xref: /dpdk/drivers/net/e1000/igb_ethdev.c (revision a3c8a446)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2016 Intel Corporation
3  */
4 
5 #include <sys/queue.h>
6 #include <stdio.h>
7 #include <errno.h>
8 #include <stdint.h>
9 #include <stdarg.h>
10 
11 #include <rte_string_fns.h>
12 #include <rte_common.h>
13 #include <rte_interrupts.h>
14 #include <rte_byteorder.h>
15 #include <rte_log.h>
16 #include <rte_debug.h>
17 #include <rte_pci.h>
18 #include <rte_bus_pci.h>
19 #include <rte_ether.h>
20 #include <ethdev_driver.h>
21 #include <ethdev_pci.h>
22 #include <rte_memory.h>
23 #include <rte_eal.h>
24 #include <rte_malloc.h>
25 #include <rte_dev.h>
26 
27 #include "e1000_logs.h"
28 #include "base/e1000_api.h"
29 #include "e1000_ethdev.h"
30 #include "igb_regs.h"
31 
32 /*
33  * Default values for port configuration
34  */
35 #define IGB_DEFAULT_RX_FREE_THRESH  32
36 
37 #define IGB_DEFAULT_RX_PTHRESH      ((hw->mac.type == e1000_i354) ? 12 : 8)
38 #define IGB_DEFAULT_RX_HTHRESH      8
39 #define IGB_DEFAULT_RX_WTHRESH      ((hw->mac.type == e1000_82576) ? 1 : 4)
40 
41 #define IGB_DEFAULT_TX_PTHRESH      ((hw->mac.type == e1000_i354) ? 20 : 8)
42 #define IGB_DEFAULT_TX_HTHRESH      1
43 #define IGB_DEFAULT_TX_WTHRESH      ((hw->mac.type == e1000_82576) ? 1 : 16)
44 
45 /* Bit shift and mask */
46 #define IGB_4_BIT_WIDTH  (CHAR_BIT / 2)
47 #define IGB_4_BIT_MASK   RTE_LEN2MASK(IGB_4_BIT_WIDTH, uint8_t)
48 #define IGB_8_BIT_WIDTH  CHAR_BIT
49 #define IGB_8_BIT_MASK   UINT8_MAX
50 
51 /* Additional timesync values. */
52 #define E1000_CYCLECOUNTER_MASK      0xffffffffffffffffULL
53 #define E1000_ETQF_FILTER_1588       3
54 #define IGB_82576_TSYNC_SHIFT        16
55 #define E1000_INCPERIOD_82576        (1 << E1000_TIMINCA_16NS_SHIFT)
56 #define E1000_INCVALUE_82576         (16 << IGB_82576_TSYNC_SHIFT)
57 #define E1000_TSAUXC_DISABLE_SYSTIME 0x80000000
58 
59 #define E1000_VTIVAR_MISC                0x01740
60 #define E1000_VTIVAR_MISC_MASK           0xFF
61 #define E1000_VTIVAR_VALID               0x80
62 #define E1000_VTIVAR_MISC_MAILBOX        0
63 #define E1000_VTIVAR_MISC_INTR_MASK      0x3
64 
65 /* External VLAN Enable bit mask */
66 #define E1000_CTRL_EXT_EXT_VLAN      (1 << 26)
67 
68 /* External VLAN Ether Type bit mask and shift */
69 #define E1000_VET_VET_EXT            0xFFFF0000
70 #define E1000_VET_VET_EXT_SHIFT      16
71 
72 /* MSI-X other interrupt vector */
73 #define IGB_MSIX_OTHER_INTR_VEC      0
74 
75 static int  eth_igb_configure(struct rte_eth_dev *dev);
76 static int  eth_igb_start(struct rte_eth_dev *dev);
77 static int  eth_igb_stop(struct rte_eth_dev *dev);
78 static int  eth_igb_dev_set_link_up(struct rte_eth_dev *dev);
79 static int  eth_igb_dev_set_link_down(struct rte_eth_dev *dev);
80 static int eth_igb_close(struct rte_eth_dev *dev);
81 static int eth_igb_reset(struct rte_eth_dev *dev);
82 static int  eth_igb_promiscuous_enable(struct rte_eth_dev *dev);
83 static int  eth_igb_promiscuous_disable(struct rte_eth_dev *dev);
84 static int  eth_igb_allmulticast_enable(struct rte_eth_dev *dev);
85 static int  eth_igb_allmulticast_disable(struct rte_eth_dev *dev);
86 static int  eth_igb_link_update(struct rte_eth_dev *dev,
87 				int wait_to_complete);
88 static int eth_igb_stats_get(struct rte_eth_dev *dev,
89 				struct rte_eth_stats *rte_stats);
90 static int eth_igb_xstats_get(struct rte_eth_dev *dev,
91 			      struct rte_eth_xstat *xstats, unsigned n);
92 static int eth_igb_xstats_get_by_id(struct rte_eth_dev *dev,
93 		const uint64_t *ids,
94 		uint64_t *values, unsigned int n);
95 static int eth_igb_xstats_get_names(struct rte_eth_dev *dev,
96 				    struct rte_eth_xstat_name *xstats_names,
97 				    unsigned int size);
98 static int eth_igb_xstats_get_names_by_id(struct rte_eth_dev *dev,
99 		struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
100 		unsigned int limit);
101 static int eth_igb_stats_reset(struct rte_eth_dev *dev);
102 static int eth_igb_xstats_reset(struct rte_eth_dev *dev);
103 static int eth_igb_fw_version_get(struct rte_eth_dev *dev,
104 				   char *fw_version, size_t fw_size);
105 static int eth_igb_infos_get(struct rte_eth_dev *dev,
106 			      struct rte_eth_dev_info *dev_info);
107 static const uint32_t *eth_igb_supported_ptypes_get(struct rte_eth_dev *dev);
108 static int eth_igbvf_infos_get(struct rte_eth_dev *dev,
109 				struct rte_eth_dev_info *dev_info);
110 static int  eth_igb_flow_ctrl_get(struct rte_eth_dev *dev,
111 				struct rte_eth_fc_conf *fc_conf);
112 static int  eth_igb_flow_ctrl_set(struct rte_eth_dev *dev,
113 				struct rte_eth_fc_conf *fc_conf);
114 static int eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
115 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev);
116 static int eth_igb_interrupt_get_status(struct rte_eth_dev *dev);
117 static int eth_igb_interrupt_action(struct rte_eth_dev *dev,
118 				    struct rte_intr_handle *handle);
119 static void eth_igb_interrupt_handler(void *param);
120 static int  igb_hardware_init(struct e1000_hw *hw);
121 static void igb_hw_control_acquire(struct e1000_hw *hw);
122 static void igb_hw_control_release(struct e1000_hw *hw);
123 static void igb_init_manageability(struct e1000_hw *hw);
124 static void igb_release_manageability(struct e1000_hw *hw);
125 
126 static int  eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
127 
128 static int eth_igb_vlan_filter_set(struct rte_eth_dev *dev,
129 		uint16_t vlan_id, int on);
130 static int eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
131 				 enum rte_vlan_type vlan_type,
132 				 uint16_t tpid_id);
133 static int eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask);
134 
135 static void igb_vlan_hw_filter_enable(struct rte_eth_dev *dev);
136 static void igb_vlan_hw_filter_disable(struct rte_eth_dev *dev);
137 static void igb_vlan_hw_strip_enable(struct rte_eth_dev *dev);
138 static void igb_vlan_hw_strip_disable(struct rte_eth_dev *dev);
139 static void igb_vlan_hw_extend_enable(struct rte_eth_dev *dev);
140 static void igb_vlan_hw_extend_disable(struct rte_eth_dev *dev);
141 
142 static int eth_igb_led_on(struct rte_eth_dev *dev);
143 static int eth_igb_led_off(struct rte_eth_dev *dev);
144 
145 static void igb_intr_disable(struct rte_eth_dev *dev);
146 static int  igb_get_rx_buffer_size(struct e1000_hw *hw);
147 static int eth_igb_rar_set(struct rte_eth_dev *dev,
148 			   struct rte_ether_addr *mac_addr,
149 			   uint32_t index, uint32_t pool);
150 static void eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index);
151 static int eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
152 		struct rte_ether_addr *addr);
153 
154 static void igbvf_intr_disable(struct e1000_hw *hw);
155 static int igbvf_dev_configure(struct rte_eth_dev *dev);
156 static int igbvf_dev_start(struct rte_eth_dev *dev);
157 static int igbvf_dev_stop(struct rte_eth_dev *dev);
158 static int igbvf_dev_close(struct rte_eth_dev *dev);
159 static int igbvf_promiscuous_enable(struct rte_eth_dev *dev);
160 static int igbvf_promiscuous_disable(struct rte_eth_dev *dev);
161 static int igbvf_allmulticast_enable(struct rte_eth_dev *dev);
162 static int igbvf_allmulticast_disable(struct rte_eth_dev *dev);
163 static int eth_igbvf_link_update(struct e1000_hw *hw);
164 static int eth_igbvf_stats_get(struct rte_eth_dev *dev,
165 				struct rte_eth_stats *rte_stats);
166 static int eth_igbvf_xstats_get(struct rte_eth_dev *dev,
167 				struct rte_eth_xstat *xstats, unsigned n);
168 static int eth_igbvf_xstats_get_names(struct rte_eth_dev *dev,
169 				      struct rte_eth_xstat_name *xstats_names,
170 				      unsigned limit);
171 static int eth_igbvf_stats_reset(struct rte_eth_dev *dev);
172 static int igbvf_vlan_filter_set(struct rte_eth_dev *dev,
173 		uint16_t vlan_id, int on);
174 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on);
175 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on);
176 static int igbvf_default_mac_addr_set(struct rte_eth_dev *dev,
177 		struct rte_ether_addr *addr);
178 static int igbvf_get_reg_length(struct rte_eth_dev *dev);
179 static int igbvf_get_regs(struct rte_eth_dev *dev,
180 		struct rte_dev_reg_info *regs);
181 
182 static int eth_igb_rss_reta_update(struct rte_eth_dev *dev,
183 				   struct rte_eth_rss_reta_entry64 *reta_conf,
184 				   uint16_t reta_size);
185 static int eth_igb_rss_reta_query(struct rte_eth_dev *dev,
186 				  struct rte_eth_rss_reta_entry64 *reta_conf,
187 				  uint16_t reta_size);
188 
189 static int igb_add_2tuple_filter(struct rte_eth_dev *dev,
190 			struct rte_eth_ntuple_filter *ntuple_filter);
191 static int igb_remove_2tuple_filter(struct rte_eth_dev *dev,
192 			struct rte_eth_ntuple_filter *ntuple_filter);
193 static int igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
194 			struct rte_eth_ntuple_filter *ntuple_filter);
195 static int igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
196 			struct rte_eth_ntuple_filter *ntuple_filter);
197 static int eth_igb_flow_ops_get(struct rte_eth_dev *dev,
198 				const struct rte_flow_ops **ops);
199 static int eth_igb_get_reg_length(struct rte_eth_dev *dev);
200 static int eth_igb_get_regs(struct rte_eth_dev *dev,
201 		struct rte_dev_reg_info *regs);
202 static int eth_igb_get_eeprom_length(struct rte_eth_dev *dev);
203 static int eth_igb_get_eeprom(struct rte_eth_dev *dev,
204 		struct rte_dev_eeprom_info *eeprom);
205 static int eth_igb_set_eeprom(struct rte_eth_dev *dev,
206 		struct rte_dev_eeprom_info *eeprom);
207 static int eth_igb_get_module_info(struct rte_eth_dev *dev,
208 				   struct rte_eth_dev_module_info *modinfo);
209 static int eth_igb_get_module_eeprom(struct rte_eth_dev *dev,
210 				     struct rte_dev_eeprom_info *info);
211 static int eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
212 				    struct rte_ether_addr *mc_addr_set,
213 				    uint32_t nb_mc_addr);
214 static int igb_timesync_enable(struct rte_eth_dev *dev);
215 static int igb_timesync_disable(struct rte_eth_dev *dev);
216 static int igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
217 					  struct timespec *timestamp,
218 					  uint32_t flags);
219 static int igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
220 					  struct timespec *timestamp);
221 static int igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
222 static int igb_timesync_read_time(struct rte_eth_dev *dev,
223 				  struct timespec *timestamp);
224 static int igb_timesync_write_time(struct rte_eth_dev *dev,
225 				   const struct timespec *timestamp);
226 static int eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev,
227 					uint16_t queue_id);
228 static int eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev,
229 					 uint16_t queue_id);
230 static void eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
231 				       uint8_t queue, uint8_t msix_vector);
232 static void eth_igb_write_ivar(struct e1000_hw *hw, uint8_t msix_vector,
233 			       uint8_t index, uint8_t offset);
234 static void eth_igb_configure_msix_intr(struct rte_eth_dev *dev);
235 static void eth_igbvf_interrupt_handler(void *param);
236 static void igbvf_mbx_process(struct rte_eth_dev *dev);
237 static int igb_filter_restore(struct rte_eth_dev *dev);
238 
239 /*
240  * Define VF Stats MACRO for Non "cleared on read" register
241  */
242 #define UPDATE_VF_STAT(reg, last, cur)            \
243 {                                                 \
244 	u32 latest = E1000_READ_REG(hw, reg);     \
245 	cur += (latest - last) & UINT_MAX;        \
246 	last = latest;                            \
247 }
248 
249 #define IGB_FC_PAUSE_TIME 0x0680
250 #define IGB_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
251 #define IGB_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
252 
253 #define IGBVF_PMD_NAME "rte_igbvf_pmd"     /* PMD name */
254 
255 static enum e1000_fc_mode igb_fc_setting = e1000_fc_full;
256 
257 /*
258  * The set of PCI devices this driver supports
259  */
260 static const struct rte_pci_id pci_id_igb_map[] = {
261 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576) },
262 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_FIBER) },
263 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_SERDES) },
264 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_QUAD_COPPER) },
265 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_QUAD_COPPER_ET2) },
266 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_NS) },
267 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_NS_SERDES) },
268 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_SERDES_QUAD) },
269 
270 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575EB_COPPER) },
271 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575EB_FIBER_SERDES) },
272 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575GB_QUAD_COPPER) },
273 
274 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_COPPER) },
275 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_FIBER) },
276 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_SERDES) },
277 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_SGMII) },
278 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_COPPER_DUAL) },
279 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_QUAD_FIBER) },
280 
281 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_COPPER) },
282 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_FIBER) },
283 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_SERDES) },
284 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_SGMII) },
285 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_DA4) },
286 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER) },
287 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_OEM1) },
288 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_IT) },
289 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_FIBER) },
290 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SERDES) },
291 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SGMII) },
292 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_FLASHLESS) },
293 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SERDES_FLASHLESS) },
294 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I211_COPPER) },
295 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
296 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_SGMII) },
297 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
298 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SGMII) },
299 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SERDES) },
300 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_BACKPLANE) },
301 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SFP) },
302 	{ .vendor_id = 0, /* sentinel */ },
303 };
304 
305 /*
306  * The set of PCI devices this driver supports (for 82576&I350 VF)
307  */
308 static const struct rte_pci_id pci_id_igbvf_map[] = {
309 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_VF) },
310 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_VF_HV) },
311 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_VF) },
312 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_VF_HV) },
313 	{ .vendor_id = 0, /* sentinel */ },
314 };
315 
316 static const struct rte_eth_desc_lim rx_desc_lim = {
317 	.nb_max = E1000_MAX_RING_DESC,
318 	.nb_min = E1000_MIN_RING_DESC,
319 	.nb_align = IGB_RXD_ALIGN,
320 };
321 
322 static const struct rte_eth_desc_lim tx_desc_lim = {
323 	.nb_max = E1000_MAX_RING_DESC,
324 	.nb_min = E1000_MIN_RING_DESC,
325 	.nb_align = IGB_RXD_ALIGN,
326 	.nb_seg_max = IGB_TX_MAX_SEG,
327 	.nb_mtu_seg_max = IGB_TX_MAX_MTU_SEG,
328 };
329 
330 static const struct eth_dev_ops eth_igb_ops = {
331 	.dev_configure        = eth_igb_configure,
332 	.dev_start            = eth_igb_start,
333 	.dev_stop             = eth_igb_stop,
334 	.dev_set_link_up      = eth_igb_dev_set_link_up,
335 	.dev_set_link_down    = eth_igb_dev_set_link_down,
336 	.dev_close            = eth_igb_close,
337 	.dev_reset            = eth_igb_reset,
338 	.promiscuous_enable   = eth_igb_promiscuous_enable,
339 	.promiscuous_disable  = eth_igb_promiscuous_disable,
340 	.allmulticast_enable  = eth_igb_allmulticast_enable,
341 	.allmulticast_disable = eth_igb_allmulticast_disable,
342 	.link_update          = eth_igb_link_update,
343 	.stats_get            = eth_igb_stats_get,
344 	.xstats_get           = eth_igb_xstats_get,
345 	.xstats_get_by_id     = eth_igb_xstats_get_by_id,
346 	.xstats_get_names_by_id = eth_igb_xstats_get_names_by_id,
347 	.xstats_get_names     = eth_igb_xstats_get_names,
348 	.stats_reset          = eth_igb_stats_reset,
349 	.xstats_reset         = eth_igb_xstats_reset,
350 	.fw_version_get       = eth_igb_fw_version_get,
351 	.dev_infos_get        = eth_igb_infos_get,
352 	.dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
353 	.mtu_set              = eth_igb_mtu_set,
354 	.vlan_filter_set      = eth_igb_vlan_filter_set,
355 	.vlan_tpid_set        = eth_igb_vlan_tpid_set,
356 	.vlan_offload_set     = eth_igb_vlan_offload_set,
357 	.rx_queue_setup       = eth_igb_rx_queue_setup,
358 	.rx_queue_intr_enable = eth_igb_rx_queue_intr_enable,
359 	.rx_queue_intr_disable = eth_igb_rx_queue_intr_disable,
360 	.rx_queue_release     = eth_igb_rx_queue_release,
361 	.tx_queue_setup       = eth_igb_tx_queue_setup,
362 	.tx_queue_release     = eth_igb_tx_queue_release,
363 	.tx_done_cleanup      = eth_igb_tx_done_cleanup,
364 	.dev_led_on           = eth_igb_led_on,
365 	.dev_led_off          = eth_igb_led_off,
366 	.flow_ctrl_get        = eth_igb_flow_ctrl_get,
367 	.flow_ctrl_set        = eth_igb_flow_ctrl_set,
368 	.mac_addr_add         = eth_igb_rar_set,
369 	.mac_addr_remove      = eth_igb_rar_clear,
370 	.mac_addr_set         = eth_igb_default_mac_addr_set,
371 	.reta_update          = eth_igb_rss_reta_update,
372 	.reta_query           = eth_igb_rss_reta_query,
373 	.rss_hash_update      = eth_igb_rss_hash_update,
374 	.rss_hash_conf_get    = eth_igb_rss_hash_conf_get,
375 	.flow_ops_get         = eth_igb_flow_ops_get,
376 	.set_mc_addr_list     = eth_igb_set_mc_addr_list,
377 	.rxq_info_get         = igb_rxq_info_get,
378 	.txq_info_get         = igb_txq_info_get,
379 	.timesync_enable      = igb_timesync_enable,
380 	.timesync_disable     = igb_timesync_disable,
381 	.timesync_read_rx_timestamp = igb_timesync_read_rx_timestamp,
382 	.timesync_read_tx_timestamp = igb_timesync_read_tx_timestamp,
383 	.get_reg              = eth_igb_get_regs,
384 	.get_eeprom_length    = eth_igb_get_eeprom_length,
385 	.get_eeprom           = eth_igb_get_eeprom,
386 	.set_eeprom           = eth_igb_set_eeprom,
387 	.get_module_info      = eth_igb_get_module_info,
388 	.get_module_eeprom    = eth_igb_get_module_eeprom,
389 	.timesync_adjust_time = igb_timesync_adjust_time,
390 	.timesync_read_time   = igb_timesync_read_time,
391 	.timesync_write_time  = igb_timesync_write_time,
392 };
393 
394 /*
395  * dev_ops for virtual function, bare necessities for basic vf
396  * operation have been implemented
397  */
398 static const struct eth_dev_ops igbvf_eth_dev_ops = {
399 	.dev_configure        = igbvf_dev_configure,
400 	.dev_start            = igbvf_dev_start,
401 	.dev_stop             = igbvf_dev_stop,
402 	.dev_close            = igbvf_dev_close,
403 	.promiscuous_enable   = igbvf_promiscuous_enable,
404 	.promiscuous_disable  = igbvf_promiscuous_disable,
405 	.allmulticast_enable  = igbvf_allmulticast_enable,
406 	.allmulticast_disable = igbvf_allmulticast_disable,
407 	.link_update          = eth_igb_link_update,
408 	.stats_get            = eth_igbvf_stats_get,
409 	.xstats_get           = eth_igbvf_xstats_get,
410 	.xstats_get_names     = eth_igbvf_xstats_get_names,
411 	.stats_reset          = eth_igbvf_stats_reset,
412 	.xstats_reset         = eth_igbvf_stats_reset,
413 	.vlan_filter_set      = igbvf_vlan_filter_set,
414 	.dev_infos_get        = eth_igbvf_infos_get,
415 	.dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
416 	.rx_queue_setup       = eth_igb_rx_queue_setup,
417 	.rx_queue_release     = eth_igb_rx_queue_release,
418 	.tx_queue_setup       = eth_igb_tx_queue_setup,
419 	.tx_queue_release     = eth_igb_tx_queue_release,
420 	.tx_done_cleanup      = eth_igb_tx_done_cleanup,
421 	.set_mc_addr_list     = eth_igb_set_mc_addr_list,
422 	.rxq_info_get         = igb_rxq_info_get,
423 	.txq_info_get         = igb_txq_info_get,
424 	.mac_addr_set         = igbvf_default_mac_addr_set,
425 	.get_reg              = igbvf_get_regs,
426 };
427 
428 /* store statistics names and its offset in stats structure */
429 struct rte_igb_xstats_name_off {
430 	char name[RTE_ETH_XSTATS_NAME_SIZE];
431 	unsigned offset;
432 };
433 
434 static const struct rte_igb_xstats_name_off rte_igb_stats_strings[] = {
435 	{"rx_crc_errors", offsetof(struct e1000_hw_stats, crcerrs)},
436 	{"rx_align_errors", offsetof(struct e1000_hw_stats, algnerrc)},
437 	{"rx_symbol_errors", offsetof(struct e1000_hw_stats, symerrs)},
438 	{"rx_missed_packets", offsetof(struct e1000_hw_stats, mpc)},
439 	{"tx_single_collision_packets", offsetof(struct e1000_hw_stats, scc)},
440 	{"tx_multiple_collision_packets", offsetof(struct e1000_hw_stats, mcc)},
441 	{"tx_excessive_collision_packets", offsetof(struct e1000_hw_stats,
442 		ecol)},
443 	{"tx_late_collisions", offsetof(struct e1000_hw_stats, latecol)},
444 	{"tx_total_collisions", offsetof(struct e1000_hw_stats, colc)},
445 	{"tx_deferred_packets", offsetof(struct e1000_hw_stats, dc)},
446 	{"tx_no_carrier_sense_packets", offsetof(struct e1000_hw_stats, tncrs)},
447 	{"rx_carrier_ext_errors", offsetof(struct e1000_hw_stats, cexterr)},
448 	{"rx_length_errors", offsetof(struct e1000_hw_stats, rlec)},
449 	{"rx_xon_packets", offsetof(struct e1000_hw_stats, xonrxc)},
450 	{"tx_xon_packets", offsetof(struct e1000_hw_stats, xontxc)},
451 	{"rx_xoff_packets", offsetof(struct e1000_hw_stats, xoffrxc)},
452 	{"tx_xoff_packets", offsetof(struct e1000_hw_stats, xofftxc)},
453 	{"rx_flow_control_unsupported_packets", offsetof(struct e1000_hw_stats,
454 		fcruc)},
455 	{"rx_size_64_packets", offsetof(struct e1000_hw_stats, prc64)},
456 	{"rx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, prc127)},
457 	{"rx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, prc255)},
458 	{"rx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, prc511)},
459 	{"rx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
460 		prc1023)},
461 	{"rx_size_1024_to_max_packets", offsetof(struct e1000_hw_stats,
462 		prc1522)},
463 	{"rx_broadcast_packets", offsetof(struct e1000_hw_stats, bprc)},
464 	{"rx_multicast_packets", offsetof(struct e1000_hw_stats, mprc)},
465 	{"rx_undersize_errors", offsetof(struct e1000_hw_stats, ruc)},
466 	{"rx_fragment_errors", offsetof(struct e1000_hw_stats, rfc)},
467 	{"rx_oversize_errors", offsetof(struct e1000_hw_stats, roc)},
468 	{"rx_jabber_errors", offsetof(struct e1000_hw_stats, rjc)},
469 	{"rx_management_packets", offsetof(struct e1000_hw_stats, mgprc)},
470 	{"rx_management_dropped", offsetof(struct e1000_hw_stats, mgpdc)},
471 	{"tx_management_packets", offsetof(struct e1000_hw_stats, mgptc)},
472 	{"rx_total_packets", offsetof(struct e1000_hw_stats, tpr)},
473 	{"tx_total_packets", offsetof(struct e1000_hw_stats, tpt)},
474 	{"rx_total_bytes", offsetof(struct e1000_hw_stats, tor)},
475 	{"tx_total_bytes", offsetof(struct e1000_hw_stats, tot)},
476 	{"tx_size_64_packets", offsetof(struct e1000_hw_stats, ptc64)},
477 	{"tx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, ptc127)},
478 	{"tx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, ptc255)},
479 	{"tx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, ptc511)},
480 	{"tx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
481 		ptc1023)},
482 	{"tx_size_1023_to_max_packets", offsetof(struct e1000_hw_stats,
483 		ptc1522)},
484 	{"tx_multicast_packets", offsetof(struct e1000_hw_stats, mptc)},
485 	{"tx_broadcast_packets", offsetof(struct e1000_hw_stats, bptc)},
486 	{"tx_tso_packets", offsetof(struct e1000_hw_stats, tsctc)},
487 	{"tx_tso_errors", offsetof(struct e1000_hw_stats, tsctfc)},
488 	{"rx_sent_to_host_packets", offsetof(struct e1000_hw_stats, rpthc)},
489 	{"tx_sent_by_host_packets", offsetof(struct e1000_hw_stats, hgptc)},
490 	{"rx_code_violation_packets", offsetof(struct e1000_hw_stats, scvpc)},
491 
492 	{"interrupt_assert_count", offsetof(struct e1000_hw_stats, iac)},
493 };
494 
495 #define IGB_NB_XSTATS (sizeof(rte_igb_stats_strings) / \
496 		sizeof(rte_igb_stats_strings[0]))
497 
498 static const struct rte_igb_xstats_name_off rte_igbvf_stats_strings[] = {
499 	{"rx_multicast_packets", offsetof(struct e1000_vf_stats, mprc)},
500 	{"rx_good_loopback_packets", offsetof(struct e1000_vf_stats, gprlbc)},
501 	{"tx_good_loopback_packets", offsetof(struct e1000_vf_stats, gptlbc)},
502 	{"rx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gorlbc)},
503 	{"tx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gotlbc)},
504 };
505 
506 #define IGBVF_NB_XSTATS (sizeof(rte_igbvf_stats_strings) / \
507 		sizeof(rte_igbvf_stats_strings[0]))
508 
509 
510 static inline void
511 igb_intr_enable(struct rte_eth_dev *dev)
512 {
513 	struct e1000_interrupt *intr =
514 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
515 	struct e1000_hw *hw =
516 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
517 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
518 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
519 
520 	if (rte_intr_allow_others(intr_handle) &&
521 		dev->data->dev_conf.intr_conf.lsc != 0) {
522 		E1000_WRITE_REG(hw, E1000_EIMS, 1 << IGB_MSIX_OTHER_INTR_VEC);
523 	}
524 
525 	E1000_WRITE_REG(hw, E1000_IMS, intr->mask);
526 	E1000_WRITE_FLUSH(hw);
527 }
528 
529 static void
530 igb_intr_disable(struct rte_eth_dev *dev)
531 {
532 	struct e1000_hw *hw =
533 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
534 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
535 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
536 
537 	if (rte_intr_allow_others(intr_handle) &&
538 		dev->data->dev_conf.intr_conf.lsc != 0) {
539 		E1000_WRITE_REG(hw, E1000_EIMC, 1 << IGB_MSIX_OTHER_INTR_VEC);
540 	}
541 
542 	E1000_WRITE_REG(hw, E1000_IMC, ~0);
543 	E1000_WRITE_FLUSH(hw);
544 }
545 
546 static inline void
547 igbvf_intr_enable(struct rte_eth_dev *dev)
548 {
549 	struct e1000_hw *hw =
550 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
551 
552 	/* only for mailbox */
553 	E1000_WRITE_REG(hw, E1000_EIAM, 1 << E1000_VTIVAR_MISC_MAILBOX);
554 	E1000_WRITE_REG(hw, E1000_EIAC, 1 << E1000_VTIVAR_MISC_MAILBOX);
555 	E1000_WRITE_REG(hw, E1000_EIMS, 1 << E1000_VTIVAR_MISC_MAILBOX);
556 	E1000_WRITE_FLUSH(hw);
557 }
558 
559 /* only for mailbox now. If RX/TX needed, should extend this function.  */
560 static void
561 igbvf_set_ivar_map(struct e1000_hw *hw, uint8_t msix_vector)
562 {
563 	uint32_t tmp = 0;
564 
565 	/* mailbox */
566 	tmp |= (msix_vector & E1000_VTIVAR_MISC_INTR_MASK);
567 	tmp |= E1000_VTIVAR_VALID;
568 	E1000_WRITE_REG(hw, E1000_VTIVAR_MISC, tmp);
569 }
570 
571 static void
572 eth_igbvf_configure_msix_intr(struct rte_eth_dev *dev)
573 {
574 	struct e1000_hw *hw =
575 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
576 
577 	/* Configure VF other cause ivar */
578 	igbvf_set_ivar_map(hw, E1000_VTIVAR_MISC_MAILBOX);
579 }
580 
581 static inline int32_t
582 igb_pf_reset_hw(struct e1000_hw *hw)
583 {
584 	uint32_t ctrl_ext;
585 	int32_t status;
586 
587 	status = e1000_reset_hw(hw);
588 
589 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
590 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
591 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
592 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
593 	E1000_WRITE_FLUSH(hw);
594 
595 	return status;
596 }
597 
598 static void
599 igb_identify_hardware(struct rte_eth_dev *dev, struct rte_pci_device *pci_dev)
600 {
601 	struct e1000_hw *hw =
602 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
603 
604 
605 	hw->vendor_id = pci_dev->id.vendor_id;
606 	hw->device_id = pci_dev->id.device_id;
607 	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
608 	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
609 
610 	e1000_set_mac_type(hw);
611 
612 	/* need to check if it is a vf device below */
613 }
614 
615 static int
616 igb_reset_swfw_lock(struct e1000_hw *hw)
617 {
618 	int ret_val;
619 
620 	/*
621 	 * Do mac ops initialization manually here, since we will need
622 	 * some function pointers set by this call.
