drivers/net: fix FW version query
[dpdk.git] / drivers / net / ixgbe / ixgbe_ethdev.c
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2017 Intel Corporation
3  */
4
5 #include <sys/queue.h>
6 #include <stdio.h>
7 #include <errno.h>
8 #include <stdint.h>
9 #include <string.h>
10 #include <unistd.h>
11 #include <stdarg.h>
12 #include <inttypes.h>
13 #include <netinet/in.h>
14 #include <rte_string_fns.h>
15 #include <rte_byteorder.h>
16 #include <rte_common.h>
17 #include <rte_cycles.h>
18
19 #include <rte_interrupts.h>
20 #include <rte_log.h>
21 #include <rte_debug.h>
22 #include <rte_pci.h>
23 #include <rte_bus_pci.h>
24 #include <rte_branch_prediction.h>
25 #include <rte_memory.h>
26 #include <rte_kvargs.h>
27 #include <rte_eal.h>
28 #include <rte_alarm.h>
29 #include <rte_ether.h>
30 #include <ethdev_driver.h>
31 #include <ethdev_pci.h>
32 #include <rte_malloc.h>
33 #include <rte_random.h>
34 #include <rte_dev.h>
35 #include <rte_hash_crc.h>
36 #ifdef RTE_LIB_SECURITY
37 #include <rte_security_driver.h>
38 #endif
39
40 #include "ixgbe_logs.h"
41 #include "base/ixgbe_api.h"
42 #include "base/ixgbe_vf.h"
43 #include "base/ixgbe_common.h"
44 #include "ixgbe_ethdev.h"
45 #include "ixgbe_bypass.h"
46 #include "ixgbe_rxtx.h"
47 #include "base/ixgbe_type.h"
48 #include "base/ixgbe_phy.h"
49 #include "ixgbe_regs.h"
50
51 /*
52  * High threshold controlling when to start sending XOFF frames. Must be at
53  * least 8 bytes less than receive packet buffer size. This value is in units
54  * of 1024 bytes.
55  */
56 #define IXGBE_FC_HI    0x80
57
58 /*
59  * Low threshold controlling when to start sending XON frames. This value is
60  * in units of 1024 bytes.
61  */
62 #define IXGBE_FC_LO    0x40
63
64 /* Timer value included in XOFF frames. */
65 #define IXGBE_FC_PAUSE 0x680
66
67 /*Default value of Max Rx Queue*/
68 #define IXGBE_MAX_RX_QUEUE_NUM 128
69
70 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
71 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
72 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
73
74 #define IXGBE_MMW_SIZE_DEFAULT        0x4
75 #define IXGBE_MMW_SIZE_JUMBO_FRAME    0x14
76 #define IXGBE_MAX_RING_DESC           4096 /* replicate define from rxtx */
77
78 /*
79  *  Default values for RX/TX configuration
80  */
81 #define IXGBE_DEFAULT_RX_FREE_THRESH  32
82 #define IXGBE_DEFAULT_RX_PTHRESH      8
83 #define IXGBE_DEFAULT_RX_HTHRESH      8
84 #define IXGBE_DEFAULT_RX_WTHRESH      0
85
86 #define IXGBE_DEFAULT_TX_FREE_THRESH  32
87 #define IXGBE_DEFAULT_TX_PTHRESH      32
88 #define IXGBE_DEFAULT_TX_HTHRESH      0
89 #define IXGBE_DEFAULT_TX_WTHRESH      0
90 #define IXGBE_DEFAULT_TX_RSBIT_THRESH 32
91
92 /* Bit shift and mask */
93 #define IXGBE_4_BIT_WIDTH  (CHAR_BIT / 2)
94 #define IXGBE_4_BIT_MASK   RTE_LEN2MASK(IXGBE_4_BIT_WIDTH, uint8_t)
95 #define IXGBE_8_BIT_WIDTH  CHAR_BIT
96 #define IXGBE_8_BIT_MASK   UINT8_MAX
97
98 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
99
100 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
101
102 /* Additional timesync values. */
103 #define NSEC_PER_SEC             1000000000L
104 #define IXGBE_INCVAL_10GB        0x66666666
105 #define IXGBE_INCVAL_1GB         0x40000000
106 #define IXGBE_INCVAL_100         0x50000000
107 #define IXGBE_INCVAL_SHIFT_10GB  28
108 #define IXGBE_INCVAL_SHIFT_1GB   24
109 #define IXGBE_INCVAL_SHIFT_100   21
110 #define IXGBE_INCVAL_SHIFT_82599 7
111 #define IXGBE_INCPER_SHIFT_82599 24
112
113 #define IXGBE_CYCLECOUNTER_MASK   0xffffffffffffffffULL
114
115 #define IXGBE_VT_CTL_POOLING_MODE_MASK         0x00030000
116 #define IXGBE_VT_CTL_POOLING_MODE_ETAG         0x00010000
117 #define IXGBE_ETAG_ETYPE                       0x00005084
118 #define IXGBE_ETAG_ETYPE_MASK                  0x0000ffff
119 #define IXGBE_ETAG_ETYPE_VALID                 0x80000000
120 #define IXGBE_RAH_ADTYPE                       0x40000000
121 #define IXGBE_RAL_ETAG_FILTER_MASK             0x00003fff
122 #define IXGBE_VMVIR_TAGA_MASK                  0x18000000
123 #define IXGBE_VMVIR_TAGA_ETAG_INSERT           0x08000000
124 #define IXGBE_VMTIR(_i) (0x00017000 + ((_i) * 4)) /* 64 of these (0-63) */
125 #define IXGBE_QDE_STRIP_TAG                    0x00000004
126 #define IXGBE_VTEICR_MASK                      0x07
127
128 #define IXGBE_EXVET_VET_EXT_SHIFT              16
129 #define IXGBE_DMATXCTL_VT_MASK                 0xFFFF0000
130
131 #define IXGBEVF_DEVARG_PFLINK_FULLCHK           "pflink_fullchk"
132
133 static const char * const ixgbevf_valid_arguments[] = {
134         IXGBEVF_DEVARG_PFLINK_FULLCHK,
135         NULL
136 };
137
138 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params);
139 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
140 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev);
141 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev);
142 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev);
143 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev);
144 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev);
145 static int ixgbe_dev_configure(struct rte_eth_dev *dev);
146 static int ixgbe_dev_start(struct rte_eth_dev *dev);
147 static int ixgbe_dev_stop(struct rte_eth_dev *dev);
148 static int ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
149 static int ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
150 static int ixgbe_dev_close(struct rte_eth_dev *dev);
151 static int ixgbe_dev_reset(struct rte_eth_dev *dev);
152 static int ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
153 static int ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
154 static int ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
155 static int ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
156 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
157                                 int wait_to_complete);
158 static int ixgbe_dev_stats_get(struct rte_eth_dev *dev,
159                                 struct rte_eth_stats *stats);
160 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
161                                 struct rte_eth_xstat *xstats, unsigned n);
162 static int ixgbevf_dev_xstats_get(struct rte_eth_dev *dev,
163                                   struct rte_eth_xstat *xstats, unsigned n);
164 static int
165 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
166                 uint64_t *values, unsigned int n);
167 static int ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
168 static int ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
169 static int ixgbe_dev_xstats_get_names(struct rte_eth_dev *dev,
170         struct rte_eth_xstat_name *xstats_names,
171         unsigned int size);
172 static int ixgbevf_dev_xstats_get_names(struct rte_eth_dev *dev,
173         struct rte_eth_xstat_name *xstats_names, unsigned limit);
174 static int ixgbe_dev_xstats_get_names_by_id(
175         struct rte_eth_dev *dev,
176         struct rte_eth_xstat_name *xstats_names,
177         const uint64_t *ids,
178         unsigned int limit);
179 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
180                                              uint16_t queue_id,
181                                              uint8_t stat_idx,
182                                              uint8_t is_rx);
183 static int ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
184                                  size_t fw_size);
185 static int ixgbe_dev_info_get(struct rte_eth_dev *dev,
186                               struct rte_eth_dev_info *dev_info);
187 static const uint32_t *ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev);
188 static int ixgbevf_dev_info_get(struct rte_eth_dev *dev,
189                                 struct rte_eth_dev_info *dev_info);
190 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
191
192 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
193                 uint16_t vlan_id, int on);
194 static int ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
195                                enum rte_vlan_type vlan_type,
196                                uint16_t tpid_id);
197 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
198                 uint16_t queue, bool on);
199 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
200                 int on);
201 static void ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev,
202                                                   int mask);
203 static int ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask);
204 static int ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
205 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
206 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
207 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
208 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
209
210 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
211 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
212 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
213                                struct rte_eth_fc_conf *fc_conf);
214 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
215                                struct rte_eth_fc_conf *fc_conf);
216 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
217                 struct rte_eth_pfc_conf *pfc_conf);
218 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
219                         struct rte_eth_rss_reta_entry64 *reta_conf,
220                         uint16_t reta_size);
221 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
222                         struct rte_eth_rss_reta_entry64 *reta_conf,
223                         uint16_t reta_size);
224 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
225 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
226 static int ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev);
227 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
228 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
229 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev);
230 static void ixgbe_dev_interrupt_handler(void *param);
231 static void ixgbe_dev_interrupt_delayed_handler(void *param);
232 static void *ixgbe_dev_setup_link_thread_handler(void *param);
233 static int ixgbe_dev_wait_setup_link_complete(struct rte_eth_dev *dev,
234                                               uint32_t timeout_ms);
235
236 static int ixgbe_add_rar(struct rte_eth_dev *dev,
237                         struct rte_ether_addr *mac_addr,
238                         uint32_t index, uint32_t pool);
239 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
240 static int ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
241                                            struct rte_ether_addr *mac_addr);
242 static void ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config);
243 static bool is_device_supported(struct rte_eth_dev *dev,
244                                 struct rte_pci_driver *drv);
245
246 /* For Virtual Function support */
247 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
248 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
249 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
250 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
251 static int ixgbevf_dev_link_update(struct rte_eth_dev *dev,
252                                    int wait_to_complete);
253 static int ixgbevf_dev_stop(struct rte_eth_dev *dev);
254 static int ixgbevf_dev_close(struct rte_eth_dev *dev);
255 static int  ixgbevf_dev_reset(struct rte_eth_dev *dev);
256 static void ixgbevf_intr_disable(struct rte_eth_dev *dev);
257 static void ixgbevf_intr_enable(struct rte_eth_dev *dev);
258 static int ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
259                 struct rte_eth_stats *stats);
260 static int ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
261 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
262                 uint16_t vlan_id, int on);
263 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
264                 uint16_t queue, int on);
265 static int ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask);
266 static int ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
267 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
268 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
269                                             uint16_t queue_id);
270 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
271                                              uint16_t queue_id);
272 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
273                                  uint8_t queue, uint8_t msix_vector);
274 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
275 static int ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev);
276 static int ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev);
277 static int ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev);
278 static int ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev);
279
280 /* For Eth VMDQ APIs support */
281 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
282                 rte_ether_addr * mac_addr, uint8_t on);
283 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on);
284 static int ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
285                 struct rte_eth_mirror_conf *mirror_conf,
286                 uint8_t rule_id, uint8_t on);
287 static int ixgbe_mirror_rule_reset(struct rte_eth_dev *dev,
288                 uint8_t rule_id);
289 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
290                                           uint16_t queue_id);
291 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
292                                            uint16_t queue_id);
293 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
294                                uint8_t queue, uint8_t msix_vector);
295 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
296
297 static int ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
298                                 struct rte_ether_addr *mac_addr,
299                                 uint32_t index, uint32_t pool);
300 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
301 static int ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
302                                              struct rte_ether_addr *mac_addr);
303 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
304                         struct ixgbe_5tuple_filter *filter);
305 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
306                         struct ixgbe_5tuple_filter *filter);
307 static int ixgbe_dev_flow_ops_get(struct rte_eth_dev *dev,
308                                   const struct rte_flow_ops **ops);
309 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
310
311 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
312                                       struct rte_ether_addr *mc_addr_set,
313                                       uint32_t nb_mc_addr);
314 static int ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
315                                    struct rte_eth_dcb_info *dcb_info);
316
317 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
318 static int ixgbe_get_regs(struct rte_eth_dev *dev,
319                             struct rte_dev_reg_info *regs);
320 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
321 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
322                                 struct rte_dev_eeprom_info *eeprom);
323 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
324                                 struct rte_dev_eeprom_info *eeprom);
325
326 static int ixgbe_get_module_info(struct rte_eth_dev *dev,
327                                  struct rte_eth_dev_module_info *modinfo);
328 static int ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
329                                    struct rte_dev_eeprom_info *info);
330
331 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
332 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
333                                 struct rte_dev_reg_info *regs);
334
335 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
336 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
337 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
338                                             struct timespec *timestamp,
339                                             uint32_t flags);
340 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
341                                             struct timespec *timestamp);
342 static int ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
343 static int ixgbe_timesync_read_time(struct rte_eth_dev *dev,
344                                    struct timespec *timestamp);
345 static int ixgbe_timesync_write_time(struct rte_eth_dev *dev,
346                                    const struct timespec *timestamp);
347 static void ixgbevf_dev_interrupt_handler(void *param);
348
349 static int ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
350                                          struct rte_eth_udp_tunnel *udp_tunnel);
351 static int ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
352                                          struct rte_eth_udp_tunnel *udp_tunnel);
353 static int ixgbe_filter_restore(struct rte_eth_dev *dev);
354 static void ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev);
355 static int ixgbe_wait_for_link_up(struct ixgbe_hw *hw);
356
357 /*
358  * Define VF Stats MACRO for Non "cleared on read" register
359  */
360 #define UPDATE_VF_STAT(reg, last, cur)                          \
361 {                                                               \
362         uint32_t latest = IXGBE_READ_REG(hw, reg);              \
363         cur += (latest - last) & UINT_MAX;                      \
364         last = latest;                                          \
365 }
366
367 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
368 {                                                                \
369         u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
370         u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
371         u64 latest = ((new_msb << 32) | new_lsb);                \
372         cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
373         last = latest;                                           \
374 }
375
376 #define IXGBE_SET_HWSTRIP(h, q) do {\
377                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
378                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
379                 (h)->bitmap[idx] |= 1 << bit;\
380         } while (0)
381
382 #define IXGBE_CLEAR_HWSTRIP(h, q) do {\
383                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
384                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
385                 (h)->bitmap[idx] &= ~(1 << bit);\
386         } while (0)
387
388 #define IXGBE_GET_HWSTRIP(h, q, r) do {\
389                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
390                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
391                 (r) = (h)->bitmap[idx] >> bit & 1;\
392         } while (0)
393
394 /*
395  * The set of PCI devices this driver supports
396  */
397 static const struct rte_pci_id pci_id_ixgbe_map[] = {
398         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598) },
399         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_BX) },
400         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT) },
401         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT) },
402         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT) },
403         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2) },
404         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM) },
405         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4) },
406         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT) },
407         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT) },
408         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM) },
409         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR) },
410         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4) },
411         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ) },
412         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KR) },
413         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE) },
414         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4) },
415         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP) },
416         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE) },
417         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE) },
418         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_EM) },
419         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2) },
420         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP) },
421         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP) },
422         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP) },
423         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM) },
424         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM) },
425         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T) },
426         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1) },
427         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP) },
428         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T) },
429         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_1G_T) },
430         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T) },
431         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1) },
432         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR) },
433         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR_L) },
434         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP_N) },
435         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII) },
436         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII_L) },
437         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_10G_T) },
438         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP) },
439         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP_N) },
440         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP) },
441         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T) },
442         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T_L) },
443         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4) },
444         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR) },
445         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_XFI) },
446 #ifdef RTE_LIBRTE_IXGBE_BYPASS
447         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BYPASS) },
448 #endif
449         { .vendor_id = 0, /* sentinel */ },
450 };
451
452 /*
453  * The set of PCI devices this driver supports (for 82599 VF)
454  */
455 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
456         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF) },
457         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF_HV) },
458         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF) },
459         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF_HV) },
460         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF_HV) },
461         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF) },
462         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF) },
463         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF_HV) },
464         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF) },
465         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF_HV) },
466         { .vendor_id = 0, /* sentinel */ },
467 };
468
469 static const struct rte_eth_desc_lim rx_desc_lim = {
470         .nb_max = IXGBE_MAX_RING_DESC,
471         .nb_min = IXGBE_MIN_RING_DESC,
472         .nb_align = IXGBE_RXD_ALIGN,
473 };
474
475 static const struct rte_eth_desc_lim tx_desc_lim = {
476         .nb_max = IXGBE_MAX_RING_DESC,
477         .nb_min = IXGBE_MIN_RING_DESC,
478         .nb_align = IXGBE_TXD_ALIGN,
479         .nb_seg_max = IXGBE_TX_MAX_SEG,
480         .nb_mtu_seg_max = IXGBE_TX_MAX_SEG,
481 };
482
483 static const struct eth_dev_ops ixgbe_eth_dev_ops = {
484         .dev_configure        = ixgbe_dev_configure,
485         .dev_start            = ixgbe_dev_start,
486         .dev_stop             = ixgbe_dev_stop,
487         .dev_set_link_up    = ixgbe_dev_set_link_up,
488         .dev_set_link_down  = ixgbe_dev_set_link_down,
489         .dev_close            = ixgbe_dev_close,
490         .dev_reset            = ixgbe_dev_reset,
491         .promiscuous_enable   = ixgbe_dev_promiscuous_enable,
492         .promiscuous_disable  = ixgbe_dev_promiscuous_disable,
493         .allmulticast_enable  = ixgbe_dev_allmulticast_enable,
494         .allmulticast_disable = ixgbe_dev_allmulticast_disable,
495         .link_update          = ixgbe_dev_link_update,
496         .stats_get            = ixgbe_dev_stats_get,
497         .xstats_get           = ixgbe_dev_xstats_get,
498         .xstats_get_by_id     = ixgbe_dev_xstats_get_by_id,
499         .stats_reset          = ixgbe_dev_stats_reset,
500         .xstats_reset         = ixgbe_dev_xstats_reset,
501         .xstats_get_names     = ixgbe_dev_xstats_get_names,
502         .xstats_get_names_by_id = ixgbe_dev_xstats_get_names_by_id,
503         .queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
504         .fw_version_get       = ixgbe_fw_version_get,
505         .dev_infos_get        = ixgbe_dev_info_get,
506         .dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
507         .mtu_set              = ixgbe_dev_mtu_set,
508         .vlan_filter_set      = ixgbe_vlan_filter_set,
509         .vlan_tpid_set        = ixgbe_vlan_tpid_set,
510         .vlan_offload_set     = ixgbe_vlan_offload_set,
511         .vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
512         .rx_queue_start       = ixgbe_dev_rx_queue_start,
513         .rx_queue_stop        = ixgbe_dev_rx_queue_stop,
514         .tx_queue_start       = ixgbe_dev_tx_queue_start,
515         .tx_queue_stop        = ixgbe_dev_tx_queue_stop,
516         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
517         .rx_queue_intr_enable = ixgbe_dev_rx_queue_intr_enable,
518         .rx_queue_intr_disable = ixgbe_dev_rx_queue_intr_disable,
519         .rx_queue_release     = ixgbe_dev_rx_queue_release,
520         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
521         .tx_queue_release     = ixgbe_dev_tx_queue_release,
522         .dev_led_on           = ixgbe_dev_led_on,
523         .dev_led_off          = ixgbe_dev_led_off,
524         .flow_ctrl_get        = ixgbe_flow_ctrl_get,
525         .flow_ctrl_set        = ixgbe_flow_ctrl_set,
526         .priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
527         .mac_addr_add         = ixgbe_add_rar,
528         .mac_addr_remove      = ixgbe_remove_rar,
529         .mac_addr_set         = ixgbe_set_default_mac_addr,
530         .uc_hash_table_set    = ixgbe_uc_hash_table_set,
531         .uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
532         .mirror_rule_set      = ixgbe_mirror_rule_set,
533         .mirror_rule_reset    = ixgbe_mirror_rule_reset,
534         .set_queue_rate_limit = ixgbe_set_queue_rate_limit,
535         .reta_update          = ixgbe_dev_rss_reta_update,
536         .reta_query           = ixgbe_dev_rss_reta_query,
537         .rss_hash_update      = ixgbe_dev_rss_hash_update,
538         .rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
539         .flow_ops_get         = ixgbe_dev_flow_ops_get,
540         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
541         .rxq_info_get         = ixgbe_rxq_info_get,
542         .txq_info_get         = ixgbe_txq_info_get,
543         .timesync_enable      = ixgbe_timesync_enable,
544         .timesync_disable     = ixgbe_timesync_disable,
545         .timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
546         .timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
547         .get_reg              = ixgbe_get_regs,
548         .get_eeprom_length    = ixgbe_get_eeprom_length,
549         .get_eeprom           = ixgbe_get_eeprom,
550         .set_eeprom           = ixgbe_set_eeprom,
551         .get_module_info      = ixgbe_get_module_info,
552         .get_module_eeprom    = ixgbe_get_module_eeprom,
553         .get_dcb_info         = ixgbe_dev_get_dcb_info,
554         .timesync_adjust_time = ixgbe_timesync_adjust_time,
555         .timesync_read_time   = ixgbe_timesync_read_time,
556         .timesync_write_time  = ixgbe_timesync_write_time,
557         .udp_tunnel_port_add  = ixgbe_dev_udp_tunnel_port_add,
558         .udp_tunnel_port_del  = ixgbe_dev_udp_tunnel_port_del,
559         .tm_ops_get           = ixgbe_tm_ops_get,
560         .tx_done_cleanup      = ixgbe_dev_tx_done_cleanup,
561         .get_monitor_addr     = ixgbe_get_monitor_addr,
562 };
563
564 /*
565  * dev_ops for virtual function, bare necessities for basic vf
566  * operation have been implemented
567  */
568 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
569         .dev_configure        = ixgbevf_dev_configure,
570         .dev_start            = ixgbevf_dev_start,
571         .dev_stop             = ixgbevf_dev_stop,
572         .link_update          = ixgbevf_dev_link_update,
573         .stats_get            = ixgbevf_dev_stats_get,
574         .xstats_get           = ixgbevf_dev_xstats_get,
575         .stats_reset          = ixgbevf_dev_stats_reset,
576         .xstats_reset         = ixgbevf_dev_stats_reset,
577         .xstats_get_names     = ixgbevf_dev_xstats_get_names,
578         .dev_close            = ixgbevf_dev_close,
579         .dev_reset            = ixgbevf_dev_reset,
580         .promiscuous_enable   = ixgbevf_dev_promiscuous_enable,
581         .promiscuous_disable  = ixgbevf_dev_promiscuous_disable,
582         .allmulticast_enable  = ixgbevf_dev_allmulticast_enable,
583         .allmulticast_disable = ixgbevf_dev_allmulticast_disable,
584         .dev_infos_get        = ixgbevf_dev_info_get,
585         .dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
586         .mtu_set              = ixgbevf_dev_set_mtu,
587         .vlan_filter_set      = ixgbevf_vlan_filter_set,
588         .vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
589         .vlan_offload_set     = ixgbevf_vlan_offload_set,
590         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
591         .rx_queue_release     = ixgbe_dev_rx_queue_release,
592         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
593         .tx_queue_release     = ixgbe_dev_tx_queue_release,
594         .rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
595         .rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
596         .mac_addr_add         = ixgbevf_add_mac_addr,
597         .mac_addr_remove      = ixgbevf_remove_mac_addr,
598         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
599         .rxq_info_get         = ixgbe_rxq_info_get,
600         .txq_info_get         = ixgbe_txq_info_get,
601         .mac_addr_set         = ixgbevf_set_default_mac_addr,
602         .get_reg              = ixgbevf_get_regs,
603         .reta_update          = ixgbe_dev_rss_reta_update,
604         .reta_query           = ixgbe_dev_rss_reta_query,
605         .rss_hash_update      = ixgbe_dev_rss_hash_update,
606         .rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
607         .tx_done_cleanup      = ixgbe_dev_tx_done_cleanup,
608 };
609
610 /* store statistics names and its offset in stats structure */
611 struct rte_ixgbe_xstats_name_off {
612         char name[RTE_ETH_XSTATS_NAME_SIZE];
613         unsigned offset;
614 };
615
616 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
617         {"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
618         {"rx_illegal_byte_errors", offsetof(struct ixgbe_hw_stats, illerrc)},
619         {"rx_error_bytes", offsetof(struct ixgbe_hw_stats, errbc)},
620         {"mac_local_errors", offsetof(struct ixgbe_hw_stats, mlfc)},
621         {"mac_remote_errors", offsetof(struct ixgbe_hw_stats, mrfc)},
622         {"rx_length_errors", offsetof(struct ixgbe_hw_stats, rlec)},
623         {"tx_xon_packets", offsetof(struct ixgbe_hw_stats, lxontxc)},
624         {"rx_xon_packets", offsetof(struct ixgbe_hw_stats, lxonrxc)},
625         {"tx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxofftxc)},
626         {"rx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
627         {"rx_size_64_packets", offsetof(struct ixgbe_hw_stats, prc64)},
628         {"rx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, prc127)},
629         {"rx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, prc255)},
630         {"rx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, prc511)},
631         {"rx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
632                 prc1023)},
633         {"rx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
634                 prc1522)},
635         {"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
636         {"rx_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
637         {"rx_fragment_errors", offsetof(struct ixgbe_hw_stats, rfc)},
638         {"rx_undersize_errors", offsetof(struct ixgbe_hw_stats, ruc)},
639         {"rx_oversize_errors", offsetof(struct ixgbe_hw_stats, roc)},
640         {"rx_jabber_errors", offsetof(struct ixgbe_hw_stats, rjc)},
641         {"rx_management_packets", offsetof(struct ixgbe_hw_stats, mngprc)},
642         {"rx_management_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
643         {"tx_management_packets", offsetof(struct ixgbe_hw_stats, mngptc)},
644         {"rx_total_packets", offsetof(struct ixgbe_hw_stats, tpr)},
645         {"rx_total_bytes", offsetof(struct ixgbe_hw_stats, tor)},
646         {"tx_total_packets", offsetof(struct ixgbe_hw_stats, tpt)},
647         {"tx_size_64_packets", offsetof(struct ixgbe_hw_stats, ptc64)},
648         {"tx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, ptc127)},
649         {"tx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, ptc255)},
650         {"tx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, ptc511)},
651         {"tx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
652                 ptc1023)},
653         {"tx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
654                 ptc1522)},
655         {"tx_multicast_packets", offsetof(struct ixgbe_hw_stats, mptc)},
656         {"tx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bptc)},
657         {"rx_mac_short_packet_dropped", offsetof(struct ixgbe_hw_stats, mspdc)},
658         {"rx_l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
659
660         {"flow_director_added_filters", offsetof(struct ixgbe_hw_stats,
661                 fdirustat_add)},
662         {"flow_director_removed_filters", offsetof(struct ixgbe_hw_stats,
663                 fdirustat_remove)},
664         {"flow_director_filter_add_errors", offsetof(struct ixgbe_hw_stats,
665                 fdirfstat_fadd)},
666         {"flow_director_filter_remove_errors", offsetof(struct ixgbe_hw_stats,
667                 fdirfstat_fremove)},
668         {"flow_director_matched_filters", offsetof(struct ixgbe_hw_stats,
669                 fdirmatch)},
670         {"flow_director_missed_filters", offsetof(struct ixgbe_hw_stats,
671                 fdirmiss)},
672
673         {"rx_fcoe_crc_errors", offsetof(struct ixgbe_hw_stats, fccrc)},
674         {"rx_fcoe_dropped", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
675         {"rx_fcoe_mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats,
676                 fclast)},
677         {"rx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeprc)},
678         {"tx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeptc)},
679         {"rx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwrc)},
680         {"tx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwtc)},
681         {"rx_fcoe_no_direct_data_placement", offsetof(struct ixgbe_hw_stats,
682                 fcoe_noddp)},
683         {"rx_fcoe_no_direct_data_placement_ext_buff",
684                 offsetof(struct ixgbe_hw_stats, fcoe_noddp_ext_buff)},
685
686         {"tx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
687                 lxontxc)},
688         {"rx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
689                 lxonrxc)},
690         {"tx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
691                 lxofftxc)},
692         {"rx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
693                 lxoffrxc)},
694         {"rx_total_missed_packets", offsetof(struct ixgbe_hw_stats, mpctotal)},
695 };
696
697 #define IXGBE_NB_HW_STATS (sizeof(rte_ixgbe_stats_strings) / \
698                            sizeof(rte_ixgbe_stats_strings[0]))
699
700 /* MACsec statistics */
701 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_macsec_strings[] = {
702         {"out_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
703                 out_pkts_untagged)},
704         {"out_pkts_encrypted", offsetof(struct ixgbe_macsec_stats,
705                 out_pkts_encrypted)},
706         {"out_pkts_protected", offsetof(struct ixgbe_macsec_stats,
707                 out_pkts_protected)},
708         {"out_octets_encrypted", offsetof(struct ixgbe_macsec_stats,
709                 out_octets_encrypted)},
710         {"out_octets_protected", offsetof(struct ixgbe_macsec_stats,
711                 out_octets_protected)},
712         {"in_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
713                 in_pkts_untagged)},
714         {"in_pkts_badtag", offsetof(struct ixgbe_macsec_stats,
715                 in_pkts_badtag)},
716         {"in_pkts_nosci", offsetof(struct ixgbe_macsec_stats,
717                 in_pkts_nosci)},
718         {"in_pkts_unknownsci", offsetof(struct ixgbe_macsec_stats,
719                 in_pkts_unknownsci)},
720         {"in_octets_decrypted", offsetof(struct ixgbe_macsec_stats,
721                 in_octets_decrypted)},
722         {"in_octets_validated", offsetof(struct ixgbe_macsec_stats,
723                 in_octets_validated)},
724         {"in_pkts_unchecked", offsetof(struct ixgbe_macsec_stats,
725                 in_pkts_unchecked)},
726         {"in_pkts_delayed", offsetof(struct ixgbe_macsec_stats,
727                 in_pkts_delayed)},
728         {"in_pkts_late", offsetof(struct ixgbe_macsec_stats,
729                 in_pkts_late)},
730         {"in_pkts_ok", offsetof(struct ixgbe_macsec_stats,
731                 in_pkts_ok)},
732         {"in_pkts_invalid", offsetof(struct ixgbe_macsec_stats,
733                 in_pkts_invalid)},
734         {"in_pkts_notvalid", offsetof(struct ixgbe_macsec_stats,
735                 in_pkts_notvalid)},
736         {"in_pkts_unusedsa", offsetof(struct ixgbe_macsec_stats,
737                 in_pkts_unusedsa)},
738         {"in_pkts_notusingsa", offsetof(struct ixgbe_macsec_stats,
739                 in_pkts_notusingsa)},
740 };
741
742 #define IXGBE_NB_MACSEC_STATS (sizeof(rte_ixgbe_macsec_strings) / \
743                            sizeof(rte_ixgbe_macsec_strings[0]))
744
745 /* Per-queue statistics */
746 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_rxq_strings[] = {
747         {"mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats, rnbc)},
748         {"dropped", offsetof(struct ixgbe_hw_stats, mpc)},
749         {"xon_packets", offsetof(struct ixgbe_hw_stats, pxonrxc)},
750         {"xoff_packets", offsetof(struct ixgbe_hw_stats, pxoffrxc)},
751 };
752
753 #define IXGBE_NB_RXQ_PRIO_STATS (sizeof(rte_ixgbe_rxq_strings) / \
754                            sizeof(rte_ixgbe_rxq_strings[0]))
755 #define IXGBE_NB_RXQ_PRIO_VALUES 8
756
757 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_txq_strings[] = {
758         {"xon_packets", offsetof(struct ixgbe_hw_stats, pxontxc)},
759         {"xoff_packets", offsetof(struct ixgbe_hw_stats, pxofftxc)},
760         {"xon_to_xoff_packets", offsetof(struct ixgbe_hw_stats,
761                 pxon2offc)},
762 };
763
764 #define IXGBE_NB_TXQ_PRIO_STATS (sizeof(rte_ixgbe_txq_strings) / \
765                            sizeof(rte_ixgbe_txq_strings[0]))
766 #define IXGBE_NB_TXQ_PRIO_VALUES 8
767
768 static const struct rte_ixgbe_xstats_name_off rte_ixgbevf_stats_strings[] = {
769         {"rx_multicast_packets", offsetof(struct ixgbevf_hw_stats, vfmprc)},
770 };
771
772 #define IXGBEVF_NB_XSTATS (sizeof(rte_ixgbevf_stats_strings) /  \
773                 sizeof(rte_ixgbevf_stats_strings[0]))
774
775 /*
776  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
777  */
778 static inline int
779 ixgbe_is_sfp(struct ixgbe_hw *hw)
780 {
781         switch (hw->phy.type) {
782         case ixgbe_phy_sfp_avago:
783         case ixgbe_phy_sfp_ftl:
784         case ixgbe_phy_sfp_intel:
785         case ixgbe_phy_sfp_unknown:
786         case ixgbe_phy_sfp_passive_tyco:
787         case ixgbe_phy_sfp_passive_unknown:
788                 return 1;
789         default:
790                 return 0;
791         }
792 }
793
794 static inline int32_t
795 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
796 {
797         uint32_t ctrl_ext;
798         int32_t status;
799
800         status = ixgbe_reset_hw(hw);
801
802         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
803         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
804         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
805         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
806         IXGBE_WRITE_FLUSH(hw);
807
808         if (status == IXGBE_ERR_SFP_NOT_PRESENT)
809                 status = IXGBE_SUCCESS;
810         return status;
811 }
812
813 static inline void
814 ixgbe_enable_intr(struct rte_eth_dev *dev)
815 {
816         struct ixgbe_interrupt *intr =
817                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
818         struct ixgbe_hw *hw =
819                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
820
821         IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
822         IXGBE_WRITE_FLUSH(hw);
823 }
824
825 /*
826  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
827  */
828 static void
829 ixgbe_disable_intr(struct ixgbe_hw *hw)
830 {
831         PMD_INIT_FUNC_TRACE();
832
833         if (hw->mac.type == ixgbe_mac_82598EB) {
834                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
835         } else {
836                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
837                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
838                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
839         }
840         IXGBE_WRITE_FLUSH(hw);
841 }
842
843 /*
844  * This function resets queue statistics mapping registers.
845  * From Niantic datasheet, Initialization of Statistics section:
846  * "...if software requires the queue counters, the RQSMR and TQSM registers
847  * must be re-programmed following a device reset.
