xref: /dpdk/drivers/net/e1000/igb_ethdev.c (revision 3e3f736e)
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 			     RTE_ETHER_ADDR_PRT_FMT,
981 			     RTE_ETHER_ADDR_BYTES(perm_addr));
982 	}
983 
984 	diag = e1000_rar_set(hw, perm_addr->addr_bytes, 0);
985 	if (diag) {
986 		rte_free(eth_dev->data->mac_addrs);
987 		eth_dev->data->mac_addrs = NULL;
988 		return diag;
989 	}
990 	/* Copy the permanent MAC address */
991 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.perm_addr,
992 			&eth_dev->data->mac_addrs[0]);
993 
994 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x "
995 		     "mac.type=%s",
996 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
997 		     pci_dev->id.device_id, "igb_mac_82576_vf");
998 
999 	intr_handle = &pci_dev->intr_handle;
1000 	rte_intr_callback_register(intr_handle,
1001 				   eth_igbvf_interrupt_handler, eth_dev);
1002 
1003 	return 0;
1004 }
1005 
1006 static int
1007 eth_igbvf_dev_uninit(struct rte_eth_dev *eth_dev)
1008 {
1009 	PMD_INIT_FUNC_TRACE();
1010 
1011 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1012 		return 0;
1013 
1014 	igbvf_dev_close(eth_dev);
1015 
1016 	return 0;
1017 }
1018 
1019 static int eth_igb_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1020 	struct rte_pci_device *pci_dev)
1021 {
1022 	return rte_eth_dev_pci_generic_probe(pci_dev,
1023 		sizeof(struct e1000_adapter), eth_igb_dev_init);
1024 }
1025 
1026 static int eth_igb_pci_remove(struct rte_pci_device *pci_dev)
1027 {
1028 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igb_dev_uninit);
1029 }
1030 
1031 static struct rte_pci_driver rte_igb_pmd = {
1032 	.id_table = pci_id_igb_map,
1033 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1034 	.probe = eth_igb_pci_probe,
1035 	.remove = eth_igb_pci_remove,
1036 };
1037 
1038 
1039 static int eth_igbvf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1040 	struct rte_pci_device *pci_dev)
1041 {
1042 	return rte_eth_dev_pci_generic_probe(pci_dev,
1043 		sizeof(struct e1000_adapter), eth_igbvf_dev_init);
1044 }
1045 
1046 static int eth_igbvf_pci_remove(struct rte_pci_device *pci_dev)
1047 {
1048 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igbvf_dev_uninit);
1049 }
1050 
1051 /*
1052  * virtual function driver struct
1053  */
1054 static struct rte_pci_driver rte_igbvf_pmd = {
1055 	.id_table = pci_id_igbvf_map,
1056 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1057 	.probe = eth_igbvf_pci_probe,
1058 	.remove = eth_igbvf_pci_remove,
1059 };
1060 
1061 static void
1062 igb_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1063 {
1064 	struct e1000_hw *hw =
1065 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1066 	/* RCTL: enable VLAN filter since VMDq always use VLAN filter */
1067 	uint32_t rctl = E1000_READ_REG(hw, E1000_RCTL);
1068 	rctl |= E1000_RCTL_VFE;
1069 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1070 }
1071 
1072 static int
1073 igb_check_mq_mode(struct rte_eth_dev *dev)
1074 {
1075 	enum rte_eth_rx_mq_mode rx_mq_mode = dev->data->dev_conf.rxmode.mq_mode;
1076 	enum rte_eth_tx_mq_mode tx_mq_mode = dev->data->dev_conf.txmode.mq_mode;
1077 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
1078 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
1079 
1080 	if ((rx_mq_mode & ETH_MQ_RX_DCB_FLAG) ||
1081 	    tx_mq_mode == ETH_MQ_TX_DCB ||
1082 	    tx_mq_mode == ETH_MQ_TX_VMDQ_DCB) {
1083 		PMD_INIT_LOG(ERR, "DCB mode is not supported.");
1084 		return -EINVAL;
1085 	}
1086 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
1087 		/* Check multi-queue mode.
1088 		 * To no break software we accept ETH_MQ_RX_NONE as this might
1089 		 * be used to turn off VLAN filter.
1090 		 */
1091 
1092 		if (rx_mq_mode == ETH_MQ_RX_NONE ||
1093 		    rx_mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1094 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
1095 			RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = 1;
1096 		} else {
1097 			/* Only support one queue on VFs.
1098 			 * RSS together with SRIOV is not supported.
1099 			 */
1100 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1101 					" wrong mq_mode rx %d.",
1102 					rx_mq_mode);
1103 			return -EINVAL;
1104 		}
1105 		/* TX mode is not used here, so mode might be ignored.*/
1106 		if (tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1107 			/* SRIOV only works in VMDq enable mode */
1108 			PMD_INIT_LOG(WARNING, "SRIOV is active,"
1109 					" TX mode %d is not supported. "
1110 					" Driver will behave as %d mode.",
1111 					tx_mq_mode, ETH_MQ_TX_VMDQ_ONLY);
1112 		}
1113 
1114 		/* check valid queue number */
1115 		if ((nb_rx_q > 1) || (nb_tx_q > 1)) {
1116 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1117 					" only support one queue on VFs.");
1118 			return -EINVAL;
1119 		}
1120 	} else {
1121 		/* To no break software that set invalid mode, only display
1122 		 * warning if invalid mode is used.
1123 		 */
1124 		if (rx_mq_mode != ETH_MQ_RX_NONE &&
1125 		    rx_mq_mode != ETH_MQ_RX_VMDQ_ONLY &&
1126 		    rx_mq_mode != ETH_MQ_RX_RSS) {
1127 			/* RSS together with VMDq not supported*/
1128 			PMD_INIT_LOG(ERR, "RX mode %d is not supported.",
1129 				     rx_mq_mode);
1130 			return -EINVAL;
1131 		}
1132 
1133 		if (tx_mq_mode != ETH_MQ_TX_NONE &&
1134 		    tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1135 			PMD_INIT_LOG(WARNING, "TX mode %d is not supported."
1136 					" Due to txmode is meaningless in this"
1137 					" driver, just ignore.",
1138 					tx_mq_mode);
1139 		}
1140 	}
1141 	return 0;
1142 }
1143 
1144 static int
1145 eth_igb_configure(struct rte_eth_dev *dev)
1146 {
1147 	struct e1000_interrupt *intr =
1148 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1149 	int ret;
1150 
1151 	PMD_INIT_FUNC_TRACE();
1152 
1153 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1154 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1155 
1156 	/* multipe queue mode checking */
1157 	ret  = igb_check_mq_mode(dev);
1158 	if (ret != 0) {
1159 		PMD_DRV_LOG(ERR, "igb_check_mq_mode fails with %d.",
1160 			    ret);
1161 		return ret;
1162 	}
1163 
1164 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1165 	PMD_INIT_FUNC_TRACE();
1166 
1167 	return 0;
1168 }
1169 
1170 static void
1171 eth_igb_rxtx_control(struct rte_eth_dev *dev,
1172 		     bool enable)
1173 {
1174 	struct e1000_hw *hw =
1175 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1176 	uint32_t tctl, rctl;
1177 
1178 	tctl = E1000_READ_REG(hw, E1000_TCTL);
1179 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1180 
1181 	if (enable) {
1182 		/* enable Tx/Rx */
1183 		tctl |= E1000_TCTL_EN;
1184 		rctl |= E1000_RCTL_EN;
1185 	} else {
1186 		/* disable Tx/Rx */
1187 		tctl &= ~E1000_TCTL_EN;
1188 		rctl &= ~E1000_RCTL_EN;
1189 	}
1190 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
1191 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1192 	E1000_WRITE_FLUSH(hw);
1193 }
1194 
1195 static int
1196 eth_igb_start(struct rte_eth_dev *dev)
1197 {
1198 	struct e1000_hw *hw =
1199 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1200 	struct e1000_adapter *adapter =
1201 		E1000_DEV_PRIVATE(dev->data->dev_private);
1202 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1203 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1204 	int ret, mask;
1205 	uint32_t intr_vector = 0;
1206 	uint32_t ctrl_ext;
1207 	uint32_t *speeds;
1208 	int num_speeds;
1209 	bool autoneg;
1210 
1211 	PMD_INIT_FUNC_TRACE();
1212 
1213 	/* disable uio/vfio intr/eventfd mapping */
1214 	rte_intr_disable(intr_handle);
1215 
1216 	/* Power up the phy. Needed to make the link go Up */
1217 	eth_igb_dev_set_link_up(dev);
1218 
1219 	/*
1220 	 * Packet Buffer Allocation (PBA)
1221 	 * Writing PBA sets the receive portion of the buffer
1222 	 * the remainder is used for the transmit buffer.
1223 	 */
1224 	if (hw->mac.type == e1000_82575) {
1225 		uint32_t pba;
1226 
1227 		pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
1228 		E1000_WRITE_REG(hw, E1000_PBA, pba);
1229 	}
1230 
1231 	/* Put the address into the Receive Address Array */
1232 	e1000_rar_set(hw, hw->mac.addr, 0);
1233 
1234 	/* Initialize the hardware */
1235 	if (igb_hardware_init(hw)) {
1236 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
1237 		return -EIO;
1238 	}
1239 	adapter->stopped = 0;
1240 
1241 	E1000_WRITE_REG(hw, E1000_VET,
1242 			RTE_ETHER_TYPE_VLAN << 16 | RTE_ETHER_TYPE_VLAN);
1243 
1244 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1245 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1246 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
1247 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
1248 	E1000_WRITE_FLUSH(hw);
1249 
1250 	/* configure PF module if SRIOV enabled */
1251 	igb_pf_host_configure(dev);
1252 
1253 	/* check and configure queue intr-vector mapping */
1254 	if ((rte_intr_cap_multiple(intr_handle) ||
1255 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
1256 	    dev->data->dev_conf.intr_conf.rxq != 0) {
1257 		intr_vector = dev->data->nb_rx_queues;
1258 		if (rte_intr_efd_enable(intr_handle, intr_vector))
1259 			return -1;
1260 	}
1261 
1262 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
1263 		intr_handle->intr_vec =
1264 			rte_zmalloc("intr_vec",
1265 				    dev->data->nb_rx_queues * sizeof(int), 0);
1266 		if (intr_handle->intr_vec == NULL) {
1267 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
1268 				     " intr_vec", dev->data->nb_rx_queues);
1269 			return -ENOMEM;
1270 		}
1271 	}
1272 
1273 	/* confiugre msix for rx interrupt */
1274 	eth_igb_configure_msix_intr(dev);
1275 
1276 	/* Configure for OS presence */
1277 	igb_init_manageability(hw);
1278 
1279 	eth_igb_tx_init(dev);
1280 
1281 	/* This can fail when allocating mbufs for descriptor rings */
1282 	ret = eth_igb_rx_init(dev);
1283 	if (ret) {
1284 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
1285 		igb_dev_clear_queues(dev);
1286 		return ret;
1287 	}
1288 
1289 	e1000_clear_hw_cntrs_base_generic(hw);
1290 
1291 	/*
1292 	 * VLAN Offload Settings
1293 	 */
1294 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
1295 			ETH_VLAN_EXTEND_MASK;
1296 	ret = eth_igb_vlan_offload_set(dev, mask);
1297 	if (ret) {
1298 		PMD_INIT_LOG(ERR, "Unable to set vlan offload");
1299 		igb_dev_clear_queues(dev);
1300 		return ret;
1301 	}
1302 
1303 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1304 		/* Enable VLAN filter since VMDq always use VLAN filter */
1305 		igb_vmdq_vlan_hw_filter_enable(dev);
1306 	}
1307 
1308 	if ((hw->mac.type == e1000_82576) || (hw->mac.type == e1000_82580) ||
1309 		(hw->mac.type == e1000_i350) || (hw->mac.type == e1000_i210) ||
1310 		(hw->mac.type == e1000_i211)) {
1311 		/* Configure EITR with the maximum possible value (0xFFFF) */
1312 		E1000_WRITE_REG(hw, E1000_EITR(0), 0xFFFF);
1313 	}
1314 
1315 	/* Setup link speed and duplex */
1316 	speeds = &dev->data->dev_conf.link_speeds;
1317 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
1318 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
1319 		hw->mac.autoneg = 1;
1320 	} else {
1321 		num_speeds = 0;
1322 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
1323 
1324 		/* Reset */
1325 		hw->phy.autoneg_advertised = 0;
1326 
1327 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1328 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1329 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
1330 			num_speeds = -1;
1331 			goto error_invalid_config;
1332 		}
1333 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
1334 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1335 			num_speeds++;
1336 		}
1337 		if (*speeds & ETH_LINK_SPEED_10M) {
1338 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1339 			num_speeds++;
1340 		}
1341 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
1342 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1343 			num_speeds++;
1344 		}
1345 		if (*speeds & ETH_LINK_SPEED_100M) {
1346 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1347 			num_speeds++;
1348 		}
1349 		if (*speeds & ETH_LINK_SPEED_1G) {
1350 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1351 			num_speeds++;
1352 		}
1353 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
1354 			goto error_invalid_config;
1355 
1356 		/* Set/reset the mac.autoneg based on the link speed,
1357 		 * fixed or not
1358 		 */
1359 		if (!autoneg) {
1360 			hw->mac.autoneg = 0;
1361 			hw->mac.forced_speed_duplex =
1362 					hw->phy.autoneg_advertised;
1363 		} else {
1364 			hw->mac.autoneg = 1;
1365 		}
1366 	}
1367 
1368 	e1000_setup_link(hw);
1369 
1370 	if (rte_intr_allow_others(intr_handle)) {
1371 		/* check if lsc interrupt is enabled */
1372 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1373 			eth_igb_lsc_interrupt_setup(dev, TRUE);
1374 		else
1375 			eth_igb_lsc_interrupt_setup(dev, FALSE);
1376 	} else {
1377 		rte_intr_callback_unregister(intr_handle,
1378 					     eth_igb_interrupt_handler,
1379 					     (void *)dev);
1380 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1381 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
1382 				     " no intr multiplex");
1383 	}
1384 
1385 	/* check if rxq interrupt is enabled */
1386 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
1387 	    rte_intr_dp_is_en(intr_handle))
1388 		eth_igb_rxq_interrupt_setup(dev);
1389 
1390 	/* enable uio/vfio intr/eventfd mapping */
1391 	rte_intr_enable(intr_handle);
1392 
1393 	/* resume enabled intr since hw reset */
1394 	igb_intr_enable(dev);
1395 
1396 	/* restore all types filter */
1397 	igb_filter_restore(dev);
1398 
1399 	eth_igb_rxtx_control(dev, true);
1400 	eth_igb_link_update(dev, 0);
1401 
1402 	PMD_INIT_LOG(DEBUG, "<<");
1403 
1404 	return 0;
1405 
1406 error_invalid_config:
1407 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
1408 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
1409 	igb_dev_clear_queues(dev);
1410 	return -EINVAL;
1411 }
1412 
1413 /*********************************************************************
1414  *
1415  *  This routine disables all traffic on the adapter by issuing a
1416  *  global reset on the MAC.
1417  *
1418  **********************************************************************/
1419 static int
1420 eth_igb_stop(struct rte_eth_dev *dev)
1421 {
1422 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1423 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1424 	struct rte_eth_link link;
1425 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1426 	struct e1000_adapter *adapter =
1427 		E1000_DEV_PRIVATE(dev->data->dev_private);
1428 
1429 	if (adapter->stopped)
1430 		return 0;
1431 
1432 	eth_igb_rxtx_control(dev, false);
1433 
1434 	igb_intr_disable(dev);
1435 
1436 	/* disable intr eventfd mapping */
1437 	rte_intr_disable(intr_handle);
1438 
1439 	igb_pf_reset_hw(hw);
1440 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1441 
1442 	/* Set bit for Go Link disconnect if PHY reset is not blocked */
1443 	if (hw->mac.type >= e1000_82580 &&
1444 	    (e1000_check_reset_block(hw) != E1000_BLK_PHY_RESET)) {
1445 		uint32_t phpm_reg;
1446 
1447 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1448 		phpm_reg |= E1000_82580_PM_GO_LINKD;
1449 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1450 	}
1451 
1452 	/* Power down the phy. Needed to make the link go Down */
1453 	eth_igb_dev_set_link_down(dev);
1454 
1455 	igb_dev_clear_queues(dev);
1456 
1457 	/* clear the recorded link status */
1458 	memset(&link, 0, sizeof(link));
1459 	rte_eth_linkstatus_set(dev, &link);
1460 
1461 	if (!rte_intr_allow_others(intr_handle))
1462 		/* resume to the default handler */
1463 		rte_intr_callback_register(intr_handle,
1464 					   eth_igb_interrupt_handler,
1465 					   (void *)dev);
1466 
1467 	/* Clean datapath event and queue/vec mapping */
1468 	rte_intr_efd_disable(intr_handle);
1469 	if (intr_handle->intr_vec != NULL) {
1470 		rte_free(intr_handle->intr_vec);
1471 		intr_handle->intr_vec = NULL;
1472 	}
1473 
1474 	adapter->stopped = true;
1475 	dev->data->dev_started = 0;
1476 
1477 	return 0;
1478 }
1479 
1480 static int
1481 eth_igb_dev_set_link_up(struct rte_eth_dev *dev)
1482 {
1483 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1484 
1485 	if (hw->phy.media_type == e1000_media_type_copper)
1486 		e1000_power_up_phy(hw);
1487 	else
1488 		e1000_power_up_fiber_serdes_link(hw);
1489 
1490 	return 0;
1491 }
1492 
1493 static int
1494 eth_igb_dev_set_link_down(struct rte_eth_dev *dev)
1495 {
1496 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1497 
1498 	if (hw->phy.media_type == e1000_media_type_copper)
1499 		e1000_power_down_phy(hw);
1500 	else
1501 		e1000_shutdown_fiber_serdes_link(hw);
1502 
1503 	return 0;
1504 }
1505 
1506 static int
1507 eth_igb_close(struct rte_eth_dev *dev)
1508 {
1509 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1510 	struct rte_eth_link link;
1511 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1512 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1513 	struct e1000_filter_info *filter_info =
1514 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
1515 	int ret;
1516 
1517 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1518 		return 0;
1519 
1520 	ret = eth_igb_stop(dev);
1521 
1522 	e1000_phy_hw_reset(hw);
1523 	igb_release_manageability(hw);
1524 	igb_hw_control_release(hw);
1525 
1526 	/* Clear bit for Go Link disconnect if PHY reset is not blocked */
1527 	if (hw->mac.type >= e1000_82580 &&
1528 	    (e1000_check_reset_block(hw) != E1000_BLK_PHY_RESET)) {
1529 		uint32_t phpm_reg;
1530 
1531 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1532 		phpm_reg &= ~E1000_82580_PM_GO_LINKD;
1533 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1534 	}
1535 
1536 	igb_dev_free_queues(dev);
1537 
1538 	if (intr_handle->intr_vec) {
1539 		rte_free(intr_handle->intr_vec);
1540 		intr_handle->intr_vec = NULL;
1541 	}
1542 
1543 	memset(&link, 0, sizeof(link));
1544 	rte_eth_linkstatus_set(dev, &link);
1545 
1546 	/* Reset any pending lock */
1547 	igb_reset_swfw_lock(hw);
1548 
1549 	/* uninitialize PF if max_vfs not zero */
1550 	igb_pf_host_uninit(dev);
1551 
1552 	rte_intr_callback_unregister(intr_handle,
1553 				     eth_igb_interrupt_handler, dev);
1554 
1555 	/* clear the SYN filter info */
1556 	filter_info->syn_info = 0;
1557 
1558 	/* clear the ethertype filters info */
1559 	filter_info->ethertype_mask = 0;
1560 	memset(filter_info->ethertype_filters, 0,
1561 		E1000_MAX_ETQF_FILTERS * sizeof(struct igb_ethertype_filter));
1562 
1563 	/* clear the rss filter info */
1564 	memset(&filter_info->rss_info, 0,
1565 		sizeof(struct igb_rte_flow_rss_conf));
1566 
1567 	/* remove all ntuple filters of the device */
1568 	igb_ntuple_filter_uninit(dev);
1569 
1570 	/* remove all flex filters of the device */
1571 	igb_flex_filter_uninit(dev);
1572 
1573 	/* clear all the filters list */
1574 	igb_filterlist_flush(dev);
1575 
1576 	return ret;
1577 }
1578 
1579 /*
1580  * Reset PF device.
