xref: /f-stack/dpdk/drivers/net/e1000/igb_ethdev.c (revision ebf5cedb)
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 <rte_ethdev_driver.h>
21 #include <rte_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 void 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 void 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 void igbvf_dev_stop(struct rte_eth_dev *dev);
158 static void 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 eth_igb_syn_filter_get(struct rte_eth_dev *dev,
190 			struct rte_eth_syn_filter *filter);
191 static int eth_igb_syn_filter_handle(struct rte_eth_dev *dev,
192 			enum rte_filter_op filter_op,
193 			void *arg);
194 static int igb_add_2tuple_filter(struct rte_eth_dev *dev,
195 			struct rte_eth_ntuple_filter *ntuple_filter);
196 static int igb_remove_2tuple_filter(struct rte_eth_dev *dev,
197 			struct rte_eth_ntuple_filter *ntuple_filter);
198 static int eth_igb_get_flex_filter(struct rte_eth_dev *dev,
199 			struct rte_eth_flex_filter *filter);
200 static int eth_igb_flex_filter_handle(struct rte_eth_dev *dev,
201 			enum rte_filter_op filter_op,
202 			void *arg);
203 static int igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
204 			struct rte_eth_ntuple_filter *ntuple_filter);
205 static int igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
206 			struct rte_eth_ntuple_filter *ntuple_filter);
207 static int igb_get_ntuple_filter(struct rte_eth_dev *dev,
208 			struct rte_eth_ntuple_filter *filter);
209 static int igb_ntuple_filter_handle(struct rte_eth_dev *dev,
210 				enum rte_filter_op filter_op,
211 				void *arg);
212 static int igb_ethertype_filter_handle(struct rte_eth_dev *dev,
213 				enum rte_filter_op filter_op,
214 				void *arg);
215 static int igb_get_ethertype_filter(struct rte_eth_dev *dev,
216 			struct rte_eth_ethertype_filter *filter);
217 static int eth_igb_filter_ctrl(struct rte_eth_dev *dev,
218 		     enum rte_filter_type filter_type,
219 		     enum rte_filter_op filter_op,
220 		     void *arg);
221 static int eth_igb_get_reg_length(struct rte_eth_dev *dev);
222 static int eth_igb_get_regs(struct rte_eth_dev *dev,
223 		struct rte_dev_reg_info *regs);
224 static int eth_igb_get_eeprom_length(struct rte_eth_dev *dev);
225 static int eth_igb_get_eeprom(struct rte_eth_dev *dev,
226 		struct rte_dev_eeprom_info *eeprom);
227 static int eth_igb_set_eeprom(struct rte_eth_dev *dev,
228 		struct rte_dev_eeprom_info *eeprom);
229 static int eth_igb_get_module_info(struct rte_eth_dev *dev,
230 				   struct rte_eth_dev_module_info *modinfo);
231 static int eth_igb_get_module_eeprom(struct rte_eth_dev *dev,
232 				     struct rte_dev_eeprom_info *info);
233 static int eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
234 				    struct rte_ether_addr *mc_addr_set,
235 				    uint32_t nb_mc_addr);
236 static int igb_timesync_enable(struct rte_eth_dev *dev);
237 static int igb_timesync_disable(struct rte_eth_dev *dev);
238 static int igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
239 					  struct timespec *timestamp,
240 					  uint32_t flags);
241 static int igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
242 					  struct timespec *timestamp);
243 static int igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
244 static int igb_timesync_read_time(struct rte_eth_dev *dev,
245 				  struct timespec *timestamp);
246 static int igb_timesync_write_time(struct rte_eth_dev *dev,
247 				   const struct timespec *timestamp);
248 static int eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev,
249 					uint16_t queue_id);
250 static int eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev,
251 					 uint16_t queue_id);
252 static void eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
253 				       uint8_t queue, uint8_t msix_vector);
254 static void eth_igb_write_ivar(struct e1000_hw *hw, uint8_t msix_vector,
255 			       uint8_t index, uint8_t offset);
256 static void eth_igb_configure_msix_intr(struct rte_eth_dev *dev);
257 static void eth_igbvf_interrupt_handler(void *param);
258 static void igbvf_mbx_process(struct rte_eth_dev *dev);
259 static int igb_filter_restore(struct rte_eth_dev *dev);
260 
261 /*
262  * Define VF Stats MACRO for Non "cleared on read" register
263  */
264 #define UPDATE_VF_STAT(reg, last, cur)            \
265 {                                                 \
266 	u32 latest = E1000_READ_REG(hw, reg);     \
267 	cur += (latest - last) & UINT_MAX;        \
268 	last = latest;                            \
269 }
270 
271 #define IGB_FC_PAUSE_TIME 0x0680
272 #define IGB_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
273 #define IGB_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
274 
275 #define IGBVF_PMD_NAME "rte_igbvf_pmd"     /* PMD name */
276 
277 static enum e1000_fc_mode igb_fc_setting = e1000_fc_full;
278 
279 /*
280  * The set of PCI devices this driver supports
281  */
282 static const struct rte_pci_id pci_id_igb_map[] = {
283 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576) },
284 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_FIBER) },
285 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_SERDES) },
286 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_QUAD_COPPER) },
287 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_QUAD_COPPER_ET2) },
288 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_NS) },
289 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_NS_SERDES) },
290 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_SERDES_QUAD) },
291 
292 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575EB_COPPER) },
293 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575EB_FIBER_SERDES) },
294 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575GB_QUAD_COPPER) },
295 
296 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_COPPER) },
297 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_FIBER) },
298 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_SERDES) },
299 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_SGMII) },
300 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_COPPER_DUAL) },
301 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_QUAD_FIBER) },
302 
303 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_COPPER) },
304 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_FIBER) },
305 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_SERDES) },
306 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_SGMII) },
307 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_DA4) },
308 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER) },
309 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_OEM1) },
310 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_IT) },
311 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_FIBER) },
312 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SERDES) },
313 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SGMII) },
314 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_FLASHLESS) },
315 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SERDES_FLASHLESS) },
316 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I211_COPPER) },
317 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
318 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_SGMII) },
319 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
320 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SGMII) },
321 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SERDES) },
322 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_BACKPLANE) },
323 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SFP) },
324 	{ .vendor_id = 0, /* sentinel */ },
325 };
326 
327 /*
328  * The set of PCI devices this driver supports (for 82576&I350 VF)
329  */
330 static const struct rte_pci_id pci_id_igbvf_map[] = {
331 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_VF) },
332 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_VF_HV) },
333 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_VF) },
334 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_VF_HV) },
335 	{ .vendor_id = 0, /* sentinel */ },
336 };
337 
338 static const struct rte_eth_desc_lim rx_desc_lim = {
339 	.nb_max = E1000_MAX_RING_DESC,
340 	.nb_min = E1000_MIN_RING_DESC,
341 	.nb_align = IGB_RXD_ALIGN,
342 };
343 
344 static const struct rte_eth_desc_lim tx_desc_lim = {
345 	.nb_max = E1000_MAX_RING_DESC,
346 	.nb_min = E1000_MIN_RING_DESC,
347 	.nb_align = IGB_RXD_ALIGN,
348 	.nb_seg_max = IGB_TX_MAX_SEG,
349 	.nb_mtu_seg_max = IGB_TX_MAX_MTU_SEG,
350 };
351 
352 static const struct eth_dev_ops eth_igb_ops = {
353 	.dev_configure        = eth_igb_configure,
354 	.dev_start            = eth_igb_start,
355 	.dev_stop             = eth_igb_stop,
356 	.dev_set_link_up      = eth_igb_dev_set_link_up,
357 	.dev_set_link_down    = eth_igb_dev_set_link_down,
358 	.dev_close            = eth_igb_close,
359 	.dev_reset            = eth_igb_reset,
360 	.promiscuous_enable   = eth_igb_promiscuous_enable,
361 	.promiscuous_disable  = eth_igb_promiscuous_disable,
362 	.allmulticast_enable  = eth_igb_allmulticast_enable,
363 	.allmulticast_disable = eth_igb_allmulticast_disable,
364 	.link_update          = eth_igb_link_update,
365 	.stats_get            = eth_igb_stats_get,
366 	.xstats_get           = eth_igb_xstats_get,
367 	.xstats_get_by_id     = eth_igb_xstats_get_by_id,
368 	.xstats_get_names_by_id = eth_igb_xstats_get_names_by_id,
369 	.xstats_get_names     = eth_igb_xstats_get_names,
370 	.stats_reset          = eth_igb_stats_reset,
371 	.xstats_reset         = eth_igb_xstats_reset,
372 	.fw_version_get       = eth_igb_fw_version_get,
373 	.dev_infos_get        = eth_igb_infos_get,
374 	.dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
375 	.mtu_set              = eth_igb_mtu_set,
376 	.vlan_filter_set      = eth_igb_vlan_filter_set,
377 	.vlan_tpid_set        = eth_igb_vlan_tpid_set,
378 	.vlan_offload_set     = eth_igb_vlan_offload_set,
379 	.rx_queue_setup       = eth_igb_rx_queue_setup,
380 	.rx_queue_intr_enable = eth_igb_rx_queue_intr_enable,
381 	.rx_queue_intr_disable = eth_igb_rx_queue_intr_disable,
382 	.rx_queue_release     = eth_igb_rx_queue_release,
383 	.rx_queue_count       = eth_igb_rx_queue_count,
384 	.rx_descriptor_done   = eth_igb_rx_descriptor_done,
385 	.rx_descriptor_status = eth_igb_rx_descriptor_status,
386 	.tx_descriptor_status = eth_igb_tx_descriptor_status,
387 	.tx_queue_setup       = eth_igb_tx_queue_setup,
388 	.tx_queue_release     = eth_igb_tx_queue_release,
389 	.tx_done_cleanup      = eth_igb_tx_done_cleanup,
390 	.dev_led_on           = eth_igb_led_on,
391 	.dev_led_off          = eth_igb_led_off,
392 	.flow_ctrl_get        = eth_igb_flow_ctrl_get,
393 	.flow_ctrl_set        = eth_igb_flow_ctrl_set,
394 	.mac_addr_add         = eth_igb_rar_set,
395 	.mac_addr_remove      = eth_igb_rar_clear,
396 	.mac_addr_set         = eth_igb_default_mac_addr_set,
397 	.reta_update          = eth_igb_rss_reta_update,
398 	.reta_query           = eth_igb_rss_reta_query,
399 	.rss_hash_update      = eth_igb_rss_hash_update,
400 	.rss_hash_conf_get    = eth_igb_rss_hash_conf_get,
401 	.filter_ctrl          = eth_igb_filter_ctrl,
402 	.set_mc_addr_list     = eth_igb_set_mc_addr_list,
403 	.rxq_info_get         = igb_rxq_info_get,
404 	.txq_info_get         = igb_txq_info_get,
405 	.timesync_enable      = igb_timesync_enable,
406 	.timesync_disable     = igb_timesync_disable,
407 	.timesync_read_rx_timestamp = igb_timesync_read_rx_timestamp,
408 	.timesync_read_tx_timestamp = igb_timesync_read_tx_timestamp,
409 	.get_reg              = eth_igb_get_regs,
410 	.get_eeprom_length    = eth_igb_get_eeprom_length,
411 	.get_eeprom           = eth_igb_get_eeprom,
412 	.set_eeprom           = eth_igb_set_eeprom,
413 	.get_module_info      = eth_igb_get_module_info,
414 	.get_module_eeprom    = eth_igb_get_module_eeprom,
415 	.timesync_adjust_time = igb_timesync_adjust_time,
416 	.timesync_read_time   = igb_timesync_read_time,
417 	.timesync_write_time  = igb_timesync_write_time,
418 };
419 
420 /*
421  * dev_ops for virtual function, bare necessities for basic vf
422  * operation have been implemented
423  */
424 static const struct eth_dev_ops igbvf_eth_dev_ops = {
425 	.dev_configure        = igbvf_dev_configure,
426 	.dev_start            = igbvf_dev_start,
427 	.dev_stop             = igbvf_dev_stop,
428 	.dev_close            = igbvf_dev_close,
429 	.promiscuous_enable   = igbvf_promiscuous_enable,
430 	.promiscuous_disable  = igbvf_promiscuous_disable,
431 	.allmulticast_enable  = igbvf_allmulticast_enable,
432 	.allmulticast_disable = igbvf_allmulticast_disable,
433 	.link_update          = eth_igb_link_update,
434 	.stats_get            = eth_igbvf_stats_get,
435 	.xstats_get           = eth_igbvf_xstats_get,
436 	.xstats_get_names     = eth_igbvf_xstats_get_names,
437 	.stats_reset          = eth_igbvf_stats_reset,
438 	.xstats_reset         = eth_igbvf_stats_reset,
439 	.vlan_filter_set      = igbvf_vlan_filter_set,
440 	.dev_infos_get        = eth_igbvf_infos_get,
441 	.dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
442 	.rx_queue_setup       = eth_igb_rx_queue_setup,
443 	.rx_queue_release     = eth_igb_rx_queue_release,
444 	.rx_descriptor_done   = eth_igb_rx_descriptor_done,
445 	.rx_descriptor_status = eth_igb_rx_descriptor_status,
446 	.tx_descriptor_status = eth_igb_tx_descriptor_status,
447 	.tx_queue_setup       = eth_igb_tx_queue_setup,
448 	.tx_queue_release     = eth_igb_tx_queue_release,
449 	.set_mc_addr_list     = eth_igb_set_mc_addr_list,
450 	.rxq_info_get         = igb_rxq_info_get,
451 	.txq_info_get         = igb_txq_info_get,
452 	.mac_addr_set         = igbvf_default_mac_addr_set,
453 	.get_reg              = igbvf_get_regs,
454 };
455 
456 /* store statistics names and its offset in stats structure */
457 struct rte_igb_xstats_name_off {
458 	char name[RTE_ETH_XSTATS_NAME_SIZE];
459 	unsigned offset;
460 };
461 
462 static const struct rte_igb_xstats_name_off rte_igb_stats_strings[] = {
463 	{"rx_crc_errors", offsetof(struct e1000_hw_stats, crcerrs)},
464 	{"rx_align_errors", offsetof(struct e1000_hw_stats, algnerrc)},
465 	{"rx_symbol_errors", offsetof(struct e1000_hw_stats, symerrs)},
466 	{"rx_missed_packets", offsetof(struct e1000_hw_stats, mpc)},
467 	{"tx_single_collision_packets", offsetof(struct e1000_hw_stats, scc)},
468 	{"tx_multiple_collision_packets", offsetof(struct e1000_hw_stats, mcc)},
469 	{"tx_excessive_collision_packets", offsetof(struct e1000_hw_stats,
470 		ecol)},
471 	{"tx_late_collisions", offsetof(struct e1000_hw_stats, latecol)},
472 	{"tx_total_collisions", offsetof(struct e1000_hw_stats, colc)},
473 	{"tx_deferred_packets", offsetof(struct e1000_hw_stats, dc)},
474 	{"tx_no_carrier_sense_packets", offsetof(struct e1000_hw_stats, tncrs)},
475 	{"rx_carrier_ext_errors", offsetof(struct e1000_hw_stats, cexterr)},
476 	{"rx_length_errors", offsetof(struct e1000_hw_stats, rlec)},
477 	{"rx_xon_packets", offsetof(struct e1000_hw_stats, xonrxc)},
478 	{"tx_xon_packets", offsetof(struct e1000_hw_stats, xontxc)},
479 	{"rx_xoff_packets", offsetof(struct e1000_hw_stats, xoffrxc)},
480 	{"tx_xoff_packets", offsetof(struct e1000_hw_stats, xofftxc)},
481 	{"rx_flow_control_unsupported_packets", offsetof(struct e1000_hw_stats,
482 		fcruc)},
483 	{"rx_size_64_packets", offsetof(struct e1000_hw_stats, prc64)},
484 	{"rx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, prc127)},
485 	{"rx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, prc255)},
486 	{"rx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, prc511)},
487 	{"rx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
488 		prc1023)},
489 	{"rx_size_1024_to_max_packets", offsetof(struct e1000_hw_stats,
490 		prc1522)},
491 	{"rx_broadcast_packets", offsetof(struct e1000_hw_stats, bprc)},
492 	{"rx_multicast_packets", offsetof(struct e1000_hw_stats, mprc)},
493 	{"rx_undersize_errors", offsetof(struct e1000_hw_stats, ruc)},
494 	{"rx_fragment_errors", offsetof(struct e1000_hw_stats, rfc)},
495 	{"rx_oversize_errors", offsetof(struct e1000_hw_stats, roc)},
496 	{"rx_jabber_errors", offsetof(struct e1000_hw_stats, rjc)},
497 	{"rx_management_packets", offsetof(struct e1000_hw_stats, mgprc)},
498 	{"rx_management_dropped", offsetof(struct e1000_hw_stats, mgpdc)},
499 	{"tx_management_packets", offsetof(struct e1000_hw_stats, mgptc)},
500 	{"rx_total_packets", offsetof(struct e1000_hw_stats, tpr)},
501 	{"tx_total_packets", offsetof(struct e1000_hw_stats, tpt)},
502 	{"rx_total_bytes", offsetof(struct e1000_hw_stats, tor)},
503 	{"tx_total_bytes", offsetof(struct e1000_hw_stats, tot)},
504 	{"tx_size_64_packets", offsetof(struct e1000_hw_stats, ptc64)},
505 	{"tx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, ptc127)},
506 	{"tx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, ptc255)},
507 	{"tx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, ptc511)},
508 	{"tx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
509 		ptc1023)},
510 	{"tx_size_1023_to_max_packets", offsetof(struct e1000_hw_stats,
511 		ptc1522)},
512 	{"tx_multicast_packets", offsetof(struct e1000_hw_stats, mptc)},
513 	{"tx_broadcast_packets", offsetof(struct e1000_hw_stats, bptc)},
514 	{"tx_tso_packets", offsetof(struct e1000_hw_stats, tsctc)},
515 	{"tx_tso_errors", offsetof(struct e1000_hw_stats, tsctfc)},
516 	{"rx_sent_to_host_packets", offsetof(struct e1000_hw_stats, rpthc)},
517 	{"tx_sent_by_host_packets", offsetof(struct e1000_hw_stats, hgptc)},
518 	{"rx_code_violation_packets", offsetof(struct e1000_hw_stats, scvpc)},
519 
520 	{"interrupt_assert_count", offsetof(struct e1000_hw_stats, iac)},
521 };
522 
523 #define IGB_NB_XSTATS (sizeof(rte_igb_stats_strings) / \
524 		sizeof(rte_igb_stats_strings[0]))
525 
526 static const struct rte_igb_xstats_name_off rte_igbvf_stats_strings[] = {
527 	{"rx_multicast_packets", offsetof(struct e1000_vf_stats, mprc)},
528 	{"rx_good_loopback_packets", offsetof(struct e1000_vf_stats, gprlbc)},
529 	{"tx_good_loopback_packets", offsetof(struct e1000_vf_stats, gptlbc)},
530 	{"rx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gorlbc)},
531 	{"tx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gotlbc)},
532 };
533 
534 #define IGBVF_NB_XSTATS (sizeof(rte_igbvf_stats_strings) / \
535 		sizeof(rte_igbvf_stats_strings[0]))
536 
537 
538 static inline void
539 igb_intr_enable(struct rte_eth_dev *dev)
540 {
541 	struct e1000_interrupt *intr =
542 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
543 	struct e1000_hw *hw =
544 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
545 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
546 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
547 
548 	if (rte_intr_allow_others(intr_handle) &&
549 		dev->data->dev_conf.intr_conf.lsc != 0) {
550 		E1000_WRITE_REG(hw, E1000_EIMS, 1 << IGB_MSIX_OTHER_INTR_VEC);
551 	}
552 
553 	E1000_WRITE_REG(hw, E1000_IMS, intr->mask);
554 	E1000_WRITE_FLUSH(hw);
555 }
556 
557 static void
558 igb_intr_disable(struct rte_eth_dev *dev)
559 {
560 	struct e1000_hw *hw =
561 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
562 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
563 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
564 
565 	if (rte_intr_allow_others(intr_handle) &&
566 		dev->data->dev_conf.intr_conf.lsc != 0) {
567 		E1000_WRITE_REG(hw, E1000_EIMC, 1 << IGB_MSIX_OTHER_INTR_VEC);
568 	}
569 
570 	E1000_WRITE_REG(hw, E1000_IMC, ~0);
571 	E1000_WRITE_FLUSH(hw);
572 }
573 
574 static inline void
575 igbvf_intr_enable(struct rte_eth_dev *dev)
576 {
577 	struct e1000_hw *hw =
578 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
579 
580 	/* only for mailbox */
581 	E1000_WRITE_REG(hw, E1000_EIAM, 1 << E1000_VTIVAR_MISC_MAILBOX);
582 	E1000_WRITE_REG(hw, E1000_EIAC, 1 << E1000_VTIVAR_MISC_MAILBOX);
583 	E1000_WRITE_REG(hw, E1000_EIMS, 1 << E1000_VTIVAR_MISC_MAILBOX);
584 	E1000_WRITE_FLUSH(hw);
585 }
586 
587 /* only for mailbox now. If RX/TX needed, should extend this function.  */
588 static void
589 igbvf_set_ivar_map(struct e1000_hw *hw, uint8_t msix_vector)
590 {
591 	uint32_t tmp = 0;
592 
593 	/* mailbox */
594 	tmp |= (msix_vector & E1000_VTIVAR_MISC_INTR_MASK);
595 	tmp |= E1000_VTIVAR_VALID;
596 	E1000_WRITE_REG(hw, E1000_VTIVAR_MISC, tmp);
597 }
598 
599 static void
600 eth_igbvf_configure_msix_intr(struct rte_eth_dev *dev)
601 {
602 	struct e1000_hw *hw =
603 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
604 
605 	/* Configure VF other cause ivar */
606 	igbvf_set_ivar_map(hw, E1000_VTIVAR_MISC_MAILBOX);
607 }
608 
609 static inline int32_t
610 igb_pf_reset_hw(struct e1000_hw *hw)
611 {
612 	uint32_t ctrl_ext;
613 	int32_t status;
614 
615 	status = e1000_reset_hw(hw);
616 
617 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
618 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
619 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
620 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
621 	E1000_WRITE_FLUSH(hw);
622 
623 	return status;
624 }
625 
626 static void
627 igb_identify_hardware(struct rte_eth_dev *dev, struct rte_pci_device *pci_dev)
628 {
629 	struct e1000_hw *hw =
630 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
631 
632 
633 	hw->vendor_id = pci_dev->id.vendor_id;
634 	hw->device_id = pci_dev->id.device_id;
635 	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
636 	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
637 
638 	e1000_set_mac_type(hw);
639 
640 	/* need to check if it is a vf device below */
641 }
642 
643 static int
644 igb_reset_swfw_lock(struct e1000_hw *hw)
645 {
646 	int ret_val;
647 
648 	/*
649 	 * Do mac ops initialization manually here, since we will need
650 	 * some function pointers set by this call.
651 	 */
652 	ret_val = e1000_init_mac_params(hw);
653 	if (ret_val)
654 		return ret_val;
655 
656 	/*
657 	 * SMBI lock should not fail in this early stage. If this is the case,
658 	 * it is due to an improper exit of the application.
659 	 * So force the release of the faulty lock.
660 	 */
661 	if (e1000_get_hw_semaphore_generic(hw) < 0) {
662 		PMD_DRV_LOG(DEBUG, "SMBI lock released");
663 	}
664 	e1000_put_hw_semaphore_generic(hw);
665 
666 	if (hw->mac.ops.acquire_swfw_sync != NULL) {
667 		uint16_t mask;
668 
669 		/*
670 		 * Phy lock should not fail in this early stage. If this is the case,
671 		 * it is due to an improper exit of the application.
672 		 * So force the release of the faulty lock.
673 		 */
674 		mask = E1000_SWFW_PHY0_SM << hw->bus.func;
675 		if (hw->bus.func > E1000_FUNC_1)
676 			mask <<= 2;
677 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
678 			PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released",
679 				    hw->bus.func);
680 		}
681 		hw->mac.ops.release_swfw_sync(hw, mask);
682 
683 		/*
684 		 * This one is more tricky since it is common to all ports; but
685 		 * swfw_sync retries last long enough (1s) to be almost sure that if
686 		 * lock can not be taken it is due to an improper lock of the
687 		 * semaphore.
