4a6e5c32eca887c4e3e85c49a2a103614931b97d
[dpdk.git] / drivers / net / ixgbe / ixgbe_ethdev.c
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2017 Intel Corporation
3  */
4
5 #include <sys/queue.h>
6 #include <stdio.h>
7 #include <errno.h>
8 #include <stdint.h>
9 #include <string.h>
10 #include <unistd.h>
11 #include <stdarg.h>
12 #include <inttypes.h>
13 #include <netinet/in.h>
14 #include <rte_string_fns.h>
15 #include <rte_byteorder.h>
16 #include <rte_common.h>
17 #include <rte_cycles.h>
18
19 #include <rte_interrupts.h>
20 #include <rte_log.h>
21 #include <rte_debug.h>
22 #include <rte_pci.h>
23 #include <rte_bus_pci.h>
24 #include <rte_branch_prediction.h>
25 #include <rte_memory.h>
26 #include <rte_kvargs.h>
27 #include <rte_eal.h>
28 #include <rte_alarm.h>
29 #include <rte_ether.h>
30 #include <rte_ethdev_driver.h>
31 #include <rte_ethdev_pci.h>
32 #include <rte_malloc.h>
33 #include <rte_random.h>
34 #include <rte_dev.h>
35 #include <rte_hash_crc.h>
36 #ifdef RTE_LIBRTE_SECURITY
37 #include <rte_security_driver.h>
38 #endif
39
40 #include "ixgbe_logs.h"
41 #include "base/ixgbe_api.h"
42 #include "base/ixgbe_vf.h"
43 #include "base/ixgbe_common.h"
44 #include "ixgbe_ethdev.h"
45 #include "ixgbe_bypass.h"
46 #include "ixgbe_rxtx.h"
47 #include "base/ixgbe_type.h"
48 #include "base/ixgbe_phy.h"
49 #include "ixgbe_regs.h"
50
51 /*
52  * High threshold controlling when to start sending XOFF frames. Must be at
53  * least 8 bytes less than receive packet buffer size. This value is in units
54  * of 1024 bytes.
55  */
56 #define IXGBE_FC_HI    0x80
57
58 /*
59  * Low threshold controlling when to start sending XON frames. This value is
60  * in units of 1024 bytes.
61  */
62 #define IXGBE_FC_LO    0x40
63
64 /* Timer value included in XOFF frames. */
65 #define IXGBE_FC_PAUSE 0x680
66
67 /*Default value of Max Rx Queue*/
68 #define IXGBE_MAX_RX_QUEUE_NUM 128
69
70 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
71 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
72 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
73
74 #define IXGBE_MMW_SIZE_DEFAULT        0x4
75 #define IXGBE_MMW_SIZE_JUMBO_FRAME    0x14
76 #define IXGBE_MAX_RING_DESC           4096 /* replicate define from rxtx */
77
78 /*
79  *  Default values for RX/TX configuration
80  */
81 #define IXGBE_DEFAULT_RX_FREE_THRESH  32
82 #define IXGBE_DEFAULT_RX_PTHRESH      8
83 #define IXGBE_DEFAULT_RX_HTHRESH      8
84 #define IXGBE_DEFAULT_RX_WTHRESH      0
85
86 #define IXGBE_DEFAULT_TX_FREE_THRESH  32
87 #define IXGBE_DEFAULT_TX_PTHRESH      32
88 #define IXGBE_DEFAULT_TX_HTHRESH      0
89 #define IXGBE_DEFAULT_TX_WTHRESH      0
90 #define IXGBE_DEFAULT_TX_RSBIT_THRESH 32
91
92 /* Bit shift and mask */
93 #define IXGBE_4_BIT_WIDTH  (CHAR_BIT / 2)
94 #define IXGBE_4_BIT_MASK   RTE_LEN2MASK(IXGBE_4_BIT_WIDTH, uint8_t)
95 #define IXGBE_8_BIT_WIDTH  CHAR_BIT
96 #define IXGBE_8_BIT_MASK   UINT8_MAX
97
98 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
99
100 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
101
102 /* Additional timesync values. */
103 #define NSEC_PER_SEC             1000000000L
104 #define IXGBE_INCVAL_10GB        0x66666666
105 #define IXGBE_INCVAL_1GB         0x40000000
106 #define IXGBE_INCVAL_100         0x50000000
107 #define IXGBE_INCVAL_SHIFT_10GB  28
108 #define IXGBE_INCVAL_SHIFT_1GB   24
109 #define IXGBE_INCVAL_SHIFT_100   21
110 #define IXGBE_INCVAL_SHIFT_82599 7
111 #define IXGBE_INCPER_SHIFT_82599 24
112
113 #define IXGBE_CYCLECOUNTER_MASK   0xffffffffffffffffULL
114
115 #define IXGBE_VT_CTL_POOLING_MODE_MASK         0x00030000
116 #define IXGBE_VT_CTL_POOLING_MODE_ETAG         0x00010000
117 #define IXGBE_ETAG_ETYPE                       0x00005084
118 #define IXGBE_ETAG_ETYPE_MASK                  0x0000ffff
119 #define IXGBE_ETAG_ETYPE_VALID                 0x80000000
120 #define IXGBE_RAH_ADTYPE                       0x40000000
121 #define IXGBE_RAL_ETAG_FILTER_MASK             0x00003fff
122 #define IXGBE_VMVIR_TAGA_MASK                  0x18000000
123 #define IXGBE_VMVIR_TAGA_ETAG_INSERT           0x08000000
124 #define IXGBE_VMTIR(_i) (0x00017000 + ((_i) * 4)) /* 64 of these (0-63) */
125 #define IXGBE_QDE_STRIP_TAG                    0x00000004
126 #define IXGBE_VTEICR_MASK                      0x07
127
128 #define IXGBE_EXVET_VET_EXT_SHIFT              16
129 #define IXGBE_DMATXCTL_VT_MASK                 0xFFFF0000
130
131 #define IXGBEVF_DEVARG_PFLINK_FULLCHK           "pflink_fullchk"
132
133 static const char * const ixgbevf_valid_arguments[] = {
134         IXGBEVF_DEVARG_PFLINK_FULLCHK,
135         NULL
136 };
137
138 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params);
139 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
140 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev);
141 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev);
142 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev);
143 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev);
144 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev);
145 static int  ixgbe_dev_configure(struct rte_eth_dev *dev);
146 static int  ixgbe_dev_start(struct rte_eth_dev *dev);
147 static void ixgbe_dev_stop(struct rte_eth_dev *dev);
148 static int  ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
149 static int  ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
150 static void ixgbe_dev_close(struct rte_eth_dev *dev);
151 static int  ixgbe_dev_reset(struct rte_eth_dev *dev);
152 static void ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
153 static void ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
154 static void ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
155 static void ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
156 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
157                                 int wait_to_complete);
158 static int ixgbe_dev_stats_get(struct rte_eth_dev *dev,
159                                 struct rte_eth_stats *stats);
160 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
161                                 struct rte_eth_xstat *xstats, unsigned n);
162 static int ixgbevf_dev_xstats_get(struct rte_eth_dev *dev,
163                                   struct rte_eth_xstat *xstats, unsigned n);
164 static int
165 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
166                 uint64_t *values, unsigned int n);
167 static void ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
168 static void ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
169 static int ixgbe_dev_xstats_get_names(struct rte_eth_dev *dev,
170         struct rte_eth_xstat_name *xstats_names,
171         unsigned int size);
172 static int ixgbevf_dev_xstats_get_names(struct rte_eth_dev *dev,
173         struct rte_eth_xstat_name *xstats_names, unsigned limit);
174 static int ixgbe_dev_xstats_get_names_by_id(
175         struct rte_eth_dev *dev,
176         struct rte_eth_xstat_name *xstats_names,
177         const uint64_t *ids,
178         unsigned int limit);
179 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
180                                              uint16_t queue_id,
181                                              uint8_t stat_idx,
182                                              uint8_t is_rx);
183 static int ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
184                                  size_t fw_size);
185 static void ixgbe_dev_info_get(struct rte_eth_dev *dev,
186                                struct rte_eth_dev_info *dev_info);
187 static const uint32_t *ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev);
188 static void ixgbevf_dev_info_get(struct rte_eth_dev *dev,
189                                  struct rte_eth_dev_info *dev_info);
190 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
191
192 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
193                 uint16_t vlan_id, int on);
194 static int ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
195                                enum rte_vlan_type vlan_type,
196                                uint16_t tpid_id);
197 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
198                 uint16_t queue, bool on);
199 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
200                 int on);
201 static void ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev,
202                                                   int mask);
203 static int ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask);
204 static int ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
205 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
206 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
207 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
208 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
209
210 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
211 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
212 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
213                                struct rte_eth_fc_conf *fc_conf);
214 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
215                                struct rte_eth_fc_conf *fc_conf);
216 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
217                 struct rte_eth_pfc_conf *pfc_conf);
218 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
219                         struct rte_eth_rss_reta_entry64 *reta_conf,
220                         uint16_t reta_size);
221 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
222                         struct rte_eth_rss_reta_entry64 *reta_conf,
223                         uint16_t reta_size);
224 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
225 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
226 static int ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev);
227 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
228 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
229 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev);
230 static void ixgbe_dev_interrupt_handler(void *param);
231 static void ixgbe_dev_interrupt_delayed_handler(void *param);
232 static void ixgbe_dev_setup_link_alarm_handler(void *param);
233
234 static int ixgbe_add_rar(struct rte_eth_dev *dev,
235                         struct rte_ether_addr *mac_addr,
236                         uint32_t index, uint32_t pool);
237 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
238 static int ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
239                                            struct rte_ether_addr *mac_addr);
240 static void ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config);
241 static bool is_device_supported(struct rte_eth_dev *dev,
242                                 struct rte_pci_driver *drv);
243
244 /* For Virtual Function support */
245 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
246 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
247 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
248 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
249 static int ixgbevf_dev_link_update(struct rte_eth_dev *dev,
250                                    int wait_to_complete);
251 static void ixgbevf_dev_stop(struct rte_eth_dev *dev);
252 static void ixgbevf_dev_close(struct rte_eth_dev *dev);
253 static int  ixgbevf_dev_reset(struct rte_eth_dev *dev);
254 static void ixgbevf_intr_disable(struct rte_eth_dev *dev);
255 static void ixgbevf_intr_enable(struct rte_eth_dev *dev);
256 static int ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
257                 struct rte_eth_stats *stats);
258 static void ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
259 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
260                 uint16_t vlan_id, int on);
261 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
262                 uint16_t queue, int on);
263 static int ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask);
264 static int ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
265 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
266 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
267                                             uint16_t queue_id);
268 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
269                                              uint16_t queue_id);
270 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
271                                  uint8_t queue, uint8_t msix_vector);
272 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
273 static void ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev);
274 static void ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev);
275 static void ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev);
276 static void ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev);
277
278 /* For Eth VMDQ APIs support */
279 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
280                 rte_ether_addr * mac_addr, uint8_t on);
281 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on);
282 static int ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
283                 struct rte_eth_mirror_conf *mirror_conf,
284                 uint8_t rule_id, uint8_t on);
285 static int ixgbe_mirror_rule_reset(struct rte_eth_dev *dev,
286                 uint8_t rule_id);
287 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
288                                           uint16_t queue_id);
289 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
290                                            uint16_t queue_id);
291 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
292                                uint8_t queue, uint8_t msix_vector);
293 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
294
295 static int ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
296                                 struct rte_ether_addr *mac_addr,
297                                 uint32_t index, uint32_t pool);
298 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
299 static int ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
300                                              struct rte_ether_addr *mac_addr);
301 static int ixgbe_syn_filter_get(struct rte_eth_dev *dev,
302                         struct rte_eth_syn_filter *filter);
303 static int ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
304                         enum rte_filter_op filter_op,
305                         void *arg);
306 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
307                         struct ixgbe_5tuple_filter *filter);
308 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
309                         struct ixgbe_5tuple_filter *filter);
310 static int ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
311                                 enum rte_filter_op filter_op,
312                                 void *arg);
313 static int ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
314                         struct rte_eth_ntuple_filter *filter);
315 static int ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
316                                 enum rte_filter_op filter_op,
317                                 void *arg);
318 static int ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
319                         struct rte_eth_ethertype_filter *filter);
320 static int ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
321                      enum rte_filter_type filter_type,
322                      enum rte_filter_op filter_op,
323                      void *arg);
324 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
325
326 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
327                                       struct rte_ether_addr *mc_addr_set,
328                                       uint32_t nb_mc_addr);
329 static int ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
330                                    struct rte_eth_dcb_info *dcb_info);
331
332 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
333 static int ixgbe_get_regs(struct rte_eth_dev *dev,
334                             struct rte_dev_reg_info *regs);
335 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
336 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
337                                 struct rte_dev_eeprom_info *eeprom);
338 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
339                                 struct rte_dev_eeprom_info *eeprom);
340
341 static int ixgbe_get_module_info(struct rte_eth_dev *dev,
342                                  struct rte_eth_dev_module_info *modinfo);
343 static int ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
344                                    struct rte_dev_eeprom_info *info);
345
346 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
347 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
348                                 struct rte_dev_reg_info *regs);
349
350 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
351 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
352 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
353                                             struct timespec *timestamp,
354                                             uint32_t flags);
355 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
356                                             struct timespec *timestamp);
357 static int ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
358 static int ixgbe_timesync_read_time(struct rte_eth_dev *dev,
359                                    struct timespec *timestamp);
360 static int ixgbe_timesync_write_time(struct rte_eth_dev *dev,
361                                    const struct timespec *timestamp);
362 static void ixgbevf_dev_interrupt_handler(void *param);
363
364 static int ixgbe_dev_l2_tunnel_eth_type_conf
365         (struct rte_eth_dev *dev, struct rte_eth_l2_tunnel_conf *l2_tunnel);
366 static int ixgbe_dev_l2_tunnel_offload_set
367         (struct rte_eth_dev *dev,
368          struct rte_eth_l2_tunnel_conf *l2_tunnel,
369          uint32_t mask,
370          uint8_t en);
371 static int ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
372                                              enum rte_filter_op filter_op,
373                                              void *arg);
374
375 static int ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
376                                          struct rte_eth_udp_tunnel *udp_tunnel);
377 static int ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
378                                          struct rte_eth_udp_tunnel *udp_tunnel);
379 static int ixgbe_filter_restore(struct rte_eth_dev *dev);
380 static void ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev);
381
382 /*
383  * Define VF Stats MACRO for Non "cleared on read" register
384  */
385 #define UPDATE_VF_STAT(reg, last, cur)                          \
386 {                                                               \
387         uint32_t latest = IXGBE_READ_REG(hw, reg);              \
388         cur += (latest - last) & UINT_MAX;                      \
389         last = latest;                                          \
390 }
391
392 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
393 {                                                                \
394         u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
395         u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
396         u64 latest = ((new_msb << 32) | new_lsb);                \
397         cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
398         last = latest;                                           \
399 }
400
401 #define IXGBE_SET_HWSTRIP(h, q) do {\
402                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
403                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
404                 (h)->bitmap[idx] |= 1 << bit;\
405         } while (0)
406
407 #define IXGBE_CLEAR_HWSTRIP(h, q) do {\
408                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
409                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
410                 (h)->bitmap[idx] &= ~(1 << bit);\
411         } while (0)
412
413 #define IXGBE_GET_HWSTRIP(h, q, r) do {\
414                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
415                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
416                 (r) = (h)->bitmap[idx] >> bit & 1;\
417         } while (0)
418
419 int ixgbe_logtype_init;
420 int ixgbe_logtype_driver;
421
422 /*
423  * The set of PCI devices this driver supports
424  */
425 static const struct rte_pci_id pci_id_ixgbe_map[] = {
426         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598) },
427         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_BX) },
428         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT) },
429         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT) },
430         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT) },
431         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2) },
432         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM) },
433         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4) },
434         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT) },
435         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT) },
436         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM) },
437         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR) },
438         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4) },
439         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ) },
440         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KR) },
441         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE) },
442         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4) },
443         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP) },
444         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE) },
445         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE) },
446         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_EM) },
447         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2) },
448         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP) },
449         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP) },
450         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP) },
451         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM) },
452         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM) },
453         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T) },
454         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1) },
455         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP) },
456         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T) },
457         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_1G_T) },
458         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T) },
459         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1) },
460         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR) },
461         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR_L) },
462         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP_N) },
463         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII) },
464         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII_L) },
465         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_10G_T) },
466         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP) },
467         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP_N) },
468         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP) },
469         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T) },
470         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T_L) },
471         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4) },
472         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR) },
473         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_XFI) },
474 #ifdef RTE_LIBRTE_IXGBE_BYPASS
475         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BYPASS) },
476 #endif
477         { .vendor_id = 0, /* sentinel */ },
478 };
479
480 /*
481  * The set of PCI devices this driver supports (for 82599 VF)
482  */
483 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
484         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF) },
485         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF_HV) },
486         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF) },
487         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF_HV) },
488         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF_HV) },
489         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF) },
490         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF) },
491         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF_HV) },
492         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF) },
493         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF_HV) },
494         { .vendor_id = 0, /* sentinel */ },
495 };
496
497 static const struct rte_eth_desc_lim rx_desc_lim = {
498         .nb_max = IXGBE_MAX_RING_DESC,
499         .nb_min = IXGBE_MIN_RING_DESC,
500         .nb_align = IXGBE_RXD_ALIGN,
501 };
502
503 static const struct rte_eth_desc_lim tx_desc_lim = {
504         .nb_max = IXGBE_MAX_RING_DESC,
505         .nb_min = IXGBE_MIN_RING_DESC,
506         .nb_align = IXGBE_TXD_ALIGN,
507         .nb_seg_max = IXGBE_TX_MAX_SEG,
508         .nb_mtu_seg_max = IXGBE_TX_MAX_SEG,
509 };
510
511 static const struct eth_dev_ops ixgbe_eth_dev_ops = {
512         .dev_configure        = ixgbe_dev_configure,
513         .dev_start            = ixgbe_dev_start,
514         .dev_stop             = ixgbe_dev_stop,
515         .dev_set_link_up    = ixgbe_dev_set_link_up,
516         .dev_set_link_down  = ixgbe_dev_set_link_down,
517         .dev_close            = ixgbe_dev_close,
518         .dev_reset            = ixgbe_dev_reset,
519         .promiscuous_enable   = ixgbe_dev_promiscuous_enable,
520         .promiscuous_disable  = ixgbe_dev_promiscuous_disable,
521         .allmulticast_enable  = ixgbe_dev_allmulticast_enable,
522         .allmulticast_disable = ixgbe_dev_allmulticast_disable,
523         .link_update          = ixgbe_dev_link_update,
524         .stats_get            = ixgbe_dev_stats_get,
525         .xstats_get           = ixgbe_dev_xstats_get,
526         .xstats_get_by_id     = ixgbe_dev_xstats_get_by_id,
527         .stats_reset          = ixgbe_dev_stats_reset,
528         .xstats_reset         = ixgbe_dev_xstats_reset,
529         .xstats_get_names     = ixgbe_dev_xstats_get_names,
530         .xstats_get_names_by_id = ixgbe_dev_xstats_get_names_by_id,
531         .queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
532         .fw_version_get       = ixgbe_fw_version_get,
533         .dev_infos_get        = ixgbe_dev_info_get,
534         .dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
535         .mtu_set              = ixgbe_dev_mtu_set,
536         .vlan_filter_set      = ixgbe_vlan_filter_set,
537         .vlan_tpid_set        = ixgbe_vlan_tpid_set,
538         .vlan_offload_set     = ixgbe_vlan_offload_set,
539         .vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
540         .rx_queue_start       = ixgbe_dev_rx_queue_start,
541         .rx_queue_stop        = ixgbe_dev_rx_queue_stop,
542         .tx_queue_start       = ixgbe_dev_tx_queue_start,
543         .tx_queue_stop        = ixgbe_dev_tx_queue_stop,
544         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
545         .rx_queue_intr_enable = ixgbe_dev_rx_queue_intr_enable,
546         .rx_queue_intr_disable = ixgbe_dev_rx_queue_intr_disable,
547         .rx_queue_release     = ixgbe_dev_rx_queue_release,
548         .rx_queue_count       = ixgbe_dev_rx_queue_count,
549         .rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
550         .rx_descriptor_status = ixgbe_dev_rx_descriptor_status,
551         .tx_descriptor_status = ixgbe_dev_tx_descriptor_status,
552         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
553         .tx_queue_release     = ixgbe_dev_tx_queue_release,
554         .dev_led_on           = ixgbe_dev_led_on,
555         .dev_led_off          = ixgbe_dev_led_off,
556         .flow_ctrl_get        = ixgbe_flow_ctrl_get,
557         .flow_ctrl_set        = ixgbe_flow_ctrl_set,
558         .priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
559         .mac_addr_add         = ixgbe_add_rar,
560         .mac_addr_remove      = ixgbe_remove_rar,
561         .mac_addr_set         = ixgbe_set_default_mac_addr,
562         .uc_hash_table_set    = ixgbe_uc_hash_table_set,
563         .uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
564         .mirror_rule_set      = ixgbe_mirror_rule_set,
565         .mirror_rule_reset    = ixgbe_mirror_rule_reset,
566         .set_queue_rate_limit = ixgbe_set_queue_rate_limit,
567         .reta_update          = ixgbe_dev_rss_reta_update,
568         .reta_query           = ixgbe_dev_rss_reta_query,
569         .rss_hash_update      = ixgbe_dev_rss_hash_update,
570         .rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
571         .filter_ctrl          = ixgbe_dev_filter_ctrl,
572         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
573         .rxq_info_get         = ixgbe_rxq_info_get,
574         .txq_info_get         = ixgbe_txq_info_get,
575         .timesync_enable      = ixgbe_timesync_enable,
576         .timesync_disable     = ixgbe_timesync_disable,
577         .timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
578         .timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
579         .get_reg              = ixgbe_get_regs,
580         .get_eeprom_length    = ixgbe_get_eeprom_length,
581         .get_eeprom           = ixgbe_get_eeprom,
582         .set_eeprom           = ixgbe_set_eeprom,
583         .get_module_info      = ixgbe_get_module_info,
584         .get_module_eeprom    = ixgbe_get_module_eeprom,
585         .get_dcb_info         = ixgbe_dev_get_dcb_info,
586         .timesync_adjust_time = ixgbe_timesync_adjust_time,
587         .timesync_read_time   = ixgbe_timesync_read_time,
588         .timesync_write_time  = ixgbe_timesync_write_time,
589         .l2_tunnel_eth_type_conf = ixgbe_dev_l2_tunnel_eth_type_conf,
590         .l2_tunnel_offload_set   = ixgbe_dev_l2_tunnel_offload_set,
591         .udp_tunnel_port_add  = ixgbe_dev_udp_tunnel_port_add,
592         .udp_tunnel_port_del  = ixgbe_dev_udp_tunnel_port_del,
593         .tm_ops_get           = ixgbe_tm_ops_get,
594 };
595
596 /*
597  * dev_ops for virtual function, bare necessities for basic vf
598  * operation have been implemented
599  */
600 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
601         .dev_configure        = ixgbevf_dev_configure,
602         .dev_start            = ixgbevf_dev_start,
603         .dev_stop             = ixgbevf_dev_stop,
604         .link_update          = ixgbevf_dev_link_update,
605         .stats_get            = ixgbevf_dev_stats_get,
606         .xstats_get           = ixgbevf_dev_xstats_get,
607         .stats_reset          = ixgbevf_dev_stats_reset,
608         .xstats_reset         = ixgbevf_dev_stats_reset,
609         .xstats_get_names     = ixgbevf_dev_xstats_get_names,
610         .dev_close            = ixgbevf_dev_close,
611         .dev_reset            = ixgbevf_dev_reset,
612         .promiscuous_enable   = ixgbevf_dev_promiscuous_enable,
613         .promiscuous_disable  = ixgbevf_dev_promiscuous_disable,
614         .allmulticast_enable  = ixgbevf_dev_allmulticast_enable,
615         .allmulticast_disable = ixgbevf_dev_allmulticast_disable,
616         .dev_infos_get        = ixgbevf_dev_info_get,
617         .dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
618         .mtu_set              = ixgbevf_dev_set_mtu,
619         .vlan_filter_set      = ixgbevf_vlan_filter_set,
620         .vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
621         .vlan_offload_set     = ixgbevf_vlan_offload_set,
622         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
623         .rx_queue_release     = ixgbe_dev_rx_queue_release,
624         .rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
625         .rx_descriptor_status = ixgbe_dev_rx_descriptor_status,
626         .tx_descriptor_status = ixgbe_dev_tx_descriptor_status,
627         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
628         .tx_queue_release     = ixgbe_dev_tx_queue_release,
629         .rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
630         .rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
631         .mac_addr_add         = ixgbevf_add_mac_addr,
632         .mac_addr_remove      = ixgbevf_remove_mac_addr,
633         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
634         .rxq_info_get         = ixgbe_rxq_info_get,
635         .txq_info_get         = ixgbe_txq_info_get,
636         .mac_addr_set         = ixgbevf_set_default_mac_addr,
637         .get_reg              = ixgbevf_get_regs,
638         .reta_update          = ixgbe_dev_rss_reta_update,
639         .reta_query           = ixgbe_dev_rss_reta_query,
640         .rss_hash_update      = ixgbe_dev_rss_hash_update,
641         .rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
642 };
643
644 /* store statistics names and its offset in stats structure */
645 struct rte_ixgbe_xstats_name_off {
646         char name[RTE_ETH_XSTATS_NAME_SIZE];
647         unsigned offset;
648 };
649
650 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
651         {"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
652         {"rx_illegal_byte_errors", offsetof(struct ixgbe_hw_stats, illerrc)},
653         {"rx_error_bytes", offsetof(struct ixgbe_hw_stats, errbc)},
654         {"mac_local_errors", offsetof(struct ixgbe_hw_stats, mlfc)},
655         {"mac_remote_errors", offsetof(struct ixgbe_hw_stats, mrfc)},
656         {"rx_length_errors", offsetof(struct ixgbe_hw_stats, rlec)},
657         {"tx_xon_packets", offsetof(struct ixgbe_hw_stats, lxontxc)},
658         {"rx_xon_packets", offsetof(struct ixgbe_hw_stats, lxonrxc)},
659         {"tx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxofftxc)},
660         {"rx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
661         {"rx_size_64_packets", offsetof(struct ixgbe_hw_stats, prc64)},
662         {"rx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, prc127)},
663         {"rx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, prc255)},
664         {"rx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, prc511)},
665         {"rx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
666                 prc1023)},
667         {"rx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
668                 prc1522)},
669         {"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
670         {"rx_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
671         {"rx_fragment_errors", offsetof(struct ixgbe_hw_stats, rfc)},
672         {"rx_undersize_errors", offsetof(struct ixgbe_hw_stats, ruc)},
673         {"rx_oversize_errors", offsetof(struct ixgbe_hw_stats, roc)},
674         {"rx_jabber_errors", offsetof(struct ixgbe_hw_stats, rjc)},
675         {"rx_management_packets", offsetof(struct ixgbe_hw_stats, mngprc)},
676         {"rx_management_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
677         {"tx_management_packets", offsetof(struct ixgbe_hw_stats, mngptc)},
678         {"rx_total_packets", offsetof(struct ixgbe_hw_stats, tpr)},
679         {"rx_total_bytes", offsetof(struct ixgbe_hw_stats, tor)},
680         {"tx_total_packets", offsetof(struct ixgbe_hw_stats, tpt)},
681         {"tx_size_64_packets", offsetof(struct ixgbe_hw_stats, ptc64)},
682         {"tx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, ptc127)},
683         {"tx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, ptc255)},
684         {"tx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, ptc511)},
685         {"tx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
686                 ptc1023)},
687         {"tx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
688                 ptc1522)},
689         {"tx_multicast_packets", offsetof(struct ixgbe_hw_stats, mptc)},
690         {"tx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bptc)},
691         {"rx_mac_short_packet_dropped", offsetof(struct ixgbe_hw_stats, mspdc)},
692         {"rx_l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
693
694         {"flow_director_added_filters", offsetof(struct ixgbe_hw_stats,
695                 fdirustat_add)},
696         {"flow_director_removed_filters", offsetof(struct ixgbe_hw_stats,
697                 fdirustat_remove)},
698         {"flow_director_filter_add_errors", offsetof(struct ixgbe_hw_stats,
699                 fdirfstat_fadd)},
700         {"flow_director_filter_remove_errors", offsetof(struct ixgbe_hw_stats,
701                 fdirfstat_fremove)},
702         {"flow_director_matched_filters", offsetof(struct ixgbe_hw_stats,
703                 fdirmatch)},
704         {"flow_director_missed_filters", offsetof(struct ixgbe_hw_stats,
705                 fdirmiss)},
706
707         {"rx_fcoe_crc_errors", offsetof(struct ixgbe_hw_stats, fccrc)},
708         {"rx_fcoe_dropped", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
709         {"rx_fcoe_mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats,
710                 fclast)},
711         {"rx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeprc)},
712         {"tx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeptc)},
713         {"rx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwrc)},
714         {"tx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwtc)},
715         {"rx_fcoe_no_direct_data_placement", offsetof(struct ixgbe_hw_stats,
716                 fcoe_noddp)},
717         {"rx_fcoe_no_direct_data_placement_ext_buff",
718                 offsetof(struct ixgbe_hw_stats, fcoe_noddp_ext_buff)},
719
720         {"tx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
721                 lxontxc)},
722         {"rx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
723                 lxonrxc)},
724         {"tx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
725                 lxofftxc)},
726         {"rx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
727                 lxoffrxc)},
728         {"rx_total_missed_packets", offsetof(struct ixgbe_hw_stats, mpctotal)},
729 };
730
731 #define IXGBE_NB_HW_STATS (sizeof(rte_ixgbe_stats_strings) / \
732                            sizeof(rte_ixgbe_stats_strings[0]))
733
734 /* MACsec statistics */
735 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_macsec_strings[] = {
736         {"out_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
737                 out_pkts_untagged)},
738         {"out_pkts_encrypted", offsetof(struct ixgbe_macsec_stats,
739                 out_pkts_encrypted)},
740         {"out_pkts_protected", offsetof(struct ixgbe_macsec_stats,
741                 out_pkts_protected)},
742         {"out_octets_encrypted", offsetof(struct ixgbe_macsec_stats,
743                 out_octets_encrypted)},
744         {"out_octets_protected", offsetof(struct ixgbe_macsec_stats,
745                 out_octets_protected)},
746         {"in_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
747                 in_pkts_untagged)},
748         {"in_pkts_badtag", offsetof(struct ixgbe_macsec_stats,
749                 in_pkts_badtag)},
750         {"in_pkts_nosci", offsetof(struct ixgbe_macsec_stats,
751                 in_pkts_nosci)},
752         {"in_pkts_unknownsci", offsetof(struct ixgbe_macsec_stats,
753                 in_pkts_unknownsci)},
754         {"in_octets_decrypted", offsetof(struct ixgbe_macsec_stats,
755                 in_octets_decrypted)},
756         {"in_octets_validated", offsetof(struct ixgbe_macsec_stats,
757                 in_octets_validated)},
758         {"in_pkts_unchecked", offsetof(struct ixgbe_macsec_stats,
759                 in_pkts_unchecked)},
760         {"in_pkts_delayed", offsetof(struct ixgbe_macsec_stats,
761                 in_pkts_delayed)},
762         {"in_pkts_late", offsetof(struct ixgbe_macsec_stats,
763                 in_pkts_late)},
764         {"in_pkts_ok", offsetof(struct ixgbe_macsec_stats,
765                 in_pkts_ok)},
766         {"in_pkts_invalid", offsetof(struct ixgbe_macsec_stats,
767                 in_pkts_invalid)},
768         {"in_pkts_notvalid", offsetof(struct ixgbe_macsec_stats,
769                 in_pkts_notvalid)},
770         {"in_pkts_unusedsa", offsetof(struct ixgbe_macsec_stats,
771                 in_pkts_unusedsa)},
772         {"in_pkts_notusingsa", offsetof(struct ixgbe_macsec_stats,
773                 in_pkts_notusingsa)},
774 };
775
776 #define IXGBE_NB_MACSEC_STATS (sizeof(rte_ixgbe_macsec_strings) / \
777                            sizeof(rte_ixgbe_macsec_strings[0]))
778
779 /* Per-queue statistics */
780 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_rxq_strings[] = {
781         {"mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats, rnbc)},
782         {"dropped", offsetof(struct ixgbe_hw_stats, mpc)},
783         {"xon_packets", offsetof(struct ixgbe_hw_stats, pxonrxc)},
784         {"xoff_packets", offsetof(struct ixgbe_hw_stats, pxoffrxc)},
785 };
786
787 #define IXGBE_NB_RXQ_PRIO_STATS (sizeof(rte_ixgbe_rxq_strings) / \
788                            sizeof(rte_ixgbe_rxq_strings[0]))
789 #define IXGBE_NB_RXQ_PRIO_VALUES 8
790
791 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_txq_strings[] = {
792         {"xon_packets", offsetof(struct ixgbe_hw_stats, pxontxc)},
793         {"xoff_packets", offsetof(struct ixgbe_hw_stats, pxofftxc)},
794         {"xon_to_xoff_packets", offsetof(struct ixgbe_hw_stats,
795                 pxon2offc)},
796 };
797
798 #define IXGBE_NB_TXQ_PRIO_STATS (sizeof(rte_ixgbe_txq_strings) / \
799                            sizeof(rte_ixgbe_txq_strings[0]))
800 #define IXGBE_NB_TXQ_PRIO_VALUES 8
801
802 static const struct rte_ixgbe_xstats_name_off rte_ixgbevf_stats_strings[] = {
803         {"rx_multicast_packets", offsetof(struct ixgbevf_hw_stats, vfmprc)},
804 };
805
806 #define IXGBEVF_NB_XSTATS (sizeof(rte_ixgbevf_stats_strings) /  \
807                 sizeof(rte_ixgbevf_stats_strings[0]))
808
809 /*
810  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
811  */
812 static inline int
813 ixgbe_is_sfp(struct ixgbe_hw *hw)
814 {
815         switch (hw->phy.type) {
816         case ixgbe_phy_sfp_avago:
817         case ixgbe_phy_sfp_ftl:
818         case ixgbe_phy_sfp_intel:
819         case ixgbe_phy_sfp_unknown:
820         case ixgbe_phy_sfp_passive_tyco:
821         case ixgbe_phy_sfp_passive_unknown:
822                 return 1;
823         default:
824                 return 0;
825         }
826 }
827
828 static inline int32_t
829 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
830 {
831         uint32_t ctrl_ext;
832         int32_t status;
833
834         status = ixgbe_reset_hw(hw);
835
836         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
837         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
838         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
839         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
840         IXGBE_WRITE_FLUSH(hw);
841
842         if (status == IXGBE_ERR_SFP_NOT_PRESENT)
843                 status = IXGBE_SUCCESS;
844         return status;
845 }
846
847 static inline void
848 ixgbe_enable_intr(struct rte_eth_dev *dev)
849 {
850         struct ixgbe_interrupt *intr =
851                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
852         struct ixgbe_hw *hw =
853                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
854
855         IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
856         IXGBE_WRITE_FLUSH(hw);
857 }
858
859 /*
860  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
861  */
862 static void
863 ixgbe_disable_intr(struct ixgbe_hw *hw)
864 {
865         PMD_INIT_FUNC_TRACE();
866
867         if (hw->mac.type == ixgbe_mac_82598EB) {
868                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
869         } else {
870                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
871                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
872                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
873         }
874         IXGBE_WRITE_FLUSH(hw);
875 }
876
877 /*
878  * This function resets queue statistics mapping registers.
