xref: /f-stack/dpdk/drivers/net/e1000/igb_ethdev.c (revision a9643ea8)
1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
5  *   All rights reserved.
6  *
7  *   Redistribution and use in source and binary forms, with or without
8  *   modification, are permitted provided that the following conditions
9  *   are met:
10  *
11  *     * Redistributions of source code must retain the above copyright
12  *       notice, this list of conditions and the following disclaimer.
13  *     * Redistributions in binary form must reproduce the above copyright
14  *       notice, this list of conditions and the following disclaimer in
15  *       the documentation and/or other materials provided with the
16  *       distribution.
17  *     * Neither the name of Intel Corporation nor the names of its
18  *       contributors may be used to endorse or promote products derived
19  *       from this software without specific prior written permission.
20  *
21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32  */
33 
34 #include <sys/queue.h>
35 #include <stdio.h>
36 #include <errno.h>
37 #include <stdint.h>
38 #include <stdarg.h>
39 
40 #include <rte_common.h>
41 #include <rte_interrupts.h>
42 #include <rte_byteorder.h>
43 #include <rte_log.h>
44 #include <rte_debug.h>
45 #include <rte_pci.h>
46 #include <rte_ether.h>
47 #include <rte_ethdev.h>
48 #include <rte_memory.h>
49 #include <rte_memzone.h>
50 #include <rte_eal.h>
51 #include <rte_atomic.h>
52 #include <rte_malloc.h>
53 #include <rte_dev.h>
54 
55 #include "e1000_logs.h"
56 #include "base/e1000_api.h"
57 #include "e1000_ethdev.h"
58 #include "igb_regs.h"
59 
60 /*
61  * Default values for port configuration
62  */
63 #define IGB_DEFAULT_RX_FREE_THRESH  32
64 
65 #define IGB_DEFAULT_RX_PTHRESH      ((hw->mac.type == e1000_i354) ? 12 : 8)
66 #define IGB_DEFAULT_RX_HTHRESH      8
67 #define IGB_DEFAULT_RX_WTHRESH      ((hw->mac.type == e1000_82576) ? 1 : 4)
68 
69 #define IGB_DEFAULT_TX_PTHRESH      ((hw->mac.type == e1000_i354) ? 20 : 8)
70 #define IGB_DEFAULT_TX_HTHRESH      1
71 #define IGB_DEFAULT_TX_WTHRESH      ((hw->mac.type == e1000_82576) ? 1 : 16)
72 
73 #define IGB_HKEY_MAX_INDEX 10
74 
75 /* Bit shift and mask */
76 #define IGB_4_BIT_WIDTH  (CHAR_BIT / 2)
77 #define IGB_4_BIT_MASK   RTE_LEN2MASK(IGB_4_BIT_WIDTH, uint8_t)
78 #define IGB_8_BIT_WIDTH  CHAR_BIT
79 #define IGB_8_BIT_MASK   UINT8_MAX
80 
81 /* Additional timesync values. */
82 #define E1000_CYCLECOUNTER_MASK      0xffffffffffffffffULL
83 #define E1000_ETQF_FILTER_1588       3
84 #define IGB_82576_TSYNC_SHIFT        16
85 #define E1000_INCPERIOD_82576        (1 << E1000_TIMINCA_16NS_SHIFT)
86 #define E1000_INCVALUE_82576         (16 << IGB_82576_TSYNC_SHIFT)
87 #define E1000_TSAUXC_DISABLE_SYSTIME 0x80000000
88 
89 #define E1000_VTIVAR_MISC                0x01740
90 #define E1000_VTIVAR_MISC_MASK           0xFF
91 #define E1000_VTIVAR_VALID               0x80
92 #define E1000_VTIVAR_MISC_MAILBOX        0
93 #define E1000_VTIVAR_MISC_INTR_MASK      0x3
94 
95 /* External VLAN Enable bit mask */
96 #define E1000_CTRL_EXT_EXT_VLAN      (1 << 26)
97 
98 /* External VLAN Ether Type bit mask and shift */
99 #define E1000_VET_VET_EXT            0xFFFF0000
100 #define E1000_VET_VET_EXT_SHIFT      16
101 
102 static int  eth_igb_configure(struct rte_eth_dev *dev);
103 static int  eth_igb_start(struct rte_eth_dev *dev);
104 static void eth_igb_stop(struct rte_eth_dev *dev);
105 static int  eth_igb_dev_set_link_up(struct rte_eth_dev *dev);
106 static int  eth_igb_dev_set_link_down(struct rte_eth_dev *dev);
107 static void eth_igb_close(struct rte_eth_dev *dev);
108 static void eth_igb_promiscuous_enable(struct rte_eth_dev *dev);
109 static void eth_igb_promiscuous_disable(struct rte_eth_dev *dev);
110 static void eth_igb_allmulticast_enable(struct rte_eth_dev *dev);
111 static void eth_igb_allmulticast_disable(struct rte_eth_dev *dev);
112 static int  eth_igb_link_update(struct rte_eth_dev *dev,
113 				int wait_to_complete);
114 static void eth_igb_stats_get(struct rte_eth_dev *dev,
115 				struct rte_eth_stats *rte_stats);
116 static int eth_igb_xstats_get(struct rte_eth_dev *dev,
117 			      struct rte_eth_xstat *xstats, unsigned n);
118 static int eth_igb_xstats_get_names(struct rte_eth_dev *dev,
119 				    struct rte_eth_xstat_name *xstats_names,
120 				    unsigned limit);
121 static void eth_igb_stats_reset(struct rte_eth_dev *dev);
122 static void eth_igb_xstats_reset(struct rte_eth_dev *dev);
123 static void eth_igb_infos_get(struct rte_eth_dev *dev,
124 			      struct rte_eth_dev_info *dev_info);
125 static const uint32_t *eth_igb_supported_ptypes_get(struct rte_eth_dev *dev);
126 static void eth_igbvf_infos_get(struct rte_eth_dev *dev,
127 				struct rte_eth_dev_info *dev_info);
128 static int  eth_igb_flow_ctrl_get(struct rte_eth_dev *dev,
129 				struct rte_eth_fc_conf *fc_conf);
130 static int  eth_igb_flow_ctrl_set(struct rte_eth_dev *dev,
131 				struct rte_eth_fc_conf *fc_conf);
132 static int eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev);
133 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev);
134 static int eth_igb_interrupt_get_status(struct rte_eth_dev *dev);
135 static int eth_igb_interrupt_action(struct rte_eth_dev *dev);
136 static void eth_igb_interrupt_handler(struct rte_intr_handle *handle,
137 							void *param);
138 static int  igb_hardware_init(struct e1000_hw *hw);
139 static void igb_hw_control_acquire(struct e1000_hw *hw);
140 static void igb_hw_control_release(struct e1000_hw *hw);
141 static void igb_init_manageability(struct e1000_hw *hw);
142 static void igb_release_manageability(struct e1000_hw *hw);
143 
144 static int  eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
145 
146 static int eth_igb_vlan_filter_set(struct rte_eth_dev *dev,
147 		uint16_t vlan_id, int on);
148 static int eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
149 				 enum rte_vlan_type vlan_type,
150 				 uint16_t tpid_id);
151 static void eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask);
152 
153 static void igb_vlan_hw_filter_enable(struct rte_eth_dev *dev);
154 static void igb_vlan_hw_filter_disable(struct rte_eth_dev *dev);
155 static void igb_vlan_hw_strip_enable(struct rte_eth_dev *dev);
156 static void igb_vlan_hw_strip_disable(struct rte_eth_dev *dev);
157 static void igb_vlan_hw_extend_enable(struct rte_eth_dev *dev);
158 static void igb_vlan_hw_extend_disable(struct rte_eth_dev *dev);
159 
160 static int eth_igb_led_on(struct rte_eth_dev *dev);
161 static int eth_igb_led_off(struct rte_eth_dev *dev);
162 
163 static void igb_intr_disable(struct e1000_hw *hw);
164 static int  igb_get_rx_buffer_size(struct e1000_hw *hw);
165 static void eth_igb_rar_set(struct rte_eth_dev *dev,
166 		struct ether_addr *mac_addr,
167 		uint32_t index, uint32_t pool);
168 static void eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index);
169 static void eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
170 		struct ether_addr *addr);
171 
172 static void igbvf_intr_disable(struct e1000_hw *hw);
173 static int igbvf_dev_configure(struct rte_eth_dev *dev);
174 static int igbvf_dev_start(struct rte_eth_dev *dev);
175 static void igbvf_dev_stop(struct rte_eth_dev *dev);
176 static void igbvf_dev_close(struct rte_eth_dev *dev);
177 static void igbvf_promiscuous_enable(struct rte_eth_dev *dev);
178 static void igbvf_promiscuous_disable(struct rte_eth_dev *dev);
179 static void igbvf_allmulticast_enable(struct rte_eth_dev *dev);
180 static void igbvf_allmulticast_disable(struct rte_eth_dev *dev);
181 static int eth_igbvf_link_update(struct e1000_hw *hw);
182 static void eth_igbvf_stats_get(struct rte_eth_dev *dev,
183 				struct rte_eth_stats *rte_stats);
184 static int eth_igbvf_xstats_get(struct rte_eth_dev *dev,
185 				struct rte_eth_xstat *xstats, unsigned n);
186 static int eth_igbvf_xstats_get_names(struct rte_eth_dev *dev,
187 				      struct rte_eth_xstat_name *xstats_names,
188 				      unsigned limit);
189 static void eth_igbvf_stats_reset(struct rte_eth_dev *dev);
190 static int igbvf_vlan_filter_set(struct rte_eth_dev *dev,
191 		uint16_t vlan_id, int on);
192 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on);
193 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on);
194 static void igbvf_default_mac_addr_set(struct rte_eth_dev *dev,
195 		struct ether_addr *addr);
196 static int igbvf_get_reg_length(struct rte_eth_dev *dev);
197 static int igbvf_get_regs(struct rte_eth_dev *dev,
198 		struct rte_dev_reg_info *regs);
199 
200 static int eth_igb_rss_reta_update(struct rte_eth_dev *dev,
201 				   struct rte_eth_rss_reta_entry64 *reta_conf,
202 				   uint16_t reta_size);
203 static int eth_igb_rss_reta_query(struct rte_eth_dev *dev,
204 				  struct rte_eth_rss_reta_entry64 *reta_conf,
205 				  uint16_t reta_size);
206 
207 static int eth_igb_syn_filter_set(struct rte_eth_dev *dev,
208 			struct rte_eth_syn_filter *filter,
209 			bool add);
210 static int eth_igb_syn_filter_get(struct rte_eth_dev *dev,
211 			struct rte_eth_syn_filter *filter);
212 static int eth_igb_syn_filter_handle(struct rte_eth_dev *dev,
213 			enum rte_filter_op filter_op,
214 			void *arg);
215 static int igb_add_2tuple_filter(struct rte_eth_dev *dev,
216 			struct rte_eth_ntuple_filter *ntuple_filter);
217 static int igb_remove_2tuple_filter(struct rte_eth_dev *dev,
218 			struct rte_eth_ntuple_filter *ntuple_filter);
219 static int eth_igb_add_del_flex_filter(struct rte_eth_dev *dev,
220 			struct rte_eth_flex_filter *filter,
221 			bool add);
222 static int eth_igb_get_flex_filter(struct rte_eth_dev *dev,
223 			struct rte_eth_flex_filter *filter);
224 static int eth_igb_flex_filter_handle(struct rte_eth_dev *dev,
225 			enum rte_filter_op filter_op,
226 			void *arg);
227 static int igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
228 			struct rte_eth_ntuple_filter *ntuple_filter);
229 static int igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
230 			struct rte_eth_ntuple_filter *ntuple_filter);
231 static int igb_add_del_ntuple_filter(struct rte_eth_dev *dev,
232 			struct rte_eth_ntuple_filter *filter,
233 			bool add);
234 static int igb_get_ntuple_filter(struct rte_eth_dev *dev,
235 			struct rte_eth_ntuple_filter *filter);
236 static int igb_ntuple_filter_handle(struct rte_eth_dev *dev,
237 				enum rte_filter_op filter_op,
238 				void *arg);
239 static int igb_add_del_ethertype_filter(struct rte_eth_dev *dev,
240 			struct rte_eth_ethertype_filter *filter,
241 			bool add);
242 static int igb_ethertype_filter_handle(struct rte_eth_dev *dev,
243 				enum rte_filter_op filter_op,
244 				void *arg);
245 static int igb_get_ethertype_filter(struct rte_eth_dev *dev,
246 			struct rte_eth_ethertype_filter *filter);
247 static int eth_igb_filter_ctrl(struct rte_eth_dev *dev,
248 		     enum rte_filter_type filter_type,
249 		     enum rte_filter_op filter_op,
250 		     void *arg);
251 static int eth_igb_get_reg_length(struct rte_eth_dev *dev);
252 static int eth_igb_get_regs(struct rte_eth_dev *dev,
253 		struct rte_dev_reg_info *regs);
254 static int eth_igb_get_eeprom_length(struct rte_eth_dev *dev);
255 static int eth_igb_get_eeprom(struct rte_eth_dev *dev,
256 		struct rte_dev_eeprom_info *eeprom);
257 static int eth_igb_set_eeprom(struct rte_eth_dev *dev,
258 		struct rte_dev_eeprom_info *eeprom);
259 static int eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
260 				    struct ether_addr *mc_addr_set,
261 				    uint32_t nb_mc_addr);
262 static int igb_timesync_enable(struct rte_eth_dev *dev);
263 static int igb_timesync_disable(struct rte_eth_dev *dev);
264 static int igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
265 					  struct timespec *timestamp,
266 					  uint32_t flags);
267 static int igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
268 					  struct timespec *timestamp);
269 static int igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
270 static int igb_timesync_read_time(struct rte_eth_dev *dev,
271 				  struct timespec *timestamp);
272 static int igb_timesync_write_time(struct rte_eth_dev *dev,
273 				   const struct timespec *timestamp);
274 static int eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev,
275 					uint16_t queue_id);
276 static int eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev,
277 					 uint16_t queue_id);
278 static void eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
279 				       uint8_t queue, uint8_t msix_vector);
280 static void eth_igb_write_ivar(struct e1000_hw *hw, uint8_t msix_vector,
281 			       uint8_t index, uint8_t offset);
282 static void eth_igb_configure_msix_intr(struct rte_eth_dev *dev);
283 static void eth_igbvf_interrupt_handler(struct rte_intr_handle *handle,
284 					void *param);
285 static void igbvf_mbx_process(struct rte_eth_dev *dev);
286 
287 /*
288  * Define VF Stats MACRO for Non "cleared on read" register
289  */
290 #define UPDATE_VF_STAT(reg, last, cur)            \
291 {                                                 \
292 	u32 latest = E1000_READ_REG(hw, reg);     \
293 	cur += (latest - last) & UINT_MAX;        \
294 	last = latest;                            \
295 }
296 
297 #define IGB_FC_PAUSE_TIME 0x0680
298 #define IGB_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
299 #define IGB_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
300 
301 #define IGBVF_PMD_NAME "rte_igbvf_pmd"     /* PMD name */
302 
303 static enum e1000_fc_mode igb_fc_setting = e1000_fc_full;
304 
305 /*
306  * The set of PCI devices this driver supports
307  */
308 static const struct rte_pci_id pci_id_igb_map[] = {
309 
310 #define RTE_PCI_DEV_ID_DECL_IGB(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
311 #include "rte_pci_dev_ids.h"
312 
313 {0},
314 };
315 
316 /*
317  * The set of PCI devices this driver supports (for 82576&I350 VF)
318  */
319 static const struct rte_pci_id pci_id_igbvf_map[] = {
320 
321 #define RTE_PCI_DEV_ID_DECL_IGBVF(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
322 #include "rte_pci_dev_ids.h"
323 
324 {0},
325 };
326 
327 static const struct rte_eth_desc_lim rx_desc_lim = {
328 	.nb_max = E1000_MAX_RING_DESC,
329 	.nb_min = E1000_MIN_RING_DESC,
330 	.nb_align = IGB_RXD_ALIGN,
331 };
332 
333 static const struct rte_eth_desc_lim tx_desc_lim = {
334 	.nb_max = E1000_MAX_RING_DESC,
335 	.nb_min = E1000_MIN_RING_DESC,
336 	.nb_align = IGB_RXD_ALIGN,
337 };
338 
339 static const struct eth_dev_ops eth_igb_ops = {
340 	.dev_configure        = eth_igb_configure,
341 	.dev_start            = eth_igb_start,
342 	.dev_stop             = eth_igb_stop,
343 	.dev_set_link_up      = eth_igb_dev_set_link_up,
344 	.dev_set_link_down    = eth_igb_dev_set_link_down,
345 	.dev_close            = eth_igb_close,
346 	.promiscuous_enable   = eth_igb_promiscuous_enable,
347 	.promiscuous_disable  = eth_igb_promiscuous_disable,
348 	.allmulticast_enable  = eth_igb_allmulticast_enable,
349 	.allmulticast_disable = eth_igb_allmulticast_disable,
350 	.link_update          = eth_igb_link_update,
351 	.stats_get            = eth_igb_stats_get,
352 	.xstats_get           = eth_igb_xstats_get,
353 	.xstats_get_names     = eth_igb_xstats_get_names,
354 	.stats_reset          = eth_igb_stats_reset,
355 	.xstats_reset         = eth_igb_xstats_reset,
356 	.dev_infos_get        = eth_igb_infos_get,
357 	.dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
358 	.mtu_set              = eth_igb_mtu_set,
359 	.vlan_filter_set      = eth_igb_vlan_filter_set,
360 	.vlan_tpid_set        = eth_igb_vlan_tpid_set,
361 	.vlan_offload_set     = eth_igb_vlan_offload_set,
362 	.rx_queue_setup       = eth_igb_rx_queue_setup,
363 	.rx_queue_intr_enable = eth_igb_rx_queue_intr_enable,
364 	.rx_queue_intr_disable = eth_igb_rx_queue_intr_disable,
365 	.rx_queue_release     = eth_igb_rx_queue_release,
366 	.rx_queue_count       = eth_igb_rx_queue_count,
367 	.rx_descriptor_done   = eth_igb_rx_descriptor_done,
368 	.tx_queue_setup       = eth_igb_tx_queue_setup,
369 	.tx_queue_release     = eth_igb_tx_queue_release,
370 	.dev_led_on           = eth_igb_led_on,
371 	.dev_led_off          = eth_igb_led_off,
372 	.flow_ctrl_get        = eth_igb_flow_ctrl_get,
373 	.flow_ctrl_set        = eth_igb_flow_ctrl_set,
374 	.mac_addr_add         = eth_igb_rar_set,
375 	.mac_addr_remove      = eth_igb_rar_clear,
376 	.mac_addr_set         = eth_igb_default_mac_addr_set,
377 	.reta_update          = eth_igb_rss_reta_update,
378 	.reta_query           = eth_igb_rss_reta_query,
379 	.rss_hash_update      = eth_igb_rss_hash_update,
380 	.rss_hash_conf_get    = eth_igb_rss_hash_conf_get,
381 	.filter_ctrl          = eth_igb_filter_ctrl,
382 	.set_mc_addr_list     = eth_igb_set_mc_addr_list,
383 	.rxq_info_get         = igb_rxq_info_get,
384 	.txq_info_get         = igb_txq_info_get,
385 	.timesync_enable      = igb_timesync_enable,
386 	.timesync_disable     = igb_timesync_disable,
387 	.timesync_read_rx_timestamp = igb_timesync_read_rx_timestamp,
388 	.timesync_read_tx_timestamp = igb_timesync_read_tx_timestamp,
389 	.get_reg              = eth_igb_get_regs,
390 	.get_eeprom_length    = eth_igb_get_eeprom_length,
391 	.get_eeprom           = eth_igb_get_eeprom,
392 	.set_eeprom           = eth_igb_set_eeprom,
393 	.timesync_adjust_time = igb_timesync_adjust_time,
394 	.timesync_read_time   = igb_timesync_read_time,
395 	.timesync_write_time  = igb_timesync_write_time,
396 };
397 
398 /*
399  * dev_ops for virtual function, bare necessities for basic vf
400  * operation have been implemented
401  */
402 static const struct eth_dev_ops igbvf_eth_dev_ops = {
403 	.dev_configure        = igbvf_dev_configure,
404 	.dev_start            = igbvf_dev_start,
405 	.dev_stop             = igbvf_dev_stop,
406 	.dev_close            = igbvf_dev_close,
407 	.promiscuous_enable   = igbvf_promiscuous_enable,
408 	.promiscuous_disable  = igbvf_promiscuous_disable,
409 	.allmulticast_enable  = igbvf_allmulticast_enable,
410 	.allmulticast_disable = igbvf_allmulticast_disable,
411 	.link_update          = eth_igb_link_update,
412 	.stats_get            = eth_igbvf_stats_get,
413 	.xstats_get           = eth_igbvf_xstats_get,
414 	.xstats_get_names     = eth_igbvf_xstats_get_names,
415 	.stats_reset          = eth_igbvf_stats_reset,
416 	.xstats_reset         = eth_igbvf_stats_reset,
417 	.vlan_filter_set      = igbvf_vlan_filter_set,
418 	.dev_infos_get        = eth_igbvf_infos_get,
419 	.dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
420 	.rx_queue_setup       = eth_igb_rx_queue_setup,
421 	.rx_queue_release     = eth_igb_rx_queue_release,
422 	.tx_queue_setup       = eth_igb_tx_queue_setup,
423 	.tx_queue_release     = eth_igb_tx_queue_release,
424 	.set_mc_addr_list     = eth_igb_set_mc_addr_list,
425 	.rxq_info_get         = igb_rxq_info_get,
426 	.txq_info_get         = igb_txq_info_get,
427 	.mac_addr_set         = igbvf_default_mac_addr_set,
428 	.get_reg              = igbvf_get_regs,
429 };
430 
431 /* store statistics names and its offset in stats structure */
432 struct rte_igb_xstats_name_off {
433 	char name[RTE_ETH_XSTATS_NAME_SIZE];
434 	unsigned offset;
435 };
436 
437 static const struct rte_igb_xstats_name_off rte_igb_stats_strings[] = {
438 	{"rx_crc_errors", offsetof(struct e1000_hw_stats, crcerrs)},
439 	{"rx_align_errors", offsetof(struct e1000_hw_stats, algnerrc)},
440 	{"rx_symbol_errors", offsetof(struct e1000_hw_stats, symerrs)},
441 	{"rx_missed_packets", offsetof(struct e1000_hw_stats, mpc)},
442 	{"tx_single_collision_packets", offsetof(struct e1000_hw_stats, scc)},
443 	{"tx_multiple_collision_packets", offsetof(struct e1000_hw_stats, mcc)},
444 	{"tx_excessive_collision_packets", offsetof(struct e1000_hw_stats,
445 		ecol)},
446 	{"tx_late_collisions", offsetof(struct e1000_hw_stats, latecol)},
447 	{"tx_total_collisions", offsetof(struct e1000_hw_stats, colc)},
448 	{"tx_deferred_packets", offsetof(struct e1000_hw_stats, dc)},
449 	{"tx_no_carrier_sense_packets", offsetof(struct e1000_hw_stats, tncrs)},
450 	{"rx_carrier_ext_errors", offsetof(struct e1000_hw_stats, cexterr)},
451 	{"rx_length_errors", offsetof(struct e1000_hw_stats, rlec)},
452 	{"rx_xon_packets", offsetof(struct e1000_hw_stats, xonrxc)},
453 	{"tx_xon_packets", offsetof(struct e1000_hw_stats, xontxc)},
454 	{"rx_xoff_packets", offsetof(struct e1000_hw_stats, xoffrxc)},
455 	{"tx_xoff_packets", offsetof(struct e1000_hw_stats, xofftxc)},
456 	{"rx_flow_control_unsupported_packets", offsetof(struct e1000_hw_stats,
457 		fcruc)},
458 	{"rx_size_64_packets", offsetof(struct e1000_hw_stats, prc64)},
459 	{"rx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, prc127)},
460 	{"rx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, prc255)},
461 	{"rx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, prc511)},
462 	{"rx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
463 		prc1023)},
464 	{"rx_size_1024_to_max_packets", offsetof(struct e1000_hw_stats,
465 		prc1522)},
466 	{"rx_broadcast_packets", offsetof(struct e1000_hw_stats, bprc)},
467 	{"rx_multicast_packets", offsetof(struct e1000_hw_stats, mprc)},
468 	{"rx_undersize_errors", offsetof(struct e1000_hw_stats, ruc)},
469 	{"rx_fragment_errors", offsetof(struct e1000_hw_stats, rfc)},
470 	{"rx_oversize_errors", offsetof(struct e1000_hw_stats, roc)},
471 	{"rx_jabber_errors", offsetof(struct e1000_hw_stats, rjc)},
472 	{"rx_management_packets", offsetof(struct e1000_hw_stats, mgprc)},
473 	{"rx_management_dropped", offsetof(struct e1000_hw_stats, mgpdc)},
474 	{"tx_management_packets", offsetof(struct e1000_hw_stats, mgptc)},
475 	{"rx_total_packets", offsetof(struct e1000_hw_stats, tpr)},
476 	{"tx_total_packets", offsetof(struct e1000_hw_stats, tpt)},
477 	{"rx_total_bytes", offsetof(struct e1000_hw_stats, tor)},
478 	{"tx_total_bytes", offsetof(struct e1000_hw_stats, tot)},
479 	{"tx_size_64_packets", offsetof(struct e1000_hw_stats, ptc64)},
480 	{"tx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, ptc127)},
481 	{"tx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, ptc255)},
482 	{"tx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, ptc511)},
483 	{"tx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
484 		ptc1023)},
485 	{"tx_size_1023_to_max_packets", offsetof(struct e1000_hw_stats,
486 		ptc1522)},
487 	{"tx_multicast_packets", offsetof(struct e1000_hw_stats, mptc)},
488 	{"tx_broadcast_packets", offsetof(struct e1000_hw_stats, bptc)},
489 	{"tx_tso_packets", offsetof(struct e1000_hw_stats, tsctc)},
490 	{"tx_tso_errors", offsetof(struct e1000_hw_stats, tsctfc)},
491 	{"rx_sent_to_host_packets", offsetof(struct e1000_hw_stats, rpthc)},
492 	{"tx_sent_by_host_packets", offsetof(struct e1000_hw_stats, hgptc)},
493 	{"rx_code_violation_packets", offsetof(struct e1000_hw_stats, scvpc)},
494 
495 	{"interrupt_assert_count", offsetof(struct e1000_hw_stats, iac)},
496 };
497 
498 #define IGB_NB_XSTATS (sizeof(rte_igb_stats_strings) / \
499 		sizeof(rte_igb_stats_strings[0]))
500 
501 static const struct rte_igb_xstats_name_off rte_igbvf_stats_strings[] = {
502 	{"rx_multicast_packets", offsetof(struct e1000_vf_stats, mprc)},
503 	{"rx_good_loopback_packets", offsetof(struct e1000_vf_stats, gprlbc)},
504 	{"tx_good_loopback_packets", offsetof(struct e1000_vf_stats, gptlbc)},
505 	{"rx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gorlbc)},
506 	{"tx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gotlbc)},
507 };
508 
509 #define IGBVF_NB_XSTATS (sizeof(rte_igbvf_stats_strings) / \
510 		sizeof(rte_igbvf_stats_strings[0]))
511 
512 /**
513  * Atomically reads the link status information from global
514  * structure rte_eth_dev.
