6b3d4fa19aa770ff08d1b64664ff32ab59f5b764
[dpdk.git] / drivers / net / ixgbe / ixgbe_ethdev.c
1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
5  *   All rights reserved.
6  *
7  *   Redistribution and use in source and binary forms, with or without
8  *   modification, are permitted provided that the following conditions
9  *   are met:
10  *
11  *     * Redistributions of source code must retain the above copyright
12  *       notice, this list of conditions and the following disclaimer.
13  *     * Redistributions in binary form must reproduce the above copyright
14  *       notice, this list of conditions and the following disclaimer in
15  *       the documentation and/or other materials provided with the
16  *       distribution.
17  *     * Neither the name of Intel Corporation nor the names of its
18  *       contributors may be used to endorse or promote products derived
19  *       from this software without specific prior written permission.
20  *
21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32  */
33
34 #include <sys/queue.h>
35 #include <stdio.h>
36 #include <errno.h>
37 #include <stdint.h>
38 #include <string.h>
39 #include <unistd.h>
40 #include <stdarg.h>
41 #include <inttypes.h>
42 #include <netinet/in.h>
43 #include <rte_byteorder.h>
44 #include <rte_common.h>
45 #include <rte_cycles.h>
46
47 #include <rte_interrupts.h>
48 #include <rte_log.h>
49 #include <rte_debug.h>
50 #include <rte_pci.h>
51 #include <rte_atomic.h>
52 #include <rte_branch_prediction.h>
53 #include <rte_memory.h>
54 #include <rte_memzone.h>
55 #include <rte_eal.h>
56 #include <rte_alarm.h>
57 #include <rte_ether.h>
58 #include <rte_ethdev.h>
59 #include <rte_atomic.h>
60 #include <rte_malloc.h>
61 #include <rte_random.h>
62 #include <rte_dev.h>
63
64 #include "ixgbe_logs.h"
65 #include "base/ixgbe_api.h"
66 #include "base/ixgbe_vf.h"
67 #include "base/ixgbe_common.h"
68 #include "ixgbe_ethdev.h"
69 #include "ixgbe_bypass.h"
70 #include "ixgbe_rxtx.h"
71 #include "base/ixgbe_type.h"
72 #include "base/ixgbe_phy.h"
73 #include "ixgbe_regs.h"
74
75 /*
76  * High threshold controlling when to start sending XOFF frames. Must be at
77  * least 8 bytes less than receive packet buffer size. This value is in units
78  * of 1024 bytes.
79  */
80 #define IXGBE_FC_HI    0x80
81
82 /*
83  * Low threshold controlling when to start sending XON frames. This value is
84  * in units of 1024 bytes.
85  */
86 #define IXGBE_FC_LO    0x40
87
88 /* Default minimum inter-interrupt interval for EITR configuration */
89 #define IXGBE_MIN_INTER_INTERRUPT_INTERVAL_DEFAULT    0x79E
90
91 /* Timer value included in XOFF frames. */
92 #define IXGBE_FC_PAUSE 0x680
93
94 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
95 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
96 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
97
98 #define IXGBE_MMW_SIZE_DEFAULT        0x4
99 #define IXGBE_MMW_SIZE_JUMBO_FRAME    0x14
100 #define IXGBE_MAX_RING_DESC           4096 /* replicate define from rxtx */
101
102 /*
103  *  Default values for RX/TX configuration
104  */
105 #define IXGBE_DEFAULT_RX_FREE_THRESH  32
106 #define IXGBE_DEFAULT_RX_PTHRESH      8
107 #define IXGBE_DEFAULT_RX_HTHRESH      8
108 #define IXGBE_DEFAULT_RX_WTHRESH      0
109
110 #define IXGBE_DEFAULT_TX_FREE_THRESH  32
111 #define IXGBE_DEFAULT_TX_PTHRESH      32
112 #define IXGBE_DEFAULT_TX_HTHRESH      0
113 #define IXGBE_DEFAULT_TX_WTHRESH      0
114 #define IXGBE_DEFAULT_TX_RSBIT_THRESH 32
115
116 /* Bit shift and mask */
117 #define IXGBE_4_BIT_WIDTH  (CHAR_BIT / 2)
118 #define IXGBE_4_BIT_MASK   RTE_LEN2MASK(IXGBE_4_BIT_WIDTH, uint8_t)
119 #define IXGBE_8_BIT_WIDTH  CHAR_BIT
120 #define IXGBE_8_BIT_MASK   UINT8_MAX
121
122 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
123
124 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
125
126 #define IXGBE_HKEY_MAX_INDEX 10
127
128 /* Additional timesync values. */
129 #define NSEC_PER_SEC             1000000000L
130 #define IXGBE_INCVAL_10GB        0x66666666
131 #define IXGBE_INCVAL_1GB         0x40000000
132 #define IXGBE_INCVAL_100         0x50000000
133 #define IXGBE_INCVAL_SHIFT_10GB  28
134 #define IXGBE_INCVAL_SHIFT_1GB   24
135 #define IXGBE_INCVAL_SHIFT_100   21
136 #define IXGBE_INCVAL_SHIFT_82599 7
137 #define IXGBE_INCPER_SHIFT_82599 24
138
139 #define IXGBE_CYCLECOUNTER_MASK   0xffffffffffffffffULL
140
141 #define IXGBE_VT_CTL_POOLING_MODE_MASK         0x00030000
142 #define IXGBE_VT_CTL_POOLING_MODE_ETAG         0x00010000
143 #define DEFAULT_ETAG_ETYPE                     0x893f
144 #define IXGBE_ETAG_ETYPE                       0x00005084
145 #define IXGBE_ETAG_ETYPE_MASK                  0x0000ffff
146 #define IXGBE_ETAG_ETYPE_VALID                 0x80000000
147 #define IXGBE_RAH_ADTYPE                       0x40000000
148 #define IXGBE_RAL_ETAG_FILTER_MASK             0x00003fff
149 #define IXGBE_VMVIR_TAGA_MASK                  0x18000000
150 #define IXGBE_VMVIR_TAGA_ETAG_INSERT           0x08000000
151 #define IXGBE_VMTIR(_i) (0x00017000 + ((_i) * 4)) /* 64 of these (0-63) */
152 #define IXGBE_QDE_STRIP_TAG                    0x00000004
153 #define IXGBE_VTEICR_MASK                      0x07
154
155 enum ixgbevf_xcast_modes {
156         IXGBEVF_XCAST_MODE_NONE = 0,
157         IXGBEVF_XCAST_MODE_MULTI,
158         IXGBEVF_XCAST_MODE_ALLMULTI,
159 };
160
161 #define IXGBE_EXVET_VET_EXT_SHIFT              16
162 #define IXGBE_DMATXCTL_VT_MASK                 0xFFFF0000
163
164 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev);
165 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
166 static int  ixgbe_dev_configure(struct rte_eth_dev *dev);
167 static int  ixgbe_dev_start(struct rte_eth_dev *dev);
168 static void ixgbe_dev_stop(struct rte_eth_dev *dev);
169 static int  ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
170 static int  ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
171 static void ixgbe_dev_close(struct rte_eth_dev *dev);
172 static void ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
173 static void ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
174 static void ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
175 static void ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
176 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
177                                 int wait_to_complete);
178 static void ixgbe_dev_stats_get(struct rte_eth_dev *dev,
179                                 struct rte_eth_stats *stats);
180 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
181                                 struct rte_eth_xstat *xstats, unsigned n);
182 static int ixgbevf_dev_xstats_get(struct rte_eth_dev *dev,
183                                   struct rte_eth_xstat *xstats, unsigned n);
184 static void ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
185 static void ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
186 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
187         struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned limit);
188 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
189         struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned limit);
190 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
191                                              uint16_t queue_id,
192                                              uint8_t stat_idx,
193                                              uint8_t is_rx);
194 static void ixgbe_dev_info_get(struct rte_eth_dev *dev,
195                                struct rte_eth_dev_info *dev_info);
196 static const uint32_t *ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev);
197 static void ixgbevf_dev_info_get(struct rte_eth_dev *dev,
198                                  struct rte_eth_dev_info *dev_info);
199 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
200
201 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
202                 uint16_t vlan_id, int on);
203 static int ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
204                                enum rte_vlan_type vlan_type,
205                                uint16_t tpid_id);
206 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
207                 uint16_t queue, bool on);
208 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
209                 int on);
210 static void ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
211 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
212 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
213 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
214 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
215
216 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
217 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
218 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
219                                struct rte_eth_fc_conf *fc_conf);
220 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
221                                struct rte_eth_fc_conf *fc_conf);
222 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
223                 struct rte_eth_pfc_conf *pfc_conf);
224 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
225                         struct rte_eth_rss_reta_entry64 *reta_conf,
226                         uint16_t reta_size);
227 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
228                         struct rte_eth_rss_reta_entry64 *reta_conf,
229                         uint16_t reta_size);
230 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
231 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev);
232 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
233 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
234 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev);
235 static void ixgbe_dev_interrupt_handler(struct rte_intr_handle *handle,
236                 void *param);
237 static void ixgbe_dev_interrupt_delayed_handler(void *param);
238 static void ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
239                 uint32_t index, uint32_t pool);
240 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
241 static void ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
242                                            struct ether_addr *mac_addr);
243 static void ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config);
244
245 /* For Virtual Function support */
246 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
247 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
248 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
249 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
250 static void ixgbevf_dev_stop(struct rte_eth_dev *dev);
251 static void ixgbevf_dev_close(struct rte_eth_dev *dev);
252 static void ixgbevf_intr_disable(struct ixgbe_hw *hw);
253 static void ixgbevf_intr_enable(struct ixgbe_hw *hw);
254 static void ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
255                 struct rte_eth_stats *stats);
256 static void ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
257 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
258                 uint16_t vlan_id, int on);
259 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
260                 uint16_t queue, int on);
261 static void ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
262 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
263 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
264                                             uint16_t queue_id);
265 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
266                                              uint16_t queue_id);
267 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
268                                  uint8_t queue, uint8_t msix_vector);
269 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
270 static void ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev);
271 static void ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev);
272
273 /* For Eth VMDQ APIs support */
274 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
275                 ether_addr * mac_addr, uint8_t on);
276 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on);
277 static int  ixgbe_set_pool_rx_mode(struct rte_eth_dev *dev,  uint16_t pool,
278                 uint16_t rx_mask, uint8_t on);
279 static int ixgbe_set_pool_rx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on);
280 static int ixgbe_set_pool_tx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on);
281 static int ixgbe_set_pool_vlan_filter(struct rte_eth_dev *dev, uint16_t vlan,
282                 uint64_t pool_mask, uint8_t vlan_on);
283 static int ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
284                 struct rte_eth_mirror_conf *mirror_conf,
285                 uint8_t rule_id, uint8_t on);
286 static int ixgbe_mirror_rule_reset(struct rte_eth_dev *dev,
287                 uint8_t rule_id);
288 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
289                                           uint16_t queue_id);
290 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
291                                            uint16_t queue_id);
292 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
293                                uint8_t queue, uint8_t msix_vector);
294 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
295
296 static int ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
297                 uint16_t queue_idx, uint16_t tx_rate);
298 static int ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
299                 uint16_t tx_rate, uint64_t q_msk);
300
301 static void ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
302                                  struct ether_addr *mac_addr,
303                                  uint32_t index, uint32_t pool);
304 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
305 static void ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
306                                              struct ether_addr *mac_addr);
307 static int ixgbe_syn_filter_set(struct rte_eth_dev *dev,
308                         struct rte_eth_syn_filter *filter,
309                         bool add);
310 static int ixgbe_syn_filter_get(struct rte_eth_dev *dev,
311                         struct rte_eth_syn_filter *filter);
312 static int ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
313                         enum rte_filter_op filter_op,
314                         void *arg);
315 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
316                         struct ixgbe_5tuple_filter *filter);
317 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
318                         struct ixgbe_5tuple_filter *filter);
319 static int ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
320                         struct rte_eth_ntuple_filter *filter,
321                         bool add);
322 static int ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
323                                 enum rte_filter_op filter_op,
324                                 void *arg);
325 static int ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
326                         struct rte_eth_ntuple_filter *filter);
327 static int ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
328                         struct rte_eth_ethertype_filter *filter,
329                         bool add);
330 static int ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
331                                 enum rte_filter_op filter_op,
332                                 void *arg);
333 static int ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
334                         struct rte_eth_ethertype_filter *filter);
335 static int ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
336                      enum rte_filter_type filter_type,
337                      enum rte_filter_op filter_op,
338                      void *arg);
339 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
340
341 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
342                                       struct ether_addr *mc_addr_set,
343                                       uint32_t nb_mc_addr);
344 static int ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
345                                    struct rte_eth_dcb_info *dcb_info);
346
347 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
348 static int ixgbe_get_regs(struct rte_eth_dev *dev,
349                             struct rte_dev_reg_info *regs);
350 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
351 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
352                                 struct rte_dev_eeprom_info *eeprom);
353 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
354                                 struct rte_dev_eeprom_info *eeprom);
355
356 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
357 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
358                                 struct rte_dev_reg_info *regs);
359
360 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
361 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
362 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
363                                             struct timespec *timestamp,
364                                             uint32_t flags);
365 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
366                                             struct timespec *timestamp);
367 static int ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
368 static int ixgbe_timesync_read_time(struct rte_eth_dev *dev,
369                                    struct timespec *timestamp);
370 static int ixgbe_timesync_write_time(struct rte_eth_dev *dev,
371                                    const struct timespec *timestamp);
372 static void ixgbevf_dev_interrupt_handler(struct rte_intr_handle *handle,
373                                           void *param);
374
375 static int ixgbe_dev_l2_tunnel_eth_type_conf
376         (struct rte_eth_dev *dev, struct rte_eth_l2_tunnel_conf *l2_tunnel);
377 static int ixgbe_dev_l2_tunnel_offload_set
378         (struct rte_eth_dev *dev,
379          struct rte_eth_l2_tunnel_conf *l2_tunnel,
380          uint32_t mask,
381          uint8_t en);
382 static int ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
383                                              enum rte_filter_op filter_op,
384                                              void *arg);
385
386 static int ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
387                                          struct rte_eth_udp_tunnel *udp_tunnel);
388 static int ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
389                                          struct rte_eth_udp_tunnel *udp_tunnel);
390
391 /*
392  * Define VF Stats MACRO for Non "cleared on read" register
393  */
394 #define UPDATE_VF_STAT(reg, last, cur)                          \
395 {                                                               \
396         uint32_t latest = IXGBE_READ_REG(hw, reg);              \
397         cur += (latest - last) & UINT_MAX;                      \
398         last = latest;                                          \
399 }
400
401 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
402 {                                                                \
403         u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
404         u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
405         u64 latest = ((new_msb << 32) | new_lsb);                \
406         cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
407         last = latest;                                           \
408 }
409
410 #define IXGBE_SET_HWSTRIP(h, q) do {\
411                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
412                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
413                 (h)->bitmap[idx] |= 1 << bit;\
414         } while (0)
415
416 #define IXGBE_CLEAR_HWSTRIP(h, q) do {\
417                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
418                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
419                 (h)->bitmap[idx] &= ~(1 << bit);\
420         } while (0)
421
422 #define IXGBE_GET_HWSTRIP(h, q, r) do {\
423                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
424                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
425                 (r) = (h)->bitmap[idx] >> bit & 1;\
426         } while (0)
427
428 /*
429  * The set of PCI devices this driver supports
430  */
431 static const struct rte_pci_id pci_id_ixgbe_map[] = {
432         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598) },
433         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_BX) },
434         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT) },
435         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT) },
436         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT) },
437         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2) },
438         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM) },
439         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4) },
440         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT) },
441         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT) },
442         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM) },
443         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR) },
444         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4) },
445         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ) },
446         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KR) },
447         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE) },
448         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ) },
449         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4) },
450         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP) },
451         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_SUBDEV_ID_82599_SFP) },
452         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_SUBDEV_ID_82599_RNDC) },
453         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_SUBDEV_ID_82599_560FLR) },
454         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_SUBDEV_ID_82599_ECNA_DP) },
455         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE) },
456         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE) },
457         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_EM) },
458         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2) },
459         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP) },
460         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP) },
461         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP) },
462         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM) },
463         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM) },
464         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_LS) },
465         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T) },
466         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1) },
467         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP) },
468         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T) },
469         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_1G_T) },
470         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T) },
471         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1) },
472         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR) },
473         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR_L) },
474         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP_N) },
475         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII) },
476         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII_L) },
477         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_10G_T) },
478         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP) },
479         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP_N) },
480         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP) },
481         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T) },
482         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T_L) },
483         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4) },
484         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR) },
485 #ifdef RTE_NIC_BYPASS
486         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BYPASS) },
487 #endif
488         { .vendor_id = 0, /* sentinel */ },
489 };
490
491 /*
492  * The set of PCI devices this driver supports (for 82599 VF)
493  */
494 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
495         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF) },
496         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF_HV) },
497         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF) },
498         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF_HV) },
499         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF_HV) },
500         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF) },
501         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF) },
502         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF_HV) },
503         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF) },
504         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF_HV) },
505         { .vendor_id = 0, /* sentinel */ },
506 };
507
508 static const struct rte_eth_desc_lim rx_desc_lim = {
509         .nb_max = IXGBE_MAX_RING_DESC,
510         .nb_min = IXGBE_MIN_RING_DESC,
511         .nb_align = IXGBE_RXD_ALIGN,
512 };
513
514 static const struct rte_eth_desc_lim tx_desc_lim = {
515         .nb_max = IXGBE_MAX_RING_DESC,
516         .nb_min = IXGBE_MIN_RING_DESC,
517         .nb_align = IXGBE_TXD_ALIGN,
518 };
519
520 static const struct eth_dev_ops ixgbe_eth_dev_ops = {
521         .dev_configure        = ixgbe_dev_configure,
522         .dev_start            = ixgbe_dev_start,
523         .dev_stop             = ixgbe_dev_stop,
524         .dev_set_link_up    = ixgbe_dev_set_link_up,
525         .dev_set_link_down  = ixgbe_dev_set_link_down,
526         .dev_close            = ixgbe_dev_close,
527         .promiscuous_enable   = ixgbe_dev_promiscuous_enable,
528         .promiscuous_disable  = ixgbe_dev_promiscuous_disable,
529         .allmulticast_enable  = ixgbe_dev_allmulticast_enable,
530         .allmulticast_disable = ixgbe_dev_allmulticast_disable,
531         .link_update          = ixgbe_dev_link_update,
532         .stats_get            = ixgbe_dev_stats_get,
533         .xstats_get           = ixgbe_dev_xstats_get,
534         .stats_reset          = ixgbe_dev_stats_reset,
535         .xstats_reset         = ixgbe_dev_xstats_reset,
536         .xstats_get_names     = ixgbe_dev_xstats_get_names,
537         .queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
538         .dev_infos_get        = ixgbe_dev_info_get,
539         .dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
540         .mtu_set              = ixgbe_dev_mtu_set,
541         .vlan_filter_set      = ixgbe_vlan_filter_set,
542         .vlan_tpid_set        = ixgbe_vlan_tpid_set,
543         .vlan_offload_set     = ixgbe_vlan_offload_set,
544         .vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
545         .rx_queue_start       = ixgbe_dev_rx_queue_start,
546         .rx_queue_stop        = ixgbe_dev_rx_queue_stop,
547         .tx_queue_start       = ixgbe_dev_tx_queue_start,
548         .tx_queue_stop        = ixgbe_dev_tx_queue_stop,
549         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
550         .rx_queue_intr_enable = ixgbe_dev_rx_queue_intr_enable,
551         .rx_queue_intr_disable = ixgbe_dev_rx_queue_intr_disable,
552         .rx_queue_release     = ixgbe_dev_rx_queue_release,
553         .rx_queue_count       = ixgbe_dev_rx_queue_count,
554         .rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
555         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
556         .tx_queue_release     = ixgbe_dev_tx_queue_release,
557         .dev_led_on           = ixgbe_dev_led_on,
558         .dev_led_off          = ixgbe_dev_led_off,
559         .flow_ctrl_get        = ixgbe_flow_ctrl_get,
560         .flow_ctrl_set        = ixgbe_flow_ctrl_set,
561         .priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
562         .mac_addr_add         = ixgbe_add_rar,
563         .mac_addr_remove      = ixgbe_remove_rar,
564         .mac_addr_set         = ixgbe_set_default_mac_addr,
565         .uc_hash_table_set    = ixgbe_uc_hash_table_set,
566         .uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
567         .mirror_rule_set      = ixgbe_mirror_rule_set,
568         .mirror_rule_reset    = ixgbe_mirror_rule_reset,
569         .set_vf_rx_mode       = ixgbe_set_pool_rx_mode,
570         .set_vf_rx            = ixgbe_set_pool_rx,
571         .set_vf_tx            = ixgbe_set_pool_tx,
572         .set_vf_vlan_filter   = ixgbe_set_pool_vlan_filter,
573         .set_queue_rate_limit = ixgbe_set_queue_rate_limit,
574         .set_vf_rate_limit    = ixgbe_set_vf_rate_limit,
575         .reta_update          = ixgbe_dev_rss_reta_update,
576         .reta_query           = ixgbe_dev_rss_reta_query,
577 #ifdef RTE_NIC_BYPASS
578         .bypass_init          = ixgbe_bypass_init,
579         .bypass_state_set     = ixgbe_bypass_state_store,
580         .bypass_state_show    = ixgbe_bypass_state_show,
581         .bypass_event_set     = ixgbe_bypass_event_store,
582         .bypass_event_show    = ixgbe_bypass_event_show,
583         .bypass_wd_timeout_set  = ixgbe_bypass_wd_timeout_store,
584         .bypass_wd_timeout_show = ixgbe_bypass_wd_timeout_show,
585         .bypass_ver_show      = ixgbe_bypass_ver_show,
586         .bypass_wd_reset      = ixgbe_bypass_wd_reset,
587 #endif /* RTE_NIC_BYPASS */
588         .rss_hash_update      = ixgbe_dev_rss_hash_update,
589         .rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
590         .filter_ctrl          = ixgbe_dev_filter_ctrl,
591         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
592         .rxq_info_get         = ixgbe_rxq_info_get,
593         .txq_info_get         = ixgbe_txq_info_get,
594         .timesync_enable      = ixgbe_timesync_enable,
595         .timesync_disable     = ixgbe_timesync_disable,
596         .timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
597         .timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
598         .get_reg              = ixgbe_get_regs,
599         .get_eeprom_length    = ixgbe_get_eeprom_length,
600         .get_eeprom           = ixgbe_get_eeprom,
601         .set_eeprom           = ixgbe_set_eeprom,
602         .get_dcb_info         = ixgbe_dev_get_dcb_info,
603         .timesync_adjust_time = ixgbe_timesync_adjust_time,
604         .timesync_read_time   = ixgbe_timesync_read_time,
605         .timesync_write_time  = ixgbe_timesync_write_time,
606         .l2_tunnel_eth_type_conf = ixgbe_dev_l2_tunnel_eth_type_conf,
607         .l2_tunnel_offload_set   = ixgbe_dev_l2_tunnel_offload_set,
608         .udp_tunnel_port_add  = ixgbe_dev_udp_tunnel_port_add,
609         .udp_tunnel_port_del  = ixgbe_dev_udp_tunnel_port_del,
610 };
611
612 /*
613  * dev_ops for virtual function, bare necessities for basic vf
614  * operation have been implemented
615  */
616 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
617         .dev_configure        = ixgbevf_dev_configure,
618         .dev_start            = ixgbevf_dev_start,
619         .dev_stop             = ixgbevf_dev_stop,
620         .link_update          = ixgbe_dev_link_update,
621         .stats_get            = ixgbevf_dev_stats_get,
622         .xstats_get           = ixgbevf_dev_xstats_get,
623         .stats_reset          = ixgbevf_dev_stats_reset,
624         .xstats_reset         = ixgbevf_dev_stats_reset,
625         .xstats_get_names     = ixgbevf_dev_xstats_get_names,
626         .dev_close            = ixgbevf_dev_close,
627         .allmulticast_enable  = ixgbevf_dev_allmulticast_enable,
628         .allmulticast_disable = ixgbevf_dev_allmulticast_disable,
629         .dev_infos_get        = ixgbevf_dev_info_get,
630         .dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
631         .mtu_set              = ixgbevf_dev_set_mtu,
632         .vlan_filter_set      = ixgbevf_vlan_filter_set,
633         .vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
634         .vlan_offload_set     = ixgbevf_vlan_offload_set,
635         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
636         .rx_queue_release     = ixgbe_dev_rx_queue_release,
637         .rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
638         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
639         .tx_queue_release     = ixgbe_dev_tx_queue_release,
640         .rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
641         .rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
642         .mac_addr_add         = ixgbevf_add_mac_addr,
643         .mac_addr_remove      = ixgbevf_remove_mac_addr,
644         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
645         .rxq_info_get         = ixgbe_rxq_info_get,
646         .txq_info_get         = ixgbe_txq_info_get,
647         .mac_addr_set         = ixgbevf_set_default_mac_addr,
648         .get_reg              = ixgbevf_get_regs,
649         .reta_update          = ixgbe_dev_rss_reta_update,
650         .reta_query           = ixgbe_dev_rss_reta_query,
651         .rss_hash_update      = ixgbe_dev_rss_hash_update,
652         .rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
653 };
654
655 /* store statistics names and its offset in stats structure */
656 struct rte_ixgbe_xstats_name_off {
657         char name[RTE_ETH_XSTATS_NAME_SIZE];
658         unsigned offset;
659 };
660
661 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
662         {"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
663         {"rx_illegal_byte_errors", offsetof(struct ixgbe_hw_stats, illerrc)},
664         {"rx_error_bytes", offsetof(struct ixgbe_hw_stats, errbc)},
665         {"mac_local_errors", offsetof(struct ixgbe_hw_stats, mlfc)},
666         {"mac_remote_errors", offsetof(struct ixgbe_hw_stats, mrfc)},
667         {"rx_length_errors", offsetof(struct ixgbe_hw_stats, rlec)},
668         {"tx_xon_packets", offsetof(struct ixgbe_hw_stats, lxontxc)},
669         {"rx_xon_packets", offsetof(struct ixgbe_hw_stats, lxonrxc)},
670         {"tx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxofftxc)},
671         {"rx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
672         {"rx_size_64_packets", offsetof(struct ixgbe_hw_stats, prc64)},
673         {"rx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, prc127)},
674         {"rx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, prc255)},
675         {"rx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, prc511)},
676         {"rx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
677                 prc1023)},
678         {"rx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
679                 prc1522)},
680         {"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
681         {"rx_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
682         {"rx_fragment_errors", offsetof(struct ixgbe_hw_stats, rfc)},
683         {"rx_undersize_errors", offsetof(struct ixgbe_hw_stats, ruc)},
684         {"rx_oversize_errors", offsetof(struct ixgbe_hw_stats, roc)},
685         {"rx_jabber_errors", offsetof(struct ixgbe_hw_stats, rjc)},
686         {"rx_management_packets", offsetof(struct ixgbe_hw_stats, mngprc)},
687         {"rx_management_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
688         {"tx_management_packets", offsetof(struct ixgbe_hw_stats, mngptc)},
689         {"rx_total_packets", offsetof(struct ixgbe_hw_stats, tpr)},
690         {"rx_total_bytes", offsetof(struct ixgbe_hw_stats, tor)},
691         {"tx_total_packets", offsetof(struct ixgbe_hw_stats, tpt)},
692         {"tx_size_64_packets", offsetof(struct ixgbe_hw_stats, ptc64)},
693         {"tx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, ptc127)},
694         {"tx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, ptc255)},
695         {"tx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, ptc511)},
696         {"tx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
697                 ptc1023)},
698         {"tx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
699                 ptc1522)},
700         {"tx_multicast_packets", offsetof(struct ixgbe_hw_stats, mptc)},
701         {"tx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bptc)},
702         {"rx_mac_short_packet_dropped", offsetof(struct ixgbe_hw_stats, mspdc)},
703         {"rx_l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
704
705         {"flow_director_added_filters", offsetof(struct ixgbe_hw_stats,
706                 fdirustat_add)},
707         {"flow_director_removed_filters", offsetof(struct ixgbe_hw_stats,
708                 fdirustat_remove)},
709         {"flow_director_filter_add_errors", offsetof(struct ixgbe_hw_stats,
710                 fdirfstat_fadd)},
711         {"flow_director_filter_remove_errors", offsetof(struct ixgbe_hw_stats,
712                 fdirfstat_fremove)},
713         {"flow_director_matched_filters", offsetof(struct ixgbe_hw_stats,
714                 fdirmatch)},
715         {"flow_director_missed_filters", offsetof(struct ixgbe_hw_stats,
716                 fdirmiss)},
717
718         {"rx_fcoe_crc_errors", offsetof(struct ixgbe_hw_stats, fccrc)},
719         {"rx_fcoe_dropped", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
720         {"rx_fcoe_mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats,
721                 fclast)},
722         {"rx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeprc)},
723         {"tx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeptc)},
724         {"rx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwrc)},
725         {"tx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwtc)},
726         {"rx_fcoe_no_direct_data_placement", offsetof(struct ixgbe_hw_stats,
727                 fcoe_noddp)},
728         {"rx_fcoe_no_direct_data_placement_ext_buff",
729                 offsetof(struct ixgbe_hw_stats, fcoe_noddp_ext_buff)},
730
731         {"tx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
732                 lxontxc)},
733         {"rx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
734                 lxonrxc)},
735         {"tx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
736                 lxofftxc)},
737         {"rx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
738                 lxoffrxc)},
739         {"rx_total_missed_packets", offsetof(struct ixgbe_hw_stats, mpctotal)},
740 };
741
742 #define IXGBE_NB_HW_STATS (sizeof(rte_ixgbe_stats_strings) / \
743                            sizeof(rte_ixgbe_stats_strings[0]))
744
745 /* Per-queue statistics */
746 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_rxq_strings[] = {
747         {"mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats, rnbc)},
748         {"dropped", offsetof(struct ixgbe_hw_stats, mpc)},
749         {"xon_packets", offsetof(struct ixgbe_hw_stats, pxonrxc)},
750         {"xoff_packets", offsetof(struct ixgbe_hw_stats, pxoffrxc)},
751 };
752
753 #define IXGBE_NB_RXQ_PRIO_STATS (sizeof(rte_ixgbe_rxq_strings) / \
754                            sizeof(rte_ixgbe_rxq_strings[0]))
755 #define IXGBE_NB_RXQ_PRIO_VALUES 8
756
757 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_txq_strings[] = {
758         {"xon_packets", offsetof(struct ixgbe_hw_stats, pxontxc)},
759         {"xoff_packets", offsetof(struct ixgbe_hw_stats, pxofftxc)},
760         {"xon_to_xoff_packets", offsetof(struct ixgbe_hw_stats,
761                 pxon2offc)},
762 };
763
764 #define IXGBE_NB_TXQ_PRIO_STATS (sizeof(rte_ixgbe_txq_strings) / \
765                            sizeof(rte_ixgbe_txq_strings[0]))
766 #define IXGBE_NB_TXQ_PRIO_VALUES 8
767
768 static const struct rte_ixgbe_xstats_name_off rte_ixgbevf_stats_strings[] = {
769         {"rx_multicast_packets", offsetof(struct ixgbevf_hw_stats, vfmprc)},
770 };
771
772 #define IXGBEVF_NB_XSTATS (sizeof(rte_ixgbevf_stats_strings) /  \
773                 sizeof(rte_ixgbevf_stats_strings[0]))
774
775 /**
776  * Atomically reads the link status information from global
777  * structure rte_eth_dev.
