root/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c

/* [<][>][^][v][top][bottom][index][help] */

DEFINITIONS

This source file includes following definitions.
  1. netif_is_ixgbe
  2. ixgbe_read_pci_cfg_word_parent
  3. ixgbe_get_parent_bus_info
  4. ixgbe_pcie_from_parent
  5. ixgbe_check_minimum_link
  6. ixgbe_service_event_schedule
  7. ixgbe_remove_adapter
  8. ixgbe_check_remove
  9. ixgbe_read_reg
  10. ixgbe_check_cfg_remove
  11. ixgbe_read_pci_cfg_word
  12. ixgbe_read_pci_cfg_dword
  13. ixgbe_write_pci_cfg_word
  14. ixgbe_service_event_complete
  15. ixgbe_regdump
  16. ixgbe_print_buffer
  17. ixgbe_dump
  18. ixgbe_release_hw_control
  19. ixgbe_get_hw_control
  20. ixgbe_set_ivar
  21. ixgbe_irq_rearm_queues
  22. ixgbe_update_xoff_rx_lfc
  23. ixgbe_update_xoff_received
  24. ixgbe_get_tx_completed
  25. ixgbe_get_tx_pending
  26. ixgbe_check_tx_hang
  27. ixgbe_tx_timeout_reset
  28. ixgbe_tx_maxrate
  29. ixgbe_clean_tx_irq
  30. ixgbe_update_tx_dca
  31. ixgbe_update_rx_dca
  32. ixgbe_update_dca
  33. ixgbe_setup_dca
  34. __ixgbe_notify_dca
  35. ixgbe_rx_hash
  36. ixgbe_rx_is_fcoe
  37. ixgbe_rx_checksum
  38. ixgbe_rx_offset
  39. ixgbe_alloc_mapped_page
  40. ixgbe_alloc_rx_buffers
  41. ixgbe_set_rsc_gso_size
  42. ixgbe_update_rsc_stats
  43. ixgbe_process_skb_fields
  44. ixgbe_rx_skb
  45. ixgbe_is_non_eop
  46. ixgbe_pull_tail
  47. ixgbe_dma_sync_frag
  48. ixgbe_cleanup_headers
  49. ixgbe_reuse_rx_page
  50. ixgbe_page_is_reserved
  51. ixgbe_can_reuse_rx_page
  52. ixgbe_add_rx_frag
  53. ixgbe_get_rx_buffer
  54. ixgbe_put_rx_buffer
  55. ixgbe_construct_skb
  56. ixgbe_build_skb
  57. ixgbe_run_xdp
  58. ixgbe_rx_buffer_flip
  59. ixgbe_clean_rx_irq
  60. ixgbe_configure_msix
  61. ixgbe_update_itr
  62. ixgbe_write_eitr
  63. ixgbe_set_itr
  64. ixgbe_check_overtemp_subtask
  65. ixgbe_check_fan_failure
  66. ixgbe_check_overtemp_event
  67. ixgbe_is_sfp
  68. ixgbe_check_sfp_event
  69. ixgbe_check_lsc
  70. ixgbe_irq_enable_queues
  71. ixgbe_irq_disable_queues
  72. ixgbe_irq_enable
  73. ixgbe_msix_other
  74. ixgbe_msix_clean_rings
  75. ixgbe_poll
  76. ixgbe_request_msix_irqs
  77. ixgbe_intr
  78. ixgbe_request_irq
  79. ixgbe_free_irq
  80. ixgbe_irq_disable
  81. ixgbe_configure_msi_and_legacy
  82. ixgbe_configure_tx_ring
  83. ixgbe_setup_mtqc
  84. ixgbe_configure_tx
  85. ixgbe_enable_rx_drop
  86. ixgbe_disable_rx_drop
  87. ixgbe_set_rx_drop_en
  88. ixgbe_configure_srrctl
  89. ixgbe_rss_indir_tbl_entries
  90. ixgbe_store_key
  91. ixgbe_init_rss_key
  92. ixgbe_store_reta
  93. ixgbe_store_vfreta
  94. ixgbe_setup_reta
  95. ixgbe_setup_vfreta
  96. ixgbe_setup_mrqc
  97. ixgbe_configure_rscctl
  98. ixgbe_rx_desc_queue_enable
  99. ixgbe_configure_rx_ring
  100. ixgbe_setup_psrtype
  101. ixgbe_configure_virtualization
  102. ixgbe_set_rx_buffer_len
  103. ixgbe_setup_rdrxctl
  104. ixgbe_configure_rx
  105. ixgbe_vlan_rx_add_vid
  106. ixgbe_find_vlvf_entry
  107. ixgbe_update_pf_promisc_vlvf
  108. ixgbe_vlan_rx_kill_vid
  109. ixgbe_vlan_strip_disable
  110. ixgbe_vlan_strip_enable
  111. ixgbe_vlan_promisc_enable
  112. ixgbe_scrub_vfta
  113. ixgbe_vlan_promisc_disable
  114. ixgbe_restore_vlan
  115. ixgbe_write_mc_addr_list
  116. ixgbe_full_sync_mac_table
  117. ixgbe_sync_mac_table
  118. ixgbe_flush_sw_mac_table
  119. ixgbe_available_rars
  120. ixgbe_mac_set_default_filter
  121. ixgbe_add_mac_filter
  122. ixgbe_del_mac_filter
  123. ixgbe_uc_sync
  124. ixgbe_uc_unsync
  125. ixgbe_set_rx_mode
  126. ixgbe_napi_enable_all
  127. ixgbe_napi_disable_all
  128. ixgbe_clear_udp_tunnel_port
  129. ixgbe_configure_dcb
  130. ixgbe_hpbthresh
  131. ixgbe_lpbthresh
  132. ixgbe_pbthresh_setup
  133. ixgbe_configure_pb
  134. ixgbe_fdir_filter_restore
  135. ixgbe_clean_rx_ring
  136. ixgbe_fwd_ring_up
  137. ixgbe_macvlan_up
  138. ixgbe_configure_dfwd
  139. ixgbe_configure
  140. ixgbe_sfp_link_config
  141. ixgbe_non_sfp_link_config
  142. ixgbe_setup_gpie
  143. ixgbe_up_complete
  144. ixgbe_reinit_locked
  145. ixgbe_up
  146. ixgbe_get_completion_timeout
  147. ixgbe_disable_rx
  148. ixgbe_disable_tx
  149. ixgbe_reset
  150. ixgbe_clean_tx_ring
  151. ixgbe_clean_all_rx_rings
  152. ixgbe_clean_all_tx_rings
  153. ixgbe_fdir_filter_exit
  154. ixgbe_down
  155. ixgbe_set_eee_capable
  156. ixgbe_tx_timeout
  157. ixgbe_init_dcb
  158. ixgbe_sw_init
  159. ixgbe_setup_tx_resources
  160. ixgbe_setup_all_tx_resources
  161. ixgbe_setup_rx_resources
  162. ixgbe_setup_all_rx_resources
  163. ixgbe_free_tx_resources
  164. ixgbe_free_all_tx_resources
  165. ixgbe_free_rx_resources
  166. ixgbe_free_all_rx_resources
  167. ixgbe_change_mtu
  168. ixgbe_open
  169. ixgbe_close_suspend
  170. ixgbe_close
  171. ixgbe_resume
  172. __ixgbe_shutdown
  173. ixgbe_suspend
  174. ixgbe_shutdown
  175. ixgbe_update_stats
  176. ixgbe_fdir_reinit_subtask
  177. ixgbe_check_hang_subtask
  178. ixgbe_watchdog_update_link
  179. ixgbe_update_default_up
  180. ixgbe_watchdog_link_is_up
  181. ixgbe_watchdog_link_is_down
  182. ixgbe_ring_tx_pending
  183. ixgbe_vf_tx_pending
  184. ixgbe_watchdog_flush_tx
  185. ixgbe_check_for_bad_vf
  186. ixgbe_spoof_check
  187. ixgbe_spoof_check
  188. ixgbe_check_for_bad_vf
  189. ixgbe_watchdog_subtask
  190. ixgbe_sfp_detection_subtask
  191. ixgbe_sfp_link_config_subtask
  192. ixgbe_service_timer
  193. ixgbe_phy_interrupt_subtask
  194. ixgbe_reset_subtask
  195. ixgbe_check_fw_error
  196. ixgbe_service_task
  197. ixgbe_tso
  198. ixgbe_ipv6_csum_is_sctp
  199. ixgbe_tx_csum
  200. ixgbe_tx_cmd_type
  201. ixgbe_tx_olinfo_status
  202. __ixgbe_maybe_stop_tx
  203. ixgbe_maybe_stop_tx
  204. ixgbe_tx_map
  205. ixgbe_atr
  206. ixgbe_select_queue
  207. ixgbe_xmit_xdp_ring
  208. ixgbe_xmit_frame_ring
  209. __ixgbe_xmit_frame
  210. ixgbe_xmit_frame
  211. ixgbe_set_mac
  212. ixgbe_mdio_read
  213. ixgbe_mdio_write
  214. ixgbe_ioctl
  215. ixgbe_add_sanmac_netdev
  216. ixgbe_del_sanmac_netdev
  217. ixgbe_get_ring_stats64
  218. ixgbe_get_stats64
  219. ixgbe_validate_rtr
  220. ixgbe_set_prio_tc_map
  221. ixgbe_reassign_macvlan_pool
  222. ixgbe_defrag_macvlan_pools
  223. ixgbe_setup_tc
  224. ixgbe_delete_clsu32
  225. ixgbe_configure_clsu32_add_hnode
  226. ixgbe_configure_clsu32_del_hnode
  227. get_macvlan_queue
  228. handle_redirect_action
  229. parse_tc_actions
  230. parse_tc_actions
  231. ixgbe_clsu32_build_input
  232. ixgbe_configure_clsu32
  233. ixgbe_setup_tc_cls_u32
  234. ixgbe_setup_tc_block_cb
  235. ixgbe_setup_tc_mqprio
  236. __ixgbe_setup_tc
  237. ixgbe_sriov_reinit
  238. ixgbe_do_reset
  239. ixgbe_fix_features
  240. ixgbe_reset_l2fw_offload
  241. ixgbe_set_features
  242. ixgbe_add_udp_tunnel_port
  243. ixgbe_del_udp_tunnel_port
  244. ixgbe_ndo_fdb_add
  245. ixgbe_configure_bridge_mode
  246. ixgbe_ndo_bridge_setlink
  247. ixgbe_ndo_bridge_getlink
  248. ixgbe_fwd_add
  249. ixgbe_fwd_del
  250. ixgbe_features_check
  251. ixgbe_xdp_setup
  252. ixgbe_xdp
  253. ixgbe_xdp_ring_update_tail
  254. ixgbe_xdp_xmit
  255. ixgbe_disable_txr_hw
  256. ixgbe_disable_txr
  257. ixgbe_disable_rxr_hw
  258. ixgbe_reset_txr_stats
  259. ixgbe_reset_rxr_stats
  260. ixgbe_txrx_ring_disable
  261. ixgbe_txrx_ring_enable
  262. ixgbe_enumerate_functions
  263. ixgbe_wol_supported
  264. ixgbe_set_fw_version
  265. ixgbe_probe
  266. ixgbe_remove
  267. ixgbe_io_error_detected
  268. ixgbe_io_slot_reset
  269. ixgbe_io_resume
  270. ixgbe_init_module
  271. ixgbe_exit_module
  272. ixgbe_notify_dca

   1 // SPDX-License-Identifier: GPL-2.0
   2 /* Copyright(c) 1999 - 2018 Intel Corporation. */
   3 
   4 #include <linux/types.h>
   5 #include <linux/module.h>
   6 #include <linux/pci.h>
   7 #include <linux/netdevice.h>
   8 #include <linux/vmalloc.h>
   9 #include <linux/string.h>
  10 #include <linux/in.h>
  11 #include <linux/interrupt.h>
  12 #include <linux/ip.h>
  13 #include <linux/tcp.h>
  14 #include <linux/sctp.h>
  15 #include <linux/pkt_sched.h>
  16 #include <linux/ipv6.h>
  17 #include <linux/slab.h>
  18 #include <net/checksum.h>
  19 #include <net/ip6_checksum.h>
  20 #include <linux/etherdevice.h>
  21 #include <linux/ethtool.h>
  22 #include <linux/if.h>
  23 #include <linux/if_vlan.h>
  24 #include <linux/if_macvlan.h>
  25 #include <linux/if_bridge.h>
  26 #include <linux/prefetch.h>
  27 #include <linux/bpf.h>
  28 #include <linux/bpf_trace.h>
  29 #include <linux/atomic.h>
  30 #include <linux/numa.h>
  31 #include <scsi/fc/fc_fcoe.h>
  32 #include <net/udp_tunnel.h>
  33 #include <net/pkt_cls.h>
  34 #include <net/tc_act/tc_gact.h>
  35 #include <net/tc_act/tc_mirred.h>
  36 #include <net/vxlan.h>
  37 #include <net/mpls.h>
  38 #include <net/xdp_sock.h>
  39 #include <net/xfrm.h>
  40 
  41 #include "ixgbe.h"
  42 #include "ixgbe_common.h"
  43 #include "ixgbe_dcb_82599.h"
  44 #include "ixgbe_phy.h"
  45 #include "ixgbe_sriov.h"
  46 #include "ixgbe_model.h"
  47 #include "ixgbe_txrx_common.h"
  48 
  49 char ixgbe_driver_name[] = "ixgbe";
  50 static const char ixgbe_driver_string[] =
  51                               "Intel(R) 10 Gigabit PCI Express Network Driver";
  52 #ifdef IXGBE_FCOE
  53 char ixgbe_default_device_descr[] =
  54                               "Intel(R) 10 Gigabit Network Connection";
  55 #else
  56 static char ixgbe_default_device_descr[] =
  57                               "Intel(R) 10 Gigabit Network Connection";
  58 #endif
  59 #define DRV_VERSION "5.1.0-k"
  60 const char ixgbe_driver_version[] = DRV_VERSION;
  61 static const char ixgbe_copyright[] =
  62                                 "Copyright (c) 1999-2016 Intel Corporation.";
  63 
  64 static const char ixgbe_overheat_msg[] = "Network adapter has been stopped because it has over heated. Restart the computer. If the problem persists, power off the system and replace the adapter";
  65 
  66 static const struct ixgbe_info *ixgbe_info_tbl[] = {
  67         [board_82598]           = &ixgbe_82598_info,
  68         [board_82599]           = &ixgbe_82599_info,
  69         [board_X540]            = &ixgbe_X540_info,
  70         [board_X550]            = &ixgbe_X550_info,
  71         [board_X550EM_x]        = &ixgbe_X550EM_x_info,
  72         [board_x550em_x_fw]     = &ixgbe_x550em_x_fw_info,
  73         [board_x550em_a]        = &ixgbe_x550em_a_info,
  74         [board_x550em_a_fw]     = &ixgbe_x550em_a_fw_info,
  75 };
  76 
  77 /* ixgbe_pci_tbl - PCI Device ID Table
  78  *
  79  * Wildcard entries (PCI_ANY_ID) should come last
  80  * Last entry must be all 0s
  81  *
  82  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
  83  *   Class, Class Mask, private data (not used) }
  84  */
  85 static const struct pci_device_id ixgbe_pci_tbl[] = {
  86         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
  87         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
  88         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
  89         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
  90         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
  91         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
  92         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
  93         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
  94         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
  95         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
  96         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
  97         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
  98         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
  99         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
 100         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
 101         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
 102         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
 103         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
 104         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
 105         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
 106         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
 107         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
 108         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
 109         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
 110         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
 111         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
 112         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
 113         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
 114         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
 115         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
 116         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
 117         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T1), board_X550},
 118         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
 119         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_XFI), board_X550EM_x},
 120         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
 121         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
 122         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_SFP), board_X550EM_x},
 123         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_1G_T), board_x550em_x_fw},
 124         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR), board_x550em_a },
 125         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR_L), board_x550em_a },
 126         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP_N), board_x550em_a },
 127         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII), board_x550em_a },
 128         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII_L), board_x550em_a },
 129         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_10G_T), board_x550em_a},
 130         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP), board_x550em_a },
 131         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T), board_x550em_a_fw },
 132         {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T_L), board_x550em_a_fw },
 133         /* required last entry */
 134         {0, }
 135 };
 136 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
 137 
 138 #ifdef CONFIG_IXGBE_DCA
 139 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
 140                             void *p);
 141 static struct notifier_block dca_notifier = {
 142         .notifier_call = ixgbe_notify_dca,
 143         .next          = NULL,
 144         .priority      = 0
 145 };
 146 #endif
 147 
 148 #ifdef CONFIG_PCI_IOV
 149 static unsigned int max_vfs;
 150 module_param(max_vfs, uint, 0);
 151 MODULE_PARM_DESC(max_vfs,
 152                  "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
 153 #endif /* CONFIG_PCI_IOV */
 154 
 155 static unsigned int allow_unsupported_sfp;
 156 module_param(allow_unsupported_sfp, uint, 0);
 157 MODULE_PARM_DESC(allow_unsupported_sfp,
 158                  "Allow unsupported and untested SFP+ modules on 82599-based adapters");
 159 
 160 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
 161 static int debug = -1;
 162 module_param(debug, int, 0);
 163 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
 164 
 165 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
 166 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
 167 MODULE_LICENSE("GPL v2");
 168 MODULE_VERSION(DRV_VERSION);
 169 
 170 static struct workqueue_struct *ixgbe_wq;
 171 
 172 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
 173 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *);
 174 
 175 static const struct net_device_ops ixgbe_netdev_ops;
 176 
 177 static bool netif_is_ixgbe(struct net_device *dev)
 178 {
 179         return dev && (dev->netdev_ops == &ixgbe_netdev_ops);
 180 }
 181 
 182 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
 183                                           u32 reg, u16 *value)
 184 {
 185         struct pci_dev *parent_dev;
 186         struct pci_bus *parent_bus;
 187 
 188         parent_bus = adapter->pdev->bus->parent;
 189         if (!parent_bus)
 190                 return -1;
 191 
 192         parent_dev = parent_bus->self;
 193         if (!parent_dev)
 194                 return -1;
 195 
 196         if (!pci_is_pcie(parent_dev))
 197                 return -1;
 198 
 199         pcie_capability_read_word(parent_dev, reg, value);
 200         if (*value == IXGBE_FAILED_READ_CFG_WORD &&
 201             ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
 202                 return -1;
 203         return 0;
 204 }
 205 
 206 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
 207 {
 208         struct ixgbe_hw *hw = &adapter->hw;
 209         u16 link_status = 0;
 210         int err;
 211 
 212         hw->bus.type = ixgbe_bus_type_pci_express;
 213 
 214         /* Get the negotiated link width and speed from PCI config space of the
 215          * parent, as this device is behind a switch
 216          */
 217         err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
 218 
 219         /* assume caller will handle error case */
 220         if (err)
 221                 return err;
 222 
 223         hw->bus.width = ixgbe_convert_bus_width(link_status);
 224         hw->bus.speed = ixgbe_convert_bus_speed(link_status);
 225 
 226         return 0;
 227 }
 228 
 229 /**
 230  * ixgbe_check_from_parent - Determine whether PCIe info should come from parent
 231  * @hw: hw specific details
 232  *
 233  * This function is used by probe to determine whether a device's PCI-Express
 234  * bandwidth details should be gathered from the parent bus instead of from the
 235  * device. Used to ensure that various locations all have the correct device ID
 236  * checks.
 237  */
 238 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
 239 {
 240         switch (hw->device_id) {
 241         case IXGBE_DEV_ID_82599_SFP_SF_QP:
 242         case IXGBE_DEV_ID_82599_QSFP_SF_QP:
 243                 return true;
 244         default:
 245                 return false;
 246         }
 247 }
 248 
 249 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
 250                                      int expected_gts)
 251 {
 252         struct ixgbe_hw *hw = &adapter->hw;
 253         struct pci_dev *pdev;
 254 
 255         /* Some devices are not connected over PCIe and thus do not negotiate
 256          * speed. These devices do not have valid bus info, and thus any report
 257          * we generate may not be correct.
 258          */
 259         if (hw->bus.type == ixgbe_bus_type_internal)
 260                 return;
 261 
 262         /* determine whether to use the parent device */
 263         if (ixgbe_pcie_from_parent(&adapter->hw))
 264                 pdev = adapter->pdev->bus->parent->self;
 265         else
 266                 pdev = adapter->pdev;
 267 
 268         pcie_print_link_status(pdev);
 269 }
 270 
 271 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
 272 {
 273         if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
 274             !test_bit(__IXGBE_REMOVING, &adapter->state) &&
 275             !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
 276                 queue_work(ixgbe_wq, &adapter->service_task);
 277 }
 278 
 279 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
 280 {
 281         struct ixgbe_adapter *adapter = hw->back;
 282 
 283         if (!hw->hw_addr)
 284                 return;
 285         hw->hw_addr = NULL;
 286         e_dev_err("Adapter removed\n");
 287         if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
 288                 ixgbe_service_event_schedule(adapter);
 289 }
 290 
 291 static u32 ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
 292 {
 293         u8 __iomem *reg_addr;
 294         u32 value;
 295         int i;
 296 
 297         reg_addr = READ_ONCE(hw->hw_addr);
 298         if (ixgbe_removed(reg_addr))
 299                 return IXGBE_FAILED_READ_REG;
 300 
 301         /* Register read of 0xFFFFFFF can indicate the adapter has been removed,
 302          * so perform several status register reads to determine if the adapter
 303          * has been removed.
 304          */
 305         for (i = 0; i < IXGBE_FAILED_READ_RETRIES; i++) {
 306                 value = readl(reg_addr + IXGBE_STATUS);
 307                 if (value != IXGBE_FAILED_READ_REG)
 308                         break;
 309                 mdelay(3);
 310         }
 311 
 312         if (value == IXGBE_FAILED_READ_REG)
 313                 ixgbe_remove_adapter(hw);
 314         else
 315                 value = readl(reg_addr + reg);
 316         return value;
 317 }
 318 
 319 /**
 320  * ixgbe_read_reg - Read from device register
 321  * @hw: hw specific details
 322  * @reg: offset of register to read
 323  *
 324  * Returns : value read or IXGBE_FAILED_READ_REG if removed
 325  *
 326  * This function is used to read device registers. It checks for device
 327  * removal by confirming any read that returns all ones by checking the
 328  * status register value for all ones. This function avoids reading from
 329  * the hardware if a removal was previously detected in which case it
 330  * returns IXGBE_FAILED_READ_REG (all ones).
 331  */
 332 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
 333 {
 334         u8 __iomem *reg_addr = READ_ONCE(hw->hw_addr);
 335         u32 value;
 336 
 337         if (ixgbe_removed(reg_addr))
 338                 return IXGBE_FAILED_READ_REG;
 339         if (unlikely(hw->phy.nw_mng_if_sel &
 340                      IXGBE_NW_MNG_IF_SEL_SGMII_ENABLE)) {
 341                 struct ixgbe_adapter *adapter;
 342                 int i;
 343 
 344                 for (i = 0; i < 200; ++i) {
 345                         value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
 346                         if (likely(!value))
 347                                 goto writes_completed;
 348                         if (value == IXGBE_FAILED_READ_REG) {
 349                                 ixgbe_remove_adapter(hw);
 350                                 return IXGBE_FAILED_READ_REG;
 351                         }
 352                         udelay(5);
 353                 }
 354 
 355                 adapter = hw->back;
 356                 e_warn(hw, "register writes incomplete %08x\n", value);
 357         }
 358 
 359 writes_completed:
 360         value = readl(reg_addr + reg);
 361         if (unlikely(value == IXGBE_FAILED_READ_REG))
 362                 value = ixgbe_check_remove(hw, reg);
 363         return value;
 364 }
 365 
 366 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
 367 {
 368         u16 value;
 369 
 370         pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
 371         if (value == IXGBE_FAILED_READ_CFG_WORD) {
 372                 ixgbe_remove_adapter(hw);
 373                 return true;
 374         }
 375         return false;
 376 }
 377 
 378 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
 379 {
 380         struct ixgbe_adapter *adapter = hw->back;
 381         u16 value;
 382 
 383         if (ixgbe_removed(hw->hw_addr))
 384                 return IXGBE_FAILED_READ_CFG_WORD;
 385         pci_read_config_word(adapter->pdev, reg, &value);
 386         if (value == IXGBE_FAILED_READ_CFG_WORD &&
 387             ixgbe_check_cfg_remove(hw, adapter->pdev))
 388                 return IXGBE_FAILED_READ_CFG_WORD;
 389         return value;
 390 }
 391 
 392 #ifdef CONFIG_PCI_IOV
 393 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
 394 {
 395         struct ixgbe_adapter *adapter = hw->back;
 396         u32 value;
 397 
 398         if (ixgbe_removed(hw->hw_addr))
 399                 return IXGBE_FAILED_READ_CFG_DWORD;
 400         pci_read_config_dword(adapter->pdev, reg, &value);
 401         if (value == IXGBE_FAILED_READ_CFG_DWORD &&
 402             ixgbe_check_cfg_remove(hw, adapter->pdev))
 403                 return IXGBE_FAILED_READ_CFG_DWORD;
 404         return value;
 405 }
 406 #endif /* CONFIG_PCI_IOV */
 407 
 408 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
 409 {
 410         struct ixgbe_adapter *adapter = hw->back;
 411 
 412         if (ixgbe_removed(hw->hw_addr))
 413                 return;
 414         pci_write_config_word(adapter->pdev, reg, value);
 415 }
 416 
 417 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
 418 {
 419         BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
 420 
 421         /* flush memory to make sure state is correct before next watchdog */
 422         smp_mb__before_atomic();
 423         clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
 424 }
 425 
 426 struct ixgbe_reg_info {
 427         u32 ofs;
 428         char *name;
 429 };
 430 
 431 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
 432 
 433         /* General Registers */
 434         {IXGBE_CTRL, "CTRL"},
 435         {IXGBE_STATUS, "STATUS"},
 436         {IXGBE_CTRL_EXT, "CTRL_EXT"},
 437 
 438         /* Interrupt Registers */
 439         {IXGBE_EICR, "EICR"},
 440 
 441         /* RX Registers */
 442         {IXGBE_SRRCTL(0), "SRRCTL"},
 443         {IXGBE_DCA_RXCTRL(0), "DRXCTL"},
 444         {IXGBE_RDLEN(0), "RDLEN"},
 445         {IXGBE_RDH(0), "RDH"},
 446         {IXGBE_RDT(0), "RDT"},
 447         {IXGBE_RXDCTL(0), "RXDCTL"},
 448         {IXGBE_RDBAL(0), "RDBAL"},
 449         {IXGBE_RDBAH(0), "RDBAH"},
 450 
 451         /* TX Registers */
 452         {IXGBE_TDBAL(0), "TDBAL"},
 453         {IXGBE_TDBAH(0), "TDBAH"},
 454         {IXGBE_TDLEN(0), "TDLEN"},
 455         {IXGBE_TDH(0), "TDH"},
 456         {IXGBE_TDT(0), "TDT"},
 457         {IXGBE_TXDCTL(0), "TXDCTL"},
 458 
 459         /* List Terminator */
 460         { .name = NULL }
 461 };
 462 
 463 
 464 /*
 465  * ixgbe_regdump - register printout routine
 466  */
 467 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
 468 {
 469         int i;
 470         char rname[16];
 471         u32 regs[64];
 472 
 473         switch (reginfo->ofs) {
 474         case IXGBE_SRRCTL(0):
 475                 for (i = 0; i < 64; i++)
 476                         regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
 477                 break;
 478         case IXGBE_DCA_RXCTRL(0):
 479                 for (i = 0; i < 64; i++)
 480                         regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
 481                 break;
 482         case IXGBE_RDLEN(0):
 483                 for (i = 0; i < 64; i++)
 484                         regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
 485                 break;
 486         case IXGBE_RDH(0):
 487                 for (i = 0; i < 64; i++)
 488                         regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
 489                 break;
 490         case IXGBE_RDT(0):
 491                 for (i = 0; i < 64; i++)
 492                         regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
 493                 break;
 494         case IXGBE_RXDCTL(0):
 495                 for (i = 0; i < 64; i++)
 496                         regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
 497                 break;
 498         case IXGBE_RDBAL(0):
 499                 for (i = 0; i < 64; i++)
 500                         regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
 501                 break;
 502         case IXGBE_RDBAH(0):
 503                 for (i = 0; i < 64; i++)
 504                         regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
 505                 break;
 506         case IXGBE_TDBAL(0):
 507                 for (i = 0; i < 64; i++)
 508                         regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
 509                 break;
 510         case IXGBE_TDBAH(0):
 511                 for (i = 0; i < 64; i++)
 512                         regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
 513                 break;
 514         case IXGBE_TDLEN(0):
 515                 for (i = 0; i < 64; i++)
 516                         regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
 517                 break;
 518         case IXGBE_TDH(0):
 519                 for (i = 0; i < 64; i++)
 520                         regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
 521                 break;
 522         case IXGBE_TDT(0):
 523                 for (i = 0; i < 64; i++)
 524                         regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
 525                 break;
 526         case IXGBE_TXDCTL(0):
 527                 for (i = 0; i < 64; i++)
 528                         regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
 529                 break;
 530         default:
 531                 pr_info("%-15s %08x\n",
 532                         reginfo->name, IXGBE_READ_REG(hw, reginfo->ofs));
 533                 return;
 534         }
 535 
 536         i = 0;
 537         while (i < 64) {
 538                 int j;
 539                 char buf[9 * 8 + 1];
 540                 char *p = buf;
 541 
 542                 snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i, i + 7);
 543                 for (j = 0; j < 8; j++)
 544                         p += sprintf(p, " %08x", regs[i++]);
 545                 pr_err("%-15s%s\n", rname, buf);
 546         }
 547 
 548 }
 549 
 550 static void ixgbe_print_buffer(struct ixgbe_ring *ring, int n)
 551 {
 552         struct ixgbe_tx_buffer *tx_buffer;
 553 
 554         tx_buffer = &ring->tx_buffer_info[ring->next_to_clean];
 555         pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
 556                 n, ring->next_to_use, ring->next_to_clean,
 557                 (u64)dma_unmap_addr(tx_buffer, dma),
 558                 dma_unmap_len(tx_buffer, len),
 559                 tx_buffer->next_to_watch,
 560                 (u64)tx_buffer->time_stamp);
 561 }
 562 
 563 /*
 564  * ixgbe_dump - Print registers, tx-rings and rx-rings
 565  */
 566 static void ixgbe_dump(struct ixgbe_adapter *adapter)
 567 {
 568         struct net_device *netdev = adapter->netdev;
 569         struct ixgbe_hw *hw = &adapter->hw;
 570         struct ixgbe_reg_info *reginfo;
 571         int n = 0;
 572         struct ixgbe_ring *ring;
 573         struct ixgbe_tx_buffer *tx_buffer;
 574         union ixgbe_adv_tx_desc *tx_desc;
 575         struct my_u0 { u64 a; u64 b; } *u0;
 576         struct ixgbe_ring *rx_ring;
 577         union ixgbe_adv_rx_desc *rx_desc;
 578         struct ixgbe_rx_buffer *rx_buffer_info;
 579         int i = 0;
 580 
 581         if (!netif_msg_hw(adapter))
 582                 return;
 583 
 584         /* Print netdevice Info */
 585         if (netdev) {
 586                 dev_info(&adapter->pdev->dev, "Net device Info\n");
 587                 pr_info("Device Name     state            "
 588                         "trans_start\n");
 589                 pr_info("%-15s %016lX %016lX\n",
 590                         netdev->name,
 591                         netdev->state,
 592                         dev_trans_start(netdev));
 593         }
 594 
 595         /* Print Registers */
 596         dev_info(&adapter->pdev->dev, "Register Dump\n");
 597         pr_info(" Register Name   Value\n");
 598         for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
 599              reginfo->name; reginfo++) {
 600                 ixgbe_regdump(hw, reginfo);
 601         }
 602 
 603         /* Print TX Ring Summary */
 604         if (!netdev || !netif_running(netdev))
 605                 return;
 606 
 607         dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
 608         pr_info(" %s     %s              %s        %s\n",
 609                 "Queue [NTU] [NTC] [bi(ntc)->dma  ]",
 610                 "leng", "ntw", "timestamp");
 611         for (n = 0; n < adapter->num_tx_queues; n++) {
 612                 ring = adapter->tx_ring[n];
 613                 ixgbe_print_buffer(ring, n);
 614         }
 615 
 616         for (n = 0; n < adapter->num_xdp_queues; n++) {
 617                 ring = adapter->xdp_ring[n];
 618                 ixgbe_print_buffer(ring, n);
 619         }
 620 
 621         /* Print TX Rings */
 622         if (!netif_msg_tx_done(adapter))
 623                 goto rx_ring_summary;
 624 
 625         dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
 626 
 627         /* Transmit Descriptor Formats
 628          *
 629          * 82598 Advanced Transmit Descriptor
 630          *   +--------------------------------------------------------------+
 631          * 0 |         Buffer Address [63:0]                                |
 632          *   +--------------------------------------------------------------+
 633          * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
 634          *   +--------------------------------------------------------------+
 635          *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
 636          *
 637          * 82598 Advanced Transmit Descriptor (Write-Back Format)
 638          *   +--------------------------------------------------------------+
 639          * 0 |                          RSV [63:0]                          |
 640          *   +--------------------------------------------------------------+
 641          * 8 |            RSV           |  STA  |          NXTSEQ           |
 642          *   +--------------------------------------------------------------+
 643          *   63                       36 35   32 31                         0
 644          *
 645          * 82599+ Advanced Transmit Descriptor
 646          *   +--------------------------------------------------------------+
 647          * 0 |         Buffer Address [63:0]                                |
 648          *   +--------------------------------------------------------------+
 649          * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
 650          *   +--------------------------------------------------------------+
 651          *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
 652          *
 653          * 82599+ Advanced Transmit Descriptor (Write-Back Format)
 654          *   +--------------------------------------------------------------+
 655          * 0 |                          RSV [63:0]                          |
 656          *   +--------------------------------------------------------------+
 657          * 8 |            RSV           |  STA  |           RSV             |
 658          *   +--------------------------------------------------------------+
 659          *   63                       36 35   32 31                         0
 660          */
 661 
 662         for (n = 0; n < adapter->num_tx_queues; n++) {
 663                 ring = adapter->tx_ring[n];
 664                 pr_info("------------------------------------\n");
 665                 pr_info("TX QUEUE INDEX = %d\n", ring->queue_index);
 666                 pr_info("------------------------------------\n");
 667                 pr_info("%s%s    %s              %s        %s          %s\n",
 668                         "T [desc]     [address 63:0  ] ",
 669                         "[PlPOIdStDDt Ln] [bi->dma       ] ",
 670                         "leng", "ntw", "timestamp", "bi->skb");
 671 
 672                 for (i = 0; ring->desc && (i < ring->count); i++) {
 673                         tx_desc = IXGBE_TX_DESC(ring, i);
 674                         tx_buffer = &ring->tx_buffer_info[i];
 675                         u0 = (struct my_u0 *)tx_desc;
 676                         if (dma_unmap_len(tx_buffer, len) > 0) {
 677                                 const char *ring_desc;
 678 
 679                                 if (i == ring->next_to_use &&
 680                                     i == ring->next_to_clean)
 681                                         ring_desc = " NTC/U";
 682                                 else if (i == ring->next_to_use)
 683                                         ring_desc = " NTU";
 684                                 else if (i == ring->next_to_clean)
 685                                         ring_desc = " NTC";
 686                                 else
 687                                         ring_desc = "";
 688                                 pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p%s",
 689                                         i,
 690                                         le64_to_cpu((__force __le64)u0->a),
 691                                         le64_to_cpu((__force __le64)u0->b),
 692                                         (u64)dma_unmap_addr(tx_buffer, dma),
 693                                         dma_unmap_len(tx_buffer, len),
 694                                         tx_buffer->next_to_watch,
 695                                         (u64)tx_buffer->time_stamp,
 696                                         tx_buffer->skb,
 697                                         ring_desc);
 698 
 699                                 if (netif_msg_pktdata(adapter) &&
 700                                     tx_buffer->skb)
 701                                         print_hex_dump(KERN_INFO, "",
 702                                                 DUMP_PREFIX_ADDRESS, 16, 1,
 703                                                 tx_buffer->skb->data,
 704                                                 dma_unmap_len(tx_buffer, len),
 705                                                 true);
 706                         }
 707                 }
 708         }
 709 
 710         /* Print RX Rings Summary */
 711 rx_ring_summary:
 712         dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
 713         pr_info("Queue [NTU] [NTC]\n");
 714         for (n = 0; n < adapter->num_rx_queues; n++) {
 715                 rx_ring = adapter->rx_ring[n];
 716                 pr_info("%5d %5X %5X\n",
 717                         n, rx_ring->next_to_use, rx_ring->next_to_clean);
 718         }
 719 
 720         /* Print RX Rings */
 721         if (!netif_msg_rx_status(adapter))
 722                 return;
 723 
 724         dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
 725 
 726         /* Receive Descriptor Formats
 727          *
 728          * 82598 Advanced Receive Descriptor (Read) Format
 729          *    63                                           1        0
 730          *    +-----------------------------------------------------+
 731          *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
 732          *    +----------------------------------------------+------+
 733          *  8 |       Header Buffer Address [63:1]           |  DD  |
 734          *    +-----------------------------------------------------+
 735          *
 736          *
 737          * 82598 Advanced Receive Descriptor (Write-Back) Format
 738          *
 739          *   63       48 47    32 31  30      21 20 16 15   4 3     0
 740          *   +------------------------------------------------------+
 741          * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
 742          *   | Packet   | IP     |   |          |     | Type | Type |
 743          *   | Checksum | Ident  |   |          |     |      |      |
 744          *   +------------------------------------------------------+
 745          * 8 | VLAN Tag | Length | Extended Error | Extended Status |
 746          *   +------------------------------------------------------+
 747          *   63       48 47    32 31            20 19               0
 748          *
 749          * 82599+ Advanced Receive Descriptor (Read) Format
 750          *    63                                           1        0
 751          *    +-----------------------------------------------------+
 752          *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
 753          *    +----------------------------------------------+------+
 754          *  8 |       Header Buffer Address [63:1]           |  DD  |
 755          *    +-----------------------------------------------------+
 756          *
 757          *
 758          * 82599+ Advanced Receive Descriptor (Write-Back) Format
 759          *
 760          *   63       48 47    32 31  30      21 20 17 16   4 3     0
 761          *   +------------------------------------------------------+
 762          * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
 763          *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
 764          *   |/ Flow Dir Flt ID  |   |          |     |      |      |
 765          *   +------------------------------------------------------+
 766          * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
 767          *   +------------------------------------------------------+
 768          *   63       48 47    32 31          20 19                 0
 769          */
 770 
 771         for (n = 0; n < adapter->num_rx_queues; n++) {
 772                 rx_ring = adapter->rx_ring[n];
 773                 pr_info("------------------------------------\n");
 774                 pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
 775                 pr_info("------------------------------------\n");
 776                 pr_info("%s%s%s\n",
 777                         "R  [desc]      [ PktBuf     A0] ",
 778                         "[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
 779                         "<-- Adv Rx Read format");
 780                 pr_info("%s%s%s\n",
 781                         "RWB[desc]      [PcsmIpSHl PtRs] ",
 782                         "[vl er S cks ln] ---------------- [bi->skb       ] ",
 783                         "<-- Adv Rx Write-Back format");
 784 
 785                 for (i = 0; i < rx_ring->count; i++) {
 786                         const char *ring_desc;
 787 
 788                         if (i == rx_ring->next_to_use)
 789                                 ring_desc = " NTU";
 790                         else if (i == rx_ring->next_to_clean)
 791                                 ring_desc = " NTC";
 792                         else
 793                                 ring_desc = "";
 794 
 795                         rx_buffer_info = &rx_ring->rx_buffer_info[i];
 796                         rx_desc = IXGBE_RX_DESC(rx_ring, i);
 797                         u0 = (struct my_u0 *)rx_desc;
 798                         if (rx_desc->wb.upper.length) {
 799                                 /* Descriptor Done */
 800                                 pr_info("RWB[0x%03X]     %016llX %016llX ---------------- %p%s\n",
 801                                         i,
 802                                         le64_to_cpu((__force __le64)u0->a),
 803                                         le64_to_cpu((__force __le64)u0->b),
 804                                         rx_buffer_info->skb,
 805                                         ring_desc);
 806                         } else {
 807                                 pr_info("R  [0x%03X]     %016llX %016llX %016llX %p%s\n",
 808                                         i,
 809                                         le64_to_cpu((__force __le64)u0->a),
 810                                         le64_to_cpu((__force __le64)u0->b),
 811                                         (u64)rx_buffer_info->dma,
 812                                         rx_buffer_info->skb,
 813                                         ring_desc);
 814 
 815                                 if (netif_msg_pktdata(adapter) &&
 816                                     rx_buffer_info->dma) {
 817                                         print_hex_dump(KERN_INFO, "",
 818                                            DUMP_PREFIX_ADDRESS, 16, 1,
 819                                            page_address(rx_buffer_info->page) +
 820                                                     rx_buffer_info->page_offset,
 821                                            ixgbe_rx_bufsz(rx_ring), true);
 822                                 }
 823                         }
 824                 }
 825         }
 826 }
 827 
 828 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
 829 {
 830         u32 ctrl_ext;
 831 
 832         /* Let firmware take over control of h/w */
 833         ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
 834         IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
 835                         ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
 836 }
 837 
 838 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
 839 {
 840         u32 ctrl_ext;
 841 
 842         /* Let firmware know the driver has taken over */
 843         ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
 844         IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
 845                         ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
 846 }
 847 
 848 /**
 849  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
 850  * @adapter: pointer to adapter struct
 851  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
 852  * @queue: queue to map the corresponding interrupt to
 853  * @msix_vector: the vector to map to the corresponding queue
 854  *
 855  */
 856 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
 857                            u8 queue, u8 msix_vector)
 858 {
 859         u32 ivar, index;
 860         struct ixgbe_hw *hw = &adapter->hw;
 861         switch (hw->mac.type) {
 862         case ixgbe_mac_82598EB:
 863                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
 864                 if (direction == -1)
 865                         direction = 0;
 866                 index = (((direction * 64) + queue) >> 2) & 0x1F;
 867                 ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
 868                 ivar &= ~(0xFF << (8 * (queue & 0x3)));
 869                 ivar |= (msix_vector << (8 * (queue & 0x3)));
 870                 IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
 871                 break;
 872         case ixgbe_mac_82599EB:
 873         case ixgbe_mac_X540:
 874         case ixgbe_mac_X550:
 875         case ixgbe_mac_X550EM_x:
 876         case ixgbe_mac_x550em_a:
 877                 if (direction == -1) {
 878                         /* other causes */
 879                         msix_vector |= IXGBE_IVAR_ALLOC_VAL;
 880                         index = ((queue & 1) * 8);
 881                         ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
 882                         ivar &= ~(0xFF << index);
 883                         ivar |= (msix_vector << index);
 884                         IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
 885                         break;
 886                 } else {
 887                         /* tx or rx causes */
 888                         msix_vector |= IXGBE_IVAR_ALLOC_VAL;
 889                         index = ((16 * (queue & 1)) + (8 * direction));
 890                         ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
 891                         ivar &= ~(0xFF << index);
 892                         ivar |= (msix_vector << index);
 893                         IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
 894                         break;
 895                 }
 896         default:
 897                 break;
 898         }
 899 }
 900 
 901 void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
 902                             u64 qmask)
 903 {
 904         u32 mask;
 905 
 906         switch (adapter->hw.mac.type) {
 907         case ixgbe_mac_82598EB:
 908                 mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
 909                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
 910                 break;
 911         case ixgbe_mac_82599EB:
 912         case ixgbe_mac_X540:
 913         case ixgbe_mac_X550:
 914         case ixgbe_mac_X550EM_x:
 915         case ixgbe_mac_x550em_a:
 916                 mask = (qmask & 0xFFFFFFFF);
 917                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
 918                 mask = (qmask >> 32);
 919                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
 920                 break;
 921         default:
 922                 break;
 923         }
 924 }
 925 
 926 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
 927 {
 928         struct ixgbe_hw *hw = &adapter->hw;
 929         struct ixgbe_hw_stats *hwstats = &adapter->stats;
 930         int i;
 931         u32 data;
 932 
 933         if ((hw->fc.current_mode != ixgbe_fc_full) &&
 934             (hw->fc.current_mode != ixgbe_fc_rx_pause))
 935                 return;
 936 
 937         switch (hw->mac.type) {
 938         case ixgbe_mac_82598EB:
 939                 data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
 940                 break;
 941         default:
 942                 data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
 943         }
 944         hwstats->lxoffrxc += data;
 945 
 946         /* refill credits (no tx hang) if we received xoff */
 947         if (!data)
 948                 return;
 949 
 950         for (i = 0; i < adapter->num_tx_queues; i++)
 951                 clear_bit(__IXGBE_HANG_CHECK_ARMED,
 952                           &adapter->tx_ring[i]->state);
 953 
 954         for (i = 0; i < adapter->num_xdp_queues; i++)
 955                 clear_bit(__IXGBE_HANG_CHECK_ARMED,
 956                           &adapter->xdp_ring[i]->state);
 957 }
 958 
 959 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
 960 {
 961         struct ixgbe_hw *hw = &adapter->hw;
 962         struct ixgbe_hw_stats *hwstats = &adapter->stats;
 963         u32 xoff[8] = {0};
 964         u8 tc;
 965         int i;
 966         bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
 967 
 968         if (adapter->ixgbe_ieee_pfc)
 969                 pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
 970 
 971         if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
 972                 ixgbe_update_xoff_rx_lfc(adapter);
 973                 return;
 974         }
 975 
 976         /* update stats for each tc, only valid with PFC enabled */
 977         for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
 978                 u32 pxoffrxc;
 979 
 980                 switch (hw->mac.type) {
 981                 case ixgbe_mac_82598EB:
 982                         pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
 983                         break;
 984                 default:
 985                         pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
 986                 }
 987                 hwstats->pxoffrxc[i] += pxoffrxc;
 988                 /* Get the TC for given UP */
 989                 tc = netdev_get_prio_tc_map(adapter->netdev, i);
 990                 xoff[tc] += pxoffrxc;
 991         }
 992 
 993         /* disarm tx queues that have received xoff frames */
 994         for (i = 0; i < adapter->num_tx_queues; i++) {
 995                 struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
 996 
 997                 tc = tx_ring->dcb_tc;
 998                 if (xoff[tc])
 999                         clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1000         }
1001 
1002         for (i = 0; i < adapter->num_xdp_queues; i++) {
1003                 struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
1004 
1005                 tc = xdp_ring->dcb_tc;
1006                 if (xoff[tc])
1007                         clear_bit(__IXGBE_HANG_CHECK_ARMED, &xdp_ring->state);
1008         }
1009 }
1010 
1011 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1012 {
1013         return ring->stats.packets;
1014 }
1015 
1016 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1017 {
1018         unsigned int head, tail;
1019 
1020         head = ring->next_to_clean;
1021         tail = ring->next_to_use;
1022 
1023         return ((head <= tail) ? tail : tail + ring->count) - head;
1024 }
1025 
1026 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1027 {
1028         u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1029         u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1030         u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1031 
1032         clear_check_for_tx_hang(tx_ring);
1033 
1034         /*
1035          * Check for a hung queue, but be thorough. This verifies
1036          * that a transmit has been completed since the previous
1037          * check AND there is at least one packet pending. The
1038          * ARMED bit is set to indicate a potential hang. The
1039          * bit is cleared if a pause frame is received to remove
1040          * false hang detection due to PFC or 802.3x frames. By
1041          * requiring this to fail twice we avoid races with
1042          * pfc clearing the ARMED bit and conditions where we
1043          * run the check_tx_hang logic with a transmit completion
1044          * pending but without time to complete it yet.
