xref: /f-stack/dpdk/drivers/net/ixgbe/ixgbe_ethdev.c (revision a9643ea8)
1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright(c) 2010-2016 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_atomic.h>
52 #include <rte_branch_prediction.h>
53 #include <rte_memory.h>
54 #include <rte_memzone.h>
55 #include <rte_eal.h>
56 #include <rte_alarm.h>
57 #include <rte_ether.h>
58 #include <rte_ethdev.h>
59 #include <rte_atomic.h>
60 #include <rte_malloc.h>
61 #include <rte_random.h>
62 #include <rte_dev.h>
63 
64 #include "ixgbe_logs.h"
65 #include "base/ixgbe_api.h"
66 #include "base/ixgbe_vf.h"
67 #include "base/ixgbe_common.h"
68 #include "ixgbe_ethdev.h"
69 #include "ixgbe_bypass.h"
70 #include "ixgbe_rxtx.h"
71 #include "base/ixgbe_type.h"
72 #include "base/ixgbe_phy.h"
73 #include "ixgbe_regs.h"
74 
75 /*
76  * High threshold controlling when to start sending XOFF frames. Must be at
77  * least 8 bytes less than receive packet buffer size. This value is in units
78  * of 1024 bytes.
79  */
80 #define IXGBE_FC_HI    0x80
81 
82 /*
83  * Low threshold controlling when to start sending XON frames. This value is
84  * in units of 1024 bytes.
85  */
86 #define IXGBE_FC_LO    0x40
87 
88 /* Default minimum inter-interrupt interval for EITR configuration */
89 #define IXGBE_MIN_INTER_INTERRUPT_INTERVAL_DEFAULT    0x79E
90 
91 /* Timer value included in XOFF frames. */
92 #define IXGBE_FC_PAUSE 0x680
93 
94 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
95 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
96 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
97 
98 #define IXGBE_MMW_SIZE_DEFAULT        0x4
99 #define IXGBE_MMW_SIZE_JUMBO_FRAME    0x14
100 #define IXGBE_MAX_RING_DESC           4096 /* replicate define from rxtx */
101 
102 /*
103  *  Default values for RX/TX configuration
104  */
105 #define IXGBE_DEFAULT_RX_FREE_THRESH  32
106 #define IXGBE_DEFAULT_RX_PTHRESH      8
107 #define IXGBE_DEFAULT_RX_HTHRESH      8
108 #define IXGBE_DEFAULT_RX_WTHRESH      0
109 
110 #define IXGBE_DEFAULT_TX_FREE_THRESH  32
111 #define IXGBE_DEFAULT_TX_PTHRESH      32
112 #define IXGBE_DEFAULT_TX_HTHRESH      0
113 #define IXGBE_DEFAULT_TX_WTHRESH      0
114 #define IXGBE_DEFAULT_TX_RSBIT_THRESH 32
115 
116 /* Bit shift and mask */
117 #define IXGBE_4_BIT_WIDTH  (CHAR_BIT / 2)
118 #define IXGBE_4_BIT_MASK   RTE_LEN2MASK(IXGBE_4_BIT_WIDTH, uint8_t)
119 #define IXGBE_8_BIT_WIDTH  CHAR_BIT
120 #define IXGBE_8_BIT_MASK   UINT8_MAX
121 
122 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
123 
124 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
125 
126 #define IXGBE_HKEY_MAX_INDEX 10
127 
128 /* Additional timesync values. */
129 #define NSEC_PER_SEC             1000000000L
130 #define IXGBE_INCVAL_10GB        0x66666666
131 #define IXGBE_INCVAL_1GB         0x40000000
132 #define IXGBE_INCVAL_100         0x50000000
133 #define IXGBE_INCVAL_SHIFT_10GB  28
134 #define IXGBE_INCVAL_SHIFT_1GB   24
135 #define IXGBE_INCVAL_SHIFT_100   21
136 #define IXGBE_INCVAL_SHIFT_82599 7
137 #define IXGBE_INCPER_SHIFT_82599 24
138 
139 #define IXGBE_CYCLECOUNTER_MASK   0xffffffffffffffffULL
140 
141 #define IXGBE_VT_CTL_POOLING_MODE_MASK         0x00030000
142 #define IXGBE_VT_CTL_POOLING_MODE_ETAG         0x00010000
143 #define DEFAULT_ETAG_ETYPE                     0x893f
144 #define IXGBE_ETAG_ETYPE                       0x00005084
145 #define IXGBE_ETAG_ETYPE_MASK                  0x0000ffff
146 #define IXGBE_ETAG_ETYPE_VALID                 0x80000000
147 #define IXGBE_RAH_ADTYPE                       0x40000000
148 #define IXGBE_RAL_ETAG_FILTER_MASK             0x00003fff
149 #define IXGBE_VMVIR_TAGA_MASK                  0x18000000
150 #define IXGBE_VMVIR_TAGA_ETAG_INSERT           0x08000000
151 #define IXGBE_VMTIR(_i) (0x00017000 + ((_i) * 4)) /* 64 of these (0-63) */
152 #define IXGBE_QDE_STRIP_TAG                    0x00000004
153 #define IXGBE_VTEICR_MASK                      0x07
154 
155 enum ixgbevf_xcast_modes {
156 	IXGBEVF_XCAST_MODE_NONE = 0,
157 	IXGBEVF_XCAST_MODE_MULTI,
158 	IXGBEVF_XCAST_MODE_ALLMULTI,
159 };
160 
161 #define IXGBE_EXVET_VET_EXT_SHIFT              16
162 #define IXGBE_DMATXCTL_VT_MASK                 0xFFFF0000
163 
164 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev);
165 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
166 static int  ixgbe_dev_configure(struct rte_eth_dev *dev);
167 static int  ixgbe_dev_start(struct rte_eth_dev *dev);
168 static void ixgbe_dev_stop(struct rte_eth_dev *dev);
169 static int  ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
170 static int  ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
171 static void ixgbe_dev_close(struct rte_eth_dev *dev);
172 static void ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
173 static void ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
174 static void ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
175 static void ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
176 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
177 				int wait_to_complete);
178 static void ixgbe_dev_stats_get(struct rte_eth_dev *dev,
179 				struct rte_eth_stats *stats);
180 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
181 				struct rte_eth_xstat *xstats, unsigned n);
182 static int ixgbevf_dev_xstats_get(struct rte_eth_dev *dev,
183 				  struct rte_eth_xstat *xstats, unsigned n);
184 static void ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
185 static void ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
186 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
187 	struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned limit);
188 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
189 	struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned limit);
190 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
191 					     uint16_t queue_id,
192 					     uint8_t stat_idx,
193 					     uint8_t is_rx);
194 static void ixgbe_dev_info_get(struct rte_eth_dev *dev,
195 			       struct rte_eth_dev_info *dev_info);
196 static const uint32_t *ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev);
197 static void ixgbevf_dev_info_get(struct rte_eth_dev *dev,
198 				 struct rte_eth_dev_info *dev_info);
199 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
200 
201 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
202 		uint16_t vlan_id, int on);
203 static int ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
204 			       enum rte_vlan_type vlan_type,
205 			       uint16_t tpid_id);
206 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
207 		uint16_t queue, bool on);
208 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
209 		int on);
210 static void ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
211 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
212 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
213 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
214 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
215 
216 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
217 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
218 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
219 			       struct rte_eth_fc_conf *fc_conf);
220 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
221 			       struct rte_eth_fc_conf *fc_conf);
222 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
223 		struct rte_eth_pfc_conf *pfc_conf);
224 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
225 			struct rte_eth_rss_reta_entry64 *reta_conf,
226 			uint16_t reta_size);
227 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
228 			struct rte_eth_rss_reta_entry64 *reta_conf,
229 			uint16_t reta_size);
230 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
231 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev);
232 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
233 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
234 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev);
235 static void ixgbe_dev_interrupt_handler(struct rte_intr_handle *handle,
236 		void *param);
237 static void ixgbe_dev_interrupt_delayed_handler(void *param);
238 static void ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
239 		uint32_t index, uint32_t pool);
240 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
241 static void ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
242 					   struct ether_addr *mac_addr);
243 static void ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config);
244 
245 /* For Virtual Function support */
246 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
247 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
248 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
249 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
250 static void ixgbevf_dev_stop(struct rte_eth_dev *dev);
251 static void ixgbevf_dev_close(struct rte_eth_dev *dev);
252 static void ixgbevf_intr_disable(struct ixgbe_hw *hw);
253 static void ixgbevf_intr_enable(struct ixgbe_hw *hw);
254 static void ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
255 		struct rte_eth_stats *stats);
256 static void ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
257 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
258 		uint16_t vlan_id, int on);
259 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
260 		uint16_t queue, int on);
261 static void ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
262 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
263 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
264 					    uint16_t queue_id);
265 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
266 					     uint16_t queue_id);
267 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
268 				 uint8_t queue, uint8_t msix_vector);
269 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
270 static void ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev);
271 static void ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev);
272 
273 /* For Eth VMDQ APIs support */
274 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
275 		ether_addr * mac_addr, uint8_t on);
276 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on);
277 static int  ixgbe_set_pool_rx_mode(struct rte_eth_dev *dev,  uint16_t pool,
278 		uint16_t rx_mask, uint8_t on);
279 static int ixgbe_set_pool_rx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on);
280 static int ixgbe_set_pool_tx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on);
281 static int ixgbe_set_pool_vlan_filter(struct rte_eth_dev *dev, uint16_t vlan,
282 		uint64_t pool_mask, uint8_t vlan_on);
283 static int ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
284 		struct rte_eth_mirror_conf *mirror_conf,
285 		uint8_t rule_id, uint8_t on);
286 static int ixgbe_mirror_rule_reset(struct rte_eth_dev *dev,
287 		uint8_t	rule_id);
288 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
289 					  uint16_t queue_id);
290 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
291 					   uint16_t queue_id);
292 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
293 			       uint8_t queue, uint8_t msix_vector);
294 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
295 
296 static int ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
297 		uint16_t queue_idx, uint16_t tx_rate);
298 static int ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
299 		uint16_t tx_rate, uint64_t q_msk);
300 
301 static void ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
302 				 struct ether_addr *mac_addr,
303 				 uint32_t index, uint32_t pool);
304 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
305 static void ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
306 					     struct ether_addr *mac_addr);
307 static int ixgbe_syn_filter_set(struct rte_eth_dev *dev,
308 			struct rte_eth_syn_filter *filter,
309 			bool add);
310 static int ixgbe_syn_filter_get(struct rte_eth_dev *dev,
311 			struct rte_eth_syn_filter *filter);
312 static int ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
313 			enum rte_filter_op filter_op,
314 			void *arg);
315 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
316 			struct ixgbe_5tuple_filter *filter);
317 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
318 			struct ixgbe_5tuple_filter *filter);
319 static int ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
320 			struct rte_eth_ntuple_filter *filter,
321 			bool add);
322 static int ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
323 				enum rte_filter_op filter_op,
324 				void *arg);
325 static int ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
326 			struct rte_eth_ntuple_filter *filter);
327 static int ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
328 			struct rte_eth_ethertype_filter *filter,
329 			bool add);
330 static int ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
331 				enum rte_filter_op filter_op,
332 				void *arg);
333 static int ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
334 			struct rte_eth_ethertype_filter *filter);
335 static int ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
336 		     enum rte_filter_type filter_type,
337 		     enum rte_filter_op filter_op,
338 		     void *arg);
339 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
340 
341 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
342 				      struct ether_addr *mc_addr_set,
343 				      uint32_t nb_mc_addr);
344 static int ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
345 				   struct rte_eth_dcb_info *dcb_info);
346 
347 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
348 static int ixgbe_get_regs(struct rte_eth_dev *dev,
349 			    struct rte_dev_reg_info *regs);
350 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
351 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
352 				struct rte_dev_eeprom_info *eeprom);
353 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
354 				struct rte_dev_eeprom_info *eeprom);
355 
356 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
357 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
358 				struct rte_dev_reg_info *regs);
359 
360 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
361 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
362 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
363 					    struct timespec *timestamp,
364 					    uint32_t flags);
365 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
366 					    struct timespec *timestamp);
367 static int ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
368 static int ixgbe_timesync_read_time(struct rte_eth_dev *dev,
369 				   struct timespec *timestamp);
370 static int ixgbe_timesync_write_time(struct rte_eth_dev *dev,
371 				   const struct timespec *timestamp);
372 static void ixgbevf_dev_interrupt_handler(struct rte_intr_handle *handle,
373 					  void *param);
374 
375 static int ixgbe_dev_l2_tunnel_eth_type_conf
376 	(struct rte_eth_dev *dev, struct rte_eth_l2_tunnel_conf *l2_tunnel);
377 static int ixgbe_dev_l2_tunnel_offload_set
378 	(struct rte_eth_dev *dev,
379 	 struct rte_eth_l2_tunnel_conf *l2_tunnel,
380 	 uint32_t mask,
381 	 uint8_t en);
382 static int ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
383 					     enum rte_filter_op filter_op,
384 					     void *arg);
385 
386 static int ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
387 					 struct rte_eth_udp_tunnel *udp_tunnel);
388 static int ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
389 					 struct rte_eth_udp_tunnel *udp_tunnel);
390 
391 /*
392  * Define VF Stats MACRO for Non "cleared on read" register
393  */
394 #define UPDATE_VF_STAT(reg, last, cur)                          \
395 {                                                               \
396 	uint32_t latest = IXGBE_READ_REG(hw, reg);              \
397 	cur += (latest - last) & UINT_MAX;                      \
398 	last = latest;                                          \
399 }
400 
401 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
402 {                                                                \
403 	u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
404 	u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
405 	u64 latest = ((new_msb << 32) | new_lsb);                \
406 	cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
407 	last = latest;                                           \
408 }
409 
410 #define IXGBE_SET_HWSTRIP(h, q) do {\
411 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
412 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
413 		(h)->bitmap[idx] |= 1 << bit;\
414 	} while (0)
415 
416 #define IXGBE_CLEAR_HWSTRIP(h, q) do {\
417 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
418 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
419 		(h)->bitmap[idx] &= ~(1 << bit);\
420 	} while (0)
421 
422 #define IXGBE_GET_HWSTRIP(h, q, r) do {\
423 		uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
424 		uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
425 		(r) = (h)->bitmap[idx] >> bit & 1;\
426 	} while (0)
427 
428 /*
429  * The set of PCI devices this driver supports
430  */
431 static const struct rte_pci_id pci_id_ixgbe_map[] = {
432 
433 #define RTE_PCI_DEV_ID_DECL_IXGBE(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
434 #include "rte_pci_dev_ids.h"
435 
436 { .vendor_id = 0, /* sentinel */ },
437 };
438 
439 
440 /*
441  * The set of PCI devices this driver supports (for 82599 VF)
442  */
443 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
444 
445 #define RTE_PCI_DEV_ID_DECL_IXGBEVF(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
446 #include "rte_pci_dev_ids.h"
447 { .vendor_id = 0, /* sentinel */ },
448 
449 };
450 
451 static const struct rte_eth_desc_lim rx_desc_lim = {
452 	.nb_max = IXGBE_MAX_RING_DESC,
453 	.nb_min = IXGBE_MIN_RING_DESC,
454 	.nb_align = IXGBE_RXD_ALIGN,
455 };
456 
457 static const struct rte_eth_desc_lim tx_desc_lim = {
458 	.nb_max = IXGBE_MAX_RING_DESC,
459 	.nb_min = IXGBE_MIN_RING_DESC,
460 	.nb_align = IXGBE_TXD_ALIGN,
461 };
462 
463 static const struct eth_dev_ops ixgbe_eth_dev_ops = {
464 	.dev_configure        = ixgbe_dev_configure,
465 	.dev_start            = ixgbe_dev_start,
466 	.dev_stop             = ixgbe_dev_stop,
467 	.dev_set_link_up    = ixgbe_dev_set_link_up,
468 	.dev_set_link_down  = ixgbe_dev_set_link_down,
469 	.dev_close            = ixgbe_dev_close,
470 	.promiscuous_enable   = ixgbe_dev_promiscuous_enable,
471 	.promiscuous_disable  = ixgbe_dev_promiscuous_disable,
472 	.allmulticast_enable  = ixgbe_dev_allmulticast_enable,
473 	.allmulticast_disable = ixgbe_dev_allmulticast_disable,
474 	.link_update          = ixgbe_dev_link_update,
475 	.stats_get            = ixgbe_dev_stats_get,
476 	.xstats_get           = ixgbe_dev_xstats_get,
477 	.stats_reset          = ixgbe_dev_stats_reset,
478 	.xstats_reset         = ixgbe_dev_xstats_reset,
479 	.xstats_get_names     = ixgbe_dev_xstats_get_names,
480 	.queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
481 	.dev_infos_get        = ixgbe_dev_info_get,
482 	.dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
483 	.mtu_set              = ixgbe_dev_mtu_set,
484 	.vlan_filter_set      = ixgbe_vlan_filter_set,
485 	.vlan_tpid_set        = ixgbe_vlan_tpid_set,
486 	.vlan_offload_set     = ixgbe_vlan_offload_set,
487 	.vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
488 	.rx_queue_start	      = ixgbe_dev_rx_queue_start,
489 	.rx_queue_stop        = ixgbe_dev_rx_queue_stop,
490 	.tx_queue_start	      = ixgbe_dev_tx_queue_start,
491 	.tx_queue_stop        = ixgbe_dev_tx_queue_stop,
492 	.rx_queue_setup       = ixgbe_dev_rx_queue_setup,
493 	.rx_queue_intr_enable = ixgbe_dev_rx_queue_intr_enable,
494 	.rx_queue_intr_disable = ixgbe_dev_rx_queue_intr_disable,
495 	.rx_queue_release     = ixgbe_dev_rx_queue_release,
496 	.rx_queue_count       = ixgbe_dev_rx_queue_count,
497 	.rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
498 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
499 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
500 	.dev_led_on           = ixgbe_dev_led_on,
501 	.dev_led_off          = ixgbe_dev_led_off,
502 	.flow_ctrl_get        = ixgbe_flow_ctrl_get,
503 	.flow_ctrl_set        = ixgbe_flow_ctrl_set,
504 	.priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
505 	.mac_addr_add         = ixgbe_add_rar,
506 	.mac_addr_remove      = ixgbe_remove_rar,
507 	.mac_addr_set         = ixgbe_set_default_mac_addr,
508 	.uc_hash_table_set    = ixgbe_uc_hash_table_set,
509 	.uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
510 	.mirror_rule_set      = ixgbe_mirror_rule_set,
511 	.mirror_rule_reset    = ixgbe_mirror_rule_reset,
512 	.set_vf_rx_mode       = ixgbe_set_pool_rx_mode,
513 	.set_vf_rx            = ixgbe_set_pool_rx,
514 	.set_vf_tx            = ixgbe_set_pool_tx,
515 	.set_vf_vlan_filter   = ixgbe_set_pool_vlan_filter,
516 	.set_queue_rate_limit = ixgbe_set_queue_rate_limit,
517 	.set_vf_rate_limit    = ixgbe_set_vf_rate_limit,
518 	.reta_update          = ixgbe_dev_rss_reta_update,
519 	.reta_query           = ixgbe_dev_rss_reta_query,
520 #ifdef RTE_NIC_BYPASS
521 	.bypass_init          = ixgbe_bypass_init,
522 	.bypass_state_set     = ixgbe_bypass_state_store,
523 	.bypass_state_show    = ixgbe_bypass_state_show,
524 	.bypass_event_set     = ixgbe_bypass_event_store,
525 	.bypass_event_show    = ixgbe_bypass_event_show,
526 	.bypass_wd_timeout_set  = ixgbe_bypass_wd_timeout_store,
527 	.bypass_wd_timeout_show = ixgbe_bypass_wd_timeout_show,
528 	.bypass_ver_show      = ixgbe_bypass_ver_show,
529 	.bypass_wd_reset      = ixgbe_bypass_wd_reset,
530 #endif /* RTE_NIC_BYPASS */
531 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
532 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
533 	.filter_ctrl          = ixgbe_dev_filter_ctrl,
534 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
535 	.rxq_info_get         = ixgbe_rxq_info_get,
536 	.txq_info_get         = ixgbe_txq_info_get,
537 	.timesync_enable      = ixgbe_timesync_enable,
538 	.timesync_disable     = ixgbe_timesync_disable,
539 	.timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
540 	.timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
541 	.get_reg              = ixgbe_get_regs,
542 	.get_eeprom_length    = ixgbe_get_eeprom_length,
543 	.get_eeprom           = ixgbe_get_eeprom,
544 	.set_eeprom           = ixgbe_set_eeprom,
545 	.get_dcb_info         = ixgbe_dev_get_dcb_info,
546 	.timesync_adjust_time = ixgbe_timesync_adjust_time,
547 	.timesync_read_time   = ixgbe_timesync_read_time,
548 	.timesync_write_time  = ixgbe_timesync_write_time,
549 	.l2_tunnel_eth_type_conf = ixgbe_dev_l2_tunnel_eth_type_conf,
550 	.l2_tunnel_offload_set   = ixgbe_dev_l2_tunnel_offload_set,
551 	.udp_tunnel_port_add  = ixgbe_dev_udp_tunnel_port_add,
552 	.udp_tunnel_port_del  = ixgbe_dev_udp_tunnel_port_del,
553 };
554 
555 /*
556  * dev_ops for virtual function, bare necessities for basic vf
557  * operation have been implemented
558  */
559 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
560 	.dev_configure        = ixgbevf_dev_configure,
561 	.dev_start            = ixgbevf_dev_start,
562 	.dev_stop             = ixgbevf_dev_stop,
563 	.link_update          = ixgbe_dev_link_update,
564 	.stats_get            = ixgbevf_dev_stats_get,
565 	.xstats_get           = ixgbevf_dev_xstats_get,
566 	.stats_reset          = ixgbevf_dev_stats_reset,
567 	.xstats_reset         = ixgbevf_dev_stats_reset,
568 	.xstats_get_names     = ixgbevf_dev_xstats_get_names,
569 	.dev_close            = ixgbevf_dev_close,
570 	.allmulticast_enable  = ixgbevf_dev_allmulticast_enable,
571 	.allmulticast_disable = ixgbevf_dev_allmulticast_disable,
572 	.dev_infos_get        = ixgbevf_dev_info_get,
573 	.dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
574 	.mtu_set              = ixgbevf_dev_set_mtu,
575 	.vlan_filter_set      = ixgbevf_vlan_filter_set,
576 	.vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
577 	.vlan_offload_set     = ixgbevf_vlan_offload_set,
578 	.rx_queue_setup       = ixgbe_dev_rx_queue_setup,
579 	.rx_queue_release     = ixgbe_dev_rx_queue_release,
580 	.rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
581 	.tx_queue_setup       = ixgbe_dev_tx_queue_setup,
582 	.tx_queue_release     = ixgbe_dev_tx_queue_release,
583 	.rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
584 	.rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
585 	.mac_addr_add         = ixgbevf_add_mac_addr,
586 	.mac_addr_remove      = ixgbevf_remove_mac_addr,
587 	.set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
588 	.rxq_info_get         = ixgbe_rxq_info_get,
589 	.txq_info_get         = ixgbe_txq_info_get,
590 	.mac_addr_set         = ixgbevf_set_default_mac_addr,
591 	.get_reg              = ixgbevf_get_regs,
592 	.reta_update          = ixgbe_dev_rss_reta_update,
593 	.reta_query           = ixgbe_dev_rss_reta_query,
594 	.rss_hash_update      = ixgbe_dev_rss_hash_update,
595 	.rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
596 };
597 
598 /* store statistics names and its offset in stats structure */
599 struct rte_ixgbe_xstats_name_off {
600 	char name[RTE_ETH_XSTATS_NAME_SIZE];
601 	unsigned offset;
602 };
603 
604 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
605 	{"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
606 	{"rx_illegal_byte_errors", offsetof(struct ixgbe_hw_stats, illerrc)},
607 	{"rx_error_bytes", offsetof(struct ixgbe_hw_stats, errbc)},
608 	{"mac_local_errors", offsetof(struct ixgbe_hw_stats, mlfc)},
609 	{"mac_remote_errors", offsetof(struct ixgbe_hw_stats, mrfc)},
610 	{"rx_length_errors", offsetof(struct ixgbe_hw_stats, rlec)},
611 	{"tx_xon_packets", offsetof(struct ixgbe_hw_stats, lxontxc)},
612 	{"rx_xon_packets", offsetof(struct ixgbe_hw_stats, lxonrxc)},
613 	{"tx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxofftxc)},
614 	{"rx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
615 	{"rx_size_64_packets", offsetof(struct ixgbe_hw_stats, prc64)},
616 	{"rx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, prc127)},
617 	{"rx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, prc255)},
618 	{"rx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, prc511)},
619 	{"rx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
620 		prc1023)},
621 	{"rx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
622 		prc1522)},
623 	{"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
624 	{"rx_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
625 	{"rx_fragment_errors", offsetof(struct ixgbe_hw_stats, rfc)},
626 	{"rx_undersize_errors", offsetof(struct ixgbe_hw_stats, ruc)},
627 	{"rx_oversize_errors", offsetof(struct ixgbe_hw_stats, roc)},
628 	{"rx_jabber_errors", offsetof(struct ixgbe_hw_stats, rjc)},
629 	{"rx_management_packets", offsetof(struct ixgbe_hw_stats, mngprc)},
630 	{"rx_management_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
631 	{"tx_management_packets", offsetof(struct ixgbe_hw_stats, mngptc)},
632 	{"rx_total_packets", offsetof(struct ixgbe_hw_stats, tpr)},
633 	{"rx_total_bytes", offsetof(struct ixgbe_hw_stats, tor)},
634 	{"tx_total_packets", offsetof(struct ixgbe_hw_stats, tpt)},
635 	{"tx_size_64_packets", offsetof(struct ixgbe_hw_stats, ptc64)},
636 	{"tx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, ptc127)},
637 	{"tx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, ptc255)},
638 	{"tx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, ptc511)},
639 	{"tx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
640 		ptc1023)},
641 	{"tx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
642 		ptc1522)},
643 	{"tx_multicast_packets", offsetof(struct ixgbe_hw_stats, mptc)},
644 	{"tx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bptc)},
645 	{"rx_mac_short_packet_dropped", offsetof(struct ixgbe_hw_stats, mspdc)},
646 	{"rx_l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
647 
648 	{"flow_director_added_filters", offsetof(struct ixgbe_hw_stats,
649 		fdirustat_add)},
650 	{"flow_director_removed_filters", offsetof(struct ixgbe_hw_stats,
651 		fdirustat_remove)},
652 	{"flow_director_filter_add_errors", offsetof(struct ixgbe_hw_stats,
653 		fdirfstat_fadd)},
654 	{"flow_director_filter_remove_errors", offsetof(struct ixgbe_hw_stats,
655 		fdirfstat_fremove)},
656 	{"flow_director_matched_filters", offsetof(struct ixgbe_hw_stats,
657 		fdirmatch)},
658 	{"flow_director_missed_filters", offsetof(struct ixgbe_hw_stats,
659 		fdirmiss)},
660 
661 	{"rx_fcoe_crc_errors", offsetof(struct ixgbe_hw_stats, fccrc)},
662 	{"rx_fcoe_dropped", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
663 	{"rx_fcoe_mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats,
664 		fclast)},
665 	{"rx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeprc)},
666 	{"tx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeptc)},
667 	{"rx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwrc)},
668 	{"tx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwtc)},
669 	{"rx_fcoe_no_direct_data_placement", offsetof(struct ixgbe_hw_stats,
670 		fcoe_noddp)},
671 	{"rx_fcoe_no_direct_data_placement_ext_buff",
672 		offsetof(struct ixgbe_hw_stats, fcoe_noddp_ext_buff)},
673 
674 	{"tx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
675 		lxontxc)},
676 	{"rx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
677 		lxonrxc)},
678 	{"tx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
679 		lxofftxc)},
680 	{"rx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
681 		lxoffrxc)},
682 	{"rx_total_missed_packets", offsetof(struct ixgbe_hw_stats, mpctotal)},
683 };
684 
685 #define IXGBE_NB_HW_STATS (sizeof(rte_ixgbe_stats_strings) / \
686 			   sizeof(rte_ixgbe_stats_strings[0]))
687 
688 /* Per-queue statistics */
689 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_rxq_strings[] = {
690 	{"mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats, rnbc)},
691 	{"dropped", offsetof(struct ixgbe_hw_stats, mpc)},
692 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxonrxc)},
693 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxoffrxc)},
694 };
695 
696 #define IXGBE_NB_RXQ_PRIO_STATS (sizeof(rte_ixgbe_rxq_strings) / \
697 			   sizeof(rte_ixgbe_rxq_strings[0]))
698 #define IXGBE_NB_RXQ_PRIO_VALUES 8
699 
700 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_txq_strings[] = {
701 	{"xon_packets", offsetof(struct ixgbe_hw_stats, pxontxc)},
702 	{"xoff_packets", offsetof(struct ixgbe_hw_stats, pxofftxc)},
703 	{"xon_to_xoff_packets", offsetof(struct ixgbe_hw_stats,
704 		pxon2offc)},
705 };
706 
707 #define IXGBE_NB_TXQ_PRIO_STATS (sizeof(rte_ixgbe_txq_strings) / \
708 			   sizeof(rte_ixgbe_txq_strings[0]))
709 #define IXGBE_NB_TXQ_PRIO_VALUES 8
710 
711 static const struct rte_ixgbe_xstats_name_off rte_ixgbevf_stats_strings[] = {
712 	{"rx_multicast_packets", offsetof(struct ixgbevf_hw_stats, vfmprc)},
713 };
714 
715 #define IXGBEVF_NB_XSTATS (sizeof(rte_ixgbevf_stats_strings) /	\
716 		sizeof(rte_ixgbevf_stats_strings[0]))
717 
718 /**
719  * Atomically reads the link status information from global
720  * structure rte_eth_dev.
721  *
722  * @param dev
723  *   - Pointer to the structure rte_eth_dev to read from.
724  *   - Pointer to the buffer to be saved with the link status.
725  *
726  * @return
727  *   - On success, zero.
728  *   - On failure, negative value.
