xref: /dpdk/drivers/net/e1000/igb_ethdev.c (revision 2addc463)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2016 Intel Corporation
3  */
4 
5 #include <sys/queue.h>
6 #include <stdio.h>
7 #include <errno.h>
8 #include <stdint.h>
9 #include <stdarg.h>
10 
11 #include <rte_string_fns.h>
12 #include <rte_common.h>
13 #include <rte_interrupts.h>
14 #include <rte_byteorder.h>
15 #include <rte_log.h>
16 #include <rte_debug.h>
17 #include <rte_pci.h>
18 #include <rte_bus_pci.h>
19 #include <rte_ether.h>
20 #include <ethdev_driver.h>
21 #include <ethdev_pci.h>
22 #include <rte_memory.h>
23 #include <rte_eal.h>
24 #include <rte_malloc.h>
25 #include <rte_dev.h>
26 
27 #include "e1000_logs.h"
28 #include "base/e1000_api.h"
29 #include "e1000_ethdev.h"
30 #include "igb_regs.h"
31 
32 /*
33  * Default values for port configuration
34  */
35 #define IGB_DEFAULT_RX_FREE_THRESH  32
36 
37 #define IGB_DEFAULT_RX_PTHRESH      ((hw->mac.type == e1000_i354) ? 12 : 8)
38 #define IGB_DEFAULT_RX_HTHRESH      8
39 #define IGB_DEFAULT_RX_WTHRESH      ((hw->mac.type == e1000_82576) ? 1 : 4)
40 
41 #define IGB_DEFAULT_TX_PTHRESH      ((hw->mac.type == e1000_i354) ? 20 : 8)
42 #define IGB_DEFAULT_TX_HTHRESH      1
43 #define IGB_DEFAULT_TX_WTHRESH      ((hw->mac.type == e1000_82576) ? 1 : 16)
44 
45 /* Bit shift and mask */
46 #define IGB_4_BIT_WIDTH  (CHAR_BIT / 2)
47 #define IGB_4_BIT_MASK   RTE_LEN2MASK(IGB_4_BIT_WIDTH, uint8_t)
48 #define IGB_8_BIT_WIDTH  CHAR_BIT
49 #define IGB_8_BIT_MASK   UINT8_MAX
50 
51 /* Additional timesync values. */
52 #define E1000_CYCLECOUNTER_MASK      0xffffffffffffffffULL
53 #define E1000_ETQF_FILTER_1588       3
54 #define IGB_82576_TSYNC_SHIFT        16
55 #define E1000_INCPERIOD_82576        (1 << E1000_TIMINCA_16NS_SHIFT)
56 #define E1000_INCVALUE_82576         (16 << IGB_82576_TSYNC_SHIFT)
57 #define E1000_TSAUXC_DISABLE_SYSTIME 0x80000000
58 
59 #define E1000_VTIVAR_MISC                0x01740
60 #define E1000_VTIVAR_MISC_MASK           0xFF
61 #define E1000_VTIVAR_VALID               0x80
62 #define E1000_VTIVAR_MISC_MAILBOX        0
63 #define E1000_VTIVAR_MISC_INTR_MASK      0x3
64 
65 /* External VLAN Enable bit mask */
66 #define E1000_CTRL_EXT_EXT_VLAN      (1 << 26)
67 
68 /* External VLAN Ether Type bit mask and shift */
69 #define E1000_VET_VET_EXT            0xFFFF0000
70 #define E1000_VET_VET_EXT_SHIFT      16
71 
72 /* MSI-X other interrupt vector */
73 #define IGB_MSIX_OTHER_INTR_VEC      0
74 
75 static int  eth_igb_configure(struct rte_eth_dev *dev);
76 static int  eth_igb_start(struct rte_eth_dev *dev);
77 static int  eth_igb_stop(struct rte_eth_dev *dev);
78 static int  eth_igb_dev_set_link_up(struct rte_eth_dev *dev);
79 static int  eth_igb_dev_set_link_down(struct rte_eth_dev *dev);
80 static int eth_igb_close(struct rte_eth_dev *dev);
81 static int eth_igb_reset(struct rte_eth_dev *dev);
82 static int  eth_igb_promiscuous_enable(struct rte_eth_dev *dev);
83 static int  eth_igb_promiscuous_disable(struct rte_eth_dev *dev);
84 static int  eth_igb_allmulticast_enable(struct rte_eth_dev *dev);
85 static int  eth_igb_allmulticast_disable(struct rte_eth_dev *dev);
86 static int  eth_igb_link_update(struct rte_eth_dev *dev,
87 				int wait_to_complete);
88 static int eth_igb_stats_get(struct rte_eth_dev *dev,
89 				struct rte_eth_stats *rte_stats);
90 static int eth_igb_xstats_get(struct rte_eth_dev *dev,
91 			      struct rte_eth_xstat *xstats, unsigned n);
92 static int eth_igb_xstats_get_by_id(struct rte_eth_dev *dev,
93 		const uint64_t *ids,
94 		uint64_t *values, unsigned int n);
95 static int eth_igb_xstats_get_names(struct rte_eth_dev *dev,
96 				    struct rte_eth_xstat_name *xstats_names,
97 				    unsigned int size);
98 static int eth_igb_xstats_get_names_by_id(struct rte_eth_dev *dev,
99 		struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
100 		unsigned int limit);
101 static int eth_igb_stats_reset(struct rte_eth_dev *dev);
102 static int eth_igb_xstats_reset(struct rte_eth_dev *dev);
103 static int eth_igb_fw_version_get(struct rte_eth_dev *dev,
104 				   char *fw_version, size_t fw_size);
105 static int eth_igb_infos_get(struct rte_eth_dev *dev,
106 			      struct rte_eth_dev_info *dev_info);
107 static const uint32_t *eth_igb_supported_ptypes_get(struct rte_eth_dev *dev);
108 static int eth_igbvf_infos_get(struct rte_eth_dev *dev,
109 				struct rte_eth_dev_info *dev_info);
110 static int  eth_igb_flow_ctrl_get(struct rte_eth_dev *dev,
111 				struct rte_eth_fc_conf *fc_conf);
112 static int  eth_igb_flow_ctrl_set(struct rte_eth_dev *dev,
113 				struct rte_eth_fc_conf *fc_conf);
114 static int eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
115 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev);
116 static int eth_igb_interrupt_get_status(struct rte_eth_dev *dev);
117 static int eth_igb_interrupt_action(struct rte_eth_dev *dev,
118 				    struct rte_intr_handle *handle);
119 static void eth_igb_interrupt_handler(void *param);
120 static int  igb_hardware_init(struct e1000_hw *hw);
121 static void igb_hw_control_acquire(struct e1000_hw *hw);
122 static void igb_hw_control_release(struct e1000_hw *hw);
123 static void igb_init_manageability(struct e1000_hw *hw);
124 static void igb_release_manageability(struct e1000_hw *hw);
125 
126 static int  eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
127 
128 static int eth_igb_vlan_filter_set(struct rte_eth_dev *dev,
129 		uint16_t vlan_id, int on);
130 static int eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
131 				 enum rte_vlan_type vlan_type,
132 				 uint16_t tpid_id);
133 static int eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask);
134 
135 static void igb_vlan_hw_filter_enable(struct rte_eth_dev *dev);
136 static void igb_vlan_hw_filter_disable(struct rte_eth_dev *dev);
137 static void igb_vlan_hw_strip_enable(struct rte_eth_dev *dev);
138 static void igb_vlan_hw_strip_disable(struct rte_eth_dev *dev);
139 static void igb_vlan_hw_extend_enable(struct rte_eth_dev *dev);
140 static void igb_vlan_hw_extend_disable(struct rte_eth_dev *dev);
141 
142 static int eth_igb_led_on(struct rte_eth_dev *dev);
143 static int eth_igb_led_off(struct rte_eth_dev *dev);
144 
145 static void igb_intr_disable(struct rte_eth_dev *dev);
146 static int  igb_get_rx_buffer_size(struct e1000_hw *hw);
147 static int eth_igb_rar_set(struct rte_eth_dev *dev,
148 			   struct rte_ether_addr *mac_addr,
149 			   uint32_t index, uint32_t pool);
150 static void eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index);
151 static int eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
152 		struct rte_ether_addr *addr);
153 
154 static void igbvf_intr_disable(struct e1000_hw *hw);
155 static int igbvf_dev_configure(struct rte_eth_dev *dev);
156 static int igbvf_dev_start(struct rte_eth_dev *dev);
157 static int igbvf_dev_stop(struct rte_eth_dev *dev);
158 static int igbvf_dev_close(struct rte_eth_dev *dev);
159 static int igbvf_promiscuous_enable(struct rte_eth_dev *dev);
160 static int igbvf_promiscuous_disable(struct rte_eth_dev *dev);
161 static int igbvf_allmulticast_enable(struct rte_eth_dev *dev);
162 static int igbvf_allmulticast_disable(struct rte_eth_dev *dev);
163 static int eth_igbvf_link_update(struct e1000_hw *hw);
164 static int eth_igbvf_stats_get(struct rte_eth_dev *dev,
165 				struct rte_eth_stats *rte_stats);
166 static int eth_igbvf_xstats_get(struct rte_eth_dev *dev,
167 				struct rte_eth_xstat *xstats, unsigned n);
168 static int eth_igbvf_xstats_get_names(struct rte_eth_dev *dev,
169 				      struct rte_eth_xstat_name *xstats_names,
170 				      unsigned limit);
171 static int eth_igbvf_stats_reset(struct rte_eth_dev *dev);
172 static int igbvf_vlan_filter_set(struct rte_eth_dev *dev,
173 		uint16_t vlan_id, int on);
174 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on);
175 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on);
176 static int igbvf_default_mac_addr_set(struct rte_eth_dev *dev,
177 		struct rte_ether_addr *addr);
178 static int igbvf_get_reg_length(struct rte_eth_dev *dev);
179 static int igbvf_get_regs(struct rte_eth_dev *dev,
180 		struct rte_dev_reg_info *regs);
181 
182 static int eth_igb_rss_reta_update(struct rte_eth_dev *dev,
183 				   struct rte_eth_rss_reta_entry64 *reta_conf,
184 				   uint16_t reta_size);
185 static int eth_igb_rss_reta_query(struct rte_eth_dev *dev,
186 				  struct rte_eth_rss_reta_entry64 *reta_conf,
187 				  uint16_t reta_size);
188 
189 static int igb_add_2tuple_filter(struct rte_eth_dev *dev,
190 			struct rte_eth_ntuple_filter *ntuple_filter);
191 static int igb_remove_2tuple_filter(struct rte_eth_dev *dev,
192 			struct rte_eth_ntuple_filter *ntuple_filter);
193 static int igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
194 			struct rte_eth_ntuple_filter *ntuple_filter);
195 static int igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
196 			struct rte_eth_ntuple_filter *ntuple_filter);
197 static int eth_igb_filter_ctrl(struct rte_eth_dev *dev,
198 		     enum rte_filter_type filter_type,
199 		     enum rte_filter_op filter_op,
200 		     void *arg);
201 static int eth_igb_get_reg_length(struct rte_eth_dev *dev);
202 static int eth_igb_get_regs(struct rte_eth_dev *dev,
203 		struct rte_dev_reg_info *regs);
204 static int eth_igb_get_eeprom_length(struct rte_eth_dev *dev);
205 static int eth_igb_get_eeprom(struct rte_eth_dev *dev,
206 		struct rte_dev_eeprom_info *eeprom);
207 static int eth_igb_set_eeprom(struct rte_eth_dev *dev,
208 		struct rte_dev_eeprom_info *eeprom);
209 static int eth_igb_get_module_info(struct rte_eth_dev *dev,
210 				   struct rte_eth_dev_module_info *modinfo);
211 static int eth_igb_get_module_eeprom(struct rte_eth_dev *dev,
212 				     struct rte_dev_eeprom_info *info);
213 static int eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
214 				    struct rte_ether_addr *mc_addr_set,
215 				    uint32_t nb_mc_addr);
216 static int igb_timesync_enable(struct rte_eth_dev *dev);
217 static int igb_timesync_disable(struct rte_eth_dev *dev);
218 static int igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
219 					  struct timespec *timestamp,
220 					  uint32_t flags);
221 static int igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
222 					  struct timespec *timestamp);
223 static int igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
224 static int igb_timesync_read_time(struct rte_eth_dev *dev,
225 				  struct timespec *timestamp);
226 static int igb_timesync_write_time(struct rte_eth_dev *dev,
227 				   const struct timespec *timestamp);
228 static int eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev,
229 					uint16_t queue_id);
230 static int eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev,
231 					 uint16_t queue_id);
232 static void eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
233 				       uint8_t queue, uint8_t msix_vector);
234 static void eth_igb_write_ivar(struct e1000_hw *hw, uint8_t msix_vector,
235 			       uint8_t index, uint8_t offset);
236 static void eth_igb_configure_msix_intr(struct rte_eth_dev *dev);
237 static void eth_igbvf_interrupt_handler(void *param);
238 static void igbvf_mbx_process(struct rte_eth_dev *dev);
239 static int igb_filter_restore(struct rte_eth_dev *dev);
240 
241 /*
242  * Define VF Stats MACRO for Non "cleared on read" register
243  */
244 #define UPDATE_VF_STAT(reg, last, cur)            \
245 {                                                 \
246 	u32 latest = E1000_READ_REG(hw, reg);     \
247 	cur += (latest - last) & UINT_MAX;        \
248 	last = latest;                            \
249 }
250 
251 #define IGB_FC_PAUSE_TIME 0x0680
252 #define IGB_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
253 #define IGB_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
254 
255 #define IGBVF_PMD_NAME "rte_igbvf_pmd"     /* PMD name */
256 
257 static enum e1000_fc_mode igb_fc_setting = e1000_fc_full;
258 
259 /*
260  * The set of PCI devices this driver supports
261  */
262 static const struct rte_pci_id pci_id_igb_map[] = {
263 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576) },
264 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_FIBER) },
265 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_SERDES) },
266 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_QUAD_COPPER) },
267 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_QUAD_COPPER_ET2) },
268 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_NS) },
269 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_NS_SERDES) },
270 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_SERDES_QUAD) },
271 
272 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575EB_COPPER) },
273 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575EB_FIBER_SERDES) },
274 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575GB_QUAD_COPPER) },
275 
276 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_COPPER) },
277 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_FIBER) },
278 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_SERDES) },
279 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_SGMII) },
280 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_COPPER_DUAL) },
281 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_QUAD_FIBER) },
282 
283 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_COPPER) },
284 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_FIBER) },
285 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_SERDES) },
286 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_SGMII) },
287 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_DA4) },
288 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER) },
289 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_OEM1) },
290 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_IT) },
291 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_FIBER) },
292 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SERDES) },
293 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SGMII) },
294 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_FLASHLESS) },
295 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SERDES_FLASHLESS) },
296 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I211_COPPER) },
297 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
298 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_SGMII) },
299 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
300 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SGMII) },
301 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SERDES) },
302 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_BACKPLANE) },
303 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SFP) },
304 	{ .vendor_id = 0, /* sentinel */ },
305 };
306 
307 /*
308  * The set of PCI devices this driver supports (for 82576&I350 VF)
309  */
310 static const struct rte_pci_id pci_id_igbvf_map[] = {
311 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_VF) },
312 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_VF_HV) },
313 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_VF) },
314 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_VF_HV) },
315 	{ .vendor_id = 0, /* sentinel */ },
316 };
317 
318 static const struct rte_eth_desc_lim rx_desc_lim = {
319 	.nb_max = E1000_MAX_RING_DESC,
320 	.nb_min = E1000_MIN_RING_DESC,
321 	.nb_align = IGB_RXD_ALIGN,
322 };
323 
324 static const struct rte_eth_desc_lim tx_desc_lim = {
325 	.nb_max = E1000_MAX_RING_DESC,
326 	.nb_min = E1000_MIN_RING_DESC,
327 	.nb_align = IGB_RXD_ALIGN,
328 	.nb_seg_max = IGB_TX_MAX_SEG,
329 	.nb_mtu_seg_max = IGB_TX_MAX_MTU_SEG,
330 };
331 
332 static const struct eth_dev_ops eth_igb_ops = {
333 	.dev_configure        = eth_igb_configure,
334 	.dev_start            = eth_igb_start,
335 	.dev_stop             = eth_igb_stop,
336 	.dev_set_link_up      = eth_igb_dev_set_link_up,
337 	.dev_set_link_down    = eth_igb_dev_set_link_down,
338 	.dev_close            = eth_igb_close,
339 	.dev_reset            = eth_igb_reset,
340 	.promiscuous_enable   = eth_igb_promiscuous_enable,
341 	.promiscuous_disable  = eth_igb_promiscuous_disable,
342 	.allmulticast_enable  = eth_igb_allmulticast_enable,
343 	.allmulticast_disable = eth_igb_allmulticast_disable,
344 	.link_update          = eth_igb_link_update,
345 	.stats_get            = eth_igb_stats_get,
346 	.xstats_get           = eth_igb_xstats_get,
347 	.xstats_get_by_id     = eth_igb_xstats_get_by_id,
348 	.xstats_get_names_by_id = eth_igb_xstats_get_names_by_id,
349 	.xstats_get_names     = eth_igb_xstats_get_names,
350 	.stats_reset          = eth_igb_stats_reset,
351 	.xstats_reset         = eth_igb_xstats_reset,
352 	.fw_version_get       = eth_igb_fw_version_get,
353 	.dev_infos_get        = eth_igb_infos_get,
354 	.dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
355 	.mtu_set              = eth_igb_mtu_set,
356 	.vlan_filter_set      = eth_igb_vlan_filter_set,
357 	.vlan_tpid_set        = eth_igb_vlan_tpid_set,
358 	.vlan_offload_set     = eth_igb_vlan_offload_set,
359 	.rx_queue_setup       = eth_igb_rx_queue_setup,
360 	.rx_queue_intr_enable = eth_igb_rx_queue_intr_enable,
361 	.rx_queue_intr_disable = eth_igb_rx_queue_intr_disable,
362 	.rx_queue_release     = eth_igb_rx_queue_release,
363 	.tx_queue_setup       = eth_igb_tx_queue_setup,
364 	.tx_queue_release     = eth_igb_tx_queue_release,
365 	.tx_done_cleanup      = eth_igb_tx_done_cleanup,
366 	.dev_led_on           = eth_igb_led_on,
367 	.dev_led_off          = eth_igb_led_off,
368 	.flow_ctrl_get        = eth_igb_flow_ctrl_get,
369 	.flow_ctrl_set        = eth_igb_flow_ctrl_set,
370 	.mac_addr_add         = eth_igb_rar_set,
371 	.mac_addr_remove      = eth_igb_rar_clear,
372 	.mac_addr_set         = eth_igb_default_mac_addr_set,
373 	.reta_update          = eth_igb_rss_reta_update,
374 	.reta_query           = eth_igb_rss_reta_query,
375 	.rss_hash_update      = eth_igb_rss_hash_update,
376 	.rss_hash_conf_get    = eth_igb_rss_hash_conf_get,
377 	.filter_ctrl          = eth_igb_filter_ctrl,
378 	.set_mc_addr_list     = eth_igb_set_mc_addr_list,
379 	.rxq_info_get         = igb_rxq_info_get,
380 	.txq_info_get         = igb_txq_info_get,
381 	.timesync_enable      = igb_timesync_enable,
382 	.timesync_disable     = igb_timesync_disable,
383 	.timesync_read_rx_timestamp = igb_timesync_read_rx_timestamp,
384 	.timesync_read_tx_timestamp = igb_timesync_read_tx_timestamp,
385 	.get_reg              = eth_igb_get_regs,
386 	.get_eeprom_length    = eth_igb_get_eeprom_length,
387 	.get_eeprom           = eth_igb_get_eeprom,
388 	.set_eeprom           = eth_igb_set_eeprom,
389 	.get_module_info      = eth_igb_get_module_info,
390 	.get_module_eeprom    = eth_igb_get_module_eeprom,
391 	.timesync_adjust_time = igb_timesync_adjust_time,
392 	.timesync_read_time   = igb_timesync_read_time,
393 	.timesync_write_time  = igb_timesync_write_time,
394 };
395 
396 /*
397  * dev_ops for virtual function, bare necessities for basic vf
398  * operation have been implemented
399  */
400 static const struct eth_dev_ops igbvf_eth_dev_ops = {
401 	.dev_configure        = igbvf_dev_configure,
402 	.dev_start            = igbvf_dev_start,
403 	.dev_stop             = igbvf_dev_stop,
404 	.dev_close            = igbvf_dev_close,
405 	.promiscuous_enable   = igbvf_promiscuous_enable,
406 	.promiscuous_disable  = igbvf_promiscuous_disable,
407 	.allmulticast_enable  = igbvf_allmulticast_enable,
408 	.allmulticast_disable = igbvf_allmulticast_disable,
409 	.link_update          = eth_igb_link_update,
410 	.stats_get            = eth_igbvf_stats_get,
411 	.xstats_get           = eth_igbvf_xstats_get,
412 	.xstats_get_names     = eth_igbvf_xstats_get_names,
413 	.stats_reset          = eth_igbvf_stats_reset,
414 	.xstats_reset         = eth_igbvf_stats_reset,
415 	.vlan_filter_set      = igbvf_vlan_filter_set,
416 	.dev_infos_get        = eth_igbvf_infos_get,
417 	.dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
418 	.rx_queue_setup       = eth_igb_rx_queue_setup,
419 	.rx_queue_release     = eth_igb_rx_queue_release,
420 	.tx_queue_setup       = eth_igb_tx_queue_setup,
421 	.tx_queue_release     = eth_igb_tx_queue_release,
422 	.tx_done_cleanup      = eth_igb_tx_done_cleanup,
423 	.set_mc_addr_list     = eth_igb_set_mc_addr_list,
424 	.rxq_info_get         = igb_rxq_info_get,
425 	.txq_info_get         = igb_txq_info_get,
426 	.mac_addr_set         = igbvf_default_mac_addr_set,
427 	.get_reg              = igbvf_get_regs,
428 };
429 
430 /* store statistics names and its offset in stats structure */
431 struct rte_igb_xstats_name_off {
432 	char name[RTE_ETH_XSTATS_NAME_SIZE];
433 	unsigned offset;
434 };
435 
436 static const struct rte_igb_xstats_name_off rte_igb_stats_strings[] = {
437 	{"rx_crc_errors", offsetof(struct e1000_hw_stats, crcerrs)},
438 	{"rx_align_errors", offsetof(struct e1000_hw_stats, algnerrc)},
439 	{"rx_symbol_errors", offsetof(struct e1000_hw_stats, symerrs)},
440 	{"rx_missed_packets", offsetof(struct e1000_hw_stats, mpc)},
441 	{"tx_single_collision_packets", offsetof(struct e1000_hw_stats, scc)},
442 	{"tx_multiple_collision_packets", offsetof(struct e1000_hw_stats, mcc)},
443 	{"tx_excessive_collision_packets", offsetof(struct e1000_hw_stats,
444 		ecol)},
445 	{"tx_late_collisions", offsetof(struct e1000_hw_stats, latecol)},
446 	{"tx_total_collisions", offsetof(struct e1000_hw_stats, colc)},
447 	{"tx_deferred_packets", offsetof(struct e1000_hw_stats, dc)},
448 	{"tx_no_carrier_sense_packets", offsetof(struct e1000_hw_stats, tncrs)},
449 	{"rx_carrier_ext_errors", offsetof(struct e1000_hw_stats, cexterr)},
450 	{"rx_length_errors", offsetof(struct e1000_hw_stats, rlec)},
451 	{"rx_xon_packets", offsetof(struct e1000_hw_stats, xonrxc)},
452 	{"tx_xon_packets", offsetof(struct e1000_hw_stats, xontxc)},
453 	{"rx_xoff_packets", offsetof(struct e1000_hw_stats, xoffrxc)},