623 	 */
624 	ret_val = e1000_init_mac_params(hw);
625 	if (ret_val)
626 		return ret_val;
627 
628 	/*
629 	 * SMBI lock should not fail in this early stage. If this is the case,
630 	 * it is due to an improper exit of the application.
631 	 * So force the release of the faulty lock.
632 	 */
633 	if (e1000_get_hw_semaphore_generic(hw) < 0) {
634 		PMD_DRV_LOG(DEBUG, "SMBI lock released");
635 	}
636 	e1000_put_hw_semaphore_generic(hw);
637 
638 	if (hw->mac.ops.acquire_swfw_sync != NULL) {
639 		uint16_t mask;
640 
641 		/*
642 		 * Phy lock should not fail in this early stage. If this is the case,
643 		 * it is due to an improper exit of the application.
644 		 * So force the release of the faulty lock.
645 		 */
646 		mask = E1000_SWFW_PHY0_SM << hw->bus.func;
647 		if (hw->bus.func > E1000_FUNC_1)
648 			mask <<= 2;
649 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
650 			PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released",
651 				    hw->bus.func);
652 		}
653 		hw->mac.ops.release_swfw_sync(hw, mask);
654 
655 		/*
656 		 * This one is more tricky since it is common to all ports; but
657 		 * swfw_sync retries last long enough (1s) to be almost sure that if
658 		 * lock can not be taken it is due to an improper lock of the
659 		 * semaphore.
660 		 */
661 		mask = E1000_SWFW_EEP_SM;
662 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
663 			PMD_DRV_LOG(DEBUG, "SWFW common locks released");
664 		}
665 		hw->mac.ops.release_swfw_sync(hw, mask);
666 	}
667 
668 	return E1000_SUCCESS;
669 }
670 
671 /* Remove all ntuple filters of the device */
672 static int igb_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
673 {
674 	struct e1000_filter_info *filter_info =
675 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
676 	struct e1000_5tuple_filter *p_5tuple;
677 	struct e1000_2tuple_filter *p_2tuple;
678 
679 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
680 		TAILQ_REMOVE(&filter_info->fivetuple_list,
681 			p_5tuple, entries);
682 			rte_free(p_5tuple);
683 	}
684 	filter_info->fivetuple_mask = 0;
685 	while ((p_2tuple = TAILQ_FIRST(&filter_info->twotuple_list))) {
686 		TAILQ_REMOVE(&filter_info->twotuple_list,
687 			p_2tuple, entries);
688 			rte_free(p_2tuple);
689 	}
690 	filter_info->twotuple_mask = 0;
691 
692 	return 0;
693 }
694 
695 /* Remove all flex filters of the device */
696 static int igb_flex_filter_uninit(struct rte_eth_dev *eth_dev)
697 {
698 	struct e1000_filter_info *filter_info =
699 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
700 	struct e1000_flex_filter *p_flex;
701 
702 	while ((p_flex = TAILQ_FIRST(&filter_info->flex_list))) {
703 		TAILQ_REMOVE(&filter_info->flex_list, p_flex, entries);
704 		rte_free(p_flex);
705 	}
706 	filter_info->flex_mask = 0;
707 
708 	return 0;
709 }
710 
711 static int
712 eth_igb_dev_init(struct rte_eth_dev *eth_dev)
713 {
714 	int error = 0;
715 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
716 	struct e1000_hw *hw =
717 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
718 	struct e1000_vfta * shadow_vfta =
719 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
720 	struct e1000_filter_info *filter_info =
721 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
722 	struct e1000_adapter *adapter =
723 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
724 
725 	uint32_t ctrl_ext;
726 
727 	eth_dev->dev_ops = &eth_igb_ops;
728 	eth_dev->rx_queue_count = eth_igb_rx_queue_count;
729 	eth_dev->rx_descriptor_done   = eth_igb_rx_descriptor_done;
730 	eth_dev->rx_descriptor_status = eth_igb_rx_descriptor_status;
731 	eth_dev->tx_descriptor_status = eth_igb_tx_descriptor_status;
732 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
733 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
734 	eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
735 
736 	/* for secondary processes, we don't initialise any further as primary
737 	 * has already done this work. Only check we don't need a different
738 	 * RX function */
739 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
740 		if (eth_dev->data->scattered_rx)
741 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
742 		return 0;
743 	}
744 
745 	rte_eth_copy_pci_info(eth_dev, pci_dev);
746 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
747 
748 	hw->hw_addr= (void *)pci_dev->mem_resource[0].addr;
749 
750 	igb_identify_hardware(eth_dev, pci_dev);
751 	if (e1000_setup_init_funcs(hw, FALSE) != E1000_SUCCESS) {
752 		error = -EIO;
753 		goto err_late;
754 	}
755 
756 	e1000_get_bus_info(hw);
757 
758 	/* Reset any pending lock */
759 	if (igb_reset_swfw_lock(hw) != E1000_SUCCESS) {
760 		error = -EIO;
761 		goto err_late;
762 	}
763 
764 	/* Finish initialization */
765 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS) {
766 		error = -EIO;
767 		goto err_late;
768 	}
769 
770 	hw->mac.autoneg = 1;
771 	hw->phy.autoneg_wait_to_complete = 0;
772 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
773 
774 	/* Copper options */
775 	if (hw->phy.media_type == e1000_media_type_copper) {
776 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
777 		hw->phy.disable_polarity_correction = 0;
778 		hw->phy.ms_type = e1000_ms_hw_default;
779 	}
780 
781 	/*
782 	 * Start from a known state, this is important in reading the nvm
783 	 * and mac from that.
784 	 */
785 	igb_pf_reset_hw(hw);
786 
787 	/* Make sure we have a good EEPROM before we read from it */
788 	if (e1000_validate_nvm_checksum(hw) < 0) {
789 		/*
790 		 * Some PCI-E parts fail the first check due to
791 		 * the link being in sleep state, call it again,
792 		 * if it fails a second time its a real issue.
793 		 */
794 		if (e1000_validate_nvm_checksum(hw) < 0) {
795 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
796 			error = -EIO;
797 			goto err_late;
798 		}
799 	}
800 
801 	/* Read the permanent MAC address out of the EEPROM */
802 	if (e1000_read_mac_addr(hw) != 0) {
803 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
804 		error = -EIO;
805 		goto err_late;
806 	}
807 
808 	/* Allocate memory for storing MAC addresses */
809 	eth_dev->data->mac_addrs = rte_zmalloc("e1000",
810 		RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count, 0);
811 	if (eth_dev->data->mac_addrs == NULL) {
812 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
813 						"store MAC addresses",
814 				RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
815 		error = -ENOMEM;
816 		goto err_late;
817 	}
818 
819 	/* Copy the permanent MAC address */
820 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
821 			&eth_dev->data->mac_addrs[0]);
822 
823 	/* initialize the vfta */
824 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
825 
826 	/* Now initialize the hardware */
827 	if (igb_hardware_init(hw) != 0) {
828 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
829 		rte_free(eth_dev->data->mac_addrs);
830 		eth_dev->data->mac_addrs = NULL;
831 		error = -ENODEV;
832 		goto err_late;
833 	}
834 	hw->mac.get_link_status = 1;
835 	adapter->stopped = 0;
836 
837 	/* Indicate SOL/IDER usage */
838 	if (e1000_check_reset_block(hw) < 0) {
839 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to"
840 					"SOL/IDER session");
841 	}
842 
843 	/* initialize PF if max_vfs not zero */
844 	igb_pf_host_init(eth_dev);
845 
846 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
847 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
848 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
849 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
850 	E1000_WRITE_FLUSH(hw);
851 
852 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
853 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
854 		     pci_dev->id.device_id);
855 
856 	rte_intr_callback_register(&pci_dev->intr_handle,
857 				   eth_igb_interrupt_handler,
858 				   (void *)eth_dev);
859 
860 	/* enable uio/vfio intr/eventfd mapping */
861 	rte_intr_enable(&pci_dev->intr_handle);
862 
863 	/* enable support intr */
864 	igb_intr_enable(eth_dev);
865 
866 	eth_igb_dev_set_link_down(eth_dev);
867 
868 	/* initialize filter info */
869 	memset(filter_info, 0,
870 	       sizeof(struct e1000_filter_info));
871 
872 	TAILQ_INIT(&filter_info->flex_list);
873 	TAILQ_INIT(&filter_info->twotuple_list);
874 	TAILQ_INIT(&filter_info->fivetuple_list);
875 
876 	TAILQ_INIT(&igb_filter_ntuple_list);
877 	TAILQ_INIT(&igb_filter_ethertype_list);
878 	TAILQ_INIT(&igb_filter_syn_list);
879 	TAILQ_INIT(&igb_filter_flex_list);
880 	TAILQ_INIT(&igb_filter_rss_list);
881 	TAILQ_INIT(&igb_flow_list);
882 
883 	return 0;
884 
885 err_late:
886 	igb_hw_control_release(hw);
887 
888 	return error;
889 }
890 
891 static int
892 eth_igb_dev_uninit(struct rte_eth_dev *eth_dev)
893 {
894 	PMD_INIT_FUNC_TRACE();
895 
896 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
897 		return 0;
898 
899 	eth_igb_close(eth_dev);
900 
901 	return 0;
902 }
903 
904 /*
905  * Virtual Function device init
906  */
907 static int
908 eth_igbvf_dev_init(struct rte_eth_dev *eth_dev)
909 {
910 	struct rte_pci_device *pci_dev;
911 	struct rte_intr_handle *intr_handle;
912 	struct e1000_adapter *adapter =
913 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
914 	struct e1000_hw *hw =
915 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
916 	int diag;
917 	struct rte_ether_addr *perm_addr =
918 		(struct rte_ether_addr *)hw->mac.perm_addr;
919 
920 	PMD_INIT_FUNC_TRACE();
921 
922 	eth_dev->dev_ops = &igbvf_eth_dev_ops;
923 	eth_dev->rx_descriptor_done   = eth_igb_rx_descriptor_done;
924 	eth_dev->rx_descriptor_status = eth_igb_rx_descriptor_status;
925 	eth_dev->tx_descriptor_status = eth_igb_tx_descriptor_status;
926 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
927 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
928 	eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
929 
930 	/* for secondary processes, we don't initialise any further as primary
931 	 * has already done this work. Only check we don't need a different
932 	 * RX function */
933 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
934 		if (eth_dev->data->scattered_rx)
935 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
936 		return 0;
937 	}
938 
939 	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
940 	rte_eth_copy_pci_info(eth_dev, pci_dev);
941 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
942 
943 	hw->device_id = pci_dev->id.device_id;
944 	hw->vendor_id = pci_dev->id.vendor_id;
945 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
946 	adapter->stopped = 0;
947 
948 	/* Initialize the shared code (base driver) */
949 	diag = e1000_setup_init_funcs(hw, TRUE);
950 	if (diag != 0) {
951 		PMD_INIT_LOG(ERR, "Shared code init failed for igbvf: %d",
952 			diag);
953 		return -EIO;
954 	}
955 
956 	/* init_mailbox_params */
957 	hw->mbx.ops.init_params(hw);
958 
959 	/* Disable the interrupts for VF */
960 	igbvf_intr_disable(hw);
961 
962 	diag = hw->mac.ops.reset_hw(hw);
963 
964 	/* Allocate memory for storing MAC addresses */
965 	eth_dev->data->mac_addrs = rte_zmalloc("igbvf", RTE_ETHER_ADDR_LEN *
966 		hw->mac.rar_entry_count, 0);
967 	if (eth_dev->data->mac_addrs == NULL) {
968 		PMD_INIT_LOG(ERR,
969 			"Failed to allocate %d bytes needed to store MAC "
970 			"addresses",
971 			RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
972 		return -ENOMEM;
973 	}
974 
975 	/* Generate a random MAC address, if none was assigned by PF. */
976 	if (rte_is_zero_ether_addr(perm_addr)) {
977 		rte_eth_random_addr(perm_addr->addr_bytes);
978 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
979 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
980 			     "%02x:%02x:%02x:%02x:%02x:%02x",
981 			     perm_addr->addr_bytes[0],
982 			     perm_addr->addr_bytes[1],
983 			     perm_addr->addr_bytes[2],
984 			     perm_addr->addr_bytes[3],
985 			     perm_addr->addr_bytes[4],
986 			     perm_addr->addr_bytes[5]);
987 	}
988 
989 	diag = e1000_rar_set(hw, perm_addr->addr_bytes, 0);
990 	if (diag) {
991 		rte_free(eth_dev->data->mac_addrs);
992 		eth_dev->data->mac_addrs = NULL;
993 		return diag;
994 	}
995 	/* Copy the permanent MAC address */
996 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.perm_addr,
997 			&eth_dev->data->mac_addrs[0]);
998 
999 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x "
1000 		     "mac.type=%s",
1001 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1002 		     pci_dev->id.device_id, "igb_mac_82576_vf");
1003 
1004 	intr_handle = &pci_dev->intr_handle;
1005 	rte_intr_callback_register(intr_handle,
1006 				   eth_igbvf_interrupt_handler, eth_dev);
1007 
1008 	return 0;
1009 }
1010 
1011 static int
1012 eth_igbvf_dev_uninit(struct rte_eth_dev *eth_dev)
1013 {
1014 	PMD_INIT_FUNC_TRACE();
1015 
1016 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1017 		return 0;
1018 
1019 	igbvf_dev_close(eth_dev);
1020 
1021 	return 0;
1022 }
1023 
1024 static int eth_igb_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1025 	struct rte_pci_device *pci_dev)
1026 {
1027 	return rte_eth_dev_pci_generic_probe(pci_dev,
1028 		sizeof(struct e1000_adapter), eth_igb_dev_init);
1029 }
1030 
1031 static int eth_igb_pci_remove(struct rte_pci_device *pci_dev)
1032 {
1033 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igb_dev_uninit);
1034 }
1035 
1036 static struct rte_pci_driver rte_igb_pmd = {
1037 	.id_table = pci_id_igb_map,
1038 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1039 	.probe = eth_igb_pci_probe,
1040 	.remove = eth_igb_pci_remove,
1041 };
1042 
1043 
1044 static int eth_igbvf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1045 	struct rte_pci_device *pci_dev)
1046 {
1047 	return rte_eth_dev_pci_generic_probe(pci_dev,
1048 		sizeof(struct e1000_adapter), eth_igbvf_dev_init);
1049 }
1050 
1051 static int eth_igbvf_pci_remove(struct rte_pci_device *pci_dev)
1052 {
1053 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igbvf_dev_uninit);
1054 }
1055 
1056 /*
1057  * virtual function driver struct
1058  */
1059 static struct rte_pci_driver rte_igbvf_pmd = {
1060 	.id_table = pci_id_igbvf_map,
1061 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1062 	.probe = eth_igbvf_pci_probe,
1063 	.remove = eth_igbvf_pci_remove,
1064 };
1065 
1066 static void
1067 igb_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1068 {
1069 	struct e1000_hw *hw =
1070 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1071 	/* RCTL: enable VLAN filter since VMDq always use VLAN filter */
1072 	uint32_t rctl = E1000_READ_REG(hw, E1000_RCTL);
1073 	rctl |= E1000_RCTL_VFE;
1074 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1075 }
1076 
1077 static int
1078 igb_check_mq_mode(struct rte_eth_dev *dev)
1079 {
1080 	enum rte_eth_rx_mq_mode rx_mq_mode = dev->data->dev_conf.rxmode.mq_mode;
1081 	enum rte_eth_tx_mq_mode tx_mq_mode = dev->data->dev_conf.txmode.mq_mode;
1082 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
1083 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
1084 
1085 	if ((rx_mq_mode & ETH_MQ_RX_DCB_FLAG) ||
1086 	    tx_mq_mode == ETH_MQ_TX_DCB ||
1087 	    tx_mq_mode == ETH_MQ_TX_VMDQ_DCB) {
1088 		PMD_INIT_LOG(ERR, "DCB mode is not supported.");
1089 		return -EINVAL;
1090 	}
1091 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
1092 		/* Check multi-queue mode.
1093 		 * To no break software we accept ETH_MQ_RX_NONE as this might
1094 		 * be used to turn off VLAN filter.
1095 		 */
1096 
1097 		if (rx_mq_mode == ETH_MQ_RX_NONE ||
1098 		    rx_mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1099 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
1100 			RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = 1;
1101 		} else {
1102 			/* Only support one queue on VFs.
1103 			 * RSS together with SRIOV is not supported.
1104 			 */
1105 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1106 					" wrong mq_mode rx %d.",
1107 					rx_mq_mode);
1108 			return -EINVAL;
1109 		}
1110 		/* TX mode is not used here, so mode might be ignored.*/
1111 		if (tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1112 			/* SRIOV only works in VMDq enable mode */
1113 			PMD_INIT_LOG(WARNING, "SRIOV is active,"
1114 					" TX mode %d is not supported. "
1115 					" Driver will behave as %d mode.",
1116 					tx_mq_mode, ETH_MQ_TX_VMDQ_ONLY);
1117 		}
1118 
1119 		/* check valid queue number */
1120 		if ((nb_rx_q > 1) || (nb_tx_q > 1)) {
1121 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1122 					" only support one queue on VFs.");
1123 			return -EINVAL;
1124 		}
1125 	} else {
1126 		/* To no break software that set invalid mode, only display
1127 		 * warning if invalid mode is used.
