Index: sys/dev/ixgbe/if_ix.c =================================================================== --- sys/dev/ixgbe/if_ix.c +++ sys/dev/ixgbe/if_ix.c @@ -33,13 +33,15 @@ /*$FreeBSD$*/ -#ifndef IXGBE_STANDALONE_BUILD #include "opt_inet.h" #include "opt_inet6.h" #include "opt_rss.h" -#endif #include "ixgbe.h" +#include "ifdi_if.h" + +#include +#include #ifdef RSS #include @@ -47,10 +49,14 @@ #endif /********************************************************************* - * Driver version + * Set this to one to display debug statistics *********************************************************************/ -char ixgbe_driver_version[] = "3.1.13-k"; +int ixgbe_display_debug_stats = 0; +/********************************************************************* + * Driver version + *********************************************************************/ +char ixgbe_driver_version[] = "3.1.0"; /********************************************************************* * PCI Device ID Table @@ -61,149 +67,140 @@ * * { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index } *********************************************************************/ - -static ixgbe_vendor_info_t ixgbe_vendor_info_array[] = +static pci_vendor_info_t ixgbe_vendor_info_array[] = { - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T, 0, 0, 0}, - {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP, 0, 0, 0}, + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4, "Intel(R) PRO/10GbE PCI-Express Network Driver"), + PVID(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T, "Intel(R) PRO/10GbE PCI-Express Network Driver"), /* required last entry */ - {0, 0, 0, 0, 0} -}; - -/********************************************************************* - * Table of branding strings - *********************************************************************/ - -static char *ixgbe_strings[] = { - "Intel(R) PRO/10GbE PCI-Express Network Driver" + PVID_END }; /********************************************************************* * Function prototypes *********************************************************************/ -static int ixgbe_probe(device_t); -static int ixgbe_attach(device_t); -static int ixgbe_detach(device_t); -static int ixgbe_shutdown(device_t); -static int ixgbe_suspend(device_t); -static int ixgbe_resume(device_t); -static int ixgbe_ioctl(struct ifnet *, u_long, caddr_t); -static void ixgbe_init(void *); -static void ixgbe_init_locked(struct adapter *); -static void ixgbe_stop(void *); +static void *ixgbe_register(device_t dev); +static int ixgbe_if_attach_pre(if_ctx_t ctx); +static int ixgbe_if_attach_post(if_ctx_t ctx); +static int ixgbe_if_detach(if_ctx_t ctx); +static int ixgbe_if_shutdown(if_ctx_t ctx); +static int ixgbe_if_suspend(if_ctx_t ctx); +static int ixgbe_if_resume(if_ctx_t ctx); + +static void ixgbe_if_stop(if_ctx_t ctx); +static void ixgbe_if_init(if_ctx_t ctx); +void ixgbe_if_enable_intr(if_ctx_t ctx); +static void ixgbe_if_disable_intr(if_ctx_t ctx); +static void ixgbe_if_queue_intr_enable(if_ctx_t ctx, uint16_t qid); +static void ixgbe_if_media_status(if_ctx_t ctx, struct ifmediareq * ifmr); +static int ixgbe_if_media_change(if_ctx_t ctx); +static int ixgbe_if_msix_intr_assign(if_ctx_t, int); +static int ixgbe_if_mtu_set(if_ctx_t ctx, uint32_t mtu); +static void ixgbe_if_crcstrip_set(if_ctx_t ctx, int onoff); +static void ixgbe_if_multi_set(if_ctx_t ctx); +static int ixgbe_if_promisc_set(if_ctx_t ctx, int flags); +static int ixgbe_if_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nqs); +static void ixgbe_if_queues_free(if_ctx_t ctx); +static void ixgbe_if_timer(if_ctx_t ctx, uint16_t); +static void ixgbe_if_update_admin_status(if_ctx_t ctx); +static void ixgbe_if_vlan_register(if_ctx_t ctx, u16 vtag); +static void ixgbe_if_vlan_unregister(if_ctx_t ctx, u16 vtag); + +int ixgbe_intr(void *arg); + #if __FreeBSD_version >= 1100036 -static uint64_t ixgbe_get_counter(struct ifnet *, ift_counter); +static uint64_t ixgbe_if_get_counter(if_ctx_t, ift_counter); #endif -static void ixgbe_add_media_types(struct adapter *); -static void ixgbe_media_status(struct ifnet *, struct ifmediareq *); -static int ixgbe_media_change(struct ifnet *); -static void ixgbe_identify_hardware(struct adapter *); -static int ixgbe_allocate_pci_resources(struct adapter *); -static void ixgbe_get_slot_info(struct adapter *); -static int ixgbe_allocate_msix(struct adapter *); -static int ixgbe_allocate_legacy(struct adapter *); -static int ixgbe_setup_msix(struct adapter *); -static void ixgbe_free_pci_resources(struct adapter *); -static void ixgbe_local_timer(void *); -static int ixgbe_setup_interface(device_t, struct adapter *); -static void ixgbe_config_gpie(struct adapter *); -static void ixgbe_config_dmac(struct adapter *); -static void ixgbe_config_delay_values(struct adapter *); -static void ixgbe_config_link(struct adapter *); -static void ixgbe_check_wol_support(struct adapter *); -static int ixgbe_setup_low_power_mode(struct adapter *); -static void ixgbe_rearm_queues(struct adapter *, u64); - -static void ixgbe_initialize_transmit_units(struct adapter *); -static void ixgbe_initialize_receive_units(struct adapter *); -static void ixgbe_enable_rx_drop(struct adapter *); -static void ixgbe_disable_rx_drop(struct adapter *); -static void ixgbe_initialize_rss_mapping(struct adapter *); - -static void ixgbe_enable_intr(struct adapter *); -static void ixgbe_disable_intr(struct adapter *); -static void ixgbe_update_stats_counters(struct adapter *); -static void ixgbe_set_promisc(struct adapter *); -static void ixgbe_set_multi(struct adapter *); -static void ixgbe_update_link_status(struct adapter *); -static void ixgbe_set_ivar(struct adapter *, u8, u8, s8); -static void ixgbe_configure_ivars(struct adapter *); -static u8 * ixgbe_mc_array_itr(struct ixgbe_hw *, u8 **, u32 *); - -static void ixgbe_setup_vlan_hw_support(struct adapter *); -static void ixgbe_register_vlan(void *, struct ifnet *, u16); -static void ixgbe_unregister_vlan(void *, struct ifnet *, u16); - -static void ixgbe_add_device_sysctls(struct adapter *); -static void ixgbe_add_hw_stats(struct adapter *); -static int ixgbe_set_flowcntl(struct adapter *, int); -static int ixgbe_set_advertise(struct adapter *, int); + +static void ixgbe_enable_queue(struct adapter *adapter, u32 vector); +static void ixgbe_disable_queue(struct adapter *adapter, u32 vector); +static void ixgbe_add_device_sysctls(if_ctx_t ctx); +static int ixgbe_allocate_pci_resources(if_ctx_t ctx); +static int ixgbe_setup_low_power_mode(if_ctx_t ctx); + +static void ixgbe_config_dmac(struct adapter *adapter); +static void ixgbe_configure_ivars(struct adapter *adapter); +static void ixgbe_set_ivar(struct adapter *adapter, u8 entry, u8 vector, s8 type); +static u8 * ixgbe_mc_array_itr(struct ixgbe_hw *, u8 **, u32 *); +static bool ixgbe_sfp_probe(if_ctx_t ctx); + +static void ixgbe_identify_hardware(if_ctx_t ctx); +static void ixgbe_free_pci_resources(if_ctx_t ctx); + +static int ixgbe_msix_link(void *arg); +static int ixgbe_msix_que(void *arg); +static void ixgbe_initialise_rss_mapping(struct adapter *adapter); +static void ixgbe_initialize_receive_units(if_ctx_t ctx); +static void ixgbe_initialize_transmit_units(if_ctx_t ctx); + +static int ixgbe_interface_setup(if_ctx_t ctx); +static void ixgbe_add_media_types(if_ctx_t ctx); +static void ixgbe_update_stats_counters(struct adapter *adapter); +static void ixgbe_config_link(struct adapter *adapter); +static void ixgbe_get_slot_info(struct ixgbe_hw *hw); +static void ixgbe_check_eee_support(struct adapter *adapter); +static void ixgbe_check_wol_support(struct adapter *adapter); +static void ixgbe_enable_rx_drop(struct adapter *); +static void ixgbe_disable_rx_drop(struct adapter *); + +static void ixgbe_add_hw_stats(struct adapter *adapter); +static void ixgbe_setup_vlan_hw_support(if_ctx_t ctx); +static void ixgbe_setup_optics(struct adapter *adapter); +static void ixgbe_config_gpie(struct adapter *adapter); +static void ixgbe_config_delay_values(struct adapter *adapter); /* Sysctl handlers */ -static void ixgbe_set_sysctl_value(struct adapter *, const char *, - const char *, int *, int); -static int ixgbe_sysctl_flowcntl(SYSCTL_HANDLER_ARGS); -static int ixgbe_sysctl_advertise(SYSCTL_HANDLER_ARGS); -static int ixgbe_sysctl_thermal_test(SYSCTL_HANDLER_ARGS); -static int ixgbe_sysctl_dmac(SYSCTL_HANDLER_ARGS); -static int ixgbe_sysctl_phy_temp(SYSCTL_HANDLER_ARGS); -static int ixgbe_sysctl_phy_overtemp_occurred(SYSCTL_HANDLER_ARGS); -#ifdef IXGBE_DEBUG -static int ixgbe_sysctl_power_state(SYSCTL_HANDLER_ARGS); -static int ixgbe_sysctl_print_rss_config(SYSCTL_HANDLER_ARGS); -#endif -static int ixgbe_sysctl_wol_enable(SYSCTL_HANDLER_ARGS); -static int ixgbe_sysctl_wufc(SYSCTL_HANDLER_ARGS); -static int ixgbe_sysctl_eee_enable(SYSCTL_HANDLER_ARGS); -static int ixgbe_sysctl_eee_negotiated(SYSCTL_HANDLER_ARGS); -static int ixgbe_sysctl_eee_rx_lpi_status(SYSCTL_HANDLER_ARGS); -static int ixgbe_sysctl_eee_tx_lpi_status(SYSCTL_HANDLER_ARGS); -static int ixgbe_sysctl_eee_tx_lpi_delay(SYSCTL_HANDLER_ARGS); +static void ixgbe_set_sysctl_value(struct adapter *adapter, const char *name, + const char *description, int *limit, int value); +static int ixgbe_set_flowcntl(SYSCTL_HANDLER_ARGS); +static int ixgbe_set_advertise(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_interrupt_rate_handler(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_thermal_test(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_dmac(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_phy_temp(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_phy_overtemp_occurred(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_rdh_handler(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_rdt_handler(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_tdt_handler(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_tdh_handler(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_wol_enable(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_wufc(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_eee_enable(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_eee_negotiated(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_eee_rx_lpi_status(SYSCTL_HANDLER_ARGS); +static int ixgbe_sysctl_eee_tx_lpi_status(SYSCTL_HANDLER_ARGS); /* Support for pluggable optic modules */ -static bool ixgbe_sfp_probe(struct adapter *); static void ixgbe_setup_optics(struct adapter *); -/* Legacy (single vector interrupt handler */ -static void ixgbe_legacy_irq(void *); - -/* The MSI/X Interrupt handlers */ -static void ixgbe_msix_que(void *); -static void ixgbe_msix_link(void *); - /* Deferred interrupt tasklets */ -static void ixgbe_handle_que(void *, int); -static void ixgbe_handle_link(void *, int); static void ixgbe_handle_msf(void *, int); static void ixgbe_handle_mod(void *, int); static void ixgbe_handle_phy(void *, int); @@ -223,19 +220,18 @@ static void ixgbe_init_vf(struct adapter *, struct ixgbe_vf *); #endif /* PCI_IOV */ - -/********************************************************************* +/********************************************************************** * FreeBSD Device Interface Entry Points *********************************************************************/ - -static device_method_t ix_methods[] = { +static device_method_t ixgbe_methods[] = { /* Device interface */ - DEVMETHOD(device_probe, ixgbe_probe), - DEVMETHOD(device_attach, ixgbe_attach), - DEVMETHOD(device_detach, ixgbe_detach), - DEVMETHOD(device_shutdown, ixgbe_shutdown), - DEVMETHOD(device_suspend, ixgbe_suspend), - DEVMETHOD(device_resume, ixgbe_resume), + DEVMETHOD(device_register, ixgbe_register), + DEVMETHOD(device_probe, iflib_device_probe), + DEVMETHOD(device_attach, iflib_device_attach), + DEVMETHOD(device_detach, iflib_device_detach), + DEVMETHOD(device_shutdown, iflib_device_shutdown), + DEVMETHOD(device_suspend, iflib_device_suspend), + DEVMETHOD(device_resume, iflib_device_resume), #ifdef PCI_IOV DEVMETHOD(pci_iov_init, ixgbe_init_iov), DEVMETHOD(pci_iov_uninit, ixgbe_uninit_iov), @@ -244,18 +240,54 @@ DEVMETHOD_END }; -static driver_t ix_driver = { - "ix", ix_methods, sizeof(struct adapter), +static driver_t ixgbe_driver = { + "ix", ixgbe_methods, sizeof(struct adapter), }; devclass_t ix_devclass; -DRIVER_MODULE(ix, pci, ix_driver, ix_devclass, 0, 0); +DRIVER_MODULE(ix, pci, ixgbe_driver, ix_devclass, 0, 0); MODULE_DEPEND(ix, pci, 1, 1, 1); MODULE_DEPEND(ix, ether, 1, 1, 1); -#ifdef DEV_NETMAP -MODULE_DEPEND(ix, netmap, 1, 1, 1); -#endif /* DEV_NETMAP */ +MODULE_DEPEND(ix, iflib, 1, 1, 1); + +static device_method_t ixgbe_if_methods[] = { + DEVMETHOD(ifdi_attach_pre, ixgbe_if_attach_pre), + DEVMETHOD(ifdi_attach_post, ixgbe_if_attach_post), + DEVMETHOD(ifdi_detach, ixgbe_if_detach), + DEVMETHOD(ifdi_shutdown, ixgbe_if_shutdown), + DEVMETHOD(ifdi_suspend, ixgbe_if_suspend), + DEVMETHOD(ifdi_resume, ixgbe_if_resume), + DEVMETHOD(ifdi_init, ixgbe_if_init), + DEVMETHOD(ifdi_stop, ixgbe_if_stop), + DEVMETHOD(ifdi_msix_intr_assign, ixgbe_if_msix_intr_assign), + DEVMETHOD(ifdi_intr_enable, ixgbe_if_enable_intr), + DEVMETHOD(ifdi_intr_disable, ixgbe_if_disable_intr), + DEVMETHOD(ifdi_queue_intr_enable, ixgbe_if_queue_intr_enable), + DEVMETHOD(ifdi_queues_alloc, ixgbe_if_queues_alloc), + DEVMETHOD(ifdi_queues_free, ixgbe_if_queues_free), + DEVMETHOD(ifdi_update_admin_status, ixgbe_if_update_admin_status), + DEVMETHOD(ifdi_multi_set, ixgbe_if_multi_set), + DEVMETHOD(ifdi_mtu_set, ixgbe_if_mtu_set), + DEVMETHOD(ifdi_crcstrip_set, ixgbe_if_crcstrip_set), + DEVMETHOD(ifdi_media_status, ixgbe_if_media_status), + DEVMETHOD(ifdi_media_change, ixgbe_if_media_change), + DEVMETHOD(ifdi_promisc_set, ixgbe_if_promisc_set), + DEVMETHOD(ifdi_timer, ixgbe_if_timer), + DEVMETHOD(ifdi_vlan_register, ixgbe_if_vlan_register), + DEVMETHOD(ifdi_vlan_unregister, ixgbe_if_vlan_unregister), + DEVMETHOD(ifdi_get_counter, ixgbe_if_get_counter), + DEVMETHOD_END +}; + +/* + * note that if (adapter->msix_mem) is replaced by: + * if (adapter->intr_type == IFLIB_INTR_MSIX) + */ + +static driver_t ixgbe_if_driver = { + "ixgbe_if", ixgbe_if_methods, sizeof(struct adapter) +}; /* ** TUNEABLE PARAMETERS: @@ -292,16 +324,6 @@ "Maximum number of sent packets to process at a time," "-1 means unlimited"); -/* Flow control setting, default to full */ -static int ixgbe_flow_control = ixgbe_fc_full; -SYSCTL_INT(_hw_ix, OID_AUTO, flow_control, CTLFLAG_RDTUN, - &ixgbe_flow_control, 0, "Default flow control used for all adapters"); - -/* Advertise Speed, default to 0 (auto) */ -static int ixgbe_advertise_speed = 0; -SYSCTL_INT(_hw_ix, OID_AUTO, advertise_speed, CTLFLAG_RDTUN, - &ixgbe_advertise_speed, 0, "Default advertised speed for all adapters"); - /* ** Smart speed setting, default to on ** this only works as a compile option @@ -325,9 +347,12 @@ * number of cpus with a max of 8. This * can be overriden manually here. */ +#if 0 +/* determined by iflib */ static int ixgbe_num_queues = 0; SYSCTL_INT(_hw_ix, OID_AUTO, num_queues, CTLFLAG_RDTUN, &ixgbe_num_queues, 0, "Number of queues to configure, 0 indicates autoconfigure"); +#endif /* ** Number of TX descriptors per ring, @@ -343,6 +368,17 @@ SYSCTL_INT(_hw_ix, OID_AUTO, rxd, CTLFLAG_RDTUN, &ixgbe_rxd, 0, "Number of receive descriptors per queue"); + +/* ix_crcstrip: 0: keep CRC in rx frames (default), 1: strip it. + * During regular operations the CRC is stripped, but on some + * hardware reception of frames not multiple of 64 is slower, + * so using crcstrip=0 helps in benchmarks. + */ +SYSCTL_DECL(_dev_netmap); +int ix_crcstrip; +SYSCTL_INT(_dev_netmap, OID_AUTO, ix_crcstrip, + CTLFLAG_RW, &ix_crcstrip, 0, "strip CRC on rx frames"); + /* ** Defining this on will allow the use ** of unsupported SFP+ modules, note that @@ -351,8 +387,10 @@ static int allow_unsupported_sfp = FALSE; TUNABLE_INT("hw.ix.unsupported_sfp", &allow_unsupported_sfp); +#if 0 /* Keep running tab on them for sanity check */ static int ixgbe_total_ports; +#endif #ifdef IXGBE_FDIR /* @@ -365,2158 +403,1591 @@ static int fdir_pballoc = 1; #endif -#ifdef DEV_NETMAP -/* - * The #ifdef DEV_NETMAP / #endif blocks in this file are meant to - * be a reference on how to implement netmap support in a driver. - * Additional comments are in ixgbe_netmap.h . - * - * contains functions for netmap support - * that extend the standard driver. - */ -#include -#endif /* DEV_NETMAP */ - static MALLOC_DEFINE(M_IXGBE, "ix", "ix driver allocations"); -/********************************************************************* - * Device identification routine - * - * ixgbe_probe determines if the driver should be loaded on - * adapter based on PCI vendor/device id of the adapter. - * - * return BUS_PROBE_DEFAULT on success, positive on failure - *********************************************************************/ - -static int -ixgbe_probe(device_t dev) -{ - ixgbe_vendor_info_t *ent; - - u16 pci_vendor_id = 0; - u16 pci_device_id = 0; - u16 pci_subvendor_id = 0; - u16 pci_subdevice_id = 0; - char adapter_name[256]; - - INIT_DEBUGOUT("ixgbe_probe: begin"); - - pci_vendor_id = pci_get_vendor(dev); - if (pci_vendor_id != IXGBE_INTEL_VENDOR_ID) - return (ENXIO); - - pci_device_id = pci_get_device(dev); - pci_subvendor_id = pci_get_subvendor(dev); - pci_subdevice_id = pci_get_subdevice(dev); - - ent = ixgbe_vendor_info_array; - while (ent->vendor_id != 0) { - if ((pci_vendor_id == ent->vendor_id) && - (pci_device_id == ent->device_id) && - - ((pci_subvendor_id == ent->subvendor_id) || - (ent->subvendor_id == 0)) && - - ((pci_subdevice_id == ent->subdevice_id) || - (ent->subdevice_id == 0))) { - sprintf(adapter_name, "%s, Version - %s", - ixgbe_strings[ent->index], - ixgbe_driver_version); - device_set_desc_copy(dev, adapter_name); - ++ixgbe_total_ports; - return (BUS_PROBE_DEFAULT); - } - ent++; - } - return (ENXIO); -} +extern struct if_txrx ixgbe_txrx; + +static struct if_shared_ctx ixgbe_sctx_init = { + .isc_magic = IFLIB_MAGIC, + .isc_q_align = PAGE_SIZE,/* max(DBA_ALIGN, PAGE_SIZE) */ + .isc_tx_maxsize = IXGBE_TSO_SIZE, + + .isc_tx_maxsegsize = PAGE_SIZE*4, + + .isc_rx_maxsize = PAGE_SIZE*4, + .isc_rx_nsegments = 1, + .isc_rx_maxsegsize = PAGE_SIZE*4, + .isc_ntxd = DEFAULT_TXD, + .isc_nrxd = DEFAULT_RXD, + .isc_nfl = 1, + .isc_qsizes[0] = roundup2((DEFAULT_TXD * sizeof(union ixgbe_adv_tx_desc)) + + sizeof(u32), DBA_ALIGN), + .isc_qsizes[1] = roundup2(DEFAULT_RXD * + sizeof(union ixgbe_adv_rx_desc), DBA_ALIGN), + .isc_nqs = 2, + + .isc_admin_intrcnt = 1, + .isc_vendor_info = ixgbe_vendor_info_array, + .isc_driver_version = ixgbe_driver_version, + .isc_txrx = &ixgbe_txrx, + .isc_driver = &ixgbe_if_driver, +}; -/********************************************************************* - * Device initialization routine - * - * The attach entry point is called when the driver is being loaded. - * This routine identifies the type of hardware, allocates all resources - * and initializes the hardware. - * - * return 0 on success, positive on failure - *********************************************************************/ +if_shared_ctx_t ixgbe_sctx = &ixgbe_sctx_init; static int -ixgbe_attach(device_t dev) +ixgbe_if_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nqs) { - struct adapter *adapter; - struct ixgbe_hw *hw; - int error = 0; - u16 csum; - u32 ctrl_ext; - - INIT_DEBUGOUT("ixgbe_attach: begin"); - - /* Allocate, clear, and link in our adapter structure */ - adapter = device_get_softc(dev); - adapter->dev = dev; - hw = &adapter->hw; - -#ifdef DEV_NETMAP - adapter->init_locked = ixgbe_init_locked; - adapter->stop_locked = ixgbe_stop; -#endif - - /* Core Lock Init*/ - IXGBE_CORE_LOCK_INIT(adapter, device_get_nameunit(dev)); - - /* Set up the timer callout */ - callout_init_mtx(&adapter->timer, &adapter->core_mtx, 0); - - /* Determine hardware revision */ - ixgbe_identify_hardware(adapter); - - /* Do base PCI setup - map BAR0 */ - if (ixgbe_allocate_pci_resources(adapter)) { - device_printf(dev, "Allocation of PCI resources failed\n"); - error = ENXIO; - goto err_out; - } - - /* Sysctls for limiting the amount of work done in the taskqueues */ - ixgbe_set_sysctl_value(adapter, "rx_processing_limit", - "max number of rx packets to process", - &adapter->rx_process_limit, ixgbe_rx_process_limit); - - ixgbe_set_sysctl_value(adapter, "tx_processing_limit", - "max number of tx packets to process", - &adapter->tx_process_limit, ixgbe_tx_process_limit); + struct adapter *adapter = iflib_get_softc(ctx); + struct ix_queue *que; + int i, error; - /* Do descriptor calc and sanity checks */ - if (((ixgbe_txd * sizeof(union ixgbe_adv_tx_desc)) % DBA_ALIGN) != 0 || - ixgbe_txd < MIN_TXD || ixgbe_txd > MAX_TXD) { - device_printf(dev, "TXD config issue, using default!\n"); - adapter->num_tx_desc = DEFAULT_TXD; - } else - adapter->num_tx_desc = ixgbe_txd; + MPASS(adapter->num_queues > 0); + MPASS(nqs == 2); - /* - ** With many RX rings it is easy to exceed the - ** system mbuf allocation. Tuning nmbclusters - ** can alleviate this. - */ - if (nmbclusters > 0) { - int s; - s = (ixgbe_rxd * adapter->num_queues) * ixgbe_total_ports; - if (s > nmbclusters) { - device_printf(dev, "RX Descriptors exceed " - "system mbuf max, using default instead!\n"); - ixgbe_rxd = DEFAULT_RXD; - } + /* Allocate queue structure memory */ + if (!(adapter->queues = + (struct ix_queue *) malloc(sizeof(struct ix_queue) * + adapter->num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) { + device_printf(iflib_get_dev(ctx), "Unable to allocate TX ring memory\n"); + return (ENOMEM); } - if (((ixgbe_rxd * sizeof(union ixgbe_adv_rx_desc)) % DBA_ALIGN) != 0 || - ixgbe_rxd < MIN_RXD || ixgbe_rxd > MAX_RXD) { - device_printf(dev, "RXD config issue, using default!\n"); - adapter->num_rx_desc = DEFAULT_RXD; - } else - adapter->num_rx_desc = ixgbe_rxd; +#ifdef PCI_IOV + iov_mode = ixgbe_get_iov_mode(adapter); + adapter->pool = ixgbe_max_vfs(iov_mode); +#else + adapter->pool = 0; +#endif - /* Allocate our TX/RX Queues */ - if (ixgbe_allocate_queues(adapter)) { - error = ENOMEM; - goto err_out; - } + for (i = 0, que = adapter->queues; i < adapter->num_queues; i++, que++) { + struct tx_ring *txr = &que->txr; + struct rx_ring *rxr = &que->rxr; - /* Allocate multicast array memory. */ - adapter->mta = malloc(sizeof(*adapter->mta) * - MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT); - if (adapter->mta == NULL) { - device_printf(dev, "Can not allocate multicast setup array\n"); + if (!(txr->tx_buffers = (struct ixgbe_tx_buf *) malloc(sizeof(struct ixgbe_tx_buf) * ixgbe_sctx->isc_ntxd, M_DEVBUF, M_NOWAIT | M_ZERO))) { + device_printf(iflib_get_dev(ctx), "failed to allocate tx_buffer memory\n"); error = ENOMEM; - goto err_late; - } - - /* Initialize the shared code */ - hw->allow_unsupported_sfp = allow_unsupported_sfp; - error = ixgbe_init_shared_code(hw); - if (error == IXGBE_ERR_SFP_NOT_PRESENT) { - /* - ** No optics in this port, set up - ** so the timer routine will probe - ** for later insertion. - */ - adapter->sfp_probe = TRUE; - error = 0; - } else if (error == IXGBE_ERR_SFP_NOT_SUPPORTED) { - device_printf(dev, "Unsupported SFP+ module detected!\n"); - error = EIO; - goto err_late; - } else if (error) { - device_printf(dev, "Unable to initialize the shared code\n"); - error = EIO; - goto err_late; - } - - /* Make sure we have a good EEPROM before we read from it */ - if (ixgbe_validate_eeprom_checksum(&adapter->hw, &csum) < 0) { - device_printf(dev, "The EEPROM Checksum Is Not Valid\n"); - error = EIO; - goto err_late; - } - - error = ixgbe_init_hw(hw); - switch (error) { - case IXGBE_ERR_EEPROM_VERSION: - device_printf(dev, "This device is a pre-production adapter/" - "LOM. Please be aware there may be issues associated " - "with your hardware.\nIf you are experiencing problems " - "please contact your Intel or hardware representative " - "who provided you with this hardware.\n"); - break; - case IXGBE_ERR_SFP_NOT_SUPPORTED: - device_printf(dev, "Unsupported SFP+ Module\n"); - error = EIO; - goto err_late; - case IXGBE_ERR_SFP_NOT_PRESENT: - device_printf(dev, "No SFP+ Module found\n"); - /* falls thru */ - default: - break; - } - - /* hw.ix defaults init */ - ixgbe_set_advertise(adapter, ixgbe_advertise_speed); - ixgbe_set_flowcntl(adapter, ixgbe_flow_control); - adapter->enable_aim = ixgbe_enable_aim; - - if ((adapter->msix > 1) && (ixgbe_enable_msix)) - error = ixgbe_allocate_msix(adapter); - else - error = ixgbe_allocate_legacy(adapter); - if (error) - goto err_late; - - /* Enable the optics for 82599 SFP+ fiber */ - ixgbe_enable_tx_laser(hw); - - /* Enable power to the phy. */ - ixgbe_set_phy_power(hw, TRUE); - - /* Setup OS specific network interface */ - if (ixgbe_setup_interface(dev, adapter) != 0) - goto err_late; - - /* Initialize statistics */ - ixgbe_update_stats_counters(adapter); + goto fail; + } +#ifdef PCI_IOV + que->me = txr->me = rxr->me = ixgbe_pf_que_index(iov_mode, i); +#else + que->me = txr->me = rxr->me = i; +#endif - /* Register for VLAN events */ - adapter->vlan_attach = EVENTHANDLER_REGISTER(vlan_config, - ixgbe_register_vlan, adapter, EVENTHANDLER_PRI_FIRST); - adapter->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig, - ixgbe_unregister_vlan, adapter, EVENTHANDLER_PRI_FIRST); + txr->adapter = rxr->adapter = que->adapter = adapter; + adapter->active_queues |= (u64)1 << que->me; - /* Check PCIE slot type/speed/width */ - ixgbe_get_slot_info(adapter); + /* get the virtual and physical address of the hardware queues */ + txr->tail = IXGBE_TDT(que->me); + txr->tx_base = (union ixgbe_adv_tx_desc *)vaddrs[i*2]; + txr->tx_paddr = paddrs[i*2]; - /* Set an initial default flow control & dmac value */ - adapter->fc = ixgbe_fc_full; - adapter->dmac = 0; - adapter->eee_enabled = 0; + rxr->tail = IXGBE_RDT(que->me); + rxr->rx_base = (union ixgbe_adv_rx_desc *)vaddrs[i*2 + 1]; + rxr->rx_paddr = paddrs[i*2 + 1]; + rxr->bytes = 0; + txr->que = rxr->que = que; + txr->tx_buffers->eop = NULL; + txr->bytes = 0; + txr->total_packets = 0; -#ifdef PCI_IOV - if ((hw->mac.type != ixgbe_mac_82598EB) && (adapter->msix > 1)) { - nvlist_t *pf_schema, *vf_schema; +#ifdef IXGBE_FDIR + /* Set the rate at which we sample packets */ + if (adapter->hw.mac.type != ixgbe_mac_82598EB) + txr->atr_sample = atr_sample_rate; +#endif - hw->mbx.ops.init_params(hw); - pf_schema = pci_iov_schema_alloc_node(); - vf_schema = pci_iov_schema_alloc_node(); - pci_iov_schema_add_unicast_mac(vf_schema, "mac-addr", 0, NULL); - pci_iov_schema_add_bool(vf_schema, "mac-anti-spoof", - IOV_SCHEMA_HASDEFAULT, TRUE); - pci_iov_schema_add_bool(vf_schema, "allow-set-mac", - IOV_SCHEMA_HASDEFAULT, FALSE); - pci_iov_schema_add_bool(vf_schema, "allow-promisc", - IOV_SCHEMA_HASDEFAULT, FALSE); - error = pci_iov_attach(dev, pf_schema, vf_schema); - if (error != 0) { - device_printf(dev, - "Error %d setting up SR-IOV\n", error); - } + txr->tx_avail = ixgbe_sctx->isc_ntxd; } -#endif /* PCI_IOV */ - - /* Check for certain supported features */ - ixgbe_check_wol_support(adapter); - - /* Add sysctls */ - ixgbe_add_device_sysctls(adapter); - ixgbe_add_hw_stats(adapter); - /* let hardware know driver is loaded */ - ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT); - ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD; - IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext); + iflib_config_gtask_init(ctx, &adapter->mod_task, ixgbe_handle_mod, "mod_task"); + iflib_config_gtask_init(ctx, &adapter->msf_task, ixgbe_handle_msf, "msf_task"); + iflib_config_gtask_init(ctx, &adapter->phy_task, ixgbe_handle_phy, "phy_task"); -#ifdef DEV_NETMAP - ixgbe_netmap_attach(adapter); -#endif /* DEV_NETMAP */ - INIT_DEBUGOUT("ixgbe_attach: end"); + device_printf(iflib_get_dev(ctx), "allocated for %d queues\n", adapter->num_queues); return (0); -err_late: - ixgbe_free_transmit_structures(adapter); - ixgbe_free_receive_structures(adapter); -err_out: - if (adapter->ifp != NULL) - if_free(adapter->ifp); - ixgbe_free_pci_resources(adapter); - free(adapter->mta, M_DEVBUF); + fail: + ixgbe_if_queues_free(ctx); return (error); } -/********************************************************************* - * Device removal routine - * - * The detach entry point is called when the driver is being removed. - * This routine stops the adapter and deallocates all the resources - * that were allocated for driver operation. - * - * return 0 on success, positive on failure - *********************************************************************/ - -static int -ixgbe_detach(device_t dev) +static void +ixgbe_if_queues_free(if_ctx_t ctx) { - struct adapter *adapter = device_get_softc(dev); + struct adapter *adapter = iflib_get_softc(ctx); struct ix_queue *que = adapter->queues; - struct tx_ring *txr = adapter->tx_rings; - u32 ctrl_ext; + int i; - INIT_DEBUGOUT("ixgbe_detach: begin"); + if (que == NULL) + return; - /* Make sure VLANS are not using driver */ - if (adapter->ifp->if_vlantrunk != NULL) { - device_printf(dev,"Vlan in use, detach first\n"); - return (EBUSY); - } + for (i = 0; i < adapter->num_queues; i++, que++) { + struct tx_ring *txr = &que->txr; + if (txr->tx_buffers == NULL) + break; -#ifdef PCI_IOV - if (pci_iov_detach(dev) != 0) { - device_printf(dev, "SR-IOV in use; detach first.\n"); - return (EBUSY); + free(txr->tx_buffers, M_DEVBUF); + txr->tx_buffers = NULL; } -#endif /* PCI_IOV */ - - ether_ifdetach(adapter->ifp); - /* Stop the adapter */ - IXGBE_CORE_LOCK(adapter); - ixgbe_setup_low_power_mode(adapter); - IXGBE_CORE_UNLOCK(adapter); + + free(adapter->queues, M_DEVBUF); + adapter->queues = NULL; +} - for (int i = 0; i < adapter->num_queues; i++, que++, txr++) { - if (que->tq) { -#ifndef IXGBE_LEGACY_TX - taskqueue_drain(que->tq, &txr->txq_task); +static void +ixgbe_initialise_rss_mapping(struct adapter *adapter) +{ + struct ixgbe_hw *hw = &adapter->hw; + u32 reta = 0, mrqc, rss_key[10]; + int queue_id, table_size, index_mult; +#ifdef RSS + u32 rss_hash_config; #endif - taskqueue_drain(que->tq, &que->que_task); - taskqueue_free(que->tq); - } - } - - /* Drain the Link queue */ - if (adapter->tq) { - taskqueue_drain(adapter->tq, &adapter->link_task); - taskqueue_drain(adapter->tq, &adapter->mod_task); - taskqueue_drain(adapter->tq, &adapter->msf_task); #ifdef PCI_IOV - taskqueue_drain(adapter->tq, &adapter->mbx_task); + enum ixgbe_iov_mode mode; #endif - taskqueue_drain(adapter->tq, &adapter->phy_task); -#ifdef IXGBE_FDIR - taskqueue_drain(adapter->tq, &adapter->fdir_task); + +#ifdef RSS + /* Fetch the configured RSS key */ + rss_getkey((uint8_t *) &rss_key); +#else + /* set up random bits */ + arc4rand(&rss_key, sizeof(rss_key), 0); #endif - taskqueue_free(adapter->tq); + + /* Set multiplier for RETA setup and table size based on MAC */ + index_mult = 0x1; + table_size = 128; + switch (adapter->hw.mac.type) { + case ixgbe_mac_82598EB: + index_mult = 0x11; + break; + case ixgbe_mac_X550: + case ixgbe_mac_X550EM_x: + table_size = 512; + break; + default: + break; } - /* let hardware know driver is unloading */ - ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT); - ctrl_ext &= ~IXGBE_CTRL_EXT_DRV_LOAD; - IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT, ctrl_ext); - - /* Unregister VLAN events */ - if (adapter->vlan_attach != NULL) - EVENTHANDLER_DEREGISTER(vlan_config, adapter->vlan_attach); - if (adapter->vlan_detach != NULL) - EVENTHANDLER_DEREGISTER(vlan_unconfig, adapter->vlan_detach); - - callout_drain(&adapter->timer); -#ifdef DEV_NETMAP - netmap_detach(adapter->ifp); -#endif /* DEV_NETMAP */ - ixgbe_free_pci_resources(adapter); - bus_generic_detach(dev); - if_free(adapter->ifp); + /* Set up the redirection table */ + for (int i = 0, j = 0; i < table_size; i++, j++) { + if (j == adapter->num_queues) j = 0; +#ifdef RSS + /* + * Fetch the RSS bucket id for the given indirection entry. + * Cap it at the number of configured buckets (which is + * num_queues.) + */ + queue_id = rss_get_indirection_to_bucket(i); + queue_id = queue_id % adapter->num_queues; +#else + queue_id = (j * index_mult); +#endif + /* + * The low 8 bits are for hash value (n+0); + * The next 8 bits are for hash value (n+1), etc. + */ + reta = reta >> 8; + reta = reta | ( ((uint32_t) queue_id) << 24); + if ((i & 3) == 3) { + if (i < 128) + IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta); + else + IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32), reta); + reta = 0; + } + } - ixgbe_free_transmit_structures(adapter); - ixgbe_free_receive_structures(adapter); - free(adapter->mta, M_DEVBUF); + /* Now fill our hash function seeds */ + for (int i = 0; i < 10; i++) + IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), rss_key[i]); - IXGBE_CORE_LOCK_DESTROY(adapter); - return (0); + /* Perform hash on these packet types */ +#ifdef RSS + mrqc = IXGBE_MRQC_RSSEN; + rss_hash_config = rss_gethashconfig(); + if (rss_hash_config & RSS_HASHTYPE_RSS_IPV4) + mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4; + if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV4) + mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_TCP; + if (rss_hash_config & RSS_HASHTYPE_RSS_IPV6) + mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6; + if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV6) + mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_TCP; + if (rss_hash_config & RSS_HASHTYPE_RSS_IPV6_EX) + mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX; + if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV6_EX) + mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX_TCP; + if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV4) + mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP; + if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV4_EX) + device_printf(adapter->dev, + "%s: RSS_HASHTYPE_RSS_UDP_IPV4_EX defined, " + "but not supported\n", __func__); + if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV6) + mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP; + if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV6_EX) + mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX_UDP; +#else + /* + * Disable UDP - IP fragments aren't currently being handled + * and so we end up with a mix of 2-tuple and 4-tuple + * traffic. + */ + mrqc = IXGBE_MRQC_RSSEN + | IXGBE_MRQC_RSS_FIELD_IPV4 + | IXGBE_MRQC_RSS_FIELD_IPV4_TCP + | IXGBE_MRQC_RSS_FIELD_IPV6_EX_TCP + | IXGBE_MRQC_RSS_FIELD_IPV6_EX + | IXGBE_MRQC_RSS_FIELD_IPV6 + | IXGBE_MRQC_RSS_FIELD_IPV6_TCP + ; +#endif /* RSS */ +#ifdef PCI_IOV + mode = ixgbe_get_iov_mode(adapter); + mrqc |= ixgbe_get_mrqc(mode); +#endif + IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc); } + /********************************************************************* * - * Shutdown entry point + * Setup receive registers and features. * **********************************************************************/ +#define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2 -static int -ixgbe_shutdown(device_t dev) +#define BSIZEPKT_ROUNDUP ((1<hw; + struct ifnet *ifp = iflib_get_ifp(ctx); - INIT_DEBUGOUT("ixgbe_shutdown: begin"); + struct ix_queue *que; + u32 bufsz, fctrl, srrctl, rxcsum; + u32 hlreg; + int i; - IXGBE_CORE_LOCK(adapter); - error = ixgbe_setup_low_power_mode(adapter); - IXGBE_CORE_UNLOCK(adapter); + /* + * Make sure receives are disabled while + * setting up the descriptor ring + */ + ixgbe_disable_rx(hw); - return (error); -} + /* Enable broadcasts */ + fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL); + fctrl |= IXGBE_FCTRL_BAM; + if (adapter->hw.mac.type == ixgbe_mac_82598EB) { + fctrl |= IXGBE_FCTRL_DPF; + fctrl |= IXGBE_FCTRL_PMCF; + } + IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl); -/** - * Methods for going from: - * D0 -> D3: ixgbe_suspend - * D3 -> D0: ixgbe_resume - */ -static int -ixgbe_suspend(device_t dev) -{ - struct adapter *adapter = device_get_softc(dev); - int error = 0; + /* Set for Jumbo Frames? */ + hlreg = IXGBE_READ_REG(hw, IXGBE_HLREG0); + if (ifp->if_mtu > ETHERMTU) + hlreg |= IXGBE_HLREG0_JUMBOEN; + else + hlreg &= ~IXGBE_HLREG0_JUMBOEN; +#ifdef DEV_NETMAP + /* crcstrip is conditional in netmap (in RDRXCTL too ?) */ + if (ifp->if_capenable & IFCAP_NETMAP && !ix_crcstrip) + hlreg &= ~IXGBE_HLREG0_RXCRCSTRP; + else + hlreg |= IXGBE_HLREG0_RXCRCSTRP; +#endif /* DEV_NETMAP */ + IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg); - INIT_DEBUGOUT("ixgbe_suspend: begin"); + bufsz = (adapter->rx_mbuf_sz + + BSIZEPKT_ROUNDUP) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT; - IXGBE_CORE_LOCK(adapter); + /* Setup the Base and Length of the Rx Descriptor Ring */ + for (i = 0, que = adapter->queues; i < adapter->num_queues; i++, que++) { + struct rx_ring *rxr = &que->rxr; + u64 rdba = rxr->rx_paddr; + int j = rxr->me; - error = ixgbe_setup_low_power_mode(adapter); + /* Setup the Base and Length of the Rx Descriptor Ring */ + IXGBE_WRITE_REG(hw, IXGBE_RDBAL(j), + (rdba & 0x00000000ffffffffULL)); + IXGBE_WRITE_REG(hw, IXGBE_RDBAH(j), (rdba >> 32)); + IXGBE_WRITE_REG(hw, IXGBE_RDLEN(j), + ixgbe_sctx->isc_nrxd * sizeof(union ixgbe_adv_rx_desc)); - IXGBE_CORE_UNLOCK(adapter); + /* Set up the SRRCTL register */ + srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(j)); + srrctl &= ~IXGBE_SRRCTL_BSIZEHDR_MASK; + srrctl &= ~IXGBE_SRRCTL_BSIZEPKT_MASK; + srrctl |= bufsz; + srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF; - return (error); -} + /* + * Set DROP_EN iff we have no flow control and >1 queue. + * Note that srrctl was cleared shortly before during reset, + * so we do not need to clear the bit, but do it just in case + * this code is moved elsewhere. + */ + if (adapter->num_queues > 1 && + adapter->hw.fc.requested_mode == ixgbe_fc_none) { + srrctl |= IXGBE_SRRCTL_DROP_EN; + } else { + srrctl &= ~IXGBE_SRRCTL_DROP_EN; + } -static int -ixgbe_resume(device_t dev) -{ - struct adapter *adapter = device_get_softc(dev); - struct ifnet *ifp = adapter->ifp; - struct ixgbe_hw *hw = &adapter->hw; - u32 wus; + IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(j), srrctl); - INIT_DEBUGOUT("ixgbe_resume: begin"); + /* Setup the HW Rx Head and Tail Descriptor Pointers */ + IXGBE_WRITE_REG(hw, IXGBE_RDH(j), 0); + IXGBE_WRITE_REG(hw, IXGBE_RDT(j), 0); - IXGBE_CORE_LOCK(adapter); + /* Set the driver rx tail address */ + rxr->tail = IXGBE_RDT(rxr->me); + } - /* Read & clear WUS register */ - wus = IXGBE_READ_REG(hw, IXGBE_WUS); - if (wus) - device_printf(dev, "Woken up by (WUS): %#010x\n", - IXGBE_READ_REG(hw, IXGBE_WUS)); - IXGBE_WRITE_REG(hw, IXGBE_WUS, 0xffffffff); - /* And clear WUFC until next low-power transition */ - IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0); + if (adapter->hw.mac.type != ixgbe_mac_82598EB) { + u32 psrtype = IXGBE_PSRTYPE_TCPHDR | + IXGBE_PSRTYPE_UDPHDR | + IXGBE_PSRTYPE_IPV4HDR | + IXGBE_PSRTYPE_IPV6HDR; + IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(0), psrtype); + } - /* - * Required after D3->D0 transition; - * will re-advertise all previous advertised speeds - */ - if (ifp->if_flags & IFF_UP) - ixgbe_init_locked(adapter); + rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM); - IXGBE_CORE_UNLOCK(adapter); + ixgbe_initialise_rss_mapping(adapter); - return (0); + if (adapter->num_queues > 1) { + /* RSS and RX IPP Checksum are mutually exclusive */ + rxcsum |= IXGBE_RXCSUM_PCSD; + } + + if (ifp->if_capenable & IFCAP_RXCSUM) + rxcsum |= IXGBE_RXCSUM_PCSD; + + if (!