ixgbe: add 82599 bypass support
[dpdk.git] / lib / librte_pmd_ixgbe / ixgbe_ethdev.c
1 /*-
2  *   BSD LICENSE
3  * 
4  *   Copyright(c) 2010-2013 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 <rte_byteorder.h>
43 #include <rte_common.h>
44 #include <rte_cycles.h>
45
46 #include <rte_interrupts.h>
47 #include <rte_log.h>
48 #include <rte_debug.h>
49 #include <rte_pci.h>
50 #include <rte_atomic.h>
51 #include <rte_branch_prediction.h>
52 #include <rte_memory.h>
53 #include <rte_memzone.h>
54 #include <rte_tailq.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
62 #include "ixgbe_logs.h"
63 #include "ixgbe/ixgbe_api.h"
64 #include "ixgbe/ixgbe_vf.h"
65 #include "ixgbe/ixgbe_common.h"
66 #include "ixgbe_ethdev.h"
67 #include "ixgbe_bypass.h"
68
69 /*
70  * High threshold controlling when to start sending XOFF frames. Must be at
71  * least 8 bytes less than receive packet buffer size. This value is in units
72  * of 1024 bytes.
73  */
74 #define IXGBE_FC_HI    0x80
75
76 /*
77  * Low threshold controlling when to start sending XON frames. This value is
78  * in units of 1024 bytes.
79  */
80 #define IXGBE_FC_LO    0x40
81
82 /* Timer value included in XOFF frames. */
83 #define IXGBE_FC_PAUSE 0x680
84
85 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
86 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
87 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
88
89
90 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
91
92 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
93
94 static int eth_ixgbe_dev_init(struct eth_driver *eth_drv,
95                 struct rte_eth_dev *eth_dev);
96 static int  ixgbe_dev_configure(struct rte_eth_dev *dev);
97 static int  ixgbe_dev_start(struct rte_eth_dev *dev);
98 static void ixgbe_dev_stop(struct rte_eth_dev *dev);
99 static void ixgbe_dev_close(struct rte_eth_dev *dev);
100 static void ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
101 static void ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
102 static void ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
103 static void ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
104 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
105                                 int wait_to_complete);
106 static void ixgbe_dev_stats_get(struct rte_eth_dev *dev,
107                                 struct rte_eth_stats *stats);
108 static void ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
109 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
110                                              uint16_t queue_id,
111                                              uint8_t stat_idx,
112                                              uint8_t is_rx);
113 static void ixgbe_dev_info_get(struct rte_eth_dev *dev,
114                                 struct rte_eth_dev_info *dev_info);
115 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
116                 uint16_t vlan_id, int on);
117 static void ixgbe_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid_id);
118 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, 
119                 uint16_t queue, bool on);
120 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
121                 int on);
122 static void ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
123 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
124 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
125 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
126 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
127
128 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
129 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
130 static int  ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
131                 struct rte_eth_fc_conf *fc_conf);
132 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
133                 struct rte_eth_pfc_conf *pfc_conf);
134 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
135                 struct rte_eth_rss_reta *reta_conf);
136 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
137                 struct rte_eth_rss_reta *reta_conf);    
138 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
139 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev);
140 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
141 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev);
142 static void ixgbe_dev_interrupt_handler(struct rte_intr_handle *handle,
143                 void *param);
144 static void ixgbe_dev_interrupt_delayed_handler(void *param);
145 static void ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
146                 uint32_t index, uint32_t pool);
147 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
148 static void ixgbe_dcb_init(struct ixgbe_hw *hw,struct ixgbe_dcb_config *dcb_config);
149
150 /* For Virtual Function support */
151 static int eth_ixgbevf_dev_init(struct eth_driver *eth_drv,
152                 struct rte_eth_dev *eth_dev);
153 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
154 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
155 static void ixgbevf_dev_stop(struct rte_eth_dev *dev);
156 static void ixgbevf_dev_close(struct rte_eth_dev *dev);
157 static void ixgbevf_intr_disable(struct ixgbe_hw *hw);
158 static void ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
159                 struct rte_eth_stats *stats);
160 static void ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
161 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, 
162                 uint16_t vlan_id, int on);
163 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
164                 uint16_t queue, int on);
165 static void ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
166 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
167
168 /* For Eth VMDQ APIs support */
169 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
170                 ether_addr* mac_addr,uint8_t on);
171 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev,uint8_t on);
172 static int  ixgbe_set_pool_rx_mode(struct rte_eth_dev *dev,  uint16_t pool, 
173                 uint16_t rx_mask, uint8_t on);
174 static int ixgbe_set_pool_rx(struct rte_eth_dev *dev,uint16_t pool,uint8_t on);
175 static int ixgbe_set_pool_tx(struct rte_eth_dev *dev,uint16_t pool,uint8_t on);
176 static int ixgbe_set_pool_vlan_filter(struct rte_eth_dev *dev, uint16_t vlan, 
177                 uint64_t pool_mask,uint8_t vlan_on);
178 static int ixgbe_mirror_rule_set(struct rte_eth_dev *dev, 
179                 struct rte_eth_vmdq_mirror_conf *mirror_conf, 
180                 uint8_t rule_id, uint8_t on);
181 static int ixgbe_mirror_rule_reset(struct rte_eth_dev *dev,
182                 uint8_t rule_id);
183
184 /*
185  * Define VF Stats MACRO for Non "cleared on read" register
186  */
187 #define UPDATE_VF_STAT(reg, last, cur)                          \
188 {                                                               \
189         u32 latest = IXGBE_READ_REG(hw, reg);                   \
190         cur += latest - last;                                   \
191         last = latest;                                          \
192 }
193
194 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
195 {                                                                \
196         u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
197         u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
198         u64 latest = ((new_msb << 32) | new_lsb);                \
199         cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
200         last = latest;                                           \
201 }
202
203 #define IXGBE_SET_HWSTRIP(h, q) do{\
204                 uint32_t idx = (q) / (sizeof ((h)->bitmap[0]) * NBBY); \
205                 uint32_t bit = (q) % (sizeof ((h)->bitmap[0]) * NBBY); \
206                 (h)->bitmap[idx] |= 1 << bit;\
207         }while(0)
208         
209 #define IXGBE_CLEAR_HWSTRIP(h, q) do{\
210                 uint32_t idx = (q) / (sizeof ((h)->bitmap[0]) * NBBY); \
211                 uint32_t bit = (q) % (sizeof ((h)->bitmap[0]) * NBBY); \
212                 (h)->bitmap[idx] &= ~(1 << bit);\
213         }while(0)
214  
215 #define IXGBE_GET_HWSTRIP(h, q, r) do{\
216                 uint32_t idx = (q) / (sizeof ((h)->bitmap[0]) * NBBY); \
217                 uint32_t bit = (q) % (sizeof ((h)->bitmap[0]) * NBBY); \
218                 (r) = (h)->bitmap[idx] >> bit & 1;\
219         }while(0)
220
221 /*
222  * The set of PCI devices this driver supports
223  */
224 static struct rte_pci_id pci_id_ixgbe_map[] = {
225
226 #define RTE_PCI_DEV_ID_DECL_IXGBE(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
227 #include "rte_pci_dev_ids.h"
228
229 { .vendor_id = 0, /* sentinel */ },
230 };
231
232
233 /*
234  * The set of PCI devices this driver supports (for 82599 VF)
235  */
236 static struct rte_pci_id pci_id_ixgbevf_map[] = {
237
238 #define RTE_PCI_DEV_ID_DECL_IXGBEVF(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
239 #include "rte_pci_dev_ids.h"
240 { .vendor_id = 0, /* sentinel */ },
241
242 };
243
244 static struct eth_dev_ops ixgbe_eth_dev_ops = {
245         .dev_configure        = ixgbe_dev_configure,
246         .dev_start            = ixgbe_dev_start,
247         .dev_stop             = ixgbe_dev_stop,
248         .dev_close            = ixgbe_dev_close,
249         .promiscuous_enable   = ixgbe_dev_promiscuous_enable,
250         .promiscuous_disable  = ixgbe_dev_promiscuous_disable,
251         .allmulticast_enable  = ixgbe_dev_allmulticast_enable,
252         .allmulticast_disable = ixgbe_dev_allmulticast_disable,
253         .link_update          = ixgbe_dev_link_update,
254         .stats_get            = ixgbe_dev_stats_get,
255         .stats_reset          = ixgbe_dev_stats_reset,
256         .queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
257         .dev_infos_get        = ixgbe_dev_info_get,
258         .vlan_filter_set      = ixgbe_vlan_filter_set,
259         .vlan_tpid_set        = ixgbe_vlan_tpid_set,
260         .vlan_offload_set     = ixgbe_vlan_offload_set,
261         .vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
262         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
263         .rx_queue_release     = ixgbe_dev_rx_queue_release,
264         .rx_queue_count       = ixgbe_dev_rx_queue_count,
265         .rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
266         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
267         .tx_queue_release     = ixgbe_dev_tx_queue_release,
268         .dev_led_on           = ixgbe_dev_led_on,
269         .dev_led_off          = ixgbe_dev_led_off,
270         .flow_ctrl_set        = ixgbe_flow_ctrl_set,
271         .priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
272         .mac_addr_add         = ixgbe_add_rar,
273         .mac_addr_remove      = ixgbe_remove_rar,
274         .uc_hash_table_set    = ixgbe_uc_hash_table_set,
275         .uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
276         .mirror_rule_set        = ixgbe_mirror_rule_set,
277         .mirror_rule_reset      = ixgbe_mirror_rule_reset,
278         .set_vf_rx_mode       = ixgbe_set_pool_rx_mode,
279         .set_vf_rx            = ixgbe_set_pool_rx,
280         .set_vf_tx            = ixgbe_set_pool_tx,
281         .set_vf_vlan_filter   = ixgbe_set_pool_vlan_filter,
282         .fdir_add_signature_filter    = ixgbe_fdir_add_signature_filter,
283         .fdir_update_signature_filter = ixgbe_fdir_update_signature_filter,
284         .fdir_remove_signature_filter = ixgbe_fdir_remove_signature_filter,
285         .fdir_infos_get               = ixgbe_fdir_info_get,
286         .fdir_add_perfect_filter      = ixgbe_fdir_add_perfect_filter,
287         .fdir_update_perfect_filter   = ixgbe_fdir_update_perfect_filter,
288         .fdir_remove_perfect_filter   = ixgbe_fdir_remove_perfect_filter,
289         .fdir_set_masks               = ixgbe_fdir_set_masks,
290         .reta_update          = ixgbe_dev_rss_reta_update,
291         .reta_query           = ixgbe_dev_rss_reta_query,
292 #ifdef RTE_NIC_BYPASS
293         .bypass_init          = ixgbe_bypass_init,
294         .bypass_state_set     = ixgbe_bypass_state_store,
295         .bypass_state_show    = ixgbe_bypass_state_show,
296         .bypass_event_set     = ixgbe_bypass_event_store,
297         .bypass_event_show    = ixgbe_bypass_event_show,
298         .bypass_wd_timeout_set  = ixgbe_bypass_wd_timeout_store,
299         .bypass_wd_timeout_show = ixgbe_bypass_wd_timeout_show,
300         .bypass_ver_show      = ixgbe_bypass_ver_show,
301         .bypass_wd_reset      = ixgbe_bypass_wd_reset,
302 #endif /* RTE_NIC_BYPASS */
303 };
304
305 /*
306  * dev_ops for virtual function, bare necessities for basic vf
307  * operation have been implemented
308  */
309 static struct eth_dev_ops ixgbevf_eth_dev_ops = {
310
311         .dev_configure        = ixgbevf_dev_configure,
312         .dev_start            = ixgbevf_dev_start,
313         .dev_stop             = ixgbevf_dev_stop,
314         .link_update          = ixgbe_dev_link_update,
315         .stats_get            = ixgbevf_dev_stats_get,
316         .stats_reset          = ixgbevf_dev_stats_reset,
317         .dev_close            = ixgbevf_dev_close,
318         .dev_infos_get        = ixgbe_dev_info_get,
319         .vlan_filter_set      = ixgbevf_vlan_filter_set,
320         .vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
321         .vlan_offload_set     = ixgbevf_vlan_offload_set,
322         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
323         .rx_queue_release     = ixgbe_dev_rx_queue_release,
324         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
325         .tx_queue_release     = ixgbe_dev_tx_queue_release,
326 };
327
328 /**
329  * Atomically reads the link status information from global
330  * structure rte_eth_dev.
