xref: /dpdk/drivers/net/e1000/igb_ethdev.c (revision ba9de463)
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_common.h>
12 #include <rte_interrupts.h>
13 #include <rte_byteorder.h>
14 #include <rte_log.h>
15 #include <rte_debug.h>
16 #include <rte_pci.h>
17 #include <rte_bus_pci.h>
18 #include <rte_ether.h>
19 #include <rte_ethdev.h>
20 #include <rte_ethdev_pci.h>
21 #include <rte_memory.h>
22 #include <rte_eal.h>
23 #include <rte_atomic.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 #define IGB_HKEY_MAX_INDEX 10
46 
47 /* Bit shift and mask */
48 #define IGB_4_BIT_WIDTH  (CHAR_BIT / 2)
49 #define IGB_4_BIT_MASK   RTE_LEN2MASK(IGB_4_BIT_WIDTH, uint8_t)
50 #define IGB_8_BIT_WIDTH  CHAR_BIT
51 #define IGB_8_BIT_MASK   UINT8_MAX
52 
53 /* Additional timesync values. */
54 #define E1000_CYCLECOUNTER_MASK      0xffffffffffffffffULL
55 #define E1000_ETQF_FILTER_1588       3
56 #define IGB_82576_TSYNC_SHIFT        16
57 #define E1000_INCPERIOD_82576        (1 << E1000_TIMINCA_16NS_SHIFT)
58 #define E1000_INCVALUE_82576         (16 << IGB_82576_TSYNC_SHIFT)
59 #define E1000_TSAUXC_DISABLE_SYSTIME 0x80000000
60 
61 #define E1000_VTIVAR_MISC                0x01740
62 #define E1000_VTIVAR_MISC_MASK           0xFF
63 #define E1000_VTIVAR_VALID               0x80
64 #define E1000_VTIVAR_MISC_MAILBOX        0
65 #define E1000_VTIVAR_MISC_INTR_MASK      0x3
66 
67 /* External VLAN Enable bit mask */
68 #define E1000_CTRL_EXT_EXT_VLAN      (1 << 26)
69 
70 /* External VLAN Ether Type bit mask and shift */
71 #define E1000_VET_VET_EXT            0xFFFF0000
72 #define E1000_VET_VET_EXT_SHIFT      16
73 
74 static int  eth_igb_configure(struct rte_eth_dev *dev);
75 static int  eth_igb_start(struct rte_eth_dev *dev);
76 static void eth_igb_stop(struct rte_eth_dev *dev);
77 static int  eth_igb_dev_set_link_up(struct rte_eth_dev *dev);
78 static int  eth_igb_dev_set_link_down(struct rte_eth_dev *dev);
79 static void eth_igb_close(struct rte_eth_dev *dev);
80 static void eth_igb_promiscuous_enable(struct rte_eth_dev *dev);
81 static void eth_igb_promiscuous_disable(struct rte_eth_dev *dev);
82 static void eth_igb_allmulticast_enable(struct rte_eth_dev *dev);
83 static void eth_igb_allmulticast_disable(struct rte_eth_dev *dev);
84 static int  eth_igb_link_update(struct rte_eth_dev *dev,
85 				int wait_to_complete);
86 static int eth_igb_stats_get(struct rte_eth_dev *dev,
87 				struct rte_eth_stats *rte_stats);
88 static int eth_igb_xstats_get(struct rte_eth_dev *dev,
89 			      struct rte_eth_xstat *xstats, unsigned n);
90 static int eth_igb_xstats_get_by_id(struct rte_eth_dev *dev,
91 		const uint64_t *ids,
92 		uint64_t *values, unsigned int n);
93 static int eth_igb_xstats_get_names(struct rte_eth_dev *dev,
94 				    struct rte_eth_xstat_name *xstats_names,
95 				    unsigned int size);
96 static int eth_igb_xstats_get_names_by_id(struct rte_eth_dev *dev,
97 		struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
98 		unsigned int limit);
99 static void eth_igb_stats_reset(struct rte_eth_dev *dev);
100 static void eth_igb_xstats_reset(struct rte_eth_dev *dev);
101 static int eth_igb_fw_version_get(struct rte_eth_dev *dev,
102 				   char *fw_version, size_t fw_size);
103 static void eth_igb_infos_get(struct rte_eth_dev *dev,
104 			      struct rte_eth_dev_info *dev_info);
105 static const uint32_t *eth_igb_supported_ptypes_get(struct rte_eth_dev *dev);
106 static void eth_igbvf_infos_get(struct rte_eth_dev *dev,
107 				struct rte_eth_dev_info *dev_info);
108 static int  eth_igb_flow_ctrl_get(struct rte_eth_dev *dev,
109 				struct rte_eth_fc_conf *fc_conf);
110 static int  eth_igb_flow_ctrl_set(struct rte_eth_dev *dev,
111 				struct rte_eth_fc_conf *fc_conf);
112 static int eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
113 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev);
114 static int eth_igb_interrupt_get_status(struct rte_eth_dev *dev);
115 static int eth_igb_interrupt_action(struct rte_eth_dev *dev,
116 				    struct rte_intr_handle *handle);
117 static void eth_igb_interrupt_handler(void *param);
118 static int  igb_hardware_init(struct e1000_hw *hw);
119 static void igb_hw_control_acquire(struct e1000_hw *hw);
120 static void igb_hw_control_release(struct e1000_hw *hw);
121 static void igb_init_manageability(struct e1000_hw *hw);
122 static void igb_release_manageability(struct e1000_hw *hw);
123 
124 static int  eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
125 
126 static int eth_igb_vlan_filter_set(struct rte_eth_dev *dev,
127 		uint16_t vlan_id, int on);
128 static int eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
129 				 enum rte_vlan_type vlan_type,
130 				 uint16_t tpid_id);
131 static int eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask);
132 
133 static void igb_vlan_hw_filter_enable(struct rte_eth_dev *dev);
134 static void igb_vlan_hw_filter_disable(struct rte_eth_dev *dev);
135 static void igb_vlan_hw_strip_enable(struct rte_eth_dev *dev);
136 static void igb_vlan_hw_strip_disable(struct rte_eth_dev *dev);
137 static void igb_vlan_hw_extend_enable(struct rte_eth_dev *dev);
138 static void igb_vlan_hw_extend_disable(struct rte_eth_dev *dev);
139 
140 static int eth_igb_led_on(struct rte_eth_dev *dev);
141 static int eth_igb_led_off(struct rte_eth_dev *dev);
142 
143 static void igb_intr_disable(struct e1000_hw *hw);
144 static int  igb_get_rx_buffer_size(struct e1000_hw *hw);
145 static int eth_igb_rar_set(struct rte_eth_dev *dev,
146 			   struct ether_addr *mac_addr,
147 			   uint32_t index, uint32_t pool);
148 static void eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index);
149 static void eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
150 		struct ether_addr *addr);
151 
152 static void igbvf_intr_disable(struct e1000_hw *hw);
153 static int igbvf_dev_configure(struct rte_eth_dev *dev);
154 static int igbvf_dev_start(struct rte_eth_dev *dev);
155 static void igbvf_dev_stop(struct rte_eth_dev *dev);
156 static void igbvf_dev_close(struct rte_eth_dev *dev);
157 static void igbvf_promiscuous_enable(struct rte_eth_dev *dev);
158 static void igbvf_promiscuous_disable(struct rte_eth_dev *dev);
159 static void igbvf_allmulticast_enable(struct rte_eth_dev *dev);
160 static void igbvf_allmulticast_disable(struct rte_eth_dev *dev);
161 static int eth_igbvf_link_update(struct e1000_hw *hw);
162 static int eth_igbvf_stats_get(struct rte_eth_dev *dev,
163 				struct rte_eth_stats *rte_stats);
164 static int eth_igbvf_xstats_get(struct rte_eth_dev *dev,
165 				struct rte_eth_xstat *xstats, unsigned n);
166 static int eth_igbvf_xstats_get_names(struct rte_eth_dev *dev,
167 				      struct rte_eth_xstat_name *xstats_names,
168 				      unsigned limit);
169 static void eth_igbvf_stats_reset(struct rte_eth_dev *dev);
170 static int igbvf_vlan_filter_set(struct rte_eth_dev *dev,
171 		uint16_t vlan_id, int on);
172 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on);
173 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on);
174 static void igbvf_default_mac_addr_set(struct rte_eth_dev *dev,
175 		struct ether_addr *addr);
176 static int igbvf_get_reg_length(struct rte_eth_dev *dev);
177 static int igbvf_get_regs(struct rte_eth_dev *dev,
178 		struct rte_dev_reg_info *regs);
179 
180 static int eth_igb_rss_reta_update(struct rte_eth_dev *dev,
181 				   struct rte_eth_rss_reta_entry64 *reta_conf,
182 				   uint16_t reta_size);
183 static int eth_igb_rss_reta_query(struct rte_eth_dev *dev,
184 				  struct rte_eth_rss_reta_entry64 *reta_conf,
185 				  uint16_t reta_size);
186 
187 static int eth_igb_syn_filter_get(struct rte_eth_dev *dev,
188 			struct rte_eth_syn_filter *filter);
189 static int eth_igb_syn_filter_handle(struct rte_eth_dev *dev,
190 			enum rte_filter_op filter_op,
191 			void *arg);
192 static int igb_add_2tuple_filter(struct rte_eth_dev *dev,
193 			struct rte_eth_ntuple_filter *ntuple_filter);
194 static int igb_remove_2tuple_filter(struct rte_eth_dev *dev,
195 			struct rte_eth_ntuple_filter *ntuple_filter);
196 static int eth_igb_get_flex_filter(struct rte_eth_dev *dev,
197 			struct rte_eth_flex_filter *filter);
198 static int eth_igb_flex_filter_handle(struct rte_eth_dev *dev,
199 			enum rte_filter_op filter_op,
200 			void *arg);
201 static int igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
202 			struct rte_eth_ntuple_filter *ntuple_filter);
203 static int igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
204 			struct rte_eth_ntuple_filter *ntuple_filter);
205 static int igb_get_ntuple_filter(struct rte_eth_dev *dev,
206 			struct rte_eth_ntuple_filter *filter);
207 static int igb_ntuple_filter_handle(struct rte_eth_dev *dev,
208 				enum rte_filter_op filter_op,
209 				void *arg);
210 static int igb_ethertype_filter_handle(struct rte_eth_dev *dev,
211 				enum rte_filter_op filter_op,
212 				void *arg);
213 static int igb_get_ethertype_filter(struct rte_eth_dev *dev,
214 			struct rte_eth_ethertype_filter *filter);
215 static int eth_igb_filter_ctrl(struct rte_eth_dev *dev,
216 		     enum rte_filter_type filter_type,
217 		     enum rte_filter_op filter_op,
218 		     void *arg);
219 static int eth_igb_get_reg_length(struct rte_eth_dev *dev);
220 static int eth_igb_get_regs(struct rte_eth_dev *dev,
221 		struct rte_dev_reg_info *regs);
222 static int eth_igb_get_eeprom_length(struct rte_eth_dev *dev);
223 static int eth_igb_get_eeprom(struct rte_eth_dev *dev,
224 		struct rte_dev_eeprom_info *eeprom);
225 static int eth_igb_set_eeprom(struct rte_eth_dev *dev,
226 		struct rte_dev_eeprom_info *eeprom);
227 static int eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
228 				    struct ether_addr *mc_addr_set,
229 				    uint32_t nb_mc_addr);
230 static int igb_timesync_enable(struct rte_eth_dev *dev);
231 static int igb_timesync_disable(struct rte_eth_dev *dev);
232 static int igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
233 					  struct timespec *timestamp,
234 					  uint32_t flags);
235 static int igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
236 					  struct timespec *timestamp);
237 static int igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
238 static int igb_timesync_read_time(struct rte_eth_dev *dev,
239 				  struct timespec *timestamp);
240 static int igb_timesync_write_time(struct rte_eth_dev *dev,
241 				   const struct timespec *timestamp);
242 static int eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev,
243 					uint16_t queue_id);
244 static int eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev,
245 					 uint16_t queue_id);
246 static void eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
247 				       uint8_t queue, uint8_t msix_vector);
248 static void eth_igb_write_ivar(struct e1000_hw *hw, uint8_t msix_vector,
249 			       uint8_t index, uint8_t offset);
250 static void eth_igb_configure_msix_intr(struct rte_eth_dev *dev);
251 static void eth_igbvf_interrupt_handler(void *param);
252 static void igbvf_mbx_process(struct rte_eth_dev *dev);
253 static int igb_filter_restore(struct rte_eth_dev *dev);
254 
255 /*
256  * Define VF Stats MACRO for Non "cleared on read" register
257  */
258 #define UPDATE_VF_STAT(reg, last, cur)            \
259 {                                                 \
260 	u32 latest = E1000_READ_REG(hw, reg);     \
261 	cur += (latest - last) & UINT_MAX;        \
262 	last = latest;                            \
263 }
264 
265 #define IGB_FC_PAUSE_TIME 0x0680
266 #define IGB_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
267 #define IGB_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
268 
269 #define IGBVF_PMD_NAME "rte_igbvf_pmd"     /* PMD name */
270 
271 static enum e1000_fc_mode igb_fc_setting = e1000_fc_full;
272 
273 /*
274  * The set of PCI devices this driver supports
275  */
276 static const struct rte_pci_id pci_id_igb_map[] = {
277 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576) },
278 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_FIBER) },
279 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_SERDES) },
280 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_QUAD_COPPER) },
281 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_QUAD_COPPER_ET2) },
282 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_NS) },
283 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_NS_SERDES) },
284 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_SERDES_QUAD) },
285 
286 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575EB_COPPER) },
287 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575EB_FIBER_SERDES) },
288 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575GB_QUAD_COPPER) },
289 
290 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_COPPER) },
291 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_FIBER) },
292 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_SERDES) },
293 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_SGMII) },
294 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_COPPER_DUAL) },
295 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_QUAD_FIBER) },
296 
297 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_COPPER) },
298 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_FIBER) },
299 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_SERDES) },
300 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_SGMII) },
301 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_DA4) },
302 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER) },
303 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_OEM1) },
304 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_IT) },
305 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_FIBER) },
306 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SERDES) },
307 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SGMII) },
308 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_FLASHLESS) },
309 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SERDES_FLASHLESS) },
310 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I211_COPPER) },
311 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
312 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_SGMII) },
313 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
314 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SGMII) },
315 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SERDES) },
316 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_BACKPLANE) },
317 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SFP) },
318 	{ .vendor_id = 0, /* sentinel */ },
319 };
320 
321 /*
322  * The set of PCI devices this driver supports (for 82576&I350 VF)
323  */
324 static const struct rte_pci_id pci_id_igbvf_map[] = {
325 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_VF) },
326 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_VF_HV) },
327 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_VF) },
328 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_VF_HV) },
329 	{ .vendor_id = 0, /* sentinel */ },
330 };
331 
332 static const struct rte_eth_desc_lim rx_desc_lim = {
333 	.nb_max = E1000_MAX_RING_DESC,
334 	.nb_min = E1000_MIN_RING_DESC,
335 	.nb_align = IGB_RXD_ALIGN,
336 };
337 
338 static const struct rte_eth_desc_lim tx_desc_lim = {
339 	.nb_max = E1000_MAX_RING_DESC,
340 	.nb_min = E1000_MIN_RING_DESC,
341 	.nb_align = IGB_RXD_ALIGN,
342 	.nb_seg_max = IGB_TX_MAX_SEG,
343 	.nb_mtu_seg_max = IGB_TX_MAX_MTU_SEG,
344 };
345 
346 static const struct eth_dev_ops eth_igb_ops = {
347 	.dev_configure        = eth_igb_configure,
348 	.dev_start            = eth_igb_start,
349 	.dev_stop             = eth_igb_stop,
350 	.dev_set_link_up      = eth_igb_dev_set_link_up,
351 	.dev_set_link_down    = eth_igb_dev_set_link_down,
352 	.dev_close            = eth_igb_close,
353 	.promiscuous_enable   = eth_igb_promiscuous_enable,
354 	.promiscuous_disable  = eth_igb_promiscuous_disable,
355 	.allmulticast_enable  = eth_igb_allmulticast_enable,
356 	.allmulticast_disable = eth_igb_allmulticast_disable,
357 	.link_update          = eth_igb_link_update,
358 	.stats_get            = eth_igb_stats_get,
359 	.xstats_get           = eth_igb_xstats_get,
360 	.xstats_get_by_id     = eth_igb_xstats_get_by_id,
361 	.xstats_get_names_by_id = eth_igb_xstats_get_names_by_id,
362 	.xstats_get_names     = eth_igb_xstats_get_names,
363 	.stats_reset          = eth_igb_stats_reset,
364 	.xstats_reset         = eth_igb_xstats_reset,
365 	.fw_version_get       = eth_igb_fw_version_get,
366 	.dev_infos_get        = eth_igb_infos_get,
367 	.dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
368 	.mtu_set              = eth_igb_mtu_set,
369 	.vlan_filter_set      = eth_igb_vlan_filter_set,
370 	.vlan_tpid_set        = eth_igb_vlan_tpid_set,
371 	.vlan_offload_set     = eth_igb_vlan_offload_set,
372 	.rx_queue_setup       = eth_igb_rx_queue_setup,
373 	.rx_queue_intr_enable = eth_igb_rx_queue_intr_enable,
374 	.rx_queue_intr_disable = eth_igb_rx_queue_intr_disable,
375 	.rx_queue_release     = eth_igb_rx_queue_release,
376 	.rx_queue_count       = eth_igb_rx_queue_count,
377 	.rx_descriptor_done   = eth_igb_rx_descriptor_done,
378 	.rx_descriptor_status = eth_igb_rx_descriptor_status,
379 	.tx_descriptor_status = eth_igb_tx_descriptor_status,
380 	.tx_queue_setup       = eth_igb_tx_queue_setup,
381 	.tx_queue_release     = eth_igb_tx_queue_release,
382 	.tx_done_cleanup      = eth_igb_tx_done_cleanup,
383 	.dev_led_on           = eth_igb_led_on,
384 	.dev_led_off          = eth_igb_led_off,
385 	.flow_ctrl_get        = eth_igb_flow_ctrl_get,
386 	.flow_ctrl_set        = eth_igb_flow_ctrl_set,
387 	.mac_addr_add         = eth_igb_rar_set,
388 	.mac_addr_remove      = eth_igb_rar_clear,
389 	.mac_addr_set         = eth_igb_default_mac_addr_set,
390 	.reta_update          = eth_igb_rss_reta_update,
391 	.reta_query           = eth_igb_rss_reta_query,
392 	.rss_hash_update      = eth_igb_rss_hash_update,
393 	.rss_hash_conf_get    = eth_igb_rss_hash_conf_get,
394 	.filter_ctrl          = eth_igb_filter_ctrl,
395 	.set_mc_addr_list     = eth_igb_set_mc_addr_list,
396 	.rxq_info_get         = igb_rxq_info_get,
397 	.txq_info_get         = igb_txq_info_get,
398 	.timesync_enable      = igb_timesync_enable,
399 	.timesync_disable     = igb_timesync_disable,
400 	.timesync_read_rx_timestamp = igb_timesync_read_rx_timestamp,
401 	.timesync_read_tx_timestamp = igb_timesync_read_tx_timestamp,
402 	.get_reg              = eth_igb_get_regs,
403 	.get_eeprom_length    = eth_igb_get_eeprom_length,
404 	.get_eeprom           = eth_igb_get_eeprom,
405 	.set_eeprom           = eth_igb_set_eeprom,
406 	.timesync_adjust_time = igb_timesync_adjust_time,
407 	.timesync_read_time   = igb_timesync_read_time,
408 	.timesync_write_time  = igb_timesync_write_time,
409 };
410 
411 /*
412  * dev_ops for virtual function, bare necessities for basic vf
413  * operation have been implemented
414  */
415 static const struct eth_dev_ops igbvf_eth_dev_ops = {
416 	.dev_configure        = igbvf_dev_configure,
417 	.dev_start            = igbvf_dev_start,
418 	.dev_stop             = igbvf_dev_stop,
419 	.dev_close            = igbvf_dev_close,
420 	.promiscuous_enable   = igbvf_promiscuous_enable,
421 	.promiscuous_disable  = igbvf_promiscuous_disable,
422 	.allmulticast_enable  = igbvf_allmulticast_enable,
423 	.allmulticast_disable = igbvf_allmulticast_disable,
424 	.link_update          = eth_igb_link_update,
425 	.stats_get            = eth_igbvf_stats_get,
426 	.xstats_get           = eth_igbvf_xstats_get,
427 	.xstats_get_names     = eth_igbvf_xstats_get_names,
428 	.stats_reset          = eth_igbvf_stats_reset,
429 	.xstats_reset         = eth_igbvf_stats_reset,
430 	.vlan_filter_set      = igbvf_vlan_filter_set,
431 	.dev_infos_get        = eth_igbvf_infos_get,
432 	.dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
433 	.rx_queue_setup       = eth_igb_rx_queue_setup,
434 	.rx_queue_release     = eth_igb_rx_queue_release,
435 	.tx_queue_setup       = eth_igb_tx_queue_setup,
436 	.tx_queue_release     = eth_igb_tx_queue_release,
437 	.set_mc_addr_list     = eth_igb_set_mc_addr_list,
438 	.rxq_info_get         = igb_rxq_info_get,
439 	.txq_info_get         = igb_txq_info_get,
440 	.mac_addr_set         = igbvf_default_mac_addr_set,
441 	.get_reg              = igbvf_get_regs,
442 };
443 
444 /* store statistics names and its offset in stats structure */
445 struct rte_igb_xstats_name_off {
446 	char name[RTE_ETH_XSTATS_NAME_SIZE];
447 	unsigned offset;
448 };
449 
450 static const struct rte_igb_xstats_name_off rte_igb_stats_strings[] = {
451 	{"rx_crc_errors", offsetof(struct e1000_hw_stats, crcerrs)},
452 	{"rx_align_errors", offsetof(struct e1000_hw_stats, algnerrc)},
453 	{"rx_symbol_errors", offsetof(struct e1000_hw_stats, symerrs)},
454 	{"rx_missed_packets", offsetof(struct e1000_hw_stats, mpc)},
455 	{"tx_single_collision_packets", offsetof(struct e1000_hw_stats, scc)},
456 	{"tx_multiple_collision_packets", offsetof(struct e1000_hw_stats, mcc)},
457 	{"tx_excessive_collision_packets", offsetof(struct e1000_hw_stats,
458 		ecol)},
459 	{"tx_late_collisions", offsetof(struct e1000_hw_stats, latecol)},
460 	{"tx_total_collisions", offsetof(struct e1000_hw_stats, colc)},
461 	{"tx_deferred_packets", offsetof(struct e1000_hw_stats, dc)},
462 	{"tx_no_carrier_sense_packets", offsetof(struct e1000_hw_stats, tncrs)},
463 	{"rx_carrier_ext_errors", offsetof(struct e1000_hw_stats, cexterr)},
464 	{"rx_length_errors", offsetof(struct e1000_hw_stats, rlec)},
465 	{"rx_xon_packets", offsetof(struct e1000_hw_stats, xonrxc)},
466 	{"tx_xon_packets", offsetof(struct e1000_hw_stats, xontxc)},
467 	{"rx_xoff_packets", offsetof(struct e1000_hw_stats, xoffrxc)},
468 	{"tx_xoff_packets", offsetof(struct e1000_hw_stats, xofftxc)},
469 	{"rx_flow_control_unsupported_packets", offsetof(struct e1000_hw_stats,
470 		fcruc)},
471 	{"rx_size_64_packets", offsetof(struct e1000_hw_stats, prc64)},
472 	{"rx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, prc127)},
473 	{"rx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, prc255)},
474 	{"rx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, prc511)},
475 	{"rx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
476 		prc1023)},
477 	{"rx_size_1024_to_max_packets", offsetof(struct e1000_hw_stats,
478 		prc1522)},
479 	{"rx_broadcast_packets", offsetof(struct e1000_hw_stats, bprc)},
480 	{"rx_multicast_packets", offsetof(struct e1000_hw_stats, mprc)},
481 	{"rx_undersize_errors", offsetof(struct e1000_hw_stats, ruc)},
482 	{"rx_fragment_errors", offsetof(struct e1000_hw_stats, rfc)},
483 	{"rx_oversize_errors", offsetof(struct e1000_hw_stats, roc)},
484 	{"rx_jabber_errors", offsetof(struct e1000_hw_stats, rjc)},
485 	{"rx_management_packets", offsetof(struct e1000_hw_stats, mgprc)},
486 	{"rx_management_dropped", offsetof(struct e1000_hw_stats, mgpdc)},
487 	{"tx_management_packets", offsetof(struct e1000_hw_stats, mgptc)},
488 	{"rx_total_packets", offsetof(struct e1000_hw_stats, tpr)},
489 	{"tx_total_packets", offsetof(struct e1000_hw_stats, tpt)},
490 	{"rx_total_bytes", offsetof(struct e1000_hw_stats, tor)},
491 	{"tx_total_bytes", offsetof(struct e1000_hw_stats, tot)},
492 	{"tx_size_64_packets", offsetof(struct e1000_hw_stats, ptc64)},
493 	{"tx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, ptc127)},
494 	{"tx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, ptc255)},
495 	{"tx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, ptc511)},
496 	{"tx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
497 		ptc1023)},
498 	{"tx_size_1023_to_max_packets", offsetof(struct e1000_hw_stats,
499 		ptc1522)},
500 	{"tx_multicast_packets", offsetof(struct e1000_hw_stats, mptc)},
501 	{"tx_broadcast_packets", offsetof(struct e1000_hw_stats, bptc)},
502 	{"tx_tso_packets", offsetof(struct e1000_hw_stats, tsctc)},
503 	{"tx_tso_errors", offsetof(struct e1000_hw_stats, tsctfc)},
504 	{"rx_sent_to_host_packets", offsetof(struct e1000_hw_stats, rpthc)},
505 	{"tx_sent_by_host_packets", offsetof(struct e1000_hw_stats, hgptc)},
506 	{"rx_code_violation_packets", offsetof(struct e1000_hw_stats, scvpc)},
507 
508 	{"interrupt_assert_count", offsetof(struct e1000_hw_stats, iac)},
509 };
510 
511 #define IGB_NB_XSTATS (sizeof(rte_igb_stats_strings) / \
512 		sizeof(rte_igb_stats_strings[0]))
513 
514 static const struct rte_igb_xstats_name_off rte_igbvf_stats_strings[] = {
515 	{"rx_multicast_packets", offsetof(struct e1000_vf_stats, mprc)},
516 	{"rx_good_loopback_packets", offsetof(struct e1000_vf_stats, gprlbc)},
517 	{"tx_good_loopback_packets", offsetof(struct e1000_vf_stats, gptlbc)},
518 	{"rx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gorlbc)},
519 	{"tx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gotlbc)},
520 };
521 
522 #define IGBVF_NB_XSTATS (sizeof(rte_igbvf_stats_strings) / \
523 		sizeof(rte_igbvf_stats_strings[0]))
524 
525 /**
526  * Atomically reads the link status information from global
527  * structure rte_eth_dev.