848  */
849 static void
850 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
851 {
852         uint32_t i;
853
854         for (i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
855                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
856                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
857         }
858 }
859
860
861 static int
862 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
863                                   uint16_t queue_id,
864                                   uint8_t stat_idx,
865                                   uint8_t is_rx)
866 {
867 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
868 #define NB_QMAP_FIELDS_PER_QSM_REG 4
869 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
870
871         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
872         struct ixgbe_stat_mapping_registers *stat_mappings =
873                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
874         uint32_t qsmr_mask = 0;
875         uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
876         uint32_t q_map;
877         uint8_t n, offset;
878
879         if ((hw->mac.type != ixgbe_mac_82599EB) &&
880                 (hw->mac.type != ixgbe_mac_X540) &&
881                 (hw->mac.type != ixgbe_mac_X550) &&
882                 (hw->mac.type != ixgbe_mac_X550EM_x) &&
883                 (hw->mac.type != ixgbe_mac_X550EM_a))
884                 return -ENOSYS;
885
886         PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
887                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
888                      queue_id, stat_idx);
889
890         n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
891         if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
892                 PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
893                 return -EIO;
894         }
895         offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
896
897         /* Now clear any previous stat_idx set */
898         clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
899         if (!is_rx)
900                 stat_mappings->tqsm[n] &= ~clearing_mask;
901         else
902                 stat_mappings->rqsmr[n] &= ~clearing_mask;
903
904         q_map = (uint32_t)stat_idx;
905         q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
906         qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
907         if (!is_rx)
908                 stat_mappings->tqsm[n] |= qsmr_mask;
909         else
910                 stat_mappings->rqsmr[n] |= qsmr_mask;
911
912         PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
913                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
914                      queue_id, stat_idx);
915         PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
916                      is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
917
918         /* Now write the mapping in the appropriate register */
919         if (is_rx) {
920                 PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
921                              stat_mappings->rqsmr[n], n);
922                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
923         } else {
924                 PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
925                              stat_mappings->tqsm[n], n);
926                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
927         }
928         return 0;
929 }
930
931 static void
932 ixgbe_restore_statistics_mapping(struct rte_eth_dev *dev)
933 {
934         struct ixgbe_stat_mapping_registers *stat_mappings =
935                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
936         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
937         int i;
938
939         /* write whatever was in stat mapping table to the NIC */
940         for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
941                 /* rx */
942                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
943
944                 /* tx */
945                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
946         }
947 }
948
949 static void
950 ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config)
951 {
952         uint8_t i;
953         struct ixgbe_dcb_tc_config *tc;
954         uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
955
956         dcb_config->num_tcs.pg_tcs = dcb_max_tc;
957         dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
958         for (i = 0; i < dcb_max_tc; i++) {
959                 tc = &dcb_config->tc_config[i];
960                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
961                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
962                                  (uint8_t)(100/dcb_max_tc + (i & 1));
963                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
964                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
965                                  (uint8_t)(100/dcb_max_tc + (i & 1));
966                 tc->pfc = ixgbe_dcb_pfc_disabled;
967         }
968
969         /* Initialize default user to priority mapping, UPx->TC0 */
970         tc = &dcb_config->tc_config[0];
971         tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
972         tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
973         for (i = 0; i < IXGBE_DCB_MAX_BW_GROUP; i++) {
974                 dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
975                 dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
976         }
977         dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
978         dcb_config->pfc_mode_enable = false;
979         dcb_config->vt_mode = true;
980         dcb_config->round_robin_enable = false;
981         /* support all DCB capabilities in 82599 */
982         dcb_config->support.capabilities = 0xFF;
983
984         /*we only support 4 Tcs for X540, X550 */
985         if (hw->mac.type == ixgbe_mac_X540 ||
986                 hw->mac.type == ixgbe_mac_X550 ||
987                 hw->mac.type == ixgbe_mac_X550EM_x ||
988                 hw->mac.type == ixgbe_mac_X550EM_a) {
989                 dcb_config->num_tcs.pg_tcs = 4;
990                 dcb_config->num_tcs.pfc_tcs = 4;
991         }
992 }
993
994 /*
995  * Ensure that all locks are released before first NVM or PHY access
996  */
997 static void
998 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
999 {
1000         uint16_t mask;
1001
1002         /*
1003          * Phy lock should not fail in this early stage. If this is the case,
1004          * it is due to an improper exit of the application.
1005          * So force the release of the faulty lock. Release of common lock
1006          * is done automatically by swfw_sync function.
1007          */
1008         mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
1009         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1010                 PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
1011         }
1012         ixgbe_release_swfw_semaphore(hw, mask);
1013
1014         /*
1015          * These ones are more tricky since they are common to all ports; but
1016          * swfw_sync retries last long enough (1s) to be almost sure that if
1017          * lock can not be taken it is due to an improper lock of the
1018          * semaphore.
1019          */
1020         mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
1021         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1022                 PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1023         }
1024         ixgbe_release_swfw_semaphore(hw, mask);
1025 }
1026
1027 /*
1028  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
1029  * It returns 0 on success.
1030  */
1031 static int
1032 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params __rte_unused)
1033 {
1034         struct ixgbe_adapter *ad = eth_dev->data->dev_private;
1035         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1036         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1037         struct ixgbe_hw *hw =
1038                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1039         struct ixgbe_vfta *shadow_vfta =
1040                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1041         struct ixgbe_hwstrip *hwstrip =
1042                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1043         struct ixgbe_dcb_config *dcb_config =
1044                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
1045         struct ixgbe_filter_info *filter_info =
1046                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1047         struct ixgbe_bw_conf *bw_conf =
1048                 IXGBE_DEV_PRIVATE_TO_BW_CONF(eth_dev->data->dev_private);
1049         uint32_t ctrl_ext;
1050         uint16_t csum;
1051         int diag, i, ret;
1052
1053         PMD_INIT_FUNC_TRACE();
1054
1055         ixgbe_dev_macsec_setting_reset(eth_dev);
1056
1057         eth_dev->dev_ops = &ixgbe_eth_dev_ops;
1058         eth_dev->rx_queue_count       = ixgbe_dev_rx_queue_count;
1059         eth_dev->rx_descriptor_done   = ixgbe_dev_rx_descriptor_done;
1060         eth_dev->rx_descriptor_status = ixgbe_dev_rx_descriptor_status;
1061         eth_dev->tx_descriptor_status = ixgbe_dev_tx_descriptor_status;
1062         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1063         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1064         eth_dev->tx_pkt_prepare = &ixgbe_prep_pkts;
1065
1066         /*
1067          * For secondary processes, we don't initialise any further as primary
1068          * has already done this work. Only check we don't need a different
1069          * RX and TX function.
1070          */
1071         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1072                 struct ixgbe_tx_queue *txq;
1073                 /* TX queue function in primary, set by last queue initialized
1074                  * Tx queue may not initialized by primary process
1075                  */
1076                 if (eth_dev->data->tx_queues) {
1077                         txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
1078                         ixgbe_set_tx_function(eth_dev, txq);
1079                 } else {
1080                         /* Use default TX function if we get here */
1081                         PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
1082                                      "Using default TX function.");
1083                 }
1084
1085                 ixgbe_set_rx_function(eth_dev);
1086
1087                 return 0;
1088         }
1089
1090         rte_atomic32_clear(&ad->link_thread_running);
1091         rte_eth_copy_pci_info(eth_dev, pci_dev);
1092         eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1093
1094         /* Vendor and Device ID need to be set before init of shared code */
1095         hw->device_id = pci_dev->id.device_id;
1096         hw->vendor_id = pci_dev->id.vendor_id;
1097         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1098         hw->allow_unsupported_sfp = 1;
1099
1100         /* Initialize the shared code (base driver) */
1101 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1102         diag = ixgbe_bypass_init_shared_code(hw);
1103 #else
1104         diag = ixgbe_init_shared_code(hw);
1105 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1106
1107         if (diag != IXGBE_SUCCESS) {
1108                 PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
1109                 return -EIO;
1110         }
1111
1112         if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
1113                 PMD_INIT_LOG(ERR, "\nERROR: "
1114                         "Firmware recovery mode detected. Limiting functionality.\n"
1115                         "Refer to the Intel(R) Ethernet Adapters and Devices "
1116                         "User Guide for details on firmware recovery mode.");
1117                 return -EIO;
1118         }
1119
1120         /* pick up the PCI bus settings for reporting later */
1121         ixgbe_get_bus_info(hw);
1122
1123         /* Unlock any pending hardware semaphore */
1124         ixgbe_swfw_lock_reset(hw);
1125
1126 #ifdef RTE_LIB_SECURITY
1127         /* Initialize security_ctx only for primary process*/
1128         if (ixgbe_ipsec_ctx_create(eth_dev))
1129                 return -ENOMEM;
1130 #endif
1131
1132         /* Initialize DCB configuration*/
1133         memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
1134         ixgbe_dcb_init(hw, dcb_config);
1135         /* Get Hardware Flow Control setting */
1136         hw->fc.requested_mode = ixgbe_fc_none;
1137         hw->fc.current_mode = ixgbe_fc_none;
1138         hw->fc.pause_time = IXGBE_FC_PAUSE;
1139         for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
1140                 hw->fc.low_water[i] = IXGBE_FC_LO;
1141                 hw->fc.high_water[i] = IXGBE_FC_HI;
1142         }
1143         hw->fc.send_xon = 1;
1144
1145         /* Make sure we have a good EEPROM before we read from it */
1146         diag = ixgbe_validate_eeprom_checksum(hw, &csum);
1147         if (diag != IXGBE_SUCCESS) {
1148                 PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
1149                 return -EIO;
1150         }
1151
1152 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1153         diag = ixgbe_bypass_init_hw(hw);
1154 #else
1155         diag = ixgbe_init_hw(hw);
1156 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1157
1158         /*
1159          * Devices with copper phys will fail to initialise if ixgbe_init_hw()
1160          * is called too soon after the kernel driver unbinding/binding occurs.
1161          * The failure occurs in ixgbe_identify_phy_generic() for all devices,
1162          * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
1163          * also called. See ixgbe_identify_phy_82599(). The reason for the
1164          * failure is not known, and only occuts when virtualisation features
1165          * are disabled in the bios. A delay of 100ms  was found to be enough by
1166          * trial-and-error, and is doubled to be safe.
1167          */
1168         if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
1169                 rte_delay_ms(200);
1170                 diag = ixgbe_init_hw(hw);
1171         }
1172
1173         if (diag == IXGBE_ERR_SFP_NOT_PRESENT)
1174                 diag = IXGBE_SUCCESS;
1175
1176         if (diag == IXGBE_ERR_EEPROM_VERSION) {
1177                 PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
1178                              "LOM.  Please be aware there may be issues associated "
1179                              "with your hardware.");
1180                 PMD_INIT_LOG(ERR, "If you are experiencing problems "
1181                              "please contact your Intel or hardware representative "
1182                              "who provided you with this hardware.");
1183         } else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
1184                 PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
1185         if (diag) {
1186                 PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
1187                 return -EIO;
1188         }
1189
1190         /* Reset the hw statistics */
1191         ixgbe_dev_stats_reset(eth_dev);
1192
1193         /* disable interrupt */
1194         ixgbe_disable_intr(hw);
1195
1196         /* reset mappings for queue statistics hw counters*/
1197         ixgbe_reset_qstat_mappings(hw);
1198
1199         /* Allocate memory for storing MAC addresses */
1200         eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", RTE_ETHER_ADDR_LEN *
1201                                                hw->mac.num_rar_entries, 0);
1202         if (eth_dev->data->mac_addrs == NULL) {
1203                 PMD_INIT_LOG(ERR,
1204                              "Failed to allocate %u bytes needed to store "
1205                              "MAC addresses",
1206                              RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1207                 return -ENOMEM;
1208         }
1209         /* Copy the permanent MAC address */
1210         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.perm_addr,
1211                         &eth_dev->data->mac_addrs[0]);
1212
1213         /* Allocate memory for storing hash filter MAC addresses */
1214         eth_dev->data->hash_mac_addrs = rte_zmalloc(
1215                 "ixgbe", RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC, 0);
1216         if (eth_dev->data->hash_mac_addrs == NULL) {
1217                 PMD_INIT_LOG(ERR,
1218                              "Failed to allocate %d bytes needed to store MAC addresses",
1219                              RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
1220                 return -ENOMEM;
1221         }
1222
1223         /* initialize the vfta */
1224         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1225
1226         /* initialize the hw strip bitmap*/
1227         memset(hwstrip, 0, sizeof(*hwstrip));
1228
1229         /* initialize PF if max_vfs not zero */
1230         ret = ixgbe_pf_host_init(eth_dev);
1231         if (ret) {
1232                 rte_free(eth_dev->data->mac_addrs);
1233                 eth_dev->data->mac_addrs = NULL;
1234                 rte_free(eth_dev->data->hash_mac_addrs);
1235                 eth_dev->data->hash_mac_addrs = NULL;
1236                 return ret;
1237         }
1238
1239         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1240         /* let hardware know driver is loaded */
1241         ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1242         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
1243         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1244         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1245         IXGBE_WRITE_FLUSH(hw);
1246
1247         if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1248                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1249                              (int) hw->mac.type, (int) hw->phy.type,
1250                              (int) hw->phy.sfp_type);
1251         else
1252                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1253                              (int) hw->mac.type, (int) hw->phy.type);
1254
1255         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1256                      eth_dev->data->port_id, pci_dev->id.vendor_id,
1257                      pci_dev->id.device_id);
1258
1259         rte_intr_callback_register(intr_handle,
1260                                    ixgbe_dev_interrupt_handler, eth_dev);
1261
1262         /* enable uio/vfio intr/eventfd mapping */
1263         rte_intr_enable(intr_handle);
1264
1265         /* enable support intr */
1266         ixgbe_enable_intr(eth_dev);
1267
1268         /* initialize filter info */
1269         memset(filter_info, 0,
1270                sizeof(struct ixgbe_filter_info));
1271
1272         /* initialize 5tuple filter list */
1273         TAILQ_INIT(&filter_info->fivetuple_list);
1274
1275         /* initialize flow director filter list & hash */
1276         ixgbe_fdir_filter_init(eth_dev);
1277
1278         /* initialize l2 tunnel filter list & hash */
1279         ixgbe_l2_tn_filter_init(eth_dev);
1280
1281         /* initialize flow filter lists */
1282         ixgbe_filterlist_init();
1283
1284         /* initialize bandwidth configuration info */
1285         memset(bw_conf, 0, sizeof(struct ixgbe_bw_conf));
1286
1287         /* initialize Traffic Manager configuration */
1288         ixgbe_tm_conf_init(eth_dev);
1289
1290         return 0;
1291 }
1292
1293 static int
1294 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1295 {
1296         PMD_INIT_FUNC_TRACE();
1297
1298         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1299                 return 0;
1300
1301         ixgbe_dev_close(eth_dev);
1302
1303         return 0;
1304 }
1305
1306 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
1307 {
1308         struct ixgbe_filter_info *filter_info =
1309                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1310         struct ixgbe_5tuple_filter *p_5tuple;
1311
1312         while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
1313                 TAILQ_REMOVE(&filter_info->fivetuple_list,
1314                              p_5tuple,
1315                              entries);
1316                 rte_free(p_5tuple);
1317         }
1318         memset(filter_info->fivetuple_mask, 0,
1319                sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1320
1321         return 0;
1322 }
1323
1324 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev)
1325 {
1326         struct ixgbe_hw_fdir_info *fdir_info =
1327                 IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1328         struct ixgbe_fdir_filter *fdir_filter;
1329
1330                 if (fdir_info->hash_map)
1331                 rte_free(fdir_info->hash_map);
1332         if (fdir_info->hash_handle)
1333                 rte_hash_free(fdir_info->hash_handle);
1334
1335         while ((fdir_filter = TAILQ_FIRST(&fdir_info->fdir_list))) {
1336                 TAILQ_REMOVE(&fdir_info->fdir_list,
1337                              fdir_filter,
1338                              entries);
1339                 rte_free(fdir_filter);
1340         }
1341
1342         return 0;
1343 }
1344
1345 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev)
1346 {
1347         struct ixgbe_l2_tn_info *l2_tn_info =
1348                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1349         struct ixgbe_l2_tn_filter *l2_tn_filter;
1350
1351         if (l2_tn_info->hash_map)
1352                 rte_free(l2_tn_info->hash_map);
1353         if (l2_tn_info->hash_handle)
1354                 rte_hash_free(l2_tn_info->hash_handle);
1355
1356         while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
1357                 TAILQ_REMOVE(&l2_tn_info->l2_tn_list,
1358                              l2_tn_filter,
1359                              entries);
1360                 rte_free(l2_tn_filter);
1361         }
1362
1363         return 0;
1364 }
1365
1366 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev)
1367 {
1368         struct ixgbe_hw_fdir_info *fdir_info =
1369                 IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1370         char fdir_hash_name[RTE_HASH_NAMESIZE];
1371         struct rte_hash_parameters fdir_hash_params = {
1372                 .name = fdir_hash_name,
1373                 .entries = IXGBE_MAX_FDIR_FILTER_NUM,
1374                 .key_len = sizeof(union ixgbe_atr_input),
1375                 .hash_func = rte_hash_crc,
1376                 .hash_func_init_val = 0,
1377                 .socket_id = rte_socket_id(),
1378         };
1379
1380         TAILQ_INIT(&fdir_info->fdir_list);
1381         snprintf(fdir_hash_name, RTE_HASH_NAMESIZE,
1382                  "fdir_%s", eth_dev->device->name);
1383         fdir_info->hash_handle = rte_hash_create(&fdir_hash_params);
1384         if (!fdir_info->hash_handle) {
1385                 PMD_INIT_LOG(ERR, "Failed to create fdir hash table!");
1386                 return -EINVAL;
1387         }
1388         fdir_info->hash_map = rte_zmalloc("ixgbe",
1389                                           sizeof(struct ixgbe_fdir_filter *) *
1390                                           IXGBE_MAX_FDIR_FILTER_NUM,
1391                                           0);
1392         if (!fdir_info->hash_map) {
1393                 PMD_INIT_LOG(ERR,
1394                              "Failed to allocate memory for fdir hash map!");
1395                 return -ENOMEM;
1396         }
1397         fdir_info->mask_added = FALSE;
1398
1399         return 0;
1400 }
1401
1402 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev)
1403 {
1404         struct ixgbe_l2_tn_info *l2_tn_info =
1405                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1406         char l2_tn_hash_name[RTE_HASH_NAMESIZE];
1407         struct rte_hash_parameters l2_tn_hash_params = {
1408                 .name = l2_tn_hash_name,
1409                 .entries = IXGBE_MAX_L2_TN_FILTER_NUM,
1410                 .key_len = sizeof(struct ixgbe_l2_tn_key),
1411                 .hash_func = rte_hash_crc,
1412                 .hash_func_init_val = 0,
1413                 .socket_id = rte_socket_id(),
1414         };
1415
1416         TAILQ_INIT(&l2_tn_info->l2_tn_list);
1417         snprintf(l2_tn_hash_name, RTE_HASH_NAMESIZE,
1418                  "l2_tn_%s", eth_dev->device->name);
1419         l2_tn_info->hash_handle = rte_hash_create(&l2_tn_hash_params);
1420         if (!l2_tn_info->hash_handle) {
1421                 PMD_INIT_LOG(ERR, "Failed to create L2 TN hash table!");
1422                 return -EINVAL;
1423         }
1424         l2_tn_info->hash_map = rte_zmalloc("ixgbe",
1425                                    sizeof(struct ixgbe_l2_tn_filter *) *
1426                                    IXGBE_MAX_L2_TN_FILTER_NUM,
1427                                    0);
1428         if (!l2_tn_info->hash_map) {
1429                 PMD_INIT_LOG(ERR,
1430                         "Failed to allocate memory for L2 TN hash map!");
1431                 return -ENOMEM;
1432         }
1433         l2_tn_info->e_tag_en = FALSE;
1434         l2_tn_info->e_tag_fwd_en = FALSE;
1435         l2_tn_info->e_tag_ether_type = RTE_ETHER_TYPE_ETAG;
1436
1437         return 0;
1438 }
1439 /*
1440  * Negotiate mailbox API version with the PF.
1441  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1442  * Then we try to negotiate starting with the most recent one.
1443  * If all negotiation attempts fail, then we will proceed with
1444  * the default one (ixgbe_mbox_api_10).
1445  */
1446 static void
1447 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1448 {
1449         int32_t i;
1450
1451         /* start with highest supported, proceed down */
1452         static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1453                 ixgbe_mbox_api_13,
1454                 ixgbe_mbox_api_12,
1455                 ixgbe_mbox_api_11,
1456                 ixgbe_mbox_api_10,
1457         };
1458
1459         for (i = 0;
1460                         i != RTE_DIM(sup_ver) &&
1461                         ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1462                         i++)
1463                 ;
1464 }
1465
1466 static void
1467 generate_random_mac_addr(struct rte_ether_addr *mac_addr)
1468 {
1469         uint64_t random;
1470
1471         /* Set Organizationally Unique Identifier (OUI) prefix. */
1472         mac_addr->addr_bytes[0] = 0x00;
1473         mac_addr->addr_bytes[1] = 0x09;
1474         mac_addr->addr_bytes[2] = 0xC0;
1475         /* Force indication of locally assigned MAC address. */
1476         mac_addr->addr_bytes[0] |= RTE_ETHER_LOCAL_ADMIN_ADDR;
1477         /* Generate the last 3 bytes of the MAC address with a random number. */
1478         random = rte_rand();
1479         memcpy(&mac_addr->addr_bytes[3], &random, 3);
1480 }
1481
1482 static int
1483 devarg_handle_int(__rte_unused const char *key, const char *value,
1484                   void *extra_args)
1485 {
1486         uint16_t *n = extra_args;
1487
1488         if (value == NULL || extra_args == NULL)
1489                 return -EINVAL;
1490
1491         *n = (uint16_t)strtoul(value, NULL, 0);
1492         if (*n == USHRT_MAX && errno == ERANGE)
1493                 return -1;
1494
1495         return 0;
1496 }
1497
1498 static void
1499 ixgbevf_parse_devargs(struct ixgbe_adapter *adapter,
1500                       struct rte_devargs *devargs)
1501 {
1502         struct rte_kvargs *kvlist;
1503         uint16_t pflink_fullchk;
1504
1505         if (devargs == NULL)
1506                 return;
1507
1508         kvlist = rte_kvargs_parse(devargs->args, ixgbevf_valid_arguments);
1509         if (kvlist == NULL)
1510                 return;
1511
1512         if (rte_kvargs_count(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK) == 1 &&
1513             rte_kvargs_process(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK,
1514                                devarg_handle_int, &pflink_fullchk) == 0 &&
1515             pflink_fullchk == 1)
1516                 adapter->pflink_fullchk = 1;
1517
1518         rte_kvargs_free(kvlist);
1519 }
1520
1521 /*
1522  * Virtual Function device init
1523  */
1524 static int
1525 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1526 {
1527         int diag;
1528         uint32_t tc, tcs;
1529         struct ixgbe_adapter *ad = eth_dev->data->dev_private;
1530         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1531         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1532         struct ixgbe_hw *hw =
1533                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1534         struct ixgbe_vfta *shadow_vfta =
1535                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1536         struct ixgbe_hwstrip *hwstrip =
1537                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1538         struct rte_ether_addr *perm_addr =
1539                 (struct rte_ether_addr *)hw->mac.perm_addr;
1540
1541         PMD_INIT_FUNC_TRACE();
1542
1543         eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1544         eth_dev->rx_descriptor_done   = ixgbe_dev_rx_descriptor_done;
1545         eth_dev->rx_descriptor_status = ixgbe_dev_rx_descriptor_status;
1546         eth_dev->tx_descriptor_status = ixgbe_dev_tx_descriptor_status;
1547         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1548         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1549
1550         /* for secondary processes, we don't initialise any further as primary
1551          * has already done this work. Only check we don't need a different
1552          * RX function
1553          */
1554         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1555                 struct ixgbe_tx_queue *txq;
1556                 /* TX queue function in primary, set by last queue initialized
1557                  * Tx queue may not initialized by primary process
1558                  */
1559                 if (eth_dev->data->tx_queues) {
1560                         txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1561                         ixgbe_set_tx_function(eth_dev, txq);
1562                 } else {
1563                         /* Use default TX function if we get here */
1564                         PMD_INIT_LOG(NOTICE,
1565                                      "No TX queues configured yet. Using default TX function.");
1566                 }
1567
1568                 ixgbe_set_rx_function(eth_dev);
1569
1570                 return 0;
1571         }
1572
1573         rte_atomic32_clear(&ad->link_thread_running);
1574         ixgbevf_parse_devargs(eth_dev->data->dev_private,
1575                               pci_dev->device.devargs);
1576
1577         rte_eth_copy_pci_info(eth_dev, pci_dev);
1578         eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1579
1580         hw->device_id = pci_dev->id.device_id;
1581         hw->vendor_id = pci_dev->id.vendor_id;
1582         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1583
1584         /* initialize the vfta */
1585         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1586
1587         /* initialize the hw strip bitmap*/
1588         memset(hwstrip, 0, sizeof(*hwstrip));
1589
1590         /* Initialize the shared code (base driver) */
1591         diag = ixgbe_init_shared_code(hw);
1592         if (diag != IXGBE_SUCCESS) {
1593                 PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1594                 return -EIO;
1595         }
1596
1597         /* init_mailbox_params */
1598         hw->mbx.ops.init_params(hw);
1599
1600         /* Reset the hw statistics */
1601         ixgbevf_dev_stats_reset(eth_dev);
1602
1603         /* Disable the interrupts for VF */
1604         ixgbevf_intr_disable(eth_dev);
1605
1606         hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1607         diag = hw->mac.ops.reset_hw(hw);
1608
1609         /*
1610          * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1611          * the underlying PF driver has not assigned a MAC address to the VF.
1612          * In this case, assign a random MAC address.