729  */
730 static inline int
731 rte_ixgbe_dev_atomic_read_link_status(struct rte_eth_dev *dev,
732 				struct rte_eth_link *link)
733 {
734 	struct rte_eth_link *dst = link;
735 	struct rte_eth_link *src = &(dev->data->dev_link);
736 
737 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
738 					*(uint64_t *)src) == 0)
739 		return -1;
740 
741 	return 0;
742 }
743 
744 /**
745  * Atomically writes the link status information into global
746  * structure rte_eth_dev.
747  *
748  * @param dev
749  *   - Pointer to the structure rte_eth_dev to read from.
750  *   - Pointer to the buffer to be saved with the link status.
751  *
752  * @return
753  *   - On success, zero.
754  *   - On failure, negative value.
755  */
756 static inline int
757 rte_ixgbe_dev_atomic_write_link_status(struct rte_eth_dev *dev,
758 				struct rte_eth_link *link)
759 {
760 	struct rte_eth_link *dst = &(dev->data->dev_link);
761 	struct rte_eth_link *src = link;
762 
763 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
764 					*(uint64_t *)src) == 0)
765 		return -1;
766 
767 	return 0;
768 }
769 
770 /*
771  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
772  */
773 static inline int
774 ixgbe_is_sfp(struct ixgbe_hw *hw)
775 {
776 	switch (hw->phy.type) {
777 	case ixgbe_phy_sfp_avago:
778 	case ixgbe_phy_sfp_ftl:
779 	case ixgbe_phy_sfp_intel:
780 	case ixgbe_phy_sfp_unknown:
781 	case ixgbe_phy_sfp_passive_tyco:
782 	case ixgbe_phy_sfp_passive_unknown:
783 		return 1;
784 	default:
785 		return 0;
786 	}
787 }
788 
789 static inline int32_t
790 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
791 {
792 	uint32_t ctrl_ext;
793 	int32_t status;
794 
795 	status = ixgbe_reset_hw(hw);
796 
797 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
798 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
799 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
800 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
801 	IXGBE_WRITE_FLUSH(hw);
802 
803 	return status;
804 }
805 
806 static inline void
807 ixgbe_enable_intr(struct rte_eth_dev *dev)
808 {
809 	struct ixgbe_interrupt *intr =
810 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
811 	struct ixgbe_hw *hw =
812 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
813 
814 	IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
815 	IXGBE_WRITE_FLUSH(hw);
816 }
817 
818 /*
819  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
820  */
821 static void
822 ixgbe_disable_intr(struct ixgbe_hw *hw)
823 {
824 	PMD_INIT_FUNC_TRACE();
825 
826 	if (hw->mac.type == ixgbe_mac_82598EB) {
827 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
828 	} else {
829 		IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
830 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
831 		IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
832 	}
833 	IXGBE_WRITE_FLUSH(hw);
834 }
835 
836 /*
837  * This function resets queue statistics mapping registers.
838  * From Niantic datasheet, Initialization of Statistics section:
839  * "...if software requires the queue counters, the RQSMR and TQSM registers
840  * must be re-programmed following a device reset.
841  */
842 static void
843 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
844 {
845 	uint32_t i;
846 
847 	for (i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
848 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
849 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
850 	}
851 }
852 
853 
854 static int
855 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
856 				  uint16_t queue_id,
857 				  uint8_t stat_idx,
858 				  uint8_t is_rx)
859 {
860 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
861 #define NB_QMAP_FIELDS_PER_QSM_REG 4
862 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
863 
864 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
865 	struct ixgbe_stat_mapping_registers *stat_mappings =
866 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
867 	uint32_t qsmr_mask = 0;
868 	uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
869 	uint32_t q_map;
870 	uint8_t n, offset;
871 
872 	if ((hw->mac.type != ixgbe_mac_82599EB) &&
873 		(hw->mac.type != ixgbe_mac_X540) &&
874 		(hw->mac.type != ixgbe_mac_X550) &&
875 		(hw->mac.type != ixgbe_mac_X550EM_x) &&
876 		(hw->mac.type != ixgbe_mac_X550EM_a))
877 		return -ENOSYS;
878 
879 	PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
880 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
881 		     queue_id, stat_idx);
882 
883 	n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
884 	if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
885 		PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
886 		return -EIO;
887 	}
888 	offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
889 
890 	/* Now clear any previous stat_idx set */
891 	clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
892 	if (!is_rx)
893 		stat_mappings->tqsm[n] &= ~clearing_mask;
894 	else
895 		stat_mappings->rqsmr[n] &= ~clearing_mask;
896 
897 	q_map = (uint32_t)stat_idx;
898 	q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
899 	qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
900 	if (!is_rx)
901 		stat_mappings->tqsm[n] |= qsmr_mask;
902 	else
903 		stat_mappings->rqsmr[n] |= qsmr_mask;
904 
905 	PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
906 		     (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
907 		     queue_id, stat_idx);
908 	PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
909 		     is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
910 
911 	/* Now write the mapping in the appropriate register */
912 	if (is_rx) {
913 		PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
914 			     stat_mappings->rqsmr[n], n);
915 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
916 	} else {
917 		PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
918 			     stat_mappings->tqsm[n], n);
919 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
920 	}
921 	return 0;
922 }
923 
924 static void
925 ixgbe_restore_statistics_mapping(struct rte_eth_dev *dev)
926 {
927 	struct ixgbe_stat_mapping_registers *stat_mappings =
928 		IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
929 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
930 	int i;
931 
932 	/* write whatever was in stat mapping table to the NIC */
933 	for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
934 		/* rx */
935 		IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
936 
937 		/* tx */
938 		IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
939 	}
940 }
941 
942 static void
943 ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config)
944 {
945 	uint8_t i;
946 	struct ixgbe_dcb_tc_config *tc;
947 	uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
948 
949 	dcb_config->num_tcs.pg_tcs = dcb_max_tc;
950 	dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
951 	for (i = 0; i < dcb_max_tc; i++) {
952 		tc = &dcb_config->tc_config[i];
953 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
954 		tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
955 				 (uint8_t)(100/dcb_max_tc + (i & 1));
956 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
957 		tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
958 				 (uint8_t)(100/dcb_max_tc + (i & 1));
959 		tc->pfc = ixgbe_dcb_pfc_disabled;
960 	}
961 
962 	/* Initialize default user to priority mapping, UPx->TC0 */
963 	tc = &dcb_config->tc_config[0];
964 	tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
965 	tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
966 	for (i = 0; i < IXGBE_DCB_MAX_BW_GROUP; i++) {
967 		dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
968 		dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
969 	}
970 	dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
971 	dcb_config->pfc_mode_enable = false;
972 	dcb_config->vt_mode = true;
973 	dcb_config->round_robin_enable = false;
974 	/* support all DCB capabilities in 82599 */
975 	dcb_config->support.capabilities = 0xFF;
976 
977 	/*we only support 4 Tcs for X540, X550 */
978 	if (hw->mac.type == ixgbe_mac_X540 ||
979 		hw->mac.type == ixgbe_mac_X550 ||
980 		hw->mac.type == ixgbe_mac_X550EM_x ||
981 		hw->mac.type == ixgbe_mac_X550EM_a) {
982 		dcb_config->num_tcs.pg_tcs = 4;
983 		dcb_config->num_tcs.pfc_tcs = 4;
984 	}
985 }
986 
987 /*
988  * Ensure that all locks are released before first NVM or PHY access
989  */
990 static void
991 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
992 {
993 	uint16_t mask;
994 
995 	/*
996 	 * Phy lock should not fail in this early stage. If this is the case,
997 	 * it is due to an improper exit of the application.
998 	 * So force the release of the faulty lock. Release of common lock
999 	 * is done automatically by swfw_sync function.
1000 	 */
1001 	mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
1002 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1003 		PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
1004 	}
1005 	ixgbe_release_swfw_semaphore(hw, mask);
1006 
1007 	/*
1008 	 * These ones are more tricky since they are common to all ports; but
1009 	 * swfw_sync retries last long enough (1s) to be almost sure that if
1010 	 * lock can not be taken it is due to an improper lock of the
1011 	 * semaphore.
1012 	 */
1013 	mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
1014 	if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1015 		PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1016 	}
1017 	ixgbe_release_swfw_semaphore(hw, mask);
1018 }
1019 
1020 /*
1021  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
1022  * It returns 0 on success.
1023  */
1024 static int
1025 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev)
1026 {
1027 	struct rte_pci_device *pci_dev;
1028 	struct ixgbe_hw *hw =
1029 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1030 	struct ixgbe_vfta *shadow_vfta =
1031 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1032 	struct ixgbe_hwstrip *hwstrip =
1033 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1034 	struct ixgbe_dcb_config *dcb_config =
1035 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
1036 	struct ixgbe_filter_info *filter_info =
1037 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1038 	uint32_t ctrl_ext;
1039 	uint16_t csum;
1040 	int diag, i;
1041 
1042 	PMD_INIT_FUNC_TRACE();
1043 
1044 	eth_dev->dev_ops = &ixgbe_eth_dev_ops;
1045 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1046 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1047 
1048 	/*
1049 	 * For secondary processes, we don't initialise any further as primary
1050 	 * has already done this work. Only check we don't need a different
1051 	 * RX and TX function.
1052 	 */
1053 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1054 		struct ixgbe_tx_queue *txq;
1055 		/* TX queue function in primary, set by last queue initialized
1056 		 * Tx queue may not initialized by primary process
1057 		 */
1058 		if (eth_dev->data->tx_queues) {
1059 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
1060 			ixgbe_set_tx_function(eth_dev, txq);
1061 		} else {
1062 			/* Use default TX function if we get here */
1063 			PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
1064 				     "Using default TX function.");
1065 		}
1066 
1067 		ixgbe_set_rx_function(eth_dev);
1068 
1069 		return 0;
1070 	}
1071 	pci_dev = eth_dev->pci_dev;
1072 
1073 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1074 
1075 	/* Vendor and Device ID need to be set before init of shared code */
1076 	hw->device_id = pci_dev->id.device_id;
1077 	hw->vendor_id = pci_dev->id.vendor_id;
1078 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1079 	hw->allow_unsupported_sfp = 1;
1080 
1081 	/* Initialize the shared code (base driver) */
1082 #ifdef RTE_NIC_BYPASS
1083 	diag = ixgbe_bypass_init_shared_code(hw);
1084 #else
1085 	diag = ixgbe_init_shared_code(hw);
1086 #endif /* RTE_NIC_BYPASS */
1087 
1088 	if (diag != IXGBE_SUCCESS) {
1089 		PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
1090 		return -EIO;
1091 	}
1092 
1093 	/* pick up the PCI bus settings for reporting later */
1094 	ixgbe_get_bus_info(hw);
1095 
1096 	/* Unlock any pending hardware semaphore */
1097 	ixgbe_swfw_lock_reset(hw);
1098 
1099 	/* Initialize DCB configuration*/
1100 	memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
1101 	ixgbe_dcb_init(hw, dcb_config);
1102 	/* Get Hardware Flow Control setting */
1103 	hw->fc.requested_mode = ixgbe_fc_full;
1104 	hw->fc.current_mode = ixgbe_fc_full;
1105 	hw->fc.pause_time = IXGBE_FC_PAUSE;
1106 	for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
1107 		hw->fc.low_water[i] = IXGBE_FC_LO;
1108 		hw->fc.high_water[i] = IXGBE_FC_HI;
1109 	}
1110 	hw->fc.send_xon = 1;
1111 
1112 	/* Make sure we have a good EEPROM before we read from it */
1113 	diag = ixgbe_validate_eeprom_checksum(hw, &csum);
1114 	if (diag != IXGBE_SUCCESS) {
1115 		PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
1116 		return -EIO;
1117 	}
1118 
1119 #ifdef RTE_NIC_BYPASS
1120 	diag = ixgbe_bypass_init_hw(hw);
1121 #else
1122 	diag = ixgbe_init_hw(hw);
1123 #endif /* RTE_NIC_BYPASS */
1124 
1125 	/*
1126 	 * Devices with copper phys will fail to initialise if ixgbe_init_hw()
1127 	 * is called too soon after the kernel driver unbinding/binding occurs.
1128 	 * The failure occurs in ixgbe_identify_phy_generic() for all devices,
1129 	 * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
1130 	 * also called. See ixgbe_identify_phy_82599(). The reason for the
1131 	 * failure is not known, and only occuts when virtualisation features
1132 	 * are disabled in the bios. A delay of 100ms  was found to be enough by
1133 	 * trial-and-error, and is doubled to be safe.
1134 	 */
1135 	if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
1136 		rte_delay_ms(200);
1137 		diag = ixgbe_init_hw(hw);
1138 	}
1139 
1140 	if (diag == IXGBE_ERR_EEPROM_VERSION) {
1141 		PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
1142 			     "LOM.  Please be aware there may be issues associated "
1143 			     "with your hardware.");
1144 		PMD_INIT_LOG(ERR, "If you are experiencing problems "
1145 			     "please contact your Intel or hardware representative "
1146 			     "who provided you with this hardware.");
1147 	} else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
1148 		PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
1149 	if (diag) {
1150 		PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
1151 		return -EIO;
1152 	}
1153 
1154 	/* Reset the hw statistics */
1155 	ixgbe_dev_stats_reset(eth_dev);
1156 
1157 	/* disable interrupt */
1158 	ixgbe_disable_intr(hw);
1159 
1160 	/* reset mappings for queue statistics hw counters*/
1161 	ixgbe_reset_qstat_mappings(hw);
1162 
1163 	/* Allocate memory for storing MAC addresses */
1164 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
1165 					       hw->mac.num_rar_entries, 0);
1166 	if (eth_dev->data->mac_addrs == NULL) {
1167 		PMD_INIT_LOG(ERR,
1168 			     "Failed to allocate %u bytes needed to store "
1169 			     "MAC addresses",
1170 			     ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1171 		return -ENOMEM;
1172 	}
1173 	/* Copy the permanent MAC address */
1174 	ether_addr_copy((struct ether_addr *) hw->mac.perm_addr,
1175 			&eth_dev->data->mac_addrs[0]);
1176 
1177 	/* Allocate memory for storing hash filter MAC addresses */
1178 	eth_dev->data->hash_mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
1179 						    IXGBE_VMDQ_NUM_UC_MAC, 0);
1180 	if (eth_dev->data->hash_mac_addrs == NULL) {
1181 		PMD_INIT_LOG(ERR,
1182 			     "Failed to allocate %d bytes needed to store MAC addresses",
1183 			     ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
1184 		return -ENOMEM;
1185 	}
1186 
1187 	/* initialize the vfta */
1188 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1189 
1190 	/* initialize the hw strip bitmap*/
1191 	memset(hwstrip, 0, sizeof(*hwstrip));
1192 
1193 	/* initialize PF if max_vfs not zero */
1194 	ixgbe_pf_host_init(eth_dev);
1195 
1196 	ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1197 	/* let hardware know driver is loaded */
1198 	ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1199 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1200 	ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1201 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1202 	IXGBE_WRITE_FLUSH(hw);
1203 
1204 	if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1205 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1206 			     (int) hw->mac.type, (int) hw->phy.type,
1207 			     (int) hw->phy.sfp_type);
1208 	else
1209 		PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1210 			     (int) hw->mac.type, (int) hw->phy.type);
1211 
1212 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1213 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1214 		     pci_dev->id.device_id);
1215 
1216 	rte_intr_callback_register(&pci_dev->intr_handle,
1217 				   ixgbe_dev_interrupt_handler,
1218 				   (void *)eth_dev);
1219 
1220 	/* enable uio/vfio intr/eventfd mapping */
1221 	rte_intr_enable(&pci_dev->intr_handle);
1222 
1223 	/* enable support intr */
1224 	ixgbe_enable_intr(eth_dev);
1225 
1226 	/* initialize 5tuple filter list */
1227 	TAILQ_INIT(&filter_info->fivetuple_list);
1228 	memset(filter_info->fivetuple_mask, 0,
1229 	       sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1230 
1231 	return 0;
1232 }
1233 
1234 static int
1235 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1236 {
1237 	struct rte_pci_device *pci_dev;
1238 	struct ixgbe_hw *hw;
1239 
1240 	PMD_INIT_FUNC_TRACE();
1241 
1242 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1243 		return -EPERM;
1244 
1245 	hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1246 	pci_dev = eth_dev->pci_dev;
1247 
1248 	if (hw->adapter_stopped == 0)
1249 		ixgbe_dev_close(eth_dev);
1250 
1251 	eth_dev->dev_ops = NULL;
1252 	eth_dev->rx_pkt_burst = NULL;
1253 	eth_dev->tx_pkt_burst = NULL;
1254 
1255 	/* Unlock any pending hardware semaphore */
1256 	ixgbe_swfw_lock_reset(hw);
1257 
1258 	/* disable uio intr before callback unregister */
1259 	rte_intr_disable(&(pci_dev->intr_handle));
1260 	rte_intr_callback_unregister(&(pci_dev->intr_handle),
1261 		ixgbe_dev_interrupt_handler, (void *)eth_dev);
1262 
1263 	/* uninitialize PF if max_vfs not zero */
1264 	ixgbe_pf_host_uninit(eth_dev);
1265 
1266 	rte_free(eth_dev->data->mac_addrs);
1267 	eth_dev->data->mac_addrs = NULL;
1268 
1269 	rte_free(eth_dev->data->hash_mac_addrs);
1270 	eth_dev->data->hash_mac_addrs = NULL;
1271 
1272 	return 0;
1273 }
1274 
1275 /*
1276  * Negotiate mailbox API version with the PF.
1277  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1278  * Then we try to negotiate starting with the most recent one.
1279  * If all negotiation attempts fail, then we will proceed with
1280  * the default one (ixgbe_mbox_api_10).
1281  */
1282 static void
1283 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1284 {
1285 	int32_t i;
1286 
1287 	/* start with highest supported, proceed down */
1288 	static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1289 		ixgbe_mbox_api_12,
1290 		ixgbe_mbox_api_11,
1291 		ixgbe_mbox_api_10,
1292 	};
1293 
1294 	for (i = 0;
1295 			i != RTE_DIM(sup_ver) &&
1296 			ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1297 			i++)
1298 		;
1299 }
1300 
1301 static void
1302 generate_random_mac_addr(struct ether_addr *mac_addr)
1303 {
1304 	uint64_t random;
1305 
1306 	/* Set Organizationally Unique Identifier (OUI) prefix. */
1307 	mac_addr->addr_bytes[0] = 0x00;
1308 	mac_addr->addr_bytes[1] = 0x09;
1309 	mac_addr->addr_bytes[2] = 0xC0;
1310 	/* Force indication of locally assigned MAC address. */
1311 	mac_addr->addr_bytes[0] |= ETHER_LOCAL_ADMIN_ADDR;
1312 	/* Generate the last 3 bytes of the MAC address with a random number. */
1313 	random = rte_rand();
1314 	memcpy(&mac_addr->addr_bytes[3], &random, 3);
1315 }
1316 
1317 /*
1318  * Virtual Function device init
1319  */
1320 static int
1321 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1322 {
1323 	int diag;
1324 	uint32_t tc, tcs;
1325 	struct rte_pci_device *pci_dev;
1326 	struct ixgbe_hw *hw =
1327 		IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1328 	struct ixgbe_vfta *shadow_vfta =
1329 		IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1330 	struct ixgbe_hwstrip *hwstrip =
1331 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1332 	struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
1333 
1334 	PMD_INIT_FUNC_TRACE();
1335 
1336 	eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1337 	eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1338 	eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1339 
1340 	/* for secondary processes, we don't initialise any further as primary
1341 	 * has already done this work. Only check we don't need a different
1342 	 * RX function
1343 	 */
1344 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1345 		struct ixgbe_tx_queue *txq;
1346 		/* TX queue function in primary, set by last queue initialized
1347 		 * Tx queue may not initialized by primary process
1348 		 */
1349 		if (eth_dev->data->tx_queues) {
1350 			txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1351 			ixgbe_set_tx_function(eth_dev, txq);
1352 		} else {
1353 			/* Use default TX function if we get here */
1354 			PMD_INIT_LOG(NOTICE,
1355 				     "No TX queues configured yet. Using default TX function.");
1356 		}
1357 
1358 		ixgbe_set_rx_function(eth_dev);
1359 
1360 		return 0;
1361 	}
1362 
1363 	pci_dev = eth_dev->pci_dev;
1364 
1365 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1366 
1367 	hw->device_id = pci_dev->id.device_id;
1368 	hw->vendor_id = pci_dev->id.vendor_id;
1369 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1370 
1371 	/* initialize the vfta */
1372 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1373 
1374 	/* initialize the hw strip bitmap*/
1375 	memset(hwstrip, 0, sizeof(*hwstrip));
1376 
1377 	/* Initialize the shared code (base driver) */
1378 	diag = ixgbe_init_shared_code(hw);
1379 	if (diag != IXGBE_SUCCESS) {
1380 		PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1381 		return -EIO;
1382 	}
1383 
1384 	/* init_mailbox_params */
1385 	hw->mbx.ops.init_params(hw);
1386 
1387 	/* Reset the hw statistics */
1388 	ixgbevf_dev_stats_reset(eth_dev);
1389 
1390 	/* Disable the interrupts for VF */
1391 	ixgbevf_intr_disable(hw);
1392 
1393 	hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1394 	diag = hw->mac.ops.reset_hw(hw);
1395 
1396 	/*
1397 	 * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1398 	 * the underlying PF driver has not assigned a MAC address to the VF.
1399 	 * In this case, assign a random MAC address.
1400 	 */
1401 	if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1402 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1403 		return diag;
1404 	}
1405 
1406 	/* negotiate mailbox API version to use with the PF. */
1407 	ixgbevf_negotiate_api(hw);
1408 
1409 	/* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1410 	ixgbevf_get_queues(hw, &tcs, &tc);
1411 
1412 	/* Allocate memory for storing MAC addresses */
1413 	eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", ETHER_ADDR_LEN *
1414 					       hw->mac.num_rar_entries, 0);
1415 	if (eth_dev->data->mac_addrs == NULL) {
1416 		PMD_INIT_LOG(ERR,
1417 			     "Failed to allocate %u bytes needed to store "
1418 			     "MAC addresses",
1419 			     ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1420 		return -ENOMEM;
1421 	}
1422 
1423 	/* Generate a random MAC address, if none was assigned by PF. */
1424 	if (is_zero_ether_addr(perm_addr)) {
1425 		generate_random_mac_addr(perm_addr);
1426 		diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1427 		if (diag) {
1428 			rte_free(eth_dev->data->mac_addrs);
1429 			eth_dev->data->mac_addrs = NULL;
1430 			return diag;
1431 		}
1432 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1433 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1434 			     "%02x:%02x:%02x:%02x:%02x:%02x",
1435 			     perm_addr->addr_bytes[0],
1436 			     perm_addr->addr_bytes[1],
1437 			     perm_addr->addr_bytes[2],
1438 			     perm_addr->addr_bytes[3],
1439 			     perm_addr->addr_bytes[4],
1440 			     perm_addr->addr_bytes[5]);
1441 	}
1442 
1443 	/* Copy the permanent MAC address */
1444 	ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1445 
1446 	/* reset the hardware with the new settings */
1447 	diag = hw->mac.ops.start_hw(hw);
1448 	switch (diag) {
1449 	case  0:
1450 		break;
1451 
1452 	default:
1453 		PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1454 		return -EIO;
1455 	}
1456 
1457 	rte_intr_callback_register(&pci_dev->intr_handle,
1458 				   ixgbevf_dev_interrupt_handler,
1459 				   (void *)eth_dev);
1460 	rte_intr_enable(&pci_dev->intr_handle);
1461 	ixgbevf_intr_enable(hw);
1462 
1463 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1464 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1465 		     pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1466 
1467 	return 0;
1468 }
1469 
1470 /* Virtual Function device uninit */
1471 
1472 static int
1473 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1474 {
1475 	struct ixgbe_hw *hw;
1476 	struct rte_pci_device *pci_dev = eth_dev->pci_dev;
1477 
1478 	PMD_INIT_FUNC_TRACE();
1479 
1480 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1481 		return -EPERM;
1482 
1483 	hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1484 
1485 	if (hw->adapter_stopped == 0)
1486 		ixgbevf_dev_close(eth_dev);
1487 
1488 	eth_dev->dev_ops = NULL;
1489 	eth_dev->rx_pkt_burst = NULL;
1490 	eth_dev->tx_pkt_burst = NULL;
1491 
1492 	/* Disable the interrupts for VF */
1493 	ixgbevf_intr_disable(hw);
1494 
1495 	rte_free(eth_dev->data->mac_addrs);
1496 	eth_dev->data->mac_addrs = NULL;
1497 
1498 	rte_intr_disable(&pci_dev->intr_handle);
1499 	rte_intr_callback_unregister(&pci_dev->intr_handle,
1500 				     ixgbevf_dev_interrupt_handler,
1501 				     (void *)eth_dev);
1502 
1503 	return 0;
1504 }
1505 
1506 static struct eth_driver rte_ixgbe_pmd = {
1507 	.pci_drv = {
1508 		.name = "rte_ixgbe_pmd",
1509 		.id_table = pci_id_ixgbe_map,
1510 		.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
1511 			RTE_PCI_DRV_DETACHABLE,
1512 	},
1513 	.eth_dev_init = eth_ixgbe_dev_init,
1514 	.eth_dev_uninit = eth_ixgbe_dev_uninit,
1515 	.dev_private_size = sizeof(struct ixgbe_adapter),
1516 };
1517 
1518 /*
1519  * virtual function driver struct
1520  */
1521 static struct eth_driver rte_ixgbevf_pmd = {
1522 	.pci_drv = {
1523 		.name = "rte_ixgbevf_pmd",
1524 		.id_table = pci_id_ixgbevf_map,
1525 		.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_DETACHABLE,
1526 	},
1527 	.eth_dev_init = eth_ixgbevf_dev_init,
1528 	.eth_dev_uninit = eth_ixgbevf_dev_uninit,
1529 	.dev_private_size = sizeof(struct ixgbe_adapter),
1530 };
1531 
1532 /*
1533  * Driver initialization routine.
1534  * Invoked once at EAL init time.
1535  * Register itself as the [Poll Mode] Driver of PCI IXGBE devices.
1536  */
1537 static int
1538 rte_ixgbe_pmd_init(const char *name __rte_unused, const char *params __rte_unused)
1539 {
1540 	PMD_INIT_FUNC_TRACE();
1541 
1542 	rte_eth_driver_register(&rte_ixgbe_pmd);
1543 	return 0;
1544 }
1545 
1546 /*
1547  * VF Driver initialization routine.
1548  * Invoked one at EAL init time.
1549  * Register itself as the [Virtual Poll Mode] Driver of PCI niantic devices.