1581  */
1582 static int
1583 eth_igb_reset(struct rte_eth_dev *dev)
1584 {
1585 	int ret;
1586 
1587 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
1588 	 * its VF to make them align with it. The detailed notification
1589 	 * mechanism is PMD specific and is currently not implemented.
1590 	 * To avoid unexpected behavior in VF, currently reset of PF with
1591 	 * SR-IOV activation is not supported. It might be supported later.
1592 	 */
1593 	if (dev->data->sriov.active)
1594 		return -ENOTSUP;
1595 
1596 	ret = eth_igb_dev_uninit(dev);
1597 	if (ret)
1598 		return ret;
1599 
1600 	ret = eth_igb_dev_init(dev);
1601 
1602 	return ret;
1603 }
1604 
1605 
1606 static int
1607 igb_get_rx_buffer_size(struct e1000_hw *hw)
1608 {
1609 	uint32_t rx_buf_size;
1610 	if (hw->mac.type == e1000_82576) {
1611 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xffff) << 10;
1612 	} else if (hw->mac.type == e1000_82580 || hw->mac.type == e1000_i350) {
1613 		/* PBS needs to be translated according to a lookup table */
1614 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xf);
1615 		rx_buf_size = (uint32_t) e1000_rxpbs_adjust_82580(rx_buf_size);
1616 		rx_buf_size = (rx_buf_size << 10);
1617 	} else if (hw->mac.type == e1000_i210 || hw->mac.type == e1000_i211) {
1618 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0x3f) << 10;
1619 	} else {
1620 		rx_buf_size = (E1000_READ_REG(hw, E1000_PBA) & 0xffff) << 10;
1621 	}
1622 
1623 	return rx_buf_size;
1624 }
1625 
1626 /*********************************************************************
1627  *
1628  *  Initialize the hardware
1629  *
1630  **********************************************************************/
1631 static int
1632 igb_hardware_init(struct e1000_hw *hw)
1633 {
1634 	uint32_t rx_buf_size;
1635 	int diag;
1636 
1637 	/* Let the firmware know the OS is in control */
1638 	igb_hw_control_acquire(hw);
1639 
1640 	/*
1641 	 * These parameters control the automatic generation (Tx) and
1642 	 * response (Rx) to Ethernet PAUSE frames.
1643 	 * - High water mark should allow for at least two standard size (1518)
1644 	 *   frames to be received after sending an XOFF.
1645 	 * - Low water mark works best when it is very near the high water mark.
1646 	 *   This allows the receiver to restart by sending XON when it has
1647 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
1648 	 *   restart after one full frame is pulled from the buffer. There
1649 	 *   could be several smaller frames in the buffer and if so they will
1650 	 *   not trigger the XON until their total number reduces the buffer
1651 	 *   by 1500.
1652 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
1653 	 */
1654 	rx_buf_size = igb_get_rx_buffer_size(hw);
1655 
1656 	hw->fc.high_water = rx_buf_size - (RTE_ETHER_MAX_LEN * 2);
1657 	hw->fc.low_water = hw->fc.high_water - 1500;
1658 	hw->fc.pause_time = IGB_FC_PAUSE_TIME;
1659 	hw->fc.send_xon = 1;
1660 
1661 	/* Set Flow control, use the tunable location if sane */
1662 	if ((igb_fc_setting != e1000_fc_none) && (igb_fc_setting < 4))
1663 		hw->fc.requested_mode = igb_fc_setting;
1664 	else
1665 		hw->fc.requested_mode = e1000_fc_none;
1666 
1667 	/* Issue a global reset */
1668 	igb_pf_reset_hw(hw);
1669 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1670 
1671 	diag = e1000_init_hw(hw);
1672 	if (diag < 0)
1673 		return diag;
1674 
1675 	E1000_WRITE_REG(hw, E1000_VET,
1676 			RTE_ETHER_TYPE_VLAN << 16 | RTE_ETHER_TYPE_VLAN);
1677 	e1000_get_phy_info(hw);
1678 	e1000_check_for_link(hw);
1679 
1680 	return 0;
1681 }
1682 
1683 /* This function is based on igb_update_stats_counters() in igb/if_igb.c */
1684 static void
1685 igb_read_stats_registers(struct e1000_hw *hw, struct e1000_hw_stats *stats)
1686 {
1687 	int pause_frames;
1688 
1689 	uint64_t old_gprc  = stats->gprc;
1690 	uint64_t old_gptc  = stats->gptc;
1691 	uint64_t old_tpr   = stats->tpr;
1692 	uint64_t old_tpt   = stats->tpt;
1693 	uint64_t old_rpthc = stats->rpthc;
1694 	uint64_t old_hgptc = stats->hgptc;
1695 
1696 	if(hw->phy.media_type == e1000_media_type_copper ||
1697 	    (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
1698 		stats->symerrs +=
1699 		    E1000_READ_REG(hw,E1000_SYMERRS);
1700 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
1701 	}
1702 
1703 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
1704 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
1705 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
1706 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
1707 
1708 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
1709 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
1710 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
1711 	stats->dc += E1000_READ_REG(hw, E1000_DC);
1712 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
1713 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
1714 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
1715 	/*
1716 	** For watchdog management we need to know if we have been
1717 	** paused during the last interval, so capture that here.
1718 	*/
1719 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
1720 	stats->xoffrxc += pause_frames;
1721 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
1722 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
1723 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
1724 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
1725 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
1726 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
1727 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
1728 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
1729 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
1730 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
1731 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
1732 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
1733 
1734 	/* For the 64-bit byte counters the low dword must be read first. */
1735 	/* Both registers clear on the read of the high dword */
1736 
1737 	/* Workaround CRC bytes included in size, take away 4 bytes/packet */
1738 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
1739 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
1740 	stats->gorc -= (stats->gprc - old_gprc) * RTE_ETHER_CRC_LEN;
1741 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
1742 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
1743 	stats->gotc -= (stats->gptc - old_gptc) * RTE_ETHER_CRC_LEN;
1744 
1745 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
1746 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
1747 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
1748 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
1749 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
1750 
1751 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
1752 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
1753 
1754 	stats->tor += E1000_READ_REG(hw, E1000_TORL);
1755 	stats->tor += ((uint64_t)E1000_READ_REG(hw, E1000_TORH) << 32);
1756 	stats->tor -= (stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
1757 	stats->tot += E1000_READ_REG(hw, E1000_TOTL);
1758 	stats->tot += ((uint64_t)E1000_READ_REG(hw, E1000_TOTH) << 32);
1759 	stats->tot -= (stats->tpt - old_tpt) * RTE_ETHER_CRC_LEN;
1760 
1761 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
1762 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
1763 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
1764 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
1765 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
1766 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
1767 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
1768 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
1769 
1770 	/* Interrupt Counts */
1771 
1772 	stats->iac += E1000_READ_REG(hw, E1000_IAC);
1773 	stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
1774 	stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
1775 	stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
1776 	stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
1777 	stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
1778 	stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
1779 	stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
1780 	stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
1781 
1782 	/* Host to Card Statistics */
1783 
1784 	stats->cbtmpc += E1000_READ_REG(hw, E1000_CBTMPC);
1785 	stats->htdpmc += E1000_READ_REG(hw, E1000_HTDPMC);
1786 	stats->cbrdpc += E1000_READ_REG(hw, E1000_CBRDPC);
1787 	stats->cbrmpc += E1000_READ_REG(hw, E1000_CBRMPC);
1788 	stats->rpthc += E1000_READ_REG(hw, E1000_RPTHC);
1789 	stats->hgptc += E1000_READ_REG(hw, E1000_HGPTC);
1790 	stats->htcbdpc += E1000_READ_REG(hw, E1000_HTCBDPC);
1791 	stats->hgorc += E1000_READ_REG(hw, E1000_HGORCL);
1792 	stats->hgorc += ((uint64_t)E1000_READ_REG(hw, E1000_HGORCH) << 32);
1793 	stats->hgorc -= (stats->rpthc - old_rpthc) * RTE_ETHER_CRC_LEN;
1794 	stats->hgotc += E1000_READ_REG(hw, E1000_HGOTCL);
1795 	stats->hgotc += ((uint64_t)E1000_READ_REG(hw, E1000_HGOTCH) << 32);
1796 	stats->hgotc -= (stats->hgptc - old_hgptc) * RTE_ETHER_CRC_LEN;
1797 	stats->lenerrs += E1000_READ_REG(hw, E1000_LENERRS);
1798 	stats->scvpc += E1000_READ_REG(hw, E1000_SCVPC);
1799 	stats->hrmpc += E1000_READ_REG(hw, E1000_HRMPC);
1800 
1801 	stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
1802 	stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
1803 	stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
1804 	stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
1805 	stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
1806 	stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
1807 }
1808 
1809 static int
1810 eth_igb_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1811 {
1812 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1813 	struct e1000_hw_stats *stats =
1814 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1815 
1816 	igb_read_stats_registers(hw, stats);
1817 
1818 	if (rte_stats == NULL)
1819 		return -EINVAL;
1820 
1821 	/* Rx Errors */
1822 	rte_stats->imissed = stats->mpc;
1823 	rte_stats->ierrors = stats->crcerrs + stats->rlec +
1824 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
1825 
1826 	/* Tx Errors */
1827 	rte_stats->oerrors = stats->ecol + stats->latecol;
1828 
1829 	rte_stats->ipackets = stats->gprc;
1830 	rte_stats->opackets = stats->gptc;
1831 	rte_stats->ibytes   = stats->gorc;
1832 	rte_stats->obytes   = stats->gotc;
1833 	return 0;
1834 }
1835 
1836 static int
1837 eth_igb_stats_reset(struct rte_eth_dev *dev)
1838 {
1839 	struct e1000_hw_stats *hw_stats =
1840 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1841 
1842 	/* HW registers are cleared on read */
1843 	eth_igb_stats_get(dev, NULL);
1844 
1845 	/* Reset software totals */
1846 	memset(hw_stats, 0, sizeof(*hw_stats));
1847 
1848 	return 0;
1849 }
1850 
1851 static int
1852 eth_igb_xstats_reset(struct rte_eth_dev *dev)
1853 {
1854 	struct e1000_hw_stats *stats =
1855 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1856 
1857 	/* HW registers are cleared on read */
1858 	eth_igb_xstats_get(dev, NULL, IGB_NB_XSTATS);
1859 
1860 	/* Reset software totals */
1861 	memset(stats, 0, sizeof(*stats));
1862 
1863 	return 0;
1864 }
1865 
1866 static int eth_igb_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1867 	struct rte_eth_xstat_name *xstats_names,
1868 	__rte_unused unsigned int size)
1869 {
1870 	unsigned i;
1871 
1872 	if (xstats_names == NULL)
1873 		return IGB_NB_XSTATS;
1874 
1875 	/* Note: limit checked in rte_eth_xstats_names() */
1876 
1877 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1878 		strlcpy(xstats_names[i].name, rte_igb_stats_strings[i].name,
1879 			sizeof(xstats_names[i].name));
1880 	}
1881 
1882 	return IGB_NB_XSTATS;
1883 }
1884 
1885 static int eth_igb_xstats_get_names_by_id(struct rte_eth_dev *dev,
1886 		struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
1887 		unsigned int limit)
1888 {
1889 	unsigned int i;
1890 
1891 	if (!ids) {
1892 		if (xstats_names == NULL)
1893 			return IGB_NB_XSTATS;
1894 
1895 		for (i = 0; i < IGB_NB_XSTATS; i++)
1896 			strlcpy(xstats_names[i].name,
1897 				rte_igb_stats_strings[i].name,
1898 				sizeof(xstats_names[i].name));
1899 
1900 		return IGB_NB_XSTATS;
1901 
1902 	} else {
1903 		struct rte_eth_xstat_name xstats_names_copy[IGB_NB_XSTATS];
1904 
1905 		eth_igb_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
1906 				IGB_NB_XSTATS);
1907 
1908 		for (i = 0; i < limit; i++) {
1909 			if (ids[i] >= IGB_NB_XSTATS) {
1910 				PMD_INIT_LOG(ERR, "id value isn't valid");
1911 				return -1;
1912 			}
1913 			strcpy(xstats_names[i].name,
1914 					xstats_names_copy[ids[i]].name);
1915 		}
1916 		return limit;
1917 	}
1918 }
1919 
1920 static int
1921 eth_igb_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1922 		   unsigned n)
1923 {
1924 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1925 	struct e1000_hw_stats *hw_stats =
1926 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1927 	unsigned i;
1928 
1929 	if (n < IGB_NB_XSTATS)
1930 		return IGB_NB_XSTATS;
1931 
1932 	igb_read_stats_registers(hw, hw_stats);
1933 
1934 	/* If this is a reset xstats is NULL, and we have cleared the
1935 	 * registers by reading them.
1936 	 */
1937 	if (!xstats)
1938 		return 0;
1939 
1940 	/* Extended stats */
1941 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1942 		xstats[i].id = i;
1943 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1944 			rte_igb_stats_strings[i].offset);
1945 	}
1946 
1947 	return IGB_NB_XSTATS;
1948 }
1949 
1950 static int
1951 eth_igb_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
1952 		uint64_t *values, unsigned int n)
1953 {
1954 	unsigned int i;
1955 
1956 	if (!ids) {
1957 		struct e1000_hw *hw =
1958 			E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1959 		struct e1000_hw_stats *hw_stats =
1960 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1961 
1962 		if (n < IGB_NB_XSTATS)
1963 			return IGB_NB_XSTATS;
1964 
1965 		igb_read_stats_registers(hw, hw_stats);
1966 
1967 		/* If this is a reset xstats is NULL, and we have cleared the
1968 		 * registers by reading them.