454 	{"tx_xoff_packets", offsetof(struct e1000_hw_stats, xofftxc)},
455 	{"rx_flow_control_unsupported_packets", offsetof(struct e1000_hw_stats,
456 		fcruc)},
457 	{"rx_size_64_packets", offsetof(struct e1000_hw_stats, prc64)},
458 	{"rx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, prc127)},
459 	{"rx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, prc255)},
460 	{"rx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, prc511)},
461 	{"rx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
462 		prc1023)},
463 	{"rx_size_1024_to_max_packets", offsetof(struct e1000_hw_stats,
464 		prc1522)},
465 	{"rx_broadcast_packets", offsetof(struct e1000_hw_stats, bprc)},
466 	{"rx_multicast_packets", offsetof(struct e1000_hw_stats, mprc)},
467 	{"rx_undersize_errors", offsetof(struct e1000_hw_stats, ruc)},
468 	{"rx_fragment_errors", offsetof(struct e1000_hw_stats, rfc)},
469 	{"rx_oversize_errors", offsetof(struct e1000_hw_stats, roc)},
470 	{"rx_jabber_errors", offsetof(struct e1000_hw_stats, rjc)},
471 	{"rx_management_packets", offsetof(struct e1000_hw_stats, mgprc)},
472 	{"rx_management_dropped", offsetof(struct e1000_hw_stats, mgpdc)},
473 	{"tx_management_packets", offsetof(struct e1000_hw_stats, mgptc)},
474 	{"rx_total_packets", offsetof(struct e1000_hw_stats, tpr)},
475 	{"tx_total_packets", offsetof(struct e1000_hw_stats, tpt)},
476 	{"rx_total_bytes", offsetof(struct e1000_hw_stats, tor)},
477 	{"tx_total_bytes", offsetof(struct e1000_hw_stats, tot)},
478 	{"tx_size_64_packets", offsetof(struct e1000_hw_stats, ptc64)},
479 	{"tx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, ptc127)},
480 	{"tx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, ptc255)},
481 	{"tx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, ptc511)},
482 	{"tx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
483 		ptc1023)},
484 	{"tx_size_1023_to_max_packets", offsetof(struct e1000_hw_stats,
485 		ptc1522)},
486 	{"tx_multicast_packets", offsetof(struct e1000_hw_stats, mptc)},
487 	{"tx_broadcast_packets", offsetof(struct e1000_hw_stats, bptc)},
488 	{"tx_tso_packets", offsetof(struct e1000_hw_stats, tsctc)},
489 	{"tx_tso_errors", offsetof(struct e1000_hw_stats, tsctfc)},
490 	{"rx_sent_to_host_packets", offsetof(struct e1000_hw_stats, rpthc)},
491 	{"tx_sent_by_host_packets", offsetof(struct e1000_hw_stats, hgptc)},
492 	{"rx_code_violation_packets", offsetof(struct e1000_hw_stats, scvpc)},
493 
494 	{"interrupt_assert_count", offsetof(struct e1000_hw_stats, iac)},
495 };
496 
497 #define IGB_NB_XSTATS (sizeof(rte_igb_stats_strings) / \
498 		sizeof(rte_igb_stats_strings[0]))
499 
500 static const struct rte_igb_xstats_name_off rte_igbvf_stats_strings[] = {
501 	{"rx_multicast_packets", offsetof(struct e1000_vf_stats, mprc)},
502 	{"rx_good_loopback_packets", offsetof(struct e1000_vf_stats, gprlbc)},
503 	{"tx_good_loopback_packets", offsetof(struct e1000_vf_stats, gptlbc)},
504 	{"rx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gorlbc)},
505 	{"tx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gotlbc)},
506 };
507 
508 #define IGBVF_NB_XSTATS (sizeof(rte_igbvf_stats_strings) / \
509 		sizeof(rte_igbvf_stats_strings[0]))
510 
511 
512 static inline void
513 igb_intr_enable(struct rte_eth_dev *dev)
514 {
515 	struct e1000_interrupt *intr =
516 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
517 	struct e1000_hw *hw =
518 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
519 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
520 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
521 
522 	if (rte_intr_allow_others(intr_handle) &&
523 		dev->data->dev_conf.intr_conf.lsc != 0) {
524 		E1000_WRITE_REG(hw, E1000_EIMS, 1 << IGB_MSIX_OTHER_INTR_VEC);
525 	}
526 
527 	E1000_WRITE_REG(hw, E1000_IMS, intr->mask);
528 	E1000_WRITE_FLUSH(hw);
529 }
530 
531 static void
532 igb_intr_disable(struct rte_eth_dev *dev)
533 {
534 	struct e1000_hw *hw =
535 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
536 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
537 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
538 
539 	if (rte_intr_allow_others(intr_handle) &&
540 		dev->data->dev_conf.intr_conf.lsc != 0) {
541 		E1000_WRITE_REG(hw, E1000_EIMC, 1 << IGB_MSIX_OTHER_INTR_VEC);
542 	}
543 
544 	E1000_WRITE_REG(hw, E1000_IMC, ~0);
545 	E1000_WRITE_FLUSH(hw);
546 }
547 
548 static inline void
549 igbvf_intr_enable(struct rte_eth_dev *dev)
550 {
551 	struct e1000_hw *hw =
552 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
553 
554 	/* only for mailbox */
555 	E1000_WRITE_REG(hw, E1000_EIAM, 1 << E1000_VTIVAR_MISC_MAILBOX);
556 	E1000_WRITE_REG(hw, E1000_EIAC, 1 << E1000_VTIVAR_MISC_MAILBOX);
557 	E1000_WRITE_REG(hw, E1000_EIMS, 1 << E1000_VTIVAR_MISC_MAILBOX);
558 	E1000_WRITE_FLUSH(hw);
559 }
560 
561 /* only for mailbox now. If RX/TX needed, should extend this function.  */
562 static void
563 igbvf_set_ivar_map(struct e1000_hw *hw, uint8_t msix_vector)
564 {
565 	uint32_t tmp = 0;
566 
567 	/* mailbox */
568 	tmp |= (msix_vector & E1000_VTIVAR_MISC_INTR_MASK);
569 	tmp |= E1000_VTIVAR_VALID;
570 	E1000_WRITE_REG(hw, E1000_VTIVAR_MISC, tmp);
571 }
572 
573 static void
574 eth_igbvf_configure_msix_intr(struct rte_eth_dev *dev)
575 {
576 	struct e1000_hw *hw =
577 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
578 
579 	/* Configure VF other cause ivar */
580 	igbvf_set_ivar_map(hw, E1000_VTIVAR_MISC_MAILBOX);
581 }
582 
583 static inline int32_t
584 igb_pf_reset_hw(struct e1000_hw *hw)
585 {
586 	uint32_t ctrl_ext;
587 	int32_t status;
588 
589 	status = e1000_reset_hw(hw);
590 
591 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
592 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
593 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
594 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
595 	E1000_WRITE_FLUSH(hw);
596 
597 	return status;
598 }
599 
600 static void
601 igb_identify_hardware(struct rte_eth_dev *dev, struct rte_pci_device *pci_dev)
602 {
603 	struct e1000_hw *hw =
604 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
605 
606 
607 	hw->vendor_id = pci_dev->id.vendor_id;
608 	hw->device_id = pci_dev->id.device_id;
609 	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
610 	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
611 
612 	e1000_set_mac_type(hw);
613 
614 	/* need to check if it is a vf device below */
615 }
616 
617 static int
618 igb_reset_swfw_lock(struct e1000_hw *hw)
619 {
620 	int ret_val;
621 
622 	/*
623 	 * Do mac ops initialization manually here, since we will need
624 	 * some function pointers set by this call.
625 	 */
626 	ret_val = e1000_init_mac_params(hw);
627 	if (ret_val)
628 		return ret_val;
629 
630 	/*
631 	 * SMBI lock should not fail in this early stage. If this is the case,
632 	 * it is due to an improper exit of the application.
633 	 * So force the release of the faulty lock.
634 	 */
635 	if (e1000_get_hw_semaphore_generic(hw) < 0) {
636 		PMD_DRV_LOG(DEBUG, "SMBI lock released");
637 	}
638 	e1000_put_hw_semaphore_generic(hw);
639 
640 	if (hw->mac.ops.acquire_swfw_sync != NULL) {
641 		uint16_t mask;
642 
643 		/*
644 		 * Phy lock should not fail in this early stage. If this is the case,
645 		 * it is due to an improper exit of the application.
646 		 * So force the release of the faulty lock.
647 		 */
648 		mask = E1000_SWFW_PHY0_SM << hw->bus.func;
649 		if (hw->bus.func > E1000_FUNC_1)
650 			mask <<= 2;
651 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
652 			PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released",
653 				    hw->bus.func);
654 		}
655 		hw->mac.ops.release_swfw_sync(hw, mask);
656 
657 		/*
658 		 * This one is more tricky since it is common to all ports; but
659 		 * swfw_sync retries last long enough (1s) to be almost sure that if
660 		 * lock can not be taken it is due to an improper lock of the
661 		 * semaphore.
662 		 */
663 		mask = E1000_SWFW_EEP_SM;
664 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
665 			PMD_DRV_LOG(DEBUG, "SWFW common locks released");
666 		}
667 		hw->mac.ops.release_swfw_sync(hw, mask);
668 	}
669 
670 	return E1000_SUCCESS;
671 }
672 
673 /* Remove all ntuple filters of the device */
674 static int igb_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
675 {
676 	struct e1000_filter_info *filter_info =
677 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
678 	struct e1000_5tuple_filter *p_5tuple;
679 	struct e1000_2tuple_filter *p_2tuple;
680 
681 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
682 		TAILQ_REMOVE(&filter_info->fivetuple_list,
683 			p_5tuple, entries);
684 			rte_free(p_5tuple);
685 	}
686 	filter_info->fivetuple_mask = 0;
687 	while ((p_2tuple = TAILQ_FIRST(&filter_info->twotuple_list))) {
688 		TAILQ_REMOVE(&filter_info->twotuple_list,
689 			p_2tuple, entries);
690 			rte_free(p_2tuple);
691 	}
692 	filter_info->twotuple_mask = 0;
693 
694 	return 0;
695 }
696 
697 /* Remove all flex filters of the device */
698 static int igb_flex_filter_uninit(struct rte_eth_dev *eth_dev)
699 {
700 	struct e1000_filter_info *filter_info =
701 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
702 	struct e1000_flex_filter *p_flex;
703 
704 	while ((p_flex = TAILQ_FIRST(&filter_info->flex_list))) {
705 		TAILQ_REMOVE(&filter_info->flex_list, p_flex, entries);
706 		rte_free(p_flex);
707 	}
708 	filter_info->flex_mask = 0;
709 
710 	return 0;
711 }
712 
713 static int
714 eth_igb_dev_init(struct rte_eth_dev *eth_dev)
715 {
716 	int error = 0;
717 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
718 	struct e1000_hw *hw =
719 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
720 	struct e1000_vfta * shadow_vfta =
721 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
722 	struct e1000_filter_info *filter_info =
723 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
724 	struct e1000_adapter *adapter =
725 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
726 
727 	uint32_t ctrl_ext;
728 
729 	eth_dev->dev_ops = &eth_igb_ops;
730 	eth_dev->rx_queue_count = eth_igb_rx_queue_count;
731 	eth_dev->rx_descriptor_done   = eth_igb_rx_descriptor_done;
732 	eth_dev->rx_descriptor_status = eth_igb_rx_descriptor_status;
733 	eth_dev->tx_descriptor_status = eth_igb_tx_descriptor_status;
734 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
735 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
736 	eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
737 
738 	/* for secondary processes, we don't initialise any further as primary
739 	 * has already done this work. Only check we don't need a different
740 	 * RX function */
741 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
742 		if (eth_dev->data->scattered_rx)
743 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
744 		return 0;
745 	}
746 
747 	rte_eth_copy_pci_info(eth_dev, pci_dev);
748 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
749 
750 	hw->hw_addr= (void *)pci_dev->mem_resource[0].addr;
751 
752 	igb_identify_hardware(eth_dev, pci_dev);
753 	if (e1000_setup_init_funcs(hw, FALSE) != E1000_SUCCESS) {
754 		error = -EIO;
755 		goto err_late;
756 	}
757 
758 	e1000_get_bus_info(hw);
759 
760 	/* Reset any pending lock */
761 	if (igb_reset_swfw_lock(hw) != E1000_SUCCESS) {
762 		error = -EIO;
763 		goto err_late;
764 	}
765 
766 	/* Finish initialization */
767 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS) {
768 		error = -EIO;
769 		goto err_late;
770 	}
771 
772 	hw->mac.autoneg = 1;
773 	hw->phy.autoneg_wait_to_complete = 0;
774 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
775 
776 	/* Copper options */
777 	if (hw->phy.media_type == e1000_media_type_copper) {
778 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
779 		hw->phy.disable_polarity_correction = 0;
780 		hw->phy.ms_type = e1000_ms_hw_default;
781 	}
782 
783 	/*
784 	 * Start from a known state, this is important in reading the nvm
785 	 * and mac from that.
786 	 */
787 	igb_pf_reset_hw(hw);
788 
789 	/* Make sure we have a good EEPROM before we read from it */
790 	if (e1000_validate_nvm_checksum(hw) < 0) {
791 		/*
792 		 * Some PCI-E parts fail the first check due to
793 		 * the link being in sleep state, call it again,
794 		 * if it fails a second time its a real issue.
795 		 */
796 		if (e1000_validate_nvm_checksum(hw) < 0) {
797 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
798 			error = -EIO;
799 			goto err_late;
800 		}
801 	}
802 
803 	/* Read the permanent MAC address out of the EEPROM */
804 	if (e1000_read_mac_addr(hw) != 0) {
805 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
806 		error = -EIO;
807 		goto err_late;
808 	}
809 
810 	/* Allocate memory for storing MAC addresses */
811 	eth_dev->data->mac_addrs = rte_zmalloc("e1000",
812 		RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count, 0);
813 	if (eth_dev->data->mac_addrs == NULL) {
814 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
815 						"store MAC addresses",
816 				RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
817 		error = -ENOMEM;
818 		goto err_late;
819 	}
820 
821 	/* Copy the permanent MAC address */
822 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
823 			&eth_dev->data->mac_addrs[0]);
824 
825 	/* initialize the vfta */
826 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
827 
828 	/* Now initialize the hardware */
829 	if (igb_hardware_init(hw) != 0) {
830 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
831 		rte_free(eth_dev->data->mac_addrs);
832 		eth_dev->data->mac_addrs = NULL;
833 		error = -ENODEV;
834 		goto err_late;
835 	}
836 	hw->mac.get_link_status = 1;
837 	adapter->stopped = 0;
838 
839 	/* Indicate SOL/IDER usage */
840 	if (e1000_check_reset_block(hw) < 0) {
841 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to"
842 					"SOL/IDER session");
843 	}
844 
845 	/* initialize PF if max_vfs not zero */
846 	igb_pf_host_init(eth_dev);
847 
848 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
849 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
850 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
851 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
852 	E1000_WRITE_FLUSH(hw);
853 
854 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
855 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
856 		     pci_dev->id.device_id);
857 
858 	rte_intr_callback_register(&pci_dev->intr_handle,
859 				   eth_igb_interrupt_handler,
860 				   (void *)eth_dev);
861 
862 	/* enable uio/vfio intr/eventfd mapping */
863 	rte_intr_enable(&pci_dev->intr_handle);
864 
865 	/* enable support intr */
866 	igb_intr_enable(eth_dev);
867 
868 	eth_igb_dev_set_link_down(eth_dev);
869 
870 	/* initialize filter info */
871 	memset(filter_info, 0,
872 	       sizeof(struct e1000_filter_info));
873 
874 	TAILQ_INIT(&filter_info->flex_list);
875 	TAILQ_INIT(&filter_info->twotuple_list);
876 	TAILQ_INIT(&filter_info->fivetuple_list);
877 
878 	TAILQ_INIT(&igb_filter_ntuple_list);
879 	TAILQ_INIT(&igb_filter_ethertype_list);
880 	TAILQ_INIT(&igb_filter_syn_list);
881 	TAILQ_INIT(&igb_filter_flex_list);
882 	TAILQ_INIT(&igb_filter_rss_list);
883 	TAILQ_INIT(&igb_flow_list);
884 
885 	return 0;
886 
887 err_late:
888 	igb_hw_control_release(hw);
889 
890 	return error;
891 }
892 
893 static int
894 eth_igb_dev_uninit(struct rte_eth_dev *eth_dev)
895 {
896 	PMD_INIT_FUNC_TRACE();
897 
898 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
899 		return 0;
900 
901 	eth_igb_close(eth_dev);
902 
903 	return 0;
904 }
905 
906 /*
907  * Virtual Function device init
908  */
909 static int
910 eth_igbvf_dev_init(struct rte_eth_dev *eth_dev)
911 {
912 	struct rte_pci_device *pci_dev;
913 	struct rte_intr_handle *intr_handle;
914 	struct e1000_adapter *adapter =
915 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
916 	struct e1000_hw *hw =
917 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
918 	int diag;
919 	struct rte_ether_addr *perm_addr =
920 		(struct rte_ether_addr *)hw->mac.perm_addr;
921 
922 	PMD_INIT_FUNC_TRACE();
923 
924 	eth_dev->dev_ops = &igbvf_eth_dev_ops;
925 	eth_dev->rx_descriptor_done   = eth_igb_rx_descriptor_done;
926 	eth_dev->rx_descriptor_status = eth_igb_rx_descriptor_status;
927 	eth_dev->tx_descriptor_status = eth_igb_tx_descriptor_status;
928 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
929 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
930 	eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
931 
932 	/* for secondary processes, we don't initialise any further as primary
933 	 * has already done this work. Only check we don't need a different
934 	 * RX function */
935 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
936 		if (eth_dev->data->scattered_rx)
937 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
938 		return 0;
939 	}
940 
941 	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
942 	rte_eth_copy_pci_info(eth_dev, pci_dev);
943 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
944 
945 	hw->device_id = pci_dev->id.device_id;
946 	hw->vendor_id = pci_dev->id.vendor_id;
947 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
948 	adapter->stopped = 0;
949 
950 	/* Initialize the shared code (base driver) */
951 	diag = e1000_setup_init_funcs(hw, TRUE);
952 	if (diag != 0) {
953 		PMD_INIT_LOG(ERR, "Shared code init failed for igbvf: %d",
954 			diag);
955 		return -EIO;
956 	}
957 
958 	/* init_mailbox_params */
959 	hw->mbx.ops.init_params(hw);
960 
961 	/* Disable the interrupts for VF */
962 	igbvf_intr_disable(hw);
963 
964 	diag = hw->mac.ops.reset_hw(hw);
965 
966 	/* Allocate memory for storing MAC addresses */
967 	eth_dev->data->mac_addrs = rte_zmalloc("igbvf", RTE_ETHER_ADDR_LEN *
968 		hw->mac.rar_entry_count, 0);
969 	if (eth_dev->data->mac_addrs == NULL) {
970 		PMD_INIT_LOG(ERR,
971 			"Failed to allocate %d bytes needed to store MAC "
972 			"addresses",
973 			RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
974 		return -ENOMEM;
975 	}
976 
977 	/* Generate a random MAC address, if none was assigned by PF. */
978 	if (rte_is_zero_ether_addr(perm_addr)) {
979 		rte_eth_random_addr(perm_addr->addr_bytes);
980 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
981 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
982 			     "%02x:%02x:%02x:%02x:%02x:%02x",
983 			     perm_addr->addr_bytes[0],
984 			     perm_addr->addr_bytes[1],
985 			     perm_addr->addr_bytes[2],
986 			     perm_addr->addr_bytes[3],
987 			     perm_addr->addr_bytes[4],
988 			     perm_addr->addr_bytes[5]);
989 	}
990 
991 	diag = e1000_rar_set(hw, perm_addr->addr_bytes, 0);
992 	if (diag) {
993 		rte_free(eth_dev->data->mac_addrs);
994 		eth_dev->data->mac_addrs = NULL;
995 		return diag;
996 	}
997 	/* Copy the permanent MAC address */
998 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.perm_addr,
999 			&eth_dev->data->mac_addrs[0]);
1000 
1001 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x "
1002 		     "mac.type=%s",
1003 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1004 		     pci_dev->id.device_id, "igb_mac_82576_vf");
1005 
1006 	intr_handle = &pci_dev->intr_handle;
1007 	rte_intr_callback_register(intr_handle,
1008 				   eth_igbvf_interrupt_handler, eth_dev);
1009 
1010 	return 0;
1011 }
1012 
1013 static int
1014 eth_igbvf_dev_uninit(struct rte_eth_dev *eth_dev)
1015 {
1016 	PMD_INIT_FUNC_TRACE();
1017 
1018 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1019 		return 0;
1020 
1021 	igbvf_dev_close(eth_dev);
1022 
1023 	return 0;
1024 }
1025 
1026 static int eth_igb_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1027 	struct rte_pci_device *pci_dev)
1028 {
1029 	return rte_eth_dev_pci_generic_probe(pci_dev,
1030 		sizeof(struct e1000_adapter), eth_igb_dev_init);
1031 }
1032 
1033 static int eth_igb_pci_remove(struct rte_pci_device *pci_dev)
1034 {
1035 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igb_dev_uninit);
1036 }
1037 
1038 static struct rte_pci_driver rte_igb_pmd = {
1039 	.id_table = pci_id_igb_map,
1040 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1041 	.probe = eth_igb_pci_probe,
1042 	.remove = eth_igb_pci_remove,
1043 };
1044 
1045 
1046 static int eth_igbvf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1047 	struct rte_pci_device *pci_dev)
1048 {
1049 	return rte_eth_dev_pci_generic_probe(pci_dev,
1050 		sizeof(struct e1000_adapter), eth_igbvf_dev_init);
1051 }
1052 
1053 static int eth_igbvf_pci_remove(struct rte_pci_device *pci_dev)
1054 {
1055 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igbvf_dev_uninit);
1056 }
1057 
1058 /*
1059  * virtual function driver struct
1060  */
1061 static struct rte_pci_driver rte_igbvf_pmd = {
1062 	.id_table = pci_id_igbvf_map,
1063 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1064 	.probe = eth_igbvf_pci_probe,
1065 	.remove = eth_igbvf_pci_remove,
1066 };
1067 
1068 static void
1069 igb_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1070 {
1071 	struct e1000_hw *hw =
1072 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1073 	/* RCTL: enable VLAN filter since VMDq always use VLAN filter */
1074 	uint32_t rctl = E1000_READ_REG(hw, E1000_RCTL);
1075 	rctl |= E1000_RCTL_VFE;
1076 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1077 }
1078 
1079 static int
1080 igb_check_mq_mode(struct rte_eth_dev *dev)
1081 {
1082 	enum rte_eth_rx_mq_mode rx_mq_mode = dev->data->dev_conf.rxmode.mq_mode;
1083 	enum rte_eth_tx_mq_mode tx_mq_mode = dev->data->dev_conf.txmode.mq_mode;
1084 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
1085 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
1086 
1087 	if ((rx_mq_mode & ETH_MQ_RX_DCB_FLAG) ||
1088 	    tx_mq_mode == ETH_MQ_TX_DCB ||
1089 	    tx_mq_mode == ETH_MQ_TX_VMDQ_DCB) {
1090 		PMD_INIT_LOG(ERR, "DCB mode is not supported.");
1091 		return -EINVAL;
1092 	}
1093 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
1094 		/* Check multi-queue mode.