688 		 */
689 		mask = E1000_SWFW_EEP_SM;
690 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
691 			PMD_DRV_LOG(DEBUG, "SWFW common locks released");
692 		}
693 		hw->mac.ops.release_swfw_sync(hw, mask);
694 	}
695 
696 	return E1000_SUCCESS;
697 }
698 
699 /* Remove all ntuple filters of the device */
700 static int igb_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
701 {
702 	struct e1000_filter_info *filter_info =
703 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
704 	struct e1000_5tuple_filter *p_5tuple;
705 	struct e1000_2tuple_filter *p_2tuple;
706 
707 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
708 		TAILQ_REMOVE(&filter_info->fivetuple_list,
709 			p_5tuple, entries);
710 			rte_free(p_5tuple);
711 	}
712 	filter_info->fivetuple_mask = 0;
713 	while ((p_2tuple = TAILQ_FIRST(&filter_info->twotuple_list))) {
714 		TAILQ_REMOVE(&filter_info->twotuple_list,
715 			p_2tuple, entries);
716 			rte_free(p_2tuple);
717 	}
718 	filter_info->twotuple_mask = 0;
719 
720 	return 0;
721 }
722 
723 /* Remove all flex filters of the device */
724 static int igb_flex_filter_uninit(struct rte_eth_dev *eth_dev)
725 {
726 	struct e1000_filter_info *filter_info =
727 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
728 	struct e1000_flex_filter *p_flex;
729 
730 	while ((p_flex = TAILQ_FIRST(&filter_info->flex_list))) {
731 		TAILQ_REMOVE(&filter_info->flex_list, p_flex, entries);
732 		rte_free(p_flex);
733 	}
734 	filter_info->flex_mask = 0;
735 
736 	return 0;
737 }
738 
739 static int
740 eth_igb_dev_init(struct rte_eth_dev *eth_dev)
741 {
742 	int error = 0;
743 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
744 	struct e1000_hw *hw =
745 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
746 	struct e1000_vfta * shadow_vfta =
747 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
748 	struct e1000_filter_info *filter_info =
749 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
750 	struct e1000_adapter *adapter =
751 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
752 
753 	uint32_t ctrl_ext;
754 
755 	eth_dev->dev_ops = &eth_igb_ops;
756 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
757 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
758 	eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
759 
760 	/* for secondary processes, we don't initialise any further as primary
761 	 * has already done this work. Only check we don't need a different
762 	 * RX function */
763 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
764 		if (eth_dev->data->scattered_rx)
765 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
766 		return 0;
767 	}
768 
769 	rte_eth_copy_pci_info(eth_dev, pci_dev);
770 
771 	hw->hw_addr= (void *)pci_dev->mem_resource[0].addr;
772 
773 	igb_identify_hardware(eth_dev, pci_dev);
774 	if (e1000_setup_init_funcs(hw, FALSE) != E1000_SUCCESS) {
775 		error = -EIO;
776 		goto err_late;
777 	}
778 
779 	e1000_get_bus_info(hw);
780 
781 	/* Reset any pending lock */
782 	if (igb_reset_swfw_lock(hw) != E1000_SUCCESS) {
783 		error = -EIO;
784 		goto err_late;
785 	}
786 
787 	/* Finish initialization */
788 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS) {
789 		error = -EIO;
790 		goto err_late;
791 	}
792 
793 	hw->mac.autoneg = 1;
794 	hw->phy.autoneg_wait_to_complete = 0;
795 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
796 
797 	/* Copper options */
798 	if (hw->phy.media_type == e1000_media_type_copper) {
799 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
800 		hw->phy.disable_polarity_correction = 0;
801 		hw->phy.ms_type = e1000_ms_hw_default;
802 	}
803 
804 	/*
805 	 * Start from a known state, this is important in reading the nvm
806 	 * and mac from that.
807 	 */
808 	igb_pf_reset_hw(hw);
809 
810 	/* Make sure we have a good EEPROM before we read from it */
811 	if (e1000_validate_nvm_checksum(hw) < 0) {
812 		/*
813 		 * Some PCI-E parts fail the first check due to
814 		 * the link being in sleep state, call it again,
815 		 * if it fails a second time its a real issue.
816 		 */
817 		if (e1000_validate_nvm_checksum(hw) < 0) {
818 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
819 			error = -EIO;
820 			goto err_late;
821 		}
822 	}
823 
824 	/* Read the permanent MAC address out of the EEPROM */
825 	if (e1000_read_mac_addr(hw) != 0) {
826 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
827 		error = -EIO;
828 		goto err_late;
829 	}
830 
831 	/* Allocate memory for storing MAC addresses */
832 	eth_dev->data->mac_addrs = rte_zmalloc("e1000",
833 		RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count, 0);
834 	if (eth_dev->data->mac_addrs == NULL) {
835 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
836 						"store MAC addresses",
837 				RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
838 		error = -ENOMEM;
839 		goto err_late;
840 	}
841 
842 	/* Copy the permanent MAC address */
843 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
844 			&eth_dev->data->mac_addrs[0]);
845 
846 	/* Pass the information to the rte_eth_dev_close() that it should also
847 	 * release the private port resources.
848 	 */
849 	eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
850 
851 	/* initialize the vfta */
852 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
853 
854 	/* Now initialize the hardware */
855 	if (igb_hardware_init(hw) != 0) {
856 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
857 		rte_free(eth_dev->data->mac_addrs);
858 		eth_dev->data->mac_addrs = NULL;
859 		error = -ENODEV;
860 		goto err_late;
861 	}
862 	hw->mac.get_link_status = 1;
863 	adapter->stopped = 0;
864 
865 	/* Indicate SOL/IDER usage */
866 	if (e1000_check_reset_block(hw) < 0) {
867 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to"
868 					"SOL/IDER session");
869 	}
870 
871 	/* initialize PF if max_vfs not zero */
872 	igb_pf_host_init(eth_dev);
873 
874 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
875 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
876 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
877 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
878 	E1000_WRITE_FLUSH(hw);
879 
880 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
881 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
882 		     pci_dev->id.device_id);
883 
884 	rte_intr_callback_register(&pci_dev->intr_handle,
885 				   eth_igb_interrupt_handler,
886 				   (void *)eth_dev);
887 
888 	/* enable uio/vfio intr/eventfd mapping */
889 	rte_intr_enable(&pci_dev->intr_handle);
890 
891 	/* enable support intr */
892 	igb_intr_enable(eth_dev);
893 
894 	eth_igb_dev_set_link_down(eth_dev);
895 
896 	/* initialize filter info */
897 	memset(filter_info, 0,
898 	       sizeof(struct e1000_filter_info));
899 
900 	TAILQ_INIT(&filter_info->flex_list);
901 	TAILQ_INIT(&filter_info->twotuple_list);
902 	TAILQ_INIT(&filter_info->fivetuple_list);
903 
904 	TAILQ_INIT(&igb_filter_ntuple_list);
905 	TAILQ_INIT(&igb_filter_ethertype_list);
906 	TAILQ_INIT(&igb_filter_syn_list);
907 	TAILQ_INIT(&igb_filter_flex_list);
908 	TAILQ_INIT(&igb_filter_rss_list);
909 	TAILQ_INIT(&igb_flow_list);
910 
911 	return 0;
912 
913 err_late:
914 	igb_hw_control_release(hw);
915 
916 	return error;
917 }
918 
919 static int
920 eth_igb_dev_uninit(struct rte_eth_dev *eth_dev)
921 {
922 	PMD_INIT_FUNC_TRACE();
923 
924 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
925 		return -EPERM;
926 
927 	eth_igb_close(eth_dev);
928 
929 	return 0;
930 }
931 
932 /*
933  * Virtual Function device init
934  */
935 static int
936 eth_igbvf_dev_init(struct rte_eth_dev *eth_dev)
937 {
938 	struct rte_pci_device *pci_dev;
939 	struct rte_intr_handle *intr_handle;
940 	struct e1000_adapter *adapter =
941 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
942 	struct e1000_hw *hw =
943 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
944 	int diag;
945 	struct rte_ether_addr *perm_addr =
946 		(struct rte_ether_addr *)hw->mac.perm_addr;
947 
948 	PMD_INIT_FUNC_TRACE();
949 
950 	eth_dev->dev_ops = &igbvf_eth_dev_ops;
951 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
952 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
953 	eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
954 
955 	/* for secondary processes, we don't initialise any further as primary
956 	 * has already done this work. Only check we don't need a different
957 	 * RX function */
958 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
959 		if (eth_dev->data->scattered_rx)
960 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
961 		return 0;
962 	}
963 
964 	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
965 	rte_eth_copy_pci_info(eth_dev, pci_dev);
966 
967 	hw->device_id = pci_dev->id.device_id;
968 	hw->vendor_id = pci_dev->id.vendor_id;
969 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
970 	adapter->stopped = 0;
971 
972 	/* Initialize the shared code (base driver) */
973 	diag = e1000_setup_init_funcs(hw, TRUE);
974 	if (diag != 0) {
975 		PMD_INIT_LOG(ERR, "Shared code init failed for igbvf: %d",
976 			diag);
977 		return -EIO;
978 	}
979 
980 	/* init_mailbox_params */
981 	hw->mbx.ops.init_params(hw);
982 
983 	/* Disable the interrupts for VF */
984 	igbvf_intr_disable(hw);
985 
986 	diag = hw->mac.ops.reset_hw(hw);
987 
988 	/* Allocate memory for storing MAC addresses */
989 	eth_dev->data->mac_addrs = rte_zmalloc("igbvf", RTE_ETHER_ADDR_LEN *
990 		hw->mac.rar_entry_count, 0);
991 	if (eth_dev->data->mac_addrs == NULL) {
992 		PMD_INIT_LOG(ERR,
993 			"Failed to allocate %d bytes needed to store MAC "
994 			"addresses",
995 			RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
996 		return -ENOMEM;
997 	}
998 
999 	/* Pass the information to the rte_eth_dev_close() that it should also
1000 	 * release the private port resources.
1001 	 */
1002 	eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
1003 
1004 	/* Generate a random MAC address, if none was assigned by PF. */
1005 	if (rte_is_zero_ether_addr(perm_addr)) {
1006 		rte_eth_random_addr(perm_addr->addr_bytes);
1007 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1008 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1009 			     "%02x:%02x:%02x:%02x:%02x:%02x",
1010 			     perm_addr->addr_bytes[0],
1011 			     perm_addr->addr_bytes[1],
1012 			     perm_addr->addr_bytes[2],
1013 			     perm_addr->addr_bytes[3],
1014 			     perm_addr->addr_bytes[4],
1015 			     perm_addr->addr_bytes[5]);
1016 	}
1017 
1018 	diag = e1000_rar_set(hw, perm_addr->addr_bytes, 0);
1019 	if (diag) {
1020 		rte_free(eth_dev->data->mac_addrs);
1021 		eth_dev->data->mac_addrs = NULL;
1022 		return diag;
1023 	}
1024 	/* Copy the permanent MAC address */
1025 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.perm_addr,
1026 			&eth_dev->data->mac_addrs[0]);
1027 
1028 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x "
1029 		     "mac.type=%s",
1030 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1031 		     pci_dev->id.device_id, "igb_mac_82576_vf");
1032 
1033 	intr_handle = &pci_dev->intr_handle;
1034 	rte_intr_callback_register(intr_handle,
1035 				   eth_igbvf_interrupt_handler, eth_dev);
1036 
1037 	return 0;
1038 }
1039 
1040 static int
1041 eth_igbvf_dev_uninit(struct rte_eth_dev *eth_dev)
1042 {
1043 	PMD_INIT_FUNC_TRACE();
1044 
1045 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1046 		return -EPERM;
1047 
1048 	igbvf_dev_close(eth_dev);
1049 
1050 	return 0;
1051 }
1052 
1053 static int eth_igb_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1054 	struct rte_pci_device *pci_dev)
1055 {
1056 	return rte_eth_dev_pci_generic_probe(pci_dev,
1057 		sizeof(struct e1000_adapter), eth_igb_dev_init);
1058 }
1059 
1060 static int eth_igb_pci_remove(struct rte_pci_device *pci_dev)
1061 {
1062 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igb_dev_uninit);
1063 }
1064 
1065 static struct rte_pci_driver rte_igb_pmd = {
1066 	.id_table = pci_id_igb_map,
1067 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1068 	.probe = eth_igb_pci_probe,
1069 	.remove = eth_igb_pci_remove,
1070 };
1071 
1072 
1073 static int eth_igbvf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1074 	struct rte_pci_device *pci_dev)
1075 {
1076 	return rte_eth_dev_pci_generic_probe(pci_dev,
1077 		sizeof(struct e1000_adapter), eth_igbvf_dev_init);
1078 }
1079 
1080 static int eth_igbvf_pci_remove(struct rte_pci_device *pci_dev)
1081 {
1082 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igbvf_dev_uninit);
1083 }
1084 
1085 /*
1086  * virtual function driver struct
1087  */
1088 static struct rte_pci_driver rte_igbvf_pmd = {
1089 	.id_table = pci_id_igbvf_map,
1090 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1091 	.probe = eth_igbvf_pci_probe,
1092 	.remove = eth_igbvf_pci_remove,
1093 };
1094 
1095 static void
1096 igb_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1097 {
1098 	struct e1000_hw *hw =
1099 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1100 	/* RCTL: enable VLAN filter since VMDq always use VLAN filter */
1101 	uint32_t rctl = E1000_READ_REG(hw, E1000_RCTL);
1102 	rctl |= E1000_RCTL_VFE;
1103 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1104 }
1105 
1106 static int
1107 igb_check_mq_mode(struct rte_eth_dev *dev)
1108 {
1109 	enum rte_eth_rx_mq_mode rx_mq_mode = dev->data->dev_conf.rxmode.mq_mode;
1110 	enum rte_eth_tx_mq_mode tx_mq_mode = dev->data->dev_conf.txmode.mq_mode;
1111 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
1112 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
1113 
1114 	if ((rx_mq_mode & ETH_MQ_RX_DCB_FLAG) ||
1115 	    tx_mq_mode == ETH_MQ_TX_DCB ||
1116 	    tx_mq_mode == ETH_MQ_TX_VMDQ_DCB) {
1117 		PMD_INIT_LOG(ERR, "DCB mode is not supported.");
1118 		return -EINVAL;
1119 	}
1120 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
1121 		/* Check multi-queue mode.
1122 		 * To no break software we accept ETH_MQ_RX_NONE as this might
1123 		 * be used to turn off VLAN filter.
1124 		 */
1125 
1126 		if (rx_mq_mode == ETH_MQ_RX_NONE ||
1127 		    rx_mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1128 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
1129 			RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = 1;
1130 		} else {
1131 			/* Only support one queue on VFs.
1132 			 * RSS together with SRIOV is not supported.
1133 			 */
1134 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1135 					" wrong mq_mode rx %d.",
1136 					rx_mq_mode);
1137 			return -EINVAL;
1138 		}
1139 		/* TX mode is not used here, so mode might be ignored.*/
1140 		if (tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1141 			/* SRIOV only works in VMDq enable mode */
1142 			PMD_INIT_LOG(WARNING, "SRIOV is active,"
1143 					" TX mode %d is not supported. "
1144 					" Driver will behave as %d mode.",
1145 					tx_mq_mode, ETH_MQ_TX_VMDQ_ONLY);
1146 		}
1147 
1148 		/* check valid queue number */
1149 		if ((nb_rx_q > 1) || (nb_tx_q > 1)) {
1150 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1151 					" only support one queue on VFs.");
1152 			return -EINVAL;
1153 		}
1154 	} else {
1155 		/* To no break software that set invalid mode, only display
1156 		 * warning if invalid mode is used.
1157 		 */
1158 		if (rx_mq_mode != ETH_MQ_RX_NONE &&
1159 		    rx_mq_mode != ETH_MQ_RX_VMDQ_ONLY &&
1160 		    rx_mq_mode != ETH_MQ_RX_RSS) {
1161 			/* RSS together with VMDq not supported*/
1162 			PMD_INIT_LOG(ERR, "RX mode %d is not supported.",
1163 				     rx_mq_mode);
1164 			return -EINVAL;
1165 		}
1166 
1167 		if (tx_mq_mode != ETH_MQ_TX_NONE &&
1168 		    tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1169 			PMD_INIT_LOG(WARNING, "TX mode %d is not supported."
1170 					" Due to txmode is meaningless in this"
1171 					" driver, just ignore.",
1172 					tx_mq_mode);
1173 		}
1174 	}
1175 	return 0;
1176 }
1177 
1178 static int
1179 eth_igb_configure(struct rte_eth_dev *dev)
1180 {
1181 	struct e1000_interrupt *intr =
1182 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1183 	int ret;
1184 
1185 	PMD_INIT_FUNC_TRACE();
1186 
1187 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1188 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1189 
1190 	/* multipe queue mode checking */
1191 	ret  = igb_check_mq_mode(dev);
1192 	if (ret != 0) {
1193 		PMD_DRV_LOG(ERR, "igb_check_mq_mode fails with %d.",
1194 			    ret);
1195 		return ret;
1196 	}
1197 
1198 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1199 	PMD_INIT_FUNC_TRACE();
1200 
1201 	return 0;
1202 }
1203 
1204 static void
1205 eth_igb_rxtx_control(struct rte_eth_dev *dev,
1206 		     bool enable)
1207 {
1208 	struct e1000_hw *hw =
1209 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1210 	uint32_t tctl, rctl;
1211 
1212 	tctl = E1000_READ_REG(hw, E1000_TCTL);
1213 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1214 
1215 	if (enable) {
1216 		/* enable Tx/Rx */
1217 		tctl |= E1000_TCTL_EN;
1218 		rctl |= E1000_RCTL_EN;
1219 	} else {
1220 		/* disable Tx/Rx */
1221 		tctl &= ~E1000_TCTL_EN;
1222 		rctl &= ~E1000_RCTL_EN;
1223 	}
1224 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
1225 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1226 	E1000_WRITE_FLUSH(hw);
1227 }
1228 
1229 static int
1230 eth_igb_start(struct rte_eth_dev *dev)
1231 {
1232 	struct e1000_hw *hw =
1233 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1234 	struct e1000_adapter *adapter =
1235 		E1000_DEV_PRIVATE(dev->data->dev_private);
1236 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1237 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1238 	int ret, mask;
1239 	uint32_t intr_vector = 0;
1240 	uint32_t ctrl_ext;
1241 	uint32_t *speeds;
1242 	int num_speeds;
1243 	bool autoneg;
1244 
1245 	PMD_INIT_FUNC_TRACE();
1246 
1247 	/* disable uio/vfio intr/eventfd mapping */
1248 	rte_intr_disable(intr_handle);
1249 
1250 	/* Power up the phy. Needed to make the link go Up */
1251 	eth_igb_dev_set_link_up(dev);
1252 
1253 	/*
1254 	 * Packet Buffer Allocation (PBA)
1255 	 * Writing PBA sets the receive portion of the buffer
1256 	 * the remainder is used for the transmit buffer.
1257 	 */
1258 	if (hw->mac.type == e1000_82575) {
1259 		uint32_t pba;
1260 
1261 		pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
1262 		E1000_WRITE_REG(hw, E1000_PBA, pba);
1263 	}
1264 
1265 	/* Put the address into the Receive Address Array */
1266 	e1000_rar_set(hw, hw->mac.addr, 0);
1267 
1268 	/* Initialize the hardware */
1269 	if (igb_hardware_init(hw)) {
1270 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
1271 		return -EIO;
1272 	}
1273 	adapter->stopped = 0;
1274 
1275 	E1000_WRITE_REG(hw, E1000_VET,
1276 			RTE_ETHER_TYPE_VLAN << 16 | RTE_ETHER_TYPE_VLAN);
1277 
1278 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1279 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1280 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
1281 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
1282 	E1000_WRITE_FLUSH(hw);
1283 
1284 	/* configure PF module if SRIOV enabled */
1285 	igb_pf_host_configure(dev);
1286 
1287 	/* check and configure queue intr-vector mapping */
1288 	if ((rte_intr_cap_multiple(intr_handle) ||
1289 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
1290 	    dev->data->dev_conf.intr_conf.rxq != 0) {
1291 		intr_vector = dev->data->nb_rx_queues;
1292 		if (rte_intr_efd_enable(intr_handle, intr_vector))
1293 			return -1;
1294 	}
1295 
1296 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
1297 		intr_handle->intr_vec =
1298 			rte_zmalloc("intr_vec",
1299 				    dev->data->nb_rx_queues * sizeof(int), 0);
1300 		if (intr_handle->intr_vec == NULL) {
1301 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
1302 				     " intr_vec", dev->data->nb_rx_queues);
1303 			return -ENOMEM;
1304 		}
1305 	}
1306 
1307 	/* confiugre msix for rx interrupt */
1308 	eth_igb_configure_msix_intr(dev);
1309 
1310 	/* Configure for OS presence */
1311 	igb_init_manageability(hw);
1312 
1313 	eth_igb_tx_init(dev);
1314 
1315 	/* This can fail when allocating mbufs for descriptor rings */
1316 	ret = eth_igb_rx_init(dev);
1317 	if (ret) {
1318 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
1319 		igb_dev_clear_queues(dev);
1320 		return ret;
1321 	}
1322 
1323 	e1000_clear_hw_cntrs_base_generic(hw);
1324 
1325 	/*
1326 	 * VLAN Offload Settings
1327 	 */
1328 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
1329 			ETH_VLAN_EXTEND_MASK;
1330 	ret = eth_igb_vlan_offload_set(dev, mask);
1331 	if (ret) {
1332 		PMD_INIT_LOG(ERR, "Unable to set vlan offload");
1333 		igb_dev_clear_queues(dev);
1334 		return ret;
1335 	}
1336 
1337 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1338 		/* Enable VLAN filter since VMDq always use VLAN filter */
1339 		igb_vmdq_vlan_hw_filter_enable(dev);
1340 	}
1341 
1342 	if ((hw->mac.type == e1000_82576) || (hw->mac.type == e1000_82580) ||
1343 		(hw->mac.type == e1000_i350) || (hw->mac.type == e1000_i210) ||
1344 		(hw->mac.type == e1000_i211)) {
1345 		/* Configure EITR with the maximum possible value (0xFFFF) */
1346 		E1000_WRITE_REG(hw, E1000_EITR(0), 0xFFFF);
1347 	}
1348 
1349 	/* Setup link speed and duplex */
1350 	speeds = &dev->data->dev_conf.link_speeds;
1351 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
1352 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
1353 		hw->mac.autoneg = 1;
1354 	} else {
1355 		num_speeds = 0;
1356 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
1357 
1358 		/* Reset */
1359 		hw->phy.autoneg_advertised = 0;
1360 
1361 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1362 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1363 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
1364 			num_speeds = -1;
1365 			goto error_invalid_config;
1366 		}
1367 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
1368 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1369 			num_speeds++;
1370 		}
1371 		if (*speeds & ETH_LINK_SPEED_10M) {
1372 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1373 			num_speeds++;
1374 		}
1375 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
1376 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1377 			num_speeds++;
1378 		}
1379 		if (*speeds & ETH_LINK_SPEED_100M) {
1380 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1381 			num_speeds++;
1382 		}
1383 		if (*speeds & ETH_LINK_SPEED_1G) {
1384 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1385 			num_speeds++;
1386 		}
1387 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
1388 			goto error_invalid_config;
1389 
1390 		/* Set/reset the mac.autoneg based on the link speed,
1391 		 * fixed or not
1392 		 */
1393 		if (!autoneg) {
1394 			hw->mac.autoneg = 0;
1395 			hw->mac.forced_speed_duplex =
1396 					hw->phy.autoneg_advertised;
1397 		} else {
1398 			hw->mac.autoneg = 1;
1399 		}
1400 	}
1401 
1402 	e1000_setup_link(hw);
1403 
1404 	if (rte_intr_allow_others(intr_handle)) {
1405 		/* check if lsc interrupt is enabled */
1406 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1407 			eth_igb_lsc_interrupt_setup(dev, TRUE);
1408 		else
1409 			eth_igb_lsc_interrupt_setup(dev, FALSE);
1410 	} else {
1411 		rte_intr_callback_unregister(intr_handle,
1412 					     eth_igb_interrupt_handler,
1413 					     (void *)dev);
1414 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1415 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
1416 				     " no intr multiplex");
1417 	}
1418 
1419 	/* check if rxq interrupt is enabled */
1420 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
1421 	    rte_intr_dp_is_en(intr_handle))
1422 		eth_igb_rxq_interrupt_setup(dev);
1423 
1424 	/* enable uio/vfio intr/eventfd mapping */
1425 	rte_intr_enable(intr_handle);
1426 
1427 	/* resume enabled intr since hw reset */
1428 	igb_intr_enable(dev);
1429 
1430 	/* restore all types filter */
1431 	igb_filter_restore(dev);
1432 
1433 	eth_igb_rxtx_control(dev, true);
1434 	eth_igb_link_update(dev, 0);
1435 
1436 	PMD_INIT_LOG(DEBUG, "<<");
1437 
1438 	return 0;
1439 
1440 error_invalid_config:
1441 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
1442 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
1443 	igb_dev_clear_queues(dev);
1444 	return -EINVAL;
1445 }
1446 
1447 /*********************************************************************
1448  *
1449  *  This routine disables all traffic on the adapter by issuing a
1450  *  global reset on the MAC.