879  * From Niantic datasheet, Initialization of Statistics section:
880  * "...if software requires the queue counters, the RQSMR and TQSM registers
881  * must be re-programmed following a device reset.
882  */
883 static void
884 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
885 {
886         uint32_t i;
887
888         for (i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
889                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
890                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
891         }
892 }
893
894
895 static int
896 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
897                                   uint16_t queue_id,
898                                   uint8_t stat_idx,
899                                   uint8_t is_rx)
900 {
901 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
902 #define NB_QMAP_FIELDS_PER_QSM_REG 4
903 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
904
905         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
906         struct ixgbe_stat_mapping_registers *stat_mappings =
907                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
908         uint32_t qsmr_mask = 0;
909         uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
910         uint32_t q_map;
911         uint8_t n, offset;
912
913         if ((hw->mac.type != ixgbe_mac_82599EB) &&
914                 (hw->mac.type != ixgbe_mac_X540) &&
915                 (hw->mac.type != ixgbe_mac_X550) &&
916                 (hw->mac.type != ixgbe_mac_X550EM_x) &&
917                 (hw->mac.type != ixgbe_mac_X550EM_a))
918                 return -ENOSYS;
919
920         PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
921                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
922                      queue_id, stat_idx);
923
924         n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
925         if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
926                 PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
927                 return -EIO;
928         }
929         offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
930
931         /* Now clear any previous stat_idx set */
932         clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
933         if (!is_rx)
934                 stat_mappings->tqsm[n] &= ~clearing_mask;
935         else
936                 stat_mappings->rqsmr[n] &= ~clearing_mask;
937
938         q_map = (uint32_t)stat_idx;
939         q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
940         qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
941         if (!is_rx)
942                 stat_mappings->tqsm[n] |= qsmr_mask;
943         else
944                 stat_mappings->rqsmr[n] |= qsmr_mask;
945
946         PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
947                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
948                      queue_id, stat_idx);
949         PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
950                      is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
951
952         /* Now write the mapping in the appropriate register */
953         if (is_rx) {
954                 PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
955                              stat_mappings->rqsmr[n], n);
956                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
957         } else {
958                 PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
959                              stat_mappings->tqsm[n], n);
960                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
961         }
962         return 0;
963 }
964
965 static void
966 ixgbe_restore_statistics_mapping(struct rte_eth_dev *dev)
967 {
968         struct ixgbe_stat_mapping_registers *stat_mappings =
969                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
970         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
971         int i;
972
973         /* write whatever was in stat mapping table to the NIC */
974         for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
975                 /* rx */
976                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
977
978                 /* tx */
979                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
980         }
981 }
982
983 static void
984 ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config)
985 {
986         uint8_t i;
987         struct ixgbe_dcb_tc_config *tc;
988         uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
989
990         dcb_config->num_tcs.pg_tcs = dcb_max_tc;
991         dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
992         for (i = 0; i < dcb_max_tc; i++) {
993                 tc = &dcb_config->tc_config[i];
994                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
995                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
996                                  (uint8_t)(100/dcb_max_tc + (i & 1));
997                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
998                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
999                                  (uint8_t)(100/dcb_max_tc + (i & 1));
1000                 tc->pfc = ixgbe_dcb_pfc_disabled;
1001         }
1002
1003         /* Initialize default user to priority mapping, UPx->TC0 */
1004         tc = &dcb_config->tc_config[0];
1005         tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
1006         tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
1007         for (i = 0; i < IXGBE_DCB_MAX_BW_GROUP; i++) {
1008                 dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
1009                 dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
1010         }
1011         dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
1012         dcb_config->pfc_mode_enable = false;
1013         dcb_config->vt_mode = true;
1014         dcb_config->round_robin_enable = false;
1015         /* support all DCB capabilities in 82599 */
1016         dcb_config->support.capabilities = 0xFF;
1017
1018         /*we only support 4 Tcs for X540, X550 */
1019         if (hw->mac.type == ixgbe_mac_X540 ||
1020                 hw->mac.type == ixgbe_mac_X550 ||
1021                 hw->mac.type == ixgbe_mac_X550EM_x ||
1022                 hw->mac.type == ixgbe_mac_X550EM_a) {
1023                 dcb_config->num_tcs.pg_tcs = 4;
1024                 dcb_config->num_tcs.pfc_tcs = 4;
1025         }
1026 }
1027
1028 /*
1029  * Ensure that all locks are released before first NVM or PHY access
1030  */
1031 static void
1032 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
1033 {
1034         uint16_t mask;
1035
1036         /*
1037          * Phy lock should not fail in this early stage. If this is the case,
1038          * it is due to an improper exit of the application.
1039          * So force the release of the faulty lock. Release of common lock
1040          * is done automatically by swfw_sync function.
1041          */
1042         mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
1043         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1044                 PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
1045         }
1046         ixgbe_release_swfw_semaphore(hw, mask);
1047
1048         /*
1049          * These ones are more tricky since they are common to all ports; but
1050          * swfw_sync retries last long enough (1s) to be almost sure that if
1051          * lock can not be taken it is due to an improper lock of the
1052          * semaphore.
1053          */
1054         mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
1055         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1056                 PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1057         }
1058         ixgbe_release_swfw_semaphore(hw, mask);
1059 }
1060
1061 /*
1062  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
1063  * It returns 0 on success.
1064  */
1065 static int
1066 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params __rte_unused)
1067 {
1068         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1069         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1070         struct ixgbe_hw *hw =
1071                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1072         struct ixgbe_vfta *shadow_vfta =
1073                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1074         struct ixgbe_hwstrip *hwstrip =
1075                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1076         struct ixgbe_dcb_config *dcb_config =
1077                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
1078         struct ixgbe_filter_info *filter_info =
1079                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1080         struct ixgbe_bw_conf *bw_conf =
1081                 IXGBE_DEV_PRIVATE_TO_BW_CONF(eth_dev->data->dev_private);
1082         uint32_t ctrl_ext;
1083         uint16_t csum;
1084         int diag, i;
1085
1086         PMD_INIT_FUNC_TRACE();
1087
1088         eth_dev->dev_ops = &ixgbe_eth_dev_ops;
1089         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1090         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1091         eth_dev->tx_pkt_prepare = &ixgbe_prep_pkts;
1092
1093         /*
1094          * For secondary processes, we don't initialise any further as primary
1095          * has already done this work. Only check we don't need a different
1096          * RX and TX function.
1097          */
1098         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1099                 struct ixgbe_tx_queue *txq;
1100                 /* TX queue function in primary, set by last queue initialized
1101                  * Tx queue may not initialized by primary process
1102                  */
1103                 if (eth_dev->data->tx_queues) {
1104                         txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
1105                         ixgbe_set_tx_function(eth_dev, txq);
1106                 } else {
1107                         /* Use default TX function if we get here */
1108                         PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
1109                                      "Using default TX function.");
1110                 }
1111
1112                 ixgbe_set_rx_function(eth_dev);
1113
1114                 return 0;
1115         }
1116
1117         rte_eth_copy_pci_info(eth_dev, pci_dev);
1118
1119         /* Vendor and Device ID need to be set before init of shared code */
1120         hw->device_id = pci_dev->id.device_id;
1121         hw->vendor_id = pci_dev->id.vendor_id;
1122         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1123         hw->allow_unsupported_sfp = 1;
1124
1125         /* Initialize the shared code (base driver) */
1126 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1127         diag = ixgbe_bypass_init_shared_code(hw);
1128 #else
1129         diag = ixgbe_init_shared_code(hw);
1130 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1131
1132         if (diag != IXGBE_SUCCESS) {
1133                 PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
1134                 return -EIO;
1135         }
1136
1137         if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
1138                 PMD_INIT_LOG(ERR, "\nERROR: "
1139                         "Firmware recovery mode detected. Limiting functionality.\n"
1140                         "Refer to the Intel(R) Ethernet Adapters and Devices "
1141                         "User Guide for details on firmware recovery mode.");
1142                 return -EIO;
1143         }
1144
1145         /* pick up the PCI bus settings for reporting later */
1146         ixgbe_get_bus_info(hw);
1147
1148         /* Unlock any pending hardware semaphore */
1149         ixgbe_swfw_lock_reset(hw);
1150
1151 #ifdef RTE_LIBRTE_SECURITY
1152         /* Initialize security_ctx only for primary process*/
1153         if (ixgbe_ipsec_ctx_create(eth_dev))
1154                 return -ENOMEM;
1155 #endif
1156
1157         /* Initialize DCB configuration*/
1158         memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
1159         ixgbe_dcb_init(hw, dcb_config);
1160         /* Get Hardware Flow Control setting */
1161         hw->fc.requested_mode = ixgbe_fc_full;
1162         hw->fc.current_mode = ixgbe_fc_full;
1163         hw->fc.pause_time = IXGBE_FC_PAUSE;
1164         for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
1165                 hw->fc.low_water[i] = IXGBE_FC_LO;
1166                 hw->fc.high_water[i] = IXGBE_FC_HI;
1167         }
1168         hw->fc.send_xon = 1;
1169
1170         /* Make sure we have a good EEPROM before we read from it */
1171         diag = ixgbe_validate_eeprom_checksum(hw, &csum);
1172         if (diag != IXGBE_SUCCESS) {
1173                 PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
1174                 return -EIO;
1175         }
1176
1177 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1178         diag = ixgbe_bypass_init_hw(hw);
1179 #else
1180         diag = ixgbe_init_hw(hw);
1181 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1182
1183         /*
1184          * Devices with copper phys will fail to initialise if ixgbe_init_hw()
1185          * is called too soon after the kernel driver unbinding/binding occurs.
1186          * The failure occurs in ixgbe_identify_phy_generic() for all devices,
1187          * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
1188          * also called. See ixgbe_identify_phy_82599(). The reason for the
1189          * failure is not known, and only occuts when virtualisation features
1190          * are disabled in the bios. A delay of 100ms  was found to be enough by
1191          * trial-and-error, and is doubled to be safe.
1192          */
1193         if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
1194                 rte_delay_ms(200);
1195                 diag = ixgbe_init_hw(hw);
1196         }
1197
1198         if (diag == IXGBE_ERR_SFP_NOT_PRESENT)
1199                 diag = IXGBE_SUCCESS;
1200
1201         if (diag == IXGBE_ERR_EEPROM_VERSION) {
1202                 PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
1203                              "LOM.  Please be aware there may be issues associated "
1204                              "with your hardware.");
1205                 PMD_INIT_LOG(ERR, "If you are experiencing problems "
1206                              "please contact your Intel or hardware representative "
1207                              "who provided you with this hardware.");
1208         } else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
1209                 PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
1210         if (diag) {
1211                 PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
1212                 return -EIO;
1213         }
1214
1215         /* Reset the hw statistics */
1216         ixgbe_dev_stats_reset(eth_dev);
1217
1218         /* disable interrupt */
1219         ixgbe_disable_intr(hw);
1220
1221         /* reset mappings for queue statistics hw counters*/
1222         ixgbe_reset_qstat_mappings(hw);
1223
1224         /* Allocate memory for storing MAC addresses */
1225         eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", RTE_ETHER_ADDR_LEN *
1226                                                hw->mac.num_rar_entries, 0);
1227         if (eth_dev->data->mac_addrs == NULL) {
1228                 PMD_INIT_LOG(ERR,
1229                              "Failed to allocate %u bytes needed to store "
1230                              "MAC addresses",
1231                              RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1232                 return -ENOMEM;
1233         }
1234         /* Copy the permanent MAC address */
1235         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.perm_addr,
1236                         &eth_dev->data->mac_addrs[0]);
1237
1238         /* Allocate memory for storing hash filter MAC addresses */
1239         eth_dev->data->hash_mac_addrs = rte_zmalloc(
1240                 "ixgbe", RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC, 0);
1241         if (eth_dev->data->hash_mac_addrs == NULL) {
1242                 PMD_INIT_LOG(ERR,
1243                              "Failed to allocate %d bytes needed to store MAC addresses",
1244                              RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
1245                 return -ENOMEM;
1246         }
1247
1248         /* initialize the vfta */
1249         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1250
1251         /* initialize the hw strip bitmap*/
1252         memset(hwstrip, 0, sizeof(*hwstrip));
1253
1254         /* initialize PF if max_vfs not zero */
1255         ixgbe_pf_host_init(eth_dev);
1256
1257         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1258         /* let hardware know driver is loaded */
1259         ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1260         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
1261         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1262         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1263         IXGBE_WRITE_FLUSH(hw);
1264
1265         if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1266                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1267                              (int) hw->mac.type, (int) hw->phy.type,
1268                              (int) hw->phy.sfp_type);
1269         else
1270                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1271                              (int) hw->mac.type, (int) hw->phy.type);
1272
1273         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1274                      eth_dev->data->port_id, pci_dev->id.vendor_id,
1275                      pci_dev->id.device_id);
1276
1277         rte_intr_callback_register(intr_handle,
1278                                    ixgbe_dev_interrupt_handler, eth_dev);
1279
1280         /* enable uio/vfio intr/eventfd mapping */
1281         rte_intr_enable(intr_handle);
1282
1283         /* enable support intr */
1284         ixgbe_enable_intr(eth_dev);
1285
1286         /* initialize filter info */
1287         memset(filter_info, 0,
1288                sizeof(struct ixgbe_filter_info));
1289
1290         /* initialize 5tuple filter list */
1291         TAILQ_INIT(&filter_info->fivetuple_list);
1292
1293         /* initialize flow director filter list & hash */
1294         ixgbe_fdir_filter_init(eth_dev);
1295
1296         /* initialize l2 tunnel filter list & hash */
1297         ixgbe_l2_tn_filter_init(eth_dev);
1298
1299         /* initialize flow filter lists */
1300         ixgbe_filterlist_init();
1301
1302         /* initialize bandwidth configuration info */
1303         memset(bw_conf, 0, sizeof(struct ixgbe_bw_conf));
1304
1305         /* initialize Traffic Manager configuration */
1306         ixgbe_tm_conf_init(eth_dev);
1307
1308         return 0;
1309 }
1310
1311 static int
1312 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1313 {
1314         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1315         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1316         struct ixgbe_hw *hw;
1317         int retries = 0;
1318         int ret;
1319
1320         PMD_INIT_FUNC_TRACE();
1321
1322         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1323                 return 0;
1324
1325         hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1326
1327         if (hw->adapter_stopped == 0)
1328                 ixgbe_dev_close(eth_dev);
1329
1330         eth_dev->dev_ops = NULL;
1331         eth_dev->rx_pkt_burst = NULL;
1332         eth_dev->tx_pkt_burst = NULL;
1333
1334         /* Unlock any pending hardware semaphore */
1335         ixgbe_swfw_lock_reset(hw);
1336
1337         /* disable uio intr before callback unregister */
1338         rte_intr_disable(intr_handle);
1339
1340         do {
1341                 ret = rte_intr_callback_unregister(intr_handle,
1342                                 ixgbe_dev_interrupt_handler, eth_dev);
1343                 if (ret >= 0) {
1344                         break;
1345                 } else if (ret != -EAGAIN) {
1346                         PMD_INIT_LOG(ERR,
1347                                 "intr callback unregister failed: %d",
1348                                 ret);
1349                         return ret;
1350                 }
1351                 rte_delay_ms(100);
1352         } while (retries++ < (10 + IXGBE_LINK_UP_TIME));
1353
1354         /* cancel the delay handler before remove dev */
1355         rte_eal_alarm_cancel(ixgbe_dev_interrupt_delayed_handler, eth_dev);
1356
1357         /* cancel the link handler before remove dev */
1358         rte_eal_alarm_cancel(ixgbe_dev_setup_link_alarm_handler, eth_dev);
1359
1360         /* uninitialize PF if max_vfs not zero */
1361         ixgbe_pf_host_uninit(eth_dev);
1362
1363         /* remove all the fdir filters & hash */
1364         ixgbe_fdir_filter_uninit(eth_dev);
1365
1366         /* remove all the L2 tunnel filters & hash */
1367         ixgbe_l2_tn_filter_uninit(eth_dev);
1368
1369         /* Remove all ntuple filters of the device */
1370         ixgbe_ntuple_filter_uninit(eth_dev);
1371
1372         /* clear all the filters list */
1373         ixgbe_filterlist_flush();
1374
1375         /* Remove all Traffic Manager configuration */
1376         ixgbe_tm_conf_uninit(eth_dev);
1377
1378 #ifdef RTE_LIBRTE_SECURITY
1379         rte_free(eth_dev->security_ctx);
1380 #endif
1381
1382         return 0;
1383 }
1384
1385 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
1386 {
1387         struct ixgbe_filter_info *filter_info =
1388                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1389         struct ixgbe_5tuple_filter *p_5tuple;
1390
1391         while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
1392                 TAILQ_REMOVE(&filter_info->fivetuple_list,
1393                              p_5tuple,
1394                              entries);
1395                 rte_free(p_5tuple);
1396         }
1397         memset(filter_info->fivetuple_mask, 0,
1398                sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1399
1400         return 0;
1401 }
1402
1403 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev)
1404 {
1405         struct ixgbe_hw_fdir_info *fdir_info =
1406                 IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1407         struct ixgbe_fdir_filter *fdir_filter;
1408
1409                 if (fdir_info->hash_map)
1410                 rte_free(fdir_info->hash_map);
1411         if (fdir_info->hash_handle)
1412                 rte_hash_free(fdir_info->hash_handle);
1413
1414         while ((fdir_filter = TAILQ_FIRST(&fdir_info->fdir_list))) {
1415                 TAILQ_REMOVE(&fdir_info->fdir_list,
1416                              fdir_filter,
1417                              entries);
1418                 rte_free(fdir_filter);
1419         }
1420
1421         return 0;
1422 }
1423
1424 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev)
1425 {
1426         struct ixgbe_l2_tn_info *l2_tn_info =
1427                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1428         struct ixgbe_l2_tn_filter *l2_tn_filter;
1429
1430         if (l2_tn_info->hash_map)
1431                 rte_free(l2_tn_info->hash_map);
1432         if (l2_tn_info->hash_handle)
1433                 rte_hash_free(l2_tn_info->hash_handle);
1434
1435         while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
1436                 TAILQ_REMOVE(&l2_tn_info->l2_tn_list,
1437                              l2_tn_filter,
1438                              entries);
1439                 rte_free(l2_tn_filter);
1440         }
1441
1442         return 0;
1443 }
1444
1445 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev)
1446 {
1447         struct ixgbe_hw_fdir_info *fdir_info =
1448                 IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1449         char fdir_hash_name[RTE_HASH_NAMESIZE];
1450         struct rte_hash_parameters fdir_hash_params = {
1451                 .name = fdir_hash_name,
1452                 .entries = IXGBE_MAX_FDIR_FILTER_NUM,
1453                 .key_len = sizeof(union ixgbe_atr_input),
1454                 .hash_func = rte_hash_crc,
1455                 .hash_func_init_val = 0,
1456                 .socket_id = rte_socket_id(),
1457         };
1458
1459         TAILQ_INIT(&fdir_info->fdir_list);
1460         snprintf(fdir_hash_name, RTE_HASH_NAMESIZE,
1461                  "fdir_%s", eth_dev->device->name);
1462         fdir_info->hash_handle = rte_hash_create(&fdir_hash_params);
1463         if (!fdir_info->hash_handle) {
1464                 PMD_INIT_LOG(ERR, "Failed to create fdir hash table!");
1465                 return -EINVAL;
1466         }
1467         fdir_info->hash_map = rte_zmalloc("ixgbe",
1468                                           sizeof(struct ixgbe_fdir_filter *) *
1469                                           IXGBE_MAX_FDIR_FILTER_NUM,
1470                                           0);
1471         if (!fdir_info->hash_map) {
1472                 PMD_INIT_LOG(ERR,
1473                              "Failed to allocate memory for fdir hash map!");
1474                 return -ENOMEM;
1475         }
1476         fdir_info->mask_added = FALSE;
1477
1478         return 0;
1479 }
1480
1481 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev)
1482 {
1483         struct ixgbe_l2_tn_info *l2_tn_info =
1484                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1485         char l2_tn_hash_name[RTE_HASH_NAMESIZE];
1486         struct rte_hash_parameters l2_tn_hash_params = {
1487                 .name = l2_tn_hash_name,
1488                 .entries = IXGBE_MAX_L2_TN_FILTER_NUM,
1489                 .key_len = sizeof(struct ixgbe_l2_tn_key),
1490                 .hash_func = rte_hash_crc,
1491                 .hash_func_init_val = 0,
1492                 .socket_id = rte_socket_id(),
1493         };
1494
1495         TAILQ_INIT(&l2_tn_info->l2_tn_list);
1496         snprintf(l2_tn_hash_name, RTE_HASH_NAMESIZE,
1497                  "l2_tn_%s", eth_dev->device->name);
1498         l2_tn_info->hash_handle = rte_hash_create(&l2_tn_hash_params);
1499         if (!l2_tn_info->hash_handle) {
1500                 PMD_INIT_LOG(ERR, "Failed to create L2 TN hash table!");
1501                 return -EINVAL;
1502         }
1503         l2_tn_info->hash_map = rte_zmalloc("ixgbe",
1504                                    sizeof(struct ixgbe_l2_tn_filter *) *
1505                                    IXGBE_MAX_L2_TN_FILTER_NUM,
1506                                    0);
1507         if (!l2_tn_info->hash_map) {
1508                 PMD_INIT_LOG(ERR,
1509                         "Failed to allocate memory for L2 TN hash map!");
1510                 return -ENOMEM;
1511         }
1512         l2_tn_info->e_tag_en = FALSE;
1513         l2_tn_info->e_tag_fwd_en = FALSE;
1514         l2_tn_info->e_tag_ether_type = RTE_ETHER_TYPE_ETAG;
1515
1516         return 0;
1517 }
1518 /*
1519  * Negotiate mailbox API version with the PF.
1520  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1521  * Then we try to negotiate starting with the most recent one.
1522  * If all negotiation attempts fail, then we will proceed with
1523  * the default one (ixgbe_mbox_api_10).
1524  */
1525 static void
1526 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1527 {
1528         int32_t i;
1529
1530         /* start with highest supported, proceed down */
1531         static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1532                 ixgbe_mbox_api_13,
1533                 ixgbe_mbox_api_12,
1534                 ixgbe_mbox_api_11,
1535                 ixgbe_mbox_api_10,
1536         };
1537
1538         for (i = 0;
1539                         i != RTE_DIM(sup_ver) &&
1540                         ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1541                         i++)
1542                 ;
1543 }
1544
1545 static void
1546 generate_random_mac_addr(struct rte_ether_addr *mac_addr)
1547 {
1548         uint64_t random;
1549
1550         /* Set Organizationally Unique Identifier (OUI) prefix. */
1551         mac_addr->addr_bytes[0] = 0x00;
1552         mac_addr->addr_bytes[1] = 0x09;
1553         mac_addr->addr_bytes[2] = 0xC0;
1554         /* Force indication of locally assigned MAC address. */
1555         mac_addr->addr_bytes[0] |= RTE_ETHER_LOCAL_ADMIN_ADDR;
1556         /* Generate the last 3 bytes of the MAC address with a random number. */
1557         random = rte_rand();
1558         memcpy(&mac_addr->addr_bytes[3], &random, 3);
1559 }
1560
1561 static int
1562 devarg_handle_int(__rte_unused const char *key, const char *value,
1563                   void *extra_args)
1564 {
1565         uint16_t *n = extra_args;
1566
1567         if (value == NULL || extra_args == NULL)
1568                 return -EINVAL;
1569
1570         *n = (uint16_t)strtoul(value, NULL, 0);
1571         if (*n == USHRT_MAX && errno == ERANGE)
1572                 return -1;
1573
1574         return 0;
1575 }
1576
1577 static void
1578 ixgbevf_parse_devargs(struct ixgbe_adapter *adapter,
1579                       struct rte_devargs *devargs)
1580 {
1581         struct rte_kvargs *kvlist;
1582         uint16_t pflink_fullchk;
1583
1584         if (devargs == NULL)
1585                 return;
1586
1587         kvlist = rte_kvargs_parse(devargs->args, ixgbevf_valid_arguments);
1588         if (kvlist == NULL)
1589                 return;
1590
1591         if (rte_kvargs_count(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK) == 1 &&
1592             rte_kvargs_process(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK,
1593                                devarg_handle_int, &pflink_fullchk) == 0 &&
1594             pflink_fullchk == 1)
1595                 adapter->pflink_fullchk = 1;
1596
1597         rte_kvargs_free(kvlist);
1598 }
1599
1600 /*
1601  * Virtual Function device init
1602  */
1603 static int
1604 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1605 {
1606         int diag;
1607         uint32_t tc, tcs;
1608         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1609         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1610         struct ixgbe_hw *hw =
1611                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1612         struct ixgbe_vfta *shadow_vfta =
1613                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1614         struct ixgbe_hwstrip *hwstrip =
1615                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1616         struct rte_ether_addr *perm_addr =
1617                 (struct rte_ether_addr *)hw->mac.perm_addr;
1618
1619         PMD_INIT_FUNC_TRACE();
1620
1621         eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1622         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1623         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1624
1625         /* for secondary processes, we don't initialise any further as primary
1626          * has already done this work. Only check we don't need a different
1627          * RX function
1628          */
1629         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1630                 struct ixgbe_tx_queue *txq;
1631                 /* TX queue function in primary, set by last queue initialized
1632                  * Tx queue may not initialized by primary process
1633                  */
1634                 if (eth_dev->data->tx_queues) {
1635                         txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1636                         ixgbe_set_tx_function(eth_dev, txq);
1637                 } else {
1638                         /* Use default TX function if we get here */
1639                         PMD_INIT_LOG(NOTICE,
1640                                      "No TX queues configured yet. Using default TX function.");
1641                 }
1642
1643                 ixgbe_set_rx_function(eth_dev);
1644
1645                 return 0;
1646         }
1647
1648         ixgbevf_parse_devargs(eth_dev->data->dev_private,
1649                               pci_dev->device.devargs);
1650
1651         rte_eth_copy_pci_info(eth_dev, pci_dev);
1652
1653         hw->device_id = pci_dev->id.device_id;
1654         hw->vendor_id = pci_dev->id.vendor_id;
1655         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1656
1657         /* initialize the vfta */
1658         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1659
1660         /* initialize the hw strip bitmap*/
1661         memset(hwstrip, 0, sizeof(*hwstrip));
1662
1663         /* Initialize the shared code (base driver) */
1664         diag = ixgbe_init_shared_code(hw);
1665         if (diag != IXGBE_SUCCESS) {
1666                 PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1667                 return -EIO;
1668         }
1669
1670         /* init_mailbox_params */
1671         hw->mbx.ops.init_params(hw);
1672
1673         /* Reset the hw statistics */
1674         ixgbevf_dev_stats_reset(eth_dev);
1675
1676         /* Disable the interrupts for VF */
1677         ixgbevf_intr_disable(eth_dev);
1678
1679         hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1680         diag = hw->mac.ops.reset_hw(hw);
1681
1682         /*
1683          * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1684          * the underlying PF driver has not assigned a MAC address to the VF.
1685          * In this case, assign a random MAC address.