1095 		 * To no break software we accept ETH_MQ_RX_NONE as this might
1096 		 * be used to turn off VLAN filter.
1097 		 */
1098 
1099 		if (rx_mq_mode == ETH_MQ_RX_NONE ||
1100 		    rx_mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1101 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
1102 			RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = 1;
1103 		} else {
1104 			/* Only support one queue on VFs.
1105 			 * RSS together with SRIOV is not supported.
1106 			 */
1107 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1108 					" wrong mq_mode rx %d.",
1109 					rx_mq_mode);
1110 			return -EINVAL;
1111 		}
1112 		/* TX mode is not used here, so mode might be ignored.*/
1113 		if (tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1114 			/* SRIOV only works in VMDq enable mode */
1115 			PMD_INIT_LOG(WARNING, "SRIOV is active,"
1116 					" TX mode %d is not supported. "
1117 					" Driver will behave as %d mode.",
1118 					tx_mq_mode, ETH_MQ_TX_VMDQ_ONLY);
1119 		}
1120 
1121 		/* check valid queue number */
1122 		if ((nb_rx_q > 1) || (nb_tx_q > 1)) {
1123 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1124 					" only support one queue on VFs.");
1125 			return -EINVAL;
1126 		}
1127 	} else {
1128 		/* To no break software that set invalid mode, only display
1129 		 * warning if invalid mode is used.
1130 		 */
1131 		if (rx_mq_mode != ETH_MQ_RX_NONE &&
1132 		    rx_mq_mode != ETH_MQ_RX_VMDQ_ONLY &&
1133 		    rx_mq_mode != ETH_MQ_RX_RSS) {
1134 			/* RSS together with VMDq not supported*/
1135 			PMD_INIT_LOG(ERR, "RX mode %d is not supported.",
1136 				     rx_mq_mode);
1137 			return -EINVAL;
1138 		}
1139 
1140 		if (tx_mq_mode != ETH_MQ_TX_NONE &&
1141 		    tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1142 			PMD_INIT_LOG(WARNING, "TX mode %d is not supported."
1143 					" Due to txmode is meaningless in this"
1144 					" driver, just ignore.",
1145 					tx_mq_mode);
1146 		}
1147 	}
1148 	return 0;
1149 }
1150 
1151 static int
1152 eth_igb_configure(struct rte_eth_dev *dev)
1153 {
1154 	struct e1000_interrupt *intr =
1155 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1156 	int ret;
1157 
1158 	PMD_INIT_FUNC_TRACE();
1159 
1160 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1161 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1162 
1163 	/* multipe queue mode checking */
1164 	ret  = igb_check_mq_mode(dev);
1165 	if (ret != 0) {
1166 		PMD_DRV_LOG(ERR, "igb_check_mq_mode fails with %d.",
1167 			    ret);
1168 		return ret;
1169 	}
1170 
1171 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1172 	PMD_INIT_FUNC_TRACE();
1173 
1174 	return 0;
1175 }
1176 
1177 static void
1178 eth_igb_rxtx_control(struct rte_eth_dev *dev,
1179 		     bool enable)
1180 {
1181 	struct e1000_hw *hw =
1182 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1183 	uint32_t tctl, rctl;
1184 
1185 	tctl = E1000_READ_REG(hw, E1000_TCTL);
1186 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1187 
1188 	if (enable) {
1189 		/* enable Tx/Rx */
1190 		tctl |= E1000_TCTL_EN;
1191 		rctl |= E1000_RCTL_EN;
1192 	} else {
1193 		/* disable Tx/Rx */
1194 		tctl &= ~E1000_TCTL_EN;
1195 		rctl &= ~E1000_RCTL_EN;
1196 	}
1197 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
1198 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1199 	E1000_WRITE_FLUSH(hw);
1200 }
1201 
1202 static int
1203 eth_igb_start(struct rte_eth_dev *dev)
1204 {
1205 	struct e1000_hw *hw =
1206 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1207 	struct e1000_adapter *adapter =
1208 		E1000_DEV_PRIVATE(dev->data->dev_private);
1209 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1210 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1211 	int ret, mask;
1212 	uint32_t intr_vector = 0;
1213 	uint32_t ctrl_ext;
1214 	uint32_t *speeds;
1215 	int num_speeds;
1216 	bool autoneg;
1217 
1218 	PMD_INIT_FUNC_TRACE();
1219 
1220 	/* disable uio/vfio intr/eventfd mapping */
1221 	rte_intr_disable(intr_handle);
1222 
1223 	/* Power up the phy. Needed to make the link go Up */
1224 	eth_igb_dev_set_link_up(dev);
1225 
1226 	/*
1227 	 * Packet Buffer Allocation (PBA)
1228 	 * Writing PBA sets the receive portion of the buffer
1229 	 * the remainder is used for the transmit buffer.
1230 	 */
1231 	if (hw->mac.type == e1000_82575) {
1232 		uint32_t pba;
1233 
1234 		pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
1235 		E1000_WRITE_REG(hw, E1000_PBA, pba);
1236 	}
1237 
1238 	/* Put the address into the Receive Address Array */
1239 	e1000_rar_set(hw, hw->mac.addr, 0);
1240 
1241 	/* Initialize the hardware */
1242 	if (igb_hardware_init(hw)) {
1243 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
1244 		return -EIO;
1245 	}
1246 	adapter->stopped = 0;
1247 
1248 	E1000_WRITE_REG(hw, E1000_VET,
1249 			RTE_ETHER_TYPE_VLAN << 16 | RTE_ETHER_TYPE_VLAN);
1250 
1251 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1252 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1253 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
1254 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
1255 	E1000_WRITE_FLUSH(hw);
1256 
1257 	/* configure PF module if SRIOV enabled */
1258 	igb_pf_host_configure(dev);
1259 
1260 	/* check and configure queue intr-vector mapping */
1261 	if ((rte_intr_cap_multiple(intr_handle) ||
1262 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
1263 	    dev->data->dev_conf.intr_conf.rxq != 0) {
1264 		intr_vector = dev->data->nb_rx_queues;
1265 		if (rte_intr_efd_enable(intr_handle, intr_vector))
1266 			return -1;
1267 	}
1268 
1269 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
1270 		intr_handle->intr_vec =
1271 			rte_zmalloc("intr_vec",
1272 				    dev->data->nb_rx_queues * sizeof(int), 0);
1273 		if (intr_handle->intr_vec == NULL) {
1274 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
1275 				     " intr_vec", dev->data->nb_rx_queues);
1276 			return -ENOMEM;
1277 		}
1278 	}
1279 
1280 	/* confiugre msix for rx interrupt */
1281 	eth_igb_configure_msix_intr(dev);
1282 
1283 	/* Configure for OS presence */
1284 	igb_init_manageability(hw);
1285 
1286 	eth_igb_tx_init(dev);
1287 
1288 	/* This can fail when allocating mbufs for descriptor rings */
1289 	ret = eth_igb_rx_init(dev);
1290 	if (ret) {
1291 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
1292 		igb_dev_clear_queues(dev);
1293 		return ret;
1294 	}
1295 
1296 	e1000_clear_hw_cntrs_base_generic(hw);
1297 
1298 	/*
1299 	 * VLAN Offload Settings
1300 	 */
1301 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
1302 			ETH_VLAN_EXTEND_MASK;
1303 	ret = eth_igb_vlan_offload_set(dev, mask);
1304 	if (ret) {
1305 		PMD_INIT_LOG(ERR, "Unable to set vlan offload");
1306 		igb_dev_clear_queues(dev);
1307 		return ret;
1308 	}
1309 
1310 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1311 		/* Enable VLAN filter since VMDq always use VLAN filter */
1312 		igb_vmdq_vlan_hw_filter_enable(dev);
1313 	}
1314 
1315 	if ((hw->mac.type == e1000_82576) || (hw->mac.type == e1000_82580) ||
1316 		(hw->mac.type == e1000_i350) || (hw->mac.type == e1000_i210) ||
1317 		(hw->mac.type == e1000_i211)) {
1318 		/* Configure EITR with the maximum possible value (0xFFFF) */
1319 		E1000_WRITE_REG(hw, E1000_EITR(0), 0xFFFF);
1320 	}
1321 
1322 	/* Setup link speed and duplex */
1323 	speeds = &dev->data->dev_conf.link_speeds;
1324 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
1325 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
1326 		hw->mac.autoneg = 1;
1327 	} else {
1328 		num_speeds = 0;
1329 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
1330 
1331 		/* Reset */
1332 		hw->phy.autoneg_advertised = 0;
1333 
1334 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1335 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1336 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
1337 			num_speeds = -1;
1338 			goto error_invalid_config;
1339 		}
1340 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
1341 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1342 			num_speeds++;
1343 		}
1344 		if (*speeds & ETH_LINK_SPEED_10M) {
1345 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1346 			num_speeds++;
1347 		}
1348 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
1349 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1350 			num_speeds++;
1351 		}
1352 		if (*speeds & ETH_LINK_SPEED_100M) {
1353 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1354 			num_speeds++;
1355 		}
1356 		if (*speeds & ETH_LINK_SPEED_1G) {
1357 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1358 			num_speeds++;
1359 		}
1360 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
1361 			goto error_invalid_config;
1362 
1363 		/* Set/reset the mac.autoneg based on the link speed,
1364 		 * fixed or not
1365 		 */
1366 		if (!autoneg) {
1367 			hw->mac.autoneg = 0;
1368 			hw->mac.forced_speed_duplex =
1369 					hw->phy.autoneg_advertised;
1370 		} else {
1371 			hw->mac.autoneg = 1;
1372 		}
1373 	}
1374 
1375 	e1000_setup_link(hw);
1376 
1377 	if (rte_intr_allow_others(intr_handle)) {
1378 		/* check if lsc interrupt is enabled */
1379 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1380 			eth_igb_lsc_interrupt_setup(dev, TRUE);
1381 		else
1382 			eth_igb_lsc_interrupt_setup(dev, FALSE);
1383 	} else {
1384 		rte_intr_callback_unregister(intr_handle,
1385 					     eth_igb_interrupt_handler,
1386 					     (void *)dev);
1387 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1388 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
1389 				     " no intr multiplex");
1390 	}
1391 
1392 	/* check if rxq interrupt is enabled */
1393 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
1394 	    rte_intr_dp_is_en(intr_handle))
1395 		eth_igb_rxq_interrupt_setup(dev);
1396 
1397 	/* enable uio/vfio intr/eventfd mapping */
1398 	rte_intr_enable(intr_handle);
1399 
1400 	/* resume enabled intr since hw reset */
1401 	igb_intr_enable(dev);
1402 
1403 	/* restore all types filter */
1404 	igb_filter_restore(dev);
1405 
1406 	eth_igb_rxtx_control(dev, true);
1407 	eth_igb_link_update(dev, 0);
1408 
1409 	PMD_INIT_LOG(DEBUG, "<<");
1410 
1411 	return 0;
1412 
1413 error_invalid_config:
1414 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
1415 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
1416 	igb_dev_clear_queues(dev);
1417 	return -EINVAL;
1418 }
1419 
1420 /*********************************************************************
1421  *
1422  *  This routine disables all traffic on the adapter by issuing a
1423  *  global reset on the MAC.
1424  *
1425  **********************************************************************/
1426 static int
1427 eth_igb_stop(struct rte_eth_dev *dev)
1428 {
1429 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1430 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1431 	struct rte_eth_link link;
1432 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1433 	struct e1000_adapter *adapter =
1434 		E1000_DEV_PRIVATE(dev->data->dev_private);
1435 
1436 	if (adapter->stopped)
1437 		return 0;
1438 
1439 	eth_igb_rxtx_control(dev, false);
1440 
1441 	igb_intr_disable(dev);
1442 
1443 	/* disable intr eventfd mapping */
1444 	rte_intr_disable(intr_handle);
1445 
1446 	igb_pf_reset_hw(hw);
1447 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1448 
1449 	/* Set bit for Go Link disconnect if PHY reset is not blocked */
1450 	if (hw->mac.type >= e1000_82580 &&
1451 	    (e1000_check_reset_block(hw) != E1000_BLK_PHY_RESET)) {
1452 		uint32_t phpm_reg;
1453 
1454 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1455 		phpm_reg |= E1000_82580_PM_GO_LINKD;
1456 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1457 	}
1458 
1459 	/* Power down the phy. Needed to make the link go Down */
1460 	eth_igb_dev_set_link_down(dev);
1461 
1462 	igb_dev_clear_queues(dev);
1463 
1464 	/* clear the recorded link status */
1465 	memset(&link, 0, sizeof(link));
1466 	rte_eth_linkstatus_set(dev, &link);
1467 
1468 	if (!rte_intr_allow_others(intr_handle))
1469 		/* resume to the default handler */
1470 		rte_intr_callback_register(intr_handle,
1471 					   eth_igb_interrupt_handler,
1472 					   (void *)dev);
1473 
1474 	/* Clean datapath event and queue/vec mapping */
1475 	rte_intr_efd_disable(intr_handle);
1476 	if (intr_handle->intr_vec != NULL) {
1477 		rte_free(intr_handle->intr_vec);
1478 		intr_handle->intr_vec = NULL;
1479 	}
1480 
1481 	adapter->stopped = true;
1482 	dev->data->dev_started = 0;
1483 
1484 	return 0;
1485 }
1486 
1487 static int
1488 eth_igb_dev_set_link_up(struct rte_eth_dev *dev)
1489 {
1490 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1491 
1492 	if (hw->phy.media_type == e1000_media_type_copper)
1493 		e1000_power_up_phy(hw);
1494 	else
1495 		e1000_power_up_fiber_serdes_link(hw);
1496 
1497 	return 0;
1498 }
1499 
1500 static int
1501 eth_igb_dev_set_link_down(struct rte_eth_dev *dev)
1502 {
1503 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1504 
1505 	if (hw->phy.media_type == e1000_media_type_copper)
1506 		e1000_power_down_phy(hw);
1507 	else
1508 		e1000_shutdown_fiber_serdes_link(hw);
1509 
1510 	return 0;
1511 }
1512 
1513 static int
1514 eth_igb_close(struct rte_eth_dev *dev)
1515 {
1516 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1517 	struct rte_eth_link link;
1518 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1519 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1520 	struct e1000_filter_info *filter_info =
1521 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
1522 	int ret;
1523 
1524 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1525 		return 0;
1526 
1527 	ret = eth_igb_stop(dev);
1528 
1529 	e1000_phy_hw_reset(hw);
1530 	igb_release_manageability(hw);
1531 	igb_hw_control_release(hw);
1532 
1533 	/* Clear bit for Go Link disconnect if PHY reset is not blocked */
1534 	if (hw->mac.type >= e1000_82580 &&
1535 	    (e1000_check_reset_block(hw) != E1000_BLK_PHY_RESET)) {
1536 		uint32_t phpm_reg;
1537 
1538 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1539 		phpm_reg &= ~E1000_82580_PM_GO_LINKD;
1540 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1541 	}
1542 
1543 	igb_dev_free_queues(dev);
1544 
1545 	if (intr_handle->intr_vec) {
1546 		rte_free(intr_handle->intr_vec);
1547 		intr_handle->intr_vec = NULL;
1548 	}
1549 
1550 	memset(&link, 0, sizeof(link));
1551 	rte_eth_linkstatus_set(dev, &link);
1552 
1553 	/* Reset any pending lock */
1554 	igb_reset_swfw_lock(hw);
1555 
1556 	/* uninitialize PF if max_vfs not zero */
1557 	igb_pf_host_uninit(dev);
1558 
1559 	rte_intr_callback_unregister(intr_handle,
1560 				     eth_igb_interrupt_handler, dev);
1561 
1562 	/* clear the SYN filter info */
1563 	filter_info->syn_info = 0;
1564 
1565 	/* clear the ethertype filters info */
1566 	filter_info->ethertype_mask = 0;
1567 	memset(filter_info->ethertype_filters, 0,
1568 		E1000_MAX_ETQF_FILTERS * sizeof(struct igb_ethertype_filter));
1569 
1570 	/* clear the rss filter info */
1571 	memset(&filter_info->rss_info, 0,
1572 		sizeof(struct igb_rte_flow_rss_conf));
1573 
1574 	/* remove all ntuple filters of the device */
1575 	igb_ntuple_filter_uninit(dev);
1576 
1577 	/* remove all flex filters of the device */
1578 	igb_flex_filter_uninit(dev);
1579 
1580 	/* clear all the filters list */
1581 	igb_filterlist_flush(dev);
1582 
1583 	return ret;
1584 }
1585 
1586 /*
1587  * Reset PF device.
1588  */
1589 static int
1590 eth_igb_reset(struct rte_eth_dev *dev)
1591 {
1592 	int ret;
1593 
1594 	/* When a DPDK PMD PF begin to reset PF port, it should notify all
1595 	 * its VF to make them align with it. The detailed notification
1596 	 * mechanism is PMD specific and is currently not implemented.
1597 	 * To avoid unexpected behavior in VF, currently reset of PF with
1598 	 * SR-IOV activation is not supported. It might be supported later.
1599 	 */
1600 	if (dev->data->sriov.active)
1601 		return -ENOTSUP;
1602 
1603 	ret = eth_igb_dev_uninit(dev);
1604 	if (ret)
1605 		return ret;
1606 
1607 	ret = eth_igb_dev_init(dev);
1608 
1609 	return ret;
1610 }
1611 
1612 
1613 static int
1614 igb_get_rx_buffer_size(struct e1000_hw *hw)
1615 {
1616 	uint32_t rx_buf_size;
1617 	if (hw->mac.type == e1000_82576) {
1618 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xffff) << 10;
1619 	} else if (hw->mac.type == e1000_82580 || hw->mac.type == e1000_i350) {
1620 		/* PBS needs to be translated according to a lookup table */
1621 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xf);
1622 		rx_buf_size = (uint32_t) e1000_rxpbs_adjust_82580(rx_buf_size);
1623 		rx_buf_size = (rx_buf_size << 10);
1624 	} else if (hw->mac.type == e1000_i210 || hw->mac.type == e1000_i211) {
1625 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0x3f) << 10;
1626 	} else {
1627 		rx_buf_size = (E1000_READ_REG(hw, E1000_PBA) & 0xffff) << 10;
1628 	}
1629 
1630 	return rx_buf_size;
1631 }
1632 
1633 /*********************************************************************
1634  *
1635  *  Initialize the hardware
1636  *
1637  **********************************************************************/
1638 static int
1639 igb_hardware_init(struct e1000_hw *hw)
1640 {
1641 	uint32_t rx_buf_size;
1642 	int diag;
1643 
1644 	/* Let the firmware know the OS is in control */
1645 	igb_hw_control_acquire(hw);
1646 
1647 	/*
1648 	 * These parameters control the automatic generation (Tx) and
1649 	 * response (Rx) to Ethernet PAUSE frames.
1650 	 * - High water mark should allow for at least two standard size (1518)
1651 	 *   frames to be received after sending an XOFF.
1652 	 * - Low water mark works best when it is very near the high water mark.
1653 	 *   This allows the receiver to restart by sending XON when it has
1654 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
1655 	 *   restart after one full frame is pulled from the buffer. There
1656 	 *   could be several smaller frames in the buffer and if so they will
1657 	 *   not trigger the XON until their total number reduces the buffer
1658 	 *   by 1500.
1659 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
1660 	 */
1661 	rx_buf_size = igb_get_rx_buffer_size(hw);
1662 
1663 	hw->fc.high_water = rx_buf_size - (RTE_ETHER_MAX_LEN * 2);
1664 	hw->fc.low_water = hw->fc.high_water - 1500;
1665 	hw->fc.pause_time = IGB_FC_PAUSE_TIME;
1666 	hw->fc.send_xon = 1;
1667 
1668 	/* Set Flow control, use the tunable location if sane */
1669 	if ((igb_fc_setting != e1000_fc_none) && (igb_fc_setting < 4))
1670 		hw->fc.requested_mode = igb_fc_setting;
1671 	else
1672 		hw->fc.requested_mode = e1000_fc_none;
1673 
1674 	/* Issue a global reset */
1675 	igb_pf_reset_hw(hw);
1676 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1677 
1678 	diag = e1000_init_hw(hw);
1679 	if (diag < 0)
1680 		return diag;
1681 
1682 	E1000_WRITE_REG(hw, E1000_VET,
1683 			RTE_ETHER_TYPE_VLAN << 16 | RTE_ETHER_TYPE_VLAN);
1684 	e1000_get_phy_info(hw);
1685 	e1000_check_for_link(hw);
1686 
1687 	return 0;
1688 }
1689 
1690 /* This function is based on igb_update_stats_counters() in igb/if_igb.c */
1691 static void
1692 igb_read_stats_registers(struct e1000_hw *hw, struct e1000_hw_stats *stats)
1693 {
1694 	int pause_frames;
1695 
1696 	uint64_t old_gprc  = stats->gprc;
1697 	uint64_t old_gptc  = stats->gptc;
1698 	uint64_t old_tpr   = stats->tpr;
1699 	uint64_t old_tpt   = stats->tpt;
1700 	uint64_t old_rpthc = stats->rpthc;
1701 	uint64_t old_hgptc = stats->hgptc;
1702 
1703 	if(hw->phy.media_type == e1000_media_type_copper ||
1704 	    (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
1705 		stats->symerrs +=
1706 		    E1000_READ_REG(hw,E1000_SYMERRS);
1707 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
1708 	}
1709 
1710 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
1711 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
1712 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
1713 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
1714 
1715 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
1716 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
1717 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
1718 	stats->dc += E1000_READ_REG(hw, E1000_DC);
1719 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
1720 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
1721 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
1722 	/*
1723 	** For watchdog management we need to know if we have been
1724 	** paused during the last interval, so capture that here.