1550  */
1551 static int
1552 rte_ixgbevf_pmd_init(const char *name __rte_unused, const char *param __rte_unused)
1553 {
1554 	PMD_INIT_FUNC_TRACE();
1555 
1556 	rte_eth_driver_register(&rte_ixgbevf_pmd);
1557 	return 0;
1558 }
1559 
1560 static int
1561 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1562 {
1563 	struct ixgbe_hw *hw =
1564 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1565 	struct ixgbe_vfta *shadow_vfta =
1566 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1567 	uint32_t vfta;
1568 	uint32_t vid_idx;
1569 	uint32_t vid_bit;
1570 
1571 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1572 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1573 	vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1574 	if (on)
1575 		vfta |= vid_bit;
1576 	else
1577 		vfta &= ~vid_bit;
1578 	IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1579 
1580 	/* update local VFTA copy */
1581 	shadow_vfta->vfta[vid_idx] = vfta;
1582 
1583 	return 0;
1584 }
1585 
1586 static void
1587 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1588 {
1589 	if (on)
1590 		ixgbe_vlan_hw_strip_enable(dev, queue);
1591 	else
1592 		ixgbe_vlan_hw_strip_disable(dev, queue);
1593 }
1594 
1595 static int
1596 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1597 		    enum rte_vlan_type vlan_type,
1598 		    uint16_t tpid)
1599 {
1600 	struct ixgbe_hw *hw =
1601 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1602 	int ret = 0;
1603 	uint32_t reg;
1604 	uint32_t qinq;
1605 
1606 	qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1607 	qinq &= IXGBE_DMATXCTL_GDV;
1608 
1609 	switch (vlan_type) {
1610 	case ETH_VLAN_TYPE_INNER:
1611 		if (qinq) {
1612 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1613 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1614 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1615 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1616 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1617 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1618 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1619 		} else {
1620 			ret = -ENOTSUP;
1621 			PMD_DRV_LOG(ERR, "Inner type is not supported"
1622 				    " by single VLAN");
1623 		}
1624 		break;
1625 	case ETH_VLAN_TYPE_OUTER:
1626 		if (qinq) {
1627 			/* Only the high 16-bits is valid */
1628 			IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1629 					IXGBE_EXVET_VET_EXT_SHIFT);
1630 		} else {
1631 			reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1632 			reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1633 			IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1634 			reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1635 			reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1636 				| ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1637 			IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1638 		}
1639 
1640 		break;
1641 	default:
1642 		ret = -EINVAL;
1643 		PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1644 		break;
1645 	}
1646 
1647 	return ret;
1648 }
1649 
1650 void
1651 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1652 {
1653 	struct ixgbe_hw *hw =
1654 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1655 	uint32_t vlnctrl;
1656 
1657 	PMD_INIT_FUNC_TRACE();
1658 
1659 	/* Filter Table Disable */
1660 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1661 	vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1662 
1663 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1664 }
1665 
1666 void
1667 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1668 {
1669 	struct ixgbe_hw *hw =
1670 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1671 	struct ixgbe_vfta *shadow_vfta =
1672 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1673 	uint32_t vlnctrl;
1674 	uint16_t i;
1675 
1676 	PMD_INIT_FUNC_TRACE();
1677 
1678 	/* Filter Table Enable */
1679 	vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1680 	vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1681 	vlnctrl |= IXGBE_VLNCTRL_VFE;
1682 
1683 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1684 
1685 	/* write whatever is in local vfta copy */
1686 	for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1687 		IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1688 }
1689 
1690 static void
1691 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1692 {
1693 	struct ixgbe_hwstrip *hwstrip =
1694 		IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1695 	struct ixgbe_rx_queue *rxq;
1696 
1697 	if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
1698 		return;
1699 
1700 	if (on)
1701 		IXGBE_SET_HWSTRIP(hwstrip, queue);
1702 	else
1703 		IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1704 
1705 	if (queue >= dev->data->nb_rx_queues)
1706 		return;
1707 
1708 	rxq = dev->data->rx_queues[queue];
1709 
1710 	if (on)
1711 		rxq->vlan_flags = PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED;
1712 	else
1713 		rxq->vlan_flags = PKT_RX_VLAN_PKT;
1714 }
1715 
1716 static void
1717 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
1718 {
1719 	struct ixgbe_hw *hw =
1720 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1721 	uint32_t ctrl;
1722 
1723 	PMD_INIT_FUNC_TRACE();
1724 
1725 	if (hw->mac.type == ixgbe_mac_82598EB) {
1726 		/* No queue level support */
1727 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1728 		return;
1729 	}
1730 
1731 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1732 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1733 	ctrl &= ~IXGBE_RXDCTL_VME;
1734 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1735 
1736 	/* record those setting for HW strip per queue */
1737 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
1738 }
1739 
1740 static void
1741 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
1742 {
1743 	struct ixgbe_hw *hw =
1744 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1745 	uint32_t ctrl;
1746 
1747 	PMD_INIT_FUNC_TRACE();
1748 
1749 	if (hw->mac.type == ixgbe_mac_82598EB) {
1750 		/* No queue level supported */
1751 		PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1752 		return;
1753 	}
1754 
1755 	/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1756 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1757 	ctrl |= IXGBE_RXDCTL_VME;
1758 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1759 
1760 	/* record those setting for HW strip per queue */
1761 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
1762 }
1763 
1764 void
1765 ixgbe_vlan_hw_strip_disable_all(struct rte_eth_dev *dev)
1766 {
1767 	struct ixgbe_hw *hw =
1768 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1769 	uint32_t ctrl;
1770 	uint16_t i;
1771 	struct ixgbe_rx_queue *rxq;
1772 
1773 	PMD_INIT_FUNC_TRACE();
1774 
1775 	if (hw->mac.type == ixgbe_mac_82598EB) {
1776 		ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1777 		ctrl &= ~IXGBE_VLNCTRL_VME;
1778 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1779 	} else {
1780 		/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1781 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
1782 			rxq = dev->data->rx_queues[i];
1783 			ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
1784 			ctrl &= ~IXGBE_RXDCTL_VME;
1785 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
1786 
1787 			/* record those setting for HW strip per queue */
1788 			ixgbe_vlan_hw_strip_bitmap_set(dev, i, 0);
1789 		}
1790 	}
1791 }
1792 
1793 void
1794 ixgbe_vlan_hw_strip_enable_all(struct rte_eth_dev *dev)
1795 {
1796 	struct ixgbe_hw *hw =
1797 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1798 	uint32_t ctrl;
1799 	uint16_t i;
1800 	struct ixgbe_rx_queue *rxq;
1801 
1802 	PMD_INIT_FUNC_TRACE();
1803 
1804 	if (hw->mac.type == ixgbe_mac_82598EB) {
1805 		ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1806 		ctrl |= IXGBE_VLNCTRL_VME;
1807 		IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1808 	} else {
1809 		/* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1810 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
1811 			rxq = dev->data->rx_queues[i];
1812 			ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
1813 			ctrl |= IXGBE_RXDCTL_VME;
1814 			IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
1815 
1816 			/* record those setting for HW strip per queue */
1817 			ixgbe_vlan_hw_strip_bitmap_set(dev, i, 1);
1818 		}
1819 	}
1820 }
1821 
1822 static void
1823 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
1824 {
1825 	struct ixgbe_hw *hw =
1826 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1827 	uint32_t ctrl;
1828 
1829 	PMD_INIT_FUNC_TRACE();
1830 
1831 	/* DMATXCTRL: Geric Double VLAN Disable */
1832 	ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1833 	ctrl &= ~IXGBE_DMATXCTL_GDV;
1834 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
1835 
1836 	/* CTRL_EXT: Global Double VLAN Disable */
1837 	ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1838 	ctrl &= ~IXGBE_EXTENDED_VLAN;
1839 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
1840 
1841 }
1842 
1843 static void
1844 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
1845 {
1846 	struct ixgbe_hw *hw =
1847 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1848 	uint32_t ctrl;
1849 
1850 	PMD_INIT_FUNC_TRACE();
1851 
1852 	/* DMATXCTRL: Geric Double VLAN Enable */
1853 	ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1854 	ctrl |= IXGBE_DMATXCTL_GDV;
1855 	IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
1856 
1857 	/* CTRL_EXT: Global Double VLAN Enable */
1858 	ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1859 	ctrl |= IXGBE_EXTENDED_VLAN;
1860 	IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
1861 
1862 	/* Clear pooling mode of PFVTCTL. It's required by X550. */
1863 	if (hw->mac.type == ixgbe_mac_X550 ||
1864 	    hw->mac.type == ixgbe_mac_X550EM_x ||
1865 	    hw->mac.type == ixgbe_mac_X550EM_a) {
1866 		ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
1867 		ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
1868 		IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
1869 	}
1870 
1871 	/*
1872 	 * VET EXT field in the EXVET register = 0x8100 by default
1873 	 * So no need to change. Same to VT field of DMATXCTL register
1874 	 */
1875 }
1876 
1877 static void
1878 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1879 {
1880 	if (mask & ETH_VLAN_STRIP_MASK) {
1881 		if (dev->data->dev_conf.rxmode.hw_vlan_strip)
1882 			ixgbe_vlan_hw_strip_enable_all(dev);
1883 		else
1884 			ixgbe_vlan_hw_strip_disable_all(dev);
1885 	}
1886 
1887 	if (mask & ETH_VLAN_FILTER_MASK) {
1888 		if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1889 			ixgbe_vlan_hw_filter_enable(dev);
1890 		else
1891 			ixgbe_vlan_hw_filter_disable(dev);
1892 	}
1893 
1894 	if (mask & ETH_VLAN_EXTEND_MASK) {
1895 		if (dev->data->dev_conf.rxmode.hw_vlan_extend)
1896 			ixgbe_vlan_hw_extend_enable(dev);
1897 		else
1898 			ixgbe_vlan_hw_extend_disable(dev);
1899 	}
1900 }
1901 
1902 static void
1903 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1904 {
1905 	struct ixgbe_hw *hw =
1906 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1907 	/* VLNCTRL: enable vlan filtering and allow all vlan tags through */
1908 	uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1909 
1910 	vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
1911 	IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
1912 }
1913 
1914 static int
1915 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
1916 {
1917 	switch (nb_rx_q) {
1918 	case 1:
1919 	case 2:
1920 		RTE_ETH_DEV_SRIOV(dev).active = ETH_64_POOLS;
1921 		break;
1922 	case 4:
1923 		RTE_ETH_DEV_SRIOV(dev).active = ETH_32_POOLS;
1924 		break;
1925 	default:
1926 		return -EINVAL;
1927 	}
1928 
1929 	RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = nb_rx_q;
1930 	RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx = dev->pci_dev->max_vfs * nb_rx_q;
1931 
1932 	return 0;
1933 }
1934 
1935 static int
1936 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
1937 {
1938 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
1939 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1940 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
1941 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
1942 
1943 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
1944 		/* check multi-queue mode */
1945 		switch (dev_conf->rxmode.mq_mode) {
1946 		case ETH_MQ_RX_VMDQ_DCB:
1947 		case ETH_MQ_RX_VMDQ_DCB_RSS:
1948 			/* DCB/RSS VMDQ in SRIOV mode, not implement yet */
1949 			PMD_INIT_LOG(ERR, "SRIOV active,"
1950 					" unsupported mq_mode rx %d.",
1951 					dev_conf->rxmode.mq_mode);
1952 			return -EINVAL;
1953 		case ETH_MQ_RX_RSS:
1954 		case ETH_MQ_RX_VMDQ_RSS:
1955 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_RSS;
1956 			if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
1957 				if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
1958 					PMD_INIT_LOG(ERR, "SRIOV is active,"
1959 						" invalid queue number"
1960 						" for VMDQ RSS, allowed"
1961 						" value are 1, 2 or 4.");
1962 					return -EINVAL;
1963 				}
1964 			break;
1965 		case ETH_MQ_RX_VMDQ_ONLY:
1966 		case ETH_MQ_RX_NONE:
1967 			/* if nothing mq mode configure, use default scheme */
1968 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
1969 			if (RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool > 1)
1970 				RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = 1;
1971 			break;
1972 		default: /* ETH_MQ_RX_DCB, ETH_MQ_RX_DCB_RSS or ETH_MQ_TX_DCB*/
1973 			/* SRIOV only works in VMDq enable mode */
1974 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1975 					" wrong mq_mode rx %d.",
1976 					dev_conf->rxmode.mq_mode);
1977 			return -EINVAL;
1978 		}
1979 
1980 		switch (dev_conf->txmode.mq_mode) {
1981 		case ETH_MQ_TX_VMDQ_DCB:
1982 			/* DCB VMDQ in SRIOV mode, not implement yet */
1983 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1984 					" unsupported VMDQ mq_mode tx %d.",
1985 					dev_conf->txmode.mq_mode);
1986 			return -EINVAL;
1987 		default: /* ETH_MQ_TX_VMDQ_ONLY or ETH_MQ_TX_NONE */
1988 			dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_ONLY;
1989 			break;
1990 		}
1991 
1992 		/* check valid queue number */
1993 		if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
1994 		    (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
1995 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1996 					" nb_rx_q=%d nb_tx_q=%d queue number"
1997 					" must be less than or equal to %d.",
1998 					nb_rx_q, nb_tx_q,
1999 					RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2000 			return -EINVAL;
2001 		}
2002 	} else {
2003 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB_RSS) {
2004 			PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2005 					  " not supported.");
2006 			return -EINVAL;
2007 		}
2008 		/* check configuration for vmdb+dcb mode */
2009 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB) {
2010 			const struct rte_eth_vmdq_dcb_conf *conf;
2011 
2012 			if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2013 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2014 						IXGBE_VMDQ_DCB_NB_QUEUES);
2015 				return -EINVAL;
2016 			}
2017 			conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2018 			if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2019 			       conf->nb_queue_pools == ETH_32_POOLS)) {
2020 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2021 						" nb_queue_pools must be %d or %d.",
2022 						ETH_16_POOLS, ETH_32_POOLS);
2023 				return -EINVAL;
2024 			}
2025 		}
2026 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_VMDQ_DCB) {
2027 			const struct rte_eth_vmdq_dcb_tx_conf *conf;
2028 
2029 			if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2030 				PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2031 						 IXGBE_VMDQ_DCB_NB_QUEUES);
2032 				return -EINVAL;
2033 			}
2034 			conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2035 			if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2036 			       conf->nb_queue_pools == ETH_32_POOLS)) {
2037 				PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2038 						" nb_queue_pools != %d and"
2039 						" nb_queue_pools != %d.",
2040 						ETH_16_POOLS, ETH_32_POOLS);
2041 				return -EINVAL;
2042 			}
2043 		}
2044 
2045 		/* For DCB mode check our configuration before we go further */
2046 		if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB) {
2047 			const struct rte_eth_dcb_rx_conf *conf;
2048 
2049 			if (nb_rx_q != IXGBE_DCB_NB_QUEUES) {
2050 				PMD_INIT_LOG(ERR, "DCB selected, nb_rx_q != %d.",
2051 						 IXGBE_DCB_NB_QUEUES);
2052 				return -EINVAL;
2053 			}
2054 			conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2055 			if (!(conf->nb_tcs == ETH_4_TCS ||
2056 			       conf->nb_tcs == ETH_8_TCS)) {
2057 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2058 						" and nb_tcs != %d.",
2059 						ETH_4_TCS, ETH_8_TCS);
2060 				return -EINVAL;
2061 			}
2062 		}
2063 
2064 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_DCB) {
2065 			const struct rte_eth_dcb_tx_conf *conf;
2066 
2067 			if (nb_tx_q != IXGBE_DCB_NB_QUEUES) {
2068 				PMD_INIT_LOG(ERR, "DCB, nb_tx_q != %d.",
2069 						 IXGBE_DCB_NB_QUEUES);
2070 				return -EINVAL;
2071 			}
2072 			conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2073 			if (!(conf->nb_tcs == ETH_4_TCS ||
2074 			       conf->nb_tcs == ETH_8_TCS)) {
2075 				PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2076 						" and nb_tcs != %d.",
2077 						ETH_4_TCS, ETH_8_TCS);
2078 				return -EINVAL;
2079 			}
2080 		}
2081 
2082 		/*
2083 		 * When DCB/VT is off, maximum number of queues changes,
2084 		 * except for 82598EB, which remains constant.
2085 		 */
2086 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
2087 				hw->mac.type != ixgbe_mac_82598EB) {
2088 			if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2089 				PMD_INIT_LOG(ERR,
2090 					     "Neither VT nor DCB are enabled, "
2091 					     "nb_tx_q > %d.",
2092 					     IXGBE_NONE_MODE_TX_NB_QUEUES);
2093 				return -EINVAL;
2094 			}
2095 		}
2096 	}
2097 	return 0;
2098 }
2099 
2100 static int
2101 ixgbe_dev_configure(struct rte_eth_dev *dev)
2102 {
2103 	struct ixgbe_interrupt *intr =
2104 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2105 	struct ixgbe_adapter *adapter =
2106 		(struct ixgbe_adapter *)dev->data->dev_private;
2107 	int ret;
2108 
2109 	PMD_INIT_FUNC_TRACE();
2110 	/* multipe queue mode checking */
2111 	ret  = ixgbe_check_mq_mode(dev);
2112 	if (ret != 0) {
2113 		PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2114 			    ret);
2115 		return ret;
2116 	}
2117 
2118 	/* set flag to update link status after init */
2119 	intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2120 
2121 	/*
2122 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2123 	 * allocation or vector Rx preconditions we will reset it.
2124 	 */
2125 	adapter->rx_bulk_alloc_allowed = true;
2126 	adapter->rx_vec_allowed = true;
2127 
2128 	return 0;
2129 }
2130 
2131 static void
2132 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2133 {
2134 	struct ixgbe_hw *hw =
2135 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2136 	struct ixgbe_interrupt *intr =
2137 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2138 	uint32_t gpie;
2139 
2140 	/* only set up it on X550EM_X */
2141 	if (hw->mac.type == ixgbe_mac_X550EM_x) {
2142 		gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2143 		gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2144 		IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2145 		if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2146 			intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2147 	}
2148 }
2149 
2150 /*
2151  * Configure device link speed and setup link.
2152  * It returns 0 on success.
2153  */
2154 static int
2155 ixgbe_dev_start(struct rte_eth_dev *dev)
2156 {
2157 	struct ixgbe_hw *hw =
2158 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2159 	struct ixgbe_vf_info *vfinfo =
2160 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2161 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
2162 	uint32_t intr_vector = 0;
2163 	int err, link_up = 0, negotiate = 0;
2164 	uint32_t speed = 0;
2165 	int mask = 0;
2166 	int status;
2167 	uint16_t vf, idx;
2168 	uint32_t *link_speeds;
2169 
2170 	PMD_INIT_FUNC_TRACE();
2171 
2172 	/* IXGBE devices don't support:
2173 	*    - half duplex (checked afterwards for valid speeds)
2174 	*    - fixed speed: TODO implement
2175 	*/
2176 	if (dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED) {
2177 		PMD_INIT_LOG(ERR, "Invalid link_speeds for port %hhu; fix speed not supported",
2178 			     dev->data->port_id);
2179 		return -EINVAL;
2180 	}
2181 
2182 	/* disable uio/vfio intr/eventfd mapping */
2183 	rte_intr_disable(intr_handle);
2184 
2185 	/* stop adapter */
2186 	hw->adapter_stopped = 0;
2187 	ixgbe_stop_adapter(hw);
2188 
2189 	/* reinitialize adapter
2190 	 * this calls reset and start
2191 	 */
2192 	status = ixgbe_pf_reset_hw(hw);
2193 	if (status != 0)
2194 		return -1;
2195 	hw->mac.ops.start_hw(hw);
2196 	hw->mac.get_link_status = true;
2197 
2198 	/* configure PF module if SRIOV enabled */
2199 	ixgbe_pf_host_configure(dev);
2200 
2201 	ixgbe_dev_phy_intr_setup(dev);
2202 
2203 	/* check and configure queue intr-vector mapping */
2204 	if ((rte_intr_cap_multiple(intr_handle) ||
2205 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
2206 	    dev->data->dev_conf.intr_conf.rxq != 0) {
2207 		intr_vector = dev->data->nb_rx_queues;
2208 		if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2209 			PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2210 					IXGBE_MAX_INTR_QUEUE_NUM);
2211 			return -ENOTSUP;
2212 		}
2213 		if (rte_intr_efd_enable(intr_handle, intr_vector))
2214 			return -1;
2215 	}
2216 
2217 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
2218 		intr_handle->intr_vec =
2219 			rte_zmalloc("intr_vec",
2220 				    dev->data->nb_rx_queues * sizeof(int), 0);
2221 		if (intr_handle->intr_vec == NULL) {
2222 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2223 				     " intr_vec\n", dev->data->nb_rx_queues);
2224 			return -ENOMEM;
2225 		}
2226 	}
2227 
2228 	/* confiugre msix for sleep until rx interrupt */
2229 	ixgbe_configure_msix(dev);
2230 
2231 	/* initialize transmission unit */
2232 	ixgbe_dev_tx_init(dev);
2233 
2234 	/* This can fail when allocating mbufs for descriptor rings */
2235 	err = ixgbe_dev_rx_init(dev);
2236 	if (err) {
2237 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2238 		goto error;
2239 	}
2240 
2241 	err = ixgbe_dev_rxtx_start(dev);
2242 	if (err < 0) {
2243 		PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2244 		goto error;
2245 	}
2246 
2247 	/* Skip link setup if loopback mode is enabled for 82599. */
2248 	if (hw->mac.type == ixgbe_mac_82599EB &&
2249 			dev->data->dev_conf.lpbk_mode == IXGBE_LPBK_82599_TX_RX)
2250 		goto skip_link_setup;
2251 
2252 	if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2253 		err = hw->mac.ops.setup_sfp(hw);
2254 		if (err)
2255 			goto error;
2256 	}
2257 
2258 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2259 		/* Turn on the copper */
2260 		ixgbe_set_phy_power(hw, true);
2261 	} else {
2262 		/* Turn on the laser */
2263 		ixgbe_enable_tx_laser(hw);
2264 	}
2265 
2266 	err = ixgbe_check_link(hw, &speed, &link_up, 0);
2267 	if (err)
2268 		goto error;
2269 	dev->data->dev_link.link_status = link_up;
2270 
2271 	err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2272 	if (err)
2273 		goto error;
2274 
2275 	link_speeds = &dev->data->dev_conf.link_speeds;
2276 	if (*link_speeds & ~(ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2277 			ETH_LINK_SPEED_10G)) {
2278 		PMD_INIT_LOG(ERR, "Invalid link setting");
2279 		goto error;
2280 	}
2281 
2282 	speed = 0x0;
2283 	if (*link_speeds == ETH_LINK_SPEED_AUTONEG) {
2284 		speed = (hw->mac.type != ixgbe_mac_82598EB) ?
2285 				IXGBE_LINK_SPEED_82599_AUTONEG :
2286 				IXGBE_LINK_SPEED_82598_AUTONEG;
2287 	} else {
2288 		if (*link_speeds & ETH_LINK_SPEED_10G)
2289 			speed |= IXGBE_LINK_SPEED_10GB_FULL;
2290 		if (*link_speeds & ETH_LINK_SPEED_1G)
2291 			speed |= IXGBE_LINK_SPEED_1GB_FULL;
2292 		if (*link_speeds & ETH_LINK_SPEED_100M)
2293 			speed |= IXGBE_LINK_SPEED_100_FULL;
2294 	}
2295 
2296 	err = ixgbe_setup_link(hw, speed, link_up);
2297 	if (err)
2298 		goto error;
2299 
2300 skip_link_setup:
2301 
2302 	if (rte_intr_allow_others(intr_handle)) {
2303 		/* check if lsc interrupt is enabled */
2304 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2305 			ixgbe_dev_lsc_interrupt_setup(dev);
2306 	} else {
2307 		rte_intr_callback_unregister(intr_handle,
2308 					     ixgbe_dev_interrupt_handler,
2309 					     (void *)dev);
2310 		if (dev->data->dev_conf.intr_conf.lsc != 0)
2311 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
2312 				     " no intr multiplex\n");
2313 	}
2314 
2315 	/* check if rxq interrupt is enabled */
2316 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2317 	    rte_intr_dp_is_en(intr_handle))
2318 		ixgbe_dev_rxq_interrupt_setup(dev);
2319 
2320 	/* enable uio/vfio intr/eventfd mapping */
2321 	rte_intr_enable(intr_handle);
2322 
2323 	/* resume enabled intr since hw reset */
2324 	ixgbe_enable_intr(dev);
2325 
2326 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
2327 		ETH_VLAN_EXTEND_MASK;
2328 	ixgbe_vlan_offload_set(dev, mask);
2329 
2330 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
2331 		/* Enable vlan filtering for VMDq */
2332 		ixgbe_vmdq_vlan_hw_filter_enable(dev);
2333 	}
2334 
2335 	/* Configure DCB hw */
2336 	ixgbe_configure_dcb(dev);
2337 
2338 	if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2339 		err = ixgbe_fdir_configure(dev);
2340 		if (err)
2341 			goto error;
2342 	}
2343 
2344 	/* Restore vf rate limit */
2345 	if (vfinfo != NULL) {
2346 		for (vf = 0; vf < dev->pci_dev->max_vfs; vf++)
2347 			for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2348 				if (vfinfo[vf].tx_rate[idx] != 0)
2349 					ixgbe_set_vf_rate_limit(dev, vf,
2350 						vfinfo[vf].tx_rate[idx],
2351 						1 << idx);
2352 	}
2353 
2354 	ixgbe_restore_statistics_mapping(dev);
2355 
2356 	return 0;
2357 
2358 error:
2359 	PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2360 	ixgbe_dev_clear_queues(dev);
2361 	return -EIO;
2362 }
2363 
2364 /*
2365  * Stop device: disable rx and tx functions to allow for reconfiguring.
2366  */
2367 static void
2368 ixgbe_dev_stop(struct rte_eth_dev *dev)
2369 {
2370 	struct rte_eth_link link;
2371 	struct ixgbe_hw *hw =
2372 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2373 	struct ixgbe_vf_info *vfinfo =
2374 		*IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2375 	struct ixgbe_filter_info *filter_info =
2376 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
2377 	struct ixgbe_5tuple_filter *p_5tuple, *p_5tuple_next;
2378 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
2379 	int vf;
2380 
2381 	PMD_INIT_FUNC_TRACE();
2382 
2383 	/* disable interrupts */
2384 	ixgbe_disable_intr(hw);
2385 
2386 	/* reset the NIC */
2387 	ixgbe_pf_reset_hw(hw);
2388 	hw->adapter_stopped = 0;
2389 
2390 	/* stop adapter */
2391 	ixgbe_stop_adapter(hw);
2392 
2393 	for (vf = 0; vfinfo != NULL &&
2394 		     vf < dev->pci_dev->max_vfs; vf++)
2395 		vfinfo[vf].clear_to_send = false;
2396 
2397 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2398 		/* Turn off the copper */
2399 		ixgbe_set_phy_power(hw, false);
2400 	} else {
2401 		/* Turn off the laser */
2402 		ixgbe_disable_tx_laser(hw);
2403 	}
2404 
2405 	ixgbe_dev_clear_queues(dev);
2406 
2407 	/* Clear stored conf */
2408 	dev->data->scattered_rx = 0;
2409 	dev->data->lro = 0;
2410 
2411 	/* Clear recorded link status */
2412 	memset(&link, 0, sizeof(link));
2413 	rte_ixgbe_dev_atomic_write_link_status(dev, &link);
2414 
2415 	/* Remove all ntuple filters of the device */
2416 	for (p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list);
2417 	     p_5tuple != NULL; p_5tuple = p_5tuple_next) {
2418 		p_5tuple_next = TAILQ_NEXT(p_5tuple, entries);
2419 		TAILQ_REMOVE(&filter_info->fivetuple_list,
2420 			     p_5tuple, entries);
2421 		rte_free(p_5tuple);
2422 	}
2423 	memset(filter_info->fivetuple_mask, 0,
2424 		sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
2425 
2426 	if (!rte_intr_allow_others(intr_handle))
2427 		/* resume to the default handler */
2428 		rte_intr_callback_register(intr_handle,
2429 					   ixgbe_dev_interrupt_handler,
2430 					   (void *)dev);
2431 
2432 	/* Clean datapath event and queue/vec mapping */
2433 	rte_intr_efd_disable(intr_handle);
2434 	if (intr_handle->intr_vec != NULL) {
2435 		rte_free(intr_handle->intr_vec);
2436 		intr_handle->intr_vec = NULL;
2437 	}
2438 }
2439 
2440 /*
2441  * Set device link up: enable tx.
2442  */
2443 static int
2444 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2445 {
2446 	struct ixgbe_hw *hw =
2447 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2448 	if (hw->mac.type == ixgbe_mac_82599EB) {
2449 #ifdef RTE_NIC_BYPASS
2450 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2451 			/* Not suported in bypass mode */
2452 			PMD_INIT_LOG(ERR, "Set link up is not supported "
2453 				     "by device id 0x%x", hw->device_id);
2454 			return -ENOTSUP;
2455 		}
2456 #endif
2457 	}
2458 
2459 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2460 		/* Turn on the copper */
2461 		ixgbe_set_phy_power(hw, true);
2462 	} else {
2463 		/* Turn on the laser */
2464 		ixgbe_enable_tx_laser(hw);
2465 	}
2466 
2467 	return 0;
2468 }
2469 
2470 /*
2471  * Set device link down: disable tx.
2472  */
2473 static int
2474 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2475 {
2476 	struct ixgbe_hw *hw =
2477 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2478 	if (hw->mac.type == ixgbe_mac_82599EB) {
2479 #ifdef RTE_NIC_BYPASS
2480 		if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2481 			/* Not suported in bypass mode */
2482 			PMD_INIT_LOG(ERR, "Set link down is not supported "
2483 				     "by device id 0x%x", hw->device_id);
2484 			return -ENOTSUP;
2485 		}
2486 #endif
2487 	}
2488 
2489 	if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2490 		/* Turn off the copper */
2491 		ixgbe_set_phy_power(hw, false);
2492 	} else {
2493 		/* Turn off the laser */
2494 		ixgbe_disable_tx_laser(hw);
2495 	}
2496 
2497 	return 0;
2498 }
2499 
2500 /*
2501  * Reest and stop device.
2502  */
2503 static void
2504 ixgbe_dev_close(struct rte_eth_dev *dev)
2505 {
2506 	struct ixgbe_hw *hw =
2507 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2508 
2509 	PMD_INIT_FUNC_TRACE();
2510 
2511 	ixgbe_pf_reset_hw(hw);
2512 
2513 	ixgbe_dev_stop(dev);
2514 	hw->adapter_stopped = 1;
2515 
2516 	ixgbe_dev_free_queues(dev);
2517 
2518 	ixgbe_disable_pcie_master(hw);
2519 
2520 	/* reprogram the RAR[0] in case user changed it. */
2521 	ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
2522 }
2523 
2524 static void
2525 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
2526 			   struct ixgbe_hw_stats *hw_stats,
2527 			   uint64_t *total_missed_rx, uint64_t *total_qbrc,
2528 			   uint64_t *total_qprc, uint64_t *total_qprdc)
2529 {
2530 	uint32_t bprc, lxon, lxoff, total;
2531 	uint32_t delta_gprc = 0;
2532 	unsigned i;
2533 	/* Workaround for RX byte count not including CRC bytes when CRC
2534 +	 * strip is enabled. CRC bytes are removed from counters when crc_strip
2535 	 * is disabled.
2536 +	 */
2537 	int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
2538 			IXGBE_HLREG0_RXCRCSTRP);
2539 
2540 	hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
2541 	hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
2542 	hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
2543 	hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
2544 
2545 	for (i = 0; i < 8; i++) {
2546 		uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
2547 
2548 		/* global total per queue */
2549 		hw_stats->mpc[i] += mp;
2550 		/* Running comprehensive total for stats display */
2551 		*total_missed_rx += hw_stats->mpc[i];
2552 		if (hw->mac.type == ixgbe_mac_82598EB) {
2553 			hw_stats->rnbc[i] +=
2554 			    IXGBE_READ_REG(hw, IXGBE_RNBC(i));
2555 			hw_stats->pxonrxc[i] +=
2556 				IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
2557 			hw_stats->pxoffrxc[i] +=
2558 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
2559 		} else {
2560 			hw_stats->pxonrxc[i] +=
2561 				IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
2562 			hw_stats->pxoffrxc[i] +=
2563 				IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
2564 			hw_stats->pxon2offc[i] +=
2565 				IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
2566 		}
2567 		hw_stats->pxontxc[i] +=
2568 		    IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
2569 		hw_stats->pxofftxc[i] +=
2570 		    IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
2571 	}
2572 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
2573 		uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
2574 		uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
2575 		uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
2576 
2577 		delta_gprc += delta_qprc;
2578 
2579 		hw_stats->qprc[i] += delta_qprc;
2580 		hw_stats->qptc[i] += delta_qptc;
2581 
2582 		hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
2583 		hw_stats->qbrc[i] +=
2584 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
2585 		if (crc_strip == 0)
2586 			hw_stats->qbrc[i] -= delta_qprc * ETHER_CRC_LEN;
2587 
2588 		hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
2589 		hw_stats->qbtc[i] +=
2590 		    ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
2591 
2592 		hw_stats->qprdc[i] += delta_qprdc;
2593 		*total_qprdc += hw_stats->qprdc[i];
2594 
2595 		*total_qprc += hw_stats->qprc[i];
2596 		*total_qbrc += hw_stats->qbrc[i];
2597 	}
2598 	hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
2599 	hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
2600 	hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
2601 
2602 	/*
2603 	 * An errata states that gprc actually counts good + missed packets:
2604 	 * Workaround to set gprc to summated queue packet receives
2605 	 */
2606 	hw_stats->gprc = *total_qprc;
2607 
2608 	if (hw->mac.type != ixgbe_mac_82598EB) {
2609 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
2610 		hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
2611 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
2612 		hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
2613 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
2614 		hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
2615 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
2616 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
2617 	} else {
2618 		hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
2619 		hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
2620 		/* 82598 only has a counter in the high register */
2621 		hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
2622 		hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
2623 		hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
2624 	}
2625 	uint64_t old_tpr = hw_stats->tpr;
2626 
2627 	hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
2628 	hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
2629 
2630 	if (crc_strip == 0)
2631 		hw_stats->gorc -= delta_gprc * ETHER_CRC_LEN;
2632 
2633 	uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
2634 	hw_stats->gptc += delta_gptc;
2635 	hw_stats->gotc -= delta_gptc * ETHER_CRC_LEN;
2636 	hw_stats->tor -= (hw_stats->tpr - old_tpr) * ETHER_CRC_LEN;
2637 
2638 	/*
2639 	 * Workaround: mprc hardware is incorrectly counting
2640 	 * broadcasts, so for now we subtract those.
