xref: /dpdk/drivers/net/e1000/igb_ethdev.c (revision c14ef1ec)
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 	.tx_done_cleanup      = eth_igb_tx_done_cleanup,
450 	.set_mc_addr_list     = eth_igb_set_mc_addr_list,
451 	.rxq_info_get         = igb_rxq_info_get,
452 	.txq_info_get         = igb_txq_info_get,
453 	.mac_addr_set         = igbvf_default_mac_addr_set,
454 	.get_reg              = igbvf_get_regs,
455 };
456 
457 /* store statistics names and its offset in stats structure */
458 struct rte_igb_xstats_name_off {
459 	char name[RTE_ETH_XSTATS_NAME_SIZE];
460 	unsigned offset;
461 };
462 
463 static const struct rte_igb_xstats_name_off rte_igb_stats_strings[] = {
464 	{"rx_crc_errors", offsetof(struct e1000_hw_stats, crcerrs)},
465 	{"rx_align_errors", offsetof(struct e1000_hw_stats, algnerrc)},
466 	{"rx_symbol_errors", offsetof(struct e1000_hw_stats, symerrs)},
467 	{"rx_missed_packets", offsetof(struct e1000_hw_stats, mpc)},
468 	{"tx_single_collision_packets", offsetof(struct e1000_hw_stats, scc)},
469 	{"tx_multiple_collision_packets", offsetof(struct e1000_hw_stats, mcc)},
470 	{"tx_excessive_collision_packets", offsetof(struct e1000_hw_stats,
471 		ecol)},
472 	{"tx_late_collisions", offsetof(struct e1000_hw_stats, latecol)},
473 	{"tx_total_collisions", offsetof(struct e1000_hw_stats, colc)},
474 	{"tx_deferred_packets", offsetof(struct e1000_hw_stats, dc)},
475 	{"tx_no_carrier_sense_packets", offsetof(struct e1000_hw_stats, tncrs)},
476 	{"rx_carrier_ext_errors", offsetof(struct e1000_hw_stats, cexterr)},
477 	{"rx_length_errors", offsetof(struct e1000_hw_stats, rlec)},
478 	{"rx_xon_packets", offsetof(struct e1000_hw_stats, xonrxc)},
479 	{"tx_xon_packets", offsetof(struct e1000_hw_stats, xontxc)},
480 	{"rx_xoff_packets", offsetof(struct e1000_hw_stats, xoffrxc)},
481 	{"tx_xoff_packets", offsetof(struct e1000_hw_stats, xofftxc)},
482 	{"rx_flow_control_unsupported_packets", offsetof(struct e1000_hw_stats,
483 		fcruc)},
484 	{"rx_size_64_packets", offsetof(struct e1000_hw_stats, prc64)},
485 	{"rx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, prc127)},
486 	{"rx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, prc255)},
487 	{"rx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, prc511)},
488 	{"rx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
489 		prc1023)},
490 	{"rx_size_1024_to_max_packets", offsetof(struct e1000_hw_stats,
491 		prc1522)},
492 	{"rx_broadcast_packets", offsetof(struct e1000_hw_stats, bprc)},
493 	{"rx_multicast_packets", offsetof(struct e1000_hw_stats, mprc)},
494 	{"rx_undersize_errors", offsetof(struct e1000_hw_stats, ruc)},
495 	{"rx_fragment_errors", offsetof(struct e1000_hw_stats, rfc)},
496 	{"rx_oversize_errors", offsetof(struct e1000_hw_stats, roc)},
497 	{"rx_jabber_errors", offsetof(struct e1000_hw_stats, rjc)},
498 	{"rx_management_packets", offsetof(struct e1000_hw_stats, mgprc)},
499 	{"rx_management_dropped", offsetof(struct e1000_hw_stats, mgpdc)},
500 	{"tx_management_packets", offsetof(struct e1000_hw_stats, mgptc)},
501 	{"rx_total_packets", offsetof(struct e1000_hw_stats, tpr)},
502 	{"tx_total_packets", offsetof(struct e1000_hw_stats, tpt)},
503 	{"rx_total_bytes", offsetof(struct e1000_hw_stats, tor)},
504 	{"tx_total_bytes", offsetof(struct e1000_hw_stats, tot)},
505 	{"tx_size_64_packets", offsetof(struct e1000_hw_stats, ptc64)},
506 	{"tx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, ptc127)},
507 	{"tx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, ptc255)},
508 	{"tx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, ptc511)},
509 	{"tx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
510 		ptc1023)},
511 	{"tx_size_1023_to_max_packets", offsetof(struct e1000_hw_stats,
512 		ptc1522)},
513 	{"tx_multicast_packets", offsetof(struct e1000_hw_stats, mptc)},
514 	{"tx_broadcast_packets", offsetof(struct e1000_hw_stats, bptc)},
515 	{"tx_tso_packets", offsetof(struct e1000_hw_stats, tsctc)},
516 	{"tx_tso_errors", offsetof(struct e1000_hw_stats, tsctfc)},
517 	{"rx_sent_to_host_packets", offsetof(struct e1000_hw_stats, rpthc)},
518 	{"tx_sent_by_host_packets", offsetof(struct e1000_hw_stats, hgptc)},
519 	{"rx_code_violation_packets", offsetof(struct e1000_hw_stats, scvpc)},
520 
521 	{"interrupt_assert_count", offsetof(struct e1000_hw_stats, iac)},
522 };
523 
524 #define IGB_NB_XSTATS (sizeof(rte_igb_stats_strings) / \
525 		sizeof(rte_igb_stats_strings[0]))
526 
527 static const struct rte_igb_xstats_name_off rte_igbvf_stats_strings[] = {
528 	{"rx_multicast_packets", offsetof(struct e1000_vf_stats, mprc)},
529 	{"rx_good_loopback_packets", offsetof(struct e1000_vf_stats, gprlbc)},
530 	{"tx_good_loopback_packets", offsetof(struct e1000_vf_stats, gptlbc)},
531 	{"rx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gorlbc)},
532 	{"tx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gotlbc)},
533 };
534 
535 #define IGBVF_NB_XSTATS (sizeof(rte_igbvf_stats_strings) / \
536 		sizeof(rte_igbvf_stats_strings[0]))
537 
538 
539 static inline void
540 igb_intr_enable(struct rte_eth_dev *dev)
541 {
542 	struct e1000_interrupt *intr =
543 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
544 	struct e1000_hw *hw =
545 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
546 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
547 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
548 
549 	if (rte_intr_allow_others(intr_handle) &&
550 		dev->data->dev_conf.intr_conf.lsc != 0) {
551 		E1000_WRITE_REG(hw, E1000_EIMS, 1 << IGB_MSIX_OTHER_INTR_VEC);
552 	}
553 
554 	E1000_WRITE_REG(hw, E1000_IMS, intr->mask);
555 	E1000_WRITE_FLUSH(hw);
556 }
557 
558 static void
559 igb_intr_disable(struct rte_eth_dev *dev)
560 {
561 	struct e1000_hw *hw =
562 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
563 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
564 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
565 
566 	if (rte_intr_allow_others(intr_handle) &&
567 		dev->data->dev_conf.intr_conf.lsc != 0) {
568 		E1000_WRITE_REG(hw, E1000_EIMC, 1 << IGB_MSIX_OTHER_INTR_VEC);
569 	}
570 
571 	E1000_WRITE_REG(hw, E1000_IMC, ~0);
572 	E1000_WRITE_FLUSH(hw);
573 }
574 
575 static inline void
576 igbvf_intr_enable(struct rte_eth_dev *dev)
577 {
578 	struct e1000_hw *hw =
579 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
580 
581 	/* only for mailbox */
582 	E1000_WRITE_REG(hw, E1000_EIAM, 1 << E1000_VTIVAR_MISC_MAILBOX);
583 	E1000_WRITE_REG(hw, E1000_EIAC, 1 << E1000_VTIVAR_MISC_MAILBOX);
584 	E1000_WRITE_REG(hw, E1000_EIMS, 1 << E1000_VTIVAR_MISC_MAILBOX);
585 	E1000_WRITE_FLUSH(hw);
586 }
587 
588 /* only for mailbox now. If RX/TX needed, should extend this function.  */
589 static void
590 igbvf_set_ivar_map(struct e1000_hw *hw, uint8_t msix_vector)
591 {
592 	uint32_t tmp = 0;
593 
594 	/* mailbox */
595 	tmp |= (msix_vector & E1000_VTIVAR_MISC_INTR_MASK);
596 	tmp |= E1000_VTIVAR_VALID;
597 	E1000_WRITE_REG(hw, E1000_VTIVAR_MISC, tmp);
598 }
599 
600 static void
601 eth_igbvf_configure_msix_intr(struct rte_eth_dev *dev)
602 {
603 	struct e1000_hw *hw =
604 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
605 
606 	/* Configure VF other cause ivar */
607 	igbvf_set_ivar_map(hw, E1000_VTIVAR_MISC_MAILBOX);
608 }
609 
610 static inline int32_t
611 igb_pf_reset_hw(struct e1000_hw *hw)
612 {
613 	uint32_t ctrl_ext;
614 	int32_t status;
615 
616 	status = e1000_reset_hw(hw);
617 
618 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
619 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
620 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
621 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
622 	E1000_WRITE_FLUSH(hw);
623 
624 	return status;
625 }
626 
627 static void
628 igb_identify_hardware(struct rte_eth_dev *dev, struct rte_pci_device *pci_dev)
629 {
630 	struct e1000_hw *hw =
631 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
632 
633 
634 	hw->vendor_id = pci_dev->id.vendor_id;
635 	hw->device_id = pci_dev->id.device_id;
636 	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
637 	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
638 
639 	e1000_set_mac_type(hw);
640 
641 	/* need to check if it is a vf device below */
642 }
643 
644 static int
645 igb_reset_swfw_lock(struct e1000_hw *hw)
646 {
647 	int ret_val;
648 
649 	/*
650 	 * Do mac ops initialization manually here, since we will need
651 	 * some function pointers set by this call.
652 	 */
653 	ret_val = e1000_init_mac_params(hw);
654 	if (ret_val)
655 		return ret_val;
656 
657 	/*
658 	 * SMBI lock should not fail in this early stage. If this is the case,
659 	 * it is due to an improper exit of the application.
660 	 * So force the release of the faulty lock.
661 	 */
662 	if (e1000_get_hw_semaphore_generic(hw) < 0) {
663 		PMD_DRV_LOG(DEBUG, "SMBI lock released");
664 	}
665 	e1000_put_hw_semaphore_generic(hw);
666 
667 	if (hw->mac.ops.acquire_swfw_sync != NULL) {
668 		uint16_t mask;
669 
670 		/*
671 		 * Phy lock should not fail in this early stage. If this is the case,
672 		 * it is due to an improper exit of the application.
673 		 * So force the release of the faulty lock.
674 		 */
675 		mask = E1000_SWFW_PHY0_SM << hw->bus.func;
676 		if (hw->bus.func > E1000_FUNC_1)
677 			mask <<= 2;
678 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
679 			PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released",
680 				    hw->bus.func);
681 		}
682 		hw->mac.ops.release_swfw_sync(hw, mask);
683 
684 		/*
685 		 * This one is more tricky since it is common to all ports; but
686 		 * swfw_sync retries last long enough (1s) to be almost sure that if
687 		 * lock can not be taken it is due to an improper lock of the
688 		 * semaphore.
689 		 */
690 		mask = E1000_SWFW_EEP_SM;
691 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
692 			PMD_DRV_LOG(DEBUG, "SWFW common locks released");
693 		}
694 		hw->mac.ops.release_swfw_sync(hw, mask);
695 	}
696 
697 	return E1000_SUCCESS;
698 }
699 
700 /* Remove all ntuple filters of the device */
701 static int igb_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
702 {
703 	struct e1000_filter_info *filter_info =
704 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
705 	struct e1000_5tuple_filter *p_5tuple;
706 	struct e1000_2tuple_filter *p_2tuple;
707 
708 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
709 		TAILQ_REMOVE(&filter_info->fivetuple_list,
710 			p_5tuple, entries);
711 			rte_free(p_5tuple);
712 	}
713 	filter_info->fivetuple_mask = 0;
714 	while ((p_2tuple = TAILQ_FIRST(&filter_info->twotuple_list))) {
715 		TAILQ_REMOVE(&filter_info->twotuple_list,
716 			p_2tuple, entries);
717 			rte_free(p_2tuple);
718 	}
719 	filter_info->twotuple_mask = 0;
720 
721 	return 0;
722 }
723 
724 /* Remove all flex filters of the device */
725 static int igb_flex_filter_uninit(struct rte_eth_dev *eth_dev)
726 {
727 	struct e1000_filter_info *filter_info =
728 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
729 	struct e1000_flex_filter *p_flex;
730 
731 	while ((p_flex = TAILQ_FIRST(&filter_info->flex_list))) {
732 		TAILQ_REMOVE(&filter_info->flex_list, p_flex, entries);
733 		rte_free(p_flex);
734 	}
735 	filter_info->flex_mask = 0;
736 
737 	return 0;
738 }
739 
740 static int
741 eth_igb_dev_init(struct rte_eth_dev *eth_dev)
742 {
743 	int error = 0;
744 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
745 	struct e1000_hw *hw =
746 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
747 	struct e1000_vfta * shadow_vfta =
748 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
749 	struct e1000_filter_info *filter_info =
750 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
751 	struct e1000_adapter *adapter =
752 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
753 
754 	uint32_t ctrl_ext;
755 
756 	eth_dev->dev_ops = &eth_igb_ops;
757 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
758 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
759 	eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
760 
761 	/* for secondary processes, we don't initialise any further as primary
762 	 * has already done this work. Only check we don't need a different
763 	 * RX function */
764 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
765 		if (eth_dev->data->scattered_rx)
766 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
767 		return 0;
768 	}
769 
770 	rte_eth_copy_pci_info(eth_dev, pci_dev);
771 
772 	hw->hw_addr= (void *)pci_dev->mem_resource[0].addr;
773 
774 	igb_identify_hardware(eth_dev, pci_dev);
775 	if (e1000_setup_init_funcs(hw, FALSE) != E1000_SUCCESS) {
776 		error = -EIO;
777 		goto err_late;
778 	}
779 
780 	e1000_get_bus_info(hw);
781 
782 	/* Reset any pending lock */
783 	if (igb_reset_swfw_lock(hw) != E1000_SUCCESS) {
784 		error = -EIO;
785 		goto err_late;
786 	}
787 
788 	/* Finish initialization */
789 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS) {
790 		error = -EIO;
791 		goto err_late;
792 	}
793 
794 	hw->mac.autoneg = 1;
795 	hw->phy.autoneg_wait_to_complete = 0;
796 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
797 
798 	/* Copper options */
799 	if (hw->phy.media_type == e1000_media_type_copper) {
800 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
801 		hw->phy.disable_polarity_correction = 0;
802 		hw->phy.ms_type = e1000_ms_hw_default;
803 	}
804 
805 	/*
806 	 * Start from a known state, this is important in reading the nvm
807 	 * and mac from that.
808 	 */
809 	igb_pf_reset_hw(hw);
810 
811 	/* Make sure we have a good EEPROM before we read from it */
812 	if (e1000_validate_nvm_checksum(hw) < 0) {
813 		/*
814 		 * Some PCI-E parts fail the first check due to
815 		 * the link being in sleep state, call it again,
816 		 * if it fails a second time its a real issue.
817 		 */
818 		if (e1000_validate_nvm_checksum(hw) < 0) {
819 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
820 			error = -EIO;
821 			goto err_late;
822 		}
823 	}
824 
825 	/* Read the permanent MAC address out of the EEPROM */
826 	if (e1000_read_mac_addr(hw) != 0) {
827 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
828 		error = -EIO;
829 		goto err_late;
830 	}
831 
832 	/* Allocate memory for storing MAC addresses */
833 	eth_dev->data->mac_addrs = rte_zmalloc("e1000",
834 		RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count, 0);
835 	if (eth_dev->data->mac_addrs == NULL) {
836 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
837 						"store MAC addresses",
838 				RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
839 		error = -ENOMEM;
840 		goto err_late;
841 	}
842 
843 	/* Copy the permanent MAC address */
844 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
845 			&eth_dev->data->mac_addrs[0]);
846 
847 	/* Pass the information to the rte_eth_dev_close() that it should also
848 	 * release the private port resources.
849 	 */
850 	eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
851 
852 	/* initialize the vfta */
853 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
854 
855 	/* Now initialize the hardware */
856 	if (igb_hardware_init(hw) != 0) {
857 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
858 		rte_free(eth_dev->data->mac_addrs);
859 		eth_dev->data->mac_addrs = NULL;
860 		error = -ENODEV;
861 		goto err_late;
862 	}
863 	hw->mac.get_link_status = 1;
864 	adapter->stopped = 0;
865 
866 	/* Indicate SOL/IDER usage */
867 	if (e1000_check_reset_block(hw) < 0) {
868 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to"
869 					"SOL/IDER session");
870 	}
871 
872 	/* initialize PF if max_vfs not zero */
873 	igb_pf_host_init(eth_dev);
874 
875 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
876 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
877 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
878 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
879 	E1000_WRITE_FLUSH(hw);
880 
881 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
882 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
883 		     pci_dev->id.device_id);
884 
885 	rte_intr_callback_register(&pci_dev->intr_handle,
886 				   eth_igb_interrupt_handler,
887 				   (void *)eth_dev);
888 
889 	/* enable uio/vfio intr/eventfd mapping */
890 	rte_intr_enable(&pci_dev->intr_handle);
891 
892 	/* enable support intr */
893 	igb_intr_enable(eth_dev);
894 
895 	eth_igb_dev_set_link_down(eth_dev);
896 
897 	/* initialize filter info */
898 	memset(filter_info, 0,
899 	       sizeof(struct e1000_filter_info));
900 
901 	TAILQ_INIT(&filter_info->flex_list);
902 	TAILQ_INIT(&filter_info->twotuple_list);
903 	TAILQ_INIT(&filter_info->fivetuple_list);
904 
905 	TAILQ_INIT(&igb_filter_ntuple_list);
906 	TAILQ_INIT(&igb_filter_ethertype_list);
907 	TAILQ_INIT(&igb_filter_syn_list);
908 	TAILQ_INIT(&igb_filter_flex_list);
909 	TAILQ_INIT(&igb_filter_rss_list);
910 	TAILQ_INIT(&igb_flow_list);
911 
912 	return 0;
913 
914 err_late:
915 	igb_hw_control_release(hw);
916 
917 	return error;
918 }
919 
920 static int
921 eth_igb_dev_uninit(struct rte_eth_dev *eth_dev)
922 {
923 	PMD_INIT_FUNC_TRACE();
924 
925 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
926 		return 0;
927 
928 	eth_igb_close(eth_dev);
929 
930 	return 0;
931 }
932 
933 /*
934  * Virtual Function device init
935  */
936 static int
937 eth_igbvf_dev_init(struct rte_eth_dev *eth_dev)
938 {
939 	struct rte_pci_device *pci_dev;
940 	struct rte_intr_handle *intr_handle;
941 	struct e1000_adapter *adapter =
942 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
943 	struct e1000_hw *hw =
944 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
945 	int diag;
946 	struct rte_ether_addr *perm_addr =
947 		(struct rte_ether_addr *)hw->mac.perm_addr;
948 
949 	PMD_INIT_FUNC_TRACE();
950 
951 	eth_dev->dev_ops = &igbvf_eth_dev_ops;
952 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
953 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
954 	eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
955 
956 	/* for secondary processes, we don't initialise any further as primary
957 	 * has already done this work. Only check we don't need a different
958 	 * RX function */
959 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
960 		if (eth_dev->data->scattered_rx)
961 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
962 		return 0;
963 	}
964 
965 	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
966 	rte_eth_copy_pci_info(eth_dev, pci_dev);
967 
968 	hw->device_id = pci_dev->id.device_id;
969 	hw->vendor_id = pci_dev->id.vendor_id;
970 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
971 	adapter->stopped = 0;
972 
973 	/* Initialize the shared code (base driver) */
974 	diag = e1000_setup_init_funcs(hw, TRUE);
975 	if (diag != 0) {
976 		PMD_INIT_LOG(ERR, "Shared code init failed for igbvf: %d",
977 			diag);
978 		return -EIO;
979 	}
980 
981 	/* init_mailbox_params */
982 	hw->mbx.ops.init_params(hw);
983 
984 	/* Disable the interrupts for VF */
985 	igbvf_intr_disable(hw);
986 
987 	diag = hw->mac.ops.reset_hw(hw);
988 
989 	/* Allocate memory for storing MAC addresses */
990 	eth_dev->data->mac_addrs = rte_zmalloc("igbvf", RTE_ETHER_ADDR_LEN *
991 		hw->mac.rar_entry_count, 0);
992 	if (eth_dev->data->mac_addrs == NULL) {
993 		PMD_INIT_LOG(ERR,
994 			"Failed to allocate %d bytes needed to store MAC "
995 			"addresses",
996 			RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
997 		return -ENOMEM;
998 	}
999 
1000 	/* Pass the information to the rte_eth_dev_close() that it should also
1001 	 * release the private port resources.