331  *
332  * @param dev
333  *   - Pointer to the structure rte_eth_dev to read from.
334  *   - Pointer to the buffer to be saved with the link status.
335  *
336  * @return
337  *   - On success, zero.
338  *   - On failure, negative value.
339  */
340 static inline int
341 rte_ixgbe_dev_atomic_read_link_status(struct rte_eth_dev *dev,
342                                 struct rte_eth_link *link)
343 {
344         struct rte_eth_link *dst = link;
345         struct rte_eth_link *src = &(dev->data->dev_link);
346
347         if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
348                                         *(uint64_t *)src) == 0)
349                 return -1;
350
351         return 0;
352 }
353
354 /**
355  * Atomically writes the link status information into global
356  * structure rte_eth_dev.
357  *
358  * @param dev
359  *   - Pointer to the structure rte_eth_dev to read from.
360  *   - Pointer to the buffer to be saved with the link status.
361  *
362  * @return
363  *   - On success, zero.
364  *   - On failure, negative value.
365  */
366 static inline int
367 rte_ixgbe_dev_atomic_write_link_status(struct rte_eth_dev *dev,
368                                 struct rte_eth_link *link)
369 {
370         struct rte_eth_link *dst = &(dev->data->dev_link);
371         struct rte_eth_link *src = link;
372
373         if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
374                                         *(uint64_t *)src) == 0)
375                 return -1;
376
377         return 0;
378 }
379
380 /*
381  * This function is the same as ixgbe_is_sfp() in ixgbe/ixgbe.h.
382  */
383 static inline int
384 ixgbe_is_sfp(struct ixgbe_hw *hw)
385 {
386         switch (hw->phy.type) {
387         case ixgbe_phy_sfp_avago:
388         case ixgbe_phy_sfp_ftl:
389         case ixgbe_phy_sfp_intel:
390         case ixgbe_phy_sfp_unknown:
391         case ixgbe_phy_sfp_passive_tyco:
392         case ixgbe_phy_sfp_passive_unknown:
393                 return 1;
394         default:
395                 return 0;
396         }
397 }
398
399 static inline int32_t
400 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
401 {
402         uint32_t ctrl_ext;
403         int32_t status;
404
405         status = ixgbe_reset_hw(hw);
406
407         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
408         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
409         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
410         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
411         IXGBE_WRITE_FLUSH(hw);
412
413         return status;
414 }
415
416 static inline void
417 ixgbe_enable_intr(struct rte_eth_dev *dev)
418 {
419         struct ixgbe_interrupt *intr =
420                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
421         struct ixgbe_hw *hw = 
422                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
423         
424         IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
425         IXGBE_WRITE_FLUSH(hw);
426 }
427
428 /*
429  * This function is based on ixgbe_disable_intr() in ixgbe/ixgbe.h.
430  */
431 static void
432 ixgbe_disable_intr(struct ixgbe_hw *hw)
433 {
434         PMD_INIT_FUNC_TRACE();
435
436         if (hw->mac.type == ixgbe_mac_82598EB) {
437                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
438         } else {
439                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
440                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
441                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
442         }
443         IXGBE_WRITE_FLUSH(hw);
444 }
445
446 /*
447  * This function resets queue statistics mapping registers.
448  * From Niantic datasheet, Initialization of Statistics section:
449  * "...if software requires the queue counters, the RQSMR and TQSM registers
450  * must be re-programmed following a device reset.
451  */
452 static void
453 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
454 {
455         uint32_t i;
456
457         for(i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
458                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
459                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
460         }
461 }
462
463
464 static int
465 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
466                                   uint16_t queue_id,
467                                   uint8_t stat_idx,
468                                   uint8_t is_rx)
469 {
470 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
471 #define NB_QMAP_FIELDS_PER_QSM_REG 4
472 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
473
474         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
475         struct ixgbe_stat_mapping_registers *stat_mappings =
476                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
477         uint32_t qsmr_mask = 0;
478         uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
479         uint32_t q_map;
480         uint8_t n, offset;
481
482         if ((hw->mac.type != ixgbe_mac_82599EB) && (hw->mac.type != ixgbe_mac_X540))
483                 return -ENOSYS;
484
485         PMD_INIT_LOG(INFO, "Setting port %d, %s queue_id %d to stat index %d\n",
486                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX", queue_id, stat_idx);
487
488         n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
489         if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
490                 PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded\n");
491                 return -EIO;
492         }
493         offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
494
495         /* Now clear any previous stat_idx set */
496         clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
497         if (!is_rx)
498                 stat_mappings->tqsm[n] &= ~clearing_mask;
499         else
500                 stat_mappings->rqsmr[n] &= ~clearing_mask;
501
502         q_map = (uint32_t)stat_idx;
503         q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
504         qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
505         if (!is_rx)
506                 stat_mappings->tqsm[n] |= qsmr_mask;
507         else
508                 stat_mappings->rqsmr[n] |= qsmr_mask;
509
510         PMD_INIT_LOG(INFO, "Set port %d, %s queue_id %d to stat index %d\n"
511                      "%s[%d] = 0x%08x\n",
512                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX", queue_id, stat_idx,
513                      is_rx ? "RQSMR" : "TQSM",n, is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
514
515         /* Now write the mapping in the appropriate register */
516         if (is_rx) {
517                 PMD_INIT_LOG(INFO, "Write 0x%x to RX IXGBE stat mapping reg:%d\n",
518                              stat_mappings->rqsmr[n], n);
519                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
520         }
521         else {
522                 PMD_INIT_LOG(INFO, "Write 0x%x to TX IXGBE stat mapping reg:%d\n",
523                              stat_mappings->tqsm[n], n);
524                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
525         }
526         return 0;
527 }
528
529 static void
530 ixgbe_restore_statistics_mapping(struct rte_eth_dev * dev)
531 {
532         struct ixgbe_stat_mapping_registers *stat_mappings =
533                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
534         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
535         int i;
536
537         /* write whatever was in stat mapping table to the NIC */
538         for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
539                 /* rx */
540                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
541
542                 /* tx */
543                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
544         }
545 }
546
547 static void
548 ixgbe_dcb_init(struct ixgbe_hw *hw,struct ixgbe_dcb_config *dcb_config)
549 {
550         uint8_t i;
551         struct ixgbe_dcb_tc_config *tc;
552         uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
553
554         dcb_config->num_tcs.pg_tcs = dcb_max_tc;
555         dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
556         for (i = 0; i < dcb_max_tc; i++) {
557                 tc = &dcb_config->tc_config[i];
558                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
559                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
560                                  (uint8_t)(100/dcb_max_tc + (i & 1));
561                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
562                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent = 
563                                  (uint8_t)(100/dcb_max_tc + (i & 1));
564                 tc->pfc = ixgbe_dcb_pfc_disabled;
565         }
566
567         /* Initialize default user to priority mapping, UPx->TC0 */
568         tc = &dcb_config->tc_config[0];
569         tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
570         tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
571         for (i = 0; i< IXGBE_DCB_MAX_BW_GROUP; i++) {
572                 dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
573                 dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
574         }
575         dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
576         dcb_config->pfc_mode_enable = false;
577         dcb_config->vt_mode = true;
578         dcb_config->round_robin_enable = false;
579         /* support all DCB capabilities in 82599 */
580         dcb_config->support.capabilities = 0xFF;
581
582         /*we only support 4 Tcs for X540*/              
583         if (hw->mac.type == ixgbe_mac_X540) {
584                 dcb_config->num_tcs.pg_tcs = 4;
585                 dcb_config->num_tcs.pfc_tcs = 4;
586         }
587
588
589 /*
590  * This function is based on code in ixgbe_attach() in ixgbe/ixgbe.c.
591  * It returns 0 on success.
592  */
593 static int
594 eth_ixgbe_dev_init(__attribute__((unused)) struct eth_driver *eth_drv,
595                      struct rte_eth_dev *eth_dev)
596 {
597         struct rte_pci_device *pci_dev;
598         struct ixgbe_hw *hw =
599                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
600         struct ixgbe_vfta * shadow_vfta =
601                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
602         struct ixgbe_hwstrip *hwstrip = 
603                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
604         struct ixgbe_dcb_config *dcb_config =
605                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
606         uint32_t ctrl_ext;
607         uint16_t csum;
608         int diag, i;
609
610         PMD_INIT_FUNC_TRACE();
611
612         eth_dev->dev_ops = &ixgbe_eth_dev_ops;
613         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
614         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
615
616         /* for secondary processes, we don't initialise any further as primary
617          * has already done this work. Only check we don't need a different
618          * RX function */
619         if (rte_eal_process_type() != RTE_PROC_PRIMARY){
620                 if (eth_dev->data->scattered_rx)
621                         eth_dev->rx_pkt_burst = ixgbe_recv_scattered_pkts;
622                 return 0;
623         }
624         pci_dev = eth_dev->pci_dev;
625
626         /* Vendor and Device ID need to be set before init of shared code */
627         hw->device_id = pci_dev->id.device_id;
628         hw->vendor_id = pci_dev->id.vendor_id;
629         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
630 #ifdef RTE_LIBRTE_IXGBE_ALLOW_UNSUPPORTED_SFP
631         hw->allow_unsupported_sfp = 1;
632 #endif
633
634         /* Initialize the shared code */
635 #ifdef RTE_NIC_BYPASS
636         diag = ixgbe_bypass_init_shared_code(hw);
637 #else
638         diag = ixgbe_init_shared_code(hw);
639 #endif /* RTE_NIC_BYPASS */
640
641         if (diag != IXGBE_SUCCESS) {
642                 PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
643                 return -EIO;
644         }
645
646         /* Initialize DCB configuration*/
647         memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
648         ixgbe_dcb_init(hw,dcb_config);
649         /* Get Hardware Flow Control setting */
650         hw->fc.requested_mode = ixgbe_fc_full;
651         hw->fc.current_mode = ixgbe_fc_full;
652         hw->fc.pause_time = IXGBE_FC_PAUSE;
653         for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
654                 hw->fc.low_water[i] = IXGBE_FC_LO;
655                 hw->fc.high_water[i] = IXGBE_FC_HI;
656         }
657         hw->fc.send_xon = 1;
658
659         /* Make sure we have a good EEPROM before we read from it */
660         diag = ixgbe_validate_eeprom_checksum(hw, &csum);
661         if (diag != IXGBE_SUCCESS) {
662                 PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
663                 return -EIO;
664         }
665
666 #ifdef RTE_NIC_BYPASS
667         diag = ixgbe_bypass_init_hw(hw);
668 #else
669         diag = ixgbe_init_hw(hw);
670 #endif /* RTE_NIC_BYPASS */
671
672         /*
673          * Devices with copper phys will fail to initialise if ixgbe_init_hw()
674          * is called too soon after the kernel driver unbinding/binding occurs.
675          * The failure occurs in ixgbe_identify_phy_generic() for all devices,
676          * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
677          * also called. See ixgbe_identify_phy_82599(). The reason for the
678          * failure is not known, and only occuts when virtualisation features
679          * are disabled in the bios. A delay of 100ms  was found to be enough by
680          * trial-and-error, and is doubled to be safe.