1451  *
1452  **********************************************************************/
1453 static void
1454 eth_igb_stop(struct rte_eth_dev *dev)
1455 {
1456 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1457 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1458 	struct rte_eth_link link;
1459 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1460 	struct e1000_adapter *adapter =
1461 		E1000_DEV_PRIVATE(dev->data->dev_private);
1462 
1463 	if (adapter->stopped)
1464 		return;
1465 
1466 	eth_igb_rxtx_control(dev, false);
1467 
1468 	igb_intr_disable(dev);
1469 
1470 	/* disable intr eventfd mapping */
1471 	rte_intr_disable(intr_handle);
1472 
1473 	igb_pf_reset_hw(hw);
1474 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1475 
1476 	/* Set bit for Go Link disconnect if PHY reset is not blocked */
1477 	if (hw->mac.type >= e1000_82580 &&
1478 	    (e1000_check_reset_block(hw) != E1000_BLK_PHY_RESET)) {
1479 		uint32_t phpm_reg;
1480 
1481 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1482 		phpm_reg |= E1000_82580_PM_GO_LINKD;
1483 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1484 	}
1485 
1486 	/* Power down the phy. Needed to make the link go Down */
1487 	eth_igb_dev_set_link_down(dev);
1488 
1489 	igb_dev_clear_queues(dev);
1490 
1491 	/* clear the recorded link status */
1492 	memset(&link, 0, sizeof(link));
1493 	rte_eth_linkstatus_set(dev, &link);
1494 
1495 	if (!rte_intr_allow_others(intr_handle))
1496 		/* resume to the default handler */
1497 		rte_intr_callback_register(intr_handle,
1498 					   eth_igb_interrupt_handler,
1499 					   (void *)dev);
1500 
1501 	/* Clean datapath event and queue/vec mapping */
1502 	rte_intr_efd_disable(intr_handle);
1503 	if (intr_handle->intr_vec != NULL) {
1504 		rte_free(intr_handle->intr_vec);
1505 		intr_handle->intr_vec = NULL;
1506 	}
1507 
1508 	adapter->stopped = true;
1509 }
1510 
1511 static int
1512 eth_igb_dev_set_link_up(struct rte_eth_dev *dev)
1513 {
1514 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1515 
1516 	if (hw->phy.media_type == e1000_media_type_copper)
1517 		e1000_power_up_phy(hw);
1518 	else
1519 		e1000_power_up_fiber_serdes_link(hw);
1520 
1521 	return 0;
1522 }
1523 
1524 static int
1525 eth_igb_dev_set_link_down(struct rte_eth_dev *dev)
1526 {
1527 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1528 
1529 	if (hw->phy.media_type == e1000_media_type_copper)
1530 		e1000_power_down_phy(hw);
1531 	else
1532 		e1000_shutdown_fiber_serdes_link(hw);
1533 
1534 	return 0;
1535 }
1536 
1537 static void
1538 eth_igb_close(struct rte_eth_dev *dev)
1539 {
1540 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1541 	struct rte_eth_link link;
1542 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1543 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1544 	struct e1000_filter_info *filter_info =
1545 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
1546 
1547 	eth_igb_stop(dev);
1548 
1549 	e1000_phy_hw_reset(hw);
1550 	igb_release_manageability(hw);
1551 	igb_hw_control_release(hw);
1552 
1553 	/* Clear bit for Go Link disconnect if PHY reset is not blocked */
1554 	if (hw->mac.type >= e1000_82580 &&
1555 	    (e1000_check_reset_block(hw) != E1000_BLK_PHY_RESET)) {
1556 		uint32_t phpm_reg;
1557 
1558 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1559 		phpm_reg &= ~E1000_82580_PM_GO_LINKD;
1560 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1561 	}
1562 
1563 	igb_dev_free_queues(dev);
1564 
1565 	if (intr_handle->intr_vec) {
1566 		rte_free(intr_handle->intr_vec);
1567 		intr_handle->intr_vec = NULL;
1568 	}
1569 
1570 	memset(&link, 0, sizeof(link));
1571 	rte_eth_linkstatus_set(dev, &link);
1572 
1573 	dev->dev_ops = NULL;
1574 	dev->rx_pkt_burst = NULL;
1575 	dev->tx_pkt_burst = NULL;
1576 
1577 	/* Reset any pending lock */
1578 	igb_reset_swfw_lock(hw);
1579 
1580 	/* uninitialize PF if max_vfs not zero */
1581 	igb_pf_host_uninit(dev);
1582 
1583 	rte_intr_callback_unregister(intr_handle,
1584 				     eth_igb_interrupt_handler, dev);
1585 
1586 	/* clear the SYN filter info */
1587 	filter_info->syn_info = 0;
1588 
1589 	/* clear the ethertype filters info */
1590 	filter_info->ethertype_mask = 0;
1591 	memset(filter_info->ethertype_filters, 0,
1592 		E1000_MAX_ETQF_FILTERS * sizeof(struct igb_ethertype_filter));
1593 
1594 	/* clear the rss filter info */
1595 	memset(&filter_info->rss_info, 0,
1596 		sizeof(struct igb_rte_flow_rss_conf));
1597 
1598 	/* remove all ntuple filters of the device */
1599 	igb_ntuple_filter_uninit(dev);
1600 
1601 	/* remove all flex filters of the device */
1602 	igb_flex_filter_uninit(dev);
1603 
1604 	/* clear all the filters list */
1605 	igb_filterlist_flush(dev);
1606 }
1607 
1608 /*
1609  * Reset PF device.
1610  */
1611 static int
1612 eth_igb_reset(struct rte_eth_dev *dev)
1613 {
1614 	int ret;
1615 
1616 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
1617 	 * its VF to make them align with it. The detailed notification
1618 	 * mechanism is PMD specific and is currently not implemented.
1619 	 * To avoid unexpected behavior in VF, currently reset of PF with
1620 	 * SR-IOV activation is not supported. It might be supported later.
1621 	 */
1622 	if (dev->data->sriov.active)
1623 		return -ENOTSUP;
1624 
1625 	ret = eth_igb_dev_uninit(dev);
1626 	if (ret)
1627 		return ret;
1628 
1629 	ret = eth_igb_dev_init(dev);
1630 
1631 	return ret;
1632 }
1633 
1634 
1635 static int
1636 igb_get_rx_buffer_size(struct e1000_hw *hw)
1637 {
1638 	uint32_t rx_buf_size;
1639 	if (hw->mac.type == e1000_82576) {
1640 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xffff) << 10;
1641 	} else if (hw->mac.type == e1000_82580 || hw->mac.type == e1000_i350) {
1642 		/* PBS needs to be translated according to a lookup table */
1643 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xf);
1644 		rx_buf_size = (uint32_t) e1000_rxpbs_adjust_82580(rx_buf_size);
1645 		rx_buf_size = (rx_buf_size << 10);
1646 	} else if (hw->mac.type == e1000_i210 || hw->mac.type == e1000_i211) {
1647 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0x3f) << 10;
1648 	} else {
1649 		rx_buf_size = (E1000_READ_REG(hw, E1000_PBA) & 0xffff) << 10;
1650 	}
1651 
1652 	return rx_buf_size;
1653 }
1654 
1655 /*********************************************************************
1656  *
1657  *  Initialize the hardware
1658  *
1659  **********************************************************************/
1660 static int
1661 igb_hardware_init(struct e1000_hw *hw)
1662 {
1663 	uint32_t rx_buf_size;
1664 	int diag;
1665 
1666 	/* Let the firmware know the OS is in control */
1667 	igb_hw_control_acquire(hw);
1668 
1669 	/*
1670 	 * These parameters control the automatic generation (Tx) and
1671 	 * response (Rx) to Ethernet PAUSE frames.
1672 	 * - High water mark should allow for at least two standard size (1518)
1673 	 *   frames to be received after sending an XOFF.
1674 	 * - Low water mark works best when it is very near the high water mark.
1675 	 *   This allows the receiver to restart by sending XON when it has
1676 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
1677 	 *   restart after one full frame is pulled from the buffer. There
1678 	 *   could be several smaller frames in the buffer and if so they will
1679 	 *   not trigger the XON until their total number reduces the buffer
1680 	 *   by 1500.
1681 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
1682 	 */
1683 	rx_buf_size = igb_get_rx_buffer_size(hw);
1684 
1685 	hw->fc.high_water = rx_buf_size - (RTE_ETHER_MAX_LEN * 2);
1686 	hw->fc.low_water = hw->fc.high_water - 1500;
1687 	hw->fc.pause_time = IGB_FC_PAUSE_TIME;
1688 	hw->fc.send_xon = 1;
1689 
1690 	/* Set Flow control, use the tunable location if sane */
1691 	if ((igb_fc_setting != e1000_fc_none) && (igb_fc_setting < 4))
1692 		hw->fc.requested_mode = igb_fc_setting;
1693 	else
1694 		hw->fc.requested_mode = e1000_fc_none;
1695 
1696 	/* Issue a global reset */
1697 	igb_pf_reset_hw(hw);
1698 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1699 
1700 	diag = e1000_init_hw(hw);
1701 	if (diag < 0)
1702 		return diag;
1703 
1704 	E1000_WRITE_REG(hw, E1000_VET,
1705 			RTE_ETHER_TYPE_VLAN << 16 | RTE_ETHER_TYPE_VLAN);
1706 	e1000_get_phy_info(hw);
1707 	e1000_check_for_link(hw);
1708 
1709 	return 0;
1710 }
1711 
1712 /* This function is based on igb_update_stats_counters() in igb/if_igb.c */
1713 static void
1714 igb_read_stats_registers(struct e1000_hw *hw, struct e1000_hw_stats *stats)
1715 {
1716 	int pause_frames;
1717 
1718 	uint64_t old_gprc  = stats->gprc;
1719 	uint64_t old_gptc  = stats->gptc;
1720 	uint64_t old_tpr   = stats->tpr;
1721 	uint64_t old_tpt   = stats->tpt;
1722 	uint64_t old_rpthc = stats->rpthc;
1723 	uint64_t old_hgptc = stats->hgptc;
1724 
1725 	if(hw->phy.media_type == e1000_media_type_copper ||
1726 	    (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
1727 		stats->symerrs +=
1728 		    E1000_READ_REG(hw,E1000_SYMERRS);
1729 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
1730 	}
1731 
1732 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
1733 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
1734 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
1735 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
1736 
1737 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
1738 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
1739 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
1740 	stats->dc += E1000_READ_REG(hw, E1000_DC);
1741 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
1742 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
1743 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
1744 	/*
1745 	** For watchdog management we need to know if we have been
1746 	** paused during the last interval, so capture that here.
1747 	*/
1748 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
1749 	stats->xoffrxc += pause_frames;
1750 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
1751 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
1752 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
1753 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
1754 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
1755 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
1756 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
1757 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
1758 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
1759 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
1760 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
1761 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
1762 
1763 	/* For the 64-bit byte counters the low dword must be read first. */
1764 	/* Both registers clear on the read of the high dword */
1765 
1766 	/* Workaround CRC bytes included in size, take away 4 bytes/packet */
1767 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
1768 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
1769 	stats->gorc -= (stats->gprc - old_gprc) * RTE_ETHER_CRC_LEN;
1770 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
1771 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
1772 	stats->gotc -= (stats->gptc - old_gptc) * RTE_ETHER_CRC_LEN;
1773 
1774 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
1775 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
1776 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
1777 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
1778 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
1779 
1780 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
1781 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
1782 
1783 	stats->tor += E1000_READ_REG(hw, E1000_TORL);
1784 	stats->tor += ((uint64_t)E1000_READ_REG(hw, E1000_TORH) << 32);
1785 	stats->tor -= (stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
1786 	stats->tot += E1000_READ_REG(hw, E1000_TOTL);
1787 	stats->tot += ((uint64_t)E1000_READ_REG(hw, E1000_TOTH) << 32);
1788 	stats->tot -= (stats->tpt - old_tpt) * RTE_ETHER_CRC_LEN;
1789 
1790 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
1791 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
1792 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
1793 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
1794 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
1795 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
1796 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
1797 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
1798 
1799 	/* Interrupt Counts */
1800 
1801 	stats->iac += E1000_READ_REG(hw, E1000_IAC);
1802 	stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
1803 	stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
1804 	stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
1805 	stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
1806 	stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
1807 	stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
1808 	stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
1809 	stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
1810 
1811 	/* Host to Card Statistics */
1812 
1813 	stats->cbtmpc += E1000_READ_REG(hw, E1000_CBTMPC);
1814 	stats->htdpmc += E1000_READ_REG(hw, E1000_HTDPMC);
1815 	stats->cbrdpc += E1000_READ_REG(hw, E1000_CBRDPC);
1816 	stats->cbrmpc += E1000_READ_REG(hw, E1000_CBRMPC);
1817 	stats->rpthc += E1000_READ_REG(hw, E1000_RPTHC);
1818 	stats->hgptc += E1000_READ_REG(hw, E1000_HGPTC);
1819 	stats->htcbdpc += E1000_READ_REG(hw, E1000_HTCBDPC);
1820 	stats->hgorc += E1000_READ_REG(hw, E1000_HGORCL);
1821 	stats->hgorc += ((uint64_t)E1000_READ_REG(hw, E1000_HGORCH) << 32);
1822 	stats->hgorc -= (stats->rpthc - old_rpthc) * RTE_ETHER_CRC_LEN;
1823 	stats->hgotc += E1000_READ_REG(hw, E1000_HGOTCL);
1824 	stats->hgotc += ((uint64_t)E1000_READ_REG(hw, E1000_HGOTCH) << 32);
1825 	stats->hgotc -= (stats->hgptc - old_hgptc) * RTE_ETHER_CRC_LEN;
1826 	stats->lenerrs += E1000_READ_REG(hw, E1000_LENERRS);
1827 	stats->scvpc += E1000_READ_REG(hw, E1000_SCVPC);
1828 	stats->hrmpc += E1000_READ_REG(hw, E1000_HRMPC);
1829 
1830 	stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
1831 	stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
1832 	stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
1833 	stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
1834 	stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
1835 	stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
1836 }
1837 
1838 static int
1839 eth_igb_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1840 {
1841 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1842 	struct e1000_hw_stats *stats =
1843 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1844 
1845 	igb_read_stats_registers(hw, stats);
1846 
1847 	if (rte_stats == NULL)
1848 		return -EINVAL;
1849 
1850 	/* Rx Errors */
1851 	rte_stats->imissed = stats->mpc;
1852 	rte_stats->ierrors = stats->crcerrs +
1853 	                     stats->rlec + stats->ruc + stats->roc +
1854 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
1855 
1856 	/* Tx Errors */
1857 	rte_stats->oerrors = stats->ecol + stats->latecol;
1858 
1859 	rte_stats->ipackets = stats->gprc;
1860 	rte_stats->opackets = stats->gptc;
1861 	rte_stats->ibytes   = stats->gorc;
1862 	rte_stats->obytes   = stats->gotc;
1863 	return 0;
1864 }
1865 
1866 static int
1867 eth_igb_stats_reset(struct rte_eth_dev *dev)
1868 {
1869 	struct e1000_hw_stats *hw_stats =
1870 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1871 
1872 	/* HW registers are cleared on read */
1873 	eth_igb_stats_get(dev, NULL);
1874 
1875 	/* Reset software totals */
1876 	memset(hw_stats, 0, sizeof(*hw_stats));
1877 
1878 	return 0;
1879 }
1880 
1881 static int
1882 eth_igb_xstats_reset(struct rte_eth_dev *dev)
1883 {
1884 	struct e1000_hw_stats *stats =
1885 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1886 
1887 	/* HW registers are cleared on read */
1888 	eth_igb_xstats_get(dev, NULL, IGB_NB_XSTATS);
1889 
1890 	/* Reset software totals */
1891 	memset(stats, 0, sizeof(*stats));
1892 
1893 	return 0;
1894 }
1895 
1896 static int eth_igb_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1897 	struct rte_eth_xstat_name *xstats_names,
1898 	__rte_unused unsigned int size)
1899 {
1900 	unsigned i;
1901 
1902 	if (xstats_names == NULL)
1903 		return IGB_NB_XSTATS;
1904 
1905 	/* Note: limit checked in rte_eth_xstats_names() */
1906 
1907 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1908 		strlcpy(xstats_names[i].name, rte_igb_stats_strings[i].name,
1909 			sizeof(xstats_names[i].name));
1910 	}
1911 
1912 	return IGB_NB_XSTATS;
1913 }
1914 
1915 static int eth_igb_xstats_get_names_by_id(struct rte_eth_dev *dev,
1916 		struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
1917 		unsigned int limit)
1918 {
1919 	unsigned int i;
1920 
1921 	if (!ids) {
1922 		if (xstats_names == NULL)
1923 			return IGB_NB_XSTATS;
1924 
1925 		for (i = 0; i < IGB_NB_XSTATS; i++)
1926 			strlcpy(xstats_names[i].name,
1927 				rte_igb_stats_strings[i].name,
1928 				sizeof(xstats_names[i].name));
1929 
1930 		return IGB_NB_XSTATS;
1931 
1932 	} else {
1933 		struct rte_eth_xstat_name xstats_names_copy[IGB_NB_XSTATS];
1934 
1935 		eth_igb_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
1936 				IGB_NB_XSTATS);
1937 
1938 		for (i = 0; i < limit; i++) {
1939 			if (ids[i] >= IGB_NB_XSTATS) {
1940 				PMD_INIT_LOG(ERR, "id value isn't valid");
1941 				return -1;
1942 			}
1943 			strcpy(xstats_names[i].name,
1944 					xstats_names_copy[ids[i]].name);
1945 		}
1946 		return limit;
1947 	}
1948 }
1949 
1950 static int
1951 eth_igb_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1952 		   unsigned n)
1953 {
1954 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1955 	struct e1000_hw_stats *hw_stats =
1956 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1957 	unsigned i;
1958 
1959 	if (n < IGB_NB_XSTATS)
1960 		return IGB_NB_XSTATS;
1961 
1962 	igb_read_stats_registers(hw, hw_stats);
1963 
1964 	/* If this is a reset xstats is NULL, and we have cleared the
1965 	 * registers by reading them.
1966 	 */
1967 	if (!xstats)
1968 		return 0;
1969 
1970 	/* Extended stats */
1971 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1972 		xstats[i].id = i;
1973 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1974 			rte_igb_stats_strings[i].offset);
1975 	}
1976 
1977 	return IGB_NB_XSTATS;
1978 }
1979 
1980 static int
1981 eth_igb_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
1982 		uint64_t *values, unsigned int n)
1983 {
1984 	unsigned int i;
1985 
1986 	if (!ids) {
1987 		struct e1000_hw *hw =
1988 			E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1989 		struct e1000_hw_stats *hw_stats =
1990 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1991 
1992 		if (n < IGB_NB_XSTATS)
1993 			return IGB_NB_XSTATS;
1994 
1995 		igb_read_stats_registers(hw, hw_stats);
1996 
1997 		/* If this is a reset xstats is NULL, and we have cleared the
1998 		 * registers by reading them.
