net/ice: fix max frame size
[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_eal.h>
27 #include <rte_alarm.h>
28 #include <rte_ether.h>
29 #include <rte_ethdev_driver.h>
30 #include <rte_ethdev_pci.h>
31 #include <rte_malloc.h>
32 #include <rte_random.h>
33 #include <rte_dev.h>
34 #include <rte_hash_crc.h>
35 #ifdef RTE_LIBRTE_SECURITY
36 #include <rte_security_driver.h>
37 #endif
38
39 #include "ixgbe_logs.h"
40 #include "base/ixgbe_api.h"
41 #include "base/ixgbe_vf.h"
42 #include "base/ixgbe_common.h"
43 #include "ixgbe_ethdev.h"
44 #include "ixgbe_bypass.h"
45 #include "ixgbe_rxtx.h"
46 #include "base/ixgbe_type.h"
47 #include "base/ixgbe_phy.h"
48 #include "ixgbe_regs.h"
49
50 /*
51  * High threshold controlling when to start sending XOFF frames. Must be at
52  * least 8 bytes less than receive packet buffer size. This value is in units
53  * of 1024 bytes.
54  */
55 #define IXGBE_FC_HI    0x80
56
57 /*
58  * Low threshold controlling when to start sending XON frames. This value is
59  * in units of 1024 bytes.
60  */
61 #define IXGBE_FC_LO    0x40
62
63 /* Timer value included in XOFF frames. */
64 #define IXGBE_FC_PAUSE 0x680
65
66 /*Default value of Max Rx Queue*/
67 #define IXGBE_MAX_RX_QUEUE_NUM 128
68
69 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
70 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
71 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
72
73 #define IXGBE_MMW_SIZE_DEFAULT        0x4
74 #define IXGBE_MMW_SIZE_JUMBO_FRAME    0x14
75 #define IXGBE_MAX_RING_DESC           4096 /* replicate define from rxtx */
76
77 /*
78  *  Default values for RX/TX configuration
79  */
80 #define IXGBE_DEFAULT_RX_FREE_THRESH  32
81 #define IXGBE_DEFAULT_RX_PTHRESH      8
82 #define IXGBE_DEFAULT_RX_HTHRESH      8
83 #define IXGBE_DEFAULT_RX_WTHRESH      0
84
85 #define IXGBE_DEFAULT_TX_FREE_THRESH  32
86 #define IXGBE_DEFAULT_TX_PTHRESH      32
87 #define IXGBE_DEFAULT_TX_HTHRESH      0
88 #define IXGBE_DEFAULT_TX_WTHRESH      0
89 #define IXGBE_DEFAULT_TX_RSBIT_THRESH 32
90
91 /* Bit shift and mask */
92 #define IXGBE_4_BIT_WIDTH  (CHAR_BIT / 2)
93 #define IXGBE_4_BIT_MASK   RTE_LEN2MASK(IXGBE_4_BIT_WIDTH, uint8_t)
94 #define IXGBE_8_BIT_WIDTH  CHAR_BIT
95 #define IXGBE_8_BIT_MASK   UINT8_MAX
96
97 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
98
99 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
100
101 /* Additional timesync values. */
102 #define NSEC_PER_SEC             1000000000L
103 #define IXGBE_INCVAL_10GB        0x66666666
104 #define IXGBE_INCVAL_1GB         0x40000000
105 #define IXGBE_INCVAL_100         0x50000000
106 #define IXGBE_INCVAL_SHIFT_10GB  28
107 #define IXGBE_INCVAL_SHIFT_1GB   24
108 #define IXGBE_INCVAL_SHIFT_100   21
109 #define IXGBE_INCVAL_SHIFT_82599 7
110 #define IXGBE_INCPER_SHIFT_82599 24
111
112 #define IXGBE_CYCLECOUNTER_MASK   0xffffffffffffffffULL
113
114 #define IXGBE_VT_CTL_POOLING_MODE_MASK         0x00030000
115 #define IXGBE_VT_CTL_POOLING_MODE_ETAG         0x00010000
116 #define IXGBE_ETAG_ETYPE                       0x00005084
117 #define IXGBE_ETAG_ETYPE_MASK                  0x0000ffff
118 #define IXGBE_ETAG_ETYPE_VALID                 0x80000000
119 #define IXGBE_RAH_ADTYPE                       0x40000000
120 #define IXGBE_RAL_ETAG_FILTER_MASK             0x00003fff
121 #define IXGBE_VMVIR_TAGA_MASK                  0x18000000
122 #define IXGBE_VMVIR_TAGA_ETAG_INSERT           0x08000000
123 #define IXGBE_VMTIR(_i) (0x00017000 + ((_i) * 4)) /* 64 of these (0-63) */
124 #define IXGBE_QDE_STRIP_TAG                    0x00000004
125 #define IXGBE_VTEICR_MASK                      0x07
126
127 #define IXGBE_EXVET_VET_EXT_SHIFT              16
128 #define IXGBE_DMATXCTL_VT_MASK                 0xFFFF0000
129
130 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params);
131 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
132 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev);
133 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev);
134 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev);
135 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev);
136 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev);
137 static int  ixgbe_dev_configure(struct rte_eth_dev *dev);
138 static int  ixgbe_dev_start(struct rte_eth_dev *dev);
139 static void ixgbe_dev_stop(struct rte_eth_dev *dev);
140 static int  ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
141 static int  ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
142 static void ixgbe_dev_close(struct rte_eth_dev *dev);
143 static int  ixgbe_dev_reset(struct rte_eth_dev *dev);
144 static void ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
145 static void ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
146 static void ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
147 static void ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
148 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
149                                 int wait_to_complete);
150 static int ixgbe_dev_stats_get(struct rte_eth_dev *dev,
151                                 struct rte_eth_stats *stats);
152 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
153                                 struct rte_eth_xstat *xstats, unsigned n);
154 static int ixgbevf_dev_xstats_get(struct rte_eth_dev *dev,
155                                   struct rte_eth_xstat *xstats, unsigned n);
156 static int
157 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
158                 uint64_t *values, unsigned int n);
159 static void ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
160 static void ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
161 static int ixgbe_dev_xstats_get_names(struct rte_eth_dev *dev,
162         struct rte_eth_xstat_name *xstats_names,
163         unsigned int size);
164 static int ixgbevf_dev_xstats_get_names(struct rte_eth_dev *dev,
165         struct rte_eth_xstat_name *xstats_names, unsigned limit);
166 static int ixgbe_dev_xstats_get_names_by_id(
167         struct rte_eth_dev *dev,
168         struct rte_eth_xstat_name *xstats_names,
169         const uint64_t *ids,
170         unsigned int limit);
171 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
172                                              uint16_t queue_id,
173                                              uint8_t stat_idx,
174                                              uint8_t is_rx);
175 static int ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
176                                  size_t fw_size);
177 static void ixgbe_dev_info_get(struct rte_eth_dev *dev,
178                                struct rte_eth_dev_info *dev_info);
179 static const uint32_t *ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev);
180 static void ixgbevf_dev_info_get(struct rte_eth_dev *dev,
181                                  struct rte_eth_dev_info *dev_info);
182 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
183
184 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
185                 uint16_t vlan_id, int on);
186 static int ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
187                                enum rte_vlan_type vlan_type,
188                                uint16_t tpid_id);
189 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
190                 uint16_t queue, bool on);
191 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
192                 int on);
193 static void ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev,
194                                                   int mask);
195 static int ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask);
196 static int ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
197 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
198 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
199 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
200 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
201
202 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
203 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
204 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
205                                struct rte_eth_fc_conf *fc_conf);
206 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
207                                struct rte_eth_fc_conf *fc_conf);
208 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
209                 struct rte_eth_pfc_conf *pfc_conf);
210 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
211                         struct rte_eth_rss_reta_entry64 *reta_conf,
212                         uint16_t reta_size);
213 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
214                         struct rte_eth_rss_reta_entry64 *reta_conf,
215                         uint16_t reta_size);
216 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
217 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
218 static int ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev);
219 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
220 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
221 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev);
222 static void ixgbe_dev_interrupt_handler(void *param);
223 static void ixgbe_dev_interrupt_delayed_handler(void *param);
224 static void ixgbe_dev_setup_link_alarm_handler(void *param);
225
226 static int ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
227                          uint32_t index, uint32_t pool);
228 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
229 static int ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
230                                            struct ether_addr *mac_addr);
231 static void ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config);
232 static bool is_device_supported(struct rte_eth_dev *dev,
233                                 struct rte_pci_driver *drv);
234
235 /* For Virtual Function support */
236 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
237 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
238 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
239 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
240 static int ixgbevf_dev_link_update(struct rte_eth_dev *dev,
241                                    int wait_to_complete);
242 static void ixgbevf_dev_stop(struct rte_eth_dev *dev);
243 static void ixgbevf_dev_close(struct rte_eth_dev *dev);
244 static int  ixgbevf_dev_reset(struct rte_eth_dev *dev);
245 static void ixgbevf_intr_disable(struct rte_eth_dev *dev);
246 static void ixgbevf_intr_enable(struct rte_eth_dev *dev);
247 static int ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
248                 struct rte_eth_stats *stats);
249 static void ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
250 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
251                 uint16_t vlan_id, int on);
252 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
253                 uint16_t queue, int on);
254 static int ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask);
255 static int ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
256 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
257 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
258                                             uint16_t queue_id);
259 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
260                                              uint16_t queue_id);
261 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
262                                  uint8_t queue, uint8_t msix_vector);
263 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
264 static void ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev);
265 static void ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev);
266 static void ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev);
267 static void ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev);
268
269 /* For Eth VMDQ APIs support */
270 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
271                 ether_addr * mac_addr, uint8_t on);
272 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on);
273 static int ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
274                 struct rte_eth_mirror_conf *mirror_conf,
275                 uint8_t rule_id, uint8_t on);
276 static int ixgbe_mirror_rule_reset(struct rte_eth_dev *dev,
277                 uint8_t rule_id);
278 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
279                                           uint16_t queue_id);
280 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
281                                            uint16_t queue_id);
282 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
283                                uint8_t queue, uint8_t msix_vector);
284 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
285
286 static int ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
287                                 struct ether_addr *mac_addr,
288                                 uint32_t index, uint32_t pool);
289 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
290 static int ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
291                                              struct ether_addr *mac_addr);
292 static int ixgbe_syn_filter_get(struct rte_eth_dev *dev,
293                         struct rte_eth_syn_filter *filter);
294 static int ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
295                         enum rte_filter_op filter_op,
296                         void *arg);
297 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
298                         struct ixgbe_5tuple_filter *filter);
299 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
300                         struct ixgbe_5tuple_filter *filter);
301 static int ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
302                                 enum rte_filter_op filter_op,
303                                 void *arg);
304 static int ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
305                         struct rte_eth_ntuple_filter *filter);
306 static int ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
307                                 enum rte_filter_op filter_op,
308                                 void *arg);
309 static int ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
310                         struct rte_eth_ethertype_filter *filter);
311 static int ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
312                      enum rte_filter_type filter_type,
313                      enum rte_filter_op filter_op,
314                      void *arg);
315 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
316
317 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
318                                       struct ether_addr *mc_addr_set,
319                                       uint32_t nb_mc_addr);
320 static int ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
321                                    struct rte_eth_dcb_info *dcb_info);
322
323 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
324 static int ixgbe_get_regs(struct rte_eth_dev *dev,
325                             struct rte_dev_reg_info *regs);
326 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
327 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
328                                 struct rte_dev_eeprom_info *eeprom);
329 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
330                                 struct rte_dev_eeprom_info *eeprom);
331
332 static int ixgbe_get_module_info(struct rte_eth_dev *dev,
333                                  struct rte_eth_dev_module_info *modinfo);
334 static int ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
335                                    struct rte_dev_eeprom_info *info);
336
337 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
338 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
339                                 struct rte_dev_reg_info *regs);
340
341 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
342 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
343 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
344                                             struct timespec *timestamp,
345                                             uint32_t flags);
346 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
347                                             struct timespec *timestamp);
348 static int ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
349 static int ixgbe_timesync_read_time(struct rte_eth_dev *dev,
350                                    struct timespec *timestamp);
351 static int ixgbe_timesync_write_time(struct rte_eth_dev *dev,
352                                    const struct timespec *timestamp);
353 static void ixgbevf_dev_interrupt_handler(void *param);
354
355 static int ixgbe_dev_l2_tunnel_eth_type_conf
356         (struct rte_eth_dev *dev, struct rte_eth_l2_tunnel_conf *l2_tunnel);
357 static int ixgbe_dev_l2_tunnel_offload_set
358         (struct rte_eth_dev *dev,
359          struct rte_eth_l2_tunnel_conf *l2_tunnel,
360          uint32_t mask,
361          uint8_t en);
362 static int ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
363                                              enum rte_filter_op filter_op,
364                                              void *arg);
365
366 static int ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
367                                          struct rte_eth_udp_tunnel *udp_tunnel);
368 static int ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
369                                          struct rte_eth_udp_tunnel *udp_tunnel);
370 static int ixgbe_filter_restore(struct rte_eth_dev *dev);
371 static void ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev);
372
373 /*
374  * Define VF Stats MACRO for Non "cleared on read" register
375  */
376 #define UPDATE_VF_STAT(reg, last, cur)                          \
377 {                                                               \
378         uint32_t latest = IXGBE_READ_REG(hw, reg);              \
379         cur += (latest - last) & UINT_MAX;                      \
380         last = latest;                                          \
381 }
382
383 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
384 {                                                                \
385         u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
386         u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
387         u64 latest = ((new_msb << 32) | new_lsb);                \
388         cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
389         last = latest;                                           \
390 }
391
392 #define IXGBE_SET_HWSTRIP(h, q) do {\
393                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
394                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
395                 (h)->bitmap[idx] |= 1 << bit;\
396         } while (0)
397
398 #define IXGBE_CLEAR_HWSTRIP(h, q) do {\
399                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
400                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
401                 (h)->bitmap[idx] &= ~(1 << bit);\
402         } while (0)
403
404 #define IXGBE_GET_HWSTRIP(h, q, r) do {\
405                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
406                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
407                 (r) = (h)->bitmap[idx] >> bit & 1;\
408         } while (0)
409
410 int ixgbe_logtype_init;
411 int ixgbe_logtype_driver;
412
413 /*
414  * The set of PCI devices this driver supports
415  */
416 static const struct rte_pci_id pci_id_ixgbe_map[] = {
417         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598) },
418         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_BX) },
419         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT) },
420         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT) },
421         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT) },
422         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2) },
423         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM) },
424         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4) },
425         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT) },
426         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT) },
427         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM) },
428         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR) },
429         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4) },
430         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ) },
431         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KR) },
432         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE) },
433         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4) },
434         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP) },
435         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE) },
436         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE) },
437         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_EM) },
438         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2) },
439         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP) },
440         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP) },
441         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP) },
442         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM) },
443         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM) },
444         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T) },
445         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1) },
446         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP) },
447         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T) },
448         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_1G_T) },
449         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T) },
450         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1) },
451         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR) },
452         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR_L) },
453         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP_N) },
454         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII) },
455         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII_L) },
456         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_10G_T) },
457         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP) },
458         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP_N) },
459         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP) },
460         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T) },
461         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T_L) },
462         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4) },
463         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR) },
464 #ifdef RTE_LIBRTE_IXGBE_BYPASS
465         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BYPASS) },
466 #endif
467         { .vendor_id = 0, /* sentinel */ },
468 };
469
470 /*
471  * The set of PCI devices this driver supports (for 82599 VF)
472  */
473 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
474         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF) },
475         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF_HV) },
476         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF) },
477         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF_HV) },
478         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF_HV) },
479         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF) },
480         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF) },
481         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF_HV) },
482         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF) },
483         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF_HV) },
484         { .vendor_id = 0, /* sentinel */ },
485 };
486
487 static const struct rte_eth_desc_lim rx_desc_lim = {
488         .nb_max = IXGBE_MAX_RING_DESC,
489         .nb_min = IXGBE_MIN_RING_DESC,
490         .nb_align = IXGBE_RXD_ALIGN,
491 };
492
493 static const struct rte_eth_desc_lim tx_desc_lim = {
494         .nb_max = IXGBE_MAX_RING_DESC,
495         .nb_min = IXGBE_MIN_RING_DESC,
496         .nb_align = IXGBE_TXD_ALIGN,
497         .nb_seg_max = IXGBE_TX_MAX_SEG,
498         .nb_mtu_seg_max = IXGBE_TX_MAX_SEG,
499 };
500
501 static const struct eth_dev_ops ixgbe_eth_dev_ops = {
502         .dev_configure        = ixgbe_dev_configure,
503         .dev_start            = ixgbe_dev_start,
504         .dev_stop             = ixgbe_dev_stop,
505         .dev_set_link_up    = ixgbe_dev_set_link_up,
506         .dev_set_link_down  = ixgbe_dev_set_link_down,
507         .dev_close            = ixgbe_dev_close,
508         .dev_reset            = ixgbe_dev_reset,
509         .promiscuous_enable   = ixgbe_dev_promiscuous_enable,
510         .promiscuous_disable  = ixgbe_dev_promiscuous_disable,
511         .allmulticast_enable  = ixgbe_dev_allmulticast_enable,
512         .allmulticast_disable = ixgbe_dev_allmulticast_disable,
513         .link_update          = ixgbe_dev_link_update,
514         .stats_get            = ixgbe_dev_stats_get,
515         .xstats_get           = ixgbe_dev_xstats_get,
516         .xstats_get_by_id     = ixgbe_dev_xstats_get_by_id,
517         .stats_reset          = ixgbe_dev_stats_reset,
518         .xstats_reset         = ixgbe_dev_xstats_reset,
519         .xstats_get_names     = ixgbe_dev_xstats_get_names,
520         .xstats_get_names_by_id = ixgbe_dev_xstats_get_names_by_id,
521         .queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
522         .fw_version_get       = ixgbe_fw_version_get,
523         .dev_infos_get        = ixgbe_dev_info_get,
524         .dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
525         .mtu_set              = ixgbe_dev_mtu_set,
526         .vlan_filter_set      = ixgbe_vlan_filter_set,
527         .vlan_tpid_set        = ixgbe_vlan_tpid_set,
528         .vlan_offload_set     = ixgbe_vlan_offload_set,
529         .vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
530         .rx_queue_start       = ixgbe_dev_rx_queue_start,
531         .rx_queue_stop        = ixgbe_dev_rx_queue_stop,
532         .tx_queue_start       = ixgbe_dev_tx_queue_start,
533         .tx_queue_stop        = ixgbe_dev_tx_queue_stop,
534         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
535         .rx_queue_intr_enable = ixgbe_dev_rx_queue_intr_enable,
536         .rx_queue_intr_disable = ixgbe_dev_rx_queue_intr_disable,
537         .rx_queue_release     = ixgbe_dev_rx_queue_release,
538         .rx_queue_count       = ixgbe_dev_rx_queue_count,
539         .rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
540         .rx_descriptor_status = ixgbe_dev_rx_descriptor_status,
541         .tx_descriptor_status = ixgbe_dev_tx_descriptor_status,
542         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
543         .tx_queue_release     = ixgbe_dev_tx_queue_release,
544         .dev_led_on           = ixgbe_dev_led_on,
545         .dev_led_off          = ixgbe_dev_led_off,
546         .flow_ctrl_get        = ixgbe_flow_ctrl_get,
547         .flow_ctrl_set        = ixgbe_flow_ctrl_set,
548         .priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
549         .mac_addr_add         = ixgbe_add_rar,
550         .mac_addr_remove      = ixgbe_remove_rar,
551         .mac_addr_set         = ixgbe_set_default_mac_addr,
552         .uc_hash_table_set    = ixgbe_uc_hash_table_set,
553         .uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
554         .mirror_rule_set      = ixgbe_mirror_rule_set,
555         .mirror_rule_reset    = ixgbe_mirror_rule_reset,
556         .set_queue_rate_limit = ixgbe_set_queue_rate_limit,
557         .reta_update          = ixgbe_dev_rss_reta_update,
558         .reta_query           = ixgbe_dev_rss_reta_query,
559         .rss_hash_update      = ixgbe_dev_rss_hash_update,
560         .rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
561         .filter_ctrl          = ixgbe_dev_filter_ctrl,
562         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
563         .rxq_info_get         = ixgbe_rxq_info_get,
564         .txq_info_get         = ixgbe_txq_info_get,
565         .timesync_enable      = ixgbe_timesync_enable,
566         .timesync_disable     = ixgbe_timesync_disable,
567         .timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
568         .timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
569         .get_reg              = ixgbe_get_regs,
570         .get_eeprom_length    = ixgbe_get_eeprom_length,
571         .get_eeprom           = ixgbe_get_eeprom,
572         .set_eeprom           = ixgbe_set_eeprom,
573         .get_module_info      = ixgbe_get_module_info,
574         .get_module_eeprom    = ixgbe_get_module_eeprom,
575         .get_dcb_info         = ixgbe_dev_get_dcb_info,
576         .timesync_adjust_time = ixgbe_timesync_adjust_time,
577         .timesync_read_time   = ixgbe_timesync_read_time,
578         .timesync_write_time  = ixgbe_timesync_write_time,
579         .l2_tunnel_eth_type_conf = ixgbe_dev_l2_tunnel_eth_type_conf,
580         .l2_tunnel_offload_set   = ixgbe_dev_l2_tunnel_offload_set,
581         .udp_tunnel_port_add  = ixgbe_dev_udp_tunnel_port_add,
582         .udp_tunnel_port_del  = ixgbe_dev_udp_tunnel_port_del,
583         .tm_ops_get           = ixgbe_tm_ops_get,
584 };
585
586 /*
587  * dev_ops for virtual function, bare necessities for basic vf
588  * operation have been implemented
589  */
590 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
591         .dev_configure        = ixgbevf_dev_configure,
592         .dev_start            = ixgbevf_dev_start,
593         .dev_stop             = ixgbevf_dev_stop,
594         .link_update          = ixgbevf_dev_link_update,
595         .stats_get            = ixgbevf_dev_stats_get,
596         .xstats_get           = ixgbevf_dev_xstats_get,
597         .stats_reset          = ixgbevf_dev_stats_reset,
598         .xstats_reset         = ixgbevf_dev_stats_reset,
599         .xstats_get_names     = ixgbevf_dev_xstats_get_names,
600         .dev_close            = ixgbevf_dev_close,
601         .dev_reset            = ixgbevf_dev_reset,
602         .promiscuous_enable   = ixgbevf_dev_promiscuous_enable,
603         .promiscuous_disable  = ixgbevf_dev_promiscuous_disable,
604         .allmulticast_enable  = ixgbevf_dev_allmulticast_enable,
605         .allmulticast_disable = ixgbevf_dev_allmulticast_disable,
606         .dev_infos_get        = ixgbevf_dev_info_get,
607         .dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
608         .mtu_set              = ixgbevf_dev_set_mtu,
609         .vlan_filter_set      = ixgbevf_vlan_filter_set,
610         .vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
611         .vlan_offload_set     = ixgbevf_vlan_offload_set,
612         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
613         .rx_queue_release     = ixgbe_dev_rx_queue_release,
614         .rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
615         .rx_descriptor_status = ixgbe_dev_rx_descriptor_status,
616         .tx_descriptor_status = ixgbe_dev_tx_descriptor_status,
617         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
618         .tx_queue_release     = ixgbe_dev_tx_queue_release,
619         .rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
620         .rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
621         .mac_addr_add         = ixgbevf_add_mac_addr,
622         .mac_addr_remove      = ixgbevf_remove_mac_addr,
623         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
624         .rxq_info_get         = ixgbe_rxq_info_get,
625         .txq_info_get         = ixgbe_txq_info_get,
626         .mac_addr_set         = ixgbevf_set_default_mac_addr,
627         .get_reg              = ixgbevf_get_regs,
628         .reta_update          = ixgbe_dev_rss_reta_update,
629         .reta_query           = ixgbe_dev_rss_reta_query,
630         .rss_hash_update      = ixgbe_dev_rss_hash_update,
631         .rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
632 };
633
634 /* store statistics names and its offset in stats structure */
635 struct rte_ixgbe_xstats_name_off {
636         char name[RTE_ETH_XSTATS_NAME_SIZE];
637         unsigned offset;
638 };
639
640 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
641         {"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
642         {"rx_illegal_byte_errors", offsetof(struct ixgbe_hw_stats, illerrc)},
643         {"rx_error_bytes", offsetof(struct ixgbe_hw_stats, errbc)},
644         {"mac_local_errors", offsetof(struct ixgbe_hw_stats, mlfc)},
645         {"mac_remote_errors", offsetof(struct ixgbe_hw_stats, mrfc)},
646         {"rx_length_errors", offsetof(struct ixgbe_hw_stats, rlec)},
647         {"tx_xon_packets", offsetof(struct ixgbe_hw_stats, lxontxc)},
648         {"rx_xon_packets", offsetof(struct ixgbe_hw_stats, lxonrxc)},
649         {"tx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxofftxc)},
650         {"rx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
651         {"rx_size_64_packets", offsetof(struct ixgbe_hw_stats, prc64)},
652         {"rx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, prc127)},
653         {"rx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, prc255)},
654         {"rx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, prc511)},
655         {"rx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
656                 prc1023)},
657         {"rx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
658                 prc1522)},
659         {"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
660         {"rx_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
661         {"rx_fragment_errors", offsetof(struct ixgbe_hw_stats, rfc)},
662         {"rx_undersize_errors", offsetof(struct ixgbe_hw_stats, ruc)},
663         {"rx_oversize_errors", offsetof(struct ixgbe_hw_stats, roc)},
664         {"rx_jabber_errors", offsetof(struct ixgbe_hw_stats, rjc)},
665         {"rx_management_packets", offsetof(struct ixgbe_hw_stats, mngprc)},
666         {"rx_management_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
667         {"tx_management_packets", offsetof(struct ixgbe_hw_stats, mngptc)},
668         {"rx_total_packets", offsetof(struct ixgbe_hw_stats, tpr)},
669         {"rx_total_bytes", offsetof(struct ixgbe_hw_stats, tor)},
670         {"tx_total_packets", offsetof(struct ixgbe_hw_stats, tpt)},
671         {"tx_size_64_packets", offsetof(struct ixgbe_hw_stats, ptc64)},
672         {"tx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, ptc127)},
673         {"tx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, ptc255)},
674         {"tx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, ptc511)},
675         {"tx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
676                 ptc1023)},
677         {"tx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
678                 ptc1522)},
679         {"tx_multicast_packets", offsetof(struct ixgbe_hw_stats, mptc)},
680         {"tx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bptc)},
681         {"rx_mac_short_packet_dropped", offsetof(struct ixgbe_hw_stats, mspdc)},
682         {"rx_l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
683
684         {"flow_director_added_filters", offsetof(struct ixgbe_hw_stats,
685                 fdirustat_add)},
686         {"flow_director_removed_filters", offsetof(struct ixgbe_hw_stats,
687                 fdirustat_remove)},
688         {"flow_director_filter_add_errors", offsetof(struct ixgbe_hw_stats,
689                 fdirfstat_fadd)},
690         {"flow_director_filter_remove_errors", offsetof(struct ixgbe_hw_stats,
691                 fdirfstat_fremove)},
692         {"flow_director_matched_filters", offsetof(struct ixgbe_hw_stats,
693                 fdirmatch)},
694         {"flow_director_missed_filters", offsetof(struct ixgbe_hw_stats,
695                 fdirmiss)},
696
697         {"rx_fcoe_crc_errors", offsetof(struct ixgbe_hw_stats, fccrc)},
698         {"rx_fcoe_dropped", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
699         {"rx_fcoe_mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats,
700                 fclast)},
701         {"rx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeprc)},
702         {"tx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeptc)},
703         {"rx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwrc)},
704         {"tx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwtc)},
705         {"rx_fcoe_no_direct_data_placement", offsetof(struct ixgbe_hw_stats,
706                 fcoe_noddp)},
707         {"rx_fcoe_no_direct_data_placement_ext_buff",
708                 offsetof(struct ixgbe_hw_stats, fcoe_noddp_ext_buff)},
709
710         {"tx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
711                 lxontxc)},
712         {"rx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
713                 lxonrxc)},
714         {"tx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
715                 lxofftxc)},
716         {"rx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
717                 lxoffrxc)},
718         {"rx_total_missed_packets", offsetof(struct ixgbe_hw_stats, mpctotal)},
719 };
720
721 #define IXGBE_NB_HW_STATS (sizeof(rte_ixgbe_stats_strings) / \
722                            sizeof(rte_ixgbe_stats_strings[0]))
723
724 /* MACsec statistics */
725 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_macsec_strings[] = {
726         {"out_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
727                 out_pkts_untagged)},
728         {"out_pkts_encrypted", offsetof(struct ixgbe_macsec_stats,
729                 out_pkts_encrypted)},
730         {"out_pkts_protected", offsetof(struct ixgbe_macsec_stats,
731                 out_pkts_protected)},
732         {"out_octets_encrypted", offsetof(struct ixgbe_macsec_stats,
733                 out_octets_encrypted)},
734         {"out_octets_protected", offsetof(struct ixgbe_macsec_stats,
735                 out_octets_protected)},
736         {"in_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
737                 in_pkts_untagged)},
738         {"in_pkts_badtag", offsetof(struct ixgbe_macsec_stats,
739                 in_pkts_badtag)},
740         {"in_pkts_nosci", offsetof(struct ixgbe_macsec_stats,
741                 in_pkts_nosci)},
742         {"in_pkts_unknownsci", offsetof(struct ixgbe_macsec_stats,
743                 in_pkts_unknownsci)},
744         {"in_octets_decrypted", offsetof(struct ixgbe_macsec_stats,
745                 in_octets_decrypted)},
746         {"in_octets_validated", offsetof(struct ixgbe_macsec_stats,
747                 in_octets_validated)},
748         {"in_pkts_unchecked", offsetof(struct ixgbe_macsec_stats,
749                 in_pkts_unchecked)},
750         {"in_pkts_delayed", offsetof(struct ixgbe_macsec_stats,
751                 in_pkts_delayed)},
752         {"in_pkts_late", offsetof(struct ixgbe_macsec_stats,
753                 in_pkts_late)},
754         {"in_pkts_ok", offsetof(struct ixgbe_macsec_stats,
755                 in_pkts_ok)},
756         {"in_pkts_invalid", offsetof(struct ixgbe_macsec_stats,
757                 in_pkts_invalid)},
758         {"in_pkts_notvalid", offsetof(struct ixgbe_macsec_stats,
759                 in_pkts_notvalid)},
760         {"in_pkts_unusedsa", offsetof(struct ixgbe_macsec_stats,
761                 in_pkts_unusedsa)},
762         {"in_pkts_notusingsa", offsetof(struct ixgbe_macsec_stats,
763                 in_pkts_notusingsa)},
764 };
765
766 #define IXGBE_NB_MACSEC_STATS (sizeof(rte_ixgbe_macsec_strings) / \
767                            sizeof(rte_ixgbe_macsec_strings[0]))
768
769 /* Per-queue statistics */
770 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_rxq_strings[] = {
771         {"mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats, rnbc)},
772         {"dropped", offsetof(struct ixgbe_hw_stats, mpc)},
773         {"xon_packets", offsetof(struct ixgbe_hw_stats, pxonrxc)},
774         {"xoff_packets", offsetof(struct ixgbe_hw_stats, pxoffrxc)},
775 };
776
777 #define IXGBE_NB_RXQ_PRIO_STATS (sizeof(rte_ixgbe_rxq_strings) / \
778                            sizeof(rte_ixgbe_rxq_strings[0]))
779 #define IXGBE_NB_RXQ_PRIO_VALUES 8
780
781 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_txq_strings[] = {
782         {"xon_packets", offsetof(struct ixgbe_hw_stats, pxontxc)},
783         {"xoff_packets", offsetof(struct ixgbe_hw_stats, pxofftxc)},
784         {"xon_to_xoff_packets", offsetof(struct ixgbe_hw_stats,
785                 pxon2offc)},
786 };
787
788 #define IXGBE_NB_TXQ_PRIO_STATS (sizeof(rte_ixgbe_txq_strings) / \
789                            sizeof(rte_ixgbe_txq_strings[0]))
790 #define IXGBE_NB_TXQ_PRIO_VALUES 8
791
792 static const struct rte_ixgbe_xstats_name_off rte_ixgbevf_stats_strings[] = {
793         {"rx_multicast_packets", offsetof(struct ixgbevf_hw_stats, vfmprc)},
794 };
795
796 #define IXGBEVF_NB_XSTATS (sizeof(rte_ixgbevf_stats_strings) /  \
797                 sizeof(rte_ixgbevf_stats_strings[0]))
798
799 /*
800  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
801  */
802 static inline int
803 ixgbe_is_sfp(struct ixgbe_hw *hw)
804 {
805         switch (hw->phy.type) {
806         case ixgbe_phy_sfp_avago:
807         case ixgbe_phy_sfp_ftl:
808         case ixgbe_phy_sfp_intel:
809         case ixgbe_phy_sfp_unknown:
810         case ixgbe_phy_sfp_passive_tyco:
811         case ixgbe_phy_sfp_passive_unknown:
812                 return 1;
813         default:
814                 return 0;
815         }
816 }
817
818 static inline int32_t
819 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
820 {
821         uint32_t ctrl_ext;
822         int32_t status;
823
824         status = ixgbe_reset_hw(hw);
825
826         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
827         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
828         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
829         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
830         IXGBE_WRITE_FLUSH(hw);
831
832         if (status == IXGBE_ERR_SFP_NOT_PRESENT)
833                 status = IXGBE_SUCCESS;
834         return status;
835 }
836
837 static inline void
838 ixgbe_enable_intr(struct rte_eth_dev *dev)
839 {
840         struct ixgbe_interrupt *intr =
841                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
842         struct ixgbe_hw *hw =
843                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
844
845         IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
846         IXGBE_WRITE_FLUSH(hw);
847 }
848
849 /*
850  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
851  */
852 static void
853 ixgbe_disable_intr(struct ixgbe_hw *hw)
854 {
855         PMD_INIT_FUNC_TRACE();
856
857         if (hw->mac.type == ixgbe_mac_82598EB) {
858                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
859         } else {
860                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
861                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
862                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
863         }
864         IXGBE_WRITE_FLUSH(hw);
865 }
866
867 /*
868  * This function resets queue statistics mapping registers.
869  * From Niantic datasheet, Initialization of Statistics section:
870  * "...if software requires the queue counters, the RQSMR and TQSM registers
871  * must be re-programmed following a device reset.