1725 	*/
1726 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
1727 	stats->xoffrxc += pause_frames;
1728 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
1729 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
1730 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
1731 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
1732 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
1733 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
1734 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
1735 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
1736 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
1737 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
1738 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
1739 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
1740 
1741 	/* For the 64-bit byte counters the low dword must be read first. */
1742 	/* Both registers clear on the read of the high dword */
1743 
1744 	/* Workaround CRC bytes included in size, take away 4 bytes/packet */
1745 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
1746 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
1747 	stats->gorc -= (stats->gprc - old_gprc) * RTE_ETHER_CRC_LEN;
1748 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
1749 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
1750 	stats->gotc -= (stats->gptc - old_gptc) * RTE_ETHER_CRC_LEN;
1751 
1752 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
1753 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
1754 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
1755 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
1756 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
1757 
1758 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
1759 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
1760 
1761 	stats->tor += E1000_READ_REG(hw, E1000_TORL);
1762 	stats->tor += ((uint64_t)E1000_READ_REG(hw, E1000_TORH) << 32);
1763 	stats->tor -= (stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
1764 	stats->tot += E1000_READ_REG(hw, E1000_TOTL);
1765 	stats->tot += ((uint64_t)E1000_READ_REG(hw, E1000_TOTH) << 32);
1766 	stats->tot -= (stats->tpt - old_tpt) * RTE_ETHER_CRC_LEN;
1767 
1768 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
1769 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
1770 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
1771 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
1772 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
1773 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
1774 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
1775 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
1776 
1777 	/* Interrupt Counts */
1778 
1779 	stats->iac += E1000_READ_REG(hw, E1000_IAC);
1780 	stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
1781 	stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
1782 	stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
1783 	stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
1784 	stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
1785 	stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
1786 	stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
1787 	stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
1788 
1789 	/* Host to Card Statistics */
1790 
1791 	stats->cbtmpc += E1000_READ_REG(hw, E1000_CBTMPC);
1792 	stats->htdpmc += E1000_READ_REG(hw, E1000_HTDPMC);
1793 	stats->cbrdpc += E1000_READ_REG(hw, E1000_CBRDPC);
1794 	stats->cbrmpc += E1000_READ_REG(hw, E1000_CBRMPC);
1795 	stats->rpthc += E1000_READ_REG(hw, E1000_RPTHC);
1796 	stats->hgptc += E1000_READ_REG(hw, E1000_HGPTC);
1797 	stats->htcbdpc += E1000_READ_REG(hw, E1000_HTCBDPC);
1798 	stats->hgorc += E1000_READ_REG(hw, E1000_HGORCL);
1799 	stats->hgorc += ((uint64_t)E1000_READ_REG(hw, E1000_HGORCH) << 32);
1800 	stats->hgorc -= (stats->rpthc - old_rpthc) * RTE_ETHER_CRC_LEN;
1801 	stats->hgotc += E1000_READ_REG(hw, E1000_HGOTCL);
1802 	stats->hgotc += ((uint64_t)E1000_READ_REG(hw, E1000_HGOTCH) << 32);
1803 	stats->hgotc -= (stats->hgptc - old_hgptc) * RTE_ETHER_CRC_LEN;
1804 	stats->lenerrs += E1000_READ_REG(hw, E1000_LENERRS);
1805 	stats->scvpc += E1000_READ_REG(hw, E1000_SCVPC);
1806 	stats->hrmpc += E1000_READ_REG(hw, E1000_HRMPC);
1807 
1808 	stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
1809 	stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
1810 	stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
1811 	stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
1812 	stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
1813 	stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
1814 }
1815 
1816 static int
1817 eth_igb_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1818 {
1819 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1820 	struct e1000_hw_stats *stats =
1821 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1822 
1823 	igb_read_stats_registers(hw, stats);
1824 
1825 	if (rte_stats == NULL)
1826 		return -EINVAL;
1827 
1828 	/* Rx Errors */
1829 	rte_stats->imissed = stats->mpc;
1830 	rte_stats->ierrors = stats->crcerrs +
1831 	                     stats->rlec + stats->ruc + stats->roc +
1832 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
1833 
1834 	/* Tx Errors */
1835 	rte_stats->oerrors = stats->ecol + stats->latecol;
1836 
1837 	rte_stats->ipackets = stats->gprc;
1838 	rte_stats->opackets = stats->gptc;
1839 	rte_stats->ibytes   = stats->gorc;
1840 	rte_stats->obytes   = stats->gotc;
1841 	return 0;
1842 }
1843 
1844 static int
1845 eth_igb_stats_reset(struct rte_eth_dev *dev)
1846 {
1847 	struct e1000_hw_stats *hw_stats =
1848 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1849 
1850 	/* HW registers are cleared on read */
1851 	eth_igb_stats_get(dev, NULL);
1852 
1853 	/* Reset software totals */
1854 	memset(hw_stats, 0, sizeof(*hw_stats));
1855 
1856 	return 0;
1857 }
1858 
1859 static int
1860 eth_igb_xstats_reset(struct rte_eth_dev *dev)
1861 {
1862 	struct e1000_hw_stats *stats =
1863 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1864 
1865 	/* HW registers are cleared on read */
1866 	eth_igb_xstats_get(dev, NULL, IGB_NB_XSTATS);
1867 
1868 	/* Reset software totals */
1869 	memset(stats, 0, sizeof(*stats));
1870 
1871 	return 0;
1872 }
1873 
1874 static int eth_igb_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1875 	struct rte_eth_xstat_name *xstats_names,
1876 	__rte_unused unsigned int size)
1877 {
1878 	unsigned i;
1879 
1880 	if (xstats_names == NULL)
1881 		return IGB_NB_XSTATS;
1882 
1883 	/* Note: limit checked in rte_eth_xstats_names() */
1884 
1885 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1886 		strlcpy(xstats_names[i].name, rte_igb_stats_strings[i].name,
1887 			sizeof(xstats_names[i].name));
1888 	}
1889 
1890 	return IGB_NB_XSTATS;
1891 }
1892 
1893 static int eth_igb_xstats_get_names_by_id(struct rte_eth_dev *dev,
1894 		struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
1895 		unsigned int limit)
1896 {
1897 	unsigned int i;
1898 
1899 	if (!ids) {
1900 		if (xstats_names == NULL)
1901 			return IGB_NB_XSTATS;
1902 
1903 		for (i = 0; i < IGB_NB_XSTATS; i++)
1904 			strlcpy(xstats_names[i].name,
1905 				rte_igb_stats_strings[i].name,
1906 				sizeof(xstats_names[i].name));
1907 
1908 		return IGB_NB_XSTATS;
1909 
1910 	} else {
1911 		struct rte_eth_xstat_name xstats_names_copy[IGB_NB_XSTATS];
1912 
1913 		eth_igb_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
1914 				IGB_NB_XSTATS);
1915 
1916 		for (i = 0; i < limit; i++) {
1917 			if (ids[i] >= IGB_NB_XSTATS) {
1918 				PMD_INIT_LOG(ERR, "id value isn't valid");
1919 				return -1;
1920 			}
1921 			strcpy(xstats_names[i].name,
1922 					xstats_names_copy[ids[i]].name);
1923 		}
1924 		return limit;
1925 	}
1926 }
1927 
1928 static int
1929 eth_igb_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1930 		   unsigned n)
1931 {
1932 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1933 	struct e1000_hw_stats *hw_stats =
1934 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1935 	unsigned i;
1936 
1937 	if (n < IGB_NB_XSTATS)
1938 		return IGB_NB_XSTATS;
1939 
1940 	igb_read_stats_registers(hw, hw_stats);
1941 
1942 	/* If this is a reset xstats is NULL, and we have cleared the
1943 	 * registers by reading them.
1944 	 */
1945 	if (!xstats)
1946 		return 0;
1947 
1948 	/* Extended stats */
1949 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1950 		xstats[i].id = i;
1951 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1952 			rte_igb_stats_strings[i].offset);
1953 	}
1954 
1955 	return IGB_NB_XSTATS;
1956 }
1957 
1958 static int
1959 eth_igb_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
1960 		uint64_t *values, unsigned int n)
1961 {
1962 	unsigned int i;
1963 
1964 	if (!ids) {
1965 		struct e1000_hw *hw =
1966 			E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1967 		struct e1000_hw_stats *hw_stats =
1968 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1969 
1970 		if (n < IGB_NB_XSTATS)
1971 			return IGB_NB_XSTATS;
1972 
1973 		igb_read_stats_registers(hw, hw_stats);
1974 
1975 		/* If this is a reset xstats is NULL, and we have cleared the
1976 		 * registers by reading them.
1977 		 */
1978 		if (!values)
1979 			return 0;
1980 
1981 		/* Extended stats */
1982 		for (i = 0; i < IGB_NB_XSTATS; i++)
1983 			values[i] = *(uint64_t *)(((char *)hw_stats) +
1984 					rte_igb_stats_strings[i].offset);
1985 
1986 		return IGB_NB_XSTATS;
1987 
1988 	} else {
1989 		uint64_t values_copy[IGB_NB_XSTATS];
1990 
1991 		eth_igb_xstats_get_by_id(dev, NULL, values_copy,
1992 				IGB_NB_XSTATS);
1993 
1994 		for (i = 0; i < n; i++) {
1995 			if (ids[i] >= IGB_NB_XSTATS) {
1996 				PMD_INIT_LOG(ERR, "id value isn't valid");
1997 				return -1;
1998 			}
1999 			values[i] = values_copy[ids[i]];
2000 		}
2001 		return n;
2002 	}
2003 }
2004 
2005 static void
2006 igbvf_read_stats_registers(struct e1000_hw *hw, struct e1000_vf_stats *hw_stats)
2007 {
2008 	/* Good Rx packets, include VF loopback */
2009 	UPDATE_VF_STAT(E1000_VFGPRC,
2010 	    hw_stats->last_gprc, hw_stats->gprc);
2011 
2012 	/* Good Rx octets, include VF loopback */
2013 	UPDATE_VF_STAT(E1000_VFGORC,
2014 	    hw_stats->last_gorc, hw_stats->gorc);
2015 
2016 	/* Good Tx packets, include VF loopback */
2017 	UPDATE_VF_STAT(E1000_VFGPTC,
2018 	    hw_stats->last_gptc, hw_stats->gptc);
2019 
2020 	/* Good Tx octets, include VF loopback */
2021 	UPDATE_VF_STAT(E1000_VFGOTC,
2022 	    hw_stats->last_gotc, hw_stats->gotc);
2023 
2024 	/* Rx Multicst packets */
2025 	UPDATE_VF_STAT(E1000_VFMPRC,
2026 	    hw_stats->last_mprc, hw_stats->mprc);
2027 
2028 	/* Good Rx loopback packets */
2029 	UPDATE_VF_STAT(E1000_VFGPRLBC,
2030 	    hw_stats->last_gprlbc, hw_stats->gprlbc);
2031 
2032 	/* Good Rx loopback octets */
2033 	UPDATE_VF_STAT(E1000_VFGORLBC,
2034 	    hw_stats->last_gorlbc, hw_stats->gorlbc);
2035 
2036 	/* Good Tx loopback packets */
2037 	UPDATE_VF_STAT(E1000_VFGPTLBC,
2038 	    hw_stats->last_gptlbc, hw_stats->gptlbc);
2039 
2040 	/* Good Tx loopback octets */
2041 	UPDATE_VF_STAT(E1000_VFGOTLBC,
2042 	    hw_stats->last_gotlbc, hw_stats->gotlbc);
2043 }
2044 
2045 static int eth_igbvf_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
2046 				     struct rte_eth_xstat_name *xstats_names,
2047 				     __rte_unused unsigned limit)
2048 {
2049 	unsigned i;
2050 
2051 	if (xstats_names != NULL)
2052 		for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2053 			strlcpy(xstats_names[i].name,
2054 				rte_igbvf_stats_strings[i].name,
2055 				sizeof(xstats_names[i].name));
2056 		}
2057 	return IGBVF_NB_XSTATS;
2058 }
2059 
2060 static int
2061 eth_igbvf_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
2062 		     unsigned n)
2063 {
2064 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2065 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2066 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2067 	unsigned i;
2068 
2069 	if (n < IGBVF_NB_XSTATS)
2070 		return IGBVF_NB_XSTATS;
2071 
2072 	igbvf_read_stats_registers(hw, hw_stats);
2073 
2074 	if (!xstats)
2075 		return 0;
2076 
2077 	for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2078 		xstats[i].id = i;
2079 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
2080 			rte_igbvf_stats_strings[i].offset);
2081 	}
2082 
2083 	return IGBVF_NB_XSTATS;
2084 }
2085 
2086 static int
2087 eth_igbvf_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
2088 {
2089 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2090 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2091 			  E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2092 
2093 	igbvf_read_stats_registers(hw, hw_stats);
2094 
2095 	if (rte_stats == NULL)
2096 		return -EINVAL;
2097 
2098 	rte_stats->ipackets = hw_stats->gprc;
2099 	rte_stats->ibytes = hw_stats->gorc;
2100 	rte_stats->opackets = hw_stats->gptc;
2101 	rte_stats->obytes = hw_stats->gotc;
2102 	return 0;
2103 }
2104 
2105 static int
2106 eth_igbvf_stats_reset(struct rte_eth_dev *dev)
2107 {
2108 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats*)
2109 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2110 
2111 	/* Sync HW register to the last stats */
2112 	eth_igbvf_stats_get(dev, NULL);
2113 
2114 	/* reset HW current stats*/
2115 	memset(&hw_stats->gprc, 0, sizeof(*hw_stats) -
2116 	       offsetof(struct e1000_vf_stats, gprc));
2117 
2118 	return 0;
2119 }
2120 
2121 static int
2122 eth_igb_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
2123 		       size_t fw_size)
2124 {
2125 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2126 	struct e1000_fw_version fw;
2127 	int ret;
2128 
2129 	e1000_get_fw_version(hw, &fw);
2130 
2131 	switch (hw->mac.type) {
2132 	case e1000_i210:
2133 	case e1000_i211:
2134 		if (!(e1000_get_flash_presence_i210(hw))) {
2135 			ret = snprintf(fw_version, fw_size,
2136 				 "%2d.%2d-%d",
2137 				 fw.invm_major, fw.invm_minor,
2138 				 fw.invm_img_type);
2139 			break;
2140 		}
2141 		/* fall through */
2142 	default:
2143 		/* if option rom is valid, display its version too */
2144 		if (fw.or_valid) {
2145 			ret = snprintf(fw_version, fw_size,
2146 				 "%d.%d, 0x%08x, %d.%d.%d",
2147 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
2148 				 fw.or_major, fw.or_build, fw.or_patch);
2149 		/* no option rom */
2150 		} else {
2151 			if (fw.etrack_id != 0X0000) {
2152 				ret = snprintf(fw_version, fw_size,
2153 					 "%d.%d, 0x%08x",
2154 					 fw.eep_major, fw.eep_minor,
2155 					 fw.etrack_id);
2156 			} else {
2157 				ret = snprintf(fw_version, fw_size,
2158 					 "%d.%d.%d",
2159 					 fw.eep_major, fw.eep_minor,
2160 					 fw.eep_build);
2161 			}
2162 		}
2163 		break;
2164 	}
2165 
2166 	ret += 1; /* add the size of '\0' */
2167 	if (fw_size < (u32)ret)
2168 		return ret;
2169 	else
2170 		return 0;
2171 }
2172 
2173 static int
2174 eth_igb_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2175 {
2176 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2177 
2178 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2179 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2180 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2181 	dev_info->rx_queue_offload_capa = igb_get_rx_queue_offloads_capa(dev);
2182 	dev_info->rx_offload_capa = igb_get_rx_port_offloads_capa(dev) |
2183 				    dev_info->rx_queue_offload_capa;
2184 	dev_info->tx_queue_offload_capa = igb_get_tx_queue_offloads_capa(dev);
2185 	dev_info->tx_offload_capa = igb_get_tx_port_offloads_capa(dev) |
2186 				    dev_info->tx_queue_offload_capa;
2187 
2188 	switch (hw->mac.type) {
2189 	case e1000_82575:
2190 		dev_info->max_rx_queues = 4;
2191 		dev_info->max_tx_queues = 4;
2192 		dev_info->max_vmdq_pools = 0;
2193 		break;
2194 
2195 	case e1000_82576:
2196 		dev_info->max_rx_queues = 16;
2197 		dev_info->max_tx_queues = 16;
2198 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2199 		dev_info->vmdq_queue_num = 16;
2200 		break;
2201 
2202 	case e1000_82580:
2203 		dev_info->max_rx_queues = 8;
2204 		dev_info->max_tx_queues = 8;
2205 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2206 		dev_info->vmdq_queue_num = 8;
2207 		break;
2208 
2209 	case e1000_i350:
2210 		dev_info->max_rx_queues = 8;
2211 		dev_info->max_tx_queues = 8;
2212 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2213 		dev_info->vmdq_queue_num = 8;
2214 		break;
2215 
2216 	case e1000_i354:
2217 		dev_info->max_rx_queues = 8;
2218 		dev_info->max_tx_queues = 8;
2219 		break;
2220 
2221 	case e1000_i210:
2222 		dev_info->max_rx_queues = 4;
2223 		dev_info->max_tx_queues = 4;
2224 		dev_info->max_vmdq_pools = 0;
2225 		break;
2226 
2227 	case e1000_i211:
2228 		dev_info->max_rx_queues = 2;
2229 		dev_info->max_tx_queues = 2;
2230 		dev_info->max_vmdq_pools = 0;
2231 		break;
2232 
2233 	default:
2234 		/* Should not happen */
2235 		return -EINVAL;
2236 	}
2237 	dev_info->hash_key_size = IGB_HKEY_MAX_INDEX * sizeof(uint32_t);
2238 	dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
2239 	dev_info->flow_type_rss_offloads = IGB_RSS_OFFLOAD_ALL;
2240 
2241 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2242 		.rx_thresh = {
2243 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2244 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2245 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2246 		},
2247 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2248 		.rx_drop_en = 0,
2249 		.offloads = 0,
2250 	};
2251 
2252 	dev_info->default_txconf = (struct rte_eth_txconf) {
2253 		.tx_thresh = {
2254 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2255 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2256 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2257 		},
2258 		.offloads = 0,
2259 	};
2260 
2261 	dev_info->rx_desc_lim = rx_desc_lim;
2262 	dev_info->tx_desc_lim = tx_desc_lim;
2263 
2264 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
2265 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
2266 			ETH_LINK_SPEED_1G;
2267 
2268 	dev_info->max_mtu = dev_info->max_rx_pktlen - E1000_ETH_OVERHEAD;
2269 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
2270 
2271 	return 0;
2272 }
2273 
2274 static const uint32_t *
2275 eth_igb_supported_ptypes_get(struct rte_eth_dev *dev)
2276 {
2277 	static const uint32_t ptypes[] = {
2278 		/* refers to igb_rxd_pkt_info_to_pkt_type() */
2279 		RTE_PTYPE_L2_ETHER,
2280 		RTE_PTYPE_L3_IPV4,
2281 		RTE_PTYPE_L3_IPV4_EXT,
2282 		RTE_PTYPE_L3_IPV6,
2283 		RTE_PTYPE_L3_IPV6_EXT,
2284 		RTE_PTYPE_L4_TCP,
2285 		RTE_PTYPE_L4_UDP,
2286 		RTE_PTYPE_L4_SCTP,
2287 		RTE_PTYPE_TUNNEL_IP,
2288 		RTE_PTYPE_INNER_L3_IPV6,
2289 		RTE_PTYPE_INNER_L3_IPV6_EXT,
2290 		RTE_PTYPE_INNER_L4_TCP,
2291 		RTE_PTYPE_INNER_L4_UDP,
2292 		RTE_PTYPE_UNKNOWN
2293 	};
2294 
2295 	if (dev->rx_pkt_burst == eth_igb_recv_pkts ||
2296 	    dev->rx_pkt_burst == eth_igb_recv_scattered_pkts)
2297 		return ptypes;
2298 	return NULL;
2299 }
2300 
2301 static int
2302 eth_igbvf_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2303 {
2304 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2305 
2306 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2307 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2308 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2309 	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
2310 				DEV_TX_OFFLOAD_IPV4_CKSUM  |
2311 				DEV_TX_OFFLOAD_UDP_CKSUM   |
2312 				DEV_TX_OFFLOAD_TCP_CKSUM   |
2313 				DEV_TX_OFFLOAD_SCTP_CKSUM  |
2314 				DEV_TX_OFFLOAD_TCP_TSO;
2315 	switch (hw->mac.type) {
2316 	case e1000_vfadapt:
2317 		dev_info->max_rx_queues = 2;
2318 		dev_info->max_tx_queues = 2;
2319 		break;
2320 	case e1000_vfadapt_i350:
2321 		dev_info->max_rx_queues = 1;
2322 		dev_info->max_tx_queues = 1;
2323 		break;
2324 	default:
2325 		/* Should not happen */
2326 		return -EINVAL;
2327 	}
2328 
2329 	dev_info->rx_queue_offload_capa = igb_get_rx_queue_offloads_capa(dev);
2330 	dev_info->rx_offload_capa = igb_get_rx_port_offloads_capa(dev) |
2331 				    dev_info->rx_queue_offload_capa;
2332 	dev_info->tx_queue_offload_capa = igb_get_tx_queue_offloads_capa(dev);
2333 	dev_info->tx_offload_capa = igb_get_tx_port_offloads_capa(dev) |
2334 				    dev_info->tx_queue_offload_capa;
2335 
2336 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2337 		.rx_thresh = {
2338 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2339 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2340 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2341 		},
2342 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2343 		.rx_drop_en = 0,
2344 		.offloads = 0,
2345 	};
2346 
2347 	dev_info->default_txconf = (struct rte_eth_txconf) {
2348 		.tx_thresh = {
2349 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2350 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2351 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2352 		},
2353 		.offloads = 0,
2354 	};
2355 
2356 	dev_info->rx_desc_lim = rx_desc_lim;
2357 	dev_info->tx_desc_lim = tx_desc_lim;
2358 
2359 	return 0;
2360 }
2361 
2362 /* return 0 means link status changed, -1 means not changed */
2363 static int
2364 eth_igb_link_update(struct rte_eth_dev *dev, int wait_to_complete)
2365 {
2366 	struct e1000_hw *hw =
2367 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2368 	struct rte_eth_link link;
2369 	int link_check, count;
2370 
2371 	link_check = 0;
2372 	hw->mac.get_link_status = 1;
2373 
2374 	/* possible wait-to-complete in up to 9 seconds */
2375 	for (count = 0; count < IGB_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
2376 		/* Read the real link status */
2377 		switch (hw->phy.media_type) {
2378 		case e1000_media_type_copper:
2379 			/* Do the work to read phy */
2380 			e1000_check_for_link(hw);
2381 			link_check = !hw->mac.get_link_status;
2382 			break;
2383 
2384 		case e1000_media_type_fiber:
2385 			e1000_check_for_link(hw);
2386 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
2387 				      E1000_STATUS_LU);
2388 			break;
2389 
2390 		case e1000_media_type_internal_serdes:
2391 			e1000_check_for_link(hw);
2392 			link_check = hw->mac.serdes_has_link;
2393 			break;
2394 
2395 		/* VF device is type_unknown */
2396 		case e1000_media_type_unknown:
2397 			eth_igbvf_link_update(hw);
2398 			link_check = !hw->mac.get_link_status;
2399 			break;
2400 
2401 		default:
2402 			break;
2403 		}
2404 		if (link_check || wait_to_complete == 0)
2405 			break;
2406 		rte_delay_ms(IGB_LINK_UPDATE_CHECK_INTERVAL);
2407 	}
2408 	memset(&link, 0, sizeof(link));
2409 
2410 	/* Now we check if a transition has happened */
2411 	if (link_check) {
2412 		uint16_t duplex, speed;
2413 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
2414 		link.link_duplex = (duplex == FULL_DUPLEX) ?