515  *
516  * @param dev
517  *   - Pointer to the structure rte_eth_dev to read from.
518  *   - Pointer to the buffer to be saved with the link status.
519  *
520  * @return
521  *   - On success, zero.
522  *   - On failure, negative value.
523  */
524 static inline int
525 rte_igb_dev_atomic_read_link_status(struct rte_eth_dev *dev,
526 				struct rte_eth_link *link)
527 {
528 	struct rte_eth_link *dst = link;
529 	struct rte_eth_link *src = &(dev->data->dev_link);
530 
531 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
532 					*(uint64_t *)src) == 0)
533 		return -1;
534 
535 	return 0;
536 }
537 
538 /**
539  * Atomically writes the link status information into global
540  * structure rte_eth_dev.
541  *
542  * @param dev
543  *   - Pointer to the structure rte_eth_dev to read from.
544  *   - Pointer to the buffer to be saved with the link status.
545  *
546  * @return
547  *   - On success, zero.
548  *   - On failure, negative value.
549  */
550 static inline int
551 rte_igb_dev_atomic_write_link_status(struct rte_eth_dev *dev,
552 				struct rte_eth_link *link)
553 {
554 	struct rte_eth_link *dst = &(dev->data->dev_link);
555 	struct rte_eth_link *src = link;
556 
557 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
558 					*(uint64_t *)src) == 0)
559 		return -1;
560 
561 	return 0;
562 }
563 
564 static inline void
565 igb_intr_enable(struct rte_eth_dev *dev)
566 {
567 	struct e1000_interrupt *intr =
568 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
569 	struct e1000_hw *hw =
570 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
571 
572 	E1000_WRITE_REG(hw, E1000_IMS, intr->mask);
573 	E1000_WRITE_FLUSH(hw);
574 }
575 
576 static void
577 igb_intr_disable(struct e1000_hw *hw)
578 {
579 	E1000_WRITE_REG(hw, E1000_IMC, ~0);
580 	E1000_WRITE_FLUSH(hw);
581 }
582 
583 static inline void
584 igbvf_intr_enable(struct rte_eth_dev *dev)
585 {
586 	struct e1000_hw *hw =
587 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
588 
589 	/* only for mailbox */
590 	E1000_WRITE_REG(hw, E1000_EIAM, 1 << E1000_VTIVAR_MISC_MAILBOX);
591 	E1000_WRITE_REG(hw, E1000_EIAC, 1 << E1000_VTIVAR_MISC_MAILBOX);
592 	E1000_WRITE_REG(hw, E1000_EIMS, 1 << E1000_VTIVAR_MISC_MAILBOX);
593 	E1000_WRITE_FLUSH(hw);
594 }
595 
596 /* only for mailbox now. If RX/TX needed, should extend this function.  */
597 static void
598 igbvf_set_ivar_map(struct e1000_hw *hw, uint8_t msix_vector)
599 {
600 	uint32_t tmp = 0;
601 
602 	/* mailbox */
603 	tmp |= (msix_vector & E1000_VTIVAR_MISC_INTR_MASK);
604 	tmp |= E1000_VTIVAR_VALID;
605 	E1000_WRITE_REG(hw, E1000_VTIVAR_MISC, tmp);
606 }
607 
608 static void
609 eth_igbvf_configure_msix_intr(struct rte_eth_dev *dev)
610 {
611 	struct e1000_hw *hw =
612 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
613 
614 	/* Configure VF other cause ivar */
615 	igbvf_set_ivar_map(hw, E1000_VTIVAR_MISC_MAILBOX);
616 }
617 
618 static inline int32_t
619 igb_pf_reset_hw(struct e1000_hw *hw)
620 {
621 	uint32_t ctrl_ext;
622 	int32_t status;
623 
624 	status = e1000_reset_hw(hw);
625 
626 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
627 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
628 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
629 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
630 	E1000_WRITE_FLUSH(hw);
631 
632 	return status;
633 }
634 
635 static void
636 igb_identify_hardware(struct rte_eth_dev *dev)
637 {
638 	struct e1000_hw *hw =
639 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
640 
641 	hw->vendor_id = dev->pci_dev->id.vendor_id;
642 	hw->device_id = dev->pci_dev->id.device_id;
643 	hw->subsystem_vendor_id = dev->pci_dev->id.subsystem_vendor_id;
644 	hw->subsystem_device_id = dev->pci_dev->id.subsystem_device_id;
645 
646 	e1000_set_mac_type(hw);
647 
648 	/* need to check if it is a vf device below */
649 }
650 
651 static int
652 igb_reset_swfw_lock(struct e1000_hw *hw)
653 {
654 	int ret_val;
655 
656 	/*
657 	 * Do mac ops initialization manually here, since we will need
658 	 * some function pointers set by this call.
659 	 */
660 	ret_val = e1000_init_mac_params(hw);
661 	if (ret_val)
662 		return ret_val;
663 
664 	/*
665 	 * SMBI lock should not fail in this early stage. If this is the case,
666 	 * it is due to an improper exit of the application.
667 	 * So force the release of the faulty lock.
668 	 */
669 	if (e1000_get_hw_semaphore_generic(hw) < 0) {
670 		PMD_DRV_LOG(DEBUG, "SMBI lock released");
671 	}
672 	e1000_put_hw_semaphore_generic(hw);
673 
674 	if (hw->mac.ops.acquire_swfw_sync != NULL) {
675 		uint16_t mask;
676 
677 		/*
678 		 * Phy lock should not fail in this early stage. If this is the case,
679 		 * it is due to an improper exit of the application.
680 		 * So force the release of the faulty lock.
681 		 */
682 		mask = E1000_SWFW_PHY0_SM << hw->bus.func;
683 		if (hw->bus.func > E1000_FUNC_1)
684 			mask <<= 2;
685 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
686 			PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released",
687 				    hw->bus.func);
688 		}
689 		hw->mac.ops.release_swfw_sync(hw, mask);
690 
691 		/*
692 		 * This one is more tricky since it is common to all ports; but
693 		 * swfw_sync retries last long enough (1s) to be almost sure that if
694 		 * lock can not be taken it is due to an improper lock of the
695 		 * semaphore.
696 		 */
697 		mask = E1000_SWFW_EEP_SM;
698 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
699 			PMD_DRV_LOG(DEBUG, "SWFW common locks released");
700 		}
701 		hw->mac.ops.release_swfw_sync(hw, mask);
702 	}
703 
704 	return E1000_SUCCESS;
705 }
706 
707 static int
708 eth_igb_dev_init(struct rte_eth_dev *eth_dev)
709 {
710 	int error = 0;
711 	struct rte_pci_device *pci_dev;
712 	struct e1000_hw *hw =
713 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
714 	struct e1000_vfta * shadow_vfta =
715 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
716 	struct e1000_filter_info *filter_info =
717 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
718 	struct e1000_adapter *adapter =
719 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
720 
721 	uint32_t ctrl_ext;
722 
723 	pci_dev = eth_dev->pci_dev;
724 
725 	eth_dev->dev_ops = &eth_igb_ops;
726 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
727 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
728 
729 	/* for secondary processes, we don't initialise any further as primary
730 	 * has already done this work. Only check we don't need a different
731 	 * RX function */
732 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
733 		if (eth_dev->data->scattered_rx)
734 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
735 		return 0;
736 	}
737 
738 	rte_eth_copy_pci_info(eth_dev, pci_dev);
739 
740 	hw->hw_addr= (void *)pci_dev->mem_resource[0].addr;
741 
742 	igb_identify_hardware(eth_dev);
743 	if (e1000_setup_init_funcs(hw, FALSE) != E1000_SUCCESS) {
744 		error = -EIO;
745 		goto err_late;
746 	}
747 
748 	e1000_get_bus_info(hw);
749 
750 	/* Reset any pending lock */
751 	if (igb_reset_swfw_lock(hw) != E1000_SUCCESS) {
752 		error = -EIO;
753 		goto err_late;
754 	}
755 
756 	/* Finish initialization */
757 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS) {
758 		error = -EIO;
759 		goto err_late;
760 	}
761 
762 	hw->mac.autoneg = 1;
763 	hw->phy.autoneg_wait_to_complete = 0;
764 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
765 
766 	/* Copper options */
767 	if (hw->phy.media_type == e1000_media_type_copper) {
768 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
769 		hw->phy.disable_polarity_correction = 0;
770 		hw->phy.ms_type = e1000_ms_hw_default;
771 	}
772 
773 	/*
774 	 * Start from a known state, this is important in reading the nvm
775 	 * and mac from that.
776 	 */
777 	igb_pf_reset_hw(hw);
778 
779 	/* Make sure we have a good EEPROM before we read from it */
780 	if (e1000_validate_nvm_checksum(hw) < 0) {
781 		/*
782 		 * Some PCI-E parts fail the first check due to
783 		 * the link being in sleep state, call it again,
784 		 * if it fails a second time its a real issue.
785 		 */
786 		if (e1000_validate_nvm_checksum(hw) < 0) {
787 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
788 			error = -EIO;
789 			goto err_late;
790 		}
791 	}
792 
793 	/* Read the permanent MAC address out of the EEPROM */
794 	if (e1000_read_mac_addr(hw) != 0) {
795 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
796 		error = -EIO;
797 		goto err_late;
798 	}
799 
800 	/* Allocate memory for storing MAC addresses */
801 	eth_dev->data->mac_addrs = rte_zmalloc("e1000",
802 		ETHER_ADDR_LEN * hw->mac.rar_entry_count, 0);
803 	if (eth_dev->data->mac_addrs == NULL) {
804 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
805 						"store MAC addresses",
806 				ETHER_ADDR_LEN * hw->mac.rar_entry_count);
807 		error = -ENOMEM;
808 		goto err_late;
809 	}
810 
811 	/* Copy the permanent MAC address */
812 	ether_addr_copy((struct ether_addr *)hw->mac.addr, &eth_dev->data->mac_addrs[0]);
813 
814 	/* initialize the vfta */
815 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
816 
817 	/* Now initialize the hardware */
818 	if (igb_hardware_init(hw) != 0) {
819 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
820 		rte_free(eth_dev->data->mac_addrs);
821 		eth_dev->data->mac_addrs = NULL;
822 		error = -ENODEV;
823 		goto err_late;
824 	}
825 	hw->mac.get_link_status = 1;
826 	adapter->stopped = 0;
827 
828 	/* Indicate SOL/IDER usage */
829 	if (e1000_check_reset_block(hw) < 0) {
830 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to"
831 					"SOL/IDER session");
832 	}
833 
834 	/* initialize PF if max_vfs not zero */
835 	igb_pf_host_init(eth_dev);
836 
837 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
838 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
839 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
840 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
841 	E1000_WRITE_FLUSH(hw);
842 
843 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
844 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
845 		     pci_dev->id.device_id);
846 
847 	rte_intr_callback_register(&pci_dev->intr_handle,
848 				   eth_igb_interrupt_handler,
849 				   (void *)eth_dev);
850 
851 	/* enable uio/vfio intr/eventfd mapping */
852 	rte_intr_enable(&pci_dev->intr_handle);
853 
854 	/* enable support intr */
855 	igb_intr_enable(eth_dev);
856 
857 	TAILQ_INIT(&filter_info->flex_list);
858 	filter_info->flex_mask = 0;
859 	TAILQ_INIT(&filter_info->twotuple_list);
860 	filter_info->twotuple_mask = 0;
861 	TAILQ_INIT(&filter_info->fivetuple_list);
862 	filter_info->fivetuple_mask = 0;
863 
864 	return 0;
865 
866 err_late:
867 	igb_hw_control_release(hw);
868 
869 	return error;
870 }
871 
872 static int
873 eth_igb_dev_uninit(struct rte_eth_dev *eth_dev)
874 {
875 	struct rte_pci_device *pci_dev;
876 	struct e1000_hw *hw;
877 	struct e1000_adapter *adapter =
878 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
879 
880 	PMD_INIT_FUNC_TRACE();
881 
882 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
883 		return -EPERM;
884 
885 	hw = E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
886 	pci_dev = eth_dev->pci_dev;
887 
888 	if (adapter->stopped == 0)
889 		eth_igb_close(eth_dev);
890 
891 	eth_dev->dev_ops = NULL;
892 	eth_dev->rx_pkt_burst = NULL;
893 	eth_dev->tx_pkt_burst = NULL;
894 
895 	/* Reset any pending lock */
896 	igb_reset_swfw_lock(hw);
897 
898 	rte_free(eth_dev->data->mac_addrs);
899 	eth_dev->data->mac_addrs = NULL;
900 
901 	/* uninitialize PF if max_vfs not zero */
902 	igb_pf_host_uninit(eth_dev);
903 
904 	/* disable uio intr before callback unregister */
905 	rte_intr_disable(&(pci_dev->intr_handle));
906 	rte_intr_callback_unregister(&(pci_dev->intr_handle),
907 		eth_igb_interrupt_handler, (void *)eth_dev);
908 
909 	return 0;
910 }
911 
912 /*
913  * Virtual Function device init
914  */
915 static int
916 eth_igbvf_dev_init(struct rte_eth_dev *eth_dev)
917 {
918 	struct rte_pci_device *pci_dev;
919 	struct e1000_adapter *adapter =
920 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
921 	struct e1000_hw *hw =
922 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
923 	int diag;
924 	struct ether_addr *perm_addr = (struct ether_addr *)hw->mac.perm_addr;
925 
926 	PMD_INIT_FUNC_TRACE();
927 
928 	eth_dev->dev_ops = &igbvf_eth_dev_ops;
929 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
930 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
931 
932 	/* for secondary processes, we don't initialise any further as primary
933 	 * has already done this work. Only check we don't need a different
934 	 * RX function */
935 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
936 		if (eth_dev->data->scattered_rx)
937 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
938 		return 0;
939 	}
940 
941 	pci_dev = eth_dev->pci_dev;
942 
943 	rte_eth_copy_pci_info(eth_dev, pci_dev);
944 
945 	hw->device_id = pci_dev->id.device_id;
946 	hw->vendor_id = pci_dev->id.vendor_id;
947 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
948 	adapter->stopped = 0;
949 
950 	/* Initialize the shared code (base driver) */
951 	diag = e1000_setup_init_funcs(hw, TRUE);
952 	if (diag != 0) {
953 		PMD_INIT_LOG(ERR, "Shared code init failed for igbvf: %d",
954 			diag);
955 		return -EIO;
956 	}
957 
958 	/* init_mailbox_params */
959 	hw->mbx.ops.init_params(hw);
960 
961 	/* Disable the interrupts for VF */
962 	igbvf_intr_disable(hw);
963 
964 	diag = hw->mac.ops.reset_hw(hw);
965 
966 	/* Allocate memory for storing MAC addresses */
967 	eth_dev->data->mac_addrs = rte_zmalloc("igbvf", ETHER_ADDR_LEN *
968 		hw->mac.rar_entry_count, 0);
969 	if (eth_dev->data->mac_addrs == NULL) {
970 		PMD_INIT_LOG(ERR,
971 			"Failed to allocate %d bytes needed to store MAC "
972 			"addresses",
973 			ETHER_ADDR_LEN * hw->mac.rar_entry_count);
974 		return -ENOMEM;
975 	}
976 
977 	/* Generate a random MAC address, if none was assigned by PF. */
978 	if (is_zero_ether_addr(perm_addr)) {
979 		eth_random_addr(perm_addr->addr_bytes);
980 		diag = e1000_rar_set(hw, perm_addr->addr_bytes, 0);
981 		if (diag) {
982 			rte_free(eth_dev->data->mac_addrs);
983 			eth_dev->data->mac_addrs = NULL;
984 			return diag;
985 		}
986 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
987 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
988 			     "%02x:%02x:%02x:%02x:%02x:%02x",
989 			     perm_addr->addr_bytes[0],
990 			     perm_addr->addr_bytes[1],
991 			     perm_addr->addr_bytes[2],
992 			     perm_addr->addr_bytes[3],
993 			     perm_addr->addr_bytes[4],
994 			     perm_addr->addr_bytes[5]);
995 	}
996 
997 	/* Copy the permanent MAC address */
998 	ether_addr_copy((struct ether_addr *) hw->mac.perm_addr,
999 			&eth_dev->data->mac_addrs[0]);
1000 
1001 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x "
1002 		     "mac.type=%s",
1003 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1004 		     pci_dev->id.device_id, "igb_mac_82576_vf");
1005 
1006 	rte_intr_callback_register(&pci_dev->intr_handle,
1007 				   eth_igbvf_interrupt_handler,
1008 				   (void *)eth_dev);
1009 
1010 	return 0;
1011 }
1012 
1013 static int
1014 eth_igbvf_dev_uninit(struct rte_eth_dev *eth_dev)
1015 {
1016 	struct e1000_adapter *adapter =
1017 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
1018 	struct rte_pci_device *pci_dev = eth_dev->pci_dev;
1019 
1020 	PMD_INIT_FUNC_TRACE();
1021 
1022 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1023 		return -EPERM;
1024 
1025 	if (adapter->stopped == 0)
1026 		igbvf_dev_close(eth_dev);
1027 
1028 	eth_dev->dev_ops = NULL;
1029 	eth_dev->rx_pkt_burst = NULL;
1030 	eth_dev->tx_pkt_burst = NULL;
1031 
1032 	rte_free(eth_dev->data->mac_addrs);
1033 	eth_dev->data->mac_addrs = NULL;
1034 
1035 	/* disable uio intr before callback unregister */
1036 	rte_intr_disable(&pci_dev->intr_handle);
1037 	rte_intr_callback_unregister(&pci_dev->intr_handle,
1038 				     eth_igbvf_interrupt_handler,
1039 				     (void *)eth_dev);
1040 
1041 	return 0;
1042 }
1043 
1044 static struct eth_driver rte_igb_pmd = {
1045 	.pci_drv = {
1046 		.name = "rte_igb_pmd",
1047 		.id_table = pci_id_igb_map,
1048 		.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
1049 			RTE_PCI_DRV_DETACHABLE,
1050 	},
1051 	.eth_dev_init = eth_igb_dev_init,
1052 	.eth_dev_uninit = eth_igb_dev_uninit,
1053 	.dev_private_size = sizeof(struct e1000_adapter),
1054 };
1055 
1056 /*
1057  * virtual function driver struct
1058  */
1059 static struct eth_driver rte_igbvf_pmd = {
1060 	.pci_drv = {
1061 		.name = "rte_igbvf_pmd",
1062 		.id_table = pci_id_igbvf_map,
1063 		.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_DETACHABLE,
1064 	},
1065 	.eth_dev_init = eth_igbvf_dev_init,
1066 	.eth_dev_uninit = eth_igbvf_dev_uninit,
1067 	.dev_private_size = sizeof(struct e1000_adapter),
1068 };
1069 
1070 static int
1071 rte_igb_pmd_init(const char *name __rte_unused, const char *params __rte_unused)
1072 {
1073 	rte_eth_driver_register(&rte_igb_pmd);
1074 	return 0;
1075 }
1076 
1077 static void
1078 igb_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1079 {
1080 	struct e1000_hw *hw =
1081 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1082 	/* RCTL: enable VLAN filter since VMDq always use VLAN filter */
1083 	uint32_t rctl = E1000_READ_REG(hw, E1000_RCTL);
1084 	rctl |= E1000_RCTL_VFE;
1085 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1086 }
1087 
1088 /*
1089  * VF Driver initialization routine.
1090  * Invoked one at EAL init time.
1091  * Register itself as the [Virtual Poll Mode] Driver of PCI IGB devices.
1092  */
1093 static int
1094 rte_igbvf_pmd_init(const char *name __rte_unused, const char *params __rte_unused)
1095 {
1096 	PMD_INIT_FUNC_TRACE();
1097 
1098 	rte_eth_driver_register(&rte_igbvf_pmd);
1099 	return 0;
1100 }
1101 
1102 static int
1103 igb_check_mq_mode(struct rte_eth_dev *dev)
1104 {
1105 	enum rte_eth_rx_mq_mode rx_mq_mode = dev->data->dev_conf.rxmode.mq_mode;
1106 	enum rte_eth_tx_mq_mode tx_mq_mode = dev->data->dev_conf.txmode.mq_mode;
1107 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
1108 	uint16_t nb_tx_q = dev->data->nb_rx_queues;
1109 
1110 	if ((rx_mq_mode & ETH_MQ_RX_DCB_FLAG) ||
1111 	    tx_mq_mode == ETH_MQ_TX_DCB ||
1112 	    tx_mq_mode == ETH_MQ_TX_VMDQ_DCB) {
1113 		PMD_INIT_LOG(ERR, "DCB mode is not supported.");
1114 		return -EINVAL;
1115 	}
1116 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
1117 		/* Check multi-queue mode.