872  */
873 static void
874 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
875 {
876         uint32_t i;
877
878         for (i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
879                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
880                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
881         }
882 }
883
884
885 static int
886 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
887                                   uint16_t queue_id,
888                                   uint8_t stat_idx,
889                                   uint8_t is_rx)
890 {
891 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
892 #define NB_QMAP_FIELDS_PER_QSM_REG 4
893 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
894
895         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
896         struct ixgbe_stat_mapping_registers *stat_mappings =
897                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
898         uint32_t qsmr_mask = 0;
899         uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
900         uint32_t q_map;
901         uint8_t n, offset;
902
903         if ((hw->mac.type != ixgbe_mac_82599EB) &&
904                 (hw->mac.type != ixgbe_mac_X540) &&
905                 (hw->mac.type != ixgbe_mac_X550) &&
906                 (hw->mac.type != ixgbe_mac_X550EM_x) &&
907                 (hw->mac.type != ixgbe_mac_X550EM_a))
908                 return -ENOSYS;
909
910         PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
911                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
912                      queue_id, stat_idx);
913
914         n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
915         if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
916                 PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
917                 return -EIO;
918         }
919         offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
920
921         /* Now clear any previous stat_idx set */
922         clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
923         if (!is_rx)
924                 stat_mappings->tqsm[n] &= ~clearing_mask;
925         else
926                 stat_mappings->rqsmr[n] &= ~clearing_mask;
927
928         q_map = (uint32_t)stat_idx;
929         q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
930         qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
931         if (!is_rx)
932                 stat_mappings->tqsm[n] |= qsmr_mask;
933         else
934                 stat_mappings->rqsmr[n] |= qsmr_mask;
935
936         PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
937                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
938                      queue_id, stat_idx);
939         PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
940                      is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
941
942         /* Now write the mapping in the appropriate register */
943         if (is_rx) {
944                 PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
945                              stat_mappings->rqsmr[n], n);
946                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
947         } else {
948                 PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
949                              stat_mappings->tqsm[n], n);
950                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
951         }
952         return 0;
953 }
954
955 static void
956 ixgbe_restore_statistics_mapping(struct rte_eth_dev *dev)
957 {
958         struct ixgbe_stat_mapping_registers *stat_mappings =
959                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
960         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
961         int i;
962
963         /* write whatever was in stat mapping table to the NIC */
964         for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
965                 /* rx */
966                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
967
968                 /* tx */
969                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
970         }
971 }
972
973 static void
974 ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config)
975 {
976         uint8_t i;
977         struct ixgbe_dcb_tc_config *tc;
978         uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
979
980         dcb_config->num_tcs.pg_tcs = dcb_max_tc;
981         dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
982         for (i = 0; i < dcb_max_tc; i++) {
983                 tc = &dcb_config->tc_config[i];
984                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
985                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
986                                  (uint8_t)(100/dcb_max_tc + (i & 1));
987                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
988                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
989                                  (uint8_t)(100/dcb_max_tc + (i & 1));
990                 tc->pfc = ixgbe_dcb_pfc_disabled;
991         }
992
993         /* Initialize default user to priority mapping, UPx->TC0 */
994         tc = &dcb_config->tc_config[0];
995         tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
996         tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
997         for (i = 0; i < IXGBE_DCB_MAX_BW_GROUP; i++) {
998                 dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
999                 dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
1000         }
1001         dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
1002         dcb_config->pfc_mode_enable = false;
1003         dcb_config->vt_mode = true;
1004         dcb_config->round_robin_enable = false;
1005         /* support all DCB capabilities in 82599 */
1006         dcb_config->support.capabilities = 0xFF;
1007
1008         /*we only support 4 Tcs for X540, X550 */
1009         if (hw->mac.type == ixgbe_mac_X540 ||
1010                 hw->mac.type == ixgbe_mac_X550 ||
1011                 hw->mac.type == ixgbe_mac_X550EM_x ||
1012                 hw->mac.type == ixgbe_mac_X550EM_a) {
1013                 dcb_config->num_tcs.pg_tcs = 4;
1014                 dcb_config->num_tcs.pfc_tcs = 4;
1015         }
1016 }
1017
1018 /*
1019  * Ensure that all locks are released before first NVM or PHY access
1020  */
1021 static void
1022 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
1023 {
1024         uint16_t mask;
1025
1026         /*
1027          * Phy lock should not fail in this early stage. If this is the case,
1028          * it is due to an improper exit of the application.
1029          * So force the release of the faulty lock. Release of common lock
1030          * is done automatically by swfw_sync function.
1031          */
1032         mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
1033         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1034                 PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
1035         }
1036         ixgbe_release_swfw_semaphore(hw, mask);
1037
1038         /*
1039          * These ones are more tricky since they are common to all ports; but
1040          * swfw_sync retries last long enough (1s) to be almost sure that if
1041          * lock can not be taken it is due to an improper lock of the
1042          * semaphore.
1043          */
1044         mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
1045         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1046                 PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1047         }
1048         ixgbe_release_swfw_semaphore(hw, mask);
1049 }
1050
1051 /*
1052  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
1053  * It returns 0 on success.
1054  */
1055 static int
1056 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params __rte_unused)
1057 {
1058         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1059         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1060         struct ixgbe_hw *hw =
1061                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1062         struct ixgbe_vfta *shadow_vfta =
1063                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1064         struct ixgbe_hwstrip *hwstrip =
1065                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1066         struct ixgbe_dcb_config *dcb_config =
1067                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
1068         struct ixgbe_filter_info *filter_info =
1069                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1070         struct ixgbe_bw_conf *bw_conf =
1071                 IXGBE_DEV_PRIVATE_TO_BW_CONF(eth_dev->data->dev_private);
1072         uint32_t ctrl_ext;
1073         uint16_t csum;
1074         int diag, i;
1075
1076         PMD_INIT_FUNC_TRACE();
1077
1078         eth_dev->dev_ops = &ixgbe_eth_dev_ops;
1079         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1080         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1081         eth_dev->tx_pkt_prepare = &ixgbe_prep_pkts;
1082
1083         /*
1084          * For secondary processes, we don't initialise any further as primary
1085          * has already done this work. Only check we don't need a different
1086          * RX and TX function.
1087          */
1088         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1089                 struct ixgbe_tx_queue *txq;
1090                 /* TX queue function in primary, set by last queue initialized
1091                  * Tx queue may not initialized by primary process
1092                  */
1093                 if (eth_dev->data->tx_queues) {
1094                         txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
1095                         ixgbe_set_tx_function(eth_dev, txq);
1096                 } else {
1097                         /* Use default TX function if we get here */
1098                         PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
1099                                      "Using default TX function.");
1100                 }
1101
1102                 ixgbe_set_rx_function(eth_dev);
1103
1104                 return 0;
1105         }
1106
1107         rte_eth_copy_pci_info(eth_dev, pci_dev);
1108
1109         /* Vendor and Device ID need to be set before init of shared code */
1110         hw->device_id = pci_dev->id.device_id;
1111         hw->vendor_id = pci_dev->id.vendor_id;
1112         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1113         hw->allow_unsupported_sfp = 1;
1114
1115         /* Initialize the shared code (base driver) */
1116 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1117         diag = ixgbe_bypass_init_shared_code(hw);
1118 #else
1119         diag = ixgbe_init_shared_code(hw);
1120 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1121
1122         if (diag != IXGBE_SUCCESS) {
1123                 PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
1124                 return -EIO;
1125         }
1126
1127         if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
1128                 PMD_INIT_LOG(ERR, "\nERROR: "
1129                         "Firmware recovery mode detected. Limiting functionality.\n"
1130                         "Refer to the Intel(R) Ethernet Adapters and Devices "
1131                         "User Guide for details on firmware recovery mode.");
1132                 return -EIO;
1133         }
1134
1135         /* pick up the PCI bus settings for reporting later */
1136         ixgbe_get_bus_info(hw);
1137
1138         /* Unlock any pending hardware semaphore */
1139         ixgbe_swfw_lock_reset(hw);
1140
1141 #ifdef RTE_LIBRTE_SECURITY
1142         /* Initialize security_ctx only for primary process*/
1143         if (ixgbe_ipsec_ctx_create(eth_dev))
1144                 return -ENOMEM;
1145 #endif
1146
1147         /* Initialize DCB configuration*/
1148         memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
1149         ixgbe_dcb_init(hw, dcb_config);
1150         /* Get Hardware Flow Control setting */
1151         hw->fc.requested_mode = ixgbe_fc_full;
1152         hw->fc.current_mode = ixgbe_fc_full;
1153         hw->fc.pause_time = IXGBE_FC_PAUSE;
1154         for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
1155                 hw->fc.low_water[i] = IXGBE_FC_LO;
1156                 hw->fc.high_water[i] = IXGBE_FC_HI;
1157         }
1158         hw->fc.send_xon = 1;
1159
1160         /* Make sure we have a good EEPROM before we read from it */
1161         diag = ixgbe_validate_eeprom_checksum(hw, &csum);
1162         if (diag != IXGBE_SUCCESS) {
1163                 PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
1164                 return -EIO;
1165         }
1166
1167 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1168         diag = ixgbe_bypass_init_hw(hw);
1169 #else
1170         diag = ixgbe_init_hw(hw);
1171 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1172
1173         /*
1174          * Devices with copper phys will fail to initialise if ixgbe_init_hw()
1175          * is called too soon after the kernel driver unbinding/binding occurs.
1176          * The failure occurs in ixgbe_identify_phy_generic() for all devices,
1177          * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
1178          * also called. See ixgbe_identify_phy_82599(). The reason for the
1179          * failure is not known, and only occuts when virtualisation features
1180          * are disabled in the bios. A delay of 100ms  was found to be enough by
1181          * trial-and-error, and is doubled to be safe.
1182          */
1183         if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
1184                 rte_delay_ms(200);
1185                 diag = ixgbe_init_hw(hw);
1186         }
1187
1188         if (diag == IXGBE_ERR_SFP_NOT_PRESENT)
1189                 diag = IXGBE_SUCCESS;
1190
1191         if (diag == IXGBE_ERR_EEPROM_VERSION) {
1192                 PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
1193                              "LOM.  Please be aware there may be issues associated "
1194                              "with your hardware.");
1195                 PMD_INIT_LOG(ERR, "If you are experiencing problems "
1196                              "please contact your Intel or hardware representative "
1197                              "who provided you with this hardware.");
1198         } else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
1199                 PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
1200         if (diag) {
1201                 PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
1202                 return -EIO;
1203         }
1204
1205         /* Reset the hw statistics */
1206         ixgbe_dev_stats_reset(eth_dev);
1207
1208         /* disable interrupt */
1209         ixgbe_disable_intr(hw);
1210
1211         /* reset mappings for queue statistics hw counters*/
1212         ixgbe_reset_qstat_mappings(hw);
1213
1214         /* Allocate memory for storing MAC addresses */
1215         eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
1216                                                hw->mac.num_rar_entries, 0);
1217         if (eth_dev->data->mac_addrs == NULL) {
1218                 PMD_INIT_LOG(ERR,
1219                              "Failed to allocate %u bytes needed to store "
1220                              "MAC addresses",
1221                              ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1222                 return -ENOMEM;
1223         }
1224         /* Copy the permanent MAC address */
1225         ether_addr_copy((struct ether_addr *) hw->mac.perm_addr,
1226                         &eth_dev->data->mac_addrs[0]);
1227
1228         /* Allocate memory for storing hash filter MAC addresses */
1229         eth_dev->data->hash_mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
1230                                                     IXGBE_VMDQ_NUM_UC_MAC, 0);
1231         if (eth_dev->data->hash_mac_addrs == NULL) {
1232                 PMD_INIT_LOG(ERR,
1233                              "Failed to allocate %d bytes needed to store MAC addresses",
1234                              ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
1235                 return -ENOMEM;
1236         }
1237
1238         /* initialize the vfta */
1239         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1240
1241         /* initialize the hw strip bitmap*/
1242         memset(hwstrip, 0, sizeof(*hwstrip));
1243
1244         /* initialize PF if max_vfs not zero */
1245         ixgbe_pf_host_init(eth_dev);
1246
1247         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1248         /* let hardware know driver is loaded */
1249         ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1250         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
1251         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1252         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1253         IXGBE_WRITE_FLUSH(hw);
1254
1255         if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1256                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1257                              (int) hw->mac.type, (int) hw->phy.type,
1258                              (int) hw->phy.sfp_type);
1259         else
1260                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1261                              (int) hw->mac.type, (int) hw->phy.type);
1262
1263         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1264                      eth_dev->data->port_id, pci_dev->id.vendor_id,
1265                      pci_dev->id.device_id);
1266
1267         rte_intr_callback_register(intr_handle,
1268                                    ixgbe_dev_interrupt_handler, eth_dev);
1269
1270         /* enable uio/vfio intr/eventfd mapping */
1271         rte_intr_enable(intr_handle);
1272
1273         /* enable support intr */
1274         ixgbe_enable_intr(eth_dev);
1275
1276         /* initialize filter info */
1277         memset(filter_info, 0,
1278                sizeof(struct ixgbe_filter_info));
1279
1280         /* initialize 5tuple filter list */
1281         TAILQ_INIT(&filter_info->fivetuple_list);
1282
1283         /* initialize flow director filter list & hash */
1284         ixgbe_fdir_filter_init(eth_dev);
1285
1286         /* initialize l2 tunnel filter list & hash */
1287         ixgbe_l2_tn_filter_init(eth_dev);
1288
1289         /* initialize flow filter lists */
1290         ixgbe_filterlist_init();
1291
1292         /* initialize bandwidth configuration info */
1293         memset(bw_conf, 0, sizeof(struct ixgbe_bw_conf));
1294
1295         /* initialize Traffic Manager configuration */
1296         ixgbe_tm_conf_init(eth_dev);
1297
1298         return 0;
1299 }
1300
1301 static int
1302 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1303 {
1304         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1305         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1306         struct ixgbe_hw *hw;
1307         int retries = 0;
1308         int ret;
1309
1310         PMD_INIT_FUNC_TRACE();
1311
1312         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1313                 return 0;
1314
1315         hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1316
1317         if (hw->adapter_stopped == 0)
1318                 ixgbe_dev_close(eth_dev);
1319
1320         eth_dev->dev_ops = NULL;
1321         eth_dev->rx_pkt_burst = NULL;
1322         eth_dev->tx_pkt_burst = NULL;
1323
1324         /* Unlock any pending hardware semaphore */
1325         ixgbe_swfw_lock_reset(hw);
1326
1327         /* disable uio intr before callback unregister */
1328         rte_intr_disable(intr_handle);
1329
1330         do {
1331                 ret = rte_intr_callback_unregister(intr_handle,
1332                                 ixgbe_dev_interrupt_handler, eth_dev);
1333                 if (ret >= 0) {
1334                         break;
1335                 } else if (ret != -EAGAIN) {
1336                         PMD_INIT_LOG(ERR,
1337                                 "intr callback unregister failed: %d",
1338                                 ret);
1339                         return ret;
1340                 }
1341                 rte_delay_ms(100);
1342         } while (retries++ < (10 + IXGBE_LINK_UP_TIME));
1343
1344         /* cancel the delay handler before remove dev */
1345         rte_eal_alarm_cancel(ixgbe_dev_interrupt_delayed_handler, eth_dev);
1346
1347         /* uninitialize PF if max_vfs not zero */
1348         ixgbe_pf_host_uninit(eth_dev);
1349
1350         /* remove all the fdir filters & hash */
1351         ixgbe_fdir_filter_uninit(eth_dev);
1352
1353         /* remove all the L2 tunnel filters & hash */
1354         ixgbe_l2_tn_filter_uninit(eth_dev);
1355
1356         /* Remove all ntuple filters of the device */
1357         ixgbe_ntuple_filter_uninit(eth_dev);
1358
1359         /* clear all the filters list */
1360         ixgbe_filterlist_flush();
1361
1362         /* Remove all Traffic Manager configuration */
1363         ixgbe_tm_conf_uninit(eth_dev);
1364
1365 #ifdef RTE_LIBRTE_SECURITY
1366         rte_free(eth_dev->security_ctx);
1367 #endif
1368
1369         return 0;
1370 }
1371
1372 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
1373 {
1374         struct ixgbe_filter_info *filter_info =
1375                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1376         struct ixgbe_5tuple_filter *p_5tuple;
1377
1378         while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
1379                 TAILQ_REMOVE(&filter_info->fivetuple_list,
1380                              p_5tuple,
1381                              entries);
1382                 rte_free(p_5tuple);
1383         }
1384         memset(filter_info->fivetuple_mask, 0,
1385                sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1386
1387         return 0;
1388 }
1389
1390 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev)
1391 {
1392         struct ixgbe_hw_fdir_info *fdir_info =
1393                 IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1394         struct ixgbe_fdir_filter *fdir_filter;
1395
1396                 if (fdir_info->hash_map)
1397                 rte_free(fdir_info->hash_map);
1398         if (fdir_info->hash_handle)
1399                 rte_hash_free(fdir_info->hash_handle);
1400
1401         while ((fdir_filter = TAILQ_FIRST(&fdir_info->fdir_list))) {
1402                 TAILQ_REMOVE(&fdir_info->fdir_list,
1403                              fdir_filter,
1404                              entries);
1405                 rte_free(fdir_filter);
1406         }
1407
1408         return 0;
1409 }
1410
1411 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev)
1412 {
1413         struct ixgbe_l2_tn_info *l2_tn_info =
1414                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1415         struct ixgbe_l2_tn_filter *l2_tn_filter;
1416
1417         if (l2_tn_info->hash_map)
1418                 rte_free(l2_tn_info->hash_map);
1419         if (l2_tn_info->hash_handle)
1420                 rte_hash_free(l2_tn_info->hash_handle);
1421
1422         while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
1423                 TAILQ_REMOVE(&l2_tn_info->l2_tn_list,
1424                              l2_tn_filter,
1425                              entries);
1426                 rte_free(l2_tn_filter);
1427         }
1428
1429         return 0;
1430 }
1431
1432 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev)
1433 {
1434         struct ixgbe_hw_fdir_info *fdir_info =
1435                 IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1436         char fdir_hash_name[RTE_HASH_NAMESIZE];
1437         struct rte_hash_parameters fdir_hash_params = {
1438                 .name = fdir_hash_name,
1439                 .entries = IXGBE_MAX_FDIR_FILTER_NUM,
1440                 .key_len = sizeof(union ixgbe_atr_input),
1441                 .hash_func = rte_hash_crc,
1442                 .hash_func_init_val = 0,
1443                 .socket_id = rte_socket_id(),
1444         };
1445
1446         TAILQ_INIT(&fdir_info->fdir_list);
1447         snprintf(fdir_hash_name, RTE_HASH_NAMESIZE,
1448                  "fdir_%s", eth_dev->device->name);
1449         fdir_info->hash_handle = rte_hash_create(&fdir_hash_params);
1450         if (!fdir_info->hash_handle) {
1451                 PMD_INIT_LOG(ERR, "Failed to create fdir hash table!");
1452                 return -EINVAL;
1453         }
1454         fdir_info->hash_map = rte_zmalloc("ixgbe",
1455                                           sizeof(struct ixgbe_fdir_filter *) *
1456                                           IXGBE_MAX_FDIR_FILTER_NUM,
1457                                           0);
1458         if (!fdir_info->hash_map) {
1459                 PMD_INIT_LOG(ERR,
1460                              "Failed to allocate memory for fdir hash map!");
1461                 return -ENOMEM;
1462         }
1463         fdir_info->mask_added = FALSE;
1464
1465         return 0;
1466 }
1467
1468 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev)
1469 {
1470         struct ixgbe_l2_tn_info *l2_tn_info =
1471                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1472         char l2_tn_hash_name[RTE_HASH_NAMESIZE];
1473         struct rte_hash_parameters l2_tn_hash_params = {
1474                 .name = l2_tn_hash_name,
1475                 .entries = IXGBE_MAX_L2_TN_FILTER_NUM,
1476                 .key_len = sizeof(struct ixgbe_l2_tn_key),
1477                 .hash_func = rte_hash_crc,
1478                 .hash_func_init_val = 0,
1479                 .socket_id = rte_socket_id(),
1480         };
1481
1482         TAILQ_INIT(&l2_tn_info->l2_tn_list);
1483         snprintf(l2_tn_hash_name, RTE_HASH_NAMESIZE,
1484                  "l2_tn_%s", eth_dev->device->name);
1485         l2_tn_info->hash_handle = rte_hash_create(&l2_tn_hash_params);
1486         if (!l2_tn_info->hash_handle) {
1487                 PMD_INIT_LOG(ERR, "Failed to create L2 TN hash table!");
1488                 return -EINVAL;
1489         }
1490         l2_tn_info->hash_map = rte_zmalloc("ixgbe",
1491                                    sizeof(struct ixgbe_l2_tn_filter *) *
1492                                    IXGBE_MAX_L2_TN_FILTER_NUM,
1493                                    0);
1494         if (!l2_tn_info->hash_map) {
1495                 PMD_INIT_LOG(ERR,
1496                         "Failed to allocate memory for L2 TN hash map!");
1497                 return -ENOMEM;
1498         }
1499         l2_tn_info->e_tag_en = FALSE;
1500         l2_tn_info->e_tag_fwd_en = FALSE;
1501         l2_tn_info->e_tag_ether_type = ETHER_TYPE_ETAG;
1502
1503         return 0;
1504 }
1505 /*
1506  * Negotiate mailbox API version with the PF.
1507  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1508  * Then we try to negotiate starting with the most recent one.
1509  * If all negotiation attempts fail, then we will proceed with
1510  * the default one (ixgbe_mbox_api_10).
1511  */
1512 static void
1513 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1514 {
1515         int32_t i;
1516
1517         /* start with highest supported, proceed down */
1518         static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1519                 ixgbe_mbox_api_13,
1520                 ixgbe_mbox_api_12,
1521                 ixgbe_mbox_api_11,
1522                 ixgbe_mbox_api_10,
1523         };
1524
1525         for (i = 0;
1526                         i != RTE_DIM(sup_ver) &&
1527                         ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1528                         i++)
1529                 ;
1530 }
1531
1532 static void
1533 generate_random_mac_addr(struct ether_addr *mac_addr)
1534 {
1535         uint64_t random;
1536
1537         /* Set Organizationally Unique Identifier (OUI) prefix. */
1538         mac_addr->addr_bytes[0] = 0x00;
1539         mac_addr->addr_bytes[1] = 0x09;
1540         mac_addr->addr_bytes[2] = 0xC0;
1541         /* Force indication of locally assigned MAC address. */
1542         mac_addr->addr_bytes[0] |= ETHER_LOCAL_ADMIN_ADDR;
1543         /* Generate the last 3 bytes of the MAC address with a random number. */
1544         random = rte_rand();
1545         memcpy(&mac_addr->addr_bytes[3], &random, 3);
1546 }
1547
1548 /*
1549  * Virtual Function device init
1550  */
1551 static int
1552 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1553 {
1554         int diag;
1555         uint32_t tc, tcs;
1556         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1557         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1558         struct ixgbe_hw *hw =
1559                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1560         struct ixgbe_vfta *shadow_vfta =
1561                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1562         struct ixgbe_hwstrip *hwstrip =
1563                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1564         struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
1565
1566         PMD_INIT_FUNC_TRACE();
1567
1568         eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1569         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1570         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1571
1572         /* for secondary processes, we don't initialise any further as primary
1573          * has already done this work. Only check we don't need a different
1574          * RX function
1575          */
1576         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1577                 struct ixgbe_tx_queue *txq;
1578                 /* TX queue function in primary, set by last queue initialized
1579                  * Tx queue may not initialized by primary process
1580                  */
1581                 if (eth_dev->data->tx_queues) {
1582                         txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1583                         ixgbe_set_tx_function(eth_dev, txq);
1584                 } else {
1585                         /* Use default TX function if we get here */
1586                         PMD_INIT_LOG(NOTICE,
1587                                      "No TX queues configured yet. Using default TX function.");
1588                 }
1589
1590                 ixgbe_set_rx_function(eth_dev);
1591
1592                 return 0;
1593         }
1594
1595         rte_eth_copy_pci_info(eth_dev, pci_dev);
1596
1597         hw->device_id = pci_dev->id.device_id;
1598         hw->vendor_id = pci_dev->id.vendor_id;
1599         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1600
1601         /* initialize the vfta */
1602         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1603
1604         /* initialize the hw strip bitmap*/
1605         memset(hwstrip, 0, sizeof(*hwstrip));
1606
1607         /* Initialize the shared code (base driver) */
1608         diag = ixgbe_init_shared_code(hw);
1609         if (diag != IXGBE_SUCCESS) {
1610                 PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1611                 return -EIO;
1612         }
1613
1614         /* init_mailbox_params */
1615         hw->mbx.ops.init_params(hw);
1616
1617         /* Reset the hw statistics */
1618         ixgbevf_dev_stats_reset(eth_dev);
1619
1620         /* Disable the interrupts for VF */
1621         ixgbevf_intr_disable(eth_dev);
1622
1623         hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1624         diag = hw->mac.ops.reset_hw(hw);
1625
1626         /*
1627          * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1628          * the underlying PF driver has not assigned a MAC address to the VF.
1629          * In this case, assign a random MAC address.