2415 				ETH_LINK_FULL_DUPLEX :
2416 				ETH_LINK_HALF_DUPLEX;
2417 		link.link_speed = speed;
2418 		link.link_status = ETH_LINK_UP;
2419 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
2420 				ETH_LINK_SPEED_FIXED);
2421 	} else if (!link_check) {
2422 		link.link_speed = 0;
2423 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
2424 		link.link_status = ETH_LINK_DOWN;
2425 		link.link_autoneg = ETH_LINK_FIXED;
2426 	}
2427 
2428 	return rte_eth_linkstatus_set(dev, &link);
2429 }
2430 
2431 /*
2432  * igb_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
2433  * For ASF and Pass Through versions of f/w this means
2434  * that the driver is loaded.
2435  */
2436 static void
2437 igb_hw_control_acquire(struct e1000_hw *hw)
2438 {
2439 	uint32_t ctrl_ext;
2440 
2441 	/* Let firmware know the driver has taken over */
2442 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2443 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2444 }
2445 
2446 /*
2447  * igb_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
2448  * For ASF and Pass Through versions of f/w this means that the
2449  * driver is no longer loaded.
2450  */
2451 static void
2452 igb_hw_control_release(struct e1000_hw *hw)
2453 {
2454 	uint32_t ctrl_ext;
2455 
2456 	/* Let firmware taken over control of h/w */
2457 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2458 	E1000_WRITE_REG(hw, E1000_CTRL_EXT,
2459 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2460 }
2461 
2462 /*
2463  * Bit of a misnomer, what this really means is
2464  * to enable OS management of the system... aka
2465  * to disable special hardware management features.
2466  */
2467 static void
2468 igb_init_manageability(struct e1000_hw *hw)
2469 {
2470 	if (e1000_enable_mng_pass_thru(hw)) {
2471 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
2472 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2473 
2474 		/* disable hardware interception of ARP */
2475 		manc &= ~(E1000_MANC_ARP_EN);
2476 
2477 		/* enable receiving management packets to the host */
2478 		manc |= E1000_MANC_EN_MNG2HOST;
2479 		manc2h |= 1 << 5;  /* Mng Port 623 */
2480 		manc2h |= 1 << 6;  /* Mng Port 664 */
2481 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
2482 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2483 	}
2484 }
2485 
2486 static void
2487 igb_release_manageability(struct e1000_hw *hw)
2488 {
2489 	if (e1000_enable_mng_pass_thru(hw)) {
2490 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2491 
2492 		manc |= E1000_MANC_ARP_EN;
2493 		manc &= ~E1000_MANC_EN_MNG2HOST;
2494 
2495 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2496 	}
2497 }
2498 
2499 static int
2500 eth_igb_promiscuous_enable(struct rte_eth_dev *dev)
2501 {
2502 	struct e1000_hw *hw =
2503 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2504 	uint32_t rctl;
2505 
2506 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2507 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
2508 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2509 
2510 	return 0;
2511 }
2512 
2513 static int
2514 eth_igb_promiscuous_disable(struct rte_eth_dev *dev)
2515 {
2516 	struct e1000_hw *hw =
2517 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2518 	uint32_t rctl;
2519 
2520 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2521 	rctl &= (~E1000_RCTL_UPE);
2522 	if (dev->data->all_multicast == 1)
2523 		rctl |= E1000_RCTL_MPE;
2524 	else
2525 		rctl &= (~E1000_RCTL_MPE);
2526 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2527 
2528 	return 0;
2529 }
2530 
2531 static int
2532 eth_igb_allmulticast_enable(struct rte_eth_dev *dev)
2533 {
2534 	struct e1000_hw *hw =
2535 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2536 	uint32_t rctl;
2537 
2538 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2539 	rctl |= E1000_RCTL_MPE;
2540 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2541 
2542 	return 0;
2543 }
2544 
2545 static int
2546 eth_igb_allmulticast_disable(struct rte_eth_dev *dev)
2547 {
2548 	struct e1000_hw *hw =
2549 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2550 	uint32_t rctl;
2551 
2552 	if (dev->data->promiscuous == 1)
2553 		return 0; /* must remain in all_multicast mode */
2554 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2555 	rctl &= (~E1000_RCTL_MPE);
2556 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2557 
2558 	return 0;
2559 }
2560 
2561 static int
2562 eth_igb_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2563 {
2564 	struct e1000_hw *hw =
2565 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2566 	struct e1000_vfta * shadow_vfta =
2567 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2568 	uint32_t vfta;
2569 	uint32_t vid_idx;
2570 	uint32_t vid_bit;
2571 
2572 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
2573 			      E1000_VFTA_ENTRY_MASK);
2574 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
2575 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
2576 	if (on)
2577 		vfta |= vid_bit;
2578 	else
2579 		vfta &= ~vid_bit;
2580 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
2581 
2582 	/* update local VFTA copy */
2583 	shadow_vfta->vfta[vid_idx] = vfta;
2584 
2585 	return 0;
2586 }
2587 
2588 static int
2589 eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
2590 		      enum rte_vlan_type vlan_type,
2591 		      uint16_t tpid)
2592 {
2593 	struct e1000_hw *hw =
2594 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2595 	uint32_t reg, qinq;
2596 
2597 	qinq = E1000_READ_REG(hw, E1000_CTRL_EXT);
2598 	qinq &= E1000_CTRL_EXT_EXT_VLAN;
2599 
2600 	/* only outer TPID of double VLAN can be configured*/
2601 	if (qinq && vlan_type == ETH_VLAN_TYPE_OUTER) {
2602 		reg = E1000_READ_REG(hw, E1000_VET);
2603 		reg = (reg & (~E1000_VET_VET_EXT)) |
2604 			((uint32_t)tpid << E1000_VET_VET_EXT_SHIFT);
2605 		E1000_WRITE_REG(hw, E1000_VET, reg);
2606 
2607 		return 0;
2608 	}
2609 
2610 	/* all other TPID values are read-only*/
2611 	PMD_DRV_LOG(ERR, "Not supported");
2612 
2613 	return -ENOTSUP;
2614 }
2615 
2616 static void
2617 igb_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2618 {
2619 	struct e1000_hw *hw =
2620 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2621 	uint32_t reg;
2622 
2623 	/* Filter Table Disable */
2624 	reg = E1000_READ_REG(hw, E1000_RCTL);
2625 	reg &= ~E1000_RCTL_CFIEN;
2626 	reg &= ~E1000_RCTL_VFE;
2627 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2628 }
2629 
2630 static void
2631 igb_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2632 {
2633 	struct e1000_hw *hw =
2634 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2635 	struct e1000_vfta * shadow_vfta =
2636 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2637 	uint32_t reg;
2638 	int i;
2639 
2640 	/* Filter Table Enable, CFI not used for packet acceptance */
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 	/* restore VFTA table */
2647 	for (i = 0; i < IGB_VFTA_SIZE; i++)
2648 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
2649 }
2650 
2651 static void
2652 igb_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2653 {
2654 	struct e1000_hw *hw =
2655 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2656 	uint32_t reg;
2657 
2658 	/* VLAN Mode Disable */
2659 	reg = E1000_READ_REG(hw, E1000_CTRL);
2660 	reg &= ~E1000_CTRL_VME;
2661 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2662 }
2663 
2664 static void
2665 igb_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2666 {
2667 	struct e1000_hw *hw =
2668 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2669 	uint32_t reg;
2670 
2671 	/* VLAN Mode Enable */
2672 	reg = E1000_READ_REG(hw, E1000_CTRL);
2673 	reg |= E1000_CTRL_VME;
2674 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2675 }
2676 
2677 static void
2678 igb_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2679 {
2680 	struct e1000_hw *hw =
2681 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2682 	uint32_t reg;
2683 
2684 	/* CTRL_EXT: Extended VLAN */
2685 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2686 	reg &= ~E1000_CTRL_EXT_EXTEND_VLAN;
2687 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2688 
2689 	/* Update maximum packet length */
2690 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME)
2691 		E1000_WRITE_REG(hw, E1000_RLPML,
2692 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2693 						VLAN_TAG_SIZE);
2694 }
2695 
2696 static void
2697 igb_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2698 {
2699 	struct e1000_hw *hw =
2700 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2701 	uint32_t reg;
2702 
2703 	/* CTRL_EXT: Extended VLAN */
2704 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2705 	reg |= E1000_CTRL_EXT_EXTEND_VLAN;
2706 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2707 
2708 	/* Update maximum packet length */
2709 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME)
2710 		E1000_WRITE_REG(hw, E1000_RLPML,
2711 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2712 						2 * VLAN_TAG_SIZE);
2713 }
2714 
2715 static int
2716 eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2717 {
2718 	struct rte_eth_rxmode *rxmode;
2719 
2720 	rxmode = &dev->data->dev_conf.rxmode;
2721 	if(mask & ETH_VLAN_STRIP_MASK){
2722 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2723 			igb_vlan_hw_strip_enable(dev);
2724 		else
2725 			igb_vlan_hw_strip_disable(dev);
2726 	}
2727 
2728 	if(mask & ETH_VLAN_FILTER_MASK){
2729 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2730 			igb_vlan_hw_filter_enable(dev);
2731 		else
2732 			igb_vlan_hw_filter_disable(dev);
2733 	}
2734 
2735 	if(mask & ETH_VLAN_EXTEND_MASK){
2736 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2737 			igb_vlan_hw_extend_enable(dev);
2738 		else
2739 			igb_vlan_hw_extend_disable(dev);
2740 	}
2741 
2742 	return 0;
2743 }
2744 
2745 
2746 /**
2747  * It enables the interrupt mask and then enable the interrupt.
2748  *
2749  * @param dev
2750  *  Pointer to struct rte_eth_dev.
2751  * @param on
2752  *  Enable or Disable
2753  *
2754  * @return
2755  *  - On success, zero.
2756  *  - On failure, a negative value.
2757  */
2758 static int
2759 eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
2760 {
2761 	struct e1000_interrupt *intr =
2762 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2763 
2764 	if (on)
2765 		intr->mask |= E1000_ICR_LSC;
2766 	else
2767 		intr->mask &= ~E1000_ICR_LSC;
2768 
2769 	return 0;
2770 }
2771 
2772 /* It clears the interrupt causes and enables the interrupt.
2773  * It will be called once only during nic initialized.
2774  *
2775  * @param dev
2776  *  Pointer to struct rte_eth_dev.
2777  *
2778  * @return
2779  *  - On success, zero.
2780  *  - On failure, a negative value.
2781  */
2782 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev)
2783 {
2784 	uint32_t mask, regval;
2785 	int ret;
2786 	struct e1000_hw *hw =
2787 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2788 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2789 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2790 	int misc_shift = rte_intr_allow_others(intr_handle) ? 1 : 0;
2791 	struct rte_eth_dev_info dev_info;
2792 
2793 	memset(&dev_info, 0, sizeof(dev_info));
2794 	ret = eth_igb_infos_get(dev, &dev_info);
2795 	if (ret != 0)
2796 		return ret;
2797 
2798 	mask = (0xFFFFFFFF >> (32 - dev_info.max_rx_queues)) << misc_shift;
2799 	regval = E1000_READ_REG(hw, E1000_EIMS);
2800 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
2801 
2802 	return 0;
2803 }
2804 
2805 /*
2806  * It reads ICR and gets interrupt causes, check it and set a bit flag
2807  * to update link status.
2808  *
2809  * @param dev
2810  *  Pointer to struct rte_eth_dev.
2811  *
2812  * @return
2813  *  - On success, zero.
2814  *  - On failure, a negative value.
2815  */
2816 static int
2817 eth_igb_interrupt_get_status(struct rte_eth_dev *dev)
2818 {
2819 	uint32_t icr;
2820 	struct e1000_hw *hw =
2821 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2822 	struct e1000_interrupt *intr =
2823 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2824 
2825 	igb_intr_disable(dev);
2826 
2827 	/* read-on-clear nic registers here */
2828 	icr = E1000_READ_REG(hw, E1000_ICR);
2829 
2830 	intr->flags = 0;
2831 	if (icr & E1000_ICR_LSC) {
2832 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
2833 	}
2834 
2835 	if (icr & E1000_ICR_VMMB)
2836 		intr->flags |= E1000_FLAG_MAILBOX;
2837 
2838 	return 0;
2839 }
2840 
2841 /*
2842  * It executes link_update after knowing an interrupt is prsent.
2843  *
2844  * @param dev
2845  *  Pointer to struct rte_eth_dev.
2846  *
2847  * @return
2848  *  - On success, zero.
2849  *  - On failure, a negative value.
2850  */
2851 static int
2852 eth_igb_interrupt_action(struct rte_eth_dev *dev,
2853 			 struct rte_intr_handle *intr_handle)
2854 {
2855 	struct e1000_hw *hw =
2856 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2857 	struct e1000_interrupt *intr =
2858 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2859 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2860 	struct rte_eth_link link;
2861 	int ret;
2862 
2863 	if (intr->flags & E1000_FLAG_MAILBOX) {
2864 		igb_pf_mbx_process(dev);
2865 		intr->flags &= ~E1000_FLAG_MAILBOX;
2866 	}
2867 
2868 	igb_intr_enable(dev);
2869 	rte_intr_ack(intr_handle);
2870 
2871 	if (intr->flags & E1000_FLAG_NEED_LINK_UPDATE) {
2872 		intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
2873 
2874 		/* set get_link_status to check register later */
2875 		hw->mac.get_link_status = 1;
2876 		ret = eth_igb_link_update(dev, 0);
2877 
2878 		/* check if link has changed */
2879 		if (ret < 0)
2880 			return 0;
2881 
2882 		rte_eth_linkstatus_get(dev, &link);
2883 		if (link.link_status) {
2884 			PMD_INIT_LOG(INFO,
2885 				     " Port %d: Link Up - speed %u Mbps - %s",
2886 				     dev->data->port_id,
2887 				     (unsigned)link.link_speed,
2888 				     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
2889 				     "full-duplex" : "half-duplex");
2890 		} else {
2891 			PMD_INIT_LOG(INFO, " Port %d: Link Down",
2892 				     dev->data->port_id);
2893 		}
2894 
2895 		PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
2896 			     pci_dev->addr.domain,
2897 			     pci_dev->addr.bus,
2898 			     pci_dev->addr.devid,
2899 			     pci_dev->addr.function);
2900 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
2901 	}
2902 
2903 	return 0;
2904 }
2905 
2906 /**
2907  * Interrupt handler which shall be registered at first.
2908  *
2909  * @param handle
2910  *  Pointer to interrupt handle.
2911  * @param param
2912  *  The address of parameter (struct rte_eth_dev *) regsitered before.
2913  *
2914  * @return
2915  *  void
2916  */
2917 static void
2918 eth_igb_interrupt_handler(void *param)
2919 {
2920 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2921 
2922 	eth_igb_interrupt_get_status(dev);
2923 	eth_igb_interrupt_action(dev, dev->intr_handle);
2924 }
2925 
2926 static int
2927 eth_igbvf_interrupt_get_status(struct rte_eth_dev *dev)
2928 {
2929 	uint32_t eicr;
2930 	struct e1000_hw *hw =
2931 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2932 	struct e1000_interrupt *intr =
2933 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2934 
2935 	igbvf_intr_disable(hw);
2936 
2937 	/* read-on-clear nic registers here */
2938 	eicr = E1000_READ_REG(hw, E1000_EICR);
2939 	intr->flags = 0;
2940 
2941 	if (eicr == E1000_VTIVAR_MISC_MAILBOX)
2942 		intr->flags |= E1000_FLAG_MAILBOX;
2943 
2944 	return 0;
2945 }
2946 
2947 void igbvf_mbx_process(struct rte_eth_dev *dev)
2948 {
2949 	struct e1000_hw *hw =
2950 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2951 	struct e1000_mbx_info *mbx = &hw->mbx;
2952 	u32 in_msg = 0;
2953 
2954 	/* peek the message first */
2955 	in_msg = E1000_READ_REG(hw, E1000_VMBMEM(0));
2956 
2957 	/* PF reset VF event */
2958 	if (in_msg == E1000_PF_CONTROL_MSG) {
2959 		/* dummy mbx read to ack pf */
2960 		if (mbx->ops.read(hw, &in_msg, 1, 0))
2961 			return;
2962 		rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
2963 					     NULL);
2964 	}
2965 }
2966 
2967 static int
2968 eth_igbvf_interrupt_action(struct rte_eth_dev *dev, struct rte_intr_handle *intr_handle)
2969 {
2970 	struct e1000_interrupt *intr =
2971 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2972 
2973 	if (intr->flags & E1000_FLAG_MAILBOX) {
2974 		igbvf_mbx_process(dev);
2975 		intr->flags &= ~E1000_FLAG_MAILBOX;
2976 	}
2977 
2978 	igbvf_intr_enable(dev);
2979 	rte_intr_ack(intr_handle);
2980 
2981 	return 0;
2982 }
2983 
2984 static void
2985 eth_igbvf_interrupt_handler(void *param)
2986 {
2987 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2988 
2989 	eth_igbvf_interrupt_get_status(dev);
2990 	eth_igbvf_interrupt_action(dev, dev->intr_handle);
2991 }
2992 
2993 static int
2994 eth_igb_led_on(struct rte_eth_dev *dev)
2995 {
2996 	struct e1000_hw *hw;
2997 
2998 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2999 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3000 }
3001 
3002 static int
3003 eth_igb_led_off(struct rte_eth_dev *dev)
3004 {
3005 	struct e1000_hw *hw;
3006 
3007 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3008 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3009 }
3010 
3011 static int
3012 eth_igb_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3013 {
3014 	struct e1000_hw *hw;
3015 	uint32_t ctrl;
3016 	int tx_pause;
3017 	int rx_pause;
3018 
3019 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3020 	fc_conf->pause_time = hw->fc.pause_time;
3021 	fc_conf->high_water = hw->fc.high_water;
3022 	fc_conf->low_water = hw->fc.low_water;
3023 	fc_conf->send_xon = hw->fc.send_xon;
3024 	fc_conf->autoneg = hw->mac.autoneg;
3025 
3026 	/*
3027 	 * Return rx_pause and tx_pause status according to actual setting of
3028 	 * the TFCE and RFCE bits in the CTRL register.