1969 		 */
1970 		if (!values)
1971 			return 0;
1972 
1973 		/* Extended stats */
1974 		for (i = 0; i < IGB_NB_XSTATS; i++)
1975 			values[i] = *(uint64_t *)(((char *)hw_stats) +
1976 					rte_igb_stats_strings[i].offset);
1977 
1978 		return IGB_NB_XSTATS;
1979 
1980 	} else {
1981 		uint64_t values_copy[IGB_NB_XSTATS];
1982 
1983 		eth_igb_xstats_get_by_id(dev, NULL, values_copy,
1984 				IGB_NB_XSTATS);
1985 
1986 		for (i = 0; i < n; i++) {
1987 			if (ids[i] >= IGB_NB_XSTATS) {
1988 				PMD_INIT_LOG(ERR, "id value isn't valid");
1989 				return -1;
1990 			}
1991 			values[i] = values_copy[ids[i]];
1992 		}
1993 		return n;
1994 	}
1995 }
1996 
1997 static void
1998 igbvf_read_stats_registers(struct e1000_hw *hw, struct e1000_vf_stats *hw_stats)
1999 {
2000 	/* Good Rx packets, include VF loopback */
2001 	UPDATE_VF_STAT(E1000_VFGPRC,
2002 	    hw_stats->last_gprc, hw_stats->gprc);
2003 
2004 	/* Good Rx octets, include VF loopback */
2005 	UPDATE_VF_STAT(E1000_VFGORC,
2006 	    hw_stats->last_gorc, hw_stats->gorc);
2007 
2008 	/* Good Tx packets, include VF loopback */
2009 	UPDATE_VF_STAT(E1000_VFGPTC,
2010 	    hw_stats->last_gptc, hw_stats->gptc);
2011 
2012 	/* Good Tx octets, include VF loopback */
2013 	UPDATE_VF_STAT(E1000_VFGOTC,
2014 	    hw_stats->last_gotc, hw_stats->gotc);
2015 
2016 	/* Rx Multicst packets */
2017 	UPDATE_VF_STAT(E1000_VFMPRC,
2018 	    hw_stats->last_mprc, hw_stats->mprc);
2019 
2020 	/* Good Rx loopback packets */
2021 	UPDATE_VF_STAT(E1000_VFGPRLBC,
2022 	    hw_stats->last_gprlbc, hw_stats->gprlbc);
2023 
2024 	/* Good Rx loopback octets */
2025 	UPDATE_VF_STAT(E1000_VFGORLBC,
2026 	    hw_stats->last_gorlbc, hw_stats->gorlbc);
2027 
2028 	/* Good Tx loopback packets */
2029 	UPDATE_VF_STAT(E1000_VFGPTLBC,
2030 	    hw_stats->last_gptlbc, hw_stats->gptlbc);
2031 
2032 	/* Good Tx loopback octets */
2033 	UPDATE_VF_STAT(E1000_VFGOTLBC,
2034 	    hw_stats->last_gotlbc, hw_stats->gotlbc);
2035 }
2036 
2037 static int eth_igbvf_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
2038 				     struct rte_eth_xstat_name *xstats_names,
2039 				     __rte_unused unsigned limit)
2040 {
2041 	unsigned i;
2042 
2043 	if (xstats_names != NULL)
2044 		for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2045 			strlcpy(xstats_names[i].name,
2046 				rte_igbvf_stats_strings[i].name,
2047 				sizeof(xstats_names[i].name));
2048 		}
2049 	return IGBVF_NB_XSTATS;
2050 }
2051 
2052 static int
2053 eth_igbvf_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
2054 		     unsigned n)
2055 {
2056 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2057 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2058 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2059 	unsigned i;
2060 
2061 	if (n < IGBVF_NB_XSTATS)
2062 		return IGBVF_NB_XSTATS;
2063 
2064 	igbvf_read_stats_registers(hw, hw_stats);
2065 
2066 	if (!xstats)
2067 		return 0;
2068 
2069 	for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2070 		xstats[i].id = i;
2071 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
2072 			rte_igbvf_stats_strings[i].offset);
2073 	}
2074 
2075 	return IGBVF_NB_XSTATS;
2076 }
2077 
2078 static int
2079 eth_igbvf_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
2080 {
2081 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2082 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2083 			  E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2084 
2085 	igbvf_read_stats_registers(hw, hw_stats);
2086 
2087 	if (rte_stats == NULL)
2088 		return -EINVAL;
2089 
2090 	rte_stats->ipackets = hw_stats->gprc;
2091 	rte_stats->ibytes = hw_stats->gorc;
2092 	rte_stats->opackets = hw_stats->gptc;
2093 	rte_stats->obytes = hw_stats->gotc;
2094 	return 0;
2095 }
2096 
2097 static int
2098 eth_igbvf_stats_reset(struct rte_eth_dev *dev)
2099 {
2100 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats*)
2101 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2102 
2103 	/* Sync HW register to the last stats */
2104 	eth_igbvf_stats_get(dev, NULL);
2105 
2106 	/* reset HW current stats*/
2107 	memset(&hw_stats->gprc, 0, sizeof(*hw_stats) -
2108 	       offsetof(struct e1000_vf_stats, gprc));
2109 
2110 	return 0;
2111 }
2112 
2113 static int
2114 eth_igb_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
2115 		       size_t fw_size)
2116 {
2117 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2118 	struct e1000_fw_version fw;
2119 	int ret;
2120 
2121 	e1000_get_fw_version(hw, &fw);
2122 
2123 	switch (hw->mac.type) {
2124 	case e1000_i210:
2125 	case e1000_i211:
2126 		if (!(e1000_get_flash_presence_i210(hw))) {
2127 			ret = snprintf(fw_version, fw_size,
2128 				 "%2d.%2d-%d",
2129 				 fw.invm_major, fw.invm_minor,
2130 				 fw.invm_img_type);
2131 			break;
2132 		}
2133 		/* fall through */
2134 	default:
2135 		/* if option rom is valid, display its version too */
2136 		if (fw.or_valid) {
2137 			ret = snprintf(fw_version, fw_size,
2138 				 "%d.%d, 0x%08x, %d.%d.%d",
2139 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
2140 				 fw.or_major, fw.or_build, fw.or_patch);
2141 		/* no option rom */
2142 		} else {
2143 			if (fw.etrack_id != 0X0000) {
2144 				ret = snprintf(fw_version, fw_size,
2145 					 "%d.%d, 0x%08x",
2146 					 fw.eep_major, fw.eep_minor,
2147 					 fw.etrack_id);
2148 			} else {
2149 				ret = snprintf(fw_version, fw_size,
2150 					 "%d.%d.%d",
2151 					 fw.eep_major, fw.eep_minor,
2152 					 fw.eep_build);
2153 			}
2154 		}
2155 		break;
2156 	}
2157 	if (ret < 0)
2158 		return -EINVAL;
2159 
2160 	ret += 1; /* add the size of '\0' */
2161 	if (fw_size < (size_t)ret)
2162 		return ret;
2163 	else
2164 		return 0;
2165 }
2166 
2167 static int
2168 eth_igb_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2169 {
2170 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2171 
2172 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2173 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2174 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2175 	dev_info->rx_queue_offload_capa = igb_get_rx_queue_offloads_capa(dev);
2176 	dev_info->rx_offload_capa = igb_get_rx_port_offloads_capa(dev) |
2177 				    dev_info->rx_queue_offload_capa;
2178 	dev_info->tx_queue_offload_capa = igb_get_tx_queue_offloads_capa(dev);
2179 	dev_info->tx_offload_capa = igb_get_tx_port_offloads_capa(dev) |
2180 				    dev_info->tx_queue_offload_capa;
2181 
2182 	switch (hw->mac.type) {
2183 	case e1000_82575:
2184 		dev_info->max_rx_queues = 4;
2185 		dev_info->max_tx_queues = 4;
2186 		dev_info->max_vmdq_pools = 0;
2187 		break;
2188 
2189 	case e1000_82576:
2190 		dev_info->max_rx_queues = 16;
2191 		dev_info->max_tx_queues = 16;
2192 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2193 		dev_info->vmdq_queue_num = 16;
2194 		break;
2195 
2196 	case e1000_82580:
2197 		dev_info->max_rx_queues = 8;
2198 		dev_info->max_tx_queues = 8;
2199 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2200 		dev_info->vmdq_queue_num = 8;
2201 		break;
2202 
2203 	case e1000_i350:
2204 		dev_info->max_rx_queues = 8;
2205 		dev_info->max_tx_queues = 8;
2206 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2207 		dev_info->vmdq_queue_num = 8;
2208 		break;
2209 
2210 	case e1000_i354:
2211 		dev_info->max_rx_queues = 8;
2212 		dev_info->max_tx_queues = 8;
2213 		break;
2214 
2215 	case e1000_i210:
2216 		dev_info->max_rx_queues = 4;
2217 		dev_info->max_tx_queues = 4;
2218 		dev_info->max_vmdq_pools = 0;
2219 		break;
2220 
2221 	case e1000_i211:
2222 		dev_info->max_rx_queues = 2;
2223 		dev_info->max_tx_queues = 2;
2224 		dev_info->max_vmdq_pools = 0;
2225 		break;
2226 
2227 	default:
2228 		/* Should not happen */
2229 		return -EINVAL;
2230 	}
2231 	dev_info->hash_key_size = IGB_HKEY_MAX_INDEX * sizeof(uint32_t);
2232 	dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
2233 	dev_info->flow_type_rss_offloads = IGB_RSS_OFFLOAD_ALL;
2234 
2235 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2236 		.rx_thresh = {
2237 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2238 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2239 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2240 		},
2241 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2242 		.rx_drop_en = 0,
2243 		.offloads = 0,
2244 	};
2245 
2246 	dev_info->default_txconf = (struct rte_eth_txconf) {
2247 		.tx_thresh = {
2248 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2249 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2250 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2251 		},
2252 		.offloads = 0,
2253 	};
2254 
2255 	dev_info->rx_desc_lim = rx_desc_lim;
2256 	dev_info->tx_desc_lim = tx_desc_lim;
2257 
2258 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
2259 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
2260 			ETH_LINK_SPEED_1G;
2261 
2262 	dev_info->max_mtu = dev_info->max_rx_pktlen - E1000_ETH_OVERHEAD;
2263 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
2264 
2265 	return 0;
2266 }
2267 
2268 static const uint32_t *
2269 eth_igb_supported_ptypes_get(struct rte_eth_dev *dev)
2270 {
2271 	static const uint32_t ptypes[] = {
2272 		/* refers to igb_rxd_pkt_info_to_pkt_type() */
2273 		RTE_PTYPE_L2_ETHER,
2274 		RTE_PTYPE_L3_IPV4,
2275 		RTE_PTYPE_L3_IPV4_EXT,
2276 		RTE_PTYPE_L3_IPV6,
2277 		RTE_PTYPE_L3_IPV6_EXT,
2278 		RTE_PTYPE_L4_TCP,
2279 		RTE_PTYPE_L4_UDP,
2280 		RTE_PTYPE_L4_SCTP,
2281 		RTE_PTYPE_TUNNEL_IP,
2282 		RTE_PTYPE_INNER_L3_IPV6,
2283 		RTE_PTYPE_INNER_L3_IPV6_EXT,
2284 		RTE_PTYPE_INNER_L4_TCP,
2285 		RTE_PTYPE_INNER_L4_UDP,
2286 		RTE_PTYPE_UNKNOWN
2287 	};
2288 
2289 	if (dev->rx_pkt_burst == eth_igb_recv_pkts ||
2290 	    dev->rx_pkt_burst == eth_igb_recv_scattered_pkts)
2291 		return ptypes;
2292 	return NULL;
2293 }
2294 
2295 static int
2296 eth_igbvf_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2297 {
2298 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2299 
2300 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2301 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2302 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2303 	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
2304 				DEV_TX_OFFLOAD_IPV4_CKSUM  |
2305 				DEV_TX_OFFLOAD_UDP_CKSUM   |
2306 				DEV_TX_OFFLOAD_TCP_CKSUM   |
2307 				DEV_TX_OFFLOAD_SCTP_CKSUM  |
2308 				DEV_TX_OFFLOAD_TCP_TSO;
2309 	switch (hw->mac.type) {
2310 	case e1000_vfadapt:
2311 		dev_info->max_rx_queues = 2;
2312 		dev_info->max_tx_queues = 2;
2313 		break;
2314 	case e1000_vfadapt_i350:
2315 		dev_info->max_rx_queues = 1;
2316 		dev_info->max_tx_queues = 1;
2317 		break;
2318 	default:
2319 		/* Should not happen */
2320 		return -EINVAL;
2321 	}
2322 
2323 	dev_info->rx_queue_offload_capa = igb_get_rx_queue_offloads_capa(dev);
2324 	dev_info->rx_offload_capa = igb_get_rx_port_offloads_capa(dev) |
2325 				    dev_info->rx_queue_offload_capa;
2326 	dev_info->tx_queue_offload_capa = igb_get_tx_queue_offloads_capa(dev);
2327 	dev_info->tx_offload_capa = igb_get_tx_port_offloads_capa(dev) |
2328 				    dev_info->tx_queue_offload_capa;
2329 
2330 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2331 		.rx_thresh = {
2332 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2333 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2334 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2335 		},
2336 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2337 		.rx_drop_en = 0,
2338 		.offloads = 0,
2339 	};
2340 
2341 	dev_info->default_txconf = (struct rte_eth_txconf) {
2342 		.tx_thresh = {
2343 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2344 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2345 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2346 		},
2347 		.offloads = 0,
2348 	};
2349 
2350 	dev_info->rx_desc_lim = rx_desc_lim;
2351 	dev_info->tx_desc_lim = tx_desc_lim;
2352 
2353 	return 0;
2354 }
2355 
2356 /* return 0 means link status changed, -1 means not changed */
2357 static int
2358 eth_igb_link_update(struct rte_eth_dev *dev, int wait_to_complete)
2359 {
2360 	struct e1000_hw *hw =
2361 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2362 	struct rte_eth_link link;
2363 	int link_check, count;
2364 
2365 	link_check = 0;
2366 	hw->mac.get_link_status = 1;
2367 
2368 	/* possible wait-to-complete in up to 9 seconds */
2369 	for (count = 0; count < IGB_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
2370 		/* Read the real link status */
2371 		switch (hw->phy.media_type) {
2372 		case e1000_media_type_copper:
2373 			/* Do the work to read phy */
2374 			e1000_check_for_link(hw);
2375 			link_check = !hw->mac.get_link_status;
2376 			break;
2377 
2378 		case e1000_media_type_fiber:
2379 			e1000_check_for_link(hw);
2380 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
2381 				      E1000_STATUS_LU);
2382 			break;
2383 
2384 		case e1000_media_type_internal_serdes:
2385 			e1000_check_for_link(hw);
2386 			link_check = hw->mac.serdes_has_link;
2387 			break;
2388 
2389 		/* VF device is type_unknown */
2390 		case e1000_media_type_unknown:
2391 			eth_igbvf_link_update(hw);
2392 			link_check = !hw->mac.get_link_status;
2393 			break;
2394 
2395 		default:
2396 			break;
2397 		}
2398 		if (link_check || wait_to_complete == 0)
2399 			break;
2400 		rte_delay_ms(IGB_LINK_UPDATE_CHECK_INTERVAL);
2401 	}
2402 	memset(&link, 0, sizeof(link));
2403 
2404 	/* Now we check if a transition has happened */
2405 	if (link_check) {
2406 		uint16_t duplex, speed;
2407 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
2408 		link.link_duplex = (duplex == FULL_DUPLEX) ?
2409 				ETH_LINK_FULL_DUPLEX :
2410 				ETH_LINK_HALF_DUPLEX;
2411 		link.link_speed = speed;
2412 		link.link_status = ETH_LINK_UP;
2413 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
2414 				ETH_LINK_SPEED_FIXED);
2415 	} else if (!link_check) {
2416 		link.link_speed = 0;
2417 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
2418 		link.link_status = ETH_LINK_DOWN;
2419 		link.link_autoneg = ETH_LINK_FIXED;
2420 	}
2421 
2422 	return rte_eth_linkstatus_set(dev, &link);
2423 }
2424 
2425 /*
2426  * igb_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
2427  * For ASF and Pass Through versions of f/w this means
2428  * that the driver is loaded.
2429  */
2430 static void
2431 igb_hw_control_acquire(struct e1000_hw *hw)
2432 {
2433 	uint32_t ctrl_ext;
2434 
2435 	/* Let firmware know the driver has taken over */
2436 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2437 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2438 }
2439 
2440 /*
2441  * igb_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
2442  * For ASF and Pass Through versions of f/w this means that the
2443  * driver is no longer loaded.
2444  */
2445 static void
2446 igb_hw_control_release(struct e1000_hw *hw)
2447 {
2448 	uint32_t ctrl_ext;
2449 
2450 	/* Let firmware taken over control of h/w */
2451 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2452 	E1000_WRITE_REG(hw, E1000_CTRL_EXT,
2453 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2454 }
2455 
2456 /*
2457  * Bit of a misnomer, what this really means is
2458  * to enable OS management of the system... aka
2459  * to disable special hardware management features.
2460  */
2461 static void
2462 igb_init_manageability(struct e1000_hw *hw)
2463 {
2464 	if (e1000_enable_mng_pass_thru(hw)) {
2465 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
2466 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2467 
2468 		/* disable hardware interception of ARP */
2469 		manc &= ~(E1000_MANC_ARP_EN);
2470 
2471 		/* enable receiving management packets to the host */
2472 		manc |= E1000_MANC_EN_MNG2HOST;
2473 		manc2h |= 1 << 5;  /* Mng Port 623 */
2474 		manc2h |= 1 << 6;  /* Mng Port 664 */
2475 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
2476 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2477 	}
2478 }
2479 
2480 static void
2481 igb_release_manageability(struct e1000_hw *hw)
2482 {
2483 	if (e1000_enable_mng_pass_thru(hw)) {
2484 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2485 
2486 		manc |= E1000_MANC_ARP_EN;
2487 		manc &= ~E1000_MANC_EN_MNG2HOST;
2488 
2489 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2490 	}
2491 }
2492 
2493 static int
2494 eth_igb_promiscuous_enable(struct rte_eth_dev *dev)
2495 {
2496 	struct e1000_hw *hw =
2497 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2498 	uint32_t rctl;
2499 
2500 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2501 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
2502 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2503 
2504 	return 0;
2505 }
2506 
2507 static int
2508 eth_igb_promiscuous_disable(struct rte_eth_dev *dev)
2509 {
2510 	struct e1000_hw *hw =
2511 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2512 	uint32_t rctl;
2513 
2514 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2515 	rctl &= (~E1000_RCTL_UPE);
2516 	if (dev->data->all_multicast == 1)
2517 		rctl |= E1000_RCTL_MPE;
2518 	else
2519 		rctl &= (~E1000_RCTL_MPE);
2520 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2521 
2522 	return 0;
2523 }
2524 
2525 static int
2526 eth_igb_allmulticast_enable(struct rte_eth_dev *dev)
2527 {
2528 	struct e1000_hw *hw =
2529 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2530 	uint32_t rctl;
2531 
2532 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2533 	rctl |= E1000_RCTL_MPE;
2534 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2535 
2536 	return 0;
2537 }
2538 
2539 static int
2540 eth_igb_allmulticast_disable(struct rte_eth_dev *dev)
2541 {
2542 	struct e1000_hw *hw =
2543 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2544 	uint32_t rctl;
2545 
2546 	if (dev->data->promiscuous == 1)
2547 		return 0; /* must remain in all_multicast mode */
2548 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2549 	rctl &= (~E1000_RCTL_MPE);
2550 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2551 
2552 	return 0;
2553 }
2554 
2555 static int
2556 eth_igb_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2557 {
2558 	struct e1000_hw *hw =
2559 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2560 	struct e1000_vfta * shadow_vfta =
2561 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2562 	uint32_t vfta;
2563 	uint32_t vid_idx;
2564 	uint32_t vid_bit;
2565 
2566 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
2567 			      E1000_VFTA_ENTRY_MASK);
2568 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
2569 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
2570 	if (on)
2571 		vfta |= vid_bit;
2572 	else
2573 		vfta &= ~vid_bit;
2574 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
2575 
2576 	/* update local VFTA copy */
2577 	shadow_vfta->vfta[vid_idx] = vfta;
2578 
2579 	return 0;
2580 }
2581 
2582 static int
2583 eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
2584 		      enum rte_vlan_type vlan_type,
2585 		      uint16_t tpid)
2586 {
2587 	struct e1000_hw *hw =
2588 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2589 	uint32_t reg, qinq;
2590 
2591 	qinq = E1000_READ_REG(hw, E1000_CTRL_EXT);
2592 	qinq &= E1000_CTRL_EXT_EXT_VLAN;
2593 
2594 	/* only outer TPID of double VLAN can be configured*/
2595 	if (qinq && vlan_type == ETH_VLAN_TYPE_OUTER) {
2596 		reg = E1000_READ_REG(hw, E1000_VET);
2597 		reg = (reg & (~E1000_VET_VET_EXT)) |
2598 			((uint32_t)tpid << E1000_VET_VET_EXT_SHIFT);
2599 		E1000_WRITE_REG(hw, E1000_VET, reg);
2600 
2601 		return 0;
2602 	}
2603 
2604 	/* all other TPID values are read-only*/
2605 	PMD_DRV_LOG(ERR, "Not supported");
2606 
2607 	return -ENOTSUP;
2608 }
2609 
2610 static void
2611 igb_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2612 {
2613 	struct e1000_hw *hw =
2614 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2615 	uint32_t reg;
2616 
2617 	/* Filter Table Disable */
2618 	reg = E1000_READ_REG(hw, E1000_RCTL);
2619 	reg &= ~E1000_RCTL_CFIEN;
2620 	reg &= ~E1000_RCTL_VFE;
2621 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2622 }
2623 
2624 static void
2625 igb_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2626 {
2627 	struct e1000_hw *hw =
2628 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2629 	struct e1000_vfta * shadow_vfta =
2630 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2631 	uint32_t reg;
2632 	int i;
2633 
2634 	/* Filter Table Enable, CFI not used for packet acceptance */
2635 	reg = E1000_READ_REG(hw, E1000_RCTL);
2636 	reg &= ~E1000_RCTL_CFIEN;
2637 	reg |= E1000_RCTL_VFE;
2638 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2639 
2640 	/* restore VFTA table */
2641 	for (i = 0; i < IGB_VFTA_SIZE; i++)
2642 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
2643 }
2644 
2645 static void
2646 igb_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2647 {
2648 	struct e1000_hw *hw =
2649 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2650 	uint32_t reg;
2651 
2652 	/* VLAN Mode Disable */
2653 	reg = E1000_READ_REG(hw, E1000_CTRL);
2654 	reg &= ~E1000_CTRL_VME;
2655 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2656 }
2657 
2658 static void
2659 igb_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2660 {
2661 	struct e1000_hw *hw =
2662 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2663 	uint32_t reg;
2664 
2665 	/* VLAN Mode Enable */
2666 	reg = E1000_READ_REG(hw, E1000_CTRL);
2667 	reg |= E1000_CTRL_VME;
2668 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2669 }
2670 
2671 static void
2672 igb_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2673 {
2674 	struct e1000_hw *hw =
2675 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2676 	uint32_t reg;
2677 
2678 	/* CTRL_EXT: Extended VLAN */
2679 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2680 	reg &= ~E1000_CTRL_EXT_EXTEND_VLAN;
2681 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2682 
2683 	/* Update maximum packet length */
2684 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME)
2685 		E1000_WRITE_REG(hw, E1000_RLPML,
2686 				dev->data->dev_conf.rxmode.max_rx_pkt_len);
2687 }
2688 
2689 static void
2690 igb_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2691 {
2692 	struct e1000_hw *hw =
2693 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2694 	uint32_t reg;
2695 
2696 	/* CTRL_EXT: Extended VLAN */
2697 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2698 	reg |= E1000_CTRL_EXT_EXTEND_VLAN;
2699 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2700 
2701 	/* Update maximum packet length */
2702 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME)
2703 		E1000_WRITE_REG(hw, E1000_RLPML,
2704 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2705 						VLAN_TAG_SIZE);
2706 }
2707 
2708 static int
2709 eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2710 {
2711 	struct rte_eth_rxmode *rxmode;
2712 
2713 	rxmode = &dev->data->dev_conf.rxmode;
2714 	if(mask & ETH_VLAN_STRIP_MASK){
2715 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2716 			igb_vlan_hw_strip_enable(dev);
2717 		else
2718 			igb_vlan_hw_strip_disable(dev);
2719 	}
2720 
2721 	if(mask & ETH_VLAN_FILTER_MASK){
2722 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2723 			igb_vlan_hw_filter_enable(dev);
2724 		else
2725 			igb_vlan_hw_filter_disable(dev);
2726 	}
2727 
2728 	if(mask & ETH_VLAN_EXTEND_MASK){
2729 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2730 			igb_vlan_hw_extend_enable(dev);
2731 		else
2732 			igb_vlan_hw_extend_disable(dev);
2733 	}
2734 
2735 	return 0;
2736 }
2737 
2738 
2739 /**
2740  * It enables the interrupt mask and then enable the interrupt.
