Back to home page

OSCL-LXR

 
 

    


0001 // SPDX-License-Identifier: GPL-2.0
0002 /* Copyright(c) 1999 - 2018 Intel Corporation. */
0003 
0004 #include <linux/types.h>
0005 #include <linux/module.h>
0006 #include <linux/pci.h>
0007 #include <linux/netdevice.h>
0008 #include <linux/vmalloc.h>
0009 #include <linux/string.h>
0010 #include <linux/in.h>
0011 #include <linux/interrupt.h>
0012 #include <linux/ip.h>
0013 #include <linux/tcp.h>
0014 #include <linux/sctp.h>
0015 #include <linux/pkt_sched.h>
0016 #include <linux/ipv6.h>
0017 #include <linux/slab.h>
0018 #include <net/checksum.h>
0019 #include <net/ip6_checksum.h>
0020 #include <linux/etherdevice.h>
0021 #include <linux/ethtool.h>
0022 #include <linux/if.h>
0023 #include <linux/if_vlan.h>
0024 #include <linux/if_macvlan.h>
0025 #include <linux/if_bridge.h>
0026 #include <linux/prefetch.h>
0027 #include <linux/bpf.h>
0028 #include <linux/bpf_trace.h>
0029 #include <linux/atomic.h>
0030 #include <linux/numa.h>
0031 #include <generated/utsrelease.h>
0032 #include <scsi/fc/fc_fcoe.h>
0033 #include <net/udp_tunnel.h>
0034 #include <net/pkt_cls.h>
0035 #include <net/tc_act/tc_gact.h>
0036 #include <net/tc_act/tc_mirred.h>
0037 #include <net/vxlan.h>
0038 #include <net/mpls.h>
0039 #include <net/xdp_sock_drv.h>
0040 #include <net/xfrm.h>
0041 
0042 #include "ixgbe.h"
0043 #include "ixgbe_common.h"
0044 #include "ixgbe_dcb_82599.h"
0045 #include "ixgbe_phy.h"
0046 #include "ixgbe_sriov.h"
0047 #include "ixgbe_model.h"
0048 #include "ixgbe_txrx_common.h"
0049 
0050 char ixgbe_driver_name[] = "ixgbe";
0051 static const char ixgbe_driver_string[] =
0052                   "Intel(R) 10 Gigabit PCI Express Network Driver";
0053 #ifdef IXGBE_FCOE
0054 char ixgbe_default_device_descr[] =
0055                   "Intel(R) 10 Gigabit Network Connection";
0056 #else
0057 static char ixgbe_default_device_descr[] =
0058                   "Intel(R) 10 Gigabit Network Connection";
0059 #endif
0060 static const char ixgbe_copyright[] =
0061                 "Copyright (c) 1999-2016 Intel Corporation.";
0062 
0063 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";
0064 
0065 static const struct ixgbe_info *ixgbe_info_tbl[] = {
0066     [board_82598]       = &ixgbe_82598_info,
0067     [board_82599]       = &ixgbe_82599_info,
0068     [board_X540]        = &ixgbe_X540_info,
0069     [board_X550]        = &ixgbe_X550_info,
0070     [board_X550EM_x]    = &ixgbe_X550EM_x_info,
0071     [board_x550em_x_fw] = &ixgbe_x550em_x_fw_info,
0072     [board_x550em_a]    = &ixgbe_x550em_a_info,
0073     [board_x550em_a_fw] = &ixgbe_x550em_a_fw_info,
0074 };
0075 
0076 /* ixgbe_pci_tbl - PCI Device ID Table
0077  *
0078  * Wildcard entries (PCI_ANY_ID) should come last
0079  * Last entry must be all 0s
0080  *
0081  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
0082  *   Class, Class Mask, private data (not used) }
0083  */
0084 static const struct pci_device_id ixgbe_pci_tbl[] = {
0085     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598), board_82598 },
0086     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_DUAL_PORT), board_82598 },
0087     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AF_SINGLE_PORT), board_82598 },
0088     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT), board_82598 },
0089     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598AT2), board_82598 },
0090     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_CX4), board_82598 },
0091     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_CX4_DUAL_PORT), board_82598 },
0092     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_DA_DUAL_PORT), board_82598 },
0093     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM), board_82598 },
0094     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_XF_LR), board_82598 },
0095     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598EB_SFP_LOM), board_82598 },
0096     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82598_BX), board_82598 },
0097     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4), board_82599 },
0098     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_XAUI_LOM), board_82599 },
0099     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KR), board_82599 },
0100     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP), board_82599 },
0101     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_EM), board_82599 },
0102     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_KX4_MEZZ), board_82599 },
0103     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_CX4), board_82599 },
0104     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_BACKPLANE_FCOE), board_82599 },
0105     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_FCOE), board_82599 },
0106     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_T3_LOM), board_82599 },
0107     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_COMBO_BACKPLANE), board_82599 },
0108     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T), board_X540 },
0109     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF2), board_82599 },
0110     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_LS), board_82599 },
0111     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_QSFP_SF_QP), board_82599 },
0112     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599EN_SFP), board_82599 },
0113     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_82599_SFP_SF_QP), board_82599 },
0114     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X540T1), board_X540 },
0115     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T), board_X550},
0116     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550T1), board_X550},
0117     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KX4), board_X550EM_x},
0118     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_XFI), board_X550EM_x},
0119     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_KR), board_X550EM_x},
0120     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_10G_T), board_X550EM_x},
0121     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_SFP), board_X550EM_x},
0122     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_X_1G_T), board_x550em_x_fw},
0123     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR), board_x550em_a },
0124     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_KR_L), board_x550em_a },
0125     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP_N), board_x550em_a },
0126     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII), board_x550em_a },
0127     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SGMII_L), board_x550em_a },
0128     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_10G_T), board_x550em_a},
0129     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_SFP), board_x550em_a },
0130     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T), board_x550em_a_fw },
0131     {PCI_VDEVICE(INTEL, IXGBE_DEV_ID_X550EM_A_1G_T_L), board_x550em_a_fw },
0132     /* required last entry */
0133     {0, }
0134 };
0135 MODULE_DEVICE_TABLE(pci, ixgbe_pci_tbl);
0136 
0137 #ifdef CONFIG_IXGBE_DCA
0138 static int ixgbe_notify_dca(struct notifier_block *, unsigned long event,
0139                 void *p);
0140 static struct notifier_block dca_notifier = {
0141     .notifier_call = ixgbe_notify_dca,
0142     .next          = NULL,
0143     .priority      = 0
0144 };
0145 #endif
0146 
0147 #ifdef CONFIG_PCI_IOV
0148 static unsigned int max_vfs;
0149 module_param(max_vfs, uint, 0);
0150 MODULE_PARM_DESC(max_vfs,
0151          "Maximum number of virtual functions to allocate per physical function - default is zero and maximum value is 63. (Deprecated)");
0152 #endif /* CONFIG_PCI_IOV */
0153 
0154 static bool allow_unsupported_sfp;
0155 module_param(allow_unsupported_sfp, bool, 0);
0156 MODULE_PARM_DESC(allow_unsupported_sfp,
0157          "Allow unsupported and untested SFP+ modules on 82599-based adapters");
0158 
0159 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
0160 static int debug = -1;
0161 module_param(debug, int, 0);
0162 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
0163 
0164 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
0165 MODULE_DESCRIPTION("Intel(R) 10 Gigabit PCI Express Network Driver");
0166 MODULE_LICENSE("GPL v2");
0167 
0168 DEFINE_STATIC_KEY_FALSE(ixgbe_xdp_locking_key);
0169 EXPORT_SYMBOL(ixgbe_xdp_locking_key);
0170 
0171 static struct workqueue_struct *ixgbe_wq;
0172 
0173 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev);
0174 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *);
0175 
0176 static const struct net_device_ops ixgbe_netdev_ops;
0177 
0178 static bool netif_is_ixgbe(struct net_device *dev)
0179 {
0180     return dev && (dev->netdev_ops == &ixgbe_netdev_ops);
0181 }
0182 
0183 static int ixgbe_read_pci_cfg_word_parent(struct ixgbe_adapter *adapter,
0184                       u32 reg, u16 *value)
0185 {
0186     struct pci_dev *parent_dev;
0187     struct pci_bus *parent_bus;
0188 
0189     parent_bus = adapter->pdev->bus->parent;
0190     if (!parent_bus)
0191         return -1;
0192 
0193     parent_dev = parent_bus->self;
0194     if (!parent_dev)
0195         return -1;
0196 
0197     if (!pci_is_pcie(parent_dev))
0198         return -1;
0199 
0200     pcie_capability_read_word(parent_dev, reg, value);
0201     if (*value == IXGBE_FAILED_READ_CFG_WORD &&
0202         ixgbe_check_cfg_remove(&adapter->hw, parent_dev))
0203         return -1;
0204     return 0;
0205 }
0206 
0207 static s32 ixgbe_get_parent_bus_info(struct ixgbe_adapter *adapter)
0208 {
0209     struct ixgbe_hw *hw = &adapter->hw;
0210     u16 link_status = 0;
0211     int err;
0212 
0213     hw->bus.type = ixgbe_bus_type_pci_express;
0214 
0215     /* Get the negotiated link width and speed from PCI config space of the
0216      * parent, as this device is behind a switch
0217      */
0218     err = ixgbe_read_pci_cfg_word_parent(adapter, 18, &link_status);
0219 
0220     /* assume caller will handle error case */
0221     if (err)
0222         return err;
0223 
0224     hw->bus.width = ixgbe_convert_bus_width(link_status);
0225     hw->bus.speed = ixgbe_convert_bus_speed(link_status);
0226 
0227     return 0;
0228 }
0229 
0230 /**
0231  * ixgbe_pcie_from_parent - Determine whether PCIe info should come from parent
0232  * @hw: hw specific details
0233  *
0234  * This function is used by probe to determine whether a device's PCI-Express
0235  * bandwidth details should be gathered from the parent bus instead of from the
0236  * device. Used to ensure that various locations all have the correct device ID
0237  * checks.
0238  */
0239 static inline bool ixgbe_pcie_from_parent(struct ixgbe_hw *hw)
0240 {
0241     switch (hw->device_id) {
0242     case IXGBE_DEV_ID_82599_SFP_SF_QP:
0243     case IXGBE_DEV_ID_82599_QSFP_SF_QP:
0244         return true;
0245     default:
0246         return false;
0247     }
0248 }
0249 
0250 static void ixgbe_check_minimum_link(struct ixgbe_adapter *adapter,
0251                      int expected_gts)
0252 {
0253     struct ixgbe_hw *hw = &adapter->hw;
0254     struct pci_dev *pdev;
0255 
0256     /* Some devices are not connected over PCIe and thus do not negotiate
0257      * speed. These devices do not have valid bus info, and thus any report
0258      * we generate may not be correct.
0259      */
0260     if (hw->bus.type == ixgbe_bus_type_internal)
0261         return;
0262 
0263     /* determine whether to use the parent device */
0264     if (ixgbe_pcie_from_parent(&adapter->hw))
0265         pdev = adapter->pdev->bus->parent->self;
0266     else
0267         pdev = adapter->pdev;
0268 
0269     pcie_print_link_status(pdev);
0270 }
0271 
0272 static void ixgbe_service_event_schedule(struct ixgbe_adapter *adapter)
0273 {
0274     if (!test_bit(__IXGBE_DOWN, &adapter->state) &&
0275         !test_bit(__IXGBE_REMOVING, &adapter->state) &&
0276         !test_and_set_bit(__IXGBE_SERVICE_SCHED, &adapter->state))
0277         queue_work(ixgbe_wq, &adapter->service_task);
0278 }
0279 
0280 static void ixgbe_remove_adapter(struct ixgbe_hw *hw)
0281 {
0282     struct ixgbe_adapter *adapter = hw->back;
0283 
0284     if (!hw->hw_addr)
0285         return;
0286     hw->hw_addr = NULL;
0287     e_dev_err("Adapter removed\n");
0288     if (test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
0289         ixgbe_service_event_schedule(adapter);
0290 }
0291 
0292 static u32 ixgbe_check_remove(struct ixgbe_hw *hw, u32 reg)
0293 {
0294     u8 __iomem *reg_addr;
0295     u32 value;
0296     int i;
0297 
0298     reg_addr = READ_ONCE(hw->hw_addr);
0299     if (ixgbe_removed(reg_addr))
0300         return IXGBE_FAILED_READ_REG;
0301 
0302     /* Register read of 0xFFFFFFF can indicate the adapter has been removed,
0303      * so perform several status register reads to determine if the adapter
0304      * has been removed.
0305      */
0306     for (i = 0; i < IXGBE_FAILED_READ_RETRIES; i++) {
0307         value = readl(reg_addr + IXGBE_STATUS);
0308         if (value != IXGBE_FAILED_READ_REG)
0309             break;
0310         mdelay(3);
0311     }
0312 
0313     if (value == IXGBE_FAILED_READ_REG)
0314         ixgbe_remove_adapter(hw);
0315     else
0316         value = readl(reg_addr + reg);
0317     return value;
0318 }
0319 
0320 /**
0321  * ixgbe_read_reg - Read from device register
0322  * @hw: hw specific details
0323  * @reg: offset of register to read
0324  *
0325  * Returns : value read or IXGBE_FAILED_READ_REG if removed
0326  *
0327  * This function is used to read device registers. It checks for device
0328  * removal by confirming any read that returns all ones by checking the
0329  * status register value for all ones. This function avoids reading from
0330  * the hardware if a removal was previously detected in which case it
0331  * returns IXGBE_FAILED_READ_REG (all ones).
0332  */
0333 u32 ixgbe_read_reg(struct ixgbe_hw *hw, u32 reg)
0334 {
0335     u8 __iomem *reg_addr = READ_ONCE(hw->hw_addr);
0336     u32 value;
0337 
0338     if (ixgbe_removed(reg_addr))
0339         return IXGBE_FAILED_READ_REG;
0340     if (unlikely(hw->phy.nw_mng_if_sel &
0341              IXGBE_NW_MNG_IF_SEL_SGMII_ENABLE)) {
0342         struct ixgbe_adapter *adapter;
0343         int i;
0344 
0345         for (i = 0; i < 200; ++i) {
0346             value = readl(reg_addr + IXGBE_MAC_SGMII_BUSY);
0347             if (likely(!value))
0348                 goto writes_completed;
0349             if (value == IXGBE_FAILED_READ_REG) {
0350                 ixgbe_remove_adapter(hw);
0351                 return IXGBE_FAILED_READ_REG;
0352             }
0353             udelay(5);
0354         }
0355 
0356         adapter = hw->back;
0357         e_warn(hw, "register writes incomplete %08x\n", value);
0358     }
0359 
0360 writes_completed:
0361     value = readl(reg_addr + reg);
0362     if (unlikely(value == IXGBE_FAILED_READ_REG))
0363         value = ixgbe_check_remove(hw, reg);
0364     return value;
0365 }
0366 
0367 static bool ixgbe_check_cfg_remove(struct ixgbe_hw *hw, struct pci_dev *pdev)
0368 {
0369     u16 value;
0370 
0371     pci_read_config_word(pdev, PCI_VENDOR_ID, &value);
0372     if (value == IXGBE_FAILED_READ_CFG_WORD) {
0373         ixgbe_remove_adapter(hw);
0374         return true;
0375     }
0376     return false;
0377 }
0378 
0379 u16 ixgbe_read_pci_cfg_word(struct ixgbe_hw *hw, u32 reg)
0380 {
0381     struct ixgbe_adapter *adapter = hw->back;
0382     u16 value;
0383 
0384     if (ixgbe_removed(hw->hw_addr))
0385         return IXGBE_FAILED_READ_CFG_WORD;
0386     pci_read_config_word(adapter->pdev, reg, &value);
0387     if (value == IXGBE_FAILED_READ_CFG_WORD &&
0388         ixgbe_check_cfg_remove(hw, adapter->pdev))
0389         return IXGBE_FAILED_READ_CFG_WORD;
0390     return value;
0391 }
0392 
0393 #ifdef CONFIG_PCI_IOV
0394 static u32 ixgbe_read_pci_cfg_dword(struct ixgbe_hw *hw, u32 reg)
0395 {
0396     struct ixgbe_adapter *adapter = hw->back;
0397     u32 value;
0398 
0399     if (ixgbe_removed(hw->hw_addr))
0400         return IXGBE_FAILED_READ_CFG_DWORD;
0401     pci_read_config_dword(adapter->pdev, reg, &value);
0402     if (value == IXGBE_FAILED_READ_CFG_DWORD &&
0403         ixgbe_check_cfg_remove(hw, adapter->pdev))
0404         return IXGBE_FAILED_READ_CFG_DWORD;
0405     return value;
0406 }
0407 #endif /* CONFIG_PCI_IOV */
0408 
0409 void ixgbe_write_pci_cfg_word(struct ixgbe_hw *hw, u32 reg, u16 value)
0410 {
0411     struct ixgbe_adapter *adapter = hw->back;
0412 
0413     if (ixgbe_removed(hw->hw_addr))
0414         return;
0415     pci_write_config_word(adapter->pdev, reg, value);
0416 }
0417 
0418 static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
0419 {
0420     BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
0421 
0422     /* flush memory to make sure state is correct before next watchdog */
0423     smp_mb__before_atomic();
0424     clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
0425 }
0426 
0427 struct ixgbe_reg_info {
0428     u32 ofs;
0429     char *name;
0430 };
0431 
0432 static const struct ixgbe_reg_info ixgbe_reg_info_tbl[] = {
0433 
0434     /* General Registers */
0435     {IXGBE_CTRL, "CTRL"},
0436     {IXGBE_STATUS, "STATUS"},
0437     {IXGBE_CTRL_EXT, "CTRL_EXT"},
0438 
0439     /* Interrupt Registers */
0440     {IXGBE_EICR, "EICR"},
0441 
0442     /* RX Registers */
0443     {IXGBE_SRRCTL(0), "SRRCTL"},
0444     {IXGBE_DCA_RXCTRL(0), "DRXCTL"},
0445     {IXGBE_RDLEN(0), "RDLEN"},
0446     {IXGBE_RDH(0), "RDH"},
0447     {IXGBE_RDT(0), "RDT"},
0448     {IXGBE_RXDCTL(0), "RXDCTL"},
0449     {IXGBE_RDBAL(0), "RDBAL"},
0450     {IXGBE_RDBAH(0), "RDBAH"},
0451 
0452     /* TX Registers */
0453     {IXGBE_TDBAL(0), "TDBAL"},
0454     {IXGBE_TDBAH(0), "TDBAH"},
0455     {IXGBE_TDLEN(0), "TDLEN"},
0456     {IXGBE_TDH(0), "TDH"},
0457     {IXGBE_TDT(0), "TDT"},
0458     {IXGBE_TXDCTL(0), "TXDCTL"},
0459 
0460     /* List Terminator */
0461     { .name = NULL }
0462 };
0463 
0464 
0465 /*
0466  * ixgbe_regdump - register printout routine
0467  */
0468 static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
0469 {
0470     int i;
0471     char rname[16];
0472     u32 regs[64];
0473 
0474     switch (reginfo->ofs) {
0475     case IXGBE_SRRCTL(0):
0476         for (i = 0; i < 64; i++)
0477             regs[i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
0478         break;
0479     case IXGBE_DCA_RXCTRL(0):
0480         for (i = 0; i < 64; i++)
0481             regs[i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
0482         break;
0483     case IXGBE_RDLEN(0):
0484         for (i = 0; i < 64; i++)
0485             regs[i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i));
0486         break;
0487     case IXGBE_RDH(0):
0488         for (i = 0; i < 64; i++)
0489             regs[i] = IXGBE_READ_REG(hw, IXGBE_RDH(i));
0490         break;
0491     case IXGBE_RDT(0):
0492         for (i = 0; i < 64; i++)
0493             regs[i] = IXGBE_READ_REG(hw, IXGBE_RDT(i));
0494         break;
0495     case IXGBE_RXDCTL(0):
0496         for (i = 0; i < 64; i++)
0497             regs[i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
0498         break;
0499     case IXGBE_RDBAL(0):
0500         for (i = 0; i < 64; i++)
0501             regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i));
0502         break;
0503     case IXGBE_RDBAH(0):
0504         for (i = 0; i < 64; i++)
0505             regs[i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i));
0506         break;
0507     case IXGBE_TDBAL(0):
0508         for (i = 0; i < 64; i++)
0509             regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i));
0510         break;
0511     case IXGBE_TDBAH(0):
0512         for (i = 0; i < 64; i++)
0513             regs[i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i));
0514         break;
0515     case IXGBE_TDLEN(0):
0516         for (i = 0; i < 64; i++)
0517             regs[i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i));
0518         break;
0519     case IXGBE_TDH(0):
0520         for (i = 0; i < 64; i++)
0521             regs[i] = IXGBE_READ_REG(hw, IXGBE_TDH(i));
0522         break;
0523     case IXGBE_TDT(0):
0524         for (i = 0; i < 64; i++)
0525             regs[i] = IXGBE_READ_REG(hw, IXGBE_TDT(i));
0526         break;
0527     case IXGBE_TXDCTL(0):
0528         for (i = 0; i < 64; i++)
0529             regs[i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
0530         break;
0531     default:
0532         pr_info("%-15s %08x\n",
0533             reginfo->name, IXGBE_READ_REG(hw, reginfo->ofs));
0534         return;
0535     }
0536 
0537     i = 0;
0538     while (i < 64) {
0539         int j;
0540         char buf[9 * 8 + 1];
0541         char *p = buf;
0542 
0543         snprintf(rname, 16, "%s[%d-%d]", reginfo->name, i, i + 7);
0544         for (j = 0; j < 8; j++)
0545             p += sprintf(p, " %08x", regs[i++]);
0546         pr_err("%-15s%s\n", rname, buf);
0547     }
0548 
0549 }
0550 
0551 static void ixgbe_print_buffer(struct ixgbe_ring *ring, int n)
0552 {
0553     struct ixgbe_tx_buffer *tx_buffer;
0554 
0555     tx_buffer = &ring->tx_buffer_info[ring->next_to_clean];
0556     pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
0557         n, ring->next_to_use, ring->next_to_clean,
0558         (u64)dma_unmap_addr(tx_buffer, dma),
0559         dma_unmap_len(tx_buffer, len),
0560         tx_buffer->next_to_watch,
0561         (u64)tx_buffer->time_stamp);
0562 }
0563 
0564 /*
0565  * ixgbe_dump - Print registers, tx-rings and rx-rings
0566  */
0567 static void ixgbe_dump(struct ixgbe_adapter *adapter)
0568 {
0569     struct net_device *netdev = adapter->netdev;
0570     struct ixgbe_hw *hw = &adapter->hw;
0571     struct ixgbe_reg_info *reginfo;
0572     int n = 0;
0573     struct ixgbe_ring *ring;
0574     struct ixgbe_tx_buffer *tx_buffer;
0575     union ixgbe_adv_tx_desc *tx_desc;
0576     struct my_u0 { u64 a; u64 b; } *u0;
0577     struct ixgbe_ring *rx_ring;
0578     union ixgbe_adv_rx_desc *rx_desc;
0579     struct ixgbe_rx_buffer *rx_buffer_info;
0580     int i = 0;
0581 
0582     if (!netif_msg_hw(adapter))
0583         return;
0584 
0585     /* Print netdevice Info */
0586     if (netdev) {
0587         dev_info(&adapter->pdev->dev, "Net device Info\n");
0588         pr_info("Device Name     state            "
0589             "trans_start\n");
0590         pr_info("%-15s %016lX %016lX\n",
0591             netdev->name,
0592             netdev->state,
0593             dev_trans_start(netdev));
0594     }
0595 
0596     /* Print Registers */
0597     dev_info(&adapter->pdev->dev, "Register Dump\n");
0598     pr_info(" Register Name   Value\n");
0599     for (reginfo = (struct ixgbe_reg_info *)ixgbe_reg_info_tbl;
0600          reginfo->name; reginfo++) {
0601         ixgbe_regdump(hw, reginfo);
0602     }
0603 
0604     /* Print TX Ring Summary */
0605     if (!netdev || !netif_running(netdev))
0606         return;
0607 
0608     dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
0609     pr_info(" %s     %s              %s        %s\n",
0610         "Queue [NTU] [NTC] [bi(ntc)->dma  ]",
0611         "leng", "ntw", "timestamp");
0612     for (n = 0; n < adapter->num_tx_queues; n++) {
0613         ring = adapter->tx_ring[n];
0614         ixgbe_print_buffer(ring, n);
0615     }
0616 
0617     for (n = 0; n < adapter->num_xdp_queues; n++) {
0618         ring = adapter->xdp_ring[n];
0619         ixgbe_print_buffer(ring, n);
0620     }
0621 
0622     /* Print TX Rings */
0623     if (!netif_msg_tx_done(adapter))
0624         goto rx_ring_summary;
0625 
0626     dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
0627 
0628     /* Transmit Descriptor Formats
0629      *
0630      * 82598 Advanced Transmit Descriptor
0631      *   +--------------------------------------------------------------+
0632      * 0 |         Buffer Address [63:0]                                |
0633      *   +--------------------------------------------------------------+
0634      * 8 |  PAYLEN  | POPTS  | IDX | STA | DCMD  |DTYP |  RSV |  DTALEN |
0635      *   +--------------------------------------------------------------+
0636      *   63       46 45    40 39 36 35 32 31   24 23 20 19              0
0637      *
0638      * 82598 Advanced Transmit Descriptor (Write-Back Format)
0639      *   +--------------------------------------------------------------+
0640      * 0 |                          RSV [63:0]                          |
0641      *   +--------------------------------------------------------------+
0642      * 8 |            RSV           |  STA  |          NXTSEQ           |
0643      *   +--------------------------------------------------------------+
0644      *   63                       36 35   32 31                         0
0645      *
0646      * 82599+ Advanced Transmit Descriptor
0647      *   +--------------------------------------------------------------+
0648      * 0 |         Buffer Address [63:0]                                |
0649      *   +--------------------------------------------------------------+
0650      * 8 |PAYLEN  |POPTS|CC|IDX  |STA  |DCMD  |DTYP |MAC  |RSV  |DTALEN |
0651      *   +--------------------------------------------------------------+
0652      *   63     46 45 40 39 38 36 35 32 31  24 23 20 19 18 17 16 15     0
0653      *
0654      * 82599+ Advanced Transmit Descriptor (Write-Back Format)
0655      *   +--------------------------------------------------------------+
0656      * 0 |                          RSV [63:0]                          |
0657      *   +--------------------------------------------------------------+
0658      * 8 |            RSV           |  STA  |           RSV             |
0659      *   +--------------------------------------------------------------+
0660      *   63                       36 35   32 31                         0
0661      */
0662 
0663     for (n = 0; n < adapter->num_tx_queues; n++) {
0664         ring = adapter->tx_ring[n];
0665         pr_info("------------------------------------\n");
0666         pr_info("TX QUEUE INDEX = %d\n", ring->queue_index);
0667         pr_info("------------------------------------\n");
0668         pr_info("%s%s    %s              %s        %s          %s\n",
0669             "T [desc]     [address 63:0  ] ",
0670             "[PlPOIdStDDt Ln] [bi->dma       ] ",
0671             "leng", "ntw", "timestamp", "bi->skb");
0672 
0673         for (i = 0; ring->desc && (i < ring->count); i++) {
0674             tx_desc = IXGBE_TX_DESC(ring, i);
0675             tx_buffer = &ring->tx_buffer_info[i];
0676             u0 = (struct my_u0 *)tx_desc;
0677             if (dma_unmap_len(tx_buffer, len) > 0) {
0678                 const char *ring_desc;
0679 
0680                 if (i == ring->next_to_use &&
0681                     i == ring->next_to_clean)
0682                     ring_desc = " NTC/U";
0683                 else if (i == ring->next_to_use)
0684                     ring_desc = " NTU";
0685                 else if (i == ring->next_to_clean)
0686                     ring_desc = " NTC";
0687                 else
0688                     ring_desc = "";
0689                 pr_info("T [0x%03X]    %016llX %016llX %016llX %08X %p %016llX %p%s",
0690                     i,
0691                     le64_to_cpu((__force __le64)u0->a),
0692                     le64_to_cpu((__force __le64)u0->b),
0693                     (u64)dma_unmap_addr(tx_buffer, dma),
0694                     dma_unmap_len(tx_buffer, len),
0695                     tx_buffer->next_to_watch,
0696                     (u64)tx_buffer->time_stamp,
0697                     tx_buffer->skb,
0698                     ring_desc);
0699 
0700                 if (netif_msg_pktdata(adapter) &&
0701                     tx_buffer->skb)
0702                     print_hex_dump(KERN_INFO, "",
0703                         DUMP_PREFIX_ADDRESS, 16, 1,
0704                         tx_buffer->skb->data,
0705                         dma_unmap_len(tx_buffer, len),
0706                         true);
0707             }
0708         }
0709     }
0710 
0711     /* Print RX Rings Summary */
0712 rx_ring_summary:
0713     dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
0714     pr_info("Queue [NTU] [NTC]\n");
0715     for (n = 0; n < adapter->num_rx_queues; n++) {
0716         rx_ring = adapter->rx_ring[n];
0717         pr_info("%5d %5X %5X\n",
0718             n, rx_ring->next_to_use, rx_ring->next_to_clean);
0719     }
0720 
0721     /* Print RX Rings */
0722     if (!netif_msg_rx_status(adapter))
0723         return;
0724 
0725     dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
0726 
0727     /* Receive Descriptor Formats
0728      *
0729      * 82598 Advanced Receive Descriptor (Read) Format
0730      *    63                                           1        0
0731      *    +-----------------------------------------------------+
0732      *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
0733      *    +----------------------------------------------+------+
0734      *  8 |       Header Buffer Address [63:1]           |  DD  |
0735      *    +-----------------------------------------------------+
0736      *
0737      *
0738      * 82598 Advanced Receive Descriptor (Write-Back) Format
0739      *
0740      *   63       48 47    32 31  30      21 20 16 15   4 3     0
0741      *   +------------------------------------------------------+
0742      * 0 |       RSS Hash /  |SPH| HDR_LEN  | RSV |Packet|  RSS |
0743      *   | Packet   | IP     |   |          |     | Type | Type |
0744      *   | Checksum | Ident  |   |          |     |      |      |
0745      *   +------------------------------------------------------+
0746      * 8 | VLAN Tag | Length | Extended Error | Extended Status |
0747      *   +------------------------------------------------------+
0748      *   63       48 47    32 31            20 19               0
0749      *
0750      * 82599+ Advanced Receive Descriptor (Read) Format
0751      *    63                                           1        0
0752      *    +-----------------------------------------------------+
0753      *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
0754      *    +----------------------------------------------+------+
0755      *  8 |       Header Buffer Address [63:1]           |  DD  |
0756      *    +-----------------------------------------------------+
0757      *
0758      *
0759      * 82599+ Advanced Receive Descriptor (Write-Back) Format
0760      *
0761      *   63       48 47    32 31  30      21 20 17 16   4 3     0
0762      *   +------------------------------------------------------+
0763      * 0 |RSS / Frag Checksum|SPH| HDR_LEN  |RSC- |Packet|  RSS |
0764      *   |/ RTT / PCoE_PARAM |   |          | CNT | Type | Type |
0765      *   |/ Flow Dir Flt ID  |   |          |     |      |      |
0766      *   +------------------------------------------------------+
0767      * 8 | VLAN Tag | Length |Extended Error| Xtnd Status/NEXTP |
0768      *   +------------------------------------------------------+
0769      *   63       48 47    32 31          20 19                 0
0770      */
0771 
0772     for (n = 0; n < adapter->num_rx_queues; n++) {
0773         rx_ring = adapter->rx_ring[n];
0774         pr_info("------------------------------------\n");
0775         pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
0776         pr_info("------------------------------------\n");
0777         pr_info("%s%s%s\n",
0778             "R  [desc]      [ PktBuf     A0] ",
0779             "[  HeadBuf   DD] [bi->dma       ] [bi->skb       ] ",
0780             "<-- Adv Rx Read format");
0781         pr_info("%s%s%s\n",
0782             "RWB[desc]      [PcsmIpSHl PtRs] ",
0783             "[vl er S cks ln] ---------------- [bi->skb       ] ",
0784             "<-- Adv Rx Write-Back format");
0785 
0786         for (i = 0; i < rx_ring->count; i++) {
0787             const char *ring_desc;
0788 
0789             if (i == rx_ring->next_to_use)
0790                 ring_desc = " NTU";
0791             else if (i == rx_ring->next_to_clean)
0792                 ring_desc = " NTC";
0793             else
0794                 ring_desc = "";
0795 
0796             rx_buffer_info = &rx_ring->rx_buffer_info[i];
0797             rx_desc = IXGBE_RX_DESC(rx_ring, i);
0798             u0 = (struct my_u0 *)rx_desc;
0799             if (rx_desc->wb.upper.length) {
0800                 /* Descriptor Done */
0801                 pr_info("RWB[0x%03X]     %016llX %016llX ---------------- %p%s\n",
0802                     i,
0803                     le64_to_cpu((__force __le64)u0->a),
0804                     le64_to_cpu((__force __le64)u0->b),
0805                     rx_buffer_info->skb,
0806                     ring_desc);
0807             } else {
0808                 pr_info("R  [0x%03X]     %016llX %016llX %016llX %p%s\n",
0809                     i,
0810                     le64_to_cpu((__force __le64)u0->a),
0811                     le64_to_cpu((__force __le64)u0->b),
0812                     (u64)rx_buffer_info->dma,
0813                     rx_buffer_info->skb,
0814                     ring_desc);
0815 
0816                 if (netif_msg_pktdata(adapter) &&
0817                     rx_buffer_info->dma) {
0818                     print_hex_dump(KERN_INFO, "",
0819                        DUMP_PREFIX_ADDRESS, 16, 1,
0820                        page_address(rx_buffer_info->page) +
0821                             rx_buffer_info->page_offset,
0822                        ixgbe_rx_bufsz(rx_ring), true);
0823                 }
0824             }
0825         }
0826     }
0827 }
0828 
0829 static void ixgbe_release_hw_control(struct ixgbe_adapter *adapter)
0830 {
0831     u32 ctrl_ext;
0832 
0833     /* Let firmware take over control of h/w */
0834     ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
0835     IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
0836             ctrl_ext & ~IXGBE_CTRL_EXT_DRV_LOAD);
0837 }
0838 
0839 static void ixgbe_get_hw_control(struct ixgbe_adapter *adapter)
0840 {
0841     u32 ctrl_ext;
0842 
0843     /* Let firmware know the driver has taken over */
0844     ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
0845     IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT,
0846             ctrl_ext | IXGBE_CTRL_EXT_DRV_LOAD);
0847 }
0848 
0849 /**
0850  * ixgbe_set_ivar - set the IVAR registers, mapping interrupt causes to vectors
0851  * @adapter: pointer to adapter struct
0852  * @direction: 0 for Rx, 1 for Tx, -1 for other causes
0853  * @queue: queue to map the corresponding interrupt to
0854  * @msix_vector: the vector to map to the corresponding queue
0855  *
0856  */
0857 static void ixgbe_set_ivar(struct ixgbe_adapter *adapter, s8 direction,
0858                u8 queue, u8 msix_vector)
0859 {
0860     u32 ivar, index;
0861     struct ixgbe_hw *hw = &adapter->hw;
0862     switch (hw->mac.type) {
0863     case ixgbe_mac_82598EB:
0864         msix_vector |= IXGBE_IVAR_ALLOC_VAL;
0865         if (direction == -1)
0866             direction = 0;
0867         index = (((direction * 64) + queue) >> 2) & 0x1F;
0868         ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
0869         ivar &= ~(0xFF << (8 * (queue & 0x3)));
0870         ivar |= (msix_vector << (8 * (queue & 0x3)));
0871         IXGBE_WRITE_REG(hw, IXGBE_IVAR(index), ivar);
0872         break;
0873     case ixgbe_mac_82599EB:
0874     case ixgbe_mac_X540:
0875     case ixgbe_mac_X550:
0876     case ixgbe_mac_X550EM_x:
0877     case ixgbe_mac_x550em_a:
0878         if (direction == -1) {
0879             /* other causes */
0880             msix_vector |= IXGBE_IVAR_ALLOC_VAL;
0881             index = ((queue & 1) * 8);
0882             ivar = IXGBE_READ_REG(&adapter->hw, IXGBE_IVAR_MISC);
0883             ivar &= ~(0xFF << index);
0884             ivar |= (msix_vector << index);
0885             IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR_MISC, ivar);
0886             break;
0887         } else {
0888             /* tx or rx causes */
0889             msix_vector |= IXGBE_IVAR_ALLOC_VAL;
0890             index = ((16 * (queue & 1)) + (8 * direction));
0891             ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
0892             ivar &= ~(0xFF << index);
0893             ivar |= (msix_vector << index);
0894             IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), ivar);
0895             break;
0896         }
0897     default:
0898         break;
0899     }
0900 }
0901 
0902 void ixgbe_irq_rearm_queues(struct ixgbe_adapter *adapter,
0903                 u64 qmask)
0904 {
0905     u32 mask;
0906 
0907     switch (adapter->hw.mac.type) {
0908     case ixgbe_mac_82598EB:
0909         mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
0910         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
0911         break;
0912     case ixgbe_mac_82599EB:
0913     case ixgbe_mac_X540:
0914     case ixgbe_mac_X550:
0915     case ixgbe_mac_X550EM_x:
0916     case ixgbe_mac_x550em_a:
0917         mask = (qmask & 0xFFFFFFFF);
0918         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
0919         mask = (qmask >> 32);
0920         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
0921         break;
0922     default:
0923         break;
0924     }
0925 }
0926 
0927 static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
0928 {
0929     struct ixgbe_hw *hw = &adapter->hw;
0930     struct ixgbe_hw_stats *hwstats = &adapter->stats;
0931     int i;
0932     u32 data;
0933 
0934     if ((hw->fc.current_mode != ixgbe_fc_full) &&
0935         (hw->fc.current_mode != ixgbe_fc_rx_pause))
0936         return;
0937 
0938     switch (hw->mac.type) {
0939     case ixgbe_mac_82598EB:
0940         data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
0941         break;
0942     default:
0943         data = IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
0944     }
0945     hwstats->lxoffrxc += data;
0946 
0947     /* refill credits (no tx hang) if we received xoff */
0948     if (!data)
0949         return;
0950 
0951     for (i = 0; i < adapter->num_tx_queues; i++)
0952         clear_bit(__IXGBE_HANG_CHECK_ARMED,
0953               &adapter->tx_ring[i]->state);
0954 
0955     for (i = 0; i < adapter->num_xdp_queues; i++)
0956         clear_bit(__IXGBE_HANG_CHECK_ARMED,
0957               &adapter->xdp_ring[i]->state);
0958 }
0959 
0960 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
0961 {
0962     struct ixgbe_hw *hw = &adapter->hw;
0963     struct ixgbe_hw_stats *hwstats = &adapter->stats;
0964     u32 xoff[8] = {0};
0965     u8 tc;
0966     int i;
0967     bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
0968 
0969     if (adapter->ixgbe_ieee_pfc)
0970         pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
0971 
0972     if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED) || !pfc_en) {
0973         ixgbe_update_xoff_rx_lfc(adapter);
0974         return;
0975     }
0976 
0977     /* update stats for each tc, only valid with PFC enabled */
0978     for (i = 0; i < MAX_TX_PACKET_BUFFERS; i++) {
0979         u32 pxoffrxc;
0980 
0981         switch (hw->mac.type) {
0982         case ixgbe_mac_82598EB:
0983             pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
0984             break;
0985         default:
0986             pxoffrxc = IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
0987         }
0988         hwstats->pxoffrxc[i] += pxoffrxc;
0989         /* Get the TC for given UP */
0990         tc = netdev_get_prio_tc_map(adapter->netdev, i);
0991         xoff[tc] += pxoffrxc;
0992     }
0993 
0994     /* disarm tx queues that have received xoff frames */
0995     for (i = 0; i < adapter->num_tx_queues; i++) {
0996         struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
0997 
0998         tc = tx_ring->dcb_tc;
0999         if (xoff[tc])
1000             clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1001     }
1002 
1003     for (i = 0; i < adapter->num_xdp_queues; i++) {
1004         struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
1005 
1006         tc = xdp_ring->dcb_tc;
1007         if (xoff[tc])
1008             clear_bit(__IXGBE_HANG_CHECK_ARMED, &xdp_ring->state);
1009     }
1010 }
1011 
1012 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
1013 {
1014     return ring->stats.packets;
1015 }
1016 
1017 static u64 ixgbe_get_tx_pending(struct ixgbe_ring *ring)
1018 {
1019     unsigned int head, tail;
1020 
1021     head = ring->next_to_clean;
1022     tail = ring->next_to_use;
1023 
1024     return ((head <= tail) ? tail : tail + ring->count) - head;
1025 }
1026 
1027 static inline bool ixgbe_check_tx_hang(struct ixgbe_ring *tx_ring)
1028 {
1029     u32 tx_done = ixgbe_get_tx_completed(tx_ring);
1030     u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1031     u32 tx_pending = ixgbe_get_tx_pending(tx_ring);
1032 
1033     clear_check_for_tx_hang(tx_ring);
1034 
1035     /*
1036      * Check for a hung queue, but be thorough. This verifies
1037      * that a transmit has been completed since the previous
1038      * check AND there is at least one packet pending. The
1039      * ARMED bit is set to indicate a potential hang. The
1040      * bit is cleared if a pause frame is received to remove
1041      * false hang detection due to PFC or 802.3x frames. By
1042      * requiring this to fail twice we avoid races with
1043      * pfc clearing the ARMED bit and conditions where we
1044      * run the check_tx_hang logic with a transmit completion
1045      * pending but without time to complete it yet.
1046      */
1047     if (tx_done_old == tx_done && tx_pending)
1048         /* make sure it is true for two checks in a row */
1049         return test_and_set_bit(__IXGBE_HANG_CHECK_ARMED,
1050                     &tx_ring->state);
1051     /* update completed stats and continue */
1052     tx_ring->tx_stats.tx_done_old = tx_done;
1053     /* reset the countdown */
1054     clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
1055 
1056     return false;
1057 }
1058 
1059 /**
1060  * ixgbe_tx_timeout_reset - initiate reset due to Tx timeout
1061  * @adapter: driver private struct
1062  **/
1063 static void ixgbe_tx_timeout_reset(struct ixgbe_adapter *adapter)
1064 {
1065 
1066     /* Do the reset outside of interrupt context */
1067     if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
1068         set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
1069         e_warn(drv, "initiating reset due to tx timeout\n");
1070         ixgbe_service_event_schedule(adapter);
1071     }
1072 }
1073 
1074 /**
1075  * ixgbe_tx_maxrate - callback to set the maximum per-queue bitrate
1076  * @netdev: network interface device structure
1077  * @queue_index: Tx queue to set
1078  * @maxrate: desired maximum transmit bitrate
1079  **/
1080 static int ixgbe_tx_maxrate(struct net_device *netdev,
1081                 int queue_index, u32 maxrate)
1082 {
1083     struct ixgbe_adapter *adapter = netdev_priv(netdev);
1084     struct ixgbe_hw *hw = &adapter->hw;
1085     u32 bcnrc_val = ixgbe_link_mbps(adapter);
1086 
1087     if (!maxrate)
1088         return 0;
1089 
1090     /* Calculate the rate factor values to set */
1091     bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1092     bcnrc_val /= maxrate;
1093 
1094     /* clear everything but the rate factor */
1095     bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1096     IXGBE_RTTBCNRC_RF_DEC_MASK;
1097 
1098     /* enable the rate scheduler */
1099     bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1100 
1101     IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_index);
1102     IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1103 
1104     return 0;
1105 }
1106 
1107 /**
1108  * ixgbe_clean_tx_irq - Reclaim resources after transmit completes
1109  * @q_vector: structure containing interrupt and ring information
1110  * @tx_ring: tx ring to clean
1111  * @napi_budget: Used to determine if we are in netpoll
1112  **/
1113 static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
1114                    struct ixgbe_ring *tx_ring, int napi_budget)
1115 {
1116     struct ixgbe_adapter *adapter = q_vector->adapter;
1117     struct ixgbe_tx_buffer *tx_buffer;
1118     union ixgbe_adv_tx_desc *tx_desc;
1119     unsigned int total_bytes = 0, total_packets = 0, total_ipsec = 0;
1120     unsigned int budget = q_vector->tx.work_limit;
1121     unsigned int i = tx_ring->next_to_clean;
1122 
1123     if (test_bit(__IXGBE_DOWN, &adapter->state))
1124         return true;
1125 
1126     tx_buffer = &tx_ring->tx_buffer_info[i];
1127     tx_desc = IXGBE_TX_DESC(tx_ring, i);
1128     i -= tx_ring->count;
1129 
1130     do {
1131         union ixgbe_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
1132 
1133         /* if next_to_watch is not set then there is no work pending */
1134         if (!eop_desc)
1135             break;
1136 
1137         /* prevent any other reads prior to eop_desc */
1138         smp_rmb();
1139 
1140         /* if DD is not set pending work has not been completed */
1141         if (!(eop_desc->wb.status & cpu_to_le32(IXGBE_TXD_STAT_DD)))
1142             break;
1143 
1144         /* clear next_to_watch to prevent false hangs */
1145         tx_buffer->next_to_watch = NULL;
1146 
1147         /* update the statistics for this packet */
1148         total_bytes += tx_buffer->bytecount;
1149         total_packets += tx_buffer->gso_segs;
1150         if (tx_buffer->tx_flags & IXGBE_TX_FLAGS_IPSEC)
1151             total_ipsec++;
1152 
1153         /* free the skb */
1154         if (ring_is_xdp(tx_ring))
1155             xdp_return_frame(tx_buffer->xdpf);
1156         else
1157             napi_consume_skb(tx_buffer->skb, napi_budget);
1158 
1159         /* unmap skb header data */
1160         dma_unmap_single(tx_ring->dev,
1161                  dma_unmap_addr(tx_buffer, dma),
1162                  dma_unmap_len(tx_buffer, len),
1163                  DMA_TO_DEVICE);
1164 
1165         /* clear tx_buffer data */
1166         dma_unmap_len_set(tx_buffer, len, 0);
1167 
1168         /* unmap remaining buffers */
1169         while (tx_desc != eop_desc) {
1170             tx_buffer++;
1171             tx_desc++;
1172             i++;
1173             if (unlikely(!i)) {
1174                 i -= tx_ring->count;
1175                 tx_buffer = tx_ring->tx_buffer_info;
1176                 tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1177             }
1178 
1179             /* unmap any remaining paged data */
1180             if (dma_unmap_len(tx_buffer, len)) {
1181                 dma_unmap_page(tx_ring->dev,
1182                            dma_unmap_addr(tx_buffer, dma),
1183                            dma_unmap_len(tx_buffer, len),
1184                            DMA_TO_DEVICE);
1185                 dma_unmap_len_set(tx_buffer, len, 0);
1186             }
1187         }
1188 
1189         /* move us one more past the eop_desc for start of next pkt */
1190         tx_buffer++;
1191         tx_desc++;
1192         i++;
1193         if (unlikely(!i)) {
1194             i -= tx_ring->count;
1195             tx_buffer = tx_ring->tx_buffer_info;
1196             tx_desc = IXGBE_TX_DESC(tx_ring, 0);
1197         }
1198 
1199         /* issue prefetch for next Tx descriptor */
1200         prefetch(tx_desc);
1201 
1202         /* update budget accounting */
1203         budget--;
1204     } while (likely(budget));
1205 
1206     i += tx_ring->count;
1207     tx_ring->next_to_clean = i;
1208     u64_stats_update_begin(&tx_ring->syncp);
1209     tx_ring->stats.bytes += total_bytes;
1210     tx_ring->stats.packets += total_packets;
1211     u64_stats_update_end(&tx_ring->syncp);
1212     q_vector->tx.total_bytes += total_bytes;
1213     q_vector->tx.total_packets += total_packets;
1214     adapter->tx_ipsec += total_ipsec;
1215 
1216     if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
1217         /* schedule immediate reset if we believe we hung */
1218         struct ixgbe_hw *hw = &adapter->hw;
1219         e_err(drv, "Detected Tx Unit Hang %s\n"
1220             "  Tx Queue             <%d>\n"
1221             "  TDH, TDT             <%x>, <%x>\n"
1222             "  next_to_use          <%x>\n"
1223             "  next_to_clean        <%x>\n"
1224             "tx_buffer_info[next_to_clean]\n"
1225             "  time_stamp           <%lx>\n"
1226             "  jiffies              <%lx>\n",
1227             ring_is_xdp(tx_ring) ? "(XDP)" : "",
1228             tx_ring->queue_index,
1229             IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
1230             IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
1231             tx_ring->next_to_use, i,
1232             tx_ring->tx_buffer_info[i].time_stamp, jiffies);
1233 
1234         if (!ring_is_xdp(tx_ring))
1235             netif_stop_subqueue(tx_ring->netdev,
1236                         tx_ring->queue_index);
1237 
1238         e_info(probe,
1239                "tx hang %d detected on queue %d, resetting adapter\n",
1240             adapter->tx_timeout_count + 1, tx_ring->queue_index);
1241 
1242         /* schedule immediate reset if we believe we hung */
1243         ixgbe_tx_timeout_reset(adapter);
1244 
1245         /* the adapter is about to reset, no point in enabling stuff */
1246         return true;
1247     }
1248 
1249     if (ring_is_xdp(tx_ring))
1250         return !!budget;
1251 
1252     netdev_tx_completed_queue(txring_txq(tx_ring),
1253                   total_packets, total_bytes);
1254 
1255 #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
1256     if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
1257              (ixgbe_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
1258         /* Make sure that anybody stopping the queue after this
1259          * sees the new next_to_clean.
1260          */
1261         smp_mb();
1262         if (__netif_subqueue_stopped(tx_ring->netdev,
1263                          tx_ring->queue_index)
1264             && !test_bit(__IXGBE_DOWN, &adapter->state)) {
1265             netif_wake_subqueue(tx_ring->netdev,
1266                         tx_ring->queue_index);
1267             ++tx_ring->tx_stats.restart_queue;
1268         }
1269     }
1270 
1271     return !!budget;
1272 }
1273 
1274 #ifdef CONFIG_IXGBE_DCA
1275 static void ixgbe_update_tx_dca(struct ixgbe_adapter *adapter,
1276                 struct ixgbe_ring *tx_ring,
1277                 int cpu)
1278 {
1279     struct ixgbe_hw *hw = &adapter->hw;
1280     u32 txctrl = 0;
1281     u16 reg_offset;
1282 
1283     if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1284         txctrl = dca3_get_tag(tx_ring->dev, cpu);
1285 
1286     switch (hw->mac.type) {
1287     case ixgbe_mac_82598EB:
1288         reg_offset = IXGBE_DCA_TXCTRL(tx_ring->reg_idx);
1289         break;
1290     case ixgbe_mac_82599EB:
1291     case ixgbe_mac_X540:
1292         reg_offset = IXGBE_DCA_TXCTRL_82599(tx_ring->reg_idx);
1293         txctrl <<= IXGBE_DCA_TXCTRL_CPUID_SHIFT_82599;
1294         break;
1295     default:
1296         /* for unknown hardware do not write register */
1297         return;
1298     }
1299 
1300     /*
1301      * We can enable relaxed ordering for reads, but not writes when
1302      * DCA is enabled.  This is due to a known issue in some chipsets
1303      * which will cause the DCA tag to be cleared.
1304      */
1305     txctrl |= IXGBE_DCA_TXCTRL_DESC_RRO_EN |
1306           IXGBE_DCA_TXCTRL_DATA_RRO_EN |
1307           IXGBE_DCA_TXCTRL_DESC_DCA_EN;
1308 
1309     IXGBE_WRITE_REG(hw, reg_offset, txctrl);
1310 }
1311 
1312 static void ixgbe_update_rx_dca(struct ixgbe_adapter *adapter,
1313                 struct ixgbe_ring *rx_ring,
1314                 int cpu)
1315 {
1316     struct ixgbe_hw *hw = &adapter->hw;
1317     u32 rxctrl = 0;
1318     u8 reg_idx = rx_ring->reg_idx;
1319 
1320     if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1321         rxctrl = dca3_get_tag(rx_ring->dev, cpu);
1322 
1323     switch (hw->mac.type) {
1324     case ixgbe_mac_82599EB:
1325     case ixgbe_mac_X540:
1326         rxctrl <<= IXGBE_DCA_RXCTRL_CPUID_SHIFT_82599;
1327         break;
1328     default:
1329         break;
1330     }
1331 
1332     /*
1333      * We can enable relaxed ordering for reads, but not writes when
1334      * DCA is enabled.  This is due to a known issue in some chipsets
1335      * which will cause the DCA tag to be cleared.
1336      */
1337     rxctrl |= IXGBE_DCA_RXCTRL_DESC_RRO_EN |
1338           IXGBE_DCA_RXCTRL_DATA_DCA_EN |
1339           IXGBE_DCA_RXCTRL_DESC_DCA_EN;
1340 
1341     IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(reg_idx), rxctrl);
1342 }
1343 
1344 static void ixgbe_update_dca(struct ixgbe_q_vector *q_vector)
1345 {
1346     struct ixgbe_adapter *adapter = q_vector->adapter;
1347     struct ixgbe_ring *ring;
1348     int cpu = get_cpu();
1349 
1350     if (q_vector->cpu == cpu)
1351         goto out_no_update;
1352 
1353     ixgbe_for_each_ring(ring, q_vector->tx)
1354         ixgbe_update_tx_dca(adapter, ring, cpu);
1355 
1356     ixgbe_for_each_ring(ring, q_vector->rx)
1357         ixgbe_update_rx_dca(adapter, ring, cpu);
1358 
1359     q_vector->cpu = cpu;
1360 out_no_update:
1361     put_cpu();
1362 }
1363 
1364 static void ixgbe_setup_dca(struct ixgbe_adapter *adapter)
1365 {
1366     int i;
1367 
1368     /* always use CB2 mode, difference is masked in the CB driver */
1369     if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1370         IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1371                 IXGBE_DCA_CTRL_DCA_MODE_CB2);
1372     else
1373         IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1374                 IXGBE_DCA_CTRL_DCA_DISABLE);
1375 
1376     for (i = 0; i < adapter->num_q_vectors; i++) {
1377         adapter->q_vector[i]->cpu = -1;
1378         ixgbe_update_dca(adapter->q_vector[i]);
1379     }
1380 }
1381 
1382 static int __ixgbe_notify_dca(struct device *dev, void *data)
1383 {
1384     struct ixgbe_adapter *adapter = dev_get_drvdata(dev);
1385     unsigned long event = *(unsigned long *)data;
1386 
1387     if (!(adapter->flags & IXGBE_FLAG_DCA_CAPABLE))
1388         return 0;
1389 
1390     switch (event) {
1391     case DCA_PROVIDER_ADD:
1392         /* if we're already enabled, don't do it again */
1393         if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
1394             break;
1395         if (dca_add_requester(dev) == 0) {
1396             adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
1397             IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1398                     IXGBE_DCA_CTRL_DCA_MODE_CB2);
1399             break;
1400         }
1401         fallthrough; /* DCA is disabled. */
1402     case DCA_PROVIDER_REMOVE:
1403         if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
1404             dca_remove_requester(dev);
1405             adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
1406             IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
1407                     IXGBE_DCA_CTRL_DCA_DISABLE);
1408         }
1409         break;
1410     }
1411 
1412     return 0;
1413 }
1414 
1415 #endif /* CONFIG_IXGBE_DCA */
1416 
1417 #define IXGBE_RSS_L4_TYPES_MASK \
1418     ((1ul << IXGBE_RXDADV_RSSTYPE_IPV4_TCP) | \
1419      (1ul << IXGBE_RXDADV_RSSTYPE_IPV4_UDP) | \
1420      (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_TCP) | \
1421      (1ul << IXGBE_RXDADV_RSSTYPE_IPV6_UDP))
1422 
1423 static inline void ixgbe_rx_hash(struct ixgbe_ring *ring,
1424                  union ixgbe_adv_rx_desc *rx_desc,
1425                  struct sk_buff *skb)
1426 {
1427     u16 rss_type;
1428 
1429     if (!(ring->netdev->features & NETIF_F_RXHASH))
1430         return;
1431 
1432     rss_type = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info) &
1433            IXGBE_RXDADV_RSSTYPE_MASK;
1434 
1435     if (!rss_type)
1436         return;
1437 
1438     skb_set_hash(skb, le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
1439              (IXGBE_RSS_L4_TYPES_MASK & (1ul << rss_type)) ?