1999 		 */
2000 		if (!values)
2001 			return 0;
2002 
2003 		/* Extended stats */
2004 		for (i = 0; i < IGB_NB_XSTATS; i++)
2005 			values[i] = *(uint64_t *)(((char *)hw_stats) +
2006 					rte_igb_stats_strings[i].offset);
2007 
2008 		return IGB_NB_XSTATS;
2009 
2010 	} else {
2011 		uint64_t values_copy[IGB_NB_XSTATS];
2012 
2013 		eth_igb_xstats_get_by_id(dev, NULL, values_copy,
2014 				IGB_NB_XSTATS);
2015 
2016 		for (i = 0; i < n; i++) {
2017 			if (ids[i] >= IGB_NB_XSTATS) {
2018 				PMD_INIT_LOG(ERR, "id value isn't valid");
2019 				return -1;
2020 			}
2021 			values[i] = values_copy[ids[i]];
2022 		}
2023 		return n;
2024 	}
2025 }
2026 
2027 static void
2028 igbvf_read_stats_registers(struct e1000_hw *hw, struct e1000_vf_stats *hw_stats)
2029 {
2030 	/* Good Rx packets, include VF loopback */
2031 	UPDATE_VF_STAT(E1000_VFGPRC,
2032 	    hw_stats->last_gprc, hw_stats->gprc);
2033 
2034 	/* Good Rx octets, include VF loopback */
2035 	UPDATE_VF_STAT(E1000_VFGORC,
2036 	    hw_stats->last_gorc, hw_stats->gorc);
2037 
2038 	/* Good Tx packets, include VF loopback */
2039 	UPDATE_VF_STAT(E1000_VFGPTC,
2040 	    hw_stats->last_gptc, hw_stats->gptc);
2041 
2042 	/* Good Tx octets, include VF loopback */
2043 	UPDATE_VF_STAT(E1000_VFGOTC,
2044 	    hw_stats->last_gotc, hw_stats->gotc);
2045 
2046 	/* Rx Multicst packets */
2047 	UPDATE_VF_STAT(E1000_VFMPRC,
2048 	    hw_stats->last_mprc, hw_stats->mprc);
2049 
2050 	/* Good Rx loopback packets */
2051 	UPDATE_VF_STAT(E1000_VFGPRLBC,
2052 	    hw_stats->last_gprlbc, hw_stats->gprlbc);
2053 
2054 	/* Good Rx loopback octets */
2055 	UPDATE_VF_STAT(E1000_VFGORLBC,
2056 	    hw_stats->last_gorlbc, hw_stats->gorlbc);
2057 
2058 	/* Good Tx loopback packets */
2059 	UPDATE_VF_STAT(E1000_VFGPTLBC,
2060 	    hw_stats->last_gptlbc, hw_stats->gptlbc);
2061 
2062 	/* Good Tx loopback octets */
2063 	UPDATE_VF_STAT(E1000_VFGOTLBC,
2064 	    hw_stats->last_gotlbc, hw_stats->gotlbc);
2065 }
2066 
2067 static int eth_igbvf_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
2068 				     struct rte_eth_xstat_name *xstats_names,
2069 				     __rte_unused unsigned limit)
2070 {
2071 	unsigned i;
2072 
2073 	if (xstats_names != NULL)
2074 		for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2075 			strlcpy(xstats_names[i].name,
2076 				rte_igbvf_stats_strings[i].name,
2077 				sizeof(xstats_names[i].name));
2078 		}
2079 	return IGBVF_NB_XSTATS;
2080 }
2081 
2082 static int
2083 eth_igbvf_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
2084 		     unsigned n)
2085 {
2086 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2087 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2088 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2089 	unsigned i;
2090 
2091 	if (n < IGBVF_NB_XSTATS)
2092 		return IGBVF_NB_XSTATS;
2093 
2094 	igbvf_read_stats_registers(hw, hw_stats);
2095 
2096 	if (!xstats)
2097 		return 0;
2098 
2099 	for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2100 		xstats[i].id = i;
2101 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
2102 			rte_igbvf_stats_strings[i].offset);
2103 	}
2104 
2105 	return IGBVF_NB_XSTATS;
2106 }
2107 
2108 static int
2109 eth_igbvf_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
2110 {
2111 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2112 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2113 			  E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2114 
2115 	igbvf_read_stats_registers(hw, hw_stats);
2116 
2117 	if (rte_stats == NULL)
2118 		return -EINVAL;
2119 
2120 	rte_stats->ipackets = hw_stats->gprc;
2121 	rte_stats->ibytes = hw_stats->gorc;
2122 	rte_stats->opackets = hw_stats->gptc;
2123 	rte_stats->obytes = hw_stats->gotc;
2124 	return 0;
2125 }
2126 
2127 static int
2128 eth_igbvf_stats_reset(struct rte_eth_dev *dev)
2129 {
2130 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats*)
2131 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2132 
2133 	/* Sync HW register to the last stats */
2134 	eth_igbvf_stats_get(dev, NULL);
2135 
2136 	/* reset HW current stats*/
2137 	memset(&hw_stats->gprc, 0, sizeof(*hw_stats) -
2138 	       offsetof(struct e1000_vf_stats, gprc));
2139 
2140 	return 0;
2141 }
2142 
2143 static int
2144 eth_igb_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
2145 		       size_t fw_size)
2146 {
2147 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2148 	struct e1000_fw_version fw;
2149 	int ret;
2150 
2151 	e1000_get_fw_version(hw, &fw);
2152 
2153 	switch (hw->mac.type) {
2154 	case e1000_i210:
2155 	case e1000_i211:
2156 		if (!(e1000_get_flash_presence_i210(hw))) {
2157 			ret = snprintf(fw_version, fw_size,
2158 				 "%2d.%2d-%d",
2159 				 fw.invm_major, fw.invm_minor,
2160 				 fw.invm_img_type);
2161 			break;
2162 		}
2163 		/* fall through */
2164 	default:
2165 		/* if option rom is valid, display its version too */
2166 		if (fw.or_valid) {
2167 			ret = snprintf(fw_version, fw_size,
2168 				 "%d.%d, 0x%08x, %d.%d.%d",
2169 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
2170 				 fw.or_major, fw.or_build, fw.or_patch);
2171 		/* no option rom */
2172 		} else {
2173 			if (fw.etrack_id != 0X0000) {
2174 				ret = snprintf(fw_version, fw_size,
2175 					 "%d.%d, 0x%08x",
2176 					 fw.eep_major, fw.eep_minor,
2177 					 fw.etrack_id);
2178 			} else {
2179 				ret = snprintf(fw_version, fw_size,
2180 					 "%d.%d.%d",
2181 					 fw.eep_major, fw.eep_minor,
2182 					 fw.eep_build);
2183 			}
2184 		}
2185 		break;
2186 	}
2187 
2188 	ret += 1; /* add the size of '\0' */
2189 	if (fw_size < (u32)ret)
2190 		return ret;
2191 	else
2192 		return 0;
2193 }
2194 
2195 static int
2196 eth_igb_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2197 {
2198 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2199 
2200 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2201 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2202 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2203 	dev_info->rx_queue_offload_capa = igb_get_rx_queue_offloads_capa(dev);
2204 	dev_info->rx_offload_capa = igb_get_rx_port_offloads_capa(dev) |
2205 				    dev_info->rx_queue_offload_capa;
2206 	dev_info->tx_queue_offload_capa = igb_get_tx_queue_offloads_capa(dev);
2207 	dev_info->tx_offload_capa = igb_get_tx_port_offloads_capa(dev) |
2208 				    dev_info->tx_queue_offload_capa;
2209 
2210 	switch (hw->mac.type) {
2211 	case e1000_82575:
2212 		dev_info->max_rx_queues = 4;
2213 		dev_info->max_tx_queues = 4;
2214 		dev_info->max_vmdq_pools = 0;
2215 		break;
2216 
2217 	case e1000_82576:
2218 		dev_info->max_rx_queues = 16;
2219 		dev_info->max_tx_queues = 16;
2220 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2221 		dev_info->vmdq_queue_num = 16;
2222 		break;
2223 
2224 	case e1000_82580:
2225 		dev_info->max_rx_queues = 8;
2226 		dev_info->max_tx_queues = 8;
2227 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2228 		dev_info->vmdq_queue_num = 8;
2229 		break;
2230 
2231 	case e1000_i350:
2232 		dev_info->max_rx_queues = 8;
2233 		dev_info->max_tx_queues = 8;
2234 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2235 		dev_info->vmdq_queue_num = 8;
2236 		break;
2237 
2238 	case e1000_i354:
2239 		dev_info->max_rx_queues = 8;
2240 		dev_info->max_tx_queues = 8;
2241 		break;
2242 
2243 	case e1000_i210:
2244 		dev_info->max_rx_queues = 4;
2245 		dev_info->max_tx_queues = 4;
2246 		dev_info->max_vmdq_pools = 0;
2247 		break;
2248 
2249 	case e1000_i211:
2250 		dev_info->max_rx_queues = 2;
2251 		dev_info->max_tx_queues = 2;
2252 		dev_info->max_vmdq_pools = 0;
2253 		break;
2254 
2255 	default:
2256 		/* Should not happen */
2257 		return -EINVAL;
2258 	}
2259 	dev_info->hash_key_size = IGB_HKEY_MAX_INDEX * sizeof(uint32_t);
2260 	dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
2261 	dev_info->flow_type_rss_offloads = IGB_RSS_OFFLOAD_ALL;
2262 
2263 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2264 		.rx_thresh = {
2265 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2266 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2267 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2268 		},
2269 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2270 		.rx_drop_en = 0,
2271 		.offloads = 0,
2272 	};
2273 
2274 	dev_info->default_txconf = (struct rte_eth_txconf) {
2275 		.tx_thresh = {
2276 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2277 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2278 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2279 		},
2280 		.offloads = 0,
2281 	};
2282 
2283 	dev_info->rx_desc_lim = rx_desc_lim;
2284 	dev_info->tx_desc_lim = tx_desc_lim;
2285 
2286 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
2287 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
2288 			ETH_LINK_SPEED_1G;
2289 
2290 	dev_info->max_mtu = dev_info->max_rx_pktlen - E1000_ETH_OVERHEAD;
2291 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
2292 
2293 	return 0;
2294 }
2295 
2296 static const uint32_t *
2297 eth_igb_supported_ptypes_get(struct rte_eth_dev *dev)
2298 {
2299 	static const uint32_t ptypes[] = {
2300 		/* refers to igb_rxd_pkt_info_to_pkt_type() */
2301 		RTE_PTYPE_L2_ETHER,
2302 		RTE_PTYPE_L3_IPV4,
2303 		RTE_PTYPE_L3_IPV4_EXT,
2304 		RTE_PTYPE_L3_IPV6,
2305 		RTE_PTYPE_L3_IPV6_EXT,
2306 		RTE_PTYPE_L4_TCP,
2307 		RTE_PTYPE_L4_UDP,
2308 		RTE_PTYPE_L4_SCTP,
2309 		RTE_PTYPE_TUNNEL_IP,
2310 		RTE_PTYPE_INNER_L3_IPV6,
2311 		RTE_PTYPE_INNER_L3_IPV6_EXT,
2312 		RTE_PTYPE_INNER_L4_TCP,
2313 		RTE_PTYPE_INNER_L4_UDP,
2314 		RTE_PTYPE_UNKNOWN
2315 	};
2316 
2317 	if (dev->rx_pkt_burst == eth_igb_recv_pkts ||
2318 	    dev->rx_pkt_burst == eth_igb_recv_scattered_pkts)
2319 		return ptypes;
2320 	return NULL;
2321 }
2322 
2323 static int
2324 eth_igbvf_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2325 {
2326 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2327 
2328 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2329 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2330 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2331 	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
2332 				DEV_TX_OFFLOAD_IPV4_CKSUM  |
2333 				DEV_TX_OFFLOAD_UDP_CKSUM   |
2334 				DEV_TX_OFFLOAD_TCP_CKSUM   |
2335 				DEV_TX_OFFLOAD_SCTP_CKSUM  |
2336 				DEV_TX_OFFLOAD_TCP_TSO;
2337 	switch (hw->mac.type) {
2338 	case e1000_vfadapt:
2339 		dev_info->max_rx_queues = 2;
2340 		dev_info->max_tx_queues = 2;
2341 		break;
2342 	case e1000_vfadapt_i350:
2343 		dev_info->max_rx_queues = 1;
2344 		dev_info->max_tx_queues = 1;
2345 		break;
2346 	default:
2347 		/* Should not happen */
2348 		return -EINVAL;
2349 	}
2350 
2351 	dev_info->rx_queue_offload_capa = igb_get_rx_queue_offloads_capa(dev);
2352 	dev_info->rx_offload_capa = igb_get_rx_port_offloads_capa(dev) |
2353 				    dev_info->rx_queue_offload_capa;
2354 	dev_info->tx_queue_offload_capa = igb_get_tx_queue_offloads_capa(dev);
2355 	dev_info->tx_offload_capa = igb_get_tx_port_offloads_capa(dev) |
2356 				    dev_info->tx_queue_offload_capa;
2357 
2358 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2359 		.rx_thresh = {
2360 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2361 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2362 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2363 		},
2364 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2365 		.rx_drop_en = 0,
2366 		.offloads = 0,
2367 	};
2368 
2369 	dev_info->default_txconf = (struct rte_eth_txconf) {
2370 		.tx_thresh = {
2371 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2372 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2373 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2374 		},
2375 		.offloads = 0,
2376 	};
2377 
2378 	dev_info->rx_desc_lim = rx_desc_lim;
2379 	dev_info->tx_desc_lim = tx_desc_lim;
2380 
2381 	return 0;
2382 }
2383 
2384 /* return 0 means link status changed, -1 means not changed */
2385 static int
2386 eth_igb_link_update(struct rte_eth_dev *dev, int wait_to_complete)
2387 {
2388 	struct e1000_hw *hw =
2389 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2390 	struct rte_eth_link link;
2391 	int link_check, count;
2392 
2393 	link_check = 0;
2394 	hw->mac.get_link_status = 1;
2395 
2396 	/* possible wait-to-complete in up to 9 seconds */
2397 	for (count = 0; count < IGB_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
2398 		/* Read the real link status */
2399 		switch (hw->phy.media_type) {
2400 		case e1000_media_type_copper:
2401 			/* Do the work to read phy */
2402 			e1000_check_for_link(hw);
2403 			link_check = !hw->mac.get_link_status;
2404 			break;
2405 
2406 		case e1000_media_type_fiber:
2407 			e1000_check_for_link(hw);
2408 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
2409 				      E1000_STATUS_LU);
2410 			break;
2411 
2412 		case e1000_media_type_internal_serdes:
2413 			e1000_check_for_link(hw);
2414 			link_check = hw->mac.serdes_has_link;
2415 			break;
2416 
2417 		/* VF device is type_unknown */
2418 		case e1000_media_type_unknown:
2419 			eth_igbvf_link_update(hw);
2420 			link_check = !hw->mac.get_link_status;
2421 			break;
2422 
2423 		default:
2424 			break;
2425 		}
2426 		if (link_check || wait_to_complete == 0)
2427 			break;
2428 		rte_delay_ms(IGB_LINK_UPDATE_CHECK_INTERVAL);
2429 	}
2430 	memset(&link, 0, sizeof(link));
2431 
2432 	/* Now we check if a transition has happened */
2433 	if (link_check) {
2434 		uint16_t duplex, speed;
2435 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
2436 		link.link_duplex = (duplex == FULL_DUPLEX) ?
2437 				ETH_LINK_FULL_DUPLEX :
2438 				ETH_LINK_HALF_DUPLEX;
2439 		link.link_speed = speed;
2440 		link.link_status = ETH_LINK_UP;
2441 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
2442 				ETH_LINK_SPEED_FIXED);
2443 	} else if (!link_check) {
2444 		link.link_speed = 0;
2445 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
2446 		link.link_status = ETH_LINK_DOWN;
2447 		link.link_autoneg = ETH_LINK_FIXED;
2448 	}
2449 
2450 	return rte_eth_linkstatus_set(dev, &link);
2451 }
2452 
2453 /*
2454  * igb_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
2455  * For ASF and Pass Through versions of f/w this means
2456  * that the driver is loaded.
2457  */
2458 static void
2459 igb_hw_control_acquire(struct e1000_hw *hw)
2460 {
2461 	uint32_t ctrl_ext;
2462 
2463 	/* Let firmware know the driver has taken over */
2464 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2465 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2466 }
2467 
2468 /*
2469  * igb_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
2470  * For ASF and Pass Through versions of f/w this means that the
2471  * driver is no longer loaded.
2472  */
2473 static void
2474 igb_hw_control_release(struct e1000_hw *hw)
2475 {
2476 	uint32_t ctrl_ext;
2477 
2478 	/* Let firmware taken over control of h/w */
2479 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2480 	E1000_WRITE_REG(hw, E1000_CTRL_EXT,
2481 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2482 }
2483 
2484 /*
2485  * Bit of a misnomer, what this really means is
2486  * to enable OS management of the system... aka
2487  * to disable special hardware management features.
2488  */
2489 static void
2490 igb_init_manageability(struct e1000_hw *hw)
2491 {
2492 	if (e1000_enable_mng_pass_thru(hw)) {
2493 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
2494 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2495 
2496 		/* disable hardware interception of ARP */
2497 		manc &= ~(E1000_MANC_ARP_EN);
2498 
2499 		/* enable receiving management packets to the host */
2500 		manc |= E1000_MANC_EN_MNG2HOST;
2501 		manc2h |= 1 << 5;  /* Mng Port 623 */
2502 		manc2h |= 1 << 6;  /* Mng Port 664 */
2503 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
2504 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2505 	}
2506 }
2507 
2508 static void
2509 igb_release_manageability(struct e1000_hw *hw)
2510 {
2511 	if (e1000_enable_mng_pass_thru(hw)) {
2512 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2513 
2514 		manc |= E1000_MANC_ARP_EN;
2515 		manc &= ~E1000_MANC_EN_MNG2HOST;
2516 
2517 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2518 	}
2519 }
2520 
2521 static int
2522 eth_igb_promiscuous_enable(struct rte_eth_dev *dev)
2523 {
2524 	struct e1000_hw *hw =
2525 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2526 	uint32_t rctl;
2527 
2528 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2529 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
2530 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2531 
2532 	return 0;
2533 }
2534 
2535 static int
2536 eth_igb_promiscuous_disable(struct rte_eth_dev *dev)
2537 {
2538 	struct e1000_hw *hw =
2539 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2540 	uint32_t rctl;
2541 
2542 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2543 	rctl &= (~E1000_RCTL_UPE);
2544 	if (dev->data->all_multicast == 1)
2545 		rctl |= E1000_RCTL_MPE;
2546 	else
2547 		rctl &= (~E1000_RCTL_MPE);
2548 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2549 
2550 	return 0;
2551 }
2552 
2553 static int
2554 eth_igb_allmulticast_enable(struct rte_eth_dev *dev)
2555 {
2556 	struct e1000_hw *hw =
2557 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2558 	uint32_t rctl;
2559 
2560 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2561 	rctl |= E1000_RCTL_MPE;
2562 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2563 
2564 	return 0;
2565 }
2566 
2567 static int
2568 eth_igb_allmulticast_disable(struct rte_eth_dev *dev)
2569 {
2570 	struct e1000_hw *hw =
2571 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2572 	uint32_t rctl;
2573 
2574 	if (dev->data->promiscuous == 1)
2575 		return 0; /* must remain in all_multicast mode */
2576 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2577 	rctl &= (~E1000_RCTL_MPE);
2578 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2579 
2580 	return 0;
2581 }
2582 
2583 static int
2584 eth_igb_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2585 {
2586 	struct e1000_hw *hw =
2587 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2588 	struct e1000_vfta * shadow_vfta =
2589 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2590 	uint32_t vfta;
2591 	uint32_t vid_idx;
2592 	uint32_t vid_bit;
2593 
2594 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
2595 			      E1000_VFTA_ENTRY_MASK);
2596 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
2597 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
2598 	if (on)
2599 		vfta |= vid_bit;
2600 	else
2601 		vfta &= ~vid_bit;
2602 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
2603 
2604 	/* update local VFTA copy */
2605 	shadow_vfta->vfta[vid_idx] = vfta;
2606 
2607 	return 0;
2608 }
2609 
2610 static int
2611 eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
2612 		      enum rte_vlan_type vlan_type,
2613 		      uint16_t tpid)
2614 {
2615 	struct e1000_hw *hw =
2616 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2617 	uint32_t reg, qinq;
2618 
2619 	qinq = E1000_READ_REG(hw, E1000_CTRL_EXT);
2620 	qinq &= E1000_CTRL_EXT_EXT_VLAN;
2621 
2622 	/* only outer TPID of double VLAN can be configured*/
2623 	if (qinq && vlan_type == ETH_VLAN_TYPE_OUTER) {
2624 		reg = E1000_READ_REG(hw, E1000_VET);
2625 		reg = (reg & (~E1000_VET_VET_EXT)) |
2626 			((uint32_t)tpid << E1000_VET_VET_EXT_SHIFT);
2627 		E1000_WRITE_REG(hw, E1000_VET, reg);
2628 
2629 		return 0;
2630 	}
2631 
2632 	/* all other TPID values are read-only*/
2633 	PMD_DRV_LOG(ERR, "Not supported");
2634 
2635 	return -ENOTSUP;
2636 }
2637 
2638 static void
2639 igb_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2640 {
2641 	struct e1000_hw *hw =
2642 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2643 	uint32_t reg;
2644 
2645 	/* Filter Table Disable */
2646 	reg = E1000_READ_REG(hw, E1000_RCTL);
2647 	reg &= ~E1000_RCTL_CFIEN;
2648 	reg &= ~E1000_RCTL_VFE;
2649 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2650 }
2651 
2652 static void
2653 igb_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2654 {
2655 	struct e1000_hw *hw =
2656 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2657 	struct e1000_vfta * shadow_vfta =
2658 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2659 	uint32_t reg;
2660 	int i;
2661 
2662 	/* Filter Table Enable, CFI not used for packet acceptance */
2663 	reg = E1000_READ_REG(hw, E1000_RCTL);
2664 	reg &= ~E1000_RCTL_CFIEN;
2665 	reg |= E1000_RCTL_VFE;
2666 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2667 
2668 	/* restore VFTA table */
2669 	for (i = 0; i < IGB_VFTA_SIZE; i++)
2670 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
2671 }
2672 
2673 static void
2674 igb_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2675 {
2676 	struct e1000_hw *hw =
2677 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2678 	uint32_t reg;
2679 
2680 	/* VLAN Mode Disable */
2681 	reg = E1000_READ_REG(hw, E1000_CTRL);
2682 	reg &= ~E1000_CTRL_VME;
2683 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2684 }
2685 
2686 static void
2687 igb_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2688 {
2689 	struct e1000_hw *hw =
2690 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2691 	uint32_t reg;
2692 
2693 	/* VLAN Mode Enable */
2694 	reg = E1000_READ_REG(hw, E1000_CTRL);
2695 	reg |= E1000_CTRL_VME;
2696 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2697 }
2698 
2699 static void
2700 igb_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2701 {
2702 	struct e1000_hw *hw =
2703 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2704 	uint32_t reg;
2705 
2706 	/* CTRL_EXT: Extended VLAN */
2707 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2708 	reg &= ~E1000_CTRL_EXT_EXTEND_VLAN;
2709 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2710 
2711 	/* Update maximum packet length */
2712 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME)
2713 		E1000_WRITE_REG(hw, E1000_RLPML,
2714 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2715 						VLAN_TAG_SIZE);
2716 }
2717 
2718 static void
2719 igb_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2720 {
2721 	struct e1000_hw *hw =
2722 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2723 	uint32_t reg;
2724 
2725 	/* CTRL_EXT: Extended VLAN */
2726 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2727 	reg |= E1000_CTRL_EXT_EXTEND_VLAN;
2728 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2729 
2730 	/* Update maximum packet length */
2731 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME)
2732 		E1000_WRITE_REG(hw, E1000_RLPML,
2733 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2734 						2 * VLAN_TAG_SIZE);
2735 }
2736 
2737 static int
2738 eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2739 {
2740 	struct rte_eth_rxmode *rxmode;
2741 
2742 	rxmode = &dev->data->dev_conf.rxmode;
2743 	if(mask & ETH_VLAN_STRIP_MASK){
2744 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2745 			igb_vlan_hw_strip_enable(dev);
2746 		else
2747 			igb_vlan_hw_strip_disable(dev);
2748 	}
2749 
2750 	if(mask & ETH_VLAN_FILTER_MASK){
2751 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2752 			igb_vlan_hw_filter_enable(dev);
2753 		else
2754 			igb_vlan_hw_filter_disable(dev);
2755 	}
2756 
2757 	if(mask & ETH_VLAN_EXTEND_MASK){
2758 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2759 			igb_vlan_hw_extend_enable(dev);
2760 		else
2761 			igb_vlan_hw_extend_disable(dev);
2762 	}
2763 
2764 	return 0;
2765 }
2766 
2767 
2768 /**
2769  * It enables the interrupt mask and then enable the interrupt.
2770  *
2771  * @param dev
2772  *  Pointer to struct rte_eth_dev.
2773  * @param on
2774  *  Enable or Disable
2775  *
2776  * @return
2777  *  - On success, zero.
2778  *  - On failure, a negative value.
2779  */
2780 static int
2781 eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
2782 {
2783 	struct e1000_interrupt *intr =
2784 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2785 
2786 	if (on)
2787 		intr->mask |= E1000_ICR_LSC;
2788 	else
2789 		intr->mask &= ~E1000_ICR_LSC;
2790 
2791 	return 0;
2792 }
2793 
2794 /* It clears the interrupt causes and enables the interrupt.
2795  * It will be called once only during nic initialized.
2796  *
2797  * @param dev
2798  *  Pointer to struct rte_eth_dev.
2799  *
2800  * @return
2801  *  - On success, zero.
2802  *  - On failure, a negative value.
2803  */
2804 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev)
2805 {
2806 	uint32_t mask, regval;
2807 	int ret;
2808 	struct e1000_hw *hw =
2809 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2810 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2811 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2812 	int misc_shift = rte_intr_allow_others(intr_handle) ? 1 : 0;
2813 	struct rte_eth_dev_info dev_info;
2814 
2815 	memset(&dev_info, 0, sizeof(dev_info));
2816 	ret = eth_igb_infos_get(dev, &dev_info);
2817 	if (ret != 0)
2818 		return ret;
2819 
2820 	mask = (0xFFFFFFFF >> (32 - dev_info.max_rx_queues)) << misc_shift;
2821 	regval = E1000_READ_REG(hw, E1000_EIMS);
2822 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
2823 
2824 	return 0;
2825 }
2826 
2827 /*
2828  * It reads ICR and gets interrupt causes, check it and set a bit flag
2829  * to update link status.
2830  *
2831  * @param dev
2832  *  Pointer to struct rte_eth_dev.
2833  *
2834  * @return
2835  *  - On success, zero.
2836  *  - On failure, a negative value.
2837  */
2838 static int
2839 eth_igb_interrupt_get_status(struct rte_eth_dev *dev)
2840 {
2841 	uint32_t icr;
2842 	struct e1000_hw *hw =
2843 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2844 	struct e1000_interrupt *intr =
2845 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2846 
2847 	igb_intr_disable(dev);
2848 
2849 	/* read-on-clear nic registers here */
2850 	icr = E1000_READ_REG(hw, E1000_ICR);
2851 
2852 	intr->flags = 0;
2853 	if (icr & E1000_ICR_LSC) {
2854 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
2855 	}
2856 
2857 	if (icr & E1000_ICR_VMMB)
2858 		intr->flags |= E1000_FLAG_MAILBOX;
2859 
2860 	return 0;
2861 }
2862 
2863 /*
2864  * It executes link_update after knowing an interrupt is prsent.
2865  *
2866  * @param dev
2867  *  Pointer to struct rte_eth_dev.
2868  *
2869  * @return
2870  *  - On success, zero.
2871  *  - On failure, a negative value.
2872  */
2873 static int
2874 eth_igb_interrupt_action(struct rte_eth_dev *dev,
2875 			 struct rte_intr_handle *intr_handle)
2876 {
2877 	struct e1000_hw *hw =
2878 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2879 	struct e1000_interrupt *intr =
2880 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2881 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2882 	struct rte_eth_link link;
2883 	int ret;
2884 
2885 	if (intr->flags & E1000_FLAG_MAILBOX) {
2886 		igb_pf_mbx_process(dev);
2887 		intr->flags &= ~E1000_FLAG_MAILBOX;
2888 	}
2889 
2890 	igb_intr_enable(dev);
2891 	rte_intr_ack(intr_handle);
2892 
2893 	if (intr->flags & E1000_FLAG_NEED_LINK_UPDATE) {
2894 		intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
2895 
2896 		/* set get_link_status to check register later */
2897 		hw->mac.get_link_status = 1;
2898 		ret = eth_igb_link_update(dev, 0);
2899 
2900 		/* check if link has changed */
2901 		if (ret < 0)
2902 			return 0;
2903 
2904 		rte_eth_linkstatus_get(dev, &link);
2905 		if (link.link_status) {
2906 			PMD_INIT_LOG(INFO,
2907 				     " Port %d: Link Up - speed %u Mbps - %s",
2908 				     dev->data->port_id,
2909 				     (unsigned)link.link_speed,
2910 				     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
2911 				     "full-duplex" : "half-duplex");
2912 		} else {
2913 			PMD_INIT_LOG(INFO, " Port %d: Link Down",
2914 				     dev->data->port_id);
2915 		}
2916 
2917 		PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
2918 			     pci_dev->addr.domain,
2919 			     pci_dev->addr.bus,
2920 			     pci_dev->addr.devid,
2921 			     pci_dev->addr.function);
2922 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC,
2923 					      NULL);
2924 	}
2925 
2926 	return 0;
2927 }
2928 
2929 /**
2930  * Interrupt handler which shall be registered at first.
2931  *
2932  * @param handle
2933  *  Pointer to interrupt handle.
2934  * @param param
2935  *  The address of parameter (struct rte_eth_dev *) regsitered before.
