xref: /f-stack/dpdk/drivers/net/e1000/igb_ethdev.c (revision 2bfe3f2e)
1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
5  *   All rights reserved.
6  *
7  *   Redistribution and use in source and binary forms, with or without
8  *   modification, are permitted provided that the following conditions
9  *   are met:
10  *
11  *     * Redistributions of source code must retain the above copyright
12  *       notice, this list of conditions and the following disclaimer.
13  *     * Redistributions in binary form must reproduce the above copyright
14  *       notice, this list of conditions and the following disclaimer in
15  *       the documentation and/or other materials provided with the
16  *       distribution.
17  *     * Neither the name of Intel Corporation nor the names of its
18  *       contributors may be used to endorse or promote products derived
19  *       from this software without specific prior written permission.
20  *
21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32  */
33 
34 #include <sys/queue.h>
35 #include <stdio.h>
36 #include <errno.h>
37 #include <stdint.h>
38 #include <stdarg.h>
39 
40 #include <rte_common.h>
41 #include <rte_interrupts.h>
42 #include <rte_byteorder.h>
43 #include <rte_log.h>
44 #include <rte_debug.h>
45 #include <rte_pci.h>
46 #include <rte_bus_pci.h>
47 #include <rte_ether.h>
48 #include <rte_ethdev.h>
49 #include <rte_ethdev_pci.h>
50 #include <rte_memory.h>
51 #include <rte_eal.h>
52 #include <rte_atomic.h>
53 #include <rte_malloc.h>
54 #include <rte_dev.h>
55 
56 #include "e1000_logs.h"
57 #include "base/e1000_api.h"
58 #include "e1000_ethdev.h"
59 #include "igb_regs.h"
60 
61 /*
62  * Default values for port configuration
63  */
64 #define IGB_DEFAULT_RX_FREE_THRESH  32
65 
66 #define IGB_DEFAULT_RX_PTHRESH      ((hw->mac.type == e1000_i354) ? 12 : 8)
67 #define IGB_DEFAULT_RX_HTHRESH      8
68 #define IGB_DEFAULT_RX_WTHRESH      ((hw->mac.type == e1000_82576) ? 1 : 4)
69 
70 #define IGB_DEFAULT_TX_PTHRESH      ((hw->mac.type == e1000_i354) ? 20 : 8)
71 #define IGB_DEFAULT_TX_HTHRESH      1
72 #define IGB_DEFAULT_TX_WTHRESH      ((hw->mac.type == e1000_82576) ? 1 : 16)
73 
74 #define IGB_HKEY_MAX_INDEX 10
75 
76 /* Bit shift and mask */
77 #define IGB_4_BIT_WIDTH  (CHAR_BIT / 2)
78 #define IGB_4_BIT_MASK   RTE_LEN2MASK(IGB_4_BIT_WIDTH, uint8_t)
79 #define IGB_8_BIT_WIDTH  CHAR_BIT
80 #define IGB_8_BIT_MASK   UINT8_MAX
81 
82 /* Additional timesync values. */
83 #define E1000_CYCLECOUNTER_MASK      0xffffffffffffffffULL
84 #define E1000_ETQF_FILTER_1588       3
85 #define IGB_82576_TSYNC_SHIFT        16
86 #define E1000_INCPERIOD_82576        (1 << E1000_TIMINCA_16NS_SHIFT)
87 #define E1000_INCVALUE_82576         (16 << IGB_82576_TSYNC_SHIFT)
88 #define E1000_TSAUXC_DISABLE_SYSTIME 0x80000000
89 
90 #define E1000_VTIVAR_MISC                0x01740
91 #define E1000_VTIVAR_MISC_MASK           0xFF
92 #define E1000_VTIVAR_VALID               0x80
93 #define E1000_VTIVAR_MISC_MAILBOX        0
94 #define E1000_VTIVAR_MISC_INTR_MASK      0x3
95 
96 /* External VLAN Enable bit mask */
97 #define E1000_CTRL_EXT_EXT_VLAN      (1 << 26)
98 
99 /* External VLAN Ether Type bit mask and shift */
100 #define E1000_VET_VET_EXT            0xFFFF0000
101 #define E1000_VET_VET_EXT_SHIFT      16
102 
103 static int  eth_igb_configure(struct rte_eth_dev *dev);
104 static int  eth_igb_start(struct rte_eth_dev *dev);
105 static void eth_igb_stop(struct rte_eth_dev *dev);
106 static int  eth_igb_dev_set_link_up(struct rte_eth_dev *dev);
107 static int  eth_igb_dev_set_link_down(struct rte_eth_dev *dev);
108 static void eth_igb_close(struct rte_eth_dev *dev);
109 static void eth_igb_promiscuous_enable(struct rte_eth_dev *dev);
110 static void eth_igb_promiscuous_disable(struct rte_eth_dev *dev);
111 static void eth_igb_allmulticast_enable(struct rte_eth_dev *dev);
112 static void eth_igb_allmulticast_disable(struct rte_eth_dev *dev);
113 static int  eth_igb_link_update(struct rte_eth_dev *dev,
114 				int wait_to_complete);
115 static int eth_igb_stats_get(struct rte_eth_dev *dev,
116 				struct rte_eth_stats *rte_stats);
117 static int eth_igb_xstats_get(struct rte_eth_dev *dev,
118 			      struct rte_eth_xstat *xstats, unsigned n);
119 static int eth_igb_xstats_get_by_id(struct rte_eth_dev *dev,
120 		const uint64_t *ids,
121 		uint64_t *values, unsigned int n);
122 static int eth_igb_xstats_get_names(struct rte_eth_dev *dev,
123 				    struct rte_eth_xstat_name *xstats_names,
124 				    unsigned int size);
125 static int eth_igb_xstats_get_names_by_id(struct rte_eth_dev *dev,
126 		struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
127 		unsigned int limit);
128 static void eth_igb_stats_reset(struct rte_eth_dev *dev);
129 static void eth_igb_xstats_reset(struct rte_eth_dev *dev);
130 static int eth_igb_fw_version_get(struct rte_eth_dev *dev,
131 				   char *fw_version, size_t fw_size);
132 static void eth_igb_infos_get(struct rte_eth_dev *dev,
133 			      struct rte_eth_dev_info *dev_info);
134 static const uint32_t *eth_igb_supported_ptypes_get(struct rte_eth_dev *dev);
135 static void eth_igbvf_infos_get(struct rte_eth_dev *dev,
136 				struct rte_eth_dev_info *dev_info);
137 static int  eth_igb_flow_ctrl_get(struct rte_eth_dev *dev,
138 				struct rte_eth_fc_conf *fc_conf);
139 static int  eth_igb_flow_ctrl_set(struct rte_eth_dev *dev,
140 				struct rte_eth_fc_conf *fc_conf);
141 static int eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
142 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev);
143 static int eth_igb_interrupt_get_status(struct rte_eth_dev *dev);
144 static int eth_igb_interrupt_action(struct rte_eth_dev *dev,
145 				    struct rte_intr_handle *handle);
146 static void eth_igb_interrupt_handler(void *param);
147 static int  igb_hardware_init(struct e1000_hw *hw);
148 static void igb_hw_control_acquire(struct e1000_hw *hw);
149 static void igb_hw_control_release(struct e1000_hw *hw);
150 static void igb_init_manageability(struct e1000_hw *hw);
151 static void igb_release_manageability(struct e1000_hw *hw);
152 
153 static int  eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
154 
155 static int eth_igb_vlan_filter_set(struct rte_eth_dev *dev,
156 		uint16_t vlan_id, int on);
157 static int eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
158 				 enum rte_vlan_type vlan_type,
159 				 uint16_t tpid_id);
160 static int eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask);
161 
162 static void igb_vlan_hw_filter_enable(struct rte_eth_dev *dev);
163 static void igb_vlan_hw_filter_disable(struct rte_eth_dev *dev);
164 static void igb_vlan_hw_strip_enable(struct rte_eth_dev *dev);
165 static void igb_vlan_hw_strip_disable(struct rte_eth_dev *dev);
166 static void igb_vlan_hw_extend_enable(struct rte_eth_dev *dev);
167 static void igb_vlan_hw_extend_disable(struct rte_eth_dev *dev);
168 
169 static int eth_igb_led_on(struct rte_eth_dev *dev);
170 static int eth_igb_led_off(struct rte_eth_dev *dev);
171 
172 static void igb_intr_disable(struct e1000_hw *hw);
173 static int  igb_get_rx_buffer_size(struct e1000_hw *hw);
174 static int eth_igb_rar_set(struct rte_eth_dev *dev,
175 			   struct ether_addr *mac_addr,
176 			   uint32_t index, uint32_t pool);
177 static void eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index);
178 static void eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
179 		struct ether_addr *addr);
180 
181 static void igbvf_intr_disable(struct e1000_hw *hw);
182 static int igbvf_dev_configure(struct rte_eth_dev *dev);
183 static int igbvf_dev_start(struct rte_eth_dev *dev);
184 static void igbvf_dev_stop(struct rte_eth_dev *dev);
185 static void igbvf_dev_close(struct rte_eth_dev *dev);
186 static void igbvf_promiscuous_enable(struct rte_eth_dev *dev);
187 static void igbvf_promiscuous_disable(struct rte_eth_dev *dev);
188 static void igbvf_allmulticast_enable(struct rte_eth_dev *dev);
189 static void igbvf_allmulticast_disable(struct rte_eth_dev *dev);
190 static int eth_igbvf_link_update(struct e1000_hw *hw);
191 static int eth_igbvf_stats_get(struct rte_eth_dev *dev,
192 				struct rte_eth_stats *rte_stats);
193 static int eth_igbvf_xstats_get(struct rte_eth_dev *dev,
194 				struct rte_eth_xstat *xstats, unsigned n);
195 static int eth_igbvf_xstats_get_names(struct rte_eth_dev *dev,
196 				      struct rte_eth_xstat_name *xstats_names,
197 				      unsigned limit);
198 static void eth_igbvf_stats_reset(struct rte_eth_dev *dev);
199 static int igbvf_vlan_filter_set(struct rte_eth_dev *dev,
200 		uint16_t vlan_id, int on);
201 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on);
202 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on);
203 static void igbvf_default_mac_addr_set(struct rte_eth_dev *dev,
204 		struct ether_addr *addr);
205 static int igbvf_get_reg_length(struct rte_eth_dev *dev);
206 static int igbvf_get_regs(struct rte_eth_dev *dev,
207 		struct rte_dev_reg_info *regs);
208 
209 static int eth_igb_rss_reta_update(struct rte_eth_dev *dev,
210 				   struct rte_eth_rss_reta_entry64 *reta_conf,
211 				   uint16_t reta_size);
212 static int eth_igb_rss_reta_query(struct rte_eth_dev *dev,
213 				  struct rte_eth_rss_reta_entry64 *reta_conf,
214 				  uint16_t reta_size);
215 
216 static int eth_igb_syn_filter_get(struct rte_eth_dev *dev,
217 			struct rte_eth_syn_filter *filter);
218 static int eth_igb_syn_filter_handle(struct rte_eth_dev *dev,
219 			enum rte_filter_op filter_op,
220 			void *arg);
221 static int igb_add_2tuple_filter(struct rte_eth_dev *dev,
222 			struct rte_eth_ntuple_filter *ntuple_filter);
223 static int igb_remove_2tuple_filter(struct rte_eth_dev *dev,
224 			struct rte_eth_ntuple_filter *ntuple_filter);
225 static int eth_igb_get_flex_filter(struct rte_eth_dev *dev,
226 			struct rte_eth_flex_filter *filter);
227 static int eth_igb_flex_filter_handle(struct rte_eth_dev *dev,
228 			enum rte_filter_op filter_op,
229 			void *arg);
230 static int igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
231 			struct rte_eth_ntuple_filter *ntuple_filter);
232 static int igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
233 			struct rte_eth_ntuple_filter *ntuple_filter);
234 static int igb_get_ntuple_filter(struct rte_eth_dev *dev,
235 			struct rte_eth_ntuple_filter *filter);
236 static int igb_ntuple_filter_handle(struct rte_eth_dev *dev,
237 				enum rte_filter_op filter_op,
238 				void *arg);
239 static int igb_ethertype_filter_handle(struct rte_eth_dev *dev,
240 				enum rte_filter_op filter_op,
241 				void *arg);
242 static int igb_get_ethertype_filter(struct rte_eth_dev *dev,
243 			struct rte_eth_ethertype_filter *filter);
244 static int eth_igb_filter_ctrl(struct rte_eth_dev *dev,
245 		     enum rte_filter_type filter_type,
246 		     enum rte_filter_op filter_op,
247 		     void *arg);
248 static int eth_igb_get_reg_length(struct rte_eth_dev *dev);
249 static int eth_igb_get_regs(struct rte_eth_dev *dev,
250 		struct rte_dev_reg_info *regs);
251 static int eth_igb_get_eeprom_length(struct rte_eth_dev *dev);
252 static int eth_igb_get_eeprom(struct rte_eth_dev *dev,
253 		struct rte_dev_eeprom_info *eeprom);
254 static int eth_igb_set_eeprom(struct rte_eth_dev *dev,
255 		struct rte_dev_eeprom_info *eeprom);
256 static int eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
257 				    struct ether_addr *mc_addr_set,
258 				    uint32_t nb_mc_addr);
259 static int igb_timesync_enable(struct rte_eth_dev *dev);
260 static int igb_timesync_disable(struct rte_eth_dev *dev);
261 static int igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
262 					  struct timespec *timestamp,
263 					  uint32_t flags);
264 static int igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
265 					  struct timespec *timestamp);
266 static int igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
267 static int igb_timesync_read_time(struct rte_eth_dev *dev,
268 				  struct timespec *timestamp);
269 static int igb_timesync_write_time(struct rte_eth_dev *dev,
270 				   const struct timespec *timestamp);
271 static int eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev,
272 					uint16_t queue_id);
273 static int eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev,
274 					 uint16_t queue_id);
275 static void eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
276 				       uint8_t queue, uint8_t msix_vector);
277 static void eth_igb_write_ivar(struct e1000_hw *hw, uint8_t msix_vector,
278 			       uint8_t index, uint8_t offset);
279 static void eth_igb_configure_msix_intr(struct rte_eth_dev *dev);
280 static void eth_igbvf_interrupt_handler(void *param);
281 static void igbvf_mbx_process(struct rte_eth_dev *dev);
282 static int igb_filter_restore(struct rte_eth_dev *dev);
283 
284 /*
285  * Define VF Stats MACRO for Non "cleared on read" register
286  */
287 #define UPDATE_VF_STAT(reg, last, cur)            \
288 {                                                 \
289 	u32 latest = E1000_READ_REG(hw, reg);     \
290 	cur += (latest - last) & UINT_MAX;        \
291 	last = latest;                            \
292 }
293 
294 #define IGB_FC_PAUSE_TIME 0x0680
295 #define IGB_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
296 #define IGB_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
297 
298 #define IGBVF_PMD_NAME "rte_igbvf_pmd"     /* PMD name */
299 
300 static enum e1000_fc_mode igb_fc_setting = e1000_fc_full;
301 
302 /*
303  * The set of PCI devices this driver supports
304  */
305 static const struct rte_pci_id pci_id_igb_map[] = {
306 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576) },
307 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_FIBER) },
308 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_SERDES) },
309 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_QUAD_COPPER) },
310 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_QUAD_COPPER_ET2) },
311 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_NS) },
312 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_NS_SERDES) },
313 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_SERDES_QUAD) },
314 
315 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575EB_COPPER) },
316 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575EB_FIBER_SERDES) },
317 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575GB_QUAD_COPPER) },
318 
319 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_COPPER) },
320 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_FIBER) },
321 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_SERDES) },
322 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_SGMII) },
323 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_COPPER_DUAL) },
324 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_QUAD_FIBER) },
325 
326 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_COPPER) },
327 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_FIBER) },
328 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_SERDES) },
329 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_SGMII) },
330 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_DA4) },
331 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER) },
332 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_OEM1) },
333 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_IT) },
334 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_FIBER) },
335 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SERDES) },
336 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SGMII) },
337 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_FLASHLESS) },
338 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SERDES_FLASHLESS) },
339 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I211_COPPER) },
340 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
341 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_SGMII) },
342 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
343 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SGMII) },
344 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SERDES) },
345 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_BACKPLANE) },
346 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SFP) },
347 	{ .vendor_id = 0, /* sentinel */ },
348 };
349 
350 /*
351  * The set of PCI devices this driver supports (for 82576&I350 VF)
352  */
353 static const struct rte_pci_id pci_id_igbvf_map[] = {
354 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_VF) },
355 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_VF_HV) },
356 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_VF) },
357 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_VF_HV) },
358 	{ .vendor_id = 0, /* sentinel */ },
359 };
360 
361 static const struct rte_eth_desc_lim rx_desc_lim = {
362 	.nb_max = E1000_MAX_RING_DESC,
363 	.nb_min = E1000_MIN_RING_DESC,
364 	.nb_align = IGB_RXD_ALIGN,
365 };
366 
367 static const struct rte_eth_desc_lim tx_desc_lim = {
368 	.nb_max = E1000_MAX_RING_DESC,
369 	.nb_min = E1000_MIN_RING_DESC,
370 	.nb_align = IGB_RXD_ALIGN,
371 	.nb_seg_max = IGB_TX_MAX_SEG,
372 	.nb_mtu_seg_max = IGB_TX_MAX_MTU_SEG,
373 };
374 
375 static const struct eth_dev_ops eth_igb_ops = {
376 	.dev_configure        = eth_igb_configure,
377 	.dev_start            = eth_igb_start,
378 	.dev_stop             = eth_igb_stop,
379 	.dev_set_link_up      = eth_igb_dev_set_link_up,
380 	.dev_set_link_down    = eth_igb_dev_set_link_down,
381 	.dev_close            = eth_igb_close,
382 	.promiscuous_enable   = eth_igb_promiscuous_enable,
383 	.promiscuous_disable  = eth_igb_promiscuous_disable,
384 	.allmulticast_enable  = eth_igb_allmulticast_enable,
385 	.allmulticast_disable = eth_igb_allmulticast_disable,
386 	.link_update          = eth_igb_link_update,
387 	.stats_get            = eth_igb_stats_get,
388 	.xstats_get           = eth_igb_xstats_get,
389 	.xstats_get_by_id     = eth_igb_xstats_get_by_id,
390 	.xstats_get_names_by_id = eth_igb_xstats_get_names_by_id,
391 	.xstats_get_names     = eth_igb_xstats_get_names,
392 	.stats_reset          = eth_igb_stats_reset,
393 	.xstats_reset         = eth_igb_xstats_reset,
394 	.fw_version_get       = eth_igb_fw_version_get,
395 	.dev_infos_get        = eth_igb_infos_get,
396 	.dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
397 	.mtu_set              = eth_igb_mtu_set,
398 	.vlan_filter_set      = eth_igb_vlan_filter_set,
399 	.vlan_tpid_set        = eth_igb_vlan_tpid_set,
400 	.vlan_offload_set     = eth_igb_vlan_offload_set,
401 	.rx_queue_setup       = eth_igb_rx_queue_setup,
402 	.rx_queue_intr_enable = eth_igb_rx_queue_intr_enable,
403 	.rx_queue_intr_disable = eth_igb_rx_queue_intr_disable,
404 	.rx_queue_release     = eth_igb_rx_queue_release,
405 	.rx_queue_count       = eth_igb_rx_queue_count,
406 	.rx_descriptor_done   = eth_igb_rx_descriptor_done,
407 	.rx_descriptor_status = eth_igb_rx_descriptor_status,
408 	.tx_descriptor_status = eth_igb_tx_descriptor_status,
409 	.tx_queue_setup       = eth_igb_tx_queue_setup,
410 	.tx_queue_release     = eth_igb_tx_queue_release,
411 	.tx_done_cleanup      = eth_igb_tx_done_cleanup,
412 	.dev_led_on           = eth_igb_led_on,
413 	.dev_led_off          = eth_igb_led_off,
414 	.flow_ctrl_get        = eth_igb_flow_ctrl_get,
415 	.flow_ctrl_set        = eth_igb_flow_ctrl_set,
416 	.mac_addr_add         = eth_igb_rar_set,
417 	.mac_addr_remove      = eth_igb_rar_clear,
418 	.mac_addr_set         = eth_igb_default_mac_addr_set,
419 	.reta_update          = eth_igb_rss_reta_update,
420 	.reta_query           = eth_igb_rss_reta_query,
421 	.rss_hash_update      = eth_igb_rss_hash_update,
422 	.rss_hash_conf_get    = eth_igb_rss_hash_conf_get,
423 	.filter_ctrl          = eth_igb_filter_ctrl,
424 	.set_mc_addr_list     = eth_igb_set_mc_addr_list,
425 	.rxq_info_get         = igb_rxq_info_get,
426 	.txq_info_get         = igb_txq_info_get,
427 	.timesync_enable      = igb_timesync_enable,
428 	.timesync_disable     = igb_timesync_disable,
429 	.timesync_read_rx_timestamp = igb_timesync_read_rx_timestamp,
430 	.timesync_read_tx_timestamp = igb_timesync_read_tx_timestamp,
431 	.get_reg              = eth_igb_get_regs,
432 	.get_eeprom_length    = eth_igb_get_eeprom_length,
433 	.get_eeprom           = eth_igb_get_eeprom,
434 	.set_eeprom           = eth_igb_set_eeprom,
435 	.timesync_adjust_time = igb_timesync_adjust_time,
436 	.timesync_read_time   = igb_timesync_read_time,
437 	.timesync_write_time  = igb_timesync_write_time,
438 };
439 
440 /*
441  * dev_ops for virtual function, bare necessities for basic vf
442  * operation have been implemented
443  */
444 static const struct eth_dev_ops igbvf_eth_dev_ops = {
445 	.dev_configure        = igbvf_dev_configure,
446 	.dev_start            = igbvf_dev_start,
447 	.dev_stop             = igbvf_dev_stop,
448 	.dev_close            = igbvf_dev_close,
449 	.promiscuous_enable   = igbvf_promiscuous_enable,
450 	.promiscuous_disable  = igbvf_promiscuous_disable,
451 	.allmulticast_enable  = igbvf_allmulticast_enable,
452 	.allmulticast_disable = igbvf_allmulticast_disable,
453 	.link_update          = eth_igb_link_update,
454 	.stats_get            = eth_igbvf_stats_get,
455 	.xstats_get           = eth_igbvf_xstats_get,
456 	.xstats_get_names     = eth_igbvf_xstats_get_names,
457 	.stats_reset          = eth_igbvf_stats_reset,
458 	.xstats_reset         = eth_igbvf_stats_reset,
459 	.vlan_filter_set      = igbvf_vlan_filter_set,
460 	.dev_infos_get        = eth_igbvf_infos_get,
461 	.dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
462 	.rx_queue_setup       = eth_igb_rx_queue_setup,
463 	.rx_queue_release     = eth_igb_rx_queue_release,
464 	.tx_queue_setup       = eth_igb_tx_queue_setup,
465 	.tx_queue_release     = eth_igb_tx_queue_release,
466 	.set_mc_addr_list     = eth_igb_set_mc_addr_list,
467 	.rxq_info_get         = igb_rxq_info_get,
468 	.txq_info_get         = igb_txq_info_get,
469 	.mac_addr_set         = igbvf_default_mac_addr_set,
470 	.get_reg              = igbvf_get_regs,
471 };
472 
473 /* store statistics names and its offset in stats structure */
474 struct rte_igb_xstats_name_off {
475 	char name[RTE_ETH_XSTATS_NAME_SIZE];
476 	unsigned offset;
477 };
478 
479 static const struct rte_igb_xstats_name_off rte_igb_stats_strings[] = {
480 	{"rx_crc_errors", offsetof(struct e1000_hw_stats, crcerrs)},
481 	{"rx_align_errors", offsetof(struct e1000_hw_stats, algnerrc)},
482 	{"rx_symbol_errors", offsetof(struct e1000_hw_stats, symerrs)},
483 	{"rx_missed_packets", offsetof(struct e1000_hw_stats, mpc)},
484 	{"tx_single_collision_packets", offsetof(struct e1000_hw_stats, scc)},
485 	{"tx_multiple_collision_packets", offsetof(struct e1000_hw_stats, mcc)},
486 	{"tx_excessive_collision_packets", offsetof(struct e1000_hw_stats,
487 		ecol)},
488 	{"tx_late_collisions", offsetof(struct e1000_hw_stats, latecol)},
489 	{"tx_total_collisions", offsetof(struct e1000_hw_stats, colc)},
490 	{"tx_deferred_packets", offsetof(struct e1000_hw_stats, dc)},
491 	{"tx_no_carrier_sense_packets", offsetof(struct e1000_hw_stats, tncrs)},
492 	{"rx_carrier_ext_errors", offsetof(struct e1000_hw_stats, cexterr)},
493 	{"rx_length_errors", offsetof(struct e1000_hw_stats, rlec)},
494 	{"rx_xon_packets", offsetof(struct e1000_hw_stats, xonrxc)},
495 	{"tx_xon_packets", offsetof(struct e1000_hw_stats, xontxc)},
496 	{"rx_xoff_packets", offsetof(struct e1000_hw_stats, xoffrxc)},
497 	{"tx_xoff_packets", offsetof(struct e1000_hw_stats, xofftxc)},
498 	{"rx_flow_control_unsupported_packets", offsetof(struct e1000_hw_stats,
499 		fcruc)},
500 	{"rx_size_64_packets", offsetof(struct e1000_hw_stats, prc64)},
501 	{"rx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, prc127)},
502 	{"rx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, prc255)},
503 	{"rx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, prc511)},
504 	{"rx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
505 		prc1023)},
506 	{"rx_size_1024_to_max_packets", offsetof(struct e1000_hw_stats,
507 		prc1522)},
508 	{"rx_broadcast_packets", offsetof(struct e1000_hw_stats, bprc)},
509 	{"rx_multicast_packets", offsetof(struct e1000_hw_stats, mprc)},
510 	{"rx_undersize_errors", offsetof(struct e1000_hw_stats, ruc)},
511 	{"rx_fragment_errors", offsetof(struct e1000_hw_stats, rfc)},
512 	{"rx_oversize_errors", offsetof(struct e1000_hw_stats, roc)},
513 	{"rx_jabber_errors", offsetof(struct e1000_hw_stats, rjc)},
514 	{"rx_management_packets", offsetof(struct e1000_hw_stats, mgprc)},
515 	{"rx_management_dropped", offsetof(struct e1000_hw_stats, mgpdc)},
516 	{"tx_management_packets", offsetof(struct e1000_hw_stats, mgptc)},
517 	{"rx_total_packets", offsetof(struct e1000_hw_stats, tpr)},
518 	{"tx_total_packets", offsetof(struct e1000_hw_stats, tpt)},
519 	{"rx_total_bytes", offsetof(struct e1000_hw_stats, tor)},
520 	{"tx_total_bytes", offsetof(struct e1000_hw_stats, tot)},
521 	{"tx_size_64_packets", offsetof(struct e1000_hw_stats, ptc64)},
522 	{"tx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, ptc127)},
523 	{"tx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, ptc255)},
524 	{"tx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, ptc511)},
525 	{"tx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
526 		ptc1023)},
527 	{"tx_size_1023_to_max_packets", offsetof(struct e1000_hw_stats,
528 		ptc1522)},
529 	{"tx_multicast_packets", offsetof(struct e1000_hw_stats, mptc)},
530 	{"tx_broadcast_packets", offsetof(struct e1000_hw_stats, bptc)},
531 	{"tx_tso_packets", offsetof(struct e1000_hw_stats, tsctc)},
532 	{"tx_tso_errors", offsetof(struct e1000_hw_stats, tsctfc)},
533 	{"rx_sent_to_host_packets", offsetof(struct e1000_hw_stats, rpthc)},
534 	{"tx_sent_by_host_packets", offsetof(struct e1000_hw_stats, hgptc)},
535 	{"rx_code_violation_packets", offsetof(struct e1000_hw_stats, scvpc)},
536 
537 	{"interrupt_assert_count", offsetof(struct e1000_hw_stats, iac)},
538 };
539 
540 #define IGB_NB_XSTATS (sizeof(rte_igb_stats_strings) / \
541 		sizeof(rte_igb_stats_strings[0]))
542 
543 static const struct rte_igb_xstats_name_off rte_igbvf_stats_strings[] = {
544 	{"rx_multicast_packets", offsetof(struct e1000_vf_stats, mprc)},
545 	{"rx_good_loopback_packets", offsetof(struct e1000_vf_stats, gprlbc)},
546 	{"tx_good_loopback_packets", offsetof(struct e1000_vf_stats, gptlbc)},
547 	{"rx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gorlbc)},
548 	{"tx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gotlbc)},
549 };
550 
551 #define IGBVF_NB_XSTATS (sizeof(rte_igbvf_stats_strings) / \
552 		sizeof(rte_igbvf_stats_strings[0]))
553 
554 /**
555  * Atomically reads the link status information from global
556  * structure rte_eth_dev.