1440              PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
1441 }
1442 
1443 #ifdef IXGBE_FCOE
1444 /**
1445  * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
1446  * @ring: structure containing ring specific data
1447  * @rx_desc: advanced rx descriptor
1448  *
1449  * Returns : true if it is FCoE pkt
1450  */
1451 static inline bool ixgbe_rx_is_fcoe(struct ixgbe_ring *ring,
1452                     union ixgbe_adv_rx_desc *rx_desc)
1453 {
1454     __le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1455 
1456     return test_bit(__IXGBE_RX_FCOE, &ring->state) &&
1457            ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_ETQF_MASK)) ==
1458         (cpu_to_le16(IXGBE_ETQF_FILTER_FCOE <<
1459                  IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT)));
1460 }
1461 
1462 #endif /* IXGBE_FCOE */
1463 /**
1464  * ixgbe_rx_checksum - indicate in skb if hw indicated a good cksum
1465  * @ring: structure containing ring specific data
1466  * @rx_desc: current Rx descriptor being processed
1467  * @skb: skb currently being received and modified
1468  **/
1469 static inline void ixgbe_rx_checksum(struct ixgbe_ring *ring,
1470                      union ixgbe_adv_rx_desc *rx_desc,
1471                      struct sk_buff *skb)
1472 {
1473     __le16 pkt_info = rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
1474     bool encap_pkt = false;
1475 
1476     skb_checksum_none_assert(skb);
1477 
1478     /* Rx csum disabled */
1479     if (!(ring->netdev->features & NETIF_F_RXCSUM))
1480         return;
1481 
1482     /* check for VXLAN and Geneve packets */
1483     if (pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_VXLAN)) {
1484         encap_pkt = true;
1485         skb->encapsulation = 1;
1486     }
1487 
1488     /* if IP and error */
1489     if (ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_IPCS) &&
1490         ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_IPE)) {
1491         ring->rx_stats.csum_err++;
1492         return;
1493     }
1494 
1495     if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_L4CS))
1496         return;
1497 
1498     if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_TCPE)) {
1499         /*
1500          * 82599 errata, UDP frames with a 0 checksum can be marked as
1501          * checksum errors.
1502          */
1503         if ((pkt_info & cpu_to_le16(IXGBE_RXDADV_PKTTYPE_UDP)) &&
1504             test_bit(__IXGBE_RX_CSUM_UDP_ZERO_ERR, &ring->state))
1505             return;
1506 
1507         ring->rx_stats.csum_err++;
1508         return;
1509     }
1510 
1511     /* It must be a TCP or UDP packet with a valid checksum */
1512     skb->ip_summed = CHECKSUM_UNNECESSARY;
1513     if (encap_pkt) {
1514         if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_OUTERIPCS))
1515             return;
1516 
1517         if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_ERR_OUTERIPER)) {
1518             skb->ip_summed = CHECKSUM_NONE;
1519             return;
1520         }
1521         /* If we checked the outer header let the stack know */
1522         skb->csum_level = 1;
1523     }
1524 }
1525 
1526 static unsigned int ixgbe_rx_offset(struct ixgbe_ring *rx_ring)
1527 {
1528     return ring_uses_build_skb(rx_ring) ? IXGBE_SKB_PAD : 0;
1529 }
1530 
1531 static bool ixgbe_alloc_mapped_page(struct ixgbe_ring *rx_ring,
1532                     struct ixgbe_rx_buffer *bi)
1533 {
1534     struct page *page = bi->page;
1535     dma_addr_t dma;
1536 
1537     /* since we are recycling buffers we should seldom need to alloc */
1538     if (likely(page))
1539         return true;
1540 
1541     /* alloc new page for storage */
1542     page = dev_alloc_pages(ixgbe_rx_pg_order(rx_ring));
1543     if (unlikely(!page)) {
1544         rx_ring->rx_stats.alloc_rx_page_failed++;
1545         return false;
1546     }
1547 
1548     /* map page for use */
1549     dma = dma_map_page_attrs(rx_ring->dev, page, 0,
1550                  ixgbe_rx_pg_size(rx_ring),
1551                  DMA_FROM_DEVICE,
1552                  IXGBE_RX_DMA_ATTR);
1553 
1554     /*
1555      * if mapping failed free memory back to system since
1556      * there isn't much point in holding memory we can't use
1557      */
1558     if (dma_mapping_error(rx_ring->dev, dma)) {
1559         __free_pages(page, ixgbe_rx_pg_order(rx_ring));
1560 
1561         rx_ring->rx_stats.alloc_rx_page_failed++;
1562         return false;
1563     }
1564 
1565     bi->dma = dma;
1566     bi->page = page;
1567     bi->page_offset = rx_ring->rx_offset;
1568     page_ref_add(page, USHRT_MAX - 1);
1569     bi->pagecnt_bias = USHRT_MAX;
1570     rx_ring->rx_stats.alloc_rx_page++;
1571 
1572     return true;
1573 }
1574 
1575 /**
1576  * ixgbe_alloc_rx_buffers - Replace used receive buffers
1577  * @rx_ring: ring to place buffers on
1578  * @cleaned_count: number of buffers to replace
1579  **/
1580 void ixgbe_alloc_rx_buffers(struct ixgbe_ring *rx_ring, u16 cleaned_count)
1581 {
1582     union ixgbe_adv_rx_desc *rx_desc;
1583     struct ixgbe_rx_buffer *bi;
1584     u16 i = rx_ring->next_to_use;
1585     u16 bufsz;
1586 
1587     /* nothing to do */
1588     if (!cleaned_count)
1589         return;
1590 
1591     rx_desc = IXGBE_RX_DESC(rx_ring, i);
1592     bi = &rx_ring->rx_buffer_info[i];
1593     i -= rx_ring->count;
1594 
1595     bufsz = ixgbe_rx_bufsz(rx_ring);
1596 
1597     do {
1598         if (!ixgbe_alloc_mapped_page(rx_ring, bi))
1599             break;
1600 
1601         /* sync the buffer for use by the device */
1602         dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
1603                          bi->page_offset, bufsz,
1604                          DMA_FROM_DEVICE);
1605 
1606         /*
1607          * Refresh the desc even if buffer_addrs didn't change
1608          * because each write-back erases this info.
1609          */
1610         rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
1611 
1612         rx_desc++;
1613         bi++;
1614         i++;
1615         if (unlikely(!i)) {
1616             rx_desc = IXGBE_RX_DESC(rx_ring, 0);
1617             bi = rx_ring->rx_buffer_info;
1618             i -= rx_ring->count;
1619         }
1620 
1621         /* clear the length for the next_to_use descriptor */
1622         rx_desc->wb.upper.length = 0;
1623 
1624         cleaned_count--;
1625     } while (cleaned_count);
1626 
1627     i += rx_ring->count;
1628 
1629     if (rx_ring->next_to_use != i) {
1630         rx_ring->next_to_use = i;
1631 
1632         /* update next to alloc since we have filled the ring */
1633         rx_ring->next_to_alloc = i;
1634 
1635         /* Force memory writes to complete before letting h/w
1636          * know there are new descriptors to fetch.  (Only
1637          * applicable for weak-ordered memory model archs,
1638          * such as IA-64).
1639          */
1640         wmb();
1641         writel(i, rx_ring->tail);
1642     }
1643 }
1644 
1645 static void ixgbe_set_rsc_gso_size(struct ixgbe_ring *ring,
1646                    struct sk_buff *skb)
1647 {
1648     u16 hdr_len = skb_headlen(skb);
1649 
1650     /* set gso_size to avoid messing up TCP MSS */
1651     skb_shinfo(skb)->gso_size = DIV_ROUND_UP((skb->len - hdr_len),
1652                          IXGBE_CB(skb)->append_cnt);
1653     skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
1654 }
1655 
1656 static void ixgbe_update_rsc_stats(struct ixgbe_ring *rx_ring,
1657                    struct sk_buff *skb)
1658 {
1659     /* if append_cnt is 0 then frame is not RSC */
1660     if (!IXGBE_CB(skb)->append_cnt)
1661         return;
1662 
1663     rx_ring->rx_stats.rsc_count += IXGBE_CB(skb)->append_cnt;
1664     rx_ring->rx_stats.rsc_flush++;
1665 
1666     ixgbe_set_rsc_gso_size(rx_ring, skb);
1667 
1668     /* gso_size is computed using append_cnt so always clear it last */
1669     IXGBE_CB(skb)->append_cnt = 0;
1670 }
1671 
1672 /**
1673  * ixgbe_process_skb_fields - Populate skb header fields from Rx descriptor
1674  * @rx_ring: rx descriptor ring packet is being transacted on
1675  * @rx_desc: pointer to the EOP Rx descriptor
1676  * @skb: pointer to current skb being populated
1677  *
1678  * This function checks the ring, descriptor, and packet information in
1679  * order to populate the hash, checksum, VLAN, timestamp, protocol, and
1680  * other fields within the skb.
1681  **/
1682 void ixgbe_process_skb_fields(struct ixgbe_ring *rx_ring,
1683                   union ixgbe_adv_rx_desc *rx_desc,
1684                   struct sk_buff *skb)
1685 {
1686     struct net_device *dev = rx_ring->netdev;
1687     u32 flags = rx_ring->q_vector->adapter->flags;
1688 
1689     ixgbe_update_rsc_stats(rx_ring, skb);
1690 
1691     ixgbe_rx_hash(rx_ring, rx_desc, skb);
1692 
1693     ixgbe_rx_checksum(rx_ring, rx_desc, skb);
1694 
1695     if (unlikely(flags & IXGBE_FLAG_RX_HWTSTAMP_ENABLED))
1696         ixgbe_ptp_rx_hwtstamp(rx_ring, rx_desc, skb);
1697 
1698     if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
1699         ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_VP)) {
1700         u16 vid = le16_to_cpu(rx_desc->wb.upper.vlan);
1701         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
1702     }
1703 
1704     if (ixgbe_test_staterr(rx_desc, IXGBE_RXDADV_STAT_SECP))
1705         ixgbe_ipsec_rx(rx_ring, rx_desc, skb);
1706 
1707     /* record Rx queue, or update MACVLAN statistics */
1708     if (netif_is_ixgbe(dev))
1709         skb_record_rx_queue(skb, rx_ring->queue_index);
1710     else
1711         macvlan_count_rx(netdev_priv(dev), skb->len + ETH_HLEN, true,
1712                  false);
1713 
1714     skb->protocol = eth_type_trans(skb, dev);
1715 }
1716 
1717 void ixgbe_rx_skb(struct ixgbe_q_vector *q_vector,
1718           struct sk_buff *skb)
1719 {
1720     napi_gro_receive(&q_vector->napi, skb);
1721 }
1722 
1723 /**
1724  * ixgbe_is_non_eop - process handling of non-EOP buffers
1725  * @rx_ring: Rx ring being processed
1726  * @rx_desc: Rx descriptor for current buffer
1727  * @skb: Current socket buffer containing buffer in progress
1728  *
1729  * This function updates next to clean.  If the buffer is an EOP buffer
1730  * this function exits returning false, otherwise it will place the
1731  * sk_buff in the next buffer to be chained and return true indicating
1732  * that this is in fact a non-EOP buffer.
1733  **/
1734 static bool ixgbe_is_non_eop(struct ixgbe_ring *rx_ring,
1735                  union ixgbe_adv_rx_desc *rx_desc,
1736                  struct sk_buff *skb)
1737 {
1738     u32 ntc = rx_ring->next_to_clean + 1;
1739 
1740     /* fetch, update, and store next to clean */
1741     ntc = (ntc < rx_ring->count) ? ntc : 0;
1742     rx_ring->next_to_clean = ntc;
1743 
1744     prefetch(IXGBE_RX_DESC(rx_ring, ntc));
1745 
1746     /* update RSC append count if present */
1747     if (ring_is_rsc_enabled(rx_ring)) {
1748         __le32 rsc_enabled = rx_desc->wb.lower.lo_dword.data &
1749                      cpu_to_le32(IXGBE_RXDADV_RSCCNT_MASK);
1750 
1751         if (unlikely(rsc_enabled)) {
1752             u32 rsc_cnt = le32_to_cpu(rsc_enabled);
1753 
1754             rsc_cnt >>= IXGBE_RXDADV_RSCCNT_SHIFT;
1755             IXGBE_CB(skb)->append_cnt += rsc_cnt - 1;
1756 
1757             /* update ntc based on RSC value */
1758             ntc = le32_to_cpu(rx_desc->wb.upper.status_error);
1759             ntc &= IXGBE_RXDADV_NEXTP_MASK;
1760             ntc >>= IXGBE_RXDADV_NEXTP_SHIFT;
1761         }
1762     }
1763 
1764     /* if we are the last buffer then there is nothing else to do */
1765     if (likely(ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)))
1766         return false;
1767 
1768     /* place skb in next buffer to be received */
1769     rx_ring->rx_buffer_info[ntc].skb = skb;
1770     rx_ring->rx_stats.non_eop_descs++;
1771 
1772     return true;
1773 }
1774 
1775 /**
1776  * ixgbe_pull_tail - ixgbe specific version of skb_pull_tail
1777  * @rx_ring: rx descriptor ring packet is being transacted on
1778  * @skb: pointer to current skb being adjusted
1779  *
1780  * This function is an ixgbe specific version of __pskb_pull_tail.  The
1781  * main difference between this version and the original function is that
1782  * this function can make several assumptions about the state of things
1783  * that allow for significant optimizations versus the standard function.
1784  * As a result we can do things like drop a frag and maintain an accurate
1785  * truesize for the skb.
1786  */
1787 static void ixgbe_pull_tail(struct ixgbe_ring *rx_ring,
1788                 struct sk_buff *skb)
1789 {
1790     skb_frag_t *frag = &skb_shinfo(skb)->frags[0];
1791     unsigned char *va;
1792     unsigned int pull_len;
1793 
1794     /*
1795      * it is valid to use page_address instead of kmap since we are
1796      * working with pages allocated out of the lomem pool per
1797      * alloc_page(GFP_ATOMIC)
1798      */
1799     va = skb_frag_address(frag);
1800 
1801     /*
1802      * we need the header to contain the greater of either ETH_HLEN or
1803      * 60 bytes if the skb->len is less than 60 for skb_pad.
1804      */
1805     pull_len = eth_get_headlen(skb->dev, va, IXGBE_RX_HDR_SIZE);
1806 
1807     /* align pull length to size of long to optimize memcpy performance */
1808     skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
1809 
1810     /* update all of the pointers */
1811     skb_frag_size_sub(frag, pull_len);
1812     skb_frag_off_add(frag, pull_len);
1813     skb->data_len -= pull_len;
1814     skb->tail += pull_len;
1815 }
1816 
1817 /**
1818  * ixgbe_dma_sync_frag - perform DMA sync for first frag of SKB
1819  * @rx_ring: rx descriptor ring packet is being transacted on
1820  * @skb: pointer to current skb being updated
1821  *
1822  * This function provides a basic DMA sync up for the first fragment of an
1823  * skb.  The reason for doing this is that the first fragment cannot be
1824  * unmapped until we have reached the end of packet descriptor for a buffer
1825  * chain.
1826  */
1827 static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
1828                 struct sk_buff *skb)
1829 {
1830     if (ring_uses_build_skb(rx_ring)) {
1831         unsigned long mask = (unsigned long)ixgbe_rx_pg_size(rx_ring) - 1;
1832         unsigned long offset = (unsigned long)(skb->data) & mask;
1833 
1834         dma_sync_single_range_for_cpu(rx_ring->dev,
1835                           IXGBE_CB(skb)->dma,
1836                           offset,
1837                           skb_headlen(skb),
1838                           DMA_FROM_DEVICE);
1839     } else {
1840         skb_frag_t *frag = &skb_shinfo(skb)->frags[0];
1841 
1842         dma_sync_single_range_for_cpu(rx_ring->dev,
1843                           IXGBE_CB(skb)->dma,
1844                           skb_frag_off(frag),
1845                           skb_frag_size(frag),
1846                           DMA_FROM_DEVICE);
1847     }
1848 
1849     /* If the page was released, just unmap it. */
1850     if (unlikely(IXGBE_CB(skb)->page_released)) {
1851         dma_unmap_page_attrs(rx_ring->dev, IXGBE_CB(skb)->dma,
1852                      ixgbe_rx_pg_size(rx_ring),
1853                      DMA_FROM_DEVICE,
1854                      IXGBE_RX_DMA_ATTR);
1855     }
1856 }
1857 
1858 /**
1859  * ixgbe_cleanup_headers - Correct corrupted or empty headers
1860  * @rx_ring: rx descriptor ring packet is being transacted on
1861  * @rx_desc: pointer to the EOP Rx descriptor
1862  * @skb: pointer to current skb being fixed
1863  *
1864  * Check if the skb is valid in the XDP case it will be an error pointer.
1865  * Return true in this case to abort processing and advance to next
1866  * descriptor.
1867  *
1868  * Check for corrupted packet headers caused by senders on the local L2
1869  * embedded NIC switch not setting up their Tx Descriptors right.  These
1870  * should be very rare.
1871  *
1872  * Also address the case where we are pulling data in on pages only
1873  * and as such no data is present in the skb header.
1874  *
1875  * In addition if skb is not at least 60 bytes we need to pad it so that
1876  * it is large enough to qualify as a valid Ethernet frame.
1877  *
1878  * Returns true if an error was encountered and skb was freed.
1879  **/
1880 bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
1881                union ixgbe_adv_rx_desc *rx_desc,
1882                struct sk_buff *skb)
1883 {
1884     struct net_device *netdev = rx_ring->netdev;
1885 
1886     /* XDP packets use error pointer so abort at this point */
1887     if (IS_ERR(skb))
1888         return true;
1889 
1890     /* Verify netdev is present, and that packet does not have any
1891      * errors that would be unacceptable to the netdev.
1892      */
1893     if (!netdev ||
1894         (unlikely(ixgbe_test_staterr(rx_desc,
1895                      IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
1896          !(netdev->features & NETIF_F_RXALL)))) {
1897         dev_kfree_skb_any(skb);
1898         return true;
1899     }
1900 
1901     /* place header in linear portion of buffer */
1902     if (!skb_headlen(skb))
1903         ixgbe_pull_tail(rx_ring, skb);
1904 
1905 #ifdef IXGBE_FCOE
1906     /* do not attempt to pad FCoE Frames as this will disrupt DDP */
1907     if (ixgbe_rx_is_fcoe(rx_ring, rx_desc))
1908         return false;
1909 
1910 #endif
1911     /* if eth_skb_pad returns an error the skb was freed */
1912     if (eth_skb_pad(skb))
1913         return true;
1914 
1915     return false;
1916 }
1917 
1918 /**
1919  * ixgbe_reuse_rx_page - page flip buffer and store it back on the ring
1920  * @rx_ring: rx descriptor ring to store buffers on
1921  * @old_buff: donor buffer to have page reused
1922  *
1923  * Synchronizes page for reuse by the adapter
1924  **/
1925 static void ixgbe_reuse_rx_page(struct ixgbe_ring *rx_ring,
1926                 struct ixgbe_rx_buffer *old_buff)
1927 {
1928     struct ixgbe_rx_buffer *new_buff;
1929     u16 nta = rx_ring->next_to_alloc;
1930 
1931     new_buff = &rx_ring->rx_buffer_info[nta];
1932 
1933     /* update, and store next to alloc */
1934     nta++;
1935     rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
1936 
1937     /* Transfer page from old buffer to new buffer.
1938      * Move each member individually to avoid possible store
1939      * forwarding stalls and unnecessary copy of skb.
1940      */
1941     new_buff->dma       = old_buff->dma;
1942     new_buff->page      = old_buff->page;
1943     new_buff->page_offset   = old_buff->page_offset;
1944     new_buff->pagecnt_bias  = old_buff->pagecnt_bias;
1945 }
1946 
1947 static bool ixgbe_can_reuse_rx_page(struct ixgbe_rx_buffer *rx_buffer,
1948                     int rx_buffer_pgcnt)
1949 {
1950     unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
1951     struct page *page = rx_buffer->page;
1952 
1953     /* avoid re-using remote and pfmemalloc pages */
1954     if (!dev_page_is_reusable(page))
1955         return false;
1956 
1957 #if (PAGE_SIZE < 8192)
1958     /* if we are only owner of page we can reuse it */
1959     if (unlikely((rx_buffer_pgcnt - pagecnt_bias) > 1))
1960         return false;
1961 #else
1962     /* The last offset is a bit aggressive in that we assume the
1963      * worst case of FCoE being enabled and using a 3K buffer.
1964      * However this should have minimal impact as the 1K extra is
1965      * still less than one buffer in size.
1966      */
1967 #define IXGBE_LAST_OFFSET \
1968     (SKB_WITH_OVERHEAD(PAGE_SIZE) - IXGBE_RXBUFFER_3K)
1969     if (rx_buffer->page_offset > IXGBE_LAST_OFFSET)
1970         return false;
1971 #endif
1972 
1973     /* If we have drained the page fragment pool we need to update
1974      * the pagecnt_bias and page count so that we fully restock the
1975      * number of references the driver holds.
1976      */
1977     if (unlikely(pagecnt_bias == 1)) {
1978         page_ref_add(page, USHRT_MAX - 1);
1979         rx_buffer->pagecnt_bias = USHRT_MAX;
1980     }
1981 
1982     return true;
1983 }
1984 
1985 /**
1986  * ixgbe_add_rx_frag - Add contents of Rx buffer to sk_buff
1987  * @rx_ring: rx descriptor ring to transact packets on
1988  * @rx_buffer: buffer containing page to add
1989  * @skb: sk_buff to place the data into
1990  * @size: size of data in rx_buffer
1991  *
1992  * This function will add the data contained in rx_buffer->page to the skb.
1993  * This is done either through a direct copy if the data in the buffer is
1994  * less than the skb header size, otherwise it will just attach the page as
1995  * a frag to the skb.
1996  *
1997  * The function will then update the page offset if necessary and return
1998  * true if the buffer can be reused by the adapter.
1999  **/
2000 static void ixgbe_add_rx_frag(struct ixgbe_ring *rx_ring,
2001                   struct ixgbe_rx_buffer *rx_buffer,
2002                   struct sk_buff *skb,
2003                   unsigned int size)
2004 {
2005 #if (PAGE_SIZE < 8192)
2006     unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2007 #else
2008     unsigned int truesize = rx_ring->rx_offset ?
2009                 SKB_DATA_ALIGN(rx_ring->rx_offset + size) :
2010                 SKB_DATA_ALIGN(size);
2011 #endif
2012     skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
2013             rx_buffer->page_offset, size, truesize);
2014 #if (PAGE_SIZE < 8192)
2015     rx_buffer->page_offset ^= truesize;
2016 #else
2017     rx_buffer->page_offset += truesize;
2018 #endif
2019 }
2020 
2021 static struct ixgbe_rx_buffer *ixgbe_get_rx_buffer(struct ixgbe_ring *rx_ring,
2022                            union ixgbe_adv_rx_desc *rx_desc,
2023                            struct sk_buff **skb,
2024                            const unsigned int size,
2025                            int *rx_buffer_pgcnt)
2026 {
2027     struct ixgbe_rx_buffer *rx_buffer;
2028 
2029     rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
2030     *rx_buffer_pgcnt =
2031 #if (PAGE_SIZE < 8192)
2032         page_count(rx_buffer->page);
2033 #else
2034         0;
2035 #endif
2036     prefetchw(rx_buffer->page);
2037     *skb = rx_buffer->skb;
2038 
2039     /* Delay unmapping of the first packet. It carries the header
2040      * information, HW may still access the header after the writeback.
2041      * Only unmap it when EOP is reached
2042      */
2043     if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP)) {
2044         if (!*skb)
2045             goto skip_sync;
2046     } else {
2047         if (*skb)
2048             ixgbe_dma_sync_frag(rx_ring, *skb);
2049     }
2050 
2051     /* we are reusing so sync this buffer for CPU use */
2052     dma_sync_single_range_for_cpu(rx_ring->dev,
2053                       rx_buffer->dma,
2054                       rx_buffer->page_offset,
2055                       size,
2056                       DMA_FROM_DEVICE);
2057 skip_sync:
2058     rx_buffer->pagecnt_bias--;
2059 
2060     return rx_buffer;
2061 }
2062 
2063 static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
2064                 struct ixgbe_rx_buffer *rx_buffer,
2065                 struct sk_buff *skb,
2066                 int rx_buffer_pgcnt)
2067 {
2068     if (ixgbe_can_reuse_rx_page(rx_buffer, rx_buffer_pgcnt)) {
2069         /* hand second half of page back to the ring */
2070         ixgbe_reuse_rx_page(rx_ring, rx_buffer);
2071     } else {
2072         if (!IS_ERR(skb) && IXGBE_CB(skb)->dma == rx_buffer->dma) {
2073             /* the page has been released from the ring */
2074             IXGBE_CB(skb)->page_released = true;
2075         } else {
2076             /* we are not reusing the buffer so unmap it */
2077             dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
2078                          ixgbe_rx_pg_size(rx_ring),
2079                          DMA_FROM_DEVICE,
2080                          IXGBE_RX_DMA_ATTR);
2081         }
2082         __page_frag_cache_drain(rx_buffer->page,
2083                     rx_buffer->pagecnt_bias);
2084     }
2085 
2086     /* clear contents of rx_buffer */
2087     rx_buffer->page = NULL;
2088     rx_buffer->skb = NULL;
2089 }
2090 
2091 static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
2092                        struct ixgbe_rx_buffer *rx_buffer,
2093                        struct xdp_buff *xdp,
2094                        union ixgbe_adv_rx_desc *rx_desc)
2095 {
2096     unsigned int size = xdp->data_end - xdp->data;
2097 #if (PAGE_SIZE < 8192)
2098     unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2099 #else
2100     unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
2101                            xdp->data_hard_start);
2102 #endif
2103     struct sk_buff *skb;
2104 
2105     /* prefetch first cache line of first page */
2106     net_prefetch(xdp->data);
2107 
2108     /* Note, we get here by enabling legacy-rx via:
2109      *
2110      *    ethtool --set-priv-flags <dev> legacy-rx on
2111      *
2112      * In this mode, we currently get 0 extra XDP headroom as
2113      * opposed to having legacy-rx off, where we process XDP
2114      * packets going to stack via ixgbe_build_skb(). The latter
2115      * provides us currently with 192 bytes of headroom.
2116      *
2117      * For ixgbe_construct_skb() mode it means that the
2118      * xdp->data_meta will always point to xdp->data, since
2119      * the helper cannot expand the head. Should this ever
2120      * change in future for legacy-rx mode on, then lets also
2121      * add xdp->data_meta handling here.
2122      */
2123 
2124     /* allocate a skb to store the frags */
2125     skb = napi_alloc_skb(&rx_ring->q_vector->napi, IXGBE_RX_HDR_SIZE);
2126     if (unlikely(!skb))
2127         return NULL;
2128 
2129     if (size > IXGBE_RX_HDR_SIZE) {
2130         if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2131             IXGBE_CB(skb)->dma = rx_buffer->dma;
2132 
2133         skb_add_rx_frag(skb, 0, rx_buffer->page,
2134                 xdp->data - page_address(rx_buffer->page),
2135                 size, truesize);
2136 #if (PAGE_SIZE < 8192)
2137         rx_buffer->page_offset ^= truesize;
2138 #else
2139         rx_buffer->page_offset += truesize;
2140 #endif
2141     } else {
2142         memcpy(__skb_put(skb, size),
2143                xdp->data, ALIGN(size, sizeof(long)));
2144         rx_buffer->pagecnt_bias++;
2145     }
2146 
2147     return skb;
2148 }
2149 
2150 static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
2151                        struct ixgbe_rx_buffer *rx_buffer,
2152                        struct xdp_buff *xdp,
2153                        union ixgbe_adv_rx_desc *rx_desc)
2154 {
2155     unsigned int metasize = xdp->data - xdp->data_meta;
2156 #if (PAGE_SIZE < 8192)
2157     unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
2158 #else
2159     unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
2160                 SKB_DATA_ALIGN(xdp->data_end -
2161                            xdp->data_hard_start);
2162 #endif
2163     struct sk_buff *skb;
2164 
2165     /* Prefetch first cache line of first page. If xdp->data_meta
2166      * is unused, this points extactly as xdp->data, otherwise we
2167      * likely have a consumer accessing first few bytes of meta
2168      * data, and then actual data.
2169      */
2170     net_prefetch(xdp->data_meta);
2171 
2172     /* build an skb to around the page buffer */
2173     skb = napi_build_skb(xdp->data_hard_start, truesize);
2174     if (unlikely(!skb))
2175         return NULL;
2176 
2177     /* update pointers within the skb to store the data */
2178     skb_reserve(skb, xdp->data - xdp->data_hard_start);
2179     __skb_put(skb, xdp->data_end - xdp->data);
2180     if (metasize)
2181         skb_metadata_set(skb, metasize);
2182 
2183     /* record DMA address if this is the start of a chain of buffers */
2184     if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
2185         IXGBE_CB(skb)->dma = rx_buffer->dma;
2186 
2187     /* update buffer offset */
2188 #if (PAGE_SIZE < 8192)
2189     rx_buffer->page_offset ^= truesize;
2190 #else
2191     rx_buffer->page_offset += truesize;
2192 #endif
2193 
2194     return skb;
2195 }
2196 
2197 static struct sk_buff *ixgbe_run_xdp(struct ixgbe_adapter *adapter,
2198                      struct ixgbe_ring *rx_ring,
2199                      struct xdp_buff *xdp)
2200 {
2201     int err, result = IXGBE_XDP_PASS;
2202     struct bpf_prog *xdp_prog;
2203     struct ixgbe_ring *ring;
2204     struct xdp_frame *xdpf;
2205     u32 act;
2206 
2207     xdp_prog = READ_ONCE(rx_ring->xdp_prog);
2208 
2209     if (!xdp_prog)
2210         goto xdp_out;
2211 
2212     prefetchw(xdp->data_hard_start); /* xdp_frame write */
2213 
2214     act = bpf_prog_run_xdp(xdp_prog, xdp);
2215     switch (act) {
2216     case XDP_PASS:
2217         break;
2218     case XDP_TX:
2219         xdpf = xdp_convert_buff_to_frame(xdp);
2220         if (unlikely(!xdpf))
2221             goto out_failure;
2222         ring = ixgbe_determine_xdp_ring(adapter);
2223         if (static_branch_unlikely(&ixgbe_xdp_locking_key))
2224             spin_lock(&ring->tx_lock);
2225         result = ixgbe_xmit_xdp_ring(ring, xdpf);
2226         if (static_branch_unlikely(&ixgbe_xdp_locking_key))
2227             spin_unlock(&ring->tx_lock);
2228         if (result == IXGBE_XDP_CONSUMED)
2229             goto out_failure;
2230         break;
2231     case XDP_REDIRECT:
2232         err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
2233         if (err)
2234             goto out_failure;
2235         result = IXGBE_XDP_REDIR;
2236         break;
2237     default:
2238         bpf_warn_invalid_xdp_action(rx_ring->netdev, xdp_prog, act);
2239         fallthrough;
2240     case XDP_ABORTED:
2241 out_failure:
2242         trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
2243         fallthrough; /* handle aborts by dropping packet */
2244     case XDP_DROP:
2245         result = IXGBE_XDP_CONSUMED;
2246         break;
2247     }
2248 xdp_out:
2249     return ERR_PTR(-result);
2250 }
2251 
2252 static unsigned int ixgbe_rx_frame_truesize(struct ixgbe_ring *rx_ring,
2253                         unsigned int size)
2254 {
2255     unsigned int truesize;
2256 
2257 #if (PAGE_SIZE < 8192)
2258     truesize = ixgbe_rx_pg_size(rx_ring) / 2; /* Must be power-of-2 */
2259 #else
2260     truesize = rx_ring->rx_offset ?
2261         SKB_DATA_ALIGN(rx_ring->rx_offset + size) +
2262         SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) :
2263         SKB_DATA_ALIGN(size);
2264 #endif
2265     return truesize;
2266 }
2267 
2268 static void ixgbe_rx_buffer_flip(struct ixgbe_ring *rx_ring,
2269                  struct ixgbe_rx_buffer *rx_buffer,
2270                  unsigned int size)
2271 {
2272     unsigned int truesize = ixgbe_rx_frame_truesize(rx_ring, size);
2273 #if (PAGE_SIZE < 8192)
2274     rx_buffer->page_offset ^= truesize;
2275 #else
2276     rx_buffer->page_offset += truesize;
2277 #endif
2278 }
2279 
2280 /**
2281  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
2282  * @q_vector: structure containing interrupt and ring information
2283  * @rx_ring: rx descriptor ring to transact packets on
2284  * @budget: Total limit on number of packets to process
2285  *
2286  * This function provides a "bounce buffer" approach to Rx interrupt
2287  * processing.  The advantage to this is that on systems that have
2288  * expensive overhead for IOMMU access this provides a means of avoiding
2289  * it by maintaining the mapping of the page to the syste.
2290  *
2291  * Returns amount of work completed
2292  **/
2293 static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
2294                    struct ixgbe_ring *rx_ring,
2295                    const int budget)
2296 {
2297     unsigned int total_rx_bytes = 0, total_rx_packets = 0, frame_sz = 0;
2298     struct ixgbe_adapter *adapter = q_vector->adapter;
2299 #ifdef IXGBE_FCOE
2300     int ddp_bytes;
2301     unsigned int mss = 0;
2302 #endif /* IXGBE_FCOE */
2303     u16 cleaned_count = ixgbe_desc_unused(rx_ring);
2304     unsigned int offset = rx_ring->rx_offset;
2305     unsigned int xdp_xmit = 0;
2306     struct xdp_buff xdp;
2307 
2308     /* Frame size depend on rx_ring setup when PAGE_SIZE=4K */
2309 #if (PAGE_SIZE < 8192)
2310     frame_sz = ixgbe_rx_frame_truesize(rx_ring, 0);
2311 #endif
2312     xdp_init_buff(&xdp, frame_sz, &rx_ring->xdp_rxq);
2313 
2314     while (likely(total_rx_packets < budget)) {
2315         union ixgbe_adv_rx_desc *rx_desc;
2316         struct ixgbe_rx_buffer *rx_buffer;
2317         struct sk_buff *skb;
2318         int rx_buffer_pgcnt;
2319         unsigned int size;
2320 
2321         /* return some buffers to hardware, one at a time is too slow */
2322         if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
2323             ixgbe_alloc_rx_buffers(rx_ring, cleaned_count);
2324             cleaned_count = 0;
2325         }
2326 
2327         rx_desc = IXGBE_RX_DESC(rx_ring, rx_ring->next_to_clean);
2328         size = le16_to_cpu(rx_desc->wb.upper.length);
2329         if (!size)
2330             break;
2331 
2332         /* This memory barrier is needed to keep us from reading
2333          * any other fields out of the rx_desc until we know the
2334          * descriptor has been written back
2335          */
2336         dma_rmb();
2337 
2338         rx_buffer = ixgbe_get_rx_buffer(rx_ring, rx_desc, &skb, size, &rx_buffer_pgcnt);
2339 
2340         /* retrieve a buffer from the ring */
2341         if (!skb) {
2342             unsigned char *hard_start;
2343 
2344             hard_start = page_address(rx_buffer->page) +
2345                      rx_buffer->page_offset - offset;
2346             xdp_prepare_buff(&xdp, hard_start, offset, size, true);
2347             xdp_buff_clear_frags_flag(&xdp);
2348 #if (PAGE_SIZE > 4096)
2349             /* At larger PAGE_SIZE, frame_sz depend on len size */
2350             xdp.frame_sz = ixgbe_rx_frame_truesize(rx_ring, size);
2351 #endif
2352             skb = ixgbe_run_xdp(adapter, rx_ring, &xdp);
2353         }
2354 
2355         if (IS_ERR(skb)) {
2356             unsigned int xdp_res = -PTR_ERR(skb);
2357 
2358             if (xdp_res & (IXGBE_XDP_TX | IXGBE_XDP_REDIR)) {
2359                 xdp_xmit |= xdp_res;
2360                 ixgbe_rx_buffer_flip(rx_ring, rx_buffer, size);
2361             } else {
2362                 rx_buffer->pagecnt_bias++;
2363             }
2364             total_rx_packets++;
2365             total_rx_bytes += size;
2366         } else if (skb) {
2367             ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
2368         } else if (ring_uses_build_skb(rx_ring)) {
2369             skb = ixgbe_build_skb(rx_ring, rx_buffer,
2370                           &xdp, rx_desc);
2371         } else {
2372             skb = ixgbe_construct_skb(rx_ring, rx_buffer,
2373                           &xdp, rx_desc);
2374         }
2375 
2376         /* exit if we failed to retrieve a buffer */
2377         if (!skb) {
2378             rx_ring->rx_stats.alloc_rx_buff_failed++;
2379             rx_buffer->pagecnt_bias++;
2380             break;
2381         }
2382 
2383         ixgbe_put_rx_buffer(rx_ring, rx_buffer, skb, rx_buffer_pgcnt);
2384         cleaned_count++;
2385 
2386         /* place incomplete frames back on ring for completion */
2387         if (ixgbe_is_non_eop(rx_ring, rx_desc, skb))
2388             continue;
2389 
2390         /* verify the packet layout is correct */
2391         if (ixgbe_cleanup_headers(rx_ring, rx_desc, skb))
2392             continue;
2393 
2394         /* probably a little skewed due to removing CRC */
2395         total_rx_bytes += skb->len;
2396 
2397         /* populate checksum, timestamp, VLAN, and protocol */
2398         ixgbe_process_skb_fields(rx_ring, rx_desc, skb);
2399 
2400 #ifdef IXGBE_FCOE
2401         /* if ddp, not passing to ULD unless for FCP_RSP or error */
2402         if (ixgbe_rx_is_fcoe(rx_ring, rx_desc)) {
2403             ddp_bytes = ixgbe_fcoe_ddp(adapter, rx_desc, skb);
2404             /* include DDPed FCoE data */
2405             if (ddp_bytes > 0) {
2406                 if (!mss) {
2407                     mss = rx_ring->netdev->mtu -
2408                         sizeof(struct fcoe_hdr) -
2409                         sizeof(struct fc_frame_header) -
2410                         sizeof(struct fcoe_crc_eof);
2411                     if (mss > 512)
2412                         mss &= ~511;
2413                 }
2414                 total_rx_bytes += ddp_bytes;
2415                 total_rx_packets += DIV_ROUND_UP(ddp_bytes,
2416                                  mss);
2417             }
2418             if (!ddp_bytes) {
2419                 dev_kfree_skb_any(skb);
2420                 continue;
2421             }
2422         }
2423 
2424 #endif /* IXGBE_FCOE */
2425         ixgbe_rx_skb(q_vector, skb);
2426 
2427         /* update budget accounting */
2428         total_rx_packets++;
2429     }
2430 
2431     if (xdp_xmit & IXGBE_XDP_REDIR)
2432         xdp_do_flush_map();
2433 
2434     if (xdp_xmit & IXGBE_XDP_TX) {
2435         struct ixgbe_ring *ring = ixgbe_determine_xdp_ring(adapter);
2436 
2437         ixgbe_xdp_ring_update_tail_locked(ring);
2438     }
2439 
2440     u64_stats_update_begin(&rx_ring->syncp);
2441     rx_ring->stats.packets += total_rx_packets;
2442     rx_ring->stats.bytes += total_rx_bytes;
2443     u64_stats_update_end(&rx_ring->syncp);
2444     q_vector->rx.total_packets += total_rx_packets;
2445     q_vector->rx.total_bytes += total_rx_bytes;
2446 
2447     return total_rx_packets;
2448 }
2449 
2450 /**
2451  * ixgbe_configure_msix - Configure MSI-X hardware
2452  * @adapter: board private structure
2453  *
2454  * ixgbe_configure_msix sets up the hardware to properly generate MSI-X
2455  * interrupts.
2456  **/
2457 static void ixgbe_configure_msix(struct ixgbe_adapter *adapter)
2458 {
2459     struct ixgbe_q_vector *q_vector;
2460     int v_idx;
2461     u32 mask;
2462 
2463     /* Populate MSIX to EITR Select */
2464     if (adapter->num_vfs > 32) {
2465         u32 eitrsel = BIT(adapter->num_vfs - 32) - 1;
2466         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, eitrsel);
2467     }
2468 
2469     /*
2470      * Populate the IVAR table and set the ITR values to the
2471      * corresponding register.
2472      */
2473     for (v_idx = 0; v_idx < adapter->num_q_vectors; v_idx++) {
2474         struct ixgbe_ring *ring;
2475         q_vector = adapter->q_vector[v_idx];
2476 
2477         ixgbe_for_each_ring(ring, q_vector->rx)
2478             ixgbe_set_ivar(adapter, 0, ring->reg_idx, v_idx);
2479 
2480         ixgbe_for_each_ring(ring, q_vector->tx)
2481             ixgbe_set_ivar(adapter, 1, ring->reg_idx, v_idx);
2482 
2483         ixgbe_write_eitr(q_vector);
2484     }
2485 
2486     switch (adapter->hw.mac.type) {
2487     case ixgbe_mac_82598EB:
2488         ixgbe_set_ivar(adapter, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
2489                    v_idx);
2490         break;
2491     case ixgbe_mac_82599EB:
2492     case ixgbe_mac_X540:
2493     case ixgbe_mac_X550:
2494     case ixgbe_mac_X550EM_x:
2495     case ixgbe_mac_x550em_a:
2496         ixgbe_set_ivar(adapter, -1, 1, v_idx);
2497         break;
2498     default:
2499         break;
2500     }
2501     IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(v_idx), 1950);
2502 
2503     /* set up to autoclear timer, and the vectors */
2504     mask = IXGBE_EIMS_ENABLE_MASK;
2505     mask &= ~(IXGBE_EIMS_OTHER |
2506           IXGBE_EIMS_MAILBOX |
2507           IXGBE_EIMS_LSC);
2508 
2509     IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, mask);
2510 }
2511 
2512 /**
2513  * ixgbe_update_itr - update the dynamic ITR value based on statistics
2514  * @q_vector: structure containing interrupt and ring information
2515  * @ring_container: structure containing ring performance data
2516  *
2517  *      Stores a new ITR value based on packets and byte
2518  *      counts during the last interrupt.  The advantage of per interrupt
2519  *      computation is faster updates and more accurate ITR for the current
2520  *      traffic pattern.  Constants in this function were computed
2521  *      based on theoretical maximum wire speed and thresholds were set based
2522  *      on testing data as well as attempting to minimize response time
2523  *      while increasing bulk throughput.
2524  **/
2525 static void ixgbe_update_itr(struct ixgbe_q_vector *q_vector,
2526                  struct ixgbe_ring_container *ring_container)
2527 {
2528     unsigned int itr = IXGBE_ITR_ADAPTIVE_MIN_USECS |
2529                IXGBE_ITR_ADAPTIVE_LATENCY;
2530     unsigned int avg_wire_size, packets, bytes;
2531     unsigned long next_update = jiffies;
2532 
2533     /* If we don't have any rings just leave ourselves set for maximum
2534      * possible latency so we take ourselves out of the equation.
2535      */
2536     if (!ring_container->ring)
2537         return;
2538 
2539     /* If we didn't update within up to 1 - 2 jiffies we can assume
2540      * that either packets are coming in so slow there hasn't been
2541      * any work, or that there is so much work that NAPI is dealing
2542      * with interrupt moderation and we don't need to do anything.
2543      */
2544     if (time_after(next_update, ring_container->next_update))
2545         goto clear_counts;
2546 
2547     packets = ring_container->total_packets;
2548 
2549     /* We have no packets to actually measure against. This means
2550      * either one of the other queues on this vector is active or
2551      * we are a Tx queue doing TSO with too high of an interrupt rate.
2552      *
2553      * When this occurs just tick up our delay by the minimum value
2554      * and hope that this extra delay will prevent us from being called
2555      * without any work on our queue.
2556      */
2557     if (!packets) {
2558         itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2559         if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2560             itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2561         itr += ring_container->itr & IXGBE_ITR_ADAPTIVE_LATENCY;
2562         goto clear_counts;
2563     }
2564 
2565     bytes = ring_container->total_bytes;
2566 
2567     /* If packets are less than 4 or bytes are less than 9000 assume
2568      * insufficient data to use bulk rate limiting approach. We are
2569      * likely latency driven.
2570      */
2571     if (packets < 4 && bytes < 9000) {
2572         itr = IXGBE_ITR_ADAPTIVE_LATENCY;
2573         goto adjust_by_size;
2574     }
2575 
2576     /* Between 4 and 48 we can assume that our current interrupt delay
2577      * is only slightly too low. As such we should increase it by a small
2578      * fixed amount.
2579      */
2580     if (packets < 48) {
2581         itr = (q_vector->itr >> 2) + IXGBE_ITR_ADAPTIVE_MIN_INC;
2582         if (itr > IXGBE_ITR_ADAPTIVE_MAX_USECS)
2583             itr = IXGBE_ITR_ADAPTIVE_MAX_USECS;
2584         goto clear_counts;
2585     }
2586 
2587     /* Between 48 and 96 is our "goldilocks" zone where we are working
2588      * out "just right". Just report that our current ITR is good for us.
2589      */
2590     if (packets < 96) {
2591         itr = q_vector->itr >> 2;
2592         goto clear_counts;
2593     }
2594 
2595     /* If packet count is 96 or greater we are likely looking at a slight
2596      * overrun of the delay we want. Try halving our delay to see if that
2597      * will cut the number of packets in half per interrupt.
2598      */
2599     if (packets < 256) {
2600         itr = q_vector->itr >> 3;
2601         if (itr < IXGBE_ITR_ADAPTIVE_MIN_USECS)
2602             itr = IXGBE_ITR_ADAPTIVE_MIN_USECS;
2603         goto clear_counts;
2604     }
2605 
2606     /* The paths below assume we are dealing with a bulk ITR since number
2607      * of packets is 256 or greater. We are just going to have to compute
2608      * a value and try to bring the count under control, though for smaller
2609      * packet sizes there isn't much we can do as NAPI polling will likely
2610      * be kicking in sooner rather than later.
2611      */
2612     itr = IXGBE_ITR_ADAPTIVE_BULK;
2613 
2614 adjust_by_size:
2615     /* If packet counts are 256 or greater we can assume we have a gross
2616      * overestimation of what the rate should be. Instead of trying to fine
2617      * tune it just use the formula below to try and dial in an exact value
2618      * give the current packet size of the frame.
2619      */
2620     avg_wire_size = bytes / packets;
2621 
2622     /* The following is a crude approximation of:
2623      *  wmem_default / (size + overhead) = desired_pkts_per_int
2624      *  rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
2625      *  (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
2626      *
2627      * Assuming wmem_default is 212992 and overhead is 640 bytes per
2628      * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
2629      * formula down to
2630      *
2631      *  (170 * (size + 24)) / (size + 640) = ITR
2632      *
2633      * We first do some math on the packet size and then finally bitshift
2634      * by 8 after rounding up. We also have to account for PCIe link speed
2635      * difference as ITR scales based on this.
2636      */
2637     if (avg_wire_size <= 60) {
2638         /* Start at 50k ints/sec */
2639         avg_wire_size = 5120;
2640     } else if (avg_wire_size <= 316) {
2641         /* 50K ints/sec to 16K ints/sec */
2642         avg_wire_size *= 40;
2643         avg_wire_size += 2720;
2644     } else if (avg_wire_size <= 1084) {
2645         /* 16K ints/sec to 9.2K ints/sec */
2646         avg_wire_size *= 15;
2647         avg_wire_size += 11452;
2648     } else if (avg_wire_size < 1968) {
2649         /* 9.2K ints/sec to 8K ints/sec */
2650         avg_wire_size *= 5;
2651         avg_wire_size += 22420;
2652     } else {
2653         /* plateau at a limit of 8K ints/sec */
2654         avg_wire_size = 32256;
2655     }
2656 
2657     /* If we are in low latency mode half our delay which doubles the rate
2658      * to somewhere between 100K to 16K ints/sec
2659      */
2660     if (itr & IXGBE_ITR_ADAPTIVE_LATENCY)
2661         avg_wire_size >>= 1;
2662 
2663     /* Resultant value is 256 times larger than it needs to be. This
2664      * gives us room to adjust the value as needed to either increase
2665      * or decrease the value based on link speeds of 10G, 2.5G, 1G, etc.
2666      *
2667      * Use addition as we have already recorded the new latency flag
2668      * for the ITR value.
2669      */
2670     switch (q_vector->adapter->link_speed) {
2671     case IXGBE_LINK_SPEED_10GB_FULL:
2672     case IXGBE_LINK_SPEED_100_FULL:
2673     default:
2674         itr += DIV_ROUND_UP(avg_wire_size,
2675                     IXGBE_ITR_ADAPTIVE_MIN_INC * 256) *
2676                IXGBE_ITR_ADAPTIVE_MIN_INC;
2677         break;
2678     case IXGBE_LINK_SPEED_2_5GB_FULL:
2679     case IXGBE_LINK_SPEED_1GB_FULL:
2680     case IXGBE_LINK_SPEED_10_FULL:
2681         if (avg_wire_size > 8064)
2682             avg_wire_size = 8064;
2683         itr += DIV_ROUND_UP(avg_wire_size,
2684                     IXGBE_ITR_ADAPTIVE_MIN_INC * 64) *
2685                IXGBE_ITR_ADAPTIVE_MIN_INC;
2686         break;
2687     }
2688 
2689 clear_counts:
2690     /* write back value */
2691     ring_container->itr = itr;
2692 
2693     /* next update should occur within next jiffy */
2694     ring_container->next_update = next_update + 1;
2695 
2696     ring_container->total_bytes = 0;
2697     ring_container->total_packets = 0;
2698 }
2699 
2700 /**
2701  * ixgbe_write_eitr - write EITR register in hardware specific way
2702  * @q_vector: structure containing interrupt and ring information
2703  *
2704  * This function is made to be called by ethtool and by the driver
2705  * when it needs to update EITR registers at runtime.  Hardware
2706  * specific quirks/differences are taken care of here.
2707  */
2708 void ixgbe_write_eitr(struct ixgbe_q_vector *q_vector)
2709 {
2710     struct ixgbe_adapter *adapter = q_vector->adapter;
2711     struct ixgbe_hw *hw = &adapter->hw;
2712     int v_idx = q_vector->v_idx;
2713     u32 itr_reg = q_vector->itr & IXGBE_MAX_EITR;
2714 
2715     switch (adapter->hw.mac.type) {
2716     case ixgbe_mac_82598EB:
2717         /* must write high and low 16 bits to reset counter */
2718         itr_reg |= (itr_reg << 16);
2719         break;
2720     case ixgbe_mac_82599EB:
2721     case ixgbe_mac_X540:
2722     case ixgbe_mac_X550:
2723     case ixgbe_mac_X550EM_x:
2724     case ixgbe_mac_x550em_a:
2725         /*
2726          * set the WDIS bit to not clear the timer bits and cause an
2727          * immediate assertion of the interrupt
2728          */
2729         itr_reg |= IXGBE_EITR_CNT_WDIS;
2730         break;
2731     default:
2732         break;
2733     }
2734     IXGBE_WRITE_REG(hw, IXGBE_EITR(v_idx), itr_reg);
2735 }
2736 
2737 static void ixgbe_set_itr(struct ixgbe_q_vector *q_vector)
2738 {
2739     u32 new_itr;
2740 
2741     ixgbe_update_itr(q_vector, &q_vector->tx);
2742     ixgbe_update_itr(q_vector, &q_vector->rx);
2743 
2744     /* use the smallest value of new ITR delay calculations */
2745     new_itr = min(q_vector->rx.itr, q_vector->tx.itr);
2746 
2747     /* Clear latency flag if set, shift into correct position */
2748     new_itr &= ~IXGBE_ITR_ADAPTIVE_LATENCY;
2749     new_itr <<= 2;
2750 
2751     if (new_itr != q_vector->itr) {
2752         /* save the algorithm value here */
2753         q_vector->itr = new_itr;
2754 
2755         ixgbe_write_eitr(q_vector);
2756     }
2757 }
2758 
2759 /**
2760  * ixgbe_check_overtemp_subtask - check for over temperature
2761  * @adapter: pointer to adapter
2762  **/
2763 static void ixgbe_check_overtemp_subtask(struct ixgbe_adapter *adapter)
2764 {
2765     struct ixgbe_hw *hw = &adapter->hw;
2766     u32 eicr = adapter->interrupt_event;
2767     s32 rc;
2768 
2769     if (test_bit(__IXGBE_DOWN, &adapter->state))
2770         return;
2771 
2772     if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_EVENT))
2773         return;
2774 
2775     adapter->flags2 &= ~IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2776 
2777     switch (hw->device_id) {
2778     case IXGBE_DEV_ID_82599_T3_LOM:
2779         /*
2780          * Since the warning interrupt is for both ports
2781          * we don't have to check if:
2782          *  - This interrupt wasn't for our port.