2936  *
2937  * @return
2938  *  void
2939  */
2940 static void
2941 eth_igb_interrupt_handler(void *param)
2942 {
2943 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2944 
2945 	eth_igb_interrupt_get_status(dev);
2946 	eth_igb_interrupt_action(dev, dev->intr_handle);
2947 }
2948 
2949 static int
2950 eth_igbvf_interrupt_get_status(struct rte_eth_dev *dev)
2951 {
2952 	uint32_t eicr;
2953 	struct e1000_hw *hw =
2954 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2955 	struct e1000_interrupt *intr =
2956 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2957 
2958 	igbvf_intr_disable(hw);
2959 
2960 	/* read-on-clear nic registers here */
2961 	eicr = E1000_READ_REG(hw, E1000_EICR);
2962 	intr->flags = 0;
2963 
2964 	if (eicr == E1000_VTIVAR_MISC_MAILBOX)
2965 		intr->flags |= E1000_FLAG_MAILBOX;
2966 
2967 	return 0;
2968 }
2969 
2970 void igbvf_mbx_process(struct rte_eth_dev *dev)
2971 {
2972 	struct e1000_hw *hw =
2973 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2974 	struct e1000_mbx_info *mbx = &hw->mbx;
2975 	u32 in_msg = 0;
2976 
2977 	/* peek the message first */
2978 	in_msg = E1000_READ_REG(hw, E1000_VMBMEM(0));
2979 
2980 	/* PF reset VF event */
2981 	if (in_msg == E1000_PF_CONTROL_MSG) {
2982 		/* dummy mbx read to ack pf */
2983 		if (mbx->ops.read(hw, &in_msg, 1, 0))
2984 			return;
2985 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
2986 					      NULL);
2987 	}
2988 }
2989 
2990 static int
2991 eth_igbvf_interrupt_action(struct rte_eth_dev *dev, struct rte_intr_handle *intr_handle)
2992 {
2993 	struct e1000_interrupt *intr =
2994 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2995 
2996 	if (intr->flags & E1000_FLAG_MAILBOX) {
2997 		igbvf_mbx_process(dev);
2998 		intr->flags &= ~E1000_FLAG_MAILBOX;
2999 	}
3000 
3001 	igbvf_intr_enable(dev);
3002 	rte_intr_ack(intr_handle);
3003 
3004 	return 0;
3005 }
3006 
3007 static void
3008 eth_igbvf_interrupt_handler(void *param)
3009 {
3010 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
3011 
3012 	eth_igbvf_interrupt_get_status(dev);
3013 	eth_igbvf_interrupt_action(dev, dev->intr_handle);
3014 }
3015 
3016 static int
3017 eth_igb_led_on(struct rte_eth_dev *dev)
3018 {
3019 	struct e1000_hw *hw;
3020 
3021 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3022 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3023 }
3024 
3025 static int
3026 eth_igb_led_off(struct rte_eth_dev *dev)
3027 {
3028 	struct e1000_hw *hw;
3029 
3030 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3031 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3032 }
3033 
3034 static int
3035 eth_igb_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3036 {
3037 	struct e1000_hw *hw;
3038 	uint32_t ctrl;
3039 	int tx_pause;
3040 	int rx_pause;
3041 
3042 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3043 	fc_conf->pause_time = hw->fc.pause_time;
3044 	fc_conf->high_water = hw->fc.high_water;
3045 	fc_conf->low_water = hw->fc.low_water;
3046 	fc_conf->send_xon = hw->fc.send_xon;
3047 	fc_conf->autoneg = hw->mac.autoneg;
3048 
3049 	/*
3050 	 * Return rx_pause and tx_pause status according to actual setting of
3051 	 * the TFCE and RFCE bits in the CTRL register.
3052 	 */
3053 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
3054 	if (ctrl & E1000_CTRL_TFCE)
3055 		tx_pause = 1;
3056 	else
3057 		tx_pause = 0;
3058 
3059 	if (ctrl & E1000_CTRL_RFCE)
3060 		rx_pause = 1;
3061 	else
3062 		rx_pause = 0;
3063 
3064 	if (rx_pause && tx_pause)
3065 		fc_conf->mode = RTE_FC_FULL;
3066 	else if (rx_pause)
3067 		fc_conf->mode = RTE_FC_RX_PAUSE;
3068 	else if (tx_pause)
3069 		fc_conf->mode = RTE_FC_TX_PAUSE;
3070 	else
3071 		fc_conf->mode = RTE_FC_NONE;
3072 
3073 	return 0;
3074 }
3075 
3076 static int
3077 eth_igb_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3078 {
3079 	struct e1000_hw *hw;
3080 	int err;
3081 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
3082 		e1000_fc_none,
3083 		e1000_fc_rx_pause,
3084 		e1000_fc_tx_pause,
3085 		e1000_fc_full
3086 	};
3087 	uint32_t rx_buf_size;
3088 	uint32_t max_high_water;
3089 	uint32_t rctl;
3090 
3091 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3092 	if (fc_conf->autoneg != hw->mac.autoneg)
3093 		return -ENOTSUP;
3094 	rx_buf_size = igb_get_rx_buffer_size(hw);
3095 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3096 
3097 	/* At least reserve one Ethernet frame for watermark */
3098 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
3099 	if ((fc_conf->high_water > max_high_water) ||
3100 	    (fc_conf->high_water < fc_conf->low_water)) {
3101 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
3102 		PMD_INIT_LOG(ERR, "high water must <=  0x%x", max_high_water);
3103 		return -EINVAL;
3104 	}
3105 
3106 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
3107 	hw->fc.pause_time     = fc_conf->pause_time;
3108 	hw->fc.high_water     = fc_conf->high_water;
3109 	hw->fc.low_water      = fc_conf->low_water;
3110 	hw->fc.send_xon	      = fc_conf->send_xon;
3111 
3112 	err = e1000_setup_link_generic(hw);
3113 	if (err == E1000_SUCCESS) {
3114 
3115 		/* check if we want to forward MAC frames - driver doesn't have native
3116 		 * capability to do that, so we'll write the registers ourselves */
3117 
3118 		rctl = E1000_READ_REG(hw, E1000_RCTL);
3119 
3120 		/* set or clear MFLCN.PMCF bit depending on configuration */
3121 		if (fc_conf->mac_ctrl_frame_fwd != 0)
3122 			rctl |= E1000_RCTL_PMCF;
3123 		else
3124 			rctl &= ~E1000_RCTL_PMCF;
3125 
3126 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3127 		E1000_WRITE_FLUSH(hw);
3128 
3129 		return 0;
3130 	}
3131 
3132 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
3133 	return -EIO;
3134 }
3135 
3136 #define E1000_RAH_POOLSEL_SHIFT      (18)
3137 static int
3138 eth_igb_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
3139 		uint32_t index, uint32_t pool)
3140 {
3141 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3142 	uint32_t rah;
3143 
3144 	e1000_rar_set(hw, mac_addr->addr_bytes, index);
3145 	rah = E1000_READ_REG(hw, E1000_RAH(index));
3146 	rah |= (0x1 << (E1000_RAH_POOLSEL_SHIFT + pool));
3147 	E1000_WRITE_REG(hw, E1000_RAH(index), rah);
3148 	return 0;
3149 }
3150 
3151 static void
3152 eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index)
3153 {
3154 	uint8_t addr[RTE_ETHER_ADDR_LEN];
3155 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3156 
3157 	memset(addr, 0, sizeof(addr));
3158 
3159 	e1000_rar_set(hw, addr, index);
3160 }
3161 
3162 static int
3163 eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
3164 				struct rte_ether_addr *addr)
3165 {
3166 	eth_igb_rar_clear(dev, 0);
3167 	eth_igb_rar_set(dev, (void *)addr, 0, 0);
3168 
3169 	return 0;
3170 }
3171 /*
3172  * Virtual Function operations
3173  */
3174 static void
3175 igbvf_intr_disable(struct e1000_hw *hw)
3176 {
3177 	PMD_INIT_FUNC_TRACE();
3178 
3179 	/* Clear interrupt mask to stop from interrupts being generated */
3180 	E1000_WRITE_REG(hw, E1000_EIMC, 0xFFFF);
3181 
3182 	E1000_WRITE_FLUSH(hw);
3183 }
3184 
3185 static void
3186 igbvf_stop_adapter(struct rte_eth_dev *dev)
3187 {
3188 	u32 reg_val;
3189 	u16 i;
3190 	struct rte_eth_dev_info dev_info;
3191 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3192 	int ret;
3193 
3194 	memset(&dev_info, 0, sizeof(dev_info));
3195 	ret = eth_igbvf_infos_get(dev, &dev_info);
3196 	if (ret != 0)
3197 		return;
3198 
3199 	/* Clear interrupt mask to stop from interrupts being generated */
3200 	igbvf_intr_disable(hw);
3201 
3202 	/* Clear any pending interrupts, flush previous writes */
3203 	E1000_READ_REG(hw, E1000_EICR);
3204 
3205 	/* Disable the transmit unit.  Each queue must be disabled. */
3206 	for (i = 0; i < dev_info.max_tx_queues; i++)
3207 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), E1000_TXDCTL_SWFLSH);
3208 
3209 	/* Disable the receive unit by stopping each queue */
3210 	for (i = 0; i < dev_info.max_rx_queues; i++) {
3211 		reg_val = E1000_READ_REG(hw, E1000_RXDCTL(i));
3212 		reg_val &= ~E1000_RXDCTL_QUEUE_ENABLE;
3213 		E1000_WRITE_REG(hw, E1000_RXDCTL(i), reg_val);
3214 		while (E1000_READ_REG(hw, E1000_RXDCTL(i)) & E1000_RXDCTL_QUEUE_ENABLE)
3215 			;
3216 	}
3217 
3218 	/* flush all queues disables */
3219 	E1000_WRITE_FLUSH(hw);
3220 	msec_delay(2);
3221 }
3222 
3223 static int eth_igbvf_link_update(struct e1000_hw *hw)
3224 {
3225 	struct e1000_mbx_info *mbx = &hw->mbx;
3226 	struct e1000_mac_info *mac = &hw->mac;
3227 	int ret_val = E1000_SUCCESS;
3228 
3229 	PMD_INIT_LOG(DEBUG, "e1000_check_for_link_vf");
3230 
3231 	/*
3232 	 * We only want to run this if there has been a rst asserted.
3233 	 * in this case that could mean a link change, device reset,
3234 	 * or a virtual function reset
3235 	 */
3236 
3237 	/* If we were hit with a reset or timeout drop the link */
3238 	if (!e1000_check_for_rst(hw, 0) || !mbx->timeout)
3239 		mac->get_link_status = TRUE;
3240 
3241 	if (!mac->get_link_status)
3242 		goto out;
3243 
3244 	/* if link status is down no point in checking to see if pf is up */
3245 	if (!(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU))
3246 		goto out;
3247 
3248 	/* if we passed all the tests above then the link is up and we no
3249 	 * longer need to check for link */
3250 	mac->get_link_status = FALSE;
3251 
3252 out:
3253 	return ret_val;
3254 }
3255 
3256 
3257 static int
3258 igbvf_dev_configure(struct rte_eth_dev *dev)
3259 {
3260 	struct rte_eth_conf* conf = &dev->data->dev_conf;
3261 
3262 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
3263 		     dev->data->port_id);
3264 
3265 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
3266 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
3267 
3268 	/*
3269 	 * VF has no ability to enable/disable HW CRC
3270 	 * Keep the persistent behavior the same as Host PF
3271 	 */
3272 #ifndef RTE_LIBRTE_E1000_PF_DISABLE_STRIP_CRC
3273 	if (conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
3274 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
3275 		conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_KEEP_CRC;
3276 	}
3277 #else
3278 	if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC)) {
3279 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
3280 		conf->rxmode.offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
3281 	}
3282 #endif
3283 
3284 	return 0;
3285 }
3286 
3287 static int
3288 igbvf_dev_start(struct rte_eth_dev *dev)
3289 {
3290 	struct e1000_hw *hw =
3291 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3292 	struct e1000_adapter *adapter =
3293 		E1000_DEV_PRIVATE(dev->data->dev_private);
3294 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3295 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3296 	int ret;
3297 	uint32_t intr_vector = 0;
3298 
3299 	PMD_INIT_FUNC_TRACE();
3300 
3301 	hw->mac.ops.reset_hw(hw);
3302 	adapter->stopped = 0;
3303 
3304 	/* Set all vfta */
3305 	igbvf_set_vfta_all(dev,1);
3306 
3307 	eth_igbvf_tx_init(dev);
3308 
3309 	/* This can fail when allocating mbufs for descriptor rings */
3310 	ret = eth_igbvf_rx_init(dev);
3311 	if (ret) {
3312 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
3313 		igb_dev_clear_queues(dev);
3314 		return ret;
3315 	}
3316 
3317 	/* check and configure queue intr-vector mapping */
3318 	if (rte_intr_cap_multiple(intr_handle) &&
3319 	    dev->data->dev_conf.intr_conf.rxq) {
3320 		intr_vector = dev->data->nb_rx_queues;
3321 		ret = rte_intr_efd_enable(intr_handle, intr_vector);
3322 		if (ret)
3323 			return ret;
3324 	}
3325 
3326 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
3327 		intr_handle->intr_vec =
3328 			rte_zmalloc("intr_vec",
3329 				    dev->data->nb_rx_queues * sizeof(int), 0);
3330 		if (!intr_handle->intr_vec) {
3331 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
3332 				     " intr_vec", dev->data->nb_rx_queues);
3333 			return -ENOMEM;
3334 		}
3335 	}
3336 
3337 	eth_igbvf_configure_msix_intr(dev);
3338 
3339 	/* enable uio/vfio intr/eventfd mapping */
3340 	rte_intr_enable(intr_handle);
3341 
3342 	/* resume enabled intr since hw reset */
3343 	igbvf_intr_enable(dev);
3344 
3345 	return 0;
3346 }
3347 
3348 static void
3349 igbvf_dev_stop(struct rte_eth_dev *dev)
3350 {
3351 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3352 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3353 	struct e1000_adapter *adapter =
3354 		E1000_DEV_PRIVATE(dev->data->dev_private);
3355 
3356 	if (adapter->stopped)
3357 		return;
3358 
3359 	PMD_INIT_FUNC_TRACE();
3360 
3361 	igbvf_stop_adapter(dev);
3362 
3363 	/*
3364 	  * Clear what we set, but we still keep shadow_vfta to
3365 	  * restore after device starts
3366 	  */
3367 	igbvf_set_vfta_all(dev,0);
3368 
3369 	igb_dev_clear_queues(dev);
3370 
3371 	/* disable intr eventfd mapping */
3372 	rte_intr_disable(intr_handle);
3373 
3374 	/* Clean datapath event and queue/vec mapping */
3375 	rte_intr_efd_disable(intr_handle);
3376 	if (intr_handle->intr_vec) {
3377 		rte_free(intr_handle->intr_vec);
3378 		intr_handle->intr_vec = NULL;
3379 	}
3380 
3381 	adapter->stopped = true;
3382 }
3383 
3384 static void
3385 igbvf_dev_close(struct rte_eth_dev *dev)
3386 {
3387 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3388 	struct rte_ether_addr addr;
3389 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3390 
3391 	PMD_INIT_FUNC_TRACE();
3392 
3393 	e1000_reset_hw(hw);
3394 
3395 	igbvf_dev_stop(dev);
3396 
3397 	igb_dev_free_queues(dev);
3398 
3399 	/**
3400 	 * reprogram the RAR with a zero mac address,
3401 	 * to ensure that the VF traffic goes to the PF
3402 	 * after stop, close and detach of the VF.
3403 	 **/
3404 
3405 	memset(&addr, 0, sizeof(addr));
3406 	igbvf_default_mac_addr_set(dev, &addr);
3407 
3408 	dev->dev_ops = NULL;
3409 	dev->rx_pkt_burst = NULL;
3410 	dev->tx_pkt_burst = NULL;
3411 
3412 	rte_intr_callback_unregister(&pci_dev->intr_handle,
3413 				     eth_igbvf_interrupt_handler,
3414 				     (void *)dev);
3415 }
3416 
3417 static int
3418 igbvf_promiscuous_enable(struct rte_eth_dev *dev)
3419 {
3420 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3421 
3422 	/* Set both unicast and multicast promisc */
3423 	e1000_promisc_set_vf(hw, e1000_promisc_enabled);
3424 
3425 	return 0;
3426 }
3427 
3428 static int
3429 igbvf_promiscuous_disable(struct rte_eth_dev *dev)
3430 {
3431 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3432 
3433 	/* If in allmulticast mode leave multicast promisc */
3434 	if (dev->data->all_multicast == 1)
3435 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3436 	else
3437 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3438 
3439 	return 0;
3440 }
3441 
3442 static int
3443 igbvf_allmulticast_enable(struct rte_eth_dev *dev)
3444 {
3445 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3446 
3447 	/* In promiscuous mode multicast promisc already set */
3448 	if (dev->data->promiscuous == 0)
3449 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3450 
3451 	return 0;
3452 }
3453 
3454 static int
3455 igbvf_allmulticast_disable(struct rte_eth_dev *dev)
3456 {
3457 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3458 
3459 	/* In promiscuous mode leave multicast promisc enabled */
3460 	if (dev->data->promiscuous == 0)
3461 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3462 
3463 	return 0;
3464 }
3465 
3466 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on)
3467 {
3468 	struct e1000_mbx_info *mbx = &hw->mbx;
3469 	uint32_t msgbuf[2];
3470 	s32 err;
3471 
3472 	/* After set vlan, vlan strip will also be enabled in igb driver*/
3473 	msgbuf[0] = E1000_VF_SET_VLAN;
3474 	msgbuf[1] = vid;
3475 	/* Setting the 8 bit field MSG INFO to TRUE indicates "add" */
3476 	if (on)
3477 		msgbuf[0] |= E1000_VF_SET_VLAN_ADD;
3478 
3479 	err = mbx->ops.write_posted(hw, msgbuf, 2, 0);
3480 	if (err)
3481 		goto mbx_err;
3482 
3483 	err = mbx->ops.read_posted(hw, msgbuf, 2, 0);
3484 	if (err)
3485 		goto mbx_err;
3486 
3487 	msgbuf[0] &= ~E1000_VT_MSGTYPE_CTS;
3488 	if (msgbuf[0] == (E1000_VF_SET_VLAN | E1000_VT_MSGTYPE_NACK))
3489 		err = -EINVAL;
3490 
3491 mbx_err:
3492 	return err;
3493 }
3494 
3495 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on)
3496 {
3497 	struct e1000_hw *hw =
3498 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3499 	struct e1000_vfta * shadow_vfta =
3500 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3501 	int i = 0, j = 0, vfta = 0, mask = 1;
3502 
3503 	for (i = 0; i < IGB_VFTA_SIZE; i++){
3504 		vfta = shadow_vfta->vfta[i];
3505 		if(vfta){
3506 			mask = 1;
3507 			for (j = 0; j < 32; j++){
3508 				if(vfta & mask)
3509 					igbvf_set_vfta(hw,
3510 						(uint16_t)((i<<5)+j), on);
3511 				mask<<=1;
3512 			}
3513 		}
3514 	}
3515 
3516 }
3517 
3518 static int
3519 igbvf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
3520 {
3521 	struct e1000_hw *hw =
3522 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3523 	struct e1000_vfta * shadow_vfta =
3524 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3525 	uint32_t vid_idx = 0;
3526 	uint32_t vid_bit = 0;
3527 	int ret = 0;
3528 
3529 	PMD_INIT_FUNC_TRACE();
3530 
3531 	/*vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf*/
3532 	ret = igbvf_set_vfta(hw, vlan_id, !!on);
3533 	if(ret){
3534 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
3535 		return ret;
3536 	}
3537 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
3538 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
3539 
3540 	/*Save what we set and retore it after device reset*/
3541 	if (on)
3542 		shadow_vfta->vfta[vid_idx] |= vid_bit;
3543 	else
3544 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
3545 
3546 	return 0;
3547 }
3548 
3549 static int
3550 igbvf_default_mac_addr_set(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
3551 {
3552 	struct e1000_hw *hw =
3553 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3554 
3555 	/* index is not used by rar_set() */
3556 	hw->mac.ops.rar_set(hw, (void *)addr, 0);
3557 	return 0;
3558 }
3559 
3560 
3561 static int
3562 eth_igb_rss_reta_update(struct rte_eth_dev *dev,
3563 			struct rte_eth_rss_reta_entry64 *reta_conf,
3564 			uint16_t reta_size)
3565 {
3566 	uint8_t i, j, mask;
3567 	uint32_t reta, r;
3568 	uint16_t idx, shift;
3569 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3570 
3571 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3572 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3573 			"(%d) doesn't match the number hardware can supported "
3574 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3575 		return -EINVAL;
3576 	}
3577 
3578 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3579 		idx = i / RTE_RETA_GROUP_SIZE;
3580 		shift = i % RTE_RETA_GROUP_SIZE;
3581 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3582 						IGB_4_BIT_MASK);
3583 		if (!mask)
3584 			continue;
3585 		if (mask == IGB_4_BIT_MASK)
3586 			r = 0;
3587 		else
3588 			r = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3589 		for (j = 0, reta = 0; j < IGB_4_BIT_WIDTH; j++) {
3590 			if (mask & (0x1 << j))
3591 				reta |= reta_conf[idx].reta[shift + j] <<
3592 							(CHAR_BIT * j);
3593 			else
3594 				reta |= r & (IGB_8_BIT_MASK << (CHAR_BIT * j));
3595 		}
3596 		E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
3597 	}
3598 
3599 	return 0;
3600 }
3601 
3602 static int
3603 eth_igb_rss_reta_query(struct rte_eth_dev *dev,
3604 		       struct rte_eth_rss_reta_entry64 *reta_conf,
3605 		       uint16_t reta_size)
3606 {
3607 	uint8_t i, j, mask;
3608 	uint32_t reta;
3609 	uint16_t idx, shift;
3610 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3611 
3612 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3613 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3614 			"(%d) doesn't match the number hardware can supported "
3615 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3616 		return -EINVAL;
3617 	}
3618 
3619 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3620 		idx = i / RTE_RETA_GROUP_SIZE;
3621 		shift = i % RTE_RETA_GROUP_SIZE;
3622 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3623 						IGB_4_BIT_MASK);
3624 		if (!mask)
3625 			continue;
3626 		reta = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3627 		for (j = 0; j < IGB_4_BIT_WIDTH; j++) {
3628 			if (mask & (0x1 << j))
3629 				reta_conf[idx].reta[shift + j] =
3630 					((reta >> (CHAR_BIT * j)) &
3631 						IGB_8_BIT_MASK);
3632 		}
3633 	}
3634 
3635 	return 0;
3636 }
3637 
3638 int
3639 eth_igb_syn_filter_set(struct rte_eth_dev *dev,
3640 			struct rte_eth_syn_filter *filter,
3641 			bool add)
3642 {
3643 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3644 	struct e1000_filter_info *filter_info =
3645 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3646 	uint32_t synqf, rfctl;
3647 
3648 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3649 		return -EINVAL;
3650 
3651 	synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3652 
3653 	if (add) {
3654 		if (synqf & E1000_SYN_FILTER_ENABLE)
3655 			return -EINVAL;
3656 
3657 		synqf = (uint32_t)(((filter->queue << E1000_SYN_FILTER_QUEUE_SHIFT) &
3658 			E1000_SYN_FILTER_QUEUE) | E1000_SYN_FILTER_ENABLE);
3659 
3660 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3661 		if (filter->hig_pri)
3662 			rfctl |= E1000_RFCTL_SYNQFP;
3663 		else
3664 			rfctl &= ~E1000_RFCTL_SYNQFP;
3665 
3666 		E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3667 	} else {
3668 		if (!(synqf & E1000_SYN_FILTER_ENABLE))
3669 			return -ENOENT;
3670 		synqf = 0;
3671 	}
3672 
3673 	filter_info->syn_info = synqf;
3674 	E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
3675 	E1000_WRITE_FLUSH(hw);
3676 	return 0;
3677 }
3678 
3679 static int
3680 eth_igb_syn_filter_get(struct rte_eth_dev *dev,
3681 			struct rte_eth_syn_filter *filter)
3682 {
3683 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3684 	uint32_t synqf, rfctl;
3685 
3686 	synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3687 	if (synqf & E1000_SYN_FILTER_ENABLE) {
3688 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3689 		filter->hig_pri = (rfctl & E1000_RFCTL_SYNQFP) ? 1 : 0;
3690 		filter->queue = (uint8_t)((synqf & E1000_SYN_FILTER_QUEUE) >>
3691 				E1000_SYN_FILTER_QUEUE_SHIFT);
3692 		return 0;
3693 	}
3694 
3695 	return -ENOENT;
3696 }
3697 
3698 static int
3699 eth_igb_syn_filter_handle(struct rte_eth_dev *dev,
3700 			enum rte_filter_op filter_op,
3701 			void *arg)
3702 {
3703 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3704 	int ret;
3705 
3706 	MAC_TYPE_FILTER_SUP(hw->mac.type);
3707 
3708 	if (filter_op == RTE_ETH_FILTER_NOP)
3709 		return 0;
3710 
3711 	if (arg == NULL) {
3712 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
3713 			    filter_op);
3714 		return -EINVAL;
3715 	}
3716 
3717 	switch (filter_op) {
3718 	case RTE_ETH_FILTER_ADD:
3719 		ret = eth_igb_syn_filter_set(dev,
3720 				(struct rte_eth_syn_filter *)arg,
3721 				TRUE);
3722 		break;
3723 	case RTE_ETH_FILTER_DELETE:
3724 		ret = eth_igb_syn_filter_set(dev,
3725 				(struct rte_eth_syn_filter *)arg,
3726 				FALSE);
3727 		break;
3728 	case RTE_ETH_FILTER_GET:
3729 		ret = eth_igb_syn_filter_get(dev,
3730 				(struct rte_eth_syn_filter *)arg);
3731 		break;
3732 	default:
3733 		PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
3734 		ret = -EINVAL;
3735 		break;
3736 	}
3737 
3738 	return ret;
3739 }
3740 
3741 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_2tuple_filter_info*/
3742 static inline int
3743 ntuple_filter_to_2tuple(struct rte_eth_ntuple_filter *filter,
3744 			struct e1000_2tuple_filter_info *filter_info)
3745 {
3746 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3747 		return -EINVAL;
3748 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
3749 		return -EINVAL;  /* filter index is out of range. */
3750 	if (filter->tcp_flags > RTE_NTUPLE_TCP_FLAGS_MASK)
3751 		return -EINVAL;  /* flags is invalid. */
3752 
3753 	switch (filter->dst_port_mask) {
3754 	case UINT16_MAX:
3755 		filter_info->dst_port_mask = 0;
3756 		filter_info->dst_port = filter->dst_port;
3757 		break;
3758 	case 0:
3759 		filter_info->dst_port_mask = 1;
3760 		break;
3761 	default:
3762 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
3763 		return -EINVAL;
3764 	}
3765 
3766 	switch (filter->proto_mask) {
3767 	case UINT8_MAX:
3768 		filter_info->proto_mask = 0;
3769 		filter_info->proto = filter->proto;
3770 		break;
3771 	case 0:
3772 		filter_info->proto_mask = 1;
3773 		break;
3774 	default:
3775 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
3776 		return -EINVAL;
3777 	}
3778 
3779 	filter_info->priority = (uint8_t)filter->priority;
3780 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
3781 		filter_info->tcp_flags = filter->tcp_flags;
3782 	else
3783 		filter_info->tcp_flags = 0;
3784 
3785 	return 0;
3786 }
3787 
3788 static inline struct e1000_2tuple_filter *
3789 igb_2tuple_filter_lookup(struct e1000_2tuple_filter_list *filter_list,
3790 			struct e1000_2tuple_filter_info *key)
3791 {
3792 	struct e1000_2tuple_filter *it;
3793 
3794 	TAILQ_FOREACH(it, filter_list, entries) {
3795 		if (memcmp(key, &it->filter_info,
3796 			sizeof(struct e1000_2tuple_filter_info)) == 0) {
3797 			return it;
3798 		}
3799 	}
3800 	return NULL;
3801 }
3802 
3803 /* inject a igb 2tuple filter to HW */
3804 static inline void
3805 igb_inject_2uple_filter(struct rte_eth_dev *dev,
3806 			   struct e1000_2tuple_filter *filter)
3807 {
3808 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3809 	uint32_t ttqf = E1000_TTQF_DISABLE_MASK;
3810 	uint32_t imir, imir_ext = E1000_IMIREXT_SIZE_BP;
3811 	int i;
3812 
3813 	i = filter->index;
3814 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
3815 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
3816 		imir |= E1000_IMIR_PORT_BP;
3817 	else
3818 		imir &= ~E1000_IMIR_PORT_BP;
3819 
3820 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
3821 
3822 	ttqf |= E1000_TTQF_QUEUE_ENABLE;
3823 	ttqf |= (uint32_t)(filter->queue << E1000_TTQF_QUEUE_SHIFT);
3824 	ttqf |= (uint32_t)(filter->filter_info.proto &
3825 						E1000_TTQF_PROTOCOL_MASK);
3826 	if (filter->filter_info.proto_mask == 0)
3827 		ttqf &= ~E1000_TTQF_MASK_ENABLE;
3828 
3829 	/* tcp flags bits setting. */
3830 	if (filter->filter_info.tcp_flags & RTE_NTUPLE_TCP_FLAGS_MASK) {
3831 		if (filter->filter_info.tcp_flags & RTE_TCP_URG_FLAG)
3832 			imir_ext |= E1000_IMIREXT_CTRL_URG;
3833 		if (filter->filter_info.tcp_flags & RTE_TCP_ACK_FLAG)
3834 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
3835 		if (filter->filter_info.tcp_flags & RTE_TCP_PSH_FLAG)
3836 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
3837 		if (filter->filter_info.tcp_flags & RTE_TCP_RST_FLAG)
3838 			imir_ext |= E1000_IMIREXT_CTRL_RST;
3839 		if (filter->filter_info.tcp_flags & RTE_TCP_SYN_FLAG)
3840 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
3841 		if (filter->filter_info.tcp_flags & RTE_TCP_FIN_FLAG)
3842 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
3843 	} else {
3844 		imir_ext |= E1000_IMIREXT_CTRL_BP;
3845 	}
3846 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
3847 	E1000_WRITE_REG(hw, E1000_TTQF(i), ttqf);
3848 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
3849 }
3850 
3851 /*
3852  * igb_add_2tuple_filter - add a 2tuple filter
3853  *
3854  * @param
3855  * dev: Pointer to struct rte_eth_dev.