2641 	 */
2642 	bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
2643 	hw_stats->bprc += bprc;
2644 	hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
2645 	if (hw->mac.type == ixgbe_mac_82598EB)
2646 		hw_stats->mprc -= bprc;
2647 
2648 	hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
2649 	hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
2650 	hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
2651 	hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
2652 	hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
2653 	hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
2654 
2655 	lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
2656 	hw_stats->lxontxc += lxon;
2657 	lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
2658 	hw_stats->lxofftxc += lxoff;
2659 	total = lxon + lxoff;
2660 
2661 	hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
2662 	hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
2663 	hw_stats->gptc -= total;
2664 	hw_stats->mptc -= total;
2665 	hw_stats->ptc64 -= total;
2666 	hw_stats->gotc -= total * ETHER_MIN_LEN;
2667 
2668 	hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
2669 	hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
2670 	hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
2671 	hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
2672 	hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
2673 	hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
2674 	hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
2675 	hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
2676 	hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
2677 	hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
2678 	hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
2679 	hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
2680 	hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
2681 	hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
2682 	hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
2683 	hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
2684 	/* Only read FCOE on 82599 */
2685 	if (hw->mac.type != ixgbe_mac_82598EB) {
2686 		hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
2687 		hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
2688 		hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
2689 		hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
2690 		hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
2691 	}
2692 
2693 	/* Flow Director Stats registers */
2694 	hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
2695 	hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
2696 }
2697 
2698 /*
2699  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
2700  */
2701 static void
2702 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
2703 {
2704 	struct ixgbe_hw *hw =
2705 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2706 	struct ixgbe_hw_stats *hw_stats =
2707 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2708 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
2709 	unsigned i;
2710 
2711 	total_missed_rx = 0;
2712 	total_qbrc = 0;
2713 	total_qprc = 0;
2714 	total_qprdc = 0;
2715 
2716 	ixgbe_read_stats_registers(hw, hw_stats, &total_missed_rx, &total_qbrc,
2717 			&total_qprc, &total_qprdc);
2718 
2719 	if (stats == NULL)
2720 		return;
2721 
2722 	/* Fill out the rte_eth_stats statistics structure */
2723 	stats->ipackets = total_qprc;
2724 	stats->ibytes = total_qbrc;
2725 	stats->opackets = hw_stats->gptc;
2726 	stats->obytes = hw_stats->gotc;
2727 
2728 	for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
2729 		stats->q_ipackets[i] = hw_stats->qprc[i];
2730 		stats->q_opackets[i] = hw_stats->qptc[i];
2731 		stats->q_ibytes[i] = hw_stats->qbrc[i];
2732 		stats->q_obytes[i] = hw_stats->qbtc[i];
2733 		stats->q_errors[i] = hw_stats->qprdc[i];
2734 	}
2735 
2736 	/* Rx Errors */
2737 	stats->imissed  = total_missed_rx;
2738 	stats->ierrors  = hw_stats->crcerrs +
2739 			  hw_stats->mspdc +
2740 			  hw_stats->rlec +
2741 			  hw_stats->ruc +
2742 			  hw_stats->roc +
2743 			  hw_stats->illerrc +
2744 			  hw_stats->errbc +
2745 			  hw_stats->rfc +
2746 			  hw_stats->fccrc +
2747 			  hw_stats->fclast;
2748 
2749 	/* Tx Errors */
2750 	stats->oerrors  = 0;
2751 }
2752 
2753 static void
2754 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
2755 {
2756 	struct ixgbe_hw_stats *stats =
2757 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2758 
2759 	/* HW registers are cleared on read */
2760 	ixgbe_dev_stats_get(dev, NULL);
2761 
2762 	/* Reset software totals */
2763 	memset(stats, 0, sizeof(*stats));
2764 }
2765 
2766 /* This function calculates the number of xstats based on the current config */
2767 static unsigned
2768 ixgbe_xstats_calc_num(void) {
2769 	return IXGBE_NB_HW_STATS +
2770 		(IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
2771 		(IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
2772 }
2773 
2774 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
2775 	struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned limit)
2776 {
2777 	const unsigned cnt_stats = ixgbe_xstats_calc_num();
2778 	unsigned stat, i, count;
2779 
2780 	if (xstats_names != NULL) {
2781 		count = 0;
2782 
2783 		/* Note: limit >= cnt_stats checked upstream
2784 		 * in rte_eth_xstats_names()
2785 		 */
2786 
2787 		/* Extended stats from ixgbe_hw_stats */
2788 		for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
2789 			snprintf(xstats_names[count].name,
2790 				sizeof(xstats_names[count].name),
2791 				"%s",
2792 				rte_ixgbe_stats_strings[i].name);
2793 			count++;
2794 		}
2795 
2796 		/* RX Priority Stats */
2797 		for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
2798 			for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
2799 				snprintf(xstats_names[count].name,
2800 					sizeof(xstats_names[count].name),
2801 					"rx_priority%u_%s", i,
2802 					rte_ixgbe_rxq_strings[stat].name);
2803 				count++;
2804 			}
2805 		}
2806 
2807 		/* TX Priority Stats */
2808 		for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
2809 			for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
2810 				snprintf(xstats_names[count].name,
2811 					sizeof(xstats_names[count].name),
2812 					"tx_priority%u_%s", i,
2813 					rte_ixgbe_txq_strings[stat].name);
2814 				count++;
2815 			}
2816 		}
2817 	}
2818 	return cnt_stats;
2819 }
2820 
2821 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
2822 	struct rte_eth_xstat_name *xstats_names, unsigned limit)
2823 {
2824 	unsigned i;
2825 
2826 	if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
2827 		return -ENOMEM;
2828 
2829 	if (xstats_names != NULL)
2830 		for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
2831 			snprintf(xstats_names[i].name,
2832 				sizeof(xstats_names[i].name),
2833 				"%s", rte_ixgbevf_stats_strings[i].name);
2834 	return IXGBEVF_NB_XSTATS;
2835 }
2836 
2837 static int
2838 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
2839 					 unsigned n)
2840 {
2841 	struct ixgbe_hw *hw =
2842 			IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2843 	struct ixgbe_hw_stats *hw_stats =
2844 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2845 	uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
2846 	unsigned i, stat, count = 0;
2847 
2848 	count = ixgbe_xstats_calc_num();
2849 
2850 	if (n < count)
2851 		return count;
2852 
2853 	total_missed_rx = 0;
2854 	total_qbrc = 0;
2855 	total_qprc = 0;
2856 	total_qprdc = 0;
2857 
2858 	ixgbe_read_stats_registers(hw, hw_stats, &total_missed_rx, &total_qbrc,
2859 				   &total_qprc, &total_qprdc);
2860 
2861 	/* If this is a reset xstats is NULL, and we have cleared the
2862 	 * registers by reading them.
2863 	 */
2864 	if (!xstats)
2865 		return 0;
2866 
2867 	/* Extended stats from ixgbe_hw_stats */
2868 	count = 0;
2869 	for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
2870 		xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
2871 				rte_ixgbe_stats_strings[i].offset);
2872 		count++;
2873 	}
2874 
2875 	/* RX Priority Stats */
2876 	for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
2877 		for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
2878 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
2879 					rte_ixgbe_rxq_strings[stat].offset +
2880 					(sizeof(uint64_t) * i));
2881 			count++;
2882 		}
2883 	}
2884 
2885 	/* TX Priority Stats */
2886 	for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
2887 		for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
2888 			xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
2889 					rte_ixgbe_txq_strings[stat].offset +
2890 					(sizeof(uint64_t) * i));
2891 			count++;
2892 		}
2893 	}
2894 	return count;
2895 }
2896 
2897 static void
2898 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
2899 {
2900 	struct ixgbe_hw_stats *stats =
2901 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2902 
2903 	unsigned count = ixgbe_xstats_calc_num();
2904 
2905 	/* HW registers are cleared on read */
2906 	ixgbe_dev_xstats_get(dev, NULL, count);
2907 
2908 	/* Reset software totals */
2909 	memset(stats, 0, sizeof(*stats));
2910 }
2911 
2912 static void
2913 ixgbevf_update_stats(struct rte_eth_dev *dev)
2914 {
2915 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2916 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
2917 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2918 
2919 	/* Good Rx packet, include VF loopback */
2920 	UPDATE_VF_STAT(IXGBE_VFGPRC,
2921 	    hw_stats->last_vfgprc, hw_stats->vfgprc);
2922 
2923 	/* Good Rx octets, include VF loopback */
2924 	UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
2925 	    hw_stats->last_vfgorc, hw_stats->vfgorc);
2926 
2927 	/* Good Tx packet, include VF loopback */
2928 	UPDATE_VF_STAT(IXGBE_VFGPTC,
2929 	    hw_stats->last_vfgptc, hw_stats->vfgptc);
2930 
2931 	/* Good Tx octets, include VF loopback */
2932 	UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
2933 	    hw_stats->last_vfgotc, hw_stats->vfgotc);
2934 
2935 	/* Rx Multicst Packet */
2936 	UPDATE_VF_STAT(IXGBE_VFMPRC,
2937 	    hw_stats->last_vfmprc, hw_stats->vfmprc);
2938 }
2939 
2940 static int
2941 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
2942 		       unsigned n)
2943 {
2944 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
2945 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2946 	unsigned i;
2947 
2948 	if (n < IXGBEVF_NB_XSTATS)
2949 		return IXGBEVF_NB_XSTATS;
2950 
2951 	ixgbevf_update_stats(dev);
2952 
2953 	if (!xstats)
2954 		return 0;
2955 
2956 	/* Extended stats */
2957 	for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
2958 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
2959 			rte_ixgbevf_stats_strings[i].offset);
2960 	}
2961 
2962 	return IXGBEVF_NB_XSTATS;
2963 }
2964 
2965 static void
2966 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
2967 {
2968 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
2969 			  IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2970 
2971 	ixgbevf_update_stats(dev);
2972 
2973 	if (stats == NULL)
2974 		return;
2975 
2976 	stats->ipackets = hw_stats->vfgprc;
2977 	stats->ibytes = hw_stats->vfgorc;
2978 	stats->opackets = hw_stats->vfgptc;
2979 	stats->obytes = hw_stats->vfgotc;
2980 }
2981 
2982 static void
2983 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
2984 {
2985 	struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
2986 			IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2987 
2988 	/* Sync HW register to the last stats */
2989 	ixgbevf_dev_stats_get(dev, NULL);
2990 
2991 	/* reset HW current stats*/
2992 	hw_stats->vfgprc = 0;
2993 	hw_stats->vfgorc = 0;
2994 	hw_stats->vfgptc = 0;
2995 	hw_stats->vfgotc = 0;
2996 }
2997 
2998 static void
2999 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3000 {
3001 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3002 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3003 
3004 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3005 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3006 	if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3007 		/*
3008 		 * When DCB/VT is off, maximum number of queues changes,
3009 		 * except for 82598EB, which remains constant.
3010 		 */
3011 		if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
3012 				hw->mac.type != ixgbe_mac_82598EB)
3013 			dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3014 	}
3015 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3016 	dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3017 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3018 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3019 	dev_info->max_vfs = dev->pci_dev->max_vfs;
3020 	if (hw->mac.type == ixgbe_mac_82598EB)
3021 		dev_info->max_vmdq_pools = ETH_16_POOLS;
3022 	else
3023 		dev_info->max_vmdq_pools = ETH_64_POOLS;
3024 	dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3025 	dev_info->rx_offload_capa =
3026 		DEV_RX_OFFLOAD_VLAN_STRIP |
3027 		DEV_RX_OFFLOAD_IPV4_CKSUM |
3028 		DEV_RX_OFFLOAD_UDP_CKSUM  |
3029 		DEV_RX_OFFLOAD_TCP_CKSUM;
3030 
3031 	/*
3032 	 * RSC is only supported by 82599 and x540 PF devices in a non-SR-IOV
3033 	 * mode.
3034 	 */
3035 	if ((hw->mac.type == ixgbe_mac_82599EB ||
3036 	     hw->mac.type == ixgbe_mac_X540) &&
3037 	    !RTE_ETH_DEV_SRIOV(dev).active)
3038 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TCP_LRO;
3039 
3040 	if (hw->mac.type == ixgbe_mac_X550 ||
3041 	    hw->mac.type == ixgbe_mac_X550EM_x ||
3042 	    hw->mac.type == ixgbe_mac_X550EM_a)
3043 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM;
3044 
3045 	dev_info->tx_offload_capa =
3046 		DEV_TX_OFFLOAD_VLAN_INSERT |
3047 		DEV_TX_OFFLOAD_IPV4_CKSUM  |
3048 		DEV_TX_OFFLOAD_UDP_CKSUM   |
3049 		DEV_TX_OFFLOAD_TCP_CKSUM   |
3050 		DEV_TX_OFFLOAD_SCTP_CKSUM  |
3051 		DEV_TX_OFFLOAD_TCP_TSO;
3052 
3053 	if (hw->mac.type == ixgbe_mac_X550 ||
3054 	    hw->mac.type == ixgbe_mac_X550EM_x ||
3055 	    hw->mac.type == ixgbe_mac_X550EM_a)
3056 		dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM;
3057 
3058 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3059 		.rx_thresh = {
3060 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3061 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3062 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3063 		},
3064 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3065 		.rx_drop_en = 0,
3066 	};
3067 
3068 	dev_info->default_txconf = (struct rte_eth_txconf) {
3069 		.tx_thresh = {
3070 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3071 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3072 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3073 		},
3074 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3075 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3076 		.txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
3077 				ETH_TXQ_FLAGS_NOOFFLOADS,
3078 	};
3079 
3080 	dev_info->rx_desc_lim = rx_desc_lim;
3081 	dev_info->tx_desc_lim = tx_desc_lim;
3082 
3083 	dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3084 	dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3085 	dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3086 
3087 	dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G;
3088 	if (hw->mac.type == ixgbe_mac_X540 ||
3089 	    hw->mac.type == ixgbe_mac_X540_vf ||
3090 	    hw->mac.type == ixgbe_mac_X550 ||
3091 	    hw->mac.type == ixgbe_mac_X550_vf) {
3092 		dev_info->speed_capa |= ETH_LINK_SPEED_100M;
3093 	}
3094 }
3095 
3096 static const uint32_t *
3097 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3098 {
3099 	static const uint32_t ptypes[] = {
3100 		/* For non-vec functions,
3101 		 * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3102 		 * for vec functions,
3103 		 * refers to _recv_raw_pkts_vec().
3104 		 */
3105 		RTE_PTYPE_L2_ETHER,
3106 		RTE_PTYPE_L3_IPV4,
3107 		RTE_PTYPE_L3_IPV4_EXT,
3108 		RTE_PTYPE_L3_IPV6,
3109 		RTE_PTYPE_L3_IPV6_EXT,
3110 		RTE_PTYPE_L4_SCTP,
3111 		RTE_PTYPE_L4_TCP,
3112 		RTE_PTYPE_L4_UDP,
3113 		RTE_PTYPE_TUNNEL_IP,
3114 		RTE_PTYPE_INNER_L3_IPV6,
3115 		RTE_PTYPE_INNER_L3_IPV6_EXT,
3116 		RTE_PTYPE_INNER_L4_TCP,
3117 		RTE_PTYPE_INNER_L4_UDP,
3118 		RTE_PTYPE_UNKNOWN
3119 	};
3120 
3121 	if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3122 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3123 	    dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3124 	    dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3125 		return ptypes;
3126 	return NULL;
3127 }
3128 
3129 static void
3130 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3131 		     struct rte_eth_dev_info *dev_info)
3132 {
3133 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3134 
3135 	dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3136 	dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3137 	dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3138 	dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS reg */
3139 	dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3140 	dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3141 	dev_info->max_vfs = dev->pci_dev->max_vfs;
3142 	if (hw->mac.type == ixgbe_mac_82598EB)
3143 		dev_info->max_vmdq_pools = ETH_16_POOLS;
3144 	else
3145 		dev_info->max_vmdq_pools = ETH_64_POOLS;
3146 	dev_info->rx_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP |
3147 				DEV_RX_OFFLOAD_IPV4_CKSUM |
3148 				DEV_RX_OFFLOAD_UDP_CKSUM  |
3149 				DEV_RX_OFFLOAD_TCP_CKSUM;
3150 	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
3151 				DEV_TX_OFFLOAD_IPV4_CKSUM  |
3152 				DEV_TX_OFFLOAD_UDP_CKSUM   |
3153 				DEV_TX_OFFLOAD_TCP_CKSUM   |
3154 				DEV_TX_OFFLOAD_SCTP_CKSUM  |
3155 				DEV_TX_OFFLOAD_TCP_TSO;
3156 
3157 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
3158 		.rx_thresh = {
3159 			.pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3160 			.hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3161 			.wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3162 		},
3163 		.rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3164 		.rx_drop_en = 0,
3165 	};
3166 
3167 	dev_info->default_txconf = (struct rte_eth_txconf) {
3168 		.tx_thresh = {
3169 			.pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3170 			.hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3171 			.wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3172 		},
3173 		.tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3174 		.tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3175 		.txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
3176 				ETH_TXQ_FLAGS_NOOFFLOADS,
3177 	};
3178 
3179 	dev_info->rx_desc_lim = rx_desc_lim;
3180 	dev_info->tx_desc_lim = tx_desc_lim;
3181 }
3182 
3183 /* return 0 means link status changed, -1 means not changed */
3184 static int
3185 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
3186 {
3187 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3188 	struct rte_eth_link link, old;
3189 	ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
3190 	int link_up;
3191 	int diag;
3192 
3193 	link.link_status = ETH_LINK_DOWN;
3194 	link.link_speed = 0;
3195 	link.link_duplex = ETH_LINK_HALF_DUPLEX;
3196 	memset(&old, 0, sizeof(old));
3197 	rte_ixgbe_dev_atomic_read_link_status(dev, &old);
3198 
3199 	hw->mac.get_link_status = true;
3200 
3201 	/* check if it needs to wait to complete, if lsc interrupt is enabled */
3202 	if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
3203 		diag = ixgbe_check_link(hw, &link_speed, &link_up, 0);
3204 	else
3205 		diag = ixgbe_check_link(hw, &link_speed, &link_up, 1);
3206 
3207 	if (diag != 0) {
3208 		link.link_speed = ETH_SPEED_NUM_100M;
3209 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
3210 		rte_ixgbe_dev_atomic_write_link_status(dev, &link);
3211 		if (link.link_status == old.link_status)
3212 			return -1;
3213 		return 0;
3214 	}
3215 
3216 	if (link_up == 0) {
3217 		rte_ixgbe_dev_atomic_write_link_status(dev, &link);
3218 		if (link.link_status == old.link_status)
3219 			return -1;
3220 		return 0;
3221 	}
3222 	link.link_status = ETH_LINK_UP;
3223 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
3224 
3225 	switch (link_speed) {
3226 	default:
3227 	case IXGBE_LINK_SPEED_UNKNOWN:
3228 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
3229 		link.link_speed = ETH_SPEED_NUM_100M;
3230 		break;
3231 
3232 	case IXGBE_LINK_SPEED_100_FULL:
3233 		link.link_speed = ETH_SPEED_NUM_100M;
3234 		break;
3235 
3236 	case IXGBE_LINK_SPEED_1GB_FULL:
3237 		link.link_speed = ETH_SPEED_NUM_1G;
3238 		break;
3239 
3240 	case IXGBE_LINK_SPEED_10GB_FULL:
3241 		link.link_speed = ETH_SPEED_NUM_10G;
3242 		break;
3243 	}
3244 	rte_ixgbe_dev_atomic_write_link_status(dev, &link);
3245 
3246 	if (link.link_status == old.link_status)
3247 		return -1;
3248 
3249 	return 0;
3250 }
3251 
3252 static void
3253 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
3254 {
3255 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3256 	uint32_t fctrl;
3257 
3258 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3259 	fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3260 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3261 }
3262 
3263 static void
3264 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
3265 {
3266 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3267 	uint32_t fctrl;
3268 
3269 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3270 	fctrl &= (~IXGBE_FCTRL_UPE);
3271 	if (dev->data->all_multicast == 1)
3272 		fctrl |= IXGBE_FCTRL_MPE;
3273 	else
3274 		fctrl &= (~IXGBE_FCTRL_MPE);
3275 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3276 }
3277 
3278 static void
3279 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
3280 {
3281 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3282 	uint32_t fctrl;
3283 
3284 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3285 	fctrl |= IXGBE_FCTRL_MPE;
3286 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3287 }
3288 
3289 static void
3290 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
3291 {
3292 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3293 	uint32_t fctrl;
3294 
3295 	if (dev->data->promiscuous == 1)
3296 		return; /* must remain in all_multicast mode */
3297 
3298 	fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3299 	fctrl &= (~IXGBE_FCTRL_MPE);
3300 	IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3301 }
3302 
3303 /**
3304  * It clears the interrupt causes and enables the interrupt.
3305  * It will be called once only during nic initialized.
3306  *
3307  * @param dev
3308  *  Pointer to struct rte_eth_dev.
3309  *
3310  * @return
3311  *  - On success, zero.
3312  *  - On failure, a negative value.
3313  */
3314 static int
3315 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev)
3316 {
3317 	struct ixgbe_interrupt *intr =
3318 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3319 
3320 	ixgbe_dev_link_status_print(dev);
3321 	intr->mask |= IXGBE_EICR_LSC;
3322 
3323 	return 0;
3324 }
3325 
3326 /**
3327  * It clears the interrupt causes and enables the interrupt.
3328  * It will be called once only during nic initialized.
3329  *
3330  * @param dev
3331  *  Pointer to struct rte_eth_dev.
3332  *
3333  * @return
3334  *  - On success, zero.
3335  *  - On failure, a negative value.
3336  */
3337 static int
3338 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
3339 {
3340 	struct ixgbe_interrupt *intr =
3341 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3342 
3343 	intr->mask |= IXGBE_EICR_RTX_QUEUE;
3344 
3345 	return 0;
3346 }
3347 
3348 /*
3349  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
3350  *
3351  * @param dev
3352  *  Pointer to struct rte_eth_dev.
3353  *
3354  * @return
3355  *  - On success, zero.
3356  *  - On failure, a negative value.
3357  */
3358 static int
3359 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
3360 {
3361 	uint32_t eicr;
3362 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3363 	struct ixgbe_interrupt *intr =
3364 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3365 
3366 	/* clear all cause mask */
3367 	ixgbe_disable_intr(hw);
3368 
3369 	/* read-on-clear nic registers here */
3370 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3371 	PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
3372 
3373 	intr->flags = 0;
3374 
3375 	/* set flag for async link update */
3376 	if (eicr & IXGBE_EICR_LSC)
3377 		intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
3378 
3379 	if (eicr & IXGBE_EICR_MAILBOX)
3380 		intr->flags |= IXGBE_FLAG_MAILBOX;
3381 
3382 	if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
3383 	    hw->phy.type == ixgbe_phy_x550em_ext_t &&
3384 	    (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
3385 		intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
3386 
3387 	return 0;
3388 }
3389 
3390 /**
3391  * It gets and then prints the link status.
3392  *
3393  * @param dev
3394  *  Pointer to struct rte_eth_dev.
3395  *
3396  * @return
3397  *  - On success, zero.
3398  *  - On failure, a negative value.
3399  */
3400 static void
3401 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
3402 {
3403 	struct rte_eth_link link;
3404 
3405 	memset(&link, 0, sizeof(link));
3406 	rte_ixgbe_dev_atomic_read_link_status(dev, &link);
3407 	if (link.link_status) {
3408 		PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
3409 					(int)(dev->data->port_id),
3410 					(unsigned)link.link_speed,
3411 			link.link_duplex == ETH_LINK_FULL_DUPLEX ?
3412 					"full-duplex" : "half-duplex");
3413 	} else {
3414 		PMD_INIT_LOG(INFO, " Port %d: Link Down",
3415 				(int)(dev->data->port_id));
3416 	}
3417 	PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
3418 				dev->pci_dev->addr.domain,
3419 				dev->pci_dev->addr.bus,
3420 				dev->pci_dev->addr.devid,
3421 				dev->pci_dev->addr.function);
3422 }
3423 
3424 /*
3425  * It executes link_update after knowing an interrupt occurred.
3426  *
3427  * @param dev
3428  *  Pointer to struct rte_eth_dev.
3429  *
3430  * @return
3431  *  - On success, zero.
3432  *  - On failure, a negative value.
3433  */
3434 static int
3435 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
3436 {
3437 	struct ixgbe_interrupt *intr =
3438 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3439 	int64_t timeout;
3440 	struct rte_eth_link link;
3441 	int intr_enable_delay = false;
3442 	struct ixgbe_hw *hw =
3443 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3444 
3445 	PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
3446 
3447 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
3448 		ixgbe_pf_mbx_process(dev);
3449 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
3450 	}
3451 
3452 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
3453 		ixgbe_handle_lasi(hw);
3454 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
3455 	}
3456 
3457 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
3458 		/* get the link status before link update, for predicting later */
3459 		memset(&link, 0, sizeof(link));
3460 		rte_ixgbe_dev_atomic_read_link_status(dev, &link);
3461 
3462 		ixgbe_dev_link_update(dev, 0);
3463 
3464 		/* likely to up */
3465 		if (!link.link_status)
3466 			/* handle it 1 sec later, wait it being stable */
3467 			timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
3468 		/* likely to down */
3469 		else
3470 			/* handle it 4 sec later, wait it being stable */
3471 			timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
3472 
3473 		ixgbe_dev_link_status_print(dev);
3474 
3475 		intr_enable_delay = true;
3476 	}
3477 
3478 	if (intr_enable_delay) {
3479 		if (rte_eal_alarm_set(timeout * 1000,
3480 				      ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
3481 			PMD_DRV_LOG(ERR, "Error setting alarm");
3482 	} else {
3483 		PMD_DRV_LOG(DEBUG, "enable intr immediately");
3484 		ixgbe_enable_intr(dev);
3485 		rte_intr_enable(&(dev->pci_dev->intr_handle));
3486 	}
3487 
3488 
3489 	return 0;
3490 }
3491 
3492 /**
3493  * Interrupt handler which shall be registered for alarm callback for delayed
3494  * handling specific interrupt to wait for the stable nic state. As the
3495  * NIC interrupt state is not stable for ixgbe after link is just down,
3496  * it needs to wait 4 seconds to get the stable status.
3497  *
3498  * @param handle
3499  *  Pointer to interrupt handle.
3500  * @param param
3501  *  The address of parameter (struct rte_eth_dev *) regsitered before.
3502  *
3503  * @return
3504  *  void
3505  */
3506 static void
3507 ixgbe_dev_interrupt_delayed_handler(void *param)
3508 {
3509 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
3510 	struct ixgbe_interrupt *intr =
3511 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3512 	struct ixgbe_hw *hw =
3513 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3514 	uint32_t eicr;
3515 
3516 	eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3517 	if (eicr & IXGBE_EICR_MAILBOX)
3518 		ixgbe_pf_mbx_process(dev);
3519 
3520 	if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
3521 		ixgbe_handle_lasi(hw);
3522 		intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
3523 	}
3524 
3525 	if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
3526 		ixgbe_dev_link_update(dev, 0);
3527 		intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
3528 		ixgbe_dev_link_status_print(dev);
3529 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC);
3530 	}
3531 
3532 	PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
3533 	ixgbe_enable_intr(dev);
3534 	rte_intr_enable(&(dev->pci_dev->intr_handle));
3535 }
3536 
3537 /**
3538  * Interrupt handler triggered by NIC  for handling
3539  * specific interrupt.
3540  *
3541  * @param handle
3542  *  Pointer to interrupt handle.
3543  * @param param
3544  *  The address of parameter (struct rte_eth_dev *) regsitered before.
3545  *
3546  * @return
3547  *  void
3548  */
3549 static void
3550 ixgbe_dev_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
3551 			    void *param)
3552 {
3553 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
3554 
3555 	ixgbe_dev_interrupt_get_status(dev);
3556 	ixgbe_dev_interrupt_action(dev);
3557 }
3558 
3559 static int
3560 ixgbe_dev_led_on(struct rte_eth_dev *dev)
3561 {
3562 	struct ixgbe_hw *hw;
3563 
3564 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3565 	return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
3566 }
3567 
3568 static int
3569 ixgbe_dev_led_off(struct rte_eth_dev *dev)
3570 {
3571 	struct ixgbe_hw *hw;
3572 
3573 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3574 	return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
3575 }
3576 
3577 static int
3578 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3579 {
3580 	struct ixgbe_hw *hw;
3581 	uint32_t mflcn_reg;
3582 	uint32_t fccfg_reg;
3583 	int rx_pause;
3584 	int tx_pause;
3585 
3586 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3587 
3588 	fc_conf->pause_time = hw->fc.pause_time;
3589 	fc_conf->high_water = hw->fc.high_water[0];
3590 	fc_conf->low_water = hw->fc.low_water[0];
3591 	fc_conf->send_xon = hw->fc.send_xon;
3592 	fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
3593 
3594 	/*
3595 	 * Return rx_pause status according to actual setting of
3596 	 * MFLCN register.