1613          */
1614         if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1615                 PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1616                 /*
1617                  * This error code will be propagated to the app by
1618                  * rte_eth_dev_reset, so use a public error code rather than
1619                  * the internal-only IXGBE_ERR_RESET_FAILED
1620                  */
1621                 return -EAGAIN;
1622         }
1623
1624         /* negotiate mailbox API version to use with the PF. */
1625         ixgbevf_negotiate_api(hw);
1626
1627         /* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1628         ixgbevf_get_queues(hw, &tcs, &tc);
1629
1630         /* Allocate memory for storing MAC addresses */
1631         eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", RTE_ETHER_ADDR_LEN *
1632                                                hw->mac.num_rar_entries, 0);
1633         if (eth_dev->data->mac_addrs == NULL) {
1634                 PMD_INIT_LOG(ERR,
1635                              "Failed to allocate %u bytes needed to store "
1636                              "MAC addresses",
1637                              RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1638                 return -ENOMEM;
1639         }
1640
1641         /* Generate a random MAC address, if none was assigned by PF. */
1642         if (rte_is_zero_ether_addr(perm_addr)) {
1643                 generate_random_mac_addr(perm_addr);
1644                 diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1645                 if (diag) {
1646                         rte_free(eth_dev->data->mac_addrs);
1647                         eth_dev->data->mac_addrs = NULL;
1648                         return diag;
1649                 }
1650                 PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1651                 PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1652                              "%02x:%02x:%02x:%02x:%02x:%02x",
1653                              perm_addr->addr_bytes[0],
1654                              perm_addr->addr_bytes[1],
1655                              perm_addr->addr_bytes[2],
1656                              perm_addr->addr_bytes[3],
1657                              perm_addr->addr_bytes[4],
1658                              perm_addr->addr_bytes[5]);
1659         }
1660
1661         /* Copy the permanent MAC address */
1662         rte_ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1663
1664         /* reset the hardware with the new settings */
1665         diag = hw->mac.ops.start_hw(hw);
1666         switch (diag) {
1667         case  0:
1668                 break;
1669
1670         default:
1671                 PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1672                 return -EIO;
1673         }
1674
1675         rte_intr_callback_register(intr_handle,
1676                                    ixgbevf_dev_interrupt_handler, eth_dev);
1677         rte_intr_enable(intr_handle);
1678         ixgbevf_intr_enable(eth_dev);
1679
1680         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1681                      eth_dev->data->port_id, pci_dev->id.vendor_id,
1682                      pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1683
1684         return 0;
1685 }
1686
1687 /* Virtual Function device uninit */
1688
1689 static int
1690 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1691 {
1692         PMD_INIT_FUNC_TRACE();
1693
1694         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1695                 return 0;
1696
1697         ixgbevf_dev_close(eth_dev);
1698
1699         return 0;
1700 }
1701
1702 static int
1703 eth_ixgbe_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1704                 struct rte_pci_device *pci_dev)
1705 {
1706         char name[RTE_ETH_NAME_MAX_LEN];
1707         struct rte_eth_dev *pf_ethdev;
1708         struct rte_eth_devargs eth_da;
1709         int i, retval;
1710
1711         if (pci_dev->device.devargs) {
1712                 retval = rte_eth_devargs_parse(pci_dev->device.devargs->args,
1713                                 &eth_da);
1714                 if (retval)
1715                         return retval;
1716         } else
1717                 memset(&eth_da, 0, sizeof(eth_da));
1718
1719         if (eth_da.nb_representor_ports > 0 &&
1720             eth_da.type != RTE_ETH_REPRESENTOR_VF) {
1721                 PMD_DRV_LOG(ERR, "unsupported representor type: %s\n",
1722                             pci_dev->device.devargs->args);
1723                 return -ENOTSUP;
1724         }
1725
1726         retval = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
1727                 sizeof(struct ixgbe_adapter),
1728                 eth_dev_pci_specific_init, pci_dev,
1729                 eth_ixgbe_dev_init, NULL);
1730
1731         if (retval || eth_da.nb_representor_ports < 1)
1732                 return retval;
1733
1734         pf_ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1735         if (pf_ethdev == NULL)
1736                 return -ENODEV;
1737
1738         /* probe VF representor ports */
1739         for (i = 0; i < eth_da.nb_representor_ports; i++) {
1740                 struct ixgbe_vf_info *vfinfo;
1741                 struct ixgbe_vf_representor representor;
1742
1743                 vfinfo = *IXGBE_DEV_PRIVATE_TO_P_VFDATA(
1744                         pf_ethdev->data->dev_private);
1745                 if (vfinfo == NULL) {
1746                         PMD_DRV_LOG(ERR,
1747                                 "no virtual functions supported by PF");
1748                         break;
1749                 }
1750
1751                 representor.vf_id = eth_da.representor_ports[i];
1752                 representor.switch_domain_id = vfinfo->switch_domain_id;
1753                 representor.pf_ethdev = pf_ethdev;
1754
1755                 /* representor port net_bdf_port */
1756                 snprintf(name, sizeof(name), "net_%s_representor_%d",
1757                         pci_dev->device.name,
1758                         eth_da.representor_ports[i]);
1759
1760                 retval = rte_eth_dev_create(&pci_dev->device, name,
1761                         sizeof(struct ixgbe_vf_representor), NULL, NULL,
1762                         ixgbe_vf_representor_init, &representor);
1763
1764                 if (retval)
1765                         PMD_DRV_LOG(ERR, "failed to create ixgbe vf "
1766                                 "representor %s.", name);
1767         }
1768
1769         return 0;
1770 }
1771
1772 static int eth_ixgbe_pci_remove(struct rte_pci_device *pci_dev)
1773 {
1774         struct rte_eth_dev *ethdev;
1775
1776         ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1777         if (!ethdev)
1778                 return 0;
1779
1780         if (ethdev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
1781                 return rte_eth_dev_pci_generic_remove(pci_dev,
1782                                         ixgbe_vf_representor_uninit);
1783         else
1784                 return rte_eth_dev_pci_generic_remove(pci_dev,
1785                                                 eth_ixgbe_dev_uninit);
1786 }
1787
1788 static struct rte_pci_driver rte_ixgbe_pmd = {
1789         .id_table = pci_id_ixgbe_map,
1790         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1791         .probe = eth_ixgbe_pci_probe,
1792         .remove = eth_ixgbe_pci_remove,
1793 };
1794
1795 static int eth_ixgbevf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1796         struct rte_pci_device *pci_dev)
1797 {
1798         return rte_eth_dev_pci_generic_probe(pci_dev,
1799                 sizeof(struct ixgbe_adapter), eth_ixgbevf_dev_init);
1800 }
1801
1802 static int eth_ixgbevf_pci_remove(struct rte_pci_device *pci_dev)
1803 {
1804         return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbevf_dev_uninit);
1805 }
1806
1807 /*
1808  * virtual function driver struct
1809  */
1810 static struct rte_pci_driver rte_ixgbevf_pmd = {
1811         .id_table = pci_id_ixgbevf_map,
1812         .drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1813         .probe = eth_ixgbevf_pci_probe,
1814         .remove = eth_ixgbevf_pci_remove,
1815 };
1816
1817 static int
1818 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1819 {
1820         struct ixgbe_hw *hw =
1821                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1822         struct ixgbe_vfta *shadow_vfta =
1823                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1824         uint32_t vfta;
1825         uint32_t vid_idx;
1826         uint32_t vid_bit;
1827
1828         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1829         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1830         vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1831         if (on)
1832                 vfta |= vid_bit;
1833         else
1834                 vfta &= ~vid_bit;
1835         IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1836
1837         /* update local VFTA copy */
1838         shadow_vfta->vfta[vid_idx] = vfta;
1839
1840         return 0;
1841 }
1842
1843 static void
1844 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1845 {
1846         if (on)
1847                 ixgbe_vlan_hw_strip_enable(dev, queue);
1848         else
1849                 ixgbe_vlan_hw_strip_disable(dev, queue);
1850 }
1851
1852 static int
1853 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1854                     enum rte_vlan_type vlan_type,
1855                     uint16_t tpid)
1856 {
1857         struct ixgbe_hw *hw =
1858                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1859         int ret = 0;
1860         uint32_t reg;
1861         uint32_t qinq;
1862
1863         qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1864         qinq &= IXGBE_DMATXCTL_GDV;
1865
1866         switch (vlan_type) {
1867         case ETH_VLAN_TYPE_INNER:
1868                 if (qinq) {
1869                         reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1870                         reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1871                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1872                         reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1873                         reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1874                                 | ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1875                         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1876                 } else {
1877                         ret = -ENOTSUP;
1878                         PMD_DRV_LOG(ERR, "Inner type is not supported"
1879                                     " by single VLAN");
1880                 }
1881                 break;
1882         case ETH_VLAN_TYPE_OUTER:
1883                 if (qinq) {
1884                         /* Only the high 16-bits is valid */
1885                         IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1886                                         IXGBE_EXVET_VET_EXT_SHIFT);
1887                 } else {
1888                         reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1889                         reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1890                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1891                         reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1892                         reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1893                                 | ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1894                         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1895                 }
1896
1897                 break;
1898         default:
1899                 ret = -EINVAL;
1900                 PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1901                 break;
1902         }
1903
1904         return ret;
1905 }
1906
1907 void
1908 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1909 {
1910         struct ixgbe_hw *hw =
1911                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1912         uint32_t vlnctrl;
1913
1914         PMD_INIT_FUNC_TRACE();
1915
1916         /* Filter Table Disable */
1917         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1918         vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1919
1920         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1921 }
1922
1923 void
1924 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1925 {
1926         struct ixgbe_hw *hw =
1927                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1928         struct ixgbe_vfta *shadow_vfta =
1929                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1930         uint32_t vlnctrl;
1931         uint16_t i;
1932
1933         PMD_INIT_FUNC_TRACE();
1934
1935         /* Filter Table Enable */
1936         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1937         vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1938         vlnctrl |= IXGBE_VLNCTRL_VFE;
1939
1940         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1941
1942         /* write whatever is in local vfta copy */
1943         for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1944                 IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1945 }
1946
1947 static void
1948 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1949 {
1950         struct ixgbe_hwstrip *hwstrip =
1951                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1952         struct ixgbe_rx_queue *rxq;
1953
1954         if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
1955                 return;
1956
1957         if (on)
1958                 IXGBE_SET_HWSTRIP(hwstrip, queue);
1959         else
1960                 IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1961
1962         if (queue >= dev->data->nb_rx_queues)
1963                 return;
1964
1965         rxq = dev->data->rx_queues[queue];
1966
1967         if (on) {
1968                 rxq->vlan_flags = PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
1969                 rxq->offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
1970         } else {
1971                 rxq->vlan_flags = PKT_RX_VLAN;
1972                 rxq->offloads &= ~DEV_RX_OFFLOAD_VLAN_STRIP;
1973         }
1974 }
1975
1976 static void
1977 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
1978 {
1979         struct ixgbe_hw *hw =
1980                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1981         uint32_t ctrl;
1982
1983         PMD_INIT_FUNC_TRACE();
1984
1985         if (hw->mac.type == ixgbe_mac_82598EB) {
1986                 /* No queue level support */
1987                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1988                 return;
1989         }
1990
1991         /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1992         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1993         ctrl &= ~IXGBE_RXDCTL_VME;
1994         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1995
1996         /* record those setting for HW strip per queue */
1997         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
1998 }
1999
2000 static void
2001 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
2002 {
2003         struct ixgbe_hw *hw =
2004                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2005         uint32_t ctrl;
2006
2007         PMD_INIT_FUNC_TRACE();
2008
2009         if (hw->mac.type == ixgbe_mac_82598EB) {
2010                 /* No queue level supported */
2011                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2012                 return;
2013         }
2014
2015         /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2016         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2017         ctrl |= IXGBE_RXDCTL_VME;
2018         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2019
2020         /* record those setting for HW strip per queue */
2021         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
2022 }
2023
2024 static void
2025 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2026 {
2027         struct ixgbe_hw *hw =
2028                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2029         uint32_t ctrl;
2030
2031         PMD_INIT_FUNC_TRACE();
2032
2033         /* DMATXCTRL: Geric Double VLAN Disable */
2034         ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2035         ctrl &= ~IXGBE_DMATXCTL_GDV;
2036         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2037
2038         /* CTRL_EXT: Global Double VLAN Disable */
2039         ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2040         ctrl &= ~IXGBE_EXTENDED_VLAN;
2041         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2042
2043 }
2044
2045 static void
2046 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2047 {
2048         struct ixgbe_hw *hw =
2049                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2050         uint32_t ctrl;
2051
2052         PMD_INIT_FUNC_TRACE();
2053
2054         /* DMATXCTRL: Geric Double VLAN Enable */
2055         ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2056         ctrl |= IXGBE_DMATXCTL_GDV;
2057         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2058
2059         /* CTRL_EXT: Global Double VLAN Enable */
2060         ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2061         ctrl |= IXGBE_EXTENDED_VLAN;
2062         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2063
2064         /* Clear pooling mode of PFVTCTL. It's required by X550. */
2065         if (hw->mac.type == ixgbe_mac_X550 ||
2066             hw->mac.type == ixgbe_mac_X550EM_x ||
2067             hw->mac.type == ixgbe_mac_X550EM_a) {
2068                 ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
2069                 ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
2070                 IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
2071         }
2072
2073         /*
2074          * VET EXT field in the EXVET register = 0x8100 by default
2075          * So no need to change. Same to VT field of DMATXCTL register
2076          */
2077 }
2078
2079 void
2080 ixgbe_vlan_hw_strip_config(struct rte_eth_dev *dev)
2081 {
2082         struct ixgbe_hw *hw =
2083                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2084         struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
2085         uint32_t ctrl;
2086         uint16_t i;
2087         struct ixgbe_rx_queue *rxq;
2088         bool on;
2089
2090         PMD_INIT_FUNC_TRACE();
2091
2092         if (hw->mac.type == ixgbe_mac_82598EB) {
2093                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP) {
2094                         ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2095                         ctrl |= IXGBE_VLNCTRL_VME;
2096                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2097                 } else {
2098                         ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2099                         ctrl &= ~IXGBE_VLNCTRL_VME;
2100                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2101                 }
2102         } else {
2103                 /*
2104                  * Other 10G NIC, the VLAN strip can be setup
2105                  * per queue in RXDCTL
2106                  */
2107                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
2108                         rxq = dev->data->rx_queues[i];
2109                         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2110                         if (rxq->offloads & DEV_RX_OFFLOAD_VLAN_STRIP) {
2111                                 ctrl |= IXGBE_RXDCTL_VME;
2112                                 on = TRUE;
2113                         } else {
2114                                 ctrl &= ~IXGBE_RXDCTL_VME;
2115                                 on = FALSE;
2116                         }
2117                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2118
2119                         /* record those setting for HW strip per queue */
2120                         ixgbe_vlan_hw_strip_bitmap_set(dev, i, on);
2121                 }
2122         }
2123 }
2124
2125 static void
2126 ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev, int mask)
2127 {
2128         uint16_t i;
2129         struct rte_eth_rxmode *rxmode;
2130         struct ixgbe_rx_queue *rxq;
2131
2132         if (mask & ETH_VLAN_STRIP_MASK) {
2133                 rxmode = &dev->data->dev_conf.rxmode;
2134                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2135                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
2136                                 rxq = dev->data->rx_queues[i];
2137                                 rxq->offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
2138                         }
2139                 else
2140                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
2141                                 rxq = dev->data->rx_queues[i];
2142                                 rxq->offloads &= ~DEV_RX_OFFLOAD_VLAN_STRIP;
2143                         }
2144         }
2145 }
2146
2147 static int
2148 ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask)
2149 {
2150         struct rte_eth_rxmode *rxmode;
2151         rxmode = &dev->data->dev_conf.rxmode;
2152
2153         if (mask & ETH_VLAN_STRIP_MASK) {
2154                 ixgbe_vlan_hw_strip_config(dev);
2155         }
2156
2157         if (mask & ETH_VLAN_FILTER_MASK) {
2158                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2159                         ixgbe_vlan_hw_filter_enable(dev);
2160                 else
2161                         ixgbe_vlan_hw_filter_disable(dev);
2162         }
2163
2164         if (mask & ETH_VLAN_EXTEND_MASK) {
2165                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2166                         ixgbe_vlan_hw_extend_enable(dev);
2167                 else
2168                         ixgbe_vlan_hw_extend_disable(dev);
2169         }
2170
2171         return 0;
2172 }
2173
2174 static int
2175 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2176 {
2177         ixgbe_config_vlan_strip_on_all_queues(dev, mask);
2178
2179         ixgbe_vlan_offload_config(dev, mask);
2180
2181         return 0;
2182 }
2183
2184 static void
2185 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2186 {
2187         struct ixgbe_hw *hw =
2188                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2189         /* VLNCTRL: enable vlan filtering and allow all vlan tags through */
2190         uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2191
2192         vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
2193         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
2194 }
2195
2196 static int
2197 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
2198 {
2199         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2200
2201         switch (nb_rx_q) {
2202         case 1:
2203         case 2:
2204                 RTE_ETH_DEV_SRIOV(dev).active = ETH_64_POOLS;
2205                 break;
2206         case 4:
2207                 RTE_ETH_DEV_SRIOV(dev).active = ETH_32_POOLS;
2208                 break;
2209         default:
2210                 return -EINVAL;
2211         }
2212
2213         RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
2214                 IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2215         RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
2216                 pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2217         return 0;
2218 }
2219
2220 static int
2221 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
2222 {
2223         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
2224         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2225         uint16_t nb_rx_q = dev->data->nb_rx_queues;
2226         uint16_t nb_tx_q = dev->data->nb_tx_queues;
2227
2228         if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
2229                 /* check multi-queue mode */
2230                 switch (dev_conf->rxmode.mq_mode) {
2231                 case ETH_MQ_RX_VMDQ_DCB:
2232                         PMD_INIT_LOG(INFO, "ETH_MQ_RX_VMDQ_DCB mode supported in SRIOV");
2233                         break;
2234                 case ETH_MQ_RX_VMDQ_DCB_RSS:
2235                         /* DCB/RSS VMDQ in SRIOV mode, not implement yet */
2236                         PMD_INIT_LOG(ERR, "SRIOV active,"
2237                                         " unsupported mq_mode rx %d.",
2238                                         dev_conf->rxmode.mq_mode);
2239                         return -EINVAL;
2240                 case ETH_MQ_RX_RSS:
2241                 case ETH_MQ_RX_VMDQ_RSS:
2242                         dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_RSS;
2243                         if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
2244                                 if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
2245                                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2246                                                 " invalid queue number"
2247                                                 " for VMDQ RSS, allowed"
2248                                                 " value are 1, 2 or 4.");
2249                                         return -EINVAL;
2250                                 }
2251                         break;
2252                 case ETH_MQ_RX_VMDQ_ONLY:
2253                 case ETH_MQ_RX_NONE:
2254                         /* if nothing mq mode configure, use default scheme */
2255                         dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
2256                         break;
2257                 default: /* ETH_MQ_RX_DCB, ETH_MQ_RX_DCB_RSS or ETH_MQ_TX_DCB*/
2258                         /* SRIOV only works in VMDq enable mode */
2259                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2260                                         " wrong mq_mode rx %d.",
2261                                         dev_conf->rxmode.mq_mode);
2262                         return -EINVAL;
2263                 }
2264
2265                 switch (dev_conf->txmode.mq_mode) {
2266                 case ETH_MQ_TX_VMDQ_DCB:
2267                         PMD_INIT_LOG(INFO, "ETH_MQ_TX_VMDQ_DCB mode supported in SRIOV");
2268                         dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_DCB;
2269                         break;
2270                 default: /* ETH_MQ_TX_VMDQ_ONLY or ETH_MQ_TX_NONE */
2271                         dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_ONLY;
2272                         break;
2273                 }
2274
2275                 /* check valid queue number */
2276                 if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2277                     (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2278                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2279                                         " nb_rx_q=%d nb_tx_q=%d queue number"
2280                                         " must be less than or equal to %d.",
2281                                         nb_rx_q, nb_tx_q,
2282                                         RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2283                         return -EINVAL;
2284                 }
2285         } else {
2286                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB_RSS) {
2287                         PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2288                                           " not supported.");
2289                         return -EINVAL;
2290                 }
2291                 /* check configuration for vmdb+dcb mode */
2292                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB) {
2293                         const struct rte_eth_vmdq_dcb_conf *conf;
2294
2295                         if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2296                                 PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2297                                                 IXGBE_VMDQ_DCB_NB_QUEUES);
2298                                 return -EINVAL;
2299                         }
2300                         conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2301                         if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2302                                conf->nb_queue_pools == ETH_32_POOLS)) {
2303                                 PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2304                                                 " nb_queue_pools must be %d or %d.",
2305                                                 ETH_16_POOLS, ETH_32_POOLS);
2306                                 return -EINVAL;
2307                         }
2308                 }
2309                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_VMDQ_DCB) {
2310                         const struct rte_eth_vmdq_dcb_tx_conf *conf;
2311
2312                         if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2313                                 PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2314                                                  IXGBE_VMDQ_DCB_NB_QUEUES);
2315                                 return -EINVAL;
2316                         }
2317                         conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2318                         if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2319                                conf->nb_queue_pools == ETH_32_POOLS)) {
2320                                 PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2321                                                 " nb_queue_pools != %d and"
2322                                                 " nb_queue_pools != %d.",
2323                                                 ETH_16_POOLS, ETH_32_POOLS);
2324                                 return -EINVAL;
2325                         }
2326                 }
2327
2328                 /* For DCB mode check our configuration before we go further */
2329                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB) {
2330                         const struct rte_eth_dcb_rx_conf *conf;
2331
2332                         conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2333                         if (!(conf->nb_tcs == ETH_4_TCS ||
2334                                conf->nb_tcs == ETH_8_TCS)) {
2335                                 PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2336                                                 " and nb_tcs != %d.",
2337                                                 ETH_4_TCS, ETH_8_TCS);
2338                                 return -EINVAL;
2339                         }
2340                 }
2341
2342                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_DCB) {
2343                         const struct rte_eth_dcb_tx_conf *conf;
2344
2345                         conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2346                         if (!(conf->nb_tcs == ETH_4_TCS ||
2347                                conf->nb_tcs == ETH_8_TCS)) {
2348                                 PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2349                                                 " and nb_tcs != %d.",
2350                                                 ETH_4_TCS, ETH_8_TCS);
2351                                 return -EINVAL;
2352                         }
2353                 }
2354
2355                 /*
2356                  * When DCB/VT is off, maximum number of queues changes,
2357                  * except for 82598EB, which remains constant.
2358                  */
2359                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
2360                                 hw->mac.type != ixgbe_mac_82598EB) {
2361                         if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2362                                 PMD_INIT_LOG(ERR,
2363                                              "Neither VT nor DCB are enabled, "
2364                                              "nb_tx_q > %d.",
2365                                              IXGBE_NONE_MODE_TX_NB_QUEUES);
2366                                 return -EINVAL;
2367                         }
2368                 }
2369         }
2370         return 0;
2371 }
2372
2373 static int
2374 ixgbe_dev_configure(struct rte_eth_dev *dev)
2375 {
2376         struct ixgbe_interrupt *intr =
2377                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2378         struct ixgbe_adapter *adapter = dev->data->dev_private;
2379         int ret;
2380
2381         PMD_INIT_FUNC_TRACE();
2382
2383         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
2384                 dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
2385
2386         /* multipe queue mode checking */
2387         ret  = ixgbe_check_mq_mode(dev);
2388         if (ret != 0) {
2389                 PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2390                             ret);
2391                 return ret;
2392         }
2393
2394         /* set flag to update link status after init */
2395         intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2396
2397         /*
2398          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2399          * allocation or vector Rx preconditions we will reset it.
2400          */
2401         adapter->rx_bulk_alloc_allowed = true;
2402         adapter->rx_vec_allowed = true;
2403
2404         return 0;
2405 }
2406
2407 static void
2408 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2409 {
2410         struct ixgbe_hw *hw =
2411                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2412         struct ixgbe_interrupt *intr =
2413                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2414         uint32_t gpie;
2415
2416         /* only set up it on X550EM_X */
2417         if (hw->mac.type == ixgbe_mac_X550EM_x) {
2418                 gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2419                 gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2420                 IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2421                 if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2422                         intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2423         }
2424 }
2425
2426 int
2427 ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
2428                         uint16_t tx_rate, uint64_t q_msk)
2429 {
2430         struct ixgbe_hw *hw;
2431         struct ixgbe_vf_info *vfinfo;
2432         struct rte_eth_link link;
2433         uint8_t  nb_q_per_pool;
2434         uint32_t queue_stride;
2435         uint32_t queue_idx, idx = 0, vf_idx;
2436         uint32_t queue_end;
2437         uint16_t total_rate = 0;
2438         struct rte_pci_device *pci_dev;
2439         int ret;
2440
2441         pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2442         ret = rte_eth_link_get_nowait(dev->data->port_id, &link);
2443         if (ret < 0)
2444                 return ret;
2445
2446         if (vf >= pci_dev->max_vfs)
2447                 return -EINVAL;
2448
2449         if (tx_rate > link.link_speed)
2450                 return -EINVAL;
2451
2452         if (q_msk == 0)
2453                 return 0;
2454
2455         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2456         vfinfo = *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
2457         nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2458         queue_stride = IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2459         queue_idx = vf * queue_stride;
2460         queue_end = queue_idx + nb_q_per_pool - 1;
2461         if (queue_end >= hw->mac.max_tx_queues)
2462                 return -EINVAL;
2463
2464         if (vfinfo) {
2465                 for (vf_idx = 0; vf_idx < pci_dev->max_vfs; vf_idx++) {
2466                         if (vf_idx == vf)
2467                                 continue;
2468                         for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
2469                                 idx++)
2470                                 total_rate += vfinfo[vf_idx].tx_rate[idx];
2471                 }
2472         } else {
2473                 return -EINVAL;
2474         }
2475
2476         /* Store tx_rate for this vf. */
2477         for (idx = 0; idx < nb_q_per_pool; idx++) {
2478                 if (((uint64_t)0x1 << idx) & q_msk) {
2479                         if (vfinfo[vf].tx_rate[idx] != tx_rate)
2480                                 vfinfo[vf].tx_rate[idx] = tx_rate;
2481                         total_rate += tx_rate;
2482                 }
2483         }
2484
2485         if (total_rate > dev->data->dev_link.link_speed) {
2486                 /* Reset stored TX rate of the VF if it causes exceed
2487                  * link speed.
2488                  */
2489                 memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
2490                 return -EINVAL;
2491         }
2492
2493         /* Set RTTBCNRC of each queue/pool for vf X  */
2494         for (; queue_idx <= queue_end; queue_idx++) {
2495                 if (0x1 & q_msk)
2496                         ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
2497                 q_msk = q_msk >> 1;
2498         }
2499
2500         return 0;
2501 }
2502
2503 static int
2504 ixgbe_flow_ctrl_enable(struct rte_eth_dev *dev, struct ixgbe_hw *hw)
2505 {
2506         struct ixgbe_adapter *adapter = dev->data->dev_private;
2507         int err;
2508         uint32_t mflcn;
2509
2510         ixgbe_setup_fc(hw);
2511
2512         err = ixgbe_fc_enable(hw);
2513
2514         /* Not negotiated is not an error case */
2515         if (err == IXGBE_SUCCESS || err == IXGBE_ERR_FC_NOT_NEGOTIATED) {
2516                 /*
2517                  *check if we want to forward MAC frames - driver doesn't
2518                  *have native capability to do that,
2519                  *so we'll write the registers ourselves
2520                  */
2521
2522                 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
2523
2524                 /* set or clear MFLCN.PMCF bit depending on configuration */
2525                 if (adapter->mac_ctrl_frame_fwd != 0)
2526                         mflcn |= IXGBE_MFLCN_PMCF;
2527                 else
2528                         mflcn &= ~IXGBE_MFLCN_PMCF;
2529
2530                 IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
2531                 IXGBE_WRITE_FLUSH(hw);
2532
2533                 return 0;
2534         }
2535         return err;
2536 }
2537
2538 /*
2539  * Configure device link speed and setup link.
2540  * It returns 0 on success.
2541  */
2542 static int
2543 ixgbe_dev_start(struct rte_eth_dev *dev)
2544 {
2545         struct ixgbe_hw *hw =
2546                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2547         struct ixgbe_vf_info *vfinfo =
2548                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2549         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2550         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2551         uint32_t intr_vector = 0;
2552         int err;
2553         bool link_up = false, negotiate = 0;
2554         uint32_t speed = 0;
2555         uint32_t allowed_speeds = 0;
2556         int mask = 0;
2557         int status;
2558         uint16_t vf, idx;
2559         uint32_t *link_speeds;
2560         struct ixgbe_tm_conf *tm_conf =
2561                 IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2562         struct ixgbe_macsec_setting *macsec_setting =
2563                 IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
2564
2565         PMD_INIT_FUNC_TRACE();
2566
2567         /* Stop the link setup handler before resetting the HW. */
2568         ixgbe_dev_wait_setup_link_complete(dev, 0);
2569
2570         /* disable uio/vfio intr/eventfd mapping */
2571         rte_intr_disable(intr_handle);
2572
2573         /* stop adapter */
2574         hw->adapter_stopped = 0;
2575         ixgbe_stop_adapter(hw);
2576
2577         /* reinitialize adapter
2578          * this calls reset and start
2579          */
2580         status = ixgbe_pf_reset_hw(hw);
2581         if (status != 0)
2582                 return -1;
2583         hw->mac.ops.start_hw(hw);
2584         hw->mac.get_link_status = true;
2585
2586         /* configure PF module if SRIOV enabled */
2587         ixgbe_pf_host_configure(dev);
2588
2589         ixgbe_dev_phy_intr_setup(dev);
2590
2591         /* check and configure queue intr-vector mapping */
2592         if ((rte_intr_cap_multiple(intr_handle) ||
2593              !RTE_ETH_DEV_SRIOV(dev).active) &&
2594             dev->data->dev_conf.intr_conf.rxq != 0) {
2595                 intr_vector = dev->data->nb_rx_queues;
2596                 if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2597                         PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2598                                         IXGBE_MAX_INTR_QUEUE_NUM);
2599                         return -ENOTSUP;
2600                 }
2601                 if (rte_intr_efd_enable(intr_handle, intr_vector))
2602                         return -1;
2603         }
2604
2605         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
2606                 intr_handle->intr_vec =
2607                         rte_zmalloc("intr_vec",
2608                                     dev->data->nb_rx_queues * sizeof(int), 0);
2609                 if (intr_handle->intr_vec == NULL) {
2610                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2611                                      " intr_vec", dev->data->nb_rx_queues);
2612                         return -ENOMEM;
2613                 }
2614         }
2615
2616         /* confiugre msix for sleep until rx interrupt */
2617         ixgbe_configure_msix(dev);
2618
2619         /* initialize transmission unit */
2620         ixgbe_dev_tx_init(dev);
2621
2622         /* This can fail when allocating mbufs for descriptor rings */
2623         err = ixgbe_dev_rx_init(dev);
2624         if (err) {
2625                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2626                 goto error;
2627         }
2628
2629         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
2630                 ETH_VLAN_EXTEND_MASK;
2631         err = ixgbe_vlan_offload_config(dev, mask);
2632         if (err) {
2633                 PMD_INIT_LOG(ERR, "Unable to set VLAN offload");
2634                 goto error;
2635         }
2636
2637         if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
2638                 /* Enable vlan filtering for VMDq */
2639                 ixgbe_vmdq_vlan_hw_filter_enable(dev);
2640         }
2641
2642         /* Configure DCB hw */
2643         ixgbe_configure_dcb(dev);
2644
2645         if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2646                 err = ixgbe_fdir_configure(dev);
2647                 if (err)
2648                         goto error;
2649         }
2650
2651         /* Restore vf rate limit */
2652         if (vfinfo != NULL) {
2653                 for (vf = 0; vf < pci_dev->max_vfs; vf++)
2654                         for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2655                                 if (vfinfo[vf].tx_rate[idx] != 0)
2656                                         ixgbe_set_vf_rate_limit(
2657                                                 dev, vf,
2658                                                 vfinfo[vf].tx_rate[idx],
2659                                                 1 << idx);
2660         }
2661
2662         ixgbe_restore_statistics_mapping(dev);
2663
2664         err = ixgbe_flow_ctrl_enable(dev, hw);
2665         if (err < 0) {
2666                 PMD_INIT_LOG(ERR, "enable flow ctrl err");
2667                 goto error;
2668         }
2669
2670         err = ixgbe_dev_rxtx_start(dev);
2671         if (err < 0) {
2672                 PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2673                 goto error;
2674         }
2675
2676         /* Skip link setup if loopback mode is enabled. */
2677         if (dev->data->dev_conf.lpbk_mode != 0) {
2678                 err = ixgbe_check_supported_loopback_mode(dev);
2679                 if (err < 0) {
2680                         PMD_INIT_LOG(ERR, "Unsupported loopback mode");
2681                         goto error;
2682                 } else {
2683                         goto skip_link_setup;
2684                 }
2685         }
2686
2687         if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2688                 err = hw->mac.ops.setup_sfp(hw);
2689                 if (err)
2690                         goto error;
2691         }
2692
2693         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2694                 /* Turn on the copper */
2695                 ixgbe_set_phy_power(hw, true);
2696         } else {
2697                 /* Turn on the laser */
2698                 ixgbe_enable_tx_laser(hw);
2699         }
2700
2701         err = ixgbe_check_link(hw, &speed, &link_up, 0);
2702         if (err)
2703                 goto error;
2704         dev->data->dev_link.link_status = link_up;
2705
2706         err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2707         if (err)
2708                 goto error;
2709
2710         switch (hw->mac.type) {
2711         case ixgbe_mac_X550:
2712         case ixgbe_mac_X550EM_x:
2713         case ixgbe_mac_X550EM_a:
2714                 allowed_speeds = ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2715                         ETH_LINK_SPEED_2_5G |  ETH_LINK_SPEED_5G |
2716                         ETH_LINK_SPEED_10G;
2717                 if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
2718                                 hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
2719                         allowed_speeds = ETH_LINK_SPEED_10M |
2720                                 ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G;
2721                 break;
2722         default:
2723                 allowed_speeds = ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2724                         ETH_LINK_SPEED_10G;
2725         }
2726
2727         link_speeds = &dev->data->dev_conf.link_speeds;
2728
2729         /* Ignore autoneg flag bit and check the validity of 
2730          * link_speed 
2731          */
2732         if (((*link_speeds) >> 1) & ~(allowed_speeds >> 1)) {
2733                 PMD_INIT_LOG(ERR, "Invalid link setting");
2734                 goto error;
2735         }
2736
2737         speed = 0x0;
2738         if (*link_speeds == ETH_LINK_SPEED_AUTONEG) {
2739                 switch (hw->mac.type) {
2740                 case ixgbe_mac_82598EB:
2741                         speed = IXGBE_LINK_SPEED_82598_AUTONEG;
2742                         break;
2743                 case ixgbe_mac_82599EB:
2744                 case ixgbe_mac_X540:
2745                         speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2746                         break;
2747                 case ixgbe_mac_X550:
2748                 case ixgbe_mac_X550EM_x:
2749                 case ixgbe_mac_X550EM_a:
2750                         speed = IXGBE_LINK_SPEED_X550_AUTONEG;
2751                         break;
2752                 default:
2753                         speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2754                 }
2755         } else {
2756                 if (*link_speeds & ETH_LINK_SPEED_10G)
2757                         speed |= IXGBE_LINK_SPEED_10GB_FULL;
2758                 if (*link_speeds & ETH_LINK_SPEED_5G)
2759                         speed |= IXGBE_LINK_SPEED_5GB_FULL;
2760                 if (*link_speeds & ETH_LINK_SPEED_2_5G)
2761                         speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
2762                 if (*link_speeds & ETH_LINK_SPEED_1G)
2763                         speed |= IXGBE_LINK_SPEED_1GB_FULL;
2764                 if (*link_speeds & ETH_LINK_SPEED_100M)
2765                         speed |= IXGBE_LINK_SPEED_100_FULL;
2766                 if (*link_speeds & ETH_LINK_SPEED_10M)
2767                         speed |= IXGBE_LINK_SPEED_10_FULL;
2768         }
2769
2770         err = ixgbe_setup_link(hw, speed, link_up);
2771         if (err)
2772                 goto error;
2773
2774 skip_link_setup:
2775
2776         if (rte_intr_allow_others(intr_handle)) {
2777                 /* check if lsc interrupt is enabled */
2778                 if (dev->data->dev_conf.intr_conf.lsc != 0)
2779                         ixgbe_dev_lsc_interrupt_setup(dev, TRUE);
2780                 else
2781                         ixgbe_dev_lsc_interrupt_setup(dev, FALSE);
2782                 ixgbe_dev_macsec_interrupt_setup(dev);
2783         } else {
2784                 rte_intr_callback_unregister(intr_handle,
2785                                              ixgbe_dev_interrupt_handler, dev);
2786                 if (dev->data->dev_conf.intr_conf.lsc != 0)
2787                         PMD_INIT_LOG(INFO, "lsc won't enable because of"
2788                                      " no intr multiplex");
2789         }
2790
2791         /* check if rxq interrupt is enabled */
2792         if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2793             rte_intr_dp_is_en(intr_handle))
2794                 ixgbe_dev_rxq_interrupt_setup(dev);
2795
2796         /* enable uio/vfio intr/eventfd mapping */
2797         rte_intr_enable(intr_handle);
2798
2799         /* resume enabled intr since hw reset */
2800         ixgbe_enable_intr(dev);
2801         ixgbe_l2_tunnel_conf(dev);
2802         ixgbe_filter_restore(dev);
2803
2804         if (tm_conf->root && !tm_conf->committed)
2805                 PMD_DRV_LOG(WARNING,
2806                             "please call hierarchy_commit() "
2807                             "before starting the port");
2808
2809         /* wait for the controller to acquire link */
2810         err = ixgbe_wait_for_link_up(hw);
2811         if (err)
2812                 goto error;
2813
2814         /*
2815          * Update link status right before return, because it may
2816          * start link configuration process in a separate thread.
2817          */
2818         ixgbe_dev_link_update(dev, 0);
2819
2820         /* setup the macsec setting register */
2821         if (macsec_setting->offload_en)
2822                 ixgbe_dev_macsec_register_enable(dev, macsec_setting);
2823
2824         return 0;
2825
2826 error:
2827         PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2828         ixgbe_dev_clear_queues(dev);
2829         return -EIO;
2830 }
2831
2832 /*
2833  * Stop device: disable rx and tx functions to allow for reconfiguring.
2834  */
2835 static int
2836 ixgbe_dev_stop(struct rte_eth_dev *dev)
2837 {
2838         struct rte_eth_link link;
2839         struct ixgbe_adapter *adapter = dev->data->dev_private;
2840         struct ixgbe_hw *hw =
2841                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2842         struct ixgbe_vf_info *vfinfo =
2843                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2844         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2845         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2846         int vf;
2847         struct ixgbe_tm_conf *tm_conf =
2848                 IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2849
2850         if (hw->adapter_stopped)
2851                 return 0;
2852
2853         PMD_INIT_FUNC_TRACE();
2854
2855         ixgbe_dev_wait_setup_link_complete(dev, 0);
2856
2857         /* disable interrupts */
2858         ixgbe_disable_intr(hw);
2859
2860         /* reset the NIC */
2861         ixgbe_pf_reset_hw(hw);
2862         hw->adapter_stopped = 0;
2863
2864         /* stop adapter */
2865         ixgbe_stop_adapter(hw);
2866
2867         for (vf = 0; vfinfo != NULL && vf < pci_dev->max_vfs; vf++)
2868                 vfinfo[vf].clear_to_send = false;
2869
2870         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2871                 /* Turn off the copper */
2872                 ixgbe_set_phy_power(hw, false);
2873         } else {
2874                 /* Turn off the laser */
2875                 ixgbe_disable_tx_laser(hw);
2876         }
2877
2878         ixgbe_dev_clear_queues(dev);
2879
2880         /* Clear stored conf */
2881         dev->data->scattered_rx = 0;
2882         dev->data->lro = 0;
2883
2884         /* Clear recorded link status */
2885         memset(&link, 0, sizeof(link));
2886         rte_eth_linkstatus_set(dev, &link);
2887
2888         if (!rte_intr_allow_others(intr_handle))
2889                 /* resume to the default handler */
2890                 rte_intr_callback_register(intr_handle,
2891                                            ixgbe_dev_interrupt_handler,
2892                                            (void *)dev);
2893
2894         /* Clean datapath event and queue/vec mapping */
2895         rte_intr_efd_disable(intr_handle);
2896         if (intr_handle->intr_vec != NULL) {
2897                 rte_free(intr_handle->intr_vec);
2898                 intr_handle->intr_vec = NULL;
2899         }
2900
2901         /* reset hierarchy commit */
2902         tm_conf->committed = false;
2903
2904         adapter->rss_reta_updated = 0;
2905
2906         hw->adapter_stopped = true;
2907         dev->data->dev_started = 0;
2908
2909         return 0;
2910 }
2911
2912 /*
2913  * Set device link up: enable tx.
2914  */
2915 static int
2916 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2917 {
2918         struct ixgbe_hw *hw =
2919                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2920         if (hw->mac.type == ixgbe_mac_82599EB) {
2921 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2922                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2923                         /* Not suported in bypass mode */
2924                         PMD_INIT_LOG(ERR, "Set link up is not supported "
2925                                      "by device id 0x%x", hw->device_id);
2926                         return -ENOTSUP;
2927                 }
2928 #endif
2929         }
2930
2931         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2932                 /* Turn on the copper */
2933                 ixgbe_set_phy_power(hw, true);
2934         } else {
2935                 /* Turn on the laser */
2936                 ixgbe_enable_tx_laser(hw);
2937                 ixgbe_dev_link_update(dev, 0);
2938         }
2939
2940         return 0;
2941 }
2942
2943 /*
2944  * Set device link down: disable tx.
2945  */
2946 static int
2947 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2948 {
2949         struct ixgbe_hw *hw =
2950                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2951         if (hw->mac.type == ixgbe_mac_82599EB) {
2952 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2953                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2954                         /* Not suported in bypass mode */
2955                         PMD_INIT_LOG(ERR, "Set link down is not supported "
2956                                      "by device id 0x%x", hw->device_id);
2957                         return -ENOTSUP;
2958                 }
2959 #endif
2960         }
2961
2962         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2963                 /* Turn off the copper */
2964                 ixgbe_set_phy_power(hw, false);
2965         } else {
2966                 /* Turn off the laser */
2967                 ixgbe_disable_tx_laser(hw);
2968                 ixgbe_dev_link_update(dev, 0);
2969         }
2970
2971         return 0;
2972 }
2973
2974 /*
2975  * Reset and stop device.
