ixgbe: remove jabber count from Rx errors
[dpdk.git] / drivers / net / ixgbe / ixgbe_ethdev.c
1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright(c) 2010-2015 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 IXGBE_TIMINCA_16NS_SHIFT 24
130 #define IXGBE_TIMINCA_INCVALUE   16000000
131 #define IXGBE_TIMINCA_INIT       ((0x02 << IXGBE_TIMINCA_16NS_SHIFT) \
132                                   | IXGBE_TIMINCA_INCVALUE)
133
134 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev);
135 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
136 static int  ixgbe_dev_configure(struct rte_eth_dev *dev);
137 static int  ixgbe_dev_start(struct rte_eth_dev *dev);
138 static void ixgbe_dev_stop(struct rte_eth_dev *dev);
139 static int  ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
140 static int  ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
141 static void ixgbe_dev_close(struct rte_eth_dev *dev);
142 static void ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
143 static void ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
144 static void ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
145 static void ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
146 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
147                                 int wait_to_complete);
148 static void ixgbe_dev_stats_get(struct rte_eth_dev *dev,
149                                 struct rte_eth_stats *stats);
150 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
151                                 struct rte_eth_xstats *xstats, unsigned n);
152 static void ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
153 static void ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
154 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
155                                              uint16_t queue_id,
156                                              uint8_t stat_idx,
157                                              uint8_t is_rx);
158 static void ixgbe_dev_info_get(struct rte_eth_dev *dev,
159                                struct rte_eth_dev_info *dev_info);
160 static void ixgbevf_dev_info_get(struct rte_eth_dev *dev,
161                                  struct rte_eth_dev_info *dev_info);
162 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
163
164 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
165                 uint16_t vlan_id, int on);
166 static void ixgbe_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid_id);
167 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
168                 uint16_t queue, bool on);
169 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
170                 int on);
171 static void ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
172 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
173 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
174 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
175 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
176
177 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
178 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
179 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
180                                struct rte_eth_fc_conf *fc_conf);
181 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
182                                struct rte_eth_fc_conf *fc_conf);
183 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
184                 struct rte_eth_pfc_conf *pfc_conf);
185 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
186                         struct rte_eth_rss_reta_entry64 *reta_conf,
187                         uint16_t reta_size);
188 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
189                         struct rte_eth_rss_reta_entry64 *reta_conf,
190                         uint16_t reta_size);
191 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
192 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev);
193 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
194 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
195 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev);
196 static void ixgbe_dev_interrupt_handler(struct rte_intr_handle *handle,
197                 void *param);
198 static void ixgbe_dev_interrupt_delayed_handler(void *param);
199 static void ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
200                 uint32_t index, uint32_t pool);
201 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
202 static void ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
203                                            struct ether_addr *mac_addr);
204 static void ixgbe_dcb_init(struct ixgbe_hw *hw,struct ixgbe_dcb_config *dcb_config);
205
206 /* For Virtual Function support */
207 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
208 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
209 static int ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev);
210 static int ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev);
211 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
212 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
213 static void ixgbevf_dev_stop(struct rte_eth_dev *dev);
214 static void ixgbevf_dev_close(struct rte_eth_dev *dev);
215 static void ixgbevf_intr_disable(struct ixgbe_hw *hw);
216 static void ixgbevf_intr_enable(struct ixgbe_hw *hw);
217 static void ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
218                 struct rte_eth_stats *stats);
219 static void ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
220 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
221                 uint16_t vlan_id, int on);
222 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
223                 uint16_t queue, int on);
224 static void ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
225 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
226 static void ixgbevf_dev_interrupt_handler(struct rte_intr_handle *handle,
227                                           void *param);
228 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
229                                             uint16_t queue_id);
230 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
231                                              uint16_t queue_id);
232 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
233                                  uint8_t queue, uint8_t msix_vector);
234 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
235
236 /* For Eth VMDQ APIs support */
237 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
238                 ether_addr* mac_addr,uint8_t on);
239 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev,uint8_t on);
240 static int  ixgbe_set_pool_rx_mode(struct rte_eth_dev *dev,  uint16_t pool,
241                 uint16_t rx_mask, uint8_t on);
242 static int ixgbe_set_pool_rx(struct rte_eth_dev *dev,uint16_t pool,uint8_t on);
243 static int ixgbe_set_pool_tx(struct rte_eth_dev *dev,uint16_t pool,uint8_t on);
244 static int ixgbe_set_pool_vlan_filter(struct rte_eth_dev *dev, uint16_t vlan,
245                 uint64_t pool_mask,uint8_t vlan_on);
246 static int ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
247                 struct rte_eth_mirror_conf *mirror_conf,
248                 uint8_t rule_id, uint8_t on);
249 static int ixgbe_mirror_rule_reset(struct rte_eth_dev *dev,
250                 uint8_t rule_id);
251 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
252                                           uint16_t queue_id);
253 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
254                                            uint16_t queue_id);
255 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
256                                uint8_t queue, uint8_t msix_vector);
257 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
258
259 static int ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
260                 uint16_t queue_idx, uint16_t tx_rate);
261 static int ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
262                 uint16_t tx_rate, uint64_t q_msk);
263
264 static void ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
265                                  struct ether_addr *mac_addr,
266                                  uint32_t index, uint32_t pool);
267 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
268 static void ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
269                                              struct ether_addr *mac_addr);
270 static int ixgbe_syn_filter_set(struct rte_eth_dev *dev,
271                         struct rte_eth_syn_filter *filter,
272                         bool add);
273 static int ixgbe_syn_filter_get(struct rte_eth_dev *dev,
274                         struct rte_eth_syn_filter *filter);
275 static int ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
276                         enum rte_filter_op filter_op,
277                         void *arg);
278 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
279                         struct ixgbe_5tuple_filter *filter);
280 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
281                         struct ixgbe_5tuple_filter *filter);
282 static int ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
283                         struct rte_eth_ntuple_filter *filter,
284                         bool add);
285 static int ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
286                                 enum rte_filter_op filter_op,
287                                 void *arg);
288 static int ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
289                         struct rte_eth_ntuple_filter *filter);
290 static int ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
291                         struct rte_eth_ethertype_filter *filter,
292                         bool add);
293 static int ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
294                                 enum rte_filter_op filter_op,
295                                 void *arg);
296 static int ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
297                         struct rte_eth_ethertype_filter *filter);
298 static int ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
299                      enum rte_filter_type filter_type,
300                      enum rte_filter_op filter_op,
301                      void *arg);
302 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
303
304 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
305                                       struct ether_addr *mc_addr_set,
306                                       uint32_t nb_mc_addr);
307
308 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
309 static int ixgbe_get_regs(struct rte_eth_dev *dev,
310                             struct rte_dev_reg_info *regs);
311 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
312 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
313                                 struct rte_dev_eeprom_info *eeprom);
314 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
315                                 struct rte_dev_eeprom_info *eeprom);
316
317 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
318 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
319                                 struct rte_dev_reg_info *regs);
320
321 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
322 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
323 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
324                                             struct timespec *timestamp,
325                                             uint32_t flags);
326 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
327                                             struct timespec *timestamp);
328
329 /*
330  * Define VF Stats MACRO for Non "cleared on read" register
331  */
332 #define UPDATE_VF_STAT(reg, last, cur)                          \
333 {                                                               \
334         uint32_t latest = IXGBE_READ_REG(hw, reg);              \
335         cur += (latest - last) & UINT_MAX;                      \
336         last = latest;                                          \
337 }
338
339 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
340 {                                                                \
341         u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
342         u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
343         u64 latest = ((new_msb << 32) | new_lsb);                \
344         cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
345         last = latest;                                           \
346 }
347
348 #define IXGBE_SET_HWSTRIP(h, q) do{\
349                 uint32_t idx = (q) / (sizeof ((h)->bitmap[0]) * NBBY); \
350                 uint32_t bit = (q) % (sizeof ((h)->bitmap[0]) * NBBY); \
351                 (h)->bitmap[idx] |= 1 << bit;\
352         }while(0)
353
354 #define IXGBE_CLEAR_HWSTRIP(h, q) do{\
355                 uint32_t idx = (q) / (sizeof ((h)->bitmap[0]) * NBBY); \
356                 uint32_t bit = (q) % (sizeof ((h)->bitmap[0]) * NBBY); \
357                 (h)->bitmap[idx] &= ~(1 << bit);\
358         }while(0)
359
360 #define IXGBE_GET_HWSTRIP(h, q, r) do{\
361                 uint32_t idx = (q) / (sizeof ((h)->bitmap[0]) * NBBY); \
362                 uint32_t bit = (q) % (sizeof ((h)->bitmap[0]) * NBBY); \
363                 (r) = (h)->bitmap[idx] >> bit & 1;\
364         }while(0)
365
366 /*
367  * The set of PCI devices this driver supports
368  */
369 static const struct rte_pci_id pci_id_ixgbe_map[] = {
370
371 #define RTE_PCI_DEV_ID_DECL_IXGBE(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
372 #include "rte_pci_dev_ids.h"
373
374 { .vendor_id = 0, /* sentinel */ },
375 };
376
377
378 /*
379  * The set of PCI devices this driver supports (for 82599 VF)
380  */
381 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
382
383 #define RTE_PCI_DEV_ID_DECL_IXGBEVF(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
384 #include "rte_pci_dev_ids.h"
385 { .vendor_id = 0, /* sentinel */ },
386
387 };
388
389 static const struct eth_dev_ops ixgbe_eth_dev_ops = {
390         .dev_configure        = ixgbe_dev_configure,
391         .dev_start            = ixgbe_dev_start,
392         .dev_stop             = ixgbe_dev_stop,
393         .dev_set_link_up    = ixgbe_dev_set_link_up,
394         .dev_set_link_down  = ixgbe_dev_set_link_down,
395         .dev_close            = ixgbe_dev_close,
396         .promiscuous_enable   = ixgbe_dev_promiscuous_enable,
397         .promiscuous_disable  = ixgbe_dev_promiscuous_disable,
398         .allmulticast_enable  = ixgbe_dev_allmulticast_enable,
399         .allmulticast_disable = ixgbe_dev_allmulticast_disable,
400         .link_update          = ixgbe_dev_link_update,
401         .stats_get            = ixgbe_dev_stats_get,
402         .xstats_get           = ixgbe_dev_xstats_get,
403         .stats_reset          = ixgbe_dev_stats_reset,
404         .xstats_reset         = ixgbe_dev_xstats_reset,
405         .queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
406         .dev_infos_get        = ixgbe_dev_info_get,
407         .mtu_set              = ixgbe_dev_mtu_set,
408         .vlan_filter_set      = ixgbe_vlan_filter_set,
409         .vlan_tpid_set        = ixgbe_vlan_tpid_set,
410         .vlan_offload_set     = ixgbe_vlan_offload_set,
411         .vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
412         .rx_queue_start       = ixgbe_dev_rx_queue_start,
413         .rx_queue_stop        = ixgbe_dev_rx_queue_stop,
414         .tx_queue_start       = ixgbe_dev_tx_queue_start,
415         .tx_queue_stop        = ixgbe_dev_tx_queue_stop,
416         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
417         .rx_queue_intr_enable = ixgbe_dev_rx_queue_intr_enable,
418         .rx_queue_intr_disable = ixgbe_dev_rx_queue_intr_disable,
419         .rx_queue_release     = ixgbe_dev_rx_queue_release,
420         .rx_queue_count       = ixgbe_dev_rx_queue_count,
421         .rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
422         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
423         .tx_queue_release     = ixgbe_dev_tx_queue_release,
424         .dev_led_on           = ixgbe_dev_led_on,
425         .dev_led_off          = ixgbe_dev_led_off,
426         .flow_ctrl_get        = ixgbe_flow_ctrl_get,
427         .flow_ctrl_set        = ixgbe_flow_ctrl_set,
428         .priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
429         .mac_addr_add         = ixgbe_add_rar,
430         .mac_addr_remove      = ixgbe_remove_rar,
431         .mac_addr_set         = ixgbe_set_default_mac_addr,
432         .uc_hash_table_set    = ixgbe_uc_hash_table_set,
433         .uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
434         .mirror_rule_set      = ixgbe_mirror_rule_set,
435         .mirror_rule_reset    = ixgbe_mirror_rule_reset,
436         .set_vf_rx_mode       = ixgbe_set_pool_rx_mode,
437         .set_vf_rx            = ixgbe_set_pool_rx,
438         .set_vf_tx            = ixgbe_set_pool_tx,
439         .set_vf_vlan_filter   = ixgbe_set_pool_vlan_filter,
440         .set_queue_rate_limit = ixgbe_set_queue_rate_limit,
441         .set_vf_rate_limit    = ixgbe_set_vf_rate_limit,
442         .reta_update          = ixgbe_dev_rss_reta_update,
443         .reta_query           = ixgbe_dev_rss_reta_query,
444 #ifdef RTE_NIC_BYPASS
445         .bypass_init          = ixgbe_bypass_init,
446         .bypass_state_set     = ixgbe_bypass_state_store,
447         .bypass_state_show    = ixgbe_bypass_state_show,
448         .bypass_event_set     = ixgbe_bypass_event_store,
449         .bypass_event_show    = ixgbe_bypass_event_show,
450         .bypass_wd_timeout_set  = ixgbe_bypass_wd_timeout_store,
451         .bypass_wd_timeout_show = ixgbe_bypass_wd_timeout_show,
452         .bypass_ver_show      = ixgbe_bypass_ver_show,
453         .bypass_wd_reset      = ixgbe_bypass_wd_reset,
454 #endif /* RTE_NIC_BYPASS */
455         .rss_hash_update      = ixgbe_dev_rss_hash_update,
456         .rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
457         .filter_ctrl          = ixgbe_dev_filter_ctrl,
458         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
459         .timesync_enable      = ixgbe_timesync_enable,
460         .timesync_disable     = ixgbe_timesync_disable,
461         .timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
462         .timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
463         .get_reg_length       = ixgbe_get_reg_length,
464         .get_reg              = ixgbe_get_regs,
465         .get_eeprom_length    = ixgbe_get_eeprom_length,
466         .get_eeprom           = ixgbe_get_eeprom,
467         .set_eeprom           = ixgbe_set_eeprom,
468 };
469
470 /*
471  * dev_ops for virtual function, bare necessities for basic vf
472  * operation have been implemented
473  */
474 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
475         .dev_configure        = ixgbevf_dev_configure,
476         .dev_start            = ixgbevf_dev_start,
477         .dev_stop             = ixgbevf_dev_stop,
478         .link_update          = ixgbe_dev_link_update,
479         .stats_get            = ixgbevf_dev_stats_get,
480         .stats_reset          = ixgbevf_dev_stats_reset,
481         .dev_close            = ixgbevf_dev_close,
482         .dev_infos_get        = ixgbevf_dev_info_get,
483         .mtu_set              = ixgbevf_dev_set_mtu,
484         .vlan_filter_set      = ixgbevf_vlan_filter_set,
485         .vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
486         .vlan_offload_set     = ixgbevf_vlan_offload_set,
487         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
488         .rx_queue_release     = ixgbe_dev_rx_queue_release,
489         .rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
490         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
491         .tx_queue_release     = ixgbe_dev_tx_queue_release,
492         .rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
493         .rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
494         .mac_addr_add         = ixgbevf_add_mac_addr,
495         .mac_addr_remove      = ixgbevf_remove_mac_addr,
496         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
497         .mac_addr_set         = ixgbevf_set_default_mac_addr,
498         .get_reg_length       = ixgbevf_get_reg_length,
499         .get_reg              = ixgbevf_get_regs,
500 };
501
502 /* store statistics names and its offset in stats structure */
503 struct rte_ixgbe_xstats_name_off {
504         char name[RTE_ETH_XSTATS_NAME_SIZE];
505         unsigned offset;
506 };
507
508 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
509         {"rx_illegal_byte_err", offsetof(struct ixgbe_hw_stats, errbc)},
510         {"rx_len_err", offsetof(struct ixgbe_hw_stats, rlec)},
511         {"rx_undersize_count", offsetof(struct ixgbe_hw_stats, ruc)},
512         {"rx_oversize_count", offsetof(struct ixgbe_hw_stats, roc)},
513         {"rx_fragment_count", offsetof(struct ixgbe_hw_stats, rfc)},
514         {"rx_jabber_count", offsetof(struct ixgbe_hw_stats, rjc)},
515         {"l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
516         {"mac_local_fault", offsetof(struct ixgbe_hw_stats, mlfc)},
517         {"mac_remote_fault", offsetof(struct ixgbe_hw_stats, mrfc)},
518         {"mac_short_pkt_discard", offsetof(struct ixgbe_hw_stats, mspdc)},
519         {"fccrc_error", offsetof(struct ixgbe_hw_stats, fccrc)},
520         {"fcoe_drop", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
521         {"fc_last_error", offsetof(struct ixgbe_hw_stats, fclast)},
522         {"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
523         {"rx_phy_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
524         {"mgmt_pkts_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
525         {"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
526         {"fdir_match", offsetof(struct ixgbe_hw_stats, fdirmatch)},
527         {"fdir_miss", offsetof(struct ixgbe_hw_stats, fdirmiss)},
528         {"tx_flow_control_xon", offsetof(struct ixgbe_hw_stats, lxontxc)},
529         {"rx_flow_control_xon", offsetof(struct ixgbe_hw_stats, lxonrxc)},
530         {"tx_flow_control_xoff", offsetof(struct ixgbe_hw_stats, lxofftxc)},
531         {"rx_flow_control_xoff", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
532 };
533
534 #define IXGBE_NB_XSTATS (sizeof(rte_ixgbe_stats_strings) /      \
535                 sizeof(rte_ixgbe_stats_strings[0]))
536
537 /**
538  * Atomically reads the link status information from global
539  * structure rte_eth_dev.
540  *
541  * @param dev
542  *   - Pointer to the structure rte_eth_dev to read from.
543  *   - Pointer to the buffer to be saved with the link status.
544  *
545  * @return
546  *   - On success, zero.
547  *   - On failure, negative value.
548  */
549 static inline int
550 rte_ixgbe_dev_atomic_read_link_status(struct rte_eth_dev *dev,
551                                 struct rte_eth_link *link)
552 {
553         struct rte_eth_link *dst = link;
554         struct rte_eth_link *src = &(dev->data->dev_link);
555
556         if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
557                                         *(uint64_t *)src) == 0)
558                 return -1;
559
560         return 0;
561 }
562
563 /**
564  * Atomically writes the link status information into global
565  * structure rte_eth_dev.
566  *
567  * @param dev
568  *   - Pointer to the structure rte_eth_dev to read from.
569  *   - Pointer to the buffer to be saved with the link status.
570  *
571  * @return
572  *   - On success, zero.
573  *   - On failure, negative value.
574  */
575 static inline int
576 rte_ixgbe_dev_atomic_write_link_status(struct rte_eth_dev *dev,
577                                 struct rte_eth_link *link)
578 {
579         struct rte_eth_link *dst = &(dev->data->dev_link);
580         struct rte_eth_link *src = link;
581
582         if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
583                                         *(uint64_t *)src) == 0)
584                 return -1;
585
586         return 0;
587 }
588
589 /*
590  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
591  */
592 static inline int
593 ixgbe_is_sfp(struct ixgbe_hw *hw)
594 {
595         switch (hw->phy.type) {
596         case ixgbe_phy_sfp_avago:
597         case ixgbe_phy_sfp_ftl:
598         case ixgbe_phy_sfp_intel:
599         case ixgbe_phy_sfp_unknown:
600         case ixgbe_phy_sfp_passive_tyco:
601         case ixgbe_phy_sfp_passive_unknown:
602                 return 1;
603         default:
604                 return 0;
605         }
606 }
607
608 static inline int32_t
609 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
610 {
611         uint32_t ctrl_ext;
612         int32_t status;
613
614         status = ixgbe_reset_hw(hw);
615
616         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
617         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
618         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
619         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
620         IXGBE_WRITE_FLUSH(hw);
621
622         return status;
623 }
624
625 static inline void
626 ixgbe_enable_intr(struct rte_eth_dev *dev)
627 {
628         struct ixgbe_interrupt *intr =
629                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
630         struct ixgbe_hw *hw =
631                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
632
633         IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
634         IXGBE_WRITE_FLUSH(hw);
635 }
636
637 /*
638  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
639  */
640 static void
641 ixgbe_disable_intr(struct ixgbe_hw *hw)
642 {
643         PMD_INIT_FUNC_TRACE();
644
645         if (hw->mac.type == ixgbe_mac_82598EB) {
646                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
647         } else {
648                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
649                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
650                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
651         }
652         IXGBE_WRITE_FLUSH(hw);
653 }
654
655 /*
656  * This function resets queue statistics mapping registers.
657  * From Niantic datasheet, Initialization of Statistics section:
658  * "...if software requires the queue counters, the RQSMR and TQSM registers
659  * must be re-programmed following a device reset.