(rxcsum & IXGBE_RXCSUM_PCSD)) + rxcsum |= IXGBE_RXCSUM_IPPCSE; + + IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum); } + /********************************************************************* - * Ioctl entry point * - * ixgbe_ioctl is called when the user wants to configure the - * interface. + * Enable transmit units. * - * return 0 on success, positive on failure **********************************************************************/ - -static int -ixgbe_ioctl(struct ifnet * ifp, u_long command, caddr_t data) +static void +ixgbe_initialize_transmit_units(if_ctx_t ctx) { - struct adapter *adapter = ifp->if_softc; - struct ifreq *ifr = (struct ifreq *) data; -#if defined(INET) || defined(INET6) - struct ifaddr *ifa = (struct ifaddr *)data; -#endif - int error = 0; - bool avoid_reset = FALSE; + struct adapter *adapter = iflib_get_softc(ctx); + struct ixgbe_hw *hw = &adapter->hw; + struct ix_queue *que; + int i; + + /* Setup the Base and Length of the Tx Descriptor Ring */ + for (i = 0, que = adapter->queues; i < adapter->num_queues; i++, que++) { + struct tx_ring *txr = &que->txr; + u64 tdba = txr->tx_paddr; + u32 txctrl = 0; + int j = txr->me; - switch (command) { + IXGBE_WRITE_REG(hw, IXGBE_TDBAL(j), + (tdba & 0x00000000ffffffffULL)); + IXGBE_WRITE_REG(hw, IXGBE_TDBAH(j), (tdba >> 32)); + IXGBE_WRITE_REG(hw, IXGBE_TDLEN(j), + ixgbe_sctx->isc_ntxd * sizeof(union ixgbe_adv_tx_desc)); - case SIOCSIFADDR: -#ifdef INET - if (ifa->ifa_addr->sa_family == AF_INET) - avoid_reset = TRUE; -#endif -#ifdef INET6 - if (ifa->ifa_addr->sa_family == AF_INET6) - avoid_reset = TRUE; -#endif - /* - ** Calling init results in link renegotiation, - ** so we avoid doing it when possible. - */ - if (avoid_reset) { - ifp->if_flags |= IFF_UP; - if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) - ixgbe_init(adapter); -#ifdef INET - if (!(ifp->if_flags & IFF_NOARP)) - arp_ifinit(ifp, ifa); -#endif - } else - error = ether_ioctl(ifp, command, data); - break; - case SIOCSIFMTU: - IOCTL_DEBUGOUT("ioctl: SIOCSIFMTU (Set Interface MTU)"); - if (ifr->ifr_mtu > IXGBE_MAX_MTU) { - error = EINVAL; - } else { - IXGBE_CORE_LOCK(adapter); - ifp->if_mtu = ifr->ifr_mtu; - adapter->max_frame_size = - ifp->if_mtu + IXGBE_MTU_HDR; - ixgbe_init_locked(adapter); -#ifdef PCI_IOV - ixgbe_recalculate_max_frame(adapter); -#endif - IXGBE_CORE_UNLOCK(adapter); - } - break; - case SIOCSIFFLAGS: - IOCTL_DEBUGOUT("ioctl: SIOCSIFFLAGS (Set Interface Flags)"); - IXGBE_CORE_LOCK(adapter); - if (ifp->if_flags & IFF_UP) { - if ((ifp->if_drv_flags & IFF_DRV_RUNNING)) { - if ((ifp->if_flags ^ adapter->if_flags) & - (IFF_PROMISC | IFF_ALLMULTI)) { - ixgbe_set_promisc(adapter); - } - } else - ixgbe_init_locked(adapter); - } else - if (ifp->if_drv_flags & IFF_DRV_RUNNING) - ixgbe_stop(adapter); - adapter->if_flags = ifp->if_flags; - IXGBE_CORE_UNLOCK(adapter); - break; - case SIOCADDMULTI: - case SIOCDELMULTI: - IOCTL_DEBUGOUT("ioctl: SIOC(ADD|DEL)MULTI"); - if (ifp->if_drv_flags & IFF_DRV_RUNNING) { - IXGBE_CORE_LOCK(adapter); - ixgbe_disable_intr(adapter); - ixgbe_set_multi(adapter); - ixgbe_enable_intr(adapter); - IXGBE_CORE_UNLOCK(adapter); - } - break; - case SIOCSIFMEDIA: - case SIOCGIFMEDIA: - IOCTL_DEBUGOUT("ioctl: SIOCxIFMEDIA (Get/Set Interface Media)"); - error = ifmedia_ioctl(ifp, ifr, &adapter->media, command); - break; - case SIOCSIFCAP: - { - IOCTL_DEBUGOUT("ioctl: SIOCSIFCAP (Set Capabilities)"); + /* Setup the HW Tx Head and Tail descriptor pointers */ + IXGBE_WRITE_REG(hw, IXGBE_TDH(j), 0); + IXGBE_WRITE_REG(hw, IXGBE_TDT(j), 0); - int mask = ifr->ifr_reqcap ^ ifp->if_capenable; - if (!mask) + /* Disable Head Writeback */ + switch (hw->mac.type) { + case ixgbe_mac_82598EB: + txctrl = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL(j)); break; - - /* HW cannot turn these on/off separately */ - if (mask & (IFCAP_RXCSUM | IFCAP_RXCSUM_IPV6)) { - ifp->if_capenable ^= IFCAP_RXCSUM; - ifp->if_capenable ^= IFCAP_RXCSUM_IPV6; - } - if (mask & IFCAP_TXCSUM) - ifp->if_capenable ^= IFCAP_TXCSUM; - if (mask & IFCAP_TXCSUM_IPV6) - ifp->if_capenable ^= IFCAP_TXCSUM_IPV6; - if (mask & IFCAP_TSO4) - ifp->if_capenable ^= IFCAP_TSO4; - if (mask & IFCAP_TSO6) - ifp->if_capenable ^= IFCAP_TSO6; - if (mask & IFCAP_LRO) - ifp->if_capenable ^= IFCAP_LRO; - if (mask & IFCAP_VLAN_HWTAGGING) - ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING; - if (mask & IFCAP_VLAN_HWFILTER) - ifp->if_capenable ^= IFCAP_VLAN_HWFILTER; - if (mask & IFCAP_VLAN_HWTSO) - ifp->if_capenable ^= IFCAP_VLAN_HWTSO; - - if (ifp->if_drv_flags & IFF_DRV_RUNNING) { - IXGBE_CORE_LOCK(adapter); - ixgbe_init_locked(adapter); - IXGBE_CORE_UNLOCK(adapter); - } - VLAN_CAPABILITIES(ifp); - break; - } -#if __FreeBSD_version >= 1100036 - case SIOCGI2C: - { - struct ixgbe_hw *hw = &adapter->hw; - struct ifi2creq i2c; - int i; - IOCTL_DEBUGOUT("ioctl: SIOCGI2C (Get I2C Data)"); - error = copyin(ifr->ifr_data, &i2c, sizeof(i2c)); - if (error != 0) + case ixgbe_mac_82599EB: + case ixgbe_mac_X540: + default: + txctrl = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL_82599(j)); break; - if (i2c.dev_addr != 0xA0 && i2c.dev_addr != 0xA2) { - error = EINVAL; + } + txctrl &= ~IXGBE_DCA_TXCTRL_DESC_WRO_EN; + switch (hw->mac.type) { + case ixgbe_mac_82598EB: + IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL(j), txctrl); break; - } - if (i2c.len > sizeof(i2c.data)) { - error = EINVAL; + case ixgbe_mac_82599EB: + case ixgbe_mac_X540: + default: + IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL_82599(j), txctrl); break; } - for (i = 0; i < i2c.len; i++) - hw->phy.ops.read_i2c_byte(hw, i2c.offset + i, - i2c.dev_addr, &i2c.data[i]); - error = copyout(&i2c, ifr->ifr_data, sizeof(i2c)); - break; } + + if (hw->mac.type != ixgbe_mac_82598EB) { + u32 dmatxctl, rttdcs; +#ifdef PCI_IOV + enum ixgbe_iov_mode mode = ixgbe_get_iov_mode(adapter); #endif - default: - IOCTL_DEBUGOUT1("ioctl: UNKNOWN (0x%X)\n", (int)command); - error = ether_ioctl(ifp, command, data); - break; + dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL); + dmatxctl |= IXGBE_DMATXCTL_TE; + IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl); + /* Disable arbiter to set MTQC */ + rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS); + rttdcs |= IXGBE_RTTDCS_ARBDIS; + IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs); +#ifdef PCI_IOV + IXGBE_WRITE_REG(hw, IXGBE_MTQC, ixgbe_get_mtqc(mode)); +#else + IXGBE_WRITE_REG(hw, IXGBE_MTQC, IXGBE_MTQC_64Q_1PB); +#endif + rttdcs &= ~IXGBE_RTTDCS_ARBDIS; + IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs); } - - return (error); } -/* - * Set the various hardware offload abilities. - * - * This takes the ifnet's if_capenable flags (e.g. set by the user using - * ifconfig) and indicates to the OS via the ifnet's if_hwassist field what - * mbuf offload flags the driver will understand. - */ -static void -ixgbe_set_if_hwassist(struct adapter *adapter) -{ - struct ifnet *ifp = adapter->ifp; - struct ixgbe_hw *hw = &adapter->hw; - ifp->if_hwassist = 0; -#if __FreeBSD_version >= 1000000 - if (ifp->if_capenable & IFCAP_TSO4) - ifp->if_hwassist |= CSUM_IP_TSO; - if (ifp->if_capenable & IFCAP_TSO6) - ifp->if_hwassist |= CSUM_IP6_TSO; - if (ifp->if_capenable & IFCAP_TXCSUM) { - ifp->if_hwassist |= (CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP); - if (hw->mac.type != ixgbe_mac_82598EB) - ifp->if_hwassist |= CSUM_IP_SCTP; - } - if (ifp->if_capenable & IFCAP_TXCSUM_IPV6) { - ifp->if_hwassist |= (CSUM_IP6_UDP | CSUM_IP6_TCP); - if (hw->mac.type != ixgbe_mac_82598EB) - ifp->if_hwassist |= CSUM_IP6_SCTP; - } -#else - if (ifp->if_capenable & IFCAP_TSO) - ifp->if_hwassist |= CSUM_TSO; - if (ifp->if_capenable & IFCAP_TXCSUM) { - ifp->if_hwassist |= (CSUM_TCP | CSUM_UDP); - if (hw->mac.type != ixgbe_mac_82598EB) - ifp->if_hwassist |= CSUM_SCTP; +static void * +ixgbe_register(device_t dev) +{ + /* Do descriptor calc and sanity checks */ + if (((ixgbe_txd * sizeof(union ixgbe_adv_tx_desc)) % DBA_ALIGN) != 0 || + ixgbe_txd < MIN_TXD || ixgbe_txd > MAX_TXD) { + device_printf(dev, "TXD config issue, using default!\n"); + ixgbe_sctx->isc_ntxd = DEFAULT_TXD; + } else { + ixgbe_sctx->isc_ntxd = ixgbe_txd; } -#endif + ixgbe_tx_process_limit = min(ixgbe_sctx->isc_ntxd, ixgbe_txd); + ixgbe_sctx->isc_qsizes[0] = roundup2((ixgbe_sctx->isc_ntxd * sizeof(union ixgbe_adv_tx_desc)) + + sizeof(u32), DBA_ALIGN); + ixgbe_sctx->isc_qsizes[1] = roundup2(ixgbe_rxd * + sizeof(union ixgbe_adv_rx_desc), DBA_ALIGN); + + return (ixgbe_sctx); } -/********************************************************************* - * Init entry point + /********************************************************************* + * Device initialization routine * - * This routine is used in two ways. It is used by the stack as - * init entry point in network interface structure. It is also used - * by the driver as a hw/sw initialization routine to get to a - * consistent state. + * The attach entry point is called when the driver is being loaded. + * This routine identifies the type of hardware, allocates all resources + * and initializes the hardware. * * return 0 on success, positive on failure - **********************************************************************/ -#define IXGBE_MHADD_MFS_SHIFT 16 + *********************************************************************/ -static void -ixgbe_init_locked(struct adapter *adapter) +static int +ixgbe_if_attach_pre(if_ctx_t ctx) { - struct ifnet *ifp = adapter->ifp; - device_t dev = adapter->dev; - struct ixgbe_hw *hw = &adapter->hw; - struct tx_ring *txr; - struct rx_ring *rxr; - u32 txdctl, mhadd; - u32 rxdctl, rxctrl; - int err = 0; -#ifdef PCI_IOV - enum ixgbe_iov_mode mode; -#endif + device_t dev; + struct adapter *adapter; + struct ixgbe_hw *hw; + uint16_t csum; + int error = 0; + + INIT_DEBUGOUT("ixgbe_attach: begin"); - mtx_assert(&adapter->core_mtx, MA_OWNED); - INIT_DEBUGOUT("ixgbe_init_locked: begin"); + /* Allocate, clear, and link in our adapter structure */ + dev = iflib_get_dev(ctx); + adapter = iflib_get_softc(ctx); + adapter->ctx = ctx; + adapter->dev = dev; + adapter->shared = iflib_get_softc_ctx(ctx); + adapter->media = iflib_get_media(ctx); + hw = &adapter->hw; - hw->adapter_stopped = FALSE; - ixgbe_stop_adapter(hw); - callout_stop(&adapter->timer); + /* Identify hardware revision */ + ixgbe_identify_hardware(ctx); -#ifdef PCI_IOV - mode = ixgbe_get_iov_mode(adapter); - adapter->pool = ixgbe_max_vfs(mode); - /* Queue indices may change with IOV mode */ - for (int i = 0; i < adapter->num_queues; i++) { - adapter->rx_rings[i].me = ixgbe_pf_que_index(mode, i); - adapter->tx_rings[i].me = ixgbe_pf_que_index(mode, i); + if (adapter->hw.mac.type == ixgbe_mac_82598EB) { + adapter->shared->isc_tx_nsegments = IXGBE_82598_SCATTER; + adapter->shared->isc_msix_bar = PCIR_BAR(MSIX_82598_BAR); + } else { + hw->phy.smart_speed = ixgbe_smart_speed; + adapter->shared->isc_tx_nsegments = IXGBE_82599_SCATTER; + adapter->shared->isc_msix_bar = PCIR_BAR(MSIX_82599_BAR); } -#endif - /* reprogram the RAR[0] in case user changed it. */ - ixgbe_set_rar(hw, 0, hw->mac.addr, adapter->pool, IXGBE_RAH_AV); - - /* Get the latest mac address, User can use a LAA */ - bcopy(IF_LLADDR(ifp), hw->mac.addr, IXGBE_ETH_LENGTH_OF_ADDRESS); - ixgbe_set_rar(hw, 0, hw->mac.addr, adapter->pool, 1); - hw->addr_ctrl.rar_used_count = 1; - /* Set hardware offload abilities from ifnet flags */ - ixgbe_set_if_hwassist(adapter); + /* Sysctls */ + ixgbe_add_device_sysctls(ctx); - /* Prepare transmit descriptors and buffers */ - if (ixgbe_setup_transmit_structures(adapter)) { - device_printf(dev, "Could not setup transmit structures\n"); - ixgbe_stop(adapter); - return; + /* Do base PCI setup - map BAR0 */ + if (ixgbe_allocate_pci_resources(ctx)) { + device_printf(dev, "Allocation of PCI resources failed\n"); + return (ENXIO); } - ixgbe_init_hw(hw); -#ifdef PCI_IOV - ixgbe_initialize_iov(adapter); -#endif - ixgbe_initialize_transmit_units(adapter); - - /* Setup Multicast table */ - ixgbe_set_multi(adapter); + /* Sysctls for limiting the amount of work done in the taskqueues */ + ixgbe_set_sysctl_value(adapter, "rx_processing_limit", + "max number of rx packets to process", + &adapter->rx_process_limit, ixgbe_rx_process_limit); - /* Determine the correct mbuf pool, based on frame size */ - if (adapter->max_frame_size <= MCLBYTES) - adapter->rx_mbuf_sz = MCLBYTES; - else - adapter->rx_mbuf_sz = MJUMPAGESIZE; + ixgbe_set_sysctl_value(adapter, "tx_processing_limit", + "max number of tx packets to process", + &adapter->tx_process_limit, ixgbe_tx_process_limit); - /* Prepare receive descriptors and buffers */ - if (ixgbe_setup_receive_structures(adapter)) { - device_printf(dev, "Could not setup receive structures\n"); - ixgbe_stop(adapter); - return; + if (((ixgbe_rxd * sizeof(union ixgbe_adv_rx_desc)) % DBA_ALIGN) != 0 || + ixgbe_rxd < MIN_RXD || ixgbe_rxd > MAX_RXD) { + device_printf(dev, "RXD config issue, using default!\n"); + ixgbe_sctx->isc_nrxd = DEFAULT_RXD; + } else { + ixgbe_sctx->isc_nrxd = ixgbe_rxd; } - /* Configure RX settings */ - ixgbe_initialize_receive_units(adapter); - - /* Enable SDP & MSIX interrupts based on adapter */ - ixgbe_config_gpie(adapter); - - /* Set MTU size */ - if (ifp->if_mtu > ETHERMTU) { - /* aka IXGBE_MAXFRS on 82599 and newer */ - mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD); - mhadd &= ~IXGBE_MHADD_MFS_MASK; - mhadd |= adapter->max_frame_size << IXGBE_MHADD_MFS_SHIFT; - IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd); + /* Allocate multicast array memory. */ + adapter->mta = malloc(sizeof(*adapter->mta) * + MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT); + if (adapter->mta == NULL) { + device_printf(dev, "Can not allocate multicast setup array\n"); + error = ENOMEM; + goto err_pci; } - - /* Now enable all the queues */ - for (int i = 0; i < adapter->num_queues; i++) { - txr = &adapter->tx_rings[i]; - txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(txr->me)); - txdctl |= IXGBE_TXDCTL_ENABLE; - /* Set WTHRESH to 8, burst writeback */ - txdctl |= (8 << 16); + + /* Initialize the shared code */ + hw->allow_unsupported_sfp = allow_unsupported_sfp; + error = ixgbe_init_shared_code(hw); + if (error == IXGBE_ERR_SFP_NOT_PRESENT) { /* - * When the internal queue falls below PTHRESH (32), - * start prefetching as long as there are at least - * HTHRESH (1) buffers ready. The values are taken - * from the Intel linux driver 3.8.21. - * Prefetching enables tx line rate even with 1 queue. - */ - txdctl |= (32 << 0) | (1 << 8); - IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(txr->me), txdctl); + ** No optics in this port, set up + ** so the timer routine will probe + ** for later insertion. + */ + adapter->sfp_probe = TRUE; + error = 0; + } else if (error == IXGBE_ERR_SFP_NOT_SUPPORTED) { + device_printf(dev,"Unsupported SFP+ module detected!\n"); + error = EIO; + goto err_late; + } else if (error) { + device_printf(dev,"Unable to initialize the shared code\n"); + error = EIO; + goto err_late; + } + + /* Make sure we have a good EEPROM before we read from it */ + if (ixgbe_validate_eeprom_checksum(&adapter->hw, &csum) < 0) { + device_printf(dev,"The EEPROM Checksum Is Not Valid\n"); + error = EIO; + goto err_late; } - for (int i = 0, j = 0; i < adapter->num_queues; i++) { - rxr = &adapter->rx_rings[i]; - rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me)); - if (hw->mac.type == ixgbe_mac_82598EB) { - /* - ** PTHRESH = 21 - ** HTHRESH = 4 - ** WTHRESH = 8 - */ - rxdctl &= ~0x3FFFFF; - rxdctl |= 0x080420; - } - rxdctl |= IXGBE_RXDCTL_ENABLE; - IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxr->me), rxdctl); - for (; j < 10; j++) { - if (IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me)) & - IXGBE_RXDCTL_ENABLE) - break; - else - msec_delay(1); - } - wmb(); -#ifdef DEV_NETMAP - /* - * In netmap mode, we must preserve the buffers made - * available to userspace before the if_init() - * (this is true by default on the TX side, because - * init makes all buffers available to userspace). - * - * netmap_reset() and the device specific routines - * (e.g. ixgbe_setup_receive_rings()) map these - * buffers at the end of the NIC ring, so here we - * must set the RDT (tail) register to make sure - * they are not overwritten. - * - * In this driver the NIC ring starts at RDH = 0, - * RDT points to the last slot available for reception (?), - * so RDT = num_rx_desc - 1 means the whole ring is available. - */ - if (ifp->if_capenable & IFCAP_NETMAP) { - struct netmap_adapter *na = NA(adapter->ifp); - struct netmap_kring *kring = &na->rx_rings[i]; - int t = na->num_rx_desc - 1 - nm_kr_rxspace(kring); + error = ixgbe_init_hw(hw); - IXGBE_WRITE_REG(hw, IXGBE_RDT(rxr->me), t); - } else -#endif /* DEV_NETMAP */ - IXGBE_WRITE_REG(hw, IXGBE_RDT(rxr->me), adapter->num_rx_desc - 1); + switch (error) { + case IXGBE_ERR_EEPROM_VERSION: + device_printf(dev, "This device is a pre-production adapter/" + "LOM. Please be aware there may be issues associated " + "with your hardware.\n If you are experiencing problems " + "please contact your Intel or hardware representative " + "who provided you with this hardware.\n"); + break; + case IXGBE_ERR_SFP_NOT_SUPPORTED: + device_printf(dev,"Unsupported SFP+ Module\n"); + error = EIO; + goto err_late; + case IXGBE_ERR_SFP_NOT_PRESENT: + device_printf(dev,"No SFP+ Module found\n"); + /* falls thru */ + default: + break; } - /* Enable Receive engine */ - rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL); - if (hw->mac.type == ixgbe_mac_82598EB) - rxctrl |= IXGBE_RXCTRL_DMBYPS; - rxctrl |= IXGBE_RXCTRL_RXEN; - ixgbe_enable_rx_dma(hw, rxctrl); + /* Detect and set physical type */ + ixgbe_setup_optics(adapter); - callout_reset(&adapter->timer, hz, ixgbe_local_timer, adapter); + iflib_set_mac(ctx, hw->mac.addr); + return (0); +err_late: + free(adapter->mta, M_DEVBUF); +err_pci: + ixgbe_free_pci_resources(ctx); - /* Set up MSI/X routing */ - if (ixgbe_enable_msix) { - ixgbe_configure_ivars(adapter); - /* Set up auto-mask */ - if (hw->mac.type == ixgbe_mac_82598EB) - IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE); - else { - IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF); - IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF); - } - } else { /* Simple settings for Legacy/MSI */ - ixgbe_set_ivar(adapter, 0, 0, 0); - ixgbe_set_ivar(adapter, 0, 0, 1); - IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE); - } + return (error); +} -#ifdef IXGBE_FDIR - /* Init Flow director */ - if (hw->mac.type != ixgbe_mac_82598EB) { - u32 hdrm = 32 << fdir_pballoc; +static int +ixgbe_if_attach_post(if_ctx_t ctx) +{ + device_t dev; + struct adapter *adapter; + struct ixgbe_hw *hw; + int error = 0; + u32 ctrl_ext; - hw->mac.ops.setup_rxpba(hw, 0, hdrm, PBA_STRATEGY_EQUAL); - ixgbe_init_fdir_signature_82599(&adapter->hw, fdir_pballoc); - } + dev = iflib_get_dev(ctx); + adapter = iflib_get_softc(ctx); + hw = &adapter->hw; + +#ifdef PCI_IOV + ixgbe_initialize_iov(adapter); #endif + error = ixgbe_interface_setup(ctx); + if (error) { + device_printf(dev, "Interface setup failed: %d\n", error); + goto err; + } - /* - * Check on any SFP devices that - * need to be kick-started - */ - if (hw->phy.type == ixgbe_phy_none) { - err = hw->phy.ops.identify(hw); - if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) { - device_printf(dev, - "Unsupported SFP+ module type was detected.\n"); - return; - } - } - - /* Set moderation on the Link interrupt */ - IXGBE_WRITE_REG(hw, IXGBE_EITR(adapter->vector), IXGBE_LINK_ITR); - - /* Configure Energy Efficient Ethernet for supported devices */ - if (hw->mac.ops.setup_eee) { - err = hw->mac.ops.setup_eee(hw, adapter->eee_enabled); - if (err) - device_printf(dev, "Error setting up EEE: %d\n", err); - } - - /* Enable power to the phy. */ - ixgbe_set_phy_power(hw, TRUE); - - /* Config/Enable Link */ - ixgbe_config_link(adapter); - - /* Hardware Packet Buffer & Flow Control setup */ - ixgbe_config_delay_values(adapter); - - /* Initialize the FC settings */ - ixgbe_start_hw(hw); - - /* Set up VLAN support and filter */ - ixgbe_setup_vlan_hw_support(adapter); - - /* Setup DMA Coalescing */ - ixgbe_config_dmac(adapter); + /* Initialize statistics */ + ixgbe_update_stats_counters(adapter); + ixgbe_add_hw_stats(adapter); + + /* Check PCIE slot type/speed/width */ + ixgbe_get_slot_info(hw); - /* And now turn on interrupts */ - ixgbe_enable_intr(adapter); + /* Set an initial default flow control value */ + adapter->fc = ixgbe_fc_full; #ifdef PCI_IOV - /* Enable the use of the MBX by the VF's */ - { - u32 reg = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT); - reg |= IXGBE_CTRL_EXT_PFRSTD; - IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, reg); - } -#endif - - /* Now inform the stack we're ready */ - ifp->if_drv_flags |= IFF_DRV_RUNNING; - - return; -} - -static void -ixgbe_init(void *arg) -{ - struct adapter *adapter = arg; - - IXGBE_CORE_LOCK(adapter); - ixgbe_init_locked(adapter); - IXGBE_CORE_UNLOCK(adapter); - return; -} - -static void -ixgbe_config_gpie(struct adapter *adapter) -{ - struct ixgbe_hw *hw = &adapter->hw; - u32 gpie; - - gpie = IXGBE_READ_REG(hw, IXGBE_GPIE); - - /* Fan Failure Interrupt */ - if (hw->device_id == IXGBE_DEV_ID_82598AT) - gpie |= IXGBE_SDP1_GPIEN; + if ((hw->mac.type != ixgbe_mac_82598EB) && (adapter->intr_type == IFLIB_INTR_MSIX)) { + nvlist_t *pf_schema, *vf_schema; - /* - * Module detection (SDP2) - * Media ready (SDP1) - */ - if (hw->mac.type == ixgbe_mac_82599EB) { - gpie |= IXGBE_SDP2_GPIEN; - if (hw->device_id != IXGBE_DEV_ID_82599_QSFP_SF_QP) - gpie |= IXGBE_SDP1_GPIEN; + hw->mbx.ops.init_params(hw); + pf_schema = pci_iov_schema_alloc_node(); + vf_schema = pci_iov_schema_alloc_node(); + pci_iov_schema_add_unicast_mac(vf_schema, "mac-addr", 0, NULL); + pci_iov_schema_add_bool(vf_schema, "mac-anti-spoof", + IOV_SCHEMA_HASDEFAULT, TRUE); + pci_iov_schema_add_bool(vf_schema, "allow-set-mac", + IOV_SCHEMA_HASDEFAULT, FALSE); + pci_iov_schema_add_bool(vf_schema, "allow-promisc", + IOV_SCHEMA_HASDEFAULT, FALSE); + error = pci_iov_attach(dev, pf_schema, vf_schema); + if (error != 0) { + device_printf(dev, + "Error %d setting up SR-IOV\n", error); + } } +#endif /* PCI_IOV */ - /* - * Thermal Failure Detection (X540) - * Link Detection (X552 SFP+, X552/X557-AT) - */ - if (hw->mac.type == ixgbe_mac_X540 || - hw->device_id == IXGBE_DEV_ID_X550EM_X_SFP || - hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) - gpie |= IXGBE_SDP0_GPIEN_X540; + /* Check for certain supported features */ + ixgbe_check_wol_support(adapter); + ixgbe_check_eee_support(adapter); - if (adapter->msix > 1) { - /* Enable Enhanced MSIX mode */ - gpie |= IXGBE_GPIE_MSIX_MODE; - gpie |= IXGBE_GPIE_EIAME | IXGBE_GPIE_PBA_SUPPORT | - IXGBE_GPIE_OCD; - } + /* let hardware know driver is loaded */ + ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT); + ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD; + IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext); - IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie); - return; + return (0); +err: + ixgbe_if_detach(ctx); + return (error); } /* - * Requires adapter->max_frame_size to be set. + * Checks whether the adapter supports Energy Efficient Ethernet + * or not, based on device ID. */ static void -ixgbe_config_delay_values(struct adapter *adapter) +ixgbe_check_eee_support(struct adapter *adapter) { struct ixgbe_hw *hw = &adapter->hw; - u32 rxpb, frame, size, tmp; - - frame = adapter->max_frame_size; - - /* Calculate High Water */ - switch (hw->mac.type) { - case ixgbe_mac_X540: - case ixgbe_mac_X550: - case ixgbe_mac_X550EM_x: - tmp = IXGBE_DV_X540(frame, frame); - break; - default: - tmp = IXGBE_DV(frame, frame); - break; - } - size = IXGBE_BT2KB(tmp); - rxpb = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0)) >> 10; - hw->fc.high_water[0] = rxpb - size; - - /* Now calculate Low Water */ - switch (hw->mac.type) { - case ixgbe_mac_X540: - case ixgbe_mac_X550: - case ixgbe_mac_X550EM_x: - tmp = IXGBE_LOW_DV_X540(frame); - break; - default: - tmp = IXGBE_LOW_DV(frame); - break; - } - hw->fc.low_water[0] = IXGBE_BT2KB(tmp); - hw->fc.requested_mode = adapter->fc; - hw->fc.pause_time = IXGBE_FC_PAUSE; - hw->fc.send_xon = TRUE; + adapter->eee_enabled = !!(hw->mac.ops.setup_eee); } /* -** -** MSIX Interrupt Handlers and Tasklets -** -*/ - -static inline void -ixgbe_enable_queue(struct adapter *adapter, u32 vector) -{ - struct ixgbe_hw *hw = &adapter->hw; - u64 queue = (u64)(1 << vector); - u32 mask; - - if (hw->mac.type == ixgbe_mac_82598EB) { - mask = (IXGBE_EIMS_RTX_QUEUE & queue); - IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask); - } else { - mask = (queue & 0xFFFFFFFF); - if (mask) - IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask); - mask = (queue >> 32); - if (mask) - IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask); - } -} - -static inline void -ixgbe_disable_queue(struct adapter *adapter, u32 vector) + * Checks whether the adapter's ports are capable of + * Wake On LAN by reading the adapter's NVM. + * + * Sets each port's hw->wol_enabled value depending + * on the value read here. + */ +static void +ixgbe_check_wol_support(struct adapter *adapter) { struct ixgbe_hw *hw = &adapter->hw; - u64 queue = (u64)(1 << vector); - u32 mask; + u16 dev_caps = 0; - if (hw->mac.type == ixgbe_mac_82598EB) { - mask = (IXGBE_EIMS_RTX_QUEUE & queue); - IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask); - } else { - mask = (queue & 0xFFFFFFFF); - if (mask) - IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask); - mask = (queue >> 32); - if (mask) - IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask); - } -} + /* Find out WoL support for port */ + adapter->wol_support = hw->wol_enabled = 0; + ixgbe_get_device_caps(hw, &dev_caps); + if ((dev_caps & IXGBE_DEVICE_CAPS_WOL_PORT0_1) || + ((dev_caps & IXGBE_DEVICE_CAPS_WOL_PORT0) && + hw->bus.func == 0)) + adapter->wol_support = hw->wol_enabled = 1; -static void -ixgbe_handle_que(void *context, int pending) -{ - struct ix_queue *que = context; - struct adapter *adapter = que->adapter; - struct tx_ring *txr = que->txr; - struct ifnet *ifp = adapter->ifp; - - if (ifp->if_drv_flags & IFF_DRV_RUNNING) { - ixgbe_rxeof(que); - IXGBE_TX_LOCK(txr); - ixgbe_txeof(txr); -#ifndef IXGBE_LEGACY_TX - if (!drbr_empty(ifp, txr->br)) - ixgbe_mq_start_locked(ifp, txr); -#else - if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) - ixgbe_start_locked(txr, ifp); -#endif - IXGBE_TX_UNLOCK(txr); - } + /* Save initial wake up filter configuration */ + adapter->wufc = IXGBE_READ_REG(hw, IXGBE_WUFC); - /* Reenable this interrupt */ - if (que->res != NULL) - ixgbe_enable_queue(adapter, que->msix); - else - ixgbe_enable_intr(adapter); return; } - - + /********************************************************************* * - * Legacy Interrupt Service routine + * Setup networking device structure and register an interface. * **********************************************************************/ - -static void -ixgbe_legacy_irq(void *arg) +static int +ixgbe_interface_setup(if_ctx_t ctx) { - struct ix_queue *que = arg; - struct adapter *adapter = que->adapter; - struct ixgbe_hw *hw = &adapter->hw; - struct ifnet *ifp = adapter->ifp; - struct tx_ring *txr = adapter->tx_rings; - bool more; - u32 reg_eicr; + struct ifnet *ifp = iflib_get_ifp(ctx); + struct adapter *adapter = iflib_get_softc(ctx); + uint64_t cap = 0; + INIT_DEBUGOUT("ixgbe_interface_setup: begin"); - reg_eicr = IXGBE_READ_REG(hw, IXGBE_EICR); + cap |= IFCAP_HWCSUM | IFCAP_TSO | IFCAP_VLAN_HWCSUM | IFCAP_JUMBO_MTU; + cap |= IFCAP_LRO | IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU; + cap |= IFCAP_HWSTATS; - ++que->irqs; - if (reg_eicr == 0) { - ixgbe_enable_intr(adapter); - return; - } + if_setifheaderlen(ifp, sizeof(struct ether_vlan_header)); + if_setcapabilitiesbit(ifp, cap, 0); + if_setcapenable(ifp, if_getcapabilities(ifp)); + if_setbaudrate(ifp, 1000000000); - more = ixgbe_rxeof(que); + adapter->max_frame_size = + ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN; - IXGBE_TX_LOCK(txr); - ixgbe_txeof(txr); -#ifdef IXGBE_LEGACY_TX - if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) - ixgbe_start_locked(txr, ifp); -#else - if (!drbr_empty(ifp, txr->br)) - ixgbe_mq_start_locked(ifp, txr); +#if __FreeBSD_version >= 1100045 + /* TSO parameters */ + ifp->if_hw_tsomax = 65518; + ifp->if_hw_tsomaxsegcount = IXGBE_82599_SCATTER; + ifp->if_hw_tsomaxsegsize = 2048; #endif - IXGBE_TX_UNLOCK(txr); - /* Check for fan failure */ - if ((hw->device_id == IXGBE_DEV_ID_82598AT) && - (reg_eicr & IXGBE_EICR_GPI_SDP1)) { - device_printf(adapter->dev, "\nCRITICAL: FAN FAILURE!! " - "REPLACE IMMEDIATELY!!\n"); - IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EICR_GPI_SDP1_BY_MAC(hw)); - } + /* + ** Don't turn this on by default, if vlans are + ** created on another pseudo device (eg. lagg) + ** then vlan events are not passed thru, breaking + ** operation, but with HW FILTER off it works. If + ** using vlans directly on the ixgbe driver you can + ** enable this and get full hardware tag filtering. + */ + ifp->if_capabilities |= IFCAP_VLAN_HWFILTER; - /* Link status change */ - if (reg_eicr & IXGBE_EICR_LSC) - taskqueue_enqueue(adapter->tq, &adapter->link_task); + ixgbe_add_media_types(ctx); - /* External PHY interrupt */ - if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T && - (reg_eicr & IXGBE_EICR_GPI_SDP0_X540)) - taskqueue_enqueue(adapter->tq, &adapter->phy_task); + /* Autoselect media by default */ + ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO); - if (more) - taskqueue_enqueue(que->tq, &que->que_task); - else - ixgbe_enable_intr(adapter); - return; + return (0); } - -/********************************************************************* - * - * MSIX Queue Interrupt Service routine - * - **********************************************************************/ -void -ixgbe_msix_que(void *arg) +#if __FreeBSD_version >= 1100036 +static uint64_t +ixgbe_if_get_counter(if_ctx_t ctx, ift_counter cnt) { - struct ix_queue *que = arg; - struct adapter *adapter = que->adapter; - struct ifnet *ifp = adapter->ifp; - struct tx_ring *txr = que->txr; - struct rx_ring *rxr = que->rxr; - bool more; - u32 newitr = 0; - + struct adapter *adapter; + if_t ifp = iflib_get_ifp(ctx); - /* Protect against spurious interrupts */ - if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) - return; + adapter = if_getsoftc(ifp); - ixgbe_disable_queue(adapter, que->msix); - ++que->irqs; + switch (cnt) { + case IFCOUNTER_IPACKETS: + return (adapter->ipackets); + case IFCOUNTER_OPACKETS: + return (adapter->opackets); + case IFCOUNTER_IBYTES: + return (adapter->ibytes); + case IFCOUNTER_OBYTES: + return (adapter->obytes); + case IFCOUNTER_IMCASTS: + return (adapter->imcasts); + case IFCOUNTER_OMCASTS: + return (adapter->omcasts); + case IFCOUNTER_COLLISIONS: + return (0); + case IFCOUNTER_IQDROPS: + return (adapter->iqdrops); + case IFCOUNTER_OQDROPS: + return (0); + case IFCOUNTER_IERRORS: + return (adapter->ierrors); + default: + return (if_get_counter_default(ifp, cnt)); + } +} +#endif - more = ixgbe_rxeof(que); +static void +ixgbe_add_media_types(if_ctx_t ctx) +{ + struct adapter *adapter = iflib_get_softc(ctx); + struct ixgbe_hw *hw = &adapter->hw; + device_t dev = iflib_get_dev(ctx); + int layer; - IXGBE_TX_LOCK(txr); - ixgbe_txeof(txr); -#ifdef IXGBE_LEGACY_TX - if (!IFQ_DRV_IS_EMPTY(ifp->if_snd)) - ixgbe_start_locked(txr, ifp); -#else - if (!drbr_empty(ifp, txr->br)) - ixgbe_mq_start_locked(ifp, txr); -#endif - IXGBE_TX_UNLOCK(txr); + layer = adapter->phy_layer = ixgbe_get_supported_physical_layer(hw); - /* Do AIM now? */ + /* Media types with matching FreeBSD media defines */ + if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T) + ifmedia_add(adapter->media, IFM_ETHER | IFM_10G_T, 0, NULL); + if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_T) + ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL); + if (layer & IXGBE_PHYSICAL_LAYER_100BASE_TX) + ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL); + + if (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU || + layer & IXGBE_PHYSICAL_LAYER_SFP_ACTIVE_DA) + ifmedia_add(adapter->media, IFM_ETHER | IFM_10G_TWINAX, 0, NULL); - if (adapter->enable_aim == FALSE) - goto no_calc; + if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LR) + ifmedia_add(adapter->media, IFM_ETHER | IFM_10G_LR, 0, NULL); + if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR) + ifmedia_add(adapter->media, IFM_ETHER | IFM_10G_SR, 0, NULL); + if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_CX4) + ifmedia_add(adapter->media, IFM_ETHER | IFM_10G_CX4, 0, NULL); + if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX) + ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_SX, 0, NULL); /* - ** Do Adaptive Interrupt Moderation: - ** - Write out last calculated setting - ** - Calculate based on average size over - ** the last interval. + ** Other (no matching FreeBSD media type): + ** To workaround this, we'll assign these completely + ** inappropriate media types. */ - if (que->eitr_setting) - IXGBE_WRITE_REG(&adapter->hw, - IXGBE_EITR(que->msix), que->eitr_setting); - - que->eitr_setting = 0; - - /* Idle, do nothing */ - if ((txr->bytes == 0) && (rxr->bytes == 0)) - goto no_calc; - - if ((txr->bytes) && (txr->packets)) - newitr = txr->bytes/txr->packets; - if ((rxr->bytes) && (rxr->packets)) - newitr = max(newitr, - (rxr->bytes / rxr->packets)); - newitr += 24; /* account for hardware frame, crc */ - - /* set an upper boundary */ - newitr = min(newitr, 3000); - - /* Be nice to the mid range */ - if ((newitr > 300) && (newitr < 1200)) - newitr = (newitr / 3); - else - newitr = (newitr / 2); - - if (adapter->hw.mac.type == ixgbe_mac_82598EB) - newitr |= newitr << 16; - else - newitr |= IXGBE_EITR_CNT_WDIS; - - /* save for next interrupt */ - que->eitr_setting = newitr; - - /* Reset state */ - txr->bytes = 0; - txr->packets = 0; - rxr->bytes = 0; - rxr->packets = 0; + if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR) { + device_printf(dev, "Media supported: 10GbaseKR\n"); + device_printf(dev, "10GbaseKR mapped to 10GbaseSR\n"); + ifmedia_add(adapter->media, IFM_ETHER | IFM_10G_SR, 0, NULL); + } + if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4) { + device_printf(dev, "Media supported: 10GbaseKX4\n"); + device_printf(dev, "10GbaseKX4 mapped to 10GbaseCX4\n"); + ifmedia_add(adapter->media, IFM_ETHER | IFM_10G_CX4, 0, NULL); + } + if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX) { + device_printf(dev, "Media supported: 1000baseKX\n"); + device_printf(dev, "1000baseKX mapped to 1000baseCX\n"); + ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_CX, 0, NULL); + } + if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_BX) { + /* Someday, someone will care about you... */ + device_printf(dev, "Media supported: 1000baseBX\n"); + } + + if (hw->device_id == IXGBE_DEV_ID_82598AT) { + ifmedia_add(adapter->media, + IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL); + ifmedia_add(adapter->media, + IFM_ETHER | IFM_1000_T, 0, NULL); + } -no_calc: - if (more) - taskqueue_enqueue(que->tq, &que->que_task); - else - ixgbe_enable_queue(adapter, que->msix); - return; + ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL); } - static void -ixgbe_msix_link(void *arg) +ixgbe_config_link(struct adapter *adapter) { - struct adapter *adapter = arg; struct ixgbe_hw *hw = &adapter->hw; - u32 reg_eicr, mod_mask; - - ++adapter->link_irq; - - /* Pause other interrupts */ - IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_OTHER); - - /* First get the cause */ - reg_eicr = IXGBE_READ_REG(hw, IXGBE_EICS); - /* Be sure the queue bits are not cleared */ - reg_eicr &= ~IXGBE_EICR_RTX_QUEUE; - /* Clear interrupt with write */ - IXGBE_WRITE_REG(hw, IXGBE_EICR, reg_eicr); - - /* Link status change */ - if (reg_eicr & IXGBE_EICR_LSC) { - IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EIMC_LSC); - taskqueue_enqueue(adapter->tq, &adapter->link_task); - } - - if (adapter->hw.mac.type != ixgbe_mac_82598EB) { -#ifdef IXGBE_FDIR - if (reg_eicr & IXGBE_EICR_FLOW_DIR) { - /* This is probably overkill :) */ - if (!atomic_cmpset_int(&adapter->fdir_reinit, 0, 1)) - return; - /* Disable the interrupt */ - IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EICR_FLOW_DIR); - taskqueue_enqueue(adapter->tq, &adapter->fdir_task); - } else -#endif - if (reg_eicr & IXGBE_EICR_ECC) { - device_printf(adapter->dev, "CRITICAL: ECC ERROR!! " - "Please Reboot!!