2783          *  - We may have missed the interrupt so always have to
2784          *    check if we  got a LSC
2785          */
2786         if (!(eicr & IXGBE_EICR_GPI_SDP0_8259X) &&
2787             !(eicr & IXGBE_EICR_LSC))
2788             return;
2789 
2790         if (!(eicr & IXGBE_EICR_LSC) && hw->mac.ops.check_link) {
2791             u32 speed;
2792             bool link_up = false;
2793 
2794             hw->mac.ops.check_link(hw, &speed, &link_up, false);
2795 
2796             if (link_up)
2797                 return;
2798         }
2799 
2800         /* Check if this is not due to overtemp */
2801         if (hw->phy.ops.check_overtemp(hw) != IXGBE_ERR_OVERTEMP)
2802             return;
2803 
2804         break;
2805     case IXGBE_DEV_ID_X550EM_A_1G_T:
2806     case IXGBE_DEV_ID_X550EM_A_1G_T_L:
2807         rc = hw->phy.ops.check_overtemp(hw);
2808         if (rc != IXGBE_ERR_OVERTEMP)
2809             return;
2810         break;
2811     default:
2812         if (adapter->hw.mac.type >= ixgbe_mac_X540)
2813             return;
2814         if (!(eicr & IXGBE_EICR_GPI_SDP0(hw)))
2815             return;
2816         break;
2817     }
2818     e_crit(drv, "%s\n", ixgbe_overheat_msg);
2819 
2820     adapter->interrupt_event = 0;
2821 }
2822 
2823 static void ixgbe_check_fan_failure(struct ixgbe_adapter *adapter, u32 eicr)
2824 {
2825     struct ixgbe_hw *hw = &adapter->hw;
2826 
2827     if ((adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) &&
2828         (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2829         e_crit(probe, "Fan has stopped, replace the adapter\n");
2830         /* write to clear the interrupt */
2831         IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2832     }
2833 }
2834 
2835 static void ixgbe_check_overtemp_event(struct ixgbe_adapter *adapter, u32 eicr)
2836 {
2837     struct ixgbe_hw *hw = &adapter->hw;
2838 
2839     if (!(adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE))
2840         return;
2841 
2842     switch (adapter->hw.mac.type) {
2843     case ixgbe_mac_82599EB:
2844         /*
2845          * Need to check link state so complete overtemp check
2846          * on service task
2847          */
2848         if (((eicr & IXGBE_EICR_GPI_SDP0(hw)) ||
2849              (eicr & IXGBE_EICR_LSC)) &&
2850             (!test_bit(__IXGBE_DOWN, &adapter->state))) {
2851             adapter->interrupt_event = eicr;
2852             adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2853             ixgbe_service_event_schedule(adapter);
2854             return;
2855         }
2856         return;
2857     case ixgbe_mac_x550em_a:
2858         if (eicr & IXGBE_EICR_GPI_SDP0_X550EM_a) {
2859             adapter->interrupt_event = eicr;
2860             adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_EVENT;
2861             ixgbe_service_event_schedule(adapter);
2862             IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC,
2863                     IXGBE_EICR_GPI_SDP0_X550EM_a);
2864             IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICR,
2865                     IXGBE_EICR_GPI_SDP0_X550EM_a);
2866         }
2867         return;
2868     case ixgbe_mac_X550:
2869     case ixgbe_mac_X540:
2870         if (!(eicr & IXGBE_EICR_TS))
2871             return;
2872         break;
2873     default:
2874         return;
2875     }
2876 
2877     e_crit(drv, "%s\n", ixgbe_overheat_msg);
2878 }
2879 
2880 static inline bool ixgbe_is_sfp(struct ixgbe_hw *hw)
2881 {
2882     switch (hw->mac.type) {
2883     case ixgbe_mac_82598EB:
2884         if (hw->phy.type == ixgbe_phy_nl)
2885             return true;
2886         return false;
2887     case ixgbe_mac_82599EB:
2888     case ixgbe_mac_X550EM_x:
2889     case ixgbe_mac_x550em_a:
2890         switch (hw->mac.ops.get_media_type(hw)) {
2891         case ixgbe_media_type_fiber:
2892         case ixgbe_media_type_fiber_qsfp:
2893             return true;
2894         default:
2895             return false;
2896         }
2897     default:
2898         return false;
2899     }
2900 }
2901 
2902 static void ixgbe_check_sfp_event(struct ixgbe_adapter *adapter, u32 eicr)
2903 {
2904     struct ixgbe_hw *hw = &adapter->hw;
2905     u32 eicr_mask = IXGBE_EICR_GPI_SDP2(hw);
2906 
2907     if (!ixgbe_is_sfp(hw))
2908         return;
2909 
2910     /* Later MAC's use different SDP */
2911     if (hw->mac.type >= ixgbe_mac_X540)
2912         eicr_mask = IXGBE_EICR_GPI_SDP0_X540;
2913 
2914     if (eicr & eicr_mask) {
2915         /* Clear the interrupt */
2916         IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr_mask);
2917         if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2918             adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
2919             adapter->sfp_poll_time = 0;
2920             ixgbe_service_event_schedule(adapter);
2921         }
2922     }
2923 
2924     if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
2925         (eicr & IXGBE_EICR_GPI_SDP1(hw))) {
2926         /* Clear the interrupt */
2927         IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1(hw));
2928         if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2929             adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
2930             ixgbe_service_event_schedule(adapter);
2931         }
2932     }
2933 }
2934 
2935 static void ixgbe_check_lsc(struct ixgbe_adapter *adapter)
2936 {
2937     struct ixgbe_hw *hw = &adapter->hw;
2938 
2939     adapter->lsc_int++;
2940     adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2941     adapter->link_check_timeout = jiffies;
2942     if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
2943         IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC);
2944         IXGBE_WRITE_FLUSH(hw);
2945         ixgbe_service_event_schedule(adapter);
2946     }
2947 }
2948 
2949 static inline void ixgbe_irq_enable_queues(struct ixgbe_adapter *adapter,
2950                        u64 qmask)
2951 {
2952     u32 mask;
2953     struct ixgbe_hw *hw = &adapter->hw;
2954 
2955     switch (hw->mac.type) {
2956     case ixgbe_mac_82598EB:
2957         mask = (IXGBE_EIMS_RTX_QUEUE & qmask);
2958         IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
2959         break;
2960     case ixgbe_mac_82599EB:
2961     case ixgbe_mac_X540:
2962     case ixgbe_mac_X550:
2963     case ixgbe_mac_X550EM_x:
2964     case ixgbe_mac_x550em_a:
2965         mask = (qmask & 0xFFFFFFFF);
2966         if (mask)
2967             IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
2968         mask = (qmask >> 32);
2969         if (mask)
2970             IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
2971         break;
2972     default:
2973         break;
2974     }
2975     /* skip the flush */
2976 }
2977 
2978 /**
2979  * ixgbe_irq_enable - Enable default interrupt generation settings
2980  * @adapter: board private structure
2981  * @queues: enable irqs for queues
2982  * @flush: flush register write
2983  **/
2984 static inline void ixgbe_irq_enable(struct ixgbe_adapter *adapter, bool queues,
2985                     bool flush)
2986 {
2987     struct ixgbe_hw *hw = &adapter->hw;
2988     u32 mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
2989 
2990     /* don't reenable LSC while waiting for link */
2991     if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
2992         mask &= ~IXGBE_EIMS_LSC;
2993 
2994     if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE)
2995         switch (adapter->hw.mac.type) {
2996         case ixgbe_mac_82599EB:
2997             mask |= IXGBE_EIMS_GPI_SDP0(hw);
2998             break;
2999         case ixgbe_mac_X540:
3000         case ixgbe_mac_X550:
3001         case ixgbe_mac_X550EM_x:
3002         case ixgbe_mac_x550em_a:
3003             mask |= IXGBE_EIMS_TS;
3004             break;
3005         default:
3006             break;
3007         }
3008     if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
3009         mask |= IXGBE_EIMS_GPI_SDP1(hw);
3010     switch (adapter->hw.mac.type) {
3011     case ixgbe_mac_82599EB:
3012         mask |= IXGBE_EIMS_GPI_SDP1(hw);
3013         mask |= IXGBE_EIMS_GPI_SDP2(hw);
3014         fallthrough;
3015     case ixgbe_mac_X540:
3016     case ixgbe_mac_X550:
3017     case ixgbe_mac_X550EM_x:
3018     case ixgbe_mac_x550em_a:
3019         if (adapter->hw.device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
3020             adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP ||
3021             adapter->hw.device_id == IXGBE_DEV_ID_X550EM_A_SFP_N)
3022             mask |= IXGBE_EIMS_GPI_SDP0(&adapter->hw);
3023         if (adapter->hw.phy.type == ixgbe_phy_x550em_ext_t)
3024             mask |= IXGBE_EICR_GPI_SDP0_X540;
3025         mask |= IXGBE_EIMS_ECC;
3026         mask |= IXGBE_EIMS_MAILBOX;
3027         break;
3028     default:
3029         break;
3030     }
3031 
3032     if ((adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) &&
3033         !(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
3034         mask |= IXGBE_EIMS_FLOW_DIR;
3035 
3036     IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, mask);
3037     if (queues)
3038         ixgbe_irq_enable_queues(adapter, ~0);
3039     if (flush)
3040         IXGBE_WRITE_FLUSH(&adapter->hw);
3041 }
3042 
3043 static irqreturn_t ixgbe_msix_other(int irq, void *data)
3044 {
3045     struct ixgbe_adapter *adapter = data;
3046     struct ixgbe_hw *hw = &adapter->hw;
3047     u32 eicr;
3048 
3049     /*
3050      * Workaround for Silicon errata.  Use clear-by-write instead
3051      * of clear-by-read.  Reading with EICS will return the
3052      * interrupt causes without clearing, which later be done
3053      * with the write to EICR.
3054      */
3055     eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
3056 
3057     /* The lower 16bits of the EICR register are for the queue interrupts
3058      * which should be masked here in order to not accidentally clear them if
3059      * the bits are high when ixgbe_msix_other is called. There is a race
3060      * condition otherwise which results in possible performance loss
3061      * especially if the ixgbe_msix_other interrupt is triggering
3062      * consistently (as it would when PPS is turned on for the X540 device)
3063      */
3064     eicr &= 0xFFFF0000;
3065 
3066     IXGBE_WRITE_REG(hw, IXGBE_EICR, eicr);
3067 
3068     if (eicr & IXGBE_EICR_LSC)
3069         ixgbe_check_lsc(adapter);
3070 
3071     if (eicr & IXGBE_EICR_MAILBOX)
3072         ixgbe_msg_task(adapter);
3073 
3074     switch (hw->mac.type) {
3075     case ixgbe_mac_82599EB:
3076     case ixgbe_mac_X540:
3077     case ixgbe_mac_X550:
3078     case ixgbe_mac_X550EM_x:
3079     case ixgbe_mac_x550em_a:
3080         if (hw->phy.type == ixgbe_phy_x550em_ext_t &&
3081             (eicr & IXGBE_EICR_GPI_SDP0_X540)) {
3082             adapter->flags2 |= IXGBE_FLAG2_PHY_INTERRUPT;
3083             ixgbe_service_event_schedule(adapter);
3084             IXGBE_WRITE_REG(hw, IXGBE_EICR,
3085                     IXGBE_EICR_GPI_SDP0_X540);
3086         }
3087         if (eicr & IXGBE_EICR_ECC) {
3088             e_info(link, "Received ECC Err, initiating reset\n");
3089             set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3090             ixgbe_service_event_schedule(adapter);
3091             IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3092         }
3093         /* Handle Flow Director Full threshold interrupt */
3094         if (eicr & IXGBE_EICR_FLOW_DIR) {
3095             int reinit_count = 0;
3096             int i;
3097             for (i = 0; i < adapter->num_tx_queues; i++) {
3098                 struct ixgbe_ring *ring = adapter->tx_ring[i];
3099                 if (test_and_clear_bit(__IXGBE_TX_FDIR_INIT_DONE,
3100                                &ring->state))
3101                     reinit_count++;
3102             }
3103             if (reinit_count) {
3104                 /* no more flow director interrupts until after init */
3105                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_FLOW_DIR);
3106                 adapter->flags2 |= IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
3107                 ixgbe_service_event_schedule(adapter);
3108             }
3109         }
3110         ixgbe_check_sfp_event(adapter, eicr);
3111         ixgbe_check_overtemp_event(adapter, eicr);
3112         break;
3113     default:
3114         break;
3115     }
3116 
3117     ixgbe_check_fan_failure(adapter, eicr);
3118 
3119     if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3120         ixgbe_ptp_check_pps_event(adapter);
3121 
3122     /* re-enable the original interrupt state, no lsc, no queues */
3123     if (!test_bit(__IXGBE_DOWN, &adapter->state))
3124         ixgbe_irq_enable(adapter, false, false);
3125 
3126     return IRQ_HANDLED;
3127 }
3128 
3129 static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
3130 {
3131     struct ixgbe_q_vector *q_vector = data;
3132 
3133     /* EIAM disabled interrupts (on this vector) for us */
3134 
3135     if (q_vector->rx.ring || q_vector->tx.ring)
3136         napi_schedule_irqoff(&q_vector->napi);
3137 
3138     return IRQ_HANDLED;
3139 }
3140 
3141 /**
3142  * ixgbe_poll - NAPI Rx polling callback
3143  * @napi: structure for representing this polling device
3144  * @budget: how many packets driver is allowed to clean
3145  *
3146  * This function is used for legacy and MSI, NAPI mode
3147  **/
3148 int ixgbe_poll(struct napi_struct *napi, int budget)
3149 {
3150     struct ixgbe_q_vector *q_vector =
3151                 container_of(napi, struct ixgbe_q_vector, napi);
3152     struct ixgbe_adapter *adapter = q_vector->adapter;
3153     struct ixgbe_ring *ring;
3154     int per_ring_budget, work_done = 0;
3155     bool clean_complete = true;
3156 
3157 #ifdef CONFIG_IXGBE_DCA
3158     if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
3159         ixgbe_update_dca(q_vector);
3160 #endif
3161 
3162     ixgbe_for_each_ring(ring, q_vector->tx) {
3163         bool wd = ring->xsk_pool ?
3164               ixgbe_clean_xdp_tx_irq(q_vector, ring, budget) :
3165               ixgbe_clean_tx_irq(q_vector, ring, budget);
3166 
3167         if (!wd)
3168             clean_complete = false;
3169     }
3170 
3171     /* Exit if we are called by netpoll */
3172     if (budget <= 0)
3173         return budget;
3174 
3175     /* attempt to distribute budget to each queue fairly, but don't allow
3176      * the budget to go below 1 because we'll exit polling */
3177     if (q_vector->rx.count > 1)
3178         per_ring_budget = max(budget/q_vector->rx.count, 1);
3179     else
3180         per_ring_budget = budget;
3181 
3182     ixgbe_for_each_ring(ring, q_vector->rx) {
3183         int cleaned = ring->xsk_pool ?
3184                   ixgbe_clean_rx_irq_zc(q_vector, ring,
3185                             per_ring_budget) :
3186                   ixgbe_clean_rx_irq(q_vector, ring,
3187                          per_ring_budget);
3188 
3189         work_done += cleaned;
3190         if (cleaned >= per_ring_budget)
3191             clean_complete = false;
3192     }
3193 
3194     /* If all work not completed, return budget and keep polling */
3195     if (!clean_complete)
3196         return budget;
3197 
3198     /* all work done, exit the polling mode */
3199     if (likely(napi_complete_done(napi, work_done))) {
3200         if (adapter->rx_itr_setting & 1)
3201             ixgbe_set_itr(q_vector);
3202         if (!test_bit(__IXGBE_DOWN, &adapter->state))
3203             ixgbe_irq_enable_queues(adapter,
3204                         BIT_ULL(q_vector->v_idx));
3205     }
3206 
3207     return min(work_done, budget - 1);
3208 }
3209 
3210 /**
3211  * ixgbe_request_msix_irqs - Initialize MSI-X interrupts
3212  * @adapter: board private structure
3213  *
3214  * ixgbe_request_msix_irqs allocates MSI-X vectors and requests
3215  * interrupts from the kernel.
3216  **/
3217 static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
3218 {
3219     struct net_device *netdev = adapter->netdev;
3220     unsigned int ri = 0, ti = 0;
3221     int vector, err;
3222 
3223     for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3224         struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3225         struct msix_entry *entry = &adapter->msix_entries[vector];
3226 
3227         if (q_vector->tx.ring && q_vector->rx.ring) {
3228             snprintf(q_vector->name, sizeof(q_vector->name),
3229                  "%s-TxRx-%u", netdev->name, ri++);
3230             ti++;
3231         } else if (q_vector->rx.ring) {
3232             snprintf(q_vector->name, sizeof(q_vector->name),
3233                  "%s-rx-%u", netdev->name, ri++);
3234         } else if (q_vector->tx.ring) {
3235             snprintf(q_vector->name, sizeof(q_vector->name),
3236                  "%s-tx-%u", netdev->name, ti++);
3237         } else {
3238             /* skip this unused q_vector */
3239             continue;
3240         }
3241         err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
3242                   q_vector->name, q_vector);
3243         if (err) {
3244             e_err(probe, "request_irq failed for MSIX interrupt "
3245                   "Error: %d\n", err);
3246             goto free_queue_irqs;
3247         }
3248         /* If Flow Director is enabled, set interrupt affinity */
3249         if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3250             /* assign the mask for this irq */
3251             irq_update_affinity_hint(entry->vector,
3252                          &q_vector->affinity_mask);
3253         }
3254     }
3255 
3256     err = request_irq(adapter->msix_entries[vector].vector,
3257               ixgbe_msix_other, 0, netdev->name, adapter);
3258     if (err) {
3259         e_err(probe, "request_irq for msix_other failed: %d\n", err);
3260         goto free_queue_irqs;
3261     }
3262 
3263     return 0;
3264 
3265 free_queue_irqs:
3266     while (vector) {
3267         vector--;
3268         irq_update_affinity_hint(adapter->msix_entries[vector].vector,
3269                      NULL);
3270         free_irq(adapter->msix_entries[vector].vector,
3271              adapter->q_vector[vector]);
3272     }
3273     adapter->flags &= ~IXGBE_FLAG_MSIX_ENABLED;
3274     pci_disable_msix(adapter->pdev);
3275     kfree(adapter->msix_entries);
3276     adapter->msix_entries = NULL;
3277     return err;
3278 }
3279 
3280 /**
3281  * ixgbe_intr - legacy mode Interrupt Handler
3282  * @irq: interrupt number
3283  * @data: pointer to a network interface device structure
3284  **/
3285 static irqreturn_t ixgbe_intr(int irq, void *data)
3286 {
3287     struct ixgbe_adapter *adapter = data;
3288     struct ixgbe_hw *hw = &adapter->hw;
3289     struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3290     u32 eicr;
3291 
3292     /*
3293      * Workaround for silicon errata #26 on 82598.  Mask the interrupt
3294      * before the read of EICR.
3295      */
3296     IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_IRQ_CLEAR_MASK);
3297 
3298     /* for NAPI, using EIAM to auto-mask tx/rx interrupt bits on read
3299      * therefore no explicit interrupt disable is necessary */
3300     eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
3301     if (!eicr) {
3302         /*
3303          * shared interrupt alert!
3304          * make sure interrupts are enabled because the read will
3305          * have disabled interrupts due to EIAM
3306          * finish the workaround of silicon errata on 82598.  Unmask
3307          * the interrupt that we masked before the EICR read.
3308          */
3309         if (!test_bit(__IXGBE_DOWN, &adapter->state))
3310             ixgbe_irq_enable(adapter, true, true);
3311         return IRQ_NONE;    /* Not our interrupt */
3312     }
3313 
3314     if (eicr & IXGBE_EICR_LSC)
3315         ixgbe_check_lsc(adapter);
3316 
3317     switch (hw->mac.type) {
3318     case ixgbe_mac_82599EB:
3319         ixgbe_check_sfp_event(adapter, eicr);
3320         fallthrough;
3321     case ixgbe_mac_X540:
3322     case ixgbe_mac_X550:
3323     case ixgbe_mac_X550EM_x:
3324     case ixgbe_mac_x550em_a:
3325         if (eicr & IXGBE_EICR_ECC) {
3326             e_info(link, "Received ECC Err, initiating reset\n");
3327             set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
3328             ixgbe_service_event_schedule(adapter);
3329             IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
3330         }
3331         ixgbe_check_overtemp_event(adapter, eicr);
3332         break;
3333     default:
3334         break;
3335     }
3336 
3337     ixgbe_check_fan_failure(adapter, eicr);
3338     if (unlikely(eicr & IXGBE_EICR_TIMESYNC))
3339         ixgbe_ptp_check_pps_event(adapter);
3340 
3341     /* would disable interrupts here but EIAM disabled it */
3342     napi_schedule_irqoff(&q_vector->napi);
3343 
3344     /*
3345      * re-enable link(maybe) and non-queue interrupts, no flush.
3346      * ixgbe_poll will re-enable the queue interrupts
3347      */
3348     if (!test_bit(__IXGBE_DOWN, &adapter->state))
3349         ixgbe_irq_enable(adapter, false, false);
3350 
3351     return IRQ_HANDLED;
3352 }
3353 
3354 /**
3355  * ixgbe_request_irq - initialize interrupts
3356  * @adapter: board private structure
3357  *
3358  * Attempts to configure interrupts using the best available
3359  * capabilities of the hardware and kernel.
3360  **/
3361 static int ixgbe_request_irq(struct ixgbe_adapter *adapter)
3362 {
3363     struct net_device *netdev = adapter->netdev;
3364     int err;
3365 
3366     if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
3367         err = ixgbe_request_msix_irqs(adapter);
3368     else if (adapter->flags & IXGBE_FLAG_MSI_ENABLED)
3369         err = request_irq(adapter->pdev->irq, ixgbe_intr, 0,
3370                   netdev->name, adapter);
3371     else
3372         err = request_irq(adapter->pdev->irq, ixgbe_intr, IRQF_SHARED,
3373                   netdev->name, adapter);
3374 
3375     if (err)
3376         e_err(probe, "request_irq failed, Error %d\n", err);
3377 
3378     return err;
3379 }
3380 
3381 static void ixgbe_free_irq(struct ixgbe_adapter *adapter)
3382 {
3383     int vector;
3384 
3385     if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
3386         free_irq(adapter->pdev->irq, adapter);
3387         return;
3388     }
3389 
3390     if (!adapter->msix_entries)
3391         return;
3392 
3393     for (vector = 0; vector < adapter->num_q_vectors; vector++) {
3394         struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
3395         struct msix_entry *entry = &adapter->msix_entries[vector];
3396 
3397         /* free only the irqs that were actually requested */
3398         if (!q_vector->rx.ring && !q_vector->tx.ring)
3399             continue;
3400 
3401         /* clear the affinity_mask in the IRQ descriptor */
3402         irq_update_affinity_hint(entry->vector, NULL);
3403 
3404         free_irq(entry->vector, q_vector);
3405     }
3406 
3407     free_irq(adapter->msix_entries[vector].vector, adapter);
3408 }
3409 
3410 /**
3411  * ixgbe_irq_disable - Mask off interrupt generation on the NIC
3412  * @adapter: board private structure
3413  **/
3414 static inline void ixgbe_irq_disable(struct ixgbe_adapter *adapter)
3415 {
3416     switch (adapter->hw.mac.type) {
3417     case ixgbe_mac_82598EB:
3418         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
3419         break;
3420     case ixgbe_mac_82599EB:
3421     case ixgbe_mac_X540:
3422     case ixgbe_mac_X550:
3423     case ixgbe_mac_X550EM_x:
3424     case ixgbe_mac_x550em_a:
3425         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
3426         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
3427         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
3428         break;
3429     default:
3430         break;
3431     }
3432     IXGBE_WRITE_FLUSH(&adapter->hw);
3433     if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
3434         int vector;
3435 
3436         for (vector = 0; vector < adapter->num_q_vectors; vector++)
3437             synchronize_irq(adapter->msix_entries[vector].vector);
3438 
3439         synchronize_irq(adapter->msix_entries[vector++].vector);
3440     } else {
3441         synchronize_irq(adapter->pdev->irq);
3442     }
3443 }
3444 
3445 /**
3446  * ixgbe_configure_msi_and_legacy - Initialize PIN (INTA...) and MSI interrupts
3447  * @adapter: board private structure
3448  *
3449  **/
3450 static void ixgbe_configure_msi_and_legacy(struct ixgbe_adapter *adapter)
3451 {
3452     struct ixgbe_q_vector *q_vector = adapter->q_vector[0];
3453 
3454     ixgbe_write_eitr(q_vector);
3455 
3456     ixgbe_set_ivar(adapter, 0, 0, 0);
3457     ixgbe_set_ivar(adapter, 1, 0, 0);
3458 
3459     e_info(hw, "Legacy interrupt IVAR setup done\n");
3460 }
3461 
3462 /**
3463  * ixgbe_configure_tx_ring - Configure 8259x Tx ring after Reset
3464  * @adapter: board private structure
3465  * @ring: structure containing ring specific data
3466  *
3467  * Configure the Tx descriptor ring after a reset.
3468  **/
3469 void ixgbe_configure_tx_ring(struct ixgbe_adapter *adapter,
3470                  struct ixgbe_ring *ring)
3471 {
3472     struct ixgbe_hw *hw = &adapter->hw;
3473     u64 tdba = ring->dma;
3474     int wait_loop = 10;
3475     u32 txdctl = IXGBE_TXDCTL_ENABLE;
3476     u8 reg_idx = ring->reg_idx;
3477 
3478     ring->xsk_pool = NULL;
3479     if (ring_is_xdp(ring))
3480         ring->xsk_pool = ixgbe_xsk_pool(adapter, ring);
3481 
3482     /* disable queue to avoid issues while updating state */
3483     IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), 0);
3484     IXGBE_WRITE_FLUSH(hw);
3485 
3486     IXGBE_WRITE_REG(hw, IXGBE_TDBAL(reg_idx),
3487             (tdba & DMA_BIT_MASK(32)));
3488     IXGBE_WRITE_REG(hw, IXGBE_TDBAH(reg_idx), (tdba >> 32));
3489     IXGBE_WRITE_REG(hw, IXGBE_TDLEN(reg_idx),
3490             ring->count * sizeof(union ixgbe_adv_tx_desc));
3491     IXGBE_WRITE_REG(hw, IXGBE_TDH(reg_idx), 0);
3492     IXGBE_WRITE_REG(hw, IXGBE_TDT(reg_idx), 0);
3493     ring->tail = adapter->io_addr + IXGBE_TDT(reg_idx);
3494 
3495     /*
3496      * set WTHRESH to encourage burst writeback, it should not be set
3497      * higher than 1 when:
3498      * - ITR is 0 as it could cause false TX hangs
3499      * - ITR is set to > 100k int/sec and BQL is enabled
3500      *
3501      * In order to avoid issues WTHRESH + PTHRESH should always be equal
3502      * to or less than the number of on chip descriptors, which is
3503      * currently 40.
3504      */
3505     if (!ring->q_vector || (ring->q_vector->itr < IXGBE_100K_ITR))
3506         txdctl |= 1u << 16; /* WTHRESH = 1 */
3507     else
3508         txdctl |= 8u << 16; /* WTHRESH = 8 */
3509 
3510     /*
3511      * Setting PTHRESH to 32 both improves performance
3512      * and avoids a TX hang with DFP enabled
3513      */
3514     txdctl |= (1u << 8) |   /* HTHRESH = 1 */
3515            32;      /* PTHRESH = 32 */
3516 
3517     /* reinitialize flowdirector state */
3518     if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
3519         ring->atr_sample_rate = adapter->atr_sample_rate;
3520         ring->atr_count = 0;
3521         set_bit(__IXGBE_TX_FDIR_INIT_DONE, &ring->state);
3522     } else {
3523         ring->atr_sample_rate = 0;
3524     }
3525 
3526     /* initialize XPS */
3527     if (!test_and_set_bit(__IXGBE_TX_XPS_INIT_DONE, &ring->state)) {
3528         struct ixgbe_q_vector *q_vector = ring->q_vector;
3529 
3530         if (q_vector)
3531             netif_set_xps_queue(ring->netdev,
3532                         &q_vector->affinity_mask,
3533                         ring->queue_index);
3534     }
3535 
3536     clear_bit(__IXGBE_HANG_CHECK_ARMED, &ring->state);
3537 
3538     /* reinitialize tx_buffer_info */
3539     memset(ring->tx_buffer_info, 0,
3540            sizeof(struct ixgbe_tx_buffer) * ring->count);
3541 
3542     /* enable queue */
3543     IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), txdctl);
3544 
3545     /* TXDCTL.EN will return 0 on 82598 if link is down, so skip it */
3546     if (hw->mac.type == ixgbe_mac_82598EB &&
3547         !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
3548         return;
3549 
3550     /* poll to verify queue is enabled */
3551     do {
3552         usleep_range(1000, 2000);
3553         txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
3554     } while (--wait_loop && !(txdctl & IXGBE_TXDCTL_ENABLE));
3555     if (!wait_loop)
3556         hw_dbg(hw, "Could not enable Tx Queue %d\n", reg_idx);
3557 }
3558 
3559 static void ixgbe_setup_mtqc(struct ixgbe_adapter *adapter)
3560 {
3561     struct ixgbe_hw *hw = &adapter->hw;
3562     u32 rttdcs, mtqc;
3563     u8 tcs = adapter->hw_tcs;
3564 
3565     if (hw->mac.type == ixgbe_mac_82598EB)
3566         return;
3567 
3568     /* disable the arbiter while setting MTQC */
3569     rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
3570     rttdcs |= IXGBE_RTTDCS_ARBDIS;
3571     IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3572 
3573     /* set transmit pool layout */
3574     if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3575         mtqc = IXGBE_MTQC_VT_ENA;
3576         if (tcs > 4)
3577             mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3578         else if (tcs > 1)
3579             mtqc |= IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3580         else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3581              IXGBE_82599_VMDQ_4Q_MASK)
3582             mtqc |= IXGBE_MTQC_32VF;
3583         else
3584             mtqc |= IXGBE_MTQC_64VF;
3585     } else {
3586         if (tcs > 4) {
3587             mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_8TC_8TQ;
3588         } else if (tcs > 1) {
3589             mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3590         } else {
3591             u8 max_txq = adapter->num_tx_queues +
3592                 adapter->num_xdp_queues;
3593             if (max_txq > 63)
3594                 mtqc = IXGBE_MTQC_RT_ENA | IXGBE_MTQC_4TC_4TQ;
3595             else
3596                 mtqc = IXGBE_MTQC_64Q_1PB;
3597         }
3598     }
3599 
3600     IXGBE_WRITE_REG(hw, IXGBE_MTQC, mtqc);
3601 
3602     /* Enable Security TX Buffer IFG for multiple pb */
3603     if (tcs) {
3604         u32 sectx = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
3605         sectx |= IXGBE_SECTX_DCB;
3606         IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, sectx);
3607     }
3608 
3609     /* re-enable the arbiter */
3610     rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
3611     IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
3612 }
3613 
3614 /**
3615  * ixgbe_configure_tx - Configure 8259x Transmit Unit after Reset
3616  * @adapter: board private structure
3617  *
3618  * Configure the Tx unit of the MAC after a reset.
3619  **/
3620 static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
3621 {
3622     struct ixgbe_hw *hw = &adapter->hw;
3623     u32 dmatxctl;
3624     u32 i;
3625 
3626     ixgbe_setup_mtqc(adapter);
3627 
3628     if (hw->mac.type != ixgbe_mac_82598EB) {
3629         /* DMATXCTL.EN must be before Tx queues are enabled */
3630         dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
3631         dmatxctl |= IXGBE_DMATXCTL_TE;
3632         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
3633     }
3634 
3635     /* Setup the HW Tx Head and Tail descriptor pointers */
3636     for (i = 0; i < adapter->num_tx_queues; i++)
3637         ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
3638     for (i = 0; i < adapter->num_xdp_queues; i++)
3639         ixgbe_configure_tx_ring(adapter, adapter->xdp_ring[i]);
3640 }
3641 
3642 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
3643                  struct ixgbe_ring *ring)
3644 {
3645     struct ixgbe_hw *hw = &adapter->hw;
3646     u8 reg_idx = ring->reg_idx;
3647     u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3648 
3649     srrctl |= IXGBE_SRRCTL_DROP_EN;
3650 
3651     IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3652 }
3653 
3654 static void ixgbe_disable_rx_drop(struct ixgbe_adapter *adapter,
3655                   struct ixgbe_ring *ring)
3656 {
3657     struct ixgbe_hw *hw = &adapter->hw;
3658     u8 reg_idx = ring->reg_idx;
3659     u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(reg_idx));
3660 
3661     srrctl &= ~IXGBE_SRRCTL_DROP_EN;
3662 
3663     IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3664 }
3665 
3666 #ifdef CONFIG_IXGBE_DCB
3667 void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3668 #else
3669 static void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter)
3670 #endif
3671 {
3672     int i;
3673     bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
3674 
3675     if (adapter->ixgbe_ieee_pfc)
3676         pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
3677 
3678     /*
3679      * We should set the drop enable bit if:
3680      *  SR-IOV is enabled
3681      *   or
3682      *  Number of Rx queues > 1 and flow control is disabled
3683      *
3684      *  This allows us to avoid head of line blocking for security
3685      *  and performance reasons.
3686      */
3687     if (adapter->num_vfs || (adapter->num_rx_queues > 1 &&
3688         !(adapter->hw.fc.current_mode & ixgbe_fc_tx_pause) && !pfc_en)) {
3689         for (i = 0; i < adapter->num_rx_queues; i++)
3690             ixgbe_enable_rx_drop(adapter, adapter->rx_ring[i]);
3691     } else {
3692         for (i = 0; i < adapter->num_rx_queues; i++)
3693             ixgbe_disable_rx_drop(adapter, adapter->rx_ring[i]);
3694     }
3695 }
3696 
3697 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
3698 
3699 static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
3700                    struct ixgbe_ring *rx_ring)
3701 {
3702     struct ixgbe_hw *hw = &adapter->hw;
3703     u32 srrctl;
3704     u8 reg_idx = rx_ring->reg_idx;
3705 
3706     if (hw->mac.type == ixgbe_mac_82598EB) {
3707         u16 mask = adapter->ring_feature[RING_F_RSS].mask;
3708 
3709         /*
3710          * if VMDq is not active we must program one srrctl register
3711          * per RSS queue since we have enabled RDRXCTL.MVMEN
3712          */
3713         reg_idx &= mask;
3714     }
3715 
3716     /* configure header buffer length, needed for RSC */
3717     srrctl = IXGBE_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT;
3718 
3719     /* configure the packet buffer length */
3720     if (rx_ring->xsk_pool) {
3721         u32 xsk_buf_len = xsk_pool_get_rx_frame_size(rx_ring->xsk_pool);
3722 
3723         /* If the MAC support setting RXDCTL.RLPML, the
3724          * SRRCTL[n].BSIZEPKT is set to PAGE_SIZE and
3725          * RXDCTL.RLPML is set to the actual UMEM buffer
3726          * size. If not, then we are stuck with a 1k buffer
3727          * size resolution. In this case frames larger than
3728          * the UMEM buffer size viewed in a 1k resolution will
3729          * be dropped.
3730          */
3731         if (hw->mac.type != ixgbe_mac_82599EB)
3732             srrctl |= PAGE_SIZE >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3733         else
3734             srrctl |= xsk_buf_len >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3735     } else if (test_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state)) {
3736         srrctl |= IXGBE_RXBUFFER_3K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3737     } else {
3738         srrctl |= IXGBE_RXBUFFER_2K >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
3739     }
3740 
3741     /* configure descriptor type */
3742     srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
3743 
3744     IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(reg_idx), srrctl);
3745 }
3746 
3747 /**
3748  * ixgbe_rss_indir_tbl_entries - Return RSS indirection table entries
3749  * @adapter: device handle
3750  *
3751  *  - 82598/82599/X540:     128
3752  *  - X550(non-SRIOV mode): 512
3753  *  - X550(SRIOV mode):     64
3754  */
3755 u32 ixgbe_rss_indir_tbl_entries(struct ixgbe_adapter *adapter)
3756 {
3757     if (adapter->hw.mac.type < ixgbe_mac_X550)
3758         return 128;
3759     else if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
3760         return 64;
3761     else
3762         return 512;
3763 }
3764 
3765 /**
3766  * ixgbe_store_key - Write the RSS key to HW
3767  * @adapter: device handle
3768  *
3769  * Write the RSS key stored in adapter.rss_key to HW.
3770  */
3771 void ixgbe_store_key(struct ixgbe_adapter *adapter)
3772 {
3773     struct ixgbe_hw *hw = &adapter->hw;
3774     int i;
3775 
3776     for (i = 0; i < 10; i++)
3777         IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), adapter->rss_key[i]);
3778 }
3779 
3780 /**
3781  * ixgbe_init_rss_key - Initialize adapter RSS key
3782  * @adapter: device handle
3783  *
3784  * Allocates and initializes the RSS key if it is not allocated.
3785  **/
3786 static inline int ixgbe_init_rss_key(struct ixgbe_adapter *adapter)
3787 {
3788     u32 *rss_key;
3789 
3790     if (!adapter->rss_key) {
3791         rss_key = kzalloc(IXGBE_RSS_KEY_SIZE, GFP_KERNEL);
3792         if (unlikely(!rss_key))
3793             return -ENOMEM;
3794 
3795         netdev_rss_key_fill(rss_key, IXGBE_RSS_KEY_SIZE);
3796         adapter->rss_key = rss_key;
3797     }
3798 
3799     return 0;
3800 }
3801 
3802 /**
3803  * ixgbe_store_reta - Write the RETA table to HW
3804  * @adapter: device handle
3805  *
3806  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3807  */
3808 void ixgbe_store_reta(struct ixgbe_adapter *adapter)
3809 {
3810     u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3811     struct ixgbe_hw *hw = &adapter->hw;
3812     u32 reta = 0;
3813     u32 indices_multi;
3814     u8 *indir_tbl = adapter->rss_indir_tbl;
3815 
3816     /* Fill out the redirection table as follows:
3817      *  - 82598:      8 bit wide entries containing pair of 4 bit RSS
3818      *    indices.
3819      *  - 82599/X540: 8 bit wide entries containing 4 bit RSS index
3820      *  - X550:       8 bit wide entries containing 6 bit RSS index
3821      */
3822     if (adapter->hw.mac.type == ixgbe_mac_82598EB)
3823         indices_multi = 0x11;
3824     else
3825         indices_multi = 0x1;
3826 
3827     /* Write redirection table to HW */
3828     for (i = 0; i < reta_entries; i++) {
3829         reta |= indices_multi * indir_tbl[i] << (i & 0x3) * 8;
3830         if ((i & 3) == 3) {
3831             if (i < 128)
3832                 IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3833             else
3834                 IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32),
3835                         reta);
3836             reta = 0;
3837         }
3838     }
3839 }
3840 
3841 /**
3842  * ixgbe_store_vfreta - Write the RETA table to HW (x550 devices in SRIOV mode)
3843  * @adapter: device handle
3844  *
3845  * Write the RSS redirection table stored in adapter.rss_indir_tbl[] to HW.
3846  */
3847 static void ixgbe_store_vfreta(struct ixgbe_adapter *adapter)
3848 {
3849     u32 i, reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3850     struct ixgbe_hw *hw = &adapter->hw;
3851     u32 vfreta = 0;
3852 
3853     /* Write redirection table to HW */
3854     for (i = 0; i < reta_entries; i++) {
3855         u16 pool = adapter->num_rx_pools;
3856 
3857         vfreta |= (u32)adapter->rss_indir_tbl[i] << (i & 0x3) * 8;
3858         if ((i & 3) != 3)
3859             continue;
3860 
3861         while (pool--)
3862             IXGBE_WRITE_REG(hw,
3863                     IXGBE_PFVFRETA(i >> 2, VMDQ_P(pool)),
3864                     vfreta);
3865         vfreta = 0;
3866     }
3867 }
3868 
3869 static void ixgbe_setup_reta(struct ixgbe_adapter *adapter)
3870 {
3871     u32 i, j;
3872     u32 reta_entries = ixgbe_rss_indir_tbl_entries(adapter);
3873     u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3874 
3875     /* Program table for at least 4 queues w/ SR-IOV so that VFs can
3876      * make full use of any rings they may have.  We will use the
3877      * PSRTYPE register to control how many rings we use within the PF.
3878      */
3879     if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) && (rss_i < 4))
3880         rss_i = 4;
3881 
3882     /* Fill out hash function seeds */
3883     ixgbe_store_key(adapter);
3884 
3885     /* Fill out redirection table */
3886     memset(adapter->rss_indir_tbl, 0, sizeof(adapter->rss_indir_tbl));
3887 
3888     for (i = 0, j = 0; i < reta_entries; i++, j++) {
3889         if (j == rss_i)
3890             j = 0;
3891 
3892         adapter->rss_indir_tbl[i] = j;
3893     }
3894 
3895     ixgbe_store_reta(adapter);
3896 }
3897 
3898 static void ixgbe_setup_vfreta(struct ixgbe_adapter *adapter)
3899 {
3900     struct ixgbe_hw *hw = &adapter->hw;
3901     u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
3902     int i, j;
3903 
3904     /* Fill out hash function seeds */
3905     for (i = 0; i < 10; i++) {
3906         u16 pool = adapter->num_rx_pools;
3907 
3908         while (pool--)
3909             IXGBE_WRITE_REG(hw,
3910                     IXGBE_PFVFRSSRK(i, VMDQ_P(pool)),
3911                     *(adapter->rss_key + i));
3912     }
3913 
3914     /* Fill out the redirection table */
3915     for (i = 0, j = 0; i < 64; i++, j++) {
3916         if (j == rss_i)
3917             j = 0;
3918 
3919         adapter->rss_indir_tbl[i] = j;
3920     }
3921 
3922     ixgbe_store_vfreta(adapter);
3923 }
3924 
3925 static void ixgbe_setup_mrqc(struct ixgbe_adapter *adapter)
3926 {
3927     struct ixgbe_hw *hw = &adapter->hw;
3928     u32 mrqc = 0, rss_field = 0, vfmrqc = 0;
3929     u32 rxcsum;
3930 
3931     /* Disable indicating checksum in descriptor, enables RSS hash */
3932     rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
3933     rxcsum |= IXGBE_RXCSUM_PCSD;
3934     IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
3935 
3936     if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
3937         if (adapter->ring_feature[RING_F_RSS].mask)
3938             mrqc = IXGBE_MRQC_RSSEN;
3939     } else {
3940         u8 tcs = adapter->hw_tcs;
3941 
3942         if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
3943             if (tcs > 4)
3944                 mrqc = IXGBE_MRQC_VMDQRT8TCEN;  /* 8 TCs */
3945             else if (tcs > 1)
3946                 mrqc = IXGBE_MRQC_VMDQRT4TCEN;  /* 4 TCs */
3947             else if (adapter->ring_feature[RING_F_VMDQ].mask ==
3948                  IXGBE_82599_VMDQ_4Q_MASK)
3949                 mrqc = IXGBE_MRQC_VMDQRSS32EN;
3950             else
3951                 mrqc = IXGBE_MRQC_VMDQRSS64EN;
3952 
3953             /* Enable L3/L4 for Tx Switched packets only for X550,
3954              * older devices do not support this feature
3955              */
3956             if (hw->mac.type >= ixgbe_mac_X550)
3957                 mrqc |= IXGBE_MRQC_L3L4TXSWEN;
3958         } else {
3959             if (tcs > 4)
3960                 mrqc = IXGBE_MRQC_RTRSS8TCEN;
3961             else if (tcs > 1)
3962                 mrqc = IXGBE_MRQC_RTRSS4TCEN;
3963             else
3964                 mrqc = IXGBE_MRQC_RSSEN;
3965         }
3966     }
3967 
3968     /* Perform hash on these packet types */
3969     rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4 |
3970              IXGBE_MRQC_RSS_FIELD_IPV4_TCP |
3971              IXGBE_MRQC_RSS_FIELD_IPV6 |
3972              IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
3973 
3974     if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV4_UDP)
3975         rss_field |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
3976     if (adapter->flags2 & IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
3977         rss_field |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
3978 
3979     if ((hw->mac.type >= ixgbe_mac_X550) &&
3980         (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)) {
3981         u16 pool = adapter->num_rx_pools;
3982 
3983         /* Enable VF RSS mode */
3984         mrqc |= IXGBE_MRQC_MULTIPLE_RSS;
3985         IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
3986 
3987         /* Setup RSS through the VF registers */
3988         ixgbe_setup_vfreta(adapter);
3989         vfmrqc = IXGBE_MRQC_RSSEN;
3990         vfmrqc |= rss_field;
3991 
3992         while (pool--)
3993             IXGBE_WRITE_REG(hw,
3994                     IXGBE_PFVFMRQC(VMDQ_P(pool)),
3995                     vfmrqc);
3996     } else {
3997         ixgbe_setup_reta(adapter);
3998         mrqc |= rss_field;
3999         IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
4000     }
4001 }
4002 
4003 /**
4004  * ixgbe_configure_rscctl - enable RSC for the indicated ring
4005  * @adapter: address of board private structure
4006  * @ring: structure containing ring specific data
4007  **/
4008 static void ixgbe_configure_rscctl(struct ixgbe_adapter *adapter,
4009                    struct ixgbe_ring *ring)
4010 {
4011     struct ixgbe_hw *hw = &adapter->hw;
4012     u32 rscctrl;
4013     u8 reg_idx = ring->reg_idx;
4014 
4015     if (!ring_is_rsc_enabled(ring))
4016         return;
4017 
4018     rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(reg_idx));
4019     rscctrl |= IXGBE_RSCCTL_RSCEN;
4020     /*
4021      * we must limit the number of descriptors so that the
4022      * total size of max desc * buf_len is not greater
4023      * than 65536
4024      */
4025     rscctrl |= IXGBE_RSCCTL_MAXDESC_16;
4026     IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(reg_idx), rscctrl);
4027 }
4028 
4029 #define IXGBE_MAX_RX_DESC_POLL 10
4030 static void ixgbe_rx_desc_queue_enable(struct ixgbe_adapter *adapter,
4031                        struct ixgbe_ring *ring)
4032 {
4033     struct ixgbe_hw *hw = &adapter->hw;
4034     int wait_loop = IXGBE_MAX_RX_DESC_POLL;
4035     u32 rxdctl;
4036     u8 reg_idx = ring->reg_idx;
4037 
4038     if (ixgbe_removed(hw->hw_addr))
4039         return;
4040     /* RXDCTL.EN will return 0 on 82598 if link is down, so skip it */
4041     if (hw->mac.type == ixgbe_mac_82598EB &&
4042         !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
4043         return;
4044 
4045     do {
4046         usleep_range(1000, 2000);
4047         rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4048     } while (--wait_loop && !(rxdctl & IXGBE_RXDCTL_ENABLE));
4049 
4050     if (!wait_loop) {
4051         e_err(drv, "RXDCTL.ENABLE on Rx queue %d not set within "
4052               "the polling period\n", reg_idx);
4053     }
4054 }
4055 
4056 void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
4057                  struct ixgbe_ring *ring)
4058 {
4059     struct ixgbe_hw *hw = &adapter->hw;
4060     union ixgbe_adv_rx_desc *rx_desc;
4061     u64 rdba = ring->dma;
4062     u32 rxdctl;
4063     u8 reg_idx = ring->reg_idx;
4064 
4065     xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
4066     ring->xsk_pool = ixgbe_xsk_pool(adapter, ring);
4067     if (ring->xsk_pool) {
4068         WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4069                            MEM_TYPE_XSK_BUFF_POOL,
4070                            NULL));
4071         xsk_pool_set_rxq_info(ring->xsk_pool, &ring->xdp_rxq);
4072     } else {
4073         WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4074                            MEM_TYPE_PAGE_SHARED, NULL));
4075     }
4076 
4077     /* disable queue to avoid use of these values while updating state */
4078     rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
4079     rxdctl &= ~IXGBE_RXDCTL_ENABLE;
4080 
4081     /* write value back with RXDCTL.ENABLE bit cleared */
4082     IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4083     IXGBE_WRITE_FLUSH(hw);
4084 
4085     IXGBE_WRITE_REG(hw, IXGBE_RDBAL(reg_idx), (rdba & DMA_BIT_MASK(32)));
4086     IXGBE_WRITE_REG(hw, IXGBE_RDBAH(reg_idx), (rdba >> 32));
4087     IXGBE_WRITE_REG(hw, IXGBE_RDLEN(reg_idx),
4088             ring->count * sizeof(union ixgbe_adv_rx_desc));
4089     /* Force flushing of IXGBE_RDLEN to prevent MDD */
4090     IXGBE_WRITE_FLUSH(hw);
4091 
4092     IXGBE_WRITE_REG(hw, IXGBE_RDH(reg_idx), 0);
4093     IXGBE_WRITE_REG(hw, IXGBE_RDT(reg_idx), 0);
4094     ring->tail = adapter->io_addr + IXGBE_RDT(reg_idx);
4095 
4096     ixgbe_configure_srrctl(adapter, ring);
4097     ixgbe_configure_rscctl(adapter, ring);
4098 
4099     if (hw->mac.type == ixgbe_mac_82598EB) {
4100         /*
4101          * enable cache line friendly hardware writes:
4102          * PTHRESH=32 descriptors (half the internal cache),
4103          * this also removes ugly rx_no_buffer_count increment
4104          * HTHRESH=4 descriptors (to minimize latency on fetch)
4105          * WTHRESH=8 burst writeback up to two cache lines
4106          */
4107         rxdctl &= ~0x3FFFFF;
4108         rxdctl |=  0x080420;
4109 #if (PAGE_SIZE < 8192)
4110     /* RXDCTL.RLPML does not work on 82599 */
4111     } else if (hw->mac.type != ixgbe_mac_82599EB) {
4112         rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4113                 IXGBE_RXDCTL_RLPML_EN);
4114 
4115         /* Limit the maximum frame size so we don't overrun the skb.
4116          * This can happen in SRIOV mode when the MTU of the VF is
4117          * higher than the MTU of the PF.
4118          */
4119         if (ring_uses_build_skb(ring) &&
4120             !test_bit(__IXGBE_RX_3K_BUFFER, &ring->state))
4121             rxdctl |= IXGBE_MAX_2K_FRAME_BUILD_SKB |
4122                   IXGBE_RXDCTL_RLPML_EN;
4123 #endif
4124     }
4125 
4126     ring->rx_offset = ixgbe_rx_offset(ring);
4127 
4128     if (ring->xsk_pool && hw->mac.type != ixgbe_mac_82599EB) {
4129         u32 xsk_buf_len = xsk_pool_get_rx_frame_size(ring->xsk_pool);
4130 
4131         rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
4132                 IXGBE_RXDCTL_RLPML_EN);
4133         rxdctl |= xsk_buf_len | IXGBE_RXDCTL_RLPML_EN;
4134 
4135         ring->rx_buf_len = xsk_buf_len;
4136     }
4137 
4138     /* initialize rx_buffer_info */
4139     memset(ring->rx_buffer_info, 0,
4140            sizeof(struct ixgbe_rx_buffer) * ring->count);
4141 
4142     /* initialize Rx descriptor 0 */
4143     rx_desc = IXGBE_RX_DESC(ring, 0);
4144     rx_desc->wb.upper.length = 0;
4145 
4146     /* enable receive descriptor ring */
4147     rxdctl |= IXGBE_RXDCTL_ENABLE;
4148     IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
4149 
4150     ixgbe_rx_desc_queue_enable(adapter, ring);
4151     if (ring->xsk_pool)
4152         ixgbe_alloc_rx_buffers_zc(ring, ixgbe_desc_unused(ring));
4153     else
4154         ixgbe_alloc_rx_buffers(ring, ixgbe_desc_unused(ring));
4155 }
4156 
4157 static void ixgbe_setup_psrtype(struct ixgbe_adapter *adapter)
4158 {
4159     struct ixgbe_hw *hw = &adapter->hw;
4160     int rss_i = adapter->ring_feature[RING_F_RSS].indices;
4161     u16 pool = adapter->num_rx_pools;
4162 
4163     /* PSRTYPE must be initialized in non 82598 adapters */
4164     u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
4165               IXGBE_PSRTYPE_UDPHDR |
4166               IXGBE_PSRTYPE_IPV4HDR |
4167               IXGBE_PSRTYPE_L2HDR |
4168               IXGBE_PSRTYPE_IPV6HDR;
4169 
4170     if (hw->mac.type == ixgbe_mac_82598EB)
4171         return;
4172 
4173     if (rss_i > 3)
4174         psrtype |= 2u << 29;
4175     else if (rss_i > 1)
4176         psrtype |= 1u << 29;
4177 
4178     while (pool--)
4179         IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(VMDQ_P(pool)), psrtype);
4180 }
4181 
4182 static void ixgbe_configure_virtualization(struct ixgbe_adapter *adapter)
4183 {
4184     struct ixgbe_hw *hw = &adapter->hw;
4185     u16 pool = adapter->num_rx_pools;
4186     u32 reg_offset, vf_shift, vmolr;
4187     u32 gcr_ext, vmdctl;
4188     int i;
4189 
4190     if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
4191         return;
4192 
4193     vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
4194     vmdctl |= IXGBE_VMD_CTL_VMDQ_EN;
4195     vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
4196     vmdctl |= VMDQ_P(0) << IXGBE_VT_CTL_POOL_SHIFT;
4197     vmdctl |= IXGBE_VT_CTL_REPLEN;
4198     IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
4199 
4200     /* accept untagged packets until a vlan tag is
4201      * specifically set for the VMDQ queue/pool
4202      */
4203     vmolr = IXGBE_VMOLR_AUPE;
4204     while (pool--)
4205         IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(pool)), vmolr);
4206 
4207     vf_shift = VMDQ_P(0) % 32;
4208     reg_offset = (VMDQ_P(0) >= 32) ? 1 : 0;
4209 
4210     /* Enable only the PF's pool for Tx/Rx */
4211     IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), GENMASK(31, vf_shift));
4212     IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset ^ 1), reg_offset - 1);
4213     IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), GENMASK(31, vf_shift));
4214     IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset ^ 1), reg_offset - 1);
4215     if (adapter->bridge_mode == BRIDGE_MODE_VEB)
4216         IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
4217 
4218     /* Map PF MAC address in RAR Entry 0 to first pool following VFs */
4219     hw->mac.ops.set_vmdq(hw, 0, VMDQ_P(0));
4220 
4221     /* clear VLAN promisc flag so VFTA will be updated if necessary */
4222     adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4223 
4224     /*
4225      * Set up VF register offsets for selected VT Mode,
4226      * i.e. 32 or 64 VFs for SR-IOV
4227      */
4228     switch (adapter->ring_feature[RING_F_VMDQ].mask) {
4229     case IXGBE_82599_VMDQ_8Q_MASK:
4230         gcr_ext = IXGBE_GCR_EXT_VT_MODE_16;
4231         break;
4232     case IXGBE_82599_VMDQ_4Q_MASK:
4233         gcr_ext = IXGBE_GCR_EXT_VT_MODE_32;
4234         break;
4235     default:
4236         gcr_ext = IXGBE_GCR_EXT_VT_MODE_64;
4237         break;
4238     }
4239 
4240     IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, gcr_ext);
4241 
4242     for (i = 0; i < adapter->num_vfs; i++) {
4243         /* configure spoof checking */
4244         ixgbe_ndo_set_vf_spoofchk(adapter->netdev, i,
4245                       adapter->vfinfo[i].spoofchk_enabled);
4246 
4247         /* Enable/Disable RSS query feature  */
4248         ixgbe_ndo_set_vf_rss_query_en(adapter->netdev, i,
4249                       adapter->vfinfo[i].rss_query_enabled);
4250     }
4251 }
4252 
4253 static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
4254 {
4255     struct ixgbe_hw *hw = &adapter->hw;
4256     struct net_device *netdev = adapter->netdev;
4257     int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4258     struct ixgbe_ring *rx_ring;
4259     int i;
4260     u32 mhadd, hlreg0;
4261 
4262 #ifdef IXGBE_FCOE
4263     /* adjust max frame to be able to do baby jumbo for FCoE */
4264     if ((adapter->flags & IXGBE_FLAG_FCOE_ENABLED) &&
4265         (max_frame < IXGBE_FCOE_JUMBO_FRAME_SIZE))
4266         max_frame = IXGBE_FCOE_JUMBO_FRAME_SIZE;
4267 
4268 #endif /* IXGBE_FCOE */
4269 
4270     /* adjust max frame to be at least the size of a standard frame */
4271     if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
4272         max_frame = (ETH_FRAME_LEN + ETH_FCS_LEN);
4273 
4274     mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
4275     if (max_frame != (mhadd >> IXGBE_MHADD_MFS_SHIFT)) {
4276         mhadd &= ~IXGBE_MHADD_MFS_MASK;
4277         mhadd |= max_frame << IXGBE_MHADD_MFS_SHIFT;
4278 
4279         IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
4280     }
4281 
4282     hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
4283     /* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
4284     hlreg0 |= IXGBE_HLREG0_JUMBOEN;
4285     IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
4286 
4287     /*
4288      * Setup the HW Rx Head and Tail Descriptor Pointers and
4289      * the Base and Length of the Rx Descriptor Ring
4290      */
4291     for (i = 0; i < adapter->num_rx_queues; i++) {
4292         rx_ring = adapter->rx_ring[i];
4293 
4294         clear_ring_rsc_enabled(rx_ring);
4295         clear_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4296         clear_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4297 
4298         if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4299             set_ring_rsc_enabled(rx_ring);
4300 
4301         if (test_bit(__IXGBE_RX_FCOE, &rx_ring->state))
4302             set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4303 
4304         if (adapter->flags2 & IXGBE_FLAG2_RX_LEGACY)
4305             continue;
4306 
4307         set_bit(__IXGBE_RX_BUILD_SKB_ENABLED, &rx_ring->state);
4308 
4309 #if (PAGE_SIZE < 8192)
4310         if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
4311             set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4312 
4313         if (IXGBE_2K_TOO_SMALL_WITH_PADDING ||
4314             (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
4315             set_bit(__IXGBE_RX_3K_BUFFER, &rx_ring->state);
4316 #endif
4317     }
4318 }
4319 
4320 static void ixgbe_setup_rdrxctl(struct ixgbe_adapter *adapter)
4321 {
4322     struct ixgbe_hw *hw = &adapter->hw;
4323     u32 rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL);
4324 
4325     switch (hw->mac.type) {
4326     case ixgbe_mac_82598EB:
4327         /*
4328          * For VMDq support of different descriptor types or
4329          * buffer sizes through the use of multiple SRRCTL
4330          * registers, RDRXCTL.MVMEN must be set to 1
4331          *
4332          * also, the manual doesn't mention it clearly but DCA hints
4333          * will only use queue 0's tags unless this bit is set.  Side
4334          * effects of setting this bit are only that SRRCTL must be
4335          * fully programmed [0..15]
4336          */
4337         rdrxctl |= IXGBE_RDRXCTL_MVMEN;
4338         break;
4339     case ixgbe_mac_X550:
4340     case ixgbe_mac_X550EM_x:
4341     case ixgbe_mac_x550em_a:
4342         if (adapter->num_vfs)
4343             rdrxctl |= IXGBE_RDRXCTL_PSP;
4344         fallthrough;
4345     case ixgbe_mac_82599EB:
4346     case ixgbe_mac_X540:
4347         /* Disable RSC for ACK packets */
4348         IXGBE_WRITE_REG(hw, IXGBE_RSCDBU,
4349            (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU)));
4350         rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE;
4351         /* hardware requires some bits to be set by default */
4352         rdrxctl |= (IXGBE_RDRXCTL_RSCACKC | IXGBE_RDRXCTL_FCOE_WRFIX);
4353         rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP;
4354         break;
4355     default:
4356         /* We should do nothing since we don't know this hardware */
4357         return;
4358     }
4359 
4360     IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl);
4361 }
4362 
4363 /**
4364  * ixgbe_configure_rx - Configure 8259x Receive Unit after Reset
4365  * @adapter: board private structure
4366  *
4367  * Configure the Rx unit of the MAC after a reset.