557  *
558  * @param dev
559  *   - Pointer to the structure rte_eth_dev to read from.
560  *   - Pointer to the buffer to be saved with the link status.
561  *
562  * @return
563  *   - On success, zero.
564  *   - On failure, negative value.
565  */
566 static inline int
567 rte_igb_dev_atomic_read_link_status(struct rte_eth_dev *dev,
568 				struct rte_eth_link *link)
569 {
570 	struct rte_eth_link *dst = link;
571 	struct rte_eth_link *src = &(dev->data->dev_link);
572 
573 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
574 					*(uint64_t *)src) == 0)
575 		return -1;
576 
577 	return 0;
578 }
579 
580 /**
581  * Atomically writes the link status information into global
582  * structure rte_eth_dev.
583  *
584  * @param dev
585  *   - Pointer to the structure rte_eth_dev to read from.
586  *   - Pointer to the buffer to be saved with the link status.
587  *
588  * @return
589  *   - On success, zero.
590  *   - On failure, negative value.
591  */
592 static inline int
593 rte_igb_dev_atomic_write_link_status(struct rte_eth_dev *dev,
594 				struct rte_eth_link *link)
595 {
596 	struct rte_eth_link *dst = &(dev->data->dev_link);
597 	struct rte_eth_link *src = link;
598 
599 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
600 					*(uint64_t *)src) == 0)
601 		return -1;
602 
603 	return 0;
604 }
605 
606 static inline void
607 igb_intr_enable(struct rte_eth_dev *dev)
608 {
609 	struct e1000_interrupt *intr =
610 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
611 	struct e1000_hw *hw =
612 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
613 
614 	E1000_WRITE_REG(hw, E1000_IMS, intr->mask);
615 	E1000_WRITE_FLUSH(hw);
616 }
617 
618 static void
619 igb_intr_disable(struct e1000_hw *hw)
620 {
621 	E1000_WRITE_REG(hw, E1000_IMC, ~0);
622 	E1000_WRITE_FLUSH(hw);
623 }
624 
625 static inline void
626 igbvf_intr_enable(struct rte_eth_dev *dev)
627 {
628 	struct e1000_hw *hw =
629 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
630 
631 	/* only for mailbox */
632 	E1000_WRITE_REG(hw, E1000_EIAM, 1 << E1000_VTIVAR_MISC_MAILBOX);
633 	E1000_WRITE_REG(hw, E1000_EIAC, 1 << E1000_VTIVAR_MISC_MAILBOX);
634 	E1000_WRITE_REG(hw, E1000_EIMS, 1 << E1000_VTIVAR_MISC_MAILBOX);
635 	E1000_WRITE_FLUSH(hw);
636 }
637 
638 /* only for mailbox now. If RX/TX needed, should extend this function.  */
639 static void
640 igbvf_set_ivar_map(struct e1000_hw *hw, uint8_t msix_vector)
641 {
642 	uint32_t tmp = 0;
643 
644 	/* mailbox */
645 	tmp |= (msix_vector & E1000_VTIVAR_MISC_INTR_MASK);
646 	tmp |= E1000_VTIVAR_VALID;
647 	E1000_WRITE_REG(hw, E1000_VTIVAR_MISC, tmp);
648 }
649 
650 static void
651 eth_igbvf_configure_msix_intr(struct rte_eth_dev *dev)
652 {
653 	struct e1000_hw *hw =
654 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
655 
656 	/* Configure VF other cause ivar */
657 	igbvf_set_ivar_map(hw, E1000_VTIVAR_MISC_MAILBOX);
658 }
659 
660 static inline int32_t
661 igb_pf_reset_hw(struct e1000_hw *hw)
662 {
663 	uint32_t ctrl_ext;
664 	int32_t status;
665 
666 	status = e1000_reset_hw(hw);
667 
668 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
669 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
670 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
671 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
672 	E1000_WRITE_FLUSH(hw);
673 
674 	return status;
675 }
676 
677 static void
678 igb_identify_hardware(struct rte_eth_dev *dev, struct rte_pci_device *pci_dev)
679 {
680 	struct e1000_hw *hw =
681 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
682 
683 
684 	hw->vendor_id = pci_dev->id.vendor_id;
685 	hw->device_id = pci_dev->id.device_id;
686 	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
687 	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
688 
689 	e1000_set_mac_type(hw);
690 
691 	/* need to check if it is a vf device below */
692 }
693 
694 static int
695 igb_reset_swfw_lock(struct e1000_hw *hw)
696 {
697 	int ret_val;
698 
699 	/*
700 	 * Do mac ops initialization manually here, since we will need
701 	 * some function pointers set by this call.
702 	 */
703 	ret_val = e1000_init_mac_params(hw);
704 	if (ret_val)
705 		return ret_val;
706 
707 	/*
708 	 * SMBI lock should not fail in this early stage. If this is the case,
709 	 * it is due to an improper exit of the application.
710 	 * So force the release of the faulty lock.
711 	 */
712 	if (e1000_get_hw_semaphore_generic(hw) < 0) {
713 		PMD_DRV_LOG(DEBUG, "SMBI lock released");
714 	}
715 	e1000_put_hw_semaphore_generic(hw);
716 
717 	if (hw->mac.ops.acquire_swfw_sync != NULL) {
718 		uint16_t mask;
719 
720 		/*
721 		 * Phy lock should not fail in this early stage. If this is the case,
722 		 * it is due to an improper exit of the application.
723 		 * So force the release of the faulty lock.
724 		 */
725 		mask = E1000_SWFW_PHY0_SM << hw->bus.func;
726 		if (hw->bus.func > E1000_FUNC_1)
727 			mask <<= 2;
728 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
729 			PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released",
730 				    hw->bus.func);
731 		}
732 		hw->mac.ops.release_swfw_sync(hw, mask);
733 
734 		/*
735 		 * This one is more tricky since it is common to all ports; but
736 		 * swfw_sync retries last long enough (1s) to be almost sure that if
737 		 * lock can not be taken it is due to an improper lock of the
738 		 * semaphore.
739 		 */
740 		mask = E1000_SWFW_EEP_SM;
741 		if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
742 			PMD_DRV_LOG(DEBUG, "SWFW common locks released");
743 		}
744 		hw->mac.ops.release_swfw_sync(hw, mask);
745 	}
746 
747 	return E1000_SUCCESS;
748 }
749 
750 /* Remove all ntuple filters of the device */
751 static int igb_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
752 {
753 	struct e1000_filter_info *filter_info =
754 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
755 	struct e1000_5tuple_filter *p_5tuple;
756 	struct e1000_2tuple_filter *p_2tuple;
757 
758 	while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
759 		TAILQ_REMOVE(&filter_info->fivetuple_list,
760 			p_5tuple, entries);
761 			rte_free(p_5tuple);
762 	}
763 	filter_info->fivetuple_mask = 0;
764 	while ((p_2tuple = TAILQ_FIRST(&filter_info->twotuple_list))) {
765 		TAILQ_REMOVE(&filter_info->twotuple_list,
766 			p_2tuple, entries);
767 			rte_free(p_2tuple);
768 	}
769 	filter_info->twotuple_mask = 0;
770 
771 	return 0;
772 }
773 
774 /* Remove all flex filters of the device */
775 static int igb_flex_filter_uninit(struct rte_eth_dev *eth_dev)
776 {
777 	struct e1000_filter_info *filter_info =
778 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
779 	struct e1000_flex_filter *p_flex;
780 
781 	while ((p_flex = TAILQ_FIRST(&filter_info->flex_list))) {
782 		TAILQ_REMOVE(&filter_info->flex_list, p_flex, entries);
783 		rte_free(p_flex);
784 	}
785 	filter_info->flex_mask = 0;
786 
787 	return 0;
788 }
789 
790 static int
791 eth_igb_dev_init(struct rte_eth_dev *eth_dev)
792 {
793 	int error = 0;
794 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
795 	struct e1000_hw *hw =
796 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
797 	struct e1000_vfta * shadow_vfta =
798 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
799 	struct e1000_filter_info *filter_info =
800 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
801 	struct e1000_adapter *adapter =
802 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
803 
804 	uint32_t ctrl_ext;
805 
806 	eth_dev->dev_ops = &eth_igb_ops;
807 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
808 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
809 	eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
810 
811 	/* for secondary processes, we don't initialise any further as primary
812 	 * has already done this work. Only check we don't need a different
813 	 * RX function */
814 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
815 		if (eth_dev->data->scattered_rx)
816 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
817 		return 0;
818 	}
819 
820 	rte_eth_copy_pci_info(eth_dev, pci_dev);
821 
822 	hw->hw_addr= (void *)pci_dev->mem_resource[0].addr;
823 
824 	igb_identify_hardware(eth_dev, pci_dev);
825 	if (e1000_setup_init_funcs(hw, FALSE) != E1000_SUCCESS) {
826 		error = -EIO;
827 		goto err_late;
828 	}
829 
830 	e1000_get_bus_info(hw);
831 
832 	/* Reset any pending lock */
833 	if (igb_reset_swfw_lock(hw) != E1000_SUCCESS) {
834 		error = -EIO;
835 		goto err_late;
836 	}
837 
838 	/* Finish initialization */
839 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS) {
840 		error = -EIO;
841 		goto err_late;
842 	}
843 
844 	hw->mac.autoneg = 1;
845 	hw->phy.autoneg_wait_to_complete = 0;
846 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
847 
848 	/* Copper options */
849 	if (hw->phy.media_type == e1000_media_type_copper) {
850 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
851 		hw->phy.disable_polarity_correction = 0;
852 		hw->phy.ms_type = e1000_ms_hw_default;
853 	}
854 
855 	/*
856 	 * Start from a known state, this is important in reading the nvm
857 	 * and mac from that.
858 	 */
859 	igb_pf_reset_hw(hw);
860 
861 	/* Make sure we have a good EEPROM before we read from it */
862 	if (e1000_validate_nvm_checksum(hw) < 0) {
863 		/*
864 		 * Some PCI-E parts fail the first check due to
865 		 * the link being in sleep state, call it again,
866 		 * if it fails a second time its a real issue.
867 		 */
868 		if (e1000_validate_nvm_checksum(hw) < 0) {
869 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
870 			error = -EIO;
871 			goto err_late;
872 		}
873 	}
874 
875 	/* Read the permanent MAC address out of the EEPROM */
876 	if (e1000_read_mac_addr(hw) != 0) {
877 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
878 		error = -EIO;
879 		goto err_late;
880 	}
881 
882 	/* Allocate memory for storing MAC addresses */
883 	eth_dev->data->mac_addrs = rte_zmalloc("e1000",
884 		ETHER_ADDR_LEN * hw->mac.rar_entry_count, 0);
885 	if (eth_dev->data->mac_addrs == NULL) {
886 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
887 						"store MAC addresses",
888 				ETHER_ADDR_LEN * hw->mac.rar_entry_count);
889 		error = -ENOMEM;
890 		goto err_late;
891 	}
892 
893 	/* Copy the permanent MAC address */
894 	ether_addr_copy((struct ether_addr *)hw->mac.addr, &eth_dev->data->mac_addrs[0]);
895 
896 	/* initialize the vfta */
897 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
898 
899 	/* Now initialize the hardware */
900 	if (igb_hardware_init(hw) != 0) {
901 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
902 		rte_free(eth_dev->data->mac_addrs);
903 		eth_dev->data->mac_addrs = NULL;
904 		error = -ENODEV;
905 		goto err_late;
906 	}
907 	hw->mac.get_link_status = 1;
908 	adapter->stopped = 0;
909 
910 	/* Indicate SOL/IDER usage */
911 	if (e1000_check_reset_block(hw) < 0) {
912 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to"
913 					"SOL/IDER session");
914 	}
915 
916 	/* initialize PF if max_vfs not zero */
917 	igb_pf_host_init(eth_dev);
918 
919 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
920 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
921 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
922 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
923 	E1000_WRITE_FLUSH(hw);
924 
925 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
926 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
927 		     pci_dev->id.device_id);
928 
929 	rte_intr_callback_register(&pci_dev->intr_handle,
930 				   eth_igb_interrupt_handler,
931 				   (void *)eth_dev);
932 
933 	/* enable uio/vfio intr/eventfd mapping */
934 	rte_intr_enable(&pci_dev->intr_handle);
935 
936 	/* enable support intr */
937 	igb_intr_enable(eth_dev);
938 
939 	/* initialize filter info */
940 	memset(filter_info, 0,
941 	       sizeof(struct e1000_filter_info));
942 
943 	TAILQ_INIT(&filter_info->flex_list);
944 	TAILQ_INIT(&filter_info->twotuple_list);
945 	TAILQ_INIT(&filter_info->fivetuple_list);
946 
947 	TAILQ_INIT(&igb_filter_ntuple_list);
948 	TAILQ_INIT(&igb_filter_ethertype_list);
949 	TAILQ_INIT(&igb_filter_syn_list);
950 	TAILQ_INIT(&igb_filter_flex_list);
951 	TAILQ_INIT(&igb_flow_list);
952 
953 	return 0;
954 
955 err_late:
956 	igb_hw_control_release(hw);
957 
958 	return error;
959 }
960 
961 static int
962 eth_igb_dev_uninit(struct rte_eth_dev *eth_dev)
963 {
964 	struct rte_pci_device *pci_dev;
965 	struct rte_intr_handle *intr_handle;
966 	struct e1000_hw *hw;
967 	struct e1000_adapter *adapter =
968 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
969 	struct e1000_filter_info *filter_info =
970 		E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
971 
972 	PMD_INIT_FUNC_TRACE();
973 
974 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
975 		return -EPERM;
976 
977 	hw = E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
978 	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
979 	intr_handle = &pci_dev->intr_handle;
980 
981 	if (adapter->stopped == 0)
982 		eth_igb_close(eth_dev);
983 
984 	eth_dev->dev_ops = NULL;
985 	eth_dev->rx_pkt_burst = NULL;
986 	eth_dev->tx_pkt_burst = NULL;
987 
988 	/* Reset any pending lock */
989 	igb_reset_swfw_lock(hw);
990 
991 	rte_free(eth_dev->data->mac_addrs);
992 	eth_dev->data->mac_addrs = NULL;
993 
994 	/* uninitialize PF if max_vfs not zero */
995 	igb_pf_host_uninit(eth_dev);
996 
997 	/* disable uio intr before callback unregister */
998 	rte_intr_disable(intr_handle);
999 	rte_intr_callback_unregister(intr_handle,
1000 				     eth_igb_interrupt_handler, eth_dev);
1001 
1002 	/* clear the SYN filter info */
1003 	filter_info->syn_info = 0;
1004 
1005 	/* clear the ethertype filters info */
1006 	filter_info->ethertype_mask = 0;
1007 	memset(filter_info->ethertype_filters, 0,
1008 		E1000_MAX_ETQF_FILTERS * sizeof(struct igb_ethertype_filter));
1009 
1010 	/* remove all ntuple filters of the device */
1011 	igb_ntuple_filter_uninit(eth_dev);
1012 
1013 	/* remove all flex filters of the device */
1014 	igb_flex_filter_uninit(eth_dev);
1015 
1016 	/* clear all the filters list */
1017 	igb_filterlist_flush(eth_dev);
1018 
1019 	return 0;
1020 }
1021 
1022 /*
1023  * Virtual Function device init
1024  */
1025 static int
1026 eth_igbvf_dev_init(struct rte_eth_dev *eth_dev)
1027 {
1028 	struct rte_pci_device *pci_dev;
1029 	struct rte_intr_handle *intr_handle;
1030 	struct e1000_adapter *adapter =
1031 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
1032 	struct e1000_hw *hw =
1033 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1034 	int diag;
1035 	struct ether_addr *perm_addr = (struct ether_addr *)hw->mac.perm_addr;
1036 
1037 	PMD_INIT_FUNC_TRACE();
1038 
1039 	eth_dev->dev_ops = &igbvf_eth_dev_ops;
1040 	eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
1041 	eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
1042 	eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
1043 
1044 	/* for secondary processes, we don't initialise any further as primary
1045 	 * has already done this work. Only check we don't need a different
1046 	 * RX function */
1047 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
1048 		if (eth_dev->data->scattered_rx)
1049 			eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
1050 		return 0;
1051 	}
1052 
1053 	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1054 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1055 
1056 	hw->device_id = pci_dev->id.device_id;
1057 	hw->vendor_id = pci_dev->id.vendor_id;
1058 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1059 	adapter->stopped = 0;
1060 
1061 	/* Initialize the shared code (base driver) */
1062 	diag = e1000_setup_init_funcs(hw, TRUE);
1063 	if (diag != 0) {
1064 		PMD_INIT_LOG(ERR, "Shared code init failed for igbvf: %d",
1065 			diag);
1066 		return -EIO;
1067 	}
1068 
1069 	/* init_mailbox_params */
1070 	hw->mbx.ops.init_params(hw);
1071 
1072 	/* Disable the interrupts for VF */
1073 	igbvf_intr_disable(hw);
1074 
1075 	diag = hw->mac.ops.reset_hw(hw);
1076 
1077 	/* Allocate memory for storing MAC addresses */
1078 	eth_dev->data->mac_addrs = rte_zmalloc("igbvf", ETHER_ADDR_LEN *
1079 		hw->mac.rar_entry_count, 0);
1080 	if (eth_dev->data->mac_addrs == NULL) {
1081 		PMD_INIT_LOG(ERR,
1082 			"Failed to allocate %d bytes needed to store MAC "
1083 			"addresses",
1084 			ETHER_ADDR_LEN * hw->mac.rar_entry_count);
1085 		return -ENOMEM;
1086 	}
1087 
1088 	/* Generate a random MAC address, if none was assigned by PF. */
1089 	if (is_zero_ether_addr(perm_addr)) {
1090 		eth_random_addr(perm_addr->addr_bytes);
1091 		PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1092 		PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1093 			     "%02x:%02x:%02x:%02x:%02x:%02x",
1094 			     perm_addr->addr_bytes[0],
1095 			     perm_addr->addr_bytes[1],
1096 			     perm_addr->addr_bytes[2],
1097 			     perm_addr->addr_bytes[3],
1098 			     perm_addr->addr_bytes[4],
1099 			     perm_addr->addr_bytes[5]);
1100 	}
1101 
1102 	diag = e1000_rar_set(hw, perm_addr->addr_bytes, 0);
1103 	if (diag) {
1104 		rte_free(eth_dev->data->mac_addrs);
1105 		eth_dev->data->mac_addrs = NULL;
1106 		return diag;
1107 	}
1108 	/* Copy the permanent MAC address */
1109 	ether_addr_copy((struct ether_addr *) hw->mac.perm_addr,
1110 			&eth_dev->data->mac_addrs[0]);
1111 
1112 	PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x "
1113 		     "mac.type=%s",
1114 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
1115 		     pci_dev->id.device_id, "igb_mac_82576_vf");
1116 
1117 	intr_handle = &pci_dev->intr_handle;
1118 	rte_intr_callback_register(intr_handle,
1119 				   eth_igbvf_interrupt_handler, eth_dev);
1120 
1121 	return 0;
1122 }
1123 
1124 static int
1125 eth_igbvf_dev_uninit(struct rte_eth_dev *eth_dev)
1126 {
1127 	struct e1000_adapter *adapter =
1128 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
1129 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1130 
1131 	PMD_INIT_FUNC_TRACE();
1132 
1133 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1134 		return -EPERM;
1135 
1136 	if (adapter->stopped == 0)
1137 		igbvf_dev_close(eth_dev);
1138 
1139 	eth_dev->dev_ops = NULL;
1140 	eth_dev->rx_pkt_burst = NULL;
1141 	eth_dev->tx_pkt_burst = NULL;
1142 
1143 	rte_free(eth_dev->data->mac_addrs);
1144 	eth_dev->data->mac_addrs = NULL;
1145 
1146 	/* disable uio intr before callback unregister */
1147 	rte_intr_disable(&pci_dev->intr_handle);
1148 	rte_intr_callback_unregister(&pci_dev->intr_handle,
1149 				     eth_igbvf_interrupt_handler,
1150 				     (void *)eth_dev);
1151 
1152 	return 0;
1153 }
1154 
1155 static int eth_igb_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1156 	struct rte_pci_device *pci_dev)
1157 {
1158 	return rte_eth_dev_pci_generic_probe(pci_dev,
1159 		sizeof(struct e1000_adapter), eth_igb_dev_init);
1160 }
1161 
1162 static int eth_igb_pci_remove(struct rte_pci_device *pci_dev)
1163 {
1164 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igb_dev_uninit);
1165 }
1166 
1167 static struct rte_pci_driver rte_igb_pmd = {
1168 	.id_table = pci_id_igb_map,
1169 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
1170 		     RTE_PCI_DRV_IOVA_AS_VA,
1171 	.probe = eth_igb_pci_probe,
1172 	.remove = eth_igb_pci_remove,
1173 };
1174 
1175 
1176 static int eth_igbvf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1177 	struct rte_pci_device *pci_dev)
1178 {
1179 	return rte_eth_dev_pci_generic_probe(pci_dev,
1180 		sizeof(struct e1000_adapter), eth_igbvf_dev_init);
1181 }
1182 
1183 static int eth_igbvf_pci_remove(struct rte_pci_device *pci_dev)
1184 {
1185 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_igbvf_dev_uninit);
1186 }
1187 
1188 /*
1189  * virtual function driver struct
1190  */
1191 static struct rte_pci_driver rte_igbvf_pmd = {
1192 	.id_table = pci_id_igbvf_map,
1193 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_IOVA_AS_VA,
1194 	.probe = eth_igbvf_pci_probe,
1195 	.remove = eth_igbvf_pci_remove,
1196 };
1197 
1198 static void
1199 igb_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1200 {
1201 	struct e1000_hw *hw =
1202 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1203 	/* RCTL: enable VLAN filter since VMDq always use VLAN filter */
1204 	uint32_t rctl = E1000_READ_REG(hw, E1000_RCTL);
1205 	rctl |= E1000_RCTL_VFE;
1206 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1207 }
1208 
1209 static int
1210 igb_check_mq_mode(struct rte_eth_dev *dev)
1211 {
1212 	enum rte_eth_rx_mq_mode rx_mq_mode = dev->data->dev_conf.rxmode.mq_mode;
1213 	enum rte_eth_tx_mq_mode tx_mq_mode = dev->data->dev_conf.txmode.mq_mode;
1214 	uint16_t nb_rx_q = dev->data->nb_rx_queues;
1215 	uint16_t nb_tx_q = dev->data->nb_tx_queues;
1216 
1217 	if ((rx_mq_mode & ETH_MQ_RX_DCB_FLAG) ||
1218 	    tx_mq_mode == ETH_MQ_TX_DCB ||
1219 	    tx_mq_mode == ETH_MQ_TX_VMDQ_DCB) {
1220 		PMD_INIT_LOG(ERR, "DCB mode is not supported.");
1221 		return -EINVAL;
1222 	}
1223 	if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
1224 		/* Check multi-queue mode.