2741  *
2742  * @param dev
2743  *  Pointer to struct rte_eth_dev.
2744  * @param on
2745  *  Enable or Disable
2746  *
2747  * @return
2748  *  - On success, zero.
2749  *  - On failure, a negative value.
2750  */
2751 static int
2752 eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
2753 {
2754 	struct e1000_interrupt *intr =
2755 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2756 
2757 	if (on)
2758 		intr->mask |= E1000_ICR_LSC;
2759 	else
2760 		intr->mask &= ~E1000_ICR_LSC;
2761 
2762 	return 0;
2763 }
2764 
2765 /* It clears the interrupt causes and enables the interrupt.
2766  * It will be called once only during nic initialized.
2767  *
2768  * @param dev
2769  *  Pointer to struct rte_eth_dev.
2770  *
2771  * @return
2772  *  - On success, zero.
2773  *  - On failure, a negative value.
2774  */
2775 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev)
2776 {
2777 	uint32_t mask, regval;
2778 	int ret;
2779 	struct e1000_hw *hw =
2780 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2781 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2782 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2783 	int misc_shift = rte_intr_allow_others(intr_handle) ? 1 : 0;
2784 	struct rte_eth_dev_info dev_info;
2785 
2786 	memset(&dev_info, 0, sizeof(dev_info));
2787 	ret = eth_igb_infos_get(dev, &dev_info);
2788 	if (ret != 0)
2789 		return ret;
2790 
2791 	mask = (0xFFFFFFFF >> (32 - dev_info.max_rx_queues)) << misc_shift;
2792 	regval = E1000_READ_REG(hw, E1000_EIMS);
2793 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
2794 
2795 	return 0;
2796 }
2797 
2798 /*
2799  * It reads ICR and gets interrupt causes, check it and set a bit flag
2800  * to update link status.
2801  *
2802  * @param dev
2803  *  Pointer to struct rte_eth_dev.
2804  *
2805  * @return
2806  *  - On success, zero.
2807  *  - On failure, a negative value.
2808  */
2809 static int
2810 eth_igb_interrupt_get_status(struct rte_eth_dev *dev)
2811 {
2812 	uint32_t icr;
2813 	struct e1000_hw *hw =
2814 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2815 	struct e1000_interrupt *intr =
2816 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2817 
2818 	igb_intr_disable(dev);
2819 
2820 	/* read-on-clear nic registers here */
2821 	icr = E1000_READ_REG(hw, E1000_ICR);
2822 
2823 	intr->flags = 0;
2824 	if (icr & E1000_ICR_LSC) {
2825 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
2826 	}
2827 
2828 	if (icr & E1000_ICR_VMMB)
2829 		intr->flags |= E1000_FLAG_MAILBOX;
2830 
2831 	return 0;
2832 }
2833 
2834 /*
2835  * It executes link_update after knowing an interrupt is prsent.
2836  *
2837  * @param dev
2838  *  Pointer to struct rte_eth_dev.
2839  *
2840  * @return
2841  *  - On success, zero.
2842  *  - On failure, a negative value.
2843  */
2844 static int
2845 eth_igb_interrupt_action(struct rte_eth_dev *dev,
2846 			 struct rte_intr_handle *intr_handle)
2847 {
2848 	struct e1000_hw *hw =
2849 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2850 	struct e1000_interrupt *intr =
2851 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2852 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2853 	struct rte_eth_link link;
2854 	int ret;
2855 
2856 	if (intr->flags & E1000_FLAG_MAILBOX) {
2857 		igb_pf_mbx_process(dev);
2858 		intr->flags &= ~E1000_FLAG_MAILBOX;
2859 	}
2860 
2861 	igb_intr_enable(dev);
2862 	rte_intr_ack(intr_handle);
2863 
2864 	if (intr->flags & E1000_FLAG_NEED_LINK_UPDATE) {
2865 		intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
2866 
2867 		/* set get_link_status to check register later */
2868 		hw->mac.get_link_status = 1;
2869 		ret = eth_igb_link_update(dev, 0);
2870 
2871 		/* check if link has changed */
2872 		if (ret < 0)
2873 			return 0;
2874 
2875 		rte_eth_linkstatus_get(dev, &link);
2876 		if (link.link_status) {
2877 			PMD_INIT_LOG(INFO,
2878 				     " Port %d: Link Up - speed %u Mbps - %s",
2879 				     dev->data->port_id,
2880 				     (unsigned)link.link_speed,
2881 				     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
2882 				     "full-duplex" : "half-duplex");
2883 		} else {
2884 			PMD_INIT_LOG(INFO, " Port %d: Link Down",
2885 				     dev->data->port_id);
2886 		}
2887 
2888 		PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
2889 			     pci_dev->addr.domain,
2890 			     pci_dev->addr.bus,
2891 			     pci_dev->addr.devid,
2892 			     pci_dev->addr.function);
2893 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
2894 	}
2895 
2896 	return 0;
2897 }
2898 
2899 /**
2900  * Interrupt handler which shall be registered at first.
2901  *
2902  * @param handle
2903  *  Pointer to interrupt handle.
2904  * @param param
2905  *  The address of parameter (struct rte_eth_dev *) regsitered before.
2906  *
2907  * @return
2908  *  void
2909  */
2910 static void
2911 eth_igb_interrupt_handler(void *param)
2912 {
2913 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2914 
2915 	eth_igb_interrupt_get_status(dev);
2916 	eth_igb_interrupt_action(dev, dev->intr_handle);
2917 }
2918 
2919 static int
2920 eth_igbvf_interrupt_get_status(struct rte_eth_dev *dev)
2921 {
2922 	uint32_t eicr;
2923 	struct e1000_hw *hw =
2924 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2925 	struct e1000_interrupt *intr =
2926 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2927 
2928 	igbvf_intr_disable(hw);
2929 
2930 	/* read-on-clear nic registers here */
2931 	eicr = E1000_READ_REG(hw, E1000_EICR);
2932 	intr->flags = 0;
2933 
2934 	if (eicr == E1000_VTIVAR_MISC_MAILBOX)
2935 		intr->flags |= E1000_FLAG_MAILBOX;
2936 
2937 	return 0;
2938 }
2939 
2940 void igbvf_mbx_process(struct rte_eth_dev *dev)
2941 {
2942 	struct e1000_hw *hw =
2943 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2944 	struct e1000_mbx_info *mbx = &hw->mbx;
2945 	u32 in_msg = 0;
2946 
2947 	/* peek the message first */
2948 	in_msg = E1000_READ_REG(hw, E1000_VMBMEM(0));
2949 
2950 	/* PF reset VF event */
2951 	if (in_msg == E1000_PF_CONTROL_MSG) {
2952 		/* dummy mbx read to ack pf */
2953 		if (mbx->ops.read(hw, &in_msg, 1, 0))
2954 			return;
2955 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
2956 					     NULL);
2957 	}
2958 }
2959 
2960 static int
2961 eth_igbvf_interrupt_action(struct rte_eth_dev *dev, struct rte_intr_handle *intr_handle)
2962 {
2963 	struct e1000_interrupt *intr =
2964 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2965 
2966 	if (intr->flags & E1000_FLAG_MAILBOX) {
2967 		igbvf_mbx_process(dev);
2968 		intr->flags &= ~E1000_FLAG_MAILBOX;
2969 	}
2970 
2971 	igbvf_intr_enable(dev);
2972 	rte_intr_ack(intr_handle);
2973 
2974 	return 0;
2975 }
2976 
2977 static void
2978 eth_igbvf_interrupt_handler(void *param)
2979 {
2980 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2981 
2982 	eth_igbvf_interrupt_get_status(dev);
2983 	eth_igbvf_interrupt_action(dev, dev->intr_handle);
2984 }
2985 
2986 static int
2987 eth_igb_led_on(struct rte_eth_dev *dev)
2988 {
2989 	struct e1000_hw *hw;
2990 
2991 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2992 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
2993 }
2994 
2995 static int
2996 eth_igb_led_off(struct rte_eth_dev *dev)
2997 {
2998 	struct e1000_hw *hw;
2999 
3000 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3001 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3002 }
3003 
3004 static int
3005 eth_igb_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3006 {
3007 	struct e1000_hw *hw;
3008 	uint32_t ctrl;
3009 	int tx_pause;
3010 	int rx_pause;
3011 
3012 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3013 	fc_conf->pause_time = hw->fc.pause_time;
3014 	fc_conf->high_water = hw->fc.high_water;
3015 	fc_conf->low_water = hw->fc.low_water;
3016 	fc_conf->send_xon = hw->fc.send_xon;
3017 	fc_conf->autoneg = hw->mac.autoneg;
3018 
3019 	/*
3020 	 * Return rx_pause and tx_pause status according to actual setting of
3021 	 * the TFCE and RFCE bits in the CTRL register.
3022 	 */
3023 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
3024 	if (ctrl & E1000_CTRL_TFCE)
3025 		tx_pause = 1;
3026 	else
3027 		tx_pause = 0;
3028 
3029 	if (ctrl & E1000_CTRL_RFCE)
3030 		rx_pause = 1;
3031 	else
3032 		rx_pause = 0;
3033 
3034 	if (rx_pause && tx_pause)
3035 		fc_conf->mode = RTE_FC_FULL;
3036 	else if (rx_pause)
3037 		fc_conf->mode = RTE_FC_RX_PAUSE;
3038 	else if (tx_pause)
3039 		fc_conf->mode = RTE_FC_TX_PAUSE;
3040 	else
3041 		fc_conf->mode = RTE_FC_NONE;
3042 
3043 	return 0;
3044 }
3045 
3046 static int
3047 eth_igb_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3048 {
3049 	struct e1000_hw *hw;
3050 	int err;
3051 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
3052 		e1000_fc_none,
3053 		e1000_fc_rx_pause,
3054 		e1000_fc_tx_pause,
3055 		e1000_fc_full
3056 	};
3057 	uint32_t rx_buf_size;
3058 	uint32_t max_high_water;
3059 	uint32_t rctl;
3060 	uint32_t ctrl;
3061 
3062 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3063 	if (fc_conf->autoneg != hw->mac.autoneg)
3064 		return -ENOTSUP;
3065 	rx_buf_size = igb_get_rx_buffer_size(hw);
3066 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3067 
3068 	/* At least reserve one Ethernet frame for watermark */
3069 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
3070 	if ((fc_conf->high_water > max_high_water) ||
3071 	    (fc_conf->high_water < fc_conf->low_water)) {
3072 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
3073 		PMD_INIT_LOG(ERR, "high water must <=  0x%x", max_high_water);
3074 		return -EINVAL;
3075 	}
3076 
3077 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
3078 	hw->fc.pause_time     = fc_conf->pause_time;
3079 	hw->fc.high_water     = fc_conf->high_water;
3080 	hw->fc.low_water      = fc_conf->low_water;
3081 	hw->fc.send_xon	      = fc_conf->send_xon;
3082 
3083 	err = e1000_setup_link_generic(hw);
3084 	if (err == E1000_SUCCESS) {
3085 
3086 		/* check if we want to forward MAC frames - driver doesn't have native
3087 		 * capability to do that, so we'll write the registers ourselves */
3088 
3089 		rctl = E1000_READ_REG(hw, E1000_RCTL);
3090 
3091 		/* set or clear MFLCN.PMCF bit depending on configuration */
3092 		if (fc_conf->mac_ctrl_frame_fwd != 0)
3093 			rctl |= E1000_RCTL_PMCF;
3094 		else
3095 			rctl &= ~E1000_RCTL_PMCF;
3096 
3097 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3098 
3099 		/*
3100 		 * check if we want to change flow control mode - driver doesn't have native
3101 		 * capability to do that, so we'll write the registers ourselves
3102 		 */
3103 		ctrl = E1000_READ_REG(hw, E1000_CTRL);
3104 
3105 		/*
3106 		 * set or clear E1000_CTRL_RFCE and E1000_CTRL_TFCE bits depending
3107 		 * on configuration
3108 		 */
3109 		switch (fc_conf->mode) {
3110 		case RTE_FC_NONE:
3111 			ctrl &= ~E1000_CTRL_RFCE & ~E1000_CTRL_TFCE;
3112 			break;
3113 		case RTE_FC_RX_PAUSE:
3114 			ctrl |= E1000_CTRL_RFCE;
3115 			ctrl &= ~E1000_CTRL_TFCE;
3116 			break;
3117 		case RTE_FC_TX_PAUSE:
3118 			ctrl |= E1000_CTRL_TFCE;
3119 			ctrl &= ~E1000_CTRL_RFCE;
3120 			break;
3121 		case RTE_FC_FULL:
3122 			ctrl |= E1000_CTRL_RFCE | E1000_CTRL_TFCE;
3123 			break;
3124 		default:
3125 			PMD_INIT_LOG(ERR, "invalid flow control mode");
3126 			return -EINVAL;
3127 		}
3128 
3129 		E1000_WRITE_REG(hw, E1000_CTRL, ctrl);
3130 
3131 		E1000_WRITE_FLUSH(hw);
3132 
3133 		return 0;
3134 	}
3135 
3136 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
3137 	return -EIO;
3138 }
3139 
3140 #define E1000_RAH_POOLSEL_SHIFT      (18)
3141 static int
3142 eth_igb_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
3143 		uint32_t index, uint32_t pool)
3144 {
3145 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3146 	uint32_t rah;
3147 
3148 	e1000_rar_set(hw, mac_addr->addr_bytes, index);
3149 	rah = E1000_READ_REG(hw, E1000_RAH(index));
3150 	rah |= (0x1 << (E1000_RAH_POOLSEL_SHIFT + pool));
3151 	E1000_WRITE_REG(hw, E1000_RAH(index), rah);
3152 	return 0;
3153 }
3154 
3155 static void
3156 eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index)
3157 {
3158 	uint8_t addr[RTE_ETHER_ADDR_LEN];
3159 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3160 
3161 	memset(addr, 0, sizeof(addr));
3162 
3163 	e1000_rar_set(hw, addr, index);
3164 }
3165 
3166 static int
3167 eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
3168 				struct rte_ether_addr *addr)
3169 {
3170 	eth_igb_rar_clear(dev, 0);
3171 	eth_igb_rar_set(dev, (void *)addr, 0, 0);
3172 
3173 	return 0;
3174 }
3175 /*
3176  * Virtual Function operations
3177  */
3178 static void
3179 igbvf_intr_disable(struct e1000_hw *hw)
3180 {
3181 	PMD_INIT_FUNC_TRACE();
3182 
3183 	/* Clear interrupt mask to stop from interrupts being generated */
3184 	E1000_WRITE_REG(hw, E1000_EIMC, 0xFFFF);
3185 
3186 	E1000_WRITE_FLUSH(hw);
3187 }
3188 
3189 static void
3190 igbvf_stop_adapter(struct rte_eth_dev *dev)
3191 {
3192 	u32 reg_val;
3193 	u16 i;
3194 	struct rte_eth_dev_info dev_info;
3195 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3196 	int ret;
3197 
3198 	memset(&dev_info, 0, sizeof(dev_info));
3199 	ret = eth_igbvf_infos_get(dev, &dev_info);
3200 	if (ret != 0)
3201 		return;
3202 
3203 	/* Clear interrupt mask to stop from interrupts being generated */
3204 	igbvf_intr_disable(hw);
3205 
3206 	/* Clear any pending interrupts, flush previous writes */
3207 	E1000_READ_REG(hw, E1000_EICR);
3208 
3209 	/* Disable the transmit unit.  Each queue must be disabled. */
3210 	for (i = 0; i < dev_info.max_tx_queues; i++)
3211 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), E1000_TXDCTL_SWFLSH);
3212 
3213 	/* Disable the receive unit by stopping each queue */
3214 	for (i = 0; i < dev_info.max_rx_queues; i++) {
3215 		reg_val = E1000_READ_REG(hw, E1000_RXDCTL(i));
3216 		reg_val &= ~E1000_RXDCTL_QUEUE_ENABLE;
3217 		E1000_WRITE_REG(hw, E1000_RXDCTL(i), reg_val);
3218 		while (E1000_READ_REG(hw, E1000_RXDCTL(i)) & E1000_RXDCTL_QUEUE_ENABLE)
3219 			;
3220 	}
3221 
3222 	/* flush all queues disables */
3223 	E1000_WRITE_FLUSH(hw);
3224 	msec_delay(2);
3225 }
3226 
3227 static int eth_igbvf_link_update(struct e1000_hw *hw)
3228 {
3229 	struct e1000_mbx_info *mbx = &hw->mbx;
3230 	struct e1000_mac_info *mac = &hw->mac;
3231 	int ret_val = E1000_SUCCESS;
3232 
3233 	PMD_INIT_LOG(DEBUG, "e1000_check_for_link_vf");
3234 
3235 	/*
3236 	 * We only want to run this if there has been a rst asserted.