2976  */
2977 static int
2978 ixgbe_dev_close(struct rte_eth_dev *dev)
2979 {
2980         struct ixgbe_hw *hw =
2981                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2982         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2983         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2984         int retries = 0;
2985         int ret;
2986
2987         PMD_INIT_FUNC_TRACE();
2988         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2989                 return 0;
2990
2991         ixgbe_pf_reset_hw(hw);
2992
2993         ret = ixgbe_dev_stop(dev);
2994
2995         ixgbe_dev_free_queues(dev);
2996
2997         ixgbe_disable_pcie_master(hw);
2998
2999         /* reprogram the RAR[0] in case user changed it. */
3000         ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
3001
3002         /* Unlock any pending hardware semaphore */
3003         ixgbe_swfw_lock_reset(hw);
3004
3005         /* disable uio intr before callback unregister */
3006         rte_intr_disable(intr_handle);
3007
3008         do {
3009                 ret = rte_intr_callback_unregister(intr_handle,
3010                                 ixgbe_dev_interrupt_handler, dev);
3011                 if (ret >= 0 || ret == -ENOENT) {
3012                         break;
3013                 } else if (ret != -EAGAIN) {
3014                         PMD_INIT_LOG(ERR,
3015                                 "intr callback unregister failed: %d",
3016                                 ret);
3017                 }
3018                 rte_delay_ms(100);
3019         } while (retries++ < (10 + IXGBE_LINK_UP_TIME));
3020
3021         /* cancel the delay handler before remove dev */
3022         rte_eal_alarm_cancel(ixgbe_dev_interrupt_delayed_handler, dev);
3023
3024         /* uninitialize PF if max_vfs not zero */
3025         ixgbe_pf_host_uninit(dev);
3026
3027         /* remove all the fdir filters & hash */
3028         ixgbe_fdir_filter_uninit(dev);
3029
3030         /* remove all the L2 tunnel filters & hash */
3031         ixgbe_l2_tn_filter_uninit(dev);
3032
3033         /* Remove all ntuple filters of the device */
3034         ixgbe_ntuple_filter_uninit(dev);
3035
3036         /* clear all the filters list */
3037         ixgbe_filterlist_flush();
3038
3039         /* Remove all Traffic Manager configuration */
3040         ixgbe_tm_conf_uninit(dev);
3041
3042 #ifdef RTE_LIB_SECURITY
3043         rte_free(dev->security_ctx);
3044 #endif
3045
3046         return ret;
3047 }
3048
3049 /*
3050  * Reset PF device.
3051  */
3052 static int
3053 ixgbe_dev_reset(struct rte_eth_dev *dev)
3054 {
3055         int ret;
3056
3057         /* When a DPDK PMD PF begin to reset PF port, it should notify all
3058          * its VF to make them align with it. The detailed notification
3059          * mechanism is PMD specific. As to ixgbe PF, it is rather complex.
3060          * To avoid unexpected behavior in VF, currently reset of PF with
3061          * SR-IOV activation is not supported. It might be supported later.
3062          */
3063         if (dev->data->sriov.active)
3064                 return -ENOTSUP;
3065
3066         ret = eth_ixgbe_dev_uninit(dev);
3067         if (ret)
3068                 return ret;
3069
3070         ret = eth_ixgbe_dev_init(dev, NULL);
3071
3072         return ret;
3073 }
3074
3075 static void
3076 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
3077                            struct ixgbe_hw_stats *hw_stats,
3078                            struct ixgbe_macsec_stats *macsec_stats,
3079                            uint64_t *total_missed_rx, uint64_t *total_qbrc,
3080                            uint64_t *total_qprc, uint64_t *total_qprdc)
3081 {
3082         uint32_t bprc, lxon, lxoff, total;
3083         uint32_t delta_gprc = 0;
3084         unsigned i;
3085         /* Workaround for RX byte count not including CRC bytes when CRC
3086          * strip is enabled. CRC bytes are removed from counters when crc_strip
3087          * is disabled.
3088          */
3089         int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
3090                         IXGBE_HLREG0_RXCRCSTRP);
3091
3092         hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
3093         hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
3094         hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
3095         hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
3096
3097         for (i = 0; i < 8; i++) {
3098                 uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
3099
3100                 /* global total per queue */
3101                 hw_stats->mpc[i] += mp;
3102                 /* Running comprehensive total for stats display */
3103                 *total_missed_rx += hw_stats->mpc[i];
3104                 if (hw->mac.type == ixgbe_mac_82598EB) {
3105                         hw_stats->rnbc[i] +=
3106                             IXGBE_READ_REG(hw, IXGBE_RNBC(i));
3107                         hw_stats->pxonrxc[i] +=
3108                                 IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
3109                         hw_stats->pxoffrxc[i] +=
3110                                 IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
3111                 } else {
3112                         hw_stats->pxonrxc[i] +=
3113                                 IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
3114                         hw_stats->pxoffrxc[i] +=
3115                                 IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
3116                         hw_stats->pxon2offc[i] +=
3117                                 IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
3118                 }
3119                 hw_stats->pxontxc[i] +=
3120                     IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
3121                 hw_stats->pxofftxc[i] +=
3122                     IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
3123         }
3124         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3125                 uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
3126                 uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
3127                 uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
3128
3129                 delta_gprc += delta_qprc;
3130
3131                 hw_stats->qprc[i] += delta_qprc;
3132                 hw_stats->qptc[i] += delta_qptc;
3133
3134                 hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
3135                 hw_stats->qbrc[i] +=
3136                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
3137                 if (crc_strip == 0)
3138                         hw_stats->qbrc[i] -= delta_qprc * RTE_ETHER_CRC_LEN;
3139
3140                 hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
3141                 hw_stats->qbtc[i] +=
3142                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
3143
3144                 hw_stats->qprdc[i] += delta_qprdc;
3145                 *total_qprdc += hw_stats->qprdc[i];
3146
3147                 *total_qprc += hw_stats->qprc[i];
3148                 *total_qbrc += hw_stats->qbrc[i];
3149         }
3150         hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
3151         hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
3152         hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
3153
3154         /*
3155          * An errata states that gprc actually counts good + missed packets:
3156          * Workaround to set gprc to summated queue packet receives
3157          */
3158         hw_stats->gprc = *total_qprc;
3159
3160         if (hw->mac.type != ixgbe_mac_82598EB) {
3161                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
3162                 hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
3163                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
3164                 hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
3165                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
3166                 hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
3167                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
3168                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
3169         } else {
3170                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
3171                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
3172                 /* 82598 only has a counter in the high register */
3173                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
3174                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
3175                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
3176         }
3177         uint64_t old_tpr = hw_stats->tpr;
3178
3179         hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
3180         hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
3181
3182         if (crc_strip == 0)
3183                 hw_stats->gorc -= delta_gprc * RTE_ETHER_CRC_LEN;
3184
3185         uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
3186         hw_stats->gptc += delta_gptc;
3187         hw_stats->gotc -= delta_gptc * RTE_ETHER_CRC_LEN;
3188         hw_stats->tor -= (hw_stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
3189
3190         /*
3191          * Workaround: mprc hardware is incorrectly counting
3192          * broadcasts, so for now we subtract those.
3193          */
3194         bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
3195         hw_stats->bprc += bprc;
3196         hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
3197         if (hw->mac.type == ixgbe_mac_82598EB)
3198                 hw_stats->mprc -= bprc;
3199
3200         hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
3201         hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
3202         hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
3203         hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
3204         hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
3205         hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
3206
3207         lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
3208         hw_stats->lxontxc += lxon;
3209         lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
3210         hw_stats->lxofftxc += lxoff;
3211         total = lxon + lxoff;
3212
3213         hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
3214         hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
3215         hw_stats->gptc -= total;
3216         hw_stats->mptc -= total;
3217         hw_stats->ptc64 -= total;
3218         hw_stats->gotc -= total * RTE_ETHER_MIN_LEN;
3219
3220         hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3221         hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3222         hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3223         hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
3224         hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
3225         hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
3226         hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
3227         hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3228         hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3229         hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3230         hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3231         hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
3232         hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3233         hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
3234         hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
3235         hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
3236         /* Only read FCOE on 82599 */
3237         if (hw->mac.type != ixgbe_mac_82598EB) {
3238                 hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
3239                 hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
3240                 hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
3241                 hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
3242                 hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
3243         }
3244
3245         /* Flow Director Stats registers */
3246         if (hw->mac.type != ixgbe_mac_82598EB) {
3247                 hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
3248                 hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
3249                 hw_stats->fdirustat_add += IXGBE_READ_REG(hw,
3250                                         IXGBE_FDIRUSTAT) & 0xFFFF;
3251                 hw_stats->fdirustat_remove += (IXGBE_READ_REG(hw,
3252                                         IXGBE_FDIRUSTAT) >> 16) & 0xFFFF;
3253                 hw_stats->fdirfstat_fadd += IXGBE_READ_REG(hw,
3254                                         IXGBE_FDIRFSTAT) & 0xFFFF;
3255                 hw_stats->fdirfstat_fremove += (IXGBE_READ_REG(hw,
3256                                         IXGBE_FDIRFSTAT) >> 16) & 0xFFFF;
3257         }
3258         /* MACsec Stats registers */
3259         macsec_stats->out_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECTXUT);
3260         macsec_stats->out_pkts_encrypted +=
3261                 IXGBE_READ_REG(hw, IXGBE_LSECTXPKTE);
3262         macsec_stats->out_pkts_protected +=
3263                 IXGBE_READ_REG(hw, IXGBE_LSECTXPKTP);
3264         macsec_stats->out_octets_encrypted +=
3265                 IXGBE_READ_REG(hw, IXGBE_LSECTXOCTE);
3266         macsec_stats->out_octets_protected +=
3267                 IXGBE_READ_REG(hw, IXGBE_LSECTXOCTP);
3268         macsec_stats->in_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECRXUT);
3269         macsec_stats->in_pkts_badtag += IXGBE_READ_REG(hw, IXGBE_LSECRXBAD);
3270         macsec_stats->in_pkts_nosci += IXGBE_READ_REG(hw, IXGBE_LSECRXNOSCI);
3271         macsec_stats->in_pkts_unknownsci +=
3272                 IXGBE_READ_REG(hw, IXGBE_LSECRXUNSCI);
3273         macsec_stats->in_octets_decrypted +=
3274                 IXGBE_READ_REG(hw, IXGBE_LSECRXOCTD);
3275         macsec_stats->in_octets_validated +=
3276                 IXGBE_READ_REG(hw, IXGBE_LSECRXOCTV);
3277         macsec_stats->in_pkts_unchecked += IXGBE_READ_REG(hw, IXGBE_LSECRXUNCH);
3278         macsec_stats->in_pkts_delayed += IXGBE_READ_REG(hw, IXGBE_LSECRXDELAY);
3279         macsec_stats->in_pkts_late += IXGBE_READ_REG(hw, IXGBE_LSECRXLATE);
3280         for (i = 0; i < 2; i++) {
3281                 macsec_stats->in_pkts_ok +=
3282                         IXGBE_READ_REG(hw, IXGBE_LSECRXOK(i));
3283                 macsec_stats->in_pkts_invalid +=
3284                         IXGBE_READ_REG(hw, IXGBE_LSECRXINV(i));
3285                 macsec_stats->in_pkts_notvalid +=
3286                         IXGBE_READ_REG(hw, IXGBE_LSECRXNV(i));
3287         }
3288         macsec_stats->in_pkts_unusedsa += IXGBE_READ_REG(hw, IXGBE_LSECRXUNSA);
3289         macsec_stats->in_pkts_notusingsa +=
3290                 IXGBE_READ_REG(hw, IXGBE_LSECRXNUSA);
3291 }
3292
3293 /*
3294  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
3295  */
3296 static int
3297 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3298 {
3299         struct ixgbe_hw *hw =
3300                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3301         struct ixgbe_hw_stats *hw_stats =
3302                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3303         struct ixgbe_macsec_stats *macsec_stats =
3304                         IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3305                                 dev->data->dev_private);
3306         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3307         unsigned i;
3308
3309         total_missed_rx = 0;
3310         total_qbrc = 0;
3311         total_qprc = 0;
3312         total_qprdc = 0;
3313
3314         ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3315                         &total_qbrc, &total_qprc, &total_qprdc);
3316
3317         if (stats == NULL)
3318                 return -EINVAL;
3319
3320         /* Fill out the rte_eth_stats statistics structure */
3321         stats->ipackets = total_qprc;
3322         stats->ibytes = total_qbrc;
3323         stats->opackets = hw_stats->gptc;
3324         stats->obytes = hw_stats->gotc;
3325
3326         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3327                 stats->q_ipackets[i] = hw_stats->qprc[i];
3328                 stats->q_opackets[i] = hw_stats->qptc[i];
3329                 stats->q_ibytes[i] = hw_stats->qbrc[i];
3330                 stats->q_obytes[i] = hw_stats->qbtc[i];
3331                 stats->q_errors[i] = hw_stats->qprdc[i];
3332         }
3333
3334         /* Rx Errors */
3335         stats->imissed  = total_missed_rx;
3336         stats->ierrors  = hw_stats->crcerrs +
3337                           hw_stats->mspdc +
3338                           hw_stats->rlec +
3339                           hw_stats->ruc +
3340                           hw_stats->roc +
3341                           hw_stats->illerrc +
3342                           hw_stats->errbc +
3343                           hw_stats->rfc +
3344                           hw_stats->fccrc +
3345                           hw_stats->fclast;
3346
3347         /*
3348          * 82599 errata, UDP frames with a 0 checksum can be marked as checksum
3349          * errors.
3350          */
3351         if (hw->mac.type != ixgbe_mac_82599EB)
3352                 stats->ierrors += hw_stats->xec;
3353
3354         /* Tx Errors */
3355         stats->oerrors  = 0;
3356         return 0;
3357 }
3358
3359 static int
3360 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
3361 {
3362         struct ixgbe_hw_stats *stats =
3363                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3364
3365         /* HW registers are cleared on read */
3366         ixgbe_dev_stats_get(dev, NULL);
3367
3368         /* Reset software totals */
3369         memset(stats, 0, sizeof(*stats));
3370
3371         return 0;
3372 }
3373
3374 /* This function calculates the number of xstats based on the current config */
3375 static unsigned
3376 ixgbe_xstats_calc_num(void) {
3377         return IXGBE_NB_HW_STATS + IXGBE_NB_MACSEC_STATS +
3378                 (IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
3379                 (IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
3380 }
3381
3382 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3383         struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned int size)
3384 {
3385         const unsigned cnt_stats = ixgbe_xstats_calc_num();
3386         unsigned stat, i, count;
3387
3388         if (xstats_names != NULL) {
3389                 count = 0;
3390
3391                 /* Note: limit >= cnt_stats checked upstream
3392                  * in rte_eth_xstats_names()
3393                  */
3394
3395                 /* Extended stats from ixgbe_hw_stats */
3396                 for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3397                         strlcpy(xstats_names[count].name,
3398                                 rte_ixgbe_stats_strings[i].name,
3399                                 sizeof(xstats_names[count].name));
3400                         count++;
3401                 }
3402
3403                 /* MACsec Stats */
3404                 for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3405                         strlcpy(xstats_names[count].name,
3406                                 rte_ixgbe_macsec_strings[i].name,
3407                                 sizeof(xstats_names[count].name));
3408                         count++;
3409                 }
3410
3411                 /* RX Priority Stats */
3412                 for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3413                         for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3414                                 snprintf(xstats_names[count].name,
3415                                         sizeof(xstats_names[count].name),
3416                                         "rx_priority%u_%s", i,
3417                                         rte_ixgbe_rxq_strings[stat].name);
3418                                 count++;
3419                         }
3420                 }
3421
3422                 /* TX Priority Stats */
3423                 for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3424                         for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3425                                 snprintf(xstats_names[count].name,
3426                                         sizeof(xstats_names[count].name),
3427                                         "tx_priority%u_%s", i,
3428                                         rte_ixgbe_txq_strings[stat].name);
3429                                 count++;
3430                         }
3431                 }
3432         }
3433         return cnt_stats;
3434 }
3435
3436 static int ixgbe_dev_xstats_get_names_by_id(
3437         struct rte_eth_dev *dev,
3438         struct rte_eth_xstat_name *xstats_names,
3439         const uint64_t *ids,
3440         unsigned int limit)
3441 {
3442         if (!ids) {
3443                 const unsigned int cnt_stats = ixgbe_xstats_calc_num();
3444                 unsigned int stat, i, count;
3445
3446                 if (xstats_names != NULL) {
3447                         count = 0;
3448
3449                         /* Note: limit >= cnt_stats checked upstream
3450                          * in rte_eth_xstats_names()
3451                          */
3452
3453                         /* Extended stats from ixgbe_hw_stats */
3454                         for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3455                                 strlcpy(xstats_names[count].name,
3456                                         rte_ixgbe_stats_strings[i].name,
3457                                         sizeof(xstats_names[count].name));
3458                                 count++;
3459                         }
3460
3461                         /* MACsec Stats */
3462                         for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3463                                 strlcpy(xstats_names[count].name,
3464                                         rte_ixgbe_macsec_strings[i].name,
3465                                         sizeof(xstats_names[count].name));
3466                                 count++;
3467                         }
3468
3469                         /* RX Priority Stats */
3470                         for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3471                                 for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3472                                         snprintf(xstats_names[count].name,
3473                                             sizeof(xstats_names[count].name),
3474                                             "rx_priority%u_%s", i,
3475                                             rte_ixgbe_rxq_strings[stat].name);
3476                                         count++;
3477                                 }
3478                         }
3479
3480                         /* TX Priority Stats */
3481                         for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3482                                 for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3483                                         snprintf(xstats_names[count].name,
3484                                             sizeof(xstats_names[count].name),
3485                                             "tx_priority%u_%s", i,
3486                                             rte_ixgbe_txq_strings[stat].name);
3487                                         count++;
3488                                 }
3489                         }
3490                 }
3491                 return cnt_stats;
3492         }
3493
3494         uint16_t i;
3495         uint16_t size = ixgbe_xstats_calc_num();
3496         struct rte_eth_xstat_name xstats_names_copy[size];
3497
3498         ixgbe_dev_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
3499                         size);
3500
3501         for (i = 0; i < limit; i++) {
3502                 if (ids[i] >= size) {
3503                         PMD_INIT_LOG(ERR, "id value isn't valid");
3504                         return -1;
3505                 }
3506                 strcpy(xstats_names[i].name,
3507                                 xstats_names_copy[ids[i]].name);
3508         }
3509         return limit;
3510 }
3511
3512 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3513         struct rte_eth_xstat_name *xstats_names, unsigned limit)
3514 {
3515         unsigned i;
3516
3517         if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
3518                 return -ENOMEM;
3519
3520         if (xstats_names != NULL)
3521                 for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
3522                         strlcpy(xstats_names[i].name,
3523                                 rte_ixgbevf_stats_strings[i].name,
3524                                 sizeof(xstats_names[i].name));
3525         return IXGBEVF_NB_XSTATS;
3526 }
3527
3528 static int
3529 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3530                                          unsigned n)
3531 {
3532         struct ixgbe_hw *hw =
3533                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3534         struct ixgbe_hw_stats *hw_stats =
3535                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3536         struct ixgbe_macsec_stats *macsec_stats =
3537                         IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3538                                 dev->data->dev_private);
3539         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3540         unsigned i, stat, count = 0;
3541
3542         count = ixgbe_xstats_calc_num();
3543
3544         if (n < count)
3545                 return count;
3546
3547         total_missed_rx = 0;
3548         total_qbrc = 0;
3549         total_qprc = 0;
3550         total_qprdc = 0;
3551
3552         ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3553                         &total_qbrc, &total_qprc, &total_qprdc);
3554
3555         /* If this is a reset xstats is NULL, and we have cleared the
3556          * registers by reading them.
3557          */
3558         if (!xstats)
3559                 return 0;
3560
3561         /* Extended stats from ixgbe_hw_stats */
3562         count = 0;
3563         for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3564                 xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3565                                 rte_ixgbe_stats_strings[i].offset);
3566                 xstats[count].id = count;
3567                 count++;
3568         }
3569
3570         /* MACsec Stats */
3571         for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3572                 xstats[count].value = *(uint64_t *)(((char *)macsec_stats) +
3573                                 rte_ixgbe_macsec_strings[i].offset);
3574                 xstats[count].id = count;
3575                 count++;
3576         }
3577
3578         /* RX Priority Stats */
3579         for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3580                 for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3581                         xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3582                                         rte_ixgbe_rxq_strings[stat].offset +
3583                                         (sizeof(uint64_t) * i));
3584                         xstats[count].id = count;
3585                         count++;
3586                 }
3587         }
3588
3589         /* TX Priority Stats */
3590         for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3591                 for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3592                         xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3593                                         rte_ixgbe_txq_strings[stat].offset +
3594                                         (sizeof(uint64_t) * i));
3595                         xstats[count].id = count;
3596                         count++;
3597                 }
3598         }
3599         return count;
3600 }
3601
3602 static int
3603 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
3604                 uint64_t *values, unsigned int n)
3605 {
3606         if (!ids) {
3607                 struct ixgbe_hw *hw =
3608                                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3609                 struct ixgbe_hw_stats *hw_stats =
3610                                 IXGBE_DEV_PRIVATE_TO_STATS(
3611                                                 dev->data->dev_private);
3612                 struct ixgbe_macsec_stats *macsec_stats =
3613                                 IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3614                                         dev->data->dev_private);
3615                 uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3616                 unsigned int i, stat, count = 0;
3617
3618                 count = ixgbe_xstats_calc_num();
3619
3620                 if (!ids && n < count)
3621                         return count;
3622
3623                 total_missed_rx = 0;
3624                 total_qbrc = 0;
3625                 total_qprc = 0;
3626                 total_qprdc = 0;
3627
3628                 ixgbe_read_stats_registers(hw, hw_stats, macsec_stats,
3629                                 &total_missed_rx, &total_qbrc, &total_qprc,
3630                                 &total_qprdc);
3631
3632                 /* If this is a reset xstats is NULL, and we have cleared the
3633                  * registers by reading them.
3634                  */
3635                 if (!ids && !values)
3636                         return 0;
3637
3638                 /* Extended stats from ixgbe_hw_stats */
3639                 count = 0;
3640                 for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3641                         values[count] = *(uint64_t *)(((char *)hw_stats) +
3642                                         rte_ixgbe_stats_strings[i].offset);
3643                         count++;
3644                 }
3645
3646                 /* MACsec Stats */
3647                 for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3648                         values[count] = *(uint64_t *)(((char *)macsec_stats) +
3649                                         rte_ixgbe_macsec_strings[i].offset);
3650                         count++;
3651                 }
3652
3653                 /* RX Priority Stats */
3654                 for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3655                         for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3656                                 values[count] =
3657                                         *(uint64_t *)(((char *)hw_stats) +
3658                                         rte_ixgbe_rxq_strings[stat].offset +
3659                                         (sizeof(uint64_t) * i));
3660                                 count++;
3661                         }
3662                 }
3663
3664                 /* TX Priority Stats */
3665                 for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3666                         for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3667                                 values[count] =
3668                                         *(uint64_t *)(((char *)hw_stats) +
3669                                         rte_ixgbe_txq_strings[stat].offset +
3670                                         (sizeof(uint64_t) * i));
3671                                 count++;
3672                         }
3673                 }
3674                 return count;
3675         }
3676
3677         uint16_t i;
3678         uint16_t size = ixgbe_xstats_calc_num();
3679         uint64_t values_copy[size];
3680
3681         ixgbe_dev_xstats_get_by_id(dev, NULL, values_copy, size);
3682
3683         for (i = 0; i < n; i++) {
3684                 if (ids[i] >= size) {
3685                         PMD_INIT_LOG(ERR, "id value isn't valid");
3686                         return -1;
3687                 }
3688                 values[i] = values_copy[ids[i]];
3689         }
3690         return n;
3691 }
3692
3693 static int
3694 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
3695 {
3696         struct ixgbe_hw_stats *stats =
3697                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3698         struct ixgbe_macsec_stats *macsec_stats =
3699                         IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3700                                 dev->data->dev_private);
3701
3702         unsigned count = ixgbe_xstats_calc_num();
3703
3704         /* HW registers are cleared on read */
3705         ixgbe_dev_xstats_get(dev, NULL, count);
3706
3707         /* Reset software totals */
3708         memset(stats, 0, sizeof(*stats));
3709         memset(macsec_stats, 0, sizeof(*macsec_stats));
3710
3711         return 0;
3712 }
3713
3714 static void
3715 ixgbevf_update_stats(struct rte_eth_dev *dev)
3716 {
3717         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3718         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3719                           IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3720
3721         /* Good Rx packet, include VF loopback */
3722         UPDATE_VF_STAT(IXGBE_VFGPRC,
3723             hw_stats->last_vfgprc, hw_stats->vfgprc);
3724
3725         /* Good Rx octets, include VF loopback */
3726         UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
3727             hw_stats->last_vfgorc, hw_stats->vfgorc);
3728
3729         /* Good Tx packet, include VF loopback */
3730         UPDATE_VF_STAT(IXGBE_VFGPTC,
3731             hw_stats->last_vfgptc, hw_stats->vfgptc);
3732
3733         /* Good Tx octets, include VF loopback */
3734         UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
3735             hw_stats->last_vfgotc, hw_stats->vfgotc);
3736
3737         /* Rx Multicst Packet */
3738         UPDATE_VF_STAT(IXGBE_VFMPRC,
3739             hw_stats->last_vfmprc, hw_stats->vfmprc);
3740 }
3741
3742 static int
3743 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3744                        unsigned n)
3745 {
3746         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3747                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3748         unsigned i;
3749
3750         if (n < IXGBEVF_NB_XSTATS)
3751                 return IXGBEVF_NB_XSTATS;
3752
3753         ixgbevf_update_stats(dev);
3754
3755         if (!xstats)
3756                 return 0;
3757
3758         /* Extended stats */
3759         for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
3760                 xstats[i].id = i;
3761                 xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
3762                         rte_ixgbevf_stats_strings[i].offset);
3763         }
3764
3765         return IXGBEVF_NB_XSTATS;
3766 }
3767
3768 static int
3769 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3770 {
3771         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3772                           IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3773
3774         ixgbevf_update_stats(dev);
3775
3776         if (stats == NULL)
3777                 return -EINVAL;
3778
3779         stats->ipackets = hw_stats->vfgprc;
3780         stats->ibytes = hw_stats->vfgorc;
3781         stats->opackets = hw_stats->vfgptc;
3782         stats->obytes = hw_stats->vfgotc;
3783         return 0;
3784 }
3785
3786 static int
3787 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3788 {
3789         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3790                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3791
3792         /* Sync HW register to the last stats */
3793         ixgbevf_dev_stats_get(dev, NULL);
3794
3795         /* reset HW current stats*/
3796         hw_stats->vfgprc = 0;
3797         hw_stats->vfgorc = 0;
3798         hw_stats->vfgptc = 0;
3799         hw_stats->vfgotc = 0;
3800
3801         return 0;
3802 }
3803
3804 static int
3805 ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
3806 {
3807         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3808         u16 eeprom_verh, eeprom_verl;
3809         u32 etrack_id;
3810         int ret;
3811
3812         ixgbe_read_eeprom(hw, 0x2e, &eeprom_verh);
3813         ixgbe_read_eeprom(hw, 0x2d, &eeprom_verl);
3814
3815         etrack_id = (eeprom_verh << 16) | eeprom_verl;
3816         ret = snprintf(fw_version, fw_size, "0x%08x", etrack_id);
3817         if (ret < 0)
3818                 return -EINVAL;
3819
3820         ret += 1; /* add the size of '\0' */
3821         if (fw_size < (size_t)ret)
3822                 return ret;
3823         else
3824                 return 0;
3825 }
3826
3827 static int
3828 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3829 {
3830         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3831         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3832         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3833
3834         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3835         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3836         if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3837                 /*
3838                  * When DCB/VT is off, maximum number of queues changes,
3839                  * except for 82598EB, which remains constant.
3840                  */
3841                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
3842                                 hw->mac.type != ixgbe_mac_82598EB)
3843                         dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3844         }
3845         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3846         dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3847         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3848         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3849         dev_info->max_vfs = pci_dev->max_vfs;
3850         if (hw->mac.type == ixgbe_mac_82598EB)
3851                 dev_info->max_vmdq_pools = ETH_16_POOLS;
3852         else
3853                 dev_info->max_vmdq_pools = ETH_64_POOLS;
3854         dev_info->max_mtu =  dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3855         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
3856         dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3857         dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3858         dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3859                                      dev_info->rx_queue_offload_capa);
3860         dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3861         dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3862
3863         dev_info->default_rxconf = (struct rte_eth_rxconf) {
3864                 .rx_thresh = {
3865                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3866                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3867                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3868                 },
3869                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3870                 .rx_drop_en = 0,
3871                 .offloads = 0,
3872         };
3873
3874         dev_info->default_txconf = (struct rte_eth_txconf) {
3875                 .tx_thresh = {
3876                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3877                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3878                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3879                 },
3880                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3881                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3882                 .offloads = 0,
3883         };
3884
3885         dev_info->rx_desc_lim = rx_desc_lim;
3886         dev_info->tx_desc_lim = tx_desc_lim;
3887
3888         dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3889         dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3890         dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3891
3892         dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G;
3893         if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
3894                         hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
3895                 dev_info->speed_capa = ETH_LINK_SPEED_10M |
3896                         ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G;
3897
3898         if (hw->mac.type == ixgbe_mac_X540 ||
3899             hw->mac.type == ixgbe_mac_X540_vf ||
3900             hw->mac.type == ixgbe_mac_X550 ||
3901             hw->mac.type == ixgbe_mac_X550_vf) {
3902                 dev_info->speed_capa |= ETH_LINK_SPEED_100M;
3903         }
3904         if (hw->mac.type == ixgbe_mac_X550) {
3905                 dev_info->speed_capa |= ETH_LINK_SPEED_2_5G;
3906                 dev_info->speed_capa |= ETH_LINK_SPEED_5G;
3907         }
3908
3909         /* Driver-preferred Rx/Tx parameters */
3910         dev_info->default_rxportconf.burst_size = 32;
3911         dev_info->default_txportconf.burst_size = 32;
3912         dev_info->default_rxportconf.nb_queues = 1;
3913         dev_info->default_txportconf.nb_queues = 1;
3914         dev_info->default_rxportconf.ring_size = 256;
3915         dev_info->default_txportconf.ring_size = 256;
3916
3917         return 0;
3918 }
3919
3920 static const uint32_t *
3921 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3922 {
3923         static const uint32_t ptypes[] = {
3924                 /* For non-vec functions,
3925                  * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3926                  * for vec functions,
3927                  * refers to _recv_raw_pkts_vec().
3928                  */
3929                 RTE_PTYPE_L2_ETHER,
3930                 RTE_PTYPE_L3_IPV4,
3931                 RTE_PTYPE_L3_IPV4_EXT,
3932                 RTE_PTYPE_L3_IPV6,
3933                 RTE_PTYPE_L3_IPV6_EXT,
3934                 RTE_PTYPE_L4_SCTP,
3935                 RTE_PTYPE_L4_TCP,
3936                 RTE_PTYPE_L4_UDP,
3937                 RTE_PTYPE_TUNNEL_IP,
3938                 RTE_PTYPE_INNER_L3_IPV6,
3939                 RTE_PTYPE_INNER_L3_IPV6_EXT,
3940                 RTE_PTYPE_INNER_L4_TCP,
3941                 RTE_PTYPE_INNER_L4_UDP,
3942                 RTE_PTYPE_UNKNOWN
3943         };
3944
3945         if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3946             dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3947             dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3948             dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3949                 return ptypes;
3950
3951 #if defined(RTE_ARCH_X86) || defined(__ARM_NEON)
3952         if (dev->rx_pkt_burst == ixgbe_recv_pkts_vec ||
3953             dev->rx_pkt_burst == ixgbe_recv_scattered_pkts_vec)
3954                 return ptypes;
3955 #endif
3956         return NULL;
3957 }
3958
3959 static int
3960 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3961                      struct rte_eth_dev_info *dev_info)
3962 {
3963         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3964         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3965
3966         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3967         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3968         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3969         dev_info->max_rx_pktlen = 9728; /* includes CRC, cf MAXFRS reg */
3970         dev_info->max_mtu = dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3971         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3972         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3973         dev_info->max_vfs = pci_dev->max_vfs;
3974         if (hw->mac.type == ixgbe_mac_82598EB)
3975                 dev_info->max_vmdq_pools = ETH_16_POOLS;
3976         else
3977                 dev_info->max_vmdq_pools = ETH_64_POOLS;
3978         dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3979         dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3980                                      dev_info->rx_queue_offload_capa);
3981         dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3982         dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3983         dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3984         dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3985         dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3986
3987         dev_info->default_rxconf = (struct rte_eth_rxconf) {
3988                 .rx_thresh = {
3989                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3990                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3991                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3992                 },
3993                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3994                 .rx_drop_en = 0,
3995                 .offloads = 0,
3996         };
3997
3998         dev_info->default_txconf = (struct rte_eth_txconf) {
3999                 .tx_thresh = {
4000                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
4001                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
4002                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
4003                 },
4004                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
4005                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
4006                 .offloads = 0,
4007         };
4008
4009         dev_info->rx_desc_lim = rx_desc_lim;
4010         dev_info->tx_desc_lim = tx_desc_lim;
4011
4012         return 0;
4013 }
4014
4015 static int
4016 ixgbevf_check_link(struct ixgbe_hw *hw, ixgbe_link_speed *speed,
4017                    bool *link_up, int wait_to_complete)
4018 {
4019         struct ixgbe_adapter *adapter = container_of(hw,
4020                                                      struct ixgbe_adapter, hw);
4021         struct ixgbe_mbx_info *mbx = &hw->mbx;
4022         struct ixgbe_mac_info *mac = &hw->mac;
4023         uint32_t links_reg, in_msg;
4024         int ret_val = 0;
4025
4026         /* If we were hit with a reset drop the link */
4027         if (!mbx->ops.check_for_rst(hw, 0) || !mbx->timeout)
4028                 mac->get_link_status = true;
4029
4030         if (!mac->get_link_status)
4031                 goto out;
4032
4033         /* if link status is down no point in checking to see if pf is up */
4034         links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4035         if (!(links_reg & IXGBE_LINKS_UP))
4036                 goto out;
4037
4038         /* for SFP+ modules and DA cables on 82599 it can take up to 500usecs
4039          * before the link status is correct
4040          */
4041         if (mac->type == ixgbe_mac_82599_vf && wait_to_complete) {
4042                 int i;
4043
4044                 for (i = 0; i < 5; i++) {
4045                         rte_delay_us(100);
4046                         links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4047
4048                         if (!(links_reg & IXGBE_LINKS_UP))
4049                                 goto out;
4050                 }
4051         }
4052
4053         switch (links_reg & IXGBE_LINKS_SPEED_82599) {
4054         case IXGBE_LINKS_SPEED_10G_82599:
4055                 *speed = IXGBE_LINK_SPEED_10GB_FULL;
4056                 if (hw->mac.type >= ixgbe_mac_X550) {
4057                         if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4058                                 *speed = IXGBE_LINK_SPEED_2_5GB_FULL;
4059                 }
4060                 break;
4061         case IXGBE_LINKS_SPEED_1G_82599:
4062                 *speed = IXGBE_LINK_SPEED_1GB_FULL;
4063                 break;
4064         case IXGBE_LINKS_SPEED_100_82599:
4065                 *speed = IXGBE_LINK_SPEED_100_FULL;
4066                 if (hw->mac.type == ixgbe_mac_X550) {
4067                         if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4068                                 *speed = IXGBE_LINK_SPEED_5GB_FULL;
4069                 }
4070                 break;
4071         case IXGBE_LINKS_SPEED_10_X550EM_A:
4072                 *speed = IXGBE_LINK_SPEED_UNKNOWN;
4073                 /* Since Reserved in older MAC's */
4074                 if (hw->mac.type >= ixgbe_mac_X550)
4075                         *speed = IXGBE_LINK_SPEED_10_FULL;
4076                 break;
4077         default:
4078                 *speed = IXGBE_LINK_SPEED_UNKNOWN;
4079         }
4080
4081         if (wait_to_complete == 0 && adapter->pflink_fullchk == 0) {
4082                 if (*speed == IXGBE_LINK_SPEED_UNKNOWN)
4083                         mac->get_link_status = true;
4084                 else
4085                         mac->get_link_status = false;
4086
4087                 goto out;
4088         }
4089
4090         /* if the read failed it could just be a mailbox collision, best wait
4091          * until we are called again and don't report an error
4092          */
4093         if (mbx->ops.read(hw, &in_msg, 1, 0))
4094                 goto out;
4095
4096         if (!(in_msg & IXGBE_VT_MSGTYPE_CTS)) {
4097                 /* msg is not CTS and is NACK we must have lost CTS status */
4098                 if (in_msg & IXGBE_VT_MSGTYPE_NACK)
4099                         mac->get_link_status = false;
4100                 goto out;
4101         }
4102
4103         /* the pf is talking, if we timed out in the past we reinit */
4104         if (!mbx->timeout) {
4105                 ret_val = -1;
4106                 goto out;
4107         }
4108
4109         /* if we passed all the tests above then the link is up and we no
4110          * longer need to check for link
4111          */
4112         mac->get_link_status = false;
4113
4114 out:
4115         *link_up = !mac->get_link_status;
4116         return ret_val;
4117 }
4118
4119 /*
4120  * If @timeout_ms was 0, it means that it will not return until link complete.