1128 		 */
1129 		if (rx_mq_mode != ETH_MQ_RX_NONE &&
1130 		    rx_mq_mode != ETH_MQ_RX_VMDQ_ONLY &&
1131 		    rx_mq_mode != ETH_MQ_RX_RSS) {
1132 			/* RSS together with VMDq not supported*/
1133 			PMD_INIT_LOG(ERR, "RX mode %d is not supported.",
1134 				     rx_mq_mode);
1135 			return -EINVAL;
1136 		}
1137 
1138 		if (tx_mq_mode != ETH_MQ_TX_NONE &&
1139 		    tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1140 			PMD_INIT_LOG(WARNING, "TX mode %d is not supported."
1141 					" Due to txmode is meaningless in this"
1142 					" driver, just ignore.",
1143 					tx_mq_mode);
1144 		}
1145 	}
1146 	return 0;
1147 }
1148 
1149 static int
1150 eth_igb_configure(struct rte_eth_dev *dev)
1151 {
1152 	struct e1000_interrupt *intr =
1153 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1154 	int ret;
1155 
1156 	PMD_INIT_FUNC_TRACE();
1157 
1158 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1159 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1160 
1161 	/* multipe queue mode checking */
1162 	ret  = igb_check_mq_mode(dev);
1163 	if (ret != 0) {
1164 		PMD_DRV_LOG(ERR, "igb_check_mq_mode fails with %d.",
1165 			    ret);
1166 		return ret;
1167 	}
1168 
1169 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1170 	PMD_INIT_FUNC_TRACE();
1171 
1172 	return 0;
1173 }
1174 
1175 static void
1176 eth_igb_rxtx_control(struct rte_eth_dev *dev,
1177 		     bool enable)
1178 {
1179 	struct e1000_hw *hw =
1180 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1181 	uint32_t tctl, rctl;
1182 
1183 	tctl = E1000_READ_REG(hw, E1000_TCTL);
1184 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1185 
1186 	if (enable) {
1187 		/* enable Tx/Rx */
1188 		tctl |= E1000_TCTL_EN;
1189 		rctl |= E1000_RCTL_EN;
1190 	} else {
1191 		/* disable Tx/Rx */
1192 		tctl &= ~E1000_TCTL_EN;
1193 		rctl &= ~E1000_RCTL_EN;
1194 	}
1195 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
1196 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1197 	E1000_WRITE_FLUSH(hw);
1198 }
1199 
1200 static int
1201 eth_igb_start(struct rte_eth_dev *dev)
1202 {
1203 	struct e1000_hw *hw =
1204 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1205 	struct e1000_adapter *adapter =
1206 		E1000_DEV_PRIVATE(dev->data->dev_private);
1207 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1208 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1209 	int ret, mask;
1210 	uint32_t intr_vector = 0;
1211 	uint32_t ctrl_ext;
1212 	uint32_t *speeds;
1213 	int num_speeds;
1214 	bool autoneg;
1215 
1216 	PMD_INIT_FUNC_TRACE();
1217 
1218 	/* disable uio/vfio intr/eventfd mapping */
1219 	rte_intr_disable(intr_handle);
1220 
1221 	/* Power up the phy. Needed to make the link go Up */
1222 	eth_igb_dev_set_link_up(dev);
1223 
1224 	/*
1225 	 * Packet Buffer Allocation (PBA)
1226 	 * Writing PBA sets the receive portion of the buffer
1227 	 * the remainder is used for the transmit buffer.
1228 	 */
1229 	if (hw->mac.type == e1000_82575) {
1230 		uint32_t pba;
1231 
1232 		pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
1233 		E1000_WRITE_REG(hw, E1000_PBA, pba);
1234 	}
1235 
1236 	/* Put the address into the Receive Address Array */
1237 	e1000_rar_set(hw, hw->mac.addr, 0);
1238 
1239 	/* Initialize the hardware */
1240 	if (igb_hardware_init(hw)) {
1241 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
1242 		return -EIO;
1243 	}
1244 	adapter->stopped = 0;
1245 
1246 	E1000_WRITE_REG(hw, E1000_VET,
1247 			RTE_ETHER_TYPE_VLAN << 16 | RTE_ETHER_TYPE_VLAN);
1248 
1249 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1250 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1251 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
1252 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
1253 	E1000_WRITE_FLUSH(hw);
1254 
1255 	/* configure PF module if SRIOV enabled */
1256 	igb_pf_host_configure(dev);
1257 
1258 	/* check and configure queue intr-vector mapping */
1259 	if ((rte_intr_cap_multiple(intr_handle) ||
1260 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
1261 	    dev->data->dev_conf.intr_conf.rxq != 0) {
1262 		intr_vector = dev->data->nb_rx_queues;
1263 		if (rte_intr_efd_enable(intr_handle, intr_vector))
1264 			return -1;
1265 	}
1266 
1267 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
1268 		intr_handle->intr_vec =
1269 			rte_zmalloc("intr_vec",
1270 				    dev->data->nb_rx_queues * sizeof(int), 0);
1271 		if (intr_handle->intr_vec == NULL) {
1272 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
1273 				     " intr_vec", dev->data->nb_rx_queues);
1274 			return -ENOMEM;
1275 		}
1276 	}
1277 
1278 	/* confiugre msix for rx interrupt */
1279 	eth_igb_configure_msix_intr(dev);
1280 
1281 	/* Configure for OS presence */
1282 	igb_init_manageability(hw);
1283 
1284 	eth_igb_tx_init(dev);
1285 
1286 	/* This can fail when allocating mbufs for descriptor rings */
1287 	ret = eth_igb_rx_init(dev);
1288 	if (ret) {
1289 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
1290 		igb_dev_clear_queues(dev);
1291 		return ret;
1292 	}
1293 
1294 	e1000_clear_hw_cntrs_base_generic(hw);
1295 
1296 	/*
1297 	 * VLAN Offload Settings
1298 	 */
1299 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
1300 			ETH_VLAN_EXTEND_MASK;
1301 	ret = eth_igb_vlan_offload_set(dev, mask);
1302 	if (ret) {
1303 		PMD_INIT_LOG(ERR, "Unable to set vlan offload");
1304 		igb_dev_clear_queues(dev);
1305 		return ret;
1306 	}
1307 
1308 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1309 		/* Enable VLAN filter since VMDq always use VLAN filter */
1310 		igb_vmdq_vlan_hw_filter_enable(dev);
1311 	}
1312 
1313 	if ((hw->mac.type == e1000_82576) || (hw->mac.type == e1000_82580) ||
1314 		(hw->mac.type == e1000_i350) || (hw->mac.type == e1000_i210) ||
1315 		(hw->mac.type == e1000_i211)) {
1316 		/* Configure EITR with the maximum possible value (0xFFFF) */
1317 		E1000_WRITE_REG(hw, E1000_EITR(0), 0xFFFF);
1318 	}
1319 
1320 	/* Setup link speed and duplex */
1321 	speeds = &dev->data->dev_conf.link_speeds;
1322 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
1323 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
1324 		hw->mac.autoneg = 1;
1325 	} else {
1326 		num_speeds = 0;
1327 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
1328 
1329 		/* Reset */
1330 		hw->phy.autoneg_advertised = 0;
1331 
1332 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1333 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1334 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
1335 			num_speeds = -1;
1336 			goto error_invalid_config;
1337 		}
1338 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
1339 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1340 			num_speeds++;
1341 		}
1342 		if (*speeds & ETH_LINK_SPEED_10M) {
1343 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1344 			num_speeds++;
1345 		}
1346 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
1347 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1348 			num_speeds++;
1349 		}
1350 		if (*speeds & ETH_LINK_SPEED_100M) {
1351 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1352 			num_speeds++;
1353 		}
1354 		if (*speeds & ETH_LINK_SPEED_1G) {
1355 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1356 			num_speeds++;
1357 		}
1358 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
1359 			goto error_invalid_config;
1360 
1361 		/* Set/reset the mac.autoneg based on the link speed,
1362 		 * fixed or not
1363 		 */
1364 		if (!autoneg) {
1365 			hw->mac.autoneg = 0;
1366 			hw->mac.forced_speed_duplex =
1367 					hw->phy.autoneg_advertised;
1368 		} else {
1369 			hw->mac.autoneg = 1;
1370 		}
1371 	}
1372 
1373 	e1000_setup_link(hw);
1374 
1375 	if (rte_intr_allow_others(intr_handle)) {
1376 		/* check if lsc interrupt is enabled */
1377 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1378 			eth_igb_lsc_interrupt_setup(dev, TRUE);
1379 		else
1380 			eth_igb_lsc_interrupt_setup(dev, FALSE);
1381 	} else {
1382 		rte_intr_callback_unregister(intr_handle,
1383 					     eth_igb_interrupt_handler,
1384 					     (void *)dev);
1385 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1386 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
1387 				     " no intr multiplex");
1388 	}
1389 
1390 	/* check if rxq interrupt is enabled */
1391 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
1392 	    rte_intr_dp_is_en(intr_handle))
1393 		eth_igb_rxq_interrupt_setup(dev);
1394 
1395 	/* enable uio/vfio intr/eventfd mapping */
1396 	rte_intr_enable(intr_handle);
1397 
1398 	/* resume enabled intr since hw reset */
1399 	igb_intr_enable(dev);
1400 
1401 	/* restore all types filter */
1402 	igb_filter_restore(dev);
1403 
1404 	eth_igb_rxtx_control(dev, true);
1405 	eth_igb_link_update(dev, 0);
1406 
1407 	PMD_INIT_LOG(DEBUG, "<<");
1408 
1409 	return 0;
1410 
1411 error_invalid_config:
1412 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
1413 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
1414 	igb_dev_clear_queues(dev);
1415 	return -EINVAL;
1416 }
1417 
1418 /*********************************************************************
1419  *
1420  *  This routine disables all traffic on the adapter by issuing a
1421  *  global reset on the MAC.
1422  *
1423  **********************************************************************/
1424 static int
1425 eth_igb_stop(struct rte_eth_dev *dev)
1426 {
1427 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1428 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1429 	struct rte_eth_link link;
1430 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1431 	struct e1000_adapter *adapter =
1432 		E1000_DEV_PRIVATE(dev->data->dev_private);
1433 
1434 	if (adapter->stopped)
1435 		return 0;
1436 
1437 	eth_igb_rxtx_control(dev, false);
1438 
1439 	igb_intr_disable(dev);
1440 
1441 	/* disable intr eventfd mapping */
1442 	rte_intr_disable(intr_handle);
1443 
1444 	igb_pf_reset_hw(hw);
1445 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1446 
1447 	/* Set bit for Go Link disconnect if PHY reset is not blocked */
1448 	if (hw->mac.type >= e1000_82580 &&
1449 	    (e1000_check_reset_block(hw) != E1000_BLK_PHY_RESET)) {
1450 		uint32_t phpm_reg;
1451 
1452 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1453 		phpm_reg |= E1000_82580_PM_GO_LINKD;
1454 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1455 	}
1456 
1457 	/* Power down the phy. Needed to make the link go Down */
1458 	eth_igb_dev_set_link_down(dev);
1459 
1460 	igb_dev_clear_queues(dev);
1461 
1462 	/* clear the recorded link status */
1463 	memset(&link, 0, sizeof(link));
1464 	rte_eth_linkstatus_set(dev, &link);
1465 
1466 	if (!rte_intr_allow_others(intr_handle))
1467 		/* resume to the default handler */
1468 		rte_intr_callback_register(intr_handle,
1469 					   eth_igb_interrupt_handler,
1470 					   (void *)dev);
1471 
1472 	/* Clean datapath event and queue/vec mapping */
1473 	rte_intr_efd_disable(intr_handle);
1474 	if (intr_handle->intr_vec != NULL) {
1475 		rte_free(intr_handle->intr_vec);
1476 		intr_handle->intr_vec = NULL;
1477 	}
1478 
1479 	adapter->stopped = true;
1480 	dev->data->dev_started = 0;
1481 
1482 	return 0;
1483 }
1484 
1485 static int
1486 eth_igb_dev_set_link_up(struct rte_eth_dev *dev)
1487 {
1488 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1489 
1490 	if (hw->phy.media_type == e1000_media_type_copper)
1491 		e1000_power_up_phy(hw);
1492 	else
1493 		e1000_power_up_fiber_serdes_link(hw);
1494 
1495 	return 0;
1496 }
1497 
1498 static int
1499 eth_igb_dev_set_link_down(struct rte_eth_dev *dev)
1500 {
1501 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1502 
1503 	if (hw->phy.media_type == e1000_media_type_copper)
1504 		e1000_power_down_phy(hw);
1505 	else
1506 		e1000_shutdown_fiber_serdes_link(hw);
1507 
1508 	return 0;
1509 }
1510 
1511 static int
1512 eth_igb_close(struct rte_eth_dev *dev)
1513 {
1514 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1515 	struct rte_eth_link link;
1516 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1517 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1518 	struct e1000_filter_info *filter_info =
1519 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
1520 	int ret;
1521 
1522 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1523 		return 0;
1524 
1525 	ret = eth_igb_stop(dev);
1526 
1527 	e1000_phy_hw_reset(hw);
1528 	igb_release_manageability(hw);
1529 	igb_hw_control_release(hw);
1530 
1531 	/* Clear bit for Go Link disconnect if PHY reset is not blocked */
1532 	if (hw->mac.type >= e1000_82580 &&
1533 	    (e1000_check_reset_block(hw) != E1000_BLK_PHY_RESET)) {
1534 		uint32_t phpm_reg;
1535 
1536 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1537 		phpm_reg &= ~E1000_82580_PM_GO_LINKD;
1538 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1539 	}
1540 
1541 	igb_dev_free_queues(dev);
1542 
1543 	if (intr_handle->intr_vec) {
1544 		rte_free(intr_handle->intr_vec);
1545 		intr_handle->intr_vec = NULL;
1546 	}
1547 
1548 	memset(&link, 0, sizeof(link));
1549 	rte_eth_linkstatus_set(dev, &link);
1550 
1551 	/* Reset any pending lock */
1552 	igb_reset_swfw_lock(hw);
1553 
1554 	/* uninitialize PF if max_vfs not zero */
1555 	igb_pf_host_uninit(dev);
1556 
1557 	rte_intr_callback_unregister(intr_handle,
1558 				     eth_igb_interrupt_handler, dev);
1559 
1560 	/* clear the SYN filter info */
1561 	filter_info->syn_info = 0;
1562 
1563 	/* clear the ethertype filters info */
1564 	filter_info->ethertype_mask = 0;
1565 	memset(filter_info->ethertype_filters, 0,
1566 		E1000_MAX_ETQF_FILTERS * sizeof(struct igb_ethertype_filter));
1567 
1568 	/* clear the rss filter info */
1569 	memset(&filter_info->rss_info, 0,
1570 		sizeof(struct igb_rte_flow_rss_conf));
1571 
1572 	/* remove all ntuple filters of the device */
1573 	igb_ntuple_filter_uninit(dev);
1574 
1575 	/* remove all flex filters of the device */
1576 	igb_flex_filter_uninit(dev);
1577 
1578 	/* clear all the filters list */
1579 	igb_filterlist_flush(dev);
1580 
1581 	return ret;
1582 }
1583 
1584 /*
1585  * Reset PF device.
1586  */
1587 static int
1588 eth_igb_reset(struct rte_eth_dev *dev)
1589 {
1590 	int ret;
1591 
1592 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
1593 	 * its VF to make them align with it. The detailed notification
1594 	 * mechanism is PMD specific and is currently not implemented.
1595 	 * To avoid unexpected behavior in VF, currently reset of PF with
1596 	 * SR-IOV activation is not supported. It might be supported later.
1597 	 */
1598 	if (dev->data->sriov.active)
1599 		return -ENOTSUP;
1600 
1601 	ret = eth_igb_dev_uninit(dev);
1602 	if (ret)
1603 		return ret;
1604 
1605 	ret = eth_igb_dev_init(dev);
1606 
1607 	return ret;
1608 }
1609 
1610 
1611 static int
1612 igb_get_rx_buffer_size(struct e1000_hw *hw)
1613 {
1614 	uint32_t rx_buf_size;
1615 	if (hw->mac.type == e1000_82576) {
1616 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xffff) << 10;
1617 	} else if (hw->mac.type == e1000_82580 || hw->mac.type == e1000_i350) {
1618 		/* PBS needs to be translated according to a lookup table */
1619 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xf);
1620 		rx_buf_size = (uint32_t) e1000_rxpbs_adjust_82580(rx_buf_size);
1621 		rx_buf_size = (rx_buf_size << 10);
1622 	} else if (hw->mac.type == e1000_i210 || hw->mac.type == e1000_i211) {
1623 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0x3f) << 10;
1624 	} else {
1625 		rx_buf_size = (E1000_READ_REG(hw, E1000_PBA) & 0xffff) << 10;
1626 	}
1627 
1628 	return rx_buf_size;
1629 }
1630 
1631 /*********************************************************************
1632  *
1633  *  Initialize the hardware
1634  *
1635  **********************************************************************/
1636 static int
1637 igb_hardware_init(struct e1000_hw *hw)
1638 {
1639 	uint32_t rx_buf_size;
1640 	int diag;
1641 
1642 	/* Let the firmware know the OS is in control */
1643 	igb_hw_control_acquire(hw);
1644 
1645 	/*
1646 	 * These parameters control the automatic generation (Tx) and
1647 	 * response (Rx) to Ethernet PAUSE frames.
1648 	 * - High water mark should allow for at least two standard size (1518)
1649 	 *   frames to be received after sending an XOFF.
1650 	 * - Low water mark works best when it is very near the high water mark.
1651 	 *   This allows the receiver to restart by sending XON when it has
1652 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
1653 	 *   restart after one full frame is pulled from the buffer. There
1654 	 *   could be several smaller frames in the buffer and if so they will
1655 	 *   not trigger the XON until their total number reduces the buffer
1656 	 *   by 1500.
1657 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
1658 	 */
1659 	rx_buf_size = igb_get_rx_buffer_size(hw);
1660 
1661 	hw->fc.high_water = rx_buf_size - (RTE_ETHER_MAX_LEN * 2);
1662 	hw->fc.low_water = hw->fc.high_water - 1500;
1663 	hw->fc.pause_time = IGB_FC_PAUSE_TIME;
1664 	hw->fc.send_xon = 1;
1665 
1666 	/* Set Flow control, use the tunable location if sane */
1667 	if ((igb_fc_setting != e1000_fc_none) && (igb_fc_setting < 4))
1668 		hw->fc.requested_mode = igb_fc_setting;
1669 	else
1670 		hw->fc.requested_mode = e1000_fc_none;
1671 
1672 	/* Issue a global reset */
1673 	igb_pf_reset_hw(hw);
1674 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1675 
1676 	diag = e1000_init_hw(hw);
1677 	if (diag < 0)
1678 		return diag;
1679 
1680 	E1000_WRITE_REG(hw, E1000_VET,
1681 			RTE_ETHER_TYPE_VLAN << 16 | RTE_ETHER_TYPE_VLAN);
1682 	e1000_get_phy_info(hw);
1683 	e1000_check_for_link(hw);
1684 
1685 	return 0;
1686 }
1687 
1688 /* This function is based on igb_update_stats_counters() in igb/if_igb.c */
1689 static void
1690 igb_read_stats_registers(struct e1000_hw *hw, struct e1000_hw_stats *stats)
1691 {
1692 	int pause_frames;
1693 
1694 	uint64_t old_gprc  = stats->gprc;
1695 	uint64_t old_gptc  = stats->gptc;
1696 	uint64_t old_tpr   = stats->tpr;
1697 	uint64_t old_tpt   = stats->tpt;
1698 	uint64_t old_rpthc = stats->rpthc;
1699 	uint64_t old_hgptc = stats->hgptc;
1700 
1701 	if(hw->phy.media_type == e1000_media_type_copper ||
1702 	    (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
1703 		stats->symerrs +=
1704 		    E1000_READ_REG(hw,E1000_SYMERRS);
1705 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
1706 	}
1707 
1708 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
1709 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
1710 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
1711 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
1712 
1713 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
1714 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
1715 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
1716 	stats->dc += E1000_READ_REG(hw, E1000_DC);
1717 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
1718 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
1719 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
1720 	/*
1721 	** For watchdog management we need to know if we have been
1722 	** paused during the last interval, so capture that here.
1723 	*/
1724 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
1725 	stats->xoffrxc += pause_frames;
1726 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
1727 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
1728 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
1729 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
1730 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
1731 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
1732 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
1733 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
1734 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
1735 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
1736 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
1737 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
1738 
1739 	/* For the 64-bit byte counters the low dword must be read first. */
1740 	/* Both registers clear on the read of the high dword */
1741 
1742 	/* Workaround CRC bytes included in size, take away 4 bytes/packet */
1743 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
1744 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
1745 	stats->gorc -= (stats->gprc - old_gprc) * RTE_ETHER_CRC_LEN;
1746 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
1747 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
1748 	stats->gotc -= (stats->gptc - old_gptc) * RTE_ETHER_CRC_LEN;
1749 
1750 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
1751 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
1752 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
1753 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
1754 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
1755 
1756 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
1757 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
1758 
1759 	stats->tor += E1000_READ_REG(hw, E1000_TORL);
1760 	stats->tor += ((uint64_t)E1000_READ_REG(hw, E1000_TORH) << 32);
1761 	stats->tor -= (stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
1762 	stats->tot += E1000_READ_REG(hw, E1000_TOTL);
1763 	stats->tot += ((uint64_t)E1000_READ_REG(hw, E1000_TOTH) << 32);
1764 	stats->tot -= (stats->tpt - old_tpt) * RTE_ETHER_CRC_LEN;
1765 
1766 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
1767 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
1768 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
1769 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
1770 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
1771 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
1772 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
1773 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
1774 
1775 	/* Interrupt Counts */
1776 
1777 	stats->iac += E1000_READ_REG(hw, E1000_IAC);
1778 	stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
1779 	stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
1780 	stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
1781 	stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
1782 	stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
1783 	stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
1784 	stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
1785 	stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
1786 
1787 	/* Host to Card Statistics */
1788 
1789 	stats->cbtmpc += E1000_READ_REG(hw, E1000_CBTMPC);
1790 	stats->htdpmc += E1000_READ_REG(hw, E1000_HTDPMC);
1791 	stats->cbrdpc += E1000_READ_REG(hw, E1000_CBRDPC);
1792 	stats->cbrmpc += E1000_READ_REG(hw, E1000_CBRMPC);
1793 	stats->rpthc += E1000_READ_REG(hw, E1000_RPTHC);
1794 	stats->hgptc += E1000_READ_REG(hw, E1000_HGPTC);
1795 	stats->htcbdpc += E1000_READ_REG(hw, E1000_HTCBDPC);
1796 	stats->hgorc += E1000_READ_REG(hw, E1000_HGORCL);
1797 	stats->hgorc += ((uint64_t)E1000_READ_REG(hw, E1000_HGORCH) << 32);
1798 	stats->hgorc -= (stats->rpthc - old_rpthc) * RTE_ETHER_CRC_LEN;
1799 	stats->hgotc += E1000_READ_REG(hw, E1000_HGOTCL);
1800 	stats->hgotc += ((uint64_t)E1000_READ_REG(hw, E1000_HGOTCH) << 32);
1801 	stats->hgotc -= (stats->hgptc - old_hgptc) * RTE_ETHER_CRC_LEN;
1802 	stats->lenerrs += E1000_READ_REG(hw, E1000_LENERRS);
1803 	stats->scvpc += E1000_READ_REG(hw, E1000_SCVPC);
1804 	stats->hrmpc += E1000_READ_REG(hw, E1000_HRMPC);
1805 
1806 	stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
1807 	stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
1808 	stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
1809 	stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
1810 	stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
1811 	stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
1812 }
1813 
1814 static int
1815 eth_igb_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1816 {
1817 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1818 	struct e1000_hw_stats *stats =
1819 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1820 
1821 	igb_read_stats_registers(hw, stats);
1822 
1823 	if (rte_stats == NULL)
1824 		return -EINVAL;
1825 
1826 	/* Rx Errors */
1827 	rte_stats->imissed = stats->mpc;
1828 	rte_stats->ierrors = stats->crcerrs + stats->rlec +
1829 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
1830 
1831 	/* Tx Errors */
1832 	rte_stats->oerrors = stats->ecol + stats->latecol;
1833 
1834 	rte_stats->ipackets = stats->gprc;
1835 	rte_stats->opackets = stats->gptc;
1836 	rte_stats->ibytes   = stats->gorc;
1837 	rte_stats->obytes   = stats->gotc;
1838 	return 0;
1839 }
1840 
1841 static int
1842 eth_igb_stats_reset(struct rte_eth_dev *dev)
1843 {
1844 	struct e1000_hw_stats *hw_stats =
1845 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1846 
1847 	/* HW registers are cleared on read */
1848 	eth_igb_stats_get(dev, NULL);
1849 
1850 	/* Reset software totals */
1851 	memset(hw_stats, 0, sizeof(*hw_stats));
1852 
1853 	return 0;
1854 }
1855 
1856 static int
1857 eth_igb_xstats_reset(struct rte_eth_dev *dev)
1858 {
1859 	struct e1000_hw_stats *stats =
1860 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1861 
1862 	/* HW registers are cleared on read */
1863 	eth_igb_xstats_get(dev, NULL, IGB_NB_XSTATS);
1864 
1865 	/* Reset software totals */
1866 	memset(stats, 0, sizeof(*stats));
1867 
1868 	return 0;
1869 }
1870 
1871 static int eth_igb_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1872 	struct rte_eth_xstat_name *xstats_names,
1873 	__rte_unused unsigned int size)
1874 {
1875 	unsigned i;
1876 
1877 	if (xstats_names == NULL)
1878 		return IGB_NB_XSTATS;
1879 
1880 	/* Note: limit checked in rte_eth_xstats_names() */
1881 
1882 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1883 		strlcpy(xstats_names[i].name, rte_igb_stats_strings[i].name,
1884 			sizeof(xstats_names[i].name));
1885 	}
1886 
1887 	return IGB_NB_XSTATS;
1888 }
1889 
1890 static int eth_igb_xstats_get_names_by_id(struct rte_eth_dev *dev,
1891 		struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
1892 		unsigned int limit)
1893 {
1894 	unsigned int i;
1895 
1896 	if (!ids) {
1897 		if (xstats_names == NULL)
1898 			return IGB_NB_XSTATS;
1899 
1900 		for (i = 0; i < IGB_NB_XSTATS; i++)
1901 			strlcpy(xstats_names[i].name,
1902 				rte_igb_stats_strings[i].name,
1903 				sizeof(xstats_names[i].name));
1904 
1905 		return IGB_NB_XSTATS;
1906 
1907 	} else {
1908 		struct rte_eth_xstat_name xstats_names_copy[IGB_NB_XSTATS];
1909 
1910 		eth_igb_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
1911 				IGB_NB_XSTATS);
1912 
1913 		for (i = 0; i < limit; i++) {
1914 			if (ids[i] >= IGB_NB_XSTATS) {
1915 				PMD_INIT_LOG(ERR, "id value isn't valid");
1916 				return -1;
1917 			}
1918 			strcpy(xstats_names[i].name,
1919 					xstats_names_copy[ids[i]].name);
1920 		}
1921 		return limit;
1922 	}
1923 }
1924 
1925 static int
1926 eth_igb_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1927 		   unsigned n)
1928 {
1929 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1930 	struct e1000_hw_stats *hw_stats =
1931 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1932 	unsigned i;
1933 
1934 	if (n < IGB_NB_XSTATS)
1935 		return IGB_NB_XSTATS;
1936 
1937 	igb_read_stats_registers(hw, hw_stats);
1938 
1939 	/* If this is a reset xstats is NULL, and we have cleared the
1940 	 * registers by reading them.