660  */
661 static void
662 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
663 {
664         uint32_t i;
665
666         for(i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
667                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
668                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
669         }
670 }
671
672
673 static int
674 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
675                                   uint16_t queue_id,
676                                   uint8_t stat_idx,
677                                   uint8_t is_rx)
678 {
679 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
680 #define NB_QMAP_FIELDS_PER_QSM_REG 4
681 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
682
683         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
684         struct ixgbe_stat_mapping_registers *stat_mappings =
685                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
686         uint32_t qsmr_mask = 0;
687         uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
688         uint32_t q_map;
689         uint8_t n, offset;
690
691         if ((hw->mac.type != ixgbe_mac_82599EB) &&
692                 (hw->mac.type != ixgbe_mac_X540) &&
693                 (hw->mac.type != ixgbe_mac_X550) &&
694                 (hw->mac.type != ixgbe_mac_X550EM_x))
695                 return -ENOSYS;
696
697         PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
698                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
699                      queue_id, stat_idx);
700
701         n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
702         if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
703                 PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
704                 return -EIO;
705         }
706         offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
707
708         /* Now clear any previous stat_idx set */
709         clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
710         if (!is_rx)
711                 stat_mappings->tqsm[n] &= ~clearing_mask;
712         else
713                 stat_mappings->rqsmr[n] &= ~clearing_mask;
714
715         q_map = (uint32_t)stat_idx;
716         q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
717         qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
718         if (!is_rx)
719                 stat_mappings->tqsm[n] |= qsmr_mask;
720         else
721                 stat_mappings->rqsmr[n] |= qsmr_mask;
722
723         PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
724                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
725                      queue_id, stat_idx);
726         PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
727                      is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
728
729         /* Now write the mapping in the appropriate register */
730         if (is_rx) {
731                 PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
732                              stat_mappings->rqsmr[n], n);
733                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
734         }
735         else {
736                 PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
737                              stat_mappings->tqsm[n], n);
738                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
739         }
740         return 0;
741 }
742
743 static void
744 ixgbe_restore_statistics_mapping(struct rte_eth_dev * dev)
745 {
746         struct ixgbe_stat_mapping_registers *stat_mappings =
747                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
748         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
749         int i;
750
751         /* write whatever was in stat mapping table to the NIC */
752         for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
753                 /* rx */
754                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
755
756                 /* tx */
757                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
758         }
759 }
760
761 static void
762 ixgbe_dcb_init(struct ixgbe_hw *hw,struct ixgbe_dcb_config *dcb_config)
763 {
764         uint8_t i;
765         struct ixgbe_dcb_tc_config *tc;
766         uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
767
768         dcb_config->num_tcs.pg_tcs = dcb_max_tc;
769         dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
770         for (i = 0; i < dcb_max_tc; i++) {
771                 tc = &dcb_config->tc_config[i];
772                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
773                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
774                                  (uint8_t)(100/dcb_max_tc + (i & 1));
775                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
776                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
777                                  (uint8_t)(100/dcb_max_tc + (i & 1));
778                 tc->pfc = ixgbe_dcb_pfc_disabled;
779         }
780
781         /* Initialize default user to priority mapping, UPx->TC0 */
782         tc = &dcb_config->tc_config[0];
783         tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
784         tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
785         for (i = 0; i< IXGBE_DCB_MAX_BW_GROUP; i++) {
786                 dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
787                 dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
788         }
789         dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
790         dcb_config->pfc_mode_enable = false;
791         dcb_config->vt_mode = true;
792         dcb_config->round_robin_enable = false;
793         /* support all DCB capabilities in 82599 */
794         dcb_config->support.capabilities = 0xFF;
795
796         /*we only support 4 Tcs for X540, X550 */
797         if (hw->mac.type == ixgbe_mac_X540 ||
798                 hw->mac.type == ixgbe_mac_X550 ||
799                 hw->mac.type == ixgbe_mac_X550EM_x) {
800                 dcb_config->num_tcs.pg_tcs = 4;
801                 dcb_config->num_tcs.pfc_tcs = 4;
802         }
803 }
804
805 /*
806  * Ensure that all locks are released before first NVM or PHY access
807  */
808 static void
809 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
810 {
811         uint16_t mask;
812
813         /*
814          * Phy lock should not fail in this early stage. If this is the case,
815          * it is due to an improper exit of the application.
816          * So force the release of the faulty lock. Release of common lock
817          * is done automatically by swfw_sync function.
818          */
819         mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
820         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
821                 PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
822         }
823         ixgbe_release_swfw_semaphore(hw, mask);
824
825         /*
826          * These ones are more tricky since they are common to all ports; but
827          * swfw_sync retries last long enough (1s) to be almost sure that if
828          * lock can not be taken it is due to an improper lock of the
829          * semaphore.
830          */
831         mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
832         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
833                 PMD_DRV_LOG(DEBUG, "SWFW common locks released");
834         }
835         ixgbe_release_swfw_semaphore(hw, mask);
836 }
837
838 /*
839  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
840  * It returns 0 on success.
841  */
842 static int
843 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev)
844 {
845         struct rte_pci_device *pci_dev;
846         struct ixgbe_hw *hw =
847                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
848         struct ixgbe_vfta * shadow_vfta =
849                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
850         struct ixgbe_hwstrip *hwstrip =
851                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
852         struct ixgbe_dcb_config *dcb_config =
853                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
854         struct ixgbe_filter_info *filter_info =
855                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
856         uint32_t ctrl_ext;
857         uint16_t csum;
858         int diag, i;
859
860         PMD_INIT_FUNC_TRACE();
861
862         eth_dev->dev_ops = &ixgbe_eth_dev_ops;
863         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
864         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
865
866         /*
867          * For secondary processes, we don't initialise any further as primary
868          * has already done this work. Only check we don't need a different
869          * RX and TX function.
870          */
871         if (rte_eal_process_type() != RTE_PROC_PRIMARY){
872                 struct ixgbe_tx_queue *txq;
873                 /* TX queue function in primary, set by last queue initialized
874                  * Tx queue may not initialized by primary process */
875                 if (eth_dev->data->tx_queues) {
876                         txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
877                         ixgbe_set_tx_function(eth_dev, txq);
878                 } else {
879                         /* Use default TX function if we get here */
880                         PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
881                                              "Using default TX function.");
882                 }
883
884                 ixgbe_set_rx_function(eth_dev);
885
886                 return 0;
887         }
888         pci_dev = eth_dev->pci_dev;
889
890         /* Vendor and Device ID need to be set before init of shared code */
891         hw->device_id = pci_dev->id.device_id;
892         hw->vendor_id = pci_dev->id.vendor_id;
893         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
894         hw->allow_unsupported_sfp = 1;
895
896         /* Initialize the shared code (base driver) */
897 #ifdef RTE_NIC_BYPASS
898         diag = ixgbe_bypass_init_shared_code(hw);
899 #else
900         diag = ixgbe_init_shared_code(hw);
901 #endif /* RTE_NIC_BYPASS */
902
903         if (diag != IXGBE_SUCCESS) {
904                 PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
905                 return -EIO;
906         }
907
908         /* pick up the PCI bus settings for reporting later */
909         ixgbe_get_bus_info(hw);
910
911         /* Unlock any pending hardware semaphore */
912         ixgbe_swfw_lock_reset(hw);
913
914         /* Initialize DCB configuration*/
915         memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
916         ixgbe_dcb_init(hw,dcb_config);
917         /* Get Hardware Flow Control setting */
918         hw->fc.requested_mode = ixgbe_fc_full;
919         hw->fc.current_mode = ixgbe_fc_full;
920         hw->fc.pause_time = IXGBE_FC_PAUSE;
921         for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
922                 hw->fc.low_water[i] = IXGBE_FC_LO;
923                 hw->fc.high_water[i] = IXGBE_FC_HI;
924         }
925         hw->fc.send_xon = 1;
926
927         /* Make sure we have a good EEPROM before we read from it */
928         diag = ixgbe_validate_eeprom_checksum(hw, &csum);
929         if (diag != IXGBE_SUCCESS) {
930                 PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
931                 return -EIO;
932         }
933
934 #ifdef RTE_NIC_BYPASS
935         diag = ixgbe_bypass_init_hw(hw);
936 #else
937         diag = ixgbe_init_hw(hw);
938 #endif /* RTE_NIC_BYPASS */
939
940         /*
941          * Devices with copper phys will fail to initialise if ixgbe_init_hw()
942          * is called too soon after the kernel driver unbinding/binding occurs.
943          * The failure occurs in ixgbe_identify_phy_generic() for all devices,
944          * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
945          * also called. See ixgbe_identify_phy_82599(). The reason for the
946          * failure is not known, and only occuts when virtualisation features
947          * are disabled in the bios. A delay of 100ms  was found to be enough by
948          * trial-and-error, and is doubled to be safe.
949          */
950         if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
951                 rte_delay_ms(200);
952                 diag = ixgbe_init_hw(hw);
953         }
954
955         if (diag == IXGBE_ERR_EEPROM_VERSION) {
956                 PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
957                     "LOM.  Please be aware there may be issues associated "
958                     "with your hardware.");
959                 PMD_INIT_LOG(ERR, "If you are experiencing problems "
960                     "please contact your Intel or hardware representative "
961                     "who provided you with this hardware.");
962         } else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
963                 PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
964         if (diag) {
965                 PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
966                 return -EIO;
967         }
968
969         /* Reset the hw statistics */
970         ixgbe_dev_stats_reset(eth_dev);
971
972         /* disable interrupt */
973         ixgbe_disable_intr(hw);
974
975         /* reset mappings for queue statistics hw counters*/
976         ixgbe_reset_qstat_mappings(hw);
977
978         /* Allocate memory for storing MAC addresses */
979         eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
980                         hw->mac.num_rar_entries, 0);
981         if (eth_dev->data->mac_addrs == NULL) {
982                 PMD_INIT_LOG(ERR,
983                         "Failed to allocate %u bytes needed to store "
984                         "MAC addresses",
985                         ETHER_ADDR_LEN * hw->mac.num_rar_entries);
986                 return -ENOMEM;
987         }
988         /* Copy the permanent MAC address */
989         ether_addr_copy((struct ether_addr *) hw->mac.perm_addr,
990                         &eth_dev->data->mac_addrs[0]);
991
992         /* Allocate memory for storing hash filter MAC addresses */
993         eth_dev->data->hash_mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
994                         IXGBE_VMDQ_NUM_UC_MAC, 0);
995         if (eth_dev->data->hash_mac_addrs == NULL) {
996                 PMD_INIT_LOG(ERR,
997                         "Failed to allocate %d bytes needed to store MAC addresses",
998                         ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
999                 return -ENOMEM;
1000         }
1001
1002         /* initialize the vfta */
1003         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1004
1005         /* initialize the hw strip bitmap*/
1006         memset(hwstrip, 0, sizeof(*hwstrip));
1007
1008         /* initialize PF if max_vfs not zero */
1009         ixgbe_pf_host_init(eth_dev);
1010
1011         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1012         /* let hardware know driver is loaded */
1013         ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1014         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
1015         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1016         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1017         IXGBE_WRITE_FLUSH(hw);
1018
1019         if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1020                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1021                              (int) hw->mac.type, (int) hw->phy.type,
1022                              (int) hw->phy.sfp_type);
1023         else
1024                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1025                              (int) hw->mac.type, (int) hw->phy.type);
1026
1027         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1028                         eth_dev->data->port_id, pci_dev->id.vendor_id,
1029                         pci_dev->id.device_id);
1030
1031         /* enable support intr */
1032         ixgbe_enable_intr(eth_dev);
1033
1034         /* initialize 5tuple filter list */
1035         TAILQ_INIT(&filter_info->fivetuple_list);
1036         memset(filter_info->fivetuple_mask, 0,
1037                 sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1038
1039         return 0;
1040 }
1041
1042 static int
1043 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1044 {
1045         struct rte_pci_device *pci_dev;
1046         struct ixgbe_hw *hw;
1047
1048         PMD_INIT_FUNC_TRACE();
1049
1050         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1051                 return -EPERM;
1052
1053         hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1054         pci_dev = eth_dev->pci_dev;
1055
1056         if (hw->adapter_stopped == 0)
1057                 ixgbe_dev_close(eth_dev);
1058
1059         eth_dev->dev_ops = NULL;
1060         eth_dev->rx_pkt_burst = NULL;
1061         eth_dev->tx_pkt_burst = NULL;
1062
1063         /* Unlock any pending hardware semaphore */
1064         ixgbe_swfw_lock_reset(hw);
1065
1066         /* disable uio intr before callback unregister */
1067         rte_intr_disable(&(pci_dev->intr_handle));
1068         rte_intr_callback_unregister(&(pci_dev->intr_handle),
1069                 ixgbe_dev_interrupt_handler, (void *)eth_dev);
1070
1071         /* uninitialize PF if max_vfs not zero */
1072         ixgbe_pf_host_uninit(eth_dev);
1073
1074         rte_free(eth_dev->data->mac_addrs);
1075         eth_dev->data->mac_addrs = NULL;
1076
1077         rte_free(eth_dev->data->hash_mac_addrs);
1078         eth_dev->data->hash_mac_addrs = NULL;
1079
1080         return 0;
1081 }
1082
1083 /*
1084  * Negotiate mailbox API version with the PF.
1085  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1086  * Then we try to negotiate starting with the most recent one.
1087  * If all negotiation attempts fail, then we will proceed with
1088  * the default one (ixgbe_mbox_api_10).
1089  */
1090 static void
1091 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1092 {
1093         int32_t i;
1094
1095         /* start with highest supported, proceed down */
1096         static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1097                 ixgbe_mbox_api_11,
1098                 ixgbe_mbox_api_10,
1099         };
1100
1101         for (i = 0;
1102                         i != RTE_DIM(sup_ver) &&
1103                         ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1104                         i++)
1105                 ;
1106 }
1107
1108 static void
1109 generate_random_mac_addr(struct ether_addr *mac_addr)
1110 {
1111         uint64_t random;
1112
1113         /* Set Organizationally Unique Identifier (OUI) prefix. */
1114         mac_addr->addr_bytes[0] = 0x00;
1115         mac_addr->addr_bytes[1] = 0x09;
1116         mac_addr->addr_bytes[2] = 0xC0;
1117         /* Force indication of locally assigned MAC address. */
1118         mac_addr->addr_bytes[0] |= ETHER_LOCAL_ADMIN_ADDR;
1119         /* Generate the last 3 bytes of the MAC address with a random number. */
1120         random = rte_rand();
1121         memcpy(&mac_addr->addr_bytes[3], &random, 3);
1122 }
1123
1124 /*
1125  * Virtual Function device init
1126  */
1127 static int
1128 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1129 {
1130         int diag;
1131         uint32_t tc, tcs;
1132         struct rte_pci_device *pci_dev;
1133         struct ixgbe_hw *hw =
1134                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1135         struct ixgbe_vfta * shadow_vfta =
1136                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1137         struct ixgbe_hwstrip *hwstrip =
1138                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1139         struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
1140
1141         PMD_INIT_FUNC_TRACE();
1142
1143         eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1144         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1145         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1146
1147         /* for secondary processes, we don't initialise any further as primary
1148          * has already done this work. Only check we don't need a different
1149          * RX function */
1150         if (rte_eal_process_type() != RTE_PROC_PRIMARY){
1151                 if (eth_dev->data->scattered_rx)
1152                         eth_dev->rx_pkt_burst = ixgbe_recv_pkts_lro_single_alloc;
1153                 return 0;
1154         }
1155
1156         pci_dev = eth_dev->pci_dev;
1157
1158         hw->device_id = pci_dev->id.device_id;
1159         hw->vendor_id = pci_dev->id.vendor_id;
1160         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1161
1162         /* initialize the vfta */
1163         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1164
1165         /* initialize the hw strip bitmap*/
1166         memset(hwstrip, 0, sizeof(*hwstrip));
1167
1168         /* Initialize the shared code (base driver) */
1169         diag = ixgbe_init_shared_code(hw);
1170         if (diag != IXGBE_SUCCESS) {
1171                 PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1172                 return -EIO;
1173         }
1174
1175         /* init_mailbox_params */
1176         hw->mbx.ops.init_params(hw);
1177
1178         /* Reset the hw statistics */
1179         ixgbevf_dev_stats_reset(eth_dev);
1180
1181         /* Disable the interrupts for VF */
1182         ixgbevf_intr_disable(hw);
1183
1184         hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1185         diag = hw->mac.ops.reset_hw(hw);
1186
1187         /*
1188          * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1189          * the underlying PF driver has not assigned a MAC address to the VF.
1190          * In this case, assign a random MAC address.
1191          */
1192         if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1193                 PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1194                 return (diag);
1195         }
1196
1197         /* negotiate mailbox API version to use with the PF. */
1198         ixgbevf_negotiate_api(hw);
1199
1200         /* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1201         ixgbevf_get_queues(hw, &tcs, &tc);
1202
1203         /* Allocate memory for storing MAC addresses */
1204         eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", ETHER_ADDR_LEN *
1205                         hw->mac.num_rar_entries, 0);
1206         if (eth_dev->data->mac_addrs == NULL) {
1207                 PMD_INIT_LOG(ERR,
1208                         "Failed to allocate %u bytes needed to store "
1209                         "MAC addresses",
1210                         ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1211                 return -ENOMEM;
1212         }
1213
1214         /* Generate a random MAC address, if none was assigned by PF. */
1215         if (is_zero_ether_addr(perm_addr)) {
1216                 generate_random_mac_addr(perm_addr);
1217                 diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1218                 if (diag) {
1219                         rte_free(eth_dev->data->mac_addrs);
1220                         eth_dev->data->mac_addrs = NULL;
1221                         return diag;
1222                 }
1223                 PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1224                 PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1225                              "%02x:%02x:%02x:%02x:%02x:%02x",
1226                              perm_addr->addr_bytes[0],
1227                              perm_addr->addr_bytes[1],
1228                              perm_addr->addr_bytes[2],
1229                              perm_addr->addr_bytes[3],
1230                              perm_addr->addr_bytes[4],
1231                              perm_addr->addr_bytes[5]);
1232         }
1233
1234         /* Copy the permanent MAC address */
1235         ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1236
1237         /* reset the hardware with the new settings */
1238         diag = hw->mac.ops.start_hw(hw);
1239         switch (diag) {
1240                 case  0:
1241                         break;
1242
1243                 default:
1244                         PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1245                         return (-EIO);
1246         }
1247
1248         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1249                      eth_dev->data->port_id, pci_dev->id.vendor_id,
1250                      pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1251
1252         return 0;
1253 }
1254
1255 /* Virtual Function device uninit */
1256
1257 static int
1258 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1259 {
1260         struct ixgbe_hw *hw;
1261         unsigned i;
1262
1263         PMD_INIT_FUNC_TRACE();
1264
1265         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1266                 return -EPERM;
1267
1268         hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1269
1270         if (hw->adapter_stopped == 0)
1271                 ixgbevf_dev_close(eth_dev);
1272
1273         eth_dev->dev_ops = NULL;
1274         eth_dev->rx_pkt_burst = NULL;
1275         eth_dev->tx_pkt_burst = NULL;
1276
1277         /* Disable the interrupts for VF */
1278         ixgbevf_intr_disable(hw);
1279
1280         for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1281                 ixgbe_dev_rx_queue_release(eth_dev->data->rx_queues[i]);
1282                 eth_dev->data->rx_queues[i] = NULL;
1283         }
1284         eth_dev->data->nb_rx_queues = 0;
1285
1286         for (i = 0; i < eth_dev->data->nb_tx_queues; i++) {
1287                 ixgbe_dev_tx_queue_release(eth_dev->data->tx_queues[i]);
1288                 eth_dev->data->tx_queues[i] = NULL;
1289         }
1290         eth_dev->data->nb_tx_queues = 0;
1291
1292         rte_free(eth_dev->data->mac_addrs);
1293         eth_dev->data->mac_addrs = NULL;
1294
1295         return 0;
1296 }
1297
1298 static struct eth_driver rte_ixgbe_pmd = {
1299         .pci_drv = {
1300                 .name = "rte_ixgbe_pmd",
1301                 .id_table = pci_id_ixgbe_map,
1302                 .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
1303                         RTE_PCI_DRV_DETACHABLE,
1304         },
1305         .eth_dev_init = eth_ixgbe_dev_init,
1306         .eth_dev_uninit = eth_ixgbe_dev_uninit,
1307         .dev_private_size = sizeof(struct ixgbe_adapter),
1308 };
1309
1310 /*
1311  * virtual function driver struct
1312  */
1313 static struct eth_driver rte_ixgbevf_pmd = {
1314         .pci_drv = {
1315                 .name = "rte_ixgbevf_pmd",
1316                 .id_table = pci_id_ixgbevf_map,
1317                 .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_DETACHABLE,
1318         },
1319         .eth_dev_init = eth_ixgbevf_dev_init,
1320         .eth_dev_uninit = eth_ixgbevf_dev_uninit,
1321         .dev_private_size = sizeof(struct ixgbe_adapter),
1322 };
1323
1324 /*
1325  * Driver initialization routine.
1326  * Invoked once at EAL init time.
1327  * Register itself as the [Poll Mode] Driver of PCI IXGBE devices.
1328  */
1329 static int
1330 rte_ixgbe_pmd_init(const char *name __rte_unused, const char *params __rte_unused)
1331 {
1332         PMD_INIT_FUNC_TRACE();
1333
1334         rte_eth_driver_register(&rte_ixgbe_pmd);
1335         return 0;
1336 }
1337
1338 /*
1339  * VF Driver initialization routine.
1340  * Invoked one at EAL init time.
1341  * Register itself as the [Virtual Poll Mode] Driver of PCI niantic devices.