4121  * It returns 1 on complete, return 0 on timeout.
4122  */
4123 static int
4124 ixgbe_dev_wait_setup_link_complete(struct rte_eth_dev *dev, uint32_t timeout_ms)
4125 {
4126 #define WARNING_TIMEOUT    9000 /* 9s  in total */
4127         struct ixgbe_adapter *ad = dev->data->dev_private;
4128         uint32_t timeout = timeout_ms ? timeout_ms : WARNING_TIMEOUT;
4129
4130         while (rte_atomic32_read(&ad->link_thread_running)) {
4131                 msec_delay(1);
4132                 timeout--;
4133
4134                 if (timeout_ms) {
4135                         if (!timeout)
4136                                 return 0;
4137                 } else if (!timeout) {
4138                         /* It will not return until link complete */
4139                         timeout = WARNING_TIMEOUT;
4140                         PMD_DRV_LOG(ERR, "IXGBE link thread not complete too long time!");
4141                 }
4142         }
4143
4144         return 1;
4145 }
4146
4147 static void *
4148 ixgbe_dev_setup_link_thread_handler(void *param)
4149 {
4150         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4151         struct ixgbe_adapter *ad = dev->data->dev_private;
4152         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4153         struct ixgbe_interrupt *intr =
4154                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4155         u32 speed;
4156         bool autoneg = false;
4157
4158         pthread_detach(pthread_self());
4159         speed = hw->phy.autoneg_advertised;
4160         if (!speed)
4161                 ixgbe_get_link_capabilities(hw, &speed, &autoneg);
4162
4163         ixgbe_setup_link(hw, speed, true);
4164
4165         intr->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4166         rte_atomic32_clear(&ad->link_thread_running);
4167         return NULL;
4168 }
4169
4170 /*
4171  * In freebsd environment, nic_uio drivers do not support interrupts,
4172  * rte_intr_callback_register() will fail to register interrupts.
4173  * We can not make link status to change from down to up by interrupt
4174  * callback. So we need to wait for the controller to acquire link
4175  * when ports start.
4176  * It returns 0 on link up.
4177  */
4178 static int
4179 ixgbe_wait_for_link_up(struct ixgbe_hw *hw)
4180 {
4181 #ifdef RTE_EXEC_ENV_FREEBSD
4182         int err, i;
4183         bool link_up = false;
4184         uint32_t speed = 0;
4185         const int nb_iter = 25;
4186
4187         for (i = 0; i < nb_iter; i++) {
4188                 err = ixgbe_check_link(hw, &speed, &link_up, 0);
4189                 if (err)
4190                         return err;
4191                 if (link_up)
4192                         return 0;
4193                 msec_delay(200);
4194         }
4195
4196         return 0;
4197 #else
4198         RTE_SET_USED(hw);
4199         return 0;
4200 #endif
4201 }
4202
4203 /* return 0 means link status changed, -1 means not changed */
4204 int
4205 ixgbe_dev_link_update_share(struct rte_eth_dev *dev,
4206                             int wait_to_complete, int vf)
4207 {
4208         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4209         struct ixgbe_adapter *ad = dev->data->dev_private;
4210         struct rte_eth_link link;
4211         ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
4212         struct ixgbe_interrupt *intr =
4213                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4214         bool link_up;
4215         int diag;
4216         int wait = 1;
4217         u32 esdp_reg;
4218
4219         memset(&link, 0, sizeof(link));
4220         link.link_status = ETH_LINK_DOWN;
4221         link.link_speed = ETH_SPEED_NUM_NONE;
4222         link.link_duplex = ETH_LINK_HALF_DUPLEX;
4223         link.link_autoneg = !(dev->data->dev_conf.link_speeds &
4224                         ETH_LINK_SPEED_FIXED);
4225
4226         hw->mac.get_link_status = true;
4227
4228         if (intr->flags & IXGBE_FLAG_NEED_LINK_CONFIG)
4229                 return rte_eth_linkstatus_set(dev, &link);
4230
4231         /* check if it needs to wait to complete, if lsc interrupt is enabled */
4232         if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
4233                 wait = 0;
4234
4235 /* BSD has no interrupt mechanism, so force NIC status synchronization. */
4236 #ifdef RTE_EXEC_ENV_FREEBSD
4237         wait = 1;
4238 #endif
4239
4240         if (vf)
4241                 diag = ixgbevf_check_link(hw, &link_speed, &link_up, wait);
4242         else
4243                 diag = ixgbe_check_link(hw, &link_speed, &link_up, wait);
4244
4245         if (diag != 0) {
4246                 link.link_speed = ETH_SPEED_NUM_100M;
4247                 link.link_duplex = ETH_LINK_FULL_DUPLEX;
4248                 return rte_eth_linkstatus_set(dev, &link);
4249         }
4250
4251         if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4252                 esdp_reg = IXGBE_READ_REG(hw, IXGBE_ESDP);
4253                 if ((esdp_reg & IXGBE_ESDP_SDP3))
4254                         link_up = 0;
4255         }
4256
4257         if (link_up == 0) {
4258                 if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4259                         ixgbe_dev_wait_setup_link_complete(dev, 0);
4260                         if (rte_atomic32_test_and_set(&ad->link_thread_running)) {
4261                                 /* To avoid race condition between threads, set
4262                                  * the IXGBE_FLAG_NEED_LINK_CONFIG flag only
4263                                  * when there is no link thread running.
4264                                  */
4265                                 intr->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
4266                                 if (rte_ctrl_thread_create(&ad->link_thread_tid,
4267                                         "ixgbe-link-handler",
4268                                         NULL,
4269                                         ixgbe_dev_setup_link_thread_handler,
4270                                         dev) < 0) {
4271                                         PMD_DRV_LOG(ERR,
4272                                                 "Create link thread failed!");
4273                                         rte_atomic32_clear(&ad->link_thread_running);
4274                                 }
4275                         } else {
4276                                 PMD_DRV_LOG(ERR,
4277                                         "Other link thread is running now!");
4278                         }
4279                 }
4280                 return rte_eth_linkstatus_set(dev, &link);
4281         }
4282
4283         link.link_status = ETH_LINK_UP;
4284         link.link_duplex = ETH_LINK_FULL_DUPLEX;
4285
4286         switch (link_speed) {
4287         default:
4288         case IXGBE_LINK_SPEED_UNKNOWN:
4289                 link.link_speed = ETH_SPEED_NUM_UNKNOWN;
4290                 break;
4291
4292         case IXGBE_LINK_SPEED_10_FULL:
4293                 link.link_speed = ETH_SPEED_NUM_10M;
4294                 break;
4295
4296         case IXGBE_LINK_SPEED_100_FULL:
4297                 link.link_speed = ETH_SPEED_NUM_100M;
4298                 break;
4299
4300         case IXGBE_LINK_SPEED_1GB_FULL:
4301                 link.link_speed = ETH_SPEED_NUM_1G;
4302                 break;
4303
4304         case IXGBE_LINK_SPEED_2_5GB_FULL:
4305                 link.link_speed = ETH_SPEED_NUM_2_5G;
4306                 break;
4307
4308         case IXGBE_LINK_SPEED_5GB_FULL:
4309                 link.link_speed = ETH_SPEED_NUM_5G;
4310                 break;
4311
4312         case IXGBE_LINK_SPEED_10GB_FULL:
4313                 link.link_speed = ETH_SPEED_NUM_10G;
4314                 break;
4315         }
4316
4317         return rte_eth_linkstatus_set(dev, &link);
4318 }
4319
4320 static int
4321 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4322 {
4323         return ixgbe_dev_link_update_share(dev, wait_to_complete, 0);
4324 }
4325
4326 static int
4327 ixgbevf_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4328 {
4329         return ixgbe_dev_link_update_share(dev, wait_to_complete, 1);
4330 }
4331
4332 static int
4333 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
4334 {
4335         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4336         uint32_t fctrl;
4337
4338         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4339         fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4340         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4341
4342         return 0;
4343 }
4344
4345 static int
4346 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
4347 {
4348         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4349         uint32_t fctrl;
4350
4351         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4352         fctrl &= (~IXGBE_FCTRL_UPE);
4353         if (dev->data->all_multicast == 1)
4354                 fctrl |= IXGBE_FCTRL_MPE;
4355         else
4356                 fctrl &= (~IXGBE_FCTRL_MPE);
4357         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4358
4359         return 0;
4360 }
4361
4362 static int
4363 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
4364 {
4365         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4366         uint32_t fctrl;
4367
4368         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4369         fctrl |= IXGBE_FCTRL_MPE;
4370         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4371
4372         return 0;
4373 }
4374
4375 static int
4376 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
4377 {
4378         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4379         uint32_t fctrl;
4380
4381         if (dev->data->promiscuous == 1)
4382                 return 0; /* must remain in all_multicast mode */
4383
4384         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4385         fctrl &= (~IXGBE_FCTRL_MPE);
4386         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4387
4388         return 0;
4389 }
4390
4391 /**
4392  * It clears the interrupt causes and enables the interrupt.
4393  * It will be called once only during nic initialized.
4394  *
4395  * @param dev
4396  *  Pointer to struct rte_eth_dev.
4397  * @param on
4398  *  Enable or Disable.
4399  *
4400  * @return
4401  *  - On success, zero.
4402  *  - On failure, a negative value.
4403  */
4404 static int
4405 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
4406 {
4407         struct ixgbe_interrupt *intr =
4408                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4409
4410         ixgbe_dev_link_status_print(dev);
4411         if (on)
4412                 intr->mask |= IXGBE_EICR_LSC;
4413         else
4414                 intr->mask &= ~IXGBE_EICR_LSC;
4415
4416         return 0;
4417 }
4418
4419 /**
4420  * It clears the interrupt causes and enables the interrupt.
4421  * It will be called once only during nic initialized.
4422  *
4423  * @param dev
4424  *  Pointer to struct rte_eth_dev.
4425  *
4426  * @return
4427  *  - On success, zero.
4428  *  - On failure, a negative value.
4429  */
4430 static int
4431 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
4432 {
4433         struct ixgbe_interrupt *intr =
4434                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4435
4436         intr->mask |= IXGBE_EICR_RTX_QUEUE;
4437
4438         return 0;
4439 }
4440
4441 /**
4442  * It clears the interrupt causes and enables the interrupt.
4443  * It will be called once only during nic initialized.
4444  *
4445  * @param dev
4446  *  Pointer to struct rte_eth_dev.
4447  *
4448  * @return
4449  *  - On success, zero.
4450  *  - On failure, a negative value.
4451  */
4452 static int
4453 ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev)
4454 {
4455         struct ixgbe_interrupt *intr =
4456                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4457
4458         intr->mask |= IXGBE_EICR_LINKSEC;
4459
4460         return 0;
4461 }
4462
4463 /*
4464  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
4465  *
4466  * @param dev
4467  *  Pointer to struct rte_eth_dev.
4468  *
4469  * @return
4470  *  - On success, zero.
4471  *  - On failure, a negative value.
4472  */
4473 static int
4474 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
4475 {
4476         uint32_t eicr;
4477         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4478         struct ixgbe_interrupt *intr =
4479                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4480
4481         /* clear all cause mask */
4482         ixgbe_disable_intr(hw);
4483
4484         /* read-on-clear nic registers here */
4485         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4486         PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
4487
4488         intr->flags = 0;
4489
4490         /* set flag for async link update */
4491         if (eicr & IXGBE_EICR_LSC)
4492                 intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4493
4494         if (eicr & IXGBE_EICR_MAILBOX)
4495                 intr->flags |= IXGBE_FLAG_MAILBOX;
4496
4497         if (eicr & IXGBE_EICR_LINKSEC)
4498                 intr->flags |= IXGBE_FLAG_MACSEC;
4499
4500         if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
4501             hw->phy.type == ixgbe_phy_x550em_ext_t &&
4502             (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
4503                 intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
4504
4505         return 0;
4506 }
4507
4508 /**
4509  * It gets and then prints the link status.
4510  *
4511  * @param dev
4512  *  Pointer to struct rte_eth_dev.
4513  *
4514  * @return
4515  *  - On success, zero.
4516  *  - On failure, a negative value.
4517  */
4518 static void
4519 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
4520 {
4521         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4522         struct rte_eth_link link;
4523
4524         rte_eth_linkstatus_get(dev, &link);
4525
4526         if (link.link_status) {
4527                 PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
4528                                         (int)(dev->data->port_id),
4529                                         (unsigned)link.link_speed,
4530                         link.link_duplex == ETH_LINK_FULL_DUPLEX ?
4531                                         "full-duplex" : "half-duplex");
4532         } else {
4533                 PMD_INIT_LOG(INFO, " Port %d: Link Down",
4534                                 (int)(dev->data->port_id));
4535         }
4536         PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
4537                                 pci_dev->addr.domain,
4538                                 pci_dev->addr.bus,
4539                                 pci_dev->addr.devid,
4540                                 pci_dev->addr.function);
4541 }
4542
4543 /*
4544  * It executes link_update after knowing an interrupt occurred.
4545  *
4546  * @param dev
4547  *  Pointer to struct rte_eth_dev.
4548  *
4549  * @return
4550  *  - On success, zero.
4551  *  - On failure, a negative value.
4552  */
4553 static int
4554 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
4555 {
4556         struct ixgbe_interrupt *intr =
4557                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4558         int64_t timeout;
4559         struct ixgbe_hw *hw =
4560                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4561
4562         PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
4563
4564         if (intr->flags & IXGBE_FLAG_MAILBOX) {
4565                 ixgbe_pf_mbx_process(dev);
4566                 intr->flags &= ~IXGBE_FLAG_MAILBOX;
4567         }
4568
4569         if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4570                 ixgbe_handle_lasi(hw);
4571                 intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4572         }
4573
4574         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4575                 struct rte_eth_link link;
4576
4577                 /* get the link status before link update, for predicting later */
4578                 rte_eth_linkstatus_get(dev, &link);
4579
4580                 ixgbe_dev_link_update(dev, 0);
4581
4582                 /* likely to up */
4583                 if (!link.link_status)
4584                         /* handle it 1 sec later, wait it being stable */
4585                         timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
4586                 /* likely to down */
4587                 else
4588                         /* handle it 4 sec later, wait it being stable */
4589                         timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
4590
4591                 ixgbe_dev_link_status_print(dev);
4592                 if (rte_eal_alarm_set(timeout * 1000,
4593                                       ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
4594                         PMD_DRV_LOG(ERR, "Error setting alarm");
4595                 else {
4596                         /* remember original mask */
4597                         intr->mask_original = intr->mask;
4598                         /* only disable lsc interrupt */
4599                         intr->mask &= ~IXGBE_EIMS_LSC;
4600                 }
4601         }
4602
4603         PMD_DRV_LOG(DEBUG, "enable intr immediately");
4604         ixgbe_enable_intr(dev);
4605
4606         return 0;
4607 }
4608
4609 /**
4610  * Interrupt handler which shall be registered for alarm callback for delayed
4611  * handling specific interrupt to wait for the stable nic state. As the
4612  * NIC interrupt state is not stable for ixgbe after link is just down,
4613  * it needs to wait 4 seconds to get the stable status.
4614  *
4615  * @param handle
4616  *  Pointer to interrupt handle.
4617  * @param param
4618  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4619  *
4620  * @return
4621  *  void
4622  */
4623 static void
4624 ixgbe_dev_interrupt_delayed_handler(void *param)
4625 {
4626         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4627         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4628         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
4629         struct ixgbe_interrupt *intr =
4630                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4631         struct ixgbe_hw *hw =
4632                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4633         uint32_t eicr;
4634
4635         ixgbe_disable_intr(hw);
4636
4637         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4638         if (eicr & IXGBE_EICR_MAILBOX)
4639                 ixgbe_pf_mbx_process(dev);
4640
4641         if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4642                 ixgbe_handle_lasi(hw);
4643                 intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4644         }
4645
4646         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4647                 ixgbe_dev_link_update(dev, 0);
4648                 intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4649                 ixgbe_dev_link_status_print(dev);
4650                 rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
4651         }
4652
4653         if (intr->flags & IXGBE_FLAG_MACSEC) {
4654                 rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_MACSEC, NULL);
4655                 intr->flags &= ~IXGBE_FLAG_MACSEC;
4656         }
4657
4658         /* restore original mask */
4659         intr->mask = intr->mask_original;
4660         intr->mask_original = 0;
4661
4662         PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
4663         ixgbe_enable_intr(dev);
4664         rte_intr_ack(intr_handle);
4665 }
4666
4667 /**
4668  * Interrupt handler triggered by NIC  for handling
4669  * specific interrupt.
4670  *
4671  * @param handle
4672  *  Pointer to interrupt handle.
4673  * @param param
4674  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4675  *
4676  * @return
4677  *  void
4678  */
4679 static void
4680 ixgbe_dev_interrupt_handler(void *param)
4681 {
4682         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4683
4684         ixgbe_dev_interrupt_get_status(dev);
4685         ixgbe_dev_interrupt_action(dev);
4686 }
4687
4688 static int
4689 ixgbe_dev_led_on(struct rte_eth_dev *dev)
4690 {
4691         struct ixgbe_hw *hw;
4692
4693         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4694         return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4695 }
4696
4697 static int
4698 ixgbe_dev_led_off(struct rte_eth_dev *dev)
4699 {
4700         struct ixgbe_hw *hw;
4701
4702         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4703         return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4704 }
4705
4706 static int
4707 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4708 {
4709         struct ixgbe_hw *hw;
4710         uint32_t mflcn_reg;
4711         uint32_t fccfg_reg;
4712         int rx_pause;
4713         int tx_pause;
4714
4715         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4716
4717         fc_conf->pause_time = hw->fc.pause_time;
4718         fc_conf->high_water = hw->fc.high_water[0];
4719         fc_conf->low_water = hw->fc.low_water[0];
4720         fc_conf->send_xon = hw->fc.send_xon;
4721         fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
4722
4723         /*
4724          * Return rx_pause status according to actual setting of
4725          * MFLCN register.
4726          */
4727         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4728         if (mflcn_reg & IXGBE_MFLCN_PMCF)
4729                 fc_conf->mac_ctrl_frame_fwd = 1;
4730         else
4731                 fc_conf->mac_ctrl_frame_fwd = 0;
4732
4733         if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
4734                 rx_pause = 1;
4735         else
4736                 rx_pause = 0;
4737
4738         /*
4739          * Return tx_pause status according to actual setting of
4740          * FCCFG register.
4741          */
4742         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4743         if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
4744                 tx_pause = 1;
4745         else
4746                 tx_pause = 0;
4747
4748         if (rx_pause && tx_pause)
4749                 fc_conf->mode = RTE_FC_FULL;
4750         else if (rx_pause)
4751                 fc_conf->mode = RTE_FC_RX_PAUSE;
4752         else if (tx_pause)
4753                 fc_conf->mode = RTE_FC_TX_PAUSE;
4754         else
4755                 fc_conf->mode = RTE_FC_NONE;
4756
4757         return 0;
4758 }
4759
4760 static int
4761 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4762 {
4763         struct ixgbe_hw *hw;
4764         struct ixgbe_adapter *adapter = dev->data->dev_private;
4765         int err;
4766         uint32_t rx_buf_size;
4767         uint32_t max_high_water;
4768         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4769                 ixgbe_fc_none,
4770                 ixgbe_fc_rx_pause,
4771                 ixgbe_fc_tx_pause,
4772                 ixgbe_fc_full
4773         };
4774
4775         PMD_INIT_FUNC_TRACE();
4776
4777         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4778         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
4779         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4780
4781         /*
4782          * At least reserve one Ethernet frame for watermark
4783          * high_water/low_water in kilo bytes for ixgbe
4784          */
4785         max_high_water = (rx_buf_size -
4786                         RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4787         if ((fc_conf->high_water > max_high_water) ||
4788                 (fc_conf->high_water < fc_conf->low_water)) {
4789                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4790                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4791                 return -EINVAL;
4792         }
4793
4794         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
4795         hw->fc.pause_time     = fc_conf->pause_time;
4796         hw->fc.high_water[0]  = fc_conf->high_water;
4797         hw->fc.low_water[0]   = fc_conf->low_water;
4798         hw->fc.send_xon       = fc_conf->send_xon;
4799         hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
4800         adapter->mac_ctrl_frame_fwd = fc_conf->mac_ctrl_frame_fwd;
4801
4802         err = ixgbe_flow_ctrl_enable(dev, hw);
4803         if (err < 0) {
4804                 PMD_INIT_LOG(ERR, "ixgbe_flow_ctrl_enable = 0x%x", err);
4805                 return -EIO;
4806         }
4807         return err;
4808 }
4809
4810 /**
4811  *  ixgbe_pfc_enable_generic - Enable flow control
4812  *  @hw: pointer to hardware structure
4813  *  @tc_num: traffic class number
4814  *  Enable flow control according to the current settings.
4815  */
4816 static int
4817 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
4818 {
4819         int ret_val = 0;
4820         uint32_t mflcn_reg, fccfg_reg;
4821         uint32_t reg;
4822         uint32_t fcrtl, fcrth;
4823         uint8_t i;
4824         uint8_t nb_rx_en;
4825
4826         /* Validate the water mark configuration */
4827         if (!hw->fc.pause_time) {
4828                 ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4829                 goto out;
4830         }
4831
4832         /* Low water mark of zero causes XOFF floods */
4833         if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
4834                  /* High/Low water can not be 0 */
4835                 if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_water[tc_num])) {
4836                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4837                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4838                         goto out;
4839                 }
4840
4841                 if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
4842                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4843                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4844                         goto out;
4845                 }
4846         }
4847         /* Negotiate the fc mode to use */
4848         ixgbe_fc_autoneg(hw);
4849
4850         /* Disable any previous flow control settings */
4851         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4852         mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
4853
4854         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4855         fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
4856
4857         switch (hw->fc.current_mode) {
4858         case ixgbe_fc_none:
4859                 /*
4860                  * If the count of enabled RX Priority Flow control >1,
4861                  * and the TX pause can not be disabled
4862                  */
4863                 nb_rx_en = 0;
4864                 for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4865                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4866                         if (reg & IXGBE_FCRTH_FCEN)
4867                                 nb_rx_en++;
4868                 }
4869                 if (nb_rx_en > 1)
4870                         fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4871                 break;
4872         case ixgbe_fc_rx_pause:
4873                 /*
4874                  * Rx Flow control is enabled and Tx Flow control is
4875                  * disabled by software override. Since there really
4876                  * isn't a way to advertise that we are capable of RX
4877                  * Pause ONLY, we will advertise that we support both
4878                  * symmetric and asymmetric Rx PAUSE.  Later, we will
4879                  * disable the adapter's ability to send PAUSE frames.
4880                  */
4881                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
4882                 /*
4883                  * If the count of enabled RX Priority Flow control >1,
4884                  * and the TX pause can not be disabled
4885                  */
4886                 nb_rx_en = 0;
4887                 for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4888                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4889                         if (reg & IXGBE_FCRTH_FCEN)
4890                                 nb_rx_en++;
4891                 }
4892                 if (nb_rx_en > 1)
4893                         fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4894                 break;
4895         case ixgbe_fc_tx_pause:
4896                 /*
4897                  * Tx Flow control is enabled, and Rx Flow control is
4898                  * disabled by software override.
4899                  */
4900                 fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4901                 break;
4902         case ixgbe_fc_full:
4903                 /* Flow control (both Rx and Tx) is enabled by SW override. */
4904                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
4905                 fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4906                 break;
4907         default:
4908                 PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
4909                 ret_val = IXGBE_ERR_CONFIG;
4910                 goto out;
4911         }
4912
4913         /* Set 802.3x based flow control settings. */
4914         mflcn_reg |= IXGBE_MFLCN_DPF;
4915         IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
4916         IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
4917
4918         /* Set up and enable Rx high/low water mark thresholds, enable XON. */
4919         if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
4920                 hw->fc.high_water[tc_num]) {
4921                 fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
4922                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
4923                 fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
4924         } else {
4925                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
4926                 /*
4927                  * In order to prevent Tx hangs when the internal Tx
4928                  * switch is enabled we must set the high water mark
4929                  * to the maximum FCRTH value.  This allows the Tx
4930                  * switch to function even under heavy Rx workloads.
4931                  */
4932                 fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
4933         }
4934         IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
4935
4936         /* Configure pause time (2 TCs per register) */
4937         reg = hw->fc.pause_time * 0x00010001;
4938         for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
4939                 IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
4940
4941         /* Configure flow control refresh threshold value */
4942         IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
4943
4944 out:
4945         return ret_val;
4946 }
4947
4948 static int
4949 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
4950 {
4951         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4952         int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
4953
4954         if (hw->mac.type != ixgbe_mac_82598EB) {
4955                 ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
4956         }
4957         return ret_val;
4958 }
4959
4960 static int
4961 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
4962 {
4963         int err;
4964         uint32_t rx_buf_size;
4965         uint32_t max_high_water;
4966         uint8_t tc_num;
4967         uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
4968         struct ixgbe_hw *hw =
4969                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4970         struct ixgbe_dcb_config *dcb_config =
4971                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
4972
4973         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4974                 ixgbe_fc_none,
4975                 ixgbe_fc_rx_pause,
4976                 ixgbe_fc_tx_pause,
4977                 ixgbe_fc_full
4978         };
4979
4980         PMD_INIT_FUNC_TRACE();
4981
4982         ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
4983         tc_num = map[pfc_conf->priority];
4984         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
4985         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4986         /*
4987          * At least reserve one Ethernet frame for watermark
4988          * high_water/low_water in kilo bytes for ixgbe
4989          */
4990         max_high_water = (rx_buf_size -
4991                         RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4992         if ((pfc_conf->fc.high_water > max_high_water) ||
4993             (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
4994                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4995                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4996                 return -EINVAL;
4997         }
4998
4999         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
5000         hw->fc.pause_time = pfc_conf->fc.pause_time;
5001         hw->fc.send_xon = pfc_conf->fc.send_xon;
5002         hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
5003         hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
5004
5005         err = ixgbe_dcb_pfc_enable(dev, tc_num);
5006
5007         /* Not negotiated is not an error case */
5008         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
5009                 return 0;
5010
5011         PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
5012         return -EIO;
5013 }
5014
5015 static int
5016 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
5017                           struct rte_eth_rss_reta_entry64 *reta_conf,
5018                           uint16_t reta_size)
5019 {
5020         uint16_t i, sp_reta_size;
5021         uint8_t j, mask;
5022         uint32_t reta, r;
5023         uint16_t idx, shift;
5024         struct ixgbe_adapter *adapter = dev->data->dev_private;
5025         struct rte_eth_dev_data *dev_data = dev->data;
5026         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5027         uint32_t reta_reg;
5028
5029         PMD_INIT_FUNC_TRACE();
5030
5031         if (!dev_data->dev_started) {
5032                 PMD_DRV_LOG(ERR,
5033                         "port %d must be started before rss reta update",
5034                          dev_data->port_id);
5035                 return -EIO;
5036         }
5037
5038         if (!ixgbe_rss_update_sp(hw->mac.type)) {
5039                 PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
5040                         "NIC.");
5041                 return -ENOTSUP;
5042         }
5043
5044         sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5045         if (reta_size != sp_reta_size) {
5046                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5047                         "(%d) doesn't match the number hardware can supported "
5048                         "(%d)", reta_size, sp_reta_size);
5049                 return -EINVAL;
5050         }
5051
5052         for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5053                 idx = i / RTE_RETA_GROUP_SIZE;
5054                 shift = i % RTE_RETA_GROUP_SIZE;
5055                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5056                                                 IXGBE_4_BIT_MASK);
5057                 if (!mask)
5058                         continue;
5059                 reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5060                 if (mask == IXGBE_4_BIT_MASK)
5061                         r = 0;
5062                 else
5063                         r = IXGBE_READ_REG(hw, reta_reg);
5064                 for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5065                         if (mask & (0x1 << j))
5066                                 reta |= reta_conf[idx].reta[shift + j] <<
5067                                                         (CHAR_BIT * j);
5068                         else
5069                                 reta |= r & (IXGBE_8_BIT_MASK <<
5070                                                 (CHAR_BIT * j));
5071                 }
5072                 IXGBE_WRITE_REG(hw, reta_reg, reta);
5073         }
5074         adapter->rss_reta_updated = 1;
5075
5076         return 0;
5077 }
5078
5079 static int
5080 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
5081                          struct rte_eth_rss_reta_entry64 *reta_conf,
5082                          uint16_t reta_size)
5083 {
5084         uint16_t i, sp_reta_size;
5085         uint8_t j, mask;
5086         uint32_t reta;
5087         uint16_t idx, shift;
5088         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5089         uint32_t reta_reg;
5090
5091         PMD_INIT_FUNC_TRACE();
5092         sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5093         if (reta_size != sp_reta_size) {
5094                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5095                         "(%d) doesn't match the number hardware can supported "
5096                         "(%d)", reta_size, sp_reta_size);
5097                 return -EINVAL;
5098         }
5099
5100         for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5101                 idx = i / RTE_RETA_GROUP_SIZE;
5102                 shift = i % RTE_RETA_GROUP_SIZE;
5103                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5104                                                 IXGBE_4_BIT_MASK);
5105                 if (!mask)
5106                         continue;
5107
5108                 reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5109                 reta = IXGBE_READ_REG(hw, reta_reg);
5110                 for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5111                         if (mask & (0x1 << j))
5112                                 reta_conf[idx].reta[shift + j] =
5113                                         ((reta >> (CHAR_BIT * j)) &
5114                                                 IXGBE_8_BIT_MASK);
5115                 }
5116         }
5117
5118         return 0;
5119 }
5120
5121 static int
5122 ixgbe_add_rar(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
5123                                 uint32_t index, uint32_t pool)
5124 {
5125         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5126         uint32_t enable_addr = 1;
5127
5128         return ixgbe_set_rar(hw, index, mac_addr->addr_bytes,
5129                              pool, enable_addr);
5130 }
5131
5132 static void
5133 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
5134 {
5135         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5136
5137         ixgbe_clear_rar(hw, index);
5138 }
5139
5140 static int
5141 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
5142 {
5143         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5144
5145         ixgbe_remove_rar(dev, 0);
5146         ixgbe_add_rar(dev, addr, 0, pci_dev->max_vfs);
5147
5148         return 0;
5149 }
5150
5151 static bool
5152 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
5153 {
5154         if (strcmp(dev->device->driver->name, drv->driver.name))
5155                 return false;
5156
5157         return true;
5158 }
5159
5160 bool
5161 is_ixgbe_supported(struct rte_eth_dev *dev)
5162 {
5163         return is_device_supported(dev, &rte_ixgbe_pmd);
5164 }
5165
5166 static int
5167 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
5168 {
5169         uint32_t hlreg0;
5170         uint32_t maxfrs;
5171         struct ixgbe_hw *hw;
5172         struct rte_eth_dev_info dev_info;
5173         uint32_t frame_size = mtu + IXGBE_ETH_OVERHEAD;
5174         struct rte_eth_dev_data *dev_data = dev->data;
5175         int ret;
5176
5177         ret = ixgbe_dev_info_get(dev, &dev_info);
5178         if (ret != 0)
5179                 return ret;
5180
5181         /* check that mtu is within the allowed range */
5182         if (mtu < RTE_ETHER_MIN_MTU || frame_size > dev_info.max_rx_pktlen)
5183                 return -EINVAL;
5184
5185         /* If device is started, refuse mtu that requires the support of
5186          * scattered packets when this feature has not been enabled before.
