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