1225 		 * To no break software we accept ETH_MQ_RX_NONE as this might
1226 		 * be used to turn off VLAN filter.
1227 		 */
1228 
1229 		if (rx_mq_mode == ETH_MQ_RX_NONE ||
1230 		    rx_mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1231 			dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
1232 			RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = 1;
1233 		} else {
1234 			/* Only support one queue on VFs.
1235 			 * RSS together with SRIOV is not supported.
1236 			 */
1237 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1238 					" wrong mq_mode rx %d.",
1239 					rx_mq_mode);
1240 			return -EINVAL;
1241 		}
1242 		/* TX mode is not used here, so mode might be ignored.*/
1243 		if (tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1244 			/* SRIOV only works in VMDq enable mode */
1245 			PMD_INIT_LOG(WARNING, "SRIOV is active,"
1246 					" TX mode %d is not supported. "
1247 					" Driver will behave as %d mode.",
1248 					tx_mq_mode, ETH_MQ_TX_VMDQ_ONLY);
1249 		}
1250 
1251 		/* check valid queue number */
1252 		if ((nb_rx_q > 1) || (nb_tx_q > 1)) {
1253 			PMD_INIT_LOG(ERR, "SRIOV is active,"
1254 					" only support one queue on VFs.");
1255 			return -EINVAL;
1256 		}
1257 	} else {
1258 		/* To no break software that set invalid mode, only display
1259 		 * warning if invalid mode is used.
1260 		 */
1261 		if (rx_mq_mode != ETH_MQ_RX_NONE &&
1262 		    rx_mq_mode != ETH_MQ_RX_VMDQ_ONLY &&
1263 		    rx_mq_mode != ETH_MQ_RX_RSS) {
1264 			/* RSS together with VMDq not supported*/
1265 			PMD_INIT_LOG(ERR, "RX mode %d is not supported.",
1266 				     rx_mq_mode);
1267 			return -EINVAL;
1268 		}
1269 
1270 		if (tx_mq_mode != ETH_MQ_TX_NONE &&
1271 		    tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1272 			PMD_INIT_LOG(WARNING, "TX mode %d is not supported."
1273 					" Due to txmode is meaningless in this"
1274 					" driver, just ignore.",
1275 					tx_mq_mode);
1276 		}
1277 	}
1278 	return 0;
1279 }
1280 
1281 static int
1282 eth_igb_configure(struct rte_eth_dev *dev)
1283 {
1284 	struct e1000_interrupt *intr =
1285 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1286 	int ret;
1287 
1288 	PMD_INIT_FUNC_TRACE();
1289 
1290 	/* multipe queue mode checking */
1291 	ret  = igb_check_mq_mode(dev);
1292 	if (ret != 0) {
1293 		PMD_DRV_LOG(ERR, "igb_check_mq_mode fails with %d.",
1294 			    ret);
1295 		return ret;
1296 	}
1297 
1298 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1299 	PMD_INIT_FUNC_TRACE();
1300 
1301 	return 0;
1302 }
1303 
1304 static int
1305 eth_igb_start(struct rte_eth_dev *dev)
1306 {
1307 	struct e1000_hw *hw =
1308 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1309 	struct e1000_adapter *adapter =
1310 		E1000_DEV_PRIVATE(dev->data->dev_private);
1311 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1312 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1313 	int ret, mask;
1314 	uint32_t intr_vector = 0;
1315 	uint32_t ctrl_ext;
1316 	uint32_t *speeds;
1317 	int num_speeds;
1318 	bool autoneg;
1319 
1320 	PMD_INIT_FUNC_TRACE();
1321 
1322 	/* disable uio/vfio intr/eventfd mapping */
1323 	rte_intr_disable(intr_handle);
1324 
1325 	/* Power up the phy. Needed to make the link go Up */
1326 	eth_igb_dev_set_link_up(dev);
1327 
1328 	/*
1329 	 * Packet Buffer Allocation (PBA)
1330 	 * Writing PBA sets the receive portion of the buffer
1331 	 * the remainder is used for the transmit buffer.
1332 	 */
1333 	if (hw->mac.type == e1000_82575) {
1334 		uint32_t pba;
1335 
1336 		pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
1337 		E1000_WRITE_REG(hw, E1000_PBA, pba);
1338 	}
1339 
1340 	/* Put the address into the Receive Address Array */
1341 	e1000_rar_set(hw, hw->mac.addr, 0);
1342 
1343 	/* Initialize the hardware */
1344 	if (igb_hardware_init(hw)) {
1345 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
1346 		return -EIO;
1347 	}
1348 	adapter->stopped = 0;
1349 
1350 	E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN << 16 | ETHER_TYPE_VLAN);
1351 
1352 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1353 	/* Set PF Reset Done bit so PF/VF Mail Ops can work */
1354 	ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
1355 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
1356 	E1000_WRITE_FLUSH(hw);
1357 
1358 	/* configure PF module if SRIOV enabled */
1359 	igb_pf_host_configure(dev);
1360 
1361 	/* check and configure queue intr-vector mapping */
1362 	if ((rte_intr_cap_multiple(intr_handle) ||
1363 	     !RTE_ETH_DEV_SRIOV(dev).active) &&
1364 	    dev->data->dev_conf.intr_conf.rxq != 0) {
1365 		intr_vector = dev->data->nb_rx_queues;
1366 		if (rte_intr_efd_enable(intr_handle, intr_vector))
1367 			return -1;
1368 	}
1369 
1370 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
1371 		intr_handle->intr_vec =
1372 			rte_zmalloc("intr_vec",
1373 				    dev->data->nb_rx_queues * sizeof(int), 0);
1374 		if (intr_handle->intr_vec == NULL) {
1375 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
1376 				     " intr_vec", dev->data->nb_rx_queues);
1377 			return -ENOMEM;
1378 		}
1379 	}
1380 
1381 	/* confiugre msix for rx interrupt */
1382 	eth_igb_configure_msix_intr(dev);
1383 
1384 	/* Configure for OS presence */
1385 	igb_init_manageability(hw);
1386 
1387 	eth_igb_tx_init(dev);
1388 
1389 	/* This can fail when allocating mbufs for descriptor rings */
1390 	ret = eth_igb_rx_init(dev);
1391 	if (ret) {
1392 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
1393 		igb_dev_clear_queues(dev);
1394 		return ret;
1395 	}
1396 
1397 	e1000_clear_hw_cntrs_base_generic(hw);
1398 
1399 	/*
1400 	 * VLAN Offload Settings
1401 	 */
1402 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
1403 			ETH_VLAN_EXTEND_MASK;
1404 	ret = eth_igb_vlan_offload_set(dev, mask);
1405 	if (ret) {
1406 		PMD_INIT_LOG(ERR, "Unable to set vlan offload");
1407 		igb_dev_clear_queues(dev);
1408 		return ret;
1409 	}
1410 
1411 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1412 		/* Enable VLAN filter since VMDq always use VLAN filter */
1413 		igb_vmdq_vlan_hw_filter_enable(dev);
1414 	}
1415 
1416 	if ((hw->mac.type == e1000_82576) || (hw->mac.type == e1000_82580) ||
1417 		(hw->mac.type == e1000_i350) || (hw->mac.type == e1000_i210) ||
1418 		(hw->mac.type == e1000_i211)) {
1419 		/* Configure EITR with the maximum possible value (0xFFFF) */
1420 		E1000_WRITE_REG(hw, E1000_EITR(0), 0xFFFF);
1421 	}
1422 
1423 	/* Setup link speed and duplex */
1424 	speeds = &dev->data->dev_conf.link_speeds;
1425 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
1426 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
1427 		hw->mac.autoneg = 1;
1428 	} else {
1429 		num_speeds = 0;
1430 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
1431 
1432 		/* Reset */
1433 		hw->phy.autoneg_advertised = 0;
1434 
1435 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1436 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1437 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
1438 			num_speeds = -1;
1439 			goto error_invalid_config;
1440 		}
1441 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
1442 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1443 			num_speeds++;
1444 		}
1445 		if (*speeds & ETH_LINK_SPEED_10M) {
1446 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1447 			num_speeds++;
1448 		}
1449 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
1450 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1451 			num_speeds++;
1452 		}
1453 		if (*speeds & ETH_LINK_SPEED_100M) {
1454 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1455 			num_speeds++;
1456 		}
1457 		if (*speeds & ETH_LINK_SPEED_1G) {
1458 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1459 			num_speeds++;
1460 		}
1461 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
1462 			goto error_invalid_config;
1463 
1464 		/* Set/reset the mac.autoneg based on the link speed,
1465 		 * fixed or not
1466 		 */
1467 		if (!autoneg) {
1468 			hw->mac.autoneg = 0;
1469 			hw->mac.forced_speed_duplex =
1470 					hw->phy.autoneg_advertised;
1471 		} else {
1472 			hw->mac.autoneg = 1;
1473 		}
1474 	}
1475 
1476 	e1000_setup_link(hw);
1477 
1478 	if (rte_intr_allow_others(intr_handle)) {
1479 		/* check if lsc interrupt is enabled */
1480 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1481 			eth_igb_lsc_interrupt_setup(dev, TRUE);
1482 		else
1483 			eth_igb_lsc_interrupt_setup(dev, FALSE);
1484 	} else {
1485 		rte_intr_callback_unregister(intr_handle,
1486 					     eth_igb_interrupt_handler,
1487 					     (void *)dev);
1488 		if (dev->data->dev_conf.intr_conf.lsc != 0)
1489 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
1490 				     " no intr multiplex");
1491 	}
1492 
1493 	/* check if rxq interrupt is enabled */
1494 	if (dev->data->dev_conf.intr_conf.rxq != 0 &&
1495 	    rte_intr_dp_is_en(intr_handle))
1496 		eth_igb_rxq_interrupt_setup(dev);
1497 
1498 	/* enable uio/vfio intr/eventfd mapping */
1499 	rte_intr_enable(intr_handle);
1500 
1501 	/* resume enabled intr since hw reset */
1502 	igb_intr_enable(dev);
1503 
1504 	/* restore all types filter */
1505 	igb_filter_restore(dev);
1506 
1507 	PMD_INIT_LOG(DEBUG, "<<");
1508 
1509 	return 0;
1510 
1511 error_invalid_config:
1512 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
1513 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
1514 	igb_dev_clear_queues(dev);
1515 	return -EINVAL;
1516 }
1517 
1518 /*********************************************************************
1519  *
1520  *  This routine disables all traffic on the adapter by issuing a
1521  *  global reset on the MAC.
1522  *
1523  **********************************************************************/
1524 static void
1525 eth_igb_stop(struct rte_eth_dev *dev)
1526 {
1527 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1528 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1529 	struct rte_eth_link link;
1530 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1531 
1532 	igb_intr_disable(hw);
1533 
1534 	/* disable intr eventfd mapping */
1535 	rte_intr_disable(intr_handle);
1536 
1537 	igb_pf_reset_hw(hw);
1538 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1539 
1540 	/* Set bit for Go Link disconnect */
1541 	if (hw->mac.type >= e1000_82580) {
1542 		uint32_t phpm_reg;
1543 
1544 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1545 		phpm_reg |= E1000_82580_PM_GO_LINKD;
1546 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1547 	}
1548 
1549 	/* Power down the phy. Needed to make the link go Down */
1550 	eth_igb_dev_set_link_down(dev);
1551 
1552 	igb_dev_clear_queues(dev);
1553 
1554 	/* clear the recorded link status */
1555 	memset(&link, 0, sizeof(link));
1556 	rte_igb_dev_atomic_write_link_status(dev, &link);
1557 
1558 	if (!rte_intr_allow_others(intr_handle))
1559 		/* resume to the default handler */
1560 		rte_intr_callback_register(intr_handle,
1561 					   eth_igb_interrupt_handler,
1562 					   (void *)dev);
1563 
1564 	/* Clean datapath event and queue/vec mapping */
1565 	rte_intr_efd_disable(intr_handle);
1566 	if (intr_handle->intr_vec != NULL) {
1567 		rte_free(intr_handle->intr_vec);
1568 		intr_handle->intr_vec = NULL;
1569 	}
1570 }
1571 
1572 static int
1573 eth_igb_dev_set_link_up(struct rte_eth_dev *dev)
1574 {
1575 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1576 
1577 	if (hw->phy.media_type == e1000_media_type_copper)
1578 		e1000_power_up_phy(hw);
1579 	else
1580 		e1000_power_up_fiber_serdes_link(hw);
1581 
1582 	return 0;
1583 }
1584 
1585 static int
1586 eth_igb_dev_set_link_down(struct rte_eth_dev *dev)
1587 {
1588 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1589 
1590 	if (hw->phy.media_type == e1000_media_type_copper)
1591 		e1000_power_down_phy(hw);
1592 	else
1593 		e1000_shutdown_fiber_serdes_link(hw);
1594 
1595 	return 0;
1596 }
1597 
1598 static void
1599 eth_igb_close(struct rte_eth_dev *dev)
1600 {
1601 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1602 	struct e1000_adapter *adapter =
1603 		E1000_DEV_PRIVATE(dev->data->dev_private);
1604 	struct rte_eth_link link;
1605 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1606 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1607 
1608 	eth_igb_stop(dev);
1609 	adapter->stopped = 1;
1610 
1611 	e1000_phy_hw_reset(hw);
1612 	igb_release_manageability(hw);
1613 	igb_hw_control_release(hw);
1614 
1615 	/* Clear bit for Go Link disconnect */
1616 	if (hw->mac.type >= e1000_82580) {
1617 		uint32_t phpm_reg;
1618 
1619 		phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1620 		phpm_reg &= ~E1000_82580_PM_GO_LINKD;
1621 		E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1622 	}
1623 
1624 	igb_dev_free_queues(dev);
1625 
1626 	if (intr_handle->intr_vec) {
1627 		rte_free(intr_handle->intr_vec);
1628 		intr_handle->intr_vec = NULL;
1629 	}
1630 
1631 	memset(&link, 0, sizeof(link));
1632 	rte_igb_dev_atomic_write_link_status(dev, &link);
1633 }
1634 
1635 static int
1636 igb_get_rx_buffer_size(struct e1000_hw *hw)
1637 {
1638 	uint32_t rx_buf_size;
1639 	if (hw->mac.type == e1000_82576) {
1640 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xffff) << 10;
1641 	} else if (hw->mac.type == e1000_82580 || hw->mac.type == e1000_i350) {
1642 		/* PBS needs to be translated according to a lookup table */
1643 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xf);
1644 		rx_buf_size = (uint32_t) e1000_rxpbs_adjust_82580(rx_buf_size);
1645 		rx_buf_size = (rx_buf_size << 10);
1646 	} else if (hw->mac.type == e1000_i210 || hw->mac.type == e1000_i211) {
1647 		rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0x3f) << 10;
1648 	} else {
1649 		rx_buf_size = (E1000_READ_REG(hw, E1000_PBA) & 0xffff) << 10;
1650 	}
1651 
1652 	return rx_buf_size;
1653 }
1654 
1655 /*********************************************************************
1656  *
1657  *  Initialize the hardware
1658  *
1659  **********************************************************************/
1660 static int
1661 igb_hardware_init(struct e1000_hw *hw)
1662 {
1663 	uint32_t rx_buf_size;
1664 	int diag;
1665 
1666 	/* Let the firmware know the OS is in control */
1667 	igb_hw_control_acquire(hw);
1668 
1669 	/*
1670 	 * These parameters control the automatic generation (Tx) and
1671 	 * response (Rx) to Ethernet PAUSE frames.
1672 	 * - High water mark should allow for at least two standard size (1518)
1673 	 *   frames to be received after sending an XOFF.
1674 	 * - Low water mark works best when it is very near the high water mark.
1675 	 *   This allows the receiver to restart by sending XON when it has
1676 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
1677 	 *   restart after one full frame is pulled from the buffer. There
1678 	 *   could be several smaller frames in the buffer and if so they will
1679 	 *   not trigger the XON until their total number reduces the buffer
1680 	 *   by 1500.
1681 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
1682 	 */
1683 	rx_buf_size = igb_get_rx_buffer_size(hw);
1684 
1685 	hw->fc.high_water = rx_buf_size - (ETHER_MAX_LEN * 2);
1686 	hw->fc.low_water = hw->fc.high_water - 1500;
1687 	hw->fc.pause_time = IGB_FC_PAUSE_TIME;
1688 	hw->fc.send_xon = 1;
1689 
1690 	/* Set Flow control, use the tunable location if sane */
1691 	if ((igb_fc_setting != e1000_fc_none) && (igb_fc_setting < 4))
1692 		hw->fc.requested_mode = igb_fc_setting;
1693 	else
1694 		hw->fc.requested_mode = e1000_fc_none;
1695 
1696 	/* Issue a global reset */
1697 	igb_pf_reset_hw(hw);
1698 	E1000_WRITE_REG(hw, E1000_WUC, 0);
1699 
1700 	diag = e1000_init_hw(hw);
1701 	if (diag < 0)
1702 		return diag;
1703 
1704 	E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN << 16 | ETHER_TYPE_VLAN);
1705 	e1000_get_phy_info(hw);
1706 	e1000_check_for_link(hw);
1707 
1708 	return 0;
1709 }
1710 
1711 /* This function is based on igb_update_stats_counters() in igb/if_igb.c */
1712 static void
1713 igb_read_stats_registers(struct e1000_hw *hw, struct e1000_hw_stats *stats)
1714 {
1715 	int pause_frames;
1716 
1717 	uint64_t old_gprc  = stats->gprc;
1718 	uint64_t old_gptc  = stats->gptc;
1719 	uint64_t old_tpr   = stats->tpr;
1720 	uint64_t old_tpt   = stats->tpt;
1721 	uint64_t old_rpthc = stats->rpthc;
1722 	uint64_t old_hgptc = stats->hgptc;
1723 
1724 	if(hw->phy.media_type == e1000_media_type_copper ||
1725 	    (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
1726 		stats->symerrs +=
1727 		    E1000_READ_REG(hw,E1000_SYMERRS);
1728 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
1729 	}
1730 
1731 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
1732 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
1733 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
1734 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
1735 
1736 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
1737 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
1738 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
1739 	stats->dc += E1000_READ_REG(hw, E1000_DC);
1740 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
1741 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
1742 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
1743 	/*
1744 	** For watchdog management we need to know if we have been
1745 	** paused during the last interval, so capture that here.
1746 	*/
1747 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
1748 	stats->xoffrxc += pause_frames;
1749 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
1750 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
1751 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
1752 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
1753 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
1754 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
1755 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
1756 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
1757 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
1758 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
1759 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
1760 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
1761 
1762 	/* For the 64-bit byte counters the low dword must be read first. */
1763 	/* Both registers clear on the read of the high dword */
1764 
1765 	/* Workaround CRC bytes included in size, take away 4 bytes/packet */
1766 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
1767 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
1768 	stats->gorc -= (stats->gprc - old_gprc) * ETHER_CRC_LEN;
1769 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
1770 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
1771 	stats->gotc -= (stats->gptc - old_gptc) * ETHER_CRC_LEN;
1772 
1773 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
1774 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
1775 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
1776 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
1777 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
1778 
1779 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
1780 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
1781 
1782 	stats->tor += E1000_READ_REG(hw, E1000_TORL);
1783 	stats->tor += ((uint64_t)E1000_READ_REG(hw, E1000_TORH) << 32);
1784 	stats->tor -= (stats->tpr - old_tpr) * ETHER_CRC_LEN;
1785 	stats->tot += E1000_READ_REG(hw, E1000_TOTL);
1786 	stats->tot += ((uint64_t)E1000_READ_REG(hw, E1000_TOTH) << 32);
1787 	stats->tot -= (stats->tpt - old_tpt) * ETHER_CRC_LEN;
1788 
1789 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
1790 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
1791 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
1792 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
1793 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
1794 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
1795 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
1796 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
1797 
1798 	/* Interrupt Counts */
1799 
1800 	stats->iac += E1000_READ_REG(hw, E1000_IAC);
1801 	stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
1802 	stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
1803 	stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
1804 	stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
1805 	stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
1806 	stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
1807 	stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
1808 	stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
1809 
1810 	/* Host to Card Statistics */
1811 
1812 	stats->cbtmpc += E1000_READ_REG(hw, E1000_CBTMPC);
1813 	stats->htdpmc += E1000_READ_REG(hw, E1000_HTDPMC);
1814 	stats->cbrdpc += E1000_READ_REG(hw, E1000_CBRDPC);
1815 	stats->cbrmpc += E1000_READ_REG(hw, E1000_CBRMPC);
1816 	stats->rpthc += E1000_READ_REG(hw, E1000_RPTHC);
1817 	stats->hgptc += E1000_READ_REG(hw, E1000_HGPTC);
1818 	stats->htcbdpc += E1000_READ_REG(hw, E1000_HTCBDPC);
1819 	stats->hgorc += E1000_READ_REG(hw, E1000_HGORCL);
1820 	stats->hgorc += ((uint64_t)E1000_READ_REG(hw, E1000_HGORCH) << 32);
1821 	stats->hgorc -= (stats->rpthc - old_rpthc) * ETHER_CRC_LEN;
1822 	stats->hgotc += E1000_READ_REG(hw, E1000_HGOTCL);
1823 	stats->hgotc += ((uint64_t)E1000_READ_REG(hw, E1000_HGOTCH) << 32);
1824 	stats->hgotc -= (stats->hgptc - old_hgptc) * ETHER_CRC_LEN;
1825 	stats->lenerrs += E1000_READ_REG(hw, E1000_LENERRS);
1826 	stats->scvpc += E1000_READ_REG(hw, E1000_SCVPC);
1827 	stats->hrmpc += E1000_READ_REG(hw, E1000_HRMPC);
1828 
1829 	stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
1830 	stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
1831 	stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
1832 	stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
1833 	stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
1834 	stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
1835 }
1836 
1837 static int
1838 eth_igb_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1839 {
1840 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1841 	struct e1000_hw_stats *stats =
1842 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1843 
1844 	igb_read_stats_registers(hw, stats);
1845 
1846 	if (rte_stats == NULL)
1847 		return -EINVAL;
1848 
1849 	/* Rx Errors */
1850 	rte_stats->imissed = stats->mpc;
1851 	rte_stats->ierrors = stats->crcerrs +
1852 	                     stats->rlec + stats->ruc + stats->roc +
1853 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
1854 
1855 	/* Tx Errors */
1856 	rte_stats->oerrors = stats->ecol + stats->latecol;
1857 
1858 	rte_stats->ipackets = stats->gprc;
1859 	rte_stats->opackets = stats->gptc;
1860 	rte_stats->ibytes   = stats->gorc;
1861 	rte_stats->obytes   = stats->gotc;
1862 	return 0;
1863 }
1864 
1865 static void
1866 eth_igb_stats_reset(struct rte_eth_dev *dev)
1867 {
1868 	struct e1000_hw_stats *hw_stats =
1869 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1870 
1871 	/* HW registers are cleared on read */
1872 	eth_igb_stats_get(dev, NULL);
1873 
1874 	/* Reset software totals */
1875 	memset(hw_stats, 0, sizeof(*hw_stats));
1876 }
1877 
1878 static void
1879 eth_igb_xstats_reset(struct rte_eth_dev *dev)
1880 {
1881 	struct e1000_hw_stats *stats =
1882 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1883 
1884 	/* HW registers are cleared on read */
1885 	eth_igb_xstats_get(dev, NULL, IGB_NB_XSTATS);
1886 
1887 	/* Reset software totals */
1888 	memset(stats, 0, sizeof(*stats));
1889 }
1890 
1891 static int eth_igb_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1892 	struct rte_eth_xstat_name *xstats_names,
1893 	__rte_unused unsigned int size)
1894 {
1895 	unsigned i;
1896 
1897 	if (xstats_names == NULL)
1898 		return IGB_NB_XSTATS;
1899 
1900 	/* Note: limit checked in rte_eth_xstats_names() */
1901 
1902 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1903 		snprintf(xstats_names[i].name, sizeof(xstats_names[i].name),
1904 			 "%s", rte_igb_stats_strings[i].name);
1905 	}
1906 
1907 	return IGB_NB_XSTATS;
1908 }
1909 
1910 static int eth_igb_xstats_get_names_by_id(struct rte_eth_dev *dev,
1911 		struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
1912 		unsigned int limit)
1913 {
1914 	unsigned int i;
1915 
1916 	if (!ids) {
1917 		if (xstats_names == NULL)
1918 			return IGB_NB_XSTATS;
1919 
1920 		for (i = 0; i < IGB_NB_XSTATS; i++)
1921 			snprintf(xstats_names[i].name,
1922 					sizeof(xstats_names[i].name),
1923 					"%s", rte_igb_stats_strings[i].name);
1924 
1925 		return IGB_NB_XSTATS;
1926 
1927 	} else {
1928 		struct rte_eth_xstat_name xstats_names_copy[IGB_NB_XSTATS];
1929 
1930 		eth_igb_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
1931 				IGB_NB_XSTATS);
1932 
1933 		for (i = 0; i < limit; i++) {
1934 			if (ids[i] >= IGB_NB_XSTATS) {
1935 				PMD_INIT_LOG(ERR, "id value isn't valid");
1936 				return -1;
1937 			}
1938 			strcpy(xstats_names[i].name,
1939 					xstats_names_copy[ids[i]].name);
1940 		}
1941 		return limit;
1942 	}
1943 }
1944 
1945 static int
1946 eth_igb_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1947 		   unsigned n)
1948 {
1949 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1950 	struct e1000_hw_stats *hw_stats =
1951 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1952 	unsigned i;
1953 
1954 	if (n < IGB_NB_XSTATS)
1955 		return IGB_NB_XSTATS;
1956 
1957 	igb_read_stats_registers(hw, hw_stats);
1958 
1959 	/* If this is a reset xstats is NULL, and we have cleared the
1960 	 * registers by reading them.