1002 	 */
1003 	eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
1004 
1005 	/* Generate a random MAC address, if none was assigned by PF. */
1006 	if (rte_is_zero_ether_addr(perm_addr)) {
1007 		rte_eth_random_addr(perm_addr->addr_bytes);
1008 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1009 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1010 			     "%02x:%02x:%02x:%02x:%02x:%02x",
1011 			     perm_addr->addr_bytes[0],
1012 			     perm_addr->addr_bytes[1],
1013 			     perm_addr->addr_bytes[2],
1014 			     perm_addr->addr_bytes[3],
1015 			     perm_addr->addr_bytes[4],
1016 			     perm_addr->addr_bytes[5]);
1017 	}
1018 
1019 	diag = e1000_rar_set(hw, perm_addr->addr_bytes, 0);
1020 	if (diag) {
1021 		rte_free(eth_dev->data->mac_addrs);
1022 		eth_dev->data->mac_addrs = NULL;
1023 		return diag;
1024 	}
1025 	/* Copy the permanent MAC address */
1026 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.perm_addr,
1027 			&eth_dev->data->mac_addrs[0]);
1028 
1029 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x "
1030 		     "mac.type=%s",
1031 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1032 		     pci_dev->id.device_id, "igb_mac_82576_vf");
1033 
1034 	intr_handle = &pci_dev->intr_handle;
1035 	rte_intr_callback_register(intr_handle,
1036 				   eth_igbvf_interrupt_handler, eth_dev);
1037 
1038 	return 0;
1039 }
1040 
1041 static int
1042 eth_igbvf_dev_uninit(struct rte_eth_dev *eth_dev)
1043 {
1044 	PMD_INIT_FUNC_TRACE();
1045 
1046 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1047 		return 0;
1048 
1049 	igbvf_dev_close(eth_dev);
1050 
1051 	return 0;
1052 }
1053 
1054 static int eth_igb_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1055 	struct rte_pci_device *pci_dev)
1056 {
1057 	return rte_eth_dev_pci_generic_probe(pci_dev,
1058 		sizeof(struct e1000_adapter), eth_igb_dev_init);
1059 }
1060 
1061 static int eth_igb_pci_remove(struct rte_pci_device *pci_dev)
1062 {
1063 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igb_dev_uninit);
1064 }
1065 
1066 static struct rte_pci_driver rte_igb_pmd = {
1067 	.id_table = pci_id_igb_map,
1068 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1069 	.probe = eth_igb_pci_probe,
1070 	.remove = eth_igb_pci_remove,
1071 };
1072 
1073 
1074 static int eth_igbvf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1075 	struct rte_pci_device *pci_dev)
1076 {
1077 	return rte_eth_dev_pci_generic_probe(pci_dev,
1078 		sizeof(struct e1000_adapter), eth_igbvf_dev_init);
1079 }
1080 
1081 static int eth_igbvf_pci_remove(struct rte_pci_device *pci_dev)
1082 {
1083 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igbvf_dev_uninit);
1084 }
1085 
1086 /*
1087  * virtual function driver struct
1088  */
1089 static struct rte_pci_driver rte_igbvf_pmd = {
1090 	.id_table = pci_id_igbvf_map,
1091 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1092 	.probe = eth_igbvf_pci_probe,
1093 	.remove = eth_igbvf_pci_remove,
1094 };
1095 
1096 static void
1097 igb_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1098 {
1099 	struct e1000_hw *hw =
1100 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1101 	/* RCTL: enable VLAN filter since VMDq always use VLAN filter */
1102 	uint32_t rctl = E1000_READ_REG(hw, E1000_RCTL);
1103 	rctl |= E1000_RCTL_VFE;
1104 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1105 }
1106 
1107 static int
1108 igb_check_mq_mode(struct rte_eth_dev *dev)
1109 {
1110 	enum rte_eth_rx_mq_mode rx_mq_mode = dev->data->dev_conf.rxmode.mq_mode;
1111 	enum rte_eth_tx_mq_mode tx_mq_mode = dev->data->dev_conf.txmode.mq_mode;
1112 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
1113 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
1114 
1115 	if ((rx_mq_mode & ETH_MQ_RX_DCB_FLAG) ||
1116 	    tx_mq_mode == ETH_MQ_TX_DCB ||
1117 	    tx_mq_mode == ETH_MQ_TX_VMDQ_DCB) {
1118 		PMD_INIT_LOG(ERR, "DCB mode is not supported.");
1119 		return -EINVAL;
1120 	}
1121 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
1122 		/* Check multi-queue mode.
1123 		 * To no break software we accept ETH_MQ_RX_NONE as this might
1124 		 * be used to turn off VLAN filter.
1125 		 */
1126 
1127 		if (rx_mq_mode == ETH_MQ_RX_NONE ||
1128 		    rx_mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1129 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
1130 			RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = 1;
1131 		} else {
1132 			/* Only support one queue on VFs.
1133 			 * RSS together with SRIOV is not supported.
1134 			 */
1135 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1136 					" wrong mq_mode rx %d.",
1137 					rx_mq_mode);
1138 			return -EINVAL;
1139 		}
1140 		/* TX mode is not used here, so mode might be ignored.*/
1141 		if (tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1142 			/* SRIOV only works in VMDq enable mode */
1143 			PMD_INIT_LOG(WARNING, "SRIOV is active,"
1144 					" TX mode %d is not supported. "
1145 					" Driver will behave as %d mode.",
1146 					tx_mq_mode, ETH_MQ_TX_VMDQ_ONLY);
1147 		}
1148 
1149 		/* check valid queue number */
1150 		if ((nb_rx_q > 1) || (nb_tx_q > 1)) {
1151 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1152 					" only support one queue on VFs.");
1153 			return -EINVAL;
1154 		}
1155 	} else {
1156 		/* To no break software that set invalid mode, only display
1157 		 * warning if invalid mode is used.
1158 		 */
1159 		if (rx_mq_mode != ETH_MQ_RX_NONE &&
1160 		    rx_mq_mode != ETH_MQ_RX_VMDQ_ONLY &&
1161 		    rx_mq_mode != ETH_MQ_RX_RSS) {
1162 			/* RSS together with VMDq not supported*/
1163 			PMD_INIT_LOG(ERR, "RX mode %d is not supported.",
1164 				     rx_mq_mode);
1165 			return -EINVAL;
1166 		}
1167 
1168 		if (tx_mq_mode != ETH_MQ_TX_NONE &&
1169 		    tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1170 			PMD_INIT_LOG(WARNING, "TX mode %d is not supported."
1171 					" Due to txmode is meaningless in this"
1172 					" driver, just ignore.",
1173 					tx_mq_mode);
1174 		}
1175 	}
1176 	return 0;
1177 }
1178 
1179 static int
1180 eth_igb_configure(struct rte_eth_dev *dev)
1181 {
1182 	struct e1000_interrupt *intr =
1183 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1184 	int ret;
1185 
1186 	PMD_INIT_FUNC_TRACE();
1187 
1188 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1189 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1190 
1191 	/* multipe queue mode checking */
1192 	ret  = igb_check_mq_mode(dev);
1193 	if (ret != 0) {
1194 		PMD_DRV_LOG(ERR, "igb_check_mq_mode fails with %d.",
1195 			    ret);
1196 		return ret;
1197 	}
1198 
1199 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1200 	PMD_INIT_FUNC_TRACE();
1201 
1202 	return 0;
1203 }
1204 
1205 static void
1206 eth_igb_rxtx_control(struct rte_eth_dev *dev,
1207 		     bool enable)
1208 {
1209 	struct e1000_hw *hw =
1210 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1211 	uint32_t tctl, rctl;
1212 
1213 	tctl = E1000_READ_REG(hw, E1000_TCTL);
1214 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1215 
1216 	if (enable) {
1217 		/* enable Tx/Rx */
1218 		tctl |= E1000_TCTL_EN;
1219 		rctl |= E1000_RCTL_EN;
1220 	} else {
1221 		/* disable Tx/Rx */
1222 		tctl &= ~E1000_TCTL_EN;
1223 		rctl &= ~E1000_RCTL_EN;
1224 	}
1225 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
1226 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1227 	E1000_WRITE_FLUSH(hw);
1228 }
1229 
1230 static int
1231 eth_igb_start(struct rte_eth_dev *dev)
1232 {
1233 	struct e1000_hw *hw =
1234 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1235 	struct e1000_adapter *adapter =
1236 		E1000_DEV_PRIVATE(dev->data->dev_private);
1237 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1238 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1239 	int ret, mask;
1240 	uint32_t intr_vector = 0;
1241 	uint32_t ctrl_ext;
1242 	uint32_t *speeds;
1243 	int num_speeds;
1244 	bool autoneg;
1245 
1246 	PMD_INIT_FUNC_TRACE();
1247 
1248 	/* disable uio/vfio intr/eventfd mapping */
1249 	rte_intr_disable(intr_handle);
1250 
1251 	/* Power up the phy. Needed to make the link go Up */
1252 	eth_igb_dev_set_link_up(dev);
1253 
1254 	/*
1255 	 * Packet Buffer Allocation (PBA)
1256 	 * Writing PBA sets the receive portion of the buffer
1257 	 * the remainder is used for the transmit buffer.
1258 	 */
1259 	if (hw->mac.type == e1000_82575) {
1260 		uint32_t pba;
1261 
1262 		pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
1263 		E1000_WRITE_REG(hw, E1000_PBA, pba);
1264 	}
1265 
1266 	/* Put the address into the Receive Address Array */
1267 	e1000_rar_set(hw, hw->mac.addr, 0);
1268 
1269 	/* Initialize the hardware */
1270 	if (igb_hardware_init(hw)) {
1271 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
1272 		return -EIO;
1273 	}
1274 	adapter->stopped = 0;
1275 
1276 	E1000_WRITE_REG(hw, E1000_VET,
1277 			RTE_ETHER_TYPE_VLAN << 16 | RTE_ETHER_TYPE_VLAN);
1278 
1279 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1280 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1281 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
1282 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
1283 	E1000_WRITE_FLUSH(hw);
1284 
1285 	/* configure PF module if SRIOV enabled */
1286 	igb_pf_host_configure(dev);
1287 
1288 	/* check and configure queue intr-vector mapping */
1289 	if ((rte_intr_cap_multiple(intr_handle) ||
1290 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
1291 	    dev->data->dev_conf.intr_conf.rxq != 0) {
1292 		intr_vector = dev->data->nb_rx_queues;
1293 		if (rte_intr_efd_enable(intr_handle, intr_vector))
1294 			return -1;
1295 	}
1296 
1297 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
1298 		intr_handle->intr_vec =
1299 			rte_zmalloc("intr_vec",
1300 				    dev->data->nb_rx_queues * sizeof(int), 0);
1301 		if (intr_handle->intr_vec == NULL) {
1302 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
1303 				     " intr_vec", dev->data->nb_rx_queues);
1304 			return -ENOMEM;
1305 		}
1306 	}
1307 
1308 	/* confiugre msix for rx interrupt */
1309 	eth_igb_configure_msix_intr(dev);
1310 
1311 	/* Configure for OS presence */
1312 	igb_init_manageability(hw);
1313 
1314 	eth_igb_tx_init(dev);
1315 
1316 	/* This can fail when allocating mbufs for descriptor rings */
1317 	ret = eth_igb_rx_init(dev);
1318 	if (ret) {
1319 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
1320 		igb_dev_clear_queues(dev);
1321 		return ret;
1322 	}
1323 
1324 	e1000_clear_hw_cntrs_base_generic(hw);
1325 
1326 	/*
1327 	 * VLAN Offload Settings
1328 	 */
1329 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
1330 			ETH_VLAN_EXTEND_MASK;
1331 	ret = eth_igb_vlan_offload_set(dev, mask);
1332 	if (ret) {
1333 		PMD_INIT_LOG(ERR, "Unable to set vlan offload");
1334 		igb_dev_clear_queues(dev);
1335 		return ret;
1336 	}
1337 
1338 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1339 		/* Enable VLAN filter since VMDq always use VLAN filter */
1340 		igb_vmdq_vlan_hw_filter_enable(dev);
1341 	}
1342 
1343 	if ((hw->mac.type == e1000_82576) || (hw->mac.type == e1000_82580) ||
1344 		(hw->mac.type == e1000_i350) || (hw->mac.type == e1000_i210) ||
1345 		(hw->mac.type == e1000_i211)) {
1346 		/* Configure EITR with the maximum possible value (0xFFFF) */
1347 		E1000_WRITE_REG(hw, E1000_EITR(0), 0xFFFF);
1348 	}
1349 
1350 	/* Setup link speed and duplex */
1351 	speeds = &dev->data->dev_conf.link_speeds;
1352 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
1353 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
1354 		hw->mac.autoneg = 1;
1355 	} else {
1356 		num_speeds = 0;
1357 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
1358 
1359 		/* Reset */
1360 		hw->phy.autoneg_advertised = 0;
1361 
1362 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1363 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1364 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
1365 			num_speeds = -1;
1366 			goto error_invalid_config;
1367 		}
1368 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
1369 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1370 			num_speeds++;
1371 		}
1372 		if (*speeds & ETH_LINK_SPEED_10M) {
1373 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1374 			num_speeds++;
1375 		}
1376 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
1377 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1378 			num_speeds++;
1379 		}
1380 		if (*speeds & ETH_LINK_SPEED_100M) {
1381 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1382 			num_speeds++;
1383 		}
1384 		if (*speeds & ETH_LINK_SPEED_1G) {
1385 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1386 			num_speeds++;
1387 		}
1388 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
1389 			goto error_invalid_config;
1390 
1391 		/* Set/reset the mac.autoneg based on the link speed,
1392 		 * fixed or not
1393 		 */
1394 		if (!autoneg) {
1395 			hw->mac.autoneg = 0;
1396 			hw->mac.forced_speed_duplex =
1397 					hw->phy.autoneg_advertised;
1398 		} else {
1399 			hw->mac.autoneg = 1;
1400 		}
1401 	}
1402 
1403 	e1000_setup_link(hw);
1404 
1405 	if (rte_intr_allow_others(intr_handle)) {
1406 		/* check if lsc interrupt is enabled */
1407 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1408 			eth_igb_lsc_interrupt_setup(dev, TRUE);
1409 		else
1410 			eth_igb_lsc_interrupt_setup(dev, FALSE);
1411 	} else {
1412 		rte_intr_callback_unregister(intr_handle,
1413 					     eth_igb_interrupt_handler,
1414 					     (void *)dev);
1415 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1416 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
1417 				     " no intr multiplex");
1418 	}
1419 
1420 	/* check if rxq interrupt is enabled */
1421 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
1422 	    rte_intr_dp_is_en(intr_handle))
1423 		eth_igb_rxq_interrupt_setup(dev);
1424 
1425 	/* enable uio/vfio intr/eventfd mapping */
1426 	rte_intr_enable(intr_handle);
1427 
1428 	/* resume enabled intr since hw reset */
1429 	igb_intr_enable(dev);
1430 
1431 	/* restore all types filter */
1432 	igb_filter_restore(dev);
1433 
1434 	eth_igb_rxtx_control(dev, true);
1435 	eth_igb_link_update(dev, 0);
1436 
1437 	PMD_INIT_LOG(DEBUG, "<<");
1438 
1439 	return 0;
1440 
1441 error_invalid_config:
1442 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
1443 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
1444 	igb_dev_clear_queues(dev);
1445 	return -EINVAL;
1446 }
1447 
1448 /*********************************************************************
1449  *
1450  *  This routine disables all traffic on the adapter by issuing a
1451  *  global reset on the MAC.
1452  *
1453  **********************************************************************/
1454 static void
1455 eth_igb_stop(struct rte_eth_dev *dev)
1456 {
1457 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1458 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1459 	struct rte_eth_link link;
1460 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1461 	struct e1000_adapter *adapter =
1462 		E1000_DEV_PRIVATE(dev->data->dev_private);
1463 
1464 	if (adapter->stopped)
1465 		return;
1466 
1467 	eth_igb_rxtx_control(dev, false);
1468 
1469 	igb_intr_disable(dev);
1470 
1471 	/* disable intr eventfd mapping */
1472 	rte_intr_disable(intr_handle);
1473 
1474 	igb_pf_reset_hw(hw);
1475 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1476 
1477 	/* Set bit for Go Link disconnect if PHY reset is not blocked */
1478 	if (hw->mac.type >= e1000_82580 &&
1479 	    (e1000_check_reset_block(hw) != E1000_BLK_PHY_RESET)) {
1480 		uint32_t phpm_reg;
1481 
1482 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1483 		phpm_reg |= E1000_82580_PM_GO_LINKD;
1484 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1485 	}
1486 
1487 	/* Power down the phy. Needed to make the link go Down */
1488 	eth_igb_dev_set_link_down(dev);
1489 
1490 	igb_dev_clear_queues(dev);
1491 
1492 	/* clear the recorded link status */
1493 	memset(&link, 0, sizeof(link));
1494 	rte_eth_linkstatus_set(dev, &link);
1495 
1496 	if (!rte_intr_allow_others(intr_handle))
1497 		/* resume to the default handler */
1498 		rte_intr_callback_register(intr_handle,
1499 					   eth_igb_interrupt_handler,
1500 					   (void *)dev);
1501 
1502 	/* Clean datapath event and queue/vec mapping */
1503 	rte_intr_efd_disable(intr_handle);
1504 	if (intr_handle->intr_vec != NULL) {
1505 		rte_free(intr_handle->intr_vec);
1506 		intr_handle->intr_vec = NULL;
1507 	}
1508 
1509 	adapter->stopped = true;
1510 }
1511 
1512 static int
1513 eth_igb_dev_set_link_up(struct rte_eth_dev *dev)
1514 {
1515 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1516 
1517 	if (hw->phy.media_type == e1000_media_type_copper)
1518 		e1000_power_up_phy(hw);
1519 	else
1520 		e1000_power_up_fiber_serdes_link(hw);
1521 
1522 	return 0;
1523 }
1524 
1525 static int
1526 eth_igb_dev_set_link_down(struct rte_eth_dev *dev)
1527 {
1528 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1529 
1530 	if (hw->phy.media_type == e1000_media_type_copper)
1531 		e1000_power_down_phy(hw);
1532 	else
1533 		e1000_shutdown_fiber_serdes_link(hw);
1534 
1535 	return 0;
1536 }
1537 
1538 static void
1539 eth_igb_close(struct rte_eth_dev *dev)
1540 {
1541 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1542 	struct rte_eth_link link;
1543 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1544 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1545 	struct e1000_filter_info *filter_info =
1546 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
1547 
1548 	eth_igb_stop(dev);
1549 
1550 	e1000_phy_hw_reset(hw);
1551 	igb_release_manageability(hw);
1552 	igb_hw_control_release(hw);
1553 
1554 	/* Clear bit for Go Link disconnect if PHY reset is not blocked */
1555 	if (hw->mac.type >= e1000_82580 &&
1556 	    (e1000_check_reset_block(hw) != E1000_BLK_PHY_RESET)) {
1557 		uint32_t phpm_reg;
1558 
1559 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1560 		phpm_reg &= ~E1000_82580_PM_GO_LINKD;
1561 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1562 	}
1563 
1564 	igb_dev_free_queues(dev);
1565 
1566 	if (intr_handle->intr_vec) {
1567 		rte_free(intr_handle->intr_vec);
1568 		intr_handle->intr_vec = NULL;
1569 	}
1570 
1571 	memset(&link, 0, sizeof(link));
1572 	rte_eth_linkstatus_set(dev, &link);
1573 
1574 	dev->dev_ops = NULL;
1575 	dev->rx_pkt_burst = NULL;
1576 	dev->tx_pkt_burst = NULL;
1577 
1578 	/* Reset any pending lock */
1579 	igb_reset_swfw_lock(hw);
1580 
1581 	/* uninitialize PF if max_vfs not zero */
1582 	igb_pf_host_uninit(dev);
1583 
1584 	rte_intr_callback_unregister(intr_handle,
1585 				     eth_igb_interrupt_handler, dev);
1586 
1587 	/* clear the SYN filter info */
1588 	filter_info->syn_info = 0;
1589 
1590 	/* clear the ethertype filters info */
1591 	filter_info->ethertype_mask = 0;
1592 	memset(filter_info->ethertype_filters, 0,
1593 		E1000_MAX_ETQF_FILTERS * sizeof(struct igb_ethertype_filter));
1594 
1595 	/* clear the rss filter info */
1596 	memset(&filter_info->rss_info, 0,
1597 		sizeof(struct igb_rte_flow_rss_conf));
1598 
1599 	/* remove all ntuple filters of the device */
1600 	igb_ntuple_filter_uninit(dev);
1601 
1602 	/* remove all flex filters of the device */
1603 	igb_flex_filter_uninit(dev);
1604 
1605 	/* clear all the filters list */
1606 	igb_filterlist_flush(dev);
1607 }
1608 
1609 /*
1610  * Reset PF device.
1611  */
1612 static int
1613 eth_igb_reset(struct rte_eth_dev *dev)
1614 {
1615 	int ret;
1616 
1617 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
1618 	 * its VF to make them align with it. The detailed notification
1619 	 * mechanism is PMD specific and is currently not implemented.
1620 	 * To avoid unexpected behavior in VF, currently reset of PF with
1621 	 * SR-IOV activation is not supported. It might be supported later.
1622 	 */
1623 	if (dev->data->sriov.active)
1624 		return -ENOTSUP;
1625 
1626 	ret = eth_igb_dev_uninit(dev);
1627 	if (ret)
1628 		return ret;
1629 
1630 	ret = eth_igb_dev_init(dev);
1631 
1632 	return ret;
1633 }
1634 
1635 
1636 static int
1637 igb_get_rx_buffer_size(struct e1000_hw *hw)
1638 {
1639 	uint32_t rx_buf_size;
1640 	if (hw->mac.type == e1000_82576) {
1641 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xffff) << 10;
1642 	} else if (hw->mac.type == e1000_82580 || hw->mac.type == e1000_i350) {
1643 		/* PBS needs to be translated according to a lookup table */
1644 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xf);
1645 		rx_buf_size = (uint32_t) e1000_rxpbs_adjust_82580(rx_buf_size);
1646 		rx_buf_size = (rx_buf_size << 10);
1647 	} else if (hw->mac.type == e1000_i210 || hw->mac.type == e1000_i211) {
1648 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0x3f) << 10;
1649 	} else {
1650 		rx_buf_size = (E1000_READ_REG(hw, E1000_PBA) & 0xffff) << 10;
1651 	}
1652 
1653 	return rx_buf_size;
1654 }
1655 
1656 /*********************************************************************
1657  *
1658  *  Initialize the hardware
1659  *
1660  **********************************************************************/
1661 static int
1662 igb_hardware_init(struct e1000_hw *hw)
1663 {
1664 	uint32_t rx_buf_size;
1665 	int diag;
1666 
1667 	/* Let the firmware know the OS is in control */
1668 	igb_hw_control_acquire(hw);
1669 
1670 	/*
1671 	 * These parameters control the automatic generation (Tx) and
1672 	 * response (Rx) to Ethernet PAUSE frames.
1673 	 * - High water mark should allow for at least two standard size (1518)
1674 	 *   frames to be received after sending an XOFF.
1675 	 * - Low water mark works best when it is very near the high water mark.
1676 	 *   This allows the receiver to restart by sending XON when it has
1677 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
1678 	 *   restart after one full frame is pulled from the buffer. There
1679 	 *   could be several smaller frames in the buffer and if so they will
1680 	 *   not trigger the XON until their total number reduces the buffer
1681 	 *   by 1500.
1682 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
1683 	 */
1684 	rx_buf_size = igb_get_rx_buffer_size(hw);
1685 
1686 	hw->fc.high_water = rx_buf_size - (RTE_ETHER_MAX_LEN * 2);
1687 	hw->fc.low_water = hw->fc.high_water - 1500;
1688 	hw->fc.pause_time = IGB_FC_PAUSE_TIME;
1689 	hw->fc.send_xon = 1;
1690 
1691 	/* Set Flow control, use the tunable location if sane */
1692 	if ((igb_fc_setting != e1000_fc_none) && (igb_fc_setting < 4))
1693 		hw->fc.requested_mode = igb_fc_setting;
1694 	else
1695 		hw->fc.requested_mode = e1000_fc_none;
1696 
1697 	/* Issue a global reset */
1698 	igb_pf_reset_hw(hw);
1699 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1700 
1701 	diag = e1000_init_hw(hw);
1702 	if (diag < 0)
1703 		return diag;
1704 
1705 	E1000_WRITE_REG(hw, E1000_VET,
1706 			RTE_ETHER_TYPE_VLAN << 16 | RTE_ETHER_TYPE_VLAN);
1707 	e1000_get_phy_info(hw);
1708 	e1000_check_for_link(hw);
1709 
1710 	return 0;
1711 }
1712 
1713 /* This function is based on igb_update_stats_counters() in igb/if_igb.c */
1714 static void
1715 igb_read_stats_registers(struct e1000_hw *hw, struct e1000_hw_stats *stats)
1716 {
1717 	int pause_frames;
1718 
1719 	uint64_t old_gprc  = stats->gprc;
1720 	uint64_t old_gptc  = stats->gptc;
1721 	uint64_t old_tpr   = stats->tpr;
1722 	uint64_t old_tpt   = stats->tpt;
1723 	uint64_t old_rpthc = stats->rpthc;
1724 	uint64_t old_hgptc = stats->hgptc;
1725 
1726 	if(hw->phy.media_type == e1000_media_type_copper ||
1727 	    (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
1728 		stats->symerrs +=
1729 		    E1000_READ_REG(hw,E1000_SYMERRS);
1730 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
1731 	}
1732 
1733 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
1734 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
1735 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
1736 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
1737 
1738 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
1739 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
1740 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
1741 	stats->dc += E1000_READ_REG(hw, E1000_DC);
1742 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
1743 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
1744 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
1745 	/*
1746 	** For watchdog management we need to know if we have been
1747 	** paused during the last interval, so capture that here.