1630          */
1631         if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1632                 PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1633                 /*
1634                  * This error code will be propagated to the app by
1635                  * rte_eth_dev_reset, so use a public error code rather than
1636                  * the internal-only IXGBE_ERR_RESET_FAILED
1637                  */
1638                 return -EAGAIN;
1639         }
1640
1641         /* negotiate mailbox API version to use with the PF. */
1642         ixgbevf_negotiate_api(hw);
1643
1644         /* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1645         ixgbevf_get_queues(hw, &tcs, &tc);
1646
1647         /* Allocate memory for storing MAC addresses */
1648         eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", ETHER_ADDR_LEN *
1649                                                hw->mac.num_rar_entries, 0);
1650         if (eth_dev->data->mac_addrs == NULL) {
1651                 PMD_INIT_LOG(ERR,
1652                              "Failed to allocate %u bytes needed to store "
1653                              "MAC addresses",
1654                              ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1655                 return -ENOMEM;
1656         }
1657
1658         /* Generate a random MAC address, if none was assigned by PF. */
1659         if (is_zero_ether_addr(perm_addr)) {
1660                 generate_random_mac_addr(perm_addr);
1661                 diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1662                 if (diag) {
1663                         rte_free(eth_dev->data->mac_addrs);
1664                         eth_dev->data->mac_addrs = NULL;
1665                         return diag;
1666                 }
1667                 PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1668                 PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1669                              "%02x:%02x:%02x:%02x:%02x:%02x",
1670                              perm_addr->addr_bytes[0],
1671                              perm_addr->addr_bytes[1],
1672                              perm_addr->addr_bytes[2],
1673                              perm_addr->addr_bytes[3],
1674                              perm_addr->addr_bytes[4],
1675                              perm_addr->addr_bytes[5]);
1676         }
1677
1678         /* Copy the permanent MAC address */
1679         ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1680
1681         /* reset the hardware with the new settings */
1682         diag = hw->mac.ops.start_hw(hw);
1683         switch (diag) {
1684         case  0:
1685                 break;
1686
1687         default:
1688                 PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1689                 return -EIO;
1690         }
1691
1692         rte_intr_callback_register(intr_handle,
1693                                    ixgbevf_dev_interrupt_handler, eth_dev);
1694         rte_intr_enable(intr_handle);
1695         ixgbevf_intr_enable(eth_dev);
1696
1697         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1698                      eth_dev->data->port_id, pci_dev->id.vendor_id,
1699                      pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1700
1701         return 0;
1702 }
1703
1704 /* Virtual Function device uninit */
1705
1706 static int
1707 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1708 {
1709         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1710         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1711         struct ixgbe_hw *hw;
1712
1713         PMD_INIT_FUNC_TRACE();
1714
1715         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1716                 return 0;
1717
1718         hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1719
1720         if (hw->adapter_stopped == 0)
1721                 ixgbevf_dev_close(eth_dev);
1722
1723         eth_dev->dev_ops = NULL;
1724         eth_dev->rx_pkt_burst = NULL;
1725         eth_dev->tx_pkt_burst = NULL;
1726
1727         /* Disable the interrupts for VF */
1728         ixgbevf_intr_disable(eth_dev);
1729
1730         rte_intr_disable(intr_handle);
1731         rte_intr_callback_unregister(intr_handle,
1732                                      ixgbevf_dev_interrupt_handler, eth_dev);
1733
1734         return 0;
1735 }
1736
1737 static int
1738 eth_ixgbe_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1739                 struct rte_pci_device *pci_dev)
1740 {
1741         char name[RTE_ETH_NAME_MAX_LEN];
1742         struct rte_eth_dev *pf_ethdev;
1743         struct rte_eth_devargs eth_da;
1744         int i, retval;
1745
1746         if (pci_dev->device.devargs) {
1747                 retval = rte_eth_devargs_parse(pci_dev->device.devargs->args,
1748                                 &eth_da);
1749                 if (retval)
1750                         return retval;
1751         } else
1752                 memset(&eth_da, 0, sizeof(eth_da));
1753
1754         retval = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
1755                 sizeof(struct ixgbe_adapter),
1756                 eth_dev_pci_specific_init, pci_dev,
1757                 eth_ixgbe_dev_init, NULL);
1758
1759         if (retval || eth_da.nb_representor_ports < 1)
1760                 return retval;
1761
1762         pf_ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1763         if (pf_ethdev == NULL)
1764                 return -ENODEV;
1765
1766         /* probe VF representor ports */
1767         for (i = 0; i < eth_da.nb_representor_ports; i++) {
1768                 struct ixgbe_vf_info *vfinfo;
1769                 struct ixgbe_vf_representor representor;
1770
1771                 vfinfo = *IXGBE_DEV_PRIVATE_TO_P_VFDATA(
1772                         pf_ethdev->data->dev_private);
1773                 if (vfinfo == NULL) {
1774                         PMD_DRV_LOG(ERR,
1775                                 "no virtual functions supported by PF");
1776                         break;
1777                 }
1778
1779                 representor.vf_id = eth_da.representor_ports[i];
1780                 representor.switch_domain_id = vfinfo->switch_domain_id;
1781                 representor.pf_ethdev = pf_ethdev;
1782
1783                 /* representor port net_bdf_port */
1784                 snprintf(name, sizeof(name), "net_%s_representor_%d",
1785                         pci_dev->device.name,
1786                         eth_da.representor_ports[i]);
1787
1788                 retval = rte_eth_dev_create(&pci_dev->device, name,
1789                         sizeof(struct ixgbe_vf_representor), NULL, NULL,
1790                         ixgbe_vf_representor_init, &representor);
1791
1792                 if (retval)
1793                         PMD_DRV_LOG(ERR, "failed to create ixgbe vf "
1794                                 "representor %s.", name);
1795         }
1796
1797         return 0;
1798 }
1799
1800 static int eth_ixgbe_pci_remove(struct rte_pci_device *pci_dev)
1801 {
1802         struct rte_eth_dev *ethdev;
1803
1804         ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1805         if (!ethdev)
1806                 return -ENODEV;
1807
1808         if (ethdev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
1809                 return rte_eth_dev_destroy(ethdev, ixgbe_vf_representor_uninit);
1810         else
1811                 return rte_eth_dev_destroy(ethdev, eth_ixgbe_dev_uninit);
1812 }
1813
1814 static struct rte_pci_driver rte_ixgbe_pmd = {
1815         .id_table = pci_id_ixgbe_map,
1816         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
1817                      RTE_PCI_DRV_IOVA_AS_VA,
1818         .probe = eth_ixgbe_pci_probe,
1819         .remove = eth_ixgbe_pci_remove,
1820 };
1821
1822 static int eth_ixgbevf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1823         struct rte_pci_device *pci_dev)
1824 {
1825         return rte_eth_dev_pci_generic_probe(pci_dev,
1826                 sizeof(struct ixgbe_adapter), eth_ixgbevf_dev_init);
1827 }
1828
1829 static int eth_ixgbevf_pci_remove(struct rte_pci_device *pci_dev)
1830 {
1831         return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbevf_dev_uninit);
1832 }
1833
1834 /*
1835  * virtual function driver struct
1836  */
1837 static struct rte_pci_driver rte_ixgbevf_pmd = {
1838         .id_table = pci_id_ixgbevf_map,
1839         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_IOVA_AS_VA,
1840         .probe = eth_ixgbevf_pci_probe,
1841         .remove = eth_ixgbevf_pci_remove,
1842 };
1843
1844 static int
1845 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1846 {
1847         struct ixgbe_hw *hw =
1848                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1849         struct ixgbe_vfta *shadow_vfta =
1850                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1851         uint32_t vfta;
1852         uint32_t vid_idx;
1853         uint32_t vid_bit;
1854
1855         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1856         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1857         vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1858         if (on)
1859                 vfta |= vid_bit;
1860         else
1861                 vfta &= ~vid_bit;
1862         IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1863
1864         /* update local VFTA copy */
1865         shadow_vfta->vfta[vid_idx] = vfta;
1866
1867         return 0;
1868 }
1869
1870 static void
1871 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1872 {
1873         if (on)
1874                 ixgbe_vlan_hw_strip_enable(dev, queue);
1875         else
1876                 ixgbe_vlan_hw_strip_disable(dev, queue);
1877 }
1878
1879 static int
1880 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1881                     enum rte_vlan_type vlan_type,
1882                     uint16_t tpid)
1883 {
1884         struct ixgbe_hw *hw =
1885                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1886         int ret = 0;
1887         uint32_t reg;
1888         uint32_t qinq;
1889
1890         qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1891         qinq &= IXGBE_DMATXCTL_GDV;
1892
1893         switch (vlan_type) {
1894         case ETH_VLAN_TYPE_INNER:
1895                 if (qinq) {
1896                         reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1897                         reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1898                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1899                         reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1900                         reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1901                                 | ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1902                         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1903                 } else {
1904                         ret = -ENOTSUP;
1905                         PMD_DRV_LOG(ERR, "Inner type is not supported"
1906                                     " by single VLAN");
1907                 }
1908                 break;
1909         case ETH_VLAN_TYPE_OUTER:
1910                 if (qinq) {
1911                         /* Only the high 16-bits is valid */
1912                         IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1913                                         IXGBE_EXVET_VET_EXT_SHIFT);
1914                 } else {
1915                         reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1916                         reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1917                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1918                         reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1919                         reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1920                                 | ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1921                         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1922                 }
1923
1924                 break;
1925         default:
1926                 ret = -EINVAL;
1927                 PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1928                 break;
1929         }
1930
1931         return ret;
1932 }
1933
1934 void
1935 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1936 {
1937         struct ixgbe_hw *hw =
1938                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1939         uint32_t vlnctrl;
1940
1941         PMD_INIT_FUNC_TRACE();
1942
1943         /* Filter Table Disable */
1944         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1945         vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1946
1947         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1948 }
1949
1950 void
1951 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1952 {
1953         struct ixgbe_hw *hw =
1954                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1955         struct ixgbe_vfta *shadow_vfta =
1956                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1957         uint32_t vlnctrl;
1958         uint16_t i;
1959
1960         PMD_INIT_FUNC_TRACE();
1961
1962         /* Filter Table Enable */
1963         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1964         vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1965         vlnctrl |= IXGBE_VLNCTRL_VFE;
1966
1967         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1968
1969         /* write whatever is in local vfta copy */
1970         for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1971                 IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1972 }
1973
1974 static void
1975 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1976 {
1977         struct ixgbe_hwstrip *hwstrip =
1978                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1979         struct ixgbe_rx_queue *rxq;
1980
1981         if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
1982                 return;
1983
1984         if (on)
1985                 IXGBE_SET_HWSTRIP(hwstrip, queue);
1986         else
1987                 IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1988
1989         if (queue >= dev->data->nb_rx_queues)
1990                 return;
1991
1992         rxq = dev->data->rx_queues[queue];
1993
1994         if (on) {
1995                 rxq->vlan_flags = PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
1996                 rxq->offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
1997         } else {
1998                 rxq->vlan_flags = PKT_RX_VLAN;
1999                 rxq->offloads &= ~DEV_RX_OFFLOAD_VLAN_STRIP;
2000         }
2001 }
2002
2003 static void
2004 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
2005 {
2006         struct ixgbe_hw *hw =
2007                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2008         uint32_t ctrl;
2009
2010         PMD_INIT_FUNC_TRACE();
2011
2012         if (hw->mac.type == ixgbe_mac_82598EB) {
2013                 /* No queue level support */
2014                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2015                 return;
2016         }
2017
2018         /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2019         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2020         ctrl &= ~IXGBE_RXDCTL_VME;
2021         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2022
2023         /* record those setting for HW strip per queue */
2024         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
2025 }
2026
2027 static void
2028 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
2029 {
2030         struct ixgbe_hw *hw =
2031                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2032         uint32_t ctrl;
2033
2034         PMD_INIT_FUNC_TRACE();
2035
2036         if (hw->mac.type == ixgbe_mac_82598EB) {
2037                 /* No queue level supported */
2038                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2039                 return;
2040         }
2041
2042         /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2043         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2044         ctrl |= IXGBE_RXDCTL_VME;
2045         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2046
2047         /* record those setting for HW strip per queue */
2048         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
2049 }
2050
2051 static void
2052 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2053 {
2054         struct ixgbe_hw *hw =
2055                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2056         uint32_t ctrl;
2057
2058         PMD_INIT_FUNC_TRACE();
2059
2060         /* DMATXCTRL: Geric Double VLAN Disable */
2061         ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2062         ctrl &= ~IXGBE_DMATXCTL_GDV;
2063         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2064
2065         /* CTRL_EXT: Global Double VLAN Disable */
2066         ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2067         ctrl &= ~IXGBE_EXTENDED_VLAN;
2068         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2069
2070 }
2071
2072 static void
2073 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2074 {
2075         struct ixgbe_hw *hw =
2076                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2077         uint32_t ctrl;
2078
2079         PMD_INIT_FUNC_TRACE();
2080
2081         /* DMATXCTRL: Geric Double VLAN Enable */
2082         ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2083         ctrl |= IXGBE_DMATXCTL_GDV;
2084         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2085
2086         /* CTRL_EXT: Global Double VLAN Enable */
2087         ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2088         ctrl |= IXGBE_EXTENDED_VLAN;
2089         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2090
2091         /* Clear pooling mode of PFVTCTL. It's required by X550. */
2092         if (hw->mac.type == ixgbe_mac_X550 ||
2093             hw->mac.type == ixgbe_mac_X550EM_x ||
2094             hw->mac.type == ixgbe_mac_X550EM_a) {
2095                 ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
2096                 ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
2097                 IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
2098         }
2099
2100         /*
2101          * VET EXT field in the EXVET register = 0x8100 by default
2102          * So no need to change. Same to VT field of DMATXCTL register
2103          */
2104 }
2105
2106 void
2107 ixgbe_vlan_hw_strip_config(struct rte_eth_dev *dev)
2108 {
2109         struct ixgbe_hw *hw =
2110                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2111         struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
2112         uint32_t ctrl;
2113         uint16_t i;
2114         struct ixgbe_rx_queue *rxq;
2115         bool on;
2116
2117         PMD_INIT_FUNC_TRACE();
2118
2119         if (hw->mac.type == ixgbe_mac_82598EB) {
2120                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP) {
2121                         ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2122                         ctrl |= IXGBE_VLNCTRL_VME;
2123                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2124                 } else {
2125                         ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2126                         ctrl &= ~IXGBE_VLNCTRL_VME;
2127                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2128                 }
2129         } else {
2130                 /*
2131                  * Other 10G NIC, the VLAN strip can be setup
2132                  * per queue in RXDCTL
2133                  */
2134                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
2135                         rxq = dev->data->rx_queues[i];
2136                         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2137                         if (rxq->offloads & DEV_RX_OFFLOAD_VLAN_STRIP) {
2138                                 ctrl |= IXGBE_RXDCTL_VME;
2139                                 on = TRUE;
2140                         } else {
2141                                 ctrl &= ~IXGBE_RXDCTL_VME;
2142                                 on = FALSE;
2143                         }
2144                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2145
2146                         /* record those setting for HW strip per queue */
2147                         ixgbe_vlan_hw_strip_bitmap_set(dev, i, on);
2148                 }
2149         }
2150 }
2151
2152 static void
2153 ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev, int mask)
2154 {
2155         uint16_t i;
2156         struct rte_eth_rxmode *rxmode;
2157         struct ixgbe_rx_queue *rxq;
2158
2159         if (mask & ETH_VLAN_STRIP_MASK) {
2160                 rxmode = &dev->data->dev_conf.rxmode;
2161                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2162                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
2163                                 rxq = dev->data->rx_queues[i];
2164                                 rxq->offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
2165                         }
2166                 else
2167                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
2168                                 rxq = dev->data->rx_queues[i];
2169                                 rxq->offloads &= ~DEV_RX_OFFLOAD_VLAN_STRIP;
2170                         }
2171         }
2172 }
2173
2174 static int
2175 ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask)
2176 {
2177         struct rte_eth_rxmode *rxmode;
2178         rxmode = &dev->data->dev_conf.rxmode;
2179
2180         if (mask & ETH_VLAN_STRIP_MASK) {
2181                 ixgbe_vlan_hw_strip_config(dev);
2182         }
2183
2184         if (mask & ETH_VLAN_FILTER_MASK) {
2185                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2186                         ixgbe_vlan_hw_filter_enable(dev);
2187                 else
2188                         ixgbe_vlan_hw_filter_disable(dev);
2189         }
2190
2191         if (mask & ETH_VLAN_EXTEND_MASK) {
2192                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2193                         ixgbe_vlan_hw_extend_enable(dev);
2194                 else
2195                         ixgbe_vlan_hw_extend_disable(dev);
2196         }
2197
2198         return 0;
2199 }
2200
2201 static int
2202 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2203 {
2204         ixgbe_config_vlan_strip_on_all_queues(dev, mask);
2205
2206         ixgbe_vlan_offload_config(dev, mask);
2207
2208         return 0;
2209 }
2210
2211 static void
2212 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2213 {
2214         struct ixgbe_hw *hw =
2215                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2216         /* VLNCTRL: enable vlan filtering and allow all vlan tags through */
2217         uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2218
2219         vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
2220         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
2221 }
2222
2223 static int
2224 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
2225 {
2226         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2227
2228         switch (nb_rx_q) {
2229         case 1:
2230         case 2:
2231                 RTE_ETH_DEV_SRIOV(dev).active = ETH_64_POOLS;
2232                 break;
2233         case 4:
2234                 RTE_ETH_DEV_SRIOV(dev).active = ETH_32_POOLS;
2235                 break;
2236         default:
2237                 return -EINVAL;
2238         }
2239
2240         RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
2241                 IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2242         RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
2243                 pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2244         return 0;
2245 }
2246
2247 static int
2248 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
2249 {
2250         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
2251         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2252         uint16_t nb_rx_q = dev->data->nb_rx_queues;
2253         uint16_t nb_tx_q = dev->data->nb_tx_queues;
2254
2255         if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
2256                 /* check multi-queue mode */
2257                 switch (dev_conf->rxmode.mq_mode) {
2258                 case ETH_MQ_RX_VMDQ_DCB:
2259                         PMD_INIT_LOG(INFO, "ETH_MQ_RX_VMDQ_DCB mode supported in SRIOV");
2260                         break;
2261                 case ETH_MQ_RX_VMDQ_DCB_RSS:
2262                         /* DCB/RSS VMDQ in SRIOV mode, not implement yet */
2263                         PMD_INIT_LOG(ERR, "SRIOV active,"
2264                                         " unsupported mq_mode rx %d.",
2265                                         dev_conf->rxmode.mq_mode);
2266                         return -EINVAL;
2267                 case ETH_MQ_RX_RSS:
2268                 case ETH_MQ_RX_VMDQ_RSS:
2269                         dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_RSS;
2270                         if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
2271                                 if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
2272                                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2273                                                 " invalid queue number"
2274                                                 " for VMDQ RSS, allowed"
2275                                                 " value are 1, 2 or 4.");
2276                                         return -EINVAL;
2277                                 }
2278                         break;
2279                 case ETH_MQ_RX_VMDQ_ONLY:
2280                 case ETH_MQ_RX_NONE:
2281                         /* if nothing mq mode configure, use default scheme */
2282                         dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
2283                         break;
2284                 default: /* ETH_MQ_RX_DCB, ETH_MQ_RX_DCB_RSS or ETH_MQ_TX_DCB*/
2285                         /* SRIOV only works in VMDq enable mode */
2286                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2287                                         " wrong mq_mode rx %d.",
2288                                         dev_conf->rxmode.mq_mode);
2289                         return -EINVAL;
2290                 }
2291
2292                 switch (dev_conf->txmode.mq_mode) {
2293                 case ETH_MQ_TX_VMDQ_DCB:
2294                         PMD_INIT_LOG(INFO, "ETH_MQ_TX_VMDQ_DCB mode supported in SRIOV");
2295                         dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_DCB;
2296                         break;
2297                 default: /* ETH_MQ_TX_VMDQ_ONLY or ETH_MQ_TX_NONE */
2298                         dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_ONLY;
2299                         break;
2300                 }
2301
2302                 /* check valid queue number */
2303                 if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2304                     (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2305                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2306                                         " nb_rx_q=%d nb_tx_q=%d queue number"
2307                                         " must be less than or equal to %d.",
2308                                         nb_rx_q, nb_tx_q,
2309                                         RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2310                         return -EINVAL;
2311                 }
2312         } else {
2313                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB_RSS) {
2314                         PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2315                                           " not supported.");
2316                         return -EINVAL;
2317                 }
2318                 /* check configuration for vmdb+dcb mode */
2319                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB) {
2320                         const struct rte_eth_vmdq_dcb_conf *conf;
2321
2322                         if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2323                                 PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2324                                                 IXGBE_VMDQ_DCB_NB_QUEUES);
2325                                 return -EINVAL;
2326                         }
2327                         conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2328                         if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2329                                conf->nb_queue_pools == ETH_32_POOLS)) {
2330                                 PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2331                                                 " nb_queue_pools must be %d or %d.",
2332                                                 ETH_16_POOLS, ETH_32_POOLS);
2333                                 return -EINVAL;
2334                         }
2335                 }
2336                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_VMDQ_DCB) {
2337                         const struct rte_eth_vmdq_dcb_tx_conf *conf;
2338
2339                         if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2340                                 PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2341                                                  IXGBE_VMDQ_DCB_NB_QUEUES);
2342                                 return -EINVAL;
2343                         }
2344                         conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2345                         if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2346                                conf->nb_queue_pools == ETH_32_POOLS)) {
2347                                 PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2348                                                 " nb_queue_pools != %d and"
2349                                                 " nb_queue_pools != %d.",
2350                                                 ETH_16_POOLS, ETH_32_POOLS);
2351                                 return -EINVAL;
2352                         }
2353                 }
2354
2355                 /* For DCB mode check our configuration before we go further */
2356                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB) {
2357                         const struct rte_eth_dcb_rx_conf *conf;
2358
2359                         conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2360                         if (!(conf->nb_tcs == ETH_4_TCS ||
2361                                conf->nb_tcs == ETH_8_TCS)) {
2362                                 PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2363                                                 " and nb_tcs != %d.",
2364                                                 ETH_4_TCS, ETH_8_TCS);
2365                                 return -EINVAL;
2366                         }
2367                 }
2368
2369                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_DCB) {
2370                         const struct rte_eth_dcb_tx_conf *conf;
2371
2372                         conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2373                         if (!(conf->nb_tcs == ETH_4_TCS ||
2374                                conf->nb_tcs == ETH_8_TCS)) {
2375                                 PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2376                                                 " and nb_tcs != %d.",
2377                                                 ETH_4_TCS, ETH_8_TCS);
2378                                 return -EINVAL;
2379                         }
2380                 }
2381
2382                 /*
2383                  * When DCB/VT is off, maximum number of queues changes,
2384                  * except for 82598EB, which remains constant.
2385                  */
2386                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
2387                                 hw->mac.type != ixgbe_mac_82598EB) {
2388                         if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2389                                 PMD_INIT_LOG(ERR,
2390                                              "Neither VT nor DCB are enabled, "
2391                                              "nb_tx_q > %d.",
2392                                              IXGBE_NONE_MODE_TX_NB_QUEUES);
2393                                 return -EINVAL;
2394                         }
2395                 }
2396         }
2397         return 0;
2398 }
2399
2400 static int
2401 ixgbe_dev_configure(struct rte_eth_dev *dev)
2402 {
2403         struct ixgbe_interrupt *intr =
2404                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2405         struct ixgbe_adapter *adapter =
2406                 (struct ixgbe_adapter *)dev->data->dev_private;
2407         int ret;
2408
2409         PMD_INIT_FUNC_TRACE();
2410         /* multipe queue mode checking */
2411         ret  = ixgbe_check_mq_mode(dev);
2412         if (ret != 0) {
2413                 PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2414                             ret);
2415                 return ret;
2416         }
2417
2418         /* set flag to update link status after init */
2419         intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2420
2421         /*
2422          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2423          * allocation or vector Rx preconditions we will reset it.
2424          */
2425         adapter->rx_bulk_alloc_allowed = true;
2426         adapter->rx_vec_allowed = true;
2427
2428         return 0;
2429 }
2430
2431 static void
2432 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2433 {
2434         struct ixgbe_hw *hw =
2435                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2436         struct ixgbe_interrupt *intr =
2437                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2438         uint32_t gpie;
2439
2440         /* only set up it on X550EM_X */
2441         if (hw->mac.type == ixgbe_mac_X550EM_x) {
2442                 gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2443                 gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2444                 IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2445                 if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2446                         intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2447         }
2448 }
2449
2450 int
2451 ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
2452                         uint16_t tx_rate, uint64_t q_msk)
2453 {
2454         struct ixgbe_hw *hw;
2455         struct ixgbe_vf_info *vfinfo;
2456         struct rte_eth_link link;
2457         uint8_t  nb_q_per_pool;
2458         uint32_t queue_stride;
2459         uint32_t queue_idx, idx = 0, vf_idx;
2460         uint32_t queue_end;
2461         uint16_t total_rate = 0;
2462         struct rte_pci_device *pci_dev;
2463
2464         pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2465         rte_eth_link_get_nowait(dev->data->port_id, &link);
2466
2467         if (vf >= pci_dev->max_vfs)
2468                 return -EINVAL;
2469
2470         if (tx_rate > link.link_speed)
2471                 return -EINVAL;
2472
2473         if (q_msk == 0)
2474                 return 0;
2475
2476         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2477         vfinfo = *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
2478         nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2479         queue_stride = IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2480         queue_idx = vf * queue_stride;
2481         queue_end = queue_idx + nb_q_per_pool - 1;
2482         if (queue_end >= hw->mac.max_tx_queues)
2483                 return -EINVAL;
2484
2485         if (vfinfo) {
2486                 for (vf_idx = 0; vf_idx < pci_dev->max_vfs; vf_idx++) {
2487                         if (vf_idx == vf)
2488                                 continue;
2489                         for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
2490                                 idx++)
2491                                 total_rate += vfinfo[vf_idx].tx_rate[idx];
2492                 }
2493         } else {
2494                 return -EINVAL;
2495         }
2496
2497         /* Store tx_rate for this vf. */
2498         for (idx = 0; idx < nb_q_per_pool; idx++) {
2499                 if (((uint64_t)0x1 << idx) & q_msk) {
2500                         if (vfinfo[vf].tx_rate[idx] != tx_rate)
2501                                 vfinfo[vf].tx_rate[idx] = tx_rate;
2502                         total_rate += tx_rate;
2503                 }
2504         }
2505
2506         if (total_rate > dev->data->dev_link.link_speed) {
2507                 /* Reset stored TX rate of the VF if it causes exceed
2508                  * link speed.
2509                  */
2510                 memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
2511                 return -EINVAL;
2512         }
2513
2514         /* Set RTTBCNRC of each queue/pool for vf X  */
2515         for (; queue_idx <= queue_end; queue_idx++) {
2516                 if (0x1 & q_msk)
2517                         ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
2518                 q_msk = q_msk >> 1;
2519         }
2520
2521         return 0;
2522 }
2523
2524 /*
2525  * Configure device link speed and setup link.
2526  * It returns 0 on success.
2527  */
2528 static int
2529 ixgbe_dev_start(struct rte_eth_dev *dev)
2530 {
2531         struct ixgbe_hw *hw =
2532                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2533         struct ixgbe_vf_info *vfinfo =
2534                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2535         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2536         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2537         uint32_t intr_vector = 0;
2538         int err, link_up = 0, negotiate = 0;
2539         uint32_t speed = 0;
2540         uint32_t allowed_speeds = 0;
2541         int mask = 0;
2542         int status;
2543         uint16_t vf, idx;
2544         uint32_t *link_speeds;
2545         struct ixgbe_tm_conf *tm_conf =
2546                 IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2547
2548         PMD_INIT_FUNC_TRACE();
2549
2550         /* IXGBE devices don't support:
2551         *    - half duplex (checked afterwards for valid speeds)
2552         *    - fixed speed: TODO implement
2553         */
2554         if (dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED) {
2555                 PMD_INIT_LOG(ERR,
2556                 "Invalid link_speeds for port %u, fix speed not supported",
2557                                 dev->data->port_id);
2558                 return -EINVAL;
2559         }
2560
2561         /* Stop the link setup handler before resetting the HW. */
2562         rte_eal_alarm_cancel(ixgbe_dev_setup_link_alarm_handler, dev);
2563
2564         /* disable uio/vfio intr/eventfd mapping */
2565         rte_intr_disable(intr_handle);
2566
2567         /* stop adapter */
2568         hw->adapter_stopped = 0;
2569         ixgbe_stop_adapter(hw);
2570
2571         /* reinitialize adapter
2572          * this calls reset and start
2573          */
2574         status = ixgbe_pf_reset_hw(hw);
2575         if (status != 0)
2576                 return -1;
2577         hw->mac.ops.start_hw(hw);
2578         hw->mac.get_link_status = true;
2579
2580         /* configure PF module if SRIOV enabled */
2581         ixgbe_pf_host_configure(dev);
2582
2583         ixgbe_dev_phy_intr_setup(dev);
2584
2585         /* check and configure queue intr-vector mapping */
2586         if ((rte_intr_cap_multiple(intr_handle) ||
2587              !RTE_ETH_DEV_SRIOV(dev).active) &&
2588             dev->data->dev_conf.intr_conf.rxq != 0) {
2589                 intr_vector = dev->data->nb_rx_queues;
2590                 if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2591                         PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2592                                         IXGBE_MAX_INTR_QUEUE_NUM);
2593                         return -ENOTSUP;
2594                 }
2595                 if (rte_intr_efd_enable(intr_handle, intr_vector))
2596                         return -1;
2597         }
2598
2599         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
2600                 intr_handle->intr_vec =
2601                         rte_zmalloc("intr_vec",
2602                                     dev->data->nb_rx_queues * sizeof(int), 0);
2603                 if (intr_handle->intr_vec == NULL) {
2604                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2605                                      " intr_vec", dev->data->nb_rx_queues);
2606                         return -ENOMEM;
2607                 }
2608         }
2609
2610         /* confiugre msix for sleep until rx interrupt */
2611         ixgbe_configure_msix(dev);
2612
2613         /* initialize transmission unit */
2614         ixgbe_dev_tx_init(dev);
2615
2616         /* This can fail when allocating mbufs for descriptor rings */
2617         err = ixgbe_dev_rx_init(dev);
2618         if (err) {
2619                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2620                 goto error;
2621         }
2622
2623         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
2624                 ETH_VLAN_EXTEND_MASK;
2625         err = ixgbe_vlan_offload_config(dev, mask);
2626         if (err) {
2627                 PMD_INIT_LOG(ERR, "Unable to set VLAN offload");
2628                 goto error;
2629         }
2630
2631         if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
2632                 /* Enable vlan filtering for VMDq */
2633                 ixgbe_vmdq_vlan_hw_filter_enable(dev);
2634         }
2635
2636         /* Configure DCB hw */
2637         ixgbe_configure_dcb(dev);
2638
2639         if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2640                 err = ixgbe_fdir_configure(dev);
2641                 if (err)
2642                         goto error;
2643         }
2644
2645         /* Restore vf rate limit */
2646         if (vfinfo != NULL) {
2647                 for (vf = 0; vf < pci_dev->max_vfs; vf++)
2648                         for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2649                                 if (vfinfo[vf].tx_rate[idx] != 0)
2650                                         ixgbe_set_vf_rate_limit(
2651                                                 dev, vf,
2652                                                 vfinfo[vf].tx_rate[idx],
2653                                                 1 << idx);
2654         }
2655
2656         ixgbe_restore_statistics_mapping(dev);
2657
2658         err = ixgbe_dev_rxtx_start(dev);
2659         if (err < 0) {
2660                 PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2661                 goto error;
2662         }
2663
2664         /* Skip link setup if loopback mode is enabled. */
2665         if (dev->data->dev_conf.lpbk_mode != 0) {
2666                 err = ixgbe_check_supported_loopback_mode(dev);
2667                 if (err < 0) {
2668                         PMD_INIT_LOG(ERR, "Unsupported loopback mode");
2669                         goto error;
2670                 } else {
2671                         goto skip_link_setup;
2672                 }
2673         }
2674
2675         if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2676                 err = hw->mac.ops.setup_sfp(hw);
2677                 if (err)
2678                         goto error;
2679         }
2680
2681         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2682                 /* Turn on the copper */
2683                 ixgbe_set_phy_power(hw, true);
2684         } else {
2685                 /* Turn on the laser */
2686                 ixgbe_enable_tx_laser(hw);
2687         }
2688
2689         err = ixgbe_check_link(hw, &speed, &link_up, 0);
2690         if (err)
2691                 goto error;
2692         dev->data->dev_link.link_status = link_up;
2693
2694         err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2695         if (err)
2696                 goto error;
2697
2698         switch (hw->mac.type) {
2699         case ixgbe_mac_X550:
2700         case ixgbe_mac_X550EM_x:
2701         case ixgbe_mac_X550EM_a:
2702                 allowed_speeds = ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2703                         ETH_LINK_SPEED_2_5G |  ETH_LINK_SPEED_5G |
2704                         ETH_LINK_SPEED_10G;
2705                 if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
2706                                 hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
2707                         allowed_speeds = ETH_LINK_SPEED_10M |
2708                                 ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G;
2709                 break;
2710         default:
2711                 allowed_speeds = ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2712                         ETH_LINK_SPEED_10G;
2713         }
2714
2715         link_speeds = &dev->data->dev_conf.link_speeds;
2716         if (*link_speeds & ~allowed_speeds) {
2717                 PMD_INIT_LOG(ERR, "Invalid link setting");
2718                 goto error;
2719         }
2720
2721         speed = 0x0;
2722         if (*link_speeds == ETH_LINK_SPEED_AUTONEG) {
2723                 switch (hw->mac.type) {
2724                 case ixgbe_mac_82598EB:
2725                         speed = IXGBE_LINK_SPEED_82598_AUTONEG;
2726                         break;
2727                 case ixgbe_mac_82599EB:
2728                 case ixgbe_mac_X540:
2729                         speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2730                         break;
2731                 case ixgbe_mac_X550:
2732                 case ixgbe_mac_X550EM_x:
2733                 case ixgbe_mac_X550EM_a:
2734                         speed = IXGBE_LINK_SPEED_X550_AUTONEG;
2735                         break;
2736                 default:
2737                         speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2738                 }
2739         } else {
2740                 if (*link_speeds & ETH_LINK_SPEED_10G)
2741                         speed |= IXGBE_LINK_SPEED_10GB_FULL;
2742                 if (*link_speeds & ETH_LINK_SPEED_5G)
2743                         speed |= IXGBE_LINK_SPEED_5GB_FULL;
2744                 if (*link_speeds & ETH_LINK_SPEED_2_5G)
2745                         speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
2746                 if (*link_speeds & ETH_LINK_SPEED_1G)
2747                         speed |= IXGBE_LINK_SPEED_1GB_FULL;
2748                 if (*link_speeds & ETH_LINK_SPEED_100M)
2749                         speed |= IXGBE_LINK_SPEED_100_FULL;
2750                 if (*link_speeds & ETH_LINK_SPEED_10M)
2751                         speed |= IXGBE_LINK_SPEED_10_FULL;
2752         }
2753
2754         err = ixgbe_setup_link(hw, speed, link_up);
2755         if (err)
2756                 goto error;
2757
2758 skip_link_setup:
2759
2760         if (rte_intr_allow_others(intr_handle)) {
2761                 /* check if lsc interrupt is enabled */
2762                 if (dev->data->dev_conf.intr_conf.lsc != 0)
2763                         ixgbe_dev_lsc_interrupt_setup(dev, TRUE);
2764                 else
2765                         ixgbe_dev_lsc_interrupt_setup(dev, FALSE);
2766                 ixgbe_dev_macsec_interrupt_setup(dev);
2767         } else {
2768                 rte_intr_callback_unregister(intr_handle,
2769                                              ixgbe_dev_interrupt_handler, dev);
2770                 if (dev->data->dev_conf.intr_conf.lsc != 0)
2771                         PMD_INIT_LOG(INFO, "lsc won't enable because of"
2772                                      " no intr multiplex");
2773         }
2774
2775         /* check if rxq interrupt is enabled */
2776         if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2777             rte_intr_dp_is_en(intr_handle))
2778                 ixgbe_dev_rxq_interrupt_setup(dev);
2779
2780         /* enable uio/vfio intr/eventfd mapping */
2781         rte_intr_enable(intr_handle);
2782
2783         /* resume enabled intr since hw reset */
2784         ixgbe_enable_intr(dev);
2785         ixgbe_l2_tunnel_conf(dev);
2786         ixgbe_filter_restore(dev);
2787
2788         if (tm_conf->root && !tm_conf->committed)
2789                 PMD_DRV_LOG(WARNING,
2790                             "please call hierarchy_commit() "
2791                             "before starting the port");
2792
2793         /*
2794          * Update link status right before return, because it may
2795          * start link configuration process in a separate thread.
2796          */
2797         ixgbe_dev_link_update(dev, 0);
2798
2799         return 0;
2800
2801 error:
2802         PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2803         ixgbe_dev_clear_queues(dev);
2804         return -EIO;
2805 }
2806
2807 /*
2808  * Stop device: disable rx and tx functions to allow for reconfiguring.
2809  */
2810 static void
2811 ixgbe_dev_stop(struct rte_eth_dev *dev)
2812 {
2813         struct rte_eth_link link;
2814         struct ixgbe_adapter *adapter =
2815                 (struct ixgbe_adapter *)dev->data->dev_private;
2816         struct ixgbe_hw *hw =
2817                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2818         struct ixgbe_vf_info *vfinfo =
2819                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2820         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2821         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2822         int vf;
2823         struct ixgbe_tm_conf *tm_conf =
2824                 IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2825
2826         PMD_INIT_FUNC_TRACE();
2827
2828         rte_eal_alarm_cancel(ixgbe_dev_setup_link_alarm_handler, dev);
2829
2830         /* disable interrupts */
2831         ixgbe_disable_intr(hw);
2832
2833         /* reset the NIC */
2834         ixgbe_pf_reset_hw(hw);
2835         hw->adapter_stopped = 0;
2836
2837         /* stop adapter */
2838         ixgbe_stop_adapter(hw);
2839
2840         for (vf = 0; vfinfo != NULL && vf < pci_dev->max_vfs; vf++)
2841                 vfinfo[vf].clear_to_send = false;
2842
2843         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2844                 /* Turn off the copper */
2845                 ixgbe_set_phy_power(hw, false);
2846         } else {
2847                 /* Turn off the laser */
2848                 ixgbe_disable_tx_laser(hw);
2849         }
2850
2851         ixgbe_dev_clear_queues(dev);
2852
2853         /* Clear stored conf */
2854         dev->data->scattered_rx = 0;
2855         dev->data->lro = 0;
2856
2857         /* Clear recorded link status */
2858         memset(&link, 0, sizeof(link));
2859         rte_eth_linkstatus_set(dev, &link);
2860
2861         if (!rte_intr_allow_others(intr_handle))
2862                 /* resume to the default handler */
2863                 rte_intr_callback_register(intr_handle,
2864                                            ixgbe_dev_interrupt_handler,
2865                                            (void *)dev);
2866
2867         /* Clean datapath event and queue/vec mapping */
2868         rte_intr_efd_disable(intr_handle);
2869         if (intr_handle->intr_vec != NULL) {
2870                 rte_free(intr_handle->intr_vec);
2871                 intr_handle->intr_vec = NULL;
2872         }
2873
2874         /* reset hierarchy commit */
2875         tm_conf->committed = false;
2876
2877         adapter->rss_reta_updated = 0;
2878 }
2879
2880 /*
2881  * Set device link up: enable tx.
2882  */
2883 static int
2884 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2885 {
2886         struct ixgbe_hw *hw =
2887                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2888         if (hw->mac.type == ixgbe_mac_82599EB) {
2889 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2890                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2891                         /* Not suported in bypass mode */
2892                         PMD_INIT_LOG(ERR, "Set link up is not supported "
2893                                      "by device id 0x%x", hw->device_id);
2894                         return -ENOTSUP;
2895                 }
2896 #endif
2897         }
2898
2899         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2900                 /* Turn on the copper */
2901                 ixgbe_set_phy_power(hw, true);
2902         } else {
2903                 /* Turn on the laser */
2904                 ixgbe_enable_tx_laser(hw);
2905         }
2906
2907         return 0;
2908 }
2909
2910 /*
2911  * Set device link down: disable tx.
2912  */
2913 static int
2914 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2915 {
2916         struct ixgbe_hw *hw =
2917                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2918         if (hw->mac.type == ixgbe_mac_82599EB) {
2919 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2920                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2921                         /* Not suported in bypass mode */
2922                         PMD_INIT_LOG(ERR, "Set link down is not supported "
2923                                      "by device id 0x%x", hw->device_id);
2924                         return -ENOTSUP;
2925                 }
2926 #endif
2927         }
2928
2929         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2930                 /* Turn off the copper */
2931                 ixgbe_set_phy_power(hw, false);
2932         } else {
2933                 /* Turn off the laser */
2934                 ixgbe_disable_tx_laser(hw);
2935         }
2936
2937         return 0;
2938 }
2939
2940 /*
2941  * Reset and stop device.
2942  */
2943 static void
2944 ixgbe_dev_close(struct rte_eth_dev *dev)
2945 {
2946         struct ixgbe_hw *hw =
2947                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2948
2949         PMD_INIT_FUNC_TRACE();
2950
2951         ixgbe_pf_reset_hw(hw);
2952
2953         ixgbe_dev_stop(dev);
2954         hw->adapter_stopped = 1;
2955
2956         ixgbe_dev_free_queues(dev);
2957
2958         ixgbe_disable_pcie_master(hw);
2959
2960         /* reprogram the RAR[0] in case user changed it. */
2961         ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
2962 }
2963
2964 /*
2965  * Reset PF device.
2966  */
2967 static int
2968 ixgbe_dev_reset(struct rte_eth_dev *dev)
2969 {
2970         int ret;
2971
2972         /* When a DPDK PMD PF begin to reset PF port, it should notify all
2973          * its VF to make them align with it. The detailed notification
2974          * mechanism is PMD specific. As to ixgbe PF, it is rather complex.
2975          * To avoid unexpected behavior in VF, currently reset of PF with
2976          * SR-IOV activation is not supported. It might be supported later.
2977          */
2978         if (dev->data->sriov.active)
2979                 return -ENOTSUP;
2980
2981         ret = eth_ixgbe_dev_uninit(dev);
2982         if (ret)
2983                 return ret;
2984
2985         ret = eth_ixgbe_dev_init(dev, NULL);
2986
2987         return ret;
2988 }
2989
2990 static void
2991 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
2992                            struct ixgbe_hw_stats *hw_stats,
2993                            struct ixgbe_macsec_stats *macsec_stats,
2994                            uint64_t *total_missed_rx, uint64_t *total_qbrc,
2995                            uint64_t *total_qprc, uint64_t *total_qprdc)
2996 {
2997         uint32_t bprc, lxon, lxoff, total;
2998         uint32_t delta_gprc = 0;
2999         unsigned i;
3000         /* Workaround for RX byte count not including CRC bytes when CRC
3001          * strip is enabled. CRC bytes are removed from counters when crc_strip
3002          * is disabled.
3003          */
3004         int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
3005                         IXGBE_HLREG0_RXCRCSTRP);
3006
3007         hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
3008         hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
3009         hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
3010         hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
3011
3012         for (i = 0; i < 8; i++) {
3013                 uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
3014
3015                 /* global total per queue */
3016                 hw_stats->mpc[i] += mp;
3017                 /* Running comprehensive total for stats display */
3018                 *total_missed_rx += hw_stats->mpc[i];
3019                 if (hw->mac.type == ixgbe_mac_82598EB) {
3020                         hw_stats->rnbc[i] +=
3021                             IXGBE_READ_REG(hw, IXGBE_RNBC(i));
3022                         hw_stats->pxonrxc[i] +=
3023                                 IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
3024                         hw_stats->pxoffrxc[i] +=
3025                                 IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
3026                 } else {
3027                         hw_stats->pxonrxc[i] +=
3028                                 IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
3029                         hw_stats->pxoffrxc[i] +=
3030                                 IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
3031                         hw_stats->pxon2offc[i] +=
3032                                 IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
3033                 }
3034                 hw_stats->pxontxc[i] +=
3035                     IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
3036                 hw_stats->pxofftxc[i] +=
3037                     IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
3038         }
3039         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3040                 uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
3041                 uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
3042                 uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
3043
3044                 delta_gprc += delta_qprc;
3045
3046                 hw_stats->qprc[i] += delta_qprc;
3047                 hw_stats->qptc[i] += delta_qptc;
3048
3049                 hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
3050                 hw_stats->qbrc[i] +=
3051                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
3052                 if (crc_strip == 0)
3053                         hw_stats->qbrc[i] -= delta_qprc * ETHER_CRC_LEN;
3054
3055                 hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
3056                 hw_stats->qbtc[i] +=
3057                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
3058
3059                 hw_stats->qprdc[i] += delta_qprdc;
3060                 *total_qprdc += hw_stats->qprdc[i];
3061
3062                 *total_qprc += hw_stats->qprc[i];
3063                 *total_qbrc += hw_stats->qbrc[i];
3064         }
3065         hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
3066         hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
3067         hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
3068
3069         /*
3070          * An errata states that gprc actually counts good + missed packets:
3071          * Workaround to set gprc to summated queue packet receives
3072          */
3073         hw_stats->gprc = *total_qprc;
3074
3075         if (hw->mac.type != ixgbe_mac_82598EB) {
3076                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
3077                 hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
3078                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
3079                 hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
3080                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
3081                 hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
3082                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
3083                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
3084         } else {
3085                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
3086                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
3087                 /* 82598 only has a counter in the high register */
3088                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
3089                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
3090                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
3091         }
3092         uint64_t old_tpr = hw_stats->tpr;
3093
3094         hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
3095         hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
3096
3097         if (crc_strip == 0)
3098                 hw_stats->gorc -= delta_gprc * ETHER_CRC_LEN;
3099
3100         uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
3101         hw_stats->gptc += delta_gptc;
3102         hw_stats->gotc -= delta_gptc * ETHER_CRC_LEN;
3103         hw_stats->tor -= (hw_stats->tpr - old_tpr) * ETHER_CRC_LEN;
3104
3105         /*
3106          * Workaround: mprc hardware is incorrectly counting
3107          * broadcasts, so for now we subtract those.