681          */
682         if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
683                 rte_delay_ms(200);
684                 diag = ixgbe_init_hw(hw);
685         }
686
687         if (diag == IXGBE_ERR_EEPROM_VERSION) {
688                 PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
689                     "LOM.  Please be aware there may be issues associated "
690                     "with your hardware.\n If you are experiencing problems "
691                     "please contact your Intel or hardware representative "
692                     "who provided you with this hardware.\n");
693         } else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
694                 PMD_INIT_LOG(ERR, "Unsupported SFP+ Module\n");
695         if (diag) {
696                 PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
697                 return -EIO;
698         }
699
700         /* disable interrupt */
701         ixgbe_disable_intr(hw);
702
703         /* pick up the PCI bus settings for reporting later */
704         ixgbe_get_bus_info(hw);
705
706         /* reset mappings for queue statistics hw counters*/
707         ixgbe_reset_qstat_mappings(hw);
708
709         /* Allocate memory for storing MAC addresses */
710         eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
711                         hw->mac.num_rar_entries, 0);
712         if (eth_dev->data->mac_addrs == NULL) {
713                 PMD_INIT_LOG(ERR,
714                         "Failed to allocate %d bytes needed to store MAC addresses",
715                         ETHER_ADDR_LEN * hw->mac.num_rar_entries);
716                 return -ENOMEM;
717         }
718         /* Copy the permanent MAC address */
719         ether_addr_copy((struct ether_addr *) hw->mac.perm_addr,
720                         &eth_dev->data->mac_addrs[0]);
721         
722         /* Allocate memory for storing hash filter MAC addresses */
723         eth_dev->data->hash_mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
724                         IXGBE_VMDQ_NUM_UC_MAC, 0);
725         if (eth_dev->data->hash_mac_addrs == NULL) {
726                 PMD_INIT_LOG(ERR,
727                         "Failed to allocate %d bytes needed to store MAC addresses",
728                         ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
729                 return -ENOMEM;
730         }
731
732         /* initialize the vfta */
733         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
734
735         /* initialize the hw strip bitmap*/
736         memset(hwstrip, 0, sizeof(*hwstrip));
737
738         /* initialize PF if max_vfs not zero */
739         ixgbe_pf_host_init(eth_dev);
740
741         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
742         /* let hardware know driver is loaded */
743         ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
744         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
745         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
746         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
747         IXGBE_WRITE_FLUSH(hw);
748
749         if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
750                 PMD_INIT_LOG(DEBUG,
751                              "MAC: %d, PHY: %d, SFP+: %d<n",
752                              (int) hw->mac.type, (int) hw->phy.type,
753                              (int) hw->phy.sfp_type);
754         else
755                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d\n",
756                              (int) hw->mac.type, (int) hw->phy.type);
757
758         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
759                         eth_dev->data->port_id, pci_dev->id.vendor_id,
760                         pci_dev->id.device_id);
761
762         rte_intr_callback_register(&(pci_dev->intr_handle),
763                 ixgbe_dev_interrupt_handler, (void *)eth_dev);
764
765         /* enable uio intr after callback register */
766         rte_intr_enable(&(pci_dev->intr_handle));
767
768         /* enable support intr */
769         ixgbe_enable_intr(eth_dev);
770
771         return 0;
772 }
773
774 /*
775  * Virtual Function device init
776  */
777 static int
778 eth_ixgbevf_dev_init(__attribute__((unused)) struct eth_driver *eth_drv,
779                      struct rte_eth_dev *eth_dev)
780 {
781         struct rte_pci_device *pci_dev;
782         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
783         int diag;
784         struct ixgbe_vfta * shadow_vfta =
785                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
786         struct ixgbe_hwstrip *hwstrip = 
787                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
788
789         PMD_INIT_LOG(DEBUG, "eth_ixgbevf_dev_init");
790
791         eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
792         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
793         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
794
795         /* for secondary processes, we don't initialise any further as primary
796          * has already done this work. Only check we don't need a different
797          * RX function */
798         if (rte_eal_process_type() != RTE_PROC_PRIMARY){
799                 if (eth_dev->data->scattered_rx)
800                         eth_dev->rx_pkt_burst = ixgbe_recv_scattered_pkts;
801                 return 0;
802         }
803
804         pci_dev = eth_dev->pci_dev;
805
806         hw->device_id = pci_dev->id.device_id;
807         hw->vendor_id = pci_dev->id.vendor_id;
808         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
809
810         /* initialize the vfta */
811         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
812
813         /* initialize the hw strip bitmap*/
814         memset(hwstrip, 0, sizeof(*hwstrip));
815
816         /* Initialize the shared code */
817         diag = ixgbe_init_shared_code(hw);
818         if (diag != IXGBE_SUCCESS) {
819                 PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
820                 return -EIO;
821         }
822
823         /* init_mailbox_params */
824         hw->mbx.ops.init_params(hw);
825
826         /* Disable the interrupts for VF */
827         ixgbevf_intr_disable(hw);
828
829         hw->mac.num_rar_entries = hw->mac.max_rx_queues;
830         diag = hw->mac.ops.reset_hw(hw);
831
832         if (diag != IXGBE_SUCCESS) {
833                 PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
834                         RTE_LOG(ERR, PMD, "\tThe MAC address is not valid.\n"
835                                         "\tThe most likely cause of this error is that the VM host\n"
836                                         "\thas not assigned a valid MAC address to this VF device.\n"
837                                         "\tPlease consult the DPDK Release Notes (FAQ section) for\n"
838                                         "\ta possible solution to this problem.\n");
839                 return (diag);
840         }
841
842         /* Allocate memory for storing MAC addresses */
843         eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", ETHER_ADDR_LEN *
844                         hw->mac.num_rar_entries, 0);
845         if (eth_dev->data->mac_addrs == NULL) {
846                 PMD_INIT_LOG(ERR,
847                         "Failed to allocate %d bytes needed to store MAC addresses",
848                         ETHER_ADDR_LEN * hw->mac.num_rar_entries);
849                 return -ENOMEM;
850         }
851
852         /* Copy the permanent MAC address */
853         ether_addr_copy((struct ether_addr *) hw->mac.perm_addr,
854                         &eth_dev->data->mac_addrs[0]);
855
856         /* reset the hardware with the new settings */
857         diag = hw->mac.ops.start_hw(hw);
858         switch (diag) {
859                 case  0:
860                         break;
861
862                 default:
863                         PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
864                         return (-EIO);
865         }
866
867         PMD_INIT_LOG(DEBUG, "\nport %d vendorID=0x%x deviceID=0x%x mac.type=%s\n",
868                          eth_dev->data->port_id, pci_dev->id.vendor_id, pci_dev->id.device_id,
869                          "ixgbe_mac_82599_vf");
870
871         return 0;
872 }
873
874 static struct eth_driver rte_ixgbe_pmd = {
875         {
876                 .name = "rte_ixgbe_pmd",
877                 .id_table = pci_id_ixgbe_map,
878 #ifdef RTE_EAL_UNBIND_PORTS
879                 .drv_flags = RTE_PCI_DRV_NEED_IGB_UIO,
880 #endif
881         },
882         .eth_dev_init = eth_ixgbe_dev_init,
883         .dev_private_size = sizeof(struct ixgbe_adapter),
884 };
885
886 /*
887  * virtual function driver struct
888  */
889 static struct eth_driver rte_ixgbevf_pmd = {
890         {
891                 .name = "rte_ixgbevf_pmd",
892                 .id_table = pci_id_ixgbevf_map,
893 #ifdef RTE_EAL_UNBIND_PORTS
894                 .drv_flags = RTE_PCI_DRV_NEED_IGB_UIO,
895 #endif
896         },
897         .eth_dev_init = eth_ixgbevf_dev_init,
898         .dev_private_size = sizeof(struct ixgbe_adapter),
899 };
900
901 /*
902  * Driver initialization routine.
903  * Invoked once at EAL init time.
904  * Register itself as the [Poll Mode] Driver of PCI IXGBE devices.
905  */
906 int
907 rte_ixgbe_pmd_init(void)
908 {
909         PMD_INIT_FUNC_TRACE();
910
911         rte_eth_driver_register(&rte_ixgbe_pmd);
912         return 0;
913 }
914
915 /*
916  * VF Driver initialization routine.
917  * Invoked one at EAL init time.
918  * Register itself as the [Virtual Poll Mode] Driver of PCI niantic devices.
919  */
920 int
921 rte_ixgbevf_pmd_init(void)
922 {
923         DEBUGFUNC("rte_ixgbevf_pmd_init");
924
925         rte_eth_driver_register(&rte_ixgbevf_pmd);
926         return (0);
927 }
928
929 static int
930 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
931 {
932         struct ixgbe_hw *hw =
933                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
934         struct ixgbe_vfta * shadow_vfta =
935                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
936         uint32_t vfta;
937         uint32_t vid_idx;
938         uint32_t vid_bit;
939
940         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
941         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
942         vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
943         if (on)
944                 vfta |= vid_bit;
945         else
946                 vfta &= ~vid_bit;
947         IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
948
949         /* update local VFTA copy */
950         shadow_vfta->vfta[vid_idx] = vfta;
951
952         return 0;
953 }
954
955 static void
956 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
957 {
958         if (on)
959                 ixgbe_vlan_hw_strip_enable(dev, queue);
960         else
961                 ixgbe_vlan_hw_strip_disable(dev, queue);
962 }
963
964 static void
965 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid)
966 {
967         struct ixgbe_hw *hw =
968                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
969
970         /* Only the high 16-bits is valid */
971         IXGBE_WRITE_REG(hw, IXGBE_EXVET, tpid << 16);
972 }
973
974 void
975 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
976 {
977         struct ixgbe_hw *hw =
978                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
979         uint32_t vlnctrl;
980
981         PMD_INIT_FUNC_TRACE();
982
983         /* Filter Table Disable */
984         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
985         vlnctrl &= ~IXGBE_VLNCTRL_VFE;
986
987         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
988 }
989
990 void
991 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
992 {
993         struct ixgbe_hw *hw =
994                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
995         struct ixgbe_vfta * shadow_vfta =
996                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
997         uint32_t vlnctrl;
998         uint16_t i;
999
1000         PMD_INIT_FUNC_TRACE();
1001
1002         /* Filter Table Enable */
1003         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1004         vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1005         vlnctrl |= IXGBE_VLNCTRL_VFE;
1006
1007         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1008
1009         /* write whatever is in local vfta copy */
1010         for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1011                 IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1012 }
1013
1014 static void 
1015 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1016 {
1017         struct ixgbe_hwstrip *hwstrip = 
1018                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1019
1020         if(queue >= IXGBE_MAX_RX_QUEUE_NUM)
1021                 return;
1022
1023         if (on)
1024                 IXGBE_SET_HWSTRIP(hwstrip, queue);
1025         else
1026                 IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1027 }
1028
1029 static void
1030 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
1031 {
1032         struct ixgbe_hw *hw =
1033                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1034         uint32_t ctrl;
1035
1036         PMD_INIT_FUNC_TRACE();
1037
1038         if (hw->mac.type == ixgbe_mac_82598EB) {
1039                 /* No queue level support */
1040                 PMD_INIT_LOG(INFO, "82598EB not support queue level hw strip");
1041                 return;
1042         }
1043         else {
1044                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1045                 ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1046                 ctrl &= ~IXGBE_RXDCTL_VME;
1047                 IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1048         }
1049         /* record those setting for HW strip per queue */
1050         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
1051 }
1052
1053 static void
1054 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
1055 {
1056         struct ixgbe_hw *hw =
1057                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1058         uint32_t ctrl;
1059
1060         PMD_INIT_FUNC_TRACE();
1061
1062         if (hw->mac.type == ixgbe_mac_82598EB) {
1063                 /* No queue level supported */
1064                 PMD_INIT_LOG(INFO, "82598EB not support queue level hw strip");
1065                 return;
1066         }
1067         else {
1068                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1069                 ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1070                 ctrl |= IXGBE_RXDCTL_VME;
1071                 IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1072         }
1073         /* record those setting for HW strip per queue */
1074         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
1075 }
1076
1077 void
1078 ixgbe_vlan_hw_strip_disable_all(struct rte_eth_dev *dev)
1079 {
1080         struct ixgbe_hw *hw =
1081                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1082         uint32_t ctrl;
1083         uint16_t i;
1084
1085         PMD_INIT_FUNC_TRACE();
1086
1087         if (hw->mac.type == ixgbe_mac_82598EB) {
1088                 ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1089                 ctrl &= ~IXGBE_VLNCTRL_VME;
1090                 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1091         }
1092         else {
1093                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1094                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
1095                         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