778  *
779  * @param dev
780  *   - Pointer to the structure rte_eth_dev to read from.
781  *   - Pointer to the buffer to be saved with the link status.
782  *
783  * @return
784  *   - On success, zero.
785  *   - On failure, negative value.
786  */
787 static inline int
788 rte_ixgbe_dev_atomic_read_link_status(struct rte_eth_dev *dev,
789                                 struct rte_eth_link *link)
790 {
791         struct rte_eth_link *dst = link;
792         struct rte_eth_link *src = &(dev->data->dev_link);
793
794         if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
795                                         *(uint64_t *)src) == 0)
796                 return -1;
797
798         return 0;
799 }
800
801 /**
802  * Atomically writes the link status information into global
803  * structure rte_eth_dev.
804  *
805  * @param dev
806  *   - Pointer to the structure rte_eth_dev to read from.
807  *   - Pointer to the buffer to be saved with the link status.
808  *
809  * @return
810  *   - On success, zero.
811  *   - On failure, negative value.
812  */
813 static inline int
814 rte_ixgbe_dev_atomic_write_link_status(struct rte_eth_dev *dev,
815                                 struct rte_eth_link *link)
816 {
817         struct rte_eth_link *dst = &(dev->data->dev_link);
818         struct rte_eth_link *src = link;
819
820         if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
821                                         *(uint64_t *)src) == 0)
822                 return -1;
823
824         return 0;
825 }
826
827 /*
828  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
829  */
830 static inline int
831 ixgbe_is_sfp(struct ixgbe_hw *hw)
832 {
833         switch (hw->phy.type) {
834         case ixgbe_phy_sfp_avago:
835         case ixgbe_phy_sfp_ftl:
836         case ixgbe_phy_sfp_intel:
837         case ixgbe_phy_sfp_unknown:
838         case ixgbe_phy_sfp_passive_tyco:
839         case ixgbe_phy_sfp_passive_unknown:
840                 return 1;
841         default:
842                 return 0;
843         }
844 }
845
846 static inline int32_t
847 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
848 {
849         uint32_t ctrl_ext;
850         int32_t status;
851
852         status = ixgbe_reset_hw(hw);
853
854         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
855         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
856         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
857         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
858         IXGBE_WRITE_FLUSH(hw);
859
860         return status;
861 }
862
863 static inline void
864 ixgbe_enable_intr(struct rte_eth_dev *dev)
865 {
866         struct ixgbe_interrupt *intr =
867                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
868         struct ixgbe_hw *hw =
869                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
870
871         IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
872         IXGBE_WRITE_FLUSH(hw);
873 }
874
875 /*
876  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
877  */
878 static void
879 ixgbe_disable_intr(struct ixgbe_hw *hw)
880 {
881         PMD_INIT_FUNC_TRACE();
882
883         if (hw->mac.type == ixgbe_mac_82598EB) {
884                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
885         } else {
886                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
887                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
888                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
889         }
890         IXGBE_WRITE_FLUSH(hw);
891 }
892
893 /*
894  * This function resets queue statistics mapping registers.
895  * From Niantic datasheet, Initialization of Statistics section:
896  * "...if software requires the queue counters, the RQSMR and TQSM registers
897  * must be re-programmed following a device reset.
898  */
899 static void
900 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
901 {
902         uint32_t i;
903
904         for (i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
905                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
906                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
907         }
908 }
909
910
911 static int
912 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
913                                   uint16_t queue_id,
914                                   uint8_t stat_idx,
915                                   uint8_t is_rx)
916 {
917 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
918 #define NB_QMAP_FIELDS_PER_QSM_REG 4
919 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
920
921         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
922         struct ixgbe_stat_mapping_registers *stat_mappings =
923                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
924         uint32_t qsmr_mask = 0;
925         uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
926         uint32_t q_map;
927         uint8_t n, offset;
928
929         if ((hw->mac.type != ixgbe_mac_82599EB) &&
930                 (hw->mac.type != ixgbe_mac_X540) &&
931                 (hw->mac.type != ixgbe_mac_X550) &&
932                 (hw->mac.type != ixgbe_mac_X550EM_x) &&
933                 (hw->mac.type != ixgbe_mac_X550EM_a))
934                 return -ENOSYS;
935
936         PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
937                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
938                      queue_id, stat_idx);
939
940         n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
941         if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
942                 PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
943                 return -EIO;
944         }
945         offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
946
947         /* Now clear any previous stat_idx set */
948         clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
949         if (!is_rx)
950                 stat_mappings->tqsm[n] &= ~clearing_mask;
951         else
952                 stat_mappings->rqsmr[n] &= ~clearing_mask;
953
954         q_map = (uint32_t)stat_idx;
955         q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
956         qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
957         if (!is_rx)
958                 stat_mappings->tqsm[n] |= qsmr_mask;
959         else
960                 stat_mappings->rqsmr[n] |= qsmr_mask;
961
962         PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
963                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
964                      queue_id, stat_idx);
965         PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
966                      is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
967
968         /* Now write the mapping in the appropriate register */
969         if (is_rx) {
970                 PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
971                              stat_mappings->rqsmr[n], n);
972                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
973         } else {
974                 PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
975                              stat_mappings->tqsm[n], n);
976                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
977         }
978         return 0;
979 }
980
981 static void
982 ixgbe_restore_statistics_mapping(struct rte_eth_dev *dev)
983 {
984         struct ixgbe_stat_mapping_registers *stat_mappings =
985                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
986         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
987         int i;
988
989         /* write whatever was in stat mapping table to the NIC */
990         for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
991                 /* rx */
992                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
993
994                 /* tx */
995                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
996         }
997 }
998
999 static void
1000 ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config)
1001 {
1002         uint8_t i;
1003         struct ixgbe_dcb_tc_config *tc;
1004         uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
1005
1006         dcb_config->num_tcs.pg_tcs = dcb_max_tc;
1007         dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
1008         for (i = 0; i < dcb_max_tc; i++) {
1009                 tc = &dcb_config->tc_config[i];
1010                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
1011                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
1012                                  (uint8_t)(100/dcb_max_tc + (i & 1));
1013                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
1014                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
1015                                  (uint8_t)(100/dcb_max_tc + (i & 1));
1016                 tc->pfc = ixgbe_dcb_pfc_disabled;
1017         }
1018
1019         /* Initialize default user to priority mapping, UPx->TC0 */
1020         tc = &dcb_config->tc_config[0];
1021         tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
1022         tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
1023         for (i = 0; i < IXGBE_DCB_MAX_BW_GROUP; i++) {
1024                 dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
1025                 dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
1026         }
1027         dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
1028         dcb_config->pfc_mode_enable = false;
1029         dcb_config->vt_mode = true;
1030         dcb_config->round_robin_enable = false;
1031         /* support all DCB capabilities in 82599 */
1032         dcb_config->support.capabilities = 0xFF;
1033
1034         /*we only support 4 Tcs for X540, X550 */
1035         if (hw->mac.type == ixgbe_mac_X540 ||
1036                 hw->mac.type == ixgbe_mac_X550 ||
1037                 hw->mac.type == ixgbe_mac_X550EM_x ||
1038                 hw->mac.type == ixgbe_mac_X550EM_a) {
1039                 dcb_config->num_tcs.pg_tcs = 4;
1040                 dcb_config->num_tcs.pfc_tcs = 4;
1041         }
1042 }
1043
1044 /*
1045  * Ensure that all locks are released before first NVM or PHY access
1046  */
1047 static void
1048 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
1049 {
1050         uint16_t mask;
1051
1052         /*
1053          * Phy lock should not fail in this early stage. If this is the case,
1054          * it is due to an improper exit of the application.
1055          * So force the release of the faulty lock. Release of common lock
1056          * is done automatically by swfw_sync function.
1057          */
1058         mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
1059         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1060                 PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
1061         }
1062         ixgbe_release_swfw_semaphore(hw, mask);
1063
1064         /*
1065          * These ones are more tricky since they are common to all ports; but
1066          * swfw_sync retries last long enough (1s) to be almost sure that if
1067          * lock can not be taken it is due to an improper lock of the
1068          * semaphore.
1069          */
1070         mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
1071         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1072                 PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1073         }
1074         ixgbe_release_swfw_semaphore(hw, mask);
1075 }
1076
1077 /*
1078  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
1079  * It returns 0 on success.
1080  */
1081 static int
1082 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev)
1083 {
1084         struct rte_pci_device *pci_dev;
1085         struct ixgbe_hw *hw =
1086                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1087         struct ixgbe_vfta *shadow_vfta =
1088                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1089         struct ixgbe_hwstrip *hwstrip =
1090                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1091         struct ixgbe_dcb_config *dcb_config =
1092                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
1093         struct ixgbe_filter_info *filter_info =
1094                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1095         uint32_t ctrl_ext;
1096         uint16_t csum;
1097         int diag, i;
1098
1099         PMD_INIT_FUNC_TRACE();
1100
1101         eth_dev->dev_ops = &ixgbe_eth_dev_ops;
1102         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1103         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1104
1105         /*
1106          * For secondary processes, we don't initialise any further as primary
1107          * has already done this work. Only check we don't need a different
1108          * RX and TX function.
1109          */
1110         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1111                 struct ixgbe_tx_queue *txq;
1112                 /* TX queue function in primary, set by last queue initialized
1113                  * Tx queue may not initialized by primary process
1114                  */
1115                 if (eth_dev->data->tx_queues) {
1116                         txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
1117                         ixgbe_set_tx_function(eth_dev, txq);
1118                 } else {
1119                         /* Use default TX function if we get here */
1120                         PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
1121                                      "Using default TX function.");
1122                 }
1123
1124                 ixgbe_set_rx_function(eth_dev);
1125
1126                 return 0;
1127         }
1128         pci_dev = eth_dev->pci_dev;
1129
1130         rte_eth_copy_pci_info(eth_dev, pci_dev);
1131
1132         /* Vendor and Device ID need to be set before init of shared code */
1133         hw->device_id = pci_dev->id.device_id;
1134         hw->vendor_id = pci_dev->id.vendor_id;
1135         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1136         hw->allow_unsupported_sfp = 1;
1137
1138         /* Initialize the shared code (base driver) */
1139 #ifdef RTE_NIC_BYPASS
1140         diag = ixgbe_bypass_init_shared_code(hw);
1141 #else
1142         diag = ixgbe_init_shared_code(hw);
1143 #endif /* RTE_NIC_BYPASS */
1144
1145         if (diag != IXGBE_SUCCESS) {
1146                 PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
1147                 return -EIO;
1148         }
1149
1150         /* pick up the PCI bus settings for reporting later */
1151         ixgbe_get_bus_info(hw);
1152
1153         /* Unlock any pending hardware semaphore */
1154         ixgbe_swfw_lock_reset(hw);
1155
1156         /* Initialize DCB configuration*/
1157         memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
1158         ixgbe_dcb_init(hw, dcb_config);
1159         /* Get Hardware Flow Control setting */
1160         hw->fc.requested_mode = ixgbe_fc_full;
1161         hw->fc.current_mode = ixgbe_fc_full;
1162         hw->fc.pause_time = IXGBE_FC_PAUSE;
1163         for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
1164                 hw->fc.low_water[i] = IXGBE_FC_LO;
1165                 hw->fc.high_water[i] = IXGBE_FC_HI;
1166         }
1167         hw->fc.send_xon = 1;
1168
1169         /* Make sure we have a good EEPROM before we read from it */
1170         diag = ixgbe_validate_eeprom_checksum(hw, &csum);
1171         if (diag != IXGBE_SUCCESS) {
1172                 PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
1173                 return -EIO;
1174         }
1175
1176 #ifdef RTE_NIC_BYPASS
1177         diag = ixgbe_bypass_init_hw(hw);
1178 #else
1179         diag = ixgbe_init_hw(hw);
1180 #endif /* RTE_NIC_BYPASS */
1181
1182         /*
1183          * Devices with copper phys will fail to initialise if ixgbe_init_hw()
1184          * is called too soon after the kernel driver unbinding/binding occurs.
1185          * The failure occurs in ixgbe_identify_phy_generic() for all devices,
1186          * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
1187          * also called. See ixgbe_identify_phy_82599(). The reason for the
1188          * failure is not known, and only occuts when virtualisation features
1189          * are disabled in the bios. A delay of 100ms  was found to be enough by
1190          * trial-and-error, and is doubled to be safe.
1191          */
1192         if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
1193                 rte_delay_ms(200);
1194                 diag = ixgbe_init_hw(hw);
1195         }
1196
1197         if (diag == IXGBE_ERR_EEPROM_VERSION) {
1198                 PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
1199                              "LOM.  Please be aware there may be issues associated "
1200                              "with your hardware.");
1201                 PMD_INIT_LOG(ERR, "If you are experiencing problems "
1202                              "please contact your Intel or hardware representative "
1203                              "who provided you with this hardware.");
1204         } else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
1205                 PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
1206         if (diag) {
1207                 PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
1208                 return -EIO;
1209         }
1210
1211         /* Reset the hw statistics */
1212         ixgbe_dev_stats_reset(eth_dev);
1213
1214         /* disable interrupt */
1215         ixgbe_disable_intr(hw);
1216
1217         /* reset mappings for queue statistics hw counters*/
1218         ixgbe_reset_qstat_mappings(hw);
1219
1220         /* Allocate memory for storing MAC addresses */
1221         eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
1222                                                hw->mac.num_rar_entries, 0);
1223         if (eth_dev->data->mac_addrs == NULL) {
1224                 PMD_INIT_LOG(ERR,
1225                              "Failed to allocate %u bytes needed to store "
1226                              "MAC addresses",
1227                              ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1228                 return -ENOMEM;
1229         }
1230         /* Copy the permanent MAC address */
1231         ether_addr_copy((struct ether_addr *) hw->mac.perm_addr,
1232                         &eth_dev->data->mac_addrs[0]);
1233
1234         /* Allocate memory for storing hash filter MAC addresses */
1235         eth_dev->data->hash_mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
1236                                                     IXGBE_VMDQ_NUM_UC_MAC, 0);
1237         if (eth_dev->data->hash_mac_addrs == NULL) {
1238                 PMD_INIT_LOG(ERR,
1239                              "Failed to allocate %d bytes needed to store MAC addresses",
1240                              ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
1241                 return -ENOMEM;
1242         }
1243
1244         /* initialize the vfta */
1245         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1246
1247         /* initialize the hw strip bitmap*/
1248         memset(hwstrip, 0, sizeof(*hwstrip));
1249
1250         /* initialize PF if max_vfs not zero */
1251         ixgbe_pf_host_init(eth_dev);
1252
1253         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1254         /* let hardware know driver is loaded */
1255         ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1256         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
1257         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1258         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1259         IXGBE_WRITE_FLUSH(hw);
1260
1261         if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1262                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1263                              (int) hw->mac.type, (int) hw->phy.type,
1264                              (int) hw->phy.sfp_type);
1265         else
1266                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1267                              (int) hw->mac.type, (int) hw->phy.type);
1268
1269         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1270                      eth_dev->data->port_id, pci_dev->id.vendor_id,
1271                      pci_dev->id.device_id);
1272
1273         rte_intr_callback_register(&pci_dev->intr_handle,
1274                                    ixgbe_dev_interrupt_handler,
1275                                    (void *)eth_dev);
1276
1277         /* enable uio/vfio intr/eventfd mapping */
1278         rte_intr_enable(&pci_dev->intr_handle);
1279
1280         /* enable support intr */
1281         ixgbe_enable_intr(eth_dev);
1282
1283         /* initialize 5tuple filter list */
1284         TAILQ_INIT(&filter_info->fivetuple_list);
1285         memset(filter_info->fivetuple_mask, 0,
1286                sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1287
1288         return 0;
1289 }
1290
1291 static int
1292 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1293 {
1294         struct rte_pci_device *pci_dev;
1295         struct ixgbe_hw *hw;
1296
1297         PMD_INIT_FUNC_TRACE();
1298
1299         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1300                 return -EPERM;
1301
1302         hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1303         pci_dev = eth_dev->pci_dev;
1304
1305         if (hw->adapter_stopped == 0)
1306                 ixgbe_dev_close(eth_dev);
1307
1308         eth_dev->dev_ops = NULL;
1309         eth_dev->rx_pkt_burst = NULL;
1310         eth_dev->tx_pkt_burst = NULL;
1311
1312         /* Unlock any pending hardware semaphore */
1313         ixgbe_swfw_lock_reset(hw);
1314
1315         /* disable uio intr before callback unregister */
1316         rte_intr_disable(&(pci_dev->intr_handle));
1317         rte_intr_callback_unregister(&(pci_dev->intr_handle),
1318                 ixgbe_dev_interrupt_handler, (void *)eth_dev);
1319
1320         /* uninitialize PF if max_vfs not zero */
1321         ixgbe_pf_host_uninit(eth_dev);
1322
1323         rte_free(eth_dev->data->mac_addrs);
1324         eth_dev->data->mac_addrs = NULL;
1325
1326         rte_free(eth_dev->data->hash_mac_addrs);
1327         eth_dev->data->hash_mac_addrs = NULL;
1328
1329         return 0;
1330 }
1331
1332 /*
1333  * Negotiate mailbox API version with the PF.
1334  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1335  * Then we try to negotiate starting with the most recent one.
1336  * If all negotiation attempts fail, then we will proceed with
1337  * the default one (ixgbe_mbox_api_10).
1338  */
1339 static void
1340 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1341 {
1342         int32_t i;
1343
1344         /* start with highest supported, proceed down */
1345         static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1346                 ixgbe_mbox_api_12,
1347                 ixgbe_mbox_api_11,
1348                 ixgbe_mbox_api_10,
1349         };
1350
1351         for (i = 0;
1352                         i != RTE_DIM(sup_ver) &&
1353                         ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1354                         i++)
1355                 ;
1356 }
1357
1358 static void
1359 generate_random_mac_addr(struct ether_addr *mac_addr)
1360 {
1361         uint64_t random;
1362
1363         /* Set Organizationally Unique Identifier (OUI) prefix. */
1364         mac_addr->addr_bytes[0] = 0x00;
1365         mac_addr->addr_bytes[1] = 0x09;
1366         mac_addr->addr_bytes[2] = 0xC0;
1367         /* Force indication of locally assigned MAC address. */
1368         mac_addr->addr_bytes[0] |= ETHER_LOCAL_ADMIN_ADDR;
1369         /* Generate the last 3 bytes of the MAC address with a random number. */
1370         random = rte_rand();
1371         memcpy(&mac_addr->addr_bytes[3], &random, 3);
1372 }
1373
1374 /*
1375  * Virtual Function device init
1376  */
1377 static int
1378 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1379 {
1380         int diag;
1381         uint32_t tc, tcs;
1382         struct rte_pci_device *pci_dev;
1383         struct ixgbe_hw *hw =
1384                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1385         struct ixgbe_vfta *shadow_vfta =
1386                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1387         struct ixgbe_hwstrip *hwstrip =
1388                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1389         struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
1390
1391         PMD_INIT_FUNC_TRACE();
1392
1393         eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1394         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1395         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1396
1397         /* for secondary processes, we don't initialise any further as primary
1398          * has already done this work. Only check we don't need a different
1399          * RX function
1400          */
1401         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1402                 struct ixgbe_tx_queue *txq;
1403                 /* TX queue function in primary, set by last queue initialized
1404                  * Tx queue may not initialized by primary process
1405                  */
1406                 if (eth_dev->data->tx_queues) {
1407                         txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1408                         ixgbe_set_tx_function(eth_dev, txq);
1409                 } else {
1410                         /* Use default TX function if we get here */
1411                         PMD_INIT_LOG(NOTICE,
1412                                      "No TX queues configured yet. Using default TX function.");
1413                 }
1414
1415                 ixgbe_set_rx_function(eth_dev);
1416
1417                 return 0;
1418         }
1419
1420         pci_dev = eth_dev->pci_dev;
1421
1422         rte_eth_copy_pci_info(eth_dev, pci_dev);
1423
1424         hw->device_id = pci_dev->id.device_id;
1425         hw->vendor_id = pci_dev->id.vendor_id;
1426         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1427
1428         /* initialize the vfta */
1429         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1430
1431         /* initialize the hw strip bitmap*/
1432         memset(hwstrip, 0, sizeof(*hwstrip));
1433
1434         /* Initialize the shared code (base driver) */
1435         diag = ixgbe_init_shared_code(hw);
1436         if (diag != IXGBE_SUCCESS) {
1437                 PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1438                 return -EIO;
1439         }
1440
1441         /* init_mailbox_params */
1442         hw->mbx.ops.init_params(hw);
1443
1444         /* Reset the hw statistics */
1445         ixgbevf_dev_stats_reset(eth_dev);
1446
1447         /* Disable the interrupts for VF */
1448         ixgbevf_intr_disable(hw);
1449
1450         hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1451         diag = hw->mac.ops.reset_hw(hw);
1452
1453         /*
1454          * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1455          * the underlying PF driver has not assigned a MAC address to the VF.
1456          * In this case, assign a random MAC address.