4368  **/
4369 static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
4370 {
4371     struct ixgbe_hw *hw = &adapter->hw;
4372     int i;
4373     u32 rxctrl, rfctl;
4374 
4375     /* disable receives while setting up the descriptors */
4376     hw->mac.ops.disable_rx(hw);
4377 
4378     ixgbe_setup_psrtype(adapter);
4379     ixgbe_setup_rdrxctl(adapter);
4380 
4381     /* RSC Setup */
4382     rfctl = IXGBE_READ_REG(hw, IXGBE_RFCTL);
4383     rfctl &= ~IXGBE_RFCTL_RSC_DIS;
4384     if (!(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED))
4385         rfctl |= IXGBE_RFCTL_RSC_DIS;
4386 
4387     /* disable NFS filtering */
4388     rfctl |= (IXGBE_RFCTL_NFSW_DIS | IXGBE_RFCTL_NFSR_DIS);
4389     IXGBE_WRITE_REG(hw, IXGBE_RFCTL, rfctl);
4390 
4391     /* Program registers for the distribution of queues */
4392     ixgbe_setup_mrqc(adapter);
4393 
4394     /* set_rx_buffer_len must be called before ring initialization */
4395     ixgbe_set_rx_buffer_len(adapter);
4396 
4397     /*
4398      * Setup the HW Rx Head and Tail Descriptor Pointers and
4399      * the Base and Length of the Rx Descriptor Ring
4400      */
4401     for (i = 0; i < adapter->num_rx_queues; i++)
4402         ixgbe_configure_rx_ring(adapter, adapter->rx_ring[i]);
4403 
4404     rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
4405     /* disable drop enable for 82598 parts */
4406     if (hw->mac.type == ixgbe_mac_82598EB)
4407         rxctrl |= IXGBE_RXCTRL_DMBYPS;
4408 
4409     /* enable all receives */
4410     rxctrl |= IXGBE_RXCTRL_RXEN;
4411     hw->mac.ops.enable_rx_dma(hw, rxctrl);
4412 }
4413 
4414 static int ixgbe_vlan_rx_add_vid(struct net_device *netdev,
4415                  __be16 proto, u16 vid)
4416 {
4417     struct ixgbe_adapter *adapter = netdev_priv(netdev);
4418     struct ixgbe_hw *hw = &adapter->hw;
4419 
4420     /* add VID to filter table */
4421     if (!vid || !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4422         hw->mac.ops.set_vfta(&adapter->hw, vid, VMDQ_P(0), true, !!vid);
4423 
4424     set_bit(vid, adapter->active_vlans);
4425 
4426     return 0;
4427 }
4428 
4429 static int ixgbe_find_vlvf_entry(struct ixgbe_hw *hw, u32 vlan)
4430 {
4431     u32 vlvf;
4432     int idx;
4433 
4434     /* short cut the special case */
4435     if (vlan == 0)
4436         return 0;
4437 
4438     /* Search for the vlan id in the VLVF entries */
4439     for (idx = IXGBE_VLVF_ENTRIES; --idx;) {
4440         vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(idx));
4441         if ((vlvf & VLAN_VID_MASK) == vlan)
4442             break;
4443     }
4444 
4445     return idx;
4446 }
4447 
4448 void ixgbe_update_pf_promisc_vlvf(struct ixgbe_adapter *adapter, u32 vid)
4449 {
4450     struct ixgbe_hw *hw = &adapter->hw;
4451     u32 bits, word;
4452     int idx;
4453 
4454     idx = ixgbe_find_vlvf_entry(hw, vid);
4455     if (!idx)
4456         return;
4457 
4458     /* See if any other pools are set for this VLAN filter
4459      * entry other than the PF.
4460      */
4461     word = idx * 2 + (VMDQ_P(0) / 32);
4462     bits = ~BIT(VMDQ_P(0) % 32);
4463     bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4464 
4465     /* Disable the filter so this falls into the default pool. */
4466     if (!bits && !IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1))) {
4467         if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4468             IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), 0);
4469         IXGBE_WRITE_REG(hw, IXGBE_VLVF(idx), 0);
4470     }
4471 }
4472 
4473 static int ixgbe_vlan_rx_kill_vid(struct net_device *netdev,
4474                   __be16 proto, u16 vid)
4475 {
4476     struct ixgbe_adapter *adapter = netdev_priv(netdev);
4477     struct ixgbe_hw *hw = &adapter->hw;
4478 
4479     /* remove VID from filter table */
4480     if (vid && !(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4481         hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), false, true);
4482 
4483     clear_bit(vid, adapter->active_vlans);
4484 
4485     return 0;
4486 }
4487 
4488 /**
4489  * ixgbe_vlan_strip_disable - helper to disable hw vlan stripping
4490  * @adapter: driver data
4491  */
4492 static void ixgbe_vlan_strip_disable(struct ixgbe_adapter *adapter)
4493 {
4494     struct ixgbe_hw *hw = &adapter->hw;
4495     u32 vlnctrl;
4496     int i, j;
4497 
4498     switch (hw->mac.type) {
4499     case ixgbe_mac_82598EB:
4500         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4501         vlnctrl &= ~IXGBE_VLNCTRL_VME;
4502         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4503         break;
4504     case ixgbe_mac_82599EB:
4505     case ixgbe_mac_X540:
4506     case ixgbe_mac_X550:
4507     case ixgbe_mac_X550EM_x:
4508     case ixgbe_mac_x550em_a:
4509         for (i = 0; i < adapter->num_rx_queues; i++) {
4510             struct ixgbe_ring *ring = adapter->rx_ring[i];
4511 
4512             if (!netif_is_ixgbe(ring->netdev))
4513                 continue;
4514 
4515             j = ring->reg_idx;
4516             vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4517             vlnctrl &= ~IXGBE_RXDCTL_VME;
4518             IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4519         }
4520         break;
4521     default:
4522         break;
4523     }
4524 }
4525 
4526 /**
4527  * ixgbe_vlan_strip_enable - helper to enable hw vlan stripping
4528  * @adapter: driver data
4529  */
4530 static void ixgbe_vlan_strip_enable(struct ixgbe_adapter *adapter)
4531 {
4532     struct ixgbe_hw *hw = &adapter->hw;
4533     u32 vlnctrl;
4534     int i, j;
4535 
4536     switch (hw->mac.type) {
4537     case ixgbe_mac_82598EB:
4538         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4539         vlnctrl |= IXGBE_VLNCTRL_VME;
4540         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4541         break;
4542     case ixgbe_mac_82599EB:
4543     case ixgbe_mac_X540:
4544     case ixgbe_mac_X550:
4545     case ixgbe_mac_X550EM_x:
4546     case ixgbe_mac_x550em_a:
4547         for (i = 0; i < adapter->num_rx_queues; i++) {
4548             struct ixgbe_ring *ring = adapter->rx_ring[i];
4549 
4550             if (!netif_is_ixgbe(ring->netdev))
4551                 continue;
4552 
4553             j = ring->reg_idx;
4554             vlnctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(j));
4555             vlnctrl |= IXGBE_RXDCTL_VME;
4556             IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(j), vlnctrl);
4557         }
4558         break;
4559     default:
4560         break;
4561     }
4562 }
4563 
4564 static void ixgbe_vlan_promisc_enable(struct ixgbe_adapter *adapter)
4565 {
4566     struct ixgbe_hw *hw = &adapter->hw;
4567     u32 vlnctrl, i;
4568 
4569     vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4570 
4571     if (adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) {
4572     /* For VMDq and SR-IOV we must leave VLAN filtering enabled */
4573         vlnctrl |= IXGBE_VLNCTRL_VFE;
4574         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4575     } else {
4576         vlnctrl &= ~IXGBE_VLNCTRL_VFE;
4577         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4578         return;
4579     }
4580 
4581     /* Nothing to do for 82598 */
4582     if (hw->mac.type == ixgbe_mac_82598EB)
4583         return;
4584 
4585     /* We are already in VLAN promisc, nothing to do */
4586     if (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC)
4587         return;
4588 
4589     /* Set flag so we don't redo unnecessary work */
4590     adapter->flags2 |= IXGBE_FLAG2_VLAN_PROMISC;
4591 
4592     /* Add PF to all active pools */
4593     for (i = IXGBE_VLVF_ENTRIES; --i;) {
4594         u32 reg_offset = IXGBE_VLVFB(i * 2 + VMDQ_P(0) / 32);
4595         u32 vlvfb = IXGBE_READ_REG(hw, reg_offset);
4596 
4597         vlvfb |= BIT(VMDQ_P(0) % 32);
4598         IXGBE_WRITE_REG(hw, reg_offset, vlvfb);
4599     }
4600 
4601     /* Set all bits in the VLAN filter table array */
4602     for (i = hw->mac.vft_size; i--;)
4603         IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), ~0U);
4604 }
4605 
4606 #define VFTA_BLOCK_SIZE 8
4607 static void ixgbe_scrub_vfta(struct ixgbe_adapter *adapter, u32 vfta_offset)
4608 {
4609     struct ixgbe_hw *hw = &adapter->hw;
4610     u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4611     u32 vid_start = vfta_offset * 32;
4612     u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4613     u32 i, vid, word, bits;
4614 
4615     for (i = IXGBE_VLVF_ENTRIES; --i;) {
4616         u32 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
4617 
4618         /* pull VLAN ID from VLVF */
4619         vid = vlvf & VLAN_VID_MASK;
4620 
4621         /* only concern outselves with a certain range */
4622         if (vid < vid_start || vid >= vid_end)
4623             continue;
4624 
4625         if (vlvf) {
4626             /* record VLAN ID in VFTA */
4627             vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4628 
4629             /* if PF is part of this then continue */
4630             if (test_bit(vid, adapter->active_vlans))
4631                 continue;
4632         }
4633 
4634         /* remove PF from the pool */
4635         word = i * 2 + VMDQ_P(0) / 32;
4636         bits = ~BIT(VMDQ_P(0) % 32);
4637         bits &= IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
4638         IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), bits);
4639     }
4640 
4641     /* extract values from active_vlans and write back to VFTA */
4642     for (i = VFTA_BLOCK_SIZE; i--;) {
4643         vid = (vfta_offset + i) * 32;
4644         word = vid / BITS_PER_LONG;
4645         bits = vid % BITS_PER_LONG;
4646 
4647         vfta[i] |= adapter->active_vlans[word] >> bits;
4648 
4649         IXGBE_WRITE_REG(hw, IXGBE_VFTA(vfta_offset + i), vfta[i]);
4650     }
4651 }
4652 
4653 static void ixgbe_vlan_promisc_disable(struct ixgbe_adapter *adapter)
4654 {
4655     struct ixgbe_hw *hw = &adapter->hw;
4656     u32 vlnctrl, i;
4657 
4658     /* Set VLAN filtering to enabled */
4659     vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
4660     vlnctrl |= IXGBE_VLNCTRL_VFE;
4661     IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
4662 
4663     if (!(adapter->flags & IXGBE_FLAG_VMDQ_ENABLED) ||
4664         hw->mac.type == ixgbe_mac_82598EB)
4665         return;
4666 
4667     /* We are not in VLAN promisc, nothing to do */
4668     if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
4669         return;
4670 
4671     /* Set flag so we don't redo unnecessary work */
4672     adapter->flags2 &= ~IXGBE_FLAG2_VLAN_PROMISC;
4673 
4674     for (i = 0; i < hw->mac.vft_size; i += VFTA_BLOCK_SIZE)
4675         ixgbe_scrub_vfta(adapter, i);
4676 }
4677 
4678 static void ixgbe_restore_vlan(struct ixgbe_adapter *adapter)
4679 {
4680     u16 vid = 1;
4681 
4682     ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
4683 
4684     for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
4685         ixgbe_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
4686 }
4687 
4688 /**
4689  * ixgbe_write_mc_addr_list - write multicast addresses to MTA
4690  * @netdev: network interface device structure
4691  *
4692  * Writes multicast address list to the MTA hash table.
4693  * Returns: -ENOMEM on failure
4694  *                0 on no addresses written
4695  *                X on writing X addresses to MTA
4696  **/
4697 static int ixgbe_write_mc_addr_list(struct net_device *netdev)
4698 {
4699     struct ixgbe_adapter *adapter = netdev_priv(netdev);
4700     struct ixgbe_hw *hw = &adapter->hw;
4701 
4702     if (!netif_running(netdev))
4703         return 0;
4704 
4705     if (hw->mac.ops.update_mc_addr_list)
4706         hw->mac.ops.update_mc_addr_list(hw, netdev);
4707     else
4708         return -ENOMEM;
4709 
4710 #ifdef CONFIG_PCI_IOV
4711     ixgbe_restore_vf_multicasts(adapter);
4712 #endif
4713 
4714     return netdev_mc_count(netdev);
4715 }
4716 
4717 #ifdef CONFIG_PCI_IOV
4718 void ixgbe_full_sync_mac_table(struct ixgbe_adapter *adapter)
4719 {
4720     struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4721     struct ixgbe_hw *hw = &adapter->hw;
4722     int i;
4723 
4724     for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4725         mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4726 
4727         if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4728             hw->mac.ops.set_rar(hw, i,
4729                         mac_table->addr,
4730                         mac_table->pool,
4731                         IXGBE_RAH_AV);
4732         else
4733             hw->mac.ops.clear_rar(hw, i);
4734     }
4735 }
4736 
4737 #endif
4738 static void ixgbe_sync_mac_table(struct ixgbe_adapter *adapter)
4739 {
4740     struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4741     struct ixgbe_hw *hw = &adapter->hw;
4742     int i;
4743 
4744     for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4745         if (!(mac_table->state & IXGBE_MAC_STATE_MODIFIED))
4746             continue;
4747 
4748         mac_table->state &= ~IXGBE_MAC_STATE_MODIFIED;
4749 
4750         if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4751             hw->mac.ops.set_rar(hw, i,
4752                         mac_table->addr,
4753                         mac_table->pool,
4754                         IXGBE_RAH_AV);
4755         else
4756             hw->mac.ops.clear_rar(hw, i);
4757     }
4758 }
4759 
4760 static void ixgbe_flush_sw_mac_table(struct ixgbe_adapter *adapter)
4761 {
4762     struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4763     struct ixgbe_hw *hw = &adapter->hw;
4764     int i;
4765 
4766     for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4767         mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4768         mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4769     }
4770 
4771     ixgbe_sync_mac_table(adapter);
4772 }
4773 
4774 static int ixgbe_available_rars(struct ixgbe_adapter *adapter, u16 pool)
4775 {
4776     struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4777     struct ixgbe_hw *hw = &adapter->hw;
4778     int i, count = 0;
4779 
4780     for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4781         /* do not count default RAR as available */
4782         if (mac_table->state & IXGBE_MAC_STATE_DEFAULT)
4783             continue;
4784 
4785         /* only count unused and addresses that belong to us */
4786         if (mac_table->state & IXGBE_MAC_STATE_IN_USE) {
4787             if (mac_table->pool != pool)
4788                 continue;
4789         }
4790 
4791         count++;
4792     }
4793 
4794     return count;
4795 }
4796 
4797 /* this function destroys the first RAR entry */
4798 static void ixgbe_mac_set_default_filter(struct ixgbe_adapter *adapter)
4799 {
4800     struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4801     struct ixgbe_hw *hw = &adapter->hw;
4802 
4803     memcpy(&mac_table->addr, hw->mac.addr, ETH_ALEN);
4804     mac_table->pool = VMDQ_P(0);
4805 
4806     mac_table->state = IXGBE_MAC_STATE_DEFAULT | IXGBE_MAC_STATE_IN_USE;
4807 
4808     hw->mac.ops.set_rar(hw, 0, mac_table->addr, mac_table->pool,
4809                 IXGBE_RAH_AV);
4810 }
4811 
4812 int ixgbe_add_mac_filter(struct ixgbe_adapter *adapter,
4813              const u8 *addr, u16 pool)
4814 {
4815     struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4816     struct ixgbe_hw *hw = &adapter->hw;
4817     int i;
4818 
4819     if (is_zero_ether_addr(addr))
4820         return -EINVAL;
4821 
4822     for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4823         if (mac_table->state & IXGBE_MAC_STATE_IN_USE)
4824             continue;
4825 
4826         ether_addr_copy(mac_table->addr, addr);
4827         mac_table->pool = pool;
4828 
4829         mac_table->state |= IXGBE_MAC_STATE_MODIFIED |
4830                     IXGBE_MAC_STATE_IN_USE;
4831 
4832         ixgbe_sync_mac_table(adapter);
4833 
4834         return i;
4835     }
4836 
4837     return -ENOMEM;
4838 }
4839 
4840 int ixgbe_del_mac_filter(struct ixgbe_adapter *adapter,
4841              const u8 *addr, u16 pool)
4842 {
4843     struct ixgbe_mac_addr *mac_table = &adapter->mac_table[0];
4844     struct ixgbe_hw *hw = &adapter->hw;
4845     int i;
4846 
4847     if (is_zero_ether_addr(addr))
4848         return -EINVAL;
4849 
4850     /* search table for addr, if found clear IN_USE flag and sync */
4851     for (i = 0; i < hw->mac.num_rar_entries; i++, mac_table++) {
4852         /* we can only delete an entry if it is in use */
4853         if (!(mac_table->state & IXGBE_MAC_STATE_IN_USE))
4854             continue;
4855         /* we only care about entries that belong to the given pool */
4856         if (mac_table->pool != pool)
4857             continue;
4858         /* we only care about a specific MAC address */
4859         if (!ether_addr_equal(addr, mac_table->addr))
4860             continue;
4861 
4862         mac_table->state |= IXGBE_MAC_STATE_MODIFIED;
4863         mac_table->state &= ~IXGBE_MAC_STATE_IN_USE;
4864 
4865         ixgbe_sync_mac_table(adapter);
4866 
4867         return 0;
4868     }
4869 
4870     return -ENOMEM;
4871 }
4872 
4873 static int ixgbe_uc_sync(struct net_device *netdev, const unsigned char *addr)
4874 {
4875     struct ixgbe_adapter *adapter = netdev_priv(netdev);
4876     int ret;
4877 
4878     ret = ixgbe_add_mac_filter(adapter, addr, VMDQ_P(0));
4879 
4880     return min_t(int, ret, 0);
4881 }
4882 
4883 static int ixgbe_uc_unsync(struct net_device *netdev, const unsigned char *addr)
4884 {
4885     struct ixgbe_adapter *adapter = netdev_priv(netdev);
4886 
4887     ixgbe_del_mac_filter(adapter, addr, VMDQ_P(0));
4888 
4889     return 0;
4890 }
4891 
4892 /**
4893  * ixgbe_set_rx_mode - Unicast, Multicast and Promiscuous mode set
4894  * @netdev: network interface device structure
4895  *
4896  * The set_rx_method entry point is called whenever the unicast/multicast
4897  * address list or the network interface flags are updated.  This routine is
4898  * responsible for configuring the hardware for proper unicast, multicast and
4899  * promiscuous mode.
4900  **/
4901 void ixgbe_set_rx_mode(struct net_device *netdev)
4902 {
4903     struct ixgbe_adapter *adapter = netdev_priv(netdev);
4904     struct ixgbe_hw *hw = &adapter->hw;
4905     u32 fctrl, vmolr = IXGBE_VMOLR_BAM | IXGBE_VMOLR_AUPE;
4906     netdev_features_t features = netdev->features;
4907     int count;
4908 
4909     /* Check for Promiscuous and All Multicast modes */
4910     fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4911 
4912     /* set all bits that we expect to always be set */
4913     fctrl &= ~IXGBE_FCTRL_SBP; /* disable store-bad-packets */
4914     fctrl |= IXGBE_FCTRL_BAM;
4915     fctrl |= IXGBE_FCTRL_DPF; /* discard pause frames when FC enabled */
4916     fctrl |= IXGBE_FCTRL_PMCF;
4917 
4918     /* clear the bits we are changing the status of */
4919     fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4920     if (netdev->flags & IFF_PROMISC) {
4921         hw->addr_ctrl.user_set_promisc = true;
4922         fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4923         vmolr |= IXGBE_VMOLR_MPE;
4924         features &= ~NETIF_F_HW_VLAN_CTAG_FILTER;
4925     } else {
4926         if (netdev->flags & IFF_ALLMULTI) {
4927             fctrl |= IXGBE_FCTRL_MPE;
4928             vmolr |= IXGBE_VMOLR_MPE;
4929         }
4930         hw->addr_ctrl.user_set_promisc = false;
4931     }
4932 
4933     /*
4934      * Write addresses to available RAR registers, if there is not
4935      * sufficient space to store all the addresses then enable
4936      * unicast promiscuous mode
4937      */
4938     if (__dev_uc_sync(netdev, ixgbe_uc_sync, ixgbe_uc_unsync)) {
4939         fctrl |= IXGBE_FCTRL_UPE;
4940         vmolr |= IXGBE_VMOLR_ROPE;
4941     }
4942 
4943     /* Write addresses to the MTA, if the attempt fails
4944      * then we should just turn on promiscuous mode so
4945      * that we can at least receive multicast traffic
4946      */
4947     count = ixgbe_write_mc_addr_list(netdev);
4948     if (count < 0) {
4949         fctrl |= IXGBE_FCTRL_MPE;
4950         vmolr |= IXGBE_VMOLR_MPE;
4951     } else if (count) {
4952         vmolr |= IXGBE_VMOLR_ROMPE;
4953     }
4954 
4955     if (hw->mac.type != ixgbe_mac_82598EB) {
4956         vmolr |= IXGBE_READ_REG(hw, IXGBE_VMOLR(VMDQ_P(0))) &
4957              ~(IXGBE_VMOLR_MPE | IXGBE_VMOLR_ROMPE |
4958                IXGBE_VMOLR_ROPE);
4959         IXGBE_WRITE_REG(hw, IXGBE_VMOLR(VMDQ_P(0)), vmolr);
4960     }
4961 
4962     /* This is useful for sniffing bad packets. */
4963     if (features & NETIF_F_RXALL) {
4964         /* UPE and MPE will be handled by normal PROMISC logic
4965          * in e1000e_set_rx_mode */
4966         fctrl |= (IXGBE_FCTRL_SBP | /* Receive bad packets */
4967               IXGBE_FCTRL_BAM | /* RX All Bcast Pkts */
4968               IXGBE_FCTRL_PMCF); /* RX All MAC Ctrl Pkts */
4969 
4970         fctrl &= ~(IXGBE_FCTRL_DPF);
4971         /* NOTE:  VLAN filtering is disabled by setting PROMISC */
4972     }
4973 
4974     IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4975 
4976     if (features & NETIF_F_HW_VLAN_CTAG_RX)
4977         ixgbe_vlan_strip_enable(adapter);
4978     else
4979         ixgbe_vlan_strip_disable(adapter);
4980 
4981     if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
4982         ixgbe_vlan_promisc_disable(adapter);
4983     else
4984         ixgbe_vlan_promisc_enable(adapter);
4985 }
4986 
4987 static void ixgbe_napi_enable_all(struct ixgbe_adapter *adapter)
4988 {
4989     int q_idx;
4990 
4991     for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
4992         napi_enable(&adapter->q_vector[q_idx]->napi);
4993 }
4994 
4995 static void ixgbe_napi_disable_all(struct ixgbe_adapter *adapter)
4996 {
4997     int q_idx;
4998 
4999     for (q_idx = 0; q_idx < adapter->num_q_vectors; q_idx++)
5000         napi_disable(&adapter->q_vector[q_idx]->napi);
5001 }
5002 
5003 static int ixgbe_udp_tunnel_sync(struct net_device *dev, unsigned int table)
5004 {
5005     struct ixgbe_adapter *adapter = netdev_priv(dev);
5006     struct ixgbe_hw *hw = &adapter->hw;
5007     struct udp_tunnel_info ti;
5008 
5009     udp_tunnel_nic_get_port(dev, table, 0, &ti);
5010     if (ti.type == UDP_TUNNEL_TYPE_VXLAN)
5011         adapter->vxlan_port = ti.port;
5012     else
5013         adapter->geneve_port = ti.port;
5014 
5015     IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL,
5016             ntohs(adapter->vxlan_port) |
5017             ntohs(adapter->geneve_port) <<
5018                 IXGBE_VXLANCTRL_GENEVE_UDPPORT_SHIFT);
5019     return 0;
5020 }
5021 
5022 static const struct udp_tunnel_nic_info ixgbe_udp_tunnels_x550 = {
5023     .sync_table = ixgbe_udp_tunnel_sync,
5024     .flags      = UDP_TUNNEL_NIC_INFO_IPV4_ONLY,
5025     .tables     = {
5026         { .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
5027     },
5028 };
5029 
5030 static const struct udp_tunnel_nic_info ixgbe_udp_tunnels_x550em_a = {
5031     .sync_table = ixgbe_udp_tunnel_sync,
5032     .flags      = UDP_TUNNEL_NIC_INFO_IPV4_ONLY,
5033     .tables     = {
5034         { .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_VXLAN,  },
5035         { .n_entries = 1, .tunnel_types = UDP_TUNNEL_TYPE_GENEVE, },
5036     },
5037 };
5038 
5039 #ifdef CONFIG_IXGBE_DCB
5040 /**
5041  * ixgbe_configure_dcb - Configure DCB hardware
5042  * @adapter: ixgbe adapter struct
5043  *
5044  * This is called by the driver on open to configure the DCB hardware.
5045  * This is also called by the gennetlink interface when reconfiguring
5046  * the DCB state.
5047  */
5048 static void ixgbe_configure_dcb(struct ixgbe_adapter *adapter)
5049 {
5050     struct ixgbe_hw *hw = &adapter->hw;
5051     int max_frame = adapter->netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
5052 
5053     if (!(adapter->flags & IXGBE_FLAG_DCB_ENABLED)) {
5054         if (hw->mac.type == ixgbe_mac_82598EB)
5055             netif_set_tso_max_size(adapter->netdev, 65536);
5056         return;
5057     }
5058 
5059     if (hw->mac.type == ixgbe_mac_82598EB)
5060         netif_set_tso_max_size(adapter->netdev, 32768);
5061 
5062 #ifdef IXGBE_FCOE
5063     if (adapter->netdev->features & NETIF_F_FCOE_MTU)
5064         max_frame = max(max_frame, IXGBE_FCOE_JUMBO_FRAME_SIZE);
5065 #endif
5066 
5067     /* reconfigure the hardware */
5068     if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE) {
5069         ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5070                         DCB_TX_CONFIG);
5071         ixgbe_dcb_calculate_tc_credits(hw, &adapter->dcb_cfg, max_frame,
5072                         DCB_RX_CONFIG);
5073         ixgbe_dcb_hw_config(hw, &adapter->dcb_cfg);
5074     } else if (adapter->ixgbe_ieee_ets && adapter->ixgbe_ieee_pfc) {
5075         ixgbe_dcb_hw_ets(&adapter->hw,
5076                  adapter->ixgbe_ieee_ets,
5077                  max_frame);
5078         ixgbe_dcb_hw_pfc_config(&adapter->hw,
5079                     adapter->ixgbe_ieee_pfc->pfc_en,
5080                     adapter->ixgbe_ieee_ets->prio_tc);
5081     }
5082 
5083     /* Enable RSS Hash per TC */
5084     if (hw->mac.type != ixgbe_mac_82598EB) {
5085         u32 msb = 0;
5086         u16 rss_i = adapter->ring_feature[RING_F_RSS].indices - 1;
5087 
5088         while (rss_i) {
5089             msb++;
5090             rss_i >>= 1;
5091         }
5092 
5093         /* write msb to all 8 TCs in one write */
5094         IXGBE_WRITE_REG(hw, IXGBE_RQTC, msb * 0x11111111);
5095     }
5096 }
5097 #endif
5098 
5099 /* Additional bittime to account for IXGBE framing */
5100 #define IXGBE_ETH_FRAMING 20
5101 
5102 /**
5103  * ixgbe_hpbthresh - calculate high water mark for flow control
5104  *
5105  * @adapter: board private structure to calculate for
5106  * @pb: packet buffer to calculate
5107  */
5108 static int ixgbe_hpbthresh(struct ixgbe_adapter *adapter, int pb)
5109 {
5110     struct ixgbe_hw *hw = &adapter->hw;
5111     struct net_device *dev = adapter->netdev;
5112     int link, tc, kb, marker;
5113     u32 dv_id, rx_pba;
5114 
5115     /* Calculate max LAN frame size */
5116     tc = link = dev->mtu + ETH_HLEN + ETH_FCS_LEN + IXGBE_ETH_FRAMING;
5117 
5118 #ifdef IXGBE_FCOE
5119     /* FCoE traffic class uses FCOE jumbo frames */
5120     if ((dev->features & NETIF_F_FCOE_MTU) &&
5121         (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5122         (pb == ixgbe_fcoe_get_tc(adapter)))
5123         tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5124 #endif
5125 
5126     /* Calculate delay value for device */
5127     switch (hw->mac.type) {
5128     case ixgbe_mac_X540:
5129     case ixgbe_mac_X550:
5130     case ixgbe_mac_X550EM_x:
5131     case ixgbe_mac_x550em_a:
5132         dv_id = IXGBE_DV_X540(link, tc);
5133         break;
5134     default:
5135         dv_id = IXGBE_DV(link, tc);
5136         break;
5137     }
5138 
5139     /* Loopback switch introduces additional latency */
5140     if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5141         dv_id += IXGBE_B2BT(tc);
5142 
5143     /* Delay value is calculated in bit times convert to KB */
5144     kb = IXGBE_BT2KB(dv_id);
5145     rx_pba = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(pb)) >> 10;
5146 
5147     marker = rx_pba - kb;
5148 
5149     /* It is possible that the packet buffer is not large enough
5150      * to provide required headroom. In this case throw an error
5151      * to user and a do the best we can.
5152      */
5153     if (marker < 0) {
5154         e_warn(drv, "Packet Buffer(%i) can not provide enough"
5155                 "headroom to support flow control."
5156                 "Decrease MTU or number of traffic classes\n", pb);
5157         marker = tc + 1;
5158     }
5159 
5160     return marker;
5161 }
5162 
5163 /**
5164  * ixgbe_lpbthresh - calculate low water mark for flow control
5165  *
5166  * @adapter: board private structure to calculate for
5167  * @pb: packet buffer to calculate
5168  */
5169 static int ixgbe_lpbthresh(struct ixgbe_adapter *adapter, int pb)
5170 {
5171     struct ixgbe_hw *hw = &adapter->hw;
5172     struct net_device *dev = adapter->netdev;
5173     int tc;
5174     u32 dv_id;
5175 
5176     /* Calculate max LAN frame size */
5177     tc = dev->mtu + ETH_HLEN + ETH_FCS_LEN;
5178 
5179 #ifdef IXGBE_FCOE
5180     /* FCoE traffic class uses FCOE jumbo frames */
5181     if ((dev->features & NETIF_F_FCOE_MTU) &&
5182         (tc < IXGBE_FCOE_JUMBO_FRAME_SIZE) &&
5183         (pb == netdev_get_prio_tc_map(dev, adapter->fcoe.up)))
5184         tc = IXGBE_FCOE_JUMBO_FRAME_SIZE;
5185 #endif
5186 
5187     /* Calculate delay value for device */
5188     switch (hw->mac.type) {
5189     case ixgbe_mac_X540:
5190     case ixgbe_mac_X550:
5191     case ixgbe_mac_X550EM_x:
5192     case ixgbe_mac_x550em_a:
5193         dv_id = IXGBE_LOW_DV_X540(tc);
5194         break;
5195     default:
5196         dv_id = IXGBE_LOW_DV(tc);
5197         break;
5198     }
5199 
5200     /* Delay value is calculated in bit times convert to KB */
5201     return IXGBE_BT2KB(dv_id);
5202 }
5203 
5204 /*
5205  * ixgbe_pbthresh_setup - calculate and setup high low water marks
5206  */
5207 static void ixgbe_pbthresh_setup(struct ixgbe_adapter *adapter)
5208 {
5209     struct ixgbe_hw *hw = &adapter->hw;
5210     int num_tc = adapter->hw_tcs;
5211     int i;
5212 
5213     if (!num_tc)
5214         num_tc = 1;
5215 
5216     for (i = 0; i < num_tc; i++) {
5217         hw->fc.high_water[i] = ixgbe_hpbthresh(adapter, i);
5218         hw->fc.low_water[i] = ixgbe_lpbthresh(adapter, i);
5219 
5220         /* Low water marks must not be larger than high water marks */
5221         if (hw->fc.low_water[i] > hw->fc.high_water[i])
5222             hw->fc.low_water[i] = 0;
5223     }
5224 
5225     for (; i < MAX_TRAFFIC_CLASS; i++)
5226         hw->fc.high_water[i] = 0;
5227 }
5228 
5229 static void ixgbe_configure_pb(struct ixgbe_adapter *adapter)
5230 {
5231     struct ixgbe_hw *hw = &adapter->hw;
5232     int hdrm;
5233     u8 tc = adapter->hw_tcs;
5234 
5235     if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE ||
5236         adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
5237         hdrm = 32 << adapter->fdir_pballoc;
5238     else
5239         hdrm = 0;
5240 
5241     hw->mac.ops.set_rxpba(hw, tc, hdrm, PBA_STRATEGY_EQUAL);
5242     ixgbe_pbthresh_setup(adapter);
5243 }
5244 
5245 static void ixgbe_fdir_filter_restore(struct ixgbe_adapter *adapter)
5246 {
5247     struct ixgbe_hw *hw = &adapter->hw;
5248     struct hlist_node *node2;
5249     struct ixgbe_fdir_filter *filter;
5250     u8 queue;
5251 
5252     spin_lock(&adapter->fdir_perfect_lock);
5253 
5254     if (!hlist_empty(&adapter->fdir_filter_list))
5255         ixgbe_fdir_set_input_mask_82599(hw, &adapter->fdir_mask);
5256 
5257     hlist_for_each_entry_safe(filter, node2,
5258                   &adapter->fdir_filter_list, fdir_node) {
5259         if (filter->action == IXGBE_FDIR_DROP_QUEUE) {
5260             queue = IXGBE_FDIR_DROP_QUEUE;
5261         } else {
5262             u32 ring = ethtool_get_flow_spec_ring(filter->action);
5263             u8 vf = ethtool_get_flow_spec_ring_vf(filter->action);
5264 
5265             if (!vf && (ring >= adapter->num_rx_queues)) {
5266                 e_err(drv, "FDIR restore failed without VF, ring: %u\n",
5267                       ring);
5268                 continue;
5269             } else if (vf &&
5270                    ((vf > adapter->num_vfs) ||
5271                      ring >= adapter->num_rx_queues_per_pool)) {
5272                 e_err(drv, "FDIR restore failed with VF, vf: %hhu, ring: %u\n",
5273                       vf, ring);
5274                 continue;
5275             }
5276 
5277             /* Map the ring onto the absolute queue index */
5278             if (!vf)
5279                 queue = adapter->rx_ring[ring]->reg_idx;
5280             else
5281                 queue = ((vf - 1) *
5282                     adapter->num_rx_queues_per_pool) + ring;
5283         }
5284 
5285         ixgbe_fdir_write_perfect_filter_82599(hw,
5286                 &filter->filter, filter->sw_idx, queue);
5287     }
5288 
5289     spin_unlock(&adapter->fdir_perfect_lock);
5290 }
5291 
5292 /**
5293  * ixgbe_clean_rx_ring - Free Rx Buffers per Queue
5294  * @rx_ring: ring to free buffers from
5295  **/
5296 static void ixgbe_clean_rx_ring(struct ixgbe_ring *rx_ring)
5297 {
5298     u16 i = rx_ring->next_to_clean;
5299     struct ixgbe_rx_buffer *rx_buffer = &rx_ring->rx_buffer_info[i];
5300 
5301     if (rx_ring->xsk_pool) {
5302         ixgbe_xsk_clean_rx_ring(rx_ring);
5303         goto skip_free;
5304     }
5305 
5306     /* Free all the Rx ring sk_buffs */
5307     while (i != rx_ring->next_to_alloc) {
5308         if (rx_buffer->skb) {
5309             struct sk_buff *skb = rx_buffer->skb;
5310             if (IXGBE_CB(skb)->page_released)
5311                 dma_unmap_page_attrs(rx_ring->dev,
5312                              IXGBE_CB(skb)->dma,
5313                              ixgbe_rx_pg_size(rx_ring),
5314                              DMA_FROM_DEVICE,
5315                              IXGBE_RX_DMA_ATTR);
5316             dev_kfree_skb(skb);
5317         }
5318 
5319         /* Invalidate cache lines that may have been written to by
5320          * device so that we avoid corrupting memory.
5321          */
5322         dma_sync_single_range_for_cpu(rx_ring->dev,
5323                           rx_buffer->dma,
5324                           rx_buffer->page_offset,
5325                           ixgbe_rx_bufsz(rx_ring),
5326                           DMA_FROM_DEVICE);
5327 
5328         /* free resources associated with mapping */
5329         dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
5330                      ixgbe_rx_pg_size(rx_ring),
5331                      DMA_FROM_DEVICE,
5332                      IXGBE_RX_DMA_ATTR);
5333         __page_frag_cache_drain(rx_buffer->page,
5334                     rx_buffer->pagecnt_bias);
5335 
5336         i++;
5337         rx_buffer++;
5338         if (i == rx_ring->count) {
5339             i = 0;
5340             rx_buffer = rx_ring->rx_buffer_info;
5341         }
5342     }
5343 
5344 skip_free:
5345     rx_ring->next_to_alloc = 0;
5346     rx_ring->next_to_clean = 0;
5347     rx_ring->next_to_use = 0;
5348 }
5349 
5350 static int ixgbe_fwd_ring_up(struct ixgbe_adapter *adapter,
5351                  struct ixgbe_fwd_adapter *accel)
5352 {
5353     u16 rss_i = adapter->ring_feature[RING_F_RSS].indices;
5354     int num_tc = netdev_get_num_tc(adapter->netdev);
5355     struct net_device *vdev = accel->netdev;
5356     int i, baseq, err;
5357 
5358     baseq = accel->pool * adapter->num_rx_queues_per_pool;
5359     netdev_dbg(vdev, "pool %i:%i queues %i:%i\n",
5360            accel->pool, adapter->num_rx_pools,
5361            baseq, baseq + adapter->num_rx_queues_per_pool);
5362 
5363     accel->rx_base_queue = baseq;
5364     accel->tx_base_queue = baseq;
5365 
5366     /* record configuration for macvlan interface in vdev */
5367     for (i = 0; i < num_tc; i++)
5368         netdev_bind_sb_channel_queue(adapter->netdev, vdev,
5369                          i, rss_i, baseq + (rss_i * i));
5370 
5371     for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5372         adapter->rx_ring[baseq + i]->netdev = vdev;
5373 
5374     /* Guarantee all rings are updated before we update the
5375      * MAC address filter.
5376      */
5377     wmb();
5378 
5379     /* ixgbe_add_mac_filter will return an index if it succeeds, so we
5380      * need to only treat it as an error value if it is negative.
5381      */
5382     err = ixgbe_add_mac_filter(adapter, vdev->dev_addr,
5383                    VMDQ_P(accel->pool));
5384     if (err >= 0)
5385         return 0;
5386 
5387     /* if we cannot add the MAC rule then disable the offload */
5388     macvlan_release_l2fw_offload(vdev);
5389 
5390     for (i = 0; i < adapter->num_rx_queues_per_pool; i++)
5391         adapter->rx_ring[baseq + i]->netdev = NULL;
5392 
5393     netdev_err(vdev, "L2FW offload disabled due to L2 filter error\n");
5394 
5395     /* unbind the queues and drop the subordinate channel config */
5396     netdev_unbind_sb_channel(adapter->netdev, vdev);
5397     netdev_set_sb_channel(vdev, 0);
5398 
5399     clear_bit(accel->pool, adapter->fwd_bitmask);
5400     kfree(accel);
5401 
5402     return err;
5403 }
5404 
5405 static int ixgbe_macvlan_up(struct net_device *vdev,
5406                 struct netdev_nested_priv *priv)
5407 {
5408     struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)priv->data;
5409     struct ixgbe_fwd_adapter *accel;
5410 
5411     if (!netif_is_macvlan(vdev))
5412         return 0;
5413 
5414     accel = macvlan_accel_priv(vdev);
5415     if (!accel)
5416         return 0;
5417 
5418     ixgbe_fwd_ring_up(adapter, accel);
5419 
5420     return 0;
5421 }
5422 
5423 static void ixgbe_configure_dfwd(struct ixgbe_adapter *adapter)
5424 {
5425     struct netdev_nested_priv priv = {
5426         .data = (void *)adapter,
5427     };
5428 
5429     netdev_walk_all_upper_dev_rcu(adapter->netdev,
5430                       ixgbe_macvlan_up, &priv);
5431 }
5432 
5433 static void ixgbe_configure(struct ixgbe_adapter *adapter)
5434 {
5435     struct ixgbe_hw *hw = &adapter->hw;
5436 
5437     ixgbe_configure_pb(adapter);
5438 #ifdef CONFIG_IXGBE_DCB
5439     ixgbe_configure_dcb(adapter);
5440 #endif
5441     /*
5442      * We must restore virtualization before VLANs or else
5443      * the VLVF registers will not be populated
5444      */
5445     ixgbe_configure_virtualization(adapter);
5446 
5447     ixgbe_set_rx_mode(adapter->netdev);
5448     ixgbe_restore_vlan(adapter);
5449     ixgbe_ipsec_restore(adapter);
5450 
5451     switch (hw->mac.type) {
5452     case ixgbe_mac_82599EB:
5453     case ixgbe_mac_X540:
5454         hw->mac.ops.disable_rx_buff(hw);
5455         break;
5456     default:
5457         break;
5458     }
5459 
5460     if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
5461         ixgbe_init_fdir_signature_82599(&adapter->hw,
5462                         adapter->fdir_pballoc);
5463     } else if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE) {
5464         ixgbe_init_fdir_perfect_82599(&adapter->hw,
5465                           adapter->fdir_pballoc);
5466         ixgbe_fdir_filter_restore(adapter);
5467     }
5468 
5469     switch (hw->mac.type) {
5470     case ixgbe_mac_82599EB:
5471     case ixgbe_mac_X540:
5472         hw->mac.ops.enable_rx_buff(hw);
5473         break;
5474     default:
5475         break;
5476     }
5477 
5478 #ifdef CONFIG_IXGBE_DCA
5479     /* configure DCA */
5480     if (adapter->flags & IXGBE_FLAG_DCA_CAPABLE)
5481         ixgbe_setup_dca(adapter);
5482 #endif /* CONFIG_IXGBE_DCA */
5483 
5484 #ifdef IXGBE_FCOE
5485     /* configure FCoE L2 filters, redirection table, and Rx control */
5486     ixgbe_configure_fcoe(adapter);
5487 
5488 #endif /* IXGBE_FCOE */
5489     ixgbe_configure_tx(adapter);
5490     ixgbe_configure_rx(adapter);
5491     ixgbe_configure_dfwd(adapter);
5492 }
5493 
5494 /**
5495  * ixgbe_sfp_link_config - set up SFP+ link
5496  * @adapter: pointer to private adapter struct
5497  **/
5498 static void ixgbe_sfp_link_config(struct ixgbe_adapter *adapter)
5499 {
5500     /*
5501      * We are assuming the worst case scenario here, and that
5502      * is that an SFP was inserted/removed after the reset
5503      * but before SFP detection was enabled.  As such the best
5504      * solution is to just start searching as soon as we start
5505      */
5506     if (adapter->hw.mac.type == ixgbe_mac_82598EB)
5507         adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
5508 
5509     adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
5510     adapter->sfp_poll_time = 0;
5511 }
5512 
5513 /**
5514  * ixgbe_non_sfp_link_config - set up non-SFP+ link
5515  * @hw: pointer to private hardware struct
5516  *
5517  * Returns 0 on success, negative on failure
5518  **/
5519 static int ixgbe_non_sfp_link_config(struct ixgbe_hw *hw)
5520 {
5521     u32 speed;
5522     bool autoneg, link_up = false;
5523     int ret = IXGBE_ERR_LINK_SETUP;
5524 
5525     if (hw->mac.ops.check_link)
5526         ret = hw->mac.ops.check_link(hw, &speed, &link_up, false);
5527 
5528     if (ret)
5529         return ret;
5530 
5531     speed = hw->phy.autoneg_advertised;
5532     if (!speed && hw->mac.ops.get_link_capabilities) {
5533         ret = hw->mac.ops.get_link_capabilities(hw, &speed,
5534                             &autoneg);
5535         /* remove NBASE-T speeds from default autonegotiation
5536          * to accommodate broken network switches in the field
5537          * which cannot cope with advertised NBASE-T speeds
5538          */
5539         speed &= ~(IXGBE_LINK_SPEED_5GB_FULL |
5540                IXGBE_LINK_SPEED_2_5GB_FULL);
5541     }
5542 
5543     if (ret)
5544         return ret;
5545 
5546     if (hw->mac.ops.setup_link)
5547         ret = hw->mac.ops.setup_link(hw, speed, link_up);
5548 
5549     return ret;
5550 }
5551 
5552 /**
5553  * ixgbe_clear_vf_stats_counters - Clear out VF stats after reset
5554  * @adapter: board private structure
5555  *
5556  * On a reset we need to clear out the VF stats or accounting gets
5557  * messed up because they're not clear on read.
5558  **/
5559 static void ixgbe_clear_vf_stats_counters(struct ixgbe_adapter *adapter)
5560 {
5561     struct ixgbe_hw *hw = &adapter->hw;
5562     int i;
5563 
5564     for (i = 0; i < adapter->num_vfs; i++) {
5565         adapter->vfinfo[i].last_vfstats.gprc =
5566             IXGBE_READ_REG(hw, IXGBE_PVFGPRC(i));
5567         adapter->vfinfo[i].saved_rst_vfstats.gprc +=
5568             adapter->vfinfo[i].vfstats.gprc;
5569         adapter->vfinfo[i].vfstats.gprc = 0;
5570         adapter->vfinfo[i].last_vfstats.gptc =
5571             IXGBE_READ_REG(hw, IXGBE_PVFGPTC(i));
5572         adapter->vfinfo[i].saved_rst_vfstats.gptc +=
5573             adapter->vfinfo[i].vfstats.gptc;
5574         adapter->vfinfo[i].vfstats.gptc = 0;
5575         adapter->vfinfo[i].last_vfstats.gorc =
5576             IXGBE_READ_REG(hw, IXGBE_PVFGORC_LSB(i));
5577         adapter->vfinfo[i].saved_rst_vfstats.gorc +=
5578             adapter->vfinfo[i].vfstats.gorc;
5579         adapter->vfinfo[i].vfstats.gorc = 0;
5580         adapter->vfinfo[i].last_vfstats.gotc =
5581             IXGBE_READ_REG(hw, IXGBE_PVFGOTC_LSB(i));
5582         adapter->vfinfo[i].saved_rst_vfstats.gotc +=
5583             adapter->vfinfo[i].vfstats.gotc;
5584         adapter->vfinfo[i].vfstats.gotc = 0;
5585         adapter->vfinfo[i].last_vfstats.mprc =
5586             IXGBE_READ_REG(hw, IXGBE_PVFMPRC(i));
5587         adapter->vfinfo[i].saved_rst_vfstats.mprc +=
5588             adapter->vfinfo[i].vfstats.mprc;
5589         adapter->vfinfo[i].vfstats.mprc = 0;
5590     }
5591 }
5592 
5593 static void ixgbe_setup_gpie(struct ixgbe_adapter *adapter)
5594 {
5595     struct ixgbe_hw *hw = &adapter->hw;
5596     u32 gpie = 0;
5597 
5598     if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED) {
5599         gpie = IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
5600                IXGBE_GPIE_OCD;
5601         gpie |= IXGBE_GPIE_EIAME;
5602         /*
5603          * use EIAM to auto-mask when MSI-X interrupt is asserted
5604          * this saves a register write for every interrupt
5605          */
5606         switch (hw->mac.type) {
5607         case ixgbe_mac_82598EB:
5608             IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5609             break;
5610         case ixgbe_mac_82599EB:
5611         case ixgbe_mac_X540:
5612         case ixgbe_mac_X550:
5613         case ixgbe_mac_X550EM_x:
5614         case ixgbe_mac_x550em_a:
5615         default:
5616             IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
5617             IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
5618             break;
5619         }
5620     } else {
5621         /* legacy interrupts, use EIAM to auto-mask when reading EICR,
5622          * specifically only auto mask tx and rx interrupts */
5623         IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
5624     }
5625 
5626     /* XXX: to interrupt immediately for EICS writes, enable this */
5627     /* gpie |= IXGBE_GPIE_EIMEN; */
5628 
5629     if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
5630         gpie &= ~IXGBE_GPIE_VTMODE_MASK;
5631 
5632         switch (adapter->ring_feature[RING_F_VMDQ].mask) {
5633         case IXGBE_82599_VMDQ_8Q_MASK:
5634             gpie |= IXGBE_GPIE_VTMODE_16;
5635             break;
5636         case IXGBE_82599_VMDQ_4Q_MASK:
5637             gpie |= IXGBE_GPIE_VTMODE_32;
5638             break;
5639         default:
5640             gpie |= IXGBE_GPIE_VTMODE_64;
5641             break;
5642         }
5643     }
5644 
5645     /* Enable Thermal over heat sensor interrupt */
5646     if (adapter->flags2 & IXGBE_FLAG2_TEMP_SENSOR_CAPABLE) {
5647         switch (adapter->hw.mac.type) {
5648         case ixgbe_mac_82599EB:
5649             gpie |= IXGBE_SDP0_GPIEN_8259X;
5650             break;
5651         default:
5652             break;
5653         }
5654     }
5655 
5656     /* Enable fan failure interrupt */
5657     if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE)
5658         gpie |= IXGBE_SDP1_GPIEN(hw);
5659 
5660     switch (hw->mac.type) {
5661     case ixgbe_mac_82599EB:
5662         gpie |= IXGBE_SDP1_GPIEN_8259X | IXGBE_SDP2_GPIEN_8259X;
5663         break;
5664     case ixgbe_mac_X550EM_x:
5665     case ixgbe_mac_x550em_a:
5666         gpie |= IXGBE_SDP0_GPIEN_X540;
5667         break;
5668     default:
5669         break;
5670     }
5671 
5672     IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
5673 }
5674 
5675 static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
5676 {
5677     struct ixgbe_hw *hw = &adapter->hw;
5678     int err;
5679     u32 ctrl_ext;
5680 
5681     ixgbe_get_hw_control(adapter);
5682     ixgbe_setup_gpie(adapter);
5683 
5684     if (adapter->flags & IXGBE_FLAG_MSIX_ENABLED)
5685         ixgbe_configure_msix(adapter);
5686     else
5687         ixgbe_configure_msi_and_legacy(adapter);
5688 
5689     /* enable the optics for 82599 SFP+ fiber */
5690     if (hw->mac.ops.enable_tx_laser)
5691         hw->mac.ops.enable_tx_laser(hw);
5692 
5693     if (hw->phy.ops.set_phy_power)
5694         hw->phy.ops.set_phy_power(hw, true);
5695 
5696     smp_mb__before_atomic();
5697     clear_bit(__IXGBE_DOWN, &adapter->state);
5698     ixgbe_napi_enable_all(adapter);
5699 
5700     if (ixgbe_is_sfp(hw)) {
5701         ixgbe_sfp_link_config(adapter);
5702     } else {
5703         err = ixgbe_non_sfp_link_config(hw);
5704         if (err)
5705             e_err(probe, "link_config FAILED %d\n", err);
5706     }
5707 
5708     /* clear any pending interrupts, may auto mask */
5709     IXGBE_READ_REG(hw, IXGBE_EICR);
5710     ixgbe_irq_enable(adapter, true, true);
5711 
5712     /*
5713      * If this adapter has a fan, check to see if we had a failure
5714      * before we enabled the interrupt.
5715      */
5716     if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
5717         u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
5718         if (esdp & IXGBE_ESDP_SDP1)
5719             e_crit(drv, "Fan has stopped, replace the adapter\n");
5720     }
5721 
5722     /* bring the link up in the watchdog, this could race with our first
5723      * link up interrupt but shouldn't be a problem */
5724     adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
5725     adapter->link_check_timeout = jiffies;
5726     mod_timer(&adapter->service_timer, jiffies);
5727 
5728     ixgbe_clear_vf_stats_counters(adapter);
5729     /* Set PF Reset Done bit so PF/VF Mail Ops can work */
5730     ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
5731     ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
5732     IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
5733 
5734     /* update setting rx tx for all active vfs */
5735     ixgbe_set_all_vfs(adapter);
5736 }
5737 
5738 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter)
5739 {
5740     /* put off any impending NetWatchDogTimeout */
5741     netif_trans_update(adapter->netdev);
5742 
5743     while (test_and_set_bit(__IXGBE_RESETTING, &adapter->state))
5744         usleep_range(1000, 2000);
5745     if (adapter->hw.phy.type == ixgbe_phy_fw)
5746         ixgbe_watchdog_link_is_down(adapter);
5747     ixgbe_down(adapter);
5748     /*
5749      * If SR-IOV enabled then wait a bit before bringing the adapter
5750      * back up to give the VFs time to respond to the reset.  The
5751      * two second wait is based upon the watchdog timer cycle in
5752      * the VF driver.
5753      */
5754     if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
5755         msleep(2000);
5756     ixgbe_up(adapter);
5757     clear_bit(__IXGBE_RESETTING, &adapter->state);
5758 }
5759 
5760 void ixgbe_up(struct ixgbe_adapter *adapter)
5761 {
5762     /* hardware has been reset, we need to reload some things */
5763     ixgbe_configure(adapter);
5764 
5765     ixgbe_up_complete(adapter);
5766 }
5767 
5768 static unsigned long ixgbe_get_completion_timeout(struct ixgbe_adapter *adapter)
5769 {
5770     u16 devctl2;
5771 
5772     pcie_capability_read_word(adapter->pdev, PCI_EXP_DEVCTL2, &devctl2);
5773 
5774     switch (devctl2 & IXGBE_PCIDEVCTRL2_TIMEO_MASK) {
5775     case IXGBE_PCIDEVCTRL2_17_34s:
5776     case IXGBE_PCIDEVCTRL2_4_8s:
5777         /* For now we cap the upper limit on delay to 2 seconds
5778          * as we end up going up to 34 seconds of delay in worst
5779          * case timeout value.
5780          */
5781     case IXGBE_PCIDEVCTRL2_1_2s:
5782         return 2000000ul;   /* 2.0 s */
5783     case IXGBE_PCIDEVCTRL2_260_520ms:
5784         return 520000ul;    /* 520 ms */
5785     case IXGBE_PCIDEVCTRL2_65_130ms:
5786         return 130000ul;    /* 130 ms */
5787     case IXGBE_PCIDEVCTRL2_16_32ms:
5788         return 32000ul;     /* 32 ms */
5789     case IXGBE_PCIDEVCTRL2_1_2ms:
5790         return 2000ul;      /* 2 ms */
5791     case IXGBE_PCIDEVCTRL2_50_100us:
5792         return 100ul;       /* 100 us */
5793     case IXGBE_PCIDEVCTRL2_16_32ms_def:
5794         return 32000ul;     /* 32 ms */
5795     default:
5796         break;
5797     }
5798 
5799     /* We shouldn't need to hit this path, but just in case default as
5800      * though completion timeout is not supported and support 32ms.
5801      */
5802     return 32000ul;
5803 }
5804 
5805 void ixgbe_disable_rx(struct ixgbe_adapter *adapter)
5806 {
5807     unsigned long wait_delay, delay_interval;
5808     struct ixgbe_hw *hw = &adapter->hw;
5809     int i, wait_loop;
5810     u32 rxdctl;
5811 
5812     /* disable receives */
5813     hw->mac.ops.disable_rx(hw);
5814 
5815     if (ixgbe_removed(hw->hw_addr))
5816         return;
5817 
5818     /* disable all enabled Rx queues */
5819     for (i = 0; i < adapter->num_rx_queues; i++) {
5820         struct ixgbe_ring *ring = adapter->rx_ring[i];
5821         u8 reg_idx = ring->reg_idx;
5822 
5823         rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
5824         rxdctl &= ~IXGBE_RXDCTL_ENABLE;
5825         rxdctl |= IXGBE_RXDCTL_SWFLSH;
5826 
5827         /* write value back with RXDCTL.ENABLE bit cleared */
5828         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
5829     }
5830 
5831     /* RXDCTL.EN may not change on 82598 if link is down, so skip it */
5832     if (hw->mac.type == ixgbe_mac_82598EB &&
5833         !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
5834         return;
5835 
5836     /* Determine our minimum delay interval. We will increase this value
5837      * with each subsequent test. This way if the device returns quickly
5838      * we should spend as little time as possible waiting, however as
5839      * the time increases we will wait for larger periods of time.