1096                         ctrl &= ~IXGBE_RXDCTL_VME;
1097                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(i), ctrl);
1098
1099                         /* record those setting for HW strip per queue */
1100                         ixgbe_vlan_hw_strip_bitmap_set(dev, i, 0);
1101                 }
1102         }
1103 }
1104
1105 void
1106 ixgbe_vlan_hw_strip_enable_all(struct rte_eth_dev *dev)
1107 {
1108         struct ixgbe_hw *hw =
1109                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1110         uint32_t ctrl;
1111         uint16_t i;
1112
1113         PMD_INIT_FUNC_TRACE();
1114
1115         if (hw->mac.type == ixgbe_mac_82598EB) {
1116                 ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1117                 ctrl |= IXGBE_VLNCTRL_VME;
1118                 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1119         }
1120         else {
1121                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1122                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
1123                         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
1124                         ctrl |= IXGBE_RXDCTL_VME;
1125                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(i), ctrl);
1126
1127                         /* record those setting for HW strip per queue */
1128                         ixgbe_vlan_hw_strip_bitmap_set(dev, i, 1);                      
1129                 }
1130         }
1131 }
1132
1133 static void
1134 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
1135 {
1136         struct ixgbe_hw *hw =
1137                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1138         uint32_t ctrl;
1139
1140         PMD_INIT_FUNC_TRACE();
1141
1142         /* DMATXCTRL: Geric Double VLAN Disable */
1143         ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1144         ctrl &= ~IXGBE_DMATXCTL_GDV;
1145         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
1146
1147         /* CTRL_EXT: Global Double VLAN Disable */
1148         ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1149         ctrl &= ~IXGBE_EXTENDED_VLAN;
1150         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
1151
1152 }
1153
1154 static void
1155 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
1156 {
1157         struct ixgbe_hw *hw =
1158                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1159         uint32_t ctrl;
1160
1161         PMD_INIT_FUNC_TRACE();
1162
1163         /* DMATXCTRL: Geric Double VLAN Enable */
1164         ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1165         ctrl |= IXGBE_DMATXCTL_GDV;
1166         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
1167
1168         /* CTRL_EXT: Global Double VLAN Enable */
1169         ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1170         ctrl |= IXGBE_EXTENDED_VLAN;
1171         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
1172
1173         /*
1174          * VET EXT field in the EXVET register = 0x8100 by default
1175          * So no need to change. Same to VT field of DMATXCTL register
1176          */
1177 }
1178
1179 static void
1180 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1181 {
1182         if(mask & ETH_VLAN_STRIP_MASK){
1183                 if (dev->data->dev_conf.rxmode.hw_vlan_strip)
1184                         ixgbe_vlan_hw_strip_enable_all(dev);
1185                 else
1186                         ixgbe_vlan_hw_strip_disable_all(dev);
1187         }
1188
1189         if(mask & ETH_VLAN_FILTER_MASK){
1190                 if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1191                         ixgbe_vlan_hw_filter_enable(dev);
1192                 else
1193                         ixgbe_vlan_hw_filter_disable(dev);
1194         }
1195
1196         if(mask & ETH_VLAN_EXTEND_MASK){
1197                 if (dev->data->dev_conf.rxmode.hw_vlan_extend)
1198                         ixgbe_vlan_hw_extend_enable(dev);
1199                 else
1200                         ixgbe_vlan_hw_extend_disable(dev);
1201         }
1202 }
1203
1204 static void
1205 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1206 {
1207         struct ixgbe_hw *hw =
1208                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1209         /* VLNCTRL: enable vlan filtering and allow all vlan tags through */
1210         uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1211         vlanctrl |= IXGBE_VLNCTRL_VFE ; /* enable vlan filters */
1212         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
1213 }
1214
1215 static int
1216 ixgbe_dev_configure(struct rte_eth_dev *dev)
1217 {
1218         struct ixgbe_interrupt *intr =
1219                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1220
1221         PMD_INIT_FUNC_TRACE();
1222
1223         /* set flag to update link status after init */
1224         intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
1225
1226         return 0;
1227 }
1228
1229 /*
1230  * Configure device link speed and setup link.
1231  * It returns 0 on success.
1232  */
1233 static int
1234 ixgbe_dev_start(struct rte_eth_dev *dev)
1235 {
1236         struct ixgbe_hw *hw =
1237                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1238         int err, link_up = 0, negotiate = 0;
1239         uint32_t speed = 0;
1240         int mask = 0;
1241         int status;
1242         
1243         PMD_INIT_FUNC_TRACE();
1244
1245         /* IXGBE devices don't support half duplex */
1246         if ((dev->data->dev_conf.link_duplex != ETH_LINK_AUTONEG_DUPLEX) &&
1247                         (dev->data->dev_conf.link_duplex != ETH_LINK_FULL_DUPLEX)) {
1248                 PMD_INIT_LOG(ERR, "Invalid link_duplex (%u) for port %u\n",
1249                                 dev->data->dev_conf.link_duplex,
1250                                 dev->data->port_id);
1251                 return -EINVAL;
1252         }
1253
1254         /* stop adapter */
1255         hw->adapter_stopped = FALSE;
1256         ixgbe_stop_adapter(hw);
1257
1258         /* reinitialize adapter
1259          * this calls reset and start */
1260         status = ixgbe_pf_reset_hw(hw);
1261         if (status != 0)
1262                 return -1;
1263         hw->mac.ops.start_hw(hw);
1264
1265         /* configure PF module if SRIOV enabled */
1266         ixgbe_pf_host_configure(dev);
1267
1268         /* initialize transmission unit */
1269         ixgbe_dev_tx_init(dev);
1270       
1271         /* This can fail when allocating mbufs for descriptor rings */
1272         err = ixgbe_dev_rx_init(dev);
1273         if (err) {
1274                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware\n");
1275                 goto error;
1276         }
1277
1278         ixgbe_dev_rxtx_start(dev);
1279
1280         if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
1281                 err = hw->mac.ops.setup_sfp(hw);
1282                 if (err)
1283                         goto error;
1284         }
1285
1286         /* Turn on the laser */
1287         ixgbe_enable_tx_laser(hw);
1288
1289         err = ixgbe_check_link(hw, &speed, &link_up, 0);
1290         if (err)
1291                 goto error;
1292         err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
1293         if (err)
1294                 goto error;
1295
1296         switch(dev->data->dev_conf.link_speed) {
1297         case ETH_LINK_SPEED_AUTONEG:
1298                 speed = (hw->mac.type != ixgbe_mac_82598EB) ?
1299                                 IXGBE_LINK_SPEED_82599_AUTONEG :
1300                                 IXGBE_LINK_SPEED_82598_AUTONEG;
1301                 break;
1302         case ETH_LINK_SPEED_100:
1303                 /*
1304                  * Invalid for 82598 but error will be detected by
1305                  * ixgbe_setup_link()
1306                  */
1307                 speed = IXGBE_LINK_SPEED_100_FULL;
1308                 break;
1309         case ETH_LINK_SPEED_1000:
1310                 speed = IXGBE_LINK_SPEED_1GB_FULL;
1311                 break;
1312         case ETH_LINK_SPEED_10000:
1313                 speed = IXGBE_LINK_SPEED_10GB_FULL;
1314                 break;
1315         default:
1316                 PMD_INIT_LOG(ERR, "Invalid link_speed (%u) for port %u\n",
1317                                 dev->data->dev_conf.link_speed, dev->data->port_id);
1318                 goto error;
1319         }
1320
1321         err = ixgbe_setup_link(hw, speed, negotiate, link_up);
1322         if (err)
1323                 goto error;
1324
1325         /* check if lsc interrupt is enabled */
1326         if (dev->data->dev_conf.intr_conf.lsc != 0)
1327                 ixgbe_dev_lsc_interrupt_setup(dev);
1328
1329         /* resume enabled intr since hw reset */
1330         ixgbe_enable_intr(dev);
1331
1332         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
1333                 ETH_VLAN_EXTEND_MASK;
1334         ixgbe_vlan_offload_set(dev, mask);
1335
1336         if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1337                 /* Enable vlan filtering for VMDq */
1338                 ixgbe_vmdq_vlan_hw_filter_enable(dev);
1339         }       
1340
1341         /* Configure DCB hw */
1342         ixgbe_configure_dcb(dev); 
1343
1344         if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
1345                 err = ixgbe_fdir_configure(dev);
1346                 if (err)
1347                         goto error;
1348         }
1349
1350         ixgbe_restore_statistics_mapping(dev);
1351
1352         return (0);
1353
1354 error:
1355         PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
1356         ixgbe_dev_clear_queues(dev);
1357         return -EIO;
1358 }
1359
1360 /*
1361  * Stop device: disable rx and tx functions to allow for reconfiguring.
1362  */
1363 static void
1364 ixgbe_dev_stop(struct rte_eth_dev *dev)
1365 {
1366         struct rte_eth_link link;
1367         struct ixgbe_hw *hw =
1368                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1369
1370         PMD_INIT_FUNC_TRACE();
1371
1372         /* disable interrupts */
1373         ixgbe_disable_intr(hw);
1374
1375         /* reset the NIC */
1376         ixgbe_pf_reset_hw(hw);
1377         hw->adapter_stopped = FALSE;
1378
1379         /* stop adapter */
1380         ixgbe_stop_adapter(hw);
1381
1382         /* Turn off the laser */
1383         ixgbe_disable_tx_laser(hw);
1384
1385         ixgbe_dev_clear_queues(dev);
1386
1387         /* Clear recorded link status */
1388         memset(&link, 0, sizeof(link));
1389         rte_ixgbe_dev_atomic_write_link_status(dev, &link);
1390 }
1391
1392 /*
1393  * Reest and stop device.
1394  */
1395 static void
1396 ixgbe_dev_close(struct rte_eth_dev *dev)
1397 {
1398         struct ixgbe_hw *hw =
1399                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1400
1401         PMD_INIT_FUNC_TRACE();
1402
1403         ixgbe_pf_reset_hw(hw);
1404
1405         ixgbe_dev_stop(dev);
1406         hw->adapter_stopped = 1;
1407
1408         ixgbe_disable_pcie_master(hw);
1409
1410         /* reprogram the RAR[0] in case user changed it. */
1411         ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
1412 }
1413
1414 /*
1415  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
1416  */
1417 static void
1418 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1419 {
1420         struct ixgbe_hw *hw =
1421                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1422         struct ixgbe_hw_stats *hw_stats =
1423                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1424         uint32_t bprc, lxon, lxoff, total;
1425         uint64_t total_missed_rx, total_qbrc, total_qprc;
1426         unsigned i;
1427
1428         total_missed_rx = 0;
1429         total_qbrc = 0;
1430         total_qprc = 0;
1431
1432         hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
1433         hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
1434         hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
1435         hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
1436
1437         for (i = 0; i < 8; i++) {
1438                 uint32_t mp;
1439                 mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
1440                 /* global total per queue */
1441                 hw_stats->mpc[i] += mp;
1442                 /* Running comprehensive total for stats display */
1443                 total_missed_rx += hw_stats->mpc[i];
1444                 if (hw->mac.type == ixgbe_mac_82598EB)
1445                         hw_stats->rnbc[i] +=
1446                             IXGBE_READ_REG(hw, IXGBE_RNBC(i));
1447                 hw_stats->pxontxc[i] +=
1448                     IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
1449                 hw_stats->pxonrxc[i] +=
1450                     IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
1451                 hw_stats->pxofftxc[i] +=
1452                     IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
1453                 hw_stats->pxoffrxc[i] +=
1454                     IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
1455                 hw_stats->pxon2offc[i] +=
1456                     IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
1457         }
1458         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
1459                 hw_stats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
1460                 hw_stats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
1461                 hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
1462                 hw_stats->qbrc[i] +=
1463                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
1464                 hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
1465                 hw_stats->qbtc[i] +=
1466                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
1467                 hw_stats->qprdc[i] += IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
1468
1469                 total_qprc += hw_stats->qprc[i];
1470                 total_qbrc += hw_stats->qbrc[i];
1471         }
1472         hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
1473         hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
1474         hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
1475
1476         /* Note that gprc counts missed packets */
1477         hw_stats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
1478
1479         if (hw->mac.type != ixgbe_mac_82598EB) {
1480                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
1481                 hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
1482                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
1483                 hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
1484                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
1485                 hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
1486                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
1487                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
1488         } else {
1489                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
1490                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
1491                 /* 82598 only has a counter in the high register */
1492                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
1493                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
1494                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
1495         }
1496
1497         /*
1498          * Workaround: mprc hardware is incorrectly counting
1499          * broadcasts, so for now we subtract those.