3237 	 * in this case that could mean a link change, device reset,
3238 	 * or a virtual function reset
3239 	 */
3240 
3241 	/* If we were hit with a reset or timeout drop the link */
3242 	if (!e1000_check_for_rst(hw, 0) || !mbx->timeout)
3243 		mac->get_link_status = TRUE;
3244 
3245 	if (!mac->get_link_status)
3246 		goto out;
3247 
3248 	/* if link status is down no point in checking to see if pf is up */
3249 	if (!(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU))
3250 		goto out;
3251 
3252 	/* if we passed all the tests above then the link is up and we no
3253 	 * longer need to check for link */
3254 	mac->get_link_status = FALSE;
3255 
3256 out:
3257 	return ret_val;
3258 }
3259 
3260 
3261 static int
3262 igbvf_dev_configure(struct rte_eth_dev *dev)
3263 {
3264 	struct rte_eth_conf* conf = &dev->data->dev_conf;
3265 
3266 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
3267 		     dev->data->port_id);
3268 
3269 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
3270 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
3271 
3272 	/*
3273 	 * VF has no ability to enable/disable HW CRC
3274 	 * Keep the persistent behavior the same as Host PF
3275 	 */
3276 #ifndef RTE_LIBRTE_E1000_PF_DISABLE_STRIP_CRC
3277 	if (conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
3278 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
3279 		conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_KEEP_CRC;
3280 	}
3281 #else
3282 	if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC)) {
3283 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
3284 		conf->rxmode.offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
3285 	}
3286 #endif
3287 
3288 	return 0;
3289 }
3290 
3291 static int
3292 igbvf_dev_start(struct rte_eth_dev *dev)
3293 {
3294 	struct e1000_hw *hw =
3295 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3296 	struct e1000_adapter *adapter =
3297 		E1000_DEV_PRIVATE(dev->data->dev_private);
3298 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3299 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3300 	int ret;
3301 	uint32_t intr_vector = 0;
3302 
3303 	PMD_INIT_FUNC_TRACE();
3304 
3305 	hw->mac.ops.reset_hw(hw);
3306 	adapter->stopped = 0;
3307 
3308 	/* Set all vfta */
3309 	igbvf_set_vfta_all(dev,1);
3310 
3311 	eth_igbvf_tx_init(dev);
3312 
3313 	/* This can fail when allocating mbufs for descriptor rings */
3314 	ret = eth_igbvf_rx_init(dev);
3315 	if (ret) {
3316 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
3317 		igb_dev_clear_queues(dev);
3318 		return ret;
3319 	}
3320 
3321 	/* check and configure queue intr-vector mapping */
3322 	if (rte_intr_cap_multiple(intr_handle) &&
3323 	    dev->data->dev_conf.intr_conf.rxq) {
3324 		intr_vector = dev->data->nb_rx_queues;
3325 		ret = rte_intr_efd_enable(intr_handle, intr_vector);
3326 		if (ret)
3327 			return ret;
3328 	}
3329 
3330 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
3331 		intr_handle->intr_vec =
3332 			rte_zmalloc("intr_vec",
3333 				    dev->data->nb_rx_queues * sizeof(int), 0);
3334 		if (!intr_handle->intr_vec) {
3335 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
3336 				     " intr_vec", dev->data->nb_rx_queues);
3337 			return -ENOMEM;
3338 		}
3339 	}
3340 
3341 	eth_igbvf_configure_msix_intr(dev);
3342 
3343 	/* enable uio/vfio intr/eventfd mapping */
3344 	rte_intr_enable(intr_handle);
3345 
3346 	/* resume enabled intr since hw reset */
3347 	igbvf_intr_enable(dev);
3348 
3349 	return 0;
3350 }
3351 
3352 static int
3353 igbvf_dev_stop(struct rte_eth_dev *dev)
3354 {
3355 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3356 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3357 	struct e1000_adapter *adapter =
3358 		E1000_DEV_PRIVATE(dev->data->dev_private);
3359 
3360 	if (adapter->stopped)
3361 		return 0;
3362 
3363 	PMD_INIT_FUNC_TRACE();
3364 
3365 	igbvf_stop_adapter(dev);
3366 
3367 	/*
3368 	  * Clear what we set, but we still keep shadow_vfta to
3369 	  * restore after device starts
3370 	  */
3371 	igbvf_set_vfta_all(dev,0);
3372 
3373 	igb_dev_clear_queues(dev);
3374 
3375 	/* disable intr eventfd mapping */
3376 	rte_intr_disable(intr_handle);
3377 
3378 	/* Clean datapath event and queue/vec mapping */
3379 	rte_intr_efd_disable(intr_handle);
3380 	if (intr_handle->intr_vec) {
3381 		rte_free(intr_handle->intr_vec);
3382 		intr_handle->intr_vec = NULL;
3383 	}
3384 
3385 	adapter->stopped = true;
3386 	dev->data->dev_started = 0;
3387 
3388 	return 0;
3389 }
3390 
3391 static int
3392 igbvf_dev_close(struct rte_eth_dev *dev)
3393 {
3394 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3395 	struct rte_ether_addr addr;
3396 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3397 	int ret;
3398 
3399 	PMD_INIT_FUNC_TRACE();
3400 
3401 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
3402 		return 0;
3403 
3404 	e1000_reset_hw(hw);
3405 
3406 	ret = igbvf_dev_stop(dev);
3407 	if (ret != 0)
3408 		return ret;
3409 
3410 	igb_dev_free_queues(dev);
3411 
3412 	/**
3413 	 * reprogram the RAR with a zero mac address,
3414 	 * to ensure that the VF traffic goes to the PF
3415 	 * after stop, close and detach of the VF.
3416 	 **/
3417 
3418 	memset(&addr, 0, sizeof(addr));
3419 	igbvf_default_mac_addr_set(dev, &addr);
3420 
3421 	rte_intr_callback_unregister(&pci_dev->intr_handle,
3422 				     eth_igbvf_interrupt_handler,
3423 				     (void *)dev);
3424 
3425 	return 0;
3426 }
3427 
3428 static int
3429 igbvf_promiscuous_enable(struct rte_eth_dev *dev)
3430 {
3431 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3432 
3433 	/* Set both unicast and multicast promisc */
3434 	e1000_promisc_set_vf(hw, e1000_promisc_enabled);
3435 
3436 	return 0;
3437 }
3438 
3439 static int
3440 igbvf_promiscuous_disable(struct rte_eth_dev *dev)
3441 {
3442 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3443 
3444 	/* If in allmulticast mode leave multicast promisc */
3445 	if (dev->data->all_multicast == 1)
3446 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3447 	else
3448 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3449 
3450 	return 0;
3451 }
3452 
3453 static int
3454 igbvf_allmulticast_enable(struct rte_eth_dev *dev)
3455 {
3456 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3457 
3458 	/* In promiscuous mode multicast promisc already set */
3459 	if (dev->data->promiscuous == 0)
3460 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3461 
3462 	return 0;
3463 }
3464 
3465 static int
3466 igbvf_allmulticast_disable(struct rte_eth_dev *dev)
3467 {
3468 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3469 
3470 	/* In promiscuous mode leave multicast promisc enabled */
3471 	if (dev->data->promiscuous == 0)
3472 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3473 
3474 	return 0;
3475 }
3476 
3477 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on)
3478 {
3479 	struct e1000_mbx_info *mbx = &hw->mbx;
3480 	uint32_t msgbuf[2];
3481 	s32 err;
3482 
3483 	/* After set vlan, vlan strip will also be enabled in igb driver*/
3484 	msgbuf[0] = E1000_VF_SET_VLAN;
3485 	msgbuf[1] = vid;
3486 	/* Setting the 8 bit field MSG INFO to TRUE indicates "add" */
3487 	if (on)
3488 		msgbuf[0] |= E1000_VF_SET_VLAN_ADD;
3489 
3490 	err = mbx->ops.write_posted(hw, msgbuf, 2, 0);
3491 	if (err)
3492 		goto mbx_err;
3493 
3494 	err = mbx->ops.read_posted(hw, msgbuf, 2, 0);
3495 	if (err)
3496 		goto mbx_err;
3497 
3498 	msgbuf[0] &= ~E1000_VT_MSGTYPE_CTS;
3499 	if (msgbuf[0] == (E1000_VF_SET_VLAN | E1000_VT_MSGTYPE_NACK))
3500 		err = -EINVAL;
3501 
3502 mbx_err:
3503 	return err;
3504 }
3505 
3506 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on)
3507 {
3508 	struct e1000_hw *hw =
3509 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3510 	struct e1000_vfta * shadow_vfta =
3511 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3512 	int i = 0, j = 0, vfta = 0, mask = 1;
3513 
3514 	for (i = 0; i < IGB_VFTA_SIZE; i++){
3515 		vfta = shadow_vfta->vfta[i];
3516 		if(vfta){
3517 			mask = 1;
3518 			for (j = 0; j < 32; j++){
3519 				if(vfta & mask)
3520 					igbvf_set_vfta(hw,
3521 						(uint16_t)((i<<5)+j), on);
3522 				mask<<=1;
3523 			}
3524 		}
3525 	}
3526 
3527 }
3528 
3529 static int
3530 igbvf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
3531 {
3532 	struct e1000_hw *hw =
3533 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3534 	struct e1000_vfta * shadow_vfta =
3535 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3536 	uint32_t vid_idx = 0;
3537 	uint32_t vid_bit = 0;
3538 	int ret = 0;
3539 
3540 	PMD_INIT_FUNC_TRACE();
3541 
3542 	/*vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf*/
3543 	ret = igbvf_set_vfta(hw, vlan_id, !!on);
3544 	if(ret){
3545 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
3546 		return ret;
3547 	}
3548 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
3549 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
3550 
3551 	/*Save what we set and retore it after device reset*/
3552 	if (on)
3553 		shadow_vfta->vfta[vid_idx] |= vid_bit;
3554 	else
3555 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
3556 
3557 	return 0;
3558 }
3559 
3560 static int
3561 igbvf_default_mac_addr_set(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
3562 {
3563 	struct e1000_hw *hw =
3564 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3565 
3566 	/* index is not used by rar_set() */
3567 	hw->mac.ops.rar_set(hw, (void *)addr, 0);
3568 	return 0;
3569 }
3570 
3571 
3572 static int
3573 eth_igb_rss_reta_update(struct rte_eth_dev *dev,
3574 			struct rte_eth_rss_reta_entry64 *reta_conf,
3575 			uint16_t reta_size)
3576 {
3577 	uint8_t i, j, mask;
3578 	uint32_t reta, r;
3579 	uint16_t idx, shift;
3580 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3581 
3582 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3583 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3584 			"(%d) doesn't match the number hardware can supported "
3585 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3586 		return -EINVAL;
3587 	}
3588 
3589 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3590 		idx = i / RTE_RETA_GROUP_SIZE;
3591 		shift = i % RTE_RETA_GROUP_SIZE;
3592 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3593 						IGB_4_BIT_MASK);
3594 		if (!mask)
3595 			continue;
3596 		if (mask == IGB_4_BIT_MASK)
3597 			r = 0;
3598 		else
3599 			r = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3600 		for (j = 0, reta = 0; j < IGB_4_BIT_WIDTH; j++) {
3601 			if (mask & (0x1 << j))
3602 				reta |= reta_conf[idx].reta[shift + j] <<
3603 							(CHAR_BIT * j);
3604 			else
3605 				reta |= r & (IGB_8_BIT_MASK << (CHAR_BIT * j));
3606 		}
3607 		E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
3608 	}
3609 
3610 	return 0;
3611 }
3612 
3613 static int
3614 eth_igb_rss_reta_query(struct rte_eth_dev *dev,
3615 		       struct rte_eth_rss_reta_entry64 *reta_conf,
3616 		       uint16_t reta_size)
3617 {
3618 	uint8_t i, j, mask;
3619 	uint32_t reta;
3620 	uint16_t idx, shift;
3621 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3622 
3623 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3624 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3625 			"(%d) doesn't match the number hardware can supported "
3626 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3627 		return -EINVAL;
3628 	}
3629 
3630 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3631 		idx = i / RTE_RETA_GROUP_SIZE;
3632 		shift = i % RTE_RETA_GROUP_SIZE;
3633 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3634 						IGB_4_BIT_MASK);
3635 		if (!mask)
3636 			continue;
3637 		reta = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3638 		for (j = 0; j < IGB_4_BIT_WIDTH; j++) {
3639 			if (mask & (0x1 << j))
3640 				reta_conf[idx].reta[shift + j] =
3641 					((reta >> (CHAR_BIT * j)) &
3642 						IGB_8_BIT_MASK);
3643 		}
3644 	}
3645 
3646 	return 0;
3647 }
3648 
3649 int
3650 eth_igb_syn_filter_set(struct rte_eth_dev *dev,
3651 			struct rte_eth_syn_filter *filter,
3652 			bool add)
3653 {
3654 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3655 	struct e1000_filter_info *filter_info =
3656 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3657 	uint32_t synqf, rfctl;
3658 
3659 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3660 		return -EINVAL;
3661 
3662 	synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3663 
3664 	if (add) {
3665 		if (synqf & E1000_SYN_FILTER_ENABLE)
3666 			return -EINVAL;
3667 
3668 		synqf = (uint32_t)(((filter->queue << E1000_SYN_FILTER_QUEUE_SHIFT) &
3669 			E1000_SYN_FILTER_QUEUE) | E1000_SYN_FILTER_ENABLE);
3670 
3671 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3672 		if (filter->hig_pri)
3673 			rfctl |= E1000_RFCTL_SYNQFP;
3674 		else
3675 			rfctl &= ~E1000_RFCTL_SYNQFP;
3676 
3677 		E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3678 	} else {
3679 		if (!(synqf & E1000_SYN_FILTER_ENABLE))
3680 			return -ENOENT;
3681 		synqf = 0;
3682 	}
3683 
3684 	filter_info->syn_info = synqf;
3685 	E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
3686 	E1000_WRITE_FLUSH(hw);
3687 	return 0;
3688 }
3689 
3690 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_2tuple_filter_info*/
3691 static inline int
3692 ntuple_filter_to_2tuple(struct rte_eth_ntuple_filter *filter,
3693 			struct e1000_2tuple_filter_info *filter_info)
3694 {
3695 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3696 		return -EINVAL;
3697 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
3698 		return -EINVAL;  /* filter index is out of range. */
3699 	if (filter->tcp_flags > RTE_NTUPLE_TCP_FLAGS_MASK)
3700 		return -EINVAL;  /* flags is invalid. */
3701 
3702 	switch (filter->dst_port_mask) {
3703 	case UINT16_MAX:
3704 		filter_info->dst_port_mask = 0;
3705 		filter_info->dst_port = filter->dst_port;
3706 		break;
3707 	case 0:
3708 		filter_info->dst_port_mask = 1;
3709 		break;
3710 	default:
3711 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
3712 		return -EINVAL;
3713 	}
3714 
3715 	switch (filter->proto_mask) {
3716 	case UINT8_MAX:
3717 		filter_info->proto_mask = 0;
3718 		filter_info->proto = filter->proto;
3719 		break;
3720 	case 0:
3721 		filter_info->proto_mask = 1;
3722 		break;
3723 	default:
3724 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
3725 		return -EINVAL;
3726 	}
3727 
3728 	filter_info->priority = (uint8_t)filter->priority;
3729 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
3730 		filter_info->tcp_flags = filter->tcp_flags;
3731 	else
3732 		filter_info->tcp_flags = 0;
3733 
3734 	return 0;
3735 }
3736 
3737 static inline struct e1000_2tuple_filter *
3738 igb_2tuple_filter_lookup(struct e1000_2tuple_filter_list *filter_list,
3739 			struct e1000_2tuple_filter_info *key)
3740 {
3741 	struct e1000_2tuple_filter *it;
3742 
3743 	TAILQ_FOREACH(it, filter_list, entries) {
3744 		if (memcmp(key, &it->filter_info,
3745 			sizeof(struct e1000_2tuple_filter_info)) == 0) {
3746 			return it;
3747 		}
3748 	}
3749 	return NULL;
3750 }
3751 
3752 /* inject a igb 2tuple filter to HW */
3753 static inline void
3754 igb_inject_2uple_filter(struct rte_eth_dev *dev,
3755 			   struct e1000_2tuple_filter *filter)
3756 {
3757 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3758 	uint32_t ttqf = E1000_TTQF_DISABLE_MASK;
3759 	uint32_t imir, imir_ext = E1000_IMIREXT_SIZE_BP;
3760 	int i;
3761 
3762 	i = filter->index;
3763 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
3764 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
3765 		imir |= E1000_IMIR_PORT_BP;
3766 	else
3767 		imir &= ~E1000_IMIR_PORT_BP;
3768 
3769 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
3770 
3771 	ttqf |= E1000_TTQF_QUEUE_ENABLE;
3772 	ttqf |= (uint32_t)(filter->queue << E1000_TTQF_QUEUE_SHIFT);
3773 	ttqf |= (uint32_t)(filter->filter_info.proto &
3774 						E1000_TTQF_PROTOCOL_MASK);
3775 	if (filter->filter_info.proto_mask == 0)
3776 		ttqf &= ~E1000_TTQF_MASK_ENABLE;
3777 
3778 	/* tcp flags bits setting. */
3779 	if (filter->filter_info.tcp_flags & RTE_NTUPLE_TCP_FLAGS_MASK) {
3780 		if (filter->filter_info.tcp_flags & RTE_TCP_URG_FLAG)
3781 			imir_ext |= E1000_IMIREXT_CTRL_URG;
3782 		if (filter->filter_info.tcp_flags & RTE_TCP_ACK_FLAG)
3783 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
3784 		if (filter->filter_info.tcp_flags & RTE_TCP_PSH_FLAG)
3785 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
3786 		if (filter->filter_info.tcp_flags & RTE_TCP_RST_FLAG)
3787 			imir_ext |= E1000_IMIREXT_CTRL_RST;
3788 		if (filter->filter_info.tcp_flags & RTE_TCP_SYN_FLAG)
3789 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
3790 		if (filter->filter_info.tcp_flags & RTE_TCP_FIN_FLAG)
3791 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
3792 	} else {
3793 		imir_ext |= E1000_IMIREXT_CTRL_BP;
3794 	}
3795 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
3796 	E1000_WRITE_REG(hw, E1000_TTQF(i), ttqf);
3797 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
3798 }
3799 
3800 /*
3801  * igb_add_2tuple_filter - add a 2tuple filter
3802  *
3803  * @param
3804  * dev: Pointer to struct rte_eth_dev.
3805  * ntuple_filter: ponter to the filter that will be added.
3806  *
3807  * @return
3808  *    - On success, zero.
3809  *    - On failure, a negative value.