3597 	 */
3598 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
3599 	if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
3600 		rx_pause = 1;
3601 	else
3602 		rx_pause = 0;
3603 
3604 	/*
3605 	 * Return tx_pause status according to actual setting of
3606 	 * FCCFG register.
3607 	 */
3608 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
3609 	if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
3610 		tx_pause = 1;
3611 	else
3612 		tx_pause = 0;
3613 
3614 	if (rx_pause && tx_pause)
3615 		fc_conf->mode = RTE_FC_FULL;
3616 	else if (rx_pause)
3617 		fc_conf->mode = RTE_FC_RX_PAUSE;
3618 	else if (tx_pause)
3619 		fc_conf->mode = RTE_FC_TX_PAUSE;
3620 	else
3621 		fc_conf->mode = RTE_FC_NONE;
3622 
3623 	return 0;
3624 }
3625 
3626 static int
3627 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3628 {
3629 	struct ixgbe_hw *hw;
3630 	int err;
3631 	uint32_t rx_buf_size;
3632 	uint32_t max_high_water;
3633 	uint32_t mflcn;
3634 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
3635 		ixgbe_fc_none,
3636 		ixgbe_fc_rx_pause,
3637 		ixgbe_fc_tx_pause,
3638 		ixgbe_fc_full
3639 	};
3640 
3641 	PMD_INIT_FUNC_TRACE();
3642 
3643 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3644 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
3645 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3646 
3647 	/*
3648 	 * At least reserve one Ethernet frame for watermark
3649 	 * high_water/low_water in kilo bytes for ixgbe
3650 	 */
3651 	max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
3652 	if ((fc_conf->high_water > max_high_water) ||
3653 		(fc_conf->high_water < fc_conf->low_water)) {
3654 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
3655 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
3656 		return -EINVAL;
3657 	}
3658 
3659 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
3660 	hw->fc.pause_time     = fc_conf->pause_time;
3661 	hw->fc.high_water[0]  = fc_conf->high_water;
3662 	hw->fc.low_water[0]   = fc_conf->low_water;
3663 	hw->fc.send_xon       = fc_conf->send_xon;
3664 	hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
3665 
3666 	err = ixgbe_fc_enable(hw);
3667 
3668 	/* Not negotiated is not an error case */
3669 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED)) {
3670 
3671 		/* check if we want to forward MAC frames - driver doesn't have native
3672 		 * capability to do that, so we'll write the registers ourselves */
3673 
3674 		mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
3675 
3676 		/* set or clear MFLCN.PMCF bit depending on configuration */
3677 		if (fc_conf->mac_ctrl_frame_fwd != 0)
3678 			mflcn |= IXGBE_MFLCN_PMCF;
3679 		else
3680 			mflcn &= ~IXGBE_MFLCN_PMCF;
3681 
3682 		IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
3683 		IXGBE_WRITE_FLUSH(hw);
3684 
3685 		return 0;
3686 	}
3687 
3688 	PMD_INIT_LOG(ERR, "ixgbe_fc_enable = 0x%x", err);
3689 	return -EIO;
3690 }
3691 
3692 /**
3693  *  ixgbe_pfc_enable_generic - Enable flow control
3694  *  @hw: pointer to hardware structure
3695  *  @tc_num: traffic class number
3696  *  Enable flow control according to the current settings.
3697  */
3698 static int
3699 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
3700 {
3701 	int ret_val = 0;
3702 	uint32_t mflcn_reg, fccfg_reg;
3703 	uint32_t reg;
3704 	uint32_t fcrtl, fcrth;
3705 	uint8_t i;
3706 	uint8_t nb_rx_en;
3707 
3708 	/* Validate the water mark configuration */
3709 	if (!hw->fc.pause_time) {
3710 		ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
3711 		goto out;
3712 	}
3713 
3714 	/* Low water mark of zero causes XOFF floods */
3715 	if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
3716 		 /* High/Low water can not be 0 */
3717 		if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_water[tc_num])) {
3718 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
3719 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
3720 			goto out;
3721 		}
3722 
3723 		if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
3724 			PMD_INIT_LOG(ERR, "Invalid water mark configuration");
3725 			ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
3726 			goto out;
3727 		}
3728 	}
3729 	/* Negotiate the fc mode to use */
3730 	ixgbe_fc_autoneg(hw);
3731 
3732 	/* Disable any previous flow control settings */
3733 	mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
3734 	mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
3735 
3736 	fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
3737 	fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
3738 
3739 	switch (hw->fc.current_mode) {
3740 	case ixgbe_fc_none:
3741 		/*
3742 		 * If the count of enabled RX Priority Flow control >1,
3743 		 * and the TX pause can not be disabled
3744 		 */
3745 		nb_rx_en = 0;
3746 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
3747 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
3748 			if (reg & IXGBE_FCRTH_FCEN)
3749 				nb_rx_en++;
3750 		}
3751 		if (nb_rx_en > 1)
3752 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
3753 		break;
3754 	case ixgbe_fc_rx_pause:
3755 		/*
3756 		 * Rx Flow control is enabled and Tx Flow control is
3757 		 * disabled by software override. Since there really
3758 		 * isn't a way to advertise that we are capable of RX
3759 		 * Pause ONLY, we will advertise that we support both
3760 		 * symmetric and asymmetric Rx PAUSE.  Later, we will
3761 		 * disable the adapter's ability to send PAUSE frames.
3762 		 */
3763 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
3764 		/*
3765 		 * If the count of enabled RX Priority Flow control >1,
3766 		 * and the TX pause can not be disabled
3767 		 */
3768 		nb_rx_en = 0;
3769 		for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
3770 			reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
3771 			if (reg & IXGBE_FCRTH_FCEN)
3772 				nb_rx_en++;
3773 		}
3774 		if (nb_rx_en > 1)
3775 			fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
3776 		break;
3777 	case ixgbe_fc_tx_pause:
3778 		/*
3779 		 * Tx Flow control is enabled, and Rx Flow control is
3780 		 * disabled by software override.
3781 		 */
3782 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
3783 		break;
3784 	case ixgbe_fc_full:
3785 		/* Flow control (both Rx and Tx) is enabled by SW override. */
3786 		mflcn_reg |= IXGBE_MFLCN_RPFCE;
3787 		fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
3788 		break;
3789 	default:
3790 		PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
3791 		ret_val = IXGBE_ERR_CONFIG;
3792 		goto out;
3793 	}
3794 
3795 	/* Set 802.3x based flow control settings. */
3796 	mflcn_reg |= IXGBE_MFLCN_DPF;
3797 	IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
3798 	IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
3799 
3800 	/* Set up and enable Rx high/low water mark thresholds, enable XON. */
3801 	if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
3802 		hw->fc.high_water[tc_num]) {
3803 		fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
3804 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
3805 		fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
3806 	} else {
3807 		IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
3808 		/*
3809 		 * In order to prevent Tx hangs when the internal Tx
3810 		 * switch is enabled we must set the high water mark
3811 		 * to the maximum FCRTH value.  This allows the Tx
3812 		 * switch to function even under heavy Rx workloads.
3813 		 */
3814 		fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
3815 	}
3816 	IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
3817 
3818 	/* Configure pause time (2 TCs per register) */
3819 	reg = hw->fc.pause_time * 0x00010001;
3820 	for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
3821 		IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
3822 
3823 	/* Configure flow control refresh threshold value */
3824 	IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
3825 
3826 out:
3827 	return ret_val;
3828 }
3829 
3830 static int
3831 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
3832 {
3833 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3834 	int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
3835 
3836 	if (hw->mac.type != ixgbe_mac_82598EB) {
3837 		ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
3838 	}
3839 	return ret_val;
3840 }
3841 
3842 static int
3843 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
3844 {
3845 	int err;
3846 	uint32_t rx_buf_size;
3847 	uint32_t max_high_water;
3848 	uint8_t tc_num;
3849 	uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
3850 	struct ixgbe_hw *hw =
3851 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3852 	struct ixgbe_dcb_config *dcb_config =
3853 		IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
3854 
3855 	enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
3856 		ixgbe_fc_none,
3857 		ixgbe_fc_rx_pause,
3858 		ixgbe_fc_tx_pause,
3859 		ixgbe_fc_full
3860 	};
3861 
3862 	PMD_INIT_FUNC_TRACE();
3863 
3864 	ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
3865 	tc_num = map[pfc_conf->priority];
3866 	rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
3867 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3868 	/*
3869 	 * At least reserve one Ethernet frame for watermark
3870 	 * high_water/low_water in kilo bytes for ixgbe
3871 	 */
3872 	max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
3873 	if ((pfc_conf->fc.high_water > max_high_water) ||
3874 	    (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
3875 		PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
3876 		PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
3877 		return -EINVAL;
3878 	}
3879 
3880 	hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
3881 	hw->fc.pause_time = pfc_conf->fc.pause_time;
3882 	hw->fc.send_xon = pfc_conf->fc.send_xon;
3883 	hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
3884 	hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
3885 
3886 	err = ixgbe_dcb_pfc_enable(dev, tc_num);
3887 
3888 	/* Not negotiated is not an error case */
3889 	if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
3890 		return 0;
3891 
3892 	PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
3893 	return -EIO;
3894 }
3895 
3896 static int
3897 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
3898 			  struct rte_eth_rss_reta_entry64 *reta_conf,
3899 			  uint16_t reta_size)
3900 {
3901 	uint16_t i, sp_reta_size;
3902 	uint8_t j, mask;
3903 	uint32_t reta, r;
3904 	uint16_t idx, shift;
3905 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3906 	uint32_t reta_reg;
3907 
3908 	PMD_INIT_FUNC_TRACE();
3909 
3910 	if (!ixgbe_rss_update_sp(hw->mac.type)) {
3911 		PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
3912 			"NIC.");
3913 		return -ENOTSUP;
3914 	}
3915 
3916 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
3917 	if (reta_size != sp_reta_size) {
3918 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3919 			"(%d) doesn't match the number hardware can supported "
3920 			"(%d)\n", reta_size, sp_reta_size);
3921 		return -EINVAL;
3922 	}
3923 
3924 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
3925 		idx = i / RTE_RETA_GROUP_SIZE;
3926 		shift = i % RTE_RETA_GROUP_SIZE;
3927 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3928 						IXGBE_4_BIT_MASK);
3929 		if (!mask)
3930 			continue;
3931 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
3932 		if (mask == IXGBE_4_BIT_MASK)
3933 			r = 0;
3934 		else
3935 			r = IXGBE_READ_REG(hw, reta_reg);
3936 		for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
3937 			if (mask & (0x1 << j))
3938 				reta |= reta_conf[idx].reta[shift + j] <<
3939 							(CHAR_BIT * j);
3940 			else
3941 				reta |= r & (IXGBE_8_BIT_MASK <<
3942 						(CHAR_BIT * j));
3943 		}
3944 		IXGBE_WRITE_REG(hw, reta_reg, reta);
3945 	}
3946 
3947 	return 0;
3948 }
3949 
3950 static int
3951 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
3952 			 struct rte_eth_rss_reta_entry64 *reta_conf,
3953 			 uint16_t reta_size)
3954 {
3955 	uint16_t i, sp_reta_size;
3956 	uint8_t j, mask;
3957 	uint32_t reta;
3958 	uint16_t idx, shift;
3959 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3960 	uint32_t reta_reg;
3961 
3962 	PMD_INIT_FUNC_TRACE();
3963 	sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
3964 	if (reta_size != sp_reta_size) {
3965 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3966 			"(%d) doesn't match the number hardware can supported "
3967 			"(%d)\n", reta_size, sp_reta_size);
3968 		return -EINVAL;
3969 	}
3970 
3971 	for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
3972 		idx = i / RTE_RETA_GROUP_SIZE;
3973 		shift = i % RTE_RETA_GROUP_SIZE;
3974 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3975 						IXGBE_4_BIT_MASK);
3976 		if (!mask)
3977 			continue;
3978 
3979 		reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
3980 		reta = IXGBE_READ_REG(hw, reta_reg);
3981 		for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
3982 			if (mask & (0x1 << j))
3983 				reta_conf[idx].reta[shift + j] =
3984 					((reta >> (CHAR_BIT * j)) &
3985 						IXGBE_8_BIT_MASK);
3986 		}
3987 	}
3988 
3989 	return 0;
3990 }
3991 
3992 static void
3993 ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
3994 				uint32_t index, uint32_t pool)
3995 {
3996 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3997 	uint32_t enable_addr = 1;
3998 
3999 	ixgbe_set_rar(hw, index, mac_addr->addr_bytes, pool, enable_addr);
4000 }
4001 
4002 static void
4003 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
4004 {
4005 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4006 
4007 	ixgbe_clear_rar(hw, index);
4008 }
4009 
4010 static void
4011 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
4012 {
4013 	ixgbe_remove_rar(dev, 0);
4014 
4015 	ixgbe_add_rar(dev, addr, 0, 0);
4016 }
4017 
4018 static int
4019 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4020 {
4021 	uint32_t hlreg0;
4022 	uint32_t maxfrs;
4023 	struct ixgbe_hw *hw;
4024 	struct rte_eth_dev_info dev_info;
4025 	uint32_t frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
4026 
4027 	ixgbe_dev_info_get(dev, &dev_info);
4028 
4029 	/* check that mtu is within the allowed range */
4030 	if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
4031 		return -EINVAL;
4032 
4033 	/* refuse mtu that requires the support of scattered packets when this
4034 	 * feature has not been enabled before.
4035 	 */
4036 	if (!dev->data->scattered_rx &&
4037 	    (frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
4038 	     dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM))
4039 		return -EINVAL;
4040 
4041 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4042 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4043 
4044 	/* switch to jumbo mode if needed */
4045 	if (frame_size > ETHER_MAX_LEN) {
4046 		dev->data->dev_conf.rxmode.jumbo_frame = 1;
4047 		hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4048 	} else {
4049 		dev->data->dev_conf.rxmode.jumbo_frame = 0;
4050 		hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
4051 	}
4052 	IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4053 
4054 	/* update max frame size */
4055 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4056 
4057 	maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
4058 	maxfrs &= 0x0000FFFF;
4059 	maxfrs |= (dev->data->dev_conf.rxmode.max_rx_pkt_len << 16);
4060 	IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
4061 
4062 	return 0;
4063 }
4064 
4065 /*
4066  * Virtual Function operations
4067  */
4068 static void
4069 ixgbevf_intr_disable(struct ixgbe_hw *hw)
4070 {
4071 	PMD_INIT_FUNC_TRACE();
4072 
4073 	/* Clear interrupt mask to stop from interrupts being generated */
4074 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
4075 
4076 	IXGBE_WRITE_FLUSH(hw);
4077 }
4078 
4079 static void
4080 ixgbevf_intr_enable(struct ixgbe_hw *hw)
4081 {
4082 	PMD_INIT_FUNC_TRACE();
4083 
4084 	/* VF enable interrupt autoclean */
4085 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
4086 	IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
4087 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
4088 
4089 	IXGBE_WRITE_FLUSH(hw);
4090 }
4091 
4092 static int
4093 ixgbevf_dev_configure(struct rte_eth_dev *dev)
4094 {
4095 	struct rte_eth_conf *conf = &dev->data->dev_conf;
4096 	struct ixgbe_adapter *adapter =
4097 			(struct ixgbe_adapter *)dev->data->dev_private;
4098 
4099 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
4100 		     dev->data->port_id);
4101 
4102 	/*
4103 	 * VF has no ability to enable/disable HW CRC
4104 	 * Keep the persistent behavior the same as Host PF
4105 	 */
4106 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
4107 	if (!conf->rxmode.hw_strip_crc) {
4108 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
4109 		conf->rxmode.hw_strip_crc = 1;
4110 	}
4111 #else
4112 	if (conf->rxmode.hw_strip_crc) {
4113 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
4114 		conf->rxmode.hw_strip_crc = 0;
4115 	}
4116 #endif
4117 
4118 	/*
4119 	 * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
4120 	 * allocation or vector Rx preconditions we will reset it.
4121 	 */
4122 	adapter->rx_bulk_alloc_allowed = true;
4123 	adapter->rx_vec_allowed = true;
4124 
4125 	return 0;
4126 }
4127 
4128 static int
4129 ixgbevf_dev_start(struct rte_eth_dev *dev)
4130 {
4131 	struct ixgbe_hw *hw =
4132 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4133 	uint32_t intr_vector = 0;
4134 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
4135 
4136 	int err, mask = 0;
4137 
4138 	PMD_INIT_FUNC_TRACE();
4139 
4140 	hw->mac.ops.reset_hw(hw);
4141 	hw->mac.get_link_status = true;
4142 
4143 	/* negotiate mailbox API version to use with the PF. */
4144 	ixgbevf_negotiate_api(hw);
4145 
4146 	ixgbevf_dev_tx_init(dev);
4147 
4148 	/* This can fail when allocating mbufs for descriptor rings */
4149 	err = ixgbevf_dev_rx_init(dev);
4150 	if (err) {
4151 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
4152 		ixgbe_dev_clear_queues(dev);
4153 		return err;
4154 	}
4155 
4156 	/* Set vfta */
4157 	ixgbevf_set_vfta_all(dev, 1);
4158 
4159 	/* Set HW strip */
4160 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
4161 		ETH_VLAN_EXTEND_MASK;
4162 	ixgbevf_vlan_offload_set(dev, mask);
4163 
4164 	ixgbevf_dev_rxtx_start(dev);
4165 
4166 	/* check and configure queue intr-vector mapping */
4167 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
4168 		intr_vector = dev->data->nb_rx_queues;
4169 		if (rte_intr_efd_enable(intr_handle, intr_vector))
4170 			return -1;
4171 	}
4172 
4173 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
4174 		intr_handle->intr_vec =
4175 			rte_zmalloc("intr_vec",
4176 				    dev->data->nb_rx_queues * sizeof(int), 0);
4177 		if (intr_handle->intr_vec == NULL) {
4178 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
4179 				     " intr_vec\n", dev->data->nb_rx_queues);
4180 			return -ENOMEM;
4181 		}
4182 	}
4183 	ixgbevf_configure_msix(dev);
4184 
4185 	rte_intr_enable(intr_handle);
4186 
4187 	/* Re-enable interrupt for VF */
4188 	ixgbevf_intr_enable(hw);
4189 
4190 	return 0;
4191 }
4192 
4193 static void
4194 ixgbevf_dev_stop(struct rte_eth_dev *dev)
4195 {
4196 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4197 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
4198 
4199 	PMD_INIT_FUNC_TRACE();
4200 
4201 	ixgbevf_intr_disable(hw);
4202 
4203 	hw->adapter_stopped = 1;
4204 	ixgbe_stop_adapter(hw);
4205 
4206 	/*
4207 	  * Clear what we set, but we still keep shadow_vfta to
4208 	  * restore after device starts
4209 	  */
4210 	ixgbevf_set_vfta_all(dev, 0);
4211 
4212 	/* Clear stored conf */
4213 	dev->data->scattered_rx = 0;
4214 
4215 	ixgbe_dev_clear_queues(dev);
4216 
4217 	/* Clean datapath event and queue/vec mapping */
4218 	rte_intr_efd_disable(intr_handle);
4219 	if (intr_handle->intr_vec != NULL) {
4220 		rte_free(intr_handle->intr_vec);
4221 		intr_handle->intr_vec = NULL;
4222 	}
4223 }
4224 
4225 static void
4226 ixgbevf_dev_close(struct rte_eth_dev *dev)
4227 {
4228 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4229 
4230 	PMD_INIT_FUNC_TRACE();
4231 
4232 	ixgbe_reset_hw(hw);
4233 
4234 	ixgbevf_dev_stop(dev);
4235 
4236 	ixgbe_dev_free_queues(dev);
4237 
4238 	/**
4239 	 * Remove the VF MAC address ro ensure
4240 	 * that the VF traffic goes to the PF
4241 	 * after stop, close and detach of the VF
4242 	 **/
4243 	ixgbevf_remove_mac_addr(dev, 0);
4244 }
4245 
4246 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
4247 {
4248 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4249 	struct ixgbe_vfta *shadow_vfta =
4250 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
4251 	int i = 0, j = 0, vfta = 0, mask = 1;
4252 
4253 	for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
4254 		vfta = shadow_vfta->vfta[i];
4255 		if (vfta) {
4256 			mask = 1;
4257 			for (j = 0; j < 32; j++) {
4258 				if (vfta & mask)
4259 					ixgbe_set_vfta(hw, (i<<5)+j, 0,
4260 						       on, false);
4261 				mask <<= 1;
4262 			}
4263 		}
4264 	}
4265 
4266 }
4267 
4268 static int
4269 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
4270 {
4271 	struct ixgbe_hw *hw =
4272 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4273 	struct ixgbe_vfta *shadow_vfta =
4274 		IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
4275 	uint32_t vid_idx = 0;
4276 	uint32_t vid_bit = 0;
4277 	int ret = 0;
4278 
4279 	PMD_INIT_FUNC_TRACE();
4280 
4281 	/* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
4282 	ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
4283 	if (ret) {
4284 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
4285 		return ret;
4286 	}
4287 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
4288 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
4289 
4290 	/* Save what we set and retore it after device reset */
4291 	if (on)
4292 		shadow_vfta->vfta[vid_idx] |= vid_bit;
4293 	else
4294 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
4295 
4296 	return 0;
4297 }
4298 
4299 static void
4300 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
4301 {
4302 	struct ixgbe_hw *hw =
4303 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4304 	uint32_t ctrl;
4305 
4306 	PMD_INIT_FUNC_TRACE();
4307 
4308 	if (queue >= hw->mac.max_rx_queues)
4309 		return;
4310 
4311 	ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
4312 	if (on)
4313 		ctrl |= IXGBE_RXDCTL_VME;
4314 	else
4315 		ctrl &= ~IXGBE_RXDCTL_VME;
4316 	IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
4317 
4318 	ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
4319 }
4320 
4321 static void
4322 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
4323 {
4324 	struct ixgbe_hw *hw =
4325 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4326 	uint16_t i;
4327 	int on = 0;
4328 
4329 	/* VF function only support hw strip feature, others are not support */
4330 	if (mask & ETH_VLAN_STRIP_MASK) {
4331 		on = !!(dev->data->dev_conf.rxmode.hw_vlan_strip);
4332 
4333 		for (i = 0; i < hw->mac.max_rx_queues; i++)
4334 			ixgbevf_vlan_strip_queue_set(dev, i, on);
4335 	}
4336 }
4337 
4338 static int
4339 ixgbe_vmdq_mode_check(struct ixgbe_hw *hw)
4340 {
4341 	uint32_t reg_val;
4342 
4343 	/* we only need to do this if VMDq is enabled */
4344 	reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
4345 	if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
4346 		PMD_INIT_LOG(ERR, "VMDq must be enabled for this setting");
4347 		return -1;
4348 	}
4349 
4350 	return 0;
4351 }
4352 
4353 static uint32_t
4354 ixgbe_uta_vector(struct ixgbe_hw *hw, struct ether_addr *uc_addr)
4355 {
4356 	uint32_t vector = 0;
4357 
4358 	switch (hw->mac.mc_filter_type) {
4359 	case 0:   /* use bits [47:36] of the address */
4360 		vector = ((uc_addr->addr_bytes[4] >> 4) |
4361 			(((uint16_t)uc_addr->addr_bytes[5]) << 4));
4362 		break;
4363 	case 1:   /* use bits [46:35] of the address */
4364 		vector = ((uc_addr->addr_bytes[4] >> 3) |
4365 			(((uint16_t)uc_addr->addr_bytes[5]) << 5));
4366 		break;
4367 	case 2:   /* use bits [45:34] of the address */
4368 		vector = ((uc_addr->addr_bytes[4] >> 2) |
4369 			(((uint16_t)uc_addr->addr_bytes[5]) << 6));
4370 		break;
4371 	case 3:   /* use bits [43:32] of the address */
4372 		vector = ((uc_addr->addr_bytes[4]) |
4373 			(((uint16_t)uc_addr->addr_bytes[5]) << 8));
4374 		break;
4375 	default:  /* Invalid mc_filter_type */
4376 		break;
4377 	}
4378 
4379 	/* vector can only be 12-bits or boundary will be exceeded */
4380 	vector &= 0xFFF;
4381 	return vector;
4382 }
4383 
4384 static int
4385 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
4386 			uint8_t on)
4387 {
4388 	uint32_t vector;
4389 	uint32_t uta_idx;
4390 	uint32_t reg_val;
4391 	uint32_t uta_shift;
4392 	uint32_t rc;
4393 	const uint32_t ixgbe_uta_idx_mask = 0x7F;
4394 	const uint32_t ixgbe_uta_bit_shift = 5;
4395 	const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
4396 	const uint32_t bit1 = 0x1;
4397 
4398 	struct ixgbe_hw *hw =
4399 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4400 	struct ixgbe_uta_info *uta_info =
4401 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
4402 
4403 	/* The UTA table only exists on 82599 hardware and newer */
4404 	if (hw->mac.type < ixgbe_mac_82599EB)
4405 		return -ENOTSUP;
4406 
4407 	vector = ixgbe_uta_vector(hw, mac_addr);
4408 	uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
4409 	uta_shift = vector & ixgbe_uta_bit_mask;
4410 
4411 	rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
4412 	if (rc == on)
4413 		return 0;
4414 
4415 	reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
4416 	if (on) {
4417 		uta_info->uta_in_use++;
4418 		reg_val |= (bit1 << uta_shift);
4419 		uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
4420 	} else {
4421 		uta_info->uta_in_use--;
4422 		reg_val &= ~(bit1 << uta_shift);
4423 		uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
4424 	}
4425 
4426 	IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
4427 
4428 	if (uta_info->uta_in_use > 0)
4429 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
4430 				IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
4431 	else
4432 		IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
4433 
4434 	return 0;
4435 }
4436 
4437 static int
4438 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
4439 {
4440 	int i;
4441 	struct ixgbe_hw *hw =
4442 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4443 	struct ixgbe_uta_info *uta_info =
4444 		IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
4445 
4446 	/* The UTA table only exists on 82599 hardware and newer */
4447 	if (hw->mac.type < ixgbe_mac_82599EB)
4448 		return -ENOTSUP;
4449 
4450 	if (on) {
4451 		for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
4452 			uta_info->uta_shadow[i] = ~0;
4453 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
4454 		}
4455 	} else {
4456 		for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
4457 			uta_info->uta_shadow[i] = 0;
4458 			IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
4459 		}
4460 	}
4461 	return 0;
4462 
4463 }
4464 
4465 uint32_t
4466 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
4467 {
4468 	uint32_t new_val = orig_val;
4469 
4470 	if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG)
4471 		new_val |= IXGBE_VMOLR_AUPE;
4472 	if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC)
4473 		new_val |= IXGBE_VMOLR_ROMPE;
4474 	if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
4475 		new_val |= IXGBE_VMOLR_ROPE;
4476 	if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
4477 		new_val |= IXGBE_VMOLR_BAM;
4478 	if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
4479 		new_val |= IXGBE_VMOLR_MPE;
4480 
4481 	return new_val;
4482 }
4483 
4484 static int
4485 ixgbe_set_pool_rx_mode(struct rte_eth_dev *dev, uint16_t pool,
4486 			       uint16_t rx_mask, uint8_t on)
4487 {
4488 	int val = 0;
4489 
4490 	struct ixgbe_hw *hw =
4491 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4492 	uint32_t vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
4493 
4494 	if (hw->mac.type == ixgbe_mac_82598EB) {
4495 		PMD_INIT_LOG(ERR, "setting VF receive mode set should be done"
4496 			     " on 82599 hardware and newer");
4497 		return -ENOTSUP;
4498 	}
4499 	if (ixgbe_vmdq_mode_check(hw) < 0)
4500 		return -ENOTSUP;
4501 
4502 	val = ixgbe_convert_vm_rx_mask_to_val(rx_mask, val);
4503 
4504 	if (on)
4505 		vmolr |= val;
4506 	else
4507 		vmolr &= ~val;
4508 
4509 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
4510 
4511 	return 0;
4512 }
4513 
4514 static int
4515 ixgbe_set_pool_rx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on)
4516 {
4517 	uint32_t reg, addr;
4518 	uint32_t val;
4519 	const uint8_t bit1 = 0x1;
4520 
4521 	struct ixgbe_hw *hw =
4522 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4523 
4524 	if (ixgbe_vmdq_mode_check(hw) < 0)
4525 		return -ENOTSUP;
4526 
4527 	if (pool >= ETH_64_POOLS)
4528 		return -EINVAL;
4529 
4530 	/* for pool >= 32, set bit in PFVFRE[1], otherwise PFVFRE[0] */
4531 	if (pool >= 32) {
4532 		addr = IXGBE_VFRE(1);
4533 		val = bit1 << (pool - 32);
4534 	} else {
4535 		addr = IXGBE_VFRE(0);
4536 		val = bit1 << pool;
4537 	}
4538 
4539 	reg = IXGBE_READ_REG(hw, addr);
4540 
4541 	if (on)
4542 		reg |= val;
4543 	else
4544 		reg &= ~val;
4545 
4546 	IXGBE_WRITE_REG(hw, addr, reg);
4547 
4548 	return 0;
4549 }
4550 
4551 static int
4552 ixgbe_set_pool_tx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on)
4553 {
4554 	uint32_t reg, addr;
4555 	uint32_t val;
4556 	const uint8_t bit1 = 0x1;
4557 
4558 	struct ixgbe_hw *hw =
4559 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4560 
4561 	if (ixgbe_vmdq_mode_check(hw) < 0)