3108          */
3109         bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
3110         hw_stats->bprc += bprc;
3111         hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
3112         if (hw->mac.type == ixgbe_mac_82598EB)
3113                 hw_stats->mprc -= bprc;
3114
3115         hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
3116         hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
3117         hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
3118         hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
3119         hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
3120         hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
3121
3122         lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
3123         hw_stats->lxontxc += lxon;
3124         lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
3125         hw_stats->lxofftxc += lxoff;
3126         total = lxon + lxoff;
3127
3128         hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
3129         hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
3130         hw_stats->gptc -= total;
3131         hw_stats->mptc -= total;
3132         hw_stats->ptc64 -= total;
3133         hw_stats->gotc -= total * ETHER_MIN_LEN;
3134
3135         hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3136         hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3137         hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3138         hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
3139         hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
3140         hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
3141         hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
3142         hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3143         hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3144         hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3145         hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3146         hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
3147         hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3148         hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
3149         hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
3150         hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
3151         /* Only read FCOE on 82599 */
3152         if (hw->mac.type != ixgbe_mac_82598EB) {
3153                 hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
3154                 hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
3155                 hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
3156                 hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
3157                 hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
3158         }
3159
3160         /* Flow Director Stats registers */
3161         if (hw->mac.type != ixgbe_mac_82598EB) {
3162                 hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
3163                 hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
3164                 hw_stats->fdirustat_add += IXGBE_READ_REG(hw,
3165                                         IXGBE_FDIRUSTAT) & 0xFFFF;
3166                 hw_stats->fdirustat_remove += (IXGBE_READ_REG(hw,
3167                                         IXGBE_FDIRUSTAT) >> 16) & 0xFFFF;
3168                 hw_stats->fdirfstat_fadd += IXGBE_READ_REG(hw,
3169                                         IXGBE_FDIRFSTAT) & 0xFFFF;
3170                 hw_stats->fdirfstat_fremove += (IXGBE_READ_REG(hw,
3171                                         IXGBE_FDIRFSTAT) >> 16) & 0xFFFF;
3172         }
3173         /* MACsec Stats registers */
3174         macsec_stats->out_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECTXUT);
3175         macsec_stats->out_pkts_encrypted +=
3176                 IXGBE_READ_REG(hw, IXGBE_LSECTXPKTE);
3177         macsec_stats->out_pkts_protected +=
3178                 IXGBE_READ_REG(hw, IXGBE_LSECTXPKTP);
3179         macsec_stats->out_octets_encrypted +=
3180                 IXGBE_READ_REG(hw, IXGBE_LSECTXOCTE);
3181         macsec_stats->out_octets_protected +=
3182                 IXGBE_READ_REG(hw, IXGBE_LSECTXOCTP);
3183         macsec_stats->in_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECRXUT);
3184         macsec_stats->in_pkts_badtag += IXGBE_READ_REG(hw, IXGBE_LSECRXBAD);
3185         macsec_stats->in_pkts_nosci += IXGBE_READ_REG(hw, IXGBE_LSECRXNOSCI);
3186         macsec_stats->in_pkts_unknownsci +=
3187                 IXGBE_READ_REG(hw, IXGBE_LSECRXUNSCI);
3188         macsec_stats->in_octets_decrypted +=
3189                 IXGBE_READ_REG(hw, IXGBE_LSECRXOCTD);
3190         macsec_stats->in_octets_validated +=
3191                 IXGBE_READ_REG(hw, IXGBE_LSECRXOCTV);
3192         macsec_stats->in_pkts_unchecked += IXGBE_READ_REG(hw, IXGBE_LSECRXUNCH);
3193         macsec_stats->in_pkts_delayed += IXGBE_READ_REG(hw, IXGBE_LSECRXDELAY);
3194         macsec_stats->in_pkts_late += IXGBE_READ_REG(hw, IXGBE_LSECRXLATE);
3195         for (i = 0; i < 2; i++) {
3196                 macsec_stats->in_pkts_ok +=
3197                         IXGBE_READ_REG(hw, IXGBE_LSECRXOK(i));
3198                 macsec_stats->in_pkts_invalid +=
3199                         IXGBE_READ_REG(hw, IXGBE_LSECRXINV(i));
3200                 macsec_stats->in_pkts_notvalid +=
3201                         IXGBE_READ_REG(hw, IXGBE_LSECRXNV(i));
3202         }
3203         macsec_stats->in_pkts_unusedsa += IXGBE_READ_REG(hw, IXGBE_LSECRXUNSA);
3204         macsec_stats->in_pkts_notusingsa +=
3205                 IXGBE_READ_REG(hw, IXGBE_LSECRXNUSA);
3206 }
3207
3208 /*
3209  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
3210  */
3211 static int
3212 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3213 {
3214         struct ixgbe_hw *hw =
3215                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3216         struct ixgbe_hw_stats *hw_stats =
3217                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3218         struct ixgbe_macsec_stats *macsec_stats =
3219                         IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3220                                 dev->data->dev_private);
3221         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3222         unsigned i;
3223
3224         total_missed_rx = 0;
3225         total_qbrc = 0;
3226         total_qprc = 0;
3227         total_qprdc = 0;
3228
3229         ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3230                         &total_qbrc, &total_qprc, &total_qprdc);
3231
3232         if (stats == NULL)
3233                 return -EINVAL;
3234
3235         /* Fill out the rte_eth_stats statistics structure */
3236         stats->ipackets = total_qprc;
3237         stats->ibytes = total_qbrc;
3238         stats->opackets = hw_stats->gptc;
3239         stats->obytes = hw_stats->gotc;
3240
3241         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3242                 stats->q_ipackets[i] = hw_stats->qprc[i];
3243                 stats->q_opackets[i] = hw_stats->qptc[i];
3244                 stats->q_ibytes[i] = hw_stats->qbrc[i];
3245                 stats->q_obytes[i] = hw_stats->qbtc[i];
3246                 stats->q_errors[i] = hw_stats->qprdc[i];
3247         }
3248
3249         /* Rx Errors */
3250         stats->imissed  = total_missed_rx;
3251         stats->ierrors  = hw_stats->crcerrs +
3252                           hw_stats->mspdc +
3253                           hw_stats->rlec +
3254                           hw_stats->ruc +
3255                           hw_stats->roc +
3256                           hw_stats->illerrc +
3257                           hw_stats->errbc +
3258                           hw_stats->rfc +
3259                           hw_stats->fccrc +
3260                           hw_stats->fclast;
3261
3262         /* Tx Errors */
3263         stats->oerrors  = 0;
3264         return 0;
3265 }
3266
3267 static void
3268 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
3269 {
3270         struct ixgbe_hw_stats *stats =
3271                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3272
3273         /* HW registers are cleared on read */
3274         ixgbe_dev_stats_get(dev, NULL);
3275
3276         /* Reset software totals */
3277         memset(stats, 0, sizeof(*stats));
3278 }
3279
3280 /* This function calculates the number of xstats based on the current config */
3281 static unsigned
3282 ixgbe_xstats_calc_num(void) {
3283         return IXGBE_NB_HW_STATS + IXGBE_NB_MACSEC_STATS +
3284                 (IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
3285                 (IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
3286 }
3287
3288 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3289         struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned int size)
3290 {
3291         const unsigned cnt_stats = ixgbe_xstats_calc_num();
3292         unsigned stat, i, count;
3293
3294         if (xstats_names != NULL) {
3295                 count = 0;
3296
3297                 /* Note: limit >= cnt_stats checked upstream
3298                  * in rte_eth_xstats_names()
3299                  */
3300
3301                 /* Extended stats from ixgbe_hw_stats */
3302                 for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3303                         strlcpy(xstats_names[count].name,
3304                                 rte_ixgbe_stats_strings[i].name,
3305                                 sizeof(xstats_names[count].name));
3306                         count++;
3307                 }
3308
3309                 /* MACsec Stats */
3310                 for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3311                         strlcpy(xstats_names[count].name,
3312                                 rte_ixgbe_macsec_strings[i].name,
3313                                 sizeof(xstats_names[count].name));
3314                         count++;
3315                 }
3316
3317                 /* RX Priority Stats */
3318                 for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3319                         for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3320                                 snprintf(xstats_names[count].name,
3321                                         sizeof(xstats_names[count].name),
3322                                         "rx_priority%u_%s", i,
3323                                         rte_ixgbe_rxq_strings[stat].name);
3324                                 count++;
3325                         }
3326                 }
3327
3328                 /* TX Priority Stats */
3329                 for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3330                         for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3331                                 snprintf(xstats_names[count].name,
3332                                         sizeof(xstats_names[count].name),
3333                                         "tx_priority%u_%s", i,
3334                                         rte_ixgbe_txq_strings[stat].name);
3335                                 count++;
3336                         }
3337                 }
3338         }
3339         return cnt_stats;
3340 }
3341
3342 static int ixgbe_dev_xstats_get_names_by_id(
3343         struct rte_eth_dev *dev,
3344         struct rte_eth_xstat_name *xstats_names,
3345         const uint64_t *ids,
3346         unsigned int limit)
3347 {
3348         if (!ids) {
3349                 const unsigned int cnt_stats = ixgbe_xstats_calc_num();
3350                 unsigned int stat, i, count;
3351
3352                 if (xstats_names != NULL) {
3353                         count = 0;
3354
3355                         /* Note: limit >= cnt_stats checked upstream
3356                          * in rte_eth_xstats_names()
3357                          */
3358
3359                         /* Extended stats from ixgbe_hw_stats */
3360                         for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3361                                 strlcpy(xstats_names[count].name,
3362                                         rte_ixgbe_stats_strings[i].name,
3363                                         sizeof(xstats_names[count].name));
3364                                 count++;
3365                         }
3366
3367                         /* MACsec Stats */
3368                         for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3369                                 strlcpy(xstats_names[count].name,
3370                                         rte_ixgbe_macsec_strings[i].name,
3371                                         sizeof(xstats_names[count].name));
3372                                 count++;
3373                         }
3374
3375                         /* RX Priority Stats */
3376                         for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3377                                 for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3378                                         snprintf(xstats_names[count].name,
3379                                             sizeof(xstats_names[count].name),
3380                                             "rx_priority%u_%s", i,
3381                                             rte_ixgbe_rxq_strings[stat].name);
3382                                         count++;
3383                                 }
3384                         }
3385
3386                         /* TX Priority Stats */
3387                         for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3388                                 for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3389                                         snprintf(xstats_names[count].name,
3390                                             sizeof(xstats_names[count].name),
3391                                             "tx_priority%u_%s", i,
3392                                             rte_ixgbe_txq_strings[stat].name);
3393                                         count++;
3394                                 }
3395                         }
3396                 }
3397                 return cnt_stats;
3398         }
3399
3400         uint16_t i;
3401         uint16_t size = ixgbe_xstats_calc_num();
3402         struct rte_eth_xstat_name xstats_names_copy[size];
3403
3404         ixgbe_dev_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
3405                         size);
3406
3407         for (i = 0; i < limit; i++) {
3408                 if (ids[i] >= size) {
3409                         PMD_INIT_LOG(ERR, "id value isn't valid");
3410                         return -1;
3411                 }
3412                 strcpy(xstats_names[i].name,
3413                                 xstats_names_copy[ids[i]].name);
3414         }
3415         return limit;
3416 }
3417
3418 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3419         struct rte_eth_xstat_name *xstats_names, unsigned limit)
3420 {
3421         unsigned i;
3422
3423         if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
3424                 return -ENOMEM;
3425
3426         if (xstats_names != NULL)
3427                 for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
3428                         strlcpy(xstats_names[i].name,
3429                                 rte_ixgbevf_stats_strings[i].name,
3430                                 sizeof(xstats_names[i].name));
3431         return IXGBEVF_NB_XSTATS;
3432 }
3433
3434 static int
3435 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3436                                          unsigned n)
3437 {
3438         struct ixgbe_hw *hw =
3439                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3440         struct ixgbe_hw_stats *hw_stats =
3441                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3442         struct ixgbe_macsec_stats *macsec_stats =
3443                         IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3444                                 dev->data->dev_private);
3445         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3446         unsigned i, stat, count = 0;
3447
3448         count = ixgbe_xstats_calc_num();
3449
3450         if (n < count)
3451                 return count;
3452
3453         total_missed_rx = 0;
3454         total_qbrc = 0;
3455         total_qprc = 0;
3456         total_qprdc = 0;
3457
3458         ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3459                         &total_qbrc, &total_qprc, &total_qprdc);
3460
3461         /* If this is a reset xstats is NULL, and we have cleared the
3462          * registers by reading them.
3463          */
3464         if (!xstats)
3465                 return 0;
3466
3467         /* Extended stats from ixgbe_hw_stats */
3468         count = 0;
3469         for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3470                 xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3471                                 rte_ixgbe_stats_strings[i].offset);
3472                 xstats[count].id = count;
3473                 count++;
3474         }
3475
3476         /* MACsec Stats */
3477         for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3478                 xstats[count].value = *(uint64_t *)(((char *)macsec_stats) +
3479                                 rte_ixgbe_macsec_strings[i].offset);
3480                 xstats[count].id = count;
3481                 count++;
3482         }
3483
3484         /* RX Priority Stats */
3485         for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3486                 for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3487                         xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3488                                         rte_ixgbe_rxq_strings[stat].offset +
3489                                         (sizeof(uint64_t) * i));
3490                         xstats[count].id = count;
3491                         count++;
3492                 }
3493         }
3494
3495         /* TX Priority Stats */
3496         for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3497                 for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3498                         xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3499                                         rte_ixgbe_txq_strings[stat].offset +
3500                                         (sizeof(uint64_t) * i));
3501                         xstats[count].id = count;
3502                         count++;
3503                 }
3504         }
3505         return count;
3506 }
3507
3508 static int
3509 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
3510                 uint64_t *values, unsigned int n)
3511 {
3512         if (!ids) {
3513                 struct ixgbe_hw *hw =
3514                                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3515                 struct ixgbe_hw_stats *hw_stats =
3516                                 IXGBE_DEV_PRIVATE_TO_STATS(
3517                                                 dev->data->dev_private);
3518                 struct ixgbe_macsec_stats *macsec_stats =
3519                                 IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3520                                         dev->data->dev_private);
3521                 uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3522                 unsigned int i, stat, count = 0;
3523
3524                 count = ixgbe_xstats_calc_num();
3525
3526                 if (!ids && n < count)
3527                         return count;
3528
3529                 total_missed_rx = 0;
3530                 total_qbrc = 0;
3531                 total_qprc = 0;
3532                 total_qprdc = 0;
3533
3534                 ixgbe_read_stats_registers(hw, hw_stats, macsec_stats,
3535                                 &total_missed_rx, &total_qbrc, &total_qprc,
3536                                 &total_qprdc);
3537
3538                 /* If this is a reset xstats is NULL, and we have cleared the
3539                  * registers by reading them.
3540                  */
3541                 if (!ids && !values)
3542                         return 0;
3543
3544                 /* Extended stats from ixgbe_hw_stats */
3545                 count = 0;
3546                 for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3547                         values[count] = *(uint64_t *)(((char *)hw_stats) +
3548                                         rte_ixgbe_stats_strings[i].offset);
3549                         count++;
3550                 }
3551
3552                 /* MACsec Stats */
3553                 for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3554                         values[count] = *(uint64_t *)(((char *)macsec_stats) +
3555                                         rte_ixgbe_macsec_strings[i].offset);
3556                         count++;
3557                 }
3558
3559                 /* RX Priority Stats */
3560                 for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3561                         for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3562                                 values[count] =
3563                                         *(uint64_t *)(((char *)hw_stats) +
3564                                         rte_ixgbe_rxq_strings[stat].offset +
3565                                         (sizeof(uint64_t) * i));
3566                                 count++;
3567                         }
3568                 }
3569
3570                 /* TX Priority Stats */
3571                 for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3572                         for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3573                                 values[count] =
3574                                         *(uint64_t *)(((char *)hw_stats) +
3575                                         rte_ixgbe_txq_strings[stat].offset +
3576                                         (sizeof(uint64_t) * i));
3577                                 count++;
3578                         }
3579                 }
3580                 return count;
3581         }
3582
3583         uint16_t i;
3584         uint16_t size = ixgbe_xstats_calc_num();
3585         uint64_t values_copy[size];
3586
3587         ixgbe_dev_xstats_get_by_id(dev, NULL, values_copy, size);
3588
3589         for (i = 0; i < n; i++) {
3590                 if (ids[i] >= size) {
3591                         PMD_INIT_LOG(ERR, "id value isn't valid");
3592                         return -1;
3593                 }
3594                 values[i] = values_copy[ids[i]];
3595         }
3596         return n;
3597 }
3598
3599 static void
3600 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
3601 {
3602         struct ixgbe_hw_stats *stats =
3603                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3604         struct ixgbe_macsec_stats *macsec_stats =
3605                         IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3606                                 dev->data->dev_private);
3607
3608         unsigned count = ixgbe_xstats_calc_num();
3609
3610         /* HW registers are cleared on read */
3611         ixgbe_dev_xstats_get(dev, NULL, count);
3612
3613         /* Reset software totals */
3614         memset(stats, 0, sizeof(*stats));
3615         memset(macsec_stats, 0, sizeof(*macsec_stats));
3616 }
3617
3618 static void
3619 ixgbevf_update_stats(struct rte_eth_dev *dev)
3620 {
3621         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3622         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3623                           IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3624
3625         /* Good Rx packet, include VF loopback */
3626         UPDATE_VF_STAT(IXGBE_VFGPRC,
3627             hw_stats->last_vfgprc, hw_stats->vfgprc);
3628
3629         /* Good Rx octets, include VF loopback */
3630         UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
3631             hw_stats->last_vfgorc, hw_stats->vfgorc);
3632
3633         /* Good Tx packet, include VF loopback */
3634         UPDATE_VF_STAT(IXGBE_VFGPTC,
3635             hw_stats->last_vfgptc, hw_stats->vfgptc);
3636
3637         /* Good Tx octets, include VF loopback */
3638         UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
3639             hw_stats->last_vfgotc, hw_stats->vfgotc);
3640
3641         /* Rx Multicst Packet */
3642         UPDATE_VF_STAT(IXGBE_VFMPRC,
3643             hw_stats->last_vfmprc, hw_stats->vfmprc);
3644 }
3645
3646 static int
3647 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3648                        unsigned n)
3649 {
3650         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3651                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3652         unsigned i;
3653
3654         if (n < IXGBEVF_NB_XSTATS)
3655                 return IXGBEVF_NB_XSTATS;
3656
3657         ixgbevf_update_stats(dev);
3658
3659         if (!xstats)
3660                 return 0;
3661
3662         /* Extended stats */
3663         for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
3664                 xstats[i].id = i;
3665                 xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
3666                         rte_ixgbevf_stats_strings[i].offset);
3667         }
3668
3669         return IXGBEVF_NB_XSTATS;
3670 }
3671
3672 static int
3673 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3674 {
3675         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3676                           IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3677
3678         ixgbevf_update_stats(dev);
3679
3680         if (stats == NULL)
3681                 return -EINVAL;
3682
3683         stats->ipackets = hw_stats->vfgprc;
3684         stats->ibytes = hw_stats->vfgorc;
3685         stats->opackets = hw_stats->vfgptc;
3686         stats->obytes = hw_stats->vfgotc;
3687         return 0;
3688 }
3689
3690 static void
3691 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3692 {
3693         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3694                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3695
3696         /* Sync HW register to the last stats */
3697         ixgbevf_dev_stats_get(dev, NULL);
3698
3699         /* reset HW current stats*/
3700         hw_stats->vfgprc = 0;
3701         hw_stats->vfgorc = 0;
3702         hw_stats->vfgptc = 0;
3703         hw_stats->vfgotc = 0;
3704 }
3705
3706 static int
3707 ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
3708 {
3709         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3710         u16 eeprom_verh, eeprom_verl;
3711         u32 etrack_id;
3712         int ret;
3713
3714         ixgbe_read_eeprom(hw, 0x2e, &eeprom_verh);
3715         ixgbe_read_eeprom(hw, 0x2d, &eeprom_verl);
3716
3717         etrack_id = (eeprom_verh << 16) | eeprom_verl;
3718         ret = snprintf(fw_version, fw_size, "0x%08x", etrack_id);
3719
3720         ret += 1; /* add the size of '\0' */
3721         if (fw_size < (u32)ret)
3722                 return ret;
3723         else
3724                 return 0;
3725 }
3726
3727 static void
3728 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3729 {
3730         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3731         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3732         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3733
3734         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3735         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3736         if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3737                 /*
3738                  * When DCB/VT is off, maximum number of queues changes,
3739                  * except for 82598EB, which remains constant.
3740                  */
3741                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
3742                                 hw->mac.type != ixgbe_mac_82598EB)
3743                         dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3744         }
3745         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3746         dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3747         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3748         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3749         dev_info->max_vfs = pci_dev->max_vfs;
3750         if (hw->mac.type == ixgbe_mac_82598EB)
3751                 dev_info->max_vmdq_pools = ETH_16_POOLS;
3752         else
3753                 dev_info->max_vmdq_pools = ETH_64_POOLS;
3754         dev_info->max_mtu =  dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3755         dev_info->min_mtu = ETHER_MIN_MTU;
3756         dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3757         dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3758         dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3759                                      dev_info->rx_queue_offload_capa);
3760         dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3761         dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3762
3763         dev_info->default_rxconf = (struct rte_eth_rxconf) {
3764                 .rx_thresh = {
3765                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3766                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3767                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3768                 },
3769                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3770                 .rx_drop_en = 0,
3771                 .offloads = 0,
3772         };
3773
3774         dev_info->default_txconf = (struct rte_eth_txconf) {
3775                 .tx_thresh = {
3776                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3777                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3778                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3779                 },
3780                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3781                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3782                 .offloads = 0,
3783         };
3784
3785         dev_info->rx_desc_lim = rx_desc_lim;
3786         dev_info->tx_desc_lim = tx_desc_lim;
3787
3788         dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3789         dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3790         dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3791
3792         dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G;
3793         if (hw->mac.type == ixgbe_mac_X540 ||
3794             hw->mac.type == ixgbe_mac_X540_vf ||
3795             hw->mac.type == ixgbe_mac_X550 ||
3796             hw->mac.type == ixgbe_mac_X550_vf) {
3797                 dev_info->speed_capa |= ETH_LINK_SPEED_100M;
3798         }
3799         if (hw->mac.type == ixgbe_mac_X550) {
3800                 dev_info->speed_capa |= ETH_LINK_SPEED_2_5G;
3801                 dev_info->speed_capa |= ETH_LINK_SPEED_5G;
3802         }
3803
3804         /* Driver-preferred Rx/Tx parameters */
3805         dev_info->default_rxportconf.burst_size = 32;
3806         dev_info->default_txportconf.burst_size = 32;
3807         dev_info->default_rxportconf.nb_queues = 1;
3808         dev_info->default_txportconf.nb_queues = 1;
3809         dev_info->default_rxportconf.ring_size = 256;
3810         dev_info->default_txportconf.ring_size = 256;
3811 }
3812
3813 static const uint32_t *
3814 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3815 {
3816         static const uint32_t ptypes[] = {
3817                 /* For non-vec functions,
3818                  * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3819                  * for vec functions,
3820                  * refers to _recv_raw_pkts_vec().
3821                  */
3822                 RTE_PTYPE_L2_ETHER,
3823                 RTE_PTYPE_L3_IPV4,
3824                 RTE_PTYPE_L3_IPV4_EXT,
3825                 RTE_PTYPE_L3_IPV6,
3826                 RTE_PTYPE_L3_IPV6_EXT,
3827                 RTE_PTYPE_L4_SCTP,
3828                 RTE_PTYPE_L4_TCP,
3829                 RTE_PTYPE_L4_UDP,
3830                 RTE_PTYPE_TUNNEL_IP,
3831                 RTE_PTYPE_INNER_L3_IPV6,
3832                 RTE_PTYPE_INNER_L3_IPV6_EXT,
3833                 RTE_PTYPE_INNER_L4_TCP,
3834                 RTE_PTYPE_INNER_L4_UDP,
3835                 RTE_PTYPE_UNKNOWN
3836         };
3837
3838         if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3839             dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3840             dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3841             dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3842                 return ptypes;
3843
3844 #if defined(RTE_ARCH_X86)
3845         if (dev->rx_pkt_burst == ixgbe_recv_pkts_vec ||
3846             dev->rx_pkt_burst == ixgbe_recv_scattered_pkts_vec)
3847                 return ptypes;
3848 #endif
3849         return NULL;
3850 }
3851
3852 static void
3853 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3854                      struct rte_eth_dev_info *dev_info)
3855 {
3856         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3857         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3858
3859         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3860         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3861         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3862         dev_info->max_rx_pktlen = 9728; /* includes CRC, cf MAXFRS reg */
3863         dev_info->max_mtu = dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3864         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3865         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3866         dev_info->max_vfs = pci_dev->max_vfs;
3867         if (hw->mac.type == ixgbe_mac_82598EB)
3868                 dev_info->max_vmdq_pools = ETH_16_POOLS;
3869         else
3870                 dev_info->max_vmdq_pools = ETH_64_POOLS;
3871         dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3872         dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3873                                      dev_info->rx_queue_offload_capa);
3874         dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3875         dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3876
3877         dev_info->default_rxconf = (struct rte_eth_rxconf) {
3878                 .rx_thresh = {
3879                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3880                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3881                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3882                 },
3883                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3884                 .rx_drop_en = 0,
3885                 .offloads = 0,
3886         };
3887
3888         dev_info->default_txconf = (struct rte_eth_txconf) {
3889                 .tx_thresh = {
3890                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3891                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3892                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3893                 },
3894                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3895                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3896                 .offloads = 0,
3897         };
3898
3899         dev_info->rx_desc_lim = rx_desc_lim;
3900         dev_info->tx_desc_lim = tx_desc_lim;
3901 }
3902
3903 static int
3904 ixgbevf_check_link(struct ixgbe_hw *hw, ixgbe_link_speed *speed,
3905                    int *link_up, int wait_to_complete)
3906 {
3907         struct ixgbe_mbx_info *mbx = &hw->mbx;
3908         struct ixgbe_mac_info *mac = &hw->mac;
3909         uint32_t links_reg, in_msg;
3910         int ret_val = 0;
3911
3912         /* If we were hit with a reset drop the link */
3913         if (!mbx->ops.check_for_rst(hw, 0) || !mbx->timeout)
3914                 mac->get_link_status = true;
3915
3916         if (!mac->get_link_status)
3917                 goto out;
3918
3919         /* if link status is down no point in checking to see if pf is up */
3920         links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
3921         if (!(links_reg & IXGBE_LINKS_UP))
3922                 goto out;
3923
3924         /* for SFP+ modules and DA cables on 82599 it can take up to 500usecs
3925          * before the link status is correct
3926          */
3927         if (mac->type == ixgbe_mac_82599_vf && wait_to_complete) {
3928                 int i;
3929
3930                 for (i = 0; i < 5; i++) {
3931                         rte_delay_us(100);
3932                         links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
3933
3934                         if (!(links_reg & IXGBE_LINKS_UP))
3935                                 goto out;
3936                 }
3937         }
3938
3939         switch (links_reg & IXGBE_LINKS_SPEED_82599) {
3940         case IXGBE_LINKS_SPEED_10G_82599:
3941                 *speed = IXGBE_LINK_SPEED_10GB_FULL;
3942                 if (hw->mac.type >= ixgbe_mac_X550) {
3943                         if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
3944                                 *speed = IXGBE_LINK_SPEED_2_5GB_FULL;
3945                 }
3946                 break;
3947         case IXGBE_LINKS_SPEED_1G_82599:
3948                 *speed = IXGBE_LINK_SPEED_1GB_FULL;
3949                 break;
3950         case IXGBE_LINKS_SPEED_100_82599:
3951                 *speed = IXGBE_LINK_SPEED_100_FULL;
3952                 if (hw->mac.type == ixgbe_mac_X550) {
3953                         if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
3954                                 *speed = IXGBE_LINK_SPEED_5GB_FULL;
3955                 }
3956                 break;
3957         case IXGBE_LINKS_SPEED_10_X550EM_A:
3958                 *speed = IXGBE_LINK_SPEED_UNKNOWN;
3959                 /* Since Reserved in older MAC's */
3960                 if (hw->mac.type >= ixgbe_mac_X550)
3961                         *speed = IXGBE_LINK_SPEED_10_FULL;
3962                 break;
3963         default:
3964                 *speed = IXGBE_LINK_SPEED_UNKNOWN;
3965         }
3966
3967         /* if the read failed it could just be a mailbox collision, best wait
3968          * until we are called again and don't report an error
3969          */
3970         if (mbx->ops.read(hw, &in_msg, 1, 0))
3971                 goto out;
3972
3973         if (!(in_msg & IXGBE_VT_MSGTYPE_CTS)) {
3974                 /* msg is not CTS and is NACK we must have lost CTS status */
3975                 if (in_msg & IXGBE_VT_MSGTYPE_NACK)
3976                         mac->get_link_status = false;
3977                 goto out;
3978         }
3979
3980         /* the pf is talking, if we timed out in the past we reinit */
3981         if (!mbx->timeout) {
3982                 ret_val = -1;
3983                 goto out;
3984         }
3985
3986         /* if we passed all the tests above then the link is up and we no
3987          * longer need to check for link
3988          */
3989         mac->get_link_status = false;
3990
3991 out:
3992         *link_up = !mac->get_link_status;
3993         return ret_val;
3994 }
3995
3996 static void
3997 ixgbe_dev_setup_link_alarm_handler(void *param)
3998 {
3999         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4000         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4001         struct ixgbe_interrupt *intr =
4002                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4003         u32 speed;
4004         bool autoneg = false;
4005
4006         speed = hw->phy.autoneg_advertised;
4007         if (!speed)
4008                 ixgbe_get_link_capabilities(hw, &speed, &autoneg);
4009
4010         ixgbe_setup_link(hw, speed, true);
4011
4012         intr->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4013 }
4014
4015 /* return 0 means link status changed, -1 means not changed */
4016 int
4017 ixgbe_dev_link_update_share(struct rte_eth_dev *dev,
4018                             int wait_to_complete, int vf)
4019 {
4020         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4021         struct rte_eth_link link;
4022         ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
4023         struct ixgbe_interrupt *intr =
4024                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4025         int link_up;
4026         int diag;
4027         int wait = 1;
4028
4029         memset(&link, 0, sizeof(link));
4030         link.link_status = ETH_LINK_DOWN;
4031         link.link_speed = ETH_SPEED_NUM_NONE;
4032         link.link_duplex = ETH_LINK_HALF_DUPLEX;
4033         link.link_autoneg = ETH_LINK_AUTONEG;
4034
4035         hw->mac.get_link_status = true;
4036
4037         if (intr->flags & IXGBE_FLAG_NEED_LINK_CONFIG)
4038                 return rte_eth_linkstatus_set(dev, &link);
4039
4040         /* check if it needs to wait to complete, if lsc interrupt is enabled */
4041         if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
4042                 wait = 0;
4043
4044         if (vf)
4045                 diag = ixgbevf_check_link(hw, &link_speed, &link_up, wait);
4046         else
4047                 diag = ixgbe_check_link(hw, &link_speed, &link_up, wait);
4048
4049         if (diag != 0) {
4050                 link.link_speed = ETH_SPEED_NUM_100M;
4051                 link.link_duplex = ETH_LINK_FULL_DUPLEX;
4052                 return rte_eth_linkstatus_set(dev, &link);
4053         }
4054
4055         if (link_up == 0) {
4056                 if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4057                         intr->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
4058                         rte_eal_alarm_set(10,
4059                                 ixgbe_dev_setup_link_alarm_handler, dev);
4060                 }
4061                 return rte_eth_linkstatus_set(dev, &link);
4062         }
4063
4064         link.link_status = ETH_LINK_UP;
4065         link.link_duplex = ETH_LINK_FULL_DUPLEX;
4066
4067         switch (link_speed) {
4068         default:
4069         case IXGBE_LINK_SPEED_UNKNOWN:
4070                 if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
4071                         hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
4072                         link.link_speed = ETH_SPEED_NUM_10M;
4073                 else
4074                         link.link_speed = ETH_SPEED_NUM_100M;
4075                 link.link_duplex = ETH_LINK_FULL_DUPLEX;
4076                 break;
4077
4078         case IXGBE_LINK_SPEED_100_FULL:
4079                 link.link_speed = ETH_SPEED_NUM_100M;
4080                 break;
4081
4082         case IXGBE_LINK_SPEED_1GB_FULL:
4083                 link.link_speed = ETH_SPEED_NUM_1G;
4084                 break;
4085
4086         case IXGBE_LINK_SPEED_2_5GB_FULL:
4087                 link.link_speed = ETH_SPEED_NUM_2_5G;
4088                 break;
4089
4090         case IXGBE_LINK_SPEED_5GB_FULL:
4091                 link.link_speed = ETH_SPEED_NUM_5G;
4092                 break;
4093
4094         case IXGBE_LINK_SPEED_10GB_FULL:
4095                 link.link_speed = ETH_SPEED_NUM_10G;
4096                 break;
4097         }
4098
4099         return rte_eth_linkstatus_set(dev, &link);
4100 }
4101
4102 static int
4103 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4104 {
4105         return ixgbe_dev_link_update_share(dev, wait_to_complete, 0);
4106 }
4107
4108 static int
4109 ixgbevf_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4110 {
4111         return ixgbe_dev_link_update_share(dev, wait_to_complete, 1);
4112 }
4113
4114 static void
4115 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
4116 {
4117         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4118         uint32_t fctrl;
4119
4120         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4121         fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4122         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4123 }
4124
4125 static void
4126 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
4127 {
4128         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4129         uint32_t fctrl;
4130
4131         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4132         fctrl &= (~IXGBE_FCTRL_UPE);
4133         if (dev->data->all_multicast == 1)
4134                 fctrl |= IXGBE_FCTRL_MPE;
4135         else
4136                 fctrl &= (~IXGBE_FCTRL_MPE);
4137         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4138 }
4139
4140 static void
4141 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
4142 {
4143         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4144         uint32_t fctrl;
4145
4146         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4147         fctrl |= IXGBE_FCTRL_MPE;
4148         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4149 }
4150
4151 static void
4152 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
4153 {
4154         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4155         uint32_t fctrl;
4156
4157         if (dev->data->promiscuous == 1)
4158                 return; /* must remain in all_multicast mode */
4159
4160         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4161         fctrl &= (~IXGBE_FCTRL_MPE);
4162         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4163 }
4164
4165 /**
4166  * It clears the interrupt causes and enables the interrupt.
