 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5185 	int i = 0, j = 0, vfta = 0, mask = 1;
5186 
5187 	for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
5188 		vfta = shadow_vfta->vfta[i];
5189 		if (vfta) {
5190 			mask = 1;
5191 			for (j = 0; j < 32; j++) {
5192 				if (vfta & mask)
5193 					ixgbe_set_vfta(hw, (i<<5)+j, 0,
5194 						       on, false);
5195 				mask <<= 1;
5196 			}
5197 		}
5198 	}
5199 
5200 }
5201 
5202 static int
5203 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
5204 {
5205 	struct ixgbe_hw *hw =
5206 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5207 	struct ixgbe_vfta *shadow_vfta =
5208 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5209 	uint32_t vid_idx = 0;
5210 	uint32_t vid_bit = 0;
5211 	int ret = 0;
5212 
5213 	PMD_INIT_FUNC_TRACE();
5214 
5215 	/* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
5216 	ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
5217 	if (ret) {
5218 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
5219 		return ret;
5220 	}
5221 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
5222 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
5223 
5224 	/* Save what we set and retore it after device reset */
5225 	if (on)
5226 		shadow_vfta->vfta[vid_idx] |= vid_bit;
5227 	else
5228 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
5229 
5230 	return 0;
5231 }
5232 
5233 static void
5234 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
5235 {
5236 	struct ixgbe_hw *hw =
5237 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5238 	uint32_t ctrl;
5239 
5240 	PMD_INIT_FUNC_TRACE();
5241 
5242 	if (queue >= hw->mac.max_rx_queues)
5243 		return;
5244 
5245 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
5246 	if (on)
5247 		ctrl |= IXGBE_RXDCTL_VME;
5248 	else
5249 		ctrl &= ~IXGBE_RXDCTL_VME;
5250 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
5251 
5252 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
5253 }
5254 
5255 static int
5256 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
5257 {
5258 	struct ixgbe_hw *hw =
5259 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5260 	uint16_t i;
5261 	int on = 0;
5262 
5263 	/* VF function only support hw strip feature, others are not support */
5264 	if (mask & ETH_VLAN_STRIP_MASK) {
5265 		on = !!(dev->data->dev_conf.rxmode.hw_vlan_strip);
5266 
5267 		for (i = 0; i < hw->mac.max_rx_queues; i++)
5268 			ixgbevf_vlan_strip_queue_set(dev, i, on);
5269 	}
5270 
5271 	return 0;
5272 }
5273 
5274 int
5275 ixgbe_vt_check(struct ixgbe_hw *hw)
5276 {
5277 	uint32_t reg_val;
5278 
5279 	/* if Virtualization Technology is enabled */
5280 	reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
5281 	if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
5282 		PMD_INIT_LOG(ERR, "VT must be enabled for this setting");
5283 		return -1;
5284 	}
5285 
5286 	return 0;
5287 }
5288 
5289 static uint32_t
5290 ixgbe_uta_vector(struct ixgbe_hw *hw, struct ether_addr *uc_addr)
5291 {
5292 	uint32_t vector = 0;
5293 
5294 	switch (hw->mac.mc_filter_type) {
5295 	case 0:   /* use bits [47:36] of the address */
5296 		vector = ((uc_addr->addr_bytes[4] >> 4) |
5297 			(((uint16_t)uc_addr->addr_bytes[5]) << 4));
5298 		break;
5299 	case 1:   /* use bits [46:35] of the address */
5300 		vector = ((uc_addr->addr_bytes[4] >> 3) |
5301 			(((uint16_t)uc_addr->addr_bytes[5]) << 5));
5302 		break;
5303 	case 2:   /* use bits [45:34] of the address */
5304 		vector = ((uc_addr->addr_bytes[4] >> 2) |
5305 			(((uint16_t)uc_addr->addr_bytes[5]) << 6));
5306 		break;
5307 	case 3:   /* use bits [43:32] of the address */
5308 		vector = ((uc_addr->addr_bytes[4]) |
5309 			(((uint16_t)uc_addr->addr_bytes[5]) << 8));
5310 		break;
5311 	default:  /* Invalid mc_filter_type */
5312 		break;
5313 	}
5314 
5315 	/* vector can only be 12-bits or boundary will be exceeded */
5316 	vector &= 0xFFF;
5317 	return vector;
5318 }
5319 
5320 static int
5321 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
5322 			uint8_t on)
5323 {
5324 	uint32_t vector;
5325 	uint32_t uta_idx;
5326 	uint32_t reg_val;
5327 	uint32_t uta_shift;
5328 	uint32_t rc;
5329 	const uint32_t ixgbe_uta_idx_mask = 0x7F;
5330 	const uint32_t ixgbe_uta_bit_shift = 5;
5331 	const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
5332 	const uint32_t bit1 = 0x1;
5333 
5334 	struct ixgbe_hw *hw =
5335 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5336 	struct ixgbe_uta_info *uta_info =
5337 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5338 
5339 	/* The UTA table only exists on 82599 hardware and newer */
5340 	if (hw->mac.type < ixgbe_mac_82599EB)
5341 		return -ENOTSUP;
5342 
5343 	vector = ixgbe_uta_vector(hw, mac_addr);
5344 	uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
5345 	uta_shift = vector & ixgbe_uta_bit_mask;
5346 
5347 	rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
5348 	if (rc == on)
5349 		return 0;
5350 
5351 	reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
5352 	if (on) {
5353 		uta_info->uta_in_use++;
5354 		reg_val |= (bit1 << uta_shift);
5355 		uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
5356 	} else {
5357 		uta_info->uta_in_use--;
5358 		reg_val &= ~(bit1 << uta_shift);
5359 		uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
5360 	}
5361 
5362 	IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
5363 
5364 	if (uta_info->uta_in_use > 0)
5365 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
5366 				IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
5367 	else
5368 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
5369 
5370 	return 0;
5371 }
5372 
5373 static int
5374 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
5375 {
5376 	int i;
5377 	struct ixgbe_hw *hw =
5378 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5379 	struct ixgbe_uta_info *uta_info =
5380 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5381 
5382 	/* The UTA table only exists on 82599 hardware and newer */
5383 	if (hw->mac.type < ixgbe_mac_82599EB)
5384 		return -ENOTSUP;
5385 
5386 	if (on) {
5387 		for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5388 			uta_info->uta_shadow[i] = ~0;
5389 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
5390 		}
5391 	} else {
5392 		for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5393 			uta_info->uta_shadow[i] = 0;
5394 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
5395 		}
5396 	}
5397 	return 0;
5398 
5399 }
5400 
5401 uint32_t
5402 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
5403 {
5404 	uint32_t new_val = orig_val;
5405 
5406 	if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG)
5407 		new_val |= IXGBE_VMOLR_AUPE;
5408 	if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC)
5409 		new_val |= IXGBE_VMOLR_ROMPE;
5410 	if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
5411 		new_val |= IXGBE_VMOLR_ROPE;
5412 	if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
5413 		new_val |= IXGBE_VMOLR_BAM;
5414 	if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
5415 		new_val |= IXGBE_VMOLR_MPE;
5416 
5417 	return new_val;
5418 }
5419 
5420 #define IXGBE_MRCTL_VPME  0x01 /* Virtual Pool Mirroring. */
5421 #define IXGBE_MRCTL_UPME  0x02 /* Uplink Port Mirroring. */
5422 #define IXGBE_MRCTL_DPME  0x04 /* Downlink Port Mirroring. */
5423 #define IXGBE_MRCTL_VLME  0x08 /* VLAN Mirroring. */
5424 #define IXGBE_INVALID_MIRROR_TYPE(mirror_type) \
5425 	((mirror_type) & ~(uint8_t)(ETH_MIRROR_VIRTUAL_POOL_UP | \
5426 	ETH_MIRROR_UPLINK_PORT | ETH_MIRROR_DOWNLINK_PORT | ETH_MIRROR_VLAN))
5427 
5428 static int
5429 ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
5430 		      struct rte_eth_mirror_conf *mirror_conf,
5431 		      uint8_t rule_id, uint8_t on)
5432 {
5433 	uint32_t mr_ctl, vlvf;
5434 	uint32_t mp_lsb = 0;
5435 	uint32_t mv_msb = 0;
5436 	uint32_t mv_lsb = 0;
5437 	uint32_t mp_msb = 0;
5438 	uint8_t i = 0;
5439 	int reg_index = 0;
5440 	uint64_t vlan_mask = 0;
5441 
5442 	const uint8_t pool_mask_offset = 32;
5443 	const uint8_t vlan_mask_offset = 32;
5444 	const uint8_t dst_pool_offset = 8;
5445 	const uint8_t rule_mr_offset  = 4;
5446 	const uint8_t mirror_rule_mask = 0x0F;
5447 
5448 	struct ixgbe_mirror_info *mr_info =
5449 			(IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5450 	struct ixgbe_hw *hw =
5451 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5452 	uint8_t mirror_type = 0;
5453 
5454 	if (ixgbe_vt_check(hw) < 0)
5455 		return -ENOTSUP;
5456 
5457 	if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5458 		return -EINVAL;
5459 
5460 	if (IXGBE_INVALID_MIRROR_TYPE(mirror_conf->rule_type)) {
5461 		PMD_DRV_LOG(ERR, "unsupported mirror type 0x%x.",
5462 			    mirror_conf->rule_type);
5463 		return -EINVAL;
5464 	}
5465 
5466 	if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5467 		mirror_type |= IXGBE_MRCTL_VLME;
5468 		/* Check if vlan id is valid and find conresponding VLAN ID
5469 		 * index in VLVF
5470 		 */
5471 		for (i = 0; i < IXGBE_VLVF_ENTRIES; i++) {
5472 			if (mirror_conf->vlan.vlan_mask & (1ULL << i)) {
5473 				/* search vlan id related pool vlan filter
5474 				 * index
5475 				 */
5476 				reg_index = ixgbe_find_vlvf_slot(
5477 						hw,
5478 						mirror_conf->vlan.vlan_id[i],
5479 						false);
5480 				if (reg_index < 0)
5481 					return -EINVAL;
5482 				vlvf = IXGBE_READ_REG(hw,
5483 						      IXGBE_VLVF(reg_index));
5484 				if ((vlvf & IXGBE_VLVF_VIEN) &&
5485 				    ((vlvf & IXGBE_VLVF_VLANID_MASK) ==
5486 				      mirror_conf->vlan.vlan_id[i]))
5487 					vlan_mask |= (1ULL << reg_index);
5488 				else
5489 					return -EINVAL;
5490 			}
5491 		}
5492 
5493 		if (on) {
5494 			mv_lsb = vlan_mask & 0xFFFFFFFF;
5495 			mv_msb = vlan_mask >> vlan_mask_offset;
5496 
5497 			mr_info->mr_conf[rule_id].vlan.vlan_mask =
5498 						mirror_conf->vlan.vlan_mask;
5499 			for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++) {
5500 				if (mirror_conf->vlan.vlan_mask & (1ULL << i))
5501 					mr_info->mr_conf[rule_id].vlan.vlan_id[i] =
5502 						mirror_conf->vlan.vlan_id[i];
5503 			}
5504 		} else {
5505 			mv_lsb = 0;
5506 			mv_msb = 0;
5507 			mr_info->mr_conf[rule_id].vlan.vlan_mask = 0;
5508 			for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++)
5509 				mr_info->mr_conf[rule_id].vlan.vlan_id[i] = 0;
5510 		}
5511 	}
5512 
5513 	/**
5514 	 * if enable pool mirror, write related pool mask register,if disable
5515 	 * pool mirror, clear PFMRVM register
5516 	 */
5517 	if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5518 		mirror_type |= IXGBE_MRCTL_VPME;
5519 		if (on) {
5520 			mp_lsb = mirror_conf->pool_mask & 0xFFFFFFFF;
5521 			mp_msb = mirror_conf->pool_mask >> pool_mask_offset;
5522 			mr_info->mr_conf[rule_id].pool_mask =
5523 					mirror_conf->pool_mask;
5524 
5525 		} else {
5526 			mp_lsb = 0;
5527 			mp_msb = 0;
5528 			mr_info->mr_conf[rule_id].pool_mask = 0;
5529 		}
5530 	}
5531 	if (mirror_conf->rule_type & ETH_MIRROR_UPLINK_PORT)
5532 		mirror_type |= IXGBE_MRCTL_UPME;
5533 	if (mirror_conf->rule_type & ETH_MIRROR_DOWNLINK_PORT)
5534 		mirror_type |= IXGBE_MRCTL_DPME;
5535 
5536 	/* read  mirror control register and recalculate it */
5537 	mr_ctl = IXGBE_READ_REG(hw, IXGBE_MRCTL(rule_id));
5538 
5539 	if (on) {
5540 		mr_ctl |= mirror_type;
5541 		mr_ctl &= mirror_rule_mask;
5542 		mr_ctl |= mirror_conf->dst_pool << dst_pool_offset;
5543 	} else {
5544 		mr_ctl &= ~(mirror_conf->rule_type & mirror_rule_mask);
5545 	}
5546 
5547 	mr_info->mr_conf[rule_id].rule_type = mirror_conf->rule_type;
5548 	mr_info->mr_conf[rule_id].dst_pool = mirror_conf->dst_pool;
5549 
5550 	/* write mirrror control  register */
5551 	IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5552 
5553 	/* write pool mirrror control  register */
5554 	if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5555 		IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), mp_lsb);
5556 		IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset),
5557 				mp_msb);
5558 	}
5559 	/* write VLAN mirrror control  register */
5560 	if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5561 		IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), mv_lsb);
5562 		IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset),
5563 				mv_msb);
5564 	}
5565 
5566 	return 0;
5567 }
5568 
5569 static int
5570 ixgbe_mirror_rule_reset(struct rte_eth_dev *dev, uint8_t rule_id)
5571 {
5572 	int mr_ctl = 0;
5573 	uint32_t lsb_val = 0;
5574 	uint32_t msb_val = 0;
5575 	const uint8_t rule_mr_offset = 4;
5576 
5577 	struct ixgbe_hw *hw =
5578 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5579 	struct ixgbe_mirror_info *mr_info =
5580 		(IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5581 