1941 	 */
1942 	if (!xstats)
1943 		return 0;
1944 
1945 	/* Extended stats */
1946 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1947 		xstats[i].id = i;
1948 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1949 			rte_igb_stats_strings[i].offset);
1950 	}
1951 
1952 	return IGB_NB_XSTATS;
1953 }
1954 
1955 static int
1956 eth_igb_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
1957 		uint64_t *values, unsigned int n)
1958 {
1959 	unsigned int i;
1960 
1961 	if (!ids) {
1962 		struct e1000_hw *hw =
1963 			E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1964 		struct e1000_hw_stats *hw_stats =
1965 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1966 
1967 		if (n < IGB_NB_XSTATS)
1968 			return IGB_NB_XSTATS;
1969 
1970 		igb_read_stats_registers(hw, hw_stats);
1971 
1972 		/* If this is a reset xstats is NULL, and we have cleared the
1973 		 * registers by reading them.
1974 		 */
1975 		if (!values)
1976 			return 0;
1977 
1978 		/* Extended stats */
1979 		for (i = 0; i < IGB_NB_XSTATS; i++)
1980 			values[i] = *(uint64_t *)(((char *)hw_stats) +
1981 					rte_igb_stats_strings[i].offset);
1982 
1983 		return IGB_NB_XSTATS;
1984 
1985 	} else {
1986 		uint64_t values_copy[IGB_NB_XSTATS];
1987 
1988 		eth_igb_xstats_get_by_id(dev, NULL, values_copy,
1989 				IGB_NB_XSTATS);
1990 
1991 		for (i = 0; i < n; i++) {
1992 			if (ids[i] >= IGB_NB_XSTATS) {
1993 				PMD_INIT_LOG(ERR, "id value isn't valid");
1994 				return -1;
1995 			}
1996 			values[i] = values_copy[ids[i]];
1997 		}
1998 		return n;
1999 	}
2000 }
2001 
2002 static void
2003 igbvf_read_stats_registers(struct e1000_hw *hw, struct e1000_vf_stats *hw_stats)
2004 {
2005 	/* Good Rx packets, include VF loopback */
2006 	UPDATE_VF_STAT(E1000_VFGPRC,
2007 	    hw_stats->last_gprc, hw_stats->gprc);
2008 
2009 	/* Good Rx octets, include VF loopback */
2010 	UPDATE_VF_STAT(E1000_VFGORC,
2011 	    hw_stats->last_gorc, hw_stats->gorc);
2012 
2013 	/* Good Tx packets, include VF loopback */
2014 	UPDATE_VF_STAT(E1000_VFGPTC,
2015 	    hw_stats->last_gptc, hw_stats->gptc);
2016 
2017 	/* Good Tx octets, include VF loopback */
2018 	UPDATE_VF_STAT(E1000_VFGOTC,
2019 	    hw_stats->last_gotc, hw_stats->gotc);
2020 
2021 	/* Rx Multicst packets */
2022 	UPDATE_VF_STAT(E1000_VFMPRC,
2023 	    hw_stats->last_mprc, hw_stats->mprc);
2024 
2025 	/* Good Rx loopback packets */
2026 	UPDATE_VF_STAT(E1000_VFGPRLBC,
2027 	    hw_stats->last_gprlbc, hw_stats->gprlbc);
2028 
2029 	/* Good Rx loopback octets */
2030 	UPDATE_VF_STAT(E1000_VFGORLBC,
2031 	    hw_stats->last_gorlbc, hw_stats->gorlbc);
2032 
2033 	/* Good Tx loopback packets */
2034 	UPDATE_VF_STAT(E1000_VFGPTLBC,
2035 	    hw_stats->last_gptlbc, hw_stats->gptlbc);
2036 
2037 	/* Good Tx loopback octets */
2038 	UPDATE_VF_STAT(E1000_VFGOTLBC,
2039 	    hw_stats->last_gotlbc, hw_stats->gotlbc);
2040 }
2041 
2042 static int eth_igbvf_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
2043 				     struct rte_eth_xstat_name *xstats_names,
2044 				     __rte_unused unsigned limit)
2045 {
2046 	unsigned i;
2047 
2048 	if (xstats_names != NULL)
2049 		for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2050 			strlcpy(xstats_names[i].name,
2051 				rte_igbvf_stats_strings[i].name,
2052 				sizeof(xstats_names[i].name));
2053 		}
2054 	return IGBVF_NB_XSTATS;
2055 }
2056 
2057 static int
2058 eth_igbvf_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
2059 		     unsigned n)
2060 {
2061 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2062 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2063 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2064 	unsigned i;
2065 
2066 	if (n < IGBVF_NB_XSTATS)
2067 		return IGBVF_NB_XSTATS;
2068 
2069 	igbvf_read_stats_registers(hw, hw_stats);
2070 
2071 	if (!xstats)
2072 		return 0;
2073 
2074 	for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2075 		xstats[i].id = i;
2076 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
2077 			rte_igbvf_stats_strings[i].offset);
2078 	}
2079 
2080 	return IGBVF_NB_XSTATS;
2081 }
2082 
2083 static int
2084 eth_igbvf_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
2085 {
2086 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2087 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2088 			  E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2089 
2090 	igbvf_read_stats_registers(hw, hw_stats);
2091 
2092 	if (rte_stats == NULL)
2093 		return -EINVAL;
2094 
2095 	rte_stats->ipackets = hw_stats->gprc;
2096 	rte_stats->ibytes = hw_stats->gorc;
2097 	rte_stats->opackets = hw_stats->gptc;
2098 	rte_stats->obytes = hw_stats->gotc;
2099 	return 0;
2100 }
2101 
2102 static int
2103 eth_igbvf_stats_reset(struct rte_eth_dev *dev)
2104 {
2105 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats*)
2106 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2107 
2108 	/* Sync HW register to the last stats */
2109 	eth_igbvf_stats_get(dev, NULL);
2110 
2111 	/* reset HW current stats*/
2112 	memset(&hw_stats->gprc, 0, sizeof(*hw_stats) -
2113 	       offsetof(struct e1000_vf_stats, gprc));
2114 
2115 	return 0;
2116 }
2117 
2118 static int
2119 eth_igb_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
2120 		       size_t fw_size)
2121 {
2122 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2123 	struct e1000_fw_version fw;
2124 	int ret;
2125 
2126 	e1000_get_fw_version(hw, &fw);
2127 
2128 	switch (hw->mac.type) {
2129 	case e1000_i210:
2130 	case e1000_i211:
2131 		if (!(e1000_get_flash_presence_i210(hw))) {
2132 			ret = snprintf(fw_version, fw_size,
2133 				 "%2d.%2d-%d",
2134 				 fw.invm_major, fw.invm_minor,
2135 				 fw.invm_img_type);
2136 			break;
2137 		}
2138 		/* fall through */
2139 	default:
2140 		/* if option rom is valid, display its version too */
2141 		if (fw.or_valid) {
2142 			ret = snprintf(fw_version, fw_size,
2143 				 "%d.%d, 0x%08x, %d.%d.%d",
2144 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
2145 				 fw.or_major, fw.or_build, fw.or_patch);
2146 		/* no option rom */
2147 		} else {
2148 			if (fw.etrack_id != 0X0000) {
2149 				ret = snprintf(fw_version, fw_size,
2150 					 "%d.%d, 0x%08x",
2151 					 fw.eep_major, fw.eep_minor,
2152 					 fw.etrack_id);
2153 			} else {
2154 				ret = snprintf(fw_version, fw_size,
2155 					 "%d.%d.%d",
2156 					 fw.eep_major, fw.eep_minor,
2157 					 fw.eep_build);
2158 			}
2159 		}
2160 		break;
2161 	}
2162 
2163 	ret += 1; /* add the size of '\0' */
2164 	if (fw_size < (u32)ret)
2165 		return ret;
2166 	else
2167 		return 0;
2168 }
2169 
2170 static int
2171 eth_igb_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2172 {
2173 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2174 
2175 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2176 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2177 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2178 	dev_info->rx_queue_offload_capa = igb_get_rx_queue_offloads_capa(dev);
2179 	dev_info->rx_offload_capa = igb_get_rx_port_offloads_capa(dev) |
2180 				    dev_info->rx_queue_offload_capa;
2181 	dev_info->tx_queue_offload_capa = igb_get_tx_queue_offloads_capa(dev);
2182 	dev_info->tx_offload_capa = igb_get_tx_port_offloads_capa(dev) |
2183 				    dev_info->tx_queue_offload_capa;
2184 
2185 	switch (hw->mac.type) {
2186 	case e1000_82575:
2187 		dev_info->max_rx_queues = 4;
2188 		dev_info->max_tx_queues = 4;
2189 		dev_info->max_vmdq_pools = 0;
2190 		break;
2191 
2192 	case e1000_82576:
2193 		dev_info->max_rx_queues = 16;
2194 		dev_info->max_tx_queues = 16;
2195 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2196 		dev_info->vmdq_queue_num = 16;
2197 		break;
2198 
2199 	case e1000_82580:
2200 		dev_info->max_rx_queues = 8;
2201 		dev_info->max_tx_queues = 8;
2202 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2203 		dev_info->vmdq_queue_num = 8;
2204 		break;
2205 
2206 	case e1000_i350:
2207 		dev_info->max_rx_queues = 8;
2208 		dev_info->max_tx_queues = 8;
2209 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2210 		dev_info->vmdq_queue_num = 8;
2211 		break;
2212 
2213 	case e1000_i354:
2214 		dev_info->max_rx_queues = 8;
2215 		dev_info->max_tx_queues = 8;
2216 		break;
2217 
2218 	case e1000_i210:
2219 		dev_info->max_rx_queues = 4;
2220 		dev_info->max_tx_queues = 4;
2221 		dev_info->max_vmdq_pools = 0;
2222 		break;
2223 
2224 	case e1000_i211:
2225 		dev_info->max_rx_queues = 2;
2226 		dev_info->max_tx_queues = 2;
2227 		dev_info->max_vmdq_pools = 0;
2228 		break;
2229 
2230 	default:
2231 		/* Should not happen */
2232 		return -EINVAL;
2233 	}
2234 	dev_info->hash_key_size = IGB_HKEY_MAX_INDEX * sizeof(uint32_t);
2235 	dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
2236 	dev_info->flow_type_rss_offloads = IGB_RSS_OFFLOAD_ALL;
2237 
2238 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2239 		.rx_thresh = {
2240 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2241 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2242 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2243 		},
2244 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2245 		.rx_drop_en = 0,
2246 		.offloads = 0,
2247 	};
2248 
2249 	dev_info->default_txconf = (struct rte_eth_txconf) {
2250 		.tx_thresh = {
2251 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2252 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2253 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2254 		},
2255 		.offloads = 0,
2256 	};
2257 
2258 	dev_info->rx_desc_lim = rx_desc_lim;
2259 	dev_info->tx_desc_lim = tx_desc_lim;
2260 
2261 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
2262 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
2263 			ETH_LINK_SPEED_1G;
2264 
2265 	dev_info->max_mtu = dev_info->max_rx_pktlen - E1000_ETH_OVERHEAD;
2266 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
2267 
2268 	return 0;
2269 }
2270 
2271 static const uint32_t *
2272 eth_igb_supported_ptypes_get(struct rte_eth_dev *dev)
2273 {
2274 	static const uint32_t ptypes[] = {
2275 		/* refers to igb_rxd_pkt_info_to_pkt_type() */
2276 		RTE_PTYPE_L2_ETHER,
2277 		RTE_PTYPE_L3_IPV4,
2278 		RTE_PTYPE_L3_IPV4_EXT,
2279 		RTE_PTYPE_L3_IPV6,
2280 		RTE_PTYPE_L3_IPV6_EXT,
2281 		RTE_PTYPE_L4_TCP,
2282 		RTE_PTYPE_L4_UDP,
2283 		RTE_PTYPE_L4_SCTP,
2284 		RTE_PTYPE_TUNNEL_IP,
2285 		RTE_PTYPE_INNER_L3_IPV6,
2286 		RTE_PTYPE_INNER_L3_IPV6_EXT,
2287 		RTE_PTYPE_INNER_L4_TCP,
2288 		RTE_PTYPE_INNER_L4_UDP,
2289 		RTE_PTYPE_UNKNOWN
2290 	};
2291 
2292 	if (dev->rx_pkt_burst == eth_igb_recv_pkts ||
2293 	    dev->rx_pkt_burst == eth_igb_recv_scattered_pkts)
2294 		return ptypes;
2295 	return NULL;
2296 }
2297 
2298 static int
2299 eth_igbvf_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2300 {
2301 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2302 
2303 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2304 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2305 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2306 	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
2307 				DEV_TX_OFFLOAD_IPV4_CKSUM  |
2308 				DEV_TX_OFFLOAD_UDP_CKSUM   |
2309 				DEV_TX_OFFLOAD_TCP_CKSUM   |
2310 				DEV_TX_OFFLOAD_SCTP_CKSUM  |
2311 				DEV_TX_OFFLOAD_TCP_TSO;
2312 	switch (hw->mac.type) {
2313 	case e1000_vfadapt:
2314 		dev_info->max_rx_queues = 2;
2315 		dev_info->max_tx_queues = 2;
2316 		break;
2317 	case e1000_vfadapt_i350:
2318 		dev_info->max_rx_queues = 1;
2319 		dev_info->max_tx_queues = 1;
2320 		break;
2321 	default:
2322 		/* Should not happen */
2323 		return -EINVAL;
2324 	}
2325 
2326 	dev_info->rx_queue_offload_capa = igb_get_rx_queue_offloads_capa(dev);
2327 	dev_info->rx_offload_capa = igb_get_rx_port_offloads_capa(dev) |
2328 				    dev_info->rx_queue_offload_capa;
2329 	dev_info->tx_queue_offload_capa = igb_get_tx_queue_offloads_capa(dev);
2330 	dev_info->tx_offload_capa = igb_get_tx_port_offloads_capa(dev) |
2331 				    dev_info->tx_queue_offload_capa;
2332 
2333 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2334 		.rx_thresh = {
2335 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2336 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2337 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2338 		},
2339 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2340 		.rx_drop_en = 0,
2341 		.offloads = 0,
2342 	};
2343 
2344 	dev_info->default_txconf = (struct rte_eth_txconf) {
2345 		.tx_thresh = {
2346 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2347 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2348 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2349 		},
2350 		.offloads = 0,
2351 	};
2352 
2353 	dev_info->rx_desc_lim = rx_desc_lim;
2354 	dev_info->tx_desc_lim = tx_desc_lim;
2355 
2356 	return 0;
2357 }
2358 
2359 /* return 0 means link status changed, -1 means not changed */
2360 static int
2361 eth_igb_link_update(struct rte_eth_dev *dev, int wait_to_complete)
2362 {
2363 	struct e1000_hw *hw =
2364 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2365 	struct rte_eth_link link;
2366 	int link_check, count;
2367 
2368 	link_check = 0;
2369 	hw->mac.get_link_status = 1;
2370 
2371 	/* possible wait-to-complete in up to 9 seconds */
2372 	for (count = 0; count < IGB_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
2373 		/* Read the real link status */
2374 		switch (hw->phy.media_type) {
2375 		case e1000_media_type_copper:
2376 			/* Do the work to read phy */
2377 			e1000_check_for_link(hw);
2378 			link_check = !hw->mac.get_link_status;
2379 			break;
2380 
2381 		case e1000_media_type_fiber:
2382 			e1000_check_for_link(hw);
2383 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
2384 				      E1000_STATUS_LU);
2385 			break;
2386 
2387 		case e1000_media_type_internal_serdes:
2388 			e1000_check_for_link(hw);
2389 			link_check = hw->mac.serdes_has_link;
2390 			break;
2391 
2392 		/* VF device is type_unknown */
2393 		case e1000_media_type_unknown:
2394 			eth_igbvf_link_update(hw);
2395 			link_check = !hw->mac.get_link_status;
2396 			break;
2397 
2398 		default:
2399 			break;
2400 		}
2401 		if (link_check || wait_to_complete == 0)
2402 			break;
2403 		rte_delay_ms(IGB_LINK_UPDATE_CHECK_INTERVAL);
2404 	}
2405 	memset(&link, 0, sizeof(link));
2406 
2407 	/* Now we check if a transition has happened */
2408 	if (link_check) {
2409 		uint16_t duplex, speed;
2410 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
2411 		link.link_duplex = (duplex == FULL_DUPLEX) ?
2412 				ETH_LINK_FULL_DUPLEX :
2413 				ETH_LINK_HALF_DUPLEX;
2414 		link.link_speed = speed;
2415 		link.link_status = ETH_LINK_UP;
2416 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
2417 				ETH_LINK_SPEED_FIXED);
2418 	} else if (!link_check) {
2419 		link.link_speed = 0;
2420 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
2421 		link.link_status = ETH_LINK_DOWN;
2422 		link.link_autoneg = ETH_LINK_FIXED;
2423 	}
2424 
2425 	return rte_eth_linkstatus_set(dev, &link);
2426 }
2427 
2428 /*
2429  * igb_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
2430  * For ASF and Pass Through versions of f/w this means
2431  * that the driver is loaded.
2432  */
2433 static void
2434 igb_hw_control_acquire(struct e1000_hw *hw)
2435 {
2436 	uint32_t ctrl_ext;
2437 
2438 	/* Let firmware know the driver has taken over */
2439 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2440 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2441 }
2442 
2443 /*
2444  * igb_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
2445  * For ASF and Pass Through versions of f/w this means that the
2446  * driver is no longer loaded.
2447  */
2448 static void
2449 igb_hw_control_release(struct e1000_hw *hw)
2450 {
2451 	uint32_t ctrl_ext;
2452 
2453 	/* Let firmware taken over control of h/w */
2454 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2455 	E1000_WRITE_REG(hw, E1000_CTRL_EXT,
2456 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2457 }
2458 
2459 /*
2460  * Bit of a misnomer, what this really means is
2461  * to enable OS management of the system... aka
2462  * to disable special hardware management features.