3810  */
3811 static int
3812 igb_add_2tuple_filter(struct rte_eth_dev *dev,
3813 			struct rte_eth_ntuple_filter *ntuple_filter)
3814 {
3815 	struct e1000_filter_info *filter_info =
3816 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3817 	struct e1000_2tuple_filter *filter;
3818 	int i, ret;
3819 
3820 	filter = rte_zmalloc("e1000_2tuple_filter",
3821 			sizeof(struct e1000_2tuple_filter), 0);
3822 	if (filter == NULL)
3823 		return -ENOMEM;
3824 
3825 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3826 				      &filter->filter_info);
3827 	if (ret < 0) {
3828 		rte_free(filter);
3829 		return ret;
3830 	}
3831 	if (igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3832 					 &filter->filter_info) != NULL) {
3833 		PMD_DRV_LOG(ERR, "filter exists.");
3834 		rte_free(filter);
3835 		return -EEXIST;
3836 	}
3837 	filter->queue = ntuple_filter->queue;
3838 
3839 	/*
3840 	 * look for an unused 2tuple filter index,
3841 	 * and insert the filter to list.
3842 	 */
3843 	for (i = 0; i < E1000_MAX_TTQF_FILTERS; i++) {
3844 		if (!(filter_info->twotuple_mask & (1 << i))) {
3845 			filter_info->twotuple_mask |= 1 << i;
3846 			filter->index = i;
3847 			TAILQ_INSERT_TAIL(&filter_info->twotuple_list,
3848 					  filter,
3849 					  entries);
3850 			break;
3851 		}
3852 	}
3853 	if (i >= E1000_MAX_TTQF_FILTERS) {
3854 		PMD_DRV_LOG(ERR, "2tuple filters are full.");
3855 		rte_free(filter);
3856 		return -ENOSYS;
3857 	}
3858 
3859 	igb_inject_2uple_filter(dev, filter);
3860 	return 0;
3861 }
3862 
3863 int
3864 igb_delete_2tuple_filter(struct rte_eth_dev *dev,
3865 			struct e1000_2tuple_filter *filter)
3866 {
3867 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3868 	struct e1000_filter_info *filter_info =
3869 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3870 
3871 	filter_info->twotuple_mask &= ~(1 << filter->index);
3872 	TAILQ_REMOVE(&filter_info->twotuple_list, filter, entries);
3873 	rte_free(filter);
3874 
3875 	E1000_WRITE_REG(hw, E1000_TTQF(filter->index), E1000_TTQF_DISABLE_MASK);
3876 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
3877 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
3878 	return 0;
3879 }
3880 
3881 /*
3882  * igb_remove_2tuple_filter - remove a 2tuple filter
3883  *
3884  * @param
3885  * dev: Pointer to struct rte_eth_dev.
3886  * ntuple_filter: ponter to the filter that will be removed.
3887  *
3888  * @return
3889  *    - On success, zero.
3890  *    - On failure, a negative value.
3891  */
3892 static int
3893 igb_remove_2tuple_filter(struct rte_eth_dev *dev,
3894 			struct rte_eth_ntuple_filter *ntuple_filter)
3895 {
3896 	struct e1000_filter_info *filter_info =
3897 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3898 	struct e1000_2tuple_filter_info filter_2tuple;
3899 	struct e1000_2tuple_filter *filter;
3900 	int ret;
3901 
3902 	memset(&filter_2tuple, 0, sizeof(struct e1000_2tuple_filter_info));
3903 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3904 				      &filter_2tuple);
3905 	if (ret < 0)
3906 		return ret;
3907 
3908 	filter = igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3909 					 &filter_2tuple);
3910 	if (filter == NULL) {
3911 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
3912 		return -ENOENT;
3913 	}
3914 
3915 	igb_delete_2tuple_filter(dev, filter);
3916 
3917 	return 0;
3918 }
3919 
3920 /* inject a igb flex filter to HW */
3921 static inline void
3922 igb_inject_flex_filter(struct rte_eth_dev *dev,
3923 			   struct e1000_flex_filter *filter)
3924 {
3925 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3926 	uint32_t wufc, queueing;
3927 	uint32_t reg_off;
3928 	uint8_t i, j = 0;
3929 
3930 	wufc = E1000_READ_REG(hw, E1000_WUFC);
3931 	if (filter->index < E1000_MAX_FHFT)
3932 		reg_off = E1000_FHFT(filter->index);
3933 	else
3934 		reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
3935 
3936 	E1000_WRITE_REG(hw, E1000_WUFC, wufc | E1000_WUFC_FLEX_HQ |
3937 			(E1000_WUFC_FLX0 << filter->index));
3938 	queueing = filter->filter_info.len |
3939 		(filter->queue << E1000_FHFT_QUEUEING_QUEUE_SHIFT) |
3940 		(filter->filter_info.priority <<
3941 			E1000_FHFT_QUEUEING_PRIO_SHIFT);
3942 	E1000_WRITE_REG(hw, reg_off + E1000_FHFT_QUEUEING_OFFSET,
3943 			queueing);
3944 
3945 	for (i = 0; i < E1000_FLEX_FILTERS_MASK_SIZE; i++) {
3946 		E1000_WRITE_REG(hw, reg_off,
3947 				filter->filter_info.dwords[j]);
3948 		reg_off += sizeof(uint32_t);
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 			(uint32_t)filter->filter_info.mask[i]);
3954 		reg_off += sizeof(uint32_t) * 2;
3955 		++j;
3956 	}
3957 }
3958 
3959 static inline struct e1000_flex_filter *
3960 eth_igb_flex_filter_lookup(struct e1000_flex_filter_list *filter_list,
3961 			struct e1000_flex_filter_info *key)
3962 {
3963 	struct e1000_flex_filter *it;
3964 
3965 	TAILQ_FOREACH(it, filter_list, entries) {
3966 		if (memcmp(key, &it->filter_info,
3967 			sizeof(struct e1000_flex_filter_info)) == 0)
3968 			return it;
3969 	}
3970 
3971 	return NULL;
3972 }
3973 
3974 /* remove a flex byte filter
3975  * @param
3976  * dev: Pointer to struct rte_eth_dev.
3977  * filter: the pointer of the filter will be removed.
3978  */
3979 void
3980 igb_remove_flex_filter(struct rte_eth_dev *dev,
3981 			struct e1000_flex_filter *filter)
3982 {
3983 	struct e1000_filter_info *filter_info =
3984 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3985 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3986 	uint32_t wufc, i;
3987 	uint32_t reg_off;
3988 
3989 	wufc = E1000_READ_REG(hw, E1000_WUFC);
3990 	if (filter->index < E1000_MAX_FHFT)
3991 		reg_off = E1000_FHFT(filter->index);
3992 	else
3993 		reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
3994 
3995 	for (i = 0; i < E1000_FHFT_SIZE_IN_DWD; i++)
3996 		E1000_WRITE_REG(hw, reg_off + i * sizeof(uint32_t), 0);
3997 
3998 	E1000_WRITE_REG(hw, E1000_WUFC, wufc &
3999 		(~(E1000_WUFC_FLX0 << filter->index)));
4000 
4001 	filter_info->flex_mask &= ~(1 << filter->index);
4002 	TAILQ_REMOVE(&filter_info->flex_list, filter, entries);
4003 	rte_free(filter);
4004 }
4005 
4006 int
4007 eth_igb_add_del_flex_filter(struct rte_eth_dev *dev,
4008 			struct igb_flex_filter *filter,
4009 			bool add)
4010 {
4011 	struct e1000_filter_info *filter_info =
4012 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4013 	struct e1000_flex_filter *flex_filter, *it;
4014 	uint32_t mask;
4015 	uint8_t shift, i;
4016 
4017 	flex_filter = rte_zmalloc("e1000_flex_filter",
4018 			sizeof(struct e1000_flex_filter), 0);
4019 	if (flex_filter == NULL)
4020 		return -ENOMEM;
4021 
4022 	flex_filter->filter_info.len = filter->len;
4023 	flex_filter->filter_info.priority = filter->priority;
4024 	memcpy(flex_filter->filter_info.dwords, filter->bytes, filter->len);
4025 	for (i = 0; i < RTE_ALIGN(filter->len, CHAR_BIT) / CHAR_BIT; i++) {
4026 		mask = 0;
4027 		/* reverse bits in flex filter's mask*/
4028 		for (shift = 0; shift < CHAR_BIT; shift++) {
4029 			if (filter->mask[i] & (0x01 << shift))
4030 				mask |= (0x80 >> shift);
4031 		}
4032 		flex_filter->filter_info.mask[i] = mask;
4033 	}
4034 
4035 	it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
4036 				&flex_filter->filter_info);
4037 	if (it == NULL && !add) {
4038 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4039 		rte_free(flex_filter);
4040 		return -ENOENT;
4041 	}
4042 	if (it != NULL && add) {
4043 		PMD_DRV_LOG(ERR, "filter exists.");
4044 		rte_free(flex_filter);
4045 		return -EEXIST;
4046 	}
4047 
4048 	if (add) {
4049 		flex_filter->queue = filter->queue;
4050 		/*
4051 		 * look for an unused flex filter index
4052 		 * and insert the filter into the list.
4053 		 */
4054 		for (i = 0; i < E1000_MAX_FLEX_FILTERS; i++) {
4055 			if (!(filter_info->flex_mask & (1 << i))) {
4056 				filter_info->flex_mask |= 1 << i;
4057 				flex_filter->index = i;
4058 				TAILQ_INSERT_TAIL(&filter_info->flex_list,
4059 					flex_filter,
4060 					entries);
4061 				break;
4062 			}
4063 		}
4064 		if (i >= E1000_MAX_FLEX_FILTERS) {
4065 			PMD_DRV_LOG(ERR, "flex filters are full.");
4066 			rte_free(flex_filter);
4067 			return -ENOSYS;
4068 		}
4069 
4070 		igb_inject_flex_filter(dev, flex_filter);
4071 
4072 	} else {
4073 		igb_remove_flex_filter(dev, it);
4074 		rte_free(flex_filter);
4075 	}
4076 
4077 	return 0;
4078 }
4079 
4080 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_5tuple_filter_info*/
4081 static inline int
4082 ntuple_filter_to_5tuple_82576(struct rte_eth_ntuple_filter *filter,
4083 			struct e1000_5tuple_filter_info *filter_info)
4084 {
4085 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM_82576)
4086 		return -EINVAL;
4087 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
4088 		return -EINVAL;  /* filter index is out of range. */
4089 	if (filter->tcp_flags > RTE_NTUPLE_TCP_FLAGS_MASK)
4090 		return -EINVAL;  /* flags is invalid. */
4091 
4092 	switch (filter->dst_ip_mask) {
4093 	case UINT32_MAX:
4094 		filter_info->dst_ip_mask = 0;
4095 		filter_info->dst_ip = filter->dst_ip;
4096 		break;
4097 	case 0:
4098 		filter_info->dst_ip_mask = 1;
4099 		break;
4100 	default:
4101 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
4102 		return -EINVAL;
4103 	}
4104 
4105 	switch (filter->src_ip_mask) {
4106 	case UINT32_MAX:
4107 		filter_info->src_ip_mask = 0;
4108 		filter_info->src_ip = filter->src_ip;
4109 		break;
4110 	case 0:
4111 		filter_info->src_ip_mask = 1;
4112 		break;
4113 	default:
4114 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
4115 		return -EINVAL;
4116 	}
4117 
4118 	switch (filter->dst_port_mask) {
4119 	case UINT16_MAX:
4120 		filter_info->dst_port_mask = 0;
4121 		filter_info->dst_port = filter->dst_port;
4122 		break;
4123 	case 0:
4124 		filter_info->dst_port_mask = 1;
4125 		break;
4126 	default:
4127 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
4128 		return -EINVAL;
4129 	}
4130 
4131 	switch (filter->src_port_mask) {
4132 	case UINT16_MAX:
4133 		filter_info->src_port_mask = 0;
4134 		filter_info->src_port = filter->src_port;
4135 		break;
4136 	case 0:
4137 		filter_info->src_port_mask = 1;
4138 		break;
4139 	default:
4140 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
4141 		return -EINVAL;
4142 	}
4143 
4144 	switch (filter->proto_mask) {
4145 	case UINT8_MAX:
4146 		filter_info->proto_mask = 0;
4147 		filter_info->proto = filter->proto;
4148 		break;
4149 	case 0:
4150 		filter_info->proto_mask = 1;
4151 		break;
4152 	default:
4153 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
4154 		return -EINVAL;
4155 	}
4156 
4157 	filter_info->priority = (uint8_t)filter->priority;
4158 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
4159 		filter_info->tcp_flags = filter->tcp_flags;
4160 	else
4161 		filter_info->tcp_flags = 0;
4162 
4163 	return 0;
4164 }
4165 
4166 static inline struct e1000_5tuple_filter *
4167 igb_5tuple_filter_lookup_82576(struct e1000_5tuple_filter_list *filter_list,
4168 			struct e1000_5tuple_filter_info *key)
4169 {
4170 	struct e1000_5tuple_filter *it;
4171 
4172 	TAILQ_FOREACH(it, filter_list, entries) {
4173 		if (memcmp(key, &it->filter_info,
4174 			sizeof(struct e1000_5tuple_filter_info)) == 0) {
4175 			return it;
4176 		}
4177 	}
4178 	return NULL;
4179 }
4180 
4181 /* inject a igb 5-tuple filter to HW */
4182 static inline void
4183 igb_inject_5tuple_filter_82576(struct rte_eth_dev *dev,
4184 			   struct e1000_5tuple_filter *filter)
4185 {
4186 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4187 	uint32_t ftqf = E1000_FTQF_VF_BP | E1000_FTQF_MASK;
4188 	uint32_t spqf, imir, imir_ext = E1000_IMIREXT_SIZE_BP;
4189 	uint8_t i;
4190 
4191 	i = filter->index;
4192 	ftqf |= filter->filter_info.proto & E1000_FTQF_PROTOCOL_MASK;
4193 	if (filter->filter_info.src_ip_mask == 0) /* 0b means compare. */
4194 		ftqf &= ~E1000_FTQF_MASK_SOURCE_ADDR_BP;
4195 	if (filter->filter_info.dst_ip_mask == 0)
4196 		ftqf &= ~E1000_FTQF_MASK_DEST_ADDR_BP;
4197 	if (filter->filter_info.src_port_mask == 0)
4198 		ftqf &= ~E1000_FTQF_MASK_SOURCE_PORT_BP;
4199 	if (filter->filter_info.proto_mask == 0)
4200 		ftqf &= ~E1000_FTQF_MASK_PROTO_BP;
4201 	ftqf |= (filter->queue << E1000_FTQF_QUEUE_SHIFT) &
4202 		E1000_FTQF_QUEUE_MASK;
4203 	ftqf |= E1000_FTQF_QUEUE_ENABLE;
4204 	E1000_WRITE_REG(hw, E1000_FTQF(i), ftqf);
4205 	E1000_WRITE_REG(hw, E1000_DAQF(i), filter->filter_info.dst_ip);
4206 	E1000_WRITE_REG(hw, E1000_SAQF(i), filter->filter_info.src_ip);
4207 
4208 	spqf = filter->filter_info.src_port & E1000_SPQF_SRCPORT;
4209 	E1000_WRITE_REG(hw, E1000_SPQF(i), spqf);
4210 
4211 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
4212 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
4213 		imir |= E1000_IMIR_PORT_BP;
4214 	else
4215 		imir &= ~E1000_IMIR_PORT_BP;
4216 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
4217 
4218 	/* tcp flags bits setting. */
4219 	if (filter->filter_info.tcp_flags & RTE_NTUPLE_TCP_FLAGS_MASK) {
4220 		if (filter->filter_info.tcp_flags & RTE_TCP_URG_FLAG)
4221 			imir_ext |= E1000_IMIREXT_CTRL_URG;
4222 		if (filter->filter_info.tcp_flags & RTE_TCP_ACK_FLAG)
4223 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
4224 		if (filter->filter_info.tcp_flags & RTE_TCP_PSH_FLAG)
4225 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
4226 		if (filter->filter_info.tcp_flags & RTE_TCP_RST_FLAG)
4227 			imir_ext |= E1000_IMIREXT_CTRL_RST;
4228 		if (filter->filter_info.tcp_flags & RTE_TCP_SYN_FLAG)
4229 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
4230 		if (filter->filter_info.tcp_flags & RTE_TCP_FIN_FLAG)
4231 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
4232 	} else {
4233 		imir_ext |= E1000_IMIREXT_CTRL_BP;
4234 	}
4235 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
4236 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
4237 }
4238 
4239 /*
4240  * igb_add_5tuple_filter_82576 - add a 5tuple filter
4241  *
4242  * @param
4243  * dev: Pointer to struct rte_eth_dev.
4244  * ntuple_filter: ponter to the filter that will be added.
4245  *
4246  * @return
4247  *    - On success, zero.
4248  *    - On failure, a negative value.
4249  */
4250 static int
4251 igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
4252 			struct rte_eth_ntuple_filter *ntuple_filter)
4253 {
4254 	struct e1000_filter_info *filter_info =
4255 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4256 	struct e1000_5tuple_filter *filter;
4257 	uint8_t i;
4258 	int ret;
4259 
4260 	filter = rte_zmalloc("e1000_5tuple_filter",
4261 			sizeof(struct e1000_5tuple_filter), 0);
4262 	if (filter == NULL)
4263 		return -ENOMEM;
4264 
4265 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4266 					    &filter->filter_info);
4267 	if (ret < 0) {
4268 		rte_free(filter);
4269 		return ret;
4270 	}
4271 
4272 	if (igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4273 					 &filter->filter_info) != NULL) {
4274 		PMD_DRV_LOG(ERR, "filter exists.");
4275 		rte_free(filter);
4276 		return -EEXIST;
4277 	}
4278 	filter->queue = ntuple_filter->queue;
4279 
4280 	/*
4281 	 * look for an unused 5tuple filter index,
4282 	 * and insert the filter to list.
4283 	 */
4284 	for (i = 0; i < E1000_MAX_FTQF_FILTERS; i++) {
4285 		if (!(filter_info->fivetuple_mask & (1 << i))) {
4286 			filter_info->fivetuple_mask |= 1 << i;
4287 			filter->index = i;
4288 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
4289 					  filter,
4290 					  entries);
4291 			break;
4292 		}
4293 	}
4294 	if (i >= E1000_MAX_FTQF_FILTERS) {
4295 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
4296 		rte_free(filter);
4297 		return -ENOSYS;
4298 	}
4299 
4300 	igb_inject_5tuple_filter_82576(dev, filter);
4301 	return 0;
4302 }
4303 
4304 int
4305 igb_delete_5tuple_filter_82576(struct rte_eth_dev *dev,
4306 				struct e1000_5tuple_filter *filter)
4307 {
4308 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4309 	struct e1000_filter_info *filter_info =
4310 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4311 
4312 	filter_info->fivetuple_mask &= ~(1 << filter->index);
4313 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
4314 	rte_free(filter);
4315 
4316 	E1000_WRITE_REG(hw, E1000_FTQF(filter->index),
4317 			E1000_FTQF_VF_BP | E1000_FTQF_MASK);
4318 	E1000_WRITE_REG(hw, E1000_DAQF(filter->index), 0);
4319 	E1000_WRITE_REG(hw, E1000_SAQF(filter->index), 0);
4320 	E1000_WRITE_REG(hw, E1000_SPQF(filter->index), 0);
4321 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
4322 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
4323 	return 0;
4324 }
4325 
4326 /*
4327  * igb_remove_5tuple_filter_82576 - remove a 5tuple filter
4328  *
4329  * @param
4330  * dev: Pointer to struct rte_eth_dev.