1686          */
1687         if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1688                 PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1689                 /*
1690                  * This error code will be propagated to the app by
1691                  * rte_eth_dev_reset, so use a public error code rather than
1692                  * the internal-only IXGBE_ERR_RESET_FAILED
1693                  */
1694                 return -EAGAIN;
1695         }
1696
1697         /* negotiate mailbox API version to use with the PF. */
1698         ixgbevf_negotiate_api(hw);
1699
1700         /* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1701         ixgbevf_get_queues(hw, &tcs, &tc);
1702
1703         /* Allocate memory for storing MAC addresses */
1704         eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", RTE_ETHER_ADDR_LEN *
1705                                                hw->mac.num_rar_entries, 0);
1706         if (eth_dev->data->mac_addrs == NULL) {
1707                 PMD_INIT_LOG(ERR,
1708                              "Failed to allocate %u bytes needed to store "
1709                              "MAC addresses",
1710                              RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1711                 return -ENOMEM;
1712         }
1713
1714         /* Generate a random MAC address, if none was assigned by PF. */
1715         if (rte_is_zero_ether_addr(perm_addr)) {
1716                 generate_random_mac_addr(perm_addr);
1717                 diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1718                 if (diag) {
1719                         rte_free(eth_dev->data->mac_addrs);
1720                         eth_dev->data->mac_addrs = NULL;
1721                         return diag;
1722                 }
1723                 PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1724                 PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1725                              "%02x:%02x:%02x:%02x:%02x:%02x",
1726                              perm_addr->addr_bytes[0],
1727                              perm_addr->addr_bytes[1],
1728                              perm_addr->addr_bytes[2],
1729                              perm_addr->addr_bytes[3],
1730                              perm_addr->addr_bytes[4],
1731                              perm_addr->addr_bytes[5]);
1732         }
1733
1734         /* Copy the permanent MAC address */
1735         rte_ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1736
1737         /* reset the hardware with the new settings */
1738         diag = hw->mac.ops.start_hw(hw);
1739         switch (diag) {
1740         case  0:
1741                 break;
1742
1743         default:
1744                 PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1745                 return -EIO;
1746         }
1747
1748         rte_intr_callback_register(intr_handle,
1749                                    ixgbevf_dev_interrupt_handler, eth_dev);
1750         rte_intr_enable(intr_handle);
1751         ixgbevf_intr_enable(eth_dev);
1752
1753         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1754                      eth_dev->data->port_id, pci_dev->id.vendor_id,
1755                      pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1756
1757         return 0;
1758 }
1759
1760 /* Virtual Function device uninit */
1761
1762 static int
1763 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1764 {
1765         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1766         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1767         struct ixgbe_hw *hw;
1768
1769         PMD_INIT_FUNC_TRACE();
1770
1771         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1772                 return 0;
1773
1774         hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1775
1776         if (hw->adapter_stopped == 0)
1777                 ixgbevf_dev_close(eth_dev);
1778
1779         eth_dev->dev_ops = NULL;
1780         eth_dev->rx_pkt_burst = NULL;
1781         eth_dev->tx_pkt_burst = NULL;
1782
1783         /* Disable the interrupts for VF */
1784         ixgbevf_intr_disable(eth_dev);
1785
1786         rte_intr_disable(intr_handle);
1787         rte_intr_callback_unregister(intr_handle,
1788                                      ixgbevf_dev_interrupt_handler, eth_dev);
1789
1790         return 0;
1791 }
1792
1793 static int
1794 eth_ixgbe_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1795                 struct rte_pci_device *pci_dev)
1796 {
1797         char name[RTE_ETH_NAME_MAX_LEN];
1798         struct rte_eth_dev *pf_ethdev;
1799         struct rte_eth_devargs eth_da;
1800         int i, retval;
1801
1802         if (pci_dev->device.devargs) {
1803                 retval = rte_eth_devargs_parse(pci_dev->device.devargs->args,
1804                                 &eth_da);
1805                 if (retval)
1806                         return retval;
1807         } else
1808                 memset(&eth_da, 0, sizeof(eth_da));
1809
1810         retval = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
1811                 sizeof(struct ixgbe_adapter),
1812                 eth_dev_pci_specific_init, pci_dev,
1813                 eth_ixgbe_dev_init, NULL);
1814
1815         if (retval || eth_da.nb_representor_ports < 1)
1816                 return retval;
1817
1818         pf_ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1819         if (pf_ethdev == NULL)
1820                 return -ENODEV;
1821
1822         /* probe VF representor ports */
1823         for (i = 0; i < eth_da.nb_representor_ports; i++) {
1824                 struct ixgbe_vf_info *vfinfo;
1825                 struct ixgbe_vf_representor representor;
1826
1827                 vfinfo = *IXGBE_DEV_PRIVATE_TO_P_VFDATA(
1828                         pf_ethdev->data->dev_private);
1829                 if (vfinfo == NULL) {
1830                         PMD_DRV_LOG(ERR,
1831                                 "no virtual functions supported by PF");
1832                         break;
1833                 }
1834
1835                 representor.vf_id = eth_da.representor_ports[i];
1836                 representor.switch_domain_id = vfinfo->switch_domain_id;
1837                 representor.pf_ethdev = pf_ethdev;
1838
1839                 /* representor port net_bdf_port */
1840                 snprintf(name, sizeof(name), "net_%s_representor_%d",
1841                         pci_dev->device.name,
1842                         eth_da.representor_ports[i]);
1843
1844                 retval = rte_eth_dev_create(&pci_dev->device, name,
1845                         sizeof(struct ixgbe_vf_representor), NULL, NULL,
1846                         ixgbe_vf_representor_init, &representor);
1847
1848                 if (retval)
1849                         PMD_DRV_LOG(ERR, "failed to create ixgbe vf "
1850                                 "representor %s.", name);
1851         }
1852
1853         return 0;
1854 }
1855
1856 static int eth_ixgbe_pci_remove(struct rte_pci_device *pci_dev)
1857 {
1858         struct rte_eth_dev *ethdev;
1859
1860         ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1861         if (!ethdev)
1862                 return -ENODEV;
1863
1864         if (ethdev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
1865                 return rte_eth_dev_destroy(ethdev, ixgbe_vf_representor_uninit);
1866         else
1867                 return rte_eth_dev_destroy(ethdev, eth_ixgbe_dev_uninit);
1868 }
1869
1870 static struct rte_pci_driver rte_ixgbe_pmd = {
1871         .id_table = pci_id_ixgbe_map,
1872         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1873         .probe = eth_ixgbe_pci_probe,
1874         .remove = eth_ixgbe_pci_remove,
1875 };
1876
1877 static int eth_ixgbevf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1878         struct rte_pci_device *pci_dev)
1879 {
1880         return rte_eth_dev_pci_generic_probe(pci_dev,
1881                 sizeof(struct ixgbe_adapter), eth_ixgbevf_dev_init);
1882 }
1883
1884 static int eth_ixgbevf_pci_remove(struct rte_pci_device *pci_dev)
1885 {
1886         return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbevf_dev_uninit);
1887 }
1888
1889 /*
1890  * virtual function driver struct
1891  */
1892 static struct rte_pci_driver rte_ixgbevf_pmd = {
1893         .id_table = pci_id_ixgbevf_map,
1894         .drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1895         .probe = eth_ixgbevf_pci_probe,
1896         .remove = eth_ixgbevf_pci_remove,
1897 };
1898
1899 static int
1900 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1901 {
1902         struct ixgbe_hw *hw =
1903                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1904         struct ixgbe_vfta *shadow_vfta =
1905                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1906         uint32_t vfta;
1907         uint32_t vid_idx;
1908         uint32_t vid_bit;
1909
1910         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1911         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1912         vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1913         if (on)
1914                 vfta |= vid_bit;
1915         else
1916                 vfta &= ~vid_bit;
1917         IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1918
1919         /* update local VFTA copy */
1920         shadow_vfta->vfta[vid_idx] = vfta;
1921
1922         return 0;
1923 }
1924
1925 static void
1926 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1927 {
1928         if (on)
1929                 ixgbe_vlan_hw_strip_enable(dev, queue);
1930         else
1931                 ixgbe_vlan_hw_strip_disable(dev, queue);
1932 }
1933
1934 static int
1935 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1936                     enum rte_vlan_type vlan_type,
1937                     uint16_t tpid)
1938 {
1939         struct ixgbe_hw *hw =
1940                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1941         int ret = 0;
1942         uint32_t reg;
1943         uint32_t qinq;
1944
1945         qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1946         qinq &= IXGBE_DMATXCTL_GDV;
1947
1948         switch (vlan_type) {
1949         case ETH_VLAN_TYPE_INNER:
1950                 if (qinq) {
1951                         reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1952                         reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1953                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1954                         reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1955                         reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1956                                 | ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1957                         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1958                 } else {
1959                         ret = -ENOTSUP;
1960                         PMD_DRV_LOG(ERR, "Inner type is not supported"
1961                                     " by single VLAN");
1962                 }
1963                 break;
1964         case ETH_VLAN_TYPE_OUTER:
1965                 if (qinq) {
1966                         /* Only the high 16-bits is valid */
1967                         IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1968                                         IXGBE_EXVET_VET_EXT_SHIFT);
1969                 } else {
1970                         reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1971                         reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1972                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1973                         reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1974                         reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1975                                 | ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1976                         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1977                 }
1978
1979                 break;
1980         default:
1981                 ret = -EINVAL;
1982                 PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1983                 break;
1984         }
1985
1986         return ret;
1987 }
1988
1989 void
1990 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1991 {
1992         struct ixgbe_hw *hw =
1993                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1994         uint32_t vlnctrl;
1995
1996         PMD_INIT_FUNC_TRACE();
1997
1998         /* Filter Table Disable */
1999         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2000         vlnctrl &= ~IXGBE_VLNCTRL_VFE;
2001
2002         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
2003 }
2004
2005 void
2006 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2007 {
2008         struct ixgbe_hw *hw =
2009                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2010         struct ixgbe_vfta *shadow_vfta =
2011                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2012         uint32_t vlnctrl;
2013         uint16_t i;
2014
2015         PMD_INIT_FUNC_TRACE();
2016
2017         /* Filter Table Enable */
2018         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2019         vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
2020         vlnctrl |= IXGBE_VLNCTRL_VFE;
2021
2022         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
2023
2024         /* write whatever is in local vfta copy */
2025         for (i = 0; i < IXGBE_VFTA_SIZE; i++)
2026                 IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
2027 }
2028
2029 static void
2030 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
2031 {
2032         struct ixgbe_hwstrip *hwstrip =
2033                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
2034         struct ixgbe_rx_queue *rxq;
2035
2036         if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
2037                 return;
2038
2039         if (on)
2040                 IXGBE_SET_HWSTRIP(hwstrip, queue);
2041         else
2042                 IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
2043
2044         if (queue >= dev->data->nb_rx_queues)
2045                 return;
2046
2047         rxq = dev->data->rx_queues[queue];
2048
2049         if (on) {
2050                 rxq->vlan_flags = PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
2051                 rxq->offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
2052         } else {
2053                 rxq->vlan_flags = PKT_RX_VLAN;
2054                 rxq->offloads &= ~DEV_RX_OFFLOAD_VLAN_STRIP;
2055         }
2056 }
2057
2058 static void
2059 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
2060 {
2061         struct ixgbe_hw *hw =
2062                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2063         uint32_t ctrl;
2064
2065         PMD_INIT_FUNC_TRACE();
2066
2067         if (hw->mac.type == ixgbe_mac_82598EB) {
2068                 /* No queue level support */
2069                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2070                 return;
2071         }
2072
2073         /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2074         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2075         ctrl &= ~IXGBE_RXDCTL_VME;
2076         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2077
2078         /* record those setting for HW strip per queue */
2079         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
2080 }
2081
2082 static void
2083 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
2084 {
2085         struct ixgbe_hw *hw =
2086                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2087         uint32_t ctrl;
2088
2089         PMD_INIT_FUNC_TRACE();
2090
2091         if (hw->mac.type == ixgbe_mac_82598EB) {
2092                 /* No queue level supported */
2093                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2094                 return;
2095         }
2096
2097         /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2098         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2099         ctrl |= IXGBE_RXDCTL_VME;
2100         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2101
2102         /* record those setting for HW strip per queue */
2103         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
2104 }
2105
2106 static void
2107 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2108 {
2109         struct ixgbe_hw *hw =
2110                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2111         uint32_t ctrl;
2112
2113         PMD_INIT_FUNC_TRACE();
2114
2115         /* DMATXCTRL: Geric Double VLAN Disable */
2116         ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2117         ctrl &= ~IXGBE_DMATXCTL_GDV;
2118         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2119
2120         /* CTRL_EXT: Global Double VLAN Disable */
2121         ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2122         ctrl &= ~IXGBE_EXTENDED_VLAN;
2123         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2124
2125 }
2126
2127 static void
2128 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2129 {
2130         struct ixgbe_hw *hw =
2131                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2132         uint32_t ctrl;
2133
2134         PMD_INIT_FUNC_TRACE();
2135
2136         /* DMATXCTRL: Geric Double VLAN Enable */
2137         ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2138         ctrl |= IXGBE_DMATXCTL_GDV;
2139         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2140
2141         /* CTRL_EXT: Global Double VLAN Enable */
2142         ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2143         ctrl |= IXGBE_EXTENDED_VLAN;
2144         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2145
2146         /* Clear pooling mode of PFVTCTL. It's required by X550. */
2147         if (hw->mac.type == ixgbe_mac_X550 ||
2148             hw->mac.type == ixgbe_mac_X550EM_x ||
2149             hw->mac.type == ixgbe_mac_X550EM_a) {
2150                 ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
2151                 ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
2152                 IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
2153         }
2154
2155         /*
2156          * VET EXT field in the EXVET register = 0x8100 by default
2157          * So no need to change. Same to VT field of DMATXCTL register
2158          */
2159 }
2160
2161 void
2162 ixgbe_vlan_hw_strip_config(struct rte_eth_dev *dev)
2163 {
2164         struct ixgbe_hw *hw =
2165                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2166         struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
2167         uint32_t ctrl;
2168         uint16_t i;
2169         struct ixgbe_rx_queue *rxq;
2170         bool on;
2171
2172         PMD_INIT_FUNC_TRACE();
2173
2174         if (hw->mac.type == ixgbe_mac_82598EB) {
2175                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP) {
2176                         ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2177                         ctrl |= IXGBE_VLNCTRL_VME;
2178                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2179                 } else {
2180                         ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2181                         ctrl &= ~IXGBE_VLNCTRL_VME;
2182                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2183                 }
2184         } else {
2185                 /*
2186                  * Other 10G NIC, the VLAN strip can be setup
2187                  * per queue in RXDCTL
2188                  */
2189                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
2190                         rxq = dev->data->rx_queues[i];
2191                         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2192                         if (rxq->offloads & DEV_RX_OFFLOAD_VLAN_STRIP) {
2193                                 ctrl |= IXGBE_RXDCTL_VME;
2194                                 on = TRUE;
2195                         } else {
2196                                 ctrl &= ~IXGBE_RXDCTL_VME;
2197                                 on = FALSE;
2198                         }
2199                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2200
2201                         /* record those setting for HW strip per queue */
2202                         ixgbe_vlan_hw_strip_bitmap_set(dev, i, on);
2203                 }
2204         }
2205 }
2206
2207 static void
2208 ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev, int mask)
2209 {
2210         uint16_t i;
2211         struct rte_eth_rxmode *rxmode;
2212         struct ixgbe_rx_queue *rxq;
2213
2214         if (mask & ETH_VLAN_STRIP_MASK) {
2215                 rxmode = &dev->data->dev_conf.rxmode;
2216                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2217                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
2218                                 rxq = dev->data->rx_queues[i];
2219                                 rxq->offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
2220                         }
2221                 else
2222                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
2223                                 rxq = dev->data->rx_queues[i];
2224                                 rxq->offloads &= ~DEV_RX_OFFLOAD_VLAN_STRIP;
2225                         }
2226         }
2227 }
2228
2229 static int
2230 ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask)
2231 {
2232         struct rte_eth_rxmode *rxmode;
2233         rxmode = &dev->data->dev_conf.rxmode;
2234
2235         if (mask & ETH_VLAN_STRIP_MASK) {
2236                 ixgbe_vlan_hw_strip_config(dev);
2237         }
2238
2239         if (mask & ETH_VLAN_FILTER_MASK) {
2240                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2241                         ixgbe_vlan_hw_filter_enable(dev);
2242                 else
2243                         ixgbe_vlan_hw_filter_disable(dev);
2244         }
2245
2246         if (mask & ETH_VLAN_EXTEND_MASK) {
2247                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2248                         ixgbe_vlan_hw_extend_enable(dev);
2249                 else
2250                         ixgbe_vlan_hw_extend_disable(dev);
2251         }
2252
2253         return 0;
2254 }
2255
2256 static int
2257 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2258 {
2259         ixgbe_config_vlan_strip_on_all_queues(dev, mask);
2260
2261         ixgbe_vlan_offload_config(dev, mask);
2262
2263         return 0;
2264 }
2265
2266 static void
2267 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2268 {
2269         struct ixgbe_hw *hw =
2270                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2271         /* VLNCTRL: enable vlan filtering and allow all vlan tags through */
2272         uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2273
2274         vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
2275         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
2276 }
2277
2278 static int
2279 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
2280 {
2281         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2282
2283         switch (nb_rx_q) {
2284         case 1:
2285         case 2:
2286                 RTE_ETH_DEV_SRIOV(dev).active = ETH_64_POOLS;
2287                 break;
2288         case 4:
2289                 RTE_ETH_DEV_SRIOV(dev).active = ETH_32_POOLS;
2290                 break;
2291         default:
2292                 return -EINVAL;
2293         }
2294
2295         RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
2296                 IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2297         RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
2298                 pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2299         return 0;
2300 }
2301
2302 static int
2303 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
2304 {
2305         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
2306         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2307         uint16_t nb_rx_q = dev->data->nb_rx_queues;
2308         uint16_t nb_tx_q = dev->data->nb_tx_queues;
2309
2310         if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
2311                 /* check multi-queue mode */
2312                 switch (dev_conf->rxmode.mq_mode) {
2313                 case ETH_MQ_RX_VMDQ_DCB:
2314                         PMD_INIT_LOG(INFO, "ETH_MQ_RX_VMDQ_DCB mode supported in SRIOV");
2315                         break;
2316                 case ETH_MQ_RX_VMDQ_DCB_RSS:
2317                         /* DCB/RSS VMDQ in SRIOV mode, not implement yet */
2318                         PMD_INIT_LOG(ERR, "SRIOV active,"
2319                                         " unsupported mq_mode rx %d.",
2320                                         dev_conf->rxmode.mq_mode);
2321                         return -EINVAL;
2322                 case ETH_MQ_RX_RSS:
2323                 case ETH_MQ_RX_VMDQ_RSS:
2324                         dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_RSS;
2325                         if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
2326                                 if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
2327                                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2328                                                 " invalid queue number"
2329                                                 " for VMDQ RSS, allowed"
2330                                                 " value are 1, 2 or 4.");
2331                                         return -EINVAL;
2332                                 }
2333                         break;
2334                 case ETH_MQ_RX_VMDQ_ONLY:
2335                 case ETH_MQ_RX_NONE:
2336                         /* if nothing mq mode configure, use default scheme */
2337                         dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
2338                         break;
2339                 default: /* ETH_MQ_RX_DCB, ETH_MQ_RX_DCB_RSS or ETH_MQ_TX_DCB*/
2340                         /* SRIOV only works in VMDq enable mode */
2341                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2342                                         " wrong mq_mode rx %d.",
2343                                         dev_conf->rxmode.mq_mode);
2344                         return -EINVAL;
2345                 }
2346
2347                 switch (dev_conf->txmode.mq_mode) {
2348                 case ETH_MQ_TX_VMDQ_DCB:
2349                         PMD_INIT_LOG(INFO, "ETH_MQ_TX_VMDQ_DCB mode supported in SRIOV");
2350                         dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_DCB;
2351                         break;
2352                 default: /* ETH_MQ_TX_VMDQ_ONLY or ETH_MQ_TX_NONE */
2353                         dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_ONLY;
2354                         break;
2355                 }
2356
2357                 /* check valid queue number */
2358                 if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2359                     (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2360                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2361                                         " nb_rx_q=%d nb_tx_q=%d queue number"
2362                                         " must be less than or equal to %d.",
2363                                         nb_rx_q, nb_tx_q,
2364                                         RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2365                         return -EINVAL;
2366                 }
2367         } else {
2368                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB_RSS) {
2369                         PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2370                                           " not supported.");
2371                         return -EINVAL;
2372                 }
2373                 /* check configuration for vmdb+dcb mode */
2374                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB) {
2375                         const struct rte_eth_vmdq_dcb_conf *conf;
2376
2377                         if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2378                                 PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2379                                                 IXGBE_VMDQ_DCB_NB_QUEUES);
2380                                 return -EINVAL;
2381                         }
2382                         conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2383                         if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2384                                conf->nb_queue_pools == ETH_32_POOLS)) {
2385                                 PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2386                                                 " nb_queue_pools must be %d or %d.",
2387                                                 ETH_16_POOLS, ETH_32_POOLS);
2388                                 return -EINVAL;
2389                         }
2390                 }
2391                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_VMDQ_DCB) {
2392                         const struct rte_eth_vmdq_dcb_tx_conf *conf;
2393
2394                         if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2395                                 PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2396                                                  IXGBE_VMDQ_DCB_NB_QUEUES);
2397                                 return -EINVAL;
2398                         }
2399                         conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2400                         if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2401                                conf->nb_queue_pools == ETH_32_POOLS)) {
2402                                 PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2403                                                 " nb_queue_pools != %d and"
2404                                                 " nb_queue_pools != %d.",
2405                                                 ETH_16_POOLS, ETH_32_POOLS);
2406                                 return -EINVAL;
2407                         }
2408                 }
2409
2410                 /* For DCB mode check our configuration before we go further */
2411                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB) {
2412                         const struct rte_eth_dcb_rx_conf *conf;
2413
2414                         conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2415                         if (!(conf->nb_tcs == ETH_4_TCS ||
2416                                conf->nb_tcs == ETH_8_TCS)) {
2417                                 PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2418                                                 " and nb_tcs != %d.",
2419                                                 ETH_4_TCS, ETH_8_TCS);
2420                                 return -EINVAL;
2421                         }
2422                 }
2423
2424                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_DCB) {
2425                         const struct rte_eth_dcb_tx_conf *conf;
2426
2427                         conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2428                         if (!(conf->nb_tcs == ETH_4_TCS ||
2429                                conf->nb_tcs == ETH_8_TCS)) {
2430                                 PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2431                                                 " and nb_tcs != %d.",
2432                                                 ETH_4_TCS, ETH_8_TCS);
2433                                 return -EINVAL;
2434                         }
2435                 }
2436
2437                 /*
2438                  * When DCB/VT is off, maximum number of queues changes,
2439                  * except for 82598EB, which remains constant.
2440                  */
2441                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
2442                                 hw->mac.type != ixgbe_mac_82598EB) {
2443                         if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2444                                 PMD_INIT_LOG(ERR,
2445                                              "Neither VT nor DCB are enabled, "
2446                                              "nb_tx_q > %d.",
2447                                              IXGBE_NONE_MODE_TX_NB_QUEUES);
2448                                 return -EINVAL;
2449                         }
2450                 }
2451         }
2452         return 0;
2453 }
2454
2455 static int
2456 ixgbe_dev_configure(struct rte_eth_dev *dev)
2457 {
2458         struct ixgbe_interrupt *intr =
2459                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2460         struct ixgbe_adapter *adapter = dev->data->dev_private;
2461         int ret;
2462
2463         PMD_INIT_FUNC_TRACE();
2464         /* multipe queue mode checking */
2465         ret  = ixgbe_check_mq_mode(dev);
2466         if (ret != 0) {
2467                 PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2468                             ret);
2469                 return ret;
2470         }
2471
2472         /* set flag to update link status after init */
2473         intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2474
2475         /*
2476          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2477          * allocation or vector Rx preconditions we will reset it.
2478          */
2479         adapter->rx_bulk_alloc_allowed = true;
2480         adapter->rx_vec_allowed = true;
2481
2482         return 0;
2483 }
2484
2485 static void
2486 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2487 {
2488         struct ixgbe_hw *hw =
2489                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2490         struct ixgbe_interrupt *intr =
2491                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2492         uint32_t gpie;
2493
2494         /* only set up it on X550EM_X */
2495         if (hw->mac.type == ixgbe_mac_X550EM_x) {
2496                 gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2497                 gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2498                 IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2499                 if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2500                         intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2501         }
2502 }
2503
2504 int
2505 ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
2506                         uint16_t tx_rate, uint64_t q_msk)
2507 {
2508         struct ixgbe_hw *hw;
2509         struct ixgbe_vf_info *vfinfo;
2510         struct rte_eth_link link;
2511         uint8_t  nb_q_per_pool;
2512         uint32_t queue_stride;
2513         uint32_t queue_idx, idx = 0, vf_idx;
2514         uint32_t queue_end;
2515         uint16_t total_rate = 0;
2516         struct rte_pci_device *pci_dev;
2517
2518         pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2519         rte_eth_link_get_nowait(dev->data->port_id, &link);
2520
2521         if (vf >= pci_dev->max_vfs)
2522                 return -EINVAL;
2523
2524         if (tx_rate > link.link_speed)
2525                 return -EINVAL;
2526
2527         if (q_msk == 0)
2528                 return 0;
2529
2530         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2531         vfinfo = *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
2532         nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2533         queue_stride = IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2534         queue_idx = vf * queue_stride;
2535         queue_end = queue_idx + nb_q_per_pool - 1;
2536         if (queue_end >= hw->mac.max_tx_queues)
2537                 return -EINVAL;
2538
2539         if (vfinfo) {
2540                 for (vf_idx = 0; vf_idx < pci_dev->max_vfs; vf_idx++) {
2541                         if (vf_idx == vf)
2542                                 continue;
2543                         for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
2544                                 idx++)
2545                                 total_rate += vfinfo[vf_idx].tx_rate[idx];
2546                 }
2547         } else {
2548                 return -EINVAL;
2549         }
2550
2551         /* Store tx_rate for this vf. */
2552         for (idx = 0; idx < nb_q_per_pool; idx++) {
2553                 if (((uint64_t)0x1 << idx) & q_msk) {
2554                         if (vfinfo[vf].tx_rate[idx] != tx_rate)
2555                                 vfinfo[vf].tx_rate[idx] = tx_rate;
2556                         total_rate += tx_rate;
2557                 }
2558         }
2559
2560         if (total_rate > dev->data->dev_link.link_speed) {
2561                 /* Reset stored TX rate of the VF if it causes exceed
2562                  * link speed.
2563                  */
2564                 memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
2565                 return -EINVAL;
2566         }
2567
2568         /* Set RTTBCNRC of each queue/pool for vf X  */
2569         for (; queue_idx <= queue_end; queue_idx++) {
2570                 if (0x1 & q_msk)
2571                         ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
2572                 q_msk = q_msk >> 1;
2573         }
2574
2575         return 0;
2576 }
2577
2578 /*
2579  * Configure device link speed and setup link.
2580  * It returns 0 on success.
2581  */
2582 static int
2583 ixgbe_dev_start(struct rte_eth_dev *dev)
2584 {
2585         struct ixgbe_hw *hw =
2586                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2587         struct ixgbe_vf_info *vfinfo =
2588                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2589         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2590         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2591         uint32_t intr_vector = 0;
2592         int err, link_up = 0, negotiate = 0;
2593         uint32_t speed = 0;
2594         uint32_t allowed_speeds = 0;
2595         int mask = 0;
2596         int status;
2597         uint16_t vf, idx;
2598         uint32_t *link_speeds;
2599         struct ixgbe_tm_conf *tm_conf =
2600                 IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2601
2602         PMD_INIT_FUNC_TRACE();
2603
2604         /* IXGBE devices don't support:
2605         *    - half duplex (checked afterwards for valid speeds)
2606         *    - fixed speed: TODO implement
2607         */
2608         if (dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED) {
2609                 PMD_INIT_LOG(ERR,
2610                 "Invalid link_speeds for port %u, fix speed not supported",
2611                                 dev->data->port_id);
2612                 return -EINVAL;
2613         }
2614
2615         /* Stop the link setup handler before resetting the HW. */
2616         rte_eal_alarm_cancel(ixgbe_dev_setup_link_alarm_handler, dev);
2617
2618         /* disable uio/vfio intr/eventfd mapping */
2619         rte_intr_disable(intr_handle);
2620
2621         /* stop adapter */
2622         hw->adapter_stopped = 0;
2623         ixgbe_stop_adapter(hw);
2624
2625         /* reinitialize adapter
2626          * this calls reset and start
2627          */
2628         status = ixgbe_pf_reset_hw(hw);
2629         if (status != 0)
2630                 return -1;
2631         hw->mac.ops.start_hw(hw);
2632         hw->mac.get_link_status = true;
2633
2634         /* configure PF module if SRIOV enabled */
2635         ixgbe_pf_host_configure(dev);
2636
2637         ixgbe_dev_phy_intr_setup(dev);
2638
2639         /* check and configure queue intr-vector mapping */
2640         if ((rte_intr_cap_multiple(intr_handle) ||
2641              !RTE_ETH_DEV_SRIOV(dev).active) &&
2642             dev->data->dev_conf.intr_conf.rxq != 0) {
2643                 intr_vector = dev->data->nb_rx_queues;
2644                 if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2645                         PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2646                                         IXGBE_MAX_INTR_QUEUE_NUM);
2647                         return -ENOTSUP;
2648                 }
2649                 if (rte_intr_efd_enable(intr_handle, intr_vector))
2650                         return -1;
2651         }
2652
2653         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
2654                 intr_handle->intr_vec =
2655                         rte_zmalloc("intr_vec",
2656                                     dev->data->nb_rx_queues * sizeof(int), 0);
2657                 if (intr_handle->intr_vec == NULL) {
2658                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2659                                      " intr_vec", dev->data->nb_rx_queues);
2660                         return -ENOMEM;
2661                 }
2662         }
2663
2664         /* confiugre msix for sleep until rx interrupt */
2665         ixgbe_configure_msix(dev);
2666
2667         /* initialize transmission unit */
2668         ixgbe_dev_tx_init(dev);
2669
2670         /* This can fail when allocating mbufs for descriptor rings */
2671         err = ixgbe_dev_rx_init(dev);
2672         if (err) {
2673                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2674                 goto error;
2675         }
2676
2677         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
2678                 ETH_VLAN_EXTEND_MASK;
2679         err = ixgbe_vlan_offload_config(dev, mask);
2680         if (err) {
2681                 PMD_INIT_LOG(ERR, "Unable to set VLAN offload");
2682                 goto error;
2683         }
2684
2685         if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
2686                 /* Enable vlan filtering for VMDq */
2687                 ixgbe_vmdq_vlan_hw_filter_enable(dev);
2688         }
2689
2690         /* Configure DCB hw */
2691         ixgbe_configure_dcb(dev);
2692
2693         if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2694                 err = ixgbe_fdir_configure(dev);
2695                 if (err)
2696                         goto error;
2697         }
2698
2699         /* Restore vf rate limit */
2700         if (vfinfo != NULL) {
2701                 for (vf = 0; vf < pci_dev->max_vfs; vf++)
2702                         for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2703                                 if (vfinfo[vf].tx_rate[idx] != 0)
2704                                         ixgbe_set_vf_rate_limit(
2705                                                 dev, vf,
2706                                                 vfinfo[vf].tx_rate[idx],
2707                                                 1 << idx);
2708         }
2709
2710         ixgbe_restore_statistics_mapping(dev);
2711
2712         err = ixgbe_dev_rxtx_start(dev);
2713         if (err < 0) {
2714                 PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2715                 goto error;
2716         }
2717
2718         /* Skip link setup if loopback mode is enabled. */
2719         if (dev->data->dev_conf.lpbk_mode != 0) {
2720                 err = ixgbe_check_supported_loopback_mode(dev);
2721                 if (err < 0) {
2722                         PMD_INIT_LOG(ERR, "Unsupported loopback mode");
2723                         goto error;
2724                 } else {
2725                         goto skip_link_setup;
2726                 }
2727         }
2728
2729         if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2730                 err = hw->mac.ops.setup_sfp(hw);
2731                 if (err)
2732                         goto error;
2733         }
2734
2735         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2736                 /* Turn on the copper */
2737                 ixgbe_set_phy_power(hw, true);
2738         } else {
2739                 /* Turn on the laser */
2740                 ixgbe_enable_tx_laser(hw);
2741         }
2742
2743         err = ixgbe_check_link(hw, &speed, &link_up, 0);
2744         if (err)
2745                 goto error;
2746         dev->data->dev_link.link_status = link_up;
2747
2748         err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2749         if (err)
2750                 goto error;
2751
2752         switch (hw->mac.type) {
2753         case ixgbe_mac_X550:
2754         case ixgbe_mac_X550EM_x:
2755         case ixgbe_mac_X550EM_a:
2756                 allowed_speeds = ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2757                         ETH_LINK_SPEED_2_5G |  ETH_LINK_SPEED_5G |
2758                         ETH_LINK_SPEED_10G;
2759                 if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
2760                                 hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
2761                         allowed_speeds = ETH_LINK_SPEED_10M |
2762                                 ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G;
2763                 break;
2764         default:
2765                 allowed_speeds = ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2766                         ETH_LINK_SPEED_10G;
2767         }
2768
2769         link_speeds = &dev->data->dev_conf.link_speeds;
2770         if (*link_speeds & ~allowed_speeds) {
2771                 PMD_INIT_LOG(ERR, "Invalid link setting");
2772                 goto error;
2773         }
2774
2775         speed = 0x0;
2776         if (*link_speeds == ETH_LINK_SPEED_AUTONEG) {
2777                 switch (hw->mac.type) {
2778                 case ixgbe_mac_82598EB:
2779                         speed = IXGBE_LINK_SPEED_82598_AUTONEG;
2780                         break;
2781                 case ixgbe_mac_82599EB:
2782                 case ixgbe_mac_X540:
2783                         speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2784                         break;
2785                 case ixgbe_mac_X550:
2786                 case ixgbe_mac_X550EM_x:
2787                 case ixgbe_mac_X550EM_a:
2788                         speed = IXGBE_LINK_SPEED_X550_AUTONEG;
2789                         break;
2790                 default:
2791                         speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2792                 }
2793         } else {
2794                 if (*link_speeds & ETH_LINK_SPEED_10G)
2795                         speed |= IXGBE_LINK_SPEED_10GB_FULL;
2796                 if (*link_speeds & ETH_LINK_SPEED_5G)
2797                         speed |= IXGBE_LINK_SPEED_5GB_FULL;
2798                 if (*link_speeds & ETH_LINK_SPEED_2_5G)
2799                         speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
2800                 if (*link_speeds & ETH_LINK_SPEED_1G)
2801                         speed |= IXGBE_LINK_SPEED_1GB_FULL;
2802                 if (*link_speeds & ETH_LINK_SPEED_100M)
2803                         speed |= IXGBE_LINK_SPEED_100_FULL;
2804                 if (*link_speeds & ETH_LINK_SPEED_10M)
2805                         speed |= IXGBE_LINK_SPEED_10_FULL;
2806         }
2807
2808         err = ixgbe_setup_link(hw, speed, link_up);
2809         if (err)
2810                 goto error;
2811
2812 skip_link_setup:
2813
2814         if (rte_intr_allow_others(intr_handle)) {
2815                 /* check if lsc interrupt is enabled */
2816                 if (dev->data->dev_conf.intr_conf.lsc != 0)
2817                         ixgbe_dev_lsc_interrupt_setup(dev, TRUE);
2818                 else
2819                         ixgbe_dev_lsc_interrupt_setup(dev, FALSE);
2820                 ixgbe_dev_macsec_interrupt_setup(dev);
2821         } else {
2822                 rte_intr_callback_unregister(intr_handle,
2823                                              ixgbe_dev_interrupt_handler, dev);
2824                 if (dev->data->dev_conf.intr_conf.lsc != 0)
2825                         PMD_INIT_LOG(INFO, "lsc won't enable because of"
2826                                      " no intr multiplex");
2827         }
2828
2829         /* check if rxq interrupt is enabled */
2830         if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2831             rte_intr_dp_is_en(intr_handle))
2832                 ixgbe_dev_rxq_interrupt_setup(dev);
2833
2834         /* enable uio/vfio intr/eventfd mapping */
2835         rte_intr_enable(intr_handle);
2836
2837         /* resume enabled intr since hw reset */
2838         ixgbe_enable_intr(dev);
2839         ixgbe_l2_tunnel_conf(dev);
2840         ixgbe_filter_restore(dev);
2841
2842         if (tm_conf->root && !tm_conf->committed)
2843                 PMD_DRV_LOG(WARNING,
2844                             "please call hierarchy_commit() "
2845                             "before starting the port");
2846
2847         /*
2848          * Update link status right before return, because it may
2849          * start link configuration process in a separate thread.
2850          */
2851         ixgbe_dev_link_update(dev, 0);
2852
2853         return 0;
2854
2855 error:
2856         PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2857         ixgbe_dev_clear_queues(dev);
2858         return -EIO;
2859 }
2860
2861 /*
2862  * Stop device: disable rx and tx functions to allow for reconfiguring.
2863  */
2864 static void
2865 ixgbe_dev_stop(struct rte_eth_dev *dev)
2866 {
2867         struct rte_eth_link link;
2868         struct ixgbe_adapter *adapter = dev->data->dev_private;
2869         struct ixgbe_hw *hw =
2870                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2871         struct ixgbe_vf_info *vfinfo =
2872                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2873         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2874         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2875         int vf;
2876         struct ixgbe_tm_conf *tm_conf =
2877                 IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2878
2879         PMD_INIT_FUNC_TRACE();
2880
2881         rte_eal_alarm_cancel(ixgbe_dev_setup_link_alarm_handler, dev);
2882
2883         /* disable interrupts */
2884         ixgbe_disable_intr(hw);
2885
2886         /* reset the NIC */
2887         ixgbe_pf_reset_hw(hw);
2888         hw->adapter_stopped = 0;
2889
2890         /* stop adapter */
2891         ixgbe_stop_adapter(hw);
2892
2893         for (vf = 0; vfinfo != NULL && vf < pci_dev->max_vfs; vf++)
2894                 vfinfo[vf].clear_to_send = false;
2895
2896         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2897                 /* Turn off the copper */
2898                 ixgbe_set_phy_power(hw, false);
2899         } else {
2900                 /* Turn off the laser */
2901                 ixgbe_disable_tx_laser(hw);
2902         }
2903
2904         ixgbe_dev_clear_queues(dev);
2905
2906         /* Clear stored conf */
2907         dev->data->scattered_rx = 0;
2908         dev->data->lro = 0;
2909
2910         /* Clear recorded link status */
2911         memset(&link, 0, sizeof(link));
2912         rte_eth_linkstatus_set(dev, &link);
2913
2914         if (!rte_intr_allow_others(intr_handle))
2915                 /* resume to the default handler */
2916                 rte_intr_callback_register(intr_handle,
2917                                            ixgbe_dev_interrupt_handler,
2918                                            (void *)dev);
2919
2920         /* Clean datapath event and queue/vec mapping */
2921         rte_intr_efd_disable(intr_handle);
2922         if (intr_handle->intr_vec != NULL) {
2923                 rte_free(intr_handle->intr_vec);
2924                 intr_handle->intr_vec = NULL;
2925         }
2926
2927         /* reset hierarchy commit */
2928         tm_conf->committed = false;
2929
2930         adapter->rss_reta_updated = 0;
2931 }
2932
2933 /*
2934  * Set device link up: enable tx.
2935  */
2936 static int
2937 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2938 {
2939         struct ixgbe_hw *hw =
2940                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2941         if (hw->mac.type == ixgbe_mac_82599EB) {
2942 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2943                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2944                         /* Not suported in bypass mode */
2945                         PMD_INIT_LOG(ERR, "Set link up is not supported "
2946                                      "by device id 0x%x", hw->device_id);
2947                         return -ENOTSUP;
2948                 }
2949 #endif
2950         }
2951
2952         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2953                 /* Turn on the copper */
2954                 ixgbe_set_phy_power(hw, true);
2955         } else {
2956                 /* Turn on the laser */
2957                 ixgbe_enable_tx_laser(hw);
2958         }
2959
2960         return 0;
2961 }
2962
2963 /*
2964  * Set device link down: disable tx.
2965  */
2966 static int
2967 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2968 {
2969         struct ixgbe_hw *hw =
2970                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2971         if (hw->mac.type == ixgbe_mac_82599EB) {
2972 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2973                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2974                         /* Not suported in bypass mode */
2975                         PMD_INIT_LOG(ERR, "Set link down is not supported "
2976                                      "by device id 0x%x", hw->device_id);
2977                         return -ENOTSUP;
2978                 }
2979 #endif
2980         }
2981
2982         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2983                 /* Turn off the copper */
2984                 ixgbe_set_phy_power(hw, false);
2985         } else {
2986                 /* Turn off the laser */
2987                 ixgbe_disable_tx_laser(hw);
2988         }
2989
2990         return 0;
2991 }
2992
2993 /*
2994  * Reset and stop device.
2995  */
2996 static void
2997 ixgbe_dev_close(struct rte_eth_dev *dev)
2998 {
2999         struct ixgbe_hw *hw =
3000                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3001
3002         PMD_INIT_FUNC_TRACE();
3003
3004         ixgbe_pf_reset_hw(hw);
3005
3006         ixgbe_dev_stop(dev);
3007         hw->adapter_stopped = 1;
3008
3009         ixgbe_dev_free_queues(dev);
3010
3011         ixgbe_disable_pcie_master(hw);
3012
3013         /* reprogram the RAR[0] in case user changed it. */
3014         ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
3015 }
3016
3017 /*
3018  * Reset PF device.
3019  */
3020 static int
3021 ixgbe_dev_reset(struct rte_eth_dev *dev)
3022 {
3023         int ret;
3024
3025         /* When a DPDK PMD PF begin to reset PF port, it should notify all
3026          * its VF to make them align with it. The detailed notification
3027          * mechanism is PMD specific. As to ixgbe PF, it is rather complex.
3028          * To avoid unexpected behavior in VF, currently reset of PF with
3029          * SR-IOV activation is not supported. It might be supported later.
3030          */
3031         if (dev->data->sriov.active)
3032                 return -ENOTSUP;
3033
3034         ret = eth_ixgbe_dev_uninit(dev);
3035         if (ret)
3036                 return ret;
3037
3038         ret = eth_ixgbe_dev_init(dev, NULL);
3039
3040         return ret;
3041 }
3042
3043 static void
3044 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
3045                            struct ixgbe_hw_stats *hw_stats,
3046                            struct ixgbe_macsec_stats *macsec_stats,
3047                            uint64_t *total_missed_rx, uint64_t *total_qbrc,
3048                            uint64_t *total_qprc, uint64_t *total_qprdc)
3049 {
3050         uint32_t bprc, lxon, lxoff, total;
3051         uint32_t delta_gprc = 0;
3052         unsigned i;
3053         /* Workaround for RX byte count not including CRC bytes when CRC
3054          * strip is enabled. CRC bytes are removed from counters when crc_strip
3055          * is disabled.