5582 	if (ixgbe_vt_check(hw) < 0)
5583 		return -ENOTSUP;
5584 
5585 	if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5586 		return -EINVAL;
5587 
5588 	memset(&mr_info->mr_conf[rule_id], 0,
5589 	       sizeof(struct rte_eth_mirror_conf));
5590 
5591 	/* clear PFVMCTL register */
5592 	IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5593 
5594 	/* clear pool mask register */
5595 	IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), lsb_val);
5596 	IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset), msb_val);
5597 
5598 	/* clear vlan mask register */
5599 	IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), lsb_val);
5600 	IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset), msb_val);
5601 
5602 	return 0;
5603 }
5604 
5605 static int
5606 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5607 {
5608 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5609 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5610 	uint32_t mask;
5611 	struct ixgbe_hw *hw =
5612 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5613 	uint32_t vec = IXGBE_MISC_VEC_ID;
5614 
5615 	mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
5616 	if (rte_intr_allow_others(intr_handle))
5617 		vec = IXGBE_RX_VEC_START;
5618 	mask |= (1 << vec);
5619 	RTE_SET_USED(queue_id);
5620 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
5621 
5622 	rte_intr_enable(intr_handle);
5623 
5624 	return 0;
5625 }
5626 
5627 static int
5628 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5629 {
5630 	uint32_t mask;
5631 	struct ixgbe_hw *hw =
5632 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5633 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5634 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5635 	uint32_t vec = IXGBE_MISC_VEC_ID;
5636 
5637 	mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
5638 	if (rte_intr_allow_others(intr_handle))
5639 		vec = IXGBE_RX_VEC_START;
5640 	mask &= ~(1 << vec);
5641 	RTE_SET_USED(queue_id);
5642 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
5643 
5644 	return 0;
5645 }
5646 
5647 static int
5648 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5649 {
5650 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5651 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5652 	uint32_t mask;
5653 	struct ixgbe_hw *hw =
5654 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5655 	struct ixgbe_interrupt *intr =
5656 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5657 
5658 	if (queue_id < 16) {
5659 		ixgbe_disable_intr(hw);
5660 		intr->mask |= (1 << queue_id);
5661 		ixgbe_enable_intr(dev);
5662 	} else if (queue_id < 32) {
5663 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5664 		mask &= (1 << queue_id);
5665 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5666 	} else if (queue_id < 64) {
5667 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5668 		mask &= (1 << (queue_id - 32));
5669 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5670 	}
5671 	rte_intr_enable(intr_handle);
5672 
5673 	return 0;
5674 }
5675 
5676 static int
5677 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5678 {
5679 	uint32_t mask;
5680 	struct ixgbe_hw *hw =
5681 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5682 	struct ixgbe_interrupt *intr =
5683 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5684 
5685 	if (queue_id < 16) {
5686 		ixgbe_disable_intr(hw);
5687 		intr->mask &= ~(1 << queue_id);
5688 		ixgbe_enable_intr(dev);
5689 	} else if (queue_id < 32) {
5690 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5691 		mask &= ~(1 << queue_id);
5692 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5693 	} else if (queue_id < 64) {
5694 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5695 		mask &= ~(1 << (queue_id - 32));
5696 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5697 	}
5698 
5699 	return 0;
5700 }
5701 
5702 static void
5703 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5704 		     uint8_t queue, uint8_t msix_vector)
5705 {
5706 	uint32_t tmp, idx;
5707 
5708 	if (direction == -1) {
5709 		/* other causes */
5710 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5711 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
5712 		tmp &= ~0xFF;
5713 		tmp |= msix_vector;
5714 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
5715 	} else {
5716 		/* rx or tx cause */
5717 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5718 		idx = ((16 * (queue & 1)) + (8 * direction));
5719 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
5720 		tmp &= ~(0xFF << idx);
5721 		tmp |= (msix_vector << idx);
5722 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
5723 	}
5724 }
5725 
5726 /**
5727  * set the IVAR registers, mapping interrupt causes to vectors
5728  * @param hw
5729  *  pointer to ixgbe_hw struct
5730  * @direction
5731  *  0 for Rx, 1 for Tx, -1 for other causes
5732  * @queue
5733  *  queue to map the corresponding interrupt to
5734  * @msix_vector
5735  *  the vector to map to the corresponding queue
5736  */
5737 static void
5738 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5739 		   uint8_t queue, uint8_t msix_vector)
5740 {
5741 	uint32_t tmp, idx;
5742 
5743 	msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5744 	if (hw->mac.type == ixgbe_mac_82598EB) {
5745 		if (direction == -1)
5746 			direction = 0;
5747 		idx = (((direction * 64) + queue) >> 2) & 0x1F;
5748 		tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
5749 		tmp &= ~(0xFF << (8 * (queue & 0x3)));
5750 		tmp |= (msix_vector << (8 * (queue & 0x3)));
5751 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
5752 	} else if ((hw->mac.type == ixgbe_mac_82599EB) ||
5753 			(hw->mac.type == ixgbe_mac_X540) ||
5754 			(hw->mac.type == ixgbe_mac_X550)) {
5755 		if (direction == -1) {
5756 			/* other causes */
5757 			idx = ((queue & 1) * 8);
5758 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
5759 			tmp &= ~(0xFF << idx);
5760 			tmp |= (msix_vector << idx);
5761 			IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
5762 		} else {
5763 			/* rx or tx causes */
5764 			idx = ((16 * (queue & 1)) + (8 * direction));
5765 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
5766 			tmp &= ~(0xFF << idx);
5767 			tmp |= (msix_vector << idx);
5768 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
5769 		}
5770 	}
5771 }
5772 
5773 static void
5774 ixgbevf_configure_msix(struct rte_eth_dev *dev)
5775 {
5776 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5777 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5778 	struct ixgbe_hw *hw =
5779 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5780 	uint32_t q_idx;
5781 	uint32_t vector_idx = IXGBE_MISC_VEC_ID;
5782 	uint32_t base = IXGBE_MISC_VEC_ID;
5783 
5784 	/* Configure VF other cause ivar */
5785 	ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
5786 
5787 	/* won't configure msix register if no mapping is done
5788 	 * between intr vector and event fd.
5789 	 */
5790 	if (!rte_intr_dp_is_en(intr_handle))
5791 		return;
5792 
5793 	if (rte_intr_allow_others(intr_handle)) {
5794 		base = IXGBE_RX_VEC_START;
5795 		vector_idx = IXGBE_RX_VEC_START;
5796 	}
5797 
5798 	/* Configure all RX queues of VF */
5799 	for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
5800 		/* Force all queue use vector 0,
5801 		 * as IXGBE_VF_MAXMSIVECOTR = 1
5802 		 */
5803 		ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
5804 		intr_handle->intr_vec[q_idx] = vector_idx;
5805 		if (vector_idx < base + intr_handle->nb_efd - 1)
5806 			vector_idx++;
5807 	}
5808 }
5809 
5810 /**
5811  * Sets up the hardware to properly generate MSI-X interrupts
5812  * @hw
5813  *  board private structure
5814  */
5815 static void
5816 ixgbe_configure_msix(struct rte_eth_dev *dev)
5817 {
5818 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5819 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5820 	struct ixgbe_hw *hw =
5821 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5822 	uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
5823 	uint32_t vec = IXGBE_MISC_VEC_ID;
5824 	uint32_t mask;
5825 	uint32_t gpie;
5826 
5827 	/* won't configure msix register if no mapping is done
5828 	 * between intr vector and event fd
5829 	 * but if misx has been enabled already, need to configure
5830 	 * auto clean, auto mask and throttling.
5831 	 */
5832 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5833 	if (!rte_intr_dp_is_en(intr_handle) &&
5834 	    !(gpie & (IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT)))
5835 		return;
5836 
5837 	if (rte_intr_allow_others(intr_handle))
5838 		vec = base = IXGBE_RX_VEC_START;
5839 
5840 	/* setup GPIE for MSI-x mode */
5841 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5842 	gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5843 		IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
5844 	/* auto clearing and auto setting corresponding bits in EIMS
5845 	 * when MSI-X interrupt is triggered
5846 	 */
5847 	if (hw->mac.type == ixgbe_mac_82598EB) {
5848 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5849 	} else {
5850 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5851 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5852 	}
5853 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5854 
5855 	/* Populate the IVAR table and set the ITR values to the
5856 	 * corresponding register.
5857 	 */
5858 	if (rte_intr_dp_is_en(intr_handle)) {
5859 		for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
5860 			queue_id++) {
5861 			/* by default, 1:1 mapping */
5862 			ixgbe_set_ivar_map(hw, 0, queue_id, vec);
5863 			intr_handle->intr_vec[queue_id] = vec;
5864 			if (vec < base + intr_handle->nb_efd - 1)
5865 				vec++;
5866 		}
5867 
5868 		switch (hw->mac.type) {
5869 		case ixgbe_mac_82598EB:
5870 			ixgbe_set_ivar_map(hw, -1,
5871 					   IXGBE_IVAR_OTHER_CAUSES_INDEX,
5872 					   IXGBE_MISC_VEC_ID);
5873 			break;
5874 		case ixgbe_mac_82599EB:
5875 		case ixgbe_mac_X540:
5876 		case ixgbe_mac_X550:
5877 			ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
5878 			break;
5879 		default:
5880 			break;
5881 		}
5882 	}
5883 	IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
5884 			IXGBE_MIN_INTER_INTERRUPT_INTERVAL_DEFAULT & 0xFFF);
5885 
5886 	/* set up to autoclear timer, and the vectors */
5887 	mask = IXGBE_EIMS_ENABLE_MASK;
5888 	mask &= ~(IXGBE_EIMS_OTHER |
5889 		  IXGBE_EIMS_MAILBOX |
5890 		  IXGBE_EIMS_LSC);
5891 
5892 	IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
5893 }
5894 
5895 int
5896 ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
5897 			   uint16_t queue_idx, uint16_t tx_rate)
5898 {
5899 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5900 	uint32_t rf_dec, rf_int;
5901 	uint32_t bcnrc_val;
5902 	uint16_t link_speed = dev->data->dev_link.link_speed;
5903 
5904 	if (queue_idx >= hw->mac.max_tx_queues)
5905 		return -EINVAL;
5906 
5907 	if (tx_rate != 0) {
5908 		/* Calculate the rate factor values to set */
5909 		rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
5910 		rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
5911 		rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
5912 
5913 		bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
5914 		bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
5915 				IXGBE_RTTBCNRC_RF_INT_MASK_M);
5916 		bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
5917 	} else {
5918 		bcnrc_val = 0;
5919 	}
5920 
5921 	/*
5922 	 * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
5923 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
5924 	 * set as 0x4.
5925 	 */
5926 	if ((dev->data->dev_conf.rxmode.jumbo_frame == 1) &&
5927 		(dev->data->dev_conf.rxmode.max_rx_pkt_len >=
5928 				IXGBE_MAX_JUMBO_FRAME_SIZE))
5929 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
5930 			IXGBE_MMW_SIZE_JUMBO_FRAME);
5931 	else
5932 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
5933 			IXGBE_MMW_SIZE_DEFAULT);
5934 
5935 	/* Set RTTBCNRC of queue X */
5936 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
5937 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
5938 	IXGBE_WRITE_FLUSH(hw);
5939 
5940 	return 0;
5941 }
5942 
5943 static int
5944 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
5945 		     __attribute__((unused)) uint32_t index,
5946 		     __attribute__((unused)) uint32_t pool)
5947 {
5948 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5949 	int diag;
5950 
5951 	/*
5952 	 * On a 82599 VF, adding again the same MAC addr is not an idempotent
5953 	 * operation. Trap this case to avoid exhausting the [very limited]
5954 	 * set of PF resources used to store VF MAC addresses.