1118 		 * To no break software we accept ETH_MQ_RX_NONE as this might
1119 		 * be used to turn off VLAN filter.
1120 		 */
1121 
1122 		if (rx_mq_mode == ETH_MQ_RX_NONE ||
1123 		    rx_mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1124 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
1125 			RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = 1;
1126 		} else {
1127 			/* Only support one queue on VFs.
1128 			 * RSS together with SRIOV is not supported.
1129 			 */
1130 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1131 					" wrong mq_mode rx %d.",
1132 					rx_mq_mode);
1133 			return -EINVAL;
1134 		}
1135 		/* TX mode is not used here, so mode might be ignored.*/
1136 		if (tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1137 			/* SRIOV only works in VMDq enable mode */
1138 			PMD_INIT_LOG(WARNING, "SRIOV is active,"
1139 					" TX mode %d is not supported. "
1140 					" Driver will behave as %d mode.",
1141 					tx_mq_mode, ETH_MQ_TX_VMDQ_ONLY);
1142 		}
1143 
1144 		/* check valid queue number */
1145 		if ((nb_rx_q > 1) || (nb_tx_q > 1)) {
1146 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1147 					" only support one queue on VFs.");
1148 			return -EINVAL;
1149 		}
1150 	} else {
1151 		/* To no break software that set invalid mode, only display
1152 		 * warning if invalid mode is used.
1153 		 */
1154 		if (rx_mq_mode != ETH_MQ_RX_NONE &&
1155 		    rx_mq_mode != ETH_MQ_RX_VMDQ_ONLY &&
1156 		    rx_mq_mode != ETH_MQ_RX_RSS) {
1157 			/* RSS together with VMDq not supported*/
1158 			PMD_INIT_LOG(ERR, "RX mode %d is not supported.",
1159 				     rx_mq_mode);
1160 			return -EINVAL;
1161 		}
1162 
1163 		if (tx_mq_mode != ETH_MQ_TX_NONE &&
1164 		    tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1165 			PMD_INIT_LOG(WARNING, "TX mode %d is not supported."
1166 					" Due to txmode is meaningless in this"
1167 					" driver, just ignore.",
1168 					tx_mq_mode);
1169 		}
1170 	}
1171 	return 0;
1172 }
1173 
1174 static int
1175 eth_igb_configure(struct rte_eth_dev *dev)
1176 {
1177 	struct e1000_interrupt *intr =
1178 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1179 	int ret;
1180 
1181 	PMD_INIT_FUNC_TRACE();
1182 
1183 	/* multipe queue mode checking */
1184 	ret  = igb_check_mq_mode(dev);
1185 	if (ret != 0) {
1186 		PMD_DRV_LOG(ERR, "igb_check_mq_mode fails with %d.",
1187 			    ret);
1188 		return ret;
1189 	}
1190 
1191 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1192 	PMD_INIT_FUNC_TRACE();
1193 
1194 	return 0;
1195 }
1196 
1197 static int
1198 eth_igb_start(struct rte_eth_dev *dev)
1199 {
1200 	struct e1000_hw *hw =
1201 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1202 	struct e1000_adapter *adapter =
1203 		E1000_DEV_PRIVATE(dev->data->dev_private);
1204 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
1205 	int ret, mask;
1206 	uint32_t intr_vector = 0;
1207 	uint32_t ctrl_ext;
1208 	uint32_t *speeds;
1209 	int num_speeds;
1210 	bool autoneg;
1211 
1212 	PMD_INIT_FUNC_TRACE();
1213 
1214 	/* disable uio/vfio intr/eventfd mapping */
1215 	rte_intr_disable(intr_handle);
1216 
1217 	/* Power up the phy. Needed to make the link go Up */
1218 	eth_igb_dev_set_link_up(dev);
1219 
1220 	/*
1221 	 * Packet Buffer Allocation (PBA)
1222 	 * Writing PBA sets the receive portion of the buffer
1223 	 * the remainder is used for the transmit buffer.
1224 	 */
1225 	if (hw->mac.type == e1000_82575) {
1226 		uint32_t pba;
1227 
1228 		pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
1229 		E1000_WRITE_REG(hw, E1000_PBA, pba);
1230 	}
1231 
1232 	/* Put the address into the Receive Address Array */
1233 	e1000_rar_set(hw, hw->mac.addr, 0);
1234 
1235 	/* Initialize the hardware */
1236 	if (igb_hardware_init(hw)) {
1237 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
1238 		return -EIO;
1239 	}
1240 	adapter->stopped = 0;
1241 
1242 	E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN << 16 | ETHER_TYPE_VLAN);
1243 
1244 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1245 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1246 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
1247 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
1248 	E1000_WRITE_FLUSH(hw);
1249 
1250 	/* configure PF module if SRIOV enabled */
1251 	igb_pf_host_configure(dev);
1252 
1253 	/* check and configure queue intr-vector mapping */
1254 	if ((rte_intr_cap_multiple(intr_handle) ||
1255 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
1256 	    dev->data->dev_conf.intr_conf.rxq != 0) {
1257 		intr_vector = dev->data->nb_rx_queues;
1258 		if (rte_intr_efd_enable(intr_handle, intr_vector))
1259 			return -1;
1260 	}
1261 
1262 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
1263 		intr_handle->intr_vec =
1264 			rte_zmalloc("intr_vec",
1265 				    dev->data->nb_rx_queues * sizeof(int), 0);
1266 		if (intr_handle->intr_vec == NULL) {
1267 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
1268 				     " intr_vec\n", dev->data->nb_rx_queues);
1269 			return -ENOMEM;
1270 		}
1271 	}
1272 
1273 	/* confiugre msix for rx interrupt */
1274 	eth_igb_configure_msix_intr(dev);
1275 
1276 	/* Configure for OS presence */
1277 	igb_init_manageability(hw);
1278 
1279 	eth_igb_tx_init(dev);
1280 
1281 	/* This can fail when allocating mbufs for descriptor rings */
1282 	ret = eth_igb_rx_init(dev);
1283 	if (ret) {
1284 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
1285 		igb_dev_clear_queues(dev);
1286 		return ret;
1287 	}
1288 
1289 	e1000_clear_hw_cntrs_base_generic(hw);
1290 
1291 	/*
1292 	 * VLAN Offload Settings
1293 	 */
1294 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
1295 			ETH_VLAN_EXTEND_MASK;
1296 	eth_igb_vlan_offload_set(dev, mask);
1297 
1298 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1299 		/* Enable VLAN filter since VMDq always use VLAN filter */
1300 		igb_vmdq_vlan_hw_filter_enable(dev);
1301 	}
1302 
1303 	if ((hw->mac.type == e1000_82576) || (hw->mac.type == e1000_82580) ||
1304 		(hw->mac.type == e1000_i350) || (hw->mac.type == e1000_i210) ||
1305 		(hw->mac.type == e1000_i211)) {
1306 		/* Configure EITR with the maximum possible value (0xFFFF) */
1307 		E1000_WRITE_REG(hw, E1000_EITR(0), 0xFFFF);
1308 	}
1309 
1310 	/* Setup link speed and duplex */
1311 	speeds = &dev->data->dev_conf.link_speeds;
1312 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
1313 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
1314 	} else {
1315 		num_speeds = 0;
1316 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
1317 
1318 		/* Reset */
1319 		hw->phy.autoneg_advertised = 0;
1320 
1321 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1322 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1323 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
1324 			num_speeds = -1;
1325 			goto error_invalid_config;
1326 		}
1327 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
1328 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1329 			num_speeds++;
1330 		}
1331 		if (*speeds & ETH_LINK_SPEED_10M) {
1332 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1333 			num_speeds++;
1334 		}
1335 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
1336 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1337 			num_speeds++;
1338 		}
1339 		if (*speeds & ETH_LINK_SPEED_100M) {
1340 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1341 			num_speeds++;
1342 		}
1343 		if (*speeds & ETH_LINK_SPEED_1G) {
1344 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1345 			num_speeds++;
1346 		}
1347 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
1348 			goto error_invalid_config;
1349 	}
1350 
1351 	e1000_setup_link(hw);
1352 
1353 	if (rte_intr_allow_others(intr_handle)) {
1354 		/* check if lsc interrupt is enabled */
1355 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1356 			eth_igb_lsc_interrupt_setup(dev);
1357 	} else {
1358 		rte_intr_callback_unregister(intr_handle,
1359 					     eth_igb_interrupt_handler,
1360 					     (void *)dev);
1361 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1362 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
1363 				     " no intr multiplex\n");
1364 	}
1365 
1366 	/* check if rxq interrupt is enabled */
1367 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
1368 	    rte_intr_dp_is_en(intr_handle))
1369 		eth_igb_rxq_interrupt_setup(dev);
1370 
1371 	/* enable uio/vfio intr/eventfd mapping */
1372 	rte_intr_enable(intr_handle);
1373 
1374 	/* resume enabled intr since hw reset */
1375 	igb_intr_enable(dev);
1376 
1377 	PMD_INIT_LOG(DEBUG, "<<");
1378 
1379 	return 0;
1380 
1381 error_invalid_config:
1382 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
1383 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
1384 	igb_dev_clear_queues(dev);
1385 	return -EINVAL;
1386 }
1387 
1388 /*********************************************************************
1389  *
1390  *  This routine disables all traffic on the adapter by issuing a
1391  *  global reset on the MAC.
1392  *
1393  **********************************************************************/
1394 static void
1395 eth_igb_stop(struct rte_eth_dev *dev)
1396 {
1397 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1398 	struct e1000_filter_info *filter_info =
1399 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
1400 	struct rte_eth_link link;
1401 	struct e1000_flex_filter *p_flex;
1402 	struct e1000_5tuple_filter *p_5tuple, *p_5tuple_next;
1403 	struct e1000_2tuple_filter *p_2tuple, *p_2tuple_next;
1404 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
1405 
1406 	igb_intr_disable(hw);
1407 
1408 	/* disable intr eventfd mapping */
1409 	rte_intr_disable(intr_handle);
1410 
1411 	igb_pf_reset_hw(hw);
1412 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1413 
1414 	/* Set bit for Go Link disconnect */
1415 	if (hw->mac.type >= e1000_82580) {
1416 		uint32_t phpm_reg;
1417 
1418 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1419 		phpm_reg |= E1000_82580_PM_GO_LINKD;
1420 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1421 	}
1422 
1423 	/* Power down the phy. Needed to make the link go Down */
1424 	eth_igb_dev_set_link_down(dev);
1425 
1426 	igb_dev_clear_queues(dev);
1427 
1428 	/* clear the recorded link status */
1429 	memset(&link, 0, sizeof(link));
1430 	rte_igb_dev_atomic_write_link_status(dev, &link);
1431 
1432 	/* Remove all flex filters of the device */
1433 	while ((p_flex = TAILQ_FIRST(&filter_info->flex_list))) {
1434 		TAILQ_REMOVE(&filter_info->flex_list, p_flex, entries);
1435 		rte_free(p_flex);
1436 	}
1437 	filter_info->flex_mask = 0;
1438 
1439 	/* Remove all ntuple filters of the device */
1440 	for (p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list);
1441 	     p_5tuple != NULL; p_5tuple = p_5tuple_next) {
1442 		p_5tuple_next = TAILQ_NEXT(p_5tuple, entries);
1443 		TAILQ_REMOVE(&filter_info->fivetuple_list,
1444 			     p_5tuple, entries);
1445 		rte_free(p_5tuple);
1446 	}
1447 	filter_info->fivetuple_mask = 0;
1448 	for (p_2tuple = TAILQ_FIRST(&filter_info->twotuple_list);
1449 	     p_2tuple != NULL; p_2tuple = p_2tuple_next) {
1450 		p_2tuple_next = TAILQ_NEXT(p_2tuple, entries);
1451 		TAILQ_REMOVE(&filter_info->twotuple_list,
1452 			     p_2tuple, entries);
1453 		rte_free(p_2tuple);
1454 	}
1455 	filter_info->twotuple_mask = 0;
1456 
1457 	if (!rte_intr_allow_others(intr_handle))
1458 		/* resume to the default handler */
1459 		rte_intr_callback_register(intr_handle,
1460 					   eth_igb_interrupt_handler,
1461 					   (void *)dev);
1462 
1463 	/* Clean datapath event and queue/vec mapping */
1464 	rte_intr_efd_disable(intr_handle);
1465 	if (intr_handle->intr_vec != NULL) {
1466 		rte_free(intr_handle->intr_vec);
1467 		intr_handle->intr_vec = NULL;
1468 	}
1469 }
1470 
1471 static int
1472 eth_igb_dev_set_link_up(struct rte_eth_dev *dev)
1473 {
1474 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1475 
1476 	if (hw->phy.media_type == e1000_media_type_copper)
1477 		e1000_power_up_phy(hw);
1478 	else
1479 		e1000_power_up_fiber_serdes_link(hw);
1480 
1481 	return 0;
1482 }
1483 
1484 static int
1485 eth_igb_dev_set_link_down(struct rte_eth_dev *dev)
1486 {
1487 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1488 
1489 	if (hw->phy.media_type == e1000_media_type_copper)
1490 		e1000_power_down_phy(hw);
1491 	else
1492 		e1000_shutdown_fiber_serdes_link(hw);
1493 
1494 	return 0;
1495 }
1496 
1497 static void
1498 eth_igb_close(struct rte_eth_dev *dev)
1499 {
1500 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1501 	struct e1000_adapter *adapter =
1502 		E1000_DEV_PRIVATE(dev->data->dev_private);
1503 	struct rte_eth_link link;
1504 	struct rte_pci_device *pci_dev;
1505 
1506 	eth_igb_stop(dev);
1507 	adapter->stopped = 1;
1508 
1509 	e1000_phy_hw_reset(hw);
1510 	igb_release_manageability(hw);
1511 	igb_hw_control_release(hw);
1512 
1513 	/* Clear bit for Go Link disconnect */
1514 	if (hw->mac.type >= e1000_82580) {
1515 		uint32_t phpm_reg;
1516 
1517 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1518 		phpm_reg &= ~E1000_82580_PM_GO_LINKD;
1519 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1520 	}
1521 
1522 	igb_dev_free_queues(dev);
1523 
1524 	pci_dev = dev->pci_dev;
1525 	if (pci_dev->intr_handle.intr_vec) {
1526 		rte_free(pci_dev->intr_handle.intr_vec);
1527 		pci_dev->intr_handle.intr_vec = NULL;
1528 	}
1529 
1530 	memset(&link, 0, sizeof(link));
1531 	rte_igb_dev_atomic_write_link_status(dev, &link);
1532 }
1533 
1534 static int
1535 igb_get_rx_buffer_size(struct e1000_hw *hw)
1536 {
1537 	uint32_t rx_buf_size;
1538 	if (hw->mac.type == e1000_82576) {
1539 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xffff) << 10;
1540 	} else if (hw->mac.type == e1000_82580 || hw->mac.type == e1000_i350) {
1541 		/* PBS needs to be translated according to a lookup table */
1542 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xf);
1543 		rx_buf_size = (uint32_t) e1000_rxpbs_adjust_82580(rx_buf_size);
1544 		rx_buf_size = (rx_buf_size << 10);
1545 	} else if (hw->mac.type == e1000_i210 || hw->mac.type == e1000_i211) {
1546 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0x3f) << 10;
1547 	} else {
1548 		rx_buf_size = (E1000_READ_REG(hw, E1000_PBA) & 0xffff) << 10;
1549 	}
1550 
1551 	return rx_buf_size;
1552 }
1553 
1554 /*********************************************************************
1555  *
1556  *  Initialize the hardware
1557  *
1558  **********************************************************************/
1559 static int
1560 igb_hardware_init(struct e1000_hw *hw)
1561 {
1562 	uint32_t rx_buf_size;
1563 	int diag;
1564 
1565 	/* Let the firmware know the OS is in control */
1566 	igb_hw_control_acquire(hw);
1567 
1568 	/*
1569 	 * These parameters control the automatic generation (Tx) and
1570 	 * response (Rx) to Ethernet PAUSE frames.
1571 	 * - High water mark should allow for at least two standard size (1518)
1572 	 *   frames to be received after sending an XOFF.
1573 	 * - Low water mark works best when it is very near the high water mark.
1574 	 *   This allows the receiver to restart by sending XON when it has
1575 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
1576 	 *   restart after one full frame is pulled from the buffer. There
1577 	 *   could be several smaller frames in the buffer and if so they will
1578 	 *   not trigger the XON until their total number reduces the buffer
1579 	 *   by 1500.
1580 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
1581 	 */
1582 	rx_buf_size = igb_get_rx_buffer_size(hw);
1583 
1584 	hw->fc.high_water = rx_buf_size - (ETHER_MAX_LEN * 2);
1585 	hw->fc.low_water = hw->fc.high_water - 1500;
1586 	hw->fc.pause_time = IGB_FC_PAUSE_TIME;
1587 	hw->fc.send_xon = 1;
1588 
1589 	/* Set Flow control, use the tunable location if sane */
1590 	if ((igb_fc_setting != e1000_fc_none) && (igb_fc_setting < 4))
1591 		hw->fc.requested_mode = igb_fc_setting;
1592 	else
1593 		hw->fc.requested_mode = e1000_fc_none;
1594 
1595 	/* Issue a global reset */
1596 	igb_pf_reset_hw(hw);
1597 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1598 
1599 	diag = e1000_init_hw(hw);
1600 	if (diag < 0)
1601 		return diag;
1602 
1603 	E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN << 16 | ETHER_TYPE_VLAN);
1604 	e1000_get_phy_info(hw);
1605 	e1000_check_for_link(hw);
1606 
1607 	return 0;
1608 }
1609 
1610 /* This function is based on igb_update_stats_counters() in igb/if_igb.c */
1611 static void
1612 igb_read_stats_registers(struct e1000_hw *hw, struct e1000_hw_stats *stats)
1613 {
1614 	int pause_frames;
1615 
1616 	uint64_t old_gprc  = stats->gprc;
1617 	uint64_t old_gptc  = stats->gptc;
1618 	uint64_t old_tpr   = stats->tpr;
1619 	uint64_t old_tpt   = stats->tpt;
1620 	uint64_t old_rpthc = stats->rpthc;
1621 	uint64_t old_hgptc = stats->hgptc;
1622 
1623 	if(hw->phy.media_type == e1000_media_type_copper ||
1624 	    (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
1625 		stats->symerrs +=
1626 		    E1000_READ_REG(hw,E1000_SYMERRS);
1627 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
1628 	}
1629 
1630 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
1631 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
1632 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
1633 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
1634 
1635 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
1636 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
1637 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
1638 	stats->dc += E1000_READ_REG(hw, E1000_DC);
1639 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
1640 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
1641 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
1642 	/*
1643 	** For watchdog management we need to know if we have been
1644 	** paused during the last interval, so capture that here.
1645 	*/
1646 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
1647 	stats->xoffrxc += pause_frames;
1648 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
1649 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
1650 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
1651 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
1652 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
1653 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
1654 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
1655 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
1656 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
1657 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
1658 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
1659 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
1660 
1661 	/* For the 64-bit byte counters the low dword must be read first. */
1662 	/* Both registers clear on the read of the high dword */
1663 
1664 	/* Workaround CRC bytes included in size, take away 4 bytes/packet */
1665 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
1666 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
1667 	stats->gorc -= (stats->gprc - old_gprc) * ETHER_CRC_LEN;
1668 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
1669 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
1670 	stats->gotc -= (stats->gptc - old_gptc) * ETHER_CRC_LEN;
1671 
1672 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
1673 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
1674 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
1675 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
1676 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
1677 
1678 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
1679 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
1680 
1681 	stats->tor += E1000_READ_REG(hw, E1000_TORL);
1682 	stats->tor += ((uint64_t)E1000_READ_REG(hw, E1000_TORH) << 32);
1683 	stats->tor -= (stats->tpr - old_tpr) * ETHER_CRC_LEN;
1684 	stats->tot += E1000_READ_REG(hw, E1000_TOTL);
1685 	stats->tot += ((uint64_t)E1000_READ_REG(hw, E1000_TOTH) << 32);
1686 	stats->tot -= (stats->tpt - old_tpt) * ETHER_CRC_LEN;
1687 
1688 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
1689 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
1690 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
1691 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
1692 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
1693 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
1694 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
1695 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
1696 
1697 	/* Interrupt Counts */
1698 
1699 	stats->iac += E1000_READ_REG(hw, E1000_IAC);
1700 	stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
1701 	stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
1702 	stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
1703 	stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
1704 	stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
1705 	stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
1706 	stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
1707 	stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
1708 
1709 	/* Host to Card Statistics */
1710 
1711 	stats->cbtmpc += E1000_READ_REG(hw, E1000_CBTMPC);
1712 	stats->htdpmc += E1000_READ_REG(hw, E1000_HTDPMC);
1713 	stats->cbrdpc += E1000_READ_REG(hw, E1000_CBRDPC);
1714 	stats->cbrmpc += E1000_READ_REG(hw, E1000_CBRMPC);
1715 	stats->rpthc += E1000_READ_REG(hw, E1000_RPTHC);
1716 	stats->hgptc += E1000_READ_REG(hw, E1000_HGPTC);
1717 	stats->htcbdpc += E1000_READ_REG(hw, E1000_HTCBDPC);
1718 	stats->hgorc += E1000_READ_REG(hw, E1000_HGORCL);
1719 	stats->hgorc += ((uint64_t)E1000_READ_REG(hw, E1000_HGORCH) << 32);
1720 	stats->hgorc -= (stats->rpthc - old_rpthc) * ETHER_CRC_LEN;
1721 	stats->hgotc += E1000_READ_REG(hw, E1000_HGOTCL);
1722 	stats->hgotc += ((uint64_t)E1000_READ_REG(hw, E1000_HGOTCH) << 32);
1723 	stats->hgotc -= (stats->hgptc - old_hgptc) * ETHER_CRC_LEN;
1724 	stats->lenerrs += E1000_READ_REG(hw, E1000_LENERRS);
1725 	stats->scvpc += E1000_READ_REG(hw, E1000_SCVPC);
1726 	stats->hrmpc += E1000_READ_REG(hw, E1000_HRMPC);
1727 
1728 	stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
1729 	stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
1730 	stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
1731 	stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
1732 	stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
1733 	stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
1734 }
1735 
1736 static void
1737 eth_igb_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1738 {
1739 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1740 	struct e1000_hw_stats *stats =
1741 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1742 
1743 	igb_read_stats_registers(hw, stats);
1744 
1745 	if (rte_stats == NULL)
1746 		return;
1747 
1748 	/* Rx Errors */
1749 	rte_stats->imissed = stats->mpc;
1750 	rte_stats->ierrors = stats->crcerrs +
1751 	                     stats->rlec + stats->ruc + stats->roc +
1752 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
1753 
1754 	/* Tx Errors */
1755 	rte_stats->oerrors = stats->ecol + stats->latecol;
1756 
1757 	rte_stats->ipackets = stats->gprc;
1758 	rte_stats->opackets = stats->gptc;
1759 	rte_stats->ibytes   = stats->gorc;
1760 	rte_stats->obytes   = stats->gotc;
1761 }
1762 
1763 static void
1764 eth_igb_stats_reset(struct rte_eth_dev *dev)
1765 {
1766 	struct e1000_hw_stats *hw_stats =
1767 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1768 
1769 	/* HW registers are cleared on read */
1770 	eth_igb_stats_get(dev, NULL);
1771 
1772 	/* Reset software totals */
1773 	memset(hw_stats, 0, sizeof(*hw_stats));
1774 }
1775 
1776 static void
1777 eth_igb_xstats_reset(struct rte_eth_dev *dev)
1778 {
1779 	struct e1000_hw_stats *stats =
1780 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1781 
1782 	/* HW registers are cleared on read */
1783 	eth_igb_xstats_get(dev, NULL, IGB_NB_XSTATS);
1784 
1785 	/* Reset software totals */
1786 	memset(stats, 0, sizeof(*stats));
1787 }
1788 
1789 static int eth_igb_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1790 	struct rte_eth_xstat_name *xstats_names,
1791 	__rte_unused unsigned limit)
1792 {
1793 	unsigned i;
1794 
1795 	if (xstats_names == NULL)
1796 		return IGB_NB_XSTATS;
1797 
1798 	/* Note: limit checked in rte_eth_xstats_names() */
1799 
1800 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1801 		snprintf(xstats_names[i].name, sizeof(xstats_names[i].name),
1802 			 "%s", rte_igb_stats_strings[i].name);
1803 	}
1804 
1805 	return IGB_NB_XSTATS;
1806 }
1807 
1808 static int
1809 eth_igb_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1810 		   unsigned n)
1811 {
1812 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1813 	struct e1000_hw_stats *hw_stats =
1814 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1815 	unsigned i;
1816 
1817 	if (n < IGB_NB_XSTATS)
1818 		return IGB_NB_XSTATS;
1819 
1820 	igb_read_stats_registers(hw, hw_stats);
1821 
1822 	/* If this is a reset xstats is NULL, and we have cleared the
1823 	 * registers by reading them.