3056          */
3057         int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
3058                         IXGBE_HLREG0_RXCRCSTRP);
3059
3060         hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
3061         hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
3062         hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
3063         hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
3064
3065         for (i = 0; i < 8; i++) {
3066                 uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
3067
3068                 /* global total per queue */
3069                 hw_stats->mpc[i] += mp;
3070                 /* Running comprehensive total for stats display */
3071                 *total_missed_rx += hw_stats->mpc[i];
3072                 if (hw->mac.type == ixgbe_mac_82598EB) {
3073                         hw_stats->rnbc[i] +=
3074                             IXGBE_READ_REG(hw, IXGBE_RNBC(i));
3075                         hw_stats->pxonrxc[i] +=
3076                                 IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
3077                         hw_stats->pxoffrxc[i] +=
3078                                 IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
3079                 } else {
3080                         hw_stats->pxonrxc[i] +=
3081                                 IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
3082                         hw_stats->pxoffrxc[i] +=
3083                                 IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
3084                         hw_stats->pxon2offc[i] +=
3085                                 IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
3086                 }
3087                 hw_stats->pxontxc[i] +=
3088                     IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
3089                 hw_stats->pxofftxc[i] +=
3090                     IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
3091         }
3092         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3093                 uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
3094                 uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
3095                 uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
3096
3097                 delta_gprc += delta_qprc;
3098
3099                 hw_stats->qprc[i] += delta_qprc;
3100                 hw_stats->qptc[i] += delta_qptc;
3101
3102                 hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
3103                 hw_stats->qbrc[i] +=
3104                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
3105                 if (crc_strip == 0)
3106                         hw_stats->qbrc[i] -= delta_qprc * RTE_ETHER_CRC_LEN;
3107
3108                 hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
3109                 hw_stats->qbtc[i] +=
3110                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
3111
3112                 hw_stats->qprdc[i] += delta_qprdc;
3113                 *total_qprdc += hw_stats->qprdc[i];
3114
3115                 *total_qprc += hw_stats->qprc[i];
3116                 *total_qbrc += hw_stats->qbrc[i];
3117         }
3118         hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
3119         hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
3120         hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
3121
3122         /*
3123          * An errata states that gprc actually counts good + missed packets:
3124          * Workaround to set gprc to summated queue packet receives
3125          */
3126         hw_stats->gprc = *total_qprc;
3127
3128         if (hw->mac.type != ixgbe_mac_82598EB) {
3129                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
3130                 hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
3131                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
3132                 hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
3133                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
3134                 hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
3135                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
3136                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
3137         } else {
3138                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
3139                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
3140                 /* 82598 only has a counter in the high register */
3141                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
3142                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
3143                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
3144         }
3145         uint64_t old_tpr = hw_stats->tpr;
3146
3147         hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
3148         hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
3149
3150         if (crc_strip == 0)
3151                 hw_stats->gorc -= delta_gprc * RTE_ETHER_CRC_LEN;
3152
3153         uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
3154         hw_stats->gptc += delta_gptc;
3155         hw_stats->gotc -= delta_gptc * RTE_ETHER_CRC_LEN;
3156         hw_stats->tor -= (hw_stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
3157
3158         /*
3159          * Workaround: mprc hardware is incorrectly counting
3160          * broadcasts, so for now we subtract those.
3161          */
3162         bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
3163         hw_stats->bprc += bprc;
3164         hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
3165         if (hw->mac.type == ixgbe_mac_82598EB)
3166                 hw_stats->mprc -= bprc;
3167
3168         hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
3169         hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
3170         hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
3171         hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
3172         hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
3173         hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
3174
3175         lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
3176         hw_stats->lxontxc += lxon;
3177         lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
3178         hw_stats->lxofftxc += lxoff;
3179         total = lxon + lxoff;
3180
3181         hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
3182         hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
3183         hw_stats->gptc -= total;
3184         hw_stats->mptc -= total;
3185         hw_stats->ptc64 -= total;
3186         hw_stats->gotc -= total * RTE_ETHER_MIN_LEN;
3187
3188         hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3189         hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3190         hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3191         hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
3192         hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
3193         hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
3194         hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
3195         hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3196         hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3197         hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3198         hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3199         hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
3200         hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3201         hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
3202         hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
3203         hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
3204         /* Only read FCOE on 82599 */
3205         if (hw->mac.type != ixgbe_mac_82598EB) {
3206                 hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
3207                 hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
3208                 hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
3209                 hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
3210                 hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
3211         }
3212
3213         /* Flow Director Stats registers */
3214         if (hw->mac.type != ixgbe_mac_82598EB) {
3215                 hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
3216                 hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
3217                 hw_stats->fdirustat_add += IXGBE_READ_REG(hw,
3218                                         IXGBE_FDIRUSTAT) & 0xFFFF;
3219                 hw_stats->fdirustat_remove += (IXGBE_READ_REG(hw,
3220                                         IXGBE_FDIRUSTAT) >> 16) & 0xFFFF;
3221                 hw_stats->fdirfstat_fadd += IXGBE_READ_REG(hw,
3222                                         IXGBE_FDIRFSTAT) & 0xFFFF;
3223                 hw_stats->fdirfstat_fremove += (IXGBE_READ_REG(hw,
3224                                         IXGBE_FDIRFSTAT) >> 16) & 0xFFFF;
3225         }
3226         /* MACsec Stats registers */
3227         macsec_stats->out_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECTXUT);
3228         macsec_stats->out_pkts_encrypted +=
3229                 IXGBE_READ_REG(hw, IXGBE_LSECTXPKTE);
3230         macsec_stats->out_pkts_protected +=
3231                 IXGBE_READ_REG(hw, IXGBE_LSECTXPKTP);
3232         macsec_stats->out_octets_encrypted +=
3233                 IXGBE_READ_REG(hw, IXGBE_LSECTXOCTE);
3234         macsec_stats->out_octets_protected +=
3235                 IXGBE_READ_REG(hw, IXGBE_LSECTXOCTP);
3236         macsec_stats->in_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECRXUT);
3237         macsec_stats->in_pkts_badtag += IXGBE_READ_REG(hw, IXGBE_LSECRXBAD);
3238         macsec_stats->in_pkts_nosci += IXGBE_READ_REG(hw, IXGBE_LSECRXNOSCI);
3239         macsec_stats->in_pkts_unknownsci +=
3240                 IXGBE_READ_REG(hw, IXGBE_LSECRXUNSCI);
3241         macsec_stats->in_octets_decrypted +=
3242                 IXGBE_READ_REG(hw, IXGBE_LSECRXOCTD);
3243         macsec_stats->in_octets_validated +=
3244                 IXGBE_READ_REG(hw, IXGBE_LSECRXOCTV);
3245         macsec_stats->in_pkts_unchecked += IXGBE_READ_REG(hw, IXGBE_LSECRXUNCH);
3246         macsec_stats->in_pkts_delayed += IXGBE_READ_REG(hw, IXGBE_LSECRXDELAY);
3247         macsec_stats->in_pkts_late += IXGBE_READ_REG(hw, IXGBE_LSECRXLATE);
3248         for (i = 0; i < 2; i++) {
3249                 macsec_stats->in_pkts_ok +=
3250                         IXGBE_READ_REG(hw, IXGBE_LSECRXOK(i));
3251                 macsec_stats->in_pkts_invalid +=
3252                         IXGBE_READ_REG(hw, IXGBE_LSECRXINV(i));
3253                 macsec_stats->in_pkts_notvalid +=
3254                         IXGBE_READ_REG(hw, IXGBE_LSECRXNV(i));
3255         }
3256         macsec_stats->in_pkts_unusedsa += IXGBE_READ_REG(hw, IXGBE_LSECRXUNSA);
3257         macsec_stats->in_pkts_notusingsa +=
3258                 IXGBE_READ_REG(hw, IXGBE_LSECRXNUSA);
3259 }
3260
3261 /*
3262  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
3263  */
3264 static int
3265 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3266 {
3267         struct ixgbe_hw *hw =
3268                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3269         struct ixgbe_hw_stats *hw_stats =
3270                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3271         struct ixgbe_macsec_stats *macsec_stats =
3272                         IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3273                                 dev->data->dev_private);
3274         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3275         unsigned i;
3276
3277         total_missed_rx = 0;
3278         total_qbrc = 0;
3279         total_qprc = 0;
3280         total_qprdc = 0;
3281
3282         ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3283                         &total_qbrc, &total_qprc, &total_qprdc);
3284
3285         if (stats == NULL)
3286                 return -EINVAL;
3287
3288         /* Fill out the rte_eth_stats statistics structure */
3289         stats->ipackets = total_qprc;
3290         stats->ibytes = total_qbrc;
3291         stats->opackets = hw_stats->gptc;
3292         stats->obytes = hw_stats->gotc;
3293
3294         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3295                 stats->q_ipackets[i] = hw_stats->qprc[i];
3296                 stats->q_opackets[i] = hw_stats->qptc[i];
3297                 stats->q_ibytes[i] = hw_stats->qbrc[i];
3298                 stats->q_obytes[i] = hw_stats->qbtc[i];
3299                 stats->q_errors[i] = hw_stats->qprdc[i];
3300         }
3301
3302         /* Rx Errors */
3303         stats->imissed  = total_missed_rx;
3304         stats->ierrors  = hw_stats->crcerrs +
3305                           hw_stats->mspdc +
3306                           hw_stats->rlec +
3307                           hw_stats->ruc +
3308                           hw_stats->roc +
3309                           hw_stats->illerrc +
3310                           hw_stats->errbc +
3311                           hw_stats->rfc +
3312                           hw_stats->fccrc +
3313                           hw_stats->fclast;
3314
3315         /* Tx Errors */
3316         stats->oerrors  = 0;
3317         return 0;
3318 }
3319
3320 static void
3321 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
3322 {
3323         struct ixgbe_hw_stats *stats =
3324                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3325
3326         /* HW registers are cleared on read */
3327         ixgbe_dev_stats_get(dev, NULL);
3328
3329         /* Reset software totals */
3330         memset(stats, 0, sizeof(*stats));
3331 }
3332
3333 /* This function calculates the number of xstats based on the current config */
3334 static unsigned
3335 ixgbe_xstats_calc_num(void) {
3336         return IXGBE_NB_HW_STATS + IXGBE_NB_MACSEC_STATS +
3337                 (IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
3338                 (IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
3339 }
3340
3341 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3342         struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned int size)
3343 {
3344         const unsigned cnt_stats = ixgbe_xstats_calc_num();
3345         unsigned stat, i, count;
3346
3347         if (xstats_names != NULL) {
3348                 count = 0;
3349
3350                 /* Note: limit >= cnt_stats checked upstream
3351                  * in rte_eth_xstats_names()
3352                  */
3353
3354                 /* Extended stats from ixgbe_hw_stats */
3355                 for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3356                         strlcpy(xstats_names[count].name,
3357                                 rte_ixgbe_stats_strings[i].name,
3358                                 sizeof(xstats_names[count].name));
3359                         count++;
3360                 }
3361
3362                 /* MACsec Stats */
3363                 for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3364                         strlcpy(xstats_names[count].name,
3365                                 rte_ixgbe_macsec_strings[i].name,
3366                                 sizeof(xstats_names[count].name));
3367                         count++;
3368                 }
3369
3370                 /* RX Priority Stats */
3371                 for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3372                         for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3373                                 snprintf(xstats_names[count].name,
3374                                         sizeof(xstats_names[count].name),
3375                                         "rx_priority%u_%s", i,
3376                                         rte_ixgbe_rxq_strings[stat].name);
3377                                 count++;
3378                         }
3379                 }
3380
3381                 /* TX Priority Stats */
3382                 for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3383                         for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3384                                 snprintf(xstats_names[count].name,
3385                                         sizeof(xstats_names[count].name),
3386                                         "tx_priority%u_%s", i,
3387                                         rte_ixgbe_txq_strings[stat].name);
3388                                 count++;
3389                         }
3390                 }
3391         }
3392         return cnt_stats;
3393 }
3394
3395 static int ixgbe_dev_xstats_get_names_by_id(
3396         struct rte_eth_dev *dev,
3397         struct rte_eth_xstat_name *xstats_names,
3398         const uint64_t *ids,
3399         unsigned int limit)
3400 {
3401         if (!ids) {
3402                 const unsigned int cnt_stats = ixgbe_xstats_calc_num();
3403                 unsigned int stat, i, count;
3404
3405                 if (xstats_names != NULL) {
3406                         count = 0;
3407
3408                         /* Note: limit >= cnt_stats checked upstream
3409                          * in rte_eth_xstats_names()
3410                          */
3411
3412                         /* Extended stats from ixgbe_hw_stats */
3413                         for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3414                                 strlcpy(xstats_names[count].name,
3415                                         rte_ixgbe_stats_strings[i].name,
3416                                         sizeof(xstats_names[count].name));
3417                                 count++;
3418                         }
3419
3420                         /* MACsec Stats */
3421                         for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3422                                 strlcpy(xstats_names[count].name,
3423                                         rte_ixgbe_macsec_strings[i].name,
3424                                         sizeof(xstats_names[count].name));
3425                                 count++;
3426                         }
3427
3428                         /* RX Priority Stats */
3429                         for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3430                                 for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3431                                         snprintf(xstats_names[count].name,
3432                                             sizeof(xstats_names[count].name),
3433                                             "rx_priority%u_%s", i,
3434                                             rte_ixgbe_rxq_strings[stat].name);
3435                                         count++;
3436                                 }
3437                         }
3438
3439                         /* TX Priority Stats */
3440                         for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3441                                 for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3442                                         snprintf(xstats_names[count].name,
3443                                             sizeof(xstats_names[count].name),
3444                                             "tx_priority%u_%s", i,
3445                                             rte_ixgbe_txq_strings[stat].name);
3446                                         count++;
3447                                 }
3448                         }
3449                 }
3450                 return cnt_stats;
3451         }
3452
3453         uint16_t i;
3454         uint16_t size = ixgbe_xstats_calc_num();
3455         struct rte_eth_xstat_name xstats_names_copy[size];
3456
3457         ixgbe_dev_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
3458                         size);
3459
3460         for (i = 0; i < limit; i++) {
3461                 if (ids[i] >= size) {
3462                         PMD_INIT_LOG(ERR, "id value isn't valid");
3463                         return -1;
3464                 }
3465                 strcpy(xstats_names[i].name,
3466                                 xstats_names_copy[ids[i]].name);
3467         }
3468         return limit;
3469 }
3470
3471 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3472         struct rte_eth_xstat_name *xstats_names, unsigned limit)
3473 {
3474         unsigned i;
3475
3476         if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
3477                 return -ENOMEM;
3478
3479         if (xstats_names != NULL)
3480                 for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
3481                         strlcpy(xstats_names[i].name,
3482                                 rte_ixgbevf_stats_strings[i].name,
3483                                 sizeof(xstats_names[i].name));
3484         return IXGBEVF_NB_XSTATS;
3485 }
3486
3487 static int
3488 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3489                                          unsigned n)
3490 {
3491         struct ixgbe_hw *hw =
3492                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3493         struct ixgbe_hw_stats *hw_stats =
3494                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3495         struct ixgbe_macsec_stats *macsec_stats =
3496                         IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3497                                 dev->data->dev_private);
3498         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3499         unsigned i, stat, count = 0;
3500
3501         count = ixgbe_xstats_calc_num();
3502
3503         if (n < count)
3504                 return count;
3505
3506         total_missed_rx = 0;
3507         total_qbrc = 0;
3508         total_qprc = 0;
3509         total_qprdc = 0;
3510
3511         ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3512                         &total_qbrc, &total_qprc, &total_qprdc);
3513
3514         /* If this is a reset xstats is NULL, and we have cleared the
3515          * registers by reading them.
3516          */
3517         if (!xstats)
3518                 return 0;
3519
3520         /* Extended stats from ixgbe_hw_stats */
3521         count = 0;
3522         for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3523                 xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3524                                 rte_ixgbe_stats_strings[i].offset);
3525                 xstats[count].id = count;
3526                 count++;
3527         }
3528
3529         /* MACsec Stats */
3530         for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3531                 xstats[count].value = *(uint64_t *)(((char *)macsec_stats) +
3532                                 rte_ixgbe_macsec_strings[i].offset);
3533                 xstats[count].id = count;
3534                 count++;
3535         }
3536
3537         /* RX Priority Stats */
3538         for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3539                 for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3540                         xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3541                                         rte_ixgbe_rxq_strings[stat].offset +
3542                                         (sizeof(uint64_t) * i));
3543                         xstats[count].id = count;
3544                         count++;
3545                 }
3546         }
3547
3548         /* TX Priority Stats */
3549         for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3550                 for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3551                         xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3552                                         rte_ixgbe_txq_strings[stat].offset +
3553                                         (sizeof(uint64_t) * i));
3554                         xstats[count].id = count;
3555                         count++;
3556                 }
3557         }
3558         return count;
3559 }
3560
3561 static int
3562 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
3563                 uint64_t *values, unsigned int n)
3564 {
3565         if (!ids) {
3566                 struct ixgbe_hw *hw =
3567                                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3568                 struct ixgbe_hw_stats *hw_stats =
3569                                 IXGBE_DEV_PRIVATE_TO_STATS(
3570                                                 dev->data->dev_private);
3571                 struct ixgbe_macsec_stats *macsec_stats =
3572                                 IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3573                                         dev->data->dev_private);
3574                 uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3575                 unsigned int i, stat, count = 0;
3576
3577                 count = ixgbe_xstats_calc_num();
3578
3579                 if (!ids && n < count)
3580                         return count;
3581
3582                 total_missed_rx = 0;
3583                 total_qbrc = 0;
3584                 total_qprc = 0;
3585                 total_qprdc = 0;
3586
3587                 ixgbe_read_stats_registers(hw, hw_stats, macsec_stats,
3588                                 &total_missed_rx, &total_qbrc, &total_qprc,
3589                                 &total_qprdc);
3590
3591                 /* If this is a reset xstats is NULL, and we have cleared the
3592                  * registers by reading them.
3593                  */
3594                 if (!ids && !values)
3595                         return 0;
3596
3597                 /* Extended stats from ixgbe_hw_stats */
3598                 count = 0;
3599                 for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3600                         values[count] = *(uint64_t *)(((char *)hw_stats) +
3601                                         rte_ixgbe_stats_strings[i].offset);
3602                         count++;
3603                 }
3604
3605                 /* MACsec Stats */
3606                 for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3607                         values[count] = *(uint64_t *)(((char *)macsec_stats) +
3608                                         rte_ixgbe_macsec_strings[i].offset);
3609                         count++;
3610                 }
3611
3612                 /* RX Priority Stats */
3613                 for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3614                         for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3615                                 values[count] =
3616                                         *(uint64_t *)(((char *)hw_stats) +
3617                                         rte_ixgbe_rxq_strings[stat].offset +
3618                                         (sizeof(uint64_t) * i));
3619                                 count++;
3620                         }
3621                 }
3622
3623                 /* TX Priority Stats */
3624                 for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3625                         for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3626                                 values[count] =
3627                                         *(uint64_t *)(((char *)hw_stats) +
3628                                         rte_ixgbe_txq_strings[stat].offset +
3629                                         (sizeof(uint64_t) * i));
3630                                 count++;
3631                         }
3632                 }
3633                 return count;
3634         }
3635
3636         uint16_t i;
3637         uint16_t size = ixgbe_xstats_calc_num();
3638         uint64_t values_copy[size];
3639
3640         ixgbe_dev_xstats_get_by_id(dev, NULL, values_copy, size);
3641
3642         for (i = 0; i < n; i++) {
3643                 if (ids[i] >= size) {
3644                         PMD_INIT_LOG(ERR, "id value isn't valid");
3645                         return -1;
3646                 }
3647                 values[i] = values_copy[ids[i]];
3648         }
3649         return n;
3650 }
3651
3652 static void
3653 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
3654 {
3655         struct ixgbe_hw_stats *stats =
3656                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3657         struct ixgbe_macsec_stats *macsec_stats =
3658                         IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3659                                 dev->data->dev_private);
3660
3661         unsigned count = ixgbe_xstats_calc_num();
3662
3663         /* HW registers are cleared on read */
3664         ixgbe_dev_xstats_get(dev, NULL, count);
3665
3666         /* Reset software totals */
3667         memset(stats, 0, sizeof(*stats));
3668         memset(macsec_stats, 0, sizeof(*macsec_stats));
3669 }
3670
3671 static void
3672 ixgbevf_update_stats(struct rte_eth_dev *dev)
3673 {
3674         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3675         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3676                           IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3677
3678         /* Good Rx packet, include VF loopback */
3679         UPDATE_VF_STAT(IXGBE_VFGPRC,
3680             hw_stats->last_vfgprc, hw_stats->vfgprc);
3681
3682         /* Good Rx octets, include VF loopback */
3683         UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
3684             hw_stats->last_vfgorc, hw_stats->vfgorc);
3685
3686         /* Good Tx packet, include VF loopback */
3687         UPDATE_VF_STAT(IXGBE_VFGPTC,
3688             hw_stats->last_vfgptc, hw_stats->vfgptc);
3689
3690         /* Good Tx octets, include VF loopback */
3691         UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
3692             hw_stats->last_vfgotc, hw_stats->vfgotc);
3693
3694         /* Rx Multicst Packet */
3695         UPDATE_VF_STAT(IXGBE_VFMPRC,
3696             hw_stats->last_vfmprc, hw_stats->vfmprc);
3697 }
3698
3699 static int
3700 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3701                        unsigned n)
3702 {
3703         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3704                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3705         unsigned i;
3706
3707         if (n < IXGBEVF_NB_XSTATS)
3708                 return IXGBEVF_NB_XSTATS;
3709
3710         ixgbevf_update_stats(dev);
3711
3712         if (!xstats)
3713                 return 0;
3714
3715         /* Extended stats */
3716         for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
3717                 xstats[i].id = i;
3718                 xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
3719                         rte_ixgbevf_stats_strings[i].offset);
3720         }
3721
3722         return IXGBEVF_NB_XSTATS;
3723 }
3724
3725 static int
3726 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3727 {
3728         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3729                           IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3730
3731         ixgbevf_update_stats(dev);
3732
3733         if (stats == NULL)
3734                 return -EINVAL;
3735
3736         stats->ipackets = hw_stats->vfgprc;
3737         stats->ibytes = hw_stats->vfgorc;
3738         stats->opackets = hw_stats->vfgptc;
3739         stats->obytes = hw_stats->vfgotc;
3740         return 0;
3741 }
3742
3743 static void
3744 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3745 {
3746         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3747                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3748
3749         /* Sync HW register to the last stats */
3750         ixgbevf_dev_stats_get(dev, NULL);
3751
3752         /* reset HW current stats*/
3753         hw_stats->vfgprc = 0;
3754         hw_stats->vfgorc = 0;
3755         hw_stats->vfgptc = 0;
3756         hw_stats->vfgotc = 0;
3757 }
3758
3759 static int
3760 ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
3761 {
3762         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3763         u16 eeprom_verh, eeprom_verl;
3764         u32 etrack_id;
3765         int ret;
3766
3767         ixgbe_read_eeprom(hw, 0x2e, &eeprom_verh);
3768         ixgbe_read_eeprom(hw, 0x2d, &eeprom_verl);
3769
3770         etrack_id = (eeprom_verh << 16) | eeprom_verl;
3771         ret = snprintf(fw_version, fw_size, "0x%08x", etrack_id);
3772
3773         ret += 1; /* add the size of '\0' */
3774         if (fw_size < (u32)ret)
3775                 return ret;
3776         else
3777                 return 0;
3778 }
3779
3780 static void
3781 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3782 {
3783         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3784         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3785         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3786
3787         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3788         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3789         if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3790                 /*
3791                  * When DCB/VT is off, maximum number of queues changes,
3792                  * except for 82598EB, which remains constant.
3793                  */
3794                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
3795                                 hw->mac.type != ixgbe_mac_82598EB)
3796                         dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3797         }
3798         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3799         dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3800         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3801         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3802         dev_info->max_vfs = pci_dev->max_vfs;
3803         if (hw->mac.type == ixgbe_mac_82598EB)
3804                 dev_info->max_vmdq_pools = ETH_16_POOLS;
3805         else
3806                 dev_info->max_vmdq_pools = ETH_64_POOLS;
3807         dev_info->max_mtu =  dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3808         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
3809         dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3810         dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3811         dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3812                                      dev_info->rx_queue_offload_capa);
3813         dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3814         dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3815
3816         dev_info->default_rxconf = (struct rte_eth_rxconf) {
3817                 .rx_thresh = {
3818                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3819                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3820                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3821                 },
3822                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3823                 .rx_drop_en = 0,
3824                 .offloads = 0,
3825         };
3826
3827         dev_info->default_txconf = (struct rte_eth_txconf) {
3828                 .tx_thresh = {
3829                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3830                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3831                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3832                 },
3833                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3834                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3835                 .offloads = 0,
3836         };
3837
3838         dev_info->rx_desc_lim = rx_desc_lim;
3839         dev_info->tx_desc_lim = tx_desc_lim;
3840
3841         dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3842         dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3843         dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3844
3845         dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G;
3846         if (hw->mac.type == ixgbe_mac_X540 ||
3847             hw->mac.type == ixgbe_mac_X540_vf ||
3848             hw->mac.type == ixgbe_mac_X550 ||
3849             hw->mac.type == ixgbe_mac_X550_vf) {
3850                 dev_info->speed_capa |= ETH_LINK_SPEED_100M;
3851         }
3852         if (hw->mac.type == ixgbe_mac_X550) {
3853                 dev_info->speed_capa |= ETH_LINK_SPEED_2_5G;
3854                 dev_info->speed_capa |= ETH_LINK_SPEED_5G;
3855         }
3856
3857         /* Driver-preferred Rx/Tx parameters */
3858         dev_info->default_rxportconf.burst_size = 32;
3859         dev_info->default_txportconf.burst_size = 32;
3860         dev_info->default_rxportconf.nb_queues = 1;
3861         dev_info->default_txportconf.nb_queues = 1;
3862         dev_info->default_rxportconf.ring_size = 256;
3863         dev_info->default_txportconf.ring_size = 256;
3864 }
3865
3866 static const uint32_t *
3867 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3868 {
3869         static const uint32_t ptypes[] = {
3870                 /* For non-vec functions,
3871                  * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3872                  * for vec functions,
3873                  * refers to _recv_raw_pkts_vec().
3874                  */
3875                 RTE_PTYPE_L2_ETHER,
3876                 RTE_PTYPE_L3_IPV4,
3877                 RTE_PTYPE_L3_IPV4_EXT,
3878                 RTE_PTYPE_L3_IPV6,
3879                 RTE_PTYPE_L3_IPV6_EXT,
3880                 RTE_PTYPE_L4_SCTP,
3881                 RTE_PTYPE_L4_TCP,
3882                 RTE_PTYPE_L4_UDP,
3883                 RTE_PTYPE_TUNNEL_IP,
3884                 RTE_PTYPE_INNER_L3_IPV6,
3885                 RTE_PTYPE_INNER_L3_IPV6_EXT,
3886                 RTE_PTYPE_INNER_L4_TCP,
3887                 RTE_PTYPE_INNER_L4_UDP,
3888                 RTE_PTYPE_UNKNOWN
3889         };
3890
3891         if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3892             dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3893             dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3894             dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3895                 return ptypes;
3896
3897 #if defined(RTE_ARCH_X86)
3898         if (dev->rx_pkt_burst == ixgbe_recv_pkts_vec ||
3899             dev->rx_pkt_burst == ixgbe_recv_scattered_pkts_vec)
3900                 return ptypes;
3901 #endif
3902         return NULL;
3903 }
3904
3905 static void
3906 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3907                      struct rte_eth_dev_info *dev_info)
3908 {
3909         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3910         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3911
3912         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3913         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3914         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3915         dev_info->max_rx_pktlen = 9728; /* includes CRC, cf MAXFRS reg */
3916         dev_info->max_mtu = dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3917         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3918         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3919         dev_info->max_vfs = pci_dev->max_vfs;
3920         if (hw->mac.type == ixgbe_mac_82598EB)
3921                 dev_info->max_vmdq_pools = ETH_16_POOLS;
3922         else
3923                 dev_info->max_vmdq_pools = ETH_64_POOLS;
3924         dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3925         dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3926                                      dev_info->rx_queue_offload_capa);
3927         dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3928         dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3929
3930         dev_info->default_rxconf = (struct rte_eth_rxconf) {
3931                 .rx_thresh = {
3932                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3933                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3934                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3935                 },
3936                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3937                 .rx_drop_en = 0,
3938                 .offloads = 0,
3939         };
3940
3941         dev_info->default_txconf = (struct rte_eth_txconf) {
3942                 .tx_thresh = {
3943                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3944                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3945                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3946                 },
3947                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3948                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3949                 .offloads = 0,
3950         };
3951
3952         dev_info->rx_desc_lim = rx_desc_lim;
3953         dev_info->tx_desc_lim = tx_desc_lim;
3954 }
3955
3956 static int
3957 ixgbevf_check_link(struct ixgbe_hw *hw, ixgbe_link_speed *speed,
3958                    int *link_up, int wait_to_complete)
3959 {
3960         struct ixgbe_adapter *adapter = container_of(hw,
3961                                                      struct ixgbe_adapter, hw);
3962         struct ixgbe_mbx_info *mbx = &hw->mbx;
3963         struct ixgbe_mac_info *mac = &hw->mac;
3964         uint32_t links_reg, in_msg;
3965         int ret_val = 0;
3966
3967         /* If we were hit with a reset drop the link */
3968         if (!mbx->ops.check_for_rst(hw, 0) || !mbx->timeout)
3969                 mac->get_link_status = true;
3970
3971         if (!mac->get_link_status)
3972                 goto out;
3973
3974         /* if link status is down no point in checking to see if pf is up */
3975         links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
3976         if (!(links_reg & IXGBE_LINKS_UP))
3977                 goto out;
3978
3979         /* for SFP+ modules and DA cables on 82599 it can take up to 500usecs
3980          * before the link status is correct
3981          */
3982         if (mac->type == ixgbe_mac_82599_vf && wait_to_complete) {
3983                 int i;
3984
3985                 for (i = 0; i < 5; i++) {
3986                         rte_delay_us(100);
3987                         links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
3988
3989                         if (!(links_reg & IXGBE_LINKS_UP))
3990                                 goto out;
3991                 }
3992         }
3993
3994         switch (links_reg & IXGBE_LINKS_SPEED_82599) {
3995         case IXGBE_LINKS_SPEED_10G_82599:
3996                 *speed = IXGBE_LINK_SPEED_10GB_FULL;
3997                 if (hw->mac.type >= ixgbe_mac_X550) {
3998                         if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
3999                                 *speed = IXGBE_LINK_SPEED_2_5GB_FULL;
4000                 }
4001                 break;
4002         case IXGBE_LINKS_SPEED_1G_82599:
4003                 *speed = IXGBE_LINK_SPEED_1GB_FULL;
4004                 break;
4005         case IXGBE_LINKS_SPEED_100_82599:
4006                 *speed = IXGBE_LINK_SPEED_100_FULL;
4007                 if (hw->mac.type == ixgbe_mac_X550) {
4008                         if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4009                                 *speed = IXGBE_LINK_SPEED_5GB_FULL;
4010                 }
4011                 break;
4012         case IXGBE_LINKS_SPEED_10_X550EM_A:
4013                 *speed = IXGBE_LINK_SPEED_UNKNOWN;
4014                 /* Since Reserved in older MAC's */
4015                 if (hw->mac.type >= ixgbe_mac_X550)
4016                         *speed = IXGBE_LINK_SPEED_10_FULL;
4017                 break;
4018         default:
4019                 *speed = IXGBE_LINK_SPEED_UNKNOWN;
4020         }
4021
4022         if (wait_to_complete == 0 && adapter->pflink_fullchk == 0) {
4023                 if (*speed == IXGBE_LINK_SPEED_UNKNOWN)
4024                         mac->get_link_status = true;
4025                 else
4026                         mac->get_link_status = false;
4027
4028                 goto out;
4029         }
4030
4031         /* if the read failed it could just be a mailbox collision, best wait
4032          * until we are called again and don't report an error
4033          */
4034         if (mbx->ops.read(hw, &in_msg, 1, 0))
4035                 goto out;
4036
4037         if (!(in_msg & IXGBE_VT_MSGTYPE_CTS)) {
4038                 /* msg is not CTS and is NACK we must have lost CTS status */
4039                 if (in_msg & IXGBE_VT_MSGTYPE_NACK)
4040                         mac->get_link_status = false;
4041                 goto out;
4042         }
4043
4044         /* the pf is talking, if we timed out in the past we reinit */
4045         if (!mbx->timeout) {
4046                 ret_val = -1;
4047                 goto out;
4048         }
4049
4050         /* if we passed all the tests above then the link is up and we no
4051          * longer need to check for link
4052          */
4053         mac->get_link_status = false;
4054
4055 out:
4056         *link_up = !mac->get_link_status;
4057         return ret_val;
4058 }
4059
4060 static void
4061 ixgbe_dev_setup_link_alarm_handler(void *param)
4062 {
4063         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4064         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4065         struct ixgbe_interrupt *intr =
4066                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4067         u32 speed;
4068         bool autoneg = false;
4069
4070         speed = hw->phy.autoneg_advertised;
4071         if (!speed)
4072                 ixgbe_get_link_capabilities(hw, &speed, &autoneg);
4073
4074         ixgbe_setup_link(hw, speed, true);
4075
4076         intr->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4077 }
4078
4079 /* return 0 means link status changed, -1 means not changed */
4080 int
4081 ixgbe_dev_link_update_share(struct rte_eth_dev *dev,
4082                             int wait_to_complete, int vf)
4083 {
4084         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4085         struct rte_eth_link link;
4086         ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
4087         struct ixgbe_interrupt *intr =
4088                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4089         int link_up;
4090         int diag;
4091         int wait = 1;
4092
4093         memset(&link, 0, sizeof(link));
4094         link.link_status = ETH_LINK_DOWN;
4095         link.link_speed = ETH_SPEED_NUM_NONE;
4096         link.link_duplex = ETH_LINK_HALF_DUPLEX;
4097         link.link_autoneg = ETH_LINK_AUTONEG;
4098
4099         hw->mac.get_link_status = true;
4100
4101         if (intr->flags & IXGBE_FLAG_NEED_LINK_CONFIG)
4102                 return rte_eth_linkstatus_set(dev, &link);
4103
4104         /* check if it needs to wait to complete, if lsc interrupt is enabled */
4105         if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
4106                 wait = 0;
4107
4108         if (vf)
4109                 diag = ixgbevf_check_link(hw, &link_speed, &link_up, wait);
4110         else
4111                 diag = ixgbe_check_link(hw, &link_speed, &link_up, wait);
4112
4113         if (diag != 0) {
4114                 link.link_speed = ETH_SPEED_NUM_100M;
4115                 link.link_duplex = ETH_LINK_FULL_DUPLEX;
4116                 return rte_eth_linkstatus_set(dev, &link);
4117         }
4118
4119         if (link_up == 0) {
4120                 if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4121                         intr->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
4122                         rte_eal_alarm_set(10,
4123                                 ixgbe_dev_setup_link_alarm_handler, dev);
4124                 }
4125                 return rte_eth_linkstatus_set(dev, &link);
4126         }
4127
4128         link.link_status = ETH_LINK_UP;
4129         link.link_duplex = ETH_LINK_FULL_DUPLEX;
4130
4131         switch (link_speed) {
4132         default:
4133         case IXGBE_LINK_SPEED_UNKNOWN:
4134                 if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
4135                         hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
4136                         link.link_speed = ETH_SPEED_NUM_10M;
4137                 else
4138                         link.link_speed = ETH_SPEED_NUM_100M;
4139                 link.link_duplex = ETH_LINK_FULL_DUPLEX;
4140                 break;
4141
4142         case IXGBE_LINK_SPEED_100_FULL:
4143                 link.link_speed = ETH_SPEED_NUM_100M;
4144                 break;
4145
4146         case IXGBE_LINK_SPEED_1GB_FULL:
4147                 link.link_speed = ETH_SPEED_NUM_1G;
4148                 break;
4149
4150         case IXGBE_LINK_SPEED_2_5GB_FULL:
4151                 link.link_speed = ETH_SPEED_NUM_2_5G;
4152                 break;
4153
4154         case IXGBE_LINK_SPEED_5GB_FULL:
4155                 link.link_speed = ETH_SPEED_NUM_5G;
4156                 break;
4157
4158         case IXGBE_LINK_SPEED_10GB_FULL:
4159                 link.link_speed = ETH_SPEED_NUM_10G;
4160                 break;
4161         }
4162
4163         return rte_eth_linkstatus_set(dev, &link);
4164 }
4165
4166 static int
4167 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4168 {
4169         return ixgbe_dev_link_update_share(dev, wait_to_complete, 0);
4170 }
4171
4172 static int
4173 ixgbevf_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4174 {
4175         return ixgbe_dev_link_update_share(dev, wait_to_complete, 1);
4176 }
4177
4178 static void
4179 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
4180 {
4181         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4182         uint32_t fctrl;
4183
4184         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4185         fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4186         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4187 }
4188
4189 static void
4190 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
4191 {
4192         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4193         uint32_t fctrl;
4194
4195         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4196         fctrl &= (~IXGBE_FCTRL_UPE);
4197         if (dev->data->all_multicast == 1)
4198                 fctrl |= IXGBE_FCTRL_MPE;
4199         else
4200                 fctrl &= (~IXGBE_FCTRL_MPE);
4201         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4202 }
4203
4204 static void
4205 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
4206 {
4207         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4208         uint32_t fctrl;
4209
4210         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4211         fctrl |= IXGBE_FCTRL_MPE;
4212         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4213 }
4214
4215 static void
4216 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
4217 {
4218         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4219         uint32_t fctrl;
4220
4221         if (dev->data->promiscuous == 1)
4222                 return; /* must remain in all_multicast mode */
4223
4224         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4225         fctrl &= (~IXGBE_FCTRL_MPE);
4226         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4227 }
4228
4229 /**
4230  * It clears the interrupt causes and enables the interrupt.