4331  * ntuple_filter: ponter to the filter that will be removed.
4332  *
4333  * @return
4334  *    - On success, zero.
4335  *    - On failure, a negative value.
4336  */
4337 static int
4338 igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
4339 				struct rte_eth_ntuple_filter *ntuple_filter)
4340 {
4341 	struct e1000_filter_info *filter_info =
4342 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4343 	struct e1000_5tuple_filter_info filter_5tuple;
4344 	struct e1000_5tuple_filter *filter;
4345 	int ret;
4346 
4347 	memset(&filter_5tuple, 0, sizeof(struct e1000_5tuple_filter_info));
4348 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4349 					    &filter_5tuple);
4350 	if (ret < 0)
4351 		return ret;
4352 
4353 	filter = igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4354 					 &filter_5tuple);
4355 	if (filter == NULL) {
4356 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4357 		return -ENOENT;
4358 	}
4359 
4360 	igb_delete_5tuple_filter_82576(dev, filter);
4361 
4362 	return 0;
4363 }
4364 
4365 static int
4366 eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4367 {
4368 	uint32_t rctl;
4369 	struct e1000_hw *hw;
4370 	struct rte_eth_dev_info dev_info;
4371 	uint32_t frame_size = mtu + E1000_ETH_OVERHEAD;
4372 	int ret;
4373 
4374 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4375 
4376 #ifdef RTE_LIBRTE_82571_SUPPORT
4377 	/* XXX: not bigger than max_rx_pktlen */
4378 	if (hw->mac.type == e1000_82571)
4379 		return -ENOTSUP;
4380 #endif
4381 	ret = eth_igb_infos_get(dev, &dev_info);
4382 	if (ret != 0)
4383 		return ret;
4384 
4385 	/* check that mtu is within the allowed range */
4386 	if (mtu < RTE_ETHER_MIN_MTU ||
4387 			frame_size > dev_info.max_rx_pktlen)
4388 		return -EINVAL;
4389 
4390 	/*
4391 	 * If device is started, refuse mtu that requires the support of
4392 	 * scattered packets when this feature has not been enabled before.
4393 	 */
4394 	if (dev->data->dev_started && !dev->data->scattered_rx &&
4395 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
4396 		PMD_INIT_LOG(ERR, "Stop port first.");
4397 		return -EINVAL;
4398 	}
4399 
4400 	rctl = E1000_READ_REG(hw, E1000_RCTL);
4401 
4402 	/* switch to jumbo mode if needed */
4403 	if (frame_size > E1000_ETH_MAX_LEN) {
4404 		dev->data->dev_conf.rxmode.offloads |=
4405 			DEV_RX_OFFLOAD_JUMBO_FRAME;
4406 		rctl |= E1000_RCTL_LPE;
4407 	} else {
4408 		dev->data->dev_conf.rxmode.offloads &=
4409 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
4410 		rctl &= ~E1000_RCTL_LPE;
4411 	}
4412 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
4413 
4414 	/* update max frame size */
4415 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4416 
4417 	E1000_WRITE_REG(hw, E1000_RLPML,
4418 			dev->data->dev_conf.rxmode.max_rx_pkt_len);
4419 
4420 	return 0;
4421 }
4422 
4423 /*
4424  * igb_add_del_ntuple_filter - add or delete a ntuple filter
4425  *
4426  * @param
4427  * dev: Pointer to struct rte_eth_dev.
4428  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4429  * add: if true, add filter, if false, remove filter
4430  *
4431  * @return
4432  *    - On success, zero.
4433  *    - On failure, a negative value.
4434  */
4435 int
4436 igb_add_del_ntuple_filter(struct rte_eth_dev *dev,
4437 			struct rte_eth_ntuple_filter *ntuple_filter,
4438 			bool add)
4439 {
4440 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4441 	int ret;
4442 
4443 	switch (ntuple_filter->flags) {
4444 	case RTE_5TUPLE_FLAGS:
4445 	case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4446 		if (hw->mac.type != e1000_82576)
4447 			return -ENOTSUP;
4448 		if (add)
4449 			ret = igb_add_5tuple_filter_82576(dev,
4450 							  ntuple_filter);
4451 		else
4452 			ret = igb_remove_5tuple_filter_82576(dev,
4453 							     ntuple_filter);
4454 		break;
4455 	case RTE_2TUPLE_FLAGS:
4456 	case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4457 		if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350 &&
4458 			hw->mac.type != e1000_i210 &&
4459 			hw->mac.type != e1000_i211)
4460 			return -ENOTSUP;
4461 		if (add)
4462 			ret = igb_add_2tuple_filter(dev, ntuple_filter);
4463 		else
4464 			ret = igb_remove_2tuple_filter(dev, ntuple_filter);
4465 		break;
4466 	default:
4467 		ret = -EINVAL;
4468 		break;
4469 	}
4470 
4471 	return ret;
4472 }
4473 
4474 static inline int
4475 igb_ethertype_filter_lookup(struct e1000_filter_info *filter_info,
4476 			uint16_t ethertype)
4477 {
4478 	int i;
4479 
4480 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4481 		if (filter_info->ethertype_filters[i].ethertype == ethertype &&
4482 		    (filter_info->ethertype_mask & (1 << i)))
4483 			return i;
4484 	}
4485 	return -1;
4486 }
4487 
4488 static inline int
4489 igb_ethertype_filter_insert(struct e1000_filter_info *filter_info,
4490 			uint16_t ethertype, uint32_t etqf)
4491 {
4492 	int i;
4493 
4494 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4495 		if (!(filter_info->ethertype_mask & (1 << i))) {
4496 			filter_info->ethertype_mask |= 1 << i;
4497 			filter_info->ethertype_filters[i].ethertype = ethertype;
4498 			filter_info->ethertype_filters[i].etqf = etqf;
4499 			return i;
4500 		}
4501 	}
4502 	return -1;
4503 }
4504 
4505 int
4506 igb_ethertype_filter_remove(struct e1000_filter_info *filter_info,
4507 			uint8_t idx)
4508 {
4509 	if (idx >= E1000_MAX_ETQF_FILTERS)
4510 		return -1;
4511 	filter_info->ethertype_mask &= ~(1 << idx);
4512 	filter_info->ethertype_filters[idx].ethertype = 0;
4513 	filter_info->ethertype_filters[idx].etqf = 0;
4514 	return idx;
4515 }
4516 
4517 
4518 int
4519 igb_add_del_ethertype_filter(struct rte_eth_dev *dev,
4520 			struct rte_eth_ethertype_filter *filter,
4521 			bool add)
4522 {
4523 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4524 	struct e1000_filter_info *filter_info =
4525 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4526 	uint32_t etqf = 0;
4527 	int ret;
4528 
4529 	if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
4530 		filter->ether_type == RTE_ETHER_TYPE_IPV6) {
4531 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
4532 			" ethertype filter.", filter->ether_type);
4533 		return -EINVAL;
4534 	}
4535 
4536 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
4537 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
4538 		return -EINVAL;
4539 	}
4540 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
4541 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
4542 		return -EINVAL;
4543 	}
4544 
4545 	ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4546 	if (ret >= 0 && add) {
4547 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
4548 			    filter->ether_type);
4549 		return -EEXIST;
4550 	}
4551 	if (ret < 0 && !add) {
4552 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4553 			    filter->ether_type);
4554 		return -ENOENT;
4555 	}
4556 
4557 	if (add) {
4558 		etqf |= E1000_ETQF_FILTER_ENABLE | E1000_ETQF_QUEUE_ENABLE;
4559 		etqf |= (uint32_t)(filter->ether_type & E1000_ETQF_ETHERTYPE);
4560 		etqf |= filter->queue << E1000_ETQF_QUEUE_SHIFT;
4561 		ret = igb_ethertype_filter_insert(filter_info,
4562 				filter->ether_type, etqf);
4563 		if (ret < 0) {
4564 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
4565 			return -ENOSYS;
4566 		}
4567 	} else {
4568 		ret = igb_ethertype_filter_remove(filter_info, (uint8_t)ret);
4569 		if (ret < 0)
4570 			return -ENOSYS;
4571 	}
4572 	E1000_WRITE_REG(hw, E1000_ETQF(ret), etqf);
4573 	E1000_WRITE_FLUSH(hw);
4574 
4575 	return 0;
4576 }
4577 
4578 static int
4579 eth_igb_flow_ops_get(struct rte_eth_dev *dev __rte_unused,
4580 		     const struct rte_flow_ops **ops)
4581 {
4582 	*ops = &igb_flow_ops;
4583 	return 0;
4584 }
4585 
4586 static int
4587 eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
4588 			 struct rte_ether_addr *mc_addr_set,
4589 			 uint32_t nb_mc_addr)
4590 {
4591 	struct e1000_hw *hw;
4592 
4593 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4594 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
4595 	return 0;
4596 }
4597 
4598 static uint64_t
4599 igb_read_systime_cyclecounter(struct rte_eth_dev *dev)
4600 {
4601 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4602 	uint64_t systime_cycles;
4603 
4604 	switch (hw->mac.type) {
4605 	case e1000_i210:
4606 	case e1000_i211:
4607 		/*
4608 		 * Need to read System Time Residue Register to be able
4609 		 * to read the other two registers.
4610 		 */
4611 		E1000_READ_REG(hw, E1000_SYSTIMR);
4612 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
4613 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4614 		systime_cycles += (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4615 				* NSEC_PER_SEC;
4616 		break;
4617 	case e1000_82580:
4618 	case e1000_i350:
4619 	case e1000_i354:
4620 		/*
4621 		 * Need to read System Time Residue Register to be able
4622 		 * to read the other two registers.
4623 		 */
4624 		E1000_READ_REG(hw, E1000_SYSTIMR);
4625 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4626 		/* Only the 8 LSB are valid. */
4627 		systime_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_SYSTIMH)
4628 				& 0xff) << 32;
4629 		break;
4630 	default:
4631 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4632 		systime_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4633 				<< 32;
4634 		break;
4635 	}
4636 
4637 	return systime_cycles;
4638 }
4639 
4640 static uint64_t
4641 igb_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
4642 {
4643 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4644 	uint64_t rx_tstamp_cycles;
4645 
4646 	switch (hw->mac.type) {
4647 	case e1000_i210:
4648 	case e1000_i211:
4649 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
4650 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4651 		rx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
4652 				* NSEC_PER_SEC;
4653 		break;
4654 	case e1000_82580:
4655 	case e1000_i350:
4656 	case e1000_i354:
4657 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4658 		/* Only the 8 LSB are valid. */
4659 		rx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_RXSTMPH)
4660 				& 0xff) << 32;
4661 		break;
4662 	default:
4663 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4664 		rx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
4665 				<< 32;
4666 		break;
4667 	}
4668 
4669 	return rx_tstamp_cycles;
4670 }
4671 
4672 static uint64_t
4673 igb_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
4674 {
4675 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4676 	uint64_t tx_tstamp_cycles;
4677 
4678 	switch (hw->mac.type) {
4679 	case e1000_i210:
4680 	case e1000_i211:
4681 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
4682 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
4683 		tx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
4684 				* NSEC_PER_SEC;
4685 		break;
4686 	case e1000_82580:
4687 	case e1000_i350:
4688 	case e1000_i354:
4689 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
4690 		/* Only the 8 LSB are valid. */
4691 		tx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_TXSTMPH)
4692 				& 0xff) << 32;
4693 		break;
4694 	default:
4695 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
4696 		tx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
4697 				<< 32;
4698 		break;
4699 	}
4700 
4701 	return tx_tstamp_cycles;
4702 }
4703 
4704 static void
4705 igb_start_timecounters(struct rte_eth_dev *dev)
4706 {
4707 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4708 	struct e1000_adapter *adapter = dev->data->dev_private;
4709 	uint32_t incval = 1;
4710 	uint32_t shift = 0;
4711 	uint64_t mask = E1000_CYCLECOUNTER_MASK;
4712 
4713 	switch (hw->mac.type) {
4714 	case e1000_82580:
4715 	case e1000_i350:
4716 	case e1000_i354:
4717 		/* 32 LSB bits + 8 MSB bits = 40 bits */
4718 		mask = (1ULL << 40) - 1;
4719 		/* fall-through */
4720 	case e1000_i210:
4721 	case e1000_i211:
4722 		/*
4723 		 * Start incrementing the register
4724 		 * used to timestamp PTP packets.
4725 		 */
4726 		E1000_WRITE_REG(hw, E1000_TIMINCA, incval);
4727 		break;
4728 	case e1000_82576:
4729 		incval = E1000_INCVALUE_82576;
4730 		shift = IGB_82576_TSYNC_SHIFT;
4731 		E1000_WRITE_REG(hw, E1000_TIMINCA,
4732 				E1000_INCPERIOD_82576 | incval);
4733 		break;
4734 	default:
4735 		/* Not supported */
4736 		return;
4737 	}
4738 
4739 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
4740 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
4741 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
4742 
4743 	adapter->systime_tc.cc_mask = mask;
4744 	adapter->systime_tc.cc_shift = shift;
4745 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
4746 
4747 	adapter->rx_tstamp_tc.cc_mask = mask;
4748 	adapter->rx_tstamp_tc.cc_shift = shift;
4749 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
4750 
4751 	adapter->tx_tstamp_tc.cc_mask = mask;
4752 	adapter->tx_tstamp_tc.cc_shift = shift;
4753 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
4754 }
4755 
4756 static int
4757 igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
4758 {
4759 	struct e1000_adapter *adapter = dev->data->dev_private;
4760 
4761 	adapter->systime_tc.nsec += delta;
4762 	adapter->rx_tstamp_tc.nsec += delta;
4763 	adapter->tx_tstamp_tc.nsec += delta;
4764 
4765 	return 0;
4766 }
4767 
4768 static int
4769 igb_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
4770 {
4771 	uint64_t ns;
4772 	struct e1000_adapter *adapter = dev->data->dev_private;
4773 
4774 	ns = rte_timespec_to_ns(ts);
4775 
4776 	/* Set the timecounters to a new value. */
4777 	adapter->systime_tc.nsec = ns;
4778 	adapter->rx_tstamp_tc.nsec = ns;
4779 	adapter->tx_tstamp_tc.nsec = ns;
4780 
4781 	return 0;
4782 }
4783 
4784 static int
4785 igb_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
4786 {
4787 	uint64_t ns, systime_cycles;
4788 	struct e1000_adapter *adapter = dev->data->dev_private;
4789 
4790 	systime_cycles = igb_read_systime_cyclecounter(dev);
4791 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
4792 	*ts = rte_ns_to_timespec(ns);
4793 
4794 	return 0;
4795 }
4796 
4797 static int
4798 igb_timesync_enable(struct rte_eth_dev *dev)
4799 {
4800 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4801 	uint32_t tsync_ctl;
4802 	uint32_t tsauxc;
4803 
4804 	/* Stop the timesync system time. */
4805 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0x0);
4806 	/* Reset the timesync system time value. */
4807 	switch (hw->mac.type) {
4808 	case e1000_82580:
4809 	case e1000_i350:
4810 	case e1000_i354:
4811 	case e1000_i210:
4812 	case e1000_i211:
4813 		E1000_WRITE_REG(hw, E1000_SYSTIMR, 0x0);
4814 		/* fall-through */
4815 	case e1000_82576:
4816 		E1000_WRITE_REG(hw, E1000_SYSTIML, 0x0);
4817 		E1000_WRITE_REG(hw, E1000_SYSTIMH, 0x0);
4818 		break;
4819 	default:
4820 		/* Not supported. */
4821 		return -ENOTSUP;
4822 	}
4823 
4824 	/* Enable system time for it isn't on by default. */
4825 	tsauxc = E1000_READ_REG(hw, E1000_TSAUXC);
4826 	tsauxc &= ~E1000_TSAUXC_DISABLE_SYSTIME;
4827 	E1000_WRITE_REG(hw, E1000_TSAUXC, tsauxc);
4828 
4829 	igb_start_timecounters(dev);
4830 
4831 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
4832 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588),
4833 			(RTE_ETHER_TYPE_1588 |
4834 			 E1000_ETQF_FILTER_ENABLE |
4835 			 E1000_ETQF_1588));
4836 
4837 	/* Enable timestamping of received PTP packets. */
4838 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
4839 	tsync_ctl |= E1000_TSYNCRXCTL_ENABLED;
4840 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
4841 
4842 	/* Enable Timestamping of transmitted PTP packets. */
4843 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
4844 	tsync_ctl |= E1000_TSYNCTXCTL_ENABLED;
4845 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
4846 
4847 	return 0;
4848 }
4849 
4850 static int
4851 igb_timesync_disable(struct rte_eth_dev *dev)
4852 {
4853 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4854 	uint32_t tsync_ctl;
4855 
4856 	/* Disable timestamping of transmitted PTP packets. */
4857 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
4858 	tsync_ctl &= ~E1000_TSYNCTXCTL_ENABLED;
4859 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
4860 
4861 	/* Disable timestamping of received PTP packets. */
4862 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
4863 	tsync_ctl &= ~E1000_TSYNCRXCTL_ENABLED;
4864 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
4865 
4866 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
4867 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588), 0);
4868 
4869 	/* Stop incrementating the System Time registers. */
4870 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0);
4871 
4872 	return 0;
4873 }
4874 
4875 static int
4876 igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
4877 			       struct timespec *timestamp,
4878 			       uint32_t flags __rte_unused)
4879 {
4880 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4881 	struct e1000_adapter *adapter = dev->data->dev_private;
4882 	uint32_t tsync_rxctl;