4562 		return -ENOTSUP;
4563 
4564 	if (pool >= ETH_64_POOLS)
4565 		return -EINVAL;
4566 
4567 	/* for pool >= 32, set bit in PFVFTE[1], otherwise PFVFTE[0] */
4568 	if (pool >= 32) {
4569 		addr = IXGBE_VFTE(1);
4570 		val = bit1 << (pool - 32);
4571 	} else {
4572 		addr = IXGBE_VFTE(0);
4573 		val = bit1 << pool;
4574 	}
4575 
4576 	reg = IXGBE_READ_REG(hw, addr);
4577 
4578 	if (on)
4579 		reg |= val;
4580 	else
4581 		reg &= ~val;
4582 
4583 	IXGBE_WRITE_REG(hw, addr, reg);
4584 
4585 	return 0;
4586 }
4587 
4588 static int
4589 ixgbe_set_pool_vlan_filter(struct rte_eth_dev *dev, uint16_t vlan,
4590 			uint64_t pool_mask, uint8_t vlan_on)
4591 {
4592 	int ret = 0;
4593 	uint16_t pool_idx;
4594 	struct ixgbe_hw *hw =
4595 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4596 
4597 	if (ixgbe_vmdq_mode_check(hw) < 0)
4598 		return -ENOTSUP;
4599 	for (pool_idx = 0; pool_idx < ETH_64_POOLS; pool_idx++) {
4600 		if (pool_mask & ((uint64_t)(1ULL << pool_idx))) {
4601 			ret = hw->mac.ops.set_vfta(hw, vlan, pool_idx,
4602 						   vlan_on, false);
4603 			if (ret < 0)
4604 				return ret;
4605 		}
4606 	}
4607 
4608 	return ret;
4609 }
4610 
4611 #define IXGBE_MRCTL_VPME  0x01 /* Virtual Pool Mirroring. */
4612 #define IXGBE_MRCTL_UPME  0x02 /* Uplink Port Mirroring. */
4613 #define IXGBE_MRCTL_DPME  0x04 /* Downlink Port Mirroring. */
4614 #define IXGBE_MRCTL_VLME  0x08 /* VLAN Mirroring. */
4615 #define IXGBE_INVALID_MIRROR_TYPE(mirror_type) \
4616 	((mirror_type) & ~(uint8_t)(ETH_MIRROR_VIRTUAL_POOL_UP | \
4617 	ETH_MIRROR_UPLINK_PORT | ETH_MIRROR_DOWNLINK_PORT | ETH_MIRROR_VLAN))
4618 
4619 static int
4620 ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
4621 			struct rte_eth_mirror_conf *mirror_conf,
4622 			uint8_t rule_id, uint8_t on)
4623 {
4624 	uint32_t mr_ctl, vlvf;
4625 	uint32_t mp_lsb = 0;
4626 	uint32_t mv_msb = 0;
4627 	uint32_t mv_lsb = 0;
4628 	uint32_t mp_msb = 0;
4629 	uint8_t i = 0;
4630 	int reg_index = 0;
4631 	uint64_t vlan_mask = 0;
4632 
4633 	const uint8_t pool_mask_offset = 32;
4634 	const uint8_t vlan_mask_offset = 32;
4635 	const uint8_t dst_pool_offset = 8;
4636 	const uint8_t rule_mr_offset  = 4;
4637 	const uint8_t mirror_rule_mask = 0x0F;
4638 
4639 	struct ixgbe_mirror_info *mr_info =
4640 			(IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
4641 	struct ixgbe_hw *hw =
4642 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4643 	uint8_t mirror_type = 0;
4644 
4645 	if (ixgbe_vmdq_mode_check(hw) < 0)
4646 		return -ENOTSUP;
4647 
4648 	if (rule_id >= IXGBE_MAX_MIRROR_RULES)
4649 		return -EINVAL;
4650 
4651 	if (IXGBE_INVALID_MIRROR_TYPE(mirror_conf->rule_type)) {
4652 		PMD_DRV_LOG(ERR, "unsupported mirror type 0x%x.",
4653 			mirror_conf->rule_type);
4654 		return -EINVAL;
4655 	}
4656 
4657 	if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
4658 		mirror_type |= IXGBE_MRCTL_VLME;
4659 		/* Check if vlan id is valid and find conresponding VLAN ID index in VLVF */
4660 		for (i = 0; i < IXGBE_VLVF_ENTRIES; i++) {
4661 			if (mirror_conf->vlan.vlan_mask & (1ULL << i)) {
4662 				/* search vlan id related pool vlan filter index */
4663 				reg_index = ixgbe_find_vlvf_slot(hw,
4664 						 mirror_conf->vlan.vlan_id[i],
4665 						 false);
4666 				if (reg_index < 0)
4667 					return -EINVAL;
4668 				vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(reg_index));
4669 				if ((vlvf & IXGBE_VLVF_VIEN) &&
4670 				    ((vlvf & IXGBE_VLVF_VLANID_MASK) ==
4671 				      mirror_conf->vlan.vlan_id[i]))
4672 					vlan_mask |= (1ULL << reg_index);
4673 				else
4674 					return -EINVAL;
4675 			}
4676 		}
4677 
4678 		if (on) {
4679 			mv_lsb = vlan_mask & 0xFFFFFFFF;
4680 			mv_msb = vlan_mask >> vlan_mask_offset;
4681 
4682 			mr_info->mr_conf[rule_id].vlan.vlan_mask =
4683 						mirror_conf->vlan.vlan_mask;
4684 			for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++) {
4685 				if (mirror_conf->vlan.vlan_mask & (1ULL << i))
4686 					mr_info->mr_conf[rule_id].vlan.vlan_id[i] =
4687 						mirror_conf->vlan.vlan_id[i];
4688 			}
4689 		} else {
4690 			mv_lsb = 0;
4691 			mv_msb = 0;
4692 			mr_info->mr_conf[rule_id].vlan.vlan_mask = 0;
4693 			for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++)
4694 				mr_info->mr_conf[rule_id].vlan.vlan_id[i] = 0;
4695 		}
4696 	}
4697 
4698 	/*
4699 	 * if enable pool mirror, write related pool mask register,if disable
4700 	 * pool mirror, clear PFMRVM register
4701 	 */
4702 	if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
4703 		mirror_type |= IXGBE_MRCTL_VPME;
4704 		if (on) {
4705 			mp_lsb = mirror_conf->pool_mask & 0xFFFFFFFF;
4706 			mp_msb = mirror_conf->pool_mask >> pool_mask_offset;
4707 			mr_info->mr_conf[rule_id].pool_mask =
4708 					mirror_conf->pool_mask;
4709 
4710 		} else {
4711 			mp_lsb = 0;
4712 			mp_msb = 0;
4713 			mr_info->mr_conf[rule_id].pool_mask = 0;
4714 		}
4715 	}
4716 	if (mirror_conf->rule_type & ETH_MIRROR_UPLINK_PORT)
4717 		mirror_type |= IXGBE_MRCTL_UPME;
4718 	if (mirror_conf->rule_type & ETH_MIRROR_DOWNLINK_PORT)
4719 		mirror_type |= IXGBE_MRCTL_DPME;
4720 
4721 	/* read  mirror control register and recalculate it */
4722 	mr_ctl = IXGBE_READ_REG(hw, IXGBE_MRCTL(rule_id));
4723 
4724 	if (on) {
4725 		mr_ctl |= mirror_type;
4726 		mr_ctl &= mirror_rule_mask;
4727 		mr_ctl |= mirror_conf->dst_pool << dst_pool_offset;
4728 	} else
4729 		mr_ctl &= ~(mirror_conf->rule_type & mirror_rule_mask);
4730 
4731 	mr_info->mr_conf[rule_id].rule_type = mirror_conf->rule_type;
4732 	mr_info->mr_conf[rule_id].dst_pool = mirror_conf->dst_pool;
4733 
4734 	/* write mirrror control  register */
4735 	IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
4736 
4737 	/* write pool mirrror control  register */
4738 	if (mirror_conf->rule_type == ETH_MIRROR_VIRTUAL_POOL_UP) {
4739 		IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), mp_lsb);
4740 		IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset),
4741 				mp_msb);
4742 	}
4743 	/* write VLAN mirrror control  register */
4744 	if (mirror_conf->rule_type == ETH_MIRROR_VLAN) {
4745 		IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), mv_lsb);
4746 		IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset),
4747 				mv_msb);
4748 	}
4749 
4750 	return 0;
4751 }
4752 
4753 static int
4754 ixgbe_mirror_rule_reset(struct rte_eth_dev *dev, uint8_t rule_id)
4755 {
4756 	int mr_ctl = 0;
4757 	uint32_t lsb_val = 0;
4758 	uint32_t msb_val = 0;
4759 	const uint8_t rule_mr_offset = 4;
4760 
4761 	struct ixgbe_hw *hw =
4762 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4763 	struct ixgbe_mirror_info *mr_info =
4764 		(IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
4765 
4766 	if (ixgbe_vmdq_mode_check(hw) < 0)
4767 		return -ENOTSUP;
4768 
4769 	memset(&mr_info->mr_conf[rule_id], 0,
4770 		sizeof(struct rte_eth_mirror_conf));
4771 
4772 	/* clear PFVMCTL register */
4773 	IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
4774 
4775 	/* clear pool mask register */
4776 	IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), lsb_val);
4777 	IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset), msb_val);
4778 
4779 	/* clear vlan mask register */
4780 	IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), lsb_val);
4781 	IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset), msb_val);
4782 
4783 	return 0;
4784 }
4785 
4786 static int
4787 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
4788 {
4789 	uint32_t mask;
4790 	struct ixgbe_hw *hw =
4791 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4792 
4793 	mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
4794 	mask |= (1 << IXGBE_MISC_VEC_ID);
4795 	RTE_SET_USED(queue_id);
4796 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
4797 
4798 	rte_intr_enable(&dev->pci_dev->intr_handle);
4799 
4800 	return 0;
4801 }
4802 
4803 static int
4804 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
4805 {
4806 	uint32_t mask;
4807 	struct ixgbe_hw *hw =
4808 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4809 
4810 	mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
4811 	mask &= ~(1 << IXGBE_MISC_VEC_ID);
4812 	RTE_SET_USED(queue_id);
4813 	IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
4814 
4815 	return 0;
4816 }
4817 
4818 static int
4819 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
4820 {
4821 	uint32_t mask;
4822 	struct ixgbe_hw *hw =
4823 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4824 	struct ixgbe_interrupt *intr =
4825 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4826 
4827 	if (queue_id < 16) {
4828 		ixgbe_disable_intr(hw);
4829 		intr->mask |= (1 << queue_id);
4830 		ixgbe_enable_intr(dev);
4831 	} else if (queue_id < 32) {
4832 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
4833 		mask &= (1 << queue_id);
4834 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
4835 	} else if (queue_id < 64) {
4836 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
4837 		mask &= (1 << (queue_id - 32));
4838 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
4839 	}
4840 	rte_intr_enable(&dev->pci_dev->intr_handle);
4841 
4842 	return 0;
4843 }
4844 
4845 static int
4846 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
4847 {
4848 	uint32_t mask;
4849 	struct ixgbe_hw *hw =
4850 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4851 	struct ixgbe_interrupt *intr =
4852 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4853 
4854 	if (queue_id < 16) {
4855 		ixgbe_disable_intr(hw);
4856 		intr->mask &= ~(1 << queue_id);
4857 		ixgbe_enable_intr(dev);
4858 	} else if (queue_id < 32) {
4859 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
4860 		mask &= ~(1 << queue_id);
4861 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
4862 	} else if (queue_id < 64) {
4863 		mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
4864 		mask &= ~(1 << (queue_id - 32));
4865 		IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
4866 	}
4867 
4868 	return 0;
4869 }
4870 
4871 static void
4872 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
4873 		     uint8_t queue, uint8_t msix_vector)
4874 {
4875 	uint32_t tmp, idx;
4876 
4877 	if (direction == -1) {
4878 		/* other causes */
4879 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
4880 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
4881 		tmp &= ~0xFF;
4882 		tmp |= msix_vector;
4883 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
4884 	} else {
4885 		/* rx or tx cause */
4886 		msix_vector |= IXGBE_IVAR_ALLOC_VAL;
4887 		idx = ((16 * (queue & 1)) + (8 * direction));
4888 		tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
4889 		tmp &= ~(0xFF << idx);
4890 		tmp |= (msix_vector << idx);
4891 		IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
4892 	}
4893 }
4894 
4895 /**
4896  * set the IVAR registers, mapping interrupt causes to vectors
4897  * @param hw
4898  *  pointer to ixgbe_hw struct
4899  * @direction
4900  *  0 for Rx, 1 for Tx, -1 for other causes
4901  * @queue
4902  *  queue to map the corresponding interrupt to
4903  * @msix_vector
4904  *  the vector to map to the corresponding queue
4905  */
4906 static void
4907 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
4908 		   uint8_t queue, uint8_t msix_vector)
4909 {
4910 	uint32_t tmp, idx;
4911 
4912 	msix_vector |= IXGBE_IVAR_ALLOC_VAL;
4913 	if (hw->mac.type == ixgbe_mac_82598EB) {
4914 		if (direction == -1)
4915 			direction = 0;
4916 		idx = (((direction * 64) + queue) >> 2) & 0x1F;
4917 		tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
4918 		tmp &= ~(0xFF << (8 * (queue & 0x3)));
4919 		tmp |= (msix_vector << (8 * (queue & 0x3)));
4920 		IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
4921 	} else if ((hw->mac.type == ixgbe_mac_82599EB) ||
4922 			(hw->mac.type == ixgbe_mac_X540)) {
4923 		if (direction == -1) {
4924 			/* other causes */
4925 			idx = ((queue & 1) * 8);
4926 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
4927 			tmp &= ~(0xFF << idx);
4928 			tmp |= (msix_vector << idx);
4929 			IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
4930 		} else {
4931 			/* rx or tx causes */
4932 			idx = ((16 * (queue & 1)) + (8 * direction));
4933 			tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
4934 			tmp &= ~(0xFF << idx);
4935 			tmp |= (msix_vector << idx);
4936 			IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
4937 		}
4938 	}
4939 }
4940 
4941 static void
4942 ixgbevf_configure_msix(struct rte_eth_dev *dev)
4943 {
4944 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
4945 	struct ixgbe_hw *hw =
4946 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4947 	uint32_t q_idx;
4948 	uint32_t vector_idx = IXGBE_MISC_VEC_ID;
4949 
4950 	/* Configure VF other cause ivar */
4951 	ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
4952 
4953 	/* won't configure msix register if no mapping is done
4954 	 * between intr vector and event fd.
4955 	 */
4956 	if (!rte_intr_dp_is_en(intr_handle))
4957 		return;
4958 
4959 	/* Configure all RX queues of VF */
4960 	for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
4961 		/* Force all queue use vector 0,
4962 		 * as IXGBE_VF_MAXMSIVECOTR = 1
4963 		 */
4964 		ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
4965 		intr_handle->intr_vec[q_idx] = vector_idx;
4966 	}
4967 }
4968 
4969 /**
4970  * Sets up the hardware to properly generate MSI-X interrupts
4971  * @hw
4972  *  board private structure
4973  */
4974 static void
4975 ixgbe_configure_msix(struct rte_eth_dev *dev)
4976 {
4977 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
4978 	struct ixgbe_hw *hw =
4979 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4980 	uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
4981 	uint32_t vec = IXGBE_MISC_VEC_ID;
4982 	uint32_t mask;
4983 	uint32_t gpie;
4984 
4985 	/* won't configure msix register if no mapping is done
4986 	 * between intr vector and event fd
4987 	 */
4988 	if (!rte_intr_dp_is_en(intr_handle))
4989 		return;
4990 
4991 	if (rte_intr_allow_others(intr_handle))
4992 		vec = base = IXGBE_RX_VEC_START;
4993 
4994 	/* setup GPIE for MSI-x mode */
4995 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
4996 	gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
4997 		IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
4998 	/* auto clearing and auto setting corresponding bits in EIMS
4999 	 * when MSI-X interrupt is triggered
5000 	 */
5001 	if (hw->mac.type == ixgbe_mac_82598EB) {
5002 		IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5003 	} else {
5004 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5005 		IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5006 	}
5007 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5008 
5009 	/* Populate the IVAR table and set the ITR values to the
5010 	 * corresponding register.
5011 	 */
5012 	for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
5013 	     queue_id++) {
5014 		/* by default, 1:1 mapping */
5015 		ixgbe_set_ivar_map(hw, 0, queue_id, vec);
5016 		intr_handle->intr_vec[queue_id] = vec;
5017 		if (vec < base + intr_handle->nb_efd - 1)
5018 			vec++;
5019 	}
5020 
5021 	switch (hw->mac.type) {
5022 	case ixgbe_mac_82598EB:
5023 		ixgbe_set_ivar_map(hw, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
5024 				   IXGBE_MISC_VEC_ID);
5025 		break;
5026 	case ixgbe_mac_82599EB:
5027 	case ixgbe_mac_X540:
5028 		ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
5029 		break;
5030 	default:
5031 		break;
5032 	}
5033 	IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
5034 			IXGBE_MIN_INTER_INTERRUPT_INTERVAL_DEFAULT & 0xFFF);
5035 
5036 	/* set up to autoclear timer, and the vectors */
5037 	mask = IXGBE_EIMS_ENABLE_MASK;
5038 	mask &= ~(IXGBE_EIMS_OTHER |
5039 		  IXGBE_EIMS_MAILBOX |
5040 		  IXGBE_EIMS_LSC);
5041 
5042 	IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
5043 }
5044 
5045 static int ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
5046 	uint16_t queue_idx, uint16_t tx_rate)
5047 {
5048 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5049 	uint32_t rf_dec, rf_int;
5050 	uint32_t bcnrc_val;
5051 	uint16_t link_speed = dev->data->dev_link.link_speed;
5052 
5053 	if (queue_idx >= hw->mac.max_tx_queues)
5054 		return -EINVAL;
5055 
5056 	if (tx_rate != 0) {
5057 		/* Calculate the rate factor values to set */
5058 		rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
5059 		rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
5060 		rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
5061 
5062 		bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
5063 		bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
5064 				IXGBE_RTTBCNRC_RF_INT_MASK_M);
5065 		bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
5066 	} else {
5067 		bcnrc_val = 0;
5068 	}
5069 
5070 	/*
5071 	 * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
5072 	 * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
5073 	 * set as 0x4.
5074 	 */
5075 	if ((dev->data->dev_conf.rxmode.jumbo_frame == 1) &&
5076 		(dev->data->dev_conf.rxmode.max_rx_pkt_len >=
5077 				IXGBE_MAX_JUMBO_FRAME_SIZE))
5078 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
5079 			IXGBE_MMW_SIZE_JUMBO_FRAME);
5080 	else
5081 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
5082 			IXGBE_MMW_SIZE_DEFAULT);
5083 
5084 	/* Set RTTBCNRC of queue X */
5085 	IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
5086 	IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
5087 	IXGBE_WRITE_FLUSH(hw);
5088 
5089 	return 0;
5090 }
5091 
5092 static int ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
5093 	uint16_t tx_rate, uint64_t q_msk)
5094 {
5095 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5096 	struct ixgbe_vf_info *vfinfo =
5097 		*(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
5098 	uint8_t  nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
5099 	uint32_t queue_stride =
5100 		IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
5101 	uint32_t queue_idx = vf * queue_stride, idx = 0, vf_idx;
5102 	uint32_t queue_end = queue_idx + nb_q_per_pool - 1;
5103 	uint16_t total_rate = 0;
5104 
5105 	if (queue_end >= hw->mac.max_tx_queues)
5106 		return -EINVAL;
5107 
5108 	if (vfinfo != NULL) {
5109 		for (vf_idx = 0; vf_idx < dev->pci_dev->max_vfs; vf_idx++) {
5110 			if (vf_idx == vf)
5111 				continue;
5112 			for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
5113 				idx++)
5114 				total_rate += vfinfo[vf_idx].tx_rate[idx];
5115 		}
5116 	} else
5117 		return -EINVAL;
5118 
5119 	/* Store tx_rate for this vf. */
5120 	for (idx = 0; idx < nb_q_per_pool; idx++) {
5121 		if (((uint64_t)0x1 << idx) & q_msk) {
5122 			if (vfinfo[vf].tx_rate[idx] != tx_rate)
5123 				vfinfo[vf].tx_rate[idx] = tx_rate;
5124 			total_rate += tx_rate;
5125 		}
5126 	}
5127 
5128 	if (total_rate > dev->data->dev_link.link_speed) {
5129 		/*
5130 		 * Reset stored TX rate of the VF if it causes exceed
5131 		 * link speed.
5132 		 */
5133 		memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
5134 		return -EINVAL;
5135 	}
5136 
5137 	/* Set RTTBCNRC of each queue/pool for vf X  */
5138 	for (; queue_idx <= queue_end; queue_idx++) {
5139 		if (0x1 & q_msk)
5140 			ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
5141 		q_msk = q_msk >> 1;
5142 	}
5143 
5144 	return 0;
5145 }
5146 
5147 static void
5148 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
5149 		     __attribute__((unused)) uint32_t index,
5150 		     __attribute__((unused)) uint32_t pool)
5151 {
5152 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5153 	int diag;
5154 
5155 	/*
5156 	 * On a 82599 VF, adding again the same MAC addr is not an idempotent
5157 	 * operation. Trap this case to avoid exhausting the [very limited]
5158 	 * set of PF resources used to store VF MAC addresses.
5159 	 */
5160 	if (memcmp(hw->mac.perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
5161 		return;
5162 	diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
5163 	if (diag == 0)
5164 		return;
5165 	PMD_DRV_LOG(ERR, "Unable to add MAC address - diag=%d", diag);
5166 }
5167 
5168 static void
5169 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
5170 {
5171 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5172 	struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
5173 	struct ether_addr *mac_addr;
5174 	uint32_t i;
5175 	int diag;
5176 
5177 	/*
5178 	 * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
5179 	 * not support the deletion of a given MAC address.
5180 	 * Instead, it imposes to delete all MAC addresses, then to add again
5181 	 * all MAC addresses with the exception of the one to be deleted.
5182 	 */
5183 	(void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
5184 
5185 	/*
5186 	 * Add again all MAC addresses, with the exception of the deleted one
5187 	 * and of the permanent MAC address.
5188 	 */
5189 	for (i = 0, mac_addr = dev->data->mac_addrs;
5190 	     i < hw->mac.num_rar_entries; i++, mac_addr++) {
5191 		/* Skip the deleted MAC address */
5192 		if (i == index)
5193 			continue;
5194 		/* Skip NULL MAC addresses */
5195 		if (is_zero_ether_addr(mac_addr))
5196 			continue;
5197 		/* Skip the permanent MAC address */
5198 		if (memcmp(perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
5199 			continue;
5200 		diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
5201 		if (diag != 0)
5202 			PMD_DRV_LOG(ERR,
5203 				    "Adding again MAC address "
5204 				    "%02x:%02x:%02x:%02x:%02x:%02x failed "
5205 				    "diag=%d",
5206 				    mac_addr->addr_bytes[0],
5207 				    mac_addr->addr_bytes[1],
5208 				    mac_addr->addr_bytes[2],
5209 				    mac_addr->addr_bytes[3],
5210 				    mac_addr->addr_bytes[4],
5211 				    mac_addr->addr_bytes[5],
5212 				    diag);
5213 	}
5214 }
5215 
5216 static void
5217 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
5218 {
5219 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5220 
5221 	hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
5222 }
5223 
5224 #define MAC_TYPE_FILTER_SUP(type)    do {\
5225 	if ((type) != ixgbe_mac_82599EB && (type) != ixgbe_mac_X540 &&\
5226 		(type) != ixgbe_mac_X550 && (type) != ixgbe_mac_X550EM_x &&\
5227 		(type) != ixgbe_mac_X550EM_a)\
5228 		return -ENOTSUP;\
5229 } while (0)
5230 
5231 static int
5232 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
5233 			struct rte_eth_syn_filter *filter,
5234 			bool add)
5235 {
5236 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5237 	uint32_t synqf;
5238 
5239 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
5240 		return -EINVAL;
5241 
5242 	synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
5243 
5244 	if (add) {
5245 		if (synqf & IXGBE_SYN_FILTER_ENABLE)
5246 			return -EINVAL;
5247 		synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
5248 			IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
5249 
5250 		if (filter->hig_pri)
5251 			synqf |= IXGBE_SYN_FILTER_SYNQFP;
5252 		else
5253 			synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
5254 	} else {
5255 		if (!(synqf & IXGBE_SYN_FILTER_ENABLE))
5256 			return -ENOENT;
5257 		synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
5258 	}
5259 	IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
5260 	IXGBE_WRITE_FLUSH(hw);
5261 	return 0;
5262 }
5263 
5264 static int
5265 ixgbe_syn_filter_get(struct rte_eth_dev *dev,
5266 			struct rte_eth_syn_filter *filter)
5267 {
5268 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5269 	uint32_t synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
5270 
5271 	if (synqf & IXGBE_SYN_FILTER_ENABLE) {
5272 		filter->hig_pri = (synqf & IXGBE_SYN_FILTER_SYNQFP) ? 1 : 0;
5273 		filter->queue = (uint16_t)((synqf & IXGBE_SYN_FILTER_QUEUE) >> 1);
5274 		return 0;
5275 	}
5276 	return -ENOENT;
5277 }
5278 
5279 static int
5280 ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
5281 			enum rte_filter_op filter_op,
5282 			void *arg)
5283 {
5284 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5285 	int ret;
5286 
5287 	MAC_TYPE_FILTER_SUP(hw->mac.type);
5288 
5289 	if (filter_op == RTE_ETH_FILTER_NOP)
5290 		return 0;
5291 
5292 	if (arg == NULL) {
5293 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
5294 			    filter_op);
5295 		return -EINVAL;
5296 	}
5297 
5298 	switch (filter_op) {
5299 	case RTE_ETH_FILTER_ADD:
5300 		ret = ixgbe_syn_filter_set(dev,
5301 				(struct rte_eth_syn_filter *)arg,
5302 				TRUE);
5303 		break;
5304 	case RTE_ETH_FILTER_DELETE:
5305 		ret = ixgbe_syn_filter_set(dev,
5306 				(struct rte_eth_syn_filter *)arg,
5307 				FALSE);
5308 		break;
5309 	case RTE_ETH_FILTER_GET:
5310 		ret = ixgbe_syn_filter_get(dev,
5311 				(struct rte_eth_syn_filter *)arg);
5312 		break;
5313 	default:
5314 		PMD_DRV_LOG(ERR, "unsupported operation %u\n", filter_op);
5315 		ret = -EINVAL;
5316 		break;
5317 	}
5318 
5319 	return ret;
5320 }
5321 
5322 
5323 static inline enum ixgbe_5tuple_protocol
5324 convert_protocol_type(uint8_t protocol_value)
5325 {
5326 	if (protocol_value == IPPROTO_TCP)
5327 		return IXGBE_FILTER_PROTOCOL_TCP;
5328 	else if (protocol_value == IPPROTO_UDP)
5329 		return IXGBE_FILTER_PROTOCOL_UDP;
5330 	else if (protocol_value == IPPROTO_SCTP)
5331 		return IXGBE_FILTER_PROTOCOL_SCTP;
5332 	else
5333 		return IXGBE_FILTER_PROTOCOL_NONE;
5334 }
5335 
5336 /*
5337  * add a 5tuple filter
5338  *
5339  * @param
5340  * dev: Pointer to struct rte_eth_dev.
5341  * index: the index the filter allocates.
5342  * filter: ponter to the filter that will be added.
5343  * rx_queue: the queue id the filter assigned to.
5344  *
5345  * @return
5346  *    - On success, zero.
5347  *    - On failure, a negative value.
5348  */
5349 static int
5350 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
5351 			struct ixgbe_5tuple_filter *filter)
5352 {
5353 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5354 	struct ixgbe_filter_info *filter_info =
5355 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5356 	int i, idx, shift;
5357 	uint32_t ftqf, sdpqf;
5358 	uint32_t l34timir = 0;
5359 	uint8_t mask = 0xff;
5360 
5361 	/*
5362 	 * look for an unused 5tuple filter index,
5363 	 * and insert the filter to list.
5364 	 */
5365 	for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
5366 		idx = i / (sizeof(uint32_t) * NBBY);
5367 		shift = i % (sizeof(uint32_t) * NBBY);
5368 		if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
5369 			filter_info->fivetuple_mask[idx] |= 1 << shift;
5370 			filter->index = i;
5371 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
5372 					  filter,
5373 					  entries);
5374 			break;
5375 		}
5376 	}
5377 	if (i >= IXGBE_MAX_FTQF_FILTERS) {
5378 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
5379 		return -ENOSYS;
5380 	}
5381 
5382 	sdpqf = (uint32_t)(filter->filter_info.dst_port <<
5383 				IXGBE_SDPQF_DSTPORT_SHIFT);
5384 	sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
5385 
5386 	ftqf = (uint32_t)(filter->filter_info.proto &
5387 		IXGBE_FTQF_PROTOCOL_MASK);
5388 	ftqf |= (uint32_t)((filter->filter_info.priority &
5389 		IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
5390 	if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
5391 		mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
5392 	if (filter->filter_info.dst_ip_mask == 0)
5393 		mask &= IXGBE_FTQF_DEST_ADDR_MASK;
5394 	if (filter->filter_info.src_port_mask == 0)
5395 		mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
5396 	if (filter->filter_info.dst_port_mask == 0)
5397 		mask &= IXGBE_FTQF_DEST_PORT_MASK;
5398 	if (filter->filter_info.proto_mask == 0)
5399 		mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
5400 	ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
5401 	ftqf |= IXGBE_FTQF_POOL_MASK_EN;
5402 	ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
5403 
5404 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
5405 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
5406 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
5407 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
5408 
5409 	l34timir |= IXGBE_L34T_IMIR_RESERVE;
5410 	l34timir |= (uint32_t)(filter->queue <<
5411 				IXGBE_L34T_IMIR_QUEUE_SHIFT);
5412 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
5413 	return 0;
5414 }
5415 
5416 /*
5417  * remove a 5tuple filter
5418  *
5419  * @param
5420  * dev: Pointer to struct rte_eth_dev.
5421  * filter: the pointer of the filter will be removed.
5422  */
5423 static void
5424 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
5425 			struct ixgbe_5tuple_filter *filter)
5426 {
5427 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5428 	struct ixgbe_filter_info *filter_info =
5429 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5430 	uint16_t index = filter->index;
5431 
5432 	filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
5433 				~(1 << (index % (sizeof(uint32_t) * NBBY)));
5434 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
5435 	rte_free(filter);
5436 
5437 	IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
5438 	IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
5439 	IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
5440 	IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
5441 	IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
5442 }
5443 
5444 static int
5445 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
5446 {
5447 	struct ixgbe_hw *hw;
5448 	uint32_t max_frame = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
5449 
5450 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5451 
5452 	if ((mtu < ETHER_MIN_MTU) || (max_frame > ETHER_MAX_JUMBO_FRAME_LEN))
5453 		return -EINVAL;
5454 
5455 	/* refuse mtu that requires the support of scattered packets when this
5456 	 * feature has not been enabled before.