3856  * ntuple_filter: ponter to the filter that will be added.
3857  *
3858  * @return
3859  *    - On success, zero.
3860  *    - On failure, a negative value.
3861  */
3862 static int
3863 igb_add_2tuple_filter(struct rte_eth_dev *dev,
3864 			struct rte_eth_ntuple_filter *ntuple_filter)
3865 {
3866 	struct e1000_filter_info *filter_info =
3867 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3868 	struct e1000_2tuple_filter *filter;
3869 	int i, ret;
3870 
3871 	filter = rte_zmalloc("e1000_2tuple_filter",
3872 			sizeof(struct e1000_2tuple_filter), 0);
3873 	if (filter == NULL)
3874 		return -ENOMEM;
3875 
3876 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3877 				      &filter->filter_info);
3878 	if (ret < 0) {
3879 		rte_free(filter);
3880 		return ret;
3881 	}
3882 	if (igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3883 					 &filter->filter_info) != NULL) {
3884 		PMD_DRV_LOG(ERR, "filter exists.");
3885 		rte_free(filter);
3886 		return -EEXIST;
3887 	}
3888 	filter->queue = ntuple_filter->queue;
3889 
3890 	/*
3891 	 * look for an unused 2tuple filter index,
3892 	 * and insert the filter to list.
3893 	 */
3894 	for (i = 0; i < E1000_MAX_TTQF_FILTERS; i++) {
3895 		if (!(filter_info->twotuple_mask & (1 << i))) {
3896 			filter_info->twotuple_mask |= 1 << i;
3897 			filter->index = i;
3898 			TAILQ_INSERT_TAIL(&filter_info->twotuple_list,
3899 					  filter,
3900 					  entries);
3901 			break;
3902 		}
3903 	}
3904 	if (i >= E1000_MAX_TTQF_FILTERS) {
3905 		PMD_DRV_LOG(ERR, "2tuple filters are full.");
3906 		rte_free(filter);
3907 		return -ENOSYS;
3908 	}
3909 
3910 	igb_inject_2uple_filter(dev, filter);
3911 	return 0;
3912 }
3913 
3914 int
3915 igb_delete_2tuple_filter(struct rte_eth_dev *dev,
3916 			struct e1000_2tuple_filter *filter)
3917 {
3918 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3919 	struct e1000_filter_info *filter_info =
3920 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3921 
3922 	filter_info->twotuple_mask &= ~(1 << filter->index);
3923 	TAILQ_REMOVE(&filter_info->twotuple_list, filter, entries);
3924 	rte_free(filter);
3925 
3926 	E1000_WRITE_REG(hw, E1000_TTQF(filter->index), E1000_TTQF_DISABLE_MASK);
3927 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
3928 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
3929 	return 0;
3930 }
3931 
3932 /*
3933  * igb_remove_2tuple_filter - remove a 2tuple filter
3934  *
3935  * @param
3936  * dev: Pointer to struct rte_eth_dev.
3937  * ntuple_filter: ponter to the filter that will be removed.
3938  *
3939  * @return
3940  *    - On success, zero.
3941  *    - On failure, a negative value.
3942  */
3943 static int
3944 igb_remove_2tuple_filter(struct rte_eth_dev *dev,
3945 			struct rte_eth_ntuple_filter *ntuple_filter)
3946 {
3947 	struct e1000_filter_info *filter_info =
3948 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3949 	struct e1000_2tuple_filter_info filter_2tuple;
3950 	struct e1000_2tuple_filter *filter;
3951 	int ret;
3952 
3953 	memset(&filter_2tuple, 0, sizeof(struct e1000_2tuple_filter_info));
3954 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3955 				      &filter_2tuple);
3956 	if (ret < 0)
3957 		return ret;
3958 
3959 	filter = igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3960 					 &filter_2tuple);
3961 	if (filter == NULL) {
3962 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
3963 		return -ENOENT;
3964 	}
3965 
3966 	igb_delete_2tuple_filter(dev, filter);
3967 
3968 	return 0;
3969 }
3970 
3971 /* inject a igb flex filter to HW */
3972 static inline void
3973 igb_inject_flex_filter(struct rte_eth_dev *dev,
3974 			   struct e1000_flex_filter *filter)
3975 {
3976 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3977 	uint32_t wufc, queueing;
3978 	uint32_t reg_off;
3979 	uint8_t i, j = 0;
3980 
3981 	wufc = E1000_READ_REG(hw, E1000_WUFC);
3982 	if (filter->index < E1000_MAX_FHFT)
3983 		reg_off = E1000_FHFT(filter->index);
3984 	else
3985 		reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
3986 
3987 	E1000_WRITE_REG(hw, E1000_WUFC, wufc | E1000_WUFC_FLEX_HQ |
3988 			(E1000_WUFC_FLX0 << filter->index));
3989 	queueing = filter->filter_info.len |
3990 		(filter->queue << E1000_FHFT_QUEUEING_QUEUE_SHIFT) |
3991 		(filter->filter_info.priority <<
3992 			E1000_FHFT_QUEUEING_PRIO_SHIFT);
3993 	E1000_WRITE_REG(hw, reg_off + E1000_FHFT_QUEUEING_OFFSET,
3994 			queueing);
3995 
3996 	for (i = 0; i < E1000_FLEX_FILTERS_MASK_SIZE; i++) {
3997 		E1000_WRITE_REG(hw, reg_off,
3998 				filter->filter_info.dwords[j]);
3999 		reg_off += sizeof(uint32_t);
4000 		E1000_WRITE_REG(hw, reg_off,
4001 				filter->filter_info.dwords[++j]);
4002 		reg_off += sizeof(uint32_t);
4003 		E1000_WRITE_REG(hw, reg_off,
4004 			(uint32_t)filter->filter_info.mask[i]);
4005 		reg_off += sizeof(uint32_t) * 2;
4006 		++j;
4007 	}
4008 }
4009 
4010 static inline struct e1000_flex_filter *
4011 eth_igb_flex_filter_lookup(struct e1000_flex_filter_list *filter_list,
4012 			struct e1000_flex_filter_info *key)
4013 {
4014 	struct e1000_flex_filter *it;
4015 
4016 	TAILQ_FOREACH(it, filter_list, entries) {
4017 		if (memcmp(key, &it->filter_info,
4018 			sizeof(struct e1000_flex_filter_info)) == 0)
4019 			return it;
4020 	}
4021 
4022 	return NULL;
4023 }
4024 
4025 /* remove a flex byte filter
4026  * @param
4027  * dev: Pointer to struct rte_eth_dev.
4028  * filter: the pointer of the filter will be removed.
4029  */
4030 void
4031 igb_remove_flex_filter(struct rte_eth_dev *dev,
4032 			struct e1000_flex_filter *filter)
4033 {
4034 	struct e1000_filter_info *filter_info =
4035 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4036 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4037 	uint32_t wufc, i;
4038 	uint32_t reg_off;
4039 
4040 	wufc = E1000_READ_REG(hw, E1000_WUFC);
4041 	if (filter->index < E1000_MAX_FHFT)
4042 		reg_off = E1000_FHFT(filter->index);
4043 	else
4044 		reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
4045 
4046 	for (i = 0; i < E1000_FHFT_SIZE_IN_DWD; i++)
4047 		E1000_WRITE_REG(hw, reg_off + i * sizeof(uint32_t), 0);
4048 
4049 	E1000_WRITE_REG(hw, E1000_WUFC, wufc &
4050 		(~(E1000_WUFC_FLX0 << filter->index)));
4051 
4052 	filter_info->flex_mask &= ~(1 << filter->index);
4053 	TAILQ_REMOVE(&filter_info->flex_list, filter, entries);
4054 	rte_free(filter);
4055 }
4056 
4057 int
4058 eth_igb_add_del_flex_filter(struct rte_eth_dev *dev,
4059 			struct rte_eth_flex_filter *filter,
4060 			bool add)
4061 {
4062 	struct e1000_filter_info *filter_info =
4063 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4064 	struct e1000_flex_filter *flex_filter, *it;
4065 	uint32_t mask;
4066 	uint8_t shift, i;
4067 
4068 	flex_filter = rte_zmalloc("e1000_flex_filter",
4069 			sizeof(struct e1000_flex_filter), 0);
4070 	if (flex_filter == NULL)
4071 		return -ENOMEM;
4072 
4073 	flex_filter->filter_info.len = filter->len;
4074 	flex_filter->filter_info.priority = filter->priority;
4075 	memcpy(flex_filter->filter_info.dwords, filter->bytes, filter->len);
4076 	for (i = 0; i < RTE_ALIGN(filter->len, CHAR_BIT) / CHAR_BIT; i++) {
4077 		mask = 0;
4078 		/* reverse bits in flex filter's mask*/
4079 		for (shift = 0; shift < CHAR_BIT; shift++) {
4080 			if (filter->mask[i] & (0x01 << shift))
4081 				mask |= (0x80 >> shift);
4082 		}
4083 		flex_filter->filter_info.mask[i] = mask;
4084 	}
4085 
4086 	it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
4087 				&flex_filter->filter_info);
4088 	if (it == NULL && !add) {
4089 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4090 		rte_free(flex_filter);
4091 		return -ENOENT;
4092 	}
4093 	if (it != NULL && add) {
4094 		PMD_DRV_LOG(ERR, "filter exists.");
4095 		rte_free(flex_filter);
4096 		return -EEXIST;
4097 	}
4098 
4099 	if (add) {
4100 		flex_filter->queue = filter->queue;
4101 		/*
4102 		 * look for an unused flex filter index
4103 		 * and insert the filter into the list.
4104 		 */
4105 		for (i = 0; i < E1000_MAX_FLEX_FILTERS; i++) {
4106 			if (!(filter_info->flex_mask & (1 << i))) {
4107 				filter_info->flex_mask |= 1 << i;
4108 				flex_filter->index = i;
4109 				TAILQ_INSERT_TAIL(&filter_info->flex_list,
4110 					flex_filter,
4111 					entries);
4112 				break;
4113 			}
4114 		}
4115 		if (i >= E1000_MAX_FLEX_FILTERS) {
4116 			PMD_DRV_LOG(ERR, "flex filters are full.");
4117 			rte_free(flex_filter);
4118 			return -ENOSYS;
4119 		}
4120 
4121 		igb_inject_flex_filter(dev, flex_filter);
4122 
4123 	} else {
4124 		igb_remove_flex_filter(dev, it);
4125 		rte_free(flex_filter);
4126 	}
4127 
4128 	return 0;
4129 }
4130 
4131 static int
4132 eth_igb_get_flex_filter(struct rte_eth_dev *dev,
4133 			struct rte_eth_flex_filter *filter)
4134 {
4135 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4136 	struct e1000_filter_info *filter_info =
4137 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4138 	struct e1000_flex_filter flex_filter, *it;
4139 	uint32_t wufc, queueing, wufc_en = 0;
4140 
4141 	memset(&flex_filter, 0, sizeof(struct e1000_flex_filter));
4142 	flex_filter.filter_info.len = filter->len;
4143 	flex_filter.filter_info.priority = filter->priority;
4144 	memcpy(flex_filter.filter_info.dwords, filter->bytes, filter->len);
4145 	memcpy(flex_filter.filter_info.mask, filter->mask,
4146 			RTE_ALIGN(filter->len, CHAR_BIT) / CHAR_BIT);
4147 
4148 	it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
4149 				&flex_filter.filter_info);
4150 	if (it == NULL) {
4151 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4152 		return -ENOENT;
4153 	}
4154 
4155 	wufc = E1000_READ_REG(hw, E1000_WUFC);
4156 	wufc_en = E1000_WUFC_FLEX_HQ | (E1000_WUFC_FLX0 << it->index);
4157 
4158 	if ((wufc & wufc_en) == wufc_en) {
4159 		uint32_t reg_off = 0;
4160 		if (it->index < E1000_MAX_FHFT)
4161 			reg_off = E1000_FHFT(it->index);
4162 		else
4163 			reg_off = E1000_FHFT_EXT(it->index - E1000_MAX_FHFT);
4164 
4165 		queueing = E1000_READ_REG(hw,
4166 				reg_off + E1000_FHFT_QUEUEING_OFFSET);
4167 		filter->len = queueing & E1000_FHFT_QUEUEING_LEN;
4168 		filter->priority = (queueing & E1000_FHFT_QUEUEING_PRIO) >>
4169 			E1000_FHFT_QUEUEING_PRIO_SHIFT;
4170 		filter->queue = (queueing & E1000_FHFT_QUEUEING_QUEUE) >>
4171 			E1000_FHFT_QUEUEING_QUEUE_SHIFT;
4172 		return 0;
4173 	}
4174 	return -ENOENT;
4175 }
4176 
4177 static int
4178 eth_igb_flex_filter_handle(struct rte_eth_dev *dev,
4179 			enum rte_filter_op filter_op,
4180 			void *arg)
4181 {
4182 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4183 	struct rte_eth_flex_filter *filter;
4184 	int ret = 0;
4185 
4186 	MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
4187 
4188 	if (filter_op == RTE_ETH_FILTER_NOP)
4189 		return ret;
4190 
4191 	if (arg == NULL) {
4192 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
4193 			    filter_op);
4194 		return -EINVAL;
4195 	}
4196 
4197 	filter = (struct rte_eth_flex_filter *)arg;
4198 	if (filter->len == 0 || filter->len > E1000_MAX_FLEX_FILTER_LEN
4199 	    || filter->len % sizeof(uint64_t) != 0) {
4200 		PMD_DRV_LOG(ERR, "filter's length is out of range");
4201 		return -EINVAL;
4202 	}
4203 	if (filter->priority > E1000_MAX_FLEX_FILTER_PRI) {
4204 		PMD_DRV_LOG(ERR, "filter's priority is out of range");
4205 		return -EINVAL;
4206 	}
4207 
4208 	switch (filter_op) {
4209 	case RTE_ETH_FILTER_ADD:
4210 		ret = eth_igb_add_del_flex_filter(dev, filter, TRUE);
4211 		break;
4212 	case RTE_ETH_FILTER_DELETE:
4213 		ret = eth_igb_add_del_flex_filter(dev, filter, FALSE);
4214 		break;
4215 	case RTE_ETH_FILTER_GET:
4216 		ret = eth_igb_get_flex_filter(dev, filter);
4217 		break;
4218 	default:
4219 		PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
4220 		ret = -EINVAL;
4221 		break;
4222 	}
4223 
4224 	return ret;
4225 }
4226 
4227 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_5tuple_filter_info*/
4228 static inline int
4229 ntuple_filter_to_5tuple_82576(struct rte_eth_ntuple_filter *filter,
4230 			struct e1000_5tuple_filter_info *filter_info)
4231 {
4232 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM_82576)
4233 		return -EINVAL;
4234 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
4235 		return -EINVAL;  /* filter index is out of range. */
4236 	if (filter->tcp_flags > RTE_NTUPLE_TCP_FLAGS_MASK)
4237 		return -EINVAL;  /* flags is invalid. */
4238 
4239 	switch (filter->dst_ip_mask) {
4240 	case UINT32_MAX:
4241 		filter_info->dst_ip_mask = 0;
4242 		filter_info->dst_ip = filter->dst_ip;
4243 		break;
4244 	case 0:
4245 		filter_info->dst_ip_mask = 1;
4246 		break;
4247 	default:
4248 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
4249 		return -EINVAL;
4250 	}
4251 
4252 	switch (filter->src_ip_mask) {
4253 	case UINT32_MAX:
4254 		filter_info->src_ip_mask = 0;
4255 		filter_info->src_ip = filter->src_ip;
4256 		break;
4257 	case 0:
4258 		filter_info->src_ip_mask = 1;
4259 		break;
4260 	default:
4261 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
4262 		return -EINVAL;
4263 	}
4264 
4265 	switch (filter->dst_port_mask) {
4266 	case UINT16_MAX:
4267 		filter_info->dst_port_mask = 0;
4268 		filter_info->dst_port = filter->dst_port;
4269 		break;
4270 	case 0:
4271 		filter_info->dst_port_mask = 1;
4272 		break;
4273 	default:
4274 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
4275 		return -EINVAL;
4276 	}
4277 
4278 	switch (filter->src_port_mask) {
4279 	case UINT16_MAX:
4280 		filter_info->src_port_mask = 0;
4281 		filter_info->src_port = filter->src_port;
4282 		break;
4283 	case 0:
4284 		filter_info->src_port_mask = 1;
4285 		break;
4286 	default:
4287 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
4288 		return -EINVAL;
4289 	}
4290 
4291 	switch (filter->proto_mask) {
4292 	case UINT8_MAX:
4293 		filter_info->proto_mask = 0;
4294 		filter_info->proto = filter->proto;
4295 		break;
4296 	case 0:
4297 		filter_info->proto_mask = 1;
4298 		break;
4299 	default:
4300 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
4301 		return -EINVAL;
4302 	}
4303 
4304 	filter_info->priority = (uint8_t)filter->priority;
4305 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
4306 		filter_info->tcp_flags = filter->tcp_flags;
4307 	else
4308 		filter_info->tcp_flags = 0;
4309 
4310 	return 0;
4311 }
4312 
4313 static inline struct e1000_5tuple_filter *
4314 igb_5tuple_filter_lookup_82576(struct e1000_5tuple_filter_list *filter_list,
4315 			struct e1000_5tuple_filter_info *key)
4316 {
4317 	struct e1000_5tuple_filter *it;
4318 
4319 	TAILQ_FOREACH(it, filter_list, entries) {
4320 		if (memcmp(key, &it->filter_info,
4321 			sizeof(struct e1000_5tuple_filter_info)) == 0) {
4322 			return it;
4323 		}
4324 	}
4325 	return NULL;
4326 }
4327 
4328 /* inject a igb 5-tuple filter to HW */
4329 static inline void
4330 igb_inject_5tuple_filter_82576(struct rte_eth_dev *dev,
4331 			   struct e1000_5tuple_filter *filter)
4332 {
4333 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4334 	uint32_t ftqf = E1000_FTQF_VF_BP | E1000_FTQF_MASK;
4335 	uint32_t spqf, imir, imir_ext = E1000_IMIREXT_SIZE_BP;
4336 	uint8_t i;
4337 
4338 	i = filter->index;
4339 	ftqf |= filter->filter_info.proto & E1000_FTQF_PROTOCOL_MASK;
4340 	if (filter->filter_info.src_ip_mask == 0) /* 0b means compare. */
4341 		ftqf &= ~E1000_FTQF_MASK_SOURCE_ADDR_BP;
4342 	if (filter->filter_info.dst_ip_mask == 0)
4343 		ftqf &= ~E1000_FTQF_MASK_DEST_ADDR_BP;
4344 	if (filter->filter_info.src_port_mask == 0)
4345 		ftqf &= ~E1000_FTQF_MASK_SOURCE_PORT_BP;
4346 	if (filter->filter_info.proto_mask == 0)
4347 		ftqf &= ~E1000_FTQF_MASK_PROTO_BP;
4348 	ftqf |= (filter->queue << E1000_FTQF_QUEUE_SHIFT) &
4349 		E1000_FTQF_QUEUE_MASK;
4350 	ftqf |= E1000_FTQF_QUEUE_ENABLE;
4351 	E1000_WRITE_REG(hw, E1000_FTQF(i), ftqf);
4352 	E1000_WRITE_REG(hw, E1000_DAQF(i), filter->filter_info.dst_ip);
4353 	E1000_WRITE_REG(hw, E1000_SAQF(i), filter->filter_info.src_ip);
4354 
4355 	spqf = filter->filter_info.src_port & E1000_SPQF_SRCPORT;
4356 	E1000_WRITE_REG(hw, E1000_SPQF(i), spqf);
4357 
4358 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
4359 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
4360 		imir |= E1000_IMIR_PORT_BP;
4361 	else
4362 		imir &= ~E1000_IMIR_PORT_BP;
4363 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
4364 
4365 	/* tcp flags bits setting. */
4366 	if (filter->filter_info.tcp_flags & RTE_NTUPLE_TCP_FLAGS_MASK) {
4367 		if (filter->filter_info.tcp_flags & RTE_TCP_URG_FLAG)
4368 			imir_ext |= E1000_IMIREXT_CTRL_URG;
4369 		if (filter->filter_info.tcp_flags & RTE_TCP_ACK_FLAG)
4370 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
4371 		if (filter->filter_info.tcp_flags & RTE_TCP_PSH_FLAG)
4372 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
4373 		if (filter->filter_info.tcp_flags & RTE_TCP_RST_FLAG)
4374 			imir_ext |= E1000_IMIREXT_CTRL_RST;
4375 		if (filter->filter_info.tcp_flags & RTE_TCP_SYN_FLAG)
4376 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
4377 		if (filter->filter_info.tcp_flags & RTE_TCP_FIN_FLAG)
4378 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
4379 	} else {
4380 		imir_ext |= E1000_IMIREXT_CTRL_BP;
4381 	}
4382 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
4383 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
4384 }
4385 
4386 /*
4387  * igb_add_5tuple_filter_82576 - add a 5tuple filter
4388  *
4389  * @param
4390  * dev: Pointer to struct rte_eth_dev.