4167  * It will be called once only during nic initialized.
4168  *
4169  * @param dev
4170  *  Pointer to struct rte_eth_dev.
4171  * @param on
4172  *  Enable or Disable.
4173  *
4174  * @return
4175  *  - On success, zero.
4176  *  - On failure, a negative value.
4177  */
4178 static int
4179 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
4180 {
4181         struct ixgbe_interrupt *intr =
4182                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4183
4184         ixgbe_dev_link_status_print(dev);
4185         if (on)
4186                 intr->mask |= IXGBE_EICR_LSC;
4187         else
4188                 intr->mask &= ~IXGBE_EICR_LSC;
4189
4190         return 0;
4191 }
4192
4193 /**
4194  * It clears the interrupt causes and enables the interrupt.
4195  * It will be called once only during nic initialized.
4196  *
4197  * @param dev
4198  *  Pointer to struct rte_eth_dev.
4199  *
4200  * @return
4201  *  - On success, zero.
4202  *  - On failure, a negative value.
4203  */
4204 static int
4205 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
4206 {
4207         struct ixgbe_interrupt *intr =
4208                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4209
4210         intr->mask |= IXGBE_EICR_RTX_QUEUE;
4211
4212         return 0;
4213 }
4214
4215 /**
4216  * It clears the interrupt causes and enables the interrupt.
4217  * It will be called once only during nic initialized.
4218  *
4219  * @param dev
4220  *  Pointer to struct rte_eth_dev.
4221  *
4222  * @return
4223  *  - On success, zero.
4224  *  - On failure, a negative value.
4225  */
4226 static int
4227 ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev)
4228 {
4229         struct ixgbe_interrupt *intr =
4230                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4231
4232         intr->mask |= IXGBE_EICR_LINKSEC;
4233
4234         return 0;
4235 }
4236
4237 /*
4238  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
4239  *
4240  * @param dev
4241  *  Pointer to struct rte_eth_dev.
4242  *
4243  * @return
4244  *  - On success, zero.
4245  *  - On failure, a negative value.
4246  */
4247 static int
4248 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
4249 {
4250         uint32_t eicr;
4251         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4252         struct ixgbe_interrupt *intr =
4253                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4254
4255         /* clear all cause mask */
4256         ixgbe_disable_intr(hw);
4257
4258         /* read-on-clear nic registers here */
4259         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4260         PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
4261
4262         intr->flags = 0;
4263
4264         /* set flag for async link update */
4265         if (eicr & IXGBE_EICR_LSC)
4266                 intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4267
4268         if (eicr & IXGBE_EICR_MAILBOX)
4269                 intr->flags |= IXGBE_FLAG_MAILBOX;
4270
4271         if (eicr & IXGBE_EICR_LINKSEC)
4272                 intr->flags |= IXGBE_FLAG_MACSEC;
4273
4274         if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
4275             hw->phy.type == ixgbe_phy_x550em_ext_t &&
4276             (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
4277                 intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
4278
4279         return 0;
4280 }
4281
4282 /**
4283  * It gets and then prints the link status.
4284  *
4285  * @param dev
4286  *  Pointer to struct rte_eth_dev.
4287  *
4288  * @return
4289  *  - On success, zero.
4290  *  - On failure, a negative value.
4291  */
4292 static void
4293 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
4294 {
4295         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4296         struct rte_eth_link link;
4297
4298         rte_eth_linkstatus_get(dev, &link);
4299
4300         if (link.link_status) {
4301                 PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
4302                                         (int)(dev->data->port_id),
4303                                         (unsigned)link.link_speed,
4304                         link.link_duplex == ETH_LINK_FULL_DUPLEX ?
4305                                         "full-duplex" : "half-duplex");
4306         } else {
4307                 PMD_INIT_LOG(INFO, " Port %d: Link Down",
4308                                 (int)(dev->data->port_id));
4309         }
4310         PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
4311                                 pci_dev->addr.domain,
4312                                 pci_dev->addr.bus,
4313                                 pci_dev->addr.devid,
4314                                 pci_dev->addr.function);
4315 }
4316
4317 /*
4318  * It executes link_update after knowing an interrupt occurred.
4319  *
4320  * @param dev
4321  *  Pointer to struct rte_eth_dev.
4322  *
4323  * @return
4324  *  - On success, zero.
4325  *  - On failure, a negative value.
4326  */
4327 static int
4328 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
4329 {
4330         struct ixgbe_interrupt *intr =
4331                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4332         int64_t timeout;
4333         struct ixgbe_hw *hw =
4334                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4335
4336         PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
4337
4338         if (intr->flags & IXGBE_FLAG_MAILBOX) {
4339                 ixgbe_pf_mbx_process(dev);
4340                 intr->flags &= ~IXGBE_FLAG_MAILBOX;
4341         }
4342
4343         if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4344                 ixgbe_handle_lasi(hw);
4345                 intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4346         }
4347
4348         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4349                 struct rte_eth_link link;
4350
4351                 /* get the link status before link update, for predicting later */
4352                 rte_eth_linkstatus_get(dev, &link);
4353
4354                 ixgbe_dev_link_update(dev, 0);
4355
4356                 /* likely to up */
4357                 if (!link.link_status)
4358                         /* handle it 1 sec later, wait it being stable */
4359                         timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
4360                 /* likely to down */
4361                 else
4362                         /* handle it 4 sec later, wait it being stable */
4363                         timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
4364
4365                 ixgbe_dev_link_status_print(dev);
4366                 if (rte_eal_alarm_set(timeout * 1000,
4367                                       ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
4368                         PMD_DRV_LOG(ERR, "Error setting alarm");
4369                 else {
4370                         /* remember original mask */
4371                         intr->mask_original = intr->mask;
4372                         /* only disable lsc interrupt */
4373                         intr->mask &= ~IXGBE_EIMS_LSC;
4374                 }
4375         }
4376
4377         PMD_DRV_LOG(DEBUG, "enable intr immediately");
4378         ixgbe_enable_intr(dev);
4379
4380         return 0;
4381 }
4382
4383 /**
4384  * Interrupt handler which shall be registered for alarm callback for delayed
4385  * handling specific interrupt to wait for the stable nic state. As the
4386  * NIC interrupt state is not stable for ixgbe after link is just down,
4387  * it needs to wait 4 seconds to get the stable status.
4388  *
4389  * @param handle
4390  *  Pointer to interrupt handle.
4391  * @param param
4392  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4393  *
4394  * @return
4395  *  void
4396  */
4397 static void
4398 ixgbe_dev_interrupt_delayed_handler(void *param)
4399 {
4400         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4401         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4402         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
4403         struct ixgbe_interrupt *intr =
4404                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4405         struct ixgbe_hw *hw =
4406                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4407         uint32_t eicr;
4408
4409         ixgbe_disable_intr(hw);
4410
4411         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4412         if (eicr & IXGBE_EICR_MAILBOX)
4413                 ixgbe_pf_mbx_process(dev);
4414
4415         if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4416                 ixgbe_handle_lasi(hw);
4417                 intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4418         }
4419
4420         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4421                 ixgbe_dev_link_update(dev, 0);
4422                 intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4423                 ixgbe_dev_link_status_print(dev);
4424                 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC,
4425                                               NULL);
4426         }
4427
4428         if (intr->flags & IXGBE_FLAG_MACSEC) {
4429                 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_MACSEC,
4430                                               NULL);
4431                 intr->flags &= ~IXGBE_FLAG_MACSEC;
4432         }
4433
4434         /* restore original mask */
4435         intr->mask = intr->mask_original;
4436         intr->mask_original = 0;
4437
4438         PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
4439         ixgbe_enable_intr(dev);
4440         rte_intr_enable(intr_handle);
4441 }
4442
4443 /**
4444  * Interrupt handler triggered by NIC  for handling
4445  * specific interrupt.
4446  *
4447  * @param handle
4448  *  Pointer to interrupt handle.
4449  * @param param
4450  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4451  *
4452  * @return
4453  *  void
4454  */
4455 static void
4456 ixgbe_dev_interrupt_handler(void *param)
4457 {
4458         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4459
4460         ixgbe_dev_interrupt_get_status(dev);
4461         ixgbe_dev_interrupt_action(dev);
4462 }
4463
4464 static int
4465 ixgbe_dev_led_on(struct rte_eth_dev *dev)
4466 {
4467         struct ixgbe_hw *hw;
4468
4469         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4470         return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4471 }
4472
4473 static int
4474 ixgbe_dev_led_off(struct rte_eth_dev *dev)
4475 {
4476         struct ixgbe_hw *hw;
4477
4478         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4479         return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4480 }
4481
4482 static int
4483 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4484 {
4485         struct ixgbe_hw *hw;
4486         uint32_t mflcn_reg;
4487         uint32_t fccfg_reg;
4488         int rx_pause;
4489         int tx_pause;
4490
4491         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4492
4493         fc_conf->pause_time = hw->fc.pause_time;
4494         fc_conf->high_water = hw->fc.high_water[0];
4495         fc_conf->low_water = hw->fc.low_water[0];
4496         fc_conf->send_xon = hw->fc.send_xon;
4497         fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
4498
4499         /*
4500          * Return rx_pause status according to actual setting of
4501          * MFLCN register.
4502          */
4503         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4504         if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
4505                 rx_pause = 1;
4506         else
4507                 rx_pause = 0;
4508
4509         /*
4510          * Return tx_pause status according to actual setting of
4511          * FCCFG register.
4512          */
4513         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4514         if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
4515                 tx_pause = 1;
4516         else
4517                 tx_pause = 0;
4518
4519         if (rx_pause && tx_pause)
4520                 fc_conf->mode = RTE_FC_FULL;
4521         else if (rx_pause)
4522                 fc_conf->mode = RTE_FC_RX_PAUSE;
4523         else if (tx_pause)
4524                 fc_conf->mode = RTE_FC_TX_PAUSE;
4525         else
4526                 fc_conf->mode = RTE_FC_NONE;
4527
4528         return 0;
4529 }
4530
4531 static int
4532 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4533 {
4534         struct ixgbe_hw *hw;
4535         int err;
4536         uint32_t rx_buf_size;
4537         uint32_t max_high_water;
4538         uint32_t mflcn;
4539         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4540                 ixgbe_fc_none,
4541                 ixgbe_fc_rx_pause,
4542                 ixgbe_fc_tx_pause,
4543                 ixgbe_fc_full
4544         };
4545
4546         PMD_INIT_FUNC_TRACE();
4547
4548         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4549         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
4550         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4551
4552         /*
4553          * At least reserve one Ethernet frame for watermark
4554          * high_water/low_water in kilo bytes for ixgbe
4555          */
4556         max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4557         if ((fc_conf->high_water > max_high_water) ||
4558                 (fc_conf->high_water < fc_conf->low_water)) {
4559                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4560                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4561                 return -EINVAL;
4562         }
4563
4564         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
4565         hw->fc.pause_time     = fc_conf->pause_time;
4566         hw->fc.high_water[0]  = fc_conf->high_water;
4567         hw->fc.low_water[0]   = fc_conf->low_water;
4568         hw->fc.send_xon       = fc_conf->send_xon;
4569         hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
4570
4571         err = ixgbe_fc_enable(hw);
4572
4573         /* Not negotiated is not an error case */
4574         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED)) {
4575
4576                 /* check if we want to forward MAC frames - driver doesn't have native
4577                  * capability to do that, so we'll write the registers ourselves */
4578
4579                 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4580
4581                 /* set or clear MFLCN.PMCF bit depending on configuration */
4582                 if (fc_conf->mac_ctrl_frame_fwd != 0)
4583                         mflcn |= IXGBE_MFLCN_PMCF;
4584                 else
4585                         mflcn &= ~IXGBE_MFLCN_PMCF;
4586
4587                 IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
4588                 IXGBE_WRITE_FLUSH(hw);
4589
4590                 return 0;
4591         }
4592
4593         PMD_INIT_LOG(ERR, "ixgbe_fc_enable = 0x%x", err);
4594         return -EIO;
4595 }
4596
4597 /**
4598  *  ixgbe_pfc_enable_generic - Enable flow control
4599  *  @hw: pointer to hardware structure
4600  *  @tc_num: traffic class number
4601  *  Enable flow control according to the current settings.
4602  */
4603 static int
4604 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
4605 {
4606         int ret_val = 0;
4607         uint32_t mflcn_reg, fccfg_reg;
4608         uint32_t reg;
4609         uint32_t fcrtl, fcrth;
4610         uint8_t i;
4611         uint8_t nb_rx_en;
4612
4613         /* Validate the water mark configuration */
4614         if (!hw->fc.pause_time) {
4615                 ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4616                 goto out;
4617         }
4618
4619         /* Low water mark of zero causes XOFF floods */
4620         if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
4621                  /* High/Low water can not be 0 */
4622                 if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_water[tc_num])) {
4623                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4624                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4625                         goto out;
4626                 }
4627
4628                 if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
4629                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4630                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4631                         goto out;
4632                 }
4633         }
4634         /* Negotiate the fc mode to use */
4635         ixgbe_fc_autoneg(hw);
4636
4637         /* Disable any previous flow control settings */
4638         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4639         mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
4640
4641         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4642         fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
4643
4644         switch (hw->fc.current_mode) {
4645         case ixgbe_fc_none:
4646                 /*
4647                  * If the count of enabled RX Priority Flow control >1,
4648                  * and the TX pause can not be disabled
4649                  */
4650                 nb_rx_en = 0;
4651                 for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4652                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4653                         if (reg & IXGBE_FCRTH_FCEN)
4654                                 nb_rx_en++;
4655                 }
4656                 if (nb_rx_en > 1)
4657                         fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4658                 break;
4659         case ixgbe_fc_rx_pause:
4660                 /*
4661                  * Rx Flow control is enabled and Tx Flow control is
4662                  * disabled by software override. Since there really
4663                  * isn't a way to advertise that we are capable of RX
4664                  * Pause ONLY, we will advertise that we support both
4665                  * symmetric and asymmetric Rx PAUSE.  Later, we will
4666                  * disable the adapter's ability to send PAUSE frames.
4667                  */
4668                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
4669                 /*
4670                  * If the count of enabled RX Priority Flow control >1,
4671                  * and the TX pause can not be disabled
4672                  */
4673                 nb_rx_en = 0;
4674                 for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4675                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4676                         if (reg & IXGBE_FCRTH_FCEN)
4677                                 nb_rx_en++;
4678                 }
4679                 if (nb_rx_en > 1)
4680                         fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4681                 break;
4682         case ixgbe_fc_tx_pause:
4683                 /*
4684                  * Tx Flow control is enabled, and Rx Flow control is
4685                  * disabled by software override.
4686                  */
4687                 fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4688                 break;
4689         case ixgbe_fc_full:
4690                 /* Flow control (both Rx and Tx) is enabled by SW override. */
4691                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
4692                 fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4693                 break;
4694         default:
4695                 PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
4696                 ret_val = IXGBE_ERR_CONFIG;
4697                 goto out;
4698         }
4699
4700         /* Set 802.3x based flow control settings. */
4701         mflcn_reg |= IXGBE_MFLCN_DPF;
4702         IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
4703         IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
4704
4705         /* Set up and enable Rx high/low water mark thresholds, enable XON. */
4706         if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
4707                 hw->fc.high_water[tc_num]) {
4708                 fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
4709                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
4710                 fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
4711         } else {
4712                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
4713                 /*
4714                  * In order to prevent Tx hangs when the internal Tx
4715                  * switch is enabled we must set the high water mark
4716                  * to the maximum FCRTH value.  This allows the Tx
4717                  * switch to function even under heavy Rx workloads.
4718                  */
4719                 fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
4720         }
4721         IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
4722
4723         /* Configure pause time (2 TCs per register) */
4724         reg = hw->fc.pause_time * 0x00010001;
4725         for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
4726                 IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
4727
4728         /* Configure flow control refresh threshold value */
4729         IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
4730
4731 out:
4732         return ret_val;
4733 }
4734
4735 static int
4736 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
4737 {
4738         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4739         int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
4740
4741         if (hw->mac.type != ixgbe_mac_82598EB) {
4742                 ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
4743         }
4744         return ret_val;
4745 }
4746
4747 static int
4748 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
4749 {
4750         int err;
4751         uint32_t rx_buf_size;
4752         uint32_t max_high_water;
4753         uint8_t tc_num;
4754         uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
4755         struct ixgbe_hw *hw =
4756                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4757         struct ixgbe_dcb_config *dcb_config =
4758                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
4759
4760         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4761                 ixgbe_fc_none,
4762                 ixgbe_fc_rx_pause,
4763                 ixgbe_fc_tx_pause,
4764                 ixgbe_fc_full
4765         };
4766
4767         PMD_INIT_FUNC_TRACE();
4768
4769         ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
4770         tc_num = map[pfc_conf->priority];
4771         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
4772         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4773         /*
4774          * At least reserve one Ethernet frame for watermark
4775          * high_water/low_water in kilo bytes for ixgbe
4776          */
4777         max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4778         if ((pfc_conf->fc.high_water > max_high_water) ||
4779             (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
4780                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4781                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4782                 return -EINVAL;
4783         }
4784
4785         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
4786         hw->fc.pause_time = pfc_conf->fc.pause_time;
4787         hw->fc.send_xon = pfc_conf->fc.send_xon;
4788         hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
4789         hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
4790
4791         err = ixgbe_dcb_pfc_enable(dev, tc_num);
4792
4793         /* Not negotiated is not an error case */
4794         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
4795                 return 0;
4796
4797         PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
4798         return -EIO;
4799 }
4800
4801 static int
4802 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
4803                           struct rte_eth_rss_reta_entry64 *reta_conf,
4804                           uint16_t reta_size)
4805 {
4806         uint16_t i, sp_reta_size;
4807         uint8_t j, mask;
4808         uint32_t reta, r;
4809         uint16_t idx, shift;
4810         struct ixgbe_adapter *adapter =
4811                 (struct ixgbe_adapter *)dev->data->dev_private;
4812         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4813         uint32_t reta_reg;
4814
4815         PMD_INIT_FUNC_TRACE();
4816
4817         if (!ixgbe_rss_update_sp(hw->mac.type)) {
4818                 PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
4819                         "NIC.");
4820                 return -ENOTSUP;
4821         }
4822
4823         sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
4824         if (reta_size != sp_reta_size) {
4825                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
4826                         "(%d) doesn't match the number hardware can supported "
4827                         "(%d)", reta_size, sp_reta_size);
4828                 return -EINVAL;
4829         }
4830
4831         for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
4832                 idx = i / RTE_RETA_GROUP_SIZE;
4833                 shift = i % RTE_RETA_GROUP_SIZE;
4834                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
4835                                                 IXGBE_4_BIT_MASK);
4836                 if (!mask)
4837                         continue;
4838                 reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
4839                 if (mask == IXGBE_4_BIT_MASK)
4840                         r = 0;
4841                 else
4842                         r = IXGBE_READ_REG(hw, reta_reg);
4843                 for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
4844                         if (mask & (0x1 << j))
4845                                 reta |= reta_conf[idx].reta[shift + j] <<
4846                                                         (CHAR_BIT * j);
4847                         else
4848                                 reta |= r & (IXGBE_8_BIT_MASK <<
4849                                                 (CHAR_BIT * j));
4850                 }
4851                 IXGBE_WRITE_REG(hw, reta_reg, reta);
4852         }
4853         adapter->rss_reta_updated = 1;
4854
4855         return 0;
4856 }
4857
4858 static int
4859 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
4860                          struct rte_eth_rss_reta_entry64 *reta_conf,
4861                          uint16_t reta_size)
4862 {
4863         uint16_t i, sp_reta_size;
4864         uint8_t j, mask;
4865         uint32_t reta;
4866         uint16_t idx, shift;
4867         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4868         uint32_t reta_reg;
4869
4870         PMD_INIT_FUNC_TRACE();
4871         sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
4872         if (reta_size != sp_reta_size) {
4873                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
4874                         "(%d) doesn't match the number hardware can supported "
4875                         "(%d)", reta_size, sp_reta_size);
4876                 return -EINVAL;
4877         }
4878
4879         for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
4880                 idx = i / RTE_RETA_GROUP_SIZE;
4881                 shift = i % RTE_RETA_GROUP_SIZE;
4882                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
4883                                                 IXGBE_4_BIT_MASK);
4884                 if (!mask)
4885                         continue;
4886
4887                 reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
4888                 reta = IXGBE_READ_REG(hw, reta_reg);
4889                 for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
4890                         if (mask & (0x1 << j))
4891                                 reta_conf[idx].reta[shift + j] =
4892                                         ((reta >> (CHAR_BIT * j)) &
4893                                                 IXGBE_8_BIT_MASK);
4894                 }
4895         }
4896
4897         return 0;
4898 }
4899
4900 static int
4901 ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
4902                                 uint32_t index, uint32_t pool)
4903 {
4904         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4905         uint32_t enable_addr = 1;
4906
4907         return ixgbe_set_rar(hw, index, mac_addr->addr_bytes,
4908                              pool, enable_addr);
4909 }
4910
4911 static void
4912 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
4913 {
4914         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4915
4916         ixgbe_clear_rar(hw, index);
4917 }
4918
4919 static int
4920 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
4921 {
4922         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4923
4924         ixgbe_remove_rar(dev, 0);
4925         ixgbe_add_rar(dev, addr, 0, pci_dev->max_vfs);
4926
4927         return 0;
4928 }
4929
4930 static bool
4931 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
4932 {
4933         if (strcmp(dev->device->driver->name, drv->driver.name))
4934                 return false;
4935
4936         return true;
4937 }
4938
4939 bool
4940 is_ixgbe_supported(struct rte_eth_dev *dev)
4941 {
4942         return is_device_supported(dev, &rte_ixgbe_pmd);
4943 }
4944
4945 static int
4946 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4947 {
4948         uint32_t hlreg0;
4949         uint32_t maxfrs;
4950         struct ixgbe_hw *hw;
4951         struct rte_eth_dev_info dev_info;
4952         uint32_t frame_size = mtu + IXGBE_ETH_OVERHEAD;
4953         struct rte_eth_dev_data *dev_data = dev->data;
4954
4955         ixgbe_dev_info_get(dev, &dev_info);
4956
4957         /* check that mtu is within the allowed range */
4958         if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
4959                 return -EINVAL;
4960
4961         /* If device is started, refuse mtu that requires the support of
4962          * scattered packets when this feature has not been enabled before.
4963          */
4964         if (dev_data->dev_started && !dev_data->scattered_rx &&
4965             (frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
4966              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
4967                 PMD_INIT_LOG(ERR, "Stop port first.");
4968                 return -EINVAL;
4969         }
4970
4971         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4972         hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4973
4974         /* switch to jumbo mode if needed */
4975         if (frame_size > ETHER_MAX_LEN) {
4976                 dev->data->dev_conf.rxmode.offloads |=
4977                         DEV_RX_OFFLOAD_JUMBO_FRAME;
4978                 hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4979         } else {
4980                 dev->data->dev_conf.rxmode.offloads &=
4981                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
4982                 hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
4983         }
4984         IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4985
4986         /* update max frame size */
4987         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4988
4989         maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
4990         maxfrs &= 0x0000FFFF;
4991         maxfrs |= (dev->data->dev_conf.rxmode.max_rx_pkt_len << 16);
4992         IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
4993
4994         return 0;
4995 }
4996
4997 /*
4998  * Virtual Function operations
4999  */
5000 static void
5001 ixgbevf_intr_disable(struct rte_eth_dev *dev)
5002 {
5003         struct ixgbe_interrupt *intr =
5004                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5005         struct ixgbe_hw *hw =
5006                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5007
5008         PMD_INIT_FUNC_TRACE();
5009
5010         /* Clear interrupt mask to stop from interrupts being generated */
5011         IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
5012
5013         IXGBE_WRITE_FLUSH(hw);
5014
5015         /* Clear mask value. */
5016         intr->mask = 0;
5017 }
5018
5019 static void
5020 ixgbevf_intr_enable(struct rte_eth_dev *dev)
5021 {
5022         struct ixgbe_interrupt *intr =
5023                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5024         struct ixgbe_hw *hw =
5025                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5026
5027         PMD_INIT_FUNC_TRACE();
5028
5029         /* VF enable interrupt autoclean */
5030         IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
5031         IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
5032         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
5033
5034         IXGBE_WRITE_FLUSH(hw);
5035
5036         /* Save IXGBE_VTEIMS value to mask. */
5037         intr->mask = IXGBE_VF_IRQ_ENABLE_MASK;
5038 }
5039
5040 static int
5041 ixgbevf_dev_configure(struct rte_eth_dev *dev)
5042 {
5043         struct rte_eth_conf *conf = &dev->data->dev_conf;
5044         struct ixgbe_adapter *adapter =
5045                         (struct ixgbe_adapter *)dev->data->dev_private;
5046
5047         PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
5048                      dev->data->port_id);
5049
5050         /*
5051          * VF has no ability to enable/disable HW CRC
5052          * Keep the persistent behavior the same as Host PF
5053          */
5054 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
5055         if (conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
5056                 PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
5057                 conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_KEEP_CRC;
5058         }
5059 #else
5060         if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC)) {
5061                 PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
5062                 conf->rxmode.offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
5063         }
5064 #endif
5065
5066         /*
5067          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
5068          * allocation or vector Rx preconditions we will reset it.
5069          */
5070         adapter->rx_bulk_alloc_allowed = true;
5071         adapter->rx_vec_allowed = true;
5072
5073         return 0;
5074 }
5075
5076 static int
5077 ixgbevf_dev_start(struct rte_eth_dev *dev)
5078 {
5079         struct ixgbe_hw *hw =
5080                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5081         uint32_t intr_vector = 0;
5082         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5083         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5084
5085         int err, mask = 0;
5086
5087         PMD_INIT_FUNC_TRACE();
5088
5089         /* Stop the link setup handler before resetting the HW. */
5090         rte_eal_alarm_cancel(ixgbe_dev_setup_link_alarm_handler, dev);
5091
5092         err = hw->mac.ops.reset_hw(hw);
5093         if (err) {
5094                 PMD_INIT_LOG(ERR, "Unable to reset vf hardware (%d)", err);
5095                 return err;
5096         }
5097         hw->mac.get_link_status = true;
5098
5099         /* negotiate mailbox API version to use with the PF. */
5100         ixgbevf_negotiate_api(hw);
5101
5102         ixgbevf_dev_tx_init(dev);
5103
5104         /* This can fail when allocating mbufs for descriptor rings */
5105         err = ixgbevf_dev_rx_init(dev);
5106         if (err) {
5107                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
5108                 ixgbe_dev_clear_queues(dev);
5109                 return err;
5110         }
5111
5112         /* Set vfta */
5113         ixgbevf_set_vfta_all(dev, 1);
5114
5115         /* Set HW strip */
5116         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
5117                 ETH_VLAN_EXTEND_MASK;
5118         err = ixgbevf_vlan_offload_config(dev, mask);
5119         if (err) {
5120                 PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
5121                 ixgbe_dev_clear_queues(dev);
5122                 return err;
5123         }
5124
5125         ixgbevf_dev_rxtx_start(dev);
5126
5127         /* check and configure queue intr-vector mapping */
5128         if (rte_intr_cap_multiple(intr_handle) &&
5129             dev->data->dev_conf.intr_conf.rxq) {
5130                 /* According to datasheet, only vector 0/1/2 can be used,
5131                  * now only one vector is used for Rx queue
5132                  */
5133                 intr_vector = 1;
5134                 if (rte_intr_efd_enable(intr_handle, intr_vector))
5135                         return -1;
5136         }
5137
5138         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
5139                 intr_handle->intr_vec =
5140                         rte_zmalloc("intr_vec",
5141                                     dev->data->nb_rx_queues * sizeof(int), 0);
5142                 if (intr_handle->intr_vec == NULL) {
5143                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
5144                                      " intr_vec", dev->data->nb_rx_queues);
5145                         return -ENOMEM;
5146                 }
5147         }
5148         ixgbevf_configure_msix(dev);
5149
5150         /* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
5151          * is mapped to VFIO vector 0 in eth_ixgbevf_dev_init( ).
5152          * If previous VFIO interrupt mapping setting in eth_ixgbevf_dev_init( )
5153          * is not cleared, it will fail when following rte_intr_enable( ) tries
5154          * to map Rx queue interrupt to other VFIO vectors.
5155          * So clear uio/vfio intr/evevnfd first to avoid failure.
5156          */
5157         rte_intr_disable(intr_handle);
5158
5159         rte_intr_enable(intr_handle);
5160
5161         /* Re-enable interrupt for VF */
5162         ixgbevf_intr_enable(dev);
5163
5164         /*
5165          * Update link status right before return, because it may
5166          * start link configuration process in a separate thread.