1457          */
1458         if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1459                 PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1460                 return diag;
1461         }
1462
1463         /* negotiate mailbox API version to use with the PF. */
1464         ixgbevf_negotiate_api(hw);
1465
1466         /* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1467         ixgbevf_get_queues(hw, &tcs, &tc);
1468
1469         /* Allocate memory for storing MAC addresses */
1470         eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", ETHER_ADDR_LEN *
1471                                                hw->mac.num_rar_entries, 0);
1472         if (eth_dev->data->mac_addrs == NULL) {
1473                 PMD_INIT_LOG(ERR,
1474                              "Failed to allocate %u bytes needed to store "
1475                              "MAC addresses",
1476                              ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1477                 return -ENOMEM;
1478         }
1479
1480         /* Generate a random MAC address, if none was assigned by PF. */
1481         if (is_zero_ether_addr(perm_addr)) {
1482                 generate_random_mac_addr(perm_addr);
1483                 diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1484                 if (diag) {
1485                         rte_free(eth_dev->data->mac_addrs);
1486                         eth_dev->data->mac_addrs = NULL;
1487                         return diag;
1488                 }
1489                 PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1490                 PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1491                              "%02x:%02x:%02x:%02x:%02x:%02x",
1492                              perm_addr->addr_bytes[0],
1493                              perm_addr->addr_bytes[1],
1494                              perm_addr->addr_bytes[2],
1495                              perm_addr->addr_bytes[3],
1496                              perm_addr->addr_bytes[4],
1497                              perm_addr->addr_bytes[5]);
1498         }
1499
1500         /* Copy the permanent MAC address */
1501         ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1502
1503         /* reset the hardware with the new settings */
1504         diag = hw->mac.ops.start_hw(hw);
1505         switch (diag) {
1506         case  0:
1507                 break;
1508
1509         default:
1510                 PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1511                 return -EIO;
1512         }
1513
1514         rte_intr_callback_register(&pci_dev->intr_handle,
1515                                    ixgbevf_dev_interrupt_handler,
1516                                    (void *)eth_dev);
1517         rte_intr_enable(&pci_dev->intr_handle);
1518         ixgbevf_intr_enable(hw);
1519
1520         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1521                      eth_dev->data->port_id, pci_dev->id.vendor_id,
1522                      pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1523
1524         return 0;
1525 }
1526
1527 /* Virtual Function device uninit */
1528
1529 static int
1530 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1531 {
1532         struct ixgbe_hw *hw;
1533         struct rte_pci_device *pci_dev = eth_dev->pci_dev;
1534
1535         PMD_INIT_FUNC_TRACE();
1536
1537         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1538                 return -EPERM;
1539
1540         hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1541
1542         if (hw->adapter_stopped == 0)
1543                 ixgbevf_dev_close(eth_dev);
1544
1545         eth_dev->dev_ops = NULL;
1546         eth_dev->rx_pkt_burst = NULL;
1547         eth_dev->tx_pkt_burst = NULL;
1548
1549         /* Disable the interrupts for VF */
1550         ixgbevf_intr_disable(hw);
1551
1552         rte_free(eth_dev->data->mac_addrs);
1553         eth_dev->data->mac_addrs = NULL;
1554
1555         rte_intr_disable(&pci_dev->intr_handle);
1556         rte_intr_callback_unregister(&pci_dev->intr_handle,
1557                                      ixgbevf_dev_interrupt_handler,
1558                                      (void *)eth_dev);
1559
1560         return 0;
1561 }
1562
1563 static struct eth_driver rte_ixgbe_pmd = {
1564         .pci_drv = {
1565                 .id_table = pci_id_ixgbe_map,
1566                 .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
1567                         RTE_PCI_DRV_DETACHABLE,
1568                 .probe = rte_eth_dev_pci_probe,
1569                 .remove = rte_eth_dev_pci_remove,
1570         },
1571         .eth_dev_init = eth_ixgbe_dev_init,
1572         .eth_dev_uninit = eth_ixgbe_dev_uninit,
1573         .dev_private_size = sizeof(struct ixgbe_adapter),
1574 };
1575
1576 /*
1577  * virtual function driver struct
1578  */
1579 static struct eth_driver rte_ixgbevf_pmd = {
1580         .pci_drv = {
1581                 .id_table = pci_id_ixgbevf_map,
1582                 .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_DETACHABLE,
1583                 .probe = rte_eth_dev_pci_probe,
1584                 .remove = rte_eth_dev_pci_remove,
1585         },
1586         .eth_dev_init = eth_ixgbevf_dev_init,
1587         .eth_dev_uninit = eth_ixgbevf_dev_uninit,
1588         .dev_private_size = sizeof(struct ixgbe_adapter),
1589 };
1590
1591 static int
1592 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1593 {
1594         struct ixgbe_hw *hw =
1595                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1596         struct ixgbe_vfta *shadow_vfta =
1597                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1598         uint32_t vfta;
1599         uint32_t vid_idx;
1600         uint32_t vid_bit;
1601
1602         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1603         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1604         vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1605         if (on)
1606                 vfta |= vid_bit;
1607         else
1608                 vfta &= ~vid_bit;
1609         IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1610
1611         /* update local VFTA copy */
1612         shadow_vfta->vfta[vid_idx] = vfta;
1613
1614         return 0;
1615 }
1616
1617 static void
1618 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1619 {
1620         if (on)
1621                 ixgbe_vlan_hw_strip_enable(dev, queue);
1622         else
1623                 ixgbe_vlan_hw_strip_disable(dev, queue);
1624 }
1625
1626 static int
1627 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1628                     enum rte_vlan_type vlan_type,
1629                     uint16_t tpid)
1630 {
1631         struct ixgbe_hw *hw =
1632                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1633         int ret = 0;
1634         uint32_t reg;
1635         uint32_t qinq;
1636
1637         qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1638         qinq &= IXGBE_DMATXCTL_GDV;
1639
1640         switch (vlan_type) {
1641         case ETH_VLAN_TYPE_INNER:
1642                 if (qinq) {
1643                         reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1644                         reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1645                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1646                         reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1647                         reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1648                                 | ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1649                         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1650                 } else {
1651                         ret = -ENOTSUP;
1652                         PMD_DRV_LOG(ERR, "Inner type is not supported"
1653                                     " by single VLAN");
1654                 }
1655                 break;
1656         case ETH_VLAN_TYPE_OUTER:
1657                 if (qinq) {
1658                         /* Only the high 16-bits is valid */
1659                         IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1660                                         IXGBE_EXVET_VET_EXT_SHIFT);
1661                 } else {
1662                         reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1663                         reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1664                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1665                         reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1666                         reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1667                                 | ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1668                         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1669                 }
1670
1671                 break;
1672         default:
1673                 ret = -EINVAL;
1674                 PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1675                 break;
1676         }
1677
1678         return ret;
1679 }
1680
1681 void
1682 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1683 {
1684         struct ixgbe_hw *hw =
1685                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1686         uint32_t vlnctrl;
1687
1688         PMD_INIT_FUNC_TRACE();
1689
1690         /* Filter Table Disable */
1691         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1692         vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1693
1694         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1695 }
1696
1697 void
1698 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1699 {
1700         struct ixgbe_hw *hw =
1701                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1702         struct ixgbe_vfta *shadow_vfta =
1703                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1704         uint32_t vlnctrl;
1705         uint16_t i;
1706
1707         PMD_INIT_FUNC_TRACE();
1708
1709         /* Filter Table Enable */
1710         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1711         vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1712         vlnctrl |= IXGBE_VLNCTRL_VFE;
1713
1714         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1715
1716         /* write whatever is in local vfta copy */
1717         for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1718                 IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1719 }
1720
1721 static void
1722 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1723 {
1724         struct ixgbe_hwstrip *hwstrip =
1725                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1726         struct ixgbe_rx_queue *rxq;
1727
1728         if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
1729                 return;
1730
1731         if (on)
1732                 IXGBE_SET_HWSTRIP(hwstrip, queue);
1733         else
1734                 IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1735
1736         if (queue >= dev->data->nb_rx_queues)
1737                 return;
1738
1739         rxq = dev->data->rx_queues[queue];
1740
1741         if (on)
1742                 rxq->vlan_flags = PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED;
1743         else
1744                 rxq->vlan_flags = PKT_RX_VLAN_PKT;
1745 }
1746
1747 static void
1748 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
1749 {
1750         struct ixgbe_hw *hw =
1751                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1752         uint32_t ctrl;
1753
1754         PMD_INIT_FUNC_TRACE();
1755
1756         if (hw->mac.type == ixgbe_mac_82598EB) {
1757                 /* No queue level support */
1758                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1759                 return;
1760         }
1761
1762         /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1763         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1764         ctrl &= ~IXGBE_RXDCTL_VME;
1765         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1766
1767         /* record those setting for HW strip per queue */
1768         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
1769 }
1770
1771 static void
1772 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
1773 {
1774         struct ixgbe_hw *hw =
1775                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1776         uint32_t ctrl;
1777
1778         PMD_INIT_FUNC_TRACE();
1779
1780         if (hw->mac.type == ixgbe_mac_82598EB) {
1781                 /* No queue level supported */
1782                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1783                 return;
1784         }
1785
1786         /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1787         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1788         ctrl |= IXGBE_RXDCTL_VME;
1789         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1790
1791         /* record those setting for HW strip per queue */
1792         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
1793 }
1794
1795 void
1796 ixgbe_vlan_hw_strip_disable_all(struct rte_eth_dev *dev)
1797 {
1798         struct ixgbe_hw *hw =
1799                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1800         uint32_t ctrl;
1801         uint16_t i;
1802         struct ixgbe_rx_queue *rxq;
1803
1804         PMD_INIT_FUNC_TRACE();
1805
1806         if (hw->mac.type == ixgbe_mac_82598EB) {
1807                 ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1808                 ctrl &= ~IXGBE_VLNCTRL_VME;
1809                 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1810         } else {
1811                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1812                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
1813                         rxq = dev->data->rx_queues[i];
1814                         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
1815                         ctrl &= ~IXGBE_RXDCTL_VME;
1816                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
1817
1818                         /* record those setting for HW strip per queue */
1819                         ixgbe_vlan_hw_strip_bitmap_set(dev, i, 0);
1820                 }
1821         }
1822 }
1823
1824 void
1825 ixgbe_vlan_hw_strip_enable_all(struct rte_eth_dev *dev)
1826 {
1827         struct ixgbe_hw *hw =
1828                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1829         uint32_t ctrl;
1830         uint16_t i;
1831         struct ixgbe_rx_queue *rxq;
1832
1833         PMD_INIT_FUNC_TRACE();
1834
1835         if (hw->mac.type == ixgbe_mac_82598EB) {
1836                 ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1837                 ctrl |= IXGBE_VLNCTRL_VME;
1838                 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1839         } else {
1840                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1841                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
1842                         rxq = dev->data->rx_queues[i];
1843                         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
1844                         ctrl |= IXGBE_RXDCTL_VME;
1845                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
1846
1847                         /* record those setting for HW strip per queue */
1848                         ixgbe_vlan_hw_strip_bitmap_set(dev, i, 1);
1849                 }
1850         }
1851 }
1852
1853 static void
1854 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
1855 {
1856         struct ixgbe_hw *hw =
1857                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1858         uint32_t ctrl;
1859
1860         PMD_INIT_FUNC_TRACE();
1861
1862         /* DMATXCTRL: Geric Double VLAN Disable */
1863         ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1864         ctrl &= ~IXGBE_DMATXCTL_GDV;
1865         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
1866
1867         /* CTRL_EXT: Global Double VLAN Disable */
1868         ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1869         ctrl &= ~IXGBE_EXTENDED_VLAN;
1870         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
1871
1872 }
1873
1874 static void
1875 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
1876 {
1877         struct ixgbe_hw *hw =
1878                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1879         uint32_t ctrl;
1880
1881         PMD_INIT_FUNC_TRACE();
1882
1883         /* DMATXCTRL: Geric Double VLAN Enable */
1884         ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1885         ctrl |= IXGBE_DMATXCTL_GDV;
1886         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
1887
1888         /* CTRL_EXT: Global Double VLAN Enable */
1889         ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1890         ctrl |= IXGBE_EXTENDED_VLAN;
1891         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
1892
1893         /* Clear pooling mode of PFVTCTL. It's required by X550. */
1894         if (hw->mac.type == ixgbe_mac_X550 ||
1895             hw->mac.type == ixgbe_mac_X550EM_x ||
1896             hw->mac.type == ixgbe_mac_X550EM_a) {
1897                 ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
1898                 ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
1899                 IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
1900         }
1901
1902         /*
1903          * VET EXT field in the EXVET register = 0x8100 by default
1904          * So no need to change. Same to VT field of DMATXCTL register
1905          */
1906 }
1907
1908 static void
1909 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1910 {
1911         if (mask & ETH_VLAN_STRIP_MASK) {
1912                 if (dev->data->dev_conf.rxmode.hw_vlan_strip)
1913                         ixgbe_vlan_hw_strip_enable_all(dev);
1914                 else
1915                         ixgbe_vlan_hw_strip_disable_all(dev);
1916         }
1917
1918         if (mask & ETH_VLAN_FILTER_MASK) {
1919                 if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1920                         ixgbe_vlan_hw_filter_enable(dev);
1921                 else
1922                         ixgbe_vlan_hw_filter_disable(dev);
1923         }
1924
1925         if (mask & ETH_VLAN_EXTEND_MASK) {
1926                 if (dev->data->dev_conf.rxmode.hw_vlan_extend)
1927                         ixgbe_vlan_hw_extend_enable(dev);
1928                 else
1929                         ixgbe_vlan_hw_extend_disable(dev);
1930         }
1931 }
1932
1933 static void
1934 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1935 {
1936         struct ixgbe_hw *hw =
1937                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1938         /* VLNCTRL: enable vlan filtering and allow all vlan tags through */
1939         uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1940
1941         vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
1942         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
1943 }
1944
1945 static int
1946 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
1947 {
1948         switch (nb_rx_q) {
1949         case 1:
1950         case 2:
1951                 RTE_ETH_DEV_SRIOV(dev).active = ETH_64_POOLS;
1952                 break;
1953         case 4:
1954                 RTE_ETH_DEV_SRIOV(dev).active = ETH_32_POOLS;
1955                 break;
1956         default:
1957                 return -EINVAL;
1958         }
1959
1960         RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = nb_rx_q;
1961         RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx = dev->pci_dev->max_vfs * nb_rx_q;
1962
1963         return 0;
1964 }
1965
1966 static int
1967 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
1968 {
1969         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
1970         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1971         uint16_t nb_rx_q = dev->data->nb_rx_queues;
1972         uint16_t nb_tx_q = dev->data->nb_tx_queues;
1973
1974         if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
1975                 /* check multi-queue mode */
1976                 switch (dev_conf->rxmode.mq_mode) {
1977                 case ETH_MQ_RX_VMDQ_DCB:
1978                 case ETH_MQ_RX_VMDQ_DCB_RSS:
1979                         /* DCB/RSS VMDQ in SRIOV mode, not implement yet */
1980                         PMD_INIT_LOG(ERR, "SRIOV active,"
1981                                         " unsupported mq_mode rx %d.",
1982                                         dev_conf->rxmode.mq_mode);
1983                         return -EINVAL;
1984                 case ETH_MQ_RX_RSS:
1985                 case ETH_MQ_RX_VMDQ_RSS:
1986                         dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_RSS;
1987                         if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
1988                                 if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
1989                                         PMD_INIT_LOG(ERR, "SRIOV is active,"
1990                                                 " invalid queue number"
1991                                                 " for VMDQ RSS, allowed"
1992                                                 " value are 1, 2 or 4.");
1993                                         return -EINVAL;
1994                                 }
1995                         break;
1996                 case ETH_MQ_RX_VMDQ_ONLY:
1997                 case ETH_MQ_RX_NONE:
1998                         /* if nothing mq mode configure, use default scheme */
1999                         dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
2000                         if (RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool > 1)
2001                                 RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = 1;
2002                         break;
2003                 default: /* ETH_MQ_RX_DCB, ETH_MQ_RX_DCB_RSS or ETH_MQ_TX_DCB*/
2004                         /* SRIOV only works in VMDq enable mode */
2005                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2006                                         " wrong mq_mode rx %d.",
2007                                         dev_conf->rxmode.mq_mode);
2008                         return -EINVAL;
2009                 }
2010
2011                 switch (dev_conf->txmode.mq_mode) {
2012                 case ETH_MQ_TX_VMDQ_DCB:
2013                         /* DCB VMDQ in SRIOV mode, not implement yet */
2014                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2015                                         " unsupported VMDQ mq_mode tx %d.",
2016                                         dev_conf->txmode.mq_mode);
2017                         return -EINVAL;
2018                 default: /* ETH_MQ_TX_VMDQ_ONLY or ETH_MQ_TX_NONE */
2019                         dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_ONLY;
2020                         break;
2021                 }
2022
2023                 /* check valid queue number */
2024                 if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2025                     (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2026                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2027                                         " nb_rx_q=%d nb_tx_q=%d queue number"
2028                                         " must be less than or equal to %d.",
2029                                         nb_rx_q, nb_tx_q,
2030                                         RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2031                         return -EINVAL;
2032                 }
2033         } else {
2034                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB_RSS) {
2035                         PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2036                                           " not supported.");
2037                         return -EINVAL;
2038                 }
2039                 /* check configuration for vmdb+dcb mode */
2040                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB) {
2041                         const struct rte_eth_vmdq_dcb_conf *conf;
2042
2043                         if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2044                                 PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2045                                                 IXGBE_VMDQ_DCB_NB_QUEUES);
2046                                 return -EINVAL;
2047                         }
2048                         conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2049                         if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2050                                conf->nb_queue_pools == ETH_32_POOLS)) {
2051                                 PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2052                                                 " nb_queue_pools must be %d or %d.",
2053                                                 ETH_16_POOLS, ETH_32_POOLS);
2054                                 return -EINVAL;
2055                         }
2056                 }
2057                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_VMDQ_DCB) {
2058                         const struct rte_eth_vmdq_dcb_tx_conf *conf;
2059
2060                         if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2061                                 PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2062                                                  IXGBE_VMDQ_DCB_NB_QUEUES);
2063                                 return -EINVAL;
2064                         }
2065                         conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2066                         if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2067                                conf->nb_queue_pools == ETH_32_POOLS)) {
2068                                 PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2069                                                 " nb_queue_pools != %d and"
2070                                                 " nb_queue_pools != %d.",
2071                                                 ETH_16_POOLS, ETH_32_POOLS);
2072                                 return -EINVAL;
2073                         }
2074                 }
2075
2076                 /* For DCB mode check our configuration before we go further */
2077                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB) {
2078                         const struct rte_eth_dcb_rx_conf *conf;
2079
2080                         if (nb_rx_q != IXGBE_DCB_NB_QUEUES) {
2081                                 PMD_INIT_LOG(ERR, "DCB selected, nb_rx_q != %d.",
2082                                                  IXGBE_DCB_NB_QUEUES);
2083                                 return -EINVAL;
2084                         }
2085                         conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2086                         if (!(conf->nb_tcs == ETH_4_TCS ||
2087                                conf->nb_tcs == ETH_8_TCS)) {
2088                                 PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2089                                                 " and nb_tcs != %d.",
2090                                                 ETH_4_TCS, ETH_8_TCS);
2091                                 return -EINVAL;
2092                         }
2093                 }
2094
2095                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_DCB) {
2096                         const struct rte_eth_dcb_tx_conf *conf;
2097
2098                         if (nb_tx_q != IXGBE_DCB_NB_QUEUES) {
2099                                 PMD_INIT_LOG(ERR, "DCB, nb_tx_q != %d.",
2100                                                  IXGBE_DCB_NB_QUEUES);
2101                                 return -EINVAL;
2102                         }
2103                         conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2104                         if (!(conf->nb_tcs == ETH_4_TCS ||
2105                                conf->nb_tcs == ETH_8_TCS)) {
2106                                 PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2107                                                 " and nb_tcs != %d.",
2108                                                 ETH_4_TCS, ETH_8_TCS);
2109                                 return -EINVAL;
2110                         }
2111                 }
2112
2113                 /*
2114                  * When DCB/VT is off, maximum number of queues changes,
2115                  * except for 82598EB, which remains constant.
2116                  */
2117                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
2118                                 hw->mac.type != ixgbe_mac_82598EB) {
2119                         if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2120                                 PMD_INIT_LOG(ERR,
2121                                              "Neither VT nor DCB are enabled, "
2122                                              "nb_tx_q > %d.",
2123                                              IXGBE_NONE_MODE_TX_NB_QUEUES);
2124                                 return -EINVAL;
2125                         }
2126                 }
2127         }
2128         return 0;
2129 }
2130
2131 static int
2132 ixgbe_dev_configure(struct rte_eth_dev *dev)
2133 {
2134         struct ixgbe_interrupt *intr =
2135                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2136         struct ixgbe_adapter *adapter =
2137                 (struct ixgbe_adapter *)dev->data->dev_private;
2138         int ret;
2139
2140         PMD_INIT_FUNC_TRACE();
2141         /* multipe queue mode checking */
2142         ret  = ixgbe_check_mq_mode(dev);
2143         if (ret != 0) {
2144                 PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2145                             ret);
2146                 return ret;
2147         }
2148
2149         /* set flag to update link status after init */
2150         intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2151
2152         /*
2153          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2154          * allocation or vector Rx preconditions we will reset it.
2155          */
2156         adapter->rx_bulk_alloc_allowed = true;
2157         adapter->rx_vec_allowed = true;
2158
2159         return 0;
2160 }
2161
2162 static void
2163 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2164 {
2165         struct ixgbe_hw *hw =
2166                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2167         struct ixgbe_interrupt *intr =
2168                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2169         uint32_t gpie;
2170
2171         /* only set up it on X550EM_X */
2172         if (hw->mac.type == ixgbe_mac_X550EM_x) {
2173                 gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2174                 gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2175                 IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2176                 if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2177                         intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2178         }
2179 }
2180
2181 /*
2182  * Configure device link speed and setup link.
2183  * It returns 0 on success.
2184  */
2185 static int
2186 ixgbe_dev_start(struct rte_eth_dev *dev)
2187 {
2188         struct ixgbe_hw *hw =
2189                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2190         struct ixgbe_vf_info *vfinfo =
2191                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2192         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
2193         uint32_t intr_vector = 0;
2194         int err, link_up = 0, negotiate = 0;
2195         uint32_t speed = 0;
2196         int mask = 0;
2197         int status;
2198         uint16_t vf, idx;
2199         uint32_t *link_speeds;
2200
2201         PMD_INIT_FUNC_TRACE();
2202
2203         /* IXGBE devices don't support:
2204         *    - half duplex (checked afterwards for valid speeds)
2205         *    - fixed speed: TODO implement
2206         */
2207         if (dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED) {
2208                 PMD_INIT_LOG(ERR, "Invalid link_speeds for port %hhu; fix speed not supported",
2209                              dev->data->port_id);
2210                 return -EINVAL;
2211         }
2212
2213         /* disable uio/vfio intr/eventfd mapping */
2214         rte_intr_disable(intr_handle);
2215
2216         /* stop adapter */
2217         hw->adapter_stopped = 0;
2218         ixgbe_stop_adapter(hw);
2219
2220         /* reinitialize adapter
2221          * this calls reset and start
2222          */
2223         status = ixgbe_pf_reset_hw(hw);
2224         if (status != 0)
2225                 return -1;
2226         hw->mac.ops.start_hw(hw);
2227         hw->mac.get_link_status = true;
2228
2229         /* configure PF module if SRIOV enabled */
2230         ixgbe_pf_host_configure(dev);
2231
2232         ixgbe_dev_phy_intr_setup(dev);
2233
2234         /* check and configure queue intr-vector mapping */
2235         if ((rte_intr_cap_multiple(intr_handle) ||
2236              !RTE_ETH_DEV_SRIOV(dev).active) &&
2237             dev->data->dev_conf.intr_conf.rxq != 0) {
2238                 intr_vector = dev->data->nb_rx_queues;
2239                 if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2240                         PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2241                                         IXGBE_MAX_INTR_QUEUE_NUM);
2242                         return -ENOTSUP;
2243                 }
2244                 if (rte_intr_efd_enable(intr_handle, intr_vector))
2245                         return -1;
2246         }
2247
2248         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
2249                 intr_handle->intr_vec =
2250                         rte_zmalloc("intr_vec",
2251                                     dev->data->nb_rx_queues * sizeof(int), 0);
2252                 if (intr_handle->intr_vec == NULL) {
2253                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2254                                      " intr_vec\n", dev->data->nb_rx_queues);
2255                         return -ENOMEM;
2256                 }
2257         }
2258
2259         /* confiugre msix for sleep until rx interrupt */
2260         ixgbe_configure_msix(dev);
2261
2262         /* initialize transmission unit */
2263         ixgbe_dev_tx_init(dev);
2264
2265         /* This can fail when allocating mbufs for descriptor rings */
2266         err = ixgbe_dev_rx_init(dev);
2267         if (err) {
2268                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2269                 goto error;
2270         }
2271
2272     mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
2273                 ETH_VLAN_EXTEND_MASK;
2274         ixgbe_vlan_offload_set(dev, mask);
2275
2276         if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
2277                 /* Enable vlan filtering for VMDq */
2278                 ixgbe_vmdq_vlan_hw_filter_enable(dev);
2279         }
2280
2281         /* Configure DCB hw */
2282         ixgbe_configure_dcb(dev);
2283
2284         if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2285                 err = ixgbe_fdir_configure(dev);
2286                 if (err)
2287                         goto error;
2288         }
2289
2290         /* Restore vf rate limit */
2291         if (vfinfo != NULL) {
2292                 for (vf = 0; vf < dev->pci_dev->max_vfs; vf++)
2293                         for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2294                                 if (vfinfo[vf].tx_rate[idx] != 0)
2295                                         ixgbe_set_vf_rate_limit(dev, vf,
2296                                                 vfinfo[vf].tx_rate[idx],
2297                                                 1 << idx);
2298         }
2299
2300         ixgbe_restore_statistics_mapping(dev);
2301
2302         err = ixgbe_dev_rxtx_start(dev);
2303         if (err < 0) {
2304                 PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2305                 goto error;
2306         }
2307
2308         /* Skip link setup if loopback mode is enabled for 82599. */
2309         if (hw->mac.type == ixgbe_mac_82599EB &&
2310                         dev->data->dev_conf.lpbk_mode == IXGBE_LPBK_82599_TX_RX)
2311                 goto skip_link_setup;
2312
2313         if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2314                 err = hw->mac.ops.setup_sfp(hw);
2315                 if (err)
2316                         goto error;
2317         }
2318
2319         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2320                 /* Turn on the copper */
2321                 ixgbe_set_phy_power(hw, true);
2322         } else {
2323                 /* Turn on the laser */
2324                 ixgbe_enable_tx_laser(hw);
2325         }
2326
2327         err = ixgbe_check_link(hw, &speed, &link_up, 0);
2328         if (err)
2329                 goto error;
2330         dev->data->dev_link.link_status = link_up;
2331
2332         err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2333         if (err)
2334                 goto error;
2335
2336         link_speeds = &dev->data->dev_conf.link_speeds;
2337         if (*link_speeds & ~(ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2338                         ETH_LINK_SPEED_10G)) {
2339                 PMD_INIT_LOG(ERR, "Invalid link setting");
2340                 goto error;
2341         }
2342
2343         speed = 0x0;
2344         if (*link_speeds == ETH_LINK_SPEED_AUTONEG) {
2345                 speed = (hw->mac.type != ixgbe_mac_82598EB) ?
2346                                 IXGBE_LINK_SPEED_82599_AUTONEG :
2347                                 IXGBE_LINK_SPEED_82598_AUTONEG;
2348         } else {
2349                 if (*link_speeds & ETH_LINK_SPEED_10G)
2350                         speed |= IXGBE_LINK_SPEED_10GB_FULL;
2351                 if (*link_speeds & ETH_LINK_SPEED_1G)
2352                         speed |= IXGBE_LINK_SPEED_1GB_FULL;
2353                 if (*link_speeds & ETH_LINK_SPEED_100M)
2354                         speed |= IXGBE_LINK_SPEED_100_FULL;
2355         }
2356
2357         err = ixgbe_setup_link(hw, speed, link_up);
2358         if (err)
2359                 goto error;
2360
2361 skip_link_setup:
2362
2363         if (rte_intr_allow_others(intr_handle)) {
2364                 /* check if lsc interrupt is enabled */
2365                 if (dev->data->dev_conf.intr_conf.lsc != 0)
2366                         ixgbe_dev_lsc_interrupt_setup(dev);
2367         } else {
2368                 rte_intr_callback_unregister(intr_handle,
2369                                              ixgbe_dev_interrupt_handler,
2370                                              (void *)dev);
2371                 if (dev->data->dev_conf.intr_conf.lsc != 0)
2372                         PMD_INIT_LOG(INFO, "lsc won't enable because of"
2373                                      " no intr multiplex\n");
2374         }
2375
2376         /* check if rxq interrupt is enabled */
2377         if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2378             rte_intr_dp_is_en(intr_handle))
2379                 ixgbe_dev_rxq_interrupt_setup(dev);
2380
2381         /* enable uio/vfio intr/eventfd mapping */
2382         rte_intr_enable(intr_handle);
2383
2384         /* resume enabled intr since hw reset */
2385         ixgbe_enable_intr(dev);
2386
2387         return 0;
2388
2389 error:
2390         PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2391         ixgbe_dev_clear_queues(dev);
2392         return -EIO;
2393 }
2394
2395 /*
2396  * Stop device: disable rx and tx functions to allow for reconfiguring.
2397  */
2398 static void
2399 ixgbe_dev_stop(struct rte_eth_dev *dev)
2400 {
2401         struct rte_eth_link link;
2402         struct ixgbe_hw *hw =
2403                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2404         struct ixgbe_vf_info *vfinfo =
2405                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2406         struct ixgbe_filter_info *filter_info =
2407                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
2408         struct ixgbe_5tuple_filter *p_5tuple, *p_5tuple_next;
2409         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
2410         int vf;
2411
2412         PMD_INIT_FUNC_TRACE();
2413
2414         /* disable interrupts */
2415         ixgbe_disable_intr(hw);
2416
2417         /* reset the NIC */
2418         ixgbe_pf_reset_hw(hw);
2419         hw->adapter_stopped = 0;
2420
2421         /* stop adapter */
2422         ixgbe_stop_adapter(hw);
2423
2424         for (vf = 0; vfinfo != NULL &&
2425                      vf < dev->pci_dev->max_vfs; vf++)
2426                 vfinfo[vf].clear_to_send = false;
2427
2428         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2429                 /* Turn off the copper */
2430                 ixgbe_set_phy_power(hw, false);
2431         } else {
2432                 /* Turn off the laser */
2433                 ixgbe_disable_tx_laser(hw);
2434         }
2435
2436         ixgbe_dev_clear_queues(dev);
2437
2438         /* Clear stored conf */
2439         dev->data->scattered_rx = 0;
2440         dev->data->lro = 0;
2441
2442         /* Clear recorded link status */
2443         memset(&link, 0, sizeof(link));
2444         rte_ixgbe_dev_atomic_write_link_status(dev, &link);
2445
2446         /* Remove all ntuple filters of the device */
2447         for (p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list);
2448              p_5tuple != NULL; p_5tuple = p_5tuple_next) {
2449                 p_5tuple_next = TAILQ_NEXT(p_5tuple, entries);
2450                 TAILQ_REMOVE(&filter_info->fivetuple_list,
2451                              p_5tuple, entries);
2452                 rte_free(p_5tuple);
2453         }
2454         memset(filter_info->fivetuple_mask, 0,
2455                 sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
2456
2457         if (!rte_intr_allow_others(intr_handle))
2458                 /* resume to the default handler */
2459                 rte_intr_callback_register(intr_handle,
2460                                            ixgbe_dev_interrupt_handler,
2461                                            (void *)dev);
2462
2463         /* Clean datapath event and queue/vec mapping */
2464         rte_intr_efd_disable(intr_handle);
2465         if (intr_handle->intr_vec != NULL) {
2466                 rte_free(intr_handle->intr_vec);
2467                 intr_handle->intr_vec = NULL;
2468         }
2469 }
2470
2471 /*
2472  * Set device link up: enable tx.
2473  */
2474 static int
2475 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2476 {
2477         struct ixgbe_hw *hw =
2478                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2479         if (hw->mac.type == ixgbe_mac_82599EB) {
2480 #ifdef RTE_NIC_BYPASS
2481                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2482                         /* Not suported in bypass mode */
2483                         PMD_INIT_LOG(ERR, "Set link up is not supported "
2484                                      "by device id 0x%x", hw->device_id);
2485                         return -ENOTSUP;
2486                 }
2487 #endif
2488         }
2489
2490         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2491                 /* Turn on the copper */
2492                 ixgbe_set_phy_power(hw, true);
2493         } else {
2494                 /* Turn on the laser */
2495                 ixgbe_enable_tx_laser(hw);
2496         }
2497
2498         return 0;
2499 }
2500
2501 /*
2502  * Set device link down: disable tx.
2503  */
2504 static int
2505 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2506 {
2507         struct ixgbe_hw *hw =
2508                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2509         if (hw->mac.type == ixgbe_mac_82599EB) {
2510 #ifdef RTE_NIC_BYPASS
2511                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2512                         /* Not suported in bypass mode */
2513                         PMD_INIT_LOG(ERR, "Set link down is not supported "
2514                                      "by device id 0x%x", hw->device_id);
2515                         return -ENOTSUP;
2516                 }
2517 #endif
2518         }
2519
2520         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2521                 /* Turn off the copper */
2522                 ixgbe_set_phy_power(hw, false);
2523         } else {
2524                 /* Turn off the laser */
2525                 ixgbe_disable_tx_laser(hw);
2526         }
2527
2528         return 0;
2529 }
2530
2531 /*
2532  * Reest and stop device.
2533  */
2534 static void
2535 ixgbe_dev_close(struct rte_eth_dev *dev)
2536 {
2537         struct ixgbe_hw *hw =
2538                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2539
2540         PMD_INIT_FUNC_TRACE();
2541
2542         ixgbe_pf_reset_hw(hw);
2543
2544         ixgbe_dev_stop(dev);
2545         hw->adapter_stopped = 1;
2546
2547         ixgbe_dev_free_queues(dev);
2548
2549         ixgbe_disable_pcie_master(hw);
2550
2551         /* reprogram the RAR[0] in case user changed it. */
2552         ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
2553 }
2554
2555 static void
2556 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
2557                            struct ixgbe_hw_stats *hw_stats,
2558                            uint64_t *total_missed_rx, uint64_t *total_qbrc,
2559                            uint64_t *total_qprc, uint64_t *total_qprdc)
2560 {
2561         uint32_t bprc, lxon, lxoff, total;
2562         uint32_t delta_gprc = 0;
2563         unsigned i;
2564         /* Workaround for RX byte count not including CRC bytes when CRC
2565 +        * strip is enabled. CRC bytes are removed from counters when crc_strip
2566          * is disabled.