5955 	 */
5956 	if (memcmp(hw->mac.perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
5957 		return -1;
5958 	diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
5959 	if (diag != 0)
5960 		PMD_DRV_LOG(ERR, "Unable to add MAC address "
5961 			    "%02x:%02x:%02x:%02x:%02x:%02x - diag=%d",
5962 			    mac_addr->addr_bytes[0],
5963 			    mac_addr->addr_bytes[1],
5964 			    mac_addr->addr_bytes[2],
5965 			    mac_addr->addr_bytes[3],
5966 			    mac_addr->addr_bytes[4],
5967 			    mac_addr->addr_bytes[5],
5968 			    diag);
5969 	return diag;
5970 }
5971 
5972 static void
5973 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
5974 {
5975 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5976 	struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
5977 	struct ether_addr *mac_addr;
5978 	uint32_t i;
5979 	int diag;
5980 
5981 	/*
5982 	 * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
5983 	 * not support the deletion of a given MAC address.
5984 	 * Instead, it imposes to delete all MAC addresses, then to add again
5985 	 * all MAC addresses with the exception of the one to be deleted.
5986 	 */
5987 	(void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
5988 
5989 	/*
5990 	 * Add again all MAC addresses, with the exception of the deleted one
5991 	 * and of the permanent MAC address.
5992 	 */
5993 	for (i = 0, mac_addr = dev->data->mac_addrs;
5994 	     i < hw->mac.num_rar_entries; i++, mac_addr++) {
5995 		/* Skip the deleted MAC address */
5996 		if (i == index)
5997 			continue;
5998 		/* Skip NULL MAC addresses */
5999 		if (is_zero_ether_addr(mac_addr))
6000 			continue;
6001 		/* Skip the permanent MAC address */
6002 		if (memcmp(perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
6003 			continue;
6004 		diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6005 		if (diag != 0)
6006 			PMD_DRV_LOG(ERR,
6007 				    "Adding again MAC address "
6008 				    "%02x:%02x:%02x:%02x:%02x:%02x failed "
6009 				    "diag=%d",
6010 				    mac_addr->addr_bytes[0],
6011 				    mac_addr->addr_bytes[1],
6012 				    mac_addr->addr_bytes[2],
6013 				    mac_addr->addr_bytes[3],
6014 				    mac_addr->addr_bytes[4],
6015 				    mac_addr->addr_bytes[5],
6016 				    diag);
6017 	}
6018 }
6019 
6020 static void
6021 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
6022 {
6023 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6024 
6025 	hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
6026 }
6027 
6028 int
6029 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
6030 			struct rte_eth_syn_filter *filter,
6031 			bool add)
6032 {
6033 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6034 	struct ixgbe_filter_info *filter_info =
6035 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6036 	uint32_t syn_info;
6037 	uint32_t synqf;
6038 
6039 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6040 		return -EINVAL;
6041 
6042 	syn_info = filter_info->syn_info;
6043 
6044 	if (add) {
6045 		if (syn_info & IXGBE_SYN_FILTER_ENABLE)
6046 			return -EINVAL;
6047 		synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
6048 			IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
6049 
6050 		if (filter->hig_pri)
6051 			synqf |= IXGBE_SYN_FILTER_SYNQFP;
6052 		else
6053 			synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
6054 	} else {
6055 		synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6056 		if (!(syn_info & IXGBE_SYN_FILTER_ENABLE))
6057 			return -ENOENT;
6058 		synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
6059 	}
6060 
6061 	filter_info->syn_info = synqf;
6062 	IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
6063 	IXGBE_WRITE_FLUSH(hw);
6064 	return 0;
6065 }
6066 
6067 static int
6068 ixgbe_syn_filter_get(struct rte_eth_dev *dev,
6069 			struct rte_eth_syn_filter *filter)
6070 {
6071 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6072 	uint32_t synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6073 
6074 	if (synqf & IXGBE_SYN_FILTER_ENABLE) {
6075 		filter->hig_pri = (synqf & IXGBE_SYN_FILTER_SYNQFP) ? 1 : 0;
6076 		filter->queue = (uint16_t)((synqf & IXGBE_SYN_FILTER_QUEUE) >> 1);
6077 		return 0;
6078 	}
6079 	return -ENOENT;
6080 }
6081 
6082 static int
6083 ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
6084 			enum rte_filter_op filter_op,
6085 			void *arg)
6086 {
6087 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6088 	int ret;
6089 
6090 	MAC_TYPE_FILTER_SUP(hw->mac.type);
6091 
6092 	if (filter_op == RTE_ETH_FILTER_NOP)
6093 		return 0;
6094 
6095 	if (arg == NULL) {
6096 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
6097 			    filter_op);
6098 		return -EINVAL;
6099 	}
6100 
6101 	switch (filter_op) {
6102 	case RTE_ETH_FILTER_ADD:
6103 		ret = ixgbe_syn_filter_set(dev,
6104 				(struct rte_eth_syn_filter *)arg,
6105 				TRUE);
6106 		break;
6107 	case RTE_ETH_FILTER_DELETE:
6108 		ret = ixgbe_syn_filter_set(dev,
6109 				(struct rte_eth_syn_filter *)arg,
6110 				FALSE);
6111 		break;
6112 	case RTE_ETH_FILTER_GET:
6113 		ret = ixgbe_syn_filter_get(dev,
6114 				(struct rte_eth_syn_filter *)arg);
6115 		break;
6116 	default:
6117 		PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
6118 		ret = -EINVAL;
6119 		break;
6120 	}
6121 
6122 	return ret;
6123 }
6124 
6125 
6126 static inline enum ixgbe_5tuple_protocol
6127 convert_protocol_type(uint8_t protocol_value)
6128 {
6129 	if (protocol_value == IPPROTO_TCP)
6130 		return IXGBE_FILTER_PROTOCOL_TCP;
6131 	else if (protocol_value == IPPROTO_UDP)
6132 		return IXGBE_FILTER_PROTOCOL_UDP;
6133 	else if (protocol_value == IPPROTO_SCTP)
6134 		return IXGBE_FILTER_PROTOCOL_SCTP;
6135 	else
6136 		return IXGBE_FILTER_PROTOCOL_NONE;
6137 }
6138 
6139 /* inject a 5-tuple filter to HW */
6140 static inline void
6141 ixgbe_inject_5tuple_filter(struct rte_eth_dev *dev,
6142 			   struct ixgbe_5tuple_filter *filter)
6143 {
6144 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6145 	int i;
6146 	uint32_t ftqf, sdpqf;
6147 	uint32_t l34timir = 0;
6148 	uint8_t mask = 0xff;
6149 
6150 	i = filter->index;
6151 
6152 	sdpqf = (uint32_t)(filter->filter_info.dst_port <<
6153 				IXGBE_SDPQF_DSTPORT_SHIFT);
6154 	sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
6155 
6156 	ftqf = (uint32_t)(filter->filter_info.proto &
6157 		IXGBE_FTQF_PROTOCOL_MASK);
6158 	ftqf |= (uint32_t)((filter->filter_info.priority &
6159 		IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
6160 	if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
6161 		mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
6162 	if (filter->filter_info.dst_ip_mask == 0)
6163 		mask &= IXGBE_FTQF_DEST_ADDR_MASK;
6164 	if (filter->filter_info.src_port_mask == 0)
6165 		mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
6166 	if (filter->filter_info.dst_port_mask == 0)
6167 		mask &= IXGBE_FTQF_DEST_PORT_MASK;
6168 	if (filter->filter_info.proto_mask == 0)
6169 		mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
6170 	ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
6171 	ftqf |= IXGBE_FTQF_POOL_MASK_EN;
6172 	ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
6173 
6174 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
6175 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
6176 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
6177 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
6178 
6179 	l34timir |= IXGBE_L34T_IMIR_RESERVE;
6180 	l34timir |= (uint32_t)(filter->queue <<
6181 				IXGBE_L34T_IMIR_QUEUE_SHIFT);
6182 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
6183 }
6184 
6185 /*
6186  * add a 5tuple filter
6187  *
6188  * @param
6189  * dev: Pointer to struct rte_eth_dev.
6190  * index: the index the filter allocates.
6191  * filter: ponter to the filter that will be added.
6192  * rx_queue: the queue id the filter assigned to.
6193  *
6194  * @return
6195  *    - On success, zero.
6196  *    - On failure, a negative value.
6197  */
6198 static int
6199 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
6200 			struct ixgbe_5tuple_filter *filter)
6201 {
6202 	struct ixgbe_filter_info *filter_info =
6203 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6204 	int i, idx, shift;
6205 
6206 	/*
6207 	 * look for an unused 5tuple filter index,
6208 	 * and insert the filter to list.
6209 	 */
6210 	for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
6211 		idx = i / (sizeof(uint32_t) * NBBY);
6212 		shift = i % (sizeof(uint32_t) * NBBY);
6213 		if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
6214 			filter_info->fivetuple_mask[idx] |= 1 << shift;
6215 			filter->index = i;
6216 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
6217 					  filter,
6218 					  entries);
6219 			break;
6220 		}
6221 	}
6222 	if (i >= IXGBE_MAX_FTQF_FILTERS) {
6223 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
6224 		return -ENOSYS;
6225 	}
6226 
6227 	ixgbe_inject_5tuple_filter(dev, filter);
6228 
6229 	return 0;
6230 }
6231 
6232 /*
6233  * remove a 5tuple filter
6234  *
6235  * @param
6236  * dev: Pointer to struct rte_eth_dev.
6237  * filter: the pointer of the filter will be removed.
6238  */
6239 static void
6240 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
6241 			struct ixgbe_5tuple_filter *filter)
6242 {
6243 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6244 	struct ixgbe_filter_info *filter_info =
6245 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6246 	uint16_t index = filter->index;
6247 
6248 	filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
6249 				~(1 << (index % (sizeof(uint32_t) * NBBY)));
6250 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
6251 	rte_free(filter);
6252 
6253 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
6254 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
6255 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
6256 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
6257 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
6258 }
6259 
6260 static int
6261 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
6262 {
6263 	struct ixgbe_hw *hw;
6264 	uint32_t max_frame = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
6265 	struct rte_eth_rxmode *rx_conf = &dev->data->dev_conf.rxmode;
6266 
6267 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6268 
6269 	if ((mtu < ETHER_MIN_MTU) || (max_frame > ETHER_MAX_JUMBO_FRAME_LEN))
6270 		return -EINVAL;
6271 
6272 	/* refuse mtu that requires the support of scattered packets when this
6273 	 * feature has not been enabled before.
6274 	 */
6275 	if (!rx_conf->enable_scatter &&
6276 	    (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
6277 	     dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM))
6278 		return -EINVAL;
6279 
6280 	/*
6281 	 * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
6282 	 * request of the version 2.0 of the mailbox API.
6283 	 * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
6284 	 * of the mailbox API.
6285 	 * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
6286 	 * prior to 3.11.33 which contains the following change:
6287 	 * "ixgbe: Enable jumbo frames support w/ SR-IOV"
6288 	 */
6289 	ixgbevf_rlpml_set_vf(hw, max_frame);
6290 
6291 	/* update max frame size */
6292 	dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame;
6293 	return 0;
6294 }
6295 
6296 static inline struct ixgbe_5tuple_filter *
6297 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
6298 			struct ixgbe_5tuple_filter_info *key)
6299 {
6300 	struct ixgbe_5tuple_filter *it;
6301 
6302 	TAILQ_FOREACH(it, filter_list, entries) {
6303 		if (memcmp(key, &it->filter_info,
6304 			sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
6305 			return it;
6306 		}
6307 	}
6308 	return NULL;
6309 }
6310 
6311 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
6312 static inline int
6313 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
6314 			struct ixgbe_5tuple_filter_info *filter_info)
6315 {
6316 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
6317 		filter->priority > IXGBE_5TUPLE_MAX_PRI ||
6318 		filter->priority < IXGBE_5TUPLE_MIN_PRI)
6319 		return -EINVAL;
6320 
6321 	switch (filter->dst_ip_mask) {
6322 	case UINT32_MAX:
6323 		filter_info->dst_ip_mask = 0;
6324 		filter_info->dst_ip = filter->dst_ip;
6325 		break;
6326 	case 0:
6327 		filter_info->dst_ip_mask = 1;
6328 		break;
6329 	default:
6330 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
6331 		return -EINVAL;
6332 	}
6333 
6334 	switch (filter->src_ip_mask) {
6335 	case UINT32_MAX:
6336 		filter_info->src_ip_mask = 0;
6337 		filter_info->src_ip = filter->src_ip;
6338 		break;
6339 	case 0:
6340 		filter_info->src_ip_mask = 1;
6341 		break;
6342 	default:
6343 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
6344 		return -EINVAL;
6345 	}
6346 
6347 	switch (filter->dst_port_mask) {
6348 	case UINT16_MAX:
6349 		filter_info->dst_port_mask = 0;
6350 		filter_info->dst_port = filter->dst_port;
6351 		break;
6352 	case 0:
6353 		filter_info->dst_port_mask = 1;
6354 		break;
6355 	default:
6356 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
6357 		return -EINVAL;
6358 	}
6359 
6360 	switch (filter->src_port_mask) {
6361 	case UINT16_MAX:
6362 		filter_info->src_port_mask = 0;
6363 		filter_info->src_port = filter->src_port;
6364 		break;
6365 	case 0:
6366 		filter_info->src_port_mask = 1;
6367 		break;
6368 	default:
6369 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
6370 		return -EINVAL;
6371 	}
6372 
6373 	switch (filter->proto_mask) {
6374 	case UINT8_MAX:
6375 		filter_info->proto_mask = 0;
6376 		filter_info->proto =
6377 			convert_protocol_type(filter->proto);
6378 		break;
6379 	case 0:
6380 		filter_info->proto_mask = 1;
6381 		break;
6382 	default:
6383 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
6384 		return -EINVAL;
6385 	}
6386 
6387 	filter_info->priority = (uint8_t)filter->priority;
6388 	return 0;
6389 }
6390 
6391 /*
6392  * add or delete a ntuple filter
6393  *
6394  * @param
6395  * dev: Pointer to struct rte_eth_dev.