5167          */
5168         ixgbevf_dev_link_update(dev, 0);
5169
5170         return 0;
5171 }
5172
5173 static void
5174 ixgbevf_dev_stop(struct rte_eth_dev *dev)
5175 {
5176         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5177         struct ixgbe_adapter *adapter =
5178                 (struct ixgbe_adapter *)dev->data->dev_private;
5179         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5180         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5181
5182         PMD_INIT_FUNC_TRACE();
5183
5184         rte_eal_alarm_cancel(ixgbe_dev_setup_link_alarm_handler, dev);
5185
5186         ixgbevf_intr_disable(dev);
5187
5188         hw->adapter_stopped = 1;
5189         ixgbe_stop_adapter(hw);
5190
5191         /*
5192           * Clear what we set, but we still keep shadow_vfta to
5193           * restore after device starts
5194           */
5195         ixgbevf_set_vfta_all(dev, 0);
5196
5197         /* Clear stored conf */
5198         dev->data->scattered_rx = 0;
5199
5200         ixgbe_dev_clear_queues(dev);
5201
5202         /* Clean datapath event and queue/vec mapping */
5203         rte_intr_efd_disable(intr_handle);
5204         if (intr_handle->intr_vec != NULL) {
5205                 rte_free(intr_handle->intr_vec);
5206                 intr_handle->intr_vec = NULL;
5207         }
5208
5209         adapter->rss_reta_updated = 0;
5210 }
5211
5212 static void
5213 ixgbevf_dev_close(struct rte_eth_dev *dev)
5214 {
5215         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5216
5217         PMD_INIT_FUNC_TRACE();
5218
5219         ixgbe_reset_hw(hw);
5220
5221         ixgbevf_dev_stop(dev);
5222
5223         ixgbe_dev_free_queues(dev);
5224
5225         /**
5226          * Remove the VF MAC address ro ensure
5227          * that the VF traffic goes to the PF
5228          * after stop, close and detach of the VF
5229          **/
5230         ixgbevf_remove_mac_addr(dev, 0);
5231 }
5232
5233 /*
5234  * Reset VF device
5235  */
5236 static int
5237 ixgbevf_dev_reset(struct rte_eth_dev *dev)
5238 {
5239         int ret;
5240
5241         ret = eth_ixgbevf_dev_uninit(dev);
5242         if (ret)
5243                 return ret;
5244
5245         ret = eth_ixgbevf_dev_init(dev);
5246
5247         return ret;
5248 }
5249
5250 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
5251 {
5252         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5253         struct ixgbe_vfta *shadow_vfta =
5254                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5255         int i = 0, j = 0, vfta = 0, mask = 1;
5256
5257         for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
5258                 vfta = shadow_vfta->vfta[i];
5259                 if (vfta) {
5260                         mask = 1;
5261                         for (j = 0; j < 32; j++) {
5262                                 if (vfta & mask)
5263                                         ixgbe_set_vfta(hw, (i<<5)+j, 0,
5264                                                        on, false);
5265                                 mask <<= 1;
5266                         }
5267                 }
5268         }
5269
5270 }
5271
5272 static int
5273 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
5274 {
5275         struct ixgbe_hw *hw =
5276                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5277         struct ixgbe_vfta *shadow_vfta =
5278                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5279         uint32_t vid_idx = 0;
5280         uint32_t vid_bit = 0;
5281         int ret = 0;
5282
5283         PMD_INIT_FUNC_TRACE();
5284
5285         /* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
5286         ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
5287         if (ret) {
5288                 PMD_INIT_LOG(ERR, "Unable to set VF vlan");
5289                 return ret;
5290         }
5291         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
5292         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
5293
5294         /* Save what we set and retore it after device reset */
5295         if (on)
5296                 shadow_vfta->vfta[vid_idx] |= vid_bit;
5297         else
5298                 shadow_vfta->vfta[vid_idx] &= ~vid_bit;
5299
5300         return 0;
5301 }
5302
5303 static void
5304 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
5305 {
5306         struct ixgbe_hw *hw =
5307                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5308         uint32_t ctrl;
5309
5310         PMD_INIT_FUNC_TRACE();
5311
5312         if (queue >= hw->mac.max_rx_queues)
5313                 return;
5314
5315         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
5316         if (on)
5317                 ctrl |= IXGBE_RXDCTL_VME;
5318         else
5319                 ctrl &= ~IXGBE_RXDCTL_VME;
5320         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
5321
5322         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
5323 }
5324
5325 static int
5326 ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask)
5327 {
5328         struct ixgbe_rx_queue *rxq;
5329         uint16_t i;
5330         int on = 0;
5331
5332         /* VF function only support hw strip feature, others are not support */
5333         if (mask & ETH_VLAN_STRIP_MASK) {
5334                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
5335                         rxq = dev->data->rx_queues[i];
5336                         on = !!(rxq->offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
5337                         ixgbevf_vlan_strip_queue_set(dev, i, on);
5338                 }
5339         }
5340
5341         return 0;
5342 }
5343
5344 static int
5345 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
5346 {
5347         ixgbe_config_vlan_strip_on_all_queues(dev, mask);
5348
5349         ixgbevf_vlan_offload_config(dev, mask);
5350
5351         return 0;
5352 }
5353
5354 int
5355 ixgbe_vt_check(struct ixgbe_hw *hw)
5356 {
5357         uint32_t reg_val;
5358
5359         /* if Virtualization Technology is enabled */
5360         reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
5361         if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
5362                 PMD_INIT_LOG(ERR, "VT must be enabled for this setting");
5363                 return -1;
5364         }
5365
5366         return 0;
5367 }
5368
5369 static uint32_t
5370 ixgbe_uta_vector(struct ixgbe_hw *hw, struct ether_addr *uc_addr)
5371 {
5372         uint32_t vector = 0;
5373
5374         switch (hw->mac.mc_filter_type) {
5375         case 0:   /* use bits [47:36] of the address */
5376                 vector = ((uc_addr->addr_bytes[4] >> 4) |
5377                         (((uint16_t)uc_addr->addr_bytes[5]) << 4));
5378                 break;
5379         case 1:   /* use bits [46:35] of the address */
5380                 vector = ((uc_addr->addr_bytes[4] >> 3) |
5381                         (((uint16_t)uc_addr->addr_bytes[5]) << 5));
5382                 break;
5383         case 2:   /* use bits [45:34] of the address */
5384                 vector = ((uc_addr->addr_bytes[4] >> 2) |
5385                         (((uint16_t)uc_addr->addr_bytes[5]) << 6));
5386                 break;
5387         case 3:   /* use bits [43:32] of the address */
5388                 vector = ((uc_addr->addr_bytes[4]) |
5389                         (((uint16_t)uc_addr->addr_bytes[5]) << 8));
5390                 break;
5391         default:  /* Invalid mc_filter_type */
5392                 break;
5393         }
5394
5395         /* vector can only be 12-bits or boundary will be exceeded */
5396         vector &= 0xFFF;
5397         return vector;
5398 }
5399
5400 static int
5401 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
5402                         uint8_t on)
5403 {
5404         uint32_t vector;
5405         uint32_t uta_idx;
5406         uint32_t reg_val;
5407         uint32_t uta_shift;
5408         uint32_t rc;
5409         const uint32_t ixgbe_uta_idx_mask = 0x7F;
5410         const uint32_t ixgbe_uta_bit_shift = 5;
5411         const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
5412         const uint32_t bit1 = 0x1;
5413
5414         struct ixgbe_hw *hw =
5415                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5416         struct ixgbe_uta_info *uta_info =
5417                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5418
5419         /* The UTA table only exists on 82599 hardware and newer */
5420         if (hw->mac.type < ixgbe_mac_82599EB)
5421                 return -ENOTSUP;
5422
5423         vector = ixgbe_uta_vector(hw, mac_addr);
5424         uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
5425         uta_shift = vector & ixgbe_uta_bit_mask;
5426
5427         rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
5428         if (rc == on)
5429                 return 0;
5430
5431         reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
5432         if (on) {
5433                 uta_info->uta_in_use++;
5434                 reg_val |= (bit1 << uta_shift);
5435                 uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
5436         } else {
5437                 uta_info->uta_in_use--;
5438                 reg_val &= ~(bit1 << uta_shift);
5439                 uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
5440         }
5441
5442         IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
5443
5444         if (uta_info->uta_in_use > 0)
5445                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
5446                                 IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
5447         else
5448                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
5449
5450         return 0;
5451 }
5452
5453 static int
5454 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
5455 {
5456         int i;
5457         struct ixgbe_hw *hw =
5458                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5459         struct ixgbe_uta_info *uta_info =
5460                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5461
5462         /* The UTA table only exists on 82599 hardware and newer */
5463         if (hw->mac.type < ixgbe_mac_82599EB)
5464                 return -ENOTSUP;
5465
5466         if (on) {
5467                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5468                         uta_info->uta_shadow[i] = ~0;
5469                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
5470                 }
5471         } else {
5472                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5473                         uta_info->uta_shadow[i] = 0;
5474                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
5475                 }
5476         }
5477         return 0;
5478
5479 }
5480
5481 uint32_t
5482 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
5483 {
5484         uint32_t new_val = orig_val;
5485
5486         if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG)
5487                 new_val |= IXGBE_VMOLR_AUPE;
5488         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC)
5489                 new_val |= IXGBE_VMOLR_ROMPE;
5490         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
5491                 new_val |= IXGBE_VMOLR_ROPE;
5492         if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
5493                 new_val |= IXGBE_VMOLR_BAM;
5494         if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
5495                 new_val |= IXGBE_VMOLR_MPE;
5496
5497         return new_val;
5498 }
5499
5500 #define IXGBE_MRCTL_VPME  0x01 /* Virtual Pool Mirroring. */
5501 #define IXGBE_MRCTL_UPME  0x02 /* Uplink Port Mirroring. */
5502 #define IXGBE_MRCTL_DPME  0x04 /* Downlink Port Mirroring. */
5503 #define IXGBE_MRCTL_VLME  0x08 /* VLAN Mirroring. */
5504 #define IXGBE_INVALID_MIRROR_TYPE(mirror_type) \
5505         ((mirror_type) & ~(uint8_t)(ETH_MIRROR_VIRTUAL_POOL_UP | \
5506         ETH_MIRROR_UPLINK_PORT | ETH_MIRROR_DOWNLINK_PORT | ETH_MIRROR_VLAN))
5507
5508 static int
5509 ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
5510                       struct rte_eth_mirror_conf *mirror_conf,
5511                       uint8_t rule_id, uint8_t on)
5512 {
5513         uint32_t mr_ctl, vlvf;
5514         uint32_t mp_lsb = 0;
5515         uint32_t mv_msb = 0;
5516         uint32_t mv_lsb = 0;
5517         uint32_t mp_msb = 0;
5518         uint8_t i = 0;
5519         int reg_index = 0;
5520         uint64_t vlan_mask = 0;
5521
5522         const uint8_t pool_mask_offset = 32;
5523         const uint8_t vlan_mask_offset = 32;
5524         const uint8_t dst_pool_offset = 8;
5525         const uint8_t rule_mr_offset  = 4;
5526         const uint8_t mirror_rule_mask = 0x0F;
5527
5528         struct ixgbe_mirror_info *mr_info =
5529                         (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5530         struct ixgbe_hw *hw =
5531                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5532         uint8_t mirror_type = 0;
5533
5534         if (ixgbe_vt_check(hw) < 0)
5535                 return -ENOTSUP;
5536
5537         if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5538                 return -EINVAL;
5539
5540         if (IXGBE_INVALID_MIRROR_TYPE(mirror_conf->rule_type)) {
5541                 PMD_DRV_LOG(ERR, "unsupported mirror type 0x%x.",
5542                             mirror_conf->rule_type);
5543                 return -EINVAL;
5544         }
5545
5546         if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5547                 mirror_type |= IXGBE_MRCTL_VLME;
5548                 /* Check if vlan id is valid and find conresponding VLAN ID
5549                  * index in VLVF
5550                  */
5551                 for (i = 0; i < IXGBE_VLVF_ENTRIES; i++) {
5552                         if (mirror_conf->vlan.vlan_mask & (1ULL << i)) {
5553                                 /* search vlan id related pool vlan filter
5554                                  * index
5555                                  */
5556                                 reg_index = ixgbe_find_vlvf_slot(
5557                                                 hw,
5558                                                 mirror_conf->vlan.vlan_id[i],
5559                                                 false);
5560                                 if (reg_index < 0)
5561                                         return -EINVAL;
5562                                 vlvf = IXGBE_READ_REG(hw,
5563                                                       IXGBE_VLVF(reg_index));
5564                                 if ((vlvf & IXGBE_VLVF_VIEN) &&
5565                                     ((vlvf & IXGBE_VLVF_VLANID_MASK) ==
5566                                       mirror_conf->vlan.vlan_id[i]))
5567                                         vlan_mask |= (1ULL << reg_index);
5568                                 else
5569                                         return -EINVAL;
5570                         }
5571                 }
5572
5573                 if (on) {
5574                         mv_lsb = vlan_mask & 0xFFFFFFFF;
5575                         mv_msb = vlan_mask >> vlan_mask_offset;
5576
5577                         mr_info->mr_conf[rule_id].vlan.vlan_mask =
5578                                                 mirror_conf->vlan.vlan_mask;
5579                         for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++) {
5580                                 if (mirror_conf->vlan.vlan_mask & (1ULL << i))
5581                                         mr_info->mr_conf[rule_id].vlan.vlan_id[i] =
5582                                                 mirror_conf->vlan.vlan_id[i];
5583                         }
5584                 } else {
5585                         mv_lsb = 0;
5586                         mv_msb = 0;
5587                         mr_info->mr_conf[rule_id].vlan.vlan_mask = 0;
5588                         for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++)
5589                                 mr_info->mr_conf[rule_id].vlan.vlan_id[i] = 0;
5590                 }
5591         }
5592
5593         /**
5594          * if enable pool mirror, write related pool mask register,if disable
5595          * pool mirror, clear PFMRVM register
5596          */
5597         if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5598                 mirror_type |= IXGBE_MRCTL_VPME;
5599                 if (on) {
5600                         mp_lsb = mirror_conf->pool_mask & 0xFFFFFFFF;
5601                         mp_msb = mirror_conf->pool_mask >> pool_mask_offset;
5602                         mr_info->mr_conf[rule_id].pool_mask =
5603                                         mirror_conf->pool_mask;
5604
5605                 } else {
5606                         mp_lsb = 0;
5607                         mp_msb = 0;
5608                         mr_info->mr_conf[rule_id].pool_mask = 0;
5609                 }
5610         }
5611         if (mirror_conf->rule_type & ETH_MIRROR_UPLINK_PORT)
5612                 mirror_type |= IXGBE_MRCTL_UPME;
5613         if (mirror_conf->rule_type & ETH_MIRROR_DOWNLINK_PORT)
5614                 mirror_type |= IXGBE_MRCTL_DPME;
5615
5616         /* read  mirror control register and recalculate it */
5617         mr_ctl = IXGBE_READ_REG(hw, IXGBE_MRCTL(rule_id));
5618
5619         if (on) {
5620                 mr_ctl |= mirror_type;
5621                 mr_ctl &= mirror_rule_mask;
5622                 mr_ctl |= mirror_conf->dst_pool << dst_pool_offset;
5623         } else {
5624                 mr_ctl &= ~(mirror_conf->rule_type & mirror_rule_mask);
5625         }
5626
5627         mr_info->mr_conf[rule_id].rule_type = mirror_conf->rule_type;
5628         mr_info->mr_conf[rule_id].dst_pool = mirror_conf->dst_pool;
5629
5630         /* write mirrror control  register */
5631         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5632
5633         /* write pool mirrror control  register */
5634         if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5635                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), mp_lsb);
5636                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset),
5637                                 mp_msb);
5638         }
5639         /* write VLAN mirrror control  register */
5640         if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5641                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), mv_lsb);
5642                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset),
5643                                 mv_msb);
5644         }
5645
5646         return 0;
5647 }
5648
5649 static int
5650 ixgbe_mirror_rule_reset(struct rte_eth_dev *dev, uint8_t rule_id)
5651 {
5652         int mr_ctl = 0;
5653         uint32_t lsb_val = 0;
5654         uint32_t msb_val = 0;
5655         const uint8_t rule_mr_offset = 4;
5656
5657         struct ixgbe_hw *hw =
5658                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5659         struct ixgbe_mirror_info *mr_info =
5660                 (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5661
5662         if (ixgbe_vt_check(hw) < 0)
5663                 return -ENOTSUP;
5664
5665         if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5666                 return -EINVAL;
5667
5668         memset(&mr_info->mr_conf[rule_id], 0,
5669                sizeof(struct rte_eth_mirror_conf));
5670
5671         /* clear PFVMCTL register */
5672         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5673
5674         /* clear pool mask register */
5675         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), lsb_val);
5676         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset), msb_val);
5677
5678         /* clear vlan mask register */
5679         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), lsb_val);
5680         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset), msb_val);
5681
5682         return 0;
5683 }
5684
5685 static int
5686 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5687 {
5688         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5689         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5690         struct ixgbe_interrupt *intr =
5691                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5692         struct ixgbe_hw *hw =
5693                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5694         uint32_t vec = IXGBE_MISC_VEC_ID;
5695
5696         if (rte_intr_allow_others(intr_handle))
5697                 vec = IXGBE_RX_VEC_START;
5698         intr->mask |= (1 << vec);
5699         RTE_SET_USED(queue_id);
5700         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5701
5702         rte_intr_enable(intr_handle);
5703
5704         return 0;
5705 }
5706
5707 static int
5708 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5709 {
5710         struct ixgbe_interrupt *intr =
5711                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5712         struct ixgbe_hw *hw =
5713                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5714         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5715         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5716         uint32_t vec = IXGBE_MISC_VEC_ID;
5717
5718         if (rte_intr_allow_others(intr_handle))
5719                 vec = IXGBE_RX_VEC_START;
5720         intr->mask &= ~(1 << vec);
5721         RTE_SET_USED(queue_id);
5722         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5723
5724         return 0;
5725 }
5726
5727 static int
5728 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5729 {
5730         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5731         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5732         uint32_t mask;
5733         struct ixgbe_hw *hw =
5734                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5735         struct ixgbe_interrupt *intr =
5736                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5737
5738         if (queue_id < 16) {
5739                 ixgbe_disable_intr(hw);
5740                 intr->mask |= (1 << queue_id);
5741                 ixgbe_enable_intr(dev);
5742         } else if (queue_id < 32) {
5743                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5744                 mask &= (1 << queue_id);
5745                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5746         } else if (queue_id < 64) {
5747                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5748                 mask &= (1 << (queue_id - 32));
5749                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5750         }
5751         rte_intr_enable(intr_handle);
5752
5753         return 0;
5754 }
5755
5756 static int
5757 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5758 {
5759         uint32_t mask;
5760         struct ixgbe_hw *hw =
5761                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5762         struct ixgbe_interrupt *intr =
5763                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5764
5765         if (queue_id < 16) {
5766                 ixgbe_disable_intr(hw);
5767                 intr->mask &= ~(1 << queue_id);
5768                 ixgbe_enable_intr(dev);
5769         } else if (queue_id < 32) {
5770                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5771                 mask &= ~(1 << queue_id);
5772                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5773         } else if (queue_id < 64) {
5774                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5775                 mask &= ~(1 << (queue_id - 32));
5776                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
5777         }
5778
5779         return 0;
5780 }
5781
5782 static void
5783 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5784                      uint8_t queue, uint8_t msix_vector)
5785 {
5786         uint32_t tmp, idx;
5787
5788         if (direction == -1) {
5789                 /* other causes */
5790                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5791                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
5792                 tmp &= ~0xFF;
5793                 tmp |= msix_vector;
5794                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
5795         } else {
5796                 /* rx or tx cause */
5797                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5798                 idx = ((16 * (queue & 1)) + (8 * direction));
5799                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
5800                 tmp &= ~(0xFF << idx);
5801                 tmp |= (msix_vector << idx);
5802                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
5803         }
5804 }
5805
5806 /**
5807  * set the IVAR registers, mapping interrupt causes to vectors
5808  * @param hw
5809  *  pointer to ixgbe_hw struct
5810  * @direction
5811  *  0 for Rx, 1 for Tx, -1 for other causes
5812  * @queue
5813  *  queue to map the corresponding interrupt to
5814  * @msix_vector
5815  *  the vector to map to the corresponding queue
5816  */
5817 static void
5818 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
5819                    uint8_t queue, uint8_t msix_vector)
5820 {
5821         uint32_t tmp, idx;
5822
5823         msix_vector |= IXGBE_IVAR_ALLOC_VAL;
5824         if (hw->mac.type == ixgbe_mac_82598EB) {
5825                 if (direction == -1)
5826                         direction = 0;
5827                 idx = (((direction * 64) + queue) >> 2) & 0x1F;
5828                 tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
5829                 tmp &= ~(0xFF << (8 * (queue & 0x3)));
5830                 tmp |= (msix_vector << (8 * (queue & 0x3)));
5831                 IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
5832         } else if ((hw->mac.type == ixgbe_mac_82599EB) ||
5833                         (hw->mac.type == ixgbe_mac_X540) ||
5834                         (hw->mac.type == ixgbe_mac_X550)) {
5835                 if (direction == -1) {
5836                         /* other causes */
5837                         idx = ((queue & 1) * 8);
5838                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
5839                         tmp &= ~(0xFF << idx);
5840                         tmp |= (msix_vector << idx);
5841                         IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
5842                 } else {
5843                         /* rx or tx causes */
5844                         idx = ((16 * (queue & 1)) + (8 * direction));
5845                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
5846                         tmp &= ~(0xFF << idx);
5847                         tmp |= (msix_vector << idx);
5848                         IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
5849                 }
5850         }
5851 }
5852
5853 static void
5854 ixgbevf_configure_msix(struct rte_eth_dev *dev)
5855 {
5856         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5857         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5858         struct ixgbe_hw *hw =
5859                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5860         uint32_t q_idx;
5861         uint32_t vector_idx = IXGBE_MISC_VEC_ID;
5862         uint32_t base = IXGBE_MISC_VEC_ID;
5863
5864         /* Configure VF other cause ivar */
5865         ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
5866
5867         /* won't configure msix register if no mapping is done
5868          * between intr vector and event fd.
5869          */
5870         if (!rte_intr_dp_is_en(intr_handle))
5871                 return;
5872
5873         if (rte_intr_allow_others(intr_handle)) {
5874                 base = IXGBE_RX_VEC_START;
5875                 vector_idx = IXGBE_RX_VEC_START;
5876         }
5877
5878         /* Configure all RX queues of VF */
5879         for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
5880                 /* Force all queue use vector 0,
5881                  * as IXGBE_VF_MAXMSIVECOTR = 1
5882                  */
5883                 ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
5884                 intr_handle->intr_vec[q_idx] = vector_idx;
5885                 if (vector_idx < base + intr_handle->nb_efd - 1)
5886                         vector_idx++;
5887         }
5888
5889         /* As RX queue setting above show, all queues use the vector 0.
5890          * Set only the ITR value of IXGBE_MISC_VEC_ID.
5891          */
5892         IXGBE_WRITE_REG(hw, IXGBE_VTEITR(IXGBE_MISC_VEC_ID),
5893                         IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
5894                         | IXGBE_EITR_CNT_WDIS);
5895 }
5896
5897 /**
5898  * Sets up the hardware to properly generate MSI-X interrupts
5899  * @hw
5900  *  board private structure
5901  */
5902 static void
5903 ixgbe_configure_msix(struct rte_eth_dev *dev)
5904 {
5905         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5906         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5907         struct ixgbe_hw *hw =
5908                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5909         uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
5910         uint32_t vec = IXGBE_MISC_VEC_ID;
5911         uint32_t mask;
5912         uint32_t gpie;
5913
5914         /* won't configure msix register if no mapping is done
5915          * between intr vector and event fd
5916          * but if misx has been enabled already, need to configure
5917          * auto clean, auto mask and throttling.
5918          */
5919         gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5920         if (!rte_intr_dp_is_en(intr_handle) &&
5921             !(gpie & (IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT)))
5922                 return;
5923
5924         if (rte_intr_allow_others(intr_handle))
5925                 vec = base = IXGBE_RX_VEC_START;
5926
5927         /* setup GPIE for MSI-x mode */
5928         gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5929         gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5930                 IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
5931         /* auto clearing and auto setting corresponding bits in EIMS
5932          * when MSI-X interrupt is triggered
5933          */
5934         if (hw->mac.type == ixgbe_mac_82598EB) {
5935                 IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5936         } else {
5937                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5938                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5939         }
5940         IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5941
5942         /* Populate the IVAR table and set the ITR values to the
5943          * corresponding register.
5944          */
5945         if (rte_intr_dp_is_en(intr_handle)) {
5946                 for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
5947                         queue_id++) {
5948                         /* by default, 1:1 mapping */
5949                         ixgbe_set_ivar_map(hw, 0, queue_id, vec);
5950                         intr_handle->intr_vec[queue_id] = vec;
5951                         if (vec < base + intr_handle->nb_efd - 1)
5952                                 vec++;
5953                 }
5954
5955                 switch (hw->mac.type) {
5956                 case ixgbe_mac_82598EB:
5957                         ixgbe_set_ivar_map(hw, -1,
5958                                            IXGBE_IVAR_OTHER_CAUSES_INDEX,
5959                                            IXGBE_MISC_VEC_ID);
5960                         break;
5961                 case ixgbe_mac_82599EB:
5962                 case ixgbe_mac_X540:
5963                 case ixgbe_mac_X550:
5964                         ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
5965                         break;
5966                 default:
5967                         break;
5968                 }
5969         }
5970         IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
5971                         IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
5972                         | IXGBE_EITR_CNT_WDIS);
5973
5974         /* set up to autoclear timer, and the vectors */
5975         mask = IXGBE_EIMS_ENABLE_MASK;
5976         mask &= ~(IXGBE_EIMS_OTHER |
5977                   IXGBE_EIMS_MAILBOX |
5978                   IXGBE_EIMS_LSC);
5979
5980         IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
5981 }
5982
5983 int
5984 ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
5985                            uint16_t queue_idx, uint16_t tx_rate)
5986 {
5987         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5988         struct rte_eth_rxmode *rxmode;
5989         uint32_t rf_dec, rf_int;
5990         uint32_t bcnrc_val;
5991         uint16_t link_speed = dev->data->dev_link.link_speed;
5992
5993         if (queue_idx >= hw->mac.max_tx_queues)
5994                 return -EINVAL;
5995
5996         if (tx_rate != 0) {
5997                 /* Calculate the rate factor values to set */
5998                 rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
5999                 rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
6000                 rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
6001
6002                 bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
6003                 bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
6004                                 IXGBE_RTTBCNRC_RF_INT_MASK_M);
6005                 bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
6006         } else {
6007                 bcnrc_val = 0;
6008         }
6009
6010         rxmode = &dev->data->dev_conf.rxmode;
6011         /*
6012          * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
6013          * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
6014          * set as 0x4.
6015          */
6016         if ((rxmode->offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) &&
6017             (rxmode->max_rx_pkt_len >= IXGBE_MAX_JUMBO_FRAME_SIZE))
6018                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
6019                         IXGBE_MMW_SIZE_JUMBO_FRAME);
6020         else
6021                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
6022                         IXGBE_MMW_SIZE_DEFAULT);
6023
6024         /* Set RTTBCNRC of queue X */
6025         IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
6026         IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
6027         IXGBE_WRITE_FLUSH(hw);
6028
6029         return 0;
6030 }
6031
6032 static int
6033 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
6034                      __attribute__((unused)) uint32_t index,
6035                      __attribute__((unused)) uint32_t pool)
6036 {
6037         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6038         int diag;
6039
6040         /*
6041          * On a 82599 VF, adding again the same MAC addr is not an idempotent
6042          * operation. Trap this case to avoid exhausting the [very limited]
6043          * set of PF resources used to store VF MAC addresses.
6044          */
6045         if (memcmp(hw->mac.perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
6046                 return -1;
6047         diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6048         if (diag != 0)
6049                 PMD_DRV_LOG(ERR, "Unable to add MAC address "
6050                             "%02x:%02x:%02x:%02x:%02x:%02x - diag=%d",
6051                             mac_addr->addr_bytes[0],
6052                             mac_addr->addr_bytes[1],
6053                             mac_addr->addr_bytes[2],
6054                             mac_addr->addr_bytes[3],
6055                             mac_addr->addr_bytes[4],
6056                             mac_addr->addr_bytes[5],
6057                             diag);
6058         return diag;
6059 }
6060
6061 static void
6062 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
6063 {
6064         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6065         struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
6066         struct ether_addr *mac_addr;
6067         uint32_t i;
6068         int diag;
6069
6070         /*
6071          * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
6072          * not support the deletion of a given MAC address.
6073          * Instead, it imposes to delete all MAC addresses, then to add again
6074          * all MAC addresses with the exception of the one to be deleted.
6075          */
6076         (void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
6077
6078         /*
6079          * Add again all MAC addresses, with the exception of the deleted one
6080          * and of the permanent MAC address.
6081          */
6082         for (i = 0, mac_addr = dev->data->mac_addrs;
6083              i < hw->mac.num_rar_entries; i++, mac_addr++) {
6084                 /* Skip the deleted MAC address */
6085                 if (i == index)
6086                         continue;
6087                 /* Skip NULL MAC addresses */
6088                 if (is_zero_ether_addr(mac_addr))
6089                         continue;
6090                 /* Skip the permanent MAC address */
6091                 if (memcmp(perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
6092                         continue;
6093                 diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6094                 if (diag != 0)
6095                         PMD_DRV_LOG(ERR,
6096                                     "Adding again MAC address "
6097                                     "%02x:%02x:%02x:%02x:%02x:%02x failed "
6098                                     "diag=%d",
6099                                     mac_addr->addr_bytes[0],
6100                                     mac_addr->addr_bytes[1],
6101                                     mac_addr->addr_bytes[2],
6102                                     mac_addr->addr_bytes[3],
6103                                     mac_addr->addr_bytes[4],
6104                                     mac_addr->addr_bytes[5],
6105                                     diag);
6106         }
6107 }
6108
6109 static int
6110 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
6111 {
6112         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6113
6114         hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
6115
6116         return 0;
6117 }
6118
6119 int
6120 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
6121                         struct rte_eth_syn_filter *filter,
6122                         bool add)
6123 {
6124         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6125         struct ixgbe_filter_info *filter_info =
6126                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6127         uint32_t syn_info;
6128         uint32_t synqf;
6129
6130         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6131                 return -EINVAL;
6132
6133         syn_info = filter_info->syn_info;
6134
6135         if (add) {
6136                 if (syn_info & IXGBE_SYN_FILTER_ENABLE)
6137                         return -EINVAL;
6138                 synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
6139                         IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
6140
6141                 if (filter->hig_pri)
6142                         synqf |= IXGBE_SYN_FILTER_SYNQFP;
6143                 else
6144                         synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
6145         } else {
6146                 synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6147                 if (!(syn_info & IXGBE_SYN_FILTER_ENABLE))
6148                         return -ENOENT;
6149                 synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
6150         }
6151
6152         filter_info->syn_info = synqf;
6153         IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
6154         IXGBE_WRITE_FLUSH(hw);
6155         return 0;
6156 }
6157
6158 static int
6159 ixgbe_syn_filter_get(struct rte_eth_dev *dev,
6160                         struct rte_eth_syn_filter *filter)
6161 {
6162         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6163         uint32_t synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6164
6165         if (synqf & IXGBE_SYN_FILTER_ENABLE) {
6166                 filter->hig_pri = (synqf & IXGBE_SYN_FILTER_SYNQFP) ? 1 : 0;
6167                 filter->queue = (uint16_t)((synqf & IXGBE_SYN_FILTER_QUEUE) >> 1);
6168                 return 0;
6169         }
6170         return -ENOENT;
6171 }
6172
6173 static int
6174 ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
6175                         enum rte_filter_op filter_op,
6176                         void *arg)
6177 {
6178         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6179         int ret;
6180
6181         MAC_TYPE_FILTER_SUP(hw->mac.type);
6182
6183         if (filter_op == RTE_ETH_FILTER_NOP)
6184                 return 0;
6185
6186         if (arg == NULL) {
6187                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
6188                             filter_op);
6189                 return -EINVAL;
6190         }
6191
6192         switch (filter_op) {
6193         case RTE_ETH_FILTER_ADD:
6194                 ret = ixgbe_syn_filter_set(dev,
6195                                 (struct rte_eth_syn_filter *)arg,
6196                                 TRUE);
6197                 break;
6198         case RTE_ETH_FILTER_DELETE:
6199                 ret = ixgbe_syn_filter_set(dev,
6200                                 (struct rte_eth_syn_filter *)arg,
6201                                 FALSE);
6202                 break;
6203         case RTE_ETH_FILTER_GET:
6204                 ret = ixgbe_syn_filter_get(dev,
6205                                 (struct rte_eth_syn_filter *)arg);
6206                 break;
6207         default:
6208                 PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
6209                 ret = -EINVAL;
6210                 break;
6211         }
6212
6213         return ret;
6214 }
6215
6216
6217 static inline enum ixgbe_5tuple_protocol
6218 convert_protocol_type(uint8_t protocol_value)
6219 {
6220         if (protocol_value == IPPROTO_TCP)
6221                 return IXGBE_FILTER_PROTOCOL_TCP;
6222         else if (protocol_value == IPPROTO_UDP)
6223                 return IXGBE_FILTER_PROTOCOL_UDP;
6224         else if (protocol_value == IPPROTO_SCTP)
6225                 return IXGBE_FILTER_PROTOCOL_SCTP;
6226         else
6227                 return IXGBE_FILTER_PROTOCOL_NONE;
6228 }
6229
6230 /* inject a 5-tuple filter to HW */
6231 static inline void
6232 ixgbe_inject_5tuple_filter(struct rte_eth_dev *dev,
6233                            struct ixgbe_5tuple_filter *filter)
6234 {
6235         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6236         int i;
6237         uint32_t ftqf, sdpqf;
6238         uint32_t l34timir = 0;
6239         uint8_t mask = 0xff;
6240
6241         i = filter->index;
6242
6243         sdpqf = (uint32_t)(filter->filter_info.dst_port <<
6244                                 IXGBE_SDPQF_DSTPORT_SHIFT);
6245         sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
6246
6247         ftqf = (uint32_t)(filter->filter_info.proto &
6248                 IXGBE_FTQF_PROTOCOL_MASK);
6249         ftqf |= (uint32_t)((filter->filter_info.priority &
6250                 IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
6251         if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
6252                 mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
6253         if (filter->filter_info.dst_ip_mask == 0)
6254                 mask &= IXGBE_FTQF_DEST_ADDR_MASK;
6255         if (filter->filter_info.src_port_mask == 0)
6256                 mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
6257         if (filter->filter_info.dst_port_mask == 0)
6258                 mask &= IXGBE_FTQF_DEST_PORT_MASK;
6259         if (filter->filter_info.proto_mask == 0)
6260                 mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
6261         ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
6262         ftqf |= IXGBE_FTQF_POOL_MASK_EN;
6263         ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
6264
6265         IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
6266         IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
6267         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
6268         IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
6269
6270         l34timir |= IXGBE_L34T_IMIR_RESERVE;
6271         l34timir |= (uint32_t)(filter->queue <<
6272                                 IXGBE_L34T_IMIR_QUEUE_SHIFT);
6273         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
6274 }
6275
6276 /*
6277  * add a 5tuple filter
6278  *
6279  * @param
6280  * dev: Pointer to struct rte_eth_dev.
6281  * index: the index the filter allocates.
6282  * filter: ponter to the filter that will be added.
6283  * rx_queue: the queue id the filter assigned to.
6284  *
6285  * @return
6286  *    - On success, zero.
6287  *    - On failure, a negative value.
6288  */
6289 static int
6290 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
6291                         struct ixgbe_5tuple_filter *filter)
6292 {
6293         struct ixgbe_filter_info *filter_info =
6294                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6295         int i, idx, shift;
6296
6297         /*
6298          * look for an unused 5tuple filter index,
6299          * and insert the filter to list.
6300          */
6301         for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
6302                 idx = i / (sizeof(uint32_t) * NBBY);
6303                 shift = i % (sizeof(uint32_t) * NBBY);
6304                 if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
6305                         filter_info->fivetuple_mask[idx] |= 1 << shift;
6306                         filter->index = i;
6307                         TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
6308                                           filter,
6309                                           entries);
6310                         break;
6311                 }
6312         }
6313         if (i >= IXGBE_MAX_FTQF_FILTERS) {
6314                 PMD_DRV_LOG(ERR, "5tuple filters are full.");
6315                 return -ENOSYS;
6316         }
6317
6318         ixgbe_inject_5tuple_filter(dev, filter);
6319
6320         return 0;
6321 }
6322
6323 /*
6324  * remove a 5tuple filter
6325  *
6326  * @param
6327  * dev: Pointer to struct rte_eth_dev.
6328  * filter: the pointer of the filter will be removed.
6329  */
6330 static void
6331 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
6332                         struct ixgbe_5tuple_filter *filter)
6333 {
6334         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6335         struct ixgbe_filter_info *filter_info =
6336                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6337         uint16_t index = filter->index;
6338
6339         filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
6340                                 ~(1 << (index % (sizeof(uint32_t) * NBBY)));
6341         TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
6342         rte_free(filter);
6343
6344         IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
6345         IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
6346         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
6347         IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
6348         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
6349 }
6350
6351 static int
6352 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
6353 {
6354         struct ixgbe_hw *hw;
6355         uint32_t max_frame = mtu + IXGBE_ETH_OVERHEAD;
6356         struct rte_eth_dev_data *dev_data = dev->data;
6357
6358         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6359
6360         if ((mtu < ETHER_MIN_MTU) || (max_frame > ETHER_MAX_JUMBO_FRAME_LEN))
6361                 return -EINVAL;
6362
6363         /* If device is started, refuse mtu that requires the support of
6364          * scattered packets when this feature has not been enabled before.
6365          */
6366         if (dev_data->dev_started && !dev_data->scattered_rx &&
6367             (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
6368              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
6369                 PMD_INIT_LOG(ERR, "Stop port first.");
6370                 return -EINVAL;
6371         }
6372
6373         /*
6374          * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
6375          * request of the version 2.0 of the mailbox API.
6376          * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
6377          * of the mailbox API.
6378          * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
6379          * prior to 3.11.33 which contains the following change:
6380          * "ixgbe: Enable jumbo frames support w/ SR-IOV"
6381          */
6382         ixgbevf_rlpml_set_vf(hw, max_frame);
6383
6384         /* update max frame size */
6385         dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame;
6386         return 0;
6387 }
6388
6389 static inline struct ixgbe_5tuple_filter *
6390 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
6391                         struct ixgbe_5tuple_filter_info *key)
6392 {
6393         struct ixgbe_5tuple_filter *it;
6394
6395         TAILQ_FOREACH(it, filter_list, entries) {
6396                 if (memcmp(key, &it->filter_info,
6397                         sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
6398                         return it;
6399                 }
6400         }
6401         return NULL;
6402 }
6403
6404 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
6405 static inline int
6406 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
6407                         struct ixgbe_5tuple_filter_info *filter_info)
6408 {
6409         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
6410                 filter->priority > IXGBE_5TUPLE_MAX_PRI ||
6411                 filter->priority < IXGBE_5TUPLE_MIN_PRI)
6412                 return -EINVAL;
6413
6414         switch (filter->dst_ip_mask) {
6415         case UINT32_MAX:
6416                 filter_info->dst_ip_mask = 0;
6417                 filter_info->dst_ip = filter->dst_ip;
6418                 break;
6419         case 0:
6420                 filter_info->dst_ip_mask = 1;
6421                 break;
6422         default:
6423                 PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
6424                 return -EINVAL;
6425         }
6426
6427         switch (filter->src_ip_mask) {
6428         case UINT32_MAX:
6429                 filter_info->src_ip_mask = 0;
6430                 filter_info->src_ip = filter->src_ip;
6431                 break;
6432         case 0:
6433                 filter_info->src_ip_mask = 1;
6434                 break;
6435         default:
6436                 PMD_DRV_LOG(ERR, "invalid src_ip mask.");
6437                 return -EINVAL;
6438         }
6439
6440         switch (filter->dst_port_mask) {
6441         case UINT16_MAX:
6442                 filter_info->dst_port_mask = 0;
6443                 filter_info->dst_port = filter->dst_port;
6444                 break;
6445         case 0:
6446                 filter_info->dst_port_mask = 1;
6447                 break;
6448         default:
6449                 PMD_DRV_LOG(ERR, "invalid dst_port mask.");
6450                 return -EINVAL;
6451         }
6452
6453         switch (filter->src_port_mask) {
6454         case UINT16_MAX:
6455                 filter_info->src_port_mask = 0;
6456                 filter_info->src_port = filter->src_port;
6457                 break;
6458         case 0:
6459                 filter_info->src_port_mask = 1;
6460                 break;
6461         default:
6462                 PMD_DRV_LOG(ERR, "invalid src_port mask.");
6463                 return -EINVAL;
6464         }
6465
6466         switch (filter->proto_mask) {
6467         case UINT8_MAX:
6468                 filter_info->proto_mask = 0;
6469                 filter_info->proto =
6470                         convert_protocol_type(filter->proto);
6471                 break;
6472         case 0:
6473                 filter_info->proto_mask = 1;
6474                 break;
6475         default:
6476                 PMD_DRV_LOG(ERR, "invalid protocol mask.");
6477                 return -EINVAL;
6478         }
6479
6480         filter_info->priority = (uint8_t)filter->priority;
6481         return 0;
6482 }
6483
6484 /*
6485  * add or delete a ntuple filter
6486  *
6487  * @param
6488  * dev: Pointer to struct rte_eth_dev.