1748 	*/
1749 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
1750 	stats->xoffrxc += pause_frames;
1751 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
1752 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
1753 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
1754 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
1755 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
1756 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
1757 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
1758 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
1759 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
1760 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
1761 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
1762 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
1763 
1764 	/* For the 64-bit byte counters the low dword must be read first. */
1765 	/* Both registers clear on the read of the high dword */
1766 
1767 	/* Workaround CRC bytes included in size, take away 4 bytes/packet */
1768 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
1769 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
1770 	stats->gorc -= (stats->gprc - old_gprc) * RTE_ETHER_CRC_LEN;
1771 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
1772 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
1773 	stats->gotc -= (stats->gptc - old_gptc) * RTE_ETHER_CRC_LEN;
1774 
1775 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
1776 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
1777 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
1778 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
1779 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
1780 
1781 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
1782 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
1783 
1784 	stats->tor += E1000_READ_REG(hw, E1000_TORL);
1785 	stats->tor += ((uint64_t)E1000_READ_REG(hw, E1000_TORH) << 32);
1786 	stats->tor -= (stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
1787 	stats->tot += E1000_READ_REG(hw, E1000_TOTL);
1788 	stats->tot += ((uint64_t)E1000_READ_REG(hw, E1000_TOTH) << 32);
1789 	stats->tot -= (stats->tpt - old_tpt) * RTE_ETHER_CRC_LEN;
1790 
1791 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
1792 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
1793 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
1794 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
1795 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
1796 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
1797 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
1798 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
1799 
1800 	/* Interrupt Counts */
1801 
1802 	stats->iac += E1000_READ_REG(hw, E1000_IAC);
1803 	stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
1804 	stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
1805 	stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
1806 	stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
1807 	stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
1808 	stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
1809 	stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
1810 	stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
1811 
1812 	/* Host to Card Statistics */
1813 
1814 	stats->cbtmpc += E1000_READ_REG(hw, E1000_CBTMPC);
1815 	stats->htdpmc += E1000_READ_REG(hw, E1000_HTDPMC);
1816 	stats->cbrdpc += E1000_READ_REG(hw, E1000_CBRDPC);
1817 	stats->cbrmpc += E1000_READ_REG(hw, E1000_CBRMPC);
1818 	stats->rpthc += E1000_READ_REG(hw, E1000_RPTHC);
1819 	stats->hgptc += E1000_READ_REG(hw, E1000_HGPTC);
1820 	stats->htcbdpc += E1000_READ_REG(hw, E1000_HTCBDPC);
1821 	stats->hgorc += E1000_READ_REG(hw, E1000_HGORCL);
1822 	stats->hgorc += ((uint64_t)E1000_READ_REG(hw, E1000_HGORCH) << 32);
1823 	stats->hgorc -= (stats->rpthc - old_rpthc) * RTE_ETHER_CRC_LEN;
1824 	stats->hgotc += E1000_READ_REG(hw, E1000_HGOTCL);
1825 	stats->hgotc += ((uint64_t)E1000_READ_REG(hw, E1000_HGOTCH) << 32);
1826 	stats->hgotc -= (stats->hgptc - old_hgptc) * RTE_ETHER_CRC_LEN;
1827 	stats->lenerrs += E1000_READ_REG(hw, E1000_LENERRS);
1828 	stats->scvpc += E1000_READ_REG(hw, E1000_SCVPC);
1829 	stats->hrmpc += E1000_READ_REG(hw, E1000_HRMPC);
1830 
1831 	stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
1832 	stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
1833 	stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
1834 	stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
1835 	stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
1836 	stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
1837 }
1838 
1839 static int
1840 eth_igb_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1841 {
1842 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1843 	struct e1000_hw_stats *stats =
1844 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1845 
1846 	igb_read_stats_registers(hw, stats);
1847 
1848 	if (rte_stats == NULL)
1849 		return -EINVAL;
1850 
1851 	/* Rx Errors */
1852 	rte_stats->imissed = stats->mpc;
1853 	rte_stats->ierrors = stats->crcerrs +
1854 	                     stats->rlec + stats->ruc + stats->roc +
1855 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
1856 
1857 	/* Tx Errors */
1858 	rte_stats->oerrors = stats->ecol + stats->latecol;
1859 
1860 	rte_stats->ipackets = stats->gprc;
1861 	rte_stats->opackets = stats->gptc;
1862 	rte_stats->ibytes   = stats->gorc;
1863 	rte_stats->obytes   = stats->gotc;
1864 	return 0;
1865 }
1866 
1867 static int
1868 eth_igb_stats_reset(struct rte_eth_dev *dev)
1869 {
1870 	struct e1000_hw_stats *hw_stats =
1871 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1872 
1873 	/* HW registers are cleared on read */
1874 	eth_igb_stats_get(dev, NULL);
1875 
1876 	/* Reset software totals */
1877 	memset(hw_stats, 0, sizeof(*hw_stats));
1878 
1879 	return 0;
1880 }
1881 
1882 static int
1883 eth_igb_xstats_reset(struct rte_eth_dev *dev)
1884 {
1885 	struct e1000_hw_stats *stats =
1886 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1887 
1888 	/* HW registers are cleared on read */
1889 	eth_igb_xstats_get(dev, NULL, IGB_NB_XSTATS);
1890 
1891 	/* Reset software totals */
1892 	memset(stats, 0, sizeof(*stats));
1893 
1894 	return 0;
1895 }
1896 
1897 static int eth_igb_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1898 	struct rte_eth_xstat_name *xstats_names,
1899 	__rte_unused unsigned int size)
1900 {
1901 	unsigned i;
1902 
1903 	if (xstats_names == NULL)
1904 		return IGB_NB_XSTATS;
1905 
1906 	/* Note: limit checked in rte_eth_xstats_names() */
1907 
1908 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1909 		strlcpy(xstats_names[i].name, rte_igb_stats_strings[i].name,
1910 			sizeof(xstats_names[i].name));
1911 	}
1912 
1913 	return IGB_NB_XSTATS;
1914 }
1915 
1916 static int eth_igb_xstats_get_names_by_id(struct rte_eth_dev *dev,
1917 		struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
1918 		unsigned int limit)
1919 {
1920 	unsigned int i;
1921 
1922 	if (!ids) {
1923 		if (xstats_names == NULL)
1924 			return IGB_NB_XSTATS;
1925 
1926 		for (i = 0; i < IGB_NB_XSTATS; i++)
1927 			strlcpy(xstats_names[i].name,
1928 				rte_igb_stats_strings[i].name,
1929 				sizeof(xstats_names[i].name));
1930 
1931 		return IGB_NB_XSTATS;
1932 
1933 	} else {
1934 		struct rte_eth_xstat_name xstats_names_copy[IGB_NB_XSTATS];
1935 
1936 		eth_igb_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
1937 				IGB_NB_XSTATS);
1938 
1939 		for (i = 0; i < limit; i++) {
1940 			if (ids[i] >= IGB_NB_XSTATS) {
1941 				PMD_INIT_LOG(ERR, "id value isn't valid");
1942 				return -1;
1943 			}
1944 			strcpy(xstats_names[i].name,
1945 					xstats_names_copy[ids[i]].name);
1946 		}
1947 		return limit;
1948 	}
1949 }
1950 
1951 static int
1952 eth_igb_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1953 		   unsigned n)
1954 {
1955 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1956 	struct e1000_hw_stats *hw_stats =
1957 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1958 	unsigned i;
1959 
1960 	if (n < IGB_NB_XSTATS)
1961 		return IGB_NB_XSTATS;
1962 
1963 	igb_read_stats_registers(hw, hw_stats);
1964 
1965 	/* If this is a reset xstats is NULL, and we have cleared the
1966 	 * registers by reading them.
1967 	 */
1968 	if (!xstats)
1969 		return 0;
1970 
1971 	/* Extended stats */
1972 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1973 		xstats[i].id = i;
1974 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1975 			rte_igb_stats_strings[i].offset);
1976 	}
1977 
1978 	return IGB_NB_XSTATS;
1979 }
1980 
1981 static int
1982 eth_igb_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
1983 		uint64_t *values, unsigned int n)
1984 {
1985 	unsigned int i;
1986 
1987 	if (!ids) {
1988 		struct e1000_hw *hw =
1989 			E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1990 		struct e1000_hw_stats *hw_stats =
1991 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1992 
1993 		if (n < IGB_NB_XSTATS)
1994 			return IGB_NB_XSTATS;
1995 
1996 		igb_read_stats_registers(hw, hw_stats);
1997 
1998 		/* If this is a reset xstats is NULL, and we have cleared the
1999 		 * registers by reading them.
2000 		 */
2001 		if (!values)
2002 			return 0;
2003 
2004 		/* Extended stats */
2005 		for (i = 0; i < IGB_NB_XSTATS; i++)
2006 			values[i] = *(uint64_t *)(((char *)hw_stats) +
2007 					rte_igb_stats_strings[i].offset);
2008 
2009 		return IGB_NB_XSTATS;
2010 
2011 	} else {
2012 		uint64_t values_copy[IGB_NB_XSTATS];
2013 
2014 		eth_igb_xstats_get_by_id(dev, NULL, values_copy,
2015 				IGB_NB_XSTATS);
2016 
2017 		for (i = 0; i < n; i++) {
2018 			if (ids[i] >= IGB_NB_XSTATS) {
2019 				PMD_INIT_LOG(ERR, "id value isn't valid");
2020 				return -1;
2021 			}
2022 			values[i] = values_copy[ids[i]];
2023 		}
2024 		return n;
2025 	}
2026 }
2027 
2028 static void
2029 igbvf_read_stats_registers(struct e1000_hw *hw, struct e1000_vf_stats *hw_stats)
2030 {
2031 	/* Good Rx packets, include VF loopback */
2032 	UPDATE_VF_STAT(E1000_VFGPRC,
2033 	    hw_stats->last_gprc, hw_stats->gprc);
2034 
2035 	/* Good Rx octets, include VF loopback */
2036 	UPDATE_VF_STAT(E1000_VFGORC,
2037 	    hw_stats->last_gorc, hw_stats->gorc);
2038 
2039 	/* Good Tx packets, include VF loopback */
2040 	UPDATE_VF_STAT(E1000_VFGPTC,
2041 	    hw_stats->last_gptc, hw_stats->gptc);
2042 
2043 	/* Good Tx octets, include VF loopback */
2044 	UPDATE_VF_STAT(E1000_VFGOTC,
2045 	    hw_stats->last_gotc, hw_stats->gotc);
2046 
2047 	/* Rx Multicst packets */
2048 	UPDATE_VF_STAT(E1000_VFMPRC,
2049 	    hw_stats->last_mprc, hw_stats->mprc);
2050 
2051 	/* Good Rx loopback packets */
2052 	UPDATE_VF_STAT(E1000_VFGPRLBC,
2053 	    hw_stats->last_gprlbc, hw_stats->gprlbc);
2054 
2055 	/* Good Rx loopback octets */
2056 	UPDATE_VF_STAT(E1000_VFGORLBC,
2057 	    hw_stats->last_gorlbc, hw_stats->gorlbc);
2058 
2059 	/* Good Tx loopback packets */
2060 	UPDATE_VF_STAT(E1000_VFGPTLBC,
2061 	    hw_stats->last_gptlbc, hw_stats->gptlbc);
2062 
2063 	/* Good Tx loopback octets */
2064 	UPDATE_VF_STAT(E1000_VFGOTLBC,
2065 	    hw_stats->last_gotlbc, hw_stats->gotlbc);
2066 }
2067 
2068 static int eth_igbvf_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
2069 				     struct rte_eth_xstat_name *xstats_names,
2070 				     __rte_unused unsigned limit)
2071 {
2072 	unsigned i;
2073 
2074 	if (xstats_names != NULL)
2075 		for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2076 			strlcpy(xstats_names[i].name,
2077 				rte_igbvf_stats_strings[i].name,
2078 				sizeof(xstats_names[i].name));
2079 		}
2080 	return IGBVF_NB_XSTATS;
2081 }
2082 
2083 static int
2084 eth_igbvf_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
2085 		     unsigned n)
2086 {
2087 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2088 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2089 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2090 	unsigned i;
2091 
2092 	if (n < IGBVF_NB_XSTATS)
2093 		return IGBVF_NB_XSTATS;
2094 
2095 	igbvf_read_stats_registers(hw, hw_stats);
2096 
2097 	if (!xstats)
2098 		return 0;
2099 
2100 	for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2101 		xstats[i].id = i;
2102 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
2103 			rte_igbvf_stats_strings[i].offset);
2104 	}
2105 
2106 	return IGBVF_NB_XSTATS;
2107 }
2108 
2109 static int
2110 eth_igbvf_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
2111 {
2112 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2113 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2114 			  E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2115 
2116 	igbvf_read_stats_registers(hw, hw_stats);
2117 
2118 	if (rte_stats == NULL)
2119 		return -EINVAL;
2120 
2121 	rte_stats->ipackets = hw_stats->gprc;
2122 	rte_stats->ibytes = hw_stats->gorc;
2123 	rte_stats->opackets = hw_stats->gptc;
2124 	rte_stats->obytes = hw_stats->gotc;
2125 	return 0;
2126 }
2127 
2128 static int
2129 eth_igbvf_stats_reset(struct rte_eth_dev *dev)
2130 {
2131 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats*)
2132 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2133 
2134 	/* Sync HW register to the last stats */
2135 	eth_igbvf_stats_get(dev, NULL);
2136 
2137 	/* reset HW current stats*/
2138 	memset(&hw_stats->gprc, 0, sizeof(*hw_stats) -
2139 	       offsetof(struct e1000_vf_stats, gprc));
2140 
2141 	return 0;
2142 }
2143 
2144 static int
2145 eth_igb_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
2146 		       size_t fw_size)
2147 {
2148 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2149 	struct e1000_fw_version fw;
2150 	int ret;
2151 
2152 	e1000_get_fw_version(hw, &fw);
2153 
2154 	switch (hw->mac.type) {
2155 	case e1000_i210:
2156 	case e1000_i211:
2157 		if (!(e1000_get_flash_presence_i210(hw))) {
2158 			ret = snprintf(fw_version, fw_size,
2159 				 "%2d.%2d-%d",
2160 				 fw.invm_major, fw.invm_minor,
2161 				 fw.invm_img_type);
2162 			break;
2163 		}
2164 		/* fall through */
2165 	default:
2166 		/* if option rom is valid, display its version too */
2167 		if (fw.or_valid) {
2168 			ret = snprintf(fw_version, fw_size,
2169 				 "%d.%d, 0x%08x, %d.%d.%d",
2170 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
2171 				 fw.or_major, fw.or_build, fw.or_patch);
2172 		/* no option rom */
2173 		} else {
2174 			if (fw.etrack_id != 0X0000) {
2175 				ret = snprintf(fw_version, fw_size,
2176 					 "%d.%d, 0x%08x",
2177 					 fw.eep_major, fw.eep_minor,
2178 					 fw.etrack_id);
2179 			} else {
2180 				ret = snprintf(fw_version, fw_size,
2181 					 "%d.%d.%d",
2182 					 fw.eep_major, fw.eep_minor,
2183 					 fw.eep_build);
2184 			}
2185 		}
2186 		break;
2187 	}
2188 
2189 	ret += 1; /* add the size of '\0' */
2190 	if (fw_size < (u32)ret)
2191 		return ret;
2192 	else
2193 		return 0;
2194 }
2195 
2196 static int
2197 eth_igb_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2198 {
2199 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2200 
2201 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2202 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2203 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2204 	dev_info->rx_queue_offload_capa = igb_get_rx_queue_offloads_capa(dev);
2205 	dev_info->rx_offload_capa = igb_get_rx_port_offloads_capa(dev) |
2206 				    dev_info->rx_queue_offload_capa;
2207 	dev_info->tx_queue_offload_capa = igb_get_tx_queue_offloads_capa(dev);
2208 	dev_info->tx_offload_capa = igb_get_tx_port_offloads_capa(dev) |
2209 				    dev_info->tx_queue_offload_capa;
2210 
2211 	switch (hw->mac.type) {
2212 	case e1000_82575:
2213 		dev_info->max_rx_queues = 4;
2214 		dev_info->max_tx_queues = 4;
2215 		dev_info->max_vmdq_pools = 0;
2216 		break;
2217 
2218 	case e1000_82576:
2219 		dev_info->max_rx_queues = 16;
2220 		dev_info->max_tx_queues = 16;
2221 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2222 		dev_info->vmdq_queue_num = 16;
2223 		break;
2224 
2225 	case e1000_82580:
2226 		dev_info->max_rx_queues = 8;
2227 		dev_info->max_tx_queues = 8;
2228 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2229 		dev_info->vmdq_queue_num = 8;
2230 		break;
2231 
2232 	case e1000_i350:
2233 		dev_info->max_rx_queues = 8;
2234 		dev_info->max_tx_queues = 8;
2235 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2236 		dev_info->vmdq_queue_num = 8;
2237 		break;
2238 
2239 	case e1000_i354:
2240 		dev_info->max_rx_queues = 8;
2241 		dev_info->max_tx_queues = 8;
2242 		break;
2243 
2244 	case e1000_i210:
2245 		dev_info->max_rx_queues = 4;
2246 		dev_info->max_tx_queues = 4;
2247 		dev_info->max_vmdq_pools = 0;
2248 		break;
2249 
2250 	case e1000_i211:
2251 		dev_info->max_rx_queues = 2;
2252 		dev_info->max_tx_queues = 2;
2253 		dev_info->max_vmdq_pools = 0;
2254 		break;
2255 
2256 	default:
2257 		/* Should not happen */
2258 		return -EINVAL;
2259 	}
2260 	dev_info->hash_key_size = IGB_HKEY_MAX_INDEX * sizeof(uint32_t);
2261 	dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
2262 	dev_info->flow_type_rss_offloads = IGB_RSS_OFFLOAD_ALL;
2263 
2264 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2265 		.rx_thresh = {
2266 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2267 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2268 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2269 		},
2270 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2271 		.rx_drop_en = 0,
2272 		.offloads = 0,
2273 	};
2274 
2275 	dev_info->default_txconf = (struct rte_eth_txconf) {
2276 		.tx_thresh = {
2277 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2278 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2279 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2280 		},
2281 		.offloads = 0,
2282 	};
2283 
2284 	dev_info->rx_desc_lim = rx_desc_lim;
2285 	dev_info->tx_desc_lim = tx_desc_lim;
2286 
2287 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
2288 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
2289 			ETH_LINK_SPEED_1G;
2290 
2291 	dev_info->max_mtu = dev_info->max_rx_pktlen - E1000_ETH_OVERHEAD;
2292 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
2293 
2294 	return 0;
2295 }
2296 
2297 static const uint32_t *
2298 eth_igb_supported_ptypes_get(struct rte_eth_dev *dev)
2299 {
2300 	static const uint32_t ptypes[] = {
2301 		/* refers to igb_rxd_pkt_info_to_pkt_type() */
2302 		RTE_PTYPE_L2_ETHER,
2303 		RTE_PTYPE_L3_IPV4,
2304 		RTE_PTYPE_L3_IPV4_EXT,
2305 		RTE_PTYPE_L3_IPV6,
2306 		RTE_PTYPE_L3_IPV6_EXT,
2307 		RTE_PTYPE_L4_TCP,
2308 		RTE_PTYPE_L4_UDP,
2309 		RTE_PTYPE_L4_SCTP,
2310 		RTE_PTYPE_TUNNEL_IP,
2311 		RTE_PTYPE_INNER_L3_IPV6,
2312 		RTE_PTYPE_INNER_L3_IPV6_EXT,
2313 		RTE_PTYPE_INNER_L4_TCP,
2314 		RTE_PTYPE_INNER_L4_UDP,
2315 		RTE_PTYPE_UNKNOWN
2316 	};
2317 
2318 	if (dev->rx_pkt_burst == eth_igb_recv_pkts ||
2319 	    dev->rx_pkt_burst == eth_igb_recv_scattered_pkts)
2320 		return ptypes;
2321 	return NULL;
2322 }
2323 
2324 static int
2325 eth_igbvf_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2326 {
2327 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2328 
2329 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2330 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2331 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2332 	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
2333 				DEV_TX_OFFLOAD_IPV4_CKSUM  |
2334 				DEV_TX_OFFLOAD_UDP_CKSUM   |
2335 				DEV_TX_OFFLOAD_TCP_CKSUM   |
2336 				DEV_TX_OFFLOAD_SCTP_CKSUM  |
2337 				DEV_TX_OFFLOAD_TCP_TSO;
2338 	switch (hw->mac.type) {
2339 	case e1000_vfadapt:
2340 		dev_info->max_rx_queues = 2;
2341 		dev_info->max_tx_queues = 2;
2342 		break;
2343 	case e1000_vfadapt_i350:
2344 		dev_info->max_rx_queues = 1;
2345 		dev_info->max_tx_queues = 1;
2346 		break;
2347 	default:
2348 		/* Should not happen */
2349 		return -EINVAL;
2350 	}
2351 
2352 	dev_info->rx_queue_offload_capa = igb_get_rx_queue_offloads_capa(dev);
2353 	dev_info->rx_offload_capa = igb_get_rx_port_offloads_capa(dev) |
2354 				    dev_info->rx_queue_offload_capa;
2355 	dev_info->tx_queue_offload_capa = igb_get_tx_queue_offloads_capa(dev);
2356 	dev_info->tx_offload_capa = igb_get_tx_port_offloads_capa(dev) |
2357 				    dev_info->tx_queue_offload_capa;
2358 
2359 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2360 		.rx_thresh = {
2361 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2362 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2363 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2364 		},
2365 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2366 		.rx_drop_en = 0,
2367 		.offloads = 0,
2368 	};
2369 
2370 	dev_info->default_txconf = (struct rte_eth_txconf) {
2371 		.tx_thresh = {
2372 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2373 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2374 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2375 		},
2376 		.offloads = 0,
2377 	};
2378 
2379 	dev_info->rx_desc_lim = rx_desc_lim;
2380 	dev_info->tx_desc_lim = tx_desc_lim;
2381 
2382 	return 0;
2383 }
2384 
2385 /* return 0 means link status changed, -1 means not changed */
2386 static int
2387 eth_igb_link_update(struct rte_eth_dev *dev, int wait_to_complete)
2388 {
2389 	struct e1000_hw *hw =
2390 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2391 	struct rte_eth_link link;
2392 	int link_check, count;
2393 
2394 	link_check = 0;
2395 	hw->mac.get_link_status = 1;
2396 
2397 	/* possible wait-to-complete in up to 9 seconds */
2398 	for (count = 0; count < IGB_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
2399 		/* Read the real link status */
2400 		switch (hw->phy.media_type) {
2401 		case e1000_media_type_copper:
2402 			/* Do the work to read phy */
2403 			e1000_check_for_link(hw);
2404 			link_check = !hw->mac.get_link_status;
2405 			break;
2406 
2407 		case e1000_media_type_fiber:
2408 			e1000_check_for_link(hw);
2409 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
2410 				      E1000_STATUS_LU);
2411 			break;
2412 
2413 		case e1000_media_type_internal_serdes:
2414 			e1000_check_for_link(hw);
2415 			link_check = hw->mac.serdes_has_link;
2416 			break;
2417 
2418 		/* VF device is type_unknown */
2419 		case e1000_media_type_unknown:
2420 			eth_igbvf_link_update(hw);
2421 			link_check = !hw->mac.get_link_status;
2422 			break;
2423 
2424 		default:
2425 			break;
2426 		}
2427 		if (link_check || wait_to_complete == 0)
2428 			break;
2429 		rte_delay_ms(IGB_LINK_UPDATE_CHECK_INTERVAL);
2430 	}
2431 	memset(&link, 0, sizeof(link));
2432 
2433 	/* Now we check if a transition has happened */
2434 	if (link_check) {
2435 		uint16_t duplex, speed;
2436 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
2437 		link.link_duplex = (duplex == FULL_DUPLEX) ?