1342  */
1343 static int
1344 rte_ixgbevf_pmd_init(const char *name __rte_unused, const char *param __rte_unused)
1345 {
1346         PMD_INIT_FUNC_TRACE();
1347
1348         rte_eth_driver_register(&rte_ixgbevf_pmd);
1349         return (0);
1350 }
1351
1352 static int
1353 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1354 {
1355         struct ixgbe_hw *hw =
1356                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1357         struct ixgbe_vfta * shadow_vfta =
1358                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1359         uint32_t vfta;
1360         uint32_t vid_idx;
1361         uint32_t vid_bit;
1362
1363         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1364         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1365         vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1366         if (on)
1367                 vfta |= vid_bit;
1368         else
1369                 vfta &= ~vid_bit;
1370         IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1371
1372         /* update local VFTA copy */
1373         shadow_vfta->vfta[vid_idx] = vfta;
1374
1375         return 0;
1376 }
1377
1378 static void
1379 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1380 {
1381         if (on)
1382                 ixgbe_vlan_hw_strip_enable(dev, queue);
1383         else
1384                 ixgbe_vlan_hw_strip_disable(dev, queue);
1385 }
1386
1387 static void
1388 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid)
1389 {
1390         struct ixgbe_hw *hw =
1391                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1392
1393         /* Only the high 16-bits is valid */
1394         IXGBE_WRITE_REG(hw, IXGBE_EXVET, tpid << 16);
1395 }
1396
1397 void
1398 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1399 {
1400         struct ixgbe_hw *hw =
1401                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1402         uint32_t vlnctrl;
1403
1404         PMD_INIT_FUNC_TRACE();
1405
1406         /* Filter Table Disable */
1407         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1408         vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1409
1410         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1411 }
1412
1413 void
1414 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1415 {
1416         struct ixgbe_hw *hw =
1417                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1418         struct ixgbe_vfta * shadow_vfta =
1419                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1420         uint32_t vlnctrl;
1421         uint16_t i;
1422
1423         PMD_INIT_FUNC_TRACE();
1424
1425         /* Filter Table Enable */
1426         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1427         vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1428         vlnctrl |= IXGBE_VLNCTRL_VFE;
1429
1430         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1431
1432         /* write whatever is in local vfta copy */
1433         for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1434                 IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1435 }
1436
1437 static void
1438 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1439 {
1440         struct ixgbe_hwstrip *hwstrip =
1441                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1442
1443         if(queue >= IXGBE_MAX_RX_QUEUE_NUM)
1444                 return;
1445
1446         if (on)
1447                 IXGBE_SET_HWSTRIP(hwstrip, queue);
1448         else
1449                 IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1450 }
1451
1452 static void
1453 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
1454 {
1455         struct ixgbe_hw *hw =
1456                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1457         uint32_t ctrl;
1458
1459         PMD_INIT_FUNC_TRACE();
1460
1461         if (hw->mac.type == ixgbe_mac_82598EB) {
1462                 /* No queue level support */
1463                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1464                 return;
1465         }
1466         else {
1467                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1468                 ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1469                 ctrl &= ~IXGBE_RXDCTL_VME;
1470                 IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1471         }
1472         /* record those setting for HW strip per queue */
1473         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
1474 }
1475
1476 static void
1477 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
1478 {
1479         struct ixgbe_hw *hw =
1480                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1481         uint32_t ctrl;
1482
1483         PMD_INIT_FUNC_TRACE();
1484
1485         if (hw->mac.type == ixgbe_mac_82598EB) {
1486                 /* No queue level supported */
1487                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1488                 return;
1489         }
1490         else {
1491                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1492                 ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1493                 ctrl |= IXGBE_RXDCTL_VME;
1494                 IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1495         }
1496         /* record those setting for HW strip per queue */
1497         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
1498 }
1499
1500 void
1501 ixgbe_vlan_hw_strip_disable_all(struct rte_eth_dev *dev)
1502 {
1503         struct ixgbe_hw *hw =
1504                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1505         uint32_t ctrl;
1506         uint16_t i;
1507
1508         PMD_INIT_FUNC_TRACE();
1509
1510         if (hw->mac.type == ixgbe_mac_82598EB) {
1511                 ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1512                 ctrl &= ~IXGBE_VLNCTRL_VME;
1513                 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1514         }
1515         else {
1516                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1517                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
1518                         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
1519                         ctrl &= ~IXGBE_RXDCTL_VME;
1520                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(i), ctrl);
1521
1522                         /* record those setting for HW strip per queue */
1523                         ixgbe_vlan_hw_strip_bitmap_set(dev, i, 0);
1524                 }
1525         }
1526 }
1527
1528 void
1529 ixgbe_vlan_hw_strip_enable_all(struct rte_eth_dev *dev)
1530 {
1531         struct ixgbe_hw *hw =
1532                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1533         uint32_t ctrl;
1534         uint16_t i;
1535
1536         PMD_INIT_FUNC_TRACE();
1537
1538         if (hw->mac.type == ixgbe_mac_82598EB) {
1539                 ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1540                 ctrl |= IXGBE_VLNCTRL_VME;
1541                 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1542         }
1543         else {
1544                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1545                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
1546                         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
1547                         ctrl |= IXGBE_RXDCTL_VME;
1548                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(i), ctrl);
1549
1550                         /* record those setting for HW strip per queue */
1551                         ixgbe_vlan_hw_strip_bitmap_set(dev, i, 1);
1552                 }
1553         }
1554 }
1555
1556 static void
1557 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
1558 {
1559         struct ixgbe_hw *hw =
1560                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1561         uint32_t ctrl;
1562
1563         PMD_INIT_FUNC_TRACE();
1564
1565         /* DMATXCTRL: Geric Double VLAN Disable */
1566         ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1567         ctrl &= ~IXGBE_DMATXCTL_GDV;
1568         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
1569
1570         /* CTRL_EXT: Global Double VLAN Disable */
1571         ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1572         ctrl &= ~IXGBE_EXTENDED_VLAN;
1573         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
1574
1575 }
1576
1577 static void
1578 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
1579 {
1580         struct ixgbe_hw *hw =
1581                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1582         uint32_t ctrl;
1583
1584         PMD_INIT_FUNC_TRACE();
1585
1586         /* DMATXCTRL: Geric Double VLAN Enable */
1587         ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1588         ctrl |= IXGBE_DMATXCTL_GDV;
1589         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
1590
1591         /* CTRL_EXT: Global Double VLAN Enable */
1592         ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1593         ctrl |= IXGBE_EXTENDED_VLAN;
1594         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
1595
1596         /*
1597          * VET EXT field in the EXVET register = 0x8100 by default
1598          * So no need to change. Same to VT field of DMATXCTL register
1599          */
1600 }
1601
1602 static void
1603 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1604 {
1605         if(mask & ETH_VLAN_STRIP_MASK){
1606                 if (dev->data->dev_conf.rxmode.hw_vlan_strip)
1607                         ixgbe_vlan_hw_strip_enable_all(dev);
1608                 else
1609                         ixgbe_vlan_hw_strip_disable_all(dev);
1610         }
1611
1612         if(mask & ETH_VLAN_FILTER_MASK){
1613                 if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1614                         ixgbe_vlan_hw_filter_enable(dev);
1615                 else
1616                         ixgbe_vlan_hw_filter_disable(dev);
1617         }
1618
1619         if(mask & ETH_VLAN_EXTEND_MASK){
1620                 if (dev->data->dev_conf.rxmode.hw_vlan_extend)
1621                         ixgbe_vlan_hw_extend_enable(dev);
1622                 else
1623                         ixgbe_vlan_hw_extend_disable(dev);
1624         }
1625 }
1626
1627 static void
1628 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1629 {
1630         struct ixgbe_hw *hw =
1631                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1632         /* VLNCTRL: enable vlan filtering and allow all vlan tags through */
1633         uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1634         vlanctrl |= IXGBE_VLNCTRL_VFE ; /* enable vlan filters */
1635         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
1636 }
1637
1638 static int
1639 ixgbe_dev_configure(struct rte_eth_dev *dev)
1640 {
1641         struct ixgbe_interrupt *intr =
1642                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1643         struct ixgbe_adapter *adapter =
1644                 (struct ixgbe_adapter *)dev->data->dev_private;
1645
1646         PMD_INIT_FUNC_TRACE();
1647
1648         /* set flag to update link status after init */
1649         intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
1650
1651         /*
1652          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
1653          * allocation or vector Rx preconditions we will reset it.
1654          */
1655         adapter->rx_bulk_alloc_allowed = true;
1656         adapter->rx_vec_allowed = true;
1657
1658         return 0;
1659 }
1660
1661 /*
1662  * Configure device link speed and setup link.
1663  * It returns 0 on success.
1664  */
1665 static int
1666 ixgbe_dev_start(struct rte_eth_dev *dev)
1667 {
1668         struct ixgbe_hw *hw =
1669                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1670         struct ixgbe_vf_info *vfinfo =
1671                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
1672         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
1673         uint32_t intr_vector = 0;
1674         int err, link_up = 0, negotiate = 0;
1675         uint32_t speed = 0;
1676         int mask = 0;
1677         int status;
1678         uint16_t vf, idx;
1679
1680         PMD_INIT_FUNC_TRACE();
1681
1682         /* IXGBE devices don't support half duplex */
1683         if ((dev->data->dev_conf.link_duplex != ETH_LINK_AUTONEG_DUPLEX) &&
1684                         (dev->data->dev_conf.link_duplex != ETH_LINK_FULL_DUPLEX)) {
1685                 PMD_INIT_LOG(ERR, "Invalid link_duplex (%hu) for port %hhu",
1686                              dev->data->dev_conf.link_duplex,
1687                              dev->data->port_id);
1688                 return -EINVAL;
1689         }
1690
1691         /* stop adapter */
1692         hw->adapter_stopped = 0;
1693         ixgbe_stop_adapter(hw);
1694
1695         /* reinitialize adapter
1696          * this calls reset and start */
1697         status = ixgbe_pf_reset_hw(hw);
1698         if (status != 0)
1699                 return -1;
1700         hw->mac.ops.start_hw(hw);
1701         hw->mac.get_link_status = true;
1702
1703         /* configure PF module if SRIOV enabled */
1704         ixgbe_pf_host_configure(dev);
1705
1706         /* check and configure queue intr-vector mapping */
1707         if (dev->data->dev_conf.intr_conf.rxq != 0)
1708                 intr_vector = dev->data->nb_rx_queues;
1709
1710         if (rte_intr_efd_enable(intr_handle, intr_vector))
1711                 return -1;
1712
1713         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
1714                 intr_handle->intr_vec =
1715                         rte_zmalloc("intr_vec",
1716                                     dev->data->nb_rx_queues * sizeof(int),
1717                                     0);
1718                 if (intr_handle->intr_vec == NULL) {
1719                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
1720                                      " intr_vec\n", dev->data->nb_rx_queues);
1721                         return -ENOMEM;
1722                 }
1723         }
1724
1725         /* confiugre msix for sleep until rx interrupt */
1726         ixgbe_configure_msix(dev);
1727
1728         /* initialize transmission unit */
1729         ixgbe_dev_tx_init(dev);
1730
1731         /* This can fail when allocating mbufs for descriptor rings */
1732         err = ixgbe_dev_rx_init(dev);
1733         if (err) {
1734                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
1735                 goto error;
1736         }
1737
1738         err = ixgbe_dev_rxtx_start(dev);
1739         if (err < 0) {
1740                 PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
1741                 goto error;
1742         }
1743
1744         /* Skip link setup if loopback mode is enabled for 82599. */
1745         if (hw->mac.type == ixgbe_mac_82599EB &&
1746                         dev->data->dev_conf.lpbk_mode == IXGBE_LPBK_82599_TX_RX)
1747                 goto skip_link_setup;
1748
1749         if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
1750                 err = hw->mac.ops.setup_sfp(hw);
1751                 if (err)
1752                         goto error;
1753         }
1754
1755         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
1756                 /* Turn on the copper */
1757                 ixgbe_set_phy_power(hw, true);
1758         } else {
1759                 /* Turn on the laser */
1760                 ixgbe_enable_tx_laser(hw);
1761         }
1762
1763         err = ixgbe_check_link(hw, &speed, &link_up, 0);
1764         if (err)
1765                 goto error;
1766         dev->data->dev_link.link_status = link_up;
1767
1768         err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
1769         if (err)
1770                 goto error;
1771
1772         switch(dev->data->dev_conf.link_speed) {
1773         case ETH_LINK_SPEED_AUTONEG:
1774                 speed = (hw->mac.type != ixgbe_mac_82598EB) ?
1775                                 IXGBE_LINK_SPEED_82599_AUTONEG :
1776                                 IXGBE_LINK_SPEED_82598_AUTONEG;
1777                 break;
1778         case ETH_LINK_SPEED_100:
1779                 /*
1780                  * Invalid for 82598 but error will be detected by
1781                  * ixgbe_setup_link()
1782                  */
1783                 speed = IXGBE_LINK_SPEED_100_FULL;
1784                 break;
1785         case ETH_LINK_SPEED_1000:
1786                 speed = IXGBE_LINK_SPEED_1GB_FULL;
1787                 break;
1788         case ETH_LINK_SPEED_10000:
1789                 speed = IXGBE_LINK_SPEED_10GB_FULL;
1790                 break;
1791         default:
1792                 PMD_INIT_LOG(ERR, "Invalid link_speed (%hu) for port %hhu",
1793                              dev->data->dev_conf.link_speed,
1794                              dev->data->port_id);
1795                 goto error;
1796         }
1797
1798         err = ixgbe_setup_link(hw, speed, link_up);
1799         if (err)
1800                 goto error;
1801
1802 skip_link_setup:
1803
1804         /* check if lsc interrupt is enabled */
1805         if (dev->data->dev_conf.intr_conf.lsc != 0) {
1806                 if (rte_intr_allow_others(intr_handle)) {
1807                         rte_intr_callback_register(intr_handle,
1808                                                    ixgbe_dev_interrupt_handler,
1809                                                    (void *)dev);
1810                         ixgbe_dev_lsc_interrupt_setup(dev);
1811                 } else
1812                         PMD_INIT_LOG(INFO, "lsc won't enable because of"
1813                                      " no intr multiplex\n");
1814         }
1815
1816         /* check if rxq interrupt is enabled */
1817         if (dev->data->dev_conf.intr_conf.rxq != 0)
1818                 ixgbe_dev_rxq_interrupt_setup(dev);
1819
1820         /* enable uio/vfio intr/eventfd mapping */
1821         rte_intr_enable(intr_handle);
1822
1823         /* resume enabled intr since hw reset */
1824         ixgbe_enable_intr(dev);
1825
1826         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
1827                 ETH_VLAN_EXTEND_MASK;
1828         ixgbe_vlan_offload_set(dev, mask);
1829
1830         if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1831                 /* Enable vlan filtering for VMDq */
1832                 ixgbe_vmdq_vlan_hw_filter_enable(dev);
1833         }
1834
1835         /* Configure DCB hw */
1836         ixgbe_configure_dcb(dev);
1837
1838         if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
1839                 err = ixgbe_fdir_configure(dev);
1840                 if (err)
1841                         goto error;
1842         }
1843
1844         /* Restore vf rate limit */
1845         if (vfinfo != NULL) {
1846                 for (vf = 0; vf < dev->pci_dev->max_vfs; vf++)
1847                         for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
1848                                 if (vfinfo[vf].tx_rate[idx] != 0)
1849                                         ixgbe_set_vf_rate_limit(dev, vf,
1850                                                 vfinfo[vf].tx_rate[idx],
1851                                                 1 << idx);
1852         }
1853
1854         ixgbe_restore_statistics_mapping(dev);
1855
1856         return (0);
1857
1858 error:
1859         PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
1860         ixgbe_dev_clear_queues(dev);
1861         return -EIO;
1862 }
1863
1864 /*
1865  * Stop device: disable rx and tx functions to allow for reconfiguring.
1866  */
1867 static void
1868 ixgbe_dev_stop(struct rte_eth_dev *dev)
1869 {
1870         struct rte_eth_link link;
1871         struct ixgbe_hw *hw =
1872                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1873         struct ixgbe_vf_info *vfinfo =
1874                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
1875         struct ixgbe_filter_info *filter_info =
1876                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
1877         struct ixgbe_5tuple_filter *p_5tuple, *p_5tuple_next;
1878         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
1879         int vf;
1880
1881         PMD_INIT_FUNC_TRACE();
1882
1883         /* disable interrupts */
1884         ixgbe_disable_intr(hw);
1885
1886         /* disable intr eventfd mapping */
1887         rte_intr_disable(intr_handle);
1888
1889         /* reset the NIC */
1890         ixgbe_pf_reset_hw(hw);
1891         hw->adapter_stopped = 0;
1892
1893         /* stop adapter */
1894         ixgbe_stop_adapter(hw);
1895
1896         for (vf = 0; vfinfo != NULL &&
1897                      vf < dev->pci_dev->max_vfs; vf++)
1898                 vfinfo[vf].clear_to_send = false;
1899
1900         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
1901                 /* Turn off the copper */
1902                 ixgbe_set_phy_power(hw, false);
1903         } else {
1904                 /* Turn off the laser */
1905                 ixgbe_disable_tx_laser(hw);
1906         }
1907
1908         ixgbe_dev_clear_queues(dev);
1909
1910         /* Clear stored conf */
1911         dev->data->scattered_rx = 0;
1912         dev->data->lro = 0;
1913
1914         /* Clear recorded link status */
1915         memset(&link, 0, sizeof(link));
1916         rte_ixgbe_dev_atomic_write_link_status(dev, &link);
1917
1918         /* Remove all ntuple filters of the device */
1919         for (p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list);
1920              p_5tuple != NULL; p_5tuple = p_5tuple_next) {
1921                 p_5tuple_next = TAILQ_NEXT(p_5tuple, entries);
1922                 TAILQ_REMOVE(&filter_info->fivetuple_list,
1923                              p_5tuple, entries);
1924                 rte_free(p_5tuple);
1925         }
1926         memset(filter_info->fivetuple_mask, 0,
1927                 sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1928
1929         /* Clean datapath event and queue/vec mapping */
1930         rte_intr_efd_disable(intr_handle);
1931         if (intr_handle->intr_vec != NULL) {
1932                 rte_free(intr_handle->intr_vec);
1933                 intr_handle->intr_vec = NULL;
1934         }
1935 }
1936
1937 /*
1938  * Set device link up: enable tx.
1939  */
1940 static int
1941 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
1942 {
1943         struct ixgbe_hw *hw =
1944                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1945         if (hw->mac.type == ixgbe_mac_82599EB) {
1946 #ifdef RTE_NIC_BYPASS
1947                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
1948                         /* Not suported in bypass mode */
1949                         PMD_INIT_LOG(ERR, "Set link up is not supported "
1950                                      "by device id 0x%x", hw->device_id);
1951                         return -ENOTSUP;
1952                 }
1953 #endif
1954         }
1955
1956         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
1957                 /* Turn on the copper */
1958                 ixgbe_set_phy_power(hw, true);
1959         } else {
1960                 /* Turn on the laser */
1961                 ixgbe_enable_tx_laser(hw);
1962         }
1963
1964         return 0;
1965 }
1966
1967 /*
1968  * Set device link down: disable tx.
1969  */
1970 static int
1971 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
1972 {
1973         struct ixgbe_hw *hw =
1974                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1975         if (hw->mac.type == ixgbe_mac_82599EB) {
1976 #ifdef RTE_NIC_BYPASS
1977                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
1978                         /* Not suported in bypass mode */
1979                         PMD_INIT_LOG(ERR, "Set link down is not supported "
1980                                      "by device id 0x%x", hw->device_id);
1981                         return -ENOTSUP;
1982                 }
1983 #endif
1984         }
1985
1986         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
1987                 /* Turn off the copper */
1988                 ixgbe_set_phy_power(hw, false);
1989         } else {
1990                 /* Turn off the laser */
1991                 ixgbe_disable_tx_laser(hw);
1992         }
1993
1994         return 0;
1995 }
1996
1997 /*
1998  * Reest and stop device.