528  *
529  * @param dev
530  *   - Pointer to the structure rte_eth_dev to read from.
531  *   - Pointer to the buffer to be saved with the link status.
532  *
533  * @return
534  *   - On success, zero.
535  *   - On failure, negative value.
536  */
537 static inline int
538 rte_igb_dev_atomic_read_link_status(struct rte_eth_dev *dev,
539 				struct rte_eth_link *link)
540 {
541 	struct rte_eth_link *dst = link;
542 	struct rte_eth_link *src = &(dev->data->dev_link);
543 
544 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
545 					*(uint64_t *)src) == 0)
546 		return -1;
547 
548 	return 0;
549 }
550 
551 /**
552  * Atomically writes the link status information into global
553  * structure rte_eth_dev.
554  *
555  * @param dev
556  *   - Pointer to the structure rte_eth_dev to read from.
557  *   - Pointer to the buffer to be saved with the link status.
558  *
559  * @return
560  *   - On success, zero.
561  *   - On failure, negative value.
562  */
563 static inline int
564 rte_igb_dev_atomic_write_link_status(struct rte_eth_dev *dev,
565 				struct rte_eth_link *link)
566 {
567 	struct rte_eth_link *dst = &(dev->data->dev_link);
568 	struct rte_eth_link *src = link;
569 
570 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
571 					*(uint64_t *)src) == 0)
572 		return -1;
573 
574 	return 0;
575 }
576 
577 static inline void
578 igb_intr_enable(struct rte_eth_dev *dev)
579 {
580 	struct e1000_interrupt *intr =
581 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
582 	struct e1000_hw *hw =
583 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
584 
585 	E1000_WRITE_REG(hw, E1000_IMS, intr->mask);
586 	E1000_WRITE_FLUSH(hw);
587 }
588 
589 static void
590 igb_intr_disable(struct e1000_hw *hw)
591 {
592 	E1000_WRITE_REG(hw, E1000_IMC, ~0);
593 	E1000_WRITE_FLUSH(hw);
594 }
595 
596 static inline void
597 igbvf_intr_enable(struct rte_eth_dev *dev)
598 {
599 	struct e1000_hw *hw =
600 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
601 
602 	/* only for mailbox */
603 	E1000_WRITE_REG(hw, E1000_EIAM, 1 << E1000_VTIVAR_MISC_MAILBOX);
604 	E1000_WRITE_REG(hw, E1000_EIAC, 1 << E1000_VTIVAR_MISC_MAILBOX);
605 	E1000_WRITE_REG(hw, E1000_EIMS, 1 << E1000_VTIVAR_MISC_MAILBOX);
606 	E1000_WRITE_FLUSH(hw);
607 }
608 
609 /* only for mailbox now. If RX/TX needed, should extend this function.  */
610 static void
611 igbvf_set_ivar_map(struct e1000_hw *hw, uint8_t msix_vector)
612 {
613 	uint32_t tmp = 0;
614 
615 	/* mailbox */
616 	tmp |= (msix_vector & E1000_VTIVAR_MISC_INTR_MASK);
617 	tmp |= E1000_VTIVAR_VALID;
618 	E1000_WRITE_REG(hw, E1000_VTIVAR_MISC, tmp);
619 }
620 
621 static void
622 eth_igbvf_configure_msix_intr(struct rte_eth_dev *dev)
623 {
624 	struct e1000_hw *hw =
625 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
626 
627 	/* Configure VF other cause ivar */
628 	igbvf_set_ivar_map(hw, E1000_VTIVAR_MISC_MAILBOX);
629 }
630 
631 static inline int32_t
632 igb_pf_reset_hw(struct e1000_hw *hw)
633 {
634 	uint32_t ctrl_ext;
635 	int32_t status;
636 
637 	status = e1000_reset_hw(hw);
638 
639 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
640 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
641 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
642 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
643 	E1000_WRITE_FLUSH(hw);
644 
645 	return status;
646 }
647 
648 static void
649 igb_identify_hardware(struct rte_eth_dev *dev, struct rte_pci_device *pci_dev)
650 {
651 	struct e1000_hw *hw =
652 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
653 
654 
655 	hw->vendor_id = pci_dev->id.vendor_id;
656 	hw->device_id = pci_dev->id.device_id;
657 	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
658 	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
659 
660 	e1000_set_mac_type(hw);
661 
662 	/* need to check if it is a vf device below */
663 }
664 
665 static int
666 igb_reset_swfw_lock(struct e1000_hw *hw)
667 {
668 	int ret_val;
669 
670 	/*
671 	 * Do mac ops initialization manually here, since we will need
672 	 * some function pointers set by this call.
673 	 */
674 	ret_val = e1000_init_mac_params(hw);
675 	if (ret_val)
676 		return ret_val;
677 
678 	/*
679 	 * SMBI lock should not fail in this early stage. If this is the case,
680 	 * it is due to an improper exit of the application.
681 	 * So force the release of the faulty lock.
682 	 */
683 	if (e1000_get_hw_semaphore_generic(hw) < 0) {
684 		PMD_DRV_LOG(DEBUG, "SMBI lock released");
685 	}
686 	e1000_put_hw_semaphore_generic(hw);
687 
688 	if (hw->mac.ops.acquire_swfw_sync != NULL) {
689 		uint16_t mask;
690 
691 		/*
692 		 * Phy lock should not fail in this early stage. If this is the case,
693 		 * it is due to an improper exit of the application.
694 		 * So force the release of the faulty lock.
695 		 */
696 		mask = E1000_SWFW_PHY0_SM << hw->bus.func;
697 		if (hw->bus.func > E1000_FUNC_1)
698 			mask <<= 2;
699 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
700 			PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released",
701 				    hw->bus.func);
702 		}
703 		hw->mac.ops.release_swfw_sync(hw, mask);
704 
705 		/*
706 		 * This one is more tricky since it is common to all ports; but
707 		 * swfw_sync retries last long enough (1s) to be almost sure that if
708 		 * lock can not be taken it is due to an improper lock of the
709 		 * semaphore.
710 		 */
711 		mask = E1000_SWFW_EEP_SM;
712 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
713 			PMD_DRV_LOG(DEBUG, "SWFW common locks released");
714 		}
715 		hw->mac.ops.release_swfw_sync(hw, mask);
716 	}
717 
718 	return E1000_SUCCESS;
719 }
720 
721 /* Remove all ntuple filters of the device */
722 static int igb_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
723 {
724 	struct e1000_filter_info *filter_info =
725 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
726 	struct e1000_5tuple_filter *p_5tuple;
727 	struct e1000_2tuple_filter *p_2tuple;
728 
729 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
730 		TAILQ_REMOVE(&filter_info->fivetuple_list,
731 			p_5tuple, entries);
732 			rte_free(p_5tuple);
733 	}
734 	filter_info->fivetuple_mask = 0;
735 	while ((p_2tuple = TAILQ_FIRST(&filter_info->twotuple_list))) {
736 		TAILQ_REMOVE(&filter_info->twotuple_list,
737 			p_2tuple, entries);
738 			rte_free(p_2tuple);
739 	}
740 	filter_info->twotuple_mask = 0;
741 
742 	return 0;
743 }
744 
745 /* Remove all flex filters of the device */
746 static int igb_flex_filter_uninit(struct rte_eth_dev *eth_dev)
747 {
748 	struct e1000_filter_info *filter_info =
749 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
750 	struct e1000_flex_filter *p_flex;
751 
752 	while ((p_flex = TAILQ_FIRST(&filter_info->flex_list))) {
753 		TAILQ_REMOVE(&filter_info->flex_list, p_flex, entries);
754 		rte_free(p_flex);
755 	}
756 	filter_info->flex_mask = 0;
757 
758 	return 0;
759 }
760 
761 static int
762 eth_igb_dev_init(struct rte_eth_dev *eth_dev)
763 {
764 	int error = 0;
765 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
766 	struct e1000_hw *hw =
767 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
768 	struct e1000_vfta * shadow_vfta =
769 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
770 	struct e1000_filter_info *filter_info =
771 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
772 	struct e1000_adapter *adapter =
773 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
774 
775 	uint32_t ctrl_ext;
776 
777 	eth_dev->dev_ops = &eth_igb_ops;
778 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
779 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
780 	eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
781 
782 	/* for secondary processes, we don't initialise any further as primary
783 	 * has already done this work. Only check we don't need a different
784 	 * RX function */
785 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
786 		if (eth_dev->data->scattered_rx)
787 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
788 		return 0;
789 	}
790 
791 	rte_eth_copy_pci_info(eth_dev, pci_dev);
792 
793 	hw->hw_addr= (void *)pci_dev->mem_resource[0].addr;
794 
795 	igb_identify_hardware(eth_dev, pci_dev);
796 	if (e1000_setup_init_funcs(hw, FALSE) != E1000_SUCCESS) {
797 		error = -EIO;
798 		goto err_late;
799 	}
800 
801 	e1000_get_bus_info(hw);
802 
803 	/* Reset any pending lock */
804 	if (igb_reset_swfw_lock(hw) != E1000_SUCCESS) {
805 		error = -EIO;
806 		goto err_late;
807 	}
808 
809 	/* Finish initialization */
810 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS) {
811 		error = -EIO;
812 		goto err_late;
813 	}
814 
815 	hw->mac.autoneg = 1;
816 	hw->phy.autoneg_wait_to_complete = 0;
817 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
818 
819 	/* Copper options */
820 	if (hw->phy.media_type == e1000_media_type_copper) {
821 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
822 		hw->phy.disable_polarity_correction = 0;
823 		hw->phy.ms_type = e1000_ms_hw_default;
824 	}
825 
826 	/*
827 	 * Start from a known state, this is important in reading the nvm
828 	 * and mac from that.
829 	 */
830 	igb_pf_reset_hw(hw);
831 
832 	/* Make sure we have a good EEPROM before we read from it */
833 	if (e1000_validate_nvm_checksum(hw) < 0) {
834 		/*
835 		 * Some PCI-E parts fail the first check due to
836 		 * the link being in sleep state, call it again,
837 		 * if it fails a second time its a real issue.
838 		 */
839 		if (e1000_validate_nvm_checksum(hw) < 0) {
840 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
841 			error = -EIO;
842 			goto err_late;
843 		}
844 	}
845 
846 	/* Read the permanent MAC address out of the EEPROM */
847 	if (e1000_read_mac_addr(hw) != 0) {
848 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
849 		error = -EIO;
850 		goto err_late;
851 	}
852 
853 	/* Allocate memory for storing MAC addresses */
854 	eth_dev->data->mac_addrs = rte_zmalloc("e1000",
855 		ETHER_ADDR_LEN * hw->mac.rar_entry_count, 0);
856 	if (eth_dev->data->mac_addrs == NULL) {
857 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
858 						"store MAC addresses",
859 				ETHER_ADDR_LEN * hw->mac.rar_entry_count);
860 		error = -ENOMEM;
861 		goto err_late;
862 	}
863 
864 	/* Copy the permanent MAC address */
865 	ether_addr_copy((struct ether_addr *)hw->mac.addr, &eth_dev->data->mac_addrs[0]);
866 
867 	/* initialize the vfta */
868 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
869 
870 	/* Now initialize the hardware */
871 	if (igb_hardware_init(hw) != 0) {
872 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
873 		rte_free(eth_dev->data->mac_addrs);
874 		eth_dev->data->mac_addrs = NULL;
875 		error = -ENODEV;
876 		goto err_late;
877 	}
878 	hw->mac.get_link_status = 1;
879 	adapter->stopped = 0;
880 
881 	/* Indicate SOL/IDER usage */
882 	if (e1000_check_reset_block(hw) < 0) {
883 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to"
884 					"SOL/IDER session");
885 	}
886 
887 	/* initialize PF if max_vfs not zero */
888 	igb_pf_host_init(eth_dev);
889 
890 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
891 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
892 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
893 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
894 	E1000_WRITE_FLUSH(hw);
895 
896 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
897 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
898 		     pci_dev->id.device_id);
899 
900 	rte_intr_callback_register(&pci_dev->intr_handle,
901 				   eth_igb_interrupt_handler,
902 				   (void *)eth_dev);
903 
904 	/* enable uio/vfio intr/eventfd mapping */
905 	rte_intr_enable(&pci_dev->intr_handle);
906 
907 	/* enable support intr */
908 	igb_intr_enable(eth_dev);
909 
910 	/* initialize filter info */
911 	memset(filter_info, 0,
912 	       sizeof(struct e1000_filter_info));
913 
914 	TAILQ_INIT(&filter_info->flex_list);
915 	TAILQ_INIT(&filter_info->twotuple_list);
916 	TAILQ_INIT(&filter_info->fivetuple_list);
917 
918 	TAILQ_INIT(&igb_filter_ntuple_list);
919 	TAILQ_INIT(&igb_filter_ethertype_list);
920 	TAILQ_INIT(&igb_filter_syn_list);
921 	TAILQ_INIT(&igb_filter_flex_list);
922 	TAILQ_INIT(&igb_filter_rss_list);
923 	TAILQ_INIT(&igb_flow_list);
924 
925 	return 0;
926 
927 err_late:
928 	igb_hw_control_release(hw);
929 
930 	return error;
931 }
932 
933 static int
934 eth_igb_dev_uninit(struct rte_eth_dev *eth_dev)
935 {
936 	struct rte_pci_device *pci_dev;
937 	struct rte_intr_handle *intr_handle;
938 	struct e1000_hw *hw;
939 	struct e1000_adapter *adapter =
940 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
941 	struct e1000_filter_info *filter_info =
942 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
943 
944 	PMD_INIT_FUNC_TRACE();
945 
946 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
947 		return -EPERM;
948 
949 	hw = E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
950 	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
951 	intr_handle = &pci_dev->intr_handle;
952 
953 	if (adapter->stopped == 0)
954 		eth_igb_close(eth_dev);
955 
956 	eth_dev->dev_ops = NULL;
957 	eth_dev->rx_pkt_burst = NULL;
958 	eth_dev->tx_pkt_burst = NULL;
959 
960 	/* Reset any pending lock */
961 	igb_reset_swfw_lock(hw);
962 
963 	rte_free(eth_dev->data->mac_addrs);
964 	eth_dev->data->mac_addrs = NULL;
965 
966 	/* uninitialize PF if max_vfs not zero */
967 	igb_pf_host_uninit(eth_dev);
968 
969 	/* disable uio intr before callback unregister */
970 	rte_intr_disable(intr_handle);
971 	rte_intr_callback_unregister(intr_handle,
972 				     eth_igb_interrupt_handler, eth_dev);
973 
974 	/* clear the SYN filter info */
975 	filter_info->syn_info = 0;
976 
977 	/* clear the ethertype filters info */
978 	filter_info->ethertype_mask = 0;
979 	memset(filter_info->ethertype_filters, 0,
980 		E1000_MAX_ETQF_FILTERS * sizeof(struct igb_ethertype_filter));
981 
982 	/* clear the rss filter info */
983 	memset(&filter_info->rss_info, 0,
984 		sizeof(struct igb_rte_flow_rss_conf));
985 
986 	/* remove all ntuple filters of the device */
987 	igb_ntuple_filter_uninit(eth_dev);
988 
989 	/* remove all flex filters of the device */
990 	igb_flex_filter_uninit(eth_dev);
991 
992 	/* clear all the filters list */
993 	igb_filterlist_flush(eth_dev);
994 
995 	return 0;
996 }
997 
998 /*
999  * Virtual Function device init
1000  */
1001 static int
1002 eth_igbvf_dev_init(struct rte_eth_dev *eth_dev)
1003 {
1004 	struct rte_pci_device *pci_dev;
1005 	struct rte_intr_handle *intr_handle;
1006 	struct e1000_adapter *adapter =
1007 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
1008 	struct e1000_hw *hw =
1009 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1010 	int diag;
1011 	struct ether_addr *perm_addr = (struct ether_addr *)hw->mac.perm_addr;
1012 
1013 	PMD_INIT_FUNC_TRACE();
1014 
1015 	eth_dev->dev_ops = &igbvf_eth_dev_ops;
1016 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
1017 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
1018 	eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
1019 
1020 	/* for secondary processes, we don't initialise any further as primary
1021 	 * has already done this work. Only check we don't need a different
1022 	 * RX function */
1023 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
1024 		if (eth_dev->data->scattered_rx)
1025 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
1026 		return 0;
1027 	}
1028 
1029 	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1030 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1031 
1032 	hw->device_id = pci_dev->id.device_id;
1033 	hw->vendor_id = pci_dev->id.vendor_id;
1034 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1035 	adapter->stopped = 0;
1036 
1037 	/* Initialize the shared code (base driver) */
1038 	diag = e1000_setup_init_funcs(hw, TRUE);
1039 	if (diag != 0) {
1040 		PMD_INIT_LOG(ERR, "Shared code init failed for igbvf: %d",
1041 			diag);
1042 		return -EIO;
1043 	}
1044 
1045 	/* init_mailbox_params */
1046 	hw->mbx.ops.init_params(hw);
1047 
1048 	/* Disable the interrupts for VF */
1049 	igbvf_intr_disable(hw);
1050 
1051 	diag = hw->mac.ops.reset_hw(hw);
1052 
1053 	/* Allocate memory for storing MAC addresses */
1054 	eth_dev->data->mac_addrs = rte_zmalloc("igbvf", ETHER_ADDR_LEN *
1055 		hw->mac.rar_entry_count, 0);
1056 	if (eth_dev->data->mac_addrs == NULL) {
1057 		PMD_INIT_LOG(ERR,
1058 			"Failed to allocate %d bytes needed to store MAC "
1059 			"addresses",
1060 			ETHER_ADDR_LEN * hw->mac.rar_entry_count);
1061 		return -ENOMEM;
1062 	}
1063 
1064 	/* Generate a random MAC address, if none was assigned by PF. */
1065 	if (is_zero_ether_addr(perm_addr)) {
1066 		eth_random_addr(perm_addr->addr_bytes);
1067 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1068 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1069 			     "%02x:%02x:%02x:%02x:%02x:%02x",
1070 			     perm_addr->addr_bytes[0],
1071 			     perm_addr->addr_bytes[1],
1072 			     perm_addr->addr_bytes[2],
1073 			     perm_addr->addr_bytes[3],
1074 			     perm_addr->addr_bytes[4],
1075 			     perm_addr->addr_bytes[5]);
1076 	}
1077 
1078 	diag = e1000_rar_set(hw, perm_addr->addr_bytes, 0);
1079 	if (diag) {
1080 		rte_free(eth_dev->data->mac_addrs);
1081 		eth_dev->data->mac_addrs = NULL;
1082 		return diag;
1083 	}
1084 	/* Copy the permanent MAC address */
1085 	ether_addr_copy((struct ether_addr *) hw->mac.perm_addr,
1086 			&eth_dev->data->mac_addrs[0]);
1087 
1088 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x "
1089 		     "mac.type=%s",
1090 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1091 		     pci_dev->id.device_id, "igb_mac_82576_vf");
1092 
1093 	intr_handle = &pci_dev->intr_handle;
1094 	rte_intr_callback_register(intr_handle,
1095 				   eth_igbvf_interrupt_handler, eth_dev);
1096 
1097 	return 0;
1098 }
1099 
1100 static int
1101 eth_igbvf_dev_uninit(struct rte_eth_dev *eth_dev)
1102 {
1103 	struct e1000_adapter *adapter =
1104 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
1105 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1106 
1107 	PMD_INIT_FUNC_TRACE();
1108 
1109 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1110 		return -EPERM;
1111 
1112 	if (adapter->stopped == 0)
1113 		igbvf_dev_close(eth_dev);
1114 
1115 	eth_dev->dev_ops = NULL;
1116 	eth_dev->rx_pkt_burst = NULL;
1117 	eth_dev->tx_pkt_burst = NULL;
1118 
1119 	rte_free(eth_dev->data->mac_addrs);
1120 	eth_dev->data->mac_addrs = NULL;
1121 
1122 	/* disable uio intr before callback unregister */
1123 	rte_intr_disable(&pci_dev->intr_handle);
1124 	rte_intr_callback_unregister(&pci_dev->intr_handle,
1125 				     eth_igbvf_interrupt_handler,
1126 				     (void *)eth_dev);
1127 
1128 	return 0;
1129 }
1130 
1131 static int eth_igb_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1132 	struct rte_pci_device *pci_dev)
1133 {
1134 	return rte_eth_dev_pci_generic_probe(pci_dev,
1135 		sizeof(struct e1000_adapter), eth_igb_dev_init);
1136 }
1137 
1138 static int eth_igb_pci_remove(struct rte_pci_device *pci_dev)
1139 {
1140 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igb_dev_uninit);
1141 }
1142 
1143 static struct rte_pci_driver rte_igb_pmd = {
1144 	.id_table = pci_id_igb_map,
1145 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
1146 		     RTE_PCI_DRV_IOVA_AS_VA,
1147 	.probe = eth_igb_pci_probe,
1148 	.remove = eth_igb_pci_remove,
1149 };
1150 
1151 
1152 static int eth_igbvf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1153 	struct rte_pci_device *pci_dev)
1154 {
1155 	return rte_eth_dev_pci_generic_probe(pci_dev,
1156 		sizeof(struct e1000_adapter), eth_igbvf_dev_init);
1157 }
1158 
1159 static int eth_igbvf_pci_remove(struct rte_pci_device *pci_dev)
1160 {
1161 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igbvf_dev_uninit);
1162 }
1163 
1164 /*
1165  * virtual function driver struct
1166  */
1167 static struct rte_pci_driver rte_igbvf_pmd = {
1168 	.id_table = pci_id_igbvf_map,
1169 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_IOVA_AS_VA,
1170 	.probe = eth_igbvf_pci_probe,
1171 	.remove = eth_igbvf_pci_remove,
1172 };
1173 
1174 static void
1175 igb_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1176 {
1177 	struct e1000_hw *hw =
1178 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1179 	/* RCTL: enable VLAN filter since VMDq always use VLAN filter */
1180 	uint32_t rctl = E1000_READ_REG(hw, E1000_RCTL);
1181 	rctl |= E1000_RCTL_VFE;
1182 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1183 }
1184 
1185 static int
1186 igb_check_mq_mode(struct rte_eth_dev *dev)
1187 {
1188 	enum rte_eth_rx_mq_mode rx_mq_mode = dev->data->dev_conf.rxmode.mq_mode;
1189 	enum rte_eth_tx_mq_mode tx_mq_mode = dev->data->dev_conf.txmode.mq_mode;
1190 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
1191 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
1192 
1193 	if ((rx_mq_mode & ETH_MQ_RX_DCB_FLAG) ||
1194 	    tx_mq_mode == ETH_MQ_TX_DCB ||
1195 	    tx_mq_mode == ETH_MQ_TX_VMDQ_DCB) {
1196 		PMD_INIT_LOG(ERR, "DCB mode is not supported.");
1197 		return -EINVAL;
1198 	}
1199 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
1200 		/* Check multi-queue mode.