\n"); - IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC); - } - - /* Check for over temp condition */ - if (reg_eicr & IXGBE_EICR_TS) { - device_printf(adapter->dev, "CRITICAL: OVER TEMP!! " - "PHY IS SHUT DOWN!!\n"); - device_printf(adapter->dev, "System shutdown required!\n"); - IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_TS); - } -#ifdef PCI_IOV - if (reg_eicr & IXGBE_EICR_MAILBOX) - taskqueue_enqueue(adapter->tq, &adapter->mbx_task); -#endif - } + u32 autoneg, err = 0; + bool sfp, negotiate; - /* Pluggable optics-related interrupt */ - if (hw->device_id == IXGBE_DEV_ID_X550EM_X_SFP) - mod_mask = IXGBE_EICR_GPI_SDP0_X540; - else - mod_mask = IXGBE_EICR_GPI_SDP2_BY_MAC(hw); + sfp = ixgbe_is_sfp(hw); - if (ixgbe_is_sfp(hw)) { - if (reg_eicr & IXGBE_EICR_GPI_SDP1_BY_MAC(hw)) { - IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1_BY_MAC(hw)); - taskqueue_enqueue(adapter->tq, &adapter->msf_task); - } else if (reg_eicr & mod_mask) { - IXGBE_WRITE_REG(hw, IXGBE_EICR, mod_mask); - taskqueue_enqueue(adapter->tq, &adapter->mod_task); + if (sfp) { + if (hw->phy.multispeed_fiber) { + hw->mac.ops.setup_sfp(hw); + ixgbe_enable_tx_laser(hw); + GROUPTASK_ENQUEUE(&adapter->msf_task); + } else { + GROUPTASK_ENQUEUE(&adapter->mod_task); } + } else { + if (hw->mac.ops.check_link) + err = ixgbe_check_link(hw, &adapter->link_speed, + &adapter->link_up, FALSE); + if (err) + return; + autoneg = hw->phy.autoneg_advertised; + if ((!autoneg) && (hw->mac.ops.get_link_capabilities)) + err = hw->mac.ops.get_link_capabilities(hw, + &autoneg, &negotiate); + if (err) + return; + if (hw->mac.ops.setup_link) + err = hw->mac.ops.setup_link(hw, + autoneg, adapter->link_up); } - /* Check for fan failure */ - if ((hw->device_id == IXGBE_DEV_ID_82598AT) && - (reg_eicr & IXGBE_EICR_GPI_SDP1)) { - IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1); - device_printf(adapter->dev, "\nCRITICAL: FAN FAILURE!! " - "REPLACE IMMEDIATELY!!\n"); - } - - /* External PHY interrupt */ - if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T && - (reg_eicr & IXGBE_EICR_GPI_SDP0_X540)) { - IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP0_X540); - taskqueue_enqueue(adapter->tq, &adapter->phy_task); - } - - /* Re-enable other interrupts */ - IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, IXGBE_EIMS_OTHER); - return; } -/********************************************************************* - * - * Media Ioctl callback + +/********************************************************************** * - * This routine is called whenever the user queries the status of - * the interface using ifconfig. + * Update the board statistics counters. * **********************************************************************/ static void -ixgbe_media_status(struct ifnet * ifp, struct ifmediareq * ifmr) +ixgbe_update_stats_counters(struct adapter *adapter) { - struct adapter *adapter = ifp->if_softc; struct ixgbe_hw *hw = &adapter->hw; - int layer; + u32 missed_rx = 0, bprc, lxon, lxoff, total; + u64 total_missed_rx = 0; - INIT_DEBUGOUT("ixgbe_media_status: begin"); - IXGBE_CORE_LOCK(adapter); - ixgbe_update_link_status(adapter); + adapter->stats.pf.crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS); + adapter->stats.pf.illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC); + adapter->stats.pf.errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC); + adapter->stats.pf.mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC); - ifmr->ifm_status = IFM_AVALID; - ifmr->ifm_active = IFM_ETHER; + for (int i = 0; i < 16; i++) { + adapter->stats.pf.qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i)); + adapter->stats.pf.qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i)); + adapter->stats.pf.qprdc[i] += IXGBE_READ_REG(hw, IXGBE_QPRDC(i)); + } + adapter->stats.pf.mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC); + adapter->stats.pf.mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC); + adapter->stats.pf.rlec += IXGBE_READ_REG(hw, IXGBE_RLEC); - if (!adapter->link_active) { - IXGBE_CORE_UNLOCK(adapter); - return; + /* Hardware workaround, gprc counts missed packets */ + adapter->stats.pf.gprc += IXGBE_READ_REG(hw, IXGBE_GPRC); + adapter->stats.pf.gprc -= missed_rx; + + if (hw->mac.type != ixgbe_mac_82598EB) { + adapter->stats.pf.gorc += IXGBE_READ_REG(hw, IXGBE_GORCL) + + ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32); + adapter->stats.pf.gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL) + + ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32); + adapter->stats.pf.tor += IXGBE_READ_REG(hw, IXGBE_TORL) + + ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32); + adapter->stats.pf.lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT); + adapter->stats.pf.lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT); + } else { + adapter->stats.pf.lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC); + adapter->stats.pf.lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC); + /* 82598 only has a counter in the high register */ + adapter->stats.pf.gorc += IXGBE_READ_REG(hw, IXGBE_GORCH); + adapter->stats.pf.gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH); + adapter->stats.pf.tor += IXGBE_READ_REG(hw, IXGBE_TORH); } - ifmr->ifm_status |= IFM_ACTIVE; - layer = adapter->phy_layer; + /* + * Workaround: mprc hardware is incorrectly counting + * broadcasts, so for now we subtract those. + */ + bprc = IXGBE_READ_REG(hw, IXGBE_BPRC); + adapter->stats.pf.bprc += bprc; + adapter->stats.pf.mprc += IXGBE_READ_REG(hw, IXGBE_MPRC); + if (hw->mac.type == ixgbe_mac_82598EB) + adapter->stats.pf.mprc -= bprc; - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T || - layer & IXGBE_PHYSICAL_LAYER_1000BASE_T || - layer & IXGBE_PHYSICAL_LAYER_100BASE_TX) - switch (adapter->link_speed) { - case IXGBE_LINK_SPEED_10GB_FULL: - ifmr->ifm_active |= IFM_10G_T | IFM_FDX; - break; - case IXGBE_LINK_SPEED_1GB_FULL: - ifmr->ifm_active |= IFM_1000_T | IFM_FDX; - break; - case IXGBE_LINK_SPEED_100_FULL: - ifmr->ifm_active |= IFM_100_TX | IFM_FDX; - break; - } - if (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU || - layer & IXGBE_PHYSICAL_LAYER_SFP_ACTIVE_DA) - switch (adapter->link_speed) { - case IXGBE_LINK_SPEED_10GB_FULL: - ifmr->ifm_active |= IFM_10G_TWINAX | IFM_FDX; - break; - } - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LR) - switch (adapter->link_speed) { - case IXGBE_LINK_SPEED_10GB_FULL: - ifmr->ifm_active |= IFM_10G_LR | IFM_FDX; - break; - case IXGBE_LINK_SPEED_1GB_FULL: - ifmr->ifm_active |= IFM_1000_LX | IFM_FDX; - break; - } - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LRM) - switch (adapter->link_speed) { - case IXGBE_LINK_SPEED_10GB_FULL: - ifmr->ifm_active |= IFM_10G_LRM | IFM_FDX; - break; - case IXGBE_LINK_SPEED_1GB_FULL: - ifmr->ifm_active |= IFM_1000_LX | IFM_FDX; - break; - } - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR || - layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX) - switch (adapter->link_speed) { - case IXGBE_LINK_SPEED_10GB_FULL: - ifmr->ifm_active |= IFM_10G_SR | IFM_FDX; - break; - case IXGBE_LINK_SPEED_1GB_FULL: - ifmr->ifm_active |= IFM_1000_SX | IFM_FDX; - break; - } - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_CX4) - switch (adapter->link_speed) { - case IXGBE_LINK_SPEED_10GB_FULL: - ifmr->ifm_active |= IFM_10G_CX4 | IFM_FDX; - break; - } - /* - ** XXX: These need to use the proper media types once - ** they're added. - */ -#ifndef IFM_ETH_XTYPE - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR) - switch (adapter->link_speed) { - case IXGBE_LINK_SPEED_10GB_FULL: - ifmr->ifm_active |= IFM_10G_SR | IFM_FDX; - break; - case IXGBE_LINK_SPEED_2_5GB_FULL: - ifmr->ifm_active |= IFM_2500_SX | IFM_FDX; - break; - case IXGBE_LINK_SPEED_1GB_FULL: - ifmr->ifm_active |= IFM_1000_CX | IFM_FDX; - break; - } - else if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4 - || layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX) - switch (adapter->link_speed) { - case IXGBE_LINK_SPEED_10GB_FULL: - ifmr->ifm_active |= IFM_10G_CX4 | IFM_FDX; - break; - case IXGBE_LINK_SPEED_2_5GB_FULL: - ifmr->ifm_active |= IFM_2500_SX | IFM_FDX; - break; - case IXGBE_LINK_SPEED_1GB_FULL: - ifmr->ifm_active |= IFM_1000_CX | IFM_FDX; - break; - } -#else - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR) - switch (adapter->link_speed) { - case IXGBE_LINK_SPEED_10GB_FULL: - ifmr->ifm_active |= IFM_10G_KR | IFM_FDX; - break; - case IXGBE_LINK_SPEED_2_5GB_FULL: - ifmr->ifm_active |= IFM_2500_KX | IFM_FDX; - break; - case IXGBE_LINK_SPEED_1GB_FULL: - ifmr->ifm_active |= IFM_1000_KX | IFM_FDX; - break; - } - else if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4 - || layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX) - switch (adapter->link_speed) { - case IXGBE_LINK_SPEED_10GB_FULL: - ifmr->ifm_active |= IFM_10G_KX4 | IFM_FDX; - break; - case IXGBE_LINK_SPEED_2_5GB_FULL: - ifmr->ifm_active |= IFM_2500_KX | IFM_FDX; - break; - case IXGBE_LINK_SPEED_1GB_FULL: - ifmr->ifm_active |= IFM_1000_KX | IFM_FDX; - break; - } -#endif - - /* If nothing is recognized... */ - if (IFM_SUBTYPE(ifmr->ifm_active) == 0) - ifmr->ifm_active |= IFM_UNKNOWN; - -#if __FreeBSD_version >= 900025 - /* Display current flow control setting used on link */ - if (hw->fc.current_mode == ixgbe_fc_rx_pause || - hw->fc.current_mode == ixgbe_fc_full) - ifmr->ifm_active |= IFM_ETH_RXPAUSE; - if (hw->fc.current_mode == ixgbe_fc_tx_pause || - hw->fc.current_mode == ixgbe_fc_full) - ifmr->ifm_active |= IFM_ETH_TXPAUSE; -#endif - - IXGBE_CORE_UNLOCK(adapter); - - return; -} - -/********************************************************************* - * - * Media Ioctl callback - * - * This routine is called when the user changes speed/duplex using - * media/mediopt option with ifconfig. - * - **********************************************************************/ -static int -ixgbe_media_change(struct ifnet * ifp) -{ - struct adapter *adapter = ifp->if_softc; - struct ifmedia *ifm = &adapter->media; - struct ixgbe_hw *hw = &adapter->hw; - ixgbe_link_speed speed = 0; - - INIT_DEBUGOUT("ixgbe_media_change: begin"); - - if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER) - return (EINVAL); + adapter->stats.pf.prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64); + adapter->stats.pf.prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127); + adapter->stats.pf.prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255); + adapter->stats.pf.prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511); + adapter->stats.pf.prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023); + adapter->stats.pf.prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522); - if (hw->phy.media_type == ixgbe_media_type_backplane) - return (ENODEV); + lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC); + adapter->stats.pf.lxontxc += lxon; + lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC); + adapter->stats.pf.lxofftxc += lxoff; + total = lxon + lxoff; - /* - ** We don't actually need to check against the supported - ** media types of the adapter; ifmedia will take care of - ** that for us. - */ -#ifndef IFM_ETH_XTYPE - switch (IFM_SUBTYPE(ifm->ifm_media)) { - case IFM_AUTO: - case IFM_10G_T: - speed |= IXGBE_LINK_SPEED_100_FULL; - case IFM_10G_LRM: - case IFM_10G_SR: /* KR, too */ - case IFM_10G_LR: - case IFM_10G_CX4: /* KX4 */ - speed |= IXGBE_LINK_SPEED_1GB_FULL; - case IFM_10G_TWINAX: - speed |= IXGBE_LINK_SPEED_10GB_FULL; - break; - case IFM_1000_T: - speed |= IXGBE_LINK_SPEED_100_FULL; - case IFM_1000_LX: - case IFM_1000_SX: - case IFM_1000_CX: /* KX */ - speed |= IXGBE_LINK_SPEED_1GB_FULL; - break; - case IFM_100_TX: - speed |= IXGBE_LINK_SPEED_100_FULL; - break; - default: - goto invalid; - } -#else - switch (IFM_SUBTYPE(ifm->ifm_media)) { - case IFM_AUTO: - case IFM_10G_T: - speed |= IXGBE_LINK_SPEED_100_FULL; - case IFM_10G_LRM: - case IFM_10G_KR: - case IFM_10G_LR: - case IFM_10G_KX4: - speed |= IXGBE_LINK_SPEED_1GB_FULL; - case IFM_10G_TWINAX: - speed |= IXGBE_LINK_SPEED_10GB_FULL; - break; - case IFM_1000_T: - speed |= IXGBE_LINK_SPEED_100_FULL; - case IFM_1000_LX: - case IFM_1000_SX: - case IFM_1000_KX: - speed |= IXGBE_LINK_SPEED_1GB_FULL; - break; - case IFM_100_TX: - speed |= IXGBE_LINK_SPEED_100_FULL; - break; - default: - goto invalid; - } -#endif + adapter->stats.pf.gptc += IXGBE_READ_REG(hw, IXGBE_GPTC); + adapter->stats.pf.mptc += IXGBE_READ_REG(hw, IXGBE_MPTC); + adapter->stats.pf.ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64); + adapter->stats.pf.gptc -= total; + adapter->stats.pf.mptc -= total; + adapter->stats.pf.ptc64 -= total; + adapter->stats.pf.gotc -= total * ETHER_MIN_LEN; - hw->mac.autotry_restart = TRUE; - hw->mac.ops.setup_link(hw, speed, TRUE); - if (IFM_SUBTYPE(ifm->ifm_media) == IFM_AUTO) { - adapter->advertise = 0; - } else { - if ((speed & IXGBE_LINK_SPEED_10GB_FULL) != 0) - adapter->advertise |= 1 << 2; - if ((speed & IXGBE_LINK_SPEED_1GB_FULL) != 0) - adapter->advertise |= 1 << 1; - if ((speed & IXGBE_LINK_SPEED_100_FULL) != 0) - adapter->advertise |= 1 << 0; + adapter->stats.pf.ruc += IXGBE_READ_REG(hw, IXGBE_RUC); + adapter->stats.pf.rfc += IXGBE_READ_REG(hw, IXGBE_RFC); + adapter->stats.pf.roc += IXGBE_READ_REG(hw, IXGBE_ROC); + adapter->stats.pf.rjc += IXGBE_READ_REG(hw, IXGBE_RJC); + adapter->stats.pf.mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC); + adapter->stats.pf.mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC); + adapter->stats.pf.mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC); + adapter->stats.pf.tpr += IXGBE_READ_REG(hw, IXGBE_TPR); + adapter->stats.pf.tpt += IXGBE_READ_REG(hw, IXGBE_TPT); + adapter->stats.pf.ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127); + adapter->stats.pf.ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255); + adapter->stats.pf.ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511); + adapter->stats.pf.ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023); + adapter->stats.pf.ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522); + adapter->stats.pf.bptc += IXGBE_READ_REG(hw, IXGBE_BPTC); + adapter->stats.pf.xec += IXGBE_READ_REG(hw, IXGBE_XEC); + adapter->stats.pf.fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC); + adapter->stats.pf.fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST); + /* Only read FCOE on 82599 */ + if (hw->mac.type != ixgbe_mac_82598EB) { + adapter->stats.pf.fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC); + adapter->stats.pf.fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC); + adapter->stats.pf.fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC); + adapter->stats.pf.fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC); + adapter->stats.pf.fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC); } - return (0); - -invalid: - device_printf(adapter->dev, "Invalid media type!\n"); - return (EINVAL); + /* Fill out the OS statistics structure */ + IXGBE_SET_IPACKETS(adapter, adapter->stats.pf.gprc); + IXGBE_SET_OPACKETS(adapter, adapter->stats.pf.gptc); + IXGBE_SET_IBYTES(adapter, adapter->stats.pf.gorc); + IXGBE_SET_OBYTES(adapter, adapter->stats.pf.gotc); + IXGBE_SET_IMCASTS(adapter, adapter->stats.pf.mprc); + IXGBE_SET_OMCASTS(adapter, adapter->stats.pf.mptc); + IXGBE_SET_COLLISIONS(adapter, 0); + IXGBE_SET_IQDROPS(adapter, total_missed_rx); + IXGBE_SET_IERRORS(adapter, adapter->stats.pf.crcerrs + + adapter->stats.pf.rlec); } +/* + * Add sysctl variables, one per statistic, to the system. + */ static void -ixgbe_set_promisc(struct adapter *adapter) +ixgbe_add_hw_stats(struct adapter *adapter) { - u_int32_t reg_rctl; - struct ifnet *ifp = adapter->ifp; - int mcnt = 0; + device_t dev = iflib_get_dev(adapter->ctx); + struct ix_queue *que; + int i; - reg_rctl = IXGBE_READ_REG(&adapter->hw, IXGBE_FCTRL); - reg_rctl &= (~IXGBE_FCTRL_UPE); - if (ifp->if_flags & IFF_ALLMULTI) - mcnt = MAX_NUM_MULTICAST_ADDRESSES; - else { - struct ifmultiaddr *ifma; -#if __FreeBSD_version < 800000 - IF_ADDR_LOCK(ifp); -#else - if_maddr_rlock(ifp); -#endif - TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { - if (ifma->ifma_addr->sa_family != AF_LINK) - continue; - if (mcnt == MAX_NUM_MULTICAST_ADDRESSES) - break; - mcnt++; - } -#if __FreeBSD_version < 800000 - IF_ADDR_UNLOCK(ifp); -#else - if_maddr_runlock(ifp); -#endif - } - if (mcnt < MAX_NUM_MULTICAST_ADDRESSES) - reg_rctl &= (~IXGBE_FCTRL_MPE); - IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCTRL, reg_rctl); + struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev); + struct sysctl_oid *tree = device_get_sysctl_tree(dev); + struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree); + struct ixgbe_hw_stats *stats = &adapter->stats.pf; - if (ifp->if_flags & IFF_PROMISC) { - reg_rctl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE); - IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCTRL, reg_rctl); - } else if (ifp->if_flags & IFF_ALLMULTI) { - reg_rctl |= IXGBE_FCTRL_MPE; - reg_rctl &= ~IXGBE_FCTRL_UPE; - IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCTRL, reg_rctl); - } - return; -} + struct sysctl_oid *stat_node, *queue_node; + struct sysctl_oid_list *stat_list, *queue_list; +#define QUEUE_NAME_LEN 32 + char namebuf[QUEUE_NAME_LEN]; -/********************************************************************* - * Multicast Update - * - * This routine is called whenever multicast address list is updated. - * - **********************************************************************/ -#define IXGBE_RAR_ENTRIES 16 + /* Driver Statistics */ + SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped", + CTLFLAG_RD, &adapter->dropped_pkts, + "Driver dropped packets"); + SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "mbuf_defrag_failed", + CTLFLAG_RD, &adapter->mbuf_defrag_failed, + "m_defrag() failed"); + SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_events", + CTLFLAG_RD, &adapter->watchdog_events, + "Watchdog timeouts"); + SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq", + CTLFLAG_RD, &adapter->link_irq, + "Link MSIX IRQ Handled"); -static void -ixgbe_set_multi(struct adapter *adapter) -{ - u32 fctrl; - u8 *update_ptr; - struct ifmultiaddr *ifma; - struct ixgbe_mc_addr *mta; - int mcnt = 0; - struct ifnet *ifp = adapter->ifp; + for (i = 0, que = adapter->queues; i < adapter->num_queues; i++, que++) { + struct tx_ring *txr = &que->txr; + snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i); + queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf, + CTLFLAG_RD, NULL, "Queue Name"); + queue_list = SYSCTL_CHILDREN(queue_node); - IOCTL_DEBUGOUT("ixgbe_set_multi: begin"); + SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "interrupt_rate", + CTLTYPE_UINT | CTLFLAG_RW, &adapter->queues[i], + sizeof(&adapter->queues[i]), + ixgbe_sysctl_interrupt_rate_handler, "IU", + "Interrupt Rate"); + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "irqs", + CTLFLAG_RD, &(adapter->queues[i].irqs), + "irqs on this queue"); + SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head", + CTLTYPE_UINT | CTLFLAG_RD, txr, sizeof(txr), + ixgbe_sysctl_tdh_handler, "IU", + "Transmit Descriptor Head"); + SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail", + CTLTYPE_UINT | CTLFLAG_RD, txr, sizeof(txr), + ixgbe_sysctl_tdt_handler, "IU", + "Transmit Descriptor Tail"); + SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tso_tx", + CTLFLAG_RD, &txr->tso_tx, + "TSO"); + SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "no_tx_dma_setup", + CTLFLAG_RD, &txr->no_tx_dma_setup, + "Driver tx dma failure in xmit"); + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "no_desc_avail", + CTLFLAG_RD, &txr->no_desc_avail, + "Queue No Descriptor Available"); + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "tx_packets", + CTLFLAG_RD, &txr->total_packets, + "Queue Packets Transmitted"); + } - mta = adapter->mta; - bzero(mta, sizeof(*mta) * MAX_NUM_MULTICAST_ADDRESSES); + for (i = 0, que = adapter->queues; i < adapter->num_queues; i++, que++) { + struct rx_ring *rxr = &que->rxr; + snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i); + queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf, + CTLFLAG_RD, NULL, "Queue Name"); + queue_list = SYSCTL_CHILDREN(queue_node); -#if __FreeBSD_version < 800000 - IF_ADDR_LOCK(ifp); -#else - if_maddr_rlock(ifp); -#endif - TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { - if (ifma->ifma_addr->sa_family != AF_LINK) - continue; - if (mcnt == MAX_NUM_MULTICAST_ADDRESSES) - break; - bcopy(LLADDR((struct sockaddr_dl *) ifma->ifma_addr), - mta[mcnt].addr, IXGBE_ETH_LENGTH_OF_ADDRESS); - mta[mcnt].vmdq = adapter->pool; - mcnt++; + snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i); + queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf, + CTLFLAG_RD, NULL, "Queue Name"); + queue_list = SYSCTL_CHILDREN(queue_node); + + SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head", + CTLTYPE_UINT | CTLFLAG_RD, rxr, sizeof(rxr), + ixgbe_sysctl_rdh_handler, "IU", + "Receive Descriptor Head"); + SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail", + CTLTYPE_UINT | CTLFLAG_RD, rxr, sizeof(rxr), + ixgbe_sysctl_rdt_handler, "IU", + "Receive Descriptor Tail"); + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_packets", + CTLFLAG_RD, &rxr->rx_packets, + "Queue Packets Received"); + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_bytes", + CTLFLAG_RD, &rxr->rx_bytes, + "Queue Bytes Received"); + SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_copies", + CTLFLAG_RD, &rxr->rx_copies, + "Copied RX Frames"); } -#if __FreeBSD_version < 800000 - IF_ADDR_UNLOCK(ifp); -#else - if_maddr_runlock(ifp); -#endif - fctrl = IXGBE_READ_REG(&adapter->hw, IXGBE_FCTRL); - fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE); - if (ifp->if_flags & IFF_PROMISC) - fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE); - else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES || - ifp->if_flags & IFF_ALLMULTI) { - fctrl |= IXGBE_FCTRL_MPE; - fctrl &= ~IXGBE_FCTRL_UPE; - } else - fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE); - - IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCTRL, fctrl); + /* MAC stats get the own sub node */ - if (mcnt < MAX_NUM_MULTICAST_ADDRESSES) { - update_ptr = (u8 *)mta; - ixgbe_update_mc_addr_list(&adapter->hw, - update_ptr, mcnt, ixgbe_mc_array_itr, TRUE); - } + stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats", + CTLFLAG_RD, NULL, "MAC Statistics"); + stat_list = SYSCTL_CHILDREN(stat_node); - return; -} + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs", + CTLFLAG_RD, &stats->crcerrs, + "CRC Errors"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "ill_errs", + CTLFLAG_RD, &stats->illerrc, + "Illegal Byte Errors"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "byte_errs", + CTLFLAG_RD, &stats->errbc, + "Byte Errors"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "short_discards", + CTLFLAG_RD, &stats->mspdc, + "MAC Short Packets Discarded"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "local_faults", + CTLFLAG_RD, &stats->mlfc, + "MAC Local Faults"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "remote_faults", + CTLFLAG_RD, &stats->mrfc, + "MAC Remote Faults"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rec_len_errs", + CTLFLAG_RD, &stats->rlec, + "Receive Length Errors"); -/* - * This is an iterator function now needed by the multicast - * shared code. It simply feeds the shared code routine the - * addresses in the array of ixgbe_set_multi() one by one. - */ -static u8 * -ixgbe_mc_array_itr(struct ixgbe_hw *hw, u8 **update_ptr, u32 *vmdq) -{ - struct ixgbe_mc_addr *mta; + /* Flow Control stats */ + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd", + CTLFLAG_RD, &stats->lxontxc, + "Link XON Transmitted"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd", + CTLFLAG_RD, &stats->lxonrxc, + "Link XON Received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd", + CTLFLAG_RD, &stats->lxofftxc, + "Link XOFF Transmitted"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd", + CTLFLAG_RD, &stats->lxoffrxc, + "Link XOFF Received"); - mta = (struct ixgbe_mc_addr *)*update_ptr; - *vmdq = mta->vmdq; + /* Packet Reception Stats */ + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_octets_rcvd", + CTLFLAG_RD, &stats->tor, + "Total Octets Received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_rcvd", + CTLFLAG_RD, &stats->gorc, + "Good Octets Received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_rcvd", + CTLFLAG_RD, &stats->tpr, + "Total Packets Received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_rcvd", + CTLFLAG_RD, &stats->gprc, + "Good Packets Received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_rcvd", + CTLFLAG_RD, &stats->mprc, + "Multicast Packets Received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_rcvd", + CTLFLAG_RD, &stats->bprc, + "Broadcast Packets Received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64", + CTLFLAG_RD, &stats->prc64, + "64 byte frames received "); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127", + CTLFLAG_RD, &stats->prc127, + "65-127 byte frames received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255", + CTLFLAG_RD, &stats->prc255, + "128-255 byte frames received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511", + CTLFLAG_RD, &stats->prc511, + "256-511 byte frames received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023", + CTLFLAG_RD, &stats->prc1023, + "512-1023 byte frames received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522", + CTLFLAG_RD, &stats->prc1522, + "1023-1522 byte frames received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersized", + CTLFLAG_RD, &stats->ruc, + "Receive Undersized"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented", + CTLFLAG_RD, &stats->rfc, + "Fragmented Packets Received "); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversized", + CTLFLAG_RD, &stats->roc, + "Oversized Packets Received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabberd", + CTLFLAG_RD, &stats->rjc, + "Received Jabber"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_rcvd", + CTLFLAG_RD, &stats->mngprc, + "Management Packets Received"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_drpd", + CTLFLAG_RD, &stats->mngptc, + "Management Packets Dropped"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "checksum_errs", + CTLFLAG_RD, &stats->xec, + "Checksum Errors"); - *update_ptr = (u8*)(mta + 1);; - return (mta->addr); + /* Packet Transmission Stats */ + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd", + CTLFLAG_RD, &stats->gotc, + "Good Octets Transmitted"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd", + CTLFLAG_RD, &stats->tpt, + "Total Packets Transmitted"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd", + CTLFLAG_RD, &stats->gptc, + "Good Packets Transmitted"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd", + CTLFLAG_RD, &stats->bptc, + "Broadcast Packets Transmitted"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd", + CTLFLAG_RD, &stats->mptc, + "Multicast Packets Transmitted"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_txd", + CTLFLAG_RD, &stats->mngptc, + "Management Packets Transmitted"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64", + CTLFLAG_RD, &stats->ptc64, + "64 byte frames transmitted "); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127", + CTLFLAG_RD, &stats->ptc127, + "65-127 byte frames transmitted"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255", + CTLFLAG_RD, &stats->ptc255, + "128-255 byte frames transmitted"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511", + CTLFLAG_RD, &stats->ptc511, + "256-511 byte frames transmitted"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023", + CTLFLAG_RD, &stats->ptc1023, + "512-1023 byte frames transmitted"); + SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522", + CTLFLAG_RD, &stats->ptc1522, + "1024-1522 byte frames transmitted"); } +/** ixgbe_sysctl_tdh_handler - Handler function + * Retrieves the TDH value from the hardware + */ +static int +ixgbe_sysctl_tdh_handler(SYSCTL_HANDLER_ARGS) +{ + int error; + + struct tx_ring *txr = ((struct tx_ring *)oidp->oid_arg1); + if (!txr) return 0; -/********************************************************************* - * Timer routine - * - * This routine checks for link status,updates statistics, - * and runs the watchdog check. - * - **********************************************************************/ + unsigned val = IXGBE_READ_REG(&txr->adapter->hw, IXGBE_TDH(txr->me)); + error = sysctl_handle_int(oidp, &val, 0, req); + if (error || !req->newptr) + return error; + return 0; +} -static void -ixgbe_local_timer(void *arg) +/** ixgbe_sysctl_tdt_handler - Handler function + * Retrieves the TDT value from the hardware + */ +static int +ixgbe_sysctl_tdt_handler(SYSCTL_HANDLER_ARGS) { - struct adapter *adapter = arg; - device_t dev = adapter->dev; - struct ix_queue *que = adapter->queues; - u64 queues = 0; - int hung = 0; + int error; - mtx_assert(&adapter->core_mtx, MA_OWNED); + struct tx_ring *txr = ((struct tx_ring *)oidp->oid_arg1); + if (!txr) return 0; - /* Check for pluggable optics */ - if (adapter->sfp_probe) - if (!ixgbe_sfp_probe(adapter)) - goto out; /* Nothing to do */ + unsigned val = IXGBE_READ_REG(&txr->adapter->hw, IXGBE_TDT(txr->me)); + error = sysctl_handle_int(oidp, &val, 0, req); + if (error || !req->newptr) + return error; + return 0; +} - ixgbe_update_link_status(adapter); - ixgbe_update_stats_counters(adapter); +/** ixgbe_sysctl_rdh_handler - Handler function + * Retrieves the RDH value from the hardware + */ +static int +ixgbe_sysctl_rdh_handler(SYSCTL_HANDLER_ARGS) +{ + int error; - /* - ** Check the TX queues status - ** - mark hung queues so we don't schedule on them - ** - watchdog only if all queues show hung - */ - for (int i = 0; i < adapter->num_queues; i++, que++) { - /* Keep track of queues with work for soft irq */ - if (que->txr->busy) - queues |= ((u64)1 << que->me); - /* - ** Each time txeof runs without cleaning, but there - ** are uncleaned descriptors it increments busy. If - ** we get to the MAX we declare it hung. - */ - if (que->busy == IXGBE_QUEUE_HUNG) { - ++hung; - /* Mark the queue as inactive */ - adapter->active_queues &= ~((u64)1 << que->me); - continue; - } else { - /* Check if we've come back from hung */ - if ((adapter->active_queues & ((u64)1 << que->me)) == 0) - adapter->active_queues |= ((u64)1 << que->me); - } - if (que->busy >= IXGBE_MAX_TX_BUSY) { - device_printf(dev,"Warning queue %d " - "appears to be hung!\n", i); - que->txr->busy = IXGBE_QUEUE_HUNG; - ++hung; - } + struct rx_ring *rxr = ((struct rx_ring *)oidp->oid_arg1); + if (!rxr) return 0; - } + unsigned val = IXGBE_READ_REG(&rxr->adapter->hw, IXGBE_RDH(rxr->me)); + error = sysctl_handle_int(oidp, &val, 0, req); + if (error || !req->newptr) + return error; + return 0; +} - /* Only truly watchdog if all queues show hung */ - if (hung == adapter->num_queues) - goto watchdog; - else if (queues != 0) { /* Force an IRQ on queues with work */ - ixgbe_rearm_queues(adapter, queues); - } +/** ixgbe_sysctl_rdt_handler - Handler function + * Retrieves the RDT value from the hardware + */ +static int +ixgbe_sysctl_rdt_handler(SYSCTL_HANDLER_ARGS) +{ + int error; -out: - callout_reset(&adapter->timer, hz, ixgbe_local_timer, adapter); - return; + struct rx_ring *rxr = ((struct rx_ring *)oidp->oid_arg1); + if (!rxr) return 0; -watchdog: - device_printf(adapter->dev, "Watchdog timeout -- resetting\n"); - adapter->ifp->if_drv_flags &= ~IFF_DRV_RUNNING; - adapter->watchdog_events++; - ixgbe_init_locked(adapter); + unsigned val = IXGBE_READ_REG(&rxr->adapter->hw, IXGBE_RDT(rxr->me)); + error = sysctl_handle_int(oidp, &val, 0, req); + if (error || !req->newptr) + return error; + return 0; } - /* -** Note: this routine updates the OS on the link state -** the real check of the hardware only happens with -** a link interrupt. +** This routine is run via an vlan config EVENT, +** it enables us to use the HW Filter table since +** we can get the vlan id. This just creates the +** entry in the soft version of the VFTA, init will +** repopulate the real table. */ static void -ixgbe_update_link_status(struct adapter *adapter) +ixgbe_if_vlan_register(if_ctx_t ctx, u16 vtag) { - struct ifnet *ifp = adapter->ifp; - device_t dev = adapter->dev; + struct adapter *adapter = iflib_get_softc(ctx); + u16 index, bit; - if (adapter->link_up){ - if (adapter->link_active == FALSE) { - if (bootverbose) - device_printf(dev,"Link is up %d Gbps %s \n", - ((adapter->link_speed == 128)? 10:1), - "Full Duplex"); - adapter->link_active = TRUE; - /* Update any Flow Control changes */ - ixgbe_fc_enable(&adapter->hw); - /* Update DMA coalescing config */ - ixgbe_config_dmac(adapter); - if_link_state_change(ifp, LINK_STATE_UP); -#ifdef PCI_IOV - ixgbe_ping_all_vfs(adapter); -#endif - } - } else { /* Link down */ - if (adapter->link_active == TRUE) { - if (bootverbose) - device_printf(dev,"Link is Down\n"); - if_link_state_change(ifp, LINK_STATE_DOWN); - adapter->link_active = FALSE; -#ifdef PCI_IOV - ixgbe_ping_all_vfs(adapter); -#endif - } - } + if ((vtag == 0) || (vtag > 4095)) /* Invalid */ + return; - return; + index = (vtag >> 5) & 0x7F; + bit = vtag & 0x1F; + adapter->shadow_vfta[index] |= (1 << bit); + ++adapter->num_vlans; + ixgbe_setup_vlan_hw_support(ctx); } - -/********************************************************************* - * - * This routine disables all traffic on the adapter by issuing a - * global reset on the MAC and deallocates TX/RX buffers. - * - **********************************************************************/ - +/* +** This routine is run via an vlan +** unconfig EVENT, remove our entry +** in the soft vfta. +*/ static void -ixgbe_stop(void *arg) +ixgbe_if_vlan_unregister(if_ctx_t ctx, u16 vtag) { - struct ifnet *ifp; - struct adapter *adapter = arg; - struct ixgbe_hw *hw = &adapter->hw; - ifp = adapter->ifp; - - mtx_assert(&adapter->core_mtx, MA_OWNED); - - INIT_DEBUGOUT("ixgbe_stop: begin\n"); - ixgbe_disable_intr(adapter); - callout_stop(&adapter->timer); - - /* Let the stack know...