1961 	 */
1962 	if (!xstats)
1963 		return 0;
1964 
1965 	/* Extended stats */
1966 	for (i = 0; i < IGB_NB_XSTATS; i++) {
1967 		xstats[i].id = i;
1968 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1969 			rte_igb_stats_strings[i].offset);
1970 	}
1971 
1972 	return IGB_NB_XSTATS;
1973 }
1974 
1975 static int
1976 eth_igb_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
1977 		uint64_t *values, unsigned int n)
1978 {
1979 	unsigned int i;
1980 
1981 	if (!ids) {
1982 		struct e1000_hw *hw =
1983 			E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1984 		struct e1000_hw_stats *hw_stats =
1985 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1986 
1987 		if (n < IGB_NB_XSTATS)
1988 			return IGB_NB_XSTATS;
1989 
1990 		igb_read_stats_registers(hw, hw_stats);
1991 
1992 		/* If this is a reset xstats is NULL, and we have cleared the
1993 		 * registers by reading them.
1994 		 */
1995 		if (!values)
1996 			return 0;
1997 
1998 		/* Extended stats */
1999 		for (i = 0; i < IGB_NB_XSTATS; i++)
2000 			values[i] = *(uint64_t *)(((char *)hw_stats) +
2001 					rte_igb_stats_strings[i].offset);
2002 
2003 		return IGB_NB_XSTATS;
2004 
2005 	} else {
2006 		uint64_t values_copy[IGB_NB_XSTATS];
2007 
2008 		eth_igb_xstats_get_by_id(dev, NULL, values_copy,
2009 				IGB_NB_XSTATS);
2010 
2011 		for (i = 0; i < n; i++) {
2012 			if (ids[i] >= IGB_NB_XSTATS) {
2013 				PMD_INIT_LOG(ERR, "id value isn't valid");
2014 				return -1;
2015 			}
2016 			values[i] = values_copy[ids[i]];
2017 		}
2018 		return n;
2019 	}
2020 }
2021 
2022 static void
2023 igbvf_read_stats_registers(struct e1000_hw *hw, struct e1000_vf_stats *hw_stats)
2024 {
2025 	/* Good Rx packets, include VF loopback */
2026 	UPDATE_VF_STAT(E1000_VFGPRC,
2027 	    hw_stats->last_gprc, hw_stats->gprc);
2028 
2029 	/* Good Rx octets, include VF loopback */
2030 	UPDATE_VF_STAT(E1000_VFGORC,
2031 	    hw_stats->last_gorc, hw_stats->gorc);
2032 
2033 	/* Good Tx packets, include VF loopback */
2034 	UPDATE_VF_STAT(E1000_VFGPTC,
2035 	    hw_stats->last_gptc, hw_stats->gptc);
2036 
2037 	/* Good Tx octets, include VF loopback */
2038 	UPDATE_VF_STAT(E1000_VFGOTC,
2039 	    hw_stats->last_gotc, hw_stats->gotc);
2040 
2041 	/* Rx Multicst packets */
2042 	UPDATE_VF_STAT(E1000_VFMPRC,
2043 	    hw_stats->last_mprc, hw_stats->mprc);
2044 
2045 	/* Good Rx loopback packets */
2046 	UPDATE_VF_STAT(E1000_VFGPRLBC,
2047 	    hw_stats->last_gprlbc, hw_stats->gprlbc);
2048 
2049 	/* Good Rx loopback octets */
2050 	UPDATE_VF_STAT(E1000_VFGORLBC,
2051 	    hw_stats->last_gorlbc, hw_stats->gorlbc);
2052 
2053 	/* Good Tx loopback packets */
2054 	UPDATE_VF_STAT(E1000_VFGPTLBC,
2055 	    hw_stats->last_gptlbc, hw_stats->gptlbc);
2056 
2057 	/* Good Tx loopback octets */
2058 	UPDATE_VF_STAT(E1000_VFGOTLBC,
2059 	    hw_stats->last_gotlbc, hw_stats->gotlbc);
2060 }
2061 
2062 static int eth_igbvf_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
2063 				     struct rte_eth_xstat_name *xstats_names,
2064 				     __rte_unused unsigned limit)
2065 {
2066 	unsigned i;
2067 
2068 	if (xstats_names != NULL)
2069 		for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2070 			snprintf(xstats_names[i].name,
2071 				sizeof(xstats_names[i].name), "%s",
2072 				rte_igbvf_stats_strings[i].name);
2073 		}
2074 	return IGBVF_NB_XSTATS;
2075 }
2076 
2077 static int
2078 eth_igbvf_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
2079 		     unsigned n)
2080 {
2081 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2082 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2083 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2084 	unsigned i;
2085 
2086 	if (n < IGBVF_NB_XSTATS)
2087 		return IGBVF_NB_XSTATS;
2088 
2089 	igbvf_read_stats_registers(hw, hw_stats);
2090 
2091 	if (!xstats)
2092 		return 0;
2093 
2094 	for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2095 		xstats[i].id = i;
2096 		xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
2097 			rte_igbvf_stats_strings[i].offset);
2098 	}
2099 
2100 	return IGBVF_NB_XSTATS;
2101 }
2102 
2103 static int
2104 eth_igbvf_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
2105 {
2106 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2107 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2108 			  E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2109 
2110 	igbvf_read_stats_registers(hw, hw_stats);
2111 
2112 	if (rte_stats == NULL)
2113 		return -EINVAL;
2114 
2115 	rte_stats->ipackets = hw_stats->gprc;
2116 	rte_stats->ibytes = hw_stats->gorc;
2117 	rte_stats->opackets = hw_stats->gptc;
2118 	rte_stats->obytes = hw_stats->gotc;
2119 	return 0;
2120 }
2121 
2122 static void
2123 eth_igbvf_stats_reset(struct rte_eth_dev *dev)
2124 {
2125 	struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats*)
2126 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2127 
2128 	/* Sync HW register to the last stats */
2129 	eth_igbvf_stats_get(dev, NULL);
2130 
2131 	/* reset HW current stats*/
2132 	memset(&hw_stats->gprc, 0, sizeof(*hw_stats) -
2133 	       offsetof(struct e1000_vf_stats, gprc));
2134 }
2135 
2136 static int
2137 eth_igb_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
2138 		       size_t fw_size)
2139 {
2140 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2141 	struct e1000_fw_version fw;
2142 	int ret;
2143 
2144 	e1000_get_fw_version(hw, &fw);
2145 
2146 	switch (hw->mac.type) {
2147 	case e1000_i210:
2148 	case e1000_i211:
2149 		if (!(e1000_get_flash_presence_i210(hw))) {
2150 			ret = snprintf(fw_version, fw_size,
2151 				 "%2d.%2d-%d",
2152 				 fw.invm_major, fw.invm_minor,
2153 				 fw.invm_img_type);
2154 			break;
2155 		}
2156 		/* fall through */
2157 	default:
2158 		/* if option rom is valid, display its version too */
2159 		if (fw.or_valid) {
2160 			ret = snprintf(fw_version, fw_size,
2161 				 "%d.%d, 0x%08x, %d.%d.%d",
2162 				 fw.eep_major, fw.eep_minor, fw.etrack_id,
2163 				 fw.or_major, fw.or_build, fw.or_patch);
2164 		/* no option rom */
2165 		} else {
2166 			if (fw.etrack_id != 0X0000) {
2167 				ret = snprintf(fw_version, fw_size,
2168 					 "%d.%d, 0x%08x",
2169 					 fw.eep_major, fw.eep_minor,
2170 					 fw.etrack_id);
2171 			} else {
2172 				ret = snprintf(fw_version, fw_size,
2173 					 "%d.%d.%d",
2174 					 fw.eep_major, fw.eep_minor,
2175 					 fw.eep_build);
2176 			}
2177 		}
2178 		break;
2179 	}
2180 
2181 	ret += 1; /* add the size of '\0' */
2182 	if (fw_size < (u32)ret)
2183 		return ret;
2184 	else
2185 		return 0;
2186 }
2187 
2188 static void
2189 eth_igb_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2190 {
2191 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2192 
2193 	dev_info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2194 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2195 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2196 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2197 	dev_info->rx_offload_capa =
2198 		DEV_RX_OFFLOAD_VLAN_STRIP |
2199 		DEV_RX_OFFLOAD_IPV4_CKSUM |
2200 		DEV_RX_OFFLOAD_UDP_CKSUM  |
2201 		DEV_RX_OFFLOAD_TCP_CKSUM;
2202 	dev_info->tx_offload_capa =
2203 		DEV_TX_OFFLOAD_VLAN_INSERT |
2204 		DEV_TX_OFFLOAD_IPV4_CKSUM  |
2205 		DEV_TX_OFFLOAD_UDP_CKSUM   |
2206 		DEV_TX_OFFLOAD_TCP_CKSUM   |
2207 		DEV_TX_OFFLOAD_SCTP_CKSUM  |
2208 		DEV_TX_OFFLOAD_TCP_TSO;
2209 
2210 	switch (hw->mac.type) {
2211 	case e1000_82575:
2212 		dev_info->max_rx_queues = 4;
2213 		dev_info->max_tx_queues = 4;
2214 		dev_info->max_vmdq_pools = 0;
2215 		break;
2216 
2217 	case e1000_82576:
2218 		dev_info->max_rx_queues = 16;
2219 		dev_info->max_tx_queues = 16;
2220 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2221 		dev_info->vmdq_queue_num = 16;
2222 		break;
2223 
2224 	case e1000_82580:
2225 		dev_info->max_rx_queues = 8;
2226 		dev_info->max_tx_queues = 8;
2227 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2228 		dev_info->vmdq_queue_num = 8;
2229 		break;
2230 
2231 	case e1000_i350:
2232 		dev_info->max_rx_queues = 8;
2233 		dev_info->max_tx_queues = 8;
2234 		dev_info->max_vmdq_pools = ETH_8_POOLS;
2235 		dev_info->vmdq_queue_num = 8;
2236 		break;
2237 
2238 	case e1000_i354:
2239 		dev_info->max_rx_queues = 8;
2240 		dev_info->max_tx_queues = 8;
2241 		break;
2242 
2243 	case e1000_i210:
2244 		dev_info->max_rx_queues = 4;
2245 		dev_info->max_tx_queues = 4;
2246 		dev_info->max_vmdq_pools = 0;
2247 		break;
2248 
2249 	case e1000_i211:
2250 		dev_info->max_rx_queues = 2;
2251 		dev_info->max_tx_queues = 2;
2252 		dev_info->max_vmdq_pools = 0;
2253 		break;
2254 
2255 	default:
2256 		/* Should not happen */
2257 		break;
2258 	}
2259 	dev_info->hash_key_size = IGB_HKEY_MAX_INDEX * sizeof(uint32_t);
2260 	dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
2261 	dev_info->flow_type_rss_offloads = IGB_RSS_OFFLOAD_ALL;
2262 
2263 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2264 		.rx_thresh = {
2265 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2266 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2267 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2268 		},
2269 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2270 		.rx_drop_en = 0,
2271 	};
2272 
2273 	dev_info->default_txconf = (struct rte_eth_txconf) {
2274 		.tx_thresh = {
2275 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2276 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2277 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2278 		},
2279 		.txq_flags = 0,
2280 	};
2281 
2282 	dev_info->rx_desc_lim = rx_desc_lim;
2283 	dev_info->tx_desc_lim = tx_desc_lim;
2284 
2285 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
2286 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
2287 			ETH_LINK_SPEED_1G;
2288 }
2289 
2290 static const uint32_t *
2291 eth_igb_supported_ptypes_get(struct rte_eth_dev *dev)
2292 {
2293 	static const uint32_t ptypes[] = {
2294 		/* refers to igb_rxd_pkt_info_to_pkt_type() */
2295 		RTE_PTYPE_L2_ETHER,
2296 		RTE_PTYPE_L3_IPV4,
2297 		RTE_PTYPE_L3_IPV4_EXT,
2298 		RTE_PTYPE_L3_IPV6,
2299 		RTE_PTYPE_L3_IPV6_EXT,
2300 		RTE_PTYPE_L4_TCP,
2301 		RTE_PTYPE_L4_UDP,
2302 		RTE_PTYPE_L4_SCTP,
2303 		RTE_PTYPE_TUNNEL_IP,
2304 		RTE_PTYPE_INNER_L3_IPV6,
2305 		RTE_PTYPE_INNER_L3_IPV6_EXT,
2306 		RTE_PTYPE_INNER_L4_TCP,
2307 		RTE_PTYPE_INNER_L4_UDP,
2308 		RTE_PTYPE_UNKNOWN
2309 	};
2310 
2311 	if (dev->rx_pkt_burst == eth_igb_recv_pkts ||
2312 	    dev->rx_pkt_burst == eth_igb_recv_scattered_pkts)
2313 		return ptypes;
2314 	return NULL;
2315 }
2316 
2317 static void
2318 eth_igbvf_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2319 {
2320 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2321 
2322 	dev_info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2323 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2324 	dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2325 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2326 	dev_info->rx_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP |
2327 				DEV_RX_OFFLOAD_IPV4_CKSUM |
2328 				DEV_RX_OFFLOAD_UDP_CKSUM  |
2329 				DEV_RX_OFFLOAD_TCP_CKSUM;
2330 	dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
2331 				DEV_TX_OFFLOAD_IPV4_CKSUM  |
2332 				DEV_TX_OFFLOAD_UDP_CKSUM   |
2333 				DEV_TX_OFFLOAD_TCP_CKSUM   |
2334 				DEV_TX_OFFLOAD_SCTP_CKSUM  |
2335 				DEV_TX_OFFLOAD_TCP_TSO;
2336 	switch (hw->mac.type) {
2337 	case e1000_vfadapt:
2338 		dev_info->max_rx_queues = 2;
2339 		dev_info->max_tx_queues = 2;
2340 		break;
2341 	case e1000_vfadapt_i350:
2342 		dev_info->max_rx_queues = 1;
2343 		dev_info->max_tx_queues = 1;
2344 		break;
2345 	default:
2346 		/* Should not happen */
2347 		break;
2348 	}
2349 
2350 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
2351 		.rx_thresh = {
2352 			.pthresh = IGB_DEFAULT_RX_PTHRESH,
2353 			.hthresh = IGB_DEFAULT_RX_HTHRESH,
2354 			.wthresh = IGB_DEFAULT_RX_WTHRESH,
2355 		},
2356 		.rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2357 		.rx_drop_en = 0,
2358 	};
2359 
2360 	dev_info->default_txconf = (struct rte_eth_txconf) {
2361 		.tx_thresh = {
2362 			.pthresh = IGB_DEFAULT_TX_PTHRESH,
2363 			.hthresh = IGB_DEFAULT_TX_HTHRESH,
2364 			.wthresh = IGB_DEFAULT_TX_WTHRESH,
2365 		},
2366 		.txq_flags = 0,
2367 	};
2368 
2369 	dev_info->rx_desc_lim = rx_desc_lim;
2370 	dev_info->tx_desc_lim = tx_desc_lim;
2371 }
2372 
2373 /* return 0 means link status changed, -1 means not changed */
2374 static int
2375 eth_igb_link_update(struct rte_eth_dev *dev, int wait_to_complete)
2376 {
2377 	struct e1000_hw *hw =
2378 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2379 	struct rte_eth_link link, old;
2380 	int link_check, count;
2381 
2382 	link_check = 0;
2383 	hw->mac.get_link_status = 1;
2384 
2385 	/* possible wait-to-complete in up to 9 seconds */
2386 	for (count = 0; count < IGB_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
2387 		/* Read the real link status */
2388 		switch (hw->phy.media_type) {
2389 		case e1000_media_type_copper:
2390 			/* Do the work to read phy */
2391 			e1000_check_for_link(hw);
2392 			link_check = !hw->mac.get_link_status;
2393 			break;
2394 
2395 		case e1000_media_type_fiber:
2396 			e1000_check_for_link(hw);
2397 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
2398 				      E1000_STATUS_LU);
2399 			break;
2400 
2401 		case e1000_media_type_internal_serdes:
2402 			e1000_check_for_link(hw);
2403 			link_check = hw->mac.serdes_has_link;
2404 			break;
2405 
2406 		/* VF device is type_unknown */
2407 		case e1000_media_type_unknown:
2408 			eth_igbvf_link_update(hw);
2409 			link_check = !hw->mac.get_link_status;
2410 			break;
2411 
2412 		default:
2413 			break;
2414 		}
2415 		if (link_check || wait_to_complete == 0)
2416 			break;
2417 		rte_delay_ms(IGB_LINK_UPDATE_CHECK_INTERVAL);
2418 	}
2419 	memset(&link, 0, sizeof(link));
2420 	rte_igb_dev_atomic_read_link_status(dev, &link);
2421 	old = link;
2422 
2423 	/* Now we check if a transition has happened */
2424 	if (link_check) {
2425 		uint16_t duplex, speed;
2426 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
2427 		link.link_duplex = (duplex == FULL_DUPLEX) ?
2428 				ETH_LINK_FULL_DUPLEX :
2429 				ETH_LINK_HALF_DUPLEX;
2430 		link.link_speed = speed;
2431 		link.link_status = ETH_LINK_UP;
2432 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
2433 				ETH_LINK_SPEED_FIXED);
2434 	} else if (!link_check) {
2435 		link.link_speed = 0;
2436 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
2437 		link.link_status = ETH_LINK_DOWN;
2438 		link.link_autoneg = ETH_LINK_FIXED;
2439 	}
2440 	rte_igb_dev_atomic_write_link_status(dev, &link);
2441 
2442 	/* not changed */
2443 	if (old.link_status == link.link_status)
2444 		return -1;
2445 
2446 	/* changed */
2447 	return 0;
2448 }
2449 
2450 /*
2451  * igb_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
2452  * For ASF and Pass Through versions of f/w this means
2453  * that the driver is loaded.
2454  */
2455 static void
2456 igb_hw_control_acquire(struct e1000_hw *hw)
2457 {
2458 	uint32_t ctrl_ext;
2459 
2460 	/* Let firmware know the driver has taken over */
2461 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2462 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2463 }
2464 
2465 /*
2466  * igb_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
2467  * For ASF and Pass Through versions of f/w this means that the
2468  * driver is no longer loaded.
2469  */
2470 static void
2471 igb_hw_control_release(struct e1000_hw *hw)
2472 {
2473 	uint32_t ctrl_ext;
2474 
2475 	/* Let firmware taken over control of h/w */
2476 	ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2477 	E1000_WRITE_REG(hw, E1000_CTRL_EXT,
2478 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2479 }
2480 
2481 /*
2482  * Bit of a misnomer, what this really means is
2483  * to enable OS management of the system... aka
2484  * to disable special hardware management features.
2485  */
2486 static void
2487 igb_init_manageability(struct e1000_hw *hw)
2488 {
2489 	if (e1000_enable_mng_pass_thru(hw)) {
2490 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
2491 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2492 
2493 		/* disable hardware interception of ARP */
2494 		manc &= ~(E1000_MANC_ARP_EN);
2495 
2496 		/* enable receiving management packets to the host */
2497 		manc |= E1000_MANC_EN_MNG2HOST;
2498 		manc2h |= 1 << 5;  /* Mng Port 623 */
2499 		manc2h |= 1 << 6;  /* Mng Port 664 */
2500 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
2501 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2502 	}
2503 }
2504 
2505 static void
2506 igb_release_manageability(struct e1000_hw *hw)
2507 {
2508 	if (e1000_enable_mng_pass_thru(hw)) {
2509 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2510 
2511 		manc |= E1000_MANC_ARP_EN;
2512 		manc &= ~E1000_MANC_EN_MNG2HOST;
2513 
2514 		E1000_WRITE_REG(hw, E1000_MANC, manc);
2515 	}
2516 }
2517 
2518 static void
2519 eth_igb_promiscuous_enable(struct rte_eth_dev *dev)
2520 {
2521 	struct e1000_hw *hw =
2522 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2523 	uint32_t rctl;
2524 
2525 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2526 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
2527 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2528 }
2529 
2530 static void
2531 eth_igb_promiscuous_disable(struct rte_eth_dev *dev)
2532 {
2533 	struct e1000_hw *hw =
2534 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2535 	uint32_t rctl;
2536 
2537 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2538 	rctl &= (~E1000_RCTL_UPE);
2539 	if (dev->data->all_multicast == 1)
2540 		rctl |= E1000_RCTL_MPE;
2541 	else
2542 		rctl &= (~E1000_RCTL_MPE);
2543 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2544 }
2545 
2546 static void
2547 eth_igb_allmulticast_enable(struct rte_eth_dev *dev)
2548 {
2549 	struct e1000_hw *hw =
2550 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2551 	uint32_t rctl;
2552 
2553 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2554 	rctl |= E1000_RCTL_MPE;
2555 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2556 }
2557 
2558 static void
2559 eth_igb_allmulticast_disable(struct rte_eth_dev *dev)
2560 {
2561 	struct e1000_hw *hw =
2562 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2563 	uint32_t rctl;
2564 
2565 	if (dev->data->promiscuous == 1)
2566 		return; /* must remain in all_multicast mode */
2567 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2568 	rctl &= (~E1000_RCTL_MPE);
2569 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2570 }
2571 
2572 static int
2573 eth_igb_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2574 {
2575 	struct e1000_hw *hw =
2576 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2577 	struct e1000_vfta * shadow_vfta =
2578 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2579 	uint32_t vfta;
2580 	uint32_t vid_idx;
2581 	uint32_t vid_bit;
2582 
2583 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
2584 			      E1000_VFTA_ENTRY_MASK);
2585 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
2586 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
2587 	if (on)
2588 		vfta |= vid_bit;
2589 	else
2590 		vfta &= ~vid_bit;
2591 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
2592 
2593 	/* update local VFTA copy */
2594 	shadow_vfta->vfta[vid_idx] = vfta;
2595 
2596 	return 0;
2597 }
2598 
2599 static int
2600 eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
2601 		      enum rte_vlan_type vlan_type,
2602 		      uint16_t tpid)
2603 {
2604 	struct e1000_hw *hw =
2605 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2606 	uint32_t reg, qinq;
2607 
2608 	qinq = E1000_READ_REG(hw, E1000_CTRL_EXT);
2609 	qinq &= E1000_CTRL_EXT_EXT_VLAN;
2610 
2611 	/* only outer TPID of double VLAN can be configured*/
2612 	if (qinq && vlan_type == ETH_VLAN_TYPE_OUTER) {
2613 		reg = E1000_READ_REG(hw, E1000_VET);
2614 		reg = (reg & (~E1000_VET_VET_EXT)) |
2615 			((uint32_t)tpid << E1000_VET_VET_EXT_SHIFT);
2616 		E1000_WRITE_REG(hw, E1000_VET, reg);
2617 
2618 		return 0;
2619 	}
2620 
2621 	/* all other TPID values are read-only*/
2622 	PMD_DRV_LOG(ERR, "Not supported");
2623 
2624 	return -ENOTSUP;
2625 }
2626 
2627 static void
2628 igb_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2629 {
2630 	struct e1000_hw *hw =
2631 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2632 	uint32_t reg;
2633 
2634 	/* Filter Table Disable */
2635 	reg = E1000_READ_REG(hw, E1000_RCTL);
2636 	reg &= ~E1000_RCTL_CFIEN;
2637 	reg &= ~E1000_RCTL_VFE;
2638 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2639 }
2640 
2641 static void
2642 igb_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2643 {
2644 	struct e1000_hw *hw =
2645 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2646 	struct e1000_vfta * shadow_vfta =
2647 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2648 	uint32_t reg;
2649 	int i;
2650 
2651 	/* Filter Table Enable, CFI not used for packet acceptance */
2652 	reg = E1000_READ_REG(hw, E1000_RCTL);
2653 	reg &= ~E1000_RCTL_CFIEN;
2654 	reg |= E1000_RCTL_VFE;
2655 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
2656 
2657 	/* restore VFTA table */
2658 	for (i = 0; i < IGB_VFTA_SIZE; i++)
2659 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
2660 }
2661 
2662 static void
2663 igb_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2664 {
2665 	struct e1000_hw *hw =
2666 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2667 	uint32_t reg;
2668 
2669 	/* VLAN Mode Disable */
2670 	reg = E1000_READ_REG(hw, E1000_CTRL);
2671 	reg &= ~E1000_CTRL_VME;
2672 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2673 }
2674 
2675 static void
2676 igb_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2677 {
2678 	struct e1000_hw *hw =
2679 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2680 	uint32_t reg;
2681 
2682 	/* VLAN Mode Enable */
2683 	reg = E1000_READ_REG(hw, E1000_CTRL);
2684 	reg |= E1000_CTRL_VME;
2685 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
2686 }
2687 
2688 static void
2689 igb_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2690 {
2691 	struct e1000_hw *hw =
2692 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2693 	uint32_t reg;
2694 
2695 	/* CTRL_EXT: Extended VLAN */
2696 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2697 	reg &= ~E1000_CTRL_EXT_EXTEND_VLAN;
2698 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2699 
2700 	/* Update maximum packet length */
2701 	if (dev->data->dev_conf.rxmode.jumbo_frame == 1)
2702 		E1000_WRITE_REG(hw, E1000_RLPML,
2703 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2704 						VLAN_TAG_SIZE);
2705 }
2706 
2707 static void
2708 igb_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2709 {
2710 	struct e1000_hw *hw =
2711 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2712 	uint32_t reg;
2713 
2714 	/* CTRL_EXT: Extended VLAN */
2715 	reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2716 	reg |= E1000_CTRL_EXT_EXTEND_VLAN;
2717 	E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2718 
2719 	/* Update maximum packet length */
2720 	if (dev->data->dev_conf.rxmode.jumbo_frame == 1)
2721 		E1000_WRITE_REG(hw, E1000_RLPML,
2722 			dev->data->dev_conf.rxmode.max_rx_pkt_len +
2723 						2 * VLAN_TAG_SIZE);
2724 }
2725 
2726 static int
2727 eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2728 {
2729 	if(mask & ETH_VLAN_STRIP_MASK){
2730 		if (dev->data->dev_conf.rxmode.hw_vlan_strip)
2731 			igb_vlan_hw_strip_enable(dev);
2732 		else
2733 			igb_vlan_hw_strip_disable(dev);
2734 	}
2735 
2736 	if(mask & ETH_VLAN_FILTER_MASK){
2737 		if (dev->data->dev_conf.rxmode.hw_vlan_filter)
2738 			igb_vlan_hw_filter_enable(dev);
2739 		else
2740 			igb_vlan_hw_filter_disable(dev);
2741 	}
2742 
2743 	if(mask & ETH_VLAN_EXTEND_MASK){
2744 		if (dev->data->dev_conf.rxmode.hw_vlan_extend)
2745 			igb_vlan_hw_extend_enable(dev);
2746 		else
2747 			igb_vlan_hw_extend_disable(dev);
2748 	}
2749 
2750 	return 0;
2751 }
2752 
2753 
2754 /**
2755  * It enables the interrupt mask and then enable the interrupt.