6489  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6490  * add: if true, add filter, if false, remove filter
6491  *
6492  * @return
6493  *    - On success, zero.
6494  *    - On failure, a negative value.
6495  */
6496 int
6497 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
6498                         struct rte_eth_ntuple_filter *ntuple_filter,
6499                         bool add)
6500 {
6501         struct ixgbe_filter_info *filter_info =
6502                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6503         struct ixgbe_5tuple_filter_info filter_5tuple;
6504         struct ixgbe_5tuple_filter *filter;
6505         int ret;
6506
6507         if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6508                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6509                 return -EINVAL;
6510         }
6511
6512         memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6513         ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6514         if (ret < 0)
6515                 return ret;
6516
6517         filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6518                                          &filter_5tuple);
6519         if (filter != NULL && add) {
6520                 PMD_DRV_LOG(ERR, "filter exists.");
6521                 return -EEXIST;
6522         }
6523         if (filter == NULL && !add) {
6524                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
6525                 return -ENOENT;
6526         }
6527
6528         if (add) {
6529                 filter = rte_zmalloc("ixgbe_5tuple_filter",
6530                                 sizeof(struct ixgbe_5tuple_filter), 0);
6531                 if (filter == NULL)
6532                         return -ENOMEM;
6533                 rte_memcpy(&filter->filter_info,
6534                                  &filter_5tuple,
6535                                  sizeof(struct ixgbe_5tuple_filter_info));
6536                 filter->queue = ntuple_filter->queue;
6537                 ret = ixgbe_add_5tuple_filter(dev, filter);
6538                 if (ret < 0) {
6539                         rte_free(filter);
6540                         return ret;
6541                 }
6542         } else
6543                 ixgbe_remove_5tuple_filter(dev, filter);
6544
6545         return 0;
6546 }
6547
6548 /*
6549  * get a ntuple filter
6550  *
6551  * @param
6552  * dev: Pointer to struct rte_eth_dev.
6553  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6554  *
6555  * @return
6556  *    - On success, zero.
6557  *    - On failure, a negative value.
6558  */
6559 static int
6560 ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
6561                         struct rte_eth_ntuple_filter *ntuple_filter)
6562 {
6563         struct ixgbe_filter_info *filter_info =
6564                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6565         struct ixgbe_5tuple_filter_info filter_5tuple;
6566         struct ixgbe_5tuple_filter *filter;
6567         int ret;
6568
6569         if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6570                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6571                 return -EINVAL;
6572         }
6573
6574         memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6575         ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6576         if (ret < 0)
6577                 return ret;
6578
6579         filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6580                                          &filter_5tuple);
6581         if (filter == NULL) {
6582                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
6583                 return -ENOENT;
6584         }
6585         ntuple_filter->queue = filter->queue;
6586         return 0;
6587 }
6588
6589 /*
6590  * ixgbe_ntuple_filter_handle - Handle operations for ntuple filter.
6591  * @dev: pointer to rte_eth_dev structure
6592  * @filter_op:operation will be taken.
6593  * @arg: a pointer to specific structure corresponding to the filter_op
6594  *
6595  * @return
6596  *    - On success, zero.
6597  *    - On failure, a negative value.
6598  */
6599 static int
6600 ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
6601                                 enum rte_filter_op filter_op,
6602                                 void *arg)
6603 {
6604         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6605         int ret;
6606
6607         MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
6608
6609         if (filter_op == RTE_ETH_FILTER_NOP)
6610                 return 0;
6611
6612         if (arg == NULL) {
6613                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
6614                             filter_op);
6615                 return -EINVAL;
6616         }
6617
6618         switch (filter_op) {
6619         case RTE_ETH_FILTER_ADD:
6620                 ret = ixgbe_add_del_ntuple_filter(dev,
6621                         (struct rte_eth_ntuple_filter *)arg,
6622                         TRUE);
6623                 break;
6624         case RTE_ETH_FILTER_DELETE:
6625                 ret = ixgbe_add_del_ntuple_filter(dev,
6626                         (struct rte_eth_ntuple_filter *)arg,
6627                         FALSE);
6628                 break;
6629         case RTE_ETH_FILTER_GET:
6630                 ret = ixgbe_get_ntuple_filter(dev,
6631                         (struct rte_eth_ntuple_filter *)arg);
6632                 break;
6633         default:
6634                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
6635                 ret = -EINVAL;
6636                 break;
6637         }
6638         return ret;
6639 }
6640
6641 int
6642 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
6643                         struct rte_eth_ethertype_filter *filter,
6644                         bool add)
6645 {
6646         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6647         struct ixgbe_filter_info *filter_info =
6648                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6649         uint32_t etqf = 0;
6650         uint32_t etqs = 0;
6651         int ret;
6652         struct ixgbe_ethertype_filter ethertype_filter;
6653
6654         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6655                 return -EINVAL;
6656
6657         if (filter->ether_type == ETHER_TYPE_IPv4 ||
6658                 filter->ether_type == ETHER_TYPE_IPv6) {
6659                 PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
6660                         " ethertype filter.", filter->ether_type);
6661                 return -EINVAL;
6662         }
6663
6664         if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
6665                 PMD_DRV_LOG(ERR, "mac compare is unsupported.");
6666                 return -EINVAL;
6667         }
6668         if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
6669                 PMD_DRV_LOG(ERR, "drop option is unsupported.");
6670                 return -EINVAL;
6671         }
6672
6673         ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6674         if (ret >= 0 && add) {
6675                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
6676                             filter->ether_type);
6677                 return -EEXIST;
6678         }
6679         if (ret < 0 && !add) {
6680                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6681                             filter->ether_type);
6682                 return -ENOENT;
6683         }
6684
6685         if (add) {
6686                 etqf = IXGBE_ETQF_FILTER_EN;
6687                 etqf |= (uint32_t)filter->ether_type;
6688                 etqs |= (uint32_t)((filter->queue <<
6689                                     IXGBE_ETQS_RX_QUEUE_SHIFT) &
6690                                     IXGBE_ETQS_RX_QUEUE);
6691                 etqs |= IXGBE_ETQS_QUEUE_EN;
6692
6693                 ethertype_filter.ethertype = filter->ether_type;
6694                 ethertype_filter.etqf = etqf;
6695                 ethertype_filter.etqs = etqs;
6696                 ethertype_filter.conf = FALSE;
6697                 ret = ixgbe_ethertype_filter_insert(filter_info,
6698                                                     &ethertype_filter);
6699                 if (ret < 0) {
6700                         PMD_DRV_LOG(ERR, "ethertype filters are full.");
6701                         return -ENOSPC;
6702                 }
6703         } else {
6704                 ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
6705                 if (ret < 0)
6706                         return -ENOSYS;
6707         }
6708         IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
6709         IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
6710         IXGBE_WRITE_FLUSH(hw);
6711
6712         return 0;
6713 }
6714
6715 static int
6716 ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
6717                         struct rte_eth_ethertype_filter *filter)
6718 {
6719         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6720         struct ixgbe_filter_info *filter_info =
6721                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6722         uint32_t etqf, etqs;
6723         int ret;
6724
6725         ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6726         if (ret < 0) {
6727                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6728                             filter->ether_type);
6729                 return -ENOENT;
6730         }
6731
6732         etqf = IXGBE_READ_REG(hw, IXGBE_ETQF(ret));
6733         if (etqf & IXGBE_ETQF_FILTER_EN) {
6734                 etqs = IXGBE_READ_REG(hw, IXGBE_ETQS(ret));
6735                 filter->ether_type = etqf & IXGBE_ETQF_ETHERTYPE;
6736                 filter->flags = 0;
6737                 filter->queue = (etqs & IXGBE_ETQS_RX_QUEUE) >>
6738                                IXGBE_ETQS_RX_QUEUE_SHIFT;
6739                 return 0;
6740         }
6741         return -ENOENT;
6742 }
6743
6744 /*
6745  * ixgbe_ethertype_filter_handle - Handle operations for ethertype filter.
6746  * @dev: pointer to rte_eth_dev structure
6747  * @filter_op:operation will be taken.
6748  * @arg: a pointer to specific structure corresponding to the filter_op
6749  */
6750 static int
6751 ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
6752                                 enum rte_filter_op filter_op,
6753                                 void *arg)
6754 {
6755         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6756         int ret;
6757
6758         MAC_TYPE_FILTER_SUP(hw->mac.type);
6759
6760         if (filter_op == RTE_ETH_FILTER_NOP)
6761                 return 0;
6762
6763         if (arg == NULL) {
6764                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
6765                             filter_op);
6766                 return -EINVAL;
6767         }
6768
6769         switch (filter_op) {
6770         case RTE_ETH_FILTER_ADD:
6771                 ret = ixgbe_add_del_ethertype_filter(dev,
6772                         (struct rte_eth_ethertype_filter *)arg,
6773                         TRUE);
6774                 break;
6775         case RTE_ETH_FILTER_DELETE:
6776                 ret = ixgbe_add_del_ethertype_filter(dev,
6777                         (struct rte_eth_ethertype_filter *)arg,
6778                         FALSE);
6779                 break;
6780         case RTE_ETH_FILTER_GET:
6781                 ret = ixgbe_get_ethertype_filter(dev,
6782                         (struct rte_eth_ethertype_filter *)arg);
6783                 break;
6784         default:
6785                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
6786                 ret = -EINVAL;
6787                 break;
6788         }
6789         return ret;
6790 }
6791
6792 static int
6793 ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
6794                      enum rte_filter_type filter_type,
6795                      enum rte_filter_op filter_op,
6796                      void *arg)
6797 {
6798         int ret = 0;
6799
6800         switch (filter_type) {
6801         case RTE_ETH_FILTER_NTUPLE:
6802                 ret = ixgbe_ntuple_filter_handle(dev, filter_op, arg);
6803                 break;
6804         case RTE_ETH_FILTER_ETHERTYPE:
6805                 ret = ixgbe_ethertype_filter_handle(dev, filter_op, arg);
6806                 break;
6807         case RTE_ETH_FILTER_SYN:
6808                 ret = ixgbe_syn_filter_handle(dev, filter_op, arg);
6809                 break;
6810         case RTE_ETH_FILTER_FDIR:
6811                 ret = ixgbe_fdir_ctrl_func(dev, filter_op, arg);
6812                 break;
6813         case RTE_ETH_FILTER_L2_TUNNEL:
6814                 ret = ixgbe_dev_l2_tunnel_filter_handle(dev, filter_op, arg);
6815                 break;
6816         case RTE_ETH_FILTER_GENERIC:
6817                 if (filter_op != RTE_ETH_FILTER_GET)
6818                         return -EINVAL;
6819                 *(const void **)arg = &ixgbe_flow_ops;
6820                 break;
6821         default:
6822                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
6823                                                         filter_type);
6824                 ret = -EINVAL;
6825                 break;
6826         }
6827
6828         return ret;
6829 }
6830
6831 static u8 *
6832 ixgbe_dev_addr_list_itr(__attribute__((unused)) struct ixgbe_hw *hw,
6833                         u8 **mc_addr_ptr, u32 *vmdq)
6834 {
6835         u8 *mc_addr;
6836
6837         *vmdq = 0;
6838         mc_addr = *mc_addr_ptr;
6839         *mc_addr_ptr = (mc_addr + sizeof(struct ether_addr));
6840         return mc_addr;
6841 }
6842
6843 static int
6844 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
6845                           struct ether_addr *mc_addr_set,
6846                           uint32_t nb_mc_addr)
6847 {
6848         struct ixgbe_hw *hw;
6849         u8 *mc_addr_list;
6850
6851         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6852         mc_addr_list = (u8 *)mc_addr_set;
6853         return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
6854                                          ixgbe_dev_addr_list_itr, TRUE);
6855 }
6856
6857 static uint64_t
6858 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
6859 {
6860         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6861         uint64_t systime_cycles;
6862
6863         switch (hw->mac.type) {
6864         case ixgbe_mac_X550:
6865         case ixgbe_mac_X550EM_x:
6866         case ixgbe_mac_X550EM_a:
6867                 /* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
6868                 systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6869                 systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6870                                 * NSEC_PER_SEC;
6871                 break;
6872         default:
6873                 systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6874                 systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6875                                 << 32;
6876         }
6877
6878         return systime_cycles;
6879 }
6880
6881 static uint64_t
6882 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6883 {
6884         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6885         uint64_t rx_tstamp_cycles;
6886
6887         switch (hw->mac.type) {
6888         case ixgbe_mac_X550:
6889         case ixgbe_mac_X550EM_x:
6890         case ixgbe_mac_X550EM_a:
6891                 /* RXSTMPL stores ns and RXSTMPH stores seconds. */
6892                 rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6893                 rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6894                                 * NSEC_PER_SEC;
6895                 break;
6896         default:
6897                 /* RXSTMPL stores ns and RXSTMPH stores seconds. */
6898                 rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6899                 rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6900                                 << 32;
6901         }
6902
6903         return rx_tstamp_cycles;
6904 }
6905
6906 static uint64_t
6907 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6908 {
6909         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6910         uint64_t tx_tstamp_cycles;
6911
6912         switch (hw->mac.type) {
6913         case ixgbe_mac_X550:
6914         case ixgbe_mac_X550EM_x:
6915         case ixgbe_mac_X550EM_a:
6916                 /* TXSTMPL stores ns and TXSTMPH stores seconds. */
6917                 tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6918                 tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6919                                 * NSEC_PER_SEC;
6920                 break;
6921         default:
6922                 /* TXSTMPL stores ns and TXSTMPH stores seconds. */
6923                 tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6924                 tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6925                                 << 32;
6926         }
6927
6928         return tx_tstamp_cycles;
6929 }
6930
6931 static void
6932 ixgbe_start_timecounters(struct rte_eth_dev *dev)
6933 {
6934         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6935         struct ixgbe_adapter *adapter =
6936                 (struct ixgbe_adapter *)dev->data->dev_private;
6937         struct rte_eth_link link;
6938         uint32_t incval = 0;
6939         uint32_t shift = 0;
6940
6941         /* Get current link speed. */
6942         ixgbe_dev_link_update(dev, 1);
6943         rte_eth_linkstatus_get(dev, &link);
6944
6945         switch (link.link_speed) {
6946         case ETH_SPEED_NUM_100M:
6947                 incval = IXGBE_INCVAL_100;
6948                 shift = IXGBE_INCVAL_SHIFT_100;
6949                 break;
6950         case ETH_SPEED_NUM_1G:
6951                 incval = IXGBE_INCVAL_1GB;
6952                 shift = IXGBE_INCVAL_SHIFT_1GB;
6953                 break;
6954         case ETH_SPEED_NUM_10G:
6955         default:
6956                 incval = IXGBE_INCVAL_10GB;
6957                 shift = IXGBE_INCVAL_SHIFT_10GB;
6958                 break;
6959         }
6960
6961         switch (hw->mac.type) {
6962         case ixgbe_mac_X550:
6963         case ixgbe_mac_X550EM_x:
6964         case ixgbe_mac_X550EM_a:
6965                 /* Independent of link speed. */
6966                 incval = 1;
6967                 /* Cycles read will be interpreted as ns. */
6968                 shift = 0;
6969                 /* Fall-through */
6970         case ixgbe_mac_X540:
6971                 IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
6972                 break;
6973         case ixgbe_mac_82599EB:
6974                 incval >>= IXGBE_INCVAL_SHIFT_82599;
6975                 shift -= IXGBE_INCVAL_SHIFT_82599;
6976                 IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
6977                                 (1 << IXGBE_INCPER_SHIFT_82599) | incval);
6978                 break;
6979         default:
6980                 /* Not supported. */
6981                 return;
6982         }
6983
6984         memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
6985         memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6986         memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6987
6988         adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6989         adapter->systime_tc.cc_shift = shift;
6990         adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
6991
6992         adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6993         adapter->rx_tstamp_tc.cc_shift = shift;
6994         adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6995
6996         adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6997         adapter->tx_tstamp_tc.cc_shift = shift;
6998         adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6999 }
7000
7001 static int
7002 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
7003 {
7004         struct ixgbe_adapter *adapter =
7005                         (struct ixgbe_adapter *)dev->data->dev_private;
7006
7007         adapter->systime_tc.nsec += delta;
7008         adapter->rx_tstamp_tc.nsec += delta;
7009         adapter->tx_tstamp_tc.nsec += delta;
7010
7011         return 0;
7012 }
7013
7014 static int
7015 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
7016 {
7017         uint64_t ns;
7018         struct ixgbe_adapter *adapter =
7019                         (struct ixgbe_adapter *)dev->data->dev_private;
7020
7021         ns = rte_timespec_to_ns(ts);
7022         /* Set the timecounters to a new value. */
7023         adapter->systime_tc.nsec = ns;
7024         adapter->rx_tstamp_tc.nsec = ns;
7025         adapter->tx_tstamp_tc.nsec = ns;
7026
7027         return 0;
7028 }
7029
7030 static int
7031 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
7032 {
7033         uint64_t ns, systime_cycles;
7034         struct ixgbe_adapter *adapter =
7035                         (struct ixgbe_adapter *)dev->data->dev_private;
7036
7037         systime_cycles = ixgbe_read_systime_cyclecounter(dev);
7038         ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
7039         *ts = rte_ns_to_timespec(ns);
7040
7041         return 0;
7042 }
7043
7044 static int
7045 ixgbe_timesync_enable(struct rte_eth_dev *dev)
7046 {
7047         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7048         uint32_t tsync_ctl;
7049         uint32_t tsauxc;
7050
7051         /* Stop the timesync system time. */
7052         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
7053         /* Reset the timesync system time value. */
7054         IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
7055         IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
7056
7057         /* Enable system time for platforms where it isn't on by default. */
7058         tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
7059         tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
7060         IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
7061
7062         ixgbe_start_timecounters(dev);
7063
7064         /* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
7065         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
7066                         (ETHER_TYPE_1588 |
7067                          IXGBE_ETQF_FILTER_EN |
7068                          IXGBE_ETQF_1588));
7069
7070         /* Enable timestamping of received PTP packets. */
7071         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7072         tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
7073         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
7074
7075         /* Enable timestamping of transmitted PTP packets. */
7076         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7077         tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
7078         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
7079
7080         IXGBE_WRITE_FLUSH(hw);
7081
7082         return 0;
7083 }
7084
7085 static int
7086 ixgbe_timesync_disable(struct rte_eth_dev *dev)
7087 {
7088         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7089         uint32_t tsync_ctl;
7090
7091         /* Disable timestamping of transmitted PTP packets. */
7092         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7093         tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
7094         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
7095
7096         /* Disable timestamping of received PTP packets. */
7097         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7098         tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
7099         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
7100
7101         /* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
7102         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
7103
7104         /* Stop incrementating the System Time registers. */
7105         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
7106
7107         return 0;
7108 }
7109
7110 static int
7111 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
7112                                  struct timespec *timestamp,
7113                                  uint32_t flags __rte_unused)
7114 {
7115         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7116         struct ixgbe_adapter *adapter =
7117                 (struct ixgbe_adapter *)dev->data->dev_private;
7118         uint32_t tsync_rxctl;
7119         uint64_t rx_tstamp_cycles;
7120         uint64_t ns;
7121
7122         tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7123         if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
7124                 return -EINVAL;
7125
7126         rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
7127         ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
7128         *timestamp = rte_ns_to_timespec(ns);
7129
7130         return  0;
7131 }
7132
7133 static int
7134 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
7135                                  struct timespec *timestamp)
7136 {
7137         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7138         struct ixgbe_adapter *adapter =
7139                 (struct ixgbe_adapter *)dev->data->dev_private;
7140         uint32_t tsync_txctl;
7141         uint64_t tx_tstamp_cycles;
7142         uint64_t ns;
7143
7144         tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7145         if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
7146                 return -EINVAL;
7147
7148         tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
7149         ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
7150         *timestamp = rte_ns_to_timespec(ns);
7151
7152         return 0;
7153 }
7154
7155 static int
7156 ixgbe_get_reg_length(struct rte_eth_dev *dev)
7157 {
7158         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7159         int count = 0;
7160         int g_ind = 0;
7161         const struct reg_info *reg_group;
7162         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7163                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7164
7165         while ((reg_group = reg_set[g_ind++]))
7166                 count += ixgbe_regs_group_count(reg_group);
7167
7168         return count;
7169 }
7170
7171 static int
7172 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
7173 {
7174         int count = 0;
7175         int g_ind = 0;
7176         const struct reg_info *reg_group;
7177
7178         while ((reg_group = ixgbevf_regs[g_ind++]))
7179                 count += ixgbe_regs_group_count(reg_group);
7180
7181         return count;
7182 }
7183
7184 static int
7185 ixgbe_get_regs(struct rte_eth_dev *dev,
7186               struct rte_dev_reg_info *regs)
7187 {
7188         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7189         uint32_t *data = regs->data;
7190         int g_ind = 0;
7191         int count = 0;
7192         const struct reg_info *reg_group;
7193         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7194                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7195
7196         if (data == NULL) {
7197                 regs->length = ixgbe_get_reg_length(dev);
7198                 regs->width = sizeof(uint32_t);
7199                 return 0;
7200         }
7201
7202         /* Support only full register dump */
7203         if ((regs->length == 0) ||
7204             (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
7205                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7206                         hw->device_id;
7207                 while ((reg_group = reg_set[g_ind++]))
7208                         count += ixgbe_read_regs_group(dev, &data[count],
7209                                 reg_group);
7210                 return 0;
7211         }
7212
7213         return -ENOTSUP;
7214 }
7215
7216 static int
7217 ixgbevf_get_regs(struct rte_eth_dev *dev,
7218                 struct rte_dev_reg_info *regs)
7219 {
7220         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7221         uint32_t *data = regs->data;
7222         int g_ind = 0;
7223         int count = 0;
7224         const struct reg_info *reg_group;
7225
7226         if (data == NULL) {
7227                 regs->length = ixgbevf_get_reg_length(dev);
7228                 regs->width = sizeof(uint32_t);
7229                 return 0;
7230         }
7231
7232         /* Support only full register dump */
7233         if ((regs->length == 0) ||
7234             (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
7235                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7236                         hw->device_id;
7237                 while ((reg_group = ixgbevf_regs[g_ind++]))
7238                         count += ixgbe_read_regs_group(dev, &data[count],
7239                                                       reg_group);
7240                 return 0;
7241         }
7242
7243         return -ENOTSUP;
7244 }
7245
7246 static int
7247 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
7248 {
7249         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7250
7251         /* Return unit is byte count */
7252         return hw->eeprom.word_size * 2;
7253 }
7254
7255 static int
7256 ixgbe_get_eeprom(struct rte_eth_dev *dev,
7257                 struct rte_dev_eeprom_info *in_eeprom)
7258 {
7259         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7260         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7261         uint16_t *data = in_eeprom->data;
7262         int first, length;
7263
7264         first = in_eeprom->offset >> 1;
7265         length = in_eeprom->length >> 1;
7266         if ((first > hw->eeprom.word_size) ||
7267             ((first + length) > hw->eeprom.word_size))
7268                 return -EINVAL;
7269
7270         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7271
7272         return eeprom->ops.read_buffer(hw, first, length, data);
7273 }
7274
7275 static int
7276 ixgbe_set_eeprom(struct rte_eth_dev *dev,
7277                 struct rte_dev_eeprom_info *in_eeprom)
7278 {
7279         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7280         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7281         uint16_t *data = in_eeprom->data;
7282         int first, length;
7283
7284         first = in_eeprom->offset >> 1;
7285         length = in_eeprom->length >> 1;
7286         if ((first > hw->eeprom.word_size) ||
7287             ((first + length) > hw->eeprom.word_size))
7288                 return -EINVAL;
7289
7290         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7291
7292         return eeprom->ops.write_buffer(hw,  first, length, data);
7293 }
7294
7295 static int
7296 ixgbe_get_module_info(struct rte_eth_dev *dev,
7297                       struct rte_eth_dev_module_info *modinfo)
7298 {
7299         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7300         uint32_t status;
7301         uint8_t sff8472_rev, addr_mode;
7302         bool page_swap = false;
7303
7304         /* Check whether we support SFF-8472 or not */
7305         status = hw->phy.ops.read_i2c_eeprom(hw,
7306                                              IXGBE_SFF_SFF_8472_COMP,
7307                                              &sff8472_rev);
7308         if (status != 0)
7309                 return -EIO;
7310
7311         /* addressing mode is not supported */
7312         status = hw->phy.ops.read_i2c_eeprom(hw,
7313                                              IXGBE_SFF_SFF_8472_SWAP,
7314                                              &addr_mode);
7315         if (status != 0)
7316                 return -EIO;
7317
7318         if (addr_mode & IXGBE_SFF_ADDRESSING_MODE) {
7319                 PMD_DRV_LOG(ERR,
7320                             "Address change required to access page 0xA2, "
7321                             "but not supported. Please report the module "
7322                             "type to the driver maintainers.");
7323                 page_swap = true;
7324         }
7325
7326         if (sff8472_rev == IXGBE_SFF_SFF_8472_UNSUP || page_swap) {
7327                 /* We have a SFP, but it does not support SFF-8472 */
7328                 modinfo->type = RTE_ETH_MODULE_SFF_8079;
7329                 modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
7330         } else {
7331                 /* We have a SFP which supports a revision of SFF-8472. */
7332                 modinfo->type = RTE_ETH_MODULE_SFF_8472;
7333                 modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
7334         }
7335
7336         return 0;
7337 }
7338
7339 static int
7340 ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
7341                         struct rte_dev_eeprom_info *info)
7342 {
7343         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7344         uint32_t status = IXGBE_ERR_PHY_ADDR_INVALID;
7345         uint8_t databyte = 0xFF;
7346         uint8_t *data = info->data;
7347         uint32_t i = 0;
7348
7349         if (info->length == 0)
7350                 return -EINVAL;
7351
7352         for (i = info->offset; i < info->offset + info->length; i++) {
7353                 if (i < RTE_ETH_MODULE_SFF_8079_LEN)
7354                         status = hw->phy.ops.read_i2c_eeprom(hw, i, &databyte);
7355                 else
7356                         status = hw->phy.ops.read_i2c_sff8472(hw, i, &databyte);
7357
7358                 if (status != 0)
7359                         return -EIO;
7360
7361                 data[i - info->offset] = databyte;
7362         }
7363
7364         return 0;
7365 }
7366
7367 uint16_t
7368 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
7369         switch (mac_type) {
7370         case ixgbe_mac_X550:
7371         case ixgbe_mac_X550EM_x:
7372         case ixgbe_mac_X550EM_a:
7373                 return ETH_RSS_RETA_SIZE_512;
7374         case ixgbe_mac_X550_vf:
7375         case ixgbe_mac_X550EM_x_vf:
7376         case ixgbe_mac_X550EM_a_vf:
7377                 return ETH_RSS_RETA_SIZE_64;
7378         default:
7379                 return ETH_RSS_RETA_SIZE_128;
7380         }
7381 }
7382
7383 uint32_t
7384 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
7385         switch (mac_type) {
7386         case ixgbe_mac_X550:
7387         case ixgbe_mac_X550EM_x:
7388         case ixgbe_mac_X550EM_a:
7389                 if (reta_idx < ETH_RSS_RETA_SIZE_128)
7390                         return IXGBE_RETA(reta_idx >> 2);
7391                 else
7392                         return IXGBE_ERETA((reta_idx - ETH_RSS_RETA_SIZE_128) >> 2);
7393         case ixgbe_mac_X550_vf:
7394         case ixgbe_mac_X550EM_x_vf:
7395         case ixgbe_mac_X550EM_a_vf:
7396                 return IXGBE_VFRETA(reta_idx >> 2);
7397         default:
7398                 return IXGBE_RETA(reta_idx >> 2);
7399         }
7400 }
7401
7402 uint32_t
7403 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
7404         switch (mac_type) {
7405         case ixgbe_mac_X550_vf:
7406         case ixgbe_mac_X550EM_x_vf:
7407         case ixgbe_mac_X550EM_a_vf:
7408                 return IXGBE_VFMRQC;
7409         default:
7410                 return IXGBE_MRQC;
7411         }
7412 }
7413
7414 uint32_t
7415 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
7416         switch (mac_type) {
7417         case ixgbe_mac_X550_vf:
7418         case ixgbe_mac_X550EM_x_vf:
7419         case ixgbe_mac_X550EM_a_vf:
7420                 return IXGBE_VFRSSRK(i);
7421         default:
7422                 return IXGBE_RSSRK(i);
7423         }
7424 }
7425
7426 bool
7427 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
7428         switch (mac_type) {
7429         case ixgbe_mac_82599_vf:
7430         case ixgbe_mac_X540_vf:
7431                 return 0;
7432         default:
7433                 return 1;
7434         }
7435 }
7436
7437 static int
7438 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
7439                         struct rte_eth_dcb_info *dcb_info)
7440 {
7441         struct ixgbe_dcb_config *dcb_config =
7442                         IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
7443         struct ixgbe_dcb_tc_config *tc;
7444         struct rte_eth_dcb_tc_queue_mapping *tc_queue;
7445         uint8_t nb_tcs;
7446         uint8_t i, j;
7447
7448         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_DCB_FLAG)
7449                 dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
7450         else
7451                 dcb_info->nb_tcs = 1;
7452
7453         tc_queue = &dcb_info->tc_queue;
7454         nb_tcs = dcb_info->nb_tcs;
7455
7456         if (dcb_config->vt_mode) { /* vt is enabled*/
7457                 struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
7458                                 &dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
7459                 for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7460                         dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
7461                 if (RTE_ETH_DEV_SRIOV(dev).active > 0) {
7462                         for (j = 0; j < nb_tcs; j++) {
7463                                 tc_queue->tc_rxq[0][j].base = j;
7464                                 tc_queue->tc_rxq[0][j].nb_queue = 1;
7465                                 tc_queue->tc_txq[0][j].base = j;
7466                                 tc_queue->tc_txq[0][j].nb_queue = 1;
7467                         }
7468                 } else {
7469                         for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
7470                                 for (j = 0; j < nb_tcs; j++) {
7471                                         tc_queue->tc_rxq[i][j].base =
7472                                                 i * nb_tcs + j;
7473                                         tc_queue->tc_rxq[i][j].nb_queue = 1;
7474                                         tc_queue->tc_txq[i][j].base =
7475                                                 i * nb_tcs + j;
7476                                         tc_queue->tc_txq[i][j].nb_queue = 1;
7477                                 }
7478                         }
7479                 }
7480         } else { /* vt is disabled*/
7481                 struct rte_eth_dcb_rx_conf *rx_conf =
7482                                 &dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
7483                 for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7484                         dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
7485                 if (dcb_info->nb_tcs == ETH_4_TCS) {
7486                         for (i = 0; i < dcb_info->nb_tcs; i++) {
7487                                 dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
7488                                 dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7489                         }
7490                         dcb_info->tc_queue.tc_txq[0][0].base = 0;
7491                         dcb_info->tc_queue.tc_txq[0][1].base = 64;
7492                         dcb_info->tc_queue.tc_txq[0][2].base = 96;
7493                         dcb_info->tc_queue.tc_txq[0][3].base = 112;
7494                         dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
7495                         dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7496                         dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7497                         dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7498                 } else if (dcb_info->nb_tcs == ETH_8_TCS) {
7499                         for (i = 0; i < dcb_info->nb_tcs; i++) {
7500                                 dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
7501                                 dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7502                         }
7503                         dcb_info->tc_queue.tc_txq[0][0].base = 0;
7504                         dcb_info->tc_queue.tc_txq[0][1].base = 32;
7505                         dcb_info->tc_queue.tc_txq[0][2].base = 64;
7506                         dcb_info->tc_queue.tc_txq[0][3].base = 80;
7507                         dcb_info->tc_queue.tc_txq[0][4].base = 96;
7508                         dcb_info->tc_queue.tc_txq[0][5].base = 104;
7509                         dcb_info->tc_queue.tc_txq[0][6].base = 112;
7510                         dcb_info->tc_queue.tc_txq[0][7].base = 120;
7511                         dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
7512                         dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7513                         dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7514                         dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7515                         dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
7516                         dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
7517                         dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
7518                         dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
7519                 }
7520         }
7521         for (i = 0; i < dcb_info->nb_tcs; i++) {
7522                 tc = &dcb_config->tc_config[i];
7523                 dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
7524         }
7525         return 0;
7526 }
7527
7528 /* Update e-tag ether type */
7529 static int
7530 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
7531                             uint16_t ether_type)
7532 {
7533         uint32_t etag_etype;
7534
7535         if (hw->mac.type != ixgbe_mac_X550 &&
7536             hw->mac.type != ixgbe_mac_X550EM_x &&
7537             hw->mac.type != ixgbe_mac_X550EM_a) {
7538                 return -ENOTSUP;
7539         }
7540
7541         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7542         etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
7543         etag_etype |= ether_type;
7544         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7545         IXGBE_WRITE_FLUSH(hw);
7546
7547         return 0;
7548 }
7549
7550 /* Config l2 tunnel ether type */
7551 static int
7552 ixgbe_dev_l2_tunnel_eth_type_conf(struct rte_eth_dev *dev,
7553                                   struct rte_eth_l2_tunnel_conf *l2_tunnel)
7554 {
7555         int ret = 0;
7556         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7557         struct ixgbe_l2_tn_info *l2_tn_info =
7558                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7559
7560         if (l2_tunnel == NULL)
7561                 return -EINVAL;
7562
7563         switch (l2_tunnel->l2_tunnel_type) {
7564         case RTE_L2_TUNNEL_TYPE_E_TAG:
7565                 l2_tn_info->e_tag_ether_type = l2_tunnel->ether_type;
7566                 ret = ixgbe_update_e_tag_eth_type(hw, l2_tunnel->ether_type);
7567                 break;
7568         default:
7569                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7570                 ret = -EINVAL;
7571                 break;
7572         }
7573
7574         return ret;
7575 }
7576
7577 /* Enable e-tag tunnel */
7578 static int
7579 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
7580 {
7581         uint32_t etag_etype;
7582
7583         if (hw->mac.type != ixgbe_mac_X550 &&
7584             hw->mac.type != ixgbe_mac_X550EM_x &&
7585             hw->mac.type != ixgbe_mac_X550EM_a) {
7586                 return -ENOTSUP;
7587         }
7588
7589         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7590         etag_etype |= IXGBE_ETAG_ETYPE_VALID;
7591         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7592         IXGBE_WRITE_FLUSH(hw);
7593
7594         return 0;
7595 }
7596
7597 /* Enable l2 tunnel */
7598 static int
7599 ixgbe_dev_l2_tunnel_enable(struct rte_eth_dev *dev,
7600                            enum rte_eth_tunnel_type l2_tunnel_type)
7601 {
7602         int ret = 0;
7603         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7604         struct ixgbe_l2_tn_info *l2_tn_info =
7605                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7606
7607         switch (l2_tunnel_type) {
7608         case RTE_L2_TUNNEL_TYPE_E_TAG:
7609                 l2_tn_info->e_tag_en = TRUE;
7610                 ret = ixgbe_e_tag_enable(hw);
7611                 break;
7612         default:
7613                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7614                 ret = -EINVAL;
7615                 break;
7616         }
7617
7618         return ret;
7619 }
7620
7621 /* Disable e-tag tunnel */
7622 static int
7623 ixgbe_e_tag_disable(struct ixgbe_hw *hw)
7624 {
7625         uint32_t etag_etype;
7626
7627         if (hw->mac.type != ixgbe_mac_X550 &&
7628             hw->mac.type != ixgbe_mac_X550EM_x &&
7629             hw->mac.type != ixgbe_mac_X550EM_a) {
7630                 return -ENOTSUP;
7631         }
7632
7633         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7634         etag_etype &= ~IXGBE_ETAG_ETYPE_VALID;
7635         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7636         IXGBE_WRITE_FLUSH(hw);
7637
7638         return 0;
7639 }
7640
7641 /* Disable l2 tunnel */
7642 static int
7643 ixgbe_dev_l2_tunnel_disable(struct rte_eth_dev *dev,
7644                             enum rte_eth_tunnel_type l2_tunnel_type)
7645 {
7646         int ret = 0;
7647         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7648         struct ixgbe_l2_tn_info *l2_tn_info =
7649                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7650
7651         switch (l2_tunnel_type) {
7652         case RTE_L2_TUNNEL_TYPE_E_TAG:
7653                 l2_tn_info->e_tag_en = FALSE;
7654                 ret = ixgbe_e_tag_disable(hw);
7655                 break;
7656         default:
7657                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7658                 ret = -EINVAL;
7659                 break;
7660         }
7661
7662         return ret;
7663 }
7664
7665 static int
7666 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
7667                        struct rte_eth_l2_tunnel_conf *l2_tunnel)
7668 {
7669         int ret = 0;
7670         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7671         uint32_t i, rar_entries;
7672         uint32_t rar_low, rar_high;
7673
7674         if (hw->mac.type != ixgbe_mac_X550 &&
7675             hw->mac.type != ixgbe_mac_X550EM_x &&
7676             hw->mac.type != ixgbe_mac_X550EM_a) {
7677                 return -ENOTSUP;
7678         }
7679
7680         rar_entries = ixgbe_get_num_rx_addrs(hw);
7681
7682         for (i = 1; i < rar_entries; i++) {
7683                 rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7684                 rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
7685                 if ((rar_high & IXGBE_RAH_AV) &&
7686                     (rar_high & IXGBE_RAH_ADTYPE) &&
7687                     ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
7688                      l2_tunnel->tunnel_id)) {
7689                         IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
7690                         IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
7691
7692                         ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
7693
7694                         return ret;
7695                 }
7696         }
7697
7698         return ret;
7699 }
7700
7701 static int
7702 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
7703                        struct rte_eth_l2_tunnel_conf *l2_tunnel)
7704 {
7705         int ret = 0;
7706         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7707         uint32_t i, rar_entries;
7708         uint32_t rar_low, rar_high;
7709
7710         if (hw->mac.type != ixgbe_mac_X550 &&
7711             hw->mac.type != ixgbe_mac_X550EM_x &&
7712             hw->mac.type != ixgbe_mac_X550EM_a) {
7713                 return -ENOTSUP;
7714         }
7715
7716         /* One entry for one tunnel. Try to remove potential existing entry. */
7717         ixgbe_e_tag_filter_del(dev, l2_tunnel);
7718
7719         rar_entries = ixgbe_get_num_rx_addrs(hw);
7720
7721         for (i = 1; i < rar_entries; i++) {
7722                 rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7723                 if (rar_high & IXGBE_RAH_AV) {
7724                         continue;
7725                 } else {
7726                         ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
7727                         rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
7728                         rar_low = l2_tunnel->tunnel_id;
7729
7730                         IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
7731                         IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
7732
7733                         return ret;
7734                 }
7735         }
7736
7737         PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
7738                      " Please remove a rule before adding a new one.");
7739         return -EINVAL;
7740 }
7741
7742 static inline struct ixgbe_l2_tn_filter *
7743 ixgbe_l2_tn_filter_lookup(struct ixgbe_l2_tn_info *l2_tn_info,
7744                           struct ixgbe_l2_tn_key *key)
7745 {
7746         int ret;
7747
7748         ret = rte_hash_lookup(l2_tn_info->hash_handle, (const void *)key);
7749         if (ret < 0)
7750                 return NULL;
7751
7752         return l2_tn_info->hash_map[ret];
7753 }
7754
7755 static inline int
7756 ixgbe_insert_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7757                           struct ixgbe_l2_tn_filter *l2_tn_filter)
7758 {
7759         int ret;
7760
7761         ret = rte_hash_add_key(l2_tn_info->hash_handle,
7762                                &l2_tn_filter->key);
7763
7764         if (ret < 0) {
7765                 PMD_DRV_LOG(ERR,
7766                             "Failed to insert L2 tunnel filter"
7767                             " to hash table %d!",
7768                             ret);
7769                 return ret;
7770         }
7771
7772         l2_tn_info->hash_map[ret] = l2_tn_filter;
7773
7774         TAILQ_INSERT_TAIL(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7775
7776         return 0;
7777 }
7778
7779 static inline int
7780 ixgbe_remove_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7781                           struct ixgbe_l2_tn_key *key)
7782 {
7783         int ret;
7784         struct ixgbe_l2_tn_filter *l2_tn_filter;
7785
7786         ret = rte_hash_del_key(l2_tn_info->hash_handle, key);
7787
7788         if (ret < 0) {
7789                 PMD_DRV_LOG(ERR,
7790                             "No such L2 tunnel filter to delete %d!",
7791                             ret);
7792                 return ret;
7793         }
7794
7795         l2_tn_filter = l2_tn_info->hash_map[ret];
7796         l2_tn_info->hash_map[ret] = NULL;
7797
7798         TAILQ_REMOVE(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7799         rte_free(l2_tn_filter);
7800
7801         return 0;
7802 }
7803
7804 /* Add l2 tunnel filter */
7805 int
7806 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
7807                                struct rte_eth_l2_tunnel_conf *l2_tunnel,
7808                                bool restore)
7809 {
7810         int ret;
7811         struct ixgbe_l2_tn_info *l2_tn_info =
7812                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7813         struct ixgbe_l2_tn_key key;
7814         struct ixgbe_l2_tn_filter *node;
7815
7816         if (!restore) {
7817                 key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7818                 key.tn_id = l2_tunnel->tunnel_id;
7819
7820                 node = ixgbe_l2_tn_filter_lookup(l2_tn_info, &key);
7821
7822                 if (node) {
7823                         PMD_DRV_LOG(ERR,
7824                                     "The L2 tunnel filter already exists!");
7825                         return -EINVAL;
7826                 }
7827
7828                 node = rte_zmalloc("ixgbe_l2_tn",
7829                                    sizeof(struct ixgbe_l2_tn_filter),
7830                                    0);
7831                 if (!node)
7832                         return -ENOMEM;
7833
7834                 rte_memcpy(&node->key,
7835                                  &key,
7836                                  sizeof(struct ixgbe_l2_tn_key));
7837                 node->pool = l2_tunnel->pool;
7838                 ret = ixgbe_insert_l2_tn_filter(l2_tn_info, node);
7839                 if (ret < 0) {
7840                         rte_free(node);
7841                         return ret;
7842                 }
7843         }
7844
7845         switch (l2_tunnel->l2_tunnel_type) {
7846         case RTE_L2_TUNNEL_TYPE_E_TAG:
7847                 ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
7848                 break;
7849         default:
7850                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7851                 ret = -EINVAL;
7852                 break;
7853         }
7854
7855         if ((!restore) && (ret < 0))
7856                 (void)ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7857
7858         return ret;
7859 }
7860
7861 /* Delete l2 tunnel filter */
7862 int
7863 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
7864                                struct rte_eth_l2_tunnel_conf *l2_tunnel)
7865 {
7866         int ret;
7867         struct ixgbe_l2_tn_info *l2_tn_info =
7868                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7869         struct ixgbe_l2_tn_key key;
7870
7871         key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7872         key.tn_id = l2_tunnel->tunnel_id;
7873         ret = ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7874         if (ret < 0)
7875                 return ret;
7876
7877         switch (l2_tunnel->l2_tunnel_type) {
7878         case RTE_L2_TUNNEL_TYPE_E_TAG:
7879                 ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
7880                 break;
7881         default:
7882                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7883                 ret = -EINVAL;
7884                 break;
7885         }
7886
7887         return ret;
7888 }
7889
7890 /**
7891  * ixgbe_dev_l2_tunnel_filter_handle - Handle operations for l2 tunnel filter.