1500          */
1501         bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
1502         hw_stats->bprc += bprc;
1503         hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
1504         if (hw->mac.type == ixgbe_mac_82598EB)
1505                 hw_stats->mprc -= bprc;
1506
1507         hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
1508         hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
1509         hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
1510         hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
1511         hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
1512         hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
1513
1514         lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
1515         hw_stats->lxontxc += lxon;
1516         lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
1517         hw_stats->lxofftxc += lxoff;
1518         total = lxon + lxoff;
1519
1520         hw_stats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
1521         hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
1522         hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
1523         hw_stats->gptc -= total;
1524         hw_stats->mptc -= total;
1525         hw_stats->ptc64 -= total;
1526         hw_stats->gotc -= total * ETHER_MIN_LEN;
1527
1528         hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
1529         hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
1530         hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
1531         hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
1532         hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
1533         hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
1534         hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
1535         hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
1536         hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
1537         hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
1538         hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
1539         hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
1540         hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
1541         hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
1542         hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
1543         hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
1544         hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
1545         hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
1546         /* Only read FCOE on 82599 */
1547         if (hw->mac.type != ixgbe_mac_82598EB) {
1548                 hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
1549                 hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
1550                 hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
1551                 hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
1552                 hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
1553         }
1554
1555         if (stats == NULL)
1556                 return;
1557
1558         /* Fill out the rte_eth_stats statistics structure */
1559         stats->ipackets = total_qprc;
1560         stats->ibytes = total_qbrc;
1561         stats->opackets = hw_stats->gptc;
1562         stats->obytes = hw_stats->gotc;
1563         stats->imcasts = hw_stats->mprc;
1564
1565         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
1566                 stats->q_ipackets[i] = hw_stats->qprc[i];
1567                 stats->q_opackets[i] = hw_stats->qptc[i];
1568                 stats->q_ibytes[i] = hw_stats->qbrc[i];
1569                 stats->q_obytes[i] = hw_stats->qbtc[i];
1570                 stats->q_errors[i] = hw_stats->qprdc[i];
1571         }
1572
1573         /* Rx Errors */
1574         stats->ierrors = total_missed_rx + hw_stats->crcerrs +
1575                 hw_stats->rlec;
1576
1577         stats->oerrors  = 0;
1578
1579         /* Flow Director Stats registers */
1580         hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
1581         hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
1582         stats->fdirmatch = hw_stats->fdirmatch;
1583         stats->fdirmiss = hw_stats->fdirmiss;
1584 }
1585
1586 static void
1587 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
1588 {
1589         struct ixgbe_hw_stats *stats =
1590                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1591
1592         /* HW registers are cleared on read */
1593         ixgbe_dev_stats_get(dev, NULL);
1594
1595         /* Reset software totals */
1596         memset(stats, 0, sizeof(*stats));
1597 }
1598
1599 static void
1600 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1601 {
1602         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1603         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats*)
1604                           IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1605
1606         /* Good Rx packet, include VF loopback */
1607         UPDATE_VF_STAT(IXGBE_VFGPRC,
1608             hw_stats->last_vfgprc, hw_stats->vfgprc);
1609
1610         /* Good Rx octets, include VF loopback */
1611         UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
1612             hw_stats->last_vfgorc, hw_stats->vfgorc);
1613
1614         /* Good Tx packet, include VF loopback */
1615         UPDATE_VF_STAT(IXGBE_VFGPTC,
1616             hw_stats->last_vfgptc, hw_stats->vfgptc);
1617
1618         /* Good Tx octets, include VF loopback */
1619         UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
1620             hw_stats->last_vfgotc, hw_stats->vfgotc);
1621
1622         /* Rx Multicst Packet */
1623         UPDATE_VF_STAT(IXGBE_VFMPRC,
1624             hw_stats->last_vfmprc, hw_stats->vfmprc);
1625
1626         if (stats == NULL)
1627                 return;
1628
1629         memset(stats, 0, sizeof(*stats));
1630         stats->ipackets = hw_stats->vfgprc;
1631         stats->ibytes = hw_stats->vfgorc;
1632         stats->opackets = hw_stats->vfgptc;
1633         stats->obytes = hw_stats->vfgotc;
1634         stats->imcasts = hw_stats->vfmprc;
1635 }
1636
1637 static void
1638 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
1639 {
1640         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats*)
1641                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1642
1643         /* Sync HW register to the last stats */
1644         ixgbevf_dev_stats_get(dev, NULL);
1645
1646         /* reset HW current stats*/
1647         hw_stats->vfgprc = 0;
1648         hw_stats->vfgorc = 0;
1649         hw_stats->vfgptc = 0;
1650         hw_stats->vfgotc = 0;
1651         hw_stats->vfmprc = 0;
1652
1653 }
1654
1655 static void
1656 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1657 {
1658         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1659
1660         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
1661         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
1662         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
1663         dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
1664         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
1665         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
1666         dev_info->max_vfs = dev->pci_dev->max_vfs;
1667         if (hw->mac.type == ixgbe_mac_82598EB)
1668                 dev_info->max_vmdq_pools = ETH_16_POOLS;
1669         else
1670                 dev_info->max_vmdq_pools = ETH_64_POOLS;
1671 }
1672
1673 /* return 0 means link status changed, -1 means not changed */
1674 static int
1675 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1676 {
1677         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1678         struct rte_eth_link link, old;
1679         ixgbe_link_speed link_speed;
1680         int link_up;
1681         int diag;
1682
1683         link.link_status = 0;
1684         link.link_speed = 0;
1685         link.link_duplex = 0;
1686         memset(&old, 0, sizeof(old));
1687         rte_ixgbe_dev_atomic_read_link_status(dev, &old);
1688
1689         /* check if it needs to wait to complete, if lsc interrupt is enabled */
1690         if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
1691                 diag = ixgbe_check_link(hw, &link_speed, &link_up, 0);
1692         else
1693                 diag = ixgbe_check_link(hw, &link_speed, &link_up, 1);
1694         if (diag != 0) {
1695                 link.link_speed = ETH_LINK_SPEED_100;
1696                 link.link_duplex = ETH_LINK_HALF_DUPLEX;
1697                 rte_ixgbe_dev_atomic_write_link_status(dev, &link);
1698                 if (link.link_status == old.link_status)
1699                         return -1;
1700                 return 0;
1701         }
1702
1703         if (link_up == 0) {
1704                 rte_ixgbe_dev_atomic_write_link_status(dev, &link);
1705                 if (link.link_status == old.link_status)
1706                         return -1;
1707                 return 0;
1708         }
1709         link.link_status = 1;
1710         link.link_duplex = ETH_LINK_FULL_DUPLEX;
1711
1712         switch (link_speed) {
1713         default:
1714         case IXGBE_LINK_SPEED_UNKNOWN:
1715                 link.link_duplex = ETH_LINK_HALF_DUPLEX;
1716                 link.link_speed = ETH_LINK_SPEED_100;
1717                 break;
1718
1719         case IXGBE_LINK_SPEED_100_FULL:
1720                 link.link_speed = ETH_LINK_SPEED_100;
1721                 break;
1722
1723         case IXGBE_LINK_SPEED_1GB_FULL:
1724                 link.link_speed = ETH_LINK_SPEED_1000;
1725                 break;
1726
1727         case IXGBE_LINK_SPEED_10GB_FULL:
1728                 link.link_speed = ETH_LINK_SPEED_10000;
1729                 break;
1730         }
1731         rte_ixgbe_dev_atomic_write_link_status(dev, &link);
1732
1733         if (link.link_status == old.link_status)
1734                 return -1;
1735
1736         return 0;
1737 }
1738
1739 static void
1740 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
1741 {
1742         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1743         uint32_t fctrl;
1744
1745         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
1746         fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
1747         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
1748 }
1749
1750 static void
1751 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
1752 {
1753         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1754         uint32_t fctrl;
1755
1756         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
1757         fctrl &= (~IXGBE_FCTRL_UPE);
1758         if (dev->data->all_multicast == 1)
1759                 fctrl |= IXGBE_FCTRL_MPE;
1760         else
1761                 fctrl &= (~IXGBE_FCTRL_MPE);
1762         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
1763 }
1764
1765 static void
1766 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
1767 {
1768         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1769         uint32_t fctrl;
1770
1771         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
1772         fctrl |= IXGBE_FCTRL_MPE;
1773         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
1774 }
1775
1776 static void
1777 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
1778 {
1779         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1780         uint32_t fctrl;
1781
1782         if (dev->data->promiscuous == 1)
1783                 return; /* must remain in all_multicast mode */
1784
1785         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
1786         fctrl &= (~IXGBE_FCTRL_MPE);
1787         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
1788 }
1789
1790 /**
1791  * It clears the interrupt causes and enables the interrupt.
1792  * It will be called once only during nic initialized.
1793  *
1794  * @param dev
1795  *  Pointer to struct rte_eth_dev.
1796  *
1797  * @return
1798  *  - On success, zero.
1799  *  - On failure, a negative value.
1800  */
1801 static int
1802 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev)
1803 {
1804         struct ixgbe_interrupt *intr =
1805                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1806
1807         ixgbe_dev_link_status_print(dev);
1808         intr->mask |= IXGBE_EICR_LSC;
1809
1810         return 0;
1811 }
1812
1813 /*
1814  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
1815  *
1816  * @param dev
1817  *  Pointer to struct rte_eth_dev.
1818  *
1819  * @return
1820  *  - On success, zero.
1821  *  - On failure, a negative value.
1822  */
1823 static int
1824 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
1825 {
1826         uint32_t eicr;
1827         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1828         struct ixgbe_interrupt *intr =
1829                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1830
1831         /* clear all cause mask */
1832         ixgbe_disable_intr(hw);
1833
1834         /* read-on-clear nic registers here */
1835         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
1836         PMD_DRV_LOG(INFO, "eicr %x", eicr);
1837         
1838         intr->flags = 0;
1839         if (eicr & IXGBE_EICR_LSC) {
1840                 /* set flag for async link update */
1841                 intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
1842         }
1843
1844         if (eicr & IXGBE_EICR_MAILBOX)
1845                 intr->flags |= IXGBE_FLAG_MAILBOX;
1846
1847         return 0;
1848 }
1849
1850 /**
1851  * It gets and then prints the link status.
1852  *
1853  * @param dev
1854  *  Pointer to struct rte_eth_dev.
1855  *
1856  * @return
1857  *  - On success, zero.
1858  *  - On failure, a negative value.
1859  */
1860 static void
1861 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
1862 {
1863         struct rte_eth_link link;
1864
1865         memset(&link, 0, sizeof(link));
1866         rte_ixgbe_dev_atomic_read_link_status(dev, &link);
1867         if (link.link_status) {
1868                 PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
1869                                         (int)(dev->data->port_id),
1870                                         (unsigned)link.link_speed,
1871                         link.link_duplex == ETH_LINK_FULL_DUPLEX ?
1872                                         "full-duplex" : "half-duplex");
1873         } else {
1874                 PMD_INIT_LOG(INFO, " Port %d: Link Down",
1875                                 (int)(dev->data->port_id));
1876         }
1877         PMD_INIT_LOG(INFO, "PCI Address: %04d:%02d:%02d:%d",
1878                                 dev->pci_dev->addr.domain,
1879                                 dev->pci_dev->addr.bus,
1880                                 dev->pci_dev->addr.devid,
1881                                 dev->pci_dev->addr.function);
1882 }
1883
1884 /*
1885  * It executes link_update after knowing an interrupt occured.