2756  *
2757  * @param dev
2758  *  Pointer to struct rte_eth_dev.
2759  * @param on
2760  *  Enable or Disable
2761  *
2762  * @return
2763  *  - On success, zero.
2764  *  - On failure, a negative value.
2765  */
2766 static int
2767 eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
2768 {
2769 	struct e1000_interrupt *intr =
2770 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2771 
2772 	if (on)
2773 		intr->mask |= E1000_ICR_LSC;
2774 	else
2775 		intr->mask &= ~E1000_ICR_LSC;
2776 
2777 	return 0;
2778 }
2779 
2780 /* It clears the interrupt causes and enables the interrupt.
2781  * It will be called once only during nic initialized.
2782  *
2783  * @param dev
2784  *  Pointer to struct rte_eth_dev.
2785  *
2786  * @return
2787  *  - On success, zero.
2788  *  - On failure, a negative value.
2789  */
2790 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev)
2791 {
2792 	uint32_t mask, regval;
2793 	struct e1000_hw *hw =
2794 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2795 	struct rte_eth_dev_info dev_info;
2796 
2797 	memset(&dev_info, 0, sizeof(dev_info));
2798 	eth_igb_infos_get(dev, &dev_info);
2799 
2800 	mask = 0xFFFFFFFF >> (32 - dev_info.max_rx_queues);
2801 	regval = E1000_READ_REG(hw, E1000_EIMS);
2802 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
2803 
2804 	return 0;
2805 }
2806 
2807 /*
2808  * It reads ICR and gets interrupt causes, check it and set a bit flag
2809  * to update link status.
2810  *
2811  * @param dev
2812  *  Pointer to struct rte_eth_dev.
2813  *
2814  * @return
2815  *  - On success, zero.
2816  *  - On failure, a negative value.
2817  */
2818 static int
2819 eth_igb_interrupt_get_status(struct rte_eth_dev *dev)
2820 {
2821 	uint32_t icr;
2822 	struct e1000_hw *hw =
2823 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2824 	struct e1000_interrupt *intr =
2825 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2826 
2827 	igb_intr_disable(hw);
2828 
2829 	/* read-on-clear nic registers here */
2830 	icr = E1000_READ_REG(hw, E1000_ICR);
2831 
2832 	intr->flags = 0;
2833 	if (icr & E1000_ICR_LSC) {
2834 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
2835 	}
2836 
2837 	if (icr & E1000_ICR_VMMB)
2838 		intr->flags |= E1000_FLAG_MAILBOX;
2839 
2840 	return 0;
2841 }
2842 
2843 /*
2844  * It executes link_update after knowing an interrupt is prsent.
2845  *
2846  * @param dev
2847  *  Pointer to struct rte_eth_dev.
2848  *
2849  * @return
2850  *  - On success, zero.
2851  *  - On failure, a negative value.
2852  */
2853 static int
2854 eth_igb_interrupt_action(struct rte_eth_dev *dev,
2855 			 struct rte_intr_handle *intr_handle)
2856 {
2857 	struct e1000_hw *hw =
2858 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2859 	struct e1000_interrupt *intr =
2860 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2861 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2862 	uint32_t tctl, rctl;
2863 	struct rte_eth_link link;
2864 	int ret;
2865 
2866 	if (intr->flags & E1000_FLAG_MAILBOX) {
2867 		igb_pf_mbx_process(dev);
2868 		intr->flags &= ~E1000_FLAG_MAILBOX;
2869 	}
2870 
2871 	igb_intr_enable(dev);
2872 	rte_intr_enable(intr_handle);
2873 
2874 	if (intr->flags & E1000_FLAG_NEED_LINK_UPDATE) {
2875 		intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
2876 
2877 		/* set get_link_status to check register later */
2878 		hw->mac.get_link_status = 1;
2879 		ret = eth_igb_link_update(dev, 0);
2880 
2881 		/* check if link has changed */
2882 		if (ret < 0)
2883 			return 0;
2884 
2885 		memset(&link, 0, sizeof(link));
2886 		rte_igb_dev_atomic_read_link_status(dev, &link);
2887 		if (link.link_status) {
2888 			PMD_INIT_LOG(INFO,
2889 				     " Port %d: Link Up - speed %u Mbps - %s",
2890 				     dev->data->port_id,
2891 				     (unsigned)link.link_speed,
2892 				     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
2893 				     "full-duplex" : "half-duplex");
2894 		} else {
2895 			PMD_INIT_LOG(INFO, " Port %d: Link Down",
2896 				     dev->data->port_id);
2897 		}
2898 
2899 		PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
2900 			     pci_dev->addr.domain,
2901 			     pci_dev->addr.bus,
2902 			     pci_dev->addr.devid,
2903 			     pci_dev->addr.function);
2904 		tctl = E1000_READ_REG(hw, E1000_TCTL);
2905 		rctl = E1000_READ_REG(hw, E1000_RCTL);
2906 		if (link.link_status) {
2907 			/* enable Tx/Rx */
2908 			tctl |= E1000_TCTL_EN;
2909 			rctl |= E1000_RCTL_EN;
2910 		} else {
2911 			/* disable Tx/Rx */
2912 			tctl &= ~E1000_TCTL_EN;
2913 			rctl &= ~E1000_RCTL_EN;
2914 		}
2915 		E1000_WRITE_REG(hw, E1000_TCTL, tctl);
2916 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2917 		E1000_WRITE_FLUSH(hw);
2918 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC,
2919 					      NULL, NULL);
2920 	}
2921 
2922 	return 0;
2923 }
2924 
2925 /**
2926  * Interrupt handler which shall be registered at first.
2927  *
2928  * @param handle
2929  *  Pointer to interrupt handle.
2930  * @param param
2931  *  The address of parameter (struct rte_eth_dev *) regsitered before.
2932  *
2933  * @return
2934  *  void
2935  */
2936 static void
2937 eth_igb_interrupt_handler(void *param)
2938 {
2939 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2940 
2941 	eth_igb_interrupt_get_status(dev);
2942 	eth_igb_interrupt_action(dev, dev->intr_handle);
2943 }
2944 
2945 static int
2946 eth_igbvf_interrupt_get_status(struct rte_eth_dev *dev)
2947 {
2948 	uint32_t eicr;
2949 	struct e1000_hw *hw =
2950 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2951 	struct e1000_interrupt *intr =
2952 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2953 
2954 	igbvf_intr_disable(hw);
2955 
2956 	/* read-on-clear nic registers here */
2957 	eicr = E1000_READ_REG(hw, E1000_EICR);
2958 	intr->flags = 0;
2959 
2960 	if (eicr == E1000_VTIVAR_MISC_MAILBOX)
2961 		intr->flags |= E1000_FLAG_MAILBOX;
2962 
2963 	return 0;
2964 }
2965 
2966 void igbvf_mbx_process(struct rte_eth_dev *dev)
2967 {
2968 	struct e1000_hw *hw =
2969 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2970 	struct e1000_mbx_info *mbx = &hw->mbx;
2971 	u32 in_msg = 0;
2972 
2973 	if (mbx->ops.read(hw, &in_msg, 1, 0))
2974 		return;
2975 
2976 	/* PF reset VF event */
2977 	if (in_msg == E1000_PF_CONTROL_MSG)
2978 		_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
2979 					      NULL, NULL);
2980 }
2981 
2982 static int
2983 eth_igbvf_interrupt_action(struct rte_eth_dev *dev, struct rte_intr_handle *intr_handle)
2984 {
2985 	struct e1000_interrupt *intr =
2986 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2987 
2988 	if (intr->flags & E1000_FLAG_MAILBOX) {
2989 		igbvf_mbx_process(dev);
2990 		intr->flags &= ~E1000_FLAG_MAILBOX;
2991 	}
2992 
2993 	igbvf_intr_enable(dev);
2994 	rte_intr_enable(intr_handle);
2995 
2996 	return 0;
2997 }
2998 
2999 static void
3000 eth_igbvf_interrupt_handler(void *param)
3001 {
3002 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
3003 
3004 	eth_igbvf_interrupt_get_status(dev);
3005 	eth_igbvf_interrupt_action(dev, dev->intr_handle);
3006 }
3007 
3008 static int
3009 eth_igb_led_on(struct rte_eth_dev *dev)
3010 {
3011 	struct e1000_hw *hw;
3012 
3013 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3014 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3015 }
3016 
3017 static int
3018 eth_igb_led_off(struct rte_eth_dev *dev)
3019 {
3020 	struct e1000_hw *hw;
3021 
3022 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3023 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3024 }
3025 
3026 static int
3027 eth_igb_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3028 {
3029 	struct e1000_hw *hw;
3030 	uint32_t ctrl;
3031 	int tx_pause;
3032 	int rx_pause;
3033 
3034 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3035 	fc_conf->pause_time = hw->fc.pause_time;
3036 	fc_conf->high_water = hw->fc.high_water;
3037 	fc_conf->low_water = hw->fc.low_water;
3038 	fc_conf->send_xon = hw->fc.send_xon;
3039 	fc_conf->autoneg = hw->mac.autoneg;
3040 
3041 	/*
3042 	 * Return rx_pause and tx_pause status according to actual setting of
3043 	 * the TFCE and RFCE bits in the CTRL register.
3044 	 */
3045 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
3046 	if (ctrl & E1000_CTRL_TFCE)
3047 		tx_pause = 1;
3048 	else
3049 		tx_pause = 0;
3050 
3051 	if (ctrl & E1000_CTRL_RFCE)
3052 		rx_pause = 1;
3053 	else
3054 		rx_pause = 0;
3055 
3056 	if (rx_pause && tx_pause)
3057 		fc_conf->mode = RTE_FC_FULL;
3058 	else if (rx_pause)
3059 		fc_conf->mode = RTE_FC_RX_PAUSE;
3060 	else if (tx_pause)
3061 		fc_conf->mode = RTE_FC_TX_PAUSE;
3062 	else
3063 		fc_conf->mode = RTE_FC_NONE;
3064 
3065 	return 0;
3066 }
3067 
3068 static int
3069 eth_igb_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3070 {
3071 	struct e1000_hw *hw;
3072 	int err;
3073 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
3074 		e1000_fc_none,
3075 		e1000_fc_rx_pause,
3076 		e1000_fc_tx_pause,
3077 		e1000_fc_full
3078 	};
3079 	uint32_t rx_buf_size;
3080 	uint32_t max_high_water;
3081 	uint32_t rctl;
3082 
3083 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3084 	if (fc_conf->autoneg != hw->mac.autoneg)
3085 		return -ENOTSUP;
3086 	rx_buf_size = igb_get_rx_buffer_size(hw);
3087 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3088 
3089 	/* At least reserve one Ethernet frame for watermark */
3090 	max_high_water = rx_buf_size - ETHER_MAX_LEN;
3091 	if ((fc_conf->high_water > max_high_water) ||
3092 	    (fc_conf->high_water < fc_conf->low_water)) {
3093 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
3094 		PMD_INIT_LOG(ERR, "high water must <=  0x%x", max_high_water);
3095 		return -EINVAL;
3096 	}
3097 
3098 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
3099 	hw->fc.pause_time     = fc_conf->pause_time;
3100 	hw->fc.high_water     = fc_conf->high_water;
3101 	hw->fc.low_water      = fc_conf->low_water;
3102 	hw->fc.send_xon	      = fc_conf->send_xon;
3103 
3104 	err = e1000_setup_link_generic(hw);
3105 	if (err == E1000_SUCCESS) {
3106 
3107 		/* check if we want to forward MAC frames - driver doesn't have native
3108 		 * capability to do that, so we'll write the registers ourselves */
3109 
3110 		rctl = E1000_READ_REG(hw, E1000_RCTL);
3111 
3112 		/* set or clear MFLCN.PMCF bit depending on configuration */
3113 		if (fc_conf->mac_ctrl_frame_fwd != 0)
3114 			rctl |= E1000_RCTL_PMCF;
3115 		else
3116 			rctl &= ~E1000_RCTL_PMCF;
3117 
3118 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3119 		E1000_WRITE_FLUSH(hw);
3120 
3121 		return 0;
3122 	}
3123 
3124 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
3125 	return -EIO;
3126 }
3127 
3128 #define E1000_RAH_POOLSEL_SHIFT      (18)
3129 static int
3130 eth_igb_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
3131 		uint32_t index, uint32_t pool)
3132 {
3133 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3134 	uint32_t rah;
3135 
3136 	e1000_rar_set(hw, mac_addr->addr_bytes, index);
3137 	rah = E1000_READ_REG(hw, E1000_RAH(index));
3138 	rah |= (0x1 << (E1000_RAH_POOLSEL_SHIFT + pool));
3139 	E1000_WRITE_REG(hw, E1000_RAH(index), rah);
3140 	return 0;
3141 }
3142 
3143 static void
3144 eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index)
3145 {
3146 	uint8_t addr[ETHER_ADDR_LEN];
3147 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3148 
3149 	memset(addr, 0, sizeof(addr));
3150 
3151 	e1000_rar_set(hw, addr, index);
3152 }
3153 
3154 static void
3155 eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
3156 				struct ether_addr *addr)
3157 {
3158 	eth_igb_rar_clear(dev, 0);
3159 
3160 	eth_igb_rar_set(dev, (void *)addr, 0, 0);
3161 }
3162 /*
3163  * Virtual Function operations
3164  */
3165 static void
3166 igbvf_intr_disable(struct e1000_hw *hw)
3167 {
3168 	PMD_INIT_FUNC_TRACE();
3169 
3170 	/* Clear interrupt mask to stop from interrupts being generated */
3171 	E1000_WRITE_REG(hw, E1000_EIMC, 0xFFFF);
3172 
3173 	E1000_WRITE_FLUSH(hw);
3174 }
3175 
3176 static void
3177 igbvf_stop_adapter(struct rte_eth_dev *dev)
3178 {
3179 	u32 reg_val;
3180 	u16 i;
3181 	struct rte_eth_dev_info dev_info;
3182 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3183 
3184 	memset(&dev_info, 0, sizeof(dev_info));
3185 	eth_igbvf_infos_get(dev, &dev_info);
3186 
3187 	/* Clear interrupt mask to stop from interrupts being generated */
3188 	igbvf_intr_disable(hw);
3189 
3190 	/* Clear any pending interrupts, flush previous writes */
3191 	E1000_READ_REG(hw, E1000_EICR);
3192 
3193 	/* Disable the transmit unit.  Each queue must be disabled. */
3194 	for (i = 0; i < dev_info.max_tx_queues; i++)
3195 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), E1000_TXDCTL_SWFLSH);
3196 
3197 	/* Disable the receive unit by stopping each queue */
3198 	for (i = 0; i < dev_info.max_rx_queues; i++) {
3199 		reg_val = E1000_READ_REG(hw, E1000_RXDCTL(i));
3200 		reg_val &= ~E1000_RXDCTL_QUEUE_ENABLE;
3201 		E1000_WRITE_REG(hw, E1000_RXDCTL(i), reg_val);
3202 		while (E1000_READ_REG(hw, E1000_RXDCTL(i)) & E1000_RXDCTL_QUEUE_ENABLE)
3203 			;
3204 	}
3205 
3206 	/* flush all queues disables */
3207 	E1000_WRITE_FLUSH(hw);
3208 	msec_delay(2);
3209 }
3210 
3211 static int eth_igbvf_link_update(struct e1000_hw *hw)
3212 {
3213 	struct e1000_mbx_info *mbx = &hw->mbx;
3214 	struct e1000_mac_info *mac = &hw->mac;
3215 	int ret_val = E1000_SUCCESS;
3216 
3217 	PMD_INIT_LOG(DEBUG, "e1000_check_for_link_vf");
3218 
3219 	/*
3220 	 * We only want to run this if there has been a rst asserted.
3221 	 * in this case that could mean a link change, device reset,
3222 	 * or a virtual function reset
3223 	 */
3224 
3225 	/* If we were hit with a reset or timeout drop the link */
3226 	if (!e1000_check_for_rst(hw, 0) || !mbx->timeout)
3227 		mac->get_link_status = TRUE;
3228 
3229 	if (!mac->get_link_status)
3230 		goto out;
3231 
3232 	/* if link status is down no point in checking to see if pf is up */
3233 	if (!(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU))
3234 		goto out;
3235 
3236 	/* if we passed all the tests above then the link is up and we no
3237 	 * longer need to check for link */
3238 	mac->get_link_status = FALSE;
3239 
3240 out:
3241 	return ret_val;
3242 }
3243 
3244 
3245 static int
3246 igbvf_dev_configure(struct rte_eth_dev *dev)
3247 {
3248 	struct rte_eth_conf* conf = &dev->data->dev_conf;
3249 
3250 	PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
3251 		     dev->data->port_id);
3252 
3253 	/*
3254 	 * VF has no ability to enable/disable HW CRC
3255 	 * Keep the persistent behavior the same as Host PF
3256 	 */
3257 #ifndef RTE_LIBRTE_E1000_PF_DISABLE_STRIP_CRC
3258 	if (!conf->rxmode.hw_strip_crc) {
3259 		PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
3260 		conf->rxmode.hw_strip_crc = 1;
3261 	}
3262 #else
3263 	if (conf->rxmode.hw_strip_crc) {
3264 		PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
3265 		conf->rxmode.hw_strip_crc = 0;
3266 	}
3267 #endif
3268 
3269 	return 0;
3270 }
3271 
3272 static int
3273 igbvf_dev_start(struct rte_eth_dev *dev)
3274 {
3275 	struct e1000_hw *hw =
3276 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3277 	struct e1000_adapter *adapter =
3278 		E1000_DEV_PRIVATE(dev->data->dev_private);
3279 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3280 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3281 	int ret;
3282 	uint32_t intr_vector = 0;
3283 
3284 	PMD_INIT_FUNC_TRACE();
3285 
3286 	hw->mac.ops.reset_hw(hw);
3287 	adapter->stopped = 0;
3288 
3289 	/* Set all vfta */
3290 	igbvf_set_vfta_all(dev,1);
3291 
3292 	eth_igbvf_tx_init(dev);
3293 
3294 	/* This can fail when allocating mbufs for descriptor rings */
3295 	ret = eth_igbvf_rx_init(dev);
3296 	if (ret) {
3297 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
3298 		igb_dev_clear_queues(dev);
3299 		return ret;
3300 	}
3301 
3302 	/* check and configure queue intr-vector mapping */
3303 	if (rte_intr_cap_multiple(intr_handle) &&
3304 	    dev->data->dev_conf.intr_conf.rxq) {
3305 		intr_vector = dev->data->nb_rx_queues;
3306 		ret = rte_intr_efd_enable(intr_handle, intr_vector);
3307 		if (ret)
3308 			return ret;
3309 	}
3310 
3311 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
3312 		intr_handle->intr_vec =
3313 			rte_zmalloc("intr_vec",
3314 				    dev->data->nb_rx_queues * sizeof(int), 0);
3315 		if (!intr_handle->intr_vec) {
3316 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
3317 				     " intr_vec", dev->data->nb_rx_queues);
3318 			return -ENOMEM;
3319 		}
3320 	}
3321 
3322 	eth_igbvf_configure_msix_intr(dev);
3323 
3324 	/* enable uio/vfio intr/eventfd mapping */
3325 	rte_intr_enable(intr_handle);
3326 
3327 	/* resume enabled intr since hw reset */
3328 	igbvf_intr_enable(dev);
3329 
3330 	return 0;
3331 }
3332 
3333 static void
3334 igbvf_dev_stop(struct rte_eth_dev *dev)
3335 {
3336 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3337 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3338 
3339 	PMD_INIT_FUNC_TRACE();
3340 
3341 	igbvf_stop_adapter(dev);
3342 
3343 	/*
3344 	  * Clear what we set, but we still keep shadow_vfta to
3345 	  * restore after device starts
3346 	  */
3347 	igbvf_set_vfta_all(dev,0);
3348 
3349 	igb_dev_clear_queues(dev);
3350 
3351 	/* disable intr eventfd mapping */
3352 	rte_intr_disable(intr_handle);
3353 
3354 	/* Clean datapath event and queue/vec mapping */
3355 	rte_intr_efd_disable(intr_handle);
3356 	if (intr_handle->intr_vec) {
3357 		rte_free(intr_handle->intr_vec);
3358 		intr_handle->intr_vec = NULL;
3359 	}
3360 }
3361 
3362 static void
3363 igbvf_dev_close(struct rte_eth_dev *dev)
3364 {
3365 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3366 	struct e1000_adapter *adapter =
3367 		E1000_DEV_PRIVATE(dev->data->dev_private);
3368 	struct ether_addr addr;
3369 
3370 	PMD_INIT_FUNC_TRACE();
3371 
3372 	e1000_reset_hw(hw);
3373 
3374 	igbvf_dev_stop(dev);
3375 	adapter->stopped = 1;
3376 	igb_dev_free_queues(dev);
3377 
3378 	/**
3379 	 * reprogram the RAR with a zero mac address,
3380 	 * to ensure that the VF traffic goes to the PF
3381 	 * after stop, close and detach of the VF.