2567 +        */
2568         int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
2569                         IXGBE_HLREG0_RXCRCSTRP);
2570
2571         hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
2572         hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
2573         hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
2574         hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
2575
2576         for (i = 0; i < 8; i++) {
2577                 uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
2578
2579                 /* global total per queue */
2580                 hw_stats->mpc[i] += mp;
2581                 /* Running comprehensive total for stats display */
2582                 *total_missed_rx += hw_stats->mpc[i];
2583                 if (hw->mac.type == ixgbe_mac_82598EB) {
2584                         hw_stats->rnbc[i] +=
2585                             IXGBE_READ_REG(hw, IXGBE_RNBC(i));
2586                         hw_stats->pxonrxc[i] +=
2587                                 IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
2588                         hw_stats->pxoffrxc[i] +=
2589                                 IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
2590                 } else {
2591                         hw_stats->pxonrxc[i] +=
2592                                 IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
2593                         hw_stats->pxoffrxc[i] +=
2594                                 IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
2595                         hw_stats->pxon2offc[i] +=
2596                                 IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
2597                 }
2598                 hw_stats->pxontxc[i] +=
2599                     IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
2600                 hw_stats->pxofftxc[i] +=
2601                     IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
2602         }
2603         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
2604                 uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
2605                 uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
2606                 uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
2607
2608                 delta_gprc += delta_qprc;
2609
2610                 hw_stats->qprc[i] += delta_qprc;
2611                 hw_stats->qptc[i] += delta_qptc;
2612
2613                 hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
2614                 hw_stats->qbrc[i] +=
2615                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
2616                 if (crc_strip == 0)
2617                         hw_stats->qbrc[i] -= delta_qprc * ETHER_CRC_LEN;
2618
2619                 hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
2620                 hw_stats->qbtc[i] +=
2621                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
2622
2623                 hw_stats->qprdc[i] += delta_qprdc;
2624                 *total_qprdc += hw_stats->qprdc[i];
2625
2626                 *total_qprc += hw_stats->qprc[i];
2627                 *total_qbrc += hw_stats->qbrc[i];
2628         }
2629         hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
2630         hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
2631         hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
2632
2633         /*
2634          * An errata states that gprc actually counts good + missed packets:
2635          * Workaround to set gprc to summated queue packet receives
2636          */
2637         hw_stats->gprc = *total_qprc;
2638
2639         if (hw->mac.type != ixgbe_mac_82598EB) {
2640                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
2641                 hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
2642                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
2643                 hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
2644                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
2645                 hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
2646                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
2647                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
2648         } else {
2649                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
2650                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
2651                 /* 82598 only has a counter in the high register */
2652                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
2653                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
2654                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
2655         }
2656         uint64_t old_tpr = hw_stats->tpr;
2657
2658         hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
2659         hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
2660
2661         if (crc_strip == 0)
2662                 hw_stats->gorc -= delta_gprc * ETHER_CRC_LEN;
2663
2664         uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
2665         hw_stats->gptc += delta_gptc;
2666         hw_stats->gotc -= delta_gptc * ETHER_CRC_LEN;
2667         hw_stats->tor -= (hw_stats->tpr - old_tpr) * ETHER_CRC_LEN;
2668
2669         /*
2670          * Workaround: mprc hardware is incorrectly counting
2671          * broadcasts, so for now we subtract those.
2672          */
2673         bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
2674         hw_stats->bprc += bprc;
2675         hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
2676         if (hw->mac.type == ixgbe_mac_82598EB)
2677                 hw_stats->mprc -= bprc;
2678
2679         hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
2680         hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
2681         hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
2682         hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
2683         hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
2684         hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
2685
2686         lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
2687         hw_stats->lxontxc += lxon;
2688         lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
2689         hw_stats->lxofftxc += lxoff;
2690         total = lxon + lxoff;
2691
2692         hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
2693         hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
2694         hw_stats->gptc -= total;
2695         hw_stats->mptc -= total;
2696         hw_stats->ptc64 -= total;
2697         hw_stats->gotc -= total * ETHER_MIN_LEN;
2698
2699         hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
2700         hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
2701         hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
2702         hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
2703         hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
2704         hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
2705         hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
2706         hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
2707         hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
2708         hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
2709         hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
2710         hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
2711         hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
2712         hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
2713         hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
2714         hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
2715         /* Only read FCOE on 82599 */
2716         if (hw->mac.type != ixgbe_mac_82598EB) {
2717                 hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
2718                 hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
2719                 hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
2720                 hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
2721                 hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
2722         }
2723
2724         /* Flow Director Stats registers */
2725         hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
2726         hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
2727 }
2728
2729 /*
2730  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
2731  */
2732 static void
2733 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
2734 {
2735         struct ixgbe_hw *hw =
2736                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2737         struct ixgbe_hw_stats *hw_stats =
2738                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2739         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
2740         unsigned i;
2741
2742         total_missed_rx = 0;
2743         total_qbrc = 0;
2744         total_qprc = 0;
2745         total_qprdc = 0;
2746
2747         ixgbe_read_stats_registers(hw, hw_stats, &total_missed_rx, &total_qbrc,
2748                         &total_qprc, &total_qprdc);
2749
2750         if (stats == NULL)
2751                 return;
2752
2753         /* Fill out the rte_eth_stats statistics structure */
2754         stats->ipackets = total_qprc;
2755         stats->ibytes = total_qbrc;
2756         stats->opackets = hw_stats->gptc;
2757         stats->obytes = hw_stats->gotc;
2758
2759         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
2760                 stats->q_ipackets[i] = hw_stats->qprc[i];
2761                 stats->q_opackets[i] = hw_stats->qptc[i];
2762                 stats->q_ibytes[i] = hw_stats->qbrc[i];
2763                 stats->q_obytes[i] = hw_stats->qbtc[i];
2764                 stats->q_errors[i] = hw_stats->qprdc[i];
2765         }
2766
2767         /* Rx Errors */
2768         stats->imissed  = total_missed_rx;
2769         stats->ierrors  = hw_stats->crcerrs +
2770                           hw_stats->mspdc +
2771                           hw_stats->rlec +
2772                           hw_stats->ruc +
2773                           hw_stats->roc +
2774                           hw_stats->illerrc +
2775                           hw_stats->errbc +
2776                           hw_stats->rfc +
2777                           hw_stats->fccrc +
2778                           hw_stats->fclast;
2779
2780         /* Tx Errors */
2781         stats->oerrors  = 0;
2782 }
2783
2784 static void
2785 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
2786 {
2787         struct ixgbe_hw_stats *stats =
2788                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2789
2790         /* HW registers are cleared on read */
2791         ixgbe_dev_stats_get(dev, NULL);
2792
2793         /* Reset software totals */
2794         memset(stats, 0, sizeof(*stats));
2795 }
2796
2797 /* This function calculates the number of xstats based on the current config */
2798 static unsigned
2799 ixgbe_xstats_calc_num(void) {
2800         return IXGBE_NB_HW_STATS +
2801                 (IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
2802                 (IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
2803 }
2804
2805 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
2806         struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned limit)
2807 {
2808         const unsigned cnt_stats = ixgbe_xstats_calc_num();
2809         unsigned stat, i, count;
2810
2811         if (xstats_names != NULL) {
2812                 count = 0;
2813
2814                 /* Note: limit >= cnt_stats checked upstream
2815                  * in rte_eth_xstats_names()
2816                  */
2817
2818                 /* Extended stats from ixgbe_hw_stats */
2819                 for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
2820                         snprintf(xstats_names[count].name,
2821                                 sizeof(xstats_names[count].name),
2822                                 "%s",
2823                                 rte_ixgbe_stats_strings[i].name);
2824                         count++;
2825                 }
2826
2827                 /* RX Priority Stats */
2828                 for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
2829                         for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
2830                                 snprintf(xstats_names[count].name,
2831                                         sizeof(xstats_names[count].name),
2832                                         "rx_priority%u_%s", i,
2833                                         rte_ixgbe_rxq_strings[stat].name);
2834                                 count++;
2835                         }
2836                 }
2837
2838                 /* TX Priority Stats */
2839                 for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
2840                         for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
2841                                 snprintf(xstats_names[count].name,
2842                                         sizeof(xstats_names[count].name),
2843                                         "tx_priority%u_%s", i,
2844                                         rte_ixgbe_txq_strings[stat].name);
2845                                 count++;
2846                         }
2847                 }
2848         }
2849         return cnt_stats;
2850 }
2851
2852 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
2853         struct rte_eth_xstat_name *xstats_names, unsigned limit)
2854 {
2855         unsigned i;
2856
2857         if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
2858                 return -ENOMEM;
2859
2860         if (xstats_names != NULL)
2861                 for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
2862                         snprintf(xstats_names[i].name,
2863                                 sizeof(xstats_names[i].name),
2864                                 "%s", rte_ixgbevf_stats_strings[i].name);
2865         return IXGBEVF_NB_XSTATS;
2866 }
2867
2868 static int
2869 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
2870                                          unsigned n)
2871 {
2872         struct ixgbe_hw *hw =
2873                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2874         struct ixgbe_hw_stats *hw_stats =
2875                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2876         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
2877         unsigned i, stat, count = 0;
2878
2879         count = ixgbe_xstats_calc_num();
2880
2881         if (n < count)
2882                 return count;
2883
2884         total_missed_rx = 0;
2885         total_qbrc = 0;
2886         total_qprc = 0;
2887         total_qprdc = 0;
2888
2889         ixgbe_read_stats_registers(hw, hw_stats, &total_missed_rx, &total_qbrc,
2890                                    &total_qprc, &total_qprdc);
2891
2892         /* If this is a reset xstats is NULL, and we have cleared the
2893          * registers by reading them.
2894          */
2895         if (!xstats)
2896                 return 0;
2897
2898         /* Extended stats from ixgbe_hw_stats */
2899         count = 0;
2900         for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
2901                 xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
2902                                 rte_ixgbe_stats_strings[i].offset);
2903                 count++;
2904         }
2905
2906         /* RX Priority Stats */
2907         for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
2908                 for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
2909                         xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
2910                                         rte_ixgbe_rxq_strings[stat].offset +
2911                                         (sizeof(uint64_t) * i));
2912                         count++;
2913                 }
2914         }
2915
2916         /* TX Priority Stats */
2917         for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
2918                 for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
2919                         xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
2920                                         rte_ixgbe_txq_strings[stat].offset +
2921                                         (sizeof(uint64_t) * i));
2922                         count++;
2923                 }
2924         }
2925         return count;
2926 }
2927
2928 static void
2929 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
2930 {
2931         struct ixgbe_hw_stats *stats =
2932                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2933
2934         unsigned count = ixgbe_xstats_calc_num();
2935
2936         /* HW registers are cleared on read */
2937         ixgbe_dev_xstats_get(dev, NULL, count);
2938
2939         /* Reset software totals */
2940         memset(stats, 0, sizeof(*stats));
2941 }
2942
2943 static void
2944 ixgbevf_update_stats(struct rte_eth_dev *dev)
2945 {
2946         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2947         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
2948                           IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2949
2950         /* Good Rx packet, include VF loopback */
2951         UPDATE_VF_STAT(IXGBE_VFGPRC,
2952             hw_stats->last_vfgprc, hw_stats->vfgprc);
2953
2954         /* Good Rx octets, include VF loopback */
2955         UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
2956             hw_stats->last_vfgorc, hw_stats->vfgorc);
2957
2958         /* Good Tx packet, include VF loopback */
2959         UPDATE_VF_STAT(IXGBE_VFGPTC,
2960             hw_stats->last_vfgptc, hw_stats->vfgptc);
2961
2962         /* Good Tx octets, include VF loopback */
2963         UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
2964             hw_stats->last_vfgotc, hw_stats->vfgotc);
2965
2966         /* Rx Multicst Packet */
2967         UPDATE_VF_STAT(IXGBE_VFMPRC,
2968             hw_stats->last_vfmprc, hw_stats->vfmprc);
2969 }
2970
2971 static int
2972 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
2973                        unsigned n)
2974 {
2975         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
2976                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2977         unsigned i;
2978
2979         if (n < IXGBEVF_NB_XSTATS)
2980                 return IXGBEVF_NB_XSTATS;
2981
2982         ixgbevf_update_stats(dev);
2983
2984         if (!xstats)
2985                 return 0;
2986
2987         /* Extended stats */
2988         for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
2989                 xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
2990                         rte_ixgbevf_stats_strings[i].offset);
2991         }
2992
2993         return IXGBEVF_NB_XSTATS;
2994 }
2995
2996 static void
2997 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
2998 {
2999         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3000                           IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3001
3002         ixgbevf_update_stats(dev);
3003
3004         if (stats == NULL)
3005                 return;
3006
3007         stats->ipackets = hw_stats->vfgprc;
3008         stats->ibytes = hw_stats->vfgorc;
3009         stats->opackets = hw_stats->vfgptc;
3010         stats->obytes = hw_stats->vfgotc;
3011 }
3012
3013 static void
3014 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3015 {
3016         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3017                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3018
3019         /* Sync HW register to the last stats */
3020         ixgbevf_dev_stats_get(dev, NULL);
3021
3022         /* reset HW current stats*/
3023         hw_stats->vfgprc = 0;
3024         hw_stats->vfgorc = 0;
3025         hw_stats->vfgptc = 0;
3026         hw_stats->vfgotc = 0;
3027 }
3028
3029 static void
3030 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3031 {
3032         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3033         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3034
3035         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3036         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3037         if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3038                 /*
3039                  * When DCB/VT is off, maximum number of queues changes,
3040                  * except for 82598EB, which remains constant.
3041                  */
3042                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
3043                                 hw->mac.type != ixgbe_mac_82598EB)
3044                         dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3045         }
3046         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3047         dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3048         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3049         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3050         dev_info->max_vfs = dev->pci_dev->max_vfs;
3051         if (hw->mac.type == ixgbe_mac_82598EB)
3052                 dev_info->max_vmdq_pools = ETH_16_POOLS;
3053         else
3054                 dev_info->max_vmdq_pools = ETH_64_POOLS;
3055         dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3056         dev_info->rx_offload_capa =
3057                 DEV_RX_OFFLOAD_VLAN_STRIP |
3058                 DEV_RX_OFFLOAD_IPV4_CKSUM |
3059                 DEV_RX_OFFLOAD_UDP_CKSUM  |
3060                 DEV_RX_OFFLOAD_TCP_CKSUM;
3061
3062         /*
3063          * RSC is only supported by 82599 and x540 PF devices in a non-SR-IOV
3064          * mode.
3065          */
3066         if ((hw->mac.type == ixgbe_mac_82599EB ||
3067              hw->mac.type == ixgbe_mac_X540) &&
3068             !RTE_ETH_DEV_SRIOV(dev).active)
3069                 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TCP_LRO;
3070
3071         if (hw->mac.type == ixgbe_mac_X550 ||
3072             hw->mac.type == ixgbe_mac_X550EM_x ||
3073             hw->mac.type == ixgbe_mac_X550EM_a)
3074                 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM;
3075
3076         dev_info->tx_offload_capa =
3077                 DEV_TX_OFFLOAD_VLAN_INSERT |
3078                 DEV_TX_OFFLOAD_IPV4_CKSUM  |
3079                 DEV_TX_OFFLOAD_UDP_CKSUM   |
3080                 DEV_TX_OFFLOAD_TCP_CKSUM   |
3081                 DEV_TX_OFFLOAD_SCTP_CKSUM  |
3082                 DEV_TX_OFFLOAD_TCP_TSO;
3083
3084         if (hw->mac.type == ixgbe_mac_X550 ||
3085             hw->mac.type == ixgbe_mac_X550EM_x ||
3086             hw->mac.type == ixgbe_mac_X550EM_a)
3087                 dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM;
3088
3089         dev_info->default_rxconf = (struct rte_eth_rxconf) {
3090                 .rx_thresh = {
3091                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3092                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3093                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3094                 },
3095                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3096                 .rx_drop_en = 0,
3097         };
3098
3099         dev_info->default_txconf = (struct rte_eth_txconf) {
3100                 .tx_thresh = {
3101                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3102                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3103                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3104                 },
3105                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3106                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3107                 .txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
3108                                 ETH_TXQ_FLAGS_NOOFFLOADS,
3109         };
3110
3111         dev_info->rx_desc_lim = rx_desc_lim;
3112         dev_info->tx_desc_lim = tx_desc_lim;
3113
3114         dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3115         dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3116         dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3117
3118         dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G;
3119         if (hw->mac.type == ixgbe_mac_X540 ||
3120             hw->mac.type == ixgbe_mac_X540_vf ||
3121             hw->mac.type == ixgbe_mac_X550 ||
3122             hw->mac.type == ixgbe_mac_X550_vf) {
3123                 dev_info->speed_capa |= ETH_LINK_SPEED_100M;
3124         }
3125 }
3126
3127 static const uint32_t *
3128 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3129 {
3130         static const uint32_t ptypes[] = {
3131                 /* For non-vec functions,
3132                  * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3133                  * for vec functions,
3134                  * refers to _recv_raw_pkts_vec().
3135                  */
3136                 RTE_PTYPE_L2_ETHER,
3137                 RTE_PTYPE_L3_IPV4,
3138                 RTE_PTYPE_L3_IPV4_EXT,
3139                 RTE_PTYPE_L3_IPV6,
3140                 RTE_PTYPE_L3_IPV6_EXT,
3141                 RTE_PTYPE_L4_SCTP,
3142                 RTE_PTYPE_L4_TCP,
3143                 RTE_PTYPE_L4_UDP,
3144                 RTE_PTYPE_TUNNEL_IP,
3145                 RTE_PTYPE_INNER_L3_IPV6,
3146                 RTE_PTYPE_INNER_L3_IPV6_EXT,
3147                 RTE_PTYPE_INNER_L4_TCP,
3148                 RTE_PTYPE_INNER_L4_UDP,
3149                 RTE_PTYPE_UNKNOWN
3150         };
3151
3152         if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3153             dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3154             dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3155             dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3156                 return ptypes;
3157         return NULL;
3158 }
3159
3160 static void
3161 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3162                      struct rte_eth_dev_info *dev_info)
3163 {
3164         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3165
3166         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3167         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3168         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3169         dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS reg */
3170         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3171         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3172         dev_info->max_vfs = dev->pci_dev->max_vfs;
3173         if (hw->mac.type == ixgbe_mac_82598EB)
3174                 dev_info->max_vmdq_pools = ETH_16_POOLS;
3175         else
3176                 dev_info->max_vmdq_pools = ETH_64_POOLS;
3177         dev_info->rx_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP |
3178                                 DEV_RX_OFFLOAD_IPV4_CKSUM |
3179                                 DEV_RX_OFFLOAD_UDP_CKSUM  |
3180                                 DEV_RX_OFFLOAD_TCP_CKSUM;
3181         dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
3182                                 DEV_TX_OFFLOAD_IPV4_CKSUM  |
3183                                 DEV_TX_OFFLOAD_UDP_CKSUM   |
3184                                 DEV_TX_OFFLOAD_TCP_CKSUM   |
3185                                 DEV_TX_OFFLOAD_SCTP_CKSUM  |
3186                                 DEV_TX_OFFLOAD_TCP_TSO;
3187
3188         dev_info->default_rxconf = (struct rte_eth_rxconf) {
3189                 .rx_thresh = {
3190                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3191                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3192                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3193                 },
3194                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3195                 .rx_drop_en = 0,
3196         };
3197
3198         dev_info->default_txconf = (struct rte_eth_txconf) {
3199                 .tx_thresh = {
3200                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3201                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3202                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3203                 },
3204                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3205                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3206                 .txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
3207                                 ETH_TXQ_FLAGS_NOOFFLOADS,
3208         };
3209
3210         dev_info->rx_desc_lim = rx_desc_lim;
3211         dev_info->tx_desc_lim = tx_desc_lim;
3212 }
3213
3214 /* return 0 means link status changed, -1 means not changed */
3215 static int
3216 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
3217 {
3218         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3219         struct rte_eth_link link, old;
3220         ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
3221         int link_up;
3222         int diag;
3223
3224         link.link_status = ETH_LINK_DOWN;
3225         link.link_speed = 0;
3226         link.link_duplex = ETH_LINK_HALF_DUPLEX;
3227         memset(&old, 0, sizeof(old));
3228         rte_ixgbe_dev_atomic_read_link_status(dev, &old);
3229
3230         hw->mac.get_link_status = true;
3231
3232         /* check if it needs to wait to complete, if lsc interrupt is enabled */
3233         if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
3234                 diag = ixgbe_check_link(hw, &link_speed, &link_up, 0);
3235         else
3236                 diag = ixgbe_check_link(hw, &link_speed, &link_up, 1);
3237
3238         if (diag != 0) {
3239                 link.link_speed = ETH_SPEED_NUM_100M;
3240                 link.link_duplex = ETH_LINK_FULL_DUPLEX;
3241                 rte_ixgbe_dev_atomic_write_link_status(dev, &link);
3242                 if (link.link_status == old.link_status)
3243                         return -1;
3244                 return 0;
3245         }
3246
3247         if (link_up == 0) {
3248                 rte_ixgbe_dev_atomic_write_link_status(dev, &link);
3249                 if (link.link_status == old.link_status)
3250                         return -1;
3251                 return 0;
3252         }
3253         link.link_status = ETH_LINK_UP;
3254         link.link_duplex = ETH_LINK_FULL_DUPLEX;
3255
3256         switch (link_speed) {
3257         default:
3258         case IXGBE_LINK_SPEED_UNKNOWN:
3259                 link.link_duplex = ETH_LINK_FULL_DUPLEX;
3260                 link.link_speed = ETH_SPEED_NUM_100M;
3261                 break;
3262
3263         case IXGBE_LINK_SPEED_100_FULL:
3264                 link.link_speed = ETH_SPEED_NUM_100M;
3265                 break;
3266
3267         case IXGBE_LINK_SPEED_1GB_FULL:
3268                 link.link_speed = ETH_SPEED_NUM_1G;
3269                 break;
3270
3271         case IXGBE_LINK_SPEED_10GB_FULL:
3272                 link.link_speed = ETH_SPEED_NUM_10G;
3273                 break;
3274         }
3275         rte_ixgbe_dev_atomic_write_link_status(dev, &link);
3276
3277         if (link.link_status == old.link_status)
3278                 return -1;
3279
3280         return 0;
3281 }
3282
3283 static void
3284 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
3285 {
3286         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3287         uint32_t fctrl;
3288
3289         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3290         fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
3291         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3292 }
3293
3294 static void
3295 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
3296 {
3297         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3298         uint32_t fctrl;
3299
3300         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3301         fctrl &= (~IXGBE_FCTRL_UPE);
3302         if (dev->data->all_multicast == 1)
3303                 fctrl |= IXGBE_FCTRL_MPE;
3304         else
3305                 fctrl &= (~IXGBE_FCTRL_MPE);
3306         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3307 }
3308
3309 static void
3310 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
3311 {
3312         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3313         uint32_t fctrl;
3314
3315         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3316         fctrl |= IXGBE_FCTRL_MPE;
3317         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3318 }
3319
3320 static void
3321 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
3322 {
3323         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3324         uint32_t fctrl;
3325
3326         if (dev->data->promiscuous == 1)
3327                 return; /* must remain in all_multicast mode */
3328
3329         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
3330         fctrl &= (~IXGBE_FCTRL_MPE);
3331         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
3332 }
3333
3334 /**
3335  * It clears the interrupt causes and enables the interrupt.
3336  * It will be called once only during nic initialized.
3337  *
3338  * @param dev
3339  *  Pointer to struct rte_eth_dev.
3340  *
3341  * @return
3342  *  - On success, zero.
3343  *  - On failure, a negative value.
3344  */
3345 static int
3346 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev)
3347 {
3348         struct ixgbe_interrupt *intr =
3349                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3350
3351         ixgbe_dev_link_status_print(dev);
3352         intr->mask |= IXGBE_EICR_LSC;
3353
3354         return 0;
3355 }
3356
3357 /**
3358  * It clears the interrupt causes and enables the interrupt.
3359  * It will be called once only during nic initialized.
3360  *
3361  * @param dev
3362  *  Pointer to struct rte_eth_dev.
3363  *
3364  * @return
3365  *  - On success, zero.
3366  *  - On failure, a negative value.
3367  */
3368 static int
3369 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
3370 {
3371         struct ixgbe_interrupt *intr =
3372                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3373
3374         intr->mask |= IXGBE_EICR_RTX_QUEUE;
3375
3376         return 0;
3377 }
3378
3379 /*
3380  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
3381  *
3382  * @param dev
3383  *  Pointer to struct rte_eth_dev.
3384  *
3385  * @return
3386  *  - On success, zero.
3387  *  - On failure, a negative value.
3388  */
3389 static int
3390 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
3391 {
3392         uint32_t eicr;
3393         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3394         struct ixgbe_interrupt *intr =
3395                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3396
3397         /* clear all cause mask */
3398         ixgbe_disable_intr(hw);
3399
3400         /* read-on-clear nic registers here */
3401         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3402         PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
3403
3404         intr->flags = 0;
3405
3406         /* set flag for async link update */
3407         if (eicr & IXGBE_EICR_LSC)
3408                 intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
3409
3410         if (eicr & IXGBE_EICR_MAILBOX)
3411                 intr->flags |= IXGBE_FLAG_MAILBOX;
3412
3413         if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
3414             hw->phy.type == ixgbe_phy_x550em_ext_t &&
3415             (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
3416                 intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
3417
3418         return 0;
3419 }
3420
3421 /**
3422  * It gets and then prints the link status.
3423  *
3424  * @param dev
3425  *  Pointer to struct rte_eth_dev.
3426  *
3427  * @return
3428  *  - On success, zero.
3429  *  - On failure, a negative value.
3430  */
3431 static void
3432 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
3433 {
3434         struct rte_eth_link link;
3435
3436         memset(&link, 0, sizeof(link));
3437         rte_ixgbe_dev_atomic_read_link_status(dev, &link);
3438         if (link.link_status) {
3439                 PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
3440                                         (int)(dev->data->port_id),
3441                                         (unsigned)link.link_speed,
3442                         link.link_duplex == ETH_LINK_FULL_DUPLEX ?
3443                                         "full-duplex" : "half-duplex");
3444         } else {
3445                 PMD_INIT_LOG(INFO, " Port %d: Link Down",
3446                                 (int)(dev->data->port_id));
3447         }
3448         PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
3449                                 dev->pci_dev->addr.domain,
3450                                 dev->pci_dev->addr.bus,
3451                                 dev->pci_dev->addr.devid,
3452                                 dev->pci_dev->addr.function);
3453 }
3454
3455 /*
3456  * It executes link_update after knowing an interrupt occurred.
3457  *
3458  * @param dev
3459  *  Pointer to struct rte_eth_dev.
3460  *
3461  * @return
3462  *  - On success, zero.
3463  *  - On failure, a negative value.
3464  */
3465 static int
3466 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
3467 {
3468         struct ixgbe_interrupt *intr =
3469                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3470         int64_t timeout;
3471         struct rte_eth_link link;
3472         int intr_enable_delay = false;
3473         struct ixgbe_hw *hw =
3474                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3475
3476         PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
3477
3478         if (intr->flags & IXGBE_FLAG_MAILBOX) {
3479                 ixgbe_pf_mbx_process(dev);
3480                 intr->flags &= ~IXGBE_FLAG_MAILBOX;
3481         }
3482
3483         if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
3484                 ixgbe_handle_lasi(hw);
3485                 intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
3486         }
3487
3488         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
3489                 /* get the link status before link update, for predicting later */
3490                 memset(&link, 0, sizeof(link));
3491                 rte_ixgbe_dev_atomic_read_link_status(dev, &link);
3492
3493                 ixgbe_dev_link_update(dev, 0);
3494
3495                 /* likely to up */
3496                 if (!link.link_status)
3497                         /* handle it 1 sec later, wait it being stable */
3498                         timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
3499                 /* likely to down */
3500                 else
3501                         /* handle it 4 sec later, wait it being stable */
3502                         timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
3503
3504                 ixgbe_dev_link_status_print(dev);
3505
3506                 intr_enable_delay = true;
3507         }
3508
3509         if (intr_enable_delay) {
3510                 if (rte_eal_alarm_set(timeout * 1000,
3511                                       ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
3512                         PMD_DRV_LOG(ERR, "Error setting alarm");
3513         } else {
3514                 PMD_DRV_LOG(DEBUG, "enable intr immediately");
3515                 ixgbe_enable_intr(dev);
3516                 rte_intr_enable(&(dev->pci_dev->intr_handle));
3517         }
3518
3519
3520         return 0;
3521 }
3522
3523 /**
3524  * Interrupt handler which shall be registered for alarm callback for delayed
3525  * handling specific interrupt to wait for the stable nic state. As the
3526  * NIC interrupt state is not stable for ixgbe after link is just down,
3527  * it needs to wait 4 seconds to get the stable status.
3528  *
3529  * @param handle
3530  *  Pointer to interrupt handle.
3531  * @param param
3532  *  The address of parameter (struct rte_eth_dev *) regsitered before.