1045          */
1046         if (tx_done_old == tx_done && tx_pending)
1047                 /* make sure it is true for two checks in a row */
1048                 return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1049                                         &tx_ring->state);
1050         /* update completed stats and continue */
1051         tx_ring->tx_stats.tx_done_old = tx_done;
1052         /* reset the countdown */
1053         clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1054 
1055         return false;
1056 }
1057 
1058 /**
1059  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1060  * @adapter: driver private struct
1061  **/
1062 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1063 {
1064 
1065         /* Do the reset outside of interrupt context */
1066         if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1067                 set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1068                 e_warn(drv, "initiating reset due to tx timeout\n");
1069                 ixgbe_service_event_schedule(adapter);
1070         }
1071 }
1072 
1073 /**
1074  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1075  * @netdev: network interface device structure
1076  * @queue_index: Tx queue to set
1077  * @maxrate: desired maximum transmit bitrate
1078  **/
1079 static int ixgbe_tx_maxrate(struct net_device *netdev,
1080                             int queue_index, u32 maxrate)
1081 {
1082         struct ixgbe_adapter *adapter = netdev_priv(netdev);
1083         struct ixgbe_hw *hw = &adapter->hw;
1084         u32 bcnrc_val = ixgbe_link_mbps(adapter);
1085 
1086         if (!maxrate)
1087                 return 0;
1088 
1089         /* Calculate the rate factor values to set */
1090         bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1091         bcnrc_val /= maxrate;
1092 
1093         /* clear everything but the rate factor */
1094         bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1095         IXGBE_RTTBCNRC_RF_DEC_MASK;
1096 
1097         /* enable the rate scheduler */
1098         bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1099 
1100         IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1101         IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1102 
1103         return 0;
1104 }
1105 
1106 /**
1107  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1108  * @q_vector: structure containing interrupt and ring information
1109  * @tx_ring: tx ring to clean
1110  * @napi_budget: Used to determine if we are in netpoll
1111  **/
1112 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1113                                struct ixgbe_ring *tx_ring, int napi_budget)
1114 {
1115         struct ixgbe_adapter *adapter = q_vector->adapter;
1116         struct ixgbe_tx_buffer *tx_buffer;
1117         union ixgbe_adv_tx_desc *tx_desc;
1118         unsigned int total_bytes = 0, total_packets = 0, total_ipsec = 0;
1119         unsigned int budget = q_vector->tx.work_limit;
1120         unsigned int i = tx_ring->next_to_clean;
1121 
1122         if (test_bit(__IXGBE_DOWN, &adapter->state))
1123                 return true;
1124 
1125         tx_buffer = &tx_ring->tx_buffer_info[i];
1126         tx_desc = IXGBE_TX_DESC(tx_ring, i);
1127         i -= tx_ring->count;
1128 
1129         do {
1130                 union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1131 
1132                 /* if next_to_watch is not set then there is no work pending */
1133                 if (!eop_desc)
1134                         break;
1135 
1136                 /* prevent any other reads prior to eop_desc */
1137                 smp_rmb();
1138 
1139                 /* if DD is not set pending work has not been completed */
1140                 if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1141                         break;
1142 
1143                 /* clear next_to_watch to prevent false hangs */
1144                 tx_buffer->next_to_watch = NULL;
1145 
1146                 /* update the statistics for this packet */
1147                 total_bytes += tx_buffer->bytecount;
1148                 total_packets += tx_buffer->gso_segs;
1149                 if (tx_buffer->tx_flags & IXGBE_TX_FLAGS_IPSEC)
1150                         total_ipsec++;
1151 
1152                 /* free the skb */
1153                 if (ring_is_xdp(tx_ring))
1154                         xdp_return_frame(tx_buffer->xdpf);
1155                 else
1156                         napi_consume_skb(tx_buffer->skb, napi_budget);
1157 
1158                 /* unmap skb header data */
1159                 dma_unmap_single(tx_ring->dev,
1160                                  dma_unmap_addr(tx_buffer, dma),
1161                                  dma_unmap_len(tx_buffer, len),
1162                                  DMA_TO_DEVICE);
1163 
1164                 /* clear tx_buffer data */
1165                 dma_unmap_len_set(tx_buffer, len, 0);
1166 
1167                 /* unmap remaining buffers */
1168                 while (tx_desc != eop_desc) {
1169                         tx_buffer++;
1170                         tx_desc++;
1171                         i++;
1172                         if (unlikely(!i)) {
1173                                 i -= tx_ring->count;
1174                                 tx_buffer = tx_ring->tx_buffer_info;
1175                                 tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1176                         }
1177 
1178                         /* unmap any remaining paged data */
1179                         if (dma_unmap_len(tx_buffer, len)) {
1180                                 dma_unmap_page(tx_ring->dev,
1181                                                dma_unmap_addr(tx_buffer, dma),
1182                                                dma_unmap_len(tx_buffer, len),
1183                                                DMA_TO_DEVICE);
1184                                 dma_unmap_len_set(tx_buffer, len, 0);
1185                         }
1186                 }
1187 
1188                 /* move us one more past the eop_desc for start of next pkt */
1189                 tx_buffer++;
1190                 tx_desc++;
1191                 i++;
1192                 if (unlikely(!i)) {
1193                         i -= tx_ring->count;
1194                         tx_buffer = tx_ring->tx_buffer_info;
1195                         tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1196                 }
1197 
1198                 /* issue prefetch for next Tx descriptor */
1199                 prefetch(tx_desc);
1200 
1201                 /* update budget accounting */
1202                 budget--;
1203         } while (likely(budget));
1204 
1205         i += tx_ring->count;
1206         tx_ring->next_to_clean = i;
1207         u64_stats_update_begin(&tx_ring->syncp);
1208         tx_ring->stats.bytes += total_bytes;
1209         tx_ring->stats.packets += total_packets;
1210         u64_stats_update_end(&tx_ring->syncp);
1211         q_vector->tx.total_bytes += total_bytes;
1212         q_vector->tx.total_packets += total_packets;
1213         adapter->tx_ipsec += total_ipsec;
1214 
1215         if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1216                 /* schedule immediate reset if we believe we hung */
1217                 struct ixgbe_hw *hw = &adapter->hw;
1218                 e_err(drv, "Detected Tx Unit Hang %s\n"
1219                         "  Tx Queue             <%d>\n"
1220                         "  TDH, TDT             <%x>, <%x>\n"
1221                         "  next_to_use          <%x>\n"
1222                         "  next_to_clean        <%x>\n"
1223                         "tx_buffer_info[next_to_clean]\n"
1224                         "  time_stamp           <%lx>\n"
1225                         "  jiffies              <%lx>\n",
1226                         ring_is_xdp(tx_ring) ? "(XDP)" : "",
1227                         tx_ring->queue_index,
1228                         IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1229                         IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1230                         tx_ring->next_to_use, i,
1231                         tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1232 
1233                 if (!ring_is_xdp(tx_ring))
1234                         netif_stop_subqueue(tx_ring->netdev,
1235                                             tx_ring->queue_index);
1236 
1237                 e_info(probe,
1238                        "tx hang %d detected on queue %d, resetting adapter\n",
1239                         adapter->tx_timeout_count + 1, tx_ring->queue_index);
1240 
1241                 /* schedule immediate reset if we believe we hung */
1242                 ixgbe_tx_timeout_reset(adapter);
1243 
1244                 /* the adapter is about to reset, no point in enabling stuff */
1245                 return true;
1246         }
1247 
1248         if (ring_is_xdp(tx_ring))
1249                 return !!budget;
1250 
1251         netdev_tx_completed_queue(txring_txq(tx_ring),
1252                                   total_packets, total_bytes);
1253 
1254 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1255         if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1256                      (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1257                 /* Make sure that anybody stopping the queue after this
1258                  * sees the new next_to_clean.
1259                  */
1260                 smp_mb();
1261                 if (__netif_subqueue_stopped(tx_ring->netdev,
1262                                              tx_ring->queue_index)
1263                     && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1264                         netif_wake_subqueue(tx_ring->netdev,
1265                                             tx_ring->queue_index);
1266                         ++tx_ring->tx_stats.restart_queue;
1267                 }
1268         }
1269 
1270         return !!budget;
1271 }
1272 
1273 #ifdef CONFIG_IXGBE_DCA
1274 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1275                                 struct ixgbe_ring *tx_ring,
1276                                 int cpu)
1277 {
1278         struct ixgbe_hw *hw = &adapter->hw;
1279         u32 txctrl = 0;
1280         u16 reg_offset;
1281 
1282         if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1283                 txctrl = dca3_get_tag(tx_ring->dev, cpu);
1284 
1285         switch (hw->mac.type) {
1286         case ixgbe_mac_82598EB:
1287                 reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1288                 break;
1289         case ixgbe_mac_82599EB:
1290         case ixgbe_mac_X540:
1291                 reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1292                 txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1293                 break;
1294         default:
1295                 /* for unknown hardware do not write register */
1296                 return;
1297         }
1298 
1299         /*
1300          * We can enable relaxed ordering for reads, but not writes when
1301          * DCA is enabled.  This is due to a known issue in some chipsets
1302          * which will cause the DCA tag to be cleared.
1303          */
1304         txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1305                   IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1306                   IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1307 
1308         IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1309 }
1310 
1311 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1312                                 struct ixgbe_ring *rx_ring,
1313                                 int cpu)
1314 {
1315         struct ixgbe_hw *hw = &adapter->hw;
1316         u32 rxctrl = 0;
1317         u8 reg_idx = rx_ring->reg_idx;
1318 
1319         if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1320                 rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1321 
1322         switch (hw->mac.type) {
1323         case ixgbe_mac_82599EB:
1324         case ixgbe_mac_X540:
1325                 rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1326                 break;
1327         default:
1328                 break;
1329         }
1330 
1331         /*
1332          * We can enable relaxed ordering for reads, but not writes when
1333          * DCA is enabled.  This is due to a known issue in some chipsets
1334          * which will cause the DCA tag to be cleared.
1335          */
1336         rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1337                   IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1338                   IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1339 
1340         IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1341 }
1342 
1343 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1344 {
1345         struct ixgbe_adapter *adapter = q_vector->adapter;
1346         struct ixgbe_ring *ring;
1347         int cpu = get_cpu();
1348 
1349         if (q_vector->cpu == cpu)
1350                 goto out_no_update;
1351 
1352         ixgbe_for_each_ring(ring, q_vector->tx)
1353                 ixgbe_update_tx_dca(adapter, ring, cpu);
1354 
1355         ixgbe_for_each_ring(ring, q_vector->rx)
1356                 ixgbe_update_rx_dca(adapter, ring, cpu);
1357 
1358         q_vector->cpu = cpu;
1359 out_no_update:
1360         put_cpu();
1361 }
1362 
1363 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1364 {
1365         int i;
1366 
1367         /* always use CB2 mode, difference is masked in the CB driver */
1368         if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1369                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1370                                 IXGBE_DCA_CTRL_DCA_MODE_CB2);
1371         else
1372                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1373                                 IXGBE_DCA_CTRL_DCA_DISABLE);
1374 
1375         for (i = 0; i < adapter->num_q_vectors; i++) {
1376                 adapter->q_vector[i]->cpu = -1;
1377                 ixgbe_update_dca(adapter->q_vector[i]);
1378         }
1379 }
1380 
1381 static int __ixgbe_notify_dca(struct device *dev, void *data)
1382 {
1383         struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1384         unsigned long event = *(unsigned long *)data;
1385 
1386         if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1387                 return 0;
1388 
1389         switch (event) {
1390         case DCA_PROVIDER_ADD:
1391                 /* if we're already enabled, don't do it again */
1392                 if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1393                         break;
1394                 if (dca_add_requester(dev) == 0) {
1395                         adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1396                         IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1397                                         IXGBE_DCA_CTRL_DCA_MODE_CB2);
1398                         break;
1399                 }
1400                 /* fall through - DCA is disabled. */
1401         case DCA_PROVIDER_REMOVE:
1402                 if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1403                         dca_remove_requester(dev);
1404                         adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1405                         IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1406                                         IXGBE_DCA_CTRL_DCA_DISABLE);
1407                 }
1408                 break;
1409         }
1410 
1411         return 0;
1412 }
1413 
1414 #endif /* CONFIG_IXGBE_DCA */
1415 
1416 #define IXGBE_RSS_L4_TYPES_MASK \
1417         ((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1418          (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1419          (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1420          (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1421 
1422 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1423                                  union ixgbe_adv_rx_desc *rx_desc,
1424                                  struct sk_buff *skb)
1425 {
1426         u16 rss_type;
1427 
1428         if (!(ring->netdev->features & NETIF_F_RXHASH))
1429                 return;
1430 
1431         rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1432                    IXGBE_RXDADV_RSSTYPE_MASK;
1433 
1434         if (!rss_type)
1435                 return;
1436 
1437         skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1438                      (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1439                      PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1440 }
1441 
1442 #ifdef IXGBE_FCOE
1443 /**
1444  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1445  * @ring: structure containing ring specific data
1446  * @rx_desc: advanced rx descriptor
1447  *
1448  * Returns : true if it is FCoE pkt
1449  */
1450 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1451                                     union ixgbe_adv_rx_desc *rx_desc)
1452 {
1453         __le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1454 
1455         return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1456                ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1457                 (cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1458                              IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1459 }
1460 
1461 #endif /* IXGBE_FCOE */
1462 /**
1463  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1464  * @ring: structure containing ring specific data
1465  * @rx_desc: current Rx descriptor being processed
1466  * @skb: skb currently being received and modified
1467  **/
1468 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1469                                      union ixgbe_adv_rx_desc *rx_desc,
1470                                      struct sk_buff *skb)
1471 {
1472         __le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1473         bool encap_pkt = false;
1474 
1475         skb_checksum_none_assert(skb);
1476 
1477         /* Rx csum disabled */
1478         if (!(ring->netdev->features & NETIF_F_RXCSUM))
1479                 return;
1480 
1481         /* check for VXLAN and Geneve packets */
1482         if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1483                 encap_pkt = true;
1484                 skb->encapsulation = 1;
1485         }
1486 
1487         /* if IP and error */
1488         if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1489             ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1490                 ring->rx_stats.csum_err++;
1491                 return;
1492         }
1493 
1494         if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1495                 return;
1496 
1497         if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1498                 /*
1499                  * 82599 errata, UDP frames with a 0 checksum can be marked as
1500                  * checksum errors.
1501                  */
1502                 if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1503                     test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1504                         return;
1505 
1506                 ring->rx_stats.csum_err++;
1507                 return;
1508         }
1509 
1510         /* It must be a TCP or UDP packet with a valid checksum */
1511         skb->ip_summed = CHECKSUM_UNNECESSARY;
1512         if (encap_pkt) {
1513                 if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1514                         return;
1515 
1516                 if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1517                         skb->ip_summed = CHECKSUM_NONE;
1518                         return;
1519                 }
1520                 /* If we checked the outer header let the stack know */
1521                 skb->csum_level = 1;
1522         }
1523 }
1524 
1525 static inline unsigned int ixgbe_rx_offset(struct ixgbe_ring *rx_ring)
1526 {
1527         return ring_uses_build_skb(rx_ring) ? IXGBE_SKB_PAD : 0;
1528 }
1529 
1530 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1531                                     struct ixgbe_rx_buffer *bi)
1532 {
1533         struct page *page = bi->page;
1534         dma_addr_t dma;
1535 
1536         /* since we are recycling buffers we should seldom need to alloc */
1537         if (likely(page))
1538                 return true;
1539 
1540         /* alloc new page for storage */
1541         page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1542         if (unlikely(!page)) {
1543                 rx_ring->rx_stats.alloc_rx_page_failed++;
1544                 return false;
1545         }
1546 
1547         /* map page for use */
1548         dma = dma_map_page_attrs(rx_ring->dev, page, 0,
1549                                  ixgbe_rx_pg_size(rx_ring),
1550                                  DMA_FROM_DEVICE,
1551                                  IXGBE_RX_DMA_ATTR);
1552 
1553         /*
1554          * if mapping failed free memory back to system since
1555          * there isn't much point in holding memory we can't use
1556          */
1557         if (dma_mapping_error(rx_ring->dev, dma)) {
1558                 __free_pages(page, ixgbe_rx_pg_order(rx_ring));
1559 
1560                 rx_ring->rx_stats.alloc_rx_page_failed++;
1561                 return false;
1562         }
1563 
1564         bi->dma = dma;
1565         bi->page = page;
1566         bi->page_offset = ixgbe_rx_offset(rx_ring);
1567         page_ref_add(page, USHRT_MAX - 1);
1568         bi->pagecnt_bias = USHRT_MAX;
1569         rx_ring->rx_stats.alloc_rx_page++;
1570 
1571         return true;
1572 }
1573 
1574 /**
1575  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1576  * @rx_ring: ring to place buffers on
1577  * @cleaned_count: number of buffers to replace
1578  **/
1579 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1580 {
1581         union ixgbe_adv_rx_desc *rx_desc;
1582         struct ixgbe_rx_buffer *bi;
1583         u16 i = rx_ring->next_to_use;
1584         u16 bufsz;
1585 
1586         /* nothing to do */
1587         if (!cleaned_count)
1588                 return;
1589 
1590         rx_desc = IXGBE_RX_DESC(rx_ring, i);
1591         bi = &rx_ring->rx_buffer_info[i];
1592         i -= rx_ring->count;
1593 
1594         bufsz = ixgbe_rx_bufsz(rx_ring);
1595 
1596         do {
1597                 if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1598                         break;
1599 
1600                 /* sync the buffer for use by the device */
1601                 dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
1602                                                  bi->page_offset, bufsz,
1603                                                  DMA_FROM_DEVICE);
1604 
1605                 /*
1606                  * Refresh the desc even if buffer_addrs didn't change
1607                  * because each write-back erases this info.
1608                  */
1609                 rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1610 
1611                 rx_desc++;
1612                 bi++;
1613                 i++;
1614                 if (unlikely(!i)) {
1615                         rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1616                         bi = rx_ring->rx_buffer_info;
1617                         i -= rx_ring->count;
1618                 }
1619 
1620                 /* clear the length for the next_to_use descriptor */
1621                 rx_desc->wb.upper.length = 0;
1622 
1623                 cleaned_count--;
1624         } while (cleaned_count);
1625 
1626         i += rx_ring->count;
1627 
1628         if (rx_ring->next_to_use != i) {
1629                 rx_ring->next_to_use = i;
1630 
1631                 /* update next to alloc since we have filled the ring */
1632                 rx_ring->next_to_alloc = i;
1633 
1634                 /* Force memory writes to complete before letting h/w
1635                  * know there are new descriptors to fetch.  (Only
1636                  * applicable for weak-ordered memory model archs,
1637                  * such as IA-64).
1638                  */
1639                 wmb();
1640                 writel(i, rx_ring->tail);
1641         }
1642 }
1643 
1644 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1645                                    struct sk_buff *skb)
1646 {
1647         u16 hdr_len = skb_headlen(skb);
1648 
1649         /* set gso_size to avoid messing up TCP MSS */
1650         skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1651                                                  IXGBE_CB(skb)->append_cnt);
1652         skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1653 }
1654 
1655 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1656                                    struct sk_buff *skb)
1657 {
1658         /* if append_cnt is 0 then frame is not RSC */
1659         if (!IXGBE_CB(skb)->append_cnt)
1660                 return;
1661 
1662         rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1663         rx_ring->rx_stats.rsc_flush++;
1664 
1665         ixgbe_set_rsc_gso_size(rx_ring, skb);
1666 
1667         /* gso_size is computed using append_cnt so always clear it last */
1668         IXGBE_CB(skb)->append_cnt = 0;
1669 }
1670 
1671 /**
1672  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1673  * @rx_ring: rx descriptor ring packet is being transacted on
1674  * @rx_desc: pointer to the EOP Rx descriptor
1675  * @skb: pointer to current skb being populated
1676  *
1677  * This function checks the ring, descriptor, and packet information in
1678  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1679  * other fields within the skb.
1680  **/
1681 void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1682                               union ixgbe_adv_rx_desc *rx_desc,
1683                               struct sk_buff *skb)
1684 {
1685         struct net_device *dev = rx_ring->netdev;
1686         u32 flags = rx_ring->q_vector->adapter->flags;
1687 
1688         ixgbe_update_rsc_stats(rx_ring, skb);
1689 
1690         ixgbe_rx_hash(rx_ring, rx_desc, skb);
1691 
1692         ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1693 
1694         if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1695                 ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1696 
1697         if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1698             ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1699                 u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1700                 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1701         }
1702 
1703         if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_SECP))
1704                 ixgbe_ipsec_rx(rx_ring, rx_desc, skb);
1705 
1706         /* record Rx queue, or update MACVLAN statistics */
1707         if (netif_is_ixgbe(dev))
1708                 skb_record_rx_queue(skb, rx_ring->queue_index);
1709         else
1710                 macvlan_count_rx(netdev_priv(dev), skb->len + ETH_HLEN, true,
1711                                  false);
1712 
1713         skb->protocol = eth_type_trans(skb, dev);
1714 }
1715 
1716 void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1717                   struct sk_buff *skb)
1718 {
1719         napi_gro_receive(&q_vector->napi, skb);
1720 }
1721 
1722 /**
1723  * ixgbe_is_non_eop - process handling of non-EOP buffers
1724  * @rx_ring: Rx ring being processed
1725  * @rx_desc: Rx descriptor for current buffer
1726  * @skb: Current socket buffer containing buffer in progress
1727  *
1728  * This function updates next to clean.  If the buffer is an EOP buffer
1729  * this function exits returning false, otherwise it will place the
1730  * sk_buff in the next buffer to be chained and return true indicating
1731  * that this is in fact a non-EOP buffer.
1732  **/
1733 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1734                              union ixgbe_adv_rx_desc *rx_desc,
1735                              struct sk_buff *skb)
1736 {
1737         u32 ntc = rx_ring->next_to_clean + 1;
1738 
1739         /* fetch, update, and store next to clean */
1740         ntc = (ntc < rx_ring->count) ? ntc : 0;
1741         rx_ring->next_to_clean = ntc;
1742 
1743         prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1744 
1745         /* update RSC append count if present */
1746         if (ring_is_rsc_enabled(rx_ring)) {
1747                 __le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1748                                      cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1749 
1750                 if (unlikely(rsc_enabled)) {
1751                         u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1752 
1753                         rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1754                         IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1755 
1756                         /* update ntc based on RSC value */
1757                         ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1758                         ntc &= IXGBE_RXDADV_NEXTP_MASK;
1759                         ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1760                 }
1761         }
1762 
1763         /* if we are the last buffer then there is nothing else to do */
1764         if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1765                 return false;
1766 
1767         /* place skb in next buffer to be received */
1768         rx_ring->rx_buffer_info[ntc].skb = skb;
1769         rx_ring->rx_stats.non_eop_descs++;
1770 
1771         return true;
1772 }
1773 
1774 /**
1775  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1776  * @rx_ring: rx descriptor ring packet is being transacted on
1777  * @skb: pointer to current skb being adjusted
1778  *
1779  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1780  * main difference between this version and the original function is that
1781  * this function can make several assumptions about the state of things
1782  * that allow for significant optimizations versus the standard function.
1783  * As a result we can do things like drop a frag and maintain an accurate
1784  * truesize for the skb.
1785  */
1786 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1787                             struct sk_buff *skb)
1788 {
1789         skb_frag_t *frag = &skb_shinfo(skb)->frags[0];
1790         unsigned char *va;
1791         unsigned int pull_len;
1792 
1793         /*
1794          * it is valid to use page_address instead of kmap since we are
1795          * working with pages allocated out of the lomem pool per
1796          * alloc_page(GFP_ATOMIC)
1797          */
1798         va = skb_frag_address(frag);
1799 
1800         /*
1801          * we need the header to contain the greater of either ETH_HLEN or
1802          * 60 bytes if the skb->len is less than 60 for skb_pad.
1803          */
1804         pull_len = eth_get_headlen(skb->dev, va, IXGBE_RX_HDR_SIZE);
1805 
1806         /* align pull length to size of long to optimize memcpy performance */
1807         skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1808 
1809         /* update all of the pointers */
1810         skb_frag_size_sub(frag, pull_len);
1811         skb_frag_off_add(frag, pull_len);
1812         skb->data_len -= pull_len;
1813         skb->tail += pull_len;
1814 }
1815 
1816 /**
1817  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1818  * @rx_ring: rx descriptor ring packet is being transacted on
1819  * @skb: pointer to current skb being updated
1820  *
1821  * This function provides a basic DMA sync up for the first fragment of an
1822  * skb.  The reason for doing this is that the first fragment cannot be
1823  * unmapped until we have reached the end of packet descriptor for a buffer
1824  * chain.
1825  */
1826 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1827                                 struct sk_buff *skb)
1828 {
1829         if (ring_uses_build_skb(rx_ring)) {
1830                 unsigned long offset = (unsigned long)(skb->data) & ~PAGE_MASK;
1831 
1832                 dma_sync_single_range_for_cpu(rx_ring->dev,
1833                                               IXGBE_CB(skb)->dma,
1834                                               offset,
1835                                               skb_headlen(skb),
1836                                               DMA_FROM_DEVICE);
1837         } else {
1838                 skb_frag_t *frag = &skb_shinfo(skb)->frags[0];
1839 
1840                 dma_sync_single_range_for_cpu(rx_ring->dev,
1841                                               IXGBE_CB(skb)->dma,
1842                                               skb_frag_off(frag),
1843                                               skb_frag_size(frag),
1844                                               DMA_FROM_DEVICE);
1845         }
1846 
1847         /* If the page was released, just unmap it. */
1848         if (unlikely(IXGBE_CB(skb)->page_released)) {
1849                 dma_unmap_page_attrs(rx_ring->dev, IXGBE_CB(skb)->dma,
1850                                      ixgbe_rx_pg_size(rx_ring),
1851                                      DMA_FROM_DEVICE,
1852                                      IXGBE_RX_DMA_ATTR);
1853         }
1854 }
1855 
1856 /**
1857  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1858  * @rx_ring: rx descriptor ring packet is being transacted on
1859  * @rx_desc: pointer to the EOP Rx descriptor
1860  * @skb: pointer to current skb being fixed
1861  *
1862  * Check if the skb is valid in the XDP case it will be an error pointer.
1863  * Return true in this case to abort processing and advance to next
1864  * descriptor.
1865  *
1866  * Check for corrupted packet headers caused by senders on the local L2
1867  * embedded NIC switch not setting up their Tx Descriptors right.  These
1868  * should be very rare.
1869  *
1870  * Also address the case where we are pulling data in on pages only
1871  * and as such no data is present in the skb header.
1872  *
1873  * In addition if skb is not at least 60 bytes we need to pad it so that
1874  * it is large enough to qualify as a valid Ethernet frame.
1875  *
1876  * Returns true if an error was encountered and skb was freed.
1877  **/
1878 bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1879                            union ixgbe_adv_rx_desc *rx_desc,
1880                            struct sk_buff *skb)
1881 {
1882         struct net_device *netdev = rx_ring->netdev;
1883 
1884         /* XDP packets use error pointer so abort at this point */
1885         if (IS_ERR(skb))
1886                 return true;
1887 
1888         /* Verify netdev is present, and that packet does not have any
1889          * errors that would be unacceptable to the netdev.
1890          */
1891         if (!netdev ||
1892             (unlikely(ixgbe_test_staterr(rx_desc,
1893                                          IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1894              !(netdev->features & NETIF_F_RXALL)))) {
1895                 dev_kfree_skb_any(skb);
1896                 return true;
1897         }
1898 
1899         /* place header in linear portion of buffer */
1900         if (!skb_headlen(skb))
1901                 ixgbe_pull_tail(rx_ring, skb);
1902 
1903 #ifdef IXGBE_FCOE
1904         /* do not attempt to pad FCoE Frames as this will disrupt DDP */
1905         if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1906                 return false;
1907 
1908 #endif
1909         /* if eth_skb_pad returns an error the skb was freed */
1910         if (eth_skb_pad(skb))
1911                 return true;
1912 
1913         return false;
1914 }
1915 
1916 /**
1917  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1918  * @rx_ring: rx descriptor ring to store buffers on
1919  * @old_buff: donor buffer to have page reused
1920  *
1921  * Synchronizes page for reuse by the adapter
1922  **/
1923 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1924                                 struct ixgbe_rx_buffer *old_buff)
1925 {
1926         struct ixgbe_rx_buffer *new_buff;
1927         u16 nta = rx_ring->next_to_alloc;
1928 
1929         new_buff = &rx_ring->rx_buffer_info[nta];
1930 
1931         /* update, and store next to alloc */
1932         nta++;
1933         rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1934 
1935         /* Transfer page from old buffer to new buffer.
1936          * Move each member individually to avoid possible store
1937          * forwarding stalls and unnecessary copy of skb.
1938          */
1939         new_buff->dma           = old_buff->dma;
1940         new_buff->page          = old_buff->page;
1941         new_buff->page_offset   = old_buff->page_offset;
1942         new_buff->pagecnt_bias  = old_buff->pagecnt_bias;
1943 }
1944 
1945 static inline bool ixgbe_page_is_reserved(struct page *page)
1946 {
1947         return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
1948 }
1949 
1950 static bool ixgbe_can_reuse_rx_page(struct ixgbe_rx_buffer *rx_buffer)
1951 {
1952         unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
1953         struct page *page = rx_buffer->page;
1954 
1955         /* avoid re-using remote pages */
1956         if (unlikely(ixgbe_page_is_reserved(page)))
1957                 return false;
1958 
1959 #if (PAGE_SIZE < 8192)
1960         /* if we are only owner of page we can reuse it */
1961         if (unlikely((page_ref_count(page) - pagecnt_bias) > 1))
1962                 return false;
1963 #else
1964         /* The last offset is a bit aggressive in that we assume the
1965          * worst case of FCoE being enabled and using a 3K buffer.
1966          * However this should have minimal impact as the 1K extra is
1967          * still less than one buffer in size.
1968          */
1969 #define IXGBE_LAST_OFFSET \
1970         (SKB_WITH_OVERHEAD(PAGE_SIZE) - IXGBE_RXBUFFER_3K)
1971         if (rx_buffer->page_offset > IXGBE_LAST_OFFSET)
1972                 return false;
1973 #endif
1974 
1975         /* If we have drained the page fragment pool we need to update
1976          * the pagecnt_bias and page count so that we fully restock the
1977          * number of references the driver holds.
1978          */
1979         if (unlikely(pagecnt_bias == 1)) {
1980                 page_ref_add(page, USHRT_MAX - 1);
1981                 rx_buffer->pagecnt_bias = USHRT_MAX;
1982         }
1983 
1984         return true;
1985 }
1986 
1987 /**
1988  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1989  * @rx_ring: rx descriptor ring to transact packets on
1990  * @rx_buffer: buffer containing page to add
1991  * @skb: sk_buff to place the data into
1992  * @size: size of data in rx_buffer
1993  *
1994  * This function will add the data contained in rx_buffer->page to the skb.
1995  * This is done either through a direct copy if the data in the buffer is
1996  * less than the skb header size, otherwise it will just attach the page as
1997  * a frag to the skb.
1998  *
1999  * The function will then update the page offset if necessary and return
2000  * true if the buffer can be reused by the adapter.
2001  **/
2002 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
2003                               struct ixgbe_rx_buffer *rx_buffer,
2004                               struct sk_buff *skb,
2005                               unsigned int size)
2006 {
2007 #if (PAGE_SIZE < 8192)
2008         unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2009 #else
2010         unsigned int truesize = ring_uses_build_skb(rx_ring) ?
2011                                 SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
2012                                 SKB_DATA_ALIGN(size);
2013 #endif
2014         skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
2015                         rx_buffer->page_offset, size, truesize);
2016 #if (PAGE_SIZE < 8192)
2017         rx_buffer->page_offset ^= truesize;
2018 #else
2019         rx_buffer->page_offset += truesize;
2020 #endif
2021 }
2022 
2023 static struct ixgbe_rx_buffer *ixgbe_get_rx_buffer(struct ixgbe_ring *rx_ring,
2024                                                    union ixgbe_adv_rx_desc *rx_desc,
2025                                                    struct sk_buff **skb,
2026                                                    const unsigned int size)
2027 {
2028         struct ixgbe_rx_buffer *rx_buffer;
2029 
2030         rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2031         prefetchw(rx_buffer->page);
2032         *skb = rx_buffer->skb;
2033 
2034         /* Delay unmapping of the first packet. It carries the header
2035          * information, HW may still access the header after the writeback.
2036          * Only unmap it when EOP is reached
2037          */
2038         if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)) {
2039                 if (!*skb)
2040                         goto skip_sync;
2041         } else {
2042                 if (*skb)
2043                         ixgbe_dma_sync_frag(rx_ring, *skb);
2044         }
2045 
2046         /* we are reusing so sync this buffer for CPU use */
2047         dma_sync_single_range_for_cpu(rx_ring->dev,
2048                                       rx_buffer->dma,
2049                                       rx_buffer->page_offset,
2050                                       size,
2051                                       DMA_FROM_DEVICE);
2052 skip_sync:
2053         rx_buffer->pagecnt_bias--;
2054 
2055         return rx_buffer;
2056 }
2057 
2058 static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
2059                                 struct ixgbe_rx_buffer *rx_buffer,
2060                                 struct sk_buff *skb)
2061 {
2062         if (ixgbe_can_reuse_rx_page(rx_buffer)) {
2063                 /* hand second half of page back to the ring */
2064                 ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2065         } else {
2066                 if (!IS_ERR(skb) && IXGBE_CB(skb)->dma == rx_buffer->dma) {
2067                         /* the page has been released from the ring */
2068                         IXGBE_CB(skb)->page_released = true;
2069                 } else {
2070                         /* we are not reusing the buffer so unmap it */
2071                         dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
2072                                              ixgbe_rx_pg_size(rx_ring),
2073                                              DMA_FROM_DEVICE,
2074                                              IXGBE_RX_DMA_ATTR);
2075                 }
2076                 __page_frag_cache_drain(rx_buffer->page,
2077                                         rx_buffer->pagecnt_bias);
2078         }
2079 
2080         /* clear contents of rx_buffer */
2081         rx_buffer->page = NULL;
2082         rx_buffer->skb = NULL;
2083 }
2084 
2085 static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
2086                                            struct ixgbe_rx_buffer *rx_buffer,
2087                                            struct xdp_buff *xdp,
2088                                            union ixgbe_adv_rx_desc *rx_desc)
2089 {
2090         unsigned int size = xdp->data_end - xdp->data;
2091 #if (PAGE_SIZE < 8192)
2092         unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2093 #else
2094         unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
2095                                                xdp->data_hard_start);
2096 #endif
2097         struct sk_buff *skb;
2098 
2099         /* prefetch first cache line of first page */
2100         prefetch(xdp->data);
2101 #if L1_CACHE_BYTES < 128
2102         prefetch(xdp->data + L1_CACHE_BYTES);
2103 #endif
2104         /* Note, we get here by enabling legacy-rx via:
2105          *
2106          *    ethtool --set-priv-flags <dev> legacy-rx on
2107          *
2108          * In this mode, we currently get 0 extra XDP headroom as
2109          * opposed to having legacy-rx off, where we process XDP
2110          * packets going to stack via ixgbe_build_skb(). The latter
2111          * provides us currently with 192 bytes of headroom.
2112          *
2113          * For ixgbe_construct_skb() mode it means that the
2114          * xdp->data_meta will always point to xdp->data, since
2115          * the helper cannot expand the head. Should this ever
2116          * change in future for legacy-rx mode on, then lets also
2117          * add xdp->data_meta handling here.
2118          */
2119 
2120         /* allocate a skb to store the frags */
2121         skb = napi_alloc_skb(&rx_ring->q_vector->napi, IXGBE_RX_HDR_SIZE);
2122         if (unlikely(!skb))
2123                 return NULL;
2124 
2125         if (size > IXGBE_RX_HDR_SIZE) {
2126                 if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2127                         IXGBE_CB(skb)->dma = rx_buffer->dma;
2128 
2129                 skb_add_rx_frag(skb, 0, rx_buffer->page,
2130                                 xdp->data - page_address(rx_buffer->page),
2131                                 size, truesize);
2132 #if (PAGE_SIZE < 8192)
2133                 rx_buffer->page_offset ^= truesize;
2134 #else
2135                 rx_buffer->page_offset += truesize;
2136 #endif
2137         } else {
2138                 memcpy(__skb_put(skb, size),
2139                        xdp->data, ALIGN(size, sizeof(long)));
2140                 rx_buffer->pagecnt_bias++;
2141         }
2142 
2143         return skb;
2144 }
2145 
2146 static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
2147                                        struct ixgbe_rx_buffer *rx_buffer,
2148                                        struct xdp_buff *xdp,
2149                                        union ixgbe_adv_rx_desc *rx_desc)
2150 {
2151         unsigned int metasize = xdp->data - xdp->data_meta;
2152 #if (PAGE_SIZE < 8192)
2153         unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2154 #else
2155         unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
2156                                 SKB_DATA_ALIGN(xdp->data_end -
2157                                                xdp->data_hard_start);
2158 #endif
2159         struct sk_buff *skb;
2160 
2161         /* Prefetch first cache line of first page. If xdp->data_meta
2162          * is unused, this points extactly as xdp->data, otherwise we
2163          * likely have a consumer accessing first few bytes of meta
2164          * data, and then actual data.
2165          */
2166         prefetch(xdp->data_meta);
2167 #if L1_CACHE_BYTES < 128
2168         prefetch(xdp->data_meta + L1_CACHE_BYTES);
2169 #endif
2170 
2171         /* build an skb to around the page buffer */
2172         skb = build_skb(xdp->data_hard_start, truesize);
2173         if (unlikely(!skb))
2174                 return NULL;
2175 
2176         /* update pointers within the skb to store the data */
2177         skb_reserve(skb, xdp->data - xdp->data_hard_start);
2178         __skb_put(skb, xdp->data_end - xdp->data);
2179         if (metasize)
2180                 skb_metadata_set(skb, metasize);
2181 
2182         /* record DMA address if this is the start of a chain of buffers */
2183         if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2184                 IXGBE_CB(skb)->dma = rx_buffer->dma;
2185 
2186         /* update buffer offset */
2187 #if (PAGE_SIZE < 8192)
2188         rx_buffer->page_offset ^= truesize;
2189 #else
2190         rx_buffer->page_offset += truesize;
2191 #endif
2192 
2193         return skb;
2194 }
2195 
2196 static struct sk_buff *ixgbe_run_xdp(struct ixgbe_adapter *adapter,
2197                                      struct ixgbe_ring *rx_ring,
2198                                      struct xdp_buff *xdp)
2199 {
2200         int err, result = IXGBE_XDP_PASS;
2201         struct bpf_prog *xdp_prog;
2202         struct xdp_frame *xdpf;
2203         u32 act;
2204 
2205         rcu_read_lock();
2206         xdp_prog = READ_ONCE(rx_ring->xdp_prog);
2207 
2208         if (!xdp_prog)
2209                 goto xdp_out;
2210 
2211         prefetchw(xdp->data_hard_start); /* xdp_frame write */
2212 
2213         act = bpf_prog_run_xdp(xdp_prog, xdp);
2214         switch (act) {
2215         case XDP_PASS:
2216                 break;
2217         case XDP_TX:
2218                 xdpf = convert_to_xdp_frame(xdp);
2219                 if (unlikely(!xdpf)) {
2220                         result = IXGBE_XDP_CONSUMED;
2221                         break;
2222                 }
2223                 result = ixgbe_xmit_xdp_ring(adapter, xdpf);
2224                 break;
2225         case XDP_REDIRECT:
2226                 err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
2227                 if (!err)
2228                         result = IXGBE_XDP_REDIR;
2229                 else
2230                         result = IXGBE_XDP_CONSUMED;
2231                 break;
2232         default:
2233                 bpf_warn_invalid_xdp_action(act);
2234                 /* fallthrough */
2235         case XDP_ABORTED:
2236                 trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
2237                 /* fallthrough -- handle aborts by dropping packet */
2238         case XDP_DROP:
2239                 result = IXGBE_XDP_CONSUMED;
2240                 break;
2241         }
2242 xdp_out:
2243         rcu_read_unlock();
2244         return ERR_PTR(-result);
2245 }
2246 
2247 static void ixgbe_rx_buffer_flip(struct ixgbe_ring *rx_ring,
2248                                  struct ixgbe_rx_buffer *rx_buffer,
2249                                  unsigned int size)
2250 {
2251 #if (PAGE_SIZE < 8192)
2252         unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2253 
2254         rx_buffer->page_offset ^= truesize;
2255 #else
2256         unsigned int truesize = ring_uses_build_skb(rx_ring) ?
2257                                 SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
2258                                 SKB_DATA_ALIGN(size);
2259 
2260         rx_buffer->page_offset += truesize;
2261 #endif
2262 }
2263 
2264 /**
2265  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2266  * @q_vector: structure containing interrupt and ring information
2267  * @rx_ring: rx descriptor ring to transact packets on
2268  * @budget: Total limit on number of packets to process
2269  *
2270  * This function provides a "bounce buffer" approach to Rx interrupt
2271  * processing.  The advantage to this is that on systems that have
2272  * expensive overhead for IOMMU access this provides a means of avoiding
2273  * it by maintaining the mapping of the page to the syste.
2274  *
2275  * Returns amount of work completed
2276  **/
2277 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2278                                struct ixgbe_ring *rx_ring,
2279                                const int budget)
2280 {
2281         unsigned int total_rx_bytes = 0, total_rx_packets = 0;
2282         struct ixgbe_adapter *adapter = q_vector->adapter;
2283 #ifdef IXGBE_FCOE
2284         int ddp_bytes;
2285         unsigned int mss = 0;
2286 #endif /* IXGBE_FCOE */
2287         u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2288         unsigned int xdp_xmit = 0;
2289         struct xdp_buff xdp;
2290 
2291         xdp.rxq = &rx_ring->xdp_rxq;
2292 
2293         while (likely(total_rx_packets < budget)) {
2294                 union ixgbe_adv_rx_desc *rx_desc;
2295                 struct ixgbe_rx_buffer *rx_buffer;
2296                 struct sk_buff *skb;
2297                 unsigned int size;
2298 
2299                 /* return some buffers to hardware, one at a time is too slow */
2300                 if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2301                         ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2302                         cleaned_count = 0;
2303                 }
2304 
2305                 rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2306                 size = le16_to_cpu(rx_desc->wb.upper.length);
2307                 if (!size)
2308                         break;
2309 
2310                 /* This memory barrier is needed to keep us from reading
2311                  * any other fields out of the rx_desc until we know the
2312                  * descriptor has been written back
2313                  */
2314                 dma_rmb();
2315 
2316                 rx_buffer = ixgbe_get_rx_buffer(rx_ring, rx_desc, &skb, size);
2317 
2318                 /* retrieve a buffer from the ring */
2319                 if (!skb) {
2320                         xdp.data = page_address(rx_buffer->page) +
2321                                    rx_buffer->page_offset;
2322                         xdp.data_meta = xdp.data;
2323                         xdp.data_hard_start = xdp.data -
2324                                               ixgbe_rx_offset(rx_ring);
2325                         xdp.data_end = xdp.data + size;
2326 
2327                         skb = ixgbe_run_xdp(adapter, rx_ring, &xdp);
2328                 }
2329 
2330                 if (IS_ERR(skb)) {
2331                         unsigned int xdp_res = -PTR_ERR(skb);
2332 
2333                         if (xdp_res & (IXGBE_XDP_TX | IXGBE_XDP_REDIR)) {
2334                                 xdp_xmit |= xdp_res;
2335                                 ixgbe_rx_buffer_flip(rx_ring, rx_buffer, size);
2336                         } else {
2337                                 rx_buffer->pagecnt_bias++;
2338                         }
2339                         total_rx_packets++;
2340                         total_rx_bytes += size;
2341                 } else if (skb) {
2342                         ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
2343                 } else if (ring_uses_build_skb(rx_ring)) {
2344                         skb = ixgbe_build_skb(rx_ring, rx_buffer,
2345                                               &xdp, rx_desc);
2346                 } else {
2347                         skb = ixgbe_construct_skb(rx_ring, rx_buffer,
2348                                                   &xdp, rx_desc);
2349                 }
2350 
2351                 /* exit if we failed to retrieve a buffer */
2352                 if (!skb) {
2353                         rx_ring->rx_stats.alloc_rx_buff_failed++;
2354                         rx_buffer->pagecnt_bias++;
2355                         break;
2356                 }
2357 
2358                 ixgbe_put_rx_buffer(rx_ring, rx_buffer, skb);
2359                 cleaned_count++;
2360 
2361                 /* place incomplete frames back on ring for completion */
2362                 if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2363                         continue;
2364 
2365                 /* verify the packet layout is correct */
2366                 if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2367                         continue;
2368 
2369                 /* probably a little skewed due to removing CRC */
2370                 total_rx_bytes += skb->len;
2371 
2372                 /* populate checksum, timestamp, VLAN, and protocol */
2373                 ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2374 
2375 #ifdef IXGBE_FCOE
2376                 /* if ddp, not passing to ULD unless for FCP_RSP or error */
2377                 if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2378                         ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2379                         /* include DDPed FCoE data */
2380                         if (ddp_bytes > 0) {
2381                                 if (!mss) {
2382                                         mss = rx_ring->netdev->mtu -
2383                                                 sizeof(struct fcoe_hdr) -
2384                                                 sizeof(struct fc_frame_header) -
2385                                                 sizeof(struct fcoe_crc_eof);
2386                                         if (mss > 512)
2387                                                 mss &= ~511;
2388                                 }
2389                                 total_rx_bytes += ddp_bytes;
2390                                 total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2391                                                                  mss);
2392                         }
2393                         if (!ddp_bytes) {
2394                                 dev_kfree_skb_any(skb);
2395                                 continue;
2396                         }
2397                 }
2398 
2399 #endif /* IXGBE_FCOE */
2400                 ixgbe_rx_skb(q_vector, skb);
2401 
2402                 /* update budget accounting */
2403                 total_rx_packets++;
2404         }
2405 
2406         if (xdp_xmit & IXGBE_XDP_REDIR)
2407                 xdp_do_flush_map();
2408 
2409         if (xdp_xmit & IXGBE_XDP_TX) {
2410                 struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
2411 
2412                 /* Force memory writes to complete before letting h/w
2413                  * know there are new descriptors to fetch.
2414                  */
2415                 wmb();
2416                 writel(ring->next_to_use, ring->tail);
2417         }
2418 
2419         u64_stats_update_begin(&rx_ring->syncp);
2420         rx_ring->stats.packets += total_rx_packets;
2421         rx_ring->stats.bytes += total_rx_bytes;
2422         u64_stats_update_end(&rx_ring->syncp);
2423         q_vector->rx.total_packets += total_rx_packets;
2424         q_vector->rx.total_bytes += total_rx_bytes;
2425 
2426         return total_rx_packets;
2427 }
2428 
2429 /**
2430  * ixgbe_configure_msix - Configure MSI-X hardware
2431  * @adapter: board private structure
2432  *
2433  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2434  * interrupts.
2435  **/
2436 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2437 {
2438         struct ixgbe_q_vector *q_vector;
2439         int v_idx;
2440         u32 mask;
2441 
2442         /* Populate MSIX to EITR Select */
2443         if (adapter->num_vfs > 32) {
2444                 u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2445                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2446         }
2447 
2448         /*
2449          * Populate the IVAR table and set the ITR values to the
2450          * corresponding register.
2451          */
2452         for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2453                 struct ixgbe_ring *ring;
2454                 q_vector = adapter->q_vector[v_idx];
2455 
2456                 ixgbe_for_each_ring(ring, q_vector->rx)
2457                         ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2458 
2459                 ixgbe_for_each_ring(ring, q_vector->tx)
2460                         ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2461 
2462                 ixgbe_write_eitr(q_vector);
2463         }
2464 
2465         switch (adapter->hw.mac.type) {
2466         case ixgbe_mac_82598EB:
2467                 ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2468                                v_idx);
2469                 break;
2470         case ixgbe_mac_82599EB:
2471         case ixgbe_mac_X540:
2472         case ixgbe_mac_X550:
2473         case ixgbe_mac_X550EM_x:
2474         case ixgbe_mac_x550em_a:
2475                 ixgbe_set_ivar(adapter, -1, 1, v_idx);
2476                 break;
2477         default:
2478                 break;
2479         }
2480         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2481 
2482         /* set up to autoclear timer, and the vectors */
2483         mask = IXGBE_EIMS_ENABLE_MASK;
2484         mask &= ~(IXGBE_EIMS_OTHER |
2485                   IXGBE_EIMS_MAILBOX |
2486                   IXGBE_EIMS_LSC);
2487 
2488         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2489 }
2490 
2491 /**
2492  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2493  * @q_vector: structure containing interrupt and ring information
2494  * @ring_container: structure containing ring performance data
2495  *
2496  *      Stores a new ITR value based on packets and byte
2497  *      counts during the last interrupt.  The advantage of per interrupt
2498  *      computation is faster updates and more accurate ITR for the current
2499  *      traffic pattern.  Constants in this function were computed
2500  *      based on theoretical maximum wire speed and thresholds were set based
2501  *      on testing data as well as attempting to minimize response time
2502  *      while increasing bulk throughput.
2503  **/
2504 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2505                              struct ixgbe_ring_container *ring_container)
2506 {
2507         unsigned int itr = IXGBE_ITR_ADAPTIVE_MIN_USECS |
2508                            IXGBE_ITR_ADAPTIVE_LATENCY;
2509         unsigned int avg_wire_size, packets, bytes;
2510         unsigned long next_update = jiffies;
2511 
2512         /* If we don't have any rings just leave ourselves set for maximum
2513          * possible latency so we take ourselves out of the equation.
2514          */
2515         if (!ring_container->ring)
2516                 return;
2517 
2518         /* If we didn't update within up to 1 - 2 jiffies we can assume
2519          * that either packets are coming in so slow there hasn't been
2520          * any work, or that there is so much work that NAPI is dealing
2521          * with interrupt moderation and we don't need to do anything.
2522          */
2523         if (time_after(next_update, ring_container->next_update))
2524                 goto clear_counts;
2525 
2526         packets = ring_container->total_packets;
2527 
2528         /* We have no packets to actually measure against. This means
2529          * either one of the other queues on this vector is active or
2530          * we are a Tx queue doing TSO with too high of an interrupt rate.
2531          *
2532          * When this occurs just tick up our delay by the minimum value
2533          * and hope that this extra delay will prevent us from being called
2534          * without any work on our queue.
2535          */
2536         if (!packets) {
2537                 itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2538                 if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2539                         itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2540                 itr += ring_container->itr & IXGBE_ITR_ADAPTIVE_LATENCY;
2541                 goto clear_counts;
2542         }
2543 
2544         bytes = ring_container->total_bytes;
2545 
2546         /* If packets are less than 4 or bytes are less than 9000 assume
2547          * insufficient data to use bulk rate limiting approach. We are
2548          * likely latency driven.
2549          */
2550         if (packets < 4 && bytes < 9000) {
2551                 itr = IXGBE_ITR_ADAPTIVE_LATENCY;
2552                 goto adjust_by_size;
2553         }
2554 
2555         /* Between 4 and 48 we can assume that our current interrupt delay
2556          * is only slightly too low. As such we should increase it by a small
2557          * fixed amount.
2558          */
2559         if (packets < 48) {
2560                 itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2561                 if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2562                         itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2563                 goto clear_counts;
2564         }
2565 
2566         /* Between 48 and 96 is our "goldilocks" zone where we are working
2567          * out "just right". Just report that our current ITR is good for us.
2568          */
2569         if (packets < 96) {
2570                 itr = q_vector->itr >> 2;
2571                 goto clear_counts;
2572         }
2573 
2574         /* If packet count is 96 or greater we are likely looking at a slight
2575          * overrun of the delay we want. Try halving our delay to see if that
2576          * will cut the number of packets in half per interrupt.
2577          */
2578         if (packets < 256) {
2579                 itr = q_vector->itr >> 3;
2580                 if (itr < IXGBE_ITR_ADAPTIVE_MIN_USECS)
2581                         itr = IXGBE_ITR_ADAPTIVE_MIN_USECS;
2582                 goto clear_counts;
2583         }
2584 
2585         /* The paths below assume we are dealing with a bulk ITR since number
2586          * of packets is 256 or greater. We are just going to have to compute
2587          * a value and try to bring the count under control, though for smaller
2588          * packet sizes there isn't much we can do as NAPI polling will likely
2589          * be kicking in sooner rather than later.