5187          */
5188         if (dev_data->dev_started && !dev_data->scattered_rx &&
5189             (frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
5190              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
5191                 PMD_INIT_LOG(ERR, "Stop port first.");
5192                 return -EINVAL;
5193         }
5194
5195         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5196         hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
5197
5198         /* switch to jumbo mode if needed */
5199         if (frame_size > IXGBE_ETH_MAX_LEN) {
5200                 dev->data->dev_conf.rxmode.offloads |=
5201                         DEV_RX_OFFLOAD_JUMBO_FRAME;
5202                 hlreg0 |= IXGBE_HLREG0_JUMBOEN;
5203         } else {
5204                 dev->data->dev_conf.rxmode.offloads &=
5205                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
5206                 hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
5207         }
5208         IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
5209
5210         /* update max frame size */
5211         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
5212
5213         maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
5214         maxfrs &= 0x0000FFFF;
5215         maxfrs |= (dev->data->dev_conf.rxmode.max_rx_pkt_len << 16);
5216         IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
5217
5218         return 0;
5219 }
5220
5221 /*
5222  * Virtual Function operations
5223  */
5224 static void
5225 ixgbevf_intr_disable(struct rte_eth_dev *dev)
5226 {
5227         struct ixgbe_interrupt *intr =
5228                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5229         struct ixgbe_hw *hw =
5230                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5231
5232         PMD_INIT_FUNC_TRACE();
5233
5234         /* Clear interrupt mask to stop from interrupts being generated */
5235         IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
5236
5237         IXGBE_WRITE_FLUSH(hw);
5238
5239         /* Clear mask value. */
5240         intr->mask = 0;
5241 }
5242
5243 static void
5244 ixgbevf_intr_enable(struct rte_eth_dev *dev)
5245 {
5246         struct ixgbe_interrupt *intr =
5247                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5248         struct ixgbe_hw *hw =
5249                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5250
5251         PMD_INIT_FUNC_TRACE();
5252
5253         /* VF enable interrupt autoclean */
5254         IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
5255         IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
5256         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
5257
5258         IXGBE_WRITE_FLUSH(hw);
5259
5260         /* Save IXGBE_VTEIMS value to mask. */
5261         intr->mask = IXGBE_VF_IRQ_ENABLE_MASK;
5262 }
5263
5264 static int
5265 ixgbevf_dev_configure(struct rte_eth_dev *dev)
5266 {
5267         struct rte_eth_conf *conf = &dev->data->dev_conf;
5268         struct ixgbe_adapter *adapter = dev->data->dev_private;
5269
5270         PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
5271                      dev->data->port_id);
5272
5273         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
5274                 dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
5275
5276         /*
5277          * VF has no ability to enable/disable HW CRC
5278          * Keep the persistent behavior the same as Host PF
5279          */
5280 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
5281         if (conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
5282                 PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
5283                 conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_KEEP_CRC;
5284         }
5285 #else
5286         if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC)) {
5287                 PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
5288                 conf->rxmode.offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
5289         }
5290 #endif
5291
5292         /*
5293          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
5294          * allocation or vector Rx preconditions we will reset it.
5295          */
5296         adapter->rx_bulk_alloc_allowed = true;
5297         adapter->rx_vec_allowed = true;
5298
5299         return 0;
5300 }
5301
5302 static int
5303 ixgbevf_dev_start(struct rte_eth_dev *dev)
5304 {
5305         struct ixgbe_hw *hw =
5306                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5307         uint32_t intr_vector = 0;
5308         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5309         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5310
5311         int err, mask = 0;
5312
5313         PMD_INIT_FUNC_TRACE();
5314
5315         /* Stop the link setup handler before resetting the HW. */
5316         ixgbe_dev_wait_setup_link_complete(dev, 0);
5317
5318         err = hw->mac.ops.reset_hw(hw);
5319
5320         /**
5321          * In this case, reuses the MAC address assigned by VF
5322          * initialization.
5323          */
5324         if (err != IXGBE_SUCCESS && err != IXGBE_ERR_INVALID_MAC_ADDR) {
5325                 PMD_INIT_LOG(ERR, "Unable to reset vf hardware (%d)", err);
5326                 return err;
5327         }
5328
5329         hw->mac.get_link_status = true;
5330
5331         /* negotiate mailbox API version to use with the PF. */
5332         ixgbevf_negotiate_api(hw);
5333
5334         ixgbevf_dev_tx_init(dev);
5335
5336         /* This can fail when allocating mbufs for descriptor rings */
5337         err = ixgbevf_dev_rx_init(dev);
5338         if (err) {
5339                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
5340                 ixgbe_dev_clear_queues(dev);
5341                 return err;
5342         }
5343
5344         /* Set vfta */
5345         ixgbevf_set_vfta_all(dev, 1);
5346
5347         /* Set HW strip */
5348         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
5349                 ETH_VLAN_EXTEND_MASK;
5350         err = ixgbevf_vlan_offload_config(dev, mask);
5351         if (err) {
5352                 PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
5353                 ixgbe_dev_clear_queues(dev);
5354                 return err;
5355         }
5356
5357         ixgbevf_dev_rxtx_start(dev);
5358
5359         /* check and configure queue intr-vector mapping */
5360         if (rte_intr_cap_multiple(intr_handle) &&
5361             dev->data->dev_conf.intr_conf.rxq) {
5362                 /* According to datasheet, only vector 0/1/2 can be used,
5363                  * now only one vector is used for Rx queue
5364                  */
5365                 intr_vector = 1;
5366                 if (rte_intr_efd_enable(intr_handle, intr_vector))
5367                         return -1;
5368         }
5369
5370         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
5371                 intr_handle->intr_vec =
5372                         rte_zmalloc("intr_vec",
5373                                     dev->data->nb_rx_queues * sizeof(int), 0);
5374                 if (intr_handle->intr_vec == NULL) {
5375                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
5376                                      " intr_vec", dev->data->nb_rx_queues);
5377                         return -ENOMEM;
5378                 }
5379         }
5380         ixgbevf_configure_msix(dev);
5381
5382         /* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
5383          * is mapped to VFIO vector 0 in eth_ixgbevf_dev_init( ).
5384          * If previous VFIO interrupt mapping setting in eth_ixgbevf_dev_init( )
5385          * is not cleared, it will fail when following rte_intr_enable( ) tries
5386          * to map Rx queue interrupt to other VFIO vectors.
5387          * So clear uio/vfio intr/evevnfd first to avoid failure.
5388          */
5389         rte_intr_disable(intr_handle);
5390
5391         rte_intr_enable(intr_handle);
5392
5393         /* Re-enable interrupt for VF */
5394         ixgbevf_intr_enable(dev);
5395
5396         /*
5397          * Update link status right before return, because it may
5398          * start link configuration process in a separate thread.
5399          */
5400         ixgbevf_dev_link_update(dev, 0);
5401
5402         hw->adapter_stopped = false;
5403
5404         return 0;
5405 }
5406
5407 static int
5408 ixgbevf_dev_stop(struct rte_eth_dev *dev)
5409 {
5410         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5411         struct ixgbe_adapter *adapter = dev->data->dev_private;
5412         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5413         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5414
5415         if (hw->adapter_stopped)
5416                 return 0;
5417
5418         PMD_INIT_FUNC_TRACE();
5419
5420         ixgbe_dev_wait_setup_link_complete(dev, 0);
5421
5422         ixgbevf_intr_disable(dev);
5423
5424         dev->data->dev_started = 0;
5425         hw->adapter_stopped = 1;
5426         ixgbe_stop_adapter(hw);
5427
5428         /*
5429           * Clear what we set, but we still keep shadow_vfta to
5430           * restore after device starts
5431           */
5432         ixgbevf_set_vfta_all(dev, 0);
5433
5434         /* Clear stored conf */
5435         dev->data->scattered_rx = 0;
5436
5437         ixgbe_dev_clear_queues(dev);
5438
5439         /* Clean datapath event and queue/vec mapping */
5440         rte_intr_efd_disable(intr_handle);
5441         if (intr_handle->intr_vec != NULL) {
5442                 rte_free(intr_handle->intr_vec);
5443                 intr_handle->intr_vec = NULL;
5444         }
5445
5446         adapter->rss_reta_updated = 0;
5447
5448         return 0;
5449 }
5450
5451 static int
5452 ixgbevf_dev_close(struct rte_eth_dev *dev)
5453 {
5454         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5455         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5456         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5457         int ret;
5458
5459         PMD_INIT_FUNC_TRACE();
5460         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5461                 return 0;
5462
5463         ixgbe_reset_hw(hw);
5464
5465         ret = ixgbevf_dev_stop(dev);
5466
5467         ixgbe_dev_free_queues(dev);
5468
5469         /**
5470          * Remove the VF MAC address ro ensure
5471          * that the VF traffic goes to the PF
5472          * after stop, close and detach of the VF
5473          **/
5474         ixgbevf_remove_mac_addr(dev, 0);
5475
5476         rte_intr_disable(intr_handle);
5477         rte_intr_callback_unregister(intr_handle,
5478                                      ixgbevf_dev_interrupt_handler, dev);
5479
5480         return ret;
5481 }
5482
5483 /*
5484  * Reset VF device
5485  */
5486 static int
5487 ixgbevf_dev_reset(struct rte_eth_dev *dev)
5488 {
5489         int ret;
5490
5491         ret = eth_ixgbevf_dev_uninit(dev);
5492         if (ret)
5493                 return ret;
5494
5495         ret = eth_ixgbevf_dev_init(dev);
5496
5497         return ret;
5498 }
5499
5500 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
5501 {
5502         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5503         struct ixgbe_vfta *shadow_vfta =
5504                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5505         int i = 0, j = 0, vfta = 0, mask = 1;
5506
5507         for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
5508                 vfta = shadow_vfta->vfta[i];
5509                 if (vfta) {
5510                         mask = 1;
5511                         for (j = 0; j < 32; j++) {
5512                                 if (vfta & mask)
5513                                         ixgbe_set_vfta(hw, (i<<5)+j, 0,
5514                                                        on, false);
5515                                 mask <<= 1;
5516                         }
5517                 }
5518         }
5519
5520 }
5521
5522 static int
5523 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
5524 {
5525         struct ixgbe_hw *hw =
5526                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5527         struct ixgbe_vfta *shadow_vfta =
5528                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5529         uint32_t vid_idx = 0;
5530         uint32_t vid_bit = 0;
5531         int ret = 0;
5532
5533         PMD_INIT_FUNC_TRACE();
5534
5535         /* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
5536         ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
5537         if (ret) {
5538                 PMD_INIT_LOG(ERR, "Unable to set VF vlan");
5539                 return ret;
5540         }
5541         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
5542         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
5543
5544         /* Save what we set and retore it after device reset */
5545         if (on)
5546                 shadow_vfta->vfta[vid_idx] |= vid_bit;
5547         else
5548                 shadow_vfta->vfta[vid_idx] &= ~vid_bit;
5549
5550         return 0;
5551 }
5552
5553 static void
5554 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
5555 {
5556         struct ixgbe_hw *hw =
5557                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5558         uint32_t ctrl;
5559
5560         PMD_INIT_FUNC_TRACE();
5561
5562         if (queue >= hw->mac.max_rx_queues)
5563                 return;
5564
5565         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
5566         if (on)
5567                 ctrl |= IXGBE_RXDCTL_VME;
5568         else
5569                 ctrl &= ~IXGBE_RXDCTL_VME;
5570         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
5571
5572         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
5573 }
5574
5575 static int
5576 ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask)
5577 {
5578         struct ixgbe_rx_queue *rxq;
5579         uint16_t i;
5580         int on = 0;
5581
5582         /* VF function only support hw strip feature, others are not support */
5583         if (mask & ETH_VLAN_STRIP_MASK) {
5584                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
5585                         rxq = dev->data->rx_queues[i];
5586                         on = !!(rxq->offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
5587                         ixgbevf_vlan_strip_queue_set(dev, i, on);
5588                 }
5589         }
5590
5591         return 0;
5592 }
5593
5594 static int
5595 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
5596 {
5597         ixgbe_config_vlan_strip_on_all_queues(dev, mask);
5598
5599         ixgbevf_vlan_offload_config(dev, mask);
5600
5601         return 0;
5602 }
5603
5604 int
5605 ixgbe_vt_check(struct ixgbe_hw *hw)
5606 {
5607         uint32_t reg_val;
5608
5609         /* if Virtualization Technology is enabled */
5610         reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
5611         if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
5612                 PMD_INIT_LOG(ERR, "VT must be enabled for this setting");
5613                 return -1;
5614         }
5615
5616         return 0;
5617 }
5618
5619 static uint32_t
5620 ixgbe_uta_vector(struct ixgbe_hw *hw, struct rte_ether_addr *uc_addr)
5621 {
5622         uint32_t vector = 0;
5623
5624         switch (hw->mac.mc_filter_type) {
5625         case 0:   /* use bits [47:36] of the address */
5626                 vector = ((uc_addr->addr_bytes[4] >> 4) |
5627                         (((uint16_t)uc_addr->addr_bytes[5]) << 4));
5628                 break;
5629         case 1:   /* use bits [46:35] of the address */
5630                 vector = ((uc_addr->addr_bytes[4] >> 3) |
5631                         (((uint16_t)uc_addr->addr_bytes[5]) << 5));
5632                 break;
5633         case 2:   /* use bits [45:34] of the address */
5634                 vector = ((uc_addr->addr_bytes[4] >> 2) |
5635                         (((uint16_t)uc_addr->addr_bytes[5]) << 6));
5636                 break;
5637         case 3:   /* use bits [43:32] of the address */
5638                 vector = ((uc_addr->addr_bytes[4]) |
5639                         (((uint16_t)uc_addr->addr_bytes[5]) << 8));
5640                 break;
5641         default:  /* Invalid mc_filter_type */
5642                 break;
5643         }
5644
5645         /* vector can only be 12-bits or boundary will be exceeded */
5646         vector &= 0xFFF;
5647         return vector;
5648 }
5649
5650 static int
5651 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev,
5652                         struct rte_ether_addr *mac_addr, uint8_t on)
5653 {
5654         uint32_t vector;
5655         uint32_t uta_idx;
5656         uint32_t reg_val;
5657         uint32_t uta_shift;
5658         uint32_t rc;
5659         const uint32_t ixgbe_uta_idx_mask = 0x7F;
5660         const uint32_t ixgbe_uta_bit_shift = 5;
5661         const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
5662         const uint32_t bit1 = 0x1;
5663
5664         struct ixgbe_hw *hw =
5665                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5666         struct ixgbe_uta_info *uta_info =
5667                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5668
5669         /* The UTA table only exists on 82599 hardware and newer */
5670         if (hw->mac.type < ixgbe_mac_82599EB)
5671                 return -ENOTSUP;
5672
5673         vector = ixgbe_uta_vector(hw, mac_addr);
5674         uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
5675         uta_shift = vector & ixgbe_uta_bit_mask;
5676
5677         rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
5678         if (rc == on)
5679                 return 0;
5680
5681         reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
5682         if (on) {
5683                 uta_info->uta_in_use++;
5684                 reg_val |= (bit1 << uta_shift);
5685                 uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
5686         } else {
5687                 uta_info->uta_in_use--;
5688                 reg_val &= ~(bit1 << uta_shift);
5689                 uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
5690         }
5691
5692         IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
5693
5694         if (uta_info->uta_in_use > 0)
5695                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
5696                                 IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
5697         else
5698                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
5699
5700         return 0;
5701 }
5702
5703 static int
5704 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
5705 {
5706         int i;
5707         struct ixgbe_hw *hw =
5708                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5709         struct ixgbe_uta_info *uta_info =
5710                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5711
5712         /* The UTA table only exists on 82599 hardware and newer */
5713         if (hw->mac.type < ixgbe_mac_82599EB)
5714                 return -ENOTSUP;
5715
5716         if (on) {
5717                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5718                         uta_info->uta_shadow[i] = ~0;
5719                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
5720                 }
5721         } else {
5722                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5723                         uta_info->uta_shadow[i] = 0;
5724                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
5725                 }
5726         }
5727         return 0;
5728
5729 }
5730
5731 uint32_t
5732 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
5733 {
5734         uint32_t new_val = orig_val;
5735
5736         if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG)
5737                 new_val |= IXGBE_VMOLR_AUPE;
5738         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC)
5739                 new_val |= IXGBE_VMOLR_ROMPE;
5740         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
5741                 new_val |= IXGBE_VMOLR_ROPE;
5742         if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
5743                 new_val |= IXGBE_VMOLR_BAM;
5744         if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
5745                 new_val |= IXGBE_VMOLR_MPE;
5746
5747         return new_val;
5748 }
5749
5750 #define IXGBE_MRCTL_VPME  0x01 /* Virtual Pool Mirroring. */
5751 #define IXGBE_MRCTL_UPME  0x02 /* Uplink Port Mirroring. */
5752 #define IXGBE_MRCTL_DPME  0x04 /* Downlink Port Mirroring. */
5753 #define IXGBE_MRCTL_VLME  0x08 /* VLAN Mirroring. */
5754 #define IXGBE_INVALID_MIRROR_TYPE(mirror_type) \
5755         ((mirror_type) & ~(uint8_t)(ETH_MIRROR_VIRTUAL_POOL_UP | \
5756         ETH_MIRROR_UPLINK_PORT | ETH_MIRROR_DOWNLINK_PORT | ETH_MIRROR_VLAN))
5757
5758 static int
5759 ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
5760                       struct rte_eth_mirror_conf *mirror_conf,
5761                       uint8_t rule_id, uint8_t on)
5762 {
5763         uint32_t mr_ctl, vlvf;
5764         uint32_t mp_lsb = 0;
5765         uint32_t mv_msb = 0;
5766         uint32_t mv_lsb = 0;
5767         uint32_t mp_msb = 0;
5768         uint8_t i = 0;
5769         int reg_index = 0;
5770         uint64_t vlan_mask = 0;
5771
5772         const uint8_t pool_mask_offset = 32;
5773         const uint8_t vlan_mask_offset = 32;
5774         const uint8_t dst_pool_offset = 8;
5775         const uint8_t rule_mr_offset  = 4;
5776         const uint8_t mirror_rule_mask = 0x0F;
5777
5778         struct ixgbe_mirror_info *mr_info =
5779                         (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5780         struct ixgbe_hw *hw =
5781                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5782         uint8_t mirror_type = 0;
5783
5784         if (ixgbe_vt_check(hw) < 0)
5785                 return -ENOTSUP;
5786
5787         if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5788                 return -EINVAL;
5789
5790         if (IXGBE_INVALID_MIRROR_TYPE(mirror_conf->rule_type)) {
5791                 PMD_DRV_LOG(ERR, "unsupported mirror type 0x%x.",
5792                             mirror_conf->rule_type);
5793                 return -EINVAL;
5794         }
5795
5796         if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5797                 mirror_type |= IXGBE_MRCTL_VLME;
5798                 /* Check if vlan id is valid and find conresponding VLAN ID
5799                  * index in VLVF
5800                  */
5801                 for (i = 0; i < IXGBE_VLVF_ENTRIES; i++) {
5802                         if (mirror_conf->vlan.vlan_mask & (1ULL << i)) {
5803                                 /* search vlan id related pool vlan filter
5804                                  * index
5805                                  */
5806                                 reg_index = ixgbe_find_vlvf_slot(
5807                                                 hw,
5808                                                 mirror_conf->vlan.vlan_id[i],
5809                                                 false);
5810                                 if (reg_index < 0)
5811                                         return -EINVAL;
5812                                 vlvf = IXGBE_READ_REG(hw,
5813                                                       IXGBE_VLVF(reg_index));
5814                                 if ((vlvf & IXGBE_VLVF_VIEN) &&
5815                                     ((vlvf & IXGBE_VLVF_VLANID_MASK) ==
5816                                       mirror_conf->vlan.vlan_id[i]))
5817                                         vlan_mask |= (1ULL << reg_index);
5818                                 else
5819                                         return -EINVAL;
5820                         }
5821                 }
5822
5823                 if (on) {
5824                         mv_lsb = vlan_mask & 0xFFFFFFFF;
5825                         mv_msb = vlan_mask >> vlan_mask_offset;
5826
5827                         mr_info->mr_conf[rule_id].vlan.vlan_mask =
5828                                                 mirror_conf->vlan.vlan_mask;
5829                         for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++) {
5830                                 if (mirror_conf->vlan.vlan_mask & (1ULL << i))
5831                                         mr_info->mr_conf[rule_id].vlan.vlan_id[i] =
5832                                                 mirror_conf->vlan.vlan_id[i];
5833                         }
5834                 } else {
5835                         mv_lsb = 0;
5836                         mv_msb = 0;
5837                         mr_info->mr_conf[rule_id].vlan.vlan_mask = 0;
5838                         for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++)
5839                                 mr_info->mr_conf[rule_id].vlan.vlan_id[i] = 0;
5840                 }
5841         }
5842
5843         /**
5844          * if enable pool mirror, write related pool mask register,if disable
5845          * pool mirror, clear PFMRVM register
5846          */
5847         if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5848                 mirror_type |= IXGBE_MRCTL_VPME;
5849                 if (on) {
5850                         mp_lsb = mirror_conf->pool_mask & 0xFFFFFFFF;
5851                         mp_msb = mirror_conf->pool_mask >> pool_mask_offset;
5852                         mr_info->mr_conf[rule_id].pool_mask =
5853                                         mirror_conf->pool_mask;
5854
5855                 } else {
5856                         mp_lsb = 0;
5857                         mp_msb = 0;
5858                         mr_info->mr_conf[rule_id].pool_mask = 0;
5859                 }
5860         }
5861         if (mirror_conf->rule_type & ETH_MIRROR_UPLINK_PORT)
5862                 mirror_type |= IXGBE_MRCTL_UPME;
5863         if (mirror_conf->rule_type & ETH_MIRROR_DOWNLINK_PORT)
5864                 mirror_type |= IXGBE_MRCTL_DPME;
5865
5866         /* read  mirror control register and recalculate it */
5867         mr_ctl = IXGBE_READ_REG(hw, IXGBE_MRCTL(rule_id));
5868
5869         if (on) {
5870                 mr_ctl |= mirror_type;
5871                 mr_ctl &= mirror_rule_mask;
5872                 mr_ctl |= mirror_conf->dst_pool << dst_pool_offset;
5873         } else {
5874                 mr_ctl &= ~(mirror_conf->rule_type & mirror_rule_mask);
5875         }
5876
5877         mr_info->mr_conf[rule_id].rule_type = mirror_conf->rule_type;
5878         mr_info->mr_conf[rule_id].dst_pool = mirror_conf->dst_pool;
5879
5880         /* write mirrror control  register */
5881         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5882
5883         /* write pool mirrror control  register */
5884         if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5885                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), mp_lsb);
5886                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset),
5887                                 mp_msb);
5888         }
5889         /* write VLAN mirrror control  register */
5890         if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5891                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), mv_lsb);
5892                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset),
5893                                 mv_msb);
5894         }
5895
5896         return 0;
5897 }
5898
5899 static int
5900 ixgbe_mirror_rule_reset(struct rte_eth_dev *dev, uint8_t rule_id)
5901 {
5902         int mr_ctl = 0;
5903         uint32_t lsb_val = 0;
5904         uint32_t msb_val = 0;
5905         const uint8_t rule_mr_offset = 4;
5906
5907         struct ixgbe_hw *hw =
5908                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5909         struct ixgbe_mirror_info *mr_info =
5910                 (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5911
5912         if (ixgbe_vt_check(hw) < 0)
5913                 return -ENOTSUP;
5914
5915         if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5916                 return -EINVAL;
5917
5918         memset(&mr_info->mr_conf[rule_id], 0,
5919                sizeof(struct rte_eth_mirror_conf));
5920
5921         /* clear PFVMCTL register */
5922         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5923
5924         /* clear pool mask register */
5925         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), lsb_val);
5926         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset), msb_val);
5927
5928         /* clear vlan mask register */
5929         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), lsb_val);
5930         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset), msb_val);
5931
5932         return 0;
5933 }
5934
5935 static int
5936 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5937 {
5938         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5939         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5940         struct ixgbe_interrupt *intr =
5941                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5942         struct ixgbe_hw *hw =
5943                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5944         uint32_t vec = IXGBE_MISC_VEC_ID;
5945
5946         if (rte_intr_allow_others(intr_handle))
5947                 vec = IXGBE_RX_VEC_START;
5948         intr->mask |= (1 << vec);
5949         RTE_SET_USED(queue_id);
5950         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5951
5952         rte_intr_ack(intr_handle);
5953
5954         return 0;
5955 }
5956
5957 static int
5958 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5959 {
5960         struct ixgbe_interrupt *intr =
5961                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5962         struct ixgbe_hw *hw =
5963                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5964         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5965         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5966         uint32_t vec = IXGBE_MISC_VEC_ID;
5967
5968         if (rte_intr_allow_others(intr_handle))
5969                 vec = IXGBE_RX_VEC_START;
5970         intr->mask &= ~(1 << vec);
5971         RTE_SET_USED(queue_id);
5972         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5973
5974         return 0;
5975 }
5976
5977 static int
5978 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5979 {
5980         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5981         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5982         uint32_t mask;
5983         struct ixgbe_hw *hw =
5984                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5985         struct ixgbe_interrupt *intr =
5986                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5987
5988         if (queue_id < 16) {
5989                 ixgbe_disable_intr(hw);
5990                 intr->mask |= (1 << queue_id);
5991                 ixgbe_enable_intr(dev);
5992         } else if (queue_id < 32) {
5993                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5994                 mask &= (1 << queue_id);
5995                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5996         } else if (queue_id < 64) {
5997                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
5998                 mask &= (1 << (queue_id - 32));
5999                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
6000         }
6001         rte_intr_ack(intr_handle);
6002
6003         return 0;
6004 }
6005
6006 static int
6007 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
6008 {
6009         uint32_t mask;
6010         struct ixgbe_hw *hw =
6011                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6012         struct ixgbe_interrupt *intr =
6013                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
6014
6015         if (queue_id < 16) {
6016                 ixgbe_disable_intr(hw);
6017                 intr->mask &= ~(1 << queue_id);
6018                 ixgbe_enable_intr(dev);
6019         } else if (queue_id < 32) {
6020                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
6021                 mask &= ~(1 << queue_id);
6022                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
6023         } else if (queue_id < 64) {
6024                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
6025                 mask &= ~(1 << (queue_id - 32));
6026                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
6027         }
6028
6029         return 0;
6030 }
6031
6032 static void
6033 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
6034                      uint8_t queue, uint8_t msix_vector)
6035 {
6036         uint32_t tmp, idx;
6037
6038         if (direction == -1) {
6039                 /* other causes */
6040                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
6041                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
6042                 tmp &= ~0xFF;
6043                 tmp |= msix_vector;
6044                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
6045         } else {
6046                 /* rx or tx cause */
6047                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
6048                 idx = ((16 * (queue & 1)) + (8 * direction));
6049                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
6050                 tmp &= ~(0xFF << idx);
6051                 tmp |= (msix_vector << idx);
6052                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
6053         }
6054 }
6055
6056 /**
6057  * set the IVAR registers, mapping interrupt causes to vectors
6058  * @param hw
6059  *  pointer to ixgbe_hw struct
6060  * @direction
6061  *  0 for Rx, 1 for Tx, -1 for other causes
6062  * @queue
6063  *  queue to map the corresponding interrupt to
6064  * @msix_vector
6065  *  the vector to map to the corresponding queue
6066  */
6067 static void
6068 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
6069                    uint8_t queue, uint8_t msix_vector)
6070 {
6071         uint32_t tmp, idx;
6072
6073         msix_vector |= IXGBE_IVAR_ALLOC_VAL;
6074         if (hw->mac.type == ixgbe_mac_82598EB) {
6075                 if (direction == -1)
6076                         direction = 0;
6077                 idx = (((direction * 64) + queue) >> 2) & 0x1F;
6078                 tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
6079                 tmp &= ~(0xFF << (8 * (queue & 0x3)));
6080                 tmp |= (msix_vector << (8 * (queue & 0x3)));
6081                 IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
6082         } else if ((hw->mac.type == ixgbe_mac_82599EB) ||
6083                         (hw->mac.type == ixgbe_mac_X540) ||
6084                         (hw->mac.type == ixgbe_mac_X550) ||
6085                         (hw->mac.type == ixgbe_mac_X550EM_x)) {
6086                 if (direction == -1) {
6087                         /* other causes */
6088                         idx = ((queue & 1) * 8);
6089                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
6090                         tmp &= ~(0xFF << idx);
6091                         tmp |= (msix_vector << idx);
6092                         IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
6093                 } else {
6094                         /* rx or tx causes */
6095                         idx = ((16 * (queue & 1)) + (8 * direction));
6096                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
6097                         tmp &= ~(0xFF << idx);
6098                         tmp |= (msix_vector << idx);
6099                         IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
6100                 }
6101         }
6102 }
6103
6104 static void
6105 ixgbevf_configure_msix(struct rte_eth_dev *dev)
6106 {
6107         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
6108         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
6109         struct ixgbe_hw *hw =
6110                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6111         uint32_t q_idx;
6112         uint32_t vector_idx = IXGBE_MISC_VEC_ID;
6113         uint32_t base = IXGBE_MISC_VEC_ID;
6114
6115         /* Configure VF other cause ivar */
6116         ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
6117
6118         /* won't configure msix register if no mapping is done
6119          * between intr vector and event fd.
6120          */
6121         if (!rte_intr_dp_is_en(intr_handle))
6122                 return;
6123
6124         if (rte_intr_allow_others(intr_handle)) {
6125                 base = IXGBE_RX_VEC_START;
6126                 vector_idx = IXGBE_RX_VEC_START;
6127         }
6128
6129         /* Configure all RX queues of VF */
6130         for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
6131                 /* Force all queue use vector 0,
6132                  * as IXGBE_VF_MAXMSIVECOTR = 1
6133                  */
6134                 ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
6135                 intr_handle->intr_vec[q_idx] = vector_idx;
6136                 if (vector_idx < base + intr_handle->nb_efd - 1)
6137                         vector_idx++;
6138         }
6139
6140         /* As RX queue setting above show, all queues use the vector 0.
6141          * Set only the ITR value of IXGBE_MISC_VEC_ID.
6142          */
6143         IXGBE_WRITE_REG(hw, IXGBE_VTEITR(IXGBE_MISC_VEC_ID),
6144                         IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
6145                         | IXGBE_EITR_CNT_WDIS);
6146 }
6147
6148 /**
6149  * Sets up the hardware to properly generate MSI-X interrupts
6150  * @hw
6151  *  board private structure
6152  */
6153 static void
6154 ixgbe_configure_msix(struct rte_eth_dev *dev)
6155 {
6156         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
6157         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
6158         struct ixgbe_hw *hw =
6159                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6160         uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
6161         uint32_t vec = IXGBE_MISC_VEC_ID;
6162         uint32_t mask;
6163         uint32_t gpie;
6164
6165         /* won't configure msix register if no mapping is done
6166          * between intr vector and event fd
6167          * but if misx has been enabled already, need to configure
6168          * auto clean, auto mask and throttling.
6169          */
6170         gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
6171         if (!rte_intr_dp_is_en(intr_handle) &&
6172             !(gpie & (IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT)))
6173                 return;
6174
6175         if (rte_intr_allow_others(intr_handle))
6176                 vec = base = IXGBE_RX_VEC_START;
6177
6178         /* setup GPIE for MSI-x mode */
6179         gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
6180         gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
6181                 IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
6182         /* auto clearing and auto setting corresponding bits in EIMS
6183          * when MSI-X interrupt is triggered
6184          */
6185         if (hw->mac.type == ixgbe_mac_82598EB) {
6186                 IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
6187         } else {
6188                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
6189                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
6190         }
6191         IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
6192
6193         /* Populate the IVAR table and set the ITR values to the
6194          * corresponding register.
6195          */
6196         if (rte_intr_dp_is_en(intr_handle)) {
6197                 for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
6198                         queue_id++) {
6199                         /* by default, 1:1 mapping */
6200                         ixgbe_set_ivar_map(hw, 0, queue_id, vec);
6201                         intr_handle->intr_vec[queue_id] = vec;
6202                         if (vec < base + intr_handle->nb_efd - 1)
6203                                 vec++;
6204                 }
6205
6206                 switch (hw->mac.type) {
6207                 case ixgbe_mac_82598EB:
6208                         ixgbe_set_ivar_map(hw, -1,
6209                                            IXGBE_IVAR_OTHER_CAUSES_INDEX,
6210                                            IXGBE_MISC_VEC_ID);
6211                         break;
6212                 case ixgbe_mac_82599EB:
6213                 case ixgbe_mac_X540:
6214                 case ixgbe_mac_X550:
6215                 case ixgbe_mac_X550EM_x:
6216                         ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
6217                         break;
6218                 default:
6219                         break;
6220                 }
6221         }
6222         IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
6223                         IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
6224                         | IXGBE_EITR_CNT_WDIS);
6225
6226         /* set up to autoclear timer, and the vectors */
6227         mask = IXGBE_EIMS_ENABLE_MASK;
6228         mask &= ~(IXGBE_EIMS_OTHER |
6229                   IXGBE_EIMS_MAILBOX |
6230                   IXGBE_EIMS_LSC);
6231
6232         IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
6233 }
6234
6235 int
6236 ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
6237                            uint16_t queue_idx, uint16_t tx_rate)
6238 {
6239         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6240         struct rte_eth_rxmode *rxmode;
6241         uint32_t rf_dec, rf_int;
6242         uint32_t bcnrc_val;
6243         uint16_t link_speed = dev->data->dev_link.link_speed;
6244
6245         if (queue_idx >= hw->mac.max_tx_queues)
6246                 return -EINVAL;
6247
6248         if (tx_rate != 0) {
6249                 /* Calculate the rate factor values to set */
6250                 rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
6251                 rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
6252                 rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
6253
6254                 bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
6255                 bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
6256                                 IXGBE_RTTBCNRC_RF_INT_MASK_M);
6257                 bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
6258         } else {
6259                 bcnrc_val = 0;
6260         }
6261
6262         rxmode = &dev->data->dev_conf.rxmode;
6263         /*
6264          * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
6265          * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
6266          * set as 0x4.