4391  * ntuple_filter: ponter to the filter that will be added.
4392  *
4393  * @return
4394  *    - On success, zero.
4395  *    - On failure, a negative value.
4396  */
4397 static int
4398 igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
4399 			struct rte_eth_ntuple_filter *ntuple_filter)
4400 {
4401 	struct e1000_filter_info *filter_info =
4402 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4403 	struct e1000_5tuple_filter *filter;
4404 	uint8_t i;
4405 	int ret;
4406 
4407 	filter = rte_zmalloc("e1000_5tuple_filter",
4408 			sizeof(struct e1000_5tuple_filter), 0);
4409 	if (filter == NULL)
4410 		return -ENOMEM;
4411 
4412 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4413 					    &filter->filter_info);
4414 	if (ret < 0) {
4415 		rte_free(filter);
4416 		return ret;
4417 	}
4418 
4419 	if (igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4420 					 &filter->filter_info) != NULL) {
4421 		PMD_DRV_LOG(ERR, "filter exists.");
4422 		rte_free(filter);
4423 		return -EEXIST;
4424 	}
4425 	filter->queue = ntuple_filter->queue;
4426 
4427 	/*
4428 	 * look for an unused 5tuple filter index,
4429 	 * and insert the filter to list.
4430 	 */
4431 	for (i = 0; i < E1000_MAX_FTQF_FILTERS; i++) {
4432 		if (!(filter_info->fivetuple_mask & (1 << i))) {
4433 			filter_info->fivetuple_mask |= 1 << i;
4434 			filter->index = i;
4435 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
4436 					  filter,
4437 					  entries);
4438 			break;
4439 		}
4440 	}
4441 	if (i >= E1000_MAX_FTQF_FILTERS) {
4442 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
4443 		rte_free(filter);
4444 		return -ENOSYS;
4445 	}
4446 
4447 	igb_inject_5tuple_filter_82576(dev, filter);
4448 	return 0;
4449 }
4450 
4451 int
4452 igb_delete_5tuple_filter_82576(struct rte_eth_dev *dev,
4453 				struct e1000_5tuple_filter *filter)
4454 {
4455 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4456 	struct e1000_filter_info *filter_info =
4457 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4458 
4459 	filter_info->fivetuple_mask &= ~(1 << filter->index);
4460 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
4461 	rte_free(filter);
4462 
4463 	E1000_WRITE_REG(hw, E1000_FTQF(filter->index),
4464 			E1000_FTQF_VF_BP | E1000_FTQF_MASK);
4465 	E1000_WRITE_REG(hw, E1000_DAQF(filter->index), 0);
4466 	E1000_WRITE_REG(hw, E1000_SAQF(filter->index), 0);
4467 	E1000_WRITE_REG(hw, E1000_SPQF(filter->index), 0);
4468 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
4469 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
4470 	return 0;
4471 }
4472 
4473 /*
4474  * igb_remove_5tuple_filter_82576 - remove a 5tuple filter
4475  *
4476  * @param
4477  * dev: Pointer to struct rte_eth_dev.
4478  * ntuple_filter: ponter to the filter that will be removed.
4479  *
4480  * @return
4481  *    - On success, zero.
4482  *    - On failure, a negative value.
4483  */
4484 static int
4485 igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
4486 				struct rte_eth_ntuple_filter *ntuple_filter)
4487 {
4488 	struct e1000_filter_info *filter_info =
4489 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4490 	struct e1000_5tuple_filter_info filter_5tuple;
4491 	struct e1000_5tuple_filter *filter;
4492 	int ret;
4493 
4494 	memset(&filter_5tuple, 0, sizeof(struct e1000_5tuple_filter_info));
4495 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4496 					    &filter_5tuple);
4497 	if (ret < 0)
4498 		return ret;
4499 
4500 	filter = igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4501 					 &filter_5tuple);
4502 	if (filter == NULL) {
4503 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4504 		return -ENOENT;
4505 	}
4506 
4507 	igb_delete_5tuple_filter_82576(dev, filter);
4508 
4509 	return 0;
4510 }
4511 
4512 static int
4513 eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4514 {
4515 	uint32_t rctl;
4516 	struct e1000_hw *hw;
4517 	struct rte_eth_dev_info dev_info;
4518 	uint32_t frame_size = mtu + E1000_ETH_OVERHEAD;
4519 	int ret;
4520 
4521 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4522 
4523 #ifdef RTE_LIBRTE_82571_SUPPORT
4524 	/* XXX: not bigger than max_rx_pktlen */
4525 	if (hw->mac.type == e1000_82571)
4526 		return -ENOTSUP;
4527 #endif
4528 	ret = eth_igb_infos_get(dev, &dev_info);
4529 	if (ret != 0)
4530 		return ret;
4531 
4532 	/* check that mtu is within the allowed range */
4533 	if (mtu < RTE_ETHER_MIN_MTU ||
4534 			frame_size > dev_info.max_rx_pktlen)
4535 		return -EINVAL;
4536 
4537 	/* refuse mtu that requires the support of scattered packets when this
4538 	 * feature has not been enabled before. */
4539 	if (!dev->data->scattered_rx &&
4540 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
4541 		return -EINVAL;
4542 
4543 	rctl = E1000_READ_REG(hw, E1000_RCTL);
4544 
4545 	/* switch to jumbo mode if needed */
4546 	if (frame_size > RTE_ETHER_MAX_LEN) {
4547 		dev->data->dev_conf.rxmode.offloads |=
4548 			DEV_RX_OFFLOAD_JUMBO_FRAME;
4549 		rctl |= E1000_RCTL_LPE;
4550 	} else {
4551 		dev->data->dev_conf.rxmode.offloads &=
4552 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
4553 		rctl &= ~E1000_RCTL_LPE;
4554 	}
4555 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
4556 
4557 	/* update max frame size */
4558 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4559 
4560 	E1000_WRITE_REG(hw, E1000_RLPML,
4561 			dev->data->dev_conf.rxmode.max_rx_pkt_len);
4562 
4563 	return 0;
4564 }
4565 
4566 /*
4567  * igb_add_del_ntuple_filter - add or delete a ntuple filter
4568  *
4569  * @param
4570  * dev: Pointer to struct rte_eth_dev.
4571  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4572  * add: if true, add filter, if false, remove filter
4573  *
4574  * @return
4575  *    - On success, zero.
4576  *    - On failure, a negative value.
4577  */
4578 int
4579 igb_add_del_ntuple_filter(struct rte_eth_dev *dev,
4580 			struct rte_eth_ntuple_filter *ntuple_filter,
4581 			bool add)
4582 {
4583 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4584 	int ret;
4585 
4586 	switch (ntuple_filter->flags) {
4587 	case RTE_5TUPLE_FLAGS:
4588 	case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4589 		if (hw->mac.type != e1000_82576)
4590 			return -ENOTSUP;
4591 		if (add)
4592 			ret = igb_add_5tuple_filter_82576(dev,
4593 							  ntuple_filter);
4594 		else
4595 			ret = igb_remove_5tuple_filter_82576(dev,
4596 							     ntuple_filter);
4597 		break;
4598 	case RTE_2TUPLE_FLAGS:
4599 	case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4600 		if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350 &&
4601 			hw->mac.type != e1000_i210 &&
4602 			hw->mac.type != e1000_i211)
4603 			return -ENOTSUP;
4604 		if (add)
4605 			ret = igb_add_2tuple_filter(dev, ntuple_filter);
4606 		else
4607 			ret = igb_remove_2tuple_filter(dev, ntuple_filter);
4608 		break;
4609 	default:
4610 		ret = -EINVAL;
4611 		break;
4612 	}
4613 
4614 	return ret;
4615 }
4616 
4617 /*
4618  * igb_get_ntuple_filter - get a ntuple filter
4619  *
4620  * @param
4621  * dev: Pointer to struct rte_eth_dev.
4622  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4623  *
4624  * @return
4625  *    - On success, zero.
4626  *    - On failure, a negative value.
4627  */
4628 static int
4629 igb_get_ntuple_filter(struct rte_eth_dev *dev,
4630 			struct rte_eth_ntuple_filter *ntuple_filter)
4631 {
4632 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4633 	struct e1000_filter_info *filter_info =
4634 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4635 	struct e1000_5tuple_filter_info filter_5tuple;
4636 	struct e1000_2tuple_filter_info filter_2tuple;
4637 	struct e1000_5tuple_filter *p_5tuple_filter;
4638 	struct e1000_2tuple_filter *p_2tuple_filter;
4639 	int ret;
4640 
4641 	switch (ntuple_filter->flags) {
4642 	case RTE_5TUPLE_FLAGS:
4643 	case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4644 		if (hw->mac.type != e1000_82576)
4645 			return -ENOTSUP;
4646 		memset(&filter_5tuple,
4647 			0,
4648 			sizeof(struct e1000_5tuple_filter_info));
4649 		ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4650 						    &filter_5tuple);
4651 		if (ret < 0)
4652 			return ret;
4653 		p_5tuple_filter = igb_5tuple_filter_lookup_82576(
4654 					&filter_info->fivetuple_list,
4655 					&filter_5tuple);
4656 		if (p_5tuple_filter == NULL) {
4657 			PMD_DRV_LOG(ERR, "filter doesn't exist.");
4658 			return -ENOENT;
4659 		}
4660 		ntuple_filter->queue = p_5tuple_filter->queue;
4661 		break;
4662 	case RTE_2TUPLE_FLAGS:
4663 	case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4664 		if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350)
4665 			return -ENOTSUP;
4666 		memset(&filter_2tuple,
4667 			0,
4668 			sizeof(struct e1000_2tuple_filter_info));
4669 		ret = ntuple_filter_to_2tuple(ntuple_filter, &filter_2tuple);
4670 		if (ret < 0)
4671 			return ret;
4672 		p_2tuple_filter = igb_2tuple_filter_lookup(
4673 					&filter_info->twotuple_list,
4674 					&filter_2tuple);
4675 		if (p_2tuple_filter == NULL) {
4676 			PMD_DRV_LOG(ERR, "filter doesn't exist.");
4677 			return -ENOENT;
4678 		}
4679 		ntuple_filter->queue = p_2tuple_filter->queue;
4680 		break;
4681 	default:
4682 		ret = -EINVAL;
4683 		break;
4684 	}
4685 
4686 	return 0;
4687 }
4688 
4689 /*
4690  * igb_ntuple_filter_handle - Handle operations for ntuple filter.
4691  * @dev: pointer to rte_eth_dev structure
4692  * @filter_op:operation will be taken.
4693  * @arg: a pointer to specific structure corresponding to the filter_op
4694  */
4695 static int
4696 igb_ntuple_filter_handle(struct rte_eth_dev *dev,
4697 				enum rte_filter_op filter_op,
4698 				void *arg)
4699 {
4700 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4701 	int ret;
4702 
4703 	MAC_TYPE_FILTER_SUP(hw->mac.type);
4704 
4705 	if (filter_op == RTE_ETH_FILTER_NOP)
4706 		return 0;
4707 
4708 	if (arg == NULL) {
4709 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4710 			    filter_op);
4711 		return -EINVAL;
4712 	}
4713 
4714 	switch (filter_op) {
4715 	case RTE_ETH_FILTER_ADD:
4716 		ret = igb_add_del_ntuple_filter(dev,
4717 			(struct rte_eth_ntuple_filter *)arg,
4718 			TRUE);
4719 		break;
4720 	case RTE_ETH_FILTER_DELETE:
4721 		ret = igb_add_del_ntuple_filter(dev,
4722 			(struct rte_eth_ntuple_filter *)arg,
4723 			FALSE);
4724 		break;
4725 	case RTE_ETH_FILTER_GET:
4726 		ret = igb_get_ntuple_filter(dev,
4727 			(struct rte_eth_ntuple_filter *)arg);
4728 		break;
4729 	default:
4730 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
4731 		ret = -EINVAL;
4732 		break;
4733 	}
4734 	return ret;
4735 }
4736 
4737 static inline int
4738 igb_ethertype_filter_lookup(struct e1000_filter_info *filter_info,
4739 			uint16_t ethertype)
4740 {
4741 	int i;
4742 
4743 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4744 		if (filter_info->ethertype_filters[i].ethertype == ethertype &&
4745 		    (filter_info->ethertype_mask & (1 << i)))
4746 			return i;
4747 	}
4748 	return -1;
4749 }
4750 
4751 static inline int
4752 igb_ethertype_filter_insert(struct e1000_filter_info *filter_info,
4753 			uint16_t ethertype, uint32_t etqf)
4754 {
4755 	int i;
4756 
4757 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4758 		if (!(filter_info->ethertype_mask & (1 << i))) {
4759 			filter_info->ethertype_mask |= 1 << i;
4760 			filter_info->ethertype_filters[i].ethertype = ethertype;
4761 			filter_info->ethertype_filters[i].etqf = etqf;
4762 			return i;
4763 		}
4764 	}
4765 	return -1;
4766 }
4767 
4768 int
4769 igb_ethertype_filter_remove(struct e1000_filter_info *filter_info,
4770 			uint8_t idx)
4771 {
4772 	if (idx >= E1000_MAX_ETQF_FILTERS)
4773 		return -1;
4774 	filter_info->ethertype_mask &= ~(1 << idx);
4775 	filter_info->ethertype_filters[idx].ethertype = 0;
4776 	filter_info->ethertype_filters[idx].etqf = 0;
4777 	return idx;
4778 }
4779 
4780 
4781 int
4782 igb_add_del_ethertype_filter(struct rte_eth_dev *dev,
4783 			struct rte_eth_ethertype_filter *filter,
4784 			bool add)
4785 {
4786 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4787 	struct e1000_filter_info *filter_info =
4788 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4789 	uint32_t etqf = 0;
4790 	int ret;
4791 
4792 	if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
4793 		filter->ether_type == RTE_ETHER_TYPE_IPV6) {
4794 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
4795 			" ethertype filter.", filter->ether_type);
4796 		return -EINVAL;
4797 	}
4798 
4799 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
4800 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
4801 		return -EINVAL;
4802 	}
4803 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
4804 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
4805 		return -EINVAL;
4806 	}
4807 
4808 	ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4809 	if (ret >= 0 && add) {
4810 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
4811 			    filter->ether_type);
4812 		return -EEXIST;
4813 	}
4814 	if (ret < 0 && !add) {
4815 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4816 			    filter->ether_type);
4817 		return -ENOENT;
4818 	}
4819 
4820 	if (add) {
4821 		etqf |= E1000_ETQF_FILTER_ENABLE | E1000_ETQF_QUEUE_ENABLE;
4822 		etqf |= (uint32_t)(filter->ether_type & E1000_ETQF_ETHERTYPE);
4823 		etqf |= filter->queue << E1000_ETQF_QUEUE_SHIFT;
4824 		ret = igb_ethertype_filter_insert(filter_info,
4825 				filter->ether_type, etqf);
4826 		if (ret < 0) {
4827 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
4828 			return -ENOSYS;
4829 		}
4830 	} else {
4831 		ret = igb_ethertype_filter_remove(filter_info, (uint8_t)ret);
4832 		if (ret < 0)
4833 			return -ENOSYS;
4834 	}
4835 	E1000_WRITE_REG(hw, E1000_ETQF(ret), etqf);
4836 	E1000_WRITE_FLUSH(hw);
4837 
4838 	return 0;
4839 }
4840 
4841 static int
4842 igb_get_ethertype_filter(struct rte_eth_dev *dev,
4843 			struct rte_eth_ethertype_filter *filter)
4844 {
4845 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4846 	struct e1000_filter_info *filter_info =
4847 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4848 	uint32_t etqf;
4849 	int ret;
4850 
4851 	ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4852 	if (ret < 0) {
4853 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4854 			    filter->ether_type);
4855 		return -ENOENT;
4856 	}
4857 
4858 	etqf = E1000_READ_REG(hw, E1000_ETQF(ret));
4859 	if (etqf & E1000_ETQF_FILTER_ENABLE) {
4860 		filter->ether_type = etqf & E1000_ETQF_ETHERTYPE;
4861 		filter->flags = 0;
4862 		filter->queue = (etqf & E1000_ETQF_QUEUE) >>
4863 				E1000_ETQF_QUEUE_SHIFT;
4864 		return 0;
4865 	}
4866 
4867 	return -ENOENT;
4868 }
4869 
4870 /*
4871  * igb_ethertype_filter_handle - Handle operations for ethertype filter.
4872  * @dev: pointer to rte_eth_dev structure
4873  * @filter_op:operation will be taken.
4874  * @arg: a pointer to specific structure corresponding to the filter_op
4875  */
4876 static int
4877 igb_ethertype_filter_handle(struct rte_eth_dev *dev,
4878 				enum rte_filter_op filter_op,
4879 				void *arg)
4880 {
4881 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4882 	int ret;
4883 
4884 	MAC_TYPE_FILTER_SUP(hw->mac.type);
4885 
4886 	if (filter_op == RTE_ETH_FILTER_NOP)
4887 		return 0;
4888 
4889 	if (arg == NULL) {
4890 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4891 			    filter_op);
4892 		return -EINVAL;
4893 	}
4894 
4895 	switch (filter_op) {
4896 	case RTE_ETH_FILTER_ADD:
4897 		ret = igb_add_del_ethertype_filter(dev,
4898 			(struct rte_eth_ethertype_filter *)arg,
4899 			TRUE);
4900 		break;
4901 	case RTE_ETH_FILTER_DELETE:
4902 		ret = igb_add_del_ethertype_filter(dev,
4903 			(struct rte_eth_ethertype_filter *)arg,
4904 			FALSE);
4905 		break;
4906 	case RTE_ETH_FILTER_GET:
4907 		ret = igb_get_ethertype_filter(dev,
4908 			(struct rte_eth_ethertype_filter *)arg);
4909 		break;
4910 	default:
4911 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
4912 		ret = -EINVAL;
4913 		break;
4914 	}
4915 	return ret;
4916 }
4917 
4918 static int
4919 eth_igb_filter_ctrl(struct rte_eth_dev *dev,
4920 		     enum rte_filter_type filter_type,
4921 		     enum rte_filter_op filter_op,
4922 		     void *arg)
4923 {
4924 	int ret = 0;
4925 
4926 	switch (filter_type) {
4927 	case RTE_ETH_FILTER_NTUPLE:
4928 		ret = igb_ntuple_filter_handle(dev, filter_op, arg);
4929 		break;
4930 	case RTE_ETH_FILTER_ETHERTYPE:
4931 		ret = igb_ethertype_filter_handle(dev, filter_op, arg);
4932 		break;
4933 	case RTE_ETH_FILTER_SYN:
4934 		ret = eth_igb_syn_filter_handle(dev, filter_op, arg);
4935 		break;
4936 	case RTE_ETH_FILTER_FLEXIBLE:
4937 		ret = eth_igb_flex_filter_handle(dev, filter_op, arg);
4938 		break;
4939 	case RTE_ETH_FILTER_GENERIC:
4940 		if (filter_op != RTE_ETH_FILTER_GET)
4941 			return -EINVAL;
4942 		*(const void **)arg = &igb_flow_ops;
4943 		break;
4944 	default:
4945 		PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
4946 							filter_type);
4947 		break;
4948 	}
4949 
4950 	return ret;
4951 }
4952 
4953 static int
4954 eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
4955 			 struct rte_ether_addr *mc_addr_set,
4956 			 uint32_t nb_mc_addr)
4957 {
4958 	struct e1000_hw *hw;
4959 
4960 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4961 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
4962 	return 0;
4963 }
4964 
4965 static uint64_t
4966 igb_read_systime_cyclecounter(struct rte_eth_dev *dev)
4967 {
4968 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4969 	uint64_t systime_cycles;
4970 
4971 	switch (hw->mac.type) {
4972 	case e1000_i210:
4973 	case e1000_i211:
4974 		/*
4975 		 * Need to read System Time Residue Register to be able
4976 		 * to read the other two registers.
4977 		 */
4978 		E1000_READ_REG(hw, E1000_SYSTIMR);
4979 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
4980 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4981 		systime_cycles += (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4982 				* NSEC_PER_SEC;
4983 		break;
4984 	case e1000_82580:
4985 	case e1000_i350:
4986 	case e1000_i354:
4987 		/*
4988 		 * Need to read System Time Residue Register to be able
4989 		 * to read the other two registers.
4990 		 */
4991 		E1000_READ_REG(hw, E1000_SYSTIMR);
4992 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4993 		/* Only the 8 LSB are valid. */
4994 		systime_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_SYSTIMH)
4995 				& 0xff) << 32;
4996 		break;
4997 	default:
4998 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4999 		systime_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
5000 				<< 32;
5001 		break;
5002 	}
5003 
5004 	return systime_cycles;
5005 }
5006 
5007 static uint64_t
5008 igb_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
5009 {
5010 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5011 	uint64_t rx_tstamp_cycles;
5012 
5013 	switch (hw->mac.type) {
5014 	case e1000_i210:
5015 	case e1000_i211:
5016 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
5017 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
5018 		rx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
5019 				* NSEC_PER_SEC;
5020 		break;
5021 	case e1000_82580:
5022 	case e1000_i350:
5023 	case e1000_i354:
5024 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
5025 		/* Only the 8 LSB are valid. */
5026 		rx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_RXSTMPH)
5027 				& 0xff) << 32;
5028 		break;
5029 	default:
5030 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
5031 		rx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
5032 				<< 32;
5033 		break;
5034 	}
5035 
5036 	return rx_tstamp_cycles;
5037 }
5038 
5039 static uint64_t
5040 igb_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
5041 {
5042 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5043 	uint64_t tx_tstamp_cycles;
5044 
5045 	switch (hw->mac.type) {
5046 	case e1000_i210:
5047 	case e1000_i211:
5048 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
5049 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5050 		tx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
5051 				* NSEC_PER_SEC;
5052 		break;
5053 	case e1000_82580:
5054 	case e1000_i350:
5055 	case e1000_i354:
5056 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5057 		/* Only the 8 LSB are valid. */
5058 		tx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_TXSTMPH)
5059 				& 0xff) << 32;
5060 		break;
5061 	default:
5062 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5063 		tx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
5064 				<< 32;
5065 		break;
5066 	}
5067 
5068 	return tx_tstamp_cycles;
5069 }
5070 
5071 static void
5072 igb_start_timecounters(struct rte_eth_dev *dev)
5073 {
5074 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5075 	struct e1000_adapter *adapter = dev->data->dev_private;
5076 	uint32_t incval = 1;
5077 	uint32_t shift = 0;
5078 	uint64_t mask = E1000_CYCLECOUNTER_MASK;
5079 
5080 	switch (hw->mac.type) {
5081 	case e1000_82580:
5082 	case e1000_i350:
5083 	case e1000_i354:
5084 		/* 32 LSB bits + 8 MSB bits = 40 bits */
5085 		mask = (1ULL << 40) - 1;
5086 		/* fall-through */
5087 	case e1000_i210:
5088 	case e1000_i211:
5089 		/*
5090 		 * Start incrementing the register
5091 		 * used to timestamp PTP packets.