2463  */
2464 static void
2465 igb_init_manageability(struct e1000_hw *hw)
2466 {
2467 	if (e1000_enable_mng_pass_thru(hw)) {
2468 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
2469 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2470 
2471 		/* disable hardware interception of ARP */
2472 		manc &= ~(E1000_MANC_ARP_EN);
2473 
2474 		/* enable receiving management packets to the host */
2475 		manc |= E1000_MANC_EN_MNG2HOST;
2476 		manc2h |= 1 << 5;  /* Mng Port 623 */
2477 		manc2h |= 1 << 6;  /* Mng Port 664 */
2478 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
2479 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2480 	}
2481 }
2482 
2483 static void
2484 igb_release_manageability(struct e1000_hw *hw)
2485 {
2486 	if (e1000_enable_mng_pass_thru(hw)) {
2487 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2488 
2489 		manc |= E1000_MANC_ARP_EN;
2490 		manc &= ~E1000_MANC_EN_MNG2HOST;
2491 
2492 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2493 	}
2494 }
2495 
2496 static int
2497 eth_igb_promiscuous_enable(struct rte_eth_dev *dev)
2498 {
2499 	struct e1000_hw *hw =
2500 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2501 	uint32_t rctl;
2502 
2503 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2504 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
2505 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2506 
2507 	return 0;
2508 }
2509 
2510 static int
2511 eth_igb_promiscuous_disable(struct rte_eth_dev *dev)
2512 {
2513 	struct e1000_hw *hw =
2514 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2515 	uint32_t rctl;
2516 
2517 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2518 	rctl &= (~E1000_RCTL_UPE);
2519 	if (dev->data->all_multicast == 1)
2520 		rctl |= E1000_RCTL_MPE;
2521 	else
2522 		rctl &= (~E1000_RCTL_MPE);
2523 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2524 
2525 	return 0;
2526 }
2527 
2528 static int
2529 eth_igb_allmulticast_enable(struct rte_eth_dev *dev)
2530 {
2531 	struct e1000_hw *hw =
2532 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2533 	uint32_t rctl;
2534 
2535 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2536 	rctl |= E1000_RCTL_MPE;
2537 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2538 
2539 	return 0;
2540 }
2541 
2542 static int
2543 eth_igb_allmulticast_disable(struct rte_eth_dev *dev)
2544 {
2545 	struct e1000_hw *hw =
2546 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2547 	uint32_t rctl;
2548 
2549 	if (dev->data->promiscuous == 1)
2550 		return 0; /* must remain in all_multicast mode */
2551 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2552 	rctl &= (~E1000_RCTL_MPE);
2553 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2554 
2555 	return 0;
2556 }
2557 
2558 static int
2559 eth_igb_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2560 {
2561 	struct e1000_hw *hw =
2562 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2563 	struct e1000_vfta * shadow_vfta =
2564 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2565 	uint32_t vfta;
2566 	uint32_t vid_idx;
2567 	uint32_t vid_bit;
2568 
2569 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
2570 			      E1000_VFTA_ENTRY_MASK);
2571 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
2572 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
2573 	if (on)
2574 		vfta |= vid_bit;
2575 	else
2576 		vfta &= ~vid_bit;
2577 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
2578 
2579 	/* update local VFTA copy */
2580 	shadow_vfta->vfta[vid_idx] = vfta;
2581 
2582 	return 0;
2583 }
2584 
2585 static int
2586 eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
2587 		      enum rte_vlan_type vlan_type,
2588 		      uint16_t tpid)
2589 {
2590 	struct e1000_hw *hw =
2591 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2592 	uint32_t reg, qinq;
2593 
2594 	qinq = E1000_READ_REG(hw, E1000_CTRL_EXT);
2595 	qinq &= E1000_CTRL_EXT_EXT_VLAN;
2596 
2597 	/* only outer TPID of double VLAN can be configured*/
2598 	if (qinq && vlan_type == ETH_VLAN_TYPE_OUTER) {
2599 		reg = E1000_READ_REG(hw, E1000_VET);
2600 		reg = (reg & (~E1000_VET_VET_EXT)) |
2601 			((uint32_t)tpid << E1000_VET_VET_EXT_SHIFT);
2602 		E1000_WRITE_REG(hw, E1000_VET, reg);
2603 
2604 		return 0;
2605 	}
2606 
2607 	/* all other TPID values are read-only*/
2608 	PMD_DRV_LOG(ERR, "Not supported");
2609 
2610 	return -ENOTSUP;
2611 }
2612 
2613 static void
2614 igb_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2615 {
2616 	struct e1000_hw *hw =
2617 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2618 	uint32_t reg;
2619 
2620 	/* Filter Table Disable */
2621 	reg = E1000_READ_REG(hw, E1000_RCTL);
2622 	reg &= ~E1000_RCTL_CFIEN;
2623 	reg &= ~E1000_RCTL_VFE;
2624 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2625 }
2626 
2627 static void
2628 igb_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2629 {
2630 	struct e1000_hw *hw =
2631 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2632 	struct e1000_vfta * shadow_vfta =
2633 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2634 	uint32_t reg;
2635 	int i;
2636 
2637 	/* Filter Table Enable, CFI not used for packet acceptance */
2638 	reg = E1000_READ_REG(hw, E1000_RCTL);
2639 	reg &= ~E1000_RCTL_CFIEN;
2640 	reg |= E1000_RCTL_VFE;
2641 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2642 
2643 	/* restore VFTA table */
2644 	for (i = 0; i < IGB_VFTA_SIZE; i++)
2645 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
2646 }
2647 
2648 static void
2649 igb_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2650 {
2651 	struct e1000_hw *hw =
2652 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2653 	uint32_t reg;
2654 
2655 	/* VLAN Mode Disable */
2656 	reg = E1000_READ_REG(hw, E1000_CTRL);
2657 	reg &= ~E1000_CTRL_VME;
2658 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2659 }
2660 
2661 static void
2662 igb_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2663 {
2664 	struct e1000_hw *hw =
2665 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2666 	uint32_t reg;
2667 
2668 	/* VLAN Mode Enable */
2669 	reg = E1000_READ_REG(hw, E1000_CTRL);
2670 	reg |= E1000_CTRL_VME;
2671 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2672 }
2673 
2674 static void
2675 igb_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2676 {
2677 	struct e1000_hw *hw =
2678 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2679 	uint32_t reg;
2680 
2681 	/* CTRL_EXT: Extended VLAN */
2682 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2683 	reg &= ~E1000_CTRL_EXT_EXTEND_VLAN;
2684 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2685 
2686 	/* Update maximum packet length */
2687 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME)
2688 		E1000_WRITE_REG(hw, E1000_RLPML,
2689 				dev->data->dev_conf.rxmode.max_rx_pkt_len);
2690 }
2691 
2692 static void
2693 igb_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2694 {
2695 	struct e1000_hw *hw =
2696 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2697 	uint32_t reg;
2698 
2699 	/* CTRL_EXT: Extended VLAN */
2700 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2701 	reg |= E1000_CTRL_EXT_EXTEND_VLAN;
2702 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2703 
2704 	/* Update maximum packet length */
2705 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME)
2706 		E1000_WRITE_REG(hw, E1000_RLPML,
2707 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2708 						VLAN_TAG_SIZE);
2709 }
2710 
2711 static int
2712 eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2713 {
2714 	struct rte_eth_rxmode *rxmode;
2715 
2716 	rxmode = &dev->data->dev_conf.rxmode;
2717 	if(mask & ETH_VLAN_STRIP_MASK){
2718 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2719 			igb_vlan_hw_strip_enable(dev);
2720 		else
2721 			igb_vlan_hw_strip_disable(dev);
2722 	}
2723 
2724 	if(mask & ETH_VLAN_FILTER_MASK){
2725 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2726 			igb_vlan_hw_filter_enable(dev);
2727 		else
2728 			igb_vlan_hw_filter_disable(dev);
2729 	}
2730 
2731 	if(mask & ETH_VLAN_EXTEND_MASK){
2732 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2733 			igb_vlan_hw_extend_enable(dev);
2734 		else
2735 			igb_vlan_hw_extend_disable(dev);
2736 	}
2737 
2738 	return 0;
2739 }
2740 
2741 
2742 /**
2743  * It enables the interrupt mask and then enable the interrupt.
2744  *
2745  * @param dev
2746  *  Pointer to struct rte_eth_dev.
2747  * @param on
2748  *  Enable or Disable
2749  *
2750  * @return
2751  *  - On success, zero.
2752  *  - On failure, a negative value.
2753  */
2754 static int
2755 eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
2756 {
2757 	struct e1000_interrupt *intr =
2758 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2759 
2760 	if (on)
2761 		intr->mask |= E1000_ICR_LSC;
2762 	else
2763 		intr->mask &= ~E1000_ICR_LSC;
2764 
2765 	return 0;
2766 }
2767 
2768 /* It clears the interrupt causes and enables the interrupt.
2769  * It will be called once only during nic initialized.
2770  *
2771  * @param dev
2772  *  Pointer to struct rte_eth_dev.
2773  *
2774  * @return
2775  *  - On success, zero.
2776  *  - On failure, a negative value.
2777  */
2778 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev)
2779 {
2780 	uint32_t mask, regval;
2781 	int ret;
2782 	struct e1000_hw *hw =
2783 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2784 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2785 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2786 	int misc_shift = rte_intr_allow_others(intr_handle) ? 1 : 0;
2787 	struct rte_eth_dev_info dev_info;
2788 
2789 	memset(&dev_info, 0, sizeof(dev_info));
2790 	ret = eth_igb_infos_get(dev, &dev_info);
2791 	if (ret != 0)
2792 		return ret;
2793 
2794 	mask = (0xFFFFFFFF >> (32 - dev_info.max_rx_queues)) << misc_shift;
2795 	regval = E1000_READ_REG(hw, E1000_EIMS);
2796 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
2797 
2798 	return 0;
2799 }
2800 
2801 /*
2802  * It reads ICR and gets interrupt causes, check it and set a bit flag
2803  * to update link status.
2804  *
2805  * @param dev
2806  *  Pointer to struct rte_eth_dev.
2807  *
2808  * @return
2809  *  - On success, zero.
2810  *  - On failure, a negative value.
2811  */
2812 static int
2813 eth_igb_interrupt_get_status(struct rte_eth_dev *dev)
2814 {
2815 	uint32_t icr;
2816 	struct e1000_hw *hw =
2817 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2818 	struct e1000_interrupt *intr =
2819 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2820 
2821 	igb_intr_disable(dev);
2822 
2823 	/* read-on-clear nic registers here */
2824 	icr = E1000_READ_REG(hw, E1000_ICR);
2825 
2826 	intr->flags = 0;
2827 	if (icr & E1000_ICR_LSC) {
2828 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
2829 	}
2830 
2831 	if (icr & E1000_ICR_VMMB)
2832 		intr->flags |= E1000_FLAG_MAILBOX;
2833 
2834 	return 0;
2835 }
2836 
2837 /*
2838  * It executes link_update after knowing an interrupt is prsent.
2839  *
2840  * @param dev
2841  *  Pointer to struct rte_eth_dev.
2842  *
2843  * @return
2844  *  - On success, zero.
2845  *  - On failure, a negative value.
2846  */
2847 static int
2848 eth_igb_interrupt_action(struct rte_eth_dev *dev,
2849 			 struct rte_intr_handle *intr_handle)
2850 {
2851 	struct e1000_hw *hw =
2852 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2853 	struct e1000_interrupt *intr =
2854 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2855 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2856 	struct rte_eth_link link;
2857 	int ret;
2858 
2859 	if (intr->flags & E1000_FLAG_MAILBOX) {
2860 		igb_pf_mbx_process(dev);
2861 		intr->flags &= ~E1000_FLAG_MAILBOX;
2862 	}
2863 
2864 	igb_intr_enable(dev);
2865 	rte_intr_ack(intr_handle);
2866 
2867 	if (intr->flags & E1000_FLAG_NEED_LINK_UPDATE) {
2868 		intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
2869 
2870 		/* set get_link_status to check register later */
2871 		hw->mac.get_link_status = 1;
2872 		ret = eth_igb_link_update(dev, 0);
2873 
2874 		/* check if link has changed */
2875 		if (ret < 0)
2876 			return 0;
2877 
2878 		rte_eth_linkstatus_get(dev, &link);
2879 		if (link.link_status) {
2880 			PMD_INIT_LOG(INFO,
2881 				     " Port %d: Link Up - speed %u Mbps - %s",
2882 				     dev->data->port_id,
2883 				     (unsigned)link.link_speed,
2884 				     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
2885 				     "full-duplex" : "half-duplex");
2886 		} else {
2887 			PMD_INIT_LOG(INFO, " Port %d: Link Down",
2888 				     dev->data->port_id);
2889 		}
2890 
2891 		PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
2892 			     pci_dev->addr.domain,
2893 			     pci_dev->addr.bus,
2894 			     pci_dev->addr.devid,
2895 			     pci_dev->addr.function);
2896 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
2897 	}
2898 
2899 	return 0;
2900 }
2901 
2902 /**
2903  * Interrupt handler which shall be registered at first.
2904  *
2905  * @param handle
2906  *  Pointer to interrupt handle.
2907  * @param param
2908  *  The address of parameter (struct rte_eth_dev *) regsitered before.
2909  *
2910  * @return
2911  *  void
2912  */
2913 static void
2914 eth_igb_interrupt_handler(void *param)
2915 {
2916 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2917 
2918 	eth_igb_interrupt_get_status(dev);
2919 	eth_igb_interrupt_action(dev, dev->intr_handle);
2920 }
2921 
2922 static int
2923 eth_igbvf_interrupt_get_status(struct rte_eth_dev *dev)
2924 {
2925 	uint32_t eicr;
2926 	struct e1000_hw *hw =
2927 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2928 	struct e1000_interrupt *intr =
2929 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2930 
2931 	igbvf_intr_disable(hw);
2932 
2933 	/* read-on-clear nic registers here */
2934 	eicr = E1000_READ_REG(hw, E1000_EICR);
2935 	intr->flags = 0;
2936 
2937 	if (eicr == E1000_VTIVAR_MISC_MAILBOX)
2938 		intr->flags |= E1000_FLAG_MAILBOX;
2939 
2940 	return 0;
2941 }
2942 
2943 void igbvf_mbx_process(struct rte_eth_dev *dev)
2944 {
2945 	struct e1000_hw *hw =
2946 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2947 	struct e1000_mbx_info *mbx = &hw->mbx;
2948 	u32 in_msg = 0;
2949 
2950 	/* peek the message first */
2951 	in_msg = E1000_READ_REG(hw, E1000_VMBMEM(0));
2952 
2953 	/* PF reset VF event */
2954 	if (in_msg == E1000_PF_CONTROL_MSG) {
2955 		/* dummy mbx read to ack pf */
2956 		if (mbx->ops.read(hw, &in_msg, 1, 0))
2957 			return;
2958 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
2959 					     NULL);
2960 	}
2961 }
2962 
2963 static int
2964 eth_igbvf_interrupt_action(struct rte_eth_dev *dev, struct rte_intr_handle *intr_handle)
2965 {
2966 	struct e1000_interrupt *intr =
2967 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2968 
2969 	if (intr->flags & E1000_FLAG_MAILBOX) {
2970 		igbvf_mbx_process(dev);
2971 		intr->flags &= ~E1000_FLAG_MAILBOX;
2972 	}
2973 
2974 	igbvf_intr_enable(dev);
2975 	rte_intr_ack(intr_handle);
2976 
2977 	return 0;
2978 }
2979 
2980 static void
2981 eth_igbvf_interrupt_handler(void *param)
2982 {
2983 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2984 
2985 	eth_igbvf_interrupt_get_status(dev);
2986 	eth_igbvf_interrupt_action(dev, dev->intr_handle);
2987 }
2988 
2989 static int
2990 eth_igb_led_on(struct rte_eth_dev *dev)
2991 {
2992 	struct e1000_hw *hw;
2993 
2994 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2995 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
2996 }
2997 
2998 static int
2999 eth_igb_led_off(struct rte_eth_dev *dev)
3000 {
3001 	struct e1000_hw *hw;
3002 
3003 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3004 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3005 }
3006 
3007 static int
3008 eth_igb_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3009 {
3010 	struct e1000_hw *hw;
3011 	uint32_t ctrl;
3012 	int tx_pause;
3013 	int rx_pause;
3014 
3015 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3016 	fc_conf->pause_time = hw->fc.pause_time;
3017 	fc_conf->high_water = hw->fc.high_water;
3018 	fc_conf->low_water = hw->fc.low_water;
3019 	fc_conf->send_xon = hw->fc.send_xon;
3020 	fc_conf->autoneg = hw->mac.autoneg;
3021 
3022 	/*
3023 	 * Return rx_pause and tx_pause status according to actual setting of
3024 	 * the TFCE and RFCE bits in the CTRL register.
3025 	 */
3026 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
3027 	if (ctrl & E1000_CTRL_TFCE)
3028 		tx_pause = 1;
3029 	else
3030 		tx_pause = 0;
3031 
3032 	if (ctrl & E1000_CTRL_RFCE)
3033 		rx_pause = 1;
3034 	else
3035 		rx_pause = 0;
3036 
3037 	if (rx_pause && tx_pause)
3038 		fc_conf->mode = RTE_FC_FULL;
3039 	else if (rx_pause)
3040 		fc_conf->mode = RTE_FC_RX_PAUSE;
3041 	else if (tx_pause)
3042 		fc_conf->mode = RTE_FC_TX_PAUSE;
3043 	else
3044 		fc_conf->mode = RTE_FC_NONE;
3045 
3046 	return 0;
3047 }
3048 
3049 static int
3050 eth_igb_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3051 {
3052 	struct e1000_hw *hw;
3053 	int err;
3054 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
3055 		e1000_fc_none,
3056 		e1000_fc_rx_pause,
3057 		e1000_fc_tx_pause,
3058 		e1000_fc_full
3059 	};
3060 	uint32_t rx_buf_size;
3061 	uint32_t max_high_water;
3062 	uint32_t rctl;
3063 	uint32_t ctrl;
3064 
3065 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3066 	if (fc_conf->autoneg != hw->mac.autoneg)
3067 		return -ENOTSUP;
3068 	rx_buf_size = igb_get_rx_buffer_size(hw);
3069 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3070 
3071 	/* At least reserve one Ethernet frame for watermark */
3072 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
3073 	if ((fc_conf->high_water > max_high_water) ||
3074 	    (fc_conf->high_water < fc_conf->low_water)) {
3075 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
3076 		PMD_INIT_LOG(ERR, "high water must <=  0x%x", max_high_water);
3077 		return -EINVAL;
3078 	}
3079 
3080 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
3081 	hw->fc.pause_time     = fc_conf->pause_time;
3082 	hw->fc.high_water     = fc_conf->high_water;
3083 	hw->fc.low_water      = fc_conf->low_water;
3084 	hw->fc.send_xon	      = fc_conf->send_xon;
3085 
3086 	err = e1000_setup_link_generic(hw);
3087 	if (err == E1000_SUCCESS) {
3088 
3089 		/* check if we want to forward MAC frames - driver doesn't have native
3090 		 * capability to do that, so we'll write the registers ourselves */
3091 
3092 		rctl = E1000_READ_REG(hw, E1000_RCTL);
3093 
3094 		/* set or clear MFLCN.PMCF bit depending on configuration */
3095 		if (fc_conf->mac_ctrl_frame_fwd != 0)
3096 			rctl |= E1000_RCTL_PMCF;
3097 		else
3098 			rctl &= ~E1000_RCTL_PMCF;
3099 
3100 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3101 
3102 		/*
3103 		 * check if we want to change flow control mode - driver doesn't have native
3104 		 * capability to do that, so we'll write the registers ourselves
3105 		 */
3106 		ctrl = E1000_READ_REG(hw, E1000_CTRL);
3107 
3108 		/*
3109 		 * set or clear E1000_CTRL_RFCE and E1000_CTRL_TFCE bits depending
3110 		 * on configuration
3111 		 */
3112 		switch (fc_conf->mode) {
3113 		case RTE_FC_NONE:
3114 			ctrl &= ~E1000_CTRL_RFCE & ~E1000_CTRL_TFCE;
3115 			break;
3116 		case RTE_FC_RX_PAUSE:
3117 			ctrl |= E1000_CTRL_RFCE;
3118 			ctrl &= ~E1000_CTRL_TFCE;
3119 			break;
3120 		case RTE_FC_TX_PAUSE:
3121 			ctrl |= E1000_CTRL_TFCE;
3122 			ctrl &= ~E1000_CTRL_RFCE;
3123 			break;
3124 		case RTE_FC_FULL:
3125 			ctrl |= E1000_CTRL_RFCE | E1000_CTRL_TFCE;
3126 			break;
3127 		default:
3128 			PMD_INIT_LOG(ERR, "invalid flow control mode");
3129 			return -EINVAL;
3130 		}
3131 
3132 		E1000_WRITE_REG(hw, E1000_CTRL, ctrl);
3133 
3134 		E1000_WRITE_FLUSH(hw);
3135 
3136 		return 0;
3137 	}
3138 
3139 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
3140 	return -EIO;
3141 }
3142 
3143 #define E1000_RAH_POOLSEL_SHIFT      (18)
3144 static int
3145 eth_igb_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
3146 		uint32_t index, uint32_t pool)
3147 {
3148 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3149 	uint32_t rah;
3150 
3151 	e1000_rar_set(hw, mac_addr->addr_bytes, index);
3152 	rah = E1000_READ_REG(hw, E1000_RAH(index));
3153 	rah |= (0x1 << (E1000_RAH_POOLSEL_SHIFT + pool));
3154 	E1000_WRITE_REG(hw, E1000_RAH(index), rah);
3155 	return 0;
3156 }
3157 
3158 static void
3159 eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index)
3160 {
3161 	uint8_t addr[RTE_ETHER_ADDR_LEN];
3162 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3163 
3164 	memset(addr, 0, sizeof(addr));
3165 
3166 	e1000_rar_set(hw, addr, index);
3167 }
3168 
3169 static int
3170 eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
3171 				struct rte_ether_addr *addr)
3172 {
3173 	eth_igb_rar_clear(dev, 0);
3174 	eth_igb_rar_set(dev, (void *)addr, 0, 0);
3175 
3176 	return 0;
3177 }
3178 /*
3179  * Virtual Function operations
3180  */
3181 static void
3182 igbvf_intr_disable(struct e1000_hw *hw)
3183 {
3184 	PMD_INIT_FUNC_TRACE();
3185 
3186 	/* Clear interrupt mask to stop from interrupts being generated */
3187 	E1000_WRITE_REG(hw, E1000_EIMC, 0xFFFF);
3188 
3189 	E1000_WRITE_FLUSH(hw);
3190 }
3191 
3192 static void
3193 igbvf_stop_adapter(struct rte_eth_dev *dev)
3194 {
3195 	u32 reg_val;
3196 	u16 i;
3197 	struct rte_eth_dev_info dev_info;
3198 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3199 	int ret;
3200 
3201 	memset(&dev_info, 0, sizeof(dev_info));
3202 	ret = eth_igbvf_infos_get(dev, &dev_info);
3203 	if (ret != 0)
3204 		return;
3205 
3206 	/* Clear interrupt mask to stop from interrupts being generated */
3207 	igbvf_intr_disable(hw);
3208 
3209 	/* Clear any pending interrupts, flush previous writes */
3210 	E1000_READ_REG(hw, E1000_EICR);
3211 
3212 	/* Disable the transmit unit.  Each queue must be disabled. */
3213 	for (i = 0; i < dev_info.max_tx_queues; i++)
3214 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), E1000_TXDCTL_SWFLSH);
3215 
3216 	/* Disable the receive unit by stopping each queue */
3217 	for (i = 0; i < dev_info.max_rx_queues; i++) {
3218 		reg_val = E1000_READ_REG(hw, E1000_RXDCTL(i));
3219 		reg_val &= ~E1000_RXDCTL_QUEUE_ENABLE;
3220 		E1000_WRITE_REG(hw, E1000_RXDCTL(i), reg_val);
3221 		while (E1000_READ_REG(hw, E1000_RXDCTL(i)) & E1000_RXDCTL_QUEUE_ENABLE)
3222 			;
3223 	}
3224 
3225 	/* flush all queues disables */
3226 	E1000_WRITE_FLUSH(hw);
3227 	msec_delay(2);
3228 }
3229 
3230 static int eth_igbvf_link_update(struct e1000_hw *hw)
3231 {
3232 	struct e1000_mbx_info *mbx = &hw->mbx;
3233 	struct e1000_mac_info *mac = &hw->mac;
3234 	int ret_val = E1000_SUCCESS;
3235 
3236 	PMD_INIT_LOG(DEBUG, "e1000_check_for_link_vf");
3237 
3238 	/*
3239 	 * We only want to run this if there has been a rst asserted.