2590          */
2591         itr = IXGBE_ITR_ADAPTIVE_BULK;
2592 
2593 adjust_by_size:
2594         /* If packet counts are 256 or greater we can assume we have a gross
2595          * overestimation of what the rate should be. Instead of trying to fine
2596          * tune it just use the formula below to try and dial in an exact value
2597          * give the current packet size of the frame.
2598          */
2599         avg_wire_size = bytes / packets;
2600 
2601         /* The following is a crude approximation of:
2602          *  wmem_default / (size + overhead) = desired_pkts_per_int
2603          *  rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
2604          *  (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
2605          *
2606          * Assuming wmem_default is 212992 and overhead is 640 bytes per
2607          * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
2608          * formula down to
2609          *
2610          *  (170 * (size + 24)) / (size + 640) = ITR
2611          *
2612          * We first do some math on the packet size and then finally bitshift
2613          * by 8 after rounding up. We also have to account for PCIe link speed
2614          * difference as ITR scales based on this.
2615          */
2616         if (avg_wire_size <= 60) {
2617                 /* Start at 50k ints/sec */
2618                 avg_wire_size = 5120;
2619         } else if (avg_wire_size <= 316) {
2620                 /* 50K ints/sec to 16K ints/sec */
2621                 avg_wire_size *= 40;
2622                 avg_wire_size += 2720;
2623         } else if (avg_wire_size <= 1084) {
2624                 /* 16K ints/sec to 9.2K ints/sec */
2625                 avg_wire_size *= 15;
2626                 avg_wire_size += 11452;
2627         } else if (avg_wire_size < 1968) {
2628                 /* 9.2K ints/sec to 8K ints/sec */
2629                 avg_wire_size *= 5;
2630                 avg_wire_size += 22420;
2631         } else {
2632                 /* plateau at a limit of 8K ints/sec */
2633                 avg_wire_size = 32256;
2634         }
2635 
2636         /* If we are in low latency mode half our delay which doubles the rate
2637          * to somewhere between 100K to 16K ints/sec
2638          */
2639         if (itr & IXGBE_ITR_ADAPTIVE_LATENCY)
2640                 avg_wire_size >>= 1;
2641 
2642         /* Resultant value is 256 times larger than it needs to be. This
2643          * gives us room to adjust the value as needed to either increase
2644          * or decrease the value based on link speeds of 10G, 2.5G, 1G, etc.
2645          *
2646          * Use addition as we have already recorded the new latency flag
2647          * for the ITR value.
2648          */
2649         switch (q_vector->adapter->link_speed) {
2650         case IXGBE_LINK_SPEED_10GB_FULL:
2651         case IXGBE_LINK_SPEED_100_FULL:
2652         default:
2653                 itr += DIV_ROUND_UP(avg_wire_size,
2654                                     IXGBE_ITR_ADAPTIVE_MIN_INC * 256) *
2655                        IXGBE_ITR_ADAPTIVE_MIN_INC;
2656                 break;
2657         case IXGBE_LINK_SPEED_2_5GB_FULL:
2658         case IXGBE_LINK_SPEED_1GB_FULL:
2659         case IXGBE_LINK_SPEED_10_FULL:
2660                 if (avg_wire_size > 8064)
2661                         avg_wire_size = 8064;
2662                 itr += DIV_ROUND_UP(avg_wire_size,
2663                                     IXGBE_ITR_ADAPTIVE_MIN_INC * 64) *
2664                        IXGBE_ITR_ADAPTIVE_MIN_INC;
2665                 break;
2666         }
2667 
2668 clear_counts:
2669         /* write back value */
2670         ring_container->itr = itr;
2671 
2672         /* next update should occur within next jiffy */
2673         ring_container->next_update = next_update + 1;
2674 
2675         ring_container->total_bytes = 0;
2676         ring_container->total_packets = 0;
2677 }
2678 
2679 /**
2680  * ixgbe_write_eitr - write EITR register in hardware specific way
2681  * @q_vector: structure containing interrupt and ring information
2682  *
2683  * This function is made to be called by ethtool and by the driver
2684  * when it needs to update EITR registers at runtime.  Hardware
2685  * specific quirks/differences are taken care of here.
2686  */
2687 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2688 {
2689         struct ixgbe_adapter *adapter = q_vector->adapter;
2690         struct ixgbe_hw *hw = &adapter->hw;
2691         int v_idx = q_vector->v_idx;
2692         u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2693 
2694         switch (adapter->hw.mac.type) {
2695         case ixgbe_mac_82598EB:
2696                 /* must write high and low 16 bits to reset counter */
2697                 itr_reg |= (itr_reg << 16);
2698                 break;
2699         case ixgbe_mac_82599EB:
2700         case ixgbe_mac_X540:
2701         case ixgbe_mac_X550:
2702         case ixgbe_mac_X550EM_x:
2703         case ixgbe_mac_x550em_a:
2704                 /*
2705                  * set the WDIS bit to not clear the timer bits and cause an
2706                  * immediate assertion of the interrupt
2707                  */
2708                 itr_reg |= IXGBE_EITR_CNT_WDIS;
2709                 break;
2710         default:
2711                 break;
2712         }
2713         IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2714 }
2715 
2716 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2717 {
2718         u32 new_itr;
2719 
2720         ixgbe_update_itr(q_vector, &q_vector->tx);
2721         ixgbe_update_itr(q_vector, &q_vector->rx);
2722 
2723         /* use the smallest value of new ITR delay calculations */
2724         new_itr = min(q_vector->rx.itr, q_vector->tx.itr);
2725 
2726         /* Clear latency flag if set, shift into correct position */
2727         new_itr &= ~IXGBE_ITR_ADAPTIVE_LATENCY;
2728         new_itr <<= 2;
2729 
2730         if (new_itr != q_vector->itr) {
2731                 /* save the algorithm value here */
2732                 q_vector->itr = new_itr;
2733 
2734                 ixgbe_write_eitr(q_vector);
2735         }
2736 }
2737 
2738 /**
2739  * ixgbe_check_overtemp_subtask - check for over temperature
2740  * @adapter: pointer to adapter
2741  **/
2742 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2743 {
2744         struct ixgbe_hw *hw = &adapter->hw;
2745         u32 eicr = adapter->interrupt_event;
2746         s32 rc;
2747 
2748         if (test_bit(__IXGBE_DOWN, &adapter->state))
2749                 return;
2750 
2751         if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2752                 return;
2753 
2754         adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2755 
2756         switch (hw->device_id) {
2757         case IXGBE_DEV_ID_82599_T3_LOM:
2758                 /*
2759                  * Since the warning interrupt is for both ports
2760                  * we don't have to check if:
2761                  *  - This interrupt wasn't for our port.
2762                  *  - We may have missed the interrupt so always have to
2763                  *    check if we  got a LSC
2764                  */
2765                 if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2766                     !(eicr & IXGBE_EICR_LSC))
2767                         return;
2768 
2769                 if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2770                         u32 speed;
2771                         bool link_up = false;
2772 
2773                         hw->mac.ops.check_link(hw, &speed, &link_up, false);
2774 
2775                         if (link_up)
2776                                 return;
2777                 }
2778 
2779                 /* Check if this is not due to overtemp */
2780                 if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2781                         return;
2782 
2783                 break;
2784         case IXGBE_DEV_ID_X550EM_A_1G_T:
2785         case IXGBE_DEV_ID_X550EM_A_1G_T_L:
2786                 rc = hw->phy.ops.check_overtemp(hw);
2787                 if (rc != IXGBE_ERR_OVERTEMP)
2788                         return;
2789                 break;
2790         default:
2791                 if (adapter->hw.mac.type >= ixgbe_mac_X540)
2792                         return;
2793                 if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2794                         return;
2795                 break;
2796         }
2797         e_crit(drv, "%s\n", ixgbe_overheat_msg);
2798 
2799         adapter->interrupt_event = 0;
2800 }
2801 
2802 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2803 {
2804         struct ixgbe_hw *hw = &adapter->hw;
2805 
2806         if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2807             (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2808                 e_crit(probe, "Fan has stopped, replace the adapter\n");
2809                 /* write to clear the interrupt */
2810                 IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2811         }
2812 }
2813 
2814 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2815 {
2816         struct ixgbe_hw *hw = &adapter->hw;
2817 
2818         if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2819                 return;
2820 
2821         switch (adapter->hw.mac.type) {
2822         case ixgbe_mac_82599EB:
2823                 /*
2824                  * Need to check link state so complete overtemp check
2825                  * on service task
2826                  */
2827                 if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2828                      (eicr & IXGBE_EICR_LSC)) &&
2829                     (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2830                         adapter->interrupt_event = eicr;
2831                         adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2832                         ixgbe_service_event_schedule(adapter);
2833                         return;
2834                 }
2835                 return;
2836         case ixgbe_mac_x550em_a:
2837                 if (eicr & IXGBE_EICR_GPI_SDP0_X550EM_a) {
2838                         adapter->interrupt_event = eicr;
2839                         adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2840                         ixgbe_service_event_schedule(adapter);
2841                         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC,
2842                                         IXGBE_EICR_GPI_SDP0_X550EM_a);
2843                         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICR,
2844                                         IXGBE_EICR_GPI_SDP0_X550EM_a);
2845                 }
2846                 return;
2847         case ixgbe_mac_X550:
2848         case ixgbe_mac_X540:
2849                 if (!(eicr & IXGBE_EICR_TS))
2850                         return;
2851                 break;
2852         default:
2853                 return;
2854         }
2855 
2856         e_crit(drv, "%s\n", ixgbe_overheat_msg);
2857 }
2858 
2859 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2860 {
2861         switch (hw->mac.type) {
2862         case ixgbe_mac_82598EB:
2863                 if (hw->phy.type == ixgbe_phy_nl)
2864                         return true;
2865                 return false;
2866         case ixgbe_mac_82599EB:
2867         case ixgbe_mac_X550EM_x:
2868         case ixgbe_mac_x550em_a:
2869                 switch (hw->mac.ops.get_media_type(hw)) {
2870                 case ixgbe_media_type_fiber:
2871                 case ixgbe_media_type_fiber_qsfp:
2872                         return true;
2873                 default:
2874                         return false;
2875                 }
2876         default:
2877                 return false;
2878         }
2879 }
2880 
2881 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2882 {
2883         struct ixgbe_hw *hw = &adapter->hw;
2884         u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2885 
2886         if (!ixgbe_is_sfp(hw))
2887                 return;
2888 
2889         /* Later MAC's use different SDP */
2890         if (hw->mac.type >= ixgbe_mac_X540)
2891                 eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2892 
2893         if (eicr & eicr_mask) {
2894                 /* Clear the interrupt */
2895                 IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2896                 if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2897                         adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2898                         adapter->sfp_poll_time = 0;
2899                         ixgbe_service_event_schedule(adapter);
2900                 }
2901         }
2902 
2903         if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2904             (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2905                 /* Clear the interrupt */
2906                 IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2907                 if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2908                         adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2909                         ixgbe_service_event_schedule(adapter);
2910                 }
2911         }
2912 }
2913 
2914 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2915 {
2916         struct ixgbe_hw *hw = &adapter->hw;
2917 
2918         adapter->lsc_int++;
2919         adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2920         adapter->link_check_timeout = jiffies;
2921         if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2922                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2923                 IXGBE_WRITE_FLUSH(hw);
2924                 ixgbe_service_event_schedule(adapter);
2925         }
2926 }
2927 
2928 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2929                                            u64 qmask)
2930 {
2931         u32 mask;
2932         struct ixgbe_hw *hw = &adapter->hw;
2933 
2934         switch (hw->mac.type) {
2935         case ixgbe_mac_82598EB:
2936                 mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2937                 IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2938                 break;
2939         case ixgbe_mac_82599EB:
2940         case ixgbe_mac_X540:
2941         case ixgbe_mac_X550:
2942         case ixgbe_mac_X550EM_x:
2943         case ixgbe_mac_x550em_a:
2944                 mask = (qmask & 0xFFFFFFFF);
2945                 if (mask)
2946                         IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2947                 mask = (qmask >> 32);
2948                 if (mask)
2949                         IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2950                 break;
2951         default:
2952                 break;
2953         }
2954         /* skip the flush */
2955 }
2956 
2957 static inline void ixgbe_irq_disable_queues(struct ixgbe_adapter *adapter,
2958                                             u64 qmask)
2959 {
2960         u32 mask;
2961         struct ixgbe_hw *hw = &adapter->hw;
2962 
2963         switch (hw->mac.type) {
2964         case ixgbe_mac_82598EB:
2965                 mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2966                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
2967                 break;
2968         case ixgbe_mac_82599EB:
2969         case ixgbe_mac_X540:
2970         case ixgbe_mac_X550:
2971         case ixgbe_mac_X550EM_x:
2972         case ixgbe_mac_x550em_a:
2973                 mask = (qmask & 0xFFFFFFFF);
2974                 if (mask)
2975                         IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
2976                 mask = (qmask >> 32);
2977                 if (mask)
2978                         IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
2979                 break;
2980         default:
2981                 break;
2982         }
2983         /* skip the flush */
2984 }
2985 
2986 /**
2987  * ixgbe_irq_enable - Enable default interrupt generation settings
2988  * @adapter: board private structure
2989  * @queues: enable irqs for queues
2990  * @flush: flush register write
2991  **/
2992 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2993                                     bool flush)
2994 {
2995         struct ixgbe_hw *hw = &adapter->hw;
2996         u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2997 
2998         /* don't reenable LSC while waiting for link */
2999         if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
3000                 mask &= ~IXGBE_EIMS_LSC;
3001 
3002         if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
3003                 switch (adapter->hw.mac.type) {
3004                 case ixgbe_mac_82599EB:
3005                         mask |= IXGBE_EIMS_GPI_SDP0(hw);
3006                         break;
3007                 case ixgbe_mac_X540:
3008                 case ixgbe_mac_X550:
3009                 case ixgbe_mac_X550EM_x:
3010                 case ixgbe_mac_x550em_a:
3011                         mask |= IXGBE_EIMS_TS;
3012                         break;
3013                 default:
3014                         break;
3015                 }
3016         if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
3017                 mask |= IXGBE_EIMS_GPI_SDP1(hw);
3018         switch (adapter->hw.mac.type) {
3019         case ixgbe_mac_82599EB:
3020                 mask |= IXGBE_EIMS_GPI_SDP1(hw);
3021                 mask |= IXGBE_EIMS_GPI_SDP2(hw);
3022                 /* fall through */
3023         case ixgbe_mac_X540:
3024         case ixgbe_mac_X550:
3025         case ixgbe_mac_X550EM_x:
3026         case ixgbe_mac_x550em_a:
3027                 if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
3028                     adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
3029                     adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
3030                         mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
3031                 if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
3032                         mask |= IXGBE_EICR_GPI_SDP0_X540;
3033                 mask |= IXGBE_EIMS_ECC;
3034                 mask |= IXGBE_EIMS_MAILBOX;
3035                 break;
3036         default:
3037                 break;
3038         }
3039 
3040         if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
3041             !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
3042                 mask |= IXGBE_EIMS_FLOW_DIR;
3043 
3044         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
3045         if (queues)
3046                 ixgbe_irq_enable_queues(adapter, ~0);
3047         if (flush)
3048                 IXGBE_WRITE_FLUSH(&adapter->hw);
3049 }
3050 
3051 static irqreturn_t ixgbe_msix_other(int irq, void *data)
3052 {
3053         struct ixgbe_adapter *adapter = data;
3054         struct ixgbe_hw *hw = &adapter->hw;
3055         u32 eicr;
3056 
3057         /*
3058          * Workaround for Silicon errata.  Use clear-by-write instead
3059          * of clear-by-read.  Reading with EICS will return the
3060          * interrupt causes without clearing, which later be done
3061          * with the write to EICR.
3062          */
3063         eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
3064 
3065         /* The lower 16bits of the EICR register are for the queue interrupts
3066          * which should be masked here in order to not accidentally clear them if
3067          * the bits are high when ixgbe_msix_other is called. There is a race
3068          * condition otherwise which results in possible performance loss
3069          * especially if the ixgbe_msix_other interrupt is triggering
3070          * consistently (as it would when PPS is turned on for the X540 device)
3071          */
3072         eicr &= 0xFFFF0000;
3073 
3074         IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
3075 
3076         if (eicr & IXGBE_EICR_LSC)
3077                 ixgbe_check_lsc(adapter);
3078 
3079         if (eicr & IXGBE_EICR_MAILBOX)
3080                 ixgbe_msg_task(adapter);
3081 
3082         switch (hw->mac.type) {
3083         case ixgbe_mac_82599EB:
3084         case ixgbe_mac_X540:
3085         case ixgbe_mac_X550:
3086         case ixgbe_mac_X550EM_x:
3087         case ixgbe_mac_x550em_a:
3088                 if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
3089                     (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
3090                         adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
3091                         ixgbe_service_event_schedule(adapter);
3092                         IXGBE_WRITE_REG(hw, IXGBE_EICR,
3093                                         IXGBE_EICR_GPI_SDP0_X540);
3094                 }
3095                 if (eicr & IXGBE_EICR_ECC) {
3096                         e_info(link, "Received ECC Err, initiating reset\n");
3097                         set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3098                         ixgbe_service_event_schedule(adapter);
3099                         IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3100                 }
3101                 /* Handle Flow Director Full threshold interrupt */
3102                 if (eicr & IXGBE_EICR_FLOW_DIR) {
3103                         int reinit_count = 0;
3104                         int i;
3105                         for (i = 0; i < adapter->num_tx_queues; i++) {
3106                                 struct ixgbe_ring *ring = adapter->tx_ring[i];
3107                                 if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
3108                                                        &ring->state))
3109                                         reinit_count++;
3110                         }
3111                         if (reinit_count) {
3112                                 /* no more flow director interrupts until after init */
3113                                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
3114                                 adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
3115                                 ixgbe_service_event_schedule(adapter);
3116                         }
3117                 }
3118                 ixgbe_check_sfp_event(adapter, eicr);
3119                 ixgbe_check_overtemp_event(adapter, eicr);
3120                 break;
3121         default:
3122                 break;
3123         }
3124 
3125         ixgbe_check_fan_failure(adapter, eicr);
3126 
3127         if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3128                 ixgbe_ptp_check_pps_event(adapter);
3129 
3130         /* re-enable the original interrupt state, no lsc, no queues */
3131         if (!test_bit(__IXGBE_DOWN, &adapter->state))
3132                 ixgbe_irq_enable(adapter, false, false);
3133 
3134         return IRQ_HANDLED;
3135 }
3136 
3137 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
3138 {
3139         struct ixgbe_q_vector *q_vector = data;
3140 
3141         /* EIAM disabled interrupts (on this vector) for us */
3142 
3143         if (q_vector->rx.ring || q_vector->tx.ring)
3144                 napi_schedule_irqoff(&q_vector->napi);
3145 
3146         return IRQ_HANDLED;
3147 }
3148 
3149 /**
3150  * ixgbe_poll - NAPI Rx polling callback
3151  * @napi: structure for representing this polling device
3152  * @budget: how many packets driver is allowed to clean
3153  *
3154  * This function is used for legacy and MSI, NAPI mode
3155  **/
3156 int ixgbe_poll(struct napi_struct *napi, int budget)
3157 {
3158         struct ixgbe_q_vector *q_vector =
3159                                 container_of(napi, struct ixgbe_q_vector, napi);
3160         struct ixgbe_adapter *adapter = q_vector->adapter;
3161         struct ixgbe_ring *ring;
3162         int per_ring_budget, work_done = 0;
3163         bool clean_complete = true;
3164 
3165 #ifdef CONFIG_IXGBE_DCA
3166         if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
3167                 ixgbe_update_dca(q_vector);
3168 #endif
3169 
3170         ixgbe_for_each_ring(ring, q_vector->tx) {
3171                 bool wd = ring->xsk_umem ?
3172                           ixgbe_clean_xdp_tx_irq(q_vector, ring, budget) :
3173                           ixgbe_clean_tx_irq(q_vector, ring, budget);
3174 
3175                 if (!wd)
3176                         clean_complete = false;
3177         }
3178 
3179         /* Exit if we are called by netpoll */
3180         if (budget <= 0)
3181                 return budget;
3182 
3183         /* attempt to distribute budget to each queue fairly, but don't allow
3184          * the budget to go below 1 because we'll exit polling */
3185         if (q_vector->rx.count > 1)
3186                 per_ring_budget = max(budget/q_vector->rx.count, 1);
3187         else
3188                 per_ring_budget = budget;
3189 
3190         ixgbe_for_each_ring(ring, q_vector->rx) {
3191                 int cleaned = ring->xsk_umem ?
3192                               ixgbe_clean_rx_irq_zc(q_vector, ring,
3193                                                     per_ring_budget) :
3194                               ixgbe_clean_rx_irq(q_vector, ring,
3195                                                  per_ring_budget);
3196 
3197                 work_done += cleaned;
3198                 if (cleaned >= per_ring_budget)
3199                         clean_complete = false;
3200         }
3201 
3202         /* If all work not completed, return budget and keep polling */
3203         if (!clean_complete)
3204                 return budget;
3205 
3206         /* all work done, exit the polling mode */
3207         if (likely(napi_complete_done(napi, work_done))) {
3208                 if (adapter->rx_itr_setting & 1)
3209                         ixgbe_set_itr(q_vector);
3210                 if (!test_bit(__IXGBE_DOWN, &adapter->state))
3211                         ixgbe_irq_enable_queues(adapter,
3212                                                 BIT_ULL(q_vector->v_idx));
3213         }
3214 
3215         return min(work_done, budget - 1);
3216 }
3217 
3218 /**
3219  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
3220  * @adapter: board private structure
3221  *
3222  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
3223  * interrupts from the kernel.
3224  **/
3225 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
3226 {
3227         struct net_device *netdev = adapter->netdev;
3228         unsigned int ri = 0, ti = 0;
3229         int vector, err;
3230 
3231         for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3232                 struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3233                 struct msix_entry *entry = &adapter->msix_entries[vector];
3234 
3235                 if (q_vector->tx.ring && q_vector->rx.ring) {
3236                         snprintf(q_vector->name, sizeof(q_vector->name),
3237                                  "%s-TxRx-%u", netdev->name, ri++);
3238                         ti++;
3239                 } else if (q_vector->rx.ring) {
3240                         snprintf(q_vector->name, sizeof(q_vector->name),
3241                                  "%s-rx-%u", netdev->name, ri++);
3242                 } else if (q_vector->tx.ring) {
3243                         snprintf(q_vector->name, sizeof(q_vector->name),
3244                                  "%s-tx-%u", netdev->name, ti++);
3245                 } else {
3246                         /* skip this unused q_vector */
3247                         continue;
3248                 }
3249                 err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
3250                                   q_vector->name, q_vector);
3251                 if (err) {
3252                         e_err(probe, "request_irq failed for MSIX interrupt "
3253                               "Error: %d\n", err);
3254                         goto free_queue_irqs;
3255                 }
3256                 /* If Flow Director is enabled, set interrupt affinity */
3257                 if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3258                         /* assign the mask for this irq */
3259                         irq_set_affinity_hint(entry->vector,
3260                                               &q_vector->affinity_mask);
3261                 }
3262         }
3263 
3264         err = request_irq(adapter->msix_entries[vector].vector,
3265                           ixgbe_msix_other, 0, netdev->name, adapter);
3266         if (err) {
3267                 e_err(probe, "request_irq for msix_other failed: %d\n", err);
3268                 goto free_queue_irqs;
3269         }
3270 
3271         return 0;
3272 
3273 free_queue_irqs:
3274         while (vector) {
3275                 vector--;
3276                 irq_set_affinity_hint(adapter->msix_entries[vector].vector,
3277                                       NULL);
3278                 free_irq(adapter->msix_entries[vector].vector,
3279                          adapter->q_vector[vector]);
3280         }
3281         adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
3282         pci_disable_msix(adapter->pdev);
3283         kfree(adapter->msix_entries);
3284         adapter->msix_entries = NULL;
3285         return err;
3286 }
3287 
3288 /**
3289  * ixgbe_intr - legacy mode Interrupt Handler
3290  * @irq: interrupt number
3291  * @data: pointer to a network interface device structure
3292  **/
3293 static irqreturn_t ixgbe_intr(int irq, void *data)
3294 {
3295         struct ixgbe_adapter *adapter = data;
3296         struct ixgbe_hw *hw = &adapter->hw;
3297         struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3298         u32 eicr;
3299 
3300         /*
3301          * Workaround for silicon errata #26 on 82598.  Mask the interrupt
3302          * before the read of EICR.
3303          */
3304         IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
3305 
3306         /* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
3307          * therefore no explicit interrupt disable is necessary */
3308         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3309         if (!eicr) {
3310                 /*
3311                  * shared interrupt alert!
3312                  * make sure interrupts are enabled because the read will
3313                  * have disabled interrupts due to EIAM
3314                  * finish the workaround of silicon errata on 82598.  Unmask
3315                  * the interrupt that we masked before the EICR read.
3316                  */
3317                 if (!test_bit(__IXGBE_DOWN, &adapter->state))
3318                         ixgbe_irq_enable(adapter, true, true);
3319                 return IRQ_NONE;        /* Not our interrupt */
3320         }
3321 
3322         if (eicr & IXGBE_EICR_LSC)
3323                 ixgbe_check_lsc(adapter);
3324 
3325         switch (hw->mac.type) {
3326         case ixgbe_mac_82599EB:
3327                 ixgbe_check_sfp_event(adapter, eicr);
3328                 /* Fall through */
3329         case ixgbe_mac_X540:
3330         case ixgbe_mac_X550:
3331         case ixgbe_mac_X550EM_x:
3332         case ixgbe_mac_x550em_a:
3333                 if (eicr & IXGBE_EICR_ECC) {
3334                         e_info(link, "Received ECC Err, initiating reset\n");
3335                         set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3336                         ixgbe_service_event_schedule(adapter);
3337                         IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3338                 }
3339                 ixgbe_check_overtemp_event(adapter, eicr);
3340                 break;
3341         default:
3342                 break;
3343         }
3344 
3345         ixgbe_check_fan_failure(adapter, eicr);
3346         if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3347                 ixgbe_ptp_check_pps_event(adapter);
3348 
3349         /* would disable interrupts here but EIAM disabled it */
3350         napi_schedule_irqoff(&q_vector->napi);
3351 
3352         /*
3353          * re-enable link(maybe) and non-queue interrupts, no flush.
3354          * ixgbe_poll will re-enable the queue interrupts
3355          */
3356         if (!test_bit(__IXGBE_DOWN, &adapter->state))
3357                 ixgbe_irq_enable(adapter, false, false);
3358 
3359         return IRQ_HANDLED;
3360 }
3361 
3362 /**
3363  * ixgbe_request_irq - initialize interrupts
3364  * @adapter: board private structure
3365  *
3366  * Attempts to configure interrupts using the best available
3367  * capabilities of the hardware and kernel.
3368  **/
3369 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3370 {
3371         struct net_device *netdev = adapter->netdev;
3372         int err;
3373 
3374         if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3375                 err = ixgbe_request_msix_irqs(adapter);
3376         else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3377                 err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3378                                   netdev->name, adapter);
3379         else
3380                 err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3381                                   netdev->name, adapter);
3382 
3383         if (err)
3384                 e_err(probe, "request_irq failed, Error %d\n", err);
3385 
3386         return err;
3387 }
3388 
3389 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3390 {
3391         int vector;
3392 
3393         if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3394                 free_irq(adapter->pdev->irq, adapter);
3395                 return;
3396         }
3397 
3398         if (!adapter->msix_entries)
3399                 return;
3400 
3401         for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3402                 struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3403                 struct msix_entry *entry = &adapter->msix_entries[vector];
3404 
3405                 /* free only the irqs that were actually requested */
3406                 if (!q_vector->rx.ring && !q_vector->tx.ring)
3407                         continue;
3408 
3409                 /* clear the affinity_mask in the IRQ descriptor */
3410                 irq_set_affinity_hint(entry->vector, NULL);
3411 
3412                 free_irq(entry->vector, q_vector);
3413         }
3414 
3415         free_irq(adapter->msix_entries[vector].vector, adapter);
3416 }
3417 
3418 /**
3419  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3420  * @adapter: board private structure
3421  **/
3422 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3423 {
3424         switch (adapter->hw.mac.type) {
3425         case ixgbe_mac_82598EB:
3426                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3427                 break;
3428         case ixgbe_mac_82599EB:
3429         case ixgbe_mac_X540:
3430         case ixgbe_mac_X550:
3431         case ixgbe_mac_X550EM_x:
3432         case ixgbe_mac_x550em_a:
3433                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3434                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3435                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3436                 break;
3437         default:
3438                 break;
3439         }
3440         IXGBE_WRITE_FLUSH(&adapter->hw);
3441         if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3442                 int vector;
3443 
3444                 for (vector = 0; vector < adapter->num_q_vectors; vector++)
3445                         synchronize_irq(adapter->msix_entries[vector].vector);
3446 
3447                 synchronize_irq(adapter->msix_entries[vector++].vector);
3448         } else {
3449                 synchronize_irq(adapter->pdev->irq);
3450         }
3451 }
3452 
3453 /**
3454  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3455  * @adapter: board private structure
3456  *
3457  **/
3458 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3459 {
3460         struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3461 
3462         ixgbe_write_eitr(q_vector);
3463 
3464         ixgbe_set_ivar(adapter, 0, 0, 0);
3465         ixgbe_set_ivar(adapter, 1, 0, 0);
3466 
3467         e_info(hw, "Legacy interrupt IVAR setup done\n");
3468 }
3469 
3470 /**
3471  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3472  * @adapter: board private structure
3473  * @ring: structure containing ring specific data
3474  *
3475  * Configure the Tx descriptor ring after a reset.
3476  **/
3477 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3478                              struct ixgbe_ring *ring)
3479 {
3480         struct ixgbe_hw *hw = &adapter->hw;
3481         u64 tdba = ring->dma;
3482         int wait_loop = 10;
3483         u32 txdctl = IXGBE_TXDCTL_ENABLE;
3484         u8 reg_idx = ring->reg_idx;
3485 
3486         ring->xsk_umem = NULL;
3487         if (ring_is_xdp(ring))
3488                 ring->xsk_umem = ixgbe_xsk_umem(adapter, ring);
3489 
3490         /* disable queue to avoid issues while updating state */
3491         IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3492         IXGBE_WRITE_FLUSH(hw);
3493 
3494         IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3495                         (tdba & DMA_BIT_MASK(32)));
3496         IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3497         IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3498                         ring->count * sizeof(union ixgbe_adv_tx_desc));
3499         IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3500         IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3501         ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3502 
3503         /*
3504          * set WTHRESH to encourage burst writeback, it should not be set
3505          * higher than 1 when:
3506          * - ITR is 0 as it could cause false TX hangs
3507          * - ITR is set to > 100k int/sec and BQL is enabled
3508          *
3509          * In order to avoid issues WTHRESH + PTHRESH should always be equal
3510          * to or less than the number of on chip descriptors, which is
3511          * currently 40.
3512          */
3513         if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3514                 txdctl |= 1u << 16;     /* WTHRESH = 1 */
3515         else
3516                 txdctl |= 8u << 16;     /* WTHRESH = 8 */
3517 
3518         /*
3519          * Setting PTHRESH to 32 both improves performance
3520          * and avoids a TX hang with DFP enabled
3521          */
3522         txdctl |= (1u << 8) |   /* HTHRESH = 1 */
3523                    32;          /* PTHRESH = 32 */
3524 
3525         /* reinitialize flowdirector state */
3526         if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3527                 ring->atr_sample_rate = adapter->atr_sample_rate;
3528                 ring->atr_count = 0;
3529                 set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3530         } else {
3531                 ring->atr_sample_rate = 0;
3532         }
3533 
3534         /* initialize XPS */
3535         if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3536                 struct ixgbe_q_vector *q_vector = ring->q_vector;
3537 
3538                 if (q_vector)
3539                         netif_set_xps_queue(ring->netdev,
3540                                             &q_vector->affinity_mask,
3541                                             ring->queue_index);
3542         }
3543 
3544         clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3545 
3546         /* reinitialize tx_buffer_info */
3547         memset(ring->tx_buffer_info, 0,
3548                sizeof(struct ixgbe_tx_buffer) * ring->count);
3549 
3550         /* enable queue */
3551         IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3552 
3553         /* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3554         if (hw->mac.type == ixgbe_mac_82598EB &&
3555             !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3556                 return;
3557 
3558         /* poll to verify queue is enabled */
3559         do {
3560                 usleep_range(1000, 2000);
3561                 txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3562         } while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3563         if (!wait_loop)
3564                 hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3565 }
3566 
3567 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3568 {
3569         struct ixgbe_hw *hw = &adapter->hw;
3570         u32 rttdcs, mtqc;
3571         u8 tcs = adapter->hw_tcs;
3572 
3573         if (hw->mac.type == ixgbe_mac_82598EB)
3574                 return;
3575 
3576         /* disable the arbiter while setting MTQC */
3577         rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3578         rttdcs |= IXGBE_RTTDCS_ARBDIS;
3579         IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3580 
3581         /* set transmit pool layout */
3582         if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3583                 mtqc = IXGBE_MTQC_VT_ENA;
3584                 if (tcs > 4)
3585                         mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3586                 else if (tcs > 1)
3587                         mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3588                 else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3589                          IXGBE_82599_VMDQ_4Q_MASK)
3590                         mtqc |= IXGBE_MTQC_32VF;
3591                 else
3592                         mtqc |= IXGBE_MTQC_64VF;
3593         } else {
3594                 if (tcs > 4) {
3595                         mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3596                 } else if (tcs > 1) {
3597                         mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3598                 } else {
3599                         u8 max_txq = adapter->num_tx_queues +
3600                                 adapter->num_xdp_queues;
3601                         if (max_txq > 63)
3602                                 mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3603                         else
3604                                 mtqc = IXGBE_MTQC_64Q_1PB;
3605                 }
3606         }
3607 
3608         IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3609 
3610         /* Enable Security TX Buffer IFG for multiple pb */
3611         if (tcs) {
3612                 u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3613                 sectx |= IXGBE_SECTX_DCB;
3614                 IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3615         }
3616 
3617         /* re-enable the arbiter */
3618         rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3619         IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3620 }
3621 
3622 /**
3623  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3624  * @adapter: board private structure
3625  *
3626  * Configure the Tx unit of the MAC after a reset.
3627  **/
3628 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3629 {
3630         struct ixgbe_hw *hw = &adapter->hw;
3631         u32 dmatxctl;
3632         u32 i;
3633 
3634         ixgbe_setup_mtqc(adapter);
3635 
3636         if (hw->mac.type != ixgbe_mac_82598EB) {
3637                 /* DMATXCTL.EN must be before Tx queues are enabled */
3638                 dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3639                 dmatxctl |= IXGBE_DMATXCTL_TE;
3640                 IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3641         }
3642 
3643         /* Setup the HW Tx Head and Tail descriptor pointers */
3644         for (i = 0; i < adapter->num_tx_queues; i++)
3645                 ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3646         for (i = 0; i < adapter->num_xdp_queues; i++)
3647                 ixgbe_configure_tx_ring(adapter, adapter->xdp_ring[i]);
3648 }
3649 
3650 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3651                                  struct ixgbe_ring *ring)
3652 {
3653         struct ixgbe_hw *hw = &adapter->hw;
3654         u8 reg_idx = ring->reg_idx;
3655         u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3656 
3657         srrctl |= IXGBE_SRRCTL_DROP_EN;
3658 
3659         IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3660 }
3661 
3662 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3663                                   struct ixgbe_ring *ring)
3664 {
3665         struct ixgbe_hw *hw = &adapter->hw;
3666         u8 reg_idx = ring->reg_idx;
3667         u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3668 
3669         srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3670 
3671         IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3672 }
3673 
3674 #ifdef CONFIG_IXGBE_DCB
3675 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3676 #else
3677 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3678 #endif
3679 {
3680         int i;
3681         bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3682 
3683         if (adapter->ixgbe_ieee_pfc)
3684                 pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3685 
3686         /*
3687          * We should set the drop enable bit if:
3688          *  SR-IOV is enabled
3689          *   or
3690          *  Number of Rx queues > 1 and flow control is disabled
3691          *
3692          *  This allows us to avoid head of line blocking for security
3693          *  and performance reasons.
3694          */
3695         if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3696             !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3697                 for (i = 0; i < adapter->num_rx_queues; i++)
3698                         ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3699         } else {
3700                 for (i = 0; i < adapter->num_rx_queues; i++)
3701                         ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3702         }
3703 }
3704 
3705 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3706 
3707 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3708                                    struct ixgbe_ring *rx_ring)
3709 {
3710         struct ixgbe_hw *hw = &adapter->hw;
3711         u32 srrctl;
3712         u8 reg_idx = rx_ring->reg_idx;
3713 
3714         if (hw->mac.type == ixgbe_mac_82598EB) {
3715                 u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3716 
3717                 /*
3718                  * if VMDq is not active we must program one srrctl register
3719                  * per RSS queue since we have enabled RDRXCTL.MVMEN
3720                  */
3721                 reg_idx &= mask;
3722         }
3723 
3724         /* configure header buffer length, needed for RSC */
3725         srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3726 
3727         /* configure the packet buffer length */
3728         if (rx_ring->xsk_umem) {
3729                 u32 xsk_buf_len = rx_ring->xsk_umem->chunk_size_nohr -
3730                                   XDP_PACKET_HEADROOM;
3731 
3732                 /* If the MAC support setting RXDCTL.RLPML, the
3733                  * SRRCTL[n].BSIZEPKT is set to PAGE_SIZE and
3734                  * RXDCTL.RLPML is set to the actual UMEM buffer
3735                  * size. If not, then we are stuck with a 1k buffer
3736                  * size resolution. In this case frames larger than
3737                  * the UMEM buffer size viewed in a 1k resolution will
3738                  * be dropped.
3739                  */
3740                 if (hw->mac.type != ixgbe_mac_82599EB)
3741                         srrctl |= PAGE_SIZE >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3742                 else
3743                         srrctl |= xsk_buf_len >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3744         } else if (test_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state)) {
3745                 srrctl |= IXGBE_RXBUFFER_3K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3746         } else {
3747                 srrctl |= IXGBE_RXBUFFER_2K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3748         }
3749 
3750         /* configure descriptor type */
3751         srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3752 
3753         IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3754 }
3755 
3756 /**
3757  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3758  * @adapter: device handle
3759  *
3760  *  - 82598/82599/X540:     128
3761  *  - X550(non-SRIOV mode): 512
3762  *  - X550(SRIOV mode):     64
3763  */
3764 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3765 {
3766         if (adapter->hw.mac.type < ixgbe_mac_X550)
3767                 return 128;
3768         else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3769                 return 64;
3770         else
3771                 return 512;
3772 }
3773 
3774 /**
3775  * ixgbe_store_key - Write the RSS key to HW
3776  * @adapter: device handle
3777  *
3778  * Write the RSS key stored in adapter.rss_key to HW.
3779  */
3780 void ixgbe_store_key(struct ixgbe_adapter *adapter)
3781 {
3782         struct ixgbe_hw *hw = &adapter->hw;
3783         int i;
3784 
3785         for (i = 0; i < 10; i++)
3786                 IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3787 }
3788 
3789 /**
3790  * ixgbe_init_rss_key - Initialize adapter RSS key
3791  * @adapter: device handle
3792  *
3793  * Allocates and initializes the RSS key if it is not allocated.
3794  **/
3795 static inline int ixgbe_init_rss_key(struct ixgbe_adapter *adapter)
3796 {
3797         u32 *rss_key;
3798 
3799         if (!adapter->rss_key) {
3800                 rss_key = kzalloc(IXGBE_RSS_KEY_SIZE, GFP_KERNEL);
3801                 if (unlikely(!rss_key))
3802                         return -ENOMEM;
3803 
3804                 netdev_rss_key_fill(rss_key, IXGBE_RSS_KEY_SIZE);
3805                 adapter->rss_key = rss_key;
3806         }
3807 
3808         return 0;
3809 }
3810 
3811 /**
3812  * ixgbe_store_reta - Write the RETA table to HW
3813  * @adapter: device handle
3814  *
3815  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3816  */
3817 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3818 {
3819         u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3820         struct ixgbe_hw *hw = &adapter->hw;
3821         u32 reta = 0;
3822         u32 indices_multi;
3823         u8 *indir_tbl = adapter->rss_indir_tbl;
3824 
3825         /* Fill out the redirection table as follows:
3826          *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3827          *    indices.
3828          *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3829          *  - X550:       8 bit wide entries containing 6 bit RSS index
3830          */
3831         if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3832                 indices_multi = 0x11;
3833         else
3834                 indices_multi = 0x1;
3835 
3836         /* Write redirection table to HW */
3837         for (i = 0; i < reta_entries; i++) {
3838                 reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3839                 if ((i & 3) == 3) {
3840                         if (i < 128)
3841                                 IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3842                         else
3843                                 IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3844                                                 reta);
3845                         reta = 0;
3846                 }
3847         }
3848 }
3849 
3850 /**
3851  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3852  * @adapter: device handle
3853  *
3854  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3855  */
3856 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3857 {
3858         u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3859         struct ixgbe_hw *hw = &adapter->hw;
3860         u32 vfreta = 0;
3861 
3862         /* Write redirection table to HW */
3863         for (i = 0; i < reta_entries; i++) {
3864                 u16 pool = adapter->num_rx_pools;
3865 
3866                 vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3867                 if ((i & 3) != 3)
3868                         continue;
3869 
3870                 while (pool--)
3871                         IXGBE_WRITE_REG(hw,
3872                                         IXGBE_PFVFRETA(i >> 2, VMDQ_P(pool)),
3873                                         vfreta);
3874                 vfreta = 0;
3875         }
3876 }
3877 
3878 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3879 {
3880         u32 i, j;
3881         u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3882         u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3883 
3884         /* Program table for at least 4 queues w/ SR-IOV so that VFs can
3885          * make full use of any rings they may have.  We will use the
3886          * PSRTYPE register to control how many rings we use within the PF.
3887          */
3888         if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
3889                 rss_i = 4;
3890 
3891         /* Fill out hash function seeds */
3892         ixgbe_store_key(adapter);
3893 
3894         /* Fill out redirection table */
3895         memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3896 
3897         for (i = 0, j = 0; i < reta_entries; i++, j++) {
3898                 if (j == rss_i)
3899                         j = 0;
3900 
3901                 adapter->rss_indir_tbl[i] = j;
3902         }
3903 
3904         ixgbe_store_reta(adapter);
3905 }
3906 
3907 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3908 {
3909         struct ixgbe_hw *hw = &adapter->hw;
3910         u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3911         int i, j;
3912 
3913         /* Fill out hash function seeds */
3914         for (i = 0; i < 10; i++) {
3915                 u16 pool = adapter->num_rx_pools;
3916 
3917                 while (pool--)
3918                         IXGBE_WRITE_REG(hw,
3919                                         IXGBE_PFVFRSSRK(i, VMDQ_P(pool)),
3920                                         *(adapter->rss_key + i));
3921         }
3922 
3923         /* Fill out the redirection table */
3924         for (i = 0, j = 0; i < 64; i++, j++) {
3925                 if (j == rss_i)
3926                         j = 0;
3927 
3928                 adapter->rss_indir_tbl[i] = j;
3929         }
3930 
3931         ixgbe_store_vfreta(adapter);
3932 }
3933 
3934 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3935 {
3936         struct ixgbe_hw *hw = &adapter->hw;
3937         u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3938         u32 rxcsum;
3939 
3940         /* Disable indicating checksum in descriptor, enables RSS hash */
3941         rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3942         rxcsum |= IXGBE_RXCSUM_PCSD;
3943         IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3944 
3945         if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3946                 if (adapter->ring_feature[RING_F_RSS].mask)
3947                         mrqc = IXGBE_MRQC_RSSEN;
3948         } else {
3949                 u8 tcs = adapter->hw_tcs;
3950 
3951                 if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3952                         if (tcs > 4)
3953                                 mrqc = IXGBE_MRQC_VMDQRT8TCEN;  /* 8 TCs */
3954                         else if (tcs > 1)
3955                                 mrqc = IXGBE_MRQC_VMDQRT4TCEN;  /* 4 TCs */
3956                         else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3957                                  IXGBE_82599_VMDQ_4Q_MASK)
3958                                 mrqc = IXGBE_MRQC_VMDQRSS32EN;
3959                         else
3960                                 mrqc = IXGBE_MRQC_VMDQRSS64EN;
3961 
3962                         /* Enable L3/L4 for Tx Switched packets only for X550,
3963                          * older devices do not support this feature
3964                          */
3965                         if (hw->mac.type >= ixgbe_mac_X550)
3966                                 mrqc |= IXGBE_MRQC_L3L4TXSWEN;
3967                 } else {
3968                         if (tcs > 4)
3969                                 mrqc = IXGBE_MRQC_RTRSS8TCEN;
3970                         else if (tcs > 1)
3971                                 mrqc = IXGBE_MRQC_RTRSS4TCEN;
3972                         else
3973                                 mrqc = IXGBE_MRQC_RSSEN;
3974                 }
3975         }
3976 
3977         /* Perform hash on these packet types */
3978         rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3979                      IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3980                      IXGBE_MRQC_RSS_FIELD_IPV6 |
3981                      IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3982 
3983         if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3984                 rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3985         if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3986                 rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3987 
3988         if ((hw->mac.type >= ixgbe_mac_X550) &&
3989             (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3990                 u16 pool = adapter->num_rx_pools;
3991 
3992                 /* Enable VF RSS mode */
3993                 mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3994                 IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3995 
3996                 /* Setup RSS through the VF registers */
3997                 ixgbe_setup_vfreta(adapter);
3998                 vfmrqc = IXGBE_MRQC_RSSEN;
3999                 vfmrqc |= rss_field;
4000 
4001                 while (pool--)
4002                         IXGBE_WRITE_REG(hw,
4003                                         IXGBE_PFVFMRQC(VMDQ_P(pool)),
4004                                         vfmrqc);
4005         } else {
4006                 ixgbe_setup_reta(adapter);
4007                 mrqc |= rss_field;
4008                 IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
4009         }
4010 }
4011 
4012 /**
4013  * ixgbe_configure_rscctl - enable RSC for the indicated ring
4014  * @adapter: address of board private structure
4015  * @ring: structure containing ring specific data
4016  **/
4017 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
4018                                    struct ixgbe_ring *ring)
4019 {
4020         struct ixgbe_hw *hw = &adapter->hw;
4021         u32 rscctrl;
4022         u8 reg_idx = ring->reg_idx;
4023 
4024         if (!ring_is_rsc_enabled(ring))
4025                 return;
4026 
4027         rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
4028         rscctrl |= IXGBE_RSCCTL_RSCEN;
4029         /*
4030          * we must limit the number of descriptors so that the
4031          * total size of max desc * buf_len is not greater
4032          * than 65536
4033          */
4034         rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
4035         IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
4036 }
4037 
4038 #define IXGBE_MAX_RX_DESC_POLL 10
4039 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
4040                                        struct ixgbe_ring *ring)
4041 {
4042         struct ixgbe_hw *hw = &adapter->hw;
4043         int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4044         u32 rxdctl;
4045         u8 reg_idx = ring->reg_idx;
4046 
4047         if (ixgbe_removed(hw->hw_addr))
4048                 return;
4049         /* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
4050         if (hw->mac.type == ixgbe_mac_82598EB &&
4051             !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4052                 return;
4053 
4054         do {
4055                 usleep_range(1000, 2000);
4056                 rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4057         } while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
4058 
4059         if (!wait_loop) {
4060                 e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
4061                       "the polling period\n", reg_idx);
4062         }
4063 }
4064 
4065 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
4066                              struct ixgbe_ring *ring)
4067 {
4068         struct ixgbe_hw *hw = &adapter->hw;
4069         union ixgbe_adv_rx_desc *rx_desc;
4070         u64 rdba = ring->dma;
4071         u32 rxdctl;
4072         u8 reg_idx = ring->reg_idx;
4073 
4074         xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
4075         ring->xsk_umem = ixgbe_xsk_umem(adapter, ring);
4076         if (ring->xsk_umem) {
4077                 ring->zca.free = ixgbe_zca_free;
4078                 WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4079                                                    MEM_TYPE_ZERO_COPY,
4080                                                    &ring->zca));
4081 
4082         } else {
4083                 WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4084                                                    MEM_TYPE_PAGE_SHARED, NULL));
4085         }
4086 
4087         /* disable queue to avoid use of these values while updating state */
4088         rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4089         rxdctl &= ~IXGBE_RXDCTL_ENABLE;
4090 
4091         /* write value back with RXDCTL.ENABLE bit cleared */
4092         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4093         IXGBE_WRITE_FLUSH(hw);
4094 
4095         IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
4096         IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
4097         IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
4098                         ring->count * sizeof(union ixgbe_adv_rx_desc));
4099         /* Force flushing of IXGBE_RDLEN to prevent MDD */
4100         IXGBE_WRITE_FLUSH(hw);
4101 
4102         IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
4103         IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
4104         ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
4105 
4106         ixgbe_configure_srrctl(adapter, ring);
4107         ixgbe_configure_rscctl(adapter, ring);
4108 
4109         if (hw->mac.type == ixgbe_mac_82598EB) {
4110                 /*
4111                  * enable cache line friendly hardware writes:
4112                  * PTHRESH=32 descriptors (half the internal cache),
4113                  * this also removes ugly rx_no_buffer_count increment
4114                  * HTHRESH=4 descriptors (to minimize latency on fetch)
4115                  * WTHRESH=8 burst writeback up to two cache lines
4116                  */
4117                 rxdctl &= ~0x3FFFFF;
4118                 rxdctl |=  0x080420;
4119 #if (PAGE_SIZE < 8192)
4120         /* RXDCTL.RLPML does not work on 82599 */
4121         } else if (hw->mac.type != ixgbe_mac_82599EB) {
4122                 rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4123                             IXGBE_RXDCTL_RLPML_EN);
4124 
4125                 /* Limit the maximum frame size so we don't overrun the skb.
4126                  * This can happen in SRIOV mode when the MTU of the VF is
4127                  * higher than the MTU of the PF.