3382 	 **/
3383 
3384 	memset(&addr, 0, sizeof(addr));
3385 	igbvf_default_mac_addr_set(dev, &addr);
3386 }
3387 
3388 static void
3389 igbvf_promiscuous_enable(struct rte_eth_dev *dev)
3390 {
3391 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3392 
3393 	/* Set both unicast and multicast promisc */
3394 	e1000_promisc_set_vf(hw, e1000_promisc_enabled);
3395 }
3396 
3397 static void
3398 igbvf_promiscuous_disable(struct rte_eth_dev *dev)
3399 {
3400 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3401 
3402 	/* If in allmulticast mode leave multicast promisc */
3403 	if (dev->data->all_multicast == 1)
3404 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3405 	else
3406 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3407 }
3408 
3409 static void
3410 igbvf_allmulticast_enable(struct rte_eth_dev *dev)
3411 {
3412 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3413 
3414 	/* In promiscuous mode multicast promisc already set */
3415 	if (dev->data->promiscuous == 0)
3416 		e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3417 }
3418 
3419 static void
3420 igbvf_allmulticast_disable(struct rte_eth_dev *dev)
3421 {
3422 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3423 
3424 	/* In promiscuous mode leave multicast promisc enabled */
3425 	if (dev->data->promiscuous == 0)
3426 		e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3427 }
3428 
3429 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on)
3430 {
3431 	struct e1000_mbx_info *mbx = &hw->mbx;
3432 	uint32_t msgbuf[2];
3433 	s32 err;
3434 
3435 	/* After set vlan, vlan strip will also be enabled in igb driver*/
3436 	msgbuf[0] = E1000_VF_SET_VLAN;
3437 	msgbuf[1] = vid;
3438 	/* Setting the 8 bit field MSG INFO to TRUE indicates "add" */
3439 	if (on)
3440 		msgbuf[0] |= E1000_VF_SET_VLAN_ADD;
3441 
3442 	err = mbx->ops.write_posted(hw, msgbuf, 2, 0);
3443 	if (err)
3444 		goto mbx_err;
3445 
3446 	err = mbx->ops.read_posted(hw, msgbuf, 2, 0);
3447 	if (err)
3448 		goto mbx_err;
3449 
3450 	msgbuf[0] &= ~E1000_VT_MSGTYPE_CTS;
3451 	if (msgbuf[0] == (E1000_VF_SET_VLAN | E1000_VT_MSGTYPE_NACK))
3452 		err = -EINVAL;
3453 
3454 mbx_err:
3455 	return err;
3456 }
3457 
3458 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on)
3459 {
3460 	struct e1000_hw *hw =
3461 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3462 	struct e1000_vfta * shadow_vfta =
3463 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3464 	int i = 0, j = 0, vfta = 0, mask = 1;
3465 
3466 	for (i = 0; i < IGB_VFTA_SIZE; i++){
3467 		vfta = shadow_vfta->vfta[i];
3468 		if(vfta){
3469 			mask = 1;
3470 			for (j = 0; j < 32; j++){
3471 				if(vfta & mask)
3472 					igbvf_set_vfta(hw,
3473 						(uint16_t)((i<<5)+j), on);
3474 				mask<<=1;
3475 			}
3476 		}
3477 	}
3478 
3479 }
3480 
3481 static int
3482 igbvf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
3483 {
3484 	struct e1000_hw *hw =
3485 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3486 	struct e1000_vfta * shadow_vfta =
3487 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3488 	uint32_t vid_idx = 0;
3489 	uint32_t vid_bit = 0;
3490 	int ret = 0;
3491 
3492 	PMD_INIT_FUNC_TRACE();
3493 
3494 	/*vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf*/
3495 	ret = igbvf_set_vfta(hw, vlan_id, !!on);
3496 	if(ret){
3497 		PMD_INIT_LOG(ERR, "Unable to set VF vlan");
3498 		return ret;
3499 	}
3500 	vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
3501 	vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
3502 
3503 	/*Save what we set and retore it after device reset*/
3504 	if (on)
3505 		shadow_vfta->vfta[vid_idx] |= vid_bit;
3506 	else
3507 		shadow_vfta->vfta[vid_idx] &= ~vid_bit;
3508 
3509 	return 0;
3510 }
3511 
3512 static void
3513 igbvf_default_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *addr)
3514 {
3515 	struct e1000_hw *hw =
3516 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3517 
3518 	/* index is not used by rar_set() */
3519 	hw->mac.ops.rar_set(hw, (void *)addr, 0);
3520 }
3521 
3522 
3523 static int
3524 eth_igb_rss_reta_update(struct rte_eth_dev *dev,
3525 			struct rte_eth_rss_reta_entry64 *reta_conf,
3526 			uint16_t reta_size)
3527 {
3528 	uint8_t i, j, mask;
3529 	uint32_t reta, r;
3530 	uint16_t idx, shift;
3531 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3532 
3533 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3534 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3535 			"(%d) doesn't match the number hardware can supported "
3536 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3537 		return -EINVAL;
3538 	}
3539 
3540 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3541 		idx = i / RTE_RETA_GROUP_SIZE;
3542 		shift = i % RTE_RETA_GROUP_SIZE;
3543 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3544 						IGB_4_BIT_MASK);
3545 		if (!mask)
3546 			continue;
3547 		if (mask == IGB_4_BIT_MASK)
3548 			r = 0;
3549 		else
3550 			r = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3551 		for (j = 0, reta = 0; j < IGB_4_BIT_WIDTH; j++) {
3552 			if (mask & (0x1 << j))
3553 				reta |= reta_conf[idx].reta[shift + j] <<
3554 							(CHAR_BIT * j);
3555 			else
3556 				reta |= r & (IGB_8_BIT_MASK << (CHAR_BIT * j));
3557 		}
3558 		E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
3559 	}
3560 
3561 	return 0;
3562 }
3563 
3564 static int
3565 eth_igb_rss_reta_query(struct rte_eth_dev *dev,
3566 		       struct rte_eth_rss_reta_entry64 *reta_conf,
3567 		       uint16_t reta_size)
3568 {
3569 	uint8_t i, j, mask;
3570 	uint32_t reta;
3571 	uint16_t idx, shift;
3572 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3573 
3574 	if (reta_size != ETH_RSS_RETA_SIZE_128) {
3575 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3576 			"(%d) doesn't match the number hardware can supported "
3577 			"(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3578 		return -EINVAL;
3579 	}
3580 
3581 	for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3582 		idx = i / RTE_RETA_GROUP_SIZE;
3583 		shift = i % RTE_RETA_GROUP_SIZE;
3584 		mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3585 						IGB_4_BIT_MASK);
3586 		if (!mask)
3587 			continue;
3588 		reta = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3589 		for (j = 0; j < IGB_4_BIT_WIDTH; j++) {
3590 			if (mask & (0x1 << j))
3591 				reta_conf[idx].reta[shift + j] =
3592 					((reta >> (CHAR_BIT * j)) &
3593 						IGB_8_BIT_MASK);
3594 		}
3595 	}
3596 
3597 	return 0;
3598 }
3599 
3600 int
3601 eth_igb_syn_filter_set(struct rte_eth_dev *dev,
3602 			struct rte_eth_syn_filter *filter,
3603 			bool add)
3604 {
3605 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3606 	struct e1000_filter_info *filter_info =
3607 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3608 	uint32_t synqf, rfctl;
3609 
3610 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3611 		return -EINVAL;
3612 
3613 	synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3614 
3615 	if (add) {
3616 		if (synqf & E1000_SYN_FILTER_ENABLE)
3617 			return -EINVAL;
3618 
3619 		synqf = (uint32_t)(((filter->queue << E1000_SYN_FILTER_QUEUE_SHIFT) &
3620 			E1000_SYN_FILTER_QUEUE) | E1000_SYN_FILTER_ENABLE);
3621 
3622 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3623 		if (filter->hig_pri)
3624 			rfctl |= E1000_RFCTL_SYNQFP;
3625 		else
3626 			rfctl &= ~E1000_RFCTL_SYNQFP;
3627 
3628 		E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3629 	} else {
3630 		if (!(synqf & E1000_SYN_FILTER_ENABLE))
3631 			return -ENOENT;
3632 		synqf = 0;
3633 	}
3634 
3635 	filter_info->syn_info = synqf;
3636 	E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
3637 	E1000_WRITE_FLUSH(hw);
3638 	return 0;
3639 }
3640 
3641 static int
3642 eth_igb_syn_filter_get(struct rte_eth_dev *dev,
3643 			struct rte_eth_syn_filter *filter)
3644 {
3645 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3646 	uint32_t synqf, rfctl;
3647 
3648 	synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3649 	if (synqf & E1000_SYN_FILTER_ENABLE) {
3650 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3651 		filter->hig_pri = (rfctl & E1000_RFCTL_SYNQFP) ? 1 : 0;
3652 		filter->queue = (uint8_t)((synqf & E1000_SYN_FILTER_QUEUE) >>
3653 				E1000_SYN_FILTER_QUEUE_SHIFT);
3654 		return 0;
3655 	}
3656 
3657 	return -ENOENT;
3658 }
3659 
3660 static int
3661 eth_igb_syn_filter_handle(struct rte_eth_dev *dev,
3662 			enum rte_filter_op filter_op,
3663 			void *arg)
3664 {
3665 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3666 	int ret;
3667 
3668 	MAC_TYPE_FILTER_SUP(hw->mac.type);
3669 
3670 	if (filter_op == RTE_ETH_FILTER_NOP)
3671 		return 0;
3672 
3673 	if (arg == NULL) {
3674 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
3675 			    filter_op);
3676 		return -EINVAL;
3677 	}
3678 
3679 	switch (filter_op) {
3680 	case RTE_ETH_FILTER_ADD:
3681 		ret = eth_igb_syn_filter_set(dev,
3682 				(struct rte_eth_syn_filter *)arg,
3683 				TRUE);
3684 		break;
3685 	case RTE_ETH_FILTER_DELETE:
3686 		ret = eth_igb_syn_filter_set(dev,
3687 				(struct rte_eth_syn_filter *)arg,
3688 				FALSE);
3689 		break;
3690 	case RTE_ETH_FILTER_GET:
3691 		ret = eth_igb_syn_filter_get(dev,
3692 				(struct rte_eth_syn_filter *)arg);
3693 		break;
3694 	default:
3695 		PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
3696 		ret = -EINVAL;
3697 		break;
3698 	}
3699 
3700 	return ret;
3701 }
3702 
3703 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_2tuple_filter_info*/
3704 static inline int
3705 ntuple_filter_to_2tuple(struct rte_eth_ntuple_filter *filter,
3706 			struct e1000_2tuple_filter_info *filter_info)
3707 {
3708 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3709 		return -EINVAL;
3710 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
3711 		return -EINVAL;  /* filter index is out of range. */
3712 	if (filter->tcp_flags > TCP_FLAG_ALL)
3713 		return -EINVAL;  /* flags is invalid. */
3714 
3715 	switch (filter->dst_port_mask) {
3716 	case UINT16_MAX:
3717 		filter_info->dst_port_mask = 0;
3718 		filter_info->dst_port = filter->dst_port;
3719 		break;
3720 	case 0:
3721 		filter_info->dst_port_mask = 1;
3722 		break;
3723 	default:
3724 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
3725 		return -EINVAL;
3726 	}
3727 
3728 	switch (filter->proto_mask) {
3729 	case UINT8_MAX:
3730 		filter_info->proto_mask = 0;
3731 		filter_info->proto = filter->proto;
3732 		break;
3733 	case 0:
3734 		filter_info->proto_mask = 1;
3735 		break;
3736 	default:
3737 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
3738 		return -EINVAL;
3739 	}
3740 
3741 	filter_info->priority = (uint8_t)filter->priority;
3742 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
3743 		filter_info->tcp_flags = filter->tcp_flags;
3744 	else
3745 		filter_info->tcp_flags = 0;
3746 
3747 	return 0;
3748 }
3749 
3750 static inline struct e1000_2tuple_filter *
3751 igb_2tuple_filter_lookup(struct e1000_2tuple_filter_list *filter_list,
3752 			struct e1000_2tuple_filter_info *key)
3753 {
3754 	struct e1000_2tuple_filter *it;
3755 
3756 	TAILQ_FOREACH(it, filter_list, entries) {
3757 		if (memcmp(key, &it->filter_info,
3758 			sizeof(struct e1000_2tuple_filter_info)) == 0) {
3759 			return it;
3760 		}
3761 	}
3762 	return NULL;
3763 }
3764 
3765 /* inject a igb 2tuple filter to HW */
3766 static inline void
3767 igb_inject_2uple_filter(struct rte_eth_dev *dev,
3768 			   struct e1000_2tuple_filter *filter)
3769 {
3770 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3771 	uint32_t ttqf = E1000_TTQF_DISABLE_MASK;
3772 	uint32_t imir, imir_ext = E1000_IMIREXT_SIZE_BP;
3773 	int i;
3774 
3775 	i = filter->index;
3776 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
3777 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
3778 		imir |= E1000_IMIR_PORT_BP;
3779 	else
3780 		imir &= ~E1000_IMIR_PORT_BP;
3781 
3782 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
3783 
3784 	ttqf |= E1000_TTQF_QUEUE_ENABLE;
3785 	ttqf |= (uint32_t)(filter->queue << E1000_TTQF_QUEUE_SHIFT);
3786 	ttqf |= (uint32_t)(filter->filter_info.proto &
3787 						E1000_TTQF_PROTOCOL_MASK);
3788 	if (filter->filter_info.proto_mask == 0)
3789 		ttqf &= ~E1000_TTQF_MASK_ENABLE;
3790 
3791 	/* tcp flags bits setting. */
3792 	if (filter->filter_info.tcp_flags & TCP_FLAG_ALL) {
3793 		if (filter->filter_info.tcp_flags & TCP_URG_FLAG)
3794 			imir_ext |= E1000_IMIREXT_CTRL_URG;
3795 		if (filter->filter_info.tcp_flags & TCP_ACK_FLAG)
3796 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
3797 		if (filter->filter_info.tcp_flags & TCP_PSH_FLAG)
3798 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
3799 		if (filter->filter_info.tcp_flags & TCP_RST_FLAG)
3800 			imir_ext |= E1000_IMIREXT_CTRL_RST;
3801 		if (filter->filter_info.tcp_flags & TCP_SYN_FLAG)
3802 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
3803 		if (filter->filter_info.tcp_flags & TCP_FIN_FLAG)
3804 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
3805 	} else {
3806 		imir_ext |= E1000_IMIREXT_CTRL_BP;
3807 	}
3808 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
3809 	E1000_WRITE_REG(hw, E1000_TTQF(i), ttqf);
3810 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
3811 }
3812 
3813 /*
3814  * igb_add_2tuple_filter - add a 2tuple filter
3815  *
3816  * @param
3817  * dev: Pointer to struct rte_eth_dev.
3818  * ntuple_filter: ponter to the filter that will be added.
3819  *
3820  * @return
3821  *    - On success, zero.
3822  *    - On failure, a negative value.
3823  */
3824 static int
3825 igb_add_2tuple_filter(struct rte_eth_dev *dev,
3826 			struct rte_eth_ntuple_filter *ntuple_filter)
3827 {
3828 	struct e1000_filter_info *filter_info =
3829 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3830 	struct e1000_2tuple_filter *filter;
3831 	int i, ret;
3832 
3833 	filter = rte_zmalloc("e1000_2tuple_filter",
3834 			sizeof(struct e1000_2tuple_filter), 0);
3835 	if (filter == NULL)
3836 		return -ENOMEM;
3837 
3838 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3839 				      &filter->filter_info);
3840 	if (ret < 0) {
3841 		rte_free(filter);
3842 		return ret;
3843 	}
3844 	if (igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3845 					 &filter->filter_info) != NULL) {
3846 		PMD_DRV_LOG(ERR, "filter exists.");
3847 		rte_free(filter);
3848 		return -EEXIST;
3849 	}
3850 	filter->queue = ntuple_filter->queue;
3851 
3852 	/*
3853 	 * look for an unused 2tuple filter index,
3854 	 * and insert the filter to list.
3855 	 */
3856 	for (i = 0; i < E1000_MAX_TTQF_FILTERS; i++) {
3857 		if (!(filter_info->twotuple_mask & (1 << i))) {
3858 			filter_info->twotuple_mask |= 1 << i;
3859 			filter->index = i;
3860 			TAILQ_INSERT_TAIL(&filter_info->twotuple_list,
3861 					  filter,
3862 					  entries);
3863 			break;
3864 		}
3865 	}
3866 	if (i >= E1000_MAX_TTQF_FILTERS) {
3867 		PMD_DRV_LOG(ERR, "2tuple filters are full.");
3868 		rte_free(filter);
3869 		return -ENOSYS;
3870 	}
3871 
3872 	igb_inject_2uple_filter(dev, filter);
3873 	return 0;
3874 }
3875 
3876 int
3877 igb_delete_2tuple_filter(struct rte_eth_dev *dev,
3878 			struct e1000_2tuple_filter *filter)
3879 {
3880 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3881 	struct e1000_filter_info *filter_info =
3882 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3883 
3884 	filter_info->twotuple_mask &= ~(1 << filter->index);
3885 	TAILQ_REMOVE(&filter_info->twotuple_list, filter, entries);
3886 	rte_free(filter);
3887 
3888 	E1000_WRITE_REG(hw, E1000_TTQF(filter->index), E1000_TTQF_DISABLE_MASK);
3889 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
3890 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
3891 	return 0;
3892 }
3893 
3894 /*
3895  * igb_remove_2tuple_filter - remove a 2tuple filter
3896  *
3897  * @param
3898  * dev: Pointer to struct rte_eth_dev.
3899  * ntuple_filter: ponter to the filter that will be removed.
3900  *
3901  * @return
3902  *    - On success, zero.
3903  *    - On failure, a negative value.
3904  */
3905 static int
3906 igb_remove_2tuple_filter(struct rte_eth_dev *dev,
3907 			struct rte_eth_ntuple_filter *ntuple_filter)
3908 {
3909 	struct e1000_filter_info *filter_info =
3910 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3911 	struct e1000_2tuple_filter_info filter_2tuple;
3912 	struct e1000_2tuple_filter *filter;
3913 	int ret;
3914 
3915 	memset(&filter_2tuple, 0, sizeof(struct e1000_2tuple_filter_info));
3916 	ret = ntuple_filter_to_2tuple(ntuple_filter,
3917 				      &filter_2tuple);
3918 	if (ret < 0)
3919 		return ret;
3920 
3921 	filter = igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3922 					 &filter_2tuple);
3923 	if (filter == NULL) {
3924 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
3925 		return -ENOENT;
3926 	}
3927 
3928 	igb_delete_2tuple_filter(dev, filter);
3929 
3930 	return 0;
3931 }
3932 
3933 /* inject a igb flex filter to HW */
3934 static inline void
3935 igb_inject_flex_filter(struct rte_eth_dev *dev,
3936 			   struct e1000_flex_filter *filter)
3937 {
3938 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3939 	uint32_t wufc, queueing;
3940 	uint32_t reg_off;
3941 	uint8_t i, j = 0;
3942 
3943 	wufc = E1000_READ_REG(hw, E1000_WUFC);
3944 	if (filter->index < E1000_MAX_FHFT)
3945 		reg_off = E1000_FHFT(filter->index);
3946 	else
3947 		reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
3948 
3949 	E1000_WRITE_REG(hw, E1000_WUFC, wufc | E1000_WUFC_FLEX_HQ |
3950 			(E1000_WUFC_FLX0 << filter->index));
3951 	queueing = filter->filter_info.len |
3952 		(filter->queue << E1000_FHFT_QUEUEING_QUEUE_SHIFT) |
3953 		(filter->filter_info.priority <<
3954 			E1000_FHFT_QUEUEING_PRIO_SHIFT);
3955 	E1000_WRITE_REG(hw, reg_off + E1000_FHFT_QUEUEING_OFFSET,
3956 			queueing);
3957 
3958 	for (i = 0; i < E1000_FLEX_FILTERS_MASK_SIZE; i++) {
3959 		E1000_WRITE_REG(hw, reg_off,
3960 				filter->filter_info.dwords[j]);
3961 		reg_off += sizeof(uint32_t);
3962 		E1000_WRITE_REG(hw, reg_off,
3963 				filter->filter_info.dwords[++j]);
3964 		reg_off += sizeof(uint32_t);
3965 		E1000_WRITE_REG(hw, reg_off,
3966 			(uint32_t)filter->filter_info.mask[i]);
3967 		reg_off += sizeof(uint32_t) * 2;
3968 		++j;
3969 	}
3970 }
3971 
3972 static inline struct e1000_flex_filter *
3973 eth_igb_flex_filter_lookup(struct e1000_flex_filter_list *filter_list,
3974 			struct e1000_flex_filter_info *key)
3975 {
3976 	struct e1000_flex_filter *it;
3977 
3978 	TAILQ_FOREACH(it, filter_list, entries) {
3979 		if (memcmp(key, &it->filter_info,
3980 			sizeof(struct e1000_flex_filter_info)) == 0)
3981 			return it;
3982 	}
3983 
3984 	return NULL;
3985 }
3986 
3987 /* remove a flex byte filter
3988  * @param
3989  * dev: Pointer to struct rte_eth_dev.
3990  * filter: the pointer of the filter will be removed.
3991  */
3992 void
3993 igb_remove_flex_filter(struct rte_eth_dev *dev,
3994 			struct e1000_flex_filter *filter)
3995 {
3996 	struct e1000_filter_info *filter_info =
3997 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3998 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3999 	uint32_t wufc, i;
4000 	uint32_t reg_off;
4001 
4002 	wufc = E1000_READ_REG(hw, E1000_WUFC);
4003 	if (filter->index < E1000_MAX_FHFT)
4004 		reg_off = E1000_FHFT(filter->index);
4005 	else
4006 		reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
4007 
4008 	for (i = 0; i < E1000_FHFT_SIZE_IN_DWD; i++)
4009 		E1000_WRITE_REG(hw, reg_off + i * sizeof(uint32_t), 0);
4010 
4011 	E1000_WRITE_REG(hw, E1000_WUFC, wufc &
4012 		(~(E1000_WUFC_FLX0 << filter->index)));
4013 
4014 	filter_info->flex_mask &= ~(1 << filter->index);
4015 	TAILQ_REMOVE(&filter_info->flex_list, filter, entries);
4016 	rte_free(filter);
4017 }
4018 
4019 int
4020 eth_igb_add_del_flex_filter(struct rte_eth_dev *dev,
4021 			struct rte_eth_flex_filter *filter,
4022 			bool add)
4023 {
4024 	struct e1000_filter_info *filter_info =
4025 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4026 	struct e1000_flex_filter *flex_filter, *it;
4027 	uint32_t mask;
4028 	uint8_t shift, i;
4029 
4030 	flex_filter = rte_zmalloc("e1000_flex_filter",
4031 			sizeof(struct e1000_flex_filter), 0);
4032 	if (flex_filter == NULL)
4033 		return -ENOMEM;
4034 
4035 	flex_filter->filter_info.len = filter->len;
4036 	flex_filter->filter_info.priority = filter->priority;
4037 	memcpy(flex_filter->filter_info.dwords, filter->bytes, filter->len);
4038 	for (i = 0; i < RTE_ALIGN(filter->len, CHAR_BIT) / CHAR_BIT; i++) {
4039 		mask = 0;
4040 		/* reverse bits in flex filter's mask*/
4041 		for (shift = 0; shift < CHAR_BIT; shift++) {
4042 			if (filter->mask[i] & (0x01 << shift))
4043 				mask |= (0x80 >> shift);
4044 		}
4045 		flex_filter->filter_info.mask[i] = mask;
4046 	}
4047 
4048 	it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
4049 				&flex_filter->filter_info);
4050 	if (it == NULL && !add) {
4051 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4052 		rte_free(flex_filter);
4053 		return -ENOENT;
4054 	}
4055 	if (it != NULL && add) {
4056 		PMD_DRV_LOG(ERR, "filter exists.");
4057 		rte_free(flex_filter);
4058 		return -EEXIST;
4059 	}
4060 
4061 	if (add) {
4062 		flex_filter->queue = filter->queue;
4063 		/*
4064 		 * look for an unused flex filter index
4065 		 * and insert the filter into the list.