1201 		 * To no break software we accept ETH_MQ_RX_NONE as this might
1202 		 * be used to turn off VLAN filter.
1203 		 */
1204 
1205 		if (rx_mq_mode == ETH_MQ_RX_NONE ||
1206 		    rx_mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1207 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
1208 			RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = 1;
1209 		} else {
1210 			/* Only support one queue on VFs.
1211 			 * RSS together with SRIOV is not supported.
1212 			 */
1213 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1214 					" wrong mq_mode rx %d.",
1215 					rx_mq_mode);
1216 			return -EINVAL;
1217 		}
1218 		/* TX mode is not used here, so mode might be ignored.*/
1219 		if (tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1220 			/* SRIOV only works in VMDq enable mode */
1221 			PMD_INIT_LOG(WARNING, "SRIOV is active,"
1222 					" TX mode %d is not supported. "
1223 					" Driver will behave as %d mode.",
1224 					tx_mq_mode, ETH_MQ_TX_VMDQ_ONLY);
1225 		}
1226 
1227 		/* check valid queue number */
1228 		if ((nb_rx_q > 1) || (nb_tx_q > 1)) {
1229 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1230 					" only support one queue on VFs.");
1231 			return -EINVAL;
1232 		}
1233 	} else {
1234 		/* To no break software that set invalid mode, only display
1235 		 * warning if invalid mode is used.
1236 		 */
1237 		if (rx_mq_mode != ETH_MQ_RX_NONE &&
1238 		    rx_mq_mode != ETH_MQ_RX_VMDQ_ONLY &&
1239 		    rx_mq_mode != ETH_MQ_RX_RSS) {
1240 			/* RSS together with VMDq not supported*/
1241 			PMD_INIT_LOG(ERR, "RX mode %d is not supported.",
1242 				     rx_mq_mode);
1243 			return -EINVAL;
1244 		}
1245 
1246 		if (tx_mq_mode != ETH_MQ_TX_NONE &&
1247 		    tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1248 			PMD_INIT_LOG(WARNING, "TX mode %d is not supported."
1249 					" Due to txmode is meaningless in this"
1250 					" driver, just ignore.",
1251 					tx_mq_mode);
1252 		}
1253 	}
1254 	return 0;
1255 }
1256 
1257 static int
1258 eth_igb_configure(struct rte_eth_dev *dev)
1259 {
1260 	struct e1000_interrupt *intr =
1261 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1262 	int ret;
1263 
1264 	PMD_INIT_FUNC_TRACE();
1265 
1266 	/* multipe queue mode checking */
1267 	ret  = igb_check_mq_mode(dev);
1268 	if (ret != 0) {
1269 		PMD_DRV_LOG(ERR, "igb_check_mq_mode fails with %d.",
1270 			    ret);
1271 		return ret;
1272 	}
1273 
1274 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1275 	PMD_INIT_FUNC_TRACE();
1276 
1277 	return 0;
1278 }
1279 
1280 static void
1281 eth_igb_rxtx_control(struct rte_eth_dev *dev,
1282 		     bool enable)
1283 {
1284 	struct e1000_hw *hw =
1285 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1286 	uint32_t tctl, rctl;
1287 
1288 	tctl = E1000_READ_REG(hw, E1000_TCTL);
1289 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1290 
1291 	if (enable) {
1292 		/* enable Tx/Rx */
1293 		tctl |= E1000_TCTL_EN;
1294 		rctl |= E1000_RCTL_EN;
1295 	} else {
1296 		/* disable Tx/Rx */
1297 		tctl &= ~E1000_TCTL_EN;
1298 		rctl &= ~E1000_RCTL_EN;
1299 	}
1300 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
1301 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1302 	E1000_WRITE_FLUSH(hw);
1303 }
1304 
1305 static int
1306 eth_igb_start(struct rte_eth_dev *dev)
1307 {
1308 	struct e1000_hw *hw =
1309 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1310 	struct e1000_adapter *adapter =
1311 		E1000_DEV_PRIVATE(dev->data->dev_private);
1312 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1313 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1314 	int ret, mask;
1315 	uint32_t intr_vector = 0;
1316 	uint32_t ctrl_ext;
1317 	uint32_t *speeds;
1318 	int num_speeds;
1319 	bool autoneg;
1320 
1321 	PMD_INIT_FUNC_TRACE();
1322 
1323 	/* disable uio/vfio intr/eventfd mapping */
1324 	rte_intr_disable(intr_handle);
1325 
1326 	/* Power up the phy. Needed to make the link go Up */
1327 	eth_igb_dev_set_link_up(dev);
1328 
1329 	/*
1330 	 * Packet Buffer Allocation (PBA)
1331 	 * Writing PBA sets the receive portion of the buffer
1332 	 * the remainder is used for the transmit buffer.
1333 	 */
1334 	if (hw->mac.type == e1000_82575) {
1335 		uint32_t pba;
1336 
1337 		pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
1338 		E1000_WRITE_REG(hw, E1000_PBA, pba);
1339 	}
1340 
1341 	/* Put the address into the Receive Address Array */
1342 	e1000_rar_set(hw, hw->mac.addr, 0);
1343 
1344 	/* Initialize the hardware */
1345 	if (igb_hardware_init(hw)) {
1346 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
1347 		return -EIO;
1348 	}
1349 	adapter->stopped = 0;
1350 
1351 	E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN << 16 | ETHER_TYPE_VLAN);
1352 
1353 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1354 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1355 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
1356 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
1357 	E1000_WRITE_FLUSH(hw);
1358 
1359 	/* configure PF module if SRIOV enabled */
1360 	igb_pf_host_configure(dev);
1361 
1362 	/* check and configure queue intr-vector mapping */
1363 	if ((rte_intr_cap_multiple(intr_handle) ||
1364 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
1365 	    dev->data->dev_conf.intr_conf.rxq != 0) {
1366 		intr_vector = dev->data->nb_rx_queues;
1367 		if (rte_intr_efd_enable(intr_handle, intr_vector))
1368 			return -1;
1369 	}
1370 
1371 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
1372 		intr_handle->intr_vec =
1373 			rte_zmalloc("intr_vec",
1374 				    dev->data->nb_rx_queues * sizeof(int), 0);
1375 		if (intr_handle->intr_vec == NULL) {
1376 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
1377 				     " intr_vec", dev->data->nb_rx_queues);
1378 			return -ENOMEM;
1379 		}
1380 	}
1381 
1382 	/* confiugre msix for rx interrupt */
1383 	eth_igb_configure_msix_intr(dev);
1384 
1385 	/* Configure for OS presence */
1386 	igb_init_manageability(hw);
1387 
1388 	eth_igb_tx_init(dev);
1389 
1390 	/* This can fail when allocating mbufs for descriptor rings */
1391 	ret = eth_igb_rx_init(dev);
1392 	if (ret) {
1393 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
1394 		igb_dev_clear_queues(dev);
1395 		return ret;
1396 	}
1397 
1398 	e1000_clear_hw_cntrs_base_generic(hw);
1399 
1400 	/*
1401 	 * VLAN Offload Settings
1402 	 */
1403 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
1404 			ETH_VLAN_EXTEND_MASK;
1405 	ret = eth_igb_vlan_offload_set(dev, mask);
1406 	if (ret) {
1407 		PMD_INIT_LOG(ERR, "Unable to set vlan offload");
1408 		igb_dev_clear_queues(dev);
1409 		return ret;
1410 	}
1411 
1412 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1413 		/* Enable VLAN filter since VMDq always use VLAN filter */
1414 		igb_vmdq_vlan_hw_filter_enable(dev);
1415 	}
1416 
1417 	if ((hw->mac.type == e1000_82576) || (hw->mac.type == e1000_82580) ||
1418 		(hw->mac.type == e1000_i350) || (hw->mac.type == e1000_i210) ||
1419 		(hw->mac.type == e1000_i211)) {
1420 		/* Configure EITR with the maximum possible value (0xFFFF) */
1421 		E1000_WRITE_REG(hw, E1000_EITR(0), 0xFFFF);
1422 	}
1423 
1424 	/* Setup link speed and duplex */
1425 	speeds = &dev->data->dev_conf.link_speeds;
1426 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
1427 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
1428 		hw->mac.autoneg = 1;
1429 	} else {
1430 		num_speeds = 0;
1431 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
1432 
1433 		/* Reset */
1434 		hw->phy.autoneg_advertised = 0;
1435 
1436 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1437 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1438 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
1439 			num_speeds = -1;
1440 			goto error_invalid_config;
1441 		}
1442 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
1443 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1444 			num_speeds++;
1445 		}
1446 		if (*speeds & ETH_LINK_SPEED_10M) {
1447 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1448 			num_speeds++;
1449 		}
1450 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
1451 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1452 			num_speeds++;
1453 		}
1454 		if (*speeds & ETH_LINK_SPEED_100M) {
1455 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1456 			num_speeds++;
1457 		}
1458 		if (*speeds & ETH_LINK_SPEED_1G) {
1459 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1460 			num_speeds++;
1461 		}
1462 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
1463 			goto error_invalid_config;
1464 
1465 		/* Set/reset the mac.autoneg based on the link speed,
1466 		 * fixed or not
1467 		 */
1468 		if (!autoneg) {
1469 			hw->mac.autoneg = 0;
1470 			hw->mac.forced_speed_duplex =
1471 					hw->phy.autoneg_advertised;
1472 		} else {
1473 			hw->mac.autoneg = 1;
1474 		}
1475 	}
1476 
1477 	e1000_setup_link(hw);
1478 
1479 	if (rte_intr_allow_others(intr_handle)) {
1480 		/* check if lsc interrupt is enabled */
1481 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1482 			eth_igb_lsc_interrupt_setup(dev, TRUE);
1483 		else
1484 			eth_igb_lsc_interrupt_setup(dev, FALSE);
1485 	} else {
1486 		rte_intr_callback_unregister(intr_handle,
1487 					     eth_igb_interrupt_handler,
1488 					     (void *)dev);
1489 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1490 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
1491 				     " no intr multiplex");
1492 	}
1493 
1494 	/* check if rxq interrupt is enabled */
1495 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
1496 	    rte_intr_dp_is_en(intr_handle))
1497 		eth_igb_rxq_interrupt_setup(dev);
1498 
1499 	/* enable uio/vfio intr/eventfd mapping */
1500 	rte_intr_enable(intr_handle);
1501 
1502 	/* resume enabled intr since hw reset */
1503 	igb_intr_enable(dev);
1504 
1505 	/* restore all types filter */
1506 	igb_filter_restore(dev);
1507 
1508 	eth_igb_rxtx_control(dev, true);
1509 	eth_igb_link_update(dev, 0);
1510 
1511 	PMD_INIT_LOG(DEBUG, "<<");
1512 
1513 	return 0;
1514 
1515 error_invalid_config:
1516 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
1517 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
1518 	igb_dev_clear_queues(dev);
1519 	return -EINVAL;
1520 }
1521 
1522 /*********************************************************************
1523  *
1524  *  This routine disables all traffic on the adapter by issuing a
1525  *  global reset on the MAC.
1526  *
1527  **********************************************************************/
1528 static void
1529 eth_igb_stop(struct rte_eth_dev *dev)
1530 {
1531 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1532 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1533 	struct rte_eth_link link;
1534 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1535 
1536 	eth_igb_rxtx_control(dev, false);
1537 
1538 	igb_intr_disable(hw);
1539 
1540 	/* disable intr eventfd mapping */
1541 	rte_intr_disable(intr_handle);
1542 
1543 	igb_pf_reset_hw(hw);
1544 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1545 
1546 	/* Set bit for Go Link disconnect */
1547 	if (hw->mac.type >= e1000_82580) {
1548 		uint32_t phpm_reg;
1549 
1550 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1551 		phpm_reg |= E1000_82580_PM_GO_LINKD;
1552 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1553 	}
1554 
1555 	/* Power down the phy. Needed to make the link go Down */
1556 	eth_igb_dev_set_link_down(dev);
1557 
1558 	igb_dev_clear_queues(dev);
1559 
1560 	/* clear the recorded link status */
1561 	memset(&link, 0, sizeof(link));
1562 	rte_igb_dev_atomic_write_link_status(dev, &link);
1563 
1564 	if (!rte_intr_allow_others(intr_handle))
1565 		/* resume to the default handler */
1566 		rte_intr_callback_register(intr_handle,
1567 					   eth_igb_interrupt_handler,
1568 					   (void *)dev);
1569 
1570 	/* Clean datapath event and queue/vec mapping */
1571 	rte_intr_efd_disable(intr_handle);
1572 	if (intr_handle->intr_vec != NULL) {
1573 		rte_free(intr_handle->intr_vec);
1574 		intr_handle->intr_vec = NULL;
1575 	}
1576 }
1577 
1578 static int
1579 eth_igb_dev_set_link_up(struct rte_eth_dev *dev)
1580 {
1581 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1582 
1583 	if (hw->phy.media_type == e1000_media_type_copper)
1584 		e1000_power_up_phy(hw);
1585 	else
1586 		e1000_power_up_fiber_serdes_link(hw);
1587 
1588 	return 0;
1589 }
1590 
1591 static int
1592 eth_igb_dev_set_link_down(struct rte_eth_dev *dev)
1593 {
1594 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1595 
1596 	if (hw->phy.media_type == e1000_media_type_copper)
1597 		e1000_power_down_phy(hw);
1598 	else
1599 		e1000_shutdown_fiber_serdes_link(hw);
1600 
1601 	return 0;
1602 }
1603 
1604 static void
1605 eth_igb_close(struct rte_eth_dev *dev)
1606 {
1607 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1608 	struct e1000_adapter *adapter =
1609 		E1000_DEV_PRIVATE(dev->data->dev_private);
1610 	struct rte_eth_link link;
1611 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1612 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1613 
1614 	eth_igb_stop(dev);
1615 	adapter->stopped = 1;
1616 
1617 	e1000_phy_hw_reset(hw);
1618 	igb_release_manageability(hw);
1619 	igb_hw_control_release(hw);
1620 
1621 	/* Clear bit for Go Link disconnect */
1622 	if (hw->mac.type >= e1000_82580) {
1623 		uint32_t phpm_reg;
1624 
1625 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1626 		phpm_reg &= ~E1000_82580_PM_GO_LINKD;
1627 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1628 	}
1629 
1630 	igb_dev_free_queues(dev);
1631 
1632 	if (intr_handle->intr_vec) {
1633 		rte_free(intr_handle->intr_vec);
1634 		intr_handle->intr_vec = NULL;
1635 	}
1636 
1637 	memset(&link, 0, sizeof(link));
1638 	rte_igb_dev_atomic_write_link_status(dev, &link);
1639 }
1640 
1641 static int
1642 igb_get_rx_buffer_size(struct e1000_hw *hw)
1643 {
1644 	uint32_t rx_buf_size;
1645 	if (hw->mac.type == e1000_82576) {
1646 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xffff) << 10;
1647 	} else if (hw->mac.type == e1000_82580 || hw->mac.type == e1000_i350) {
1648 		/* PBS needs to be translated according to a lookup table */
1649 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xf);
1650 		rx_buf_size = (uint32_t) e1000_rxpbs_adjust_82580(rx_buf_size);
1651 		rx_buf_size = (rx_buf_size << 10);
1652 	} else if (hw->mac.type == e1000_i210 || hw->mac.type == e1000_i211) {
1653 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0x3f) << 10;
1654 	} else {
1655 		rx_buf_size = (E1000_READ_REG(hw, E1000_PBA) & 0xffff) << 10;
1656 	}
1657 
1658 	return rx_buf_size;
1659 }
1660 
1661 /*********************************************************************
1662  *
1663  *  Initialize the hardware
1664  *
1665  **********************************************************************/
1666 static int
1667 igb_hardware_init(struct e1000_hw *hw)
1668 {
1669 	uint32_t rx_buf_size;
1670 	int diag;
1671 
1672 	/* Let the firmware know the OS is in control */
1673 	igb_hw_control_acquire(hw);
1674 
1675 	/*
1676 	 * These parameters control the automatic generation (Tx) and
1677 	 * response (Rx) to Ethernet PAUSE frames.
1678 	 * - High water mark should allow for at least two standard size (1518)
1679 	 *   frames to be received after sending an XOFF.
1680 	 * - Low water mark works best when it is very near the high water mark.
1681 	 *   This allows the receiver to restart by sending XON when it has
1682 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
1683 	 *   restart after one full frame is pulled from the buffer. There
1684 	 *   could be several smaller frames in the buffer and if so they will
1685 	 *   not trigger the XON until their total number reduces the buffer
1686 	 *   by 1500.
1687 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
1688 	 */
1689 	rx_buf_size = igb_get_rx_buffer_size(hw);
1690 
1691 	hw->fc.high_water = rx_buf_size - (ETHER_MAX_LEN * 2);
1692 	hw->fc.low_water = hw->fc.high_water - 1500;
1693 	hw->fc.pause_time = IGB_FC_PAUSE_TIME;
1694 	hw->fc.send_xon = 1;
1695 
1696 	/* Set Flow control, use the tunable location if sane */
1697 	if ((igb_fc_setting != e1000_fc_none) && (igb_fc_setting < 4))
1698 		hw->fc.requested_mode = igb_fc_setting;
1699 	else
1700 		hw->fc.requested_mode = e1000_fc_none;
1701 
1702 	/* Issue a global reset */
1703 	igb_pf_reset_hw(hw);
1704 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1705 
1706 	diag = e1000_init_hw(hw);
1707 	if (diag < 0)
1708 		return diag;
1709 
1710 	E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN << 16 | ETHER_TYPE_VLAN);
1711 	e1000_get_phy_info(hw);
1712 	e1000_check_for_link(hw);
1713 
1714 	return 0;
1715 }
1716 
1717 /* This function is based on igb_update_stats_counters() in igb/if_igb.c */
1718 static void
1719 igb_read_stats_registers(struct e1000_hw *hw, struct e1000_hw_stats *stats)
1720 {
1721 	int pause_frames;
1722 
1723 	uint64_t old_gprc  = stats->gprc;
1724 	uint64_t old_gptc  = stats->gptc;
1725 	uint64_t old_tpr   = stats->tpr;
1726 	uint64_t old_tpt   = stats->tpt;
1727 	uint64_t old_rpthc = stats->rpthc;
1728 	uint64_t old_hgptc = stats->hgptc;
1729 
1730 	if(hw->phy.media_type == e1000_media_type_copper ||
1731 	    (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
1732 		stats->symerrs +=
1733 		    E1000_READ_REG(hw,E1000_SYMERRS);
1734 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
1735 	}
1736 
1737 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
1738 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
1739 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
1740 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
1741 
1742 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
1743 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
1744 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
1745 	stats->dc += E1000_READ_REG(hw, E1000_DC);
1746 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
1747 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
1748 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
1749 	/*
1750 	** For watchdog management we need to know if we have been
1751 	** paused during the last interval, so capture that here.