4231  * It will be called once only during nic initialized.
4232  *
4233  * @param dev
4234  *  Pointer to struct rte_eth_dev.
4235  * @param on
4236  *  Enable or Disable.
4237  *
4238  * @return
4239  *  - On success, zero.
4240  *  - On failure, a negative value.
4241  */
4242 static int
4243 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
4244 {
4245         struct ixgbe_interrupt *intr =
4246                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4247
4248         ixgbe_dev_link_status_print(dev);
4249         if (on)
4250                 intr->mask |= IXGBE_EICR_LSC;
4251         else
4252                 intr->mask &= ~IXGBE_EICR_LSC;
4253
4254         return 0;
4255 }
4256
4257 /**
4258  * It clears the interrupt causes and enables the interrupt.
4259  * It will be called once only during nic initialized.
4260  *
4261  * @param dev
4262  *  Pointer to struct rte_eth_dev.
4263  *
4264  * @return
4265  *  - On success, zero.
4266  *  - On failure, a negative value.
4267  */
4268 static int
4269 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
4270 {
4271         struct ixgbe_interrupt *intr =
4272                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4273
4274         intr->mask |= IXGBE_EICR_RTX_QUEUE;
4275
4276         return 0;
4277 }
4278
4279 /**
4280  * It clears the interrupt causes and enables the interrupt.
4281  * It will be called once only during nic initialized.
4282  *
4283  * @param dev
4284  *  Pointer to struct rte_eth_dev.
4285  *
4286  * @return
4287  *  - On success, zero.
4288  *  - On failure, a negative value.
4289  */
4290 static int
4291 ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev)
4292 {
4293         struct ixgbe_interrupt *intr =
4294                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4295
4296         intr->mask |= IXGBE_EICR_LINKSEC;
4297
4298         return 0;
4299 }
4300
4301 /*
4302  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
4303  *
4304  * @param dev
4305  *  Pointer to struct rte_eth_dev.
4306  *
4307  * @return
4308  *  - On success, zero.
4309  *  - On failure, a negative value.
4310  */
4311 static int
4312 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
4313 {
4314         uint32_t eicr;
4315         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4316         struct ixgbe_interrupt *intr =
4317                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4318
4319         /* clear all cause mask */
4320         ixgbe_disable_intr(hw);
4321
4322         /* read-on-clear nic registers here */
4323         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4324         PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
4325
4326         intr->flags = 0;
4327
4328         /* set flag for async link update */
4329         if (eicr & IXGBE_EICR_LSC)
4330                 intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4331
4332         if (eicr & IXGBE_EICR_MAILBOX)
4333                 intr->flags |= IXGBE_FLAG_MAILBOX;
4334
4335         if (eicr & IXGBE_EICR_LINKSEC)
4336                 intr->flags |= IXGBE_FLAG_MACSEC;
4337
4338         if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
4339             hw->phy.type == ixgbe_phy_x550em_ext_t &&
4340             (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
4341                 intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
4342
4343         return 0;
4344 }
4345
4346 /**
4347  * It gets and then prints the link status.
4348  *
4349  * @param dev
4350  *  Pointer to struct rte_eth_dev.
4351  *
4352  * @return
4353  *  - On success, zero.
4354  *  - On failure, a negative value.
4355  */
4356 static void
4357 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
4358 {
4359         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4360         struct rte_eth_link link;
4361
4362         rte_eth_linkstatus_get(dev, &link);
4363
4364         if (link.link_status) {
4365                 PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
4366                                         (int)(dev->data->port_id),
4367                                         (unsigned)link.link_speed,
4368                         link.link_duplex == ETH_LINK_FULL_DUPLEX ?
4369                                         "full-duplex" : "half-duplex");
4370         } else {
4371                 PMD_INIT_LOG(INFO, " Port %d: Link Down",
4372                                 (int)(dev->data->port_id));
4373         }
4374         PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
4375                                 pci_dev->addr.domain,
4376                                 pci_dev->addr.bus,
4377                                 pci_dev->addr.devid,
4378                                 pci_dev->addr.function);
4379 }
4380
4381 /*
4382  * It executes link_update after knowing an interrupt occurred.
4383  *
4384  * @param dev
4385  *  Pointer to struct rte_eth_dev.
4386  *
4387  * @return
4388  *  - On success, zero.
4389  *  - On failure, a negative value.
4390  */
4391 static int
4392 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
4393 {
4394         struct ixgbe_interrupt *intr =
4395                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4396         int64_t timeout;
4397         struct ixgbe_hw *hw =
4398                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4399
4400         PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
4401
4402         if (intr->flags & IXGBE_FLAG_MAILBOX) {
4403                 ixgbe_pf_mbx_process(dev);
4404                 intr->flags &= ~IXGBE_FLAG_MAILBOX;
4405         }
4406
4407         if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4408                 ixgbe_handle_lasi(hw);
4409                 intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4410         }
4411
4412         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4413                 struct rte_eth_link link;
4414
4415                 /* get the link status before link update, for predicting later */
4416                 rte_eth_linkstatus_get(dev, &link);
4417
4418                 ixgbe_dev_link_update(dev, 0);
4419
4420                 /* likely to up */
4421                 if (!link.link_status)
4422                         /* handle it 1 sec later, wait it being stable */
4423                         timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
4424                 /* likely to down */
4425                 else
4426                         /* handle it 4 sec later, wait it being stable */
4427                         timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
4428
4429                 ixgbe_dev_link_status_print(dev);
4430                 if (rte_eal_alarm_set(timeout * 1000,
4431                                       ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
4432                         PMD_DRV_LOG(ERR, "Error setting alarm");
4433                 else {
4434                         /* remember original mask */
4435                         intr->mask_original = intr->mask;
4436                         /* only disable lsc interrupt */
4437                         intr->mask &= ~IXGBE_EIMS_LSC;
4438                 }
4439         }
4440
4441         PMD_DRV_LOG(DEBUG, "enable intr immediately");
4442         ixgbe_enable_intr(dev);
4443
4444         return 0;
4445 }
4446
4447 /**
4448  * Interrupt handler which shall be registered for alarm callback for delayed
4449  * handling specific interrupt to wait for the stable nic state. As the
4450  * NIC interrupt state is not stable for ixgbe after link is just down,
4451  * it needs to wait 4 seconds to get the stable status.
4452  *
4453  * @param handle
4454  *  Pointer to interrupt handle.
4455  * @param param
4456  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4457  *
4458  * @return
4459  *  void
4460  */
4461 static void
4462 ixgbe_dev_interrupt_delayed_handler(void *param)
4463 {
4464         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4465         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4466         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
4467         struct ixgbe_interrupt *intr =
4468                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4469         struct ixgbe_hw *hw =
4470                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4471         uint32_t eicr;
4472
4473         ixgbe_disable_intr(hw);
4474
4475         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4476         if (eicr & IXGBE_EICR_MAILBOX)
4477                 ixgbe_pf_mbx_process(dev);
4478
4479         if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4480                 ixgbe_handle_lasi(hw);
4481                 intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4482         }
4483
4484         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4485                 ixgbe_dev_link_update(dev, 0);
4486                 intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4487                 ixgbe_dev_link_status_print(dev);
4488                 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC,
4489                                               NULL);
4490         }
4491
4492         if (intr->flags & IXGBE_FLAG_MACSEC) {
4493                 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_MACSEC,
4494                                               NULL);
4495                 intr->flags &= ~IXGBE_FLAG_MACSEC;
4496         }
4497
4498         /* restore original mask */
4499         intr->mask = intr->mask_original;
4500         intr->mask_original = 0;
4501
4502         PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
4503         ixgbe_enable_intr(dev);
4504         rte_intr_enable(intr_handle);
4505 }
4506
4507 /**
4508  * Interrupt handler triggered by NIC  for handling
4509  * specific interrupt.
4510  *
4511  * @param handle
4512  *  Pointer to interrupt handle.
4513  * @param param
4514  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4515  *
4516  * @return
4517  *  void
4518  */
4519 static void
4520 ixgbe_dev_interrupt_handler(void *param)
4521 {
4522         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4523
4524         ixgbe_dev_interrupt_get_status(dev);
4525         ixgbe_dev_interrupt_action(dev);
4526 }
4527
4528 static int
4529 ixgbe_dev_led_on(struct rte_eth_dev *dev)
4530 {
4531         struct ixgbe_hw *hw;
4532
4533         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4534         return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4535 }
4536
4537 static int
4538 ixgbe_dev_led_off(struct rte_eth_dev *dev)
4539 {
4540         struct ixgbe_hw *hw;
4541
4542         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4543         return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4544 }
4545
4546 static int
4547 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4548 {
4549         struct ixgbe_hw *hw;
4550         uint32_t mflcn_reg;
4551         uint32_t fccfg_reg;
4552         int rx_pause;
4553         int tx_pause;
4554
4555         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4556
4557         fc_conf->pause_time = hw->fc.pause_time;
4558         fc_conf->high_water = hw->fc.high_water[0];
4559         fc_conf->low_water = hw->fc.low_water[0];
4560         fc_conf->send_xon = hw->fc.send_xon;
4561         fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
4562
4563         /*
4564          * Return rx_pause status according to actual setting of
4565          * MFLCN register.
4566          */
4567         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4568         if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
4569                 rx_pause = 1;
4570         else
4571                 rx_pause = 0;
4572
4573         /*
4574          * Return tx_pause status according to actual setting of
4575          * FCCFG register.
4576          */
4577         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4578         if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
4579                 tx_pause = 1;
4580         else
4581                 tx_pause = 0;
4582
4583         if (rx_pause && tx_pause)
4584                 fc_conf->mode = RTE_FC_FULL;
4585         else if (rx_pause)
4586                 fc_conf->mode = RTE_FC_RX_PAUSE;
4587         else if (tx_pause)
4588                 fc_conf->mode = RTE_FC_TX_PAUSE;
4589         else
4590                 fc_conf->mode = RTE_FC_NONE;
4591
4592         return 0;
4593 }
4594
4595 static int
4596 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4597 {
4598         struct ixgbe_hw *hw;
4599         int err;
4600         uint32_t rx_buf_size;
4601         uint32_t max_high_water;
4602         uint32_t mflcn;
4603         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4604                 ixgbe_fc_none,
4605                 ixgbe_fc_rx_pause,
4606                 ixgbe_fc_tx_pause,
4607                 ixgbe_fc_full
4608         };
4609
4610         PMD_INIT_FUNC_TRACE();
4611
4612         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4613         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
4614         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4615
4616         /*
4617          * At least reserve one Ethernet frame for watermark
4618          * high_water/low_water in kilo bytes for ixgbe
4619          */
4620         max_high_water = (rx_buf_size -
4621                         RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4622         if ((fc_conf->high_water > max_high_water) ||
4623                 (fc_conf->high_water < fc_conf->low_water)) {
4624                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4625                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4626                 return -EINVAL;
4627         }
4628
4629         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
4630         hw->fc.pause_time     = fc_conf->pause_time;
4631         hw->fc.high_water[0]  = fc_conf->high_water;
4632         hw->fc.low_water[0]   = fc_conf->low_water;
4633         hw->fc.send_xon       = fc_conf->send_xon;
4634         hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
4635
4636         err = ixgbe_fc_enable(hw);
4637
4638         /* Not negotiated is not an error case */
4639         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED)) {
4640
4641                 /* check if we want to forward MAC frames - driver doesn't have native
4642                  * capability to do that, so we'll write the registers ourselves */
4643
4644                 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4645
4646                 /* set or clear MFLCN.PMCF bit depending on configuration */
4647                 if (fc_conf->mac_ctrl_frame_fwd != 0)
4648                         mflcn |= IXGBE_MFLCN_PMCF;
4649                 else
4650                         mflcn &= ~IXGBE_MFLCN_PMCF;
4651
4652                 IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
4653                 IXGBE_WRITE_FLUSH(hw);
4654
4655                 return 0;
4656         }
4657
4658         PMD_INIT_LOG(ERR, "ixgbe_fc_enable = 0x%x", err);
4659         return -EIO;
4660 }
4661
4662 /**
4663  *  ixgbe_pfc_enable_generic - Enable flow control
4664  *  @hw: pointer to hardware structure
4665  *  @tc_num: traffic class number
4666  *  Enable flow control according to the current settings.
4667  */
4668 static int
4669 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
4670 {
4671         int ret_val = 0;
4672         uint32_t mflcn_reg, fccfg_reg;
4673         uint32_t reg;
4674         uint32_t fcrtl, fcrth;
4675         uint8_t i;
4676         uint8_t nb_rx_en;
4677
4678         /* Validate the water mark configuration */
4679         if (!hw->fc.pause_time) {
4680                 ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4681                 goto out;
4682         }
4683
4684         /* Low water mark of zero causes XOFF floods */
4685         if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
4686                  /* High/Low water can not be 0 */
4687                 if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_water[tc_num])) {
4688                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4689                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4690                         goto out;
4691                 }
4692
4693                 if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
4694                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4695                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4696                         goto out;
4697                 }
4698         }
4699         /* Negotiate the fc mode to use */
4700         ixgbe_fc_autoneg(hw);
4701
4702         /* Disable any previous flow control settings */
4703         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4704         mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
4705
4706         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4707         fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
4708
4709         switch (hw->fc.current_mode) {
4710         case ixgbe_fc_none:
4711                 /*
4712                  * If the count of enabled RX Priority Flow control >1,
4713                  * and the TX pause can not be disabled
4714                  */
4715                 nb_rx_en = 0;
4716                 for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4717                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4718                         if (reg & IXGBE_FCRTH_FCEN)
4719                                 nb_rx_en++;
4720                 }
4721                 if (nb_rx_en > 1)
4722                         fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4723                 break;
4724         case ixgbe_fc_rx_pause:
4725                 /*
4726                  * Rx Flow control is enabled and Tx Flow control is
4727                  * disabled by software override. Since there really
4728                  * isn't a way to advertise that we are capable of RX
4729                  * Pause ONLY, we will advertise that we support both
4730                  * symmetric and asymmetric Rx PAUSE.  Later, we will
4731                  * disable the adapter's ability to send PAUSE frames.
4732                  */
4733                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
4734                 /*
4735                  * If the count of enabled RX Priority Flow control >1,
4736                  * and the TX pause can not be disabled
4737                  */
4738                 nb_rx_en = 0;
4739                 for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4740                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4741                         if (reg & IXGBE_FCRTH_FCEN)
4742                                 nb_rx_en++;
4743                 }
4744                 if (nb_rx_en > 1)
4745                         fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4746                 break;
4747         case ixgbe_fc_tx_pause:
4748                 /*
4749                  * Tx Flow control is enabled, and Rx Flow control is
4750                  * disabled by software override.
4751                  */
4752                 fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4753                 break;
4754         case ixgbe_fc_full:
4755                 /* Flow control (both Rx and Tx) is enabled by SW override. */
4756                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
4757                 fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4758                 break;
4759         default:
4760                 PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
4761                 ret_val = IXGBE_ERR_CONFIG;
4762                 goto out;
4763         }
4764
4765         /* Set 802.3x based flow control settings. */
4766         mflcn_reg |= IXGBE_MFLCN_DPF;
4767         IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
4768         IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
4769
4770         /* Set up and enable Rx high/low water mark thresholds, enable XON. */
4771         if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
4772                 hw->fc.high_water[tc_num]) {
4773                 fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
4774                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
4775                 fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
4776         } else {
4777                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
4778                 /*
4779                  * In order to prevent Tx hangs when the internal Tx
4780                  * switch is enabled we must set the high water mark
4781                  * to the maximum FCRTH value.  This allows the Tx
4782                  * switch to function even under heavy Rx workloads.
4783                  */
4784                 fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
4785         }
4786         IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
4787
4788         /* Configure pause time (2 TCs per register) */
4789         reg = hw->fc.pause_time * 0x00010001;
4790         for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
4791                 IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
4792
4793         /* Configure flow control refresh threshold value */
4794         IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
4795
4796 out:
4797         return ret_val;
4798 }
4799
4800 static int
4801 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
4802 {
4803         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4804         int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
4805
4806         if (hw->mac.type != ixgbe_mac_82598EB) {
4807                 ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
4808         }
4809         return ret_val;
4810 }
4811
4812 static int
4813 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
4814 {
4815         int err;
4816         uint32_t rx_buf_size;
4817         uint32_t max_high_water;
4818         uint8_t tc_num;
4819         uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
4820         struct ixgbe_hw *hw =
4821                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4822         struct ixgbe_dcb_config *dcb_config =
4823                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
4824
4825         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4826                 ixgbe_fc_none,
4827                 ixgbe_fc_rx_pause,
4828                 ixgbe_fc_tx_pause,
4829                 ixgbe_fc_full
4830         };
4831
4832         PMD_INIT_FUNC_TRACE();
4833
4834         ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
4835         tc_num = map[pfc_conf->priority];
4836         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
4837         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4838         /*
4839          * At least reserve one Ethernet frame for watermark
4840          * high_water/low_water in kilo bytes for ixgbe
4841          */
4842         max_high_water = (rx_buf_size -
4843                         RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4844         if ((pfc_conf->fc.high_water > max_high_water) ||
4845             (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
4846                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4847                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4848                 return -EINVAL;
4849         }
4850
4851         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
4852         hw->fc.pause_time = pfc_conf->fc.pause_time;
4853         hw->fc.send_xon = pfc_conf->fc.send_xon;
4854         hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
4855         hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
4856
4857         err = ixgbe_dcb_pfc_enable(dev, tc_num);
4858
4859         /* Not negotiated is not an error case */
4860         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
4861                 return 0;
4862
4863         PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
4864         return -EIO;
4865 }
4866
4867 static int
4868 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
4869                           struct rte_eth_rss_reta_entry64 *reta_conf,
4870                           uint16_t reta_size)
4871 {
4872         uint16_t i, sp_reta_size;
4873         uint8_t j, mask;
4874         uint32_t reta, r;
4875         uint16_t idx, shift;
4876         struct ixgbe_adapter *adapter = dev->data->dev_private;
4877         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4878         uint32_t reta_reg;
4879
4880         PMD_INIT_FUNC_TRACE();
4881
4882         if (!ixgbe_rss_update_sp(hw->mac.type)) {
4883                 PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
4884                         "NIC.");
4885                 return -ENOTSUP;
4886         }
4887
4888         sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
4889         if (reta_size != sp_reta_size) {
4890                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
4891                         "(%d) doesn't match the number hardware can supported "
4892                         "(%d)", reta_size, sp_reta_size);
4893                 return -EINVAL;
4894         }
4895
4896         for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
4897                 idx = i / RTE_RETA_GROUP_SIZE;
4898                 shift = i % RTE_RETA_GROUP_SIZE;
4899                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
4900                                                 IXGBE_4_BIT_MASK);
4901                 if (!mask)
4902                         continue;
4903                 reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
4904                 if (mask == IXGBE_4_BIT_MASK)
4905                         r = 0;
4906                 else
4907                         r = IXGBE_READ_REG(hw, reta_reg);
4908                 for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
4909                         if (mask & (0x1 << j))
4910                                 reta |= reta_conf[idx].reta[shift + j] <<
4911                                                         (CHAR_BIT * j);
4912                         else
4913                                 reta |= r & (IXGBE_8_BIT_MASK <<
4914                                                 (CHAR_BIT * j));
4915                 }
4916                 IXGBE_WRITE_REG(hw, reta_reg, reta);
4917         }
4918         adapter->rss_reta_updated = 1;
4919
4920         return 0;
4921 }
4922
4923 static int
4924 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
4925                          struct rte_eth_rss_reta_entry64 *reta_conf,
4926                          uint16_t reta_size)
4927 {
4928         uint16_t i, sp_reta_size;
4929         uint8_t j, mask;
4930         uint32_t reta;
4931         uint16_t idx, shift;
4932         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4933         uint32_t reta_reg;
4934
4935         PMD_INIT_FUNC_TRACE();
4936         sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
4937         if (reta_size != sp_reta_size) {
4938                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
4939                         "(%d) doesn't match the number hardware can supported "
4940                         "(%d)", reta_size, sp_reta_size);
4941                 return -EINVAL;
4942         }
4943
4944         for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
4945                 idx = i / RTE_RETA_GROUP_SIZE;
4946                 shift = i % RTE_RETA_GROUP_SIZE;
4947                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
4948                                                 IXGBE_4_BIT_MASK);
4949                 if (!mask)
4950                         continue;
4951
4952                 reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
4953                 reta = IXGBE_READ_REG(hw, reta_reg);
4954                 for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
4955                         if (mask & (0x1 << j))
4956                                 reta_conf[idx].reta[shift + j] =
4957                                         ((reta >> (CHAR_BIT * j)) &
4958                                                 IXGBE_8_BIT_MASK);
4959                 }
4960         }
4961
4962         return 0;
4963 }
4964
4965 static int
4966 ixgbe_add_rar(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
4967                                 uint32_t index, uint32_t pool)
4968 {
4969         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4970         uint32_t enable_addr = 1;
4971
4972         return ixgbe_set_rar(hw, index, mac_addr->addr_bytes,
4973                              pool, enable_addr);
4974 }
4975
4976 static void
4977 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
4978 {
4979         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4980
4981         ixgbe_clear_rar(hw, index);
4982 }
4983
4984 static int
4985 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
4986 {
4987         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4988
4989         ixgbe_remove_rar(dev, 0);
4990         ixgbe_add_rar(dev, addr, 0, pci_dev->max_vfs);
4991
4992         return 0;
4993 }
4994
4995 static bool
4996 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
4997 {
4998         if (strcmp(dev->device->driver->name, drv->driver.name))
4999                 return false;
5000
5001         return true;
5002 }
5003
5004 bool
5005 is_ixgbe_supported(struct rte_eth_dev *dev)
5006 {
5007         return is_device_supported(dev, &rte_ixgbe_pmd);
5008 }
5009
5010 static int
5011 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
5012 {
5013         uint32_t hlreg0;
5014         uint32_t maxfrs;
5015         struct ixgbe_hw *hw;
5016         struct rte_eth_dev_info dev_info;
5017         uint32_t frame_size = mtu + IXGBE_ETH_OVERHEAD;
5018         struct rte_eth_dev_data *dev_data = dev->data;
5019
5020         ixgbe_dev_info_get(dev, &dev_info);
5021
5022         /* check that mtu is within the allowed range */
5023         if (mtu < RTE_ETHER_MIN_MTU || frame_size > dev_info.max_rx_pktlen)
5024                 return -EINVAL;
5025
5026         /* If device is started, refuse mtu that requires the support of
5027          * scattered packets when this feature has not been enabled before.
5028          */
5029         if (dev_data->dev_started && !dev_data->scattered_rx &&
5030             (frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
5031              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
5032                 PMD_INIT_LOG(ERR, "Stop port first.");
5033                 return -EINVAL;
5034         }
5035
5036         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5037         hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
5038
5039         /* switch to jumbo mode if needed */
5040         if (frame_size > RTE_ETHER_MAX_LEN) {
5041                 dev->data->dev_conf.rxmode.offloads |=
5042                         DEV_RX_OFFLOAD_JUMBO_FRAME;
5043                 hlreg0 |= IXGBE_HLREG0_JUMBOEN;
5044         } else {
5045                 dev->data->dev_conf.rxmode.offloads &=
5046                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
5047                 hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
5048         }
5049         IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
5050
5051         /* update max frame size */
5052         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
5053
5054         maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
5055         maxfrs &= 0x0000FFFF;
5056         maxfrs |= (dev->data->dev_conf.rxmode.max_rx_pkt_len << 16);
5057         IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
5058
5059         return 0;
5060 }
5061
5062 /*
5063  * Virtual Function operations
5064  */
5065 static void
5066 ixgbevf_intr_disable(struct rte_eth_dev *dev)
5067 {
5068         struct ixgbe_interrupt *intr =
5069                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5070         struct ixgbe_hw *hw =
5071                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5072
5073         PMD_INIT_FUNC_TRACE();
5074
5075         /* Clear interrupt mask to stop from interrupts being generated */
5076         IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
5077
5078         IXGBE_WRITE_FLUSH(hw);
5079
5080         /* Clear mask value. */
5081         intr->mask = 0;
5082 }
5083
5084 static void
5085 ixgbevf_intr_enable(struct rte_eth_dev *dev)
5086 {
5087         struct ixgbe_interrupt *intr =
5088                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5089         struct ixgbe_hw *hw =
5090                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5091
5092         PMD_INIT_FUNC_TRACE();
5093
5094         /* VF enable interrupt autoclean */
5095         IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
5096         IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
5097         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
5098
5099         IXGBE_WRITE_FLUSH(hw);
5100
5101         /* Save IXGBE_VTEIMS value to mask. */
5102         intr->mask = IXGBE_VF_IRQ_ENABLE_MASK;
5103 }
5104
5105 static int
5106 ixgbevf_dev_configure(struct rte_eth_dev *dev)
5107 {
5108         struct rte_eth_conf *conf = &dev->data->dev_conf;
5109         struct ixgbe_adapter *adapter = dev->data->dev_private;
5110
5111         PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
5112                      dev->data->port_id);
5113
5114         /*
5115          * VF has no ability to enable/disable HW CRC
5116          * Keep the persistent behavior the same as Host PF
5117          */
5118 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
5119         if (conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
5120                 PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
5121                 conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_KEEP_CRC;
5122         }
5123 #else
5124         if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC)) {
5125                 PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
5126                 conf->rxmode.offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
5127         }
5128 #endif
5129
5130         /*
5131          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
5132          * allocation or vector Rx preconditions we will reset it.
5133          */
5134         adapter->rx_bulk_alloc_allowed = true;
5135         adapter->rx_vec_allowed = true;
5136
5137         return 0;
5138 }
5139
5140 static int
5141 ixgbevf_dev_start(struct rte_eth_dev *dev)
5142 {
5143         struct ixgbe_hw *hw =
5144                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5145         uint32_t intr_vector = 0;
5146         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5147         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5148
5149         int err, mask = 0;
5150
5151         PMD_INIT_FUNC_TRACE();
5152
5153         /* Stop the link setup handler before resetting the HW. */
5154         rte_eal_alarm_cancel(ixgbe_dev_setup_link_alarm_handler, dev);
5155
5156         err = hw->mac.ops.reset_hw(hw);
5157         if (err) {
5158                 PMD_INIT_LOG(ERR, "Unable to reset vf hardware (%d)", err);
5159                 return err;
5160         }
5161         hw->mac.get_link_status = true;
5162
5163         /* negotiate mailbox API version to use with the PF. */
5164         ixgbevf_negotiate_api(hw);
5165
5166         ixgbevf_dev_tx_init(dev);
5167
5168         /* This can fail when allocating mbufs for descriptor rings */
5169         err = ixgbevf_dev_rx_init(dev);
5170         if (err) {
5171                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
5172                 ixgbe_dev_clear_queues(dev);
5173                 return err;
5174         }
5175
5176         /* Set vfta */
5177         ixgbevf_set_vfta_all(dev, 1);
5178
5179         /* Set HW strip */
5180         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
5181                 ETH_VLAN_EXTEND_MASK;
5182         err = ixgbevf_vlan_offload_config(dev, mask);
5183         if (err) {
5184                 PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
5185                 ixgbe_dev_clear_queues(dev);
5186                 return err;
5187         }
5188
5189         ixgbevf_dev_rxtx_start(dev);
5190
5191         /* check and configure queue intr-vector mapping */
5192         if (rte_intr_cap_multiple(intr_handle) &&
5193             dev->data->dev_conf.intr_conf.rxq) {
5194                 /* According to datasheet, only vector 0/1/2 can be used,
5195                  * now only one vector is used for Rx queue
5196                  */
5197                 intr_vector = 1;
5198                 if (rte_intr_efd_enable(intr_handle, intr_vector))
5199                         return -1;
5200         }
5201
5202         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
5203                 intr_handle->intr_vec =
5204                         rte_zmalloc("intr_vec",
5205                                     dev->data->nb_rx_queues * sizeof(int), 0);
5206                 if (intr_handle->intr_vec == NULL) {
5207                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
5208                                      " intr_vec", dev->data->nb_rx_queues);
5209                         return -ENOMEM;
5210                 }
5211         }
5212         ixgbevf_configure_msix(dev);
5213
5214         /* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
5215          * is mapped to VFIO vector 0 in eth_ixgbevf_dev_init( ).
5216          * If previous VFIO interrupt mapping setting in eth_ixgbevf_dev_init( )
5217          * is not cleared, it will fail when following rte_intr_enable( ) tries
5218          * to map Rx queue interrupt to other VFIO vectors.
5219          * So clear uio/vfio intr/evevnfd first to avoid failure.
5220          */
5221         rte_intr_disable(intr_handle);
5222
5223         rte_intr_enable(intr_handle);
5224
5225         /* Re-enable interrupt for VF */
5226         ixgbevf_intr_enable(dev);
5227
5228         /*
5229          * Update link status right before return, because it may
5230          * start link configuration process in a separate thread.