4128                  */
4129                 if (ring_uses_build_skb(ring) &&
4130                     !test_bit(__IXGBE_RX_3K_BUFFER, &ring->state))
4131                         rxdctl |= IXGBE_MAX_2K_FRAME_BUILD_SKB |
4132                                   IXGBE_RXDCTL_RLPML_EN;
4133 #endif
4134         }
4135 
4136         if (ring->xsk_umem && hw->mac.type != ixgbe_mac_82599EB) {
4137                 u32 xsk_buf_len = ring->xsk_umem->chunk_size_nohr -
4138                                   XDP_PACKET_HEADROOM;
4139 
4140                 rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4141                             IXGBE_RXDCTL_RLPML_EN);
4142                 rxdctl |= xsk_buf_len | IXGBE_RXDCTL_RLPML_EN;
4143 
4144                 ring->rx_buf_len = xsk_buf_len;
4145         }
4146 
4147         /* initialize rx_buffer_info */
4148         memset(ring->rx_buffer_info, 0,
4149                sizeof(struct ixgbe_rx_buffer) * ring->count);
4150 
4151         /* initialize Rx descriptor 0 */
4152         rx_desc = IXGBE_RX_DESC(ring, 0);
4153         rx_desc->wb.upper.length = 0;
4154 
4155         /* enable receive descriptor ring */
4156         rxdctl |= IXGBE_RXDCTL_ENABLE;
4157         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4158 
4159         ixgbe_rx_desc_queue_enable(adapter, ring);
4160         if (ring->xsk_umem)
4161                 ixgbe_alloc_rx_buffers_zc(ring, ixgbe_desc_unused(ring));
4162         else
4163                 ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
4164 }
4165 
4166 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
4167 {
4168         struct ixgbe_hw *hw = &adapter->hw;
4169         int rss_i = adapter->ring_feature[RING_F_RSS].indices;
4170         u16 pool = adapter->num_rx_pools;
4171 
4172         /* PSRTYPE must be initialized in non 82598 adapters */
4173         u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4174                       IXGBE_PSRTYPE_UDPHDR |
4175                       IXGBE_PSRTYPE_IPV4HDR |
4176                       IXGBE_PSRTYPE_L2HDR |
4177                       IXGBE_PSRTYPE_IPV6HDR;
4178 
4179         if (hw->mac.type == ixgbe_mac_82598EB)
4180                 return;
4181 
4182         if (rss_i > 3)
4183                 psrtype |= 2u << 29;
4184         else if (rss_i > 1)
4185                 psrtype |= 1u << 29;
4186 
4187         while (pool--)
4188                 IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4189 }
4190 
4191 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
4192 {
4193         struct ixgbe_hw *hw = &adapter->hw;
4194         u16 pool = adapter->num_rx_pools;
4195         u32 reg_offset, vf_shift, vmolr;
4196         u32 gcr_ext, vmdctl;
4197         int i;
4198 
4199         if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
4200                 return;
4201 
4202         vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
4203         vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
4204         vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
4205         vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
4206         vmdctl |= IXGBE_VT_CTL_REPLEN;
4207         IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
4208 
4209         /* accept untagged packets until a vlan tag is
4210          * specifically set for the VMDQ queue/pool
4211          */
4212         vmolr = IXGBE_VMOLR_AUPE;
4213         while (pool--)
4214                 IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(pool)), vmolr);
4215 
4216         vf_shift = VMDQ_P(0) % 32;
4217         reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
4218 
4219         /* Enable only the PF's pool for Tx/Rx */
4220         IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
4221         IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
4222         IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
4223         IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
4224         if (adapter->bridge_mode == BRIDGE_MODE_VEB)
4225                 IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
4226 
4227         /* Map PF MAC address in RAR Entry 0 to first pool following VFs */
4228         hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
4229 
4230         /* clear VLAN promisc flag so VFTA will be updated if necessary */
4231         adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4232 
4233         /*
4234          * Set up VF register offsets for selected VT Mode,
4235          * i.e. 32 or 64 VFs for SR-IOV
4236          */
4237         switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4238         case IXGBE_82599_VMDQ_8Q_MASK:
4239                 gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
4240                 break;
4241         case IXGBE_82599_VMDQ_4Q_MASK:
4242                 gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
4243                 break;
4244         default:
4245                 gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
4246                 break;
4247         }
4248 
4249         IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
4250 
4251         for (i = 0; i < adapter->num_vfs; i++) {
4252                 /* configure spoof checking */
4253                 ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
4254                                           adapter->vfinfo[i].spoofchk_enabled);
4255 
4256                 /* Enable/Disable RSS query feature  */
4257                 ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
4258                                           adapter->vfinfo[i].rss_query_enabled);
4259         }
4260 }
4261 
4262 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
4263 {
4264         struct ixgbe_hw *hw = &adapter->hw;
4265         struct net_device *netdev = adapter->netdev;
4266         int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4267         struct ixgbe_ring *rx_ring;
4268         int i;
4269         u32 mhadd, hlreg0;
4270 
4271 #ifdef IXGBE_FCOE
4272         /* adjust max frame to be able to do baby jumbo for FCoE */
4273         if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
4274             (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
4275                 max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4276 
4277 #endif /* IXGBE_FCOE */
4278 
4279         /* adjust max frame to be at least the size of a standard frame */
4280         if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
4281                 max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
4282 
4283         mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
4284         if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
4285                 mhadd &= ~IXGBE_MHADD_MFS_MASK;
4286                 mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
4287 
4288                 IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
4289         }
4290 
4291         hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4292         /* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
4293         hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4294         IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4295 
4296         /*
4297          * Setup the HW Rx Head and Tail Descriptor Pointers and
4298          * the Base and Length of the Rx Descriptor Ring
4299          */
4300         for (i = 0; i < adapter->num_rx_queues; i++) {
4301                 rx_ring = adapter->rx_ring[i];
4302 
4303                 clear_ring_rsc_enabled(rx_ring);
4304                 clear_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4305                 clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4306 
4307                 if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4308                         set_ring_rsc_enabled(rx_ring);
4309 
4310                 if (test_bit(__IXGBE_RX_FCOE, &rx_ring->state))
4311                         set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4312 
4313                 if (adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
4314                         continue;
4315 
4316                 set_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4317 
4318 #if (PAGE_SIZE < 8192)
4319                 if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4320                         set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4321 
4322                 if (IXGBE_2K_TOO_SMALL_WITH_PADDING ||
4323                     (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
4324                         set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4325 #endif
4326         }
4327 }
4328 
4329 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
4330 {
4331         struct ixgbe_hw *hw = &adapter->hw;
4332         u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
4333 
4334         switch (hw->mac.type) {
4335         case ixgbe_mac_82598EB:
4336                 /*
4337                  * For VMDq support of different descriptor types or
4338                  * buffer sizes through the use of multiple SRRCTL
4339                  * registers, RDRXCTL.MVMEN must be set to 1
4340                  *
4341                  * also, the manual doesn't mention it clearly but DCA hints
4342                  * will only use queue 0's tags unless this bit is set.  Side
4343                  * effects of setting this bit are only that SRRCTL must be
4344                  * fully programmed [0..15]
4345                  */
4346                 rdrxctl |= IXGBE_RDRXCTL_MVMEN;
4347                 break;
4348         case ixgbe_mac_X550:
4349         case ixgbe_mac_X550EM_x:
4350         case ixgbe_mac_x550em_a:
4351                 if (adapter->num_vfs)
4352                         rdrxctl |= IXGBE_RDRXCTL_PSP;
4353                 /* fall through */
4354         case ixgbe_mac_82599EB:
4355         case ixgbe_mac_X540:
4356                 /* Disable RSC for ACK packets */
4357                 IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
4358                    (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
4359                 rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
4360                 /* hardware requires some bits to be set by default */
4361                 rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
4362                 rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
4363                 break;
4364         default:
4365                 /* We should do nothing since we don't know this hardware */
4366                 return;
4367         }
4368 
4369         IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
4370 }
4371 
4372 /**
4373  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
4374  * @adapter: board private structure
4375  *
4376  * Configure the Rx unit of the MAC after a reset.
4377  **/
4378 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
4379 {
4380         struct ixgbe_hw *hw = &adapter->hw;
4381         int i;
4382         u32 rxctrl, rfctl;
4383 
4384         /* disable receives while setting up the descriptors */
4385         hw->mac.ops.disable_rx(hw);
4386 
4387         ixgbe_setup_psrtype(adapter);
4388         ixgbe_setup_rdrxctl(adapter);
4389 
4390         /* RSC Setup */
4391         rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
4392         rfctl &= ~IXGBE_RFCTL_RSC_DIS;
4393         if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
4394                 rfctl |= IXGBE_RFCTL_RSC_DIS;
4395 
4396         /* disable NFS filtering */
4397         rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
4398         IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
4399 
4400         /* Program registers for the distribution of queues */
4401         ixgbe_setup_mrqc(adapter);
4402 
4403         /* set_rx_buffer_len must be called before ring initialization */
4404         ixgbe_set_rx_buffer_len(adapter);
4405 
4406         /*
4407          * Setup the HW Rx Head and Tail Descriptor Pointers and
4408          * the Base and Length of the Rx Descriptor Ring
4409          */
4410         for (i = 0; i < adapter->num_rx_queues; i++)
4411                 ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
4412 
4413         rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4414         /* disable drop enable for 82598 parts */
4415         if (hw->mac.type == ixgbe_mac_82598EB)
4416                 rxctrl |= IXGBE_RXCTRL_DMBYPS;
4417 
4418         /* enable all receives */
4419         rxctrl |= IXGBE_RXCTRL_RXEN;
4420         hw->mac.ops.enable_rx_dma(hw, rxctrl);
4421 }
4422 
4423 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
4424                                  __be16 proto, u16 vid)
4425 {
4426         struct ixgbe_adapter *adapter = netdev_priv(netdev);
4427         struct ixgbe_hw *hw = &adapter->hw;
4428 
4429         /* add VID to filter table */
4430         if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4431                 hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
4432 
4433         set_bit(vid, adapter->active_vlans);
4434 
4435         return 0;
4436 }
4437 
4438 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
4439 {
4440         u32 vlvf;
4441         int idx;
4442 
4443         /* short cut the special case */
4444         if (vlan == 0)
4445                 return 0;
4446 
4447         /* Search for the vlan id in the VLVF entries */
4448         for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
4449                 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
4450                 if ((vlvf & VLAN_VID_MASK) == vlan)
4451                         break;
4452         }
4453 
4454         return idx;
4455 }
4456 
4457 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
4458 {
4459         struct ixgbe_hw *hw = &adapter->hw;
4460         u32 bits, word;
4461         int idx;
4462 
4463         idx = ixgbe_find_vlvf_entry(hw, vid);
4464         if (!idx)
4465                 return;
4466 
4467         /* See if any other pools are set for this VLAN filter
4468          * entry other than the PF.
4469          */
4470         word = idx * 2 + (VMDQ_P(0) / 32);
4471         bits = ~BIT(VMDQ_P(0) % 32);
4472         bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4473 
4474         /* Disable the filter so this falls into the default pool. */
4475         if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4476                 if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4477                         IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4478                 IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4479         }
4480 }
4481 
4482 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4483                                   __be16 proto, u16 vid)
4484 {
4485         struct ixgbe_adapter *adapter = netdev_priv(netdev);
4486         struct ixgbe_hw *hw = &adapter->hw;
4487 
4488         /* remove VID from filter table */
4489         if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4490                 hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4491 
4492         clear_bit(vid, adapter->active_vlans);
4493 
4494         return 0;
4495 }
4496 
4497 /**
4498  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4499  * @adapter: driver data
4500  */
4501 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4502 {
4503         struct ixgbe_hw *hw = &adapter->hw;
4504         u32 vlnctrl;
4505         int i, j;
4506 
4507         switch (hw->mac.type) {
4508         case ixgbe_mac_82598EB:
4509                 vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4510                 vlnctrl &= ~IXGBE_VLNCTRL_VME;
4511                 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4512                 break;
4513         case ixgbe_mac_82599EB:
4514         case ixgbe_mac_X540:
4515         case ixgbe_mac_X550:
4516         case ixgbe_mac_X550EM_x:
4517         case ixgbe_mac_x550em_a:
4518                 for (i = 0; i < adapter->num_rx_queues; i++) {
4519                         struct ixgbe_ring *ring = adapter->rx_ring[i];
4520 
4521                         if (!netif_is_ixgbe(ring->netdev))
4522                                 continue;
4523 
4524                         j = ring->reg_idx;
4525                         vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4526                         vlnctrl &= ~IXGBE_RXDCTL_VME;
4527                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4528                 }
4529                 break;
4530         default:
4531                 break;
4532         }
4533 }
4534 
4535 /**
4536  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4537  * @adapter: driver data
4538  */
4539 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4540 {
4541         struct ixgbe_hw *hw = &adapter->hw;
4542         u32 vlnctrl;
4543         int i, j;
4544 
4545         switch (hw->mac.type) {
4546         case ixgbe_mac_82598EB:
4547                 vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4548                 vlnctrl |= IXGBE_VLNCTRL_VME;
4549                 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4550                 break;
4551         case ixgbe_mac_82599EB:
4552         case ixgbe_mac_X540:
4553         case ixgbe_mac_X550:
4554         case ixgbe_mac_X550EM_x:
4555         case ixgbe_mac_x550em_a:
4556                 for (i = 0; i < adapter->num_rx_queues; i++) {
4557                         struct ixgbe_ring *ring = adapter->rx_ring[i];
4558 
4559                         if (!netif_is_ixgbe(ring->netdev))
4560                                 continue;
4561 
4562                         j = ring->reg_idx;
4563                         vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4564                         vlnctrl |= IXGBE_RXDCTL_VME;
4565                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4566                 }
4567                 break;
4568         default:
4569                 break;
4570         }
4571 }
4572 
4573 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4574 {
4575         struct ixgbe_hw *hw = &adapter->hw;
4576         u32 vlnctrl, i;
4577 
4578         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4579 
4580         if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
4581         /* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4582                 vlnctrl |= IXGBE_VLNCTRL_VFE;
4583                 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4584         } else {
4585                 vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4586                 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4587                 return;
4588         }
4589 
4590         /* Nothing to do for 82598 */
4591         if (hw->mac.type == ixgbe_mac_82598EB)
4592                 return;
4593 
4594         /* We are already in VLAN promisc, nothing to do */
4595         if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4596                 return;
4597 
4598         /* Set flag so we don't redo unnecessary work */
4599         adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4600 
4601         /* Add PF to all active pools */
4602         for (i = IXGBE_VLVF_ENTRIES; --i;) {
4603                 u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4604                 u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4605 
4606                 vlvfb |= BIT(VMDQ_P(0) % 32);
4607                 IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4608         }
4609 
4610         /* Set all bits in the VLAN filter table array */
4611         for (i = hw->mac.vft_size; i--;)
4612                 IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4613 }
4614 
4615 #define VFTA_BLOCK_SIZE 8
4616 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4617 {
4618         struct ixgbe_hw *hw = &adapter->hw;
4619         u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4620         u32 vid_start = vfta_offset * 32;
4621         u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4622         u32 i, vid, word, bits;
4623 
4624         for (i = IXGBE_VLVF_ENTRIES; --i;) {
4625                 u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4626 
4627                 /* pull VLAN ID from VLVF */
4628                 vid = vlvf & VLAN_VID_MASK;
4629 
4630                 /* only concern outselves with a certain range */
4631                 if (vid < vid_start || vid >= vid_end)
4632                         continue;
4633 
4634                 if (vlvf) {
4635                         /* record VLAN ID in VFTA */
4636                         vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4637 
4638                         /* if PF is part of this then continue */
4639                         if (test_bit(vid, adapter->active_vlans))
4640                                 continue;
4641                 }
4642 
4643                 /* remove PF from the pool */
4644                 word = i * 2 + VMDQ_P(0) / 32;
4645                 bits = ~BIT(VMDQ_P(0) % 32);
4646                 bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4647                 IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4648         }
4649 
4650         /* extract values from active_vlans and write back to VFTA */
4651         for (i = VFTA_BLOCK_SIZE; i--;) {
4652                 vid = (vfta_offset + i) * 32;
4653                 word = vid / BITS_PER_LONG;
4654                 bits = vid % BITS_PER_LONG;
4655 
4656                 vfta[i] |= adapter->active_vlans[word] >> bits;
4657 
4658                 IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4659         }
4660 }
4661 
4662 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4663 {
4664         struct ixgbe_hw *hw = &adapter->hw;
4665         u32 vlnctrl, i;
4666 
4667         /* Set VLAN filtering to enabled */
4668         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4669         vlnctrl |= IXGBE_VLNCTRL_VFE;
4670         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4671 
4672         if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
4673             hw->mac.type == ixgbe_mac_82598EB)
4674                 return;
4675 
4676         /* We are not in VLAN promisc, nothing to do */
4677         if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4678                 return;
4679 
4680         /* Set flag so we don't redo unnecessary work */
4681         adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4682 
4683         for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4684                 ixgbe_scrub_vfta(adapter, i);
4685 }
4686 
4687 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4688 {
4689         u16 vid = 1;
4690 
4691         ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4692 
4693         for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4694                 ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4695 }
4696 
4697 /**
4698  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4699  * @netdev: network interface device structure
4700  *
4701  * Writes multicast address list to the MTA hash table.
4702  * Returns: -ENOMEM on failure
4703  *                0 on no addresses written
4704  *                X on writing X addresses to MTA
4705  **/
4706 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4707 {
4708         struct ixgbe_adapter *adapter = netdev_priv(netdev);
4709         struct ixgbe_hw *hw = &adapter->hw;
4710 
4711         if (!netif_running(netdev))
4712                 return 0;
4713 
4714         if (hw->mac.ops.update_mc_addr_list)
4715                 hw->mac.ops.update_mc_addr_list(hw, netdev);
4716         else
4717                 return -ENOMEM;
4718 
4719 #ifdef CONFIG_PCI_IOV
4720         ixgbe_restore_vf_multicasts(adapter);
4721 #endif
4722 
4723         return netdev_mc_count(netdev);
4724 }
4725 
4726 #ifdef CONFIG_PCI_IOV
4727 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4728 {
4729         struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4730         struct ixgbe_hw *hw = &adapter->hw;
4731         int i;
4732 
4733         for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4734                 mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4735 
4736                 if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4737                         hw->mac.ops.set_rar(hw, i,
4738                                             mac_table->addr,
4739                                             mac_table->pool,
4740                                             IXGBE_RAH_AV);
4741                 else
4742                         hw->mac.ops.clear_rar(hw, i);
4743         }
4744 }
4745 
4746 #endif
4747 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4748 {
4749         struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4750         struct ixgbe_hw *hw = &adapter->hw;
4751         int i;
4752 
4753         for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4754                 if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4755                         continue;
4756 
4757                 mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4758 
4759                 if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4760                         hw->mac.ops.set_rar(hw, i,
4761                                             mac_table->addr,
4762                                             mac_table->pool,
4763                                             IXGBE_RAH_AV);
4764                 else
4765                         hw->mac.ops.clear_rar(hw, i);
4766         }
4767 }
4768 
4769 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4770 {
4771         struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4772         struct ixgbe_hw *hw = &adapter->hw;
4773         int i;
4774 
4775         for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4776                 mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4777                 mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4778         }
4779 
4780         ixgbe_sync_mac_table(adapter);
4781 }
4782 
4783 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4784 {
4785         struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4786         struct ixgbe_hw *hw = &adapter->hw;
4787         int i, count = 0;
4788 
4789         for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4790                 /* do not count default RAR as available */
4791                 if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4792                         continue;
4793 
4794                 /* only count unused and addresses that belong to us */
4795                 if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4796                         if (mac_table->pool != pool)
4797                                 continue;
4798                 }
4799 
4800                 count++;
4801         }
4802 
4803         return count;
4804 }
4805 
4806 /* this function destroys the first RAR entry */
4807 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4808 {
4809         struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4810         struct ixgbe_hw *hw = &adapter->hw;
4811 
4812         memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4813         mac_table->pool = VMDQ_P(0);
4814 
4815         mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4816 
4817         hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4818                             IXGBE_RAH_AV);
4819 }
4820 
4821 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4822                          const u8 *addr, u16 pool)
4823 {
4824         struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4825         struct ixgbe_hw *hw = &adapter->hw;
4826         int i;
4827 
4828         if (is_zero_ether_addr(addr))
4829                 return -EINVAL;
4830 
4831         for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4832                 if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4833                         continue;
4834 
4835                 ether_addr_copy(mac_table->addr, addr);
4836                 mac_table->pool = pool;
4837 
4838                 mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4839                                     IXGBE_MAC_STATE_IN_USE;
4840 
4841                 ixgbe_sync_mac_table(adapter);
4842 
4843                 return i;
4844         }
4845 
4846         return -ENOMEM;
4847 }
4848 
4849 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4850                          const u8 *addr, u16 pool)
4851 {
4852         struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4853         struct ixgbe_hw *hw = &adapter->hw;
4854         int i;
4855 
4856         if (is_zero_ether_addr(addr))
4857                 return -EINVAL;
4858 
4859         /* search table for addr, if found clear IN_USE flag and sync */
4860         for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4861                 /* we can only delete an entry if it is in use */
4862                 if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4863                         continue;
4864                 /* we only care about entries that belong to the given pool */
4865                 if (mac_table->pool != pool)
4866                         continue;
4867                 /* we only care about a specific MAC address */
4868                 if (!ether_addr_equal(addr, mac_table->addr))
4869                         continue;
4870 
4871                 mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4872                 mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4873 
4874                 ixgbe_sync_mac_table(adapter);
4875 
4876                 return 0;
4877         }
4878 
4879         return -ENOMEM;
4880 }
4881 
4882 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4883 {
4884         struct ixgbe_adapter *adapter = netdev_priv(netdev);
4885         int ret;
4886 
4887         ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4888 
4889         return min_t(int, ret, 0);
4890 }
4891 
4892 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4893 {
4894         struct ixgbe_adapter *adapter = netdev_priv(netdev);
4895 
4896         ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4897 
4898         return 0;
4899 }
4900 
4901 /**
4902  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4903  * @netdev: network interface device structure
4904  *
4905  * The set_rx_method entry point is called whenever the unicast/multicast
4906  * address list or the network interface flags are updated.  This routine is
4907  * responsible for configuring the hardware for proper unicast, multicast and
4908  * promiscuous mode.
4909  **/
4910 void ixgbe_set_rx_mode(struct net_device *netdev)
4911 {
4912         struct ixgbe_adapter *adapter = netdev_priv(netdev);
4913         struct ixgbe_hw *hw = &adapter->hw;
4914         u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4915         netdev_features_t features = netdev->features;
4916         int count;
4917 
4918         /* Check for Promiscuous and All Multicast modes */
4919         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4920 
4921         /* set all bits that we expect to always be set */
4922         fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4923         fctrl |= IXGBE_FCTRL_BAM;
4924         fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4925         fctrl |= IXGBE_FCTRL_PMCF;
4926 
4927         /* clear the bits we are changing the status of */
4928         fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4929         if (netdev->flags & IFF_PROMISC) {
4930                 hw->addr_ctrl.user_set_promisc = true;
4931                 fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4932                 vmolr |= IXGBE_VMOLR_MPE;
4933                 features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4934         } else {
4935                 if (netdev->flags & IFF_ALLMULTI) {
4936                         fctrl |= IXGBE_FCTRL_MPE;
4937                         vmolr |= IXGBE_VMOLR_MPE;
4938                 }
4939                 hw->addr_ctrl.user_set_promisc = false;
4940         }
4941 
4942         /*
4943          * Write addresses to available RAR registers, if there is not
4944          * sufficient space to store all the addresses then enable
4945          * unicast promiscuous mode
4946          */
4947         if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4948                 fctrl |= IXGBE_FCTRL_UPE;
4949                 vmolr |= IXGBE_VMOLR_ROPE;
4950         }
4951 
4952         /* Write addresses to the MTA, if the attempt fails
4953          * then we should just turn on promiscuous mode so
4954          * that we can at least receive multicast traffic
4955          */
4956         count = ixgbe_write_mc_addr_list(netdev);
4957         if (count < 0) {
4958                 fctrl |= IXGBE_FCTRL_MPE;
4959                 vmolr |= IXGBE_VMOLR_MPE;
4960         } else if (count) {
4961                 vmolr |= IXGBE_VMOLR_ROMPE;
4962         }
4963 
4964         if (hw->mac.type != ixgbe_mac_82598EB) {
4965                 vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4966                          ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4967                            IXGBE_VMOLR_ROPE);
4968                 IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4969         }
4970 
4971         /* This is useful for sniffing bad packets. */
4972         if (features & NETIF_F_RXALL) {
4973                 /* UPE and MPE will be handled by normal PROMISC logic
4974                  * in e1000e_set_rx_mode */
4975                 fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4976                           IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4977                           IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4978 
4979                 fctrl &= ~(IXGBE_FCTRL_DPF);
4980                 /* NOTE:  VLAN filtering is disabled by setting PROMISC */
4981         }
4982 
4983         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4984 
4985         if (features & NETIF_F_HW_VLAN_CTAG_RX)
4986                 ixgbe_vlan_strip_enable(adapter);
4987         else
4988                 ixgbe_vlan_strip_disable(adapter);
4989 
4990         if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
4991                 ixgbe_vlan_promisc_disable(adapter);
4992         else
4993                 ixgbe_vlan_promisc_enable(adapter);
4994 }
4995 
4996 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4997 {
4998         int q_idx;
4999 
5000         for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
5001                 napi_enable(&adapter->q_vector[q_idx]->napi);
5002 }
5003 
5004 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
5005 {
5006         int q_idx;
5007 
5008         for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
5009                 napi_disable(&adapter->q_vector[q_idx]->napi);
5010 }
5011 
5012 static void ixgbe_clear_udp_tunnel_port(struct ixgbe_adapter *adapter, u32 mask)
5013 {
5014         struct ixgbe_hw *hw = &adapter->hw;
5015         u32 vxlanctrl;
5016 
5017         if (!(adapter->flags & (IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE |
5018                                 IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)))
5019                 return;
5020 
5021         vxlanctrl = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) & ~mask;
5022         IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, vxlanctrl);
5023 
5024         if (mask & IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK)
5025                 adapter->vxlan_port = 0;
5026 
5027         if (mask & IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK)
5028                 adapter->geneve_port = 0;
5029 }
5030 
5031 #ifdef CONFIG_IXGBE_DCB
5032 /**
5033  * ixgbe_configure_dcb - Configure DCB hardware
5034  * @adapter: ixgbe adapter struct
5035  *
5036  * This is called by the driver on open to configure the DCB hardware.
5037  * This is also called by the gennetlink interface when reconfiguring
5038  * the DCB state.
5039  */
5040 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
5041 {
5042         struct ixgbe_hw *hw = &adapter->hw;
5043         int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
5044 
5045         if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
5046                 if (hw->mac.type == ixgbe_mac_82598EB)
5047                         netif_set_gso_max_size(adapter->netdev, 65536);
5048                 return;
5049         }
5050 
5051         if (hw->mac.type == ixgbe_mac_82598EB)
5052                 netif_set_gso_max_size(adapter->netdev, 32768);
5053 
5054 #ifdef IXGBE_FCOE
5055         if (adapter->netdev->features & NETIF_F_FCOE_MTU)
5056                 max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
5057 #endif
5058 
5059         /* reconfigure the hardware */
5060         if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
5061                 ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5062                                                 DCB_TX_CONFIG);
5063                 ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5064                                                 DCB_RX_CONFIG);
5065                 ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
5066         } else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
5067                 ixgbe_dcb_hw_ets(&adapter->hw,
5068                                  adapter->ixgbe_ieee_ets,
5069                                  max_frame);
5070                 ixgbe_dcb_hw_pfc_config(&adapter->hw,
5071                                         adapter->ixgbe_ieee_pfc->pfc_en,
5072                                         adapter->ixgbe_ieee_ets->prio_tc);
5073         }
5074 
5075         /* Enable RSS Hash per TC */
5076         if (hw->mac.type != ixgbe_mac_82598EB) {
5077                 u32 msb = 0;
5078                 u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
5079 
5080                 while (rss_i) {
5081                         msb++;
5082                         rss_i >>= 1;
5083                 }
5084 
5085                 /* write msb to all 8 TCs in one write */
5086                 IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
5087         }
5088 }
5089 #endif
5090 
5091 /* Additional bittime to account for IXGBE framing */
5092 #define IXGBE_ETH_FRAMING 20
5093 
5094 /**
5095  * ixgbe_hpbthresh - calculate high water mark for flow control
5096  *
5097  * @adapter: board private structure to calculate for
5098  * @pb: packet buffer to calculate
5099  */
5100 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
5101 {
5102         struct ixgbe_hw *hw = &adapter->hw;
5103         struct net_device *dev = adapter->netdev;
5104         int link, tc, kb, marker;
5105         u32 dv_id, rx_pba;
5106 
5107         /* Calculate max LAN frame size */
5108         tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
5109 
5110 #ifdef IXGBE_FCOE
5111         /* FCoE traffic class uses FCOE jumbo frames */
5112         if ((dev->features & NETIF_F_FCOE_MTU) &&
5113             (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5114             (pb == ixgbe_fcoe_get_tc(adapter)))
5115                 tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5116 #endif
5117 
5118         /* Calculate delay value for device */
5119         switch (hw->mac.type) {
5120         case ixgbe_mac_X540:
5121         case ixgbe_mac_X550:
5122         case ixgbe_mac_X550EM_x:
5123         case ixgbe_mac_x550em_a:
5124                 dv_id = IXGBE_DV_X540(link, tc);
5125                 break;
5126         default:
5127                 dv_id = IXGBE_DV(link, tc);
5128                 break;
5129         }
5130 
5131         /* Loopback switch introduces additional latency */
5132         if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5133                 dv_id += IXGBE_B2BT(tc);
5134 
5135         /* Delay value is calculated in bit times convert to KB */
5136         kb = IXGBE_BT2KB(dv_id);
5137         rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
5138 
5139         marker = rx_pba - kb;
5140 
5141         /* It is possible that the packet buffer is not large enough
5142          * to provide required headroom. In this case throw an error
5143          * to user and a do the best we can.
5144          */
5145         if (marker < 0) {
5146                 e_warn(drv, "Packet Buffer(%i) can not provide enough"
5147                             "headroom to support flow control."
5148                             "Decrease MTU or number of traffic classes\n", pb);
5149                 marker = tc + 1;
5150         }
5151 
5152         return marker;
5153 }
5154 
5155 /**
5156  * ixgbe_lpbthresh - calculate low water mark for for flow control
5157  *
5158  * @adapter: board private structure to calculate for
5159  * @pb: packet buffer to calculate
5160  */
5161 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
5162 {
5163         struct ixgbe_hw *hw = &adapter->hw;
5164         struct net_device *dev = adapter->netdev;
5165         int tc;
5166         u32 dv_id;
5167 
5168         /* Calculate max LAN frame size */
5169         tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
5170 
5171 #ifdef IXGBE_FCOE
5172         /* FCoE traffic class uses FCOE jumbo frames */
5173         if ((dev->features & NETIF_F_FCOE_MTU) &&
5174             (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5175             (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
5176                 tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5177 #endif
5178 
5179         /* Calculate delay value for device */
5180         switch (hw->mac.type) {
5181         case ixgbe_mac_X540:
5182         case ixgbe_mac_X550:
5183         case ixgbe_mac_X550EM_x:
5184         case ixgbe_mac_x550em_a:
5185                 dv_id = IXGBE_LOW_DV_X540(tc);
5186                 break;
5187         default:
5188                 dv_id = IXGBE_LOW_DV(tc);
5189                 break;
5190         }
5191 
5192         /* Delay value is calculated in bit times convert to KB */
5193         return IXGBE_BT2KB(dv_id);
5194 }
5195 
5196 /*
5197  * ixgbe_pbthresh_setup - calculate and setup high low water marks
5198  */
5199 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
5200 {
5201         struct ixgbe_hw *hw = &adapter->hw;
5202         int num_tc = adapter->hw_tcs;
5203         int i;
5204 
5205         if (!num_tc)
5206                 num_tc = 1;
5207 
5208         for (i = 0; i < num_tc; i++) {
5209                 hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
5210                 hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
5211 
5212                 /* Low water marks must not be larger than high water marks */
5213                 if (hw->fc.low_water[i] > hw->fc.high_water[i])
5214                         hw->fc.low_water[i] = 0;
5215         }
5216 
5217         for (; i < MAX_TRAFFIC_CLASS; i++)
5218                 hw->fc.high_water[i] = 0;
5219 }
5220 
5221 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
5222 {
5223         struct ixgbe_hw *hw = &adapter->hw;
5224         int hdrm;
5225         u8 tc = adapter->hw_tcs;
5226 
5227         if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
5228             adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
5229                 hdrm = 32 << adapter->fdir_pballoc;
5230         else
5231                 hdrm = 0;
5232 
5233         hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
5234         ixgbe_pbthresh_setup(adapter);
5235 }
5236 
5237 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
5238 {
5239         struct ixgbe_hw *hw = &adapter->hw;
5240         struct hlist_node *node2;
5241         struct ixgbe_fdir_filter *filter;
5242         u8 queue;
5243 
5244         spin_lock(&adapter->fdir_perfect_lock);
5245 
5246         if (!hlist_empty(&adapter->fdir_filter_list))
5247                 ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
5248 
5249         hlist_for_each_entry_safe(filter, node2,
5250                                   &adapter->fdir_filter_list, fdir_node) {
5251                 if (filter->action == IXGBE_FDIR_DROP_QUEUE) {
5252                         queue = IXGBE_FDIR_DROP_QUEUE;
5253                 } else {
5254                         u32 ring = ethtool_get_flow_spec_ring(filter->action);
5255                         u8 vf = ethtool_get_flow_spec_ring_vf(filter->action);
5256 
5257                         if (!vf && (ring >= adapter->num_rx_queues)) {
5258                                 e_err(drv, "FDIR restore failed without VF, ring: %u\n",
5259                                       ring);
5260                                 continue;
5261                         } else if (vf &&
5262                                    ((vf > adapter->num_vfs) ||
5263                                      ring >= adapter->num_rx_queues_per_pool)) {
5264                                 e_err(drv, "FDIR restore failed with VF, vf: %hhu, ring: %u\n",
5265                                       vf, ring);
5266                                 continue;
5267                         }
5268 
5269                         /* Map the ring onto the absolute queue index */
5270                         if (!vf)
5271                                 queue = adapter->rx_ring[ring]->reg_idx;
5272                         else
5273                                 queue = ((vf - 1) *
5274                                         adapter->num_rx_queues_per_pool) + ring;
5275                 }
5276 
5277                 ixgbe_fdir_write_perfect_filter_82599(hw,
5278                                 &filter->filter, filter->sw_idx, queue);
5279         }
5280 
5281         spin_unlock(&adapter->fdir_perfect_lock);
5282 }
5283 
5284 /**
5285  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
5286  * @rx_ring: ring to free buffers from
5287  **/
5288 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
5289 {
5290         u16 i = rx_ring->next_to_clean;
5291         struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
5292 
5293         if (rx_ring->xsk_umem) {
5294                 ixgbe_xsk_clean_rx_ring(rx_ring);
5295                 goto skip_free;
5296         }
5297 
5298         /* Free all the Rx ring sk_buffs */
5299         while (i != rx_ring->next_to_alloc) {
5300                 if (rx_buffer->skb) {
5301                         struct sk_buff *skb = rx_buffer->skb;
5302                         if (IXGBE_CB(skb)->page_released)
5303                                 dma_unmap_page_attrs(rx_ring->dev,
5304                                                      IXGBE_CB(skb)->dma,
5305                                                      ixgbe_rx_pg_size(rx_ring),
5306                                                      DMA_FROM_DEVICE,
5307                                                      IXGBE_RX_DMA_ATTR);
5308                         dev_kfree_skb(skb);
5309                 }
5310 
5311                 /* Invalidate cache lines that may have been written to by
5312                  * device so that we avoid corrupting memory.
5313                  */
5314                 dma_sync_single_range_for_cpu(rx_ring->dev,
5315                                               rx_buffer->dma,
5316                                               rx_buffer->page_offset,
5317                                               ixgbe_rx_bufsz(rx_ring),
5318                                               DMA_FROM_DEVICE);
5319 
5320                 /* free resources associated with mapping */
5321                 dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
5322                                      ixgbe_rx_pg_size(rx_ring),
5323                                      DMA_FROM_DEVICE,
5324                                      IXGBE_RX_DMA_ATTR);
5325                 __page_frag_cache_drain(rx_buffer->page,
5326                                         rx_buffer->pagecnt_bias);
5327 
5328                 i++;
5329                 rx_buffer++;
5330                 if (i == rx_ring->count) {
5331                         i = 0;
5332                         rx_buffer = rx_ring->rx_buffer_info;
5333                 }
5334         }
5335 
5336 skip_free:
5337         rx_ring->next_to_alloc = 0;
5338         rx_ring->next_to_clean = 0;
5339         rx_ring->next_to_use = 0;
5340 }
5341 
5342 static int ixgbe_fwd_ring_up(struct ixgbe_adapter *adapter,
5343                              struct ixgbe_fwd_adapter *accel)
5344 {
5345         u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
5346         int num_tc = netdev_get_num_tc(adapter->netdev);
5347         struct net_device *vdev = accel->netdev;
5348         int i, baseq, err;
5349 
5350         baseq = accel->pool * adapter->num_rx_queues_per_pool;
5351         netdev_dbg(vdev, "pool %i:%i queues %i:%i\n",
5352                    accel->pool, adapter->num_rx_pools,
5353                    baseq, baseq + adapter->num_rx_queues_per_pool);
5354 
5355         accel->rx_base_queue = baseq;
5356         accel->tx_base_queue = baseq;
5357 
5358         /* record configuration for macvlan interface in vdev */
5359         for (i = 0; i < num_tc; i++)
5360                 netdev_bind_sb_channel_queue(adapter->netdev, vdev,
5361                                              i, rss_i, baseq + (rss_i * i));
5362 
5363         for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5364                 adapter->rx_ring[baseq + i]->netdev = vdev;
5365 
5366         /* Guarantee all rings are updated before we update the
5367          * MAC address filter.
5368          */
5369         wmb();
5370 
5371         /* ixgbe_add_mac_filter will return an index if it succeeds, so we
5372          * need to only treat it as an error value if it is negative.
5373          */
5374         err = ixgbe_add_mac_filter(adapter, vdev->dev_addr,
5375                                    VMDQ_P(accel->pool));
5376         if (err >= 0)
5377                 return 0;
5378 
5379         /* if we cannot add the MAC rule then disable the offload */
5380         macvlan_release_l2fw_offload(vdev);
5381 
5382         for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5383                 adapter->rx_ring[baseq + i]->netdev = NULL;
5384 
5385         netdev_err(vdev, "L2FW offload disabled due to L2 filter error\n");
5386 
5387         /* unbind the queues and drop the subordinate channel config */
5388         netdev_unbind_sb_channel(adapter->netdev, vdev);
5389         netdev_set_sb_channel(vdev, 0);
5390 
5391         clear_bit(accel->pool, adapter->fwd_bitmask);
5392         kfree(accel);
5393 
5394         return err;
5395 }
5396 
5397 static int ixgbe_macvlan_up(struct net_device *vdev, void *data)
5398 {
5399         struct ixgbe_adapter *adapter = data;
5400         struct ixgbe_fwd_adapter *accel;
5401 
5402         if (!netif_is_macvlan(vdev))
5403                 return 0;
5404 
5405         accel = macvlan_accel_priv(vdev);
5406         if (!accel)
5407                 return 0;
5408 
5409         ixgbe_fwd_ring_up(adapter, accel);
5410 
5411         return 0;
5412 }
5413 
5414 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5415 {
5416         netdev_walk_all_upper_dev_rcu(adapter->netdev,
5417                                       ixgbe_macvlan_up, adapter);
5418 }
5419 
5420 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5421 {
5422         struct ixgbe_hw *hw = &adapter->hw;
5423 
5424         ixgbe_configure_pb(adapter);
5425 #ifdef CONFIG_IXGBE_DCB
5426         ixgbe_configure_dcb(adapter);
5427 #endif
5428         /*
5429          * We must restore virtualization before VLANs or else
5430          * the VLVF registers will not be populated
5431          */
5432         ixgbe_configure_virtualization(adapter);
5433 
5434         ixgbe_set_rx_mode(adapter->netdev);
5435         ixgbe_restore_vlan(adapter);
5436         ixgbe_ipsec_restore(adapter);
5437 
5438         switch (hw->mac.type) {
5439         case ixgbe_mac_82599EB:
5440         case ixgbe_mac_X540:
5441                 hw->mac.ops.disable_rx_buff(hw);
5442                 break;
5443         default:
5444                 break;
5445         }
5446 
5447         if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5448                 ixgbe_init_fdir_signature_82599(&adapter->hw,
5449                                                 adapter->fdir_pballoc);
5450         } else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5451                 ixgbe_init_fdir_perfect_82599(&adapter->hw,
5452                                               adapter->fdir_pballoc);
5453                 ixgbe_fdir_filter_restore(adapter);
5454         }
5455 
5456         switch (hw->mac.type) {
5457         case ixgbe_mac_82599EB:
5458         case ixgbe_mac_X540:
5459                 hw->mac.ops.enable_rx_buff(hw);
5460                 break;
5461         default:
5462                 break;
5463         }
5464 
5465 #ifdef CONFIG_IXGBE_DCA
5466         /* configure DCA */
5467         if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5468                 ixgbe_setup_dca(adapter);
5469 #endif /* CONFIG_IXGBE_DCA */
5470 
5471 #ifdef IXGBE_FCOE
5472         /* configure FCoE L2 filters, redirection table, and Rx control */
5473         ixgbe_configure_fcoe(adapter);
5474 
5475 #endif /* IXGBE_FCOE */
5476         ixgbe_configure_tx(adapter);
5477         ixgbe_configure_rx(adapter);
5478         ixgbe_configure_dfwd(adapter);
5479 }
5480 
5481 /**
5482  * ixgbe_sfp_link_config - set up SFP+ link
5483  * @adapter: pointer to private adapter struct
5484  **/
5485 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5486 {
5487         /*
5488          * We are assuming the worst case scenario here, and that
5489          * is that an SFP was inserted/removed after the reset
5490          * but before SFP detection was enabled.  As such the best
5491          * solution is to just start searching as soon as we start
5492          */
5493         if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5494                 adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5495 
5496         adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5497         adapter->sfp_poll_time = 0;
5498 }
5499 
5500 /**
5501  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5502  * @hw: pointer to private hardware struct
5503  *
5504  * Returns 0 on success, negative on failure
5505  **/
5506 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5507 {
5508         u32 speed;
5509         bool autoneg, link_up = false;
5510         int ret = IXGBE_ERR_LINK_SETUP;
5511 
5512         if (hw->mac.ops.check_link)
5513                 ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5514 
5515         if (ret)
5516                 return ret;
5517 
5518         speed = hw->phy.autoneg_advertised;
5519         if ((!speed) && (hw->mac.ops.get_link_capabilities))
5520                 ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5521                                                         &autoneg);
5522         if (ret)
5523                 return ret;
5524 
5525         if (hw->mac.ops.setup_link)
5526                 ret = hw->mac.ops.setup_link(hw, speed, link_up);
5527 
5528         return ret;
5529 }
5530 
5531 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5532 {
5533         struct ixgbe_hw *hw = &adapter->hw;
5534         u32 gpie = 0;
5535 
5536         if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5537                 gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5538                        IXGBE_GPIE_OCD;
5539                 gpie |= IXGBE_GPIE_EIAME;
5540                 /*
5541                  * use EIAM to auto-mask when MSI-X interrupt is asserted
5542                  * this saves a register write for every interrupt
5543                  */
5544                 switch (hw->mac.type) {
5545                 case ixgbe_mac_82598EB:
5546                         IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5547                         break;
5548                 case ixgbe_mac_82599EB:
5549                 case ixgbe_mac_X540:
5550                 case ixgbe_mac_X550:
5551                 case ixgbe_mac_X550EM_x:
5552                 case ixgbe_mac_x550em_a:
5553                 default:
5554                         IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5555                         IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5556                         break;
5557                 }
5558         } else {
5559                 /* legacy interrupts, use EIAM to auto-mask when reading EICR,
5560                  * specifically only auto mask tx and rx interrupts */
5561                 IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5562         }
5563 
5564         /* XXX: to interrupt immediately for EICS writes, enable this */
5565         /* gpie |= IXGBE_GPIE_EIMEN; */
5566 
5567         if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5568                 gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5569 
5570                 switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5571                 case IXGBE_82599_VMDQ_8Q_MASK:
5572                         gpie |= IXGBE_GPIE_VTMODE_16;
5573                         break;
5574                 case IXGBE_82599_VMDQ_4Q_MASK:
5575                         gpie |= IXGBE_GPIE_VTMODE_32;
5576                         break;
5577                 default:
5578                         gpie |= IXGBE_GPIE_VTMODE_64;
5579                         break;
5580                 }
5581         }
5582 
5583         /* Enable Thermal over heat sensor interrupt */
5584         if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5585                 switch (adapter->hw.mac.type) {
5586                 case ixgbe_mac_82599EB:
5587                         gpie |= IXGBE_SDP0_GPIEN_8259X;
5588                         break;
5589                 default:
5590                         break;
5591                 }
5592         }
5593 
5594         /* Enable fan failure interrupt */
5595         if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5596                 gpie |= IXGBE_SDP1_GPIEN(hw);
5597 
5598         switch (hw->mac.type) {
5599         case ixgbe_mac_82599EB:
5600                 gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5601                 break;
5602         case ixgbe_mac_X550EM_x:
5603         case ixgbe_mac_x550em_a:
5604                 gpie |= IXGBE_SDP0_GPIEN_X540;
5605                 break;
5606         default:
5607                 break;
5608         }
5609 
5610         IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5611 }
5612 
5613 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5614 {
5615         struct ixgbe_hw *hw = &adapter->hw;
5616         int err;
5617         u32 ctrl_ext;
5618 
5619         ixgbe_get_hw_control(adapter);
5620         ixgbe_setup_gpie(adapter);
5621 
5622         if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5623                 ixgbe_configure_msix(adapter);
5624         else
5625                 ixgbe_configure_msi_and_legacy(adapter);
5626 
5627         /* enable the optics for 82599 SFP+ fiber */
5628         if (hw->mac.ops.enable_tx_laser)
5629                 hw->mac.ops.enable_tx_laser(hw);
5630 
5631         if (hw->phy.ops.set_phy_power)
5632                 hw->phy.ops.set_phy_power(hw, true);
5633 
5634         smp_mb__before_atomic();
5635         clear_bit(__IXGBE_DOWN, &adapter->state);
5636         ixgbe_napi_enable_all(adapter);
5637 
5638         if (ixgbe_is_sfp(hw)) {
5639                 ixgbe_sfp_link_config(adapter);
5640         } else {
5641                 err = ixgbe_non_sfp_link_config(hw);
5642                 if (err)
5643                         e_err(probe, "link_config FAILED %d\n", err);
5644         }
5645 
5646         /* clear any pending interrupts, may auto mask */
5647         IXGBE_READ_REG(hw, IXGBE_EICR);
5648         ixgbe_irq_enable(adapter, true, true);
5649 
5650         /*
5651          * If this adapter has a fan, check to see if we had a failure
5652          * before we enabled the interrupt.
5653          */
5654         if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5655                 u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5656                 if (esdp & IXGBE_ESDP_SDP1)
5657                         e_crit(drv, "Fan has stopped, replace the adapter\n");
5658         }
5659 
5660         /* bring the link up in the watchdog, this could race with our first
5661          * link up interrupt but shouldn't be a problem */
5662         adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5663         adapter->link_check_timeout = jiffies;
5664         mod_timer(&adapter->service_timer, jiffies);
5665 
5666         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
5667         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5668         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5669         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5670 }
5671 
5672 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5673 {
5674         WARN_ON(in_interrupt());
5675         /* put off any impending NetWatchDogTimeout */
5676         netif_trans_update(adapter->netdev);
5677 
5678         while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5679                 usleep_range(1000, 2000);
5680         if (adapter->hw.phy.type == ixgbe_phy_fw)
5681                 ixgbe_watchdog_link_is_down(adapter);
5682         ixgbe_down(adapter);
5683         /*
5684          * If SR-IOV enabled then wait a bit before bringing the adapter
5685          * back up to give the VFs time to respond to the reset.  The
5686          * two second wait is based upon the watchdog timer cycle in
5687          * the VF driver.
5688          */
5689         if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5690                 msleep(2000);
5691         ixgbe_up(adapter);
5692         clear_bit(__IXGBE_RESETTING, &adapter->state);
5693 }
5694 
5695 void ixgbe_up(struct ixgbe_adapter *adapter)
5696 {
5697         /* hardware has been reset, we need to reload some things */
5698         ixgbe_configure(adapter);
5699 
5700         ixgbe_up_complete(adapter);
5701 }
5702 
5703 static unsigned long ixgbe_get_completion_timeout(struct ixgbe_adapter *adapter)
5704 {
5705         u16 devctl2;
5706 
5707         pcie_capability_read_word(adapter->pdev, PCI_EXP_DEVCTL2, &devctl2);
5708 
5709         switch (devctl2 & IXGBE_PCIDEVCTRL2_TIMEO_MASK) {
5710         case IXGBE_PCIDEVCTRL2_17_34s:
5711         case IXGBE_PCIDEVCTRL2_4_8s:
5712                 /* For now we cap the upper limit on delay to 2 seconds
5713                  * as we end up going up to 34 seconds of delay in worst
5714                  * case timeout value.
5715                  */
5716         case IXGBE_PCIDEVCTRL2_1_2s:
5717                 return 2000000ul;       /* 2.0 s */
5718         case IXGBE_PCIDEVCTRL2_260_520ms:
5719                 return 520000ul;        /* 520 ms */
5720         case IXGBE_PCIDEVCTRL2_65_130ms:
5721                 return 130000ul;        /* 130 ms */
5722         case IXGBE_PCIDEVCTRL2_16_32ms:
5723                 return 32000ul;         /* 32 ms */
5724         case IXGBE_PCIDEVCTRL2_1_2ms:
5725                 return 2000ul;          /* 2 ms */
5726         case IXGBE_PCIDEVCTRL2_50_100us:
5727                 return 100ul;           /* 100 us */
5728         case IXGBE_PCIDEVCTRL2_16_32ms_def:
5729                 return 32000ul;         /* 32 ms */
5730         default:
5731                 break;
5732         }
5733 
5734         /* We shouldn't need to hit this path, but just in case default as
5735          * though completion timeout is not supported and support 32ms.