1752 	*/
1753 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
1754 	stats->xoffrxc += pause_frames;
1755 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
1756 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
1757 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
1758 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
1759 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
1760 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
1761 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
1762 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
1763 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
1764 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
1765 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
1766 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
1767 
1768 	/* For the 64-bit byte counters the low dword must be read first. */
1769 	/* Both registers clear on the read of the high dword */
1770 
1771 	/* Workaround CRC bytes included in size, take away 4 bytes/packet */
1772 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
1773 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
1774 	stats->gorc -= (stats->gprc - old_gprc) * ETHER_CRC_LEN;
1775 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
1776 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
1777 	stats->gotc -= (stats->gptc - old_gptc) * ETHER_CRC_LEN;
1778 
1779 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
1780 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
1781 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
1782 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
1783 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
1784 
1785 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
1786 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
1787 
1788 	stats->tor += E1000_READ_REG(hw, E1000_TORL);
1789 	stats->tor += ((uint64_t)E1000_READ_REG(hw, E1000_TORH) << 32);
1790 	stats->tor -= (stats->tpr - old_tpr) * ETHER_CRC_LEN;
1791 	stats->tot += E1000_READ_REG(hw, E1000_TOTL);
1792 	stats->tot += ((uint64_t)E1000_READ_REG(hw, E1000_TOTH) << 32);
1793 	stats->tot -= (stats->tpt - old_tpt) * ETHER_CRC_LEN;
1794 
1795 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
1796 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
1797 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
1798 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
1799 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
1800 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
1801 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
1802 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
1803 
1804 	/* Interrupt Counts */
1805 
1806 	stats->iac += E1000_READ_REG(hw, E1000_IAC);
1807 	stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
1808 	stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
1809 	stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
1810 	stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
1811 	stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
1812 	stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
1813 	stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
1814 	stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
1815 
1816 	/* Host to Card Statistics */
1817 
1818 	stats->cbtmpc += E1000_READ_REG(hw, E1000_CBTMPC);
1819 	stats->htdpmc += E1000_READ_REG(hw, E1000_HTDPMC);
1820 	stats->cbrdpc += E1000_READ_REG(hw, E1000_CBRDPC);
1821 	stats->cbrmpc += E1000_READ_REG(hw, E1000_CBRMPC);
1822 	stats->rpthc += E1000_READ_REG(hw, E1000_RPTHC);
1823 	stats->hgptc += E1000_READ_REG(hw, E1000_HGPTC);
1824 	stats->htcbdpc += E1000_READ_REG(hw, E1000_HTCBDPC);
1825 	stats->hgorc += E1000_READ_REG(hw, E1000_HGORCL);
1826 	stats->hgorc += ((uint64_t)E1000_READ_REG(hw, E1000_HGORCH) << 32);
1827 	stats->hgorc -= (stats->rpthc - old_rpthc) * ETHER_CRC_LEN;
1828 	stats->hgotc += E1000_READ_REG(hw, E1000_HGOTCL);
1829 	stats->hgotc += ((uint64_t)E1000_READ_REG(hw, E1000_HGOTCH) << 32);
1830 	stats->hgotc -= (stats->hgptc - old_hgptc) * ETHER_CRC_LEN;
1831 	stats->lenerrs += E1000_READ_REG(hw, E1000_LENERRS);
1832 	stats->scvpc += E1000_READ_REG(hw, E1000_SCVPC);
1833 	stats->hrmpc += E1000_READ_REG(hw, E1000_HRMPC);
1834 
1835 	stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
1836 	stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
1837 	stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
1838 	stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
1839 	stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
1840 	stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
1841 }
1842 
1843 static int
1844 eth_igb_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1845 {
1846 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1847 	struct e1000_hw_stats *stats =
1848 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1849 
1850 	igb_read_stats_registers(hw, stats);
1851 
1852 	if (rte_stats == NULL)
1853 		return -EINVAL;
1854 
1855 	/* Rx Errors */
1856 	rte_stats->imissed = stats->mpc;
1857 	rte_stats->ierrors = stats->crcerrs +
1858 	                     stats->rlec + stats->ruc + stats->roc +
1859 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
1860 
1861 	/* Tx Errors */
1862 	rte_stats->oerrors = stats->ecol + stats->latecol;
1863 
1864 	rte_stats->ipackets = stats->gprc;
1865 	rte_stats->opackets = stats->gptc;
1866 	rte_stats->ibytes   = stats->gorc;
1867 	rte_stats->obytes   = stats->gotc;
1868 	return 0;
1869 }
1870 
1871 static void
1872 eth_igb_stats_reset(struct rte_eth_dev *dev)
1873 {
1874 	struct e1000_hw_stats *hw_stats =
1875 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1876 
1877 	/* HW registers are cleared on read */
1878 	eth_igb_stats_get(dev, NULL);
1879 
1880 	/* Reset software totals */
1881 	memset(hw_stats, 0, sizeof(*hw_stats));
1882 }
1883 
1884 static void
1885 eth_igb_xstats_reset(struct rte_eth_dev *dev)
1886 {
1887 	struct e1000_hw_stats *stats =
1888 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1889 
1890 	/* HW registers are cleared on read */
1891 	eth_igb_xstats_get(dev, NULL, IGB_NB_XSTATS);
1892 
1893 	/* Reset software totals */
1894 	memset(stats, 0, sizeof(*stats));
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 		snprintf(xstats_names[i].name, sizeof(xstats_names[i].name),
1910 			 "%s", rte_igb_stats_strings[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 			snprintf(xstats_names[i].name,
1928 					sizeof(xstats_names[i].name),
1929 					"%s", rte_igb_stats_strings[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 			snprintf(xstats_names[i].name,
2077 				sizeof(xstats_names[i].name), "%s",
2078 				rte_igbvf_stats_strings[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 void
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 
2142 static int
2143 eth_igb_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
2144 		       size_t fw_size)
2145 {
2146 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2147 	struct e1000_fw_version fw;
2148 	int ret;
2149 
2150 	e1000_get_fw_version(hw, &fw);
2151 
2152 	switch (hw->mac.type) {
2153 	case e1000_i210:
2154 	case e1000_i211:
2155 		if (!(e1000_get_flash_presence_i210(hw))) {
2156 			ret = snprintf(fw_version, fw_size,
2157 				 "%2d.%2d-%d",
2158 				 fw.invm_major, fw.invm_minor,
2159 				 fw.invm_img_type);
2160 			break;
2161 		}
2162 		/* fall through */
2163 	default:
2164 		/* if option rom is valid, display its version too */
2165 		if (fw.or_valid) {
2166 			ret = snprintf(fw_version, fw_size,
2167 				 "%d.%d, 0x%08x, %d.%d.%d",
2168 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
2169 				 fw.or_major, fw.or_build, fw.or_patch);
2170 		/* no option rom */
2171 		} else {
2172 			if (fw.etrack_id != 0X0000) {
2173 				ret = snprintf(fw_version, fw_size,
2174 					 "%d.%d, 0x%08x",
2175 					 fw.eep_major, fw.eep_minor,
2176 					 fw.etrack_id);
2177 			} else {
2178 				ret = snprintf(fw_version, fw_size,
2179 					 "%d.%d.%d",
2180 					 fw.eep_major, fw.eep_minor,
2181 					 fw.eep_build);
2182 			}
2183 		}
2184 		break;
2185 	}
2186 
2187 	ret += 1; /* add the size of '\0' */
2188 	if (fw_size < (u32)ret)
2189 		return ret;
2190 	else
2191 		return 0;
2192 }
2193 
2194 static void
2195 eth_igb_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2196 {
2197 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2198 
2199 	dev_info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2200 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2201 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2202 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2203 	dev_info->rx_offload_capa =
2204 		DEV_RX_OFFLOAD_VLAN_STRIP |
2205 		DEV_RX_OFFLOAD_IPV4_CKSUM |
2206 		DEV_RX_OFFLOAD_UDP_CKSUM  |
2207 		DEV_RX_OFFLOAD_TCP_CKSUM;
2208 	dev_info->tx_offload_capa =
2209 		DEV_TX_OFFLOAD_VLAN_INSERT |
2210 		DEV_TX_OFFLOAD_IPV4_CKSUM  |
2211 		DEV_TX_OFFLOAD_UDP_CKSUM   |
2212 		DEV_TX_OFFLOAD_TCP_CKSUM   |
2213 		DEV_TX_OFFLOAD_SCTP_CKSUM  |
2214 		DEV_TX_OFFLOAD_TCP_TSO;
2215 
2216 	switch (hw->mac.type) {
2217 	case e1000_82575:
2218 		dev_info->max_rx_queues = 4;
2219 		dev_info->max_tx_queues = 4;
2220 		dev_info->max_vmdq_pools = 0;
2221 		break;
2222 
2223 	case e1000_82576:
2224 		dev_info->max_rx_queues = 16;
2225 		dev_info->max_tx_queues = 16;
2226 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2227 		dev_info->vmdq_queue_num = 16;
2228 		break;
2229 
2230 	case e1000_82580:
2231 		dev_info->max_rx_queues = 8;
2232 		dev_info->max_tx_queues = 8;
2233 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2234 		dev_info->vmdq_queue_num = 8;
2235 		break;
2236 
2237 	case e1000_i350:
2238 		dev_info->max_rx_queues = 8;
2239 		dev_info->max_tx_queues = 8;
2240 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2241 		dev_info->vmdq_queue_num = 8;
2242 		break;
2243 
2244 	case e1000_i354:
2245 		dev_info->max_rx_queues = 8;
2246 		dev_info->max_tx_queues = 8;
2247 		break;
2248 
2249 	case e1000_i210:
2250 		dev_info->max_rx_queues = 4;
2251 		dev_info->max_tx_queues = 4;
2252 		dev_info->max_vmdq_pools = 0;
2253 		break;
2254 
2255 	case e1000_i211:
2256 		dev_info->max_rx_queues = 2;
2257 		dev_info->max_tx_queues = 2;
2258 		dev_info->max_vmdq_pools = 0;
2259 		break;
2260 
2261 	default:
2262 		/* Should not happen */
2263 		break;
2264 	}
2265 	dev_info->hash_key_size = IGB_HKEY_MAX_INDEX * sizeof(uint32_t);
2266 	dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
2267 	dev_info->flow_type_rss_offloads = IGB_RSS_OFFLOAD_ALL;
2268 
2269 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2270 		.rx_thresh = {
2271 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2272 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2273 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2274 		},
2275 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2276 		.rx_drop_en = 0,
2277 	};
2278 
2279 	dev_info->default_txconf = (struct rte_eth_txconf) {
2280 		.tx_thresh = {
2281 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2282 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2283 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2284 		},
2285 		.txq_flags = 0,
2286 	};
2287 
2288 	dev_info->rx_desc_lim = rx_desc_lim;
2289 	dev_info->tx_desc_lim = tx_desc_lim;
2290 
2291 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
2292 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
2293 			ETH_LINK_SPEED_1G;
2294 }
2295 
2296 static const uint32_t *
2297 eth_igb_supported_ptypes_get(struct rte_eth_dev *dev)
2298 {
2299 	static const uint32_t ptypes[] = {
2300 		/* refers to igb_rxd_pkt_info_to_pkt_type() */
2301 		RTE_PTYPE_L2_ETHER,
2302 		RTE_PTYPE_L3_IPV4,
2303 		RTE_PTYPE_L3_IPV4_EXT,
2304 		RTE_PTYPE_L3_IPV6,
2305 		RTE_PTYPE_L3_IPV6_EXT,
2306 		RTE_PTYPE_L4_TCP,
2307 		RTE_PTYPE_L4_UDP,
2308 		RTE_PTYPE_L4_SCTP,
2309 		RTE_PTYPE_TUNNEL_IP,
2310 		RTE_PTYPE_INNER_L3_IPV6,
2311 		RTE_PTYPE_INNER_L3_IPV6_EXT,
2312 		RTE_PTYPE_INNER_L4_TCP,
2313 		RTE_PTYPE_INNER_L4_UDP,
2314 		RTE_PTYPE_UNKNOWN
2315 	};
2316 
2317 	if (dev->rx_pkt_burst == eth_igb_recv_pkts ||
2318 	    dev->rx_pkt_burst == eth_igb_recv_scattered_pkts)
2319 		return ptypes;
2320 	return NULL;
2321 }
2322 
2323 static void
2324 eth_igbvf_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2325 {
2326 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2327 
2328 	dev_info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
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->rx_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP |
2333 				DEV_RX_OFFLOAD_IPV4_CKSUM |
2334 				DEV_RX_OFFLOAD_UDP_CKSUM  |
2335 				DEV_RX_OFFLOAD_TCP_CKSUM;
2336 	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
2337 				DEV_TX_OFFLOAD_IPV4_CKSUM  |
2338 				DEV_TX_OFFLOAD_UDP_CKSUM   |
2339 				DEV_TX_OFFLOAD_TCP_CKSUM   |
2340 				DEV_TX_OFFLOAD_SCTP_CKSUM  |
2341 				DEV_TX_OFFLOAD_TCP_TSO;
2342 	switch (hw->mac.type) {
2343 	case e1000_vfadapt:
2344 		dev_info->max_rx_queues = 2;
2345 		dev_info->max_tx_queues = 2;
2346 		break;
2347 	case e1000_vfadapt_i350:
2348 		dev_info->max_rx_queues = 1;
2349 		dev_info->max_tx_queues = 1;
2350 		break;
2351 	default:
2352 		/* Should not happen */
2353 		break;
2354 	}
2355 
2356 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2357 		.rx_thresh = {
2358 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2359 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2360 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2361 		},
2362 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2363 		.rx_drop_en = 0,
2364 	};
2365 
2366 	dev_info->default_txconf = (struct rte_eth_txconf) {
2367 		.tx_thresh = {
2368 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2369 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2370 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2371 		},
2372 		.txq_flags = 0,
2373 	};
2374 
2375 	dev_info->rx_desc_lim = rx_desc_lim;
2376 	dev_info->tx_desc_lim = tx_desc_lim;
2377 }
2378 
2379 /* return 0 means link status changed, -1 means not changed */
2380 static int
2381 eth_igb_link_update(struct rte_eth_dev *dev, int wait_to_complete)
2382 {
2383 	struct e1000_hw *hw =
2384 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2385 	struct rte_eth_link link, old;
2386 	int link_check, count;
2387 
2388 	link_check = 0;
2389 	hw->mac.get_link_status = 1;
2390 
2391 	/* possible wait-to-complete in up to 9 seconds */
2392 	for (count = 0; count < IGB_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
2393 		/* Read the real link status */
2394 		switch (hw->phy.media_type) {
2395 		case e1000_media_type_copper:
2396 			/* Do the work to read phy */
2397 			e1000_check_for_link(hw);
2398 			link_check = !hw->mac.get_link_status;
2399 			break;
2400 
2401 		case e1000_media_type_fiber:
2402 			e1000_check_for_link(hw);
2403 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
2404 				      E1000_STATUS_LU);
2405 			break;
2406 
2407 		case e1000_media_type_internal_serdes:
2408 			e1000_check_for_link(hw);
2409 			link_check = hw->mac.serdes_has_link;
2410 			break;
2411 
2412 		/* VF device is type_unknown */
2413 		case e1000_media_type_unknown:
2414 			eth_igbvf_link_update(hw);
2415 			link_check = !hw->mac.get_link_status;
2416 			break;
2417 
2418 		default:
2419 			break;
2420 		}
2421 		if (link_check || wait_to_complete == 0)
2422 			break;
2423 		rte_delay_ms(IGB_LINK_UPDATE_CHECK_INTERVAL);
2424 	}
2425 	memset(&link, 0, sizeof(link));
2426 	rte_igb_dev_atomic_read_link_status(dev, &link);
2427 	old = link;
2428 
2429 	/* Now we check if a transition has happened */
2430 	if (link_check) {
2431 		uint16_t duplex, speed;
2432 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
2433 		link.link_duplex = (duplex == FULL_DUPLEX) ?
2434 				ETH_LINK_FULL_DUPLEX :
2435 				ETH_LINK_HALF_DUPLEX;
2436 		link.link_speed = speed;
2437 		link.link_status = ETH_LINK_UP;
2438 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
2439 				ETH_LINK_SPEED_FIXED);
2440 	} else if (!link_check) {
2441 		link.link_speed = 0;
2442 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
2443 		link.link_status = ETH_LINK_DOWN;
2444 		link.link_autoneg = ETH_LINK_FIXED;
2445 	}
2446 	rte_igb_dev_atomic_write_link_status(dev, &link);
2447 
2448 	/* not changed */
2449 	if (old.link_status == link.link_status)
2450 		return -1;
2451 
2452 	/* changed */
2453 	return 0;
2454 }
2455 
2456 /*
2457  * igb_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
2458  * For ASF and Pass Through versions of f/w this means
2459  * that the driver is loaded.
2460  */
2461 static void
2462 igb_hw_control_acquire(struct e1000_hw *hw)
2463 {
2464 	uint32_t ctrl_ext;
2465 
2466 	/* Let firmware know the driver has taken over */
2467 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2468 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2469 }
2470 
2471 /*
2472  * igb_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
2473  * For ASF and Pass Through versions of f/w this means that the
2474  * driver is no longer loaded.
2475  */
2476 static void
2477 igb_hw_control_release(struct e1000_hw *hw)
2478 {
2479 	uint32_t ctrl_ext;
2480 
2481 	/* Let firmware taken over control of h/w */
2482 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2483 	E1000_WRITE_REG(hw, E1000_CTRL_EXT,
2484 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2485 }
2486 
2487 /*
2488  * Bit of a misnomer, what this really means is
2489  * to enable OS management of the system... aka
2490  * to disable special hardware management features.
2491  */
2492 static void
2493 igb_init_manageability(struct e1000_hw *hw)
2494 {
2495 	if (e1000_enable_mng_pass_thru(hw)) {
2496 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
2497 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2498 
2499 		/* disable hardware interception of ARP */
2500 		manc &= ~(E1000_MANC_ARP_EN);
2501 
2502 		/* enable receiving management packets to the host */
2503 		manc |= E1000_MANC_EN_MNG2HOST;
2504 		manc2h |= 1 << 5;  /* Mng Port 623 */
2505 		manc2h |= 1 << 6;  /* Mng Port 664 */
2506 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
2507 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2508 	}
2509 }
2510 
2511 static void
2512 igb_release_manageability(struct e1000_hw *hw)
2513 {
2514 	if (e1000_enable_mng_pass_thru(hw)) {
2515 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2516 
2517 		manc |= E1000_MANC_ARP_EN;
2518 		manc &= ~E1000_MANC_EN_MNG2HOST;
2519 
2520 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2521 	}
2522 }
2523 
2524 static void
2525 eth_igb_promiscuous_enable(struct rte_eth_dev *dev)
2526 {
2527 	struct e1000_hw *hw =
2528 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2529 	uint32_t rctl;
2530 
2531 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2532 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
2533 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2534 }
2535 
2536 static void
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 
2552 static void
2553 eth_igb_allmulticast_enable(struct rte_eth_dev *dev)
2554 {
2555 	struct e1000_hw *hw =
2556 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2557 	uint32_t rctl;
2558 
2559 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2560 	rctl |= E1000_RCTL_MPE;
2561 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2562 }
2563 
2564 static void
2565 eth_igb_allmulticast_disable(struct rte_eth_dev *dev)
2566 {
2567 	struct e1000_hw *hw =
2568 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2569 	uint32_t rctl;
2570 
2571 	if (dev->data->promiscuous == 1)
2572 		return; /* must remain in all_multicast mode */
2573 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2574 	rctl &= (~E1000_RCTL_MPE);
2575 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2576 }
2577 
2578 static int
2579 eth_igb_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2580 {
2581 	struct e1000_hw *hw =
2582 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2583 	struct e1000_vfta * shadow_vfta =
2584 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2585 	uint32_t vfta;
2586 	uint32_t vid_idx;
2587 	uint32_t vid_bit;
2588 
2589 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
2590 			      E1000_VFTA_ENTRY_MASK);
2591 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
2592 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
2593 	if (on)
2594 		vfta |= vid_bit;
2595 	else
2596 		vfta &= ~vid_bit;
2597 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
2598 
2599 	/* update local VFTA copy */
2600 	shadow_vfta->vfta[vid_idx] = vfta;
2601 
2602 	return 0;
2603 }
2604 
2605 static int
2606 eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
2607 		      enum rte_vlan_type vlan_type,
2608 		      uint16_t tpid)
2609 {
2610 	struct e1000_hw *hw =
2611 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2612 	uint32_t reg, qinq;
2613 
2614 	qinq = E1000_READ_REG(hw, E1000_CTRL_EXT);
2615 	qinq &= E1000_CTRL_EXT_EXT_VLAN;
2616 
2617 	/* only outer TPID of double VLAN can be configured*/
2618 	if (qinq && vlan_type == ETH_VLAN_TYPE_OUTER) {
2619 		reg = E1000_READ_REG(hw, E1000_VET);
2620 		reg = (reg & (~E1000_VET_VET_EXT)) |
2621 			((uint32_t)tpid << E1000_VET_VET_EXT_SHIFT);
2622 		E1000_WRITE_REG(hw, E1000_VET, reg);
2623 
2624 		return 0;
2625 	}
2626 
2627 	/* all other TPID values are read-only*/
2628 	PMD_DRV_LOG(ERR, "Not supported");
2629 
2630 	return -ENOTSUP;
2631 }
2632 
2633 static void
2634 igb_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2635 {
2636 	struct e1000_hw *hw =
2637 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2638 	uint32_t reg;
2639 
2640 	/* Filter Table Disable */
2641 	reg = E1000_READ_REG(hw, E1000_RCTL);
2642 	reg &= ~E1000_RCTL_CFIEN;
2643 	reg &= ~E1000_RCTL_VFE;
2644 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2645 }
2646 
2647 static void
2648 igb_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2649 {
2650 	struct e1000_hw *hw =
2651 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2652 	struct e1000_vfta * shadow_vfta =
2653 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2654 	uint32_t reg;
2655 	int i;
2656 
2657 	/* Filter Table Enable, CFI not used for packet acceptance */
2658 	reg = E1000_READ_REG(hw, E1000_RCTL);
2659 	reg &= ~E1000_RCTL_CFIEN;
2660 	reg |= E1000_RCTL_VFE;
2661 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2662 
2663 	/* restore VFTA table */
2664 	for (i = 0; i < IGB_VFTA_SIZE; i++)
2665 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
2666 }
2667 
2668 static void
2669 igb_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2670 {
2671 	struct e1000_hw *hw =
2672 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2673 	uint32_t reg;
2674 
2675 	/* VLAN Mode Disable */
2676 	reg = E1000_READ_REG(hw, E1000_CTRL);
2677 	reg &= ~E1000_CTRL_VME;
2678 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2679 }
2680 
2681 static void
2682 igb_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2683 {
2684 	struct e1000_hw *hw =
2685 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2686 	uint32_t reg;
2687 
2688 	/* VLAN Mode Enable */
2689 	reg = E1000_READ_REG(hw, E1000_CTRL);
2690 	reg |= E1000_CTRL_VME;
2691 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2692 }
2693 
2694 static void
2695 igb_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2696 {
2697 	struct e1000_hw *hw =
2698 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2699 	uint32_t reg;
2700 
2701 	/* CTRL_EXT: Extended VLAN */
2702 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2703 	reg &= ~E1000_CTRL_EXT_EXTEND_VLAN;
2704 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2705 
2706 	/* Update maximum packet length */
2707 	if (dev->data->dev_conf.rxmode.jumbo_frame == 1)
2708 		E1000_WRITE_REG(hw, E1000_RLPML,
2709 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2710 						VLAN_TAG_SIZE);
2711 }
2712 
2713 static void
2714 igb_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2715 {
2716 	struct e1000_hw *hw =
2717 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2718 	uint32_t reg;
2719 
2720 	/* CTRL_EXT: Extended VLAN */
2721 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2722 	reg |= E1000_CTRL_EXT_EXTEND_VLAN;
2723 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2724 
2725 	/* Update maximum packet length */
2726 	if (dev->data->dev_conf.rxmode.jumbo_frame == 1)
2727 		E1000_WRITE_REG(hw, E1000_RLPML,
2728 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2729 						2 * VLAN_TAG_SIZE);
2730 }
2731 
2732 static int
2733 eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2734 {
2735 	if(mask & ETH_VLAN_STRIP_MASK){
2736 		if (dev->data->dev_conf.rxmode.hw_vlan_strip)
2737 			igb_vlan_hw_strip_enable(dev);
2738 		else
2739 			igb_vlan_hw_strip_disable(dev);
2740 	}
2741 
2742 	if(mask & ETH_VLAN_FILTER_MASK){
2743 		if (dev->data->dev_conf.rxmode.hw_vlan_filter)
2744 			igb_vlan_hw_filter_enable(dev);
2745 		else
2746 			igb_vlan_hw_filter_disable(dev);
2747 	}
2748 
2749 	if(mask & ETH_VLAN_EXTEND_MASK){
2750 		if (dev->data->dev_conf.rxmode.hw_vlan_extend)
2751 			igb_vlan_hw_extend_enable(dev);
2752 		else
2753 			igb_vlan_hw_extend_disable(dev);
2754 	}
2755 
2756 	return 0;
2757 }
2758 
2759 
2760 /**
2761  * It enables the interrupt mask and then enable the interrupt.
2762  *
2763  * @param dev
2764  *  Pointer to struct rte_eth_dev.
2765  * @param on
2766  *  Enable or Disable
2767  *
2768  * @return
2769  *  - On success, zero.
2770  *  - On failure, a negative value.
2771  */
2772 static int
2773 eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
2774 {
2775 	struct e1000_interrupt *intr =
2776 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2777 
2778 	if (on)
2779 		intr->mask |= E1000_ICR_LSC;
2780 	else
2781 		intr->mask &= ~E1000_ICR_LSC;
2782 
2783 	return 0;
2784 }
2785 
2786 /* It clears the interrupt causes and enables the interrupt.
2787  * It will be called once only during nic initialized.
2788  *
2789  * @param dev
2790  *  Pointer to struct rte_eth_dev.
2791  *
2792  * @return
2793  *  - On success, zero.
2794  *  - On failure, a negative value.
2795  */
2796 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev)
2797 {
2798 	uint32_t mask, regval;
2799 	struct e1000_hw *hw =
2800 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2801 	struct rte_eth_dev_info dev_info;
2802 
2803 	memset(&dev_info, 0, sizeof(dev_info));
2804 	eth_igb_infos_get(dev, &dev_info);
2805 
2806 	mask = 0xFFFFFFFF >> (32 - dev_info.max_rx_queues);
2807 	regval = E1000_READ_REG(hw, E1000_EIMS);
2808 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
2809 
2810 	return 0;
2811 }
2812 
2813 /*
2814  * It reads ICR and gets interrupt causes, check it and set a bit flag
2815  * to update link status.
2816  *
2817  * @param dev
2818  *  Pointer to struct rte_eth_dev.
2819  *
2820  * @return
2821  *  - On success, zero.
2822  *  - On failure, a negative value.
2823  */
2824 static int
2825 eth_igb_interrupt_get_status(struct rte_eth_dev *dev)
2826 {
2827 	uint32_t icr;
2828 	struct e1000_hw *hw =
2829 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2830 	struct e1000_interrupt *intr =
2831 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2832 
2833 	igb_intr_disable(hw);
2834 
2835 	/* read-on-clear nic registers here */
2836 	icr = E1000_READ_REG(hw, E1000_ICR);
2837 
2838 	intr->flags = 0;
2839 	if (icr & E1000_ICR_LSC) {
2840 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
2841 	}
2842 
2843 	if (icr & E1000_ICR_VMMB)
2844 		intr->flags |= E1000_FLAG_MAILBOX;
2845 
2846 	return 0;
2847 }
2848 
2849 /*
2850  * It executes link_update after knowing an interrupt is prsent.
2851  *
2852  * @param dev
2853  *  Pointer to struct rte_eth_dev.
2854  *
2855  * @return
2856  *  - On success, zero.
2857  *  - On failure, a negative value.