5840      *
5841      * The trick here is that we increase the interval using the
5842      * following pattern: 1x 3x 5x 7x 9x 11x 13x 15x 17x 19x. The result
5843      * of that wait is that it totals up to 100x whatever interval we
5844      * choose. Since our minimum wait is 100us we can just divide the
5845      * total timeout by 100 to get our minimum delay interval.
5846      */
5847     delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
5848 
5849     wait_loop = IXGBE_MAX_RX_DESC_POLL;
5850     wait_delay = delay_interval;
5851 
5852     while (wait_loop--) {
5853         usleep_range(wait_delay, wait_delay + 10);
5854         wait_delay += delay_interval * 2;
5855         rxdctl = 0;
5856 
5857         /* OR together the reading of all the active RXDCTL registers,
5858          * and then test the result. We need the disable to complete
5859          * before we start freeing the memory and invalidating the
5860          * DMA mappings.
5861          */
5862         for (i = 0; i < adapter->num_rx_queues; i++) {
5863             struct ixgbe_ring *ring = adapter->rx_ring[i];
5864             u8 reg_idx = ring->reg_idx;
5865 
5866             rxdctl |= IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
5867         }
5868 
5869         if (!(rxdctl & IXGBE_RXDCTL_ENABLE))
5870             return;
5871     }
5872 
5873     e_err(drv,
5874           "RXDCTL.ENABLE for one or more queues not cleared within the polling period\n");
5875 }
5876 
5877 void ixgbe_disable_tx(struct ixgbe_adapter *adapter)
5878 {
5879     unsigned long wait_delay, delay_interval;
5880     struct ixgbe_hw *hw = &adapter->hw;
5881     int i, wait_loop;
5882     u32 txdctl;
5883 
5884     if (ixgbe_removed(hw->hw_addr))
5885         return;
5886 
5887     /* disable all enabled Tx queues */
5888     for (i = 0; i < adapter->num_tx_queues; i++) {
5889         struct ixgbe_ring *ring = adapter->tx_ring[i];
5890         u8 reg_idx = ring->reg_idx;
5891 
5892         IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5893     }
5894 
5895     /* disable all enabled XDP Tx queues */
5896     for (i = 0; i < adapter->num_xdp_queues; i++) {
5897         struct ixgbe_ring *ring = adapter->xdp_ring[i];
5898         u8 reg_idx = ring->reg_idx;
5899 
5900         IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
5901     }
5902 
5903     /* If the link is not up there shouldn't be much in the way of
5904      * pending transactions. Those that are left will be flushed out
5905      * when the reset logic goes through the flush sequence to clean out
5906      * the pending Tx transactions.
5907      */
5908     if (!(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
5909         goto dma_engine_disable;
5910 
5911     /* Determine our minimum delay interval. We will increase this value
5912      * with each subsequent test. This way if the device returns quickly
5913      * we should spend as little time as possible waiting, however as
5914      * the time increases we will wait for larger periods of time.
5915      *
5916      * The trick here is that we increase the interval using the
5917      * following pattern: 1x 3x 5x 7x 9x 11x 13x 15x 17x 19x. The result
5918      * of that wait is that it totals up to 100x whatever interval we
5919      * choose. Since our minimum wait is 100us we can just divide the
5920      * total timeout by 100 to get our minimum delay interval.
5921      */
5922     delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
5923 
5924     wait_loop = IXGBE_MAX_RX_DESC_POLL;
5925     wait_delay = delay_interval;
5926 
5927     while (wait_loop--) {
5928         usleep_range(wait_delay, wait_delay + 10);
5929         wait_delay += delay_interval * 2;
5930         txdctl = 0;
5931 
5932         /* OR together the reading of all the active TXDCTL registers,
5933          * and then test the result. We need the disable to complete
5934          * before we start freeing the memory and invalidating the
5935          * DMA mappings.
5936          */
5937         for (i = 0; i < adapter->num_tx_queues; i++) {
5938             struct ixgbe_ring *ring = adapter->tx_ring[i];
5939             u8 reg_idx = ring->reg_idx;
5940 
5941             txdctl |= IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
5942         }
5943         for (i = 0; i < adapter->num_xdp_queues; i++) {
5944             struct ixgbe_ring *ring = adapter->xdp_ring[i];
5945             u8 reg_idx = ring->reg_idx;
5946 
5947             txdctl |= IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
5948         }
5949 
5950         if (!(txdctl & IXGBE_TXDCTL_ENABLE))
5951             goto dma_engine_disable;
5952     }
5953 
5954     e_err(drv,
5955           "TXDCTL.ENABLE for one or more queues not cleared within the polling period\n");
5956 
5957 dma_engine_disable:
5958     /* Disable the Tx DMA engine on 82599 and later MAC */
5959     switch (hw->mac.type) {
5960     case ixgbe_mac_82599EB:
5961     case ixgbe_mac_X540:
5962     case ixgbe_mac_X550:
5963     case ixgbe_mac_X550EM_x:
5964     case ixgbe_mac_x550em_a:
5965         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL,
5966                 (IXGBE_READ_REG(hw, IXGBE_DMATXCTL) &
5967                  ~IXGBE_DMATXCTL_TE));
5968         fallthrough;
5969     default:
5970         break;
5971     }
5972 }
5973 
5974 void ixgbe_reset(struct ixgbe_adapter *adapter)
5975 {
5976     struct ixgbe_hw *hw = &adapter->hw;
5977     struct net_device *netdev = adapter->netdev;
5978     int err;
5979 
5980     if (ixgbe_removed(hw->hw_addr))
5981         return;
5982     /* lock SFP init bit to prevent race conditions with the watchdog */
5983     while (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
5984         usleep_range(1000, 2000);
5985 
5986     /* clear all SFP and link config related flags while holding SFP_INIT */
5987     adapter->flags2 &= ~(IXGBE_FLAG2_SEARCH_FOR_SFP |
5988                  IXGBE_FLAG2_SFP_NEEDS_RESET);
5989     adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
5990 
5991     err = hw->mac.ops.init_hw(hw);
5992     switch (err) {
5993     case 0:
5994     case IXGBE_ERR_SFP_NOT_PRESENT:
5995     case IXGBE_ERR_SFP_NOT_SUPPORTED:
5996         break;
5997     case IXGBE_ERR_PRIMARY_REQUESTS_PENDING:
5998         e_dev_err("primary disable timed out\n");
5999         break;
6000     case IXGBE_ERR_EEPROM_VERSION:
6001         /* We are running on a pre-production device, log a warning */
6002         e_dev_warn("This device is a pre-production adapter/LOM. "
6003                "Please be aware there may be issues associated with "
6004                "your hardware.  If you are experiencing problems "
6005                "please contact your Intel or hardware "
6006                "representative who provided you with this "
6007                "hardware.\n");
6008         break;
6009     default:
6010         e_dev_err("Hardware Error: %d\n", err);
6011     }
6012 
6013     clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
6014 
6015     /* flush entries out of MAC table */
6016     ixgbe_flush_sw_mac_table(adapter);
6017     __dev_uc_unsync(netdev, NULL);
6018 
6019     /* do not flush user set addresses */
6020     ixgbe_mac_set_default_filter(adapter);
6021 
6022     /* update SAN MAC vmdq pool selection */
6023     if (hw->mac.san_mac_rar_index)
6024         hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
6025 
6026     if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
6027         ixgbe_ptp_reset(adapter);
6028 
6029     if (hw->phy.ops.set_phy_power) {
6030         if (!netif_running(adapter->netdev) && !adapter->wol)
6031             hw->phy.ops.set_phy_power(hw, false);
6032         else
6033             hw->phy.ops.set_phy_power(hw, true);
6034     }
6035 }
6036 
6037 /**
6038  * ixgbe_clean_tx_ring - Free Tx Buffers
6039  * @tx_ring: ring to be cleaned
6040  **/
6041 static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
6042 {
6043     u16 i = tx_ring->next_to_clean;
6044     struct ixgbe_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
6045 
6046     if (tx_ring->xsk_pool) {
6047         ixgbe_xsk_clean_tx_ring(tx_ring);
6048         goto out;
6049     }
6050 
6051     while (i != tx_ring->next_to_use) {
6052         union ixgbe_adv_tx_desc *eop_desc, *tx_desc;
6053 
6054         /* Free all the Tx ring sk_buffs */
6055         if (ring_is_xdp(tx_ring))
6056             xdp_return_frame(tx_buffer->xdpf);
6057         else
6058             dev_kfree_skb_any(tx_buffer->skb);
6059 
6060         /* unmap skb header data */
6061         dma_unmap_single(tx_ring->dev,
6062                  dma_unmap_addr(tx_buffer, dma),
6063                  dma_unmap_len(tx_buffer, len),
6064                  DMA_TO_DEVICE);
6065 
6066         /* check for eop_desc to determine the end of the packet */
6067         eop_desc = tx_buffer->next_to_watch;
6068         tx_desc = IXGBE_TX_DESC(tx_ring, i);
6069 
6070         /* unmap remaining buffers */
6071         while (tx_desc != eop_desc) {
6072             tx_buffer++;
6073             tx_desc++;
6074             i++;
6075             if (unlikely(i == tx_ring->count)) {
6076                 i = 0;
6077                 tx_buffer = tx_ring->tx_buffer_info;
6078                 tx_desc = IXGBE_TX_DESC(tx_ring, 0);
6079             }
6080 
6081             /* unmap any remaining paged data */
6082             if (dma_unmap_len(tx_buffer, len))
6083                 dma_unmap_page(tx_ring->dev,
6084                            dma_unmap_addr(tx_buffer, dma),
6085                            dma_unmap_len(tx_buffer, len),
6086                            DMA_TO_DEVICE);
6087         }
6088 
6089         /* move us one more past the eop_desc for start of next pkt */
6090         tx_buffer++;
6091         i++;
6092         if (unlikely(i == tx_ring->count)) {
6093             i = 0;
6094             tx_buffer = tx_ring->tx_buffer_info;
6095         }
6096     }
6097 
6098     /* reset BQL for queue */
6099     if (!ring_is_xdp(tx_ring))
6100         netdev_tx_reset_queue(txring_txq(tx_ring));
6101 
6102 out:
6103     /* reset next_to_use and next_to_clean */
6104     tx_ring->next_to_use = 0;
6105     tx_ring->next_to_clean = 0;
6106 }
6107 
6108 /**
6109  * ixgbe_clean_all_rx_rings - Free Rx Buffers for all queues
6110  * @adapter: board private structure
6111  **/
6112 static void ixgbe_clean_all_rx_rings(struct ixgbe_adapter *adapter)
6113 {
6114     int i;
6115 
6116     for (i = 0; i < adapter->num_rx_queues; i++)
6117         ixgbe_clean_rx_ring(adapter->rx_ring[i]);
6118 }
6119 
6120 /**
6121  * ixgbe_clean_all_tx_rings - Free Tx Buffers for all queues
6122  * @adapter: board private structure
6123  **/
6124 static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
6125 {
6126     int i;
6127 
6128     for (i = 0; i < adapter->num_tx_queues; i++)
6129         ixgbe_clean_tx_ring(adapter->tx_ring[i]);
6130     for (i = 0; i < adapter->num_xdp_queues; i++)
6131         ixgbe_clean_tx_ring(adapter->xdp_ring[i]);
6132 }
6133 
6134 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
6135 {
6136     struct hlist_node *node2;
6137     struct ixgbe_fdir_filter *filter;
6138 
6139     spin_lock(&adapter->fdir_perfect_lock);
6140 
6141     hlist_for_each_entry_safe(filter, node2,
6142                   &adapter->fdir_filter_list, fdir_node) {
6143         hlist_del(&filter->fdir_node);
6144         kfree(filter);
6145     }
6146     adapter->fdir_filter_count = 0;
6147 
6148     spin_unlock(&adapter->fdir_perfect_lock);
6149 }
6150 
6151 void ixgbe_down(struct ixgbe_adapter *adapter)
6152 {
6153     struct net_device *netdev = adapter->netdev;
6154     struct ixgbe_hw *hw = &adapter->hw;
6155     int i;
6156 
6157     /* signal that we are down to the interrupt handler */
6158     if (test_and_set_bit(__IXGBE_DOWN, &adapter->state))
6159         return; /* do nothing if already down */
6160 
6161     /* Shut off incoming Tx traffic */
6162     netif_tx_stop_all_queues(netdev);
6163 
6164     /* call carrier off first to avoid false dev_watchdog timeouts */
6165     netif_carrier_off(netdev);
6166     netif_tx_disable(netdev);
6167 
6168     /* Disable Rx */
6169     ixgbe_disable_rx(adapter);
6170 
6171     /* synchronize_rcu() needed for pending XDP buffers to drain */
6172     if (adapter->xdp_ring[0])
6173         synchronize_rcu();
6174 
6175     ixgbe_irq_disable(adapter);
6176 
6177     ixgbe_napi_disable_all(adapter);
6178 
6179     clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
6180     adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
6181     adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
6182 
6183     del_timer_sync(&adapter->service_timer);
6184 
6185     if (adapter->num_vfs) {
6186         /* Clear EITR Select mapping */
6187         IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITRSEL, 0);
6188 
6189         /* Mark all the VFs as inactive */
6190         for (i = 0 ; i < adapter->num_vfs; i++)
6191             adapter->vfinfo[i].clear_to_send = false;
6192 
6193         /* update setting rx tx for all active vfs */
6194         ixgbe_set_all_vfs(adapter);
6195     }
6196 
6197     /* disable transmits in the hardware now that interrupts are off */
6198     ixgbe_disable_tx(adapter);
6199 
6200     if (!pci_channel_offline(adapter->pdev))
6201         ixgbe_reset(adapter);
6202 
6203     /* power down the optics for 82599 SFP+ fiber */
6204     if (hw->mac.ops.disable_tx_laser)
6205         hw->mac.ops.disable_tx_laser(hw);
6206 
6207     ixgbe_clean_all_tx_rings(adapter);
6208     ixgbe_clean_all_rx_rings(adapter);
6209 }
6210 
6211 /**
6212  * ixgbe_set_eee_capable - helper function to determine EEE support on X550
6213  * @adapter: board private structure
6214  */
6215 static void ixgbe_set_eee_capable(struct ixgbe_adapter *adapter)
6216 {
6217     struct ixgbe_hw *hw = &adapter->hw;
6218 
6219     switch (hw->device_id) {
6220     case IXGBE_DEV_ID_X550EM_A_1G_T:
6221     case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6222         if (!hw->phy.eee_speeds_supported)
6223             break;
6224         adapter->flags2 |= IXGBE_FLAG2_EEE_CAPABLE;
6225         if (!hw->phy.eee_speeds_advertised)
6226             break;
6227         adapter->flags2 |= IXGBE_FLAG2_EEE_ENABLED;
6228         break;
6229     default:
6230         adapter->flags2 &= ~IXGBE_FLAG2_EEE_CAPABLE;
6231         adapter->flags2 &= ~IXGBE_FLAG2_EEE_ENABLED;
6232         break;
6233     }
6234 }
6235 
6236 /**
6237  * ixgbe_tx_timeout - Respond to a Tx Hang
6238  * @netdev: network interface device structure
6239  * @txqueue: queue number that timed out
6240  **/
6241 static void ixgbe_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
6242 {
6243     struct ixgbe_adapter *adapter = netdev_priv(netdev);
6244 
6245     /* Do the reset outside of interrupt context */
6246     ixgbe_tx_timeout_reset(adapter);
6247 }
6248 
6249 #ifdef CONFIG_IXGBE_DCB
6250 static void ixgbe_init_dcb(struct ixgbe_adapter *adapter)
6251 {
6252     struct ixgbe_hw *hw = &adapter->hw;
6253     struct tc_configuration *tc;
6254     int j;
6255 
6256     switch (hw->mac.type) {
6257     case ixgbe_mac_82598EB:
6258     case ixgbe_mac_82599EB:
6259         adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
6260         adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
6261         break;
6262     case ixgbe_mac_X540:
6263     case ixgbe_mac_X550:
6264         adapter->dcb_cfg.num_tcs.pg_tcs = X540_TRAFFIC_CLASS;
6265         adapter->dcb_cfg.num_tcs.pfc_tcs = X540_TRAFFIC_CLASS;
6266         break;
6267     case ixgbe_mac_X550EM_x:
6268     case ixgbe_mac_x550em_a:
6269     default:
6270         adapter->dcb_cfg.num_tcs.pg_tcs = DEF_TRAFFIC_CLASS;
6271         adapter->dcb_cfg.num_tcs.pfc_tcs = DEF_TRAFFIC_CLASS;
6272         break;
6273     }
6274 
6275     /* Configure DCB traffic classes */
6276     for (j = 0; j < MAX_TRAFFIC_CLASS; j++) {
6277         tc = &adapter->dcb_cfg.tc_config[j];
6278         tc->path[DCB_TX_CONFIG].bwg_id = 0;
6279         tc->path[DCB_TX_CONFIG].bwg_percent = 12 + (j & 1);
6280         tc->path[DCB_RX_CONFIG].bwg_id = 0;
6281         tc->path[DCB_RX_CONFIG].bwg_percent = 12 + (j & 1);
6282         tc->dcb_pfc = pfc_disabled;
6283     }
6284 
6285     /* Initialize default user to priority mapping, UPx->TC0 */
6286     tc = &adapter->dcb_cfg.tc_config[0];
6287     tc->path[DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
6288     tc->path[DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
6289 
6290     adapter->dcb_cfg.bw_percentage[DCB_TX_CONFIG][0] = 100;
6291     adapter->dcb_cfg.bw_percentage[DCB_RX_CONFIG][0] = 100;
6292     adapter->dcb_cfg.pfc_mode_enable = false;
6293     adapter->dcb_set_bitmap = 0x00;
6294     if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
6295         adapter->dcbx_cap = DCB_CAP_DCBX_HOST | DCB_CAP_DCBX_VER_CEE;
6296     memcpy(&adapter->temp_dcb_cfg, &adapter->dcb_cfg,
6297            sizeof(adapter->temp_dcb_cfg));
6298 }
6299 #endif
6300 
6301 /**
6302  * ixgbe_sw_init - Initialize general software structures (struct ixgbe_adapter)
6303  * @adapter: board private structure to initialize
6304  * @ii: pointer to ixgbe_info for device
6305  *
6306  * ixgbe_sw_init initializes the Adapter private data structure.
6307  * Fields are initialized based on PCI device information and
6308  * OS network device settings (MTU size).
6309  **/
6310 static int ixgbe_sw_init(struct ixgbe_adapter *adapter,
6311              const struct ixgbe_info *ii)
6312 {
6313     struct ixgbe_hw *hw = &adapter->hw;
6314     struct pci_dev *pdev = adapter->pdev;
6315     unsigned int rss, fdir;
6316     u32 fwsm;
6317     int i;
6318 
6319     /* PCI config space info */
6320 
6321     hw->vendor_id = pdev->vendor;
6322     hw->device_id = pdev->device;
6323     hw->revision_id = pdev->revision;
6324     hw->subsystem_vendor_id = pdev->subsystem_vendor;
6325     hw->subsystem_device_id = pdev->subsystem_device;
6326 
6327     /* get_invariants needs the device IDs */
6328     ii->get_invariants(hw);
6329 
6330     /* Set common capability flags and settings */
6331     rss = min_t(int, ixgbe_max_rss_indices(adapter), num_online_cpus());
6332     adapter->ring_feature[RING_F_RSS].limit = rss;
6333     adapter->flags2 |= IXGBE_FLAG2_RSC_CAPABLE;
6334     adapter->max_q_vectors = MAX_Q_VECTORS_82599;
6335     adapter->atr_sample_rate = 20;
6336     fdir = min_t(int, IXGBE_MAX_FDIR_INDICES, num_online_cpus());
6337     adapter->ring_feature[RING_F_FDIR].limit = fdir;
6338     adapter->fdir_pballoc = IXGBE_FDIR_PBALLOC_64K;
6339     adapter->ring_feature[RING_F_VMDQ].limit = 1;
6340 #ifdef CONFIG_IXGBE_DCA
6341     adapter->flags |= IXGBE_FLAG_DCA_CAPABLE;
6342 #endif
6343 #ifdef CONFIG_IXGBE_DCB
6344     adapter->flags |= IXGBE_FLAG_DCB_CAPABLE;
6345     adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
6346 #endif
6347 #ifdef IXGBE_FCOE
6348     adapter->flags |= IXGBE_FLAG_FCOE_CAPABLE;
6349     adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6350 #ifdef CONFIG_IXGBE_DCB
6351     /* Default traffic class to use for FCoE */
6352     adapter->fcoe.up = IXGBE_FCOE_DEFTC;
6353 #endif /* CONFIG_IXGBE_DCB */
6354 #endif /* IXGBE_FCOE */
6355 
6356     /* initialize static ixgbe jump table entries */
6357     adapter->jump_tables[0] = kzalloc(sizeof(*adapter->jump_tables[0]),
6358                       GFP_KERNEL);
6359     if (!adapter->jump_tables[0])
6360         return -ENOMEM;
6361     adapter->jump_tables[0]->mat = ixgbe_ipv4_fields;
6362 
6363     for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++)
6364         adapter->jump_tables[i] = NULL;
6365 
6366     adapter->mac_table = kcalloc(hw->mac.num_rar_entries,
6367                      sizeof(struct ixgbe_mac_addr),
6368                      GFP_KERNEL);
6369     if (!adapter->mac_table)
6370         return -ENOMEM;
6371 
6372     if (ixgbe_init_rss_key(adapter))
6373         return -ENOMEM;
6374 
6375     adapter->af_xdp_zc_qps = bitmap_zalloc(IXGBE_MAX_XDP_QS, GFP_KERNEL);
6376     if (!adapter->af_xdp_zc_qps)
6377         return -ENOMEM;
6378 
6379     /* Set MAC specific capability flags and exceptions */
6380     switch (hw->mac.type) {
6381     case ixgbe_mac_82598EB:
6382         adapter->flags2 &= ~IXGBE_FLAG2_RSC_CAPABLE;
6383 
6384         if (hw->device_id == IXGBE_DEV_ID_82598AT)
6385             adapter->flags |= IXGBE_FLAG_FAN_FAIL_CAPABLE;
6386 
6387         adapter->max_q_vectors = MAX_Q_VECTORS_82598;
6388         adapter->ring_feature[RING_F_FDIR].limit = 0;
6389         adapter->atr_sample_rate = 0;
6390         adapter->fdir_pballoc = 0;
6391 #ifdef IXGBE_FCOE
6392         adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6393         adapter->flags &= ~IXGBE_FLAG_FCOE_ENABLED;
6394 #ifdef CONFIG_IXGBE_DCB
6395         adapter->fcoe.up = 0;
6396 #endif /* IXGBE_DCB */
6397 #endif /* IXGBE_FCOE */
6398         break;
6399     case ixgbe_mac_82599EB:
6400         if (hw->device_id == IXGBE_DEV_ID_82599_T3_LOM)
6401             adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6402         break;
6403     case ixgbe_mac_X540:
6404         fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
6405         if (fwsm & IXGBE_FWSM_TS_ENABLED)
6406             adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6407         break;
6408     case ixgbe_mac_x550em_a:
6409         switch (hw->device_id) {
6410         case IXGBE_DEV_ID_X550EM_A_1G_T:
6411         case IXGBE_DEV_ID_X550EM_A_1G_T_L:
6412             adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6413             break;
6414         default:
6415             break;
6416         }
6417         fallthrough;
6418     case ixgbe_mac_X550EM_x:
6419 #ifdef CONFIG_IXGBE_DCB
6420         adapter->flags &= ~IXGBE_FLAG_DCB_CAPABLE;
6421 #endif
6422 #ifdef IXGBE_FCOE
6423         adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
6424 #ifdef CONFIG_IXGBE_DCB
6425         adapter->fcoe.up = 0;
6426 #endif /* IXGBE_DCB */
6427 #endif /* IXGBE_FCOE */
6428         fallthrough;
6429     case ixgbe_mac_X550:
6430         if (hw->mac.type == ixgbe_mac_X550)
6431             adapter->flags2 |= IXGBE_FLAG2_TEMP_SENSOR_CAPABLE;
6432 #ifdef CONFIG_IXGBE_DCA
6433         adapter->flags &= ~IXGBE_FLAG_DCA_CAPABLE;
6434 #endif
6435         break;
6436     default:
6437         break;
6438     }
6439 
6440 #ifdef IXGBE_FCOE
6441     /* FCoE support exists, always init the FCoE lock */
6442     spin_lock_init(&adapter->fcoe.lock);
6443 
6444 #endif
6445     /* n-tuple support exists, always init our spinlock */
6446     spin_lock_init(&adapter->fdir_perfect_lock);
6447 
6448     /* init spinlock to avoid concurrency of VF resources */
6449     spin_lock_init(&adapter->vfs_lock);
6450 
6451 #ifdef CONFIG_IXGBE_DCB
6452     ixgbe_init_dcb(adapter);
6453 #endif
6454     ixgbe_init_ipsec_offload(adapter);
6455 
6456     /* default flow control settings */
6457     hw->fc.requested_mode = ixgbe_fc_full;
6458     hw->fc.current_mode = ixgbe_fc_full;    /* init for ethtool output */
6459     ixgbe_pbthresh_setup(adapter);
6460     hw->fc.pause_time = IXGBE_DEFAULT_FCPAUSE;
6461     hw->fc.send_xon = true;
6462     hw->fc.disable_fc_autoneg = ixgbe_device_supports_autoneg_fc(hw);
6463 
6464 #ifdef CONFIG_PCI_IOV
6465     if (max_vfs > 0)
6466         e_dev_warn("Enabling SR-IOV VFs using the max_vfs module parameter is deprecated - please use the pci sysfs interface instead.\n");
6467 
6468     /* assign number of SR-IOV VFs */
6469     if (hw->mac.type != ixgbe_mac_82598EB) {
6470         if (max_vfs > IXGBE_MAX_VFS_DRV_LIMIT) {
6471             max_vfs = 0;
6472             e_dev_warn("max_vfs parameter out of range. Not assigning any SR-IOV VFs\n");
6473         }
6474     }
6475 #endif /* CONFIG_PCI_IOV */
6476 
6477     /* enable itr by default in dynamic mode */
6478     adapter->rx_itr_setting = 1;
6479     adapter->tx_itr_setting = 1;
6480 
6481     /* set default ring sizes */
6482     adapter->tx_ring_count = IXGBE_DEFAULT_TXD;
6483     adapter->rx_ring_count = IXGBE_DEFAULT_RXD;
6484 
6485     /* set default work limits */
6486     adapter->tx_work_limit = IXGBE_DEFAULT_TX_WORK;
6487 
6488     /* initialize eeprom parameters */
6489     if (ixgbe_init_eeprom_params_generic(hw)) {
6490         e_dev_err("EEPROM initialization failed\n");
6491         return -EIO;
6492     }
6493 
6494     /* PF holds first pool slot */
6495     set_bit(0, adapter->fwd_bitmask);
6496     set_bit(__IXGBE_DOWN, &adapter->state);
6497 
6498     return 0;
6499 }
6500 
6501 /**
6502  * ixgbe_setup_tx_resources - allocate Tx resources (Descriptors)
6503  * @tx_ring:    tx descriptor ring (for a specific queue) to setup
6504  *
6505  * Return 0 on success, negative on failure
6506  **/
6507 int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
6508 {
6509     struct device *dev = tx_ring->dev;
6510     int orig_node = dev_to_node(dev);
6511     int ring_node = NUMA_NO_NODE;
6512     int size;
6513 
6514     size = sizeof(struct ixgbe_tx_buffer) * tx_ring->count;
6515 
6516     if (tx_ring->q_vector)
6517         ring_node = tx_ring->q_vector->numa_node;
6518 
6519     tx_ring->tx_buffer_info = vmalloc_node(size, ring_node);
6520     if (!tx_ring->tx_buffer_info)
6521         tx_ring->tx_buffer_info = vmalloc(size);
6522     if (!tx_ring->tx_buffer_info)
6523         goto err;
6524 
6525     /* round up to nearest 4K */
6526     tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
6527     tx_ring->size = ALIGN(tx_ring->size, 4096);
6528 
6529     set_dev_node(dev, ring_node);
6530     tx_ring->desc = dma_alloc_coherent(dev,
6531                        tx_ring->size,
6532                        &tx_ring->dma,
6533                        GFP_KERNEL);
6534     set_dev_node(dev, orig_node);
6535     if (!tx_ring->desc)
6536         tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
6537                            &tx_ring->dma, GFP_KERNEL);
6538     if (!tx_ring->desc)
6539         goto err;
6540 
6541     tx_ring->next_to_use = 0;
6542     tx_ring->next_to_clean = 0;
6543     return 0;
6544 
6545 err:
6546     vfree(tx_ring->tx_buffer_info);
6547     tx_ring->tx_buffer_info = NULL;
6548     dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
6549     return -ENOMEM;
6550 }
6551 
6552 /**
6553  * ixgbe_setup_all_tx_resources - allocate all queues Tx resources
6554  * @adapter: board private structure
6555  *
6556  * If this function returns with an error, then it's possible one or
6557  * more of the rings is populated (while the rest are not).  It is the
6558  * callers duty to clean those orphaned rings.
6559  *
6560  * Return 0 on success, negative on failure
6561  **/
6562 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
6563 {
6564     int i, j = 0, err = 0;
6565 
6566     for (i = 0; i < adapter->num_tx_queues; i++) {
6567         err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
6568         if (!err)
6569             continue;
6570 
6571         e_err(probe, "Allocation for Tx Queue %u failed\n", i);
6572         goto err_setup_tx;
6573     }
6574     for (j = 0; j < adapter->num_xdp_queues; j++) {
6575         err = ixgbe_setup_tx_resources(adapter->xdp_ring[j]);
6576         if (!err)
6577             continue;
6578 
6579         e_err(probe, "Allocation for Tx Queue %u failed\n", j);
6580         goto err_setup_tx;
6581     }
6582 
6583     return 0;
6584 err_setup_tx:
6585     /* rewind the index freeing the rings as we go */
6586     while (j--)
6587         ixgbe_free_tx_resources(adapter->xdp_ring[j]);
6588     while (i--)
6589         ixgbe_free_tx_resources(adapter->tx_ring[i]);
6590     return err;
6591 }
6592 
6593 static int ixgbe_rx_napi_id(struct ixgbe_ring *rx_ring)
6594 {
6595     struct ixgbe_q_vector *q_vector = rx_ring->q_vector;
6596 
6597     return q_vector ? q_vector->napi.napi_id : 0;
6598 }
6599 
6600 /**
6601  * ixgbe_setup_rx_resources - allocate Rx resources (Descriptors)
6602  * @adapter: pointer to ixgbe_adapter
6603  * @rx_ring:    rx descriptor ring (for a specific queue) to setup
6604  *
6605  * Returns 0 on success, negative on failure
6606  **/
6607 int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
6608                  struct ixgbe_ring *rx_ring)
6609 {
6610     struct device *dev = rx_ring->dev;
6611     int orig_node = dev_to_node(dev);
6612     int ring_node = NUMA_NO_NODE;
6613     int size;
6614 
6615     size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
6616 
6617     if (rx_ring->q_vector)
6618         ring_node = rx_ring->q_vector->numa_node;
6619 
6620     rx_ring->rx_buffer_info = vmalloc_node(size, ring_node);
6621     if (!rx_ring->rx_buffer_info)
6622         rx_ring->rx_buffer_info = vmalloc(size);
6623     if (!rx_ring->rx_buffer_info)
6624         goto err;
6625 
6626     /* Round up to nearest 4K */
6627     rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
6628     rx_ring->size = ALIGN(rx_ring->size, 4096);
6629 
6630     set_dev_node(dev, ring_node);
6631     rx_ring->desc = dma_alloc_coherent(dev,
6632                        rx_ring->size,
6633                        &rx_ring->dma,
6634                        GFP_KERNEL);
6635     set_dev_node(dev, orig_node);
6636     if (!rx_ring->desc)
6637         rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
6638                            &rx_ring->dma, GFP_KERNEL);
6639     if (!rx_ring->desc)
6640         goto err;
6641 
6642     rx_ring->next_to_clean = 0;
6643     rx_ring->next_to_use = 0;
6644 
6645     /* XDP RX-queue info */
6646     if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, adapter->netdev,
6647                  rx_ring->queue_index, ixgbe_rx_napi_id(rx_ring)) < 0)
6648         goto err;
6649 
6650     rx_ring->xdp_prog = adapter->xdp_prog;
6651 
6652     return 0;
6653 err:
6654     vfree(rx_ring->rx_buffer_info);
6655     rx_ring->rx_buffer_info = NULL;
6656     dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
6657     return -ENOMEM;
6658 }
6659 
6660 /**
6661  * ixgbe_setup_all_rx_resources - allocate all queues Rx resources
6662  * @adapter: board private structure
6663  *
6664  * If this function returns with an error, then it's possible one or
6665  * more of the rings is populated (while the rest are not).  It is the
6666  * callers duty to clean those orphaned rings.
6667  *
6668  * Return 0 on success, negative on failure
6669  **/
6670 static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
6671 {
6672     int i, err = 0;
6673 
6674     for (i = 0; i < adapter->num_rx_queues; i++) {
6675         err = ixgbe_setup_rx_resources(adapter, adapter->rx_ring[i]);
6676         if (!err)
6677             continue;
6678 
6679         e_err(probe, "Allocation for Rx Queue %u failed\n", i);
6680         goto err_setup_rx;
6681     }
6682 
6683 #ifdef IXGBE_FCOE
6684     err = ixgbe_setup_fcoe_ddp_resources(adapter);
6685     if (!err)
6686 #endif
6687         return 0;
6688 err_setup_rx:
6689     /* rewind the index freeing the rings as we go */
6690     while (i--)
6691         ixgbe_free_rx_resources(adapter->rx_ring[i]);
6692     return err;
6693 }
6694 
6695 /**
6696  * ixgbe_free_tx_resources - Free Tx Resources per Queue
6697  * @tx_ring: Tx descriptor ring for a specific queue
6698  *
6699  * Free all transmit software resources
6700  **/
6701 void ixgbe_free_tx_resources(struct ixgbe_ring *tx_ring)
6702 {
6703     ixgbe_clean_tx_ring(tx_ring);
6704 
6705     vfree(tx_ring->tx_buffer_info);
6706     tx_ring->tx_buffer_info = NULL;
6707 
6708     /* if not set, then don't free */
6709     if (!tx_ring->desc)
6710         return;
6711 
6712     dma_free_coherent(tx_ring->dev, tx_ring->size,
6713               tx_ring->desc, tx_ring->dma);
6714 
6715     tx_ring->desc = NULL;
6716 }
6717 
6718 /**
6719  * ixgbe_free_all_tx_resources - Free Tx Resources for All Queues
6720  * @adapter: board private structure
6721  *
6722  * Free all transmit software resources
6723  **/
6724 static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
6725 {
6726     int i;
6727 
6728     for (i = 0; i < adapter->num_tx_queues; i++)
6729         if (adapter->tx_ring[i]->desc)
6730             ixgbe_free_tx_resources(adapter->tx_ring[i]);
6731     for (i = 0; i < adapter->num_xdp_queues; i++)
6732         if (adapter->xdp_ring[i]->desc)
6733             ixgbe_free_tx_resources(adapter->xdp_ring[i]);
6734 }
6735 
6736 /**
6737  * ixgbe_free_rx_resources - Free Rx Resources
6738  * @rx_ring: ring to clean the resources from
6739  *
6740  * Free all receive software resources
6741  **/
6742 void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
6743 {
6744     ixgbe_clean_rx_ring(rx_ring);
6745 
6746     rx_ring->xdp_prog = NULL;
6747     xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
6748     vfree(rx_ring->rx_buffer_info);
6749     rx_ring->rx_buffer_info = NULL;
6750 
6751     /* if not set, then don't free */
6752     if (!rx_ring->desc)
6753         return;
6754 
6755     dma_free_coherent(rx_ring->dev, rx_ring->size,
6756               rx_ring->desc, rx_ring->dma);
6757 
6758     rx_ring->desc = NULL;
6759 }
6760 
6761 /**
6762  * ixgbe_free_all_rx_resources - Free Rx Resources for All Queues
6763  * @adapter: board private structure
6764  *
6765  * Free all receive software resources
6766  **/
6767 static void ixgbe_free_all_rx_resources(struct ixgbe_adapter *adapter)
6768 {
6769     int i;
6770 
6771 #ifdef IXGBE_FCOE
6772     ixgbe_free_fcoe_ddp_resources(adapter);
6773 
6774 #endif
6775     for (i = 0; i < adapter->num_rx_queues; i++)
6776         if (adapter->rx_ring[i]->desc)
6777             ixgbe_free_rx_resources(adapter->rx_ring[i]);
6778 }
6779 
6780 /**
6781  * ixgbe_change_mtu - Change the Maximum Transfer Unit
6782  * @netdev: network interface device structure
6783  * @new_mtu: new value for maximum frame size
6784  *
6785  * Returns 0 on success, negative on failure
6786  **/
6787 static int ixgbe_change_mtu(struct net_device *netdev, int new_mtu)
6788 {
6789     struct ixgbe_adapter *adapter = netdev_priv(netdev);
6790 
6791     if (adapter->xdp_prog) {
6792         int new_frame_size = new_mtu + ETH_HLEN + ETH_FCS_LEN +
6793                      VLAN_HLEN;
6794         int i;
6795 
6796         for (i = 0; i < adapter->num_rx_queues; i++) {
6797             struct ixgbe_ring *ring = adapter->rx_ring[i];
6798 
6799             if (new_frame_size > ixgbe_rx_bufsz(ring)) {
6800                 e_warn(probe, "Requested MTU size is not supported with XDP\n");
6801                 return -EINVAL;
6802             }
6803         }
6804     }
6805 
6806     /*
6807      * For 82599EB we cannot allow legacy VFs to enable their receive
6808      * paths when MTU greater than 1500 is configured.  So display a
6809      * warning that legacy VFs will be disabled.
6810      */
6811     if ((adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) &&
6812         (adapter->hw.mac.type == ixgbe_mac_82599EB) &&
6813         (new_mtu > ETH_DATA_LEN))
6814         e_warn(probe, "Setting MTU > 1500 will disable legacy VFs\n");
6815 
6816     netdev_dbg(netdev, "changing MTU from %d to %d\n",
6817            netdev->mtu, new_mtu);
6818 
6819     /* must set new MTU before calling down or up */
6820     netdev->mtu = new_mtu;
6821 
6822     if (netif_running(netdev))
6823         ixgbe_reinit_locked(adapter);
6824 
6825     return 0;
6826 }
6827 
6828 /**
6829  * ixgbe_open - Called when a network interface is made active
6830  * @netdev: network interface device structure
6831  *
6832  * Returns 0 on success, negative value on failure
6833  *
6834  * The open entry point is called when a network interface is made
6835  * active by the system (IFF_UP).  At this point all resources needed
6836  * for transmit and receive operations are allocated, the interrupt
6837  * handler is registered with the OS, the watchdog timer is started,
6838  * and the stack is notified that the interface is ready.
6839  **/
6840 int ixgbe_open(struct net_device *netdev)
6841 {
6842     struct ixgbe_adapter *adapter = netdev_priv(netdev);
6843     struct ixgbe_hw *hw = &adapter->hw;
6844     int err, queues;
6845 
6846     /* disallow open during test */
6847     if (test_bit(__IXGBE_TESTING, &adapter->state))
6848         return -EBUSY;
6849 
6850     netif_carrier_off(netdev);
6851 
6852     /* allocate transmit descriptors */
6853     err = ixgbe_setup_all_tx_resources(adapter);
6854     if (err)
6855         goto err_setup_tx;
6856 
6857     /* allocate receive descriptors */
6858     err = ixgbe_setup_all_rx_resources(adapter);
6859     if (err)
6860         goto err_setup_rx;
6861 
6862     ixgbe_configure(adapter);
6863 
6864     err = ixgbe_request_irq(adapter);
6865     if (err)
6866         goto err_req_irq;
6867 
6868     /* Notify the stack of the actual queue counts. */
6869     queues = adapter->num_tx_queues;
6870     err = netif_set_real_num_tx_queues(netdev, queues);
6871     if (err)
6872         goto err_set_queues;
6873 
6874     queues = adapter->num_rx_queues;
6875     err = netif_set_real_num_rx_queues(netdev, queues);
6876     if (err)
6877         goto err_set_queues;
6878 
6879     ixgbe_ptp_init(adapter);
6880 
6881     ixgbe_up_complete(adapter);
6882 
6883     udp_tunnel_nic_reset_ntf(netdev);
6884 
6885     return 0;
6886 
6887 err_set_queues:
6888     ixgbe_free_irq(adapter);
6889 err_req_irq:
6890     ixgbe_free_all_rx_resources(adapter);
6891     if (hw->phy.ops.set_phy_power && !adapter->wol)
6892         hw->phy.ops.set_phy_power(&adapter->hw, false);
6893 err_setup_rx:
6894     ixgbe_free_all_tx_resources(adapter);
6895 err_setup_tx:
6896     ixgbe_reset(adapter);
6897 
6898     return err;
6899 }
6900 
6901 static void ixgbe_close_suspend(struct ixgbe_adapter *adapter)
6902 {
6903     ixgbe_ptp_suspend(adapter);
6904 
6905     if (adapter->hw.phy.ops.enter_lplu) {
6906         adapter->hw.phy.reset_disable = true;
6907         ixgbe_down(adapter);
6908         adapter->hw.phy.ops.enter_lplu(&adapter->hw);
6909         adapter->hw.phy.reset_disable = false;
6910     } else {
6911         ixgbe_down(adapter);
6912     }
6913 
6914     ixgbe_free_irq(adapter);
6915 
6916     ixgbe_free_all_tx_resources(adapter);
6917     ixgbe_free_all_rx_resources(adapter);
6918 }
6919 
6920 /**
6921  * ixgbe_close - Disables a network interface
6922  * @netdev: network interface device structure
6923  *
6924  * Returns 0, this is not allowed to fail
6925  *
6926  * The close entry point is called when an interface is de-activated
6927  * by the OS.  The hardware is still under the drivers control, but
6928  * needs to be disabled.  A global MAC reset is issued to stop the
6929  * hardware, and all transmit and receive resources are freed.
6930  **/
6931 int ixgbe_close(struct net_device *netdev)
6932 {
6933     struct ixgbe_adapter *adapter = netdev_priv(netdev);
6934 
6935     ixgbe_ptp_stop(adapter);
6936 
6937     if (netif_device_present(netdev))
6938         ixgbe_close_suspend(adapter);
6939 
6940     ixgbe_fdir_filter_exit(adapter);
6941 
6942     ixgbe_release_hw_control(adapter);
6943 
6944     return 0;
6945 }
6946 
6947 static int __maybe_unused ixgbe_resume(struct device *dev_d)
6948 {
6949     struct pci_dev *pdev = to_pci_dev(dev_d);
6950     struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6951     struct net_device *netdev = adapter->netdev;
6952     u32 err;
6953 
6954     adapter->hw.hw_addr = adapter->io_addr;
6955 
6956     err = pci_enable_device_mem(pdev);
6957     if (err) {
6958         e_dev_err("Cannot enable PCI device from suspend\n");
6959         return err;
6960     }
6961     smp_mb__before_atomic();
6962     clear_bit(__IXGBE_DISABLED, &adapter->state);
6963     pci_set_master(pdev);
6964 
6965     device_wakeup_disable(dev_d);
6966 
6967     ixgbe_reset(adapter);
6968 
6969     IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
6970 
6971     rtnl_lock();
6972     err = ixgbe_init_interrupt_scheme(adapter);
6973     if (!err && netif_running(netdev))
6974         err = ixgbe_open(netdev);
6975 
6976 
6977     if (!err)
6978         netif_device_attach(netdev);
6979     rtnl_unlock();
6980 
6981     return err;
6982 }
6983 
6984 static int __ixgbe_shutdown(struct pci_dev *pdev, bool *enable_wake)
6985 {
6986     struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
6987     struct net_device *netdev = adapter->netdev;
6988     struct ixgbe_hw *hw = &adapter->hw;
6989     u32 ctrl;
6990     u32 wufc = adapter->wol;
6991 
6992     rtnl_lock();
6993     netif_device_detach(netdev);
6994 
6995     if (netif_running(netdev))
6996         ixgbe_close_suspend(adapter);
6997 
6998     ixgbe_clear_interrupt_scheme(adapter);
6999     rtnl_unlock();
7000 
7001     if (hw->mac.ops.stop_link_on_d3)
7002         hw->mac.ops.stop_link_on_d3(hw);
7003 
7004     if (wufc) {
7005         u32 fctrl;
7006 
7007         ixgbe_set_rx_mode(netdev);
7008 
7009         /* enable the optics for 82599 SFP+ fiber as we can WoL */
7010         if (hw->mac.ops.enable_tx_laser)
7011             hw->mac.ops.enable_tx_laser(hw);
7012 
7013         /* enable the reception of multicast packets */
7014         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7015         fctrl |= IXGBE_FCTRL_MPE;
7016         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
7017 
7018         ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL);
7019         ctrl |= IXGBE_CTRL_GIO_DIS;
7020         IXGBE_WRITE_REG(hw, IXGBE_CTRL, ctrl);
7021 
7022         IXGBE_WRITE_REG(hw, IXGBE_WUFC, wufc);
7023     } else {
7024         IXGBE_WRITE_REG(hw, IXGBE_WUC, 0);
7025         IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
7026     }
7027 
7028     switch (hw->mac.type) {
7029     case ixgbe_mac_82598EB:
7030         pci_wake_from_d3(pdev, false);
7031         break;
7032     case ixgbe_mac_82599EB:
7033     case ixgbe_mac_X540:
7034     case ixgbe_mac_X550:
7035     case ixgbe_mac_X550EM_x:
7036     case ixgbe_mac_x550em_a:
7037         pci_wake_from_d3(pdev, !!wufc);
7038         break;
7039     default:
7040         break;
7041     }
7042 
7043     *enable_wake = !!wufc;
7044     if (hw->phy.ops.set_phy_power && !*enable_wake)
7045         hw->phy.ops.set_phy_power(hw, false);
7046 
7047     ixgbe_release_hw_control(adapter);
7048 
7049     if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
7050         pci_disable_device(pdev);
7051 
7052     return 0;
7053 }
7054 
7055 static int __maybe_unused ixgbe_suspend(struct device *dev_d)
7056 {
7057     struct pci_dev *pdev = to_pci_dev(dev_d);
7058     int retval;
7059     bool wake;
7060 
7061     retval = __ixgbe_shutdown(pdev, &wake);
7062 
7063     device_set_wakeup_enable(dev_d, wake);
7064 
7065     return retval;
7066 }
7067 
7068 static void ixgbe_shutdown(struct pci_dev *pdev)
7069 {
7070     bool wake;
7071 
7072     __ixgbe_shutdown(pdev, &wake);
7073 
7074     if (system_state == SYSTEM_POWER_OFF) {
7075         pci_wake_from_d3(pdev, wake);
7076         pci_set_power_state(pdev, PCI_D3hot);
7077     }
7078 }
7079 
7080 /**
7081  * ixgbe_update_stats - Update the board statistics counters.
7082  * @adapter: board private structure
7083  **/
7084 void ixgbe_update_stats(struct ixgbe_adapter *adapter)
7085 {
7086     struct net_device *netdev = adapter->netdev;
7087     struct ixgbe_hw *hw = &adapter->hw;
7088     struct ixgbe_hw_stats *hwstats = &adapter->stats;
7089     u64 total_mpc = 0;
7090     u32 i, missed_rx = 0, mpc, bprc, lxon, lxoff, xon_off_tot;
7091     u64 non_eop_descs = 0, restart_queue = 0, tx_busy = 0;
7092     u64 alloc_rx_page_failed = 0, alloc_rx_buff_failed = 0;
7093     u64 alloc_rx_page = 0;
7094     u64 bytes = 0, packets = 0, hw_csum_rx_error = 0;
7095 
7096     if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7097         test_bit(__IXGBE_RESETTING, &adapter->state))
7098         return;
7099 
7100     if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED) {
7101         u64 rsc_count = 0;
7102         u64 rsc_flush = 0;
7103         for (i = 0; i < adapter->num_rx_queues; i++) {
7104             rsc_count += adapter->rx_ring[i]->rx_stats.rsc_count;
7105             rsc_flush += adapter->rx_ring[i]->rx_stats.rsc_flush;
7106         }
7107         adapter->rsc_total_count = rsc_count;
7108         adapter->rsc_total_flush = rsc_flush;
7109     }
7110 
7111     for (i = 0; i < adapter->num_rx_queues; i++) {
7112         struct ixgbe_ring *rx_ring = READ_ONCE(adapter->rx_ring[i]);
7113 
7114         if (!rx_ring)
7115             continue;
7116         non_eop_descs += rx_ring->rx_stats.non_eop_descs;
7117         alloc_rx_page += rx_ring->rx_stats.alloc_rx_page;
7118         alloc_rx_page_failed += rx_ring->rx_stats.alloc_rx_page_failed;
7119         alloc_rx_buff_failed += rx_ring->rx_stats.alloc_rx_buff_failed;
7120         hw_csum_rx_error += rx_ring->rx_stats.csum_err;
7121         bytes += rx_ring->stats.bytes;
7122         packets += rx_ring->stats.packets;
7123     }
7124     adapter->non_eop_descs = non_eop_descs;
7125     adapter->alloc_rx_page = alloc_rx_page;
7126     adapter->alloc_rx_page_failed = alloc_rx_page_failed;
7127     adapter->alloc_rx_buff_failed = alloc_rx_buff_failed;
7128     adapter->hw_csum_rx_error = hw_csum_rx_error;
7129     netdev->stats.rx_bytes = bytes;
7130     netdev->stats.rx_packets = packets;
7131 
7132     bytes = 0;
7133     packets = 0;
7134     /* gather some stats to the adapter struct that are per queue */
7135     for (i = 0; i < adapter->num_tx_queues; i++) {
7136         struct ixgbe_ring *tx_ring = READ_ONCE(adapter->tx_ring[i]);
7137 
7138         if (!tx_ring)
7139             continue;
7140         restart_queue += tx_ring->tx_stats.restart_queue;
7141         tx_busy += tx_ring->tx_stats.tx_busy;
7142         bytes += tx_ring->stats.bytes;
7143         packets += tx_ring->stats.packets;
7144     }
7145     for (i = 0; i < adapter->num_xdp_queues; i++) {
7146         struct ixgbe_ring *xdp_ring = READ_ONCE(adapter->xdp_ring[i]);
7147 
7148         if (!xdp_ring)
7149             continue;
7150         restart_queue += xdp_ring->tx_stats.restart_queue;
7151         tx_busy += xdp_ring->tx_stats.tx_busy;
7152         bytes += xdp_ring->stats.bytes;
7153         packets += xdp_ring->stats.packets;
7154     }
7155     adapter->restart_queue = restart_queue;
7156     adapter->tx_busy = tx_busy;
7157     netdev->stats.tx_bytes = bytes;
7158     netdev->stats.tx_packets = packets;
7159 
7160     hwstats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
7161 
7162     /* 8 register reads */
7163     for (i = 0; i < 8; i++) {
7164         /* for packet buffers not used, the register should read 0 */
7165         mpc = IXGBE_READ_REG(hw, IXGBE_MPC(i));
7166         missed_rx += mpc;
7167         hwstats->mpc[i] += mpc;
7168         total_mpc += hwstats->mpc[i];
7169         hwstats->pxontxc[i] += IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
7170         hwstats->pxofftxc[i] += IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
7171         switch (hw->mac.type) {
7172         case ixgbe_mac_82598EB:
7173             hwstats->rnbc[i] += IXGBE_READ_REG(hw, IXGBE_RNBC(i));
7174             hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC(i));
7175             hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC(i));
7176             hwstats->pxonrxc[i] +=
7177                 IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
7178             break;
7179         case ixgbe_mac_82599EB:
7180         case ixgbe_mac_X540:
7181         case ixgbe_mac_X550:
7182         case ixgbe_mac_X550EM_x:
7183         case ixgbe_mac_x550em_a:
7184             hwstats->pxonrxc[i] +=
7185                 IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
7186             break;
7187         default:
7188             break;
7189         }
7190     }
7191 
7192     /*16 register reads */
7193     for (i = 0; i < 16; i++) {
7194         hwstats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
7195         hwstats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
7196         if ((hw->mac.type == ixgbe_mac_82599EB) ||
7197             (hw->mac.type == ixgbe_mac_X540) ||
7198             (hw->mac.type == ixgbe_mac_X550) ||
7199             (hw->mac.type == ixgbe_mac_X550EM_x) ||
7200             (hw->mac.type == ixgbe_mac_x550em_a)) {
7201             hwstats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
7202             IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)); /* to clear */
7203             hwstats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
7204             IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)); /* to clear */
7205         }
7206     }
7207 
7208     hwstats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
7209     /* work around hardware counting issue */
7210     hwstats->gprc -= missed_rx;
7211 
7212     ixgbe_update_xoff_received(adapter);
7213 
7214     /* 82598 hardware only has a 32 bit counter in the high register */
7215     switch (hw->mac.type) {
7216     case ixgbe_mac_82598EB:
7217         hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
7218         hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
7219         hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
7220         hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
7221         break;
7222     case ixgbe_mac_X540:
7223     case ixgbe_mac_X550:
7224     case ixgbe_mac_X550EM_x:
7225     case ixgbe_mac_x550em_a:
7226         /* OS2BMC stats are X540 and later */
7227         hwstats->o2bgptc += IXGBE_READ_REG(hw, IXGBE_O2BGPTC);
7228         hwstats->o2bspc += IXGBE_READ_REG(hw, IXGBE_O2BSPC);
7229         hwstats->b2ospc += IXGBE_READ_REG(hw, IXGBE_B2OSPC);
7230         hwstats->b2ogprc += IXGBE_READ_REG(hw, IXGBE_B2OGPRC);
7231         fallthrough;
7232     case ixgbe_mac_82599EB:
7233         for (i = 0; i < 16; i++)
7234             adapter->hw_rx_no_dma_resources +=
7235                          IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
7236         hwstats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
7237         IXGBE_READ_REG(hw, IXGBE_GORCH); /* to clear */
7238         hwstats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
7239         IXGBE_READ_REG(hw, IXGBE_GOTCH); /* to clear */
7240         hwstats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
7241         IXGBE_READ_REG(hw, IXGBE_TORH); /* to clear */
7242         hwstats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
7243         hwstats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
7244         hwstats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
7245 #ifdef IXGBE_FCOE
7246         hwstats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
7247         hwstats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
7248         hwstats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
7249         hwstats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
7250         hwstats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
7251         hwstats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
7252         /* Add up per cpu counters for total ddp aloc fail */
7253         if (adapter->fcoe.ddp_pool) {
7254             struct ixgbe_fcoe *fcoe = &adapter->fcoe;
7255             struct ixgbe_fcoe_ddp_pool *ddp_pool;
7256             unsigned int cpu;
7257             u64 noddp = 0, noddp_ext_buff = 0;
7258             for_each_possible_cpu(cpu) {
7259                 ddp_pool = per_cpu_ptr(fcoe->ddp_pool, cpu);
7260                 noddp += ddp_pool->noddp;
7261                 noddp_ext_buff += ddp_pool->noddp_ext_buff;
7262             }
7263             hwstats->fcoe_noddp = noddp;
7264             hwstats->fcoe_noddp_ext_buff = noddp_ext_buff;
7265         }
7266 #endif /* IXGBE_FCOE */
7267         break;
7268     default:
7269         break;
7270     }
7271     bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
7272     hwstats->bprc += bprc;
7273     hwstats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
7274     if (hw->mac.type == ixgbe_mac_82598EB)
7275         hwstats->mprc -= bprc;
7276     hwstats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
7277     hwstats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
7278     hwstats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
7279     hwstats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
7280     hwstats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
7281     hwstats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
7282     hwstats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
7283     hwstats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
7284     lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
7285     hwstats->lxontxc += lxon;
7286     lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
7287     hwstats->lxofftxc += lxoff;
7288     hwstats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
7289     hwstats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
7290     /*
7291      * 82598 errata - tx of flow control packets is included in tx counters
7292      */
7293     xon_off_tot = lxon + lxoff;
7294     hwstats->gptc -= xon_off_tot;
7295     hwstats->mptc -= xon_off_tot;
7296     hwstats->gotc -= (xon_off_tot * (ETH_ZLEN + ETH_FCS_LEN));
7297     hwstats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
7298     hwstats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
7299     hwstats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
7300     hwstats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
7301     hwstats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
7302     hwstats->ptc64 -= xon_off_tot;
7303     hwstats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
7304     hwstats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
7305     hwstats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
7306     hwstats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
7307     hwstats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
7308     hwstats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
7309 
7310     /* Fill out the OS statistics structure */
7311     netdev->stats.multicast = hwstats->mprc;
7312 
7313     /* Rx Errors */
7314     netdev->stats.rx_errors = hwstats->crcerrs + hwstats->rlec;
7315     netdev->stats.rx_dropped = 0;
7316     netdev->stats.rx_length_errors = hwstats->rlec;
7317     netdev->stats.rx_crc_errors = hwstats->crcerrs;
7318     netdev->stats.rx_missed_errors = total_mpc;
7319 
7320     /* VF Stats Collection - skip while resetting because these
7321      * are not clear on read and otherwise you'll sometimes get
7322      * crazy values.