3533  *
3534  * @return
3535  *  void
3536  */
3537 static void
3538 ixgbe_dev_interrupt_delayed_handler(void *param)
3539 {
3540         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
3541         struct ixgbe_interrupt *intr =
3542                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3543         struct ixgbe_hw *hw =
3544                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3545         uint32_t eicr;
3546
3547         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3548         if (eicr & IXGBE_EICR_MAILBOX)
3549                 ixgbe_pf_mbx_process(dev);
3550
3551         if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
3552                 ixgbe_handle_lasi(hw);
3553                 intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
3554         }
3555
3556         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
3557                 ixgbe_dev_link_update(dev, 0);
3558                 intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
3559                 ixgbe_dev_link_status_print(dev);
3560                 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC);
3561         }
3562
3563         PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
3564         ixgbe_enable_intr(dev);
3565         rte_intr_enable(&(dev->pci_dev->intr_handle));
3566 }
3567
3568 /**
3569  * Interrupt handler triggered by NIC  for handling
3570  * specific interrupt.
3571  *
3572  * @param handle
3573  *  Pointer to interrupt handle.
3574  * @param param
3575  *  The address of parameter (struct rte_eth_dev *) regsitered before.
3576  *
3577  * @return
3578  *  void
3579  */
3580 static void
3581 ixgbe_dev_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
3582                             void *param)
3583 {
3584         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
3585
3586         ixgbe_dev_interrupt_get_status(dev);
3587         ixgbe_dev_interrupt_action(dev);
3588 }
3589
3590 static int
3591 ixgbe_dev_led_on(struct rte_eth_dev *dev)
3592 {
3593         struct ixgbe_hw *hw;
3594
3595         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3596         return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
3597 }
3598
3599 static int
3600 ixgbe_dev_led_off(struct rte_eth_dev *dev)
3601 {
3602         struct ixgbe_hw *hw;
3603
3604         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3605         return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
3606 }
3607
3608 static int
3609 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3610 {
3611         struct ixgbe_hw *hw;
3612         uint32_t mflcn_reg;
3613         uint32_t fccfg_reg;
3614         int rx_pause;
3615         int tx_pause;
3616
3617         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3618
3619         fc_conf->pause_time = hw->fc.pause_time;
3620         fc_conf->high_water = hw->fc.high_water[0];
3621         fc_conf->low_water = hw->fc.low_water[0];
3622         fc_conf->send_xon = hw->fc.send_xon;
3623         fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
3624
3625         /*
3626          * Return rx_pause status according to actual setting of
3627          * MFLCN register.
3628          */
3629         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
3630         if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
3631                 rx_pause = 1;
3632         else
3633                 rx_pause = 0;
3634
3635         /*
3636          * Return tx_pause status according to actual setting of
3637          * FCCFG register.
3638          */
3639         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
3640         if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
3641                 tx_pause = 1;
3642         else
3643                 tx_pause = 0;
3644
3645         if (rx_pause && tx_pause)
3646                 fc_conf->mode = RTE_FC_FULL;
3647         else if (rx_pause)
3648                 fc_conf->mode = RTE_FC_RX_PAUSE;
3649         else if (tx_pause)
3650                 fc_conf->mode = RTE_FC_TX_PAUSE;
3651         else
3652                 fc_conf->mode = RTE_FC_NONE;
3653
3654         return 0;
3655 }
3656
3657 static int
3658 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3659 {
3660         struct ixgbe_hw *hw;
3661         int err;
3662         uint32_t rx_buf_size;
3663         uint32_t max_high_water;
3664         uint32_t mflcn;
3665         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
3666                 ixgbe_fc_none,
3667                 ixgbe_fc_rx_pause,
3668                 ixgbe_fc_tx_pause,
3669                 ixgbe_fc_full
3670         };
3671
3672         PMD_INIT_FUNC_TRACE();
3673
3674         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3675         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
3676         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3677
3678         /*
3679          * At least reserve one Ethernet frame for watermark
3680          * high_water/low_water in kilo bytes for ixgbe
3681          */
3682         max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
3683         if ((fc_conf->high_water > max_high_water) ||
3684                 (fc_conf->high_water < fc_conf->low_water)) {
3685                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
3686                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
3687                 return -EINVAL;
3688         }
3689
3690         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
3691         hw->fc.pause_time     = fc_conf->pause_time;
3692         hw->fc.high_water[0]  = fc_conf->high_water;
3693         hw->fc.low_water[0]   = fc_conf->low_water;
3694         hw->fc.send_xon       = fc_conf->send_xon;
3695         hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
3696
3697         err = ixgbe_fc_enable(hw);
3698
3699         /* Not negotiated is not an error case */
3700         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED)) {
3701
3702                 /* check if we want to forward MAC frames - driver doesn't have native
3703                  * capability to do that, so we'll write the registers ourselves */
3704
3705                 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
3706
3707                 /* set or clear MFLCN.PMCF bit depending on configuration */
3708                 if (fc_conf->mac_ctrl_frame_fwd != 0)
3709                         mflcn |= IXGBE_MFLCN_PMCF;
3710                 else
3711                         mflcn &= ~IXGBE_MFLCN_PMCF;
3712
3713                 IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
3714                 IXGBE_WRITE_FLUSH(hw);
3715
3716                 return 0;
3717         }
3718
3719         PMD_INIT_LOG(ERR, "ixgbe_fc_enable = 0x%x", err);
3720         return -EIO;
3721 }
3722
3723 /**
3724  *  ixgbe_pfc_enable_generic - Enable flow control
3725  *  @hw: pointer to hardware structure
3726  *  @tc_num: traffic class number
3727  *  Enable flow control according to the current settings.
3728  */
3729 static int
3730 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
3731 {
3732         int ret_val = 0;
3733         uint32_t mflcn_reg, fccfg_reg;
3734         uint32_t reg;
3735         uint32_t fcrtl, fcrth;
3736         uint8_t i;
3737         uint8_t nb_rx_en;
3738
3739         /* Validate the water mark configuration */
3740         if (!hw->fc.pause_time) {
3741                 ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
3742                 goto out;
3743         }
3744
3745         /* Low water mark of zero causes XOFF floods */
3746         if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
3747                  /* High/Low water can not be 0 */
3748                 if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_water[tc_num])) {
3749                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
3750                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
3751                         goto out;
3752                 }
3753
3754                 if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
3755                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
3756                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
3757                         goto out;
3758                 }
3759         }
3760         /* Negotiate the fc mode to use */
3761         ixgbe_fc_autoneg(hw);
3762
3763         /* Disable any previous flow control settings */
3764         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
3765         mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
3766
3767         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
3768         fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
3769
3770         switch (hw->fc.current_mode) {
3771         case ixgbe_fc_none:
3772                 /*
3773                  * If the count of enabled RX Priority Flow control >1,
3774                  * and the TX pause can not be disabled
3775                  */
3776                 nb_rx_en = 0;
3777                 for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
3778                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
3779                         if (reg & IXGBE_FCRTH_FCEN)
3780                                 nb_rx_en++;
3781                 }
3782                 if (nb_rx_en > 1)
3783                         fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
3784                 break;
3785         case ixgbe_fc_rx_pause:
3786                 /*
3787                  * Rx Flow control is enabled and Tx Flow control is
3788                  * disabled by software override. Since there really
3789                  * isn't a way to advertise that we are capable of RX
3790                  * Pause ONLY, we will advertise that we support both
3791                  * symmetric and asymmetric Rx PAUSE.  Later, we will
3792                  * disable the adapter's ability to send PAUSE frames.
3793                  */
3794                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
3795                 /*
3796                  * If the count of enabled RX Priority Flow control >1,
3797                  * and the TX pause can not be disabled
3798                  */
3799                 nb_rx_en = 0;
3800                 for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
3801                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
3802                         if (reg & IXGBE_FCRTH_FCEN)
3803                                 nb_rx_en++;
3804                 }
3805                 if (nb_rx_en > 1)
3806                         fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
3807                 break;
3808         case ixgbe_fc_tx_pause:
3809                 /*
3810                  * Tx Flow control is enabled, and Rx Flow control is
3811                  * disabled by software override.
3812                  */
3813                 fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
3814                 break;
3815         case ixgbe_fc_full:
3816                 /* Flow control (both Rx and Tx) is enabled by SW override. */
3817                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
3818                 fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
3819                 break;
3820         default:
3821                 PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
3822                 ret_val = IXGBE_ERR_CONFIG;
3823                 goto out;
3824         }
3825
3826         /* Set 802.3x based flow control settings. */
3827         mflcn_reg |= IXGBE_MFLCN_DPF;
3828         IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
3829         IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
3830
3831         /* Set up and enable Rx high/low water mark thresholds, enable XON. */
3832         if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
3833                 hw->fc.high_water[tc_num]) {
3834                 fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
3835                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
3836                 fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
3837         } else {
3838                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
3839                 /*
3840                  * In order to prevent Tx hangs when the internal Tx
3841                  * switch is enabled we must set the high water mark
3842                  * to the maximum FCRTH value.  This allows the Tx
3843                  * switch to function even under heavy Rx workloads.
3844                  */
3845                 fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
3846         }
3847         IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
3848
3849         /* Configure pause time (2 TCs per register) */
3850         reg = hw->fc.pause_time * 0x00010001;
3851         for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
3852                 IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
3853
3854         /* Configure flow control refresh threshold value */
3855         IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
3856
3857 out:
3858         return ret_val;
3859 }
3860
3861 static int
3862 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
3863 {
3864         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3865         int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
3866
3867         if (hw->mac.type != ixgbe_mac_82598EB) {
3868                 ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
3869         }
3870         return ret_val;
3871 }
3872
3873 static int
3874 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
3875 {
3876         int err;
3877         uint32_t rx_buf_size;
3878         uint32_t max_high_water;
3879         uint8_t tc_num;
3880         uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
3881         struct ixgbe_hw *hw =
3882                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3883         struct ixgbe_dcb_config *dcb_config =
3884                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
3885
3886         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
3887                 ixgbe_fc_none,
3888                 ixgbe_fc_rx_pause,
3889                 ixgbe_fc_tx_pause,
3890                 ixgbe_fc_full
3891         };
3892
3893         PMD_INIT_FUNC_TRACE();
3894
3895         ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
3896         tc_num = map[pfc_conf->priority];
3897         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
3898         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3899         /*
3900          * At least reserve one Ethernet frame for watermark
3901          * high_water/low_water in kilo bytes for ixgbe
3902          */
3903         max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
3904         if ((pfc_conf->fc.high_water > max_high_water) ||
3905             (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
3906                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
3907                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
3908                 return -EINVAL;
3909         }
3910
3911         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
3912         hw->fc.pause_time = pfc_conf->fc.pause_time;
3913         hw->fc.send_xon = pfc_conf->fc.send_xon;
3914         hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
3915         hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
3916
3917         err = ixgbe_dcb_pfc_enable(dev, tc_num);
3918
3919         /* Not negotiated is not an error case */
3920         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
3921                 return 0;
3922
3923         PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
3924         return -EIO;
3925 }
3926
3927 static int
3928 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
3929                           struct rte_eth_rss_reta_entry64 *reta_conf,
3930                           uint16_t reta_size)
3931 {
3932         uint16_t i, sp_reta_size;
3933         uint8_t j, mask;
3934         uint32_t reta, r;
3935         uint16_t idx, shift;
3936         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3937         uint32_t reta_reg;
3938
3939         PMD_INIT_FUNC_TRACE();
3940
3941         if (!ixgbe_rss_update_sp(hw->mac.type)) {
3942                 PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
3943                         "NIC.");
3944                 return -ENOTSUP;
3945         }
3946
3947         sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
3948         if (reta_size != sp_reta_size) {
3949                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3950                         "(%d) doesn't match the number hardware can supported "
3951                         "(%d)\n", reta_size, sp_reta_size);
3952                 return -EINVAL;
3953         }
3954
3955         for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
3956                 idx = i / RTE_RETA_GROUP_SIZE;
3957                 shift = i % RTE_RETA_GROUP_SIZE;
3958                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3959                                                 IXGBE_4_BIT_MASK);
3960                 if (!mask)
3961                         continue;
3962                 reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
3963                 if (mask == IXGBE_4_BIT_MASK)
3964                         r = 0;
3965                 else
3966                         r = IXGBE_READ_REG(hw, reta_reg);
3967                 for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
3968                         if (mask & (0x1 << j))
3969                                 reta |= reta_conf[idx].reta[shift + j] <<
3970                                                         (CHAR_BIT * j);
3971                         else
3972                                 reta |= r & (IXGBE_8_BIT_MASK <<
3973                                                 (CHAR_BIT * j));
3974                 }
3975                 IXGBE_WRITE_REG(hw, reta_reg, reta);
3976         }
3977
3978         return 0;
3979 }
3980
3981 static int
3982 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
3983                          struct rte_eth_rss_reta_entry64 *reta_conf,
3984                          uint16_t reta_size)
3985 {
3986         uint16_t i, sp_reta_size;
3987         uint8_t j, mask;
3988         uint32_t reta;
3989         uint16_t idx, shift;
3990         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3991         uint32_t reta_reg;
3992
3993         PMD_INIT_FUNC_TRACE();
3994         sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
3995         if (reta_size != sp_reta_size) {
3996                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3997                         "(%d) doesn't match the number hardware can supported "
3998                         "(%d)\n", reta_size, sp_reta_size);
3999                 return -EINVAL;
4000         }
4001
4002         for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
4003                 idx = i / RTE_RETA_GROUP_SIZE;
4004                 shift = i % RTE_RETA_GROUP_SIZE;
4005                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
4006                                                 IXGBE_4_BIT_MASK);
4007                 if (!mask)
4008                         continue;
4009
4010                 reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
4011                 reta = IXGBE_READ_REG(hw, reta_reg);
4012                 for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
4013                         if (mask & (0x1 << j))
4014                                 reta_conf[idx].reta[shift + j] =
4015                                         ((reta >> (CHAR_BIT * j)) &
4016                                                 IXGBE_8_BIT_MASK);
4017                 }
4018         }
4019
4020         return 0;
4021 }
4022
4023 static void
4024 ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
4025                                 uint32_t index, uint32_t pool)
4026 {
4027         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4028         uint32_t enable_addr = 1;
4029
4030         ixgbe_set_rar(hw, index, mac_addr->addr_bytes, pool, enable_addr);
4031 }
4032
4033 static void
4034 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
4035 {
4036         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4037
4038         ixgbe_clear_rar(hw, index);
4039 }
4040
4041 static void
4042 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
4043 {
4044         ixgbe_remove_rar(dev, 0);
4045
4046         ixgbe_add_rar(dev, addr, 0, 0);
4047 }
4048
4049 static int
4050 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4051 {
4052         uint32_t hlreg0;
4053         uint32_t maxfrs;
4054         struct ixgbe_hw *hw;
4055         struct rte_eth_dev_info dev_info;
4056         uint32_t frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
4057
4058         ixgbe_dev_info_get(dev, &dev_info);
4059
4060         /* check that mtu is within the allowed range */
4061         if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
4062                 return -EINVAL;
4063
4064         /* refuse mtu that requires the support of scattered packets when this
4065          * feature has not been enabled before.
4066          */
4067         if (!dev->data->scattered_rx &&
4068             (frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
4069              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM))
4070                 return -EINVAL;
4071
4072         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4073         hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4074
4075         /* switch to jumbo mode if needed */
4076         if (frame_size > ETHER_MAX_LEN) {
4077                 dev->data->dev_conf.rxmode.jumbo_frame = 1;
4078                 hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4079         } else {
4080                 dev->data->dev_conf.rxmode.jumbo_frame = 0;
4081                 hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
4082         }
4083         IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4084
4085         /* update max frame size */
4086         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4087
4088         maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
4089         maxfrs &= 0x0000FFFF;
4090         maxfrs |= (dev->data->dev_conf.rxmode.max_rx_pkt_len << 16);
4091         IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
4092
4093         return 0;
4094 }
4095
4096 /*
4097  * Virtual Function operations
4098  */
4099 static void
4100 ixgbevf_intr_disable(struct ixgbe_hw *hw)
4101 {
4102         PMD_INIT_FUNC_TRACE();
4103
4104         /* Clear interrupt mask to stop from interrupts being generated */
4105         IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
4106
4107         IXGBE_WRITE_FLUSH(hw);
4108 }
4109
4110 static void
4111 ixgbevf_intr_enable(struct ixgbe_hw *hw)
4112 {
4113         PMD_INIT_FUNC_TRACE();
4114
4115         /* VF enable interrupt autoclean */
4116         IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
4117         IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
4118         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
4119
4120         IXGBE_WRITE_FLUSH(hw);
4121 }
4122
4123 static int
4124 ixgbevf_dev_configure(struct rte_eth_dev *dev)
4125 {
4126         struct rte_eth_conf *conf = &dev->data->dev_conf;
4127         struct ixgbe_adapter *adapter =
4128                         (struct ixgbe_adapter *)dev->data->dev_private;
4129
4130         PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
4131                      dev->data->port_id);
4132
4133         /*
4134          * VF has no ability to enable/disable HW CRC
4135          * Keep the persistent behavior the same as Host PF
4136          */
4137 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
4138         if (!conf->rxmode.hw_strip_crc) {
4139                 PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
4140                 conf->rxmode.hw_strip_crc = 1;
4141         }
4142 #else
4143         if (conf->rxmode.hw_strip_crc) {
4144                 PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
4145                 conf->rxmode.hw_strip_crc = 0;
4146         }
4147 #endif
4148
4149         /*
4150          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
4151          * allocation or vector Rx preconditions we will reset it.
4152          */
4153         adapter->rx_bulk_alloc_allowed = true;
4154         adapter->rx_vec_allowed = true;
4155
4156         return 0;
4157 }
4158
4159 static int
4160 ixgbevf_dev_start(struct rte_eth_dev *dev)
4161 {
4162         struct ixgbe_hw *hw =
4163                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4164         uint32_t intr_vector = 0;
4165         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
4166
4167         int err, mask = 0;
4168
4169         PMD_INIT_FUNC_TRACE();
4170
4171         hw->mac.ops.reset_hw(hw);
4172         hw->mac.get_link_status = true;
4173
4174         /* negotiate mailbox API version to use with the PF. */
4175         ixgbevf_negotiate_api(hw);
4176
4177         ixgbevf_dev_tx_init(dev);
4178
4179         /* This can fail when allocating mbufs for descriptor rings */
4180         err = ixgbevf_dev_rx_init(dev);
4181         if (err) {
4182                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
4183                 ixgbe_dev_clear_queues(dev);
4184                 return err;
4185         }
4186
4187         /* Set vfta */
4188         ixgbevf_set_vfta_all(dev, 1);
4189
4190         /* Set HW strip */
4191         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
4192                 ETH_VLAN_EXTEND_MASK;
4193         ixgbevf_vlan_offload_set(dev, mask);
4194
4195         ixgbevf_dev_rxtx_start(dev);
4196
4197         /* check and configure queue intr-vector mapping */
4198         if (dev->data->dev_conf.intr_conf.rxq != 0) {
4199                 intr_vector = dev->data->nb_rx_queues;
4200                 if (rte_intr_efd_enable(intr_handle, intr_vector))
4201                         return -1;
4202         }
4203
4204         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
4205                 intr_handle->intr_vec =
4206                         rte_zmalloc("intr_vec",
4207                                     dev->data->nb_rx_queues * sizeof(int), 0);
4208                 if (intr_handle->intr_vec == NULL) {
4209                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
4210                                      " intr_vec\n", dev->data->nb_rx_queues);
4211                         return -ENOMEM;
4212                 }
4213         }
4214         ixgbevf_configure_msix(dev);
4215
4216         rte_intr_enable(intr_handle);
4217
4218         /* Re-enable interrupt for VF */
4219         ixgbevf_intr_enable(hw);
4220
4221         return 0;
4222 }
4223
4224 static void
4225 ixgbevf_dev_stop(struct rte_eth_dev *dev)
4226 {
4227         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4228         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
4229
4230         PMD_INIT_FUNC_TRACE();
4231
4232         ixgbevf_intr_disable(hw);
4233
4234         hw->adapter_stopped = 1;
4235         ixgbe_stop_adapter(hw);
4236
4237         /*
4238           * Clear what we set, but we still keep shadow_vfta to
4239           * restore after device starts
4240           */
4241         ixgbevf_set_vfta_all(dev, 0);
4242
4243         /* Clear stored conf */
4244         dev->data->scattered_rx = 0;
4245
4246         ixgbe_dev_clear_queues(dev);
4247
4248         /* Clean datapath event and queue/vec mapping */
4249         rte_intr_efd_disable(intr_handle);
4250         if (intr_handle->intr_vec != NULL) {
4251                 rte_free(intr_handle->intr_vec);
4252                 intr_handle->intr_vec = NULL;
4253         }
4254 }
4255
4256 static void
4257 ixgbevf_dev_close(struct rte_eth_dev *dev)
4258 {
4259         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4260
4261         PMD_INIT_FUNC_TRACE();
4262
4263         ixgbe_reset_hw(hw);
4264
4265         ixgbevf_dev_stop(dev);
4266
4267         ixgbe_dev_free_queues(dev);
4268
4269         /**
4270          * Remove the VF MAC address ro ensure
4271          * that the VF traffic goes to the PF
4272          * after stop, close and detach of the VF
4273          **/
4274         ixgbevf_remove_mac_addr(dev, 0);
4275 }
4276
4277 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
4278 {
4279         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4280         struct ixgbe_vfta *shadow_vfta =
4281                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
4282         int i = 0, j = 0, vfta = 0, mask = 1;
4283
4284         for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
4285                 vfta = shadow_vfta->vfta[i];
4286                 if (vfta) {
4287                         mask = 1;
4288                         for (j = 0; j < 32; j++) {
4289                                 if (vfta & mask)
4290                                         ixgbe_set_vfta(hw, (i<<5)+j, 0,
4291                                                        on, false);
4292                                 mask <<= 1;
4293                         }
4294                 }
4295         }
4296
4297 }
4298
4299 static int
4300 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
4301 {
4302         struct ixgbe_hw *hw =
4303                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4304         struct ixgbe_vfta *shadow_vfta =
4305                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
4306         uint32_t vid_idx = 0;
4307         uint32_t vid_bit = 0;
4308         int ret = 0;
4309
4310         PMD_INIT_FUNC_TRACE();
4311
4312         /* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
4313         ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
4314         if (ret) {
4315                 PMD_INIT_LOG(ERR, "Unable to set VF vlan");
4316                 return ret;
4317         }
4318         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
4319         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
4320
4321         /* Save what we set and retore it after device reset */
4322         if (on)
4323                 shadow_vfta->vfta[vid_idx] |= vid_bit;
4324         else
4325                 shadow_vfta->vfta[vid_idx] &= ~vid_bit;
4326
4327         return 0;
4328 }
4329
4330 static void
4331 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
4332 {
4333         struct ixgbe_hw *hw =
4334                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4335         uint32_t ctrl;
4336
4337         PMD_INIT_FUNC_TRACE();
4338
4339         if (queue >= hw->mac.max_rx_queues)
4340                 return;
4341
4342         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
4343         if (on)
4344                 ctrl |= IXGBE_RXDCTL_VME;
4345         else
4346                 ctrl &= ~IXGBE_RXDCTL_VME;
4347         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
4348
4349         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
4350 }
4351
4352 static void
4353 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
4354 {
4355         struct ixgbe_hw *hw =
4356                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4357         uint16_t i;
4358         int on = 0;
4359
4360         /* VF function only support hw strip feature, others are not support */
4361         if (mask & ETH_VLAN_STRIP_MASK) {
4362                 on = !!(dev->data->dev_conf.rxmode.hw_vlan_strip);
4363
4364                 for (i = 0; i < hw->mac.max_rx_queues; i++)
4365                         ixgbevf_vlan_strip_queue_set(dev, i, on);
4366         }
4367 }
4368
4369 static int
4370 ixgbe_vmdq_mode_check(struct ixgbe_hw *hw)
4371 {
4372         uint32_t reg_val;
4373
4374         /* we only need to do this if VMDq is enabled */
4375         reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
4376         if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
4377                 PMD_INIT_LOG(ERR, "VMDq must be enabled for this setting");
4378                 return -1;
4379         }
4380
4381         return 0;
4382 }
4383
4384 static uint32_t
4385 ixgbe_uta_vector(struct ixgbe_hw *hw, struct ether_addr *uc_addr)
4386 {
4387         uint32_t vector = 0;
4388
4389         switch (hw->mac.mc_filter_type) {
4390         case 0:   /* use bits [47:36] of the address */
4391                 vector = ((uc_addr->addr_bytes[4] >> 4) |
4392                         (((uint16_t)uc_addr->addr_bytes[5]) << 4));
4393                 break;
4394         case 1:   /* use bits [46:35] of the address */
4395                 vector = ((uc_addr->addr_bytes[4] >> 3) |
4396                         (((uint16_t)uc_addr->addr_bytes[5]) << 5));
4397                 break;
4398         case 2:   /* use bits [45:34] of the address */
4399                 vector = ((uc_addr->addr_bytes[4] >> 2) |
4400                         (((uint16_t)uc_addr->addr_bytes[5]) << 6));
4401                 break;
4402         case 3:   /* use bits [43:32] of the address */
4403                 vector = ((uc_addr->addr_bytes[4]) |
4404                         (((uint16_t)uc_addr->addr_bytes[5]) << 8));
4405                 break;
4406         default:  /* Invalid mc_filter_type */
4407                 break;
4408         }
4409
4410         /* vector can only be 12-bits or boundary will be exceeded */
4411         vector &= 0xFFF;
4412         return vector;
4413 }
4414
4415 static int
4416 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
4417                         uint8_t on)
4418 {
4419         uint32_t vector;
4420         uint32_t uta_idx;
4421         uint32_t reg_val;
4422         uint32_t uta_shift;
4423         uint32_t rc;
4424         const uint32_t ixgbe_uta_idx_mask = 0x7F;
4425         const uint32_t ixgbe_uta_bit_shift = 5;
4426         const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
4427         const uint32_t bit1 = 0x1;
4428
4429         struct ixgbe_hw *hw =
4430                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4431         struct ixgbe_uta_info *uta_info =
4432                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
4433
4434         /* The UTA table only exists on 82599 hardware and newer */
4435         if (hw->mac.type < ixgbe_mac_82599EB)
4436                 return -ENOTSUP;
4437
4438         vector = ixgbe_uta_vector(hw, mac_addr);
4439         uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
4440         uta_shift = vector & ixgbe_uta_bit_mask;
4441
4442         rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
4443         if (rc == on)
4444                 return 0;
4445
4446         reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
4447         if (on) {
4448                 uta_info->uta_in_use++;
4449                 reg_val |= (bit1 << uta_shift);
4450                 uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
4451         } else {
4452                 uta_info->uta_in_use--;
4453                 reg_val &= ~(bit1 << uta_shift);
4454                 uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
4455         }
4456
4457         IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
4458
4459         if (uta_info->uta_in_use > 0)
4460                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
4461                                 IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
4462         else
4463                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
4464
4465         return 0;
4466 }
4467
4468 static int
4469 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
4470 {
4471         int i;
4472         struct ixgbe_hw *hw =
4473                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4474         struct ixgbe_uta_info *uta_info =
4475                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
4476
4477         /* The UTA table only exists on 82599 hardware and newer */
4478         if (hw->mac.type < ixgbe_mac_82599EB)
4479                 return -ENOTSUP;
4480
4481         if (on) {
4482                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
4483                         uta_info->uta_shadow[i] = ~0;
4484                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
4485                 }
4486         } else {
4487                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