*/ - ifp->if_drv_flags &= ~IFF_DRV_RUNNING; - - ixgbe_reset_hw(hw); - hw->adapter_stopped = FALSE; - ixgbe_stop_adapter(hw); - if (hw->mac.type == ixgbe_mac_82599EB) - ixgbe_stop_mac_link_on_d3_82599(hw); - /* Turn off the laser - noop with no optics */ - ixgbe_disable_tx_laser(hw); - - /* Update the stack */ - adapter->link_up = FALSE; - ixgbe_update_link_status(adapter); + struct adapter *adapter = iflib_get_softc(ctx); + u16 index, bit; - /* reprogram the RAR[0] in case user changed it. */ - ixgbe_set_rar(&adapter->hw, 0, adapter->hw.mac.addr, 0, IXGBE_RAH_AV); + if ((vtag == 0) || (vtag > 4095)) /* Invalid */ + return; - return; + index = (vtag >> 5) & 0x7F; + bit = vtag & 0x1F; + adapter->shadow_vfta[index] &= ~(1 << bit); + --adapter->num_vlans; + /* Re-init to load the changes */ + ixgbe_setup_vlan_hw_support(ctx); } - -/********************************************************************* - * - * Determine hardware revision. - * - **********************************************************************/ static void -ixgbe_identify_hardware(struct adapter *adapter) +ixgbe_setup_vlan_hw_support(if_ctx_t ctx) { - device_t dev = adapter->dev; + struct adapter *adapter = iflib_get_softc(ctx); struct ixgbe_hw *hw = &adapter->hw; + struct ifnet *ifp = iflib_get_ifp(ctx); + struct rx_ring *rxr; + u32 ctrl; - /* Save off the information about this board */ - hw->vendor_id = pci_get_vendor(dev); - hw->device_id = pci_get_device(dev); - hw->revision_id = pci_read_config(dev, PCIR_REVID, 1); - hw->subsystem_vendor_id = - pci_read_config(dev, PCIR_SUBVEND_0, 2); - hw->subsystem_device_id = - pci_read_config(dev, PCIR_SUBDEV_0, 2); - - /* - ** Make sure BUSMASTER is set + /* We get here thru init_locked, meaning + ** a soft reset, this has already cleared + ** the VFTA and other state, so if there + ** have been no vlan's registered do nothing. */ - pci_enable_busmaster(dev); + if (adapter->num_vlans == 0) + return; - /* We need this here to set the num_segs below */ - ixgbe_set_mac_type(hw); + /* Setup the queues for vlans */ + for (int i = 0; i < adapter->num_queues; i++) { + rxr = &adapter->queues[i].rxr; + /* On 82599 the VLAN enable is per/queue in RXDCTL */ + if (hw->mac.type != ixgbe_mac_82598EB) { + ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me)); + ctrl |= IXGBE_RXDCTL_VME; + IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxr->me), ctrl); + } + rxr->vtag_strip = TRUE; + } - /* Pick up the 82599 settings */ - if (hw->mac.type != ixgbe_mac_82598EB) { - hw->phy.smart_speed = ixgbe_smart_speed; - adapter->num_segs = IXGBE_82599_SCATTER; - } else - adapter->num_segs = IXGBE_82598_SCATTER; + if ((ifp->if_capenable & IFCAP_VLAN_HWFILTER) == 0) + return; - return; -} + /* A soft reset zero's out the VFTA, so + ** we need to repopulate it now. + */ + for (int i = 0; i < IXGBE_VFTA_SIZE; i++) + if (adapter->shadow_vfta[i] != 0) + IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), + adapter->shadow_vfta[i]); -/********************************************************************* - * - * Determine optic type - * - **********************************************************************/ + ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL); + /* Enable the Filter Table if enabled */ + if (ifp->if_capenable & IFCAP_VLAN_HWFILTER) { + ctrl &= ~IXGBE_VLNCTRL_CFIEN; + ctrl |= IXGBE_VLNCTRL_VFE; + } + if (hw->mac.type == ixgbe_mac_82598EB) + ctrl |= IXGBE_VLNCTRL_VME; + IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl); +} + +/* +** Get the width and transaction speed of +** the slot this adapter is plugged into. +*/ static void -ixgbe_setup_optics(struct adapter *adapter) +ixgbe_get_slot_info(struct ixgbe_hw *hw) { - struct ixgbe_hw *hw = &adapter->hw; - int layer; - - layer = adapter->phy_layer = ixgbe_get_supported_physical_layer(hw); + device_t dev = ((struct adapter *)hw->back)->dev; + struct ixgbe_mac_info *mac = &hw->mac; + u16 link; + u32 offset; - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T) { - adapter->optics = IFM_10G_T; - return; + /* For most devices simply call the shared code routine */ + if (hw->device_id != IXGBE_DEV_ID_82599_SFP_SF_QP) { + ixgbe_get_bus_info(hw); + /* These devices don't use PCI-E */ + switch (hw->mac.type) { + case ixgbe_mac_X550EM_x: + return; + default: + goto display; + } } - - if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_T) { - adapter->optics = IFM_1000_T; - return; + /* + ** For the Quad port adapter we need to parse back + ** up the PCI tree to find the speed of the expansion + ** slot into which this adapter is plugged. A bit more work. + */ + dev = device_get_parent(device_get_parent(dev)); +#ifdef IXGBE_DEBUG + device_printf(dev, "parent pcib = %x,%x,%x\n", + pci_get_bus(dev), pci_get_slot(dev), pci_get_function(dev)); +#endif + dev = device_get_parent(device_get_parent(dev)); +#ifdef IXGBE_DEBUG + device_printf(dev, "slot pcib = %x,%x,%x\n", + pci_get_bus(dev), pci_get_slot(dev), pci_get_function(dev)); +#endif + /* Now get the PCI Express Capabilities offset */ + pci_find_cap(dev, PCIY_EXPRESS, &offset); + /* ...and read the Link Status Register */ + link = pci_read_config(dev, offset + PCIER_LINK_STA, 2); + switch (link & IXGBE_PCI_LINK_WIDTH) { + case IXGBE_PCI_LINK_WIDTH_1: + hw->bus.width = ixgbe_bus_width_pcie_x1; + break; + case IXGBE_PCI_LINK_WIDTH_2: + hw->bus.width = ixgbe_bus_width_pcie_x2; + break; + case IXGBE_PCI_LINK_WIDTH_4: + hw->bus.width = ixgbe_bus_width_pcie_x4; + break; + case IXGBE_PCI_LINK_WIDTH_8: + hw->bus.width = ixgbe_bus_width_pcie_x8; + break; + default: + hw->bus.width = ixgbe_bus_width_unknown; + break; } - - if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX) { - adapter->optics = IFM_1000_SX; - return; + switch (link & IXGBE_PCI_LINK_SPEED) { + case IXGBE_PCI_LINK_SPEED_2500: + hw->bus.speed = ixgbe_bus_speed_2500; + break; + case IXGBE_PCI_LINK_SPEED_5000: + hw->bus.speed = ixgbe_bus_speed_5000; + break; + case IXGBE_PCI_LINK_SPEED_8000: + hw->bus.speed = ixgbe_bus_speed_8000; + break; + default: + hw->bus.speed = ixgbe_bus_speed_unknown; + break; } - if (layer & (IXGBE_PHYSICAL_LAYER_10GBASE_LR | - IXGBE_PHYSICAL_LAYER_10GBASE_LRM)) { - adapter->optics = IFM_10G_LR; - return; - } + mac->ops.set_lan_id(hw); - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR) { - adapter->optics = IFM_10G_SR; - return; - } +display: + device_printf(dev,"PCI Express Bus: Speed %s %s\n", + ((hw->bus.speed == ixgbe_bus_speed_8000) ? "8.0GT/s": + (hw->bus.speed == ixgbe_bus_speed_5000) ? "5.0GT/s": + (hw->bus.speed == ixgbe_bus_speed_2500) ? "2.5GT/s":"Unknown"), + (hw->bus.width == ixgbe_bus_width_pcie_x8) ? "Width x8" : + (hw->bus.width == ixgbe_bus_width_pcie_x4) ? "Width x4" : + (hw->bus.width == ixgbe_bus_width_pcie_x1) ? "Width x1" : + ("Unknown")); - if (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU) { - adapter->optics = IFM_10G_TWINAX; - return; + if ((hw->device_id != IXGBE_DEV_ID_82599_SFP_SF_QP) && + ((hw->bus.width <= ixgbe_bus_width_pcie_x4) && + (hw->bus.speed == ixgbe_bus_speed_2500))) { + device_printf(dev, "PCI-Express bandwidth available" + " for this card\n is not sufficient for" + " optimal performance.\n"); + device_printf(dev, "For optimal performance a x8 " + "PCIE, or x4 PCIE Gen2 slot is required.\n"); } - - if (layer & (IXGBE_PHYSICAL_LAYER_10GBASE_KX4 | - IXGBE_PHYSICAL_LAYER_10GBASE_CX4)) { - adapter->optics = IFM_10G_CX4; - return; + if ((hw->device_id == IXGBE_DEV_ID_82599_SFP_SF_QP) && + ((hw->bus.width <= ixgbe_bus_width_pcie_x8) && + (hw->bus.speed < ixgbe_bus_speed_8000))) { + device_printf(dev, "PCI-Express bandwidth available" + " for this card\n is not sufficient for" + " optimal performance.\n"); + device_printf(dev, "For optimal performance a x8 " + "PCIE Gen3 slot is required.\n"); } - /* If we get here just set the default */ - adapter->optics = IFM_ETHER | IFM_AUTO; return; } - -/********************************************************************* - * - * Setup the Legacy or MSI Interrupt handler - * - **********************************************************************/ -static int -ixgbe_allocate_legacy(struct adapter *adapter) -{ - device_t dev = adapter->dev; - struct ix_queue *que = adapter->queues; -#ifndef IXGBE_LEGACY_TX - struct tx_ring *txr = adapter->tx_rings; -#endif - int error, rid = 0; - - /* MSI RID at 1 */ - if (adapter->msix == 1) - rid = 1; - - /* We allocate a single interrupt resource */ - adapter->res = bus_alloc_resource_any(dev, - SYS_RES_IRQ, &rid, RF_SHAREABLE | RF_ACTIVE); - if (adapter->res == NULL) { - device_printf(dev, "Unable to allocate bus resource: " - "interrupt\n"); - return (ENXIO); - } - - /* - * Try allocating a fast interrupt and the associated deferred - * processing contexts. - */ -#ifndef IXGBE_LEGACY_TX - TASK_INIT(&txr->txq_task, 0, ixgbe_deferred_mq_start, txr); -#endif - TASK_INIT(&que->que_task, 0, ixgbe_handle_que, que); - que->tq = taskqueue_create_fast("ixgbe_que", M_NOWAIT, - taskqueue_thread_enqueue, &que->tq); - taskqueue_start_threads(&que->tq, 1, PI_NET, "%s ixq", - device_get_nameunit(adapter->dev)); - - /* Tasklets for Link, SFP and Multispeed Fiber */ - TASK_INIT(&adapter->link_task, 0, ixgbe_handle_link, adapter); - TASK_INIT(&adapter->mod_task, 0, ixgbe_handle_mod, adapter); - TASK_INIT(&adapter->msf_task, 0, ixgbe_handle_msf, adapter); - TASK_INIT(&adapter->phy_task, 0, ixgbe_handle_phy, adapter); -#ifdef IXGBE_FDIR - TASK_INIT(&adapter->fdir_task, 0, ixgbe_reinit_fdir, adapter); -#endif - adapter->tq = taskqueue_create_fast("ixgbe_link", M_NOWAIT, - taskqueue_thread_enqueue, &adapter->tq); - taskqueue_start_threads(&adapter->tq, 1, PI_NET, "%s linkq", - device_get_nameunit(adapter->dev)); - - if ((error = bus_setup_intr(dev, adapter->res, - INTR_TYPE_NET | INTR_MPSAFE, NULL, ixgbe_legacy_irq, - que, &adapter->tag)) != 0) { - device_printf(dev, "Failed to register fast interrupt " - "handler: %d\n", error); - taskqueue_free(que->tq); - taskqueue_free(adapter->tq); - que->tq = NULL; - adapter->tq = NULL; - return (error); - } - /* For simplicity in the handlers */ - adapter->active_queues = IXGBE_EIMS_ENABLE_MASK; - - return (0); -} - - + /********************************************************************* * * Setup MSIX Interrupt resources and handlers * **********************************************************************/ static int -ixgbe_allocate_msix(struct adapter *adapter) +ixgbe_if_msix_intr_assign(if_ctx_t ctx, int msix) { - device_t dev = adapter->dev; + struct adapter *adapter = iflib_get_softc(ctx); struct ix_queue *que = adapter->queues; - struct tx_ring *txr = adapter->tx_rings; int error, rid, vector = 0; int cpu_id = 0; -#ifdef RSS - cpuset_t cpu_mask; -#endif - -#ifdef RSS - /* - * If we're doing RSS, the number of queues needs to - * match the number of RSS buckets that are configured. - * - * + If there's more queues than RSS buckets, we'll end - * up with queues that get no traffic. - * - * + If there's more RSS buckets than queues, we'll end - * up having multiple RSS buckets map to the same queue, - * so there'll be some contention. - */ - if (adapter->num_queues != rss_getnumbuckets()) { - device_printf(dev, - "%s: number of queues (%d) != number of RSS buckets (%d)" - "; performance will be impacted.\n", - __func__, - adapter->num_queues, - rss_getnumbuckets()); - } -#endif - for (int i = 0; i < adapter->num_queues; i++, vector++, que++, txr++) { + /* Admin Que is vector 0*/ + rid = vector + 1; + for (int i = 0; i < adapter->num_queues; i++, vector++, que++) { + char buf[16]; rid = vector + 1; - que->res = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid, - RF_SHAREABLE | RF_ACTIVE); - if (que->res == NULL) { - device_printf(dev,"Unable to allocate" - " bus resource: que interrupt [%d]\n", vector); - return (ENXIO); - } - /* Set the handler function */ - error = bus_setup_intr(dev, que->res, - INTR_TYPE_NET | INTR_MPSAFE, NULL, - ixgbe_msix_que, que, &que->tag); + + snprintf(buf, sizeof(buf), "rxq%d", i); + error = iflib_irq_alloc_generic(ctx, &que->que_irq, rid, IFLIB_INTR_RX, + ixgbe_msix_que, que, que->me, buf); + if (error) { - que->res = NULL; - device_printf(dev, "Failed to register QUE handler"); - return (error); + device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error); + adapter->num_queues = i + 1; + goto fail; } + + snprintf(buf, sizeof(buf), "txq%d", i); + iflib_softirq_alloc_generic(ctx, rid, IFLIB_INTR_TX, que, que->me, buf); + + /* #if __FreeBSD_version >= 800504 - bus_describe_intr(dev, que->res, que->tag, "q%d", i); + bus_describe_intr(dev, que->res, que->tag, "que %d", i); #endif + */ que->msix = vector; adapter->active_queues |= (u64)(1 << que->msix); #ifdef RSS @@ -2537,2130 +2008,1838 @@ if (adapter->num_queues > 1) cpu_id = i; #endif - if (adapter->num_queues > 1) - bus_bind_intr(dev, que->res, cpu_id); -#ifdef IXGBE_DEBUG -#ifdef RSS - device_printf(dev, - "Bound RSS bucket %d to CPU %d\n", - i, cpu_id); -#else - device_printf(dev, - "Bound queue %d to cpu %d\n", - i, cpu_id); -#endif -#endif /* IXGBE_DEBUG */ - -#ifndef IXGBE_LEGACY_TX - TASK_INIT(&txr->txq_task, 0, ixgbe_deferred_mq_start, txr); -#endif - TASK_INIT(&que->que_task, 0, ixgbe_handle_que, que); - que->tq = taskqueue_create_fast("ixgbe_que", M_NOWAIT, - taskqueue_thread_enqueue, &que->tq); -#ifdef RSS - CPU_SETOF(cpu_id, &cpu_mask); - taskqueue_start_threads_cpuset(&que->tq, 1, PI_NET, - &cpu_mask, - "%s (bucket %d)", - device_get_nameunit(adapter->dev), - cpu_id); -#else - taskqueue_start_threads(&que->tq, 1, PI_NET, "%s:q%d", - device_get_nameunit(adapter->dev), i); -#endif } - - /* and Link */ rid = vector + 1; - adapter->res = bus_alloc_resource_any(dev, - SYS_RES_IRQ, &rid, RF_SHAREABLE | RF_ACTIVE); - if (!adapter->res) { - device_printf(dev,"Unable to allocate" - " bus resource: Link interrupt [%d]\n", rid); - return (ENXIO); - } - /* Set the link handler function */ - error = bus_setup_intr(dev, adapter->res, - INTR_TYPE_NET | INTR_MPSAFE, NULL, - ixgbe_msix_link, adapter, &adapter->tag); + error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, + ixgbe_msix_link, adapter, 0, "aq"); if (error) { - adapter->res = NULL; - device_printf(dev, "Failed to register LINK handler"); + device_printf(iflib_get_dev(ctx), "Failed to register admin handler"); return (error); } -#if __FreeBSD_version >= 800504 - bus_describe_intr(dev, adapter->res, adapter->tag, "link"); -#endif - adapter->vector = vector; - /* Tasklets for Link, SFP and Multispeed Fiber */ - TASK_INIT(&adapter->link_task, 0, ixgbe_handle_link, adapter); - TASK_INIT(&adapter->mod_task, 0, ixgbe_handle_mod, adapter); - TASK_INIT(&adapter->msf_task, 0, ixgbe_handle_msf, adapter); -#ifdef PCI_IOV - TASK_INIT(&adapter->mbx_task, 0, ixgbe_handle_mbx, adapter); -#endif - TASK_INIT(&adapter->phy_task, 0, ixgbe_handle_phy, adapter); -#ifdef IXGBE_FDIR - TASK_INIT(&adapter->fdir_task, 0, ixgbe_reinit_fdir, adapter); -#endif - adapter->tq = taskqueue_create_fast("ixgbe_link", M_NOWAIT, - taskqueue_thread_enqueue, &adapter->tq); - taskqueue_start_threads(&adapter->tq, 1, PI_NET, "%s linkq", - device_get_nameunit(adapter->dev)); - - return (0); -} - -/* - * Setup Either MSI/X or MSI - */ -static int -ixgbe_setup_msix(struct adapter *adapter) -{ - device_t dev = adapter->dev; - int rid, want, queues, msgs; - - /* Override by tuneable */ - if (ixgbe_enable_msix == 0) - goto msi; - - /* First try MSI/X */ - msgs = pci_msix_count(dev); - if (msgs == 0) - goto msi; - rid = PCIR_BAR(MSIX_82598_BAR); - adapter->msix_mem = bus_alloc_resource_any(dev, - SYS_RES_MEMORY, &rid, RF_ACTIVE); - if (adapter->msix_mem == NULL) { - rid += 4; /* 82599 maps in higher BAR */ - adapter->msix_mem = bus_alloc_resource_any(dev, - SYS_RES_MEMORY, &rid, RF_ACTIVE); - } - if (adapter->msix_mem == NULL) { - /* May not be enabled */ - device_printf(adapter->dev, - "Unable to map MSIX table \n"); - goto msi; - } - - /* Figure out a reasonable auto config value */ - queues = (mp_ncpus > (msgs - 1)) ? (msgs - 1) : mp_ncpus; - -#ifdef RSS - /* If we're doing RSS, clamp at the number of RSS buckets */ - if (queues > rss_getnumbuckets()) - queues = rss_getnumbuckets(); -#endif - - if (ixgbe_num_queues != 0) - queues = ixgbe_num_queues; - /* Set max queues to 8 when autoconfiguring */ - else if ((ixgbe_num_queues == 0) && (queues > 8)) - queues = 8; - - /* reflect correct sysctl value */ - ixgbe_num_queues = queues; - - /* - ** Want one vector (RX/TX pair) per queue - ** plus an additional for Link. - */ - want = queues + 1; - if (msgs >= want) - msgs = want; - else { - device_printf(adapter->dev, - "MSIX Configuration Problem, " - "%d vectors but %d queues wanted!\n", - msgs, want); - goto msi; - } - if ((pci_alloc_msix(dev, &msgs) == 0) && (msgs == want)) { - device_printf(adapter->dev, - "Using MSIX interrupts with %d vectors\n", msgs); - adapter->num_queues = queues; - return (msgs); - } - /* - ** If MSIX alloc failed or provided us with - ** less than needed, free and fall through to MSI - */ - pci_release_msi(dev); - -msi: - if (adapter->msix_mem != NULL) { - bus_release_resource(dev, SYS_RES_MEMORY, - rid, adapter->msix_mem); - adapter->msix_mem = NULL; - } - msgs = 1; - if (pci_alloc_msi(dev, &msgs) == 0) { - device_printf(adapter->dev, "Using an MSI interrupt\n"); - return (msgs); - } - device_printf(adapter->dev, "Using a Legacy interrupt\n"); - return (0); -} - - -static int -ixgbe_allocate_pci_resources(struct adapter *adapter) -{ - int rid; - device_t dev = adapter->dev; - - rid = PCIR_BAR(0); - adapter->pci_mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY, - &rid, RF_ACTIVE); - - if (!(adapter->pci_mem)) { - device_printf(dev, "Unable to allocate bus resource: memory\n"); - return (ENXIO); - } - /* Save bus_space values for READ/WRITE_REG macros */ - adapter->osdep.mem_bus_space_tag = - rman_get_bustag(adapter->pci_mem); - adapter->osdep.mem_bus_space_handle = - rman_get_bushandle(adapter->pci_mem); - /* Set hw values for shared code */ - adapter->hw.hw_addr = (u8 *) &adapter->osdep.mem_bus_space_handle; - adapter->hw.back = adapter; - /* Default to 1 queue if MSI-X setup fails */ - adapter->num_queues = 1; - - /* - ** Now setup MSI or MSI-X, should - ** return us the number of supported - ** vectors. (Will be 1 for MSI) - */ - adapter->msix = ixgbe_setup_msix(adapter); + adapter->vector = vector; return (0); -} - -static void -ixgbe_free_pci_resources(struct adapter * adapter) -{ - struct ix_queue *que = adapter->queues; - device_t dev = adapter->dev; - int rid, memrid; - - if (adapter->hw.mac.type == ixgbe_mac_82598EB) - memrid = PCIR_BAR(MSIX_82598_BAR); - else - memrid = PCIR_BAR(MSIX_82599_BAR); - - /* - ** There is a slight possibility of a failure mode - ** in attach that will result in entering this function - ** before interrupt resources have been initialized, and - ** in that case we do not want to execute the loops below - ** We can detect this reliably by the state of the adapter - ** res pointer. - */ - if (adapter->res == NULL) - goto mem; - - /* - ** Release all msix queue resources: - */ - for (int i = 0; i < adapter->num_queues; i++, que++) { - rid = que->msix + 1; - if (que->tag != NULL) { - bus_teardown_intr(dev, que->res, que->tag); - que->tag = NULL; - } - if (que->res != NULL) - bus_release_resource(dev, SYS_RES_IRQ, rid, que->res); - } - - - /* Clean the Legacy or Link interrupt last */ - if (adapter->vector) /* we are doing MSIX */ - rid = adapter->vector + 1; - else - (adapter->msix != 0) ? (rid = 1):(rid = 0); - - if (adapter->tag != NULL) { - bus_teardown_intr(dev, adapter->res, adapter->tag); - adapter->tag = NULL; - } - if (adapter->res != NULL) - bus_release_resource(dev, SYS_RES_IRQ, rid, adapter->res); - -mem: - if (adapter->msix) - pci_release_msi(dev); - - if (adapter->msix_mem != NULL) - bus_release_resource(dev, SYS_RES_MEMORY, - memrid, adapter->msix_mem); - - if (adapter->pci_mem != NULL) - bus_release_resource(dev, SYS_RES_MEMORY, - PCIR_BAR(0), adapter->pci_mem); - - return; +fail: + iflib_irq_free(ctx, &adapter->irq); + que = adapter->queues; + for (int i = 0; i < adapter->num_queues; i++, que++) + iflib_irq_free(ctx, &que->que_irq); + return (error); } /********************************************************************* * - * Setup networking device structure and register an interface. + * MSIX Queue Interrupt Service routine * **********************************************************************/ static int -ixgbe_setup_interface(device_t dev, struct adapter *adapter) +ixgbe_msix_que(void *arg) { - struct ifnet *ifp; + struct ix_queue *que = arg; + struct adapter *adapter = que->adapter; + struct tx_ring *txr = &que->txr; + struct rx_ring *rxr = &que->rxr; + struct ifnet *ifp = iflib_get_ifp(que->adapter->ctx); + u32 newitr = 0; - INIT_DEBUGOUT("ixgbe_setup_interface: begin"); + /* Protect against spurious interrupts */ + if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) + return 0; - ifp = adapter->ifp = if_alloc(IFT_ETHER); - if (ifp == NULL) { - device_printf(dev, "can not allocate ifnet structure\n"); - return (-1); + ixgbe_disable_queue(adapter, que->msix); + ++que->irqs; + + if (ixgbe_enable_aim == FALSE) + goto no_calc; + /* + ** Do Adaptive Interrupt Moderation: + ** - Write out last calculated setting + ** - Calculate based on average size over + ** the last interval. + */ + if (que->eitr_setting) { + IXGBE_WRITE_REG(&adapter->hw, IXGBE_EITR(que->msix), que->eitr_setting); } - if_initname(ifp, device_get_name(dev), device_get_unit(dev)); - ifp->if_baudrate = IF_Gbps(10); - ifp->if_init = ixgbe_init; - ifp->if_softc = adapter; - ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; - ifp->if_ioctl = ixgbe_ioctl; -#if __FreeBSD_version >= 1100036 - if_setgetcounterfn(ifp, ixgbe_get_counter); -#endif -#if __FreeBSD_version >= 1100045 - /* TSO parameters */ - ifp->if_hw_tsomax = 65518; - ifp->if_hw_tsomaxsegcount = IXGBE_82599_SCATTER; - ifp->if_hw_tsomaxsegsize = 2048; -#endif -#ifndef IXGBE_LEGACY_TX - ifp->if_transmit = ixgbe_mq_start; - ifp->if_qflush = ixgbe_qflush; -#else - ifp->if_start = ixgbe_start; - IFQ_SET_MAXLEN(&ifp->if_snd, adapter->num_tx_desc - 2); - ifp->if_snd.ifq_drv_maxlen = adapter->num_tx_desc - 2; - IFQ_SET_READY(&ifp->if_snd); -#endif - - ether_ifattach(ifp, adapter->hw.mac.addr); - - adapter->max_frame_size = - ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN; + + que->eitr_setting = 0; - /* - * Tell the upper layer(s) we support long frames. - */ - ifp->if_hdrlen = sizeof(struct ether_vlan_header); - - /* Set capability flags */ - ifp->if_capabilities |= IFCAP_RXCSUM - | IFCAP_TXCSUM - | IFCAP_RXCSUM_IPV6 - | IFCAP_TXCSUM_IPV6 - | IFCAP_TSO4 - | IFCAP_TSO6 - | IFCAP_LRO - | IFCAP_VLAN_HWTAGGING - | IFCAP_VLAN_HWTSO - | IFCAP_VLAN_HWCSUM - | IFCAP_JUMBO_MTU - | IFCAP_VLAN_MTU - | IFCAP_HWSTATS; - - /* Enable the above capabilities by default */ - ifp->if_capenable = ifp->if_capabilities; + /* Idle, do nothing */ + if ((txr->bytes == 0) && (rxr->bytes == 0)) + goto no_calc; + + if ((txr->bytes) && (txr->packets)) + newitr = txr->bytes/txr->packets; + if ((rxr->bytes) && (rxr->packets)) + newitr = max(newitr, + (rxr->bytes / rxr->packets)); + newitr += 24; /* account for hardware frame, crc */ - /* - ** Don't turn this on by default, if vlans are - ** created on another pseudo device (eg. lagg) - ** then vlan events are not passed thru, breaking - ** operation, but with HW FILTER off it works. If - ** using vlans directly on the ixgbe driver you can - ** enable this and get full hardware tag filtering. - */ - ifp->if_capabilities |= IFCAP_VLAN_HWFILTER; + /* set an upper boundary */ + newitr = min(newitr, 3000); - /* - * Specify the media types supported by this adapter and register - * callbacks to update media and link information - */ - ifmedia_init(&adapter->media, IFM_IMASK, ixgbe_media_change, - ixgbe_media_status); + /* Be nice to the mid range */ + if ((newitr > 300) && (newitr < 1200)) + newitr = (newitr / 3); + else + newitr = (newitr / 2); - adapter->phy_layer = ixgbe_get_supported_physical_layer(&adapter->hw); - ixgbe_add_media_types(adapter); + if (adapter->hw.mac.type == ixgbe_mac_82598EB) + newitr |= newitr << 16; + else + newitr |= IXGBE_EITR_CNT_WDIS; + + /* save for next interrupt */ + que->eitr_setting = newitr; - /* Set autoselect media by default */ - ifmedia_set(&adapter->media, IFM_ETHER | IFM_AUTO); + /* Reset state */ + txr->bytes = 0; + txr->packets = 0; + rxr->bytes = 0; + rxr->packets = 0; - return (0); +no_calc: + return (FILTER_SCHEDULE_THREAD); } +/********************************************************************* + * + * Media Ioctl callback + * + * This routine is called whenever the user queries the status of + * the interface using ifconfig. + * + **********************************************************************/ static void -ixgbe_add_media_types(struct adapter *adapter) -{ - struct ixgbe_hw *hw = &adapter->hw; - device_t dev = adapter->dev; +ixgbe_if_media_status(if_ctx_t ctx, struct ifmediareq * ifmr) +{ + struct adapter *adapter = iflib_get_softc(ctx); + struct ixgbe_hw *hw = &adapter->hw; int layer; + INIT_DEBUGOUT("ixgbe_if_media_status: begin"); + ixgbe_if_update_admin_status(ctx); + + ifmr->ifm_status = IFM_AVALID; + ifmr->ifm_active = IFM_ETHER; + + if (!adapter->link_active) { + return; + } + + ifmr->ifm_status |= IFM_ACTIVE; layer = adapter->phy_layer; - /* Media types with matching FreeBSD media defines */ - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T) - ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_T, 0, NULL); - if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_T) - ifmedia_add(&adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL); - if (layer & IXGBE_PHYSICAL_LAYER_100BASE_TX) - ifmedia_add(&adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL); - + if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T || + layer & IXGBE_PHYSICAL_LAYER_1000BASE_T || + layer & IXGBE_PHYSICAL_LAYER_100BASE_TX) + switch (adapter->link_speed) { + case IXGBE_LINK_SPEED_10GB_FULL: + ifmr->ifm_active |= IFM_10G_T | IFM_FDX; + break; + case IXGBE_LINK_SPEED_1GB_FULL: + ifmr->ifm_active |= IFM_1000_T | IFM_FDX; + break; + case IXGBE_LINK_SPEED_100_FULL: + ifmr->ifm_active |= IFM_100_TX | IFM_FDX; + break; + } if (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU || layer & IXGBE_PHYSICAL_LAYER_SFP_ACTIVE_DA) - ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_TWINAX, 0, NULL); - - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LR) { - ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_LR, 0, NULL); - if (hw->phy.multispeed_fiber) - ifmedia_add(&adapter->media, IFM_ETHER | IFM_1000_LX, 0, NULL); - } - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR) { - ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_SR, 0, NULL); - if (hw->phy.multispeed_fiber) - ifmedia_add(&adapter->media, IFM_ETHER | IFM_1000_SX, 0, NULL); - } else if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX) - ifmedia_add(&adapter->media, IFM_ETHER | IFM_1000_SX, 0, NULL); + switch (adapter->link_speed) { + case IXGBE_LINK_SPEED_10GB_FULL: + ifmr->ifm_active |= IFM_10G_TWINAX | IFM_FDX; + break; + } + if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LR) + switch (adapter->link_speed) { + case IXGBE_LINK_SPEED_10GB_FULL: + ifmr->ifm_active |= IFM_10G_LR | IFM_FDX; + break; + case IXGBE_LINK_SPEED_1GB_FULL: + ifmr->ifm_active |= IFM_1000_LX | IFM_FDX; + break; + } + if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LRM) + switch (adapter->link_speed) { + case IXGBE_LINK_SPEED_10GB_FULL: + ifmr->ifm_active |= IFM_10G_LRM | IFM_FDX; + break; + case IXGBE_LINK_SPEED_1GB_FULL: + ifmr->ifm_active |= IFM_1000_LX | IFM_FDX; + break; + } + if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR || + layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX) + switch (adapter->link_speed) { + case IXGBE_LINK_SPEED_10GB_FULL: + ifmr->ifm_active |= IFM_10G_SR | IFM_FDX; + break; + case IXGBE_LINK_SPEED_1GB_FULL: + ifmr->ifm_active |= IFM_1000_SX | IFM_FDX; + break; + } if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_CX4) - ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_CX4, 0, NULL); - -#ifdef IFM_ETH_XTYPE + switch (adapter->link_speed) { + case IXGBE_LINK_SPEED_10GB_FULL: + ifmr->ifm_active |= IFM_10G_CX4 | IFM_FDX; + break; + } + /* + ** XXX: These need to use the proper media types once + ** they're added. + */ if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR) - ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_KR, 0, NULL); - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4) - ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_KX4, 0, NULL); - if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX) - ifmedia_add(&adapter->media, IFM_ETHER | IFM_1000_KX, 0, NULL); -#else - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR) { - device_printf(dev, "Media supported: 10GbaseKR\n"); - device_printf(dev, "10GbaseKR mapped to 10GbaseSR\n"); - ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_SR, 0, NULL); - } - if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4) { - device_printf(dev, "Media supported: 10GbaseKX4\n"); - device_printf(dev, "10GbaseKX4 mapped to 10GbaseCX4\n"); - ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_CX4, 0, NULL); - } - if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX) { - device_printf(dev, "Media supported: 1000baseKX\n"); - device_printf(dev, "1000baseKX mapped to 1000baseCX\n"); - ifmedia_add(&adapter->media, IFM_ETHER | IFM_1000_CX, 0, NULL); - } -#endif - if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_BX) - device_printf(dev, "Media supported: 1000baseBX\n"); + switch (adapter->link_speed) { + case IXGBE_LINK_SPEED_10GB_FULL: + ifmr->ifm_active |= IFM_10G_SR | IFM_FDX; + break; + case IXGBE_LINK_SPEED_2_5GB_FULL: + ifmr->ifm_active |= IFM_2500_SX | IFM_FDX; + break; + case IXGBE_LINK_SPEED_1GB_FULL: + ifmr->ifm_active |= IFM_1000_CX | IFM_FDX; + break; + } + else if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4 + || layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX) + switch (adapter->link_speed) { + case IXGBE_LINK_SPEED_10GB_FULL: + ifmr->ifm_active |= IFM_10G_CX4 | IFM_FDX; + break; + case IXGBE_LINK_SPEED_2_5GB_FULL: + ifmr->ifm_active |= IFM_2500_SX | IFM_FDX; + break; + case IXGBE_LINK_SPEED_1GB_FULL: + ifmr->ifm_active |= IFM_1000_CX | IFM_FDX; + break; + } - if (hw->device_id == IXGBE_DEV_ID_82598AT) { - ifmedia_add(&adapter->media, - IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL); - ifmedia_add(&adapter->media, - IFM_ETHER | IFM_1000_T, 0, NULL); - } - - ifmedia_add(&adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL); -} - -static void -ixgbe_config_link(struct adapter *adapter) -{ - struct ixgbe_hw *hw = &adapter->hw; - u32 autoneg, err = 0; - bool sfp, negotiate; - - sfp = ixgbe_is_sfp(hw); - - if (sfp) { - taskqueue_enqueue(adapter->tq, &adapter->mod_task); - } else { - if (hw->mac.ops.check_link) - err = ixgbe_check_link(hw, &adapter->link_speed, - &adapter->link_up, FALSE); - if (err) - goto out; - autoneg = hw->phy.autoneg_advertised; - if ((!autoneg) && (hw->mac.ops.get_link_capabilities)) - err = hw->mac.ops.get_link_capabilities(hw, - &autoneg, &negotiate); - if (err) - goto out; - if (hw->mac.ops.setup_link) - err = hw->mac.ops.setup_link(hw, - autoneg, adapter->link_up); - } -out: - return; -} - + /* If nothing is recognized... */ + if (IFM_SUBTYPE(ifmr->ifm_active) == 0) + ifmr->ifm_active |= IFM_UNKNOWN; + +#if __FreeBSD_version >= 900025 + /* Display current flow control setting used on link */ + if (hw->fc.current_mode == ixgbe_fc_rx_pause || + hw->fc.current_mode == ixgbe_fc_full) + ifmr->ifm_active |= IFM_ETH_RXPAUSE; + if (hw->fc.current_mode == ixgbe_fc_tx_pause || + hw->fc.current_mode == ixgbe_fc_full) + ifmr->ifm_active |= IFM_ETH_TXPAUSE; +#endif +} /********************************************************************* * - * Enable transmit units. + * Media Ioctl callback + * + * This routine is called when the user changes speed/duplex using + * media/mediopt option with ifconfig. * **********************************************************************/ -static void -ixgbe_initialize_transmit_units(struct adapter *adapter) +static int +ixgbe_if_media_change(if_ctx_t ctx) { - struct tx_ring *txr = adapter->tx_rings; - struct ixgbe_hw *hw = &adapter->hw; + struct adapter *adapter = iflib_get_softc(ctx); + struct ifmedia *ifm = iflib_get_media(ctx); + struct ixgbe_hw *hw = &adapter->hw; + ixgbe_link_speed speed = 0; - /* Setup the Base and Length of the Tx Descriptor Ring */ - for (int i = 0; i < adapter->num_queues; i++, txr++) { - u64 tdba = txr->txdma.dma_paddr; - u32 txctrl = 0; - int j = txr->me; + INIT_DEBUGOUT("ixgbe_if_media_change: begin"); - IXGBE_WRITE_REG(hw, IXGBE_TDBAL(j), - (tdba & 0x00000000ffffffffULL)); - IXGBE_WRITE_REG(hw, IXGBE_TDBAH(j), (tdba >> 32)); - IXGBE_WRITE_REG(hw, IXGBE_TDLEN(j), - adapter->num_tx_desc * sizeof(union ixgbe_adv_tx_desc)); + if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER) + return (EINVAL); - /* Setup the HW Tx Head and Tail descriptor pointers */ - IXGBE_WRITE_REG(hw, IXGBE_TDH(j), 0); - IXGBE_WRITE_REG(hw, IXGBE_TDT(j), 0); + if (hw->phy.media_type == ixgbe_media_type_backplane) + return (EPERM); - /* Cache the tail address */ - txr->tail = IXGBE_TDT(j); + /* + ** We don't actually need to check against the supported + ** media types of the adapter; ifmedia will take care of + ** that for us. + */ + switch (IFM_SUBTYPE(ifm->ifm_media)) { + case IFM_AUTO: + case IFM_10G_T: + speed |= IXGBE_LINK_SPEED_100_FULL; + case IFM_10G_LRM: + case IFM_10G_SR: /* KR, too */ + case IFM_10G_LR: + case IFM_10G_CX4: /* KX4 */ + speed |= IXGBE_LINK_SPEED_1GB_FULL; + case IFM_10G_TWINAX: + speed |= IXGBE_LINK_SPEED_10GB_FULL; + break; + case IFM_1000_T: + speed |= IXGBE_LINK_SPEED_100_FULL; + case IFM_1000_LX: + case IFM_1000_SX: + case IFM_1000_CX: /* KX */ + speed |= IXGBE_LINK_SPEED_1GB_FULL; + break; + case IFM_100_TX: + speed |= IXGBE_LINK_SPEED_100_FULL; + break; + default: + goto invalid; + } - /* Disable Head Writeback */ - /* - * Note: for X550 series devices, these registers are actually - * prefixed with TPH_ isntead of DCA_, but the addresses and - * fields remain the same. - */ - switch (hw->mac.type) { - case ixgbe_mac_82598EB: - txctrl = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL(j)); - break; - default: - txctrl = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL_82599(j)); - break; - } - txctrl &= ~IXGBE_DCA_TXCTRL_DESC_WRO_EN; - switch (hw->mac.type) { - case ixgbe_mac_82598EB: - IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL(j), txctrl); - break; - default: - IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL_82599(j), txctrl); - break; - } + hw->mac.autotry_restart = TRUE; + hw->mac.ops.setup_link(hw, speed, TRUE); + adapter->advertise = + ((speed & IXGBE_LINK_SPEED_10GB_FULL) << 2) | + ((speed & IXGBE_LINK_SPEED_1GB_FULL) << 1) | + ((speed & IXGBE_LINK_SPEED_100_FULL) << 0); - } + return (0); - if (hw->mac.type != ixgbe_mac_82598EB) { - u32 dmatxctl, rttdcs; -#ifdef PCI_IOV - enum ixgbe_iov_mode mode = ixgbe_get_iov_mode(adapter); -#endif - dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL); - dmatxctl |= IXGBE_DMATXCTL_TE; - IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl); - /* Disable arbiter to set MTQC */ - rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS); - rttdcs |= IXGBE_RTTDCS_ARBDIS; - IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs); -#ifdef PCI_IOV - IXGBE_WRITE_REG(hw, IXGBE_MTQC, ixgbe_get_mtqc(mode)); -#else - IXGBE_WRITE_REG(hw, IXGBE_MTQC, IXGBE_MTQC_64Q_1PB); -#endif - rttdcs &= ~IXGBE_RTTDCS_ARBDIS; - IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs); - } +invalid: + device_printf(iflib_get_dev(ctx), "Invalid media type!\n"); + return (EINVAL); - return; } - -static void -ixgbe_initialize_rss_mapping(struct adapter *adapter) + +static int +ixgbe_if_promisc_set(if_ctx_t ctx, int flags) { - struct ixgbe_hw *hw = &adapter->hw; - u32 reta = 0, mrqc, rss_key[10]; - int queue_id, table_size, index_mult; -#ifdef RSS - u32 rss_hash_config; -#endif -#ifdef PCI_IOV - enum ixgbe_iov_mode mode; -#endif - -#ifdef RSS - /* Fetch the configured RSS key */ - rss_getkey((uint8_t *) &rss_key); -#else - /* set up random bits */ - arc4rand(&rss_key, sizeof(rss_key), 0); -#endif + struct adapter *adapter = iflib_get_softc(ctx); + struct ifnet *ifp = iflib_get_ifp(ctx); + u_int32_t reg_rctl; + int mcnt = 0; - /* Set multiplier for RETA setup and table size based on MAC */ - index_mult = 0x1; - table_size = 128; - switch (adapter->hw.mac.type) { - case ixgbe_mac_82598EB: - index_mult = 0x11; - break; - case ixgbe_mac_X550: - case ixgbe_mac_X550EM_x: - table_size = 512; - break; - default: - break; + reg_rctl = IXGBE_READ_REG(&adapter->hw, IXGBE_FCTRL); + reg_rctl &= (~IXGBE_FCTRL_UPE); + if (ifp->if_flags & IFF_ALLMULTI) + mcnt = MAX_NUM_MULTICAST_ADDRESSES; + else { + mcnt = if_multiaddr_count(ifp, MAX_NUM_MULTICAST_ADDRESSES); } + if (mcnt < MAX_NUM_MULTICAST_ADDRESSES) + reg_rctl &= (~IXGBE_FCTRL_MPE); - /* Set up the redirection table */ - for (int i = 0, j = 0; i < table_size; i++, j++) { - if (j == adapter->num_queues) j = 0; -#ifdef RSS - /* - * Fetch the RSS bucket id for the given indirection entry. - * Cap it at the number of configured buckets (which is - * num_queues.) - */ - queue_id = rss_get_indirection_to_bucket(i); - queue_id = queue_id % adapter->num_queues; -#else - queue_id = (j * index_mult); -#endif - /* - * The low 8 bits are for hash value (n+0); - * The next 8 bits are for hash value (n+1), etc. - */ - reta = reta >> 8; - reta = reta | ( ((uint32_t) queue_id) << 24); - if ((i & 3) == 3) { - if (i < 128) - IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta); - else - IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32), reta); - reta = 0; - } + /* clear promiscuous mode and multicast filters before enabling */ + IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCTRL, reg_rctl); + + if (ifp->if_flags & IFF_PROMISC) { + reg_rctl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE); + IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCTRL, reg_rctl); + } else if (ifp->if_flags & IFF_ALLMULTI) { + reg_rctl |= IXGBE_FCTRL_MPE; + reg_rctl &= ~IXGBE_FCTRL_UPE; + IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCTRL, reg_rctl); } + + return (0); +} - /* Now fill our hash function seeds */ - for (int i = 0; i < 10; i++) - IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), rss_key[i]); +static int +ixgbe_msix_link(void *arg) +{ + struct adapter *adapter = arg; + struct ixgbe_hw *hw = &adapter->hw; + u32 reg_eicr, mod_mask; - /* Perform hash on these packet types */ -#ifdef RSS - mrqc = IXGBE_MRQC_RSSEN; - rss_hash_config = rss_gethashconfig(); - if (rss_hash_config & RSS_HASHTYPE_RSS_IPV4) - mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4; - if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV4) - mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_TCP; - if (rss_hash_config & RSS_HASHTYPE_RSS_IPV6) - mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6; - if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV6) - mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_TCP; - if (rss_hash_config & RSS_HASHTYPE_RSS_IPV6_EX) - mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX; - if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV6_EX) - mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX_TCP; - if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV4) - mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP; - if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV4_EX) - device_printf(adapter->dev, - "%s: RSS_HASHTYPE_RSS_UDP_IPV4_EX defined, " - "but not supported\n", __func__); - if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV6) - mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP; - if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV6_EX) - mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX_UDP; -#else - /* - * Disable UDP - IP fragments aren't currently being handled - * and so we end up with a mix of 2-tuple and 4-tuple - * traffic. - */ - mrqc = IXGBE_MRQC_RSSEN - | IXGBE_MRQC_RSS_FIELD_IPV4 - | IXGBE_MRQC_RSS_FIELD_IPV4_TCP - | IXGBE_MRQC_RSS_FIELD_IPV6_EX_TCP - | IXGBE_MRQC_RSS_FIELD_IPV6_EX - | IXGBE_MRQC_RSS_FIELD_IPV6 - | IXGBE_MRQC_RSS_FIELD_IPV6_TCP - ; -#endif /* RSS */ -#ifdef PCI_IOV - mode = ixgbe_get_iov_mode(adapter); - mrqc |= ixgbe_get_mrqc(mode); -#endif - IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc); -} + ++adapter->link_irq; + MPASS(hw->back != NULL); + /* First get the cause */ + reg_eicr = IXGBE_READ_REG(hw, IXGBE_EICS); -/********************************************************************* - * - * Setup receive registers and features. - * - **********************************************************************/ -#define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2 + /* Be sure the queue bits are not cleared */ + reg_eicr &= ~IXGBE_EICR_RTX_QUEUE; + /* Clear interrupt with write */ + IXGBE_WRITE_REG(hw, IXGBE_EICR, reg_eicr); -#define BSIZEPKT_ROUNDUP ((1<ctx); -static void -ixgbe_initialize_receive_units(struct adapter *adapter) -{ - struct rx_ring *rxr = adapter->rx_rings; - struct ixgbe_hw *hw = &adapter->hw; - struct ifnet *ifp = adapter->ifp; - u32 bufsz, fctrl, srrctl, rxcsum; - u32 hlreg; + if (adapter->hw.mac.type != ixgbe_mac_82598EB) { +#ifdef IXGBE_FDIR + if (reg_eicr & IXGBE_EICR_FLOW_DIR) { + /* This is probably overkill :) */ + if (!atomic_cmpset_int(&adapter->fdir_reinit, 0, 1)) + return; + /* Disable the interrupt */ + IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EICR_FLOW_DIR); + GROUPTASK_ENQUEUE(&adapter->fdir_task); + } else +#endif + if (reg_eicr & IXGBE_EICR_ECC) { + device_printf(iflib_get_dev(adapter->ctx), "\nCRITICAL: ECC ERROR!! " + "Please Reboot!!