6396  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6397  * add: if true, add filter, if false, remove filter
6398  *
6399  * @return
6400  *    - On success, zero.
6401  *    - On failure, a negative value.
6402  */
6403 int
6404 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
6405 			struct rte_eth_ntuple_filter *ntuple_filter,
6406 			bool add)
6407 {
6408 	struct ixgbe_filter_info *filter_info =
6409 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6410 	struct ixgbe_5tuple_filter_info filter_5tuple;
6411 	struct ixgbe_5tuple_filter *filter;
6412 	int ret;
6413 
6414 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6415 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6416 		return -EINVAL;
6417 	}
6418 
6419 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6420 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6421 	if (ret < 0)
6422 		return ret;
6423 
6424 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6425 					 &filter_5tuple);
6426 	if (filter != NULL && add) {
6427 		PMD_DRV_LOG(ERR, "filter exists.");
6428 		return -EEXIST;
6429 	}
6430 	if (filter == NULL && !add) {
6431 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
6432 		return -ENOENT;
6433 	}
6434 
6435 	if (add) {
6436 		filter = rte_zmalloc("ixgbe_5tuple_filter",
6437 				sizeof(struct ixgbe_5tuple_filter), 0);
6438 		if (filter == NULL)
6439 			return -ENOMEM;
6440 		rte_memcpy(&filter->filter_info,
6441 				 &filter_5tuple,
6442 				 sizeof(struct ixgbe_5tuple_filter_info));
6443 		filter->queue = ntuple_filter->queue;
6444 		ret = ixgbe_add_5tuple_filter(dev, filter);
6445 		if (ret < 0) {
6446 			rte_free(filter);
6447 			return ret;
6448 		}
6449 	} else
6450 		ixgbe_remove_5tuple_filter(dev, filter);
6451 
6452 	return 0;
6453 }
6454 
6455 /*
6456  * get a ntuple filter
6457  *
6458  * @param
6459  * dev: Pointer to struct rte_eth_dev.
6460  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6461  *
6462  * @return
6463  *    - On success, zero.
6464  *    - On failure, a negative value.
6465  */
6466 static int
6467 ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
6468 			struct rte_eth_ntuple_filter *ntuple_filter)
6469 {
6470 	struct ixgbe_filter_info *filter_info =
6471 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6472 	struct ixgbe_5tuple_filter_info filter_5tuple;
6473 	struct ixgbe_5tuple_filter *filter;
6474 	int ret;
6475 
6476 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6477 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6478 		return -EINVAL;
6479 	}
6480 
6481 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6482 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6483 	if (ret < 0)
6484 		return ret;
6485 
6486 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6487 					 &filter_5tuple);
6488 	if (filter == NULL) {
6489 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
6490 		return -ENOENT;
6491 	}
6492 	ntuple_filter->queue = filter->queue;
6493 	return 0;
6494 }
6495 
6496 /*
6497  * ixgbe_ntuple_filter_handle - Handle operations for ntuple filter.
6498  * @dev: pointer to rte_eth_dev structure
6499  * @filter_op:operation will be taken.
6500  * @arg: a pointer to specific structure corresponding to the filter_op
6501  *
6502  * @return
6503  *    - On success, zero.
6504  *    - On failure, a negative value.
6505  */
6506 static int
6507 ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
6508 				enum rte_filter_op filter_op,
6509 				void *arg)
6510 {
6511 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6512 	int ret;
6513 
6514 	MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
6515 
6516 	if (filter_op == RTE_ETH_FILTER_NOP)
6517 		return 0;
6518 
6519 	if (arg == NULL) {
6520 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
6521 			    filter_op);
6522 		return -EINVAL;
6523 	}
6524 
6525 	switch (filter_op) {
6526 	case RTE_ETH_FILTER_ADD:
6527 		ret = ixgbe_add_del_ntuple_filter(dev,
6528 			(struct rte_eth_ntuple_filter *)arg,
6529 			TRUE);
6530 		break;
6531 	case RTE_ETH_FILTER_DELETE:
6532 		ret = ixgbe_add_del_ntuple_filter(dev,
6533 			(struct rte_eth_ntuple_filter *)arg,
6534 			FALSE);
6535 		break;
6536 	case RTE_ETH_FILTER_GET:
6537 		ret = ixgbe_get_ntuple_filter(dev,
6538 			(struct rte_eth_ntuple_filter *)arg);
6539 		break;
6540 	default:
6541 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
6542 		ret = -EINVAL;
6543 		break;
6544 	}
6545 	return ret;
6546 }
6547 
6548 int
6549 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
6550 			struct rte_eth_ethertype_filter *filter,
6551 			bool add)
6552 {
6553 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6554 	struct ixgbe_filter_info *filter_info =
6555 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6556 	uint32_t etqf = 0;
6557 	uint32_t etqs = 0;
6558 	int ret;
6559 	struct ixgbe_ethertype_filter ethertype_filter;
6560 
6561 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6562 		return -EINVAL;
6563 
6564 	if (filter->ether_type == ETHER_TYPE_IPv4 ||
6565 		filter->ether_type == ETHER_TYPE_IPv6) {
6566 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
6567 			" ethertype filter.", filter->ether_type);
6568 		return -EINVAL;
6569 	}
6570 
6571 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
6572 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
6573 		return -EINVAL;
6574 	}
6575 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
6576 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
6577 		return -EINVAL;
6578 	}
6579 
6580 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6581 	if (ret >= 0 && add) {
6582 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
6583 			    filter->ether_type);
6584 		return -EEXIST;
6585 	}
6586 	if (ret < 0 && !add) {
6587 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6588 			    filter->ether_type);
6589 		return -ENOENT;
6590 	}
6591 
6592 	if (add) {
6593 		etqf = IXGBE_ETQF_FILTER_EN;
6594 		etqf |= (uint32_t)filter->ether_type;
6595 		etqs |= (uint32_t)((filter->queue <<
6596 				    IXGBE_ETQS_RX_QUEUE_SHIFT) &
6597 				    IXGBE_ETQS_RX_QUEUE);
6598 		etqs |= IXGBE_ETQS_QUEUE_EN;
6599 
6600 		ethertype_filter.ethertype = filter->ether_type;
6601 		ethertype_filter.etqf = etqf;
6602 		ethertype_filter.etqs = etqs;
6603 		ethertype_filter.conf = FALSE;
6604 		ret = ixgbe_ethertype_filter_insert(filter_info,
6605 						    &ethertype_filter);
6606 		if (ret < 0) {
6607 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
6608 			return -ENOSPC;
6609 		}
6610 	} else {
6611 		ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
6612 		if (ret < 0)
6613 			return -ENOSYS;
6614 	}
6615 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
6616 	IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
6617 	IXGBE_WRITE_FLUSH(hw);
6618 
6619 	return 0;
6620 }
6621 
6622 static int
6623 ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
6624 			struct rte_eth_ethertype_filter *filter)
6625 {
6626 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6627 	struct ixgbe_filter_info *filter_info =
6628 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6629 	uint32_t etqf, etqs;
6630 	int ret;
6631 
6632 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6633 	if (ret < 0) {
6634 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6635 			    filter->ether_type);
6636 		return -ENOENT;
6637 	}
6638 
6639 	etqf = IXGBE_READ_REG(hw, IXGBE_ETQF(ret));
6640 	if (etqf & IXGBE_ETQF_FILTER_EN) {
6641 		etqs = IXGBE_READ_REG(hw, IXGBE_ETQS(ret));
6642 		filter->ether_type = etqf & IXGBE_ETQF_ETHERTYPE;
6643 		filter->flags = 0;
6644 		filter->queue = (etqs & IXGBE_ETQS_RX_QUEUE) >>
6645 			       IXGBE_ETQS_RX_QUEUE_SHIFT;
6646 		return 0;
6647 	}
6648 	return -ENOENT;
6649 }
6650 
6651 /*
6652  * ixgbe_ethertype_filter_handle - Handle operations for ethertype filter.
6653  * @dev: pointer to rte_eth_dev structure
6654  * @filter_op:operation will be taken.
6655  * @arg: a pointer to specific structure corresponding to the filter_op
6656  */
6657 static int
6658 ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
6659 				enum rte_filter_op filter_op,
6660 				void *arg)
6661 {
6662 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6663 	int ret;
6664 
6665 	MAC_TYPE_FILTER_SUP(hw->mac.type);
6666 
6667 	if (filter_op == RTE_ETH_FILTER_NOP)
6668 		return 0;
6669 
6670 	if (arg == NULL) {
6671 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
6672 			    filter_op);
6673 		return -EINVAL;
6674 	}
6675 
6676 	switch (filter_op) {
6677 	case RTE_ETH_FILTER_ADD:
6678 		ret = ixgbe_add_del_ethertype_filter(dev,
6679 			(struct rte_eth_ethertype_filter *)arg,
6680 			TRUE);
6681 		break;
6682 	case RTE_ETH_FILTER_DELETE:
6683 		ret = ixgbe_add_del_ethertype_filter(dev,
6684 			(struct rte_eth_ethertype_filter *)arg,
6685 			FALSE);
6686 		break;
6687 	case RTE_ETH_FILTER_GET:
6688 		ret = ixgbe_get_ethertype_filter(dev,
6689 			(struct rte_eth_ethertype_filter *)arg);
6690 		break;
6691 	default:
6692 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
6693 		ret = -EINVAL;
6694 		break;
6695 	}
6696 	return ret;
6697 }
6698 
6699 static int
6700 ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
6701 		     enum rte_filter_type filter_type,
6702 		     enum rte_filter_op filter_op,
6703 		     void *arg)
6704 {
6705 	int ret = 0;
6706 
6707 	switch (filter_type) {
6708 	case RTE_ETH_FILTER_NTUPLE:
6709 		ret = ixgbe_ntuple_filter_handle(dev, filter_op, arg);
6710 		break;
6711 	case RTE_ETH_FILTER_ETHERTYPE:
6712 		ret = ixgbe_ethertype_filter_handle(dev, filter_op, arg);
6713 		break;
6714 	case RTE_ETH_FILTER_SYN:
6715 		ret = ixgbe_syn_filter_handle(dev, filter_op, arg);
6716 		break;
6717 	case RTE_ETH_FILTER_FDIR:
6718 		ret = ixgbe_fdir_ctrl_func(dev, filter_op, arg);
6719 		break;
6720 	case RTE_ETH_FILTER_L2_TUNNEL:
6721 		ret = ixgbe_dev_l2_tunnel_filter_handle(dev, filter_op, arg);
6722 		break;
6723 	case RTE_ETH_FILTER_GENERIC:
6724 		if (filter_op != RTE_ETH_FILTER_GET)
6725 			return -EINVAL;
6726 		*(const void **)arg = &ixgbe_flow_ops;
6727 		break;
6728 	default:
6729 		PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
6730 							filter_type);
6731 		ret = -EINVAL;
6732 		break;
6733 	}
6734 
6735 	return ret;
6736 }
6737 
6738 static u8 *
6739 ixgbe_dev_addr_list_itr(__attribute__((unused)) struct ixgbe_hw *hw,
6740 			u8 **mc_addr_ptr, u32 *vmdq)
6741 {
6742 	u8 *mc_addr;
6743 
6744 	*vmdq = 0;
6745 	mc_addr = *mc_addr_ptr;
6746 	*mc_addr_ptr = (mc_addr + sizeof(struct ether_addr));
6747 	return mc_addr;
6748 }
6749 
6750 static int
6751 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
6752 			  struct ether_addr *mc_addr_set,
6753 			  uint32_t nb_mc_addr)
6754 {
6755 	struct ixgbe_hw *hw;
6756 	u8 *mc_addr_list;
6757 
6758 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6759 	mc_addr_list = (u8 *)mc_addr_set;
6760 	return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
6761 					 ixgbe_dev_addr_list_itr, TRUE);
6762 }
6763 
6764 static uint64_t
6765 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
6766 {
6767 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6768 	uint64_t systime_cycles;
6769 
6770 	switch (hw->mac.type) {
6771 	case ixgbe_mac_X550:
6772 	case ixgbe_mac_X550EM_x:
6773 	case ixgbe_mac_X550EM_a:
6774 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
6775 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6776 		systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6777 				* NSEC_PER_SEC;
6778 		break;
6779 	default:
6780 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6781 		systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6782 				<< 32;
6783 	}
6784 
6785 	return systime_cycles;
6786 }
6787 
6788 static uint64_t
6789 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6790 {
6791 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6792 	uint64_t rx_tstamp_cycles;
6793 
6794 	switch (hw->mac.type) {
6795 	case ixgbe_mac_X550:
6796 	case ixgbe_mac_X550EM_x:
6797 	case ixgbe_mac_X550EM_a:
6798 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6799 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6800 		rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6801 				* NSEC_PER_SEC;
6802 		break;
6803 	default:
6804 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6805 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6806 		rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6807 				<< 32;
6808 	}
6809 
6810 	return rx_tstamp_cycles;
6811 }
6812 
6813 static uint64_t
6814 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6815 {
6816 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6817 	uint64_t tx_tstamp_cycles;
6818 
6819 	switch (hw->mac.type) {
6820 	case ixgbe_mac_X550:
6821 	case ixgbe_mac_X550EM_x:
6822 	case ixgbe_mac_X550EM_a:
6823 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6824 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6825 		tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6826 				* NSEC_PER_SEC;
6827 		break;
6828 	default:
6829 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6830 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6831 		tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6832 				<< 32;
6833 	}
6834 
6835 	return tx_tstamp_cycles;
6836 }
6837 
6838 static void
6839 ixgbe_start_timecounters(struct rte_eth_dev *dev)
6840 {
6841 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6842 	struct ixgbe_adapter *adapter =
6843 		(struct ixgbe_adapter *)dev->data->dev_private;
6844 	struct rte_eth_link link;
6845 	uint32_t incval = 0;