4488                         uta_info->uta_shadow[i] = 0;
4489                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
4490                 }
4491         }
4492         return 0;
4493
4494 }
4495
4496 uint32_t
4497 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
4498 {
4499         uint32_t new_val = orig_val;
4500
4501         if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG)
4502                 new_val |= IXGBE_VMOLR_AUPE;
4503         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC)
4504                 new_val |= IXGBE_VMOLR_ROMPE;
4505         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
4506                 new_val |= IXGBE_VMOLR_ROPE;
4507         if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
4508                 new_val |= IXGBE_VMOLR_BAM;
4509         if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
4510                 new_val |= IXGBE_VMOLR_MPE;
4511
4512         return new_val;
4513 }
4514
4515 static int
4516 ixgbe_set_pool_rx_mode(struct rte_eth_dev *dev, uint16_t pool,
4517                                uint16_t rx_mask, uint8_t on)
4518 {
4519         int val = 0;
4520
4521         struct ixgbe_hw *hw =
4522                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4523         uint32_t vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
4524
4525         if (hw->mac.type == ixgbe_mac_82598EB) {
4526                 PMD_INIT_LOG(ERR, "setting VF receive mode set should be done"
4527                              " on 82599 hardware and newer");
4528                 return -ENOTSUP;
4529         }
4530         if (ixgbe_vmdq_mode_check(hw) < 0)
4531                 return -ENOTSUP;
4532
4533         val = ixgbe_convert_vm_rx_mask_to_val(rx_mask, val);
4534
4535         if (on)
4536                 vmolr |= val;
4537         else
4538                 vmolr &= ~val;
4539
4540         IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
4541
4542         return 0;
4543 }
4544
4545 static int
4546 ixgbe_set_pool_rx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on)
4547 {
4548         uint32_t reg, addr;
4549         uint32_t val;
4550         const uint8_t bit1 = 0x1;
4551
4552         struct ixgbe_hw *hw =
4553                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4554
4555         if (ixgbe_vmdq_mode_check(hw) < 0)
4556                 return -ENOTSUP;
4557
4558         if (pool >= ETH_64_POOLS)
4559                 return -EINVAL;
4560
4561         /* for pool >= 32, set bit in PFVFRE[1], otherwise PFVFRE[0] */
4562         if (pool >= 32) {
4563                 addr = IXGBE_VFRE(1);
4564                 val = bit1 << (pool - 32);
4565         } else {
4566                 addr = IXGBE_VFRE(0);
4567                 val = bit1 << pool;
4568         }
4569
4570         reg = IXGBE_READ_REG(hw, addr);
4571
4572         if (on)
4573                 reg |= val;
4574         else
4575                 reg &= ~val;
4576
4577         IXGBE_WRITE_REG(hw, addr, reg);
4578
4579         return 0;
4580 }
4581
4582 static int
4583 ixgbe_set_pool_tx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on)
4584 {
4585         uint32_t reg, addr;
4586         uint32_t val;
4587         const uint8_t bit1 = 0x1;
4588
4589         struct ixgbe_hw *hw =
4590                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4591
4592         if (ixgbe_vmdq_mode_check(hw) < 0)
4593                 return -ENOTSUP;
4594
4595         if (pool >= ETH_64_POOLS)
4596                 return -EINVAL;
4597
4598         /* for pool >= 32, set bit in PFVFTE[1], otherwise PFVFTE[0] */
4599         if (pool >= 32) {
4600                 addr = IXGBE_VFTE(1);
4601                 val = bit1 << (pool - 32);
4602         } else {
4603                 addr = IXGBE_VFTE(0);
4604                 val = bit1 << pool;
4605         }
4606
4607         reg = IXGBE_READ_REG(hw, addr);
4608
4609         if (on)
4610                 reg |= val;
4611         else
4612                 reg &= ~val;
4613
4614         IXGBE_WRITE_REG(hw, addr, reg);
4615
4616         return 0;
4617 }
4618
4619 static int
4620 ixgbe_set_pool_vlan_filter(struct rte_eth_dev *dev, uint16_t vlan,
4621                         uint64_t pool_mask, uint8_t vlan_on)
4622 {
4623         int ret = 0;
4624         uint16_t pool_idx;
4625         struct ixgbe_hw *hw =
4626                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4627
4628         if (ixgbe_vmdq_mode_check(hw) < 0)
4629                 return -ENOTSUP;
4630         for (pool_idx = 0; pool_idx < ETH_64_POOLS; pool_idx++) {
4631                 if (pool_mask & ((uint64_t)(1ULL << pool_idx))) {
4632                         ret = hw->mac.ops.set_vfta(hw, vlan, pool_idx,
4633                                                    vlan_on, false);
4634                         if (ret < 0)
4635                                 return ret;
4636                 }
4637         }
4638
4639         return ret;
4640 }
4641
4642 #define IXGBE_MRCTL_VPME  0x01 /* Virtual Pool Mirroring. */
4643 #define IXGBE_MRCTL_UPME  0x02 /* Uplink Port Mirroring. */
4644 #define IXGBE_MRCTL_DPME  0x04 /* Downlink Port Mirroring. */
4645 #define IXGBE_MRCTL_VLME  0x08 /* VLAN Mirroring. */
4646 #define IXGBE_INVALID_MIRROR_TYPE(mirror_type) \
4647         ((mirror_type) & ~(uint8_t)(ETH_MIRROR_VIRTUAL_POOL_UP | \
4648         ETH_MIRROR_UPLINK_PORT | ETH_MIRROR_DOWNLINK_PORT | ETH_MIRROR_VLAN))
4649
4650 static int
4651 ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
4652                         struct rte_eth_mirror_conf *mirror_conf,
4653                         uint8_t rule_id, uint8_t on)
4654 {
4655         uint32_t mr_ctl, vlvf;
4656         uint32_t mp_lsb = 0;
4657         uint32_t mv_msb = 0;
4658         uint32_t mv_lsb = 0;
4659         uint32_t mp_msb = 0;
4660         uint8_t i = 0;
4661         int reg_index = 0;
4662         uint64_t vlan_mask = 0;
4663
4664         const uint8_t pool_mask_offset = 32;
4665         const uint8_t vlan_mask_offset = 32;
4666         const uint8_t dst_pool_offset = 8;
4667         const uint8_t rule_mr_offset  = 4;
4668         const uint8_t mirror_rule_mask = 0x0F;
4669
4670         struct ixgbe_mirror_info *mr_info =
4671                         (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
4672         struct ixgbe_hw *hw =
4673                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4674         uint8_t mirror_type = 0;
4675
4676         if (ixgbe_vmdq_mode_check(hw) < 0)
4677                 return -ENOTSUP;
4678
4679         if (rule_id >= IXGBE_MAX_MIRROR_RULES)
4680                 return -EINVAL;
4681
4682         if (IXGBE_INVALID_MIRROR_TYPE(mirror_conf->rule_type)) {
4683                 PMD_DRV_LOG(ERR, "unsupported mirror type 0x%x.",
4684                         mirror_conf->rule_type);
4685                 return -EINVAL;
4686         }
4687
4688         if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
4689                 mirror_type |= IXGBE_MRCTL_VLME;
4690                 /* Check if vlan id is valid and find conresponding VLAN ID index in VLVF */
4691                 for (i = 0; i < IXGBE_VLVF_ENTRIES; i++) {
4692                         if (mirror_conf->vlan.vlan_mask & (1ULL << i)) {
4693                                 /* search vlan id related pool vlan filter index */
4694                                 reg_index = ixgbe_find_vlvf_slot(hw,
4695                                                  mirror_conf->vlan.vlan_id[i],
4696                                                  false);
4697                                 if (reg_index < 0)
4698                                         return -EINVAL;
4699                                 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(reg_index));
4700                                 if ((vlvf & IXGBE_VLVF_VIEN) &&
4701                                     ((vlvf & IXGBE_VLVF_VLANID_MASK) ==
4702                                       mirror_conf->vlan.vlan_id[i]))
4703                                         vlan_mask |= (1ULL << reg_index);
4704                                 else
4705                                         return -EINVAL;
4706                         }
4707                 }
4708
4709                 if (on) {
4710                         mv_lsb = vlan_mask & 0xFFFFFFFF;
4711                         mv_msb = vlan_mask >> vlan_mask_offset;
4712
4713                         mr_info->mr_conf[rule_id].vlan.vlan_mask =
4714                                                 mirror_conf->vlan.vlan_mask;
4715                         for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++) {
4716                                 if (mirror_conf->vlan.vlan_mask & (1ULL << i))
4717                                         mr_info->mr_conf[rule_id].vlan.vlan_id[i] =
4718                                                 mirror_conf->vlan.vlan_id[i];
4719                         }
4720                 } else {
4721                         mv_lsb = 0;
4722                         mv_msb = 0;
4723                         mr_info->mr_conf[rule_id].vlan.vlan_mask = 0;
4724                         for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++)
4725                                 mr_info->mr_conf[rule_id].vlan.vlan_id[i] = 0;
4726                 }
4727         }
4728
4729         /*
4730          * if enable pool mirror, write related pool mask register,if disable
4731          * pool mirror, clear PFMRVM register
4732          */
4733         if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
4734                 mirror_type |= IXGBE_MRCTL_VPME;
4735                 if (on) {
4736                         mp_lsb = mirror_conf->pool_mask & 0xFFFFFFFF;
4737                         mp_msb = mirror_conf->pool_mask >> pool_mask_offset;
4738                         mr_info->mr_conf[rule_id].pool_mask =
4739                                         mirror_conf->pool_mask;
4740
4741                 } else {
4742                         mp_lsb = 0;
4743                         mp_msb = 0;
4744                         mr_info->mr_conf[rule_id].pool_mask = 0;
4745                 }
4746         }
4747         if (mirror_conf->rule_type & ETH_MIRROR_UPLINK_PORT)
4748                 mirror_type |= IXGBE_MRCTL_UPME;
4749         if (mirror_conf->rule_type & ETH_MIRROR_DOWNLINK_PORT)
4750                 mirror_type |= IXGBE_MRCTL_DPME;
4751
4752         /* read  mirror control register and recalculate it */
4753         mr_ctl = IXGBE_READ_REG(hw, IXGBE_MRCTL(rule_id));
4754
4755         if (on) {
4756                 mr_ctl |= mirror_type;
4757                 mr_ctl &= mirror_rule_mask;
4758                 mr_ctl |= mirror_conf->dst_pool << dst_pool_offset;
4759         } else
4760                 mr_ctl &= ~(mirror_conf->rule_type & mirror_rule_mask);
4761
4762         mr_info->mr_conf[rule_id].rule_type = mirror_conf->rule_type;
4763         mr_info->mr_conf[rule_id].dst_pool = mirror_conf->dst_pool;
4764
4765         /* write mirrror control  register */
4766         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
4767
4768         /* write pool mirrror control  register */
4769         if (mirror_conf->rule_type == ETH_MIRROR_VIRTUAL_POOL_UP) {
4770                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), mp_lsb);
4771                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset),
4772                                 mp_msb);
4773         }
4774         /* write VLAN mirrror control  register */
4775         if (mirror_conf->rule_type == ETH_MIRROR_VLAN) {
4776                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), mv_lsb);
4777                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset),
4778                                 mv_msb);
4779         }
4780
4781         return 0;
4782 }
4783
4784 static int
4785 ixgbe_mirror_rule_reset(struct rte_eth_dev *dev, uint8_t rule_id)
4786 {
4787         int mr_ctl = 0;
4788         uint32_t lsb_val = 0;
4789         uint32_t msb_val = 0;
4790         const uint8_t rule_mr_offset = 4;
4791
4792         struct ixgbe_hw *hw =
4793                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4794         struct ixgbe_mirror_info *mr_info =
4795                 (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
4796
4797         if (ixgbe_vmdq_mode_check(hw) < 0)
4798                 return -ENOTSUP;
4799
4800         memset(&mr_info->mr_conf[rule_id], 0,
4801                 sizeof(struct rte_eth_mirror_conf));
4802
4803         /* clear PFVMCTL register */
4804         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
4805
4806         /* clear pool mask register */
4807         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), lsb_val);
4808         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset), msb_val);
4809
4810         /* clear vlan mask register */
4811         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), lsb_val);
4812         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset), msb_val);
4813
4814         return 0;
4815 }
4816
4817 static int
4818 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
4819 {
4820         uint32_t mask;
4821         struct ixgbe_hw *hw =
4822                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4823
4824         mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
4825         mask |= (1 << IXGBE_MISC_VEC_ID);
4826         RTE_SET_USED(queue_id);
4827         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
4828
4829         rte_intr_enable(&dev->pci_dev->intr_handle);
4830
4831         return 0;
4832 }
4833
4834 static int
4835 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
4836 {
4837         uint32_t mask;
4838         struct ixgbe_hw *hw =
4839                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4840
4841         mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
4842         mask &= ~(1 << IXGBE_MISC_VEC_ID);
4843         RTE_SET_USED(queue_id);
4844         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
4845
4846         return 0;
4847 }
4848
4849 static int
4850 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
4851 {
4852         uint32_t mask;
4853         struct ixgbe_hw *hw =
4854                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4855         struct ixgbe_interrupt *intr =
4856                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4857
4858         if (queue_id < 16) {
4859                 ixgbe_disable_intr(hw);
4860                 intr->mask |= (1 << queue_id);
4861                 ixgbe_enable_intr(dev);
4862         } else if (queue_id < 32) {
4863                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
4864                 mask &= (1 << queue_id);
4865                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
4866         } else if (queue_id < 64) {
4867                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
4868                 mask &= (1 << (queue_id - 32));
4869                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
4870         }
4871         rte_intr_enable(&dev->pci_dev->intr_handle);
4872
4873         return 0;
4874 }
4875
4876 static int
4877 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
4878 {
4879         uint32_t mask;
4880         struct ixgbe_hw *hw =
4881                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4882         struct ixgbe_interrupt *intr =
4883                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4884
4885         if (queue_id < 16) {
4886                 ixgbe_disable_intr(hw);
4887                 intr->mask &= ~(1 << queue_id);
4888                 ixgbe_enable_intr(dev);
4889         } else if (queue_id < 32) {
4890                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
4891                 mask &= ~(1 << queue_id);
4892                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
4893         } else if (queue_id < 64) {
4894                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
4895                 mask &= ~(1 << (queue_id - 32));
4896                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
4897         }
4898
4899         return 0;
4900 }
4901
4902 static void
4903 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
4904                      uint8_t queue, uint8_t msix_vector)
4905 {
4906         uint32_t tmp, idx;
4907
4908         if (direction == -1) {
4909                 /* other causes */
4910                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
4911                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
4912                 tmp &= ~0xFF;
4913                 tmp |= msix_vector;
4914                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
4915         } else {
4916                 /* rx or tx cause */
4917                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
4918                 idx = ((16 * (queue & 1)) + (8 * direction));
4919                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
4920                 tmp &= ~(0xFF << idx);
4921                 tmp |= (msix_vector << idx);
4922                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
4923         }
4924 }
4925
4926 /**
4927  * set the IVAR registers, mapping interrupt causes to vectors
4928  * @param hw
4929  *  pointer to ixgbe_hw struct
4930  * @direction
4931  *  0 for Rx, 1 for Tx, -1 for other causes
4932  * @queue
4933  *  queue to map the corresponding interrupt to
4934  * @msix_vector
4935  *  the vector to map to the corresponding queue
4936  */
4937 static void
4938 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
4939                    uint8_t queue, uint8_t msix_vector)
4940 {
4941         uint32_t tmp, idx;
4942
4943         msix_vector |= IXGBE_IVAR_ALLOC_VAL;
4944         if (hw->mac.type == ixgbe_mac_82598EB) {
4945                 if (direction == -1)
4946                         direction = 0;
4947                 idx = (((direction * 64) + queue) >> 2) & 0x1F;
4948                 tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
4949                 tmp &= ~(0xFF << (8 * (queue & 0x3)));
4950                 tmp |= (msix_vector << (8 * (queue & 0x3)));
4951                 IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
4952         } else if ((hw->mac.type == ixgbe_mac_82599EB) ||
4953                         (hw->mac.type == ixgbe_mac_X540)) {
4954                 if (direction == -1) {
4955                         /* other causes */
4956                         idx = ((queue & 1) * 8);
4957                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
4958                         tmp &= ~(0xFF << idx);
4959                         tmp |= (msix_vector << idx);
4960                         IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
4961                 } else {
4962                         /* rx or tx causes */
4963                         idx = ((16 * (queue & 1)) + (8 * direction));
4964                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
4965                         tmp &= ~(0xFF << idx);
4966                         tmp |= (msix_vector << idx);
4967                         IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
4968                 }
4969         }
4970 }
4971
4972 static void
4973 ixgbevf_configure_msix(struct rte_eth_dev *dev)
4974 {
4975         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
4976         struct ixgbe_hw *hw =
4977                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4978         uint32_t q_idx;
4979         uint32_t vector_idx = IXGBE_MISC_VEC_ID;
4980
4981         /* Configure VF other cause ivar */
4982         ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
4983
4984         /* won't configure msix register if no mapping is done
4985          * between intr vector and event fd.
4986          */
4987         if (!rte_intr_dp_is_en(intr_handle))
4988                 return;
4989
4990         /* Configure all RX queues of VF */
4991         for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
4992                 /* Force all queue use vector 0,
4993                  * as IXGBE_VF_MAXMSIVECOTR = 1
4994                  */
4995                 ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
4996                 intr_handle->intr_vec[q_idx] = vector_idx;
4997         }
4998 }
4999
5000 /**
5001  * Sets up the hardware to properly generate MSI-X interrupts
5002  * @hw
5003  *  board private structure
5004  */
5005 static void
5006 ixgbe_configure_msix(struct rte_eth_dev *dev)
5007 {
5008         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
5009         struct ixgbe_hw *hw =
5010                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5011         uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
5012         uint32_t vec = IXGBE_MISC_VEC_ID;
5013         uint32_t mask;
5014         uint32_t gpie;
5015
5016         /* won't configure msix register if no mapping is done
5017          * between intr vector and event fd
5018          */
5019         if (!rte_intr_dp_is_en(intr_handle))
5020                 return;
5021
5022         if (rte_intr_allow_others(intr_handle))
5023                 vec = base = IXGBE_RX_VEC_START;
5024
5025         /* setup GPIE for MSI-x mode */
5026         gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
5027         gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5028                 IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
5029         /* auto clearing and auto setting corresponding bits in EIMS
5030          * when MSI-X interrupt is triggered
5031          */
5032         if (hw->mac.type == ixgbe_mac_82598EB) {
5033                 IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5034         } else {
5035                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5036                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5037         }
5038         IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5039
5040         /* Populate the IVAR table and set the ITR values to the
5041          * corresponding register.
5042          */
5043         for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
5044              queue_id++) {
5045                 /* by default, 1:1 mapping */
5046                 ixgbe_set_ivar_map(hw, 0, queue_id, vec);
5047                 intr_handle->intr_vec[queue_id] = vec;
5048                 if (vec < base + intr_handle->nb_efd - 1)
5049                         vec++;
5050         }
5051
5052         switch (hw->mac.type) {
5053         case ixgbe_mac_82598EB:
5054                 ixgbe_set_ivar_map(hw, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
5055                                    IXGBE_MISC_VEC_ID);
5056                 break;
5057         case ixgbe_mac_82599EB:
5058         case ixgbe_mac_X540:
5059                 ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
5060                 break;
5061         default:
5062                 break;
5063         }
5064         IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
5065                         IXGBE_MIN_INTER_INTERRUPT_INTERVAL_DEFAULT & 0xFFF);
5066
5067         /* set up to autoclear timer, and the vectors */
5068         mask = IXGBE_EIMS_ENABLE_MASK;
5069         mask &= ~(IXGBE_EIMS_OTHER |
5070                   IXGBE_EIMS_MAILBOX |
5071                   IXGBE_EIMS_LSC);
5072
5073         IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
5074 }
5075
5076 static int ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
5077         uint16_t queue_idx, uint16_t tx_rate)
5078 {
5079         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5080         uint32_t rf_dec, rf_int;
5081         uint32_t bcnrc_val;
5082         uint16_t link_speed = dev->data->dev_link.link_speed;
5083
5084         if (queue_idx >= hw->mac.max_tx_queues)
5085                 return -EINVAL;
5086
5087         if (tx_rate != 0) {
5088                 /* Calculate the rate factor values to set */
5089                 rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
5090                 rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
5091                 rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
5092
5093                 bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
5094                 bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
5095                                 IXGBE_RTTBCNRC_RF_INT_MASK_M);
5096                 bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
5097         } else {
5098                 bcnrc_val = 0;
5099         }
5100
5101         /*
5102          * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
5103          * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
5104          * set as 0x4.
5105          */
5106         if ((dev->data->dev_conf.rxmode.jumbo_frame == 1) &&
5107                 (dev->data->dev_conf.rxmode.max_rx_pkt_len >=
5108                                 IXGBE_MAX_JUMBO_FRAME_SIZE))
5109                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
5110                         IXGBE_MMW_SIZE_JUMBO_FRAME);
5111         else
5112                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
5113                         IXGBE_MMW_SIZE_DEFAULT);
5114
5115         /* Set RTTBCNRC of queue X */
5116         IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
5117         IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
5118         IXGBE_WRITE_FLUSH(hw);
5119
5120         return 0;
5121 }
5122
5123 static int ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
5124         uint16_t tx_rate, uint64_t q_msk)
5125 {
5126         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5127         struct ixgbe_vf_info *vfinfo =
5128                 *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
5129         uint8_t  nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
5130         uint32_t queue_stride =
5131                 IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
5132         uint32_t queue_idx = vf * queue_stride, idx = 0, vf_idx;
5133         uint32_t queue_end = queue_idx + nb_q_per_pool - 1;
5134         uint16_t total_rate = 0;
5135
5136         if (queue_end >= hw->mac.max_tx_queues)
5137                 return -EINVAL;
5138
5139         if (vfinfo != NULL) {
5140                 for (vf_idx = 0; vf_idx < dev->pci_dev->max_vfs; vf_idx++) {
5141                         if (vf_idx == vf)
5142                                 continue;
5143                         for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
5144                                 idx++)
5145                                 total_rate += vfinfo[vf_idx].tx_rate[idx];
5146                 }
5147         } else
5148                 return -EINVAL;
5149
5150         /* Store tx_rate for this vf. */
5151         for (idx = 0; idx < nb_q_per_pool; idx++) {
5152                 if (((uint64_t)0x1 << idx) & q_msk) {
5153                         if (vfinfo[vf].tx_rate[idx] != tx_rate)
5154                                 vfinfo[vf].tx_rate[idx] = tx_rate;
5155                         total_rate += tx_rate;
5156                 }
5157         }
5158
5159         if (total_rate > dev->data->dev_link.link_speed) {
5160                 /*
5161                  * Reset stored TX rate of the VF if it causes exceed
5162                  * link speed.
5163                  */
5164                 memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
5165                 return -EINVAL;
5166         }
5167
5168         /* Set RTTBCNRC of each queue/pool for vf X  */
5169         for (; queue_idx <= queue_end; queue_idx++) {
5170                 if (0x1 & q_msk)
5171                         ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
5172                 q_msk = q_msk >> 1;
5173         }
5174
5175         return 0;
5176 }
5177
5178 static void
5179 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
5180                      __attribute__((unused)) uint32_t index,
5181                      __attribute__((unused)) uint32_t pool)
5182 {
5183         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5184         int diag;
5185
5186         /*
5187          * On a 82599 VF, adding again the same MAC addr is not an idempotent
5188          * operation. Trap this case to avoid exhausting the [very limited]
5189          * set of PF resources used to store VF MAC addresses.
5190          */
5191         if (memcmp(hw->mac.perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
5192                 return;
5193         diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
5194         if (diag == 0)
5195                 return;
5196         PMD_DRV_LOG(ERR, "Unable to add MAC address - diag=%d", diag);
5197 }
5198
5199 static void
5200 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
5201 {
5202         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5203         struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
5204         struct ether_addr *mac_addr;
5205         uint32_t i;
5206         int diag;
5207
5208         /*
5209          * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
5210          * not support the deletion of a given MAC address.
5211          * Instead, it imposes to delete all MAC addresses, then to add again
5212          * all MAC addresses with the exception of the one to be deleted.
5213          */
5214         (void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
5215
5216         /*
5217          * Add again all MAC addresses, with the exception of the deleted one
5218          * and of the permanent MAC address.
5219          */
5220         for (i = 0, mac_addr = dev->data->mac_addrs;
5221              i < hw->mac.num_rar_entries; i++, mac_addr++) {
5222                 /* Skip the deleted MAC address */
5223                 if (i == index)
5224                         continue;
5225                 /* Skip NULL MAC addresses */
5226                 if (is_zero_ether_addr(mac_addr))
5227                         continue;
5228                 /* Skip the permanent MAC address */
5229                 if (memcmp(perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
5230                         continue;
5231                 diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
5232                 if (diag != 0)
5233                         PMD_DRV_LOG(ERR,
5234                                     "Adding again MAC address "
5235                                     "%02x:%02x:%02x:%02x:%02x:%02x failed "
5236                                     "diag=%d",
5237                                     mac_addr->addr_bytes[0],
5238                                     mac_addr->addr_bytes[1],
5239                                     mac_addr->addr_bytes[2],
5240                                     mac_addr->addr_bytes[3],
5241                                     mac_addr->addr_bytes[4],
5242                                     mac_addr->addr_bytes[5],
5243                                     diag);
5244         }
5245 }
5246
5247 static void
5248 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
5249 {
5250         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5251
5252         hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
5253 }
5254
5255 #define MAC_TYPE_FILTER_SUP(type)    do {\
5256         if ((type) != ixgbe_mac_82599EB && (type) != ixgbe_mac_X540 &&\
5257                 (type) != ixgbe_mac_X550 && (type) != ixgbe_mac_X550EM_x &&\
5258                 (type) != ixgbe_mac_X550EM_a)\
5259                 return -ENOTSUP;\
5260 } while (0)
5261
5262 static int
5263 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
5264                         struct rte_eth_syn_filter *filter,
5265                         bool add)
5266 {
5267         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5268         uint32_t synqf;
5269
5270         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
5271                 return -EINVAL;
5272
5273         synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
5274
5275         if (add) {
5276                 if (synqf & IXGBE_SYN_FILTER_ENABLE)
5277                         return -EINVAL;
5278                 synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
5279                         IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
5280
5281                 if (filter->hig_pri)
5282                         synqf |= IXGBE_SYN_FILTER_SYNQFP;
5283                 else
5284                         synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
5285         } else {
5286                 if (!(synqf & IXGBE_SYN_FILTER_ENABLE))
5287                         return -ENOENT;
5288                 synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
5289         }
5290         IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
5291         IXGBE_WRITE_FLUSH(hw);
5292         return 0;
5293 }
5294
5295 static int
5296 ixgbe_syn_filter_get(struct rte_eth_dev *dev,
5297                         struct rte_eth_syn_filter *filter)
5298 {
5299         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5300         uint32_t synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
5301
5302         if (synqf & IXGBE_SYN_FILTER_ENABLE) {
5303                 filter->hig_pri = (synqf & IXGBE_SYN_FILTER_SYNQFP) ? 1 : 0;
5304                 filter->queue = (uint16_t)((synqf & IXGBE_SYN_FILTER_QUEUE) >> 1);
5305                 return 0;
5306         }
5307         return -ENOENT;
5308 }
5309
5310 static int
5311 ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
5312                         enum rte_filter_op filter_op,
5313                         void *arg)
5314 {
5315         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5316         int ret;
5317
5318         MAC_TYPE_FILTER_SUP(hw->mac.type);
5319
5320         if (filter_op == RTE_ETH_FILTER_NOP)
5321                 return 0;
5322
5323         if (arg == NULL) {
5324                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
5325                             filter_op);
5326                 return -EINVAL;
5327         }
5328
5329         switch (filter_op) {
5330         case RTE_ETH_FILTER_ADD:
5331                 ret = ixgbe_syn_filter_set(dev,
5332                                 (struct rte_eth_syn_filter *)arg,
5333                                 TRUE);
5334                 break;
5335         case RTE_ETH_FILTER_DELETE:
5336                 ret = ixgbe_syn_filter_set(dev,
5337                                 (struct rte_eth_syn_filter *)arg,
5338                                 FALSE);
5339                 break;
5340         case RTE_ETH_FILTER_GET:
5341                 ret = ixgbe_syn_filter_get(dev,
5342                                 (struct rte_eth_syn_filter *)arg);
5343                 break;
5344         default:
5345                 PMD_DRV_LOG(ERR, "unsupported operation %u\n", filter_op);
5346                 ret = -EINVAL;
5347                 break;
5348         }
5349
5350         return ret;
5351 }
5352
5353
5354 static inline enum ixgbe_5tuple_protocol
5355 convert_protocol_type(uint8_t protocol_value)
5356 {
5357         if (protocol_value == IPPROTO_TCP)
5358                 return IXGBE_FILTER_PROTOCOL_TCP;
5359         else if (protocol_value == IPPROTO_UDP)
5360                 return IXGBE_FILTER_PROTOCOL_UDP;
5361         else if (protocol_value == IPPROTO_SCTP)
5362                 return IXGBE_FILTER_PROTOCOL_SCTP;
5363         else
5364                 return IXGBE_FILTER_PROTOCOL_NONE;
5365 }
5366
5367 /*
5368  * add a 5tuple filter
5369  *
5370  * @param
5371  * dev: Pointer to struct rte_eth_dev.
5372  * index: the index the filter allocates.
5373  * filter: ponter to the filter that will be added.
5374  * rx_queue: the queue id the filter assigned to.
5375  *
5376  * @return
5377  *    - On success, zero.
5378  *    - On failure, a negative value.
5379  */
5380 static int
5381 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
5382                         struct ixgbe_5tuple_filter *filter)
5383 {
5384         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5385         struct ixgbe_filter_info *filter_info =
5386                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5387         int i, idx, shift;
5388         uint32_t ftqf, sdpqf;
5389         uint32_t l34timir = 0;
5390         uint8_t mask = 0xff;
5391
5392         /*
5393          * look for an unused 5tuple filter index,
5394          * and insert the filter to list.