\n"); + IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC); + } - /* - * Make sure receives are disabled while - * setting up the descriptor ring - */ - ixgbe_disable_rx(hw); + /* Check for over temp condition */ + if (reg_eicr & IXGBE_EICR_TS) { + device_printf(iflib_get_dev(adapter->ctx), "\nCRITICAL: OVER TEMP!! " + "PHY IS SHUT DOWN!!\n"); + device_printf(iflib_get_dev(adapter->ctx), "System shutdown required!\n"); + IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_TS); + } - /* Enable broadcasts */ - fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL); - fctrl |= IXGBE_FCTRL_BAM; - if (adapter->hw.mac.type == ixgbe_mac_82598EB) { - fctrl |= IXGBE_FCTRL_DPF; - fctrl |= IXGBE_FCTRL_PMCF; +#ifdef PCI_IOV + if (reg_eicr & IXGBE_EICR_MAILBOX) + /* taskqueue_enqueue(adapter->tq, &adapter->mbx_task); */ +#endif + } - IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl); - - /* Set for Jumbo Frames? */ - hlreg = IXGBE_READ_REG(hw, IXGBE_HLREG0); - if (ifp->if_mtu > ETHERMTU) - hlreg |= IXGBE_HLREG0_JUMBOEN; - else - hlreg &= ~IXGBE_HLREG0_JUMBOEN; -#ifdef DEV_NETMAP - /* crcstrip is conditional in netmap (in RDRXCTL too ?) */ - if (ifp->if_capenable & IFCAP_NETMAP && !ix_crcstrip) - hlreg &= ~IXGBE_HLREG0_RXCRCSTRP; + /* Pluggable optics-related interrupt */ + if (hw->device_id == IXGBE_DEV_ID_X550EM_X_SFP) + mod_mask = IXGBE_EICR_GPI_SDP0_X540; else - hlreg |= IXGBE_HLREG0_RXCRCSTRP; -#endif /* DEV_NETMAP */ - IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg); + mod_mask = IXGBE_EICR_GPI_SDP2_BY_MAC(hw); - bufsz = (adapter->rx_mbuf_sz + - BSIZEPKT_ROUNDUP) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT; + if (ixgbe_is_sfp(hw)) { + if (reg_eicr & IXGBE_EICR_GPI_SDP1_BY_MAC(hw)) { + IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1_BY_MAC(hw)); + GROUPTASK_ENQUEUE(&adapter->msf_task); + } else if (reg_eicr & mod_mask) { + IXGBE_WRITE_REG(hw, IXGBE_EICR, mod_mask); + GROUPTASK_ENQUEUE(&adapter->mod_task); + } + } + + /* Check for fan failure */ + if ((hw->device_id == IXGBE_DEV_ID_82598AT) && + (reg_eicr & IXGBE_EICR_GPI_SDP1)) { + IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1); + device_printf(iflib_get_dev(adapter->ctx), "\nCRITICAL: FAN FAILURE!! " + "REPLACE IMMEDIATELY!!\n"); + } + + /* External PHY interrupt */ + if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T && + (reg_eicr & IXGBE_EICR_GPI_SDP0_X540)) { + IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP0_X540); + GROUPTASK_ENQUEUE(&adapter->phy_task); + } + + IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, IXGBE_EIMS_OTHER); + return (FILTER_HANDLED); +} + +static int +ixgbe_sysctl_interrupt_rate_handler(SYSCTL_HANDLER_ARGS) +{ + int error; + struct ix_queue *que = ((struct ix_queue *)oidp->oid_arg1); + unsigned int reg, usec, rate; - for (int i = 0; i < adapter->num_queues; i++, rxr++) { - u64 rdba = rxr->rxdma.dma_paddr; - int j = rxr->me; + reg = IXGBE_READ_REG(&que->adapter->hw, IXGBE_EITR(que->msix)); + usec = ((reg & 0x0FF8) >> 3); + if (usec > 0) + rate = 500000 / usec; + else + rate = 0; + error = sysctl_handle_int(oidp, &rate, 0, req); + if (error || !req->newptr) + return error; + reg &= ~0xfff; /* default, no limitation */ + ixgbe_max_interrupt_rate = 0; + if (rate > 0 && rate < 500000) { + if (rate < 1000) + rate = 1000; + ixgbe_max_interrupt_rate = rate; + reg |= ((4000000/rate) & 0xff8 ); + } + IXGBE_WRITE_REG(&que->adapter->hw, IXGBE_EITR(que->msix), reg); + return 0; +} - /* Setup the Base and Length of the Rx Descriptor Ring */ - IXGBE_WRITE_REG(hw, IXGBE_RDBAL(j), - (rdba & 0x00000000ffffffffULL)); - IXGBE_WRITE_REG(hw, IXGBE_RDBAH(j), (rdba >> 32)); - IXGBE_WRITE_REG(hw, IXGBE_RDLEN(j), - adapter->num_rx_desc * sizeof(union ixgbe_adv_rx_desc)); +static void +ixgbe_add_device_sysctls(if_ctx_t ctx) +{ + device_t dev = iflib_get_dev(ctx); + struct adapter *adapter = iflib_get_softc(ctx); + struct ixgbe_hw *hw = &adapter->hw; + struct sysctl_oid_list *child; + struct sysctl_ctx_list *ctx_list; - /* Set up the SRRCTL register */ - srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(j)); - srrctl &= ~IXGBE_SRRCTL_BSIZEHDR_MASK; - srrctl &= ~IXGBE_SRRCTL_BSIZEPKT_MASK; - srrctl |= bufsz; - srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF; + ctx_list = device_get_sysctl_ctx(dev); + child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev)); - /* - * Set DROP_EN iff we have no flow control and >1 queue. - * Note that srrctl was cleared shortly before during reset, - * so we do not need to clear the bit, but do it just in case - * this code is moved elsewhere. - */ - if (adapter->num_queues > 1 && - adapter->hw.fc.requested_mode == ixgbe_fc_none) { - srrctl |= IXGBE_SRRCTL_DROP_EN; - } else { - srrctl &= ~IXGBE_SRRCTL_DROP_EN; - } + /* Sysctls for all devices */ + SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fc", + CTLTYPE_INT | CTLFLAG_RW, adapter, 0, + ixgbe_set_flowcntl, "I", IXGBE_SYSCTL_DESC_SET_FC); - IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(j), srrctl); + SYSCTL_ADD_INT(ctx_list, child, OID_AUTO, "enable_aim", + CTLFLAG_RW, + &ixgbe_enable_aim, 1, "Interrupt Moderation"); - /* Setup the HW Rx Head and Tail Descriptor Pointers */ - IXGBE_WRITE_REG(hw, IXGBE_RDH(j), 0); - IXGBE_WRITE_REG(hw, IXGBE_RDT(j), 0); + SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "advertise_speed", + CTLTYPE_INT | CTLFLAG_RW, adapter, 0, + ixgbe_set_advertise, "I", IXGBE_SYSCTL_DESC_ADV_SPEED); - /* Set the driver rx tail address */ - rxr->tail = IXGBE_RDT(rxr->me); - } + SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "thermal_test", + CTLTYPE_INT | CTLFLAG_RW, adapter, 0, + ixgbe_sysctl_thermal_test, "I", "Thermal Test"); - if (adapter->hw.mac.type != ixgbe_mac_82598EB) { - u32 psrtype = IXGBE_PSRTYPE_TCPHDR | - IXGBE_PSRTYPE_UDPHDR | - IXGBE_PSRTYPE_IPV4HDR | - IXGBE_PSRTYPE_IPV6HDR; - IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(0), psrtype); - } + SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "reg_dump", + CTLTYPE_STRING | CTLFLAG_RD, adapter, 0, + ixgbe_get_regs, "A", "Dump Registers"); - rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM); + /* for X550 devices */ + if (hw->mac.type >= ixgbe_mac_X550) + SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "dmac", + CTLTYPE_INT | CTLFLAG_RW, adapter, 0, + ixgbe_sysctl_dmac, "I", "DMA Coalesce"); - ixgbe_initialize_rss_mapping(adapter); + /* for X550T and X550EM backplane devices */ + if (hw->mac.ops.setup_eee) { + struct sysctl_oid *eee_node; + struct sysctl_oid_list *eee_list; - if (adapter->num_queues > 1) { - /* RSS and RX IPP Checksum are mutually exclusive */ - rxcsum |= IXGBE_RXCSUM_PCSD; - } + eee_node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, "eee", + CTLFLAG_RD, NULL, + "Energy Efficient Ethernet sysctls"); + eee_list = SYSCTL_CHILDREN(eee_node); - if (ifp->if_capenable & IFCAP_RXCSUM) - rxcsum |= IXGBE_RXCSUM_PCSD; + SYSCTL_ADD_PROC(ctx_list, eee_list, OID_AUTO, "enable", + CTLTYPE_INT | CTLFLAG_RW, adapter, 0, + ixgbe_sysctl_eee_enable, "I", + "Enable or Disable EEE"); + + SYSCTL_ADD_PROC(ctx_list, eee_list, OID_AUTO, "negotiated", + CTLTYPE_INT | CTLFLAG_RD, adapter, 0, + ixgbe_sysctl_eee_negotiated, "I", + "EEE negotiated on link"); + + SYSCTL_ADD_PROC(ctx_list, eee_list, OID_AUTO, "tx_lpi_status", + CTLTYPE_INT | CTLFLAG_RD, adapter, 0, + ixgbe_sysctl_eee_tx_lpi_status, "I", + "Whether or not TX link is in LPI state"); + + SYSCTL_ADD_PROC(ctx_list, eee_list, OID_AUTO, "rx_lpi_status", + CTLTYPE_INT | CTLFLAG_RD, adapter, 0, + ixgbe_sysctl_eee_rx_lpi_status, "I", + "Whether or not RX link is in LPI state"); + } + + /* for certain 10GBaseT devices */ + if (hw->device_id == IXGBE_DEV_ID_X550T || + hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) { + SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "wol_enable", + CTLTYPE_INT | CTLFLAG_RW, adapter, 0, + ixgbe_sysctl_wol_enable, "I", + "Enable/Disable Wake on LAN"); + + SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "wufc", + CTLTYPE_INT | CTLFLAG_RW, adapter, 0, + ixgbe_sysctl_wufc, "I", + "Enable/Disable Wake Up Filters"); + } + /* for X550EM 10GBaseT devices */ + if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) { + struct sysctl_oid *phy_node; + struct sysctl_oid_list *phy_list; - /* This is useful for calculating UDP/IP fragment checksums */ - if (!(rxcsum & IXGBE_RXCSUM_PCSD)) - rxcsum |= IXGBE_RXCSUM_IPPCSE; + phy_node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, "phy", + CTLFLAG_RD, NULL, + "External PHY sysctls"); + phy_list = SYSCTL_CHILDREN(phy_node); - IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum); + SYSCTL_ADD_PROC(ctx_list, phy_list, OID_AUTO, "temp", + CTLTYPE_INT | CTLFLAG_RD, adapter, 0, + ixgbe_sysctl_phy_temp, "I", + "Current External PHY Temperature (Celsius)"); - return; + SYSCTL_ADD_PROC(ctx_list, phy_list, OID_AUTO, "overtemp_occurred", + CTLTYPE_INT | CTLFLAG_RD, adapter, 0, + ixgbe_sysctl_phy_overtemp_occurred, "I", + "External PHY High Temperature Event Occurred"); + } } - -/* -** This routine is run via an vlan config EVENT, -** it enables us to use the HW Filter table since -** we can get the vlan id. This just creates the -** entry in the soft version of the VFTA, init will -** repopulate the real table. -*/ +/********************************************************************* + * + * Determine hardware revision. + * + **********************************************************************/ static void -ixgbe_register_vlan(void *arg, struct ifnet *ifp, u16 vtag) +ixgbe_identify_hardware(if_ctx_t ctx) { - struct adapter *adapter = ifp->if_softc; - u16 index, bit; - - if (ifp->if_softc != arg) /* Not our event */ - return; + struct adapter *adapter = iflib_get_softc(ctx); + device_t dev = iflib_get_dev(ctx); + struct ixgbe_hw *hw = &adapter->hw; - if ((vtag == 0) || (vtag > 4095)) /* Invalid */ - return; + /* Save off the information about this board */ + hw->vendor_id = pci_get_vendor(dev); + hw->device_id = pci_get_device(dev); + hw->revision_id = pci_read_config(dev, PCIR_REVID, 1); + hw->subsystem_vendor_id = + pci_read_config(dev, PCIR_SUBVEND_0, 2); + hw->subsystem_device_id = + pci_read_config(dev, PCIR_SUBDEV_0, 2); - IXGBE_CORE_LOCK(adapter); - index = (vtag >> 5) & 0x7F; - bit = vtag & 0x1F; - adapter->shadow_vfta[index] |= (1 << bit); - ++adapter->num_vlans; - ixgbe_setup_vlan_hw_support(adapter); - IXGBE_CORE_UNLOCK(adapter); + /* We need this here to set the num_segs in the code that follows */ + ixgbe_set_mac_type(hw); } -/* -** This routine is run via an vlan -** unconfig EVENT, remove our entry -** in the soft vfta. -*/ +/********************************************************************* + * + * Determine optic type + * + **********************************************************************/ static void -ixgbe_unregister_vlan(void *arg, struct ifnet *ifp, u16 vtag) +ixgbe_setup_optics(struct adapter *adapter) { - struct adapter *adapter = ifp->if_softc; - u16 index, bit; + struct ixgbe_hw *hw = &adapter->hw; + int layer; - if (ifp->if_softc != arg) - return; + layer = adapter->phy_layer = ixgbe_get_supported_physical_layer(hw); - if ((vtag == 0) || (vtag > 4095)) /* Invalid */ + if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T) { + adapter->optics = IFM_10G_T; return; + } - IXGBE_CORE_LOCK(adapter); - index = (vtag >> 5) & 0x7F; - bit = vtag & 0x1F; - adapter->shadow_vfta[index] &= ~(1 << bit); - --adapter->num_vlans; - /* Re-init to load the changes */ - ixgbe_setup_vlan_hw_support(adapter); - IXGBE_CORE_UNLOCK(adapter); -} + if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_T) { + adapter->optics = IFM_1000_T; + return; -static void -ixgbe_setup_vlan_hw_support(struct adapter *adapter) -{ - struct ifnet *ifp = adapter->ifp; - struct ixgbe_hw *hw = &adapter->hw; - struct rx_ring *rxr; - u32 ctrl; + } + if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX) { + adapter->optics = IFM_1000_SX; + return; + } - /* - ** We get here thru init_locked, meaning - ** a soft reset, this has already cleared - ** the VFTA and other state, so if there - ** have been no vlan's registered do nothing. - */ - if (adapter->num_vlans == 0) + if (layer & (IXGBE_PHYSICAL_LAYER_10GBASE_LR | + IXGBE_PHYSICAL_LAYER_10GBASE_LRM)) { + adapter->optics = IFM_10G_LR; return; + } - /* Setup the queues for vlans */ - for (int i = 0; i < adapter->num_queues; i++) { - rxr = &adapter->rx_rings[i]; - /* On 82599 the VLAN enable is per/queue in RXDCTL */ - if (hw->mac.type != ixgbe_mac_82598EB) { - ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me)); - ctrl |= IXGBE_RXDCTL_VME; - IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxr->me), ctrl); - } - rxr->vtag_strip = TRUE; + if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR) { + adapter->optics = IFM_10G_SR; + return; } - if ((ifp->if_capenable & IFCAP_VLAN_HWFILTER) == 0) + if (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU) { + adapter->optics = IFM_10G_TWINAX; return; - /* - ** A soft reset zero's out the VFTA, so - ** we need to repopulate it now. - */ - for (int i = 0; i < IXGBE_VFTA_SIZE; i++) - if (adapter->shadow_vfta[i] != 0) - IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), - adapter->shadow_vfta[i]); + } - ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL); - /* Enable the Filter Table if enabled */ - if (ifp->if_capenable & IFCAP_VLAN_HWFILTER) { - ctrl &= ~IXGBE_VLNCTRL_CFIEN; - ctrl |= IXGBE_VLNCTRL_VFE; + if (layer & (IXGBE_PHYSICAL_LAYER_10GBASE_KX4 | + IXGBE_PHYSICAL_LAYER_10GBASE_CX4)) { + adapter->optics = IFM_10G_CX4; + return; } - if (hw->mac.type == ixgbe_mac_82598EB) - ctrl |= IXGBE_VLNCTRL_VME; - IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl); -} -static void -ixgbe_enable_intr(struct adapter *adapter) + /* If we get here just set the default */ + adapter->optics = IFM_ETHER | IFM_AUTO; +} + +static int +ixgbe_allocate_pci_resources(if_ctx_t ctx) { - struct ixgbe_hw *hw = &adapter->hw; - struct ix_queue *que = adapter->queues; - u32 mask, fwsm; + int rid; + struct adapter *adapter = iflib_get_softc(ctx); + device_t dev = iflib_get_dev(ctx); - mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE); - /* Enable Fan Failure detection */ - if (hw->device_id == IXGBE_DEV_ID_82598AT) - mask |= IXGBE_EIMS_GPI_SDP1; + rid = PCIR_BAR(0); + adapter->pci_mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY, + &rid, RF_ACTIVE); - switch (adapter->hw.mac.type) { - case ixgbe_mac_82599EB: - mask |= IXGBE_EIMS_ECC; - /* Temperature sensor on some adapters */ - mask |= IXGBE_EIMS_GPI_SDP0; - /* SFP+ (RX_LOS_N & MOD_ABS_N) */ - mask |= IXGBE_EIMS_GPI_SDP1; - mask |= IXGBE_EIMS_GPI_SDP2; -#ifdef IXGBE_FDIR - mask |= IXGBE_EIMS_FLOW_DIR; -#endif -#ifdef PCI_IOV - mask |= IXGBE_EIMS_MAILBOX; -#endif - break; - case ixgbe_mac_X540: - /* Detect if Thermal Sensor is enabled */ - fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM); - if (fwsm & IXGBE_FWSM_TS_ENABLED) - mask |= IXGBE_EIMS_TS; - mask |= IXGBE_EIMS_ECC; -#ifdef IXGBE_FDIR - mask |= IXGBE_EIMS_FLOW_DIR; -#endif - break; - case ixgbe_mac_X550: - case ixgbe_mac_X550EM_x: - /* MAC thermal sensor is automatically enabled */ - mask |= IXGBE_EIMS_TS; - /* Some devices use SDP0 for important information */ - if (hw->device_id == IXGBE_DEV_ID_X550EM_X_SFP || - hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) - mask |= IXGBE_EIMS_GPI_SDP0_BY_MAC(hw); - mask |= IXGBE_EIMS_ECC; -#ifdef IXGBE_FDIR - mask |= IXGBE_EIMS_FLOW_DIR; -#endif -#ifdef PCI_IOV - mask |= IXGBE_EIMS_MAILBOX; -#endif - /* falls through */ - default: - break; + if (!(adapter->pci_mem)) { + device_printf(dev,"Unable to allocate bus resource: memory\n"); + return (ENXIO); } - IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask); + adapter->osdep.mem_bus_space_tag = + rman_get_bustag(adapter->pci_mem); + adapter->osdep.mem_bus_space_handle = + rman_get_bushandle(adapter->pci_mem); + adapter->hw.hw_addr = (u8 *) &adapter->osdep.mem_bus_space_handle; - /* With MSI-X we use auto clear */ - if (adapter->msix_mem) { - mask = IXGBE_EIMS_ENABLE_MASK; - /* Don't autoclear Link */ - mask &= ~IXGBE_EIMS_OTHER; - mask &= ~IXGBE_EIMS_LSC; -#ifdef PCI_IOV - mask &= ~IXGBE_EIMS_MAILBOX; -#endif - IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask); - } + adapter->hw.back = adapter; + return (0); +} - /* - ** Now enable all queues, this is done separately to - ** allow for handling the extended (beyond 32) MSIX - ** vectors that can be used by 82599 - */ - for (int i = 0; i < adapter->num_queues; i++, que++) - ixgbe_enable_queue(adapter, que->msix); +/********************************************************************* + * Device removal routine + * + * The detach entry point is called when the driver is being removed. + * This routine stops the adapter and deallocates all the resources + * that were allocated for driver operation. + * + * return 0 on success, positive on failure + *********************************************************************/ +static int +ixgbe_if_detach(if_ctx_t ctx) +{ + struct adapter *adapter = iflib_get_softc(ctx); + u32 ctrl_ext; + + INIT_DEBUGOUT("ixgbe_detach: begin"); - IXGBE_WRITE_FLUSH(hw); +#ifdef PCI_IOV + if (pci_iov_detach(dev) != 0) { + device_printf(dev, "SR-IOV in use; detach first.\n"); + return (EBUSY); + } +#endif /* PCI_IOV */ + + ixgbe_setup_low_power_mode(ctx); - return; + /* let hardware know driver is unloading */ + ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT); + ctrl_ext &= ~IXGBE_CTRL_EXT_DRV_LOAD; + IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT, ctrl_ext); + + ixgbe_free_pci_resources(ctx); + free(adapter->mta, M_DEVBUF); + + return (0); } -static void -ixgbe_disable_intr(struct adapter *adapter) +static int +ixgbe_setup_low_power_mode(if_ctx_t ctx) { - if (adapter->msix_mem) - IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, 0); - if (adapter->hw.mac.type == ixgbe_mac_82598EB) { - IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0); - } else { - IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000); - IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0); - IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0); - } - IXGBE_WRITE_FLUSH(&adapter->hw); - return; + struct adapter *adapter = iflib_get_softc(ctx); + struct ixgbe_hw *hw = &adapter->hw; + device_t dev = iflib_get_dev(ctx); + s32 error = 0; + + /* Limit power management flow to X550EM baseT */ + if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T + && hw->phy.ops.enter_lplu) { + /* Turn off support for APM wakeup. (Using ACPI instead) */ + IXGBE_WRITE_REG(hw, IXGBE_GRC, + IXGBE_READ_REG(hw, IXGBE_GRC) & ~(u32)2); + + /* + * Clear Wake Up Status register to prevent any previous wakeup + * events from waking us up immediately after we suspend. + */ + IXGBE_WRITE_REG(hw, IXGBE_WUS, 0xffffffff); + + /* + * Program the Wakeup Filter Control register with user filter + * settings + */ + IXGBE_WRITE_REG(hw, IXGBE_WUFC, adapter->wufc); + + /* Enable wakeups and power management in Wakeup Control */ + IXGBE_WRITE_REG(hw, IXGBE_WUC, + IXGBE_WUC_WKEN | IXGBE_WUC_PME_EN); + + /* X550EM baseT adapters need a special LPLU flow */ + hw->phy.reset_disable = true; + error = hw->phy.ops.enter_lplu(hw); + if (error) + device_printf(dev, + "Error entering LPLU: %d\n", error); + hw->phy.reset_disable = false; + } + + return error; } -/* -** Get the width and transaction speed of -** the slot this adapter is plugged into. -*/ -static void -ixgbe_get_slot_info(struct adapter *adapter) +/********************************************************************* + * + * Shutdown entry point + * + **********************************************************************/ +static int +ixgbe_if_shutdown(if_ctx_t ctx) { - device_t dev = adapter->dev; - struct ixgbe_hw *hw = &adapter->hw; - struct ixgbe_mac_info *mac = &hw->mac; - u16 link; - u32 offset; + int error = 0; - /* For most devices simply call the shared code routine */ - if (hw->device_id != IXGBE_DEV_ID_82599_SFP_SF_QP) { - ixgbe_get_bus_info(hw); - /* These devices don't use PCI-E */ - switch (hw->mac.type) { - case ixgbe_mac_X550EM_x: - return; - default: - goto display; - } - } + INIT_DEBUGOUT("ixgbe_shutdown: begin"); - /* - ** For the Quad port adapter we need to parse back - ** up the PCI tree to find the speed of the expansion - ** slot into which this adapter is plugged. A bit more work. - */ - dev = device_get_parent(device_get_parent(dev)); -#ifdef IXGBE_DEBUG - device_printf(dev, "parent pcib = %x,%x,%x\n", - pci_get_bus(dev), pci_get_slot(dev), pci_get_function(dev)); -#endif - dev = device_get_parent(device_get_parent(dev)); -#ifdef IXGBE_DEBUG - device_printf(dev, "slot pcib = %x,%x,%x\n", - pci_get_bus(dev), pci_get_slot(dev), pci_get_function(dev)); -#endif - /* Now get the PCI Express Capabilities offset */ - pci_find_cap(dev, PCIY_EXPRESS, &offset); - /* ...and read the Link Status Register */ - link = pci_read_config(dev, offset + PCIER_LINK_STA, 2); - switch (link & IXGBE_PCI_LINK_WIDTH) { - case IXGBE_PCI_LINK_WIDTH_1: - hw->bus.width = ixgbe_bus_width_pcie_x1; - break; - case IXGBE_PCI_LINK_WIDTH_2: - hw->bus.width = ixgbe_bus_width_pcie_x2; - break; - case IXGBE_PCI_LINK_WIDTH_4: - hw->bus.width = ixgbe_bus_width_pcie_x4; - break; - case IXGBE_PCI_LINK_WIDTH_8: - hw->bus.width = ixgbe_bus_width_pcie_x8; - break; - default: - hw->bus.width = ixgbe_bus_width_unknown; - break; - } + error = ixgbe_setup_low_power_mode(ctx); + return (error); +} - switch (link & IXGBE_PCI_LINK_SPEED) { - case IXGBE_PCI_LINK_SPEED_2500: - hw->bus.speed = ixgbe_bus_speed_2500; - break; - case IXGBE_PCI_LINK_SPEED_5000: - hw->bus.speed = ixgbe_bus_speed_5000; - break; - case IXGBE_PCI_LINK_SPEED_8000: - hw->bus.speed = ixgbe_bus_speed_8000; - break; - default: - hw->bus.speed = ixgbe_bus_speed_unknown; - break; - } +static int +ixgbe_if_suspend(if_ctx_t ctx) +{ + device_t dev = iflib_get_dev(ctx); + int error = 0; - mac->ops.set_lan_id(hw); + INIT_DEBUGOUT("ixgbe_suspend: begin"); -display: - device_printf(dev,"PCI Express Bus: Speed %s %s\n", - ((hw->bus.speed == ixgbe_bus_speed_8000) ? "8.0GT/s": - (hw->bus.speed == ixgbe_bus_speed_5000) ? "5.0GT/s": - (hw->bus.speed == ixgbe_bus_speed_2500) ? "2.5GT/s":"Unknown"), - (hw->bus.width == ixgbe_bus_width_pcie_x8) ? "Width x8" : - (hw->bus.width == ixgbe_bus_width_pcie_x4) ? "Width x4" : - (hw->bus.width == ixgbe_bus_width_pcie_x1) ? "Width x1" : - ("Unknown")); + error = ixgbe_setup_low_power_mode(ctx); - if ((hw->device_id != IXGBE_DEV_ID_82599_SFP_SF_QP) && - ((hw->bus.width <= ixgbe_bus_width_pcie_x4) && - (hw->bus.speed == ixgbe_bus_speed_2500))) { - device_printf(dev, "PCI-Express bandwidth available" - " for this card\n is not sufficient for" - " optimal performance.\n"); - device_printf(dev, "For optimal performance a x8 " - "PCIE, or x4 PCIE Gen2 slot is required.\n"); - } - if ((hw->device_id == IXGBE_DEV_ID_82599_SFP_SF_QP) && - ((hw->bus.width <= ixgbe_bus_width_pcie_x8) && - (hw->bus.speed < ixgbe_bus_speed_8000))) { - device_printf(dev, "PCI-Express bandwidth available" - " for this card\n is not sufficient for" - " optimal performance.\n"); - device_printf(dev, "For optimal performance a x8 " - "PCIE Gen3 slot is required.\n"); - } + /* Save state and power down */ + pci_save_state(dev); + pci_set_powerstate(dev, PCI_POWERSTATE_D3); - return; + return (error); } - -/* -** Setup the correct IVAR register for a particular MSIX interrupt -** (yes this is all very magic and confusing :) -** - entry is the register array entry -** - vector is the MSIX vector for this queue -** - type is RX/TX/MISC -*/ -static void -ixgbe_set_ivar(struct adapter *adapter, u8 entry, u8 vector, s8 type) +static int +ixgbe_if_resume(if_ctx_t ctx) { + struct adapter *adapter = iflib_get_softc(ctx); + device_t dev = iflib_get_dev(ctx); + struct ifnet *ifp = iflib_get_ifp(ctx); struct ixgbe_hw *hw = &adapter->hw; - u32 ivar, index; + u32 wus; - vector |= IXGBE_IVAR_ALLOC_VAL; + INIT_DEBUGOUT("ixgbe_resume: begin"); - switch (hw->mac.type) { + pci_set_powerstate(dev, PCI_POWERSTATE_D0); + pci_restore_state(dev); - case ixgbe_mac_82598EB: - if (type == -1) - entry = IXGBE_IVAR_OTHER_CAUSES_INDEX; - else - entry += (type * 64); - index = (entry >> 2) & 0x1F; - ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index)); - ivar &= ~(0xFF << (8 * (entry & 0x3))); - ivar |= (vector << (8 * (entry & 0x3))); - IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR(index), ivar); - break; + /* Read & clear WUS register */ + wus = IXGBE_READ_REG(hw, IXGBE_WUS); + if (wus) + device_printf(dev, "Woken up by (WUS): %#010x\n", + IXGBE_READ_REG(hw, IXGBE_WUS)); + IXGBE_WRITE_REG(hw, IXGBE_WUS, 0xffffffff); + /* And clear WUFC until next low-power transition */ + IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0); - case ixgbe_mac_82599EB: - case ixgbe_mac_X540: - case ixgbe_mac_X550: - case ixgbe_mac_X550EM_x: - if (type == -1) { /* MISC IVAR */ - index = (entry & 1) * 8; - ivar = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC); - ivar &= ~(0xFF << index); - ivar |= (vector << index); - IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, ivar); - } else { /* RX/TX IVARS */ - index = (16 * (entry & 1)) + (8 * type); - ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(entry >> 1)); - ivar &= ~(0xFF << index); - ivar |= (vector << index); - IXGBE_WRITE_REG(hw, IXGBE_IVAR(entry >> 1), ivar); - } + /* + * Required after D3->D0 transition; + * will re-advertise all previous advertised speeds + */ + if (ifp->if_flags & IFF_UP) + ixgbe_if_init(ctx); - default: - break; - } + INIT_DEBUGOUT("ixgbe_resume: end"); + return (0); } -static void -ixgbe_configure_ivars(struct adapter *adapter) +/********************************************************************* + * Ioctl mtu entry point + * + * + * return 0 on success, EINVAL on failure + **********************************************************************/ +static int +ixgbe_if_mtu_set(if_ctx_t ctx, uint32_t mtu) { - struct ix_queue *que = adapter->queues; - u32 newitr; + int error = 0; + struct adapter *adapter = iflib_get_softc(ctx); - if (ixgbe_max_interrupt_rate > 0) - newitr = (4000000 / ixgbe_max_interrupt_rate) & 0x0FF8; - else { - /* - ** Disable DMA coalescing if interrupt moderation is - ** disabled. - */ - adapter->dmac = 0; - newitr = 0; - } + IOCTL_DEBUGOUT("ioctl: SIOCIFMTU (Set Interface MTU)"); - for (int i = 0; i < adapter->num_queues; i++, que++) { - struct rx_ring *rxr = &adapter->rx_rings[i]; - struct tx_ring *txr = &adapter->tx_rings[i]; - /* First the RX queue entry */ - ixgbe_set_ivar(adapter, rxr->me, que->msix, 0); - /* ... and the TX */ - ixgbe_set_ivar(adapter, txr->me, que->msix, 1); - /* Set an Initial EITR value */ - IXGBE_WRITE_REG(&adapter->hw, - IXGBE_EITR(que->msix), newitr); + if (mtu > IXGBE_MAX_MTU) { + error = EINVAL; + } else { + adapter->max_frame_size = mtu + IXGBE_MTU_HDR; } - - /* For the Link interrupt */ - ixgbe_set_ivar(adapter, 1, adapter->vector, -1); + + return error; } -/* -** ixgbe_sfp_probe - called in the local timer to -** determine if a port had optics inserted. -*/ -static bool -ixgbe_sfp_probe(struct adapter *adapter) -{ - struct ixgbe_hw *hw = &adapter->hw; - device_t dev = adapter->dev; - bool result = FALSE; - - if ((hw->phy.type == ixgbe_phy_nl) && - (hw->phy.sfp_type == ixgbe_sfp_type_not_present)) { - s32 ret = hw->phy.ops.identify_sfp(hw); - if (ret) - goto out; - ret = hw->phy.ops.reset(hw); - if (ret == IXGBE_ERR_SFP_NOT_SUPPORTED) { - device_printf(dev, "Unsupported SFP+ module detected!"); - device_printf(dev, "Reload driver with supported module.\n"); - adapter->sfp_probe = FALSE; - goto out; - } else - device_printf(dev, "SFP+ module detected!\n"); - /* We now have supported optics */ - adapter->sfp_probe = FALSE; - /* Set the optics type so system reports correctly */ - ixgbe_setup_optics(adapter); - result = TRUE; - } -out: - return (result); -} -/* -** Tasklet handler for MSIX Link interrupts -** - do outside interrupt since it might sleep -*/ static void -ixgbe_handle_link(void *context, int pending) +ixgbe_if_crcstrip_set(if_ctx_t ctx, int onoff) { - struct adapter *adapter = context; - struct ixgbe_hw *hw = &adapter->hw; + struct adapter *sc = iflib_get_softc(ctx); + struct ixgbe_hw *hw = &sc->hw; + /* crc stripping is set in two places: + * IXGBE_HLREG0 (modified on init_locked and hw reset) + * IXGBE_RDRXCTL (set by the original driver in + * ixgbe_setup_hw_rsc() called in init_locked. + * We disable the setting when netmap is compiled in). + * We update the values here, but also in ixgbe.c because + * init_locked sometimes is called outside our control. + */ + uint32_t hl, rxc; + + hl = IXGBE_READ_REG(hw, IXGBE_HLREG0); + rxc = IXGBE_READ_REG(hw, IXGBE_RDRXCTL); + if (netmap_verbose) + D("%s read HLREG 0x%x rxc 0x%x", + onoff ? "enter" : "exit", hl, rxc); + /* hw requirements ... */ + rxc &= ~IXGBE_RDRXCTL_RSCFRSTSIZE; + rxc |= IXGBE_RDRXCTL_RSCACKC; + if (onoff && !ix_crcstrip) { + /* keep the crc. Fast rx */ + hl &= ~IXGBE_HLREG0_RXCRCSTRP; + rxc &= ~IXGBE_RDRXCTL_CRCSTRIP; + } else { + /* reset default mode */ + hl |= IXGBE_HLREG0_RXCRCSTRP; + rxc |= IXGBE_RDRXCTL_CRCSTRIP; + } + if (netmap_verbose) + D("%s write HLREG 0x%x rxc 0x%x", + onoff ? "enter" : "exit", hl, rxc); + IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hl); + IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rxc); +} - ixgbe_check_link(hw, - &adapter->link_speed, &adapter->link_up, 0); - ixgbe_update_link_status(adapter); - /* Re-enable link interrupts */ - IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_LSC); -} -/* -** Tasklet for handling SFP module interrupts -*/ +/********************************************************************* + * Init entry point + * + * This routine is used in two ways. It is used by the stack as + * init entry point in network interface structure. It is also used + * by the driver as a hw/sw initialization routine to get to a + * consistent state. + * + * return 0 on success, positive on failure + **********************************************************************/ +#define IXGBE_MHADD_MFS_SHIFT 16 + static void -ixgbe_handle_mod(void *context, int pending) +ixgbe_if_init(if_ctx_t ctx) { - struct adapter *adapter = context; + struct adapter *adapter = iflib_get_softc(ctx); + struct ifnet *ifp = iflib_get_ifp(ctx); + device_t dev = iflib_get_dev(ctx); struct ixgbe_hw *hw = &adapter->hw; - enum ixgbe_phy_type orig_type = hw->phy.type; - device_t dev = adapter->dev; - u32 err; - - IXGBE_CORE_LOCK(adapter); + struct ix_queue *que; - /* Check to see if the PHY type changed */ - if (hw->phy.ops.identify) { - hw->phy.type = ixgbe_phy_unknown; - hw->phy.ops.identify(hw); - } + u32 txdctl, mhadd; + u32 rxdctl, rxctrl; + int i; - if (hw->phy.type != orig_type) { - device_printf(dev, "Detected phy_type %d\n", hw->phy.type); +#ifdef PCI_IOV + enum ixgbe_iov_mode mode; +#endif - if (hw->phy.type == ixgbe_phy_none) { - hw->phy.sfp_type = ixgbe_sfp_type_unknown; - goto out; - } + INIT_DEBUGOUT("ixgbe_if_init: begin"); - /* Try to do the initialization that was skipped before */ - if (hw->phy.ops.init) - hw->phy.ops.init(hw); - if (hw->phy.ops.reset) - hw->phy.ops.reset(hw); +#ifdef PCI_IOV + mode = ixgbe_get_iov_mode(adapter); + adapter->pool = ixgbe_max_vfs(mode); + /* Queue indices may change with IOV mode */ + for (int i = 0; i < adapter->num_queues; i++) { + adapter->queues[i].rxr.me = ixgbe_pf_que_index(mode, i); + adapter->queues[i].txr.me = ixgbe_pf_que_index(mode, i); } +#endif + /* reprogram the RAR[0] in case user changed it. */ + ixgbe_set_rar(hw, 0, hw->mac.addr, adapter->pool, IXGBE_RAH_AV); - err = hw->phy.ops.identify_sfp(hw); - if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) { - device_printf(dev, - "Unsupported SFP+ module type was detected.\n"); - goto out; - } + /* Get the latest mac address, User can use a LAA */ + bcopy(IF_LLADDR(ifp), hw->mac.addr, IXGBE_ETH_LENGTH_OF_ADDRESS); + ixgbe_set_rar(hw, 0, hw->mac.addr, adapter->pool, 1); + hw->addr_ctrl.rar_used_count = 1; - err = hw->mac.ops.setup_sfp(hw); - if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) { - device_printf(dev, - "Setup failure - unsupported SFP+ module type.\n"); - goto out; - } - if (hw->phy.multispeed_fiber) - taskqueue_enqueue(adapter->tq, &adapter->msf_task); -out: - /* Update media type */ - switch (hw->mac.ops.get_media_type(hw)) { - case ixgbe_media_type_fiber: - adapter->optics = IFM_10G_SR; - break; - case ixgbe_media_type_copper: - adapter->optics = IFM_10G_TWINAX; - break; - case ixgbe_media_type_cx4: - adapter->optics = IFM_10G_CX4; - break; - default: - adapter->optics = 0; - break; - } +#if __FreeBSD_version >= 800000 + if (hw->mac.type != ixgbe_mac_82598EB) + ifp->if_hwassist |= CSUM_SCTP; +#endif + ixgbe_init_hw(hw); - IXGBE_CORE_UNLOCK(adapter); - return; -} +#ifdef PCI_IOV + ixgbe_initialize_iov(adapter); +#endif + ixgbe_initialize_transmit_units(ctx); -/* -** Tasklet for handling MSF (multispeed fiber) interrupts -*/ -static void -ixgbe_handle_msf(void *context, int pending) -{ - struct adapter *adapter = context; - struct ixgbe_hw *hw = &adapter->hw; - u32 autoneg; - bool negotiate; + /* Setup Multicast table */ + ixgbe_if_multi_set(ctx); - IXGBE_CORE_LOCK(adapter); - /* get_supported_phy_layer will call hw->phy.ops.identify_sfp() */ - adapter->phy_layer = ixgbe_get_supported_physical_layer(hw); + /* + ** Determine the correct mbuf pool + ** for doing jumbo frames + */ + if (adapter->max_frame_size <= MCLBYTES) + adapter->rx_mbuf_sz = MCLBYTES; + else + adapter->rx_mbuf_sz = MJUMPAGESIZE; - autoneg = hw->phy.autoneg_advertised; - if ((!autoneg) && (hw->mac.ops.get_link_capabilities)) - hw->mac.ops.get_link_capabilities(hw, &autoneg, &negotiate); - if (hw->mac.ops.setup_link) - hw->mac.ops.setup_link(hw, autoneg, TRUE); + /* Configure RX settings */ + ixgbe_initialize_receive_units(ctx); + + /* Enable SDP & MSIX interrupts based on adapter */ + ixgbe_config_gpie(adapter); - /* Adjust media types shown in ifconfig */ - ifmedia_removeall(&adapter->media); - ixgbe_add_media_types(adapter); - IXGBE_CORE_UNLOCK(adapter); - return; -} + /* Set MTU size */ + if (ifp->if_mtu > ETHERMTU) { + /* aka IXGBE_MAXFRS on 82599 and newer */ + mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD); + mhadd &= ~IXGBE_MHADD_MFS_MASK; + mhadd |= adapter->max_frame_size << IXGBE_MHADD_MFS_SHIFT; + IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd); + } -/* -** Tasklet for handling interrupts from an external PHY -*/ -static void -ixgbe_handle_phy(void *context, int pending) -{ - struct adapter *adapter = context; - struct ixgbe_hw *hw = &adapter->hw; - int error; + /* Now enable all the queues */ + for (i = 0, que = adapter->queues; i < adapter->num_queues; i++, que++) { + struct tx_ring *txr = &que->txr; + txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(txr->me)); + txdctl |= IXGBE_TXDCTL_ENABLE; + /* Set WTHRESH to 8, burst writeback */ + txdctl |= (8 << 16); + /* + * When the internal queue falls below PTHRESH (32), + * start prefetching as long as there are at least + * HTHRESH (1) buffers ready. The values are taken + * from the Intel linux driver 3.8.21. + * Prefetching enables tx line rate even with 1 queue. + */ + txdctl |= (32 << 0) | (1 << 8); + IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(txr->me), txdctl); + } - error = hw->phy.ops.handle_lasi(hw); - if (error == IXGBE_ERR_OVERTEMP) - device_printf(adapter->dev, - "CRITICAL: EXTERNAL PHY OVER TEMP!! " - " PHY will downshift to lower power state!