1886  *
1887  * @param dev
1888  *  Pointer to struct rte_eth_dev.
1889  *
1890  * @return
1891  *  - On success, zero.
1892  *  - On failure, a negative value.
1893  */
1894 static int
1895 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
1896 {
1897         struct ixgbe_interrupt *intr =
1898                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1899         int64_t timeout;
1900         struct rte_eth_link link;
1901         int intr_enable_delay = false;  
1902
1903         PMD_DRV_LOG(DEBUG, "intr action type %d\n", intr->flags);
1904
1905         if (intr->flags & IXGBE_FLAG_MAILBOX) {
1906                 ixgbe_pf_mbx_process(dev);
1907                 intr->flags &= ~IXGBE_FLAG_MAILBOX;
1908         } 
1909
1910         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
1911                 /* get the link status before link update, for predicting later */
1912                 memset(&link, 0, sizeof(link));
1913                 rte_ixgbe_dev_atomic_read_link_status(dev, &link);
1914
1915                 ixgbe_dev_link_update(dev, 0);
1916
1917                 /* likely to up */
1918                 if (!link.link_status)
1919                         /* handle it 1 sec later, wait it being stable */
1920                         timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
1921                 /* likely to down */
1922                 else
1923                         /* handle it 4 sec later, wait it being stable */
1924                         timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
1925                 
1926                 ixgbe_dev_link_status_print(dev);
1927
1928                 intr_enable_delay = true;
1929         } 
1930
1931         if (intr_enable_delay) {
1932                 if (rte_eal_alarm_set(timeout * 1000,
1933                                       ixgbe_dev_interrupt_delayed_handler, (void*)dev) < 0)
1934                         PMD_DRV_LOG(ERR, "Error setting alarm");
1935         } else {
1936                 PMD_DRV_LOG(DEBUG, "enable intr immediately");
1937                 ixgbe_enable_intr(dev);
1938                 rte_intr_enable(&(dev->pci_dev->intr_handle));
1939         }
1940                         
1941
1942         return 0;
1943 }
1944
1945 /**
1946  * Interrupt handler which shall be registered for alarm callback for delayed
1947  * handling specific interrupt to wait for the stable nic state. As the
1948  * NIC interrupt state is not stable for ixgbe after link is just down,
1949  * it needs to wait 4 seconds to get the stable status.
1950  *
1951  * @param handle
1952  *  Pointer to interrupt handle.
1953  * @param param
1954  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1955  *
1956  * @return
1957  *  void
1958  */
1959 static void
1960 ixgbe_dev_interrupt_delayed_handler(void *param)
1961 {
1962         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1963         struct ixgbe_interrupt *intr =
1964                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1965         struct ixgbe_hw *hw =
1966                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1967         uint32_t eicr;
1968
1969         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
1970         if (eicr & IXGBE_EICR_MAILBOX)
1971                 ixgbe_pf_mbx_process(dev);
1972
1973         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
1974                 ixgbe_dev_link_update(dev, 0);
1975                 intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
1976                 ixgbe_dev_link_status_print(dev);
1977                 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC);
1978         }
1979
1980         PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]\n", eicr);
1981         ixgbe_enable_intr(dev);
1982         rte_intr_enable(&(dev->pci_dev->intr_handle));
1983 }
1984
1985 /**
1986  * Interrupt handler triggered by NIC  for handling
1987  * specific interrupt.
1988  *
1989  * @param handle
1990  *  Pointer to interrupt handle.
1991  * @param param
1992  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1993  *
1994  * @return
1995  *  void
1996  */
1997 static void
1998 ixgbe_dev_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
1999                                                         void *param)
2000 {
2001         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2002         ixgbe_dev_interrupt_get_status(dev);
2003         ixgbe_dev_interrupt_action(dev);
2004 }
2005
2006 static int
2007 ixgbe_dev_led_on(struct rte_eth_dev *dev)
2008 {
2009         struct ixgbe_hw *hw;
2010
2011         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2012         return (ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP);
2013 }
2014
2015 static int
2016 ixgbe_dev_led_off(struct rte_eth_dev *dev)
2017 {
2018         struct ixgbe_hw *hw;
2019
2020         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2021         return (ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP);
2022 }
2023
2024 static int
2025 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2026 {
2027         struct ixgbe_hw *hw;
2028         int err;
2029         uint32_t rx_buf_size;
2030         uint32_t max_high_water;
2031         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
2032                 ixgbe_fc_none,
2033                 ixgbe_fc_rx_pause,
2034                 ixgbe_fc_tx_pause,
2035                 ixgbe_fc_full
2036         };
2037
2038         PMD_INIT_FUNC_TRACE();
2039
2040         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2041         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
2042         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x \n", rx_buf_size);
2043
2044         /*
2045          * At least reserve one Ethernet frame for watermark
2046          * high_water/low_water in kilo bytes for ixgbe
2047          */
2048         max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
2049         if ((fc_conf->high_water > max_high_water) ||
2050                 (fc_conf->high_water < fc_conf->low_water)) {
2051                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB\n");
2052                 PMD_INIT_LOG(ERR, "High_water must <=  0x%x\n", max_high_water);
2053                 return (-EINVAL);
2054         }
2055
2056         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
2057         hw->fc.pause_time     = fc_conf->pause_time;
2058         hw->fc.high_water[0]  = fc_conf->high_water;
2059         hw->fc.low_water[0]   = fc_conf->low_water;
2060         hw->fc.send_xon       = fc_conf->send_xon;
2061
2062         err = ixgbe_fc_enable(hw);
2063         /* Not negotiated is not an error case */
2064         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED)) {
2065                 return 0;
2066         }
2067
2068         PMD_INIT_LOG(ERR, "ixgbe_fc_enable = 0x%x \n", err);
2069         return -EIO;
2070 }
2071
2072 /**
2073  *  ixgbe_pfc_enable_generic - Enable flow control
2074  *  @hw: pointer to hardware structure
2075  *  @tc_num: traffic class number
2076  *  Enable flow control according to the current settings.
2077  */
2078 static int 
2079 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw,uint8_t tc_num)
2080 {
2081         int ret_val = 0;
2082         uint32_t mflcn_reg, fccfg_reg;
2083         uint32_t reg;
2084         uint32_t fcrtl, fcrth;
2085         uint8_t i;
2086         uint8_t nb_rx_en;
2087         
2088         /* Validate the water mark configuration */
2089         if (!hw->fc.pause_time) {
2090                 ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
2091                 goto out;
2092         }
2093
2094         /* Low water mark of zero causes XOFF floods */
2095         if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
2096                  /* High/Low water can not be 0 */
2097                 if( (!hw->fc.high_water[tc_num])|| (!hw->fc.low_water[tc_num])) {
2098                         PMD_INIT_LOG(ERR,"Invalid water mark configuration\n");
2099                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
2100                         goto out;
2101                 }
2102  
2103                 if(hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
2104                         PMD_INIT_LOG(ERR,"Invalid water mark configuration\n");
2105                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
2106                         goto out;
2107                 }
2108         }
2109         /* Negotiate the fc mode to use */
2110         ixgbe_fc_autoneg(hw);
2111
2112         /* Disable any previous flow control settings */
2113         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
2114         mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
2115
2116         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
2117         fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
2118
2119         switch (hw->fc.current_mode) {
2120         case ixgbe_fc_none:
2121                 /*
2122                  * If the count of enabled RX Priority Flow control >1,
2123                  * and the TX pause can not be disabled 
2124                  */
2125                 nb_rx_en = 0;
2126                 for (i =0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
2127                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
2128                         if (reg & IXGBE_FCRTH_FCEN)
2129                                 nb_rx_en++;
2130                 }
2131                 if (nb_rx_en > 1)
2132                         fccfg_reg |=IXGBE_FCCFG_TFCE_PRIORITY;
2133                 break;
2134         case ixgbe_fc_rx_pause:
2135                 /*
2136                  * Rx Flow control is enabled and Tx Flow control is
2137                  * disabled by software override. Since there really
2138                  * isn't a way to advertise that we are capable of RX
2139                  * Pause ONLY, we will advertise that we support both
2140                  * symmetric and asymmetric Rx PAUSE.  Later, we will
2141                  * disable the adapter's ability to send PAUSE frames.
2142                  */
2143                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
2144                 /*
2145                  * If the count of enabled RX Priority Flow control >1,
2146                  * and the TX pause can not be disabled
2147                  */
2148                 nb_rx_en = 0;
2149                 for (i =0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
2150                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
2151                         if (reg & IXGBE_FCRTH_FCEN)
2152                                 nb_rx_en++;
2153                 }
2154                 if (nb_rx_en > 1)
2155                         fccfg_reg |=IXGBE_FCCFG_TFCE_PRIORITY;
2156                 break;
2157         case ixgbe_fc_tx_pause:
2158                 /*
2159                  * Tx Flow control is enabled, and Rx Flow control is
2160                  * disabled by software override.
2161                  */
2162                 fccfg_reg |=IXGBE_FCCFG_TFCE_PRIORITY;
2163                 break;
2164         case ixgbe_fc_full:
2165                 /* Flow control (both Rx and Tx) is enabled by SW override. */
2166                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
2167                 fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
2168                 break;
2169         default:
2170                 DEBUGOUT("Flow control param set incorrectly\n");
2171                 ret_val = IXGBE_ERR_CONFIG;
2172                 goto out;
2173                 break;
2174         }
2175
2176         /* Set 802.3x based flow control settings. */
2177         mflcn_reg |= IXGBE_MFLCN_DPF;
2178         IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
2179         IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
2180
2181         /* Set up and enable Rx high/low water mark thresholds, enable XON. */
2182         if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
2183                 hw->fc.high_water[tc_num]) {
2184                 fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
2185                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
2186                 fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
2187         } else {
2188                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
2189                 /*
2190                  * In order to prevent Tx hangs when the internal Tx
2191                  * switch is enabled we must set the high water mark
2192                  * to the maximum FCRTH value.  This allows the Tx
2193                  * switch to function even under heavy Rx workloads.