3029 	 */
3030 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
3031 	if (ctrl & E1000_CTRL_TFCE)
3032 		tx_pause = 1;
3033 	else
3034 		tx_pause = 0;
3035 
3036 	if (ctrl & E1000_CTRL_RFCE)
3037 		rx_pause = 1;
3038 	else
3039 		rx_pause = 0;
3040 
3041 	if (rx_pause && tx_pause)
3042 		fc_conf->mode = RTE_FC_FULL;
3043 	else if (rx_pause)
3044 		fc_conf->mode = RTE_FC_RX_PAUSE;
3045 	else if (tx_pause)
3046 		fc_conf->mode = RTE_FC_TX_PAUSE;
3047 	else
3048 		fc_conf->mode = RTE_FC_NONE;
3049 
3050 	return 0;
3051 }
3052 
3053 static int
3054 eth_igb_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3055 {
3056 	struct e1000_hw *hw;
3057 	int err;
3058 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
3059 		e1000_fc_none,
3060 		e1000_fc_rx_pause,
3061 		e1000_fc_tx_pause,
3062 		e1000_fc_full
3063 	};
3064 	uint32_t rx_buf_size;
3065 	uint32_t max_high_water;
3066 	uint32_t rctl;
3067 	uint32_t ctrl;
3068 
3069 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3070 	if (fc_conf->autoneg != hw->mac.autoneg)
3071 		return -ENOTSUP;
3072 	rx_buf_size = igb_get_rx_buffer_size(hw);
3073 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3074 
3075 	/* At least reserve one Ethernet frame for watermark */
3076 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
3077 	if ((fc_conf->high_water > max_high_water) ||
3078 	    (fc_conf->high_water < fc_conf->low_water)) {
3079 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
3080 		PMD_INIT_LOG(ERR, "high water must <=  0x%x", max_high_water);
3081 		return -EINVAL;
3082 	}
3083 
3084 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
3085 	hw->fc.pause_time     = fc_conf->pause_time;
3086 	hw->fc.high_water     = fc_conf->high_water;
3087 	hw->fc.low_water      = fc_conf->low_water;
3088 	hw->fc.send_xon	      = fc_conf->send_xon;
3089 
3090 	err = e1000_setup_link_generic(hw);
3091 	if (err == E1000_SUCCESS) {
3092 
3093 		/* check if we want to forward MAC frames - driver doesn't have native
3094 		 * capability to do that, so we'll write the registers ourselves */
3095 
3096 		rctl = E1000_READ_REG(hw, E1000_RCTL);
3097 
3098 		/* set or clear MFLCN.PMCF bit depending on configuration */
3099 		if (fc_conf->mac_ctrl_frame_fwd != 0)
3100 			rctl |= E1000_RCTL_PMCF;
3101 		else
3102 			rctl &= ~E1000_RCTL_PMCF;
3103 
3104 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3105 
3106 		/*
3107 		 * check if we want to change flow control mode - driver doesn't have native
3108 		 * capability to do that, so we'll write the registers ourselves
3109 		 */
3110 		ctrl = E1000_READ_REG(hw, E1000_CTRL);
3111 
3112 		/*
3113 		 * set or clear E1000_CTRL_RFCE and E1000_CTRL_TFCE bits depending
3114 		 * on configuration
3115 		 */
3116 		switch (fc_conf->mode) {
3117 		case RTE_FC_NONE:
3118 			ctrl &= ~E1000_CTRL_RFCE & ~E1000_CTRL_TFCE;
3119 			break;
3120 		case RTE_FC_RX_PAUSE:
3121 			ctrl |= E1000_CTRL_RFCE;
3122 			ctrl &= ~E1000_CTRL_TFCE;
3123 			break;
3124 		case RTE_FC_TX_PAUSE:
3125 			ctrl |= E1000_CTRL_TFCE;
3126 			ctrl &= ~E1000_CTRL_RFCE;
3127 			break;
3128 		case RTE_FC_FULL:
3129 			ctrl |= E1000_CTRL_RFCE | E1000_CTRL_TFCE;
3130 			break;
3131 		default:
3132 			PMD_INIT_LOG(ERR, "invalid flow control mode");
3133 			return -EINVAL;
3134 		}
3135 
3136 		E1000_WRITE_REG(hw, E1000_CTRL, ctrl);
3137 
3138 		E1000_WRITE_FLUSH(hw);
3139 
3140 		return 0;
3141 	}
3142 
3143 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
3144 	return -EIO;
3145 }
3146 
3147 #define E1000_RAH_POOLSEL_SHIFT      (18)
3148 static int
3149 eth_igb_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
3150 		uint32_t index, uint32_t pool)
3151 {
3152 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3153 	uint32_t rah;
3154 
3155 	e1000_rar_set(hw, mac_addr->addr_bytes, index);
3156 	rah = E1000_READ_REG(hw, E1000_RAH(index));
3157 	rah |= (0x1 << (E1000_RAH_POOLSEL_SHIFT + pool));
3158 	E1000_WRITE_REG(hw, E1000_RAH(index), rah);
3159 	return 0;
3160 }
3161 
3162 static void
3163 eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index)
3164 {
3165 	uint8_t addr[RTE_ETHER_ADDR_LEN];
3166 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3167 
3168 	memset(addr, 0, sizeof(addr));
3169 
3170 	e1000_rar_set(hw, addr, index);
3171 }
3172 
3173 static int
3174 eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
3175 				struct rte_ether_addr *addr)
3176 {
3177 	eth_igb_rar_clear(dev, 0);
3178 	eth_igb_rar_set(dev, (void *)addr, 0, 0);
3179 
3180 	return 0;
3181 }
3182 /*
3183  * Virtual Function operations
3184  */
3185 static void
3186 igbvf_intr_disable(struct e1000_hw *hw)
3187 {
3188 	PMD_INIT_FUNC_TRACE();
3189 
3190 	/* Clear interrupt mask to stop from interrupts being generated */
3191 	E1000_WRITE_REG(hw, E1000_EIMC, 0xFFFF);
3192 
3193 	E1000_WRITE_FLUSH(hw);
3194 }
3195 
3196 static void
3197 igbvf_stop_adapter(struct rte_eth_dev *dev)
3198 {
3199 	u32 reg_val;
3200 	u16 i;
3201 	struct rte_eth_dev_info dev_info;
3202 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3203 	int ret;
3204 
3205 	memset(&dev_info, 0, sizeof(dev_info));
3206 	ret = eth_igbvf_infos_get(dev, &dev_info);
3207 	if (ret != 0)
3208 		return;
3209 
3210 	/* Clear interrupt mask to stop from interrupts being generated */
3211 	igbvf_intr_disable(hw);
3212 
3213 	/* Clear any pending interrupts, flush previous writes */
3214 	E1000_READ_REG(hw, E1000_EICR);
3215 
3216 	/* Disable the transmit unit.  Each queue must be disabled. */
3217 	for (i = 0; i < dev_info.max_tx_queues; i++)
3218 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), E1000_TXDCTL_SWFLSH);
3219 
3220 	/* Disable the receive unit by stopping each queue */
3221 	for (i = 0; i < dev_info.max_rx_queues; i++) {
3222 		reg_val = E1000_READ_REG(hw, E1000_RXDCTL(i));
3223 		reg_val &= ~E1000_RXDCTL_QUEUE_ENABLE;
3224 		E1000_WRITE_REG(hw, E1000_RXDCTL(i), reg_val);
3225 		while (E1000_READ_REG(hw, E1000_RXDCTL(i)) & E1000_RXDCTL_QUEUE_ENABLE)
3226 			;
3227 	}
3228 
3229 	/* flush all queues disables */
3230 	E1000_WRITE_FLUSH(hw);
3231 	msec_delay(2);
3232 }
3233 
3234 static int eth_igbvf_link_update(struct e1000_hw *hw)
3235 {
3236 	struct e1000_mbx_info *mbx = &hw->mbx;
3237 	struct e1000_mac_info *mac = &hw->mac;
3238 	int ret_val = E1000_SUCCESS;
3239 
3240 	PMD_INIT_LOG(DEBUG, "e1000_check_for_link_vf");
3241 
3242 	/*
3243 	 * We only want to run this if there has been a rst asserted.
3244 	 * in this case that could mean a link change, device reset,
3245 	 * or a virtual function reset
3246 	 */
3247 
3248 	/* If we were hit with a reset or timeout drop the link */
3249 	if (!e1000_check_for_rst(hw, 0) || !mbx->timeout)
3250 		mac->get_link_status = TRUE;
3251 
3252 	if (!mac->get_link_status)
3253 		goto out;
3254 
3255 	/* if link status is down no point in checking to see if pf is up */
3256 	if (!(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU))
3257 		goto out;
3258 
3259 	/* if we passed all the tests above then the link is up and we no
3260 	 * longer need to check for link */
3261 	mac->get_link_status = FALSE;
3262 
3263 out:
3264 	return ret_val;
3265 }
3266 
3267 
3268 static int
3269 igbvf_dev_configure(struct rte_eth_dev *dev)
3270 {
3271 	struct rte_eth_conf* conf = &dev->data->dev_conf;
3272 
3273 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
3274 		     dev->data->port_id);
3275 
3276 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
3277 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
3278 
3279 	/*
3280 	 * VF has no ability to enable/disable HW CRC
3281 	 * Keep the persistent behavior the same as Host PF
3282 	 */
3283 #ifndef RTE_LIBRTE_E1000_PF_DISABLE_STRIP_CRC
3284 	if (conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
3285 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
3286 		conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_KEEP_CRC;
3287 	}
3288 #else
3289 	if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC)) {
3290 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
3291 		conf->rxmode.offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
3292 	}
3293 #endif
3294 
3295 	return 0;
3296 }
3297 
3298 static int
3299 igbvf_dev_start(struct rte_eth_dev *dev)
3300 {
3301 	struct e1000_hw *hw =
3302 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3303 	struct e1000_adapter *adapter =
3304 		E1000_DEV_PRIVATE(dev->data->dev_private);
3305 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3306 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3307 	int ret;
3308 	uint32_t intr_vector = 0;
3309 
3310 	PMD_INIT_FUNC_TRACE();
3311 
3312 	hw->mac.ops.reset_hw(hw);
3313 	adapter->stopped = 0;
3314 
3315 	/* Set all vfta */
3316 	igbvf_set_vfta_all(dev,1);
3317 
3318 	eth_igbvf_tx_init(dev);
3319 
3320 	/* This can fail when allocating mbufs for descriptor rings */
3321 	ret = eth_igbvf_rx_init(dev);
3322 	if (ret) {
3323 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
3324 		igb_dev_clear_queues(dev);
3325 		return ret;
3326 	}
3327 
3328 	/* check and configure queue intr-vector mapping */
3329 	if (rte_intr_cap_multiple(intr_handle) &&
3330 	    dev->data->dev_conf.intr_conf.rxq) {
3331 		intr_vector = dev->data->nb_rx_queues;
3332 		ret = rte_intr_efd_enable(intr_handle, intr_vector);
3333 		if (ret)
3334 			return ret;
3335 	}
3336 
3337 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
3338 		intr_handle->intr_vec =
3339 			rte_zmalloc("intr_vec",
3340 				    dev->data->nb_rx_queues * sizeof(int), 0);
3341 		if (!intr_handle->intr_vec) {
3342 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
3343 				     " intr_vec", dev->data->nb_rx_queues);
3344 			return -ENOMEM;
3345 		}
3346 	}
3347 
3348 	eth_igbvf_configure_msix_intr(dev);
3349 
3350 	/* enable uio/vfio intr/eventfd mapping */
3351 	rte_intr_enable(intr_handle);
3352 
3353 	/* resume enabled intr since hw reset */
3354 	igbvf_intr_enable(dev);
3355 
3356 	return 0;
3357 }
3358 
3359 static int
3360 igbvf_dev_stop(struct rte_eth_dev *dev)
3361 {
3362 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3363 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3364 	struct e1000_adapter *adapter =
3365 		E1000_DEV_PRIVATE(dev->data->dev_private);
3366 
3367 	if (adapter->stopped)
3368 		return 0;
3369 
3370 	PMD_INIT_FUNC_TRACE();
3371 
3372 	igbvf_stop_adapter(dev);
3373 
3374 	/*
3375 	  * Clear what we set, but we still keep shadow_vfta to
3376 	  * restore after device starts
3377 	  */
3378 	igbvf_set_vfta_all(dev,0);
3379 
3380 	igb_dev_clear_queues(dev);
3381 
3382 	/* disable intr eventfd mapping */
3383 	rte_intr_disable(intr_handle);
3384 
3385 	/* Clean datapath event and queue/vec mapping */
3386 	rte_intr_efd_disable(intr_handle);
3387 	if (intr_handle->intr_vec) {
3388 		rte_free(intr_handle->intr_vec);
3389 		intr_handle->intr_vec = NULL;
3390 	}
3391 
3392 	adapter->stopped = true;
3393 	dev->data->dev_started = 0;
3394 
3395 	return 0;
3396 }
3397 
3398 static int
3399 igbvf_dev_close(struct rte_eth_dev *dev)
3400 {
3401 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3402 	struct rte_ether_addr addr;
3403 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3404 	int ret;
3405 
3406 	PMD_INIT_FUNC_TRACE();
3407 
3408 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
3409 		return 0;
3410 
3411 	e1000_reset_hw(hw);
3412 
3413 	ret = igbvf_dev_stop(dev);
3414 	if (ret != 0)
3415 		return ret;
3416 
3417 	igb_dev_free_queues(dev);
3418 
3419 	/**
3420 	 * reprogram the RAR with a zero mac address,
3421 	 * to ensure that the VF traffic goes to the PF
3422 	 * after stop, close and detach of the VF.
3423 	 **/
3424 
3425 	memset(&addr, 0, sizeof(addr));
3426 	igbvf_default_mac_addr_set(dev, &addr);
3427 
3428 	rte_intr_callback_unregister(&pci_dev->intr_handle,
3429 				     eth_igbvf_interrupt_handler,
3430 				     (void *)dev);
3431 
3432 	return 0;
3433 }
3434 
3435 static int
3436 igbvf_promiscuous_enable(struct rte_eth_dev *dev)
3437 {
3438 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3439 
3440 	/* Set both unicast and multicast promisc */
3441 	e1000_promisc_set_vf(hw, e1000_promisc_enabled);
3442 
3443 	return 0;
3444 }
3445 
3446 static int
3447 igbvf_promiscuous_disable(struct rte_eth_dev *dev)
3448 {
3449 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3450 
3451 	/* If in allmulticast mode leave multicast promisc */
3452 	if (dev->data->all_multicast == 1)
3453 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3454 	else
3455 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3456 
3457 	return 0;
3458 }
3459 
3460 static int
3461 igbvf_allmulticast_enable(struct rte_eth_dev *dev)
3462 {
3463 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3464 
3465 	/* In promiscuous mode multicast promisc already set */
3466 	if (dev->data->promiscuous == 0)
3467 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3468 
3469 	return 0;
3470 }
3471 
3472 static int
3473 igbvf_allmulticast_disable(struct rte_eth_dev *dev)
3474 {
3475 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3476 
3477 	/* In promiscuous mode leave multicast promisc enabled */
3478 	if (dev->data->promiscuous == 0)
3479 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3480 
3481 	return 0;
3482 }
3483 
3484 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on)
3485 {
3486 	struct e1000_mbx_info *mbx = &hw->mbx;
3487 	uint32_t msgbuf[2];
3488 	s32 err;
3489 
3490 	/* After set vlan, vlan strip will also be enabled in igb driver*/
3491 	msgbuf[0] = E1000_VF_SET_VLAN;
3492 	msgbuf[1] = vid;
3493 	/* Setting the 8 bit field MSG INFO to TRUE indicates "add" */
3494 	if (on)
3495 		msgbuf[0] |= E1000_VF_SET_VLAN_ADD;
3496 
3497 	err = mbx->ops.write_posted(hw, msgbuf, 2, 0);
3498 	if (err)
3499 		goto mbx_err;
3500 
3501 	err = mbx->ops.read_posted(hw, msgbuf, 2, 0);
3502 	if (err)
3503 		goto mbx_err;
3504 
3505 	msgbuf[0] &= ~E1000_VT_MSGTYPE_CTS;
3506 	if (msgbuf[0] == (E1000_VF_SET_VLAN | E1000_VT_MSGTYPE_NACK))
3507 		err = -EINVAL;
3508 
3509 mbx_err:
3510 	return err;
3511 }
3512 
3513 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on)
3514 {
3515 	struct e1000_hw *hw =
3516 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3517 	struct e1000_vfta * shadow_vfta =
3518 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3519 	int i = 0, j = 0, vfta = 0, mask = 1;
3520 
3521 	for (i = 0; i < IGB_VFTA_SIZE; i++){
3522 		vfta = shadow_vfta->vfta[i];
3523 		if(vfta){
3524 			mask = 1;
3525 			for (j = 0; j < 32; j++){
3526 				if(vfta & mask)
3527 					igbvf_set_vfta(hw,
3528 						(uint16_t)((i<<5)+j), on);
3529 				mask<<=1;
3530 			}
3531 		}
3532 	}
3533 
3534 }
3535 
3536 static int
3537 igbvf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
3538 {
3539 	struct e1000_hw *hw =
3540 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3541 	struct e1000_vfta * shadow_vfta =
3542 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3543 	uint32_t vid_idx = 0;
3544 	uint32_t vid_bit = 0;
3545 	int ret = 0;
3546 
3547 	PMD_INIT_FUNC_TRACE();
3548 
3549 	/*vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf*/
3550 	ret = igbvf_set_vfta(hw, vlan_id, !!on);
3551 	if(ret){
3552 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
3553 		return ret;
3554 	}
3555 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
3556 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
3557 
3558 	/*Save what we set and retore it after device reset*/
3559 	if (on)
3560 		shadow_vfta->vfta[vid_idx] |= vid_bit;
3561 	else
3562 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
3563 
3564 	return 0;
3565 }
3566 
3567 static int
3568 igbvf_default_mac_addr_set(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
3569 {
3570 	struct e1000_hw *hw =
3571 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3572 
3573 	/* index is not used by rar_set() */
3574 	hw->mac.ops.rar_set(hw, (void *)addr, 0);
3575 	return 0;
3576 }
3577 
3578 
3579 static int
3580 eth_igb_rss_reta_update(struct rte_eth_dev *dev,
3581 			struct rte_eth_rss_reta_entry64 *reta_conf,
3582 			uint16_t reta_size)
3583 {
3584 	uint8_t i, j, mask;
3585 	uint32_t reta, r;
3586 	uint16_t idx, shift;
3587 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3588 
3589 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3590 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3591 			"(%d) doesn't match the number hardware can supported "
3592 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3593 		return -EINVAL;
3594 	}
3595 
3596 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3597 		idx = i / RTE_RETA_GROUP_SIZE;
3598 		shift = i % RTE_RETA_GROUP_SIZE;
3599 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3600 						IGB_4_BIT_MASK);
3601 		if (!mask)
3602 			continue;
3603 		if (mask == IGB_4_BIT_MASK)
3604 			r = 0;
3605 		else
3606 			r = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3607 		for (j = 0, reta = 0; j < IGB_4_BIT_WIDTH; j++) {
3608 			if (mask & (0x1 << j))
3609 				reta |= reta_conf[idx].reta[shift + j] <<
3610 							(CHAR_BIT * j);
3611 			else
3612 				reta |= r & (IGB_8_BIT_MASK << (CHAR_BIT * j));
3613 		}
3614 		E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
3615 	}
3616 
3617 	return 0;
3618 }
3619 
3620 static int
3621 eth_igb_rss_reta_query(struct rte_eth_dev *dev,
3622 		       struct rte_eth_rss_reta_entry64 *reta_conf,
3623 		       uint16_t reta_size)
3624 {
3625 	uint8_t i, j, mask;
3626 	uint32_t reta;
3627 	uint16_t idx, shift;
3628 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3629 
3630 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3631 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3632 			"(%d) doesn't match the number hardware can supported "
3633 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3634 		return -EINVAL;
3635 	}
3636 
3637 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3638 		idx = i / RTE_RETA_GROUP_SIZE;
3639 		shift = i % RTE_RETA_GROUP_SIZE;
3640 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3641 						IGB_4_BIT_MASK);
3642 		if (!mask)
3643 			continue;
3644 		reta = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3645 		for (j = 0; j < IGB_4_BIT_WIDTH; j++) {
3646 			if (mask & (0x1 << j))
3647 				reta_conf[idx].reta[shift + j] =
3648 					((reta >> (CHAR_BIT * j)) &
3649 						IGB_8_BIT_MASK);
3650 		}
3651 	}
3652 
3653 	return 0;
3654 }
3655 
3656 int
3657 eth_igb_syn_filter_set(struct rte_eth_dev *dev,
3658 			struct rte_eth_syn_filter *filter,
3659 			bool add)
3660 {
3661 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3662 	struct e1000_filter_info *filter_info =
3663 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3664 	uint32_t synqf, rfctl;
3665 
3666 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3667 		return -EINVAL;
3668 
3669 	synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3670 
3671 	if (add) {
3672 		if (synqf & E1000_SYN_FILTER_ENABLE)
3673 			return -EINVAL;
3674 
3675 		synqf = (uint32_t)(((filter->queue << E1000_SYN_FILTER_QUEUE_SHIFT) &
3676 			E1000_SYN_FILTER_QUEUE) | E1000_SYN_FILTER_ENABLE);
3677 
3678 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3679 		if (filter->hig_pri)
3680 			rfctl |= E1000_RFCTL_SYNQFP;
3681 		else
3682 			rfctl &= ~E1000_RFCTL_SYNQFP;
3683 
3684 		E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3685 	} else {
3686 		if (!(synqf & E1000_SYN_FILTER_ENABLE))
3687 			return -ENOENT;
3688 		synqf = 0;
3689 	}
3690 
3691 	filter_info->syn_info = synqf;
3692 	E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
3693 	E1000_WRITE_FLUSH(hw);
3694 	return 0;
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 > RTE_NTUPLE_TCP_FLAGS_MASK)
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 & RTE_NTUPLE_TCP_FLAGS_MASK) {
3787 		if (filter->filter_info.tcp_flags & RTE_TCP_URG_FLAG)
3788 			imir_ext |= E1000_IMIREXT_CTRL_URG;
3789 		if (filter->filter_info.tcp_flags & RTE_TCP_ACK_FLAG)
3790 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
3791 		if (filter->filter_info.tcp_flags & RTE_TCP_PSH_FLAG)
3792 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
3793 		if (filter->filter_info.tcp_flags & RTE_TCP_RST_FLAG)
3794 			imir_ext |= E1000_IMIREXT_CTRL_RST;
3795 		if (filter->filter_info.tcp_flags & RTE_TCP_SYN_FLAG)
3796 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
3797 		if (filter->filter_info.tcp_flags & RTE_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 igb_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 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_5tuple_filter_info*/
4088 static inline int
4089 ntuple_filter_to_5tuple_82576(struct rte_eth_ntuple_filter *filter,
4090 			struct e1000_5tuple_filter_info *filter_info)
4091 {
4092 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM_82576)
4093 		return -EINVAL;
4094 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
4095 		return -EINVAL;  /* filter index is out of range. */
4096 	if (filter->tcp_flags > RTE_NTUPLE_TCP_FLAGS_MASK)
4097 		return -EINVAL;  /* flags is invalid. */
4098 
4099 	switch (filter->dst_ip_mask) {
4100 	case UINT32_MAX:
4101 		filter_info->dst_ip_mask = 0;
4102 		filter_info->dst_ip = filter->dst_ip;
4103 		break;
4104 	case 0:
4105 		filter_info->dst_ip_mask = 1;
4106 		break;
4107 	default:
4108 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
4109 		return -EINVAL;
4110 	}
4111 
4112 	switch (filter->src_ip_mask) {
4113 	case UINT32_MAX:
4114 		filter_info->src_ip_mask = 0;
4115 		filter_info->src_ip = filter->src_ip;
4116 		break;
4117 	case 0:
4118 		filter_info->src_ip_mask = 1;
4119 		break;
4120 	default:
4121 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
4122 		return -EINVAL;
4123 	}
4124 
4125 	switch (filter->dst_port_mask) {
4126 	case UINT16_MAX:
4127 		filter_info->dst_port_mask = 0;
4128 		filter_info->dst_port = filter->dst_port;
4129 		break;
4130 	case 0:
4131 		filter_info->dst_port_mask = 1;
4132 		break;
4133 	default:
4134 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
4135 		return -EINVAL;
4136 	}
4137 
4138 	switch (filter->src_port_mask) {
4139 	case UINT16_MAX:
4140 		filter_info->src_port_mask = 0;
4141 		filter_info->src_port = filter->src_port;
4142 		break;
4143 	case 0:
4144 		filter_info->src_port_mask = 1;
4145 		break;
4146 	default:
4147 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
4148 		return -EINVAL;
4149 	}
4150 
4151 	switch (filter->proto_mask) {
4152 	case UINT8_MAX:
4153 		filter_info->proto_mask = 0;
4154 		filter_info->proto = filter->proto;
4155 		break;
4156 	case 0:
4157 		filter_info->proto_mask = 1;
4158 		break;
4159 	default:
4160 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
4161 		return -EINVAL;
4162 	}
4163 
4164 	filter_info->priority = (uint8_t)filter->priority;
4165 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
4166 		filter_info->tcp_flags = filter->tcp_flags;
4167 	else
4168 		filter_info->tcp_flags = 0;
4169 
4170 	return 0;
4171 }
4172 
4173 static inline struct e1000_5tuple_filter *
4174 igb_5tuple_filter_lookup_82576(struct e1000_5tuple_filter_list *filter_list,
4175 			struct e1000_5tuple_filter_info *key)
4176 {
4177 	struct e1000_5tuple_filter *it;
4178 
4179 	TAILQ_FOREACH(it, filter_list, entries) {
4180 		if (memcmp(key, &it->filter_info,
4181 			sizeof(struct e1000_5tuple_filter_info)) == 0) {
4182 			return it;
4183 		}
4184 	}
4185 	return NULL;
4186 }
4187 
4188 /* inject a igb 5-tuple filter to HW */
4189 static inline void
4190 igb_inject_5tuple_filter_82576(struct rte_eth_dev *dev,
4191 			   struct e1000_5tuple_filter *filter)
4192 {
4193 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4194 	uint32_t ftqf = E1000_FTQF_VF_BP | E1000_FTQF_MASK;
4195 	uint32_t spqf, imir, imir_ext = E1000_IMIREXT_SIZE_BP;
4196 	uint8_t i;
4197 
4198 	i = filter->index;
4199 	ftqf |= filter->filter_info.proto & E1000_FTQF_PROTOCOL_MASK;
4200 	if (filter->filter_info.src_ip_mask == 0) /* 0b means compare. */
4201 		ftqf &= ~E1000_FTQF_MASK_SOURCE_ADDR_BP;
4202 	if (filter->filter_info.dst_ip_mask == 0)
4203 		ftqf &= ~E1000_FTQF_MASK_DEST_ADDR_BP;
4204 	if (filter->filter_info.src_port_mask == 0)
4205 		ftqf &= ~E1000_FTQF_MASK_SOURCE_PORT_BP;
4206 	if (filter->filter_info.proto_mask == 0)
4207 		ftqf &= ~E1000_FTQF_MASK_PROTO_BP;
4208 	ftqf |= (filter->queue << E1000_FTQF_QUEUE_SHIFT) &
4209 		E1000_FTQF_QUEUE_MASK;
4210 	ftqf |= E1000_FTQF_QUEUE_ENABLE;
4211 	E1000_WRITE_REG(hw, E1000_FTQF(i), ftqf);
4212 	E1000_WRITE_REG(hw, E1000_DAQF(i), filter->filter_info.dst_ip);
4213 	E1000_WRITE_REG(hw, E1000_SAQF(i), filter->filter_info.src_ip);
4214 
4215 	spqf = filter->filter_info.src_port & E1000_SPQF_SRCPORT;
4216 	E1000_WRITE_REG(hw, E1000_SPQF(i), spqf);
4217 
4218 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
4219 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
4220 		imir |= E1000_IMIR_PORT_BP;
4221 	else
4222 		imir &= ~E1000_IMIR_PORT_BP;
4223 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
4224 
4225 	/* tcp flags bits setting. */
4226 	if (filter->filter_info.tcp_flags & RTE_NTUPLE_TCP_FLAGS_MASK) {
4227 		if (filter->filter_info.tcp_flags & RTE_TCP_URG_FLAG)
4228 			imir_ext |= E1000_IMIREXT_CTRL_URG;
4229 		if (filter->filter_info.tcp_flags & RTE_TCP_ACK_FLAG)
4230 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
4231 		if (filter->filter_info.tcp_flags & RTE_TCP_PSH_FLAG)
4232 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
4233 		if (filter->filter_info.tcp_flags & RTE_TCP_RST_FLAG)
4234 			imir_ext |= E1000_IMIREXT_CTRL_RST;
4235 		if (filter->filter_info.tcp_flags & RTE_TCP_SYN_FLAG)
4236 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
4237 		if (filter->filter_info.tcp_flags & RTE_TCP_FIN_FLAG)
4238 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
4239 	} else {
4240 		imir_ext |= E1000_IMIREXT_CTRL_BP;
4241 	}
4242 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
4243 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
4244 }
4245 
4246 /*
4247  * igb_add_5tuple_filter_82576 - add a 5tuple filter
4248  *
4249  * @param
4250  * dev: Pointer to struct rte_eth_dev.