6846 	uint32_t shift = 0;
6847 
6848 	/* Get current link speed. */
6849 	memset(&link, 0, sizeof(link));
6850 	ixgbe_dev_link_update(dev, 1);
6851 	rte_ixgbe_dev_atomic_read_link_status(dev, &link);
6852 
6853 	switch (link.link_speed) {
6854 	case ETH_SPEED_NUM_100M:
6855 		incval = IXGBE_INCVAL_100;
6856 		shift = IXGBE_INCVAL_SHIFT_100;
6857 		break;
6858 	case ETH_SPEED_NUM_1G:
6859 		incval = IXGBE_INCVAL_1GB;
6860 		shift = IXGBE_INCVAL_SHIFT_1GB;
6861 		break;
6862 	case ETH_SPEED_NUM_10G:
6863 	default:
6864 		incval = IXGBE_INCVAL_10GB;
6865 		shift = IXGBE_INCVAL_SHIFT_10GB;
6866 		break;
6867 	}
6868 
6869 	switch (hw->mac.type) {
6870 	case ixgbe_mac_X550:
6871 	case ixgbe_mac_X550EM_x:
6872 	case ixgbe_mac_X550EM_a:
6873 		/* Independent of link speed. */
6874 		incval = 1;
6875 		/* Cycles read will be interpreted as ns. */
6876 		shift = 0;
6877 		/* Fall-through */
6878 	case ixgbe_mac_X540:
6879 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
6880 		break;
6881 	case ixgbe_mac_82599EB:
6882 		incval >>= IXGBE_INCVAL_SHIFT_82599;
6883 		shift -= IXGBE_INCVAL_SHIFT_82599;
6884 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
6885 				(1 << IXGBE_INCPER_SHIFT_82599) | incval);
6886 		break;
6887 	default:
6888 		/* Not supported. */
6889 		return;
6890 	}
6891 
6892 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
6893 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6894 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6895 
6896 	adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6897 	adapter->systime_tc.cc_shift = shift;
6898 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
6899 
6900 	adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6901 	adapter->rx_tstamp_tc.cc_shift = shift;
6902 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6903 
6904 	adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6905 	adapter->tx_tstamp_tc.cc_shift = shift;
6906 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6907 }
6908 
6909 static int
6910 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
6911 {
6912 	struct ixgbe_adapter *adapter =
6913 			(struct ixgbe_adapter *)dev->data->dev_private;
6914 
6915 	adapter->systime_tc.nsec += delta;
6916 	adapter->rx_tstamp_tc.nsec += delta;
6917 	adapter->tx_tstamp_tc.nsec += delta;
6918 
6919 	return 0;
6920 }
6921 
6922 static int
6923 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
6924 {
6925 	uint64_t ns;
6926 	struct ixgbe_adapter *adapter =
6927 			(struct ixgbe_adapter *)dev->data->dev_private;
6928 
6929 	ns = rte_timespec_to_ns(ts);
6930 	/* Set the timecounters to a new value. */
6931 	adapter->systime_tc.nsec = ns;
6932 	adapter->rx_tstamp_tc.nsec = ns;
6933 	adapter->tx_tstamp_tc.nsec = ns;
6934 
6935 	return 0;
6936 }
6937 
6938 static int
6939 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
6940 {
6941 	uint64_t ns, systime_cycles;
6942 	struct ixgbe_adapter *adapter =
6943 			(struct ixgbe_adapter *)dev->data->dev_private;
6944 
6945 	systime_cycles = ixgbe_read_systime_cyclecounter(dev);
6946 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
6947 	*ts = rte_ns_to_timespec(ns);
6948 
6949 	return 0;
6950 }
6951 
6952 static int
6953 ixgbe_timesync_enable(struct rte_eth_dev *dev)
6954 {
6955 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6956 	uint32_t tsync_ctl;
6957 	uint32_t tsauxc;
6958 
6959 	/* Stop the timesync system time. */
6960 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
6961 	/* Reset the timesync system time value. */
6962 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
6963 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
6964 
6965 	/* Enable system time for platforms where it isn't on by default. */
6966 	tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
6967 	tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
6968 	IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
6969 
6970 	ixgbe_start_timecounters(dev);
6971 
6972 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6973 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
6974 			(ETHER_TYPE_1588 |
6975 			 IXGBE_ETQF_FILTER_EN |
6976 			 IXGBE_ETQF_1588));
6977 
6978 	/* Enable timestamping of received PTP packets. */
6979 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6980 	tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
6981 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6982 
6983 	/* Enable timestamping of transmitted PTP packets. */
6984 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6985 	tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
6986 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6987 
6988 	IXGBE_WRITE_FLUSH(hw);
6989 
6990 	return 0;
6991 }
6992 
6993 static int
6994 ixgbe_timesync_disable(struct rte_eth_dev *dev)
6995 {
6996 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6997 	uint32_t tsync_ctl;
6998 
6999 	/* Disable timestamping of transmitted PTP packets. */
7000 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7001 	tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
7002 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
7003 
7004 	/* Disable timestamping of received PTP packets. */
7005 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7006 	tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
7007 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
7008 
7009 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
7010 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
7011 
7012 	/* Stop incrementating the System Time registers. */
7013 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
7014 
7015 	return 0;
7016 }
7017 
7018 static int
7019 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
7020 				 struct timespec *timestamp,
7021 				 uint32_t flags __rte_unused)
7022 {
7023 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7024 	struct ixgbe_adapter *adapter =
7025 		(struct ixgbe_adapter *)dev->data->dev_private;
7026 	uint32_t tsync_rxctl;
7027 	uint64_t rx_tstamp_cycles;
7028 	uint64_t ns;
7029 
7030 	tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7031 	if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
7032 		return -EINVAL;
7033 
7034 	rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
7035 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
7036 	*timestamp = rte_ns_to_timespec(ns);
7037 
7038 	return  0;
7039 }
7040 
7041 static int
7042 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
7043 				 struct timespec *timestamp)
7044 {
7045 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7046 	struct ixgbe_adapter *adapter =
7047 		(struct ixgbe_adapter *)dev->data->dev_private;
7048 	uint32_t tsync_txctl;
7049 	uint64_t tx_tstamp_cycles;
7050 	uint64_t ns;
7051 
7052 	tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7053 	if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
7054 		return -EINVAL;
7055 
7056 	tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
7057 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
7058 	*timestamp = rte_ns_to_timespec(ns);
7059 
7060 	return 0;
7061 }
7062 
7063 static int
7064 ixgbe_get_reg_length(struct rte_eth_dev *dev)
7065 {
7066 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7067 	int count = 0;
7068 	int g_ind = 0;
7069 	const struct reg_info *reg_group;
7070 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7071 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7072 
7073 	while ((reg_group = reg_set[g_ind++]))
7074 		count += ixgbe_regs_group_count(reg_group);
7075 
7076 	return count;
7077 }
7078 
7079 static int
7080 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
7081 {
7082 	int count = 0;
7083 	int g_ind = 0;
7084 	const struct reg_info *reg_group;
7085 
7086 	while ((reg_group = ixgbevf_regs[g_ind++]))
7087 		count += ixgbe_regs_group_count(reg_group);
7088 
7089 	return count;
7090 }
7091 
7092 static int
7093 ixgbe_get_regs(struct rte_eth_dev *dev,
7094 	      struct rte_dev_reg_info *regs)
7095 {
7096 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7097 	uint32_t *data = regs->data;
7098 	int g_ind = 0;
7099 	int count = 0;
7100 	const struct reg_info *reg_group;
7101 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7102 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7103 
7104 	if (data == NULL) {
7105 		regs->length = ixgbe_get_reg_length(dev);
7106 		regs->width = sizeof(uint32_t);
7107 		return 0;
7108 	}
7109 
7110 	/* Support only full register dump */
7111 	if ((regs->length == 0) ||
7112 	    (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
7113 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7114 			hw->device_id;
7115 		while ((reg_group = reg_set[g_ind++]))
7116 			count += ixgbe_read_regs_group(dev, &data[count],
7117 				reg_group);
7118 		return 0;
7119 	}
7120 
7121 	return -ENOTSUP;
7122 }
7123 
7124 static int
7125 ixgbevf_get_regs(struct rte_eth_dev *dev,
7126 		struct rte_dev_reg_info *regs)
7127 {
7128 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7129 	uint32_t *data = regs->data;
7130 	int g_ind = 0;
7131 	int count = 0;
7132 	const struct reg_info *reg_group;
7133 
7134 	if (data == NULL) {
7135 		regs->length = ixgbevf_get_reg_length(dev);
7136 		regs->width = sizeof(uint32_t);
7137 		return 0;
7138 	}
7139 
7140 	/* Support only full register dump */
7141 	if ((regs->length == 0) ||
7142 	    (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
7143 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7144 			hw->device_id;
7145 		while ((reg_group = ixgbevf_regs[g_ind++]))
7146 			count += ixgbe_read_regs_group(dev, &data[count],
7147 						      reg_group);
7148 		return 0;
7149 	}
7150 
7151 	return -ENOTSUP;
7152 }
7153 
7154 static int
7155 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
7156 {
7157 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7158 
7159 	/* Return unit is byte count */
7160 	return hw->eeprom.word_size * 2;
7161 }
7162 
7163 static int
7164 ixgbe_get_eeprom(struct rte_eth_dev *dev,
7165 		struct rte_dev_eeprom_info *in_eeprom)
7166 {
7167 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7168 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7169 	uint16_t *data = in_eeprom->data;
7170 	int first, length;
7171 
7172 	first = in_eeprom->offset >> 1;
7173 	length = in_eeprom->length >> 1;
7174 	if ((first > hw->eeprom.word_size) ||
7175 	    ((first + length) > hw->eeprom.word_size))
7176 		return -EINVAL;
7177 
7178 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7179 
7180 	return eeprom->ops.read_buffer(hw, first, length, data);
7181 }
7182 
7183 static int
7184 ixgbe_set_eeprom(struct rte_eth_dev *dev,
7185 		struct rte_dev_eeprom_info *in_eeprom)
7186 {
7187 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7188 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7189 	uint16_t *data = in_eeprom->data;
7190 	int first, length;
7191 
7192 	first = in_eeprom->offset >> 1;
7193 	length = in_eeprom->length >> 1;
7194 	if ((first > hw->eeprom.word_size) ||
7195 	    ((first + length) > hw->eeprom.word_size))
7196 		return -EINVAL;
7197 
7198 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7199 
7200 	return eeprom->ops.write_buffer(hw,  first, length, data);
7201 }
7202 
7203 uint16_t
7204 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
7205 	switch (mac_type) {
7206 	case ixgbe_mac_X550:
7207 	case ixgbe_mac_X550EM_x:
7208 	case ixgbe_mac_X550EM_a:
7209 		return ETH_RSS_RETA_SIZE_512;
7210 	case ixgbe_mac_X550_vf:
7211 	case ixgbe_mac_X550EM_x_vf:
7212 	case ixgbe_mac_X550EM_a_vf:
7213 		return ETH_RSS_RETA_SIZE_64;
7214 	default:
7215 		return ETH_RSS_RETA_SIZE_128;
7216 	}
7217 }
7218 
7219 uint32_t
7220 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
7221 	switch (mac_type) {
7222 	case ixgbe_mac_X550:
7223 	case ixgbe_mac_X550EM_x:
7224 	case ixgbe_mac_X550EM_a:
7225 		if (reta_idx < ETH_RSS_RETA_SIZE_128)
7226 			return IXGBE_RETA(reta_idx >> 2);
7227 		else
7228 			return IXGBE_ERETA((reta_idx - ETH_RSS_RETA_SIZE_128) >> 2);
7229 	case ixgbe_mac_X550_vf:
7230 	case ixgbe_mac_X550EM_x_vf:
7231 	case ixgbe_mac_X550EM_a_vf:
7232 		return IXGBE_VFRETA(reta_idx >> 2);
7233 	default:
7234 		return IXGBE_RETA(reta_idx >> 2);
7235 	}
7236 }
7237 
7238 uint32_t
7239 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
7240 	switch (mac_type) {
7241 	case ixgbe_mac_X550_vf:
7242 	case ixgbe_mac_X550EM_x_vf:
7243 	case ixgbe_mac_X550EM_a_vf:
7244 		return IXGBE_VFMRQC;
7245 	default:
7246 		return IXGBE_MRQC;
7247 	}
7248 }
7249 
7250 uint32_t
7251 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
7252 	switch (mac_type) {
7253 	case ixgbe_mac_X550_vf:
7254 	case ixgbe_mac_X550EM_x_vf:
7255 	case ixgbe_mac_X550EM_a_vf:
7256 		return IXGBE_VFRSSRK(i);
7257 	default:
7258 		return IXGBE_RSSRK(i);
7259 	}
7260 }
7261 
7262 bool
7263 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
7264 	switch (mac_type) {
7265 	case ixgbe_mac_82599_vf:
7266 	case ixgbe_mac_X540_vf:
7267 		return 0;
7268 	default:
7269 		return 1;
7270 	}
7271 }
7272 
7273 static int
7274 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
7275 			struct rte_eth_dcb_info *dcb_info)
7276 {
7277 	struct ixgbe_dcb_config *dcb_config =
7278 			IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
7279 	struct ixgbe_dcb_tc_config *tc;
7280 	struct rte_eth_dcb_tc_queue_mapping *tc_queue;
7281 	uint8_t nb_tcs;
7282 	uint8_t i, j;
7283 
7284 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_DCB_FLAG)
7285 		dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
7286 	else
7287 		dcb_info->nb_tcs = 1;