5736          */
5737         return 32000ul;
5738 }
5739 
5740 void ixgbe_disable_rx(struct ixgbe_adapter *adapter)
5741 {
5742         unsigned long wait_delay, delay_interval;
5743         struct ixgbe_hw *hw = &adapter->hw;
5744         int i, wait_loop;
5745         u32 rxdctl;
5746 
5747         /* disable receives */
5748         hw->mac.ops.disable_rx(hw);
5749 
5750         if (ixgbe_removed(hw->hw_addr))
5751                 return;
5752 
5753         /* disable all enabled Rx queues */
5754         for (i = 0; i < adapter->num_rx_queues; i++) {
5755                 struct ixgbe_ring *ring = adapter->rx_ring[i];
5756                 u8 reg_idx = ring->reg_idx;
5757 
5758                 rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
5759                 rxdctl &= ~IXGBE_RXDCTL_ENABLE;
5760                 rxdctl |= IXGBE_RXDCTL_SWFLSH;
5761 
5762                 /* write value back with RXDCTL.ENABLE bit cleared */
5763                 IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
5764         }
5765 
5766         /* RXDCTL.EN may not change on 82598 if link is down, so skip it */
5767         if (hw->mac.type == ixgbe_mac_82598EB &&
5768             !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
5769                 return;
5770 
5771         /* Determine our minimum delay interval. We will increase this value
5772          * with each subsequent test. This way if the device returns quickly
5773          * we should spend as little time as possible waiting, however as
5774          * the time increases we will wait for larger periods of time.
5775          *
5776          * The trick here is that we increase the interval using the
5777          * following pattern: 1x 3x 5x 7x 9x 11x 13x 15x 17x 19x. The result
5778          * of that wait is that it totals up to 100x whatever interval we
5779          * choose. Since our minimum wait is 100us we can just divide the
5780          * total timeout by 100 to get our minimum delay interval.
5781          */
5782         delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
5783 
5784         wait_loop = IXGBE_MAX_RX_DESC_POLL;
5785         wait_delay = delay_interval;
5786 
5787         while (wait_loop--) {
5788                 usleep_range(wait_delay, wait_delay + 10);
5789                 wait_delay += delay_interval * 2;
5790                 rxdctl = 0;
5791 
5792                 /* OR together the reading of all the active RXDCTL registers,
5793                  * and then test the result. We need the disable to complete
5794                  * before we start freeing the memory and invalidating the
5795                  * DMA mappings.
5796                  */
5797                 for (i = 0; i < adapter->num_rx_queues; i++) {
5798                         struct ixgbe_ring *ring = adapter->rx_ring[i];
5799                         u8 reg_idx = ring->reg_idx;
5800 
5801                         rxdctl |= IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
5802                 }
5803 
5804                 if (!(rxdctl & IXGBE_RXDCTL_ENABLE))
5805                         return;
5806         }
5807 
5808         e_err(drv,
5809               "RXDCTL.ENABLE for one or more queues not cleared within the polling period\n");
5810 }
5811 
5812 void ixgbe_disable_tx(struct ixgbe_adapter *adapter)
5813 {
5814         unsigned long wait_delay, delay_interval;
5815         struct ixgbe_hw *hw = &adapter->hw;
5816         int i, wait_loop;
5817         u32 txdctl;
5818 
5819         if (ixgbe_removed(hw->hw_addr))
5820                 return;
5821 
5822         /* disable all enabled Tx queues */
5823         for (i = 0; i < adapter->num_tx_queues; i++) {
5824                 struct ixgbe_ring *ring = adapter->tx_ring[i];
5825                 u8 reg_idx = ring->reg_idx;
5826 
5827                 IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5828         }
5829 
5830         /* disable all enabled XDP Tx queues */
5831         for (i = 0; i < adapter->num_xdp_queues; i++) {
5832                 struct ixgbe_ring *ring = adapter->xdp_ring[i];
5833                 u8 reg_idx = ring->reg_idx;
5834 
5835                 IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5836         }
5837 
5838         /* If the link is not up there shouldn't be much in the way of
5839          * pending transactions. Those that are left will be flushed out
5840          * when the reset logic goes through the flush sequence to clean out
5841          * the pending Tx transactions.
5842          */
5843         if (!(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
5844                 goto dma_engine_disable;
5845 
5846         /* Determine our minimum delay interval. We will increase this value
5847          * with each subsequent test. This way if the device returns quickly
5848          * we should spend as little time as possible waiting, however as
5849          * the time increases we will wait for larger periods of time.
5850          *
5851          * The trick here is that we increase the interval using the
5852          * following pattern: 1x 3x 5x 7x 9x 11x 13x 15x 17x 19x. The result
5853          * of that wait is that it totals up to 100x whatever interval we
5854          * choose. Since our minimum wait is 100us we can just divide the
5855          * total timeout by 100 to get our minimum delay interval.
5856          */
5857         delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
5858 
5859         wait_loop = IXGBE_MAX_RX_DESC_POLL;
5860         wait_delay = delay_interval;
5861 
5862         while (wait_loop--) {
5863                 usleep_range(wait_delay, wait_delay + 10);
5864                 wait_delay += delay_interval * 2;
5865                 txdctl = 0;
5866 
5867                 /* OR together the reading of all the active TXDCTL registers,
5868                  * and then test the result. We need the disable to complete
5869                  * before we start freeing the memory and invalidating the
5870                  * DMA mappings.
5871                  */
5872                 for (i = 0; i < adapter->num_tx_queues; i++) {
5873                         struct ixgbe_ring *ring = adapter->tx_ring[i];
5874                         u8 reg_idx = ring->reg_idx;
5875 
5876                         txdctl |= IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
5877                 }
5878                 for (i = 0; i < adapter->num_xdp_queues; i++) {
5879                         struct ixgbe_ring *ring = adapter->xdp_ring[i];
5880                         u8 reg_idx = ring->reg_idx;
5881 
5882                         txdctl |= IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
5883                 }
5884 
5885                 if (!(txdctl & IXGBE_TXDCTL_ENABLE))
5886                         goto dma_engine_disable;
5887         }
5888 
5889         e_err(drv,
5890               "TXDCTL.ENABLE for one or more queues not cleared within the polling period\n");
5891 
5892 dma_engine_disable:
5893         /* Disable the Tx DMA engine on 82599 and later MAC */
5894         switch (hw->mac.type) {
5895         case ixgbe_mac_82599EB:
5896         case ixgbe_mac_X540:
5897         case ixgbe_mac_X550:
5898         case ixgbe_mac_X550EM_x:
5899         case ixgbe_mac_x550em_a:
5900                 IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5901                                 (IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5902                                  ~IXGBE_DMATXCTL_TE));
5903                 /* fall through */
5904         default:
5905                 break;
5906         }
5907 }
5908 
5909 void ixgbe_reset(struct ixgbe_adapter *adapter)
5910 {
5911         struct ixgbe_hw *hw = &adapter->hw;
5912         struct net_device *netdev = adapter->netdev;
5913         int err;
5914 
5915         if (ixgbe_removed(hw->hw_addr))
5916                 return;
5917         /* lock SFP init bit to prevent race conditions with the watchdog */
5918         while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5919                 usleep_range(1000, 2000);
5920 
5921         /* clear all SFP and link config related flags while holding SFP_INIT */
5922         adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5923                              IXGBE_FLAG2_SFP_NEEDS_RESET);
5924         adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5925 
5926         err = hw->mac.ops.init_hw(hw);
5927         switch (err) {
5928         case 0:
5929         case IXGBE_ERR_SFP_NOT_PRESENT:
5930         case IXGBE_ERR_SFP_NOT_SUPPORTED:
5931                 break;
5932         case IXGBE_ERR_MASTER_REQUESTS_PENDING:
5933                 e_dev_err("master disable timed out\n");
5934                 break;
5935         case IXGBE_ERR_EEPROM_VERSION:
5936                 /* We are running on a pre-production device, log a warning */
5937                 e_dev_warn("This device is a pre-production adapter/LOM. "
5938                            "Please be aware there may be issues associated with "
5939                            "your hardware.  If you are experiencing problems "
5940                            "please contact your Intel or hardware "
5941                            "representative who provided you with this "
5942                            "hardware.\n");
5943                 break;
5944         default:
5945                 e_dev_err("Hardware Error: %d\n", err);
5946         }
5947 
5948         clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
5949 
5950         /* flush entries out of MAC table */
5951         ixgbe_flush_sw_mac_table(adapter);
5952         __dev_uc_unsync(netdev, NULL);
5953 
5954         /* do not flush user set addresses */
5955         ixgbe_mac_set_default_filter(adapter);
5956 
5957         /* update SAN MAC vmdq pool selection */
5958         if (hw->mac.san_mac_rar_index)
5959                 hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
5960 
5961         if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
5962                 ixgbe_ptp_reset(adapter);
5963 
5964         if (hw->phy.ops.set_phy_power) {
5965                 if (!netif_running(adapter->netdev) && !adapter->wol)
5966                         hw->phy.ops.set_phy_power(hw, false);
5967                 else
5968                         hw->phy.ops.set_phy_power(hw, true);
5969         }
5970 }
5971 
5972 /**
5973  * ixgbe_clean_tx_ring - Free Tx Buffers
5974  * @tx_ring: ring to be cleaned
5975  **/
5976 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
5977 {
5978         u16 i = tx_ring->next_to_clean;
5979         struct ixgbe_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
5980 
5981         if (tx_ring->xsk_umem) {
5982                 ixgbe_xsk_clean_tx_ring(tx_ring);
5983                 goto out;
5984         }
5985 
5986         while (i != tx_ring->next_to_use) {
5987                 union ixgbe_adv_tx_desc *eop_desc, *tx_desc;
5988 
5989                 /* Free all the Tx ring sk_buffs */
5990                 if (ring_is_xdp(tx_ring))
5991                         xdp_return_frame(tx_buffer->xdpf);
5992                 else
5993                         dev_kfree_skb_any(tx_buffer->skb);
5994 
5995                 /* unmap skb header data */
5996                 dma_unmap_single(tx_ring->dev,
5997                                  dma_unmap_addr(tx_buffer, dma),
5998                                  dma_unmap_len(tx_buffer, len),
5999                                  DMA_TO_DEVICE);
6000 
6001                 /* check for eop_desc to determine the end of the packet */
6002                 eop_desc = tx_buffer->next_to_watch;
6003                 tx_desc = IXGBE_TX_DESC(tx_ring, i);
6004 
6005                 /* unmap remaining buffers */
6006                 while (tx_desc != eop_desc) {
6007                         tx_buffer++;
6008                         tx_desc++;
6009                         i++;
6010                         if (unlikely(i == tx_ring->count)) {
6011                                 i = 0;
6012                                 tx_buffer = tx_ring->tx_buffer_info;
6013                                 tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6014                         }
6015 
6016                         /* unmap any remaining paged data */
6017                         if (dma_unmap_len(tx_buffer, len))
6018                                 dma_unmap_page(tx_ring->dev,
6019                                                dma_unmap_addr(tx_buffer, dma),
6020                                                dma_unmap_len(tx_buffer, len),
6021                                                DMA_TO_DEVICE);
6022                 }
6023 
6024                 /* move us one more past the eop_desc for start of next pkt */
6025                 tx_buffer++;
6026                 i++;
6027                 if (unlikely(i == tx_ring->count)) {
6028                         i = 0;
6029                         tx_buffer = tx_ring->tx_buffer_info;
6030                 }
6031         }
6032 
6033         /* reset BQL for queue */
6034         if (!ring_is_xdp(tx_ring))
6035                 netdev_tx_reset_queue(txring_txq(tx_ring));
6036 
6037 out:
6038         /* reset next_to_use and next_to_clean */
6039         tx_ring->next_to_use = 0;
6040         tx_ring->next_to_clean = 0;
6041 }
6042 
6043 /**
6044  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
6045  * @adapter: board private structure
6046  **/
6047 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
6048 {
6049         int i;
6050 
6051         for (i = 0; i < adapter->num_rx_queues; i++)
6052                 ixgbe_clean_rx_ring(adapter->rx_ring[i]);
6053 }
6054 
6055 /**
6056  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
6057  * @adapter: board private structure
6058  **/
6059 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
6060 {
6061         int i;
6062 
6063         for (i = 0; i < adapter->num_tx_queues; i++)
6064                 ixgbe_clean_tx_ring(adapter->tx_ring[i]);
6065         for (i = 0; i < adapter->num_xdp_queues; i++)
6066                 ixgbe_clean_tx_ring(adapter->xdp_ring[i]);
6067 }
6068 
6069 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
6070 {
6071         struct hlist_node *node2;
6072         struct ixgbe_fdir_filter *filter;
6073 
6074         spin_lock(&adapter->fdir_perfect_lock);
6075 
6076         hlist_for_each_entry_safe(filter, node2,
6077                                   &adapter->fdir_filter_list, fdir_node) {
6078                 hlist_del(&filter->fdir_node);
6079                 kfree(filter);
6080         }
6081         adapter->fdir_filter_count = 0;
6082 
6083         spin_unlock(&adapter->fdir_perfect_lock);
6084 }
6085 
6086 void ixgbe_down(struct ixgbe_adapter *adapter)
6087 {
6088         struct net_device *netdev = adapter->netdev;
6089         struct ixgbe_hw *hw = &adapter->hw;
6090         int i;
6091 
6092         /* signal that we are down to the interrupt handler */
6093         if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
6094                 return; /* do nothing if already down */
6095 
6096         /* Shut off incoming Tx traffic */
6097         netif_tx_stop_all_queues(netdev);
6098 
6099         /* call carrier off first to avoid false dev_watchdog timeouts */
6100         netif_carrier_off(netdev);
6101         netif_tx_disable(netdev);
6102 
6103         /* Disable Rx */
6104         ixgbe_disable_rx(adapter);
6105 
6106         /* synchronize_rcu() needed for pending XDP buffers to drain */
6107         if (adapter->xdp_ring[0])
6108                 synchronize_rcu();
6109 
6110         ixgbe_irq_disable(adapter);
6111 
6112         ixgbe_napi_disable_all(adapter);
6113 
6114         clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
6115         adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6116         adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6117 
6118         del_timer_sync(&adapter->service_timer);
6119 
6120         if (adapter->num_vfs) {
6121                 /* Clear EITR Select mapping */
6122                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
6123 
6124                 /* Mark all the VFs as inactive */
6125                 for (i = 0 ; i < adapter->num_vfs; i++)
6126                         adapter->vfinfo[i].clear_to_send = false;
6127 
6128                 /* ping all the active vfs to let them know we are going down */
6129                 ixgbe_ping_all_vfs(adapter);
6130 
6131                 /* Disable all VFTE/VFRE TX/RX */
6132                 ixgbe_disable_tx_rx(adapter);
6133         }
6134 
6135         /* disable transmits in the hardware now that interrupts are off */
6136         ixgbe_disable_tx(adapter);
6137 
6138         if (!pci_channel_offline(adapter->pdev))
6139                 ixgbe_reset(adapter);
6140 
6141         /* power down the optics for 82599 SFP+ fiber */
6142         if (hw->mac.ops.disable_tx_laser)
6143                 hw->mac.ops.disable_tx_laser(hw);
6144 
6145         ixgbe_clean_all_tx_rings(adapter);
6146         ixgbe_clean_all_rx_rings(adapter);
6147 }
6148 
6149 /**
6150  * ixgbe_eee_capable - helper function to determine EEE support on X550
6151  * @adapter: board private structure
6152  */
6153 static void ixgbe_set_eee_capable(struct ixgbe_adapter *adapter)
6154 {
6155         struct ixgbe_hw *hw = &adapter->hw;
6156 
6157         switch (hw->device_id) {
6158         case IXGBE_DEV_ID_X550EM_A_1G_T:
6159         case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6160                 if (!hw->phy.eee_speeds_supported)
6161                         break;
6162                 adapter->flags2 |= IXGBE_FLAG2_EEE_CAPABLE;
6163                 if (!hw->phy.eee_speeds_advertised)
6164                         break;
6165                 adapter->flags2 |= IXGBE_FLAG2_EEE_ENABLED;
6166                 break;
6167         default:
6168                 adapter->flags2 &= ~IXGBE_FLAG2_EEE_CAPABLE;
6169                 adapter->flags2 &= ~IXGBE_FLAG2_EEE_ENABLED;
6170                 break;
6171         }
6172 }
6173 
6174 /**
6175  * ixgbe_tx_timeout - Respond to a Tx Hang
6176  * @netdev: network interface device structure
6177  **/
6178 static void ixgbe_tx_timeout(struct net_device *netdev)
6179 {
6180         struct ixgbe_adapter *adapter = netdev_priv(netdev);
6181 
6182         /* Do the reset outside of interrupt context */
6183         ixgbe_tx_timeout_reset(adapter);
6184 }
6185 
6186 #ifdef CONFIG_IXGBE_DCB
6187 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
6188 {
6189         struct ixgbe_hw *hw = &adapter->hw;
6190         struct tc_configuration *tc;
6191         int j;
6192 
6193         switch (hw->mac.type) {
6194         case ixgbe_mac_82598EB:
6195         case ixgbe_mac_82599EB:
6196                 adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
6197                 adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
6198                 break;
6199         case ixgbe_mac_X540:
6200         case ixgbe_mac_X550:
6201                 adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
6202                 adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
6203                 break;
6204         case ixgbe_mac_X550EM_x:
6205         case ixgbe_mac_x550em_a:
6206         default:
6207                 adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
6208                 adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
6209                 break;
6210         }
6211 
6212         /* Configure DCB traffic classes */
6213         for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
6214                 tc = &adapter->dcb_cfg.tc_config[j];
6215                 tc->path[DCB_TX_CONFIG].bwg_id = 0;
6216                 tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
6217                 tc->path[DCB_RX_CONFIG].bwg_id = 0;
6218                 tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
6219                 tc->dcb_pfc = pfc_disabled;
6220         }
6221 
6222         /* Initialize default user to priority mapping, UPx->TC0 */
6223         tc = &adapter->dcb_cfg.tc_config[0];
6224         tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
6225         tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
6226 
6227         adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
6228         adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
6229         adapter->dcb_cfg.pfc_mode_enable = false;
6230         adapter->dcb_set_bitmap = 0x00;
6231         if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
6232                 adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
6233         memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
6234                sizeof(adapter->temp_dcb_cfg));
6235 }
6236 #endif
6237 
6238 /**
6239  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
6240  * @adapter: board private structure to initialize
6241  * @ii: pointer to ixgbe_info for device
6242  *
6243  * ixgbe_sw_init initializes the Adapter private data structure.
6244  * Fields are initialized based on PCI device information and
6245  * OS network device settings (MTU size).
6246  **/
6247 static int ixgbe_sw_init(struct ixgbe_adapter *adapter,
6248                          const struct ixgbe_info *ii)
6249 {
6250         struct ixgbe_hw *hw = &adapter->hw;
6251         struct pci_dev *pdev = adapter->pdev;
6252         unsigned int rss, fdir;
6253         u32 fwsm;
6254         int i;
6255 
6256         /* PCI config space info */
6257 
6258         hw->vendor_id = pdev->vendor;
6259         hw->device_id = pdev->device;
6260         hw->revision_id = pdev->revision;
6261         hw->subsystem_vendor_id = pdev->subsystem_vendor;
6262         hw->subsystem_device_id = pdev->subsystem_device;
6263 
6264         /* get_invariants needs the device IDs */
6265         ii->get_invariants(hw);
6266 
6267         /* Set common capability flags and settings */
6268         rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
6269         adapter->ring_feature[RING_F_RSS].limit = rss;
6270         adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
6271         adapter->max_q_vectors = MAX_Q_VECTORS_82599;
6272         adapter->atr_sample_rate = 20;
6273         fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
6274         adapter->ring_feature[RING_F_FDIR].limit = fdir;
6275         adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
6276         adapter->ring_feature[RING_F_VMDQ].limit = 1;
6277 #ifdef CONFIG_IXGBE_DCA
6278         adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
6279 #endif
6280 #ifdef CONFIG_IXGBE_DCB
6281         adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
6282         adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6283 #endif
6284 #ifdef IXGBE_FCOE
6285         adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
6286         adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6287 #ifdef CONFIG_IXGBE_DCB
6288         /* Default traffic class to use for FCoE */
6289         adapter->fcoe.up = IXGBE_FCOE_DEFTC;
6290 #endif /* CONFIG_IXGBE_DCB */
6291 #endif /* IXGBE_FCOE */
6292 
6293         /* initialize static ixgbe jump table entries */
6294         adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
6295                                           GFP_KERNEL);
6296         if (!adapter->jump_tables[0])
6297                 return -ENOMEM;
6298         adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
6299 
6300         for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
6301                 adapter->jump_tables[i] = NULL;
6302 
6303         adapter->mac_table = kcalloc(hw->mac.num_rar_entries,
6304                                      sizeof(struct ixgbe_mac_addr),
6305                                      GFP_KERNEL);
6306         if (!adapter->mac_table)
6307                 return -ENOMEM;
6308 
6309         if (ixgbe_init_rss_key(adapter))
6310                 return -ENOMEM;
6311 
6312         adapter->af_xdp_zc_qps = bitmap_zalloc(MAX_XDP_QUEUES, GFP_KERNEL);
6313         if (!adapter->af_xdp_zc_qps)
6314                 return -ENOMEM;
6315 
6316         /* Set MAC specific capability flags and exceptions */
6317         switch (hw->mac.type) {
6318         case ixgbe_mac_82598EB:
6319                 adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
6320 
6321                 if (hw->device_id == IXGBE_DEV_ID_82598AT)
6322                         adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
6323 
6324                 adapter->max_q_vectors = MAX_Q_VECTORS_82598;
6325                 adapter->ring_feature[RING_F_FDIR].limit = 0;
6326                 adapter->atr_sample_rate = 0;
6327                 adapter->fdir_pballoc = 0;
6328 #ifdef IXGBE_FCOE
6329                 adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6330                 adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6331 #ifdef CONFIG_IXGBE_DCB
6332                 adapter->fcoe.up = 0;
6333 #endif /* IXGBE_DCB */
6334 #endif /* IXGBE_FCOE */
6335                 break;
6336         case ixgbe_mac_82599EB:
6337                 if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
6338                         adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6339                 break;
6340         case ixgbe_mac_X540:
6341                 fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
6342                 if (fwsm & IXGBE_FWSM_TS_ENABLED)
6343                         adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6344                 break;
6345         case ixgbe_mac_x550em_a:
6346                 adapter->flags |= IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE;
6347                 switch (hw->device_id) {
6348                 case IXGBE_DEV_ID_X550EM_A_1G_T:
6349                 case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6350                         adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6351                         break;
6352                 default:
6353                         break;
6354                 }
6355         /* fall through */
6356         case ixgbe_mac_X550EM_x:
6357 #ifdef CONFIG_IXGBE_DCB
6358                 adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
6359 #endif
6360 #ifdef IXGBE_FCOE
6361                 adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6362 #ifdef CONFIG_IXGBE_DCB
6363                 adapter->fcoe.up = 0;
6364 #endif /* IXGBE_DCB */
6365 #endif /* IXGBE_FCOE */
6366         /* Fall Through */
6367         case ixgbe_mac_X550:
6368                 if (hw->mac.type == ixgbe_mac_X550)
6369                         adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6370 #ifdef CONFIG_IXGBE_DCA
6371                 adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
6372 #endif
6373                 adapter->flags |= IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE;
6374                 break;
6375         default:
6376                 break;
6377         }
6378 
6379 #ifdef IXGBE_FCOE
6380         /* FCoE support exists, always init the FCoE lock */
6381         spin_lock_init(&adapter->fcoe.lock);
6382 
6383 #endif
6384         /* n-tuple support exists, always init our spinlock */
6385         spin_lock_init(&adapter->fdir_perfect_lock);
6386 
6387 #ifdef CONFIG_IXGBE_DCB
6388         ixgbe_init_dcb(adapter);
6389 #endif
6390         ixgbe_init_ipsec_offload(adapter);
6391 
6392         /* default flow control settings */
6393         hw->fc.requested_mode = ixgbe_fc_full;
6394         hw->fc.current_mode = ixgbe_fc_full;    /* init for ethtool output */
6395         ixgbe_pbthresh_setup(adapter);
6396         hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
6397         hw->fc.send_xon = true;
6398         hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
6399 
6400 #ifdef CONFIG_PCI_IOV
6401         if (max_vfs > 0)
6402                 e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
6403 
6404         /* assign number of SR-IOV VFs */
6405         if (hw->mac.type != ixgbe_mac_82598EB) {
6406                 if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
6407                         max_vfs = 0;
6408                         e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
6409                 }
6410         }
6411 #endif /* CONFIG_PCI_IOV */
6412 
6413         /* enable itr by default in dynamic mode */
6414         adapter->rx_itr_setting = 1;
6415         adapter->tx_itr_setting = 1;
6416 
6417         /* set default ring sizes */
6418         adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
6419         adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
6420 
6421         /* set default work limits */
6422         adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
6423 
6424         /* initialize eeprom parameters */
6425         if (ixgbe_init_eeprom_params_generic(hw)) {
6426                 e_dev_err("EEPROM initialization failed\n");
6427                 return -EIO;
6428         }
6429 
6430         /* PF holds first pool slot */
6431         set_bit(0, adapter->fwd_bitmask);
6432         set_bit(__IXGBE_DOWN, &adapter->state);
6433 
6434         return 0;
6435 }
6436 
6437 /**
6438  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
6439  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
6440  *
6441  * Return 0 on success, negative on failure
6442  **/
6443 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
6444 {
6445         struct device *dev = tx_ring->dev;
6446         int orig_node = dev_to_node(dev);
6447         int ring_node = NUMA_NO_NODE;
6448         int size;
6449 
6450         size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
6451 
6452         if (tx_ring->q_vector)
6453                 ring_node = tx_ring->q_vector->numa_node;
6454 
6455         tx_ring->tx_buffer_info = vmalloc_node(size, ring_node);
6456         if (!tx_ring->tx_buffer_info)
6457                 tx_ring->tx_buffer_info = vmalloc(size);
6458         if (!tx_ring->tx_buffer_info)
6459                 goto err;
6460 
6461         /* round up to nearest 4K */
6462         tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
6463         tx_ring->size = ALIGN(tx_ring->size, 4096);
6464 
6465         set_dev_node(dev, ring_node);
6466         tx_ring->desc = dma_alloc_coherent(dev,
6467                                            tx_ring->size,
6468                                            &tx_ring->dma,
6469                                            GFP_KERNEL);
6470         set_dev_node(dev, orig_node);
6471         if (!tx_ring->desc)
6472                 tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
6473                                                    &tx_ring->dma, GFP_KERNEL);
6474         if (!tx_ring->desc)
6475                 goto err;
6476 
6477         tx_ring->next_to_use = 0;
6478         tx_ring->next_to_clean = 0;
6479         return 0;
6480 
6481 err:
6482         vfree(tx_ring->tx_buffer_info);
6483         tx_ring->tx_buffer_info = NULL;
6484         dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
6485         return -ENOMEM;
6486 }
6487 
6488 /**
6489  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
6490  * @adapter: board private structure
6491  *
6492  * If this function returns with an error, then it's possible one or
6493  * more of the rings is populated (while the rest are not).  It is the
6494  * callers duty to clean those orphaned rings.
6495  *
6496  * Return 0 on success, negative on failure
6497  **/
6498 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
6499 {
6500         int i, j = 0, err = 0;
6501 
6502         for (i = 0; i < adapter->num_tx_queues; i++) {
6503                 err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
6504                 if (!err)
6505                         continue;
6506 
6507                 e_err(probe, "Allocation for Tx Queue %u failed\n", i);
6508                 goto err_setup_tx;
6509         }
6510         for (j = 0; j < adapter->num_xdp_queues; j++) {
6511                 err = ixgbe_setup_tx_resources(adapter->xdp_ring[j]);
6512                 if (!err)
6513                         continue;
6514 
6515                 e_err(probe, "Allocation for Tx Queue %u failed\n", j);
6516                 goto err_setup_tx;
6517         }
6518 
6519         return 0;
6520 err_setup_tx:
6521         /* rewind the index freeing the rings as we go */
6522         while (j--)
6523                 ixgbe_free_tx_resources(adapter->xdp_ring[j]);
6524         while (i--)
6525                 ixgbe_free_tx_resources(adapter->tx_ring[i]);
6526         return err;
6527 }
6528 
6529 /**
6530  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
6531  * @adapter: pointer to ixgbe_adapter
6532  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
6533  *
6534  * Returns 0 on success, negative on failure
6535  **/
6536 int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
6537                              struct ixgbe_ring *rx_ring)
6538 {
6539         struct device *dev = rx_ring->dev;
6540         int orig_node = dev_to_node(dev);
6541         int ring_node = NUMA_NO_NODE;
6542         int size;
6543 
6544         size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
6545 
6546         if (rx_ring->q_vector)
6547                 ring_node = rx_ring->q_vector->numa_node;
6548 
6549         rx_ring->rx_buffer_info = vmalloc_node(size, ring_node);
6550         if (!rx_ring->rx_buffer_info)
6551                 rx_ring->rx_buffer_info = vmalloc(size);
6552         if (!rx_ring->rx_buffer_info)
6553                 goto err;
6554 
6555         /* Round up to nearest 4K */
6556         rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
6557         rx_ring->size = ALIGN(rx_ring->size, 4096);
6558 
6559         set_dev_node(dev, ring_node);
6560         rx_ring->desc = dma_alloc_coherent(dev,
6561                                            rx_ring->size,
6562                                            &rx_ring->dma,
6563                                            GFP_KERNEL);
6564         set_dev_node(dev, orig_node);
6565         if (!rx_ring->desc)
6566                 rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
6567                                                    &rx_ring->dma, GFP_KERNEL);
6568         if (!rx_ring->desc)
6569                 goto err;
6570 
6571         rx_ring->next_to_clean = 0;
6572         rx_ring->next_to_use = 0;
6573 
6574         /* XDP RX-queue info */
6575         if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, adapter->netdev,
6576                              rx_ring->queue_index) < 0)
6577                 goto err;
6578 
6579         rx_ring->xdp_prog = adapter->xdp_prog;
6580 
6581         return 0;
6582 err:
6583         vfree(rx_ring->rx_buffer_info);
6584         rx_ring->rx_buffer_info = NULL;
6585         dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
6586         return -ENOMEM;
6587 }
6588 
6589 /**
6590  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
6591  * @adapter: board private structure
6592  *
6593  * If this function returns with an error, then it's possible one or
6594  * more of the rings is populated (while the rest are not).  It is the
6595  * callers duty to clean those orphaned rings.
6596  *
6597  * Return 0 on success, negative on failure
6598  **/
6599 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
6600 {
6601         int i, err = 0;
6602 
6603         for (i = 0; i < adapter->num_rx_queues; i++) {
6604                 err = ixgbe_setup_rx_resources(adapter, adapter->rx_ring[i]);
6605                 if (!err)
6606                         continue;
6607 
6608                 e_err(probe, "Allocation for Rx Queue %u failed\n", i);
6609                 goto err_setup_rx;
6610         }
6611 
6612 #ifdef IXGBE_FCOE
6613         err = ixgbe_setup_fcoe_ddp_resources(adapter);
6614         if (!err)
6615 #endif
6616                 return 0;
6617 err_setup_rx:
6618         /* rewind the index freeing the rings as we go */
6619         while (i--)
6620                 ixgbe_free_rx_resources(adapter->rx_ring[i]);
6621         return err;
6622 }
6623 
6624 /**
6625  * ixgbe_free_tx_resources - Free Tx Resources per Queue
6626  * @tx_ring: Tx descriptor ring for a specific queue
6627  *
6628  * Free all transmit software resources
6629  **/
6630 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
6631 {
6632         ixgbe_clean_tx_ring(tx_ring);
6633 
6634         vfree(tx_ring->tx_buffer_info);
6635         tx_ring->tx_buffer_info = NULL;
6636 
6637         /* if not set, then don't free */
6638         if (!tx_ring->desc)
6639                 return;
6640 
6641         dma_free_coherent(tx_ring->dev, tx_ring->size,
6642                           tx_ring->desc, tx_ring->dma);
6643 
6644         tx_ring->desc = NULL;
6645 }
6646 
6647 /**
6648  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
6649  * @adapter: board private structure
6650  *
6651  * Free all transmit software resources
6652  **/
6653 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
6654 {
6655         int i;
6656 
6657         for (i = 0; i < adapter->num_tx_queues; i++)
6658                 if (adapter->tx_ring[i]->desc)
6659                         ixgbe_free_tx_resources(adapter->tx_ring[i]);
6660         for (i = 0; i < adapter->num_xdp_queues; i++)
6661                 if (adapter->xdp_ring[i]->desc)
6662                         ixgbe_free_tx_resources(adapter->xdp_ring[i]);
6663 }
6664 
6665 /**
6666  * ixgbe_free_rx_resources - Free Rx Resources
6667  * @rx_ring: ring to clean the resources from
6668  *
6669  * Free all receive software resources
6670  **/
6671 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6672 {
6673         ixgbe_clean_rx_ring(rx_ring);
6674 
6675         rx_ring->xdp_prog = NULL;
6676         xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
6677         vfree(rx_ring->rx_buffer_info);
6678         rx_ring->rx_buffer_info = NULL;
6679 
6680         /* if not set, then don't free */
6681         if (!rx_ring->desc)
6682                 return;
6683 
6684         dma_free_coherent(rx_ring->dev, rx_ring->size,
6685                           rx_ring->desc, rx_ring->dma);
6686 
6687         rx_ring->desc = NULL;
6688 }
6689 
6690 /**
6691  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6692  * @adapter: board private structure
6693  *
6694  * Free all receive software resources
6695  **/
6696 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6697 {
6698         int i;
6699 
6700 #ifdef IXGBE_FCOE
6701         ixgbe_free_fcoe_ddp_resources(adapter);
6702 
6703 #endif
6704         for (i = 0; i < adapter->num_rx_queues; i++)
6705                 if (adapter->rx_ring[i]->desc)
6706                         ixgbe_free_rx_resources(adapter->rx_ring[i]);
6707 }
6708 
6709 /**
6710  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6711  * @netdev: network interface device structure
6712  * @new_mtu: new value for maximum frame size
6713  *
6714  * Returns 0 on success, negative on failure
6715  **/
6716 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6717 {
6718         struct ixgbe_adapter *adapter = netdev_priv(netdev);
6719 
6720         if (adapter->xdp_prog) {
6721                 int new_frame_size = new_mtu + ETH_HLEN + ETH_FCS_LEN +
6722                                      VLAN_HLEN;
6723                 int i;
6724 
6725                 for (i = 0; i < adapter->num_rx_queues; i++) {
6726                         struct ixgbe_ring *ring = adapter->rx_ring[i];
6727 
6728                         if (new_frame_size > ixgbe_rx_bufsz(ring)) {
6729                                 e_warn(probe, "Requested MTU size is not supported with XDP\n");
6730                                 return -EINVAL;
6731                         }
6732                 }
6733         }
6734 
6735         /*
6736          * For 82599EB we cannot allow legacy VFs to enable their receive
6737          * paths when MTU greater than 1500 is configured.  So display a
6738          * warning that legacy VFs will be disabled.
6739          */
6740         if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6741             (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6742             (new_mtu > ETH_DATA_LEN))
6743                 e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6744 
6745         e_info(probe, "changing MTU from %d to %d\n", netdev->mtu, new_mtu);
6746 
6747         /* must set new MTU before calling down or up */
6748         netdev->mtu = new_mtu;
6749 
6750         if (netif_running(netdev))
6751                 ixgbe_reinit_locked(adapter);
6752 
6753         return 0;
6754 }
6755 
6756 /**
6757  * ixgbe_open - Called when a network interface is made active
6758  * @netdev: network interface device structure
6759  *
6760  * Returns 0 on success, negative value on failure
6761  *
6762  * The open entry point is called when a network interface is made
6763  * active by the system (IFF_UP).  At this point all resources needed
6764  * for transmit and receive operations are allocated, the interrupt
6765  * handler is registered with the OS, the watchdog timer is started,
6766  * and the stack is notified that the interface is ready.
6767  **/
6768 int ixgbe_open(struct net_device *netdev)
6769 {
6770         struct ixgbe_adapter *adapter = netdev_priv(netdev);
6771         struct ixgbe_hw *hw = &adapter->hw;
6772         int err, queues;
6773 
6774         /* disallow open during test */
6775         if (test_bit(__IXGBE_TESTING, &adapter->state))
6776                 return -EBUSY;
6777 
6778         netif_carrier_off(netdev);
6779 
6780         /* allocate transmit descriptors */
6781         err = ixgbe_setup_all_tx_resources(adapter);
6782         if (err)
6783                 goto err_setup_tx;
6784 
6785         /* allocate receive descriptors */
6786         err = ixgbe_setup_all_rx_resources(adapter);
6787         if (err)
6788                 goto err_setup_rx;
6789 
6790         ixgbe_configure(adapter);
6791 
6792         err = ixgbe_request_irq(adapter);
6793         if (err)
6794                 goto err_req_irq;
6795 
6796         /* Notify the stack of the actual queue counts. */
6797         queues = adapter->num_tx_queues;
6798         err = netif_set_real_num_tx_queues(netdev, queues);
6799         if (err)
6800                 goto err_set_queues;
6801 
6802         queues = adapter->num_rx_queues;
6803         err = netif_set_real_num_rx_queues(netdev, queues);
6804         if (err)
6805                 goto err_set_queues;
6806 
6807         ixgbe_ptp_init(adapter);
6808 
6809         ixgbe_up_complete(adapter);
6810 
6811         ixgbe_clear_udp_tunnel_port(adapter, IXGBE_VXLANCTRL_ALL_UDPPORT_MASK);
6812         udp_tunnel_get_rx_info(netdev);
6813 
6814         return 0;
6815 
6816 err_set_queues:
6817         ixgbe_free_irq(adapter);
6818 err_req_irq:
6819         ixgbe_free_all_rx_resources(adapter);
6820         if (hw->phy.ops.set_phy_power && !adapter->wol)
6821                 hw->phy.ops.set_phy_power(&adapter->hw, false);
6822 err_setup_rx:
6823         ixgbe_free_all_tx_resources(adapter);
6824 err_setup_tx:
6825         ixgbe_reset(adapter);
6826 
6827         return err;
6828 }
6829 
6830 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6831 {
6832         ixgbe_ptp_suspend(adapter);
6833 
6834         if (adapter->hw.phy.ops.enter_lplu) {
6835                 adapter->hw.phy.reset_disable = true;
6836                 ixgbe_down(adapter);
6837                 adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6838                 adapter->hw.phy.reset_disable = false;
6839         } else {
6840                 ixgbe_down(adapter);
6841         }
6842 
6843         ixgbe_free_irq(adapter);
6844 
6845         ixgbe_free_all_tx_resources(adapter);
6846         ixgbe_free_all_rx_resources(adapter);
6847 }
6848 
6849 /**
6850  * ixgbe_close - Disables a network interface
6851  * @netdev: network interface device structure
6852  *
6853  * Returns 0, this is not allowed to fail
6854  *
6855  * The close entry point is called when an interface is de-activated
6856  * by the OS.  The hardware is still under the drivers control, but
6857  * needs to be disabled.  A global MAC reset is issued to stop the
6858  * hardware, and all transmit and receive resources are freed.
6859  **/
6860 int ixgbe_close(struct net_device *netdev)
6861 {
6862         struct ixgbe_adapter *adapter = netdev_priv(netdev);
6863 
6864         ixgbe_ptp_stop(adapter);
6865 
6866         if (netif_device_present(netdev))
6867                 ixgbe_close_suspend(adapter);
6868 
6869         ixgbe_fdir_filter_exit(adapter);
6870 
6871         ixgbe_release_hw_control(adapter);
6872 
6873         return 0;
6874 }
6875 
6876 #ifdef CONFIG_PM
6877 static int ixgbe_resume(struct pci_dev *pdev)
6878 {
6879         struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6880         struct net_device *netdev = adapter->netdev;
6881         u32 err;
6882 
6883         adapter->hw.hw_addr = adapter->io_addr;
6884         pci_set_power_state(pdev, PCI_D0);
6885         pci_restore_state(pdev);
6886         /*
6887          * pci_restore_state clears dev->state_saved so call
6888          * pci_save_state to restore it.
6889          */
6890         pci_save_state(pdev);
6891 
6892         err = pci_enable_device_mem(pdev);
6893         if (err) {
6894                 e_dev_err("Cannot enable PCI device from suspend\n");
6895                 return err;
6896         }
6897         smp_mb__before_atomic();
6898         clear_bit(__IXGBE_DISABLED, &adapter->state);
6899         pci_set_master(pdev);
6900 
6901         pci_wake_from_d3(pdev, false);
6902 
6903         ixgbe_reset(adapter);
6904 
6905         IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6906 
6907         rtnl_lock();
6908         err = ixgbe_init_interrupt_scheme(adapter);
6909         if (!err && netif_running(netdev))
6910                 err = ixgbe_open(netdev);
6911 
6912 
6913         if (!err)
6914                 netif_device_attach(netdev);
6915         rtnl_unlock();
6916 
6917         return err;
6918 }
6919 #endif /* CONFIG_PM */
6920 
6921 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6922 {
6923         struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6924         struct net_device *netdev = adapter->netdev;
6925         struct ixgbe_hw *hw = &adapter->hw;
6926         u32 ctrl;
6927         u32 wufc = adapter->wol;
6928 #ifdef CONFIG_PM
6929         int retval = 0;
6930 #endif
6931 
6932         rtnl_lock();
6933         netif_device_detach(netdev);
6934 
6935         if (netif_running(netdev))
6936                 ixgbe_close_suspend(adapter);
6937 
6938         ixgbe_clear_interrupt_scheme(adapter);
6939         rtnl_unlock();
6940 
6941 #ifdef CONFIG_PM
6942         retval = pci_save_state(pdev);
6943         if (retval)
6944                 return retval;
6945 
6946 #endif
6947         if (hw->mac.ops.stop_link_on_d3)
6948                 hw->mac.ops.stop_link_on_d3(hw);
6949 
6950         if (wufc) {
6951                 u32 fctrl;
6952 
6953                 ixgbe_set_rx_mode(netdev);
6954 
6955                 /* enable the optics for 82599 SFP+ fiber as we can WoL */
6956                 if (hw->mac.ops.enable_tx_laser)
6957                         hw->mac.ops.enable_tx_laser(hw);
6958 
6959                 /* enable the reception of multicast packets */
6960                 fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
6961                 fctrl |= IXGBE_FCTRL_MPE;
6962                 IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
6963 
6964                 ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
6965                 ctrl |= IXGBE_CTRL_GIO_DIS;
6966                 IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
6967 
6968                 IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
6969         } else {
6970                 IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
6971                 IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
6972         }
6973 
6974         switch (hw->mac.type) {
6975         case ixgbe_mac_82598EB:
6976                 pci_wake_from_d3(pdev, false);
6977                 break;
6978         case ixgbe_mac_82599EB:
6979         case ixgbe_mac_X540:
6980         case ixgbe_mac_X550:
6981         case ixgbe_mac_X550EM_x:
6982         case ixgbe_mac_x550em_a:
6983                 pci_wake_from_d3(pdev, !!wufc);
6984                 break;
6985         default:
6986                 break;
6987         }
6988 
6989         *enable_wake = !!wufc;
6990         if (hw->phy.ops.set_phy_power && !*enable_wake)
6991                 hw->phy.ops.set_phy_power(hw, false);
6992 
6993         ixgbe_release_hw_control(adapter);
6994 
6995         if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
6996                 pci_disable_device(pdev);
6997 
6998         return 0;
6999 }
7000 
7001 #ifdef CONFIG_PM
7002 static int ixgbe_suspend(struct pci_dev *pdev, pm_message_t state)
7003 {
7004         int retval;
7005         bool wake;
7006 
7007         retval = __ixgbe_shutdown(pdev, &wake);
7008         if (retval)
7009                 return retval;
7010 
7011         if (wake) {
7012                 pci_prepare_to_sleep(pdev);
7013         } else {
7014                 pci_wake_from_d3(pdev, false);
7015                 pci_set_power_state(pdev, PCI_D3hot);
7016         }
7017 
7018         return 0;
7019 }
7020 #endif /* CONFIG_PM */
7021 
7022 static void ixgbe_shutdown(struct pci_dev *pdev)
7023 {
7024         bool wake;
7025 
7026         __ixgbe_shutdown(pdev, &wake);
7027 
7028         if (system_state == SYSTEM_POWER_OFF) {
7029                 pci_wake_from_d3(pdev, wake);
7030                 pci_set_power_state(pdev, PCI_D3hot);
7031         }
7032 }
7033 
7034 /**
7035  * ixgbe_update_stats - Update the board statistics counters.
7036  * @adapter: board private structure
7037  **/
7038 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
7039 {
7040         struct net_device *netdev = adapter->netdev;
7041         struct ixgbe_hw *hw = &adapter->hw;
7042         struct ixgbe_hw_stats *hwstats = &adapter->stats;
7043         u64 total_mpc = 0;
7044         u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
7045         u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
7046         u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
7047         u64 alloc_rx_page = 0;
7048         u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
7049 
7050         if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7051             test_bit(__IXGBE_RESETTING, &adapter->state))
7052                 return;
7053 
7054         if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
7055                 u64 rsc_count = 0;
7056                 u64 rsc_flush = 0;
7057                 for (i = 0; i < adapter->num_rx_queues; i++) {
7058                         rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
7059                         rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
7060                 }
7061                 adapter->rsc_total_count = rsc_count;
7062                 adapter->rsc_total_flush = rsc_flush;
7063         }
7064 
7065         for (i = 0; i < adapter->num_rx_queues; i++) {
7066                 struct ixgbe_ring *rx_ring = adapter->rx_ring[i];
7067                 non_eop_descs += rx_ring->rx_stats.non_eop_descs;
7068                 alloc_rx_page += rx_ring->rx_stats.alloc_rx_page;
7069                 alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
7070                 alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
7071                 hw_csum_rx_error += rx_ring->rx_stats.csum_err;
7072                 bytes += rx_ring->stats.bytes;
7073                 packets += rx_ring->stats.packets;
7074         }
7075         adapter->non_eop_descs = non_eop_descs;
7076         adapter->alloc_rx_page = alloc_rx_page;
7077         adapter->alloc_rx_page_failed = alloc_rx_page_failed;
7078         adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
7079         adapter->hw_csum_rx_error = hw_csum_rx_error;
7080         netdev->stats.rx_bytes = bytes;
7081         netdev->stats.rx_packets = packets;
7082 
7083         bytes = 0;
7084         packets = 0;
7085         /* gather some stats to the adapter struct that are per queue */
7086         for (i = 0; i < adapter->num_tx_queues; i++) {
7087                 struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
7088                 restart_queue += tx_ring->tx_stats.restart_queue;
7089                 tx_busy += tx_ring->tx_stats.tx_busy;
7090                 bytes += tx_ring->stats.bytes;
7091                 packets += tx_ring->stats.packets;
7092         }
7093         for (i = 0; i < adapter->num_xdp_queues; i++) {
7094                 struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
7095 
7096                 restart_queue += xdp_ring->tx_stats.restart_queue;
7097                 tx_busy += xdp_ring->tx_stats.tx_busy;
7098                 bytes += xdp_ring->stats.bytes;
7099                 packets += xdp_ring->stats.packets;
7100         }
7101         adapter->restart_queue = restart_queue;
7102         adapter->tx_busy = tx_busy;
7103         netdev->stats.tx_bytes = bytes;
7104         netdev->stats.tx_packets = packets;
7105 
7106         hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
7107 
7108         /* 8 register reads */
7109         for (i = 0; i < 8; i++) {
7110                 /* for packet buffers not used, the register should read 0 */
7111                 mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
7112                 missed_rx += mpc;
7113                 hwstats->mpc[i] += mpc;
7114                 total_mpc += hwstats->mpc[i];
7115                 hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
7116                 hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
7117                 switch (hw->mac.type) {
7118                 case ixgbe_mac_82598EB:
7119                         hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
7120                         hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
7121                         hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
7122                         hwstats->pxonrxc[i] +=
7123                                 IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
7124                         break;
7125                 case ixgbe_mac_82599EB:
7126                 case ixgbe_mac_X540:
7127                 case ixgbe_mac_X550:
7128                 case ixgbe_mac_X550EM_x:
7129                 case ixgbe_mac_x550em_a:
7130                         hwstats->pxonrxc[i] +=
7131                                 IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
7132                         break;
7133                 default:
7134                         break;
7135                 }
7136         }
7137 
7138         /*16 register reads */
7139         for (i = 0; i < 16; i++) {
7140                 hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
7141                 hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
7142                 if ((hw->mac.type == ixgbe_mac_82599EB) ||
7143                     (hw->mac.type == ixgbe_mac_X540) ||
7144                     (hw->mac.type == ixgbe_mac_X550) ||
7145                     (hw->mac.type == ixgbe_mac_X550EM_x) ||
7146                     (hw->mac.type == ixgbe_mac_x550em_a)) {
7147                         hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
7148                         IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
7149                         hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
7150                         IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
7151                 }
7152         }
7153 
7154         hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
7155         /* work around hardware counting issue */
7156         hwstats->gprc -= missed_rx;
7157 
7158         ixgbe_update_xoff_received(adapter);
7159 
7160         /* 82598 hardware only has a 32 bit counter in the high register */
7161         switch (hw->mac.type) {
7162         case ixgbe_mac_82598EB:
7163                 hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
7164                 hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
7165                 hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
7166                 hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
7167                 break;
7168         case ixgbe_mac_X540:
7169         case ixgbe_mac_X550:
7170         case ixgbe_mac_X550EM_x:
7171         case ixgbe_mac_x550em_a:
7172                 /* OS2BMC stats are X540 and later */
7173                 hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
7174                 hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
7175                 hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
7176                 hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
7177                 /* fall through */
7178         case ixgbe_mac_82599EB:
7179                 for (i = 0; i < 16; i++)
7180                         adapter->hw_rx_no_dma_resources +=
7181                                              IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
7182                 hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
7183                 IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
7184                 hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
7185                 IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
7186                 hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
7187                 IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
7188                 hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
7189                 hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
7190                 hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
7191 #ifdef IXGBE_FCOE
7192                 hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
7193                 hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
7194                 hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
7195                 hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
7196                 hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
7197                 hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
7198                 /* Add up per cpu counters for total ddp aloc fail */
7199                 if (adapter->fcoe.ddp_pool) {
7200                         struct ixgbe_fcoe *fcoe = &adapter->fcoe;
7201                         struct ixgbe_fcoe_ddp_pool *ddp_pool;
7202                         unsigned int cpu;
7203                         u64 noddp = 0, noddp_ext_buff = 0;
7204                         for_each_possible_cpu(cpu) {
7205                                 ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
7206                                 noddp += ddp_pool->noddp;
7207                                 noddp_ext_buff += ddp_pool->noddp_ext_buff;
7208                         }
7209                         hwstats->fcoe_noddp = noddp;
7210                         hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
7211                 }
7212 #endif /* IXGBE_FCOE */
7213                 break;
7214         default:
7215                 break;
7216         }
7217         bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
7218         hwstats->bprc += bprc;
7219         hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
7220         if (hw->mac.type == ixgbe_mac_82598EB)
7221                 hwstats->mprc -= bprc;
7222         hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
7223         hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
7224         hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
7225         hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
7226         hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
7227         hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
7228         hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
7229         hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
7230         lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
7231         hwstats->lxontxc += lxon;
7232         lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
7233         hwstats->lxofftxc += lxoff;
7234         hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
7235         hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
7236         /*
7237          * 82598 errata - tx of flow control packets is included in tx counters
7238          */
7239         xon_off_tot = lxon + lxoff;
7240         hwstats->gptc -= xon_off_tot;
7241         hwstats->mptc -= xon_off_tot;
7242         hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
7243         hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
7244         hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
7245         hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
7246         hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
7247         hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
7248         hwstats->ptc64 -= xon_off_tot;
7249         hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
7250         hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
7251         hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
7252         hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
7253         hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
7254         hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
7255 
7256         /* Fill out the OS statistics structure */
7257         netdev->stats.multicast = hwstats->mprc;
7258 
7259         /* Rx Errors */
7260         netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
7261         netdev->stats.rx_dropped = 0;
7262         netdev->stats.rx_length_errors = hwstats->rlec;
7263         netdev->stats.rx_crc_errors = hwstats->crcerrs;
7264         netdev->stats.rx_missed_errors = total_mpc;
7265 }
7266 
7267 /**
7268  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
7269  * @adapter: pointer to the device adapter structure
7270  **/
7271 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
7272 {
7273         struct ixgbe_hw *hw = &adapter->hw;
7274         int i;
7275 
7276         if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
7277                 return;
7278 
7279         adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
7280 
7281         /* if interface is down do nothing */
7282         if (test_bit(__IXGBE_DOWN, &adapter->state))
7283                 return;
7284 
7285         /* do nothing if we are not using signature filters */
7286         if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
7287                 return;
7288 
7289         adapter->fdir_overflow++;
7290 
7291         if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
7292                 for (i = 0; i < adapter->num_tx_queues; i++)
7293                         set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7294                                 &(adapter->tx_ring[i]->state));
7295                 for (i = 0; i < adapter->num_xdp_queues; i++)
7296                         set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7297                                 &adapter->xdp_ring[i]->state);
7298                 /* re-enable flow director interrupts */
7299                 IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
7300         } else {
7301                 e_err(probe, "failed to finish FDIR re-initialization, "
7302                       "ignored adding FDIR ATR filters\n");
7303         }
7304 }
7305 
7306 /**
7307  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
7308  * @adapter: pointer to the device adapter structure
7309  *
7310  * This function serves two purposes.  First it strobes the interrupt lines
7311  * in order to make certain interrupts are occurring.  Secondly it sets the
7312  * bits needed to check for TX hangs.  As a result we should immediately
7313  * determine if a hang has occurred.