5457 	 */
5458 	if (!dev->data->scattered_rx &&
5459 	    (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
5460 	     dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM))
5461 		return -EINVAL;
5462 
5463 	/*
5464 	 * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
5465 	 * request of the version 2.0 of the mailbox API.
5466 	 * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
5467 	 * of the mailbox API.
5468 	 * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
5469 	 * prior to 3.11.33 which contains the following change:
5470 	 * "ixgbe: Enable jumbo frames support w/ SR-IOV"
5471 	 */
5472 	ixgbevf_rlpml_set_vf(hw, max_frame);
5473 
5474 	/* update max frame size */
5475 	dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame;
5476 	return 0;
5477 }
5478 
5479 #define MAC_TYPE_FILTER_SUP_EXT(type)    do {\
5480 	if ((type) != ixgbe_mac_82599EB && (type) != ixgbe_mac_X540)\
5481 		return -ENOTSUP;\
5482 } while (0)
5483 
5484 static inline struct ixgbe_5tuple_filter *
5485 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
5486 			struct ixgbe_5tuple_filter_info *key)
5487 {
5488 	struct ixgbe_5tuple_filter *it;
5489 
5490 	TAILQ_FOREACH(it, filter_list, entries) {
5491 		if (memcmp(key, &it->filter_info,
5492 			sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
5493 			return it;
5494 		}
5495 	}
5496 	return NULL;
5497 }
5498 
5499 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
5500 static inline int
5501 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
5502 			struct ixgbe_5tuple_filter_info *filter_info)
5503 {
5504 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
5505 		filter->priority > IXGBE_5TUPLE_MAX_PRI ||
5506 		filter->priority < IXGBE_5TUPLE_MIN_PRI)
5507 		return -EINVAL;
5508 
5509 	switch (filter->dst_ip_mask) {
5510 	case UINT32_MAX:
5511 		filter_info->dst_ip_mask = 0;
5512 		filter_info->dst_ip = filter->dst_ip;
5513 		break;
5514 	case 0:
5515 		filter_info->dst_ip_mask = 1;
5516 		break;
5517 	default:
5518 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
5519 		return -EINVAL;
5520 	}
5521 
5522 	switch (filter->src_ip_mask) {
5523 	case UINT32_MAX:
5524 		filter_info->src_ip_mask = 0;
5525 		filter_info->src_ip = filter->src_ip;
5526 		break;
5527 	case 0:
5528 		filter_info->src_ip_mask = 1;
5529 		break;
5530 	default:
5531 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
5532 		return -EINVAL;
5533 	}
5534 
5535 	switch (filter->dst_port_mask) {
5536 	case UINT16_MAX:
5537 		filter_info->dst_port_mask = 0;
5538 		filter_info->dst_port = filter->dst_port;
5539 		break;
5540 	case 0:
5541 		filter_info->dst_port_mask = 1;
5542 		break;
5543 	default:
5544 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
5545 		return -EINVAL;
5546 	}
5547 
5548 	switch (filter->src_port_mask) {
5549 	case UINT16_MAX:
5550 		filter_info->src_port_mask = 0;
5551 		filter_info->src_port = filter->src_port;
5552 		break;
5553 	case 0:
5554 		filter_info->src_port_mask = 1;
5555 		break;
5556 	default:
5557 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
5558 		return -EINVAL;
5559 	}
5560 
5561 	switch (filter->proto_mask) {
5562 	case UINT8_MAX:
5563 		filter_info->proto_mask = 0;
5564 		filter_info->proto =
5565 			convert_protocol_type(filter->proto);
5566 		break;
5567 	case 0:
5568 		filter_info->proto_mask = 1;
5569 		break;
5570 	default:
5571 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
5572 		return -EINVAL;
5573 	}
5574 
5575 	filter_info->priority = (uint8_t)filter->priority;
5576 	return 0;
5577 }
5578 
5579 /*
5580  * add or delete a ntuple filter
5581  *
5582  * @param
5583  * dev: Pointer to struct rte_eth_dev.
5584  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
5585  * add: if true, add filter, if false, remove filter
5586  *
5587  * @return
5588  *    - On success, zero.
5589  *    - On failure, a negative value.
5590  */
5591 static int
5592 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
5593 			struct rte_eth_ntuple_filter *ntuple_filter,
5594 			bool add)
5595 {
5596 	struct ixgbe_filter_info *filter_info =
5597 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5598 	struct ixgbe_5tuple_filter_info filter_5tuple;
5599 	struct ixgbe_5tuple_filter *filter;
5600 	int ret;
5601 
5602 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
5603 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
5604 		return -EINVAL;
5605 	}
5606 
5607 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
5608 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
5609 	if (ret < 0)
5610 		return ret;
5611 
5612 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
5613 					 &filter_5tuple);
5614 	if (filter != NULL && add) {
5615 		PMD_DRV_LOG(ERR, "filter exists.");
5616 		return -EEXIST;
5617 	}
5618 	if (filter == NULL && !add) {
5619 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
5620 		return -ENOENT;
5621 	}
5622 
5623 	if (add) {
5624 		filter = rte_zmalloc("ixgbe_5tuple_filter",
5625 				sizeof(struct ixgbe_5tuple_filter), 0);
5626 		if (filter == NULL)
5627 			return -ENOMEM;
5628 		(void)rte_memcpy(&filter->filter_info,
5629 				 &filter_5tuple,
5630 				 sizeof(struct ixgbe_5tuple_filter_info));
5631 		filter->queue = ntuple_filter->queue;
5632 		ret = ixgbe_add_5tuple_filter(dev, filter);
5633 		if (ret < 0) {
5634 			rte_free(filter);
5635 			return ret;
5636 		}
5637 	} else
5638 		ixgbe_remove_5tuple_filter(dev, filter);
5639 
5640 	return 0;
5641 }
5642 
5643 /*
5644  * get a ntuple filter
5645  *
5646  * @param
5647  * dev: Pointer to struct rte_eth_dev.
5648  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
5649  *
5650  * @return
5651  *    - On success, zero.
5652  *    - On failure, a negative value.
5653  */
5654 static int
5655 ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
5656 			struct rte_eth_ntuple_filter *ntuple_filter)
5657 {
5658 	struct ixgbe_filter_info *filter_info =
5659 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5660 	struct ixgbe_5tuple_filter_info filter_5tuple;
5661 	struct ixgbe_5tuple_filter *filter;
5662 	int ret;
5663 
5664 	if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
5665 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
5666 		return -EINVAL;
5667 	}
5668 
5669 	memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
5670 	ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
5671 	if (ret < 0)
5672 		return ret;
5673 
5674 	filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
5675 					 &filter_5tuple);
5676 	if (filter == NULL) {
5677 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
5678 		return -ENOENT;
5679 	}
5680 	ntuple_filter->queue = filter->queue;
5681 	return 0;
5682 }
5683 
5684 /*
5685  * ixgbe_ntuple_filter_handle - Handle operations for ntuple filter.
5686  * @dev: pointer to rte_eth_dev structure
5687  * @filter_op:operation will be taken.
5688  * @arg: a pointer to specific structure corresponding to the filter_op
5689  *
5690  * @return
5691  *    - On success, zero.
5692  *    - On failure, a negative value.
5693  */
5694 static int
5695 ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
5696 				enum rte_filter_op filter_op,
5697 				void *arg)
5698 {
5699 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5700 	int ret;
5701 
5702 	MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
5703 
5704 	if (filter_op == RTE_ETH_FILTER_NOP)
5705 		return 0;
5706 
5707 	if (arg == NULL) {
5708 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
5709 			    filter_op);
5710 		return -EINVAL;
5711 	}
5712 
5713 	switch (filter_op) {
5714 	case RTE_ETH_FILTER_ADD:
5715 		ret = ixgbe_add_del_ntuple_filter(dev,
5716 			(struct rte_eth_ntuple_filter *)arg,
5717 			TRUE);
5718 		break;
5719 	case RTE_ETH_FILTER_DELETE:
5720 		ret = ixgbe_add_del_ntuple_filter(dev,
5721 			(struct rte_eth_ntuple_filter *)arg,
5722 			FALSE);
5723 		break;
5724 	case RTE_ETH_FILTER_GET:
5725 		ret = ixgbe_get_ntuple_filter(dev,
5726 			(struct rte_eth_ntuple_filter *)arg);
5727 		break;
5728 	default:
5729 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
5730 		ret = -EINVAL;
5731 		break;
5732 	}
5733 	return ret;
5734 }
5735 
5736 static inline int
5737 ixgbe_ethertype_filter_lookup(struct ixgbe_filter_info *filter_info,
5738 			uint16_t ethertype)
5739 {
5740 	int i;
5741 
5742 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
5743 		if (filter_info->ethertype_filters[i] == ethertype &&
5744 		    (filter_info->ethertype_mask & (1 << i)))
5745 			return i;
5746 	}
5747 	return -1;
5748 }
5749 
5750 static inline int
5751 ixgbe_ethertype_filter_insert(struct ixgbe_filter_info *filter_info,
5752 			uint16_t ethertype)
5753 {
5754 	int i;
5755 
5756 	for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
5757 		if (!(filter_info->ethertype_mask & (1 << i))) {
5758 			filter_info->ethertype_mask |= 1 << i;
5759 			filter_info->ethertype_filters[i] = ethertype;
5760 			return i;
5761 		}
5762 	}
5763 	return -1;
5764 }
5765 
5766 static inline int
5767 ixgbe_ethertype_filter_remove(struct ixgbe_filter_info *filter_info,
5768 			uint8_t idx)
5769 {
5770 	if (idx >= IXGBE_MAX_ETQF_FILTERS)
5771 		return -1;
5772 	filter_info->ethertype_mask &= ~(1 << idx);
5773 	filter_info->ethertype_filters[idx] = 0;
5774 	return idx;
5775 }
5776 
5777 static int
5778 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
5779 			struct rte_eth_ethertype_filter *filter,
5780 			bool add)
5781 {
5782 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5783 	struct ixgbe_filter_info *filter_info =
5784 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5785 	uint32_t etqf = 0;
5786 	uint32_t etqs = 0;
5787 	int ret;
5788 
5789 	if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
5790 		return -EINVAL;
5791 
5792 	if (filter->ether_type == ETHER_TYPE_IPv4 ||
5793 		filter->ether_type == ETHER_TYPE_IPv6) {
5794 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
5795 			" ethertype filter.", filter->ether_type);
5796 		return -EINVAL;
5797 	}
5798 
5799 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
5800 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
5801 		return -EINVAL;
5802 	}
5803 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
5804 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
5805 		return -EINVAL;
5806 	}
5807 
5808 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
5809 	if (ret >= 0 && add) {
5810 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
5811 			    filter->ether_type);
5812 		return -EEXIST;
5813 	}
5814 	if (ret < 0 && !add) {
5815 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
5816 			    filter->ether_type);
5817 		return -ENOENT;
5818 	}
5819 
5820 	if (add) {
5821 		ret = ixgbe_ethertype_filter_insert(filter_info,
5822 			filter->ether_type);
5823 		if (ret < 0) {
5824 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
5825 			return -ENOSYS;
5826 		}
5827 		etqf = IXGBE_ETQF_FILTER_EN;
5828 		etqf |= (uint32_t)filter->ether_type;
5829 		etqs |= (uint32_t)((filter->queue <<
5830 				    IXGBE_ETQS_RX_QUEUE_SHIFT) &
5831 				    IXGBE_ETQS_RX_QUEUE);
5832 		etqs |= IXGBE_ETQS_QUEUE_EN;
5833 	} else {
5834 		ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
5835 		if (ret < 0)
5836 			return -ENOSYS;
5837 	}
5838 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
5839 	IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
5840 	IXGBE_WRITE_FLUSH(hw);
5841 
5842 	return 0;
5843 }
5844 
5845 static int
5846 ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
5847 			struct rte_eth_ethertype_filter *filter)
5848 {
5849 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5850 	struct ixgbe_filter_info *filter_info =
5851 		IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5852 	uint32_t etqf, etqs;
5853 	int ret;
5854 
5855 	ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
5856 	if (ret < 0) {
5857 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
5858 			    filter->ether_type);
5859 		return -ENOENT;
5860 	}
5861 
5862 	etqf = IXGBE_READ_REG(hw, IXGBE_ETQF(ret));
5863 	if (etqf & IXGBE_ETQF_FILTER_EN) {
5864 		etqs = IXGBE_READ_REG(hw, IXGBE_ETQS(ret));
5865 		filter->ether_type = etqf & IXGBE_ETQF_ETHERTYPE;
5866 		filter->flags = 0;
5867 		filter->queue = (etqs & IXGBE_ETQS_RX_QUEUE) >>
5868 			       IXGBE_ETQS_RX_QUEUE_SHIFT;
5869 		return 0;
5870 	}
5871 	return -ENOENT;
5872 }
5873 
5874 /*
5875  * ixgbe_ethertype_filter_handle - Handle operations for ethertype filter.
5876  * @dev: pointer to rte_eth_dev structure
5877  * @filter_op:operation will be taken.
5878  * @arg: a pointer to specific structure corresponding to the filter_op
5879  */
5880 static int
5881 ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
5882 				enum rte_filter_op filter_op,
5883 				void *arg)
5884 {
5885 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5886 	int ret;
5887 
5888 	MAC_TYPE_FILTER_SUP(hw->mac.type);
5889 
5890 	if (filter_op == RTE_ETH_FILTER_NOP)
5891 		return 0;
5892 
5893 	if (arg == NULL) {
5894 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
5895 			    filter_op);
5896 		return -EINVAL;
5897 	}
5898 
5899 	switch (filter_op) {
5900 	case RTE_ETH_FILTER_ADD:
5901 		ret = ixgbe_add_del_ethertype_filter(dev,
5902 			(struct rte_eth_ethertype_filter *)arg,
5903 			TRUE);
5904 		break;
5905 	case RTE_ETH_FILTER_DELETE:
5906 		ret = ixgbe_add_del_ethertype_filter(dev,
5907 			(struct rte_eth_ethertype_filter *)arg,
5908 			FALSE);
5909 		break;
5910 	case RTE_ETH_FILTER_GET:
5911 		ret = ixgbe_get_ethertype_filter(dev,
5912 			(struct rte_eth_ethertype_filter *)arg);
5913 		break;
5914 	default:
5915 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
5916 		ret = -EINVAL;
5917 		break;
5918 	}
5919 	return ret;
5920 }
5921 
5922 static int
5923 ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
5924 		     enum rte_filter_type filter_type,
5925 		     enum rte_filter_op filter_op,
5926 		     void *arg)
5927 {
5928 	int ret = -EINVAL;
5929 
5930 	switch (filter_type) {
5931 	case RTE_ETH_FILTER_NTUPLE:
5932 		ret = ixgbe_ntuple_filter_handle(dev, filter_op, arg);
5933 		break;
5934 	case RTE_ETH_FILTER_ETHERTYPE:
5935 		ret = ixgbe_ethertype_filter_handle(dev, filter_op, arg);
5936 		break;
5937 	case RTE_ETH_FILTER_SYN:
5938 		ret = ixgbe_syn_filter_handle(dev, filter_op, arg);
5939 		break;
5940 	case RTE_ETH_FILTER_FDIR:
5941 		ret = ixgbe_fdir_ctrl_func(dev, filter_op, arg);
5942 		break;
5943 	case RTE_ETH_FILTER_L2_TUNNEL:
5944 		ret = ixgbe_dev_l2_tunnel_filter_handle(dev, filter_op, arg);
5945 		break;
5946 	default:
5947 		PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
5948 							filter_type);
5949 		break;
5950 	}
5951 
5952 	return ret;
5953 }
5954 
5955 static u8 *
5956 ixgbe_dev_addr_list_itr(__attribute__((unused)) struct ixgbe_hw *hw,
5957 			u8 **mc_addr_ptr, u32 *vmdq)
5958 {
5959 	u8 *mc_addr;
5960 
5961 	*vmdq = 0;
5962 	mc_addr = *mc_addr_ptr;
5963 	*mc_addr_ptr = (mc_addr + sizeof(struct ether_addr));
5964 	return mc_addr;
5965 }
5966 
5967 static int
5968 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
5969 			  struct ether_addr *mc_addr_set,
5970 			  uint32_t nb_mc_addr)
5971 {
5972 	struct ixgbe_hw *hw;
5973 	u8 *mc_addr_list;
5974 
5975 	hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5976 	mc_addr_list = (u8 *)mc_addr_set;
5977 	return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
5978 					 ixgbe_dev_addr_list_itr, TRUE);
5979 }
5980 
5981 static uint64_t
5982 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
5983 {
5984 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5985 	uint64_t systime_cycles;
5986 
5987 	switch (hw->mac.type) {
5988 	case ixgbe_mac_X550:
5989 	case ixgbe_mac_X550EM_x:
5990 	case ixgbe_mac_X550EM_a:
5991 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
5992 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
5993 		systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
5994 				* NSEC_PER_SEC;
5995 		break;
5996 	default:
5997 		systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
5998 		systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
5999 				<< 32;
6000 	}
6001 
6002 	return systime_cycles;
6003 }
6004 
6005 static uint64_t
6006 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6007 {
6008 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6009 	uint64_t rx_tstamp_cycles;
6010 
6011 	switch (hw->mac.type) {
6012 	case ixgbe_mac_X550:
6013 	case ixgbe_mac_X550EM_x:
6014 	case ixgbe_mac_X550EM_a:
6015 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6016 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6017 		rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6018 				* NSEC_PER_SEC;
6019 		break;
6020 	default:
6021 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
6022 		rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6023 		rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6024 				<< 32;
6025 	}
6026 
6027 	return rx_tstamp_cycles;
6028 }
6029 
6030 static uint64_t
6031 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6032 {
6033 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6034 	uint64_t tx_tstamp_cycles;
6035 
6036 	switch (hw->mac.type) {
6037 	case ixgbe_mac_X550:
6038 	case ixgbe_mac_X550EM_x:
6039 	case ixgbe_mac_X550EM_a:
6040 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6041 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6042 		tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6043 				* NSEC_PER_SEC;
6044 		break;
6045 	default:
6046 		/* TXSTMPL stores ns and TXSTMPH stores seconds. */
6047 		tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6048 		tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6049 				<< 32;
6050 	}
6051 
6052 	return tx_tstamp_cycles;
6053 }
6054 
6055 static void
6056 ixgbe_start_timecounters(struct rte_eth_dev *dev)
6057 {
6058 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6059 	struct ixgbe_adapter *adapter =
6060 		(struct ixgbe_adapter *)dev->data->dev_private;
6061 	struct rte_eth_link link;
6062 	uint32_t incval = 0;
6063 	uint32_t shift = 0;
6064 
6065 	/* Get current link speed. */
6066 	memset(&link, 0, sizeof(link));
6067 	ixgbe_dev_link_update(dev, 1);
6068 	rte_ixgbe_dev_atomic_read_link_status(dev, &link);
6069 
6070 	switch (link.link_speed) {
6071 	case ETH_SPEED_NUM_100M:
6072 		incval = IXGBE_INCVAL_100;
6073 		shift = IXGBE_INCVAL_SHIFT_100;
6074 		break;
6075 	case ETH_SPEED_NUM_1G:
6076 		incval = IXGBE_INCVAL_1GB;
6077 		shift = IXGBE_INCVAL_SHIFT_1GB;
6078 		break;
6079 	case ETH_SPEED_NUM_10G:
6080 	default:
6081 		incval = IXGBE_INCVAL_10GB;
6082 		shift = IXGBE_INCVAL_SHIFT_10GB;
6083 		break;
6084 	}
6085 
6086 	switch (hw->mac.type) {
6087 	case ixgbe_mac_X550:
6088 	case ixgbe_mac_X550EM_x:
6089 	case ixgbe_mac_X550EM_a:
6090 		/* Independent of link speed. */
6091 		incval = 1;
6092 		/* Cycles read will be interpreted as ns. */
6093 		shift = 0;
6094 		/* Fall-through */
6095 	case ixgbe_mac_X540:
6096 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
6097 		break;
6098 	case ixgbe_mac_82599EB:
6099 		incval >>= IXGBE_INCVAL_SHIFT_82599;
6100 		shift -= IXGBE_INCVAL_SHIFT_82599;
6101 		IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
6102 				(1 << IXGBE_INCPER_SHIFT_82599) | incval);
6103 		break;
6104 	default:
6105 		/* Not supported. */
6106 		return;
6107 	}
6108 
6109 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
6110 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6111 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6112 
6113 	adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6114 	adapter->systime_tc.cc_shift = shift;
6115 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
6116 
6117 	adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6118 	adapter->rx_tstamp_tc.cc_shift = shift;
6119 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6120 
6121 	adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6122 	adapter->tx_tstamp_tc.cc_shift = shift;
6123 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6124 }
6125 
6126 static int
6127 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
6128 {
6129 	struct ixgbe_adapter *adapter =
6130 			(struct ixgbe_adapter *)dev->data->dev_private;
6131 
6132 	adapter->systime_tc.nsec += delta;
6133 	adapter->rx_tstamp_tc.nsec += delta;
6134 	adapter->tx_tstamp_tc.nsec += delta;
6135 
6136 	return 0;
6137 }
6138 
6139 static int
6140 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
6141 {
6142 	uint64_t ns;
6143 	struct ixgbe_adapter *adapter =
6144 			(struct ixgbe_adapter *)dev->data->dev_private;
6145 
6146 	ns = rte_timespec_to_ns(ts);
6147 	/* Set the timecounters to a new value. */
6148 	adapter->systime_tc.nsec = ns;
6149 	adapter->rx_tstamp_tc.nsec = ns;
6150 	adapter->tx_tstamp_tc.nsec = ns;
6151 
6152 	return 0;
6153 }
6154 
6155 static int
6156 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
6157 {
6158 	uint64_t ns, systime_cycles;
6159 	struct ixgbe_adapter *adapter =
6160 			(struct ixgbe_adapter *)dev->data->dev_private;
6161 
6162 	systime_cycles = ixgbe_read_systime_cyclecounter(dev);
6163 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
6164 	*ts = rte_ns_to_timespec(ns);
6165 
6166 	return 0;
6167 }
6168 
6169 static int
6170 ixgbe_timesync_enable(struct rte_eth_dev *dev)
6171 {
6172 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6173 	uint32_t tsync_ctl;
6174 	uint32_t tsauxc;
6175 
6176 	/* Stop the timesync system time. */
6177 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
6178 	/* Reset the timesync system time value. */
6179 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
6180 	IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
6181 
6182 	/* Enable system time for platforms where it isn't on by default. */
6183 	tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
6184 	tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
6185 	IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
6186 
6187 	ixgbe_start_timecounters(dev);
6188 
6189 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6190 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
6191 			(ETHER_TYPE_1588 |
6192 			 IXGBE_ETQF_FILTER_EN |
6193 			 IXGBE_ETQF_1588));
6194 
6195 	/* Enable timestamping of received PTP packets. */
6196 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6197 	tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
6198 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6199 
6200 	/* Enable timestamping of transmitted PTP packets. */
6201 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6202 	tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
6203 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6204 
6205 	IXGBE_WRITE_FLUSH(hw);
6206 
6207 	return 0;
6208 }
6209 
6210 static int
6211 ixgbe_timesync_disable(struct rte_eth_dev *dev)
6212 {
6213 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6214 	uint32_t tsync_ctl;
6215 
6216 	/* Disable timestamping of transmitted PTP packets. */
6217 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6218 	tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
6219 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6220 
6221 	/* Disable timestamping of received PTP packets. */
6222 	tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6223 	tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
6224 	IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6225 
6226 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6227 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
6228 
6229 	/* Stop incrementating the System Time registers. */
6230 	IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
6231 
6232 	return 0;
6233 }
6234 
6235 static int
6236 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
6237 				 struct timespec *timestamp,
6238 				 uint32_t flags __rte_unused)
6239 {
6240 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6241 	struct ixgbe_adapter *adapter =
6242 		(struct ixgbe_adapter *)dev->data->dev_private;
6243 	uint32_t tsync_rxctl;
6244 	uint64_t rx_tstamp_cycles;
6245 	uint64_t ns;
6246 
6247 	tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6248 	if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
6249 		return -EINVAL;
6250 
6251 	rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
6252 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
6253 	*timestamp = rte_ns_to_timespec(ns);
6254 
6255 	return  0;
6256 }
6257 
6258 static int
6259 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
6260 				 struct timespec *timestamp)
6261 {
6262 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6263 	struct ixgbe_adapter *adapter =
6264 		(struct ixgbe_adapter *)dev->data->dev_private;
6265 	uint32_t tsync_txctl;
6266 	uint64_t tx_tstamp_cycles;
6267 	uint64_t ns;
6268 
6269 	tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6270 	if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
6271 		return -EINVAL;
6272 
6273 	tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
6274 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
6275 	*timestamp = rte_ns_to_timespec(ns);
6276 
6277 	return 0;
6278 }
6279 
6280 static int
6281 ixgbe_get_reg_length(struct rte_eth_dev *dev)
6282 {
6283 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6284 	int count = 0;
6285 	int g_ind = 0;
6286 	const struct reg_info *reg_group;
6287 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6288 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6289 
6290 	while ((reg_group = reg_set[g_ind++]))
6291 		count += ixgbe_regs_group_count(reg_group);
6292 
6293 	return count;
6294 }
6295 
6296 static int
6297 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
6298 {
6299 	int count = 0;
6300 	int g_ind = 0;
6301 	const struct reg_info *reg_group;
6302 
6303 	while ((reg_group = ixgbevf_regs[g_ind++]))
6304 		count += ixgbe_regs_group_count(reg_group);
6305 
6306 	return count;
6307 }
6308 
6309 static int
6310 ixgbe_get_regs(struct rte_eth_dev *dev,
6311 	      struct rte_dev_reg_info *regs)
6312 {
6313 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6314 	uint32_t *data = regs->data;
6315 	int g_ind = 0;
6316 	int count = 0;
6317 	const struct reg_info *reg_group;
6318 	const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6319 				    ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6320 
6321 	if (data == NULL) {
6322 		regs->length = ixgbe_get_reg_length(dev);
6323 		regs->width = sizeof(uint32_t);
6324 		return 0;
6325 	}
6326 
6327 	/* Support only full register dump */
6328 	if ((regs->length == 0) ||
6329 	    (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
6330 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
6331 			hw->device_id;
6332 		while ((reg_group = reg_set[g_ind++]))
6333 			count += ixgbe_read_regs_group(dev, &data[count],
6334 				reg_group);
6335 		return 0;
6336 	}
6337 
6338 	return -ENOTSUP;
6339 }
6340 
6341 static int
6342 ixgbevf_get_regs(struct rte_eth_dev *dev,
6343 		struct rte_dev_reg_info *regs)
6344 {
6345 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6346 	uint32_t *data = regs->data;
6347 	int g_ind = 0;
6348 	int count = 0;
6349 	const struct reg_info *reg_group;
6350 
6351 	if (data == NULL) {
6352 		regs->length = ixgbevf_get_reg_length(dev);
6353 		regs->width = sizeof(uint32_t);
6354 		return 0;
6355 	}
6356 
6357 	/* Support only full register dump */
6358 	if ((regs->length == 0) ||
6359 	    (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
6360 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
6361 			hw->device_id;
6362 		while ((reg_group = ixgbevf_regs[g_ind++]))
6363 			count += ixgbe_read_regs_group(dev, &data[count],
6364 						      reg_group);
6365 		return 0;
6366 	}
6367 
6368 	return -ENOTSUP;
6369 }
6370 
6371 static int
6372 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
6373 {
6374 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6375 
6376 	/* Return unit is byte count */
6377 	return hw->eeprom.word_size * 2;
6378 }
6379 
6380 static int
6381 ixgbe_get_eeprom(struct rte_eth_dev *dev,
6382 		struct rte_dev_eeprom_info *in_eeprom)
6383 {
6384 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6385 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