1999  */
2000 static void
2001 ixgbe_dev_close(struct rte_eth_dev *dev)
2002 {
2003         struct ixgbe_hw *hw =
2004                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2005
2006         PMD_INIT_FUNC_TRACE();
2007
2008         ixgbe_pf_reset_hw(hw);
2009
2010         ixgbe_dev_stop(dev);
2011         hw->adapter_stopped = 1;
2012
2013         ixgbe_dev_free_queues(dev);
2014
2015         ixgbe_disable_pcie_master(hw);
2016
2017         /* reprogram the RAR[0] in case user changed it. */
2018         ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
2019 }
2020
2021 static void
2022 ixgbe_read_stats_registers(struct ixgbe_hw *hw, struct ixgbe_hw_stats
2023                                                    *hw_stats, uint64_t *total_missed_rx,
2024                                                    uint64_t *total_qbrc, uint64_t *total_qprc,
2025                                                    uint64_t *total_qprdc)
2026 {
2027         uint32_t bprc, lxon, lxoff, total;
2028         unsigned i;
2029
2030         hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
2031         hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
2032         hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
2033         hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
2034
2035         for (i = 0; i < 8; i++) {
2036                 uint32_t mp;
2037                 mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
2038                 /* global total per queue */
2039                 hw_stats->mpc[i] += mp;
2040                 /* Running comprehensive total for stats display */
2041                 *total_missed_rx += hw_stats->mpc[i];
2042                 if (hw->mac.type == ixgbe_mac_82598EB) {
2043                         hw_stats->rnbc[i] +=
2044                             IXGBE_READ_REG(hw, IXGBE_RNBC(i));
2045                         hw_stats->pxonrxc[i] +=
2046                                 IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
2047                         hw_stats->pxoffrxc[i] +=
2048                                 IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
2049                 } else {
2050                         hw_stats->pxonrxc[i] +=
2051                                 IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
2052                         hw_stats->pxoffrxc[i] +=
2053                                 IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
2054                         hw_stats->pxon2offc[i] +=
2055                                 IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
2056                 }
2057                 hw_stats->pxontxc[i] +=
2058                     IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
2059                 hw_stats->pxofftxc[i] +=
2060                     IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
2061         }
2062         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
2063                 hw_stats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
2064                 hw_stats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
2065                 hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
2066                 hw_stats->qbrc[i] +=
2067                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
2068                 hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
2069                 hw_stats->qbtc[i] +=
2070                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
2071                 *total_qprdc += hw_stats->qprdc[i] +=
2072                                 IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
2073
2074                 *total_qprc += hw_stats->qprc[i];
2075                 *total_qbrc += hw_stats->qbrc[i];
2076         }
2077         hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
2078         hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
2079         hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
2080
2081         /* Note that gprc counts missed packets */
2082         hw_stats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
2083
2084         if (hw->mac.type != ixgbe_mac_82598EB) {
2085                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
2086                 hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
2087                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
2088                 hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
2089                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
2090                 hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
2091                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
2092                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
2093         } else {
2094                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
2095                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
2096                 /* 82598 only has a counter in the high register */
2097                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
2098                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
2099                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
2100         }
2101
2102         /*
2103          * Workaround: mprc hardware is incorrectly counting
2104          * broadcasts, so for now we subtract those.
2105          */
2106         bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
2107         hw_stats->bprc += bprc;
2108         hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
2109         if (hw->mac.type == ixgbe_mac_82598EB)
2110                 hw_stats->mprc -= bprc;
2111
2112         hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
2113         hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
2114         hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
2115         hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
2116         hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
2117         hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
2118
2119         lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
2120         hw_stats->lxontxc += lxon;
2121         lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
2122         hw_stats->lxofftxc += lxoff;
2123         total = lxon + lxoff;
2124
2125         hw_stats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
2126         hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
2127         hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
2128         hw_stats->gptc -= total;
2129         hw_stats->mptc -= total;
2130         hw_stats->ptc64 -= total;
2131         hw_stats->gotc -= total * ETHER_MIN_LEN;
2132
2133         hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
2134         hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
2135         hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
2136         hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
2137         hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
2138         hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
2139         hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
2140         hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
2141         hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
2142         hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
2143         hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
2144         hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
2145         hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
2146         hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
2147         hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
2148         hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
2149         hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
2150         hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
2151         /* Only read FCOE on 82599 */
2152         if (hw->mac.type != ixgbe_mac_82598EB) {
2153                 hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
2154                 hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
2155                 hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
2156                 hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
2157                 hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
2158         }
2159
2160         /* Flow Director Stats registers */
2161         hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
2162         hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
2163 }
2164
2165 /*
2166  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
2167  */
2168 static void
2169 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
2170 {
2171         struct ixgbe_hw *hw =
2172                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2173         struct ixgbe_hw_stats *hw_stats =
2174                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2175         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
2176         unsigned i;
2177
2178         total_missed_rx = 0;
2179         total_qbrc = 0;
2180         total_qprc = 0;
2181         total_qprdc = 0;
2182
2183         ixgbe_read_stats_registers(hw, hw_stats, &total_missed_rx, &total_qbrc,
2184                         &total_qprc, &total_qprdc);
2185
2186         if (stats == NULL)
2187                 return;
2188
2189         /* Fill out the rte_eth_stats statistics structure */
2190         stats->ipackets = total_qprc;
2191         stats->ibytes = total_qbrc;
2192         stats->opackets = hw_stats->gptc;
2193         stats->obytes = hw_stats->gotc;
2194
2195         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
2196                 stats->q_ipackets[i] = hw_stats->qprc[i];
2197                 stats->q_opackets[i] = hw_stats->qptc[i];
2198                 stats->q_ibytes[i] = hw_stats->qbrc[i];
2199                 stats->q_obytes[i] = hw_stats->qbtc[i];
2200                 stats->q_errors[i] = hw_stats->qprdc[i];
2201         }
2202
2203         /* Rx Errors */
2204         stats->ierrors  = hw_stats->crcerrs +
2205                           hw_stats->rlec +
2206                           hw_stats->ruc +
2207                           hw_stats->roc +
2208                           total_missed_rx +
2209                           hw_stats->illerrc +
2210                           hw_stats->errbc +
2211                           hw_stats->xec +
2212                           hw_stats->mlfc +
2213                           hw_stats->mrfc +
2214                           hw_stats->rfc +
2215                           hw_stats->fccrc +
2216                           hw_stats->fclast;
2217
2218         /* Tx Errors */
2219         stats->oerrors  = 0;
2220 }
2221
2222 static void
2223 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
2224 {
2225         struct ixgbe_hw_stats *stats =
2226                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2227
2228         /* HW registers are cleared on read */
2229         ixgbe_dev_stats_get(dev, NULL);
2230
2231         /* Reset software totals */
2232         memset(stats, 0, sizeof(*stats));
2233 }
2234
2235 static int
2236 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstats *xstats,
2237                                          unsigned n)
2238 {
2239         struct ixgbe_hw *hw =
2240                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2241         struct ixgbe_hw_stats *hw_stats =
2242                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2243         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
2244         unsigned i, count = IXGBE_NB_XSTATS;
2245
2246         if (n < count)
2247                 return count;
2248
2249         total_missed_rx = 0;
2250         total_qbrc = 0;
2251         total_qprc = 0;
2252         total_qprdc = 0;
2253
2254         ixgbe_read_stats_registers(hw, hw_stats, &total_missed_rx, &total_qbrc,
2255                                                            &total_qprc, &total_qprdc);
2256
2257         /* If this is a reset xstats is NULL, and we have cleared the
2258          * registers by reading them.
2259          */
2260         if (!xstats)
2261                 return 0;
2262
2263         /* Extended stats */
2264         for (i = 0; i < IXGBE_NB_XSTATS; i++) {
2265                 snprintf(xstats[i].name, sizeof(xstats[i].name),
2266                                 "%s", rte_ixgbe_stats_strings[i].name);
2267                 xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
2268                                                         rte_ixgbe_stats_strings[i].offset);
2269         }
2270
2271         return count;
2272 }
2273
2274 static void
2275 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
2276 {
2277         struct ixgbe_hw_stats *stats =
2278                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2279
2280         /* HW registers are cleared on read */
2281         ixgbe_dev_xstats_get(dev, NULL, IXGBE_NB_XSTATS);
2282
2283         /* Reset software totals */
2284         memset(stats, 0, sizeof(*stats));
2285 }
2286
2287 static void
2288 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
2289 {
2290         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2291         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats*)
2292                           IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2293
2294         /* Good Rx packet, include VF loopback */
2295         UPDATE_VF_STAT(IXGBE_VFGPRC,
2296             hw_stats->last_vfgprc, hw_stats->vfgprc);
2297
2298         /* Good Rx octets, include VF loopback */
2299         UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
2300             hw_stats->last_vfgorc, hw_stats->vfgorc);
2301
2302         /* Good Tx packet, include VF loopback */
2303         UPDATE_VF_STAT(IXGBE_VFGPTC,
2304             hw_stats->last_vfgptc, hw_stats->vfgptc);
2305
2306         /* Good Tx octets, include VF loopback */
2307         UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
2308             hw_stats->last_vfgotc, hw_stats->vfgotc);
2309
2310         /* Rx Multicst Packet */
2311         UPDATE_VF_STAT(IXGBE_VFMPRC,
2312             hw_stats->last_vfmprc, hw_stats->vfmprc);
2313
2314         if (stats == NULL)
2315                 return;
2316
2317         stats->ipackets = hw_stats->vfgprc;
2318         stats->ibytes = hw_stats->vfgorc;
2319         stats->opackets = hw_stats->vfgptc;
2320         stats->obytes = hw_stats->vfgotc;
2321         stats->imcasts = hw_stats->vfmprc;
2322         /* stats->imcasts should be removed as imcasts is deprecated */
2323 }
2324
2325 static void
2326 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
2327 {
2328         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats*)
2329                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2330
2331         /* Sync HW register to the last stats */
2332         ixgbevf_dev_stats_get(dev, NULL);
2333
2334         /* reset HW current stats*/
2335         hw_stats->vfgprc = 0;
2336         hw_stats->vfgorc = 0;
2337         hw_stats->vfgptc = 0;
2338         hw_stats->vfgotc = 0;
2339         hw_stats->vfmprc = 0;
2340
2341 }
2342
2343 static void
2344 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2345 {
2346         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2347
2348         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
2349         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
2350         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
2351         dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
2352         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
2353         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
2354         dev_info->max_vfs = dev->pci_dev->max_vfs;
2355         if (hw->mac.type == ixgbe_mac_82598EB)
2356                 dev_info->max_vmdq_pools = ETH_16_POOLS;
2357         else
2358                 dev_info->max_vmdq_pools = ETH_64_POOLS;
2359         dev_info->vmdq_queue_num = dev_info->max_rx_queues;
2360         dev_info->rx_offload_capa =
2361                 DEV_RX_OFFLOAD_VLAN_STRIP |
2362                 DEV_RX_OFFLOAD_IPV4_CKSUM |
2363                 DEV_RX_OFFLOAD_UDP_CKSUM  |
2364                 DEV_RX_OFFLOAD_TCP_CKSUM;
2365
2366         /*
2367          * RSC is only supported by 82599 and x540 PF devices in a non-SR-IOV
2368          * mode.
2369          */
2370         if ((hw->mac.type == ixgbe_mac_82599EB ||
2371              hw->mac.type == ixgbe_mac_X540) &&
2372             !RTE_ETH_DEV_SRIOV(dev).active)
2373                 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TCP_LRO;
2374
2375         dev_info->tx_offload_capa =
2376                 DEV_TX_OFFLOAD_VLAN_INSERT |
2377                 DEV_TX_OFFLOAD_IPV4_CKSUM  |
2378                 DEV_TX_OFFLOAD_UDP_CKSUM   |
2379                 DEV_TX_OFFLOAD_TCP_CKSUM   |
2380                 DEV_TX_OFFLOAD_SCTP_CKSUM  |
2381                 DEV_TX_OFFLOAD_TCP_TSO;
2382
2383         dev_info->default_rxconf = (struct rte_eth_rxconf) {
2384                 .rx_thresh = {
2385                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
2386                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
2387                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
2388                 },
2389                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
2390                 .rx_drop_en = 0,
2391         };
2392
2393         dev_info->default_txconf = (struct rte_eth_txconf) {
2394                 .tx_thresh = {
2395                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
2396                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
2397                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
2398                 },
2399                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
2400                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
2401                 .txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
2402                                 ETH_TXQ_FLAGS_NOOFFLOADS,
2403         };
2404         dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
2405         dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
2406         dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
2407 }
2408
2409 static void
2410 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
2411                      struct rte_eth_dev_info *dev_info)
2412 {
2413         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2414
2415         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
2416         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
2417         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
2418         dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS reg */
2419         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
2420         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
2421         dev_info->max_vfs = dev->pci_dev->max_vfs;
2422         if (hw->mac.type == ixgbe_mac_82598EB)
2423                 dev_info->max_vmdq_pools = ETH_16_POOLS;
2424         else
2425                 dev_info->max_vmdq_pools = ETH_64_POOLS;
2426         dev_info->rx_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP |
2427                                 DEV_RX_OFFLOAD_IPV4_CKSUM |
2428                                 DEV_RX_OFFLOAD_UDP_CKSUM  |
2429                                 DEV_RX_OFFLOAD_TCP_CKSUM;
2430         dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
2431                                 DEV_TX_OFFLOAD_IPV4_CKSUM  |
2432                                 DEV_TX_OFFLOAD_UDP_CKSUM   |
2433                                 DEV_TX_OFFLOAD_TCP_CKSUM   |
2434                                 DEV_TX_OFFLOAD_SCTP_CKSUM  |
2435                                 DEV_TX_OFFLOAD_TCP_TSO;
2436
2437         dev_info->default_rxconf = (struct rte_eth_rxconf) {
2438                 .rx_thresh = {
2439                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
2440                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
2441                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
2442                 },
2443                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
2444                 .rx_drop_en = 0,
2445         };
2446
2447         dev_info->default_txconf = (struct rte_eth_txconf) {
2448                 .tx_thresh = {
2449                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
2450                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
2451                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
2452                 },
2453                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
2454                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
2455                 .txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
2456                                 ETH_TXQ_FLAGS_NOOFFLOADS,
2457         };
2458 }
2459
2460 /* return 0 means link status changed, -1 means not changed */
2461 static int
2462 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
2463 {
2464         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2465         struct rte_eth_link link, old;
2466         ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
2467         int link_up;
2468         int diag;
2469
2470         link.link_status = 0;
2471         link.link_speed = 0;
2472         link.link_duplex = 0;
2473         memset(&old, 0, sizeof(old));
2474         rte_ixgbe_dev_atomic_read_link_status(dev, &old);
2475
2476         hw->mac.get_link_status = true;
2477
2478         /* check if it needs to wait to complete, if lsc interrupt is enabled */
2479         if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
2480                 diag = ixgbe_check_link(hw, &link_speed, &link_up, 0);
2481         else
2482                 diag = ixgbe_check_link(hw, &link_speed, &link_up, 1);
2483
2484         if (diag != 0) {
2485                 link.link_speed = ETH_LINK_SPEED_100;
2486                 link.link_duplex = ETH_LINK_HALF_DUPLEX;
2487                 rte_ixgbe_dev_atomic_write_link_status(dev, &link);
2488                 if (link.link_status == old.link_status)
2489                         return -1;
2490                 return 0;
2491         }
2492
2493         if (link_up == 0) {
2494                 rte_ixgbe_dev_atomic_write_link_status(dev, &link);
2495                 if (link.link_status == old.link_status)
2496                         return -1;
2497                 return 0;
2498         }
2499         link.link_status = 1;
2500         link.link_duplex = ETH_LINK_FULL_DUPLEX;
2501
2502         switch (link_speed) {
2503         default:
2504         case IXGBE_LINK_SPEED_UNKNOWN:
2505                 link.link_duplex = ETH_LINK_HALF_DUPLEX;
2506                 link.link_speed = ETH_LINK_SPEED_100;
2507                 break;
2508
2509         case IXGBE_LINK_SPEED_100_FULL:
2510                 link.link_speed = ETH_LINK_SPEED_100;
2511                 break;
2512
2513         case IXGBE_LINK_SPEED_1GB_FULL:
2514                 link.link_speed = ETH_LINK_SPEED_1000;
2515                 break;
2516
2517         case IXGBE_LINK_SPEED_10GB_FULL:
2518                 link.link_speed = ETH_LINK_SPEED_10000;
2519                 break;
2520         }
2521         rte_ixgbe_dev_atomic_write_link_status(dev, &link);
2522
2523         if (link.link_status == old.link_status)
2524                 return -1;
2525
2526         return 0;
2527 }
2528
2529 static void
2530 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
2531 {
2532         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2533         uint32_t fctrl;
2534
2535         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
2536         fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
2537         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
2538 }
2539
2540 static void
2541 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
2542 {
2543         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2544         uint32_t fctrl;
2545
2546         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
2547         fctrl &= (~IXGBE_FCTRL_UPE);
2548         if (dev->data->all_multicast == 1)
2549                 fctrl |= IXGBE_FCTRL_MPE;
2550         else
2551                 fctrl &= (~IXGBE_FCTRL_MPE);
2552         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
2553 }
2554
2555 static void
2556 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
2557 {
2558         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2559         uint32_t fctrl;
2560
2561         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
2562         fctrl |= IXGBE_FCTRL_MPE;
2563         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
2564 }
2565
2566 static void
2567 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
2568 {
2569         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2570         uint32_t fctrl;
2571
2572         if (dev->data->promiscuous == 1)
2573                 return; /* must remain in all_multicast mode */
2574
2575         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
2576         fctrl &= (~IXGBE_FCTRL_MPE);
2577         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
2578 }
2579
2580 /**
2581  * It clears the interrupt causes and enables the interrupt.
2582  * It will be called once only during nic initialized.
2583  *
2584  * @param dev
2585  *  Pointer to struct rte_eth_dev.
2586  *
2587  * @return
2588  *  - On success, zero.
2589  *  - On failure, a negative value.
2590  */
2591 static int
2592 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev)
2593 {
2594         struct ixgbe_interrupt *intr =
2595                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2596
2597         ixgbe_dev_link_status_print(dev);
2598         intr->mask |= IXGBE_EICR_LSC;
2599
2600         return 0;
2601 }
2602
2603 /**
2604  * It clears the interrupt causes and enables the interrupt.
2605  * It will be called once only during nic initialized.
2606  *
2607  * @param dev
2608  *  Pointer to struct rte_eth_dev.
2609  *
2610  * @return
2611  *  - On success, zero.
2612  *  - On failure, a negative value.
2613  */
2614 static int
2615 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
2616 {
2617         struct ixgbe_interrupt *intr =
2618                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2619
2620         intr->mask |= IXGBE_EICR_RTX_QUEUE;
2621
2622         return 0;
2623 }
2624
2625 /*
2626  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
2627  *
2628  * @param dev
2629  *  Pointer to struct rte_eth_dev.
2630  *
2631  * @return
2632  *  - On success, zero.
2633  *  - On failure, a negative value.
2634  */
2635 static int
2636 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
2637 {
2638         uint32_t eicr;
2639         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2640         struct ixgbe_interrupt *intr =
2641                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2642
2643         /* clear all cause mask */
2644         ixgbe_disable_intr(hw);
2645
2646         /* read-on-clear nic registers here */
2647         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2648         PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
2649
2650         intr->flags = 0;
2651
2652         /* set flag for async link update */
2653         if (eicr & IXGBE_EICR_LSC)
2654                 intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2655
2656         if (eicr & IXGBE_EICR_MAILBOX)
2657                 intr->flags |= IXGBE_FLAG_MAILBOX;
2658
2659         return 0;
2660 }
2661
2662 static int
2663 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
2664 {
2665         uint32_t eicr;
2666         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2667         struct ixgbe_interrupt *intr =
2668                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2669
2670         /* clear all cause mask */
2671         ixgbevf_intr_disable(hw);
2672
2673         /* read-on-clear nic registers here */
2674         eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
2675         PMD_DRV_LOG(INFO, "eicr %x", eicr);
2676
2677         intr->flags = 0;
2678
2679         /* set flag for async link update */
2680         if (eicr & IXGBE_EICR_LSC)
2681                 intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2682
2683         return 0;
2684 }
2685
2686 /**
2687  * It gets and then prints the link status.
2688  *
2689  * @param dev
2690  *  Pointer to struct rte_eth_dev.
2691  *
2692  * @return
2693  *  - On success, zero.
2694  *  - On failure, a negative value.
2695  */
2696 static void
2697 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
2698 {
2699         struct rte_eth_link link;
2700
2701         memset(&link, 0, sizeof(link));
2702         rte_ixgbe_dev_atomic_read_link_status(dev, &link);
2703         if (link.link_status) {
2704                 PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
2705                                         (int)(dev->data->port_id),
2706                                         (unsigned)link.link_speed,
2707                         link.link_duplex == ETH_LINK_FULL_DUPLEX ?
2708                                         "full-duplex" : "half-duplex");
2709         } else {
2710                 PMD_INIT_LOG(INFO, " Port %d: Link Down",
2711                                 (int)(dev->data->port_id));
2712         }
2713         PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
2714                                 dev->pci_dev->addr.domain,
2715                                 dev->pci_dev->addr.bus,
2716                                 dev->pci_dev->addr.devid,
2717                                 dev->pci_dev->addr.function);
2718 }
2719
2720 /*
2721  * It executes link_update after knowing an interrupt occurred.
2722  *
2723  * @param dev
2724  *  Pointer to struct rte_eth_dev.
2725  *
2726  * @return
2727  *  - On success, zero.
2728  *  - On failure, a negative value.
2729  */
2730 static int
2731 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
2732 {
2733         struct ixgbe_interrupt *intr =
2734                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2735         int64_t timeout;
2736         struct rte_eth_link link;
2737         int intr_enable_delay = false;
2738
2739         PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
2740
2741         if (intr->flags & IXGBE_FLAG_MAILBOX) {
2742                 ixgbe_pf_mbx_process(dev);
2743                 intr->flags &= ~IXGBE_FLAG_MAILBOX;
2744         }
2745
2746         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
2747                 /* get the link status before link update, for predicting later */
2748                 memset(&link, 0, sizeof(link));
2749                 rte_ixgbe_dev_atomic_read_link_status(dev, &link);
2750
2751                 ixgbe_dev_link_update(dev, 0);
2752
2753                 /* likely to up */
2754                 if (!link.link_status)
2755                         /* handle it 1 sec later, wait it being stable */
2756                         timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
2757                 /* likely to down */
2758                 else
2759                         /* handle it 4 sec later, wait it being stable */
2760                         timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
2761
2762                 ixgbe_dev_link_status_print(dev);
2763
2764                 intr_enable_delay = true;
2765         }
2766
2767         if (intr_enable_delay) {
2768                 if (rte_eal_alarm_set(timeout * 1000,
2769                                       ixgbe_dev_interrupt_delayed_handler, (void*)dev) < 0)
2770                         PMD_DRV_LOG(ERR, "Error setting alarm");
2771         } else {
2772                 PMD_DRV_LOG(DEBUG, "enable intr immediately");
2773                 ixgbe_enable_intr(dev);
2774                 rte_intr_enable(&(dev->pci_dev->intr_handle));
2775         }
2776
2777
2778         return 0;
2779 }
2780
2781 static int
2782 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
2783 {
2784         struct ixgbe_hw *hw =
2785                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2786
2787         PMD_DRV_LOG(DEBUG, "enable intr immediately");
2788         ixgbevf_intr_enable(hw);
2789         rte_intr_enable(&dev->pci_dev->intr_handle);
2790         return 0;
2791 }
2792
2793 /**
2794  * Interrupt handler which shall be registered for alarm callback for delayed
2795  * handling specific interrupt to wait for the stable nic state. As the
2796  * NIC interrupt state is not stable for ixgbe after link is just down,
2797  * it needs to wait 4 seconds to get the stable status.