4251  * ntuple_filter: ponter to the filter that will be added.
4252  *
4253  * @return
4254  *    - On success, zero.
4255  *    - On failure, a negative value.
4256  */
4257 static int
4258 igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
4259 			struct rte_eth_ntuple_filter *ntuple_filter)
4260 {
4261 	struct e1000_filter_info *filter_info =
4262 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4263 	struct e1000_5tuple_filter *filter;
4264 	uint8_t i;
4265 	int ret;
4266 
4267 	filter = rte_zmalloc("e1000_5tuple_filter",
4268 			sizeof(struct e1000_5tuple_filter), 0);
4269 	if (filter == NULL)
4270 		return -ENOMEM;
4271 
4272 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4273 					    &filter->filter_info);
4274 	if (ret < 0) {
4275 		rte_free(filter);
4276 		return ret;
4277 	}
4278 
4279 	if (igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4280 					 &filter->filter_info) != NULL) {
4281 		PMD_DRV_LOG(ERR, "filter exists.");
4282 		rte_free(filter);
4283 		return -EEXIST;
4284 	}
4285 	filter->queue = ntuple_filter->queue;
4286 
4287 	/*
4288 	 * look for an unused 5tuple filter index,
4289 	 * and insert the filter to list.
4290 	 */
4291 	for (i = 0; i < E1000_MAX_FTQF_FILTERS; i++) {
4292 		if (!(filter_info->fivetuple_mask & (1 << i))) {
4293 			filter_info->fivetuple_mask |= 1 << i;
4294 			filter->index = i;
4295 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
4296 					  filter,
4297 					  entries);
4298 			break;
4299 		}
4300 	}
4301 	if (i >= E1000_MAX_FTQF_FILTERS) {
4302 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
4303 		rte_free(filter);
4304 		return -ENOSYS;
4305 	}
4306 
4307 	igb_inject_5tuple_filter_82576(dev, filter);
4308 	return 0;
4309 }
4310 
4311 int
4312 igb_delete_5tuple_filter_82576(struct rte_eth_dev *dev,
4313 				struct e1000_5tuple_filter *filter)
4314 {
4315 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4316 	struct e1000_filter_info *filter_info =
4317 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4318 
4319 	filter_info->fivetuple_mask &= ~(1 << filter->index);
4320 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
4321 	rte_free(filter);
4322 
4323 	E1000_WRITE_REG(hw, E1000_FTQF(filter->index),
4324 			E1000_FTQF_VF_BP | E1000_FTQF_MASK);
4325 	E1000_WRITE_REG(hw, E1000_DAQF(filter->index), 0);
4326 	E1000_WRITE_REG(hw, E1000_SAQF(filter->index), 0);
4327 	E1000_WRITE_REG(hw, E1000_SPQF(filter->index), 0);
4328 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
4329 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
4330 	return 0;
4331 }
4332 
4333 /*
4334  * igb_remove_5tuple_filter_82576 - remove a 5tuple filter
4335  *
4336  * @param
4337  * dev: Pointer to struct rte_eth_dev.
4338  * ntuple_filter: ponter to the filter that will be removed.
4339  *
4340  * @return
4341  *    - On success, zero.
4342  *    - On failure, a negative value.
4343  */
4344 static int
4345 igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
4346 				struct rte_eth_ntuple_filter *ntuple_filter)
4347 {
4348 	struct e1000_filter_info *filter_info =
4349 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4350 	struct e1000_5tuple_filter_info filter_5tuple;
4351 	struct e1000_5tuple_filter *filter;
4352 	int ret;
4353 
4354 	memset(&filter_5tuple, 0, sizeof(struct e1000_5tuple_filter_info));
4355 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4356 					    &filter_5tuple);
4357 	if (ret < 0)
4358 		return ret;
4359 
4360 	filter = igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4361 					 &filter_5tuple);
4362 	if (filter == NULL) {
4363 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4364 		return -ENOENT;
4365 	}
4366 
4367 	igb_delete_5tuple_filter_82576(dev, filter);
4368 
4369 	return 0;
4370 }
4371 
4372 static int
4373 eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4374 {
4375 	uint32_t rctl;
4376 	struct e1000_hw *hw;
4377 	struct rte_eth_dev_info dev_info;
4378 	uint32_t frame_size = mtu + E1000_ETH_OVERHEAD;
4379 	int ret;
4380 
4381 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4382 
4383 #ifdef RTE_LIBRTE_82571_SUPPORT
4384 	/* XXX: not bigger than max_rx_pktlen */
4385 	if (hw->mac.type == e1000_82571)
4386 		return -ENOTSUP;
4387 #endif
4388 	ret = eth_igb_infos_get(dev, &dev_info);
4389 	if (ret != 0)
4390 		return ret;
4391 
4392 	/* check that mtu is within the allowed range */
4393 	if (mtu < RTE_ETHER_MIN_MTU ||
4394 			frame_size > dev_info.max_rx_pktlen)
4395 		return -EINVAL;
4396 
4397 	/* refuse mtu that requires the support of scattered packets when this
4398 	 * feature has not been enabled before. */
4399 	if (!dev->data->scattered_rx &&
4400 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
4401 		return -EINVAL;
4402 
4403 	rctl = E1000_READ_REG(hw, E1000_RCTL);
4404 
4405 	/* switch to jumbo mode if needed */
4406 	if (frame_size > E1000_ETH_MAX_LEN) {
4407 		dev->data->dev_conf.rxmode.offloads |=
4408 			DEV_RX_OFFLOAD_JUMBO_FRAME;
4409 		rctl |= E1000_RCTL_LPE;
4410 	} else {
4411 		dev->data->dev_conf.rxmode.offloads &=
4412 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
4413 		rctl &= ~E1000_RCTL_LPE;
4414 	}
4415 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
4416 
4417 	/* update max frame size */
4418 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4419 
4420 	E1000_WRITE_REG(hw, E1000_RLPML,
4421 			dev->data->dev_conf.rxmode.max_rx_pkt_len);
4422 
4423 	return 0;
4424 }
4425 
4426 /*
4427  * igb_add_del_ntuple_filter - add or delete a ntuple filter
4428  *
4429  * @param
4430  * dev: Pointer to struct rte_eth_dev.
4431  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4432  * add: if true, add filter, if false, remove filter
4433  *
4434  * @return
4435  *    - On success, zero.
4436  *    - On failure, a negative value.
4437  */
4438 int
4439 igb_add_del_ntuple_filter(struct rte_eth_dev *dev,
4440 			struct rte_eth_ntuple_filter *ntuple_filter,
4441 			bool add)
4442 {
4443 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4444 	int ret;
4445 
4446 	switch (ntuple_filter->flags) {
4447 	case RTE_5TUPLE_FLAGS:
4448 	case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4449 		if (hw->mac.type != e1000_82576)
4450 			return -ENOTSUP;
4451 		if (add)
4452 			ret = igb_add_5tuple_filter_82576(dev,
4453 							  ntuple_filter);
4454 		else
4455 			ret = igb_remove_5tuple_filter_82576(dev,
4456 							     ntuple_filter);
4457 		break;
4458 	case RTE_2TUPLE_FLAGS:
4459 	case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4460 		if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350 &&
4461 			hw->mac.type != e1000_i210 &&
4462 			hw->mac.type != e1000_i211)
4463 			return -ENOTSUP;
4464 		if (add)
4465 			ret = igb_add_2tuple_filter(dev, ntuple_filter);
4466 		else
4467 			ret = igb_remove_2tuple_filter(dev, ntuple_filter);
4468 		break;
4469 	default:
4470 		ret = -EINVAL;
4471 		break;
4472 	}
4473 
4474 	return ret;
4475 }
4476 
4477 static inline int
4478 igb_ethertype_filter_lookup(struct e1000_filter_info *filter_info,
4479 			uint16_t ethertype)
4480 {
4481 	int i;
4482 
4483 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4484 		if (filter_info->ethertype_filters[i].ethertype == ethertype &&
4485 		    (filter_info->ethertype_mask & (1 << i)))
4486 			return i;
4487 	}
4488 	return -1;
4489 }
4490 
4491 static inline int
4492 igb_ethertype_filter_insert(struct e1000_filter_info *filter_info,
4493 			uint16_t ethertype, uint32_t etqf)
4494 {
4495 	int i;
4496 
4497 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4498 		if (!(filter_info->ethertype_mask & (1 << i))) {
4499 			filter_info->ethertype_mask |= 1 << i;
4500 			filter_info->ethertype_filters[i].ethertype = ethertype;
4501 			filter_info->ethertype_filters[i].etqf = etqf;
4502 			return i;
4503 		}
4504 	}
4505 	return -1;
4506 }
4507 
4508 int
4509 igb_ethertype_filter_remove(struct e1000_filter_info *filter_info,
4510 			uint8_t idx)
4511 {
4512 	if (idx >= E1000_MAX_ETQF_FILTERS)
4513 		return -1;
4514 	filter_info->ethertype_mask &= ~(1 << idx);
4515 	filter_info->ethertype_filters[idx].ethertype = 0;
4516 	filter_info->ethertype_filters[idx].etqf = 0;
4517 	return idx;
4518 }
4519 
4520 
4521 int
4522 igb_add_del_ethertype_filter(struct rte_eth_dev *dev,
4523 			struct rte_eth_ethertype_filter *filter,
4524 			bool add)
4525 {
4526 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4527 	struct e1000_filter_info *filter_info =
4528 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4529 	uint32_t etqf = 0;
4530 	int ret;
4531 
4532 	if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
4533 		filter->ether_type == RTE_ETHER_TYPE_IPV6) {
4534 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
4535 			" ethertype filter.", filter->ether_type);
4536 		return -EINVAL;
4537 	}
4538 
4539 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
4540 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
4541 		return -EINVAL;
4542 	}
4543 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
4544 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
4545 		return -EINVAL;
4546 	}
4547 
4548 	ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4549 	if (ret >= 0 && add) {
4550 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
4551 			    filter->ether_type);
4552 		return -EEXIST;
4553 	}
4554 	if (ret < 0 && !add) {
4555 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4556 			    filter->ether_type);
4557 		return -ENOENT;
4558 	}
4559 
4560 	if (add) {
4561 		etqf |= E1000_ETQF_FILTER_ENABLE | E1000_ETQF_QUEUE_ENABLE;
4562 		etqf |= (uint32_t)(filter->ether_type & E1000_ETQF_ETHERTYPE);
4563 		etqf |= filter->queue << E1000_ETQF_QUEUE_SHIFT;
4564 		ret = igb_ethertype_filter_insert(filter_info,
4565 				filter->ether_type, etqf);
4566 		if (ret < 0) {
4567 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
4568 			return -ENOSYS;
4569 		}
4570 	} else {
4571 		ret = igb_ethertype_filter_remove(filter_info, (uint8_t)ret);
4572 		if (ret < 0)
4573 			return -ENOSYS;
4574 	}
4575 	E1000_WRITE_REG(hw, E1000_ETQF(ret), etqf);
4576 	E1000_WRITE_FLUSH(hw);
4577 
4578 	return 0;
4579 }
4580 
4581 static int
4582 eth_igb_filter_ctrl(struct rte_eth_dev *dev __rte_unused,
4583 		     enum rte_filter_type filter_type,
4584 		     enum rte_filter_op filter_op,
4585 		     void *arg)
4586 {
4587 	int ret = 0;
4588 
4589 	switch (filter_type) {
4590 	case RTE_ETH_FILTER_GENERIC:
4591 		if (filter_op != RTE_ETH_FILTER_GET)
4592 			return -EINVAL;
4593 		*(const void **)arg = &igb_flow_ops;
4594 		break;
4595 	default:
4596 		PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
4597 							filter_type);
4598 		break;
4599 	}
4600 
4601 	return ret;
4602 }
4603 
4604 static int
4605 eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
4606 			 struct rte_ether_addr *mc_addr_set,
4607 			 uint32_t nb_mc_addr)
4608 {
4609 	struct e1000_hw *hw;
4610 
4611 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4612 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
4613 	return 0;
4614 }
4615 
4616 static uint64_t
4617 igb_read_systime_cyclecounter(struct rte_eth_dev *dev)
4618 {
4619 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4620 	uint64_t systime_cycles;
4621 
4622 	switch (hw->mac.type) {
4623 	case e1000_i210:
4624 	case e1000_i211:
4625 		/*
4626 		 * Need to read System Time Residue Register to be able
4627 		 * to read the other two registers.
4628 		 */
4629 		E1000_READ_REG(hw, E1000_SYSTIMR);
4630 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
4631 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4632 		systime_cycles += (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4633 				* NSEC_PER_SEC;
4634 		break;
4635 	case e1000_82580:
4636 	case e1000_i350:
4637 	case e1000_i354:
4638 		/*
4639 		 * Need to read System Time Residue Register to be able
4640 		 * to read the other two registers.
4641 		 */
4642 		E1000_READ_REG(hw, E1000_SYSTIMR);
4643 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4644 		/* Only the 8 LSB are valid. */
4645 		systime_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_SYSTIMH)
4646 				& 0xff) << 32;
4647 		break;
4648 	default:
4649 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4650 		systime_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4651 				<< 32;
4652 		break;
4653 	}
4654 
4655 	return systime_cycles;
4656 }
4657 
4658 static uint64_t
4659 igb_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
4660 {
4661 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4662 	uint64_t rx_tstamp_cycles;
4663 
4664 	switch (hw->mac.type) {
4665 	case e1000_i210:
4666 	case e1000_i211:
4667 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
4668 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4669 		rx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
4670 				* NSEC_PER_SEC;
4671 		break;
4672 	case e1000_82580:
4673 	case e1000_i350:
4674 	case e1000_i354:
4675 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4676 		/* Only the 8 LSB are valid. */
4677 		rx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_RXSTMPH)
4678 				& 0xff) << 32;
4679 		break;
4680 	default:
4681 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4682 		rx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
4683 				<< 32;
4684 		break;
4685 	}
4686 
4687 	return rx_tstamp_cycles;
4688 }
4689 
4690 static uint64_t
4691 igb_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
4692 {
4693 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4694 	uint64_t tx_tstamp_cycles;
4695 
4696 	switch (hw->mac.type) {
4697 	case e1000_i210:
4698 	case e1000_i211:
4699 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
4700 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
4701 		tx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
4702 				* NSEC_PER_SEC;
4703 		break;
4704 	case e1000_82580:
4705 	case e1000_i350:
4706 	case e1000_i354:
4707 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
4708 		/* Only the 8 LSB are valid. */
4709 		tx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_TXSTMPH)
4710 				& 0xff) << 32;
4711 		break;
4712 	default:
4713 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
4714 		tx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
4715 				<< 32;
4716 		break;
4717 	}
4718 
4719 	return tx_tstamp_cycles;
4720 }
4721 
4722 static void
4723 igb_start_timecounters(struct rte_eth_dev *dev)
4724 {
4725 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4726 	struct e1000_adapter *adapter = dev->data->dev_private;
4727 	uint32_t incval = 1;
4728 	uint32_t shift = 0;
4729 	uint64_t mask = E1000_CYCLECOUNTER_MASK;
4730 
4731 	switch (hw->mac.type) {
4732 	case e1000_82580:
4733 	case e1000_i350:
4734 	case e1000_i354:
4735 		/* 32 LSB bits + 8 MSB bits = 40 bits */
4736 		mask = (1ULL << 40) - 1;
4737 		/* fall-through */
4738 	case e1000_i210:
4739 	case e1000_i211:
4740 		/*
4741 		 * Start incrementing the register
4742 		 * used to timestamp PTP packets.