5395          */
5396         for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
5397                 idx = i / (sizeof(uint32_t) * NBBY);
5398                 shift = i % (sizeof(uint32_t) * NBBY);
5399                 if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
5400                         filter_info->fivetuple_mask[idx] |= 1 << shift;
5401                         filter->index = i;
5402                         TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
5403                                           filter,
5404                                           entries);
5405                         break;
5406                 }
5407         }
5408         if (i >= IXGBE_MAX_FTQF_FILTERS) {
5409                 PMD_DRV_LOG(ERR, "5tuple filters are full.");
5410                 return -ENOSYS;
5411         }
5412
5413         sdpqf = (uint32_t)(filter->filter_info.dst_port <<
5414                                 IXGBE_SDPQF_DSTPORT_SHIFT);
5415         sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
5416
5417         ftqf = (uint32_t)(filter->filter_info.proto &
5418                 IXGBE_FTQF_PROTOCOL_MASK);
5419         ftqf |= (uint32_t)((filter->filter_info.priority &
5420                 IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
5421         if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
5422                 mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
5423         if (filter->filter_info.dst_ip_mask == 0)
5424                 mask &= IXGBE_FTQF_DEST_ADDR_MASK;
5425         if (filter->filter_info.src_port_mask == 0)
5426                 mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
5427         if (filter->filter_info.dst_port_mask == 0)
5428                 mask &= IXGBE_FTQF_DEST_PORT_MASK;
5429         if (filter->filter_info.proto_mask == 0)
5430                 mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
5431         ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
5432         ftqf |= IXGBE_FTQF_POOL_MASK_EN;
5433         ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
5434
5435         IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
5436         IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
5437         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
5438         IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
5439
5440         l34timir |= IXGBE_L34T_IMIR_RESERVE;
5441         l34timir |= (uint32_t)(filter->queue <<
5442                                 IXGBE_L34T_IMIR_QUEUE_SHIFT);
5443         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
5444         return 0;
5445 }
5446
5447 /*
5448  * remove a 5tuple filter
5449  *
5450  * @param
5451  * dev: Pointer to struct rte_eth_dev.
5452  * filter: the pointer of the filter will be removed.
5453  */
5454 static void
5455 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
5456                         struct ixgbe_5tuple_filter *filter)
5457 {
5458         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5459         struct ixgbe_filter_info *filter_info =
5460                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5461         uint16_t index = filter->index;
5462
5463         filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
5464                                 ~(1 << (index % (sizeof(uint32_t) * NBBY)));
5465         TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
5466         rte_free(filter);
5467
5468         IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
5469         IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
5470         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
5471         IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
5472         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
5473 }
5474
5475 static int
5476 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
5477 {
5478         struct ixgbe_hw *hw;
5479         uint32_t max_frame = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
5480
5481         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5482
5483         if ((mtu < ETHER_MIN_MTU) || (max_frame > ETHER_MAX_JUMBO_FRAME_LEN))
5484                 return -EINVAL;
5485
5486         /* refuse mtu that requires the support of scattered packets when this
5487          * feature has not been enabled before.
5488          */
5489         if (!dev->data->scattered_rx &&
5490             (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
5491              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM))
5492                 return -EINVAL;
5493
5494         /*
5495          * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
5496          * request of the version 2.0 of the mailbox API.
5497          * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
5498          * of the mailbox API.
5499          * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
5500          * prior to 3.11.33 which contains the following change:
5501          * "ixgbe: Enable jumbo frames support w/ SR-IOV"
5502          */
5503         ixgbevf_rlpml_set_vf(hw, max_frame);
5504
5505         /* update max frame size */
5506         dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame;
5507         return 0;
5508 }
5509
5510 #define MAC_TYPE_FILTER_SUP_EXT(type)    do {\
5511         if ((type) != ixgbe_mac_82599EB && (type) != ixgbe_mac_X540)\
5512                 return -ENOTSUP;\
5513 } while (0)
5514
5515 static inline struct ixgbe_5tuple_filter *
5516 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
5517                         struct ixgbe_5tuple_filter_info *key)
5518 {
5519         struct ixgbe_5tuple_filter *it;
5520
5521         TAILQ_FOREACH(it, filter_list, entries) {
5522                 if (memcmp(key, &it->filter_info,
5523                         sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
5524                         return it;
5525                 }
5526         }
5527         return NULL;
5528 }
5529
5530 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
5531 static inline int
5532 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
5533                         struct ixgbe_5tuple_filter_info *filter_info)
5534 {
5535         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
5536                 filter->priority > IXGBE_5TUPLE_MAX_PRI ||
5537                 filter->priority < IXGBE_5TUPLE_MIN_PRI)
5538                 return -EINVAL;
5539
5540         switch (filter->dst_ip_mask) {
5541         case UINT32_MAX:
5542                 filter_info->dst_ip_mask = 0;
5543                 filter_info->dst_ip = filter->dst_ip;
5544                 break;
5545         case 0:
5546                 filter_info->dst_ip_mask = 1;
5547                 break;
5548         default:
5549                 PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
5550                 return -EINVAL;
5551         }
5552
5553         switch (filter->src_ip_mask) {
5554         case UINT32_MAX:
5555                 filter_info->src_ip_mask = 0;
5556                 filter_info->src_ip = filter->src_ip;
5557                 break;
5558         case 0:
5559                 filter_info->src_ip_mask = 1;
5560                 break;
5561         default:
5562                 PMD_DRV_LOG(ERR, "invalid src_ip mask.");
5563                 return -EINVAL;
5564         }
5565
5566         switch (filter->dst_port_mask) {
5567         case UINT16_MAX:
5568                 filter_info->dst_port_mask = 0;
5569                 filter_info->dst_port = filter->dst_port;
5570                 break;
5571         case 0:
5572                 filter_info->dst_port_mask = 1;
5573                 break;
5574         default:
5575                 PMD_DRV_LOG(ERR, "invalid dst_port mask.");
5576                 return -EINVAL;
5577         }
5578
5579         switch (filter->src_port_mask) {
5580         case UINT16_MAX:
5581                 filter_info->src_port_mask = 0;
5582                 filter_info->src_port = filter->src_port;
5583                 break;
5584         case 0:
5585                 filter_info->src_port_mask = 1;
5586                 break;
5587         default:
5588                 PMD_DRV_LOG(ERR, "invalid src_port mask.");
5589                 return -EINVAL;
5590         }
5591
5592         switch (filter->proto_mask) {
5593         case UINT8_MAX:
5594                 filter_info->proto_mask = 0;
5595                 filter_info->proto =
5596                         convert_protocol_type(filter->proto);
5597                 break;
5598         case 0:
5599                 filter_info->proto_mask = 1;
5600                 break;
5601         default:
5602                 PMD_DRV_LOG(ERR, "invalid protocol mask.");
5603                 return -EINVAL;
5604         }
5605
5606         filter_info->priority = (uint8_t)filter->priority;
5607         return 0;
5608 }
5609
5610 /*
5611  * add or delete a ntuple filter
5612  *
5613  * @param
5614  * dev: Pointer to struct rte_eth_dev.
5615  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
5616  * add: if true, add filter, if false, remove filter
5617  *
5618  * @return
5619  *    - On success, zero.
5620  *    - On failure, a negative value.
5621  */
5622 static int
5623 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
5624                         struct rte_eth_ntuple_filter *ntuple_filter,
5625                         bool add)
5626 {
5627         struct ixgbe_filter_info *filter_info =
5628                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5629         struct ixgbe_5tuple_filter_info filter_5tuple;
5630         struct ixgbe_5tuple_filter *filter;
5631         int ret;
5632
5633         if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
5634                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
5635                 return -EINVAL;
5636         }
5637
5638         memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
5639         ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
5640         if (ret < 0)
5641                 return ret;
5642
5643         filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
5644                                          &filter_5tuple);
5645         if (filter != NULL && add) {
5646                 PMD_DRV_LOG(ERR, "filter exists.");
5647                 return -EEXIST;
5648         }
5649         if (filter == NULL && !add) {
5650                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
5651                 return -ENOENT;
5652         }
5653
5654         if (add) {
5655                 filter = rte_zmalloc("ixgbe_5tuple_filter",
5656                                 sizeof(struct ixgbe_5tuple_filter), 0);
5657                 if (filter == NULL)
5658                         return -ENOMEM;
5659                 (void)rte_memcpy(&filter->filter_info,
5660                                  &filter_5tuple,
5661                                  sizeof(struct ixgbe_5tuple_filter_info));
5662                 filter->queue = ntuple_filter->queue;
5663                 ret = ixgbe_add_5tuple_filter(dev, filter);
5664                 if (ret < 0) {
5665                         rte_free(filter);
5666                         return ret;
5667                 }
5668         } else
5669                 ixgbe_remove_5tuple_filter(dev, filter);
5670
5671         return 0;
5672 }
5673
5674 /*
5675  * get a ntuple filter
5676  *
5677  * @param
5678  * dev: Pointer to struct rte_eth_dev.
5679  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
5680  *
5681  * @return
5682  *    - On success, zero.
5683  *    - On failure, a negative value.
5684  */
5685 static int
5686 ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
5687                         struct rte_eth_ntuple_filter *ntuple_filter)
5688 {
5689         struct ixgbe_filter_info *filter_info =
5690                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5691         struct ixgbe_5tuple_filter_info filter_5tuple;
5692         struct ixgbe_5tuple_filter *filter;
5693         int ret;
5694
5695         if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
5696                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
5697                 return -EINVAL;
5698         }
5699
5700         memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
5701         ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
5702         if (ret < 0)
5703                 return ret;
5704
5705         filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
5706                                          &filter_5tuple);
5707         if (filter == NULL) {
5708                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
5709                 return -ENOENT;
5710         }
5711         ntuple_filter->queue = filter->queue;
5712         return 0;
5713 }
5714
5715 /*
5716  * ixgbe_ntuple_filter_handle - Handle operations for ntuple filter.
5717  * @dev: pointer to rte_eth_dev structure
5718  * @filter_op:operation will be taken.
5719  * @arg: a pointer to specific structure corresponding to the filter_op
5720  *
5721  * @return
5722  *    - On success, zero.
5723  *    - On failure, a negative value.
5724  */
5725 static int
5726 ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
5727                                 enum rte_filter_op filter_op,
5728                                 void *arg)
5729 {
5730         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5731         int ret;
5732
5733         MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
5734
5735         if (filter_op == RTE_ETH_FILTER_NOP)
5736                 return 0;
5737
5738         if (arg == NULL) {
5739                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
5740                             filter_op);
5741                 return -EINVAL;
5742         }
5743
5744         switch (filter_op) {
5745         case RTE_ETH_FILTER_ADD:
5746                 ret = ixgbe_add_del_ntuple_filter(dev,
5747                         (struct rte_eth_ntuple_filter *)arg,
5748                         TRUE);
5749                 break;
5750         case RTE_ETH_FILTER_DELETE:
5751                 ret = ixgbe_add_del_ntuple_filter(dev,
5752                         (struct rte_eth_ntuple_filter *)arg,
5753                         FALSE);
5754                 break;
5755         case RTE_ETH_FILTER_GET:
5756                 ret = ixgbe_get_ntuple_filter(dev,
5757                         (struct rte_eth_ntuple_filter *)arg);
5758                 break;
5759         default:
5760                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
5761                 ret = -EINVAL;
5762                 break;
5763         }
5764         return ret;
5765 }
5766
5767 static inline int
5768 ixgbe_ethertype_filter_lookup(struct ixgbe_filter_info *filter_info,
5769                         uint16_t ethertype)
5770 {
5771         int i;
5772
5773         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
5774                 if (filter_info->ethertype_filters[i] == ethertype &&
5775                     (filter_info->ethertype_mask & (1 << i)))
5776                         return i;
5777         }
5778         return -1;
5779 }
5780
5781 static inline int
5782 ixgbe_ethertype_filter_insert(struct ixgbe_filter_info *filter_info,
5783                         uint16_t ethertype)
5784 {
5785         int i;
5786
5787         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
5788                 if (!(filter_info->ethertype_mask & (1 << i))) {
5789                         filter_info->ethertype_mask |= 1 << i;
5790                         filter_info->ethertype_filters[i] = ethertype;
5791                         return i;
5792                 }
5793         }
5794         return -1;
5795 }
5796
5797 static inline int
5798 ixgbe_ethertype_filter_remove(struct ixgbe_filter_info *filter_info,
5799                         uint8_t idx)
5800 {
5801         if (idx >= IXGBE_MAX_ETQF_FILTERS)
5802                 return -1;
5803         filter_info->ethertype_mask &= ~(1 << idx);
5804         filter_info->ethertype_filters[idx] = 0;
5805         return idx;
5806 }
5807
5808 static int
5809 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
5810                         struct rte_eth_ethertype_filter *filter,
5811                         bool add)
5812 {
5813         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5814         struct ixgbe_filter_info *filter_info =
5815                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5816         uint32_t etqf = 0;
5817         uint32_t etqs = 0;
5818         int ret;
5819
5820         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
5821                 return -EINVAL;
5822
5823         if (filter->ether_type == ETHER_TYPE_IPv4 ||
5824                 filter->ether_type == ETHER_TYPE_IPv6) {
5825                 PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
5826                         " ethertype filter.", filter->ether_type);
5827                 return -EINVAL;
5828         }
5829
5830         if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
5831                 PMD_DRV_LOG(ERR, "mac compare is unsupported.");
5832                 return -EINVAL;
5833         }
5834         if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
5835                 PMD_DRV_LOG(ERR, "drop option is unsupported.");
5836                 return -EINVAL;
5837         }
5838
5839         ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
5840         if (ret >= 0 && add) {
5841                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
5842                             filter->ether_type);
5843                 return -EEXIST;
5844         }
5845         if (ret < 0 && !add) {
5846                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
5847                             filter->ether_type);
5848                 return -ENOENT;
5849         }
5850
5851         if (add) {
5852                 ret = ixgbe_ethertype_filter_insert(filter_info,
5853                         filter->ether_type);
5854                 if (ret < 0) {
5855                         PMD_DRV_LOG(ERR, "ethertype filters are full.");
5856                         return -ENOSYS;
5857                 }
5858                 etqf = IXGBE_ETQF_FILTER_EN;
5859                 etqf |= (uint32_t)filter->ether_type;
5860                 etqs |= (uint32_t)((filter->queue <<
5861                                     IXGBE_ETQS_RX_QUEUE_SHIFT) &
5862                                     IXGBE_ETQS_RX_QUEUE);
5863                 etqs |= IXGBE_ETQS_QUEUE_EN;
5864         } else {
5865                 ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
5866                 if (ret < 0)
5867                         return -ENOSYS;
5868         }
5869         IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
5870         IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
5871         IXGBE_WRITE_FLUSH(hw);
5872
5873         return 0;
5874 }
5875
5876 static int
5877 ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
5878                         struct rte_eth_ethertype_filter *filter)
5879 {
5880         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5881         struct ixgbe_filter_info *filter_info =
5882                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5883         uint32_t etqf, etqs;
5884         int ret;
5885
5886         ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
5887         if (ret < 0) {
5888                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
5889                             filter->ether_type);
5890                 return -ENOENT;
5891         }
5892
5893         etqf = IXGBE_READ_REG(hw, IXGBE_ETQF(ret));
5894         if (etqf & IXGBE_ETQF_FILTER_EN) {
5895                 etqs = IXGBE_READ_REG(hw, IXGBE_ETQS(ret));
5896                 filter->ether_type = etqf & IXGBE_ETQF_ETHERTYPE;
5897                 filter->flags = 0;
5898                 filter->queue = (etqs & IXGBE_ETQS_RX_QUEUE) >>
5899                                IXGBE_ETQS_RX_QUEUE_SHIFT;
5900                 return 0;
5901         }
5902         return -ENOENT;
5903 }
5904
5905 /*
5906  * ixgbe_ethertype_filter_handle - Handle operations for ethertype filter.
5907  * @dev: pointer to rte_eth_dev structure
5908  * @filter_op:operation will be taken.
5909  * @arg: a pointer to specific structure corresponding to the filter_op
5910  */
5911 static int
5912 ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
5913                                 enum rte_filter_op filter_op,
5914                                 void *arg)
5915 {
5916         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5917         int ret;
5918
5919         MAC_TYPE_FILTER_SUP(hw->mac.type);
5920
5921         if (filter_op == RTE_ETH_FILTER_NOP)
5922                 return 0;
5923
5924         if (arg == NULL) {
5925                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
5926                             filter_op);
5927                 return -EINVAL;
5928         }
5929
5930         switch (filter_op) {
5931         case RTE_ETH_FILTER_ADD:
5932                 ret = ixgbe_add_del_ethertype_filter(dev,
5933                         (struct rte_eth_ethertype_filter *)arg,
5934                         TRUE);
5935                 break;
5936         case RTE_ETH_FILTER_DELETE:
5937                 ret = ixgbe_add_del_ethertype_filter(dev,
5938                         (struct rte_eth_ethertype_filter *)arg,
5939                         FALSE);
5940                 break;
5941         case RTE_ETH_FILTER_GET:
5942                 ret = ixgbe_get_ethertype_filter(dev,
5943                         (struct rte_eth_ethertype_filter *)arg);
5944                 break;
5945         default:
5946                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
5947                 ret = -EINVAL;
5948                 break;
5949         }
5950         return ret;
5951 }
5952
5953 static int
5954 ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
5955                      enum rte_filter_type filter_type,
5956                      enum rte_filter_op filter_op,
5957                      void *arg)
5958 {
5959         int ret = -EINVAL;
5960
5961         switch (filter_type) {
5962         case RTE_ETH_FILTER_NTUPLE:
5963                 ret = ixgbe_ntuple_filter_handle(dev, filter_op, arg);
5964                 break;
5965         case RTE_ETH_FILTER_ETHERTYPE:
5966                 ret = ixgbe_ethertype_filter_handle(dev, filter_op, arg);
5967                 break;
5968         case RTE_ETH_FILTER_SYN:
5969                 ret = ixgbe_syn_filter_handle(dev, filter_op, arg);
5970                 break;
5971         case RTE_ETH_FILTER_FDIR:
5972                 ret = ixgbe_fdir_ctrl_func(dev, filter_op, arg);
5973                 break;
5974         case RTE_ETH_FILTER_L2_TUNNEL:
5975                 ret = ixgbe_dev_l2_tunnel_filter_handle(dev, filter_op, arg);
5976                 break;
5977         default:
5978                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
5979                                                         filter_type);
5980                 break;
5981         }
5982
5983         return ret;
5984 }
5985
5986 static u8 *
5987 ixgbe_dev_addr_list_itr(__attribute__((unused)) struct ixgbe_hw *hw,
5988                         u8 **mc_addr_ptr, u32 *vmdq)
5989 {
5990         u8 *mc_addr;
5991
5992         *vmdq = 0;
5993         mc_addr = *mc_addr_ptr;
5994         *mc_addr_ptr = (mc_addr + sizeof(struct ether_addr));
5995         return mc_addr;
5996 }
5997
5998 static int
5999 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
6000                           struct ether_addr *mc_addr_set,
6001                           uint32_t nb_mc_addr)
6002 {
6003         struct ixgbe_hw *hw;
6004         u8 *mc_addr_list;
6005
6006         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6007         mc_addr_list = (u8 *)mc_addr_set;
6008         return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
6009                                          ixgbe_dev_addr_list_itr, TRUE);
6010 }
6011
6012 static uint64_t
6013 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
6014 {
6015         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6016         uint64_t systime_cycles;
6017
6018         switch (hw->mac.type) {
6019         case ixgbe_mac_X550:
6020         case ixgbe_mac_X550EM_x:
6021         case ixgbe_mac_X550EM_a:
6022                 /* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
6023                 systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6024                 systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6025                                 * NSEC_PER_SEC;
6026                 break;
6027         default:
6028                 systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
6029                 systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
6030                                 << 32;
6031         }
6032
6033         return systime_cycles;
6034 }
6035
6036 static uint64_t
6037 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6038 {
6039         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6040         uint64_t rx_tstamp_cycles;
6041
6042         switch (hw->mac.type) {
6043         case ixgbe_mac_X550:
6044         case ixgbe_mac_X550EM_x:
6045         case ixgbe_mac_X550EM_a:
6046                 /* RXSTMPL stores ns and RXSTMPH stores seconds. */
6047                 rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6048                 rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6049                                 * NSEC_PER_SEC;
6050                 break;
6051         default:
6052                 /* RXSTMPL stores ns and RXSTMPH stores seconds. */
6053                 rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
6054                 rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
6055                                 << 32;
6056         }
6057
6058         return rx_tstamp_cycles;
6059 }
6060
6061 static uint64_t
6062 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
6063 {
6064         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6065         uint64_t tx_tstamp_cycles;
6066
6067         switch (hw->mac.type) {
6068         case ixgbe_mac_X550:
6069         case ixgbe_mac_X550EM_x:
6070         case ixgbe_mac_X550EM_a:
6071                 /* TXSTMPL stores ns and TXSTMPH stores seconds. */
6072                 tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6073                 tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6074                                 * NSEC_PER_SEC;
6075                 break;
6076         default:
6077                 /* TXSTMPL stores ns and TXSTMPH stores seconds. */
6078                 tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
6079                 tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
6080                                 << 32;
6081         }
6082
6083         return tx_tstamp_cycles;
6084 }
6085
6086 static void
6087 ixgbe_start_timecounters(struct rte_eth_dev *dev)
6088 {
6089         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6090         struct ixgbe_adapter *adapter =
6091                 (struct ixgbe_adapter *)dev->data->dev_private;
6092         struct rte_eth_link link;
6093         uint32_t incval = 0;
6094         uint32_t shift = 0;
6095
6096         /* Get current link speed. */
6097         memset(&link, 0, sizeof(link));
6098         ixgbe_dev_link_update(dev, 1);
6099         rte_ixgbe_dev_atomic_read_link_status(dev, &link);
6100
6101         switch (link.link_speed) {
6102         case ETH_SPEED_NUM_100M:
6103                 incval = IXGBE_INCVAL_100;
6104                 shift = IXGBE_INCVAL_SHIFT_100;
6105                 break;
6106         case ETH_SPEED_NUM_1G:
6107                 incval = IXGBE_INCVAL_1GB;
6108                 shift = IXGBE_INCVAL_SHIFT_1GB;
6109                 break;
6110         case ETH_SPEED_NUM_10G:
6111         default:
6112                 incval = IXGBE_INCVAL_10GB;
6113                 shift = IXGBE_INCVAL_SHIFT_10GB;
6114                 break;
6115         }
6116
6117         switch (hw->mac.type) {
6118         case ixgbe_mac_X550:
6119         case ixgbe_mac_X550EM_x:
6120         case ixgbe_mac_X550EM_a:
6121                 /* Independent of link speed. */
6122                 incval = 1;
6123                 /* Cycles read will be interpreted as ns. */
6124                 shift = 0;
6125                 /* Fall-through */
6126         case ixgbe_mac_X540:
6127                 IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
6128                 break;
6129         case ixgbe_mac_82599EB:
6130                 incval >>= IXGBE_INCVAL_SHIFT_82599;
6131                 shift -= IXGBE_INCVAL_SHIFT_82599;
6132                 IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
6133                                 (1 << IXGBE_INCPER_SHIFT_82599) | incval);
6134                 break;
6135         default:
6136                 /* Not supported. */
6137                 return;
6138         }
6139
6140         memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
6141         memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6142         memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
6143
6144         adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6145         adapter->systime_tc.cc_shift = shift;
6146         adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
6147
6148         adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6149         adapter->rx_tstamp_tc.cc_shift = shift;
6150         adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6151
6152         adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
6153         adapter->tx_tstamp_tc.cc_shift = shift;
6154         adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
6155 }
6156
6157 static int
6158 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
6159 {
6160         struct ixgbe_adapter *adapter =
6161                         (struct ixgbe_adapter *)dev->data->dev_private;
6162
6163         adapter->systime_tc.nsec += delta;
6164         adapter->rx_tstamp_tc.nsec += delta;
6165         adapter->tx_tstamp_tc.nsec += delta;
6166
6167         return 0;
6168 }
6169
6170 static int
6171 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
6172 {
6173         uint64_t ns;
6174         struct ixgbe_adapter *adapter =
6175                         (struct ixgbe_adapter *)dev->data->dev_private;
6176
6177         ns = rte_timespec_to_ns(ts);
6178         /* Set the timecounters to a new value. */
6179         adapter->systime_tc.nsec = ns;
6180         adapter->rx_tstamp_tc.nsec = ns;
6181         adapter->tx_tstamp_tc.nsec = ns;
6182
6183         return 0;
6184 }
6185
6186 static int
6187 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
6188 {
6189         uint64_t ns, systime_cycles;
6190         struct ixgbe_adapter *adapter =
6191                         (struct ixgbe_adapter *)dev->data->dev_private;
6192
6193         systime_cycles = ixgbe_read_systime_cyclecounter(dev);
6194         ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
6195         *ts = rte_ns_to_timespec(ns);
6196
6197         return 0;
6198 }
6199
6200 static int
6201 ixgbe_timesync_enable(struct rte_eth_dev *dev)
6202 {
6203         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6204         uint32_t tsync_ctl;
6205         uint32_t tsauxc;
6206
6207         /* Stop the timesync system time. */
6208         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
6209         /* Reset the timesync system time value. */
6210         IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
6211         IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
6212
6213         /* Enable system time for platforms where it isn't on by default. */
6214         tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
6215         tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
6216         IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
6217
6218         ixgbe_start_timecounters(dev);
6219
6220         /* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6221         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
6222                         (ETHER_TYPE_1588 |
6223                          IXGBE_ETQF_FILTER_EN |
6224                          IXGBE_ETQF_1588));
6225
6226         /* Enable timestamping of received PTP packets. */
6227         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6228         tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
6229         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6230
6231         /* Enable timestamping of transmitted PTP packets. */
6232         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6233         tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
6234         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6235
6236         IXGBE_WRITE_FLUSH(hw);
6237
6238         return 0;
6239 }
6240
6241 static int
6242 ixgbe_timesync_disable(struct rte_eth_dev *dev)
6243 {
6244         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6245         uint32_t tsync_ctl;
6246
6247         /* Disable timestamping of transmitted PTP packets. */
6248         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6249         tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
6250         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
6251
6252         /* Disable timestamping of received PTP packets. */
6253         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6254         tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
6255         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
6256
6257         /* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
6258         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
6259
6260         /* Stop incrementating the System Time registers. */
6261         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
6262
6263         return 0;
6264 }
6265
6266 static int
6267 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
6268                                  struct timespec *timestamp,
6269                                  uint32_t flags __rte_unused)
6270 {
6271         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6272         struct ixgbe_adapter *adapter =
6273                 (struct ixgbe_adapter *)dev->data->dev_private;
6274         uint32_t tsync_rxctl;
6275         uint64_t rx_tstamp_cycles;
6276         uint64_t ns;
6277
6278         tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
6279         if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
6280                 return -EINVAL;
6281
6282         rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
6283         ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
6284         *timestamp = rte_ns_to_timespec(ns);
6285
6286         return  0;
6287 }
6288
6289 static int
6290 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
6291                                  struct timespec *timestamp)
6292 {
6293         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6294         struct ixgbe_adapter *adapter =
6295                 (struct ixgbe_adapter *)dev->data->dev_private;
6296         uint32_t tsync_txctl;
6297         uint64_t tx_tstamp_cycles;
6298         uint64_t ns;
6299
6300         tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
6301         if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
6302                 return -EINVAL;
6303
6304         tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
6305         ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
6306         *timestamp = rte_ns_to_timespec(ns);
6307
6308         return 0;
6309 }
6310
6311 static int
6312 ixgbe_get_reg_length(struct rte_eth_dev *dev)
6313 {
6314         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6315         int count = 0;
6316         int g_ind = 0;
6317         const struct reg_info *reg_group;
6318         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6319                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6320
6321         while ((reg_group = reg_set[g_ind++]))
6322                 count += ixgbe_regs_group_count(reg_group);
6323
6324         return count;
6325 }
6326
6327 static int
6328 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
6329 {
6330         int count = 0;
6331         int g_ind = 0;
6332         const struct reg_info *reg_group;
6333
6334         while ((reg_group = ixgbevf_regs[g_ind++]))
6335                 count += ixgbe_regs_group_count(reg_group);
6336
6337         return count;
6338 }
6339
6340 static int