2798  *
2799  * @param handle
2800  *  Pointer to interrupt handle.
2801  * @param param
2802  *  The address of parameter (struct rte_eth_dev *) regsitered before.
2803  *
2804  * @return
2805  *  void
2806  */
2807 static void
2808 ixgbe_dev_interrupt_delayed_handler(void *param)
2809 {
2810         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2811         struct ixgbe_interrupt *intr =
2812                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2813         struct ixgbe_hw *hw =
2814                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2815         uint32_t eicr;
2816
2817         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2818         if (eicr & IXGBE_EICR_MAILBOX)
2819                 ixgbe_pf_mbx_process(dev);
2820
2821         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
2822                 ixgbe_dev_link_update(dev, 0);
2823                 intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
2824                 ixgbe_dev_link_status_print(dev);
2825                 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC);
2826         }
2827
2828         PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
2829         ixgbe_enable_intr(dev);
2830         rte_intr_enable(&(dev->pci_dev->intr_handle));
2831 }
2832
2833 /**
2834  * Interrupt handler triggered by NIC  for handling
2835  * specific interrupt.
2836  *
2837  * @param handle
2838  *  Pointer to interrupt handle.
2839  * @param param
2840  *  The address of parameter (struct rte_eth_dev *) regsitered before.
2841  *
2842  * @return
2843  *  void
2844  */
2845 static void
2846 ixgbe_dev_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
2847                             void *param)
2848 {
2849         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2850
2851         ixgbe_dev_interrupt_get_status(dev);
2852         ixgbe_dev_interrupt_action(dev);
2853 }
2854
2855 static void
2856 ixgbevf_dev_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
2857                               void *param)
2858 {
2859         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2860
2861         ixgbevf_dev_interrupt_get_status(dev);
2862         ixgbevf_dev_interrupt_action(dev);
2863 }
2864
2865 static int
2866 ixgbe_dev_led_on(struct rte_eth_dev *dev)
2867 {
2868         struct ixgbe_hw *hw;
2869
2870         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2871         return (ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP);
2872 }
2873
2874 static int
2875 ixgbe_dev_led_off(struct rte_eth_dev *dev)
2876 {
2877         struct ixgbe_hw *hw;
2878
2879         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2880         return (ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP);
2881 }
2882
2883 static int
2884 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2885 {
2886         struct ixgbe_hw *hw;
2887         uint32_t mflcn_reg;
2888         uint32_t fccfg_reg;
2889         int rx_pause;
2890         int tx_pause;
2891
2892         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2893
2894         fc_conf->pause_time = hw->fc.pause_time;
2895         fc_conf->high_water = hw->fc.high_water[0];
2896         fc_conf->low_water = hw->fc.low_water[0];
2897         fc_conf->send_xon = hw->fc.send_xon;
2898         fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
2899
2900         /*
2901          * Return rx_pause status according to actual setting of
2902          * MFLCN register.
2903          */
2904         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
2905         if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
2906                 rx_pause = 1;
2907         else
2908                 rx_pause = 0;
2909
2910         /*
2911          * Return tx_pause status according to actual setting of
2912          * FCCFG register.
2913          */
2914         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
2915         if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
2916                 tx_pause = 1;
2917         else
2918                 tx_pause = 0;
2919
2920         if (rx_pause && tx_pause)
2921                 fc_conf->mode = RTE_FC_FULL;
2922         else if (rx_pause)
2923                 fc_conf->mode = RTE_FC_RX_PAUSE;
2924         else if (tx_pause)
2925                 fc_conf->mode = RTE_FC_TX_PAUSE;
2926         else
2927                 fc_conf->mode = RTE_FC_NONE;
2928
2929         return 0;
2930 }
2931
2932 static int
2933 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2934 {
2935         struct ixgbe_hw *hw;
2936         int err;
2937         uint32_t rx_buf_size;
2938         uint32_t max_high_water;
2939         uint32_t mflcn;
2940         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
2941                 ixgbe_fc_none,
2942                 ixgbe_fc_rx_pause,
2943                 ixgbe_fc_tx_pause,
2944                 ixgbe_fc_full
2945         };
2946
2947         PMD_INIT_FUNC_TRACE();
2948
2949         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2950         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
2951         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
2952
2953         /*
2954          * At least reserve one Ethernet frame for watermark
2955          * high_water/low_water in kilo bytes for ixgbe
2956          */
2957         max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
2958         if ((fc_conf->high_water > max_high_water) ||
2959                 (fc_conf->high_water < fc_conf->low_water)) {
2960                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
2961                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
2962                 return (-EINVAL);
2963         }
2964
2965         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
2966         hw->fc.pause_time     = fc_conf->pause_time;
2967         hw->fc.high_water[0]  = fc_conf->high_water;
2968         hw->fc.low_water[0]   = fc_conf->low_water;
2969         hw->fc.send_xon       = fc_conf->send_xon;
2970         hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
2971
2972         err = ixgbe_fc_enable(hw);
2973
2974         /* Not negotiated is not an error case */
2975         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED)) {
2976
2977                 /* check if we want to forward MAC frames - driver doesn't have native
2978                  * capability to do that, so we'll write the registers ourselves */
2979
2980                 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
2981
2982                 /* set or clear MFLCN.PMCF bit depending on configuration */
2983                 if (fc_conf->mac_ctrl_frame_fwd != 0)
2984                         mflcn |= IXGBE_MFLCN_PMCF;
2985                 else
2986                         mflcn &= ~IXGBE_MFLCN_PMCF;
2987
2988                 IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
2989                 IXGBE_WRITE_FLUSH(hw);
2990
2991                 return 0;
2992         }
2993
2994         PMD_INIT_LOG(ERR, "ixgbe_fc_enable = 0x%x", err);
2995         return -EIO;
2996 }
2997
2998 /**
2999  *  ixgbe_pfc_enable_generic - Enable flow control
3000  *  @hw: pointer to hardware structure
3001  *  @tc_num: traffic class number
3002  *  Enable flow control according to the current settings.
3003  */
3004 static int
3005 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw,uint8_t tc_num)
3006 {
3007         int ret_val = 0;
3008         uint32_t mflcn_reg, fccfg_reg;
3009         uint32_t reg;
3010         uint32_t fcrtl, fcrth;
3011         uint8_t i;
3012         uint8_t nb_rx_en;
3013
3014         /* Validate the water mark configuration */
3015         if (!hw->fc.pause_time) {
3016                 ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
3017                 goto out;
3018         }
3019
3020         /* Low water mark of zero causes XOFF floods */
3021         if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
3022                  /* High/Low water can not be 0 */
3023                 if( (!hw->fc.high_water[tc_num])|| (!hw->fc.low_water[tc_num])) {
3024                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
3025                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
3026                         goto out;
3027                 }
3028
3029                 if(hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
3030                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
3031                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
3032                         goto out;
3033                 }
3034         }
3035         /* Negotiate the fc mode to use */
3036         ixgbe_fc_autoneg(hw);
3037
3038         /* Disable any previous flow control settings */
3039         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
3040         mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
3041
3042         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
3043         fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
3044
3045         switch (hw->fc.current_mode) {
3046         case ixgbe_fc_none:
3047                 /*
3048                  * If the count of enabled RX Priority Flow control >1,
3049                  * and the TX pause can not be disabled
3050                  */
3051                 nb_rx_en = 0;
3052                 for (i =0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
3053                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
3054                         if (reg & IXGBE_FCRTH_FCEN)
3055                                 nb_rx_en++;
3056                 }
3057                 if (nb_rx_en > 1)
3058                         fccfg_reg |=IXGBE_FCCFG_TFCE_PRIORITY;
3059                 break;
3060         case ixgbe_fc_rx_pause:
3061                 /*
3062                  * Rx Flow control is enabled and Tx Flow control is
3063                  * disabled by software override. Since there really
3064                  * isn't a way to advertise that we are capable of RX
3065                  * Pause ONLY, we will advertise that we support both
3066                  * symmetric and asymmetric Rx PAUSE.  Later, we will
3067                  * disable the adapter's ability to send PAUSE frames.
3068                  */
3069                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
3070                 /*
3071                  * If the count of enabled RX Priority Flow control >1,
3072                  * and the TX pause can not be disabled
3073                  */
3074                 nb_rx_en = 0;
3075                 for (i =0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
3076                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
3077                         if (reg & IXGBE_FCRTH_FCEN)
3078                                 nb_rx_en++;
3079                 }
3080                 if (nb_rx_en > 1)
3081                         fccfg_reg |=IXGBE_FCCFG_TFCE_PRIORITY;
3082                 break;
3083         case ixgbe_fc_tx_pause:
3084                 /*
3085                  * Tx Flow control is enabled, and Rx Flow control is
3086                  * disabled by software override.
3087                  */
3088                 fccfg_reg |=IXGBE_FCCFG_TFCE_PRIORITY;
3089                 break;
3090         case ixgbe_fc_full:
3091                 /* Flow control (both Rx and Tx) is enabled by SW override. */
3092                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
3093                 fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
3094                 break;
3095         default:
3096                 PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
3097                 ret_val = IXGBE_ERR_CONFIG;
3098                 goto out;
3099                 break;
3100         }
3101
3102         /* Set 802.3x based flow control settings. */
3103         mflcn_reg |= IXGBE_MFLCN_DPF;
3104         IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
3105         IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
3106
3107         /* Set up and enable Rx high/low water mark thresholds, enable XON. */
3108         if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
3109                 hw->fc.high_water[tc_num]) {
3110                 fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
3111                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
3112                 fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
3113         } else {
3114                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
3115                 /*
3116                  * In order to prevent Tx hangs when the internal Tx
3117                  * switch is enabled we must set the high water mark
3118                  * to the maximum FCRTH value.  This allows the Tx
3119                  * switch to function even under heavy Rx workloads.
3120                  */
3121                 fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
3122         }
3123         IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
3124
3125         /* Configure pause time (2 TCs per register) */
3126         reg = hw->fc.pause_time * 0x00010001;
3127         for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
3128                 IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
3129
3130         /* Configure flow control refresh threshold value */
3131         IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
3132
3133 out:
3134         return ret_val;
3135 }
3136
3137 static int
3138 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev,uint8_t tc_num)
3139 {
3140         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3141         int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
3142
3143         if(hw->mac.type != ixgbe_mac_82598EB) {
3144                 ret_val = ixgbe_dcb_pfc_enable_generic(hw,tc_num);
3145         }
3146         return ret_val;
3147 }
3148
3149 static int
3150 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
3151 {
3152         int err;
3153         uint32_t rx_buf_size;
3154         uint32_t max_high_water;
3155         uint8_t tc_num;
3156         uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
3157         struct ixgbe_hw *hw =
3158                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3159         struct ixgbe_dcb_config *dcb_config =
3160                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
3161
3162         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
3163                 ixgbe_fc_none,
3164                 ixgbe_fc_rx_pause,
3165                 ixgbe_fc_tx_pause,
3166                 ixgbe_fc_full
3167         };
3168
3169         PMD_INIT_FUNC_TRACE();
3170
3171         ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
3172         tc_num = map[pfc_conf->priority];
3173         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
3174         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3175         /*
3176          * At least reserve one Ethernet frame for watermark
3177          * high_water/low_water in kilo bytes for ixgbe
3178          */
3179         max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
3180         if ((pfc_conf->fc.high_water > max_high_water) ||
3181             (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
3182                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
3183                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
3184                 return (-EINVAL);
3185         }
3186
3187         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
3188         hw->fc.pause_time = pfc_conf->fc.pause_time;
3189         hw->fc.send_xon = pfc_conf->fc.send_xon;
3190         hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
3191         hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
3192
3193         err = ixgbe_dcb_pfc_enable(dev,tc_num);
3194
3195         /* Not negotiated is not an error case */
3196         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
3197                 return 0;
3198
3199         PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
3200         return -EIO;
3201 }
3202
3203 static int
3204 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
3205                           struct rte_eth_rss_reta_entry64 *reta_conf,
3206                           uint16_t reta_size)
3207 {
3208         uint8_t i, j, mask;
3209         uint32_t reta, r;
3210         uint16_t idx, shift;
3211         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3212
3213         PMD_INIT_FUNC_TRACE();
3214         if (reta_size != ETH_RSS_RETA_SIZE_128) {
3215                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3216                         "(%d) doesn't match the number hardware can supported "
3217                         "(%d)\n", reta_size, ETH_RSS_RETA_SIZE_128);
3218                 return -EINVAL;
3219         }
3220
3221         for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
3222                 idx = i / RTE_RETA_GROUP_SIZE;
3223                 shift = i % RTE_RETA_GROUP_SIZE;
3224                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3225                                                 IXGBE_4_BIT_MASK);
3226                 if (!mask)
3227                         continue;
3228                 if (mask == IXGBE_4_BIT_MASK)
3229                         r = 0;
3230                 else
3231                         r = IXGBE_READ_REG(hw, IXGBE_RETA(i >> 2));
3232                 for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
3233                         if (mask & (0x1 << j))
3234                                 reta |= reta_conf[idx].reta[shift + j] <<
3235                                                         (CHAR_BIT * j);
3236                         else
3237                                 reta |= r & (IXGBE_8_BIT_MASK <<
3238                                                 (CHAR_BIT * j));
3239                 }
3240                 IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3241         }
3242
3243         return 0;
3244 }
3245
3246 static int
3247 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
3248                          struct rte_eth_rss_reta_entry64 *reta_conf,
3249                          uint16_t reta_size)
3250 {
3251         uint8_t i, j, mask;
3252         uint32_t reta;
3253         uint16_t idx, shift;
3254         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3255
3256         PMD_INIT_FUNC_TRACE();
3257         if (reta_size != ETH_RSS_RETA_SIZE_128) {
3258                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3259                         "(%d) doesn't match the number hardware can supported "
3260                                 "(%d)\n", reta_size, ETH_RSS_RETA_SIZE_128);
3261                 return -EINVAL;
3262         }
3263
3264         for (i = 0; i < ETH_RSS_RETA_SIZE_128; i += IXGBE_4_BIT_WIDTH) {
3265                 idx = i / RTE_RETA_GROUP_SIZE;
3266                 shift = i % RTE_RETA_GROUP_SIZE;
3267                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3268                                                 IXGBE_4_BIT_MASK);
3269                 if (!mask)
3270                         continue;
3271
3272                 reta = IXGBE_READ_REG(hw, IXGBE_RETA(i >> 2));
3273                 for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
3274                         if (mask & (0x1 << j))
3275                                 reta_conf[idx].reta[shift + j] =
3276                                         ((reta >> (CHAR_BIT * j)) &
3277                                                 IXGBE_8_BIT_MASK);
3278                 }
3279         }
3280
3281         return 0;
3282 }
3283
3284 static void
3285 ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
3286                                 uint32_t index, uint32_t pool)
3287 {
3288         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3289         uint32_t enable_addr = 1;
3290
3291         ixgbe_set_rar(hw, index, mac_addr->addr_bytes, pool, enable_addr);
3292 }
3293
3294 static void
3295 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
3296 {
3297         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3298
3299         ixgbe_clear_rar(hw, index);
3300 }
3301
3302 static void
3303 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
3304 {
3305         ixgbe_remove_rar(dev, 0);
3306
3307         ixgbe_add_rar(dev, addr, 0, 0);
3308 }
3309
3310 static int
3311 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
3312 {
3313         uint32_t hlreg0;
3314         uint32_t maxfrs;
3315         struct ixgbe_hw *hw;
3316         struct rte_eth_dev_info dev_info;
3317         uint32_t frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
3318
3319         ixgbe_dev_info_get(dev, &dev_info);
3320
3321         /* check that mtu is within the allowed range */
3322         if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
3323                 return -EINVAL;
3324
3325         /* refuse mtu that requires the support of scattered packets when this
3326          * feature has not been enabled before. */
3327         if (!dev->data->scattered_rx &&
3328             (frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
3329              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM))
3330                 return -EINVAL;
3331
3332         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3333         hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3334
3335         /* switch to jumbo mode if needed */
3336         if (frame_size > ETHER_MAX_LEN) {
3337                 dev->data->dev_conf.rxmode.jumbo_frame = 1;
3338                 hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3339         } else {
3340                 dev->data->dev_conf.rxmode.jumbo_frame = 0;
3341                 hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
3342         }
3343         IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3344
3345         /* update max frame size */
3346         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
3347
3348         maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
3349         maxfrs &= 0x0000FFFF;
3350         maxfrs |= (dev->data->dev_conf.rxmode.max_rx_pkt_len << 16);
3351         IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
3352
3353         return 0;
3354 }
3355
3356 /*
3357  * Virtual Function operations
3358  */
3359 static void
3360 ixgbevf_intr_disable(struct ixgbe_hw *hw)
3361 {
3362         PMD_INIT_FUNC_TRACE();
3363
3364         /* Clear interrupt mask to stop from interrupts being generated */
3365         IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
3366
3367         IXGBE_WRITE_FLUSH(hw);
3368 }
3369
3370 static void
3371 ixgbevf_intr_enable(struct ixgbe_hw *hw)
3372 {
3373         PMD_INIT_FUNC_TRACE();
3374
3375         /* VF enable interrupt autoclean */
3376         IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
3377         IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
3378         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
3379
3380         IXGBE_WRITE_FLUSH(hw);
3381 }
3382
3383 static int
3384 ixgbevf_dev_configure(struct rte_eth_dev *dev)
3385 {
3386         struct rte_eth_conf* conf = &dev->data->dev_conf;
3387         struct ixgbe_adapter *adapter =
3388                         (struct ixgbe_adapter *)dev->data->dev_private;
3389
3390         PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
3391                      dev->data->port_id);
3392
3393         /*
3394          * VF has no ability to enable/disable HW CRC
3395          * Keep the persistent behavior the same as Host PF
3396          */
3397 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
3398         if (!conf->rxmode.hw_strip_crc) {
3399                 PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
3400                 conf->rxmode.hw_strip_crc = 1;
3401         }
3402 #else
3403         if (conf->rxmode.hw_strip_crc) {
3404                 PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
3405                 conf->rxmode.hw_strip_crc = 0;
3406         }
3407 #endif
3408
3409         /*
3410          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
3411          * allocation or vector Rx preconditions we will reset it.