2438 				ETH_LINK_FULL_DUPLEX :
2439 				ETH_LINK_HALF_DUPLEX;
2440 		link.link_speed = speed;
2441 		link.link_status = ETH_LINK_UP;
2442 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
2443 				ETH_LINK_SPEED_FIXED);
2444 	} else if (!link_check) {
2445 		link.link_speed = 0;
2446 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
2447 		link.link_status = ETH_LINK_DOWN;
2448 		link.link_autoneg = ETH_LINK_FIXED;
2449 	}
2450 
2451 	return rte_eth_linkstatus_set(dev, &link);
2452 }
2453 
2454 /*
2455  * igb_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
2456  * For ASF and Pass Through versions of f/w this means
2457  * that the driver is loaded.
2458  */
2459 static void
2460 igb_hw_control_acquire(struct e1000_hw *hw)
2461 {
2462 	uint32_t ctrl_ext;
2463 
2464 	/* Let firmware know the driver has taken over */
2465 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2466 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2467 }
2468 
2469 /*
2470  * igb_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
2471  * For ASF and Pass Through versions of f/w this means that the
2472  * driver is no longer loaded.
2473  */
2474 static void
2475 igb_hw_control_release(struct e1000_hw *hw)
2476 {
2477 	uint32_t ctrl_ext;
2478 
2479 	/* Let firmware taken over control of h/w */
2480 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2481 	E1000_WRITE_REG(hw, E1000_CTRL_EXT,
2482 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2483 }
2484 
2485 /*
2486  * Bit of a misnomer, what this really means is
2487  * to enable OS management of the system... aka
2488  * to disable special hardware management features.
2489  */
2490 static void
2491 igb_init_manageability(struct e1000_hw *hw)
2492 {
2493 	if (e1000_enable_mng_pass_thru(hw)) {
2494 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
2495 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2496 
2497 		/* disable hardware interception of ARP */
2498 		manc &= ~(E1000_MANC_ARP_EN);
2499 
2500 		/* enable receiving management packets to the host */
2501 		manc |= E1000_MANC_EN_MNG2HOST;
2502 		manc2h |= 1 << 5;  /* Mng Port 623 */
2503 		manc2h |= 1 << 6;  /* Mng Port 664 */
2504 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
2505 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2506 	}
2507 }
2508 
2509 static void
2510 igb_release_manageability(struct e1000_hw *hw)
2511 {
2512 	if (e1000_enable_mng_pass_thru(hw)) {
2513 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2514 
2515 		manc |= E1000_MANC_ARP_EN;
2516 		manc &= ~E1000_MANC_EN_MNG2HOST;
2517 
2518 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2519 	}
2520 }
2521 
2522 static int
2523 eth_igb_promiscuous_enable(struct rte_eth_dev *dev)
2524 {
2525 	struct e1000_hw *hw =
2526 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2527 	uint32_t rctl;
2528 
2529 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2530 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
2531 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2532 
2533 	return 0;
2534 }
2535 
2536 static int
2537 eth_igb_promiscuous_disable(struct rte_eth_dev *dev)
2538 {
2539 	struct e1000_hw *hw =
2540 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2541 	uint32_t rctl;
2542 
2543 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2544 	rctl &= (~E1000_RCTL_UPE);
2545 	if (dev->data->all_multicast == 1)
2546 		rctl |= E1000_RCTL_MPE;
2547 	else
2548 		rctl &= (~E1000_RCTL_MPE);
2549 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2550 
2551 	return 0;
2552 }
2553 
2554 static int
2555 eth_igb_allmulticast_enable(struct rte_eth_dev *dev)
2556 {
2557 	struct e1000_hw *hw =
2558 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2559 	uint32_t rctl;
2560 
2561 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2562 	rctl |= E1000_RCTL_MPE;
2563 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2564 
2565 	return 0;
2566 }
2567 
2568 static int
2569 eth_igb_allmulticast_disable(struct rte_eth_dev *dev)
2570 {
2571 	struct e1000_hw *hw =
2572 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2573 	uint32_t rctl;
2574 
2575 	if (dev->data->promiscuous == 1)
2576 		return 0; /* must remain in all_multicast mode */
2577 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2578 	rctl &= (~E1000_RCTL_MPE);
2579 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2580 
2581 	return 0;
2582 }
2583 
2584 static int
2585 eth_igb_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2586 {
2587 	struct e1000_hw *hw =
2588 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2589 	struct e1000_vfta * shadow_vfta =
2590 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2591 	uint32_t vfta;
2592 	uint32_t vid_idx;
2593 	uint32_t vid_bit;
2594 
2595 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
2596 			      E1000_VFTA_ENTRY_MASK);
2597 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
2598 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
2599 	if (on)
2600 		vfta |= vid_bit;
2601 	else
2602 		vfta &= ~vid_bit;
2603 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
2604 
2605 	/* update local VFTA copy */
2606 	shadow_vfta->vfta[vid_idx] = vfta;
2607 
2608 	return 0;
2609 }
2610 
2611 static int
2612 eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
2613 		      enum rte_vlan_type vlan_type,
2614 		      uint16_t tpid)
2615 {
2616 	struct e1000_hw *hw =
2617 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2618 	uint32_t reg, qinq;
2619 
2620 	qinq = E1000_READ_REG(hw, E1000_CTRL_EXT);
2621 	qinq &= E1000_CTRL_EXT_EXT_VLAN;
2622 
2623 	/* only outer TPID of double VLAN can be configured*/
2624 	if (qinq && vlan_type == ETH_VLAN_TYPE_OUTER) {
2625 		reg = E1000_READ_REG(hw, E1000_VET);
2626 		reg = (reg & (~E1000_VET_VET_EXT)) |
2627 			((uint32_t)tpid << E1000_VET_VET_EXT_SHIFT);
2628 		E1000_WRITE_REG(hw, E1000_VET, reg);
2629 
2630 		return 0;
2631 	}
2632 
2633 	/* all other TPID values are read-only*/
2634 	PMD_DRV_LOG(ERR, "Not supported");
2635 
2636 	return -ENOTSUP;
2637 }
2638 
2639 static void
2640 igb_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2641 {
2642 	struct e1000_hw *hw =
2643 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2644 	uint32_t reg;
2645 
2646 	/* Filter Table Disable */
2647 	reg = E1000_READ_REG(hw, E1000_RCTL);
2648 	reg &= ~E1000_RCTL_CFIEN;
2649 	reg &= ~E1000_RCTL_VFE;
2650 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2651 }
2652 
2653 static void
2654 igb_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2655 {
2656 	struct e1000_hw *hw =
2657 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2658 	struct e1000_vfta * shadow_vfta =
2659 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2660 	uint32_t reg;
2661 	int i;
2662 
2663 	/* Filter Table Enable, CFI not used for packet acceptance */
2664 	reg = E1000_READ_REG(hw, E1000_RCTL);
2665 	reg &= ~E1000_RCTL_CFIEN;
2666 	reg |= E1000_RCTL_VFE;
2667 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2668 
2669 	/* restore VFTA table */
2670 	for (i = 0; i < IGB_VFTA_SIZE; i++)
2671 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
2672 }
2673 
2674 static void
2675 igb_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2676 {
2677 	struct e1000_hw *hw =
2678 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2679 	uint32_t reg;
2680 
2681 	/* VLAN Mode Disable */
2682 	reg = E1000_READ_REG(hw, E1000_CTRL);
2683 	reg &= ~E1000_CTRL_VME;
2684 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2685 }
2686 
2687 static void
2688 igb_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2689 {
2690 	struct e1000_hw *hw =
2691 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2692 	uint32_t reg;
2693 
2694 	/* VLAN Mode Enable */
2695 	reg = E1000_READ_REG(hw, E1000_CTRL);
2696 	reg |= E1000_CTRL_VME;
2697 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2698 }
2699 
2700 static void
2701 igb_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2702 {
2703 	struct e1000_hw *hw =
2704 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2705 	uint32_t reg;
2706 
2707 	/* CTRL_EXT: Extended VLAN */
2708 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2709 	reg &= ~E1000_CTRL_EXT_EXTEND_VLAN;
2710 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2711 
2712 	/* Update maximum packet length */
2713 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME)
2714 		E1000_WRITE_REG(hw, E1000_RLPML,
2715 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2716 						VLAN_TAG_SIZE);
2717 }
2718 
2719 static void
2720 igb_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2721 {
2722 	struct e1000_hw *hw =
2723 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2724 	uint32_t reg;
2725 
2726 	/* CTRL_EXT: Extended VLAN */
2727 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2728 	reg |= E1000_CTRL_EXT_EXTEND_VLAN;
2729 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2730 
2731 	/* Update maximum packet length */
2732 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME)
2733 		E1000_WRITE_REG(hw, E1000_RLPML,
2734 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2735 						2 * VLAN_TAG_SIZE);
2736 }
2737 
2738 static int
2739 eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2740 {
2741 	struct rte_eth_rxmode *rxmode;
2742 
2743 	rxmode = &dev->data->dev_conf.rxmode;
2744 	if(mask & ETH_VLAN_STRIP_MASK){
2745 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2746 			igb_vlan_hw_strip_enable(dev);
2747 		else
2748 			igb_vlan_hw_strip_disable(dev);
2749 	}
2750 
2751 	if(mask & ETH_VLAN_FILTER_MASK){
2752 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2753 			igb_vlan_hw_filter_enable(dev);
2754 		else
2755 			igb_vlan_hw_filter_disable(dev);
2756 	}
2757 
2758 	if(mask & ETH_VLAN_EXTEND_MASK){
2759 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2760 			igb_vlan_hw_extend_enable(dev);
2761 		else
2762 			igb_vlan_hw_extend_disable(dev);
2763 	}
2764 
2765 	return 0;
2766 }
2767 
2768 
2769 /**
2770  * It enables the interrupt mask and then enable the interrupt.
2771  *
2772  * @param dev
2773  *  Pointer to struct rte_eth_dev.
2774  * @param on
2775  *  Enable or Disable
2776  *
2777  * @return
2778  *  - On success, zero.
2779  *  - On failure, a negative value.
2780  */
2781 static int
2782 eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
2783 {
2784 	struct e1000_interrupt *intr =
2785 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2786 
2787 	if (on)
2788 		intr->mask |= E1000_ICR_LSC;
2789 	else
2790 		intr->mask &= ~E1000_ICR_LSC;
2791 
2792 	return 0;
2793 }
2794 
2795 /* It clears the interrupt causes and enables the interrupt.
2796  * It will be called once only during nic initialized.
2797  *
2798  * @param dev
2799  *  Pointer to struct rte_eth_dev.
2800  *
2801  * @return
2802  *  - On success, zero.
2803  *  - On failure, a negative value.
2804  */
2805 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev)
2806 {
2807 	uint32_t mask, regval;
2808 	int ret;
2809 	struct e1000_hw *hw =
2810 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2811 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2812 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2813 	int misc_shift = rte_intr_allow_others(intr_handle) ? 1 : 0;
2814 	struct rte_eth_dev_info dev_info;
2815 
2816 	memset(&dev_info, 0, sizeof(dev_info));
2817 	ret = eth_igb_infos_get(dev, &dev_info);
2818 	if (ret != 0)
2819 		return ret;
2820 
2821 	mask = (0xFFFFFFFF >> (32 - dev_info.max_rx_queues)) << misc_shift;
2822 	regval = E1000_READ_REG(hw, E1000_EIMS);
2823 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
2824 
2825 	return 0;
2826 }
2827 
2828 /*
2829  * It reads ICR and gets interrupt causes, check it and set a bit flag
2830  * to update link status.
2831  *
2832  * @param dev
2833  *  Pointer to struct rte_eth_dev.
2834  *
2835  * @return
2836  *  - On success, zero.
2837  *  - On failure, a negative value.
2838  */
2839 static int
2840 eth_igb_interrupt_get_status(struct rte_eth_dev *dev)
2841 {
2842 	uint32_t icr;
2843 	struct e1000_hw *hw =
2844 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2845 	struct e1000_interrupt *intr =
2846 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2847 
2848 	igb_intr_disable(dev);
2849 
2850 	/* read-on-clear nic registers here */
2851 	icr = E1000_READ_REG(hw, E1000_ICR);
2852 
2853 	intr->flags = 0;
2854 	if (icr & E1000_ICR_LSC) {
2855 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
2856 	}
2857 
2858 	if (icr & E1000_ICR_VMMB)
2859 		intr->flags |= E1000_FLAG_MAILBOX;
2860 
2861 	return 0;
2862 }
2863 
2864 /*
2865  * It executes link_update after knowing an interrupt is prsent.
2866  *
2867  * @param dev
2868  *  Pointer to struct rte_eth_dev.
2869  *
2870  * @return
2871  *  - On success, zero.
2872  *  - On failure, a negative value.
2873  */
2874 static int
2875 eth_igb_interrupt_action(struct rte_eth_dev *dev,
2876 			 struct rte_intr_handle *intr_handle)
2877 {
2878 	struct e1000_hw *hw =
2879 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2880 	struct e1000_interrupt *intr =
2881 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2882 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2883 	struct rte_eth_link link;
2884 	int ret;
2885 
2886 	if (intr->flags & E1000_FLAG_MAILBOX) {
2887 		igb_pf_mbx_process(dev);
2888 		intr->flags &= ~E1000_FLAG_MAILBOX;
2889 	}
2890 
2891 	igb_intr_enable(dev);
2892 	rte_intr_ack(intr_handle);
2893 
2894 	if (intr->flags & E1000_FLAG_NEED_LINK_UPDATE) {
2895 		intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
2896 
2897 		/* set get_link_status to check register later */
2898 		hw->mac.get_link_status = 1;
2899 		ret = eth_igb_link_update(dev, 0);
2900 
2901 		/* check if link has changed */
2902 		if (ret < 0)
2903 			return 0;
2904 
2905 		rte_eth_linkstatus_get(dev, &link);
2906 		if (link.link_status) {
2907 			PMD_INIT_LOG(INFO,
2908 				     " Port %d: Link Up - speed %u Mbps - %s",
2909 				     dev->data->port_id,
2910 				     (unsigned)link.link_speed,
2911 				     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
2912 				     "full-duplex" : "half-duplex");
2913 		} else {
2914 			PMD_INIT_LOG(INFO, " Port %d: Link Down",
2915 				     dev->data->port_id);
2916 		}
2917 
2918 		PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
2919 			     pci_dev->addr.domain,
2920 			     pci_dev->addr.bus,
2921 			     pci_dev->addr.devid,
2922 			     pci_dev->addr.function);
2923 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC,
2924 					      NULL);
2925 	}
2926 
2927 	return 0;
2928 }
2929 
2930 /**
2931  * Interrupt handler which shall be registered at first.
2932  *
2933  * @param handle
2934  *  Pointer to interrupt handle.
2935  * @param param
2936  *  The address of parameter (struct rte_eth_dev *) regsitered before.
2937  *
2938  * @return
2939  *  void
2940  */
2941 static void
2942 eth_igb_interrupt_handler(void *param)
2943 {
2944 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2945 
2946 	eth_igb_interrupt_get_status(dev);
2947 	eth_igb_interrupt_action(dev, dev->intr_handle);
2948 }
2949 
2950 static int
2951 eth_igbvf_interrupt_get_status(struct rte_eth_dev *dev)
2952 {
2953 	uint32_t eicr;
2954 	struct e1000_hw *hw =
2955 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2956 	struct e1000_interrupt *intr =
2957 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2958 
2959 	igbvf_intr_disable(hw);
2960 
2961 	/* read-on-clear nic registers here */
2962 	eicr = E1000_READ_REG(hw, E1000_EICR);
2963 	intr->flags = 0;
2964 
2965 	if (eicr == E1000_VTIVAR_MISC_MAILBOX)
2966 		intr->flags |= E1000_FLAG_MAILBOX;
2967 
2968 	return 0;
2969 }
2970 
2971 void igbvf_mbx_process(struct rte_eth_dev *dev)
2972 {
2973 	struct e1000_hw *hw =
2974 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2975 	struct e1000_mbx_info *mbx = &hw->mbx;
2976 	u32 in_msg = 0;
2977 
2978 	/* peek the message first */
2979 	in_msg = E1000_READ_REG(hw, E1000_VMBMEM(0));
2980 
2981 	/* PF reset VF event */
2982 	if (in_msg == E1000_PF_CONTROL_MSG) {
2983 		/* dummy mbx read to ack pf */
2984 		if (mbx->ops.read(hw, &in_msg, 1, 0))
2985 			return;
2986 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
2987 					      NULL);
2988 	}
2989 }
2990 
2991 static int
2992 eth_igbvf_interrupt_action(struct rte_eth_dev *dev, struct rte_intr_handle *intr_handle)
2993 {
2994 	struct e1000_interrupt *intr =
2995 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2996 
2997 	if (intr->flags & E1000_FLAG_MAILBOX) {
2998 		igbvf_mbx_process(dev);
2999 		intr->flags &= ~E1000_FLAG_MAILBOX;
3000 	}
3001 
3002 	igbvf_intr_enable(dev);
3003 	rte_intr_ack(intr_handle);
3004 
3005 	return 0;
3006 }
3007 
3008 static void
3009 eth_igbvf_interrupt_handler(void *param)
3010 {
3011 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
3012 
3013 	eth_igbvf_interrupt_get_status(dev);
3014 	eth_igbvf_interrupt_action(dev, dev->intr_handle);
3015 }
3016 
3017 static int
3018 eth_igb_led_on(struct rte_eth_dev *dev)
3019 {
3020 	struct e1000_hw *hw;
3021 
3022 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3023 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3024 }
3025 
3026 static int
3027 eth_igb_led_off(struct rte_eth_dev *dev)
3028 {
3029 	struct e1000_hw *hw;
3030 
3031 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3032 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3033 }
3034 
3035 static int
3036 eth_igb_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3037 {
3038 	struct e1000_hw *hw;
3039 	uint32_t ctrl;
3040 	int tx_pause;
3041 	int rx_pause;
3042 
3043 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3044 	fc_conf->pause_time = hw->fc.pause_time;
3045 	fc_conf->high_water = hw->fc.high_water;
3046 	fc_conf->low_water = hw->fc.low_water;
3047 	fc_conf->send_xon = hw->fc.send_xon;
3048 	fc_conf->autoneg = hw->mac.autoneg;
3049 
3050 	/*
3051 	 * Return rx_pause and tx_pause status according to actual setting of
3052 	 * the TFCE and RFCE bits in the CTRL register.
3053 	 */
3054 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
3055 	if (ctrl & E1000_CTRL_TFCE)
3056 		tx_pause = 1;
3057 	else
3058 		tx_pause = 0;
3059 
3060 	if (ctrl & E1000_CTRL_RFCE)
3061 		rx_pause = 1;
3062 	else
3063 		rx_pause = 0;
3064 
3065 	if (rx_pause && tx_pause)
3066 		fc_conf->mode = RTE_FC_FULL;
3067 	else if (rx_pause)
3068 		fc_conf->mode = RTE_FC_RX_PAUSE;
3069 	else if (tx_pause)
3070 		fc_conf->mode = RTE_FC_TX_PAUSE;
3071 	else
3072 		fc_conf->mode = RTE_FC_NONE;
3073 
3074 	return 0;
3075 }
3076 
3077 static int
3078 eth_igb_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3079 {
3080 	struct e1000_hw *hw;
3081 	int err;
3082 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
3083 		e1000_fc_none,
3084 		e1000_fc_rx_pause,
3085 		e1000_fc_tx_pause,
3086 		e1000_fc_full
3087 	};
3088 	uint32_t rx_buf_size;
3089 	uint32_t max_high_water;
3090 	uint32_t rctl;
3091 
3092 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3093 	if (fc_conf->autoneg != hw->mac.autoneg)
3094 		return -ENOTSUP;
3095 	rx_buf_size = igb_get_rx_buffer_size(hw);
3096 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3097 
3098 	/* At least reserve one Ethernet frame for watermark */
3099 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
3100 	if ((fc_conf->high_water > max_high_water) ||
3101 	    (fc_conf->high_water < fc_conf->low_water)) {
3102 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
3103 		PMD_INIT_LOG(ERR, "high water must <=  0x%x", max_high_water);
3104 		return -EINVAL;
3105 	}
3106 
3107 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
3108 	hw->fc.pause_time     = fc_conf->pause_time;
3109 	hw->fc.high_water     = fc_conf->high_water;
3110 	hw->fc.low_water      = fc_conf->low_water;
3111 	hw->fc.send_xon	      = fc_conf->send_xon;
3112 
3113 	err = e1000_setup_link_generic(hw);
3114 	if (err == E1000_SUCCESS) {
3115 
3116 		/* check if we want to forward MAC frames - driver doesn't have native
3117 		 * capability to do that, so we'll write the registers ourselves */
3118 
3119 		rctl = E1000_READ_REG(hw, E1000_RCTL);
3120 
3121 		/* set or clear MFLCN.PMCF bit depending on configuration */
3122 		if (fc_conf->mac_ctrl_frame_fwd != 0)
3123 			rctl |= E1000_RCTL_PMCF;
3124 		else
3125 			rctl &= ~E1000_RCTL_PMCF;
3126 
3127 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3128 		E1000_WRITE_FLUSH(hw);
3129 
3130 		return 0;
3131 	}
3132 
3133 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
3134 	return -EIO;
3135 }
3136 
3137 #define E1000_RAH_POOLSEL_SHIFT      (18)
3138 static int
3139 eth_igb_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
3140 		uint32_t index, uint32_t pool)
3141 {
3142 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3143 	uint32_t rah;
3144 
3145 	e1000_rar_set(hw, mac_addr->addr_bytes, index);
3146 	rah = E1000_READ_REG(hw, E1000_RAH(index));
3147 	rah |= (0x1 << (E1000_RAH_POOLSEL_SHIFT + pool));
3148 	E1000_WRITE_REG(hw, E1000_RAH(index), rah);
3149 	return 0;
3150 }
3151 
3152 static void
3153 eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index)
3154 {
3155 	uint8_t addr[RTE_ETHER_ADDR_LEN];
3156 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3157 
3158 	memset(addr, 0, sizeof(addr));
3159 
3160 	e1000_rar_set(hw, addr, index);
3161 }
3162 
3163 static int
3164 eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
3165 				struct rte_ether_addr *addr)
3166 {
3167 	eth_igb_rar_clear(dev, 0);
3168 	eth_igb_rar_set(dev, (void *)addr, 0, 0);
3169 
3170 	return 0;
3171 }
3172 /*
3173  * Virtual Function operations
3174  */
3175 static void
3176 igbvf_intr_disable(struct e1000_hw *hw)
3177 {
3178 	PMD_INIT_FUNC_TRACE();
3179 
3180 	/* Clear interrupt mask to stop from interrupts being generated */
3181 	E1000_WRITE_REG(hw, E1000_EIMC, 0xFFFF);
3182 
3183 	E1000_WRITE_FLUSH(hw);
3184 }
3185 
3186 static void
3187 igbvf_stop_adapter(struct rte_eth_dev *dev)
3188 {
3189 	u32 reg_val;
3190 	u16 i;
3191 	struct rte_eth_dev_info dev_info;
3192 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3193 	int ret;
3194 
3195 	memset(&dev_info, 0, sizeof(dev_info));
3196 	ret = eth_igbvf_infos_get(dev, &dev_info);
3197 	if (ret != 0)
3198 		return;
3199 
3200 	/* Clear interrupt mask to stop from interrupts being generated */
3201 	igbvf_intr_disable(hw);
3202 
3203 	/* Clear any pending interrupts, flush previous writes */
3204 	E1000_READ_REG(hw, E1000_EICR);
3205 
3206 	/* Disable the transmit unit.  Each queue must be disabled. */
3207 	for (i = 0; i < dev_info.max_tx_queues; i++)
3208 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), E1000_TXDCTL_SWFLSH);
3209 
3210 	/* Disable the receive unit by stopping each queue */
3211 	for (i = 0; i < dev_info.max_rx_queues; i++) {
3212 		reg_val = E1000_READ_REG(hw, E1000_RXDCTL(i));
3213 		reg_val &= ~E1000_RXDCTL_QUEUE_ENABLE;
3214 		E1000_WRITE_REG(hw, E1000_RXDCTL(i), reg_val);
3215 		while (E1000_READ_REG(hw, E1000_RXDCTL(i)) & E1000_RXDCTL_QUEUE_ENABLE)
3216 			;
3217 	}
3218 
3219 	/* flush all queues disables */
3220 	E1000_WRITE_FLUSH(hw);
3221 	msec_delay(2);
3222 }
3223 
3224 static int eth_igbvf_link_update(struct e1000_hw *hw)
3225 {
3226 	struct e1000_mbx_info *mbx = &hw->mbx;
3227 	struct e1000_mac_info *mac = &hw->mac;
3228 	int ret_val = E1000_SUCCESS;
3229 
3230 	PMD_INIT_LOG(DEBUG, "e1000_check_for_link_vf");
3231 
3232 	/*
3233 	 * We only want to run this if there has been a rst asserted.