2858  */
2859 static int
2860 eth_igb_interrupt_action(struct rte_eth_dev *dev,
2861 			 struct rte_intr_handle *intr_handle)
2862 {
2863 	struct e1000_hw *hw =
2864 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2865 	struct e1000_interrupt *intr =
2866 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2867 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2868 	struct rte_eth_link link;
2869 	int ret;
2870 
2871 	if (intr->flags & E1000_FLAG_MAILBOX) {
2872 		igb_pf_mbx_process(dev);
2873 		intr->flags &= ~E1000_FLAG_MAILBOX;
2874 	}
2875 
2876 	igb_intr_enable(dev);
2877 	rte_intr_enable(intr_handle);
2878 
2879 	if (intr->flags & E1000_FLAG_NEED_LINK_UPDATE) {
2880 		intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
2881 
2882 		/* set get_link_status to check register later */
2883 		hw->mac.get_link_status = 1;
2884 		ret = eth_igb_link_update(dev, 0);
2885 
2886 		/* check if link has changed */
2887 		if (ret < 0)
2888 			return 0;
2889 
2890 		memset(&link, 0, sizeof(link));
2891 		rte_igb_dev_atomic_read_link_status(dev, &link);
2892 		if (link.link_status) {
2893 			PMD_INIT_LOG(INFO,
2894 				     " Port %d: Link Up - speed %u Mbps - %s",
2895 				     dev->data->port_id,
2896 				     (unsigned)link.link_speed,
2897 				     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
2898 				     "full-duplex" : "half-duplex");
2899 		} else {
2900 			PMD_INIT_LOG(INFO, " Port %d: Link Down",
2901 				     dev->data->port_id);
2902 		}
2903 
2904 		PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
2905 			     pci_dev->addr.domain,
2906 			     pci_dev->addr.bus,
2907 			     pci_dev->addr.devid,
2908 			     pci_dev->addr.function);
2909 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC,
2910 					      NULL);
2911 	}
2912 
2913 	return 0;
2914 }
2915 
2916 /**
2917  * Interrupt handler which shall be registered at first.
2918  *
2919  * @param handle
2920  *  Pointer to interrupt handle.
2921  * @param param
2922  *  The address of parameter (struct rte_eth_dev *) regsitered before.
2923  *
2924  * @return
2925  *  void
2926  */
2927 static void
2928 eth_igb_interrupt_handler(void *param)
2929 {
2930 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2931 
2932 	eth_igb_interrupt_get_status(dev);
2933 	eth_igb_interrupt_action(dev, dev->intr_handle);
2934 }
2935 
2936 static int
2937 eth_igbvf_interrupt_get_status(struct rte_eth_dev *dev)
2938 {
2939 	uint32_t eicr;
2940 	struct e1000_hw *hw =
2941 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2942 	struct e1000_interrupt *intr =
2943 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2944 
2945 	igbvf_intr_disable(hw);
2946 
2947 	/* read-on-clear nic registers here */
2948 	eicr = E1000_READ_REG(hw, E1000_EICR);
2949 	intr->flags = 0;
2950 
2951 	if (eicr == E1000_VTIVAR_MISC_MAILBOX)
2952 		intr->flags |= E1000_FLAG_MAILBOX;
2953 
2954 	return 0;
2955 }
2956 
2957 void igbvf_mbx_process(struct rte_eth_dev *dev)
2958 {
2959 	struct e1000_hw *hw =
2960 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2961 	struct e1000_mbx_info *mbx = &hw->mbx;
2962 	u32 in_msg = 0;
2963 
2964 	/* peek the message first */
2965 	in_msg = E1000_READ_REG(hw, E1000_VMBMEM(0));
2966 
2967 	/* PF reset VF event */
2968 	if (in_msg == E1000_PF_CONTROL_MSG) {
2969 		/* dummy mbx read to ack pf */
2970 		if (mbx->ops.read(hw, &in_msg, 1, 0))
2971 			return;
2972 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
2973 					      NULL);
2974 	}
2975 }
2976 
2977 static int
2978 eth_igbvf_interrupt_action(struct rte_eth_dev *dev, struct rte_intr_handle *intr_handle)
2979 {
2980 	struct e1000_interrupt *intr =
2981 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2982 
2983 	if (intr->flags & E1000_FLAG_MAILBOX) {
2984 		igbvf_mbx_process(dev);
2985 		intr->flags &= ~E1000_FLAG_MAILBOX;
2986 	}
2987 
2988 	igbvf_intr_enable(dev);
2989 	rte_intr_enable(intr_handle);
2990 
2991 	return 0;
2992 }
2993 
2994 static void
2995 eth_igbvf_interrupt_handler(void *param)
2996 {
2997 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2998 
2999 	eth_igbvf_interrupt_get_status(dev);
3000 	eth_igbvf_interrupt_action(dev, dev->intr_handle);
3001 }
3002 
3003 static int
3004 eth_igb_led_on(struct rte_eth_dev *dev)
3005 {
3006 	struct e1000_hw *hw;
3007 
3008 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3009 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3010 }
3011 
3012 static int
3013 eth_igb_led_off(struct rte_eth_dev *dev)
3014 {
3015 	struct e1000_hw *hw;
3016 
3017 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3018 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3019 }
3020 
3021 static int
3022 eth_igb_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3023 {
3024 	struct e1000_hw *hw;
3025 	uint32_t ctrl;
3026 	int tx_pause;
3027 	int rx_pause;
3028 
3029 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3030 	fc_conf->pause_time = hw->fc.pause_time;
3031 	fc_conf->high_water = hw->fc.high_water;
3032 	fc_conf->low_water = hw->fc.low_water;
3033 	fc_conf->send_xon = hw->fc.send_xon;
3034 	fc_conf->autoneg = hw->mac.autoneg;
3035 
3036 	/*
3037 	 * Return rx_pause and tx_pause status according to actual setting of
3038 	 * the TFCE and RFCE bits in the CTRL register.
3039 	 */
3040 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
3041 	if (ctrl & E1000_CTRL_TFCE)
3042 		tx_pause = 1;
3043 	else
3044 		tx_pause = 0;
3045 
3046 	if (ctrl & E1000_CTRL_RFCE)
3047 		rx_pause = 1;
3048 	else
3049 		rx_pause = 0;
3050 
3051 	if (rx_pause && tx_pause)
3052 		fc_conf->mode = RTE_FC_FULL;
3053 	else if (rx_pause)
3054 		fc_conf->mode = RTE_FC_RX_PAUSE;
3055 	else if (tx_pause)
3056 		fc_conf->mode = RTE_FC_TX_PAUSE;
3057 	else
3058 		fc_conf->mode = RTE_FC_NONE;
3059 
3060 	return 0;
3061 }
3062 
3063 static int
3064 eth_igb_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3065 {
3066 	struct e1000_hw *hw;
3067 	int err;
3068 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
3069 		e1000_fc_none,
3070 		e1000_fc_rx_pause,
3071 		e1000_fc_tx_pause,
3072 		e1000_fc_full
3073 	};
3074 	uint32_t rx_buf_size;
3075 	uint32_t max_high_water;
3076 	uint32_t rctl;
3077 
3078 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3079 	if (fc_conf->autoneg != hw->mac.autoneg)
3080 		return -ENOTSUP;
3081 	rx_buf_size = igb_get_rx_buffer_size(hw);
3082 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3083 
3084 	/* At least reserve one Ethernet frame for watermark */
3085 	max_high_water = rx_buf_size - ETHER_MAX_LEN;
3086 	if ((fc_conf->high_water > max_high_water) ||
3087 	    (fc_conf->high_water < fc_conf->low_water)) {
3088 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
3089 		PMD_INIT_LOG(ERR, "high water must <=  0x%x", max_high_water);
3090 		return -EINVAL;
3091 	}
3092 
3093 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
3094 	hw->fc.pause_time     = fc_conf->pause_time;
3095 	hw->fc.high_water     = fc_conf->high_water;
3096 	hw->fc.low_water      = fc_conf->low_water;
3097 	hw->fc.send_xon	      = fc_conf->send_xon;
3098 
3099 	err = e1000_setup_link_generic(hw);
3100 	if (err == E1000_SUCCESS) {
3101 
3102 		/* check if we want to forward MAC frames - driver doesn't have native
3103 		 * capability to do that, so we'll write the registers ourselves */
3104 
3105 		rctl = E1000_READ_REG(hw, E1000_RCTL);
3106 
3107 		/* set or clear MFLCN.PMCF bit depending on configuration */
3108 		if (fc_conf->mac_ctrl_frame_fwd != 0)
3109 			rctl |= E1000_RCTL_PMCF;
3110 		else
3111 			rctl &= ~E1000_RCTL_PMCF;
3112 
3113 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3114 		E1000_WRITE_FLUSH(hw);
3115 
3116 		return 0;
3117 	}
3118 
3119 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
3120 	return -EIO;
3121 }
3122 
3123 #define E1000_RAH_POOLSEL_SHIFT      (18)
3124 static int
3125 eth_igb_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
3126 		uint32_t index, uint32_t pool)
3127 {
3128 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3129 	uint32_t rah;
3130 
3131 	e1000_rar_set(hw, mac_addr->addr_bytes, index);
3132 	rah = E1000_READ_REG(hw, E1000_RAH(index));
3133 	rah |= (0x1 << (E1000_RAH_POOLSEL_SHIFT + pool));
3134 	E1000_WRITE_REG(hw, E1000_RAH(index), rah);
3135 	return 0;
3136 }
3137 
3138 static void
3139 eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index)
3140 {
3141 	uint8_t addr[ETHER_ADDR_LEN];
3142 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3143 
3144 	memset(addr, 0, sizeof(addr));
3145 
3146 	e1000_rar_set(hw, addr, index);
3147 }
3148 
3149 static void
3150 eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
3151 				struct ether_addr *addr)
3152 {
3153 	eth_igb_rar_clear(dev, 0);
3154 
3155 	eth_igb_rar_set(dev, (void *)addr, 0, 0);
3156 }
3157 /*
3158  * Virtual Function operations
3159  */
3160 static void
3161 igbvf_intr_disable(struct e1000_hw *hw)
3162 {
3163 	PMD_INIT_FUNC_TRACE();
3164 
3165 	/* Clear interrupt mask to stop from interrupts being generated */
3166 	E1000_WRITE_REG(hw, E1000_EIMC, 0xFFFF);
3167 
3168 	E1000_WRITE_FLUSH(hw);
3169 }
3170 
3171 static void
3172 igbvf_stop_adapter(struct rte_eth_dev *dev)
3173 {
3174 	u32 reg_val;
3175 	u16 i;
3176 	struct rte_eth_dev_info dev_info;
3177 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3178 
3179 	memset(&dev_info, 0, sizeof(dev_info));
3180 	eth_igbvf_infos_get(dev, &dev_info);
3181 
3182 	/* Clear interrupt mask to stop from interrupts being generated */
3183 	igbvf_intr_disable(hw);
3184 
3185 	/* Clear any pending interrupts, flush previous writes */
3186 	E1000_READ_REG(hw, E1000_EICR);
3187 
3188 	/* Disable the transmit unit.  Each queue must be disabled. */
3189 	for (i = 0; i < dev_info.max_tx_queues; i++)
3190 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), E1000_TXDCTL_SWFLSH);
3191 
3192 	/* Disable the receive unit by stopping each queue */
3193 	for (i = 0; i < dev_info.max_rx_queues; i++) {
3194 		reg_val = E1000_READ_REG(hw, E1000_RXDCTL(i));
3195 		reg_val &= ~E1000_RXDCTL_QUEUE_ENABLE;
3196 		E1000_WRITE_REG(hw, E1000_RXDCTL(i), reg_val);
3197 		while (E1000_READ_REG(hw, E1000_RXDCTL(i)) & E1000_RXDCTL_QUEUE_ENABLE)
3198 			;
3199 	}
3200 
3201 	/* flush all queues disables */
3202 	E1000_WRITE_FLUSH(hw);
3203 	msec_delay(2);
3204 }
3205 
3206 static int eth_igbvf_link_update(struct e1000_hw *hw)
3207 {
3208 	struct e1000_mbx_info *mbx = &hw->mbx;
3209 	struct e1000_mac_info *mac = &hw->mac;
3210 	int ret_val = E1000_SUCCESS;
3211 
3212 	PMD_INIT_LOG(DEBUG, "e1000_check_for_link_vf");
3213 
3214 	/*
3215 	 * We only want to run this if there has been a rst asserted.
3216 	 * in this case that could mean a link change, device reset,
3217 	 * or a virtual function reset
3218 	 */
3219 
3220 	/* If we were hit with a reset or timeout drop the link */
3221 	if (!e1000_check_for_rst(hw, 0) || !mbx->timeout)
3222 		mac->get_link_status = TRUE;
3223 
3224 	if (!mac->get_link_status)
3225 		goto out;
3226 
3227 	/* if link status is down no point in checking to see if pf is up */
3228 	if (!(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU))
3229 		goto out;
3230 
3231 	/* if we passed all the tests above then the link is up and we no
3232 	 * longer need to check for link */
3233 	mac->get_link_status = FALSE;
3234 
3235 out:
3236 	return ret_val;
3237 }
3238 
3239 
3240 static int
3241 igbvf_dev_configure(struct rte_eth_dev *dev)
3242 {
3243 	struct rte_eth_conf* conf = &dev->data->dev_conf;
3244 
3245 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
3246 		     dev->data->port_id);
3247 
3248 	/*
3249 	 * VF has no ability to enable/disable HW CRC
3250 	 * Keep the persistent behavior the same as Host PF
3251 	 */
3252 #ifndef RTE_LIBRTE_E1000_PF_DISABLE_STRIP_CRC
3253 	if (!conf->rxmode.hw_strip_crc) {
3254 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
3255 		conf->rxmode.hw_strip_crc = 1;
3256 	}
3257 #else
3258 	if (conf->rxmode.hw_strip_crc) {
3259 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
3260 		conf->rxmode.hw_strip_crc = 0;
3261 	}
3262 #endif
3263 
3264 	return 0;
3265 }
3266 
3267 static int
3268 igbvf_dev_start(struct rte_eth_dev *dev)
3269 {
3270 	struct e1000_hw *hw =
3271 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3272 	struct e1000_adapter *adapter =
3273 		E1000_DEV_PRIVATE(dev->data->dev_private);
3274 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3275 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3276 	int ret;
3277 	uint32_t intr_vector = 0;
3278 
3279 	PMD_INIT_FUNC_TRACE();
3280 
3281 	hw->mac.ops.reset_hw(hw);
3282 	adapter->stopped = 0;
3283 
3284 	/* Set all vfta */
3285 	igbvf_set_vfta_all(dev,1);
3286 
3287 	eth_igbvf_tx_init(dev);
3288 
3289 	/* This can fail when allocating mbufs for descriptor rings */
3290 	ret = eth_igbvf_rx_init(dev);
3291 	if (ret) {
3292 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
3293 		igb_dev_clear_queues(dev);
3294 		return ret;
3295 	}
3296 
3297 	/* check and configure queue intr-vector mapping */
3298 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
3299 		intr_vector = dev->data->nb_rx_queues;
3300 		ret = rte_intr_efd_enable(intr_handle, intr_vector);
3301 		if (ret)
3302 			return ret;
3303 	}
3304 
3305 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
3306 		intr_handle->intr_vec =
3307 			rte_zmalloc("intr_vec",
3308 				    dev->data->nb_rx_queues * sizeof(int), 0);
3309 		if (!intr_handle->intr_vec) {
3310 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
3311 				     " intr_vec", dev->data->nb_rx_queues);
3312 			return -ENOMEM;
3313 		}
3314 	}
3315 
3316 	eth_igbvf_configure_msix_intr(dev);
3317 
3318 	/* enable uio/vfio intr/eventfd mapping */
3319 	rte_intr_enable(intr_handle);
3320 
3321 	/* resume enabled intr since hw reset */
3322 	igbvf_intr_enable(dev);
3323 
3324 	return 0;
3325 }
3326 
3327 static void
3328 igbvf_dev_stop(struct rte_eth_dev *dev)
3329 {
3330 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3331 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3332 
3333 	PMD_INIT_FUNC_TRACE();
3334 
3335 	igbvf_stop_adapter(dev);
3336 
3337 	/*
3338 	  * Clear what we set, but we still keep shadow_vfta to
3339 	  * restore after device starts
3340 	  */
3341 	igbvf_set_vfta_all(dev,0);
3342 
3343 	igb_dev_clear_queues(dev);
3344 
3345 	/* disable intr eventfd mapping */
3346 	rte_intr_disable(intr_handle);
3347 
3348 	/* Clean datapath event and queue/vec mapping */
3349 	rte_intr_efd_disable(intr_handle);
3350 	if (intr_handle->intr_vec) {
3351 		rte_free(intr_handle->intr_vec);
3352 		intr_handle->intr_vec = NULL;
3353 	}
3354 }
3355 
3356 static void
3357 igbvf_dev_close(struct rte_eth_dev *dev)
3358 {
3359 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3360 	struct e1000_adapter *adapter =
3361 		E1000_DEV_PRIVATE(dev->data->dev_private);
3362 	struct ether_addr addr;
3363 
3364 	PMD_INIT_FUNC_TRACE();
3365 
3366 	e1000_reset_hw(hw);
3367 
3368 	igbvf_dev_stop(dev);
3369 	adapter->stopped = 1;
3370 	igb_dev_free_queues(dev);
3371 
3372 	/**
3373 	 * reprogram the RAR with a zero mac address,
3374 	 * to ensure that the VF traffic goes to the PF
3375 	 * after stop, close and detach of the VF.