3412          */
3413         adapter->rx_bulk_alloc_allowed = true;
3414         adapter->rx_vec_allowed = true;
3415
3416         return 0;
3417 }
3418
3419 static int
3420 ixgbevf_dev_start(struct rte_eth_dev *dev)
3421 {
3422         struct ixgbe_hw *hw =
3423                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3424         uint32_t intr_vector = 0;
3425         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
3426
3427         int err, mask = 0;
3428
3429         PMD_INIT_FUNC_TRACE();
3430
3431         hw->mac.ops.reset_hw(hw);
3432         hw->mac.get_link_status = true;
3433
3434         /* negotiate mailbox API version to use with the PF. */
3435         ixgbevf_negotiate_api(hw);
3436
3437         ixgbevf_dev_tx_init(dev);
3438
3439         /* This can fail when allocating mbufs for descriptor rings */
3440         err = ixgbevf_dev_rx_init(dev);
3441         if (err) {
3442                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
3443                 ixgbe_dev_clear_queues(dev);
3444                 return err;
3445         }
3446
3447         /* Set vfta */
3448         ixgbevf_set_vfta_all(dev,1);
3449
3450         /* Set HW strip */
3451         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
3452                 ETH_VLAN_EXTEND_MASK;
3453         ixgbevf_vlan_offload_set(dev, mask);
3454
3455         ixgbevf_dev_rxtx_start(dev);
3456
3457         /* check and configure queue intr-vector mapping */
3458         if (dev->data->dev_conf.intr_conf.rxq != 0)
3459                 intr_vector = dev->data->nb_rx_queues;
3460
3461         if (rte_intr_efd_enable(intr_handle, intr_vector))
3462                 return -1;
3463
3464         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
3465                 intr_handle->intr_vec =
3466                         rte_zmalloc("intr_vec",
3467                                     dev->data->nb_rx_queues * sizeof(int), 0);
3468                 if (intr_handle->intr_vec == NULL) {
3469                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
3470                                      " intr_vec\n", dev->data->nb_rx_queues);
3471                         return -ENOMEM;
3472                 }
3473         }
3474         ixgbevf_configure_msix(dev);
3475
3476         if (dev->data->dev_conf.intr_conf.lsc != 0) {
3477                 if (rte_intr_allow_others(intr_handle))
3478                         rte_intr_callback_register(intr_handle,
3479                                         ixgbevf_dev_interrupt_handler,
3480                                         (void *)dev);
3481                 else
3482                         PMD_INIT_LOG(INFO, "lsc won't enable because of"
3483                                      " no intr multiplex\n");
3484         }
3485
3486         rte_intr_enable(intr_handle);
3487
3488         /* Re-enable interrupt for VF */
3489         ixgbevf_intr_enable(hw);
3490
3491         return 0;
3492 }
3493
3494 static void
3495 ixgbevf_dev_stop(struct rte_eth_dev *dev)
3496 {
3497         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3498         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
3499
3500         PMD_INIT_FUNC_TRACE();
3501
3502         hw->adapter_stopped = 1;
3503         ixgbe_stop_adapter(hw);
3504
3505         /*
3506           * Clear what we set, but we still keep shadow_vfta to
3507           * restore after device starts
3508           */
3509         ixgbevf_set_vfta_all(dev,0);
3510
3511         /* Clear stored conf */
3512         dev->data->scattered_rx = 0;
3513
3514         ixgbe_dev_clear_queues(dev);
3515
3516         /* disable intr eventfd mapping */
3517         rte_intr_disable(intr_handle);
3518
3519         /* Clean datapath event and queue/vec mapping */
3520         rte_intr_efd_disable(intr_handle);
3521         if (intr_handle->intr_vec != NULL) {
3522                 rte_free(intr_handle->intr_vec);
3523                 intr_handle->intr_vec = NULL;
3524         }
3525 }
3526
3527 static void
3528 ixgbevf_dev_close(struct rte_eth_dev *dev)
3529 {
3530         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3531         struct rte_pci_device *pci_dev;
3532
3533         PMD_INIT_FUNC_TRACE();
3534
3535         ixgbe_reset_hw(hw);
3536
3537         ixgbevf_dev_stop(dev);
3538
3539         ixgbe_dev_free_queues(dev);
3540
3541         /* reprogram the RAR[0] in case user changed it. */
3542         ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
3543
3544         pci_dev = dev->pci_dev;
3545         if (pci_dev->intr_handle.intr_vec) {
3546                 rte_free(pci_dev->intr_handle.intr_vec);
3547                 pci_dev->intr_handle.intr_vec = NULL;
3548         }
3549 }
3550
3551 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
3552 {
3553         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3554         struct ixgbe_vfta * shadow_vfta =
3555                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3556         int i = 0, j = 0, vfta = 0, mask = 1;
3557
3558         for (i = 0; i < IXGBE_VFTA_SIZE; i++){
3559                 vfta = shadow_vfta->vfta[i];
3560                 if(vfta){
3561                         mask = 1;
3562                         for (j = 0; j < 32; j++){
3563                                 if(vfta & mask)
3564                                         ixgbe_set_vfta(hw, (i<<5)+j, 0, on);
3565                                 mask<<=1;
3566                         }
3567                 }
3568         }
3569
3570 }
3571
3572 static int
3573 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
3574 {
3575         struct ixgbe_hw *hw =
3576                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3577         struct ixgbe_vfta * shadow_vfta =
3578                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3579         uint32_t vid_idx = 0;
3580         uint32_t vid_bit = 0;
3581         int ret = 0;
3582
3583         PMD_INIT_FUNC_TRACE();
3584
3585         /* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
3586         ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on);
3587         if(ret){
3588                 PMD_INIT_LOG(ERR, "Unable to set VF vlan");
3589                 return ret;
3590         }
3591         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
3592         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
3593
3594         /* Save what we set and retore it after device reset */
3595         if (on)
3596                 shadow_vfta->vfta[vid_idx] |= vid_bit;
3597         else
3598                 shadow_vfta->vfta[vid_idx] &= ~vid_bit;
3599
3600         return 0;
3601 }
3602
3603 static void
3604 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
3605 {
3606         struct ixgbe_hw *hw =
3607                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3608         uint32_t ctrl;
3609
3610         PMD_INIT_FUNC_TRACE();
3611
3612         if(queue >= hw->mac.max_rx_queues)
3613                 return;
3614
3615         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
3616         if(on)
3617                 ctrl |= IXGBE_RXDCTL_VME;
3618         else
3619                 ctrl &= ~IXGBE_RXDCTL_VME;
3620         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
3621
3622         ixgbe_vlan_hw_strip_bitmap_set( dev, queue, on);
3623 }
3624
3625 static void
3626 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
3627 {
3628         struct ixgbe_hw *hw =
3629                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3630         uint16_t i;
3631         int on = 0;
3632
3633         /* VF function only support hw strip feature, others are not support */
3634         if(mask & ETH_VLAN_STRIP_MASK){
3635                 on = !!(dev->data->dev_conf.rxmode.hw_vlan_strip);
3636
3637                 for(i=0; i < hw->mac.max_rx_queues; i++)
3638                         ixgbevf_vlan_strip_queue_set(dev,i,on);
3639         }
3640 }
3641
3642 static int
3643 ixgbe_vmdq_mode_check(struct ixgbe_hw *hw)
3644 {
3645         uint32_t reg_val;
3646
3647         /* we only need to do this if VMDq is enabled */
3648         reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3649         if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
3650                 PMD_INIT_LOG(ERR, "VMDq must be enabled for this setting");
3651                 return (-1);
3652         }
3653
3654         return 0;
3655 }
3656
3657 static uint32_t
3658 ixgbe_uta_vector(struct ixgbe_hw *hw, struct ether_addr* uc_addr)
3659 {
3660         uint32_t vector = 0;
3661         switch (hw->mac.mc_filter_type) {
3662         case 0:   /* use bits [47:36] of the address */
3663                 vector = ((uc_addr->addr_bytes[4] >> 4) |
3664                         (((uint16_t)uc_addr->addr_bytes[5]) << 4));
3665                 break;
3666         case 1:   /* use bits [46:35] of the address */
3667                 vector = ((uc_addr->addr_bytes[4] >> 3) |
3668                         (((uint16_t)uc_addr->addr_bytes[5]) << 5));
3669                 break;
3670         case 2:   /* use bits [45:34] of the address */
3671                 vector = ((uc_addr->addr_bytes[4] >> 2) |
3672                         (((uint16_t)uc_addr->addr_bytes[5]) << 6));
3673                 break;
3674         case 3:   /* use bits [43:32] of the address */
3675                 vector = ((uc_addr->addr_bytes[4]) |
3676                         (((uint16_t)uc_addr->addr_bytes[5]) << 8));
3677                 break;
3678         default:  /* Invalid mc_filter_type */
3679                 break;
3680         }
3681
3682         /* vector can only be 12-bits or boundary will be exceeded */
3683         vector &= 0xFFF;
3684         return vector;
3685 }
3686
3687 static int
3688 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev,struct ether_addr* mac_addr,
3689                                uint8_t on)
3690 {
3691         uint32_t vector;
3692         uint32_t uta_idx;
3693         uint32_t reg_val;
3694         uint32_t uta_shift;
3695         uint32_t rc;
3696         const uint32_t ixgbe_uta_idx_mask = 0x7F;
3697         const uint32_t ixgbe_uta_bit_shift = 5;
3698         const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
3699         const uint32_t bit1 = 0x1;
3700
3701         struct ixgbe_hw *hw =
3702                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3703         struct ixgbe_uta_info *uta_info =
3704                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
3705
3706         /* The UTA table only exists on 82599 hardware and newer */
3707         if (hw->mac.type < ixgbe_mac_82599EB)
3708                 return (-ENOTSUP);
3709
3710         vector = ixgbe_uta_vector(hw,mac_addr);
3711         uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
3712         uta_shift = vector & ixgbe_uta_bit_mask;
3713
3714         rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
3715         if(rc == on)
3716                 return 0;
3717
3718         reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
3719         if (on) {
3720                 uta_info->uta_in_use++;
3721                 reg_val |= (bit1 << uta_shift);
3722                 uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
3723         } else {
3724                 uta_info->uta_in_use--;
3725                 reg_val &= ~(bit1 << uta_shift);
3726                 uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
3727         }
3728
3729         IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
3730
3731         if (uta_info->uta_in_use > 0)
3732                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
3733                                 IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
3734         else
3735                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,hw->mac.mc_filter_type);
3736
3737         return 0;
3738 }
3739
3740 static int
3741 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
3742 {
3743         int i;
3744         struct ixgbe_hw *hw =
3745                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3746         struct ixgbe_uta_info *uta_info =
3747                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
3748
3749         /* The UTA table only exists on 82599 hardware and newer */
3750         if (hw->mac.type < ixgbe_mac_82599EB)
3751                 return (-ENOTSUP);
3752
3753         if(on) {
3754                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
3755                         uta_info->uta_shadow[i] = ~0;
3756                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
3757                 }
3758         } else {
3759                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
3760                         uta_info->uta_shadow[i] = 0;
3761                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
3762                 }
3763         }
3764         return 0;
3765
3766 }
3767
3768 uint32_t
3769 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
3770 {
3771         uint32_t new_val = orig_val;
3772
3773         if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG)
3774                 new_val |= IXGBE_VMOLR_AUPE;
3775         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC)
3776                 new_val |= IXGBE_VMOLR_ROMPE;
3777         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
3778                 new_val |= IXGBE_VMOLR_ROPE;
3779         if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
3780                 new_val |= IXGBE_VMOLR_BAM;
3781         if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
3782                 new_val |= IXGBE_VMOLR_MPE;
3783
3784         return new_val;
3785 }
3786
3787 static int
3788 ixgbe_set_pool_rx_mode(struct rte_eth_dev *dev, uint16_t pool,
3789                                uint16_t rx_mask, uint8_t on)
3790 {
3791         int val = 0;
3792
3793         struct ixgbe_hw *hw =
3794                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3795         uint32_t vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
3796
3797         if (hw->mac.type == ixgbe_mac_82598EB) {
3798                 PMD_INIT_LOG(ERR, "setting VF receive mode set should be done"
3799                              " on 82599 hardware and newer");
3800                 return (-ENOTSUP);
3801         }
3802         if (ixgbe_vmdq_mode_check(hw) < 0)
3803                 return (-ENOTSUP);
3804
3805         val = ixgbe_convert_vm_rx_mask_to_val(rx_mask, val);
3806
3807         if (on)
3808                 vmolr |= val;
3809         else
3810                 vmolr &= ~val;
3811
3812         IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
3813
3814         return 0;
3815 }
3816
3817 static int
3818 ixgbe_set_pool_rx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on)
3819 {
3820         uint32_t reg,addr;
3821         uint32_t val;
3822         const uint8_t bit1 = 0x1;
3823
3824         struct ixgbe_hw *hw =
3825                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3826
3827         if (ixgbe_vmdq_mode_check(hw) < 0)
3828                 return (-ENOTSUP);
3829
3830         addr = IXGBE_VFRE(pool >= ETH_64_POOLS/2);
3831         reg = IXGBE_READ_REG(hw, addr);
3832         val = bit1 << pool;
3833
3834         if (on)
3835                 reg |= val;
3836         else
3837                 reg &= ~val;
3838
3839         IXGBE_WRITE_REG(hw, addr,reg);
3840
3841         return 0;
3842 }
3843
3844 static int
3845 ixgbe_set_pool_tx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on)
3846 {
3847         uint32_t reg,addr;
3848         uint32_t val;
3849         const uint8_t bit1 = 0x1;
3850
3851         struct ixgbe_hw *hw =
3852                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3853
3854         if (ixgbe_vmdq_mode_check(hw) < 0)
3855                 return (-ENOTSUP);
3856
3857         addr = IXGBE_VFTE(pool >= ETH_64_POOLS/2);
3858         reg = IXGBE_READ_REG(hw, addr);
3859         val = bit1 << pool;
3860
3861         if (on)
3862                 reg |= val;
3863         else
3864                 reg &= ~val;
3865
3866         IXGBE_WRITE_REG(hw, addr,reg);
3867
3868         return 0;
3869 }
3870
3871 static int
3872 ixgbe_set_pool_vlan_filter(struct rte_eth_dev *dev, uint16_t vlan,
3873                         uint64_t pool_mask, uint8_t vlan_on)
3874 {
3875         int ret = 0;
3876         uint16_t pool_idx;
3877         struct ixgbe_hw *hw =
3878                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3879
3880         if (ixgbe_vmdq_mode_check(hw) < 0)
3881                 return (-ENOTSUP);
3882         for (pool_idx = 0; pool_idx < ETH_64_POOLS; pool_idx++) {
3883                 if (pool_mask & ((uint64_t)(1ULL << pool_idx)))
3884                         ret = hw->mac.ops.set_vfta(hw,vlan,pool_idx,vlan_on);
3885                         if (ret < 0)
3886                                 return ret;
3887         }
3888
3889         return ret;
3890 }
3891
3892 #define IXGBE_MRCTL_VPME  0x01 /* Virtual Pool Mirroring. */
3893 #define IXGBE_MRCTL_UPME  0x02 /* Uplink Port Mirroring. */
3894 #define IXGBE_MRCTL_DPME  0x04 /* Downlink Port Mirroring. */
3895 #define IXGBE_MRCTL_VLME  0x08 /* VLAN Mirroring. */
3896 #define IXGBE_INVALID_MIRROR_TYPE(mirror_type) \
3897         ((mirror_type) & ~(uint8_t)(ETH_MIRROR_VIRTUAL_POOL_UP | \
3898         ETH_MIRROR_UPLINK_PORT | ETH_MIRROR_DOWNLINK_PORT | ETH_MIRROR_VLAN))
3899
3900 static int
3901 ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
3902                         struct rte_eth_mirror_conf *mirror_conf,
3903                         uint8_t rule_id, uint8_t on)
3904 {
3905         uint32_t mr_ctl,vlvf;
3906         uint32_t mp_lsb = 0;
3907         uint32_t mv_msb = 0;
3908         uint32_t mv_lsb = 0;
3909         uint32_t mp_msb = 0;
3910         uint8_t i = 0;
3911         int reg_index = 0;
3912         uint64_t vlan_mask = 0;
3913
3914         const uint8_t pool_mask_offset = 32;
3915         const uint8_t vlan_mask_offset = 32;
3916         const uint8_t dst_pool_offset = 8;
3917         const uint8_t rule_mr_offset  = 4;
3918         const uint8_t mirror_rule_mask= 0x0F;
3919
3920         struct ixgbe_mirror_info *mr_info =
3921                         (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
3922         struct ixgbe_hw *hw =
3923                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3924         uint8_t mirror_type = 0;
3925
3926         if (ixgbe_vmdq_mode_check(hw) < 0)
3927                 return -ENOTSUP;
3928
3929         if (rule_id >= IXGBE_MAX_MIRROR_RULES)
3930                 return -EINVAL;
3931
3932         if (IXGBE_INVALID_MIRROR_TYPE(mirror_conf->rule_type)) {
3933                 PMD_DRV_LOG(ERR, "unsupported mirror type 0x%x.",
3934                         mirror_conf->rule_type);
3935                 return -EINVAL;
3936         }
3937
3938         if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
3939                 mirror_type |= IXGBE_MRCTL_VLME;
3940                 /* Check if vlan id is valid and find conresponding VLAN ID index in VLVF */
3941                 for (i = 0;i < IXGBE_VLVF_ENTRIES; i++) {
3942                         if (mirror_conf->vlan.vlan_mask & (1ULL << i)) {
3943                                 /* search vlan id related pool vlan filter index */
3944                                 reg_index = ixgbe_find_vlvf_slot(hw,
3945                                                 mirror_conf->vlan.vlan_id[i]);
3946                                 if(reg_index < 0)
3947                                         return -EINVAL;
3948                                 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(reg_index));
3949                                 if ((vlvf & IXGBE_VLVF_VIEN) &&
3950                                     ((vlvf & IXGBE_VLVF_VLANID_MASK) ==
3951                                       mirror_conf->vlan.vlan_id[i]))
3952                                         vlan_mask |= (1ULL << reg_index);
3953                                 else
3954                                         return -EINVAL;
3955                         }
3956                 }
3957
3958                 if (on) {
3959                         mv_lsb = vlan_mask & 0xFFFFFFFF;
3960                         mv_msb = vlan_mask >> vlan_mask_offset;
3961
3962                         mr_info->mr_conf[rule_id].vlan.vlan_mask =
3963                                                 mirror_conf->vlan.vlan_mask;
3964                         for(i = 0 ;i < ETH_VMDQ_MAX_VLAN_FILTERS; i++) {
3965                                 if(mirror_conf->vlan.vlan_mask & (1ULL << i))
3966                                         mr_info->mr_conf[rule_id].vlan.vlan_id[i] =
3967                                                 mirror_conf->vlan.vlan_id[i];
3968                         }
3969                 } else {
3970                         mv_lsb = 0;
3971                         mv_msb = 0;
3972                         mr_info->mr_conf[rule_id].vlan.vlan_mask = 0;
3973                         for(i = 0 ;i < ETH_VMDQ_MAX_VLAN_FILTERS; i++)
3974                                 mr_info->mr_conf[rule_id].vlan.vlan_id[i] = 0;
3975                 }
3976         }
3977
3978         /*
3979          * if enable pool mirror, write related pool mask register,if disable
3980          * pool mirror, clear PFMRVM register
3981          */
3982         if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
3983                 mirror_type |= IXGBE_MRCTL_VPME;
3984                 if (on) {
3985                         mp_lsb = mirror_conf->pool_mask & 0xFFFFFFFF;
3986                         mp_msb = mirror_conf->pool_mask >> pool_mask_offset;
3987                         mr_info->mr_conf[rule_id].pool_mask =
3988                                         mirror_conf->pool_mask;
3989
3990                 } else {
3991                         mp_lsb = 0;
3992                         mp_msb = 0;
3993                         mr_info->mr_conf[rule_id].pool_mask = 0;
3994                 }
3995         }
3996         if (mirror_conf->rule_type & ETH_MIRROR_UPLINK_PORT)
3997                 mirror_type |= IXGBE_MRCTL_UPME;
3998         if (mirror_conf->rule_type & ETH_MIRROR_DOWNLINK_PORT)
3999                 mirror_type |= IXGBE_MRCTL_DPME;
4000
4001         /* read  mirror control register and recalculate it */
4002         mr_ctl = IXGBE_READ_REG(hw, IXGBE_MRCTL(rule_id));
4003
4004         if (on) {
4005                 mr_ctl |= mirror_type;
4006                 mr_ctl &= mirror_rule_mask;
4007                 mr_ctl |= mirror_conf->dst_pool << dst_pool_offset;
4008         } else
4009                 mr_ctl &= ~(mirror_conf->rule_type & mirror_rule_mask);
4010
4011         mr_info->mr_conf[rule_id].rule_type = mirror_conf->rule_type;
4012         mr_info->mr_conf[rule_id].dst_pool = mirror_conf->dst_pool;
4013
4014         /* write mirrror control  register */
4015         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
4016
4017         /* write pool mirrror control  register */
4018         if (mirror_conf->rule_type == ETH_MIRROR_VIRTUAL_POOL_UP) {
4019                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), mp_lsb);
4020                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset),
4021                                 mp_msb);
4022         }
4023         /* write VLAN mirrror control  register */
4024         if (mirror_conf->rule_type == ETH_MIRROR_VLAN) {
4025                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), mv_lsb);
4026                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset),
4027                                 mv_msb);
4028         }
4029
4030         return 0;
4031 }
4032
4033 static int
4034 ixgbe_mirror_rule_reset(struct rte_eth_dev *dev, uint8_t rule_id)
4035 {
4036         int mr_ctl = 0;
4037         uint32_t lsb_val = 0;
4038         uint32_t msb_val = 0;
4039         const uint8_t rule_mr_offset = 4;
4040
4041         struct ixgbe_hw *hw =
4042                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4043         struct ixgbe_mirror_info *mr_info =
4044                 (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
4045
4046         if (ixgbe_vmdq_mode_check(hw) < 0)
4047                 return (-ENOTSUP);
4048
4049         memset(&mr_info->mr_conf[rule_id], 0,
4050                 sizeof(struct rte_eth_mirror_conf));
4051
4052         /* clear PFVMCTL register */
4053         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
4054
4055         /* clear pool mask register */
4056         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), lsb_val);
4057         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset), msb_val);
4058
4059         /* clear vlan mask register */
4060         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), lsb_val);
4061         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset), msb_val);
4062
4063         return 0;
4064 }
4065
4066 static int
4067 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
4068 {
4069         uint32_t mask;
4070         struct ixgbe_hw *hw =
4071                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4072
4073         mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
4074         mask |= (1 << queue_id);
4075         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
4076
4077         rte_intr_enable(&dev->pci_dev->intr_handle);
4078
4079         return 0;
4080 }
4081
4082 static int
4083 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
4084 {
4085         uint32_t mask;
4086         struct ixgbe_hw *hw =
4087                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4088
4089         mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
4090         mask &= ~(1 << queue_id);
4091         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
4092
4093         return 0;
4094 }
4095
4096 static int
4097 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
4098 {
4099         uint32_t mask;
4100         struct ixgbe_hw *hw =
4101                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4102         struct ixgbe_interrupt *intr =
4103                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4104
4105         if (queue_id < 16) {
4106                 ixgbe_disable_intr(hw);
4107                 intr->mask |= (1 << queue_id);
4108                 ixgbe_enable_intr(dev);
4109         } else if (queue_id < 32) {
4110                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
4111                 mask &= (1 << queue_id);
4112                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
4113         } else if (queue_id < 64) {
4114                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
4115                 mask &= (1 << (queue_id - 32));
4116                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
4117         }
4118         rte_intr_enable(&dev->pci_dev->intr_handle);
4119
4120         return 0;
4121 }
4122
4123 static int
4124 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
4125 {
4126         uint32_t mask;
4127         struct ixgbe_hw *hw =
4128                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4129         struct ixgbe_interrupt *intr =
4130                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4131
4132         if (queue_id < 16) {
4133                 ixgbe_disable_intr(hw);
4134                 intr->mask &= ~(1 << queue_id);
4135                 ixgbe_enable_intr(dev);
4136         } else if (queue_id < 32) {
4137                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
4138                 mask &= ~(1 << queue_id);
4139                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
4140         } else if (queue_id < 64) {
4141                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
4142                 mask &= ~(1 << (queue_id - 32));
4143                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
4144         }
4145
4146         return 0;
4147 }
4148
4149 static void
4150 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
4151                      uint8_t queue, uint8_t msix_vector)
4152 {
4153         uint32_t tmp, idx;
4154
4155         if (direction == -1) {
4156                 /* other causes */
4157                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
4158                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
4159                 tmp &= ~0xFF;
4160                 tmp |= msix_vector;
4161                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
4162         } else {
4163                 /* rx or tx cause */
4164                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
4165                 idx = ((16 * (queue & 1)) + (8 * direction));
4166                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
4167                 tmp &= ~(0xFF << idx);
4168                 tmp |= (msix_vector << idx);
4169                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
4170         }
4171 }
4172
4173 /**
4174  * set the IVAR registers, mapping interrupt causes to vectors
4175  * @param hw
4176  *  pointer to ixgbe_hw struct
4177  * @direction
4178  *  0 for Rx, 1 for Tx, -1 for other causes
4179  * @queue
4180  *  queue to map the corresponding interrupt to
4181  * @msix_vector
4182  *  the vector to map to the corresponding queue
4183  */
4184 static void
4185 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
4186                    uint8_t queue, uint8_t msix_vector)
4187 {
4188         uint32_t tmp, idx;
4189
4190         msix_vector |= IXGBE_IVAR_ALLOC_VAL;
4191         if (hw->mac.type == ixgbe_mac_82598EB) {
4192                 if (direction == -1)
4193                         direction = 0;
4194                 idx = (((direction * 64) + queue) >> 2) & 0x1F;
4195                 tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
4196                 tmp &= ~(0xFF << (8 * (queue & 0x3)));
4197                 tmp |= (msix_vector << (8 * (queue & 0x3)));
4198                 IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
4199         } else if ((hw->mac.type == ixgbe_mac_82599EB) ||
4200                         (hw->mac.type == ixgbe_mac_X540)) {
4201                 if (direction == -1) {
4202                         /* other causes */
4203                         idx = ((queue & 1) * 8);
4204                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
4205                         tmp &= ~(0xFF << idx);
4206                         tmp |= (msix_vector << idx);
4207                         IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
4208                 } else {
4209                         /* rx or tx causes */
4210                         idx = ((16 * (queue & 1)) + (8 * direction));
4211                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
4212                         tmp &= ~(0xFF << idx);
4213                         tmp |= (msix_vector << idx);
4214                         IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
4215                 }
4216         }
4217 }
4218
4219 static void
4220 ixgbevf_configure_msix(struct rte_eth_dev *dev)
4221 {
4222         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
4223         struct ixgbe_hw *hw =
4224                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4225         uint32_t q_idx;
4226         uint32_t vector_idx = 0;
4227
4228         /* won't configure msix register if no mapping is done
4229          * between intr vector and event fd.