2194                  */
2195                 fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
2196         }
2197         IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
2198
2199         /* Configure pause time (2 TCs per register) */
2200         reg = hw->fc.pause_time * 0x00010001;
2201         for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
2202                 IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
2203
2204         /* Configure flow control refresh threshold value */
2205         IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
2206
2207 out:
2208         return ret_val;
2209 }
2210
2211 static int 
2212 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev,uint8_t tc_num)
2213 {
2214         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2215         int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
2216
2217         if(hw->mac.type != ixgbe_mac_82598EB) {
2218                 ret_val = ixgbe_dcb_pfc_enable_generic(hw,tc_num);
2219         }
2220         return ret_val;
2221 }
2222
2223 static int 
2224 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
2225 {
2226         int err;
2227         uint32_t rx_buf_size;
2228         uint32_t max_high_water;
2229         uint8_t tc_num;
2230         uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
2231         struct ixgbe_hw *hw =
2232                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2233         struct ixgbe_dcb_config *dcb_config =
2234                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
2235         
2236         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
2237                 ixgbe_fc_none,
2238                 ixgbe_fc_rx_pause,
2239                 ixgbe_fc_tx_pause,
2240                 ixgbe_fc_full
2241         };
2242         
2243         PMD_INIT_FUNC_TRACE();
2244         
2245         ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
2246         tc_num = map[pfc_conf->priority];
2247         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
2248         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x \n", rx_buf_size);
2249         /*
2250          * At least reserve one Ethernet frame for watermark
2251          * high_water/low_water in kilo bytes for ixgbe
2252          */
2253         max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
2254         if ((pfc_conf->fc.high_water > max_high_water) ||
2255                 (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
2256                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB\n");
2257                 PMD_INIT_LOG(ERR, "High_water must <=  0x%x\n", max_high_water);
2258                 return (-EINVAL);
2259         }
2260
2261         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
2262         hw->fc.pause_time = pfc_conf->fc.pause_time;
2263         hw->fc.send_xon = pfc_conf->fc.send_xon;
2264         hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
2265         hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
2266                 
2267         err = ixgbe_dcb_pfc_enable(dev,tc_num);
2268         
2269         /* Not negotiated is not an error case */
2270         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED)) 
2271                 return 0;
2272
2273         PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x \n", err);
2274         return -EIO;
2275 }       
2276
2277 static int 
2278 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
2279                                 struct rte_eth_rss_reta *reta_conf)
2280 {       
2281         uint8_t i,j,mask;
2282         uint32_t reta;
2283         struct ixgbe_hw *hw = 
2284                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2285
2286         PMD_INIT_FUNC_TRACE();
2287         /*  
2288         * Update Redirection Table RETA[n],n=0...31,The redirection table has 
2289         * 128-entries in 32 registers
2290          */ 
2291         for(i = 0; i < ETH_RSS_RETA_NUM_ENTRIES; i += 4) {
2292                 if (i < ETH_RSS_RETA_NUM_ENTRIES/2) 
2293                         mask = (uint8_t)((reta_conf->mask_lo >> i) & 0xF);
2294                 else
2295                         mask = (uint8_t)((reta_conf->mask_hi >> 
2296                                 (i - ETH_RSS_RETA_NUM_ENTRIES/2)) & 0xF);
2297                 if (mask != 0) {
2298                         reta = 0;
2299                         if (mask != 0xF)
2300                                 reta = IXGBE_READ_REG(hw,IXGBE_RETA(i >> 2));
2301
2302                         for (j = 0; j < 4; j++) {
2303                                 if (mask & (0x1 << j)) {
2304                                         if (mask != 0xF)
2305                                                 reta &= ~(0xFF << 8 * j);
2306                                         reta |= reta_conf->reta[i + j] << 8*j;
2307                                 }
2308                         }
2309                         IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2),reta);
2310                 }
2311         }
2312
2313         return 0;
2314 }
2315
2316 static int
2317 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
2318                                 struct rte_eth_rss_reta *reta_conf)
2319 {
2320         uint8_t i,j,mask;
2321         uint32_t reta;
2322         struct ixgbe_hw *hw =
2323                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2324         
2325         PMD_INIT_FUNC_TRACE();
2326         /* 
2327          * Read Redirection Table RETA[n],n=0...31,The redirection table has 
2328          * 128-entries in 32 registers
2329          */
2330         for(i = 0; i < ETH_RSS_RETA_NUM_ENTRIES; i += 4) {
2331                 if (i < ETH_RSS_RETA_NUM_ENTRIES/2)
2332                         mask = (uint8_t)((reta_conf->mask_lo >> i) & 0xF);
2333                 else
2334                         mask = (uint8_t)((reta_conf->mask_hi >> 
2335                                 (i - ETH_RSS_RETA_NUM_ENTRIES/2)) & 0xF);
2336
2337                 if (mask != 0) {
2338                         reta = IXGBE_READ_REG(hw,IXGBE_RETA(i >> 2));
2339                         for (j = 0; j < 4; j++) {
2340                                 if (mask & (0x1 << j))
2341                                         reta_conf->reta[i + j] = 
2342                                                 (uint8_t)((reta >> 8 * j) & 0xFF);
2343                         } 
2344                 }
2345         }
2346
2347         return 0;               
2348 }
2349
2350 static void
2351 ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
2352                                 uint32_t index, uint32_t pool)
2353 {
2354         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2355         uint32_t enable_addr = 1;
2356
2357         ixgbe_set_rar(hw, index, mac_addr->addr_bytes, pool, enable_addr);
2358 }
2359
2360 static void
2361 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
2362 {
2363         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2364
2365         ixgbe_clear_rar(hw, index);
2366 }
2367
2368 /*
2369  * Virtual Function operations
2370  */
2371 static void
2372 ixgbevf_intr_disable(struct ixgbe_hw *hw)
2373 {
2374         PMD_INIT_LOG(DEBUG, "ixgbevf_intr_disable");
2375
2376         /* Clear interrupt mask to stop from interrupts being generated */
2377         IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
2378
2379         IXGBE_WRITE_FLUSH(hw);
2380 }
2381
2382 static int
2383 ixgbevf_dev_configure(struct rte_eth_dev *dev)
2384 {
2385         struct rte_eth_conf* conf = &dev->data->dev_conf;
2386
2387         PMD_INIT_LOG(DEBUG, "\nConfigured Virtual Function port id: %d\n",
2388                 dev->data->port_id);
2389
2390         /*
2391          * VF has no ability to enable/disable HW CRC
2392          * Keep the persistent behavior the same as Host PF
2393          */
2394 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
2395         if (!conf->rxmode.hw_strip_crc) {
2396                 PMD_INIT_LOG(INFO, "VF can't disable HW CRC Strip\n");
2397                 conf->rxmode.hw_strip_crc = 1;
2398         }
2399 #else
2400         if (conf->rxmode.hw_strip_crc) {
2401                 PMD_INIT_LOG(INFO, "VF can't enable HW CRC Strip\n");
2402                 conf->rxmode.hw_strip_crc = 0;
2403         }
2404 #endif
2405
2406         return 0;
2407 }
2408
2409 static int
2410 ixgbevf_dev_start(struct rte_eth_dev *dev)
2411 {
2412         struct ixgbe_hw *hw = 
2413                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2414         int err, mask = 0;
2415         
2416         PMD_INIT_LOG(DEBUG, "ixgbevf_dev_start");
2417
2418         hw->mac.ops.reset_hw(hw);
2419
2420         ixgbevf_dev_tx_init(dev);
2421
2422         /* This can fail when allocating mbufs for descriptor rings */
2423         err = ixgbevf_dev_rx_init(dev);
2424         if (err) {
2425                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)\n", err);
2426                 ixgbe_dev_clear_queues(dev);
2427                 return err;
2428         }
2429         
2430         /* Set vfta */
2431         ixgbevf_set_vfta_all(dev,1);
2432
2433         /* Set HW strip */
2434         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
2435                 ETH_VLAN_EXTEND_MASK;
2436         ixgbevf_vlan_offload_set(dev, mask);
2437
2438         ixgbevf_dev_rxtx_start(dev);
2439
2440         return 0;
2441 }
2442
2443 static void
2444 ixgbevf_dev_stop(struct rte_eth_dev *dev)
2445 {
2446         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2447
2448         PMD_INIT_LOG(DEBUG, "ixgbevf_dev_stop");
2449                 
2450         hw->adapter_stopped = TRUE;
2451         ixgbe_stop_adapter(hw);
2452
2453         /* 
2454           * Clear what we set, but we still keep shadow_vfta to 
2455           * restore after device starts
2456           */
2457         ixgbevf_set_vfta_all(dev,0);
2458
2459         ixgbe_dev_clear_queues(dev);
2460 }
2461
2462 static void
2463 ixgbevf_dev_close(struct rte_eth_dev *dev)
2464 {
2465         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2466
2467         PMD_INIT_LOG(DEBUG, "ixgbevf_dev_close");
2468
2469         ixgbe_reset_hw(hw);
2470
2471         ixgbevf_dev_stop(dev);
2472
2473         /* reprogram the RAR[0] in case user changed it. */
2474         ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
2475 }
2476
2477 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
2478 {
2479         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2480         struct ixgbe_vfta * shadow_vfta =
2481                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2482         int i = 0, j = 0, vfta = 0, mask = 1;
2483
2484         for (i = 0; i < IXGBE_VFTA_SIZE; i++){
2485                 vfta = shadow_vfta->vfta[i];
2486                 if(vfta){
2487                         mask = 1;
2488                         for (j = 0; j < 32; j++){
2489                                 if(vfta & mask)
2490                                         ixgbe_set_vfta(hw, (i<<5)+j, 0, on);
2491                                 mask<<=1;
2492                         }
2493                 }
2494         }
2495
2496 }
2497
2498 static int
2499 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2500 {
2501         struct ixgbe_hw *hw =
2502                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2503         struct ixgbe_vfta * shadow_vfta =
2504                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2505         uint32_t vid_idx = 0;
2506         uint32_t vid_bit = 0;
2507         int ret = 0;
2508         
2509         PMD_INIT_FUNC_TRACE();
2510
2511         /* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
2512         ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on);
2513         if(ret){
2514                 PMD_INIT_LOG(ERR, "Unable to set VF vlan");
2515                 return ret;
2516         }
2517         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
2518         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
2519
2520         /* Save what we set and retore it after device reset */
2521         if (on)
2522                 shadow_vfta->vfta[vid_idx] |= vid_bit;
2523         else
2524                 shadow_vfta->vfta[vid_idx] &= ~vid_bit;
2525
2526         return 0;
2527 }
2528
2529 static void
2530 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
2531 {
2532         struct ixgbe_hw *hw =
2533                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2534         uint32_t ctrl;
2535
2536         PMD_INIT_FUNC_TRACE();
2537         
2538         if(queue >= hw->mac.max_rx_queues)
2539                 return;
2540
2541         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2542         if(on)
2543                 ctrl |= IXGBE_RXDCTL_VME;
2544         else 
2545                 ctrl &= ~IXGBE_RXDCTL_VME;
2546         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2547
2548         ixgbe_vlan_hw_strip_bitmap_set( dev, queue, on);
2549 }
2550
2551 static void
2552 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2553 {
2554         struct ixgbe_hw *hw =
2555                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2556         uint16_t i;
2557         int on = 0;
2558
2559         /* VF function only support hw strip feature, others are not support */
2560         if(mask & ETH_VLAN_STRIP_MASK){
2561                 on = !!(dev->data->dev_conf.rxmode.hw_vlan_strip);
2562
2563                 for(i=0; i < hw->mac.max_rx_queues; i++)
2564                         ixgbevf_vlan_strip_queue_set(dev,i,on);
2565         }
2566 }
2567
2568 static int
2569 ixgbe_vmdq_mode_check(struct ixgbe_hw *hw)
2570 {
2571         uint32_t reg_val;
2572         
2573         /* we only need to do this if VMDq is enabled */
2574         reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
2575         if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
2576                 PMD_INIT_LOG(ERR, "VMDq must be enabled for this setting\n");
2577                 return (-1);
2578         }
2579         
2580         return 0;
2581 }
2582
2583 static uint32_t 
2584 ixgbe_uta_vector(struct ixgbe_hw *hw, struct ether_addr* uc_addr)
2585 {
2586         uint32_t vector = 0;
2587         switch (hw->mac.mc_filter_type) {
2588         case 0:   /* use bits [47:36] of the address */
2589                 vector = ((uc_addr->addr_bytes[4] >> 4) | 
2590                         (((uint16_t)uc_addr->addr_bytes[5]) << 4));
2591                 break;
2592         case 1:   /* use bits [46:35] of the address */
2593                 vector = ((uc_addr->addr_bytes[4] >> 3) | 
2594                         (((uint16_t)uc_addr->addr_bytes[5]) << 5));
2595                 break;
2596         case 2:   /* use bits [45:34] of the address */
2597                 vector = ((uc_addr->addr_bytes[4] >> 2) | 
2598                         (((uint16_t)uc_addr->addr_bytes[5]) << 6));