7892  * @dev: pointer to rte_eth_dev structure
7893  * @filter_op:operation will be taken.
7894  * @arg: a pointer to specific structure corresponding to the filter_op
7895  */
7896 static int
7897 ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
7898                                   enum rte_filter_op filter_op,
7899                                   void *arg)
7900 {
7901         int ret;
7902
7903         if (filter_op == RTE_ETH_FILTER_NOP)
7904                 return 0;
7905
7906         if (arg == NULL) {
7907                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
7908                             filter_op);
7909                 return -EINVAL;
7910         }
7911
7912         switch (filter_op) {
7913         case RTE_ETH_FILTER_ADD:
7914                 ret = ixgbe_dev_l2_tunnel_filter_add
7915                         (dev,
7916                          (struct rte_eth_l2_tunnel_conf *)arg,
7917                          FALSE);
7918                 break;
7919         case RTE_ETH_FILTER_DELETE:
7920                 ret = ixgbe_dev_l2_tunnel_filter_del
7921                         (dev,
7922                          (struct rte_eth_l2_tunnel_conf *)arg);
7923                 break;
7924         default:
7925                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
7926                 ret = -EINVAL;
7927                 break;
7928         }
7929         return ret;
7930 }
7931
7932 static int
7933 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
7934 {
7935         int ret = 0;
7936         uint32_t ctrl;
7937         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7938
7939         if (hw->mac.type != ixgbe_mac_X550 &&
7940             hw->mac.type != ixgbe_mac_X550EM_x &&
7941             hw->mac.type != ixgbe_mac_X550EM_a) {
7942                 return -ENOTSUP;
7943         }
7944
7945         ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
7946         ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
7947         if (en)
7948                 ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
7949         IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
7950
7951         return ret;
7952 }
7953
7954 /* Enable l2 tunnel forwarding */
7955 static int
7956 ixgbe_dev_l2_tunnel_forwarding_enable
7957         (struct rte_eth_dev *dev,
7958          enum rte_eth_tunnel_type l2_tunnel_type)
7959 {
7960         struct ixgbe_l2_tn_info *l2_tn_info =
7961                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7962         int ret = 0;
7963
7964         switch (l2_tunnel_type) {
7965         case RTE_L2_TUNNEL_TYPE_E_TAG:
7966                 l2_tn_info->e_tag_fwd_en = TRUE;
7967                 ret = ixgbe_e_tag_forwarding_en_dis(dev, 1);
7968                 break;
7969         default:
7970                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7971                 ret = -EINVAL;
7972                 break;
7973         }
7974
7975         return ret;
7976 }
7977
7978 /* Disable l2 tunnel forwarding */
7979 static int
7980 ixgbe_dev_l2_tunnel_forwarding_disable
7981         (struct rte_eth_dev *dev,
7982          enum rte_eth_tunnel_type l2_tunnel_type)
7983 {
7984         struct ixgbe_l2_tn_info *l2_tn_info =
7985                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7986         int ret = 0;
7987
7988         switch (l2_tunnel_type) {
7989         case RTE_L2_TUNNEL_TYPE_E_TAG:
7990                 l2_tn_info->e_tag_fwd_en = FALSE;
7991                 ret = ixgbe_e_tag_forwarding_en_dis(dev, 0);
7992                 break;
7993         default:
7994                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7995                 ret = -EINVAL;
7996                 break;
7997         }
7998
7999         return ret;
8000 }
8001
8002 static int
8003 ixgbe_e_tag_insertion_en_dis(struct rte_eth_dev *dev,
8004                              struct rte_eth_l2_tunnel_conf *l2_tunnel,
8005                              bool en)
8006 {
8007         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
8008         int ret = 0;
8009         uint32_t vmtir, vmvir;
8010         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8011
8012         if (l2_tunnel->vf_id >= pci_dev->max_vfs) {
8013                 PMD_DRV_LOG(ERR,
8014                             "VF id %u should be less than %u",
8015                             l2_tunnel->vf_id,
8016                             pci_dev->max_vfs);
8017                 return -EINVAL;
8018         }
8019
8020         if (hw->mac.type != ixgbe_mac_X550 &&
8021             hw->mac.type != ixgbe_mac_X550EM_x &&
8022             hw->mac.type != ixgbe_mac_X550EM_a) {
8023                 return -ENOTSUP;
8024         }
8025
8026         if (en)
8027                 vmtir = l2_tunnel->tunnel_id;
8028         else
8029                 vmtir = 0;
8030
8031         IXGBE_WRITE_REG(hw, IXGBE_VMTIR(l2_tunnel->vf_id), vmtir);
8032
8033         vmvir = IXGBE_READ_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id));
8034         vmvir &= ~IXGBE_VMVIR_TAGA_MASK;
8035         if (en)
8036                 vmvir |= IXGBE_VMVIR_TAGA_ETAG_INSERT;
8037         IXGBE_WRITE_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id), vmvir);
8038
8039         return ret;
8040 }
8041
8042 /* Enable l2 tunnel tag insertion */
8043 static int
8044 ixgbe_dev_l2_tunnel_insertion_enable(struct rte_eth_dev *dev,
8045                                      struct rte_eth_l2_tunnel_conf *l2_tunnel)
8046 {
8047         int ret = 0;
8048
8049         switch (l2_tunnel->l2_tunnel_type) {
8050         case RTE_L2_TUNNEL_TYPE_E_TAG:
8051                 ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 1);
8052                 break;
8053         default:
8054                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8055                 ret = -EINVAL;
8056                 break;
8057         }
8058
8059         return ret;
8060 }
8061
8062 /* Disable l2 tunnel tag insertion */
8063 static int
8064 ixgbe_dev_l2_tunnel_insertion_disable
8065         (struct rte_eth_dev *dev,
8066          struct rte_eth_l2_tunnel_conf *l2_tunnel)
8067 {
8068         int ret = 0;
8069
8070         switch (l2_tunnel->l2_tunnel_type) {
8071         case RTE_L2_TUNNEL_TYPE_E_TAG:
8072                 ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 0);
8073                 break;
8074         default:
8075                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8076                 ret = -EINVAL;
8077                 break;
8078         }
8079
8080         return ret;
8081 }
8082
8083 static int
8084 ixgbe_e_tag_stripping_en_dis(struct rte_eth_dev *dev,
8085                              bool en)
8086 {
8087         int ret = 0;
8088         uint32_t qde;
8089         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8090
8091         if (hw->mac.type != ixgbe_mac_X550 &&
8092             hw->mac.type != ixgbe_mac_X550EM_x &&
8093             hw->mac.type != ixgbe_mac_X550EM_a) {
8094                 return -ENOTSUP;
8095         }
8096
8097         qde = IXGBE_READ_REG(hw, IXGBE_QDE);
8098         if (en)
8099                 qde |= IXGBE_QDE_STRIP_TAG;
8100         else
8101                 qde &= ~IXGBE_QDE_STRIP_TAG;
8102         qde &= ~IXGBE_QDE_READ;
8103         qde |= IXGBE_QDE_WRITE;
8104         IXGBE_WRITE_REG(hw, IXGBE_QDE, qde);
8105
8106         return ret;
8107 }
8108
8109 /* Enable l2 tunnel tag stripping */
8110 static int
8111 ixgbe_dev_l2_tunnel_stripping_enable
8112         (struct rte_eth_dev *dev,
8113          enum rte_eth_tunnel_type l2_tunnel_type)
8114 {
8115         int ret = 0;
8116
8117         switch (l2_tunnel_type) {
8118         case RTE_L2_TUNNEL_TYPE_E_TAG:
8119                 ret = ixgbe_e_tag_stripping_en_dis(dev, 1);
8120                 break;
8121         default:
8122                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8123                 ret = -EINVAL;
8124                 break;
8125         }
8126
8127         return ret;
8128 }
8129
8130 /* Disable l2 tunnel tag stripping */
8131 static int
8132 ixgbe_dev_l2_tunnel_stripping_disable
8133         (struct rte_eth_dev *dev,
8134          enum rte_eth_tunnel_type l2_tunnel_type)
8135 {
8136         int ret = 0;
8137
8138         switch (l2_tunnel_type) {
8139         case RTE_L2_TUNNEL_TYPE_E_TAG:
8140                 ret = ixgbe_e_tag_stripping_en_dis(dev, 0);
8141                 break;
8142         default:
8143                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8144                 ret = -EINVAL;
8145                 break;
8146         }
8147
8148         return ret;
8149 }
8150
8151 /* Enable/disable l2 tunnel offload functions */
8152 static int
8153 ixgbe_dev_l2_tunnel_offload_set
8154         (struct rte_eth_dev *dev,
8155          struct rte_eth_l2_tunnel_conf *l2_tunnel,
8156          uint32_t mask,
8157          uint8_t en)
8158 {
8159         int ret = 0;
8160
8161         if (l2_tunnel == NULL)
8162                 return -EINVAL;
8163
8164         ret = -EINVAL;
8165         if (mask & ETH_L2_TUNNEL_ENABLE_MASK) {
8166                 if (en)
8167                         ret = ixgbe_dev_l2_tunnel_enable(
8168                                 dev,
8169                                 l2_tunnel->l2_tunnel_type);
8170                 else
8171                         ret = ixgbe_dev_l2_tunnel_disable(
8172                                 dev,
8173                                 l2_tunnel->l2_tunnel_type);
8174         }
8175
8176         if (mask & ETH_L2_TUNNEL_INSERTION_MASK) {
8177                 if (en)
8178                         ret = ixgbe_dev_l2_tunnel_insertion_enable(
8179                                 dev,
8180                                 l2_tunnel);
8181                 else
8182                         ret = ixgbe_dev_l2_tunnel_insertion_disable(
8183                                 dev,
8184                                 l2_tunnel);
8185         }
8186
8187         if (mask & ETH_L2_TUNNEL_STRIPPING_MASK) {
8188                 if (en)
8189                         ret = ixgbe_dev_l2_tunnel_stripping_enable(
8190                                 dev,
8191                                 l2_tunnel->l2_tunnel_type);
8192                 else
8193                         ret = ixgbe_dev_l2_tunnel_stripping_disable(
8194                                 dev,
8195                                 l2_tunnel->l2_tunnel_type);
8196         }
8197
8198         if (mask & ETH_L2_TUNNEL_FORWARDING_MASK) {
8199                 if (en)
8200                         ret = ixgbe_dev_l2_tunnel_forwarding_enable(
8201                                 dev,
8202                                 l2_tunnel->l2_tunnel_type);
8203                 else
8204                         ret = ixgbe_dev_l2_tunnel_forwarding_disable(
8205                                 dev,
8206                                 l2_tunnel->l2_tunnel_type);
8207         }
8208
8209         return ret;
8210 }
8211
8212 static int
8213 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
8214                         uint16_t port)
8215 {
8216         IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
8217         IXGBE_WRITE_FLUSH(hw);
8218
8219         return 0;
8220 }
8221
8222 /* There's only one register for VxLAN UDP port.
8223  * So, we cannot add several ports. Will update it.
8224  */
8225 static int
8226 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
8227                      uint16_t port)
8228 {
8229         if (port == 0) {
8230                 PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
8231                 return -EINVAL;
8232         }
8233
8234         return ixgbe_update_vxlan_port(hw, port);
8235 }
8236
8237 /* We cannot delete the VxLAN port. For there's a register for VxLAN
8238  * UDP port, it must have a value.
8239  * So, will reset it to the original value 0.
8240  */
8241 static int
8242 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
8243                      uint16_t port)
8244 {
8245         uint16_t cur_port;
8246
8247         cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
8248
8249         if (cur_port != port) {
8250                 PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
8251                 return -EINVAL;
8252         }
8253
8254         return ixgbe_update_vxlan_port(hw, 0);
8255 }
8256
8257 /* Add UDP tunneling port */
8258 static int
8259 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
8260                               struct rte_eth_udp_tunnel *udp_tunnel)
8261 {
8262         int ret = 0;
8263         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8264
8265         if (hw->mac.type != ixgbe_mac_X550 &&
8266             hw->mac.type != ixgbe_mac_X550EM_x &&
8267             hw->mac.type != ixgbe_mac_X550EM_a) {
8268                 return -ENOTSUP;
8269         }
8270
8271         if (udp_tunnel == NULL)
8272                 return -EINVAL;
8273
8274         switch (udp_tunnel->prot_type) {
8275         case RTE_TUNNEL_TYPE_VXLAN:
8276                 ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
8277                 break;
8278
8279         case RTE_TUNNEL_TYPE_GENEVE:
8280         case RTE_TUNNEL_TYPE_TEREDO:
8281                 PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
8282                 ret = -EINVAL;
8283                 break;
8284
8285         default:
8286                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8287                 ret = -EINVAL;
8288                 break;
8289         }
8290
8291         return ret;
8292 }
8293
8294 /* Remove UDP tunneling port */
8295 static int
8296 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
8297                               struct rte_eth_udp_tunnel *udp_tunnel)
8298 {
8299         int ret = 0;
8300         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8301
8302         if (hw->mac.type != ixgbe_mac_X550 &&
8303             hw->mac.type != ixgbe_mac_X550EM_x &&
8304             hw->mac.type != ixgbe_mac_X550EM_a) {
8305                 return -ENOTSUP;
8306         }
8307
8308         if (udp_tunnel == NULL)
8309                 return -EINVAL;
8310
8311         switch (udp_tunnel->prot_type) {
8312         case RTE_TUNNEL_TYPE_VXLAN:
8313                 ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
8314                 break;
8315         case RTE_TUNNEL_TYPE_GENEVE:
8316         case RTE_TUNNEL_TYPE_TEREDO:
8317                 PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
8318                 ret = -EINVAL;
8319                 break;
8320         default:
8321                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8322                 ret = -EINVAL;
8323                 break;
8324         }
8325
8326         return ret;
8327 }
8328
8329 static void
8330 ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev)
8331 {
8332         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8333
8334         hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_PROMISC);
8335 }
8336
8337 static void
8338 ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev)
8339 {
8340         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8341
8342         hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_NONE);
8343 }
8344
8345 static void
8346 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
8347 {
8348         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8349
8350         hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_ALLMULTI);
8351 }
8352
8353 static void
8354 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
8355 {
8356         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8357
8358         hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_MULTI);
8359 }
8360
8361 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
8362 {
8363         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8364         u32 in_msg = 0;
8365
8366         /* peek the message first */
8367         in_msg = IXGBE_READ_REG(hw, IXGBE_VFMBMEM);
8368
8369         /* PF reset VF event */
8370         if (in_msg == IXGBE_PF_CONTROL_MSG) {
8371                 /* dummy mbx read to ack pf */
8372                 if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
8373                         return;
8374                 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
8375                                               NULL);
8376         }
8377 }
8378
8379 static int
8380 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
8381 {
8382         uint32_t eicr;
8383         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8384         struct ixgbe_interrupt *intr =
8385                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8386         ixgbevf_intr_disable(dev);
8387
8388         /* read-on-clear nic registers here */
8389         eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
8390         intr->flags = 0;
8391
8392         /* only one misc vector supported - mailbox */
8393         eicr &= IXGBE_VTEICR_MASK;
8394         if (eicr == IXGBE_MISC_VEC_ID)
8395                 intr->flags |= IXGBE_FLAG_MAILBOX;
8396
8397         return 0;
8398 }
8399
8400 static int
8401 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
8402 {
8403         struct ixgbe_interrupt *intr =
8404                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8405
8406         if (intr->flags & IXGBE_FLAG_MAILBOX) {
8407                 ixgbevf_mbx_process(dev);
8408                 intr->flags &= ~IXGBE_FLAG_MAILBOX;
8409         }
8410
8411         ixgbevf_intr_enable(dev);
8412
8413         return 0;
8414 }
8415
8416 static void
8417 ixgbevf_dev_interrupt_handler(void *param)
8418 {
8419         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
8420
8421         ixgbevf_dev_interrupt_get_status(dev);
8422         ixgbevf_dev_interrupt_action(dev);
8423 }
8424
8425 /**
8426  *  ixgbe_disable_sec_tx_path_generic - Stops the transmit data path
8427  *  @hw: pointer to hardware structure
8428  *
8429  *  Stops the transmit data path and waits for the HW to internally empty
8430  *  the Tx security block
8431  **/
8432 int ixgbe_disable_sec_tx_path_generic(struct ixgbe_hw *hw)
8433 {
8434 #define IXGBE_MAX_SECTX_POLL 40
8435
8436         int i;
8437         int sectxreg;
8438
8439         sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8440         sectxreg |= IXGBE_SECTXCTRL_TX_DIS;
8441         IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8442         for (i = 0; i < IXGBE_MAX_SECTX_POLL; i++) {
8443                 sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXSTAT);
8444                 if (sectxreg & IXGBE_SECTXSTAT_SECTX_RDY)
8445                         break;
8446                 /* Use interrupt-safe sleep just in case */
8447                 usec_delay(1000);
8448         }
8449
8450         /* For informational purposes only */
8451         if (i >= IXGBE_MAX_SECTX_POLL)
8452                 PMD_DRV_LOG(DEBUG, "Tx unit being enabled before security "
8453                          "path fully disabled.  Continuing with init.");
8454
8455         return IXGBE_SUCCESS;
8456 }
8457
8458 /**
8459  *  ixgbe_enable_sec_tx_path_generic - Enables the transmit data path
8460  *  @hw: pointer to hardware structure
8461  *
8462  *  Enables the transmit data path.
8463  **/
8464 int ixgbe_enable_sec_tx_path_generic(struct ixgbe_hw *hw)
8465 {
8466         uint32_t sectxreg;
8467
8468         sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8469         sectxreg &= ~IXGBE_SECTXCTRL_TX_DIS;
8470         IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8471         IXGBE_WRITE_FLUSH(hw);
8472
8473         return IXGBE_SUCCESS;
8474 }
8475
8476 /* restore n-tuple filter */
8477 static inline void
8478 ixgbe_ntuple_filter_restore(struct rte_eth_dev *dev)
8479 {
8480         struct ixgbe_filter_info *filter_info =
8481                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8482         struct ixgbe_5tuple_filter *node;
8483
8484         TAILQ_FOREACH(node, &filter_info->fivetuple_list, entries) {
8485                 ixgbe_inject_5tuple_filter(dev, node);
8486         }
8487 }
8488
8489 /* restore ethernet type filter */
8490 static inline void
8491 ixgbe_ethertype_filter_restore(struct rte_eth_dev *dev)
8492 {
8493         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8494         struct ixgbe_filter_info *filter_info =
8495                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8496         int i;
8497
8498         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8499                 if (filter_info->ethertype_mask & (1 << i)) {
8500                         IXGBE_WRITE_REG(hw, IXGBE_ETQF(i),
8501                                         filter_info->ethertype_filters[i].etqf);
8502                         IXGBE_WRITE_REG(hw, IXGBE_ETQS(i),
8503                                         filter_info->ethertype_filters[i].etqs);
8504                         IXGBE_WRITE_FLUSH(hw);
8505                 }
8506         }
8507 }
8508
8509 /* restore SYN filter */
8510 static inline void
8511 ixgbe_syn_filter_restore(struct rte_eth_dev *dev)
8512 {
8513         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8514         struct ixgbe_filter_info *filter_info =
8515                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8516         uint32_t synqf;
8517
8518         synqf = filter_info->syn_info;
8519
8520         if (synqf & IXGBE_SYN_FILTER_ENABLE) {
8521                 IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
8522                 IXGBE_WRITE_FLUSH(hw);
8523         }
8524 }
8525
8526 /* restore L2 tunnel filter */
8527 static inline void
8528 ixgbe_l2_tn_filter_restore(struct rte_eth_dev *dev)
8529 {
8530         struct ixgbe_l2_tn_info *l2_tn_info =
8531                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8532         struct ixgbe_l2_tn_filter *node;
8533         struct rte_eth_l2_tunnel_conf l2_tn_conf;
8534
8535         TAILQ_FOREACH(node, &l2_tn_info->l2_tn_list, entries) {
8536                 l2_tn_conf.l2_tunnel_type = node->key.l2_tn_type;
8537                 l2_tn_conf.tunnel_id      = node->key.tn_id;
8538                 l2_tn_conf.pool           = node->pool;
8539                 (void)ixgbe_dev_l2_tunnel_filter_add(dev, &l2_tn_conf, TRUE);
8540         }
8541 }
8542
8543 /* restore rss filter */
8544 static inline void
8545 ixgbe_rss_filter_restore(struct rte_eth_dev *dev)
8546 {
8547         struct ixgbe_filter_info *filter_info =
8548                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8549
8550         if (filter_info->rss_info.conf.queue_num)
8551                 ixgbe_config_rss_filter(dev,
8552                         &filter_info->rss_info, TRUE);
8553 }
8554
8555 static int
8556 ixgbe_filter_restore(struct rte_eth_dev *dev)
8557 {
8558         ixgbe_ntuple_filter_restore(dev);
8559         ixgbe_ethertype_filter_restore(dev);
8560         ixgbe_syn_filter_restore(dev);
8561         ixgbe_fdir_filter_restore(dev);
8562         ixgbe_l2_tn_filter_restore(dev);
8563         ixgbe_rss_filter_restore(dev);
8564
8565         return 0;
8566 }
8567
8568 static void
8569 ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev)
8570 {
8571         struct ixgbe_l2_tn_info *l2_tn_info =
8572                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8573         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8574
8575         if (l2_tn_info->e_tag_en)
8576                 (void)ixgbe_e_tag_enable(hw);
8577
8578         if (l2_tn_info->e_tag_fwd_en)
8579                 (void)ixgbe_e_tag_forwarding_en_dis(dev, 1);
8580
8581         (void)ixgbe_update_e_tag_eth_type(hw, l2_tn_info->e_tag_ether_type);
8582 }
8583
8584 /* remove all the n-tuple filters */
8585 void
8586 ixgbe_clear_all_ntuple_filter(struct rte_eth_dev *dev)
8587 {
8588         struct ixgbe_filter_info *filter_info =
8589                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8590         struct ixgbe_5tuple_filter *p_5tuple;
8591
8592         while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
8593                 ixgbe_remove_5tuple_filter(dev, p_5tuple);
8594 }
8595
8596 /* remove all the ether type filters */
8597 void
8598 ixgbe_clear_all_ethertype_filter(struct rte_eth_dev *dev)
8599 {
8600         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8601         struct ixgbe_filter_info *filter_info =
8602                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8603         int i;
8604
8605         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8606                 if (filter_info->ethertype_mask & (1 << i) &&
8607                     !filter_info->ethertype_filters[i].conf) {
8608                         (void)ixgbe_ethertype_filter_remove(filter_info,
8609                                                             (uint8_t)i);
8610                         IXGBE_WRITE_REG(hw, IXGBE_ETQF(i), 0);
8611                         IXGBE_WRITE_REG(hw, IXGBE_ETQS(i), 0);
8612                         IXGBE_WRITE_FLUSH(hw);
8613                 }
8614         }
8615 }
8616
8617 /* remove the SYN filter */
8618 void
8619 ixgbe_clear_syn_filter(struct rte_eth_dev *dev)
8620 {
8621         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8622         struct ixgbe_filter_info *filter_info =
8623                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8624
8625         if (filter_info->syn_info & IXGBE_SYN_FILTER_ENABLE) {
8626                 filter_info->syn_info = 0;
8627
8628                 IXGBE_WRITE_REG(hw, IXGBE_SYNQF, 0);
8629                 IXGBE_WRITE_FLUSH(hw);
8630         }
8631 }
8632
8633 /* remove all the L2 tunnel filters */
8634 int
8635 ixgbe_clear_all_l2_tn_filter(struct rte_eth_dev *dev)
8636 {
8637         struct ixgbe_l2_tn_info *l2_tn_info =
8638                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8639         struct ixgbe_l2_tn_filter *l2_tn_filter;
8640         struct rte_eth_l2_tunnel_conf l2_tn_conf;
8641         int ret = 0;
8642
8643         while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
8644                 l2_tn_conf.l2_tunnel_type = l2_tn_filter->key.l2_tn_type;
8645                 l2_tn_conf.tunnel_id      = l2_tn_filter->key.tn_id;
8646                 l2_tn_conf.pool           = l2_tn_filter->pool;
8647                 ret = ixgbe_dev_l2_tunnel_filter_del(dev, &l2_tn_conf);
8648                 if (ret < 0)
8649                         return ret;
8650         }
8651
8652         return 0;
8653 }
8654
8655 RTE_PMD_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd);
8656 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
8657 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe, "* igb_uio | uio_pci_generic | vfio-pci");
8658 RTE_PMD_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd);
8659 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);
8660 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe_vf, "* igb_uio | vfio-pci");
8661
8662 RTE_INIT(ixgbe_init_log)
8663 {
8664         ixgbe_logtype_init = rte_log_register("pmd.net.ixgbe.init");
8665         if (ixgbe_logtype_init >= 0)
8666                 rte_log_set_level(ixgbe_logtype_init, RTE_LOG_NOTICE);
8667         ixgbe_logtype_driver = rte_log_register("pmd.net.ixgbe.driver");
8668         if (ixgbe_logtype_driver >= 0)
8669                 rte_log_set_level(ixgbe_logtype_driver, RTE_LOG_NOTICE);
8670 }