3240 	 * in this case that could mean a link change, device reset,
3241 	 * or a virtual function reset
3242 	 */
3243 
3244 	/* If we were hit with a reset or timeout drop the link */
3245 	if (!e1000_check_for_rst(hw, 0) || !mbx->timeout)
3246 		mac->get_link_status = TRUE;
3247 
3248 	if (!mac->get_link_status)
3249 		goto out;
3250 
3251 	/* if link status is down no point in checking to see if pf is up */
3252 	if (!(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU))
3253 		goto out;
3254 
3255 	/* if we passed all the tests above then the link is up and we no
3256 	 * longer need to check for link */
3257 	mac->get_link_status = FALSE;
3258 
3259 out:
3260 	return ret_val;
3261 }
3262 
3263 
3264 static int
3265 igbvf_dev_configure(struct rte_eth_dev *dev)
3266 {
3267 	struct rte_eth_conf* conf = &dev->data->dev_conf;
3268 
3269 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
3270 		     dev->data->port_id);
3271 
3272 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
3273 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
3274 
3275 	/*
3276 	 * VF has no ability to enable/disable HW CRC
3277 	 * Keep the persistent behavior the same as Host PF
3278 	 */
3279 #ifndef RTE_LIBRTE_E1000_PF_DISABLE_STRIP_CRC
3280 	if (conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
3281 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
3282 		conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_KEEP_CRC;
3283 	}
3284 #else
3285 	if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC)) {
3286 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
3287 		conf->rxmode.offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
3288 	}
3289 #endif
3290 
3291 	return 0;
3292 }
3293 
3294 static int
3295 igbvf_dev_start(struct rte_eth_dev *dev)
3296 {
3297 	struct e1000_hw *hw =
3298 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3299 	struct e1000_adapter *adapter =
3300 		E1000_DEV_PRIVATE(dev->data->dev_private);
3301 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3302 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3303 	int ret;
3304 	uint32_t intr_vector = 0;
3305 
3306 	PMD_INIT_FUNC_TRACE();
3307 
3308 	hw->mac.ops.reset_hw(hw);
3309 	adapter->stopped = 0;
3310 
3311 	/* Set all vfta */
3312 	igbvf_set_vfta_all(dev,1);
3313 
3314 	eth_igbvf_tx_init(dev);
3315 
3316 	/* This can fail when allocating mbufs for descriptor rings */
3317 	ret = eth_igbvf_rx_init(dev);
3318 	if (ret) {
3319 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
3320 		igb_dev_clear_queues(dev);
3321 		return ret;
3322 	}
3323 
3324 	/* check and configure queue intr-vector mapping */
3325 	if (rte_intr_cap_multiple(intr_handle) &&
3326 	    dev->data->dev_conf.intr_conf.rxq) {
3327 		intr_vector = dev->data->nb_rx_queues;
3328 		ret = rte_intr_efd_enable(intr_handle, intr_vector);
3329 		if (ret)
3330 			return ret;
3331 	}
3332 
3333 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
3334 		intr_handle->intr_vec =
3335 			rte_zmalloc("intr_vec",
3336 				    dev->data->nb_rx_queues * sizeof(int), 0);
3337 		if (!intr_handle->intr_vec) {
3338 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
3339 				     " intr_vec", dev->data->nb_rx_queues);
3340 			return -ENOMEM;
3341 		}
3342 	}
3343 
3344 	eth_igbvf_configure_msix_intr(dev);
3345 
3346 	/* enable uio/vfio intr/eventfd mapping */
3347 	rte_intr_enable(intr_handle);
3348 
3349 	/* resume enabled intr since hw reset */
3350 	igbvf_intr_enable(dev);
3351 
3352 	return 0;
3353 }
3354 
3355 static int
3356 igbvf_dev_stop(struct rte_eth_dev *dev)
3357 {
3358 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3359 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3360 	struct e1000_adapter *adapter =
3361 		E1000_DEV_PRIVATE(dev->data->dev_private);
3362 
3363 	if (adapter->stopped)
3364 		return 0;
3365 
3366 	PMD_INIT_FUNC_TRACE();
3367 
3368 	igbvf_stop_adapter(dev);
3369 
3370 	/*
3371 	  * Clear what we set, but we still keep shadow_vfta to
3372 	  * restore after device starts
3373 	  */
3374 	igbvf_set_vfta_all(dev,0);
3375 
3376 	igb_dev_clear_queues(dev);
3377 
3378 	/* disable intr eventfd mapping */
3379 	rte_intr_disable(intr_handle);
3380 
3381 	/* Clean datapath event and queue/vec mapping */
3382 	rte_intr_efd_disable(intr_handle);
3383 	if (intr_handle->intr_vec) {
3384 		rte_free(intr_handle->intr_vec);
3385 		intr_handle->intr_vec = NULL;
3386 	}
3387 
3388 	adapter->stopped = true;
3389 	dev->data->dev_started = 0;
3390 
3391 	return 0;
3392 }
3393 
3394 static int
3395 igbvf_dev_close(struct rte_eth_dev *dev)
3396 {
3397 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3398 	struct rte_ether_addr addr;
3399 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3400 	int ret;
3401 
3402 	PMD_INIT_FUNC_TRACE();
3403 
3404 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
3405 		return 0;
3406 
3407 	e1000_reset_hw(hw);
3408 
3409 	ret = igbvf_dev_stop(dev);
3410 	if (ret != 0)
3411 		return ret;
3412 
3413 	igb_dev_free_queues(dev);
3414 
3415 	/**
3416 	 * reprogram the RAR with a zero mac address,
3417 	 * to ensure that the VF traffic goes to the PF
3418 	 * after stop, close and detach of the VF.
3419 	 **/
3420 
3421 	memset(&addr, 0, sizeof(addr));
3422 	igbvf_default_mac_addr_set(dev, &addr);
3423 
3424 	rte_intr_callback_unregister(&pci_dev->intr_handle,
3425 				     eth_igbvf_interrupt_handler,
3426 				     (void *)dev);
3427 
3428 	return 0;
3429 }
3430 
3431 static int
3432 igbvf_promiscuous_enable(struct rte_eth_dev *dev)
3433 {
3434 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3435 
3436 	/* Set both unicast and multicast promisc */
3437 	e1000_promisc_set_vf(hw, e1000_promisc_enabled);
3438 
3439 	return 0;
3440 }
3441 
3442 static int
3443 igbvf_promiscuous_disable(struct rte_eth_dev *dev)
3444 {
3445 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3446 
3447 	/* If in allmulticast mode leave multicast promisc */
3448 	if (dev->data->all_multicast == 1)
3449 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3450 	else
3451 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3452 
3453 	return 0;
3454 }
3455 
3456 static int
3457 igbvf_allmulticast_enable(struct rte_eth_dev *dev)
3458 {
3459 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3460 
3461 	/* In promiscuous mode multicast promisc already set */
3462 	if (dev->data->promiscuous == 0)
3463 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3464 
3465 	return 0;
3466 }
3467 
3468 static int
3469 igbvf_allmulticast_disable(struct rte_eth_dev *dev)
3470 {
3471 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3472 
3473 	/* In promiscuous mode leave multicast promisc enabled */
3474 	if (dev->data->promiscuous == 0)
3475 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3476 
3477 	return 0;
3478 }
3479 
3480 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on)
3481 {
3482 	struct e1000_mbx_info *mbx = &hw->mbx;
3483 	uint32_t msgbuf[2];
3484 	s32 err;
3485 
3486 	/* After set vlan, vlan strip will also be enabled in igb driver*/
3487 	msgbuf[0] = E1000_VF_SET_VLAN;
3488 	msgbuf[1] = vid;
3489 	/* Setting the 8 bit field MSG INFO to TRUE indicates "add" */
3490 	if (on)
3491 		msgbuf[0] |= E1000_VF_SET_VLAN_ADD;
3492 
3493 	err = mbx->ops.write_posted(hw, msgbuf, 2, 0);
3494 	if (err)
3495 		goto mbx_err;
3496 
3497 	err = mbx->ops.read_posted(hw, msgbuf, 2, 0);
3498 	if (err)
3499 		goto mbx_err;
3500 
3501 	msgbuf[0] &= ~E1000_VT_MSGTYPE_CTS;
3502 	if (msgbuf[0] == (E1000_VF_SET_VLAN | E1000_VT_MSGTYPE_NACK))
3503 		err = -EINVAL;
3504 
3505 mbx_err:
3506 	return err;
3507 }
3508 
3509 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on)
3510 {
3511 	struct e1000_hw *hw =
3512 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3513 	struct e1000_vfta * shadow_vfta =
3514 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3515 	int i = 0, j = 0, vfta = 0, mask = 1;
3516 
3517 	for (i = 0; i < IGB_VFTA_SIZE; i++){
3518 		vfta = shadow_vfta->vfta[i];
3519 		if(vfta){
3520 			mask = 1;
3521 			for (j = 0; j < 32; j++){
3522 				if(vfta & mask)
3523 					igbvf_set_vfta(hw,
3524 						(uint16_t)((i<<5)+j), on);
3525 				mask<<=1;
3526 			}
3527 		}
3528 	}
3529 
3530 }
3531 
3532 static int
3533 igbvf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
3534 {
3535 	struct e1000_hw *hw =
3536 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3537 	struct e1000_vfta * shadow_vfta =
3538 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3539 	uint32_t vid_idx = 0;
3540 	uint32_t vid_bit = 0;
3541 	int ret = 0;
3542 
3543 	PMD_INIT_FUNC_TRACE();
3544 
3545 	/*vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf*/
3546 	ret = igbvf_set_vfta(hw, vlan_id, !!on);
3547 	if(ret){
3548 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
3549 		return ret;
3550 	}
3551 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
3552 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
3553 
3554 	/*Save what we set and retore it after device reset*/
3555 	if (on)
3556 		shadow_vfta->vfta[vid_idx] |= vid_bit;
3557 	else
3558 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
3559 
3560 	return 0;
3561 }
3562 
3563 static int
3564 igbvf_default_mac_addr_set(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
3565 {
3566 	struct e1000_hw *hw =
3567 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3568 
3569 	/* index is not used by rar_set() */
3570 	hw->mac.ops.rar_set(hw, (void *)addr, 0);
3571 	return 0;
3572 }
3573 
3574 
3575 static int
3576 eth_igb_rss_reta_update(struct rte_eth_dev *dev,
3577 			struct rte_eth_rss_reta_entry64 *reta_conf,
3578 			uint16_t reta_size)
3579 {
3580 	uint8_t i, j, mask;
3581 	uint32_t reta, r;
3582 	uint16_t idx, shift;
3583 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3584 
3585 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3586 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3587 			"(%d) doesn't match the number hardware can supported "
3588 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3589 		return -EINVAL;
3590 	}
3591 
3592 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3593 		idx = i / RTE_RETA_GROUP_SIZE;
3594 		shift = i % RTE_RETA_GROUP_SIZE;
3595 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3596 						IGB_4_BIT_MASK);
3597 		if (!mask)
3598 			continue;
3599 		if (mask == IGB_4_BIT_MASK)
3600 			r = 0;
3601 		else
3602 			r = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3603 		for (j = 0, reta = 0; j < IGB_4_BIT_WIDTH; j++) {
3604 			if (mask & (0x1 << j))
3605 				reta |= reta_conf[idx].reta[shift + j] <<
3606 							(CHAR_BIT * j);
3607 			else
3608 				reta |= r & (IGB_8_BIT_MASK << (CHAR_BIT * j));
3609 		}
3610 		E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
3611 	}
3612 
3613 	return 0;
3614 }
3615 
3616 static int
3617 eth_igb_rss_reta_query(struct rte_eth_dev *dev,
3618 		       struct rte_eth_rss_reta_entry64 *reta_conf,
3619 		       uint16_t reta_size)
3620 {
3621 	uint8_t i, j, mask;
3622 	uint32_t reta;
3623 	uint16_t idx, shift;
3624 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3625 
3626 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3627 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3628 			"(%d) doesn't match the number hardware can supported "
3629 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3630 		return -EINVAL;
3631 	}
3632 
3633 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3634 		idx = i / RTE_RETA_GROUP_SIZE;
3635 		shift = i % RTE_RETA_GROUP_SIZE;
3636 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3637 						IGB_4_BIT_MASK);
3638 		if (!mask)
3639 			continue;
3640 		reta = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3641 		for (j = 0; j < IGB_4_BIT_WIDTH; j++) {
3642 			if (mask & (0x1 << j))
3643 				reta_conf[idx].reta[shift + j] =
3644 					((reta >> (CHAR_BIT * j)) &
3645 						IGB_8_BIT_MASK);
3646 		}
3647 	}
3648 
3649 	return 0;
3650 }
3651 
3652 int
3653 eth_igb_syn_filter_set(struct rte_eth_dev *dev,
3654 			struct rte_eth_syn_filter *filter,
3655 			bool add)
3656 {
3657 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3658 	struct e1000_filter_info *filter_info =
3659 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3660 	uint32_t synqf, rfctl;
3661 
3662 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3663 		return -EINVAL;
3664 
3665 	synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3666 
3667 	if (add) {
3668 		if (synqf & E1000_SYN_FILTER_ENABLE)
3669 			return -EINVAL;
3670 
3671 		synqf = (uint32_t)(((filter->queue << E1000_SYN_FILTER_QUEUE_SHIFT) &
3672 			E1000_SYN_FILTER_QUEUE) | E1000_SYN_FILTER_ENABLE);
3673 
3674 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3675 		if (filter->hig_pri)
3676 			rfctl |= E1000_RFCTL_SYNQFP;
3677 		else
3678 			rfctl &= ~E1000_RFCTL_SYNQFP;
3679 
3680 		E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3681 	} else {
3682 		if (!(synqf & E1000_SYN_FILTER_ENABLE))
3683 			return -ENOENT;
3684 		synqf = 0;
3685 	}
3686 
3687 	filter_info->syn_info = synqf;
3688 	E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
3689 	E1000_WRITE_FLUSH(hw);
3690 	return 0;
3691 }
3692 
3693 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_2tuple_filter_info*/
3694 static inline int
3695 ntuple_filter_to_2tuple(struct rte_eth_ntuple_filter *filter,
3696 			struct e1000_2tuple_filter_info *filter_info)
3697 {
3698 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3699 		return -EINVAL;
3700 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
3701 		return -EINVAL;  /* filter index is out of range. */
3702 	if (filter->tcp_flags > RTE_NTUPLE_TCP_FLAGS_MASK)
3703 		return -EINVAL;  /* flags is invalid. */
3704 
3705 	switch (filter->dst_port_mask) {
3706 	case UINT16_MAX:
3707 		filter_info->dst_port_mask = 0;
3708 		filter_info->dst_port = filter->dst_port;
3709 		break;
3710 	case 0:
3711 		filter_info->dst_port_mask = 1;
3712 		break;
3713 	default:
3714 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
3715 		return -EINVAL;
3716 	}
3717 
3718 	switch (filter->proto_mask) {
3719 	case UINT8_MAX:
3720 		filter_info->proto_mask = 0;
3721 		filter_info->proto = filter->proto;
3722 		break;
3723 	case 0:
3724 		filter_info->proto_mask = 1;
3725 		break;
3726 	default:
3727 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
3728 		return -EINVAL;
3729 	}
3730 
3731 	filter_info->priority = (uint8_t)filter->priority;
3732 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
3733 		filter_info->tcp_flags = filter->tcp_flags;
3734 	else
3735 		filter_info->tcp_flags = 0;
3736 
3737 	return 0;
3738 }
3739 
3740 static inline struct e1000_2tuple_filter *
3741 igb_2tuple_filter_lookup(struct e1000_2tuple_filter_list *filter_list,
3742 			struct e1000_2tuple_filter_info *key)
3743 {
3744 	struct e1000_2tuple_filter *it;
3745 
3746 	TAILQ_FOREACH(it, filter_list, entries) {
3747 		if (memcmp(key, &it->filter_info,
3748 			sizeof(struct e1000_2tuple_filter_info)) == 0) {
3749 			return it;
3750 		}
3751 	}
3752 	return NULL;
3753 }
3754 
3755 /* inject a igb 2tuple filter to HW */
3756 static inline void
3757 igb_inject_2uple_filter(struct rte_eth_dev *dev,
3758 			   struct e1000_2tuple_filter *filter)
3759 {
3760 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3761 	uint32_t ttqf = E1000_TTQF_DISABLE_MASK;
3762 	uint32_t imir, imir_ext = E1000_IMIREXT_SIZE_BP;
3763 	int i;
3764 
3765 	i = filter->index;
3766 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
3767 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
3768 		imir |= E1000_IMIR_PORT_BP;
3769 	else
3770 		imir &= ~E1000_IMIR_PORT_BP;
3771 
3772 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
3773 
3774 	ttqf |= E1000_TTQF_QUEUE_ENABLE;
3775 	ttqf |= (uint32_t)(filter->queue << E1000_TTQF_QUEUE_SHIFT);
3776 	ttqf |= (uint32_t)(filter->filter_info.proto &
3777 						E1000_TTQF_PROTOCOL_MASK);
3778 	if (filter->filter_info.proto_mask == 0)
3779 		ttqf &= ~E1000_TTQF_MASK_ENABLE;
3780 
3781 	/* tcp flags bits setting. */
3782 	if (filter->filter_info.tcp_flags & RTE_NTUPLE_TCP_FLAGS_MASK) {
3783 		if (filter->filter_info.tcp_flags & RTE_TCP_URG_FLAG)
3784 			imir_ext |= E1000_IMIREXT_CTRL_URG;
3785 		if (filter->filter_info.tcp_flags & RTE_TCP_ACK_FLAG)
3786 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
3787 		if (filter->filter_info.tcp_flags & RTE_TCP_PSH_FLAG)
3788 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
3789 		if (filter->filter_info.tcp_flags & RTE_TCP_RST_FLAG)
3790 			imir_ext |= E1000_IMIREXT_CTRL_RST;
3791 		if (filter->filter_info.tcp_flags & RTE_TCP_SYN_FLAG)
3792 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
3793 		if (filter->filter_info.tcp_flags & RTE_TCP_FIN_FLAG)
3794 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
3795 	} else {
3796 		imir_ext |= E1000_IMIREXT_CTRL_BP;
3797 	}
3798 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
3799 	E1000_WRITE_REG(hw, E1000_TTQF(i), ttqf);
3800 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
3801 }
3802 
3803 /*
3804  * igb_add_2tuple_filter - add a 2tuple filter
3805  *
3806  * @param
3807  * dev: Pointer to struct rte_eth_dev.
3808  * ntuple_filter: ponter to the filter that will be added.
3809  *
3810  * @return
3811  *    - On success, zero.
3812  *    - On failure, a negative value.
3813  */
3814 static int
3815 igb_add_2tuple_filter(struct rte_eth_dev *dev,
3816 			struct rte_eth_ntuple_filter *ntuple_filter)
3817 {
3818 	struct e1000_filter_info *filter_info =
3819 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3820 	struct e1000_2tuple_filter *filter;
3821 	int i, ret;
3822 
3823 	filter = rte_zmalloc("e1000_2tuple_filter",
3824 			sizeof(struct e1000_2tuple_filter), 0);
3825 	if (filter == NULL)
3826 		return -ENOMEM;
3827 
3828 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3829 				      &filter->filter_info);
3830 	if (ret < 0) {
3831 		rte_free(filter);
3832 		return ret;
3833 	}
3834 	if (igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3835 					 &filter->filter_info) != NULL) {
3836 		PMD_DRV_LOG(ERR, "filter exists.");
3837 		rte_free(filter);
3838 		return -EEXIST;
3839 	}
3840 	filter->queue = ntuple_filter->queue;
3841 
3842 	/*
3843 	 * look for an unused 2tuple filter index,
3844 	 * and insert the filter to list.
3845 	 */
3846 	for (i = 0; i < E1000_MAX_TTQF_FILTERS; i++) {
3847 		if (!(filter_info->twotuple_mask & (1 << i))) {
3848 			filter_info->twotuple_mask |= 1 << i;
3849 			filter->index = i;
3850 			TAILQ_INSERT_TAIL(&filter_info->twotuple_list,
3851 					  filter,
3852 					  entries);
3853 			break;
3854 		}
3855 	}
3856 	if (i >= E1000_MAX_TTQF_FILTERS) {
3857 		PMD_DRV_LOG(ERR, "2tuple filters are full.");
3858 		rte_free(filter);
3859 		return -ENOSYS;
3860 	}
3861 
3862 	igb_inject_2uple_filter(dev, filter);
3863 	return 0;
3864 }
3865 
3866 int
3867 igb_delete_2tuple_filter(struct rte_eth_dev *dev,
3868 			struct e1000_2tuple_filter *filter)
3869 {
3870 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3871 	struct e1000_filter_info *filter_info =
3872 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3873 
3874 	filter_info->twotuple_mask &= ~(1 << filter->index);
3875 	TAILQ_REMOVE(&filter_info->twotuple_list, filter, entries);
3876 	rte_free(filter);
3877 
3878 	E1000_WRITE_REG(hw, E1000_TTQF(filter->index), E1000_TTQF_DISABLE_MASK);
3879 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
3880 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
3881 	return 0;
3882 }
3883 
3884 /*
3885  * igb_remove_2tuple_filter - remove a 2tuple filter
3886  *
3887  * @param
3888  * dev: Pointer to struct rte_eth_dev.
3889  * ntuple_filter: ponter to the filter that will be removed.
3890  *
3891  * @return
3892  *    - On success, zero.
3893  *    - On failure, a negative value.
3894  */
3895 static int
3896 igb_remove_2tuple_filter(struct rte_eth_dev *dev,
3897 			struct rte_eth_ntuple_filter *ntuple_filter)
3898 {
3899 	struct e1000_filter_info *filter_info =
3900 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3901 	struct e1000_2tuple_filter_info filter_2tuple;
3902 	struct e1000_2tuple_filter *filter;
3903 	int ret;
3904 
3905 	memset(&filter_2tuple, 0, sizeof(struct e1000_2tuple_filter_info));
3906 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3907 				      &filter_2tuple);
3908 	if (ret < 0)
3909 		return ret;
3910 
3911 	filter = igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3912 					 &filter_2tuple);
3913 	if (filter == NULL) {
3914 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
3915 		return -ENOENT;
3916 	}
3917 
3918 	igb_delete_2tuple_filter(dev, filter);
3919 
3920 	return 0;
3921 }
3922 
3923 /* inject a igb flex filter to HW */
3924 static inline void
3925 igb_inject_flex_filter(struct rte_eth_dev *dev,
3926 			   struct e1000_flex_filter *filter)
3927 {
3928 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3929 	uint32_t wufc, queueing;
3930 	uint32_t reg_off;
3931 	uint8_t i, j = 0;
3932 
3933 	wufc = E1000_READ_REG(hw, E1000_WUFC);
3934 	if (filter->index < E1000_MAX_FHFT)
3935 		reg_off = E1000_FHFT(filter->index);
3936 	else
3937 		reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
3938 
3939 	E1000_WRITE_REG(hw, E1000_WUFC, wufc | E1000_WUFC_FLEX_HQ |
3940 			(E1000_WUFC_FLX0 << filter->index));
3941 	queueing = filter->filter_info.len |
3942 		(filter->queue << E1000_FHFT_QUEUEING_QUEUE_SHIFT) |
3943 		(filter->filter_info.priority <<
3944 			E1000_FHFT_QUEUEING_PRIO_SHIFT);
3945 	E1000_WRITE_REG(hw, reg_off + E1000_FHFT_QUEUEING_OFFSET,
3946 			queueing);
3947 
3948 	for (i = 0; i < E1000_FLEX_FILTERS_MASK_SIZE; i++) {
3949 		E1000_WRITE_REG(hw, reg_off,
3950 				filter->filter_info.dwords[j]);
3951 		reg_off += sizeof(uint32_t);
3952 		E1000_WRITE_REG(hw, reg_off,
3953 				filter->filter_info.dwords[++j]);
3954 		reg_off += sizeof(uint32_t);
3955 		E1000_WRITE_REG(hw, reg_off,
3956 			(uint32_t)filter->filter_info.mask[i]);
3957 		reg_off += sizeof(uint32_t) * 2;
3958 		++j;
3959 	}
3960 }
3961 
3962 static inline struct e1000_flex_filter *
3963 eth_igb_flex_filter_lookup(struct e1000_flex_filter_list *filter_list,
3964 			struct e1000_flex_filter_info *key)
3965 {
3966 	struct e1000_flex_filter *it;
3967 
3968 	TAILQ_FOREACH(it, filter_list, entries) {
3969 		if (memcmp(key, &it->filter_info,
3970 			sizeof(struct e1000_flex_filter_info)) == 0)
3971 			return it;
3972 	}
3973 
3974 	return NULL;
3975 }
3976 
3977 /* remove a flex byte filter
3978  * @param
3979  * dev: Pointer to struct rte_eth_dev.
3980  * filter: the pointer of the filter will be removed.
3981  */
3982 void
3983 igb_remove_flex_filter(struct rte_eth_dev *dev,
3984 			struct e1000_flex_filter *filter)
3985 {
3986 	struct e1000_filter_info *filter_info =
3987 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3988 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3989 	uint32_t wufc, i;
3990 	uint32_t reg_off;
3991 
3992 	wufc = E1000_READ_REG(hw, E1000_WUFC);
3993 	if (filter->index < E1000_MAX_FHFT)
3994 		reg_off = E1000_FHFT(filter->index);
3995 	else
3996 		reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
3997 
3998 	for (i = 0; i < E1000_FHFT_SIZE_IN_DWD; i++)
3999 		E1000_WRITE_REG(hw, reg_off + i * sizeof(uint32_t), 0);
4000 
4001 	E1000_WRITE_REG(hw, E1000_WUFC, wufc &
4002 		(~(E1000_WUFC_FLX0 << filter->index)));
4003 
4004 	filter_info->flex_mask &= ~(1 << filter->index);
4005 	TAILQ_REMOVE(&filter_info->flex_list, filter, entries);
4006 	rte_free(filter);
4007 }
4008 
4009 int
4010 eth_igb_add_del_flex_filter(struct rte_eth_dev *dev,
4011 			struct igb_flex_filter *filter,
4012 			bool add)
4013 {
4014 	struct e1000_filter_info *filter_info =
4015 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4016 	struct e1000_flex_filter *flex_filter, *it;
4017 	uint32_t mask;
4018 	uint8_t shift, i;
4019 
4020 	flex_filter = rte_zmalloc("e1000_flex_filter",
4021 			sizeof(struct e1000_flex_filter), 0);
4022 	if (flex_filter == NULL)
4023 		return -ENOMEM;
4024 
4025 	flex_filter->filter_info.len = filter->len;
4026 	flex_filter->filter_info.priority = filter->priority;
4027 	memcpy(flex_filter->filter_info.dwords, filter->bytes, filter->len);
4028 	for (i = 0; i < RTE_ALIGN(filter->len, CHAR_BIT) / CHAR_BIT; i++) {
4029 		mask = 0;
4030 		/* reverse bits in flex filter's mask*/
4031 		for (shift = 0; shift < CHAR_BIT; shift++) {
4032 			if (filter->mask[i] & (0x01 << shift))
4033 				mask |= (0x80 >> shift);
4034 		}
4035 		flex_filter->filter_info.mask[i] = mask;
4036 	}
4037 
4038 	it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
4039 				&flex_filter->filter_info);
4040 	if (it == NULL && !add) {
4041 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4042 		rte_free(flex_filter);
4043 		return -ENOENT;
4044 	}
4045 	if (it != NULL && add) {
4046 		PMD_DRV_LOG(ERR, "filter exists.");
4047 		rte_free(flex_filter);
4048 		return -EEXIST;
4049 	}
4050 
4051 	if (add) {
4052 		flex_filter->queue = filter->queue;
4053 		/*
4054 		 * look for an unused flex filter index
4055 		 * and insert the filter into the list.