7314  */
7315 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
7316 {
7317         struct ixgbe_hw *hw = &adapter->hw;
7318         u64 eics = 0;
7319         int i;
7320 
7321         /* If we're down, removing or resetting, just bail */
7322         if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7323             test_bit(__IXGBE_REMOVING, &adapter->state) ||
7324             test_bit(__IXGBE_RESETTING, &adapter->state))
7325                 return;
7326 
7327         /* Force detection of hung controller */
7328         if (netif_carrier_ok(adapter->netdev)) {
7329                 for (i = 0; i < adapter->num_tx_queues; i++)
7330                         set_check_for_tx_hang(adapter->tx_ring[i]);
7331                 for (i = 0; i < adapter->num_xdp_queues; i++)
7332                         set_check_for_tx_hang(adapter->xdp_ring[i]);
7333         }
7334 
7335         if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
7336                 /*
7337                  * for legacy and MSI interrupts don't set any bits
7338                  * that are enabled for EIAM, because this operation
7339                  * would set *both* EIMS and EICS for any bit in EIAM
7340                  */
7341                 IXGBE_WRITE_REG(hw, IXGBE_EICS,
7342                         (IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
7343         } else {
7344                 /* get one bit for every active tx/rx interrupt vector */
7345                 for (i = 0; i < adapter->num_q_vectors; i++) {
7346                         struct ixgbe_q_vector *qv = adapter->q_vector[i];
7347                         if (qv->rx.ring || qv->tx.ring)
7348                                 eics |= BIT_ULL(i);
7349                 }
7350         }
7351 
7352         /* Cause software interrupt to ensure rings are cleaned */
7353         ixgbe_irq_rearm_queues(adapter, eics);
7354 }
7355 
7356 /**
7357  * ixgbe_watchdog_update_link - update the link status
7358  * @adapter: pointer to the device adapter structure
7359  **/
7360 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
7361 {
7362         struct ixgbe_hw *hw = &adapter->hw;
7363         u32 link_speed = adapter->link_speed;
7364         bool link_up = adapter->link_up;
7365         bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
7366 
7367         if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
7368                 return;
7369 
7370         if (hw->mac.ops.check_link) {
7371                 hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
7372         } else {
7373                 /* always assume link is up, if no check link function */
7374                 link_speed = IXGBE_LINK_SPEED_10GB_FULL;
7375                 link_up = true;
7376         }
7377 
7378         if (adapter->ixgbe_ieee_pfc)
7379                 pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
7380 
7381         if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
7382                 hw->mac.ops.fc_enable(hw);
7383                 ixgbe_set_rx_drop_en(adapter);
7384         }
7385 
7386         if (link_up ||
7387             time_after(jiffies, (adapter->link_check_timeout +
7388                                  IXGBE_TRY_LINK_TIMEOUT))) {
7389                 adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
7390                 IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
7391                 IXGBE_WRITE_FLUSH(hw);
7392         }
7393 
7394         adapter->link_up = link_up;
7395         adapter->link_speed = link_speed;
7396 }
7397 
7398 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
7399 {
7400 #ifdef CONFIG_IXGBE_DCB
7401         struct net_device *netdev = adapter->netdev;
7402         struct dcb_app app = {
7403                               .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
7404                               .protocol = 0,
7405                              };
7406         u8 up = 0;
7407 
7408         if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
7409                 up = dcb_ieee_getapp_mask(netdev, &app);
7410 
7411         adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
7412 #endif
7413 }
7414 
7415 /**
7416  * ixgbe_watchdog_link_is_up - update netif_carrier status and
7417  *                             print link up message
7418  * @adapter: pointer to the device adapter structure
7419  **/
7420 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
7421 {
7422         struct net_device *netdev = adapter->netdev;
7423         struct ixgbe_hw *hw = &adapter->hw;
7424         u32 link_speed = adapter->link_speed;
7425         const char *speed_str;
7426         bool flow_rx, flow_tx;
7427 
7428         /* only continue if link was previously down */
7429         if (netif_carrier_ok(netdev))
7430                 return;
7431 
7432         adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7433 
7434         switch (hw->mac.type) {
7435         case ixgbe_mac_82598EB: {
7436                 u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7437                 u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
7438                 flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
7439                 flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
7440         }
7441                 break;
7442         case ixgbe_mac_X540:
7443         case ixgbe_mac_X550:
7444         case ixgbe_mac_X550EM_x:
7445         case ixgbe_mac_x550em_a:
7446         case ixgbe_mac_82599EB: {
7447                 u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
7448                 u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
7449                 flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
7450                 flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
7451         }
7452                 break;
7453         default:
7454                 flow_tx = false;
7455                 flow_rx = false;
7456                 break;
7457         }
7458 
7459         adapter->last_rx_ptp_check = jiffies;
7460 
7461         if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7462                 ixgbe_ptp_start_cyclecounter(adapter);
7463 
7464         switch (link_speed) {
7465         case IXGBE_LINK_SPEED_10GB_FULL:
7466                 speed_str = "10 Gbps";
7467                 break;
7468         case IXGBE_LINK_SPEED_5GB_FULL:
7469                 speed_str = "5 Gbps";
7470                 break;
7471         case IXGBE_LINK_SPEED_2_5GB_FULL:
7472                 speed_str = "2.5 Gbps";
7473                 break;
7474         case IXGBE_LINK_SPEED_1GB_FULL:
7475                 speed_str = "1 Gbps";
7476                 break;
7477         case IXGBE_LINK_SPEED_100_FULL:
7478                 speed_str = "100 Mbps";
7479                 break;
7480         case IXGBE_LINK_SPEED_10_FULL:
7481                 speed_str = "10 Mbps";
7482                 break;
7483         default:
7484                 speed_str = "unknown speed";
7485                 break;
7486         }
7487         e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
7488                ((flow_rx && flow_tx) ? "RX/TX" :
7489                (flow_rx ? "RX" :
7490                (flow_tx ? "TX" : "None"))));
7491 
7492         netif_carrier_on(netdev);
7493         ixgbe_check_vf_rate_limit(adapter);
7494 
7495         /* enable transmits */
7496         netif_tx_wake_all_queues(adapter->netdev);
7497 
7498         /* update the default user priority for VFs */
7499         ixgbe_update_default_up(adapter);
7500 
7501         /* ping all the active vfs to let them know link has changed */
7502         ixgbe_ping_all_vfs(adapter);
7503 }
7504 
7505 /**
7506  * ixgbe_watchdog_link_is_down - update netif_carrier status and
7507  *                               print link down message
7508  * @adapter: pointer to the adapter structure
7509  **/
7510 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
7511 {
7512         struct net_device *netdev = adapter->netdev;
7513         struct ixgbe_hw *hw = &adapter->hw;
7514 
7515         adapter->link_up = false;
7516         adapter->link_speed = 0;
7517 
7518         /* only continue if link was up previously */
7519         if (!netif_carrier_ok(netdev))
7520                 return;
7521 
7522         /* poll for SFP+ cable when link is down */
7523         if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
7524                 adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
7525 
7526         if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7527                 ixgbe_ptp_start_cyclecounter(adapter);
7528 
7529         e_info(drv, "NIC Link is Down\n");
7530         netif_carrier_off(netdev);
7531 
7532         /* ping all the active vfs to let them know link has changed */
7533         ixgbe_ping_all_vfs(adapter);
7534 }
7535 
7536 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
7537 {
7538         int i;
7539 
7540         for (i = 0; i < adapter->num_tx_queues; i++) {
7541                 struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
7542 
7543                 if (tx_ring->next_to_use != tx_ring->next_to_clean)
7544                         return true;
7545         }
7546 
7547         for (i = 0; i < adapter->num_xdp_queues; i++) {
7548                 struct ixgbe_ring *ring = adapter->xdp_ring[i];
7549 
7550                 if (ring->next_to_use != ring->next_to_clean)
7551                         return true;
7552         }
7553 
7554         return false;
7555 }
7556 
7557 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
7558 {
7559         struct ixgbe_hw *hw = &adapter->hw;
7560         struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
7561         u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
7562 
7563         int i, j;
7564 
7565         if (!adapter->num_vfs)
7566                 return false;
7567 
7568         /* resetting the PF is only needed for MAC before X550 */
7569         if (hw->mac.type >= ixgbe_mac_X550)
7570                 return false;
7571 
7572         for (i = 0; i < adapter->num_vfs; i++) {
7573                 for (j = 0; j < q_per_pool; j++) {
7574                         u32 h, t;
7575 
7576                         h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
7577                         t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
7578 
7579                         if (h != t)
7580                                 return true;
7581                 }
7582         }
7583 
7584         return false;
7585 }
7586 
7587 /**
7588  * ixgbe_watchdog_flush_tx - flush queues on link down
7589  * @adapter: pointer to the device adapter structure
7590  **/
7591 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
7592 {
7593         if (!netif_carrier_ok(adapter->netdev)) {
7594                 if (ixgbe_ring_tx_pending(adapter) ||
7595                     ixgbe_vf_tx_pending(adapter)) {
7596                         /* We've lost link, so the controller stops DMA,
7597                          * but we've got queued Tx work that's never going
7598                          * to get done, so reset controller to flush Tx.
7599                          * (Do the reset outside of interrupt context).
7600                          */
7601                         e_warn(drv, "initiating reset to clear Tx work after link loss\n");
7602                         set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
7603                 }
7604         }
7605 }
7606 
7607 #ifdef CONFIG_PCI_IOV
7608 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
7609 {
7610         struct ixgbe_hw *hw = &adapter->hw;
7611         struct pci_dev *pdev = adapter->pdev;
7612         unsigned int vf;
7613         u32 gpc;
7614 
7615         if (!(netif_carrier_ok(adapter->netdev)))
7616                 return;
7617 
7618         gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
7619         if (gpc) /* If incrementing then no need for the check below */
7620                 return;
7621         /* Check to see if a bad DMA write target from an errant or
7622          * malicious VF has caused a PCIe error.  If so then we can
7623          * issue a VFLR to the offending VF(s) and then resume without
7624          * requesting a full slot reset.
7625          */
7626 
7627         if (!pdev)
7628                 return;
7629 
7630         /* check status reg for all VFs owned by this PF */
7631         for (vf = 0; vf < adapter->num_vfs; ++vf) {
7632                 struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
7633                 u16 status_reg;
7634 
7635                 if (!vfdev)
7636                         continue;
7637                 pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
7638                 if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
7639                     status_reg & PCI_STATUS_REC_MASTER_ABORT)
7640                         pcie_flr(vfdev);
7641         }
7642 }
7643 
7644 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
7645 {
7646         u32 ssvpc;
7647 
7648         /* Do not perform spoof check for 82598 or if not in IOV mode */
7649         if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7650             adapter->num_vfs == 0)
7651                 return;
7652 
7653         ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
7654 
7655         /*
7656          * ssvpc register is cleared on read, if zero then no
7657          * spoofed packets in the last interval.
7658          */
7659         if (!ssvpc)
7660                 return;
7661 
7662         e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
7663 }
7664 #else
7665 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7666 {
7667 }
7668 
7669 static void
7670 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7671 {
7672 }
7673 #endif /* CONFIG_PCI_IOV */
7674 
7675 
7676 /**
7677  * ixgbe_watchdog_subtask - check and bring link up
7678  * @adapter: pointer to the device adapter structure
7679  **/
7680 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7681 {
7682         /* if interface is down, removing or resetting, do nothing */
7683         if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7684             test_bit(__IXGBE_REMOVING, &adapter->state) ||
7685             test_bit(__IXGBE_RESETTING, &adapter->state))
7686                 return;
7687 
7688         ixgbe_watchdog_update_link(adapter);
7689 
7690         if (adapter->link_up)
7691                 ixgbe_watchdog_link_is_up(adapter);
7692         else
7693                 ixgbe_watchdog_link_is_down(adapter);
7694 
7695         ixgbe_check_for_bad_vf(adapter);
7696         ixgbe_spoof_check(adapter);
7697         ixgbe_update_stats(adapter);
7698 
7699         ixgbe_watchdog_flush_tx(adapter);
7700 }
7701 
7702 /**
7703  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7704  * @adapter: the ixgbe adapter structure
7705  **/
7706 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7707 {
7708         struct ixgbe_hw *hw = &adapter->hw;
7709         s32 err;
7710 
7711         /* not searching for SFP so there is nothing to do here */
7712         if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7713             !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7714                 return;
7715 
7716         if (adapter->sfp_poll_time &&
7717             time_after(adapter->sfp_poll_time, jiffies))
7718                 return; /* If not yet time to poll for SFP */
7719 
7720         /* someone else is in init, wait until next service event */
7721         if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7722                 return;
7723 
7724         adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7725 
7726         err = hw->phy.ops.identify_sfp(hw);
7727         if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7728                 goto sfp_out;
7729 
7730         if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
7731                 /* If no cable is present, then we need to reset
7732                  * the next time we find a good cable. */
7733                 adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7734         }
7735 
7736         /* exit on error */
7737         if (err)
7738                 goto sfp_out;
7739 
7740         /* exit if reset not needed */
7741         if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7742                 goto sfp_out;
7743 
7744         adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7745 
7746         /*
7747          * A module may be identified correctly, but the EEPROM may not have
7748          * support for that module.  setup_sfp() will fail in that case, so
7749          * we should not allow that module to load.
7750          */
7751         if (hw->mac.type == ixgbe_mac_82598EB)
7752                 err = hw->phy.ops.reset(hw);
7753         else
7754                 err = hw->mac.ops.setup_sfp(hw);
7755 
7756         if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7757                 goto sfp_out;
7758 
7759         adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7760         e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7761 
7762 sfp_out:
7763         clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7764 
7765         if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
7766             (adapter->netdev->reg_state == NETREG_REGISTERED)) {
7767                 e_dev_err("failed to initialize because an unsupported "
7768                           "SFP+ module type was detected.\n");
7769                 e_dev_err("Reload the driver after installing a "
7770                           "supported module.\n");
7771                 unregister_netdev(adapter->netdev);
7772         }
7773 }
7774 
7775 /**
7776  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7777  * @adapter: the ixgbe adapter structure
7778  **/
7779 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7780 {
7781         struct ixgbe_hw *hw = &adapter->hw;
7782         u32 cap_speed;
7783         u32 speed;
7784         bool autoneg = false;
7785 
7786         if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7787                 return;
7788 
7789         /* someone else is in init, wait until next service event */
7790         if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7791                 return;
7792 
7793         adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7794 
7795         hw->mac.ops.get_link_capabilities(hw, &cap_speed, &autoneg);
7796 
7797         /* advertise highest capable link speed */
7798         if (!autoneg && (cap_speed & IXGBE_LINK_SPEED_10GB_FULL))
7799                 speed = IXGBE_LINK_SPEED_10GB_FULL;
7800         else
7801                 speed = cap_speed & (IXGBE_LINK_SPEED_10GB_FULL |
7802                                      IXGBE_LINK_SPEED_1GB_FULL);
7803 
7804         if (hw->mac.ops.setup_link)
7805                 hw->mac.ops.setup_link(hw, speed, true);
7806 
7807         adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7808         adapter->link_check_timeout = jiffies;
7809         clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7810 }
7811 
7812 /**
7813  * ixgbe_service_timer - Timer Call-back
7814  * @t: pointer to timer_list structure
7815  **/
7816 static void ixgbe_service_timer(struct timer_list *t)
7817 {
7818         struct ixgbe_adapter *adapter = from_timer(adapter, t, service_timer);
7819         unsigned long next_event_offset;
7820 
7821         /* poll faster when waiting for link */
7822         if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7823                 next_event_offset = HZ / 10;
7824         else
7825                 next_event_offset = HZ * 2;
7826 
7827         /* Reset the timer */
7828         mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7829 
7830         ixgbe_service_event_schedule(adapter);
7831 }
7832 
7833 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7834 {
7835         struct ixgbe_hw *hw = &adapter->hw;
7836         u32 status;
7837 
7838         if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7839                 return;
7840 
7841         adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7842 
7843         if (!hw->phy.ops.handle_lasi)
7844                 return;
7845 
7846         status = hw->phy.ops.handle_lasi(&adapter->hw);
7847         if (status != IXGBE_ERR_OVERTEMP)
7848                 return;
7849 
7850         e_crit(drv, "%s\n", ixgbe_overheat_msg);
7851 }
7852 
7853 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7854 {
7855         if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7856                 return;
7857 
7858         rtnl_lock();
7859         /* If we're already down, removing or resetting, just bail */
7860         if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7861             test_bit(__IXGBE_REMOVING, &adapter->state) ||
7862             test_bit(__IXGBE_RESETTING, &adapter->state)) {
7863                 rtnl_unlock();
7864                 return;
7865         }
7866 
7867         ixgbe_dump(adapter);
7868         netdev_err(adapter->netdev, "Reset adapter\n");
7869         adapter->tx_timeout_count++;
7870 
7871         ixgbe_reinit_locked(adapter);
7872         rtnl_unlock();
7873 }
7874 
7875 /**
7876  * ixgbe_check_fw_error - Check firmware for errors
7877  * @adapter: the adapter private structure
7878  *
7879  * Check firmware errors in register FWSM
7880  */
7881 static bool ixgbe_check_fw_error(struct ixgbe_adapter *adapter)
7882 {
7883         struct ixgbe_hw *hw = &adapter->hw;
7884         u32 fwsm;
7885 
7886         /* read fwsm.ext_err_ind register and log errors */
7887         fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
7888 
7889         if (fwsm & IXGBE_FWSM_EXT_ERR_IND_MASK ||
7890             !(fwsm & IXGBE_FWSM_FW_VAL_BIT))
7891                 e_dev_warn("Warning firmware error detected FWSM: 0x%08X\n",
7892                            fwsm);
7893 
7894         if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
7895                 e_dev_err("Firmware recovery mode detected. Limiting functionality. Refer to the Intel(R) Ethernet Adapters and Devices User Guide for details on firmware recovery mode.\n");
7896                 return true;
7897         }
7898 
7899         return false;
7900 }
7901 
7902 /**
7903  * ixgbe_service_task - manages and runs subtasks
7904  * @work: pointer to work_struct containing our data
7905  **/
7906 static void ixgbe_service_task(struct work_struct *work)
7907 {
7908         struct ixgbe_adapter *adapter = container_of(work,
7909                                                      struct ixgbe_adapter,
7910                                                      service_task);
7911         if (ixgbe_removed(adapter->hw.hw_addr)) {
7912                 if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
7913                         rtnl_lock();
7914                         ixgbe_down(adapter);
7915                         rtnl_unlock();
7916                 }
7917                 ixgbe_service_event_complete(adapter);
7918                 return;
7919         }
7920         if (ixgbe_check_fw_error(adapter)) {
7921                 if (!test_bit(__IXGBE_DOWN, &adapter->state))
7922                         unregister_netdev(adapter->netdev);
7923                 ixgbe_service_event_complete(adapter);
7924                 return;
7925         }
7926         if (adapter->flags2 & IXGBE_FLAG2_UDP_TUN_REREG_NEEDED) {
7927                 rtnl_lock();
7928                 adapter->flags2 &= ~IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
7929                 udp_tunnel_get_rx_info(adapter->netdev);
7930                 rtnl_unlock();
7931         }
7932         ixgbe_reset_subtask(adapter);
7933         ixgbe_phy_interrupt_subtask(adapter);
7934         ixgbe_sfp_detection_subtask(adapter);
7935         ixgbe_sfp_link_config_subtask(adapter);
7936         ixgbe_check_overtemp_subtask(adapter);
7937         ixgbe_watchdog_subtask(adapter);
7938         ixgbe_fdir_reinit_subtask(adapter);
7939         ixgbe_check_hang_subtask(adapter);
7940 
7941         if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
7942                 ixgbe_ptp_overflow_check(adapter);
7943                 if (adapter->flags & IXGBE_FLAG_RX_HWTSTAMP_IN_REGISTER)
7944                         ixgbe_ptp_rx_hang(adapter);
7945                 ixgbe_ptp_tx_hang(adapter);
7946         }
7947 
7948         ixgbe_service_event_complete(adapter);
7949 }
7950 
7951 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
7952                      struct ixgbe_tx_buffer *first,
7953                      u8 *hdr_len,
7954                      struct ixgbe_ipsec_tx_data *itd)
7955 {
7956         u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
7957         struct sk_buff *skb = first->skb;
7958         union {
7959                 struct iphdr *v4;
7960                 struct ipv6hdr *v6;
7961                 unsigned char *hdr;
7962         } ip;
7963         union {
7964                 struct tcphdr *tcp;
7965                 unsigned char *hdr;
7966         } l4;
7967         u32 paylen, l4_offset;
7968         u32 fceof_saidx = 0;
7969         int err;
7970 
7971         if (skb->ip_summed != CHECKSUM_PARTIAL)
7972                 return 0;
7973 
7974         if (!skb_is_gso(skb))
7975                 return 0;
7976 
7977         err = skb_cow_head(skb, 0);
7978         if (err < 0)
7979                 return err;
7980 
7981         if (eth_p_mpls(first->protocol))
7982                 ip.hdr = skb_inner_network_header(skb);
7983         else
7984                 ip.hdr = skb_network_header(skb);
7985         l4.hdr = skb_checksum_start(skb);
7986 
7987         /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
7988         type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
7989 
7990         /* initialize outer IP header fields */
7991         if (ip.v4->version == 4) {
7992                 unsigned char *csum_start = skb_checksum_start(skb);
7993                 unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
7994                 int len = csum_start - trans_start;
7995 
7996                 /* IP header will have to cancel out any data that
7997                  * is not a part of the outer IP header, so set to
7998                  * a reverse csum if needed, else init check to 0.
7999                  */
8000                 ip.v4->check = (skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL) ?
8001                                            csum_fold(csum_partial(trans_start,
8002                                                                   len, 0)) : 0;
8003                 type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
8004 
8005                 ip.v4->tot_len = 0;
8006                 first->tx_flags |= IXGBE_TX_FLAGS_TSO |
8007                                    IXGBE_TX_FLAGS_CSUM |
8008                                    IXGBE_TX_FLAGS_IPV4;
8009         } else {
8010                 ip.v6->payload_len = 0;
8011                 first->tx_flags |= IXGBE_TX_FLAGS_TSO |
8012                                    IXGBE_TX_FLAGS_CSUM;
8013         }
8014 
8015         /* determine offset of inner transport header */
8016         l4_offset = l4.hdr - skb->data;
8017 
8018         /* compute length of segmentation header */
8019         *hdr_len = (l4.tcp->doff * 4) + l4_offset;
8020 
8021         /* remove payload length from inner checksum */
8022         paylen = skb->len - l4_offset;
8023         csum_replace_by_diff(&l4.tcp->check, (__force __wsum)htonl(paylen));
8024 
8025         /* update gso size and bytecount with header size */
8026         first->gso_segs = skb_shinfo(skb)->gso_segs;
8027         first->bytecount += (first->gso_segs - 1) * *hdr_len;
8028 
8029         /* mss_l4len_id: use 0 as index for TSO */
8030         mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
8031         mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
8032 
8033         fceof_saidx |= itd->sa_idx;
8034         type_tucmd |= itd->flags | itd->trailer_len;
8035 
8036         /* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
8037         vlan_macip_lens = l4.hdr - ip.hdr;
8038         vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
8039         vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
8040 
8041         ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd,
8042                           mss_l4len_idx);
8043 
8044         return 1;
8045 }
8046 
8047 static inline bool ixgbe_ipv6_csum_is_sctp(struct sk_buff *skb)
8048 {
8049         unsigned int offset = 0;
8050 
8051         ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
8052 
8053         return offset == skb_checksum_start_offset(skb);
8054 }
8055 
8056 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
8057                           struct ixgbe_tx_buffer *first,
8058                           struct ixgbe_ipsec_tx_data *itd)
8059 {
8060         struct sk_buff *skb = first->skb;
8061         u32 vlan_macip_lens = 0;
8062         u32 fceof_saidx = 0;
8063         u32 type_tucmd = 0;
8064 
8065         if (skb->ip_summed != CHECKSUM_PARTIAL) {
8066 csum_failed:
8067                 if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
8068                                          IXGBE_TX_FLAGS_CC)))
8069                         return;
8070                 goto no_csum;
8071         }
8072 
8073         switch (skb->csum_offset) {
8074         case offsetof(struct tcphdr, check):
8075                 type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
8076                 /* fall through */
8077         case offsetof(struct udphdr, check):
8078                 break;
8079         case offsetof(struct sctphdr, checksum):
8080                 /* validate that this is actually an SCTP request */
8081                 if (((first->protocol == htons(ETH_P_IP)) &&
8082                      (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
8083                     ((first->protocol == htons(ETH_P_IPV6)) &&
8084                      ixgbe_ipv6_csum_is_sctp(skb))) {
8085                         type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
8086                         break;
8087                 }
8088                 /* fall through */
8089         default:
8090                 skb_checksum_help(skb);
8091                 goto csum_failed;
8092         }
8093 
8094         /* update TX checksum flag */
8095         first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
8096         vlan_macip_lens = skb_checksum_start_offset(skb) -
8097                           skb_network_offset(skb);
8098 no_csum:
8099         /* vlan_macip_lens: MACLEN, VLAN tag */
8100         vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
8101         vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
8102 
8103         fceof_saidx |= itd->sa_idx;
8104         type_tucmd |= itd->flags | itd->trailer_len;
8105 
8106         ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd, 0);
8107 }
8108 
8109 #define IXGBE_SET_FLAG(_input, _flag, _result) \
8110         ((_flag <= _result) ? \
8111          ((u32)(_input & _flag) * (_result / _flag)) : \
8112          ((u32)(_input & _flag) / (_flag / _result)))
8113 
8114 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
8115 {
8116         /* set type for advanced descriptor with frame checksum insertion */
8117         u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
8118                        IXGBE_ADVTXD_DCMD_DEXT |
8119                        IXGBE_ADVTXD_DCMD_IFCS;
8120 
8121         /* set HW vlan bit if vlan is present */
8122         cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
8123                                    IXGBE_ADVTXD_DCMD_VLE);
8124 
8125         /* set segmentation enable bits for TSO/FSO */
8126         cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
8127                                    IXGBE_ADVTXD_DCMD_TSE);
8128 
8129         /* set timestamp bit if present */
8130         cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
8131                                    IXGBE_ADVTXD_MAC_TSTAMP);
8132 
8133         /* insert frame checksum */
8134         cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
8135 
8136         return cmd_type;
8137 }
8138 
8139 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
8140                                    u32 tx_flags, unsigned int paylen)
8141 {
8142         u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
8143 
8144         /* enable L4 checksum for TSO and TX checksum offload */
8145         olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8146                                         IXGBE_TX_FLAGS_CSUM,
8147                                         IXGBE_ADVTXD_POPTS_TXSM);
8148 
8149         /* enable IPv4 checksum for TSO */
8150         olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8151                                         IXGBE_TX_FLAGS_IPV4,
8152                                         IXGBE_ADVTXD_POPTS_IXSM);
8153 
8154         /* enable IPsec */
8155         olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8156                                         IXGBE_TX_FLAGS_IPSEC,
8157                                         IXGBE_ADVTXD_POPTS_IPSEC);
8158 
8159         /*
8160          * Check Context must be set if Tx switch is enabled, which it
8161          * always is for case where virtual functions are running
8162          */
8163         olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8164                                         IXGBE_TX_FLAGS_CC,
8165                                         IXGBE_ADVTXD_CC);
8166 
8167         tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
8168 }
8169 
8170 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
8171 {
8172         netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
8173 
8174         /* Herbert's original patch had:
8175          *  smp_mb__after_netif_stop_queue();
8176          * but since that doesn't exist yet, just open code it.
8177          */
8178         smp_mb();
8179 
8180         /* We need to check again in a case another CPU has just
8181          * made room available.
8182          */
8183         if (likely(ixgbe_desc_unused(tx_ring) < size))
8184                 return -EBUSY;
8185 
8186         /* A reprieve! - use start_queue because it doesn't call schedule */
8187         netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
8188         ++tx_ring->tx_stats.restart_queue;
8189         return 0;
8190 }
8191 
8192 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
8193 {
8194         if (likely(ixgbe_desc_unused(tx_ring) >= size))
8195                 return 0;
8196 
8197         return __ixgbe_maybe_stop_tx(tx_ring, size);
8198 }
8199 
8200 static int ixgbe_tx_map(struct ixgbe_ring *tx_ring,
8201                         struct ixgbe_tx_buffer *first,
8202                         const u8 hdr_len)
8203 {
8204         struct sk_buff *skb = first->skb;
8205         struct ixgbe_tx_buffer *tx_buffer;
8206         union ixgbe_adv_tx_desc *tx_desc;
8207         skb_frag_t *frag;
8208         dma_addr_t dma;
8209         unsigned int data_len, size;
8210         u32 tx_flags = first->tx_flags;
8211         u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
8212         u16 i = tx_ring->next_to_use;
8213 
8214         tx_desc = IXGBE_TX_DESC(tx_ring, i);
8215 
8216         ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
8217 
8218         size = skb_headlen(skb);
8219         data_len = skb->data_len;
8220 
8221 #ifdef IXGBE_FCOE
8222         if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
8223                 if (data_len < sizeof(struct fcoe_crc_eof)) {
8224                         size -= sizeof(struct fcoe_crc_eof) - data_len;
8225                         data_len = 0;
8226                 } else {
8227                         data_len -= sizeof(struct fcoe_crc_eof);
8228                 }
8229         }
8230 
8231 #endif
8232         dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
8233 
8234         tx_buffer = first;
8235 
8236         for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
8237                 if (dma_mapping_error(tx_ring->dev, dma))
8238                         goto dma_error;
8239 
8240                 /* record length, and DMA address */
8241                 dma_unmap_len_set(tx_buffer, len, size);
8242                 dma_unmap_addr_set(tx_buffer, dma, dma);
8243 
8244                 tx_desc->read.buffer_addr = cpu_to_le64(dma);
8245 
8246                 while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
8247                         tx_desc->read.cmd_type_len =
8248                                 cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
8249 
8250                         i++;
8251                         tx_desc++;
8252                         if (i == tx_ring->count) {
8253                                 tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8254                                 i = 0;
8255                         }
8256                         tx_desc->read.olinfo_status = 0;
8257 
8258                         dma += IXGBE_MAX_DATA_PER_TXD;
8259                         size -= IXGBE_MAX_DATA_PER_TXD;
8260 
8261                         tx_desc->read.buffer_addr = cpu_to_le64(dma);
8262                 }
8263 
8264                 if (likely(!data_len))
8265                         break;
8266 
8267                 tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
8268 
8269                 i++;
8270                 tx_desc++;
8271                 if (i == tx_ring->count) {
8272                         tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8273                         i = 0;
8274                 }
8275                 tx_desc->read.olinfo_status = 0;
8276 
8277 #ifdef IXGBE_FCOE
8278                 size = min_t(unsigned int, data_len, skb_frag_size(frag));
8279 #else
8280                 size = skb_frag_size(frag);
8281 #endif
8282                 data_len -= size;
8283 
8284                 dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
8285                                        DMA_TO_DEVICE);
8286 
8287                 tx_buffer = &tx_ring->tx_buffer_info[i];
8288         }
8289 
8290         /* write last descriptor with RS and EOP bits */
8291         cmd_type |= size | IXGBE_TXD_CMD;
8292         tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8293 
8294         netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
8295 
8296         /* set the timestamp */
8297         first->time_stamp = jiffies;
8298 
8299         skb_tx_timestamp(skb);
8300 
8301         /*
8302          * Force memory writes to complete before letting h/w know there
8303          * are new descriptors to fetch.  (Only applicable for weak-ordered
8304          * memory model archs, such as IA-64).
8305          *
8306          * We also need this memory barrier to make certain all of the
8307          * status bits have been updated before next_to_watch is written.
8308          */
8309         wmb();
8310 
8311         /* set next_to_watch value indicating a packet is present */
8312         first->next_to_watch = tx_desc;
8313 
8314         i++;
8315         if (i == tx_ring->count)
8316                 i = 0;
8317 
8318         tx_ring->next_to_use = i;
8319 
8320         ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
8321 
8322         if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more()) {
8323                 writel(i, tx_ring->tail);
8324         }
8325 
8326         return 0;
8327 dma_error:
8328         dev_err(tx_ring->dev, "TX DMA map failed\n");
8329 
8330         /* clear dma mappings for failed tx_buffer_info map */
8331         for (;;) {
8332                 tx_buffer = &tx_ring->tx_buffer_info[i];
8333                 if (dma_unmap_len(tx_buffer, len))
8334                         dma_unmap_page(tx_ring->dev,
8335                                        dma_unmap_addr(tx_buffer, dma),
8336                                        dma_unmap_len(tx_buffer, len),
8337                                        DMA_TO_DEVICE);
8338                 dma_unmap_len_set(tx_buffer, len, 0);
8339                 if (tx_buffer == first)
8340                         break;
8341                 if (i == 0)
8342                         i += tx_ring->count;
8343                 i--;
8344         }
8345 
8346         dev_kfree_skb_any(first->skb);
8347         first->skb = NULL;
8348 
8349         tx_ring->next_to_use = i;
8350 
8351         return -1;
8352 }
8353 
8354 static void ixgbe_atr(struct ixgbe_ring *ring,
8355                       struct ixgbe_tx_buffer *first)
8356 {
8357         struct ixgbe_q_vector *q_vector = ring->q_vector;
8358         union ixgbe_atr_hash_dword input = { .dword = 0 };
8359         union ixgbe_atr_hash_dword common = { .dword = 0 };
8360         union {
8361                 unsigned char *network;
8362                 struct iphdr *ipv4;
8363                 struct ipv6hdr *ipv6;
8364         } hdr;
8365         struct tcphdr *th;
8366         unsigned int hlen;
8367         struct sk_buff *skb;
8368         __be16 vlan_id;
8369         int l4_proto;
8370 
8371         /* if ring doesn't have a interrupt vector, cannot perform ATR */
8372         if (!q_vector)
8373                 return;
8374 
8375         /* do nothing if sampling is disabled */
8376         if (!ring->atr_sample_rate)
8377                 return;
8378 
8379         ring->atr_count++;
8380 
8381         /* currently only IPv4/IPv6 with TCP is supported */
8382         if ((first->protocol != htons(ETH_P_IP)) &&
8383             (first->protocol != htons(ETH_P_IPV6)))
8384                 return;
8385 
8386         /* snag network header to get L4 type and address */
8387         skb = first->skb;
8388         hdr.network = skb_network_header(skb);
8389         if (unlikely(hdr.network <= skb->data))
8390                 return;
8391         if (skb->encapsulation &&
8392             first->protocol == htons(ETH_P_IP) &&
8393             hdr.ipv4->protocol == IPPROTO_UDP) {
8394                 struct ixgbe_adapter *adapter = q_vector->adapter;
8395 
8396                 if (unlikely(skb_tail_pointer(skb) < hdr.network +
8397                              VXLAN_HEADROOM))
8398                         return;
8399 
8400                 /* verify the port is recognized as VXLAN */
8401                 if (adapter->vxlan_port &&
8402                     udp_hdr(skb)->dest == adapter->vxlan_port)
8403                         hdr.network = skb_inner_network_header(skb);
8404 
8405                 if (adapter->geneve_port &&
8406                     udp_hdr(skb)->dest == adapter->geneve_port)
8407                         hdr.network = skb_inner_network_header(skb);
8408         }
8409 
8410         /* Make sure we have at least [minimum IPv4 header + TCP]
8411          * or [IPv6 header] bytes
8412          */
8413         if (unlikely(skb_tail_pointer(skb) < hdr.network + 40))
8414                 return;
8415 
8416         /* Currently only IPv4/IPv6 with TCP is supported */
8417         switch (hdr.ipv4->version) {
8418         case IPVERSION:
8419                 /* access ihl as u8 to avoid unaligned access on ia64 */
8420                 hlen = (hdr.network[0] & 0x0F) << 2;
8421                 l4_proto = hdr.ipv4->protocol;
8422                 break;
8423         case 6:
8424                 hlen = hdr.network - skb->data;
8425                 l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
8426                 hlen -= hdr.network - skb->data;
8427                 break;
8428         default:
8429                 return;
8430         }
8431 
8432         if (l4_proto != IPPROTO_TCP)
8433                 return;
8434 
8435         if (unlikely(skb_tail_pointer(skb) < hdr.network +
8436                      hlen + sizeof(struct tcphdr)))
8437                 return;
8438 
8439         th = (struct tcphdr *)(hdr.network + hlen);
8440 
8441         /* skip this packet since the socket is closing */
8442         if (th->fin)
8443                 return;
8444 
8445         /* sample on all syn packets or once every atr sample count */
8446         if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
8447                 return;
8448 
8449         /* reset sample count */
8450         ring->atr_count = 0;
8451 
8452         vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
8453 
8454         /*
8455          * src and dst are inverted, think how the receiver sees them
8456          *
8457          * The input is broken into two sections, a non-compressed section
8458          * containing vm_pool, vlan_id, and flow_type.  The rest of the data
8459          * is XORed together and stored in the compressed dword.
8460          */
8461         input.formatted.vlan_id = vlan_id;
8462 
8463         /*
8464          * since src port and flex bytes occupy the same word XOR them together
8465          * and write the value to source port portion of compressed dword
8466          */
8467         if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
8468                 common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
8469         else
8470                 common.port.src ^= th->dest ^ first->protocol;
8471         common.port.dst ^= th->source;
8472 
8473         switch (hdr.ipv4->version) {
8474         case IPVERSION:
8475                 input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
8476                 common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
8477                 break;
8478         case 6:
8479                 input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
8480                 common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
8481                              hdr.ipv6->saddr.s6_addr32[1] ^
8482                              hdr.ipv6->saddr.s6_addr32[2] ^
8483                              hdr.ipv6->saddr.s6_addr32[3] ^
8484                              hdr.ipv6->daddr.s6_addr32[0] ^
8485                              hdr.ipv6->daddr.s6_addr32[1] ^
8486                              hdr.ipv6->daddr.s6_addr32[2] ^
8487                              hdr.ipv6->daddr.s6_addr32[3];
8488                 break;
8489         default:
8490                 break;
8491         }
8492 
8493         if (hdr.network != skb_network_header(skb))
8494                 input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
8495 
8496         /* This assumes the Rx queue and Tx queue are bound to the same CPU */
8497         ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
8498                                               input, common, ring->queue_index);
8499 }
8500 
8501 #ifdef IXGBE_FCOE
8502 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
8503                               struct net_device *sb_dev)
8504 {
8505         struct ixgbe_adapter *adapter;
8506         struct ixgbe_ring_feature *f;
8507         int txq;
8508 
8509         if (sb_dev) {
8510                 u8 tc = netdev_get_prio_tc_map(dev, skb->priority);
8511                 struct net_device *vdev = sb_dev;
8512 
8513                 txq = vdev->tc_to_txq[tc].offset;
8514                 txq += reciprocal_scale(skb_get_hash(skb),
8515                                         vdev->tc_to_txq[tc].count);
8516 
8517                 return txq;
8518         }
8519 
8520         /*
8521          * only execute the code below if protocol is FCoE
8522          * or FIP and we have FCoE enabled on the adapter
8523          */
8524         switch (vlan_get_protocol(skb)) {
8525         case htons(ETH_P_FCOE):
8526         case htons(ETH_P_FIP):
8527                 adapter = netdev_priv(dev);
8528 
8529                 if (!sb_dev && (adapter->flags & IXGBE_FLAG_FCOE_ENABLED))
8530                         break;
8531                 /* fall through */
8532         default:
8533                 return netdev_pick_tx(dev, skb, sb_dev);
8534         }
8535 
8536         f = &adapter->ring_feature[RING_F_FCOE];
8537 
8538         txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
8539                                            smp_processor_id();
8540 
8541         while (txq >= f->indices)
8542                 txq -= f->indices;
8543 
8544         return txq + f->offset;
8545 }
8546 
8547 #endif
8548 int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
8549                         struct xdp_frame *xdpf)
8550 {
8551         struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
8552         struct ixgbe_tx_buffer *tx_buffer;
8553         union ixgbe_adv_tx_desc *tx_desc;
8554         u32 len, cmd_type;
8555         dma_addr_t dma;
8556         u16 i;
8557 
8558         len = xdpf->len;
8559 
8560         if (unlikely(!ixgbe_desc_unused(ring)))
8561                 return IXGBE_XDP_CONSUMED;
8562 
8563         dma = dma_map_single(ring->dev, xdpf->data, len, DMA_TO_DEVICE);
8564         if (dma_mapping_error(ring->dev, dma))
8565                 return IXGBE_XDP_CONSUMED;
8566 
8567         /* record the location of the first descriptor for this packet */
8568         tx_buffer = &ring->tx_buffer_info[ring->next_to_use];
8569         tx_buffer->bytecount = len;
8570         tx_buffer->gso_segs = 1;
8571         tx_buffer->protocol = 0;
8572 
8573         i = ring->next_to_use;
8574         tx_desc = IXGBE_TX_DESC(ring, i);
8575 
8576         dma_unmap_len_set(tx_buffer, len, len);
8577         dma_unmap_addr_set(tx_buffer, dma, dma);
8578         tx_buffer->xdpf = xdpf;
8579 
8580         tx_desc->read.buffer_addr = cpu_to_le64(dma);
8581 
8582         /* put descriptor type bits */
8583         cmd_type = IXGBE_ADVTXD_DTYP_DATA |
8584                    IXGBE_ADVTXD_DCMD_DEXT |
8585                    IXGBE_ADVTXD_DCMD_IFCS;
8586         cmd_type |= len | IXGBE_TXD_CMD;
8587         tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8588         tx_desc->read.olinfo_status =
8589                 cpu_to_le32(len << IXGBE_ADVTXD_PAYLEN_SHIFT);
8590 
8591         /* Avoid any potential race with xdp_xmit and cleanup */
8592         smp_wmb();
8593 
8594         /* set next_to_watch value indicating a packet is present */
8595         i++;
8596         if (i == ring->count)
8597                 i = 0;
8598 
8599         tx_buffer->next_to_watch = tx_desc;
8600         ring->next_to_use = i;
8601 
8602         return IXGBE_XDP_TX;
8603 }
8604 
8605 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
8606                           struct ixgbe_adapter *adapter,
8607                           struct ixgbe_ring *tx_ring)
8608 {
8609         struct ixgbe_tx_buffer *first;
8610         int tso;
8611         u32 tx_flags = 0;
8612         unsigned short f;
8613         u16 count = TXD_USE_COUNT(skb_headlen(skb));
8614         struct ixgbe_ipsec_tx_data ipsec_tx = { 0 };
8615         __be16 protocol = skb->protocol;
8616         u8 hdr_len = 0;
8617 
8618         /*
8619          * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
8620          *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
8621          *       + 2 desc gap to keep tail from touching head,
8622          *       + 1 desc for context descriptor,
8623          * otherwise try next time
8624          */
8625         for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
8626                 count += TXD_USE_COUNT(skb_frag_size(
8627                                                 &skb_shinfo(skb)->frags[f]));
8628 
8629         if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
8630                 tx_ring->tx_stats.tx_busy++;
8631                 return NETDEV_TX_BUSY;
8632         }
8633 
8634         /* record the location of the first descriptor for this packet */
8635         first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
8636         first->skb = skb;
8637         first->bytecount = skb->len;
8638         first->gso_segs = 1;
8639 
8640         /* if we have a HW VLAN tag being added default to the HW one */
8641         if (skb_vlan_tag_present(skb)) {
8642                 tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
8643                 tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8644         /* else if it is a SW VLAN check the next protocol and store the tag */
8645         } else if (protocol == htons(ETH_P_8021Q)) {
8646                 struct vlan_hdr *vhdr, _vhdr;
8647                 vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
8648                 if (!vhdr)
8649                         goto out_drop;
8650 
8651                 tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
8652                                   IXGBE_TX_FLAGS_VLAN_SHIFT;
8653                 tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
8654         }
8655         protocol = vlan_get_protocol(skb);
8656 
8657         if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
8658             adapter->ptp_clock) {
8659                 if (adapter->tstamp_config.tx_type == HWTSTAMP_TX_ON &&
8660                     !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
8661                                            &adapter->state)) {
8662                         skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
8663                         tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
8664 
8665                         /* schedule check for Tx timestamp */
8666                         adapter->ptp_tx_skb = skb_get(skb);
8667                         adapter->ptp_tx_start = jiffies;
8668                         schedule_work(&adapter->ptp_tx_work);
8669                 } else {
8670                         adapter->tx_hwtstamp_skipped++;
8671                 }
8672         }
8673 
8674 #ifdef CONFIG_PCI_IOV
8675         /*
8676          * Use the l2switch_enable flag - would be false if the DMA
8677          * Tx switch had been disabled.
8678          */
8679         if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
8680                 tx_flags |= IXGBE_TX_FLAGS_CC;
8681 
8682 #endif
8683         /* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
8684         if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8685             ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
8686              (skb->priority != TC_PRIO_CONTROL))) {
8687                 tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
8688                 tx_flags |= (skb->priority & 0x7) <<
8689                                         IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
8690                 if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
8691                         struct vlan_ethhdr *vhdr;
8692 
8693                         if (skb_cow_head(skb, 0))
8694                                 goto out_drop;
8695                         vhdr = (struct vlan_ethhdr *)skb->data;
8696                         vhdr->h_vlan_TCI = htons(tx_flags >>
8697                                                  IXGBE_TX_FLAGS_VLAN_SHIFT);
8698                 } else {
8699                         tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8700                 }
8701         }
8702 
8703         /* record initial flags and protocol */
8704         first->tx_flags = tx_flags;
8705         first->protocol = protocol;
8706 
8707 #ifdef IXGBE_FCOE
8708         /* setup tx offload for FCoE */
8709         if ((protocol == htons(ETH_P_FCOE)) &&
8710             (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
8711                 tso = ixgbe_fso(tx_ring, first, &hdr_len);
8712                 if (tso < 0)
8713                         goto out_drop;
8714 
8715                 goto xmit_fcoe;
8716         }
8717 
8718 #endif /* IXGBE_FCOE */
8719 
8720 #ifdef CONFIG_IXGBE_IPSEC
8721         if (xfrm_offload(skb) &&
8722             !ixgbe_ipsec_tx(tx_ring, first, &ipsec_tx))
8723                 goto out_drop;
8724 #endif
8725         tso = ixgbe_tso(tx_ring, first, &hdr_len, &ipsec_tx);
8726         if (tso < 0)
8727                 goto out_drop;
8728         else if (!tso)
8729                 ixgbe_tx_csum(tx_ring, first, &ipsec_tx);
8730 
8731         /* add the ATR filter if ATR is on */
8732         if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
8733                 ixgbe_atr(tx_ring, first);
8734 
8735 #ifdef IXGBE_FCOE
8736 xmit_fcoe:
8737 #endif /* IXGBE_FCOE */
8738         if (ixgbe_tx_map(tx_ring, first, hdr_len))
8739                 goto cleanup_tx_timestamp;
8740 
8741         return NETDEV_TX_OK;
8742 
8743 out_drop:
8744         dev_kfree_skb_any(first->skb);
8745         first->skb = NULL;
8746 cleanup_tx_timestamp:
8747         if (unlikely(tx_flags & IXGBE_TX_FLAGS_TSTAMP)) {
8748                 dev_kfree_skb_any(adapter->ptp_tx_skb);
8749                 adapter->ptp_tx_skb = NULL;
8750                 cancel_work_sync(&adapter->ptp_tx_work);
8751                 clear_bit_unlock(__IXGBE_PTP_TX_IN_PROGRESS, &adapter->state);
8752         }
8753 
8754         return NETDEV_TX_OK;
8755 }
8756 
8757 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
8758                                       struct net_device *netdev,
8759                                       struct ixgbe_ring *ring)
8760 {
8761         struct ixgbe_adapter *adapter = netdev_priv(netdev);
8762         struct ixgbe_ring *tx_ring;
8763 
8764         /*
8765          * The minimum packet size for olinfo paylen is 17 so pad the skb
8766          * in order to meet this minimum size requirement.