7323      */
7324     if (!test_bit(__IXGBE_RESETTING, &adapter->state)) {
7325         for (i = 0; i < adapter->num_vfs; i++) {
7326             UPDATE_VF_COUNTER_32bit(IXGBE_PVFGPRC(i),
7327                         adapter->vfinfo[i].last_vfstats.gprc,
7328                         adapter->vfinfo[i].vfstats.gprc);
7329             UPDATE_VF_COUNTER_32bit(IXGBE_PVFGPTC(i),
7330                         adapter->vfinfo[i].last_vfstats.gptc,
7331                         adapter->vfinfo[i].vfstats.gptc);
7332             UPDATE_VF_COUNTER_36bit(IXGBE_PVFGORC_LSB(i),
7333                         IXGBE_PVFGORC_MSB(i),
7334                         adapter->vfinfo[i].last_vfstats.gorc,
7335                         adapter->vfinfo[i].vfstats.gorc);
7336             UPDATE_VF_COUNTER_36bit(IXGBE_PVFGOTC_LSB(i),
7337                         IXGBE_PVFGOTC_MSB(i),
7338                         adapter->vfinfo[i].last_vfstats.gotc,
7339                         adapter->vfinfo[i].vfstats.gotc);
7340             UPDATE_VF_COUNTER_32bit(IXGBE_PVFMPRC(i),
7341                         adapter->vfinfo[i].last_vfstats.mprc,
7342                         adapter->vfinfo[i].vfstats.mprc);
7343         }
7344     }
7345 }
7346 
7347 /**
7348  * ixgbe_fdir_reinit_subtask - worker thread to reinit FDIR filter table
7349  * @adapter: pointer to the device adapter structure
7350  **/
7351 static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
7352 {
7353     struct ixgbe_hw *hw = &adapter->hw;
7354     int i;
7355 
7356     if (!(adapter->flags2 & IXGBE_FLAG2_FDIR_REQUIRES_REINIT))
7357         return;
7358 
7359     adapter->flags2 &= ~IXGBE_FLAG2_FDIR_REQUIRES_REINIT;
7360 
7361     /* if interface is down do nothing */
7362     if (test_bit(__IXGBE_DOWN, &adapter->state))
7363         return;
7364 
7365     /* do nothing if we are not using signature filters */
7366     if (!(adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE))
7367         return;
7368 
7369     adapter->fdir_overflow++;
7370 
7371     if (ixgbe_reinit_fdir_tables_82599(hw) == 0) {
7372         for (i = 0; i < adapter->num_tx_queues; i++)
7373             set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7374                 &(adapter->tx_ring[i]->state));
7375         for (i = 0; i < adapter->num_xdp_queues; i++)
7376             set_bit(__IXGBE_TX_FDIR_INIT_DONE,
7377                 &adapter->xdp_ring[i]->state);
7378         /* re-enable flow director interrupts */
7379         IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
7380     } else {
7381         e_err(probe, "failed to finish FDIR re-initialization, "
7382               "ignored adding FDIR ATR filters\n");
7383     }
7384 }
7385 
7386 /**
7387  * ixgbe_check_hang_subtask - check for hung queues and dropped interrupts
7388  * @adapter: pointer to the device adapter structure
7389  *
7390  * This function serves two purposes.  First it strobes the interrupt lines
7391  * in order to make certain interrupts are occurring.  Secondly it sets the
7392  * bits needed to check for TX hangs.  As a result we should immediately
7393  * determine if a hang has occurred.
7394  */
7395 static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
7396 {
7397     struct ixgbe_hw *hw = &adapter->hw;
7398     u64 eics = 0;
7399     int i;
7400 
7401     /* If we're down, removing or resetting, just bail */
7402     if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7403         test_bit(__IXGBE_REMOVING, &adapter->state) ||
7404         test_bit(__IXGBE_RESETTING, &adapter->state))
7405         return;
7406 
7407     /* Force detection of hung controller */
7408     if (netif_carrier_ok(adapter->netdev)) {
7409         for (i = 0; i < adapter->num_tx_queues; i++)
7410             set_check_for_tx_hang(adapter->tx_ring[i]);
7411         for (i = 0; i < adapter->num_xdp_queues; i++)
7412             set_check_for_tx_hang(adapter->xdp_ring[i]);
7413     }
7414 
7415     if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
7416         /*
7417          * for legacy and MSI interrupts don't set any bits
7418          * that are enabled for EIAM, because this operation
7419          * would set *both* EIMS and EICS for any bit in EIAM
7420          */
7421         IXGBE_WRITE_REG(hw, IXGBE_EICS,
7422             (IXGBE_EICS_TCP_TIMER | IXGBE_EICS_OTHER));
7423     } else {
7424         /* get one bit for every active tx/rx interrupt vector */
7425         for (i = 0; i < adapter->num_q_vectors; i++) {
7426             struct ixgbe_q_vector *qv = adapter->q_vector[i];
7427             if (qv->rx.ring || qv->tx.ring)
7428                 eics |= BIT_ULL(i);
7429         }
7430     }
7431 
7432     /* Cause software interrupt to ensure rings are cleaned */
7433     ixgbe_irq_rearm_queues(adapter, eics);
7434 }
7435 
7436 /**
7437  * ixgbe_watchdog_update_link - update the link status
7438  * @adapter: pointer to the device adapter structure
7439  **/
7440 static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
7441 {
7442     struct ixgbe_hw *hw = &adapter->hw;
7443     u32 link_speed = adapter->link_speed;
7444     bool link_up = adapter->link_up;
7445     bool pfc_en = adapter->dcb_cfg.pfc_mode_enable;
7446 
7447     if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE))
7448         return;
7449 
7450     if (hw->mac.ops.check_link) {
7451         hw->mac.ops.check_link(hw, &link_speed, &link_up, false);
7452     } else {
7453         /* always assume link is up, if no check link function */
7454         link_speed = IXGBE_LINK_SPEED_10GB_FULL;
7455         link_up = true;
7456     }
7457 
7458     if (adapter->ixgbe_ieee_pfc)
7459         pfc_en |= !!(adapter->ixgbe_ieee_pfc->pfc_en);
7460 
7461     if (link_up && !((adapter->flags & IXGBE_FLAG_DCB_ENABLED) && pfc_en)) {
7462         hw->mac.ops.fc_enable(hw);
7463         ixgbe_set_rx_drop_en(adapter);
7464     }
7465 
7466     if (link_up ||
7467         time_after(jiffies, (adapter->link_check_timeout +
7468                  IXGBE_TRY_LINK_TIMEOUT))) {
7469         adapter->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
7470         IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMC_LSC);
7471         IXGBE_WRITE_FLUSH(hw);
7472     }
7473 
7474     adapter->link_up = link_up;
7475     adapter->link_speed = link_speed;
7476 }
7477 
7478 static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
7479 {
7480 #ifdef CONFIG_IXGBE_DCB
7481     struct net_device *netdev = adapter->netdev;
7482     struct dcb_app app = {
7483                   .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
7484                   .protocol = 0,
7485                  };
7486     u8 up = 0;
7487 
7488     if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
7489         up = dcb_ieee_getapp_mask(netdev, &app);
7490 
7491     adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
7492 #endif
7493 }
7494 
7495 /**
7496  * ixgbe_watchdog_link_is_up - update netif_carrier status and
7497  *                             print link up message
7498  * @adapter: pointer to the device adapter structure
7499  **/
7500 static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
7501 {
7502     struct net_device *netdev = adapter->netdev;
7503     struct ixgbe_hw *hw = &adapter->hw;
7504     u32 link_speed = adapter->link_speed;
7505     const char *speed_str;
7506     bool flow_rx, flow_tx;
7507 
7508     /* only continue if link was previously down */
7509     if (netif_carrier_ok(netdev))
7510         return;
7511 
7512     adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
7513 
7514     switch (hw->mac.type) {
7515     case ixgbe_mac_82598EB: {
7516         u32 frctl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
7517         u32 rmcs = IXGBE_READ_REG(hw, IXGBE_RMCS);
7518         flow_rx = !!(frctl & IXGBE_FCTRL_RFCE);
7519         flow_tx = !!(rmcs & IXGBE_RMCS_TFCE_802_3X);
7520     }
7521         break;
7522     case ixgbe_mac_X540:
7523     case ixgbe_mac_X550:
7524     case ixgbe_mac_X550EM_x:
7525     case ixgbe_mac_x550em_a:
7526     case ixgbe_mac_82599EB: {
7527         u32 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
7528         u32 fccfg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
7529         flow_rx = !!(mflcn & IXGBE_MFLCN_RFCE);
7530         flow_tx = !!(fccfg & IXGBE_FCCFG_TFCE_802_3X);
7531     }
7532         break;
7533     default:
7534         flow_tx = false;
7535         flow_rx = false;
7536         break;
7537     }
7538 
7539     adapter->last_rx_ptp_check = jiffies;
7540 
7541     if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7542         ixgbe_ptp_start_cyclecounter(adapter);
7543 
7544     switch (link_speed) {
7545     case IXGBE_LINK_SPEED_10GB_FULL:
7546         speed_str = "10 Gbps";
7547         break;
7548     case IXGBE_LINK_SPEED_5GB_FULL:
7549         speed_str = "5 Gbps";
7550         break;
7551     case IXGBE_LINK_SPEED_2_5GB_FULL:
7552         speed_str = "2.5 Gbps";
7553         break;
7554     case IXGBE_LINK_SPEED_1GB_FULL:
7555         speed_str = "1 Gbps";
7556         break;
7557     case IXGBE_LINK_SPEED_100_FULL:
7558         speed_str = "100 Mbps";
7559         break;
7560     case IXGBE_LINK_SPEED_10_FULL:
7561         speed_str = "10 Mbps";
7562         break;
7563     default:
7564         speed_str = "unknown speed";
7565         break;
7566     }
7567     e_info(drv, "NIC Link is Up %s, Flow Control: %s\n", speed_str,
7568            ((flow_rx && flow_tx) ? "RX/TX" :
7569            (flow_rx ? "RX" :
7570            (flow_tx ? "TX" : "None"))));
7571 
7572     netif_carrier_on(netdev);
7573     ixgbe_check_vf_rate_limit(adapter);
7574 
7575     /* enable transmits */
7576     netif_tx_wake_all_queues(adapter->netdev);
7577 
7578     /* update the default user priority for VFs */
7579     ixgbe_update_default_up(adapter);
7580 
7581     /* ping all the active vfs to let them know link has changed */
7582     ixgbe_ping_all_vfs(adapter);
7583 }
7584 
7585 /**
7586  * ixgbe_watchdog_link_is_down - update netif_carrier status and
7587  *                               print link down message
7588  * @adapter: pointer to the adapter structure
7589  **/
7590 static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
7591 {
7592     struct net_device *netdev = adapter->netdev;
7593     struct ixgbe_hw *hw = &adapter->hw;
7594 
7595     adapter->link_up = false;
7596     adapter->link_speed = 0;
7597 
7598     /* only continue if link was up previously */
7599     if (!netif_carrier_ok(netdev))
7600         return;
7601 
7602     /* poll for SFP+ cable when link is down */
7603     if (ixgbe_is_sfp(hw) && hw->mac.type == ixgbe_mac_82598EB)
7604         adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
7605 
7606     if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
7607         ixgbe_ptp_start_cyclecounter(adapter);
7608 
7609     e_info(drv, "NIC Link is Down\n");
7610     netif_carrier_off(netdev);
7611 
7612     /* ping all the active vfs to let them know link has changed */
7613     ixgbe_ping_all_vfs(adapter);
7614 }
7615 
7616 static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
7617 {
7618     int i;
7619 
7620     for (i = 0; i < adapter->num_tx_queues; i++) {
7621         struct ixgbe_ring *tx_ring = adapter->tx_ring[i];
7622 
7623         if (tx_ring->next_to_use != tx_ring->next_to_clean)
7624             return true;
7625     }
7626 
7627     for (i = 0; i < adapter->num_xdp_queues; i++) {
7628         struct ixgbe_ring *ring = adapter->xdp_ring[i];
7629 
7630         if (ring->next_to_use != ring->next_to_clean)
7631             return true;
7632     }
7633 
7634     return false;
7635 }
7636 
7637 static bool ixgbe_vf_tx_pending(struct ixgbe_adapter *adapter)
7638 {
7639     struct ixgbe_hw *hw = &adapter->hw;
7640     struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
7641     u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
7642 
7643     int i, j;
7644 
7645     if (!adapter->num_vfs)
7646         return false;
7647 
7648     /* resetting the PF is only needed for MAC before X550 */
7649     if (hw->mac.type >= ixgbe_mac_X550)
7650         return false;
7651 
7652     for (i = 0; i < adapter->num_vfs; i++) {
7653         for (j = 0; j < q_per_pool; j++) {
7654             u32 h, t;
7655 
7656             h = IXGBE_READ_REG(hw, IXGBE_PVFTDHN(q_per_pool, i, j));
7657             t = IXGBE_READ_REG(hw, IXGBE_PVFTDTN(q_per_pool, i, j));
7658 
7659             if (h != t)
7660                 return true;
7661         }
7662     }
7663 
7664     return false;
7665 }
7666 
7667 /**
7668  * ixgbe_watchdog_flush_tx - flush queues on link down
7669  * @adapter: pointer to the device adapter structure
7670  **/
7671 static void ixgbe_watchdog_flush_tx(struct ixgbe_adapter *adapter)
7672 {
7673     if (!netif_carrier_ok(adapter->netdev)) {
7674         if (ixgbe_ring_tx_pending(adapter) ||
7675             ixgbe_vf_tx_pending(adapter)) {
7676             /* We've lost link, so the controller stops DMA,
7677              * but we've got queued Tx work that's never going
7678              * to get done, so reset controller to flush Tx.
7679              * (Do the reset outside of interrupt context).
7680              */
7681             e_warn(drv, "initiating reset to clear Tx work after link loss\n");
7682             set_bit(__IXGBE_RESET_REQUESTED, &adapter->state);
7683         }
7684     }
7685 }
7686 
7687 #ifdef CONFIG_PCI_IOV
7688 static void ixgbe_bad_vf_abort(struct ixgbe_adapter *adapter, u32 vf)
7689 {
7690     struct ixgbe_hw *hw = &adapter->hw;
7691 
7692     if (adapter->hw.mac.type == ixgbe_mac_82599EB &&
7693         adapter->flags2 & IXGBE_FLAG2_AUTO_DISABLE_VF) {
7694         adapter->vfinfo[vf].primary_abort_count++;
7695         if (adapter->vfinfo[vf].primary_abort_count ==
7696             IXGBE_PRIMARY_ABORT_LIMIT) {
7697             ixgbe_set_vf_link_state(adapter, vf,
7698                         IFLA_VF_LINK_STATE_DISABLE);
7699             adapter->vfinfo[vf].primary_abort_count = 0;
7700 
7701             e_info(drv,
7702                    "Malicious Driver Detection event detected on PF %d VF %d MAC: %pM mdd-disable-vf=on",
7703                    hw->bus.func, vf,
7704                    adapter->vfinfo[vf].vf_mac_addresses);
7705         }
7706     }
7707 }
7708 
7709 static void ixgbe_check_for_bad_vf(struct ixgbe_adapter *adapter)
7710 {
7711     struct ixgbe_hw *hw = &adapter->hw;
7712     struct pci_dev *pdev = adapter->pdev;
7713     unsigned int vf;
7714     u32 gpc;
7715 
7716     if (!(netif_carrier_ok(adapter->netdev)))
7717         return;
7718 
7719     gpc = IXGBE_READ_REG(hw, IXGBE_TXDGPC);
7720     if (gpc) /* If incrementing then no need for the check below */
7721         return;
7722     /* Check to see if a bad DMA write target from an errant or
7723      * malicious VF has caused a PCIe error.  If so then we can
7724      * issue a VFLR to the offending VF(s) and then resume without
7725      * requesting a full slot reset.
7726      */
7727 
7728     if (!pdev)
7729         return;
7730 
7731     /* check status reg for all VFs owned by this PF */
7732     for (vf = 0; vf < adapter->num_vfs; ++vf) {
7733         struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
7734         u16 status_reg;
7735 
7736         if (!vfdev)
7737             continue;
7738         pci_read_config_word(vfdev, PCI_STATUS, &status_reg);
7739         if (status_reg != IXGBE_FAILED_READ_CFG_WORD &&
7740             status_reg & PCI_STATUS_REC_MASTER_ABORT) {
7741             ixgbe_bad_vf_abort(adapter, vf);
7742             pcie_flr(vfdev);
7743         }
7744     }
7745 }
7746 
7747 static void ixgbe_spoof_check(struct ixgbe_adapter *adapter)
7748 {
7749     u32 ssvpc;
7750 
7751     /* Do not perform spoof check for 82598 or if not in IOV mode */
7752     if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
7753         adapter->num_vfs == 0)
7754         return;
7755 
7756     ssvpc = IXGBE_READ_REG(&adapter->hw, IXGBE_SSVPC);
7757 
7758     /*
7759      * ssvpc register is cleared on read, if zero then no
7760      * spoofed packets in the last interval.
7761      */
7762     if (!ssvpc)
7763         return;
7764 
7765     e_warn(drv, "%u Spoofed packets detected\n", ssvpc);
7766 }
7767 #else
7768 static void ixgbe_spoof_check(struct ixgbe_adapter __always_unused *adapter)
7769 {
7770 }
7771 
7772 static void
7773 ixgbe_check_for_bad_vf(struct ixgbe_adapter __always_unused *adapter)
7774 {
7775 }
7776 #endif /* CONFIG_PCI_IOV */
7777 
7778 
7779 /**
7780  * ixgbe_watchdog_subtask - check and bring link up
7781  * @adapter: pointer to the device adapter structure
7782  **/
7783 static void ixgbe_watchdog_subtask(struct ixgbe_adapter *adapter)
7784 {
7785     /* if interface is down, removing or resetting, do nothing */
7786     if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7787         test_bit(__IXGBE_REMOVING, &adapter->state) ||
7788         test_bit(__IXGBE_RESETTING, &adapter->state))
7789         return;
7790 
7791     ixgbe_watchdog_update_link(adapter);
7792 
7793     if (adapter->link_up)
7794         ixgbe_watchdog_link_is_up(adapter);
7795     else
7796         ixgbe_watchdog_link_is_down(adapter);
7797 
7798     ixgbe_check_for_bad_vf(adapter);
7799     ixgbe_spoof_check(adapter);
7800     ixgbe_update_stats(adapter);
7801 
7802     ixgbe_watchdog_flush_tx(adapter);
7803 }
7804 
7805 /**
7806  * ixgbe_sfp_detection_subtask - poll for SFP+ cable
7807  * @adapter: the ixgbe adapter structure
7808  **/
7809 static void ixgbe_sfp_detection_subtask(struct ixgbe_adapter *adapter)
7810 {
7811     struct ixgbe_hw *hw = &adapter->hw;
7812     s32 err;
7813 
7814     /* not searching for SFP so there is nothing to do here */
7815     if (!(adapter->flags2 & IXGBE_FLAG2_SEARCH_FOR_SFP) &&
7816         !(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7817         return;
7818 
7819     if (adapter->sfp_poll_time &&
7820         time_after(adapter->sfp_poll_time, jiffies))
7821         return; /* If not yet time to poll for SFP */
7822 
7823     /* someone else is in init, wait until next service event */
7824     if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7825         return;
7826 
7827     adapter->sfp_poll_time = jiffies + IXGBE_SFP_POLL_JIFFIES - 1;
7828 
7829     err = hw->phy.ops.identify_sfp(hw);
7830     if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7831         goto sfp_out;
7832 
7833     if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
7834         /* If no cable is present, then we need to reset
7835          * the next time we find a good cable. */
7836         adapter->flags2 |= IXGBE_FLAG2_SFP_NEEDS_RESET;
7837     }
7838 
7839     /* exit on error */
7840     if (err)
7841         goto sfp_out;
7842 
7843     /* exit if reset not needed */
7844     if (!(adapter->flags2 & IXGBE_FLAG2_SFP_NEEDS_RESET))
7845         goto sfp_out;
7846 
7847     adapter->flags2 &= ~IXGBE_FLAG2_SFP_NEEDS_RESET;
7848 
7849     /*
7850      * A module may be identified correctly, but the EEPROM may not have
7851      * support for that module.  setup_sfp() will fail in that case, so
7852      * we should not allow that module to load.
7853      */
7854     if (hw->mac.type == ixgbe_mac_82598EB)
7855         err = hw->phy.ops.reset(hw);
7856     else
7857         err = hw->mac.ops.setup_sfp(hw);
7858 
7859     if (err == IXGBE_ERR_SFP_NOT_SUPPORTED)
7860         goto sfp_out;
7861 
7862     adapter->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
7863     e_info(probe, "detected SFP+: %d\n", hw->phy.sfp_type);
7864 
7865 sfp_out:
7866     clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7867 
7868     if ((err == IXGBE_ERR_SFP_NOT_SUPPORTED) &&
7869         (adapter->netdev->reg_state == NETREG_REGISTERED)) {
7870         e_dev_err("failed to initialize because an unsupported "
7871               "SFP+ module type was detected.\n");
7872         e_dev_err("Reload the driver after installing a "
7873               "supported module.\n");
7874         unregister_netdev(adapter->netdev);
7875     }
7876 }
7877 
7878 /**
7879  * ixgbe_sfp_link_config_subtask - set up link SFP after module install
7880  * @adapter: the ixgbe adapter structure
7881  **/
7882 static void ixgbe_sfp_link_config_subtask(struct ixgbe_adapter *adapter)
7883 {
7884     struct ixgbe_hw *hw = &adapter->hw;
7885     u32 cap_speed;
7886     u32 speed;
7887     bool autoneg = false;
7888 
7889     if (!(adapter->flags & IXGBE_FLAG_NEED_LINK_CONFIG))
7890         return;
7891 
7892     /* someone else is in init, wait until next service event */
7893     if (test_and_set_bit(__IXGBE_IN_SFP_INIT, &adapter->state))
7894         return;
7895 
7896     adapter->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
7897 
7898     hw->mac.ops.get_link_capabilities(hw, &cap_speed, &autoneg);
7899 
7900     /* advertise highest capable link speed */
7901     if (!autoneg && (cap_speed & IXGBE_LINK_SPEED_10GB_FULL))
7902         speed = IXGBE_LINK_SPEED_10GB_FULL;
7903     else
7904         speed = cap_speed & (IXGBE_LINK_SPEED_10GB_FULL |
7905                      IXGBE_LINK_SPEED_1GB_FULL);
7906 
7907     if (hw->mac.ops.setup_link)
7908         hw->mac.ops.setup_link(hw, speed, true);
7909 
7910     adapter->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
7911     adapter->link_check_timeout = jiffies;
7912     clear_bit(__IXGBE_IN_SFP_INIT, &adapter->state);
7913 }
7914 
7915 /**
7916  * ixgbe_service_timer - Timer Call-back
7917  * @t: pointer to timer_list structure
7918  **/
7919 static void ixgbe_service_timer(struct timer_list *t)
7920 {
7921     struct ixgbe_adapter *adapter = from_timer(adapter, t, service_timer);
7922     unsigned long next_event_offset;
7923 
7924     /* poll faster when waiting for link */
7925     if (adapter->flags & IXGBE_FLAG_NEED_LINK_UPDATE)
7926         next_event_offset = HZ / 10;
7927     else
7928         next_event_offset = HZ * 2;
7929 
7930     /* Reset the timer */
7931     mod_timer(&adapter->service_timer, next_event_offset + jiffies);
7932 
7933     ixgbe_service_event_schedule(adapter);
7934 }
7935 
7936 static void ixgbe_phy_interrupt_subtask(struct ixgbe_adapter *adapter)
7937 {
7938     struct ixgbe_hw *hw = &adapter->hw;
7939     u32 status;
7940 
7941     if (!(adapter->flags2 & IXGBE_FLAG2_PHY_INTERRUPT))
7942         return;
7943 
7944     adapter->flags2 &= ~IXGBE_FLAG2_PHY_INTERRUPT;
7945 
7946     if (!hw->phy.ops.handle_lasi)
7947         return;
7948 
7949     status = hw->phy.ops.handle_lasi(&adapter->hw);
7950     if (status != IXGBE_ERR_OVERTEMP)
7951         return;
7952 
7953     e_crit(drv, "%s\n", ixgbe_overheat_msg);
7954 }
7955 
7956 static void ixgbe_reset_subtask(struct ixgbe_adapter *adapter)
7957 {
7958     if (!test_and_clear_bit(__IXGBE_RESET_REQUESTED, &adapter->state))
7959         return;
7960 
7961     rtnl_lock();
7962     /* If we're already down, removing or resetting, just bail */
7963     if (test_bit(__IXGBE_DOWN, &adapter->state) ||
7964         test_bit(__IXGBE_REMOVING, &adapter->state) ||
7965         test_bit(__IXGBE_RESETTING, &adapter->state)) {
7966         rtnl_unlock();
7967         return;
7968     }
7969 
7970     ixgbe_dump(adapter);
7971     netdev_err(adapter->netdev, "Reset adapter\n");
7972     adapter->tx_timeout_count++;
7973 
7974     ixgbe_reinit_locked(adapter);
7975     rtnl_unlock();
7976 }
7977 
7978 /**
7979  * ixgbe_check_fw_error - Check firmware for errors
7980  * @adapter: the adapter private structure
7981  *
7982  * Check firmware errors in register FWSM
7983  */
7984 static bool ixgbe_check_fw_error(struct ixgbe_adapter *adapter)
7985 {
7986     struct ixgbe_hw *hw = &adapter->hw;
7987     u32 fwsm;
7988 
7989     /* read fwsm.ext_err_ind register and log errors */
7990     fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));
7991 
7992     if (fwsm & IXGBE_FWSM_EXT_ERR_IND_MASK ||
7993         !(fwsm & IXGBE_FWSM_FW_VAL_BIT))
7994         e_dev_warn("Warning firmware error detected FWSM: 0x%08X\n",
7995                fwsm);
7996 
7997     if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
7998         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");
7999         return true;
8000     }
8001 
8002     return false;
8003 }
8004 
8005 /**
8006  * ixgbe_service_task - manages and runs subtasks
8007  * @work: pointer to work_struct containing our data
8008  **/
8009 static void ixgbe_service_task(struct work_struct *work)
8010 {
8011     struct ixgbe_adapter *adapter = container_of(work,
8012                              struct ixgbe_adapter,
8013                              service_task);
8014     if (ixgbe_removed(adapter->hw.hw_addr)) {
8015         if (!test_bit(__IXGBE_DOWN, &adapter->state)) {
8016             rtnl_lock();
8017             ixgbe_down(adapter);
8018             rtnl_unlock();
8019         }
8020         ixgbe_service_event_complete(adapter);
8021         return;
8022     }
8023     if (ixgbe_check_fw_error(adapter)) {
8024         if (!test_bit(__IXGBE_DOWN, &adapter->state))
8025             unregister_netdev(adapter->netdev);
8026         ixgbe_service_event_complete(adapter);
8027         return;
8028     }
8029     ixgbe_reset_subtask(adapter);
8030     ixgbe_phy_interrupt_subtask(adapter);
8031     ixgbe_sfp_detection_subtask(adapter);
8032     ixgbe_sfp_link_config_subtask(adapter);
8033     ixgbe_check_overtemp_subtask(adapter);
8034     ixgbe_watchdog_subtask(adapter);
8035     ixgbe_fdir_reinit_subtask(adapter);
8036     ixgbe_check_hang_subtask(adapter);
8037 
8038     if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state)) {
8039         ixgbe_ptp_overflow_check(adapter);
8040         if (adapter->flags & IXGBE_FLAG_RX_HWTSTAMP_IN_REGISTER)
8041             ixgbe_ptp_rx_hang(adapter);
8042         ixgbe_ptp_tx_hang(adapter);
8043     }
8044 
8045     ixgbe_service_event_complete(adapter);
8046 }
8047 
8048 static int ixgbe_tso(struct ixgbe_ring *tx_ring,
8049              struct ixgbe_tx_buffer *first,
8050              u8 *hdr_len,
8051              struct ixgbe_ipsec_tx_data *itd)
8052 {
8053     u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
8054     struct sk_buff *skb = first->skb;
8055     union {
8056         struct iphdr *v4;
8057         struct ipv6hdr *v6;
8058         unsigned char *hdr;
8059     } ip;
8060     union {
8061         struct tcphdr *tcp;
8062         struct udphdr *udp;
8063         unsigned char *hdr;
8064     } l4;
8065     u32 paylen, l4_offset;
8066     u32 fceof_saidx = 0;
8067     int err;
8068 
8069     if (skb->ip_summed != CHECKSUM_PARTIAL)
8070         return 0;
8071 
8072     if (!skb_is_gso(skb))
8073         return 0;
8074 
8075     err = skb_cow_head(skb, 0);
8076     if (err < 0)
8077         return err;
8078 
8079     if (eth_p_mpls(first->protocol))
8080         ip.hdr = skb_inner_network_header(skb);
8081     else
8082         ip.hdr = skb_network_header(skb);
8083     l4.hdr = skb_checksum_start(skb);
8084 
8085     /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
8086     type_tucmd = (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) ?
8087               IXGBE_ADVTXD_TUCMD_L4T_UDP : IXGBE_ADVTXD_TUCMD_L4T_TCP;
8088 
8089     /* initialize outer IP header fields */
8090     if (ip.v4->version == 4) {
8091         unsigned char *csum_start = skb_checksum_start(skb);
8092         unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
8093         int len = csum_start - trans_start;
8094 
8095         /* IP header will have to cancel out any data that
8096          * is not a part of the outer IP header, so set to
8097          * a reverse csum if needed, else init check to 0.
8098          */
8099         ip.v4->check = (skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL) ?
8100                        csum_fold(csum_partial(trans_start,
8101                                   len, 0)) : 0;
8102         type_tucmd |= IXGBE_ADVTXD_TUCMD_IPV4;
8103 
8104         ip.v4->tot_len = 0;
8105         first->tx_flags |= IXGBE_TX_FLAGS_TSO |
8106                    IXGBE_TX_FLAGS_CSUM |
8107                    IXGBE_TX_FLAGS_IPV4;
8108     } else {
8109         ip.v6->payload_len = 0;
8110         first->tx_flags |= IXGBE_TX_FLAGS_TSO |
8111                    IXGBE_TX_FLAGS_CSUM;
8112     }
8113 
8114     /* determine offset of inner transport header */
8115     l4_offset = l4.hdr - skb->data;
8116 
8117     /* remove payload length from inner checksum */
8118     paylen = skb->len - l4_offset;
8119 
8120     if (type_tucmd & IXGBE_ADVTXD_TUCMD_L4T_TCP) {
8121         /* compute length of segmentation header */
8122         *hdr_len = (l4.tcp->doff * 4) + l4_offset;
8123         csum_replace_by_diff(&l4.tcp->check,
8124                      (__force __wsum)htonl(paylen));
8125     } else {
8126         /* compute length of segmentation header */
8127         *hdr_len = sizeof(*l4.udp) + l4_offset;
8128         csum_replace_by_diff(&l4.udp->check,
8129                      (__force __wsum)htonl(paylen));
8130     }
8131 
8132     /* update gso size and bytecount with header size */
8133     first->gso_segs = skb_shinfo(skb)->gso_segs;
8134     first->bytecount += (first->gso_segs - 1) * *hdr_len;
8135 
8136     /* mss_l4len_id: use 0 as index for TSO */
8137     mss_l4len_idx = (*hdr_len - l4_offset) << IXGBE_ADVTXD_L4LEN_SHIFT;
8138     mss_l4len_idx |= skb_shinfo(skb)->gso_size << IXGBE_ADVTXD_MSS_SHIFT;
8139 
8140     fceof_saidx |= itd->sa_idx;
8141     type_tucmd |= itd->flags | itd->trailer_len;
8142 
8143     /* vlan_macip_lens: HEADLEN, MACLEN, VLAN tag */
8144     vlan_macip_lens = l4.hdr - ip.hdr;
8145     vlan_macip_lens |= (ip.hdr - skb->data) << IXGBE_ADVTXD_MACLEN_SHIFT;
8146     vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
8147 
8148     ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd,
8149               mss_l4len_idx);
8150 
8151     return 1;
8152 }
8153 
8154 static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
8155               struct ixgbe_tx_buffer *first,
8156               struct ixgbe_ipsec_tx_data *itd)
8157 {
8158     struct sk_buff *skb = first->skb;
8159     u32 vlan_macip_lens = 0;
8160     u32 fceof_saidx = 0;
8161     u32 type_tucmd = 0;
8162 
8163     if (skb->ip_summed != CHECKSUM_PARTIAL) {
8164 csum_failed:
8165         if (!(first->tx_flags & (IXGBE_TX_FLAGS_HW_VLAN |
8166                      IXGBE_TX_FLAGS_CC)))
8167             return;
8168         goto no_csum;
8169     }
8170 
8171     switch (skb->csum_offset) {
8172     case offsetof(struct tcphdr, check):
8173         type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_TCP;
8174         fallthrough;
8175     case offsetof(struct udphdr, check):
8176         break;
8177     case offsetof(struct sctphdr, checksum):
8178         /* validate that this is actually an SCTP request */
8179         if (skb_csum_is_sctp(skb)) {
8180             type_tucmd = IXGBE_ADVTXD_TUCMD_L4T_SCTP;
8181             break;
8182         }
8183         fallthrough;
8184     default:
8185         skb_checksum_help(skb);
8186         goto csum_failed;
8187     }
8188 
8189     /* update TX checksum flag */
8190     first->tx_flags |= IXGBE_TX_FLAGS_CSUM;
8191     vlan_macip_lens = skb_checksum_start_offset(skb) -
8192               skb_network_offset(skb);
8193 no_csum:
8194     /* vlan_macip_lens: MACLEN, VLAN tag */
8195     vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
8196     vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
8197 
8198     fceof_saidx |= itd->sa_idx;
8199     type_tucmd |= itd->flags | itd->trailer_len;
8200 
8201     ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd, 0);
8202 }
8203 
8204 #define IXGBE_SET_FLAG(_input, _flag, _result) \
8205     ((_flag <= _result) ? \
8206      ((u32)(_input & _flag) * (_result / _flag)) : \
8207      ((u32)(_input & _flag) / (_flag / _result)))
8208 
8209 static u32 ixgbe_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
8210 {
8211     /* set type for advanced descriptor with frame checksum insertion */
8212     u32 cmd_type = IXGBE_ADVTXD_DTYP_DATA |
8213                IXGBE_ADVTXD_DCMD_DEXT |
8214                IXGBE_ADVTXD_DCMD_IFCS;
8215 
8216     /* set HW vlan bit if vlan is present */
8217     cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_HW_VLAN,
8218                    IXGBE_ADVTXD_DCMD_VLE);
8219 
8220     /* set segmentation enable bits for TSO/FSO */
8221     cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSO,
8222                    IXGBE_ADVTXD_DCMD_TSE);
8223 
8224     /* set timestamp bit if present */
8225     cmd_type |= IXGBE_SET_FLAG(tx_flags, IXGBE_TX_FLAGS_TSTAMP,
8226                    IXGBE_ADVTXD_MAC_TSTAMP);
8227 
8228     /* insert frame checksum */
8229     cmd_type ^= IXGBE_SET_FLAG(skb->no_fcs, 1, IXGBE_ADVTXD_DCMD_IFCS);
8230 
8231     return cmd_type;
8232 }
8233 
8234 static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
8235                    u32 tx_flags, unsigned int paylen)
8236 {
8237     u32 olinfo_status = paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
8238 
8239     /* enable L4 checksum for TSO and TX checksum offload */
8240     olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8241                     IXGBE_TX_FLAGS_CSUM,
8242                     IXGBE_ADVTXD_POPTS_TXSM);
8243 
8244     /* enable IPv4 checksum for TSO */
8245     olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8246                     IXGBE_TX_FLAGS_IPV4,
8247                     IXGBE_ADVTXD_POPTS_IXSM);
8248 
8249     /* enable IPsec */
8250     olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8251                     IXGBE_TX_FLAGS_IPSEC,
8252                     IXGBE_ADVTXD_POPTS_IPSEC);
8253 
8254     /*
8255      * Check Context must be set if Tx switch is enabled, which it
8256      * always is for case where virtual functions are running
8257      */
8258     olinfo_status |= IXGBE_SET_FLAG(tx_flags,
8259                     IXGBE_TX_FLAGS_CC,
8260                     IXGBE_ADVTXD_CC);
8261 
8262     tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
8263 }
8264 
8265 static int __ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
8266 {
8267     netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
8268 
8269     /* Herbert's original patch had:
8270      *  smp_mb__after_netif_stop_queue();
8271      * but since that doesn't exist yet, just open code it.
8272      */
8273     smp_mb();
8274 
8275     /* We need to check again in a case another CPU has just
8276      * made room available.
8277      */
8278     if (likely(ixgbe_desc_unused(tx_ring) < size))
8279         return -EBUSY;
8280 
8281     /* A reprieve! - use start_queue because it doesn't call schedule */
8282     netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
8283     ++tx_ring->tx_stats.restart_queue;
8284     return 0;
8285 }
8286 
8287 static inline int ixgbe_maybe_stop_tx(struct ixgbe_ring *tx_ring, u16 size)
8288 {
8289     if (likely(ixgbe_desc_unused(tx_ring) >= size))
8290         return 0;
8291 
8292     return __ixgbe_maybe_stop_tx(tx_ring, size);
8293 }
8294 
8295 static int ixgbe_tx_map(struct ixgbe_ring *tx_ring,
8296             struct ixgbe_tx_buffer *first,
8297             const u8 hdr_len)
8298 {
8299     struct sk_buff *skb = first->skb;
8300     struct ixgbe_tx_buffer *tx_buffer;
8301     union ixgbe_adv_tx_desc *tx_desc;
8302     skb_frag_t *frag;
8303     dma_addr_t dma;
8304     unsigned int data_len, size;
8305     u32 tx_flags = first->tx_flags;
8306     u32 cmd_type = ixgbe_tx_cmd_type(skb, tx_flags);
8307     u16 i = tx_ring->next_to_use;
8308 
8309     tx_desc = IXGBE_TX_DESC(tx_ring, i);
8310 
8311     ixgbe_tx_olinfo_status(tx_desc, tx_flags, skb->len - hdr_len);
8312 
8313     size = skb_headlen(skb);
8314     data_len = skb->data_len;
8315 
8316 #ifdef IXGBE_FCOE
8317     if (tx_flags & IXGBE_TX_FLAGS_FCOE) {
8318         if (data_len < sizeof(struct fcoe_crc_eof)) {
8319             size -= sizeof(struct fcoe_crc_eof) - data_len;
8320             data_len = 0;
8321         } else {
8322             data_len -= sizeof(struct fcoe_crc_eof);
8323         }
8324     }
8325 
8326 #endif
8327     dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
8328 
8329     tx_buffer = first;
8330 
8331     for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
8332         if (dma_mapping_error(tx_ring->dev, dma))
8333             goto dma_error;
8334 
8335         /* record length, and DMA address */
8336         dma_unmap_len_set(tx_buffer, len, size);
8337         dma_unmap_addr_set(tx_buffer, dma, dma);
8338 
8339         tx_desc->read.buffer_addr = cpu_to_le64(dma);
8340 
8341         while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
8342             tx_desc->read.cmd_type_len =
8343                 cpu_to_le32(cmd_type ^ IXGBE_MAX_DATA_PER_TXD);
8344 
8345             i++;
8346             tx_desc++;
8347             if (i == tx_ring->count) {
8348                 tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8349                 i = 0;
8350             }
8351             tx_desc->read.olinfo_status = 0;
8352 
8353             dma += IXGBE_MAX_DATA_PER_TXD;
8354             size -= IXGBE_MAX_DATA_PER_TXD;
8355 
8356             tx_desc->read.buffer_addr = cpu_to_le64(dma);
8357         }
8358 
8359         if (likely(!data_len))
8360             break;
8361 
8362         tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
8363 
8364         i++;
8365         tx_desc++;
8366         if (i == tx_ring->count) {
8367             tx_desc = IXGBE_TX_DESC(tx_ring, 0);
8368             i = 0;
8369         }
8370         tx_desc->read.olinfo_status = 0;
8371 
8372 #ifdef IXGBE_FCOE
8373         size = min_t(unsigned int, data_len, skb_frag_size(frag));
8374 #else
8375         size = skb_frag_size(frag);
8376 #endif
8377         data_len -= size;
8378 
8379         dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
8380                        DMA_TO_DEVICE);
8381 
8382         tx_buffer = &tx_ring->tx_buffer_info[i];
8383     }
8384 
8385     /* write last descriptor with RS and EOP bits */
8386     cmd_type |= size | IXGBE_TXD_CMD;
8387     tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8388 
8389     netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
8390 
8391     /* set the timestamp */
8392     first->time_stamp = jiffies;
8393 
8394     skb_tx_timestamp(skb);
8395 
8396     /*
8397      * Force memory writes to complete before letting h/w know there
8398      * are new descriptors to fetch.  (Only applicable for weak-ordered
8399      * memory model archs, such as IA-64).
8400      *
8401      * We also need this memory barrier to make certain all of the
8402      * status bits have been updated before next_to_watch is written.
8403      */
8404     wmb();
8405 
8406     /* set next_to_watch value indicating a packet is present */
8407     first->next_to_watch = tx_desc;
8408 
8409     i++;
8410     if (i == tx_ring->count)
8411         i = 0;
8412 
8413     tx_ring->next_to_use = i;
8414 
8415     ixgbe_maybe_stop_tx(tx_ring, DESC_NEEDED);
8416 
8417     if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more()) {
8418         writel(i, tx_ring->tail);
8419     }
8420 
8421     return 0;
8422 dma_error:
8423     dev_err(tx_ring->dev, "TX DMA map failed\n");
8424 
8425     /* clear dma mappings for failed tx_buffer_info map */
8426     for (;;) {
8427         tx_buffer = &tx_ring->tx_buffer_info[i];
8428         if (dma_unmap_len(tx_buffer, len))
8429             dma_unmap_page(tx_ring->dev,
8430                        dma_unmap_addr(tx_buffer, dma),
8431                        dma_unmap_len(tx_buffer, len),
8432                        DMA_TO_DEVICE);
8433         dma_unmap_len_set(tx_buffer, len, 0);
8434         if (tx_buffer == first)
8435             break;
8436         if (i == 0)
8437             i += tx_ring->count;
8438         i--;
8439     }
8440 
8441     dev_kfree_skb_any(first->skb);
8442     first->skb = NULL;
8443 
8444     tx_ring->next_to_use = i;
8445 
8446     return -1;
8447 }
8448 
8449 static void ixgbe_atr(struct ixgbe_ring *ring,
8450               struct ixgbe_tx_buffer *first)
8451 {
8452     struct ixgbe_q_vector *q_vector = ring->q_vector;
8453     union ixgbe_atr_hash_dword input = { .dword = 0 };
8454     union ixgbe_atr_hash_dword common = { .dword = 0 };
8455     union {
8456         unsigned char *network;
8457         struct iphdr *ipv4;
8458         struct ipv6hdr *ipv6;
8459     } hdr;
8460     struct tcphdr *th;
8461     unsigned int hlen;
8462     struct sk_buff *skb;
8463     __be16 vlan_id;
8464     int l4_proto;
8465 
8466     /* if ring doesn't have a interrupt vector, cannot perform ATR */
8467     if (!q_vector)
8468         return;
8469 
8470     /* do nothing if sampling is disabled */
8471     if (!ring->atr_sample_rate)
8472         return;
8473 
8474     ring->atr_count++;
8475 
8476     /* currently only IPv4/IPv6 with TCP is supported */
8477     if ((first->protocol != htons(ETH_P_IP)) &&
8478         (first->protocol != htons(ETH_P_IPV6)))
8479         return;
8480 
8481     /* snag network header to get L4 type and address */
8482     skb = first->skb;
8483     hdr.network = skb_network_header(skb);
8484     if (unlikely(hdr.network <= skb->data))
8485         return;
8486     if (skb->encapsulation &&
8487         first->protocol == htons(ETH_P_IP) &&
8488         hdr.ipv4->protocol == IPPROTO_UDP) {
8489         struct ixgbe_adapter *adapter = q_vector->adapter;
8490 
8491         if (unlikely(skb_tail_pointer(skb) < hdr.network +
8492                  VXLAN_HEADROOM))
8493             return;
8494 
8495         /* verify the port is recognized as VXLAN */
8496         if (adapter->vxlan_port &&
8497             udp_hdr(skb)->dest == adapter->vxlan_port)
8498             hdr.network = skb_inner_network_header(skb);
8499 
8500         if (adapter->geneve_port &&
8501             udp_hdr(skb)->dest == adapter->geneve_port)
8502             hdr.network = skb_inner_network_header(skb);
8503     }
8504 
8505     /* Make sure we have at least [minimum IPv4 header + TCP]
8506      * or [IPv6 header] bytes
8507      */
8508     if (unlikely(skb_tail_pointer(skb) < hdr.network + 40))
8509         return;
8510 
8511     /* Currently only IPv4/IPv6 with TCP is supported */
8512     switch (hdr.ipv4->version) {
8513     case IPVERSION:
8514         /* access ihl as u8 to avoid unaligned access on ia64 */
8515         hlen = (hdr.network[0] & 0x0F) << 2;
8516         l4_proto = hdr.ipv4->protocol;
8517         break;
8518     case 6:
8519         hlen = hdr.network - skb->data;
8520         l4_proto = ipv6_find_hdr(skb, &hlen, IPPROTO_TCP, NULL, NULL);
8521         hlen -= hdr.network - skb->data;
8522         break;
8523     default:
8524         return;
8525     }
8526 
8527     if (l4_proto != IPPROTO_TCP)
8528         return;
8529 
8530     if (unlikely(skb_tail_pointer(skb) < hdr.network +
8531              hlen + sizeof(struct tcphdr)))
8532         return;
8533 
8534     th = (struct tcphdr *)(hdr.network + hlen);
8535 
8536     /* skip this packet since the socket is closing */
8537     if (th->fin)
8538         return;
8539 
8540     /* sample on all syn packets or once every atr sample count */
8541     if (!th->syn && (ring->atr_count < ring->atr_sample_rate))
8542         return;
8543 
8544     /* reset sample count */
8545     ring->atr_count = 0;
8546 
8547     vlan_id = htons(first->tx_flags >> IXGBE_TX_FLAGS_VLAN_SHIFT);
8548 
8549     /*
8550      * src and dst are inverted, think how the receiver sees them
8551      *
8552      * The input is broken into two sections, a non-compressed section
8553      * containing vm_pool, vlan_id, and flow_type.  The rest of the data
8554      * is XORed together and stored in the compressed dword.
8555      */
8556     input.formatted.vlan_id = vlan_id;
8557 
8558     /*
8559      * since src port and flex bytes occupy the same word XOR them together
8560      * and write the value to source port portion of compressed dword
8561      */
8562     if (first->tx_flags & (IXGBE_TX_FLAGS_SW_VLAN | IXGBE_TX_FLAGS_HW_VLAN))
8563         common.port.src ^= th->dest ^ htons(ETH_P_8021Q);
8564     else
8565         common.port.src ^= th->dest ^ first->protocol;
8566     common.port.dst ^= th->source;
8567 
8568     switch (hdr.ipv4->version) {
8569     case IPVERSION:
8570         input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV4;
8571         common.ip ^= hdr.ipv4->saddr ^ hdr.ipv4->daddr;
8572         break;
8573     case 6:
8574         input.formatted.flow_type = IXGBE_ATR_FLOW_TYPE_TCPV6;
8575         common.ip ^= hdr.ipv6->saddr.s6_addr32[0] ^
8576                  hdr.ipv6->saddr.s6_addr32[1] ^
8577                  hdr.ipv6->saddr.s6_addr32[2] ^
8578                  hdr.ipv6->saddr.s6_addr32[3] ^
8579                  hdr.ipv6->daddr.s6_addr32[0] ^
8580                  hdr.ipv6->daddr.s6_addr32[1] ^
8581                  hdr.ipv6->daddr.s6_addr32[2] ^
8582                  hdr.ipv6->daddr.s6_addr32[3];
8583         break;
8584     default:
8585         break;
8586     }
8587 
8588     if (hdr.network != skb_network_header(skb))
8589         input.formatted.flow_type |= IXGBE_ATR_L4TYPE_TUNNEL_MASK;
8590 
8591     /* This assumes the Rx queue and Tx queue are bound to the same CPU */
8592     ixgbe_fdir_add_signature_filter_82599(&q_vector->adapter->hw,
8593                           input, common, ring->queue_index);
8594 }
8595 
8596 #ifdef IXGBE_FCOE
8597 static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
8598                   struct net_device *sb_dev)
8599 {
8600     struct ixgbe_adapter *adapter;
8601     struct ixgbe_ring_feature *f;
8602     int txq;
8603 
8604     if (sb_dev) {
8605         u8 tc = netdev_get_prio_tc_map(dev, skb->priority);
8606         struct net_device *vdev = sb_dev;
8607 
8608         txq = vdev->tc_to_txq[tc].offset;
8609         txq += reciprocal_scale(skb_get_hash(skb),
8610                     vdev->tc_to_txq[tc].count);
8611 
8612         return txq;
8613     }
8614 
8615     /*
8616      * only execute the code below if protocol is FCoE
8617      * or FIP and we have FCoE enabled on the adapter
8618      */
8619     switch (vlan_get_protocol(skb)) {
8620     case htons(ETH_P_FCOE):
8621     case htons(ETH_P_FIP):
8622         adapter = netdev_priv(dev);
8623 
8624         if (!sb_dev && (adapter->flags & IXGBE_FLAG_FCOE_ENABLED))
8625             break;
8626         fallthrough;
8627     default:
8628         return netdev_pick_tx(dev, skb, sb_dev);
8629     }
8630 
8631     f = &adapter->ring_feature[RING_F_FCOE];
8632 
8633     txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) :
8634                        smp_processor_id();
8635 
8636     while (txq >= f->indices)
8637         txq -= f->indices;
8638 
8639     return txq + f->offset;
8640 }
8641 
8642 #endif
8643 int ixgbe_xmit_xdp_ring(struct ixgbe_ring *ring,
8644             struct xdp_frame *xdpf)
8645 {
8646     struct skb_shared_info *sinfo = xdp_get_shared_info_from_frame(xdpf);
8647     u8 nr_frags = unlikely(xdp_frame_has_frags(xdpf)) ? sinfo->nr_frags : 0;
8648     u16 i = 0, index = ring->next_to_use;
8649     struct ixgbe_tx_buffer *tx_head = &ring->tx_buffer_info[index];
8650     struct ixgbe_tx_buffer *tx_buff = tx_head;
8651     union ixgbe_adv_tx_desc *tx_desc = IXGBE_TX_DESC(ring, index);
8652     u32 cmd_type, len = xdpf->len;
8653     void *data = xdpf->data;
8654 
8655     if (unlikely(ixgbe_desc_unused(ring) < 1 + nr_frags))
8656         return IXGBE_XDP_CONSUMED;
8657 
8658     tx_head->bytecount = xdp_get_frame_len(xdpf);
8659     tx_head->gso_segs = 1;
8660     tx_head->xdpf = xdpf;
8661 
8662     tx_desc->read.olinfo_status =
8663         cpu_to_le32(tx_head->bytecount << IXGBE_ADVTXD_PAYLEN_SHIFT);
8664 
8665     for (;;) {
8666         dma_addr_t dma;
8667 
8668         dma = dma_map_single(ring->dev, data, len, DMA_TO_DEVICE);
8669         if (dma_mapping_error(ring->dev, dma))
8670             goto unmap;
8671 
8672         dma_unmap_len_set(tx_buff, len, len);
8673         dma_unmap_addr_set(tx_buff, dma, dma);
8674 
8675         cmd_type = IXGBE_ADVTXD_DTYP_DATA | IXGBE_ADVTXD_DCMD_DEXT |
8676                IXGBE_ADVTXD_DCMD_IFCS | len;
8677         tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
8678         tx_desc->read.buffer_addr = cpu_to_le64(dma);
8679         tx_buff->protocol = 0;
8680 
8681         if (++index == ring->count)
8682             index = 0;
8683 
8684         if (i == nr_frags)
8685             break;
8686 
8687         tx_buff = &ring->tx_buffer_info[index];
8688         tx_desc = IXGBE_TX_DESC(ring, index);
8689         tx_desc->read.olinfo_status = 0;
8690 
8691         data = skb_frag_address(&sinfo->frags[i]);
8692         len = skb_frag_size(&sinfo->frags[i]);
8693         i++;
8694     }
8695     /* put descriptor type bits */
8696     tx_desc->read.cmd_type_len |= cpu_to_le32(IXGBE_TXD_CMD);
8697 
8698     /* Avoid any potential race with xdp_xmit and cleanup */
8699     smp_wmb();
8700 
8701     tx_head->next_to_watch = tx_desc;
8702     ring->next_to_use = index;
8703 
8704     return IXGBE_XDP_TX;
8705 
8706 unmap:
8707     for (;;) {
8708         tx_buff = &ring->tx_buffer_info[index];
8709         if (dma_unmap_len(tx_buff, len))
8710             dma_unmap_page(ring->dev, dma_unmap_addr(tx_buff, dma),
8711                        dma_unmap_len(tx_buff, len),
8712                        DMA_TO_DEVICE);
8713         dma_unmap_len_set(tx_buff, len, 0);
8714         if (tx_buff == tx_head)
8715             break;
8716 
8717         if (!index)
8718             index += ring->count;
8719         index--;
8720     }
8721 
8722     return IXGBE_XDP_CONSUMED;
8723 }
8724 
8725 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
8726               struct ixgbe_adapter *adapter,
8727               struct ixgbe_ring *tx_ring)
8728 {
8729     struct ixgbe_tx_buffer *first;
8730     int tso;
8731     u32 tx_flags = 0;
8732     unsigned short f;
8733     u16 count = TXD_USE_COUNT(skb_headlen(skb));
8734     struct ixgbe_ipsec_tx_data ipsec_tx = { 0 };
8735     __be16 protocol = skb->protocol;
8736     u8 hdr_len = 0;
8737 
8738     /*
8739      * need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD,
8740      *       + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD,
8741      *       + 2 desc gap to keep tail from touching head,
8742      *       + 1 desc for context descriptor,
8743      * otherwise try next time
8744      */
8745     for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
8746         count += TXD_USE_COUNT(skb_frag_size(
8747                         &skb_shinfo(skb)->frags[f]));
8748 
8749     if (ixgbe_maybe_stop_tx(tx_ring, count + 3)) {
8750         tx_ring->tx_stats.tx_busy++;
8751         return NETDEV_TX_BUSY;
8752     }
8753 
8754     /* record the location of the first descriptor for this packet */
8755     first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
8756     first->skb = skb;
8757     first->bytecount = skb->len;
8758     first->gso_segs = 1;
8759 
8760     /* if we have a HW VLAN tag being added default to the HW one */
8761     if (skb_vlan_tag_present(skb)) {
8762         tx_flags |= skb_vlan_tag_get(skb) << IXGBE_TX_FLAGS_VLAN_SHIFT;
8763         tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8764     /* else if it is a SW VLAN check the next protocol and store the tag */
8765     } else if (protocol == htons(ETH_P_8021Q)) {
8766         struct vlan_hdr *vhdr, _vhdr;
8767         vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
8768         if (!vhdr)
8769             goto out_drop;
8770 
8771         tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
8772                   IXGBE_TX_FLAGS_VLAN_SHIFT;
8773         tx_flags |= IXGBE_TX_FLAGS_SW_VLAN;
8774     }
8775     protocol = vlan_get_protocol(skb);
8776 
8777     if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
8778         adapter->ptp_clock) {
8779         if (adapter->tstamp_config.tx_type == HWTSTAMP_TX_ON &&
8780             !test_and_set_bit_lock(__IXGBE_PTP_TX_IN_PROGRESS,
8781                        &adapter->state)) {
8782             skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
8783             tx_flags |= IXGBE_TX_FLAGS_TSTAMP;
8784 
8785             /* schedule check for Tx timestamp */
8786             adapter->ptp_tx_skb = skb_get(skb);
8787             adapter->ptp_tx_start = jiffies;
8788             schedule_work(&adapter->ptp_tx_work);
8789         } else {
8790             adapter->tx_hwtstamp_skipped++;
8791         }
8792     }
8793 
8794 #ifdef CONFIG_PCI_IOV
8795     /*
8796      * Use the l2switch_enable flag - would be false if the DMA
8797      * Tx switch had been disabled.