4230          */
4231         if (!rte_intr_dp_is_en(intr_handle))
4232                 return;
4233
4234         /* Configure all RX queues of VF */
4235         for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
4236                 /* Force all queue use vector 0,
4237                  * as IXGBE_VF_MAXMSIVECOTR = 1
4238                  */
4239                 ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
4240                 intr_handle->intr_vec[q_idx] = vector_idx;
4241         }
4242
4243         /* Configure VF Rx queue ivar */
4244         ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
4245 }
4246
4247 /**
4248  * Sets up the hardware to properly generate MSI-X interrupts
4249  * @hw
4250  *  board private structure
4251  */
4252 static void
4253 ixgbe_configure_msix(struct rte_eth_dev *dev)
4254 {
4255         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
4256         struct ixgbe_hw *hw =
4257                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4258         uint32_t queue_id, vec = 0;
4259         uint32_t mask;
4260         uint32_t gpie;
4261
4262         /* won't configure msix register if no mapping is done
4263          * between intr vector and event fd
4264          */
4265         if (!rte_intr_dp_is_en(intr_handle))
4266                 return;
4267
4268         /* setup GPIE for MSI-x mode */
4269         gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
4270         gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
4271                 IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
4272         /* auto clearing and auto setting corresponding bits in EIMS
4273          * when MSI-X interrupt is triggered
4274          */
4275         if (hw->mac.type == ixgbe_mac_82598EB) {
4276                 IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4277         } else {
4278                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4279                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4280         }
4281         IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
4282
4283         /* Populate the IVAR table and set the ITR values to the
4284          * corresponding register.
4285          */
4286         for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
4287              queue_id++) {
4288                 /* by default, 1:1 mapping */
4289                 ixgbe_set_ivar_map(hw, 0, queue_id, vec);
4290                 intr_handle->intr_vec[queue_id] = vec;
4291                 if (vec < intr_handle->nb_efd - 1)
4292                         vec++;
4293         }
4294
4295         switch (hw->mac.type) {
4296         case ixgbe_mac_82598EB:
4297                 ixgbe_set_ivar_map(hw, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
4298                                    intr_handle->max_intr - 1);
4299                 break;
4300         case ixgbe_mac_82599EB:
4301         case ixgbe_mac_X540:
4302                 ixgbe_set_ivar_map(hw, -1, 1, intr_handle->max_intr - 1);
4303                 break;
4304         default:
4305                 break;
4306         }
4307         IXGBE_WRITE_REG(hw, IXGBE_EITR(queue_id),
4308                         IXGBE_MIN_INTER_INTERRUPT_INTERVAL_DEFAULT & 0xFFF);
4309
4310         /* set up to autoclear timer, and the vectors */
4311         mask = IXGBE_EIMS_ENABLE_MASK;
4312         mask &= ~(IXGBE_EIMS_OTHER |
4313                   IXGBE_EIMS_MAILBOX |
4314                   IXGBE_EIMS_LSC);
4315
4316         IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
4317 }
4318
4319 static int ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
4320         uint16_t queue_idx, uint16_t tx_rate)
4321 {
4322         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4323         uint32_t rf_dec, rf_int;
4324         uint32_t bcnrc_val;
4325         uint16_t link_speed = dev->data->dev_link.link_speed;
4326
4327         if (queue_idx >= hw->mac.max_tx_queues)
4328                 return -EINVAL;
4329
4330         if (tx_rate != 0) {
4331                 /* Calculate the rate factor values to set */
4332                 rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
4333                 rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
4334                 rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
4335
4336                 bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
4337                 bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
4338                                 IXGBE_RTTBCNRC_RF_INT_MASK_M);
4339                 bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
4340         } else {
4341                 bcnrc_val = 0;
4342         }
4343
4344         /*
4345          * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
4346          * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
4347          * set as 0x4.
4348          */
4349         if ((dev->data->dev_conf.rxmode.jumbo_frame == 1) &&
4350                 (dev->data->dev_conf.rxmode.max_rx_pkt_len >=
4351                                 IXGBE_MAX_JUMBO_FRAME_SIZE))
4352                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
4353                         IXGBE_MMW_SIZE_JUMBO_FRAME);
4354         else
4355                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
4356                         IXGBE_MMW_SIZE_DEFAULT);
4357
4358         /* Set RTTBCNRC of queue X */
4359         IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
4360         IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
4361         IXGBE_WRITE_FLUSH(hw);
4362
4363         return 0;
4364 }
4365
4366 static int ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
4367         uint16_t tx_rate, uint64_t q_msk)
4368 {
4369         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4370         struct ixgbe_vf_info *vfinfo =
4371                 *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
4372         uint8_t  nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
4373         uint32_t queue_stride =
4374                 IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
4375         uint32_t queue_idx = vf * queue_stride, idx = 0, vf_idx;
4376         uint32_t queue_end = queue_idx + nb_q_per_pool - 1;
4377         uint16_t total_rate = 0;
4378
4379         if (queue_end >= hw->mac.max_tx_queues)
4380                 return -EINVAL;
4381
4382         if (vfinfo != NULL) {
4383                 for (vf_idx = 0; vf_idx < dev->pci_dev->max_vfs; vf_idx++) {
4384                         if (vf_idx == vf)
4385                                 continue;
4386                         for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
4387                                 idx++)
4388                                 total_rate += vfinfo[vf_idx].tx_rate[idx];
4389                 }
4390         } else
4391                 return -EINVAL;
4392
4393         /* Store tx_rate for this vf. */
4394         for (idx = 0; idx < nb_q_per_pool; idx++) {
4395                 if (((uint64_t)0x1 << idx) & q_msk) {
4396                         if (vfinfo[vf].tx_rate[idx] != tx_rate)
4397                                 vfinfo[vf].tx_rate[idx] = tx_rate;
4398                         total_rate += tx_rate;
4399                 }
4400         }
4401
4402         if (total_rate > dev->data->dev_link.link_speed) {
4403                 /*
4404                  * Reset stored TX rate of the VF if it causes exceed
4405                  * link speed.
4406                  */
4407                 memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
4408                 return -EINVAL;
4409         }
4410
4411         /* Set RTTBCNRC of each queue/pool for vf X  */
4412         for (; queue_idx <= queue_end; queue_idx++) {
4413                 if (0x1 & q_msk)
4414                         ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
4415                 q_msk = q_msk >> 1;
4416         }
4417
4418         return 0;
4419 }
4420
4421 static void
4422 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
4423                      __attribute__((unused)) uint32_t index,
4424                      __attribute__((unused)) uint32_t pool)
4425 {
4426         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4427         int diag;
4428
4429         /*
4430          * On a 82599 VF, adding again the same MAC addr is not an idempotent
4431          * operation. Trap this case to avoid exhausting the [very limited]
4432          * set of PF resources used to store VF MAC addresses.
4433          */
4434         if (memcmp(hw->mac.perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
4435                 return;
4436         diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
4437         if (diag == 0)
4438                 return;
4439         PMD_DRV_LOG(ERR, "Unable to add MAC address - diag=%d", diag);
4440 }
4441
4442 static void
4443 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
4444 {
4445         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4446         struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
4447         struct ether_addr *mac_addr;
4448         uint32_t i;
4449         int diag;
4450
4451         /*
4452          * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
4453          * not support the deletion of a given MAC address.
4454          * Instead, it imposes to delete all MAC addresses, then to add again
4455          * all MAC addresses with the exception of the one to be deleted.
4456          */
4457         (void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
4458
4459         /*
4460          * Add again all MAC addresses, with the exception of the deleted one
4461          * and of the permanent MAC address.
4462          */
4463         for (i = 0, mac_addr = dev->data->mac_addrs;
4464              i < hw->mac.num_rar_entries; i++, mac_addr++) {
4465                 /* Skip the deleted MAC address */
4466                 if (i == index)
4467                         continue;
4468                 /* Skip NULL MAC addresses */
4469                 if (is_zero_ether_addr(mac_addr))
4470                         continue;
4471                 /* Skip the permanent MAC address */
4472                 if (memcmp(perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
4473                         continue;
4474                 diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
4475                 if (diag != 0)
4476                         PMD_DRV_LOG(ERR,
4477                                     "Adding again MAC address "
4478                                     "%02x:%02x:%02x:%02x:%02x:%02x failed "
4479                                     "diag=%d",
4480                                     mac_addr->addr_bytes[0],
4481                                     mac_addr->addr_bytes[1],
4482                                     mac_addr->addr_bytes[2],
4483                                     mac_addr->addr_bytes[3],
4484                                     mac_addr->addr_bytes[4],
4485                                     mac_addr->addr_bytes[5],
4486                                     diag);
4487         }
4488 }
4489
4490 static void
4491 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
4492 {
4493         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4494
4495         hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
4496 }
4497
4498 #define MAC_TYPE_FILTER_SUP(type)    do {\
4499         if ((type) != ixgbe_mac_82599EB && (type) != ixgbe_mac_X540 &&\
4500                 (type) != ixgbe_mac_X550)\
4501                 return -ENOTSUP;\
4502 } while (0)
4503
4504 static int
4505 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
4506                         struct rte_eth_syn_filter *filter,
4507                         bool add)
4508 {
4509         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4510         uint32_t synqf;
4511
4512         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
4513                 return -EINVAL;
4514
4515         synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
4516
4517         if (add) {
4518                 if (synqf & IXGBE_SYN_FILTER_ENABLE)
4519                         return -EINVAL;
4520                 synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
4521                         IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
4522
4523                 if (filter->hig_pri)
4524                         synqf |= IXGBE_SYN_FILTER_SYNQFP;
4525                 else
4526                         synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
4527         } else {
4528                 if (!(synqf & IXGBE_SYN_FILTER_ENABLE))
4529                         return -ENOENT;
4530                 synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
4531         }
4532         IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
4533         IXGBE_WRITE_FLUSH(hw);
4534         return 0;
4535 }
4536
4537 static int
4538 ixgbe_syn_filter_get(struct rte_eth_dev *dev,
4539                         struct rte_eth_syn_filter *filter)
4540 {
4541         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4542         uint32_t synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
4543
4544         if (synqf & IXGBE_SYN_FILTER_ENABLE) {
4545                 filter->hig_pri = (synqf & IXGBE_SYN_FILTER_SYNQFP) ? 1 : 0;
4546                 filter->queue = (uint16_t)((synqf & IXGBE_SYN_FILTER_QUEUE) >> 1);
4547                 return 0;
4548         }
4549         return -ENOENT;
4550 }
4551
4552 static int
4553 ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
4554                         enum rte_filter_op filter_op,
4555                         void *arg)
4556 {
4557         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4558         int ret;
4559
4560         MAC_TYPE_FILTER_SUP(hw->mac.type);
4561
4562         if (filter_op == RTE_ETH_FILTER_NOP)
4563                 return 0;
4564
4565         if (arg == NULL) {
4566                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
4567                             filter_op);
4568                 return -EINVAL;
4569         }
4570
4571         switch (filter_op) {
4572         case RTE_ETH_FILTER_ADD:
4573                 ret = ixgbe_syn_filter_set(dev,
4574                                 (struct rte_eth_syn_filter *)arg,
4575                                 TRUE);
4576                 break;
4577         case RTE_ETH_FILTER_DELETE:
4578                 ret = ixgbe_syn_filter_set(dev,
4579                                 (struct rte_eth_syn_filter *)arg,
4580                                 FALSE);
4581                 break;
4582         case RTE_ETH_FILTER_GET:
4583                 ret = ixgbe_syn_filter_get(dev,
4584                                 (struct rte_eth_syn_filter *)arg);
4585                 break;
4586         default:
4587                 PMD_DRV_LOG(ERR, "unsupported operation %u\n", filter_op);
4588                 ret = -EINVAL;
4589                 break;
4590         }
4591
4592         return ret;
4593 }
4594
4595
4596 static inline enum ixgbe_5tuple_protocol
4597 convert_protocol_type(uint8_t protocol_value)
4598 {
4599         if (protocol_value == IPPROTO_TCP)
4600                 return IXGBE_FILTER_PROTOCOL_TCP;
4601         else if (protocol_value == IPPROTO_UDP)
4602                 return IXGBE_FILTER_PROTOCOL_UDP;
4603         else if (protocol_value == IPPROTO_SCTP)
4604                 return IXGBE_FILTER_PROTOCOL_SCTP;
4605         else
4606                 return IXGBE_FILTER_PROTOCOL_NONE;
4607 }
4608
4609 /*
4610  * add a 5tuple filter
4611  *
4612  * @param
4613  * dev: Pointer to struct rte_eth_dev.
4614  * index: the index the filter allocates.
4615  * filter: ponter to the filter that will be added.
4616  * rx_queue: the queue id the filter assigned to.
4617  *
4618  * @return
4619  *    - On success, zero.
4620  *    - On failure, a negative value.
4621  */
4622 static int
4623 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
4624                         struct ixgbe_5tuple_filter *filter)
4625 {
4626         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4627         struct ixgbe_filter_info *filter_info =
4628                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4629         int i, idx, shift;
4630         uint32_t ftqf, sdpqf;
4631         uint32_t l34timir = 0;
4632         uint8_t mask = 0xff;
4633
4634         /*
4635          * look for an unused 5tuple filter index,
4636          * and insert the filter to list.
4637          */
4638         for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
4639                 idx = i / (sizeof(uint32_t) * NBBY);
4640                 shift = i % (sizeof(uint32_t) * NBBY);
4641                 if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
4642                         filter_info->fivetuple_mask[idx] |= 1 << shift;
4643                         filter->index = i;
4644                         TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
4645                                           filter,
4646                                           entries);
4647                         break;
4648                 }
4649         }
4650         if (i >= IXGBE_MAX_FTQF_FILTERS) {
4651                 PMD_DRV_LOG(ERR, "5tuple filters are full.");
4652                 return -ENOSYS;
4653         }
4654
4655         sdpqf = (uint32_t)(filter->filter_info.dst_port <<
4656                                 IXGBE_SDPQF_DSTPORT_SHIFT);
4657         sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
4658
4659         ftqf = (uint32_t)(filter->filter_info.proto &
4660                 IXGBE_FTQF_PROTOCOL_MASK);
4661         ftqf |= (uint32_t)((filter->filter_info.priority &
4662                 IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
4663         if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
4664                 mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
4665         if (filter->filter_info.dst_ip_mask == 0)
4666                 mask &= IXGBE_FTQF_DEST_ADDR_MASK;
4667         if (filter->filter_info.src_port_mask == 0)
4668                 mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
4669         if (filter->filter_info.dst_port_mask == 0)
4670                 mask &= IXGBE_FTQF_DEST_PORT_MASK;
4671         if (filter->filter_info.proto_mask == 0)
4672                 mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
4673         ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
4674         ftqf |= IXGBE_FTQF_POOL_MASK_EN;
4675         ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
4676
4677         IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
4678         IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
4679         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
4680         IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
4681
4682         l34timir |= IXGBE_L34T_IMIR_RESERVE;
4683         l34timir |= (uint32_t)(filter->queue <<
4684                                 IXGBE_L34T_IMIR_QUEUE_SHIFT);
4685         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
4686         return 0;
4687 }
4688
4689 /*
4690  * remove a 5tuple filter
4691  *
4692  * @param
4693  * dev: Pointer to struct rte_eth_dev.
4694  * filter: the pointer of the filter will be removed.
4695  */
4696 static void
4697 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
4698                         struct ixgbe_5tuple_filter *filter)
4699 {
4700         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4701         struct ixgbe_filter_info *filter_info =
4702                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4703         uint16_t index = filter->index;
4704
4705         filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
4706                                 ~(1 << (index % (sizeof(uint32_t) * NBBY)));
4707         TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
4708         rte_free(filter);
4709
4710         IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
4711         IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
4712         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
4713         IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
4714         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
4715 }
4716
4717 static int
4718 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
4719 {
4720         struct ixgbe_hw *hw;
4721         uint32_t max_frame = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
4722
4723         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4724
4725         if ((mtu < ETHER_MIN_MTU) || (max_frame > ETHER_MAX_JUMBO_FRAME_LEN))
4726                 return -EINVAL;
4727
4728         /* refuse mtu that requires the support of scattered packets when this
4729          * feature has not been enabled before. */
4730         if (!dev->data->scattered_rx &&
4731             (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
4732              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM))
4733                 return -EINVAL;
4734
4735         /*
4736          * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
4737          * request of the version 2.0 of the mailbox API.
4738          * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
4739          * of the mailbox API.