3234 	 * in this case that could mean a link change, device reset,
3235 	 * or a virtual function reset
3236 	 */
3237 
3238 	/* If we were hit with a reset or timeout drop the link */
3239 	if (!e1000_check_for_rst(hw, 0) || !mbx->timeout)
3240 		mac->get_link_status = TRUE;
3241 
3242 	if (!mac->get_link_status)
3243 		goto out;
3244 
3245 	/* if link status is down no point in checking to see if pf is up */
3246 	if (!(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU))
3247 		goto out;
3248 
3249 	/* if we passed all the tests above then the link is up and we no
3250 	 * longer need to check for link */
3251 	mac->get_link_status = FALSE;
3252 
3253 out:
3254 	return ret_val;
3255 }
3256 
3257 
3258 static int
3259 igbvf_dev_configure(struct rte_eth_dev *dev)
3260 {
3261 	struct rte_eth_conf* conf = &dev->data->dev_conf;
3262 
3263 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
3264 		     dev->data->port_id);
3265 
3266 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
3267 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
3268 
3269 	/*
3270 	 * VF has no ability to enable/disable HW CRC
3271 	 * Keep the persistent behavior the same as Host PF
3272 	 */
3273 #ifndef RTE_LIBRTE_E1000_PF_DISABLE_STRIP_CRC
3274 	if (conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
3275 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
3276 		conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_KEEP_CRC;
3277 	}
3278 #else
3279 	if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC)) {
3280 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
3281 		conf->rxmode.offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
3282 	}
3283 #endif
3284 
3285 	return 0;
3286 }
3287 
3288 static int
3289 igbvf_dev_start(struct rte_eth_dev *dev)
3290 {
3291 	struct e1000_hw *hw =
3292 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3293 	struct e1000_adapter *adapter =
3294 		E1000_DEV_PRIVATE(dev->data->dev_private);
3295 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3296 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3297 	int ret;
3298 	uint32_t intr_vector = 0;
3299 
3300 	PMD_INIT_FUNC_TRACE();
3301 
3302 	hw->mac.ops.reset_hw(hw);
3303 	adapter->stopped = 0;
3304 
3305 	/* Set all vfta */
3306 	igbvf_set_vfta_all(dev,1);
3307 
3308 	eth_igbvf_tx_init(dev);
3309 
3310 	/* This can fail when allocating mbufs for descriptor rings */
3311 	ret = eth_igbvf_rx_init(dev);
3312 	if (ret) {
3313 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
3314 		igb_dev_clear_queues(dev);
3315 		return ret;
3316 	}
3317 
3318 	/* check and configure queue intr-vector mapping */
3319 	if (rte_intr_cap_multiple(intr_handle) &&
3320 	    dev->data->dev_conf.intr_conf.rxq) {
3321 		intr_vector = dev->data->nb_rx_queues;
3322 		ret = rte_intr_efd_enable(intr_handle, intr_vector);
3323 		if (ret)
3324 			return ret;
3325 	}
3326 
3327 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
3328 		intr_handle->intr_vec =
3329 			rte_zmalloc("intr_vec",
3330 				    dev->data->nb_rx_queues * sizeof(int), 0);
3331 		if (!intr_handle->intr_vec) {
3332 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
3333 				     " intr_vec", dev->data->nb_rx_queues);
3334 			return -ENOMEM;
3335 		}
3336 	}
3337 
3338 	eth_igbvf_configure_msix_intr(dev);
3339 
3340 	/* enable uio/vfio intr/eventfd mapping */
3341 	rte_intr_enable(intr_handle);
3342 
3343 	/* resume enabled intr since hw reset */
3344 	igbvf_intr_enable(dev);
3345 
3346 	return 0;
3347 }
3348 
3349 static void
3350 igbvf_dev_stop(struct rte_eth_dev *dev)
3351 {
3352 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3353 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3354 	struct e1000_adapter *adapter =
3355 		E1000_DEV_PRIVATE(dev->data->dev_private);
3356 
3357 	if (adapter->stopped)
3358 		return;
3359 
3360 	PMD_INIT_FUNC_TRACE();
3361 
3362 	igbvf_stop_adapter(dev);
3363 
3364 	/*
3365 	  * Clear what we set, but we still keep shadow_vfta to
3366 	  * restore after device starts
3367 	  */
3368 	igbvf_set_vfta_all(dev,0);
3369 
3370 	igb_dev_clear_queues(dev);
3371 
3372 	/* disable intr eventfd mapping */
3373 	rte_intr_disable(intr_handle);
3374 
3375 	/* Clean datapath event and queue/vec mapping */
3376 	rte_intr_efd_disable(intr_handle);
3377 	if (intr_handle->intr_vec) {
3378 		rte_free(intr_handle->intr_vec);
3379 		intr_handle->intr_vec = NULL;
3380 	}
3381 
3382 	adapter->stopped = true;
3383 }
3384 
3385 static void
3386 igbvf_dev_close(struct rte_eth_dev *dev)
3387 {
3388 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3389 	struct rte_ether_addr addr;
3390 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3391 
3392 	PMD_INIT_FUNC_TRACE();
3393 
3394 	e1000_reset_hw(hw);
3395 
3396 	igbvf_dev_stop(dev);
3397 
3398 	igb_dev_free_queues(dev);
3399 
3400 	/**
3401 	 * reprogram the RAR with a zero mac address,
3402 	 * to ensure that the VF traffic goes to the PF
3403 	 * after stop, close and detach of the VF.
3404 	 **/
3405 
3406 	memset(&addr, 0, sizeof(addr));
3407 	igbvf_default_mac_addr_set(dev, &addr);
3408 
3409 	dev->dev_ops = NULL;
3410 	dev->rx_pkt_burst = NULL;
3411 	dev->tx_pkt_burst = NULL;
3412 
3413 	rte_intr_callback_unregister(&pci_dev->intr_handle,
3414 				     eth_igbvf_interrupt_handler,
3415 				     (void *)dev);
3416 }
3417 
3418 static int
3419 igbvf_promiscuous_enable(struct rte_eth_dev *dev)
3420 {
3421 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3422 
3423 	/* Set both unicast and multicast promisc */
3424 	e1000_promisc_set_vf(hw, e1000_promisc_enabled);
3425 
3426 	return 0;
3427 }
3428 
3429 static int
3430 igbvf_promiscuous_disable(struct rte_eth_dev *dev)
3431 {
3432 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3433 
3434 	/* If in allmulticast mode leave multicast promisc */
3435 	if (dev->data->all_multicast == 1)
3436 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3437 	else
3438 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3439 
3440 	return 0;
3441 }
3442 
3443 static int
3444 igbvf_allmulticast_enable(struct rte_eth_dev *dev)
3445 {
3446 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3447 
3448 	/* In promiscuous mode multicast promisc already set */
3449 	if (dev->data->promiscuous == 0)
3450 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3451 
3452 	return 0;
3453 }
3454 
3455 static int
3456 igbvf_allmulticast_disable(struct rte_eth_dev *dev)
3457 {
3458 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3459 
3460 	/* In promiscuous mode leave multicast promisc enabled */
3461 	if (dev->data->promiscuous == 0)
3462 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3463 
3464 	return 0;
3465 }
3466 
3467 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on)
3468 {
3469 	struct e1000_mbx_info *mbx = &hw->mbx;
3470 	uint32_t msgbuf[2];
3471 	s32 err;
3472 
3473 	/* After set vlan, vlan strip will also be enabled in igb driver*/
3474 	msgbuf[0] = E1000_VF_SET_VLAN;
3475 	msgbuf[1] = vid;
3476 	/* Setting the 8 bit field MSG INFO to TRUE indicates "add" */
3477 	if (on)
3478 		msgbuf[0] |= E1000_VF_SET_VLAN_ADD;
3479 
3480 	err = mbx->ops.write_posted(hw, msgbuf, 2, 0);
3481 	if (err)
3482 		goto mbx_err;
3483 
3484 	err = mbx->ops.read_posted(hw, msgbuf, 2, 0);
3485 	if (err)
3486 		goto mbx_err;
3487 
3488 	msgbuf[0] &= ~E1000_VT_MSGTYPE_CTS;
3489 	if (msgbuf[0] == (E1000_VF_SET_VLAN | E1000_VT_MSGTYPE_NACK))
3490 		err = -EINVAL;
3491 
3492 mbx_err:
3493 	return err;
3494 }
3495 
3496 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on)
3497 {
3498 	struct e1000_hw *hw =
3499 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3500 	struct e1000_vfta * shadow_vfta =
3501 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3502 	int i = 0, j = 0, vfta = 0, mask = 1;
3503 
3504 	for (i = 0; i < IGB_VFTA_SIZE; i++){
3505 		vfta = shadow_vfta->vfta[i];
3506 		if(vfta){
3507 			mask = 1;
3508 			for (j = 0; j < 32; j++){
3509 				if(vfta & mask)
3510 					igbvf_set_vfta(hw,
3511 						(uint16_t)((i<<5)+j), on);
3512 				mask<<=1;
3513 			}
3514 		}
3515 	}
3516 
3517 }
3518 
3519 static int
3520 igbvf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
3521 {
3522 	struct e1000_hw *hw =
3523 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3524 	struct e1000_vfta * shadow_vfta =
3525 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3526 	uint32_t vid_idx = 0;
3527 	uint32_t vid_bit = 0;
3528 	int ret = 0;
3529 
3530 	PMD_INIT_FUNC_TRACE();
3531 
3532 	/*vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf*/
3533 	ret = igbvf_set_vfta(hw, vlan_id, !!on);
3534 	if(ret){
3535 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
3536 		return ret;
3537 	}
3538 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
3539 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
3540 
3541 	/*Save what we set and retore it after device reset*/
3542 	if (on)
3543 		shadow_vfta->vfta[vid_idx] |= vid_bit;
3544 	else
3545 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
3546 
3547 	return 0;
3548 }
3549 
3550 static int
3551 igbvf_default_mac_addr_set(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
3552 {
3553 	struct e1000_hw *hw =
3554 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3555 
3556 	/* index is not used by rar_set() */
3557 	hw->mac.ops.rar_set(hw, (void *)addr, 0);
3558 	return 0;
3559 }
3560 
3561 
3562 static int
3563 eth_igb_rss_reta_update(struct rte_eth_dev *dev,
3564 			struct rte_eth_rss_reta_entry64 *reta_conf,
3565 			uint16_t reta_size)
3566 {
3567 	uint8_t i, j, mask;
3568 	uint32_t reta, r;
3569 	uint16_t idx, shift;
3570 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3571 
3572 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3573 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3574 			"(%d) doesn't match the number hardware can supported "
3575 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3576 		return -EINVAL;
3577 	}
3578 
3579 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3580 		idx = i / RTE_RETA_GROUP_SIZE;
3581 		shift = i % RTE_RETA_GROUP_SIZE;
3582 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3583 						IGB_4_BIT_MASK);
3584 		if (!mask)
3585 			continue;
3586 		if (mask == IGB_4_BIT_MASK)
3587 			r = 0;
3588 		else
3589 			r = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3590 		for (j = 0, reta = 0; j < IGB_4_BIT_WIDTH; j++) {
3591 			if (mask & (0x1 << j))
3592 				reta |= reta_conf[idx].reta[shift + j] <<
3593 							(CHAR_BIT * j);
3594 			else
3595 				reta |= r & (IGB_8_BIT_MASK << (CHAR_BIT * j));
3596 		}
3597 		E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
3598 	}
3599 
3600 	return 0;
3601 }
3602 
3603 static int
3604 eth_igb_rss_reta_query(struct rte_eth_dev *dev,
3605 		       struct rte_eth_rss_reta_entry64 *reta_conf,
3606 		       uint16_t reta_size)
3607 {
3608 	uint8_t i, j, mask;
3609 	uint32_t reta;
3610 	uint16_t idx, shift;
3611 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3612 
3613 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3614 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3615 			"(%d) doesn't match the number hardware can supported "
3616 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3617 		return -EINVAL;
3618 	}
3619 
3620 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3621 		idx = i / RTE_RETA_GROUP_SIZE;
3622 		shift = i % RTE_RETA_GROUP_SIZE;
3623 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3624 						IGB_4_BIT_MASK);
3625 		if (!mask)
3626 			continue;
3627 		reta = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3628 		for (j = 0; j < IGB_4_BIT_WIDTH; j++) {
3629 			if (mask & (0x1 << j))
3630 				reta_conf[idx].reta[shift + j] =
3631 					((reta >> (CHAR_BIT * j)) &
3632 						IGB_8_BIT_MASK);
3633 		}
3634 	}
3635 
3636 	return 0;
3637 }
3638 
3639 int
3640 eth_igb_syn_filter_set(struct rte_eth_dev *dev,
3641 			struct rte_eth_syn_filter *filter,
3642 			bool add)
3643 {
3644 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3645 	struct e1000_filter_info *filter_info =
3646 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3647 	uint32_t synqf, rfctl;
3648 
3649 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3650 		return -EINVAL;
3651 
3652 	synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3653 
3654 	if (add) {
3655 		if (synqf & E1000_SYN_FILTER_ENABLE)
3656 			return -EINVAL;
3657 
3658 		synqf = (uint32_t)(((filter->queue << E1000_SYN_FILTER_QUEUE_SHIFT) &
3659 			E1000_SYN_FILTER_QUEUE) | E1000_SYN_FILTER_ENABLE);
3660 
3661 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3662 		if (filter->hig_pri)
3663 			rfctl |= E1000_RFCTL_SYNQFP;
3664 		else
3665 			rfctl &= ~E1000_RFCTL_SYNQFP;
3666 
3667 		E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3668 	} else {
3669 		if (!(synqf & E1000_SYN_FILTER_ENABLE))
3670 			return -ENOENT;
3671 		synqf = 0;
3672 	}
3673 
3674 	filter_info->syn_info = synqf;
3675 	E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
3676 	E1000_WRITE_FLUSH(hw);
3677 	return 0;
3678 }
3679 
3680 static int
3681 eth_igb_syn_filter_get(struct rte_eth_dev *dev,
3682 			struct rte_eth_syn_filter *filter)
3683 {
3684 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3685 	uint32_t synqf, rfctl;
3686 
3687 	synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3688 	if (synqf & E1000_SYN_FILTER_ENABLE) {
3689 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3690 		filter->hig_pri = (rfctl & E1000_RFCTL_SYNQFP) ? 1 : 0;
3691 		filter->queue = (uint8_t)((synqf & E1000_SYN_FILTER_QUEUE) >>
3692 				E1000_SYN_FILTER_QUEUE_SHIFT);
3693 		return 0;
3694 	}
3695 
3696 	return -ENOENT;
3697 }
3698 
3699 static int
3700 eth_igb_syn_filter_handle(struct rte_eth_dev *dev,
3701 			enum rte_filter_op filter_op,
3702 			void *arg)
3703 {
3704 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3705 	int ret;
3706 
3707 	MAC_TYPE_FILTER_SUP(hw->mac.type);
3708 
3709 	if (filter_op == RTE_ETH_FILTER_NOP)
3710 		return 0;
3711 
3712 	if (arg == NULL) {
3713 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
3714 			    filter_op);
3715 		return -EINVAL;
3716 	}
3717 
3718 	switch (filter_op) {
3719 	case RTE_ETH_FILTER_ADD:
3720 		ret = eth_igb_syn_filter_set(dev,
3721 				(struct rte_eth_syn_filter *)arg,
3722 				TRUE);
3723 		break;
3724 	case RTE_ETH_FILTER_DELETE:
3725 		ret = eth_igb_syn_filter_set(dev,
3726 				(struct rte_eth_syn_filter *)arg,
3727 				FALSE);
3728 		break;
3729 	case RTE_ETH_FILTER_GET:
3730 		ret = eth_igb_syn_filter_get(dev,
3731 				(struct rte_eth_syn_filter *)arg);
3732 		break;
3733 	default:
3734 		PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
3735 		ret = -EINVAL;
3736 		break;
3737 	}
3738 
3739 	return ret;
3740 }
3741 
3742 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_2tuple_filter_info*/
3743 static inline int
3744 ntuple_filter_to_2tuple(struct rte_eth_ntuple_filter *filter,
3745 			struct e1000_2tuple_filter_info *filter_info)
3746 {
3747 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3748 		return -EINVAL;
3749 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
3750 		return -EINVAL;  /* filter index is out of range. */
3751 	if (filter->tcp_flags > RTE_NTUPLE_TCP_FLAGS_MASK)
3752 		return -EINVAL;  /* flags is invalid. */
3753 
3754 	switch (filter->dst_port_mask) {
3755 	case UINT16_MAX:
3756 		filter_info->dst_port_mask = 0;
3757 		filter_info->dst_port = filter->dst_port;
3758 		break;
3759 	case 0:
3760 		filter_info->dst_port_mask = 1;
3761 		break;
3762 	default:
3763 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
3764 		return -EINVAL;
3765 	}
3766 
3767 	switch (filter->proto_mask) {
3768 	case UINT8_MAX:
3769 		filter_info->proto_mask = 0;
3770 		filter_info->proto = filter->proto;
3771 		break;
3772 	case 0:
3773 		filter_info->proto_mask = 1;
3774 		break;
3775 	default:
3776 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
3777 		return -EINVAL;
3778 	}
3779 
3780 	filter_info->priority = (uint8_t)filter->priority;
3781 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
3782 		filter_info->tcp_flags = filter->tcp_flags;
3783 	else
3784 		filter_info->tcp_flags = 0;
3785 
3786 	return 0;
3787 }
3788 
3789 static inline struct e1000_2tuple_filter *
3790 igb_2tuple_filter_lookup(struct e1000_2tuple_filter_list *filter_list,
3791 			struct e1000_2tuple_filter_info *key)
3792 {
3793 	struct e1000_2tuple_filter *it;
3794 
3795 	TAILQ_FOREACH(it, filter_list, entries) {
3796 		if (memcmp(key, &it->filter_info,
3797 			sizeof(struct e1000_2tuple_filter_info)) == 0) {
3798 			return it;
3799 		}
3800 	}
3801 	return NULL;
3802 }
3803 
3804 /* inject a igb 2tuple filter to HW */
3805 static inline void
3806 igb_inject_2uple_filter(struct rte_eth_dev *dev,
3807 			   struct e1000_2tuple_filter *filter)
3808 {
3809 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3810 	uint32_t ttqf = E1000_TTQF_DISABLE_MASK;
3811 	uint32_t imir, imir_ext = E1000_IMIREXT_SIZE_BP;
3812 	int i;
3813 
3814 	i = filter->index;
3815 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
3816 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
3817 		imir |= E1000_IMIR_PORT_BP;
3818 	else
3819 		imir &= ~E1000_IMIR_PORT_BP;
3820 
3821 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
3822 
3823 	ttqf |= E1000_TTQF_QUEUE_ENABLE;
3824 	ttqf |= (uint32_t)(filter->queue << E1000_TTQF_QUEUE_SHIFT);
3825 	ttqf |= (uint32_t)(filter->filter_info.proto &
3826 						E1000_TTQF_PROTOCOL_MASK);
3827 	if (filter->filter_info.proto_mask == 0)
3828 		ttqf &= ~E1000_TTQF_MASK_ENABLE;
3829 
3830 	/* tcp flags bits setting. */
3831 	if (filter->filter_info.tcp_flags & RTE_NTUPLE_TCP_FLAGS_MASK) {
3832 		if (filter->filter_info.tcp_flags & RTE_TCP_URG_FLAG)
3833 			imir_ext |= E1000_IMIREXT_CTRL_URG;
3834 		if (filter->filter_info.tcp_flags & RTE_TCP_ACK_FLAG)
3835 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
3836 		if (filter->filter_info.tcp_flags & RTE_TCP_PSH_FLAG)
3837 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
3838 		if (filter->filter_info.tcp_flags & RTE_TCP_RST_FLAG)
3839 			imir_ext |= E1000_IMIREXT_CTRL_RST;
3840 		if (filter->filter_info.tcp_flags & RTE_TCP_SYN_FLAG)
3841 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
3842 		if (filter->filter_info.tcp_flags & RTE_TCP_FIN_FLAG)
3843 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
3844 	} else {
3845 		imir_ext |= E1000_IMIREXT_CTRL_BP;
3846 	}
3847 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
3848 	E1000_WRITE_REG(hw, E1000_TTQF(i), ttqf);
3849 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
3850 }
3851 
3852 /*
3853  * igb_add_2tuple_filter - add a 2tuple filter
3854  *
3855  * @param
3856  * dev: Pointer to struct rte_eth_dev.