2599                 break;
2600         case 3:   /* use bits [43:32] of the address */
2601                 vector = ((uc_addr->addr_bytes[4]) | 
2602                         (((uint16_t)uc_addr->addr_bytes[5]) << 8));
2603                 break;
2604         default:  /* Invalid mc_filter_type */
2605                 break;
2606         }
2607
2608         /* vector can only be 12-bits or boundary will be exceeded */
2609         vector &= 0xFFF;
2610         return vector;
2611 }
2612
2613 static int 
2614 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev,struct ether_addr* mac_addr,
2615                                uint8_t on)
2616 {
2617         uint32_t vector;
2618         uint32_t uta_idx;
2619         uint32_t reg_val;
2620         uint32_t uta_shift;
2621         uint32_t rc;
2622         const uint32_t ixgbe_uta_idx_mask = 0x7F;
2623         const uint32_t ixgbe_uta_bit_shift = 5;
2624         const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
2625         const uint32_t bit1 = 0x1;
2626         
2627         struct ixgbe_hw *hw =
2628                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2629         struct ixgbe_uta_info *uta_info =
2630                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
2631         
2632         /* The UTA table only exists on 82599 hardware and newer */
2633         if (hw->mac.type < ixgbe_mac_82599EB)
2634                 return (-ENOTSUP);
2635         
2636         vector = ixgbe_uta_vector(hw,mac_addr);
2637         uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
2638         uta_shift = vector & ixgbe_uta_bit_mask;
2639         
2640         rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
2641         if(rc == on)
2642                 return 0;
2643         
2644         reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
2645         if (on) {
2646                 uta_info->uta_in_use++;
2647                 reg_val |= (bit1 << uta_shift);
2648                 uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
2649         } else {
2650                 uta_info->uta_in_use--;
2651                 reg_val &= ~(bit1 << uta_shift);
2652                 uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
2653         }
2654         
2655         IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
2656         
2657         if (uta_info->uta_in_use > 0)
2658                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
2659                                 IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
2660         else
2661                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,hw->mac.mc_filter_type);
2662         
2663         return 0;
2664 }
2665
2666 static int
2667 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
2668 {
2669         int i;
2670         struct ixgbe_hw *hw =
2671                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2672         struct ixgbe_uta_info *uta_info =
2673                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
2674
2675         /* The UTA table only exists on 82599 hardware and newer */
2676         if (hw->mac.type < ixgbe_mac_82599EB)
2677                 return (-ENOTSUP);
2678         
2679         if(on) {
2680                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
2681                         uta_info->uta_shadow[i] = ~0;
2682                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
2683                 }
2684         } else {
2685                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
2686                         uta_info->uta_shadow[i] = 0;
2687                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
2688                 }
2689         }
2690         return 0;
2691         
2692 }
2693 static int
2694 ixgbe_set_pool_rx_mode(struct rte_eth_dev *dev, uint16_t pool,
2695                                uint16_t rx_mask, uint8_t on)
2696 {
2697         int val = 0;
2698         
2699         struct ixgbe_hw *hw =
2700                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2701         uint32_t vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
2702         
2703         if (hw->mac.type == ixgbe_mac_82598EB) {
2704                 PMD_INIT_LOG(ERR, "setting VF receive mode set should be done"
2705                         " on 82599 hardware and newer\n");
2706                 return (-ENOTSUP);
2707         }
2708         if (ixgbe_vmdq_mode_check(hw) < 0)
2709                 return (-ENOTSUP);
2710
2711         if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG )
2712                 val |= IXGBE_VMOLR_AUPE;
2713         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC )
2714                 val |= IXGBE_VMOLR_ROMPE;
2715         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
2716                 val |= IXGBE_VMOLR_ROPE;
2717         if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
2718                 val |= IXGBE_VMOLR_BAM;
2719         if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
2720                 val |= IXGBE_VMOLR_MPE;
2721
2722         if (on)
2723                 vmolr |= val;
2724         else 
2725                 vmolr &= ~val;
2726
2727         IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
2728         
2729         return 0;
2730 }
2731
2732 static int
2733 ixgbe_set_pool_rx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on)
2734 {
2735         uint32_t reg,addr;
2736         uint32_t val;
2737         const uint8_t bit1 = 0x1;
2738         
2739         struct ixgbe_hw *hw =
2740                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2741
2742         if (ixgbe_vmdq_mode_check(hw) < 0)
2743                 return (-ENOTSUP);
2744         
2745         addr = IXGBE_VFRE(pool >= ETH_64_POOLS/2);
2746         reg = IXGBE_READ_REG(hw, addr);
2747         val = bit1 << pool;
2748
2749         if (on)
2750                 reg |= val;
2751         else
2752                 reg &= ~val;
2753         
2754         IXGBE_WRITE_REG(hw, addr,reg);
2755         
2756         return 0;
2757 }
2758
2759 static int
2760 ixgbe_set_pool_tx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on)
2761 {
2762         uint32_t reg,addr;
2763         uint32_t val;
2764         const uint8_t bit1 = 0x1;
2765         
2766         struct ixgbe_hw *hw =
2767                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2768
2769         if (ixgbe_vmdq_mode_check(hw) < 0)
2770                 return (-ENOTSUP);
2771         
2772         addr = IXGBE_VFTE(pool >= ETH_64_POOLS/2);
2773         reg = IXGBE_READ_REG(hw, addr);
2774         val = bit1 << pool;
2775
2776         if (on)
2777                 reg |= val;
2778         else
2779                 reg &= ~val;
2780         
2781         IXGBE_WRITE_REG(hw, addr,reg);
2782         
2783         return 0;
2784 }
2785
2786 static int 
2787 ixgbe_set_pool_vlan_filter(struct rte_eth_dev *dev, uint16_t vlan,
2788                         uint64_t pool_mask, uint8_t vlan_on)
2789 {
2790         int ret = 0;
2791         uint16_t pool_idx;
2792         struct ixgbe_hw *hw =
2793                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2794         
2795         if (ixgbe_vmdq_mode_check(hw) < 0)
2796                 return (-ENOTSUP);
2797         for (pool_idx = 0; pool_idx < ETH_64_POOLS; pool_idx++) {
2798                 if (pool_mask & ((uint64_t)(1ULL << pool_idx))) 
2799                         ret = hw->mac.ops.set_vfta(hw,vlan,pool_idx,vlan_on);
2800                         if (ret < 0) 
2801                                 return ret;     
2802         }
2803
2804         return ret;
2805 }
2806
2807 static int
2808 ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
2809                         struct rte_eth_vmdq_mirror_conf *mirror_conf, 
2810                         uint8_t rule_id, uint8_t on)
2811 {
2812         uint32_t mr_ctl,vlvf;
2813         uint32_t mp_lsb = 0;
2814         uint32_t mv_msb = 0;
2815         uint32_t mv_lsb = 0;
2816         uint32_t mp_msb = 0;
2817         uint8_t i = 0;
2818         int reg_index = 0;
2819         uint64_t vlan_mask = 0;
2820         
2821         const uint8_t pool_mask_offset = 32;
2822         const uint8_t vlan_mask_offset = 32;
2823         const uint8_t dst_pool_offset = 8;
2824         const uint8_t rule_mr_offset  = 4;
2825         const uint8_t mirror_rule_mask= 0x0F;
2826
2827         struct ixgbe_mirror_info *mr_info =
2828                         (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
2829         struct ixgbe_hw *hw =
2830                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2831
2832         if (ixgbe_vmdq_mode_check(hw) < 0)
2833                 return (-ENOTSUP);
2834
2835         /* Check if vlan mask is valid */
2836         if ((mirror_conf->rule_type_mask & ETH_VMDQ_VLAN_MIRROR) && (on)) {
2837                 if (mirror_conf->vlan.vlan_mask == 0)
2838                         return (-EINVAL);
2839         }
2840
2841         /* Check if vlan id is valid and find conresponding VLAN ID index in VLVF */
2842         if (mirror_conf->rule_type_mask & ETH_VMDQ_VLAN_MIRROR) {
2843                 for (i = 0;i < IXGBE_VLVF_ENTRIES; i++) {
2844                         if (mirror_conf->vlan.vlan_mask & (1ULL << i)) {
2845                                 /* search vlan id related pool vlan filter index */
2846                                 reg_index = ixgbe_find_vlvf_slot(hw,
2847                                                 mirror_conf->vlan.vlan_id[i]);
2848                                 if(reg_index < 0)
2849                                         return (-EINVAL);
2850                                 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(reg_index));
2851                                 if ((vlvf & IXGBE_VLVF_VIEN) &&
2852                                         ((vlvf & IXGBE_VLVF_VLANID_MASK)
2853                                                 == mirror_conf->vlan.vlan_id[i]))
2854                                         vlan_mask |= (1ULL << reg_index);
2855                                 else
2856                                         return (-EINVAL);
2857                         }
2858                 }
2859
2860                 if (on) {
2861                         mv_lsb = vlan_mask & 0xFFFFFFFF;
2862                         mv_msb = vlan_mask >> vlan_mask_offset;
2863                         
2864                         mr_info->mr_conf[rule_id].vlan.vlan_mask =
2865                                                 mirror_conf->vlan.vlan_mask;
2866                         for(i = 0 ;i < ETH_VMDQ_MAX_VLAN_FILTERS; i++) {
2867                                 if(mirror_conf->vlan.vlan_mask & (1ULL << i))
2868                                         mr_info->mr_conf[rule_id].vlan.vlan_id[i] =
2869                                                 mirror_conf->vlan.vlan_id[i];
2870                         }
2871                 } else {
2872                         mv_lsb = 0;
2873                         mv_msb = 0;
2874                         mr_info->mr_conf[rule_id].vlan.vlan_mask = 0;
2875                         for(i = 0 ;i < ETH_VMDQ_MAX_VLAN_FILTERS; i++)
2876                                 mr_info->mr_conf[rule_id].vlan.vlan_id[i] = 0;
2877                 }
2878         }
2879
2880         /*
2881          * if enable pool mirror, write related pool mask register,if disable 
2882          * pool mirror, clear PFMRVM register
2883          */
2884         if (mirror_conf->rule_type_mask & ETH_VMDQ_POOL_MIRROR) {
2885                 if (on) { 
2886                         mp_lsb = mirror_conf->pool_mask & 0xFFFFFFFF;
2887                         mp_msb = mirror_conf->pool_mask >> pool_mask_offset;
2888                         mr_info->mr_conf[rule_id].pool_mask = 
2889                                         mirror_conf->pool_mask;
2890                         
2891                 } else {
2892                         mp_lsb = 0;
2893                         mp_msb = 0;
2894                         mr_info->mr_conf[rule_id].pool_mask = 0;
2895                 }
2896         }
2897         
2898         /* read  mirror control register and recalculate it */
2899         mr_ctl = IXGBE_READ_REG(hw,IXGBE_MRCTL(rule_id));
2900
2901         if (on) {
2902                 mr_ctl |= mirror_conf->rule_type_mask;
2903                 mr_ctl &= mirror_rule_mask;
2904                 mr_ctl |= mirror_conf->dst_pool << dst_pool_offset;
2905         } else
2906                 mr_ctl &= ~(mirror_conf->rule_type_mask & mirror_rule_mask);
2907
2908         mr_info->mr_conf[rule_id].rule_type_mask = (uint8_t)(mr_ctl & mirror_rule_mask);
2909         mr_info->mr_conf[rule_id].dst_pool = mirror_conf->dst_pool;
2910
2911         /* write mirrror control  register */
2912         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
2913         
2914         /* write pool mirrror control  register */
2915         if (mirror_conf->rule_type_mask & ETH_VMDQ_POOL_MIRROR) {
2916                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), mp_lsb);
2917                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset),
2918                                 mp_msb);
2919         }
2920         /* write VLAN mirrror control  register */
2921         if (mirror_conf->rule_type_mask & ETH_VMDQ_VLAN_MIRROR) {
2922                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), mv_lsb);
2923                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset),
2924                                 mv_msb);
2925         }
2926
2927         return 0;
2928 }
2929
2930 static int 
2931 ixgbe_mirror_rule_reset(struct rte_eth_dev *dev, uint8_t rule_id)
2932 {
2933         int mr_ctl = 0;
2934         uint32_t lsb_val = 0;
2935         uint32_t msb_val = 0;
2936         const uint8_t rule_mr_offset = 4;
2937         
2938         struct ixgbe_hw *hw =
2939                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2940         struct ixgbe_mirror_info *mr_info = 
2941                 (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
2942         
2943         if (ixgbe_vmdq_mode_check(hw) < 0)
2944                 return (-ENOTSUP);
2945
2946         memset(&mr_info->mr_conf[rule_id], 0,
2947                 sizeof(struct rte_eth_vmdq_mirror_conf));
2948
2949         /* clear PFVMCTL register */
2950         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
2951
2952         /* clear pool mask register */
2953         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), lsb_val);
2954         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset), msb_val);
2955
2956         /* clear vlan mask register */
2957         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), lsb_val);
2958         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset), msb_val);
2959
2960         return 0;
2961 }