5092 		 */
5093 		E1000_WRITE_REG(hw, E1000_TIMINCA, incval);
5094 		break;
5095 	case e1000_82576:
5096 		incval = E1000_INCVALUE_82576;
5097 		shift = IGB_82576_TSYNC_SHIFT;
5098 		E1000_WRITE_REG(hw, E1000_TIMINCA,
5099 				E1000_INCPERIOD_82576 | incval);
5100 		break;
5101 	default:
5102 		/* Not supported */
5103 		return;
5104 	}
5105 
5106 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
5107 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
5108 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
5109 
5110 	adapter->systime_tc.cc_mask = mask;
5111 	adapter->systime_tc.cc_shift = shift;
5112 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
5113 
5114 	adapter->rx_tstamp_tc.cc_mask = mask;
5115 	adapter->rx_tstamp_tc.cc_shift = shift;
5116 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
5117 
5118 	adapter->tx_tstamp_tc.cc_mask = mask;
5119 	adapter->tx_tstamp_tc.cc_shift = shift;
5120 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
5121 }
5122 
5123 static int
5124 igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
5125 {
5126 	struct e1000_adapter *adapter = dev->data->dev_private;
5127 
5128 	adapter->systime_tc.nsec += delta;
5129 	adapter->rx_tstamp_tc.nsec += delta;
5130 	adapter->tx_tstamp_tc.nsec += delta;
5131 
5132 	return 0;
5133 }
5134 
5135 static int
5136 igb_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
5137 {
5138 	uint64_t ns;
5139 	struct e1000_adapter *adapter = dev->data->dev_private;
5140 
5141 	ns = rte_timespec_to_ns(ts);
5142 
5143 	/* Set the timecounters to a new value. */
5144 	adapter->systime_tc.nsec = ns;
5145 	adapter->rx_tstamp_tc.nsec = ns;
5146 	adapter->tx_tstamp_tc.nsec = ns;
5147 
5148 	return 0;
5149 }
5150 
5151 static int
5152 igb_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
5153 {
5154 	uint64_t ns, systime_cycles;
5155 	struct e1000_adapter *adapter = dev->data->dev_private;
5156 
5157 	systime_cycles = igb_read_systime_cyclecounter(dev);
5158 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
5159 	*ts = rte_ns_to_timespec(ns);
5160 
5161 	return 0;
5162 }
5163 
5164 static int
5165 igb_timesync_enable(struct rte_eth_dev *dev)
5166 {
5167 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5168 	uint32_t tsync_ctl;
5169 	uint32_t tsauxc;
5170 
5171 	/* Stop the timesync system time. */
5172 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0x0);
5173 	/* Reset the timesync system time value. */
5174 	switch (hw->mac.type) {
5175 	case e1000_82580:
5176 	case e1000_i350:
5177 	case e1000_i354:
5178 	case e1000_i210:
5179 	case e1000_i211:
5180 		E1000_WRITE_REG(hw, E1000_SYSTIMR, 0x0);
5181 		/* fall-through */
5182 	case e1000_82576:
5183 		E1000_WRITE_REG(hw, E1000_SYSTIML, 0x0);
5184 		E1000_WRITE_REG(hw, E1000_SYSTIMH, 0x0);
5185 		break;
5186 	default:
5187 		/* Not supported. */
5188 		return -ENOTSUP;
5189 	}
5190 
5191 	/* Enable system time for it isn't on by default. */
5192 	tsauxc = E1000_READ_REG(hw, E1000_TSAUXC);
5193 	tsauxc &= ~E1000_TSAUXC_DISABLE_SYSTIME;
5194 	E1000_WRITE_REG(hw, E1000_TSAUXC, tsauxc);
5195 
5196 	igb_start_timecounters(dev);
5197 
5198 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5199 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588),
5200 			(RTE_ETHER_TYPE_1588 |
5201 			 E1000_ETQF_FILTER_ENABLE |
5202 			 E1000_ETQF_1588));
5203 
5204 	/* Enable timestamping of received PTP packets. */
5205 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5206 	tsync_ctl |= E1000_TSYNCRXCTL_ENABLED;
5207 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
5208 
5209 	/* Enable Timestamping of transmitted PTP packets. */
5210 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5211 	tsync_ctl |= E1000_TSYNCTXCTL_ENABLED;
5212 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
5213 
5214 	return 0;
5215 }
5216 
5217 static int
5218 igb_timesync_disable(struct rte_eth_dev *dev)
5219 {
5220 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5221 	uint32_t tsync_ctl;
5222 
5223 	/* Disable timestamping of transmitted PTP packets. */
5224 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5225 	tsync_ctl &= ~E1000_TSYNCTXCTL_ENABLED;
5226 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
5227 
5228 	/* Disable timestamping of received PTP packets. */
5229 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5230 	tsync_ctl &= ~E1000_TSYNCRXCTL_ENABLED;
5231 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
5232 
5233 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5234 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588), 0);
5235 
5236 	/* Stop incrementating the System Time registers. */
5237 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0);
5238 
5239 	return 0;
5240 }
5241 
5242 static int
5243 igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
5244 			       struct timespec *timestamp,
5245 			       uint32_t flags __rte_unused)
5246 {
5247 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5248 	struct e1000_adapter *adapter = dev->data->dev_private;
5249 	uint32_t tsync_rxctl;
5250 	uint64_t rx_tstamp_cycles;
5251 	uint64_t ns;
5252 
5253 	tsync_rxctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5254 	if ((tsync_rxctl & E1000_TSYNCRXCTL_VALID) == 0)
5255 		return -EINVAL;
5256 
5257 	rx_tstamp_cycles = igb_read_rx_tstamp_cyclecounter(dev);
5258 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
5259 	*timestamp = rte_ns_to_timespec(ns);
5260 
5261 	return  0;
5262 }
5263 
5264 static int
5265 igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
5266 			       struct timespec *timestamp)
5267 {
5268 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5269 	struct e1000_adapter *adapter = dev->data->dev_private;
5270 	uint32_t tsync_txctl;
5271 	uint64_t tx_tstamp_cycles;
5272 	uint64_t ns;
5273 
5274 	tsync_txctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5275 	if ((tsync_txctl & E1000_TSYNCTXCTL_VALID) == 0)
5276 		return -EINVAL;
5277 
5278 	tx_tstamp_cycles = igb_read_tx_tstamp_cyclecounter(dev);
5279 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
5280 	*timestamp = rte_ns_to_timespec(ns);
5281 
5282 	return  0;
5283 }
5284 
5285 static int
5286 eth_igb_get_reg_length(struct rte_eth_dev *dev __rte_unused)
5287 {
5288 	int count = 0;
5289 	int g_ind = 0;
5290 	const struct reg_info *reg_group;
5291 
5292 	while ((reg_group = igb_regs[g_ind++]))
5293 		count += igb_reg_group_count(reg_group);
5294 
5295 	return count;
5296 }
5297 
5298 static int
5299 igbvf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
5300 {
5301 	int count = 0;
5302 	int g_ind = 0;
5303 	const struct reg_info *reg_group;
5304 
5305 	while ((reg_group = igbvf_regs[g_ind++]))
5306 		count += igb_reg_group_count(reg_group);
5307 
5308 	return count;
5309 }
5310 
5311 static int
5312 eth_igb_get_regs(struct rte_eth_dev *dev,
5313 	struct rte_dev_reg_info *regs)
5314 {
5315 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5316 	uint32_t *data = regs->data;
5317 	int g_ind = 0;
5318 	int count = 0;
5319 	const struct reg_info *reg_group;
5320 
5321 	if (data == NULL) {
5322 		regs->length = eth_igb_get_reg_length(dev);
5323 		regs->width = sizeof(uint32_t);
5324 		return 0;
5325 	}
5326 
5327 	/* Support only full register dump */
5328 	if ((regs->length == 0) ||
5329 	    (regs->length == (uint32_t)eth_igb_get_reg_length(dev))) {
5330 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5331 			hw->device_id;
5332 		while ((reg_group = igb_regs[g_ind++]))
5333 			count += igb_read_regs_group(dev, &data[count],
5334 							reg_group);
5335 		return 0;
5336 	}
5337 
5338 	return -ENOTSUP;
5339 }
5340 
5341 static int
5342 igbvf_get_regs(struct rte_eth_dev *dev,
5343 	struct rte_dev_reg_info *regs)
5344 {
5345 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5346 	uint32_t *data = regs->data;
5347 	int g_ind = 0;
5348 	int count = 0;
5349 	const struct reg_info *reg_group;
5350 
5351 	if (data == NULL) {
5352 		regs->length = igbvf_get_reg_length(dev);
5353 		regs->width = sizeof(uint32_t);
5354 		return 0;
5355 	}
5356 
5357 	/* Support only full register dump */
5358 	if ((regs->length == 0) ||
5359 	    (regs->length == (uint32_t)igbvf_get_reg_length(dev))) {
5360 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5361 			hw->device_id;
5362 		while ((reg_group = igbvf_regs[g_ind++]))
5363 			count += igb_read_regs_group(dev, &data[count],
5364 							reg_group);
5365 		return 0;
5366 	}
5367 
5368 	return -ENOTSUP;
5369 }
5370 
5371 static int
5372 eth_igb_get_eeprom_length(struct rte_eth_dev *dev)
5373 {
5374 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5375 
5376 	/* Return unit is byte count */
5377 	return hw->nvm.word_size * 2;
5378 }
5379 
5380 static int
5381 eth_igb_get_eeprom(struct rte_eth_dev *dev,
5382 	struct rte_dev_eeprom_info *in_eeprom)
5383 {
5384 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5385 	struct e1000_nvm_info *nvm = &hw->nvm;
5386 	uint16_t *data = in_eeprom->data;
5387 	int first, length;
5388 
5389 	first = in_eeprom->offset >> 1;
5390 	length = in_eeprom->length >> 1;
5391 	if ((first >= hw->nvm.word_size) ||
5392 	    ((first + length) >= hw->nvm.word_size))
5393 		return -EINVAL;
5394 
5395 	in_eeprom->magic = hw->vendor_id |
5396 		((uint32_t)hw->device_id << 16);
5397 
5398 	if ((nvm->ops.read) == NULL)
5399 		return -ENOTSUP;
5400 
5401 	return nvm->ops.read(hw, first, length, data);
5402 }
5403 
5404 static int
5405 eth_igb_set_eeprom(struct rte_eth_dev *dev,
5406 	struct rte_dev_eeprom_info *in_eeprom)
5407 {
5408 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5409 	struct e1000_nvm_info *nvm = &hw->nvm;
5410 	uint16_t *data = in_eeprom->data;
5411 	int first, length;
5412 
5413 	first = in_eeprom->offset >> 1;
5414 	length = in_eeprom->length >> 1;
5415 	if ((first >= hw->nvm.word_size) ||
5416 	    ((first + length) >= hw->nvm.word_size))
5417 		return -EINVAL;
5418 
5419 	in_eeprom->magic = (uint32_t)hw->vendor_id |
5420 		((uint32_t)hw->device_id << 16);
5421 
5422 	if ((nvm->ops.write) == NULL)
5423 		return -ENOTSUP;
5424 	return nvm->ops.write(hw,  first, length, data);
5425 }
5426 
5427 static int
5428 eth_igb_get_module_info(struct rte_eth_dev *dev,
5429 			struct rte_eth_dev_module_info *modinfo)
5430 {
5431 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5432 
5433 	uint32_t status = 0;
5434 	uint16_t sff8472_rev, addr_mode;
5435 	bool page_swap = false;
5436 
5437 	if (hw->phy.media_type == e1000_media_type_copper ||
5438 	    hw->phy.media_type == e1000_media_type_unknown)
5439 		return -EOPNOTSUPP;
5440 
5441 	/* Check whether we support SFF-8472 or not */
5442 	status = e1000_read_phy_reg_i2c(hw, IGB_SFF_8472_COMP, &sff8472_rev);
5443 	if (status)
5444 		return -EIO;
5445 
5446 	/* addressing mode is not supported */
5447 	status = e1000_read_phy_reg_i2c(hw, IGB_SFF_8472_SWAP, &addr_mode);
5448 	if (status)
5449 		return -EIO;
5450 
5451 	/* addressing mode is not supported */
5452 	if ((addr_mode & 0xFF) & IGB_SFF_ADDRESSING_MODE) {
5453 		PMD_DRV_LOG(ERR,
5454 			    "Address change required to access page 0xA2, "
5455 			    "but not supported. Please report the module "
5456 			    "type to the driver maintainers.\n");
5457 		page_swap = true;
5458 	}
5459 
5460 	if ((sff8472_rev & 0xFF) == IGB_SFF_8472_UNSUP || page_swap) {
5461 		/* We have an SFP, but it does not support SFF-8472 */
5462 		modinfo->type = RTE_ETH_MODULE_SFF_8079;
5463 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
5464 	} else {
5465 		/* We have an SFP which supports a revision of SFF-8472 */
5466 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
5467 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
5468 	}
5469 
5470 	return 0;
5471 }
5472 
5473 static int
5474 eth_igb_get_module_eeprom(struct rte_eth_dev *dev,
5475 			  struct rte_dev_eeprom_info *info)
5476 {
5477 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5478 
5479 	uint32_t status = 0;
5480 	uint16_t dataword[RTE_ETH_MODULE_SFF_8472_LEN / 2 + 1];
5481 	u16 first_word, last_word;
5482 	int i = 0;
5483 
5484 	if (info->length == 0)
5485 		return -EINVAL;
5486 
5487 	first_word = info->offset >> 1;
5488 	last_word = (info->offset + info->length - 1) >> 1;
5489 
5490 	/* Read EEPROM block, SFF-8079/SFF-8472, word at a time */
5491 	for (i = 0; i < last_word - first_word + 1; i++) {
5492 		status = e1000_read_phy_reg_i2c(hw, (first_word + i) * 2,
5493 						&dataword[i]);
5494 		if (status) {
5495 			/* Error occurred while reading module */
5496 			return -EIO;
5497 		}
5498 
5499 		dataword[i] = rte_be_to_cpu_16(dataword[i]);
5500 	}
5501 
5502 	memcpy(info->data, (u8 *)dataword + (info->offset & 1), info->length);
5503 
5504 	return 0;
5505 }
5506 
5507 static int
5508 eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5509 {
5510 	struct e1000_hw *hw =
5511 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5512 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5513 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5514 	uint32_t vec = E1000_MISC_VEC_ID;
5515 
5516 	if (rte_intr_allow_others(intr_handle))
5517 		vec = E1000_RX_VEC_START;
5518 
5519 	uint32_t mask = 1 << (queue_id + vec);
5520 
5521 	E1000_WRITE_REG(hw, E1000_EIMC, mask);
5522 	E1000_WRITE_FLUSH(hw);
5523 
5524 	return 0;
5525 }
5526 
5527 static int
5528 eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5529 {
5530 	struct e1000_hw *hw =
5531 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5532 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5533 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5534 	uint32_t vec = E1000_MISC_VEC_ID;
5535 
5536 	if (rte_intr_allow_others(intr_handle))
5537 		vec = E1000_RX_VEC_START;
5538 
5539 	uint32_t mask = 1 << (queue_id + vec);
5540 	uint32_t regval;
5541 
5542 	regval = E1000_READ_REG(hw, E1000_EIMS);
5543 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
5544 	E1000_WRITE_FLUSH(hw);
5545 
5546 	rte_intr_ack(intr_handle);
5547 
5548 	return 0;
5549 }
5550 
5551 static void
5552 eth_igb_write_ivar(struct e1000_hw *hw, uint8_t  msix_vector,
5553 		   uint8_t index, uint8_t offset)
5554 {
5555 	uint32_t val = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
5556 
5557 	/* clear bits */
5558 	val &= ~((uint32_t)0xFF << offset);
5559 
5560 	/* write vector and valid bit */
5561 	val |= (msix_vector | E1000_IVAR_VALID) << offset;
5562 
5563 	E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, val);
5564 }
5565 
5566 static void
5567 eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
5568 			   uint8_t queue, uint8_t msix_vector)
5569 {
5570 	uint32_t tmp = 0;
5571 
5572 	if (hw->mac.type == e1000_82575) {
5573 		if (direction == 0)
5574 			tmp = E1000_EICR_RX_QUEUE0 << queue;
5575 		else if (direction == 1)
5576 			tmp = E1000_EICR_TX_QUEUE0 << queue;
5577 		E1000_WRITE_REG(hw, E1000_MSIXBM(msix_vector), tmp);
5578 	} else if (hw->mac.type == e1000_82576) {
5579 		if ((direction == 0) || (direction == 1))
5580 			eth_igb_write_ivar(hw, msix_vector, queue & 0x7,
5581 					   ((queue & 0x8) << 1) +
5582 					   8 * direction);
5583 	} else if ((hw->mac.type == e1000_82580) ||
5584 			(hw->mac.type == e1000_i350) ||
5585 			(hw->mac.type == e1000_i354) ||
5586 			(hw->mac.type == e1000_i210) ||
5587 			(hw->mac.type == e1000_i211)) {
5588 		if ((direction == 0) || (direction == 1))
5589 			eth_igb_write_ivar(hw, msix_vector,
5590 					   queue >> 1,
5591 					   ((queue & 0x1) << 4) +
5592 					   8 * direction);
5593 	}
5594 }
5595 
5596 /* Sets up the hardware to generate MSI-X interrupts properly
5597  * @hw
5598  *  board private structure
5599  */
5600 static void
5601 eth_igb_configure_msix_intr(struct rte_eth_dev *dev)
5602 {
5603 	int queue_id;
5604 	uint32_t tmpval, regval, intr_mask;
5605 	struct e1000_hw *hw =
5606 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5607 	uint32_t vec = E1000_MISC_VEC_ID;
5608 	uint32_t base = E1000_MISC_VEC_ID;
5609 	uint32_t misc_shift = 0;
5610 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5611 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5612 
5613 	/* won't configure msix register if no mapping is done
5614 	 * between intr vector and event fd
5615 	 */
5616 	if (!rte_intr_dp_is_en(intr_handle))
5617 		return;
5618 
5619 	if (rte_intr_allow_others(intr_handle)) {
5620 		vec = base = E1000_RX_VEC_START;
5621 		misc_shift = 1;
5622 	}
5623 
5624 	/* set interrupt vector for other causes */
5625 	if (hw->mac.type == e1000_82575) {
5626 		tmpval = E1000_READ_REG(hw, E1000_CTRL_EXT);
5627 		/* enable MSI-X PBA support */
5628 		tmpval |= E1000_CTRL_EXT_PBA_CLR;
5629 
5630 		/* Auto-Mask interrupts upon ICR read */
5631 		tmpval |= E1000_CTRL_EXT_EIAME;
5632 		tmpval |= E1000_CTRL_EXT_IRCA;
5633 
5634 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmpval);
5635 
5636 		/* enable msix_other interrupt */
5637 		E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0), 0, E1000_EIMS_OTHER);
5638 		regval = E1000_READ_REG(hw, E1000_EIAC);
5639 		E1000_WRITE_REG(hw, E1000_EIAC, regval | E1000_EIMS_OTHER);
5640 		regval = E1000_READ_REG(hw, E1000_EIAM);
5641 		E1000_WRITE_REG(hw, E1000_EIMS, regval | E1000_EIMS_OTHER);
5642 	} else if ((hw->mac.type == e1000_82576) ||
5643 			(hw->mac.type == e1000_82580) ||
5644 			(hw->mac.type == e1000_i350) ||
5645 			(hw->mac.type == e1000_i354) ||
5646 			(hw->mac.type == e1000_i210) ||
5647 			(hw->mac.type == e1000_i211)) {
5648 		/* turn on MSI-X capability first */
5649 		E1000_WRITE_REG(hw, E1000_GPIE, E1000_GPIE_MSIX_MODE |
5650 					E1000_GPIE_PBA | E1000_GPIE_EIAME |
5651 					E1000_GPIE_NSICR);
5652 		intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5653 			misc_shift;
5654 
5655 		if (dev->data->dev_conf.intr_conf.lsc != 0)
5656 			intr_mask |= (1 << IGB_MSIX_OTHER_INTR_VEC);
5657 
5658 		regval = E1000_READ_REG(hw, E1000_EIAC);
5659 		E1000_WRITE_REG(hw, E1000_EIAC, regval | intr_mask);
5660 
5661 		/* enable msix_other interrupt */
5662 		regval = E1000_READ_REG(hw, E1000_EIMS);
5663 		E1000_WRITE_REG(hw, E1000_EIMS, regval | intr_mask);
5664 		tmpval = (IGB_MSIX_OTHER_INTR_VEC | E1000_IVAR_VALID) << 8;
5665 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, tmpval);
5666 	}
5667 
5668 	/* use EIAM to auto-mask when MSI-X interrupt
5669 	 * is asserted, this saves a register write for every interrupt
5670 	 */
5671 	intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5672 		misc_shift;
5673 
5674 	if (dev->data->dev_conf.intr_conf.lsc != 0)
5675 		intr_mask |= (1 << IGB_MSIX_OTHER_INTR_VEC);
5676 
5677 	regval = E1000_READ_REG(hw, E1000_EIAM);
5678 	E1000_WRITE_REG(hw, E1000_EIAM, regval | intr_mask);
5679 
5680 	for (queue_id = 0; queue_id < dev->data->nb_rx_queues; queue_id++) {
5681 		eth_igb_assign_msix_vector(hw, 0, queue_id, vec);
5682 		intr_handle->intr_vec[queue_id] = vec;
5683 		if (vec < base + intr_handle->nb_efd - 1)
5684 			vec++;
5685 	}
5686 
5687 	E1000_WRITE_FLUSH(hw);
5688 }
5689 
5690 /* restore n-tuple filter */
5691 static inline void
5692 igb_ntuple_filter_restore(struct rte_eth_dev *dev)
5693 {
5694 	struct e1000_filter_info *filter_info =
5695 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5696 	struct e1000_5tuple_filter *p_5tuple;
5697 	struct e1000_2tuple_filter *p_2tuple;
5698 
5699 	TAILQ_FOREACH(p_5tuple, &filter_info->fivetuple_list, entries) {
5700 		igb_inject_5tuple_filter_82576(dev, p_5tuple);
5701 	}
5702 
5703 	TAILQ_FOREACH(p_2tuple, &filter_info->twotuple_list, entries) {
5704 		igb_inject_2uple_filter(dev, p_2tuple);
5705 	}
5706 }
5707 
5708 /* restore SYN filter */
5709 static inline void
5710 igb_syn_filter_restore(struct rte_eth_dev *dev)
5711 {
5712 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5713 	struct e1000_filter_info *filter_info =
5714 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5715 	uint32_t synqf;
5716 
5717 	synqf = filter_info->syn_info;
5718 
5719 	if (synqf & E1000_SYN_FILTER_ENABLE) {
5720 		E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
5721 		E1000_WRITE_FLUSH(hw);
5722 	}
5723 }
5724 
5725 /* restore ethernet type filter */
5726 static inline void
5727 igb_ethertype_filter_restore(struct rte_eth_dev *dev)
5728 {
5729 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5730 	struct e1000_filter_info *filter_info =
5731 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5732 	int i;
5733 
5734 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
5735 		if (filter_info->ethertype_mask & (1 << i)) {
5736 			E1000_WRITE_REG(hw, E1000_ETQF(i),
5737 				filter_info->ethertype_filters[i].etqf);
5738 			E1000_WRITE_FLUSH(hw);
5739 		}
5740 	}
5741 }
5742 
5743 /* restore flex byte filter */
5744 static inline void
5745 igb_flex_filter_restore(struct rte_eth_dev *dev)
5746 {
5747 	struct e1000_filter_info *filter_info =
5748 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5749 	struct e1000_flex_filter *flex_filter;
5750 
5751 	TAILQ_FOREACH(flex_filter, &filter_info->flex_list, entries) {
5752 		igb_inject_flex_filter(dev, flex_filter);
5753 	}
5754 }
5755 
5756 /* restore rss filter */
5757 static inline void
5758 igb_rss_filter_restore(struct rte_eth_dev *dev)
5759 {
5760 	struct e1000_filter_info *filter_info =
5761 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5762 
5763 	if (filter_info->rss_info.conf.queue_num)
5764 		igb_config_rss_filter(dev, &filter_info->rss_info, TRUE);
5765 }
5766 
5767 /* restore all types filter */
5768 static int
5769 igb_filter_restore(struct rte_eth_dev *dev)
5770 {
5771 	igb_ntuple_filter_restore(dev);
5772 	igb_ethertype_filter_restore(dev);
5773 	igb_syn_filter_restore(dev);
5774 	igb_flex_filter_restore(dev);
5775 	igb_rss_filter_restore(dev);
5776 
5777 	return 0;
5778 }
5779 
5780 RTE_PMD_REGISTER_PCI(net_e1000_igb, rte_igb_pmd);
5781 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb, pci_id_igb_map);
5782 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb, "* igb_uio | uio_pci_generic | vfio-pci");
5783 RTE_PMD_REGISTER_PCI(net_e1000_igb_vf, rte_igbvf_pmd);
5784 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb_vf, pci_id_igbvf_map);
5785 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb_vf, "* igb_uio | vfio-pci");
5786 
5787 /* see e1000_logs.c */
5788 RTE_INIT(e1000_init_log)
5789 {
5790 	e1000_igb_init_log();
5791 }
5792