4056 		 */
4057 		for (i = 0; i < E1000_MAX_FLEX_FILTERS; i++) {
4058 			if (!(filter_info->flex_mask & (1 << i))) {
4059 				filter_info->flex_mask |= 1 << i;
4060 				flex_filter->index = i;
4061 				TAILQ_INSERT_TAIL(&filter_info->flex_list,
4062 					flex_filter,
4063 					entries);
4064 				break;
4065 			}
4066 		}
4067 		if (i >= E1000_MAX_FLEX_FILTERS) {
4068 			PMD_DRV_LOG(ERR, "flex filters are full.");
4069 			rte_free(flex_filter);
4070 			return -ENOSYS;
4071 		}
4072 
4073 		igb_inject_flex_filter(dev, flex_filter);
4074 
4075 	} else {
4076 		igb_remove_flex_filter(dev, it);
4077 		rte_free(flex_filter);
4078 	}
4079 
4080 	return 0;
4081 }
4082 
4083 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_5tuple_filter_info*/
4084 static inline int
4085 ntuple_filter_to_5tuple_82576(struct rte_eth_ntuple_filter *filter,
4086 			struct e1000_5tuple_filter_info *filter_info)
4087 {
4088 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM_82576)
4089 		return -EINVAL;
4090 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
4091 		return -EINVAL;  /* filter index is out of range. */
4092 	if (filter->tcp_flags > RTE_NTUPLE_TCP_FLAGS_MASK)
4093 		return -EINVAL;  /* flags is invalid. */
4094 
4095 	switch (filter->dst_ip_mask) {
4096 	case UINT32_MAX:
4097 		filter_info->dst_ip_mask = 0;
4098 		filter_info->dst_ip = filter->dst_ip;
4099 		break;
4100 	case 0:
4101 		filter_info->dst_ip_mask = 1;
4102 		break;
4103 	default:
4104 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
4105 		return -EINVAL;
4106 	}
4107 
4108 	switch (filter->src_ip_mask) {
4109 	case UINT32_MAX:
4110 		filter_info->src_ip_mask = 0;
4111 		filter_info->src_ip = filter->src_ip;
4112 		break;
4113 	case 0:
4114 		filter_info->src_ip_mask = 1;
4115 		break;
4116 	default:
4117 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
4118 		return -EINVAL;
4119 	}
4120 
4121 	switch (filter->dst_port_mask) {
4122 	case UINT16_MAX:
4123 		filter_info->dst_port_mask = 0;
4124 		filter_info->dst_port = filter->dst_port;
4125 		break;
4126 	case 0:
4127 		filter_info->dst_port_mask = 1;
4128 		break;
4129 	default:
4130 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
4131 		return -EINVAL;
4132 	}
4133 
4134 	switch (filter->src_port_mask) {
4135 	case UINT16_MAX:
4136 		filter_info->src_port_mask = 0;
4137 		filter_info->src_port = filter->src_port;
4138 		break;
4139 	case 0:
4140 		filter_info->src_port_mask = 1;
4141 		break;
4142 	default:
4143 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
4144 		return -EINVAL;
4145 	}
4146 
4147 	switch (filter->proto_mask) {
4148 	case UINT8_MAX:
4149 		filter_info->proto_mask = 0;
4150 		filter_info->proto = filter->proto;
4151 		break;
4152 	case 0:
4153 		filter_info->proto_mask = 1;
4154 		break;
4155 	default:
4156 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
4157 		return -EINVAL;
4158 	}
4159 
4160 	filter_info->priority = (uint8_t)filter->priority;
4161 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
4162 		filter_info->tcp_flags = filter->tcp_flags;
4163 	else
4164 		filter_info->tcp_flags = 0;
4165 
4166 	return 0;
4167 }
4168 
4169 static inline struct e1000_5tuple_filter *
4170 igb_5tuple_filter_lookup_82576(struct e1000_5tuple_filter_list *filter_list,
4171 			struct e1000_5tuple_filter_info *key)
4172 {
4173 	struct e1000_5tuple_filter *it;
4174 
4175 	TAILQ_FOREACH(it, filter_list, entries) {
4176 		if (memcmp(key, &it->filter_info,
4177 			sizeof(struct e1000_5tuple_filter_info)) == 0) {
4178 			return it;
4179 		}
4180 	}
4181 	return NULL;
4182 }
4183 
4184 /* inject a igb 5-tuple filter to HW */
4185 static inline void
4186 igb_inject_5tuple_filter_82576(struct rte_eth_dev *dev,
4187 			   struct e1000_5tuple_filter *filter)
4188 {
4189 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4190 	uint32_t ftqf = E1000_FTQF_VF_BP | E1000_FTQF_MASK;
4191 	uint32_t spqf, imir, imir_ext = E1000_IMIREXT_SIZE_BP;
4192 	uint8_t i;
4193 
4194 	i = filter->index;
4195 	ftqf |= filter->filter_info.proto & E1000_FTQF_PROTOCOL_MASK;
4196 	if (filter->filter_info.src_ip_mask == 0) /* 0b means compare. */
4197 		ftqf &= ~E1000_FTQF_MASK_SOURCE_ADDR_BP;
4198 	if (filter->filter_info.dst_ip_mask == 0)
4199 		ftqf &= ~E1000_FTQF_MASK_DEST_ADDR_BP;
4200 	if (filter->filter_info.src_port_mask == 0)
4201 		ftqf &= ~E1000_FTQF_MASK_SOURCE_PORT_BP;
4202 	if (filter->filter_info.proto_mask == 0)
4203 		ftqf &= ~E1000_FTQF_MASK_PROTO_BP;
4204 	ftqf |= (filter->queue << E1000_FTQF_QUEUE_SHIFT) &
4205 		E1000_FTQF_QUEUE_MASK;
4206 	ftqf |= E1000_FTQF_QUEUE_ENABLE;
4207 	E1000_WRITE_REG(hw, E1000_FTQF(i), ftqf);
4208 	E1000_WRITE_REG(hw, E1000_DAQF(i), filter->filter_info.dst_ip);
4209 	E1000_WRITE_REG(hw, E1000_SAQF(i), filter->filter_info.src_ip);
4210 
4211 	spqf = filter->filter_info.src_port & E1000_SPQF_SRCPORT;
4212 	E1000_WRITE_REG(hw, E1000_SPQF(i), spqf);
4213 
4214 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
4215 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
4216 		imir |= E1000_IMIR_PORT_BP;
4217 	else
4218 		imir &= ~E1000_IMIR_PORT_BP;
4219 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
4220 
4221 	/* tcp flags bits setting. */
4222 	if (filter->filter_info.tcp_flags & RTE_NTUPLE_TCP_FLAGS_MASK) {
4223 		if (filter->filter_info.tcp_flags & RTE_TCP_URG_FLAG)
4224 			imir_ext |= E1000_IMIREXT_CTRL_URG;
4225 		if (filter->filter_info.tcp_flags & RTE_TCP_ACK_FLAG)
4226 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
4227 		if (filter->filter_info.tcp_flags & RTE_TCP_PSH_FLAG)
4228 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
4229 		if (filter->filter_info.tcp_flags & RTE_TCP_RST_FLAG)
4230 			imir_ext |= E1000_IMIREXT_CTRL_RST;
4231 		if (filter->filter_info.tcp_flags & RTE_TCP_SYN_FLAG)
4232 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
4233 		if (filter->filter_info.tcp_flags & RTE_TCP_FIN_FLAG)
4234 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
4235 	} else {
4236 		imir_ext |= E1000_IMIREXT_CTRL_BP;
4237 	}
4238 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
4239 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
4240 }
4241 
4242 /*
4243  * igb_add_5tuple_filter_82576 - add a 5tuple filter
4244  *
4245  * @param
4246  * dev: Pointer to struct rte_eth_dev.
4247  * ntuple_filter: ponter to the filter that will be added.
4248  *
4249  * @return
4250  *    - On success, zero.
4251  *    - On failure, a negative value.
4252  */
4253 static int
4254 igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
4255 			struct rte_eth_ntuple_filter *ntuple_filter)
4256 {
4257 	struct e1000_filter_info *filter_info =
4258 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4259 	struct e1000_5tuple_filter *filter;
4260 	uint8_t i;
4261 	int ret;
4262 
4263 	filter = rte_zmalloc("e1000_5tuple_filter",
4264 			sizeof(struct e1000_5tuple_filter), 0);
4265 	if (filter == NULL)
4266 		return -ENOMEM;
4267 
4268 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4269 					    &filter->filter_info);
4270 	if (ret < 0) {
4271 		rte_free(filter);
4272 		return ret;
4273 	}
4274 
4275 	if (igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4276 					 &filter->filter_info) != NULL) {
4277 		PMD_DRV_LOG(ERR, "filter exists.");
4278 		rte_free(filter);
4279 		return -EEXIST;
4280 	}
4281 	filter->queue = ntuple_filter->queue;
4282 
4283 	/*
4284 	 * look for an unused 5tuple filter index,
4285 	 * and insert the filter to list.
4286 	 */
4287 	for (i = 0; i < E1000_MAX_FTQF_FILTERS; i++) {
4288 		if (!(filter_info->fivetuple_mask & (1 << i))) {
4289 			filter_info->fivetuple_mask |= 1 << i;
4290 			filter->index = i;
4291 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
4292 					  filter,
4293 					  entries);
4294 			break;
4295 		}
4296 	}
4297 	if (i >= E1000_MAX_FTQF_FILTERS) {
4298 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
4299 		rte_free(filter);
4300 		return -ENOSYS;
4301 	}
4302 
4303 	igb_inject_5tuple_filter_82576(dev, filter);
4304 	return 0;
4305 }
4306 
4307 int
4308 igb_delete_5tuple_filter_82576(struct rte_eth_dev *dev,
4309 				struct e1000_5tuple_filter *filter)
4310 {
4311 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4312 	struct e1000_filter_info *filter_info =
4313 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4314 
4315 	filter_info->fivetuple_mask &= ~(1 << filter->index);
4316 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
4317 	rte_free(filter);
4318 
4319 	E1000_WRITE_REG(hw, E1000_FTQF(filter->index),
4320 			E1000_FTQF_VF_BP | E1000_FTQF_MASK);
4321 	E1000_WRITE_REG(hw, E1000_DAQF(filter->index), 0);
4322 	E1000_WRITE_REG(hw, E1000_SAQF(filter->index), 0);
4323 	E1000_WRITE_REG(hw, E1000_SPQF(filter->index), 0);
4324 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
4325 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
4326 	return 0;
4327 }
4328 
4329 /*
4330  * igb_remove_5tuple_filter_82576 - remove a 5tuple filter
4331  *
4332  * @param
4333  * dev: Pointer to struct rte_eth_dev.
4334  * ntuple_filter: ponter to the filter that will be removed.
4335  *
4336  * @return
4337  *    - On success, zero.
4338  *    - On failure, a negative value.
4339  */
4340 static int
4341 igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
4342 				struct rte_eth_ntuple_filter *ntuple_filter)
4343 {
4344 	struct e1000_filter_info *filter_info =
4345 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4346 	struct e1000_5tuple_filter_info filter_5tuple;
4347 	struct e1000_5tuple_filter *filter;
4348 	int ret;
4349 
4350 	memset(&filter_5tuple, 0, sizeof(struct e1000_5tuple_filter_info));
4351 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4352 					    &filter_5tuple);
4353 	if (ret < 0)
4354 		return ret;
4355 
4356 	filter = igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4357 					 &filter_5tuple);
4358 	if (filter == NULL) {
4359 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4360 		return -ENOENT;
4361 	}
4362 
4363 	igb_delete_5tuple_filter_82576(dev, filter);
4364 
4365 	return 0;
4366 }
4367 
4368 static int
4369 eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4370 {
4371 	uint32_t rctl;
4372 	struct e1000_hw *hw;
4373 	struct rte_eth_dev_info dev_info;
4374 	uint32_t frame_size = mtu + E1000_ETH_OVERHEAD;
4375 	int ret;
4376 
4377 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4378 
4379 #ifdef RTE_LIBRTE_82571_SUPPORT
4380 	/* XXX: not bigger than max_rx_pktlen */
4381 	if (hw->mac.type == e1000_82571)
4382 		return -ENOTSUP;
4383 #endif
4384 	ret = eth_igb_infos_get(dev, &dev_info);
4385 	if (ret != 0)
4386 		return ret;
4387 
4388 	/* check that mtu is within the allowed range */
4389 	if (mtu < RTE_ETHER_MIN_MTU ||
4390 			frame_size > dev_info.max_rx_pktlen)
4391 		return -EINVAL;
4392 
4393 	/*
4394 	 * If device is started, refuse mtu that requires the support of
4395 	 * scattered packets when this feature has not been enabled before.
4396 	 */
4397 	if (dev->data->dev_started && !dev->data->scattered_rx &&
4398 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
4399 		PMD_INIT_LOG(ERR, "Stop port first.");
4400 		return -EINVAL;
4401 	}
4402 
4403 	rctl = E1000_READ_REG(hw, E1000_RCTL);
4404 
4405 	/* switch to jumbo mode if needed */
4406 	if (frame_size > E1000_ETH_MAX_LEN) {
4407 		dev->data->dev_conf.rxmode.offloads |=
4408 			DEV_RX_OFFLOAD_JUMBO_FRAME;
4409 		rctl |= E1000_RCTL_LPE;
4410 	} else {
4411 		dev->data->dev_conf.rxmode.offloads &=
4412 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
4413 		rctl &= ~E1000_RCTL_LPE;
4414 	}
4415 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
4416 
4417 	/* update max frame size */
4418 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4419 
4420 	E1000_WRITE_REG(hw, E1000_RLPML,
4421 			dev->data->dev_conf.rxmode.max_rx_pkt_len);
4422 
4423 	return 0;
4424 }
4425 
4426 /*
4427  * igb_add_del_ntuple_filter - add or delete a ntuple filter
4428  *
4429  * @param
4430  * dev: Pointer to struct rte_eth_dev.
4431  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4432  * add: if true, add filter, if false, remove filter
4433  *
4434  * @return
4435  *    - On success, zero.
4436  *    - On failure, a negative value.
4437  */
4438 int
4439 igb_add_del_ntuple_filter(struct rte_eth_dev *dev,
4440 			struct rte_eth_ntuple_filter *ntuple_filter,
4441 			bool add)
4442 {
4443 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4444 	int ret;
4445 
4446 	switch (ntuple_filter->flags) {
4447 	case RTE_5TUPLE_FLAGS:
4448 	case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4449 		if (hw->mac.type != e1000_82576)
4450 			return -ENOTSUP;
4451 		if (add)
4452 			ret = igb_add_5tuple_filter_82576(dev,
4453 							  ntuple_filter);
4454 		else
4455 			ret = igb_remove_5tuple_filter_82576(dev,
4456 							     ntuple_filter);
4457 		break;
4458 	case RTE_2TUPLE_FLAGS:
4459 	case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4460 		if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350 &&
4461 			hw->mac.type != e1000_i210 &&
4462 			hw->mac.type != e1000_i211)
4463 			return -ENOTSUP;
4464 		if (add)
4465 			ret = igb_add_2tuple_filter(dev, ntuple_filter);
4466 		else
4467 			ret = igb_remove_2tuple_filter(dev, ntuple_filter);
4468 		break;
4469 	default:
4470 		ret = -EINVAL;
4471 		break;
4472 	}
4473 
4474 	return ret;
4475 }
4476 
4477 static inline int
4478 igb_ethertype_filter_lookup(struct e1000_filter_info *filter_info,
4479 			uint16_t ethertype)
4480 {
4481 	int i;
4482 
4483 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4484 		if (filter_info->ethertype_filters[i].ethertype == ethertype &&
4485 		    (filter_info->ethertype_mask & (1 << i)))
4486 			return i;
4487 	}
4488 	return -1;
4489 }
4490 
4491 static inline int
4492 igb_ethertype_filter_insert(struct e1000_filter_info *filter_info,
4493 			uint16_t ethertype, uint32_t etqf)
4494 {
4495 	int i;
4496 
4497 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4498 		if (!(filter_info->ethertype_mask & (1 << i))) {
4499 			filter_info->ethertype_mask |= 1 << i;
4500 			filter_info->ethertype_filters[i].ethertype = ethertype;
4501 			filter_info->ethertype_filters[i].etqf = etqf;
4502 			return i;
4503 		}
4504 	}
4505 	return -1;
4506 }
4507 
4508 int
4509 igb_ethertype_filter_remove(struct e1000_filter_info *filter_info,
4510 			uint8_t idx)
4511 {
4512 	if (idx >= E1000_MAX_ETQF_FILTERS)
4513 		return -1;
4514 	filter_info->ethertype_mask &= ~(1 << idx);
4515 	filter_info->ethertype_filters[idx].ethertype = 0;
4516 	filter_info->ethertype_filters[idx].etqf = 0;
4517 	return idx;
4518 }
4519 
4520 
4521 int
4522 igb_add_del_ethertype_filter(struct rte_eth_dev *dev,
4523 			struct rte_eth_ethertype_filter *filter,
4524 			bool add)
4525 {
4526 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4527 	struct e1000_filter_info *filter_info =
4528 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4529 	uint32_t etqf = 0;
4530 	int ret;
4531 
4532 	if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
4533 		filter->ether_type == RTE_ETHER_TYPE_IPV6) {
4534 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
4535 			" ethertype filter.", filter->ether_type);
4536 		return -EINVAL;
4537 	}
4538 
4539 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
4540 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
4541 		return -EINVAL;
4542 	}
4543 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
4544 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
4545 		return -EINVAL;
4546 	}
4547 
4548 	ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4549 	if (ret >= 0 && add) {
4550 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
4551 			    filter->ether_type);
4552 		return -EEXIST;
4553 	}
4554 	if (ret < 0 && !add) {
4555 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4556 			    filter->ether_type);
4557 		return -ENOENT;
4558 	}
4559 
4560 	if (add) {
4561 		etqf |= E1000_ETQF_FILTER_ENABLE | E1000_ETQF_QUEUE_ENABLE;
4562 		etqf |= (uint32_t)(filter->ether_type & E1000_ETQF_ETHERTYPE);
4563 		etqf |= filter->queue << E1000_ETQF_QUEUE_SHIFT;
4564 		ret = igb_ethertype_filter_insert(filter_info,
4565 				filter->ether_type, etqf);
4566 		if (ret < 0) {
4567 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
4568 			return -ENOSYS;
4569 		}
4570 	} else {
4571 		ret = igb_ethertype_filter_remove(filter_info, (uint8_t)ret);
4572 		if (ret < 0)
4573 			return -ENOSYS;
4574 	}
4575 	E1000_WRITE_REG(hw, E1000_ETQF(ret), etqf);
4576 	E1000_WRITE_FLUSH(hw);
4577 
4578 	return 0;
4579 }
4580 
4581 static int
4582 eth_igb_flow_ops_get(struct rte_eth_dev *dev __rte_unused,
4583 		     const struct rte_flow_ops **ops)
4584 {
4585 	*ops = &igb_flow_ops;
4586 	return 0;
4587 }
4588 
4589 static int
4590 eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
4591 			 struct rte_ether_addr *mc_addr_set,
4592 			 uint32_t nb_mc_addr)
4593 {
4594 	struct e1000_hw *hw;
4595 
4596 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4597 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
4598 	return 0;
4599 }
4600 
4601 static uint64_t
4602 igb_read_systime_cyclecounter(struct rte_eth_dev *dev)
4603 {
4604 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4605 	uint64_t systime_cycles;
4606 
4607 	switch (hw->mac.type) {
4608 	case e1000_i210:
4609 	case e1000_i211:
4610 		/*
4611 		 * Need to read System Time Residue Register to be able
4612 		 * to read the other two registers.
4613 		 */
4614 		E1000_READ_REG(hw, E1000_SYSTIMR);
4615 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
4616 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4617 		systime_cycles += (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4618 				* NSEC_PER_SEC;
4619 		break;
4620 	case e1000_82580:
4621 	case e1000_i350:
4622 	case e1000_i354:
4623 		/*
4624 		 * Need to read System Time Residue Register to be able
4625 		 * to read the other two registers.
4626 		 */
4627 		E1000_READ_REG(hw, E1000_SYSTIMR);
4628 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4629 		/* Only the 8 LSB are valid. */
4630 		systime_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_SYSTIMH)
4631 				& 0xff) << 32;
4632 		break;
4633 	default:
4634 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4635 		systime_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4636 				<< 32;
4637 		break;
4638 	}
4639 
4640 	return systime_cycles;
4641 }
4642 
4643 static uint64_t
4644 igb_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
4645 {
4646 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4647 	uint64_t rx_tstamp_cycles;
4648 
4649 	switch (hw->mac.type) {
4650 	case e1000_i210:
4651 	case e1000_i211:
4652 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
4653 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4654 		rx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
4655 				* NSEC_PER_SEC;
4656 		break;
4657 	case e1000_82580:
4658 	case e1000_i350:
4659 	case e1000_i354:
4660 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4661 		/* Only the 8 LSB are valid. */
4662 		rx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_RXSTMPH)
4663 				& 0xff) << 32;
4664 		break;
4665 	default:
4666 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4667 		rx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
4668 				<< 32;
4669 		break;
4670 	}
4671 
4672 	return rx_tstamp_cycles;
4673 }
4674 
4675 static uint64_t
4676 igb_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
4677 {
4678 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4679 	uint64_t tx_tstamp_cycles;
4680 
4681 	switch (hw->mac.type) {
4682 	case e1000_i210:
4683 	case e1000_i211:
4684 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
4685 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
4686 		tx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
4687 				* NSEC_PER_SEC;
4688 		break;
4689 	case e1000_82580:
4690 	case e1000_i350:
4691 	case e1000_i354:
4692 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
4693 		/* Only the 8 LSB are valid. */
4694 		tx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_TXSTMPH)
4695 				& 0xff) << 32;
4696 		break;
4697 	default:
4698 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
4699 		tx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
4700 				<< 32;
4701 		break;
4702 	}
4703 
4704 	return tx_tstamp_cycles;
4705 }
4706 
4707 static void
4708 igb_start_timecounters(struct rte_eth_dev *dev)
4709 {
4710 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4711 	struct e1000_adapter *adapter = dev->data->dev_private;
4712 	uint32_t incval = 1;
4713 	uint32_t shift = 0;
4714 	uint64_t mask = E1000_CYCLECOUNTER_MASK;
4715 
4716 	switch (hw->mac.type) {
4717 	case e1000_82580:
4718 	case e1000_i350:
4719 	case e1000_i354:
4720 		/* 32 LSB bits + 8 MSB bits = 40 bits */
4721 		mask = (1ULL << 40) - 1;
4722 		/* fall-through */
4723 	case e1000_i210:
4724 	case e1000_i211:
4725 		/*
4726 		 * Start incrementing the register
4727 		 * used to timestamp PTP packets.