8767          */
8768         if (skb_put_padto(skb, 17))
8769                 return NETDEV_TX_OK;
8770 
8771         tx_ring = ring ? ring : adapter->tx_ring[skb_get_queue_mapping(skb)];
8772         if (unlikely(test_bit(__IXGBE_TX_DISABLED, &tx_ring->state)))
8773                 return NETDEV_TX_BUSY;
8774 
8775         return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
8776 }
8777 
8778 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
8779                                     struct net_device *netdev)
8780 {
8781         return __ixgbe_xmit_frame(skb, netdev, NULL);
8782 }
8783 
8784 /**
8785  * ixgbe_set_mac - Change the Ethernet Address of the NIC
8786  * @netdev: network interface device structure
8787  * @p: pointer to an address structure
8788  *
8789  * Returns 0 on success, negative on failure
8790  **/
8791 static int ixgbe_set_mac(struct net_device *netdev, void *p)
8792 {
8793         struct ixgbe_adapter *adapter = netdev_priv(netdev);
8794         struct ixgbe_hw *hw = &adapter->hw;
8795         struct sockaddr *addr = p;
8796 
8797         if (!is_valid_ether_addr(addr->sa_data))
8798                 return -EADDRNOTAVAIL;
8799 
8800         memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
8801         memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
8802 
8803         ixgbe_mac_set_default_filter(adapter);
8804 
8805         return 0;
8806 }
8807 
8808 static int
8809 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
8810 {
8811         struct ixgbe_adapter *adapter = netdev_priv(netdev);
8812         struct ixgbe_hw *hw = &adapter->hw;
8813         u16 value;
8814         int rc;
8815 
8816         if (adapter->mii_bus) {
8817                 int regnum = addr;
8818 
8819                 if (devad != MDIO_DEVAD_NONE)
8820                         regnum |= (devad << 16) | MII_ADDR_C45;
8821 
8822                 return mdiobus_read(adapter->mii_bus, prtad, regnum);
8823         }
8824 
8825         if (prtad != hw->phy.mdio.prtad)
8826                 return -EINVAL;
8827         rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
8828         if (!rc)
8829                 rc = value;
8830         return rc;
8831 }
8832 
8833 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
8834                             u16 addr, u16 value)
8835 {
8836         struct ixgbe_adapter *adapter = netdev_priv(netdev);
8837         struct ixgbe_hw *hw = &adapter->hw;
8838 
8839         if (adapter->mii_bus) {
8840                 int regnum = addr;
8841 
8842                 if (devad != MDIO_DEVAD_NONE)
8843                         regnum |= (devad << 16) | MII_ADDR_C45;
8844 
8845                 return mdiobus_write(adapter->mii_bus, prtad, regnum, value);
8846         }
8847 
8848         if (prtad != hw->phy.mdio.prtad)
8849                 return -EINVAL;
8850         return hw->phy.ops.write_reg(hw, addr, devad, value);
8851 }
8852 
8853 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8854 {
8855         struct ixgbe_adapter *adapter = netdev_priv(netdev);
8856 
8857         switch (cmd) {
8858         case SIOCSHWTSTAMP:
8859                 return ixgbe_ptp_set_ts_config(adapter, req);
8860         case SIOCGHWTSTAMP:
8861                 return ixgbe_ptp_get_ts_config(adapter, req);
8862         case SIOCGMIIPHY:
8863                 if (!adapter->hw.phy.ops.read_reg)
8864                         return -EOPNOTSUPP;
8865                 /* fall through */
8866         default:
8867                 return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8868         }
8869 }
8870 
8871 /**
8872  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8873  * netdev->dev_addrs
8874  * @dev: network interface device structure
8875  *
8876  * Returns non-zero on failure
8877  **/
8878 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8879 {
8880         int err = 0;
8881         struct ixgbe_adapter *adapter = netdev_priv(dev);
8882         struct ixgbe_hw *hw = &adapter->hw;
8883 
8884         if (is_valid_ether_addr(hw->mac.san_addr)) {
8885                 rtnl_lock();
8886                 err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
8887                 rtnl_unlock();
8888 
8889                 /* update SAN MAC vmdq pool selection */
8890                 hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
8891         }
8892         return err;
8893 }
8894 
8895 /**
8896  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
8897  * netdev->dev_addrs
8898  * @dev: network interface device structure
8899  *
8900  * Returns non-zero on failure
8901  **/
8902 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
8903 {
8904         int err = 0;
8905         struct ixgbe_adapter *adapter = netdev_priv(dev);
8906         struct ixgbe_mac_info *mac = &adapter->hw.mac;
8907 
8908         if (is_valid_ether_addr(mac->san_addr)) {
8909                 rtnl_lock();
8910                 err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
8911                 rtnl_unlock();
8912         }
8913         return err;
8914 }
8915 
8916 static void ixgbe_get_ring_stats64(struct rtnl_link_stats64 *stats,
8917                                    struct ixgbe_ring *ring)
8918 {
8919         u64 bytes, packets;
8920         unsigned int start;
8921 
8922         if (ring) {
8923                 do {
8924                         start = u64_stats_fetch_begin_irq(&ring->syncp);
8925                         packets = ring->stats.packets;
8926                         bytes   = ring->stats.bytes;
8927                 } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8928                 stats->tx_packets += packets;
8929                 stats->tx_bytes   += bytes;
8930         }
8931 }
8932 
8933 static void ixgbe_get_stats64(struct net_device *netdev,
8934                               struct rtnl_link_stats64 *stats)
8935 {
8936         struct ixgbe_adapter *adapter = netdev_priv(netdev);
8937         int i;
8938 
8939         rcu_read_lock();
8940         for (i = 0; i < adapter->num_rx_queues; i++) {
8941                 struct ixgbe_ring *ring = READ_ONCE(adapter->rx_ring[i]);
8942                 u64 bytes, packets;
8943                 unsigned int start;
8944 
8945                 if (ring) {
8946                         do {
8947                                 start = u64_stats_fetch_begin_irq(&ring->syncp);
8948                                 packets = ring->stats.packets;
8949                                 bytes   = ring->stats.bytes;
8950                         } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
8951                         stats->rx_packets += packets;
8952                         stats->rx_bytes   += bytes;
8953                 }
8954         }
8955 
8956         for (i = 0; i < adapter->num_tx_queues; i++) {
8957                 struct ixgbe_ring *ring = READ_ONCE(adapter->tx_ring[i]);
8958 
8959                 ixgbe_get_ring_stats64(stats, ring);
8960         }
8961         for (i = 0; i < adapter->num_xdp_queues; i++) {
8962                 struct ixgbe_ring *ring = READ_ONCE(adapter->xdp_ring[i]);
8963 
8964                 ixgbe_get_ring_stats64(stats, ring);
8965         }
8966         rcu_read_unlock();
8967 
8968         /* following stats updated by ixgbe_watchdog_task() */
8969         stats->multicast        = netdev->stats.multicast;
8970         stats->rx_errors        = netdev->stats.rx_errors;
8971         stats->rx_length_errors = netdev->stats.rx_length_errors;
8972         stats->rx_crc_errors    = netdev->stats.rx_crc_errors;
8973         stats->rx_missed_errors = netdev->stats.rx_missed_errors;
8974 }
8975 
8976 #ifdef CONFIG_IXGBE_DCB
8977 /**
8978  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
8979  * @adapter: pointer to ixgbe_adapter
8980  * @tc: number of traffic classes currently enabled
8981  *
8982  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
8983  * 802.1Q priority maps to a packet buffer that exists.
8984  */
8985 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
8986 {
8987         struct ixgbe_hw *hw = &adapter->hw;
8988         u32 reg, rsave;
8989         int i;
8990 
8991         /* 82598 have a static priority to TC mapping that can not
8992          * be changed so no validation is needed.
8993          */
8994         if (hw->mac.type == ixgbe_mac_82598EB)
8995                 return;
8996 
8997         reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
8998         rsave = reg;
8999 
9000         for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
9001                 u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
9002 
9003                 /* If up2tc is out of bounds default to zero */
9004                 if (up2tc > tc)
9005                         reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
9006         }
9007 
9008         if (reg != rsave)
9009                 IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
9010 
9011         return;
9012 }
9013 
9014 /**
9015  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
9016  * @adapter: Pointer to adapter struct
9017  *
9018  * Populate the netdev user priority to tc map
9019  */
9020 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
9021 {
9022         struct net_device *dev = adapter->netdev;
9023         struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
9024         struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
9025         u8 prio;
9026 
9027         for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
9028                 u8 tc = 0;
9029 
9030                 if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
9031                         tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
9032                 else if (ets)
9033                         tc = ets->prio_tc[prio];
9034 
9035                 netdev_set_prio_tc_map(dev, prio, tc);
9036         }
9037 }
9038 
9039 #endif /* CONFIG_IXGBE_DCB */
9040 static int ixgbe_reassign_macvlan_pool(struct net_device *vdev, void *data)
9041 {
9042         struct ixgbe_adapter *adapter = data;
9043         struct ixgbe_fwd_adapter *accel;
9044         int pool;
9045 
9046         /* we only care about macvlans... */
9047         if (!netif_is_macvlan(vdev))
9048                 return 0;
9049 
9050         /* that have hardware offload enabled... */
9051         accel = macvlan_accel_priv(vdev);
9052         if (!accel)
9053                 return 0;
9054 
9055         /* If we can relocate to a different bit do so */
9056         pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
9057         if (pool < adapter->num_rx_pools) {
9058                 set_bit(pool, adapter->fwd_bitmask);
9059                 accel->pool = pool;
9060                 return 0;
9061         }
9062 
9063         /* if we cannot find a free pool then disable the offload */
9064         netdev_err(vdev, "L2FW offload disabled due to lack of queue resources\n");
9065         macvlan_release_l2fw_offload(vdev);
9066 
9067         /* unbind the queues and drop the subordinate channel config */
9068         netdev_unbind_sb_channel(adapter->netdev, vdev);
9069         netdev_set_sb_channel(vdev, 0);
9070 
9071         kfree(accel);
9072 
9073         return 0;
9074 }
9075 
9076 static void ixgbe_defrag_macvlan_pools(struct net_device *dev)
9077 {
9078         struct ixgbe_adapter *adapter = netdev_priv(dev);
9079 
9080         /* flush any stale bits out of the fwd bitmask */
9081         bitmap_clear(adapter->fwd_bitmask, 1, 63);
9082 
9083         /* walk through upper devices reassigning pools */
9084         netdev_walk_all_upper_dev_rcu(dev, ixgbe_reassign_macvlan_pool,
9085                                       adapter);
9086 }
9087 
9088 /**
9089  * ixgbe_setup_tc - configure net_device for multiple traffic classes
9090  *
9091  * @dev: net device to configure
9092  * @tc: number of traffic classes to enable
9093  */
9094 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
9095 {
9096         struct ixgbe_adapter *adapter = netdev_priv(dev);
9097         struct ixgbe_hw *hw = &adapter->hw;
9098 
9099         /* Hardware supports up to 8 traffic classes */
9100         if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
9101                 return -EINVAL;
9102 
9103         if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
9104                 return -EINVAL;
9105 
9106         /* Hardware has to reinitialize queues and interrupts to
9107          * match packet buffer alignment. Unfortunately, the
9108          * hardware is not flexible enough to do this dynamically.
9109          */
9110         if (netif_running(dev))
9111                 ixgbe_close(dev);
9112         else
9113                 ixgbe_reset(adapter);
9114 
9115         ixgbe_clear_interrupt_scheme(adapter);
9116 
9117 #ifdef CONFIG_IXGBE_DCB
9118         if (tc) {
9119                 if (adapter->xdp_prog) {
9120                         e_warn(probe, "DCB is not supported with XDP\n");
9121 
9122                         ixgbe_init_interrupt_scheme(adapter);
9123                         if (netif_running(dev))
9124                                 ixgbe_open(dev);
9125                         return -EINVAL;
9126                 }
9127 
9128                 netdev_set_num_tc(dev, tc);
9129                 ixgbe_set_prio_tc_map(adapter);
9130 
9131                 adapter->hw_tcs = tc;
9132                 adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
9133 
9134                 if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
9135                         adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
9136                         adapter->hw.fc.requested_mode = ixgbe_fc_none;
9137                 }
9138         } else {
9139                 netdev_reset_tc(dev);
9140 
9141                 if (adapter->hw.mac.type == ixgbe_mac_82598EB)
9142                         adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
9143 
9144                 adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
9145                 adapter->hw_tcs = tc;
9146 
9147                 adapter->temp_dcb_cfg.pfc_mode_enable = false;
9148                 adapter->dcb_cfg.pfc_mode_enable = false;
9149         }
9150 
9151         ixgbe_validate_rtr(adapter, tc);
9152 
9153 #endif /* CONFIG_IXGBE_DCB */
9154         ixgbe_init_interrupt_scheme(adapter);
9155 
9156         ixgbe_defrag_macvlan_pools(dev);
9157 
9158         if (netif_running(dev))
9159                 return ixgbe_open(dev);
9160 
9161         return 0;
9162 }
9163 
9164 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
9165                                struct tc_cls_u32_offload *cls)
9166 {
9167         u32 hdl = cls->knode.handle;
9168         u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
9169         u32 loc = cls->knode.handle & 0xfffff;
9170         int err = 0, i, j;
9171         struct ixgbe_jump_table *jump = NULL;
9172 
9173         if (loc > IXGBE_MAX_HW_ENTRIES)
9174                 return -EINVAL;
9175 
9176         if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
9177                 return -EINVAL;
9178 
9179         /* Clear this filter in the link data it is associated with */
9180         if (uhtid != 0x800) {
9181                 jump = adapter->jump_tables[uhtid];
9182                 if (!jump)
9183                         return -EINVAL;
9184                 if (!test_bit(loc - 1, jump->child_loc_map))
9185                         return -EINVAL;
9186                 clear_bit(loc - 1, jump->child_loc_map);
9187         }
9188 
9189         /* Check if the filter being deleted is a link */
9190         for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9191                 jump = adapter->jump_tables[i];
9192                 if (jump && jump->link_hdl == hdl) {
9193                         /* Delete filters in the hardware in the child hash
9194                          * table associated with this link
9195                          */
9196                         for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
9197                                 if (!test_bit(j, jump->child_loc_map))
9198                                         continue;
9199                                 spin_lock(&adapter->fdir_perfect_lock);
9200                                 err = ixgbe_update_ethtool_fdir_entry(adapter,
9201                                                                       NULL,
9202                                                                       j + 1);
9203                                 spin_unlock(&adapter->fdir_perfect_lock);
9204                                 clear_bit(j, jump->child_loc_map);
9205                         }
9206                         /* Remove resources for this link */
9207                         kfree(jump->input);
9208                         kfree(jump->mask);
9209                         kfree(jump);
9210                         adapter->jump_tables[i] = NULL;
9211                         return err;
9212                 }
9213         }
9214 
9215         spin_lock(&adapter->fdir_perfect_lock);
9216         err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
9217         spin_unlock(&adapter->fdir_perfect_lock);
9218         return err;
9219 }
9220 
9221 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
9222                                             struct tc_cls_u32_offload *cls)
9223 {
9224         u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
9225 
9226         if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9227                 return -EINVAL;
9228 
9229         /* This ixgbe devices do not support hash tables at the moment
9230          * so abort when given hash tables.
9231          */
9232         if (cls->hnode.divisor > 0)
9233                 return -EINVAL;
9234 
9235         set_bit(uhtid - 1, &adapter->tables);
9236         return 0;
9237 }
9238 
9239 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
9240                                             struct tc_cls_u32_offload *cls)
9241 {
9242         u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
9243 
9244         if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9245                 return -EINVAL;
9246 
9247         clear_bit(uhtid - 1, &adapter->tables);
9248         return 0;
9249 }
9250 
9251 #ifdef CONFIG_NET_CLS_ACT
9252 struct upper_walk_data {
9253         struct ixgbe_adapter *adapter;
9254         u64 action;
9255         int ifindex;
9256         u8 queue;
9257 };
9258 
9259 static int get_macvlan_queue(struct net_device *upper, void *_data)
9260 {
9261         if (netif_is_macvlan(upper)) {
9262                 struct ixgbe_fwd_adapter *vadapter = macvlan_accel_priv(upper);
9263                 struct upper_walk_data *data = _data;
9264                 struct ixgbe_adapter *adapter = data->adapter;
9265                 int ifindex = data->ifindex;
9266 
9267                 if (vadapter && upper->ifindex == ifindex) {
9268                         data->queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
9269                         data->action = data->queue;
9270                         return 1;
9271                 }
9272         }
9273 
9274         return 0;
9275 }
9276 
9277 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
9278                                   u8 *queue, u64 *action)
9279 {
9280         struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
9281         unsigned int num_vfs = adapter->num_vfs, vf;
9282         struct upper_walk_data data;
9283         struct net_device *upper;
9284 
9285         /* redirect to a SRIOV VF */
9286         for (vf = 0; vf < num_vfs; ++vf) {
9287                 upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
9288                 if (upper->ifindex == ifindex) {
9289                         *queue = vf * __ALIGN_MASK(1, ~vmdq->mask);
9290                         *action = vf + 1;
9291                         *action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
9292                         return 0;
9293                 }
9294         }
9295 
9296         /* redirect to a offloaded macvlan netdev */
9297         data.adapter = adapter;
9298         data.ifindex = ifindex;
9299         data.action = 0;
9300         data.queue = 0;
9301         if (netdev_walk_all_upper_dev_rcu(adapter->netdev,
9302                                           get_macvlan_queue, &data)) {
9303                 *action = data.action;
9304                 *queue = data.queue;
9305 
9306                 return 0;
9307         }
9308 
9309         return -EINVAL;
9310 }
9311 
9312 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9313                             struct tcf_exts *exts, u64 *action, u8 *queue)
9314 {
9315         const struct tc_action *a;
9316         int i;
9317 
9318         if (!tcf_exts_has_actions(exts))
9319                 return -EINVAL;
9320 
9321         tcf_exts_for_each_action(i, a, exts) {
9322                 /* Drop action */
9323                 if (is_tcf_gact_shot(a)) {
9324                         *action = IXGBE_FDIR_DROP_QUEUE;
9325                         *queue = IXGBE_FDIR_DROP_QUEUE;
9326                         return 0;
9327                 }
9328 
9329                 /* Redirect to a VF or a offloaded macvlan */
9330                 if (is_tcf_mirred_egress_redirect(a)) {
9331                         struct net_device *dev = tcf_mirred_dev(a);
9332 
9333                         if (!dev)
9334                                 return -EINVAL;
9335                         return handle_redirect_action(adapter, dev->ifindex,
9336                                                       queue, action);
9337                 }
9338 
9339                 return -EINVAL;
9340         }
9341 
9342         return -EINVAL;
9343 }
9344 #else
9345 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9346                             struct tcf_exts *exts, u64 *action, u8 *queue)
9347 {
9348         return -EINVAL;
9349 }
9350 #endif /* CONFIG_NET_CLS_ACT */
9351 
9352 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
9353                                     union ixgbe_atr_input *mask,
9354                                     struct tc_cls_u32_offload *cls,
9355                                     struct ixgbe_mat_field *field_ptr,
9356                                     struct ixgbe_nexthdr *nexthdr)
9357 {
9358         int i, j, off;
9359         __be32 val, m;
9360         bool found_entry = false, found_jump_field = false;
9361 
9362         for (i = 0; i < cls->knode.sel->nkeys; i++) {
9363                 off = cls->knode.sel->keys[i].off;
9364                 val = cls->knode.sel->keys[i].val;
9365                 m = cls->knode.sel->keys[i].mask;
9366 
9367                 for (j = 0; field_ptr[j].val; j++) {
9368                         if (field_ptr[j].off == off) {
9369                                 field_ptr[j].val(input, mask, (__force u32)val,
9370                                                  (__force u32)m);
9371                                 input->filter.formatted.flow_type |=
9372                                         field_ptr[j].type;
9373                                 found_entry = true;
9374                                 break;
9375                         }
9376                 }
9377                 if (nexthdr) {
9378                         if (nexthdr->off == cls->knode.sel->keys[i].off &&
9379                             nexthdr->val ==
9380                             (__force u32)cls->knode.sel->keys[i].val &&
9381                             nexthdr->mask ==
9382                             (__force u32)cls->knode.sel->keys[i].mask)
9383                                 found_jump_field = true;
9384                         else
9385                                 continue;
9386                 }
9387         }
9388 
9389         if (nexthdr && !found_jump_field)
9390                 return -EINVAL;
9391 
9392         if (!found_entry)
9393                 return 0;
9394 
9395         mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
9396                                     IXGBE_ATR_L4TYPE_MASK;
9397 
9398         if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
9399                 mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
9400 
9401         return 0;
9402 }
9403 
9404 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
9405                                   struct tc_cls_u32_offload *cls)
9406 {
9407         __be16 protocol = cls->common.protocol;
9408         u32 loc = cls->knode.handle & 0xfffff;
9409         struct ixgbe_hw *hw = &adapter->hw;
9410         struct ixgbe_mat_field *field_ptr;
9411         struct ixgbe_fdir_filter *input = NULL;
9412         union ixgbe_atr_input *mask = NULL;
9413         struct ixgbe_jump_table *jump = NULL;
9414         int i, err = -EINVAL;
9415         u8 queue;
9416         u32 uhtid, link_uhtid;
9417 
9418         uhtid = TC_U32_USERHTID(cls->knode.handle);
9419         link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
9420 
9421         /* At the moment cls_u32 jumps to network layer and skips past
9422          * L2 headers. The canonical method to match L2 frames is to use
9423          * negative values. However this is error prone at best but really
9424          * just broken because there is no way to "know" what sort of hdr
9425          * is in front of the network layer. Fix cls_u32 to support L2
9426          * headers when needed.
9427          */
9428         if (protocol != htons(ETH_P_IP))
9429                 return err;
9430 
9431         if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
9432                 e_err(drv, "Location out of range\n");
9433                 return err;
9434         }
9435 
9436         /* cls u32 is a graph starting at root node 0x800. The driver tracks
9437          * links and also the fields used to advance the parser across each
9438          * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
9439          * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
9440          * To add support for new nodes update ixgbe_model.h parse structures
9441          * this function _should_ be generic try not to hardcode values here.
9442          */
9443         if (uhtid == 0x800) {
9444                 field_ptr = (adapter->jump_tables[0])->mat;
9445         } else {
9446                 if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9447                         return err;
9448                 if (!adapter->jump_tables[uhtid])
9449                         return err;
9450                 field_ptr = (adapter->jump_tables[uhtid])->mat;
9451         }
9452 
9453         if (!field_ptr)
9454                 return err;
9455 
9456         /* At this point we know the field_ptr is valid and need to either
9457          * build cls_u32 link or attach filter. Because adding a link to
9458          * a handle that does not exist is invalid and the same for adding
9459          * rules to handles that don't exist.
9460          */
9461 
9462         if (link_uhtid) {
9463                 struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
9464 
9465                 if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
9466                         return err;
9467 
9468                 if (!test_bit(link_uhtid - 1, &adapter->tables))
9469                         return err;
9470 
9471                 /* Multiple filters as links to the same hash table are not
9472                  * supported. To add a new filter with the same next header
9473                  * but different match/jump conditions, create a new hash table
9474                  * and link to it.
9475                  */
9476                 if (adapter->jump_tables[link_uhtid] &&
9477                     (adapter->jump_tables[link_uhtid])->link_hdl) {
9478                         e_err(drv, "Link filter exists for link: %x\n",
9479                               link_uhtid);
9480                         return err;
9481                 }
9482 
9483                 for (i = 0; nexthdr[i].jump; i++) {
9484                         if (nexthdr[i].o != cls->knode.sel->offoff ||
9485                             nexthdr[i].s != cls->knode.sel->offshift ||
9486                             nexthdr[i].m !=
9487                             (__force u32)cls->knode.sel->offmask)
9488                                 return err;
9489 
9490                         jump = kzalloc(sizeof(*jump), GFP_KERNEL);
9491                         if (!jump)
9492                                 return -ENOMEM;
9493                         input = kzalloc(sizeof(*input), GFP_KERNEL);
9494                         if (!input) {
9495                                 err = -ENOMEM;
9496                                 goto free_jump;
9497                         }
9498                         mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9499                         if (!mask) {
9500                                 err = -ENOMEM;
9501                                 goto free_input;
9502                         }
9503                         jump->input = input;
9504                         jump->mask = mask;
9505                         jump->link_hdl = cls->knode.handle;
9506 
9507                         err = ixgbe_clsu32_build_input(input, mask, cls,
9508                                                        field_ptr, &nexthdr[i]);
9509                         if (!err) {
9510                                 jump->mat = nexthdr[i].jump;
9511                                 adapter->jump_tables[link_uhtid] = jump;
9512                                 break;
9513                         } else {
9514                                 kfree(mask);
9515                                 kfree(input);
9516                                 kfree(jump);
9517                         }
9518                 }
9519                 return 0;
9520         }
9521 
9522         input = kzalloc(sizeof(*input), GFP_KERNEL);
9523         if (!input)
9524                 return -ENOMEM;
9525         mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9526         if (!mask) {
9527                 err = -ENOMEM;
9528                 goto free_input;
9529         }
9530 
9531         if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
9532                 if ((adapter->jump_tables[uhtid])->input)
9533                         memcpy(input, (adapter->jump_tables[uhtid])->input,
9534                                sizeof(*input));
9535                 if ((adapter->jump_tables[uhtid])->mask)
9536                         memcpy(mask, (adapter->jump_tables[uhtid])->mask,
9537                                sizeof(*mask));
9538 
9539                 /* Lookup in all child hash tables if this location is already
9540                  * filled with a filter
9541                  */
9542                 for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9543                         struct ixgbe_jump_table *link = adapter->jump_tables[i];
9544 
9545                         if (link && (test_bit(loc - 1, link->child_loc_map))) {
9546                                 e_err(drv, "Filter exists in location: %x\n",
9547                                       loc);
9548                                 err = -EINVAL;
9549                                 goto err_out;
9550                         }
9551                 }
9552         }
9553         err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
9554         if (err)
9555                 goto err_out;
9556 
9557         err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
9558                                &queue);
9559         if (err < 0)
9560                 goto err_out;
9561 
9562         input->sw_idx = loc;
9563 
9564         spin_lock(&adapter->fdir_perfect_lock);
9565 
9566         if (hlist_empty(&adapter->fdir_filter_list)) {
9567                 memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
9568                 err = ixgbe_fdir_set_input_mask_82599(hw, mask);
9569                 if (err)
9570                         goto err_out_w_lock;
9571         } else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
9572                 err = -EINVAL;
9573                 goto err_out_w_lock;
9574         }
9575 
9576         ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
9577         err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
9578                                                     input->sw_idx, queue);
9579         if (!err)
9580                 ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
9581         spin_unlock(&adapter->fdir_perfect_lock);
9582 
9583         if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
9584                 set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
9585 
9586         kfree(mask);
9587         return err;
9588 err_out_w_lock:
9589         spin_unlock(&adapter->fdir_perfect_lock);
9590 err_out:
9591         kfree(mask);
9592 free_input:
9593         kfree(input);
9594 free_jump:
9595         kfree(jump);
9596         return err;
9597 }
9598 
9599 static int ixgbe_setup_tc_cls_u32(struct ixgbe_adapter *adapter,
9600                                   struct tc_cls_u32_offload *cls_u32)
9601 {
9602         switch (cls_u32->command) {
9603         case TC_CLSU32_NEW_KNODE:
9604         case TC_CLSU32_REPLACE_KNODE:
9605                 return ixgbe_configure_clsu32(adapter, cls_u32);
9606         case TC_CLSU32_DELETE_KNODE:
9607                 return ixgbe_delete_clsu32(adapter, cls_u32);
9608         case TC_CLSU32_NEW_HNODE:
9609         case TC_CLSU32_REPLACE_HNODE:
9610                 return ixgbe_configure_clsu32_add_hnode(adapter, cls_u32);
9611         case TC_CLSU32_DELETE_HNODE:
9612                 return ixgbe_configure_clsu32_del_hnode(adapter, cls_u32);
9613         default:
9614                 return -EOPNOTSUPP;
9615         }
9616 }
9617 
9618 static int ixgbe_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
9619                                    void *cb_priv)
9620 {
9621         struct ixgbe_adapter *adapter = cb_priv;
9622 
9623         if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
9624                 return -EOPNOTSUPP;
9625 
9626         switch (type) {
9627         case TC_SETUP_CLSU32:
9628                 return ixgbe_setup_tc_cls_u32(adapter, type_data);
9629         default:
9630                 return -EOPNOTSUPP;
9631         }
9632 }
9633 
9634 static int ixgbe_setup_tc_mqprio(struct net_device *dev,
9635                                  struct tc_mqprio_qopt *mqprio)
9636 {
9637         mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
9638         return ixgbe_setup_tc(dev, mqprio->num_tc);
9639 }
9640 
9641 static LIST_HEAD(ixgbe_block_cb_list);
9642 
9643 static int __ixgbe_setup_tc(struct net_device *dev, enum tc_setup_type type,
9644                             void *type_data)
9645 {
9646         struct ixgbe_adapter *adapter = netdev_priv(dev);
9647 
9648         switch (type) {
9649         case TC_SETUP_BLOCK:
9650                 return flow_block_cb_setup_simple(type_data,
9651                                                   &ixgbe_block_cb_list,
9652                                                   ixgbe_setup_tc_block_cb,
9653                                                   adapter, adapter, true);
9654         case TC_SETUP_QDISC_MQPRIO:
9655                 return ixgbe_setup_tc_mqprio(dev, type_data);
9656         default:
9657                 return -EOPNOTSUPP;
9658         }
9659 }
9660 
9661 #ifdef CONFIG_PCI_IOV
9662 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
9663 {
9664         struct net_device *netdev = adapter->netdev;
9665 
9666         rtnl_lock();
9667         ixgbe_setup_tc(netdev, adapter->hw_tcs);
9668         rtnl_unlock();
9669 }
9670 
9671 #endif
9672 void ixgbe_do_reset(struct net_device *netdev)
9673 {
9674         struct ixgbe_adapter *adapter = netdev_priv(netdev);
9675 
9676         if (netif_running(netdev))
9677                 ixgbe_reinit_locked(adapter);
9678         else
9679                 ixgbe_reset(adapter);
9680 }
9681 
9682 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
9683                                             netdev_features_t features)
9684 {
9685         struct ixgbe_adapter *adapter = netdev_priv(netdev);
9686 
9687         /* If Rx checksum is disabled, then RSC/LRO should also be disabled */
9688         if (!(features & NETIF_F_RXCSUM))
9689                 features &= ~NETIF_F_LRO;
9690 
9691         /* Turn off LRO if not RSC capable */
9692         if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
9693                 features &= ~NETIF_F_LRO;
9694 
9695         if (adapter->xdp_prog && (features & NETIF_F_LRO)) {
9696                 e_dev_err("LRO is not supported with XDP\n");
9697                 features &= ~NETIF_F_LRO;
9698         }
9699 
9700         return features;
9701 }
9702 
9703 static void ixgbe_reset_l2fw_offload(struct ixgbe_adapter *adapter)
9704 {
9705         int rss = min_t(int, ixgbe_max_rss_indices(adapter),
9706                         num_online_cpus());
9707 
9708         /* go back to full RSS if we're not running SR-IOV */
9709         if (!adapter->ring_feature[RING_F_VMDQ].offset)
9710                 adapter->flags &= ~(IXGBE_FLAG_VMDQ_ENABLED |
9711                                     IXGBE_FLAG_SRIOV_ENABLED);
9712 
9713         adapter->ring_feature[RING_F_RSS].limit = rss;
9714         adapter->ring_feature[RING_F_VMDQ].limit = 1;
9715 
9716         ixgbe_setup_tc(adapter->netdev, adapter->hw_tcs);
9717 }
9718 
9719 static int ixgbe_set_features(struct net_device *netdev,
9720                               netdev_features_t features)
9721 {
9722         struct ixgbe_adapter *adapter = netdev_priv(netdev);
9723         netdev_features_t changed = netdev->features ^ features;
9724         bool need_reset = false;
9725 
9726         /* Make sure RSC matches LRO, reset if change */
9727         if (!(features & NETIF_F_LRO)) {
9728                 if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9729                         need_reset = true;
9730                 adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
9731         } else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
9732                    !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
9733                 if (adapter->rx_itr_setting == 1 ||
9734                     adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
9735                         adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
9736                         need_reset = true;
9737                 } else if ((changed ^ features) & NETIF_F_LRO) {
9738                         e_info(probe, "rx-usecs set too low, "
9739                                "disabling RSC\n");
9740                 }
9741         }
9742 
9743         /*
9744          * Check if Flow Director n-tuple support or hw_tc support was
9745          * enabled or disabled.  If the state changed, we need to reset.
9746          */
9747         if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
9748                 /* turn off ATR, enable perfect filters and reset */
9749                 if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
9750                         need_reset = true;
9751 
9752                 adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
9753                 adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9754         } else {
9755                 /* turn off perfect filters, enable ATR and reset */
9756                 if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
9757                         need_reset = true;
9758 
9759                 adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9760 
9761                 /* We cannot enable ATR if SR-IOV is enabled */
9762                 if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
9763                     /* We cannot enable ATR if we have 2 or more tcs */
9764                     (adapter->hw_tcs > 1) ||
9765                     /* We cannot enable ATR if RSS is disabled */
9766                     (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
9767                     /* A sample rate of 0 indicates ATR disabled */
9768                     (!adapter->atr_sample_rate))
9769                         ; /* do nothing not supported */
9770                 else /* otherwise supported and set the flag */
9771                         adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
9772         }
9773 
9774         if (changed & NETIF_F_RXALL)
9775                 need_reset = true;
9776 
9777         netdev->features = features;
9778 
9779         if ((adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE)) {
9780                 if (features & NETIF_F_RXCSUM) {
9781                         adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9782                 } else {
9783                         u32 port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9784 
9785                         ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9786                 }
9787         }
9788 
9789         if ((adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE)) {
9790                 if (features & NETIF_F_RXCSUM) {
9791                         adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9792                 } else {
9793                         u32 port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9794 
9795                         ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9796                 }
9797         }
9798 
9799         if ((changed & NETIF_F_HW_L2FW_DOFFLOAD) && adapter->num_rx_pools > 1)
9800                 ixgbe_reset_l2fw_offload(adapter);
9801         else if (need_reset)
9802                 ixgbe_do_reset(netdev);
9803         else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
9804                             NETIF_F_HW_VLAN_CTAG_FILTER))
9805                 ixgbe_set_rx_mode(netdev);
9806 
9807         return 1;
9808 }
9809 
9810 /**
9811  * ixgbe_add_udp_tunnel_port - Get notifications about adding UDP tunnel ports
9812  * @dev: The port's netdev
9813  * @ti: Tunnel endpoint information
9814  **/
9815 static void ixgbe_add_udp_tunnel_port(struct net_device *dev,
9816                                       struct udp_tunnel_info *ti)
9817 {
9818         struct ixgbe_adapter *adapter = netdev_priv(dev);
9819         struct ixgbe_hw *hw = &adapter->hw;
9820         __be16 port = ti->port;
9821         u32 port_shift = 0;
9822         u32 reg;
9823 
9824         if (ti->sa_family != AF_INET)
9825                 return;
9826 
9827         switch (ti->type) {
9828         case UDP_TUNNEL_TYPE_VXLAN:
9829                 if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9830                         return;
9831 
9832                 if (adapter->vxlan_port == port)
9833                         return;
9834 
9835                 if (adapter->vxlan_port) {
9836                         netdev_info(dev,
9837                                     "VXLAN port %d set, not adding port %d\n",
9838                                     ntohs(adapter->vxlan_port),
9839                                     ntohs(port));
9840                         return;
9841                 }
9842 
9843                 adapter->vxlan_port = port;
9844                 break;
9845         case UDP_TUNNEL_TYPE_GENEVE:
9846                 if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9847                         return;
9848 
9849                 if (adapter->geneve_port == port)
9850                         return;
9851 
9852                 if (adapter->geneve_port) {
9853                         netdev_info(dev,
9854                                     "GENEVE port %d set, not adding port %d\n",
9855                                     ntohs(adapter->geneve_port),
9856                                     ntohs(port));
9857                         return;
9858                 }
9859 
9860                 port_shift = IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT;
9861                 adapter->geneve_port = port;
9862                 break;
9863         default:
9864                 return;
9865         }
9866 
9867         reg = IXGBE_READ_REG(hw, IXGBE_VXLANCTRL) | ntohs(port) << port_shift;
9868         IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, reg);
9869 }
9870 
9871 /**
9872  * ixgbe_del_udp_tunnel_port - Get notifications about removing UDP tunnel ports
9873  * @dev: The port's netdev
9874  * @ti: Tunnel endpoint information
9875  **/
9876 static void ixgbe_del_udp_tunnel_port(struct net_device *dev,
9877                                       struct udp_tunnel_info *ti)
9878 {
9879         struct ixgbe_adapter *adapter = netdev_priv(dev);
9880         u32 port_mask;
9881 
9882         if (ti->type != UDP_TUNNEL_TYPE_VXLAN &&
9883             ti->type != UDP_TUNNEL_TYPE_GENEVE)
9884                 return;
9885 
9886         if (ti->sa_family != AF_INET)
9887                 return;
9888 
9889         switch (ti->type) {
9890         case UDP_TUNNEL_TYPE_VXLAN:
9891                 if (!(adapter->flags & IXGBE_FLAG_VXLAN_OFFLOAD_CAPABLE))
9892                         return;
9893 
9894                 if (adapter->vxlan_port != ti->port) {
9895                         netdev_info(dev, "VXLAN port %d not found\n",
9896                                     ntohs(ti->port));
9897                         return;
9898                 }
9899 
9900                 port_mask = IXGBE_VXLANCTRL_VXLAN_UDPPORT_MASK;
9901                 break;
9902         case UDP_TUNNEL_TYPE_GENEVE:
9903                 if (!(adapter->flags & IXGBE_FLAG_GENEVE_OFFLOAD_CAPABLE))
9904                         return;
9905 
9906                 if (adapter->geneve_port != ti->port) {
9907                         netdev_info(dev, "GENEVE port %d not found\n",
9908                                     ntohs(ti->port));
9909                         return;
9910                 }
9911 
9912                 port_mask = IXGBE_VXLANCTRL_GENEVE_UDPPORT_MASK;
9913                 break;
9914         default:
9915                 return;
9916         }
9917 
9918         ixgbe_clear_udp_tunnel_port(adapter, port_mask);
9919         adapter->flags2 |= IXGBE_FLAG2_UDP_TUN_REREG_NEEDED;
9920 }
9921 
9922 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9923                              struct net_device *dev,
9924                              const unsigned char *addr, u16 vid,
9925                              u16 flags,
9926                              struct netlink_ext_ack *extack)
9927 {
9928         /* guarantee we can provide a unique filter for the unicast address */
9929         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
9930                 struct ixgbe_adapter *adapter = netdev_priv(dev);
9931                 u16 pool = VMDQ_P(0);
9932 
9933                 if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
9934                         return -ENOMEM;
9935         }
9936 
9937         return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
9938 }
9939 
9940 /**
9941  * ixgbe_configure_bridge_mode - set various bridge modes
9942  * @adapter: the private structure
9943  * @mode: requested bridge mode
9944  *
9945  * Configure some settings require for various bridge modes.
9946  **/
9947 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
9948                                        __u16 mode)
9949 {
9950         struct ixgbe_hw *hw = &adapter->hw;
9951         unsigned int p, num_pools;
9952         u32 vmdctl;
9953 
9954         switch (mode) {
9955         case BRIDGE_MODE_VEPA:
9956                 /* disable Tx loopback, rely on switch hairpin mode */
9957                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
9958 
9959                 /* must enable Rx switching replication to allow multicast
9960                  * packet reception on all VFs, and to enable source address
9961                  * pruning.
9962                  */
9963                 vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9964                 vmdctl |= IXGBE_VT_CTL_REPLEN;
9965                 IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9966 
9967                 /* enable Rx source address pruning. Note, this requires
9968                  * replication to be enabled or else it does nothing.
9969                  */
9970                 num_pools = adapter->num_vfs + adapter->num_rx_pools;
9971                 for (p = 0; p < num_pools; p++) {
9972                         if (hw->mac.ops.set_source_address_pruning)
9973                                 hw->mac.ops.set_source_address_pruning(hw,
9974                                                                        true,
9975                                                                        p);
9976                 }
9977                 break;
9978         case BRIDGE_MODE_VEB:
9979                 /* enable Tx loopback for internal VF/PF communication */
9980                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
9981                                 IXGBE_PFDTXGSWC_VT_LBEN);
9982 
9983                 /* disable Rx switching replication unless we have SR-IOV
9984                  * virtual functions
9985                  */
9986                 vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9987                 if (!adapter->num_vfs)
9988                         vmdctl &= ~IXGBE_VT_CTL_REPLEN;
9989                 IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9990 
9991                 /* disable Rx source address pruning, since we don't expect to
9992                  * be receiving external loopback of our transmitted frames.
9993                  */
9994                 num_pools = adapter->num_vfs + adapter->num_rx_pools;
9995                 for (p = 0; p < num_pools; p++) {
9996                         if (hw->mac.ops.set_source_address_pruning)
9997                                 hw->mac.ops.set_source_address_pruning(hw,
9998                                                                        false,
9999                                                                        p);
10000                 }
10001                 break;
10002         default:
10003                 return -EINVAL;
10004         }
10005 
10006         adapter->bridge_mode = mode;
10007 
10008         e_info(drv, "enabling bridge mode: %s\n",
10009                mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
10010 
10011         return 0;
10012 }
10013 
10014 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
10015                                     struct nlmsghdr *nlh, u16 flags,
10016                                     struct netlink_ext_ack *extack)
10017 {
10018         struct ixgbe_adapter *adapter = netdev_priv(dev);
10019         struct nlattr *attr, *br_spec;
10020         int rem;
10021 
10022         if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
10023                 return -EOPNOTSUPP;
10024 
10025         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
10026         if (!br_spec)
10027                 return -EINVAL;
10028 
10029         nla_for_each_nested(attr, br_spec, rem) {
10030                 int status;
10031                 __u16 mode;
10032 
10033                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
10034                         continue;
10035 
10036                 if (nla_len(attr) < sizeof(mode))
10037                         return -EINVAL;
10038 
10039                 mode = nla_get_u16(attr);
10040                 status = ixgbe_configure_bridge_mode(adapter, mode);
10041                 if (status)
10042                         return status;
10043 
10044                 break;
10045         }
10046 
10047         return 0;
10048 }
10049 
10050 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
10051                                     struct net_device *dev,
10052                                     u32 filter_mask, int nlflags)
10053 {
10054         struct ixgbe_adapter *adapter = netdev_priv(dev);
10055 
10056         if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
10057                 return 0;
10058 
10059         return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
10060                                        adapter->bridge_mode, 0, 0, nlflags,
10061                                        filter_mask, NULL);
10062 }
10063 
10064 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
10065 {
10066         struct ixgbe_adapter *adapter = netdev_priv(pdev);
10067         struct ixgbe_fwd_adapter *accel;
10068         int tcs = adapter->hw_tcs ? : 1;
10069         int pool, err;
10070 
10071         if (adapter->xdp_prog) {
10072                 e_warn(probe, "L2FW offload is not supported with XDP\n");
10073                 return ERR_PTR(-EINVAL);
10074         }
10075 
10076         /* The hardware supported by ixgbe only filters on the destination MAC
10077          * address. In order to avoid issues we only support offloading modes
10078          * where the hardware can actually provide the functionality.
10079          */
10080         if (!macvlan_supports_dest_filter(vdev))
10081                 return ERR_PTR(-EMEDIUMTYPE);
10082 
10083         /* We need to lock down the macvlan to be a single queue device so that
10084          * we can reuse the tc_to_txq field in the macvlan netdev to represent
10085          * the queue mapping to our netdev.
10086          */
10087         if (netif_is_multiqueue(vdev))
10088                 return ERR_PTR(-ERANGE);
10089 
10090         pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
10091         if (pool == adapter->num_rx_pools) {
10092                 u16 used_pools = adapter->num_vfs + adapter->num_rx_pools;
10093                 u16 reserved_pools;
10094 
10095                 if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
10096                      adapter->num_rx_pools >= (MAX_TX_QUEUES / tcs)) ||
10097                     adapter->num_rx_pools > IXGBE_MAX_MACVLANS)
10098                         return ERR_PTR(-EBUSY);
10099 
10100                 /* Hardware has a limited number of available pools. Each VF,
10101                  * and the PF require a pool. Check to ensure we don't
10102                  * attempt to use more then the available number of pools.
10103                  */
10104                 if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
10105                         return ERR_PTR(-EBUSY);
10106 
10107                 /* Enable VMDq flag so device will be set in VM mode */
10108                 adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED |
10109                                   IXGBE_FLAG_SRIOV_ENABLED;
10110 
10111                 /* Try to reserve as many queues per pool as possible,
10112                  * we start with the configurations that support 4 queues
10113                  * per pools, followed by 2, and then by just 1 per pool.
10114                  */
10115                 if (used_pools < 32 && adapter->num_rx_pools < 16)
10116                         reserved_pools = min_t(u16,
10117                                                32 - used_pools,
10118                                                16 - adapter->num_rx_pools);
10119                 else if (adapter->num_rx_pools < 32)
10120                         reserved_pools = min_t(u16,
10121                                                64 - used_pools,
10122                                                32 - adapter->num_rx_pools);
10123                 else
10124                         reserved_pools = 64 - used_pools;
10125 
10126 
10127                 if (!reserved_pools)
10128                         return ERR_PTR(-EBUSY);
10129 
10130                 adapter->ring_feature[RING_F_VMDQ].limit += reserved_pools;
10131 
10132                 /* Force reinit of ring allocation with VMDQ enabled */
10133                 err = ixgbe_setup_tc(pdev, adapter->hw_tcs);
10134                 if (err)
10135                         return ERR_PTR(err);
10136 
10137                 if (pool >= adapter->num_rx_pools)
10138                         return ERR_PTR(-ENOMEM);
10139         }
10140 
10141         accel = kzalloc(sizeof(*accel), GFP_KERNEL);
10142         if (!accel)
10143                 return ERR_PTR(-ENOMEM);
10144 
10145         set_bit(pool, adapter->fwd_bitmask);
10146         netdev_set_sb_channel(vdev, pool);
10147         accel->pool = pool;
10148         accel->netdev = vdev;
10149 
10150         if (!netif_running(pdev))
10151                 return accel;
10152 
10153         err = ixgbe_fwd_ring_up(adapter, accel);
10154         if (err)
10155                 return ERR_PTR(err);
10156 
10157         return accel;
10158 }
10159 
10160 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
10161 {
10162         struct ixgbe_fwd_adapter *accel = priv;
10163         struct ixgbe_adapter *adapter = netdev_priv(pdev);
10164         unsigned int rxbase = accel->rx_base_queue;
10165         unsigned int i;
10166 
10167         /* delete unicast filter associated with offloaded interface */
10168         ixgbe_del_mac_filter(adapter, accel->netdev->dev_addr,
10169                              VMDQ_P(accel->pool));
10170 
10171         /* Allow remaining Rx packets to get flushed out of the
10172          * Rx FIFO before we drop the netdev for the ring.
10173          */
10174         usleep_range(10000, 20000);
10175 
10176         for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
10177                 struct ixgbe_ring *ring = adapter->rx_ring[rxbase + i];
10178                 struct ixgbe_q_vector *qv = ring->q_vector;
10179 
10180                 /* Make sure we aren't processing any packets and clear
10181                  * netdev to shut down the ring.
10182                  */
10183                 if (netif_running(adapter->netdev))
10184                         napi_synchronize(&qv->napi);
10185                 ring->netdev = NULL;
10186         }
10187 
10188         /* unbind the queues and drop the subordinate channel config */
10189         netdev_unbind_sb_channel(pdev, accel->netdev);
10190         netdev_set_sb_channel(accel->netdev, 0);
10191 
10192         clear_bit(accel->pool, adapter->fwd_bitmask);
10193         kfree(accel);
10194 }
10195 
10196 #define IXGBE_MAX_MAC_HDR_LEN           127
10197 #define IXGBE_MAX_NETWORK_HDR_LEN       511
10198 
10199 static netdev_features_t
10200 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
10201                      netdev_features_t features)
10202 {
10203         unsigned int network_hdr_len, mac_hdr_len;
10204 
10205         /* Make certain the headers can be described by a context descriptor */
10206         mac_hdr_len = skb_network_header(skb) - skb->data;
10207         if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
10208                 return features & ~(NETIF_F_HW_CSUM |
10209                                     NETIF_F_SCTP_CRC |
10210                                     NETIF_F_HW_VLAN_CTAG_TX |
10211                                     NETIF_F_TSO |
10212                                     NETIF_F_TSO6);
10213 
10214         network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
10215         if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
10216                 return features & ~(NETIF_F_HW_CSUM |
10217                                     NETIF_F_SCTP_CRC |
10218                                     NETIF_F_TSO |
10219                                     NETIF_F_TSO6);
10220 
10221         /* We can only support IPV4 TSO in tunnels if we can mangle the
10222          * inner IP ID field, so strip TSO if MANGLEID is not supported.