3376 	 **/
3377 
3378 	memset(&addr, 0, sizeof(addr));
3379 	igbvf_default_mac_addr_set(dev, &addr);
3380 }
3381 
3382 static void
3383 igbvf_promiscuous_enable(struct rte_eth_dev *dev)
3384 {
3385 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3386 
3387 	/* Set both unicast and multicast promisc */
3388 	e1000_promisc_set_vf(hw, e1000_promisc_enabled);
3389 }
3390 
3391 static void
3392 igbvf_promiscuous_disable(struct rte_eth_dev *dev)
3393 {
3394 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3395 
3396 	/* If in allmulticast mode leave multicast promisc */
3397 	if (dev->data->all_multicast == 1)
3398 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3399 	else
3400 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3401 }
3402 
3403 static void
3404 igbvf_allmulticast_enable(struct rte_eth_dev *dev)
3405 {
3406 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3407 
3408 	/* In promiscuous mode multicast promisc already set */
3409 	if (dev->data->promiscuous == 0)
3410 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3411 }
3412 
3413 static void
3414 igbvf_allmulticast_disable(struct rte_eth_dev *dev)
3415 {
3416 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3417 
3418 	/* In promiscuous mode leave multicast promisc enabled */
3419 	if (dev->data->promiscuous == 0)
3420 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3421 }
3422 
3423 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on)
3424 {
3425 	struct e1000_mbx_info *mbx = &hw->mbx;
3426 	uint32_t msgbuf[2];
3427 	s32 err;
3428 
3429 	/* After set vlan, vlan strip will also be enabled in igb driver*/
3430 	msgbuf[0] = E1000_VF_SET_VLAN;
3431 	msgbuf[1] = vid;
3432 	/* Setting the 8 bit field MSG INFO to TRUE indicates "add" */
3433 	if (on)
3434 		msgbuf[0] |= E1000_VF_SET_VLAN_ADD;
3435 
3436 	err = mbx->ops.write_posted(hw, msgbuf, 2, 0);
3437 	if (err)
3438 		goto mbx_err;
3439 
3440 	err = mbx->ops.read_posted(hw, msgbuf, 2, 0);
3441 	if (err)
3442 		goto mbx_err;
3443 
3444 	msgbuf[0] &= ~E1000_VT_MSGTYPE_CTS;
3445 	if (msgbuf[0] == (E1000_VF_SET_VLAN | E1000_VT_MSGTYPE_NACK))
3446 		err = -EINVAL;
3447 
3448 mbx_err:
3449 	return err;
3450 }
3451 
3452 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on)
3453 {
3454 	struct e1000_hw *hw =
3455 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3456 	struct e1000_vfta * shadow_vfta =
3457 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3458 	int i = 0, j = 0, vfta = 0, mask = 1;
3459 
3460 	for (i = 0; i < IGB_VFTA_SIZE; i++){
3461 		vfta = shadow_vfta->vfta[i];
3462 		if(vfta){
3463 			mask = 1;
3464 			for (j = 0; j < 32; j++){
3465 				if(vfta & mask)
3466 					igbvf_set_vfta(hw,
3467 						(uint16_t)((i<<5)+j), on);
3468 				mask<<=1;
3469 			}
3470 		}
3471 	}
3472 
3473 }
3474 
3475 static int
3476 igbvf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
3477 {
3478 	struct e1000_hw *hw =
3479 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3480 	struct e1000_vfta * shadow_vfta =
3481 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3482 	uint32_t vid_idx = 0;
3483 	uint32_t vid_bit = 0;
3484 	int ret = 0;
3485 
3486 	PMD_INIT_FUNC_TRACE();
3487 
3488 	/*vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf*/
3489 	ret = igbvf_set_vfta(hw, vlan_id, !!on);
3490 	if(ret){
3491 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
3492 		return ret;
3493 	}
3494 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
3495 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
3496 
3497 	/*Save what we set and retore it after device reset*/
3498 	if (on)
3499 		shadow_vfta->vfta[vid_idx] |= vid_bit;
3500 	else
3501 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
3502 
3503 	return 0;
3504 }
3505 
3506 static void
3507 igbvf_default_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *addr)
3508 {
3509 	struct e1000_hw *hw =
3510 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3511 
3512 	/* index is not used by rar_set() */
3513 	hw->mac.ops.rar_set(hw, (void *)addr, 0);
3514 }
3515 
3516 
3517 static int
3518 eth_igb_rss_reta_update(struct rte_eth_dev *dev,
3519 			struct rte_eth_rss_reta_entry64 *reta_conf,
3520 			uint16_t reta_size)
3521 {
3522 	uint8_t i, j, mask;
3523 	uint32_t reta, r;
3524 	uint16_t idx, shift;
3525 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3526 
3527 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3528 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3529 			"(%d) doesn't match the number hardware can supported "
3530 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3531 		return -EINVAL;
3532 	}
3533 
3534 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3535 		idx = i / RTE_RETA_GROUP_SIZE;
3536 		shift = i % RTE_RETA_GROUP_SIZE;
3537 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3538 						IGB_4_BIT_MASK);
3539 		if (!mask)
3540 			continue;
3541 		if (mask == IGB_4_BIT_MASK)
3542 			r = 0;
3543 		else
3544 			r = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3545 		for (j = 0, reta = 0; j < IGB_4_BIT_WIDTH; j++) {
3546 			if (mask & (0x1 << j))
3547 				reta |= reta_conf[idx].reta[shift + j] <<
3548 							(CHAR_BIT * j);
3549 			else
3550 				reta |= r & (IGB_8_BIT_MASK << (CHAR_BIT * j));
3551 		}
3552 		E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
3553 	}
3554 
3555 	return 0;
3556 }
3557 
3558 static int
3559 eth_igb_rss_reta_query(struct rte_eth_dev *dev,
3560 		       struct rte_eth_rss_reta_entry64 *reta_conf,
3561 		       uint16_t reta_size)
3562 {
3563 	uint8_t i, j, mask;
3564 	uint32_t reta;
3565 	uint16_t idx, shift;
3566 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3567 
3568 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3569 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3570 			"(%d) doesn't match the number hardware can supported "
3571 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3572 		return -EINVAL;
3573 	}
3574 
3575 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3576 		idx = i / RTE_RETA_GROUP_SIZE;
3577 		shift = i % RTE_RETA_GROUP_SIZE;
3578 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3579 						IGB_4_BIT_MASK);
3580 		if (!mask)
3581 			continue;
3582 		reta = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3583 		for (j = 0; j < IGB_4_BIT_WIDTH; j++) {
3584 			if (mask & (0x1 << j))
3585 				reta_conf[idx].reta[shift + j] =
3586 					((reta >> (CHAR_BIT * j)) &
3587 						IGB_8_BIT_MASK);
3588 		}
3589 	}
3590 
3591 	return 0;
3592 }
3593 
3594 int
3595 eth_igb_syn_filter_set(struct rte_eth_dev *dev,
3596 			struct rte_eth_syn_filter *filter,
3597 			bool add)
3598 {
3599 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3600 	struct e1000_filter_info *filter_info =
3601 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3602 	uint32_t synqf, rfctl;
3603 
3604 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3605 		return -EINVAL;
3606 
3607 	synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3608 
3609 	if (add) {
3610 		if (synqf & E1000_SYN_FILTER_ENABLE)
3611 			return -EINVAL;
3612 
3613 		synqf = (uint32_t)(((filter->queue << E1000_SYN_FILTER_QUEUE_SHIFT) &
3614 			E1000_SYN_FILTER_QUEUE) | E1000_SYN_FILTER_ENABLE);
3615 
3616 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3617 		if (filter->hig_pri)
3618 			rfctl |= E1000_RFCTL_SYNQFP;
3619 		else
3620 			rfctl &= ~E1000_RFCTL_SYNQFP;
3621 
3622 		E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3623 	} else {
3624 		if (!(synqf & E1000_SYN_FILTER_ENABLE))
3625 			return -ENOENT;
3626 		synqf = 0;
3627 	}
3628 
3629 	filter_info->syn_info = synqf;
3630 	E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
3631 	E1000_WRITE_FLUSH(hw);
3632 	return 0;
3633 }
3634 
3635 static int
3636 eth_igb_syn_filter_get(struct rte_eth_dev *dev,
3637 			struct rte_eth_syn_filter *filter)
3638 {
3639 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3640 	uint32_t synqf, rfctl;
3641 
3642 	synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3643 	if (synqf & E1000_SYN_FILTER_ENABLE) {
3644 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3645 		filter->hig_pri = (rfctl & E1000_RFCTL_SYNQFP) ? 1 : 0;
3646 		filter->queue = (uint8_t)((synqf & E1000_SYN_FILTER_QUEUE) >>
3647 				E1000_SYN_FILTER_QUEUE_SHIFT);
3648 		return 0;
3649 	}
3650 
3651 	return -ENOENT;
3652 }
3653 
3654 static int
3655 eth_igb_syn_filter_handle(struct rte_eth_dev *dev,
3656 			enum rte_filter_op filter_op,
3657 			void *arg)
3658 {
3659 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3660 	int ret;
3661 
3662 	MAC_TYPE_FILTER_SUP(hw->mac.type);
3663 
3664 	if (filter_op == RTE_ETH_FILTER_NOP)
3665 		return 0;
3666 
3667 	if (arg == NULL) {
3668 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
3669 			    filter_op);
3670 		return -EINVAL;
3671 	}
3672 
3673 	switch (filter_op) {
3674 	case RTE_ETH_FILTER_ADD:
3675 		ret = eth_igb_syn_filter_set(dev,
3676 				(struct rte_eth_syn_filter *)arg,
3677 				TRUE);
3678 		break;
3679 	case RTE_ETH_FILTER_DELETE:
3680 		ret = eth_igb_syn_filter_set(dev,
3681 				(struct rte_eth_syn_filter *)arg,
3682 				FALSE);
3683 		break;
3684 	case RTE_ETH_FILTER_GET:
3685 		ret = eth_igb_syn_filter_get(dev,
3686 				(struct rte_eth_syn_filter *)arg);
3687 		break;
3688 	default:
3689 		PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
3690 		ret = -EINVAL;
3691 		break;
3692 	}
3693 
3694 	return ret;
3695 }
3696 
3697 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_2tuple_filter_info*/
3698 static inline int
3699 ntuple_filter_to_2tuple(struct rte_eth_ntuple_filter *filter,
3700 			struct e1000_2tuple_filter_info *filter_info)
3701 {
3702 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3703 		return -EINVAL;
3704 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
3705 		return -EINVAL;  /* filter index is out of range. */
3706 	if (filter->tcp_flags > TCP_FLAG_ALL)
3707 		return -EINVAL;  /* flags is invalid. */
3708 
3709 	switch (filter->dst_port_mask) {
3710 	case UINT16_MAX:
3711 		filter_info->dst_port_mask = 0;
3712 		filter_info->dst_port = filter->dst_port;
3713 		break;
3714 	case 0:
3715 		filter_info->dst_port_mask = 1;
3716 		break;
3717 	default:
3718 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
3719 		return -EINVAL;
3720 	}
3721 
3722 	switch (filter->proto_mask) {
3723 	case UINT8_MAX:
3724 		filter_info->proto_mask = 0;
3725 		filter_info->proto = filter->proto;
3726 		break;
3727 	case 0:
3728 		filter_info->proto_mask = 1;
3729 		break;
3730 	default:
3731 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
3732 		return -EINVAL;
3733 	}
3734 
3735 	filter_info->priority = (uint8_t)filter->priority;
3736 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
3737 		filter_info->tcp_flags = filter->tcp_flags;
3738 	else
3739 		filter_info->tcp_flags = 0;
3740 
3741 	return 0;
3742 }
3743 
3744 static inline struct e1000_2tuple_filter *
3745 igb_2tuple_filter_lookup(struct e1000_2tuple_filter_list *filter_list,
3746 			struct e1000_2tuple_filter_info *key)
3747 {
3748 	struct e1000_2tuple_filter *it;
3749 
3750 	TAILQ_FOREACH(it, filter_list, entries) {
3751 		if (memcmp(key, &it->filter_info,
3752 			sizeof(struct e1000_2tuple_filter_info)) == 0) {
3753 			return it;
3754 		}
3755 	}
3756 	return NULL;
3757 }
3758 
3759 /* inject a igb 2tuple filter to HW */
3760 static inline void
3761 igb_inject_2uple_filter(struct rte_eth_dev *dev,
3762 			   struct e1000_2tuple_filter *filter)
3763 {
3764 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3765 	uint32_t ttqf = E1000_TTQF_DISABLE_MASK;
3766 	uint32_t imir, imir_ext = E1000_IMIREXT_SIZE_BP;
3767 	int i;
3768 
3769 	i = filter->index;
3770 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
3771 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
3772 		imir |= E1000_IMIR_PORT_BP;
3773 	else
3774 		imir &= ~E1000_IMIR_PORT_BP;
3775 
3776 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
3777 
3778 	ttqf |= E1000_TTQF_QUEUE_ENABLE;
3779 	ttqf |= (uint32_t)(filter->queue << E1000_TTQF_QUEUE_SHIFT);
3780 	ttqf |= (uint32_t)(filter->filter_info.proto &
3781 						E1000_TTQF_PROTOCOL_MASK);
3782 	if (filter->filter_info.proto_mask == 0)
3783 		ttqf &= ~E1000_TTQF_MASK_ENABLE;
3784 
3785 	/* tcp flags bits setting. */
3786 	if (filter->filter_info.tcp_flags & TCP_FLAG_ALL) {
3787 		if (filter->filter_info.tcp_flags & TCP_URG_FLAG)
3788 			imir_ext |= E1000_IMIREXT_CTRL_URG;
3789 		if (filter->filter_info.tcp_flags & TCP_ACK_FLAG)
3790 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
3791 		if (filter->filter_info.tcp_flags & TCP_PSH_FLAG)
3792 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
3793 		if (filter->filter_info.tcp_flags & TCP_RST_FLAG)
3794 			imir_ext |= E1000_IMIREXT_CTRL_RST;
3795 		if (filter->filter_info.tcp_flags & TCP_SYN_FLAG)
3796 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
3797 		if (filter->filter_info.tcp_flags & TCP_FIN_FLAG)
3798 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
3799 	} else {
3800 		imir_ext |= E1000_IMIREXT_CTRL_BP;
3801 	}
3802 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
3803 	E1000_WRITE_REG(hw, E1000_TTQF(i), ttqf);
3804 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
3805 }
3806 
3807 /*
3808  * igb_add_2tuple_filter - add a 2tuple filter
3809  *
3810  * @param
3811  * dev: Pointer to struct rte_eth_dev.
3812  * ntuple_filter: ponter to the filter that will be added.
3813  *
3814  * @return
3815  *    - On success, zero.
3816  *    - On failure, a negative value.
3817  */
3818 static int
3819 igb_add_2tuple_filter(struct rte_eth_dev *dev,
3820 			struct rte_eth_ntuple_filter *ntuple_filter)
3821 {
3822 	struct e1000_filter_info *filter_info =
3823 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3824 	struct e1000_2tuple_filter *filter;
3825 	int i, ret;
3826 
3827 	filter = rte_zmalloc("e1000_2tuple_filter",
3828 			sizeof(struct e1000_2tuple_filter), 0);
3829 	if (filter == NULL)
3830 		return -ENOMEM;
3831 
3832 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3833 				      &filter->filter_info);
3834 	if (ret < 0) {
3835 		rte_free(filter);
3836 		return ret;
3837 	}
3838 	if (igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3839 					 &filter->filter_info) != NULL) {
3840 		PMD_DRV_LOG(ERR, "filter exists.");
3841 		rte_free(filter);
3842 		return -EEXIST;
3843 	}
3844 	filter->queue = ntuple_filter->queue;
3845 
3846 	/*
3847 	 * look for an unused 2tuple filter index,
3848 	 * and insert the filter to list.
3849 	 */
3850 	for (i = 0; i < E1000_MAX_TTQF_FILTERS; i++) {
3851 		if (!(filter_info->twotuple_mask & (1 << i))) {
3852 			filter_info->twotuple_mask |= 1 << i;
3853 			filter->index = i;
3854 			TAILQ_INSERT_TAIL(&filter_info->twotuple_list,
3855 					  filter,
3856 					  entries);
3857 			break;
3858 		}
3859 	}
3860 	if (i >= E1000_MAX_TTQF_FILTERS) {
3861 		PMD_DRV_LOG(ERR, "2tuple filters are full.");
3862 		rte_free(filter);
3863 		return -ENOSYS;
3864 	}
3865 
3866 	igb_inject_2uple_filter(dev, filter);
3867 	return 0;
3868 }
3869 
3870 int
3871 igb_delete_2tuple_filter(struct rte_eth_dev *dev,
3872 			struct e1000_2tuple_filter *filter)
3873 {
3874 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3875 	struct e1000_filter_info *filter_info =
3876 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3877 
3878 	filter_info->twotuple_mask &= ~(1 << filter->index);
3879 	TAILQ_REMOVE(&filter_info->twotuple_list, filter, entries);
3880 	rte_free(filter);
3881 
3882 	E1000_WRITE_REG(hw, E1000_TTQF(filter->index), E1000_TTQF_DISABLE_MASK);
3883 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
3884 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
3885 	return 0;
3886 }
3887 
3888 /*
3889  * igb_remove_2tuple_filter - remove a 2tuple filter
3890  *
3891  * @param
3892  * dev: Pointer to struct rte_eth_dev.
3893  * ntuple_filter: ponter to the filter that will be removed.
3894  *
3895  * @return
3896  *    - On success, zero.
3897  *    - On failure, a negative value.
3898  */
3899 static int
3900 igb_remove_2tuple_filter(struct rte_eth_dev *dev,
3901 			struct rte_eth_ntuple_filter *ntuple_filter)
3902 {
3903 	struct e1000_filter_info *filter_info =
3904 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3905 	struct e1000_2tuple_filter_info filter_2tuple;
3906 	struct e1000_2tuple_filter *filter;
3907 	int ret;
3908 
3909 	memset(&filter_2tuple, 0, sizeof(struct e1000_2tuple_filter_info));
3910 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3911 				      &filter_2tuple);
3912 	if (ret < 0)
3913 		return ret;
3914 
3915 	filter = igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3916 					 &filter_2tuple);
3917 	if (filter == NULL) {
3918 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
3919 		return -ENOENT;
3920 	}
3921 
3922 	igb_delete_2tuple_filter(dev, filter);
3923 
3924 	return 0;
3925 }
3926 
3927 /* inject a igb flex filter to HW */
3928 static inline void
3929 igb_inject_flex_filter(struct rte_eth_dev *dev,
3930 			   struct e1000_flex_filter *filter)
3931 {
3932 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3933 	uint32_t wufc, queueing;
3934 	uint32_t reg_off;
3935 	uint8_t i, j = 0;
3936 
3937 	wufc = E1000_READ_REG(hw, E1000_WUFC);
3938 	if (filter->index < E1000_MAX_FHFT)
3939 		reg_off = E1000_FHFT(filter->index);
3940 	else
3941 		reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
3942 
3943 	E1000_WRITE_REG(hw, E1000_WUFC, wufc | E1000_WUFC_FLEX_HQ |
3944 			(E1000_WUFC_FLX0 << filter->index));
3945 	queueing = filter->filter_info.len |
3946 		(filter->queue << E1000_FHFT_QUEUEING_QUEUE_SHIFT) |
3947 		(filter->filter_info.priority <<
3948 			E1000_FHFT_QUEUEING_PRIO_SHIFT);
3949 	E1000_WRITE_REG(hw, reg_off + E1000_FHFT_QUEUEING_OFFSET,
3950 			queueing);
3951 
3952 	for (i = 0; i < E1000_FLEX_FILTERS_MASK_SIZE; i++) {
3953 		E1000_WRITE_REG(hw, reg_off,
3954 				filter->filter_info.dwords[j]);
3955 		reg_off += sizeof(uint32_t);
3956 		E1000_WRITE_REG(hw, reg_off,
3957 				filter->filter_info.dwords[++j]);
3958 		reg_off += sizeof(uint32_t);
3959 		E1000_WRITE_REG(hw, reg_off,
3960 			(uint32_t)filter->filter_info.mask[i]);
3961 		reg_off += sizeof(uint32_t) * 2;
3962 		++j;
3963 	}
3964 }
3965 
3966 static inline struct e1000_flex_filter *
3967 eth_igb_flex_filter_lookup(struct e1000_flex_filter_list *filter_list,
3968 			struct e1000_flex_filter_info *key)
3969 {
3970 	struct e1000_flex_filter *it;
3971 
3972 	TAILQ_FOREACH(it, filter_list, entries) {
3973 		if (memcmp(key, &it->filter_info,
3974 			sizeof(struct e1000_flex_filter_info)) == 0)
3975 			return it;
3976 	}
3977 
3978 	return NULL;
3979 }
3980 
3981 /* remove a flex byte filter
3982  * @param
3983  * dev: Pointer to struct rte_eth_dev.
3984  * filter: the pointer of the filter will be removed.
3985  */
3986 void
3987 igb_remove_flex_filter(struct rte_eth_dev *dev,
3988 			struct e1000_flex_filter *filter)
3989 {
3990 	struct e1000_filter_info *filter_info =
3991 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3992 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3993 	uint32_t wufc, i;
3994 	uint32_t reg_off;
3995 
3996 	wufc = E1000_READ_REG(hw, E1000_WUFC);
3997 	if (filter->index < E1000_MAX_FHFT)
3998 		reg_off = E1000_FHFT(filter->index);
3999 	else
4000 		reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
4001 
4002 	for (i = 0; i < E1000_FHFT_SIZE_IN_DWD; i++)
4003 		E1000_WRITE_REG(hw, reg_off + i * sizeof(uint32_t), 0);
4004 
4005 	E1000_WRITE_REG(hw, E1000_WUFC, wufc &
4006 		(~(E1000_WUFC_FLX0 << filter->index)));
4007 
4008 	filter_info->flex_mask &= ~(1 << filter->index);
4009 	TAILQ_REMOVE(&filter_info->flex_list, filter, entries);
4010 	rte_free(filter);
4011 }
4012 
4013 int
4014 eth_igb_add_del_flex_filter(struct rte_eth_dev *dev,
4015 			struct rte_eth_flex_filter *filter,
4016 			bool add)
4017 {
4018 	struct e1000_filter_info *filter_info =
4019 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4020 	struct e1000_flex_filter *flex_filter, *it;
4021 	uint32_t mask;
4022 	uint8_t shift, i;
4023 
4024 	flex_filter = rte_zmalloc("e1000_flex_filter",
4025 			sizeof(struct e1000_flex_filter), 0);
4026 	if (flex_filter == NULL)
4027 		return -ENOMEM;
4028 
4029 	flex_filter->filter_info.len = filter->len;
4030 	flex_filter->filter_info.priority = filter->priority;
4031 	memcpy(flex_filter->filter_info.dwords, filter->bytes, filter->len);
4032 	for (i = 0; i < RTE_ALIGN(filter->len, CHAR_BIT) / CHAR_BIT; i++) {
4033 		mask = 0;
4034 		/* reverse bits in flex filter's mask*/
4035 		for (shift = 0; shift < CHAR_BIT; shift++) {
4036 			if (filter->mask[i] & (0x01 << shift))
4037 				mask |= (0x80 >> shift);
4038 		}
4039 		flex_filter->filter_info.mask[i] = mask;
4040 	}
4041 
4042 	it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
4043 				&flex_filter->filter_info);
4044 	if (it == NULL && !add) {
4045 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4046 		rte_free(flex_filter);
4047 		return -ENOENT;
4048 	}
4049 	if (it != NULL && add) {
4050 		PMD_DRV_LOG(ERR, "filter exists.");
4051 		rte_free(flex_filter);
4052 		return -EEXIST;
4053 	}
4054 
4055 	if (add) {
4056 		flex_filter->queue = filter->queue;
4057 		/*
4058 		 * look for an unused flex filter index
4059 		 * and insert the filter into the list.
4060 		 */
4061 		for (i = 0; i < E1000_MAX_FLEX_FILTERS; i++) {
4062 			if (!(filter_info->flex_mask & (1 << i))) {
4063 				filter_info->flex_mask |= 1 << i;
4064 				flex_filter->index = i;
4065 				TAILQ_INSERT_TAIL(&filter_info->flex_list,
4066 					flex_filter,
4067 					entries);
4068 				break;
4069 			}
4070 		}
4071 		if (i >= E1000_MAX_FLEX_FILTERS) {
4072 			PMD_DRV_LOG(ERR, "flex filters are full.");
4073 			rte_free(flex_filter);
4074 			return -ENOSYS;
4075 		}
4076 
4077 		igb_inject_flex_filter(dev, flex_filter);
4078 
4079 	} else {
4080 		igb_remove_flex_filter(dev, it);
4081 		rte_free(flex_filter);
4082 	}
4083 
4084 	return 0;
4085 }
4086 
4087 static int
4088 eth_igb_get_flex_filter(struct rte_eth_dev *dev,
4089 			struct rte_eth_flex_filter *filter)
4090 {
4091 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4092 	struct e1000_filter_info *filter_info =
4093 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4094 	struct e1000_flex_filter flex_filter, *it;
4095 	uint32_t wufc, queueing, wufc_en = 0;
4096 
4097 	memset(&flex_filter, 0, sizeof(struct e1000_flex_filter));
4098 	flex_filter.filter_info.len = filter->len;
4099 	flex_filter.filter_info.priority = filter->priority;
4100 	memcpy(flex_filter.filter_info.dwords, filter->bytes, filter->len);
4101 	memcpy(flex_filter.filter_info.mask, filter->mask,
4102 			RTE_ALIGN(filter->len, CHAR_BIT) / CHAR_BIT);
4103 
4104 	it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
4105 				&flex_filter.filter_info);
4106 	if (it == NULL) {
4107 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4108 		return -ENOENT;
4109 	}
4110 
4111 	wufc = E1000_READ_REG(hw, E1000_WUFC);
4112 	wufc_en = E1000_WUFC_FLEX_HQ | (E1000_WUFC_FLX0 << it->index);
4113 
4114 	if ((wufc & wufc_en) == wufc_en) {
4115 		uint32_t reg_off = 0;
4116 		if (it->index < E1000_MAX_FHFT)
4117 			reg_off = E1000_FHFT(it->index);
4118 		else
4119 			reg_off = E1000_FHFT_EXT(it->index - E1000_MAX_FHFT);
4120 
4121 		queueing = E1000_READ_REG(hw,
4122 				reg_off + E1000_FHFT_QUEUEING_OFFSET);
4123 		filter->len = queueing & E1000_FHFT_QUEUEING_LEN;
4124 		filter->priority = (queueing & E1000_FHFT_QUEUEING_PRIO) >>
4125 			E1000_FHFT_QUEUEING_PRIO_SHIFT;
4126 		filter->queue = (queueing & E1000_FHFT_QUEUEING_QUEUE) >>
4127 			E1000_FHFT_QUEUEING_QUEUE_SHIFT;
4128 		return 0;
4129 	}
4130 	return -ENOENT;
4131 }
4132 
4133 static int
4134 eth_igb_flex_filter_handle(struct rte_eth_dev *dev,
4135 			enum rte_filter_op filter_op,
4136 			void *arg)
4137 {
4138 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4139 	struct rte_eth_flex_filter *filter;
4140 	int ret = 0;
4141 
4142 	MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
4143 
4144 	if (filter_op == RTE_ETH_FILTER_NOP)
4145 		return ret;
4146 
4147 	if (arg == NULL) {
4148 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
4149 			    filter_op);
4150 		return -EINVAL;
4151 	}
4152 
4153 	filter = (struct rte_eth_flex_filter *)arg;
4154 	if (filter->len == 0 || filter->len > E1000_MAX_FLEX_FILTER_LEN
4155 	    || filter->len % sizeof(uint64_t) != 0) {
4156 		PMD_DRV_LOG(ERR, "filter's length is out of range");
4157 		return -EINVAL;
4158 	}
4159 	if (filter->priority > E1000_MAX_FLEX_FILTER_PRI) {
4160 		PMD_DRV_LOG(ERR, "filter's priority is out of range");
4161 		return -EINVAL;
4162 	}
4163 
4164 	switch (filter_op) {
4165 	case RTE_ETH_FILTER_ADD:
4166 		ret = eth_igb_add_del_flex_filter(dev, filter, TRUE);
4167 		break;
4168 	case RTE_ETH_FILTER_DELETE:
4169 		ret = eth_igb_add_del_flex_filter(dev, filter, FALSE);
4170 		break;
4171 	case RTE_ETH_FILTER_GET:
4172 		ret = eth_igb_get_flex_filter(dev, filter);
4173 		break;
4174 	default:
4175 		PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
4176 		ret = -EINVAL;
4177 		break;
4178 	}
4179 
4180 	return ret;
4181 }
4182 
4183 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_5tuple_filter_info*/
4184 static inline int
4185 ntuple_filter_to_5tuple_82576(struct rte_eth_ntuple_filter *filter,
4186 			struct e1000_5tuple_filter_info *filter_info)
4187 {
4188 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM_82576)
4189 		return -EINVAL;
4190 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
4191 		return -EINVAL;  /* filter index is out of range. */
4192 	if (filter->tcp_flags > TCP_FLAG_ALL)
4193 		return -EINVAL;  /* flags is invalid. */
4194 
4195 	switch (filter->dst_ip_mask) {
4196 	case UINT32_MAX:
4197 		filter_info->dst_ip_mask = 0;
4198 		filter_info->dst_ip = filter->dst_ip;
4199 		break;
4200 	case 0:
4201 		filter_info->dst_ip_mask = 1;
4202 		break;
4203 	default:
4204 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
4205 		return -EINVAL;
4206 	}
4207 
4208 	switch (filter->src_ip_mask) {
4209 	case UINT32_MAX:
4210 		filter_info->src_ip_mask = 0;
4211 		filter_info->src_ip = filter->src_ip;
4212 		break;
4213 	case 0:
4214 		filter_info->src_ip_mask = 1;
4215 		break;
4216 	default:
4217 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
4218 		return -EINVAL;
4219 	}
4220 
4221 	switch (filter->dst_port_mask) {
4222 	case UINT16_MAX:
4223 		filter_info->dst_port_mask = 0;
4224 		filter_info->dst_port = filter->dst_port;
4225 		break;
4226 	case 0:
4227 		filter_info->dst_port_mask = 1;
4228 		break;
4229 	default:
4230 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
4231 		return -EINVAL;
4232 	}
4233 
4234 	switch (filter->src_port_mask) {
4235 	case UINT16_MAX:
4236 		filter_info->src_port_mask = 0;
4237 		filter_info->src_port = filter->src_port;
4238 		break;
4239 	case 0:
4240 		filter_info->src_port_mask = 1;
4241 		break;
4242 	default:
4243 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
4244 		return -EINVAL;
4245 	}
4246 
4247 	switch (filter->proto_mask) {
4248 	case UINT8_MAX:
4249 		filter_info->proto_mask = 0;
4250 		filter_info->proto = filter->proto;
4251 		break;
4252 	case 0:
4253 		filter_info->proto_mask = 1;
4254 		break;
4255 	default:
4256 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
4257 		return -EINVAL;
4258 	}
4259 
4260 	filter_info->priority = (uint8_t)filter->priority;
4261 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
4262 		filter_info->tcp_flags = filter->tcp_flags;
4263 	else
4264 		filter_info->tcp_flags = 0;
4265 
4266 	return 0;
4267 }
4268 
4269 static inline struct e1000_5tuple_filter *
4270 igb_5tuple_filter_lookup_82576(struct e1000_5tuple_filter_list *filter_list,
4271 			struct e1000_5tuple_filter_info *key)
4272 {
4273 	struct e1000_5tuple_filter *it;
4274 
4275 	TAILQ_FOREACH(it, filter_list, entries) {
4276 		if (memcmp(key, &it->filter_info,
4277 			sizeof(struct e1000_5tuple_filter_info)) == 0) {
4278 			return it;
4279 		}
4280 	}
4281 	return NULL;
4282 }
4283 
4284 /* inject a igb 5-tuple filter to HW */
4285 static inline void
4286 igb_inject_5tuple_filter_82576(struct rte_eth_dev *dev,
4287 			   struct e1000_5tuple_filter *filter)
4288 {
4289 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4290 	uint32_t ftqf = E1000_FTQF_VF_BP | E1000_FTQF_MASK;
4291 	uint32_t spqf, imir, imir_ext = E1000_IMIREXT_SIZE_BP;
4292 	uint8_t i;
4293 
4294 	i = filter->index;
4295 	ftqf |= filter->filter_info.proto & E1000_FTQF_PROTOCOL_MASK;
4296 	if (filter->filter_info.src_ip_mask == 0) /* 0b means compare. */
4297 		ftqf &= ~E1000_FTQF_MASK_SOURCE_ADDR_BP;
4298 	if (filter->filter_info.dst_ip_mask == 0)
4299 		ftqf &= ~E1000_FTQF_MASK_DEST_ADDR_BP;
4300 	if (filter->filter_info.src_port_mask == 0)
4301 		ftqf &= ~E1000_FTQF_MASK_SOURCE_PORT_BP;
4302 	if (filter->filter_info.proto_mask == 0)
4303 		ftqf &= ~E1000_FTQF_MASK_PROTO_BP;
4304 	ftqf |= (filter->queue << E1000_FTQF_QUEUE_SHIFT) &
4305 		E1000_FTQF_QUEUE_MASK;
4306 	ftqf |= E1000_FTQF_QUEUE_ENABLE;
4307 	E1000_WRITE_REG(hw, E1000_FTQF(i), ftqf);
4308 	E1000_WRITE_REG(hw, E1000_DAQF(i), filter->filter_info.dst_ip);
4309 	E1000_WRITE_REG(hw, E1000_SAQF(i), filter->filter_info.src_ip);
4310 
4311 	spqf = filter->filter_info.src_port & E1000_SPQF_SRCPORT;
4312 	E1000_WRITE_REG(hw, E1000_SPQF(i), spqf);
4313 
4314 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
4315 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
4316 		imir |= E1000_IMIR_PORT_BP;
4317 	else
4318 		imir &= ~E1000_IMIR_PORT_BP;
4319 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
4320 
4321 	/* tcp flags bits setting. */
4322 	if (filter->filter_info.tcp_flags & TCP_FLAG_ALL) {
4323 		if (filter->filter_info.tcp_flags & TCP_URG_FLAG)
4324 			imir_ext |= E1000_IMIREXT_CTRL_URG;
4325 		if (filter->filter_info.tcp_flags & TCP_ACK_FLAG)
4326 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
4327 		if (filter->filter_info.tcp_flags & TCP_PSH_FLAG)
4328 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
4329 		if (filter->filter_info.tcp_flags & TCP_RST_FLAG)
4330 			imir_ext |= E1000_IMIREXT_CTRL_RST;
4331 		if (filter->filter_info.tcp_flags & TCP_SYN_FLAG)
4332 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
4333 		if (filter->filter_info.tcp_flags & TCP_FIN_FLAG)
4334 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
4335 	} else {
4336 		imir_ext |= E1000_IMIREXT_CTRL_BP;
4337 	}
4338 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
4339 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
4340 }
4341 
4342 /*
4343  * igb_add_5tuple_filter_82576 - add a 5tuple filter
4344  *
4345  * @param
4346  * dev: Pointer to struct rte_eth_dev.