6341 ixgbe_get_regs(struct rte_eth_dev *dev,
6342               struct rte_dev_reg_info *regs)
6343 {
6344         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6345         uint32_t *data = regs->data;
6346         int g_ind = 0;
6347         int count = 0;
6348         const struct reg_info *reg_group;
6349         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
6350                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
6351
6352         if (data == NULL) {
6353                 regs->length = ixgbe_get_reg_length(dev);
6354                 regs->width = sizeof(uint32_t);
6355                 return 0;
6356         }
6357
6358         /* Support only full register dump */
6359         if ((regs->length == 0) ||
6360             (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
6361                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
6362                         hw->device_id;
6363                 while ((reg_group = reg_set[g_ind++]))
6364                         count += ixgbe_read_regs_group(dev, &data[count],
6365                                 reg_group);
6366                 return 0;
6367         }
6368
6369         return -ENOTSUP;
6370 }
6371
6372 static int
6373 ixgbevf_get_regs(struct rte_eth_dev *dev,
6374                 struct rte_dev_reg_info *regs)
6375 {
6376         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6377         uint32_t *data = regs->data;
6378         int g_ind = 0;
6379         int count = 0;
6380         const struct reg_info *reg_group;
6381
6382         if (data == NULL) {
6383                 regs->length = ixgbevf_get_reg_length(dev);
6384                 regs->width = sizeof(uint32_t);
6385                 return 0;
6386         }
6387
6388         /* Support only full register dump */
6389         if ((regs->length == 0) ||
6390             (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
6391                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
6392                         hw->device_id;
6393                 while ((reg_group = ixgbevf_regs[g_ind++]))
6394                         count += ixgbe_read_regs_group(dev, &data[count],
6395                                                       reg_group);
6396                 return 0;
6397         }
6398
6399         return -ENOTSUP;
6400 }
6401
6402 static int
6403 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
6404 {
6405         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6406
6407         /* Return unit is byte count */
6408         return hw->eeprom.word_size * 2;
6409 }
6410
6411 static int
6412 ixgbe_get_eeprom(struct rte_eth_dev *dev,
6413                 struct rte_dev_eeprom_info *in_eeprom)
6414 {
6415         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6416         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
6417         uint16_t *data = in_eeprom->data;
6418         int first, length;
6419
6420         first = in_eeprom->offset >> 1;
6421         length = in_eeprom->length >> 1;
6422         if ((first > hw->eeprom.word_size) ||
6423             ((first + length) > hw->eeprom.word_size))
6424                 return -EINVAL;
6425
6426         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
6427
6428         return eeprom->ops.read_buffer(hw, first, length, data);
6429 }
6430
6431 static int
6432 ixgbe_set_eeprom(struct rte_eth_dev *dev,
6433                 struct rte_dev_eeprom_info *in_eeprom)
6434 {
6435         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6436         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
6437         uint16_t *data = in_eeprom->data;
6438         int first, length;
6439
6440         first = in_eeprom->offset >> 1;
6441         length = in_eeprom->length >> 1;
6442         if ((first > hw->eeprom.word_size) ||
6443             ((first + length) > hw->eeprom.word_size))
6444                 return -EINVAL;
6445
6446         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
6447
6448         return eeprom->ops.write_buffer(hw,  first, length, data);
6449 }
6450
6451 uint16_t
6452 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
6453         switch (mac_type) {
6454         case ixgbe_mac_X550:
6455         case ixgbe_mac_X550EM_x:
6456         case ixgbe_mac_X550EM_a:
6457                 return ETH_RSS_RETA_SIZE_512;
6458         case ixgbe_mac_X550_vf:
6459         case ixgbe_mac_X550EM_x_vf:
6460         case ixgbe_mac_X550EM_a_vf:
6461                 return ETH_RSS_RETA_SIZE_64;
6462         default:
6463                 return ETH_RSS_RETA_SIZE_128;
6464         }
6465 }
6466
6467 uint32_t
6468 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
6469         switch (mac_type) {
6470         case ixgbe_mac_X550:
6471         case ixgbe_mac_X550EM_x:
6472         case ixgbe_mac_X550EM_a:
6473                 if (reta_idx < ETH_RSS_RETA_SIZE_128)
6474                         return IXGBE_RETA(reta_idx >> 2);
6475                 else
6476                         return IXGBE_ERETA((reta_idx - ETH_RSS_RETA_SIZE_128) >> 2);
6477         case ixgbe_mac_X550_vf:
6478         case ixgbe_mac_X550EM_x_vf:
6479         case ixgbe_mac_X550EM_a_vf:
6480                 return IXGBE_VFRETA(reta_idx >> 2);
6481         default:
6482                 return IXGBE_RETA(reta_idx >> 2);
6483         }
6484 }
6485
6486 uint32_t
6487 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
6488         switch (mac_type) {
6489         case ixgbe_mac_X550_vf:
6490         case ixgbe_mac_X550EM_x_vf:
6491         case ixgbe_mac_X550EM_a_vf:
6492                 return IXGBE_VFMRQC;
6493         default:
6494                 return IXGBE_MRQC;
6495         }
6496 }
6497
6498 uint32_t
6499 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
6500         switch (mac_type) {
6501         case ixgbe_mac_X550_vf:
6502         case ixgbe_mac_X550EM_x_vf:
6503         case ixgbe_mac_X550EM_a_vf:
6504                 return IXGBE_VFRSSRK(i);
6505         default:
6506                 return IXGBE_RSSRK(i);
6507         }
6508 }
6509
6510 bool
6511 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
6512         switch (mac_type) {
6513         case ixgbe_mac_82599_vf:
6514         case ixgbe_mac_X540_vf:
6515                 return 0;
6516         default:
6517                 return 1;
6518         }
6519 }
6520
6521 static int
6522 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
6523                         struct rte_eth_dcb_info *dcb_info)
6524 {
6525         struct ixgbe_dcb_config *dcb_config =
6526                         IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
6527         struct ixgbe_dcb_tc_config *tc;
6528         uint8_t i, j;
6529
6530         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_DCB_FLAG)
6531                 dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
6532         else
6533                 dcb_info->nb_tcs = 1;
6534
6535         if (dcb_config->vt_mode) { /* vt is enabled*/
6536                 struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
6537                                 &dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
6538                 for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
6539                         dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
6540                 for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
6541                         for (j = 0; j < dcb_info->nb_tcs; j++) {
6542                                 dcb_info->tc_queue.tc_rxq[i][j].base =
6543                                                 i * dcb_info->nb_tcs + j;
6544                                 dcb_info->tc_queue.tc_rxq[i][j].nb_queue = 1;
6545                                 dcb_info->tc_queue.tc_txq[i][j].base =
6546                                                 i * dcb_info->nb_tcs + j;
6547                                 dcb_info->tc_queue.tc_txq[i][j].nb_queue = 1;
6548                         }
6549                 }
6550         } else { /* vt is disabled*/
6551                 struct rte_eth_dcb_rx_conf *rx_conf =
6552                                 &dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
6553                 for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
6554                         dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
6555                 if (dcb_info->nb_tcs == ETH_4_TCS) {
6556                         for (i = 0; i < dcb_info->nb_tcs; i++) {
6557                                 dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
6558                                 dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
6559                         }
6560                         dcb_info->tc_queue.tc_txq[0][0].base = 0;
6561                         dcb_info->tc_queue.tc_txq[0][1].base = 64;
6562                         dcb_info->tc_queue.tc_txq[0][2].base = 96;
6563                         dcb_info->tc_queue.tc_txq[0][3].base = 112;
6564                         dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
6565                         dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
6566                         dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
6567                         dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
6568                 } else if (dcb_info->nb_tcs == ETH_8_TCS) {
6569                         for (i = 0; i < dcb_info->nb_tcs; i++) {
6570                                 dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
6571                                 dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
6572                         }
6573                         dcb_info->tc_queue.tc_txq[0][0].base = 0;
6574                         dcb_info->tc_queue.tc_txq[0][1].base = 32;
6575                         dcb_info->tc_queue.tc_txq[0][2].base = 64;
6576                         dcb_info->tc_queue.tc_txq[0][3].base = 80;
6577                         dcb_info->tc_queue.tc_txq[0][4].base = 96;
6578                         dcb_info->tc_queue.tc_txq[0][5].base = 104;
6579                         dcb_info->tc_queue.tc_txq[0][6].base = 112;
6580                         dcb_info->tc_queue.tc_txq[0][7].base = 120;
6581                         dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
6582                         dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
6583                         dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
6584                         dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
6585                         dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
6586                         dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
6587                         dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
6588                         dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
6589                 }
6590         }
6591         for (i = 0; i < dcb_info->nb_tcs; i++) {
6592                 tc = &dcb_config->tc_config[i];
6593                 dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
6594         }
6595         return 0;
6596 }
6597
6598 /* Update e-tag ether type */
6599 static int
6600 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
6601                             uint16_t ether_type)
6602 {
6603         uint32_t etag_etype;
6604
6605         if (hw->mac.type != ixgbe_mac_X550 &&
6606             hw->mac.type != ixgbe_mac_X550EM_x &&
6607             hw->mac.type != ixgbe_mac_X550EM_a) {
6608                 return -ENOTSUP;
6609         }
6610
6611         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
6612         etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
6613         etag_etype |= ether_type;
6614         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
6615         IXGBE_WRITE_FLUSH(hw);
6616
6617         return 0;
6618 }
6619
6620 /* Config l2 tunnel ether type */
6621 static int
6622 ixgbe_dev_l2_tunnel_eth_type_conf(struct rte_eth_dev *dev,
6623                                   struct rte_eth_l2_tunnel_conf *l2_tunnel)
6624 {
6625         int ret = 0;
6626         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6627
6628         if (l2_tunnel == NULL)
6629                 return -EINVAL;
6630
6631         switch (l2_tunnel->l2_tunnel_type) {
6632         case RTE_L2_TUNNEL_TYPE_E_TAG:
6633                 ret = ixgbe_update_e_tag_eth_type(hw, l2_tunnel->ether_type);
6634                 break;
6635         default:
6636                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
6637                 ret = -EINVAL;
6638                 break;
6639         }
6640
6641         return ret;
6642 }
6643
6644 /* Enable e-tag tunnel */
6645 static int
6646 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
6647 {
6648         uint32_t etag_etype;
6649
6650         if (hw->mac.type != ixgbe_mac_X550 &&
6651             hw->mac.type != ixgbe_mac_X550EM_x &&
6652             hw->mac.type != ixgbe_mac_X550EM_a) {
6653                 return -ENOTSUP;
6654         }
6655
6656         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
6657         etag_etype |= IXGBE_ETAG_ETYPE_VALID;
6658         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
6659         IXGBE_WRITE_FLUSH(hw);
6660
6661         return 0;
6662 }
6663
6664 /* Enable l2 tunnel */
6665 static int
6666 ixgbe_dev_l2_tunnel_enable(struct rte_eth_dev *dev,
6667                            enum rte_eth_tunnel_type l2_tunnel_type)
6668 {
6669         int ret = 0;
6670         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6671
6672         switch (l2_tunnel_type) {
6673         case RTE_L2_TUNNEL_TYPE_E_TAG:
6674                 ret = ixgbe_e_tag_enable(hw);
6675                 break;
6676         default:
6677                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
6678                 ret = -EINVAL;
6679                 break;
6680         }
6681
6682         return ret;
6683 }
6684
6685 /* Disable e-tag tunnel */
6686 static int
6687 ixgbe_e_tag_disable(struct ixgbe_hw *hw)
6688 {
6689         uint32_t etag_etype;
6690
6691         if (hw->mac.type != ixgbe_mac_X550 &&
6692             hw->mac.type != ixgbe_mac_X550EM_x &&
6693             hw->mac.type != ixgbe_mac_X550EM_a) {
6694                 return -ENOTSUP;
6695         }
6696
6697         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
6698         etag_etype &= ~IXGBE_ETAG_ETYPE_VALID;
6699         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
6700         IXGBE_WRITE_FLUSH(hw);
6701
6702         return 0;
6703 }
6704
6705 /* Disable l2 tunnel */
6706 static int
6707 ixgbe_dev_l2_tunnel_disable(struct rte_eth_dev *dev,
6708                             enum rte_eth_tunnel_type l2_tunnel_type)
6709 {
6710         int ret = 0;
6711         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6712
6713         switch (l2_tunnel_type) {
6714         case RTE_L2_TUNNEL_TYPE_E_TAG:
6715                 ret = ixgbe_e_tag_disable(hw);
6716                 break;
6717         default:
6718                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
6719                 ret = -EINVAL;
6720                 break;
6721         }
6722
6723         return ret;
6724 }
6725
6726 static int
6727 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
6728                        struct rte_eth_l2_tunnel_conf *l2_tunnel)
6729 {
6730         int ret = 0;
6731         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6732         uint32_t i, rar_entries;
6733         uint32_t rar_low, rar_high;
6734
6735         if (hw->mac.type != ixgbe_mac_X550 &&
6736             hw->mac.type != ixgbe_mac_X550EM_x &&
6737             hw->mac.type != ixgbe_mac_X550EM_a) {
6738                 return -ENOTSUP;
6739         }
6740
6741         rar_entries = ixgbe_get_num_rx_addrs(hw);
6742
6743         for (i = 1; i < rar_entries; i++) {
6744                 rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
6745                 rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
6746                 if ((rar_high & IXGBE_RAH_AV) &&
6747                     (rar_high & IXGBE_RAH_ADTYPE) &&
6748                     ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
6749                      l2_tunnel->tunnel_id)) {
6750                         IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
6751                         IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
6752
6753                         ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
6754
6755                         return ret;
6756                 }
6757         }
6758
6759         return ret;
6760 }
6761
6762 static int
6763 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
6764                        struct rte_eth_l2_tunnel_conf *l2_tunnel)
6765 {
6766         int ret = 0;
6767         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6768         uint32_t i, rar_entries;
6769         uint32_t rar_low, rar_high;
6770
6771         if (hw->mac.type != ixgbe_mac_X550 &&
6772             hw->mac.type != ixgbe_mac_X550EM_x &&
6773             hw->mac.type != ixgbe_mac_X550EM_a) {
6774                 return -ENOTSUP;
6775         }
6776
6777         /* One entry for one tunnel. Try to remove potential existing entry. */
6778         ixgbe_e_tag_filter_del(dev, l2_tunnel);
6779
6780         rar_entries = ixgbe_get_num_rx_addrs(hw);
6781
6782         for (i = 1; i < rar_entries; i++) {
6783                 rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
6784                 if (rar_high & IXGBE_RAH_AV) {
6785                         continue;
6786                 } else {
6787                         ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
6788                         rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
6789                         rar_low = l2_tunnel->tunnel_id;
6790
6791                         IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
6792                         IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
6793
6794                         return ret;
6795                 }
6796         }
6797
6798         PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
6799                      " Please remove a rule before adding a new one.");
6800         return -EINVAL;
6801 }
6802
6803 /* Add l2 tunnel filter */
6804 static int
6805 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
6806                                struct rte_eth_l2_tunnel_conf *l2_tunnel)
6807 {
6808         int ret = 0;
6809
6810         switch (l2_tunnel->l2_tunnel_type) {
6811         case RTE_L2_TUNNEL_TYPE_E_TAG:
6812                 ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
6813                 break;
6814         default:
6815                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
6816                 ret = -EINVAL;
6817                 break;
6818         }
6819
6820         return ret;
6821 }
6822
6823 /* Delete l2 tunnel filter */
6824 static int
6825 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
6826                                struct rte_eth_l2_tunnel_conf *l2_tunnel)
6827 {
6828         int ret = 0;
6829
6830         switch (l2_tunnel->l2_tunnel_type) {
6831         case RTE_L2_TUNNEL_TYPE_E_TAG:
6832                 ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
6833                 break;
6834         default:
6835                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
6836                 ret = -EINVAL;
6837                 break;
6838         }
6839
6840         return ret;
6841 }
6842
6843 /**
6844  * ixgbe_dev_l2_tunnel_filter_handle - Handle operations for l2 tunnel filter.
6845  * @dev: pointer to rte_eth_dev structure
6846  * @filter_op:operation will be taken.
6847  * @arg: a pointer to specific structure corresponding to the filter_op
6848  */
6849 static int
6850 ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
6851                                   enum rte_filter_op filter_op,
6852                                   void *arg)
6853 {
6854         int ret = 0;
6855
6856         if (filter_op == RTE_ETH_FILTER_NOP)
6857                 return 0;
6858
6859         if (arg == NULL) {
6860                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
6861                             filter_op);
6862                 return -EINVAL;
6863         }
6864
6865         switch (filter_op) {
6866         case RTE_ETH_FILTER_ADD:
6867                 ret = ixgbe_dev_l2_tunnel_filter_add
6868                         (dev,
6869                          (struct rte_eth_l2_tunnel_conf *)arg);
6870                 break;
6871         case RTE_ETH_FILTER_DELETE:
6872                 ret = ixgbe_dev_l2_tunnel_filter_del
6873                         (dev,
6874                          (struct rte_eth_l2_tunnel_conf *)arg);
6875                 break;
6876         default:
6877                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
6878                 ret = -EINVAL;
6879                 break;
6880         }
6881         return ret;
6882 }
6883
6884 static int
6885 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
6886 {
6887         int ret = 0;
6888         uint32_t ctrl;
6889         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6890
6891         if (hw->mac.type != ixgbe_mac_X550 &&
6892             hw->mac.type != ixgbe_mac_X550EM_x &&
6893             hw->mac.type != ixgbe_mac_X550EM_a) {
6894                 return -ENOTSUP;
6895         }
6896
6897         ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
6898         ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
6899         if (en)
6900                 ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
6901         IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
6902
6903         return ret;
6904 }
6905
6906 /* Enable l2 tunnel forwarding */
6907 static int
6908 ixgbe_dev_l2_tunnel_forwarding_enable
6909         (struct rte_eth_dev *dev,
6910          enum rte_eth_tunnel_type l2_tunnel_type)
6911 {
6912         int ret = 0;
6913
6914         switch (l2_tunnel_type) {
6915         case RTE_L2_TUNNEL_TYPE_E_TAG:
6916                 ret = ixgbe_e_tag_forwarding_en_dis(dev, 1);
6917                 break;
6918         default:
6919                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
6920                 ret = -EINVAL;
6921                 break;
6922         }
6923
6924         return ret;
6925 }
6926
6927 /* Disable l2 tunnel forwarding */
6928 static int
6929 ixgbe_dev_l2_tunnel_forwarding_disable
6930         (struct rte_eth_dev *dev,
6931          enum rte_eth_tunnel_type l2_tunnel_type)
6932 {
6933         int ret = 0;
6934
6935         switch (l2_tunnel_type) {
6936         case RTE_L2_TUNNEL_TYPE_E_TAG:
6937                 ret = ixgbe_e_tag_forwarding_en_dis(dev, 0);
6938                 break;
6939         default:
6940                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
6941                 ret = -EINVAL;
6942                 break;
6943         }
6944
6945         return ret;
6946 }
6947
6948 static int
6949 ixgbe_e_tag_insertion_en_dis(struct rte_eth_dev *dev,
6950                              struct rte_eth_l2_tunnel_conf *l2_tunnel,
6951                              bool en)
6952 {
6953         int ret = 0;
6954         uint32_t vmtir, vmvir;
6955         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6956
6957         if (l2_tunnel->vf_id >= dev->pci_dev->max_vfs) {
6958                 PMD_DRV_LOG(ERR,
6959                             "VF id %u should be less than %u",
6960                             l2_tunnel->vf_id,
6961                             dev->pci_dev->max_vfs);
6962                 return -EINVAL;
6963         }
6964
6965         if (hw->mac.type != ixgbe_mac_X550 &&
6966             hw->mac.type != ixgbe_mac_X550EM_x &&
6967             hw->mac.type != ixgbe_mac_X550EM_a) {
6968                 return -ENOTSUP;
6969         }
6970
6971         if (en)
6972                 vmtir = l2_tunnel->tunnel_id;
6973         else
6974                 vmtir = 0;
6975
6976         IXGBE_WRITE_REG(hw, IXGBE_VMTIR(l2_tunnel->vf_id), vmtir);
6977
6978         vmvir = IXGBE_READ_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id));
6979         vmvir &= ~IXGBE_VMVIR_TAGA_MASK;
6980         if (en)
6981                 vmvir |= IXGBE_VMVIR_TAGA_ETAG_INSERT;
6982         IXGBE_WRITE_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id), vmvir);
6983
6984         return ret;
6985 }
6986
6987 /* Enable l2 tunnel tag insertion */
6988 static int
6989 ixgbe_dev_l2_tunnel_insertion_enable(struct rte_eth_dev *dev,
6990                                      struct rte_eth_l2_tunnel_conf *l2_tunnel)
6991 {
6992         int ret = 0;
6993
6994         switch (l2_tunnel->l2_tunnel_type) {
6995         case RTE_L2_TUNNEL_TYPE_E_TAG:
6996                 ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 1);
6997                 break;
6998         default:
6999                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7000                 ret = -EINVAL;
7001                 break;
7002         }
7003
7004         return ret;
7005 }
7006
7007 /* Disable l2 tunnel tag insertion */
7008 static int
7009 ixgbe_dev_l2_tunnel_insertion_disable
7010         (struct rte_eth_dev *dev,
7011          struct rte_eth_l2_tunnel_conf *l2_tunnel)
7012 {
7013         int ret = 0;
7014
7015         switch (l2_tunnel->l2_tunnel_type) {
7016         case RTE_L2_TUNNEL_TYPE_E_TAG:
7017                 ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 0);
7018                 break;
7019         default:
7020                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7021                 ret = -EINVAL;
7022                 break;
7023         }
7024
7025         return ret;
7026 }
7027
7028 static int
7029 ixgbe_e_tag_stripping_en_dis(struct rte_eth_dev *dev,
7030                              bool en)
7031 {
7032         int ret = 0;
7033         uint32_t qde;
7034         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7035
7036         if (hw->mac.type != ixgbe_mac_X550 &&
7037             hw->mac.type != ixgbe_mac_X550EM_x &&
7038             hw->mac.type != ixgbe_mac_X550EM_a) {
7039                 return -ENOTSUP;
7040         }
7041
7042         qde = IXGBE_READ_REG(hw, IXGBE_QDE);
7043         if (en)
7044                 qde |= IXGBE_QDE_STRIP_TAG;
7045         else
7046                 qde &= ~IXGBE_QDE_STRIP_TAG;
7047         qde &= ~IXGBE_QDE_READ;
7048         qde |= IXGBE_QDE_WRITE;
7049         IXGBE_WRITE_REG(hw, IXGBE_QDE, qde);
7050
7051         return ret;
7052 }
7053
7054 /* Enable l2 tunnel tag stripping */
7055 static int
7056 ixgbe_dev_l2_tunnel_stripping_enable
7057         (struct rte_eth_dev *dev,
7058          enum rte_eth_tunnel_type l2_tunnel_type)
7059 {
7060         int ret = 0;
7061
7062         switch (l2_tunnel_type) {
7063         case RTE_L2_TUNNEL_TYPE_E_TAG:
7064                 ret = ixgbe_e_tag_stripping_en_dis(dev, 1);
7065                 break;
7066         default:
7067                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7068                 ret = -EINVAL;
7069                 break;
7070         }
7071
7072         return ret;
7073 }
7074
7075 /* Disable l2 tunnel tag stripping */
7076 static int
7077 ixgbe_dev_l2_tunnel_stripping_disable
7078         (struct rte_eth_dev *dev,
7079          enum rte_eth_tunnel_type l2_tunnel_type)
7080 {
7081         int ret = 0;
7082
7083         switch (l2_tunnel_type) {
7084         case RTE_L2_TUNNEL_TYPE_E_TAG:
7085                 ret = ixgbe_e_tag_stripping_en_dis(dev, 0);
7086                 break;
7087         default:
7088                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7089                 ret = -EINVAL;
7090                 break;
7091         }
7092
7093         return ret;
7094 }
7095
7096 /* Enable/disable l2 tunnel offload functions */
7097 static int
7098 ixgbe_dev_l2_tunnel_offload_set
7099         (struct rte_eth_dev *dev,
7100          struct rte_eth_l2_tunnel_conf *l2_tunnel,
7101          uint32_t mask,
7102          uint8_t en)
7103 {
7104         int ret = 0;
7105
7106         if (l2_tunnel == NULL)
7107                 return -EINVAL;
7108
7109         ret = -EINVAL;
7110         if (mask & ETH_L2_TUNNEL_ENABLE_MASK) {
7111                 if (en)
7112                         ret = ixgbe_dev_l2_tunnel_enable(
7113                                 dev,
7114                                 l2_tunnel->l2_tunnel_type);
7115                 else
7116                         ret = ixgbe_dev_l2_tunnel_disable(
7117                                 dev,
7118                                 l2_tunnel->l2_tunnel_type);
7119         }
7120
7121         if (mask & ETH_L2_TUNNEL_INSERTION_MASK) {
7122                 if (en)
7123                         ret = ixgbe_dev_l2_tunnel_insertion_enable(
7124                                 dev,
7125                                 l2_tunnel);
7126                 else
7127                         ret = ixgbe_dev_l2_tunnel_insertion_disable(
7128                                 dev,
7129                                 l2_tunnel);
7130         }
7131
7132         if (mask & ETH_L2_TUNNEL_STRIPPING_MASK) {
7133                 if (en)
7134                         ret = ixgbe_dev_l2_tunnel_stripping_enable(
7135                                 dev,
7136                                 l2_tunnel->l2_tunnel_type);
7137                 else
7138                         ret = ixgbe_dev_l2_tunnel_stripping_disable(
7139                                 dev,
7140                                 l2_tunnel->l2_tunnel_type);
7141         }
7142
7143         if (mask & ETH_L2_TUNNEL_FORWARDING_MASK) {
7144                 if (en)
7145                         ret = ixgbe_dev_l2_tunnel_forwarding_enable(
7146                                 dev,
7147                                 l2_tunnel->l2_tunnel_type);
7148                 else
7149                         ret = ixgbe_dev_l2_tunnel_forwarding_disable(
7150                                 dev,
7151                                 l2_tunnel->l2_tunnel_type);
7152         }
7153
7154         return ret;
7155 }
7156
7157 static int
7158 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
7159                         uint16_t port)
7160 {
7161         IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
7162         IXGBE_WRITE_FLUSH(hw);
7163
7164         return 0;
7165 }
7166
7167 /* There's only one register for VxLAN UDP port.
7168  * So, we cannot add several ports. Will update it.
7169  */
7170 static int
7171 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
7172                      uint16_t port)
7173 {
7174         if (port == 0) {
7175                 PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
7176                 return -EINVAL;
7177         }
7178
7179         return ixgbe_update_vxlan_port(hw, port);
7180 }
7181
7182 /* We cannot delete the VxLAN port. For there's a register for VxLAN
7183  * UDP port, it must have a value.
7184  * So, will reset it to the original value 0.
7185  */
7186 static int
7187 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
7188                      uint16_t port)
7189 {
7190         uint16_t cur_port;
7191
7192         cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
7193
7194         if (cur_port != port) {
7195                 PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
7196                 return -EINVAL;
7197         }
7198
7199         return ixgbe_update_vxlan_port(hw, 0);
7200 }
7201
7202 /* Add UDP tunneling port */
7203 static int
7204 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
7205                               struct rte_eth_udp_tunnel *udp_tunnel)
7206 {
7207         int ret = 0;
7208         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7209
7210         if (hw->mac.type != ixgbe_mac_X550 &&
7211             hw->mac.type != ixgbe_mac_X550EM_x &&
7212             hw->mac.type != ixgbe_mac_X550EM_a) {
7213                 return -ENOTSUP;
7214         }
7215
7216         if (udp_tunnel == NULL)
7217                 return -EINVAL;
7218
7219         switch (udp_tunnel->prot_type) {
7220         case RTE_TUNNEL_TYPE_VXLAN:
7221                 ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
7222                 break;
7223
7224         case RTE_TUNNEL_TYPE_GENEVE:
7225         case RTE_TUNNEL_TYPE_TEREDO:
7226                 PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7227                 ret = -EINVAL;
7228                 break;
7229
7230         default:
7231                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7232                 ret = -EINVAL;
7233                 break;
7234         }
7235
7236         return ret;
7237 }
7238
7239 /* Remove UDP tunneling port */
7240 static int
7241 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
7242                               struct rte_eth_udp_tunnel *udp_tunnel)
7243 {
7244         int ret = 0;
7245         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7246
7247         if (hw->mac.type != ixgbe_mac_X550 &&
7248             hw->mac.type != ixgbe_mac_X550EM_x &&
7249             hw->mac.type != ixgbe_mac_X550EM_a) {
7250                 return -ENOTSUP;
7251         }
7252
7253         if (udp_tunnel == NULL)
7254                 return -EINVAL;
7255
7256         switch (udp_tunnel->prot_type) {
7257         case RTE_TUNNEL_TYPE_VXLAN:
7258                 ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
7259                 break;
7260         case RTE_TUNNEL_TYPE_GENEVE:
7261         case RTE_TUNNEL_TYPE_TEREDO:
7262                 PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
7263                 ret = -EINVAL;
7264                 break;
7265         default:
7266                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7267                 ret = -EINVAL;
7268                 break;
7269         }
7270
7271         return ret;
7272 }
7273
7274 static void
7275 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
7276 {
7277         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7278
7279         hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_ALLMULTI);
7280 }
7281
7282 static void
7283 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
7284 {
7285         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7286
7287         hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_NONE);
7288 }
7289
7290 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
7291 {
7292         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7293         u32 in_msg = 0;
7294
7295         if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
7296                 return;
7297
7298         /* PF reset VF event */
7299         if (in_msg == IXGBE_PF_CONTROL_MSG)
7300                 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET);
7301 }
7302
7303 static int
7304 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
7305 {
7306         uint32_t eicr;
7307         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7308         struct ixgbe_interrupt *intr =
7309                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7310         ixgbevf_intr_disable(hw);
7311
7312         /* read-on-clear nic registers here */
7313         eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
7314         intr->flags = 0;
7315
7316         /* only one misc vector supported - mailbox */
7317         eicr &= IXGBE_VTEICR_MASK;
7318         if (eicr == IXGBE_MISC_VEC_ID)
7319                 intr->flags |= IXGBE_FLAG_MAILBOX;
7320
7321         return 0;
7322 }
7323
7324 static int
7325 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
7326 {
7327         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7328         struct ixgbe_interrupt *intr =
7329                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
7330
7331         if (intr->flags & IXGBE_FLAG_MAILBOX) {
7332                 ixgbevf_mbx_process(dev);
7333                 intr->flags &= ~IXGBE_FLAG_MAILBOX;
7334         }
7335
7336         ixgbevf_intr_enable(hw);
7337
7338         return 0;
7339 }
7340
7341 static void
7342 ixgbevf_dev_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
7343                               void *param)
7344 {
7345         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
7346
7347         ixgbevf_dev_interrupt_get_status(dev);
7348         ixgbevf_dev_interrupt_action(dev);
7349 }
7350
7351 DRIVER_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd.pci_drv);
7352 DRIVER_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
7353 DRIVER_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd.pci_drv);
7354 DRIVER_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);