1824 	 */
1825 	if (!xstats)
1826 		return 0;
1827 
1828 	/* Extended stats */
1829 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1830 		xstats[i].id = i;
1831 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1832 			rte_igb_stats_strings[i].offset);
1833 	}
1834 
1835 	return IGB_NB_XSTATS;
1836 }
1837 
1838 static void
1839 igbvf_read_stats_registers(struct e1000_hw *hw, struct e1000_vf_stats *hw_stats)
1840 {
1841 	/* Good Rx packets, include VF loopback */
1842 	UPDATE_VF_STAT(E1000_VFGPRC,
1843 	    hw_stats->last_gprc, hw_stats->gprc);
1844 
1845 	/* Good Rx octets, include VF loopback */
1846 	UPDATE_VF_STAT(E1000_VFGORC,
1847 	    hw_stats->last_gorc, hw_stats->gorc);
1848 
1849 	/* Good Tx packets, include VF loopback */
1850 	UPDATE_VF_STAT(E1000_VFGPTC,
1851 	    hw_stats->last_gptc, hw_stats->gptc);
1852 
1853 	/* Good Tx octets, include VF loopback */
1854 	UPDATE_VF_STAT(E1000_VFGOTC,
1855 	    hw_stats->last_gotc, hw_stats->gotc);
1856 
1857 	/* Rx Multicst packets */
1858 	UPDATE_VF_STAT(E1000_VFMPRC,
1859 	    hw_stats->last_mprc, hw_stats->mprc);
1860 
1861 	/* Good Rx loopback packets */
1862 	UPDATE_VF_STAT(E1000_VFGPRLBC,
1863 	    hw_stats->last_gprlbc, hw_stats->gprlbc);
1864 
1865 	/* Good Rx loopback octets */
1866 	UPDATE_VF_STAT(E1000_VFGORLBC,
1867 	    hw_stats->last_gorlbc, hw_stats->gorlbc);
1868 
1869 	/* Good Tx loopback packets */
1870 	UPDATE_VF_STAT(E1000_VFGPTLBC,
1871 	    hw_stats->last_gptlbc, hw_stats->gptlbc);
1872 
1873 	/* Good Tx loopback octets */
1874 	UPDATE_VF_STAT(E1000_VFGOTLBC,
1875 	    hw_stats->last_gotlbc, hw_stats->gotlbc);
1876 }
1877 
1878 static int eth_igbvf_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1879 				     struct rte_eth_xstat_name *xstats_names,
1880 				     __rte_unused unsigned limit)
1881 {
1882 	unsigned i;
1883 
1884 	if (xstats_names != NULL)
1885 		for (i = 0; i < IGBVF_NB_XSTATS; i++) {
1886 			snprintf(xstats_names[i].name,
1887 				sizeof(xstats_names[i].name), "%s",
1888 				rte_igbvf_stats_strings[i].name);
1889 		}
1890 	return IGBVF_NB_XSTATS;
1891 }
1892 
1893 static int
1894 eth_igbvf_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1895 		     unsigned n)
1896 {
1897 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1898 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
1899 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1900 	unsigned i;
1901 
1902 	if (n < IGBVF_NB_XSTATS)
1903 		return IGBVF_NB_XSTATS;
1904 
1905 	igbvf_read_stats_registers(hw, hw_stats);
1906 
1907 	if (!xstats)
1908 		return 0;
1909 
1910 	for (i = 0; i < IGBVF_NB_XSTATS; i++) {
1911 		xstats[i].id = i;
1912 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1913 			rte_igbvf_stats_strings[i].offset);
1914 	}
1915 
1916 	return IGBVF_NB_XSTATS;
1917 }
1918 
1919 static void
1920 eth_igbvf_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1921 {
1922 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1923 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
1924 			  E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1925 
1926 	igbvf_read_stats_registers(hw, hw_stats);
1927 
1928 	if (rte_stats == NULL)
1929 		return;
1930 
1931 	rte_stats->ipackets = hw_stats->gprc;
1932 	rte_stats->ibytes = hw_stats->gorc;
1933 	rte_stats->opackets = hw_stats->gptc;
1934 	rte_stats->obytes = hw_stats->gotc;
1935 }
1936 
1937 static void
1938 eth_igbvf_stats_reset(struct rte_eth_dev *dev)
1939 {
1940 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats*)
1941 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1942 
1943 	/* Sync HW register to the last stats */
1944 	eth_igbvf_stats_get(dev, NULL);
1945 
1946 	/* reset HW current stats*/
1947 	memset(&hw_stats->gprc, 0, sizeof(*hw_stats) -
1948 	       offsetof(struct e1000_vf_stats, gprc));
1949 }
1950 
1951 static void
1952 eth_igb_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1953 {
1954 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1955 
1956 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1957 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
1958 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1959 	dev_info->rx_offload_capa =
1960 		DEV_RX_OFFLOAD_VLAN_STRIP |
1961 		DEV_RX_OFFLOAD_IPV4_CKSUM |
1962 		DEV_RX_OFFLOAD_UDP_CKSUM  |
1963 		DEV_RX_OFFLOAD_TCP_CKSUM;
1964 	dev_info->tx_offload_capa =
1965 		DEV_TX_OFFLOAD_VLAN_INSERT |
1966 		DEV_TX_OFFLOAD_IPV4_CKSUM  |
1967 		DEV_TX_OFFLOAD_UDP_CKSUM   |
1968 		DEV_TX_OFFLOAD_TCP_CKSUM   |
1969 		DEV_TX_OFFLOAD_SCTP_CKSUM  |
1970 		DEV_TX_OFFLOAD_TCP_TSO;
1971 
1972 	switch (hw->mac.type) {
1973 	case e1000_82575:
1974 		dev_info->max_rx_queues = 4;
1975 		dev_info->max_tx_queues = 4;
1976 		dev_info->max_vmdq_pools = 0;
1977 		break;
1978 
1979 	case e1000_82576:
1980 		dev_info->max_rx_queues = 16;
1981 		dev_info->max_tx_queues = 16;
1982 		dev_info->max_vmdq_pools = ETH_8_POOLS;
1983 		dev_info->vmdq_queue_num = 16;
1984 		break;
1985 
1986 	case e1000_82580:
1987 		dev_info->max_rx_queues = 8;
1988 		dev_info->max_tx_queues = 8;
1989 		dev_info->max_vmdq_pools = ETH_8_POOLS;
1990 		dev_info->vmdq_queue_num = 8;
1991 		break;
1992 
1993 	case e1000_i350:
1994 		dev_info->max_rx_queues = 8;
1995 		dev_info->max_tx_queues = 8;
1996 		dev_info->max_vmdq_pools = ETH_8_POOLS;
1997 		dev_info->vmdq_queue_num = 8;
1998 		break;
1999 
2000 	case e1000_i354:
2001 		dev_info->max_rx_queues = 8;
2002 		dev_info->max_tx_queues = 8;
2003 		break;
2004 
2005 	case e1000_i210:
2006 		dev_info->max_rx_queues = 4;
2007 		dev_info->max_tx_queues = 4;
2008 		dev_info->max_vmdq_pools = 0;
2009 		break;
2010 
2011 	case e1000_i211:
2012 		dev_info->max_rx_queues = 2;
2013 		dev_info->max_tx_queues = 2;
2014 		dev_info->max_vmdq_pools = 0;
2015 		break;
2016 
2017 	default:
2018 		/* Should not happen */
2019 		break;
2020 	}
2021 	dev_info->hash_key_size = IGB_HKEY_MAX_INDEX * sizeof(uint32_t);
2022 	dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
2023 	dev_info->flow_type_rss_offloads = IGB_RSS_OFFLOAD_ALL;
2024 
2025 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2026 		.rx_thresh = {
2027 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2028 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2029 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2030 		},
2031 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2032 		.rx_drop_en = 0,
2033 	};
2034 
2035 	dev_info->default_txconf = (struct rte_eth_txconf) {
2036 		.tx_thresh = {
2037 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2038 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2039 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2040 		},
2041 		.txq_flags = 0,
2042 	};
2043 
2044 	dev_info->rx_desc_lim = rx_desc_lim;
2045 	dev_info->tx_desc_lim = tx_desc_lim;
2046 
2047 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
2048 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
2049 			ETH_LINK_SPEED_1G;
2050 }
2051 
2052 static const uint32_t *
2053 eth_igb_supported_ptypes_get(struct rte_eth_dev *dev)
2054 {
2055 	static const uint32_t ptypes[] = {
2056 		/* refers to igb_rxd_pkt_info_to_pkt_type() */
2057 		RTE_PTYPE_L2_ETHER,
2058 		RTE_PTYPE_L3_IPV4,
2059 		RTE_PTYPE_L3_IPV4_EXT,
2060 		RTE_PTYPE_L3_IPV6,
2061 		RTE_PTYPE_L3_IPV6_EXT,
2062 		RTE_PTYPE_L4_TCP,
2063 		RTE_PTYPE_L4_UDP,
2064 		RTE_PTYPE_L4_SCTP,
2065 		RTE_PTYPE_TUNNEL_IP,
2066 		RTE_PTYPE_INNER_L3_IPV6,
2067 		RTE_PTYPE_INNER_L3_IPV6_EXT,
2068 		RTE_PTYPE_INNER_L4_TCP,
2069 		RTE_PTYPE_INNER_L4_UDP,
2070 		RTE_PTYPE_UNKNOWN
2071 	};
2072 
2073 	if (dev->rx_pkt_burst == eth_igb_recv_pkts ||
2074 	    dev->rx_pkt_burst == eth_igb_recv_scattered_pkts)
2075 		return ptypes;
2076 	return NULL;
2077 }
2078 
2079 static void
2080 eth_igbvf_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2081 {
2082 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2083 
2084 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2085 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2086 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2087 	dev_info->rx_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP |
2088 				DEV_RX_OFFLOAD_IPV4_CKSUM |
2089 				DEV_RX_OFFLOAD_UDP_CKSUM  |
2090 				DEV_RX_OFFLOAD_TCP_CKSUM;
2091 	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
2092 				DEV_TX_OFFLOAD_IPV4_CKSUM  |
2093 				DEV_TX_OFFLOAD_UDP_CKSUM   |
2094 				DEV_TX_OFFLOAD_TCP_CKSUM   |
2095 				DEV_TX_OFFLOAD_SCTP_CKSUM  |
2096 				DEV_TX_OFFLOAD_TCP_TSO;
2097 	switch (hw->mac.type) {
2098 	case e1000_vfadapt:
2099 		dev_info->max_rx_queues = 2;
2100 		dev_info->max_tx_queues = 2;
2101 		break;
2102 	case e1000_vfadapt_i350:
2103 		dev_info->max_rx_queues = 1;
2104 		dev_info->max_tx_queues = 1;
2105 		break;
2106 	default:
2107 		/* Should not happen */
2108 		break;
2109 	}
2110 
2111 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2112 		.rx_thresh = {
2113 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2114 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2115 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2116 		},
2117 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2118 		.rx_drop_en = 0,
2119 	};
2120 
2121 	dev_info->default_txconf = (struct rte_eth_txconf) {
2122 		.tx_thresh = {
2123 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2124 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2125 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2126 		},
2127 		.txq_flags = 0,
2128 	};
2129 
2130 	dev_info->rx_desc_lim = rx_desc_lim;
2131 	dev_info->tx_desc_lim = tx_desc_lim;
2132 }
2133 
2134 /* return 0 means link status changed, -1 means not changed */
2135 static int
2136 eth_igb_link_update(struct rte_eth_dev *dev, int wait_to_complete)
2137 {
2138 	struct e1000_hw *hw =
2139 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2140 	struct rte_eth_link link, old;
2141 	int link_check, count;
2142 
2143 	link_check = 0;
2144 	hw->mac.get_link_status = 1;
2145 
2146 	/* possible wait-to-complete in up to 9 seconds */
2147 	for (count = 0; count < IGB_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
2148 		/* Read the real link status */
2149 		switch (hw->phy.media_type) {
2150 		case e1000_media_type_copper:
2151 			/* Do the work to read phy */
2152 			e1000_check_for_link(hw);
2153 			link_check = !hw->mac.get_link_status;
2154 			break;
2155 
2156 		case e1000_media_type_fiber:
2157 			e1000_check_for_link(hw);
2158 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
2159 				      E1000_STATUS_LU);
2160 			break;
2161 
2162 		case e1000_media_type_internal_serdes:
2163 			e1000_check_for_link(hw);
2164 			link_check = hw->mac.serdes_has_link;
2165 			break;
2166 
2167 		/* VF device is type_unknown */
2168 		case e1000_media_type_unknown:
2169 			eth_igbvf_link_update(hw);
2170 			link_check = !hw->mac.get_link_status;
2171 			break;
2172 
2173 		default:
2174 			break;
2175 		}
2176 		if (link_check || wait_to_complete == 0)
2177 			break;
2178 		rte_delay_ms(IGB_LINK_UPDATE_CHECK_INTERVAL);
2179 	}
2180 	memset(&link, 0, sizeof(link));
2181 	rte_igb_dev_atomic_read_link_status(dev, &link);
2182 	old = link;
2183 
2184 	/* Now we check if a transition has happened */
2185 	if (link_check) {
2186 		uint16_t duplex, speed;
2187 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
2188 		link.link_duplex = (duplex == FULL_DUPLEX) ?
2189 				ETH_LINK_FULL_DUPLEX :
2190 				ETH_LINK_HALF_DUPLEX;
2191 		link.link_speed = speed;
2192 		link.link_status = ETH_LINK_UP;
2193 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
2194 				ETH_LINK_SPEED_FIXED);
2195 	} else if (!link_check) {
2196 		link.link_speed = 0;
2197 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
2198 		link.link_status = ETH_LINK_DOWN;
2199 		link.link_autoneg = ETH_LINK_SPEED_FIXED;
2200 	}
2201 	rte_igb_dev_atomic_write_link_status(dev, &link);
2202 
2203 	/* not changed */
2204 	if (old.link_status == link.link_status)
2205 		return -1;
2206 
2207 	/* changed */
2208 	return 0;
2209 }
2210 
2211 /*
2212  * igb_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
2213  * For ASF and Pass Through versions of f/w this means
2214  * that the driver is loaded.
2215  */
2216 static void
2217 igb_hw_control_acquire(struct e1000_hw *hw)
2218 {
2219 	uint32_t ctrl_ext;
2220 
2221 	/* Let firmware know the driver has taken over */
2222 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2223 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2224 }
2225 
2226 /*
2227  * igb_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
2228  * For ASF and Pass Through versions of f/w this means that the
2229  * driver is no longer loaded.
2230  */
2231 static void
2232 igb_hw_control_release(struct e1000_hw *hw)
2233 {
2234 	uint32_t ctrl_ext;
2235 
2236 	/* Let firmware taken over control of h/w */
2237 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2238 	E1000_WRITE_REG(hw, E1000_CTRL_EXT,
2239 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2240 }
2241 
2242 /*
2243  * Bit of a misnomer, what this really means is
2244  * to enable OS management of the system... aka
2245  * to disable special hardware management features.
2246  */
2247 static void
2248 igb_init_manageability(struct e1000_hw *hw)
2249 {
2250 	if (e1000_enable_mng_pass_thru(hw)) {
2251 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
2252 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2253 
2254 		/* disable hardware interception of ARP */
2255 		manc &= ~(E1000_MANC_ARP_EN);
2256 
2257 		/* enable receiving management packets to the host */
2258 		manc |= E1000_MANC_EN_MNG2HOST;
2259 		manc2h |= 1 << 5;  /* Mng Port 623 */
2260 		manc2h |= 1 << 6;  /* Mng Port 664 */
2261 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
2262 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2263 	}
2264 }
2265 
2266 static void
2267 igb_release_manageability(struct e1000_hw *hw)
2268 {
2269 	if (e1000_enable_mng_pass_thru(hw)) {
2270 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2271 
2272 		manc |= E1000_MANC_ARP_EN;
2273 		manc &= ~E1000_MANC_EN_MNG2HOST;
2274 
2275 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2276 	}
2277 }
2278 
2279 static void
2280 eth_igb_promiscuous_enable(struct rte_eth_dev *dev)
2281 {
2282 	struct e1000_hw *hw =
2283 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2284 	uint32_t rctl;
2285 
2286 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2287 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
2288 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2289 }
2290 
2291 static void
2292 eth_igb_promiscuous_disable(struct rte_eth_dev *dev)
2293 {
2294 	struct e1000_hw *hw =
2295 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2296 	uint32_t rctl;
2297 
2298 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2299 	rctl &= (~E1000_RCTL_UPE);
2300 	if (dev->data->all_multicast == 1)
2301 		rctl |= E1000_RCTL_MPE;
2302 	else
2303 		rctl &= (~E1000_RCTL_MPE);
2304 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2305 }
2306 
2307 static void
2308 eth_igb_allmulticast_enable(struct rte_eth_dev *dev)
2309 {
2310 	struct e1000_hw *hw =
2311 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2312 	uint32_t rctl;
2313 
2314 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2315 	rctl |= E1000_RCTL_MPE;
2316 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2317 }
2318 
2319 static void
2320 eth_igb_allmulticast_disable(struct rte_eth_dev *dev)
2321 {
2322 	struct e1000_hw *hw =
2323 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2324 	uint32_t rctl;
2325 
2326 	if (dev->data->promiscuous == 1)
2327 		return; /* must remain in all_multicast mode */
2328 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2329 	rctl &= (~E1000_RCTL_MPE);
2330 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2331 }
2332 
2333 static int
2334 eth_igb_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2335 {
2336 	struct e1000_hw *hw =
2337 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2338 	struct e1000_vfta * shadow_vfta =
2339 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2340 	uint32_t vfta;
2341 	uint32_t vid_idx;
2342 	uint32_t vid_bit;
2343 
2344 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
2345 			      E1000_VFTA_ENTRY_MASK);
2346 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
2347 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
2348 	if (on)
2349 		vfta |= vid_bit;
2350 	else
2351 		vfta &= ~vid_bit;
2352 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
2353 
2354 	/* update local VFTA copy */
2355 	shadow_vfta->vfta[vid_idx] = vfta;
2356 
2357 	return 0;
2358 }
2359 
2360 static int
2361 eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
2362 		      enum rte_vlan_type vlan_type,
2363 		      uint16_t tpid)
2364 {
2365 	struct e1000_hw *hw =
2366 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2367 	uint32_t reg, qinq;
2368 
2369 	qinq = E1000_READ_REG(hw, E1000_CTRL_EXT);
2370 	qinq &= E1000_CTRL_EXT_EXT_VLAN;
2371 
2372 	/* only outer TPID of double VLAN can be configured*/
2373 	if (qinq && vlan_type == ETH_VLAN_TYPE_OUTER) {
2374 		reg = E1000_READ_REG(hw, E1000_VET);
2375 		reg = (reg & (~E1000_VET_VET_EXT)) |
2376 			((uint32_t)tpid << E1000_VET_VET_EXT_SHIFT);
2377 		E1000_WRITE_REG(hw, E1000_VET, reg);
2378 
2379 		return 0;
2380 	}
2381 
2382 	/* all other TPID values are read-only*/
2383 	PMD_DRV_LOG(ERR, "Not supported");
2384 
2385 	return -ENOTSUP;
2386 }
2387 
2388 static void
2389 igb_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2390 {
2391 	struct e1000_hw *hw =
2392 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2393 	uint32_t reg;
2394 
2395 	/* Filter Table Disable */
2396 	reg = E1000_READ_REG(hw, E1000_RCTL);
2397 	reg &= ~E1000_RCTL_CFIEN;
2398 	reg &= ~E1000_RCTL_VFE;
2399 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2400 }
2401 
2402 static void
2403 igb_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2404 {
2405 	struct e1000_hw *hw =
2406 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2407 	struct e1000_vfta * shadow_vfta =
2408 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2409 	uint32_t reg;
2410 	int i;
2411 
2412 	/* Filter Table Enable, CFI not used for packet acceptance */
2413 	reg = E1000_READ_REG(hw, E1000_RCTL);
2414 	reg &= ~E1000_RCTL_CFIEN;
2415 	reg |= E1000_RCTL_VFE;
2416 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2417 
2418 	/* restore VFTA table */
2419 	for (i = 0; i < IGB_VFTA_SIZE; i++)
2420 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
2421 }
2422 
2423 static void
2424 igb_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2425 {
2426 	struct e1000_hw *hw =
2427 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2428 	uint32_t reg;
2429 
2430 	/* VLAN Mode Disable */
2431 	reg = E1000_READ_REG(hw, E1000_CTRL);
2432 	reg &= ~E1000_CTRL_VME;
2433 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2434 }
2435 
2436 static void
2437 igb_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2438 {
2439 	struct e1000_hw *hw =
2440 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2441 	uint32_t reg;
2442 
2443 	/* VLAN Mode Enable */
2444 	reg = E1000_READ_REG(hw, E1000_CTRL);
2445 	reg |= E1000_CTRL_VME;
2446 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2447 }
2448 
2449 static void
2450 igb_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2451 {
2452 	struct e1000_hw *hw =
2453 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2454 	uint32_t reg;
2455 
2456 	/* CTRL_EXT: Extended VLAN */
2457 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2458 	reg &= ~E1000_CTRL_EXT_EXTEND_VLAN;
2459 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2460 
2461 	/* Update maximum packet length */
2462 	if (dev->data->dev_conf.rxmode.jumbo_frame == 1)
2463 		E1000_WRITE_REG(hw, E1000_RLPML,
2464 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2465 						VLAN_TAG_SIZE);
2466 }
2467 
2468 static void
2469 igb_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2470 {
2471 	struct e1000_hw *hw =
2472 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2473 	uint32_t reg;
2474 
2475 	/* CTRL_EXT: Extended VLAN */
2476 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2477 	reg |= E1000_CTRL_EXT_EXTEND_VLAN;
2478 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2479 
2480 	/* Update maximum packet length */
2481 	if (dev->data->dev_conf.rxmode.jumbo_frame == 1)
2482 		E1000_WRITE_REG(hw, E1000_RLPML,
2483 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2484 						2 * VLAN_TAG_SIZE);
2485 }
2486 
2487 static void
2488 eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2489 {
2490 	if(mask & ETH_VLAN_STRIP_MASK){
2491 		if (dev->data->dev_conf.rxmode.hw_vlan_strip)
2492 			igb_vlan_hw_strip_enable(dev);
2493 		else
2494 			igb_vlan_hw_strip_disable(dev);
2495 	}
2496 
2497 	if(mask & ETH_VLAN_FILTER_MASK){
2498 		if (dev->data->dev_conf.rxmode.hw_vlan_filter)
2499 			igb_vlan_hw_filter_enable(dev);
2500 		else
2501 			igb_vlan_hw_filter_disable(dev);
2502 	}
2503 
2504 	if(mask & ETH_VLAN_EXTEND_MASK){
2505 		if (dev->data->dev_conf.rxmode.hw_vlan_extend)
2506 			igb_vlan_hw_extend_enable(dev);
2507 		else
2508 			igb_vlan_hw_extend_disable(dev);
2509 	}
2510 }
2511 
2512 
2513 /**
2514  * It enables the interrupt mask and then enable the interrupt.