5231          */
5232         ixgbevf_dev_link_update(dev, 0);
5233
5234         return 0;
5235 }
5236
5237 static void
5238 ixgbevf_dev_stop(struct rte_eth_dev *dev)
5239 {
5240         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5241         struct ixgbe_adapter *adapter = dev->data->dev_private;
5242         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5243         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5244
5245         PMD_INIT_FUNC_TRACE();
5246
5247         rte_eal_alarm_cancel(ixgbe_dev_setup_link_alarm_handler, dev);
5248
5249         ixgbevf_intr_disable(dev);
5250
5251         hw->adapter_stopped = 1;
5252         ixgbe_stop_adapter(hw);
5253
5254         /*
5255           * Clear what we set, but we still keep shadow_vfta to
5256           * restore after device starts
5257           */
5258         ixgbevf_set_vfta_all(dev, 0);
5259
5260         /* Clear stored conf */
5261         dev->data->scattered_rx = 0;
5262
5263         ixgbe_dev_clear_queues(dev);
5264
5265         /* Clean datapath event and queue/vec mapping */
5266         rte_intr_efd_disable(intr_handle);
5267         if (intr_handle->intr_vec != NULL) {
5268                 rte_free(intr_handle->intr_vec);
5269                 intr_handle->intr_vec = NULL;
5270         }
5271
5272         adapter->rss_reta_updated = 0;
5273 }
5274
5275 static void
5276 ixgbevf_dev_close(struct rte_eth_dev *dev)
5277 {
5278         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5279
5280         PMD_INIT_FUNC_TRACE();
5281
5282         ixgbe_reset_hw(hw);
5283
5284         ixgbevf_dev_stop(dev);
5285
5286         ixgbe_dev_free_queues(dev);
5287
5288         /**
5289          * Remove the VF MAC address ro ensure
5290          * that the VF traffic goes to the PF
5291          * after stop, close and detach of the VF
5292          **/
5293         ixgbevf_remove_mac_addr(dev, 0);
5294 }
5295
5296 /*
5297  * Reset VF device
5298  */
5299 static int
5300 ixgbevf_dev_reset(struct rte_eth_dev *dev)
5301 {
5302         int ret;
5303
5304         ret = eth_ixgbevf_dev_uninit(dev);
5305         if (ret)
5306                 return ret;
5307
5308         ret = eth_ixgbevf_dev_init(dev);
5309
5310         return ret;
5311 }
5312
5313 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
5314 {
5315         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5316         struct ixgbe_vfta *shadow_vfta =
5317                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5318         int i = 0, j = 0, vfta = 0, mask = 1;
5319
5320         for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
5321                 vfta = shadow_vfta->vfta[i];
5322                 if (vfta) {
5323                         mask = 1;
5324                         for (j = 0; j < 32; j++) {
5325                                 if (vfta & mask)
5326                                         ixgbe_set_vfta(hw, (i<<5)+j, 0,
5327                                                        on, false);
5328                                 mask <<= 1;
5329                         }
5330                 }
5331         }
5332
5333 }
5334
5335 static int
5336 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
5337 {
5338         struct ixgbe_hw *hw =
5339                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5340         struct ixgbe_vfta *shadow_vfta =
5341                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5342         uint32_t vid_idx = 0;
5343         uint32_t vid_bit = 0;
5344         int ret = 0;
5345
5346         PMD_INIT_FUNC_TRACE();
5347
5348         /* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
5349         ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
5350         if (ret) {
5351                 PMD_INIT_LOG(ERR, "Unable to set VF vlan");
5352                 return ret;
5353         }
5354         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
5355         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
5356
5357         /* Save what we set and retore it after device reset */
5358         if (on)
5359                 shadow_vfta->vfta[vid_idx] |= vid_bit;
5360         else
5361                 shadow_vfta->vfta[vid_idx] &= ~vid_bit;
5362
5363         return 0;
5364 }
5365
5366 static void
5367 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
5368 {
5369         struct ixgbe_hw *hw =
5370                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5371         uint32_t ctrl;
5372
5373         PMD_INIT_FUNC_TRACE();
5374
5375         if (queue >= hw->mac.max_rx_queues)
5376                 return;
5377
5378         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
5379         if (on)
5380                 ctrl |= IXGBE_RXDCTL_VME;
5381         else
5382                 ctrl &= ~IXGBE_RXDCTL_VME;
5383         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
5384
5385         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
5386 }
5387
5388 static int
5389 ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask)
5390 {
5391         struct ixgbe_rx_queue *rxq;
5392         uint16_t i;
5393         int on = 0;
5394
5395         /* VF function only support hw strip feature, others are not support */
5396         if (mask & ETH_VLAN_STRIP_MASK) {
5397                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
5398                         rxq = dev->data->rx_queues[i];
5399                         on = !!(rxq->offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
5400                         ixgbevf_vlan_strip_queue_set(dev, i, on);
5401                 }
5402         }
5403
5404         return 0;
5405 }
5406
5407 static int
5408 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
5409 {
5410         ixgbe_config_vlan_strip_on_all_queues(dev, mask);
5411
5412         ixgbevf_vlan_offload_config(dev, mask);
5413
5414         return 0;
5415 }
5416
5417 int
5418 ixgbe_vt_check(struct ixgbe_hw *hw)
5419 {
5420         uint32_t reg_val;
5421
5422         /* if Virtualization Technology is enabled */
5423         reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
5424         if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
5425                 PMD_INIT_LOG(ERR, "VT must be enabled for this setting");
5426                 return -1;
5427         }
5428
5429         return 0;
5430 }
5431
5432 static uint32_t
5433 ixgbe_uta_vector(struct ixgbe_hw *hw, struct rte_ether_addr *uc_addr)
5434 {
5435         uint32_t vector = 0;
5436
5437         switch (hw->mac.mc_filter_type) {
5438         case 0:   /* use bits [47:36] of the address */
5439                 vector = ((uc_addr->addr_bytes[4] >> 4) |
5440                         (((uint16_t)uc_addr->addr_bytes[5]) << 4));
5441                 break;
5442         case 1:   /* use bits [46:35] of the address */
5443                 vector = ((uc_addr->addr_bytes[4] >> 3) |
5444                         (((uint16_t)uc_addr->addr_bytes[5]) << 5));
5445                 break;
5446         case 2:   /* use bits [45:34] of the address */
5447                 vector = ((uc_addr->addr_bytes[4] >> 2) |
5448                         (((uint16_t)uc_addr->addr_bytes[5]) << 6));
5449                 break;
5450         case 3:   /* use bits [43:32] of the address */
5451                 vector = ((uc_addr->addr_bytes[4]) |
5452                         (((uint16_t)uc_addr->addr_bytes[5]) << 8));
5453                 break;
5454         default:  /* Invalid mc_filter_type */
5455                 break;
5456         }
5457
5458         /* vector can only be 12-bits or boundary will be exceeded */
5459         vector &= 0xFFF;
5460         return vector;
5461 }
5462
5463 static int
5464 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev,
5465                         struct rte_ether_addr *mac_addr, uint8_t on)
5466 {
5467         uint32_t vector;
5468         uint32_t uta_idx;
5469         uint32_t reg_val;
5470         uint32_t uta_shift;
5471         uint32_t rc;
5472         const uint32_t ixgbe_uta_idx_mask = 0x7F;
5473         const uint32_t ixgbe_uta_bit_shift = 5;
5474         const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
5475         const uint32_t bit1 = 0x1;
5476
5477         struct ixgbe_hw *hw =
5478                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5479         struct ixgbe_uta_info *uta_info =
5480                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5481
5482         /* The UTA table only exists on 82599 hardware and newer */
5483         if (hw->mac.type < ixgbe_mac_82599EB)
5484                 return -ENOTSUP;
5485
5486         vector = ixgbe_uta_vector(hw, mac_addr);
5487         uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
5488         uta_shift = vector & ixgbe_uta_bit_mask;
5489
5490         rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
5491         if (rc == on)
5492                 return 0;
5493
5494         reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
5495         if (on) {
5496                 uta_info->uta_in_use++;
5497                 reg_val |= (bit1 << uta_shift);
5498                 uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
5499         } else {
5500                 uta_info->uta_in_use--;
5501                 reg_val &= ~(bit1 << uta_shift);
5502                 uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
5503         }
5504
5505         IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
5506
5507         if (uta_info->uta_in_use > 0)
5508                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
5509                                 IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
5510         else
5511                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
5512
5513         return 0;
5514 }
5515
5516 static int
5517 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
5518 {
5519         int i;
5520         struct ixgbe_hw *hw =
5521                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5522         struct ixgbe_uta_info *uta_info =
5523                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5524
5525         /* The UTA table only exists on 82599 hardware and newer */
5526         if (hw->mac.type < ixgbe_mac_82599EB)
5527                 return -ENOTSUP;
5528
5529         if (on) {
5530                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5531                         uta_info->uta_shadow[i] = ~0;
5532                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
5533                 }
5534         } else {
5535                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5536                         uta_info->uta_shadow[i] = 0;
5537                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
5538                 }
5539         }
5540         return 0;
5541
5542 }
5543
5544 uint32_t
5545 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
5546 {
5547         uint32_t new_val = orig_val;
5548
5549         if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG)
5550                 new_val |= IXGBE_VMOLR_AUPE;
5551         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC)
5552                 new_val |= IXGBE_VMOLR_ROMPE;
5553         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
5554                 new_val |= IXGBE_VMOLR_ROPE;
5555         if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
5556                 new_val |= IXGBE_VMOLR_BAM;
5557         if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
5558                 new_val |= IXGBE_VMOLR_MPE;
5559
5560         return new_val;
5561 }
5562
5563 #define IXGBE_MRCTL_VPME  0x01 /* Virtual Pool Mirroring. */
5564 #define IXGBE_MRCTL_UPME  0x02 /* Uplink Port Mirroring. */
5565 #define IXGBE_MRCTL_DPME  0x04 /* Downlink Port Mirroring. */
5566 #define IXGBE_MRCTL_VLME  0x08 /* VLAN Mirroring. */
5567 #define IXGBE_INVALID_MIRROR_TYPE(mirror_type) \
5568         ((mirror_type) & ~(uint8_t)(ETH_MIRROR_VIRTUAL_POOL_UP | \
5569         ETH_MIRROR_UPLINK_PORT | ETH_MIRROR_DOWNLINK_PORT | ETH_MIRROR_VLAN))
5570
5571 static int
5572 ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
5573                       struct rte_eth_mirror_conf *mirror_conf,
5574                       uint8_t rule_id, uint8_t on)
5575 {
5576         uint32_t mr_ctl, vlvf;
5577         uint32_t mp_lsb = 0;
5578         uint32_t mv_msb = 0;
5579         uint32_t mv_lsb = 0;
5580         uint32_t mp_msb = 0;
5581         uint8_t i = 0;
5582         int reg_index = 0;
5583         uint64_t vlan_mask = 0;
5584
5585         const uint8_t pool_mask_offset = 32;
5586         const uint8_t vlan_mask_offset = 32;
5587         const uint8_t dst_pool_offset = 8;
5588         const uint8_t rule_mr_offset  = 4;
5589         const uint8_t mirror_rule_mask = 0x0F;
5590
5591         struct ixgbe_mirror_info *mr_info =
5592                         (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5593         struct ixgbe_hw *hw =
5594                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5595         uint8_t mirror_type = 0;
5596
5597         if (ixgbe_vt_check(hw) < 0)
5598                 return -ENOTSUP;
5599
5600         if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5601                 return -EINVAL;
5602
5603         if (IXGBE_INVALID_MIRROR_TYPE(mirror_conf->rule_type)) {
5604                 PMD_DRV_LOG(ERR, "unsupported mirror type 0x%x.",
5605                             mirror_conf->rule_type);
5606                 return -EINVAL;
5607         }
5608
5609         if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5610                 mirror_type |= IXGBE_MRCTL_VLME;
5611                 /* Check if vlan id is valid and find conresponding VLAN ID
5612                  * index in VLVF
5613                  */
5614                 for (i = 0; i < IXGBE_VLVF_ENTRIES; i++) {
5615                         if (mirror_conf->vlan.vlan_mask & (1ULL << i)) {
5616                                 /* search vlan id related pool vlan filter
5617                                  * index
5618                                  */
5619                                 reg_index = ixgbe_find_vlvf_slot(
5620                                                 hw,
5621                                                 mirror_conf->vlan.vlan_id[i],
5622                                                 false);
5623                                 if (reg_index < 0)
5624                                         return -EINVAL;
5625                                 vlvf = IXGBE_READ_REG(hw,
5626                                                       IXGBE_VLVF(reg_index));
5627                                 if ((vlvf & IXGBE_VLVF_VIEN) &&
5628                                     ((vlvf & IXGBE_VLVF_VLANID_MASK) ==
5629                                       mirror_conf->vlan.vlan_id[i]))
5630                                         vlan_mask |= (1ULL << reg_index);
5631                                 else
5632                                         return -EINVAL;
5633                         }
5634                 }
5635
5636                 if (on) {
5637                         mv_lsb = vlan_mask & 0xFFFFFFFF;
5638                         mv_msb = vlan_mask >> vlan_mask_offset;
5639
5640                         mr_info->mr_conf[rule_id].vlan.vlan_mask =
5641                                                 mirror_conf->vlan.vlan_mask;
5642                         for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++) {
5643                                 if (mirror_conf->vlan.vlan_mask & (1ULL << i))
5644                                         mr_info->mr_conf[rule_id].vlan.vlan_id[i] =
5645                                                 mirror_conf->vlan.vlan_id[i];
5646                         }
5647                 } else {
5648                         mv_lsb = 0;
5649                         mv_msb = 0;
5650                         mr_info->mr_conf[rule_id].vlan.vlan_mask = 0;
5651                         for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++)
5652                                 mr_info->mr_conf[rule_id].vlan.vlan_id[i] = 0;
5653                 }
5654         }
5655
5656         /**
5657          * if enable pool mirror, write related pool mask register,if disable
5658          * pool mirror, clear PFMRVM register
5659          */
5660         if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5661                 mirror_type |= IXGBE_MRCTL_VPME;
5662                 if (on) {
5663                         mp_lsb = mirror_conf->pool_mask & 0xFFFFFFFF;
5664                         mp_msb = mirror_conf->pool_mask >> pool_mask_offset;
5665                         mr_info->mr_conf[rule_id].pool_mask =
5666                                         mirror_conf->pool_mask;
5667
5668                 } else {
5669                         mp_lsb = 0;
5670                         mp_msb = 0;
5671                         mr_info->mr_conf[rule_id].pool_mask = 0;
5672                 }
5673         }
5674         if (mirror_conf->rule_type & ETH_MIRROR_UPLINK_PORT)
5675                 mirror_type |= IXGBE_MRCTL_UPME;
5676         if (mirror_conf->rule_type & ETH_MIRROR_DOWNLINK_PORT)
5677                 mirror_type |= IXGBE_MRCTL_DPME;
5678
5679         /* read  mirror control register and recalculate it */
5680         mr_ctl = IXGBE_READ_REG(hw, IXGBE_MRCTL(rule_id));
5681
5682         if (on) {
5683                 mr_ctl |= mirror_type;
5684                 mr_ctl &= mirror_rule_mask;
5685                 mr_ctl |= mirror_conf->dst_pool << dst_pool_offset;
5686         } else {
5687                 mr_ctl &= ~(mirror_conf->rule_type & mirror_rule_mask);
5688         }
5689
5690         mr_info->mr_conf[rule_id].rule_type = mirror_conf->rule_type;
5691         mr_info->mr_conf[rule_id].dst_pool = mirror_conf->dst_pool;
5692
5693         /* write mirrror control  register */
5694         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5695
5696         /* write pool mirrror control  register */
5697         if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5698                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), mp_lsb);
5699                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset),
5700                                 mp_msb);
5701         }
5702         /* write VLAN mirrror control  register */
5703         if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5704                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), mv_lsb);
5705                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset),
5706                                 mv_msb);
5707         }
5708
5709         return 0;
5710 }
5711
5712 static int
5713 ixgbe_mirror_rule_reset(struct rte_eth_dev *dev, uint8_t rule_id)
5714 {
5715         int mr_ctl = 0;
5716         uint32_t lsb_val = 0;
5717         uint32_t msb_val = 0;
5718         const uint8_t rule_mr_offset = 4;
5719
5720         struct ixgbe_hw *hw =
5721                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5722         struct ixgbe_mirror_info *mr_info =
5723                 (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5724
5725         if (ixgbe_vt_check(hw) < 0)
5726                 return -ENOTSUP;
5727
5728         if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5729                 return -EINVAL;
5730
5731         memset(&mr_info->mr_conf[rule_id], 0,
5732                sizeof(struct rte_eth_mirror_conf));
5733
5734         /* clear PFVMCTL register */
5735         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5736
5737         /* clear pool mask register */
5738         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), lsb_val);
5739         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset), msb_val);
5740
5741         /* clear vlan mask register */
5742         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), lsb_val);
5743         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset), msb_val);
5744
5745         return 0;
5746 }
5747
5748 static int
5749 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5750 {
5751         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5752         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5753         struct ixgbe_interrupt *intr =
5754                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5755         struct ixgbe_hw *hw =
5756                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5757         uint32_t vec = IXGBE_MISC_VEC_ID;
5758
5759         if (rte_intr_allow_others(intr_handle))
5760                 vec = IXGBE_RX_VEC_START;
5761         intr->mask |= (1 << vec);
5762         RTE_SET_USED(queue_id);
5763         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5764
5765         rte_intr_enable(intr_handle);
5766
5767         return 0;
5768 }
5769
5770 static int
5771 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5772 {
5773         struct ixgbe_interrupt *intr =
5774                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5775         struct ixgbe_hw *hw =
5776                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5777         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5778         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5779         uint32_t vec = IXGBE_MISC_VEC_ID;
5780
5781         if (rte_intr_allow_others(intr_handle))
5782                 vec = IXGBE_RX_VEC_START;
5783         intr->mask &= ~(1 << vec);
5784         RTE_SET_USED(queue_id);
5785         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5786
5787         return 0;
5788 }
5789
5790 static int
5791 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5792 {
5793         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5794         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5795         uint32_t mask;
5796         struct ixgbe_hw *hw =
5797                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5798         struct ixgbe_interrupt *intr =
5799                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5800
5801         if (queue_id < 16) {
5802                 ixgbe_disable_intr(hw);
5803                 intr->mask |= (1 << queue_id);
5804                 ixgbe_enable_intr(dev);
5805         } else if (queue_id < 32) {
5806                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5807                 mask &= (1 << queue_id);
5808                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5809         } else if (queue_id < 64) {
5810                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5811                 mask &= (1 << (queue_id - 32));
5812                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5813         }
5814         rte_intr_enable(intr_handle);
5815
5816         return 0;
5817 }
5818
5819 static int
5820 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5821 {
5822         uint32_t mask;
5823         struct ixgbe_hw *hw =
5824                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5825         struct ixgbe_interrupt *intr =
5826                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5827
5828         if (queue_id < 16) {
5829                 ixgbe_disable_intr(hw);
5830                 intr->mask &= ~(1 << queue_id);
5831                 ixgbe_enable_intr(dev);
5832         } else if (queue_id < 32) {
5833                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5834                 mask &= ~(1 << queue_id);
5835                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5836         } else if (queue_id < 64) {
5837                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5838                 mask &= ~(1 << (queue_id - 32));
5839                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5840         }
5841
5842         return 0;
5843 }
5844
5845 static void
5846 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5847                      uint8_t queue, uint8_t msix_vector)
5848 {
5849         uint32_t tmp, idx;
5850
5851         if (direction == -1) {
5852                 /* other causes */
5853                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5854                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
5855                 tmp &= ~0xFF;
5856                 tmp |= msix_vector;
5857                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
5858         } else {
5859                 /* rx or tx cause */
5860                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5861                 idx = ((16 * (queue & 1)) + (8 * direction));
5862                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
5863                 tmp &= ~(0xFF << idx);
5864                 tmp |= (msix_vector << idx);
5865                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
5866         }
5867 }
5868
5869 /**
5870  * set the IVAR registers, mapping interrupt causes to vectors
5871  * @param hw
5872  *  pointer to ixgbe_hw struct
5873  * @direction
5874  *  0 for Rx, 1 for Tx, -1 for other causes
5875  * @queue
5876  *  queue to map the corresponding interrupt to
5877  * @msix_vector
5878  *  the vector to map to the corresponding queue
5879  */
5880 static void
5881 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5882                    uint8_t queue, uint8_t msix_vector)
5883 {
5884         uint32_t tmp, idx;
5885
5886         msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5887         if (hw->mac.type == ixgbe_mac_82598EB) {
5888                 if (direction == -1)
5889                         direction = 0;
5890                 idx = (((direction * 64) + queue) >> 2) & 0x1F;
5891                 tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
5892                 tmp &= ~(0xFF << (8 * (queue & 0x3)));
5893                 tmp |= (msix_vector << (8 * (queue & 0x3)));
5894                 IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
5895         } else if ((hw->mac.type == ixgbe_mac_82599EB) ||
5896                         (hw->mac.type == ixgbe_mac_X540) ||
5897                         (hw->mac.type == ixgbe_mac_X550)) {
5898                 if (direction == -1) {
5899                         /* other causes */
5900                         idx = ((queue & 1) * 8);
5901                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
5902                         tmp &= ~(0xFF << idx);
5903                         tmp |= (msix_vector << idx);
5904                         IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
5905                 } else {
5906                         /* rx or tx causes */
5907                         idx = ((16 * (queue & 1)) + (8 * direction));
5908                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
5909                         tmp &= ~(0xFF << idx);
5910                         tmp |= (msix_vector << idx);
5911                         IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
5912                 }
5913         }
5914 }
5915
5916 static void
5917 ixgbevf_configure_msix(struct rte_eth_dev *dev)
5918 {
5919         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5920         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5921         struct ixgbe_hw *hw =
5922                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5923         uint32_t q_idx;
5924         uint32_t vector_idx = IXGBE_MISC_VEC_ID;
5925         uint32_t base = IXGBE_MISC_VEC_ID;
5926
5927         /* Configure VF other cause ivar */
5928         ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
5929
5930         /* won't configure msix register if no mapping is done
5931          * between intr vector and event fd.
5932          */
5933         if (!rte_intr_dp_is_en(intr_handle))
5934                 return;
5935
5936         if (rte_intr_allow_others(intr_handle)) {
5937                 base = IXGBE_RX_VEC_START;
5938                 vector_idx = IXGBE_RX_VEC_START;
5939         }
5940
5941         /* Configure all RX queues of VF */
5942         for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
5943                 /* Force all queue use vector 0,
5944                  * as IXGBE_VF_MAXMSIVECOTR = 1
5945                  */
5946                 ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
5947                 intr_handle->intr_vec[q_idx] = vector_idx;
5948                 if (vector_idx < base + intr_handle->nb_efd - 1)
5949                         vector_idx++;
5950         }
5951
5952         /* As RX queue setting above show, all queues use the vector 0.
5953          * Set only the ITR value of IXGBE_MISC_VEC_ID.
5954          */
5955         IXGBE_WRITE_REG(hw, IXGBE_VTEITR(IXGBE_MISC_VEC_ID),
5956                         IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
5957                         | IXGBE_EITR_CNT_WDIS);
5958 }
5959
5960 /**
5961  * Sets up the hardware to properly generate MSI-X interrupts
5962  * @hw
5963  *  board private structure
5964  */
5965 static void
5966 ixgbe_configure_msix(struct rte_eth_dev *dev)
5967 {
5968         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5969         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5970         struct ixgbe_hw *hw =
5971                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5972         uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
5973         uint32_t vec = IXGBE_MISC_VEC_ID;
5974         uint32_t mask;
5975         uint32_t gpie;
5976
5977         /* won't configure msix register if no mapping is done
5978          * between intr vector and event fd
5979          * but if misx has been enabled already, need to configure
5980          * auto clean, auto mask and throttling.
5981          */
5982         gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5983         if (!rte_intr_dp_is_en(intr_handle) &&
5984             !(gpie & (IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT)))
5985                 return;
5986
5987         if (rte_intr_allow_others(intr_handle))
5988                 vec = base = IXGBE_RX_VEC_START;
5989
5990         /* setup GPIE for MSI-x mode */
5991         gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5992         gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5993                 IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
5994         /* auto clearing and auto setting corresponding bits in EIMS
5995          * when MSI-X interrupt is triggered
5996          */
5997         if (hw->mac.type == ixgbe_mac_82598EB) {
5998                 IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5999         } else {
6000                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
6001                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
6002         }
6003         IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
6004
6005         /* Populate the IVAR table and set the ITR values to the
6006          * corresponding register.
6007          */
6008         if (rte_intr_dp_is_en(intr_handle)) {
6009                 for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
6010                         queue_id++) {
6011                         /* by default, 1:1 mapping */
6012                         ixgbe_set_ivar_map(hw, 0, queue_id, vec);
6013                         intr_handle->intr_vec[queue_id] = vec;
6014                         if (vec < base + intr_handle->nb_efd - 1)
6015                                 vec++;
6016                 }
6017
6018                 switch (hw->mac.type) {
6019                 case ixgbe_mac_82598EB:
6020                         ixgbe_set_ivar_map(hw, -1,
6021                                            IXGBE_IVAR_OTHER_CAUSES_INDEX,
6022                                            IXGBE_MISC_VEC_ID);
6023                         break;
6024                 case ixgbe_mac_82599EB:
6025                 case ixgbe_mac_X540:
6026                 case ixgbe_mac_X550:
6027                         ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
6028                         break;
6029                 default:
6030                         break;
6031                 }
6032         }
6033         IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
6034                         IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
6035                         | IXGBE_EITR_CNT_WDIS);
6036
6037         /* set up to autoclear timer, and the vectors */
6038         mask = IXGBE_EIMS_ENABLE_MASK;
6039         mask &= ~(IXGBE_EIMS_OTHER |
6040                   IXGBE_EIMS_MAILBOX |
6041                   IXGBE_EIMS_LSC);
6042
6043         IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
6044 }
6045
6046 int
6047 ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
6048                            uint16_t queue_idx, uint16_t tx_rate)
6049 {
6050         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6051         struct rte_eth_rxmode *rxmode;
6052         uint32_t rf_dec, rf_int;
6053         uint32_t bcnrc_val;
6054         uint16_t link_speed = dev->data->dev_link.link_speed;
6055
6056         if (queue_idx >= hw->mac.max_tx_queues)
6057                 return -EINVAL;
6058
6059         if (tx_rate != 0) {
6060                 /* Calculate the rate factor values to set */
6061                 rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
6062                 rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
6063                 rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
6064
6065                 bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
6066                 bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
6067                                 IXGBE_RTTBCNRC_RF_INT_MASK_M);
6068                 bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
6069         } else {
6070                 bcnrc_val = 0;
6071         }
6072
6073         rxmode = &dev->data->dev_conf.rxmode;
6074         /*
6075          * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
6076          * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
6077          * set as 0x4.
6078          */
6079         if ((rxmode->offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) &&
6080             (rxmode->max_rx_pkt_len >= IXGBE_MAX_JUMBO_FRAME_SIZE))
6081                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
6082                         IXGBE_MMW_SIZE_JUMBO_FRAME);
6083         else
6084                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
6085                         IXGBE_MMW_SIZE_DEFAULT);
6086
6087         /* Set RTTBCNRC of queue X */
6088         IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
6089         IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
6090         IXGBE_WRITE_FLUSH(hw);
6091
6092         return 0;
6093 }
6094
6095 static int
6096 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
6097                      __attribute__((unused)) uint32_t index,
6098                      __attribute__((unused)) uint32_t pool)
6099 {
6100         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6101         int diag;
6102
6103         /*
6104          * On a 82599 VF, adding again the same MAC addr is not an idempotent
6105          * operation. Trap this case to avoid exhausting the [very limited]
6106          * set of PF resources used to store VF MAC addresses.
6107          */
6108         if (memcmp(hw->mac.perm_addr, mac_addr,
6109                         sizeof(struct rte_ether_addr)) == 0)
6110                 return -1;
6111         diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6112         if (diag != 0)
6113                 PMD_DRV_LOG(ERR, "Unable to add MAC address "
6114                             "%02x:%02x:%02x:%02x:%02x:%02x - diag=%d",
6115                             mac_addr->addr_bytes[0],
6116                             mac_addr->addr_bytes[1],
6117                             mac_addr->addr_bytes[2],
6118                             mac_addr->addr_bytes[3],
6119                             mac_addr->addr_bytes[4],
6120                             mac_addr->addr_bytes[5],
6121                             diag);
6122         return diag;
6123 }
6124
6125 static void
6126 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
6127 {
6128         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6129         struct rte_ether_addr *perm_addr =
6130                 (struct rte_ether_addr *)hw->mac.perm_addr;
6131         struct rte_ether_addr *mac_addr;
6132         uint32_t i;
6133         int diag;
6134
6135         /*
6136          * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
6137          * not support the deletion of a given MAC address.
6138          * Instead, it imposes to delete all MAC addresses, then to add again
6139          * all MAC addresses with the exception of the one to be deleted.
6140          */
6141         (void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
6142
6143         /*
6144          * Add again all MAC addresses, with the exception of the deleted one
6145          * and of the permanent MAC address.
6146          */
6147         for (i = 0, mac_addr = dev->data->mac_addrs;
6148              i < hw->mac.num_rar_entries; i++, mac_addr++) {
6149                 /* Skip the deleted MAC address */
6150                 if (i == index)
6151                         continue;
6152                 /* Skip NULL MAC addresses */
6153                 if (rte_is_zero_ether_addr(mac_addr))
6154                         continue;
6155                 /* Skip the permanent MAC address */
6156                 if (memcmp(perm_addr, mac_addr,
6157                                 sizeof(struct rte_ether_addr)) == 0)
6158                         continue;
6159                 diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6160                 if (diag != 0)
6161                         PMD_DRV_LOG(ERR,
6162                                     "Adding again MAC address "
6163                                     "%02x:%02x:%02x:%02x:%02x:%02x failed "
6164                                     "diag=%d",
6165                                     mac_addr->addr_bytes[0],
6166                                     mac_addr->addr_bytes[1],
6167                                     mac_addr->addr_bytes[2],
6168                                     mac_addr->addr_bytes[3],
6169                                     mac_addr->addr_bytes[4],
6170                                     mac_addr->addr_bytes[5],
6171                                     diag);
6172         }
6173 }
6174
6175 static int
6176 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
6177                         struct rte_ether_addr *addr)
6178 {
6179         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6180
6181         hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
6182
6183         return 0;
6184 }
6185
6186 int
6187 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
6188                         struct rte_eth_syn_filter *filter,
6189                         bool add)
6190 {
6191         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6192         struct ixgbe_filter_info *filter_info =
6193                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6194         uint32_t syn_info;
6195         uint32_t synqf;
6196
6197         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6198                 return -EINVAL;
6199
6200         syn_info = filter_info->syn_info;
6201
6202         if (add) {
6203                 if (syn_info & IXGBE_SYN_FILTER_ENABLE)
6204                         return -EINVAL;
6205                 synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
6206                         IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
6207
6208                 if (filter->hig_pri)
6209                         synqf |= IXGBE_SYN_FILTER_SYNQFP;
6210                 else
6211                         synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
6212         } else {
6213                 synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6214                 if (!(syn_info & IXGBE_SYN_FILTER_ENABLE))
6215                         return -ENOENT;
6216                 synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
6217         }
6218
6219         filter_info->syn_info = synqf;
6220         IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
6221         IXGBE_WRITE_FLUSH(hw);
6222         return 0;
6223 }
6224
6225 static int
6226 ixgbe_syn_filter_get(struct rte_eth_dev *dev,
6227                         struct rte_eth_syn_filter *filter)
6228 {
6229         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6230         uint32_t synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6231
6232         if (synqf & IXGBE_SYN_FILTER_ENABLE) {
6233                 filter->hig_pri = (synqf & IXGBE_SYN_FILTER_SYNQFP) ? 1 : 0;
6234                 filter->queue = (uint16_t)((synqf & IXGBE_SYN_FILTER_QUEUE) >> 1);
6235                 return 0;
6236         }
6237         return -ENOENT;
6238 }
6239
6240 static int
6241 ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
6242                         enum rte_filter_op filter_op,
6243                         void *arg)
6244 {
6245         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6246         int ret;
6247
6248         MAC_TYPE_FILTER_SUP(hw->mac.type);
6249
6250         if (filter_op == RTE_ETH_FILTER_NOP)
6251                 return 0;
6252
6253         if (arg == NULL) {
6254                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
6255                             filter_op);
6256                 return -EINVAL;
6257         }
6258
6259         switch (filter_op) {
6260         case RTE_ETH_FILTER_ADD:
6261                 ret = ixgbe_syn_filter_set(dev,
6262                                 (struct rte_eth_syn_filter *)arg,
6263                                 TRUE);
6264                 break;
6265         case RTE_ETH_FILTER_DELETE:
6266                 ret = ixgbe_syn_filter_set(dev,
6267                                 (struct rte_eth_syn_filter *)arg,
6268                                 FALSE);
6269                 break;
6270         case RTE_ETH_FILTER_GET:
6271                 ret = ixgbe_syn_filter_get(dev,
6272                                 (struct rte_eth_syn_filter *)arg);
6273                 break;
6274         default:
6275                 PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
6276                 ret = -EINVAL;
6277                 break;
6278         }
6279
6280         return ret;
6281 }
6282
6283
6284 static inline enum ixgbe_5tuple_protocol
6285 convert_protocol_type(uint8_t protocol_value)
6286 {
6287         if (protocol_value == IPPROTO_TCP)
6288                 return IXGBE_FILTER_PROTOCOL_TCP;
6289         else if (protocol_value == IPPROTO_UDP)
6290                 return IXGBE_FILTER_PROTOCOL_UDP;
6291         else if (protocol_value == IPPROTO_SCTP)
6292                 return IXGBE_FILTER_PROTOCOL_SCTP;
6293         else
6294                 return IXGBE_FILTER_PROTOCOL_NONE;
6295 }
6296
6297 /* inject a 5-tuple filter to HW */
6298 static inline void
6299 ixgbe_inject_5tuple_filter(struct rte_eth_dev *dev,
6300                            struct ixgbe_5tuple_filter *filter)
6301 {
6302         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6303         int i;
6304         uint32_t ftqf, sdpqf;
6305         uint32_t l34timir = 0;
6306         uint8_t mask = 0xff;
6307
6308         i = filter->index;
6309
6310         sdpqf = (uint32_t)(filter->filter_info.dst_port <<
6311                                 IXGBE_SDPQF_DSTPORT_SHIFT);
6312         sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
6313
6314         ftqf = (uint32_t)(filter->filter_info.proto &
6315                 IXGBE_FTQF_PROTOCOL_MASK);
6316         ftqf |= (uint32_t)((filter->filter_info.priority &
6317                 IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
6318         if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
6319                 mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
6320         if (filter->filter_info.dst_ip_mask == 0)
6321                 mask &= IXGBE_FTQF_DEST_ADDR_MASK;
6322         if (filter->filter_info.src_port_mask == 0)
6323                 mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
6324         if (filter->filter_info.dst_port_mask == 0)
6325                 mask &= IXGBE_FTQF_DEST_PORT_MASK;
6326         if (filter->filter_info.proto_mask == 0)
6327                 mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
6328         ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
6329         ftqf |= IXGBE_FTQF_POOL_MASK_EN;
6330         ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
6331
6332         IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
6333         IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
6334         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
6335         IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
6336
6337         l34timir |= IXGBE_L34T_IMIR_RESERVE;
6338         l34timir |= (uint32_t)(filter->queue <<
6339                                 IXGBE_L34T_IMIR_QUEUE_SHIFT);
6340         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
6341 }
6342
6343 /*
6344  * add a 5tuple filter
6345  *
6346  * @param
6347  * dev: Pointer to struct rte_eth_dev.
6348  * index: the index the filter allocates.
6349  * filter: ponter to the filter that will be added.
6350  * rx_queue: the queue id the filter assigned to.
6351  *
6352  * @return
6353  *    - On success, zero.
6354  *    - On failure, a negative value.
6355  */
6356 static int
6357 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
6358                         struct ixgbe_5tuple_filter *filter)
6359 {
6360         struct ixgbe_filter_info *filter_info =
6361                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6362         int i, idx, shift;
6363
6364         /*
6365          * look for an unused 5tuple filter index,
6366          * and insert the filter to list.
6367          */
6368         for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
6369                 idx = i / (sizeof(uint32_t) * NBBY);
6370                 shift = i % (sizeof(uint32_t) * NBBY);
6371                 if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
6372                         filter_info->fivetuple_mask[idx] |= 1 << shift;
6373                         filter->index = i;
6374                         TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
6375                                           filter,
6376                                           entries);
6377                         break;
6378                 }
6379         }
6380         if (i >= IXGBE_MAX_FTQF_FILTERS) {
6381                 PMD_DRV_LOG(ERR, "5tuple filters are full.");
6382                 return -ENOSYS;
6383         }
6384
6385         ixgbe_inject_5tuple_filter(dev, filter);
6386
6387         return 0;
6388 }
6389
6390 /*
6391  * remove a 5tuple filter
6392  *
6393  * @param
6394  * dev: Pointer to struct rte_eth_dev.
6395  * filter: the pointer of the filter will be removed.
6396  */
6397 static void
6398 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
6399                         struct ixgbe_5tuple_filter *filter)
6400 {
6401         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6402         struct ixgbe_filter_info *filter_info =
6403                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6404         uint16_t index = filter->index;
6405
6406         filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
6407                                 ~(1 << (index % (sizeof(uint32_t) * NBBY)));
6408         TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
6409         rte_free(filter);
6410
6411         IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
6412         IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
6413         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
6414         IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
6415         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
6416 }
6417
6418 static int
6419 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
6420 {
6421         struct ixgbe_hw *hw;
6422         uint32_t max_frame = mtu + IXGBE_ETH_OVERHEAD;
6423         struct rte_eth_dev_data *dev_data = dev->data;
6424
6425         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6426
6427         if (mtu < RTE_ETHER_MIN_MTU ||
6428                         max_frame > RTE_ETHER_MAX_JUMBO_FRAME_LEN)
6429                 return -EINVAL;
6430
6431         /* If device is started, refuse mtu that requires the support of
6432          * scattered packets when this feature has not been enabled before.
6433          */
6434         if (dev_data->dev_started && !dev_data->scattered_rx &&
6435             (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
6436              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
6437                 PMD_INIT_LOG(ERR, "Stop port first.");
6438                 return -EINVAL;
6439         }
6440
6441         /*
6442          * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
6443          * request of the version 2.0 of the mailbox API.
6444          * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
6445          * of the mailbox API.