4066 		 */
4067 		for (i = 0; i < E1000_MAX_FLEX_FILTERS; i++) {
4068 			if (!(filter_info->flex_mask & (1 << i))) {
4069 				filter_info->flex_mask |= 1 << i;
4070 				flex_filter->index = i;
4071 				TAILQ_INSERT_TAIL(&filter_info->flex_list,
4072 					flex_filter,
4073 					entries);
4074 				break;
4075 			}
4076 		}
4077 		if (i >= E1000_MAX_FLEX_FILTERS) {
4078 			PMD_DRV_LOG(ERR, "flex filters are full.");
4079 			rte_free(flex_filter);
4080 			return -ENOSYS;
4081 		}
4082 
4083 		igb_inject_flex_filter(dev, flex_filter);
4084 
4085 	} else {
4086 		igb_remove_flex_filter(dev, it);
4087 		rte_free(flex_filter);
4088 	}
4089 
4090 	return 0;
4091 }
4092 
4093 static int
4094 eth_igb_get_flex_filter(struct rte_eth_dev *dev,
4095 			struct rte_eth_flex_filter *filter)
4096 {
4097 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4098 	struct e1000_filter_info *filter_info =
4099 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4100 	struct e1000_flex_filter flex_filter, *it;
4101 	uint32_t wufc, queueing, wufc_en = 0;
4102 
4103 	memset(&flex_filter, 0, sizeof(struct e1000_flex_filter));
4104 	flex_filter.filter_info.len = filter->len;
4105 	flex_filter.filter_info.priority = filter->priority;
4106 	memcpy(flex_filter.filter_info.dwords, filter->bytes, filter->len);
4107 	memcpy(flex_filter.filter_info.mask, filter->mask,
4108 			RTE_ALIGN(filter->len, CHAR_BIT) / CHAR_BIT);
4109 
4110 	it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
4111 				&flex_filter.filter_info);
4112 	if (it == NULL) {
4113 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4114 		return -ENOENT;
4115 	}
4116 
4117 	wufc = E1000_READ_REG(hw, E1000_WUFC);
4118 	wufc_en = E1000_WUFC_FLEX_HQ | (E1000_WUFC_FLX0 << it->index);
4119 
4120 	if ((wufc & wufc_en) == wufc_en) {
4121 		uint32_t reg_off = 0;
4122 		if (it->index < E1000_MAX_FHFT)
4123 			reg_off = E1000_FHFT(it->index);
4124 		else
4125 			reg_off = E1000_FHFT_EXT(it->index - E1000_MAX_FHFT);
4126 
4127 		queueing = E1000_READ_REG(hw,
4128 				reg_off + E1000_FHFT_QUEUEING_OFFSET);
4129 		filter->len = queueing & E1000_FHFT_QUEUEING_LEN;
4130 		filter->priority = (queueing & E1000_FHFT_QUEUEING_PRIO) >>
4131 			E1000_FHFT_QUEUEING_PRIO_SHIFT;
4132 		filter->queue = (queueing & E1000_FHFT_QUEUEING_QUEUE) >>
4133 			E1000_FHFT_QUEUEING_QUEUE_SHIFT;
4134 		return 0;
4135 	}
4136 	return -ENOENT;
4137 }
4138 
4139 static int
4140 eth_igb_flex_filter_handle(struct rte_eth_dev *dev,
4141 			enum rte_filter_op filter_op,
4142 			void *arg)
4143 {
4144 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4145 	struct rte_eth_flex_filter *filter;
4146 	int ret = 0;
4147 
4148 	MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
4149 
4150 	if (filter_op == RTE_ETH_FILTER_NOP)
4151 		return ret;
4152 
4153 	if (arg == NULL) {
4154 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
4155 			    filter_op);
4156 		return -EINVAL;
4157 	}
4158 
4159 	filter = (struct rte_eth_flex_filter *)arg;
4160 	if (filter->len == 0 || filter->len > E1000_MAX_FLEX_FILTER_LEN
4161 	    || filter->len % sizeof(uint64_t) != 0) {
4162 		PMD_DRV_LOG(ERR, "filter's length is out of range");
4163 		return -EINVAL;
4164 	}
4165 	if (filter->priority > E1000_MAX_FLEX_FILTER_PRI) {
4166 		PMD_DRV_LOG(ERR, "filter's priority is out of range");
4167 		return -EINVAL;
4168 	}
4169 
4170 	switch (filter_op) {
4171 	case RTE_ETH_FILTER_ADD:
4172 		ret = eth_igb_add_del_flex_filter(dev, filter, TRUE);
4173 		break;
4174 	case RTE_ETH_FILTER_DELETE:
4175 		ret = eth_igb_add_del_flex_filter(dev, filter, FALSE);
4176 		break;
4177 	case RTE_ETH_FILTER_GET:
4178 		ret = eth_igb_get_flex_filter(dev, filter);
4179 		break;
4180 	default:
4181 		PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
4182 		ret = -EINVAL;
4183 		break;
4184 	}
4185 
4186 	return ret;
4187 }
4188 
4189 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_5tuple_filter_info*/
4190 static inline int
4191 ntuple_filter_to_5tuple_82576(struct rte_eth_ntuple_filter *filter,
4192 			struct e1000_5tuple_filter_info *filter_info)
4193 {
4194 	if (filter->queue >= IGB_MAX_RX_QUEUE_NUM_82576)
4195 		return -EINVAL;
4196 	if (filter->priority > E1000_2TUPLE_MAX_PRI)
4197 		return -EINVAL;  /* filter index is out of range. */
4198 	if (filter->tcp_flags > TCP_FLAG_ALL)
4199 		return -EINVAL;  /* flags is invalid. */
4200 
4201 	switch (filter->dst_ip_mask) {
4202 	case UINT32_MAX:
4203 		filter_info->dst_ip_mask = 0;
4204 		filter_info->dst_ip = filter->dst_ip;
4205 		break;
4206 	case 0:
4207 		filter_info->dst_ip_mask = 1;
4208 		break;
4209 	default:
4210 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
4211 		return -EINVAL;
4212 	}
4213 
4214 	switch (filter->src_ip_mask) {
4215 	case UINT32_MAX:
4216 		filter_info->src_ip_mask = 0;
4217 		filter_info->src_ip = filter->src_ip;
4218 		break;
4219 	case 0:
4220 		filter_info->src_ip_mask = 1;
4221 		break;
4222 	default:
4223 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
4224 		return -EINVAL;
4225 	}
4226 
4227 	switch (filter->dst_port_mask) {
4228 	case UINT16_MAX:
4229 		filter_info->dst_port_mask = 0;
4230 		filter_info->dst_port = filter->dst_port;
4231 		break;
4232 	case 0:
4233 		filter_info->dst_port_mask = 1;
4234 		break;
4235 	default:
4236 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
4237 		return -EINVAL;
4238 	}
4239 
4240 	switch (filter->src_port_mask) {
4241 	case UINT16_MAX:
4242 		filter_info->src_port_mask = 0;
4243 		filter_info->src_port = filter->src_port;
4244 		break;
4245 	case 0:
4246 		filter_info->src_port_mask = 1;
4247 		break;
4248 	default:
4249 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
4250 		return -EINVAL;
4251 	}
4252 
4253 	switch (filter->proto_mask) {
4254 	case UINT8_MAX:
4255 		filter_info->proto_mask = 0;
4256 		filter_info->proto = filter->proto;
4257 		break;
4258 	case 0:
4259 		filter_info->proto_mask = 1;
4260 		break;
4261 	default:
4262 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
4263 		return -EINVAL;
4264 	}
4265 
4266 	filter_info->priority = (uint8_t)filter->priority;
4267 	if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
4268 		filter_info->tcp_flags = filter->tcp_flags;
4269 	else
4270 		filter_info->tcp_flags = 0;
4271 
4272 	return 0;
4273 }
4274 
4275 static inline struct e1000_5tuple_filter *
4276 igb_5tuple_filter_lookup_82576(struct e1000_5tuple_filter_list *filter_list,
4277 			struct e1000_5tuple_filter_info *key)
4278 {
4279 	struct e1000_5tuple_filter *it;
4280 
4281 	TAILQ_FOREACH(it, filter_list, entries) {
4282 		if (memcmp(key, &it->filter_info,
4283 			sizeof(struct e1000_5tuple_filter_info)) == 0) {
4284 			return it;
4285 		}
4286 	}
4287 	return NULL;
4288 }
4289 
4290 /* inject a igb 5-tuple filter to HW */
4291 static inline void
4292 igb_inject_5tuple_filter_82576(struct rte_eth_dev *dev,
4293 			   struct e1000_5tuple_filter *filter)
4294 {
4295 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4296 	uint32_t ftqf = E1000_FTQF_VF_BP | E1000_FTQF_MASK;
4297 	uint32_t spqf, imir, imir_ext = E1000_IMIREXT_SIZE_BP;
4298 	uint8_t i;
4299 
4300 	i = filter->index;
4301 	ftqf |= filter->filter_info.proto & E1000_FTQF_PROTOCOL_MASK;
4302 	if (filter->filter_info.src_ip_mask == 0) /* 0b means compare. */
4303 		ftqf &= ~E1000_FTQF_MASK_SOURCE_ADDR_BP;
4304 	if (filter->filter_info.dst_ip_mask == 0)
4305 		ftqf &= ~E1000_FTQF_MASK_DEST_ADDR_BP;
4306 	if (filter->filter_info.src_port_mask == 0)
4307 		ftqf &= ~E1000_FTQF_MASK_SOURCE_PORT_BP;
4308 	if (filter->filter_info.proto_mask == 0)
4309 		ftqf &= ~E1000_FTQF_MASK_PROTO_BP;
4310 	ftqf |= (filter->queue << E1000_FTQF_QUEUE_SHIFT) &
4311 		E1000_FTQF_QUEUE_MASK;
4312 	ftqf |= E1000_FTQF_QUEUE_ENABLE;
4313 	E1000_WRITE_REG(hw, E1000_FTQF(i), ftqf);
4314 	E1000_WRITE_REG(hw, E1000_DAQF(i), filter->filter_info.dst_ip);
4315 	E1000_WRITE_REG(hw, E1000_SAQF(i), filter->filter_info.src_ip);
4316 
4317 	spqf = filter->filter_info.src_port & E1000_SPQF_SRCPORT;
4318 	E1000_WRITE_REG(hw, E1000_SPQF(i), spqf);
4319 
4320 	imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
4321 	if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
4322 		imir |= E1000_IMIR_PORT_BP;
4323 	else
4324 		imir &= ~E1000_IMIR_PORT_BP;
4325 	imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
4326 
4327 	/* tcp flags bits setting. */
4328 	if (filter->filter_info.tcp_flags & TCP_FLAG_ALL) {
4329 		if (filter->filter_info.tcp_flags & TCP_URG_FLAG)
4330 			imir_ext |= E1000_IMIREXT_CTRL_URG;
4331 		if (filter->filter_info.tcp_flags & TCP_ACK_FLAG)
4332 			imir_ext |= E1000_IMIREXT_CTRL_ACK;
4333 		if (filter->filter_info.tcp_flags & TCP_PSH_FLAG)
4334 			imir_ext |= E1000_IMIREXT_CTRL_PSH;
4335 		if (filter->filter_info.tcp_flags & TCP_RST_FLAG)
4336 			imir_ext |= E1000_IMIREXT_CTRL_RST;
4337 		if (filter->filter_info.tcp_flags & TCP_SYN_FLAG)
4338 			imir_ext |= E1000_IMIREXT_CTRL_SYN;
4339 		if (filter->filter_info.tcp_flags & TCP_FIN_FLAG)
4340 			imir_ext |= E1000_IMIREXT_CTRL_FIN;
4341 	} else {
4342 		imir_ext |= E1000_IMIREXT_CTRL_BP;
4343 	}
4344 	E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
4345 	E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
4346 }
4347 
4348 /*
4349  * igb_add_5tuple_filter_82576 - add a 5tuple filter
4350  *
4351  * @param
4352  * dev: Pointer to struct rte_eth_dev.
4353  * ntuple_filter: ponter to the filter that will be added.
4354  *
4355  * @return
4356  *    - On success, zero.
4357  *    - On failure, a negative value.
4358  */
4359 static int
4360 igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
4361 			struct rte_eth_ntuple_filter *ntuple_filter)
4362 {
4363 	struct e1000_filter_info *filter_info =
4364 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4365 	struct e1000_5tuple_filter *filter;
4366 	uint8_t i;
4367 	int ret;
4368 
4369 	filter = rte_zmalloc("e1000_5tuple_filter",
4370 			sizeof(struct e1000_5tuple_filter), 0);
4371 	if (filter == NULL)
4372 		return -ENOMEM;
4373 
4374 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4375 					    &filter->filter_info);
4376 	if (ret < 0) {
4377 		rte_free(filter);
4378 		return ret;
4379 	}
4380 
4381 	if (igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4382 					 &filter->filter_info) != NULL) {
4383 		PMD_DRV_LOG(ERR, "filter exists.");
4384 		rte_free(filter);
4385 		return -EEXIST;
4386 	}
4387 	filter->queue = ntuple_filter->queue;
4388 
4389 	/*
4390 	 * look for an unused 5tuple filter index,
4391 	 * and insert the filter to list.
4392 	 */
4393 	for (i = 0; i < E1000_MAX_FTQF_FILTERS; i++) {
4394 		if (!(filter_info->fivetuple_mask & (1 << i))) {
4395 			filter_info->fivetuple_mask |= 1 << i;
4396 			filter->index = i;
4397 			TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
4398 					  filter,
4399 					  entries);
4400 			break;
4401 		}
4402 	}
4403 	if (i >= E1000_MAX_FTQF_FILTERS) {
4404 		PMD_DRV_LOG(ERR, "5tuple filters are full.");
4405 		rte_free(filter);
4406 		return -ENOSYS;
4407 	}
4408 
4409 	igb_inject_5tuple_filter_82576(dev, filter);
4410 	return 0;
4411 }
4412 
4413 int
4414 igb_delete_5tuple_filter_82576(struct rte_eth_dev *dev,
4415 				struct e1000_5tuple_filter *filter)
4416 {
4417 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4418 	struct e1000_filter_info *filter_info =
4419 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4420 
4421 	filter_info->fivetuple_mask &= ~(1 << filter->index);
4422 	TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
4423 	rte_free(filter);
4424 
4425 	E1000_WRITE_REG(hw, E1000_FTQF(filter->index),
4426 			E1000_FTQF_VF_BP | E1000_FTQF_MASK);
4427 	E1000_WRITE_REG(hw, E1000_DAQF(filter->index), 0);
4428 	E1000_WRITE_REG(hw, E1000_SAQF(filter->index), 0);
4429 	E1000_WRITE_REG(hw, E1000_SPQF(filter->index), 0);
4430 	E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
4431 	E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
4432 	return 0;
4433 }
4434 
4435 /*
4436  * igb_remove_5tuple_filter_82576 - remove a 5tuple filter
4437  *
4438  * @param
4439  * dev: Pointer to struct rte_eth_dev.
4440  * ntuple_filter: ponter to the filter that will be removed.
4441  *
4442  * @return
4443  *    - On success, zero.
4444  *    - On failure, a negative value.
4445  */
4446 static int
4447 igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
4448 				struct rte_eth_ntuple_filter *ntuple_filter)
4449 {
4450 	struct e1000_filter_info *filter_info =
4451 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4452 	struct e1000_5tuple_filter_info filter_5tuple;
4453 	struct e1000_5tuple_filter *filter;
4454 	int ret;
4455 
4456 	memset(&filter_5tuple, 0, sizeof(struct e1000_5tuple_filter_info));
4457 	ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4458 					    &filter_5tuple);
4459 	if (ret < 0)
4460 		return ret;
4461 
4462 	filter = igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4463 					 &filter_5tuple);
4464 	if (filter == NULL) {
4465 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
4466 		return -ENOENT;
4467 	}
4468 
4469 	igb_delete_5tuple_filter_82576(dev, filter);
4470 
4471 	return 0;
4472 }
4473 
4474 static int
4475 eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4476 {
4477 	uint32_t rctl;
4478 	struct e1000_hw *hw;
4479 	struct rte_eth_dev_info dev_info;
4480 	uint32_t frame_size = mtu + (ETHER_HDR_LEN + ETHER_CRC_LEN +
4481 				     VLAN_TAG_SIZE);
4482 
4483 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4484 
4485 #ifdef RTE_LIBRTE_82571_SUPPORT
4486 	/* XXX: not bigger than max_rx_pktlen */
4487 	if (hw->mac.type == e1000_82571)
4488 		return -ENOTSUP;
4489 #endif
4490 	eth_igb_infos_get(dev, &dev_info);
4491 
4492 	/* check that mtu is within the allowed range */
4493 	if ((mtu < ETHER_MIN_MTU) ||
4494 	    (frame_size > dev_info.max_rx_pktlen))
4495 		return -EINVAL;
4496 
4497 	/* refuse mtu that requires the support of scattered packets when this
4498 	 * feature has not been enabled before. */
4499 	if (!dev->data->scattered_rx &&
4500 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
4501 		return -EINVAL;
4502 
4503 	rctl = E1000_READ_REG(hw, E1000_RCTL);
4504 
4505 	/* switch to jumbo mode if needed */
4506 	if (frame_size > ETHER_MAX_LEN) {
4507 		dev->data->dev_conf.rxmode.jumbo_frame = 1;
4508 		rctl |= E1000_RCTL_LPE;
4509 	} else {
4510 		dev->data->dev_conf.rxmode.jumbo_frame = 0;
4511 		rctl &= ~E1000_RCTL_LPE;
4512 	}
4513 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
4514 
4515 	/* update max frame size */
4516 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4517 
4518 	E1000_WRITE_REG(hw, E1000_RLPML,
4519 			dev->data->dev_conf.rxmode.max_rx_pkt_len);
4520 
4521 	return 0;
4522 }
4523 
4524 /*
4525  * igb_add_del_ntuple_filter - add or delete a ntuple filter
4526  *
4527  * @param
4528  * dev: Pointer to struct rte_eth_dev.
4529  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4530  * add: if true, add filter, if false, remove filter
4531  *
4532  * @return
4533  *    - On success, zero.
4534  *    - On failure, a negative value.
4535  */
4536 int
4537 igb_add_del_ntuple_filter(struct rte_eth_dev *dev,
4538 			struct rte_eth_ntuple_filter *ntuple_filter,
4539 			bool add)
4540 {
4541 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4542 	int ret;
4543 
4544 	switch (ntuple_filter->flags) {
4545 	case RTE_5TUPLE_FLAGS:
4546 	case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4547 		if (hw->mac.type != e1000_82576)
4548 			return -ENOTSUP;
4549 		if (add)
4550 			ret = igb_add_5tuple_filter_82576(dev,
4551 							  ntuple_filter);
4552 		else
4553 			ret = igb_remove_5tuple_filter_82576(dev,
4554 							     ntuple_filter);
4555 		break;
4556 	case RTE_2TUPLE_FLAGS:
4557 	case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4558 		if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350 &&
4559 			hw->mac.type != e1000_i210 &&
4560 			hw->mac.type != e1000_i211)
4561 			return -ENOTSUP;
4562 		if (add)
4563 			ret = igb_add_2tuple_filter(dev, ntuple_filter);
4564 		else
4565 			ret = igb_remove_2tuple_filter(dev, ntuple_filter);
4566 		break;
4567 	default:
4568 		ret = -EINVAL;
4569 		break;
4570 	}
4571 
4572 	return ret;
4573 }
4574 
4575 /*
4576  * igb_get_ntuple_filter - get a ntuple filter
4577  *
4578  * @param
4579  * dev: Pointer to struct rte_eth_dev.
4580  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4581  *
4582  * @return
4583  *    - On success, zero.
4584  *    - On failure, a negative value.
4585  */
4586 static int
4587 igb_get_ntuple_filter(struct rte_eth_dev *dev,
4588 			struct rte_eth_ntuple_filter *ntuple_filter)
4589 {
4590 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4591 	struct e1000_filter_info *filter_info =
4592 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4593 	struct e1000_5tuple_filter_info filter_5tuple;
4594 	struct e1000_2tuple_filter_info filter_2tuple;
4595 	struct e1000_5tuple_filter *p_5tuple_filter;
4596 	struct e1000_2tuple_filter *p_2tuple_filter;
4597 	int ret;
4598 
4599 	switch (ntuple_filter->flags) {
4600 	case RTE_5TUPLE_FLAGS:
4601 	case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4602 		if (hw->mac.type != e1000_82576)
4603 			return -ENOTSUP;
4604 		memset(&filter_5tuple,
4605 			0,
4606 			sizeof(struct e1000_5tuple_filter_info));
4607 		ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4608 						    &filter_5tuple);
4609 		if (ret < 0)
4610 			return ret;
4611 		p_5tuple_filter = igb_5tuple_filter_lookup_82576(
4612 					&filter_info->fivetuple_list,
4613 					&filter_5tuple);
4614 		if (p_5tuple_filter == NULL) {
4615 			PMD_DRV_LOG(ERR, "filter doesn't exist.");
4616 			return -ENOENT;
4617 		}
4618 		ntuple_filter->queue = p_5tuple_filter->queue;
4619 		break;
4620 	case RTE_2TUPLE_FLAGS:
4621 	case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4622 		if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350)
4623 			return -ENOTSUP;
4624 		memset(&filter_2tuple,
4625 			0,
4626 			sizeof(struct e1000_2tuple_filter_info));
4627 		ret = ntuple_filter_to_2tuple(ntuple_filter, &filter_2tuple);
4628 		if (ret < 0)
4629 			return ret;
4630 		p_2tuple_filter = igb_2tuple_filter_lookup(
4631 					&filter_info->twotuple_list,
4632 					&filter_2tuple);
4633 		if (p_2tuple_filter == NULL) {
4634 			PMD_DRV_LOG(ERR, "filter doesn't exist.");
4635 			return -ENOENT;
4636 		}
4637 		ntuple_filter->queue = p_2tuple_filter->queue;
4638 		break;
4639 	default:
4640 		ret = -EINVAL;
4641 		break;
4642 	}
4643 
4644 	return 0;
4645 }
4646 
4647 /*
4648  * igb_ntuple_filter_handle - Handle operations for ntuple filter.
4649  * @dev: pointer to rte_eth_dev structure
4650  * @filter_op:operation will be taken.
4651  * @arg: a pointer to specific structure corresponding to the filter_op
4652  */
4653 static int
4654 igb_ntuple_filter_handle(struct rte_eth_dev *dev,
4655 				enum rte_filter_op filter_op,
4656 				void *arg)
4657 {
4658 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4659 	int ret;
4660 
4661 	MAC_TYPE_FILTER_SUP(hw->mac.type);
4662 
4663 	if (filter_op == RTE_ETH_FILTER_NOP)
4664 		return 0;
4665 
4666 	if (arg == NULL) {
4667 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4668 			    filter_op);
4669 		return -EINVAL;
4670 	}
4671 
4672 	switch (filter_op) {
4673 	case RTE_ETH_FILTER_ADD:
4674 		ret = igb_add_del_ntuple_filter(dev,
4675 			(struct rte_eth_ntuple_filter *)arg,
4676 			TRUE);
4677 		break;
4678 	case RTE_ETH_FILTER_DELETE:
4679 		ret = igb_add_del_ntuple_filter(dev,
4680 			(struct rte_eth_ntuple_filter *)arg,
4681 			FALSE);
4682 		break;
4683 	case RTE_ETH_FILTER_GET:
4684 		ret = igb_get_ntuple_filter(dev,
4685 			(struct rte_eth_ntuple_filter *)arg);
4686 		break;
4687 	default:
4688 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
4689 		ret = -EINVAL;
4690 		break;
4691 	}
4692 	return ret;
4693 }
4694 
4695 static inline int
4696 igb_ethertype_filter_lookup(struct e1000_filter_info *filter_info,
4697 			uint16_t ethertype)
4698 {
4699 	int i;
4700 
4701 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4702 		if (filter_info->ethertype_filters[i].ethertype == ethertype &&
4703 		    (filter_info->ethertype_mask & (1 << i)))
4704 			return i;
4705 	}
4706 	return -1;
4707 }
4708 
4709 static inline int
4710 igb_ethertype_filter_insert(struct e1000_filter_info *filter_info,
4711 			uint16_t ethertype, uint32_t etqf)
4712 {
4713 	int i;
4714 
4715 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4716 		if (!(filter_info->ethertype_mask & (1 << i))) {
4717 			filter_info->ethertype_mask |= 1 << i;
4718 			filter_info->ethertype_filters[i].ethertype = ethertype;
4719 			filter_info->ethertype_filters[i].etqf = etqf;
4720 			return i;
4721 		}
4722 	}
4723 	return -1;
4724 }
4725 
4726 int
4727 igb_ethertype_filter_remove(struct e1000_filter_info *filter_info,
4728 			uint8_t idx)
4729 {
4730 	if (idx >= E1000_MAX_ETQF_FILTERS)
4731 		return -1;
4732 	filter_info->ethertype_mask &= ~(1 << idx);
4733 	filter_info->ethertype_filters[idx].ethertype = 0;
4734 	filter_info->ethertype_filters[idx].etqf = 0;
4735 	return idx;
4736 }
4737 
4738 
4739 int
4740 igb_add_del_ethertype_filter(struct rte_eth_dev *dev,
4741 			struct rte_eth_ethertype_filter *filter,
4742 			bool add)
4743 {
4744 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4745 	struct e1000_filter_info *filter_info =
4746 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4747 	uint32_t etqf = 0;
4748 	int ret;
4749 
4750 	if (filter->ether_type == ETHER_TYPE_IPv4 ||
4751 		filter->ether_type == ETHER_TYPE_IPv6) {
4752 		PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
4753 			" ethertype filter.", filter->ether_type);
4754 		return -EINVAL;
4755 	}
4756 
4757 	if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
4758 		PMD_DRV_LOG(ERR, "mac compare is unsupported.");
4759 		return -EINVAL;
4760 	}
4761 	if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
4762 		PMD_DRV_LOG(ERR, "drop option is unsupported.");
4763 		return -EINVAL;
4764 	}
4765 
4766 	ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4767 	if (ret >= 0 && add) {
4768 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
4769 			    filter->ether_type);
4770 		return -EEXIST;
4771 	}
4772 	if (ret < 0 && !add) {
4773 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4774 			    filter->ether_type);
4775 		return -ENOENT;
4776 	}
4777 
4778 	if (add) {
4779 		etqf |= E1000_ETQF_FILTER_ENABLE | E1000_ETQF_QUEUE_ENABLE;
4780 		etqf |= (uint32_t)(filter->ether_type & E1000_ETQF_ETHERTYPE);
4781 		etqf |= filter->queue << E1000_ETQF_QUEUE_SHIFT;
4782 		ret = igb_ethertype_filter_insert(filter_info,
4783 				filter->ether_type, etqf);
4784 		if (ret < 0) {
4785 			PMD_DRV_LOG(ERR, "ethertype filters are full.");
4786 			return -ENOSYS;
4787 		}
4788 	} else {
4789 		ret = igb_ethertype_filter_remove(filter_info, (uint8_t)ret);
4790 		if (ret < 0)
4791 			return -ENOSYS;
4792 	}
4793 	E1000_WRITE_REG(hw, E1000_ETQF(ret), etqf);
4794 	E1000_WRITE_FLUSH(hw);
4795 
4796 	return 0;
4797 }
4798 
4799 static int
4800 igb_get_ethertype_filter(struct rte_eth_dev *dev,
4801 			struct rte_eth_ethertype_filter *filter)
4802 {
4803 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4804 	struct e1000_filter_info *filter_info =
4805 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4806 	uint32_t etqf;
4807 	int ret;
4808 
4809 	ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4810 	if (ret < 0) {
4811 		PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4812 			    filter->ether_type);
4813 		return -ENOENT;
4814 	}
4815 
4816 	etqf = E1000_READ_REG(hw, E1000_ETQF(ret));
4817 	if (etqf & E1000_ETQF_FILTER_ENABLE) {
4818 		filter->ether_type = etqf & E1000_ETQF_ETHERTYPE;
4819 		filter->flags = 0;
4820 		filter->queue = (etqf & E1000_ETQF_QUEUE) >>
4821 				E1000_ETQF_QUEUE_SHIFT;
4822 		return 0;
4823 	}
4824 
4825 	return -ENOENT;
4826 }
4827 
4828 /*
4829  * igb_ethertype_filter_handle - Handle operations for ethertype filter.
4830  * @dev: pointer to rte_eth_dev structure
4831  * @filter_op:operation will be taken.
4832  * @arg: a pointer to specific structure corresponding to the filter_op
4833  */
4834 static int
4835 igb_ethertype_filter_handle(struct rte_eth_dev *dev,
4836 				enum rte_filter_op filter_op,
4837 				void *arg)
4838 {
4839 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4840 	int ret;
4841 
4842 	MAC_TYPE_FILTER_SUP(hw->mac.type);
4843 
4844 	if (filter_op == RTE_ETH_FILTER_NOP)
4845 		return 0;
4846 
4847 	if (arg == NULL) {
4848 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4849 			    filter_op);
4850 		return -EINVAL;
4851 	}
4852 
4853 	switch (filter_op) {
4854 	case RTE_ETH_FILTER_ADD:
4855 		ret = igb_add_del_ethertype_filter(dev,
4856 			(struct rte_eth_ethertype_filter *)arg,
4857 			TRUE);
4858 		break;
4859 	case RTE_ETH_FILTER_DELETE:
4860 		ret = igb_add_del_ethertype_filter(dev,
4861 			(struct rte_eth_ethertype_filter *)arg,
4862 			FALSE);
4863 		break;
4864 	case RTE_ETH_FILTER_GET:
4865 		ret = igb_get_ethertype_filter(dev,
4866 			(struct rte_eth_ethertype_filter *)arg);
4867 		break;
4868 	default:
4869 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
4870 		ret = -EINVAL;
4871 		break;
4872 	}
4873 	return ret;
4874 }
4875 
4876 static int
4877 eth_igb_filter_ctrl(struct rte_eth_dev *dev,
4878 		     enum rte_filter_type filter_type,
4879 		     enum rte_filter_op filter_op,
4880 		     void *arg)
4881 {
4882 	int ret = 0;
4883 
4884 	switch (filter_type) {
4885 	case RTE_ETH_FILTER_NTUPLE:
4886 		ret = igb_ntuple_filter_handle(dev, filter_op, arg);
4887 		break;
4888 	case RTE_ETH_FILTER_ETHERTYPE:
4889 		ret = igb_ethertype_filter_handle(dev, filter_op, arg);
4890 		break;
4891 	case RTE_ETH_FILTER_SYN:
4892 		ret = eth_igb_syn_filter_handle(dev, filter_op, arg);
4893 		break;
4894 	case RTE_ETH_FILTER_FLEXIBLE:
4895 		ret = eth_igb_flex_filter_handle(dev, filter_op, arg);
4896 		break;
4897 	case RTE_ETH_FILTER_GENERIC:
4898 		if (filter_op != RTE_ETH_FILTER_GET)
4899 			return -EINVAL;
4900 		*(const void **)arg = &igb_flow_ops;
4901 		break;
4902 	default:
4903 		PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
4904 							filter_type);
4905 		break;
4906 	}
4907 
4908 	return ret;
4909 }
4910 
4911 static int
4912 eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
4913 			 struct ether_addr *mc_addr_set,
4914 			 uint32_t nb_mc_addr)
4915 {
4916 	struct e1000_hw *hw;
4917 
4918 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4919 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
4920 	return 0;
4921 }
4922 
4923 static uint64_t
4924 igb_read_systime_cyclecounter(struct rte_eth_dev *dev)
4925 {
4926 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4927 	uint64_t systime_cycles;
4928 
4929 	switch (hw->mac.type) {
4930 	case e1000_i210:
4931 	case e1000_i211:
4932 		/*
4933 		 * Need to read System Time Residue Register to be able
4934 		 * to read the other two registers.