2515  *
2516  * @param dev
2517  *  Pointer to struct rte_eth_dev.
2518  *
2519  * @return
2520  *  - On success, zero.
2521  *  - On failure, a negative value.
2522  */
2523 static int
2524 eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev)
2525 {
2526 	struct e1000_interrupt *intr =
2527 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2528 
2529 	intr->mask |= E1000_ICR_LSC;
2530 
2531 	return 0;
2532 }
2533 
2534 /* It clears the interrupt causes and enables the interrupt.
2535  * It will be called once only during nic initialized.
2536  *
2537  * @param dev
2538  *  Pointer to struct rte_eth_dev.
2539  *
2540  * @return
2541  *  - On success, zero.
2542  *  - On failure, a negative value.
2543  */
2544 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev)
2545 {
2546 	uint32_t mask, regval;
2547 	struct e1000_hw *hw =
2548 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2549 	struct rte_eth_dev_info dev_info;
2550 
2551 	memset(&dev_info, 0, sizeof(dev_info));
2552 	eth_igb_infos_get(dev, &dev_info);
2553 
2554 	mask = 0xFFFFFFFF >> (32 - dev_info.max_rx_queues);
2555 	regval = E1000_READ_REG(hw, E1000_EIMS);
2556 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
2557 
2558 	return 0;
2559 }
2560 
2561 /*
2562  * It reads ICR and gets interrupt causes, check it and set a bit flag
2563  * to update link status.
2564  *
2565  * @param dev
2566  *  Pointer to struct rte_eth_dev.
2567  *
2568  * @return
2569  *  - On success, zero.
2570  *  - On failure, a negative value.
2571  */
2572 static int
2573 eth_igb_interrupt_get_status(struct rte_eth_dev *dev)
2574 {
2575 	uint32_t icr;
2576 	struct e1000_hw *hw =
2577 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2578 	struct e1000_interrupt *intr =
2579 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2580 
2581 	igb_intr_disable(hw);
2582 
2583 	/* read-on-clear nic registers here */
2584 	icr = E1000_READ_REG(hw, E1000_ICR);
2585 
2586 	intr->flags = 0;
2587 	if (icr & E1000_ICR_LSC) {
2588 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
2589 	}
2590 
2591 	if (icr & E1000_ICR_VMMB)
2592 		intr->flags |= E1000_FLAG_MAILBOX;
2593 
2594 	return 0;
2595 }
2596 
2597 /*
2598  * It executes link_update after knowing an interrupt is prsent.
2599  *
2600  * @param dev
2601  *  Pointer to struct rte_eth_dev.
2602  *
2603  * @return
2604  *  - On success, zero.
2605  *  - On failure, a negative value.
2606  */
2607 static int
2608 eth_igb_interrupt_action(struct rte_eth_dev *dev)
2609 {
2610 	struct e1000_hw *hw =
2611 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2612 	struct e1000_interrupt *intr =
2613 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2614 	uint32_t tctl, rctl;
2615 	struct rte_eth_link link;
2616 	int ret;
2617 
2618 	if (intr->flags & E1000_FLAG_MAILBOX) {
2619 		igb_pf_mbx_process(dev);
2620 		intr->flags &= ~E1000_FLAG_MAILBOX;
2621 	}
2622 
2623 	igb_intr_enable(dev);
2624 	rte_intr_enable(&(dev->pci_dev->intr_handle));
2625 
2626 	if (intr->flags & E1000_FLAG_NEED_LINK_UPDATE) {
2627 		intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
2628 
2629 		/* set get_link_status to check register later */
2630 		hw->mac.get_link_status = 1;
2631 		ret = eth_igb_link_update(dev, 0);
2632 
2633 		/* check if link has changed */
2634 		if (ret < 0)
2635 			return 0;
2636 
2637 		memset(&link, 0, sizeof(link));
2638 		rte_igb_dev_atomic_read_link_status(dev, &link);
2639 		if (link.link_status) {
2640 			PMD_INIT_LOG(INFO,
2641 				     " Port %d: Link Up - speed %u Mbps - %s",
2642 				     dev->data->port_id,
2643 				     (unsigned)link.link_speed,
2644 				     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
2645 				     "full-duplex" : "half-duplex");
2646 		} else {
2647 			PMD_INIT_LOG(INFO, " Port %d: Link Down",
2648 				     dev->data->port_id);
2649 		}
2650 
2651 		PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
2652 			     dev->pci_dev->addr.domain,
2653 			     dev->pci_dev->addr.bus,
2654 			     dev->pci_dev->addr.devid,
2655 			     dev->pci_dev->addr.function);
2656 		tctl = E1000_READ_REG(hw, E1000_TCTL);
2657 		rctl = E1000_READ_REG(hw, E1000_RCTL);
2658 		if (link.link_status) {
2659 			/* enable Tx/Rx */
2660 			tctl |= E1000_TCTL_EN;
2661 			rctl |= E1000_RCTL_EN;
2662 		} else {
2663 			/* disable Tx/Rx */
2664 			tctl &= ~E1000_TCTL_EN;
2665 			rctl &= ~E1000_RCTL_EN;
2666 		}
2667 		E1000_WRITE_REG(hw, E1000_TCTL, tctl);
2668 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2669 		E1000_WRITE_FLUSH(hw);
2670 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC);
2671 	}
2672 
2673 	return 0;
2674 }
2675 
2676 /**
2677  * Interrupt handler which shall be registered at first.
2678  *
2679  * @param handle
2680  *  Pointer to interrupt handle.
2681  * @param param
2682  *  The address of parameter (struct rte_eth_dev *) regsitered before.
2683  *
2684  * @return
2685  *  void
2686  */
2687 static void
2688 eth_igb_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
2689 							void *param)
2690 {
2691 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2692 
2693 	eth_igb_interrupt_get_status(dev);
2694 	eth_igb_interrupt_action(dev);
2695 }
2696 
2697 static int
2698 eth_igbvf_interrupt_get_status(struct rte_eth_dev *dev)
2699 {
2700 	uint32_t eicr;
2701 	struct e1000_hw *hw =
2702 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2703 	struct e1000_interrupt *intr =
2704 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2705 
2706 	igbvf_intr_disable(hw);
2707 
2708 	/* read-on-clear nic registers here */
2709 	eicr = E1000_READ_REG(hw, E1000_EICR);
2710 	intr->flags = 0;
2711 
2712 	if (eicr == E1000_VTIVAR_MISC_MAILBOX)
2713 		intr->flags |= E1000_FLAG_MAILBOX;
2714 
2715 	return 0;
2716 }
2717 
2718 void igbvf_mbx_process(struct rte_eth_dev *dev)
2719 {
2720 	struct e1000_hw *hw =
2721 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2722 	struct e1000_mbx_info *mbx = &hw->mbx;
2723 	u32 in_msg = 0;
2724 
2725 	if (mbx->ops.read(hw, &in_msg, 1, 0))
2726 		return;
2727 
2728 	/* PF reset VF event */
2729 	if (in_msg == E1000_PF_CONTROL_MSG)
2730 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET);
2731 }
2732 
2733 static int
2734 eth_igbvf_interrupt_action(struct rte_eth_dev *dev)
2735 {
2736 	struct e1000_interrupt *intr =
2737 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2738 
2739 	if (intr->flags & E1000_FLAG_MAILBOX) {
2740 		igbvf_mbx_process(dev);
2741 		intr->flags &= ~E1000_FLAG_MAILBOX;
2742 	}
2743 
2744 	igbvf_intr_enable(dev);
2745 	rte_intr_enable(&dev->pci_dev->intr_handle);
2746 
2747 	return 0;
2748 }
2749 
2750 static void
2751 eth_igbvf_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
2752 			    void *param)
2753 {
2754 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2755 
2756 	eth_igbvf_interrupt_get_status(dev);
2757 	eth_igbvf_interrupt_action(dev);
2758 }
2759 
2760 static int
2761 eth_igb_led_on(struct rte_eth_dev *dev)
2762 {
2763 	struct e1000_hw *hw;
2764 
2765 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2766 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
2767 }
2768 
2769 static int
2770 eth_igb_led_off(struct rte_eth_dev *dev)
2771 {
2772 	struct e1000_hw *hw;
2773 
2774 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2775 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
2776 }
2777 
2778 static int
2779 eth_igb_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2780 {
2781 	struct e1000_hw *hw;
2782 	uint32_t ctrl;
2783 	int tx_pause;
2784 	int rx_pause;
2785 
2786 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2787 	fc_conf->pause_time = hw->fc.pause_time;
2788 	fc_conf->high_water = hw->fc.high_water;
2789 	fc_conf->low_water = hw->fc.low_water;
2790 	fc_conf->send_xon = hw->fc.send_xon;
2791 	fc_conf->autoneg = hw->mac.autoneg;
2792 
2793 	/*
2794 	 * Return rx_pause and tx_pause status according to actual setting of
2795 	 * the TFCE and RFCE bits in the CTRL register.
2796 	 */
2797 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
2798 	if (ctrl & E1000_CTRL_TFCE)
2799 		tx_pause = 1;
2800 	else
2801 		tx_pause = 0;
2802 
2803 	if (ctrl & E1000_CTRL_RFCE)
2804 		rx_pause = 1;
2805 	else
2806 		rx_pause = 0;
2807 
2808 	if (rx_pause && tx_pause)
2809 		fc_conf->mode = RTE_FC_FULL;
2810 	else if (rx_pause)
2811 		fc_conf->mode = RTE_FC_RX_PAUSE;
2812 	else if (tx_pause)
2813 		fc_conf->mode = RTE_FC_TX_PAUSE;
2814 	else
2815 		fc_conf->mode = RTE_FC_NONE;
2816 
2817 	return 0;
2818 }
2819 
2820 static int
2821 eth_igb_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2822 {
2823 	struct e1000_hw *hw;
2824 	int err;
2825 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
2826 		e1000_fc_none,
2827 		e1000_fc_rx_pause,
2828 		e1000_fc_tx_pause,
2829 		e1000_fc_full
2830 	};
2831 	uint32_t rx_buf_size;
2832 	uint32_t max_high_water;
2833 	uint32_t rctl;
2834 
2835 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2836 	if (fc_conf->autoneg != hw->mac.autoneg)
2837 		return -ENOTSUP;
2838 	rx_buf_size = igb_get_rx_buffer_size(hw);
2839 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
2840 
2841 	/* At least reserve one Ethernet frame for watermark */
2842 	max_high_water = rx_buf_size - ETHER_MAX_LEN;
2843 	if ((fc_conf->high_water > max_high_water) ||
2844 	    (fc_conf->high_water < fc_conf->low_water)) {
2845 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
2846 		PMD_INIT_LOG(ERR, "high water must <=  0x%x", max_high_water);
2847 		return -EINVAL;
2848 	}
2849 
2850 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
2851 	hw->fc.pause_time     = fc_conf->pause_time;
2852 	hw->fc.high_water     = fc_conf->high_water;
2853 	hw->fc.low_water      = fc_conf->low_water;
2854 	hw->fc.send_xon	      = fc_conf->send_xon;
2855 
2856 	err = e1000_setup_link_generic(hw);
2857 	if (err == E1000_SUCCESS) {
2858 
2859 		/* check if we want to forward MAC frames - driver doesn't have native
2860 		 * capability to do that, so we'll write the registers ourselves */
2861 
2862 		rctl = E1000_READ_REG(hw, E1000_RCTL);
2863 
2864 		/* set or clear MFLCN.PMCF bit depending on configuration */
2865 		if (fc_conf->mac_ctrl_frame_fwd != 0)
2866 			rctl |= E1000_RCTL_PMCF;
2867 		else
2868 			rctl &= ~E1000_RCTL_PMCF;
2869 
2870 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2871 		E1000_WRITE_FLUSH(hw);
2872 
2873 		return 0;
2874 	}
2875 
2876 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
2877 	return -EIO;
2878 }
2879 
2880 #define E1000_RAH_POOLSEL_SHIFT      (18)
2881 static void
2882 eth_igb_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
2883 	        uint32_t index, __rte_unused uint32_t pool)
2884 {
2885 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2886 	uint32_t rah;
2887 
2888 	e1000_rar_set(hw, mac_addr->addr_bytes, index);
2889 	rah = E1000_READ_REG(hw, E1000_RAH(index));
2890 	rah |= (0x1 << (E1000_RAH_POOLSEL_SHIFT + pool));
2891 	E1000_WRITE_REG(hw, E1000_RAH(index), rah);
2892 }
2893 
2894 static void
2895 eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index)
2896 {
2897 	uint8_t addr[ETHER_ADDR_LEN];
2898 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2899 
2900 	memset(addr, 0, sizeof(addr));
2901 
2902 	e1000_rar_set(hw, addr, index);
2903 }
2904 
2905 static void
2906 eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
2907 				struct ether_addr *addr)
2908 {
2909 	eth_igb_rar_clear(dev, 0);
2910 
2911 	eth_igb_rar_set(dev, (void *)addr, 0, 0);
2912 }
2913 /*
2914  * Virtual Function operations
2915  */
2916 static void
2917 igbvf_intr_disable(struct e1000_hw *hw)
2918 {
2919 	PMD_INIT_FUNC_TRACE();
2920 
2921 	/* Clear interrupt mask to stop from interrupts being generated */
2922 	E1000_WRITE_REG(hw, E1000_EIMC, 0xFFFF);
2923 
2924 	E1000_WRITE_FLUSH(hw);
2925 }
2926 
2927 static void
2928 igbvf_stop_adapter(struct rte_eth_dev *dev)
2929 {
2930 	u32 reg_val;
2931 	u16 i;
2932 	struct rte_eth_dev_info dev_info;
2933 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2934 
2935 	memset(&dev_info, 0, sizeof(dev_info));
2936 	eth_igbvf_infos_get(dev, &dev_info);
2937 
2938 	/* Clear interrupt mask to stop from interrupts being generated */
2939 	igbvf_intr_disable(hw);
2940 
2941 	/* Clear any pending interrupts, flush previous writes */
2942 	E1000_READ_REG(hw, E1000_EICR);
2943 
2944 	/* Disable the transmit unit.  Each queue must be disabled. */
2945 	for (i = 0; i < dev_info.max_tx_queues; i++)
2946 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), E1000_TXDCTL_SWFLSH);
2947 
2948 	/* Disable the receive unit by stopping each queue */
2949 	for (i = 0; i < dev_info.max_rx_queues; i++) {
2950 		reg_val = E1000_READ_REG(hw, E1000_RXDCTL(i));
2951 		reg_val &= ~E1000_RXDCTL_QUEUE_ENABLE;
2952 		E1000_WRITE_REG(hw, E1000_RXDCTL(i), reg_val);
2953 		while (E1000_READ_REG(hw, E1000_RXDCTL(i)) & E1000_RXDCTL_QUEUE_ENABLE)
2954 			;
2955 	}
2956 
2957 	/* flush all queues disables */
2958 	E1000_WRITE_FLUSH(hw);
2959 	msec_delay(2);
2960 }
2961 
2962 static int eth_igbvf_link_update(struct e1000_hw *hw)
2963 {
2964 	struct e1000_mbx_info *mbx = &hw->mbx;
2965 	struct e1000_mac_info *mac = &hw->mac;
2966 	int ret_val = E1000_SUCCESS;
2967 
2968 	PMD_INIT_LOG(DEBUG, "e1000_check_for_link_vf");
2969 
2970 	/*
2971 	 * We only want to run this if there has been a rst asserted.
2972 	 * in this case that could mean a link change, device reset,
2973 	 * or a virtual function reset
2974 	 */
2975 
2976 	/* If we were hit with a reset or timeout drop the link */
2977 	if (!e1000_check_for_rst(hw, 0) || !mbx->timeout)
2978 		mac->get_link_status = TRUE;
2979 
2980 	if (!mac->get_link_status)
2981 		goto out;
2982 
2983 	/* if link status is down no point in checking to see if pf is up */
2984 	if (!(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU))
2985 		goto out;
2986 
2987 	/* if we passed all the tests above then the link is up and we no
2988 	 * longer need to check for link */
2989 	mac->get_link_status = FALSE;
2990 
2991 out:
2992 	return ret_val;
2993 }
2994 
2995 
2996 static int
2997 igbvf_dev_configure(struct rte_eth_dev *dev)
2998 {
2999 	struct rte_eth_conf* conf = &dev->data->dev_conf;
3000 
3001 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
3002 		     dev->data->port_id);
3003 
3004 	/*
3005 	 * VF has no ability to enable/disable HW CRC
3006 	 * Keep the persistent behavior the same as Host PF
3007 	 */
3008 #ifndef RTE_LIBRTE_E1000_PF_DISABLE_STRIP_CRC
3009 	if (!conf->rxmode.hw_strip_crc) {
3010 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
3011 		conf->rxmode.hw_strip_crc = 1;
3012 	}
3013 #else
3014 	if (conf->rxmode.hw_strip_crc) {
3015 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
3016 		conf->rxmode.hw_strip_crc = 0;
3017 	}
3018 #endif
3019 
3020 	return 0;
3021 }
3022 
3023 static int
3024 igbvf_dev_start(struct rte_eth_dev *dev)
3025 {
3026 	struct e1000_hw *hw =
3027 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3028 	struct e1000_adapter *adapter =
3029 		E1000_DEV_PRIVATE(dev->data->dev_private);
3030 	int ret;
3031 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
3032 	uint32_t intr_vector = 0;
3033 
3034 	PMD_INIT_FUNC_TRACE();
3035 
3036 	hw->mac.ops.reset_hw(hw);
3037 	adapter->stopped = 0;
3038 
3039 	/* Set all vfta */
3040 	igbvf_set_vfta_all(dev,1);
3041 
3042 	eth_igbvf_tx_init(dev);
3043 
3044 	/* This can fail when allocating mbufs for descriptor rings */
3045 	ret = eth_igbvf_rx_init(dev);
3046 	if (ret) {
3047 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
3048 		igb_dev_clear_queues(dev);
3049 		return ret;
3050 	}
3051 
3052 	/* check and configure queue intr-vector mapping */
3053 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
3054 		intr_vector = dev->data->nb_rx_queues;
3055 		ret = rte_intr_efd_enable(intr_handle, intr_vector);
3056 		if (ret)
3057 			return ret;
3058 	}
3059 
3060 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
3061 		intr_handle->intr_vec =
3062 			rte_zmalloc("intr_vec",
3063 				    dev->data->nb_rx_queues * sizeof(int), 0);
3064 		if (!intr_handle->intr_vec) {
3065 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
3066 				     " intr_vec\n", dev->data->nb_rx_queues);
3067 			return -ENOMEM;
3068 		}
3069 	}
3070 
3071 	eth_igbvf_configure_msix_intr(dev);
3072 
3073 	/* enable uio/vfio intr/eventfd mapping */
3074 	rte_intr_enable(intr_handle);
3075 
3076 	/* resume enabled intr since hw reset */
3077 	igbvf_intr_enable(dev);
3078 
3079 	return 0;
3080 }
3081 
3082 static void
3083 igbvf_dev_stop(struct rte_eth_dev *dev)
3084 {
3085 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
3086 
3087 	PMD_INIT_FUNC_TRACE();
3088 
3089 	igbvf_stop_adapter(dev);
3090 
3091 	/*
3092 	  * Clear what we set, but we still keep shadow_vfta to
3093 	  * restore after device starts
3094 	  */
3095 	igbvf_set_vfta_all(dev,0);
3096 
3097 	igb_dev_clear_queues(dev);
3098 
3099 	/* disable intr eventfd mapping */
3100 	rte_intr_disable(intr_handle);
3101 
3102 	/* Clean datapath event and queue/vec mapping */
3103 	rte_intr_efd_disable(intr_handle);
3104 	if (intr_handle->intr_vec) {
3105 		rte_free(intr_handle->intr_vec);
3106 		intr_handle->intr_vec = NULL;
3107 	}
3108 }
3109 
3110 static void
3111 igbvf_dev_close(struct rte_eth_dev *dev)
3112 {
3113 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3114 	struct e1000_adapter *adapter =
3115 		E1000_DEV_PRIVATE(dev->data->dev_private);
3116 	struct ether_addr addr;
3117 
3118 	PMD_INIT_FUNC_TRACE();
3119 
3120 	e1000_reset_hw(hw);
3121 
3122 	igbvf_dev_stop(dev);
3123 	adapter->stopped = 1;
3124 	igb_dev_free_queues(dev);
3125 
3126 	/**
3127 	 * reprogram the RAR with a zero mac address,
3128 	 * to ensure that the VF traffic goes to the PF
3129 	 * after stop, close and detach of the VF.