4740          * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
4741          * prior to 3.11.33 which contains the following change:
4742          * "ixgbe: Enable jumbo frames support w/ SR-IOV"
4743          */
4744         ixgbevf_rlpml_set_vf(hw, max_frame);
4745
4746         /* update max frame size */
4747         dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame;
4748         return 0;
4749 }
4750
4751 #define MAC_TYPE_FILTER_SUP_EXT(type)    do {\
4752         if ((type) != ixgbe_mac_82599EB && (type) != ixgbe_mac_X540)\
4753                 return -ENOTSUP;\
4754 } while (0)
4755
4756 static inline struct ixgbe_5tuple_filter *
4757 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
4758                         struct ixgbe_5tuple_filter_info *key)
4759 {
4760         struct ixgbe_5tuple_filter *it;
4761
4762         TAILQ_FOREACH(it, filter_list, entries) {
4763                 if (memcmp(key, &it->filter_info,
4764                         sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
4765                         return it;
4766                 }
4767         }
4768         return NULL;
4769 }
4770
4771 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
4772 static inline int
4773 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
4774                         struct ixgbe_5tuple_filter_info *filter_info)
4775 {
4776         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
4777                 filter->priority > IXGBE_5TUPLE_MAX_PRI ||
4778                 filter->priority < IXGBE_5TUPLE_MIN_PRI)
4779                 return -EINVAL;
4780
4781         switch (filter->dst_ip_mask) {
4782         case UINT32_MAX:
4783                 filter_info->dst_ip_mask = 0;
4784                 filter_info->dst_ip = filter->dst_ip;
4785                 break;
4786         case 0:
4787                 filter_info->dst_ip_mask = 1;
4788                 break;
4789         default:
4790                 PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
4791                 return -EINVAL;
4792         }
4793
4794         switch (filter->src_ip_mask) {
4795         case UINT32_MAX:
4796                 filter_info->src_ip_mask = 0;
4797                 filter_info->src_ip = filter->src_ip;
4798                 break;
4799         case 0:
4800                 filter_info->src_ip_mask = 1;
4801                 break;
4802         default:
4803                 PMD_DRV_LOG(ERR, "invalid src_ip mask.");
4804                 return -EINVAL;
4805         }
4806
4807         switch (filter->dst_port_mask) {
4808         case UINT16_MAX:
4809                 filter_info->dst_port_mask = 0;
4810                 filter_info->dst_port = filter->dst_port;
4811                 break;
4812         case 0:
4813                 filter_info->dst_port_mask = 1;
4814                 break;
4815         default:
4816                 PMD_DRV_LOG(ERR, "invalid dst_port mask.");
4817                 return -EINVAL;
4818         }
4819
4820         switch (filter->src_port_mask) {
4821         case UINT16_MAX:
4822                 filter_info->src_port_mask = 0;
4823                 filter_info->src_port = filter->src_port;
4824                 break;
4825         case 0:
4826                 filter_info->src_port_mask = 1;
4827                 break;
4828         default:
4829                 PMD_DRV_LOG(ERR, "invalid src_port mask.");
4830                 return -EINVAL;
4831         }
4832
4833         switch (filter->proto_mask) {
4834         case UINT8_MAX:
4835                 filter_info->proto_mask = 0;
4836                 filter_info->proto =
4837                         convert_protocol_type(filter->proto);
4838                 break;
4839         case 0:
4840                 filter_info->proto_mask = 1;
4841                 break;
4842         default:
4843                 PMD_DRV_LOG(ERR, "invalid protocol mask.");
4844                 return -EINVAL;
4845         }
4846
4847         filter_info->priority = (uint8_t)filter->priority;
4848         return 0;
4849 }
4850
4851 /*
4852  * add or delete a ntuple filter
4853  *
4854  * @param
4855  * dev: Pointer to struct rte_eth_dev.
4856  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4857  * add: if true, add filter, if false, remove filter
4858  *
4859  * @return
4860  *    - On success, zero.
4861  *    - On failure, a negative value.
4862  */
4863 static int
4864 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
4865                         struct rte_eth_ntuple_filter *ntuple_filter,
4866                         bool add)
4867 {
4868         struct ixgbe_filter_info *filter_info =
4869                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4870         struct ixgbe_5tuple_filter_info filter_5tuple;
4871         struct ixgbe_5tuple_filter *filter;
4872         int ret;
4873
4874         if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
4875                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
4876                 return -EINVAL;
4877         }
4878
4879         memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
4880         ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
4881         if (ret < 0)
4882                 return ret;
4883
4884         filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
4885                                          &filter_5tuple);
4886         if (filter != NULL && add) {
4887                 PMD_DRV_LOG(ERR, "filter exists.");
4888                 return -EEXIST;
4889         }
4890         if (filter == NULL && !add) {
4891                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
4892                 return -ENOENT;
4893         }
4894
4895         if (add) {
4896                 filter = rte_zmalloc("ixgbe_5tuple_filter",
4897                                 sizeof(struct ixgbe_5tuple_filter), 0);
4898                 if (filter == NULL)
4899                         return -ENOMEM;
4900                 (void)rte_memcpy(&filter->filter_info,
4901                                  &filter_5tuple,
4902                                  sizeof(struct ixgbe_5tuple_filter_info));
4903                 filter->queue = ntuple_filter->queue;
4904                 ret = ixgbe_add_5tuple_filter(dev, filter);
4905                 if (ret < 0) {
4906                         rte_free(filter);
4907                         return ret;
4908                 }
4909         } else
4910                 ixgbe_remove_5tuple_filter(dev, filter);
4911
4912         return 0;
4913 }
4914
4915 /*
4916  * get a ntuple filter
4917  *
4918  * @param
4919  * dev: Pointer to struct rte_eth_dev.
4920  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4921  *
4922  * @return
4923  *    - On success, zero.
4924  *    - On failure, a negative value.
4925  */
4926 static int
4927 ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
4928                         struct rte_eth_ntuple_filter *ntuple_filter)
4929 {
4930         struct ixgbe_filter_info *filter_info =
4931                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4932         struct ixgbe_5tuple_filter_info filter_5tuple;
4933         struct ixgbe_5tuple_filter *filter;
4934         int ret;
4935
4936         if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
4937                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
4938                 return -EINVAL;
4939         }
4940
4941         memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
4942         ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
4943         if (ret < 0)
4944                 return ret;
4945
4946         filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
4947                                          &filter_5tuple);
4948         if (filter == NULL) {
4949                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
4950                 return -ENOENT;
4951         }
4952         ntuple_filter->queue = filter->queue;
4953         return 0;
4954 }
4955
4956 /*
4957  * ixgbe_ntuple_filter_handle - Handle operations for ntuple filter.
4958  * @dev: pointer to rte_eth_dev structure
4959  * @filter_op:operation will be taken.
4960  * @arg: a pointer to specific structure corresponding to the filter_op
4961  *
4962  * @return
4963  *    - On success, zero.
4964  *    - On failure, a negative value.
4965  */
4966 static int
4967 ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
4968                                 enum rte_filter_op filter_op,
4969                                 void *arg)
4970 {
4971         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4972         int ret;
4973
4974         MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
4975
4976         if (filter_op == RTE_ETH_FILTER_NOP)
4977                 return 0;
4978
4979         if (arg == NULL) {
4980                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4981                             filter_op);
4982                 return -EINVAL;
4983         }
4984
4985         switch (filter_op) {
4986         case RTE_ETH_FILTER_ADD:
4987                 ret = ixgbe_add_del_ntuple_filter(dev,
4988                         (struct rte_eth_ntuple_filter *)arg,
4989                         TRUE);
4990                 break;
4991         case RTE_ETH_FILTER_DELETE:
4992                 ret = ixgbe_add_del_ntuple_filter(dev,
4993                         (struct rte_eth_ntuple_filter *)arg,
4994                         FALSE);
4995                 break;
4996         case RTE_ETH_FILTER_GET:
4997                 ret = ixgbe_get_ntuple_filter(dev,
4998                         (struct rte_eth_ntuple_filter *)arg);
4999                 break;
5000         default:
5001                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
5002                 ret = -EINVAL;
5003                 break;
5004         }
5005         return ret;
5006 }
5007
5008 static inline int
5009 ixgbe_ethertype_filter_lookup(struct ixgbe_filter_info *filter_info,
5010                         uint16_t ethertype)
5011 {
5012         int i;
5013
5014         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
5015                 if (filter_info->ethertype_filters[i] == ethertype &&
5016                     (filter_info->ethertype_mask & (1 << i)))
5017                         return i;
5018         }
5019         return -1;
5020 }
5021
5022 static inline int
5023 ixgbe_ethertype_filter_insert(struct ixgbe_filter_info *filter_info,
5024                         uint16_t ethertype)
5025 {
5026         int i;
5027
5028         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
5029                 if (!(filter_info->ethertype_mask & (1 << i))) {
5030                         filter_info->ethertype_mask |= 1 << i;
5031                         filter_info->ethertype_filters[i] = ethertype;
5032                         return i;
5033                 }
5034         }
5035         return -1;
5036 }
5037
5038 static inline int
5039 ixgbe_ethertype_filter_remove(struct ixgbe_filter_info *filter_info,
5040                         uint8_t idx)
5041 {
5042         if (idx >= IXGBE_MAX_ETQF_FILTERS)
5043                 return -1;
5044         filter_info->ethertype_mask &= ~(1 << idx);
5045         filter_info->ethertype_filters[idx] = 0;
5046         return idx;
5047 }
5048
5049 static int
5050 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
5051                         struct rte_eth_ethertype_filter *filter,
5052                         bool add)
5053 {
5054         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5055         struct ixgbe_filter_info *filter_info =
5056                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5057         uint32_t etqf = 0;
5058         uint32_t etqs = 0;
5059         int ret;
5060
5061         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
5062                 return -EINVAL;
5063
5064         if (filter->ether_type == ETHER_TYPE_IPv4 ||
5065                 filter->ether_type == ETHER_TYPE_IPv6) {
5066                 PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
5067                         " ethertype filter.", filter->ether_type);
5068                 return -EINVAL;
5069         }
5070
5071         if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
5072                 PMD_DRV_LOG(ERR, "mac compare is unsupported.");
5073                 return -EINVAL;
5074         }
5075         if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
5076                 PMD_DRV_LOG(ERR, "drop option is unsupported.");
5077                 return -EINVAL;
5078         }
5079
5080         ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
5081         if (ret >= 0 && add) {
5082                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
5083                             filter->ether_type);
5084                 return -EEXIST;
5085         }
5086         if (ret < 0 && !add) {
5087                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
5088                             filter->ether_type);
5089                 return -ENOENT;
5090         }
5091
5092         if (add) {
5093                 ret = ixgbe_ethertype_filter_insert(filter_info,
5094                         filter->ether_type);
5095                 if (ret < 0) {
5096                         PMD_DRV_LOG(ERR, "ethertype filters are full.");
5097                         return -ENOSYS;
5098                 }
5099                 etqf = IXGBE_ETQF_FILTER_EN;
5100                 etqf |= (uint32_t)filter->ether_type;
5101                 etqs |= (uint32_t)((filter->queue <<
5102                                     IXGBE_ETQS_RX_QUEUE_SHIFT) &
5103                                     IXGBE_ETQS_RX_QUEUE);
5104                 etqs |= IXGBE_ETQS_QUEUE_EN;
5105         } else {
5106                 ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
5107                 if (ret < 0)
5108                         return -ENOSYS;
5109         }
5110         IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
5111         IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
5112         IXGBE_WRITE_FLUSH(hw);
5113
5114         return 0;
5115 }
5116
5117 static int
5118 ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
5119                         struct rte_eth_ethertype_filter *filter)
5120 {
5121         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5122         struct ixgbe_filter_info *filter_info =
5123                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5124         uint32_t etqf, etqs;
5125         int ret;
5126
5127         ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
5128         if (ret < 0) {
5129                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
5130                             filter->ether_type);
5131                 return -ENOENT;
5132         }
5133
5134         etqf = IXGBE_READ_REG(hw, IXGBE_ETQF(ret));
5135         if (etqf & IXGBE_ETQF_FILTER_EN) {
5136                 etqs = IXGBE_READ_REG(hw, IXGBE_ETQS(ret));
5137                 filter->ether_type = etqf & IXGBE_ETQF_ETHERTYPE;
5138                 filter->flags = 0;
5139                 filter->queue = (etqs & IXGBE_ETQS_RX_QUEUE) >>
5140                                IXGBE_ETQS_RX_QUEUE_SHIFT;
5141                 return 0;
5142         }
5143         return -ENOENT;
5144 }
5145
5146 /*
5147  * ixgbe_ethertype_filter_handle - Handle operations for ethertype filter.
5148  * @dev: pointer to rte_eth_dev structure
5149  * @filter_op:operation will be taken.
5150  * @arg: a pointer to specific structure corresponding to the filter_op
5151  */
5152 static int
5153 ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
5154                                 enum rte_filter_op filter_op,
5155                                 void *arg)
5156 {
5157         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5158         int ret;
5159
5160         MAC_TYPE_FILTER_SUP(hw->mac.type);
5161
5162         if (filter_op == RTE_ETH_FILTER_NOP)
5163                 return 0;
5164
5165         if (arg == NULL) {
5166                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
5167                             filter_op);
5168                 return -EINVAL;
5169         }
5170
5171         switch (filter_op) {
5172         case RTE_ETH_FILTER_ADD:
5173                 ret = ixgbe_add_del_ethertype_filter(dev,
5174                         (struct rte_eth_ethertype_filter *)arg,
5175                         TRUE);
5176                 break;
5177         case RTE_ETH_FILTER_DELETE:
5178                 ret = ixgbe_add_del_ethertype_filter(dev,
5179                         (struct rte_eth_ethertype_filter *)arg,
5180                         FALSE);
5181                 break;
5182         case RTE_ETH_FILTER_GET:
5183                 ret = ixgbe_get_ethertype_filter(dev,
5184                         (struct rte_eth_ethertype_filter *)arg);
5185                 break;
5186         default:
5187                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
5188                 ret = -EINVAL;
5189                 break;
5190         }
5191         return ret;
5192 }
5193
5194 static int
5195 ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
5196                      enum rte_filter_type filter_type,
5197                      enum rte_filter_op filter_op,
5198                      void *arg)
5199 {
5200         int ret = -EINVAL;
5201
5202         switch (filter_type) {
5203         case RTE_ETH_FILTER_NTUPLE:
5204                 ret = ixgbe_ntuple_filter_handle(dev, filter_op, arg);
5205                 break;
5206         case RTE_ETH_FILTER_ETHERTYPE:
5207                 ret = ixgbe_ethertype_filter_handle(dev, filter_op, arg);
5208                 break;
5209         case RTE_ETH_FILTER_SYN:
5210                 ret = ixgbe_syn_filter_handle(dev, filter_op, arg);
5211                 break;
5212         case RTE_ETH_FILTER_FDIR:
5213                 ret = ixgbe_fdir_ctrl_func(dev, filter_op, arg);
5214                 break;
5215         default:
5216                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
5217                                                         filter_type);
5218                 break;
5219         }
5220
5221         return ret;
5222 }
5223
5224 static u8 *
5225 ixgbe_dev_addr_list_itr(__attribute__((unused)) struct ixgbe_hw *hw,
5226                         u8 **mc_addr_ptr, u32 *vmdq)
5227 {
5228         u8 *mc_addr;
5229
5230         *vmdq = 0;
5231         mc_addr = *mc_addr_ptr;
5232         *mc_addr_ptr = (mc_addr + sizeof(struct ether_addr));
5233         return mc_addr;
5234 }
5235
5236 static int
5237 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
5238                           struct ether_addr *mc_addr_set,
5239                           uint32_t nb_mc_addr)
5240 {
5241         struct ixgbe_hw *hw;
5242         u8 *mc_addr_list;
5243
5244         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5245         mc_addr_list = (u8 *)mc_addr_set;
5246         return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
5247                                          ixgbe_dev_addr_list_itr, TRUE);
5248 }
5249
5250 static int
5251 ixgbe_timesync_enable(struct rte_eth_dev *dev)
5252 {
5253         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5254         uint32_t tsync_ctl;
5255         uint32_t tsauxc;
5256
5257         /* Enable system time for platforms where it isn't on by default. */
5258         tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
5259         tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
5260         IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
5261
5262         /* Start incrementing the register used to timestamp PTP packets. */
5263         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, IXGBE_TIMINCA_INIT);
5264
5265         /* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5266         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
5267                         (ETHER_TYPE_1588 |
5268                          IXGBE_ETQF_FILTER_EN |
5269                          IXGBE_ETQF_1588));
5270
5271         /* Enable timestamping of received PTP packets. */
5272         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
5273         tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
5274         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
5275
5276         /* Enable timestamping of transmitted PTP packets. */
5277         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
5278         tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
5279         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
5280
5281         return 0;
5282 }
5283
5284 static int
5285 ixgbe_timesync_disable(struct rte_eth_dev *dev)
5286 {
5287         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5288         uint32_t tsync_ctl;
5289
5290         /* Disable timestamping of transmitted PTP packets. */
5291         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
5292         tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
5293         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
5294
5295         /* Disable timestamping of received PTP packets. */
5296         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
5297         tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
5298         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
5299
5300         /* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5301         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
5302
5303         /* Stop incrementating the System Time registers. */
5304         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
5305
5306         return 0;
5307 }
5308
5309 static int
5310 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
5311                                  struct timespec *timestamp,
5312                                  uint32_t flags __rte_unused)
5313 {
5314         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5315         uint32_t tsync_rxctl;
5316         uint32_t rx_stmpl;
5317         uint32_t rx_stmph;
5318
5319         tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
5320         if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
5321                 return -EINVAL;
5322
5323         rx_stmpl = IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
5324         rx_stmph = IXGBE_READ_REG(hw, IXGBE_RXSTMPH);
5325
5326         timestamp->tv_sec = (uint64_t)(((uint64_t)rx_stmph << 32) | rx_stmpl);
5327         timestamp->tv_nsec = 0;
5328
5329         return  0;
5330 }
5331
5332 static int
5333 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
5334                                  struct timespec *timestamp)
5335 {
5336         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5337         uint32_t tsync_txctl;
5338         uint32_t tx_stmpl;
5339         uint32_t tx_stmph;
5340
5341         tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
5342         if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
5343                 return -EINVAL;
5344
5345         tx_stmpl = IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
5346         tx_stmph = IXGBE_READ_REG(hw, IXGBE_TXSTMPH);
5347
5348         timestamp->tv_sec = (uint64_t)(((uint64_t)tx_stmph << 32) | tx_stmpl);
5349         timestamp->tv_nsec = 0;
5350
5351         return  0;
5352 }
5353
5354 static int
5355 ixgbe_get_reg_length(struct rte_eth_dev *dev)
5356 {
5357         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5358         int count = 0;
5359         int g_ind = 0;
5360         const struct reg_info *reg_group;
5361         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
5362                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
5363
5364         while ((reg_group = reg_set[g_ind++]))
5365                 count += ixgbe_regs_group_count(reg_group);
5366
5367         return count;
5368 }
5369
5370 static int
5371 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
5372 {
5373         int count = 0;
5374         int g_ind = 0;
5375         const struct reg_info *reg_group;
5376
5377         while ((reg_group = ixgbevf_regs[g_ind++]))
5378                 count += ixgbe_regs_group_count(reg_group);
5379
5380         return count;
5381 }
5382
5383 static int
5384 ixgbe_get_regs(struct rte_eth_dev *dev,
5385               struct rte_dev_reg_info *regs)
5386 {
5387         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5388         uint32_t *data = regs->data;
5389         int g_ind = 0;
5390         int count = 0;
5391         const struct reg_info *reg_group;
5392         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
5393                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
5394
5395         /* Support only full register dump */
5396         if ((regs->length == 0) ||
5397             (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
5398                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5399                         hw->device_id;
5400                 while ((reg_group = reg_set[g_ind++]))
5401                         count += ixgbe_read_regs_group(dev, &data[count],
5402                                 reg_group);
5403                 return 0;
5404         }
5405
5406         return -ENOTSUP;
5407 }
5408
5409 static int
5410 ixgbevf_get_regs(struct rte_eth_dev *dev,
5411                 struct rte_dev_reg_info *regs)
5412 {
5413         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5414         uint32_t *data = regs->data;
5415         int g_ind = 0;
5416         int count = 0;
5417         const struct reg_info *reg_group;
5418
5419         /* Support only full register dump */
5420         if ((regs->length == 0) ||
5421             (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
5422                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5423                         hw->device_id;
5424                 while ((reg_group = ixgbevf_regs[g_ind++]))
5425                         count += ixgbe_read_regs_group(dev, &data[count],
5426                                                       reg_group);
5427                 return 0;
5428         }
5429
5430         return -ENOTSUP;
5431 }
5432
5433 static int
5434 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
5435 {
5436         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5437
5438         /* Return unit is byte count */
5439         return hw->eeprom.word_size * 2;
5440 }
5441
5442 static int
5443 ixgbe_get_eeprom(struct rte_eth_dev *dev,
5444                 struct rte_dev_eeprom_info *in_eeprom)
5445 {
5446         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5447         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
5448         uint16_t *data = in_eeprom->data;
5449         int first, length;
5450
5451         first = in_eeprom->offset >> 1;
5452         length = in_eeprom->length >> 1;
5453         if ((first >= hw->eeprom.word_size) ||
5454             ((first + length) >= hw->eeprom.word_size))
5455                 return -EINVAL;
5456
5457         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
5458
5459         return eeprom->ops.read_buffer(hw, first, length, data);
5460 }
5461
5462 static int
5463 ixgbe_set_eeprom(struct rte_eth_dev *dev,
5464                 struct rte_dev_eeprom_info *in_eeprom)
5465 {
5466         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5467         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
5468         uint16_t *data = in_eeprom->data;
5469         int first, length;
5470
5471         first = in_eeprom->offset >> 1;
5472         length = in_eeprom->length >> 1;
5473         if ((first >= hw->eeprom.word_size) ||
5474             ((first + length) >= hw->eeprom.word_size))
5475                 return -EINVAL;
5476
5477         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
5478
5479         return eeprom->ops.write_buffer(hw,  first, length, data);
5480 }
5481
5482 static struct rte_driver rte_ixgbe_driver = {
5483         .type = PMD_PDEV,
5484         .init = rte_ixgbe_pmd_init,
5485 };
5486
5487 static struct rte_driver rte_ixgbevf_driver = {
5488         .type = PMD_PDEV,
5489         .init = rte_ixgbevf_pmd_init,
5490 };
5491
5492 PMD_REGISTER_DRIVER(rte_ixgbe_driver);
5493 PMD_REGISTER_DRIVER(rte_ixgbevf_driver);