6446          * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
6447          * prior to 3.11.33 which contains the following change:
6448          * "ixgbe: Enable jumbo frames support w/ SR-IOV"
6449          */
6450         ixgbevf_rlpml_set_vf(hw, max_frame);
6451
6452         /* update max frame size */
6453         dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame;
6454         return 0;
6455 }
6456
6457 static inline struct ixgbe_5tuple_filter *
6458 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
6459                         struct ixgbe_5tuple_filter_info *key)
6460 {
6461         struct ixgbe_5tuple_filter *it;
6462
6463         TAILQ_FOREACH(it, filter_list, entries) {
6464                 if (memcmp(key, &it->filter_info,
6465                         sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
6466                         return it;
6467                 }
6468         }
6469         return NULL;
6470 }
6471
6472 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
6473 static inline int
6474 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
6475                         struct ixgbe_5tuple_filter_info *filter_info)
6476 {
6477         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
6478                 filter->priority > IXGBE_5TUPLE_MAX_PRI ||
6479                 filter->priority < IXGBE_5TUPLE_MIN_PRI)
6480                 return -EINVAL;
6481
6482         switch (filter->dst_ip_mask) {
6483         case UINT32_MAX:
6484                 filter_info->dst_ip_mask = 0;
6485                 filter_info->dst_ip = filter->dst_ip;
6486                 break;
6487         case 0:
6488                 filter_info->dst_ip_mask = 1;
6489                 break;
6490         default:
6491                 PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
6492                 return -EINVAL;
6493         }
6494
6495         switch (filter->src_ip_mask) {
6496         case UINT32_MAX:
6497                 filter_info->src_ip_mask = 0;
6498                 filter_info->src_ip = filter->src_ip;
6499                 break;
6500         case 0:
6501                 filter_info->src_ip_mask = 1;
6502                 break;
6503         default:
6504                 PMD_DRV_LOG(ERR, "invalid src_ip mask.");
6505                 return -EINVAL;
6506         }
6507
6508         switch (filter->dst_port_mask) {
6509         case UINT16_MAX:
6510                 filter_info->dst_port_mask = 0;
6511                 filter_info->dst_port = filter->dst_port;
6512                 break;
6513         case 0:
6514                 filter_info->dst_port_mask = 1;
6515                 break;
6516         default:
6517                 PMD_DRV_LOG(ERR, "invalid dst_port mask.");
6518                 return -EINVAL;
6519         }
6520
6521         switch (filter->src_port_mask) {
6522         case UINT16_MAX:
6523                 filter_info->src_port_mask = 0;
6524                 filter_info->src_port = filter->src_port;
6525                 break;
6526         case 0:
6527                 filter_info->src_port_mask = 1;
6528                 break;
6529         default:
6530                 PMD_DRV_LOG(ERR, "invalid src_port mask.");
6531                 return -EINVAL;
6532         }
6533
6534         switch (filter->proto_mask) {
6535         case UINT8_MAX:
6536                 filter_info->proto_mask = 0;
6537                 filter_info->proto =
6538                         convert_protocol_type(filter->proto);
6539                 break;
6540         case 0:
6541                 filter_info->proto_mask = 1;
6542                 break;
6543         default:
6544                 PMD_DRV_LOG(ERR, "invalid protocol mask.");
6545                 return -EINVAL;
6546         }
6547
6548         filter_info->priority = (uint8_t)filter->priority;
6549         return 0;
6550 }
6551
6552 /*
6553  * add or delete a ntuple filter
6554  *
6555  * @param
6556  * dev: Pointer to struct rte_eth_dev.
6557  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6558  * add: if true, add filter, if false, remove filter
6559  *
6560  * @return
6561  *    - On success, zero.
6562  *    - On failure, a negative value.
6563  */
6564 int
6565 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
6566                         struct rte_eth_ntuple_filter *ntuple_filter,
6567                         bool add)
6568 {
6569         struct ixgbe_filter_info *filter_info =
6570                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6571         struct ixgbe_5tuple_filter_info filter_5tuple;
6572         struct ixgbe_5tuple_filter *filter;
6573         int ret;
6574
6575         if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6576                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6577                 return -EINVAL;
6578         }
6579
6580         memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6581         ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6582         if (ret < 0)
6583                 return ret;
6584
6585         filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6586                                          &filter_5tuple);
6587         if (filter != NULL && add) {
6588                 PMD_DRV_LOG(ERR, "filter exists.");
6589                 return -EEXIST;
6590         }
6591         if (filter == NULL && !add) {
6592                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
6593                 return -ENOENT;
6594         }
6595
6596         if (add) {
6597                 filter = rte_zmalloc("ixgbe_5tuple_filter",
6598                                 sizeof(struct ixgbe_5tuple_filter), 0);
6599                 if (filter == NULL)
6600                         return -ENOMEM;
6601                 rte_memcpy(&filter->filter_info,
6602                                  &filter_5tuple,
6603                                  sizeof(struct ixgbe_5tuple_filter_info));
6604                 filter->queue = ntuple_filter->queue;
6605                 ret = ixgbe_add_5tuple_filter(dev, filter);
6606                 if (ret < 0) {
6607                         rte_free(filter);
6608                         return ret;
6609                 }
6610         } else
6611                 ixgbe_remove_5tuple_filter(dev, filter);
6612
6613         return 0;
6614 }
6615
6616 /*
6617  * get a ntuple filter
6618  *
6619  * @param
6620  * dev: Pointer to struct rte_eth_dev.
6621  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6622  *
6623  * @return
6624  *    - On success, zero.
6625  *    - On failure, a negative value.
6626  */
6627 static int
6628 ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
6629                         struct rte_eth_ntuple_filter *ntuple_filter)
6630 {
6631         struct ixgbe_filter_info *filter_info =
6632                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6633         struct ixgbe_5tuple_filter_info filter_5tuple;
6634         struct ixgbe_5tuple_filter *filter;
6635         int ret;
6636
6637         if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6638                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6639                 return -EINVAL;
6640         }
6641
6642         memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6643         ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6644         if (ret < 0)
6645                 return ret;
6646
6647         filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6648                                          &filter_5tuple);
6649         if (filter == NULL) {
6650                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
6651                 return -ENOENT;
6652         }
6653         ntuple_filter->queue = filter->queue;
6654         return 0;
6655 }
6656
6657 /*
6658  * ixgbe_ntuple_filter_handle - Handle operations for ntuple filter.
6659  * @dev: pointer to rte_eth_dev structure
6660  * @filter_op:operation will be taken.
6661  * @arg: a pointer to specific structure corresponding to the filter_op
6662  *
6663  * @return
6664  *    - On success, zero.
6665  *    - On failure, a negative value.
6666  */
6667 static int
6668 ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
6669                                 enum rte_filter_op filter_op,
6670                                 void *arg)
6671 {
6672         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6673         int ret;
6674
6675         MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
6676
6677         if (filter_op == RTE_ETH_FILTER_NOP)
6678                 return 0;
6679
6680         if (arg == NULL) {
6681                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
6682                             filter_op);
6683                 return -EINVAL;
6684         }
6685
6686         switch (filter_op) {
6687         case RTE_ETH_FILTER_ADD:
6688                 ret = ixgbe_add_del_ntuple_filter(dev,
6689                         (struct rte_eth_ntuple_filter *)arg,
6690                         TRUE);
6691                 break;
6692         case RTE_ETH_FILTER_DELETE:
6693                 ret = ixgbe_add_del_ntuple_filter(dev,
6694                         (struct rte_eth_ntuple_filter *)arg,
6695                         FALSE);
6696                 break;
6697         case RTE_ETH_FILTER_GET:
6698                 ret = ixgbe_get_ntuple_filter(dev,
6699                         (struct rte_eth_ntuple_filter *)arg);
6700                 break;
6701         default:
6702                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
6703                 ret = -EINVAL;
6704                 break;
6705         }
6706         return ret;
6707 }
6708
6709 int
6710 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
6711                         struct rte_eth_ethertype_filter *filter,
6712                         bool add)
6713 {
6714         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6715         struct ixgbe_filter_info *filter_info =
6716                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6717         uint32_t etqf = 0;
6718         uint32_t etqs = 0;
6719         int ret;
6720         struct ixgbe_ethertype_filter ethertype_filter;
6721
6722         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6723                 return -EINVAL;
6724
6725         if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
6726                 filter->ether_type == RTE_ETHER_TYPE_IPV6) {
6727                 PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
6728                         " ethertype filter.", filter->ether_type);
6729                 return -EINVAL;
6730         }
6731
6732         if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
6733                 PMD_DRV_LOG(ERR, "mac compare is unsupported.");
6734                 return -EINVAL;
6735         }
6736         if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
6737                 PMD_DRV_LOG(ERR, "drop option is unsupported.");
6738                 return -EINVAL;
6739         }
6740
6741         ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6742         if (ret >= 0 && add) {
6743                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
6744                             filter->ether_type);
6745                 return -EEXIST;
6746         }
6747         if (ret < 0 && !add) {
6748                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6749                             filter->ether_type);
6750                 return -ENOENT;
6751         }
6752
6753         if (add) {
6754                 etqf = IXGBE_ETQF_FILTER_EN;
6755                 etqf |= (uint32_t)filter->ether_type;
6756                 etqs |= (uint32_t)((filter->queue <<
6757                                     IXGBE_ETQS_RX_QUEUE_SHIFT) &
6758                                     IXGBE_ETQS_RX_QUEUE);
6759                 etqs |= IXGBE_ETQS_QUEUE_EN;
6760
6761                 ethertype_filter.ethertype = filter->ether_type;
6762                 ethertype_filter.etqf = etqf;
6763                 ethertype_filter.etqs = etqs;
6764                 ethertype_filter.conf = FALSE;
6765                 ret = ixgbe_ethertype_filter_insert(filter_info,
6766                                                     &ethertype_filter);
6767                 if (ret < 0) {
6768                         PMD_DRV_LOG(ERR, "ethertype filters are full.");
6769                         return -ENOSPC;
6770                 }
6771         } else {
6772                 ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
6773                 if (ret < 0)
6774                         return -ENOSYS;
6775         }
6776         IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
6777         IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
6778         IXGBE_WRITE_FLUSH(hw);
6779
6780         return 0;
6781 }
6782
6783 static int
6784 ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
6785                         struct rte_eth_ethertype_filter *filter)
6786 {
6787         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6788         struct ixgbe_filter_info *filter_info =
6789                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6790         uint32_t etqf, etqs;
6791         int ret;
6792
6793         ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6794         if (ret < 0) {
6795                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6796                             filter->ether_type);
6797                 return -ENOENT;
6798         }
6799
6800         etqf = IXGBE_READ_REG(hw, IXGBE_ETQF(ret));
6801         if (etqf & IXGBE_ETQF_FILTER_EN) {
6802                 etqs = IXGBE_READ_REG(hw, IXGBE_ETQS(ret));
6803                 filter->ether_type = etqf & IXGBE_ETQF_ETHERTYPE;
6804                 filter->flags = 0;
6805                 filter->queue = (etqs & IXGBE_ETQS_RX_QUEUE) >>
6806                                IXGBE_ETQS_RX_QUEUE_SHIFT;
6807                 return 0;
6808         }
6809         return -ENOENT;
6810 }
6811
6812 /*
6813  * ixgbe_ethertype_filter_handle - Handle operations for ethertype 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_ethertype_filter_handle(struct rte_eth_dev *dev,
6820                                 enum rte_filter_op filter_op,
6821                                 void *arg)
6822 {
6823         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6824         int ret;
6825
6826         MAC_TYPE_FILTER_SUP(hw->mac.type);
6827
6828         if (filter_op == RTE_ETH_FILTER_NOP)
6829                 return 0;
6830
6831         if (arg == NULL) {
6832                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
6833                             filter_op);
6834                 return -EINVAL;
6835         }
6836
6837         switch (filter_op) {
6838         case RTE_ETH_FILTER_ADD:
6839                 ret = ixgbe_add_del_ethertype_filter(dev,
6840                         (struct rte_eth_ethertype_filter *)arg,
6841                         TRUE);
6842                 break;
6843         case RTE_ETH_FILTER_DELETE:
6844                 ret = ixgbe_add_del_ethertype_filter(dev,
6845                         (struct rte_eth_ethertype_filter *)arg,
6846                         FALSE);
6847                 break;
6848         case RTE_ETH_FILTER_GET:
6849                 ret = ixgbe_get_ethertype_filter(dev,
6850                         (struct rte_eth_ethertype_filter *)arg);
6851                 break;
6852         default:
6853                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
6854                 ret = -EINVAL;
6855                 break;
6856         }
6857         return ret;
6858 }
6859
6860 static int
6861 ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
6862                      enum rte_filter_type filter_type,
6863                      enum rte_filter_op filter_op,
6864                      void *arg)
6865 {
6866         int ret = 0;
6867
6868         switch (filter_type) {
6869         case RTE_ETH_FILTER_NTUPLE:
6870                 ret = ixgbe_ntuple_filter_handle(dev, filter_op, arg);
6871                 break;
6872         case RTE_ETH_FILTER_ETHERTYPE:
6873                 ret = ixgbe_ethertype_filter_handle(dev, filter_op, arg);
6874                 break;
6875         case RTE_ETH_FILTER_SYN:
6876                 ret = ixgbe_syn_filter_handle(dev, filter_op, arg);
6877                 break;
6878         case RTE_ETH_FILTER_FDIR:
6879                 ret = ixgbe_fdir_ctrl_func(dev, filter_op, arg);
6880                 break;
6881         case RTE_ETH_FILTER_L2_TUNNEL:
6882                 ret = ixgbe_dev_l2_tunnel_filter_handle(dev, filter_op, arg);
6883                 break;
6884         case RTE_ETH_FILTER_GENERIC:
6885                 if (filter_op != RTE_ETH_FILTER_GET)
6886                         return -EINVAL;
6887                 *(const void **)arg = &ixgbe_flow_ops;
6888                 break;
6889         default:
6890                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
6891                                                         filter_type);
6892                 ret = -EINVAL;
6893                 break;
6894         }
6895
6896         return ret;
6897 }
6898
6899 static u8 *
6900 ixgbe_dev_addr_list_itr(__attribute__((unused)) struct ixgbe_hw *hw,
6901                         u8 **mc_addr_ptr, u32 *vmdq)
6902 {
6903         u8 *mc_addr;
6904
6905         *vmdq = 0;
6906         mc_addr = *mc_addr_ptr;
6907         *mc_addr_ptr = (mc_addr + sizeof(struct rte_ether_addr));
6908         return mc_addr;
6909 }
6910
6911 static int
6912 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
6913                           struct rte_ether_addr *mc_addr_set,
6914                           uint32_t nb_mc_addr)
6915 {
6916         struct ixgbe_hw *hw;
6917         u8 *mc_addr_list;
6918
6919         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6920         mc_addr_list = (u8 *)mc_addr_set;
6921         return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
6922                                          ixgbe_dev_addr_list_itr, TRUE);
6923 }
6924
6925 static uint64_t
6926 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
6927 {
6928         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6929         uint64_t systime_cycles;
6930
6931         switch (hw->mac.type) {
6932         case ixgbe_mac_X550:
6933         case ixgbe_mac_X550EM_x:
6934         case ixgbe_mac_X550EM_a:
6935                 /* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
6936                 systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6937                 systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6938                                 * NSEC_PER_SEC;
6939                 break;
6940         default:
6941                 systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6942                 systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6943                                 << 32;
6944         }
6945
6946         return systime_cycles;
6947 }
6948
6949 static uint64_t
6950 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6951 {
6952         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6953         uint64_t rx_tstamp_cycles;
6954
6955         switch (hw->mac.type) {
6956         case ixgbe_mac_X550:
6957         case ixgbe_mac_X550EM_x:
6958         case ixgbe_mac_X550EM_a:
6959                 /* RXSTMPL stores ns and RXSTMPH stores seconds. */
6960                 rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6961                 rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6962                                 * NSEC_PER_SEC;
6963                 break;
6964         default:
6965                 /* RXSTMPL stores ns and RXSTMPH stores seconds. */
6966                 rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6967                 rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6968                                 << 32;
6969         }
6970
6971         return rx_tstamp_cycles;
6972 }
6973
6974 static uint64_t
6975 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6976 {
6977         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6978         uint64_t tx_tstamp_cycles;
6979
6980         switch (hw->mac.type) {
6981         case ixgbe_mac_X550:
6982         case ixgbe_mac_X550EM_x:
6983         case ixgbe_mac_X550EM_a:
6984                 /* TXSTMPL stores ns and TXSTMPH stores seconds. */
6985                 tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6986                 tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6987                                 * NSEC_PER_SEC;
6988                 break;
6989         default:
6990                 /* TXSTMPL stores ns and TXSTMPH stores seconds. */
6991                 tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6992                 tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6993                                 << 32;
6994         }
6995
6996         return tx_tstamp_cycles;
6997 }
6998
6999 static void
7000 ixgbe_start_timecounters(struct rte_eth_dev *dev)
7001 {
7002         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7003         struct ixgbe_adapter *adapter = dev->data->dev_private;
7004         struct rte_eth_link link;
7005         uint32_t incval = 0;
7006         uint32_t shift = 0;
7007
7008         /* Get current link speed. */
7009         ixgbe_dev_link_update(dev, 1);
7010         rte_eth_linkstatus_get(dev, &link);
7011
7012         switch (link.link_speed) {
7013         case ETH_SPEED_NUM_100M:
7014                 incval = IXGBE_INCVAL_100;
7015                 shift = IXGBE_INCVAL_SHIFT_100;
7016                 break;
7017         case ETH_SPEED_NUM_1G:
7018                 incval = IXGBE_INCVAL_1GB;
7019                 shift = IXGBE_INCVAL_SHIFT_1GB;
7020                 break;
7021         case ETH_SPEED_NUM_10G:
7022         default:
7023                 incval = IXGBE_INCVAL_10GB;
7024                 shift = IXGBE_INCVAL_SHIFT_10GB;
7025                 break;
7026         }
7027
7028         switch (hw->mac.type) {
7029         case ixgbe_mac_X550:
7030         case ixgbe_mac_X550EM_x:
7031         case ixgbe_mac_X550EM_a:
7032                 /* Independent of link speed. */
7033                 incval = 1;
7034                 /* Cycles read will be interpreted as ns. */
7035                 shift = 0;
7036                 /* Fall-through */
7037         case ixgbe_mac_X540:
7038                 IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
7039                 break;
7040         case ixgbe_mac_82599EB:
7041                 incval >>= IXGBE_INCVAL_SHIFT_82599;
7042                 shift -= IXGBE_INCVAL_SHIFT_82599;
7043                 IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
7044                                 (1 << IXGBE_INCPER_SHIFT_82599) | incval);
7045                 break;
7046         default:
7047                 /* Not supported. */
7048                 return;
7049         }
7050
7051         memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
7052         memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
7053         memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
7054
7055         adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
7056         adapter->systime_tc.cc_shift = shift;
7057         adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
7058
7059         adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
7060         adapter->rx_tstamp_tc.cc_shift = shift;
7061         adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
7062
7063         adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
7064         adapter->tx_tstamp_tc.cc_shift = shift;
7065         adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
7066 }
7067
7068 static int
7069 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
7070 {
7071         struct ixgbe_adapter *adapter = dev->data->dev_private;
7072
7073         adapter->systime_tc.nsec += delta;
7074         adapter->rx_tstamp_tc.nsec += delta;
7075         adapter->tx_tstamp_tc.nsec += delta;
7076
7077         return 0;
7078 }
7079
7080 static int
7081 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
7082 {
7083         uint64_t ns;
7084         struct ixgbe_adapter *adapter = dev->data->dev_private;
7085
7086         ns = rte_timespec_to_ns(ts);
7087         /* Set the timecounters to a new value. */
7088         adapter->systime_tc.nsec = ns;
7089         adapter->rx_tstamp_tc.nsec = ns;
7090         adapter->tx_tstamp_tc.nsec = ns;
7091
7092         return 0;
7093 }
7094
7095 static int
7096 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
7097 {
7098         uint64_t ns, systime_cycles;
7099         struct ixgbe_adapter *adapter = dev->data->dev_private;
7100
7101         systime_cycles = ixgbe_read_systime_cyclecounter(dev);
7102         ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
7103         *ts = rte_ns_to_timespec(ns);
7104
7105         return 0;
7106 }
7107
7108 static int
7109 ixgbe_timesync_enable(struct rte_eth_dev *dev)
7110 {
7111         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7112         uint32_t tsync_ctl;
7113         uint32_t tsauxc;
7114
7115         /* Stop the timesync system time. */
7116         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
7117         /* Reset the timesync system time value. */
7118         IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
7119         IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
7120
7121         /* Enable system time for platforms where it isn't on by default. */
7122         tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
7123         tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
7124         IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
7125
7126         ixgbe_start_timecounters(dev);
7127
7128         /* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
7129         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
7130                         (RTE_ETHER_TYPE_1588 |
7131                          IXGBE_ETQF_FILTER_EN |
7132                          IXGBE_ETQF_1588));
7133
7134         /* Enable timestamping of received PTP packets. */
7135         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7136         tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
7137         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
7138
7139         /* Enable timestamping of transmitted PTP packets. */
7140         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7141         tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
7142         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
7143
7144         IXGBE_WRITE_FLUSH(hw);
7145
7146         return 0;
7147 }
7148
7149 static int
7150 ixgbe_timesync_disable(struct rte_eth_dev *dev)
7151 {
7152         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7153         uint32_t tsync_ctl;
7154
7155         /* Disable timestamping of transmitted PTP packets. */
7156         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7157         tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
7158         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
7159
7160         /* Disable timestamping of received PTP packets. */
7161         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7162         tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
7163         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
7164
7165         /* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
7166         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
7167
7168         /* Stop incrementating the System Time registers. */
7169         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
7170
7171         return 0;
7172 }
7173
7174 static int
7175 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
7176                                  struct timespec *timestamp,
7177                                  uint32_t flags __rte_unused)
7178 {
7179         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7180         struct ixgbe_adapter *adapter = dev->data->dev_private;
7181         uint32_t tsync_rxctl;
7182         uint64_t rx_tstamp_cycles;
7183         uint64_t ns;
7184
7185         tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7186         if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
7187                 return -EINVAL;
7188
7189         rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
7190         ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
7191         *timestamp = rte_ns_to_timespec(ns);
7192
7193         return  0;
7194 }
7195
7196 static int
7197 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
7198                                  struct timespec *timestamp)
7199 {
7200         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7201         struct ixgbe_adapter *adapter = dev->data->dev_private;
7202         uint32_t tsync_txctl;
7203         uint64_t tx_tstamp_cycles;
7204         uint64_t ns;
7205
7206         tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7207         if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
7208                 return -EINVAL;
7209
7210         tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
7211         ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
7212         *timestamp = rte_ns_to_timespec(ns);
7213
7214         return 0;
7215 }
7216
7217 static int
7218 ixgbe_get_reg_length(struct rte_eth_dev *dev)
7219 {
7220         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7221         int count = 0;
7222         int g_ind = 0;
7223         const struct reg_info *reg_group;
7224         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7225                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7226
7227         while ((reg_group = reg_set[g_ind++]))
7228                 count += ixgbe_regs_group_count(reg_group);
7229
7230         return count;
7231 }
7232
7233 static int
7234 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
7235 {
7236         int count = 0;
7237         int g_ind = 0;
7238         const struct reg_info *reg_group;
7239
7240         while ((reg_group = ixgbevf_regs[g_ind++]))
7241                 count += ixgbe_regs_group_count(reg_group);
7242
7243         return count;
7244 }
7245
7246 static int
7247 ixgbe_get_regs(struct rte_eth_dev *dev,
7248               struct rte_dev_reg_info *regs)
7249 {
7250         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7251         uint32_t *data = regs->data;
7252         int g_ind = 0;
7253         int count = 0;
7254         const struct reg_info *reg_group;
7255         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7256                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7257
7258         if (data == NULL) {
7259                 regs->length = ixgbe_get_reg_length(dev);
7260                 regs->width = sizeof(uint32_t);
7261                 return 0;
7262         }
7263
7264         /* Support only full register dump */
7265         if ((regs->length == 0) ||
7266             (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
7267                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7268                         hw->device_id;
7269                 while ((reg_group = reg_set[g_ind++]))
7270                         count += ixgbe_read_regs_group(dev, &data[count],
7271                                 reg_group);
7272                 return 0;
7273         }
7274
7275         return -ENOTSUP;
7276 }
7277
7278 static int
7279 ixgbevf_get_regs(struct rte_eth_dev *dev,
7280                 struct rte_dev_reg_info *regs)
7281 {
7282         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7283         uint32_t *data = regs->data;
7284         int g_ind = 0;
7285         int count = 0;
7286         const struct reg_info *reg_group;
7287
7288         if (data == NULL) {
7289                 regs->length = ixgbevf_get_reg_length(dev);
7290                 regs->width = sizeof(uint32_t);
7291                 return 0;
7292         }
7293
7294         /* Support only full register dump */
7295         if ((regs->length == 0) ||
7296             (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
7297                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7298                         hw->device_id;
7299                 while ((reg_group = ixgbevf_regs[g_ind++]))
7300                         count += ixgbe_read_regs_group(dev, &data[count],
7301                                                       reg_group);
7302                 return 0;
7303         }
7304
7305         return -ENOTSUP;
7306 }
7307
7308 static int
7309 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
7310 {
7311         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7312
7313         /* Return unit is byte count */
7314         return hw->eeprom.word_size * 2;
7315 }
7316
7317 static int
7318 ixgbe_get_eeprom(struct rte_eth_dev *dev,
7319                 struct rte_dev_eeprom_info *in_eeprom)
7320 {
7321         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7322         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7323         uint16_t *data = in_eeprom->data;
7324         int first, length;
7325
7326         first = in_eeprom->offset >> 1;
7327         length = in_eeprom->length >> 1;
7328         if ((first > hw->eeprom.word_size) ||
7329             ((first + length) > hw->eeprom.word_size))
7330                 return -EINVAL;
7331
7332         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7333
7334         return eeprom->ops.read_buffer(hw, first, length, data);
7335 }
7336
7337 static int
7338 ixgbe_set_eeprom(struct rte_eth_dev *dev,
7339                 struct rte_dev_eeprom_info *in_eeprom)
7340 {
7341         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7342         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7343         uint16_t *data = in_eeprom->data;
7344         int first, length;
7345
7346         first = in_eeprom->offset >> 1;
7347         length = in_eeprom->length >> 1;
7348         if ((first > hw->eeprom.word_size) ||
7349             ((first + length) > hw->eeprom.word_size))
7350                 return -EINVAL;
7351
7352         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7353
7354         return eeprom->ops.write_buffer(hw,  first, length, data);
7355 }
7356
7357 static int
7358 ixgbe_get_module_info(struct rte_eth_dev *dev,
7359                       struct rte_eth_dev_module_info *modinfo)
7360 {
7361         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7362         uint32_t status;
7363         uint8_t sff8472_rev, addr_mode;
7364         bool page_swap = false;
7365
7366         /* Check whether we support SFF-8472 or not */
7367         status = hw->phy.ops.read_i2c_eeprom(hw,
7368                                              IXGBE_SFF_SFF_8472_COMP,
7369                                              &sff8472_rev);
7370         if (status != 0)
7371                 return -EIO;
7372
7373         /* addressing mode is not supported */
7374         status = hw->phy.ops.read_i2c_eeprom(hw,
7375                                              IXGBE_SFF_SFF_8472_SWAP,
7376                                              &addr_mode);
7377         if (status != 0)
7378                 return -EIO;
7379
7380         if (addr_mode & IXGBE_SFF_ADDRESSING_MODE) {
7381                 PMD_DRV_LOG(ERR,
7382                             "Address change required to access page 0xA2, "
7383                             "but not supported. Please report the module "
7384                             "type to the driver maintainers.");
7385                 page_swap = true;
7386         }
7387
7388         if (sff8472_rev == IXGBE_SFF_SFF_8472_UNSUP || page_swap) {
7389                 /* We have a SFP, but it does not support SFF-8472 */
7390                 modinfo->type = RTE_ETH_MODULE_SFF_8079;
7391                 modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
7392         } else {
7393                 /* We have a SFP which supports a revision of SFF-8472. */
7394                 modinfo->type = RTE_ETH_MODULE_SFF_8472;
7395                 modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
7396         }
7397
7398         return 0;
7399 }
7400
7401 static int
7402 ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
7403                         struct rte_dev_eeprom_info *info)
7404 {
7405         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7406         uint32_t status = IXGBE_ERR_PHY_ADDR_INVALID;
7407         uint8_t databyte = 0xFF;
7408         uint8_t *data = info->data;
7409         uint32_t i = 0;
7410
7411         if (info->length == 0)
7412                 return -EINVAL;
7413
7414         for (i = info->offset; i < info->offset + info->length; i++) {
7415                 if (i < RTE_ETH_MODULE_SFF_8079_LEN)
7416                         status = hw->phy.ops.read_i2c_eeprom(hw, i, &databyte);
7417                 else
7418                         status = hw->phy.ops.read_i2c_sff8472(hw, i, &databyte);
7419
7420                 if (status != 0)
7421                         return -EIO;
7422
7423                 data[i - info->offset] = databyte;
7424         }
7425
7426         return 0;
7427 }
7428
7429 uint16_t
7430 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
7431         switch (mac_type) {
7432         case ixgbe_mac_X550:
7433         case ixgbe_mac_X550EM_x:
7434         case ixgbe_mac_X550EM_a:
7435                 return ETH_RSS_RETA_SIZE_512;
7436         case ixgbe_mac_X550_vf:
7437         case ixgbe_mac_X550EM_x_vf:
7438         case ixgbe_mac_X550EM_a_vf:
7439                 return ETH_RSS_RETA_SIZE_64;
7440         default:
7441                 return ETH_RSS_RETA_SIZE_128;
7442         }
7443 }
7444
7445 uint32_t
7446 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
7447         switch (mac_type) {
7448         case ixgbe_mac_X550:
7449         case ixgbe_mac_X550EM_x:
7450         case ixgbe_mac_X550EM_a:
7451                 if (reta_idx < ETH_RSS_RETA_SIZE_128)
7452                         return IXGBE_RETA(reta_idx >> 2);
7453                 else
7454                         return IXGBE_ERETA((reta_idx - ETH_RSS_RETA_SIZE_128) >> 2);
7455         case ixgbe_mac_X550_vf:
7456         case ixgbe_mac_X550EM_x_vf:
7457         case ixgbe_mac_X550EM_a_vf:
7458                 return IXGBE_VFRETA(reta_idx >> 2);
7459         default:
7460                 return IXGBE_RETA(reta_idx >> 2);
7461         }
7462 }
7463
7464 uint32_t
7465 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
7466         switch (mac_type) {
7467         case ixgbe_mac_X550_vf:
7468         case ixgbe_mac_X550EM_x_vf:
7469         case ixgbe_mac_X550EM_a_vf:
7470                 return IXGBE_VFMRQC;
7471         default:
7472                 return IXGBE_MRQC;
7473         }
7474 }
7475
7476 uint32_t
7477 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
7478         switch (mac_type) {
7479         case ixgbe_mac_X550_vf:
7480         case ixgbe_mac_X550EM_x_vf:
7481         case ixgbe_mac_X550EM_a_vf:
7482                 return IXGBE_VFRSSRK(i);
7483         default:
7484                 return IXGBE_RSSRK(i);
7485         }
7486 }
7487
7488 bool
7489 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
7490         switch (mac_type) {
7491         case ixgbe_mac_82599_vf:
7492         case ixgbe_mac_X540_vf:
7493                 return 0;
7494         default:
7495                 return 1;
7496         }
7497 }
7498
7499 static int
7500 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
7501                         struct rte_eth_dcb_info *dcb_info)
7502 {
7503         struct ixgbe_dcb_config *dcb_config =
7504                         IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
7505         struct ixgbe_dcb_tc_config *tc;
7506         struct rte_eth_dcb_tc_queue_mapping *tc_queue;
7507         uint8_t nb_tcs;
7508         uint8_t i, j;
7509
7510         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_DCB_FLAG)
7511                 dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
7512         else
7513                 dcb_info->nb_tcs = 1;
7514
7515         tc_queue = &dcb_info->tc_queue;
7516         nb_tcs = dcb_info->nb_tcs;
7517
7518         if (dcb_config->vt_mode) { /* vt is enabled*/
7519                 struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
7520                                 &dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
7521                 for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7522                         dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
7523                 if (RTE_ETH_DEV_SRIOV(dev).active > 0) {
7524                         for (j = 0; j < nb_tcs; j++) {
7525                                 tc_queue->tc_rxq[0][j].base = j;
7526                                 tc_queue->tc_rxq[0][j].nb_queue = 1;
7527                                 tc_queue->tc_txq[0][j].base = j;
7528                                 tc_queue->tc_txq[0][j].nb_queue = 1;
7529                         }
7530                 } else {
7531                         for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
7532                                 for (j = 0; j < nb_tcs; j++) {
7533                                         tc_queue->tc_rxq[i][j].base =
7534                                                 i * nb_tcs + j;
7535                                         tc_queue->tc_rxq[i][j].nb_queue = 1;
7536                                         tc_queue->tc_txq[i][j].base =
7537                                                 i * nb_tcs + j;
7538                                         tc_queue->tc_txq[i][j].nb_queue = 1;
7539                                 }
7540                         }
7541                 }
7542         } else { /* vt is disabled*/
7543                 struct rte_eth_dcb_rx_conf *rx_conf =
7544                                 &dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
7545                 for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7546                         dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
7547                 if (dcb_info->nb_tcs == ETH_4_TCS) {
7548                         for (i = 0; i < dcb_info->nb_tcs; i++) {
7549                                 dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
7550                                 dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7551                         }
7552                         dcb_info->tc_queue.tc_txq[0][0].base = 0;
7553                         dcb_info->tc_queue.tc_txq[0][1].base = 64;
7554                         dcb_info->tc_queue.tc_txq[0][2].base = 96;
7555                         dcb_info->tc_queue.tc_txq[0][3].base = 112;
7556                         dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
7557                         dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7558                         dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7559                         dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7560                 } else if (dcb_info->nb_tcs == ETH_8_TCS) {
7561                         for (i = 0; i < dcb_info->nb_tcs; i++) {
7562                                 dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
7563                                 dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7564                         }
7565                         dcb_info->tc_queue.tc_txq[0][0].base = 0;
7566                         dcb_info->tc_queue.tc_txq[0][1].base = 32;
7567                         dcb_info->tc_queue.tc_txq[0][2].base = 64;
7568                         dcb_info->tc_queue.tc_txq[0][3].base = 80;
7569                         dcb_info->tc_queue.tc_txq[0][4].base = 96;
7570                         dcb_info->tc_queue.tc_txq[0][5].base = 104;
7571                         dcb_info->tc_queue.tc_txq[0][6].base = 112;
7572                         dcb_info->tc_queue.tc_txq[0][7].base = 120;
7573                         dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
7574                         dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7575                         dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7576                         dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7577                         dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
7578                         dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
7579                         dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
7580                         dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
7581                 }
7582         }
7583         for (i = 0; i < dcb_info->nb_tcs; i++) {
7584                 tc = &dcb_config->tc_config[i];
7585                 dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
7586         }
7587         return 0;
7588 }
7589
7590 /* Update e-tag ether type */
7591 static int
7592 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
7593                             uint16_t ether_type)
7594 {
7595         uint32_t etag_etype;
7596
7597         if (hw->mac.type != ixgbe_mac_X550 &&
7598             hw->mac.type != ixgbe_mac_X550EM_x &&
7599             hw->mac.type != ixgbe_mac_X550EM_a) {
7600                 return -ENOTSUP;
7601         }
7602
7603         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7604         etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
7605         etag_etype |= ether_type;
7606         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7607         IXGBE_WRITE_FLUSH(hw);
7608
7609         return 0;
7610 }
7611
7612 /* Config l2 tunnel ether type */
7613 static int
7614 ixgbe_dev_l2_tunnel_eth_type_conf(struct rte_eth_dev *dev,
7615                                   struct rte_eth_l2_tunnel_conf *l2_tunnel)
7616 {
7617         int ret = 0;
7618         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7619         struct ixgbe_l2_tn_info *l2_tn_info =
7620                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7621
7622         if (l2_tunnel == NULL)
7623                 return -EINVAL;
7624
7625         switch (l2_tunnel->l2_tunnel_type) {
7626         case RTE_L2_TUNNEL_TYPE_E_TAG:
7627                 l2_tn_info->e_tag_ether_type = l2_tunnel->ether_type;
7628                 ret = ixgbe_update_e_tag_eth_type(hw, l2_tunnel->ether_type);
7629                 break;
7630         default:
7631                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7632                 ret = -EINVAL;
7633                 break;
7634         }
7635
7636         return ret;
7637 }
7638
7639 /* Enable e-tag tunnel */
7640 static int
7641 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
7642 {
7643         uint32_t etag_etype;
7644
7645         if (hw->mac.type != ixgbe_mac_X550 &&
7646             hw->mac.type != ixgbe_mac_X550EM_x &&
7647             hw->mac.type != ixgbe_mac_X550EM_a) {
7648                 return -ENOTSUP;
7649         }
7650
7651         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7652         etag_etype |= IXGBE_ETAG_ETYPE_VALID;
7653         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7654         IXGBE_WRITE_FLUSH(hw);
7655
7656         return 0;
7657 }
7658
7659 /* Enable l2 tunnel */
7660 static int
7661 ixgbe_dev_l2_tunnel_enable(struct rte_eth_dev *dev,
7662                            enum rte_eth_tunnel_type l2_tunnel_type)
7663 {
7664         int ret = 0;
7665         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7666         struct ixgbe_l2_tn_info *l2_tn_info =
7667                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7668
7669         switch (l2_tunnel_type) {
7670         case RTE_L2_TUNNEL_TYPE_E_TAG:
7671                 l2_tn_info->e_tag_en = TRUE;
7672                 ret = ixgbe_e_tag_enable(hw);
7673                 break;
7674         default:
7675                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7676                 ret = -EINVAL;
7677                 break;
7678         }
7679
7680         return ret;
7681 }
7682
7683 /* Disable e-tag tunnel */
7684 static int
7685 ixgbe_e_tag_disable(struct ixgbe_hw *hw)
7686 {
7687         uint32_t etag_etype;
7688
7689         if (hw->mac.type != ixgbe_mac_X550 &&
7690             hw->mac.type != ixgbe_mac_X550EM_x &&
7691             hw->mac.type != ixgbe_mac_X550EM_a) {
7692                 return -ENOTSUP;
7693         }
7694
7695         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7696         etag_etype &= ~IXGBE_ETAG_ETYPE_VALID;
7697         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7698         IXGBE_WRITE_FLUSH(hw);
7699
7700         return 0;
7701 }
7702
7703 /* Disable l2 tunnel */
7704 static int
7705 ixgbe_dev_l2_tunnel_disable(struct rte_eth_dev *dev,
7706                             enum rte_eth_tunnel_type l2_tunnel_type)
7707 {
7708         int ret = 0;
7709         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7710         struct ixgbe_l2_tn_info *l2_tn_info =
7711                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7712
7713         switch (l2_tunnel_type) {
7714         case RTE_L2_TUNNEL_TYPE_E_TAG:
7715                 l2_tn_info->e_tag_en = FALSE;
7716                 ret = ixgbe_e_tag_disable(hw);
7717                 break;
7718         default:
7719                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7720                 ret = -EINVAL;
7721                 break;
7722         }
7723
7724         return ret;
7725 }
7726
7727 static int
7728 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
7729                        struct rte_eth_l2_tunnel_conf *l2_tunnel)
7730 {
7731         int ret = 0;
7732         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7733         uint32_t i, rar_entries;
7734         uint32_t rar_low, rar_high;
7735
7736         if (hw->mac.type != ixgbe_mac_X550 &&
7737             hw->mac.type != ixgbe_mac_X550EM_x &&
7738             hw->mac.type != ixgbe_mac_X550EM_a) {
7739                 return -ENOTSUP;
7740         }
7741
7742         rar_entries = ixgbe_get_num_rx_addrs(hw);
7743
7744         for (i = 1; i < rar_entries; i++) {
7745                 rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7746                 rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
7747                 if ((rar_high & IXGBE_RAH_AV) &&
7748                     (rar_high & IXGBE_RAH_ADTYPE) &&
7749                     ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
7750                      l2_tunnel->tunnel_id)) {
7751                         IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
7752                         IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
7753
7754                         ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
7755
7756                         return ret;
7757                 }
7758         }
7759
7760         return ret;
7761 }
7762
7763 static int
7764 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
7765                        struct rte_eth_l2_tunnel_conf *l2_tunnel)
7766 {
7767         int ret = 0;
7768         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7769         uint32_t i, rar_entries;
7770         uint32_t rar_low, rar_high;
7771
7772         if (hw->mac.type != ixgbe_mac_X550 &&
7773             hw->mac.type != ixgbe_mac_X550EM_x &&
7774             hw->mac.type != ixgbe_mac_X550EM_a) {
7775                 return -ENOTSUP;
7776         }
7777
7778         /* One entry for one tunnel. Try to remove potential existing entry. */
7779         ixgbe_e_tag_filter_del(dev, l2_tunnel);
7780
7781         rar_entries = ixgbe_get_num_rx_addrs(hw);
7782
7783         for (i = 1; i < rar_entries; i++) {
7784                 rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7785                 if (rar_high & IXGBE_RAH_AV) {
7786                         continue;
7787                 } else {
7788                         ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
7789                         rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
7790                         rar_low = l2_tunnel->tunnel_id;
7791
7792                         IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
7793                         IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
7794
7795                         return ret;
7796                 }
7797         }
7798
7799         PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
7800                      " Please remove a rule before adding a new one.");
7801         return -EINVAL;
7802 }
7803
7804 static inline struct ixgbe_l2_tn_filter *
7805 ixgbe_l2_tn_filter_lookup(struct ixgbe_l2_tn_info *l2_tn_info,
7806                           struct ixgbe_l2_tn_key *key)
7807 {
7808         int ret;
7809
7810         ret = rte_hash_lookup(l2_tn_info->hash_handle, (const void *)key);
7811         if (ret < 0)
7812                 return NULL;
7813
7814         return l2_tn_info->hash_map[ret];
7815 }
7816
7817 static inline int
7818 ixgbe_insert_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7819                           struct ixgbe_l2_tn_filter *l2_tn_filter)
7820 {
7821         int ret;
7822
7823         ret = rte_hash_add_key(l2_tn_info->hash_handle,
7824                                &l2_tn_filter->key);
7825
7826         if (ret < 0) {
7827                 PMD_DRV_LOG(ERR,
7828                             "Failed to insert L2 tunnel filter"
7829                             " to hash table %d!",
7830                             ret);
7831                 return ret;
7832         }
7833
7834         l2_tn_info->hash_map[ret] = l2_tn_filter;
7835
7836         TAILQ_INSERT_TAIL(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7837
7838         return 0;
7839 }
7840
7841 static inline int
7842 ixgbe_remove_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7843                           struct ixgbe_l2_tn_key *key)
7844 {
7845         int ret;
7846         struct ixgbe_l2_tn_filter *l2_tn_filter;
7847
7848         ret = rte_hash_del_key(l2_tn_info->hash_handle, key);
7849
7850         if (ret < 0) {
7851                 PMD_DRV_LOG(ERR,
7852                             "No such L2 tunnel filter to delete %d!",
7853                             ret);
7854                 return ret;
7855         }
7856
7857         l2_tn_filter = l2_tn_info->hash_map[ret];
7858         l2_tn_info->hash_map[ret] = NULL;
7859
7860         TAILQ_REMOVE(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7861         rte_free(l2_tn_filter);
7862
7863         return 0;
7864 }
7865
7866 /* Add l2 tunnel filter */
7867 int
7868 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
7869                                struct rte_eth_l2_tunnel_conf *l2_tunnel,
7870                                bool restore)
7871 {
7872         int ret;
7873         struct ixgbe_l2_tn_info *l2_tn_info =
7874                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7875         struct ixgbe_l2_tn_key key;
7876         struct ixgbe_l2_tn_filter *node;
7877
7878         if (!restore) {
7879                 key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7880                 key.tn_id = l2_tunnel->tunnel_id;
7881
7882                 node = ixgbe_l2_tn_filter_lookup(l2_tn_info, &key);
7883
7884                 if (node) {
7885                         PMD_DRV_LOG(ERR,
7886                                     "The L2 tunnel filter already exists!");
7887                         return -EINVAL;
7888                 }
7889
7890                 node = rte_zmalloc("ixgbe_l2_tn",
7891                                    sizeof(struct ixgbe_l2_tn_filter),
7892                                    0);
7893                 if (!node)
7894                         return -ENOMEM;
7895
7896                 rte_memcpy(&node->key,
7897                                  &key,
7898                                  sizeof(struct ixgbe_l2_tn_key));
7899                 node->pool = l2_tunnel->pool;
7900                 ret = ixgbe_insert_l2_tn_filter(l2_tn_info, node);
7901                 if (ret < 0) {
7902                         rte_free(node);
7903                         return ret;
7904                 }
7905         }
7906
7907         switch (l2_tunnel->l2_tunnel_type) {
7908         case RTE_L2_TUNNEL_TYPE_E_TAG:
7909                 ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
7910                 break;
7911         default:
7912                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7913                 ret = -EINVAL;
7914                 break;
7915         }
7916
7917         if ((!restore) && (ret < 0))
7918                 (void)ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7919
7920         return ret;
7921 }
7922
7923 /* Delete l2 tunnel filter */
7924 int
7925 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
7926                                struct rte_eth_l2_tunnel_conf *l2_tunnel)
7927 {
7928         int ret;
7929         struct ixgbe_l2_tn_info *l2_tn_info =
7930                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7931         struct ixgbe_l2_tn_key key;
7932
7933         key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7934         key.tn_id = l2_tunnel->tunnel_id;
7935         ret = ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7936         if (ret < 0)
7937                 return ret;
7938
7939         switch (l2_tunnel->l2_tunnel_type) {
7940         case RTE_L2_TUNNEL_TYPE_E_TAG:
7941                 ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
7942                 break;
7943         default:
7944                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7945                 ret = -EINVAL;
7946                 break;
7947         }
7948
7949         return ret;
7950 }
7951
7952 /**
7953  * ixgbe_dev_l2_tunnel_filter_handle - Handle operations for l2 tunnel filter.