4728 		 */
4729 		E1000_WRITE_REG(hw, E1000_TIMINCA, incval);
4730 		break;
4731 	case e1000_82576:
4732 		incval = E1000_INCVALUE_82576;
4733 		shift = IGB_82576_TSYNC_SHIFT;
4734 		E1000_WRITE_REG(hw, E1000_TIMINCA,
4735 				E1000_INCPERIOD_82576 | incval);
4736 		break;
4737 	default:
4738 		/* Not supported */
4739 		return;
4740 	}
4741 
4742 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
4743 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
4744 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
4745 
4746 	adapter->systime_tc.cc_mask = mask;
4747 	adapter->systime_tc.cc_shift = shift;
4748 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
4749 
4750 	adapter->rx_tstamp_tc.cc_mask = mask;
4751 	adapter->rx_tstamp_tc.cc_shift = shift;
4752 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
4753 
4754 	adapter->tx_tstamp_tc.cc_mask = mask;
4755 	adapter->tx_tstamp_tc.cc_shift = shift;
4756 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
4757 }
4758 
4759 static int
4760 igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
4761 {
4762 	struct e1000_adapter *adapter = dev->data->dev_private;
4763 
4764 	adapter->systime_tc.nsec += delta;
4765 	adapter->rx_tstamp_tc.nsec += delta;
4766 	adapter->tx_tstamp_tc.nsec += delta;
4767 
4768 	return 0;
4769 }
4770 
4771 static int
4772 igb_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
4773 {
4774 	uint64_t ns;
4775 	struct e1000_adapter *adapter = dev->data->dev_private;
4776 
4777 	ns = rte_timespec_to_ns(ts);
4778 
4779 	/* Set the timecounters to a new value. */
4780 	adapter->systime_tc.nsec = ns;
4781 	adapter->rx_tstamp_tc.nsec = ns;
4782 	adapter->tx_tstamp_tc.nsec = ns;
4783 
4784 	return 0;
4785 }
4786 
4787 static int
4788 igb_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
4789 {
4790 	uint64_t ns, systime_cycles;
4791 	struct e1000_adapter *adapter = dev->data->dev_private;
4792 
4793 	systime_cycles = igb_read_systime_cyclecounter(dev);
4794 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
4795 	*ts = rte_ns_to_timespec(ns);
4796 
4797 	return 0;
4798 }
4799 
4800 static int
4801 igb_timesync_enable(struct rte_eth_dev *dev)
4802 {
4803 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4804 	uint32_t tsync_ctl;
4805 	uint32_t tsauxc;
4806 
4807 	/* Stop the timesync system time. */
4808 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0x0);
4809 	/* Reset the timesync system time value. */
4810 	switch (hw->mac.type) {
4811 	case e1000_82580:
4812 	case e1000_i350:
4813 	case e1000_i354:
4814 	case e1000_i210:
4815 	case e1000_i211:
4816 		E1000_WRITE_REG(hw, E1000_SYSTIMR, 0x0);
4817 		/* fall-through */
4818 	case e1000_82576:
4819 		E1000_WRITE_REG(hw, E1000_SYSTIML, 0x0);
4820 		E1000_WRITE_REG(hw, E1000_SYSTIMH, 0x0);
4821 		break;
4822 	default:
4823 		/* Not supported. */
4824 		return -ENOTSUP;
4825 	}
4826 
4827 	/* Enable system time for it isn't on by default. */
4828 	tsauxc = E1000_READ_REG(hw, E1000_TSAUXC);
4829 	tsauxc &= ~E1000_TSAUXC_DISABLE_SYSTIME;
4830 	E1000_WRITE_REG(hw, E1000_TSAUXC, tsauxc);
4831 
4832 	igb_start_timecounters(dev);
4833 
4834 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
4835 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588),
4836 			(RTE_ETHER_TYPE_1588 |
4837 			 E1000_ETQF_FILTER_ENABLE |
4838 			 E1000_ETQF_1588));
4839 
4840 	/* Enable timestamping of received PTP packets. */
4841 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
4842 	tsync_ctl |= E1000_TSYNCRXCTL_ENABLED;
4843 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
4844 
4845 	/* Enable Timestamping of transmitted PTP packets. */
4846 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
4847 	tsync_ctl |= E1000_TSYNCTXCTL_ENABLED;
4848 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
4849 
4850 	return 0;
4851 }
4852 
4853 static int
4854 igb_timesync_disable(struct rte_eth_dev *dev)
4855 {
4856 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4857 	uint32_t tsync_ctl;
4858 
4859 	/* Disable timestamping of transmitted PTP packets. */
4860 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
4861 	tsync_ctl &= ~E1000_TSYNCTXCTL_ENABLED;
4862 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
4863 
4864 	/* Disable timestamping of received PTP packets. */
4865 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
4866 	tsync_ctl &= ~E1000_TSYNCRXCTL_ENABLED;
4867 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
4868 
4869 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
4870 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588), 0);
4871 
4872 	/* Stop incrementating the System Time registers. */
4873 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0);
4874 
4875 	return 0;
4876 }
4877 
4878 static int
4879 igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
4880 			       struct timespec *timestamp,
4881 			       uint32_t flags __rte_unused)
4882 {
4883 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4884 	struct e1000_adapter *adapter = dev->data->dev_private;
4885 	uint32_t tsync_rxctl;
4886 	uint64_t rx_tstamp_cycles;
4887 	uint64_t ns;
4888 
4889 	tsync_rxctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
4890 	if ((tsync_rxctl & E1000_TSYNCRXCTL_VALID) == 0)
4891 		return -EINVAL;
4892 
4893 	rx_tstamp_cycles = igb_read_rx_tstamp_cyclecounter(dev);
4894 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
4895 	*timestamp = rte_ns_to_timespec(ns);
4896 
4897 	return  0;
4898 }
4899 
4900 static int
4901 igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
4902 			       struct timespec *timestamp)
4903 {
4904 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4905 	struct e1000_adapter *adapter = dev->data->dev_private;
4906 	uint32_t tsync_txctl;
4907 	uint64_t tx_tstamp_cycles;
4908 	uint64_t ns;
4909 
4910 	tsync_txctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
4911 	if ((tsync_txctl & E1000_TSYNCTXCTL_VALID) == 0)
4912 		return -EINVAL;
4913 
4914 	tx_tstamp_cycles = igb_read_tx_tstamp_cyclecounter(dev);
4915 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
4916 	*timestamp = rte_ns_to_timespec(ns);
4917 
4918 	return  0;
4919 }
4920 
4921 static int
4922 eth_igb_get_reg_length(struct rte_eth_dev *dev __rte_unused)
4923 {
4924 	int count = 0;
4925 	int g_ind = 0;
4926 	const struct reg_info *reg_group;
4927 
4928 	while ((reg_group = igb_regs[g_ind++]))
4929 		count += igb_reg_group_count(reg_group);
4930 
4931 	return count;
4932 }
4933 
4934 static int
4935 igbvf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
4936 {
4937 	int count = 0;
4938 	int g_ind = 0;
4939 	const struct reg_info *reg_group;
4940 
4941 	while ((reg_group = igbvf_regs[g_ind++]))
4942 		count += igb_reg_group_count(reg_group);
4943 
4944 	return count;
4945 }
4946 
4947 static int
4948 eth_igb_get_regs(struct rte_eth_dev *dev,
4949 	struct rte_dev_reg_info *regs)
4950 {
4951 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4952 	uint32_t *data = regs->data;
4953 	int g_ind = 0;
4954 	int count = 0;
4955 	const struct reg_info *reg_group;
4956 
4957 	if (data == NULL) {
4958 		regs->length = eth_igb_get_reg_length(dev);
4959 		regs->width = sizeof(uint32_t);
4960 		return 0;
4961 	}
4962 
4963 	/* Support only full register dump */
4964 	if ((regs->length == 0) ||
4965 	    (regs->length == (uint32_t)eth_igb_get_reg_length(dev))) {
4966 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
4967 			hw->device_id;
4968 		while ((reg_group = igb_regs[g_ind++]))
4969 			count += igb_read_regs_group(dev, &data[count],
4970 							reg_group);
4971 		return 0;
4972 	}
4973 
4974 	return -ENOTSUP;
4975 }
4976 
4977 static int
4978 igbvf_get_regs(struct rte_eth_dev *dev,
4979 	struct rte_dev_reg_info *regs)
4980 {
4981 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4982 	uint32_t *data = regs->data;
4983 	int g_ind = 0;
4984 	int count = 0;
4985 	const struct reg_info *reg_group;
4986 
4987 	if (data == NULL) {
4988 		regs->length = igbvf_get_reg_length(dev);
4989 		regs->width = sizeof(uint32_t);
4990 		return 0;
4991 	}
4992 
4993 	/* Support only full register dump */
4994 	if ((regs->length == 0) ||
4995 	    (regs->length == (uint32_t)igbvf_get_reg_length(dev))) {
4996 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
4997 			hw->device_id;
4998 		while ((reg_group = igbvf_regs[g_ind++]))
4999 			count += igb_read_regs_group(dev, &data[count],
5000 							reg_group);
5001 		return 0;
5002 	}
5003 
5004 	return -ENOTSUP;
5005 }
5006 
5007 static int
5008 eth_igb_get_eeprom_length(struct rte_eth_dev *dev)
5009 {
5010 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5011 
5012 	/* Return unit is byte count */
5013 	return hw->nvm.word_size * 2;
5014 }
5015 
5016 static int
5017 eth_igb_get_eeprom(struct rte_eth_dev *dev,
5018 	struct rte_dev_eeprom_info *in_eeprom)
5019 {
5020 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5021 	struct e1000_nvm_info *nvm = &hw->nvm;
5022 	uint16_t *data = in_eeprom->data;
5023 	int first, length;
5024 
5025 	first = in_eeprom->offset >> 1;
5026 	length = in_eeprom->length >> 1;
5027 	if ((first >= hw->nvm.word_size) ||
5028 	    ((first + length) >= hw->nvm.word_size))
5029 		return -EINVAL;
5030 
5031 	in_eeprom->magic = hw->vendor_id |
5032 		((uint32_t)hw->device_id << 16);
5033 
5034 	if ((nvm->ops.read) == NULL)
5035 		return -ENOTSUP;
5036 
5037 	return nvm->ops.read(hw, first, length, data);
5038 }
5039 
5040 static int
5041 eth_igb_set_eeprom(struct rte_eth_dev *dev,
5042 	struct rte_dev_eeprom_info *in_eeprom)
5043 {
5044 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5045 	struct e1000_nvm_info *nvm = &hw->nvm;
5046 	uint16_t *data = in_eeprom->data;
5047 	int first, length;
5048 
5049 	first = in_eeprom->offset >> 1;
5050 	length = in_eeprom->length >> 1;
5051 	if ((first >= hw->nvm.word_size) ||
5052 	    ((first + length) >= hw->nvm.word_size))
5053 		return -EINVAL;
5054 
5055 	in_eeprom->magic = (uint32_t)hw->vendor_id |
5056 		((uint32_t)hw->device_id << 16);
5057 
5058 	if ((nvm->ops.write) == NULL)
5059 		return -ENOTSUP;
5060 	return nvm->ops.write(hw,  first, length, data);
5061 }
5062 
5063 static int
5064 eth_igb_get_module_info(struct rte_eth_dev *dev,
5065 			struct rte_eth_dev_module_info *modinfo)
5066 {
5067 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5068 
5069 	uint32_t status = 0;
5070 	uint16_t sff8472_rev, addr_mode;
5071 	bool page_swap = false;
5072 
5073 	if (hw->phy.media_type == e1000_media_type_copper ||
5074 	    hw->phy.media_type == e1000_media_type_unknown)
5075 		return -EOPNOTSUPP;
5076 
5077 	/* Check whether we support SFF-8472 or not */
5078 	status = e1000_read_phy_reg_i2c(hw, IGB_SFF_8472_COMP, &sff8472_rev);
5079 	if (status)
5080 		return -EIO;
5081 
5082 	/* addressing mode is not supported */
5083 	status = e1000_read_phy_reg_i2c(hw, IGB_SFF_8472_SWAP, &addr_mode);
5084 	if (status)
5085 		return -EIO;
5086 
5087 	/* addressing mode is not supported */
5088 	if ((addr_mode & 0xFF) & IGB_SFF_ADDRESSING_MODE) {
5089 		PMD_DRV_LOG(ERR,
5090 			    "Address change required to access page 0xA2, "
5091 			    "but not supported. Please report the module "
5092 			    "type to the driver maintainers.\n");
5093 		page_swap = true;
5094 	}
5095 
5096 	if ((sff8472_rev & 0xFF) == IGB_SFF_8472_UNSUP || page_swap) {
5097 		/* We have an SFP, but it does not support SFF-8472 */
5098 		modinfo->type = RTE_ETH_MODULE_SFF_8079;
5099 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
5100 	} else {
5101 		/* We have an SFP which supports a revision of SFF-8472 */
5102 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
5103 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
5104 	}
5105 
5106 	return 0;
5107 }
5108 
5109 static int
5110 eth_igb_get_module_eeprom(struct rte_eth_dev *dev,
5111 			  struct rte_dev_eeprom_info *info)
5112 {
5113 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5114 
5115 	uint32_t status = 0;
5116 	uint16_t dataword[RTE_ETH_MODULE_SFF_8472_LEN / 2 + 1];
5117 	u16 first_word, last_word;
5118 	int i = 0;
5119 
5120 	first_word = info->offset >> 1;
5121 	last_word = (info->offset + info->length - 1) >> 1;
5122 
5123 	/* Read EEPROM block, SFF-8079/SFF-8472, word at a time */
5124 	for (i = 0; i < last_word - first_word + 1; i++) {
5125 		status = e1000_read_phy_reg_i2c(hw, (first_word + i) * 2,
5126 						&dataword[i]);
5127 		if (status) {
5128 			/* Error occurred while reading module */
5129 			return -EIO;
5130 		}
5131 
5132 		dataword[i] = rte_be_to_cpu_16(dataword[i]);
5133 	}
5134 
5135 	memcpy(info->data, (u8 *)dataword + (info->offset & 1), info->length);
5136 
5137 	return 0;
5138 }
5139 
5140 static int
5141 eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5142 {
5143 	struct e1000_hw *hw =
5144 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5145 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5146 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5147 	uint32_t vec = E1000_MISC_VEC_ID;
5148 
5149 	if (rte_intr_allow_others(intr_handle))
5150 		vec = E1000_RX_VEC_START;
5151 
5152 	uint32_t mask = 1 << (queue_id + vec);
5153 
5154 	E1000_WRITE_REG(hw, E1000_EIMC, mask);
5155 	E1000_WRITE_FLUSH(hw);
5156 
5157 	return 0;
5158 }
5159 
5160 static int
5161 eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5162 {
5163 	struct e1000_hw *hw =
5164 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5165 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5166 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5167 	uint32_t vec = E1000_MISC_VEC_ID;
5168 
5169 	if (rte_intr_allow_others(intr_handle))
5170 		vec = E1000_RX_VEC_START;
5171 
5172 	uint32_t mask = 1 << (queue_id + vec);
5173 	uint32_t regval;
5174 
5175 	regval = E1000_READ_REG(hw, E1000_EIMS);
5176 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
5177 	E1000_WRITE_FLUSH(hw);
5178 
5179 	rte_intr_ack(intr_handle);
5180 
5181 	return 0;
5182 }
5183 
5184 static void
5185 eth_igb_write_ivar(struct e1000_hw *hw, uint8_t  msix_vector,
5186 		   uint8_t index, uint8_t offset)
5187 {
5188 	uint32_t val = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
5189 
5190 	/* clear bits */
5191 	val &= ~((uint32_t)0xFF << offset);
5192 
5193 	/* write vector and valid bit */
5194 	val |= (msix_vector | E1000_IVAR_VALID) << offset;
5195 
5196 	E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, val);
5197 }
5198 
5199 static void
5200 eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
5201 			   uint8_t queue, uint8_t msix_vector)
5202 {
5203 	uint32_t tmp = 0;
5204 
5205 	if (hw->mac.type == e1000_82575) {
5206 		if (direction == 0)
5207 			tmp = E1000_EICR_RX_QUEUE0 << queue;
5208 		else if (direction == 1)
5209 			tmp = E1000_EICR_TX_QUEUE0 << queue;
5210 		E1000_WRITE_REG(hw, E1000_MSIXBM(msix_vector), tmp);
5211 	} else if (hw->mac.type == e1000_82576) {
5212 		if ((direction == 0) || (direction == 1))
5213 			eth_igb_write_ivar(hw, msix_vector, queue & 0x7,
5214 					   ((queue & 0x8) << 1) +
5215 					   8 * direction);
5216 	} else if ((hw->mac.type == e1000_82580) ||
5217 			(hw->mac.type == e1000_i350) ||
5218 			(hw->mac.type == e1000_i354) ||
5219 			(hw->mac.type == e1000_i210) ||
5220 			(hw->mac.type == e1000_i211)) {
5221 		if ((direction == 0) || (direction == 1))
5222 			eth_igb_write_ivar(hw, msix_vector,
5223 					   queue >> 1,
5224 					   ((queue & 0x1) << 4) +
5225 					   8 * direction);
5226 	}
5227 }
5228 
5229 /* Sets up the hardware to generate MSI-X interrupts properly
5230  * @hw
5231  *  board private structure
5232  */
5233 static void
5234 eth_igb_configure_msix_intr(struct rte_eth_dev *dev)
5235 {
5236 	int queue_id;
5237 	uint32_t tmpval, regval, intr_mask;
5238 	struct e1000_hw *hw =
5239 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5240 	uint32_t vec = E1000_MISC_VEC_ID;
5241 	uint32_t base = E1000_MISC_VEC_ID;
5242 	uint32_t misc_shift = 0;
5243 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5244 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5245 
5246 	/* won't configure msix register if no mapping is done
5247 	 * between intr vector and event fd
5248 	 */
5249 	if (!rte_intr_dp_is_en(intr_handle))
5250 		return;
5251 
5252 	if (rte_intr_allow_others(intr_handle)) {
5253 		vec = base = E1000_RX_VEC_START;
5254 		misc_shift = 1;
5255 	}
5256 
5257 	/* set interrupt vector for other causes */
5258 	if (hw->mac.type == e1000_82575) {
5259 		tmpval = E1000_READ_REG(hw, E1000_CTRL_EXT);
5260 		/* enable MSI-X PBA support */
5261 		tmpval |= E1000_CTRL_EXT_PBA_CLR;
5262 
5263 		/* Auto-Mask interrupts upon ICR read */
5264 		tmpval |= E1000_CTRL_EXT_EIAME;
5265 		tmpval |= E1000_CTRL_EXT_IRCA;
5266 
5267 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmpval);
5268 
5269 		/* enable msix_other interrupt */
5270 		E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0), 0, E1000_EIMS_OTHER);
5271 		regval = E1000_READ_REG(hw, E1000_EIAC);
5272 		E1000_WRITE_REG(hw, E1000_EIAC, regval | E1000_EIMS_OTHER);
5273 		regval = E1000_READ_REG(hw, E1000_EIAM);
5274 		E1000_WRITE_REG(hw, E1000_EIMS, regval | E1000_EIMS_OTHER);
5275 	} else if ((hw->mac.type == e1000_82576) ||
5276 			(hw->mac.type == e1000_82580) ||
5277 			(hw->mac.type == e1000_i350) ||
5278 			(hw->mac.type == e1000_i354) ||
5279 			(hw->mac.type == e1000_i210) ||
5280 			(hw->mac.type == e1000_i211)) {
5281 		/* turn on MSI-X capability first */
5282 		E1000_WRITE_REG(hw, E1000_GPIE, E1000_GPIE_MSIX_MODE |
5283 					E1000_GPIE_PBA | E1000_GPIE_EIAME |
5284 					E1000_GPIE_NSICR);
5285 		intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5286 			misc_shift;
5287 
5288 		if (dev->data->dev_conf.intr_conf.lsc != 0)
5289 			intr_mask |= (1 << IGB_MSIX_OTHER_INTR_VEC);
5290 
5291 		regval = E1000_READ_REG(hw, E1000_EIAC);
5292 		E1000_WRITE_REG(hw, E1000_EIAC, regval | intr_mask);
5293 
5294 		/* enable msix_other interrupt */
5295 		regval = E1000_READ_REG(hw, E1000_EIMS);
5296 		E1000_WRITE_REG(hw, E1000_EIMS, regval | intr_mask);
5297 		tmpval = (IGB_MSIX_OTHER_INTR_VEC | E1000_IVAR_VALID) << 8;
5298 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, tmpval);
5299 	}
5300 
5301 	/* use EIAM to auto-mask when MSI-X interrupt
5302 	 * is asserted, this saves a register write for every interrupt
5303 	 */
5304 	intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5305 		misc_shift;
5306 
5307 	if (dev->data->dev_conf.intr_conf.lsc != 0)
5308 		intr_mask |= (1 << IGB_MSIX_OTHER_INTR_VEC);
5309 
5310 	regval = E1000_READ_REG(hw, E1000_EIAM);
5311 	E1000_WRITE_REG(hw, E1000_EIAM, regval | intr_mask);
5312 
5313 	for (queue_id = 0; queue_id < dev->data->nb_rx_queues; queue_id++) {
5314 		eth_igb_assign_msix_vector(hw, 0, queue_id, vec);
5315 		intr_handle->intr_vec[queue_id] = vec;
5316 		if (vec < base + intr_handle->nb_efd - 1)
5317 			vec++;
5318 	}
5319 
5320 	E1000_WRITE_FLUSH(hw);
5321 }
5322 
5323 /* restore n-tuple filter */
5324 static inline void
5325 igb_ntuple_filter_restore(struct rte_eth_dev *dev)
5326 {
5327 	struct e1000_filter_info *filter_info =
5328 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5329 	struct e1000_5tuple_filter *p_5tuple;
5330 	struct e1000_2tuple_filter *p_2tuple;
5331 
5332 	TAILQ_FOREACH(p_5tuple, &filter_info->fivetuple_list, entries) {
5333 		igb_inject_5tuple_filter_82576(dev, p_5tuple);
5334 	}
5335 
5336 	TAILQ_FOREACH(p_2tuple, &filter_info->twotuple_list, entries) {
5337 		igb_inject_2uple_filter(dev, p_2tuple);
5338 	}
5339 }
5340 
5341 /* restore SYN filter */
5342 static inline void
5343 igb_syn_filter_restore(struct rte_eth_dev *dev)
5344 {
5345 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5346 	struct e1000_filter_info *filter_info =
5347 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5348 	uint32_t synqf;
5349 
5350 	synqf = filter_info->syn_info;
5351 
5352 	if (synqf & E1000_SYN_FILTER_ENABLE) {
5353 		E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
5354 		E1000_WRITE_FLUSH(hw);
5355 	}
5356 }
5357 
5358 /* restore ethernet type filter */
5359 static inline void
5360 igb_ethertype_filter_restore(struct rte_eth_dev *dev)
5361 {
5362 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5363 	struct e1000_filter_info *filter_info =
5364 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5365 	int i;
5366 
5367 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
5368 		if (filter_info->ethertype_mask & (1 << i)) {
5369 			E1000_WRITE_REG(hw, E1000_ETQF(i),
5370 				filter_info->ethertype_filters[i].etqf);
5371 			E1000_WRITE_FLUSH(hw);
5372 		}
5373 	}
5374 }
5375 
5376 /* restore flex byte filter */
5377 static inline void
5378 igb_flex_filter_restore(struct rte_eth_dev *dev)
5379 {
5380 	struct e1000_filter_info *filter_info =
5381 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5382 	struct e1000_flex_filter *flex_filter;
5383 
5384 	TAILQ_FOREACH(flex_filter, &filter_info->flex_list, entries) {
5385 		igb_inject_flex_filter(dev, flex_filter);
5386 	}
5387 }
5388 
5389 /* restore rss filter */
5390 static inline void
5391 igb_rss_filter_restore(struct rte_eth_dev *dev)
5392 {
5393 	struct e1000_filter_info *filter_info =
5394 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5395 
5396 	if (filter_info->rss_info.conf.queue_num)
5397 		igb_config_rss_filter(dev, &filter_info->rss_info, TRUE);
5398 }
5399 
5400 /* restore all types filter */
5401 static int
5402 igb_filter_restore(struct rte_eth_dev *dev)
5403 {
5404 	igb_ntuple_filter_restore(dev);
5405 	igb_ethertype_filter_restore(dev);
5406 	igb_syn_filter_restore(dev);
5407 	igb_flex_filter_restore(dev);
5408 	igb_rss_filter_restore(dev);
5409 
5410 	return 0;
5411 }
5412 
5413 RTE_PMD_REGISTER_PCI(net_e1000_igb, rte_igb_pmd);
5414 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb, pci_id_igb_map);
5415 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb, "* igb_uio | uio_pci_generic | vfio-pci");
5416 RTE_PMD_REGISTER_PCI(net_e1000_igb_vf, rte_igbvf_pmd);
5417 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb_vf, pci_id_igbvf_map);
5418 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb_vf, "* igb_uio | vfio-pci");
5419