4347  * ntuple_filter: ponter to the filter that will be added.
4348  *
4349  * @return
4350  *    - On success, zero.
4351  *    - On failure, a negative value.
4352  */
4353 static int
4354 igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
4355 			struct rte_eth_ntuple_filter *ntuple_filter)
4356 {
4357 	struct e1000_filter_info *filter_info =
4358 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4359 	struct e1000_5tuple_filter *filter;
4360 	uint8_t i;
4361 	int ret;
4362 
4363 	filter = rte_zmalloc("e1000_5tuple_filter",
4364 			sizeof(struct e1000_5tuple_filter), 0);
4365 	if (filter == NULL)
4366 		return -ENOMEM;
4367 
4368 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4369 					    &filter->filter_info);
4370 	if (ret < 0) {
4371 		rte_free(filter);
4372 		return ret;
4373 	}
4374 
4375 	if (igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4376 					 &filter->filter_info) != NULL) {
4377 		PMD_DRV_LOG(ERR, "filter exists.");
4378 		rte_free(filter);
4379 		return -EEXIST;
4380 	}
4381 	filter->queue = ntuple_filter->queue;
4382 
4383 	/*
4384 	 * look for an unused 5tuple filter index,
4385 	 * and insert the filter to list.
4386 	 */
4387 	for (i = 0; i < E1000_MAX_FTQF_FILTERS; i++) {
4388 		if (!(filter_info->fivetuple_mask & (1 << i))) {
4389 			filter_info->fivetuple_mask |= 1 << i;
4390 			filter->index = i;
4391 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
4392 					  filter,
4393 					  entries);
4394 			break;
4395 		}
4396 	}
4397 	if (i >= E1000_MAX_FTQF_FILTERS) {
4398 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
4399 		rte_free(filter);
4400 		return -ENOSYS;
4401 	}
4402 
4403 	igb_inject_5tuple_filter_82576(dev, filter);
4404 	return 0;
4405 }
4406 
4407 int
4408 igb_delete_5tuple_filter_82576(struct rte_eth_dev *dev,
4409 				struct e1000_5tuple_filter *filter)
4410 {
4411 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4412 	struct e1000_filter_info *filter_info =
4413 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4414 
4415 	filter_info->fivetuple_mask &= ~(1 << filter->index);
4416 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
4417 	rte_free(filter);
4418 
4419 	E1000_WRITE_REG(hw, E1000_FTQF(filter->index),
4420 			E1000_FTQF_VF_BP | E1000_FTQF_MASK);
4421 	E1000_WRITE_REG(hw, E1000_DAQF(filter->index), 0);
4422 	E1000_WRITE_REG(hw, E1000_SAQF(filter->index), 0);
4423 	E1000_WRITE_REG(hw, E1000_SPQF(filter->index), 0);
4424 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
4425 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
4426 	return 0;
4427 }
4428 
4429 /*
4430  * igb_remove_5tuple_filter_82576 - remove a 5tuple filter
4431  *
4432  * @param
4433  * dev: Pointer to struct rte_eth_dev.
4434  * ntuple_filter: ponter to the filter that will be removed.
4435  *
4436  * @return
4437  *    - On success, zero.
4438  *    - On failure, a negative value.
4439  */
4440 static int
4441 igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
4442 				struct rte_eth_ntuple_filter *ntuple_filter)
4443 {
4444 	struct e1000_filter_info *filter_info =
4445 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4446 	struct e1000_5tuple_filter_info filter_5tuple;
4447 	struct e1000_5tuple_filter *filter;
4448 	int ret;
4449 
4450 	memset(&filter_5tuple, 0, sizeof(struct e1000_5tuple_filter_info));
4451 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4452 					    &filter_5tuple);
4453 	if (ret < 0)
4454 		return ret;
4455 
4456 	filter = igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4457 					 &filter_5tuple);
4458 	if (filter == NULL) {
4459 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4460 		return -ENOENT;
4461 	}
4462 
4463 	igb_delete_5tuple_filter_82576(dev, filter);
4464 
4465 	return 0;
4466 }
4467 
4468 static int
4469 eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4470 {
4471 	uint32_t rctl;
4472 	struct e1000_hw *hw;
4473 	struct rte_eth_dev_info dev_info;
4474 	uint32_t frame_size = mtu + (ETHER_HDR_LEN + ETHER_CRC_LEN +
4475 				     VLAN_TAG_SIZE);
4476 
4477 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4478 
4479 #ifdef RTE_LIBRTE_82571_SUPPORT
4480 	/* XXX: not bigger than max_rx_pktlen */
4481 	if (hw->mac.type == e1000_82571)
4482 		return -ENOTSUP;
4483 #endif
4484 	eth_igb_infos_get(dev, &dev_info);
4485 
4486 	/* check that mtu is within the allowed range */
4487 	if ((mtu < ETHER_MIN_MTU) ||
4488 	    (frame_size > dev_info.max_rx_pktlen))
4489 		return -EINVAL;
4490 
4491 	/* refuse mtu that requires the support of scattered packets when this
4492 	 * feature has not been enabled before. */
4493 	if (!dev->data->scattered_rx &&
4494 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
4495 		return -EINVAL;
4496 
4497 	rctl = E1000_READ_REG(hw, E1000_RCTL);
4498 
4499 	/* switch to jumbo mode if needed */
4500 	if (frame_size > ETHER_MAX_LEN) {
4501 		dev->data->dev_conf.rxmode.jumbo_frame = 1;
4502 		rctl |= E1000_RCTL_LPE;
4503 	} else {
4504 		dev->data->dev_conf.rxmode.jumbo_frame = 0;
4505 		rctl &= ~E1000_RCTL_LPE;
4506 	}
4507 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
4508 
4509 	/* update max frame size */
4510 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4511 
4512 	E1000_WRITE_REG(hw, E1000_RLPML,
4513 			dev->data->dev_conf.rxmode.max_rx_pkt_len);
4514 
4515 	return 0;
4516 }
4517 
4518 /*
4519  * igb_add_del_ntuple_filter - add or delete a ntuple filter
4520  *
4521  * @param
4522  * dev: Pointer to struct rte_eth_dev.
4523  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4524  * add: if true, add filter, if false, remove filter
4525  *
4526  * @return
4527  *    - On success, zero.
4528  *    - On failure, a negative value.
4529  */
4530 int
4531 igb_add_del_ntuple_filter(struct rte_eth_dev *dev,
4532 			struct rte_eth_ntuple_filter *ntuple_filter,
4533 			bool add)
4534 {
4535 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4536 	int ret;
4537 
4538 	switch (ntuple_filter->flags) {
4539 	case RTE_5TUPLE_FLAGS:
4540 	case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4541 		if (hw->mac.type != e1000_82576)
4542 			return -ENOTSUP;
4543 		if (add)
4544 			ret = igb_add_5tuple_filter_82576(dev,
4545 							  ntuple_filter);
4546 		else
4547 			ret = igb_remove_5tuple_filter_82576(dev,
4548 							     ntuple_filter);
4549 		break;
4550 	case RTE_2TUPLE_FLAGS:
4551 	case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4552 		if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350 &&
4553 			hw->mac.type != e1000_i210 &&
4554 			hw->mac.type != e1000_i211)
4555 			return -ENOTSUP;
4556 		if (add)
4557 			ret = igb_add_2tuple_filter(dev, ntuple_filter);
4558 		else
4559 			ret = igb_remove_2tuple_filter(dev, ntuple_filter);
4560 		break;
4561 	default:
4562 		ret = -EINVAL;
4563 		break;
4564 	}
4565 
4566 	return ret;
4567 }
4568 
4569 /*
4570  * igb_get_ntuple_filter - get a ntuple filter
4571  *
4572  * @param
4573  * dev: Pointer to struct rte_eth_dev.
4574  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4575  *
4576  * @return
4577  *    - On success, zero.
4578  *    - On failure, a negative value.
4579  */
4580 static int
4581 igb_get_ntuple_filter(struct rte_eth_dev *dev,
4582 			struct rte_eth_ntuple_filter *ntuple_filter)
4583 {
4584 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4585 	struct e1000_filter_info *filter_info =
4586 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4587 	struct e1000_5tuple_filter_info filter_5tuple;
4588 	struct e1000_2tuple_filter_info filter_2tuple;
4589 	struct e1000_5tuple_filter *p_5tuple_filter;
4590 	struct e1000_2tuple_filter *p_2tuple_filter;
4591 	int ret;
4592 
4593 	switch (ntuple_filter->flags) {
4594 	case RTE_5TUPLE_FLAGS:
4595 	case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4596 		if (hw->mac.type != e1000_82576)
4597 			return -ENOTSUP;
4598 		memset(&filter_5tuple,
4599 			0,
4600 			sizeof(struct e1000_5tuple_filter_info));
4601 		ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4602 						    &filter_5tuple);
4603 		if (ret < 0)
4604 			return ret;
4605 		p_5tuple_filter = igb_5tuple_filter_lookup_82576(
4606 					&filter_info->fivetuple_list,
4607 					&filter_5tuple);
4608 		if (p_5tuple_filter == NULL) {
4609 			PMD_DRV_LOG(ERR, "filter doesn't exist.");
4610 			return -ENOENT;
4611 		}
4612 		ntuple_filter->queue = p_5tuple_filter->queue;
4613 		break;
4614 	case RTE_2TUPLE_FLAGS:
4615 	case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4616 		if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350)
4617 			return -ENOTSUP;
4618 		memset(&filter_2tuple,
4619 			0,
4620 			sizeof(struct e1000_2tuple_filter_info));
4621 		ret = ntuple_filter_to_2tuple(ntuple_filter, &filter_2tuple);
4622 		if (ret < 0)
4623 			return ret;
4624 		p_2tuple_filter = igb_2tuple_filter_lookup(
4625 					&filter_info->twotuple_list,
4626 					&filter_2tuple);
4627 		if (p_2tuple_filter == NULL) {
4628 			PMD_DRV_LOG(ERR, "filter doesn't exist.");
4629 			return -ENOENT;
4630 		}
4631 		ntuple_filter->queue = p_2tuple_filter->queue;
4632 		break;
4633 	default:
4634 		ret = -EINVAL;
4635 		break;
4636 	}
4637 
4638 	return 0;
4639 }
4640 
4641 /*
4642  * igb_ntuple_filter_handle - Handle operations for ntuple filter.
4643  * @dev: pointer to rte_eth_dev structure
4644  * @filter_op:operation will be taken.
4645  * @arg: a pointer to specific structure corresponding to the filter_op
4646  */
4647 static int
4648 igb_ntuple_filter_handle(struct rte_eth_dev *dev,
4649 				enum rte_filter_op filter_op,
4650 				void *arg)
4651 {
4652 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4653 	int ret;
4654 
4655 	MAC_TYPE_FILTER_SUP(hw->mac.type);
4656 
4657 	if (filter_op == RTE_ETH_FILTER_NOP)
4658 		return 0;
4659 
4660 	if (arg == NULL) {
4661 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4662 			    filter_op);
4663 		return -EINVAL;
4664 	}
4665 
4666 	switch (filter_op) {
4667 	case RTE_ETH_FILTER_ADD:
4668 		ret = igb_add_del_ntuple_filter(dev,
4669 			(struct rte_eth_ntuple_filter *)arg,
4670 			TRUE);
4671 		break;
4672 	case RTE_ETH_FILTER_DELETE:
4673 		ret = igb_add_del_ntuple_filter(dev,
4674 			(struct rte_eth_ntuple_filter *)arg,
4675 			FALSE);
4676 		break;
4677 	case RTE_ETH_FILTER_GET:
4678 		ret = igb_get_ntuple_filter(dev,
4679 			(struct rte_eth_ntuple_filter *)arg);
4680 		break;
4681 	default:
4682 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
4683 		ret = -EINVAL;
4684 		break;
4685 	}
4686 	return ret;
4687 }
4688 
4689 static inline int
4690 igb_ethertype_filter_lookup(struct e1000_filter_info *filter_info,
4691 			uint16_t ethertype)
4692 {
4693 	int i;
4694 
4695 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4696 		if (filter_info->ethertype_filters[i].ethertype == ethertype &&
4697 		    (filter_info->ethertype_mask & (1 << i)))
4698 			return i;
4699 	}
4700 	return -1;
4701 }
4702 
4703 static inline int
4704 igb_ethertype_filter_insert(struct e1000_filter_info *filter_info,
4705 			uint16_t ethertype, uint32_t etqf)
4706 {
4707 	int i;
4708 
4709 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4710 		if (!(filter_info->ethertype_mask & (1 << i))) {
4711 			filter_info->ethertype_mask |= 1 << i;
4712 			filter_info->ethertype_filters[i].ethertype = ethertype;
4713 			filter_info->ethertype_filters[i].etqf = etqf;
4714 			return i;
4715 		}
4716 	}
4717 	return -1;
4718 }
4719 
4720 int
4721 igb_ethertype_filter_remove(struct e1000_filter_info *filter_info,
4722 			uint8_t idx)
4723 {
4724 	if (idx >= E1000_MAX_ETQF_FILTERS)
4725 		return -1;
4726 	filter_info->ethertype_mask &= ~(1 << idx);
4727 	filter_info->ethertype_filters[idx].ethertype = 0;
4728 	filter_info->ethertype_filters[idx].etqf = 0;
4729 	return idx;
4730 }
4731 
4732 
4733 int
4734 igb_add_del_ethertype_filter(struct rte_eth_dev *dev,
4735 			struct rte_eth_ethertype_filter *filter,
4736 			bool add)
4737 {
4738 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4739 	struct e1000_filter_info *filter_info =
4740 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4741 	uint32_t etqf = 0;
4742 	int ret;
4743 
4744 	if (filter->ether_type == ETHER_TYPE_IPv4 ||
4745 		filter->ether_type == ETHER_TYPE_IPv6) {
4746 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
4747 			" ethertype filter.", filter->ether_type);
4748 		return -EINVAL;
4749 	}
4750 
4751 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
4752 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
4753 		return -EINVAL;
4754 	}
4755 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
4756 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
4757 		return -EINVAL;
4758 	}
4759 
4760 	ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4761 	if (ret >= 0 && add) {
4762 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
4763 			    filter->ether_type);
4764 		return -EEXIST;
4765 	}
4766 	if (ret < 0 && !add) {
4767 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4768 			    filter->ether_type);
4769 		return -ENOENT;
4770 	}
4771 
4772 	if (add) {
4773 		etqf |= E1000_ETQF_FILTER_ENABLE | E1000_ETQF_QUEUE_ENABLE;
4774 		etqf |= (uint32_t)(filter->ether_type & E1000_ETQF_ETHERTYPE);
4775 		etqf |= filter->queue << E1000_ETQF_QUEUE_SHIFT;
4776 		ret = igb_ethertype_filter_insert(filter_info,
4777 				filter->ether_type, etqf);
4778 		if (ret < 0) {
4779 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
4780 			return -ENOSYS;
4781 		}
4782 	} else {
4783 		ret = igb_ethertype_filter_remove(filter_info, (uint8_t)ret);
4784 		if (ret < 0)
4785 			return -ENOSYS;
4786 	}
4787 	E1000_WRITE_REG(hw, E1000_ETQF(ret), etqf);
4788 	E1000_WRITE_FLUSH(hw);
4789 
4790 	return 0;
4791 }
4792 
4793 static int
4794 igb_get_ethertype_filter(struct rte_eth_dev *dev,
4795 			struct rte_eth_ethertype_filter *filter)
4796 {
4797 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4798 	struct e1000_filter_info *filter_info =
4799 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4800 	uint32_t etqf;
4801 	int ret;
4802 
4803 	ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4804 	if (ret < 0) {
4805 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4806 			    filter->ether_type);
4807 		return -ENOENT;
4808 	}
4809 
4810 	etqf = E1000_READ_REG(hw, E1000_ETQF(ret));
4811 	if (etqf & E1000_ETQF_FILTER_ENABLE) {
4812 		filter->ether_type = etqf & E1000_ETQF_ETHERTYPE;
4813 		filter->flags = 0;
4814 		filter->queue = (etqf & E1000_ETQF_QUEUE) >>
4815 				E1000_ETQF_QUEUE_SHIFT;
4816 		return 0;
4817 	}
4818 
4819 	return -ENOENT;
4820 }
4821 
4822 /*
4823  * igb_ethertype_filter_handle - Handle operations for ethertype filter.
4824  * @dev: pointer to rte_eth_dev structure
4825  * @filter_op:operation will be taken.
4826  * @arg: a pointer to specific structure corresponding to the filter_op
4827  */
4828 static int
4829 igb_ethertype_filter_handle(struct rte_eth_dev *dev,
4830 				enum rte_filter_op filter_op,
4831 				void *arg)
4832 {
4833 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4834 	int ret;
4835 
4836 	MAC_TYPE_FILTER_SUP(hw->mac.type);
4837 
4838 	if (filter_op == RTE_ETH_FILTER_NOP)
4839 		return 0;
4840 
4841 	if (arg == NULL) {
4842 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4843 			    filter_op);
4844 		return -EINVAL;
4845 	}
4846 
4847 	switch (filter_op) {
4848 	case RTE_ETH_FILTER_ADD:
4849 		ret = igb_add_del_ethertype_filter(dev,
4850 			(struct rte_eth_ethertype_filter *)arg,
4851 			TRUE);
4852 		break;
4853 	case RTE_ETH_FILTER_DELETE:
4854 		ret = igb_add_del_ethertype_filter(dev,
4855 			(struct rte_eth_ethertype_filter *)arg,
4856 			FALSE);
4857 		break;
4858 	case RTE_ETH_FILTER_GET:
4859 		ret = igb_get_ethertype_filter(dev,
4860 			(struct rte_eth_ethertype_filter *)arg);
4861 		break;
4862 	default:
4863 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
4864 		ret = -EINVAL;
4865 		break;
4866 	}
4867 	return ret;
4868 }
4869 
4870 static int
4871 eth_igb_filter_ctrl(struct rte_eth_dev *dev,
4872 		     enum rte_filter_type filter_type,
4873 		     enum rte_filter_op filter_op,
4874 		     void *arg)
4875 {
4876 	int ret = 0;
4877 
4878 	switch (filter_type) {
4879 	case RTE_ETH_FILTER_NTUPLE:
4880 		ret = igb_ntuple_filter_handle(dev, filter_op, arg);
4881 		break;
4882 	case RTE_ETH_FILTER_ETHERTYPE:
4883 		ret = igb_ethertype_filter_handle(dev, filter_op, arg);
4884 		break;
4885 	case RTE_ETH_FILTER_SYN:
4886 		ret = eth_igb_syn_filter_handle(dev, filter_op, arg);
4887 		break;
4888 	case RTE_ETH_FILTER_FLEXIBLE:
4889 		ret = eth_igb_flex_filter_handle(dev, filter_op, arg);
4890 		break;
4891 	case RTE_ETH_FILTER_GENERIC:
4892 		if (filter_op != RTE_ETH_FILTER_GET)
4893 			return -EINVAL;
4894 		*(const void **)arg = &igb_flow_ops;
4895 		break;
4896 	default:
4897 		PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
4898 							filter_type);
4899 		break;
4900 	}
4901 
4902 	return ret;
4903 }
4904 
4905 static int
4906 eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
4907 			 struct ether_addr *mc_addr_set,
4908 			 uint32_t nb_mc_addr)
4909 {
4910 	struct e1000_hw *hw;
4911 
4912 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4913 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
4914 	return 0;
4915 }
4916 
4917 static uint64_t
4918 igb_read_systime_cyclecounter(struct rte_eth_dev *dev)
4919 {
4920 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4921 	uint64_t systime_cycles;
4922 
4923 	switch (hw->mac.type) {
4924 	case e1000_i210:
4925 	case e1000_i211:
4926 		/*
4927 		 * Need to read System Time Residue Register to be able
4928 		 * to read the other two registers.
4929 		 */
4930 		E1000_READ_REG(hw, E1000_SYSTIMR);
4931 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
4932 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4933 		systime_cycles += (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4934 				* NSEC_PER_SEC;
4935 		break;
4936 	case e1000_82580:
4937 	case e1000_i350:
4938 	case e1000_i354:
4939 		/*
4940 		 * Need to read System Time Residue Register to be able
4941 		 * to read the other two registers.