6267          */
6268         if ((rxmode->offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) &&
6269             (rxmode->max_rx_pkt_len >= IXGBE_MAX_JUMBO_FRAME_SIZE))
6270                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
6271                         IXGBE_MMW_SIZE_JUMBO_FRAME);
6272         else
6273                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
6274                         IXGBE_MMW_SIZE_DEFAULT);
6275
6276         /* Set RTTBCNRC of queue X */
6277         IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
6278         IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
6279         IXGBE_WRITE_FLUSH(hw);
6280
6281         return 0;
6282 }
6283
6284 static int
6285 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
6286                      __rte_unused uint32_t index,
6287                      __rte_unused uint32_t pool)
6288 {
6289         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6290         int diag;
6291
6292         /*
6293          * On a 82599 VF, adding again the same MAC addr is not an idempotent
6294          * operation. Trap this case to avoid exhausting the [very limited]
6295          * set of PF resources used to store VF MAC addresses.
6296          */
6297         if (memcmp(hw->mac.perm_addr, mac_addr,
6298                         sizeof(struct rte_ether_addr)) == 0)
6299                 return -1;
6300         diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6301         if (diag != 0)
6302                 PMD_DRV_LOG(ERR, "Unable to add MAC address "
6303                             "%02x:%02x:%02x:%02x:%02x:%02x - diag=%d",
6304                             mac_addr->addr_bytes[0],
6305                             mac_addr->addr_bytes[1],
6306                             mac_addr->addr_bytes[2],
6307                             mac_addr->addr_bytes[3],
6308                             mac_addr->addr_bytes[4],
6309                             mac_addr->addr_bytes[5],
6310                             diag);
6311         return diag;
6312 }
6313
6314 static void
6315 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
6316 {
6317         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6318         struct rte_ether_addr *perm_addr =
6319                 (struct rte_ether_addr *)hw->mac.perm_addr;
6320         struct rte_ether_addr *mac_addr;
6321         uint32_t i;
6322         int diag;
6323
6324         /*
6325          * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
6326          * not support the deletion of a given MAC address.
6327          * Instead, it imposes to delete all MAC addresses, then to add again
6328          * all MAC addresses with the exception of the one to be deleted.
6329          */
6330         (void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
6331
6332         /*
6333          * Add again all MAC addresses, with the exception of the deleted one
6334          * and of the permanent MAC address.
6335          */
6336         for (i = 0, mac_addr = dev->data->mac_addrs;
6337              i < hw->mac.num_rar_entries; i++, mac_addr++) {
6338                 /* Skip the deleted MAC address */
6339                 if (i == index)
6340                         continue;
6341                 /* Skip NULL MAC addresses */
6342                 if (rte_is_zero_ether_addr(mac_addr))
6343                         continue;
6344                 /* Skip the permanent MAC address */
6345                 if (memcmp(perm_addr, mac_addr,
6346                                 sizeof(struct rte_ether_addr)) == 0)
6347                         continue;
6348                 diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6349                 if (diag != 0)
6350                         PMD_DRV_LOG(ERR,
6351                                     "Adding again MAC address "
6352                                     "%02x:%02x:%02x:%02x:%02x:%02x failed "
6353                                     "diag=%d",
6354                                     mac_addr->addr_bytes[0],
6355                                     mac_addr->addr_bytes[1],
6356                                     mac_addr->addr_bytes[2],
6357                                     mac_addr->addr_bytes[3],
6358                                     mac_addr->addr_bytes[4],
6359                                     mac_addr->addr_bytes[5],
6360                                     diag);
6361         }
6362 }
6363
6364 static int
6365 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
6366                         struct rte_ether_addr *addr)
6367 {
6368         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6369
6370         hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
6371
6372         return 0;
6373 }
6374
6375 int
6376 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
6377                         struct rte_eth_syn_filter *filter,
6378                         bool add)
6379 {
6380         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6381         struct ixgbe_filter_info *filter_info =
6382                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6383         uint32_t syn_info;
6384         uint32_t synqf;
6385
6386         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6387                 return -EINVAL;
6388
6389         syn_info = filter_info->syn_info;
6390
6391         if (add) {
6392                 if (syn_info & IXGBE_SYN_FILTER_ENABLE)
6393                         return -EINVAL;
6394                 synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
6395                         IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
6396
6397                 if (filter->hig_pri)
6398                         synqf |= IXGBE_SYN_FILTER_SYNQFP;
6399                 else
6400                         synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
6401         } else {
6402                 synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6403                 if (!(syn_info & IXGBE_SYN_FILTER_ENABLE))
6404                         return -ENOENT;
6405                 synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
6406         }
6407
6408         filter_info->syn_info = synqf;
6409         IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
6410         IXGBE_WRITE_FLUSH(hw);
6411         return 0;
6412 }
6413
6414
6415 static inline enum ixgbe_5tuple_protocol
6416 convert_protocol_type(uint8_t protocol_value)
6417 {
6418         if (protocol_value == IPPROTO_TCP)
6419                 return IXGBE_FILTER_PROTOCOL_TCP;
6420         else if (protocol_value == IPPROTO_UDP)
6421                 return IXGBE_FILTER_PROTOCOL_UDP;
6422         else if (protocol_value == IPPROTO_SCTP)
6423                 return IXGBE_FILTER_PROTOCOL_SCTP;
6424         else
6425                 return IXGBE_FILTER_PROTOCOL_NONE;
6426 }
6427
6428 /* inject a 5-tuple filter to HW */
6429 static inline void
6430 ixgbe_inject_5tuple_filter(struct rte_eth_dev *dev,
6431                            struct ixgbe_5tuple_filter *filter)
6432 {
6433         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6434         int i;
6435         uint32_t ftqf, sdpqf;
6436         uint32_t l34timir = 0;
6437         uint8_t mask = 0xff;
6438
6439         i = filter->index;
6440
6441         sdpqf = (uint32_t)(filter->filter_info.dst_port <<
6442                                 IXGBE_SDPQF_DSTPORT_SHIFT);
6443         sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
6444
6445         ftqf = (uint32_t)(filter->filter_info.proto &
6446                 IXGBE_FTQF_PROTOCOL_MASK);
6447         ftqf |= (uint32_t)((filter->filter_info.priority &
6448                 IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
6449         if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
6450                 mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
6451         if (filter->filter_info.dst_ip_mask == 0)
6452                 mask &= IXGBE_FTQF_DEST_ADDR_MASK;
6453         if (filter->filter_info.src_port_mask == 0)
6454                 mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
6455         if (filter->filter_info.dst_port_mask == 0)
6456                 mask &= IXGBE_FTQF_DEST_PORT_MASK;
6457         if (filter->filter_info.proto_mask == 0)
6458                 mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
6459         ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
6460         ftqf |= IXGBE_FTQF_POOL_MASK_EN;
6461         ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
6462
6463         IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
6464         IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
6465         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
6466         IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
6467
6468         l34timir |= IXGBE_L34T_IMIR_RESERVE;
6469         l34timir |= (uint32_t)(filter->queue <<
6470                                 IXGBE_L34T_IMIR_QUEUE_SHIFT);
6471         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
6472 }
6473
6474 /*
6475  * add a 5tuple filter
6476  *
6477  * @param
6478  * dev: Pointer to struct rte_eth_dev.
6479  * index: the index the filter allocates.
6480  * filter: ponter to the filter that will be added.
6481  * rx_queue: the queue id the filter assigned to.
6482  *
6483  * @return
6484  *    - On success, zero.
6485  *    - On failure, a negative value.
6486  */
6487 static int
6488 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
6489                         struct ixgbe_5tuple_filter *filter)
6490 {
6491         struct ixgbe_filter_info *filter_info =
6492                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6493         int i, idx, shift;
6494
6495         /*
6496          * look for an unused 5tuple filter index,
6497          * and insert the filter to list.
6498          */
6499         for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
6500                 idx = i / (sizeof(uint32_t) * NBBY);
6501                 shift = i % (sizeof(uint32_t) * NBBY);
6502                 if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
6503                         filter_info->fivetuple_mask[idx] |= 1 << shift;
6504                         filter->index = i;
6505                         TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
6506                                           filter,
6507                                           entries);
6508                         break;
6509                 }
6510         }
6511         if (i >= IXGBE_MAX_FTQF_FILTERS) {
6512                 PMD_DRV_LOG(ERR, "5tuple filters are full.");
6513                 return -ENOSYS;
6514         }
6515
6516         ixgbe_inject_5tuple_filter(dev, filter);
6517
6518         return 0;
6519 }
6520
6521 /*
6522  * remove a 5tuple filter
6523  *
6524  * @param
6525  * dev: Pointer to struct rte_eth_dev.
6526  * filter: the pointer of the filter will be removed.
6527  */
6528 static void
6529 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
6530                         struct ixgbe_5tuple_filter *filter)
6531 {
6532         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6533         struct ixgbe_filter_info *filter_info =
6534                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6535         uint16_t index = filter->index;
6536
6537         filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
6538                                 ~(1 << (index % (sizeof(uint32_t) * NBBY)));
6539         TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
6540         rte_free(filter);
6541
6542         IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
6543         IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
6544         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
6545         IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
6546         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
6547 }
6548
6549 static int
6550 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
6551 {
6552         struct ixgbe_hw *hw;
6553         uint32_t max_frame = mtu + IXGBE_ETH_OVERHEAD;
6554         struct rte_eth_dev_data *dev_data = dev->data;
6555
6556         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6557
6558         if (mtu < RTE_ETHER_MIN_MTU ||
6559                         max_frame > RTE_ETHER_MAX_JUMBO_FRAME_LEN)
6560                 return -EINVAL;
6561
6562         /* If device is started, refuse mtu that requires the support of
6563          * scattered packets when this feature has not been enabled before.
6564          */
6565         if (dev_data->dev_started && !dev_data->scattered_rx &&
6566             (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
6567              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
6568                 PMD_INIT_LOG(ERR, "Stop port first.");
6569                 return -EINVAL;
6570         }
6571
6572         /*
6573          * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
6574          * request of the version 2.0 of the mailbox API.
6575          * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
6576          * of the mailbox API.
6577          * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
6578          * prior to 3.11.33 which contains the following change:
6579          * "ixgbe: Enable jumbo frames support w/ SR-IOV"
6580          */
6581         if (ixgbevf_rlpml_set_vf(hw, max_frame))
6582                 return -EINVAL;
6583
6584         /* update max frame size */
6585         dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame;
6586         return 0;
6587 }
6588
6589 static inline struct ixgbe_5tuple_filter *
6590 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
6591                         struct ixgbe_5tuple_filter_info *key)
6592 {
6593         struct ixgbe_5tuple_filter *it;
6594
6595         TAILQ_FOREACH(it, filter_list, entries) {
6596                 if (memcmp(key, &it->filter_info,
6597                         sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
6598                         return it;
6599                 }
6600         }
6601         return NULL;
6602 }
6603
6604 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
6605 static inline int
6606 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
6607                         struct ixgbe_5tuple_filter_info *filter_info)
6608 {
6609         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
6610                 filter->priority > IXGBE_5TUPLE_MAX_PRI ||
6611                 filter->priority < IXGBE_5TUPLE_MIN_PRI)
6612                 return -EINVAL;
6613
6614         switch (filter->dst_ip_mask) {
6615         case UINT32_MAX:
6616                 filter_info->dst_ip_mask = 0;
6617                 filter_info->dst_ip = filter->dst_ip;
6618                 break;
6619         case 0:
6620                 filter_info->dst_ip_mask = 1;
6621                 break;
6622         default:
6623                 PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
6624                 return -EINVAL;
6625         }
6626
6627         switch (filter->src_ip_mask) {
6628         case UINT32_MAX:
6629                 filter_info->src_ip_mask = 0;
6630                 filter_info->src_ip = filter->src_ip;
6631                 break;
6632         case 0:
6633                 filter_info->src_ip_mask = 1;
6634                 break;
6635         default:
6636                 PMD_DRV_LOG(ERR, "invalid src_ip mask.");
6637                 return -EINVAL;
6638         }
6639
6640         switch (filter->dst_port_mask) {
6641         case UINT16_MAX:
6642                 filter_info->dst_port_mask = 0;
6643                 filter_info->dst_port = filter->dst_port;
6644                 break;
6645         case 0:
6646                 filter_info->dst_port_mask = 1;
6647                 break;
6648         default:
6649                 PMD_DRV_LOG(ERR, "invalid dst_port mask.");
6650                 return -EINVAL;
6651         }
6652
6653         switch (filter->src_port_mask) {
6654         case UINT16_MAX:
6655                 filter_info->src_port_mask = 0;
6656                 filter_info->src_port = filter->src_port;
6657                 break;
6658         case 0:
6659                 filter_info->src_port_mask = 1;
6660                 break;
6661         default:
6662                 PMD_DRV_LOG(ERR, "invalid src_port mask.");
6663                 return -EINVAL;
6664         }
6665
6666         switch (filter->proto_mask) {
6667         case UINT8_MAX:
6668                 filter_info->proto_mask = 0;
6669                 filter_info->proto =
6670                         convert_protocol_type(filter->proto);
6671                 break;
6672         case 0:
6673                 filter_info->proto_mask = 1;
6674                 break;
6675         default:
6676                 PMD_DRV_LOG(ERR, "invalid protocol mask.");
6677                 return -EINVAL;
6678         }
6679
6680         filter_info->priority = (uint8_t)filter->priority;
6681         return 0;
6682 }
6683
6684 /*
6685  * add or delete a ntuple filter
6686  *
6687  * @param
6688  * dev: Pointer to struct rte_eth_dev.
6689  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6690  * add: if true, add filter, if false, remove filter
6691  *
6692  * @return
6693  *    - On success, zero.
6694  *    - On failure, a negative value.
6695  */
6696 int
6697 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
6698                         struct rte_eth_ntuple_filter *ntuple_filter,
6699                         bool add)
6700 {
6701         struct ixgbe_filter_info *filter_info =
6702                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6703         struct ixgbe_5tuple_filter_info filter_5tuple;
6704         struct ixgbe_5tuple_filter *filter;
6705         int ret;
6706
6707         if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6708                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6709                 return -EINVAL;
6710         }
6711
6712         memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6713         ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6714         if (ret < 0)
6715                 return ret;
6716
6717         filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6718                                          &filter_5tuple);
6719         if (filter != NULL && add) {
6720                 PMD_DRV_LOG(ERR, "filter exists.");
6721                 return -EEXIST;
6722         }
6723         if (filter == NULL && !add) {
6724                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
6725                 return -ENOENT;
6726         }
6727
6728         if (add) {
6729                 filter = rte_zmalloc("ixgbe_5tuple_filter",
6730                                 sizeof(struct ixgbe_5tuple_filter), 0);
6731                 if (filter == NULL)
6732                         return -ENOMEM;
6733                 rte_memcpy(&filter->filter_info,
6734                                  &filter_5tuple,
6735                                  sizeof(struct ixgbe_5tuple_filter_info));
6736                 filter->queue = ntuple_filter->queue;
6737                 ret = ixgbe_add_5tuple_filter(dev, filter);
6738                 if (ret < 0) {
6739                         rte_free(filter);
6740                         return ret;
6741                 }
6742         } else
6743                 ixgbe_remove_5tuple_filter(dev, filter);
6744
6745         return 0;
6746 }
6747
6748 int
6749 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
6750                         struct rte_eth_ethertype_filter *filter,
6751                         bool add)
6752 {
6753         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6754         struct ixgbe_filter_info *filter_info =
6755                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6756         uint32_t etqf = 0;
6757         uint32_t etqs = 0;
6758         int ret;
6759         struct ixgbe_ethertype_filter ethertype_filter;
6760
6761         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6762                 return -EINVAL;
6763
6764         if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
6765                 filter->ether_type == RTE_ETHER_TYPE_IPV6) {
6766                 PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
6767                         " ethertype filter.", filter->ether_type);
6768                 return -EINVAL;
6769         }
6770
6771         if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
6772                 PMD_DRV_LOG(ERR, "mac compare is unsupported.");
6773                 return -EINVAL;
6774         }
6775         if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
6776                 PMD_DRV_LOG(ERR, "drop option is unsupported.");
6777                 return -EINVAL;
6778         }
6779
6780         ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6781         if (ret >= 0 && add) {
6782                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
6783                             filter->ether_type);
6784                 return -EEXIST;
6785         }
6786         if (ret < 0 && !add) {
6787                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6788                             filter->ether_type);
6789                 return -ENOENT;
6790         }
6791
6792         if (add) {
6793                 etqf = IXGBE_ETQF_FILTER_EN;
6794                 etqf |= (uint32_t)filter->ether_type;
6795                 etqs |= (uint32_t)((filter->queue <<
6796                                     IXGBE_ETQS_RX_QUEUE_SHIFT) &
6797                                     IXGBE_ETQS_RX_QUEUE);
6798                 etqs |= IXGBE_ETQS_QUEUE_EN;
6799
6800                 ethertype_filter.ethertype = filter->ether_type;
6801                 ethertype_filter.etqf = etqf;
6802                 ethertype_filter.etqs = etqs;
6803                 ethertype_filter.conf = FALSE;
6804                 ret = ixgbe_ethertype_filter_insert(filter_info,
6805                                                     &ethertype_filter);
6806                 if (ret < 0) {
6807                         PMD_DRV_LOG(ERR, "ethertype filters are full.");
6808                         return -ENOSPC;
6809                 }
6810         } else {
6811                 ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
6812                 if (ret < 0)
6813                         return -ENOSYS;
6814         }
6815         IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
6816         IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
6817         IXGBE_WRITE_FLUSH(hw);
6818
6819         return 0;
6820 }
6821
6822 static int
6823 ixgbe_dev_flow_ops_get(__rte_unused struct rte_eth_dev *dev,
6824                        const struct rte_flow_ops **ops)
6825 {
6826         *ops = &ixgbe_flow_ops;
6827         return 0;
6828 }
6829
6830 static u8 *
6831 ixgbe_dev_addr_list_itr(__rte_unused struct ixgbe_hw *hw,
6832                         u8 **mc_addr_ptr, u32 *vmdq)
6833 {
6834         u8 *mc_addr;
6835
6836         *vmdq = 0;
6837         mc_addr = *mc_addr_ptr;
6838         *mc_addr_ptr = (mc_addr + sizeof(struct rte_ether_addr));
6839         return mc_addr;
6840 }
6841
6842 static int
6843 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
6844                           struct rte_ether_addr *mc_addr_set,
6845                           uint32_t nb_mc_addr)
6846 {
6847         struct ixgbe_hw *hw;
6848         u8 *mc_addr_list;
6849
6850         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6851         mc_addr_list = (u8 *)mc_addr_set;
6852         return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
6853                                          ixgbe_dev_addr_list_itr, TRUE);
6854 }
6855
6856 static uint64_t
6857 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
6858 {
6859         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6860         uint64_t systime_cycles;
6861
6862         switch (hw->mac.type) {
6863         case ixgbe_mac_X550:
6864         case ixgbe_mac_X550EM_x:
6865         case ixgbe_mac_X550EM_a:
6866                 /* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
6867                 systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6868                 systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6869                                 * NSEC_PER_SEC;
6870                 break;
6871         default:
6872                 systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6873                 systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6874                                 << 32;
6875         }
6876
6877         return systime_cycles;
6878 }
6879
6880 static uint64_t
6881 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6882 {
6883         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6884         uint64_t rx_tstamp_cycles;
6885
6886         switch (hw->mac.type) {
6887         case ixgbe_mac_X550:
6888         case ixgbe_mac_X550EM_x:
6889         case ixgbe_mac_X550EM_a:
6890                 /* RXSTMPL stores ns and RXSTMPH stores seconds. */
6891                 rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6892                 rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6893                                 * NSEC_PER_SEC;
6894                 break;
6895         default:
6896                 /* RXSTMPL stores ns and RXSTMPH stores seconds. */
6897                 rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6898                 rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6899                                 << 32;
6900         }
6901
6902         return rx_tstamp_cycles;
6903 }
6904
6905 static uint64_t
6906 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6907 {
6908         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6909         uint64_t tx_tstamp_cycles;
6910
6911         switch (hw->mac.type) {
6912         case ixgbe_mac_X550:
6913         case ixgbe_mac_X550EM_x:
6914         case ixgbe_mac_X550EM_a:
6915                 /* TXSTMPL stores ns and TXSTMPH stores seconds. */
6916                 tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6917                 tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6918                                 * NSEC_PER_SEC;
6919                 break;
6920         default:
6921                 /* TXSTMPL stores ns and TXSTMPH stores seconds. */
6922                 tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6923                 tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6924                                 << 32;
6925         }
6926
6927         return tx_tstamp_cycles;
6928 }
6929
6930 static void
6931 ixgbe_start_timecounters(struct rte_eth_dev *dev)
6932 {
6933         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6934         struct ixgbe_adapter *adapter = dev->data->dev_private;
6935         struct rte_eth_link link;
6936         uint32_t incval = 0;
6937         uint32_t shift = 0;
6938
6939         /* Get current link speed. */
6940         ixgbe_dev_link_update(dev, 1);
6941         rte_eth_linkstatus_get(dev, &link);
6942
6943         switch (link.link_speed) {
6944         case ETH_SPEED_NUM_100M:
6945                 incval = IXGBE_INCVAL_100;
6946                 shift = IXGBE_INCVAL_SHIFT_100;
6947                 break;
6948         case ETH_SPEED_NUM_1G:
6949                 incval = IXGBE_INCVAL_1GB;
6950                 shift = IXGBE_INCVAL_SHIFT_1GB;
6951                 break;
6952         case ETH_SPEED_NUM_10G:
6953         default:
6954                 incval = IXGBE_INCVAL_10GB;
6955                 shift = IXGBE_INCVAL_SHIFT_10GB;
6956                 break;
6957         }
6958
6959         switch (hw->mac.type) {
6960         case ixgbe_mac_X550:
6961         case ixgbe_mac_X550EM_x:
6962         case ixgbe_mac_X550EM_a:
6963                 /* Independent of link speed. */
6964                 incval = 1;
6965                 /* Cycles read will be interpreted as ns. */
6966                 shift = 0;
6967                 /* Fall-through */
6968         case ixgbe_mac_X540:
6969                 IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
6970                 break;
6971         case ixgbe_mac_82599EB:
6972                 incval >>= IXGBE_INCVAL_SHIFT_82599;
6973                 shift -= IXGBE_INCVAL_SHIFT_82599;
6974                 IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
6975                                 (1 << IXGBE_INCPER_SHIFT_82599) | incval);
6976                 break;
6977         default:
6978                 /* Not supported. */
6979                 return;
6980         }
6981
6982         memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
6983         memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6984         memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6985
6986         adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6987         adapter->systime_tc.cc_shift = shift;
6988         adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
6989
6990         adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6991         adapter->rx_tstamp_tc.cc_shift = shift;
6992         adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6993
6994         adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6995         adapter->tx_tstamp_tc.cc_shift = shift;
6996         adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6997 }
6998
6999 static int
7000 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
7001 {
7002         struct ixgbe_adapter *adapter = dev->data->dev_private;
7003
7004         adapter->systime_tc.nsec += delta;
7005         adapter->rx_tstamp_tc.nsec += delta;
7006         adapter->tx_tstamp_tc.nsec += delta;
7007
7008         return 0;
7009 }
7010
7011 static int
7012 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
7013 {
7014         uint64_t ns;
7015         struct ixgbe_adapter *adapter = dev->data->dev_private;
7016
7017         ns = rte_timespec_to_ns(ts);
7018         /* Set the timecounters to a new value. */
7019         adapter->systime_tc.nsec = ns;
7020         adapter->rx_tstamp_tc.nsec = ns;
7021         adapter->tx_tstamp_tc.nsec = ns;
7022
7023         return 0;
7024 }
7025
7026 static int
7027 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
7028 {
7029         uint64_t ns, systime_cycles;
7030         struct ixgbe_adapter *adapter = dev->data->dev_private;
7031
7032         systime_cycles = ixgbe_read_systime_cyclecounter(dev);
7033         ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
7034         *ts = rte_ns_to_timespec(ns);
7035
7036         return 0;
7037 }
7038
7039 static int
7040 ixgbe_timesync_enable(struct rte_eth_dev *dev)
7041 {
7042         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7043         uint32_t tsync_ctl;
7044         uint32_t tsauxc;
7045
7046         /* Stop the timesync system time. */
7047         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
7048         /* Reset the timesync system time value. */
7049         IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
7050         IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
7051
7052         /* Enable system time for platforms where it isn't on by default. */
7053         tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
7054         tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
7055         IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
7056
7057         ixgbe_start_timecounters(dev);
7058
7059         /* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
7060         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
7061                         (RTE_ETHER_TYPE_1588 |
7062                          IXGBE_ETQF_FILTER_EN |
7063                          IXGBE_ETQF_1588));
7064
7065         /* Enable timestamping of received PTP packets. */
7066         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7067         tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
7068         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
7069
7070         /* Enable timestamping of transmitted PTP packets. */
7071         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7072         tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
7073         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
7074
7075         IXGBE_WRITE_FLUSH(hw);
7076
7077         return 0;
7078 }
7079
7080 static int
7081 ixgbe_timesync_disable(struct rte_eth_dev *dev)
7082 {
7083         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7084         uint32_t tsync_ctl;
7085
7086         /* Disable timestamping of transmitted PTP packets. */
7087         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7088         tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
7089         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
7090
7091         /* Disable timestamping of received PTP packets. */
7092         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7093         tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
7094         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
7095
7096         /* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
7097         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
7098
7099         /* Stop incrementating the System Time registers. */
7100         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
7101
7102         return 0;
7103 }
7104
7105 static int
7106 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
7107                                  struct timespec *timestamp,
7108                                  uint32_t flags __rte_unused)
7109 {
7110         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7111         struct ixgbe_adapter *adapter = dev->data->dev_private;
7112         uint32_t tsync_rxctl;
7113         uint64_t rx_tstamp_cycles;
7114         uint64_t ns;
7115
7116         tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7117         if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
7118                 return -EINVAL;
7119
7120         rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
7121         ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
7122         *timestamp = rte_ns_to_timespec(ns);
7123
7124         return  0;
7125 }
7126
7127 static int
7128 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
7129                                  struct timespec *timestamp)
7130 {
7131         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7132         struct ixgbe_adapter *adapter = dev->data->dev_private;
7133         uint32_t tsync_txctl;
7134         uint64_t tx_tstamp_cycles;
7135         uint64_t ns;
7136
7137         tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7138         if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
7139                 return -EINVAL;
7140
7141         tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
7142         ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
7143         *timestamp = rte_ns_to_timespec(ns);
7144
7145         return 0;
7146 }
7147
7148 static int
7149 ixgbe_get_reg_length(struct rte_eth_dev *dev)
7150 {
7151         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7152         int count = 0;
7153         int g_ind = 0;
7154         const struct reg_info *reg_group;
7155         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7156                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7157
7158         while ((reg_group = reg_set[g_ind++]))
7159                 count += ixgbe_regs_group_count(reg_group);
7160
7161         return count;
7162 }
7163
7164 static int
7165 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
7166 {
7167         int count = 0;
7168         int g_ind = 0;
7169         const struct reg_info *reg_group;
7170
7171         while ((reg_group = ixgbevf_regs[g_ind++]))
7172                 count += ixgbe_regs_group_count(reg_group);
7173
7174         return count;
7175 }
7176
7177 static int
7178 ixgbe_get_regs(struct rte_eth_dev *dev,
7179               struct rte_dev_reg_info *regs)
7180 {
7181         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7182         uint32_t *data = regs->data;
7183         int g_ind = 0;
7184         int count = 0;
7185         const struct reg_info *reg_group;
7186         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7187                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7188
7189         if (data == NULL) {
7190                 regs->length = ixgbe_get_reg_length(dev);
7191                 regs->width = sizeof(uint32_t);
7192                 return 0;
7193         }
7194
7195         /* Support only full register dump */
7196         if ((regs->length == 0) ||
7197             (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
7198                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7199                         hw->device_id;
7200                 while ((reg_group = reg_set[g_ind++]))
7201                         count += ixgbe_read_regs_group(dev, &data[count],
7202                                 reg_group);
7203                 return 0;
7204         }
7205
7206         return -ENOTSUP;
7207 }
7208
7209 static int
7210 ixgbevf_get_regs(struct rte_eth_dev *dev,
7211                 struct rte_dev_reg_info *regs)
7212 {
7213         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7214         uint32_t *data = regs->data;
7215         int g_ind = 0;
7216         int count = 0;
7217         const struct reg_info *reg_group;
7218
7219         if (data == NULL) {
7220                 regs->length = ixgbevf_get_reg_length(dev);
7221                 regs->width = sizeof(uint32_t);
7222                 return 0;
7223         }
7224
7225         /* Support only full register dump */
7226         if ((regs->length == 0) ||
7227             (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
7228                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7229                         hw->device_id;
7230                 while ((reg_group = ixgbevf_regs[g_ind++]))
7231                         count += ixgbe_read_regs_group(dev, &data[count],
7232                                                       reg_group);
7233                 return 0;
7234         }
7235
7236         return -ENOTSUP;
7237 }
7238
7239 static int
7240 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
7241 {
7242         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7243
7244         /* Return unit is byte count */
7245         return hw->eeprom.word_size * 2;
7246 }
7247
7248 static int
7249 ixgbe_get_eeprom(struct rte_eth_dev *dev,
7250                 struct rte_dev_eeprom_info *in_eeprom)
7251 {
7252         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7253         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7254         uint16_t *data = in_eeprom->data;
7255         int first, length;
7256
7257         first = in_eeprom->offset >> 1;
7258         length = in_eeprom->length >> 1;
7259         if ((first > hw->eeprom.word_size) ||
7260             ((first + length) > hw->eeprom.word_size))
7261                 return -EINVAL;
7262
7263         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7264
7265         return eeprom->ops.read_buffer(hw, first, length, data);
7266 }
7267
7268 static int
7269 ixgbe_set_eeprom(struct rte_eth_dev *dev,
7270                 struct rte_dev_eeprom_info *in_eeprom)
7271 {
7272         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7273         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7274         uint16_t *data = in_eeprom->data;
7275         int first, length;
7276
7277         first = in_eeprom->offset >> 1;
7278         length = in_eeprom->length >> 1;
7279         if ((first > hw->eeprom.word_size) ||
7280             ((first + length) > hw->eeprom.word_size))
7281                 return -EINVAL;
7282
7283         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7284
7285         return eeprom->ops.write_buffer(hw,  first, length, data);
7286 }
7287
7288 static int
7289 ixgbe_get_module_info(struct rte_eth_dev *dev,
7290                       struct rte_eth_dev_module_info *modinfo)
7291 {
7292         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7293         uint32_t status;
7294         uint8_t sff8472_rev, addr_mode;
7295         bool page_swap = false;
7296
7297         /* Check whether we support SFF-8472 or not */
7298         status = hw->phy.ops.read_i2c_eeprom(hw,
7299                                              IXGBE_SFF_SFF_8472_COMP,
7300                                              &sff8472_rev);
7301         if (status != 0)
7302                 return -EIO;
7303
7304         /* addressing mode is not supported */
7305         status = hw->phy.ops.read_i2c_eeprom(hw,
7306                                              IXGBE_SFF_SFF_8472_SWAP,
7307                                              &addr_mode);
7308         if (status != 0)
7309                 return -EIO;
7310
7311         if (addr_mode & IXGBE_SFF_ADDRESSING_MODE) {
7312                 PMD_DRV_LOG(ERR,
7313                             "Address change required to access page 0xA2, "
7314                             "but not supported. Please report the module "
7315                             "type to the driver maintainers.");
7316                 page_swap = true;
7317         }
7318
7319         if (sff8472_rev == IXGBE_SFF_SFF_8472_UNSUP || page_swap) {
7320                 /* We have a SFP, but it does not support SFF-8472 */
7321                 modinfo->type = RTE_ETH_MODULE_SFF_8079;
7322                 modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
7323         } else {
7324                 /* We have a SFP which supports a revision of SFF-8472. */
7325                 modinfo->type = RTE_ETH_MODULE_SFF_8472;
7326                 modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
7327         }
7328
7329         return 0;
7330 }
7331
7332 static int
7333 ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
7334                         struct rte_dev_eeprom_info *info)
7335 {
7336         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7337         uint32_t status = IXGBE_ERR_PHY_ADDR_INVALID;
7338         uint8_t databyte = 0xFF;
7339         uint8_t *data = info->data;
7340         uint32_t i = 0;
7341
7342         for (i = info->offset; i < info->offset + info->length; i++) {
7343                 if (i < RTE_ETH_MODULE_SFF_8079_LEN)
7344                         status = hw->phy.ops.read_i2c_eeprom(hw, i, &databyte);
7345                 else