4883 	uint64_t rx_tstamp_cycles;
4884 	uint64_t ns;
4885 
4886 	tsync_rxctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
4887 	if ((tsync_rxctl & E1000_TSYNCRXCTL_VALID) == 0)
4888 		return -EINVAL;
4889 
4890 	rx_tstamp_cycles = igb_read_rx_tstamp_cyclecounter(dev);
4891 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
4892 	*timestamp = rte_ns_to_timespec(ns);
4893 
4894 	return  0;
4895 }
4896 
4897 static int
4898 igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
4899 			       struct timespec *timestamp)
4900 {
4901 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4902 	struct e1000_adapter *adapter = dev->data->dev_private;
4903 	uint32_t tsync_txctl;
4904 	uint64_t tx_tstamp_cycles;
4905 	uint64_t ns;
4906 
4907 	tsync_txctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
4908 	if ((tsync_txctl & E1000_TSYNCTXCTL_VALID) == 0)
4909 		return -EINVAL;
4910 
4911 	tx_tstamp_cycles = igb_read_tx_tstamp_cyclecounter(dev);
4912 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
4913 	*timestamp = rte_ns_to_timespec(ns);
4914 
4915 	return  0;
4916 }
4917 
4918 static int
4919 eth_igb_get_reg_length(struct rte_eth_dev *dev __rte_unused)
4920 {
4921 	int count = 0;
4922 	int g_ind = 0;
4923 	const struct reg_info *reg_group;
4924 
4925 	while ((reg_group = igb_regs[g_ind++]))
4926 		count += igb_reg_group_count(reg_group);
4927 
4928 	return count;
4929 }
4930 
4931 static int
4932 igbvf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
4933 {
4934 	int count = 0;
4935 	int g_ind = 0;
4936 	const struct reg_info *reg_group;
4937 
4938 	while ((reg_group = igbvf_regs[g_ind++]))
4939 		count += igb_reg_group_count(reg_group);
4940 
4941 	return count;
4942 }
4943 
4944 static int
4945 eth_igb_get_regs(struct rte_eth_dev *dev,
4946 	struct rte_dev_reg_info *regs)
4947 {
4948 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4949 	uint32_t *data = regs->data;
4950 	int g_ind = 0;
4951 	int count = 0;
4952 	const struct reg_info *reg_group;
4953 
4954 	if (data == NULL) {
4955 		regs->length = eth_igb_get_reg_length(dev);
4956 		regs->width = sizeof(uint32_t);
4957 		return 0;
4958 	}
4959 
4960 	/* Support only full register dump */
4961 	if ((regs->length == 0) ||
4962 	    (regs->length == (uint32_t)eth_igb_get_reg_length(dev))) {
4963 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
4964 			hw->device_id;
4965 		while ((reg_group = igb_regs[g_ind++]))
4966 			count += igb_read_regs_group(dev, &data[count],
4967 							reg_group);
4968 		return 0;
4969 	}
4970 
4971 	return -ENOTSUP;
4972 }
4973 
4974 static int
4975 igbvf_get_regs(struct rte_eth_dev *dev,
4976 	struct rte_dev_reg_info *regs)
4977 {
4978 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4979 	uint32_t *data = regs->data;
4980 	int g_ind = 0;
4981 	int count = 0;
4982 	const struct reg_info *reg_group;
4983 
4984 	if (data == NULL) {
4985 		regs->length = igbvf_get_reg_length(dev);
4986 		regs->width = sizeof(uint32_t);
4987 		return 0;
4988 	}
4989 
4990 	/* Support only full register dump */
4991 	if ((regs->length == 0) ||
4992 	    (regs->length == (uint32_t)igbvf_get_reg_length(dev))) {
4993 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
4994 			hw->device_id;
4995 		while ((reg_group = igbvf_regs[g_ind++]))
4996 			count += igb_read_regs_group(dev, &data[count],
4997 							reg_group);
4998 		return 0;
4999 	}
5000 
5001 	return -ENOTSUP;
5002 }
5003 
5004 static int
5005 eth_igb_get_eeprom_length(struct rte_eth_dev *dev)
5006 {
5007 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5008 
5009 	/* Return unit is byte count */
5010 	return hw->nvm.word_size * 2;
5011 }
5012 
5013 static int
5014 eth_igb_get_eeprom(struct rte_eth_dev *dev,
5015 	struct rte_dev_eeprom_info *in_eeprom)
5016 {
5017 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5018 	struct e1000_nvm_info *nvm = &hw->nvm;
5019 	uint16_t *data = in_eeprom->data;
5020 	int first, length;
5021 
5022 	first = in_eeprom->offset >> 1;
5023 	length = in_eeprom->length >> 1;
5024 	if ((first >= hw->nvm.word_size) ||
5025 	    ((first + length) >= hw->nvm.word_size))
5026 		return -EINVAL;
5027 
5028 	in_eeprom->magic = hw->vendor_id |
5029 		((uint32_t)hw->device_id << 16);
5030 
5031 	if ((nvm->ops.read) == NULL)
5032 		return -ENOTSUP;
5033 
5034 	return nvm->ops.read(hw, first, length, data);
5035 }
5036 
5037 static int
5038 eth_igb_set_eeprom(struct rte_eth_dev *dev,
5039 	struct rte_dev_eeprom_info *in_eeprom)
5040 {
5041 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5042 	struct e1000_nvm_info *nvm = &hw->nvm;
5043 	uint16_t *data = in_eeprom->data;
5044 	int first, length;
5045 
5046 	first = in_eeprom->offset >> 1;
5047 	length = in_eeprom->length >> 1;
5048 	if ((first >= hw->nvm.word_size) ||
5049 	    ((first + length) >= hw->nvm.word_size))
5050 		return -EINVAL;
5051 
5052 	in_eeprom->magic = (uint32_t)hw->vendor_id |
5053 		((uint32_t)hw->device_id << 16);
5054 
5055 	if ((nvm->ops.write) == NULL)
5056 		return -ENOTSUP;
5057 	return nvm->ops.write(hw,  first, length, data);
5058 }
5059 
5060 static int
5061 eth_igb_get_module_info(struct rte_eth_dev *dev,
5062 			struct rte_eth_dev_module_info *modinfo)
5063 {
5064 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5065 
5066 	uint32_t status = 0;
5067 	uint16_t sff8472_rev, addr_mode;
5068 	bool page_swap = false;
5069 
5070 	if (hw->phy.media_type == e1000_media_type_copper ||
5071 	    hw->phy.media_type == e1000_media_type_unknown)
5072 		return -EOPNOTSUPP;
5073 
5074 	/* Check whether we support SFF-8472 or not */
5075 	status = e1000_read_phy_reg_i2c(hw, IGB_SFF_8472_COMP, &sff8472_rev);
5076 	if (status)
5077 		return -EIO;
5078 
5079 	/* addressing mode is not supported */
5080 	status = e1000_read_phy_reg_i2c(hw, IGB_SFF_8472_SWAP, &addr_mode);
5081 	if (status)
5082 		return -EIO;
5083 
5084 	/* addressing mode is not supported */
5085 	if ((addr_mode & 0xFF) & IGB_SFF_ADDRESSING_MODE) {
5086 		PMD_DRV_LOG(ERR,
5087 			    "Address change required to access page 0xA2, "
5088 			    "but not supported. Please report the module "
5089 			    "type to the driver maintainers.\n");
5090 		page_swap = true;
5091 	}
5092 
5093 	if ((sff8472_rev & 0xFF) == IGB_SFF_8472_UNSUP || page_swap) {
5094 		/* We have an SFP, but it does not support SFF-8472 */
5095 		modinfo->type = RTE_ETH_MODULE_SFF_8079;
5096 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
5097 	} else {
5098 		/* We have an SFP which supports a revision of SFF-8472 */
5099 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
5100 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
5101 	}
5102 
5103 	return 0;
5104 }
5105 
5106 static int
5107 eth_igb_get_module_eeprom(struct rte_eth_dev *dev,
5108 			  struct rte_dev_eeprom_info *info)
5109 {
5110 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5111 
5112 	uint32_t status = 0;
5113 	uint16_t dataword[RTE_ETH_MODULE_SFF_8472_LEN / 2 + 1];
5114 	u16 first_word, last_word;
5115 	int i = 0;
5116 
5117 	first_word = info->offset >> 1;
5118 	last_word = (info->offset + info->length - 1) >> 1;
5119 
5120 	/* Read EEPROM block, SFF-8079/SFF-8472, word at a time */
5121 	for (i = 0; i < last_word - first_word + 1; i++) {
5122 		status = e1000_read_phy_reg_i2c(hw, (first_word + i) * 2,
5123 						&dataword[i]);
5124 		if (status) {
5125 			/* Error occurred while reading module */
5126 			return -EIO;
5127 		}
5128 
5129 		dataword[i] = rte_be_to_cpu_16(dataword[i]);
5130 	}
5131 
5132 	memcpy(info->data, (u8 *)dataword + (info->offset & 1), info->length);
5133 
5134 	return 0;
5135 }
5136 
5137 static int
5138 eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5139 {
5140 	struct e1000_hw *hw =
5141 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5142 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5143 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5144 	uint32_t vec = E1000_MISC_VEC_ID;
5145 
5146 	if (rte_intr_allow_others(intr_handle))
5147 		vec = E1000_RX_VEC_START;
5148 
5149 	uint32_t mask = 1 << (queue_id + vec);
5150 
5151 	E1000_WRITE_REG(hw, E1000_EIMC, mask);
5152 	E1000_WRITE_FLUSH(hw);
5153 
5154 	return 0;
5155 }
5156 
5157 static int
5158 eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5159 {
5160 	struct e1000_hw *hw =
5161 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5162 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5163 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5164 	uint32_t vec = E1000_MISC_VEC_ID;
5165 
5166 	if (rte_intr_allow_others(intr_handle))
5167 		vec = E1000_RX_VEC_START;
5168 
5169 	uint32_t mask = 1 << (queue_id + vec);
5170 	uint32_t regval;
5171 
5172 	regval = E1000_READ_REG(hw, E1000_EIMS);
5173 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
5174 	E1000_WRITE_FLUSH(hw);
5175 
5176 	rte_intr_ack(intr_handle);
5177 
5178 	return 0;
5179 }
5180 
5181 static void
5182 eth_igb_write_ivar(struct e1000_hw *hw, uint8_t  msix_vector,
5183 		   uint8_t index, uint8_t offset)
5184 {
5185 	uint32_t val = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
5186 
5187 	/* clear bits */
5188 	val &= ~((uint32_t)0xFF << offset);
5189 
5190 	/* write vector and valid bit */
5191 	val |= (msix_vector | E1000_IVAR_VALID) << offset;
5192 
5193 	E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, val);
5194 }
5195 
5196 static void
5197 eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
5198 			   uint8_t queue, uint8_t msix_vector)
5199 {
5200 	uint32_t tmp = 0;
5201 
5202 	if (hw->mac.type == e1000_82575) {
5203 		if (direction == 0)
5204 			tmp = E1000_EICR_RX_QUEUE0 << queue;
5205 		else if (direction == 1)
5206 			tmp = E1000_EICR_TX_QUEUE0 << queue;
5207 		E1000_WRITE_REG(hw, E1000_MSIXBM(msix_vector), tmp);
5208 	} else if (hw->mac.type == e1000_82576) {
5209 		if ((direction == 0) || (direction == 1))
5210 			eth_igb_write_ivar(hw, msix_vector, queue & 0x7,
5211 					   ((queue & 0x8) << 1) +
5212 					   8 * direction);
5213 	} else if ((hw->mac.type == e1000_82580) ||
5214 			(hw->mac.type == e1000_i350) ||
5215 			(hw->mac.type == e1000_i354) ||
5216 			(hw->mac.type == e1000_i210) ||
5217 			(hw->mac.type == e1000_i211)) {
5218 		if ((direction == 0) || (direction == 1))
5219 			eth_igb_write_ivar(hw, msix_vector,
5220 					   queue >> 1,
5221 					   ((queue & 0x1) << 4) +
5222 					   8 * direction);
5223 	}
5224 }
5225 
5226 /* Sets up the hardware to generate MSI-X interrupts properly
5227  * @hw
5228  *  board private structure
5229  */
5230 static void
5231 eth_igb_configure_msix_intr(struct rte_eth_dev *dev)
5232 {
5233 	int queue_id;
5234 	uint32_t tmpval, regval, intr_mask;
5235 	struct e1000_hw *hw =
5236 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5237 	uint32_t vec = E1000_MISC_VEC_ID;
5238 	uint32_t base = E1000_MISC_VEC_ID;
5239 	uint32_t misc_shift = 0;
5240 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5241 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5242 
5243 	/* won't configure msix register if no mapping is done
5244 	 * between intr vector and event fd
5245 	 */
5246 	if (!rte_intr_dp_is_en(intr_handle))
5247 		return;
5248 
5249 	if (rte_intr_allow_others(intr_handle)) {
5250 		vec = base = E1000_RX_VEC_START;
5251 		misc_shift = 1;
5252 	}
5253 
5254 	/* set interrupt vector for other causes */
5255 	if (hw->mac.type == e1000_82575) {
5256 		tmpval = E1000_READ_REG(hw, E1000_CTRL_EXT);
5257 		/* enable MSI-X PBA support */
5258 		tmpval |= E1000_CTRL_EXT_PBA_CLR;
5259 
5260 		/* Auto-Mask interrupts upon ICR read */
5261 		tmpval |= E1000_CTRL_EXT_EIAME;
5262 		tmpval |= E1000_CTRL_EXT_IRCA;
5263 
5264 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmpval);
5265 
5266 		/* enable msix_other interrupt */
5267 		E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0), 0, E1000_EIMS_OTHER);
5268 		regval = E1000_READ_REG(hw, E1000_EIAC);
5269 		E1000_WRITE_REG(hw, E1000_EIAC, regval | E1000_EIMS_OTHER);
5270 		regval = E1000_READ_REG(hw, E1000_EIAM);
5271 		E1000_WRITE_REG(hw, E1000_EIMS, regval | E1000_EIMS_OTHER);
5272 	} else if ((hw->mac.type == e1000_82576) ||
5273 			(hw->mac.type == e1000_82580) ||
5274 			(hw->mac.type == e1000_i350) ||
5275 			(hw->mac.type == e1000_i354) ||
5276 			(hw->mac.type == e1000_i210) ||
5277 			(hw->mac.type == e1000_i211)) {
5278 		/* turn on MSI-X capability first */
5279 		E1000_WRITE_REG(hw, E1000_GPIE, E1000_GPIE_MSIX_MODE |
5280 					E1000_GPIE_PBA | E1000_GPIE_EIAME |
5281 					E1000_GPIE_NSICR);
5282 		intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5283 			misc_shift;
5284 
5285 		if (dev->data->dev_conf.intr_conf.lsc != 0)
5286 			intr_mask |= (1 << IGB_MSIX_OTHER_INTR_VEC);
5287 
5288 		regval = E1000_READ_REG(hw, E1000_EIAC);
5289 		E1000_WRITE_REG(hw, E1000_EIAC, regval | intr_mask);
5290 
5291 		/* enable msix_other interrupt */
5292 		regval = E1000_READ_REG(hw, E1000_EIMS);
5293 		E1000_WRITE_REG(hw, E1000_EIMS, regval | intr_mask);
5294 		tmpval = (IGB_MSIX_OTHER_INTR_VEC | E1000_IVAR_VALID) << 8;
5295 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, tmpval);
5296 	}
5297 
5298 	/* use EIAM to auto-mask when MSI-X interrupt
5299 	 * is asserted, this saves a register write for every interrupt
5300 	 */
5301 	intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5302 		misc_shift;
5303 
5304 	if (dev->data->dev_conf.intr_conf.lsc != 0)
5305 		intr_mask |= (1 << IGB_MSIX_OTHER_INTR_VEC);
5306 
5307 	regval = E1000_READ_REG(hw, E1000_EIAM);
5308 	E1000_WRITE_REG(hw, E1000_EIAM, regval | intr_mask);
5309 
5310 	for (queue_id = 0; queue_id < dev->data->nb_rx_queues; queue_id++) {
5311 		eth_igb_assign_msix_vector(hw, 0, queue_id, vec);
5312 		intr_handle->intr_vec[queue_id] = vec;
5313 		if (vec < base + intr_handle->nb_efd - 1)
5314 			vec++;
5315 	}
5316 
5317 	E1000_WRITE_FLUSH(hw);
5318 }
5319 
5320 /* restore n-tuple filter */
5321 static inline void
5322 igb_ntuple_filter_restore(struct rte_eth_dev *dev)
5323 {
5324 	struct e1000_filter_info *filter_info =
5325 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5326 	struct e1000_5tuple_filter *p_5tuple;
5327 	struct e1000_2tuple_filter *p_2tuple;
5328 
5329 	TAILQ_FOREACH(p_5tuple, &filter_info->fivetuple_list, entries) {
5330 		igb_inject_5tuple_filter_82576(dev, p_5tuple);
5331 	}
5332 
5333 	TAILQ_FOREACH(p_2tuple, &filter_info->twotuple_list, entries) {
5334 		igb_inject_2uple_filter(dev, p_2tuple);
5335 	}
5336 }
5337 
5338 /* restore SYN filter */
5339 static inline void
5340 igb_syn_filter_restore(struct rte_eth_dev *dev)
5341 {
5342 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5343 	struct e1000_filter_info *filter_info =
5344 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5345 	uint32_t synqf;
5346 
5347 	synqf = filter_info->syn_info;
5348 
5349 	if (synqf & E1000_SYN_FILTER_ENABLE) {
5350 		E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
5351 		E1000_WRITE_FLUSH(hw);
5352 	}
5353 }
5354 
5355 /* restore ethernet type filter */
5356 static inline void
5357 igb_ethertype_filter_restore(struct rte_eth_dev *dev)
5358 {
5359 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5360 	struct e1000_filter_info *filter_info =
5361 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5362 	int i;
5363 
5364 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
5365 		if (filter_info->ethertype_mask & (1 << i)) {
5366 			E1000_WRITE_REG(hw, E1000_ETQF(i),
5367 				filter_info->ethertype_filters[i].etqf);
5368 			E1000_WRITE_FLUSH(hw);
5369 		}
5370 	}
5371 }
5372 
5373 /* restore flex byte filter */
5374 static inline void
5375 igb_flex_filter_restore(struct rte_eth_dev *dev)
5376 {
5377 	struct e1000_filter_info *filter_info =
5378 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5379 	struct e1000_flex_filter *flex_filter;
5380 
5381 	TAILQ_FOREACH(flex_filter, &filter_info->flex_list, entries) {
5382 		igb_inject_flex_filter(dev, flex_filter);
5383 	}
5384 }
5385 
5386 /* restore rss filter */
5387 static inline void
5388 igb_rss_filter_restore(struct rte_eth_dev *dev)
5389 {
5390 	struct e1000_filter_info *filter_info =
5391 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5392 
5393 	if (filter_info->rss_info.conf.queue_num)
5394 		igb_config_rss_filter(dev, &filter_info->rss_info, TRUE);
5395 }
5396 
5397 /* restore all types filter */
5398 static int
5399 igb_filter_restore(struct rte_eth_dev *dev)
5400 {
5401 	igb_ntuple_filter_restore(dev);
5402 	igb_ethertype_filter_restore(dev);
5403 	igb_syn_filter_restore(dev);
5404 	igb_flex_filter_restore(dev);
5405 	igb_rss_filter_restore(dev);
5406 
5407 	return 0;
5408 }
5409 
5410 RTE_PMD_REGISTER_PCI(net_e1000_igb, rte_igb_pmd);
5411 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb, pci_id_igb_map);
5412 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb, "* igb_uio | uio_pci_generic | vfio-pci");
5413 RTE_PMD_REGISTER_PCI(net_e1000_igb_vf, rte_igbvf_pmd);
5414 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb_vf, pci_id_igbvf_map);
5415 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb_vf, "* igb_uio | vfio-pci");
5416