10223          * IPsec offoad sets skb->encapsulation but still can handle
10224          * the TSO, so it's the exception.
10225          */
10226         if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID)) {
10227 #ifdef CONFIG_IXGBE_IPSEC
10228                 if (!secpath_exists(skb))
10229 #endif
10230                         features &= ~NETIF_F_TSO;
10231         }
10232 
10233         return features;
10234 }
10235 
10236 static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
10237 {
10238         int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
10239         struct ixgbe_adapter *adapter = netdev_priv(dev);
10240         struct bpf_prog *old_prog;
10241         bool need_reset;
10242 
10243         if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
10244                 return -EINVAL;
10245 
10246         if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
10247                 return -EINVAL;
10248 
10249         /* verify ixgbe ring attributes are sufficient for XDP */
10250         for (i = 0; i < adapter->num_rx_queues; i++) {
10251                 struct ixgbe_ring *ring = adapter->rx_ring[i];
10252 
10253                 if (ring_is_rsc_enabled(ring))
10254                         return -EINVAL;
10255 
10256                 if (frame_size > ixgbe_rx_bufsz(ring))
10257                         return -EINVAL;
10258         }
10259 
10260         if (nr_cpu_ids > MAX_XDP_QUEUES)
10261                 return -ENOMEM;
10262 
10263         old_prog = xchg(&adapter->xdp_prog, prog);
10264         need_reset = (!!prog != !!old_prog);
10265 
10266         /* If transitioning XDP modes reconfigure rings */
10267         if (need_reset) {
10268                 int err;
10269 
10270                 if (!prog)
10271                         /* Wait until ndo_xsk_wakeup completes. */
10272                         synchronize_rcu();
10273                 err = ixgbe_setup_tc(dev, adapter->hw_tcs);
10274 
10275                 if (err) {
10276                         rcu_assign_pointer(adapter->xdp_prog, old_prog);
10277                         return -EINVAL;
10278                 }
10279         } else {
10280                 for (i = 0; i < adapter->num_rx_queues; i++)
10281                         (void)xchg(&adapter->rx_ring[i]->xdp_prog,
10282                             adapter->xdp_prog);
10283         }
10284 
10285         if (old_prog)
10286                 bpf_prog_put(old_prog);
10287 
10288         /* Kick start the NAPI context if there is an AF_XDP socket open
10289          * on that queue id. This so that receiving will start.
10290          */
10291         if (need_reset && prog)
10292                 for (i = 0; i < adapter->num_rx_queues; i++)
10293                         if (adapter->xdp_ring[i]->xsk_umem)
10294                                 (void)ixgbe_xsk_wakeup(adapter->netdev, i,
10295                                                        XDP_WAKEUP_RX);
10296 
10297         return 0;
10298 }
10299 
10300 static int ixgbe_xdp(struct net_device *dev, struct netdev_bpf *xdp)
10301 {
10302         struct ixgbe_adapter *adapter = netdev_priv(dev);
10303 
10304         switch (xdp->command) {
10305         case XDP_SETUP_PROG:
10306                 return ixgbe_xdp_setup(dev, xdp->prog);
10307         case XDP_QUERY_PROG:
10308                 xdp->prog_id = adapter->xdp_prog ?
10309                         adapter->xdp_prog->aux->id : 0;
10310                 return 0;
10311         case XDP_SETUP_XSK_UMEM:
10312                 return ixgbe_xsk_umem_setup(adapter, xdp->xsk.umem,
10313                                             xdp->xsk.queue_id);
10314 
10315         default:
10316                 return -EINVAL;
10317         }
10318 }
10319 
10320 void ixgbe_xdp_ring_update_tail(struct ixgbe_ring *ring)
10321 {
10322         /* Force memory writes to complete before letting h/w know there
10323          * are new descriptors to fetch.
10324          */
10325         wmb();
10326         writel(ring->next_to_use, ring->tail);
10327 }
10328 
10329 static int ixgbe_xdp_xmit(struct net_device *dev, int n,
10330                           struct xdp_frame **frames, u32 flags)
10331 {
10332         struct ixgbe_adapter *adapter = netdev_priv(dev);
10333         struct ixgbe_ring *ring;
10334         int drops = 0;
10335         int i;
10336 
10337         if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
10338                 return -ENETDOWN;
10339 
10340         if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
10341                 return -EINVAL;
10342 
10343         /* During program transitions its possible adapter->xdp_prog is assigned
10344          * but ring has not been configured yet. In this case simply abort xmit.
10345          */
10346         ring = adapter->xdp_prog ? adapter->xdp_ring[smp_processor_id()] : NULL;
10347         if (unlikely(!ring))
10348                 return -ENXIO;
10349 
10350         if (unlikely(test_bit(__IXGBE_TX_DISABLED, &ring->state)))
10351                 return -ENXIO;
10352 
10353         for (i = 0; i < n; i++) {
10354                 struct xdp_frame *xdpf = frames[i];
10355                 int err;
10356 
10357                 err = ixgbe_xmit_xdp_ring(adapter, xdpf);
10358                 if (err != IXGBE_XDP_TX) {
10359                         xdp_return_frame_rx_napi(xdpf);
10360                         drops++;
10361                 }
10362         }
10363 
10364         if (unlikely(flags & XDP_XMIT_FLUSH))
10365                 ixgbe_xdp_ring_update_tail(ring);
10366 
10367         return n - drops;
10368 }
10369 
10370 static const struct net_device_ops ixgbe_netdev_ops = {
10371         .ndo_open               = ixgbe_open,
10372         .ndo_stop               = ixgbe_close,
10373         .ndo_start_xmit         = ixgbe_xmit_frame,
10374         .ndo_set_rx_mode        = ixgbe_set_rx_mode,
10375         .ndo_validate_addr      = eth_validate_addr,
10376         .ndo_set_mac_address    = ixgbe_set_mac,
10377         .ndo_change_mtu         = ixgbe_change_mtu,
10378         .ndo_tx_timeout         = ixgbe_tx_timeout,
10379         .ndo_set_tx_maxrate     = ixgbe_tx_maxrate,
10380         .ndo_vlan_rx_add_vid    = ixgbe_vlan_rx_add_vid,
10381         .ndo_vlan_rx_kill_vid   = ixgbe_vlan_rx_kill_vid,
10382         .ndo_do_ioctl           = ixgbe_ioctl,
10383         .ndo_set_vf_mac         = ixgbe_ndo_set_vf_mac,
10384         .ndo_set_vf_vlan        = ixgbe_ndo_set_vf_vlan,
10385         .ndo_set_vf_rate        = ixgbe_ndo_set_vf_bw,
10386         .ndo_set_vf_spoofchk    = ixgbe_ndo_set_vf_spoofchk,
10387         .ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
10388         .ndo_set_vf_trust       = ixgbe_ndo_set_vf_trust,
10389         .ndo_get_vf_config      = ixgbe_ndo_get_vf_config,
10390         .ndo_get_stats64        = ixgbe_get_stats64,
10391         .ndo_setup_tc           = __ixgbe_setup_tc,
10392 #ifdef IXGBE_FCOE
10393         .ndo_select_queue       = ixgbe_select_queue,
10394         .ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
10395         .ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
10396         .ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
10397         .ndo_fcoe_enable = ixgbe_fcoe_enable,
10398         .ndo_fcoe_disable = ixgbe_fcoe_disable,
10399         .ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
10400         .ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
10401 #endif /* IXGBE_FCOE */
10402         .ndo_set_features = ixgbe_set_features,
10403         .ndo_fix_features = ixgbe_fix_features,
10404         .ndo_fdb_add            = ixgbe_ndo_fdb_add,
10405         .ndo_bridge_setlink     = ixgbe_ndo_bridge_setlink,
10406         .ndo_bridge_getlink     = ixgbe_ndo_bridge_getlink,
10407         .ndo_dfwd_add_station   = ixgbe_fwd_add,
10408         .ndo_dfwd_del_station   = ixgbe_fwd_del,
10409         .ndo_udp_tunnel_add     = ixgbe_add_udp_tunnel_port,
10410         .ndo_udp_tunnel_del     = ixgbe_del_udp_tunnel_port,
10411         .ndo_features_check     = ixgbe_features_check,
10412         .ndo_bpf                = ixgbe_xdp,
10413         .ndo_xdp_xmit           = ixgbe_xdp_xmit,
10414         .ndo_xsk_wakeup         = ixgbe_xsk_wakeup,
10415 };
10416 
10417 static void ixgbe_disable_txr_hw(struct ixgbe_adapter *adapter,
10418                                  struct ixgbe_ring *tx_ring)
10419 {
10420         unsigned long wait_delay, delay_interval;
10421         struct ixgbe_hw *hw = &adapter->hw;
10422         u8 reg_idx = tx_ring->reg_idx;
10423         int wait_loop;
10424         u32 txdctl;
10425 
10426         IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
10427 
10428         /* delay mechanism from ixgbe_disable_tx */
10429         delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
10430 
10431         wait_loop = IXGBE_MAX_RX_DESC_POLL;
10432         wait_delay = delay_interval;
10433 
10434         while (wait_loop--) {
10435                 usleep_range(wait_delay, wait_delay + 10);
10436                 wait_delay += delay_interval * 2;
10437                 txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
10438 
10439                 if (!(txdctl & IXGBE_TXDCTL_ENABLE))
10440                         return;
10441         }
10442 
10443         e_err(drv, "TXDCTL.ENABLE not cleared within the polling period\n");
10444 }
10445 
10446 static void ixgbe_disable_txr(struct ixgbe_adapter *adapter,
10447                               struct ixgbe_ring *tx_ring)
10448 {
10449         set_bit(__IXGBE_TX_DISABLED, &tx_ring->state);
10450         ixgbe_disable_txr_hw(adapter, tx_ring);
10451 }
10452 
10453 static void ixgbe_disable_rxr_hw(struct ixgbe_adapter *adapter,
10454                                  struct ixgbe_ring *rx_ring)
10455 {
10456         unsigned long wait_delay, delay_interval;
10457         struct ixgbe_hw *hw = &adapter->hw;
10458         u8 reg_idx = rx_ring->reg_idx;
10459         int wait_loop;
10460         u32 rxdctl;
10461 
10462         rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
10463         rxdctl &= ~IXGBE_RXDCTL_ENABLE;
10464         rxdctl |= IXGBE_RXDCTL_SWFLSH;
10465 
10466         /* write value back with RXDCTL.ENABLE bit cleared */
10467         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
10468 
10469         /* RXDCTL.EN may not change on 82598 if link is down, so skip it */
10470         if (hw->mac.type == ixgbe_mac_82598EB &&
10471             !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
10472                 return;
10473 
10474         /* delay mechanism from ixgbe_disable_rx */
10475         delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
10476 
10477         wait_loop = IXGBE_MAX_RX_DESC_POLL;
10478         wait_delay = delay_interval;
10479 
10480         while (wait_loop--) {
10481                 usleep_range(wait_delay, wait_delay + 10);
10482                 wait_delay += delay_interval * 2;
10483                 rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
10484 
10485                 if (!(rxdctl & IXGBE_RXDCTL_ENABLE))
10486                         return;
10487         }
10488 
10489         e_err(drv, "RXDCTL.ENABLE not cleared within the polling period\n");
10490 }
10491 
10492 static void ixgbe_reset_txr_stats(struct ixgbe_ring *tx_ring)
10493 {
10494         memset(&tx_ring->stats, 0, sizeof(tx_ring->stats));
10495         memset(&tx_ring->tx_stats, 0, sizeof(tx_ring->tx_stats));
10496 }
10497 
10498 static void ixgbe_reset_rxr_stats(struct ixgbe_ring *rx_ring)
10499 {
10500         memset(&rx_ring->stats, 0, sizeof(rx_ring->stats));
10501         memset(&rx_ring->rx_stats, 0, sizeof(rx_ring->rx_stats));
10502 }
10503 
10504 /**
10505  * ixgbe_txrx_ring_disable - Disable Rx/Tx/XDP Tx rings
10506  * @adapter: adapter structure
10507  * @ring: ring index
10508  *
10509  * This function disables a certain Rx/Tx/XDP Tx ring. The function
10510  * assumes that the netdev is running.
10511  **/
10512 void ixgbe_txrx_ring_disable(struct ixgbe_adapter *adapter, int ring)
10513 {
10514         struct ixgbe_ring *rx_ring, *tx_ring, *xdp_ring;
10515 
10516         rx_ring = adapter->rx_ring[ring];
10517         tx_ring = adapter->tx_ring[ring];
10518         xdp_ring = adapter->xdp_ring[ring];
10519 
10520         ixgbe_disable_txr(adapter, tx_ring);
10521         if (xdp_ring)
10522                 ixgbe_disable_txr(adapter, xdp_ring);
10523         ixgbe_disable_rxr_hw(adapter, rx_ring);
10524 
10525         if (xdp_ring)
10526                 synchronize_rcu();
10527 
10528         /* Rx/Tx/XDP Tx share the same napi context. */
10529         napi_disable(&rx_ring->q_vector->napi);
10530 
10531         ixgbe_clean_tx_ring(tx_ring);
10532         if (xdp_ring)
10533                 ixgbe_clean_tx_ring(xdp_ring);
10534         ixgbe_clean_rx_ring(rx_ring);
10535 
10536         ixgbe_reset_txr_stats(tx_ring);
10537         if (xdp_ring)
10538                 ixgbe_reset_txr_stats(xdp_ring);
10539         ixgbe_reset_rxr_stats(rx_ring);
10540 }
10541 
10542 /**
10543  * ixgbe_txrx_ring_enable - Enable Rx/Tx/XDP Tx rings
10544  * @adapter: adapter structure
10545  * @ring: ring index
10546  *
10547  * This function enables a certain Rx/Tx/XDP Tx ring. The function
10548  * assumes that the netdev is running.
10549  **/
10550 void ixgbe_txrx_ring_enable(struct ixgbe_adapter *adapter, int ring)
10551 {
10552         struct ixgbe_ring *rx_ring, *tx_ring, *xdp_ring;
10553 
10554         rx_ring = adapter->rx_ring[ring];
10555         tx_ring = adapter->tx_ring[ring];
10556         xdp_ring = adapter->xdp_ring[ring];
10557 
10558         /* Rx/Tx/XDP Tx share the same napi context. */
10559         napi_enable(&rx_ring->q_vector->napi);
10560 
10561         ixgbe_configure_tx_ring(adapter, tx_ring);
10562         if (xdp_ring)
10563                 ixgbe_configure_tx_ring(adapter, xdp_ring);
10564         ixgbe_configure_rx_ring(adapter, rx_ring);
10565 
10566         clear_bit(__IXGBE_TX_DISABLED, &tx_ring->state);
10567         if (xdp_ring)
10568                 clear_bit(__IXGBE_TX_DISABLED, &xdp_ring->state);
10569 }
10570 
10571 /**
10572  * ixgbe_enumerate_functions - Get the number of ports this device has
10573  * @adapter: adapter structure
10574  *
10575  * This function enumerates the phsyical functions co-located on a single slot,
10576  * in order to determine how many ports a device has. This is most useful in
10577  * determining the required GT/s of PCIe bandwidth necessary for optimal
10578  * performance.
10579  **/
10580 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
10581 {
10582         struct pci_dev *entry, *pdev = adapter->pdev;
10583         int physfns = 0;
10584 
10585         /* Some cards can not use the generic count PCIe functions method,
10586          * because they are behind a parent switch, so we hardcode these with
10587          * the correct number of functions.
10588          */
10589         if (ixgbe_pcie_from_parent(&adapter->hw))
10590                 physfns = 4;
10591 
10592         list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
10593                 /* don't count virtual functions */
10594                 if (entry->is_virtfn)
10595                         continue;
10596 
10597                 /* When the devices on the bus don't all match our device ID,
10598                  * we can't reliably determine the correct number of
10599                  * functions. This can occur if a function has been direct
10600                  * attached to a virtual machine using VT-d, for example. In
10601                  * this case, simply return -1 to indicate this.
10602                  */
10603                 if ((entry->vendor != pdev->vendor) ||
10604                     (entry->device != pdev->device))
10605                         return -1;
10606 
10607                 physfns++;
10608         }
10609 
10610         return physfns;
10611 }
10612 
10613 /**
10614  * ixgbe_wol_supported - Check whether device supports WoL
10615  * @adapter: the adapter private structure
10616  * @device_id: the device ID
10617  * @subdevice_id: the subsystem device ID
10618  *
10619  * This function is used by probe and ethtool to determine
10620  * which devices have WoL support
10621  *
10622  **/
10623 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
10624                          u16 subdevice_id)
10625 {
10626         struct ixgbe_hw *hw = &adapter->hw;
10627         u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
10628 
10629         /* WOL not supported on 82598 */
10630         if (hw->mac.type == ixgbe_mac_82598EB)
10631                 return false;
10632 
10633         /* check eeprom to see if WOL is enabled for X540 and newer */
10634         if (hw->mac.type >= ixgbe_mac_X540) {
10635                 if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
10636                     ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
10637                      (hw->bus.func == 0)))
10638                         return true;
10639         }
10640 
10641         /* WOL is determined based on device IDs for 82599 MACs */
10642         switch (device_id) {
10643         case IXGBE_DEV_ID_82599_SFP:
10644                 /* Only these subdevices could supports WOL */
10645                 switch (subdevice_id) {
10646                 case IXGBE_SUBDEV_ID_82599_560FLR:
10647                 case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
10648                 case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
10649                 case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
10650                         /* only support first port */
10651                         if (hw->bus.func != 0)
10652                                 break;
10653                         /* fall through */
10654                 case IXGBE_SUBDEV_ID_82599_SP_560FLR:
10655                 case IXGBE_SUBDEV_ID_82599_SFP:
10656                 case IXGBE_SUBDEV_ID_82599_RNDC:
10657                 case IXGBE_SUBDEV_ID_82599_ECNA_DP:
10658                 case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
10659                 case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
10660                 case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
10661                         return true;
10662                 }
10663                 break;
10664         case IXGBE_DEV_ID_82599EN_SFP:
10665                 /* Only these subdevices support WOL */
10666                 switch (subdevice_id) {
10667                 case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
10668                         return true;
10669                 }
10670                 break;
10671         case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
10672                 /* All except this subdevice support WOL */
10673                 if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
10674                         return true;
10675                 break;
10676         case IXGBE_DEV_ID_82599_KX4:
10677                 return  true;
10678         default:
10679                 break;
10680         }
10681 
10682         return false;
10683 }
10684 
10685 /**
10686  * ixgbe_set_fw_version - Set FW version
10687  * @adapter: the adapter private structure
10688  *
10689  * This function is used by probe and ethtool to determine the FW version to
10690  * format to display. The FW version is taken from the EEPROM/NVM.
10691  */
10692 static void ixgbe_set_fw_version(struct ixgbe_adapter *adapter)
10693 {
10694         struct ixgbe_hw *hw = &adapter->hw;
10695         struct ixgbe_nvm_version nvm_ver;
10696 
10697         ixgbe_get_oem_prod_version(hw, &nvm_ver);
10698         if (nvm_ver.oem_valid) {
10699                 snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10700                          "%x.%x.%x", nvm_ver.oem_major, nvm_ver.oem_minor,
10701                          nvm_ver.oem_release);
10702                 return;
10703         }
10704 
10705         ixgbe_get_etk_id(hw, &nvm_ver);
10706         ixgbe_get_orom_version(hw, &nvm_ver);
10707 
10708         if (nvm_ver.or_valid) {
10709                 snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10710                          "0x%08x, %d.%d.%d", nvm_ver.etk_id, nvm_ver.or_major,
10711                          nvm_ver.or_build, nvm_ver.or_patch);
10712                 return;
10713         }
10714 
10715         /* Set ETrack ID format */
10716         snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10717                  "0x%08x", nvm_ver.etk_id);
10718 }
10719 
10720 /**
10721  * ixgbe_probe - Device Initialization Routine
10722  * @pdev: PCI device information struct
10723  * @ent: entry in ixgbe_pci_tbl
10724  *
10725  * Returns 0 on success, negative on failure
10726  *
10727  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
10728  * The OS initialization, configuring of the adapter private structure,
10729  * and a hardware reset occur.
10730  **/
10731 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10732 {
10733         struct net_device *netdev;
10734         struct ixgbe_adapter *adapter = NULL;
10735         struct ixgbe_hw *hw;
10736         const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
10737         int i, err, pci_using_dac, expected_gts;
10738         unsigned int indices = MAX_TX_QUEUES;
10739         u8 part_str[IXGBE_PBANUM_LENGTH];
10740         bool disable_dev = false;
10741 #ifdef IXGBE_FCOE
10742         u16 device_caps;
10743 #endif
10744         u32 eec;
10745 
10746         /* Catch broken hardware that put the wrong VF device ID in
10747          * the PCIe SR-IOV capability.
10748          */
10749         if (pdev->is_virtfn) {
10750                 WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
10751                      pci_name(pdev), pdev->vendor, pdev->device);
10752                 return -EINVAL;
10753         }
10754 
10755         err = pci_enable_device_mem(pdev);
10756         if (err)
10757                 return err;
10758 
10759         if (!dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64))) {
10760                 pci_using_dac = 1;
10761         } else {
10762                 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10763                 if (err) {
10764                         dev_err(&pdev->dev,
10765                                 "No usable DMA configuration, aborting\n");
10766                         goto err_dma;
10767                 }
10768                 pci_using_dac = 0;
10769         }
10770 
10771         err = pci_request_mem_regions(pdev, ixgbe_driver_name);
10772         if (err) {
10773                 dev_err(&pdev->dev,
10774                         "pci_request_selected_regions failed 0x%x\n", err);
10775                 goto err_pci_reg;
10776         }
10777 
10778         pci_enable_pcie_error_reporting(pdev);
10779 
10780         pci_set_master(pdev);
10781         pci_save_state(pdev);
10782 
10783         if (ii->mac == ixgbe_mac_82598EB) {
10784 #ifdef CONFIG_IXGBE_DCB
10785                 /* 8 TC w/ 4 queues per TC */
10786                 indices = 4 * MAX_TRAFFIC_CLASS;
10787 #else
10788                 indices = IXGBE_MAX_RSS_INDICES;
10789 #endif
10790         }
10791 
10792         netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
10793         if (!netdev) {
10794                 err = -ENOMEM;
10795                 goto err_alloc_etherdev;
10796         }
10797 
10798         SET_NETDEV_DEV(netdev, &pdev->dev);
10799 
10800         adapter = netdev_priv(netdev);
10801 
10802         adapter->netdev = netdev;
10803         adapter->pdev = pdev;
10804         hw = &adapter->hw;
10805         hw->back = adapter;
10806         adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
10807 
10808         hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
10809                               pci_resource_len(pdev, 0));
10810         adapter->io_addr = hw->hw_addr;
10811         if (!hw->hw_addr) {
10812                 err = -EIO;
10813                 goto err_ioremap;
10814         }
10815 
10816         netdev->netdev_ops = &ixgbe_netdev_ops;
10817         ixgbe_set_ethtool_ops(netdev);
10818         netdev->watchdog_timeo = 5 * HZ;
10819         strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
10820 
10821         /* Setup hw api */
10822         hw->mac.ops   = *ii->mac_ops;
10823         hw->mac.type  = ii->mac;
10824         hw->mvals     = ii->mvals;
10825         if (ii->link_ops)
10826                 hw->link.ops  = *ii->link_ops;
10827 
10828         /* EEPROM */
10829         hw->eeprom.ops = *ii->eeprom_ops;
10830         eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
10831         if (ixgbe_removed(hw->hw_addr)) {
10832                 err = -EIO;
10833                 goto err_ioremap;
10834         }
10835         /* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
10836         if (!(eec & BIT(8)))
10837                 hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
10838 
10839         /* PHY */
10840         hw->phy.ops = *ii->phy_ops;
10841         hw->phy.sfp_type = ixgbe_sfp_type_unknown;
10842         /* ixgbe_identify_phy_generic will set prtad and mmds properly */
10843         hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
10844         hw->phy.mdio.mmds = 0;
10845         hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
10846         hw->phy.mdio.dev = netdev;
10847         hw->phy.mdio.mdio_read = ixgbe_mdio_read;
10848         hw->phy.mdio.mdio_write = ixgbe_mdio_write;
10849 
10850         /* setup the private structure */
10851         err = ixgbe_sw_init(adapter, ii);
10852         if (err)
10853                 goto err_sw_init;
10854 
10855         /* Make sure the SWFW semaphore is in a valid state */
10856         if (hw->mac.ops.init_swfw_sync)
10857                 hw->mac.ops.init_swfw_sync(hw);
10858 
10859         /* Make it possible the adapter to be woken up via WOL */
10860         switch (adapter->hw.mac.type) {
10861         case ixgbe_mac_82599EB:
10862         case ixgbe_mac_X540:
10863         case ixgbe_mac_X550:
10864         case ixgbe_mac_X550EM_x:
10865         case ixgbe_mac_x550em_a:
10866                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10867                 break;
10868         default:
10869                 break;
10870         }
10871 
10872         /*
10873          * If there is a fan on this device and it has failed log the
10874          * failure.
10875          */
10876         if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
10877                 u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
10878                 if (esdp & IXGBE_ESDP_SDP1)
10879                         e_crit(probe, "Fan has stopped, replace the adapter\n");
10880         }
10881 
10882         if (allow_unsupported_sfp)
10883                 hw->allow_unsupported_sfp = allow_unsupported_sfp;
10884 
10885         /* reset_hw fills in the perm_addr as well */
10886         hw->phy.reset_if_overtemp = true;
10887         err = hw->mac.ops.reset_hw(hw);
10888         hw->phy.reset_if_overtemp = false;
10889         ixgbe_set_eee_capable(adapter);
10890         if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
10891                 err = 0;
10892         } else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
10893                 e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
10894                 e_dev_err("Reload the driver after installing a supported module.\n");
10895                 goto err_sw_init;
10896         } else if (err) {
10897                 e_dev_err("HW Init failed: %d\n", err);
10898                 goto err_sw_init;
10899         }
10900 
10901 #ifdef CONFIG_PCI_IOV
10902         /* SR-IOV not supported on the 82598 */
10903         if (adapter->hw.mac.type == ixgbe_mac_82598EB)
10904                 goto skip_sriov;
10905         /* Mailbox */
10906         ixgbe_init_mbx_params_pf(hw);
10907         hw->mbx.ops = ii->mbx_ops;
10908         pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
10909         ixgbe_enable_sriov(adapter, max_vfs);
10910 skip_sriov:
10911 
10912 #endif
10913         netdev->features = NETIF_F_SG |
10914                            NETIF_F_TSO |
10915                            NETIF_F_TSO6 |
10916                            NETIF_F_RXHASH |
10917                            NETIF_F_RXCSUM |
10918                            NETIF_F_HW_CSUM;
10919 
10920 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
10921                                     NETIF_F_GSO_GRE_CSUM | \
10922                                     NETIF_F_GSO_IPXIP4 | \
10923                                     NETIF_F_GSO_IPXIP6 | \
10924                                     NETIF_F_GSO_UDP_TUNNEL | \
10925                                     NETIF_F_GSO_UDP_TUNNEL_CSUM)
10926 
10927         netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
10928         netdev->features |= NETIF_F_GSO_PARTIAL |
10929                             IXGBE_GSO_PARTIAL_FEATURES;
10930 
10931         if (hw->mac.type >= ixgbe_mac_82599EB)
10932                 netdev->features |= NETIF_F_SCTP_CRC;
10933 
10934 #ifdef CONFIG_IXGBE_IPSEC
10935 #define IXGBE_ESP_FEATURES      (NETIF_F_HW_ESP | \
10936                                  NETIF_F_HW_ESP_TX_CSUM | \
10937                                  NETIF_F_GSO_ESP)
10938 
10939         if (adapter->ipsec)
10940                 netdev->features |= IXGBE_ESP_FEATURES;
10941 #endif
10942         /* copy netdev features into list of user selectable features */
10943         netdev->hw_features |= netdev->features |
10944                                NETIF_F_HW_VLAN_CTAG_FILTER |
10945                                NETIF_F_HW_VLAN_CTAG_RX |
10946                                NETIF_F_HW_VLAN_CTAG_TX |
10947                                NETIF_F_RXALL |
10948                                NETIF_F_HW_L2FW_DOFFLOAD;
10949 
10950         if (hw->mac.type >= ixgbe_mac_82599EB)
10951                 netdev->hw_features |= NETIF_F_NTUPLE |
10952                                        NETIF_F_HW_TC;
10953 
10954         if (pci_using_dac)
10955                 netdev->features |= NETIF_F_HIGHDMA;
10956 
10957         netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
10958         netdev->hw_enc_features |= netdev->vlan_features;
10959         netdev->mpls_features |= NETIF_F_SG |
10960                                  NETIF_F_TSO |
10961                                  NETIF_F_TSO6 |
10962                                  NETIF_F_HW_CSUM;
10963         netdev->mpls_features |= IXGBE_GSO_PARTIAL_FEATURES;
10964 
10965         /* set this bit last since it cannot be part of vlan_features */
10966         netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
10967                             NETIF_F_HW_VLAN_CTAG_RX |
10968                             NETIF_F_HW_VLAN_CTAG_TX;
10969 
10970         netdev->priv_flags |= IFF_UNICAST_FLT;
10971         netdev->priv_flags |= IFF_SUPP_NOFCS;
10972 
10973         /* MTU range: 68 - 9710 */
10974         netdev->min_mtu = ETH_MIN_MTU;
10975         netdev->max_mtu = IXGBE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN);
10976 
10977 #ifdef CONFIG_IXGBE_DCB
10978         if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
10979                 netdev->dcbnl_ops = &ixgbe_dcbnl_ops;
10980 #endif
10981 
10982 #ifdef IXGBE_FCOE
10983         if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
10984                 unsigned int fcoe_l;
10985 
10986                 if (hw->mac.ops.get_device_caps) {
10987                         hw->mac.ops.get_device_caps(hw, &device_caps);
10988                         if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
10989                                 adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
10990                 }
10991 
10992 
10993                 fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
10994                 adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
10995 
10996                 netdev->features |= NETIF_F_FSO |
10997                                     NETIF_F_FCOE_CRC;
10998 
10999                 netdev->vlan_features |= NETIF_F_FSO |
11000                                          NETIF_F_FCOE_CRC |
11001                                          NETIF_F_FCOE_MTU;
11002         }
11003 #endif /* IXGBE_FCOE */
11004         if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
11005                 netdev->hw_features |= NETIF_F_LRO;
11006         if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
11007                 netdev->features |= NETIF_F_LRO;
11008 
11009         if (ixgbe_check_fw_error(adapter)) {
11010                 err = -EIO;
11011                 goto err_sw_init;
11012         }
11013 
11014         /* make sure the EEPROM is good */
11015         if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
11016                 e_dev_err("The EEPROM Checksum Is Not Valid\n");
11017                 err = -EIO;
11018                 goto err_sw_init;
11019         }
11020 
11021         eth_platform_get_mac_address(&adapter->pdev->dev,
11022                                      adapter->hw.mac.perm_addr);
11023 
11024         memcpy(netdev->dev_addr, hw->mac.perm_addr, netdev->addr_len);
11025 
11026         if (!is_valid_ether_addr(netdev->dev_addr)) {
11027                 e_dev_err("invalid MAC address\n");
11028                 err = -EIO;
11029                 goto err_sw_init;
11030         }
11031 
11032         /* Set hw->mac.addr to permanent MAC address */
11033         ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
11034         ixgbe_mac_set_default_filter(adapter);
11035 
11036         timer_setup(&adapter->service_timer, ixgbe_service_timer, 0);
11037 
11038         if (ixgbe_removed(hw->hw_addr)) {
11039                 err = -EIO;
11040                 goto err_sw_init;
11041         }
11042         INIT_WORK(&adapter->service_task, ixgbe_service_task);
11043         set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
11044         clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
11045 
11046         err = ixgbe_init_interrupt_scheme(adapter);
11047         if (err)
11048                 goto err_sw_init;
11049 
11050         for (i = 0; i < adapter->num_rx_queues; i++)
11051                 u64_stats_init(&adapter->rx_ring[i]->syncp);
11052         for (i = 0; i < adapter->num_tx_queues; i++)
11053                 u64_stats_init(&adapter->tx_ring[i]->syncp);
11054         for (i = 0; i < adapter->num_xdp_queues; i++)
11055                 u64_stats_init(&adapter->xdp_ring[i]->syncp);
11056 
11057         /* WOL not supported for all devices */
11058         adapter->wol = 0;
11059         hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
11060         hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
11061                                                 pdev->subsystem_device);
11062         if (hw->wol_enabled)
11063                 adapter->wol = IXGBE_WUFC_MAG;
11064 
11065         device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
11066 
11067         /* save off EEPROM version number */
11068         ixgbe_set_fw_version(adapter);
11069 
11070         /* pick up the PCI bus settings for reporting later */
11071         if (ixgbe_pcie_from_parent(hw))
11072                 ixgbe_get_parent_bus_info(adapter);
11073         else
11074                  hw->mac.ops.get_bus_info(hw);
11075 
11076         /* calculate the expected PCIe bandwidth required for optimal
11077          * performance. Note that some older parts will never have enough
11078          * bandwidth due to being older generation PCIe parts. We clamp these
11079          * parts to ensure no warning is displayed if it can't be fixed.
11080          */
11081         switch (hw->mac.type) {
11082         case ixgbe_mac_82598EB:
11083                 expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
11084                 break;
11085         default:
11086                 expected_gts = ixgbe_enumerate_functions(adapter) * 10;
11087                 break;
11088         }
11089 
11090         /* don't check link if we failed to enumerate functions */
11091         if (expected_gts > 0)
11092                 ixgbe_check_minimum_link(adapter, expected_gts);
11093 
11094         err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
11095         if (err)
11096                 strlcpy(part_str, "Unknown", sizeof(part_str));
11097         if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
11098                 e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
11099                            hw->mac.type, hw->phy.type, hw->phy.sfp_type,
11100                            part_str);
11101         else
11102                 e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
11103                            hw->mac.type, hw->phy.type, part_str);
11104 
11105         e_dev_info("%pM\n", netdev->dev_addr);
11106 
11107         /* reset the hardware with the new settings */
11108         err = hw->mac.ops.start_hw(hw);
11109         if (err == IXGBE_ERR_EEPROM_VERSION) {
11110                 /* We are running on a pre-production device, log a warning */
11111                 e_dev_warn("This device is a pre-production adapter/LOM. "
11112                            "Please be aware there may be issues associated "
11113                            "with your hardware.  If you are experiencing "
11114                            "problems please contact your Intel or hardware "
11115                            "representative who provided you with this "
11116                            "hardware.\n");
11117         }
11118         strcpy(netdev->name, "eth%d");
11119         pci_set_drvdata(pdev, adapter);
11120         err = register_netdev(netdev);
11121         if (err)
11122                 goto err_register;
11123 
11124 
11125         /* power down the optics for 82599 SFP+ fiber */
11126         if (hw->mac.ops.disable_tx_laser)
11127                 hw->mac.ops.disable_tx_laser(hw);
11128 
11129         /* carrier off reporting is important to ethtool even BEFORE open */
11130         netif_carrier_off(netdev);
11131 
11132 #ifdef CONFIG_IXGBE_DCA
11133         if (dca_add_requester(&pdev->dev) == 0) {
11134                 adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
11135                 ixgbe_setup_dca(adapter);
11136         }
11137 #endif
11138         if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
11139                 e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
11140                 for (i = 0; i < adapter->num_vfs; i++)
11141                         ixgbe_vf_configuration(pdev, (i | 0x10000000));
11142         }
11143 
11144         /* firmware requires driver version to be 0xFFFFFFFF
11145          * since os does not support feature
11146          */
11147         if (hw->mac.ops.set_fw_drv_ver)
11148                 hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF, 0xFF,
11149                                            sizeof(ixgbe_driver_version) - 1,
11150                                            ixgbe_driver_version);
11151 
11152         /* add san mac addr to netdev */
11153         ixgbe_add_sanmac_netdev(netdev);
11154 
11155         e_dev_info("%s\n", ixgbe_default_device_descr);
11156 
11157 #ifdef CONFIG_IXGBE_HWMON
11158         if (ixgbe_sysfs_init(adapter))
11159                 e_err(probe, "failed to allocate sysfs resources\n");
11160 #endif /* CONFIG_IXGBE_HWMON */
11161 
11162         ixgbe_dbg_adapter_init(adapter);
11163 
11164         /* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
11165         if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
11166                 hw->mac.ops.setup_link(hw,
11167                         IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
11168                         true);
11169 
11170         ixgbe_mii_bus_init(hw);
11171 
11172         return 0;
11173 
11174 err_register:
11175         ixgbe_release_hw_control(adapter);
11176         ixgbe_clear_interrupt_scheme(adapter);
11177 err_sw_init:
11178         ixgbe_disable_sriov(adapter);
11179         adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
11180         iounmap(adapter->io_addr);
11181         kfree(adapter->jump_tables[0]);
11182         kfree(adapter->mac_table);
11183         kfree(adapter->rss_key);
11184         bitmap_free(adapter->af_xdp_zc_qps);
11185 err_ioremap:
11186         disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
11187         free_netdev(netdev);
11188 err_alloc_etherdev:
11189         pci_release_mem_regions(pdev);
11190 err_pci_reg:
11191 err_dma:
11192         if (!adapter || disable_dev)
11193                 pci_disable_device(pdev);
11194         return err;
11195 }
11196 
11197 /**
11198  * ixgbe_remove - Device Removal Routine
11199  * @pdev: PCI device information struct
11200  *
11201  * ixgbe_remove is called by the PCI subsystem to alert the driver
11202  * that it should release a PCI device.  The could be caused by a
11203  * Hot-Plug event, or because the driver is going to be removed from
11204  * memory.
11205  **/
11206 static void ixgbe_remove(struct pci_dev *pdev)
11207 {
11208         struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11209         struct net_device *netdev;
11210         bool disable_dev;
11211         int i;
11212 
11213         /* if !adapter then we already cleaned up in probe */
11214         if (!adapter)
11215                 return;
11216 
11217         netdev  = adapter->netdev;
11218         ixgbe_dbg_adapter_exit(adapter);
11219 
11220         set_bit(__IXGBE_REMOVING, &adapter->state);
11221         cancel_work_sync(&adapter->service_task);
11222 
11223         if (adapter->mii_bus)
11224                 mdiobus_unregister(adapter->mii_bus);
11225 
11226 #ifdef CONFIG_IXGBE_DCA
11227         if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
11228                 adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
11229                 dca_remove_requester(&pdev->dev);
11230                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
11231                                 IXGBE_DCA_CTRL_DCA_DISABLE);
11232         }
11233 
11234 #endif
11235 #ifdef CONFIG_IXGBE_HWMON
11236         ixgbe_sysfs_exit(adapter);
11237 #endif /* CONFIG_IXGBE_HWMON */
11238 
11239         /* remove the added san mac */
11240         ixgbe_del_sanmac_netdev(netdev);
11241 
11242 #ifdef CONFIG_PCI_IOV
11243         ixgbe_disable_sriov(adapter);
11244 #endif
11245         if (netdev->reg_state == NETREG_REGISTERED)
11246                 unregister_netdev(netdev);
11247 
11248         ixgbe_stop_ipsec_offload(adapter);
11249         ixgbe_clear_interrupt_scheme(adapter);
11250 
11251         ixgbe_release_hw_control(adapter);
11252 
11253 #ifdef CONFIG_DCB
11254         kfree(adapter->ixgbe_ieee_pfc);
11255         kfree(adapter->ixgbe_ieee_ets);
11256 
11257 #endif
11258         iounmap(adapter->io_addr);
11259         pci_release_mem_regions(pdev);
11260 
11261         e_dev_info("complete\n");
11262 
11263         for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
11264                 if (adapter->jump_tables[i]) {
11265                         kfree(adapter->jump_tables[i]->input);
11266                         kfree(adapter->jump_tables[i]->mask);
11267                 }
11268                 kfree(adapter->jump_tables[i]);
11269         }
11270 
11271         kfree(adapter->mac_table);
11272         kfree(adapter->rss_key);
11273         bitmap_free(adapter->af_xdp_zc_qps);
11274         disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
11275         free_netdev(netdev);
11276 
11277         pci_disable_pcie_error_reporting(pdev);
11278 
11279         if (disable_dev)
11280                 pci_disable_device(pdev);
11281 }
11282 
11283 /**
11284  * ixgbe_io_error_detected - called when PCI error is detected
11285  * @pdev: Pointer to PCI device
11286  * @state: The current pci connection state
11287  *
11288  * This function is called after a PCI bus error affecting
11289  * this device has been detected.
11290  */
11291 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
11292                                                 pci_channel_state_t state)
11293 {
11294         struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11295         struct net_device *netdev = adapter->netdev;
11296 
11297 #ifdef CONFIG_PCI_IOV
11298         struct ixgbe_hw *hw = &adapter->hw;
11299         struct pci_dev *bdev, *vfdev;
11300         u32 dw0, dw1, dw2, dw3;
11301         int vf, pos;
11302         u16 req_id, pf_func;
11303 
11304         if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
11305             adapter->num_vfs == 0)
11306                 goto skip_bad_vf_detection;
11307 
11308         bdev = pdev->bus->self;
11309         while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
11310                 bdev = bdev->bus->self;
11311 
11312         if (!bdev)
11313                 goto skip_bad_vf_detection;
11314 
11315         pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
11316         if (!pos)
11317                 goto skip_bad_vf_detection;
11318 
11319         dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
11320         dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
11321         dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
11322         dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
11323         if (ixgbe_removed(hw->hw_addr))
11324                 goto skip_bad_vf_detection;
11325 
11326         req_id = dw1 >> 16;
11327         /* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
11328         if (!(req_id & 0x0080))
11329                 goto skip_bad_vf_detection;
11330 
11331         pf_func = req_id & 0x01;
11332         if ((pf_func & 1) == (pdev->devfn & 1)) {
11333                 unsigned int device_id;
11334 
11335                 vf = (req_id & 0x7F) >> 1;
11336                 e_dev_err("VF %d has caused a PCIe error\n", vf);
11337                 e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
11338                                 "%8.8x\tdw3: %8.8x\n",
11339                 dw0, dw1, dw2, dw3);
11340                 switch (adapter->hw.mac.type) {
11341                 case ixgbe_mac_82599EB:
11342                         device_id = IXGBE_82599_VF_DEVICE_ID;
11343                         break;
11344                 case ixgbe_mac_X540:
11345                         device_id = IXGBE_X540_VF_DEVICE_ID;
11346                         break;
11347                 case ixgbe_mac_X550:
11348                         device_id = IXGBE_DEV_ID_X550_VF;
11349                         break;
11350                 case ixgbe_mac_X550EM_x:
11351                         device_id = IXGBE_DEV_ID_X550EM_X_VF;
11352                         break;
11353                 case ixgbe_mac_x550em_a:
11354                         device_id = IXGBE_DEV_ID_X550EM_A_VF;
11355                         break;
11356                 default:
11357                         device_id = 0;
11358                         break;
11359                 }
11360 
11361                 /* Find the pci device of the offending VF */
11362                 vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
11363                 while (vfdev) {
11364                         if (vfdev->devfn == (req_id & 0xFF))
11365                                 break;
11366                         vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
11367                                                device_id, vfdev);
11368                 }
11369                 /*
11370                  * There's a slim chance the VF could have been hot plugged,
11371                  * so if it is no longer present we don't need to issue the
11372                  * VFLR.  Just clean up the AER in that case.
11373                  */
11374                 if (vfdev) {
11375                         pcie_flr(vfdev);
11376                         /* Free device reference count */
11377                         pci_dev_put(vfdev);
11378                 }
11379         }
11380 
11381         /*
11382          * Even though the error may have occurred on the other port
11383          * we still need to increment the vf error reference count for
11384          * both ports because the I/O resume function will be called
11385          * for both of them.
11386          */
11387         adapter->vferr_refcount++;
11388 
11389         return PCI_ERS_RESULT_RECOVERED;
11390 
11391 skip_bad_vf_detection:
11392 #endif /* CONFIG_PCI_IOV */
11393         if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
11394                 return PCI_ERS_RESULT_DISCONNECT;
11395 
11396         if (!netif_device_present(netdev))
11397                 return PCI_ERS_RESULT_DISCONNECT;
11398 
11399         rtnl_lock();
11400         netif_device_detach(netdev);
11401 
11402         if (netif_running(netdev))
11403                 ixgbe_close_suspend(adapter);
11404 
11405         if (state == pci_channel_io_perm_failure) {
11406                 rtnl_unlock();
11407                 return PCI_ERS_RESULT_DISCONNECT;
11408         }
11409 
11410         if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
11411                 pci_disable_device(pdev);
11412         rtnl_unlock();
11413 
11414         /* Request a slot reset. */
11415         return PCI_ERS_RESULT_NEED_RESET;
11416 }
11417 
11418 /**
11419  * ixgbe_io_slot_reset - called after the pci bus has been reset.
11420  * @pdev: Pointer to PCI device
11421  *
11422  * Restart the card from scratch, as if from a cold-boot.
11423  */
11424 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
11425 {
11426         struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11427         pci_ers_result_t result;
11428 
11429         if (pci_enable_device_mem(pdev)) {
11430                 e_err(probe, "Cannot re-enable PCI device after reset.\n");
11431                 result = PCI_ERS_RESULT_DISCONNECT;
11432         } else {
11433                 smp_mb__before_atomic();
11434                 clear_bit(__IXGBE_DISABLED, &adapter->state);
11435                 adapter->hw.hw_addr = adapter->io_addr;
11436                 pci_set_master(pdev);
11437                 pci_restore_state(pdev);
11438                 pci_save_state(pdev);
11439 
11440                 pci_wake_from_d3(pdev, false);
11441 
11442                 ixgbe_reset(adapter);
11443                 IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
11444                 result = PCI_ERS_RESULT_RECOVERED;
11445         }
11446 
11447         return result;
11448 }
11449 
11450 /**
11451  * ixgbe_io_resume - called when traffic can start flowing again.
11452  * @pdev: Pointer to PCI device
11453  *
11454  * This callback is called when the error recovery driver tells us that
11455  * its OK to resume normal operation.
11456  */
11457 static void ixgbe_io_resume(struct pci_dev *pdev)
11458 {
11459         struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11460         struct net_device *netdev = adapter->netdev;
11461 
11462 #ifdef CONFIG_PCI_IOV
11463         if (adapter->vferr_refcount) {
11464                 e_info(drv, "Resuming after VF err\n");
11465                 adapter->vferr_refcount--;
11466                 return;
11467         }
11468 
11469 #endif
11470         rtnl_lock();
11471         if (netif_running(netdev))
11472                 ixgbe_open(netdev);
11473 
11474         netif_device_attach(netdev);
11475         rtnl_unlock();
11476 }
11477 
11478 static const struct pci_error_handlers ixgbe_err_handler = {
11479         .error_detected = ixgbe_io_error_detected,
11480         .slot_reset = ixgbe_io_slot_reset,
11481         .resume = ixgbe_io_resume,
11482 };
11483 
11484 static struct pci_driver ixgbe_driver = {
11485         .name     = ixgbe_driver_name,
11486         .id_table = ixgbe_pci_tbl,
11487         .probe    = ixgbe_probe,
11488         .remove   = ixgbe_remove,
11489 #ifdef CONFIG_PM
11490         .suspend  = ixgbe_suspend,
11491         .resume   = ixgbe_resume,
11492 #endif
11493         .shutdown = ixgbe_shutdown,
11494         .sriov_configure = ixgbe_pci_sriov_configure,
11495         .err_handler = &ixgbe_err_handler
11496 };
11497 
11498 /**
11499  * ixgbe_init_module - Driver Registration Routine
11500  *
11501  * ixgbe_init_module is the first routine called when the driver is
11502  * loaded. All it does is register with the PCI subsystem.
11503  **/
11504 static int __init ixgbe_init_module(void)
11505 {
11506         int ret;
11507         pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
11508         pr_info("%s\n", ixgbe_copyright);
11509 
11510         ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
11511         if (!ixgbe_wq) {
11512                 pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
11513                 return -ENOMEM;
11514         }
11515 
11516         ixgbe_dbg_init();
11517 
11518         ret = pci_register_driver(&ixgbe_driver);
11519         if (ret) {
11520                 destroy_workqueue(ixgbe_wq);
11521                 ixgbe_dbg_exit();
11522                 return ret;
11523         }
11524 
11525 #ifdef CONFIG_IXGBE_DCA
11526         dca_register_notify(&dca_notifier);
11527 #endif
11528 
11529         return 0;
11530 }
11531 
11532 module_init(ixgbe_init_module);
11533 
11534 /**
11535  * ixgbe_exit_module - Driver Exit Cleanup Routine
11536  *
11537  * ixgbe_exit_module is called just before the driver is removed
11538  * from memory.
11539  **/
11540 static void __exit ixgbe_exit_module(void)
11541 {
11542 #ifdef CONFIG_IXGBE_DCA
11543         dca_unregister_notify(&dca_notifier);
11544 #endif
11545         pci_unregister_driver(&ixgbe_driver);
11546 
11547         ixgbe_dbg_exit();
11548         if (ixgbe_wq) {
11549                 destroy_workqueue(ixgbe_wq);
11550                 ixgbe_wq = NULL;
11551         }
11552 }
11553 
11554 #ifdef CONFIG_IXGBE_DCA
11555 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
11556                             void *p)
11557 {
11558         int ret_val;
11559 
11560         ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
11561                                          __ixgbe_notify_dca);
11562 
11563         return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
11564 }
11565 
11566 #endif /* CONFIG_IXGBE_DCA */
11567 
11568 module_exit(ixgbe_exit_module);
11569 
11570 /* ixgbe_main.c */

/* [<][>][^][v][top][bottom][index][help] */