8798      */
8799     if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
8800         tx_flags |= IXGBE_TX_FLAGS_CC;
8801 
8802 #endif
8803     /* DCB maps skb priorities 0-7 onto 3 bit PCP of VLAN tag. */
8804     if ((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
8805         ((tx_flags & (IXGBE_TX_FLAGS_HW_VLAN | IXGBE_TX_FLAGS_SW_VLAN)) ||
8806          (skb->priority != TC_PRIO_CONTROL))) {
8807         tx_flags &= ~IXGBE_TX_FLAGS_VLAN_PRIO_MASK;
8808         tx_flags |= (skb->priority & 0x7) <<
8809                     IXGBE_TX_FLAGS_VLAN_PRIO_SHIFT;
8810         if (tx_flags & IXGBE_TX_FLAGS_SW_VLAN) {
8811             struct vlan_ethhdr *vhdr;
8812 
8813             if (skb_cow_head(skb, 0))
8814                 goto out_drop;
8815             vhdr = (struct vlan_ethhdr *)skb->data;
8816             vhdr->h_vlan_TCI = htons(tx_flags >>
8817                          IXGBE_TX_FLAGS_VLAN_SHIFT);
8818         } else {
8819             tx_flags |= IXGBE_TX_FLAGS_HW_VLAN;
8820         }
8821     }
8822 
8823     /* record initial flags and protocol */
8824     first->tx_flags = tx_flags;
8825     first->protocol = protocol;
8826 
8827 #ifdef IXGBE_FCOE
8828     /* setup tx offload for FCoE */
8829     if ((protocol == htons(ETH_P_FCOE)) &&
8830         (tx_ring->netdev->features & (NETIF_F_FSO | NETIF_F_FCOE_CRC))) {
8831         tso = ixgbe_fso(tx_ring, first, &hdr_len);
8832         if (tso < 0)
8833             goto out_drop;
8834 
8835         goto xmit_fcoe;
8836     }
8837 
8838 #endif /* IXGBE_FCOE */
8839 
8840 #ifdef CONFIG_IXGBE_IPSEC
8841     if (xfrm_offload(skb) &&
8842         !ixgbe_ipsec_tx(tx_ring, first, &ipsec_tx))
8843         goto out_drop;
8844 #endif
8845     tso = ixgbe_tso(tx_ring, first, &hdr_len, &ipsec_tx);
8846     if (tso < 0)
8847         goto out_drop;
8848     else if (!tso)
8849         ixgbe_tx_csum(tx_ring, first, &ipsec_tx);
8850 
8851     /* add the ATR filter if ATR is on */
8852     if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
8853         ixgbe_atr(tx_ring, first);
8854 
8855 #ifdef IXGBE_FCOE
8856 xmit_fcoe:
8857 #endif /* IXGBE_FCOE */
8858     if (ixgbe_tx_map(tx_ring, first, hdr_len))
8859         goto cleanup_tx_timestamp;
8860 
8861     return NETDEV_TX_OK;
8862 
8863 out_drop:
8864     dev_kfree_skb_any(first->skb);
8865     first->skb = NULL;
8866 cleanup_tx_timestamp:
8867     if (unlikely(tx_flags & IXGBE_TX_FLAGS_TSTAMP)) {
8868         dev_kfree_skb_any(adapter->ptp_tx_skb);
8869         adapter->ptp_tx_skb = NULL;
8870         cancel_work_sync(&adapter->ptp_tx_work);
8871         clear_bit_unlock(__IXGBE_PTP_TX_IN_PROGRESS, &adapter->state);
8872     }
8873 
8874     return NETDEV_TX_OK;
8875 }
8876 
8877 static netdev_tx_t __ixgbe_xmit_frame(struct sk_buff *skb,
8878                       struct net_device *netdev,
8879                       struct ixgbe_ring *ring)
8880 {
8881     struct ixgbe_adapter *adapter = netdev_priv(netdev);
8882     struct ixgbe_ring *tx_ring;
8883 
8884     /*
8885      * The minimum packet size for olinfo paylen is 17 so pad the skb
8886      * in order to meet this minimum size requirement.
8887      */
8888     if (skb_put_padto(skb, 17))
8889         return NETDEV_TX_OK;
8890 
8891     tx_ring = ring ? ring : adapter->tx_ring[skb_get_queue_mapping(skb)];
8892     if (unlikely(test_bit(__IXGBE_TX_DISABLED, &tx_ring->state)))
8893         return NETDEV_TX_BUSY;
8894 
8895     return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
8896 }
8897 
8898 static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
8899                     struct net_device *netdev)
8900 {
8901     return __ixgbe_xmit_frame(skb, netdev, NULL);
8902 }
8903 
8904 /**
8905  * ixgbe_set_mac - Change the Ethernet Address of the NIC
8906  * @netdev: network interface device structure
8907  * @p: pointer to an address structure
8908  *
8909  * Returns 0 on success, negative on failure
8910  **/
8911 static int ixgbe_set_mac(struct net_device *netdev, void *p)
8912 {
8913     struct ixgbe_adapter *adapter = netdev_priv(netdev);
8914     struct ixgbe_hw *hw = &adapter->hw;
8915     struct sockaddr *addr = p;
8916 
8917     if (!is_valid_ether_addr(addr->sa_data))
8918         return -EADDRNOTAVAIL;
8919 
8920     eth_hw_addr_set(netdev, addr->sa_data);
8921     memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
8922 
8923     ixgbe_mac_set_default_filter(adapter);
8924 
8925     return 0;
8926 }
8927 
8928 static int
8929 ixgbe_mdio_read(struct net_device *netdev, int prtad, int devad, u16 addr)
8930 {
8931     struct ixgbe_adapter *adapter = netdev_priv(netdev);
8932     struct ixgbe_hw *hw = &adapter->hw;
8933     u16 value;
8934     int rc;
8935 
8936     if (adapter->mii_bus) {
8937         int regnum = addr;
8938 
8939         if (devad != MDIO_DEVAD_NONE)
8940             regnum |= (devad << 16) | MII_ADDR_C45;
8941 
8942         return mdiobus_read(adapter->mii_bus, prtad, regnum);
8943     }
8944 
8945     if (prtad != hw->phy.mdio.prtad)
8946         return -EINVAL;
8947     rc = hw->phy.ops.read_reg(hw, addr, devad, &value);
8948     if (!rc)
8949         rc = value;
8950     return rc;
8951 }
8952 
8953 static int ixgbe_mdio_write(struct net_device *netdev, int prtad, int devad,
8954                 u16 addr, u16 value)
8955 {
8956     struct ixgbe_adapter *adapter = netdev_priv(netdev);
8957     struct ixgbe_hw *hw = &adapter->hw;
8958 
8959     if (adapter->mii_bus) {
8960         int regnum = addr;
8961 
8962         if (devad != MDIO_DEVAD_NONE)
8963             regnum |= (devad << 16) | MII_ADDR_C45;
8964 
8965         return mdiobus_write(adapter->mii_bus, prtad, regnum, value);
8966     }
8967 
8968     if (prtad != hw->phy.mdio.prtad)
8969         return -EINVAL;
8970     return hw->phy.ops.write_reg(hw, addr, devad, value);
8971 }
8972 
8973 static int ixgbe_ioctl(struct net_device *netdev, struct ifreq *req, int cmd)
8974 {
8975     struct ixgbe_adapter *adapter = netdev_priv(netdev);
8976 
8977     switch (cmd) {
8978     case SIOCSHWTSTAMP:
8979         return ixgbe_ptp_set_ts_config(adapter, req);
8980     case SIOCGHWTSTAMP:
8981         return ixgbe_ptp_get_ts_config(adapter, req);
8982     case SIOCGMIIPHY:
8983         if (!adapter->hw.phy.ops.read_reg)
8984             return -EOPNOTSUPP;
8985         fallthrough;
8986     default:
8987         return mdio_mii_ioctl(&adapter->hw.phy.mdio, if_mii(req), cmd);
8988     }
8989 }
8990 
8991 /**
8992  * ixgbe_add_sanmac_netdev - Add the SAN MAC address to the corresponding
8993  * netdev->dev_addrs
8994  * @dev: network interface device structure
8995  *
8996  * Returns non-zero on failure
8997  **/
8998 static int ixgbe_add_sanmac_netdev(struct net_device *dev)
8999 {
9000     int err = 0;
9001     struct ixgbe_adapter *adapter = netdev_priv(dev);
9002     struct ixgbe_hw *hw = &adapter->hw;
9003 
9004     if (is_valid_ether_addr(hw->mac.san_addr)) {
9005         rtnl_lock();
9006         err = dev_addr_add(dev, hw->mac.san_addr, NETDEV_HW_ADDR_T_SAN);
9007         rtnl_unlock();
9008 
9009         /* update SAN MAC vmdq pool selection */
9010         hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
9011     }
9012     return err;
9013 }
9014 
9015 /**
9016  * ixgbe_del_sanmac_netdev - Removes the SAN MAC address to the corresponding
9017  * netdev->dev_addrs
9018  * @dev: network interface device structure
9019  *
9020  * Returns non-zero on failure
9021  **/
9022 static int ixgbe_del_sanmac_netdev(struct net_device *dev)
9023 {
9024     int err = 0;
9025     struct ixgbe_adapter *adapter = netdev_priv(dev);
9026     struct ixgbe_mac_info *mac = &adapter->hw.mac;
9027 
9028     if (is_valid_ether_addr(mac->san_addr)) {
9029         rtnl_lock();
9030         err = dev_addr_del(dev, mac->san_addr, NETDEV_HW_ADDR_T_SAN);
9031         rtnl_unlock();
9032     }
9033     return err;
9034 }
9035 
9036 static void ixgbe_get_ring_stats64(struct rtnl_link_stats64 *stats,
9037                    struct ixgbe_ring *ring)
9038 {
9039     u64 bytes, packets;
9040     unsigned int start;
9041 
9042     if (ring) {
9043         do {
9044             start = u64_stats_fetch_begin_irq(&ring->syncp);
9045             packets = ring->stats.packets;
9046             bytes   = ring->stats.bytes;
9047         } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
9048         stats->tx_packets += packets;
9049         stats->tx_bytes   += bytes;
9050     }
9051 }
9052 
9053 static void ixgbe_get_stats64(struct net_device *netdev,
9054                   struct rtnl_link_stats64 *stats)
9055 {
9056     struct ixgbe_adapter *adapter = netdev_priv(netdev);
9057     int i;
9058 
9059     rcu_read_lock();
9060     for (i = 0; i < adapter->num_rx_queues; i++) {
9061         struct ixgbe_ring *ring = READ_ONCE(adapter->rx_ring[i]);
9062         u64 bytes, packets;
9063         unsigned int start;
9064 
9065         if (ring) {
9066             do {
9067                 start = u64_stats_fetch_begin_irq(&ring->syncp);
9068                 packets = ring->stats.packets;
9069                 bytes   = ring->stats.bytes;
9070             } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
9071             stats->rx_packets += packets;
9072             stats->rx_bytes   += bytes;
9073         }
9074     }
9075 
9076     for (i = 0; i < adapter->num_tx_queues; i++) {
9077         struct ixgbe_ring *ring = READ_ONCE(adapter->tx_ring[i]);
9078 
9079         ixgbe_get_ring_stats64(stats, ring);
9080     }
9081     for (i = 0; i < adapter->num_xdp_queues; i++) {
9082         struct ixgbe_ring *ring = READ_ONCE(adapter->xdp_ring[i]);
9083 
9084         ixgbe_get_ring_stats64(stats, ring);
9085     }
9086     rcu_read_unlock();
9087 
9088     /* following stats updated by ixgbe_watchdog_task() */
9089     stats->multicast    = netdev->stats.multicast;
9090     stats->rx_errors    = netdev->stats.rx_errors;
9091     stats->rx_length_errors = netdev->stats.rx_length_errors;
9092     stats->rx_crc_errors    = netdev->stats.rx_crc_errors;
9093     stats->rx_missed_errors = netdev->stats.rx_missed_errors;
9094 }
9095 
9096 static int ixgbe_ndo_get_vf_stats(struct net_device *netdev, int vf,
9097                   struct ifla_vf_stats *vf_stats)
9098 {
9099     struct ixgbe_adapter *adapter = netdev_priv(netdev);
9100 
9101     if (vf < 0 || vf >= adapter->num_vfs)
9102         return -EINVAL;
9103 
9104     vf_stats->rx_packets = adapter->vfinfo[vf].vfstats.gprc;
9105     vf_stats->rx_bytes   = adapter->vfinfo[vf].vfstats.gorc;
9106     vf_stats->tx_packets = adapter->vfinfo[vf].vfstats.gptc;
9107     vf_stats->tx_bytes   = adapter->vfinfo[vf].vfstats.gotc;
9108     vf_stats->multicast  = adapter->vfinfo[vf].vfstats.mprc;
9109 
9110     return 0;
9111 }
9112 
9113 #ifdef CONFIG_IXGBE_DCB
9114 /**
9115  * ixgbe_validate_rtr - verify 802.1Qp to Rx packet buffer mapping is valid.
9116  * @adapter: pointer to ixgbe_adapter
9117  * @tc: number of traffic classes currently enabled
9118  *
9119  * Configure a valid 802.1Qp to Rx packet buffer mapping ie confirm
9120  * 802.1Q priority maps to a packet buffer that exists.
9121  */
9122 static void ixgbe_validate_rtr(struct ixgbe_adapter *adapter, u8 tc)
9123 {
9124     struct ixgbe_hw *hw = &adapter->hw;
9125     u32 reg, rsave;
9126     int i;
9127 
9128     /* 82598 have a static priority to TC mapping that can not
9129      * be changed so no validation is needed.
9130      */
9131     if (hw->mac.type == ixgbe_mac_82598EB)
9132         return;
9133 
9134     reg = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC);
9135     rsave = reg;
9136 
9137     for (i = 0; i < MAX_TRAFFIC_CLASS; i++) {
9138         u8 up2tc = reg >> (i * IXGBE_RTRUP2TC_UP_SHIFT);
9139 
9140         /* If up2tc is out of bounds default to zero */
9141         if (up2tc > tc)
9142             reg &= ~(0x7 << IXGBE_RTRUP2TC_UP_SHIFT);
9143     }
9144 
9145     if (reg != rsave)
9146         IXGBE_WRITE_REG(hw, IXGBE_RTRUP2TC, reg);
9147 
9148     return;
9149 }
9150 
9151 /**
9152  * ixgbe_set_prio_tc_map - Configure netdev prio tc map
9153  * @adapter: Pointer to adapter struct
9154  *
9155  * Populate the netdev user priority to tc map
9156  */
9157 static void ixgbe_set_prio_tc_map(struct ixgbe_adapter *adapter)
9158 {
9159     struct net_device *dev = adapter->netdev;
9160     struct ixgbe_dcb_config *dcb_cfg = &adapter->dcb_cfg;
9161     struct ieee_ets *ets = adapter->ixgbe_ieee_ets;
9162     u8 prio;
9163 
9164     for (prio = 0; prio < MAX_USER_PRIORITY; prio++) {
9165         u8 tc = 0;
9166 
9167         if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_CEE)
9168             tc = ixgbe_dcb_get_tc_from_up(dcb_cfg, 0, prio);
9169         else if (ets)
9170             tc = ets->prio_tc[prio];
9171 
9172         netdev_set_prio_tc_map(dev, prio, tc);
9173     }
9174 }
9175 
9176 #endif /* CONFIG_IXGBE_DCB */
9177 static int ixgbe_reassign_macvlan_pool(struct net_device *vdev,
9178                        struct netdev_nested_priv *priv)
9179 {
9180     struct ixgbe_adapter *adapter = (struct ixgbe_adapter *)priv->data;
9181     struct ixgbe_fwd_adapter *accel;
9182     int pool;
9183 
9184     /* we only care about macvlans... */
9185     if (!netif_is_macvlan(vdev))
9186         return 0;
9187 
9188     /* that have hardware offload enabled... */
9189     accel = macvlan_accel_priv(vdev);
9190     if (!accel)
9191         return 0;
9192 
9193     /* If we can relocate to a different bit do so */
9194     pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
9195     if (pool < adapter->num_rx_pools) {
9196         set_bit(pool, adapter->fwd_bitmask);
9197         accel->pool = pool;
9198         return 0;
9199     }
9200 
9201     /* if we cannot find a free pool then disable the offload */
9202     netdev_err(vdev, "L2FW offload disabled due to lack of queue resources\n");
9203     macvlan_release_l2fw_offload(vdev);
9204 
9205     /* unbind the queues and drop the subordinate channel config */
9206     netdev_unbind_sb_channel(adapter->netdev, vdev);
9207     netdev_set_sb_channel(vdev, 0);
9208 
9209     kfree(accel);
9210 
9211     return 0;
9212 }
9213 
9214 static void ixgbe_defrag_macvlan_pools(struct net_device *dev)
9215 {
9216     struct ixgbe_adapter *adapter = netdev_priv(dev);
9217     struct netdev_nested_priv priv = {
9218         .data = (void *)adapter,
9219     };
9220 
9221     /* flush any stale bits out of the fwd bitmask */
9222     bitmap_clear(adapter->fwd_bitmask, 1, 63);
9223 
9224     /* walk through upper devices reassigning pools */
9225     netdev_walk_all_upper_dev_rcu(dev, ixgbe_reassign_macvlan_pool,
9226                       &priv);
9227 }
9228 
9229 /**
9230  * ixgbe_setup_tc - configure net_device for multiple traffic classes
9231  *
9232  * @dev: net device to configure
9233  * @tc: number of traffic classes to enable
9234  */
9235 int ixgbe_setup_tc(struct net_device *dev, u8 tc)
9236 {
9237     struct ixgbe_adapter *adapter = netdev_priv(dev);
9238     struct ixgbe_hw *hw = &adapter->hw;
9239 
9240     /* Hardware supports up to 8 traffic classes */
9241     if (tc > adapter->dcb_cfg.num_tcs.pg_tcs)
9242         return -EINVAL;
9243 
9244     if (hw->mac.type == ixgbe_mac_82598EB && tc && tc < MAX_TRAFFIC_CLASS)
9245         return -EINVAL;
9246 
9247     /* Hardware has to reinitialize queues and interrupts to
9248      * match packet buffer alignment. Unfortunately, the
9249      * hardware is not flexible enough to do this dynamically.
9250      */
9251     if (netif_running(dev))
9252         ixgbe_close(dev);
9253     else
9254         ixgbe_reset(adapter);
9255 
9256     ixgbe_clear_interrupt_scheme(adapter);
9257 
9258 #ifdef CONFIG_IXGBE_DCB
9259     if (tc) {
9260         if (adapter->xdp_prog) {
9261             e_warn(probe, "DCB is not supported with XDP\n");
9262 
9263             ixgbe_init_interrupt_scheme(adapter);
9264             if (netif_running(dev))
9265                 ixgbe_open(dev);
9266             return -EINVAL;
9267         }
9268 
9269         netdev_set_num_tc(dev, tc);
9270         ixgbe_set_prio_tc_map(adapter);
9271 
9272         adapter->hw_tcs = tc;
9273         adapter->flags |= IXGBE_FLAG_DCB_ENABLED;
9274 
9275         if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
9276             adapter->last_lfc_mode = adapter->hw.fc.requested_mode;
9277             adapter->hw.fc.requested_mode = ixgbe_fc_none;
9278         }
9279     } else {
9280         netdev_reset_tc(dev);
9281 
9282         if (adapter->hw.mac.type == ixgbe_mac_82598EB)
9283             adapter->hw.fc.requested_mode = adapter->last_lfc_mode;
9284 
9285         adapter->flags &= ~IXGBE_FLAG_DCB_ENABLED;
9286         adapter->hw_tcs = tc;
9287 
9288         adapter->temp_dcb_cfg.pfc_mode_enable = false;
9289         adapter->dcb_cfg.pfc_mode_enable = false;
9290     }
9291 
9292     ixgbe_validate_rtr(adapter, tc);
9293 
9294 #endif /* CONFIG_IXGBE_DCB */
9295     ixgbe_init_interrupt_scheme(adapter);
9296 
9297     ixgbe_defrag_macvlan_pools(dev);
9298 
9299     if (netif_running(dev))
9300         return ixgbe_open(dev);
9301 
9302     return 0;
9303 }
9304 
9305 static int ixgbe_delete_clsu32(struct ixgbe_adapter *adapter,
9306                    struct tc_cls_u32_offload *cls)
9307 {
9308     u32 hdl = cls->knode.handle;
9309     u32 uhtid = TC_U32_USERHTID(cls->knode.handle);
9310     u32 loc = cls->knode.handle & 0xfffff;
9311     int err = 0, i, j;
9312     struct ixgbe_jump_table *jump = NULL;
9313 
9314     if (loc > IXGBE_MAX_HW_ENTRIES)
9315         return -EINVAL;
9316 
9317     if ((uhtid != 0x800) && (uhtid >= IXGBE_MAX_LINK_HANDLE))
9318         return -EINVAL;
9319 
9320     /* Clear this filter in the link data it is associated with */
9321     if (uhtid != 0x800) {
9322         jump = adapter->jump_tables[uhtid];
9323         if (!jump)
9324             return -EINVAL;
9325         if (!test_bit(loc - 1, jump->child_loc_map))
9326             return -EINVAL;
9327         clear_bit(loc - 1, jump->child_loc_map);
9328     }
9329 
9330     /* Check if the filter being deleted is a link */
9331     for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9332         jump = adapter->jump_tables[i];
9333         if (jump && jump->link_hdl == hdl) {
9334             /* Delete filters in the hardware in the child hash
9335              * table associated with this link
9336              */
9337             for (j = 0; j < IXGBE_MAX_HW_ENTRIES; j++) {
9338                 if (!test_bit(j, jump->child_loc_map))
9339                     continue;
9340                 spin_lock(&adapter->fdir_perfect_lock);
9341                 err = ixgbe_update_ethtool_fdir_entry(adapter,
9342                                       NULL,
9343                                       j + 1);
9344                 spin_unlock(&adapter->fdir_perfect_lock);
9345                 clear_bit(j, jump->child_loc_map);
9346             }
9347             /* Remove resources for this link */
9348             kfree(jump->input);
9349             kfree(jump->mask);
9350             kfree(jump);
9351             adapter->jump_tables[i] = NULL;
9352             return err;
9353         }
9354     }
9355 
9356     spin_lock(&adapter->fdir_perfect_lock);
9357     err = ixgbe_update_ethtool_fdir_entry(adapter, NULL, loc);
9358     spin_unlock(&adapter->fdir_perfect_lock);
9359     return err;
9360 }
9361 
9362 static int ixgbe_configure_clsu32_add_hnode(struct ixgbe_adapter *adapter,
9363                         struct tc_cls_u32_offload *cls)
9364 {
9365     u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
9366 
9367     if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9368         return -EINVAL;
9369 
9370     /* This ixgbe devices do not support hash tables at the moment
9371      * so abort when given hash tables.
9372      */
9373     if (cls->hnode.divisor > 0)
9374         return -EINVAL;
9375 
9376     set_bit(uhtid - 1, &adapter->tables);
9377     return 0;
9378 }
9379 
9380 static int ixgbe_configure_clsu32_del_hnode(struct ixgbe_adapter *adapter,
9381                         struct tc_cls_u32_offload *cls)
9382 {
9383     u32 uhtid = TC_U32_USERHTID(cls->hnode.handle);
9384 
9385     if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9386         return -EINVAL;
9387 
9388     clear_bit(uhtid - 1, &adapter->tables);
9389     return 0;
9390 }
9391 
9392 #ifdef CONFIG_NET_CLS_ACT
9393 struct upper_walk_data {
9394     struct ixgbe_adapter *adapter;
9395     u64 action;
9396     int ifindex;
9397     u8 queue;
9398 };
9399 
9400 static int get_macvlan_queue(struct net_device *upper,
9401                  struct netdev_nested_priv *priv)
9402 {
9403     if (netif_is_macvlan(upper)) {
9404         struct ixgbe_fwd_adapter *vadapter = macvlan_accel_priv(upper);
9405         struct ixgbe_adapter *adapter;
9406         struct upper_walk_data *data;
9407         int ifindex;
9408 
9409         data = (struct upper_walk_data *)priv->data;
9410         ifindex = data->ifindex;
9411         adapter = data->adapter;
9412         if (vadapter && upper->ifindex == ifindex) {
9413             data->queue = adapter->rx_ring[vadapter->rx_base_queue]->reg_idx;
9414             data->action = data->queue;
9415             return 1;
9416         }
9417     }
9418 
9419     return 0;
9420 }
9421 
9422 static int handle_redirect_action(struct ixgbe_adapter *adapter, int ifindex,
9423                   u8 *queue, u64 *action)
9424 {
9425     struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
9426     unsigned int num_vfs = adapter->num_vfs, vf;
9427     struct netdev_nested_priv priv;
9428     struct upper_walk_data data;
9429     struct net_device *upper;
9430 
9431     /* redirect to a SRIOV VF */
9432     for (vf = 0; vf < num_vfs; ++vf) {
9433         upper = pci_get_drvdata(adapter->vfinfo[vf].vfdev);
9434         if (upper->ifindex == ifindex) {
9435             *queue = vf * __ALIGN_MASK(1, ~vmdq->mask);
9436             *action = vf + 1;
9437             *action <<= ETHTOOL_RX_FLOW_SPEC_RING_VF_OFF;
9438             return 0;
9439         }
9440     }
9441 
9442     /* redirect to a offloaded macvlan netdev */
9443     data.adapter = adapter;
9444     data.ifindex = ifindex;
9445     data.action = 0;
9446     data.queue = 0;
9447     priv.data = (void *)&data;
9448     if (netdev_walk_all_upper_dev_rcu(adapter->netdev,
9449                       get_macvlan_queue, &priv)) {
9450         *action = data.action;
9451         *queue = data.queue;
9452 
9453         return 0;
9454     }
9455 
9456     return -EINVAL;
9457 }
9458 
9459 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9460                 struct tcf_exts *exts, u64 *action, u8 *queue)
9461 {
9462     const struct tc_action *a;
9463     int i;
9464 
9465     if (!tcf_exts_has_actions(exts))
9466         return -EINVAL;
9467 
9468     tcf_exts_for_each_action(i, a, exts) {
9469         /* Drop action */
9470         if (is_tcf_gact_shot(a)) {
9471             *action = IXGBE_FDIR_DROP_QUEUE;
9472             *queue = IXGBE_FDIR_DROP_QUEUE;
9473             return 0;
9474         }
9475 
9476         /* Redirect to a VF or a offloaded macvlan */
9477         if (is_tcf_mirred_egress_redirect(a)) {
9478             struct net_device *dev = tcf_mirred_dev(a);
9479 
9480             if (!dev)
9481                 return -EINVAL;
9482             return handle_redirect_action(adapter, dev->ifindex,
9483                               queue, action);
9484         }
9485 
9486         return -EINVAL;
9487     }
9488 
9489     return -EINVAL;
9490 }
9491 #else
9492 static int parse_tc_actions(struct ixgbe_adapter *adapter,
9493                 struct tcf_exts *exts, u64 *action, u8 *queue)
9494 {
9495     return -EINVAL;
9496 }
9497 #endif /* CONFIG_NET_CLS_ACT */
9498 
9499 static int ixgbe_clsu32_build_input(struct ixgbe_fdir_filter *input,
9500                     union ixgbe_atr_input *mask,
9501                     struct tc_cls_u32_offload *cls,
9502                     struct ixgbe_mat_field *field_ptr,
9503                     struct ixgbe_nexthdr *nexthdr)
9504 {
9505     int i, j, off;
9506     __be32 val, m;
9507     bool found_entry = false, found_jump_field = false;
9508 
9509     for (i = 0; i < cls->knode.sel->nkeys; i++) {
9510         off = cls->knode.sel->keys[i].off;
9511         val = cls->knode.sel->keys[i].val;
9512         m = cls->knode.sel->keys[i].mask;
9513 
9514         for (j = 0; field_ptr[j].val; j++) {
9515             if (field_ptr[j].off == off) {
9516                 field_ptr[j].val(input, mask, (__force u32)val,
9517                          (__force u32)m);
9518                 input->filter.formatted.flow_type |=
9519                     field_ptr[j].type;
9520                 found_entry = true;
9521                 break;
9522             }
9523         }
9524         if (nexthdr) {
9525             if (nexthdr->off == cls->knode.sel->keys[i].off &&
9526                 nexthdr->val ==
9527                 (__force u32)cls->knode.sel->keys[i].val &&
9528                 nexthdr->mask ==
9529                 (__force u32)cls->knode.sel->keys[i].mask)
9530                 found_jump_field = true;
9531             else
9532                 continue;
9533         }
9534     }
9535 
9536     if (nexthdr && !found_jump_field)
9537         return -EINVAL;
9538 
9539     if (!found_entry)
9540         return 0;
9541 
9542     mask->formatted.flow_type = IXGBE_ATR_L4TYPE_IPV6_MASK |
9543                     IXGBE_ATR_L4TYPE_MASK;
9544 
9545     if (input->filter.formatted.flow_type == IXGBE_ATR_FLOW_TYPE_IPV4)
9546         mask->formatted.flow_type &= IXGBE_ATR_L4TYPE_IPV6_MASK;
9547 
9548     return 0;
9549 }
9550 
9551 static int ixgbe_configure_clsu32(struct ixgbe_adapter *adapter,
9552                   struct tc_cls_u32_offload *cls)
9553 {
9554     __be16 protocol = cls->common.protocol;
9555     u32 loc = cls->knode.handle & 0xfffff;
9556     struct ixgbe_hw *hw = &adapter->hw;
9557     struct ixgbe_mat_field *field_ptr;
9558     struct ixgbe_fdir_filter *input = NULL;
9559     union ixgbe_atr_input *mask = NULL;
9560     struct ixgbe_jump_table *jump = NULL;
9561     int i, err = -EINVAL;
9562     u8 queue;
9563     u32 uhtid, link_uhtid;
9564 
9565     uhtid = TC_U32_USERHTID(cls->knode.handle);
9566     link_uhtid = TC_U32_USERHTID(cls->knode.link_handle);
9567 
9568     /* At the moment cls_u32 jumps to network layer and skips past
9569      * L2 headers. The canonical method to match L2 frames is to use
9570      * negative values. However this is error prone at best but really
9571      * just broken because there is no way to "know" what sort of hdr
9572      * is in front of the network layer. Fix cls_u32 to support L2
9573      * headers when needed.
9574      */
9575     if (protocol != htons(ETH_P_IP))
9576         return err;
9577 
9578     if (loc >= ((1024 << adapter->fdir_pballoc) - 2)) {
9579         e_err(drv, "Location out of range\n");
9580         return err;
9581     }
9582 
9583     /* cls u32 is a graph starting at root node 0x800. The driver tracks
9584      * links and also the fields used to advance the parser across each
9585      * link (e.g. nexthdr/eat parameters from 'tc'). This way we can map
9586      * the u32 graph onto the hardware parse graph denoted in ixgbe_model.h
9587      * To add support for new nodes update ixgbe_model.h parse structures
9588      * this function _should_ be generic try not to hardcode values here.
9589      */
9590     if (uhtid == 0x800) {
9591         field_ptr = (adapter->jump_tables[0])->mat;
9592     } else {
9593         if (uhtid >= IXGBE_MAX_LINK_HANDLE)
9594             return err;
9595         if (!adapter->jump_tables[uhtid])
9596             return err;
9597         field_ptr = (adapter->jump_tables[uhtid])->mat;
9598     }
9599 
9600     if (!field_ptr)
9601         return err;
9602 
9603     /* At this point we know the field_ptr is valid and need to either
9604      * build cls_u32 link or attach filter. Because adding a link to
9605      * a handle that does not exist is invalid and the same for adding
9606      * rules to handles that don't exist.
9607      */
9608 
9609     if (link_uhtid) {
9610         struct ixgbe_nexthdr *nexthdr = ixgbe_ipv4_jumps;
9611 
9612         if (link_uhtid >= IXGBE_MAX_LINK_HANDLE)
9613             return err;
9614 
9615         if (!test_bit(link_uhtid - 1, &adapter->tables))
9616             return err;
9617 
9618         /* Multiple filters as links to the same hash table are not
9619          * supported. To add a new filter with the same next header
9620          * but different match/jump conditions, create a new hash table
9621          * and link to it.
9622          */
9623         if (adapter->jump_tables[link_uhtid] &&
9624             (adapter->jump_tables[link_uhtid])->link_hdl) {
9625             e_err(drv, "Link filter exists for link: %x\n",
9626                   link_uhtid);
9627             return err;
9628         }
9629 
9630         for (i = 0; nexthdr[i].jump; i++) {
9631             if (nexthdr[i].o != cls->knode.sel->offoff ||
9632                 nexthdr[i].s != cls->knode.sel->offshift ||
9633                 nexthdr[i].m !=
9634                 (__force u32)cls->knode.sel->offmask)
9635                 return err;
9636 
9637             jump = kzalloc(sizeof(*jump), GFP_KERNEL);
9638             if (!jump)
9639                 return -ENOMEM;
9640             input = kzalloc(sizeof(*input), GFP_KERNEL);
9641             if (!input) {
9642                 err = -ENOMEM;
9643                 goto free_jump;
9644             }
9645             mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9646             if (!mask) {
9647                 err = -ENOMEM;
9648                 goto free_input;
9649             }
9650             jump->input = input;
9651             jump->mask = mask;
9652             jump->link_hdl = cls->knode.handle;
9653 
9654             err = ixgbe_clsu32_build_input(input, mask, cls,
9655                                field_ptr, &nexthdr[i]);
9656             if (!err) {
9657                 jump->mat = nexthdr[i].jump;
9658                 adapter->jump_tables[link_uhtid] = jump;
9659                 break;
9660             } else {
9661                 kfree(mask);
9662                 kfree(input);
9663                 kfree(jump);
9664             }
9665         }
9666         return 0;
9667     }
9668 
9669     input = kzalloc(sizeof(*input), GFP_KERNEL);
9670     if (!input)
9671         return -ENOMEM;
9672     mask = kzalloc(sizeof(*mask), GFP_KERNEL);
9673     if (!mask) {
9674         err = -ENOMEM;
9675         goto free_input;
9676     }
9677 
9678     if ((uhtid != 0x800) && (adapter->jump_tables[uhtid])) {
9679         if ((adapter->jump_tables[uhtid])->input)
9680             memcpy(input, (adapter->jump_tables[uhtid])->input,
9681                    sizeof(*input));
9682         if ((adapter->jump_tables[uhtid])->mask)
9683             memcpy(mask, (adapter->jump_tables[uhtid])->mask,
9684                    sizeof(*mask));
9685 
9686         /* Lookup in all child hash tables if this location is already
9687          * filled with a filter
9688          */
9689         for (i = 1; i < IXGBE_MAX_LINK_HANDLE; i++) {
9690             struct ixgbe_jump_table *link = adapter->jump_tables[i];
9691 
9692             if (link && (test_bit(loc - 1, link->child_loc_map))) {
9693                 e_err(drv, "Filter exists in location: %x\n",
9694                       loc);
9695                 err = -EINVAL;
9696                 goto err_out;
9697             }
9698         }
9699     }
9700     err = ixgbe_clsu32_build_input(input, mask, cls, field_ptr, NULL);
9701     if (err)
9702         goto err_out;
9703 
9704     err = parse_tc_actions(adapter, cls->knode.exts, &input->action,
9705                    &queue);
9706     if (err < 0)
9707         goto err_out;
9708 
9709     input->sw_idx = loc;
9710 
9711     spin_lock(&adapter->fdir_perfect_lock);
9712 
9713     if (hlist_empty(&adapter->fdir_filter_list)) {
9714         memcpy(&adapter->fdir_mask, mask, sizeof(*mask));
9715         err = ixgbe_fdir_set_input_mask_82599(hw, mask);
9716         if (err)
9717             goto err_out_w_lock;
9718     } else if (memcmp(&adapter->fdir_mask, mask, sizeof(*mask))) {
9719         err = -EINVAL;
9720         goto err_out_w_lock;
9721     }
9722 
9723     ixgbe_atr_compute_perfect_hash_82599(&input->filter, mask);
9724     err = ixgbe_fdir_write_perfect_filter_82599(hw, &input->filter,
9725                             input->sw_idx, queue);
9726     if (err)
9727         goto err_out_w_lock;
9728 
9729     ixgbe_update_ethtool_fdir_entry(adapter, input, input->sw_idx);
9730     spin_unlock(&adapter->fdir_perfect_lock);
9731 
9732     if ((uhtid != 0x800) && (adapter->jump_tables[uhtid]))
9733         set_bit(loc - 1, (adapter->jump_tables[uhtid])->child_loc_map);
9734 
9735     kfree(mask);
9736     return err;
9737 err_out_w_lock:
9738     spin_unlock(&adapter->fdir_perfect_lock);
9739 err_out:
9740     kfree(mask);
9741 free_input:
9742     kfree(input);
9743 free_jump:
9744     kfree(jump);
9745     return err;
9746 }
9747 
9748 static int ixgbe_setup_tc_cls_u32(struct ixgbe_adapter *adapter,
9749                   struct tc_cls_u32_offload *cls_u32)
9750 {
9751     switch (cls_u32->command) {
9752     case TC_CLSU32_NEW_KNODE:
9753     case TC_CLSU32_REPLACE_KNODE:
9754         return ixgbe_configure_clsu32(adapter, cls_u32);
9755     case TC_CLSU32_DELETE_KNODE:
9756         return ixgbe_delete_clsu32(adapter, cls_u32);
9757     case TC_CLSU32_NEW_HNODE:
9758     case TC_CLSU32_REPLACE_HNODE:
9759         return ixgbe_configure_clsu32_add_hnode(adapter, cls_u32);
9760     case TC_CLSU32_DELETE_HNODE:
9761         return ixgbe_configure_clsu32_del_hnode(adapter, cls_u32);
9762     default:
9763         return -EOPNOTSUPP;
9764     }
9765 }
9766 
9767 static int ixgbe_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
9768                    void *cb_priv)
9769 {
9770     struct ixgbe_adapter *adapter = cb_priv;
9771 
9772     if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
9773         return -EOPNOTSUPP;
9774 
9775     switch (type) {
9776     case TC_SETUP_CLSU32:
9777         return ixgbe_setup_tc_cls_u32(adapter, type_data);
9778     default:
9779         return -EOPNOTSUPP;
9780     }
9781 }
9782 
9783 static int ixgbe_setup_tc_mqprio(struct net_device *dev,
9784                  struct tc_mqprio_qopt *mqprio)
9785 {
9786     mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
9787     return ixgbe_setup_tc(dev, mqprio->num_tc);
9788 }
9789 
9790 static LIST_HEAD(ixgbe_block_cb_list);
9791 
9792 static int __ixgbe_setup_tc(struct net_device *dev, enum tc_setup_type type,
9793                 void *type_data)
9794 {
9795     struct ixgbe_adapter *adapter = netdev_priv(dev);
9796 
9797     switch (type) {
9798     case TC_SETUP_BLOCK:
9799         return flow_block_cb_setup_simple(type_data,
9800                           &ixgbe_block_cb_list,
9801                           ixgbe_setup_tc_block_cb,
9802                           adapter, adapter, true);
9803     case TC_SETUP_QDISC_MQPRIO:
9804         return ixgbe_setup_tc_mqprio(dev, type_data);
9805     default:
9806         return -EOPNOTSUPP;
9807     }
9808 }
9809 
9810 #ifdef CONFIG_PCI_IOV
9811 void ixgbe_sriov_reinit(struct ixgbe_adapter *adapter)
9812 {
9813     struct net_device *netdev = adapter->netdev;
9814 
9815     rtnl_lock();
9816     ixgbe_setup_tc(netdev, adapter->hw_tcs);
9817     rtnl_unlock();
9818 }
9819 
9820 #endif
9821 void ixgbe_do_reset(struct net_device *netdev)
9822 {
9823     struct ixgbe_adapter *adapter = netdev_priv(netdev);
9824 
9825     if (netif_running(netdev))
9826         ixgbe_reinit_locked(adapter);
9827     else
9828         ixgbe_reset(adapter);
9829 }
9830 
9831 static netdev_features_t ixgbe_fix_features(struct net_device *netdev,
9832                         netdev_features_t features)
9833 {
9834     struct ixgbe_adapter *adapter = netdev_priv(netdev);
9835 
9836     /* If Rx checksum is disabled, then RSC/LRO should also be disabled */
9837     if (!(features & NETIF_F_RXCSUM))
9838         features &= ~NETIF_F_LRO;
9839 
9840     /* Turn off LRO if not RSC capable */
9841     if (!(adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE))
9842         features &= ~NETIF_F_LRO;
9843 
9844     if (adapter->xdp_prog && (features & NETIF_F_LRO)) {
9845         e_dev_err("LRO is not supported with XDP\n");
9846         features &= ~NETIF_F_LRO;
9847     }
9848 
9849     return features;
9850 }
9851 
9852 static void ixgbe_reset_l2fw_offload(struct ixgbe_adapter *adapter)
9853 {
9854     int rss = min_t(int, ixgbe_max_rss_indices(adapter),
9855             num_online_cpus());
9856 
9857     /* go back to full RSS if we're not running SR-IOV */
9858     if (!adapter->ring_feature[RING_F_VMDQ].offset)
9859         adapter->flags &= ~(IXGBE_FLAG_VMDQ_ENABLED |
9860                     IXGBE_FLAG_SRIOV_ENABLED);
9861 
9862     adapter->ring_feature[RING_F_RSS].limit = rss;
9863     adapter->ring_feature[RING_F_VMDQ].limit = 1;
9864 
9865     ixgbe_setup_tc(adapter->netdev, adapter->hw_tcs);
9866 }
9867 
9868 static int ixgbe_set_features(struct net_device *netdev,
9869                   netdev_features_t features)
9870 {
9871     struct ixgbe_adapter *adapter = netdev_priv(netdev);
9872     netdev_features_t changed = netdev->features ^ features;
9873     bool need_reset = false;
9874 
9875     /* Make sure RSC matches LRO, reset if change */
9876     if (!(features & NETIF_F_LRO)) {
9877         if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
9878             need_reset = true;
9879         adapter->flags2 &= ~IXGBE_FLAG2_RSC_ENABLED;
9880     } else if ((adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE) &&
9881            !(adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)) {
9882         if (adapter->rx_itr_setting == 1 ||
9883             adapter->rx_itr_setting > IXGBE_MIN_RSC_ITR) {
9884             adapter->flags2 |= IXGBE_FLAG2_RSC_ENABLED;
9885             need_reset = true;
9886         } else if ((changed ^ features) & NETIF_F_LRO) {
9887             e_info(probe, "rx-usecs set too low, "
9888                    "disabling RSC\n");
9889         }
9890     }
9891 
9892     /*
9893      * Check if Flow Director n-tuple support or hw_tc support was
9894      * enabled or disabled.  If the state changed, we need to reset.
9895      */
9896     if ((features & NETIF_F_NTUPLE) || (features & NETIF_F_HW_TC)) {
9897         /* turn off ATR, enable perfect filters and reset */
9898         if (!(adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE))
9899             need_reset = true;
9900 
9901         adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
9902         adapter->flags |= IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9903     } else {
9904         /* turn off perfect filters, enable ATR and reset */
9905         if (adapter->flags & IXGBE_FLAG_FDIR_PERFECT_CAPABLE)
9906             need_reset = true;
9907 
9908         adapter->flags &= ~IXGBE_FLAG_FDIR_PERFECT_CAPABLE;
9909 
9910         /* We cannot enable ATR if SR-IOV is enabled */
9911         if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED ||
9912             /* We cannot enable ATR if we have 2 or more tcs */
9913             (adapter->hw_tcs > 1) ||
9914             /* We cannot enable ATR if RSS is disabled */
9915             (adapter->ring_feature[RING_F_RSS].limit <= 1) ||
9916             /* A sample rate of 0 indicates ATR disabled */
9917             (!adapter->atr_sample_rate))
9918             ; /* do nothing not supported */
9919         else /* otherwise supported and set the flag */
9920             adapter->flags |= IXGBE_FLAG_FDIR_HASH_CAPABLE;
9921     }
9922 
9923     if (changed & NETIF_F_RXALL)
9924         need_reset = true;
9925 
9926     netdev->features = features;
9927 
9928     if ((changed & NETIF_F_HW_L2FW_DOFFLOAD) && adapter->num_rx_pools > 1)
9929         ixgbe_reset_l2fw_offload(adapter);
9930     else if (need_reset)
9931         ixgbe_do_reset(netdev);
9932     else if (changed & (NETIF_F_HW_VLAN_CTAG_RX |
9933                 NETIF_F_HW_VLAN_CTAG_FILTER))
9934         ixgbe_set_rx_mode(netdev);
9935 
9936     return 1;
9937 }
9938 
9939 static int ixgbe_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9940                  struct net_device *dev,
9941                  const unsigned char *addr, u16 vid,
9942                  u16 flags,
9943                  struct netlink_ext_ack *extack)
9944 {
9945     /* guarantee we can provide a unique filter for the unicast address */
9946     if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
9947         struct ixgbe_adapter *adapter = netdev_priv(dev);
9948         u16 pool = VMDQ_P(0);
9949 
9950         if (netdev_uc_count(dev) >= ixgbe_available_rars(adapter, pool))
9951             return -ENOMEM;
9952     }
9953 
9954     return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
9955 }
9956 
9957 /**
9958  * ixgbe_configure_bridge_mode - set various bridge modes
9959  * @adapter: the private structure
9960  * @mode: requested bridge mode
9961  *
9962  * Configure some settings require for various bridge modes.
9963  **/
9964 static int ixgbe_configure_bridge_mode(struct ixgbe_adapter *adapter,
9965                        __u16 mode)
9966 {
9967     struct ixgbe_hw *hw = &adapter->hw;
9968     unsigned int p, num_pools;
9969     u32 vmdctl;
9970 
9971     switch (mode) {
9972     case BRIDGE_MODE_VEPA:
9973         /* disable Tx loopback, rely on switch hairpin mode */
9974         IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC, 0);
9975 
9976         /* must enable Rx switching replication to allow multicast
9977          * packet reception on all VFs, and to enable source address
9978          * pruning.
9979          */
9980         vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
9981         vmdctl |= IXGBE_VT_CTL_REPLEN;
9982         IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
9983 
9984         /* enable Rx source address pruning. Note, this requires
9985          * replication to be enabled or else it does nothing.
9986          */
9987         num_pools = adapter->num_vfs + adapter->num_rx_pools;
9988         for (p = 0; p < num_pools; p++) {
9989             if (hw->mac.ops.set_source_address_pruning)
9990                 hw->mac.ops.set_source_address_pruning(hw,
9991                                        true,
9992                                        p);
9993         }
9994         break;
9995     case BRIDGE_MODE_VEB:
9996         /* enable Tx loopback for internal VF/PF communication */
9997         IXGBE_WRITE_REG(&adapter->hw, IXGBE_PFDTXGSWC,
9998                 IXGBE_PFDTXGSWC_VT_LBEN);
9999 
10000         /* disable Rx switching replication unless we have SR-IOV
10001          * virtual functions
10002          */
10003         vmdctl = IXGBE_READ_REG(hw, IXGBE_VMD_CTL);
10004         if (!adapter->num_vfs)
10005             vmdctl &= ~IXGBE_VT_CTL_REPLEN;
10006         IXGBE_WRITE_REG(hw, IXGBE_VMD_CTL, vmdctl);
10007 
10008         /* disable Rx source address pruning, since we don't expect to
10009          * be receiving external loopback of our transmitted frames.
10010          */
10011         num_pools = adapter->num_vfs + adapter->num_rx_pools;
10012         for (p = 0; p < num_pools; p++) {
10013             if (hw->mac.ops.set_source_address_pruning)
10014                 hw->mac.ops.set_source_address_pruning(hw,
10015                                        false,
10016                                        p);
10017         }
10018         break;
10019     default:
10020         return -EINVAL;
10021     }
10022 
10023     adapter->bridge_mode = mode;
10024 
10025     e_info(drv, "enabling bridge mode: %s\n",
10026            mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
10027 
10028     return 0;
10029 }
10030 
10031 static int ixgbe_ndo_bridge_setlink(struct net_device *dev,
10032                     struct nlmsghdr *nlh, u16 flags,
10033                     struct netlink_ext_ack *extack)
10034 {
10035     struct ixgbe_adapter *adapter = netdev_priv(dev);
10036     struct nlattr *attr, *br_spec;
10037     int rem;
10038 
10039     if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
10040         return -EOPNOTSUPP;
10041 
10042     br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
10043     if (!br_spec)
10044         return -EINVAL;
10045 
10046     nla_for_each_nested(attr, br_spec, rem) {
10047         int status;
10048         __u16 mode;
10049 
10050         if (nla_type(attr) != IFLA_BRIDGE_MODE)
10051             continue;
10052 
10053         if (nla_len(attr) < sizeof(mode))
10054             return -EINVAL;
10055 
10056         mode = nla_get_u16(attr);
10057         status = ixgbe_configure_bridge_mode(adapter, mode);
10058         if (status)
10059             return status;
10060 
10061         break;
10062     }
10063 
10064     return 0;
10065 }
10066 
10067 static int ixgbe_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
10068                     struct net_device *dev,
10069                     u32 filter_mask, int nlflags)
10070 {
10071     struct ixgbe_adapter *adapter = netdev_priv(dev);
10072 
10073     if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
10074         return 0;
10075 
10076     return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
10077                        adapter->bridge_mode, 0, 0, nlflags,
10078                        filter_mask, NULL);
10079 }
10080 
10081 static void *ixgbe_fwd_add(struct net_device *pdev, struct net_device *vdev)
10082 {
10083     struct ixgbe_adapter *adapter = netdev_priv(pdev);
10084     struct ixgbe_fwd_adapter *accel;
10085     int tcs = adapter->hw_tcs ? : 1;
10086     int pool, err;
10087 
10088     if (adapter->xdp_prog) {
10089         e_warn(probe, "L2FW offload is not supported with XDP\n");
10090         return ERR_PTR(-EINVAL);
10091     }
10092 
10093     /* The hardware supported by ixgbe only filters on the destination MAC
10094      * address. In order to avoid issues we only support offloading modes
10095      * where the hardware can actually provide the functionality.
10096      */
10097     if (!macvlan_supports_dest_filter(vdev))
10098         return ERR_PTR(-EMEDIUMTYPE);
10099 
10100     /* We need to lock down the macvlan to be a single queue device so that
10101      * we can reuse the tc_to_txq field in the macvlan netdev to represent
10102      * the queue mapping to our netdev.
10103      */
10104     if (netif_is_multiqueue(vdev))
10105         return ERR_PTR(-ERANGE);
10106 
10107     pool = find_first_zero_bit(adapter->fwd_bitmask, adapter->num_rx_pools);
10108     if (pool == adapter->num_rx_pools) {
10109         u16 used_pools = adapter->num_vfs + adapter->num_rx_pools;
10110         u16 reserved_pools;
10111 
10112         if (((adapter->flags & IXGBE_FLAG_DCB_ENABLED) &&
10113              adapter->num_rx_pools >= (MAX_TX_QUEUES / tcs)) ||
10114             adapter->num_rx_pools > IXGBE_MAX_MACVLANS)
10115             return ERR_PTR(-EBUSY);
10116 
10117         /* Hardware has a limited number of available pools. Each VF,
10118          * and the PF require a pool. Check to ensure we don't
10119          * attempt to use more then the available number of pools.
10120          */
10121         if (used_pools >= IXGBE_MAX_VF_FUNCTIONS)
10122             return ERR_PTR(-EBUSY);
10123 
10124         /* Enable VMDq flag so device will be set in VM mode */
10125         adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED |
10126                   IXGBE_FLAG_SRIOV_ENABLED;
10127 
10128         /* Try to reserve as many queues per pool as possible,
10129          * we start with the configurations that support 4 queues
10130          * per pools, followed by 2, and then by just 1 per pool.
10131          */
10132         if (used_pools < 32 && adapter->num_rx_pools < 16)
10133             reserved_pools = min_t(u16,
10134                            32 - used_pools,
10135                            16 - adapter->num_rx_pools);
10136         else if (adapter->num_rx_pools < 32)
10137             reserved_pools = min_t(u16,
10138                            64 - used_pools,
10139                            32 - adapter->num_rx_pools);
10140         else
10141             reserved_pools = 64 - used_pools;
10142 
10143 
10144         if (!reserved_pools)
10145             return ERR_PTR(-EBUSY);
10146 
10147         adapter->ring_feature[RING_F_VMDQ].limit += reserved_pools;
10148 
10149         /* Force reinit of ring allocation with VMDQ enabled */
10150         err = ixgbe_setup_tc(pdev, adapter->hw_tcs);
10151         if (err)
10152             return ERR_PTR(err);
10153 
10154         if (pool >= adapter->num_rx_pools)
10155             return ERR_PTR(-ENOMEM);
10156     }
10157 
10158     accel = kzalloc(sizeof(*accel), GFP_KERNEL);
10159     if (!accel)
10160         return ERR_PTR(-ENOMEM);
10161 
10162     set_bit(pool, adapter->fwd_bitmask);
10163     netdev_set_sb_channel(vdev, pool);
10164     accel->pool = pool;
10165     accel->netdev = vdev;
10166 
10167     if (!netif_running(pdev))
10168         return accel;
10169 
10170     err = ixgbe_fwd_ring_up(adapter, accel);
10171     if (err)
10172         return ERR_PTR(err);
10173 
10174     return accel;
10175 }
10176 
10177 static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
10178 {
10179     struct ixgbe_fwd_adapter *accel = priv;
10180     struct ixgbe_adapter *adapter = netdev_priv(pdev);
10181     unsigned int rxbase = accel->rx_base_queue;
10182     unsigned int i;
10183 
10184     /* delete unicast filter associated with offloaded interface */
10185     ixgbe_del_mac_filter(adapter, accel->netdev->dev_addr,
10186                  VMDQ_P(accel->pool));
10187 
10188     /* Allow remaining Rx packets to get flushed out of the
10189      * Rx FIFO before we drop the netdev for the ring.
10190      */
10191     usleep_range(10000, 20000);
10192 
10193     for (i = 0; i < adapter->num_rx_queues_per_pool; i++) {
10194         struct ixgbe_ring *ring = adapter->rx_ring[rxbase + i];
10195         struct ixgbe_q_vector *qv = ring->q_vector;
10196 
10197         /* Make sure we aren't processing any packets and clear
10198          * netdev to shut down the ring.
10199          */
10200         if (netif_running(adapter->netdev))
10201             napi_synchronize(&qv->napi);
10202         ring->netdev = NULL;
10203     }
10204 
10205     /* unbind the queues and drop the subordinate channel config */
10206     netdev_unbind_sb_channel(pdev, accel->netdev);
10207     netdev_set_sb_channel(accel->netdev, 0);
10208 
10209     clear_bit(accel->pool, adapter->fwd_bitmask);
10210     kfree(accel);
10211 }
10212 
10213 #define IXGBE_MAX_MAC_HDR_LEN       127
10214 #define IXGBE_MAX_NETWORK_HDR_LEN   511
10215 
10216 static netdev_features_t
10217 ixgbe_features_check(struct sk_buff *skb, struct net_device *dev,
10218              netdev_features_t features)
10219 {
10220     unsigned int network_hdr_len, mac_hdr_len;
10221 
10222     /* Make certain the headers can be described by a context descriptor */
10223     mac_hdr_len = skb_network_header(skb) - skb->data;
10224     if (unlikely(mac_hdr_len > IXGBE_MAX_MAC_HDR_LEN))
10225         return features & ~(NETIF_F_HW_CSUM |
10226                     NETIF_F_SCTP_CRC |
10227                     NETIF_F_GSO_UDP_L4 |
10228                     NETIF_F_HW_VLAN_CTAG_TX |
10229                     NETIF_F_TSO |
10230                     NETIF_F_TSO6);
10231 
10232     network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
10233     if (unlikely(network_hdr_len >  IXGBE_MAX_NETWORK_HDR_LEN))
10234         return features & ~(NETIF_F_HW_CSUM |
10235                     NETIF_F_SCTP_CRC |
10236                     NETIF_F_GSO_UDP_L4 |
10237                     NETIF_F_TSO |
10238                     NETIF_F_TSO6);
10239 
10240     /* We can only support IPV4 TSO in tunnels if we can mangle the
10241      * inner IP ID field, so strip TSO if MANGLEID is not supported.