7288 
7289 	tc_queue = &dcb_info->tc_queue;
7290 	nb_tcs = dcb_info->nb_tcs;
7291 
7292 	if (dcb_config->vt_mode) { /* vt is enabled*/
7293 		struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
7294 				&dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
7295 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7296 			dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
7297 		if (RTE_ETH_DEV_SRIOV(dev).active > 0) {
7298 			for (j = 0; j < nb_tcs; j++) {
7299 				tc_queue->tc_rxq[0][j].base = j;
7300 				tc_queue->tc_rxq[0][j].nb_queue = 1;
7301 				tc_queue->tc_txq[0][j].base = j;
7302 				tc_queue->tc_txq[0][j].nb_queue = 1;
7303 			}
7304 		} else {
7305 			for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
7306 				for (j = 0; j < nb_tcs; j++) {
7307 					tc_queue->tc_rxq[i][j].base =
7308 						i * nb_tcs + j;
7309 					tc_queue->tc_rxq[i][j].nb_queue = 1;
7310 					tc_queue->tc_txq[i][j].base =
7311 						i * nb_tcs + j;
7312 					tc_queue->tc_txq[i][j].nb_queue = 1;
7313 				}
7314 			}
7315 		}
7316 	} else { /* vt is disabled*/
7317 		struct rte_eth_dcb_rx_conf *rx_conf =
7318 				&dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
7319 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7320 			dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
7321 		if (dcb_info->nb_tcs == ETH_4_TCS) {
7322 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7323 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
7324 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7325 			}
7326 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7327 			dcb_info->tc_queue.tc_txq[0][1].base = 64;
7328 			dcb_info->tc_queue.tc_txq[0][2].base = 96;
7329 			dcb_info->tc_queue.tc_txq[0][3].base = 112;
7330 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
7331 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7332 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7333 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7334 		} else if (dcb_info->nb_tcs == ETH_8_TCS) {
7335 			for (i = 0; i < dcb_info->nb_tcs; i++) {
7336 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
7337 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7338 			}
7339 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
7340 			dcb_info->tc_queue.tc_txq[0][1].base = 32;
7341 			dcb_info->tc_queue.tc_txq[0][2].base = 64;
7342 			dcb_info->tc_queue.tc_txq[0][3].base = 80;
7343 			dcb_info->tc_queue.tc_txq[0][4].base = 96;
7344 			dcb_info->tc_queue.tc_txq[0][5].base = 104;
7345 			dcb_info->tc_queue.tc_txq[0][6].base = 112;
7346 			dcb_info->tc_queue.tc_txq[0][7].base = 120;
7347 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
7348 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7349 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7350 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7351 			dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
7352 			dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
7353 			dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
7354 			dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
7355 		}
7356 	}
7357 	for (i = 0; i < dcb_info->nb_tcs; i++) {
7358 		tc = &dcb_config->tc_config[i];
7359 		dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
7360 	}
7361 	return 0;
7362 }
7363 
7364 /* Update e-tag ether type */
7365 static int
7366 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
7367 			    uint16_t ether_type)
7368 {
7369 	uint32_t etag_etype;
7370 
7371 	if (hw->mac.type != ixgbe_mac_X550 &&
7372 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7373 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7374 		return -ENOTSUP;
7375 	}
7376 
7377 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7378 	etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
7379 	etag_etype |= ether_type;
7380 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7381 	IXGBE_WRITE_FLUSH(hw);
7382 
7383 	return 0;
7384 }
7385 
7386 /* Config l2 tunnel ether type */
7387 static int
7388 ixgbe_dev_l2_tunnel_eth_type_conf(struct rte_eth_dev *dev,
7389 				  struct rte_eth_l2_tunnel_conf *l2_tunnel)
7390 {
7391 	int ret = 0;
7392 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7393 	struct ixgbe_l2_tn_info *l2_tn_info =
7394 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7395 
7396 	if (l2_tunnel == NULL)
7397 		return -EINVAL;
7398 
7399 	switch (l2_tunnel->l2_tunnel_type) {
7400 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7401 		l2_tn_info->e_tag_ether_type = l2_tunnel->ether_type;
7402 		ret = ixgbe_update_e_tag_eth_type(hw, l2_tunnel->ether_type);
7403 		break;
7404 	default:
7405 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7406 		ret = -EINVAL;
7407 		break;
7408 	}
7409 
7410 	return ret;
7411 }
7412 
7413 /* Enable e-tag tunnel */
7414 static int
7415 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
7416 {
7417 	uint32_t etag_etype;
7418 
7419 	if (hw->mac.type != ixgbe_mac_X550 &&
7420 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7421 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7422 		return -ENOTSUP;
7423 	}
7424 
7425 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7426 	etag_etype |= IXGBE_ETAG_ETYPE_VALID;
7427 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7428 	IXGBE_WRITE_FLUSH(hw);
7429 
7430 	return 0;
7431 }
7432 
7433 /* Enable l2 tunnel */
7434 static int
7435 ixgbe_dev_l2_tunnel_enable(struct rte_eth_dev *dev,
7436 			   enum rte_eth_tunnel_type l2_tunnel_type)
7437 {
7438 	int ret = 0;
7439 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7440 	struct ixgbe_l2_tn_info *l2_tn_info =
7441 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7442 
7443 	switch (l2_tunnel_type) {
7444 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7445 		l2_tn_info->e_tag_en = TRUE;
7446 		ret = ixgbe_e_tag_enable(hw);
7447 		break;
7448 	default:
7449 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7450 		ret = -EINVAL;
7451 		break;
7452 	}
7453 
7454 	return ret;
7455 }
7456 
7457 /* Disable e-tag tunnel */
7458 static int
7459 ixgbe_e_tag_disable(struct ixgbe_hw *hw)
7460 {
7461 	uint32_t etag_etype;
7462 
7463 	if (hw->mac.type != ixgbe_mac_X550 &&
7464 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7465 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7466 		return -ENOTSUP;
7467 	}
7468 
7469 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7470 	etag_etype &= ~IXGBE_ETAG_ETYPE_VALID;
7471 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7472 	IXGBE_WRITE_FLUSH(hw);
7473 
7474 	return 0;
7475 }
7476 
7477 /* Disable l2 tunnel */
7478 static int
7479 ixgbe_dev_l2_tunnel_disable(struct rte_eth_dev *dev,
7480 			    enum rte_eth_tunnel_type l2_tunnel_type)
7481 {
7482 	int ret = 0;
7483 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7484 	struct ixgbe_l2_tn_info *l2_tn_info =
7485 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7486 
7487 	switch (l2_tunnel_type) {
7488 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7489 		l2_tn_info->e_tag_en = FALSE;
7490 		ret = ixgbe_e_tag_disable(hw);
7491 		break;
7492 	default:
7493 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7494 		ret = -EINVAL;
7495 		break;
7496 	}
7497 
7498 	return ret;
7499 }
7500 
7501 static int
7502 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
7503 		       struct rte_eth_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 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7517 
7518 	for (i = 1; i < rar_entries; i++) {
7519 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7520 		rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
7521 		if ((rar_high & IXGBE_RAH_AV) &&
7522 		    (rar_high & IXGBE_RAH_ADTYPE) &&
7523 		    ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
7524 		     l2_tunnel->tunnel_id)) {
7525 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
7526 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
7527 
7528 			ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
7529 
7530 			return ret;
7531 		}
7532 	}
7533 
7534 	return ret;
7535 }
7536 
7537 static int
7538 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
7539 		       struct rte_eth_l2_tunnel_conf *l2_tunnel)
7540 {
7541 	int ret = 0;
7542 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7543 	uint32_t i, rar_entries;
7544 	uint32_t rar_low, rar_high;
7545 
7546 	if (hw->mac.type != ixgbe_mac_X550 &&
7547 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7548 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7549 		return -ENOTSUP;
7550 	}
7551 
7552 	/* One entry for one tunnel. Try to remove potential existing entry. */
7553 	ixgbe_e_tag_filter_del(dev, l2_tunnel);
7554 
7555 	rar_entries = ixgbe_get_num_rx_addrs(hw);
7556 
7557 	for (i = 1; i < rar_entries; i++) {
7558 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7559 		if (rar_high & IXGBE_RAH_AV) {
7560 			continue;
7561 		} else {
7562 			ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
7563 			rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
7564 			rar_low = l2_tunnel->tunnel_id;
7565 
7566 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
7567 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
7568 
7569 			return ret;
7570 		}
7571 	}
7572 
7573 	PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
7574 		     " Please remove a rule before adding a new one.");
7575 	return -EINVAL;
7576 }
7577 
7578 static inline struct ixgbe_l2_tn_filter *
7579 ixgbe_l2_tn_filter_lookup(struct ixgbe_l2_tn_info *l2_tn_info,
7580 			  struct ixgbe_l2_tn_key *key)
7581 {
7582 	int ret;
7583 
7584 	ret = rte_hash_lookup(l2_tn_info->hash_handle, (const void *)key);
7585 	if (ret < 0)
7586 		return NULL;
7587 
7588 	return l2_tn_info->hash_map[ret];
7589 }
7590 
7591 static inline int
7592 ixgbe_insert_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7593 			  struct ixgbe_l2_tn_filter *l2_tn_filter)
7594 {
7595 	int ret;
7596 
7597 	ret = rte_hash_add_key(l2_tn_info->hash_handle,
7598 			       &l2_tn_filter->key);
7599 
7600 	if (ret < 0) {
7601 		PMD_DRV_LOG(ERR,
7602 			    "Failed to insert L2 tunnel filter"
7603 			    " to hash table %d!",
7604 			    ret);
7605 		return ret;
7606 	}
7607 
7608 	l2_tn_info->hash_map[ret] = l2_tn_filter;
7609 
7610 	TAILQ_INSERT_TAIL(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7611 
7612 	return 0;
7613 }
7614 
7615 static inline int
7616 ixgbe_remove_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7617 			  struct ixgbe_l2_tn_key *key)
7618 {
7619 	int ret;
7620 	struct ixgbe_l2_tn_filter *l2_tn_filter;
7621 
7622 	ret = rte_hash_del_key(l2_tn_info->hash_handle, key);
7623 
7624 	if (ret < 0) {
7625 		PMD_DRV_LOG(ERR,
7626 			    "No such L2 tunnel filter to delete %d!",
7627 			    ret);
7628 		return ret;
7629 	}
7630 
7631 	l2_tn_filter = l2_tn_info->hash_map[ret];
7632 	l2_tn_info->hash_map[ret] = NULL;
7633 
7634 	TAILQ_REMOVE(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7635 	rte_free(l2_tn_filter);
7636 
7637 	return 0;
7638 }
7639 
7640 /* Add l2 tunnel filter */
7641 int
7642 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
7643 			       struct rte_eth_l2_tunnel_conf *l2_tunnel,
7644 			       bool restore)
7645 {
7646 	int ret;
7647 	struct ixgbe_l2_tn_info *l2_tn_info =
7648 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7649 	struct ixgbe_l2_tn_key key;
7650 	struct ixgbe_l2_tn_filter *node;
7651 
7652 	if (!restore) {
7653 		key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7654 		key.tn_id = l2_tunnel->tunnel_id;
7655 
7656 		node = ixgbe_l2_tn_filter_lookup(l2_tn_info, &key);
7657 
7658 		if (node) {
7659 			PMD_DRV_LOG(ERR,
7660 				    "The L2 tunnel filter already exists!");
7661 			return -EINVAL;
7662 		}
7663 
7664 		node = rte_zmalloc("ixgbe_l2_tn",
7665 				   sizeof(struct ixgbe_l2_tn_filter),
7666 				   0);
7667 		if (!node)
7668 			return -ENOMEM;
7669 
7670 		rte_memcpy(&node->key,
7671 				 &key,
7672 				 sizeof(struct ixgbe_l2_tn_key));
7673 		node->pool = l2_tunnel->pool;
7674 		ret = ixgbe_insert_l2_tn_filter(l2_tn_info, node);
7675 		if (ret < 0) {
7676 			rte_free(node);
7677 			return ret;
7678 		}
7679 	}
7680 
7681 	switch (l2_tunnel->l2_tunnel_type) {
7682 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7683 		ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
7684 		break;
7685 	default:
7686 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7687 		ret = -EINVAL;
7688 		break;
7689 	}
7690 
7691 	if ((!restore) && (ret < 0))
7692 		(void)ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7693 
7694 	return ret;
7695 }
7696 
7697 /* Delete l2 tunnel filter */
7698 int
7699 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
7700 			       struct rte_eth_l2_tunnel_conf *l2_tunnel)
7701 {
7702 	int ret;
7703 	struct ixgbe_l2_tn_info *l2_tn_info =
7704 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7705 	struct ixgbe_l2_tn_key key;
7706 
7707 	key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7708 	key.tn_id = l2_tunnel->tunnel_id;
7709 	ret = ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7710 	if (ret < 0)
7711 		return ret;
7712 
7713 	switch (l2_tunnel->l2_tunnel_type) {
7714 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7715 		ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
7716 		break;
7717 	default:
7718 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7719 		ret = -EINVAL;
7720 		break;
7721 	}
7722 
7723 	return ret;
7724 }
7725 
7726 /**
7727  * ixgbe_dev_l2_tunnel_filter_handle - Handle operations for l2 tunnel filter.