3130 	 **/
3131 
3132 	memset(&addr, 0, sizeof(addr));
3133 	igbvf_default_mac_addr_set(dev, &addr);
3134 }
3135 
3136 static void
3137 igbvf_promiscuous_enable(struct rte_eth_dev *dev)
3138 {
3139 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3140 
3141 	/* Set both unicast and multicast promisc */
3142 	e1000_promisc_set_vf(hw, e1000_promisc_enabled);
3143 }
3144 
3145 static void
3146 igbvf_promiscuous_disable(struct rte_eth_dev *dev)
3147 {
3148 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3149 
3150 	/* If in allmulticast mode leave multicast promisc */
3151 	if (dev->data->all_multicast == 1)
3152 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3153 	else
3154 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3155 }
3156 
3157 static void
3158 igbvf_allmulticast_enable(struct rte_eth_dev *dev)
3159 {
3160 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3161 
3162 	/* In promiscuous mode multicast promisc already set */
3163 	if (dev->data->promiscuous == 0)
3164 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3165 }
3166 
3167 static void
3168 igbvf_allmulticast_disable(struct rte_eth_dev *dev)
3169 {
3170 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3171 
3172 	/* In promiscuous mode leave multicast promisc enabled */
3173 	if (dev->data->promiscuous == 0)
3174 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3175 }
3176 
3177 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on)
3178 {
3179 	struct e1000_mbx_info *mbx = &hw->mbx;
3180 	uint32_t msgbuf[2];
3181 	s32 err;
3182 
3183 	/* After set vlan, vlan strip will also be enabled in igb driver*/
3184 	msgbuf[0] = E1000_VF_SET_VLAN;
3185 	msgbuf[1] = vid;
3186 	/* Setting the 8 bit field MSG INFO to TRUE indicates "add" */
3187 	if (on)
3188 		msgbuf[0] |= E1000_VF_SET_VLAN_ADD;
3189 
3190 	err = mbx->ops.write_posted(hw, msgbuf, 2, 0);
3191 	if (err)
3192 		goto mbx_err;
3193 
3194 	err = mbx->ops.read_posted(hw, msgbuf, 2, 0);
3195 	if (err)
3196 		goto mbx_err;
3197 
3198 	msgbuf[0] &= ~E1000_VT_MSGTYPE_CTS;
3199 	if (msgbuf[0] == (E1000_VF_SET_VLAN | E1000_VT_MSGTYPE_NACK))
3200 		err = -EINVAL;
3201 
3202 mbx_err:
3203 	return err;
3204 }
3205 
3206 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on)
3207 {
3208 	struct e1000_hw *hw =
3209 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3210 	struct e1000_vfta * shadow_vfta =
3211 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3212 	int i = 0, j = 0, vfta = 0, mask = 1;
3213 
3214 	for (i = 0; i < IGB_VFTA_SIZE; i++){
3215 		vfta = shadow_vfta->vfta[i];
3216 		if(vfta){
3217 			mask = 1;
3218 			for (j = 0; j < 32; j++){
3219 				if(vfta & mask)
3220 					igbvf_set_vfta(hw,
3221 						(uint16_t)((i<<5)+j), on);
3222 				mask<<=1;
3223 			}
3224 		}
3225 	}
3226 
3227 }
3228 
3229 static int
3230 igbvf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
3231 {
3232 	struct e1000_hw *hw =
3233 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3234 	struct e1000_vfta * shadow_vfta =
3235 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3236 	uint32_t vid_idx = 0;
3237 	uint32_t vid_bit = 0;
3238 	int ret = 0;
3239 
3240 	PMD_INIT_FUNC_TRACE();
3241 
3242 	/*vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf*/
3243 	ret = igbvf_set_vfta(hw, vlan_id, !!on);
3244 	if(ret){
3245 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
3246 		return ret;
3247 	}
3248 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
3249 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
3250 
3251 	/*Save what we set and retore it after device reset*/
3252 	if (on)
3253 		shadow_vfta->vfta[vid_idx] |= vid_bit;
3254 	else
3255 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
3256 
3257 	return 0;
3258 }
3259 
3260 static void
3261 igbvf_default_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *addr)
3262 {
3263 	struct e1000_hw *hw =
3264 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3265 
3266 	/* index is not used by rar_set() */
3267 	hw->mac.ops.rar_set(hw, (void *)addr, 0);
3268 }
3269 
3270 
3271 static int
3272 eth_igb_rss_reta_update(struct rte_eth_dev *dev,
3273 			struct rte_eth_rss_reta_entry64 *reta_conf,
3274 			uint16_t reta_size)
3275 {
3276 	uint8_t i, j, mask;
3277 	uint32_t reta, r;
3278 	uint16_t idx, shift;
3279 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3280 
3281 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3282 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3283 			"(%d) doesn't match the number hardware can supported "
3284 			"(%d)\n", reta_size, ETH_RSS_RETA_SIZE_128);
3285 		return -EINVAL;
3286 	}
3287 
3288 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3289 		idx = i / RTE_RETA_GROUP_SIZE;
3290 		shift = i % RTE_RETA_GROUP_SIZE;
3291 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3292 						IGB_4_BIT_MASK);
3293 		if (!mask)
3294 			continue;
3295 		if (mask == IGB_4_BIT_MASK)
3296 			r = 0;
3297 		else
3298 			r = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3299 		for (j = 0, reta = 0; j < IGB_4_BIT_WIDTH; j++) {
3300 			if (mask & (0x1 << j))
3301 				reta |= reta_conf[idx].reta[shift + j] <<
3302 							(CHAR_BIT * j);
3303 			else
3304 				reta |= r & (IGB_8_BIT_MASK << (CHAR_BIT * j));
3305 		}
3306 		E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
3307 	}
3308 
3309 	return 0;
3310 }
3311 
3312 static int
3313 eth_igb_rss_reta_query(struct rte_eth_dev *dev,
3314 		       struct rte_eth_rss_reta_entry64 *reta_conf,
3315 		       uint16_t reta_size)
3316 {
3317 	uint8_t i, j, mask;
3318 	uint32_t reta;
3319 	uint16_t idx, shift;
3320 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3321 
3322 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3323 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3324 			"(%d) doesn't match the number hardware can supported "
3325 			"(%d)\n", reta_size, ETH_RSS_RETA_SIZE_128);
3326 		return -EINVAL;
3327 	}
3328 
3329 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3330 		idx = i / RTE_RETA_GROUP_SIZE;
3331 		shift = i % RTE_RETA_GROUP_SIZE;
3332 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3333 						IGB_4_BIT_MASK);
3334 		if (!mask)
3335 			continue;
3336 		reta = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3337 		for (j = 0; j < IGB_4_BIT_WIDTH; j++) {
3338 			if (mask & (0x1 << j))
3339 				reta_conf[idx].reta[shift + j] =
3340 					((reta >> (CHAR_BIT * j)) &
3341 						IGB_8_BIT_MASK);
3342 		}
3343 	}
3344 
3345 	return 0;
3346 }
3347 
3348 #define MAC_TYPE_FILTER_SUP(type)    do {\
3349 	if ((type) != e1000_82580 && (type) != e1000_i350 &&\
3350 		(type) != e1000_82576)\
3351 		return -ENOTSUP;\
3352 } while (0)
3353 
3354 static int
3355 eth_igb_syn_filter_set(struct rte_eth_dev *dev,
3356 			struct rte_eth_syn_filter *filter,
3357 			bool add)
3358 {
3359 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3360 	uint32_t synqf, rfctl;
3361 
3362 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3363 		return -EINVAL;
3364 
3365 	synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3366 
3367 	if (add) {
3368 		if (synqf & E1000_SYN_FILTER_ENABLE)
3369 			return -EINVAL;
3370 
3371 		synqf = (uint32_t)(((filter->queue << E1000_SYN_FILTER_QUEUE_SHIFT) &
3372 			E1000_SYN_FILTER_QUEUE) | E1000_SYN_FILTER_ENABLE);
3373 
3374 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3375 		if (filter->hig_pri)
3376 			rfctl |= E1000_RFCTL_SYNQFP;
3377 		else
3378 			rfctl &= ~E1000_RFCTL_SYNQFP;
3379 
3380 		E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3381 	} else {
3382 		if (!(synqf & E1000_SYN_FILTER_ENABLE))
3383 			return -ENOENT;
3384 		synqf = 0;
3385 	}
3386 
3387 	E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
3388 	E1000_WRITE_FLUSH(hw);
3389 	return 0;
3390 }
3391 
3392 static int
3393 eth_igb_syn_filter_get(struct rte_eth_dev *dev,
3394 			struct rte_eth_syn_filter *filter)
3395 {
3396 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3397 	uint32_t synqf, rfctl;
3398 
3399 	synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3400 	if (synqf & E1000_SYN_FILTER_ENABLE) {
3401 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3402 		filter->hig_pri = (rfctl & E1000_RFCTL_SYNQFP) ? 1 : 0;
3403 		filter->queue = (uint8_t)((synqf & E1000_SYN_FILTER_QUEUE) >>
3404 				E1000_SYN_FILTER_QUEUE_SHIFT);
3405 		return 0;
3406 	}
3407 
3408 	return -ENOENT;
3409 }
3410 
3411 static int
3412 eth_igb_syn_filter_handle(struct rte_eth_dev *dev,
3413 			enum rte_filter_op filter_op,
3414 			void *arg)
3415 {
3416 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3417 	int ret;
3418 
3419 	MAC_TYPE_FILTER_SUP(hw->mac.type);
3420 
3421 	if (filter_op == RTE_ETH_FILTER_NOP)
3422 		return 0;
3423 
3424 	if (arg == NULL) {
3425 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
3426 			    filter_op);
3427 		return -EINVAL;
3428 	}
3429 
3430 	switch (filter_op) {
3431 	case RTE_ETH_FILTER_ADD:
3432 		ret = eth_igb_syn_filter_set(dev,
3433 				(struct rte_eth_syn_filter *)arg,
3434 				TRUE);
3435 		break;
3436 	case RTE_ETH_FILTER_DELETE:
3437 		ret = eth_igb_syn_filter_set(dev,
3438 				(struct rte_eth_syn_filter *)arg,
3439 				FALSE);
3440 		break;
3441 	case RTE_ETH_FILTER_GET:
3442 		ret = eth_igb_syn_filter_get(dev,
3443 				(struct rte_eth_syn_filter *)arg);
3444 		break;
3445 	default:
3446 		PMD_DRV_LOG(ERR, "unsupported operation %u\n", filter_op);
3447 		ret = -EINVAL;
3448 		break;
3449 	}
3450 
3451 	return ret;
3452 }
3453 
3454 #define MAC_TYPE_FILTER_SUP_EXT(type)    do {\
3455 	if ((type) != e1000_82580 && (type) != e1000_i350)\
3456 		return -ENOSYS; \
3457 } while (0)
3458 
3459 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_2tuple_filter_info*/
3460 static inline int
3461 ntuple_filter_to_2tuple(struct rte_eth_ntuple_filter *filter,
3462 			struct e1000_2tuple_filter_info *filter_info)
3463 {
3464 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3465 		return -EINVAL;
3466 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
3467 		return -EINVAL;  /* filter index is out of range. */
3468 	if (filter->tcp_flags > TCP_FLAG_ALL)
3469 		return -EINVAL;  /* flags is invalid. */
3470 
3471 	switch (filter->dst_port_mask) {
3472 	case UINT16_MAX:
3473 		filter_info->dst_port_mask = 0;
3474 		filter_info->dst_port = filter->dst_port;
3475 		break;
3476 	case 0:
3477 		filter_info->dst_port_mask = 1;
3478 		break;
3479 	default:
3480 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
3481 		return -EINVAL;
3482 	}
3483 
3484 	switch (filter->proto_mask) {
3485 	case UINT8_MAX:
3486 		filter_info->proto_mask = 0;
3487 		filter_info->proto = filter->proto;
3488 		break;
3489 	case 0:
3490 		filter_info->proto_mask = 1;
3491 		break;
3492 	default:
3493 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
3494 		return -EINVAL;
3495 	}
3496 
3497 	filter_info->priority = (uint8_t)filter->priority;
3498 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
3499 		filter_info->tcp_flags = filter->tcp_flags;
3500 	else
3501 		filter_info->tcp_flags = 0;
3502 
3503 	return 0;
3504 }
3505 
3506 static inline struct e1000_2tuple_filter *
3507 igb_2tuple_filter_lookup(struct e1000_2tuple_filter_list *filter_list,
3508 			struct e1000_2tuple_filter_info *key)
3509 {
3510 	struct e1000_2tuple_filter *it;
3511 
3512 	TAILQ_FOREACH(it, filter_list, entries) {
3513 		if (memcmp(key, &it->filter_info,
3514 			sizeof(struct e1000_2tuple_filter_info)) == 0) {
3515 			return it;
3516 		}
3517 	}
3518 	return NULL;
3519 }
3520 
3521 /*
3522  * igb_add_2tuple_filter - add a 2tuple filter
3523  *
3524  * @param
3525  * dev: Pointer to struct rte_eth_dev.
3526  * ntuple_filter: ponter to the filter that will be added.
3527  *
3528  * @return
3529  *    - On success, zero.
3530  *    - On failure, a negative value.
3531  */
3532 static int
3533 igb_add_2tuple_filter(struct rte_eth_dev *dev,
3534 			struct rte_eth_ntuple_filter *ntuple_filter)
3535 {
3536 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3537 	struct e1000_filter_info *filter_info =
3538 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3539 	struct e1000_2tuple_filter *filter;
3540 	uint32_t ttqf = E1000_TTQF_DISABLE_MASK;
3541 	uint32_t imir, imir_ext = E1000_IMIREXT_SIZE_BP;
3542 	int i, ret;
3543 
3544 	filter = rte_zmalloc("e1000_2tuple_filter",
3545 			sizeof(struct e1000_2tuple_filter), 0);
3546 	if (filter == NULL)
3547 		return -ENOMEM;
3548 
3549 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3550 				      &filter->filter_info);
3551 	if (ret < 0) {
3552 		rte_free(filter);
3553 		return ret;
3554 	}
3555 	if (igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3556 					 &filter->filter_info) != NULL) {
3557 		PMD_DRV_LOG(ERR, "filter exists.");
3558 		rte_free(filter);
3559 		return -EEXIST;
3560 	}
3561 	filter->queue = ntuple_filter->queue;
3562 
3563 	/*
3564 	 * look for an unused 2tuple filter index,
3565 	 * and insert the filter to list.
3566 	 */
3567 	for (i = 0; i < E1000_MAX_TTQF_FILTERS; i++) {
3568 		if (!(filter_info->twotuple_mask & (1 << i))) {
3569 			filter_info->twotuple_mask |= 1 << i;
3570 			filter->index = i;
3571 			TAILQ_INSERT_TAIL(&filter_info->twotuple_list,
3572 					  filter,
3573 					  entries);
3574 			break;
3575 		}
3576 	}
3577 	if (i >= E1000_MAX_TTQF_FILTERS) {
3578 		PMD_DRV_LOG(ERR, "2tuple filters are full.");
3579 		rte_free(filter);
3580 		return -ENOSYS;
3581 	}
3582 
3583 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
3584 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
3585 		imir |= E1000_IMIR_PORT_BP;
3586 	else
3587 		imir &= ~E1000_IMIR_PORT_BP;
3588 
3589 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
3590 
3591 	ttqf |= E1000_TTQF_QUEUE_ENABLE;
3592 	ttqf |= (uint32_t)(filter->queue << E1000_TTQF_QUEUE_SHIFT);
3593 	ttqf |= (uint32_t)(filter->filter_info.proto & E1000_TTQF_PROTOCOL_MASK);
3594 	if (filter->filter_info.proto_mask == 0)
3595 		ttqf &= ~E1000_TTQF_MASK_ENABLE;
3596 
3597 	/* tcp flags bits setting. */
3598 	if (filter->filter_info.tcp_flags & TCP_FLAG_ALL) {
3599 		if (filter->filter_info.tcp_flags & TCP_URG_FLAG)
3600 			imir_ext |= E1000_IMIREXT_CTRL_URG;
3601 		if (filter->filter_info.tcp_flags & TCP_ACK_FLAG)
3602 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
3603 		if (filter->filter_info.tcp_flags & TCP_PSH_FLAG)
3604 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
3605 		if (filter->filter_info.tcp_flags & TCP_RST_FLAG)
3606 			imir_ext |= E1000_IMIREXT_CTRL_RST;
3607 		if (filter->filter_info.tcp_flags & TCP_SYN_FLAG)
3608 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
3609 		if (filter->filter_info.tcp_flags & TCP_FIN_FLAG)
3610 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
3611 	} else
3612 		imir_ext |= E1000_IMIREXT_CTRL_BP;
3613 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
3614 	E1000_WRITE_REG(hw, E1000_TTQF(i), ttqf);
3615 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
3616 	return 0;
3617 }
3618 
3619 /*
3620  * igb_remove_2tuple_filter - remove a 2tuple filter
3621  *
3622  * @param
3623  * dev: Pointer to struct rte_eth_dev.
3624  * ntuple_filter: ponter to the filter that will be removed.
3625  *
3626  * @return
3627  *    - On success, zero.
3628  *    - On failure, a negative value.
3629  */
3630 static int
3631 igb_remove_2tuple_filter(struct rte_eth_dev *dev,
3632 			struct rte_eth_ntuple_filter *ntuple_filter)
3633 {
3634 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3635 	struct e1000_filter_info *filter_info =
3636 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3637 	struct e1000_2tuple_filter_info filter_2tuple;
3638 	struct e1000_2tuple_filter *filter;
3639 	int ret;
3640 
3641 	memset(&filter_2tuple, 0, sizeof(struct e1000_2tuple_filter_info));
3642 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3643 				      &filter_2tuple);
3644 	if (ret < 0)
3645 		return ret;
3646 
3647 	filter = igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3648 					 &filter_2tuple);
3649 	if (filter == NULL) {
3650 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
3651 		return -ENOENT;
3652 	}
3653 
3654 	filter_info->twotuple_mask &= ~(1 << filter->index);
3655 	TAILQ_REMOVE(&filter_info->twotuple_list, filter, entries);
3656 	rte_free(filter);
3657 
3658 	E1000_WRITE_REG(hw, E1000_TTQF(filter->index), E1000_TTQF_DISABLE_MASK);
3659 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
3660 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
3661 	return 0;
3662 }
3663 
3664 static inline struct e1000_flex_filter *
3665 eth_igb_flex_filter_lookup(struct e1000_flex_filter_list *filter_list,
3666 			struct e1000_flex_filter_info *key)
3667 {
3668 	struct e1000_flex_filter *it;
3669 
3670 	TAILQ_FOREACH(it, filter_list, entries) {
3671 		if (memcmp(key, &it->filter_info,
3672 			sizeof(struct e1000_flex_filter_info)) == 0)
3673 			return it;
3674 	}
3675 
3676 	return NULL;
3677 }
3678 
3679 static int
3680 eth_igb_add_del_flex_filter(struct rte_eth_dev *dev,
3681 			struct rte_eth_flex_filter *filter,
3682 			bool add)
3683 {
3684 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3685 	struct e1000_filter_info *filter_info =
3686 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3687 	struct e1000_flex_filter *flex_filter, *it;
3688 	uint32_t wufc, queueing, mask;
3689 	uint32_t reg_off;
3690 	uint8_t shift, i, j = 0;
3691 
3692 	flex_filter = rte_zmalloc("e1000_flex_filter",
3693 			sizeof(struct e1000_flex_filter), 0);
3694 	if (flex_filter == NULL)
3695 		return -ENOMEM;
3696 
3697 	flex_filter->filter_info.len = filter->len;
3698 	flex_filter->filter_info.priority = filter->priority;
3699 	memcpy(flex_filter->filter_info.dwords, filter->bytes, filter->len);
3700 	for (i = 0; i < RTE_ALIGN(filter->len, CHAR_BIT) / CHAR_BIT; i++) {
3701 		mask = 0;
3702 		/* reverse bits in flex filter's mask*/
3703 		for (shift = 0; shift < CHAR_BIT; shift++) {
3704 			if (filter->mask[i] & (0x01 << shift))
3705 				mask |= (0x80 >> shift);
3706 		}
3707 		flex_filter->filter_info.mask[i] = mask;
3708 	}
3709 
3710 	wufc = E1000_READ_REG(hw, E1000_WUFC);
3711 	if (flex_filter->index < E1000_MAX_FHFT)
3712 		reg_off = E1000_FHFT(flex_filter->index);
3713 	else
3714 		reg_off = E1000_FHFT_EXT(flex_filter->index - E1000_MAX_FHFT);
3715 
3716 	if (add) {
3717 		if (eth_igb_flex_filter_lookup(&filter_info->flex_list,
3718 				&flex_filter->filter_info) != NULL) {
3719 			PMD_DRV_LOG(ERR, "filter exists.");
3720 			rte_free(flex_filter);
3721 			return -EEXIST;
3722 		}
3723 		flex_filter->queue = filter->queue;
3724 		/*
3725 		 * look for an unused flex filter index
3726 		 * and insert the filter into the list.
3727 		 */
3728 		for (i = 0; i < E1000_MAX_FLEX_FILTERS; i++) {
3729 			if (!(filter_info->flex_mask & (1 << i))) {
3730 				filter_info->flex_mask |= 1 << i;
3731 				flex_filter->index = i;
3732 				TAILQ_INSERT_TAIL(&filter_info->flex_list,
3733 					flex_filter,
3734 					entries);
3735 				break;
3736 			}
3737 		}
3738 		if (i >= E1000_MAX_FLEX_FILTERS) {
3739 			PMD_DRV_LOG(ERR, "flex filters are full.");
3740 			rte_free(flex_filter);
3741 			return -ENOSYS;
3742 		}
3743 
3744 		E1000_WRITE_REG(hw, E1000_WUFC, wufc | E1000_WUFC_FLEX_HQ |
3745 				(E1000_WUFC_FLX0 << flex_filter->index));
3746 		queueing = filter->len |
3747 			(filter->queue << E1000_FHFT_QUEUEING_QUEUE_SHIFT) |
3748 			(filter->priority << E1000_FHFT_QUEUEING_PRIO_SHIFT);
3749 		E1000_WRITE_REG(hw, reg_off + E1000_FHFT_QUEUEING_OFFSET,
3750 				queueing);
3751 		for (i = 0; i < E1000_FLEX_FILTERS_MASK_SIZE; i++) {
3752 			E1000_WRITE_REG(hw, reg_off,
3753 					flex_filter->filter_info.dwords[j]);
3754 			reg_off += sizeof(uint32_t);
3755 			E1000_WRITE_REG(hw, reg_off,
3756 					flex_filter->filter_info.dwords[++j]);
3757 			reg_off += sizeof(uint32_t);
3758 			E1000_WRITE_REG(hw, reg_off,
3759 				(uint32_t)flex_filter->filter_info.mask[i]);
3760 			reg_off += sizeof(uint32_t) * 2;
3761 			++j;
3762 		}
3763 	} else {
3764 		it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
3765 				&flex_filter->filter_info);
3766 		if (it == NULL) {
3767 			PMD_DRV_LOG(ERR, "filter doesn't exist.");
3768 			rte_free(flex_filter);
3769 			return -ENOENT;
3770 		}
3771 
3772 		for (i = 0; i < E1000_FHFT_SIZE_IN_DWD; i++)
3773 			E1000_WRITE_REG(hw, reg_off + i * sizeof(uint32_t), 0);
3774 		E1000_WRITE_REG(hw, E1000_WUFC, wufc &
3775 			(~(E1000_WUFC_FLX0 << it->index)));
3776 
3777 		filter_info->flex_mask &= ~(1 << it->index);
3778 		TAILQ_REMOVE(&filter_info->flex_list, it, entries);
3779 		rte_free(it);
3780 		rte_free(flex_filter);
3781 	}
3782 
3783 	return 0;
3784 }
3785 
3786 static int
3787 eth_igb_get_flex_filter(struct rte_eth_dev *dev,
3788 			struct rte_eth_flex_filter *filter)
3789 {
3790 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3791 	struct e1000_filter_info *filter_info =
3792 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3793 	struct e1000_flex_filter flex_filter, *it;
3794 	uint32_t wufc, queueing, wufc_en = 0;
3795 
3796 	memset(&flex_filter, 0, sizeof(struct e1000_flex_filter));
3797 	flex_filter.filter_info.len = filter->len;
3798 	flex_filter.filter_info.priority = filter->priority;
3799 	memcpy(flex_filter.filter_info.dwords, filter->bytes, filter->len);
3800 	memcpy(flex_filter.filter_info.mask, filter->mask,
3801 			RTE_ALIGN(filter->len, sizeof(char)) / sizeof(char));
3802 
3803 	it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
3804 				&flex_filter.filter_info);
3805 	if (it == NULL) {
3806 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
3807 		return -ENOENT;
3808 	}
3809 
3810 	wufc = E1000_READ_REG(hw, E1000_WUFC);
3811 	wufc_en = E1000_WUFC_FLEX_HQ | (E1000_WUFC_FLX0 << it->index);
3812 
3813 	if ((wufc & wufc_en) == wufc_en) {
3814 		uint32_t reg_off = 0;
3815 		if (it->index < E1000_MAX_FHFT)
3816 			reg_off = E1000_FHFT(it->index);
3817 		else
3818 			reg_off = E1000_FHFT_EXT(it->index - E1000_MAX_FHFT);
3819 
3820 		queueing = E1000_READ_REG(hw,
3821 				reg_off + E1000_FHFT_QUEUEING_OFFSET);
3822 		filter->len = queueing & E1000_FHFT_QUEUEING_LEN;
3823 		filter->priority = (queueing & E1000_FHFT_QUEUEING_PRIO) >>
3824 			E1000_FHFT_QUEUEING_PRIO_SHIFT;
3825 		filter->queue = (queueing & E1000_FHFT_QUEUEING_QUEUE) >>
3826 			E1000_FHFT_QUEUEING_QUEUE_SHIFT;
3827 		return 0;
3828 	}
3829 	return -ENOENT;
3830 }
3831 
3832 static int
3833 eth_igb_flex_filter_handle(struct rte_eth_dev *dev,
3834 			enum rte_filter_op filter_op,
3835 			void *arg)
3836 {
3837 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3838 	struct rte_eth_flex_filter *filter;
3839 	int ret = 0;
3840 
3841 	MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
3842 
3843 	if (filter_op == RTE_ETH_FILTER_NOP)
3844 		return ret;
3845 
3846 	if (arg == NULL) {
3847 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
3848 			    filter_op);
3849 		return -EINVAL;
3850 	}
3851 
3852 	filter = (struct rte_eth_flex_filter *)arg;
3853 	if (filter->len == 0 || filter->len > E1000_MAX_FLEX_FILTER_LEN
3854 	    || filter->len % sizeof(uint64_t) != 0) {
3855 		PMD_DRV_LOG(ERR, "filter's length is out of range");
3856 		return -EINVAL;
3857 	}
3858 	if (filter->priority > E1000_MAX_FLEX_FILTER_PRI) {
3859 		PMD_DRV_LOG(ERR, "filter's priority is out of range");
3860 		return -EINVAL;
3861 	}
3862 
3863 	switch (filter_op) {
3864 	case RTE_ETH_FILTER_ADD:
3865 		ret = eth_igb_add_del_flex_filter(dev, filter, TRUE);
3866 		break;
3867 	case RTE_ETH_FILTER_DELETE:
3868 		ret = eth_igb_add_del_flex_filter(dev, filter, FALSE);
3869 		break;
3870 	case RTE_ETH_FILTER_GET:
3871 		ret = eth_igb_get_flex_filter(dev, filter);
3872 		break;
3873 	default:
3874 		PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
3875 		ret = -EINVAL;
3876 		break;
3877 	}
3878 
3879 	return ret;
3880 }
3881 
3882 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_5tuple_filter_info*/
3883 static inline int
3884 ntuple_filter_to_5tuple_82576(struct rte_eth_ntuple_filter *filter,
3885 			struct e1000_5tuple_filter_info *filter_info)
3886 {
3887 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM_82576)
3888 		return -EINVAL;
3889 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
3890 		return -EINVAL;  /* filter index is out of range. */
3891 	if (filter->tcp_flags > TCP_FLAG_ALL)
3892 		return -EINVAL;  /* flags is invalid. */
3893 
3894 	switch (filter->dst_ip_mask) {
3895 	case UINT32_MAX:
3896 		filter_info->dst_ip_mask = 0;
3897 		filter_info->dst_ip = filter->dst_ip;
3898 		break;
3899 	case 0:
3900 		filter_info->dst_ip_mask = 1;
3901 		break;
3902 	default:
3903 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
3904 		return -EINVAL;
3905 	}
3906 
3907 	switch (filter->src_ip_mask) {
3908 	case UINT32_MAX:
3909 		filter_info->src_ip_mask = 0;
3910 		filter_info->src_ip = filter->src_ip;
3911 		break;
3912 	case 0:
3913 		filter_info->src_ip_mask = 1;
3914 		break;
3915 	default:
3916 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
3917 		return -EINVAL;
3918 	}
3919 
3920 	switch (filter->dst_port_mask) {
3921 	case UINT16_MAX:
3922 		filter_info->dst_port_mask = 0;
3923 		filter_info->dst_port = filter->dst_port;
3924 		break;
3925 	case 0:
3926 		filter_info->dst_port_mask = 1;
3927 		break;
3928 	default:
3929 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
3930 		return -EINVAL;
3931 	}
3932 
3933 	switch (filter->src_port_mask) {
3934 	case UINT16_MAX:
3935 		filter_info->src_port_mask = 0;
3936 		filter_info->src_port = filter->src_port;
3937 		break;
3938 	case 0:
3939 		filter_info->src_port_mask = 1;
3940 		break;
3941 	default:
3942 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
3943 		return -EINVAL;
3944 	}
3945 
3946 	switch (filter->proto_mask) {
3947 	case UINT8_MAX:
3948 		filter_info->proto_mask = 0;
3949 		filter_info->proto = filter->proto;
3950 		break;
3951 	case 0:
3952 		filter_info->proto_mask = 1;
3953 		break;
3954 	default:
3955 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
3956 		return -EINVAL;
3957 	}
3958 
3959 	filter_info->priority = (uint8_t)filter->priority;
3960 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
3961 		filter_info->tcp_flags = filter->tcp_flags;
3962 	else
3963 		filter_info->tcp_flags = 0;
3964 
3965 	return 0;
3966 }
3967 
3968 static inline struct e1000_5tuple_filter *
3969 igb_5tuple_filter_lookup_82576(struct e1000_5tuple_filter_list *filter_list,
3970 			struct e1000_5tuple_filter_info *key)
3971 {
3972 	struct e1000_5tuple_filter *it;
3973 
3974 	TAILQ_FOREACH(it, filter_list, entries) {
3975 		if (memcmp(key, &it->filter_info,
3976 			sizeof(struct e1000_5tuple_filter_info)) == 0) {
3977 			return it;
3978 		}
3979 	}
3980 	return NULL;
3981 }
3982 
3983 /*
3984  * igb_add_5tuple_filter_82576 - add a 5tuple filter
3985  *
3986  * @param
3987  * dev: Pointer to struct rte_eth_dev.