\n"); - else if (error) - device_printf(adapter->dev, - "Error handling LASI interrupt: %d\n", - error); - return; -} + for (i = 0, que = adapter->queues; i < adapter->num_queues; i++, que++) { + struct rx_ring *rxr = &que->rxr; + rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me)); + if (hw->mac.type == ixgbe_mac_82598EB) { + /* + ** PTHRESH = 21 + ** HTHRESH = 4 + ** WTHRESH = 8 + */ + rxdctl &= ~0x3FFFFF; + rxdctl |= 0x080420; + } + rxdctl |= IXGBE_RXDCTL_ENABLE; + IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxr->me), rxdctl); + for (int j = 0; j < 10; j++) { + if (IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxr->me)) & + IXGBE_RXDCTL_ENABLE) + break; + else + msec_delay(1); + } + wmb(); + } -#ifdef IXGBE_FDIR -/* -** Tasklet for reinitializing the Flow Director filter table -*/ -static void -ixgbe_reinit_fdir(void *context, int pending) -{ - struct adapter *adapter = context; - struct ifnet *ifp = adapter->ifp; + /* Enable Receive engine */ + rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL); + if (hw->mac.type == ixgbe_mac_82598EB) + rxctrl |= IXGBE_RXCTRL_DMBYPS; + rxctrl |= IXGBE_RXCTRL_RXEN; + ixgbe_enable_rx_dma(hw, rxctrl); - if (adapter->fdir_reinit != 1) /* Shouldn't happen */ - return; - ixgbe_reinit_fdir_tables_82599(&adapter->hw); - adapter->fdir_reinit = 0; - /* re-enable flow director interrupts */ - IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR); - /* Restart the interface */ - ifp->if_drv_flags |= IFF_DRV_RUNNING; - return; -} + /* Set up MSI/X routing */ + if (ixgbe_enable_msix) { + ixgbe_configure_ivars(adapter); + /* Set up auto-mask */ + if (hw->mac.type == ixgbe_mac_82598EB) + IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE); + else { + IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF); + IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF); + } + } else { /* Simple settings for Legacy/MSI */ + ixgbe_set_ivar(adapter, 0, 0, 0); + ixgbe_set_ivar(adapter, 0, 0, 1); + IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE); + } + +#ifdef IXGBE_FDIR + /* Init Flow director */ + if (hw->mac.type != ixgbe_mac_82598EB) { + u32 hdrm = 32 << fdir_pballoc; + + hw->mac.ops.setup_rxpba(hw, 0, hdrm, PBA_STRATEGY_EQUAL); + ixgbe_init_fdir_signature_82599(&adapter->hw, fdir_pballoc); + } #endif + /* + * Check on any SFP devices that + * need to be kick-started + */ + if (hw->phy.type == ixgbe_phy_none) { + int err = hw->phy.ops.identify(hw); + if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) { + device_printf(dev, + "Unsupported SFP+ module type was detected.\n"); + return; + } + } -/********************************************************************* - * - * Configure DMA Coalescing - * - **********************************************************************/ -static void -ixgbe_config_dmac(struct adapter *adapter) -{ - struct ixgbe_hw *hw = &adapter->hw; - struct ixgbe_dmac_config *dcfg = &hw->mac.dmac_config; + /* Set moderation on the Link interrupt */ + IXGBE_WRITE_REG(hw, IXGBE_EITR(adapter->vector), IXGBE_LINK_ITR); - if (hw->mac.type < ixgbe_mac_X550 || - !hw->mac.ops.dmac_config) - return; + /* Configure Energy Efficient Ethernet for supported devices */ + ixgbe_setup_eee(hw, adapter->eee_enabled); - if (dcfg->watchdog_timer ^ adapter->dmac || - dcfg->link_speed ^ adapter->link_speed) { - dcfg->watchdog_timer = adapter->dmac; - dcfg->fcoe_en = false; - dcfg->link_speed = adapter->link_speed; - dcfg->num_tcs = 1; - - INIT_DEBUGOUT2("dmac settings: watchdog %d, link speed %d\n", - dcfg->watchdog_timer, dcfg->link_speed); + /* Config/Enable Link */ + ixgbe_config_link(adapter); - hw->mac.ops.dmac_config(hw); - } -} + /* Hardware Packet Buffer & Flow Control setup */ + ixgbe_config_delay_values(adapter); -/* - * Checks whether the adapter's ports are capable of - * Wake On LAN by reading the adapter's NVM. - * - * Sets each port's hw->wol_enabled value depending - * on the value read here. - */ -static void -ixgbe_check_wol_support(struct adapter *adapter) -{ - struct ixgbe_hw *hw = &adapter->hw; - u16 dev_caps = 0; + /* Initialize the FC settings */ + ixgbe_start_hw(hw); - /* Find out WoL support for port */ - adapter->wol_support = hw->wol_enabled = 0; - ixgbe_get_device_caps(hw, &dev_caps); - if ((dev_caps & IXGBE_DEVICE_CAPS_WOL_PORT0_1) || - ((dev_caps & IXGBE_DEVICE_CAPS_WOL_PORT0) && - hw->bus.func == 0)) - adapter->wol_support = hw->wol_enabled = 1; + /* Set up VLAN support and filter */ + ixgbe_setup_vlan_hw_support(ctx); - /* Save initial wake up filter configuration */ - adapter->wufc = IXGBE_READ_REG(hw, IXGBE_WUFC); + /* Setup DMA Coalescing */ + ixgbe_config_dmac(adapter); - return; + /* And now turn on interrupts */ + ixgbe_if_enable_intr(ctx); + +#ifdef PCI_IOV + /* Enable the use of the MBX by the VF's */ + { + u32 reg = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT); + reg |= IXGBE_CTRL_EXT_PFRSTD; + IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, reg); + } +#endif + + /* Now inform the stack we're ready */ + ifp->if_drv_flags |= IFF_DRV_RUNNING; } /* - * Prepare the adapter/port for LPLU and/or WoL - */ -static int -ixgbe_setup_low_power_mode(struct adapter *adapter) +** Setup the correct IVAR register for a particular MSIX interrupt +** (yes this is all very magic and confusing :) +** - entry is the register array entry +** - vector is the MSIX vector for this queue +** - type is RX/TX/MISC +*/ +static void +ixgbe_set_ivar(struct adapter *adapter, u8 entry, u8 vector, s8 type) { struct ixgbe_hw *hw = &adapter->hw; - device_t dev = adapter->dev; - s32 error = 0; + u32 ivar, index; - mtx_assert(&adapter->core_mtx, MA_OWNED); + vector |= IXGBE_IVAR_ALLOC_VAL; - if (!hw->wol_enabled) - ixgbe_set_phy_power(hw, FALSE); + switch (hw->mac.type) { - /* Limit power management flow to X550EM baseT */ - if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T - && hw->phy.ops.enter_lplu) { - /* Turn off support for APM wakeup. (Using ACPI instead) */ - IXGBE_WRITE_REG(hw, IXGBE_GRC, - IXGBE_READ_REG(hw, IXGBE_GRC) & ~(u32)2); + case ixgbe_mac_82598EB: + if (type == -1) + entry = IXGBE_IVAR_OTHER_CAUSES_INDEX; + else + entry += (type * 64); + index = (entry >> 2) & 0x1F; + ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index)); + ivar &= ~(0xFF << (8 * (entry & 0x3))); + ivar |= (vector << (8 * (entry & 0x3))); + IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR(index), ivar); + break; - /* - * Clear Wake Up Status register to prevent any previous wakeup - * events from waking us up immediately after we suspend. - */ - IXGBE_WRITE_REG(hw, IXGBE_WUS, 0xffffffff); + case ixgbe_mac_82599EB: + case ixgbe_mac_X540: + case ixgbe_mac_X550: + case ixgbe_mac_X550EM_x: + if (type == -1) { /* MISC IVAR */ + index = (entry & 1) * 8; + ivar = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC); + ivar &= ~(0xFF << index); + ivar |= (vector << index); + IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, ivar); + } else { /* RX/TX IVARS */ + index = (16 * (entry & 1)) + (8 * type); + ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(entry >> 1)); + ivar &= ~(0xFF << index); + ivar |= (vector << index); + IXGBE_WRITE_REG(hw, IXGBE_IVAR(entry >> 1), ivar); + } - /* - * Program the Wakeup Filter Control register with user filter - * settings - */ - IXGBE_WRITE_REG(hw, IXGBE_WUFC, adapter->wufc); + default: + break; + } +} - /* Enable wakeups and power management in Wakeup Control */ - IXGBE_WRITE_REG(hw, IXGBE_WUC, - IXGBE_WUC_WKEN | IXGBE_WUC_PME_EN); +static void +ixgbe_configure_ivars(struct adapter *adapter) +{ + struct ix_queue *que = adapter->queues; + u32 newitr; - /* X550EM baseT adapters need a special LPLU flow */ - hw->phy.reset_disable = true; - ixgbe_stop(adapter); - error = hw->phy.ops.enter_lplu(hw); - if (error) - device_printf(dev, - "Error entering LPLU: %d\n", error); - hw->phy.reset_disable = false; - } else { - /* Just stop for other adapters */ - ixgbe_stop(adapter); + if (ixgbe_max_interrupt_rate > 0) + newitr = (4000000 / ixgbe_max_interrupt_rate) & 0x0FF8; + else { + /* + ** Disable DMA coalescing if interrupt moderation is + ** disabled. + */ + adapter->dmac = 0; + newitr = 0; } - return error; -} + for (int i = 0; i < adapter->num_queues; i++, que++) { + struct rx_ring *rxr = &adapter->queues[i].rxr; + struct tx_ring *txr = &adapter->queues[i].txr; + /* First the RX queue entry */ + ixgbe_set_ivar(adapter, rxr->me, que->msix, 0); + /* ... and the TX */ + ixgbe_set_ivar(adapter, txr->me, que->msix, 1); + /* Set an Initial EITR value */ + IXGBE_WRITE_REG(&adapter->hw, + IXGBE_EITR(que->msix), newitr); + } -/********************************************************************** - * - * Update the board statistics counters. - * - **********************************************************************/ + /* For the Link interrupt */ + ixgbe_set_ivar(adapter, 1, adapter->vector, -1); +} + static void -ixgbe_update_stats_counters(struct adapter *adapter) +ixgbe_config_gpie(struct adapter *adapter) { struct ixgbe_hw *hw = &adapter->hw; - u32 missed_rx = 0, bprc, lxon, lxoff, total; - u64 total_missed_rx = 0; - - adapter->stats.pf.crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS); - adapter->stats.pf.illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC); - adapter->stats.pf.errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC); - adapter->stats.pf.mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC); + u32 gpie; - for (int i = 0; i < 16; i++) { - adapter->stats.pf.qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i)); - adapter->stats.pf.qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i)); - adapter->stats.pf.qprdc[i] += IXGBE_READ_REG(hw, IXGBE_QPRDC(i)); - } - adapter->stats.pf.mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC); - adapter->stats.pf.mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC); - adapter->stats.pf.rlec += IXGBE_READ_REG(hw, IXGBE_RLEC); + gpie = IXGBE_READ_REG(hw, IXGBE_GPIE); - /* Hardware workaround, gprc counts missed packets */ - adapter->stats.pf.gprc += IXGBE_READ_REG(hw, IXGBE_GPRC); - adapter->stats.pf.gprc -= missed_rx; + /* Fan Failure Interrupt */ + if (hw->device_id == IXGBE_DEV_ID_82598AT) + gpie |= IXGBE_SDP1_GPIEN; - if (hw->mac.type != ixgbe_mac_82598EB) { - adapter->stats.pf.gorc += IXGBE_READ_REG(hw, IXGBE_GORCL) + - ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32); - adapter->stats.pf.gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL) + - ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32); - adapter->stats.pf.tor += IXGBE_READ_REG(hw, IXGBE_TORL) + - ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32); - adapter->stats.pf.lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT); - adapter->stats.pf.lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT); - } else { - adapter->stats.pf.lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC); - adapter->stats.pf.lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC); - /* 82598 only has a counter in the high register */ - adapter->stats.pf.gorc += IXGBE_READ_REG(hw, IXGBE_GORCH); - adapter->stats.pf.gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH); - adapter->stats.pf.tor += IXGBE_READ_REG(hw, IXGBE_TORH); + /* + * Module detection (SDP2) + * Media ready (SDP1) + */ + if (hw->mac.type == ixgbe_mac_82599EB) { + gpie |= IXGBE_SDP2_GPIEN; + if (hw->device_id != IXGBE_DEV_ID_82599_QSFP_SF_QP) + gpie |= IXGBE_SDP1_GPIEN; } /* - * Workaround: mprc hardware is incorrectly counting - * broadcasts, so for now we subtract those. + * Thermal Failure Detection (X540) + * Link Detection (X557) */ - bprc = IXGBE_READ_REG(hw, IXGBE_BPRC); - adapter->stats.pf.bprc += bprc; - adapter->stats.pf.mprc += IXGBE_READ_REG(hw, IXGBE_MPRC); - if (hw->mac.type == ixgbe_mac_82598EB) - adapter->stats.pf.mprc -= bprc; - - adapter->stats.pf.prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64); - adapter->stats.pf.prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127); - adapter->stats.pf.prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255); - adapter->stats.pf.prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511); - adapter->stats.pf.prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023); - adapter->stats.pf.prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522); + if (hw->mac.type == ixgbe_mac_X540 || + hw->device_id == IXGBE_DEV_ID_X550EM_X_SFP || + hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) + gpie |= IXGBE_SDP0_GPIEN_X540; - lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC); - adapter->stats.pf.lxontxc += lxon; - lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC); - adapter->stats.pf.lxofftxc += lxoff; - total = lxon + lxoff; + if (adapter->intr_type == IFLIB_INTR_MSIX) { + /* Enable Enhanced MSIX mode */ + gpie |= IXGBE_GPIE_MSIX_MODE; + gpie |= IXGBE_GPIE_EIAME | IXGBE_GPIE_PBA_SUPPORT | + IXGBE_GPIE_OCD; + } - adapter->stats.pf.gptc += IXGBE_READ_REG(hw, IXGBE_GPTC); - adapter->stats.pf.mptc += IXGBE_READ_REG(hw, IXGBE_MPTC); - adapter->stats.pf.ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64); - adapter->stats.pf.gptc -= total; - adapter->stats.pf.mptc -= total; - adapter->stats.pf.ptc64 -= total; - adapter->stats.pf.gotc -= total * ETHER_MIN_LEN; + IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie); +} - adapter->stats.pf.ruc += IXGBE_READ_REG(hw, IXGBE_RUC); - adapter->stats.pf.rfc += IXGBE_READ_REG(hw, IXGBE_RFC); - adapter->stats.pf.roc += IXGBE_READ_REG(hw, IXGBE_ROC); - adapter->stats.pf.rjc += IXGBE_READ_REG(hw, IXGBE_RJC); - adapter->stats.pf.mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC); - adapter->stats.pf.mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC); - adapter->stats.pf.mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC); - adapter->stats.pf.tpr += IXGBE_READ_REG(hw, IXGBE_TPR); - adapter->stats.pf.tpt += IXGBE_READ_REG(hw, IXGBE_TPT); - adapter->stats.pf.ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127); - adapter->stats.pf.ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255); - adapter->stats.pf.ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511); - adapter->stats.pf.ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023); - adapter->stats.pf.ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522); - adapter->stats.pf.bptc += IXGBE_READ_REG(hw, IXGBE_BPTC); - adapter->stats.pf.xec += IXGBE_READ_REG(hw, IXGBE_XEC); - adapter->stats.pf.fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC); - adapter->stats.pf.fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST); - /* Only read FCOE on 82599 */ - if (hw->mac.type != ixgbe_mac_82598EB) { - adapter->stats.pf.fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC); - adapter->stats.pf.fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC); - adapter->stats.pf.fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC); - adapter->stats.pf.fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC); - adapter->stats.pf.fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC); +/* + * Requires adapter->max_frame_size to be set. + */ +static void +ixgbe_config_delay_values(struct adapter *adapter) +{ + struct ixgbe_hw *hw = &adapter->hw; + u32 rxpb, frame, size, tmp; + + frame = adapter->max_frame_size; + + /* Calculate High Water */ + switch (hw->mac.type) { + case ixgbe_mac_X540: + case ixgbe_mac_X550: + case ixgbe_mac_X550EM_x: + tmp = IXGBE_DV_X540(frame, frame); + break; + default: + tmp = IXGBE_DV(frame, frame); + break; + } + size = IXGBE_BT2KB(tmp); + rxpb = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0)) >> 10; + hw->fc.high_water[0] = rxpb - size; + + /* Now calculate Low Water */ + switch (hw->mac.type) { + case ixgbe_mac_X540: + case ixgbe_mac_X550: + case ixgbe_mac_X550EM_x: + tmp = IXGBE_LOW_DV_X540(frame); + break; + default: + tmp = IXGBE_LOW_DV(frame); + break; } + hw->fc.low_water[0] = IXGBE_BT2KB(tmp); + + hw->fc.requested_mode = adapter->fc; + hw->fc.pause_time = IXGBE_FC_PAUSE; + hw->fc.send_xon = TRUE; +} - /* Fill out the OS statistics structure */ - IXGBE_SET_IPACKETS(adapter, adapter->stats.pf.gprc); - IXGBE_SET_OPACKETS(adapter, adapter->stats.pf.gptc); - IXGBE_SET_IBYTES(adapter, adapter->stats.pf.gorc); - IXGBE_SET_OBYTES(adapter, adapter->stats.pf.gotc); - IXGBE_SET_IMCASTS(adapter, adapter->stats.pf.mprc); - IXGBE_SET_OMCASTS(adapter, adapter->stats.pf.mptc); - IXGBE_SET_COLLISIONS(adapter, 0); - IXGBE_SET_IQDROPS(adapter, total_missed_rx); - IXGBE_SET_IERRORS(adapter, adapter->stats.pf.crcerrs - + adapter->stats.pf.rlec); +/********************************************************************* + * Multicast Update + * + * This routine is called whenever multicast address list is updated. + * + **********************************************************************/ +#define IXGBE_RAR_ENTRIES 16 + +static int +ixgbe_mc_filter_apply(void *arg, struct ifmultiaddr *ifma, int count) +{ + struct adapter *adapter = arg; + struct ixgbe_mc_addr *mta = adapter->mta; + + if (ifma->ifma_addr->sa_family != AF_LINK) + return (0); + if (count == MAX_NUM_MULTICAST_ADDRESSES) + return (0); + bcopy(LLADDR((struct sockaddr_dl *) ifma->ifma_addr), + mta[count].addr, IXGBE_ETH_LENGTH_OF_ADDRESS); + mta[count].vmdq = adapter->pool; + return (1); } -#if __FreeBSD_version >= 1100036 -static uint64_t -ixgbe_get_counter(struct ifnet *ifp, ift_counter cnt) +static void +ixgbe_if_multi_set(if_ctx_t ctx) { - struct adapter *adapter; - struct tx_ring *txr; - uint64_t rv; + u32 fctrl; + u8 *update_ptr; + struct adapter *adapter = iflib_get_softc(ctx); + struct ixgbe_mc_addr *mta; + int mcnt = 0; + struct ifnet *ifp = iflib_get_ifp(ctx); + + IOCTL_DEBUGOUT("ixgbe_if_multi_set: begin"); - adapter = if_getsoftc(ifp); + mta = adapter->mta; + bzero(mta, sizeof(*mta) * MAX_NUM_MULTICAST_ADDRESSES); - switch (cnt) { - case IFCOUNTER_IPACKETS: - return (adapter->ipackets); - case IFCOUNTER_OPACKETS: - return (adapter->opackets); - case IFCOUNTER_IBYTES: - return (adapter->ibytes); - case IFCOUNTER_OBYTES: - return (adapter->obytes); - case IFCOUNTER_IMCASTS: - return (adapter->imcasts); - case IFCOUNTER_OMCASTS: - return (adapter->omcasts); - case IFCOUNTER_COLLISIONS: - return (0); - case IFCOUNTER_IQDROPS: - return (adapter->iqdrops); - case IFCOUNTER_OQDROPS: - rv = 0; - txr = adapter->tx_rings; - for (int i = 0; i < adapter->num_queues; i++, txr++) - rv += txr->br->br_drops; - return (rv); - case IFCOUNTER_IERRORS: - return (adapter->ierrors); - default: - return (if_get_counter_default(ifp, cnt)); + mcnt = if_multi_apply(iflib_get_ifp(ctx), ixgbe_mc_filter_apply, adapter); + + fctrl = IXGBE_READ_REG(&adapter->hw, IXGBE_FCTRL); + fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE); + if (ifp->if_flags & IFF_PROMISC) + fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE); + else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES || + ifp->if_flags & IFF_ALLMULTI) { + fctrl |= IXGBE_FCTRL_MPE; + fctrl &= ~IXGBE_FCTRL_UPE; + } else + fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE); + + IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCTRL, fctrl); + + if (mcnt < MAX_NUM_MULTICAST_ADDRESSES) { + update_ptr = (u8 *)mta; + ixgbe_update_mc_addr_list(&adapter->hw, update_ptr, mcnt, ixgbe_mc_array_itr, TRUE); } + + IOCTL_DEBUGOUT("ixgbe_if_multi_set: end"); } -#endif -/** ixgbe_sysctl_tdh_handler - Handler function - * Retrieves the TDH value from the hardware +/* + * This is an iterator function now needed by the multicast + * shared code. It simply feeds the shared code routine the + * addresses in the array of ixgbe_set_multi() one by one. */ -static int -ixgbe_sysctl_tdh_handler(SYSCTL_HANDLER_ARGS) +static u8 * +ixgbe_mc_array_itr(struct ixgbe_hw *hw, u8 **update_ptr, u32 *vmdq) { - int error; + struct ixgbe_mc_addr *mta; - struct tx_ring *txr = ((struct tx_ring *)oidp->oid_arg1); - if (!txr) return 0; + mta = (struct ixgbe_mc_addr *)*update_ptr; + *vmdq = mta->vmdq; - unsigned val = IXGBE_READ_REG(&txr->adapter->hw, IXGBE_TDH(txr->me)); - error = sysctl_handle_int(oidp, &val, 0, req); - if (error || !req->newptr) - return error; - return 0; + *update_ptr = (u8*)(mta + 1);; + return (mta->addr); } -/** ixgbe_sysctl_tdt_handler - Handler function - * Retrieves the TDT value from the hardware - */ -static int -ixgbe_sysctl_tdt_handler(SYSCTL_HANDLER_ARGS) +/********************************************************************* + * Timer routine + * + * This routine checks for link status,updates statistics, + * and runs the watchdog check. + * + **********************************************************************/ + +static void +ixgbe_if_timer(if_ctx_t ctx, uint16_t qid) { - int error; + struct adapter *adapter = iflib_get_softc(ctx); + struct ix_queue *que = &adapter->queues[qid]; + u64 queues = 0; - struct tx_ring *txr = ((struct tx_ring *)oidp->oid_arg1); - if (!txr) return 0; + /* Keep track of queues with work for soft irq */ + if (que->txr.busy) + queues |= ((u64)1 << que->me); - unsigned val = IXGBE_READ_REG(&txr->adapter->hw, IXGBE_TDT(txr->me)); - error = sysctl_handle_int(oidp, &val, 0, req); - if (error || !req->newptr) - return error; - return 0; + if (qid != 0) + return; + + /* Check for pluggable optics */ + if (adapter->sfp_probe) + if (!ixgbe_sfp_probe(ctx)) + return; /* Nothing to do */ + + ixgbe_check_link(&adapter->hw, + &adapter->link_speed, &adapter->link_up, 0); + ixgbe_if_update_admin_status(ctx); + ixgbe_update_stats_counters(adapter); + + + /* Fire off the adminq task */ + iflib_admin_intr_deferred(ctx); } -/** ixgbe_sysctl_rdh_handler - Handler function - * Retrieves the RDH value from the hardware - */ -static int -ixgbe_sysctl_rdh_handler(SYSCTL_HANDLER_ARGS) +/* +** ixgbe_sfp_probe - called in the local timer to +** determine if a port had optics inserted. +*/ +static bool +ixgbe_sfp_probe(if_ctx_t ctx) { - int error; - - struct rx_ring *rxr = ((struct rx_ring *)oidp->oid_arg1); - if (!rxr) return 0; + struct adapter *adapter = iflib_get_softc(ctx); + struct ixgbe_hw *hw = &adapter->hw; + device_t dev = iflib_get_dev(ctx); + bool result = FALSE; - unsigned val = IXGBE_READ_REG(&rxr->adapter->hw, IXGBE_RDH(rxr->me)); - error = sysctl_handle_int(oidp, &val, 0, req); - if (error || !req->newptr) - return error; - return 0; + if ((hw->phy.type == ixgbe_phy_nl) && + (hw->phy.sfp_type == ixgbe_sfp_type_not_present)) { + s32 ret = hw->phy.ops.identify_sfp(hw); + if (ret) + goto out; + ret = hw->phy.ops.reset(hw); + if (ret == IXGBE_ERR_SFP_NOT_SUPPORTED) { + device_printf(dev,"Unsupported SFP+ module detected!"); + printf(" Reload driver with supported module.\n"); + adapter->sfp_probe = FALSE; + goto out; + } else + device_printf(dev,"SFP+ module detected!\n"); + /* We now have supported optics */ + adapter->sfp_probe = FALSE; + /* Set the optics type so system reports correctly */ + ixgbe_setup_optics(adapter); + result = TRUE; + } +out: + return (result); } -/** ixgbe_sysctl_rdt_handler - Handler function - * Retrieves the RDT value from the hardware - */ -static int -ixgbe_sysctl_rdt_handler(SYSCTL_HANDLER_ARGS) +/************************************************************************* + * + * Tasklet handler - ixgbe_handle_mod, ixgbe_handle_msf, ixgbe_handle_phy + * + *************************************************************************/ +/* +** Tasklet for handling SFP module interrupts +*/ +static void +ixgbe_handle_mod(void *context, int pending) { - int error; + struct adapter *adapter = context; + struct ixgbe_hw *hw = &adapter->hw; + device_t dev = adapter->dev; + u32 err; - struct rx_ring *rxr = ((struct rx_ring *)oidp->oid_arg1); - if (!rxr) return 0; + err = hw->phy.ops.identify_sfp(hw); + if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) { + device_printf(dev, + "Unsupported SFP+ module type was detected.\n"); + return; + } - unsigned val = IXGBE_READ_REG(&rxr->adapter->hw, IXGBE_RDT(rxr->me)); - error = sysctl_handle_int(oidp, &val, 0, req); - if (error || !req->newptr) - return error; - return 0; + err = hw->mac.ops.setup_sfp(hw); + if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) { + device_printf(dev, + "Setup failure - unsupported SFP+ module type.\n"); + return; + } + GROUPTASK_ENQUEUE(&adapter->msf_task); + return; } -static int -ixgbe_sysctl_interrupt_rate_handler(SYSCTL_HANDLER_ARGS) + +/* Tasklet for handling MSF (multispeed fiber) interrupts */ + +static void +ixgbe_handle_msf(void *context, int pending) { - int error; - struct ix_queue *que = ((struct ix_queue *)oidp->oid_arg1); - unsigned int reg, usec, rate; + struct adapter *adapter = context; + struct ixgbe_hw *hw = &adapter->hw; + u32 autoneg; + bool negotiate; + int err; - reg = IXGBE_READ_REG(&que->adapter->hw, IXGBE_EITR(que->msix)); - usec = ((reg & 0x0FF8) >> 3); - if (usec > 0) - rate = 500000 / usec; - else - rate = 0; - error = sysctl_handle_int(oidp, &rate, 0, req); - if (error || !req->newptr) - return error; - reg &= ~0xfff; /* default, no limitation */ - ixgbe_max_interrupt_rate = 0; - if (rate > 0 && rate < 500000) { - if (rate < 1000) - rate = 1000; - ixgbe_max_interrupt_rate = rate; - reg |= ((4000000/rate) & 0xff8 ); - } - IXGBE_WRITE_REG(&que->adapter->hw, IXGBE_EITR(que->msix), reg); - return 0; + err = hw->phy.ops.identify_sfp(hw); + if (!err) { + ixgbe_setup_optics(adapter); + INIT_DEBUGOUT1("ixgbe_sfp_probe: flags: %X\n", adapter->optics); + } + + autoneg = hw->phy.autoneg_advertised; + if ((!autoneg) && (hw->mac.ops.get_link_capabilities)) + hw->mac.ops.get_link_capabilities(hw, &autoneg, &negotiate); + if (hw->mac.ops.setup_link) + hw->mac.ops.setup_link(hw, autoneg, TRUE); + + ifmedia_removeall(adapter->media); + ixgbe_add_media_types(adapter->ctx); + return; } + +/* Tasklet for handling interrupts from an external PHY */ + static void -ixgbe_add_device_sysctls(struct adapter *adapter) +ixgbe_handle_phy(void *context, int pending) { - device_t dev = adapter->dev; + struct adapter *adapter = context; struct ixgbe_hw *hw = &adapter->hw; - struct sysctl_oid_list *child; - struct sysctl_ctx_list *ctx; - - ctx = device_get_sysctl_ctx(dev); - child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev)); + int error; - /* Sysctls for all devices */ - SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "fc", - CTLTYPE_INT | CTLFLAG_RW, adapter, 0, - ixgbe_sysctl_flowcntl, "I", IXGBE_SYSCTL_DESC_SET_FC); + error = hw->phy.ops.handle_lasi(hw); + if (error == IXGBE_ERR_OVERTEMP) + device_printf(adapter->dev, + "CRITICAL: EXTERNAL PHY OVER TEMP!! " + " PHY will downshift to lower power state!\n"); + else if (error) + device_printf(adapter->dev, + "Error handling LASI interrupt: %d\n", + error); + return; +} - SYSCTL_ADD_INT(ctx, child, OID_AUTO, "enable_aim", - CTLFLAG_RW, - &ixgbe_enable_aim, 1, "Interrupt Moderation"); - SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "advertise_speed", - CTLTYPE_INT | CTLFLAG_RW, adapter, 0, - ixgbe_sysctl_advertise, "I", IXGBE_SYSCTL_DESC_ADV_SPEED); +/********************************************************************* + * + * This routine disables all traffic on the adapter by issuing a + * global reset on the MAC and deallocates TX/RX buffers. + * + **********************************************************************/ +static void +ixgbe_if_stop(if_ctx_t ctx) +{ + struct ifnet *ifp; + struct adapter *adapter = iflib_get_softc(ctx); - SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "thermal_test", - CTLTYPE_INT | CTLFLAG_RW, adapter, 0, - ixgbe_sysctl_thermal_test, "I", "Thermal Test"); + struct ixgbe_hw *hw = &adapter->hw; + ifp = iflib_get_ifp(ctx); + + INIT_DEBUGOUT("ixgbe_stop: begin\n"); -#ifdef IXGBE_DEBUG - /* testing sysctls (for all devices) */ - SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "power_state", - CTLTYPE_INT | CTLFLAG_RW, adapter, 0, - ixgbe_sysctl_power_state, "I", "PCI Power State"); - - SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "print_rss_config", - CTLTYPE_STRING | CTLFLAG_RD, adapter, 0, - ixgbe_sysctl_print_rss_config, "A", "Prints RSS Configuration"); -#endif - /* for X550 series devices */ - if (hw->mac.type >= ixgbe_mac_X550) - SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "dmac", - CTLTYPE_INT | CTLFLAG_RW, adapter, 0, - ixgbe_sysctl_dmac, "I", "DMA Coalesce"); + ixgbe_reset_hw(hw); + hw->adapter_stopped = FALSE; + ixgbe_stop_adapter(hw); + if (hw->mac.type == ixgbe_mac_82599EB) + ixgbe_stop_mac_link_on_d3_82599(hw); + /* Turn off the laser - noop with no optics */ + ixgbe_disable_tx_laser(hw); + + /* Update the stack */ + adapter->link_up = FALSE; + ixgbe_if_update_admin_status(ctx); + + /* reprogram the RAR[0] in case user changed it. */ + ixgbe_set_rar(&adapter->hw, 0, adapter->hw.mac.addr, 0, IXGBE_RAH_AV); + + return; +} - /* for X552 backplane devices */ - if (hw->device_id == IXGBE_DEV_ID_X550EM_X_KR) { - struct sysctl_oid *eee_node; - struct sysctl_oid_list *eee_list; +/* +** Note: this routine updates the OS on the link state +** the real check of the hardware only happens with +** a link interrupt. +*/ +static void +ixgbe_if_update_admin_status(if_ctx_t ctx) +{ + struct adapter *adapter = iflib_get_softc(ctx); + struct ixgbe_hw *hw = &adapter->hw; + device_t dev = iflib_get_dev(ctx); + + if (adapter->link_up){ + if (adapter->link_active == FALSE) { + if (bootverbose) + device_printf(dev,"Link is up %d Gbps %s \n", + ((adapter->link_speed == 128)? 10:1), + "Full Duplex"); + adapter->link_active = TRUE; + /* Update any Flow Control changes */ + ixgbe_fc_enable(hw); + /* Update DMA coalescing config */ + ixgbe_config_dmac(adapter); + iflib_link_state_change(ctx, LINK_STATE_UP); - eee_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "eee", - CTLFLAG_RD, NULL, - "Energy Efficient Ethernet sysctls"); - eee_list = SYSCTL_CHILDREN(eee_node); +#ifdef PCI_IOV + ixgbe_ping_all_vfs(adapter); +#endif + + } + } else { /* Link down */ + if (adapter->link_active == TRUE) { + if (bootverbose) + device_printf(dev,"Link is Down\n"); + iflib_link_state_change(ctx, LINK_STATE_DOWN); + adapter->link_active = FALSE; +#ifdef PCI_IOV + ixgbe_ping_all_vfs(adapter); +#endif + } + } + return; +} - SYSCTL_ADD_PROC(ctx, eee_list, OID_AUTO, "enable", - CTLTYPE_INT | CTLFLAG_RW, adapter, 0, - ixgbe_sysctl_eee_enable, "I", - "Enable or Disable EEE"); +#ifdef IXGBE_FDIR +/* +** Tasklet for reinitializing the Flow Director filter table +*/ +static void +ixgbe_reinit_fdir(void *context, int pending) +{ + struct adapter *adapter = context; + struct ifnet *ifp = iflib_get_ifp(adapter->ctx); - SYSCTL_ADD_PROC(ctx, eee_list, OID_AUTO, "negotiated", - CTLTYPE_INT | CTLFLAG_RD, adapter, 0, - ixgbe_sysctl_eee_negotiated, "I", - "EEE negotiated on link"); + if (adapter->fdir_reinit != 1) /* Shouldn't happen */ + return; + ixgbe_reinit_fdir_tables_82599(&adapter->hw); + adapter->fdir_reinit = 0; + /* re-enable flow director interrupts */ + IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR); + /* Restart the interface */ + ifp->if_drv_flags |= IFF_DRV_RUNNING; +} +#endif + - SYSCTL_ADD_PROC(ctx, eee_list, OID_AUTO, "tx_lpi_status", - CTLTYPE_INT | CTLFLAG_RD, adapter, 0, - ixgbe_sysctl_eee_tx_lpi_status, "I", - "Whether or not TX link is in LPI state"); +/********************************************************************* + * + * Configure DMA Coalescing + * + **********************************************************************/ +static void +ixgbe_config_dmac(struct adapter *adapter) +{ + struct ixgbe_hw *hw = &adapter->hw; + struct ixgbe_dmac_config *dcfg = &hw->mac.dmac_config; - SYSCTL_ADD_PROC(ctx, eee_list, OID_AUTO, "rx_lpi_status", - CTLTYPE_INT | CTLFLAG_RD, adapter, 0, - ixgbe_sysctl_eee_rx_lpi_status, "I", - "Whether or not RX link is in LPI state"); + if (hw->mac.type < ixgbe_mac_X550 || + !hw->mac.ops.dmac_config) + return; - SYSCTL_ADD_PROC(ctx, eee_list, OID_AUTO, "tx_lpi_delay", - CTLTYPE_INT | CTLFLAG_RD, adapter, 0, - ixgbe_sysctl_eee_tx_lpi_delay, "I", - "TX LPI entry delay in microseconds"); + if (dcfg->watchdog_timer ^ adapter->dmac || + dcfg->link_speed ^ adapter->link_speed) { + dcfg->watchdog_timer = adapter->dmac; + dcfg->fcoe_en = false; + dcfg->link_speed = adapter->link_speed; + dcfg->num_tcs = 1; + + INIT_DEBUGOUT2("dmac settings: watchdog %d, link speed %d\n", + dcfg->watchdog_timer, dcfg->link_speed); + + hw->mac.ops.dmac_config(hw); } +} - /* for WoL-capable devices */ - if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) { - SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "wol_enable", - CTLTYPE_INT | CTLFLAG_RW, adapter, 0, - ixgbe_sysctl_wol_enable, "I", - "Enable/Disable Wake on LAN"); +void +ixgbe_if_enable_intr(if_ctx_t ctx) +{ + struct adapter *adapter = iflib_get_softc(ctx); + struct ixgbe_hw *hw = &adapter->hw; + struct ix_queue *que = adapter->queues; + u32 mask, fwsm; - SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "wufc", - CTLTYPE_INT | CTLFLAG_RW, adapter, 0, - ixgbe_sysctl_wufc, "I", - "Enable/Disable Wake Up Filters"); + mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE); + /* Enable Fan Failure detection */ + if (hw->device_id == IXGBE_DEV_ID_82598AT) + mask |= IXGBE_EIMS_GPI_SDP1; + + switch (adapter->hw.mac.type) { + case ixgbe_mac_82599EB: + mask |= IXGBE_EIMS_ECC; + /* Temperature sensor on some adapters */ + mask |= IXGBE_EIMS_GPI_SDP0; + /* SFP+ (RX_LOS_N & MOD_ABS_N) */ + mask |= IXGBE_EIMS_GPI_SDP1; + mask |= IXGBE_EIMS_GPI_SDP2; +#ifdef IXGBE_FDIR + mask |= IXGBE_EIMS_FLOW_DIR; +#endif +#ifdef PCI_IOV + mask |= IXGBE_EIMS_MAILBOX; +#endif + break; + case ixgbe_mac_X540: + /* Detect if Thermal Sensor is enabled */ + fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM); + if (fwsm & IXGBE_FWSM_TS_ENABLED) + mask |= IXGBE_EIMS_TS; + mask |= IXGBE_EIMS_ECC; +#ifdef IXGBE_FDIR + mask |= IXGBE_EIMS_FLOW_DIR; +#endif + break; + case ixgbe_mac_X550: + case ixgbe_mac_X550EM_x: + /* MAC thermal sensor is automatically enabled */ + mask |= IXGBE_EIMS_TS; + /* Some devices use SDP0 for important information */ + if (hw->device_id == IXGBE_DEV_ID_X550EM_X_SFP || + hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) + mask |= IXGBE_EIMS_GPI_SDP0_BY_MAC(hw); + mask |= IXGBE_EIMS_ECC; +#ifdef IXGBE_FDIR + mask |= IXGBE_EIMS_FLOW_DIR; +#endif +#ifdef PCI_IOV + mask |= IXGBE_EIMS_MAILBOX; +#endif + /* falls through */ + default: + break; } - /* for X552/X557-AT devices */ - if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) { - struct sysctl_oid *phy_node; - struct sysctl_oid_list *phy_list; + IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask); - phy_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "phy", - CTLFLAG_RD, NULL, - "External PHY sysctls"); - phy_list = SYSCTL_CHILDREN(phy_node); + /* With MSI-X we use auto clear */ + if (adapter->intr_type == IFLIB_INTR_MSIX) { + mask = IXGBE_EIMS_ENABLE_MASK; + /* Don't autoclear Link */ + mask &= ~IXGBE_EIMS_OTHER; + mask &= ~IXGBE_EIMS_LSC; +#ifdef PCI_IOV + mask &= ~IXGBE_EIMS_MAILBOX; +#endif + IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask); + } - SYSCTL_ADD_PROC(ctx, phy_list, OID_AUTO, "temp", - CTLTYPE_INT | CTLFLAG_RD, adapter, 0, - ixgbe_sysctl_phy_temp, "I", - "Current External PHY Temperature (Celsius)"); + /* + ** Now enable all queues, this is done separately to + ** allow for handling the extended (beyond 32) MSIX + ** vectors that can be used by 82599 + */ + for (int i = 0; i < adapter->num_queues; i++, que++) + ixgbe_enable_queue(adapter, que->msix); - SYSCTL_ADD_PROC(ctx, phy_list, OID_AUTO, "overtemp_occurred", - CTLTYPE_INT | CTLFLAG_RD, adapter, 0, - ixgbe_sysctl_phy_overtemp_occurred, "I", - "External PHY High Temperature Event Occurred"); - } + IXGBE_WRITE_FLUSH(hw); } - -/* - * Add sysctl variables, one per statistic, to the system. - */ + static void -ixgbe_add_hw_stats(struct adapter *adapter) +ixgbe_if_disable_intr(if_ctx_t ctx) { - device_t dev = adapter->dev; - - struct tx_ring *txr = adapter->tx_rings; - struct rx_ring *rxr = adapter->rx_rings; - - struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev); - struct sysctl_oid *tree = device_get_sysctl_tree(dev); - struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree); - struct ixgbe_hw_stats *stats = &adapter->stats.pf; - - struct sysctl_oid *stat_node, *queue_node; - struct sysctl_oid_list *stat_list, *queue_list; + struct adapter *adapter = iflib_get_softc(ctx); + if (adapter->intr_type == IFLIB_INTR_MSIX) + IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, 0); + if (adapter->hw.mac.type == ixgbe_mac_82598EB) { + IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0); + } else { + IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000); + IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0); + IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0); + } + IXGBE_WRITE_FLUSH(&adapter->hw); +} -#define QUEUE_NAME_LEN 32 - char namebuf[QUEUE_NAME_LEN]; - /* Driver Statistics */ - SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped", - CTLFLAG_RD, &adapter->dropped_pkts, - "Driver dropped packets"); - SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "mbuf_defrag_failed", - CTLFLAG_RD, &adapter->mbuf_defrag_failed, - "m_defrag() failed"); - SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_events", - CTLFLAG_RD, &adapter->watchdog_events, - "Watchdog timeouts"); - SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq", - CTLFLAG_RD, &adapter->link_irq, - "Link MSIX IRQ Handled"); +static void +ixgbe_if_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid) +{ + struct adapter *adapter = iflib_get_softc(ctx); + struct ix_queue *que = &adapter->queues[rxqid]; - for (int i = 0; i < adapter->num_queues; i++, txr++) { - snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i); - queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf, - CTLFLAG_RD, NULL, "Queue Name"); - queue_list = SYSCTL_CHILDREN(queue_node); + ixgbe_enable_queue(adapter, que->me); +} + +/* +** +** MSIX Interrupt Handlers and Tasklets +** +*/ - SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "interrupt_rate", - CTLTYPE_UINT | CTLFLAG_RW, &adapter->queues[i], - sizeof(&adapter->queues[i]), - ixgbe_sysctl_interrupt_rate_handler, "IU", - "Interrupt Rate"); - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "irqs", - CTLFLAG_RD, &(adapter->queues[i].irqs), - "irqs on this queue"); - SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head", - CTLTYPE_UINT | CTLFLAG_RD, txr, sizeof(txr), - ixgbe_sysctl_tdh_handler, "IU", - "Transmit Descriptor Head"); - SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail", - CTLTYPE_UINT | CTLFLAG_RD, txr, sizeof(txr), - ixgbe_sysctl_tdt_handler, "IU", - "Transmit Descriptor Tail"); - SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tso_tx", - CTLFLAG_RD, &txr->tso_tx, - "TSO"); - SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "no_tx_dma_setup", - CTLFLAG_RD, &txr->no_tx_dma_setup, - "Driver tx dma failure in xmit"); - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "no_desc_avail", - CTLFLAG_RD, &txr->no_desc_avail, - "Queue No Descriptor Available"); - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "tx_packets", - CTLFLAG_RD, &txr->total_packets, - "Queue Packets Transmitted"); - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "br_drops", - CTLFLAG_RD, &txr->br->br_drops, - "Packets dropped in buf_ring"); +static void +ixgbe_enable_queue(struct adapter *adapter, u32 vector) +{ + struct ixgbe_hw *hw = &adapter->hw; + u64 queue = (u64)(1 << vector); + u32 mask; + + if (hw->mac.type == ixgbe_mac_82598EB) { + mask = (IXGBE_EIMS_RTX_QUEUE & queue); + IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask); + } else { + mask = (queue & 0xFFFFFFFF); + if (mask) + IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask); + mask = (queue >> 32); + if (mask) + IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask); } +} - for (int i = 0; i < adapter->num_queues; i++, rxr++) { - snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i); - queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf, - CTLFLAG_RD, NULL, "Queue Name"); - queue_list = SYSCTL_CHILDREN(queue_node); +static void +ixgbe_disable_queue(struct adapter *adapter, u32 vector) +{ + struct ixgbe_hw *hw = &adapter->hw; + u64 queue = (u64)(1 << vector); + u32 mask; - struct lro_ctrl *lro = &rxr->lro; + if (hw->mac.type == ixgbe_mac_82598EB) { + mask = (IXGBE_EIMS_RTX_QUEUE & queue); + IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask); + } else { + mask = (queue & 0xFFFFFFFF); + if (mask) + IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask); + mask = (queue >> 32); + if (mask) + IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask); + } +} - snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i); - queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf, - CTLFLAG_RD, NULL, "Queue Name"); - queue_list = SYSCTL_CHILDREN(queue_node); - SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head", - CTLTYPE_UINT | CTLFLAG_RD, rxr, sizeof(rxr), - ixgbe_sysctl_rdh_handler, "IU", - "Receive Descriptor Head"); - SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail", - CTLTYPE_UINT | CTLFLAG_RD, rxr, sizeof(rxr), - ixgbe_sysctl_rdt_handler, "IU", - "Receive Descriptor Tail"); - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_packets", - CTLFLAG_RD, &rxr->rx_packets, - "Queue Packets Received"); - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_bytes", - CTLFLAG_RD, &rxr->rx_bytes, - "Queue Bytes Received"); - SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_copies", - CTLFLAG_RD, &rxr->rx_copies, - "Copied RX Frames"); - SYSCTL_ADD_U64(ctx, queue_list, OID_AUTO, "lro_queued", - CTLFLAG_RD, &lro->lro_queued, 0, - "LRO Queued"); - SYSCTL_ADD_U64(ctx, queue_list, OID_AUTO, "lro_flushed", - CTLFLAG_RD, &lro->lro_flushed, 0, - "LRO Flushed"); +/********************************************************************* + * + * Legacy Interrupt Service routine + * + **********************************************************************/ +int +ixgbe_intr(void *arg) +{ + struct ix_queue *que = arg; + struct adapter *adapter = que->adapter; + struct ixgbe_hw *hw = &adapter->hw; + if_ctx_t ctx = adapter->ctx; + u32 reg_eicr; + + reg_eicr = IXGBE_READ_REG(hw, IXGBE_EICR); + + ++que->irqs; + if (reg_eicr == 0) { + ixgbe_if_enable_intr(ctx); + return (FILTER_HANDLED); } - /* MAC stats get the own sub node */ + /* Check for fan failure */ + if ((hw->device_id == IXGBE_DEV_ID_82598AT) && + (reg_eicr & IXGBE_EICR_GPI_SDP1)) { + device_printf(adapter->dev, "\nCRITICAL: FAN FAILURE!! " + "REPLACE IMMEDIATELY!!