7346                         status = hw->phy.ops.read_i2c_sff8472(hw, i, &databyte);
7347
7348                 if (status != 0)
7349                         return -EIO;
7350
7351                 data[i - info->offset] = databyte;
7352         }
7353
7354         return 0;
7355 }
7356
7357 uint16_t
7358 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
7359         switch (mac_type) {
7360         case ixgbe_mac_X550:
7361         case ixgbe_mac_X550EM_x:
7362         case ixgbe_mac_X550EM_a:
7363                 return ETH_RSS_RETA_SIZE_512;
7364         case ixgbe_mac_X550_vf:
7365         case ixgbe_mac_X550EM_x_vf:
7366         case ixgbe_mac_X550EM_a_vf:
7367                 return ETH_RSS_RETA_SIZE_64;
7368         case ixgbe_mac_X540_vf:
7369         case ixgbe_mac_82599_vf:
7370                 return 0;
7371         default:
7372                 return ETH_RSS_RETA_SIZE_128;
7373         }
7374 }
7375
7376 uint32_t
7377 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
7378         switch (mac_type) {
7379         case ixgbe_mac_X550:
7380         case ixgbe_mac_X550EM_x:
7381         case ixgbe_mac_X550EM_a:
7382                 if (reta_idx < ETH_RSS_RETA_SIZE_128)
7383                         return IXGBE_RETA(reta_idx >> 2);
7384                 else
7385                         return IXGBE_ERETA((reta_idx - ETH_RSS_RETA_SIZE_128) >> 2);
7386         case ixgbe_mac_X550_vf:
7387         case ixgbe_mac_X550EM_x_vf:
7388         case ixgbe_mac_X550EM_a_vf:
7389                 return IXGBE_VFRETA(reta_idx >> 2);
7390         default:
7391                 return IXGBE_RETA(reta_idx >> 2);
7392         }
7393 }
7394
7395 uint32_t
7396 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
7397         switch (mac_type) {
7398         case ixgbe_mac_X550_vf:
7399         case ixgbe_mac_X550EM_x_vf:
7400         case ixgbe_mac_X550EM_a_vf:
7401                 return IXGBE_VFMRQC;
7402         default:
7403                 return IXGBE_MRQC;
7404         }
7405 }
7406
7407 uint32_t
7408 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
7409         switch (mac_type) {
7410         case ixgbe_mac_X550_vf:
7411         case ixgbe_mac_X550EM_x_vf:
7412         case ixgbe_mac_X550EM_a_vf:
7413                 return IXGBE_VFRSSRK(i);
7414         default:
7415                 return IXGBE_RSSRK(i);
7416         }
7417 }
7418
7419 bool
7420 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
7421         switch (mac_type) {
7422         case ixgbe_mac_82599_vf:
7423         case ixgbe_mac_X540_vf:
7424                 return 0;
7425         default:
7426                 return 1;
7427         }
7428 }
7429
7430 static int
7431 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
7432                         struct rte_eth_dcb_info *dcb_info)
7433 {
7434         struct ixgbe_dcb_config *dcb_config =
7435                         IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
7436         struct ixgbe_dcb_tc_config *tc;
7437         struct rte_eth_dcb_tc_queue_mapping *tc_queue;
7438         uint8_t nb_tcs;
7439         uint8_t i, j;
7440
7441         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_DCB_FLAG)
7442                 dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
7443         else
7444                 dcb_info->nb_tcs = 1;
7445
7446         tc_queue = &dcb_info->tc_queue;
7447         nb_tcs = dcb_info->nb_tcs;
7448
7449         if (dcb_config->vt_mode) { /* vt is enabled*/
7450                 struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
7451                                 &dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
7452                 for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7453                         dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
7454                 if (RTE_ETH_DEV_SRIOV(dev).active > 0) {
7455                         for (j = 0; j < nb_tcs; j++) {
7456                                 tc_queue->tc_rxq[0][j].base = j;
7457                                 tc_queue->tc_rxq[0][j].nb_queue = 1;
7458                                 tc_queue->tc_txq[0][j].base = j;
7459                                 tc_queue->tc_txq[0][j].nb_queue = 1;
7460                         }
7461                 } else {
7462                         for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
7463                                 for (j = 0; j < nb_tcs; j++) {
7464                                         tc_queue->tc_rxq[i][j].base =
7465                                                 i * nb_tcs + j;
7466                                         tc_queue->tc_rxq[i][j].nb_queue = 1;
7467                                         tc_queue->tc_txq[i][j].base =
7468                                                 i * nb_tcs + j;
7469                                         tc_queue->tc_txq[i][j].nb_queue = 1;
7470                                 }
7471                         }
7472                 }
7473         } else { /* vt is disabled*/
7474                 struct rte_eth_dcb_rx_conf *rx_conf =
7475                                 &dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
7476                 for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7477                         dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
7478                 if (dcb_info->nb_tcs == ETH_4_TCS) {
7479                         for (i = 0; i < dcb_info->nb_tcs; i++) {
7480                                 dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
7481                                 dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7482                         }
7483                         dcb_info->tc_queue.tc_txq[0][0].base = 0;
7484                         dcb_info->tc_queue.tc_txq[0][1].base = 64;
7485                         dcb_info->tc_queue.tc_txq[0][2].base = 96;
7486                         dcb_info->tc_queue.tc_txq[0][3].base = 112;
7487                         dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
7488                         dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7489                         dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7490                         dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7491                 } else if (dcb_info->nb_tcs == ETH_8_TCS) {
7492                         for (i = 0; i < dcb_info->nb_tcs; i++) {
7493                                 dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
7494                                 dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7495                         }
7496                         dcb_info->tc_queue.tc_txq[0][0].base = 0;
7497                         dcb_info->tc_queue.tc_txq[0][1].base = 32;
7498                         dcb_info->tc_queue.tc_txq[0][2].base = 64;
7499                         dcb_info->tc_queue.tc_txq[0][3].base = 80;
7500                         dcb_info->tc_queue.tc_txq[0][4].base = 96;
7501                         dcb_info->tc_queue.tc_txq[0][5].base = 104;
7502                         dcb_info->tc_queue.tc_txq[0][6].base = 112;
7503                         dcb_info->tc_queue.tc_txq[0][7].base = 120;
7504                         dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
7505                         dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7506                         dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7507                         dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7508                         dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
7509                         dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
7510                         dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
7511                         dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
7512                 }
7513         }
7514         for (i = 0; i < dcb_info->nb_tcs; i++) {
7515                 tc = &dcb_config->tc_config[i];
7516                 dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
7517         }
7518         return 0;
7519 }
7520
7521 /* Update e-tag ether type */
7522 static int
7523 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
7524                             uint16_t ether_type)
7525 {
7526         uint32_t etag_etype;
7527
7528         if (hw->mac.type != ixgbe_mac_X550 &&
7529             hw->mac.type != ixgbe_mac_X550EM_x &&
7530             hw->mac.type != ixgbe_mac_X550EM_a) {
7531                 return -ENOTSUP;
7532         }
7533
7534         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7535         etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
7536         etag_etype |= ether_type;
7537         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7538         IXGBE_WRITE_FLUSH(hw);
7539
7540         return 0;
7541 }
7542
7543 /* Enable e-tag tunnel */
7544 static int
7545 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
7546 {
7547         uint32_t etag_etype;
7548
7549         if (hw->mac.type != ixgbe_mac_X550 &&
7550             hw->mac.type != ixgbe_mac_X550EM_x &&
7551             hw->mac.type != ixgbe_mac_X550EM_a) {
7552                 return -ENOTSUP;
7553         }
7554
7555         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7556         etag_etype |= IXGBE_ETAG_ETYPE_VALID;
7557         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7558         IXGBE_WRITE_FLUSH(hw);
7559
7560         return 0;
7561 }
7562
7563 static int
7564 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
7565                        struct ixgbe_l2_tunnel_conf *l2_tunnel)
7566 {
7567         int ret = 0;
7568         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7569         uint32_t i, rar_entries;
7570         uint32_t rar_low, rar_high;
7571
7572         if (hw->mac.type != ixgbe_mac_X550 &&
7573             hw->mac.type != ixgbe_mac_X550EM_x &&
7574             hw->mac.type != ixgbe_mac_X550EM_a) {
7575                 return -ENOTSUP;
7576         }
7577
7578         rar_entries = ixgbe_get_num_rx_addrs(hw);
7579
7580         for (i = 1; i < rar_entries; i++) {
7581                 rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7582                 rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
7583                 if ((rar_high & IXGBE_RAH_AV) &&
7584                     (rar_high & IXGBE_RAH_ADTYPE) &&
7585                     ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
7586                      l2_tunnel->tunnel_id)) {
7587                         IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
7588                         IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
7589
7590                         ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
7591
7592                         return ret;
7593                 }
7594         }
7595
7596         return ret;
7597 }
7598
7599 static int
7600 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
7601                        struct ixgbe_l2_tunnel_conf *l2_tunnel)
7602 {
7603         int ret = 0;
7604         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7605         uint32_t i, rar_entries;
7606         uint32_t rar_low, rar_high;
7607
7608         if (hw->mac.type != ixgbe_mac_X550 &&
7609             hw->mac.type != ixgbe_mac_X550EM_x &&
7610             hw->mac.type != ixgbe_mac_X550EM_a) {
7611                 return -ENOTSUP;
7612         }
7613
7614         /* One entry for one tunnel. Try to remove potential existing entry. */
7615         ixgbe_e_tag_filter_del(dev, l2_tunnel);
7616
7617         rar_entries = ixgbe_get_num_rx_addrs(hw);
7618
7619         for (i = 1; i < rar_entries; i++) {
7620                 rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7621                 if (rar_high & IXGBE_RAH_AV) {
7622                         continue;
7623                 } else {
7624                         ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
7625                         rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
7626                         rar_low = l2_tunnel->tunnel_id;
7627
7628                         IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
7629                         IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
7630
7631                         return ret;
7632                 }
7633         }
7634
7635         PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
7636                      " Please remove a rule before adding a new one.");
7637         return -EINVAL;
7638 }
7639
7640 static inline struct ixgbe_l2_tn_filter *
7641 ixgbe_l2_tn_filter_lookup(struct ixgbe_l2_tn_info *l2_tn_info,
7642                           struct ixgbe_l2_tn_key *key)
7643 {
7644         int ret;
7645
7646         ret = rte_hash_lookup(l2_tn_info->hash_handle, (const void *)key);
7647         if (ret < 0)
7648                 return NULL;
7649
7650         return l2_tn_info->hash_map[ret];
7651 }
7652
7653 static inline int
7654 ixgbe_insert_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7655                           struct ixgbe_l2_tn_filter *l2_tn_filter)
7656 {
7657         int ret;
7658
7659         ret = rte_hash_add_key(l2_tn_info->hash_handle,
7660                                &l2_tn_filter->key);
7661
7662         if (ret < 0) {
7663                 PMD_DRV_LOG(ERR,
7664                             "Failed to insert L2 tunnel filter"
7665                             " to hash table %d!",
7666                             ret);
7667                 return ret;
7668         }
7669
7670         l2_tn_info->hash_map[ret] = l2_tn_filter;
7671
7672         TAILQ_INSERT_TAIL(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7673
7674         return 0;
7675 }
7676
7677 static inline int
7678 ixgbe_remove_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
7679                           struct ixgbe_l2_tn_key *key)
7680 {
7681         int ret;
7682         struct ixgbe_l2_tn_filter *l2_tn_filter;
7683
7684         ret = rte_hash_del_key(l2_tn_info->hash_handle, key);
7685
7686         if (ret < 0) {
7687                 PMD_DRV_LOG(ERR,
7688                             "No such L2 tunnel filter to delete %d!",
7689                             ret);
7690                 return ret;
7691         }
7692
7693         l2_tn_filter = l2_tn_info->hash_map[ret];
7694         l2_tn_info->hash_map[ret] = NULL;
7695
7696         TAILQ_REMOVE(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
7697         rte_free(l2_tn_filter);
7698
7699         return 0;
7700 }
7701
7702 /* Add l2 tunnel filter */
7703 int
7704 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
7705                                struct ixgbe_l2_tunnel_conf *l2_tunnel,
7706                                bool restore)
7707 {
7708         int ret;
7709         struct ixgbe_l2_tn_info *l2_tn_info =
7710                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7711         struct ixgbe_l2_tn_key key;
7712         struct ixgbe_l2_tn_filter *node;
7713
7714         if (!restore) {
7715                 key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7716                 key.tn_id = l2_tunnel->tunnel_id;
7717
7718                 node = ixgbe_l2_tn_filter_lookup(l2_tn_info, &key);
7719
7720                 if (node) {
7721                         PMD_DRV_LOG(ERR,
7722                                     "The L2 tunnel filter already exists!");
7723                         return -EINVAL;
7724                 }
7725
7726                 node = rte_zmalloc("ixgbe_l2_tn",
7727                                    sizeof(struct ixgbe_l2_tn_filter),
7728                                    0);
7729                 if (!node)
7730                         return -ENOMEM;
7731
7732                 rte_memcpy(&node->key,
7733                                  &key,
7734                                  sizeof(struct ixgbe_l2_tn_key));
7735                 node->pool = l2_tunnel->pool;
7736                 ret = ixgbe_insert_l2_tn_filter(l2_tn_info, node);
7737                 if (ret < 0) {
7738                         rte_free(node);
7739                         return ret;
7740                 }
7741         }
7742
7743         switch (l2_tunnel->l2_tunnel_type) {
7744         case RTE_L2_TUNNEL_TYPE_E_TAG:
7745                 ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
7746                 break;
7747         default:
7748                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7749                 ret = -EINVAL;
7750                 break;
7751         }
7752
7753         if ((!restore) && (ret < 0))
7754                 (void)ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7755
7756         return ret;
7757 }
7758
7759 /* Delete l2 tunnel filter */
7760 int
7761 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
7762                                struct ixgbe_l2_tunnel_conf *l2_tunnel)
7763 {
7764         int ret;
7765         struct ixgbe_l2_tn_info *l2_tn_info =
7766                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7767         struct ixgbe_l2_tn_key key;
7768
7769         key.l2_tn_type = l2_tunnel->l2_tunnel_type;
7770         key.tn_id = l2_tunnel->tunnel_id;
7771         ret = ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
7772         if (ret < 0)
7773                 return ret;
7774
7775         switch (l2_tunnel->l2_tunnel_type) {
7776         case RTE_L2_TUNNEL_TYPE_E_TAG:
7777                 ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
7778                 break;
7779         default:
7780                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7781                 ret = -EINVAL;
7782                 break;
7783         }
7784
7785         return ret;
7786 }
7787
7788 static int
7789 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
7790 {
7791         int ret = 0;
7792         uint32_t ctrl;
7793         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7794
7795         if (hw->mac.type != ixgbe_mac_X550 &&
7796             hw->mac.type != ixgbe_mac_X550EM_x &&
7797             hw->mac.type != ixgbe_mac_X550EM_a) {
7798                 return -ENOTSUP;
7799         }
7800
7801         ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
7802         ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
7803         if (en)
7804                 ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
7805         IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
7806
7807         return ret;
7808 }
7809
7810 static int
7811 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
7812                         uint16_t port)
7813 {
7814         IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
7815         IXGBE_WRITE_FLUSH(hw);
7816
7817         return 0;
7818 }
7819
7820 /* There's only one register for VxLAN UDP port.
7821  * So, we cannot add several ports. Will update it.
7822  */
7823 static int
7824 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
7825                      uint16_t port)
7826 {
7827         if (port == 0) {
7828                 PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
7829                 return -EINVAL;
7830         }
7831
7832         return ixgbe_update_vxlan_port(hw, port);
7833 }
7834
7835 /* We cannot delete the VxLAN port. For there's a register for VxLAN
7836  * UDP port, it must have a value.
7837  * So, will reset it to the original value 0.
7838  */
7839 static int
7840 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
7841                      uint16_t port)
7842 {
7843         uint16_t cur_port;
7844
7845         cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
7846
7847         if (cur_port != port) {
7848                 PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
7849                 return -EINVAL;
7850         }
7851
7852         return ixgbe_update_vxlan_port(hw, 0);
7853 }
7854
7855 /* Add UDP tunneling port */
7856 static int
7857 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
7858                               struct rte_eth_udp_tunnel *udp_tunnel)
7859 {
7860         int ret = 0;
7861         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7862
7863         if (hw->mac.type != ixgbe_mac_X550 &&
7864             hw->mac.type != ixgbe_mac_X550EM_x &&
7865             hw->mac.type != ixgbe_mac_X550EM_a) {
7866                 return -ENOTSUP;
7867         }
7868
7869         if (udp_tunnel == NULL)
7870                 return -EINVAL;
7871
7872         switch (udp_tunnel->prot_type) {
7873         case RTE_TUNNEL_TYPE_VXLAN:
7874                 ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
7875                 break;
7876
7877         case RTE_TUNNEL_TYPE_GENEVE:
7878         case RTE_TUNNEL_TYPE_TEREDO:
7879                 PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7880                 ret = -EINVAL;
7881                 break;
7882
7883         default:
7884                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7885                 ret = -EINVAL;
7886                 break;
7887         }
7888
7889         return ret;
7890 }
7891
7892 /* Remove UDP tunneling port */
7893 static int
7894 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
7895                               struct rte_eth_udp_tunnel *udp_tunnel)
7896 {
7897         int ret = 0;
7898         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7899
7900         if (hw->mac.type != ixgbe_mac_X550 &&
7901             hw->mac.type != ixgbe_mac_X550EM_x &&
7902             hw->mac.type != ixgbe_mac_X550EM_a) {
7903                 return -ENOTSUP;
7904         }
7905
7906         if (udp_tunnel == NULL)
7907                 return -EINVAL;
7908
7909         switch (udp_tunnel->prot_type) {
7910         case RTE_TUNNEL_TYPE_VXLAN:
7911                 ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
7912                 break;
7913         case RTE_TUNNEL_TYPE_GENEVE:
7914         case RTE_TUNNEL_TYPE_TEREDO:
7915                 PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7916                 ret = -EINVAL;
7917                 break;
7918         default:
7919                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7920                 ret = -EINVAL;
7921                 break;
7922         }
7923
7924         return ret;
7925 }
7926
7927 static int
7928 ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev)
7929 {
7930         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7931         int ret;
7932
7933         switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_PROMISC)) {
7934         case IXGBE_SUCCESS:
7935                 ret = 0;
7936                 break;
7937         case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7938                 ret = -ENOTSUP;
7939                 break;
7940         default:
7941                 ret = -EAGAIN;
7942                 break;
7943         }
7944
7945         return ret;
7946 }
7947
7948 static int
7949 ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev)
7950 {
7951         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7952         int ret;
7953
7954         switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_NONE)) {
7955         case IXGBE_SUCCESS:
7956                 ret = 0;
7957                 break;
7958         case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7959                 ret = -ENOTSUP;
7960                 break;
7961         default:
7962                 ret = -EAGAIN;
7963                 break;
7964         }
7965
7966         return ret;
7967 }
7968
7969 static int
7970 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
7971 {
7972         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7973         int ret;
7974         int mode = IXGBEVF_XCAST_MODE_ALLMULTI;
7975
7976         switch (hw->mac.ops.update_xcast_mode(hw, mode)) {
7977         case IXGBE_SUCCESS:
7978                 ret = 0;
7979                 break;
7980         case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
7981                 ret = -ENOTSUP;
7982                 break;
7983         default:
7984                 ret = -EAGAIN;
7985                 break;
7986         }
7987
7988         return ret;
7989 }
7990
7991 static int
7992 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
7993 {
7994         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7995         int ret;
7996
7997         switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_MULTI)) {
7998         case IXGBE_SUCCESS:
7999                 ret = 0;
8000                 break;
8001         case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
8002                 ret = -ENOTSUP;
8003                 break;
8004         default:
8005                 ret = -EAGAIN;
8006                 break;
8007         }
8008
8009         return ret;
8010 }
8011
8012 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
8013 {
8014         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8015         u32 in_msg = 0;
8016
8017         /* peek the message first */
8018         in_msg = IXGBE_READ_REG(hw, IXGBE_VFMBMEM);
8019
8020         /* PF reset VF event */
8021         if (in_msg == IXGBE_PF_CONTROL_MSG) {
8022                 /* dummy mbx read to ack pf */
8023                 if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
8024                         return;
8025                 rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
8026                                              NULL);
8027         }
8028 }
8029
8030 static int
8031 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
8032 {
8033         uint32_t eicr;
8034         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8035         struct ixgbe_interrupt *intr =
8036                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8037         ixgbevf_intr_disable(dev);
8038
8039         /* read-on-clear nic registers here */
8040         eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
8041         intr->flags = 0;
8042
8043         /* only one misc vector supported - mailbox */
8044         eicr &= IXGBE_VTEICR_MASK;
8045         if (eicr == IXGBE_MISC_VEC_ID)
8046                 intr->flags |= IXGBE_FLAG_MAILBOX;
8047
8048         return 0;
8049 }
8050
8051 static int
8052 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
8053 {
8054         struct ixgbe_interrupt *intr =
8055                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8056
8057         if (intr->flags & IXGBE_FLAG_MAILBOX) {
8058                 ixgbevf_mbx_process(dev);
8059                 intr->flags &= ~IXGBE_FLAG_MAILBOX;
8060         }
8061
8062         ixgbevf_intr_enable(dev);
8063
8064         return 0;
8065 }
8066
8067 static void
8068 ixgbevf_dev_interrupt_handler(void *param)
8069 {
8070         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
8071
8072         ixgbevf_dev_interrupt_get_status(dev);
8073         ixgbevf_dev_interrupt_action(dev);
8074 }
8075
8076 /**
8077  *  ixgbe_disable_sec_tx_path_generic - Stops the transmit data path
8078  *  @hw: pointer to hardware structure
8079  *
8080  *  Stops the transmit data path and waits for the HW to internally empty
8081  *  the Tx security block
8082  **/
8083 int ixgbe_disable_sec_tx_path_generic(struct ixgbe_hw *hw)
8084 {
8085 #define IXGBE_MAX_SECTX_POLL 40
8086
8087         int i;
8088         int sectxreg;
8089
8090         sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8091         sectxreg |= IXGBE_SECTXCTRL_TX_DIS;
8092         IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8093         for (i = 0; i < IXGBE_MAX_SECTX_POLL; i++) {
8094                 sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXSTAT);
8095                 if (sectxreg & IXGBE_SECTXSTAT_SECTX_RDY)
8096                         break;
8097                 /* Use interrupt-safe sleep just in case */
8098                 usec_delay(1000);
8099         }
8100
8101         /* For informational purposes only */
8102         if (i >= IXGBE_MAX_SECTX_POLL)
8103                 PMD_DRV_LOG(DEBUG, "Tx unit being enabled before security "
8104                          "path fully disabled.  Continuing with init.");
8105
8106         return IXGBE_SUCCESS;
8107 }
8108
8109 /**
8110  *  ixgbe_enable_sec_tx_path_generic - Enables the transmit data path
8111  *  @hw: pointer to hardware structure
8112  *
8113  *  Enables the transmit data path.
8114  **/
8115 int ixgbe_enable_sec_tx_path_generic(struct ixgbe_hw *hw)
8116 {
8117         uint32_t sectxreg;
8118
8119         sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8120         sectxreg &= ~IXGBE_SECTXCTRL_TX_DIS;
8121         IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8122         IXGBE_WRITE_FLUSH(hw);
8123
8124         return IXGBE_SUCCESS;
8125 }
8126
8127 /* restore n-tuple filter */
8128 static inline void
8129 ixgbe_ntuple_filter_restore(struct rte_eth_dev *dev)
8130 {
8131         struct ixgbe_filter_info *filter_info =
8132                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8133         struct ixgbe_5tuple_filter *node;
8134
8135         TAILQ_FOREACH(node, &filter_info->fivetuple_list, entries) {
8136                 ixgbe_inject_5tuple_filter(dev, node);
8137         }
8138 }
8139
8140 /* restore ethernet type filter */
8141 static inline void
8142 ixgbe_ethertype_filter_restore(struct rte_eth_dev *dev)
8143 {
8144         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8145         struct ixgbe_filter_info *filter_info =
8146                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8147         int i;
8148
8149         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8150                 if (filter_info->ethertype_mask & (1 << i)) {
8151                         IXGBE_WRITE_REG(hw, IXGBE_ETQF(i),
8152                                         filter_info->ethertype_filters[i].etqf);
8153                         IXGBE_WRITE_REG(hw, IXGBE_ETQS(i),
8154                                         filter_info->ethertype_filters[i].etqs);
8155                         IXGBE_WRITE_FLUSH(hw);
8156                 }
8157         }
8158 }
8159
8160 /* restore SYN filter */
8161 static inline void
8162 ixgbe_syn_filter_restore(struct rte_eth_dev *dev)
8163 {
8164         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8165         struct ixgbe_filter_info *filter_info =
8166                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8167         uint32_t synqf;
8168
8169         synqf = filter_info->syn_info;
8170
8171         if (synqf & IXGBE_SYN_FILTER_ENABLE) {
8172                 IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
8173                 IXGBE_WRITE_FLUSH(hw);
8174         }
8175 }
8176
8177 /* restore L2 tunnel filter */
8178 static inline void
8179 ixgbe_l2_tn_filter_restore(struct rte_eth_dev *dev)
8180 {
8181         struct ixgbe_l2_tn_info *l2_tn_info =
8182                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8183         struct ixgbe_l2_tn_filter *node;
8184         struct ixgbe_l2_tunnel_conf l2_tn_conf;
8185
8186         TAILQ_FOREACH(node, &l2_tn_info->l2_tn_list, entries) {
8187                 l2_tn_conf.l2_tunnel_type = node->key.l2_tn_type;
8188                 l2_tn_conf.tunnel_id      = node->key.tn_id;
8189                 l2_tn_conf.pool           = node->pool;
8190                 (void)ixgbe_dev_l2_tunnel_filter_add(dev, &l2_tn_conf, TRUE);
8191         }
8192 }
8193
8194 /* restore rss filter */
8195 static inline void
8196 ixgbe_rss_filter_restore(struct rte_eth_dev *dev)
8197 {
8198         struct ixgbe_filter_info *filter_info =
8199                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8200
8201         if (filter_info->rss_info.conf.queue_num)
8202                 ixgbe_config_rss_filter(dev,
8203                         &filter_info->rss_info, TRUE);
8204 }
8205
8206 static int
8207 ixgbe_filter_restore(struct rte_eth_dev *dev)
8208 {
8209         ixgbe_ntuple_filter_restore(dev);
8210         ixgbe_ethertype_filter_restore(dev);
8211         ixgbe_syn_filter_restore(dev);
8212         ixgbe_fdir_filter_restore(dev);
8213         ixgbe_l2_tn_filter_restore(dev);
8214         ixgbe_rss_filter_restore(dev);
8215
8216         return 0;
8217 }
8218
8219 static void
8220 ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev)
8221 {
8222         struct ixgbe_l2_tn_info *l2_tn_info =
8223                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8224         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8225
8226         if (l2_tn_info->e_tag_en)
8227                 (void)ixgbe_e_tag_enable(hw);
8228
8229         if (l2_tn_info->e_tag_fwd_en)
8230                 (void)ixgbe_e_tag_forwarding_en_dis(dev, 1);
8231
8232         (void)ixgbe_update_e_tag_eth_type(hw, l2_tn_info->e_tag_ether_type);
8233 }
8234
8235 /* remove all the n-tuple filters */
8236 void
8237 ixgbe_clear_all_ntuple_filter(struct rte_eth_dev *dev)
8238 {
8239         struct ixgbe_filter_info *filter_info =
8240                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8241         struct ixgbe_5tuple_filter *p_5tuple;
8242
8243         while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
8244                 ixgbe_remove_5tuple_filter(dev, p_5tuple);
8245 }
8246
8247 /* remove all the ether type filters */
8248 void
8249 ixgbe_clear_all_ethertype_filter(struct rte_eth_dev *dev)
8250 {
8251         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8252         struct ixgbe_filter_info *filter_info =
8253                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8254         int i;
8255
8256         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8257                 if (filter_info->ethertype_mask & (1 << i) &&
8258                     !filter_info->ethertype_filters[i].conf) {
8259                         (void)ixgbe_ethertype_filter_remove(filter_info,
8260                                                             (uint8_t)i);
8261                         IXGBE_WRITE_REG(hw, IXGBE_ETQF(i), 0);
8262                         IXGBE_WRITE_REG(hw, IXGBE_ETQS(i), 0);
8263                         IXGBE_WRITE_FLUSH(hw);
8264                 }
8265         }
8266 }
8267
8268 /* remove the SYN filter */
8269 void
8270 ixgbe_clear_syn_filter(struct rte_eth_dev *dev)
8271 {
8272         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8273         struct ixgbe_filter_info *filter_info =
8274                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8275
8276         if (filter_info->syn_info & IXGBE_SYN_FILTER_ENABLE) {
8277                 filter_info->syn_info = 0;
8278
8279                 IXGBE_WRITE_REG(hw, IXGBE_SYNQF, 0);
8280                 IXGBE_WRITE_FLUSH(hw);
8281         }
8282 }
8283
8284 /* remove all the L2 tunnel filters */
8285 int
8286 ixgbe_clear_all_l2_tn_filter(struct rte_eth_dev *dev)
8287 {
8288         struct ixgbe_l2_tn_info *l2_tn_info =
8289                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8290         struct ixgbe_l2_tn_filter *l2_tn_filter;
8291         struct ixgbe_l2_tunnel_conf l2_tn_conf;
8292         int ret = 0;
8293
8294         while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
8295                 l2_tn_conf.l2_tunnel_type = l2_tn_filter->key.l2_tn_type;
8296                 l2_tn_conf.tunnel_id      = l2_tn_filter->key.tn_id;
8297                 l2_tn_conf.pool           = l2_tn_filter->pool;
8298                 ret = ixgbe_dev_l2_tunnel_filter_del(dev, &l2_tn_conf);
8299                 if (ret < 0)
8300                         return ret;
8301         }
8302
8303         return 0;
8304 }
8305
8306 void
8307 ixgbe_dev_macsec_setting_save(struct rte_eth_dev *dev,
8308                                 struct ixgbe_macsec_setting *macsec_setting)
8309 {
8310         struct ixgbe_macsec_setting *macsec =
8311                 IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8312
8313         macsec->offload_en = macsec_setting->offload_en;
8314         macsec->encrypt_en = macsec_setting->encrypt_en;
8315         macsec->replayprotect_en = macsec_setting->replayprotect_en;
8316 }
8317
8318 void
8319 ixgbe_dev_macsec_setting_reset(struct rte_eth_dev *dev)
8320 {
8321         struct ixgbe_macsec_setting *macsec =
8322                 IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8323
8324         macsec->offload_en = 0;
8325         macsec->encrypt_en = 0;
8326         macsec->replayprotect_en = 0;
8327 }
8328
8329 void
8330 ixgbe_dev_macsec_register_enable(struct rte_eth_dev *dev,
8331                                 struct ixgbe_macsec_setting *macsec_setting)
8332 {
8333         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8334         uint32_t ctrl;
8335         uint8_t en = macsec_setting->encrypt_en;
8336         uint8_t rp = macsec_setting->replayprotect_en;
8337
8338         /**
8339          * Workaround:
8340          * As no ixgbe_disable_sec_rx_path equivalent is
8341          * implemented for tx in the base code, and we are
8342          * not allowed to modify the base code in DPDK, so
8343          * just call the hand-written one directly for now.
8344          * The hardware support has been checked by
8345          * ixgbe_disable_sec_rx_path().
8346          */
8347         ixgbe_disable_sec_tx_path_generic(hw);
8348
8349         /* Enable Ethernet CRC (required by MACsec offload) */
8350         ctrl = IXGBE_READ_REG(hw, IXGBE_HLREG0);
8351         ctrl |= IXGBE_HLREG0_TXCRCEN | IXGBE_HLREG0_RXCRCSTRP;
8352         IXGBE_WRITE_REG(hw, IXGBE_HLREG0, ctrl);
8353
8354         /* Enable the TX and RX crypto engines */
8355         ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8356         ctrl &= ~IXGBE_SECTXCTRL_SECTX_DIS;
8357         IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8358
8359         ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8360         ctrl &= ~IXGBE_SECRXCTRL_SECRX_DIS;
8361         IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8362
8363         ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
8364         ctrl &= ~IXGBE_SECTX_MINSECIFG_MASK;
8365         ctrl |= 0x3;
8366         IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, ctrl);
8367
8368         /* Enable SA lookup */
8369         ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8370         ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8371         ctrl |= en ? IXGBE_LSECTXCTRL_AUTH_ENCRYPT :
8372                      IXGBE_LSECTXCTRL_AUTH;
8373         ctrl |= IXGBE_LSECTXCTRL_AISCI;
8374         ctrl &= ~IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8375         ctrl |= IXGBE_MACSEC_PNTHRSH & IXGBE_LSECTXCTRL_PNTHRSH_MASK;
8376         IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8377
8378         ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8379         ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8380         ctrl |= IXGBE_LSECRXCTRL_STRICT << IXGBE_LSECRXCTRL_EN_SHIFT;
8381         ctrl &= ~IXGBE_LSECRXCTRL_PLSH;
8382         if (rp)
8383                 ctrl |= IXGBE_LSECRXCTRL_RP;
8384         else
8385                 ctrl &= ~IXGBE_LSECRXCTRL_RP;
8386         IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8387
8388         /* Start the data paths */
8389         ixgbe_enable_sec_rx_path(hw);
8390         /**
8391          * Workaround:
8392          * As no ixgbe_enable_sec_rx_path equivalent is
8393          * implemented for tx in the base code, and we are
8394          * not allowed to modify the base code in DPDK, so
8395          * just call the hand-written one directly for now.
8396          */
8397         ixgbe_enable_sec_tx_path_generic(hw);
8398 }
8399
8400 void
8401 ixgbe_dev_macsec_register_disable(struct rte_eth_dev *dev)
8402 {
8403         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8404         uint32_t ctrl;
8405
8406         /**
8407          * Workaround:
8408          * As no ixgbe_disable_sec_rx_path equivalent is
8409          * implemented for tx in the base code, and we are
8410          * not allowed to modify the base code in DPDK, so
8411          * just call the hand-written one directly for now.
8412          * The hardware support has been checked by
8413          * ixgbe_disable_sec_rx_path().
8414          */
8415         ixgbe_disable_sec_tx_path_generic(hw);
8416
8417         /* Disable the TX and RX crypto engines */
8418         ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8419         ctrl |= IXGBE_SECTXCTRL_SECTX_DIS;
8420         IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
8421
8422         ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
8423         ctrl |= IXGBE_SECRXCTRL_SECRX_DIS;
8424         IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
8425
8426         /* Disable SA lookup */
8427         ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
8428         ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
8429         ctrl |= IXGBE_LSECTXCTRL_DISABLE;
8430         IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
8431
8432         ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
8433         ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
8434         ctrl |= IXGBE_LSECRXCTRL_DISABLE << IXGBE_LSECRXCTRL_EN_SHIFT;
8435         IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
8436
8437         /* Start the data paths */
8438         ixgbe_enable_sec_rx_path(hw);
8439         /**
8440          * Workaround:
8441          * As no ixgbe_enable_sec_rx_path equivalent is
8442          * implemented for tx in the base code, and we are
8443          * not allowed to modify the base code in DPDK, so
8444          * just call the hand-written one directly for now.
8445          */
8446         ixgbe_enable_sec_tx_path_generic(hw);
8447 }
8448
8449 RTE_PMD_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd);
8450 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
8451 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe, "* igb_uio | uio_pci_generic | vfio-pci");
8452 RTE_PMD_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd);
8453 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);
8454 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe_vf, "* igb_uio | vfio-pci");
8455 RTE_PMD_REGISTER_PARAM_STRING(net_ixgbe_vf,
8456                               IXGBEVF_DEVARG_PFLINK_FULLCHK "=<0|1>");
8457
8458 RTE_LOG_REGISTER(ixgbe_logtype_init, pmd.net.ixgbe.init, NOTICE);
8459 RTE_LOG_REGISTER(ixgbe_logtype_driver, pmd.net.ixgbe.driver, NOTICE);
8460
8461 #ifdef RTE_ETHDEV_DEBUG_RX
8462 RTE_LOG_REGISTER(ixgbe_logtype_rx, pmd.net.ixgbe.rx, DEBUG);
8463 #endif
8464 #ifdef RTE_ETHDEV_DEBUG_TX
8465 RTE_LOG_REGISTER(ixgbe_logtype_tx, pmd.net.ixgbe.tx, DEBUG);
8466 #endif