4743 		 */
4744 		E1000_WRITE_REG(hw, E1000_TIMINCA, incval);
4745 		break;
4746 	case e1000_82576:
4747 		incval = E1000_INCVALUE_82576;
4748 		shift = IGB_82576_TSYNC_SHIFT;
4749 		E1000_WRITE_REG(hw, E1000_TIMINCA,
4750 				E1000_INCPERIOD_82576 | incval);
4751 		break;
4752 	default:
4753 		/* Not supported */
4754 		return;
4755 	}
4756 
4757 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
4758 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
4759 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
4760 
4761 	adapter->systime_tc.cc_mask = mask;
4762 	adapter->systime_tc.cc_shift = shift;
4763 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
4764 
4765 	adapter->rx_tstamp_tc.cc_mask = mask;
4766 	adapter->rx_tstamp_tc.cc_shift = shift;
4767 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
4768 
4769 	adapter->tx_tstamp_tc.cc_mask = mask;
4770 	adapter->tx_tstamp_tc.cc_shift = shift;
4771 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
4772 }
4773 
4774 static int
4775 igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
4776 {
4777 	struct e1000_adapter *adapter = dev->data->dev_private;
4778 
4779 	adapter->systime_tc.nsec += delta;
4780 	adapter->rx_tstamp_tc.nsec += delta;
4781 	adapter->tx_tstamp_tc.nsec += delta;
4782 
4783 	return 0;
4784 }
4785 
4786 static int
4787 igb_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
4788 {
4789 	uint64_t ns;
4790 	struct e1000_adapter *adapter = dev->data->dev_private;
4791 
4792 	ns = rte_timespec_to_ns(ts);
4793 
4794 	/* Set the timecounters to a new value. */
4795 	adapter->systime_tc.nsec = ns;
4796 	adapter->rx_tstamp_tc.nsec = ns;
4797 	adapter->tx_tstamp_tc.nsec = ns;
4798 
4799 	return 0;
4800 }
4801 
4802 static int
4803 igb_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
4804 {
4805 	uint64_t ns, systime_cycles;
4806 	struct e1000_adapter *adapter = dev->data->dev_private;
4807 
4808 	systime_cycles = igb_read_systime_cyclecounter(dev);
4809 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
4810 	*ts = rte_ns_to_timespec(ns);
4811 
4812 	return 0;
4813 }
4814 
4815 static int
4816 igb_timesync_enable(struct rte_eth_dev *dev)
4817 {
4818 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4819 	uint32_t tsync_ctl;
4820 	uint32_t tsauxc;
4821 
4822 	/* Stop the timesync system time. */
4823 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0x0);
4824 	/* Reset the timesync system time value. */
4825 	switch (hw->mac.type) {
4826 	case e1000_82580:
4827 	case e1000_i350:
4828 	case e1000_i354:
4829 	case e1000_i210:
4830 	case e1000_i211:
4831 		E1000_WRITE_REG(hw, E1000_SYSTIMR, 0x0);
4832 		/* fall-through */
4833 	case e1000_82576:
4834 		E1000_WRITE_REG(hw, E1000_SYSTIML, 0x0);
4835 		E1000_WRITE_REG(hw, E1000_SYSTIMH, 0x0);
4836 		break;
4837 	default:
4838 		/* Not supported. */
4839 		return -ENOTSUP;
4840 	}
4841 
4842 	/* Enable system time for it isn't on by default. */
4843 	tsauxc = E1000_READ_REG(hw, E1000_TSAUXC);
4844 	tsauxc &= ~E1000_TSAUXC_DISABLE_SYSTIME;
4845 	E1000_WRITE_REG(hw, E1000_TSAUXC, tsauxc);
4846 
4847 	igb_start_timecounters(dev);
4848 
4849 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
4850 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588),
4851 			(RTE_ETHER_TYPE_1588 |
4852 			 E1000_ETQF_FILTER_ENABLE |
4853 			 E1000_ETQF_1588));
4854 
4855 	/* Enable timestamping of received PTP packets. */
4856 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
4857 	tsync_ctl |= E1000_TSYNCRXCTL_ENABLED;
4858 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
4859 
4860 	/* Enable Timestamping of transmitted PTP packets. */
4861 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
4862 	tsync_ctl |= E1000_TSYNCTXCTL_ENABLED;
4863 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
4864 
4865 	return 0;
4866 }
4867 
4868 static int
4869 igb_timesync_disable(struct rte_eth_dev *dev)
4870 {
4871 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4872 	uint32_t tsync_ctl;
4873 
4874 	/* Disable timestamping of transmitted PTP packets. */
4875 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
4876 	tsync_ctl &= ~E1000_TSYNCTXCTL_ENABLED;
4877 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
4878 
4879 	/* Disable timestamping of received PTP packets. */
4880 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
4881 	tsync_ctl &= ~E1000_TSYNCRXCTL_ENABLED;
4882 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
4883 
4884 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
4885 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588), 0);
4886 
4887 	/* Stop incrementating the System Time registers. */
4888 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0);
4889 
4890 	return 0;
4891 }
4892 
4893 static int
4894 igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
4895 			       struct timespec *timestamp,
4896 			       uint32_t flags __rte_unused)
4897 {
4898 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4899 	struct e1000_adapter *adapter = dev->data->dev_private;
4900 	uint32_t tsync_rxctl;
4901 	uint64_t rx_tstamp_cycles;
4902 	uint64_t ns;
4903 
4904 	tsync_rxctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
4905 	if ((tsync_rxctl & E1000_TSYNCRXCTL_VALID) == 0)
4906 		return -EINVAL;
4907 
4908 	rx_tstamp_cycles = igb_read_rx_tstamp_cyclecounter(dev);
4909 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
4910 	*timestamp = rte_ns_to_timespec(ns);
4911 
4912 	return  0;
4913 }
4914 
4915 static int
4916 igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
4917 			       struct timespec *timestamp)
4918 {
4919 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4920 	struct e1000_adapter *adapter = dev->data->dev_private;
4921 	uint32_t tsync_txctl;
4922 	uint64_t tx_tstamp_cycles;
4923 	uint64_t ns;
4924 
4925 	tsync_txctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
4926 	if ((tsync_txctl & E1000_TSYNCTXCTL_VALID) == 0)
4927 		return -EINVAL;
4928 
4929 	tx_tstamp_cycles = igb_read_tx_tstamp_cyclecounter(dev);
4930 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
4931 	*timestamp = rte_ns_to_timespec(ns);
4932 
4933 	return  0;
4934 }
4935 
4936 static int
4937 eth_igb_get_reg_length(struct rte_eth_dev *dev __rte_unused)
4938 {
4939 	int count = 0;
4940 	int g_ind = 0;
4941 	const struct reg_info *reg_group;
4942 
4943 	while ((reg_group = igb_regs[g_ind++]))
4944 		count += igb_reg_group_count(reg_group);
4945 
4946 	return count;
4947 }
4948 
4949 static int
4950 igbvf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
4951 {
4952 	int count = 0;
4953 	int g_ind = 0;
4954 	const struct reg_info *reg_group;
4955 
4956 	while ((reg_group = igbvf_regs[g_ind++]))
4957 		count += igb_reg_group_count(reg_group);
4958 
4959 	return count;
4960 }
4961 
4962 static int
4963 eth_igb_get_regs(struct rte_eth_dev *dev,
4964 	struct rte_dev_reg_info *regs)
4965 {
4966 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4967 	uint32_t *data = regs->data;
4968 	int g_ind = 0;
4969 	int count = 0;
4970 	const struct reg_info *reg_group;
4971 
4972 	if (data == NULL) {
4973 		regs->length = eth_igb_get_reg_length(dev);
4974 		regs->width = sizeof(uint32_t);
4975 		return 0;
4976 	}
4977 
4978 	/* Support only full register dump */
4979 	if ((regs->length == 0) ||
4980 	    (regs->length == (uint32_t)eth_igb_get_reg_length(dev))) {
4981 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
4982 			hw->device_id;
4983 		while ((reg_group = igb_regs[g_ind++]))
4984 			count += igb_read_regs_group(dev, &data[count],
4985 							reg_group);
4986 		return 0;
4987 	}
4988 
4989 	return -ENOTSUP;
4990 }
4991 
4992 static int
4993 igbvf_get_regs(struct rte_eth_dev *dev,
4994 	struct rte_dev_reg_info *regs)
4995 {
4996 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4997 	uint32_t *data = regs->data;
4998 	int g_ind = 0;
4999 	int count = 0;
5000 	const struct reg_info *reg_group;
5001 
5002 	if (data == NULL) {
5003 		regs->length = igbvf_get_reg_length(dev);
5004 		regs->width = sizeof(uint32_t);
5005 		return 0;
5006 	}
5007 
5008 	/* Support only full register dump */
5009 	if ((regs->length == 0) ||
5010 	    (regs->length == (uint32_t)igbvf_get_reg_length(dev))) {
5011 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5012 			hw->device_id;
5013 		while ((reg_group = igbvf_regs[g_ind++]))
5014 			count += igb_read_regs_group(dev, &data[count],
5015 							reg_group);
5016 		return 0;
5017 	}
5018 
5019 	return -ENOTSUP;
5020 }
5021 
5022 static int
5023 eth_igb_get_eeprom_length(struct rte_eth_dev *dev)
5024 {
5025 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5026 
5027 	/* Return unit is byte count */
5028 	return hw->nvm.word_size * 2;
5029 }
5030 
5031 static int
5032 eth_igb_get_eeprom(struct rte_eth_dev *dev,
5033 	struct rte_dev_eeprom_info *in_eeprom)
5034 {
5035 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5036 	struct e1000_nvm_info *nvm = &hw->nvm;
5037 	uint16_t *data = in_eeprom->data;
5038 	int first, length;
5039 
5040 	first = in_eeprom->offset >> 1;
5041 	length = in_eeprom->length >> 1;
5042 	if ((first >= hw->nvm.word_size) ||
5043 	    ((first + length) >= hw->nvm.word_size))
5044 		return -EINVAL;
5045 
5046 	in_eeprom->magic = hw->vendor_id |
5047 		((uint32_t)hw->device_id << 16);
5048 
5049 	if ((nvm->ops.read) == NULL)
5050 		return -ENOTSUP;
5051 
5052 	return nvm->ops.read(hw, first, length, data);
5053 }
5054 
5055 static int
5056 eth_igb_set_eeprom(struct rte_eth_dev *dev,
5057 	struct rte_dev_eeprom_info *in_eeprom)
5058 {
5059 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5060 	struct e1000_nvm_info *nvm = &hw->nvm;
5061 	uint16_t *data = in_eeprom->data;
5062 	int first, length;
5063 
5064 	first = in_eeprom->offset >> 1;
5065 	length = in_eeprom->length >> 1;
5066 	if ((first >= hw->nvm.word_size) ||
5067 	    ((first + length) >= hw->nvm.word_size))
5068 		return -EINVAL;
5069 
5070 	in_eeprom->magic = (uint32_t)hw->vendor_id |
5071 		((uint32_t)hw->device_id << 16);
5072 
5073 	if ((nvm->ops.write) == NULL)
5074 		return -ENOTSUP;
5075 	return nvm->ops.write(hw,  first, length, data);
5076 }
5077 
5078 static int
5079 eth_igb_get_module_info(struct rte_eth_dev *dev,
5080 			struct rte_eth_dev_module_info *modinfo)
5081 {
5082 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5083 
5084 	uint32_t status = 0;
5085 	uint16_t sff8472_rev, addr_mode;
5086 	bool page_swap = false;
5087 
5088 	if (hw->phy.media_type == e1000_media_type_copper ||
5089 	    hw->phy.media_type == e1000_media_type_unknown)
5090 		return -EOPNOTSUPP;
5091 
5092 	/* Check whether we support SFF-8472 or not */
5093 	status = e1000_read_phy_reg_i2c(hw, IGB_SFF_8472_COMP, &sff8472_rev);
5094 	if (status)
5095 		return -EIO;
5096 
5097 	/* addressing mode is not supported */
5098 	status = e1000_read_phy_reg_i2c(hw, IGB_SFF_8472_SWAP, &addr_mode);
5099 	if (status)
5100 		return -EIO;
5101 
5102 	/* addressing mode is not supported */
5103 	if ((addr_mode & 0xFF) & IGB_SFF_ADDRESSING_MODE) {
5104 		PMD_DRV_LOG(ERR,
5105 			    "Address change required to access page 0xA2, "
5106 			    "but not supported. Please report the module "
5107 			    "type to the driver maintainers.\n");
5108 		page_swap = true;
5109 	}
5110 
5111 	if ((sff8472_rev & 0xFF) == IGB_SFF_8472_UNSUP || page_swap) {
5112 		/* We have an SFP, but it does not support SFF-8472 */
5113 		modinfo->type = RTE_ETH_MODULE_SFF_8079;
5114 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
5115 	} else {
5116 		/* We have an SFP which supports a revision of SFF-8472 */
5117 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
5118 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
5119 	}
5120 
5121 	return 0;
5122 }
5123 
5124 static int
5125 eth_igb_get_module_eeprom(struct rte_eth_dev *dev,
5126 			  struct rte_dev_eeprom_info *info)
5127 {
5128 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5129 
5130 	uint32_t status = 0;
5131 	uint16_t dataword[RTE_ETH_MODULE_SFF_8472_LEN / 2 + 1];
5132 	u16 first_word, last_word;
5133 	int i = 0;
5134 
5135 	if (info->length == 0)
5136 		return -EINVAL;
5137 
5138 	first_word = info->offset >> 1;
5139 	last_word = (info->offset + info->length - 1) >> 1;
5140 
5141 	/* Read EEPROM block, SFF-8079/SFF-8472, word at a time */
5142 	for (i = 0; i < last_word - first_word + 1; i++) {
5143 		status = e1000_read_phy_reg_i2c(hw, (first_word + i) * 2,
5144 						&dataword[i]);
5145 		if (status) {
5146 			/* Error occurred while reading module */
5147 			return -EIO;
5148 		}
5149 
5150 		dataword[i] = rte_be_to_cpu_16(dataword[i]);
5151 	}
5152 
5153 	memcpy(info->data, (u8 *)dataword + (info->offset & 1), info->length);
5154 
5155 	return 0;
5156 }
5157 
5158 static int
5159 eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5160 {
5161 	struct e1000_hw *hw =
5162 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5163 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5164 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5165 	uint32_t vec = E1000_MISC_VEC_ID;
5166 
5167 	if (rte_intr_allow_others(intr_handle))
5168 		vec = E1000_RX_VEC_START;
5169 
5170 	uint32_t mask = 1 << (queue_id + vec);
5171 
5172 	E1000_WRITE_REG(hw, E1000_EIMC, mask);
5173 	E1000_WRITE_FLUSH(hw);
5174 
5175 	return 0;
5176 }
5177 
5178 static int
5179 eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5180 {
5181 	struct e1000_hw *hw =
5182 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5183 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5184 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5185 	uint32_t vec = E1000_MISC_VEC_ID;
5186 
5187 	if (rte_intr_allow_others(intr_handle))
5188 		vec = E1000_RX_VEC_START;
5189 
5190 	uint32_t mask = 1 << (queue_id + vec);
5191 	uint32_t regval;
5192 
5193 	regval = E1000_READ_REG(hw, E1000_EIMS);
5194 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
5195 	E1000_WRITE_FLUSH(hw);
5196 
5197 	rte_intr_ack(intr_handle);
5198 
5199 	return 0;
5200 }
5201 
5202 static void
5203 eth_igb_write_ivar(struct e1000_hw *hw, uint8_t  msix_vector,
5204 		   uint8_t index, uint8_t offset)
5205 {
5206 	uint32_t val = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
5207 
5208 	/* clear bits */
5209 	val &= ~((uint32_t)0xFF << offset);
5210 
5211 	/* write vector and valid bit */
5212 	val |= (msix_vector | E1000_IVAR_VALID) << offset;
5213 
5214 	E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, val);
5215 }
5216 
5217 static void
5218 eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
5219 			   uint8_t queue, uint8_t msix_vector)
5220 {
5221 	uint32_t tmp = 0;
5222 
5223 	if (hw->mac.type == e1000_82575) {
5224 		if (direction == 0)
5225 			tmp = E1000_EICR_RX_QUEUE0 << queue;
5226 		else if (direction == 1)
5227 			tmp = E1000_EICR_TX_QUEUE0 << queue;
5228 		E1000_WRITE_REG(hw, E1000_MSIXBM(msix_vector), tmp);
5229 	} else if (hw->mac.type == e1000_82576) {
5230 		if ((direction == 0) || (direction == 1))
5231 			eth_igb_write_ivar(hw, msix_vector, queue & 0x7,
5232 					   ((queue & 0x8) << 1) +
5233 					   8 * direction);
5234 	} else if ((hw->mac.type == e1000_82580) ||
5235 			(hw->mac.type == e1000_i350) ||
5236 			(hw->mac.type == e1000_i354) ||
5237 			(hw->mac.type == e1000_i210) ||
5238 			(hw->mac.type == e1000_i211)) {
5239 		if ((direction == 0) || (direction == 1))
5240 			eth_igb_write_ivar(hw, msix_vector,
5241 					   queue >> 1,
5242 					   ((queue & 0x1) << 4) +
5243 					   8 * direction);
5244 	}
5245 }
5246 
5247 /* Sets up the hardware to generate MSI-X interrupts properly
5248  * @hw
5249  *  board private structure
5250  */
5251 static void
5252 eth_igb_configure_msix_intr(struct rte_eth_dev *dev)
5253 {
5254 	int queue_id;
5255 	uint32_t tmpval, regval, intr_mask;
5256 	struct e1000_hw *hw =
5257 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5258 	uint32_t vec = E1000_MISC_VEC_ID;
5259 	uint32_t base = E1000_MISC_VEC_ID;
5260 	uint32_t misc_shift = 0;
5261 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5262 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5263 
5264 	/* won't configure msix register if no mapping is done
5265 	 * between intr vector and event fd
5266 	 */
5267 	if (!rte_intr_dp_is_en(intr_handle))
5268 		return;
5269 
5270 	if (rte_intr_allow_others(intr_handle)) {
5271 		vec = base = E1000_RX_VEC_START;
5272 		misc_shift = 1;
5273 	}
5274 
5275 	/* set interrupt vector for other causes */
5276 	if (hw->mac.type == e1000_82575) {
5277 		tmpval = E1000_READ_REG(hw, E1000_CTRL_EXT);
5278 		/* enable MSI-X PBA support */
5279 		tmpval |= E1000_CTRL_EXT_PBA_CLR;
5280 
5281 		/* Auto-Mask interrupts upon ICR read */
5282 		tmpval |= E1000_CTRL_EXT_EIAME;
5283 		tmpval |= E1000_CTRL_EXT_IRCA;
5284 
5285 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmpval);
5286 
5287 		/* enable msix_other interrupt */
5288 		E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0), 0, E1000_EIMS_OTHER);
5289 		regval = E1000_READ_REG(hw, E1000_EIAC);
5290 		E1000_WRITE_REG(hw, E1000_EIAC, regval | E1000_EIMS_OTHER);
5291 		regval = E1000_READ_REG(hw, E1000_EIAM);
5292 		E1000_WRITE_REG(hw, E1000_EIMS, regval | E1000_EIMS_OTHER);
5293 	} else if ((hw->mac.type == e1000_82576) ||
5294 			(hw->mac.type == e1000_82580) ||
5295 			(hw->mac.type == e1000_i350) ||
5296 			(hw->mac.type == e1000_i354) ||
5297 			(hw->mac.type == e1000_i210) ||
5298 			(hw->mac.type == e1000_i211)) {
5299 		/* turn on MSI-X capability first */
5300 		E1000_WRITE_REG(hw, E1000_GPIE, E1000_GPIE_MSIX_MODE |
5301 					E1000_GPIE_PBA | E1000_GPIE_EIAME |
5302 					E1000_GPIE_NSICR);
5303 		intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5304 			misc_shift;
5305 
5306 		if (dev->data->dev_conf.intr_conf.lsc != 0)
5307 			intr_mask |= (1 << IGB_MSIX_OTHER_INTR_VEC);
5308 
5309 		regval = E1000_READ_REG(hw, E1000_EIAC);
5310 		E1000_WRITE_REG(hw, E1000_EIAC, regval | intr_mask);
5311 
5312 		/* enable msix_other interrupt */
5313 		regval = E1000_READ_REG(hw, E1000_EIMS);
5314 		E1000_WRITE_REG(hw, E1000_EIMS, regval | intr_mask);
5315 		tmpval = (IGB_MSIX_OTHER_INTR_VEC | E1000_IVAR_VALID) << 8;
5316 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, tmpval);
5317 	}
5318 
5319 	/* use EIAM to auto-mask when MSI-X interrupt
5320 	 * is asserted, this saves a register write for every interrupt
5321 	 */
5322 	intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5323 		misc_shift;
5324 
5325 	if (dev->data->dev_conf.intr_conf.lsc != 0)
5326 		intr_mask |= (1 << IGB_MSIX_OTHER_INTR_VEC);
5327 
5328 	regval = E1000_READ_REG(hw, E1000_EIAM);
5329 	E1000_WRITE_REG(hw, E1000_EIAM, regval | intr_mask);
5330 
5331 	for (queue_id = 0; queue_id < dev->data->nb_rx_queues; queue_id++) {
5332 		eth_igb_assign_msix_vector(hw, 0, queue_id, vec);
5333 		intr_handle->intr_vec[queue_id] = vec;
5334 		if (vec < base + intr_handle->nb_efd - 1)
5335 			vec++;
5336 	}
5337 
5338 	E1000_WRITE_FLUSH(hw);
5339 }
5340 
5341 /* restore n-tuple filter */
5342 static inline void
5343 igb_ntuple_filter_restore(struct rte_eth_dev *dev)
5344 {
5345 	struct e1000_filter_info *filter_info =
5346 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5347 	struct e1000_5tuple_filter *p_5tuple;
5348 	struct e1000_2tuple_filter *p_2tuple;
5349 
5350 	TAILQ_FOREACH(p_5tuple, &filter_info->fivetuple_list, entries) {
5351 		igb_inject_5tuple_filter_82576(dev, p_5tuple);
5352 	}
5353 
5354 	TAILQ_FOREACH(p_2tuple, &filter_info->twotuple_list, entries) {
5355 		igb_inject_2uple_filter(dev, p_2tuple);
5356 	}
5357 }
5358 
5359 /* restore SYN filter */
5360 static inline void
5361 igb_syn_filter_restore(struct rte_eth_dev *dev)
5362 {
5363 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5364 	struct e1000_filter_info *filter_info =
5365 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5366 	uint32_t synqf;
5367 
5368 	synqf = filter_info->syn_info;
5369 
5370 	if (synqf & E1000_SYN_FILTER_ENABLE) {
5371 		E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
5372 		E1000_WRITE_FLUSH(hw);
5373 	}
5374 }
5375 
5376 /* restore ethernet type filter */
5377 static inline void
5378 igb_ethertype_filter_restore(struct rte_eth_dev *dev)
5379 {
5380 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5381 	struct e1000_filter_info *filter_info =
5382 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5383 	int i;
5384 
5385 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
5386 		if (filter_info->ethertype_mask & (1 << i)) {
5387 			E1000_WRITE_REG(hw, E1000_ETQF(i),
5388 				filter_info->ethertype_filters[i].etqf);
5389 			E1000_WRITE_FLUSH(hw);
5390 		}
5391 	}
5392 }
5393 
5394 /* restore flex byte filter */
5395 static inline void
5396 igb_flex_filter_restore(struct rte_eth_dev *dev)
5397 {
5398 	struct e1000_filter_info *filter_info =
5399 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5400 	struct e1000_flex_filter *flex_filter;
5401 
5402 	TAILQ_FOREACH(flex_filter, &filter_info->flex_list, entries) {
5403 		igb_inject_flex_filter(dev, flex_filter);
5404 	}
5405 }
5406 
5407 /* restore rss filter */
5408 static inline void
5409 igb_rss_filter_restore(struct rte_eth_dev *dev)
5410 {
5411 	struct e1000_filter_info *filter_info =
5412 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5413 
5414 	if (filter_info->rss_info.conf.queue_num)
5415 		igb_config_rss_filter(dev, &filter_info->rss_info, TRUE);
5416 }
5417 
5418 /* restore all types filter */
5419 static int
5420 igb_filter_restore(struct rte_eth_dev *dev)
5421 {
5422 	igb_ntuple_filter_restore(dev);
5423 	igb_ethertype_filter_restore(dev);
5424 	igb_syn_filter_restore(dev);
5425 	igb_flex_filter_restore(dev);
5426 	igb_rss_filter_restore(dev);
5427 
5428 	return 0;
5429 }
5430 
5431 RTE_PMD_REGISTER_PCI(net_e1000_igb, rte_igb_pmd);
5432 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb, pci_id_igb_map);
5433 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb, "* igb_uio | uio_pci_generic | vfio-pci");
5434 RTE_PMD_REGISTER_PCI(net_e1000_igb_vf, rte_igbvf_pmd);
5435 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb_vf, pci_id_igbvf_map);
5436 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb_vf, "* igb_uio | vfio-pci");
5437 
5438 /* see e1000_logs.c */
5439 RTE_INIT(e1000_init_log)
5440 {
5441 	e1000_igb_init_log();
5442 }
5443