3857  * ntuple_filter: ponter to the filter that will be added.
3858  *
3859  * @return
3860  *    - On success, zero.
3861  *    - On failure, a negative value.
3862  */
3863 static int
3864 igb_add_2tuple_filter(struct rte_eth_dev *dev,
3865 			struct rte_eth_ntuple_filter *ntuple_filter)
3866 {
3867 	struct e1000_filter_info *filter_info =
3868 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3869 	struct e1000_2tuple_filter *filter;
3870 	int i, ret;
3871 
3872 	filter = rte_zmalloc("e1000_2tuple_filter",
3873 			sizeof(struct e1000_2tuple_filter), 0);
3874 	if (filter == NULL)
3875 		return -ENOMEM;
3876 
3877 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3878 				      &filter->filter_info);
3879 	if (ret < 0) {
3880 		rte_free(filter);
3881 		return ret;
3882 	}
3883 	if (igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3884 					 &filter->filter_info) != NULL) {
3885 		PMD_DRV_LOG(ERR, "filter exists.");
3886 		rte_free(filter);
3887 		return -EEXIST;
3888 	}
3889 	filter->queue = ntuple_filter->queue;
3890 
3891 	/*
3892 	 * look for an unused 2tuple filter index,
3893 	 * and insert the filter to list.
3894 	 */
3895 	for (i = 0; i < E1000_MAX_TTQF_FILTERS; i++) {
3896 		if (!(filter_info->twotuple_mask & (1 << i))) {
3897 			filter_info->twotuple_mask |= 1 << i;
3898 			filter->index = i;
3899 			TAILQ_INSERT_TAIL(&filter_info->twotuple_list,
3900 					  filter,
3901 					  entries);
3902 			break;
3903 		}
3904 	}
3905 	if (i >= E1000_MAX_TTQF_FILTERS) {
3906 		PMD_DRV_LOG(ERR, "2tuple filters are full.");
3907 		rte_free(filter);
3908 		return -ENOSYS;
3909 	}
3910 
3911 	igb_inject_2uple_filter(dev, filter);
3912 	return 0;
3913 }
3914 
3915 int
3916 igb_delete_2tuple_filter(struct rte_eth_dev *dev,
3917 			struct e1000_2tuple_filter *filter)
3918 {
3919 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3920 	struct e1000_filter_info *filter_info =
3921 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3922 
3923 	filter_info->twotuple_mask &= ~(1 << filter->index);
3924 	TAILQ_REMOVE(&filter_info->twotuple_list, filter, entries);
3925 	rte_free(filter);
3926 
3927 	E1000_WRITE_REG(hw, E1000_TTQF(filter->index), E1000_TTQF_DISABLE_MASK);
3928 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
3929 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
3930 	return 0;
3931 }
3932 
3933 /*
3934  * igb_remove_2tuple_filter - remove a 2tuple filter
3935  *
3936  * @param
3937  * dev: Pointer to struct rte_eth_dev.
3938  * ntuple_filter: ponter to the filter that will be removed.
3939  *
3940  * @return
3941  *    - On success, zero.
3942  *    - On failure, a negative value.
3943  */
3944 static int
3945 igb_remove_2tuple_filter(struct rte_eth_dev *dev,
3946 			struct rte_eth_ntuple_filter *ntuple_filter)
3947 {
3948 	struct e1000_filter_info *filter_info =
3949 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3950 	struct e1000_2tuple_filter_info filter_2tuple;
3951 	struct e1000_2tuple_filter *filter;
3952 	int ret;
3953 
3954 	memset(&filter_2tuple, 0, sizeof(struct e1000_2tuple_filter_info));
3955 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3956 				      &filter_2tuple);
3957 	if (ret < 0)
3958 		return ret;
3959 
3960 	filter = igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3961 					 &filter_2tuple);
3962 	if (filter == NULL) {
3963 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
3964 		return -ENOENT;
3965 	}
3966 
3967 	igb_delete_2tuple_filter(dev, filter);
3968 
3969 	return 0;
3970 }
3971 
3972 /* inject a igb flex filter to HW */
3973 static inline void
3974 igb_inject_flex_filter(struct rte_eth_dev *dev,
3975 			   struct e1000_flex_filter *filter)
3976 {
3977 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3978 	uint32_t wufc, queueing;
3979 	uint32_t reg_off;
3980 	uint8_t i, j = 0;
3981 
3982 	wufc = E1000_READ_REG(hw, E1000_WUFC);
3983 	if (filter->index < E1000_MAX_FHFT)
3984 		reg_off = E1000_FHFT(filter->index);
3985 	else
3986 		reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
3987 
3988 	E1000_WRITE_REG(hw, E1000_WUFC, wufc | E1000_WUFC_FLEX_HQ |
3989 			(E1000_WUFC_FLX0 << filter->index));
3990 	queueing = filter->filter_info.len |
3991 		(filter->queue << E1000_FHFT_QUEUEING_QUEUE_SHIFT) |
3992 		(filter->filter_info.priority <<
3993 			E1000_FHFT_QUEUEING_PRIO_SHIFT);
3994 	E1000_WRITE_REG(hw, reg_off + E1000_FHFT_QUEUEING_OFFSET,
3995 			queueing);
3996 
3997 	for (i = 0; i < E1000_FLEX_FILTERS_MASK_SIZE; i++) {
3998 		E1000_WRITE_REG(hw, reg_off,
3999 				filter->filter_info.dwords[j]);
4000 		reg_off += sizeof(uint32_t);
4001 		E1000_WRITE_REG(hw, reg_off,
4002 				filter->filter_info.dwords[++j]);
4003 		reg_off += sizeof(uint32_t);
4004 		E1000_WRITE_REG(hw, reg_off,
4005 			(uint32_t)filter->filter_info.mask[i]);
4006 		reg_off += sizeof(uint32_t) * 2;
4007 		++j;
4008 	}
4009 }
4010 
4011 static inline struct e1000_flex_filter *
4012 eth_igb_flex_filter_lookup(struct e1000_flex_filter_list *filter_list,
4013 			struct e1000_flex_filter_info *key)
4014 {
4015 	struct e1000_flex_filter *it;
4016 
4017 	TAILQ_FOREACH(it, filter_list, entries) {
4018 		if (memcmp(key, &it->filter_info,
4019 			sizeof(struct e1000_flex_filter_info)) == 0)
4020 			return it;
4021 	}
4022 
4023 	return NULL;
4024 }
4025 
4026 /* remove a flex byte filter
4027  * @param
4028  * dev: Pointer to struct rte_eth_dev.
4029  * filter: the pointer of the filter will be removed.
4030  */
4031 void
4032 igb_remove_flex_filter(struct rte_eth_dev *dev,
4033 			struct e1000_flex_filter *filter)
4034 {
4035 	struct e1000_filter_info *filter_info =
4036 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4037 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4038 	uint32_t wufc, i;
4039 	uint32_t reg_off;
4040 
4041 	wufc = E1000_READ_REG(hw, E1000_WUFC);
4042 	if (filter->index < E1000_MAX_FHFT)
4043 		reg_off = E1000_FHFT(filter->index);
4044 	else
4045 		reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
4046 
4047 	for (i = 0; i < E1000_FHFT_SIZE_IN_DWD; i++)
4048 		E1000_WRITE_REG(hw, reg_off + i * sizeof(uint32_t), 0);
4049 
4050 	E1000_WRITE_REG(hw, E1000_WUFC, wufc &
4051 		(~(E1000_WUFC_FLX0 << filter->index)));
4052 
4053 	filter_info->flex_mask &= ~(1 << filter->index);
4054 	TAILQ_REMOVE(&filter_info->flex_list, filter, entries);
4055 	rte_free(filter);
4056 }
4057 
4058 int
4059 eth_igb_add_del_flex_filter(struct rte_eth_dev *dev,
4060 			struct rte_eth_flex_filter *filter,
4061 			bool add)
4062 {
4063 	struct e1000_filter_info *filter_info =
4064 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4065 	struct e1000_flex_filter *flex_filter, *it;
4066 	uint32_t mask;
4067 	uint8_t shift, i;
4068 
4069 	flex_filter = rte_zmalloc("e1000_flex_filter",
4070 			sizeof(struct e1000_flex_filter), 0);
4071 	if (flex_filter == NULL)
4072 		return -ENOMEM;
4073 
4074 	flex_filter->filter_info.len = filter->len;
4075 	flex_filter->filter_info.priority = filter->priority;
4076 	memcpy(flex_filter->filter_info.dwords, filter->bytes, filter->len);
4077 	for (i = 0; i < RTE_ALIGN(filter->len, CHAR_BIT) / CHAR_BIT; i++) {
4078 		mask = 0;
4079 		/* reverse bits in flex filter's mask*/
4080 		for (shift = 0; shift < CHAR_BIT; shift++) {
4081 			if (filter->mask[i] & (0x01 << shift))
4082 				mask |= (0x80 >> shift);
4083 		}
4084 		flex_filter->filter_info.mask[i] = mask;
4085 	}
4086 
4087 	it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
4088 				&flex_filter->filter_info);
4089 	if (it == NULL && !add) {
4090 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4091 		rte_free(flex_filter);
4092 		return -ENOENT;
4093 	}
4094 	if (it != NULL && add) {
4095 		PMD_DRV_LOG(ERR, "filter exists.");
4096 		rte_free(flex_filter);
4097 		return -EEXIST;
4098 	}
4099 
4100 	if (add) {
4101 		flex_filter->queue = filter->queue;
4102 		/*
4103 		 * look for an unused flex filter index
4104 		 * and insert the filter into the list.
4105 		 */
4106 		for (i = 0; i < E1000_MAX_FLEX_FILTERS; i++) {
4107 			if (!(filter_info->flex_mask & (1 << i))) {
4108 				filter_info->flex_mask |= 1 << i;
4109 				flex_filter->index = i;
4110 				TAILQ_INSERT_TAIL(&filter_info->flex_list,
4111 					flex_filter,
4112 					entries);
4113 				break;
4114 			}
4115 		}
4116 		if (i >= E1000_MAX_FLEX_FILTERS) {
4117 			PMD_DRV_LOG(ERR, "flex filters are full.");
4118 			rte_free(flex_filter);
4119 			return -ENOSYS;
4120 		}
4121 
4122 		igb_inject_flex_filter(dev, flex_filter);
4123 
4124 	} else {
4125 		igb_remove_flex_filter(dev, it);
4126 		rte_free(flex_filter);
4127 	}
4128 
4129 	return 0;
4130 }
4131 
4132 static int
4133 eth_igb_get_flex_filter(struct rte_eth_dev *dev,
4134 			struct rte_eth_flex_filter *filter)
4135 {
4136 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4137 	struct e1000_filter_info *filter_info =
4138 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4139 	struct e1000_flex_filter flex_filter, *it;
4140 	uint32_t wufc, queueing, wufc_en = 0;
4141 
4142 	memset(&flex_filter, 0, sizeof(struct e1000_flex_filter));
4143 	flex_filter.filter_info.len = filter->len;
4144 	flex_filter.filter_info.priority = filter->priority;
4145 	memcpy(flex_filter.filter_info.dwords, filter->bytes, filter->len);
4146 	memcpy(flex_filter.filter_info.mask, filter->mask,
4147 			RTE_ALIGN(filter->len, CHAR_BIT) / CHAR_BIT);
4148 
4149 	it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
4150 				&flex_filter.filter_info);
4151 	if (it == NULL) {
4152 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4153 		return -ENOENT;
4154 	}
4155 
4156 	wufc = E1000_READ_REG(hw, E1000_WUFC);
4157 	wufc_en = E1000_WUFC_FLEX_HQ | (E1000_WUFC_FLX0 << it->index);
4158 
4159 	if ((wufc & wufc_en) == wufc_en) {
4160 		uint32_t reg_off = 0;
4161 		if (it->index < E1000_MAX_FHFT)
4162 			reg_off = E1000_FHFT(it->index);
4163 		else
4164 			reg_off = E1000_FHFT_EXT(it->index - E1000_MAX_FHFT);
4165 
4166 		queueing = E1000_READ_REG(hw,
4167 				reg_off + E1000_FHFT_QUEUEING_OFFSET);
4168 		filter->len = queueing & E1000_FHFT_QUEUEING_LEN;
4169 		filter->priority = (queueing & E1000_FHFT_QUEUEING_PRIO) >>
4170 			E1000_FHFT_QUEUEING_PRIO_SHIFT;
4171 		filter->queue = (queueing & E1000_FHFT_QUEUEING_QUEUE) >>
4172 			E1000_FHFT_QUEUEING_QUEUE_SHIFT;
4173 		return 0;
4174 	}
4175 	return -ENOENT;
4176 }
4177 
4178 static int
4179 eth_igb_flex_filter_handle(struct rte_eth_dev *dev,
4180 			enum rte_filter_op filter_op,
4181 			void *arg)
4182 {
4183 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4184 	struct rte_eth_flex_filter *filter;
4185 	int ret = 0;
4186 
4187 	MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
4188 
4189 	if (filter_op == RTE_ETH_FILTER_NOP)
4190 		return ret;
4191 
4192 	if (arg == NULL) {
4193 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
4194 			    filter_op);
4195 		return -EINVAL;
4196 	}
4197 
4198 	filter = (struct rte_eth_flex_filter *)arg;
4199 	if (filter->len == 0 || filter->len > E1000_MAX_FLEX_FILTER_LEN
4200 	    || filter->len % sizeof(uint64_t) != 0) {
4201 		PMD_DRV_LOG(ERR, "filter's length is out of range");
4202 		return -EINVAL;
4203 	}
4204 	if (filter->priority > E1000_MAX_FLEX_FILTER_PRI) {
4205 		PMD_DRV_LOG(ERR, "filter's priority is out of range");
4206 		return -EINVAL;
4207 	}
4208 
4209 	switch (filter_op) {
4210 	case RTE_ETH_FILTER_ADD:
4211 		ret = eth_igb_add_del_flex_filter(dev, filter, TRUE);
4212 		break;
4213 	case RTE_ETH_FILTER_DELETE:
4214 		ret = eth_igb_add_del_flex_filter(dev, filter, FALSE);
4215 		break;
4216 	case RTE_ETH_FILTER_GET:
4217 		ret = eth_igb_get_flex_filter(dev, filter);
4218 		break;
4219 	default:
4220 		PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
4221 		ret = -EINVAL;
4222 		break;
4223 	}
4224 
4225 	return ret;
4226 }
4227 
4228 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_5tuple_filter_info*/
4229 static inline int
4230 ntuple_filter_to_5tuple_82576(struct rte_eth_ntuple_filter *filter,
4231 			struct e1000_5tuple_filter_info *filter_info)
4232 {
4233 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM_82576)
4234 		return -EINVAL;
4235 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
4236 		return -EINVAL;  /* filter index is out of range. */
4237 	if (filter->tcp_flags > RTE_NTUPLE_TCP_FLAGS_MASK)
4238 		return -EINVAL;  /* flags is invalid. */
4239 
4240 	switch (filter->dst_ip_mask) {
4241 	case UINT32_MAX:
4242 		filter_info->dst_ip_mask = 0;
4243 		filter_info->dst_ip = filter->dst_ip;
4244 		break;
4245 	case 0:
4246 		filter_info->dst_ip_mask = 1;
4247 		break;
4248 	default:
4249 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
4250 		return -EINVAL;
4251 	}
4252 
4253 	switch (filter->src_ip_mask) {
4254 	case UINT32_MAX:
4255 		filter_info->src_ip_mask = 0;
4256 		filter_info->src_ip = filter->src_ip;
4257 		break;
4258 	case 0:
4259 		filter_info->src_ip_mask = 1;
4260 		break;
4261 	default:
4262 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
4263 		return -EINVAL;
4264 	}
4265 
4266 	switch (filter->dst_port_mask) {
4267 	case UINT16_MAX:
4268 		filter_info->dst_port_mask = 0;
4269 		filter_info->dst_port = filter->dst_port;
4270 		break;
4271 	case 0:
4272 		filter_info->dst_port_mask = 1;
4273 		break;
4274 	default:
4275 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
4276 		return -EINVAL;
4277 	}
4278 
4279 	switch (filter->src_port_mask) {
4280 	case UINT16_MAX:
4281 		filter_info->src_port_mask = 0;
4282 		filter_info->src_port = filter->src_port;
4283 		break;
4284 	case 0:
4285 		filter_info->src_port_mask = 1;
4286 		break;
4287 	default:
4288 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
4289 		return -EINVAL;
4290 	}
4291 
4292 	switch (filter->proto_mask) {
4293 	case UINT8_MAX:
4294 		filter_info->proto_mask = 0;
4295 		filter_info->proto = filter->proto;
4296 		break;
4297 	case 0:
4298 		filter_info->proto_mask = 1;
4299 		break;
4300 	default:
4301 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
4302 		return -EINVAL;
4303 	}
4304 
4305 	filter_info->priority = (uint8_t)filter->priority;
4306 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
4307 		filter_info->tcp_flags = filter->tcp_flags;
4308 	else
4309 		filter_info->tcp_flags = 0;
4310 
4311 	return 0;
4312 }
4313 
4314 static inline struct e1000_5tuple_filter *
4315 igb_5tuple_filter_lookup_82576(struct e1000_5tuple_filter_list *filter_list,
4316 			struct e1000_5tuple_filter_info *key)
4317 {
4318 	struct e1000_5tuple_filter *it;
4319 
4320 	TAILQ_FOREACH(it, filter_list, entries) {
4321 		if (memcmp(key, &it->filter_info,
4322 			sizeof(struct e1000_5tuple_filter_info)) == 0) {
4323 			return it;
4324 		}
4325 	}
4326 	return NULL;
4327 }
4328 
4329 /* inject a igb 5-tuple filter to HW */
4330 static inline void
4331 igb_inject_5tuple_filter_82576(struct rte_eth_dev *dev,
4332 			   struct e1000_5tuple_filter *filter)
4333 {
4334 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4335 	uint32_t ftqf = E1000_FTQF_VF_BP | E1000_FTQF_MASK;
4336 	uint32_t spqf, imir, imir_ext = E1000_IMIREXT_SIZE_BP;
4337 	uint8_t i;
4338 
4339 	i = filter->index;
4340 	ftqf |= filter->filter_info.proto & E1000_FTQF_PROTOCOL_MASK;
4341 	if (filter->filter_info.src_ip_mask == 0) /* 0b means compare. */
4342 		ftqf &= ~E1000_FTQF_MASK_SOURCE_ADDR_BP;
4343 	if (filter->filter_info.dst_ip_mask == 0)
4344 		ftqf &= ~E1000_FTQF_MASK_DEST_ADDR_BP;
4345 	if (filter->filter_info.src_port_mask == 0)
4346 		ftqf &= ~E1000_FTQF_MASK_SOURCE_PORT_BP;
4347 	if (filter->filter_info.proto_mask == 0)
4348 		ftqf &= ~E1000_FTQF_MASK_PROTO_BP;
4349 	ftqf |= (filter->queue << E1000_FTQF_QUEUE_SHIFT) &
4350 		E1000_FTQF_QUEUE_MASK;
4351 	ftqf |= E1000_FTQF_QUEUE_ENABLE;
4352 	E1000_WRITE_REG(hw, E1000_FTQF(i), ftqf);
4353 	E1000_WRITE_REG(hw, E1000_DAQF(i), filter->filter_info.dst_ip);
4354 	E1000_WRITE_REG(hw, E1000_SAQF(i), filter->filter_info.src_ip);
4355 
4356 	spqf = filter->filter_info.src_port & E1000_SPQF_SRCPORT;
4357 	E1000_WRITE_REG(hw, E1000_SPQF(i), spqf);
4358 
4359 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
4360 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
4361 		imir |= E1000_IMIR_PORT_BP;
4362 	else
4363 		imir &= ~E1000_IMIR_PORT_BP;
4364 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
4365 
4366 	/* tcp flags bits setting. */
4367 	if (filter->filter_info.tcp_flags & RTE_NTUPLE_TCP_FLAGS_MASK) {
4368 		if (filter->filter_info.tcp_flags & RTE_TCP_URG_FLAG)
4369 			imir_ext |= E1000_IMIREXT_CTRL_URG;
4370 		if (filter->filter_info.tcp_flags & RTE_TCP_ACK_FLAG)
4371 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
4372 		if (filter->filter_info.tcp_flags & RTE_TCP_PSH_FLAG)
4373 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
4374 		if (filter->filter_info.tcp_flags & RTE_TCP_RST_FLAG)
4375 			imir_ext |= E1000_IMIREXT_CTRL_RST;
4376 		if (filter->filter_info.tcp_flags & RTE_TCP_SYN_FLAG)
4377 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
4378 		if (filter->filter_info.tcp_flags & RTE_TCP_FIN_FLAG)
4379 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
4380 	} else {
4381 		imir_ext |= E1000_IMIREXT_CTRL_BP;
4382 	}
4383 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
4384 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
4385 }
4386 
4387 /*
4388  * igb_add_5tuple_filter_82576 - add a 5tuple filter
4389  *
4390  * @param
4391  * dev: Pointer to struct rte_eth_dev.
4392  * ntuple_filter: ponter to the filter that will be added.
4393  *
4394  * @return
4395  *    - On success, zero.
4396  *    - On failure, a negative value.
4397  */
4398 static int
4399 igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
4400 			struct rte_eth_ntuple_filter *ntuple_filter)
4401 {
4402 	struct e1000_filter_info *filter_info =
4403 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4404 	struct e1000_5tuple_filter *filter;
4405 	uint8_t i;
4406 	int ret;
4407 
4408 	filter = rte_zmalloc("e1000_5tuple_filter",
4409 			sizeof(struct e1000_5tuple_filter), 0);
4410 	if (filter == NULL)
4411 		return -ENOMEM;
4412 
4413 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4414 					    &filter->filter_info);
4415 	if (ret < 0) {
4416 		rte_free(filter);
4417 		return ret;
4418 	}
4419 
4420 	if (igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4421 					 &filter->filter_info) != NULL) {
4422 		PMD_DRV_LOG(ERR, "filter exists.");
4423 		rte_free(filter);
4424 		return -EEXIST;
4425 	}
4426 	filter->queue = ntuple_filter->queue;
4427 
4428 	/*
4429 	 * look for an unused 5tuple filter index,
4430 	 * and insert the filter to list.