3988  * ntuple_filter: ponter to the filter that will be added.
3989  *
3990  * @return
3991  *    - On success, zero.
3992  *    - On failure, a negative value.
3993  */
3994 static int
3995 igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
3996 			struct rte_eth_ntuple_filter *ntuple_filter)
3997 {
3998 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3999 	struct e1000_filter_info *filter_info =
4000 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4001 	struct e1000_5tuple_filter *filter;
4002 	uint32_t ftqf = E1000_FTQF_VF_BP | E1000_FTQF_MASK;
4003 	uint32_t spqf, imir, imir_ext = E1000_IMIREXT_SIZE_BP;
4004 	uint8_t i;
4005 	int ret;
4006 
4007 	filter = rte_zmalloc("e1000_5tuple_filter",
4008 			sizeof(struct e1000_5tuple_filter), 0);
4009 	if (filter == NULL)
4010 		return -ENOMEM;
4011 
4012 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4013 					    &filter->filter_info);
4014 	if (ret < 0) {
4015 		rte_free(filter);
4016 		return ret;
4017 	}
4018 
4019 	if (igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4020 					 &filter->filter_info) != NULL) {
4021 		PMD_DRV_LOG(ERR, "filter exists.");
4022 		rte_free(filter);
4023 		return -EEXIST;
4024 	}
4025 	filter->queue = ntuple_filter->queue;
4026 
4027 	/*
4028 	 * look for an unused 5tuple filter index,
4029 	 * and insert the filter to list.
4030 	 */
4031 	for (i = 0; i < E1000_MAX_FTQF_FILTERS; i++) {
4032 		if (!(filter_info->fivetuple_mask & (1 << i))) {
4033 			filter_info->fivetuple_mask |= 1 << i;
4034 			filter->index = i;
4035 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
4036 					  filter,
4037 					  entries);
4038 			break;
4039 		}
4040 	}
4041 	if (i >= E1000_MAX_FTQF_FILTERS) {
4042 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
4043 		rte_free(filter);
4044 		return -ENOSYS;
4045 	}
4046 
4047 	ftqf |= filter->filter_info.proto & E1000_FTQF_PROTOCOL_MASK;
4048 	if (filter->filter_info.src_ip_mask == 0) /* 0b means compare. */
4049 		ftqf &= ~E1000_FTQF_MASK_SOURCE_ADDR_BP;
4050 	if (filter->filter_info.dst_ip_mask == 0)
4051 		ftqf &= ~E1000_FTQF_MASK_DEST_ADDR_BP;
4052 	if (filter->filter_info.src_port_mask == 0)
4053 		ftqf &= ~E1000_FTQF_MASK_SOURCE_PORT_BP;
4054 	if (filter->filter_info.proto_mask == 0)
4055 		ftqf &= ~E1000_FTQF_MASK_PROTO_BP;
4056 	ftqf |= (filter->queue << E1000_FTQF_QUEUE_SHIFT) &
4057 		E1000_FTQF_QUEUE_MASK;
4058 	ftqf |= E1000_FTQF_QUEUE_ENABLE;
4059 	E1000_WRITE_REG(hw, E1000_FTQF(i), ftqf);
4060 	E1000_WRITE_REG(hw, E1000_DAQF(i), filter->filter_info.dst_ip);
4061 	E1000_WRITE_REG(hw, E1000_SAQF(i), filter->filter_info.src_ip);
4062 
4063 	spqf = filter->filter_info.src_port & E1000_SPQF_SRCPORT;
4064 	E1000_WRITE_REG(hw, E1000_SPQF(i), spqf);
4065 
4066 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
4067 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
4068 		imir |= E1000_IMIR_PORT_BP;
4069 	else
4070 		imir &= ~E1000_IMIR_PORT_BP;
4071 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
4072 
4073 	/* tcp flags bits setting. */
4074 	if (filter->filter_info.tcp_flags & TCP_FLAG_ALL) {
4075 		if (filter->filter_info.tcp_flags & TCP_URG_FLAG)
4076 			imir_ext |= E1000_IMIREXT_CTRL_URG;
4077 		if (filter->filter_info.tcp_flags & TCP_ACK_FLAG)
4078 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
4079 		if (filter->filter_info.tcp_flags & TCP_PSH_FLAG)
4080 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
4081 		if (filter->filter_info.tcp_flags & TCP_RST_FLAG)
4082 			imir_ext |= E1000_IMIREXT_CTRL_RST;
4083 		if (filter->filter_info.tcp_flags & TCP_SYN_FLAG)
4084 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
4085 		if (filter->filter_info.tcp_flags & TCP_FIN_FLAG)
4086 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
4087 	} else
4088 		imir_ext |= E1000_IMIREXT_CTRL_BP;
4089 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
4090 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
4091 	return 0;
4092 }
4093 
4094 /*
4095  * igb_remove_5tuple_filter_82576 - remove a 5tuple filter
4096  *
4097  * @param
4098  * dev: Pointer to struct rte_eth_dev.
4099  * ntuple_filter: ponter to the filter that will be removed.
4100  *
4101  * @return
4102  *    - On success, zero.
4103  *    - On failure, a negative value.
4104  */
4105 static int
4106 igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
4107 				struct rte_eth_ntuple_filter *ntuple_filter)
4108 {
4109 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4110 	struct e1000_filter_info *filter_info =
4111 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4112 	struct e1000_5tuple_filter_info filter_5tuple;
4113 	struct e1000_5tuple_filter *filter;
4114 	int ret;
4115 
4116 	memset(&filter_5tuple, 0, sizeof(struct e1000_5tuple_filter_info));
4117 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4118 					    &filter_5tuple);
4119 	if (ret < 0)
4120 		return ret;
4121 
4122 	filter = igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4123 					 &filter_5tuple);
4124 	if (filter == NULL) {
4125 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4126 		return -ENOENT;
4127 	}
4128 
4129 	filter_info->fivetuple_mask &= ~(1 << filter->index);
4130 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
4131 	rte_free(filter);
4132 
4133 	E1000_WRITE_REG(hw, E1000_FTQF(filter->index),
4134 			E1000_FTQF_VF_BP | E1000_FTQF_MASK);
4135 	E1000_WRITE_REG(hw, E1000_DAQF(filter->index), 0);
4136 	E1000_WRITE_REG(hw, E1000_SAQF(filter->index), 0);
4137 	E1000_WRITE_REG(hw, E1000_SPQF(filter->index), 0);
4138 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
4139 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
4140 	return 0;
4141 }
4142 
4143 static int
4144 eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4145 {
4146 	uint32_t rctl;
4147 	struct e1000_hw *hw;
4148 	struct rte_eth_dev_info dev_info;
4149 	uint32_t frame_size = mtu + (ETHER_HDR_LEN + ETHER_CRC_LEN +
4150 				     VLAN_TAG_SIZE);
4151 
4152 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4153 
4154 #ifdef RTE_LIBRTE_82571_SUPPORT
4155 	/* XXX: not bigger than max_rx_pktlen */
4156 	if (hw->mac.type == e1000_82571)
4157 		return -ENOTSUP;
4158 #endif
4159 	eth_igb_infos_get(dev, &dev_info);
4160 
4161 	/* check that mtu is within the allowed range */
4162 	if ((mtu < ETHER_MIN_MTU) ||
4163 	    (frame_size > dev_info.max_rx_pktlen))
4164 		return -EINVAL;
4165 
4166 	/* refuse mtu that requires the support of scattered packets when this
4167 	 * feature has not been enabled before. */
4168 	if (!dev->data->scattered_rx &&
4169 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
4170 		return -EINVAL;
4171 
4172 	rctl = E1000_READ_REG(hw, E1000_RCTL);
4173 
4174 	/* switch to jumbo mode if needed */
4175 	if (frame_size > ETHER_MAX_LEN) {
4176 		dev->data->dev_conf.rxmode.jumbo_frame = 1;
4177 		rctl |= E1000_RCTL_LPE;
4178 	} else {
4179 		dev->data->dev_conf.rxmode.jumbo_frame = 0;
4180 		rctl &= ~E1000_RCTL_LPE;
4181 	}
4182 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
4183 
4184 	/* update max frame size */
4185 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4186 
4187 	E1000_WRITE_REG(hw, E1000_RLPML,
4188 			dev->data->dev_conf.rxmode.max_rx_pkt_len);
4189 
4190 	return 0;
4191 }
4192 
4193 /*
4194  * igb_add_del_ntuple_filter - add or delete a ntuple filter
4195  *
4196  * @param
4197  * dev: Pointer to struct rte_eth_dev.
4198  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4199  * add: if true, add filter, if false, remove filter
4200  *
4201  * @return
4202  *    - On success, zero.
4203  *    - On failure, a negative value.
4204  */
4205 static int
4206 igb_add_del_ntuple_filter(struct rte_eth_dev *dev,
4207 			struct rte_eth_ntuple_filter *ntuple_filter,
4208 			bool add)
4209 {
4210 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4211 	int ret;
4212 
4213 	switch (ntuple_filter->flags) {
4214 	case RTE_5TUPLE_FLAGS:
4215 	case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4216 		if (hw->mac.type != e1000_82576)
4217 			return -ENOTSUP;
4218 		if (add)
4219 			ret = igb_add_5tuple_filter_82576(dev,
4220 							  ntuple_filter);
4221 		else
4222 			ret = igb_remove_5tuple_filter_82576(dev,
4223 							     ntuple_filter);
4224 		break;
4225 	case RTE_2TUPLE_FLAGS:
4226 	case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4227 		if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350)
4228 			return -ENOTSUP;
4229 		if (add)
4230 			ret = igb_add_2tuple_filter(dev, ntuple_filter);
4231 		else
4232 			ret = igb_remove_2tuple_filter(dev, ntuple_filter);
4233 		break;
4234 	default:
4235 		ret = -EINVAL;
4236 		break;
4237 	}
4238 
4239 	return ret;
4240 }
4241 
4242 /*
4243  * igb_get_ntuple_filter - get a ntuple filter
4244  *
4245  * @param
4246  * dev: Pointer to struct rte_eth_dev.
4247  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4248  *
4249  * @return
4250  *    - On success, zero.
4251  *    - On failure, a negative value.
4252  */
4253 static int
4254 igb_get_ntuple_filter(struct rte_eth_dev *dev,
4255 			struct rte_eth_ntuple_filter *ntuple_filter)
4256 {
4257 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4258 	struct e1000_filter_info *filter_info =
4259 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4260 	struct e1000_5tuple_filter_info filter_5tuple;
4261 	struct e1000_2tuple_filter_info filter_2tuple;
4262 	struct e1000_5tuple_filter *p_5tuple_filter;
4263 	struct e1000_2tuple_filter *p_2tuple_filter;
4264 	int ret;
4265 
4266 	switch (ntuple_filter->flags) {
4267 	case RTE_5TUPLE_FLAGS:
4268 	case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4269 		if (hw->mac.type != e1000_82576)
4270 			return -ENOTSUP;
4271 		memset(&filter_5tuple,
4272 			0,
4273 			sizeof(struct e1000_5tuple_filter_info));
4274 		ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4275 						    &filter_5tuple);
4276 		if (ret < 0)
4277 			return ret;
4278 		p_5tuple_filter = igb_5tuple_filter_lookup_82576(
4279 					&filter_info->fivetuple_list,
4280 					&filter_5tuple);
4281 		if (p_5tuple_filter == NULL) {
4282 			PMD_DRV_LOG(ERR, "filter doesn't exist.");
4283 			return -ENOENT;
4284 		}
4285 		ntuple_filter->queue = p_5tuple_filter->queue;
4286 		break;
4287 	case RTE_2TUPLE_FLAGS:
4288 	case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4289 		if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350)
4290 			return -ENOTSUP;
4291 		memset(&filter_2tuple,
4292 			0,
4293 			sizeof(struct e1000_2tuple_filter_info));
4294 		ret = ntuple_filter_to_2tuple(ntuple_filter, &filter_2tuple);
4295 		if (ret < 0)
4296 			return ret;
4297 		p_2tuple_filter = igb_2tuple_filter_lookup(
4298 					&filter_info->twotuple_list,
4299 					&filter_2tuple);
4300 		if (p_2tuple_filter == NULL) {
4301 			PMD_DRV_LOG(ERR, "filter doesn't exist.");
4302 			return -ENOENT;
4303 		}
4304 		ntuple_filter->queue = p_2tuple_filter->queue;
4305 		break;
4306 	default:
4307 		ret = -EINVAL;
4308 		break;
4309 	}
4310 
4311 	return 0;
4312 }
4313 
4314 /*
4315  * igb_ntuple_filter_handle - Handle operations for ntuple filter.
4316  * @dev: pointer to rte_eth_dev structure
4317  * @filter_op:operation will be taken.
4318  * @arg: a pointer to specific structure corresponding to the filter_op
4319  */
4320 static int
4321 igb_ntuple_filter_handle(struct rte_eth_dev *dev,
4322 				enum rte_filter_op filter_op,
4323 				void *arg)
4324 {
4325 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4326 	int ret;
4327 
4328 	MAC_TYPE_FILTER_SUP(hw->mac.type);
4329 
4330 	if (filter_op == RTE_ETH_FILTER_NOP)
4331 		return 0;
4332 
4333 	if (arg == NULL) {
4334 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4335 			    filter_op);
4336 		return -EINVAL;
4337 	}
4338 
4339 	switch (filter_op) {
4340 	case RTE_ETH_FILTER_ADD:
4341 		ret = igb_add_del_ntuple_filter(dev,
4342 			(struct rte_eth_ntuple_filter *)arg,
4343 			TRUE);
4344 		break;
4345 	case RTE_ETH_FILTER_DELETE:
4346 		ret = igb_add_del_ntuple_filter(dev,
4347 			(struct rte_eth_ntuple_filter *)arg,
4348 			FALSE);
4349 		break;
4350 	case RTE_ETH_FILTER_GET:
4351 		ret = igb_get_ntuple_filter(dev,
4352 			(struct rte_eth_ntuple_filter *)arg);
4353 		break;
4354 	default:
4355 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
4356 		ret = -EINVAL;
4357 		break;
4358 	}
4359 	return ret;
4360 }
4361 
4362 static inline int
4363 igb_ethertype_filter_lookup(struct e1000_filter_info *filter_info,
4364 			uint16_t ethertype)
4365 {
4366 	int i;
4367 
4368 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4369 		if (filter_info->ethertype_filters[i] == ethertype &&
4370 		    (filter_info->ethertype_mask & (1 << i)))
4371 			return i;
4372 	}
4373 	return -1;
4374 }
4375 
4376 static inline int
4377 igb_ethertype_filter_insert(struct e1000_filter_info *filter_info,
4378 			uint16_t ethertype)
4379 {
4380 	int i;
4381 
4382 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4383 		if (!(filter_info->ethertype_mask & (1 << i))) {
4384 			filter_info->ethertype_mask |= 1 << i;
4385 			filter_info->ethertype_filters[i] = ethertype;
4386 			return i;
4387 		}
4388 	}
4389 	return -1;
4390 }
4391 
4392 static inline int
4393 igb_ethertype_filter_remove(struct e1000_filter_info *filter_info,
4394 			uint8_t idx)
4395 {
4396 	if (idx >= E1000_MAX_ETQF_FILTERS)
4397 		return -1;
4398 	filter_info->ethertype_mask &= ~(1 << idx);
4399 	filter_info->ethertype_filters[idx] = 0;
4400 	return idx;
4401 }
4402 
4403 
4404 static int
4405 igb_add_del_ethertype_filter(struct rte_eth_dev *dev,
4406 			struct rte_eth_ethertype_filter *filter,
4407 			bool add)
4408 {
4409 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4410 	struct e1000_filter_info *filter_info =
4411 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4412 	uint32_t etqf = 0;
4413 	int ret;
4414 
4415 	if (filter->ether_type == ETHER_TYPE_IPv4 ||
4416 		filter->ether_type == ETHER_TYPE_IPv6) {
4417 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
4418 			" ethertype filter.", filter->ether_type);
4419 		return -EINVAL;
4420 	}
4421 
4422 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
4423 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
4424 		return -EINVAL;
4425 	}
4426 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
4427 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
4428 		return -EINVAL;
4429 	}
4430 
4431 	ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4432 	if (ret >= 0 && add) {
4433 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
4434 			    filter->ether_type);
4435 		return -EEXIST;
4436 	}
4437 	if (ret < 0 && !add) {
4438 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4439 			    filter->ether_type);
4440 		return -ENOENT;
4441 	}
4442 
4443 	if (add) {
4444 		ret = igb_ethertype_filter_insert(filter_info,
4445 			filter->ether_type);
4446 		if (ret < 0) {
4447 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
4448 			return -ENOSYS;
4449 		}
4450 
4451 		etqf |= E1000_ETQF_FILTER_ENABLE | E1000_ETQF_QUEUE_ENABLE;
4452 		etqf |= (uint32_t)(filter->ether_type & E1000_ETQF_ETHERTYPE);
4453 		etqf |= filter->queue << E1000_ETQF_QUEUE_SHIFT;
4454 	} else {
4455 		ret = igb_ethertype_filter_remove(filter_info, (uint8_t)ret);
4456 		if (ret < 0)
4457 			return -ENOSYS;
4458 	}
4459 	E1000_WRITE_REG(hw, E1000_ETQF(ret), etqf);
4460 	E1000_WRITE_FLUSH(hw);
4461 
4462 	return 0;
4463 }
4464 
4465 static int
4466 igb_get_ethertype_filter(struct rte_eth_dev *dev,
4467 			struct rte_eth_ethertype_filter *filter)
4468 {
4469 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4470 	struct e1000_filter_info *filter_info =
4471 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4472 	uint32_t etqf;
4473 	int ret;
4474 
4475 	ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4476 	if (ret < 0) {
4477 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4478 			    filter->ether_type);
4479 		return -ENOENT;
4480 	}
4481 
4482 	etqf = E1000_READ_REG(hw, E1000_ETQF(ret));
4483 	if (etqf & E1000_ETQF_FILTER_ENABLE) {
4484 		filter->ether_type = etqf & E1000_ETQF_ETHERTYPE;
4485 		filter->flags = 0;
4486 		filter->queue = (etqf & E1000_ETQF_QUEUE) >>
4487 				E1000_ETQF_QUEUE_SHIFT;
4488 		return 0;
4489 	}
4490 
4491 	return -ENOENT;
4492 }
4493 
4494 /*
4495  * igb_ethertype_filter_handle - Handle operations for ethertype filter.
4496  * @dev: pointer to rte_eth_dev structure
4497  * @filter_op:operation will be taken.