4942 		 */
4943 		E1000_READ_REG(hw, E1000_SYSTIMR);
4944 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4945 		/* Only the 8 LSB are valid. */
4946 		systime_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_SYSTIMH)
4947 				& 0xff) << 32;
4948 		break;
4949 	default:
4950 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4951 		systime_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4952 				<< 32;
4953 		break;
4954 	}
4955 
4956 	return systime_cycles;
4957 }
4958 
4959 static uint64_t
4960 igb_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
4961 {
4962 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4963 	uint64_t rx_tstamp_cycles;
4964 
4965 	switch (hw->mac.type) {
4966 	case e1000_i210:
4967 	case e1000_i211:
4968 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
4969 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4970 		rx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
4971 				* NSEC_PER_SEC;
4972 		break;
4973 	case e1000_82580:
4974 	case e1000_i350:
4975 	case e1000_i354:
4976 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4977 		/* Only the 8 LSB are valid. */
4978 		rx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_RXSTMPH)
4979 				& 0xff) << 32;
4980 		break;
4981 	default:
4982 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4983 		rx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
4984 				<< 32;
4985 		break;
4986 	}
4987 
4988 	return rx_tstamp_cycles;
4989 }
4990 
4991 static uint64_t
4992 igb_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
4993 {
4994 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4995 	uint64_t tx_tstamp_cycles;
4996 
4997 	switch (hw->mac.type) {
4998 	case e1000_i210:
4999 	case e1000_i211:
5000 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
5001 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5002 		tx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
5003 				* NSEC_PER_SEC;
5004 		break;
5005 	case e1000_82580:
5006 	case e1000_i350:
5007 	case e1000_i354:
5008 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5009 		/* Only the 8 LSB are valid. */
5010 		tx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_TXSTMPH)
5011 				& 0xff) << 32;
5012 		break;
5013 	default:
5014 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5015 		tx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
5016 				<< 32;
5017 		break;
5018 	}
5019 
5020 	return tx_tstamp_cycles;
5021 }
5022 
5023 static void
5024 igb_start_timecounters(struct rte_eth_dev *dev)
5025 {
5026 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5027 	struct e1000_adapter *adapter =
5028 		(struct e1000_adapter *)dev->data->dev_private;
5029 	uint32_t incval = 1;
5030 	uint32_t shift = 0;
5031 	uint64_t mask = E1000_CYCLECOUNTER_MASK;
5032 
5033 	switch (hw->mac.type) {
5034 	case e1000_82580:
5035 	case e1000_i350:
5036 	case e1000_i354:
5037 		/* 32 LSB bits + 8 MSB bits = 40 bits */
5038 		mask = (1ULL << 40) - 1;
5039 		/* fall-through */
5040 	case e1000_i210:
5041 	case e1000_i211:
5042 		/*
5043 		 * Start incrementing the register
5044 		 * used to timestamp PTP packets.
5045 		 */
5046 		E1000_WRITE_REG(hw, E1000_TIMINCA, incval);
5047 		break;
5048 	case e1000_82576:
5049 		incval = E1000_INCVALUE_82576;
5050 		shift = IGB_82576_TSYNC_SHIFT;
5051 		E1000_WRITE_REG(hw, E1000_TIMINCA,
5052 				E1000_INCPERIOD_82576 | incval);
5053 		break;
5054 	default:
5055 		/* Not supported */
5056 		return;
5057 	}
5058 
5059 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
5060 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
5061 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
5062 
5063 	adapter->systime_tc.cc_mask = mask;
5064 	adapter->systime_tc.cc_shift = shift;
5065 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
5066 
5067 	adapter->rx_tstamp_tc.cc_mask = mask;
5068 	adapter->rx_tstamp_tc.cc_shift = shift;
5069 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
5070 
5071 	adapter->tx_tstamp_tc.cc_mask = mask;
5072 	adapter->tx_tstamp_tc.cc_shift = shift;
5073 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
5074 }
5075 
5076 static int
5077 igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
5078 {
5079 	struct e1000_adapter *adapter =
5080 			(struct e1000_adapter *)dev->data->dev_private;
5081 
5082 	adapter->systime_tc.nsec += delta;
5083 	adapter->rx_tstamp_tc.nsec += delta;
5084 	adapter->tx_tstamp_tc.nsec += delta;
5085 
5086 	return 0;
5087 }
5088 
5089 static int
5090 igb_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
5091 {
5092 	uint64_t ns;
5093 	struct e1000_adapter *adapter =
5094 			(struct e1000_adapter *)dev->data->dev_private;
5095 
5096 	ns = rte_timespec_to_ns(ts);
5097 
5098 	/* Set the timecounters to a new value. */
5099 	adapter->systime_tc.nsec = ns;
5100 	adapter->rx_tstamp_tc.nsec = ns;
5101 	adapter->tx_tstamp_tc.nsec = ns;
5102 
5103 	return 0;
5104 }
5105 
5106 static int
5107 igb_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
5108 {
5109 	uint64_t ns, systime_cycles;
5110 	struct e1000_adapter *adapter =
5111 			(struct e1000_adapter *)dev->data->dev_private;
5112 
5113 	systime_cycles = igb_read_systime_cyclecounter(dev);
5114 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
5115 	*ts = rte_ns_to_timespec(ns);
5116 
5117 	return 0;
5118 }
5119 
5120 static int
5121 igb_timesync_enable(struct rte_eth_dev *dev)
5122 {
5123 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5124 	uint32_t tsync_ctl;
5125 	uint32_t tsauxc;
5126 
5127 	/* Stop the timesync system time. */
5128 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0x0);
5129 	/* Reset the timesync system time value. */
5130 	switch (hw->mac.type) {
5131 	case e1000_82580:
5132 	case e1000_i350:
5133 	case e1000_i354:
5134 	case e1000_i210:
5135 	case e1000_i211:
5136 		E1000_WRITE_REG(hw, E1000_SYSTIMR, 0x0);
5137 		/* fall-through */
5138 	case e1000_82576:
5139 		E1000_WRITE_REG(hw, E1000_SYSTIML, 0x0);
5140 		E1000_WRITE_REG(hw, E1000_SYSTIMH, 0x0);
5141 		break;
5142 	default:
5143 		/* Not supported. */
5144 		return -ENOTSUP;
5145 	}
5146 
5147 	/* Enable system time for it isn't on by default. */
5148 	tsauxc = E1000_READ_REG(hw, E1000_TSAUXC);
5149 	tsauxc &= ~E1000_TSAUXC_DISABLE_SYSTIME;
5150 	E1000_WRITE_REG(hw, E1000_TSAUXC, tsauxc);
5151 
5152 	igb_start_timecounters(dev);
5153 
5154 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5155 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588),
5156 			(ETHER_TYPE_1588 |
5157 			 E1000_ETQF_FILTER_ENABLE |
5158 			 E1000_ETQF_1588));
5159 
5160 	/* Enable timestamping of received PTP packets. */
5161 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5162 	tsync_ctl |= E1000_TSYNCRXCTL_ENABLED;
5163 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
5164 
5165 	/* Enable Timestamping of transmitted PTP packets. */
5166 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5167 	tsync_ctl |= E1000_TSYNCTXCTL_ENABLED;
5168 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
5169 
5170 	return 0;
5171 }
5172 
5173 static int
5174 igb_timesync_disable(struct rte_eth_dev *dev)
5175 {
5176 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5177 	uint32_t tsync_ctl;
5178 
5179 	/* Disable timestamping of transmitted PTP packets. */
5180 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5181 	tsync_ctl &= ~E1000_TSYNCTXCTL_ENABLED;
5182 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
5183 
5184 	/* Disable timestamping of received PTP packets. */
5185 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5186 	tsync_ctl &= ~E1000_TSYNCRXCTL_ENABLED;
5187 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
5188 
5189 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5190 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588), 0);
5191 
5192 	/* Stop incrementating the System Time registers. */
5193 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0);
5194 
5195 	return 0;
5196 }
5197 
5198 static int
5199 igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
5200 			       struct timespec *timestamp,
5201 			       uint32_t flags __rte_unused)
5202 {
5203 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5204 	struct e1000_adapter *adapter =
5205 			(struct e1000_adapter *)dev->data->dev_private;
5206 	uint32_t tsync_rxctl;
5207 	uint64_t rx_tstamp_cycles;
5208 	uint64_t ns;
5209 
5210 	tsync_rxctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5211 	if ((tsync_rxctl & E1000_TSYNCRXCTL_VALID) == 0)
5212 		return -EINVAL;
5213 
5214 	rx_tstamp_cycles = igb_read_rx_tstamp_cyclecounter(dev);
5215 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
5216 	*timestamp = rte_ns_to_timespec(ns);
5217 
5218 	return  0;
5219 }
5220 
5221 static int
5222 igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
5223 			       struct timespec *timestamp)
5224 {
5225 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5226 	struct e1000_adapter *adapter =
5227 			(struct e1000_adapter *)dev->data->dev_private;
5228 	uint32_t tsync_txctl;
5229 	uint64_t tx_tstamp_cycles;
5230 	uint64_t ns;
5231 
5232 	tsync_txctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5233 	if ((tsync_txctl & E1000_TSYNCTXCTL_VALID) == 0)
5234 		return -EINVAL;
5235 
5236 	tx_tstamp_cycles = igb_read_tx_tstamp_cyclecounter(dev);
5237 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
5238 	*timestamp = rte_ns_to_timespec(ns);
5239 
5240 	return  0;
5241 }
5242 
5243 static int
5244 eth_igb_get_reg_length(struct rte_eth_dev *dev __rte_unused)
5245 {
5246 	int count = 0;
5247 	int g_ind = 0;
5248 	const struct reg_info *reg_group;
5249 
5250 	while ((reg_group = igb_regs[g_ind++]))
5251 		count += igb_reg_group_count(reg_group);
5252 
5253 	return count;
5254 }
5255 
5256 static int
5257 igbvf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
5258 {
5259 	int count = 0;
5260 	int g_ind = 0;
5261 	const struct reg_info *reg_group;
5262 
5263 	while ((reg_group = igbvf_regs[g_ind++]))
5264 		count += igb_reg_group_count(reg_group);
5265 
5266 	return count;
5267 }
5268 
5269 static int
5270 eth_igb_get_regs(struct rte_eth_dev *dev,
5271 	struct rte_dev_reg_info *regs)
5272 {
5273 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5274 	uint32_t *data = regs->data;
5275 	int g_ind = 0;
5276 	int count = 0;
5277 	const struct reg_info *reg_group;
5278 
5279 	if (data == NULL) {
5280 		regs->length = eth_igb_get_reg_length(dev);
5281 		regs->width = sizeof(uint32_t);
5282 		return 0;
5283 	}
5284 
5285 	/* Support only full register dump */
5286 	if ((regs->length == 0) ||
5287 	    (regs->length == (uint32_t)eth_igb_get_reg_length(dev))) {
5288 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5289 			hw->device_id;
5290 		while ((reg_group = igb_regs[g_ind++]))
5291 			count += igb_read_regs_group(dev, &data[count],
5292 							reg_group);
5293 		return 0;
5294 	}
5295 
5296 	return -ENOTSUP;
5297 }
5298 
5299 static int
5300 igbvf_get_regs(struct rte_eth_dev *dev,
5301 	struct rte_dev_reg_info *regs)
5302 {
5303 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5304 	uint32_t *data = regs->data;
5305 	int g_ind = 0;
5306 	int count = 0;
5307 	const struct reg_info *reg_group;
5308 
5309 	if (data == NULL) {
5310 		regs->length = igbvf_get_reg_length(dev);
5311 		regs->width = sizeof(uint32_t);
5312 		return 0;
5313 	}
5314 
5315 	/* Support only full register dump */
5316 	if ((regs->length == 0) ||
5317 	    (regs->length == (uint32_t)igbvf_get_reg_length(dev))) {
5318 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5319 			hw->device_id;
5320 		while ((reg_group = igbvf_regs[g_ind++]))
5321 			count += igb_read_regs_group(dev, &data[count],
5322 							reg_group);
5323 		return 0;
5324 	}
5325 
5326 	return -ENOTSUP;
5327 }
5328 
5329 static int
5330 eth_igb_get_eeprom_length(struct rte_eth_dev *dev)
5331 {
5332 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5333 
5334 	/* Return unit is byte count */
5335 	return hw->nvm.word_size * 2;
5336 }
5337 
5338 static int
5339 eth_igb_get_eeprom(struct rte_eth_dev *dev,
5340 	struct rte_dev_eeprom_info *in_eeprom)
5341 {
5342 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5343 	struct e1000_nvm_info *nvm = &hw->nvm;
5344 	uint16_t *data = in_eeprom->data;
5345 	int first, length;
5346 
5347 	first = in_eeprom->offset >> 1;
5348 	length = in_eeprom->length >> 1;
5349 	if ((first >= hw->nvm.word_size) ||
5350 	    ((first + length) >= hw->nvm.word_size))
5351 		return -EINVAL;
5352 
5353 	in_eeprom->magic = hw->vendor_id |
5354 		((uint32_t)hw->device_id << 16);
5355 
5356 	if ((nvm->ops.read) == NULL)
5357 		return -ENOTSUP;
5358 
5359 	return nvm->ops.read(hw, first, length, data);
5360 }
5361 
5362 static int
5363 eth_igb_set_eeprom(struct rte_eth_dev *dev,
5364 	struct rte_dev_eeprom_info *in_eeprom)
5365 {
5366 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5367 	struct e1000_nvm_info *nvm = &hw->nvm;
5368 	uint16_t *data = in_eeprom->data;
5369 	int first, length;
5370 
5371 	first = in_eeprom->offset >> 1;
5372 	length = in_eeprom->length >> 1;
5373 	if ((first >= hw->nvm.word_size) ||
5374 	    ((first + length) >= hw->nvm.word_size))
5375 		return -EINVAL;
5376 
5377 	in_eeprom->magic = (uint32_t)hw->vendor_id |
5378 		((uint32_t)hw->device_id << 16);
5379 
5380 	if ((nvm->ops.write) == NULL)
5381 		return -ENOTSUP;
5382 	return nvm->ops.write(hw,  first, length, data);
5383 }
5384 
5385 static int
5386 eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5387 {
5388 	struct e1000_hw *hw =
5389 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5390 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5391 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5392 	uint32_t vec = E1000_MISC_VEC_ID;
5393 
5394 	if (rte_intr_allow_others(intr_handle))
5395 		vec = E1000_RX_VEC_START;
5396 
5397 	uint32_t mask = 1 << (queue_id + vec);
5398 
5399 	E1000_WRITE_REG(hw, E1000_EIMC, mask);
5400 	E1000_WRITE_FLUSH(hw);
5401 
5402 	return 0;
5403 }
5404 
5405 static int
5406 eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5407 {
5408 	struct e1000_hw *hw =
5409 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5410 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5411 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5412 	uint32_t vec = E1000_MISC_VEC_ID;
5413 
5414 	if (rte_intr_allow_others(intr_handle))
5415 		vec = E1000_RX_VEC_START;
5416 
5417 	uint32_t mask = 1 << (queue_id + vec);
5418 	uint32_t regval;
5419 
5420 	regval = E1000_READ_REG(hw, E1000_EIMS);
5421 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
5422 	E1000_WRITE_FLUSH(hw);
5423 
5424 	rte_intr_enable(intr_handle);
5425 
5426 	return 0;
5427 }
5428 
5429 static void
5430 eth_igb_write_ivar(struct e1000_hw *hw, uint8_t  msix_vector,
5431 		   uint8_t index, uint8_t offset)
5432 {
5433 	uint32_t val = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
5434 
5435 	/* clear bits */
5436 	val &= ~((uint32_t)0xFF << offset);
5437 
5438 	/* write vector and valid bit */
5439 	val |= (msix_vector | E1000_IVAR_VALID) << offset;
5440 
5441 	E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, val);
5442 }
5443 
5444 static void
5445 eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
5446 			   uint8_t queue, uint8_t msix_vector)
5447 {
5448 	uint32_t tmp = 0;
5449 
5450 	if (hw->mac.type == e1000_82575) {
5451 		if (direction == 0)
5452 			tmp = E1000_EICR_RX_QUEUE0 << queue;
5453 		else if (direction == 1)
5454 			tmp = E1000_EICR_TX_QUEUE0 << queue;
5455 		E1000_WRITE_REG(hw, E1000_MSIXBM(msix_vector), tmp);
5456 	} else if (hw->mac.type == e1000_82576) {
5457 		if ((direction == 0) || (direction == 1))
5458 			eth_igb_write_ivar(hw, msix_vector, queue & 0x7,
5459 					   ((queue & 0x8) << 1) +
5460 					   8 * direction);
5461 	} else if ((hw->mac.type == e1000_82580) ||
5462 			(hw->mac.type == e1000_i350) ||
5463 			(hw->mac.type == e1000_i354) ||
5464 			(hw->mac.type == e1000_i210) ||
5465 			(hw->mac.type == e1000_i211)) {
5466 		if ((direction == 0) || (direction == 1))
5467 			eth_igb_write_ivar(hw, msix_vector,
5468 					   queue >> 1,
5469 					   ((queue & 0x1) << 4) +
5470 					   8 * direction);
5471 	}
5472 }
5473 
5474 /* Sets up the hardware to generate MSI-X interrupts properly
5475  * @hw
5476  *  board private structure
5477  */
5478 static void
5479 eth_igb_configure_msix_intr(struct rte_eth_dev *dev)
5480 {
5481 	int queue_id;
5482 	uint32_t tmpval, regval, intr_mask;
5483 	struct e1000_hw *hw =
5484 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5485 	uint32_t vec = E1000_MISC_VEC_ID;
5486 	uint32_t base = E1000_MISC_VEC_ID;
5487 	uint32_t misc_shift = 0;
5488 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5489 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5490 
5491 	/* won't configure msix register if no mapping is done
5492 	 * between intr vector and event fd
5493 	 */
5494 	if (!rte_intr_dp_is_en(intr_handle))
5495 		return;
5496 
5497 	if (rte_intr_allow_others(intr_handle)) {
5498 		vec = base = E1000_RX_VEC_START;
5499 		misc_shift = 1;
5500 	}
5501 
5502 	/* set interrupt vector for other causes */
5503 	if (hw->mac.type == e1000_82575) {
5504 		tmpval = E1000_READ_REG(hw, E1000_CTRL_EXT);
5505 		/* enable MSI-X PBA support */
5506 		tmpval |= E1000_CTRL_EXT_PBA_CLR;
5507 
5508 		/* Auto-Mask interrupts upon ICR read */
5509 		tmpval |= E1000_CTRL_EXT_EIAME;
5510 		tmpval |= E1000_CTRL_EXT_IRCA;
5511 
5512 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmpval);
5513 
5514 		/* enable msix_other interrupt */
5515 		E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0), 0, E1000_EIMS_OTHER);
5516 		regval = E1000_READ_REG(hw, E1000_EIAC);
5517 		E1000_WRITE_REG(hw, E1000_EIAC, regval | E1000_EIMS_OTHER);
5518 		regval = E1000_READ_REG(hw, E1000_EIAM);
5519 		E1000_WRITE_REG(hw, E1000_EIMS, regval | E1000_EIMS_OTHER);
5520 	} else if ((hw->mac.type == e1000_82576) ||
5521 			(hw->mac.type == e1000_82580) ||
5522 			(hw->mac.type == e1000_i350) ||
5523 			(hw->mac.type == e1000_i354) ||
5524 			(hw->mac.type == e1000_i210) ||
5525 			(hw->mac.type == e1000_i211)) {
5526 		/* turn on MSI-X capability first */
5527 		E1000_WRITE_REG(hw, E1000_GPIE, E1000_GPIE_MSIX_MODE |
5528 					E1000_GPIE_PBA | E1000_GPIE_EIAME |
5529 					E1000_GPIE_NSICR);
5530 		intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5531 			misc_shift;
5532 		regval = E1000_READ_REG(hw, E1000_EIAC);
5533 		E1000_WRITE_REG(hw, E1000_EIAC, regval | intr_mask);
5534 
5535 		/* enable msix_other interrupt */
5536 		regval = E1000_READ_REG(hw, E1000_EIMS);
5537 		E1000_WRITE_REG(hw, E1000_EIMS, regval | intr_mask);
5538 		tmpval = (dev->data->nb_rx_queues | E1000_IVAR_VALID) << 8;
5539 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, tmpval);
5540 	}
5541 
5542 	/* use EIAM to auto-mask when MSI-X interrupt
5543 	 * is asserted, this saves a register write for every interrupt
5544 	 */
5545 	intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5546 		misc_shift;
5547 	regval = E1000_READ_REG(hw, E1000_EIAM);
5548 	E1000_WRITE_REG(hw, E1000_EIAM, regval | intr_mask);
5549 
5550 	for (queue_id = 0; queue_id < dev->data->nb_rx_queues; queue_id++) {
5551 		eth_igb_assign_msix_vector(hw, 0, queue_id, vec);
5552 		intr_handle->intr_vec[queue_id] = vec;
5553 		if (vec < base + intr_handle->nb_efd - 1)
5554 			vec++;
5555 	}
5556 
5557 	E1000_WRITE_FLUSH(hw);
5558 }
5559 
5560 /* restore n-tuple filter */
5561 static inline void
5562 igb_ntuple_filter_restore(struct rte_eth_dev *dev)
5563 {
5564 	struct e1000_filter_info *filter_info =
5565 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5566 	struct e1000_5tuple_filter *p_5tuple;
5567 	struct e1000_2tuple_filter *p_2tuple;
5568 
5569 	TAILQ_FOREACH(p_5tuple, &filter_info->fivetuple_list, entries) {
5570 		igb_inject_5tuple_filter_82576(dev, p_5tuple);
5571 	}
5572 
5573 	TAILQ_FOREACH(p_2tuple, &filter_info->twotuple_list, entries) {
5574 		igb_inject_2uple_filter(dev, p_2tuple);
5575 	}
5576 }
5577 
5578 /* restore SYN filter */
5579 static inline void
5580 igb_syn_filter_restore(struct rte_eth_dev *dev)
5581 {
5582 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5583 	struct e1000_filter_info *filter_info =
5584 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5585 	uint32_t synqf;
5586 
5587 	synqf = filter_info->syn_info;
5588 
5589 	if (synqf & E1000_SYN_FILTER_ENABLE) {
5590 		E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
5591 		E1000_WRITE_FLUSH(hw);
5592 	}
5593 }
5594 
5595 /* restore ethernet type filter */
5596 static inline void
5597 igb_ethertype_filter_restore(struct rte_eth_dev *dev)
5598 {
5599 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5600 	struct e1000_filter_info *filter_info =
5601 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5602 	int i;
5603 
5604 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
5605 		if (filter_info->ethertype_mask & (1 << i)) {
5606 			E1000_WRITE_REG(hw, E1000_ETQF(i),
5607 				filter_info->ethertype_filters[i].etqf);
5608 			E1000_WRITE_FLUSH(hw);
5609 		}
5610 	}
5611 }
5612 
5613 /* restore flex byte filter */
5614 static inline void
5615 igb_flex_filter_restore(struct rte_eth_dev *dev)
5616 {
5617 	struct e1000_filter_info *filter_info =
5618 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5619 	struct e1000_flex_filter *flex_filter;
5620 
5621 	TAILQ_FOREACH(flex_filter, &filter_info->flex_list, entries) {
5622 		igb_inject_flex_filter(dev, flex_filter);
5623 	}
5624 }
5625 
5626 /* restore rss filter */
5627 static inline void
5628 igb_rss_filter_restore(struct rte_eth_dev *dev)
5629 {
5630 	struct e1000_filter_info *filter_info =
5631 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5632 
5633 	if (filter_info->rss_info.num)
5634 		igb_config_rss_filter(dev, &filter_info->rss_info, TRUE);
5635 }
5636 
5637 /* restore all types filter */
5638 static int
5639 igb_filter_restore(struct rte_eth_dev *dev)
5640 {
5641 	igb_ntuple_filter_restore(dev);
5642 	igb_ethertype_filter_restore(dev);
5643 	igb_syn_filter_restore(dev);
5644 	igb_flex_filter_restore(dev);
5645 	igb_rss_filter_restore(dev);
5646 
5647 	return 0;
5648 }
5649 
5650 RTE_PMD_REGISTER_PCI(net_e1000_igb, rte_igb_pmd);
5651 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb, pci_id_igb_map);
5652 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb, "* igb_uio | uio_pci_generic | vfio-pci");
5653 RTE_PMD_REGISTER_PCI(net_e1000_igb_vf, rte_igbvf_pmd);
5654 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb_vf, pci_id_igbvf_map);
5655 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb_vf, "* igb_uio | vfio-pci");
5656