\n"); + IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EICR_GPI_SDP1_BY_MAC(hw)); + } + + /* Link status change */ + if (reg_eicr & IXGBE_EICR_LSC) + iflib_admin_intr_deferred(ctx); - stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats", - CTLFLAG_RD, NULL, "MAC Statistics"); - stat_list = SYSCTL_CHILDREN(stat_node); + /* External PHY interrupt */ + if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T && + (reg_eicr & IXGBE_EICR_GPI_SDP0_X540)) { + GROUPTASK_ENQUEUE(&adapter->phy_task); + } - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs", - CTLFLAG_RD, &stats->crcerrs, - "CRC Errors"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "ill_errs", - CTLFLAG_RD, &stats->illerrc, - "Illegal Byte Errors"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "byte_errs", - CTLFLAG_RD, &stats->errbc, - "Byte Errors"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "short_discards", - CTLFLAG_RD, &stats->mspdc, - "MAC Short Packets Discarded"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "local_faults", - CTLFLAG_RD, &stats->mlfc, - "MAC Local Faults"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "remote_faults", - CTLFLAG_RD, &stats->mrfc, - "MAC Remote Faults"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rec_len_errs", - CTLFLAG_RD, &stats->rlec, - "Receive Length Errors"); + return (FILTER_SCHEDULE_THREAD); +} - /* Flow Control stats */ - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd", - CTLFLAG_RD, &stats->lxontxc, - "Link XON Transmitted"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd", - CTLFLAG_RD, &stats->lxonrxc, - "Link XON Received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd", - CTLFLAG_RD, &stats->lxofftxc, - "Link XOFF Transmitted"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd", - CTLFLAG_RD, &stats->lxoffrxc, - "Link XOFF Received"); +static void +ixgbe_free_pci_resources(if_ctx_t ctx) +{ + struct adapter *adapter = iflib_get_softc(ctx); + struct ix_queue *que = adapter->queues; + device_t dev = iflib_get_dev(ctx); - /* Packet Reception Stats */ - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_octets_rcvd", - CTLFLAG_RD, &stats->tor, - "Total Octets Received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_rcvd", - CTLFLAG_RD, &stats->gorc, - "Good Octets Received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_rcvd", - CTLFLAG_RD, &stats->tpr, - "Total Packets Received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_rcvd", - CTLFLAG_RD, &stats->gprc, - "Good Packets Received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_rcvd", - CTLFLAG_RD, &stats->mprc, - "Multicast Packets Received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_rcvd", - CTLFLAG_RD, &stats->bprc, - "Broadcast Packets Received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64", - CTLFLAG_RD, &stats->prc64, - "64 byte frames received "); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127", - CTLFLAG_RD, &stats->prc127, - "65-127 byte frames received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255", - CTLFLAG_RD, &stats->prc255, - "128-255 byte frames received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511", - CTLFLAG_RD, &stats->prc511, - "256-511 byte frames received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023", - CTLFLAG_RD, &stats->prc1023, - "512-1023 byte frames received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522", - CTLFLAG_RD, &stats->prc1522, - "1023-1522 byte frames received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersized", - CTLFLAG_RD, &stats->ruc, - "Receive Undersized"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented", - CTLFLAG_RD, &stats->rfc, - "Fragmented Packets Received "); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversized", - CTLFLAG_RD, &stats->roc, - "Oversized Packets Received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabberd", - CTLFLAG_RD, &stats->rjc, - "Received Jabber"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_rcvd", - CTLFLAG_RD, &stats->mngprc, - "Management Packets Received"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_drpd", - CTLFLAG_RD, &stats->mngptc, - "Management Packets Dropped"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "checksum_errs", - CTLFLAG_RD, &stats->xec, - "Checksum Errors"); + /* Release all msix queue resources */ + if (adapter->intr_type == IFLIB_INTR_MSIX) + iflib_irq_free(ctx, &adapter->irq); - /* Packet Transmission Stats */ - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd", - CTLFLAG_RD, &stats->gotc, - "Good Octets Transmitted"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd", - CTLFLAG_RD, &stats->tpt, - "Total Packets Transmitted"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd", - CTLFLAG_RD, &stats->gptc, - "Good Packets Transmitted"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd", - CTLFLAG_RD, &stats->bptc, - "Broadcast Packets Transmitted"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd", - CTLFLAG_RD, &stats->mptc, - "Multicast Packets Transmitted"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_txd", - CTLFLAG_RD, &stats->mngptc, - "Management Packets Transmitted"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64", - CTLFLAG_RD, &stats->ptc64, - "64 byte frames transmitted "); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127", - CTLFLAG_RD, &stats->ptc127, - "65-127 byte frames transmitted"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255", - CTLFLAG_RD, &stats->ptc255, - "128-255 byte frames transmitted"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511", - CTLFLAG_RD, &stats->ptc511, - "256-511 byte frames transmitted"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023", - CTLFLAG_RD, &stats->ptc1023, - "512-1023 byte frames transmitted"); - SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522", - CTLFLAG_RD, &stats->ptc1522, - "1024-1522 byte frames transmitted"); + for (int i = 0; i < adapter->num_queues; i++, que++) { + iflib_irq_free(ctx, &que->que_irq); + } + + /* + * Free link/admin interrupt + */ + if (adapter->pci_mem != NULL) + bus_release_resource(dev, SYS_RES_MEMORY, + PCIR_BAR(0), adapter->pci_mem); } static void @@ -4673,6 +3852,7 @@ OID_AUTO, name, CTLFLAG_RW, limit, value, description); } +/* Sysctls */ /* ** Set flow control using sysctl: ** Flow control values: @@ -4682,31 +3862,21 @@ ** 3 - full */ static int -ixgbe_sysctl_flowcntl(SYSCTL_HANDLER_ARGS) +ixgbe_set_flowcntl(SYSCTL_HANDLER_ARGS) { - int error, fc; - struct adapter *adapter; - - adapter = (struct adapter *) arg1; - fc = adapter->fc; + int error, last; + struct adapter *adapter = (struct adapter *) arg1; - error = sysctl_handle_int(oidp, &fc, 0, req); + last = adapter->fc; + error = sysctl_handle_int(oidp, &adapter->fc, 0, req); if ((error) || (req->newptr == NULL)) return (error); /* Don't bother if it's not changed */ - if (adapter->fc == fc) + if (adapter->fc == last) return (0); - return ixgbe_set_flowcntl(adapter, fc); -} - - -static int -ixgbe_set_flowcntl(struct adapter *adapter, int fc) -{ - - switch (fc) { + switch (adapter->fc) { case ixgbe_fc_rx_pause: case ixgbe_fc_tx_pause: case ixgbe_fc_full: @@ -4720,71 +3890,100 @@ ixgbe_enable_rx_drop(adapter); break; default: + adapter->fc = last; return (EINVAL); } - adapter->fc = fc; /* Don't autoneg if forcing a value */ adapter->hw.fc.disable_fc_autoneg = TRUE; ixgbe_fc_enable(&adapter->hw); - return (0); + return error; } /* -** Control advertised link speed: -** Flags: -** 0x1 - advertise 100 Mb -** 0x2 - advertise 1G -** 0x4 - advertise 10G +** Enable the hardware to drop packets when the buffer is +** full. This is useful when multiqueue,so that no single +** queue being full stalls the entire RX engine. We only +** enable this when Multiqueue AND when Flow Control is +** disabled. */ -static int -ixgbe_sysctl_advertise(SYSCTL_HANDLER_ARGS) +static void +ixgbe_enable_rx_drop(struct adapter *adapter) { - int error, advertise; - struct adapter *adapter; - - adapter = (struct adapter *) arg1; - advertise = adapter->advertise; + struct ixgbe_hw *hw = &adapter->hw; - error = sysctl_handle_int(oidp, &advertise, 0, req); - if ((error) || (req->newptr == NULL)) - return (error); + for (int i = 0; i < adapter->num_queues; i++) { + struct rx_ring *rxr = &adapter->queues[i].rxr; + u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(rxr->me)); + srrctl |= IXGBE_SRRCTL_DROP_EN; + IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(rxr->me), srrctl); + } +#ifdef PCI_IOV + /* enable drop for each vf */ + for (int i = 0; i < adapter->num_vfs; i++) { + IXGBE_WRITE_REG(hw, IXGBE_QDE, + (IXGBE_QDE_WRITE | (i << IXGBE_QDE_IDX_SHIFT) | + IXGBE_QDE_ENABLE)); + } +#endif +} - /* Checks to validate new value */ - if (adapter->advertise == advertise) /* no change */ - return (0); +static void +ixgbe_disable_rx_drop(struct adapter *adapter) +{ + struct ixgbe_hw *hw = &adapter->hw; - return ixgbe_set_advertise(adapter, advertise); + for (int i = 0; i < adapter->num_queues; i++) { + struct rx_ring *rxr = &adapter->queues[i].rxr; + u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(rxr->me)); + srrctl &= ~IXGBE_SRRCTL_DROP_EN; + IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(rxr->me), srrctl); + } +#ifdef PCI_IOV + /* disable drop for each vf */ + for (int i = 0; i < adapter->num_vfs; i++) { + IXGBE_WRITE_REG(hw, IXGBE_QDE, + (IXGBE_QDE_WRITE | (i << IXGBE_QDE_IDX_SHIFT))); + } +#endif } static int -ixgbe_set_advertise(struct adapter *adapter, int advertise) +ixgbe_set_advertise(SYSCTL_HANDLER_ARGS) { + int error = 0, requested; + struct adapter *adapter; device_t dev; struct ixgbe_hw *hw; - ixgbe_link_speed speed; + ixgbe_link_speed speed = 0; + adapter = (struct adapter *) arg1; + dev = iflib_get_dev(adapter->ctx); hw = &adapter->hw; - dev = adapter->dev; - /* No speed changes for backplane media */ - if (hw->phy.media_type == ixgbe_media_type_backplane) - return (ENODEV); + requested = adapter->advertise; + error = sysctl_handle_int(oidp, &requested, 0, req); + if ((error) || (req->newptr == NULL)) + return (error); + + /* Checks to validate new value */ + if (adapter->advertise == requested) /* no change */ + return (0); if (!((hw->phy.media_type == ixgbe_media_type_copper) || - (hw->phy.multispeed_fiber))) { + (hw->phy.multispeed_fiber))) { device_printf(dev, - "Advertised speed can only be set on copper or " - "multispeed fiber media types.\n"); + "Advertised speed can only be set on copper or " + "multispeed fiber media types.\n"); return (EINVAL); } - if (advertise < 0x1 || advertise > 0x7) { + if (requested < 0x1 || requested > 0x7) { device_printf(dev, - "Invalid advertised speed; valid modes are 0x1 through 0x7\n"); + "Invalid advertised speed; valid modes are 0x1 through 0x7\n"); return (EINVAL); } - if ((advertise & 0x1) + if ((requested & 0x1) && (hw->mac.type != ixgbe_mac_X540) && (hw->mac.type != ixgbe_mac_X550)) { device_printf(dev, "Set Advertise: 100Mb on X540/X550 only\n"); @@ -4792,88 +3991,24 @@ } /* Set new value and report new advertised mode */ - speed = 0; - if (advertise & 0x1) + if (requested & 0x1) speed |= IXGBE_LINK_SPEED_100_FULL; - if (advertise & 0x2) + if (requested & 0x2) speed |= IXGBE_LINK_SPEED_1GB_FULL; - if (advertise & 0x4) + if (requested & 0x4) speed |= IXGBE_LINK_SPEED_10GB_FULL; - adapter->advertise = advertise; hw->mac.autotry_restart = TRUE; hw->mac.ops.setup_link(hw, speed, TRUE); + adapter->advertise = requested; - return (0); -} - -/* - * The following two sysctls are for X552/X557-AT devices; - * they deal with the external PHY used in them. - */ -static int -ixgbe_sysctl_phy_temp(SYSCTL_HANDLER_ARGS) -{ - struct adapter *adapter = (struct adapter *) arg1; - struct ixgbe_hw *hw = &adapter->hw; - u16 reg; - - if (hw->device_id != IXGBE_DEV_ID_X550EM_X_10G_T) { - device_printf(adapter->dev, - "Device has no supported external thermal sensor.\n"); - return (ENODEV); - } - - if (hw->phy.ops.read_reg(hw, IXGBE_PHY_CURRENT_TEMP, - IXGBE_MDIO_VENDOR_SPECIFIC_1_DEV_TYPE, - ®)) { - device_printf(adapter->dev, - "Error reading from PHY's current temperature register\n"); - return (EAGAIN); - } - - /* Shift temp for output */ - reg = reg >> 8; - - return (sysctl_handle_int(oidp, NULL, reg, req)); + return (error); } -/* - * Reports whether the current PHY temperature is over - * the overtemp threshold. - * - This is reported directly from the PHY +/** Thermal Shutdown Trigger (internal MAC) + ** - Set this to 1 to cause an overtemp event to occur */ static int -ixgbe_sysctl_phy_overtemp_occurred(SYSCTL_HANDLER_ARGS) -{ - struct adapter *adapter = (struct adapter *) arg1; - struct ixgbe_hw *hw = &adapter->hw; - u16 reg; - - if (hw->device_id != IXGBE_DEV_ID_X550EM_X_10G_T) { - device_printf(adapter->dev, - "Device has no supported external thermal sensor.\n"); - return (ENODEV); - } - - if (hw->phy.ops.read_reg(hw, IXGBE_PHY_OVERTEMP_STATUS, - IXGBE_MDIO_VENDOR_SPECIFIC_1_DEV_TYPE, - ®)) { - device_printf(adapter->dev, - "Error reading from PHY's temperature status register\n"); - return (EAGAIN); - } - - /* Get occurrence bit */ - reg = !!(reg & 0x4000); - return (sysctl_handle_int(oidp, 0, reg, req)); -} - -/* -** Thermal Shutdown Trigger (internal MAC) -** - Set this to 1 to cause an overtemp event to occur -*/ -static int ixgbe_sysctl_thermal_test(SYSCTL_HANDLER_ARGS) { struct adapter *adapter = (struct adapter *) arg1; @@ -4885,6 +4020,7 @@ return (error); if (fire) { + printf("Warning: Thermal Shutdown trigger\n"); u32 reg = IXGBE_READ_REG(hw, IXGBE_EICS); reg |= IXGBE_EICR_TS; IXGBE_WRITE_REG(hw, IXGBE_EICS, reg); @@ -4893,8 +4029,7 @@ return (0); } -/* -** Manage DMA Coalescing. +/* Manage DMA Coalescing. ** Control values: ** 0/1 - off / on (use default value of 1000) ** @@ -4907,22 +4042,31 @@ ixgbe_sysctl_dmac(SYSCTL_HANDLER_ARGS) { struct adapter *adapter = (struct adapter *) arg1; - struct ifnet *ifp = adapter->ifp; + struct ixgbe_hw *hw = &adapter->hw; + struct ifnet *ifp = iflib_get_ifp(adapter->ctx); int error; - u32 newval; + u16 oldval; - newval = adapter->dmac; - error = sysctl_handle_int(oidp, &newval, 0, req); + oldval = adapter->dmac; + error = sysctl_handle_int(oidp, &adapter->dmac, 0, req); if ((error) || (req->newptr == NULL)) return (error); - switch (newval) { + switch (hw->mac.type) { + case ixgbe_mac_X550: + case ixgbe_mac_X550EM_x: + break; + default: + device_printf(iflib_get_dev(adapter->ctx), + "DMA Coalescing is only supported on X550 devices\n"); + return (ENODEV); + } + + switch (adapter->dmac) { case 0: /* Disabled */ - adapter->dmac = 0; break; - case 1: - /* Enable and use default */ + case 1: /* Enable and use default */ adapter->dmac = 1000; break; case 50: @@ -4934,56 +4078,20 @@ case 5000: case 10000: /* Legal values - allow */ - adapter->dmac = newval; break; default: /* Do nothing, illegal value */ + adapter->dmac = oldval; return (EINVAL); } /* Re-initialize hardware if it's already running */ if (ifp->if_drv_flags & IFF_DRV_RUNNING) - ixgbe_init(adapter); + ifp->if_init(ifp); return (0); } -#ifdef IXGBE_DEBUG -/** - * Sysctl to test power states - * Values: - * 0 - set device to D0 - * 3 - set device to D3 - * (none) - get current device power state - */ -static int -ixgbe_sysctl_power_state(SYSCTL_HANDLER_ARGS) -{ - struct adapter *adapter = (struct adapter *) arg1; - device_t dev = adapter->dev; - int curr_ps, new_ps, error = 0; - - curr_ps = new_ps = pci_get_powerstate(dev); - - error = sysctl_handle_int(oidp, &new_ps, 0, req); - if ((error) || (req->newptr == NULL)) - return (error); - - if (new_ps == curr_ps) - return (0); - - if (new_ps == 3 && curr_ps == 0) - error = DEVICE_SUSPEND(dev); - else if (new_ps == 0 && curr_ps == 3) - error = DEVICE_RESUME(dev); - else - return (EINVAL); - - device_printf(dev, "New state: %d\n", pci_get_powerstate(dev)); - - return (error); -} -#endif /* * Sysctl to enable/disable the WoL capability, if supported by the adapter. * Values: @@ -5002,49 +4110,13 @@ error = sysctl_handle_int(oidp, &new_wol_enabled, 0, req); if ((error) || (req->newptr == NULL)) return (error); - new_wol_enabled = !!(new_wol_enabled); if (new_wol_enabled == hw->wol_enabled) return (0); if (new_wol_enabled > 0 && !adapter->wol_support) return (ENODEV); else - hw->wol_enabled = new_wol_enabled; - - return (0); -} - -/* - * Sysctl to enable/disable the Energy Efficient Ethernet capability, - * if supported by the adapter. - * Values: - * 0 - disabled - * 1 - enabled - */ -static int -ixgbe_sysctl_eee_enable(SYSCTL_HANDLER_ARGS) -{ - struct adapter *adapter = (struct adapter *) arg1; - struct ixgbe_hw *hw = &adapter->hw; - struct ifnet *ifp = adapter->ifp; - int new_eee_enabled, error = 0; - - new_eee_enabled = adapter->eee_enabled; - error = sysctl_handle_int(oidp, &new_eee_enabled, 0, req); - if ((error) || (req->newptr == NULL)) - return (error); - new_eee_enabled = !!(new_eee_enabled); - if (new_eee_enabled == adapter->eee_enabled) - return (0); - - if (new_eee_enabled > 0 && !hw->mac.ops.setup_eee) - return (ENODEV); - else - adapter->eee_enabled = new_eee_enabled; - - /* Re-initialize hardware if it's already running */ - if (ifp->if_drv_flags & IFF_DRV_RUNNING) - ixgbe_init(adapter); + hw->wol_enabled = !!(new_wol_enabled); return (0); } @@ -5076,7 +4148,7 @@ bool status; status = !!(IXGBE_READ_REG(hw, IXGBE_EEE_STAT) & - IXGBE_EEE_RX_LPI_STATUS); + IXGBE_EEE_RX_LPI_STATUS); return (sysctl_handle_int(oidp, 0, status, req)); } @@ -5092,27 +4164,12 @@ bool status; status = !!(IXGBE_READ_REG(hw, IXGBE_EEE_STAT) & - IXGBE_EEE_TX_LPI_STATUS); + IXGBE_EEE_TX_LPI_STATUS); return (sysctl_handle_int(oidp, 0, status, req)); } /* - * Read-only sysctl indicating TX Link LPI delay - */ -static int -ixgbe_sysctl_eee_tx_lpi_delay(SYSCTL_HANDLER_ARGS) -{ - struct adapter *adapter = (struct adapter *) arg1; - struct ixgbe_hw *hw = &adapter->hw; - u32 reg; - - reg = IXGBE_READ_REG(hw, IXGBE_EEE_SU); - - return (sysctl_handle_int(oidp, 0, reg >> 26, req)); -} - -/* * Sysctl to enable/disable the types of packets that the * adapter will wake up on upon receipt. * WUFC - Wake Up Filter Control @@ -5155,128 +4212,100 @@ return (0); } -#ifdef IXGBE_DEBUG +/* + * The following two sysctls are for X550 BaseT devices; + * they deal with the external PHY used in them. + */ static int -ixgbe_sysctl_print_rss_config(SYSCTL_HANDLER_ARGS) +ixgbe_sysctl_phy_temp(SYSCTL_HANDLER_ARGS) { - struct adapter *adapter = (struct adapter *)arg1; + struct adapter *adapter = (struct adapter *) arg1; struct ixgbe_hw *hw = &adapter->hw; - device_t dev = adapter->dev; - int error = 0, reta_size; - struct sbuf *buf; - u32 reg; - - buf = sbuf_new_for_sysctl(NULL, NULL, 128, req); - if (!buf) { - device_printf(dev, "Could not allocate sbuf for output.\n"); - return (ENOMEM); - } + u16 reg; - // TODO: use sbufs to make a string to print out - /* Set multiplier for RETA setup and table size based on MAC */ - switch (adapter->hw.mac.type) { - case ixgbe_mac_X550: - case ixgbe_mac_X550EM_x: - reta_size = 128; - break; - default: - reta_size = 32; - break; + if (hw->device_id != IXGBE_DEV_ID_X550EM_X_10G_T) { + device_printf(iflib_get_dev(adapter->ctx), + "Device has no supported external thermal sensor.\n"); + return (ENODEV); } - /* Print out the redirection table */ - sbuf_cat(buf, "\n"); - for (int i = 0; i < reta_size; i++) { - if (i < 32) { - reg = IXGBE_READ_REG(hw, IXGBE_RETA(i)); - sbuf_printf(buf, "RETA(%2d): 0x%08x\n", i, reg); - } else { - reg = IXGBE_READ_REG(hw, IXGBE_ERETA(i - 32)); - sbuf_printf(buf, "ERETA(%2d): 0x%08x\n", i - 32, reg); - } + if (hw->phy.ops.read_reg(hw, IXGBE_PHY_CURRENT_TEMP, + IXGBE_MDIO_VENDOR_SPECIFIC_1_DEV_TYPE, + ®)) { + device_printf(iflib_get_dev(adapter->ctx), + "Error reading from PHY's current temperature register\n"); + return (EAGAIN); } - // TODO: print more config - - error = sbuf_finish(buf); - if (error) - device_printf(dev, "Error finishing sbuf: %d\n", error); + /* Shift temp for output */ + reg = reg >> 8; - sbuf_delete(buf); - return (0); -} -#endif /* IXGBE_DEBUG */ + return (sysctl_handle_int(oidp, NULL, reg, req)); +} /* -** Enable the hardware to drop packets when the buffer is -** full. This is useful when multiqueue,so that no single -** queue being full stalls the entire RX engine. We only -** enable this when Multiqueue AND when Flow Control is -** disabled. -*/ -static void -ixgbe_enable_rx_drop(struct adapter *adapter) + * Reports whether the current PHY temperature is over + * the overtemp threshold. + * - This is reported directly from the PHY + */ +static int +ixgbe_sysctl_phy_overtemp_occurred(SYSCTL_HANDLER_ARGS) { - struct ixgbe_hw *hw = &adapter->hw; + struct adapter *adapter = (struct adapter *) arg1; + struct ixgbe_hw *hw = &adapter->hw; + u16 reg; - for (int i = 0; i < adapter->num_queues; i++) { - struct rx_ring *rxr = &adapter->rx_rings[i]; - u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(rxr->me)); - srrctl |= IXGBE_SRRCTL_DROP_EN; - IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(rxr->me), srrctl); + if (hw->device_id != IXGBE_DEV_ID_X550EM_X_10G_T) { + device_printf(iflib_get_dev(adapter->ctx), + "Device has no supported external thermal sensor.\n"); + return (ENODEV); } -#ifdef PCI_IOV - /* enable drop for each vf */ - for (int i = 0; i < adapter->num_vfs; i++) { - IXGBE_WRITE_REG(hw, IXGBE_QDE, - (IXGBE_QDE_WRITE | (i << IXGBE_QDE_IDX_SHIFT) | - IXGBE_QDE_ENABLE)); + + if (hw->phy.ops.read_reg(hw, IXGBE_PHY_OVERTEMP_STATUS, + IXGBE_MDIO_VENDOR_SPECIFIC_1_DEV_TYPE, + ®)) { + device_printf(iflib_get_dev(adapter->ctx), + "Error reading from PHY's temperature status register\n"); + return (EAGAIN); } -#endif -} -static void -ixgbe_disable_rx_drop(struct adapter *adapter) + /* Get occurrence bit */ + reg = !!(reg & 0x4000); + return (sysctl_handle_int(oidp, 0, reg, req)); +} + +/* + * Sysctl to enable/disable the Energy Efficient Ethernet capability, + * if supported by the adapter. + * Values: + * 0 - disabled + * 1 - enabled + */ +static int +ixgbe_sysctl_eee_enable(SYSCTL_HANDLER_ARGS) { - struct ixgbe_hw *hw = &adapter->hw; + struct adapter *adapter = (struct adapter *) arg1; + struct ixgbe_hw *hw = &adapter->hw; + struct ifnet *ifp = iflib_get_ifp(adapter->ctx); + int new_eee_enabled, error = 0; - for (int i = 0; i < adapter->num_queues; i++) { - struct rx_ring *rxr = &adapter->rx_rings[i]; - u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(rxr->me)); - srrctl &= ~IXGBE_SRRCTL_DROP_EN; - IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(rxr->me), srrctl); - } -#ifdef PCI_IOV - /* disable drop for each vf */ - for (int i = 0; i < adapter->num_vfs; i++) { - IXGBE_WRITE_REG(hw, IXGBE_QDE, - (IXGBE_QDE_WRITE | (i << IXGBE_QDE_IDX_SHIFT))); - } -#endif -} + new_eee_enabled = adapter->eee_enabled; + error = sysctl_handle_int(oidp, &new_eee_enabled, 0, req); + if ((error) || (req->newptr == NULL)) + return (error); + if (new_eee_enabled == adapter->eee_enabled) + return (0); -static void -ixgbe_rearm_queues(struct adapter *adapter, u64 queues) -{ - u32 mask; + if (new_eee_enabled > 0 && !hw->mac.ops.setup_eee) + return (ENODEV); + else + adapter->eee_enabled = !!(new_eee_enabled); - switch (adapter->hw.mac.type) { - case ixgbe_mac_82598EB: - mask = (IXGBE_EIMS_RTX_QUEUE & queues); - IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask); - break; - case ixgbe_mac_82599EB: - case ixgbe_mac_X540: - case ixgbe_mac_X550: - case ixgbe_mac_X550EM_x: - mask = (queues & 0xFFFFFFFF); - IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask); - mask = (queues >> 32); - IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask); - break; - default: - break; - } + /* Re-initialize hardware if it's already running */ + if (ifp->if_drv_flags & IFF_DRV_RUNNING) + ifp->if_init(ifp); + + return (0); } #ifdef PCI_IOV @@ -5300,7 +4329,7 @@ static void ixgbe_vf_set_default_vlan(struct adapter *adapter, struct ixgbe_vf *vf, - uint16_t tag) + uint16_t tag) { struct ixgbe_hw *hw; uint32_t vmolr, vmvir; @@ -5445,7 +4474,7 @@ if (ixgbe_validate_mac_addr(vf->ether_addr) == 0) { ixgbe_set_rar(&adapter->hw, vf->rar_index, - vf->ether_addr, vf->pool, TRUE); + vf->ether_addr, vf->pool, TRUE); ack = IXGBE_VT_MSGTYPE_ACK; } else ack = IXGBE_VT_MSGTYPE_NACK; @@ -5483,7 +4512,7 @@ bcopy(mac, vf->ether_addr, ETHER_ADDR_LEN); ixgbe_set_rar(&adapter->hw, vf->rar_index, vf->ether_addr, - vf->pool, TRUE); + vf->pool, TRUE); ixgbe_send_vf_ack(adapter, vf, msg[0]); } @@ -5511,16 +4540,15 @@ for (int i = 0; i < entries; i++) { vf->mc_hash[i] = list[i]; vec_reg = (vf->mc_hash[i] >> 5) & 0x7F; - vec_bit = vf->mc_hash[i] & 0x1F; - mta_reg = IXGBE_READ_REG(&adapter->hw, IXGBE_MTA(vec_reg)); - mta_reg |= (1 << vec_bit); - IXGBE_WRITE_REG(&adapter->hw, IXGBE_MTA(vec_reg), mta_reg); - } + vec_bit = vf->mc_hash[i] & 0x1F; + mta_reg = IXGBE_READ_REG(&adapter->hw, IXGBE_MTA(vec_reg)); + mta_reg |= (1 << vec_bit); + IXGBE_WRITE_REG(&adapter->hw, IXGBE_MTA(vec_reg), mta_reg); + } vmolr |= IXGBE_VMOLR_ROMPE; IXGBE_WRITE_REG(&adapter->hw, IXGBE_VMOLR(vf->pool), vmolr); ixgbe_send_vf_ack(adapter, vf, msg[0]); - return; } @@ -5598,7 +4626,7 @@ static void ixgbe_vf_set_macvlan(struct adapter *adapter, struct ixgbe_vf *vf, - uint32_t *msg) + uint32_t *msg) { //XXX implement this ixgbe_send_vf_nack(adapter, vf, msg[0]); @@ -5607,7 +4635,7 @@ static void ixgbe_vf_api_negotiate(struct adapter *adapter, struct ixgbe_vf *vf, - uint32_t *msg) + uint32_t *msg) { switch (msg[1]) { @@ -5626,7 +4654,7 @@ static void ixgbe_vf_get_queues(struct adapter *adapter, struct ixgbe_vf *vf, - uint32_t *msg) + uint32_t *msg) { struct ixgbe_hw *hw; uint32_t resp[IXGBE_VF_GET_QUEUES_RESP_LEN]; @@ -5670,7 +4698,7 @@ return; CTR3(KTR_MALLOC, "%s: received msg %x from %d", - adapter->ifp->if_xname, msg[0], vf->pool); + adapter->ifp->if_xname, msg[0], vf->pool); if (msg[0] == IXGBE_VF_RESET) { ixgbe_vf_reset_msg(adapter, vf, msg); return; @@ -5723,7 +4751,6 @@ adapter = context; hw = &adapter->hw; - IXGBE_CORE_LOCK(adapter); for (i = 0; i < adapter->num_vfs; i++) { vf = &adapter->vfs[i]; @@ -5738,9 +4765,8 @@ ixgbe_process_vf_ack(adapter, vf); } } - IXGBE_CORE_UNLOCK(adapter); -} +} static int ixgbe_init_iov(device_t dev, u16 num_vfs, const nvlist_t *config) @@ -5757,20 +4783,16 @@ return (ENOSPC); } - IXGBE_CORE_LOCK(adapter); adapter->vfs = malloc(sizeof(*adapter->vfs) * num_vfs, M_IXGBE, - M_NOWAIT | M_ZERO); + M_NOWAIT | M_ZERO); if (adapter->vfs == NULL) { adapter->num_vfs = 0; - IXGBE_CORE_UNLOCK(adapter); return (ENOMEM); } - - ixgbe_init_locked(adapter); - - IXGBE_CORE_UNLOCK(adapter); + + ixgbe_if_init(adapter->ctx); return (0); } @@ -5786,14 +4808,12 @@ adapter = device_get_softc(dev); hw = &adapter->hw; - IXGBE_CORE_LOCK(adapter); - /* Enable rx/tx for the PF and disable it for all VFs. */ pf_reg = IXGBE_VF_INDEX(adapter->pool); IXGBE_WRITE_REG(hw, IXGBE_VFRE(pf_reg), - IXGBE_VF_BIT(adapter->pool)); + IXGBE_VF_BIT(adapter->pool)); IXGBE_WRITE_REG(hw, IXGBE_VFTE(pf_reg), - IXGBE_VF_BIT(adapter->pool)); + IXGBE_VF_BIT(adapter->pool)); if (pf_reg == 0) vf_reg = 1; @@ -5808,7 +4828,6 @@ adapter->vfs = NULL; adapter->num_vfs = 0; - IXGBE_CORE_UNLOCK(adapter); } @@ -5824,8 +4843,6 @@ if (mode == IXGBE_NO_VM) return; - IXGBE_CORE_LOCK_ASSERT(adapter); - mrqc = IXGBE_READ_REG(hw, IXGBE_MRQC); mrqc &= ~IXGBE_MRQC_MRQE_MASK; @@ -5888,9 +4905,9 @@ /* Enable rx/tx for the PF. */ vf_reg = IXGBE_VF_INDEX(adapter->pool); IXGBE_WRITE_REG(hw, IXGBE_VFRE(vf_reg), - IXGBE_VF_BIT(adapter->pool)); + IXGBE_VF_BIT(adapter->pool)); IXGBE_WRITE_REG(hw, IXGBE_VFTE(vf_reg), - IXGBE_VF_BIT(adapter->pool)); + IXGBE_VF_BIT(adapter->pool)); /* Allow VM-to-VM communication. */ IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN); @@ -5912,8 +4929,6 @@ { struct ixgbe_vf *vf; - IXGBE_CORE_LOCK_ASSERT(adapter); - for (int i = 0; i < adapter->num_vfs; i++) { vf = &adapter->vfs[i]; if (vf->flags & IXGBE_VF_ACTIVE) @@ -5928,8 +4943,6 @@ struct ixgbe_hw *hw; uint32_t vf_index, pfmbimr; - IXGBE_CORE_LOCK_ASSERT(adapter); - hw = &adapter->hw; if (!