4498  * @arg: a pointer to specific structure corresponding to the filter_op
4499  */
4500 static int
4501 igb_ethertype_filter_handle(struct rte_eth_dev *dev,
4502 				enum rte_filter_op filter_op,
4503 				void *arg)
4504 {
4505 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4506 	int ret;
4507 
4508 	MAC_TYPE_FILTER_SUP(hw->mac.type);
4509 
4510 	if (filter_op == RTE_ETH_FILTER_NOP)
4511 		return 0;
4512 
4513 	if (arg == NULL) {
4514 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4515 			    filter_op);
4516 		return -EINVAL;
4517 	}
4518 
4519 	switch (filter_op) {
4520 	case RTE_ETH_FILTER_ADD:
4521 		ret = igb_add_del_ethertype_filter(dev,
4522 			(struct rte_eth_ethertype_filter *)arg,
4523 			TRUE);
4524 		break;
4525 	case RTE_ETH_FILTER_DELETE:
4526 		ret = igb_add_del_ethertype_filter(dev,
4527 			(struct rte_eth_ethertype_filter *)arg,
4528 			FALSE);
4529 		break;
4530 	case RTE_ETH_FILTER_GET:
4531 		ret = igb_get_ethertype_filter(dev,
4532 			(struct rte_eth_ethertype_filter *)arg);
4533 		break;
4534 	default:
4535 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
4536 		ret = -EINVAL;
4537 		break;
4538 	}
4539 	return ret;
4540 }
4541 
4542 static int
4543 eth_igb_filter_ctrl(struct rte_eth_dev *dev,
4544 		     enum rte_filter_type filter_type,
4545 		     enum rte_filter_op filter_op,
4546 		     void *arg)
4547 {
4548 	int ret = -EINVAL;
4549 
4550 	switch (filter_type) {
4551 	case RTE_ETH_FILTER_NTUPLE:
4552 		ret = igb_ntuple_filter_handle(dev, filter_op, arg);
4553 		break;
4554 	case RTE_ETH_FILTER_ETHERTYPE:
4555 		ret = igb_ethertype_filter_handle(dev, filter_op, arg);
4556 		break;
4557 	case RTE_ETH_FILTER_SYN:
4558 		ret = eth_igb_syn_filter_handle(dev, filter_op, arg);
4559 		break;
4560 	case RTE_ETH_FILTER_FLEXIBLE:
4561 		ret = eth_igb_flex_filter_handle(dev, filter_op, arg);
4562 		break;
4563 	default:
4564 		PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
4565 							filter_type);
4566 		break;
4567 	}
4568 
4569 	return ret;
4570 }
4571 
4572 static int
4573 eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
4574 			 struct ether_addr *mc_addr_set,
4575 			 uint32_t nb_mc_addr)
4576 {
4577 	struct e1000_hw *hw;
4578 
4579 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4580 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
4581 	return 0;
4582 }
4583 
4584 static uint64_t
4585 igb_read_systime_cyclecounter(struct rte_eth_dev *dev)
4586 {
4587 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4588 	uint64_t systime_cycles;
4589 
4590 	switch (hw->mac.type) {
4591 	case e1000_i210:
4592 	case e1000_i211:
4593 		/*
4594 		 * Need to read System Time Residue Register to be able
4595 		 * to read the other two registers.
4596 		 */
4597 		E1000_READ_REG(hw, E1000_SYSTIMR);
4598 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
4599 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4600 		systime_cycles += (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4601 				* NSEC_PER_SEC;
4602 		break;
4603 	case e1000_82580:
4604 	case e1000_i350:
4605 	case e1000_i354:
4606 		/*
4607 		 * Need to read System Time Residue Register to be able
4608 		 * to read the other two registers.
4609 		 */
4610 		E1000_READ_REG(hw, E1000_SYSTIMR);
4611 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4612 		/* Only the 8 LSB are valid. */
4613 		systime_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_SYSTIMH)
4614 				& 0xff) << 32;
4615 		break;
4616 	default:
4617 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4618 		systime_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4619 				<< 32;
4620 		break;
4621 	}
4622 
4623 	return systime_cycles;
4624 }
4625 
4626 static uint64_t
4627 igb_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
4628 {
4629 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4630 	uint64_t rx_tstamp_cycles;
4631 
4632 	switch (hw->mac.type) {
4633 	case e1000_i210:
4634 	case e1000_i211:
4635 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
4636 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4637 		rx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
4638 				* NSEC_PER_SEC;
4639 		break;
4640 	case e1000_82580:
4641 	case e1000_i350:
4642 	case e1000_i354:
4643 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4644 		/* Only the 8 LSB are valid. */
4645 		rx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_RXSTMPH)
4646 				& 0xff) << 32;
4647 		break;
4648 	default:
4649 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4650 		rx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
4651 				<< 32;
4652 		break;
4653 	}
4654 
4655 	return rx_tstamp_cycles;
4656 }
4657 
4658 static uint64_t
4659 igb_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
4660 {
4661 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4662 	uint64_t tx_tstamp_cycles;
4663 
4664 	switch (hw->mac.type) {
4665 	case e1000_i210:
4666 	case e1000_i211:
4667 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
4668 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
4669 		tx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
4670 				* NSEC_PER_SEC;
4671 		break;
4672 	case e1000_82580:
4673 	case e1000_i350:
4674 	case e1000_i354:
4675 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
4676 		/* Only the 8 LSB are valid. */
4677 		tx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_TXSTMPH)
4678 				& 0xff) << 32;
4679 		break;
4680 	default:
4681 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
4682 		tx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
4683 				<< 32;
4684 		break;
4685 	}
4686 
4687 	return tx_tstamp_cycles;
4688 }
4689 
4690 static void
4691 igb_start_timecounters(struct rte_eth_dev *dev)
4692 {
4693 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4694 	struct e1000_adapter *adapter =
4695 		(struct e1000_adapter *)dev->data->dev_private;
4696 	uint32_t incval = 1;
4697 	uint32_t shift = 0;
4698 	uint64_t mask = E1000_CYCLECOUNTER_MASK;
4699 
4700 	switch (hw->mac.type) {
4701 	case e1000_82580:
4702 	case e1000_i350:
4703 	case e1000_i354:
4704 		/* 32 LSB bits + 8 MSB bits = 40 bits */
4705 		mask = (1ULL << 40) - 1;
4706 		/* fall-through */
4707 	case e1000_i210:
4708 	case e1000_i211:
4709 		/*
4710 		 * Start incrementing the register
4711 		 * used to timestamp PTP packets.
4712 		 */
4713 		E1000_WRITE_REG(hw, E1000_TIMINCA, incval);
4714 		break;
4715 	case e1000_82576:
4716 		incval = E1000_INCVALUE_82576;
4717 		shift = IGB_82576_TSYNC_SHIFT;
4718 		E1000_WRITE_REG(hw, E1000_TIMINCA,
4719 				E1000_INCPERIOD_82576 | incval);
4720 		break;
4721 	default:
4722 		/* Not supported */
4723 		return;
4724 	}
4725 
4726 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
4727 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
4728 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
4729 
4730 	adapter->systime_tc.cc_mask = mask;
4731 	adapter->systime_tc.cc_shift = shift;
4732 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
4733 
4734 	adapter->rx_tstamp_tc.cc_mask = mask;
4735 	adapter->rx_tstamp_tc.cc_shift = shift;
4736 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
4737 
4738 	adapter->tx_tstamp_tc.cc_mask = mask;
4739 	adapter->tx_tstamp_tc.cc_shift = shift;
4740 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
4741 }
4742 
4743 static int
4744 igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
4745 {
4746 	struct e1000_adapter *adapter =
4747 			(struct e1000_adapter *)dev->data->dev_private;
4748 
4749 	adapter->systime_tc.nsec += delta;
4750 	adapter->rx_tstamp_tc.nsec += delta;
4751 	adapter->tx_tstamp_tc.nsec += delta;
4752 
4753 	return 0;
4754 }
4755 
4756 static int
4757 igb_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
4758 {
4759 	uint64_t ns;
4760 	struct e1000_adapter *adapter =
4761 			(struct e1000_adapter *)dev->data->dev_private;
4762 
4763 	ns = rte_timespec_to_ns(ts);
4764 
4765 	/* Set the timecounters to a new value. */
4766 	adapter->systime_tc.nsec = ns;
4767 	adapter->rx_tstamp_tc.nsec = ns;
4768 	adapter->tx_tstamp_tc.nsec = ns;
4769 
4770 	return 0;
4771 }
4772 
4773 static int
4774 igb_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
4775 {
4776 	uint64_t ns, systime_cycles;
4777 	struct e1000_adapter *adapter =
4778 			(struct e1000_adapter *)dev->data->dev_private;
4779 
4780 	systime_cycles = igb_read_systime_cyclecounter(dev);
4781 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
4782 	*ts = rte_ns_to_timespec(ns);
4783 
4784 	return 0;
4785 }
4786 
4787 static int
4788 igb_timesync_enable(struct rte_eth_dev *dev)
4789 {
4790 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4791 	uint32_t tsync_ctl;
4792 	uint32_t tsauxc;
4793 
4794 	/* Stop the timesync system time. */
4795 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0x0);
4796 	/* Reset the timesync system time value. */
4797 	switch (hw->mac.type) {
4798 	case e1000_82580:
4799 	case e1000_i350:
4800 	case e1000_i354:
4801 	case e1000_i210:
4802 	case e1000_i211:
4803 		E1000_WRITE_REG(hw, E1000_SYSTIMR, 0x0);
4804 		/* fall-through */
4805 	case e1000_82576:
4806 		E1000_WRITE_REG(hw, E1000_SYSTIML, 0x0);
4807 		E1000_WRITE_REG(hw, E1000_SYSTIMH, 0x0);
4808 		break;
4809 	default:
4810 		/* Not supported. */
4811 		return -ENOTSUP;
4812 	}
4813 
4814 	/* Enable system time for it isn't on by default. */
4815 	tsauxc = E1000_READ_REG(hw, E1000_TSAUXC);
4816 	tsauxc &= ~E1000_TSAUXC_DISABLE_SYSTIME;
4817 	E1000_WRITE_REG(hw, E1000_TSAUXC, tsauxc);
4818 
4819 	igb_start_timecounters(dev);
4820 
4821 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
4822 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588),
4823 			(ETHER_TYPE_1588 |
4824 			 E1000_ETQF_FILTER_ENABLE |
4825 			 E1000_ETQF_1588));
4826 
4827 	/* Enable timestamping of received PTP packets. */
4828 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
4829 	tsync_ctl |= E1000_TSYNCRXCTL_ENABLED;
4830 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
4831 
4832 	/* Enable Timestamping of transmitted PTP packets. */
4833 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
4834 	tsync_ctl |= E1000_TSYNCTXCTL_ENABLED;
4835 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
4836 
4837 	return 0;
4838 }
4839 
4840 static int
4841 igb_timesync_disable(struct rte_eth_dev *dev)
4842 {
4843 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4844 	uint32_t tsync_ctl;
4845 
4846 	/* Disable timestamping of transmitted PTP packets. */
4847 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
4848 	tsync_ctl &= ~E1000_TSYNCTXCTL_ENABLED;
4849 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
4850 
4851 	/* Disable timestamping of received PTP packets. */
4852 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
4853 	tsync_ctl &= ~E1000_TSYNCRXCTL_ENABLED;
4854 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
4855 
4856 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
4857 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588), 0);
4858 
4859 	/* Stop incrementating the System Time registers. */
4860 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0);
4861 
4862 	return 0;
4863 }
4864 
4865 static int
4866 igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
4867 			       struct timespec *timestamp,
4868 			       uint32_t flags __rte_unused)
4869 {
4870 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4871 	struct e1000_adapter *adapter =
4872 			(struct e1000_adapter *)dev->data->dev_private;
4873 	uint32_t tsync_rxctl;
4874 	uint64_t rx_tstamp_cycles;
4875 	uint64_t ns;
4876 
4877 	tsync_rxctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
4878 	if ((tsync_rxctl & E1000_TSYNCRXCTL_VALID) == 0)
4879 		return -EINVAL;
4880 
4881 	rx_tstamp_cycles = igb_read_rx_tstamp_cyclecounter(dev);
4882 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
4883 	*timestamp = rte_ns_to_timespec(ns);
4884 
4885 	return  0;
4886 }
4887 
4888 static int
4889 igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
4890 			       struct timespec *timestamp)
4891 {
4892 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4893 	struct e1000_adapter *adapter =
4894 			(struct e1000_adapter *)dev->data->dev_private;
4895 	uint32_t tsync_txctl;
4896 	uint64_t tx_tstamp_cycles;
4897 	uint64_t ns;
4898 
4899 	tsync_txctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
4900 	if ((tsync_txctl & E1000_TSYNCTXCTL_VALID) == 0)
4901 		return -EINVAL;
4902 
4903 	tx_tstamp_cycles = igb_read_tx_tstamp_cyclecounter(dev);
4904 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
4905 	*timestamp = rte_ns_to_timespec(ns);
4906 
4907 	return  0;
4908 }
4909 
4910 static int
4911 eth_igb_get_reg_length(struct rte_eth_dev *dev __rte_unused)
4912 {
4913 	int count = 0;
4914 	int g_ind = 0;
4915 	const struct reg_info *reg_group;
4916 
4917 	while ((reg_group = igb_regs[g_ind++]))
4918 		count += igb_reg_group_count(reg_group);
4919 
4920 	return count;
4921 }
4922 
4923 static int
4924 igbvf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
4925 {
4926 	int count = 0;
4927 	int g_ind = 0;
4928 	const struct reg_info *reg_group;
4929 
4930 	while ((reg_group = igbvf_regs[g_ind++]))
4931 		count += igb_reg_group_count(reg_group);
4932 
4933 	return count;
4934 }
4935 
4936 static int
4937 eth_igb_get_regs(struct rte_eth_dev *dev,
4938 	struct rte_dev_reg_info *regs)
4939 {
4940 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4941 	uint32_t *data = regs->data;
4942 	int g_ind = 0;
4943 	int count = 0;
4944 	const struct reg_info *reg_group;
4945 
4946 	if (data == NULL) {
4947 		regs->length = eth_igb_get_reg_length(dev);
4948 		regs->width = sizeof(uint32_t);
4949 		return 0;
4950 	}
4951 
4952 	/* Support only full register dump */
4953 	if ((regs->length == 0) ||
4954 	    (regs->length == (uint32_t)eth_igb_get_reg_length(dev))) {
4955 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
4956 			hw->device_id;
4957 		while ((reg_group = igb_regs[g_ind++]))
4958 			count += igb_read_regs_group(dev, &data[count],
4959 							reg_group);
4960 		return 0;
4961 	}
4962 
4963 	return -ENOTSUP;
4964 }
4965 
4966 static int
4967 igbvf_get_regs(struct rte_eth_dev *dev,
4968 	struct rte_dev_reg_info *regs)
4969 {
4970 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4971 	uint32_t *data = regs->data;
4972 	int g_ind = 0;
4973 	int count = 0;
4974 	const struct reg_info *reg_group;
4975 
4976 	if (data == NULL) {
4977 		regs->length = igbvf_get_reg_length(dev);
4978 		regs->width = sizeof(uint32_t);
4979 		return 0;
4980 	}
4981 
4982 	/* Support only full register dump */
4983 	if ((regs->length == 0) ||
4984 	    (regs->length == (uint32_t)igbvf_get_reg_length(dev))) {
4985 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
4986 			hw->device_id;
4987 		while ((reg_group = igbvf_regs[g_ind++]))
4988 			count += igb_read_regs_group(dev, &data[count],
4989 							reg_group);
4990 		return 0;
4991 	}
4992 
4993 	return -ENOTSUP;
4994 }
4995 
4996 static int
4997 eth_igb_get_eeprom_length(struct rte_eth_dev *dev)
4998 {
4999 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5000 
5001 	/* Return unit is byte count */
5002 	return hw->nvm.word_size * 2;
5003 }
5004 
5005 static int
5006 eth_igb_get_eeprom(struct rte_eth_dev *dev,
5007 	struct rte_dev_eeprom_info *in_eeprom)
5008 {
5009 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5010 	struct e1000_nvm_info *nvm = &hw->nvm;
5011 	uint16_t *data = in_eeprom->data;
5012 	int first, length;
5013 
5014 	first = in_eeprom->offset >> 1;
5015 	length = in_eeprom->length >> 1;
5016 	if ((first >= hw->nvm.word_size) ||
5017 	    ((first + length) >= hw->nvm.word_size))
5018 		return -EINVAL;
5019 
5020 	in_eeprom->magic = hw->vendor_id |
5021 		((uint32_t)hw->device_id << 16);
5022 
5023 	if ((nvm->ops.read) == NULL)
5024 		return -ENOTSUP;
5025 
5026 	return nvm->ops.read(hw, first, length, data);
5027 }
5028 
5029 static int
5030 eth_igb_set_eeprom(struct rte_eth_dev *dev,
5031 	struct rte_dev_eeprom_info *in_eeprom)
5032 {
5033 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5034 	struct e1000_nvm_info *nvm = &hw->nvm;
5035 	uint16_t *data = in_eeprom->data;
5036 	int first, length;
5037 
5038 	first = in_eeprom->offset >> 1;
5039 	length = in_eeprom->length >> 1;
5040 	if ((first >= hw->nvm.word_size) ||
5041 	    ((first + length) >= hw->nvm.word_size))
5042 		return -EINVAL;
5043 
5044 	in_eeprom->magic = (uint32_t)hw->vendor_id |
5045 		((uint32_t)hw->device_id << 16);
5046 
5047 	if ((nvm->ops.write) == NULL)
5048 		return -ENOTSUP;
5049 	return nvm->ops.write(hw,  first, length, data);
5050 }
5051 
5052 static struct rte_driver pmd_igb_drv = {
5053 	.type = PMD_PDEV,
5054 	.init = rte_igb_pmd_init,
5055 };
5056 
5057 static struct rte_driver pmd_igbvf_drv = {
5058 	.type = PMD_PDEV,
5059 	.init = rte_igbvf_pmd_init,
5060 };
5061 
5062 static int
5063 eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5064 {
5065 	struct e1000_hw *hw =
5066 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5067 	uint32_t mask = 1 << queue_id;
5068 
5069 	E1000_WRITE_REG(hw, E1000_EIMC, mask);
5070 	E1000_WRITE_FLUSH(hw);
5071 
5072 	return 0;
5073 }
5074 
5075 static int
5076 eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5077 {
5078 	struct e1000_hw *hw =
5079 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5080 	uint32_t mask = 1 << queue_id;
5081 	uint32_t regval;
5082 
5083 	regval = E1000_READ_REG(hw, E1000_EIMS);
5084 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
5085 	E1000_WRITE_FLUSH(hw);
5086 
5087 	rte_intr_enable(&dev->pci_dev->intr_handle);
5088 
5089 	return 0;
5090 }
5091 
5092 static void
5093 eth_igb_write_ivar(struct e1000_hw *hw, uint8_t  msix_vector,
5094 		   uint8_t index, uint8_t offset)
5095 {
5096 	uint32_t val = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
5097 
5098 	/* clear bits */
5099 	val &= ~((uint32_t)0xFF << offset);
5100 
5101 	/* write vector and valid bit */
5102 	val |= (msix_vector | E1000_IVAR_VALID) << offset;
5103 
5104 	E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, val);
5105 }
5106 
5107 static void
5108 eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
5109 			   uint8_t queue, uint8_t msix_vector)
5110 {
5111 	uint32_t tmp = 0;
5112 
5113 	if (hw->mac.type == e1000_82575) {
5114 		if (direction == 0)
5115 			tmp = E1000_EICR_RX_QUEUE0 << queue;
5116 		else if (direction == 1)
5117 			tmp = E1000_EICR_TX_QUEUE0 << queue;
5118 		E1000_WRITE_REG(hw, E1000_MSIXBM(msix_vector), tmp);
5119 	} else if (hw->mac.type == e1000_82576) {
5120 		if ((direction == 0) || (direction == 1))
5121 			eth_igb_write_ivar(hw, msix_vector, queue & 0x7,
5122 					   ((queue & 0x8) << 1) +
5123 					   8 * direction);
5124 	} else if ((hw->mac.type == e1000_82580) ||
5125 			(hw->mac.type == e1000_i350) ||
5126 			(hw->mac.type == e1000_i354) ||
5127 			(hw->mac.type == e1000_i210) ||
5128 			(hw->mac.type == e1000_i211)) {
5129 		if ((direction == 0) || (direction == 1))
5130 			eth_igb_write_ivar(hw, msix_vector,
5131 					   queue >> 1,
5132 					   ((queue & 0x1) << 4) +
5133 					   8 * direction);
5134 	}
5135 }
5136 
5137 /* Sets up the hardware to generate MSI-X interrupts properly
5138  * @hw
5139  *  board private structure
5140  */
5141 static void
5142 eth_igb_configure_msix_intr(struct rte_eth_dev *dev)
5143 {
5144 	int queue_id;
5145 	uint32_t tmpval, regval, intr_mask;
5146 	struct e1000_hw *hw =
5147 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5148 	uint32_t vec = E1000_MISC_VEC_ID;
5149 	uint32_t base = E1000_MISC_VEC_ID;
5150 	uint32_t misc_shift = 0;
5151 
5152 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
5153 
5154 	/* won't configure msix register if no mapping is done
5155 	 * between intr vector and event fd
5156 	 */
5157 	if (!rte_intr_dp_is_en(intr_handle))
5158 		return;
5159 
5160 	if (rte_intr_allow_others(intr_handle)) {
5161 		vec = base = E1000_RX_VEC_START;
5162 		misc_shift = 1;
5163 	}
5164 
5165 	/* set interrupt vector for other causes */
5166 	if (hw->mac.type == e1000_82575) {
5167 		tmpval = E1000_READ_REG(hw, E1000_CTRL_EXT);
5168 		/* enable MSI-X PBA support */
5169 		tmpval |= E1000_CTRL_EXT_PBA_CLR;
5170 
5171 		/* Auto-Mask interrupts upon ICR read */
5172 		tmpval |= E1000_CTRL_EXT_EIAME;
5173 		tmpval |= E1000_CTRL_EXT_IRCA;
5174 
5175 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmpval);
5176 
5177 		/* enable msix_other interrupt */
5178 		E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0), 0, E1000_EIMS_OTHER);
5179 		regval = E1000_READ_REG(hw, E1000_EIAC);
5180 		E1000_WRITE_REG(hw, E1000_EIAC, regval | E1000_EIMS_OTHER);
5181 		regval = E1000_READ_REG(hw, E1000_EIAM);
5182 		E1000_WRITE_REG(hw, E1000_EIMS, regval | E1000_EIMS_OTHER);
5183 	} else if ((hw->mac.type == e1000_82576) ||
5184 			(hw->mac.type == e1000_82580) ||
5185 			(hw->mac.type == e1000_i350) ||
5186 			(hw->mac.type == e1000_i354) ||
5187 			(hw->mac.type == e1000_i210) ||
5188 			(hw->mac.type == e1000_i211)) {
5189 		/* turn on MSI-X capability first */
5190 		E1000_WRITE_REG(hw, E1000_GPIE, E1000_GPIE_MSIX_MODE |
5191 					E1000_GPIE_PBA | E1000_GPIE_EIAME |
5192 					E1000_GPIE_NSICR);
5193 		intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5194 			misc_shift;
5195 		regval = E1000_READ_REG(hw, E1000_EIAC);
5196 		E1000_WRITE_REG(hw, E1000_EIAC, regval | intr_mask);
5197 
5198 		/* enable msix_other interrupt */
5199 		regval = E1000_READ_REG(hw, E1000_EIMS);
5200 		E1000_WRITE_REG(hw, E1000_EIMS, regval | intr_mask);
5201 		tmpval = (dev->data->nb_rx_queues | E1000_IVAR_VALID) << 8;
5202 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, tmpval);
5203 	}
5204 
5205 	/* use EIAM to auto-mask when MSI-X interrupt
5206 	 * is asserted, this saves a register write for every interrupt
5207 	 */
5208 	intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5209 		misc_shift;
5210 	regval = E1000_READ_REG(hw, E1000_EIAM);
5211 	E1000_WRITE_REG(hw, E1000_EIAM, regval | intr_mask);
5212 
5213 	for (queue_id = 0; queue_id < dev->data->nb_rx_queues; queue_id++) {
5214 		eth_igb_assign_msix_vector(hw, 0, queue_id, vec);
5215 		intr_handle->intr_vec[queue_id] = vec;
5216 		if (vec < base + intr_handle->nb_efd - 1)
5217 			vec++;
5218 	}
5219 
5220 	E1000_WRITE_FLUSH(hw);
5221 }
5222 
5223 PMD_REGISTER_DRIVER(pmd_igb_drv, igb);
5224 DRIVER_REGISTER_PCI_TABLE(igb, pci_id_igb_map);
5225 PMD_REGISTER_DRIVER(pmd_igbvf_drv, igbvf);
5226 DRIVER_REGISTER_PCI_TABLE(igbvf, pci_id_igbvf_map);
5227