7728  * @dev: pointer to rte_eth_dev structure
7729  * @filter_op:operation will be taken.
7730  * @arg: a pointer to specific structure corresponding to the filter_op
7731  */
7732 static int
7733 ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
7734 				  enum rte_filter_op filter_op,
7735 				  void *arg)
7736 {
7737 	int ret;
7738 
7739 	if (filter_op == RTE_ETH_FILTER_NOP)
7740 		return 0;
7741 
7742 	if (arg == NULL) {
7743 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
7744 			    filter_op);
7745 		return -EINVAL;
7746 	}
7747 
7748 	switch (filter_op) {
7749 	case RTE_ETH_FILTER_ADD:
7750 		ret = ixgbe_dev_l2_tunnel_filter_add
7751 			(dev,
7752 			 (struct rte_eth_l2_tunnel_conf *)arg,
7753 			 FALSE);
7754 		break;
7755 	case RTE_ETH_FILTER_DELETE:
7756 		ret = ixgbe_dev_l2_tunnel_filter_del
7757 			(dev,
7758 			 (struct rte_eth_l2_tunnel_conf *)arg);
7759 		break;
7760 	default:
7761 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
7762 		ret = -EINVAL;
7763 		break;
7764 	}
7765 	return ret;
7766 }
7767 
7768 static int
7769 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
7770 {
7771 	int ret = 0;
7772 	uint32_t ctrl;
7773 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7774 
7775 	if (hw->mac.type != ixgbe_mac_X550 &&
7776 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7777 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7778 		return -ENOTSUP;
7779 	}
7780 
7781 	ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
7782 	ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
7783 	if (en)
7784 		ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
7785 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
7786 
7787 	return ret;
7788 }
7789 
7790 /* Enable l2 tunnel forwarding */
7791 static int
7792 ixgbe_dev_l2_tunnel_forwarding_enable
7793 	(struct rte_eth_dev *dev,
7794 	 enum rte_eth_tunnel_type l2_tunnel_type)
7795 {
7796 	struct ixgbe_l2_tn_info *l2_tn_info =
7797 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7798 	int ret = 0;
7799 
7800 	switch (l2_tunnel_type) {
7801 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7802 		l2_tn_info->e_tag_fwd_en = TRUE;
7803 		ret = ixgbe_e_tag_forwarding_en_dis(dev, 1);
7804 		break;
7805 	default:
7806 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7807 		ret = -EINVAL;
7808 		break;
7809 	}
7810 
7811 	return ret;
7812 }
7813 
7814 /* Disable l2 tunnel forwarding */
7815 static int
7816 ixgbe_dev_l2_tunnel_forwarding_disable
7817 	(struct rte_eth_dev *dev,
7818 	 enum rte_eth_tunnel_type l2_tunnel_type)
7819 {
7820 	struct ixgbe_l2_tn_info *l2_tn_info =
7821 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7822 	int ret = 0;
7823 
7824 	switch (l2_tunnel_type) {
7825 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7826 		l2_tn_info->e_tag_fwd_en = FALSE;
7827 		ret = ixgbe_e_tag_forwarding_en_dis(dev, 0);
7828 		break;
7829 	default:
7830 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7831 		ret = -EINVAL;
7832 		break;
7833 	}
7834 
7835 	return ret;
7836 }
7837 
7838 static int
7839 ixgbe_e_tag_insertion_en_dis(struct rte_eth_dev *dev,
7840 			     struct rte_eth_l2_tunnel_conf *l2_tunnel,
7841 			     bool en)
7842 {
7843 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
7844 	int ret = 0;
7845 	uint32_t vmtir, vmvir;
7846 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7847 
7848 	if (l2_tunnel->vf_id >= pci_dev->max_vfs) {
7849 		PMD_DRV_LOG(ERR,
7850 			    "VF id %u should be less than %u",
7851 			    l2_tunnel->vf_id,
7852 			    pci_dev->max_vfs);
7853 		return -EINVAL;
7854 	}
7855 
7856 	if (hw->mac.type != ixgbe_mac_X550 &&
7857 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7858 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7859 		return -ENOTSUP;
7860 	}
7861 
7862 	if (en)
7863 		vmtir = l2_tunnel->tunnel_id;
7864 	else
7865 		vmtir = 0;
7866 
7867 	IXGBE_WRITE_REG(hw, IXGBE_VMTIR(l2_tunnel->vf_id), vmtir);
7868 
7869 	vmvir = IXGBE_READ_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id));
7870 	vmvir &= ~IXGBE_VMVIR_TAGA_MASK;
7871 	if (en)
7872 		vmvir |= IXGBE_VMVIR_TAGA_ETAG_INSERT;
7873 	IXGBE_WRITE_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id), vmvir);
7874 
7875 	return ret;
7876 }
7877 
7878 /* Enable l2 tunnel tag insertion */
7879 static int
7880 ixgbe_dev_l2_tunnel_insertion_enable(struct rte_eth_dev *dev,
7881 				     struct rte_eth_l2_tunnel_conf *l2_tunnel)
7882 {
7883 	int ret = 0;
7884 
7885 	switch (l2_tunnel->l2_tunnel_type) {
7886 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7887 		ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 1);
7888 		break;
7889 	default:
7890 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7891 		ret = -EINVAL;
7892 		break;
7893 	}
7894 
7895 	return ret;
7896 }
7897 
7898 /* Disable l2 tunnel tag insertion */
7899 static int
7900 ixgbe_dev_l2_tunnel_insertion_disable
7901 	(struct rte_eth_dev *dev,
7902 	 struct rte_eth_l2_tunnel_conf *l2_tunnel)
7903 {
7904 	int ret = 0;
7905 
7906 	switch (l2_tunnel->l2_tunnel_type) {
7907 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7908 		ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 0);
7909 		break;
7910 	default:
7911 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7912 		ret = -EINVAL;
7913 		break;
7914 	}
7915 
7916 	return ret;
7917 }
7918 
7919 static int
7920 ixgbe_e_tag_stripping_en_dis(struct rte_eth_dev *dev,
7921 			     bool en)
7922 {
7923 	int ret = 0;
7924 	uint32_t qde;
7925 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7926 
7927 	if (hw->mac.type != ixgbe_mac_X550 &&
7928 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7929 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7930 		return -ENOTSUP;
7931 	}
7932 
7933 	qde = IXGBE_READ_REG(hw, IXGBE_QDE);
7934 	if (en)
7935 		qde |= IXGBE_QDE_STRIP_TAG;
7936 	else
7937 		qde &= ~IXGBE_QDE_STRIP_TAG;
7938 	qde &= ~IXGBE_QDE_READ;
7939 	qde |= IXGBE_QDE_WRITE;
7940 	IXGBE_WRITE_REG(hw, IXGBE_QDE, qde);
7941 
7942 	return ret;
7943 }
7944 
7945 /* Enable l2 tunnel tag stripping */
7946 static int
7947 ixgbe_dev_l2_tunnel_stripping_enable
7948 	(struct rte_eth_dev *dev,
7949 	 enum rte_eth_tunnel_type l2_tunnel_type)
7950 {
7951 	int ret = 0;
7952 
7953 	switch (l2_tunnel_type) {
7954 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7955 		ret = ixgbe_e_tag_stripping_en_dis(dev, 1);
7956 		break;
7957 	default:
7958 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7959 		ret = -EINVAL;
7960 		break;
7961 	}
7962 
7963 	return ret;
7964 }
7965 
7966 /* Disable l2 tunnel tag stripping */
7967 static int
7968 ixgbe_dev_l2_tunnel_stripping_disable
7969 	(struct rte_eth_dev *dev,
7970 	 enum rte_eth_tunnel_type l2_tunnel_type)
7971 {
7972 	int ret = 0;
7973 
7974 	switch (l2_tunnel_type) {
7975 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7976 		ret = ixgbe_e_tag_stripping_en_dis(dev, 0);
7977 		break;
7978 	default:
7979 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7980 		ret = -EINVAL;
7981 		break;
7982 	}
7983 
7984 	return ret;
7985 }
7986 
7987 /* Enable/disable l2 tunnel offload functions */
7988 static int
7989 ixgbe_dev_l2_tunnel_offload_set
7990 	(struct rte_eth_dev *dev,
7991 	 struct rte_eth_l2_tunnel_conf *l2_tunnel,
7992 	 uint32_t mask,
7993 	 uint8_t en)
7994 {
7995 	int ret = 0;
7996 
7997 	if (l2_tunnel == NULL)
7998 		return -EINVAL;
7999 
8000 	ret = -EINVAL;
8001 	if (mask & ETH_L2_TUNNEL_ENABLE_MASK) {
8002 		if (en)
8003 			ret = ixgbe_dev_l2_tunnel_enable(
8004 				dev,
8005 				l2_tunnel->l2_tunnel_type);
8006 		else
8007 			ret = ixgbe_dev_l2_tunnel_disable(
8008 				dev,
8009 				l2_tunnel->l2_tunnel_type);
8010 	}
8011 
8012 	if (mask & ETH_L2_TUNNEL_INSERTION_MASK) {
8013 		if (en)
8014 			ret = ixgbe_dev_l2_tunnel_insertion_enable(
8015 				dev,
8016 				l2_tunnel);
8017 		else
8018 			ret = ixgbe_dev_l2_tunnel_insertion_disable(
8019 				dev,
8020 				l2_tunnel);
8021 	}
8022 
8023 	if (mask & ETH_L2_TUNNEL_STRIPPING_MASK) {
8024 		if (en)
8025 			ret = ixgbe_dev_l2_tunnel_stripping_enable(
8026 				dev,
8027 				l2_tunnel->l2_tunnel_type);
8028 		else
8029 			ret = ixgbe_dev_l2_tunnel_stripping_disable(
8030 				dev,
8031 				l2_tunnel->l2_tunnel_type);
8032 	}
8033 
8034 	if (mask & ETH_L2_TUNNEL_FORWARDING_MASK) {
8035 		if (en)
8036 			ret = ixgbe_dev_l2_tunnel_forwarding_enable(
8037 				dev,
8038 				l2_tunnel->l2_tunnel_type);
8039 		else
8040 			ret = ixgbe_dev_l2_tunnel_forwarding_disable(
8041 				dev,
8042 				l2_tunnel->l2_tunnel_type);
8043 	}
8044 
8045 	return ret;
8046 }
8047 
8048 static int
8049 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
8050 			uint16_t port)
8051 {
8052 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
8053 	IXGBE_WRITE_FLUSH(hw);
8054 
8055 	return 0;
8056 }
8057 
8058 /* There's only one register for VxLAN UDP port.
8059  * So, we cannot add several ports. Will update it.