10242      * IPsec offoad sets skb->encapsulation but still can handle
10243      * the TSO, so it's the exception.
10244      */
10245     if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID)) {
10246 #ifdef CONFIG_IXGBE_IPSEC
10247         if (!secpath_exists(skb))
10248 #endif
10249             features &= ~NETIF_F_TSO;
10250     }
10251 
10252     return features;
10253 }
10254 
10255 static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
10256 {
10257     int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
10258     struct ixgbe_adapter *adapter = netdev_priv(dev);
10259     struct bpf_prog *old_prog;
10260     bool need_reset;
10261     int num_queues;
10262 
10263     if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
10264         return -EINVAL;
10265 
10266     if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
10267         return -EINVAL;
10268 
10269     /* verify ixgbe ring attributes are sufficient for XDP */
10270     for (i = 0; i < adapter->num_rx_queues; i++) {
10271         struct ixgbe_ring *ring = adapter->rx_ring[i];
10272 
10273         if (ring_is_rsc_enabled(ring))
10274             return -EINVAL;
10275 
10276         if (frame_size > ixgbe_rx_bufsz(ring))
10277             return -EINVAL;
10278     }
10279 
10280     /* if the number of cpus is much larger than the maximum of queues,
10281      * we should stop it and then return with ENOMEM like before.
10282      */
10283     if (nr_cpu_ids > IXGBE_MAX_XDP_QS * 2)
10284         return -ENOMEM;
10285     else if (nr_cpu_ids > IXGBE_MAX_XDP_QS)
10286         static_branch_inc(&ixgbe_xdp_locking_key);
10287 
10288     old_prog = xchg(&adapter->xdp_prog, prog);
10289     need_reset = (!!prog != !!old_prog);
10290 
10291     /* If transitioning XDP modes reconfigure rings */
10292     if (need_reset) {
10293         int err;
10294 
10295         if (!prog)
10296             /* Wait until ndo_xsk_wakeup completes. */
10297             synchronize_rcu();
10298         err = ixgbe_setup_tc(dev, adapter->hw_tcs);
10299 
10300         if (err) {
10301             rcu_assign_pointer(adapter->xdp_prog, old_prog);
10302             return -EINVAL;
10303         }
10304     } else {
10305         for (i = 0; i < adapter->num_rx_queues; i++)
10306             (void)xchg(&adapter->rx_ring[i]->xdp_prog,
10307                 adapter->xdp_prog);
10308     }
10309 
10310     if (old_prog)
10311         bpf_prog_put(old_prog);
10312 
10313     /* Kick start the NAPI context if there is an AF_XDP socket open
10314      * on that queue id. This so that receiving will start.
10315      */
10316     if (need_reset && prog) {
10317         num_queues = min_t(int, adapter->num_rx_queues,
10318                    adapter->num_xdp_queues);
10319         for (i = 0; i < num_queues; i++)
10320             if (adapter->xdp_ring[i]->xsk_pool)
10321                 (void)ixgbe_xsk_wakeup(adapter->netdev, i,
10322                                XDP_WAKEUP_RX);
10323     }
10324 
10325     return 0;
10326 }
10327 
10328 static int ixgbe_xdp(struct net_device *dev, struct netdev_bpf *xdp)
10329 {
10330     struct ixgbe_adapter *adapter = netdev_priv(dev);
10331 
10332     switch (xdp->command) {
10333     case XDP_SETUP_PROG:
10334         return ixgbe_xdp_setup(dev, xdp->prog);
10335     case XDP_SETUP_XSK_POOL:
10336         return ixgbe_xsk_pool_setup(adapter, xdp->xsk.pool,
10337                         xdp->xsk.queue_id);
10338 
10339     default:
10340         return -EINVAL;
10341     }
10342 }
10343 
10344 void ixgbe_xdp_ring_update_tail(struct ixgbe_ring *ring)
10345 {
10346     /* Force memory writes to complete before letting h/w know there
10347      * are new descriptors to fetch.
10348      */
10349     wmb();
10350     writel(ring->next_to_use, ring->tail);
10351 }
10352 
10353 void ixgbe_xdp_ring_update_tail_locked(struct ixgbe_ring *ring)
10354 {
10355     if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10356         spin_lock(&ring->tx_lock);
10357     ixgbe_xdp_ring_update_tail(ring);
10358     if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10359         spin_unlock(&ring->tx_lock);
10360 }
10361 
10362 static int ixgbe_xdp_xmit(struct net_device *dev, int n,
10363               struct xdp_frame **frames, u32 flags)
10364 {
10365     struct ixgbe_adapter *adapter = netdev_priv(dev);
10366     struct ixgbe_ring *ring;
10367     int nxmit = 0;
10368     int i;
10369 
10370     if (unlikely(test_bit(__IXGBE_DOWN, &adapter->state)))
10371         return -ENETDOWN;
10372 
10373     if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
10374         return -EINVAL;
10375 
10376     /* During program transitions its possible adapter->xdp_prog is assigned
10377      * but ring has not been configured yet. In this case simply abort xmit.
10378      */
10379     ring = adapter->xdp_prog ? ixgbe_determine_xdp_ring(adapter) : NULL;
10380     if (unlikely(!ring))
10381         return -ENXIO;
10382 
10383     if (unlikely(test_bit(__IXGBE_TX_DISABLED, &ring->state)))
10384         return -ENXIO;
10385 
10386     if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10387         spin_lock(&ring->tx_lock);
10388 
10389     for (i = 0; i < n; i++) {
10390         struct xdp_frame *xdpf = frames[i];
10391         int err;
10392 
10393         err = ixgbe_xmit_xdp_ring(ring, xdpf);
10394         if (err != IXGBE_XDP_TX)
10395             break;
10396         nxmit++;
10397     }
10398 
10399     if (unlikely(flags & XDP_XMIT_FLUSH))
10400         ixgbe_xdp_ring_update_tail(ring);
10401 
10402     if (static_branch_unlikely(&ixgbe_xdp_locking_key))
10403         spin_unlock(&ring->tx_lock);
10404 
10405     return nxmit;
10406 }
10407 
10408 static const struct net_device_ops ixgbe_netdev_ops = {
10409     .ndo_open       = ixgbe_open,
10410     .ndo_stop       = ixgbe_close,
10411     .ndo_start_xmit     = ixgbe_xmit_frame,
10412     .ndo_set_rx_mode    = ixgbe_set_rx_mode,
10413     .ndo_validate_addr  = eth_validate_addr,
10414     .ndo_set_mac_address    = ixgbe_set_mac,
10415     .ndo_change_mtu     = ixgbe_change_mtu,
10416     .ndo_tx_timeout     = ixgbe_tx_timeout,
10417     .ndo_set_tx_maxrate = ixgbe_tx_maxrate,
10418     .ndo_vlan_rx_add_vid    = ixgbe_vlan_rx_add_vid,
10419     .ndo_vlan_rx_kill_vid   = ixgbe_vlan_rx_kill_vid,
10420     .ndo_eth_ioctl      = ixgbe_ioctl,
10421     .ndo_set_vf_mac     = ixgbe_ndo_set_vf_mac,
10422     .ndo_set_vf_vlan    = ixgbe_ndo_set_vf_vlan,
10423     .ndo_set_vf_rate    = ixgbe_ndo_set_vf_bw,
10424     .ndo_set_vf_spoofchk    = ixgbe_ndo_set_vf_spoofchk,
10425     .ndo_set_vf_link_state  = ixgbe_ndo_set_vf_link_state,
10426     .ndo_set_vf_rss_query_en = ixgbe_ndo_set_vf_rss_query_en,
10427     .ndo_set_vf_trust   = ixgbe_ndo_set_vf_trust,
10428     .ndo_get_vf_config  = ixgbe_ndo_get_vf_config,
10429     .ndo_get_vf_stats   = ixgbe_ndo_get_vf_stats,
10430     .ndo_get_stats64    = ixgbe_get_stats64,
10431     .ndo_setup_tc       = __ixgbe_setup_tc,
10432 #ifdef IXGBE_FCOE
10433     .ndo_select_queue   = ixgbe_select_queue,
10434     .ndo_fcoe_ddp_setup = ixgbe_fcoe_ddp_get,
10435     .ndo_fcoe_ddp_target = ixgbe_fcoe_ddp_target,
10436     .ndo_fcoe_ddp_done = ixgbe_fcoe_ddp_put,
10437     .ndo_fcoe_enable = ixgbe_fcoe_enable,
10438     .ndo_fcoe_disable = ixgbe_fcoe_disable,
10439     .ndo_fcoe_get_wwn = ixgbe_fcoe_get_wwn,
10440     .ndo_fcoe_get_hbainfo = ixgbe_fcoe_get_hbainfo,
10441 #endif /* IXGBE_FCOE */
10442     .ndo_set_features = ixgbe_set_features,
10443     .ndo_fix_features = ixgbe_fix_features,
10444     .ndo_fdb_add        = ixgbe_ndo_fdb_add,
10445     .ndo_bridge_setlink = ixgbe_ndo_bridge_setlink,
10446     .ndo_bridge_getlink = ixgbe_ndo_bridge_getlink,
10447     .ndo_dfwd_add_station   = ixgbe_fwd_add,
10448     .ndo_dfwd_del_station   = ixgbe_fwd_del,
10449     .ndo_features_check = ixgbe_features_check,
10450     .ndo_bpf        = ixgbe_xdp,
10451     .ndo_xdp_xmit       = ixgbe_xdp_xmit,
10452     .ndo_xsk_wakeup         = ixgbe_xsk_wakeup,
10453 };
10454 
10455 static void ixgbe_disable_txr_hw(struct ixgbe_adapter *adapter,
10456                  struct ixgbe_ring *tx_ring)
10457 {
10458     unsigned long wait_delay, delay_interval;
10459     struct ixgbe_hw *hw = &adapter->hw;
10460     u8 reg_idx = tx_ring->reg_idx;
10461     int wait_loop;
10462     u32 txdctl;
10463 
10464     IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
10465 
10466     /* delay mechanism from ixgbe_disable_tx */
10467     delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
10468 
10469     wait_loop = IXGBE_MAX_RX_DESC_POLL;
10470     wait_delay = delay_interval;
10471 
10472     while (wait_loop--) {
10473         usleep_range(wait_delay, wait_delay + 10);
10474         wait_delay += delay_interval * 2;
10475         txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(reg_idx));
10476 
10477         if (!(txdctl & IXGBE_TXDCTL_ENABLE))
10478             return;
10479     }
10480 
10481     e_err(drv, "TXDCTL.ENABLE not cleared within the polling period\n");
10482 }
10483 
10484 static void ixgbe_disable_txr(struct ixgbe_adapter *adapter,
10485                   struct ixgbe_ring *tx_ring)
10486 {
10487     set_bit(__IXGBE_TX_DISABLED, &tx_ring->state);
10488     ixgbe_disable_txr_hw(adapter, tx_ring);
10489 }
10490 
10491 static void ixgbe_disable_rxr_hw(struct ixgbe_adapter *adapter,
10492                  struct ixgbe_ring *rx_ring)
10493 {
10494     unsigned long wait_delay, delay_interval;
10495     struct ixgbe_hw *hw = &adapter->hw;
10496     u8 reg_idx = rx_ring->reg_idx;
10497     int wait_loop;
10498     u32 rxdctl;
10499 
10500     rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
10501     rxdctl &= ~IXGBE_RXDCTL_ENABLE;
10502     rxdctl |= IXGBE_RXDCTL_SWFLSH;
10503 
10504     /* write value back with RXDCTL.ENABLE bit cleared */
10505     IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(reg_idx), rxdctl);
10506 
10507     /* RXDCTL.EN may not change on 82598 if link is down, so skip it */
10508     if (hw->mac.type == ixgbe_mac_82598EB &&
10509         !(IXGBE_READ_REG(hw, IXGBE_LINKS) & IXGBE_LINKS_UP))
10510         return;
10511 
10512     /* delay mechanism from ixgbe_disable_rx */
10513     delay_interval = ixgbe_get_completion_timeout(adapter) / 100;
10514 
10515     wait_loop = IXGBE_MAX_RX_DESC_POLL;
10516     wait_delay = delay_interval;
10517 
10518     while (wait_loop--) {
10519         usleep_range(wait_delay, wait_delay + 10);
10520         wait_delay += delay_interval * 2;
10521         rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(reg_idx));
10522 
10523         if (!(rxdctl & IXGBE_RXDCTL_ENABLE))
10524             return;
10525     }
10526 
10527     e_err(drv, "RXDCTL.ENABLE not cleared within the polling period\n");
10528 }
10529 
10530 static void ixgbe_reset_txr_stats(struct ixgbe_ring *tx_ring)
10531 {
10532     memset(&tx_ring->stats, 0, sizeof(tx_ring->stats));
10533     memset(&tx_ring->tx_stats, 0, sizeof(tx_ring->tx_stats));
10534 }
10535 
10536 static void ixgbe_reset_rxr_stats(struct ixgbe_ring *rx_ring)
10537 {
10538     memset(&rx_ring->stats, 0, sizeof(rx_ring->stats));
10539     memset(&rx_ring->rx_stats, 0, sizeof(rx_ring->rx_stats));
10540 }
10541 
10542 /**
10543  * ixgbe_txrx_ring_disable - Disable Rx/Tx/XDP Tx rings
10544  * @adapter: adapter structure
10545  * @ring: ring index
10546  *
10547  * This function disables a certain Rx/Tx/XDP Tx ring. The function
10548  * assumes that the netdev is running.
10549  **/
10550 void ixgbe_txrx_ring_disable(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     ixgbe_disable_txr(adapter, tx_ring);
10559     if (xdp_ring)
10560         ixgbe_disable_txr(adapter, xdp_ring);
10561     ixgbe_disable_rxr_hw(adapter, rx_ring);
10562 
10563     if (xdp_ring)
10564         synchronize_rcu();
10565 
10566     /* Rx/Tx/XDP Tx share the same napi context. */
10567     napi_disable(&rx_ring->q_vector->napi);
10568 
10569     ixgbe_clean_tx_ring(tx_ring);
10570     if (xdp_ring)
10571         ixgbe_clean_tx_ring(xdp_ring);
10572     ixgbe_clean_rx_ring(rx_ring);
10573 
10574     ixgbe_reset_txr_stats(tx_ring);
10575     if (xdp_ring)
10576         ixgbe_reset_txr_stats(xdp_ring);
10577     ixgbe_reset_rxr_stats(rx_ring);
10578 }
10579 
10580 /**
10581  * ixgbe_txrx_ring_enable - Enable Rx/Tx/XDP Tx rings
10582  * @adapter: adapter structure
10583  * @ring: ring index
10584  *
10585  * This function enables a certain Rx/Tx/XDP Tx ring. The function
10586  * assumes that the netdev is running.
10587  **/
10588 void ixgbe_txrx_ring_enable(struct ixgbe_adapter *adapter, int ring)
10589 {
10590     struct ixgbe_ring *rx_ring, *tx_ring, *xdp_ring;
10591 
10592     rx_ring = adapter->rx_ring[ring];
10593     tx_ring = adapter->tx_ring[ring];
10594     xdp_ring = adapter->xdp_ring[ring];
10595 
10596     /* Rx/Tx/XDP Tx share the same napi context. */
10597     napi_enable(&rx_ring->q_vector->napi);
10598 
10599     ixgbe_configure_tx_ring(adapter, tx_ring);
10600     if (xdp_ring)
10601         ixgbe_configure_tx_ring(adapter, xdp_ring);
10602     ixgbe_configure_rx_ring(adapter, rx_ring);
10603 
10604     clear_bit(__IXGBE_TX_DISABLED, &tx_ring->state);
10605     if (xdp_ring)
10606         clear_bit(__IXGBE_TX_DISABLED, &xdp_ring->state);
10607 }
10608 
10609 /**
10610  * ixgbe_enumerate_functions - Get the number of ports this device has
10611  * @adapter: adapter structure
10612  *
10613  * This function enumerates the phsyical functions co-located on a single slot,
10614  * in order to determine how many ports a device has. This is most useful in
10615  * determining the required GT/s of PCIe bandwidth necessary for optimal
10616  * performance.
10617  **/
10618 static inline int ixgbe_enumerate_functions(struct ixgbe_adapter *adapter)
10619 {
10620     struct pci_dev *entry, *pdev = adapter->pdev;
10621     int physfns = 0;
10622 
10623     /* Some cards can not use the generic count PCIe functions method,
10624      * because they are behind a parent switch, so we hardcode these with
10625      * the correct number of functions.
10626      */
10627     if (ixgbe_pcie_from_parent(&adapter->hw))
10628         physfns = 4;
10629 
10630     list_for_each_entry(entry, &adapter->pdev->bus->devices, bus_list) {
10631         /* don't count virtual functions */
10632         if (entry->is_virtfn)
10633             continue;
10634 
10635         /* When the devices on the bus don't all match our device ID,
10636          * we can't reliably determine the correct number of
10637          * functions. This can occur if a function has been direct
10638          * attached to a virtual machine using VT-d, for example. In
10639          * this case, simply return -1 to indicate this.
10640          */
10641         if ((entry->vendor != pdev->vendor) ||
10642             (entry->device != pdev->device))
10643             return -1;
10644 
10645         physfns++;
10646     }
10647 
10648     return physfns;
10649 }
10650 
10651 /**
10652  * ixgbe_wol_supported - Check whether device supports WoL
10653  * @adapter: the adapter private structure
10654  * @device_id: the device ID
10655  * @subdevice_id: the subsystem device ID
10656  *
10657  * This function is used by probe and ethtool to determine
10658  * which devices have WoL support
10659  *
10660  **/
10661 bool ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
10662              u16 subdevice_id)
10663 {
10664     struct ixgbe_hw *hw = &adapter->hw;
10665     u16 wol_cap = adapter->eeprom_cap & IXGBE_DEVICE_CAPS_WOL_MASK;
10666 
10667     /* WOL not supported on 82598 */
10668     if (hw->mac.type == ixgbe_mac_82598EB)
10669         return false;
10670 
10671     /* check eeprom to see if WOL is enabled for X540 and newer */
10672     if (hw->mac.type >= ixgbe_mac_X540) {
10673         if ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
10674             ((wol_cap == IXGBE_DEVICE_CAPS_WOL_PORT0) &&
10675              (hw->bus.func == 0)))
10676             return true;
10677     }
10678 
10679     /* WOL is determined based on device IDs for 82599 MACs */
10680     switch (device_id) {
10681     case IXGBE_DEV_ID_82599_SFP:
10682         /* Only these subdevices could supports WOL */
10683         switch (subdevice_id) {
10684         case IXGBE_SUBDEV_ID_82599_560FLR:
10685         case IXGBE_SUBDEV_ID_82599_LOM_SNAP6:
10686         case IXGBE_SUBDEV_ID_82599_SFP_WOL0:
10687         case IXGBE_SUBDEV_ID_82599_SFP_2OCP:
10688             /* only support first port */
10689             if (hw->bus.func != 0)
10690                 break;
10691             fallthrough;
10692         case IXGBE_SUBDEV_ID_82599_SP_560FLR:
10693         case IXGBE_SUBDEV_ID_82599_SFP:
10694         case IXGBE_SUBDEV_ID_82599_RNDC:
10695         case IXGBE_SUBDEV_ID_82599_ECNA_DP:
10696         case IXGBE_SUBDEV_ID_82599_SFP_1OCP:
10697         case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM1:
10698         case IXGBE_SUBDEV_ID_82599_SFP_LOM_OEM2:
10699             return true;
10700         }
10701         break;
10702     case IXGBE_DEV_ID_82599EN_SFP:
10703         /* Only these subdevices support WOL */
10704         switch (subdevice_id) {
10705         case IXGBE_SUBDEV_ID_82599EN_SFP_OCP1:
10706             return true;
10707         }
10708         break;
10709     case IXGBE_DEV_ID_82599_COMBO_BACKPLANE:
10710         /* All except this subdevice support WOL */
10711         if (subdevice_id != IXGBE_SUBDEV_ID_82599_KX4_KR_MEZZ)
10712             return true;
10713         break;
10714     case IXGBE_DEV_ID_82599_KX4:
10715         return  true;
10716     default:
10717         break;
10718     }
10719 
10720     return false;
10721 }
10722 
10723 /**
10724  * ixgbe_set_fw_version - Set FW version
10725  * @adapter: the adapter private structure
10726  *
10727  * This function is used by probe and ethtool to determine the FW version to
10728  * format to display. The FW version is taken from the EEPROM/NVM.
10729  */
10730 static void ixgbe_set_fw_version(struct ixgbe_adapter *adapter)
10731 {
10732     struct ixgbe_hw *hw = &adapter->hw;
10733     struct ixgbe_nvm_version nvm_ver;
10734 
10735     ixgbe_get_oem_prod_version(hw, &nvm_ver);
10736     if (nvm_ver.oem_valid) {
10737         snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10738              "%x.%x.%x", nvm_ver.oem_major, nvm_ver.oem_minor,
10739              nvm_ver.oem_release);
10740         return;
10741     }
10742 
10743     ixgbe_get_etk_id(hw, &nvm_ver);
10744     ixgbe_get_orom_version(hw, &nvm_ver);
10745 
10746     if (nvm_ver.or_valid) {
10747         snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10748              "0x%08x, %d.%d.%d", nvm_ver.etk_id, nvm_ver.or_major,
10749              nvm_ver.or_build, nvm_ver.or_patch);
10750         return;
10751     }
10752 
10753     /* Set ETrack ID format */
10754     snprintf(adapter->eeprom_id, sizeof(adapter->eeprom_id),
10755          "0x%08x", nvm_ver.etk_id);
10756 }
10757 
10758 /**
10759  * ixgbe_probe - Device Initialization Routine
10760  * @pdev: PCI device information struct
10761  * @ent: entry in ixgbe_pci_tbl
10762  *
10763  * Returns 0 on success, negative on failure
10764  *
10765  * ixgbe_probe initializes an adapter identified by a pci_dev structure.
10766  * The OS initialization, configuring of the adapter private structure,
10767  * and a hardware reset occur.
10768  **/
10769 static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10770 {
10771     struct net_device *netdev;
10772     struct ixgbe_adapter *adapter = NULL;
10773     struct ixgbe_hw *hw;
10774     const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];
10775     unsigned int indices = MAX_TX_QUEUES;
10776     u8 part_str[IXGBE_PBANUM_LENGTH];
10777     int i, err, expected_gts;
10778     bool disable_dev = false;
10779 #ifdef IXGBE_FCOE
10780     u16 device_caps;
10781 #endif
10782     u32 eec;
10783 
10784     /* Catch broken hardware that put the wrong VF device ID in
10785      * the PCIe SR-IOV capability.
10786      */
10787     if (pdev->is_virtfn) {
10788         WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
10789              pci_name(pdev), pdev->vendor, pdev->device);
10790         return -EINVAL;
10791     }
10792 
10793     err = pci_enable_device_mem(pdev);
10794     if (err)
10795         return err;
10796 
10797     err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10798     if (err) {
10799         dev_err(&pdev->dev,
10800             "No usable DMA configuration, aborting\n");
10801         goto err_dma;
10802     }
10803 
10804     err = pci_request_mem_regions(pdev, ixgbe_driver_name);
10805     if (err) {
10806         dev_err(&pdev->dev,
10807             "pci_request_selected_regions failed 0x%x\n", err);
10808         goto err_pci_reg;
10809     }
10810 
10811     pci_enable_pcie_error_reporting(pdev);
10812 
10813     pci_set_master(pdev);
10814     pci_save_state(pdev);
10815 
10816     if (ii->mac == ixgbe_mac_82598EB) {
10817 #ifdef CONFIG_IXGBE_DCB
10818         /* 8 TC w/ 4 queues per TC */
10819         indices = 4 * MAX_TRAFFIC_CLASS;
10820 #else
10821         indices = IXGBE_MAX_RSS_INDICES;
10822 #endif
10823     }
10824 
10825     netdev = alloc_etherdev_mq(sizeof(struct ixgbe_adapter), indices);
10826     if (!netdev) {
10827         err = -ENOMEM;
10828         goto err_alloc_etherdev;
10829     }
10830 
10831     SET_NETDEV_DEV(netdev, &pdev->dev);
10832 
10833     adapter = netdev_priv(netdev);
10834 
10835     adapter->netdev = netdev;
10836     adapter->pdev = pdev;
10837     hw = &adapter->hw;
10838     hw->back = adapter;
10839     adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
10840 
10841     hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
10842                   pci_resource_len(pdev, 0));
10843     adapter->io_addr = hw->hw_addr;
10844     if (!hw->hw_addr) {
10845         err = -EIO;
10846         goto err_ioremap;
10847     }
10848 
10849     netdev->netdev_ops = &ixgbe_netdev_ops;
10850     ixgbe_set_ethtool_ops(netdev);
10851     netdev->watchdog_timeo = 5 * HZ;
10852     strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
10853 
10854     /* Setup hw api */
10855     hw->mac.ops   = *ii->mac_ops;
10856     hw->mac.type  = ii->mac;
10857     hw->mvals     = ii->mvals;
10858     if (ii->link_ops)
10859         hw->link.ops  = *ii->link_ops;
10860 
10861     /* EEPROM */
10862     hw->eeprom.ops = *ii->eeprom_ops;
10863     eec = IXGBE_READ_REG(hw, IXGBE_EEC(hw));
10864     if (ixgbe_removed(hw->hw_addr)) {
10865         err = -EIO;
10866         goto err_ioremap;
10867     }
10868     /* If EEPROM is valid (bit 8 = 1), use default otherwise use bit bang */
10869     if (!(eec & BIT(8)))
10870         hw->eeprom.ops.read = &ixgbe_read_eeprom_bit_bang_generic;
10871 
10872     /* PHY */
10873     hw->phy.ops = *ii->phy_ops;
10874     hw->phy.sfp_type = ixgbe_sfp_type_unknown;
10875     /* ixgbe_identify_phy_generic will set prtad and mmds properly */
10876     hw->phy.mdio.prtad = MDIO_PRTAD_NONE;
10877     hw->phy.mdio.mmds = 0;
10878     hw->phy.mdio.mode_support = MDIO_SUPPORTS_C45 | MDIO_EMULATE_C22;
10879     hw->phy.mdio.dev = netdev;
10880     hw->phy.mdio.mdio_read = ixgbe_mdio_read;
10881     hw->phy.mdio.mdio_write = ixgbe_mdio_write;
10882 
10883     /* setup the private structure */
10884     err = ixgbe_sw_init(adapter, ii);
10885     if (err)
10886         goto err_sw_init;
10887 
10888     if (adapter->hw.mac.type == ixgbe_mac_82599EB)
10889         adapter->flags2 |= IXGBE_FLAG2_AUTO_DISABLE_VF;
10890 
10891     switch (adapter->hw.mac.type) {
10892     case ixgbe_mac_X550:
10893     case ixgbe_mac_X550EM_x:
10894         netdev->udp_tunnel_nic_info = &ixgbe_udp_tunnels_x550;
10895         break;
10896     case ixgbe_mac_x550em_a:
10897         netdev->udp_tunnel_nic_info = &ixgbe_udp_tunnels_x550em_a;
10898         break;
10899     default:
10900         break;
10901     }
10902 
10903     /* Make sure the SWFW semaphore is in a valid state */
10904     if (hw->mac.ops.init_swfw_sync)
10905         hw->mac.ops.init_swfw_sync(hw);
10906 
10907     /* Make it possible the adapter to be woken up via WOL */
10908     switch (adapter->hw.mac.type) {
10909     case ixgbe_mac_82599EB:
10910     case ixgbe_mac_X540:
10911     case ixgbe_mac_X550:
10912     case ixgbe_mac_X550EM_x:
10913     case ixgbe_mac_x550em_a:
10914         IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
10915         break;
10916     default:
10917         break;
10918     }
10919 
10920     /*
10921      * If there is a fan on this device and it has failed log the
10922      * failure.
10923      */
10924     if (adapter->flags & IXGBE_FLAG_FAN_FAIL_CAPABLE) {
10925         u32 esdp = IXGBE_READ_REG(hw, IXGBE_ESDP);
10926         if (esdp & IXGBE_ESDP_SDP1)
10927             e_crit(probe, "Fan has stopped, replace the adapter\n");
10928     }
10929 
10930     if (allow_unsupported_sfp)
10931         hw->allow_unsupported_sfp = allow_unsupported_sfp;
10932 
10933     /* reset_hw fills in the perm_addr as well */
10934     hw->phy.reset_if_overtemp = true;
10935     err = hw->mac.ops.reset_hw(hw);
10936     hw->phy.reset_if_overtemp = false;
10937     ixgbe_set_eee_capable(adapter);
10938     if (err == IXGBE_ERR_SFP_NOT_PRESENT) {
10939         err = 0;
10940     } else if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
10941         e_dev_err("failed to load because an unsupported SFP+ or QSFP module type was detected.\n");
10942         e_dev_err("Reload the driver after installing a supported module.\n");
10943         goto err_sw_init;
10944     } else if (err) {
10945         e_dev_err("HW Init failed: %d\n", err);
10946         goto err_sw_init;
10947     }
10948 
10949 #ifdef CONFIG_PCI_IOV
10950     /* SR-IOV not supported on the 82598 */
10951     if (adapter->hw.mac.type == ixgbe_mac_82598EB)
10952         goto skip_sriov;
10953     /* Mailbox */
10954     ixgbe_init_mbx_params_pf(hw);
10955     hw->mbx.ops = ii->mbx_ops;
10956     pci_sriov_set_totalvfs(pdev, IXGBE_MAX_VFS_DRV_LIMIT);
10957     ixgbe_enable_sriov(adapter, max_vfs);
10958 skip_sriov:
10959 
10960 #endif
10961     netdev->features = NETIF_F_SG |
10962                NETIF_F_TSO |
10963                NETIF_F_TSO6 |
10964                NETIF_F_RXHASH |
10965                NETIF_F_RXCSUM |
10966                NETIF_F_HW_CSUM;
10967 
10968 #define IXGBE_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
10969                     NETIF_F_GSO_GRE_CSUM | \
10970                     NETIF_F_GSO_IPXIP4 | \
10971                     NETIF_F_GSO_IPXIP6 | \
10972                     NETIF_F_GSO_UDP_TUNNEL | \
10973                     NETIF_F_GSO_UDP_TUNNEL_CSUM)
10974 
10975     netdev->gso_partial_features = IXGBE_GSO_PARTIAL_FEATURES;
10976     netdev->features |= NETIF_F_GSO_PARTIAL |
10977                 IXGBE_GSO_PARTIAL_FEATURES;
10978 
10979     if (hw->mac.type >= ixgbe_mac_82599EB)
10980         netdev->features |= NETIF_F_SCTP_CRC | NETIF_F_GSO_UDP_L4;
10981 
10982 #ifdef CONFIG_IXGBE_IPSEC
10983 #define IXGBE_ESP_FEATURES  (NETIF_F_HW_ESP | \
10984                  NETIF_F_HW_ESP_TX_CSUM | \
10985                  NETIF_F_GSO_ESP)
10986 
10987     if (adapter->ipsec)
10988         netdev->features |= IXGBE_ESP_FEATURES;
10989 #endif
10990     /* copy netdev features into list of user selectable features */
10991     netdev->hw_features |= netdev->features |
10992                    NETIF_F_HW_VLAN_CTAG_FILTER |
10993                    NETIF_F_HW_VLAN_CTAG_RX |
10994                    NETIF_F_HW_VLAN_CTAG_TX |
10995                    NETIF_F_RXALL |
10996                    NETIF_F_HW_L2FW_DOFFLOAD;
10997 
10998     if (hw->mac.type >= ixgbe_mac_82599EB)
10999         netdev->hw_features |= NETIF_F_NTUPLE |
11000                        NETIF_F_HW_TC;
11001 
11002     netdev->features |= NETIF_F_HIGHDMA;
11003 
11004     netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
11005     netdev->hw_enc_features |= netdev->vlan_features;
11006     netdev->mpls_features |= NETIF_F_SG |
11007                  NETIF_F_TSO |
11008                  NETIF_F_TSO6 |
11009                  NETIF_F_HW_CSUM;
11010     netdev->mpls_features |= IXGBE_GSO_PARTIAL_FEATURES;
11011 
11012     /* set this bit last since it cannot be part of vlan_features */
11013     netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
11014                 NETIF_F_HW_VLAN_CTAG_RX |
11015                 NETIF_F_HW_VLAN_CTAG_TX;
11016 
11017     netdev->priv_flags |= IFF_UNICAST_FLT;
11018     netdev->priv_flags |= IFF_SUPP_NOFCS;
11019 
11020     /* MTU range: 68 - 9710 */
11021     netdev->min_mtu = ETH_MIN_MTU;
11022     netdev->max_mtu = IXGBE_MAX_JUMBO_FRAME_SIZE - (ETH_HLEN + ETH_FCS_LEN);
11023 
11024 #ifdef CONFIG_IXGBE_DCB
11025     if (adapter->flags & IXGBE_FLAG_DCB_CAPABLE)
11026         netdev->dcbnl_ops = &ixgbe_dcbnl_ops;
11027 #endif
11028 
11029 #ifdef IXGBE_FCOE
11030     if (adapter->flags & IXGBE_FLAG_FCOE_CAPABLE) {
11031         unsigned int fcoe_l;
11032 
11033         if (hw->mac.ops.get_device_caps) {
11034             hw->mac.ops.get_device_caps(hw, &device_caps);
11035             if (device_caps & IXGBE_DEVICE_CAPS_FCOE_OFFLOADS)
11036                 adapter->flags &= ~IXGBE_FLAG_FCOE_CAPABLE;
11037         }
11038 
11039 
11040         fcoe_l = min_t(int, IXGBE_FCRETA_SIZE, num_online_cpus());
11041         adapter->ring_feature[RING_F_FCOE].limit = fcoe_l;
11042 
11043         netdev->features |= NETIF_F_FSO |
11044                     NETIF_F_FCOE_CRC;
11045 
11046         netdev->vlan_features |= NETIF_F_FSO |
11047                      NETIF_F_FCOE_CRC |
11048                      NETIF_F_FCOE_MTU;
11049     }
11050 #endif /* IXGBE_FCOE */
11051     if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
11052         netdev->hw_features |= NETIF_F_LRO;
11053     if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
11054         netdev->features |= NETIF_F_LRO;
11055 
11056     if (ixgbe_check_fw_error(adapter)) {
11057         err = -EIO;
11058         goto err_sw_init;
11059     }
11060 
11061     /* make sure the EEPROM is good */
11062     if (hw->eeprom.ops.validate_checksum(hw, NULL) < 0) {
11063         e_dev_err("The EEPROM Checksum Is Not Valid\n");
11064         err = -EIO;
11065         goto err_sw_init;
11066     }
11067 
11068     eth_platform_get_mac_address(&adapter->pdev->dev,
11069                      adapter->hw.mac.perm_addr);
11070 
11071     eth_hw_addr_set(netdev, hw->mac.perm_addr);
11072 
11073     if (!is_valid_ether_addr(netdev->dev_addr)) {
11074         e_dev_err("invalid MAC address\n");
11075         err = -EIO;
11076         goto err_sw_init;
11077     }
11078 
11079     /* Set hw->mac.addr to permanent MAC address */
11080     ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
11081     ixgbe_mac_set_default_filter(adapter);
11082 
11083     timer_setup(&adapter->service_timer, ixgbe_service_timer, 0);
11084 
11085     if (ixgbe_removed(hw->hw_addr)) {
11086         err = -EIO;
11087         goto err_sw_init;
11088     }
11089     INIT_WORK(&adapter->service_task, ixgbe_service_task);
11090     set_bit(__IXGBE_SERVICE_INITED, &adapter->state);
11091     clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
11092 
11093     err = ixgbe_init_interrupt_scheme(adapter);
11094     if (err)
11095         goto err_sw_init;
11096 
11097     for (i = 0; i < adapter->num_rx_queues; i++)
11098         u64_stats_init(&adapter->rx_ring[i]->syncp);
11099     for (i = 0; i < adapter->num_tx_queues; i++)
11100         u64_stats_init(&adapter->tx_ring[i]->syncp);
11101     for (i = 0; i < adapter->num_xdp_queues; i++)
11102         u64_stats_init(&adapter->xdp_ring[i]->syncp);
11103 
11104     /* WOL not supported for all devices */
11105     adapter->wol = 0;
11106     hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);
11107     hw->wol_enabled = ixgbe_wol_supported(adapter, pdev->device,
11108                         pdev->subsystem_device);
11109     if (hw->wol_enabled)
11110         adapter->wol = IXGBE_WUFC_MAG;
11111 
11112     device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
11113 
11114     /* save off EEPROM version number */
11115     ixgbe_set_fw_version(adapter);
11116 
11117     /* pick up the PCI bus settings for reporting later */
11118     if (ixgbe_pcie_from_parent(hw))
11119         ixgbe_get_parent_bus_info(adapter);
11120     else
11121          hw->mac.ops.get_bus_info(hw);
11122 
11123     /* calculate the expected PCIe bandwidth required for optimal
11124      * performance. Note that some older parts will never have enough
11125      * bandwidth due to being older generation PCIe parts. We clamp these
11126      * parts to ensure no warning is displayed if it can't be fixed.
11127      */
11128     switch (hw->mac.type) {
11129     case ixgbe_mac_82598EB:
11130         expected_gts = min(ixgbe_enumerate_functions(adapter) * 10, 16);
11131         break;
11132     default:
11133         expected_gts = ixgbe_enumerate_functions(adapter) * 10;
11134         break;
11135     }
11136 
11137     /* don't check link if we failed to enumerate functions */
11138     if (expected_gts > 0)
11139         ixgbe_check_minimum_link(adapter, expected_gts);
11140 
11141     err = ixgbe_read_pba_string_generic(hw, part_str, sizeof(part_str));
11142     if (err)
11143         strlcpy(part_str, "Unknown", sizeof(part_str));
11144     if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
11145         e_dev_info("MAC: %d, PHY: %d, SFP+: %d, PBA No: %s\n",
11146                hw->mac.type, hw->phy.type, hw->phy.sfp_type,
11147                part_str);
11148     else
11149         e_dev_info("MAC: %d, PHY: %d, PBA No: %s\n",
11150                hw->mac.type, hw->phy.type, part_str);
11151 
11152     e_dev_info("%pM\n", netdev->dev_addr);
11153 
11154     /* reset the hardware with the new settings */
11155     err = hw->mac.ops.start_hw(hw);
11156     if (err == IXGBE_ERR_EEPROM_VERSION) {
11157         /* We are running on a pre-production device, log a warning */
11158         e_dev_warn("This device is a pre-production adapter/LOM. "
11159                "Please be aware there may be issues associated "
11160                "with your hardware.  If you are experiencing "
11161                "problems please contact your Intel or hardware "
11162                "representative who provided you with this "
11163                "hardware.\n");
11164     }
11165     strcpy(netdev->name, "eth%d");
11166     pci_set_drvdata(pdev, adapter);
11167     err = register_netdev(netdev);
11168     if (err)
11169         goto err_register;
11170 
11171 
11172     /* power down the optics for 82599 SFP+ fiber */
11173     if (hw->mac.ops.disable_tx_laser)
11174         hw->mac.ops.disable_tx_laser(hw);
11175 
11176     /* carrier off reporting is important to ethtool even BEFORE open */
11177     netif_carrier_off(netdev);
11178 
11179 #ifdef CONFIG_IXGBE_DCA
11180     if (dca_add_requester(&pdev->dev) == 0) {
11181         adapter->flags |= IXGBE_FLAG_DCA_ENABLED;
11182         ixgbe_setup_dca(adapter);
11183     }
11184 #endif
11185     if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED) {
11186         e_info(probe, "IOV is enabled with %d VFs\n", adapter->num_vfs);
11187         for (i = 0; i < adapter->num_vfs; i++)
11188             ixgbe_vf_configuration(pdev, (i | 0x10000000));
11189     }
11190 
11191     /* firmware requires driver version to be 0xFFFFFFFF
11192      * since os does not support feature
11193      */
11194     if (hw->mac.ops.set_fw_drv_ver)
11195         hw->mac.ops.set_fw_drv_ver(hw, 0xFF, 0xFF, 0xFF, 0xFF,
11196                        sizeof(UTS_RELEASE) - 1,
11197                        UTS_RELEASE);
11198 
11199     /* add san mac addr to netdev */
11200     ixgbe_add_sanmac_netdev(netdev);
11201 
11202     e_dev_info("%s\n", ixgbe_default_device_descr);
11203 
11204 #ifdef CONFIG_IXGBE_HWMON
11205     if (ixgbe_sysfs_init(adapter))
11206         e_err(probe, "failed to allocate sysfs resources\n");
11207 #endif /* CONFIG_IXGBE_HWMON */
11208 
11209     ixgbe_dbg_adapter_init(adapter);
11210 
11211     /* setup link for SFP devices with MNG FW, else wait for IXGBE_UP */
11212     if (ixgbe_mng_enabled(hw) && ixgbe_is_sfp(hw) && hw->mac.ops.setup_link)
11213         hw->mac.ops.setup_link(hw,
11214             IXGBE_LINK_SPEED_10GB_FULL | IXGBE_LINK_SPEED_1GB_FULL,
11215             true);
11216 
11217     err = ixgbe_mii_bus_init(hw);
11218     if (err)
11219         goto err_netdev;
11220 
11221     return 0;
11222 
11223 err_netdev:
11224     unregister_netdev(netdev);
11225 err_register:
11226     ixgbe_release_hw_control(adapter);
11227     ixgbe_clear_interrupt_scheme(adapter);
11228 err_sw_init:
11229     ixgbe_disable_sriov(adapter);
11230     adapter->flags2 &= ~IXGBE_FLAG2_SEARCH_FOR_SFP;
11231     iounmap(adapter->io_addr);
11232     kfree(adapter->jump_tables[0]);
11233     kfree(adapter->mac_table);
11234     kfree(adapter->rss_key);
11235     bitmap_free(adapter->af_xdp_zc_qps);
11236 err_ioremap:
11237     disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
11238     free_netdev(netdev);
11239 err_alloc_etherdev:
11240     pci_disable_pcie_error_reporting(pdev);
11241     pci_release_mem_regions(pdev);
11242 err_pci_reg:
11243 err_dma:
11244     if (!adapter || disable_dev)
11245         pci_disable_device(pdev);
11246     return err;
11247 }
11248 
11249 /**
11250  * ixgbe_remove - Device Removal Routine
11251  * @pdev: PCI device information struct
11252  *
11253  * ixgbe_remove is called by the PCI subsystem to alert the driver
11254  * that it should release a PCI device.  The could be caused by a
11255  * Hot-Plug event, or because the driver is going to be removed from
11256  * memory.
11257  **/
11258 static void ixgbe_remove(struct pci_dev *pdev)
11259 {
11260     struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11261     struct net_device *netdev;
11262     bool disable_dev;
11263     int i;
11264 
11265     /* if !adapter then we already cleaned up in probe */
11266     if (!adapter)
11267         return;
11268 
11269     netdev  = adapter->netdev;
11270     ixgbe_dbg_adapter_exit(adapter);
11271 
11272     set_bit(__IXGBE_REMOVING, &adapter->state);
11273     cancel_work_sync(&adapter->service_task);
11274 
11275     if (adapter->mii_bus)
11276         mdiobus_unregister(adapter->mii_bus);
11277 
11278 #ifdef CONFIG_IXGBE_DCA
11279     if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
11280         adapter->flags &= ~IXGBE_FLAG_DCA_ENABLED;
11281         dca_remove_requester(&pdev->dev);
11282         IXGBE_WRITE_REG(&adapter->hw, IXGBE_DCA_CTRL,
11283                 IXGBE_DCA_CTRL_DCA_DISABLE);
11284     }
11285 
11286 #endif
11287 #ifdef CONFIG_IXGBE_HWMON
11288     ixgbe_sysfs_exit(adapter);
11289 #endif /* CONFIG_IXGBE_HWMON */
11290 
11291     /* remove the added san mac */
11292     ixgbe_del_sanmac_netdev(netdev);
11293 
11294 #ifdef CONFIG_PCI_IOV
11295     ixgbe_disable_sriov(adapter);
11296 #endif
11297     if (netdev->reg_state == NETREG_REGISTERED)
11298         unregister_netdev(netdev);
11299 
11300     ixgbe_stop_ipsec_offload(adapter);
11301     ixgbe_clear_interrupt_scheme(adapter);
11302 
11303     ixgbe_release_hw_control(adapter);
11304 
11305 #ifdef CONFIG_DCB
11306     kfree(adapter->ixgbe_ieee_pfc);
11307     kfree(adapter->ixgbe_ieee_ets);
11308 
11309 #endif
11310     iounmap(adapter->io_addr);
11311     pci_release_mem_regions(pdev);
11312 
11313     e_dev_info("complete\n");
11314 
11315     for (i = 0; i < IXGBE_MAX_LINK_HANDLE; i++) {
11316         if (adapter->jump_tables[i]) {
11317             kfree(adapter->jump_tables[i]->input);
11318             kfree(adapter->jump_tables[i]->mask);
11319         }
11320         kfree(adapter->jump_tables[i]);
11321     }
11322 
11323     kfree(adapter->mac_table);
11324     kfree(adapter->rss_key);
11325     bitmap_free(adapter->af_xdp_zc_qps);
11326     disable_dev = !test_and_set_bit(__IXGBE_DISABLED, &adapter->state);
11327     free_netdev(netdev);
11328 
11329     pci_disable_pcie_error_reporting(pdev);
11330 
11331     if (disable_dev)
11332         pci_disable_device(pdev);
11333 }
11334 
11335 /**
11336  * ixgbe_io_error_detected - called when PCI error is detected
11337  * @pdev: Pointer to PCI device
11338  * @state: The current pci connection state
11339  *
11340  * This function is called after a PCI bus error affecting
11341  * this device has been detected.
11342  */
11343 static pci_ers_result_t ixgbe_io_error_detected(struct pci_dev *pdev,
11344                         pci_channel_state_t state)
11345 {
11346     struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11347     struct net_device *netdev = adapter->netdev;
11348 
11349 #ifdef CONFIG_PCI_IOV
11350     struct ixgbe_hw *hw = &adapter->hw;
11351     struct pci_dev *bdev, *vfdev;
11352     u32 dw0, dw1, dw2, dw3;
11353     int vf, pos;
11354     u16 req_id, pf_func;
11355 
11356     if (adapter->hw.mac.type == ixgbe_mac_82598EB ||
11357         adapter->num_vfs == 0)
11358         goto skip_bad_vf_detection;
11359 
11360     bdev = pdev->bus->self;
11361     while (bdev && (pci_pcie_type(bdev) != PCI_EXP_TYPE_ROOT_PORT))
11362         bdev = bdev->bus->self;
11363 
11364     if (!bdev)
11365         goto skip_bad_vf_detection;
11366 
11367     pos = pci_find_ext_capability(bdev, PCI_EXT_CAP_ID_ERR);
11368     if (!pos)
11369         goto skip_bad_vf_detection;
11370 
11371     dw0 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG);
11372     dw1 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 4);
11373     dw2 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 8);
11374     dw3 = ixgbe_read_pci_cfg_dword(hw, pos + PCI_ERR_HEADER_LOG + 12);
11375     if (ixgbe_removed(hw->hw_addr))
11376         goto skip_bad_vf_detection;
11377 
11378     req_id = dw1 >> 16;
11379     /* On the 82599 if bit 7 of the requestor ID is set then it's a VF */
11380     if (!(req_id & 0x0080))
11381         goto skip_bad_vf_detection;
11382 
11383     pf_func = req_id & 0x01;
11384     if ((pf_func & 1) == (pdev->devfn & 1)) {
11385         unsigned int device_id;
11386 
11387         vf = (req_id & 0x7F) >> 1;
11388         e_dev_err("VF %d has caused a PCIe error\n", vf);
11389         e_dev_err("TLP: dw0: %8.8x\tdw1: %8.8x\tdw2: "
11390                 "%8.8x\tdw3: %8.8x\n",
11391         dw0, dw1, dw2, dw3);
11392         switch (adapter->hw.mac.type) {
11393         case ixgbe_mac_82599EB:
11394             device_id = IXGBE_82599_VF_DEVICE_ID;
11395             break;
11396         case ixgbe_mac_X540:
11397             device_id = IXGBE_X540_VF_DEVICE_ID;
11398             break;
11399         case ixgbe_mac_X550:
11400             device_id = IXGBE_DEV_ID_X550_VF;
11401             break;
11402         case ixgbe_mac_X550EM_x:
11403             device_id = IXGBE_DEV_ID_X550EM_X_VF;
11404             break;
11405         case ixgbe_mac_x550em_a:
11406             device_id = IXGBE_DEV_ID_X550EM_A_VF;
11407             break;
11408         default:
11409             device_id = 0;
11410             break;
11411         }
11412 
11413         /* Find the pci device of the offending VF */
11414         vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, device_id, NULL);
11415         while (vfdev) {
11416             if (vfdev->devfn == (req_id & 0xFF))
11417                 break;
11418             vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
11419                            device_id, vfdev);
11420         }
11421         /*
11422          * There's a slim chance the VF could have been hot plugged,
11423          * so if it is no longer present we don't need to issue the
11424          * VFLR.  Just clean up the AER in that case.
11425          */
11426         if (vfdev) {
11427             pcie_flr(vfdev);
11428             /* Free device reference count */
11429             pci_dev_put(vfdev);
11430         }
11431     }
11432 
11433     /*
11434      * Even though the error may have occurred on the other port
11435      * we still need to increment the vf error reference count for
11436      * both ports because the I/O resume function will be called
11437      * for both of them.
11438      */
11439     adapter->vferr_refcount++;
11440 
11441     return PCI_ERS_RESULT_RECOVERED;
11442 
11443 skip_bad_vf_detection:
11444 #endif /* CONFIG_PCI_IOV */
11445     if (!test_bit(__IXGBE_SERVICE_INITED, &adapter->state))
11446         return PCI_ERS_RESULT_DISCONNECT;
11447 
11448     if (!netif_device_present(netdev))
11449         return PCI_ERS_RESULT_DISCONNECT;
11450 
11451     rtnl_lock();
11452     netif_device_detach(netdev);
11453 
11454     if (netif_running(netdev))
11455         ixgbe_close_suspend(adapter);
11456 
11457     if (state == pci_channel_io_perm_failure) {
11458         rtnl_unlock();
11459         return PCI_ERS_RESULT_DISCONNECT;
11460     }
11461 
11462     if (!test_and_set_bit(__IXGBE_DISABLED, &adapter->state))
11463         pci_disable_device(pdev);
11464     rtnl_unlock();
11465 
11466     /* Request a slot reset. */
11467     return PCI_ERS_RESULT_NEED_RESET;
11468 }
11469 
11470 /**
11471  * ixgbe_io_slot_reset - called after the pci bus has been reset.
11472  * @pdev: Pointer to PCI device
11473  *
11474  * Restart the card from scratch, as if from a cold-boot.
11475  */
11476 static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
11477 {
11478     struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11479     pci_ers_result_t result;
11480 
11481     if (pci_enable_device_mem(pdev)) {
11482         e_err(probe, "Cannot re-enable PCI device after reset.\n");
11483         result = PCI_ERS_RESULT_DISCONNECT;
11484     } else {
11485         smp_mb__before_atomic();
11486         clear_bit(__IXGBE_DISABLED, &adapter->state);
11487         adapter->hw.hw_addr = adapter->io_addr;
11488         pci_set_master(pdev);
11489         pci_restore_state(pdev);
11490         pci_save_state(pdev);
11491 
11492         pci_wake_from_d3(pdev, false);
11493 
11494         ixgbe_reset(adapter);
11495         IXGBE_WRITE_REG(&adapter->hw, IXGBE_WUS, ~0);
11496         result = PCI_ERS_RESULT_RECOVERED;
11497     }
11498 
11499     return result;
11500 }
11501 
11502 /**
11503  * ixgbe_io_resume - called when traffic can start flowing again.
11504  * @pdev: Pointer to PCI device
11505  *
11506  * This callback is called when the error recovery driver tells us that
11507  * its OK to resume normal operation.
11508  */
11509 static void ixgbe_io_resume(struct pci_dev *pdev)
11510 {
11511     struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
11512     struct net_device *netdev = adapter->netdev;
11513 
11514 #ifdef CONFIG_PCI_IOV
11515     if (adapter->vferr_refcount) {
11516         e_info(drv, "Resuming after VF err\n");
11517         adapter->vferr_refcount--;
11518         return;
11519     }
11520 
11521 #endif
11522     rtnl_lock();
11523     if (netif_running(netdev))
11524         ixgbe_open(netdev);
11525 
11526     netif_device_attach(netdev);
11527     rtnl_unlock();
11528 }
11529 
11530 static const struct pci_error_handlers ixgbe_err_handler = {
11531     .error_detected = ixgbe_io_error_detected,
11532     .slot_reset = ixgbe_io_slot_reset,
11533     .resume = ixgbe_io_resume,
11534 };
11535 
11536 static SIMPLE_DEV_PM_OPS(ixgbe_pm_ops, ixgbe_suspend, ixgbe_resume);
11537 
11538 static struct pci_driver ixgbe_driver = {
11539     .name      = ixgbe_driver_name,
11540     .id_table  = ixgbe_pci_tbl,
11541     .probe     = ixgbe_probe,
11542     .remove    = ixgbe_remove,
11543     .driver.pm = &ixgbe_pm_ops,
11544     .shutdown  = ixgbe_shutdown,
11545     .sriov_configure = ixgbe_pci_sriov_configure,
11546     .err_handler = &ixgbe_err_handler
11547 };
11548 
11549 /**
11550  * ixgbe_init_module - Driver Registration Routine
11551  *
11552  * ixgbe_init_module is the first routine called when the driver is
11553  * loaded. All it does is register with the PCI subsystem.
11554  **/
11555 static int __init ixgbe_init_module(void)
11556 {
11557     int ret;
11558     pr_info("%s\n", ixgbe_driver_string);
11559     pr_info("%s\n", ixgbe_copyright);
11560 
11561     ixgbe_wq = create_singlethread_workqueue(ixgbe_driver_name);
11562     if (!ixgbe_wq) {
11563         pr_err("%s: Failed to create workqueue\n", ixgbe_driver_name);
11564         return -ENOMEM;
11565     }
11566 
11567     ixgbe_dbg_init();
11568 
11569     ret = pci_register_driver(&ixgbe_driver);
11570     if (ret) {
11571         destroy_workqueue(ixgbe_wq);
11572         ixgbe_dbg_exit();
11573         return ret;
11574     }
11575 
11576 #ifdef CONFIG_IXGBE_DCA
11577     dca_register_notify(&dca_notifier);
11578 #endif
11579 
11580     return 0;
11581 }
11582 
11583 module_init(ixgbe_init_module);
11584 
11585 /**
11586  * ixgbe_exit_module - Driver Exit Cleanup Routine
11587  *
11588  * ixgbe_exit_module is called just before the driver is removed
11589  * from memory.
11590  **/
11591 static void __exit ixgbe_exit_module(void)
11592 {
11593 #ifdef CONFIG_IXGBE_DCA
11594     dca_unregister_notify(&dca_notifier);
11595 #endif
11596     pci_unregister_driver(&ixgbe_driver);
11597 
11598     ixgbe_dbg_exit();
11599     if (ixgbe_wq) {
11600         destroy_workqueue(ixgbe_wq);
11601         ixgbe_wq = NULL;
11602     }
11603 }
11604 
11605 #ifdef CONFIG_IXGBE_DCA
11606 static int ixgbe_notify_dca(struct notifier_block *nb, unsigned long event,
11607                 void *p)
11608 {
11609     int ret_val;
11610 
11611     ret_val = driver_for_each_device(&ixgbe_driver.driver, NULL, &event,
11612                      __ixgbe_notify_dca);
11613 
11614     return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
11615 }
11616 
11617 #endif /* CONFIG_IXGBE_DCA */
11618 
11619 module_exit(ixgbe_exit_module);
11620 
11621 /* ixgbe_main.c */