7954  * @dev: pointer to rte_eth_dev structure
7955  * @filter_op:operation will be taken.
7956  * @arg: a pointer to specific structure corresponding to the filter_op
7957  */
7958 static int
7959 ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
7960                                   enum rte_filter_op filter_op,
7961                                   void *arg)
7962 {
7963         int ret;
7964
7965         if (filter_op == RTE_ETH_FILTER_NOP)
7966                 return 0;
7967
7968         if (arg == NULL) {
7969                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
7970                             filter_op);
7971                 return -EINVAL;
7972         }
7973
7974         switch (filter_op) {
7975         case RTE_ETH_FILTER_ADD:
7976                 ret = ixgbe_dev_l2_tunnel_filter_add
7977                         (dev,
7978                          (struct rte_eth_l2_tunnel_conf *)arg,
7979                          FALSE);
7980                 break;
7981         case RTE_ETH_FILTER_DELETE:
7982                 ret = ixgbe_dev_l2_tunnel_filter_del
7983                         (dev,
7984                          (struct rte_eth_l2_tunnel_conf *)arg);
7985                 break;
7986         default:
7987                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
7988                 ret = -EINVAL;
7989                 break;
7990         }
7991         return ret;
7992 }
7993
7994 static int
7995 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
7996 {
7997         int ret = 0;
7998         uint32_t ctrl;
7999         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8000
8001         if (hw->mac.type != ixgbe_mac_X550 &&
8002             hw->mac.type != ixgbe_mac_X550EM_x &&
8003             hw->mac.type != ixgbe_mac_X550EM_a) {
8004                 return -ENOTSUP;
8005         }
8006
8007         ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
8008         ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
8009         if (en)
8010                 ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
8011         IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
8012
8013         return ret;
8014 }
8015
8016 /* Enable l2 tunnel forwarding */
8017 static int
8018 ixgbe_dev_l2_tunnel_forwarding_enable
8019         (struct rte_eth_dev *dev,
8020          enum rte_eth_tunnel_type l2_tunnel_type)
8021 {
8022         struct ixgbe_l2_tn_info *l2_tn_info =
8023                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8024         int ret = 0;
8025
8026         switch (l2_tunnel_type) {
8027         case RTE_L2_TUNNEL_TYPE_E_TAG:
8028                 l2_tn_info->e_tag_fwd_en = TRUE;
8029                 ret = ixgbe_e_tag_forwarding_en_dis(dev, 1);
8030                 break;
8031         default:
8032                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8033                 ret = -EINVAL;
8034                 break;
8035         }
8036
8037         return ret;
8038 }
8039
8040 /* Disable l2 tunnel forwarding */
8041 static int
8042 ixgbe_dev_l2_tunnel_forwarding_disable
8043         (struct rte_eth_dev *dev,
8044          enum rte_eth_tunnel_type l2_tunnel_type)
8045 {
8046         struct ixgbe_l2_tn_info *l2_tn_info =
8047                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8048         int ret = 0;
8049
8050         switch (l2_tunnel_type) {
8051         case RTE_L2_TUNNEL_TYPE_E_TAG:
8052                 l2_tn_info->e_tag_fwd_en = FALSE;
8053                 ret = ixgbe_e_tag_forwarding_en_dis(dev, 0);
8054                 break;
8055         default:
8056                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8057                 ret = -EINVAL;
8058                 break;
8059         }
8060
8061         return ret;
8062 }
8063
8064 static int
8065 ixgbe_e_tag_insertion_en_dis(struct rte_eth_dev *dev,
8066                              struct rte_eth_l2_tunnel_conf *l2_tunnel,
8067                              bool en)
8068 {
8069         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
8070         int ret = 0;
8071         uint32_t vmtir, vmvir;
8072         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8073
8074         if (l2_tunnel->vf_id >= pci_dev->max_vfs) {
8075                 PMD_DRV_LOG(ERR,
8076                             "VF id %u should be less than %u",
8077                             l2_tunnel->vf_id,
8078                             pci_dev->max_vfs);
8079                 return -EINVAL;
8080         }
8081
8082         if (hw->mac.type != ixgbe_mac_X550 &&
8083             hw->mac.type != ixgbe_mac_X550EM_x &&
8084             hw->mac.type != ixgbe_mac_X550EM_a) {
8085                 return -ENOTSUP;
8086         }
8087
8088         if (en)
8089                 vmtir = l2_tunnel->tunnel_id;
8090         else
8091                 vmtir = 0;
8092
8093         IXGBE_WRITE_REG(hw, IXGBE_VMTIR(l2_tunnel->vf_id), vmtir);
8094
8095         vmvir = IXGBE_READ_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id));
8096         vmvir &= ~IXGBE_VMVIR_TAGA_MASK;
8097         if (en)
8098                 vmvir |= IXGBE_VMVIR_TAGA_ETAG_INSERT;
8099         IXGBE_WRITE_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id), vmvir);
8100
8101         return ret;
8102 }
8103
8104 /* Enable l2 tunnel tag insertion */
8105 static int
8106 ixgbe_dev_l2_tunnel_insertion_enable(struct rte_eth_dev *dev,
8107                                      struct rte_eth_l2_tunnel_conf *l2_tunnel)
8108 {
8109         int ret = 0;
8110
8111         switch (l2_tunnel->l2_tunnel_type) {
8112         case RTE_L2_TUNNEL_TYPE_E_TAG:
8113                 ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 1);
8114                 break;
8115         default:
8116                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8117                 ret = -EINVAL;
8118                 break;
8119         }
8120
8121         return ret;
8122 }
8123
8124 /* Disable l2 tunnel tag insertion */
8125 static int
8126 ixgbe_dev_l2_tunnel_insertion_disable
8127         (struct rte_eth_dev *dev,
8128          struct rte_eth_l2_tunnel_conf *l2_tunnel)
8129 {
8130         int ret = 0;
8131
8132         switch (l2_tunnel->l2_tunnel_type) {
8133         case RTE_L2_TUNNEL_TYPE_E_TAG:
8134                 ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 0);
8135                 break;
8136         default:
8137                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8138                 ret = -EINVAL;
8139                 break;
8140         }
8141
8142         return ret;
8143 }
8144
8145 static int
8146 ixgbe_e_tag_stripping_en_dis(struct rte_eth_dev *dev,
8147                              bool en)
8148 {
8149         int ret = 0;
8150         uint32_t qde;
8151         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8152
8153         if (hw->mac.type != ixgbe_mac_X550 &&
8154             hw->mac.type != ixgbe_mac_X550EM_x &&
8155             hw->mac.type != ixgbe_mac_X550EM_a) {
8156                 return -ENOTSUP;
8157         }
8158
8159         qde = IXGBE_READ_REG(hw, IXGBE_QDE);
8160         if (en)
8161                 qde |= IXGBE_QDE_STRIP_TAG;
8162         else
8163                 qde &= ~IXGBE_QDE_STRIP_TAG;
8164         qde &= ~IXGBE_QDE_READ;
8165         qde |= IXGBE_QDE_WRITE;
8166         IXGBE_WRITE_REG(hw, IXGBE_QDE, qde);
8167
8168         return ret;
8169 }
8170
8171 /* Enable l2 tunnel tag stripping */
8172 static int
8173 ixgbe_dev_l2_tunnel_stripping_enable
8174         (struct rte_eth_dev *dev,
8175          enum rte_eth_tunnel_type l2_tunnel_type)
8176 {
8177         int ret = 0;
8178
8179         switch (l2_tunnel_type) {
8180         case RTE_L2_TUNNEL_TYPE_E_TAG:
8181                 ret = ixgbe_e_tag_stripping_en_dis(dev, 1);
8182                 break;
8183         default:
8184                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8185                 ret = -EINVAL;
8186                 break;
8187         }
8188
8189         return ret;
8190 }
8191
8192 /* Disable l2 tunnel tag stripping */
8193 static int
8194 ixgbe_dev_l2_tunnel_stripping_disable
8195         (struct rte_eth_dev *dev,
8196          enum rte_eth_tunnel_type l2_tunnel_type)
8197 {
8198         int ret = 0;
8199
8200         switch (l2_tunnel_type) {
8201         case RTE_L2_TUNNEL_TYPE_E_TAG:
8202                 ret = ixgbe_e_tag_stripping_en_dis(dev, 0);
8203                 break;
8204         default:
8205                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8206                 ret = -EINVAL;
8207                 break;
8208         }
8209
8210         return ret;
8211 }
8212
8213 /* Enable/disable l2 tunnel offload functions */
8214 static int
8215 ixgbe_dev_l2_tunnel_offload_set
8216         (struct rte_eth_dev *dev,
8217          struct rte_eth_l2_tunnel_conf *l2_tunnel,
8218          uint32_t mask,
8219          uint8_t en)
8220 {
8221         int ret = 0;
8222
8223         if (l2_tunnel == NULL)
8224                 return -EINVAL;
8225
8226         ret = -EINVAL;
8227         if (mask & ETH_L2_TUNNEL_ENABLE_MASK) {
8228                 if (en)
8229                         ret = ixgbe_dev_l2_tunnel_enable(
8230                                 dev,
8231                                 l2_tunnel->l2_tunnel_type);
8232                 else
8233                         ret = ixgbe_dev_l2_tunnel_disable(
8234                                 dev,
8235                                 l2_tunnel->l2_tunnel_type);
8236         }
8237
8238         if (mask & ETH_L2_TUNNEL_INSERTION_MASK) {
8239                 if (en)
8240                         ret = ixgbe_dev_l2_tunnel_insertion_enable(
8241                                 dev,
8242                                 l2_tunnel);
8243                 else
8244                         ret = ixgbe_dev_l2_tunnel_insertion_disable(
8245                                 dev,
8246                                 l2_tunnel);
8247         }
8248
8249         if (mask & ETH_L2_TUNNEL_STRIPPING_MASK) {
8250                 if (en)
8251                         ret = ixgbe_dev_l2_tunnel_stripping_enable(
8252                                 dev,
8253                                 l2_tunnel->l2_tunnel_type);
8254                 else
8255                         ret = ixgbe_dev_l2_tunnel_stripping_disable(
8256                                 dev,
8257                                 l2_tunnel->l2_tunnel_type);
8258         }
8259
8260         if (mask & ETH_L2_TUNNEL_FORWARDING_MASK) {
8261                 if (en)
8262                         ret = ixgbe_dev_l2_tunnel_forwarding_enable(
8263                                 dev,
8264                                 l2_tunnel->l2_tunnel_type);
8265                 else
8266                         ret = ixgbe_dev_l2_tunnel_forwarding_disable(
8267                                 dev,
8268                                 l2_tunnel->l2_tunnel_type);
8269         }
8270
8271         return ret;
8272 }
8273
8274 static int
8275 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
8276                         uint16_t port)
8277 {
8278         IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
8279         IXGBE_WRITE_FLUSH(hw);
8280
8281         return 0;
8282 }
8283
8284 /* There's only one register for VxLAN UDP port.
8285  * So, we cannot add several ports. Will update it.
8286  */
8287 static int
8288 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
8289                      uint16_t port)
8290 {
8291         if (port == 0) {
8292                 PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
8293                 return -EINVAL;
8294         }
8295
8296         return ixgbe_update_vxlan_port(hw, port);
8297 }
8298
8299 /* We cannot delete the VxLAN port. For there's a register for VxLAN
8300  * UDP port, it must have a value.
8301  * So, will reset it to the original value 0.
8302  */
8303 static int
8304 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
8305                      uint16_t port)
8306 {
8307         uint16_t cur_port;
8308
8309         cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
8310
8311         if (cur_port != port) {
8312                 PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
8313                 return -EINVAL;
8314         }
8315
8316         return ixgbe_update_vxlan_port(hw, 0);
8317 }
8318
8319 /* Add UDP tunneling port */
8320 static int
8321 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
8322                               struct rte_eth_udp_tunnel *udp_tunnel)
8323 {
8324         int ret = 0;
8325         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8326
8327         if (hw->mac.type != ixgbe_mac_X550 &&
8328             hw->mac.type != ixgbe_mac_X550EM_x &&
8329             hw->mac.type != ixgbe_mac_X550EM_a) {
8330                 return -ENOTSUP;
8331         }
8332
8333         if (udp_tunnel == NULL)
8334                 return -EINVAL;
8335
8336         switch (udp_tunnel->prot_type) {
8337         case RTE_TUNNEL_TYPE_VXLAN:
8338                 ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
8339                 break;
8340
8341         case RTE_TUNNEL_TYPE_GENEVE:
8342         case RTE_TUNNEL_TYPE_TEREDO:
8343                 PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
8344                 ret = -EINVAL;
8345                 break;
8346
8347         default:
8348                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8349                 ret = -EINVAL;
8350                 break;
8351         }
8352
8353         return ret;
8354 }
8355
8356 /* Remove UDP tunneling port */
8357 static int
8358 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
8359                               struct rte_eth_udp_tunnel *udp_tunnel)
8360 {
8361         int ret = 0;
8362         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8363
8364         if (hw->mac.type != ixgbe_mac_X550 &&
8365             hw->mac.type != ixgbe_mac_X550EM_x &&
8366             hw->mac.type != ixgbe_mac_X550EM_a) {
8367                 return -ENOTSUP;
8368         }
8369
8370         if (udp_tunnel == NULL)
8371                 return -EINVAL;
8372
8373         switch (udp_tunnel->prot_type) {
8374         case RTE_TUNNEL_TYPE_VXLAN:
8375                 ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
8376                 break;
8377         case RTE_TUNNEL_TYPE_GENEVE:
8378         case RTE_TUNNEL_TYPE_TEREDO:
8379                 PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
8380                 ret = -EINVAL;
8381                 break;
8382         default:
8383                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8384                 ret = -EINVAL;
8385                 break;
8386         }
8387
8388         return ret;
8389 }
8390
8391 static void
8392 ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev)
8393 {
8394         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8395
8396         hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_PROMISC);
8397 }
8398
8399 static void
8400 ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev)
8401 {
8402         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8403
8404         hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_NONE);
8405 }
8406
8407 static void
8408 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
8409 {
8410         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8411
8412         hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_ALLMULTI);
8413 }
8414
8415 static void
8416 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
8417 {
8418         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8419
8420         hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_MULTI);
8421 }
8422
8423 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
8424 {
8425         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8426         u32 in_msg = 0;
8427
8428         /* peek the message first */
8429         in_msg = IXGBE_READ_REG(hw, IXGBE_VFMBMEM);
8430
8431         /* PF reset VF event */
8432         if (in_msg == IXGBE_PF_CONTROL_MSG) {
8433                 /* dummy mbx read to ack pf */
8434                 if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
8435                         return;
8436                 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
8437                                               NULL);
8438         }
8439 }
8440
8441 static int
8442 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
8443 {
8444         uint32_t eicr;
8445         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8446         struct ixgbe_interrupt *intr =
8447                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8448         ixgbevf_intr_disable(dev);
8449
8450         /* read-on-clear nic registers here */
8451         eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
8452         intr->flags = 0;
8453
8454         /* only one misc vector supported - mailbox */
8455         eicr &= IXGBE_VTEICR_MASK;
8456         if (eicr == IXGBE_MISC_VEC_ID)
8457                 intr->flags |= IXGBE_FLAG_MAILBOX;
8458
8459         return 0;
8460 }
8461
8462 static int
8463 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
8464 {
8465         struct ixgbe_interrupt *intr =
8466                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8467
8468         if (intr->flags & IXGBE_FLAG_MAILBOX) {
8469                 ixgbevf_mbx_process(dev);
8470                 intr->flags &= ~IXGBE_FLAG_MAILBOX;
8471         }
8472
8473         ixgbevf_intr_enable(dev);
8474
8475         return 0;
8476 }
8477
8478 static void
8479 ixgbevf_dev_interrupt_handler(void *param)
8480 {
8481         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
8482
8483         ixgbevf_dev_interrupt_get_status(dev);
8484         ixgbevf_dev_interrupt_action(dev);
8485 }
8486
8487 /**
8488  *  ixgbe_disable_sec_tx_path_generic - Stops the transmit data path
8489  *  @hw: pointer to hardware structure
8490  *
8491  *  Stops the transmit data path and waits for the HW to internally empty
8492  *  the Tx security block
8493  **/
8494 int ixgbe_disable_sec_tx_path_generic(struct ixgbe_hw *hw)
8495 {
8496 #define IXGBE_MAX_SECTX_POLL 40
8497
8498         int i;
8499         int sectxreg;
8500
8501         sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8502         sectxreg |= IXGBE_SECTXCTRL_TX_DIS;
8503         IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8504         for (i = 0; i < IXGBE_MAX_SECTX_POLL; i++) {
8505                 sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXSTAT);
8506                 if (sectxreg & IXGBE_SECTXSTAT_SECTX_RDY)
8507                         break;
8508                 /* Use interrupt-safe sleep just in case */
8509                 usec_delay(1000);
8510         }
8511
8512         /* For informational purposes only */
8513         if (i >= IXGBE_MAX_SECTX_POLL)
8514                 PMD_DRV_LOG(DEBUG, "Tx unit being enabled before security "
8515                          "path fully disabled.  Continuing with init.");
8516
8517         return IXGBE_SUCCESS;
8518 }
8519
8520 /**
8521  *  ixgbe_enable_sec_tx_path_generic - Enables the transmit data path
8522  *  @hw: pointer to hardware structure
8523  *
8524  *  Enables the transmit data path.
8525  **/
8526 int ixgbe_enable_sec_tx_path_generic(struct ixgbe_hw *hw)
8527 {
8528         uint32_t sectxreg;
8529
8530         sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8531         sectxreg &= ~IXGBE_SECTXCTRL_TX_DIS;
8532         IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8533         IXGBE_WRITE_FLUSH(hw);
8534
8535         return IXGBE_SUCCESS;
8536 }
8537
8538 /* restore n-tuple filter */
8539 static inline void
8540 ixgbe_ntuple_filter_restore(struct rte_eth_dev *dev)
8541 {
8542         struct ixgbe_filter_info *filter_info =
8543                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8544         struct ixgbe_5tuple_filter *node;
8545
8546         TAILQ_FOREACH(node, &filter_info->fivetuple_list, entries) {
8547                 ixgbe_inject_5tuple_filter(dev, node);
8548         }
8549 }
8550
8551 /* restore ethernet type filter */
8552 static inline void
8553 ixgbe_ethertype_filter_restore(struct rte_eth_dev *dev)
8554 {
8555         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8556         struct ixgbe_filter_info *filter_info =
8557                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8558         int i;
8559
8560         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8561                 if (filter_info->ethertype_mask & (1 << i)) {
8562                         IXGBE_WRITE_REG(hw, IXGBE_ETQF(i),
8563                                         filter_info->ethertype_filters[i].etqf);
8564                         IXGBE_WRITE_REG(hw, IXGBE_ETQS(i),
8565                                         filter_info->ethertype_filters[i].etqs);
8566                         IXGBE_WRITE_FLUSH(hw);
8567                 }
8568         }
8569 }
8570
8571 /* restore SYN filter */
8572 static inline void
8573 ixgbe_syn_filter_restore(struct rte_eth_dev *dev)
8574 {
8575         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8576         struct ixgbe_filter_info *filter_info =
8577                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8578         uint32_t synqf;
8579
8580         synqf = filter_info->syn_info;
8581
8582         if (synqf & IXGBE_SYN_FILTER_ENABLE) {
8583                 IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
8584                 IXGBE_WRITE_FLUSH(hw);
8585         }
8586 }
8587
8588 /* restore L2 tunnel filter */
8589 static inline void
8590 ixgbe_l2_tn_filter_restore(struct rte_eth_dev *dev)
8591 {
8592         struct ixgbe_l2_tn_info *l2_tn_info =
8593                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8594         struct ixgbe_l2_tn_filter *node;
8595         struct rte_eth_l2_tunnel_conf l2_tn_conf;
8596
8597         TAILQ_FOREACH(node, &l2_tn_info->l2_tn_list, entries) {
8598                 l2_tn_conf.l2_tunnel_type = node->key.l2_tn_type;
8599                 l2_tn_conf.tunnel_id      = node->key.tn_id;
8600                 l2_tn_conf.pool           = node->pool;
8601                 (void)ixgbe_dev_l2_tunnel_filter_add(dev, &l2_tn_conf, TRUE);
8602         }
8603 }
8604
8605 /* restore rss filter */
8606 static inline void
8607 ixgbe_rss_filter_restore(struct rte_eth_dev *dev)
8608 {
8609         struct ixgbe_filter_info *filter_info =
8610                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8611
8612         if (filter_info->rss_info.conf.queue_num)
8613                 ixgbe_config_rss_filter(dev,
8614                         &filter_info->rss_info, TRUE);
8615 }
8616
8617 static int
8618 ixgbe_filter_restore(struct rte_eth_dev *dev)
8619 {
8620         ixgbe_ntuple_filter_restore(dev);
8621         ixgbe_ethertype_filter_restore(dev);
8622         ixgbe_syn_filter_restore(dev);
8623         ixgbe_fdir_filter_restore(dev);
8624         ixgbe_l2_tn_filter_restore(dev);
8625         ixgbe_rss_filter_restore(dev);
8626
8627         return 0;
8628 }
8629
8630 static void
8631 ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev)
8632 {
8633         struct ixgbe_l2_tn_info *l2_tn_info =
8634                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8635         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8636
8637         if (l2_tn_info->e_tag_en)
8638                 (void)ixgbe_e_tag_enable(hw);
8639
8640         if (l2_tn_info->e_tag_fwd_en)
8641                 (void)ixgbe_e_tag_forwarding_en_dis(dev, 1);
8642
8643         (void)ixgbe_update_e_tag_eth_type(hw, l2_tn_info->e_tag_ether_type);
8644 }
8645
8646 /* remove all the n-tuple filters */
8647 void
8648 ixgbe_clear_all_ntuple_filter(struct rte_eth_dev *dev)
8649 {
8650         struct ixgbe_filter_info *filter_info =
8651                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8652         struct ixgbe_5tuple_filter *p_5tuple;
8653
8654         while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
8655                 ixgbe_remove_5tuple_filter(dev, p_5tuple);
8656 }
8657
8658 /* remove all the ether type filters */
8659 void
8660 ixgbe_clear_all_ethertype_filter(struct rte_eth_dev *dev)
8661 {
8662         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8663         struct ixgbe_filter_info *filter_info =
8664                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8665         int i;
8666
8667         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8668                 if (filter_info->ethertype_mask & (1 << i) &&
8669                     !filter_info->ethertype_filters[i].conf) {
8670                         (void)ixgbe_ethertype_filter_remove(filter_info,
8671                                                             (uint8_t)i);
8672                         IXGBE_WRITE_REG(hw, IXGBE_ETQF(i), 0);
8673                         IXGBE_WRITE_REG(hw, IXGBE_ETQS(i), 0);
8674                         IXGBE_WRITE_FLUSH(hw);
8675                 }
8676         }
8677 }
8678
8679 /* remove the SYN filter */
8680 void
8681 ixgbe_clear_syn_filter(struct rte_eth_dev *dev)
8682 {
8683         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8684         struct ixgbe_filter_info *filter_info =
8685                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8686
8687         if (filter_info->syn_info & IXGBE_SYN_FILTER_ENABLE) {
8688                 filter_info->syn_info = 0;
8689
8690                 IXGBE_WRITE_REG(hw, IXGBE_SYNQF, 0);
8691                 IXGBE_WRITE_FLUSH(hw);
8692         }
8693 }
8694
8695 /* remove all the L2 tunnel filters */
8696 int
8697 ixgbe_clear_all_l2_tn_filter(struct rte_eth_dev *dev)
8698 {
8699         struct ixgbe_l2_tn_info *l2_tn_info =
8700                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8701         struct ixgbe_l2_tn_filter *l2_tn_filter;
8702         struct rte_eth_l2_tunnel_conf l2_tn_conf;
8703         int ret = 0;
8704
8705         while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
8706                 l2_tn_conf.l2_tunnel_type = l2_tn_filter->key.l2_tn_type;
8707                 l2_tn_conf.tunnel_id      = l2_tn_filter->key.tn_id;
8708                 l2_tn_conf.pool           = l2_tn_filter->pool;
8709                 ret = ixgbe_dev_l2_tunnel_filter_del(dev, &l2_tn_conf);
8710                 if (ret < 0)
8711                         return ret;
8712         }
8713
8714         return 0;
8715 }
8716
8717 RTE_PMD_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd);
8718 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
8719 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe, "* igb_uio | uio_pci_generic | vfio-pci");
8720 RTE_PMD_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd);
8721 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);
8722 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe_vf, "* igb_uio | vfio-pci");
8723 RTE_PMD_REGISTER_PARAM_STRING(net_ixgbe_vf,
8724                               IXGBEVF_DEVARG_PFLINK_FULLCHK "=<0|1>");
8725
8726 RTE_INIT(ixgbe_init_log)
8727 {
8728         ixgbe_logtype_init = rte_log_register("pmd.net.ixgbe.init");
8729         if (ixgbe_logtype_init >= 0)
8730                 rte_log_set_level(ixgbe_logtype_init, RTE_LOG_NOTICE);
8731         ixgbe_logtype_driver = rte_log_register("pmd.net.ixgbe.driver");
8732         if (ixgbe_logtype_driver >= 0)
8733                 rte_log_set_level(ixgbe_logtype_driver, RTE_LOG_NOTICE);
8734 }