4431 	 */
4432 	for (i = 0; i < E1000_MAX_FTQF_FILTERS; i++) {
4433 		if (!(filter_info->fivetuple_mask & (1 << i))) {
4434 			filter_info->fivetuple_mask |= 1 << i;
4435 			filter->index = i;
4436 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
4437 					  filter,
4438 					  entries);
4439 			break;
4440 		}
4441 	}
4442 	if (i >= E1000_MAX_FTQF_FILTERS) {
4443 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
4444 		rte_free(filter);
4445 		return -ENOSYS;
4446 	}
4447 
4448 	igb_inject_5tuple_filter_82576(dev, filter);
4449 	return 0;
4450 }
4451 
4452 int
4453 igb_delete_5tuple_filter_82576(struct rte_eth_dev *dev,
4454 				struct e1000_5tuple_filter *filter)
4455 {
4456 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4457 	struct e1000_filter_info *filter_info =
4458 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4459 
4460 	filter_info->fivetuple_mask &= ~(1 << filter->index);
4461 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
4462 	rte_free(filter);
4463 
4464 	E1000_WRITE_REG(hw, E1000_FTQF(filter->index),
4465 			E1000_FTQF_VF_BP | E1000_FTQF_MASK);
4466 	E1000_WRITE_REG(hw, E1000_DAQF(filter->index), 0);
4467 	E1000_WRITE_REG(hw, E1000_SAQF(filter->index), 0);
4468 	E1000_WRITE_REG(hw, E1000_SPQF(filter->index), 0);
4469 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
4470 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
4471 	return 0;
4472 }
4473 
4474 /*
4475  * igb_remove_5tuple_filter_82576 - remove a 5tuple filter
4476  *
4477  * @param
4478  * dev: Pointer to struct rte_eth_dev.
4479  * ntuple_filter: ponter to the filter that will be removed.
4480  *
4481  * @return
4482  *    - On success, zero.
4483  *    - On failure, a negative value.
4484  */
4485 static int
4486 igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
4487 				struct rte_eth_ntuple_filter *ntuple_filter)
4488 {
4489 	struct e1000_filter_info *filter_info =
4490 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4491 	struct e1000_5tuple_filter_info filter_5tuple;
4492 	struct e1000_5tuple_filter *filter;
4493 	int ret;
4494 
4495 	memset(&filter_5tuple, 0, sizeof(struct e1000_5tuple_filter_info));
4496 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4497 					    &filter_5tuple);
4498 	if (ret < 0)
4499 		return ret;
4500 
4501 	filter = igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4502 					 &filter_5tuple);
4503 	if (filter == NULL) {
4504 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4505 		return -ENOENT;
4506 	}
4507 
4508 	igb_delete_5tuple_filter_82576(dev, filter);
4509 
4510 	return 0;
4511 }
4512 
4513 static int
4514 eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4515 {
4516 	uint32_t rctl;
4517 	struct e1000_hw *hw;
4518 	struct rte_eth_dev_info dev_info;
4519 	uint32_t frame_size = mtu + E1000_ETH_OVERHEAD;
4520 	int ret;
4521 
4522 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4523 
4524 #ifdef RTE_LIBRTE_82571_SUPPORT
4525 	/* XXX: not bigger than max_rx_pktlen */
4526 	if (hw->mac.type == e1000_82571)
4527 		return -ENOTSUP;
4528 #endif
4529 	ret = eth_igb_infos_get(dev, &dev_info);
4530 	if (ret != 0)
4531 		return ret;
4532 
4533 	/* check that mtu is within the allowed range */
4534 	if (mtu < RTE_ETHER_MIN_MTU ||
4535 			frame_size > dev_info.max_rx_pktlen)
4536 		return -EINVAL;
4537 
4538 	/* refuse mtu that requires the support of scattered packets when this
4539 	 * feature has not been enabled before. */
4540 	if (!dev->data->scattered_rx &&
4541 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
4542 		return -EINVAL;
4543 
4544 	rctl = E1000_READ_REG(hw, E1000_RCTL);
4545 
4546 	/* switch to jumbo mode if needed */
4547 	if (frame_size > RTE_ETHER_MAX_LEN) {
4548 		dev->data->dev_conf.rxmode.offloads |=
4549 			DEV_RX_OFFLOAD_JUMBO_FRAME;
4550 		rctl |= E1000_RCTL_LPE;
4551 	} else {
4552 		dev->data->dev_conf.rxmode.offloads &=
4553 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
4554 		rctl &= ~E1000_RCTL_LPE;
4555 	}
4556 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
4557 
4558 	/* update max frame size */
4559 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4560 
4561 	E1000_WRITE_REG(hw, E1000_RLPML,
4562 			dev->data->dev_conf.rxmode.max_rx_pkt_len);
4563 
4564 	return 0;
4565 }
4566 
4567 /*
4568  * igb_add_del_ntuple_filter - add or delete a ntuple filter
4569  *
4570  * @param
4571  * dev: Pointer to struct rte_eth_dev.
4572  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4573  * add: if true, add filter, if false, remove filter
4574  *
4575  * @return
4576  *    - On success, zero.
4577  *    - On failure, a negative value.
4578  */
4579 int
4580 igb_add_del_ntuple_filter(struct rte_eth_dev *dev,
4581 			struct rte_eth_ntuple_filter *ntuple_filter,
4582 			bool add)
4583 {
4584 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4585 	int ret;
4586 
4587 	switch (ntuple_filter->flags) {
4588 	case RTE_5TUPLE_FLAGS:
4589 	case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4590 		if (hw->mac.type != e1000_82576)
4591 			return -ENOTSUP;
4592 		if (add)
4593 			ret = igb_add_5tuple_filter_82576(dev,
4594 							  ntuple_filter);
4595 		else
4596 			ret = igb_remove_5tuple_filter_82576(dev,
4597 							     ntuple_filter);
4598 		break;
4599 	case RTE_2TUPLE_FLAGS:
4600 	case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4601 		if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350 &&
4602 			hw->mac.type != e1000_i210 &&
4603 			hw->mac.type != e1000_i211)
4604 			return -ENOTSUP;
4605 		if (add)
4606 			ret = igb_add_2tuple_filter(dev, ntuple_filter);
4607 		else
4608 			ret = igb_remove_2tuple_filter(dev, ntuple_filter);
4609 		break;
4610 	default:
4611 		ret = -EINVAL;
4612 		break;
4613 	}
4614 
4615 	return ret;
4616 }
4617 
4618 /*
4619  * igb_get_ntuple_filter - get a ntuple filter
4620  *
4621  * @param
4622  * dev: Pointer to struct rte_eth_dev.
4623  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4624  *
4625  * @return
4626  *    - On success, zero.
4627  *    - On failure, a negative value.
4628  */
4629 static int
4630 igb_get_ntuple_filter(struct rte_eth_dev *dev,
4631 			struct rte_eth_ntuple_filter *ntuple_filter)
4632 {
4633 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4634 	struct e1000_filter_info *filter_info =
4635 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4636 	struct e1000_5tuple_filter_info filter_5tuple;
4637 	struct e1000_2tuple_filter_info filter_2tuple;
4638 	struct e1000_5tuple_filter *p_5tuple_filter;
4639 	struct e1000_2tuple_filter *p_2tuple_filter;
4640 	int ret;
4641 
4642 	switch (ntuple_filter->flags) {
4643 	case RTE_5TUPLE_FLAGS:
4644 	case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4645 		if (hw->mac.type != e1000_82576)
4646 			return -ENOTSUP;
4647 		memset(&filter_5tuple,
4648 			0,
4649 			sizeof(struct e1000_5tuple_filter_info));
4650 		ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4651 						    &filter_5tuple);
4652 		if (ret < 0)
4653 			return ret;
4654 		p_5tuple_filter = igb_5tuple_filter_lookup_82576(
4655 					&filter_info->fivetuple_list,
4656 					&filter_5tuple);
4657 		if (p_5tuple_filter == NULL) {
4658 			PMD_DRV_LOG(ERR, "filter doesn't exist.");
4659 			return -ENOENT;
4660 		}
4661 		ntuple_filter->queue = p_5tuple_filter->queue;
4662 		break;
4663 	case RTE_2TUPLE_FLAGS:
4664 	case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4665 		if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350)
4666 			return -ENOTSUP;
4667 		memset(&filter_2tuple,
4668 			0,
4669 			sizeof(struct e1000_2tuple_filter_info));
4670 		ret = ntuple_filter_to_2tuple(ntuple_filter, &filter_2tuple);
4671 		if (ret < 0)
4672 			return ret;
4673 		p_2tuple_filter = igb_2tuple_filter_lookup(
4674 					&filter_info->twotuple_list,
4675 					&filter_2tuple);
4676 		if (p_2tuple_filter == NULL) {
4677 			PMD_DRV_LOG(ERR, "filter doesn't exist.");
4678 			return -ENOENT;
4679 		}
4680 		ntuple_filter->queue = p_2tuple_filter->queue;
4681 		break;
4682 	default:
4683 		ret = -EINVAL;
4684 		break;
4685 	}
4686 
4687 	return 0;
4688 }
4689 
4690 /*
4691  * igb_ntuple_filter_handle - Handle operations for ntuple filter.
4692  * @dev: pointer to rte_eth_dev structure
4693  * @filter_op:operation will be taken.
4694  * @arg: a pointer to specific structure corresponding to the filter_op
4695  */
4696 static int
4697 igb_ntuple_filter_handle(struct rte_eth_dev *dev,
4698 				enum rte_filter_op filter_op,
4699 				void *arg)
4700 {
4701 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4702 	int ret;
4703 
4704 	MAC_TYPE_FILTER_SUP(hw->mac.type);
4705 
4706 	if (filter_op == RTE_ETH_FILTER_NOP)
4707 		return 0;
4708 
4709 	if (arg == NULL) {
4710 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4711 			    filter_op);
4712 		return -EINVAL;
4713 	}
4714 
4715 	switch (filter_op) {
4716 	case RTE_ETH_FILTER_ADD:
4717 		ret = igb_add_del_ntuple_filter(dev,
4718 			(struct rte_eth_ntuple_filter *)arg,
4719 			TRUE);
4720 		break;
4721 	case RTE_ETH_FILTER_DELETE:
4722 		ret = igb_add_del_ntuple_filter(dev,
4723 			(struct rte_eth_ntuple_filter *)arg,
4724 			FALSE);
4725 		break;
4726 	case RTE_ETH_FILTER_GET:
4727 		ret = igb_get_ntuple_filter(dev,
4728 			(struct rte_eth_ntuple_filter *)arg);
4729 		break;
4730 	default:
4731 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
4732 		ret = -EINVAL;
4733 		break;
4734 	}
4735 	return ret;
4736 }
4737 
4738 static inline int
4739 igb_ethertype_filter_lookup(struct e1000_filter_info *filter_info,
4740 			uint16_t ethertype)
4741 {
4742 	int i;
4743 
4744 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4745 		if (filter_info->ethertype_filters[i].ethertype == ethertype &&
4746 		    (filter_info->ethertype_mask & (1 << i)))
4747 			return i;
4748 	}
4749 	return -1;
4750 }
4751 
4752 static inline int
4753 igb_ethertype_filter_insert(struct e1000_filter_info *filter_info,
4754 			uint16_t ethertype, uint32_t etqf)
4755 {
4756 	int i;
4757 
4758 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4759 		if (!(filter_info->ethertype_mask & (1 << i))) {
4760 			filter_info->ethertype_mask |= 1 << i;
4761 			filter_info->ethertype_filters[i].ethertype = ethertype;
4762 			filter_info->ethertype_filters[i].etqf = etqf;
4763 			return i;
4764 		}
4765 	}
4766 	return -1;
4767 }
4768 
4769 int
4770 igb_ethertype_filter_remove(struct e1000_filter_info *filter_info,
4771 			uint8_t idx)
4772 {
4773 	if (idx >= E1000_MAX_ETQF_FILTERS)
4774 		return -1;
4775 	filter_info->ethertype_mask &= ~(1 << idx);
4776 	filter_info->ethertype_filters[idx].ethertype = 0;
4777 	filter_info->ethertype_filters[idx].etqf = 0;
4778 	return idx;
4779 }
4780 
4781 
4782 int
4783 igb_add_del_ethertype_filter(struct rte_eth_dev *dev,
4784 			struct rte_eth_ethertype_filter *filter,
4785 			bool add)
4786 {
4787 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4788 	struct e1000_filter_info *filter_info =
4789 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4790 	uint32_t etqf = 0;
4791 	int ret;
4792 
4793 	if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
4794 		filter->ether_type == RTE_ETHER_TYPE_IPV6) {
4795 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
4796 			" ethertype filter.", filter->ether_type);
4797 		return -EINVAL;
4798 	}
4799 
4800 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
4801 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
4802 		return -EINVAL;
4803 	}
4804 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
4805 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
4806 		return -EINVAL;
4807 	}
4808 
4809 	ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4810 	if (ret >= 0 && add) {
4811 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
4812 			    filter->ether_type);
4813 		return -EEXIST;
4814 	}
4815 	if (ret < 0 && !add) {
4816 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4817 			    filter->ether_type);
4818 		return -ENOENT;
4819 	}
4820 
4821 	if (add) {
4822 		etqf |= E1000_ETQF_FILTER_ENABLE | E1000_ETQF_QUEUE_ENABLE;
4823 		etqf |= (uint32_t)(filter->ether_type & E1000_ETQF_ETHERTYPE);
4824 		etqf |= filter->queue << E1000_ETQF_QUEUE_SHIFT;
4825 		ret = igb_ethertype_filter_insert(filter_info,
4826 				filter->ether_type, etqf);
4827 		if (ret < 0) {
4828 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
4829 			return -ENOSYS;
4830 		}
4831 	} else {
4832 		ret = igb_ethertype_filter_remove(filter_info, (uint8_t)ret);
4833 		if (ret < 0)
4834 			return -ENOSYS;
4835 	}
4836 	E1000_WRITE_REG(hw, E1000_ETQF(ret), etqf);
4837 	E1000_WRITE_FLUSH(hw);
4838 
4839 	return 0;
4840 }
4841 
4842 static int
4843 igb_get_ethertype_filter(struct rte_eth_dev *dev,
4844 			struct rte_eth_ethertype_filter *filter)
4845 {
4846 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4847 	struct e1000_filter_info *filter_info =
4848 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4849 	uint32_t etqf;
4850 	int ret;
4851 
4852 	ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4853 	if (ret < 0) {
4854 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4855 			    filter->ether_type);
4856 		return -ENOENT;
4857 	}
4858 
4859 	etqf = E1000_READ_REG(hw, E1000_ETQF(ret));
4860 	if (etqf & E1000_ETQF_FILTER_ENABLE) {
4861 		filter->ether_type = etqf & E1000_ETQF_ETHERTYPE;
4862 		filter->flags = 0;
4863 		filter->queue = (etqf & E1000_ETQF_QUEUE) >>
4864 				E1000_ETQF_QUEUE_SHIFT;
4865 		return 0;
4866 	}
4867 
4868 	return -ENOENT;
4869 }
4870 
4871 /*
4872  * igb_ethertype_filter_handle - Handle operations for ethertype filter.
4873  * @dev: pointer to rte_eth_dev structure
4874  * @filter_op:operation will be taken.
4875  * @arg: a pointer to specific structure corresponding to the filter_op
4876  */
4877 static int
4878 igb_ethertype_filter_handle(struct rte_eth_dev *dev,
4879 				enum rte_filter_op filter_op,
4880 				void *arg)
4881 {
4882 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4883 	int ret;
4884 
4885 	MAC_TYPE_FILTER_SUP(hw->mac.type);
4886 
4887 	if (filter_op == RTE_ETH_FILTER_NOP)
4888 		return 0;
4889 
4890 	if (arg == NULL) {
4891 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4892 			    filter_op);
4893 		return -EINVAL;
4894 	}
4895 
4896 	switch (filter_op) {
4897 	case RTE_ETH_FILTER_ADD:
4898 		ret = igb_add_del_ethertype_filter(dev,
4899 			(struct rte_eth_ethertype_filter *)arg,
4900 			TRUE);
4901 		break;
4902 	case RTE_ETH_FILTER_DELETE:
4903 		ret = igb_add_del_ethertype_filter(dev,
4904 			(struct rte_eth_ethertype_filter *)arg,
4905 			FALSE);
4906 		break;
4907 	case RTE_ETH_FILTER_GET:
4908 		ret = igb_get_ethertype_filter(dev,
4909 			(struct rte_eth_ethertype_filter *)arg);
4910 		break;
4911 	default:
4912 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
4913 		ret = -EINVAL;
4914 		break;
4915 	}
4916 	return ret;
4917 }
4918 
4919 static int
4920 eth_igb_filter_ctrl(struct rte_eth_dev *dev,
4921 		     enum rte_filter_type filter_type,
4922 		     enum rte_filter_op filter_op,
4923 		     void *arg)
4924 {
4925 	int ret = 0;
4926 
4927 	switch (filter_type) {
4928 	case RTE_ETH_FILTER_NTUPLE:
4929 		ret = igb_ntuple_filter_handle(dev, filter_op, arg);
4930 		break;
4931 	case RTE_ETH_FILTER_ETHERTYPE:
4932 		ret = igb_ethertype_filter_handle(dev, filter_op, arg);
4933 		break;
4934 	case RTE_ETH_FILTER_SYN:
4935 		ret = eth_igb_syn_filter_handle(dev, filter_op, arg);
4936 		break;
4937 	case RTE_ETH_FILTER_FLEXIBLE:
4938 		ret = eth_igb_flex_filter_handle(dev, filter_op, arg);
4939 		break;
4940 	case RTE_ETH_FILTER_GENERIC:
4941 		if (filter_op != RTE_ETH_FILTER_GET)
4942 			return -EINVAL;
4943 		*(const void **)arg = &igb_flow_ops;
4944 		break;
4945 	default:
4946 		PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
4947 							filter_type);
4948 		break;
4949 	}
4950 
4951 	return ret;
4952 }
4953 
4954 static int
4955 eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
4956 			 struct rte_ether_addr *mc_addr_set,
4957 			 uint32_t nb_mc_addr)
4958 {
4959 	struct e1000_hw *hw;
4960 
4961 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4962 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
4963 	return 0;
4964 }
4965 
4966 static uint64_t
4967 igb_read_systime_cyclecounter(struct rte_eth_dev *dev)
4968 {
4969 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4970 	uint64_t systime_cycles;
4971 
4972 	switch (hw->mac.type) {
4973 	case e1000_i210:
4974 	case e1000_i211:
4975 		/*
4976 		 * Need to read System Time Residue Register to be able
4977 		 * to read the other two registers.
4978 		 */
4979 		E1000_READ_REG(hw, E1000_SYSTIMR);
4980 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
4981 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4982 		systime_cycles += (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4983 				* NSEC_PER_SEC;
4984 		break;
4985 	case e1000_82580:
4986 	case e1000_i350:
4987 	case e1000_i354:
4988 		/*
4989 		 * Need to read System Time Residue Register to be able
4990 		 * to read the other two registers.
4991 		 */
4992 		E1000_READ_REG(hw, E1000_SYSTIMR);
4993 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4994 		/* Only the 8 LSB are valid. */
4995 		systime_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_SYSTIMH)
4996 				& 0xff) << 32;
4997 		break;
4998 	default:
4999 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
5000 		systime_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
5001 				<< 32;
5002 		break;
5003 	}
5004 
5005 	return systime_cycles;
5006 }
5007 
5008 static uint64_t
5009 igb_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
5010 {
5011 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5012 	uint64_t rx_tstamp_cycles;
5013 
5014 	switch (hw->mac.type) {
5015 	case e1000_i210:
5016 	case e1000_i211:
5017 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
5018 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
5019 		rx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
5020 				* NSEC_PER_SEC;
5021 		break;
5022 	case e1000_82580:
5023 	case e1000_i350:
5024 	case e1000_i354:
5025 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
5026 		/* Only the 8 LSB are valid. */
5027 		rx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_RXSTMPH)
5028 				& 0xff) << 32;
5029 		break;
5030 	default:
5031 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
5032 		rx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
5033 				<< 32;
5034 		break;
5035 	}
5036 
5037 	return rx_tstamp_cycles;
5038 }
5039 
5040 static uint64_t
5041 igb_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
5042 {
5043 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5044 	uint64_t tx_tstamp_cycles;
5045 
5046 	switch (hw->mac.type) {
5047 	case e1000_i210:
5048 	case e1000_i211:
5049 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
5050 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5051 		tx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
5052 				* NSEC_PER_SEC;
5053 		break;
5054 	case e1000_82580:
5055 	case e1000_i350:
5056 	case e1000_i354:
5057 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5058 		/* Only the 8 LSB are valid. */
5059 		tx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_TXSTMPH)
5060 				& 0xff) << 32;
5061 		break;
5062 	default:
5063 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5064 		tx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
5065 				<< 32;
5066 		break;
5067 	}
5068 
5069 	return tx_tstamp_cycles;
5070 }
5071 
5072 static void
5073 igb_start_timecounters(struct rte_eth_dev *dev)
5074 {
5075 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5076 	struct e1000_adapter *adapter = dev->data->dev_private;
5077 	uint32_t incval = 1;
5078 	uint32_t shift = 0;
5079 	uint64_t mask = E1000_CYCLECOUNTER_MASK;
5080 
5081 	switch (hw->mac.type) {
5082 	case e1000_82580:
5083 	case e1000_i350:
5084 	case e1000_i354:
5085 		/* 32 LSB bits + 8 MSB bits = 40 bits */
5086 		mask = (1ULL << 40) - 1;
5087 		/* fall-through */
5088 	case e1000_i210:
5089 	case e1000_i211:
5090 		/*
5091 		 * Start incrementing the register
5092 		 * used to timestamp PTP packets.