4935 		 */
4936 		E1000_READ_REG(hw, E1000_SYSTIMR);
4937 		/* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
4938 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4939 		systime_cycles += (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4940 				* NSEC_PER_SEC;
4941 		break;
4942 	case e1000_82580:
4943 	case e1000_i350:
4944 	case e1000_i354:
4945 		/*
4946 		 * Need to read System Time Residue Register to be able
4947 		 * to read the other two registers.
4948 		 */
4949 		E1000_READ_REG(hw, E1000_SYSTIMR);
4950 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4951 		/* Only the 8 LSB are valid. */
4952 		systime_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_SYSTIMH)
4953 				& 0xff) << 32;
4954 		break;
4955 	default:
4956 		systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4957 		systime_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4958 				<< 32;
4959 		break;
4960 	}
4961 
4962 	return systime_cycles;
4963 }
4964 
4965 static uint64_t
4966 igb_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
4967 {
4968 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4969 	uint64_t rx_tstamp_cycles;
4970 
4971 	switch (hw->mac.type) {
4972 	case e1000_i210:
4973 	case e1000_i211:
4974 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
4975 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4976 		rx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
4977 				* NSEC_PER_SEC;
4978 		break;
4979 	case e1000_82580:
4980 	case e1000_i350:
4981 	case e1000_i354:
4982 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4983 		/* Only the 8 LSB are valid. */
4984 		rx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_RXSTMPH)
4985 				& 0xff) << 32;
4986 		break;
4987 	default:
4988 		rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
4989 		rx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
4990 				<< 32;
4991 		break;
4992 	}
4993 
4994 	return rx_tstamp_cycles;
4995 }
4996 
4997 static uint64_t
4998 igb_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
4999 {
5000 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5001 	uint64_t tx_tstamp_cycles;
5002 
5003 	switch (hw->mac.type) {
5004 	case e1000_i210:
5005 	case e1000_i211:
5006 		/* RXSTMPL stores ns and RXSTMPH stores seconds. */
5007 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5008 		tx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
5009 				* NSEC_PER_SEC;
5010 		break;
5011 	case e1000_82580:
5012 	case e1000_i350:
5013 	case e1000_i354:
5014 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5015 		/* Only the 8 LSB are valid. */
5016 		tx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_TXSTMPH)
5017 				& 0xff) << 32;
5018 		break;
5019 	default:
5020 		tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5021 		tx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
5022 				<< 32;
5023 		break;
5024 	}
5025 
5026 	return tx_tstamp_cycles;
5027 }
5028 
5029 static void
5030 igb_start_timecounters(struct rte_eth_dev *dev)
5031 {
5032 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5033 	struct e1000_adapter *adapter =
5034 		(struct e1000_adapter *)dev->data->dev_private;
5035 	uint32_t incval = 1;
5036 	uint32_t shift = 0;
5037 	uint64_t mask = E1000_CYCLECOUNTER_MASK;
5038 
5039 	switch (hw->mac.type) {
5040 	case e1000_82580:
5041 	case e1000_i350:
5042 	case e1000_i354:
5043 		/* 32 LSB bits + 8 MSB bits = 40 bits */
5044 		mask = (1ULL << 40) - 1;
5045 		/* fall-through */
5046 	case e1000_i210:
5047 	case e1000_i211:
5048 		/*
5049 		 * Start incrementing the register
5050 		 * used to timestamp PTP packets.
5051 		 */
5052 		E1000_WRITE_REG(hw, E1000_TIMINCA, incval);
5053 		break;
5054 	case e1000_82576:
5055 		incval = E1000_INCVALUE_82576;
5056 		shift = IGB_82576_TSYNC_SHIFT;
5057 		E1000_WRITE_REG(hw, E1000_TIMINCA,
5058 				E1000_INCPERIOD_82576 | incval);
5059 		break;
5060 	default:
5061 		/* Not supported */
5062 		return;
5063 	}
5064 
5065 	memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
5066 	memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
5067 	memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
5068 
5069 	adapter->systime_tc.cc_mask = mask;
5070 	adapter->systime_tc.cc_shift = shift;
5071 	adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
5072 
5073 	adapter->rx_tstamp_tc.cc_mask = mask;
5074 	adapter->rx_tstamp_tc.cc_shift = shift;
5075 	adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
5076 
5077 	adapter->tx_tstamp_tc.cc_mask = mask;
5078 	adapter->tx_tstamp_tc.cc_shift = shift;
5079 	adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
5080 }
5081 
5082 static int
5083 igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
5084 {
5085 	struct e1000_adapter *adapter =
5086 			(struct e1000_adapter *)dev->data->dev_private;
5087 
5088 	adapter->systime_tc.nsec += delta;
5089 	adapter->rx_tstamp_tc.nsec += delta;
5090 	adapter->tx_tstamp_tc.nsec += delta;
5091 
5092 	return 0;
5093 }
5094 
5095 static int
5096 igb_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
5097 {
5098 	uint64_t ns;
5099 	struct e1000_adapter *adapter =
5100 			(struct e1000_adapter *)dev->data->dev_private;
5101 
5102 	ns = rte_timespec_to_ns(ts);
5103 
5104 	/* Set the timecounters to a new value. */
5105 	adapter->systime_tc.nsec = ns;
5106 	adapter->rx_tstamp_tc.nsec = ns;
5107 	adapter->tx_tstamp_tc.nsec = ns;
5108 
5109 	return 0;
5110 }
5111 
5112 static int
5113 igb_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
5114 {
5115 	uint64_t ns, systime_cycles;
5116 	struct e1000_adapter *adapter =
5117 			(struct e1000_adapter *)dev->data->dev_private;
5118 
5119 	systime_cycles = igb_read_systime_cyclecounter(dev);
5120 	ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
5121 	*ts = rte_ns_to_timespec(ns);
5122 
5123 	return 0;
5124 }
5125 
5126 static int
5127 igb_timesync_enable(struct rte_eth_dev *dev)
5128 {
5129 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5130 	uint32_t tsync_ctl;
5131 	uint32_t tsauxc;
5132 
5133 	/* Stop the timesync system time. */
5134 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0x0);
5135 	/* Reset the timesync system time value. */
5136 	switch (hw->mac.type) {
5137 	case e1000_82580:
5138 	case e1000_i350:
5139 	case e1000_i354:
5140 	case e1000_i210:
5141 	case e1000_i211:
5142 		E1000_WRITE_REG(hw, E1000_SYSTIMR, 0x0);
5143 		/* fall-through */
5144 	case e1000_82576:
5145 		E1000_WRITE_REG(hw, E1000_SYSTIML, 0x0);
5146 		E1000_WRITE_REG(hw, E1000_SYSTIMH, 0x0);
5147 		break;
5148 	default:
5149 		/* Not supported. */
5150 		return -ENOTSUP;
5151 	}
5152 
5153 	/* Enable system time for it isn't on by default. */
5154 	tsauxc = E1000_READ_REG(hw, E1000_TSAUXC);
5155 	tsauxc &= ~E1000_TSAUXC_DISABLE_SYSTIME;
5156 	E1000_WRITE_REG(hw, E1000_TSAUXC, tsauxc);
5157 
5158 	igb_start_timecounters(dev);
5159 
5160 	/* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5161 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588),
5162 			(ETHER_TYPE_1588 |
5163 			 E1000_ETQF_FILTER_ENABLE |
5164 			 E1000_ETQF_1588));
5165 
5166 	/* Enable timestamping of received PTP packets. */
5167 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5168 	tsync_ctl |= E1000_TSYNCRXCTL_ENABLED;
5169 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
5170 
5171 	/* Enable Timestamping of transmitted PTP packets. */
5172 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5173 	tsync_ctl |= E1000_TSYNCTXCTL_ENABLED;
5174 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
5175 
5176 	return 0;
5177 }
5178 
5179 static int
5180 igb_timesync_disable(struct rte_eth_dev *dev)
5181 {
5182 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5183 	uint32_t tsync_ctl;
5184 
5185 	/* Disable timestamping of transmitted PTP packets. */
5186 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5187 	tsync_ctl &= ~E1000_TSYNCTXCTL_ENABLED;
5188 	E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
5189 
5190 	/* Disable timestamping of received PTP packets. */
5191 	tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5192 	tsync_ctl &= ~E1000_TSYNCRXCTL_ENABLED;
5193 	E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
5194 
5195 	/* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5196 	E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588), 0);
5197 
5198 	/* Stop incrementating the System Time registers. */
5199 	E1000_WRITE_REG(hw, E1000_TIMINCA, 0);
5200 
5201 	return 0;
5202 }
5203 
5204 static int
5205 igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
5206 			       struct timespec *timestamp,
5207 			       uint32_t flags __rte_unused)
5208 {
5209 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5210 	struct e1000_adapter *adapter =
5211 			(struct e1000_adapter *)dev->data->dev_private;
5212 	uint32_t tsync_rxctl;
5213 	uint64_t rx_tstamp_cycles;
5214 	uint64_t ns;
5215 
5216 	tsync_rxctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5217 	if ((tsync_rxctl & E1000_TSYNCRXCTL_VALID) == 0)
5218 		return -EINVAL;
5219 
5220 	rx_tstamp_cycles = igb_read_rx_tstamp_cyclecounter(dev);
5221 	ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
5222 	*timestamp = rte_ns_to_timespec(ns);
5223 
5224 	return  0;
5225 }
5226 
5227 static int
5228 igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
5229 			       struct timespec *timestamp)
5230 {
5231 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5232 	struct e1000_adapter *adapter =
5233 			(struct e1000_adapter *)dev->data->dev_private;
5234 	uint32_t tsync_txctl;
5235 	uint64_t tx_tstamp_cycles;
5236 	uint64_t ns;
5237 
5238 	tsync_txctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5239 	if ((tsync_txctl & E1000_TSYNCTXCTL_VALID) == 0)
5240 		return -EINVAL;
5241 
5242 	tx_tstamp_cycles = igb_read_tx_tstamp_cyclecounter(dev);
5243 	ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
5244 	*timestamp = rte_ns_to_timespec(ns);
5245 
5246 	return  0;
5247 }
5248 
5249 static int
5250 eth_igb_get_reg_length(struct rte_eth_dev *dev __rte_unused)
5251 {
5252 	int count = 0;
5253 	int g_ind = 0;
5254 	const struct reg_info *reg_group;
5255 
5256 	while ((reg_group = igb_regs[g_ind++]))
5257 		count += igb_reg_group_count(reg_group);
5258 
5259 	return count;
5260 }
5261 
5262 static int
5263 igbvf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
5264 {
5265 	int count = 0;
5266 	int g_ind = 0;
5267 	const struct reg_info *reg_group;
5268 
5269 	while ((reg_group = igbvf_regs[g_ind++]))
5270 		count += igb_reg_group_count(reg_group);
5271 
5272 	return count;
5273 }
5274 
5275 static int
5276 eth_igb_get_regs(struct rte_eth_dev *dev,
5277 	struct rte_dev_reg_info *regs)
5278 {
5279 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5280 	uint32_t *data = regs->data;
5281 	int g_ind = 0;
5282 	int count = 0;
5283 	const struct reg_info *reg_group;
5284 
5285 	if (data == NULL) {
5286 		regs->length = eth_igb_get_reg_length(dev);
5287 		regs->width = sizeof(uint32_t);
5288 		return 0;
5289 	}
5290 
5291 	/* Support only full register dump */
5292 	if ((regs->length == 0) ||
5293 	    (regs->length == (uint32_t)eth_igb_get_reg_length(dev))) {
5294 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5295 			hw->device_id;
5296 		while ((reg_group = igb_regs[g_ind++]))
5297 			count += igb_read_regs_group(dev, &data[count],
5298 							reg_group);
5299 		return 0;
5300 	}
5301 
5302 	return -ENOTSUP;
5303 }
5304 
5305 static int
5306 igbvf_get_regs(struct rte_eth_dev *dev,
5307 	struct rte_dev_reg_info *regs)
5308 {
5309 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5310 	uint32_t *data = regs->data;
5311 	int g_ind = 0;
5312 	int count = 0;
5313 	const struct reg_info *reg_group;
5314 
5315 	if (data == NULL) {
5316 		regs->length = igbvf_get_reg_length(dev);
5317 		regs->width = sizeof(uint32_t);
5318 		return 0;
5319 	}
5320 
5321 	/* Support only full register dump */
5322 	if ((regs->length == 0) ||
5323 	    (regs->length == (uint32_t)igbvf_get_reg_length(dev))) {
5324 		regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5325 			hw->device_id;
5326 		while ((reg_group = igbvf_regs[g_ind++]))
5327 			count += igb_read_regs_group(dev, &data[count],
5328 							reg_group);
5329 		return 0;
5330 	}
5331 
5332 	return -ENOTSUP;
5333 }
5334 
5335 static int
5336 eth_igb_get_eeprom_length(struct rte_eth_dev *dev)
5337 {
5338 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5339 
5340 	/* Return unit is byte count */
5341 	return hw->nvm.word_size * 2;
5342 }
5343 
5344 static int
5345 eth_igb_get_eeprom(struct rte_eth_dev *dev,
5346 	struct rte_dev_eeprom_info *in_eeprom)
5347 {
5348 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5349 	struct e1000_nvm_info *nvm = &hw->nvm;
5350 	uint16_t *data = in_eeprom->data;
5351 	int first, length;
5352 
5353 	first = in_eeprom->offset >> 1;
5354 	length = in_eeprom->length >> 1;
5355 	if ((first >= hw->nvm.word_size) ||
5356 	    ((first + length) >= hw->nvm.word_size))
5357 		return -EINVAL;
5358 
5359 	in_eeprom->magic = hw->vendor_id |
5360 		((uint32_t)hw->device_id << 16);
5361 
5362 	if ((nvm->ops.read) == NULL)
5363 		return -ENOTSUP;
5364 
5365 	return nvm->ops.read(hw, first, length, data);
5366 }
5367 
5368 static int
5369 eth_igb_set_eeprom(struct rte_eth_dev *dev,
5370 	struct rte_dev_eeprom_info *in_eeprom)
5371 {
5372 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5373 	struct e1000_nvm_info *nvm = &hw->nvm;
5374 	uint16_t *data = in_eeprom->data;
5375 	int first, length;
5376 
5377 	first = in_eeprom->offset >> 1;
5378 	length = in_eeprom->length >> 1;
5379 	if ((first >= hw->nvm.word_size) ||
5380 	    ((first + length) >= hw->nvm.word_size))
5381 		return -EINVAL;
5382 
5383 	in_eeprom->magic = (uint32_t)hw->vendor_id |
5384 		((uint32_t)hw->device_id << 16);
5385 
5386 	if ((nvm->ops.write) == NULL)
5387 		return -ENOTSUP;
5388 	return nvm->ops.write(hw,  first, length, data);
5389 }
5390 
5391 static int
5392 eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5393 {
5394 	struct e1000_hw *hw =
5395 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5396 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5397 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5398 	uint32_t vec = E1000_MISC_VEC_ID;
5399 
5400 	if (rte_intr_allow_others(intr_handle))
5401 		vec = E1000_RX_VEC_START;
5402 
5403 	uint32_t mask = 1 << (queue_id + vec);
5404 
5405 	E1000_WRITE_REG(hw, E1000_EIMC, mask);
5406 	E1000_WRITE_FLUSH(hw);
5407 
5408 	return 0;
5409 }
5410 
5411 static int
5412 eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5413 {
5414 	struct e1000_hw *hw =
5415 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5416 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5417 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5418 	uint32_t vec = E1000_MISC_VEC_ID;
5419 
5420 	if (rte_intr_allow_others(intr_handle))
5421 		vec = E1000_RX_VEC_START;
5422 
5423 	uint32_t mask = 1 << (queue_id + vec);
5424 	uint32_t regval;
5425 
5426 	regval = E1000_READ_REG(hw, E1000_EIMS);
5427 	E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
5428 	E1000_WRITE_FLUSH(hw);
5429 
5430 	rte_intr_enable(intr_handle);
5431 
5432 	return 0;
5433 }
5434 
5435 static void
5436 eth_igb_write_ivar(struct e1000_hw *hw, uint8_t  msix_vector,
5437 		   uint8_t index, uint8_t offset)
5438 {
5439 	uint32_t val = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
5440 
5441 	/* clear bits */
5442 	val &= ~((uint32_t)0xFF << offset);
5443 
5444 	/* write vector and valid bit */
5445 	val |= (msix_vector | E1000_IVAR_VALID) << offset;
5446 
5447 	E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, val);
5448 }
5449 
5450 static void
5451 eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
5452 			   uint8_t queue, uint8_t msix_vector)
5453 {
5454 	uint32_t tmp = 0;
5455 
5456 	if (hw->mac.type == e1000_82575) {
5457 		if (direction == 0)
5458 			tmp = E1000_EICR_RX_QUEUE0 << queue;
5459 		else if (direction == 1)
5460 			tmp = E1000_EICR_TX_QUEUE0 << queue;
5461 		E1000_WRITE_REG(hw, E1000_MSIXBM(msix_vector), tmp);
5462 	} else if (hw->mac.type == e1000_82576) {
5463 		if ((direction == 0) || (direction == 1))
5464 			eth_igb_write_ivar(hw, msix_vector, queue & 0x7,
5465 					   ((queue & 0x8) << 1) +
5466 					   8 * direction);
5467 	} else if ((hw->mac.type == e1000_82580) ||
5468 			(hw->mac.type == e1000_i350) ||
5469 			(hw->mac.type == e1000_i354) ||
5470 			(hw->mac.type == e1000_i210) ||
5471 			(hw->mac.type == e1000_i211)) {
5472 		if ((direction == 0) || (direction == 1))
5473 			eth_igb_write_ivar(hw, msix_vector,
5474 					   queue >> 1,
5475 					   ((queue & 0x1) << 4) +
5476 					   8 * direction);
5477 	}
5478 }
5479 
5480 /* Sets up the hardware to generate MSI-X interrupts properly
5481  * @hw
5482  *  board private structure
5483  */
5484 static void
5485 eth_igb_configure_msix_intr(struct rte_eth_dev *dev)
5486 {
5487 	int queue_id;
5488 	uint32_t tmpval, regval, intr_mask;
5489 	struct e1000_hw *hw =
5490 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5491 	uint32_t vec = E1000_MISC_VEC_ID;
5492 	uint32_t base = E1000_MISC_VEC_ID;
5493 	uint32_t misc_shift = 0;
5494 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5495 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5496 
5497 	/* won't configure msix register if no mapping is done
5498 	 * between intr vector and event fd
5499 	 */
5500 	if (!rte_intr_dp_is_en(intr_handle))
5501 		return;
5502 
5503 	if (rte_intr_allow_others(intr_handle)) {
5504 		vec = base = E1000_RX_VEC_START;
5505 		misc_shift = 1;
5506 	}
5507 
5508 	/* set interrupt vector for other causes */
5509 	if (hw->mac.type == e1000_82575) {
5510 		tmpval = E1000_READ_REG(hw, E1000_CTRL_EXT);
5511 		/* enable MSI-X PBA support */
5512 		tmpval |= E1000_CTRL_EXT_PBA_CLR;
5513 
5514 		/* Auto-Mask interrupts upon ICR read */
5515 		tmpval |= E1000_CTRL_EXT_EIAME;
5516 		tmpval |= E1000_CTRL_EXT_IRCA;
5517 
5518 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmpval);
5519 
5520 		/* enable msix_other interrupt */
5521 		E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0), 0, E1000_EIMS_OTHER);
5522 		regval = E1000_READ_REG(hw, E1000_EIAC);
5523 		E1000_WRITE_REG(hw, E1000_EIAC, regval | E1000_EIMS_OTHER);
5524 		regval = E1000_READ_REG(hw, E1000_EIAM);
5525 		E1000_WRITE_REG(hw, E1000_EIMS, regval | E1000_EIMS_OTHER);
5526 	} else if ((hw->mac.type == e1000_82576) ||
5527 			(hw->mac.type == e1000_82580) ||
5528 			(hw->mac.type == e1000_i350) ||
5529 			(hw->mac.type == e1000_i354) ||
5530 			(hw->mac.type == e1000_i210) ||
5531 			(hw->mac.type == e1000_i211)) {
5532 		/* turn on MSI-X capability first */
5533 		E1000_WRITE_REG(hw, E1000_GPIE, E1000_GPIE_MSIX_MODE |
5534 					E1000_GPIE_PBA | E1000_GPIE_EIAME |
5535 					E1000_GPIE_NSICR);
5536 		intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5537 			misc_shift;
5538 		regval = E1000_READ_REG(hw, E1000_EIAC);
5539 		E1000_WRITE_REG(hw, E1000_EIAC, regval | intr_mask);
5540 
5541 		/* enable msix_other interrupt */
5542 		regval = E1000_READ_REG(hw, E1000_EIMS);
5543 		E1000_WRITE_REG(hw, E1000_EIMS, regval | intr_mask);
5544 		tmpval = (dev->data->nb_rx_queues | E1000_IVAR_VALID) << 8;
5545 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, tmpval);
5546 	}
5547 
5548 	/* use EIAM to auto-mask when MSI-X interrupt
5549 	 * is asserted, this saves a register write for every interrupt
5550 	 */
5551 	intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5552 		misc_shift;
5553 	regval = E1000_READ_REG(hw, E1000_EIAM);
5554 	E1000_WRITE_REG(hw, E1000_EIAM, regval | intr_mask);
5555 
5556 	for (queue_id = 0; queue_id < dev->data->nb_rx_queues; queue_id++) {
5557 		eth_igb_assign_msix_vector(hw, 0, queue_id, vec);
5558 		intr_handle->intr_vec[queue_id] = vec;
5559 		if (vec < base + intr_handle->nb_efd - 1)
5560 			vec++;
5561 	}
5562 
5563 	E1000_WRITE_FLUSH(hw);
5564 }
5565 
5566 /* restore n-tuple filter */
5567 static inline void
5568 igb_ntuple_filter_restore(struct rte_eth_dev *dev)
5569 {
5570 	struct e1000_filter_info *filter_info =
5571 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5572 	struct e1000_5tuple_filter *p_5tuple;
5573 	struct e1000_2tuple_filter *p_2tuple;
5574 
5575 	TAILQ_FOREACH(p_5tuple, &filter_info->fivetuple_list, entries) {
5576 		igb_inject_5tuple_filter_82576(dev, p_5tuple);
5577 	}
5578 
5579 	TAILQ_FOREACH(p_2tuple, &filter_info->twotuple_list, entries) {
5580 		igb_inject_2uple_filter(dev, p_2tuple);
5581 	}
5582 }
5583 
5584 /* restore SYN filter */
5585 static inline void
5586 igb_syn_filter_restore(struct rte_eth_dev *dev)
5587 {
5588 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5589 	struct e1000_filter_info *filter_info =
5590 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5591 	uint32_t synqf;
5592 
5593 	synqf = filter_info->syn_info;
5594 
5595 	if (synqf & E1000_SYN_FILTER_ENABLE) {
5596 		E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
5597 		E1000_WRITE_FLUSH(hw);
5598 	}
5599 }
5600 
5601 /* restore ethernet type filter */
5602 static inline void
5603 igb_ethertype_filter_restore(struct rte_eth_dev *dev)
5604 {
5605 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5606 	struct e1000_filter_info *filter_info =
5607 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5608 	int i;
5609 
5610 	for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
5611 		if (filter_info->ethertype_mask & (1 << i)) {
5612 			E1000_WRITE_REG(hw, E1000_ETQF(i),
5613 				filter_info->ethertype_filters[i].etqf);
5614 			E1000_WRITE_FLUSH(hw);
5615 		}
5616 	}
5617 }
5618 
5619 /* restore flex byte filter */
5620 static inline void
5621 igb_flex_filter_restore(struct rte_eth_dev *dev)
5622 {
5623 	struct e1000_filter_info *filter_info =
5624 		E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5625 	struct e1000_flex_filter *flex_filter;
5626 
5627 	TAILQ_FOREACH(flex_filter, &filter_info->flex_list, entries) {
5628 		igb_inject_flex_filter(dev, flex_filter);
5629 	}
5630 }
5631 
5632 /* restore all types filter */
5633 static int
5634 igb_filter_restore(struct rte_eth_dev *dev)
5635 {
5636 	igb_ntuple_filter_restore(dev);
5637 	igb_ethertype_filter_restore(dev);
5638 	igb_syn_filter_restore(dev);
5639 	igb_flex_filter_restore(dev);
5640 
5641 	return 0;
5642 }
5643 
5644 RTE_PMD_REGISTER_PCI(net_e1000_igb, rte_igb_pmd);
5645 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb, pci_id_igb_map);
5646 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb, "* igb_uio | uio_pci_generic | vfio-pci");
5647 RTE_PMD_REGISTER_PCI(net_e1000_igb_vf, rte_igbvf_pmd);
5648 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb_vf, pci_id_igbvf_map);
5649 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb_vf, "* igb_uio | vfio-pci");
5650