6386 	uint16_t *data = in_eeprom->data;
6387 	int first, length;
6388 
6389 	first = in_eeprom->offset >> 1;
6390 	length = in_eeprom->length >> 1;
6391 	if ((first > hw->eeprom.word_size) ||
6392 	    ((first + length) > hw->eeprom.word_size))
6393 		return -EINVAL;
6394 
6395 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
6396 
6397 	return eeprom->ops.read_buffer(hw, first, length, data);
6398 }
6399 
6400 static int
6401 ixgbe_set_eeprom(struct rte_eth_dev *dev,
6402 		struct rte_dev_eeprom_info *in_eeprom)
6403 {
6404 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6405 	struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
6406 	uint16_t *data = in_eeprom->data;
6407 	int first, length;
6408 
6409 	first = in_eeprom->offset >> 1;
6410 	length = in_eeprom->length >> 1;
6411 	if ((first > hw->eeprom.word_size) ||
6412 	    ((first + length) > hw->eeprom.word_size))
6413 		return -EINVAL;
6414 
6415 	in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
6416 
6417 	return eeprom->ops.write_buffer(hw,  first, length, data);
6418 }
6419 
6420 uint16_t
6421 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
6422 	switch (mac_type) {
6423 	case ixgbe_mac_X550:
6424 	case ixgbe_mac_X550EM_x:
6425 	case ixgbe_mac_X550EM_a:
6426 		return ETH_RSS_RETA_SIZE_512;
6427 	case ixgbe_mac_X550_vf:
6428 	case ixgbe_mac_X550EM_x_vf:
6429 	case ixgbe_mac_X550EM_a_vf:
6430 		return ETH_RSS_RETA_SIZE_64;
6431 	default:
6432 		return ETH_RSS_RETA_SIZE_128;
6433 	}
6434 }
6435 
6436 uint32_t
6437 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
6438 	switch (mac_type) {
6439 	case ixgbe_mac_X550:
6440 	case ixgbe_mac_X550EM_x:
6441 	case ixgbe_mac_X550EM_a:
6442 		if (reta_idx < ETH_RSS_RETA_SIZE_128)
6443 			return IXGBE_RETA(reta_idx >> 2);
6444 		else
6445 			return IXGBE_ERETA((reta_idx - ETH_RSS_RETA_SIZE_128) >> 2);
6446 	case ixgbe_mac_X550_vf:
6447 	case ixgbe_mac_X550EM_x_vf:
6448 	case ixgbe_mac_X550EM_a_vf:
6449 		return IXGBE_VFRETA(reta_idx >> 2);
6450 	default:
6451 		return IXGBE_RETA(reta_idx >> 2);
6452 	}
6453 }
6454 
6455 uint32_t
6456 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
6457 	switch (mac_type) {
6458 	case ixgbe_mac_X550_vf:
6459 	case ixgbe_mac_X550EM_x_vf:
6460 	case ixgbe_mac_X550EM_a_vf:
6461 		return IXGBE_VFMRQC;
6462 	default:
6463 		return IXGBE_MRQC;
6464 	}
6465 }
6466 
6467 uint32_t
6468 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
6469 	switch (mac_type) {
6470 	case ixgbe_mac_X550_vf:
6471 	case ixgbe_mac_X550EM_x_vf:
6472 	case ixgbe_mac_X550EM_a_vf:
6473 		return IXGBE_VFRSSRK(i);
6474 	default:
6475 		return IXGBE_RSSRK(i);
6476 	}
6477 }
6478 
6479 bool
6480 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
6481 	switch (mac_type) {
6482 	case ixgbe_mac_82599_vf:
6483 	case ixgbe_mac_X540_vf:
6484 		return 0;
6485 	default:
6486 		return 1;
6487 	}
6488 }
6489 
6490 static int
6491 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
6492 			struct rte_eth_dcb_info *dcb_info)
6493 {
6494 	struct ixgbe_dcb_config *dcb_config =
6495 			IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
6496 	struct ixgbe_dcb_tc_config *tc;
6497 	uint8_t i, j;
6498 
6499 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_DCB_FLAG)
6500 		dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
6501 	else
6502 		dcb_info->nb_tcs = 1;
6503 
6504 	if (dcb_config->vt_mode) { /* vt is enabled*/
6505 		struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
6506 				&dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
6507 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
6508 			dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
6509 		for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
6510 			for (j = 0; j < dcb_info->nb_tcs; j++) {
6511 				dcb_info->tc_queue.tc_rxq[i][j].base =
6512 						i * dcb_info->nb_tcs + j;
6513 				dcb_info->tc_queue.tc_rxq[i][j].nb_queue = 1;
6514 				dcb_info->tc_queue.tc_txq[i][j].base =
6515 						i * dcb_info->nb_tcs + j;
6516 				dcb_info->tc_queue.tc_txq[i][j].nb_queue = 1;
6517 			}
6518 		}
6519 	} else { /* vt is disabled*/
6520 		struct rte_eth_dcb_rx_conf *rx_conf =
6521 				&dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
6522 		for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
6523 			dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
6524 		if (dcb_info->nb_tcs == ETH_4_TCS) {
6525 			for (i = 0; i < dcb_info->nb_tcs; i++) {
6526 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
6527 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
6528 			}
6529 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
6530 			dcb_info->tc_queue.tc_txq[0][1].base = 64;
6531 			dcb_info->tc_queue.tc_txq[0][2].base = 96;
6532 			dcb_info->tc_queue.tc_txq[0][3].base = 112;
6533 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
6534 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
6535 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
6536 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
6537 		} else if (dcb_info->nb_tcs == ETH_8_TCS) {
6538 			for (i = 0; i < dcb_info->nb_tcs; i++) {
6539 				dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
6540 				dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
6541 			}
6542 			dcb_info->tc_queue.tc_txq[0][0].base = 0;
6543 			dcb_info->tc_queue.tc_txq[0][1].base = 32;
6544 			dcb_info->tc_queue.tc_txq[0][2].base = 64;
6545 			dcb_info->tc_queue.tc_txq[0][3].base = 80;
6546 			dcb_info->tc_queue.tc_txq[0][4].base = 96;
6547 			dcb_info->tc_queue.tc_txq[0][5].base = 104;
6548 			dcb_info->tc_queue.tc_txq[0][6].base = 112;
6549 			dcb_info->tc_queue.tc_txq[0][7].base = 120;
6550 			dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
6551 			dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
6552 			dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
6553 			dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
6554 			dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
6555 			dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
6556 			dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
6557 			dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
6558 		}
6559 	}
6560 	for (i = 0; i < dcb_info->nb_tcs; i++) {
6561 		tc = &dcb_config->tc_config[i];
6562 		dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
6563 	}
6564 	return 0;
6565 }
6566 
6567 /* Update e-tag ether type */
6568 static int
6569 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
6570 			    uint16_t ether_type)
6571 {
6572 	uint32_t etag_etype;
6573 
6574 	if (hw->mac.type != ixgbe_mac_X550 &&
6575 	    hw->mac.type != ixgbe_mac_X550EM_x &&
6576 	    hw->mac.type != ixgbe_mac_X550EM_a) {
6577 		return -ENOTSUP;
6578 	}
6579 
6580 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
6581 	etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
6582 	etag_etype |= ether_type;
6583 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
6584 	IXGBE_WRITE_FLUSH(hw);
6585 
6586 	return 0;
6587 }
6588 
6589 /* Config l2 tunnel ether type */
6590 static int
6591 ixgbe_dev_l2_tunnel_eth_type_conf(struct rte_eth_dev *dev,
6592 				  struct rte_eth_l2_tunnel_conf *l2_tunnel)
6593 {
6594 	int ret = 0;
6595 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6596 
6597 	if (l2_tunnel == NULL)
6598 		return -EINVAL;
6599 
6600 	switch (l2_tunnel->l2_tunnel_type) {
6601 	case RTE_L2_TUNNEL_TYPE_E_TAG:
6602 		ret = ixgbe_update_e_tag_eth_type(hw, l2_tunnel->ether_type);
6603 		break;
6604 	default:
6605 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
6606 		ret = -EINVAL;
6607 		break;
6608 	}
6609 
6610 	return ret;
6611 }
6612 
6613 /* Enable e-tag tunnel */
6614 static int
6615 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
6616 {
6617 	uint32_t etag_etype;
6618 
6619 	if (hw->mac.type != ixgbe_mac_X550 &&
6620 	    hw->mac.type != ixgbe_mac_X550EM_x &&
6621 	    hw->mac.type != ixgbe_mac_X550EM_a) {
6622 		return -ENOTSUP;
6623 	}
6624 
6625 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
6626 	etag_etype |= IXGBE_ETAG_ETYPE_VALID;
6627 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
6628 	IXGBE_WRITE_FLUSH(hw);
6629 
6630 	return 0;
6631 }
6632 
6633 /* Enable l2 tunnel */
6634 static int
6635 ixgbe_dev_l2_tunnel_enable(struct rte_eth_dev *dev,
6636 			   enum rte_eth_tunnel_type l2_tunnel_type)
6637 {
6638 	int ret = 0;
6639 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6640 
6641 	switch (l2_tunnel_type) {
6642 	case RTE_L2_TUNNEL_TYPE_E_TAG:
6643 		ret = ixgbe_e_tag_enable(hw);
6644 		break;
6645 	default:
6646 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
6647 		ret = -EINVAL;
6648 		break;
6649 	}
6650 
6651 	return ret;
6652 }
6653 
6654 /* Disable e-tag tunnel */
6655 static int
6656 ixgbe_e_tag_disable(struct ixgbe_hw *hw)
6657 {
6658 	uint32_t etag_etype;
6659 
6660 	if (hw->mac.type != ixgbe_mac_X550 &&
6661 	    hw->mac.type != ixgbe_mac_X550EM_x &&
6662 	    hw->mac.type != ixgbe_mac_X550EM_a) {
6663 		return -ENOTSUP;
6664 	}
6665 
6666 	etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
6667 	etag_etype &= ~IXGBE_ETAG_ETYPE_VALID;
6668 	IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
6669 	IXGBE_WRITE_FLUSH(hw);
6670 
6671 	return 0;
6672 }
6673 
6674 /* Disable l2 tunnel */
6675 static int
6676 ixgbe_dev_l2_tunnel_disable(struct rte_eth_dev *dev,
6677 			    enum rte_eth_tunnel_type l2_tunnel_type)
6678 {
6679 	int ret = 0;
6680 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6681 
6682 	switch (l2_tunnel_type) {
6683 	case RTE_L2_TUNNEL_TYPE_E_TAG:
6684 		ret = ixgbe_e_tag_disable(hw);
6685 		break;
6686 	default:
6687 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
6688 		ret = -EINVAL;
6689 		break;
6690 	}
6691 
6692 	return ret;
6693 }
6694 
6695 static int
6696 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
6697 		       struct rte_eth_l2_tunnel_conf *l2_tunnel)
6698 {
6699 	int ret = 0;
6700 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6701 	uint32_t i, rar_entries;
6702 	uint32_t rar_low, rar_high;
6703 
6704 	if (hw->mac.type != ixgbe_mac_X550 &&
6705 	    hw->mac.type != ixgbe_mac_X550EM_x &&
6706 	    hw->mac.type != ixgbe_mac_X550EM_a) {
6707 		return -ENOTSUP;
6708 	}
6709 
6710 	rar_entries = ixgbe_get_num_rx_addrs(hw);
6711 
6712 	for (i = 1; i < rar_entries; i++) {
6713 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
6714 		rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
6715 		if ((rar_high & IXGBE_RAH_AV) &&
6716 		    (rar_high & IXGBE_RAH_ADTYPE) &&
6717 		    ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
6718 		     l2_tunnel->tunnel_id)) {
6719 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
6720 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
6721 
6722 			ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
6723 
6724 			return ret;
6725 		}
6726 	}
6727 
6728 	return ret;
6729 }
6730 
6731 static int
6732 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
6733 		       struct rte_eth_l2_tunnel_conf *l2_tunnel)
6734 {
6735 	int ret = 0;
6736 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6737 	uint32_t i, rar_entries;
6738 	uint32_t rar_low, rar_high;
6739 
6740 	if (hw->mac.type != ixgbe_mac_X550 &&
6741 	    hw->mac.type != ixgbe_mac_X550EM_x &&
6742 	    hw->mac.type != ixgbe_mac_X550EM_a) {
6743 		return -ENOTSUP;
6744 	}
6745 
6746 	/* One entry for one tunnel. Try to remove potential existing entry. */
6747 	ixgbe_e_tag_filter_del(dev, l2_tunnel);
6748 
6749 	rar_entries = ixgbe_get_num_rx_addrs(hw);
6750 
6751 	for (i = 1; i < rar_entries; i++) {
6752 		rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
6753 		if (rar_high & IXGBE_RAH_AV) {
6754 			continue;
6755 		} else {
6756 			ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
6757 			rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
6758 			rar_low = l2_tunnel->tunnel_id;
6759 
6760 			IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
6761 			IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
6762 
6763 			return ret;
6764 		}
6765 	}
6766 
6767 	PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
6768 		     " Please remove a rule before adding a new one.");
6769 	return -EINVAL;
6770 }
6771 
6772 /* Add l2 tunnel filter */
6773 static int
6774 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
6775 			       struct rte_eth_l2_tunnel_conf *l2_tunnel)
6776 {
6777 	int ret = 0;
6778 
6779 	switch (l2_tunnel->l2_tunnel_type) {
6780 	case RTE_L2_TUNNEL_TYPE_E_TAG:
6781 		ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
6782 		break;
6783 	default:
6784 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
6785 		ret = -EINVAL;
6786 		break;
6787 	}
6788 
6789 	return ret;
6790 }
6791 
6792 /* Delete l2 tunnel filter */
6793 static int
6794 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
6795 			       struct rte_eth_l2_tunnel_conf *l2_tunnel)
6796 {
6797 	int ret = 0;
6798 
6799 	switch (l2_tunnel->l2_tunnel_type) {
6800 	case RTE_L2_TUNNEL_TYPE_E_TAG:
6801 		ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
6802 		break;
6803 	default:
6804 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
6805 		ret = -EINVAL;
6806 		break;
6807 	}
6808 
6809 	return ret;
6810 }
6811 
6812 /**
6813  * ixgbe_dev_l2_tunnel_filter_handle - Handle operations for l2 tunnel filter.
6814  * @dev: pointer to rte_eth_dev structure
6815  * @filter_op:operation will be taken.
6816  * @arg: a pointer to specific structure corresponding to the filter_op
6817  */
6818 static int
6819 ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
6820 				  enum rte_filter_op filter_op,
6821 				  void *arg)
6822 {
6823 	int ret = 0;
6824 
6825 	if (filter_op == RTE_ETH_FILTER_NOP)
6826 		return 0;
6827 
6828 	if (arg == NULL) {
6829 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
6830 			    filter_op);
6831 		return -EINVAL;
6832 	}
6833 
6834 	switch (filter_op) {
6835 	case RTE_ETH_FILTER_ADD:
6836 		ret = ixgbe_dev_l2_tunnel_filter_add
6837 			(dev,
6838 			 (struct rte_eth_l2_tunnel_conf *)arg);
6839 		break;
6840 	case RTE_ETH_FILTER_DELETE:
6841 		ret = ixgbe_dev_l2_tunnel_filter_del
6842 			(dev,
6843 			 (struct rte_eth_l2_tunnel_conf *)arg);
6844 		break;
6845 	default:
6846 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
6847 		ret = -EINVAL;
6848 		break;
6849 	}
6850 	return ret;
6851 }
6852 
6853 static int
6854 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
6855 {
6856 	int ret = 0;
6857 	uint32_t ctrl;
6858 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6859 
6860 	if (hw->mac.type != ixgbe_mac_X550 &&
6861 	    hw->mac.type != ixgbe_mac_X550EM_x &&
6862 	    hw->mac.type != ixgbe_mac_X550EM_a) {
6863 		return -ENOTSUP;
6864 	}
6865 
6866 	ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
6867 	ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
6868 	if (en)
6869 		ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
6870 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
6871 
6872 	return ret;
6873 }
6874 
6875 /* Enable l2 tunnel forwarding */
6876 static int
6877 ixgbe_dev_l2_tunnel_forwarding_enable
6878 	(struct rte_eth_dev *dev,
6879 	 enum rte_eth_tunnel_type l2_tunnel_type)
6880 {
6881 	int ret = 0;
6882 
6883 	switch (l2_tunnel_type) {
6884 	case RTE_L2_TUNNEL_TYPE_E_TAG:
6885 		ret = ixgbe_e_tag_forwarding_en_dis(dev, 1);
6886 		break;
6887 	default:
6888 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
6889 		ret = -EINVAL;
6890 		break;
6891 	}
6892 
6893 	return ret;
6894 }
6895 
6896 /* Disable l2 tunnel forwarding */
6897 static int
6898 ixgbe_dev_l2_tunnel_forwarding_disable
6899 	(struct rte_eth_dev *dev,
6900 	 enum rte_eth_tunnel_type l2_tunnel_type)
6901 {
6902 	int ret = 0;
6903 
6904 	switch (l2_tunnel_type) {
6905 	case RTE_L2_TUNNEL_TYPE_E_TAG:
6906 		ret = ixgbe_e_tag_forwarding_en_dis(dev, 0);
6907 		break;
6908 	default:
6909 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
6910 		ret = -EINVAL;
6911 		break;
6912 	}
6913 
6914 	return ret;
6915 }
6916 
6917 static int
6918 ixgbe_e_tag_insertion_en_dis(struct rte_eth_dev *dev,
6919 			     struct rte_eth_l2_tunnel_conf *l2_tunnel,
6920 			     bool en)
6921 {
6922 	int ret = 0;
6923 	uint32_t vmtir, vmvir;
6924 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6925 
6926 	if (l2_tunnel->vf_id >= dev->pci_dev->max_vfs) {
6927 		PMD_DRV_LOG(ERR,
6928 			    "VF id %u should be less than %u",
6929 			    l2_tunnel->vf_id,
6930 			    dev->pci_dev->max_vfs);
6931 		return -EINVAL;
6932 	}
6933 
6934 	if (hw->mac.type != ixgbe_mac_X550 &&
6935 	    hw->mac.type != ixgbe_mac_X550EM_x &&
6936 	    hw->mac.type != ixgbe_mac_X550EM_a) {
6937 		return -ENOTSUP;
6938 	}
6939 
6940 	if (en)
6941 		vmtir = l2_tunnel->tunnel_id;
6942 	else
6943 		vmtir = 0;
6944 
6945 	IXGBE_WRITE_REG(hw, IXGBE_VMTIR(l2_tunnel->vf_id), vmtir);
6946 
6947 	vmvir = IXGBE_READ_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id));
6948 	vmvir &= ~IXGBE_VMVIR_TAGA_MASK;
6949 	if (en)
6950 		vmvir |= IXGBE_VMVIR_TAGA_ETAG_INSERT;
6951 	IXGBE_WRITE_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id), vmvir);
6952 
6953 	return ret;
6954 }
6955 
6956 /* Enable l2 tunnel tag insertion */
6957 static int
6958 ixgbe_dev_l2_tunnel_insertion_enable(struct rte_eth_dev *dev,
6959 				     struct rte_eth_l2_tunnel_conf *l2_tunnel)
6960 {
6961 	int ret = 0;
6962 
6963 	switch (l2_tunnel->l2_tunnel_type) {
6964 	case RTE_L2_TUNNEL_TYPE_E_TAG:
6965 		ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 1);
6966 		break;
6967 	default:
6968 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
6969 		ret = -EINVAL;
6970 		break;
6971 	}
6972 
6973 	return ret;
6974 }
6975 
6976 /* Disable l2 tunnel tag insertion */
6977 static int
6978 ixgbe_dev_l2_tunnel_insertion_disable
6979 	(struct rte_eth_dev *dev,
6980 	 struct rte_eth_l2_tunnel_conf *l2_tunnel)
6981 {
6982 	int ret = 0;
6983 
6984 	switch (l2_tunnel->l2_tunnel_type) {
6985 	case RTE_L2_TUNNEL_TYPE_E_TAG:
6986 		ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 0);
6987 		break;
6988 	default:
6989 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
6990 		ret = -EINVAL;
6991 		break;
6992 	}
6993 
6994 	return ret;
6995 }
6996 
6997 static int
6998 ixgbe_e_tag_stripping_en_dis(struct rte_eth_dev *dev,
6999 			     bool en)
7000 {
7001 	int ret = 0;
7002 	uint32_t qde;
7003 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7004 
7005 	if (hw->mac.type != ixgbe_mac_X550 &&
7006 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7007 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7008 		return -ENOTSUP;
7009 	}
7010 
7011 	qde = IXGBE_READ_REG(hw, IXGBE_QDE);
7012 	if (en)
7013 		qde |= IXGBE_QDE_STRIP_TAG;
7014 	else
7015 		qde &= ~IXGBE_QDE_STRIP_TAG;
7016 	qde &= ~IXGBE_QDE_READ;
7017 	qde |= IXGBE_QDE_WRITE;
7018 	IXGBE_WRITE_REG(hw, IXGBE_QDE, qde);
7019 
7020 	return ret;
7021 }
7022 
7023 /* Enable l2 tunnel tag stripping */
7024 static int
7025 ixgbe_dev_l2_tunnel_stripping_enable
7026 	(struct rte_eth_dev *dev,
7027 	 enum rte_eth_tunnel_type l2_tunnel_type)
7028 {
7029 	int ret = 0;
7030 
7031 	switch (l2_tunnel_type) {
7032 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7033 		ret = ixgbe_e_tag_stripping_en_dis(dev, 1);
7034 		break;
7035 	default:
7036 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7037 		ret = -EINVAL;
7038 		break;
7039 	}
7040 
7041 	return ret;
7042 }
7043 
7044 /* Disable l2 tunnel tag stripping */
7045 static int
7046 ixgbe_dev_l2_tunnel_stripping_disable
7047 	(struct rte_eth_dev *dev,
7048 	 enum rte_eth_tunnel_type l2_tunnel_type)
7049 {
7050 	int ret = 0;
7051 
7052 	switch (l2_tunnel_type) {
7053 	case RTE_L2_TUNNEL_TYPE_E_TAG:
7054 		ret = ixgbe_e_tag_stripping_en_dis(dev, 0);
7055 		break;
7056 	default:
7057 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7058 		ret = -EINVAL;
7059 		break;
7060 	}
7061 
7062 	return ret;
7063 }
7064 
7065 /* Enable/disable l2 tunnel offload functions */
7066 static int
7067 ixgbe_dev_l2_tunnel_offload_set
7068 	(struct rte_eth_dev *dev,
7069 	 struct rte_eth_l2_tunnel_conf *l2_tunnel,
7070 	 uint32_t mask,
7071 	 uint8_t en)
7072 {
7073 	int ret = 0;
7074 
7075 	if (l2_tunnel == NULL)
7076 		return -EINVAL;
7077 
7078 	ret = -EINVAL;
7079 	if (mask & ETH_L2_TUNNEL_ENABLE_MASK) {
7080 		if (en)
7081 			ret = ixgbe_dev_l2_tunnel_enable(
7082 				dev,
7083 				l2_tunnel->l2_tunnel_type);
7084 		else
7085 			ret = ixgbe_dev_l2_tunnel_disable(
7086 				dev,
7087 				l2_tunnel->l2_tunnel_type);
7088 	}
7089 
7090 	if (mask & ETH_L2_TUNNEL_INSERTION_MASK) {
7091 		if (en)
7092 			ret = ixgbe_dev_l2_tunnel_insertion_enable(
7093 				dev,
7094 				l2_tunnel);
7095 		else
7096 			ret = ixgbe_dev_l2_tunnel_insertion_disable(
7097 				dev,
7098 				l2_tunnel);
7099 	}
7100 
7101 	if (mask & ETH_L2_TUNNEL_STRIPPING_MASK) {
7102 		if (en)
7103 			ret = ixgbe_dev_l2_tunnel_stripping_enable(
7104 				dev,
7105 				l2_tunnel->l2_tunnel_type);
7106 		else
7107 			ret = ixgbe_dev_l2_tunnel_stripping_disable(
7108 				dev,
7109 				l2_tunnel->l2_tunnel_type);
7110 	}
7111 
7112 	if (mask & ETH_L2_TUNNEL_FORWARDING_MASK) {
7113 		if (en)
7114 			ret = ixgbe_dev_l2_tunnel_forwarding_enable(
7115 				dev,
7116 				l2_tunnel->l2_tunnel_type);
7117 		else
7118 			ret = ixgbe_dev_l2_tunnel_forwarding_disable(
7119 				dev,
7120 				l2_tunnel->l2_tunnel_type);
7121 	}
7122 
7123 	return ret;
7124 }
7125 
7126 static int
7127 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
7128 			uint16_t port)
7129 {
7130 	IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
7131 	IXGBE_WRITE_FLUSH(hw);
7132 
7133 	return 0;
7134 }
7135 
7136 /* There's only one register for VxLAN UDP port.
7137  * So, we cannot add several ports. Will update it.
7138  */
7139 static int
7140 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
7141 		     uint16_t port)
7142 {
7143 	if (port == 0) {
7144 		PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
7145 		return -EINVAL;
7146 	}
7147 
7148 	return ixgbe_update_vxlan_port(hw, port);
7149 }
7150 
7151 /* We cannot delete the VxLAN port. For there's a register for VxLAN
7152  * UDP port, it must have a value.
7153  * So, will reset it to the original value 0.
7154  */
7155 static int
7156 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
7157 		     uint16_t port)
7158 {
7159 	uint16_t cur_port;
7160 
7161 	cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
7162 
7163 	if (cur_port != port) {
7164 		PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
7165 		return -EINVAL;
7166 	}
7167 
7168 	return ixgbe_update_vxlan_port(hw, 0);
7169 }
7170 
7171 /* Add UDP tunneling port */
7172 static int
7173 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
7174 			      struct rte_eth_udp_tunnel *udp_tunnel)
7175 {
7176 	int ret = 0;
7177 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7178 
7179 	if (hw->mac.type != ixgbe_mac_X550 &&
7180 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7181 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7182 		return -ENOTSUP;
7183 	}
7184 
7185 	if (udp_tunnel == NULL)
7186 		return -EINVAL;
7187 
7188 	switch (udp_tunnel->prot_type) {
7189 	case RTE_TUNNEL_TYPE_VXLAN:
7190 		ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
7191 		break;
7192 
7193 	case RTE_TUNNEL_TYPE_GENEVE:
7194 	case RTE_TUNNEL_TYPE_TEREDO:
7195 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7196 		ret = -EINVAL;
7197 		break;
7198 
7199 	default:
7200 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7201 		ret = -EINVAL;
7202 		break;
7203 	}
7204 
7205 	return ret;
7206 }
7207 
7208 /* Remove UDP tunneling port */
7209 static int
7210 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
7211 			      struct rte_eth_udp_tunnel *udp_tunnel)
7212 {
7213 	int ret = 0;
7214 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7215 
7216 	if (hw->mac.type != ixgbe_mac_X550 &&
7217 	    hw->mac.type != ixgbe_mac_X550EM_x &&
7218 	    hw->mac.type != ixgbe_mac_X550EM_a) {
7219 		return -ENOTSUP;
7220 	}
7221 
7222 	if (udp_tunnel == NULL)
7223 		return -EINVAL;
7224 
7225 	switch (udp_tunnel->prot_type) {
7226 	case RTE_TUNNEL_TYPE_VXLAN:
7227 		ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
7228 		break;
7229 	case RTE_TUNNEL_TYPE_GENEVE:
7230 	case RTE_TUNNEL_TYPE_TEREDO:
7231 		PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7232 		ret = -EINVAL;
7233 		break;
7234 	default:
7235 		PMD_DRV_LOG(ERR, "Invalid tunnel type");
7236 		ret = -EINVAL;
7237 		break;
7238 	}
7239 
7240 	return ret;
7241 }
7242 
7243 /* ixgbevf_update_xcast_mode - Update Multicast mode
7244  * @hw: pointer to the HW structure
7245  * @netdev: pointer to net device structure
7246  * @xcast_mode: new multicast mode
7247  *
7248  * Updates the Multicast Mode of VF.
7249  */
7250 static int ixgbevf_update_xcast_mode(struct ixgbe_hw *hw,
7251 				     int xcast_mode)
7252 {
7253 	struct ixgbe_mbx_info *mbx = &hw->mbx;
7254 	u32 msgbuf[2];
7255 	s32 err;
7256 
7257 	switch (hw->api_version) {
7258 	case ixgbe_mbox_api_12:
7259 		break;
7260 	default:
7261 		return -EOPNOTSUPP;
7262 	}
7263 
7264 	msgbuf[0] = IXGBE_VF_UPDATE_XCAST_MODE;
7265 	msgbuf[1] = xcast_mode;
7266 
7267 	err = mbx->ops.write_posted(hw, msgbuf, 2, 0);
7268 	if (err)
7269 		return err;
7270 
7271 	err = mbx->ops.read_posted(hw, msgbuf, 2, 0);
7272 	if (err)
7273 		return err;
7274 
7275 	msgbuf[0] &= ~IXGBE_VT_MSGTYPE_CTS;
7276 	if (msgbuf[0] == (IXGBE_VF_UPDATE_XCAST_MODE | IXGBE_VT_MSGTYPE_NACK))
7277 		return -EPERM;
7278 
7279 	return 0;
7280 }
7281 
7282 static void
7283 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
7284 {
7285 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7286 
7287 	ixgbevf_update_xcast_mode(hw, IXGBEVF_XCAST_MODE_ALLMULTI);
7288 }
7289 
7290 static void
7291 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
7292 {
7293 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7294 
7295 	ixgbevf_update_xcast_mode(hw, IXGBEVF_XCAST_MODE_NONE);
7296 }
7297 
7298 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
7299 {
7300 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7301 	u32 in_msg = 0;
7302 
7303 	if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
7304 		return;
7305 
7306 	/* PF reset VF event */
7307 	if (in_msg == IXGBE_PF_CONTROL_MSG)
7308 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET);
7309 }
7310 
7311 static int
7312 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
7313 {
7314 	uint32_t eicr;
7315 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7316 	struct ixgbe_interrupt *intr =
7317 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7318 	ixgbevf_intr_disable(hw);
7319 
7320 	/* read-on-clear nic registers here */
7321 	eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
7322 	intr->flags = 0;
7323 
7324 	/* only one misc vector supported - mailbox */
7325 	eicr &= IXGBE_VTEICR_MASK;
7326 	if (eicr == IXGBE_MISC_VEC_ID)
7327 		intr->flags |= IXGBE_FLAG_MAILBOX;
7328 
7329 	return 0;
7330 }
7331 
7332 static int
7333 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
7334 {
7335 	struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7336 	struct ixgbe_interrupt *intr =
7337 		IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7338 
7339 	if (intr->flags & IXGBE_FLAG_MAILBOX) {
7340 		ixgbevf_mbx_process(dev);
7341 		intr->flags &= ~IXGBE_FLAG_MAILBOX;
7342 	}
7343 
7344 	ixgbevf_intr_enable(hw);
7345 
7346 	return 0;
7347 }
7348 
7349 static void
7350 ixgbevf_dev_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
7351 			      void *param)
7352 {
7353 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
7354 
7355 	ixgbevf_dev_interrupt_get_status(dev);
7356 	ixgbevf_dev_interrupt_action(dev);
7357 }
7358 
7359 static struct rte_driver rte_ixgbe_driver = {
7360 	.type = PMD_PDEV,
7361 	.init = rte_ixgbe_pmd_init,
7362 };
7363 
7364 static struct rte_driver rte_ixgbevf_driver = {
7365 	.type = PMD_PDEV,
7366 	.init = rte_ixgbevf_pmd_init,
7367 };
7368 
7369 PMD_REGISTER_DRIVER(rte_ixgbe_driver, ixgbe);
7370 DRIVER_REGISTER_PCI_TABLE(ixgbe, pci_id_ixgbe_map);
7371 PMD_REGISTER_DRIVER(rte_ixgbevf_driver, ixgbevf);
7372 DRIVER_REGISTER_PCI_TABLE(ixgbevf, pci_id_ixgbevf_map);
7373