8060  */
8061 static int
8062 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
8063 		     uint16_t port)
8064 {
8065 	if (port == 0) {
8066 		PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
8067 		return -EINVAL;
8068 	}
8069 
8070 	return ixgbe_update_vxlan_port(hw, port);
8071 }
8072 
8073 /* We cannot delete the VxLAN port. For there's a register for VxLAN
8074  * UDP port, it must have a value.
8075  * So, will reset it to the original value 0.
8076  */
8077 static int
8078 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
8079 		     uint16_t port)
8080 {
8081 	uint16_t cur_port;
8082 
8083 	cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
8084 
8085 	if (cur_port != port) {
8086 		PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
8087 		return -EINVAL;
8088 	}
8089 
8090 	return ixgbe_update_vxlan_port(hw, 0);
8091 }
8092 
8093 /* Add UDP tunneling port */
8094 static int
8095 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
8096 			      struct rte_eth_udp_tunnel *udp_tunnel)
8097 {
8098 	int ret = 0;
8099 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8100 
8101 	if (hw->mac.type != ixgbe_mac_X550 &&
8102 	    hw->mac.type != ixgbe_mac_X550EM_x &&
8103 	    hw->mac.type != ixgbe_mac_X550EM_a) {
8104 		return -ENOTSUP;
8105 	}
8106 
8107 	if (udp_tunnel == NULL)
8108 		return -EINVAL;
8109 
8110 	switch (udp_tunnel->prot_type) {
8111 	case RTE_TUNNEL_TYPE_VXLAN:
8112 		ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
8113 		break;
8114 
8115 	case RTE_TUNNEL_TYPE_GENEVE:
8116 	case RTE_TUNNEL_TYPE_TEREDO:
8117 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
8118 		ret = -EINVAL;
8119 		break;
8120 
8121 	default:
8122 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
8123 		ret = -EINVAL;
8124 		break;
8125 	}
8126 
8127 	return ret;
8128 }
8129 
8130 /* Remove UDP tunneling port */
8131 static int
8132 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
8133 			      struct rte_eth_udp_tunnel *udp_tunnel)
8134 {
8135 	int ret = 0;
8136 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8137 
8138 	if (hw->mac.type != ixgbe_mac_X550 &&
8139 	    hw->mac.type != ixgbe_mac_X550EM_x &&
8140 	    hw->mac.type != ixgbe_mac_X550EM_a) {
8141 		return -ENOTSUP;
8142 	}
8143 
8144 	if (udp_tunnel == NULL)
8145 		return -EINVAL;
8146 
8147 	switch (udp_tunnel->prot_type) {
8148 	case RTE_TUNNEL_TYPE_VXLAN:
8149 		ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
8150 		break;
8151 	case RTE_TUNNEL_TYPE_GENEVE:
8152 	case RTE_TUNNEL_TYPE_TEREDO:
8153 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
8154 		ret = -EINVAL;
8155 		break;
8156 	default:
8157 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
8158 		ret = -EINVAL;
8159 		break;
8160 	}
8161 
8162 	return ret;
8163 }
8164 
8165 static void
8166 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
8167 {
8168 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8169 
8170 	hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_ALLMULTI);
8171 }
8172 
8173 static void
8174 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
8175 {
8176 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8177 
8178 	hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_MULTI);
8179 }
8180 
8181 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
8182 {
8183 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8184 	u32 in_msg = 0;
8185 
8186 	if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
8187 		return;
8188 
8189 	/* PF reset VF event */
8190 	if (in_msg == IXGBE_PF_CONTROL_MSG)
8191 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
8192 					      NULL, NULL);
8193 }
8194 
8195 static int
8196 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
8197 {
8198 	uint32_t eicr;
8199 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8200 	struct ixgbe_interrupt *intr =
8201 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8202 	ixgbevf_intr_disable(hw);
8203 
8204 	/* read-on-clear nic registers here */
8205 	eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
8206 	intr->flags = 0;
8207 
8208 	/* only one misc vector supported - mailbox */
8209 	eicr &= IXGBE_VTEICR_MASK;
8210 	if (eicr == IXGBE_MISC_VEC_ID)
8211 		intr->flags |= IXGBE_FLAG_MAILBOX;
8212 
8213 	return 0;
8214 }
8215 
8216 static int
8217 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
8218 {
8219 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8220 	struct ixgbe_interrupt *intr =
8221 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8222 
8223 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
8224 		ixgbevf_mbx_process(dev);
8225 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
8226 	}
8227 
8228 	ixgbevf_intr_enable(hw);
8229 
8230 	return 0;
8231 }
8232 
8233 static void
8234 ixgbevf_dev_interrupt_handler(void *param)
8235 {
8236 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
8237 
8238 	ixgbevf_dev_interrupt_get_status(dev);
8239 	ixgbevf_dev_interrupt_action(dev);
8240 }
8241 
8242 /**
8243  *  ixgbe_disable_sec_tx_path_generic - Stops the transmit data path
8244  *  @hw: pointer to hardware structure
8245  *
8246  *  Stops the transmit data path and waits for the HW to internally empty
8247  *  the Tx security block
8248  **/
8249 int ixgbe_disable_sec_tx_path_generic(struct ixgbe_hw *hw)
8250 {
8251 #define IXGBE_MAX_SECTX_POLL 40
8252 
8253 	int i;
8254 	int sectxreg;
8255 
8256 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8257 	sectxreg |= IXGBE_SECTXCTRL_TX_DIS;
8258 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8259 	for (i = 0; i < IXGBE_MAX_SECTX_POLL; i++) {
8260 		sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXSTAT);
8261 		if (sectxreg & IXGBE_SECTXSTAT_SECTX_RDY)
8262 			break;
8263 		/* Use interrupt-safe sleep just in case */
8264 		usec_delay(1000);
8265 	}
8266 
8267 	/* For informational purposes only */
8268 	if (i >= IXGBE_MAX_SECTX_POLL)
8269 		PMD_DRV_LOG(DEBUG, "Tx unit being enabled before security "
8270 			 "path fully disabled.  Continuing with init.");
8271 
8272 	return IXGBE_SUCCESS;
8273 }
8274 
8275 /**
8276  *  ixgbe_enable_sec_tx_path_generic - Enables the transmit data path
8277  *  @hw: pointer to hardware structure
8278  *
8279  *  Enables the transmit data path.
8280  **/
8281 int ixgbe_enable_sec_tx_path_generic(struct ixgbe_hw *hw)
8282 {
8283 	uint32_t sectxreg;
8284 
8285 	sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8286 	sectxreg &= ~IXGBE_SECTXCTRL_TX_DIS;
8287 	IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8288 	IXGBE_WRITE_FLUSH(hw);
8289 
8290 	return IXGBE_SUCCESS;
8291 }
8292 
8293 /* restore n-tuple filter */
8294 static inline void
8295 ixgbe_ntuple_filter_restore(struct rte_eth_dev *dev)
8296 {
8297 	struct ixgbe_filter_info *filter_info =
8298 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8299 	struct ixgbe_5tuple_filter *node;
8300 
8301 	TAILQ_FOREACH(node, &filter_info->fivetuple_list, entries) {
8302 		ixgbe_inject_5tuple_filter(dev, node);
8303 	}
8304 }
8305 
8306 /* restore ethernet type filter */
8307 static inline void
8308 ixgbe_ethertype_filter_restore(struct rte_eth_dev *dev)
8309 {
8310 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8311 	struct ixgbe_filter_info *filter_info =
8312 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8313 	int i;
8314 
8315 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8316 		if (filter_info->ethertype_mask & (1 << i)) {
8317 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i),
8318 					filter_info->ethertype_filters[i].etqf);
8319 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i),
8320 					filter_info->ethertype_filters[i].etqs);
8321 			IXGBE_WRITE_FLUSH(hw);
8322 		}
8323 	}
8324 }
8325 
8326 /* restore SYN filter */
8327 static inline void
8328 ixgbe_syn_filter_restore(struct rte_eth_dev *dev)
8329 {
8330 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8331 	struct ixgbe_filter_info *filter_info =
8332 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8333 	uint32_t synqf;
8334 
8335 	synqf = filter_info->syn_info;
8336 
8337 	if (synqf & IXGBE_SYN_FILTER_ENABLE) {
8338 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
8339 		IXGBE_WRITE_FLUSH(hw);
8340 	}
8341 }
8342 
8343 /* restore L2 tunnel filter */
8344 static inline void
8345 ixgbe_l2_tn_filter_restore(struct rte_eth_dev *dev)
8346 {
8347 	struct ixgbe_l2_tn_info *l2_tn_info =
8348 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8349 	struct ixgbe_l2_tn_filter *node;
8350 	struct rte_eth_l2_tunnel_conf l2_tn_conf;
8351 
8352 	TAILQ_FOREACH(node, &l2_tn_info->l2_tn_list, entries) {
8353 		l2_tn_conf.l2_tunnel_type = node->key.l2_tn_type;
8354 		l2_tn_conf.tunnel_id      = node->key.tn_id;
8355 		l2_tn_conf.pool           = node->pool;
8356 		(void)ixgbe_dev_l2_tunnel_filter_add(dev, &l2_tn_conf, TRUE);
8357 	}
8358 }
8359 
8360 static int
8361 ixgbe_filter_restore(struct rte_eth_dev *dev)
8362 {
8363 	ixgbe_ntuple_filter_restore(dev);
8364 	ixgbe_ethertype_filter_restore(dev);
8365 	ixgbe_syn_filter_restore(dev);
8366 	ixgbe_fdir_filter_restore(dev);
8367 	ixgbe_l2_tn_filter_restore(dev);
8368 
8369 	return 0;
8370 }
8371 
8372 static void
8373 ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev)
8374 {
8375 	struct ixgbe_l2_tn_info *l2_tn_info =
8376 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8377 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8378 
8379 	if (l2_tn_info->e_tag_en)
8380 		(void)ixgbe_e_tag_enable(hw);
8381 
8382 	if (l2_tn_info->e_tag_fwd_en)
8383 		(void)ixgbe_e_tag_forwarding_en_dis(dev, 1);
8384 
8385 	(void)ixgbe_update_e_tag_eth_type(hw, l2_tn_info->e_tag_ether_type);
8386 }
8387 
8388 /* remove all the n-tuple filters */
8389 void
8390 ixgbe_clear_all_ntuple_filter(struct rte_eth_dev *dev)
8391 {
8392 	struct ixgbe_filter_info *filter_info =
8393 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8394 	struct ixgbe_5tuple_filter *p_5tuple;
8395 
8396 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
8397 		ixgbe_remove_5tuple_filter(dev, p_5tuple);
8398 }
8399 
8400 /* remove all the ether type filters */
8401 void
8402 ixgbe_clear_all_ethertype_filter(struct rte_eth_dev *dev)
8403 {
8404 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8405 	struct ixgbe_filter_info *filter_info =
8406 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8407 	int i;
8408 
8409 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8410 		if (filter_info->ethertype_mask & (1 << i) &&
8411 		    !filter_info->ethertype_filters[i].conf) {
8412 			(void)ixgbe_ethertype_filter_remove(filter_info,
8413 							    (uint8_t)i);
8414 			IXGBE_WRITE_REG(hw, IXGBE_ETQF(i), 0);
8415 			IXGBE_WRITE_REG(hw, IXGBE_ETQS(i), 0);
8416 			IXGBE_WRITE_FLUSH(hw);
8417 		}
8418 	}
8419 }
8420 
8421 /* remove the SYN filter */
8422 void
8423 ixgbe_clear_syn_filter(struct rte_eth_dev *dev)
8424 {
8425 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8426 	struct ixgbe_filter_info *filter_info =
8427 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8428 
8429 	if (filter_info->syn_info & IXGBE_SYN_FILTER_ENABLE) {
8430 		filter_info->syn_info = 0;
8431 
8432 		IXGBE_WRITE_REG(hw, IXGBE_SYNQF, 0);
8433 		IXGBE_WRITE_FLUSH(hw);
8434 	}
8435 }
8436 
8437 /* remove all the L2 tunnel filters */
8438 int
8439 ixgbe_clear_all_l2_tn_filter(struct rte_eth_dev *dev)
8440 {
8441 	struct ixgbe_l2_tn_info *l2_tn_info =
8442 		IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8443 	struct ixgbe_l2_tn_filter *l2_tn_filter;
8444 	struct rte_eth_l2_tunnel_conf l2_tn_conf;
8445 	int ret = 0;
8446 
8447 	while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
8448 		l2_tn_conf.l2_tunnel_type = l2_tn_filter->key.l2_tn_type;
8449 		l2_tn_conf.tunnel_id      = l2_tn_filter->key.tn_id;
8450 		l2_tn_conf.pool           = l2_tn_filter->pool;
8451 		ret = ixgbe_dev_l2_tunnel_filter_del(dev, &l2_tn_conf);
8452 		if (ret < 0)
8453 			return ret;
8454 	}
8455 
8456 	return 0;
8457 }
8458 
8459 RTE_PMD_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd);
8460 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
8461 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe, "* igb_uio | uio_pci_generic | vfio-pci");
8462 RTE_PMD_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd);
8463 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);
8464 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe_vf, "* igb_uio | vfio-pci");
8465