5093 		 */
5094 		E1000_WRITE_REG(hw, E1000_TIMINCA, incval);
5095 		break;
5096 	case e1000_82576:
5097 		incval = E1000_INCVALUE_82576;
5098 		shift = IGB_82576_TSYNC_SHIFT;
5099 		E1000_WRITE_REG(hw, E1000_TIMINCA,
5100 				E1000_INCPERIOD_82576 | incval);
5101 		break;
5102 	default:
5103 		/* Not supported */
5104 		return;
5105 	}
5106 
5107 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
5108 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
5109 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
5110 
5111 	adapter->systime_tc.cc_mask = mask;
5112 	adapter->systime_tc.cc_shift = shift;
5113 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
5114 
5115 	adapter->rx_tstamp_tc.cc_mask = mask;
5116 	adapter->rx_tstamp_tc.cc_shift = shift;
5117 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
5118 
5119 	adapter->tx_tstamp_tc.cc_mask = mask;
5120 	adapter->tx_tstamp_tc.cc_shift = shift;
5121 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
5122 }
5123 
5124 static int
5125 igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
5126 {
5127 	struct e1000_adapter *adapter = dev->data->dev_private;
5128 
5129 	adapter->systime_tc.nsec += delta;
5130 	adapter->rx_tstamp_tc.nsec += delta;
5131 	adapter->tx_tstamp_tc.nsec += delta;
5132 
5133 	return 0;
5134 }
5135 
5136 static int
5137 igb_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
5138 {
5139 	uint64_t ns;
5140 	struct e1000_adapter *adapter = dev->data->dev_private;
5141 
5142 	ns = rte_timespec_to_ns(ts);
5143 
5144 	/* Set the timecounters to a new value. */
5145 	adapter->systime_tc.nsec = ns;
5146 	adapter->rx_tstamp_tc.nsec = ns;
5147 	adapter->tx_tstamp_tc.nsec = ns;
5148 
5149 	return 0;
5150 }
5151 
5152 static int
5153 igb_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
5154 {
5155 	uint64_t ns, systime_cycles;
5156 	struct e1000_adapter *adapter = dev->data->dev_private;
5157 
5158 	systime_cycles = igb_read_systime_cyclecounter(dev);
5159 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
5160 	*ts = rte_ns_to_timespec(ns);
5161 
5162 	return 0;
5163 }
5164 
5165 static int
5166 igb_timesync_enable(struct rte_eth_dev *dev)
5167 {
5168 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5169 	uint32_t tsync_ctl;
5170 	uint32_t tsauxc;
5171 
5172 	/* Stop the timesync system time. */
5173 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0x0);
5174 	/* Reset the timesync system time value. */
5175 	switch (hw->mac.type) {
5176 	case e1000_82580:
5177 	case e1000_i350:
5178 	case e1000_i354:
5179 	case e1000_i210:
5180 	case e1000_i211:
5181 		E1000_WRITE_REG(hw, E1000_SYSTIMR, 0x0);
5182 		/* fall-through */
5183 	case e1000_82576:
5184 		E1000_WRITE_REG(hw, E1000_SYSTIML, 0x0);
5185 		E1000_WRITE_REG(hw, E1000_SYSTIMH, 0x0);
5186 		break;
5187 	default:
5188 		/* Not supported. */
5189 		return -ENOTSUP;
5190 	}
5191 
5192 	/* Enable system time for it isn't on by default. */
5193 	tsauxc = E1000_READ_REG(hw, E1000_TSAUXC);
5194 	tsauxc &= ~E1000_TSAUXC_DISABLE_SYSTIME;
5195 	E1000_WRITE_REG(hw, E1000_TSAUXC, tsauxc);
5196 
5197 	igb_start_timecounters(dev);
5198 
5199 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5200 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588),
5201 			(RTE_ETHER_TYPE_1588 |
5202 			 E1000_ETQF_FILTER_ENABLE |
5203 			 E1000_ETQF_1588));
5204 
5205 	/* Enable timestamping of received PTP packets. */
5206 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5207 	tsync_ctl |= E1000_TSYNCRXCTL_ENABLED;
5208 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
5209 
5210 	/* Enable Timestamping of transmitted PTP packets. */
5211 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5212 	tsync_ctl |= E1000_TSYNCTXCTL_ENABLED;
5213 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
5214 
5215 	return 0;
5216 }
5217 
5218 static int
5219 igb_timesync_disable(struct rte_eth_dev *dev)
5220 {
5221 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5222 	uint32_t tsync_ctl;
5223 
5224 	/* Disable timestamping of transmitted PTP packets. */
5225 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5226 	tsync_ctl &= ~E1000_TSYNCTXCTL_ENABLED;
5227 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
5228 
5229 	/* Disable timestamping of received PTP packets. */
5230 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5231 	tsync_ctl &= ~E1000_TSYNCRXCTL_ENABLED;
5232 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
5233 
5234 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5235 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588), 0);
5236 
5237 	/* Stop incrementating the System Time registers. */
5238 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0);
5239 
5240 	return 0;
5241 }
5242 
5243 static int
5244 igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
5245 			       struct timespec *timestamp,
5246 			       uint32_t flags __rte_unused)
5247 {
5248 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5249 	struct e1000_adapter *adapter = dev->data->dev_private;
5250 	uint32_t tsync_rxctl;
5251 	uint64_t rx_tstamp_cycles;
5252 	uint64_t ns;
5253 
5254 	tsync_rxctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5255 	if ((tsync_rxctl & E1000_TSYNCRXCTL_VALID) == 0)
5256 		return -EINVAL;
5257 
5258 	rx_tstamp_cycles = igb_read_rx_tstamp_cyclecounter(dev);
5259 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
5260 	*timestamp = rte_ns_to_timespec(ns);
5261 
5262 	return  0;
5263 }
5264 
5265 static int
5266 igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
5267 			       struct timespec *timestamp)
5268 {
5269 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5270 	struct e1000_adapter *adapter = dev->data->dev_private;
5271 	uint32_t tsync_txctl;
5272 	uint64_t tx_tstamp_cycles;
5273 	uint64_t ns;
5274 
5275 	tsync_txctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5276 	if ((tsync_txctl & E1000_TSYNCTXCTL_VALID) == 0)
5277 		return -EINVAL;
5278 
5279 	tx_tstamp_cycles = igb_read_tx_tstamp_cyclecounter(dev);
5280 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
5281 	*timestamp = rte_ns_to_timespec(ns);
5282 
5283 	return  0;
5284 }
5285 
5286 static int
5287 eth_igb_get_reg_length(struct rte_eth_dev *dev __rte_unused)
5288 {
5289 	int count = 0;
5290 	int g_ind = 0;
5291 	const struct reg_info *reg_group;
5292 
5293 	while ((reg_group = igb_regs[g_ind++]))
5294 		count += igb_reg_group_count(reg_group);
5295 
5296 	return count;
5297 }
5298 
5299 static int
5300 igbvf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
5301 {
5302 	int count = 0;
5303 	int g_ind = 0;
5304 	const struct reg_info *reg_group;
5305 
5306 	while ((reg_group = igbvf_regs[g_ind++]))
5307 		count += igb_reg_group_count(reg_group);
5308 
5309 	return count;
5310 }
5311 
5312 static int
5313 eth_igb_get_regs(struct rte_eth_dev *dev,
5314 	struct rte_dev_reg_info *regs)
5315 {
5316 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5317 	uint32_t *data = regs->data;
5318 	int g_ind = 0;
5319 	int count = 0;
5320 	const struct reg_info *reg_group;
5321 
5322 	if (data == NULL) {
5323 		regs->length = eth_igb_get_reg_length(dev);
5324 		regs->width = sizeof(uint32_t);
5325 		return 0;
5326 	}
5327 
5328 	/* Support only full register dump */
5329 	if ((regs->length == 0) ||
5330 	    (regs->length == (uint32_t)eth_igb_get_reg_length(dev))) {
5331 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5332 			hw->device_id;
5333 		while ((reg_group = igb_regs[g_ind++]))
5334 			count += igb_read_regs_group(dev, &data[count],
5335 							reg_group);
5336 		return 0;
5337 	}
5338 
5339 	return -ENOTSUP;
5340 }
5341 
5342 static int
5343 igbvf_get_regs(struct rte_eth_dev *dev,
5344 	struct rte_dev_reg_info *regs)
5345 {
5346 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5347 	uint32_t *data = regs->data;
5348 	int g_ind = 0;
5349 	int count = 0;
5350 	const struct reg_info *reg_group;
5351 
5352 	if (data == NULL) {
5353 		regs->length = igbvf_get_reg_length(dev);
5354 		regs->width = sizeof(uint32_t);
5355 		return 0;
5356 	}
5357 
5358 	/* Support only full register dump */
5359 	if ((regs->length == 0) ||
5360 	    (regs->length == (uint32_t)igbvf_get_reg_length(dev))) {
5361 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5362 			hw->device_id;
5363 		while ((reg_group = igbvf_regs[g_ind++]))
5364 			count += igb_read_regs_group(dev, &data[count],
5365 							reg_group);
5366 		return 0;
5367 	}
5368 
5369 	return -ENOTSUP;
5370 }
5371 
5372 static int
5373 eth_igb_get_eeprom_length(struct rte_eth_dev *dev)
5374 {
5375 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5376 
5377 	/* Return unit is byte count */
5378 	return hw->nvm.word_size * 2;
5379 }
5380 
5381 static int
5382 eth_igb_get_eeprom(struct rte_eth_dev *dev,
5383 	struct rte_dev_eeprom_info *in_eeprom)
5384 {
5385 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5386 	struct e1000_nvm_info *nvm = &hw->nvm;
5387 	uint16_t *data = in_eeprom->data;
5388 	int first, length;
5389 
5390 	first = in_eeprom->offset >> 1;
5391 	length = in_eeprom->length >> 1;
5392 	if ((first >= hw->nvm.word_size) ||
5393 	    ((first + length) >= hw->nvm.word_size))
5394 		return -EINVAL;
5395 
5396 	in_eeprom->magic = hw->vendor_id |
5397 		((uint32_t)hw->device_id << 16);
5398 
5399 	if ((nvm->ops.read) == NULL)
5400 		return -ENOTSUP;
5401 
5402 	return nvm->ops.read(hw, first, length, data);
5403 }
5404 
5405 static int
5406 eth_igb_set_eeprom(struct rte_eth_dev *dev,
5407 	struct rte_dev_eeprom_info *in_eeprom)
5408 {
5409 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5410 	struct e1000_nvm_info *nvm = &hw->nvm;
5411 	uint16_t *data = in_eeprom->data;
5412 	int first, length;
5413 
5414 	first = in_eeprom->offset >> 1;
5415 	length = in_eeprom->length >> 1;
5416 	if ((first >= hw->nvm.word_size) ||
5417 	    ((first + length) >= hw->nvm.word_size))
5418 		return -EINVAL;
5419 
5420 	in_eeprom->magic = (uint32_t)hw->vendor_id |
5421 		((uint32_t)hw->device_id << 16);
5422 
5423 	if ((nvm->ops.write) == NULL)
5424 		return -ENOTSUP;
5425 	return nvm->ops.write(hw,  first, length, data);
5426 }
5427 
5428 static int
5429 eth_igb_get_module_info(struct rte_eth_dev *dev,
5430 			struct rte_eth_dev_module_info *modinfo)
5431 {
5432 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5433 
5434 	uint32_t status = 0;
5435 	uint16_t sff8472_rev, addr_mode;
5436 	bool page_swap = false;
5437 
5438 	if (hw->phy.media_type == e1000_media_type_copper ||
5439 	    hw->phy.media_type == e1000_media_type_unknown)
5440 		return -EOPNOTSUPP;
5441 
5442 	/* Check whether we support SFF-8472 or not */
5443 	status = e1000_read_phy_reg_i2c(hw, IGB_SFF_8472_COMP, &sff8472_rev);
5444 	if (status)
5445 		return -EIO;
5446 
5447 	/* addressing mode is not supported */
5448 	status = e1000_read_phy_reg_i2c(hw, IGB_SFF_8472_SWAP, &addr_mode);
5449 	if (status)
5450 		return -EIO;
5451 
5452 	/* addressing mode is not supported */
5453 	if ((addr_mode & 0xFF) & IGB_SFF_ADDRESSING_MODE) {
5454 		PMD_DRV_LOG(ERR,
5455 			    "Address change required to access page 0xA2, "
5456 			    "but not supported. Please report the module "
5457 			    "type to the driver maintainers.\n");
5458 		page_swap = true;
5459 	}
5460 
5461 	if ((sff8472_rev & 0xFF) == IGB_SFF_8472_UNSUP || page_swap) {
5462 		/* We have an SFP, but it does not support SFF-8472 */
5463 		modinfo->type = RTE_ETH_MODULE_SFF_8079;
5464 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
5465 	} else {
5466 		/* We have an SFP which supports a revision of SFF-8472 */
5467 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
5468 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
5469 	}
5470 
5471 	return 0;
5472 }
5473 
5474 static int
5475 eth_igb_get_module_eeprom(struct rte_eth_dev *dev,
5476 			  struct rte_dev_eeprom_info *info)
5477 {
5478 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5479 
5480 	uint32_t status = 0;
5481 	uint16_t dataword[RTE_ETH_MODULE_SFF_8472_LEN / 2 + 1];
5482 	u16 first_word, last_word;
5483 	int i = 0;
5484 
5485 	if (info->length == 0)
5486 		return -EINVAL;
5487 
5488 	first_word = info->offset >> 1;
5489 	last_word = (info->offset + info->length - 1) >> 1;
5490 
5491 	/* Read EEPROM block, SFF-8079/SFF-8472, word at a time */
5492 	for (i = 0; i < last_word - first_word + 1; i++) {
5493 		status = e1000_read_phy_reg_i2c(hw, (first_word + i) * 2,
5494 						&dataword[i]);
5495 		if (status) {
5496 			/* Error occurred while reading module */
5497 			return -EIO;
5498 		}
5499 
5500 		dataword[i] = rte_be_to_cpu_16(dataword[i]);
5501 	}
5502 
5503 	memcpy(info->data, (u8 *)dataword + (info->offset & 1), info->length);
5504 
5505 	return 0;
5506 }
5507 
5508 static int
5509 eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5510 {
5511 	struct e1000_hw *hw =
5512 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5513 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5514 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5515 	uint32_t vec = E1000_MISC_VEC_ID;
5516 
5517 	if (rte_intr_allow_others(intr_handle))
5518 		vec = E1000_RX_VEC_START;
5519 
5520 	uint32_t mask = 1 << (queue_id + vec);
5521 
5522 	E1000_WRITE_REG(hw, E1000_EIMC, mask);
5523 	E1000_WRITE_FLUSH(hw);
5524 
5525 	return 0;
5526 }
5527 
5528 static int
5529 eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5530 {
5531 	struct e1000_hw *hw =
5532 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5533 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5534 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5535 	uint32_t vec = E1000_MISC_VEC_ID;
5536 
5537 	if (rte_intr_allow_others(intr_handle))
5538 		vec = E1000_RX_VEC_START;
5539 
5540 	uint32_t mask = 1 << (queue_id + vec);
5541 	uint32_t regval;
5542 
5543 	regval = E1000_READ_REG(hw, E1000_EIMS);
5544 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
5545 	E1000_WRITE_FLUSH(hw);
5546 
5547 	rte_intr_ack(intr_handle);
5548 
5549 	return 0;
5550 }
5551 
5552 static void
5553 eth_igb_write_ivar(struct e1000_hw *hw, uint8_t  msix_vector,
5554 		   uint8_t index, uint8_t offset)
5555 {
5556 	uint32_t val = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
5557 
5558 	/* clear bits */
5559 	val &= ~((uint32_t)0xFF << offset);
5560 
5561 	/* write vector and valid bit */
5562 	val |= (msix_vector | E1000_IVAR_VALID) << offset;
5563 
5564 	E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, val);
5565 }
5566 
5567 static void
5568 eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
5569 			   uint8_t queue, uint8_t msix_vector)
5570 {
5571 	uint32_t tmp = 0;
5572 
5573 	if (hw->mac.type == e1000_82575) {
5574 		if (direction == 0)
5575 			tmp = E1000_EICR_RX_QUEUE0 << queue;
5576 		else if (direction == 1)
5577 			tmp = E1000_EICR_TX_QUEUE0 << queue;
5578 		E1000_WRITE_REG(hw, E1000_MSIXBM(msix_vector), tmp);
5579 	} else if (hw->mac.type == e1000_82576) {
5580 		if ((direction == 0) || (direction == 1))
5581 			eth_igb_write_ivar(hw, msix_vector, queue & 0x7,
5582 					   ((queue & 0x8) << 1) +
5583 					   8 * direction);
5584 	} else if ((hw->mac.type == e1000_82580) ||
5585 			(hw->mac.type == e1000_i350) ||
5586 			(hw->mac.type == e1000_i354) ||
5587 			(hw->mac.type == e1000_i210) ||
5588 			(hw->mac.type == e1000_i211)) {
5589 		if ((direction == 0) || (direction == 1))
5590 			eth_igb_write_ivar(hw, msix_vector,
5591 					   queue >> 1,
5592 					   ((queue & 0x1) << 4) +
5593 					   8 * direction);
5594 	}
5595 }
5596 
5597 /* Sets up the hardware to generate MSI-X interrupts properly
5598  * @hw
5599  *  board private structure
5600  */
5601 static void
5602 eth_igb_configure_msix_intr(struct rte_eth_dev *dev)
5603 {
5604 	int queue_id;
5605 	uint32_t tmpval, regval, intr_mask;
5606 	struct e1000_hw *hw =
5607 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5608 	uint32_t vec = E1000_MISC_VEC_ID;
5609 	uint32_t base = E1000_MISC_VEC_ID;
5610 	uint32_t misc_shift = 0;
5611 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5612 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5613 
5614 	/* won't configure msix register if no mapping is done
5615 	 * between intr vector and event fd
5616 	 */
5617 	if (!rte_intr_dp_is_en(intr_handle))
5618 		return;
5619 
5620 	if (rte_intr_allow_others(intr_handle)) {
5621 		vec = base = E1000_RX_VEC_START;
5622 		misc_shift = 1;
5623 	}
5624 
5625 	/* set interrupt vector for other causes */
5626 	if (hw->mac.type == e1000_82575) {
5627 		tmpval = E1000_READ_REG(hw, E1000_CTRL_EXT);
5628 		/* enable MSI-X PBA support */
5629 		tmpval |= E1000_CTRL_EXT_PBA_CLR;
5630 
5631 		/* Auto-Mask interrupts upon ICR read */
5632 		tmpval |= E1000_CTRL_EXT_EIAME;
5633 		tmpval |= E1000_CTRL_EXT_IRCA;
5634 
5635 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmpval);
5636 
5637 		/* enable msix_other interrupt */
5638 		E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0), 0, E1000_EIMS_OTHER);
5639 		regval = E1000_READ_REG(hw, E1000_EIAC);
5640 		E1000_WRITE_REG(hw, E1000_EIAC, regval | E1000_EIMS_OTHER);
5641 		regval = E1000_READ_REG(hw, E1000_EIAM);
5642 		E1000_WRITE_REG(hw, E1000_EIMS, regval | E1000_EIMS_OTHER);
5643 	} else if ((hw->mac.type == e1000_82576) ||
5644 			(hw->mac.type == e1000_82580) ||
5645 			(hw->mac.type == e1000_i350) ||
5646 			(hw->mac.type == e1000_i354) ||
5647 			(hw->mac.type == e1000_i210) ||
5648 			(hw->mac.type == e1000_i211)) {
5649 		/* turn on MSI-X capability first */
5650 		E1000_WRITE_REG(hw, E1000_GPIE, E1000_GPIE_MSIX_MODE |
5651 					E1000_GPIE_PBA | E1000_GPIE_EIAME |
5652 					E1000_GPIE_NSICR);
5653 		intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5654 			misc_shift;
5655 
5656 		if (dev->data->dev_conf.intr_conf.lsc != 0)
5657 			intr_mask |= (1 << IGB_MSIX_OTHER_INTR_VEC);
5658 
5659 		regval = E1000_READ_REG(hw, E1000_EIAC);
5660 		E1000_WRITE_REG(hw, E1000_EIAC, regval | intr_mask);
5661 
5662 		/* enable msix_other interrupt */
5663 		regval = E1000_READ_REG(hw, E1000_EIMS);
5664 		E1000_WRITE_REG(hw, E1000_EIMS, regval | intr_mask);
5665 		tmpval = (IGB_MSIX_OTHER_INTR_VEC | E1000_IVAR_VALID) << 8;
5666 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, tmpval);
5667 	}
5668 
5669 	/* use EIAM to auto-mask when MSI-X interrupt
5670 	 * is asserted, this saves a register write for every interrupt
5671 	 */
5672 	intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5673 		misc_shift;
5674 
5675 	if (dev->data->dev_conf.intr_conf.lsc != 0)
5676 		intr_mask |= (1 << IGB_MSIX_OTHER_INTR_VEC);
5677 
5678 	regval = E1000_READ_REG(hw, E1000_EIAM);
5679 	E1000_WRITE_REG(hw, E1000_EIAM, regval | intr_mask);
5680 
5681 	for (queue_id = 0; queue_id < dev->data->nb_rx_queues; queue_id++) {
5682 		eth_igb_assign_msix_vector(hw, 0, queue_id, vec);
5683 		intr_handle->intr_vec[queue_id] = vec;
5684 		if (vec < base + intr_handle->nb_efd - 1)
5685 			vec++;
5686 	}
5687 
5688 	E1000_WRITE_FLUSH(hw);
5689 }
5690 
5691 /* restore n-tuple filter */
5692 static inline void
5693 igb_ntuple_filter_restore(struct rte_eth_dev *dev)
5694 {
5695 	struct e1000_filter_info *filter_info =
5696 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5697 	struct e1000_5tuple_filter *p_5tuple;
5698 	struct e1000_2tuple_filter *p_2tuple;
5699 
5700 	TAILQ_FOREACH(p_5tuple, &filter_info->fivetuple_list, entries) {
5701 		igb_inject_5tuple_filter_82576(dev, p_5tuple);
5702 	}
5703 
5704 	TAILQ_FOREACH(p_2tuple, &filter_info->twotuple_list, entries) {
5705 		igb_inject_2uple_filter(dev, p_2tuple);
5706 	}
5707 }
5708 
5709 /* restore SYN filter */
5710 static inline void
5711 igb_syn_filter_restore(struct rte_eth_dev *dev)
5712 {
5713 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5714 	struct e1000_filter_info *filter_info =
5715 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5716 	uint32_t synqf;
5717 
5718 	synqf = filter_info->syn_info;
5719 
5720 	if (synqf & E1000_SYN_FILTER_ENABLE) {
5721 		E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
5722 		E1000_WRITE_FLUSH(hw);
5723 	}
5724 }
5725 
5726 /* restore ethernet type filter */
5727 static inline void
5728 igb_ethertype_filter_restore(struct rte_eth_dev *dev)
5729 {
5730 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5731 	struct e1000_filter_info *filter_info =
5732 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5733 	int i;
5734 
5735 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
5736 		if (filter_info->ethertype_mask & (1 << i)) {
5737 			E1000_WRITE_REG(hw, E1000_ETQF(i),
5738 				filter_info->ethertype_filters[i].etqf);
5739 			E1000_WRITE_FLUSH(hw);
5740 		}
5741 	}
5742 }
5743 
5744 /* restore flex byte filter */
5745 static inline void
5746 igb_flex_filter_restore(struct rte_eth_dev *dev)
5747 {
5748 	struct e1000_filter_info *filter_info =
5749 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5750 	struct e1000_flex_filter *flex_filter;
5751 
5752 	TAILQ_FOREACH(flex_filter, &filter_info->flex_list, entries) {
5753 		igb_inject_flex_filter(dev, flex_filter);
5754 	}
5755 }
5756 
5757 /* restore rss filter */
5758 static inline void
5759 igb_rss_filter_restore(struct rte_eth_dev *dev)
5760 {
5761 	struct e1000_filter_info *filter_info =
5762 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5763 
5764 	if (filter_info->rss_info.conf.queue_num)
5765 		igb_config_rss_filter(dev, &filter_info->rss_info, TRUE);
5766 }
5767 
5768 /* restore all types filter */
5769 static int
5770 igb_filter_restore(struct rte_eth_dev *dev)
5771 {
5772 	igb_ntuple_filter_restore(dev);
5773 	igb_ethertype_filter_restore(dev);
5774 	igb_syn_filter_restore(dev);
5775 	igb_flex_filter_restore(dev);
5776 	igb_rss_filter_restore(dev);
5777 
5778 	return 0;
5779 }
5780 
5781 RTE_PMD_REGISTER_PCI(net_e1000_igb, rte_igb_pmd);
5782 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb, pci_id_igb_map);
5783 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb, "* igb_uio | uio_pci_generic | vfio-pci");
5784 RTE_PMD_REGISTER_PCI(net_e1000_igb_vf, rte_igbvf_pmd);
5785 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb_vf, pci_id_igbvf_map);
5786 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb_vf, "* igb_uio | vfio-pci");
5787 
5788 /* see e1000_logs.c */
5789 RTE_INIT(e1000_init_log)
5790 {
5791 	e1000_igb_init_log();
5792 }
5793