(vf->flags & IXGBE_VF_ACTIVE)) @@ -5946,7 +4959,7 @@ if (ixgbe_validate_mac_addr(vf->ether_addr) == 0) { ixgbe_set_rar(&adapter->hw, vf->rar_index, - vf->ether_addr, vf->pool, TRUE); + vf->ether_addr, vf->pool, TRUE); } ixgbe_vf_enable_transmit(adapter, vf); @@ -5965,9 +4978,8 @@ adapter = device_get_softc(dev); KASSERT(vfnum < adapter->num_vfs, ("VF index %d is out of range %d", - vfnum, adapter->num_vfs)); + vfnum, adapter->num_vfs)); - IXGBE_CORE_LOCK(adapter); vf = &adapter->vfs[vfnum]; vf->pool= vfnum; @@ -5992,7 +5004,6 @@ vf->flags = IXGBE_VF_ACTIVE; ixgbe_init_vf(adapter, vf); - IXGBE_CORE_UNLOCK(adapter); return (0); } Index: sys/dev/ixgbe/ix_txrx.c =================================================================== --- sys/dev/ixgbe/ix_txrx.c +++ sys/dev/ixgbe/ix_txrx.c @@ -46,671 +46,78 @@ #include #endif -#ifdef DEV_NETMAP -#include -#include -#include - -extern int ix_crcstrip; -#endif - -/* -** HW RSC control: -** this feature only works with -** IPv4, and only on 82599 and later. -** Also this will cause IP forwarding to -** fail and that can't be controlled by -** the stack as LRO can. For all these -** reasons I've deemed it best to leave -** this off and not bother with a tuneable -** interface, this would need to be compiled -** to enable. -*/ -static bool ixgbe_rsc_enable = FALSE; - -#ifdef IXGBE_FDIR -/* -** For Flow Director: this is the -** number of TX packets we sample -** for the filter pool, this means -** every 20th packet will be probed. -** -** This feature can be disabled by -** setting this to 0. -*/ -static int atr_sample_rate = 20; -#endif - /********************************************************************* * Local Function prototypes *********************************************************************/ -static void ixgbe_setup_transmit_ring(struct tx_ring *); -static void ixgbe_free_transmit_buffers(struct tx_ring *); -static int ixgbe_setup_receive_ring(struct rx_ring *); -static void ixgbe_free_receive_buffers(struct rx_ring *); - -static void ixgbe_rx_checksum(u32, struct mbuf *, u32); -static void ixgbe_refresh_mbufs(struct rx_ring *, int); -static int ixgbe_xmit(struct tx_ring *, struct mbuf **); -static int ixgbe_tx_ctx_setup(struct tx_ring *, - struct mbuf *, u32 *, u32 *); -static int ixgbe_tso_setup(struct tx_ring *, - struct mbuf *, u32 *, u32 *); -#ifdef IXGBE_FDIR -static void ixgbe_atr(struct tx_ring *, struct mbuf *); -#endif -static __inline void ixgbe_rx_discard(struct rx_ring *, int); -static __inline void ixgbe_rx_input(struct rx_ring *, struct ifnet *, - struct mbuf *, u32); - -#ifdef IXGBE_LEGACY_TX -/********************************************************************* - * Transmit entry point - * - * ixgbe_start is called by the stack to initiate a transmit. - * The driver will remain in this routine as long as there are - * packets to transmit and transmit resources are available. - * In case resources are not available stack is notified and - * the packet is requeued. - **********************************************************************/ - -void -ixgbe_start_locked(struct tx_ring *txr, struct ifnet * ifp) -{ - struct mbuf *m_head; - struct adapter *adapter = txr->adapter; - - IXGBE_TX_LOCK_ASSERT(txr); - - if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) - return; - if (!adapter->link_active) - return; - - while (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) { - if (txr->tx_avail <= IXGBE_QUEUE_MIN_FREE) - break; - - IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head); - if (m_head == NULL) - break; - - if (ixgbe_xmit(txr, &m_head)) { - if (m_head != NULL) - IFQ_DRV_PREPEND(&ifp->if_snd, m_head); - break; - } - /* Send a copy of the frame to the BPF listener */ - ETHER_BPF_MTAP(ifp, m_head); - } - return; -} - -/* - * Legacy TX start - called by the stack, this - * always uses the first tx ring, and should - * not be used with multiqueue tx enabled. - */ -void -ixgbe_start(struct ifnet *ifp) -{ - struct adapter *adapter = ifp->if_softc; - struct tx_ring *txr = adapter->tx_rings; - - if (ifp->if_drv_flags & IFF_DRV_RUNNING) { - IXGBE_TX_LOCK(txr); - ixgbe_start_locked(txr, ifp); - IXGBE_TX_UNLOCK(txr); - } - return; -} - -#else /* ! IXGBE_LEGACY_TX */ - -/* -** Multiqueue Transmit Entry Point -** (if_transmit function) -*/ -int -ixgbe_mq_start(struct ifnet *ifp, struct mbuf *m) -{ - struct adapter *adapter = ifp->if_softc; - struct ix_queue *que; - struct tx_ring *txr; - int i, err = 0; -#ifdef RSS - uint32_t bucket_id; -#endif - - /* - * When doing RSS, map it to the same outbound queue - * as the incoming flow would be mapped to. - * - * If everything is setup correctly, it should be the - * same bucket that the current CPU we're on is. - */ - if (M_HASHTYPE_GET(m) != M_HASHTYPE_NONE) { -#ifdef RSS - if (rss_hash2bucket(m->m_pkthdr.flowid, - M_HASHTYPE_GET(m), &bucket_id) == 0) { - i = bucket_id % adapter->num_queues; -#ifdef IXGBE_DEBUG - if (bucket_id > adapter->num_queues) - if_printf(ifp, "bucket_id (%d) > num_queues " - "(%d)\n", bucket_id, adapter->num_queues); -#endif - } else -#endif - i = m->m_pkthdr.flowid % adapter->num_queues; - } else - i = curcpu % adapter->num_queues; - - /* Check for a hung queue and pick alternative */ - if (((1 << i) & adapter->active_queues) == 0) - i = ffsl(adapter->active_queues); - - txr = &adapter->tx_rings[i]; - que = &adapter->queues[i]; - - err = drbr_enqueue(ifp, txr->br, m); - if (err) - return (err); - if (IXGBE_TX_TRYLOCK(txr)) { - ixgbe_mq_start_locked(ifp, txr); - IXGBE_TX_UNLOCK(txr); - } else - taskqueue_enqueue(que->tq, &txr->txq_task); - - return (0); -} - -int -ixgbe_mq_start_locked(struct ifnet *ifp, struct tx_ring *txr) -{ - struct adapter *adapter = txr->adapter; - struct mbuf *next; - int enqueued = 0, err = 0; - - if (((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) || - adapter->link_active == 0) - return (ENETDOWN); - - /* Process the queue */ -#if __FreeBSD_version < 901504 - next = drbr_dequeue(ifp, txr->br); - while (next != NULL) { - if ((err = ixgbe_xmit(txr, &next)) != 0) { - if (next != NULL) - err = drbr_enqueue(ifp, txr->br, next); -#else - while ((next = drbr_peek(ifp, txr->br)) != NULL) { - if ((err = ixgbe_xmit(txr, &next)) != 0) { - if (next == NULL) { - drbr_advance(ifp, txr->br); - } else { - drbr_putback(ifp, txr->br, next); - } -#endif - break; - } -#if __FreeBSD_version >= 901504 - drbr_advance(ifp, txr->br); -#endif - enqueued++; -#if 0 // this is VF-only -#if __FreeBSD_version >= 1100036 - /* - * Since we're looking at the tx ring, we can check - * to see if we're a VF by examing our tail register - * address. - */ - if (txr->tail < IXGBE_TDT(0) && next->m_flags & M_MCAST) - if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); -#endif -#endif - /* Send a copy of the frame to the BPF listener */ - ETHER_BPF_MTAP(ifp, next); - if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) - break; -#if __FreeBSD_version < 901504 - next = drbr_dequeue(ifp, txr->br); -#endif - } - - if (txr->tx_avail < IXGBE_TX_CLEANUP_THRESHOLD) - ixgbe_txeof(txr); - - return (err); -} - -/* - * Called from a taskqueue to drain queued transmit packets. - */ -void -ixgbe_deferred_mq_start(void *arg, int pending) -{ - struct tx_ring *txr = arg; - struct adapter *adapter = txr->adapter; - struct ifnet *ifp = adapter->ifp; - - IXGBE_TX_LOCK(txr); - if (!drbr_empty(ifp, txr->br)) - ixgbe_mq_start_locked(ifp, txr); - IXGBE_TX_UNLOCK(txr); -} +static int ixgbe_isc_txd_encap(void *arg, if_pkt_info_t pi); +static void ixgbe_isc_txd_flush(void *arg, uint16_t txqid, uint32_t pidx); +static int ixgbe_isc_txd_credits_update(void *arg, uint16_t txqid, uint32_t cidx); + +static void ixgbe_isc_rxd_refill(void *arg, uint16_t rxqid, uint8_t flid __unused, + uint32_t pidx, uint64_t *paddrs, caddr_t *vaddrs __unused, uint16_t count); +static void ixgbe_isc_rxd_flush(void *arg, uint16_t rxqid, uint8_t flid __unused, uint32_t pidx); +static int ixgbe_isc_rxd_available(void *arg, uint16_t rxqid, uint32_t idx); +static int ixgbe_isc_rxd_pkt_get(void *arg, if_rxd_info_t ri); + +static void ixgbe_rx_checksum(u32 staterr, if_rxd_info_t ri, u32 ptype); + +extern void ixgbe_if_enable_intr(if_ctx_t ctx); +extern int ixgbe_intr(void *arg); + +struct if_txrx ixgbe_txrx = { + ixgbe_isc_txd_encap, + ixgbe_isc_txd_flush, + ixgbe_isc_txd_credits_update, + ixgbe_isc_rxd_available, + ixgbe_isc_rxd_pkt_get, + ixgbe_isc_rxd_refill, + ixgbe_isc_rxd_flush, + ixgbe_intr +}; + +extern if_shared_ctx_t ixgbe_sctx; -/* - * Flush all ring buffers - */ -void -ixgbe_qflush(struct ifnet *ifp) -{ - struct adapter *adapter = ifp->if_softc; - struct tx_ring *txr = adapter->tx_rings; - struct mbuf *m; - - for (int i = 0; i < adapter->num_queues; i++, txr++) { - IXGBE_TX_LOCK(txr); - while ((m = buf_ring_dequeue_sc(txr->br)) != NULL) - m_freem(m); - IXGBE_TX_UNLOCK(txr); - } - if_qflush(ifp); -} -#endif /* IXGBE_LEGACY_TX */ - - -/********************************************************************* +/********************************************************************** * - * This routine maps the mbufs to tx descriptors, allowing the - * TX engine to transmit the packets. - * - return 0 on success, positive on failure + * Setup work for hardware segmentation offload (TSO) on + * adapters using advanced tx descriptors * **********************************************************************/ - static int -ixgbe_xmit(struct tx_ring *txr, struct mbuf **m_headp) +ixgbe_tso_setup(struct ixgbe_adv_tx_context_desc * TXD, if_pkt_info_t pi) { - struct adapter *adapter = txr->adapter; - u32 olinfo_status = 0, cmd_type_len; - int i, j, error, nsegs; - int first; - bool remap = TRUE; - struct mbuf *m_head; - bus_dma_segment_t segs[adapter->num_segs]; - bus_dmamap_t map; - struct ixgbe_tx_buf *txbuf; - union ixgbe_adv_tx_desc *txd = NULL; - - m_head = *m_headp; - - /* Basic descriptor defines */ - cmd_type_len = (IXGBE_ADVTXD_DTYP_DATA | - IXGBE_ADVTXD_DCMD_IFCS | IXGBE_ADVTXD_DCMD_DEXT); - - if (m_head->m_flags & M_VLANTAG) - cmd_type_len |= IXGBE_ADVTXD_DCMD_VLE; - - /* - * Important to capture the first descriptor - * used because it will contain the index of - * the one we tell the hardware to report back - */ - first = txr->next_avail_desc; - txbuf = &txr->tx_buffers[first]; - map = txbuf->map; - - /* - * Map the packet for DMA. - */ -retry: - error = bus_dmamap_load_mbuf_sg(txr->txtag, map, - *m_headp, segs, &nsegs, BUS_DMA_NOWAIT); - - if (__predict_false(error)) { - struct mbuf *m; - - switch (error) { - case EFBIG: - /* Try it again? - one try */ - if (remap == TRUE) { - remap = FALSE; - /* - * XXX: m_defrag will choke on - * non-MCLBYTES-sized clusters - */ - m = m_defrag(*m_headp, M_NOWAIT); - if (m == NULL) { - adapter->mbuf_defrag_failed++; - m_freem(*m_headp); - *m_headp = NULL; - return (ENOBUFS); - } - *m_headp = m; - goto retry; - } else - return (error); - case ENOMEM: - txr->no_tx_dma_setup++; - return (error); - default: - txr->no_tx_dma_setup++; - m_freem(*m_headp); - *m_headp = NULL; - return (error); - } - } - - /* Make certain there are enough descriptors */ - if (nsegs > txr->tx_avail - 2) { - txr->no_desc_avail++; - bus_dmamap_unload(txr->txtag, map); - return (ENOBUFS); - } - m_head = *m_headp; - - /* - * Set up the appropriate offload context - * this will consume the first descriptor - */ - error = ixgbe_tx_ctx_setup(txr, m_head, &cmd_type_len, &olinfo_status); - if (__predict_false(error)) { - if (error == ENOBUFS) - *m_headp = NULL; - return (error); - } - -#ifdef IXGBE_FDIR - /* Do the flow director magic */ - if ((txr->atr_sample) && (!adapter->fdir_reinit)) { - ++txr->atr_count; - if (txr->atr_count >= atr_sample_rate) { - ixgbe_atr(txr, m_head); - txr->atr_count = 0; - } - } -#endif + u32 vlan_macip_lens, type_tucmd_mlhl; + u32 mss_l4len_idx, paylen; + u32 olinfo_status; - olinfo_status |= IXGBE_ADVTXD_CC; - i = txr->next_avail_desc; - for (j = 0; j < nsegs; j++) { - bus_size_t seglen; - bus_addr_t segaddr; - - txbuf = &txr->tx_buffers[i]; - txd = &txr->tx_base[i]; - seglen = segs[j].ds_len; - segaddr = htole64(segs[j].ds_addr); - - txd->read.buffer_addr = segaddr; - txd->read.cmd_type_len = htole32(txr->txd_cmd | - cmd_type_len |seglen); - txd->read.olinfo_status = htole32(olinfo_status); - - if (++i == txr->num_desc) - i = 0; - } - - txd->read.cmd_type_len |= - htole32(IXGBE_TXD_CMD_EOP | IXGBE_TXD_CMD_RS); - txr->tx_avail -= nsegs; - txr->next_avail_desc = i; - - txbuf->m_head = m_head; - /* - * Here we swap the map so the last descriptor, - * which gets the completion interrupt has the - * real map, and the first descriptor gets the - * unused map from this descriptor. - */ - txr->tx_buffers[first].map = txbuf->map; - txbuf->map = map; - bus_dmamap_sync(txr->txtag, map, BUS_DMASYNC_PREWRITE); - - /* Set the EOP descriptor that will be marked done */ - txbuf = &txr->tx_buffers[first]; - txbuf->eop = txd; - - bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, - BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); - /* - * Advance the Transmit Descriptor Tail (Tdt), this tells the - * hardware that this frame is available to transmit. - */ - ++txr->total_packets; - IXGBE_WRITE_REG(&adapter->hw, txr->tail, i); - - /* Mark queue as having work */ - if (txr->busy == 0) - txr->busy = 1; - - return (0); -} - - -/********************************************************************* - * - * Allocate memory for tx_buffer structures. The tx_buffer stores all - * the information needed to transmit a packet on the wire. This is - * called only once at attach, setup is done every reset. - * - **********************************************************************/ -int -ixgbe_allocate_transmit_buffers(struct tx_ring *txr) -{ - struct adapter *adapter = txr->adapter; - device_t dev = adapter->dev; - struct ixgbe_tx_buf *txbuf; - int error, i; - - /* - * Setup DMA descriptor areas. - */ - if ((error = bus_dma_tag_create( - bus_get_dma_tag(adapter->dev), /* parent */ - 1, 0, /* alignment, bounds */ - BUS_SPACE_MAXADDR, /* lowaddr */ - BUS_SPACE_MAXADDR, /* highaddr */ - NULL, NULL, /* filter, filterarg */ - IXGBE_TSO_SIZE, /* maxsize */ - adapter->num_segs, /* nsegments */ - PAGE_SIZE, /* maxsegsize */ - 0, /* flags */ - NULL, /* lockfunc */ - NULL, /* lockfuncarg */ - &txr->txtag))) { - device_printf(dev,"Unable to allocate TX DMA tag\n"); - goto fail; - } - - if (!(txr->tx_buffers = - (struct ixgbe_tx_buf *) malloc(sizeof(struct ixgbe_tx_buf) * - adapter->num_tx_desc, M_DEVBUF, M_NOWAIT | M_ZERO))) { - device_printf(dev, "Unable to allocate tx_buffer memory\n"); - error = ENOMEM; - goto fail; - } - - /* Create the descriptor buffer dma maps */ - txbuf = txr->tx_buffers; - for (i = 0; i < adapter->num_tx_desc; i++, txbuf++) { - error = bus_dmamap_create(txr->txtag, 0, &txbuf->map); - if (error != 0) { - device_printf(dev, "Unable to create TX DMA map\n"); - goto fail; - } - } - - return 0; -fail: - /* We free all, it handles case where we are in the middle */ - ixgbe_free_transmit_structures(adapter); - return (error); -} - -/********************************************************************* - * - * Initialize a transmit ring. - * - **********************************************************************/ -static void -ixgbe_setup_transmit_ring(struct tx_ring *txr) -{ - struct adapter *adapter = txr->adapter; - struct ixgbe_tx_buf *txbuf; -#ifdef DEV_NETMAP - struct netmap_adapter *na = NA(adapter->ifp); - struct netmap_slot *slot; -#endif /* DEV_NETMAP */ - - /* Clear the old ring contents */ - IXGBE_TX_LOCK(txr); -#ifdef DEV_NETMAP - /* - * (under lock): if in netmap mode, do some consistency - * checks and set slot to entry 0 of the netmap ring. - */ - slot = netmap_reset(na, NR_TX, txr->me, 0); -#endif /* DEV_NETMAP */ - bzero((void *)txr->tx_base, - (sizeof(union ixgbe_adv_tx_desc)) * adapter->num_tx_desc); - /* Reset indices */ - txr->next_avail_desc = 0; - txr->next_to_clean = 0; - - /* Free any existing tx buffers. */ - txbuf = txr->tx_buffers; - for (int i = 0; i < txr->num_desc; i++, txbuf++) { - if (txbuf->m_head != NULL) { - bus_dmamap_sync(txr->txtag, txbuf->map, - BUS_DMASYNC_POSTWRITE); - bus_dmamap_unload(txr->txtag, txbuf->map); - m_freem(txbuf->m_head); - txbuf->m_head = NULL; - } -#ifdef DEV_NETMAP - /* - * In netmap mode, set the map for the packet buffer. - * NOTE: Some drivers (not this one) also need to set - * the physical buffer address in the NIC ring. - * Slots in the netmap ring (indexed by "si") are - * kring->nkr_hwofs positions "ahead" wrt the - * corresponding slot in the NIC ring. In some drivers - * (not here) nkr_hwofs can be negative. Function - * netmap_idx_n2k() handles wraparounds properly. - */ - if (slot) { - int si = netmap_idx_n2k(&na->tx_rings[txr->me], i); - netmap_load_map(na, txr->txtag, - txbuf->map, NMB(na, slot + si)); - } -#endif /* DEV_NETMAP */ - /* Clear the EOP descriptor pointer */ - txbuf->eop = NULL; - } - -#ifdef IXGBE_FDIR - /* Set the rate at which we sample packets */ - if (adapter->hw.mac.type != ixgbe_mac_82598EB) - txr->atr_sample = atr_sample_rate; -#endif + mss_l4len_idx = vlan_macip_lens = type_tucmd_mlhl = olinfo_status = 0; + if (pi->ipi_flags & IPI_TX_IPV4) + type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV4; + else if (pi->ipi_flags & IPI_TX_IPV6) + type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV6; + else + panic("CSUM_TSO but no supported IP version"); - /* Set number of descriptors available */ - txr->tx_avail = adapter->num_tx_desc; - bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, - BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); - IXGBE_TX_UNLOCK(txr); -} + /* This is used in the transmit desc in encap */ + paylen = pi->ipi_len - pi->ipi_ehdrlen - pi->ipi_ip_hlen - pi->ipi_tcp_hlen; -/********************************************************************* - * - * Initialize all transmit rings. - * - **********************************************************************/ -int -ixgbe_setup_transmit_structures(struct adapter *adapter) -{ - struct tx_ring *txr = adapter->tx_rings; + /* ADV DTYPE TUCMD */ + type_tucmd_mlhl |= IXGBE_ADVTXD_DCMD_DEXT | IXGBE_ADVTXD_DTYP_CTXT; + type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_TCP; + TXD->type_tucmd_mlhl = htole32(type_tucmd_mlhl); - for (int i = 0; i < adapter->num_queues; i++, txr++) - ixgbe_setup_transmit_ring(txr); + /* MSS L4LEN IDX */ + mss_l4len_idx |= (pi->ipi_tso_segsz << IXGBE_ADVTXD_MSS_SHIFT); + mss_l4len_idx |= (pi->ipi_tcp_hlen << IXGBE_ADVTXD_L4LEN_SHIFT); + TXD->mss_l4len_idx = htole32(mss_l4len_idx); - return (0); -} + TXD->seqnum_seed = htole32(0); -/********************************************************************* - * - * Free all transmit rings. - * - **********************************************************************/ -void -ixgbe_free_transmit_structures(struct adapter *adapter) -{ - struct tx_ring *txr = adapter->tx_rings; - - for (int i = 0; i < adapter->num_queues; i++, txr++) { - IXGBE_TX_LOCK(txr); - ixgbe_free_transmit_buffers(txr); - ixgbe_dma_free(adapter, &txr->txdma); - IXGBE_TX_UNLOCK(txr); - IXGBE_TX_LOCK_DESTROY(txr); - } - free(adapter->tx_rings, M_DEVBUF); -} + olinfo_status |= IXGBE_TXD_POPTS_TXSM << 8; + olinfo_status |= paylen << IXGBE_ADVTXD_PAYLEN_SHIFT; -/********************************************************************* - * - * Free transmit ring related data structures. - * - **********************************************************************/ -static void -ixgbe_free_transmit_buffers(struct tx_ring *txr) -{ - struct adapter *adapter = txr->adapter; - struct ixgbe_tx_buf *tx_buffer; - int i; - - INIT_DEBUGOUT("ixgbe_free_transmit_ring: begin"); - - if (txr->tx_buffers == NULL) - return; - - tx_buffer = txr->tx_buffers; - for (i = 0; i < adapter->num_tx_desc; i++, tx_buffer++) { - if (tx_buffer->m_head != NULL) { - bus_dmamap_sync(txr->txtag, tx_buffer->map, - BUS_DMASYNC_POSTWRITE); - bus_dmamap_unload(txr->txtag, - tx_buffer->map); - m_freem(tx_buffer->m_head); - tx_buffer->m_head = NULL; - if (tx_buffer->map != NULL) { - bus_dmamap_destroy(txr->txtag, - tx_buffer->map); - tx_buffer->map = NULL; - } - } else if (tx_buffer->map != NULL) { - bus_dmamap_unload(txr->txtag, - tx_buffer->map); - bus_dmamap_destroy(txr->txtag, - tx_buffer->map); - tx_buffer->map = NULL; - } - } -#ifdef IXGBE_LEGACY_TX - if (txr->br != NULL) - buf_ring_free(txr->br, M_DEVBUF); -#endif - if (txr->tx_buffers != NULL) { - free(txr->tx_buffers, M_DEVBUF); - txr->tx_buffers = NULL; - } - if (txr->txtag != NULL) { - bus_dma_tag_destroy(txr->txtag); - txr->txtag = NULL; - } - return; + return (olinfo_status); } /********************************************************************* @@ -718,140 +125,52 @@ * Advanced Context Descriptor setup for VLAN, CSUM or TSO * **********************************************************************/ - static int -ixgbe_tx_ctx_setup(struct tx_ring *txr, struct mbuf *mp, - u32 *cmd_type_len, u32 *olinfo_status) +ixgbe_tx_ctx_setup(struct ixgbe_adv_tx_context_desc *TXD, if_pkt_info_t pi) { - struct adapter *adapter = txr->adapter; - struct ixgbe_adv_tx_context_desc *TXD; - struct ether_vlan_header *eh; -#ifdef INET - struct ip *ip; -#endif -#ifdef INET6 - struct ip6_hdr *ip6; -#endif - u32 vlan_macip_lens = 0, type_tucmd_mlhl = 0; - int ehdrlen, ip_hlen = 0; - u16 etype; - u8 ipproto = 0; - int offload = TRUE; - int ctxd = txr->next_avail_desc; - u16 vtag = 0; - caddr_t l3d; + u32 vlan_macip_lens, type_tucmd_mlhl; + u32 olinfo_status; + vlan_macip_lens = type_tucmd_mlhl = 0; + /* VLAN MACLEN IPLEN */ + vlan_macip_lens |= (htole16(pi->ipi_vtag) << IXGBE_ADVTXD_VLAN_SHIFT); + vlan_macip_lens |= pi->ipi_ehdrlen << IXGBE_ADVTXD_MACLEN_SHIFT; + vlan_macip_lens |= pi->ipi_ip_hlen; + TXD->vlan_macip_lens = htole32(vlan_macip_lens); /* First check if TSO is to be used */ - if (mp->m_pkthdr.csum_flags & (CSUM_IP_TSO|CSUM_IP6_TSO)) - return (ixgbe_tso_setup(txr, mp, cmd_type_len, olinfo_status)); + if (pi->ipi_csum_flags & CSUM_TSO) + return (ixgbe_tso_setup(TXD, pi)); - if ((mp->m_pkthdr.csum_flags & CSUM_OFFLOAD) == 0) - offload = FALSE; - - /* Indicate the whole packet as payload when not doing TSO */ - *olinfo_status |= mp->m_pkthdr.len << IXGBE_ADVTXD_PAYLEN_SHIFT; - - /* Now ready a context descriptor */ - TXD = (struct ixgbe_adv_tx_context_desc *) &txr->tx_base[ctxd]; - - /* - ** In advanced descriptors the vlan tag must - ** be placed into the context descriptor. Hence - ** we need to make one even if not doing offloads. - */ - if (mp->m_flags & M_VLANTAG) { - vtag = htole16(mp->m_pkthdr.ether_vtag); - vlan_macip_lens |= (vtag << IXGBE_ADVTXD_VLAN_SHIFT); - } else if (!IXGBE_IS_X550VF(adapter) && (offload == FALSE)) - return (0); - - /* - * Determine where frame payload starts. - * Jump over vlan headers if already present, - * helpful for QinQ too. - */ - eh = mtod(mp, struct ether_vlan_header *); - if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) { - etype = ntohs(eh->evl_proto); - ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN; - } else { - etype = ntohs(eh->evl_encap_proto); - ehdrlen = ETHER_HDR_LEN; - } - - /* Set the ether header length */ - vlan_macip_lens |= ehdrlen << IXGBE_ADVTXD_MACLEN_SHIFT; - - if (offload == FALSE) + olinfo_status = pi->ipi_len << IXGBE_ADVTXD_PAYLEN_SHIFT; + if (pi->ipi_flags & IPI_TX_IPV4) + type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV4; + else if (pi->ipi_flags & IPI_TX_IPV6) + type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV6; + else goto no_offloads; - /* - * If the first mbuf only includes the ethernet header, jump to the next one - * XXX: This assumes the stack splits mbufs containing headers on header boundaries - * XXX: And assumes the entire IP header is contained in one mbuf - */ - if (mp->m_len == ehdrlen && mp->m_next) - l3d = mtod(mp->m_next, caddr_t); - else - l3d = mtod(mp, caddr_t) + ehdrlen; - - switch (etype) { -#ifdef INET - case ETHERTYPE_IP: - ip = (struct ip *)(l3d); - ip_hlen = ip->ip_hl << 2; - ipproto = ip->ip_p; - type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV4; - /* Insert IPv4 checksum into data descriptors */ - if (mp->m_pkthdr.csum_flags & CSUM_IP) { - ip->ip_sum = 0; - *olinfo_status |= IXGBE_TXD_POPTS_IXSM << 8; - } - break; -#endif -#ifdef INET6 - case ETHERTYPE_IPV6: - ip6 = (struct ip6_hdr *)(l3d); - ip_hlen = sizeof(struct ip6_hdr); - ipproto = ip6->ip6_nxt; - type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV6; - break; + switch (pi->ipi_ipproto) { + case IPPROTO_TCP: + if (pi->ipi_csum_flags & CSUM_TCP) + type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_TCP; + break; + case IPPROTO_UDP: + if (pi->ipi_csum_flags & CSUM_UDP) + type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_UDP; + break; +#if __FreeBSD_version >= 800000 + case IPPROTO_SCTP: + if (pi->ipi_csum_flags & CSUM_SCTP) + type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_SCTP; + break; #endif - default: - offload = FALSE; - break; - } - - vlan_macip_lens |= ip_hlen; - - /* No support for offloads for non-L4 next headers */ - switch (ipproto) { - case IPPROTO_TCP: - if (mp->m_pkthdr.csum_flags & (CSUM_IP_TCP | CSUM_IP6_TCP)) - type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_TCP; - else - offload = false; - break; - case IPPROTO_UDP: - if (mp->m_pkthdr.csum_flags & (CSUM_IP_UDP | CSUM_IP6_UDP)) - type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_UDP; - else - offload = false; - break; - case IPPROTO_SCTP: - if (mp->m_pkthdr.csum_flags & (CSUM_IP_SCTP | CSUM_IP6_SCTP)) - type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_SCTP; - else - offload = false; - break; - default: - offload = false; - break; + default: + goto no_offloads; + break; } - if (offload) /* Insert L4 checksum into data descriptors */ - *olinfo_status |= IXGBE_TXD_POPTS_TXSM << 8; + olinfo_status |= IXGBE_TXD_POPTS_TXSM << 8; no_offloads: type_tucmd_mlhl |= IXGBE_ADVTXD_DCMD_DEXT | IXGBE_ADVTXD_DTYP_CTXT; @@ -862,1164 +181,293 @@ TXD->seqnum_seed = htole32(0); TXD->mss_l4len_idx = htole32(0); - /* We've consumed the first desc, adjust counters */ - if (++ctxd == txr->num_desc) - ctxd = 0; - txr->next_avail_desc = ctxd; - --txr->tx_avail; - - return (0); + return (olinfo_status); } -/********************************************************************** - * - * Setup work for hardware segmentation offload (TSO) on - * adapters using advanced tx descriptors - * - **********************************************************************/ static int -ixgbe_tso_setup(struct tx_ring *txr, struct mbuf *mp, - u32 *cmd_type_len, u32 *olinfo_status) -{ +ixgbe_isc_txd_encap(void *arg, if_pkt_info_t pi) +{ + struct adapter *sc = arg; + struct ix_queue *que = &sc->queues[pi->ipi_qsidx]; + struct tx_ring *txr = &que->txr; + struct ixgbe_tx_buf *buf; + int nsegs = pi->ipi_nsegs; + bus_dma_segment_t *segs = pi->ipi_segs; + union ixgbe_adv_tx_desc *txd = NULL; struct ixgbe_adv_tx_context_desc *TXD; - u32 vlan_macip_lens = 0, type_tucmd_mlhl = 0; - u32 mss_l4len_idx = 0, paylen; - u16 vtag = 0, eh_type; - int ctxd, ehdrlen, ip_hlen, tcp_hlen; - struct ether_vlan_header *eh; -#ifdef INET6 - struct ip6_hdr *ip6; -#endif -#ifdef INET - struct ip *ip; -#endif - struct tcphdr *th; - - /* - * Determine where frame payload starts. - * Jump over vlan headers if already present - */ - eh = mtod(mp, struct ether_vlan_header *); - if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) { - ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN; - eh_type = eh->evl_proto; - } else { - ehdrlen = ETHER_HDR_LEN; - eh_type = eh->evl_encap_proto; - } + int i, j, first; + u32 olinfo_status, cmd, flags = 0; + + cmd = (IXGBE_ADVTXD_DTYP_DATA | + IXGBE_ADVTXD_DCMD_IFCS | IXGBE_ADVTXD_DCMD_DEXT); + olinfo_status = 0; + if (pi->ipi_mflags & M_VLANTAG) + cmd |= IXGBE_ADVTXD_DCMD_VLE; + + i = first = pi->ipi_pidx; + if (pi->ipi_flags & IPI_TX_INTR) + flags = IXGBE_TXD_CMD_RS; - switch (ntohs(eh_type)) { -#ifdef INET6 - case ETHERTYPE_IPV6: - ip6 = (struct ip6_hdr *)(mp->m_data + ehdrlen); - /* XXX-BZ For now we do not pretend to support ext. hdrs. */ - if (ip6->ip6_nxt != IPPROTO_TCP) - return (ENXIO); - ip_hlen = sizeof(struct ip6_hdr); - ip6 = (struct ip6_hdr *)(mp->m_data + ehdrlen); - th = (struct tcphdr *)((caddr_t)ip6 + ip_hlen); - th->th_sum = in6_cksum_pseudo(ip6, 0, IPPROTO_TCP, 0); - type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV6; - break; -#endif -#ifdef INET - case ETHERTYPE_IP: - ip = (struct ip *)(mp->m_data + ehdrlen); - if (ip->ip_p != IPPROTO_TCP) - return (ENXIO); - ip->ip_sum = 0; - ip_hlen = ip->ip_hl << 2; - th = (struct tcphdr *)((caddr_t)ip + ip_hlen); - th->th_sum = in_pseudo(ip->ip_src.s_addr, - ip->ip_dst.s_addr, htons(IPPROTO_TCP)); - type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV4; - /* Tell transmit desc to also do IPv4 checksum. */ - *olinfo_status |= IXGBE_TXD_POPTS_IXSM << 8; - break; -#endif - default: - panic("%s: CSUM_TSO but no supported IP version (0x%04x)", - __func__, ntohs(eh_type)); - break; + /* Indicate the whole packet as payload when not doing TSO */ + TXD = (struct ixgbe_adv_tx_context_desc *) &txr->tx_base[first]; + + olinfo_status |= pi->ipi_len << IXGBE_ADVTXD_PAYLEN_SHIFT; + if (pi->ipi_csum_flags & CSUM_OFFLOAD || IXGBE_IS_X550VF(sc)) { + /********************************************* + * Set up the appropriate offload context + * this will consume the first descriptor + *********************************************/ + olinfo_status = ixgbe_tx_ctx_setup(TXD, pi); + if (pi->ipi_csum_flags & CSUM_TSO) { + cmd |= IXGBE_ADVTXD_DCMD_TSE; + ++txr->tso_tx; + } + ++i; } - ctxd = txr->next_avail_desc; - TXD = (struct ixgbe_adv_tx_context_desc *) &txr->tx_base[ctxd]; + for (j = 0; j < nsegs; j++) { + bus_size_t seglen; + bus_addr_t segaddr; - tcp_hlen = th->th_off << 2; + txd = &txr->tx_base[i]; + buf = &txr->tx_buffers[i]; + seglen = segs[j].ds_len; + segaddr = htole64(segs[j].ds_addr); - /* This is used in the transmit desc in encap */ - paylen = mp->m_pkthdr.len - ehdrlen - ip_hlen - tcp_hlen; + txd->read.buffer_addr = segaddr; + txd->read.cmd_type_len = htole32(flags | cmd | seglen); + txd->read.olinfo_status = htole32(olinfo_status); - /* VLAN MACLEN IPLEN */ - if (mp->m_flags & M_VLANTAG) { - vtag = htole16(mp->m_pkthdr.ether_vtag); - vlan_macip_lens |= (vtag << IXGBE_ADVTXD_VLAN_SHIFT); + if (++i == ixgbe_sctx->isc_ntxd) { + i = 0; + } } - vlan_macip_lens |= ehdrlen << IXGBE_ADVTXD_MACLEN_SHIFT; - vlan_macip_lens |= ip_hlen; - TXD->vlan_macip_lens = htole32(vlan_macip_lens); - - /* ADV DTYPE TUCMD */ - type_tucmd_mlhl |= IXGBE_ADVTXD_DCMD_DEXT | IXGBE_ADVTXD_DTYP_CTXT; - type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_TCP; - TXD->type_tucmd_mlhl = htole32(type_tucmd_mlhl); + txd->read.cmd_type_len |= + htole32(IXGBE_TXD_CMD_EOP | IXGBE_TXD_CMD_RS); - /* MSS L4LEN IDX */ - mss_l4len_idx |= (mp->m_pkthdr.tso_segsz << IXGBE_ADVTXD_MSS_SHIFT); - mss_l4len_idx |= (tcp_hlen << IXGBE_ADVTXD_L4LEN_SHIFT); - TXD->mss_l4len_idx = htole32(mss_l4len_idx); + /* Set the EOP descriptor that will be marked done */ + buf = &txr->tx_buffers[first]; + buf->eop = txd; - TXD->seqnum_seed = htole32(0); + txr->bytes += pi->ipi_len; + pi->ipi_new_pidx = i; - if (++ctxd == txr->num_desc) - ctxd = 0; - - txr->tx_avail--; - txr->next_avail_desc = ctxd; - *cmd_type_len |= IXGBE_ADVTXD_DCMD_TSE; - *olinfo_status |= IXGBE_TXD_POPTS_TXSM << 8; - *olinfo_status |= paylen << IXGBE_ADVTXD_PAYLEN_SHIFT; - ++txr->tso_tx; + ++txr->total_packets; + return (0); } + +static void +ixgbe_isc_txd_flush(void *arg, uint16_t txqid, uint32_t pidx) +{ + struct adapter *sc = arg; + struct ix_queue *que = &sc->queues[txqid]; + struct tx_ring *txr = &que->txr; + + IXGBE_WRITE_REG(&sc->hw, txr->tail, pidx); +} - -/********************************************************************** - * - * Examine each tx_buffer in the used queue. If the hardware is done - * processing the packet then free associated resources. The - * tx_buffer is put back on the free queue. - * - **********************************************************************/ -void -ixgbe_txeof(struct tx_ring *txr) +static int +ixgbe_isc_txd_credits_update(void *arg, uint16_t txqid, uint32_t cidx_init) { - struct adapter *adapter = txr->adapter; -#ifdef DEV_NETMAP - struct ifnet *ifp = adapter->ifp; -#endif - u32 work, processed = 0; - u32 limit = adapter->tx_process_limit; + struct adapter *sc = arg; + struct ix_queue *que = &sc->queues[txqid]; + struct tx_ring *txr = &que->txr; + + u32 cidx, processed = 0; + u32 limit = sc->tx_process_limit; + struct ixgbe_tx_buf *buf; union ixgbe_adv_tx_desc *txd; - mtx_assert(&txr->tx_mtx, MA_OWNED); - -#ifdef DEV_NETMAP - if (ifp->if_capenable & IFCAP_NETMAP) { - struct netmap_adapter *na = NA(ifp); - struct netmap_kring *kring = &na->tx_rings[txr->me]; - txd = txr->tx_base; - bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, - BUS_DMASYNC_POSTREAD); - /* - * In netmap mode, all the work is done in the context - * of the client thread. Interrupt handlers only wake up - * clients, which may be sleeping on individual rings - * or on a global resource for all rings. - * To implement tx interrupt mitigation, we wake up the client - * thread roughly every half ring, even if the NIC interrupts - * more frequently. This is implemented as follows: - * - ixgbe_txsync() sets kring->nr_kflags with the index of - * the slot that should wake up the thread (nkr_num_slots - * means the user thread should not be woken up); - * - the driver ignores tx interrupts unless netmap_mitigate=0 - * or the slot has the DD bit set. - */ - if (!netmap_mitigate || - (kring->nr_kflags < kring->nkr_num_slots && - txd[kring->nr_kflags].wb.status & IXGBE_TXD_STAT_DD)) { - netmap_tx_irq(ifp, txr->me); - } - return; - } -#endif /* DEV_NETMAP */ + cidx = cidx_init; - if (txr->tx_avail == txr->num_desc) { - txr->busy = 0; - return; - } - - /* Get work starting point */ - work = txr->next_to_clean; - buf = &txr->tx_buffers[work]; - txd = &txr->tx_base[work]; - work -= txr->num_desc; /* The distance to ring end */ - bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, - BUS_DMASYNC_POSTREAD); + buf = &txr->tx_buffers[cidx]; + txd = &txr->tx_base[cidx]; do { + union ixgbe_adv_tx_desc *eop = buf->eop; - if (eop == NULL) /* No work */ + if (eop == NULL) { /* No work */ break; - - if ((eop->wb.status & IXGBE_TXD_STAT_DD) == 0) + } + + if ((eop->wb.status & IXGBE_TXD_STAT_DD) == 0) { break; /* I/O not complete */ - - if (buf->m_head) { - txr->bytes += - buf->m_head->m_pkthdr.len; - bus_dmamap_sync(txr->txtag, - buf->map, - BUS_DMASYNC_POSTWRITE); - bus_dmamap_unload(txr->txtag, - buf->map); - m_freem(buf->m_head); - buf->m_head = NULL; } - buf->eop = NULL; - ++txr->tx_avail; - + + buf->eop = NULL; /* clear indicate processed */ + /* We clean the range if multi segment */ while (txd != eop) { ++txd; ++buf; - ++work; + cidx++; + processed++; /* wrap the ring? */ - if (__predict_false(!work)) { - work -= txr->num_desc; + if (cidx == ixgbe_sctx->isc_ntxd) { buf = txr->tx_buffers; txd = txr->tx_base; + cidx = 0; } - if (buf->m_head) { - txr->bytes += - buf->m_head->m_pkthdr.len; - bus_dmamap_sync(txr->txtag, - buf->map, - BUS_DMASYNC_POSTWRITE); - bus_dmamap_unload(txr->txtag, - buf->map); - m_freem(buf->m_head); - buf->m_head = NULL; - } - ++txr->tx_avail; buf->eop = NULL; - } ++txr->packets; ++processed; - + /* Try the next packet */ - ++txd; - ++buf; - ++work; + txd++; + buf++; + cidx++; /* reset with a wrap */ - if (__predict_false(!work)) { - work -= txr->num_desc; + if (__predict_false(cidx == ixgbe_sctx->isc_ntxd)) { + cidx = 0; buf = txr->tx_buffers; txd = txr->tx_base; } prefetch(txd); - } while (__predict_true(--limit)); - - bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, - BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); - - work += txr->num_desc; - txr->next_to_clean = work; - - /* - ** Queue Hang detection, we know there's - ** work outstanding or the first return - ** would have been taken, so increment busy - ** if nothing managed to get cleaned, then - ** in local_timer it will be checked and - ** marked as HUNG if it exceeds a MAX attempt. - */ - if ((processed == 0) && (txr->busy != IXGBE_QUEUE_HUNG)) - ++txr->busy; - /* - ** If anything gets cleaned we reset state to 1, - ** note this will turn off HUNG if its set. - */ - if (processed) - txr->busy = 1; - - if (txr->tx_avail == txr->num_desc) - txr->busy = 0; - - return; -} - - -#ifdef IXGBE_FDIR -/* -** This routine parses packet headers so that Flow -** Director can make a hashed filter table entry -** allowing traffic flows to be identified and kept -** on the same cpu. This would be a performance -** hit, but we only do it at IXGBE_FDIR_RATE of -** packets. -*/ -static void -ixgbe_atr(struct tx_ring *txr, struct mbuf *mp) -{ - struct adapter *adapter = txr->adapter; - struct ix_queue *que; - struct ip *ip; - struct tcphdr *th; - struct udphdr *uh; - struct ether_vlan_header *eh; - union ixgbe_atr_hash_dword input = {.dword = 0}; - union ixgbe_atr_hash_dword common = {.dword = 0}; - int ehdrlen, ip_hlen; - u16 etype; - - eh = mtod(mp, struct ether_vlan_header *); - if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) { - ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN; - etype = eh->evl_proto; - } else { - ehdrlen = ETHER_HDR_LEN; - etype = eh->evl_encap_proto; - } - - /* Only handling IPv4 */ - if (etype != htons(ETHERTYPE_IP)) - return; - - ip = (struct ip *)(mp->m_data + ehdrlen); - ip_hlen = ip->ip_hl << 2; + } while (__predict_true(--limit) && cidx != cidx_init); - /* check if we're UDP or TCP */ - switch (ip->ip_p) { - case IPPROTO_TCP: - th = (struct tcphdr *)((caddr_t)ip + ip_hlen); - /* src and dst are inverted */ - common.port.dst ^= th->th_sport; - common.port.src ^= th->th_dport; - input.formatted.flow_type ^= IXGBE_ATR_FLOW_TYPE_TCPV4; - break; - case IPPROTO_UDP: - uh = (struct udphdr *)((caddr_t)ip + ip_hlen); - /* src and dst are inverted */ - common.port.dst ^= uh->uh_sport; - common.port.src ^= uh->uh_dport; - input.formatted.flow_type ^= IXGBE_ATR_FLOW_TYPE_UDPV4; - break; - default: - return; - } - - input.formatted.vlan_id = htobe16(mp->m_pkthdr.ether_vtag); - if (mp->m_pkthdr.ether_vtag) - common.flex_bytes ^= htons(ETHERTYPE_VLAN); - else - common.flex_bytes ^= etype; - common.ip ^= ip->ip_src.s_addr ^ ip->ip_dst.s_addr; - - que = &adapter->queues[txr->me]; - /* - ** This assumes the Rx queue and Tx - ** queue are bound to the same CPU - */ - ixgbe_fdir_add_signature_filter_82599(&adapter->hw, - input, common, que->msix); -} -#endif /* IXGBE_FDIR */ - -/* -** Used to detect a descriptor that has -** been merged by Hardware RSC. -*/ -static inline u32 -ixgbe_rsc_count(union ixgbe_adv_rx_desc *rx) -{ - return (le32toh(rx->wb.lower.lo_dword.data) & - IXGBE_RXDADV_RSCCNT_MASK) >> IXGBE_RXDADV_RSCCNT_SHIFT; + return (processed); } -/********************************************************************* - * - * Initialize Hardware RSC (LRO) feature on 82599 - * for an RX ring, this is toggled by the LRO capability - * even though it is transparent to the stack. - * - * NOTE: since this HW feature only works with IPV4 and - * our testing has shown soft LRO to be as effective - * I have decided to disable this by default. - * - **********************************************************************/ -static void -ixgbe_setup_hw_rsc(struct rx_ring *rxr) +static void ixgbe_isc_rxd_refill(void *arg, uint16_t rxqid, uint8_t flid __unused, + uint32_t pidx, uint64_t *paddrs, caddr_t *vaddrs __unused, uint16_t count) { - struct adapter *adapter = rxr->adapter; - struct ixgbe_hw *hw = &adapter->hw; - u32 rscctrl, rdrxctl; - - /* If turning LRO/RSC off we need to disable it */ - if ((adapter->ifp->if_capenable & IFCAP_LRO) == 0) { - rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(rxr->me)); - rscctrl &= ~IXGBE_RSCCTL_RSCEN; - return; - } - - rdrxctl = IXGBE_READ_REG(hw, IXGBE_RDRXCTL); - rdrxctl &= ~IXGBE_RDRXCTL_RSCFRSTSIZE; -#ifdef DEV_NETMAP /* crcstrip is optional in netmap */ - if (adapter->ifp->if_capenable & IFCAP_NETMAP && !ix_crcstrip) -#endif /* DEV_NETMAP */ - rdrxctl |= IXGBE_RDRXCTL_CRCSTRIP; - rdrxctl |= IXGBE_RDRXCTL_RSCACKC; - IXGBE_WRITE_REG(hw, IXGBE_RDRXCTL, rdrxctl); - - rscctrl = IXGBE_READ_REG(hw, IXGBE_RSCCTL(rxr->me)); - rscctrl |= IXGBE_RSCCTL_RSCEN; - /* - ** Limit the total number of descriptors that - ** can be combined, so it does not exceed 64K - */ - if (rxr->mbuf_sz == MCLBYTES) - rscctrl |= IXGBE_RSCCTL_MAXDESC_16; - else if (rxr->mbuf_sz == MJUMPAGESIZE) - rscctrl |= IXGBE_RSCCTL_MAXDESC_8; - else if (rxr->mbuf_sz == MJUM9BYTES) - rscctrl |= IXGBE_RSCCTL_MAXDESC_4; - else /* Using 16K cluster */ - rscctrl |= IXGBE_RSCCTL_MAXDESC_1; - - IXGBE_WRITE_REG(hw, IXGBE_RSCCTL(rxr->me), rscctrl); - - /* Enable TCP header recognition */ - IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(0), - (IXGBE_READ_REG(hw, IXGBE_PSRTYPE(0)) | - IXGBE_PSRTYPE_TCPHDR)); - - /* Disable RSC for ACK packets */ - IXGBE_WRITE_REG(hw, IXGBE_RSCDBU, - (IXGBE_RSCDBU_RSCACKDIS | IXGBE_READ_REG(hw, IXGBE_RSCDBU))); - - rxr->hw_rsc = TRUE; -} + struct adapter *sc = arg; + struct ix_queue *que = &sc->queues[rxqid]; + struct rx_ring *rxr = &que->rxr; + int i; + uint32_t next_pidx; -/********************************************************************* - * - * Refresh mbuf buffers for RX descriptor rings - * - now keeps its own state so discards due to resource - * exhaustion are unnecessary, if an mbuf cannot be obtained - * it just returns, keeping its placeholder, thus it can simply - * be recalled to try again. - * - **********************************************************************/ -static void -ixgbe_refresh_mbufs(struct rx_ring *rxr, int limit) -{ - struct adapter *adapter = rxr->adapter; - bus_dma_segment_t seg[1]; - struct ixgbe_rx_buf *rxbuf; - struct mbuf *mp; - int i, j, nsegs, error; - bool refreshed = FALSE; - - i = j = rxr->next_to_refresh; - /* Control the loop with one beyond */ - if (++j == rxr->num_desc) - j = 0; - - while (j != limit) { - rxbuf = &rxr->rx_buffers[i]; - if (rxbuf->buf == NULL) { - mp = m_getjcl(M_NOWAIT, MT_DATA, - M_PKTHDR, rxr->mbuf_sz); - if (mp == NULL) - goto update; - if (adapter->max_frame_size <= (MCLBYTES - ETHER_ALIGN)) - m_adj(mp, ETHER_ALIGN); - } else - mp = rxbuf->buf; - - mp->m_pkthdr.len = mp->m_len = rxr->mbuf_sz; - - /* If we're dealing with an mbuf that was copied rather - * than replaced, there's no need to go through busdma. - */ - if ((rxbuf->flags & IXGBE_RX_COPY) == 0) { - /* Get the memory mapping */ - bus_dmamap_unload(rxr->ptag, rxbuf->pmap); - error = bus_dmamap_load_mbuf_sg(rxr->ptag, - rxbuf->pmap, mp, seg, &nsegs, BUS_DMA_NOWAIT); - if (error != 0) { - printf("Refresh mbufs: payload dmamap load" - " failure - %d\n", error); - m_free(mp); - rxbuf->buf = NULL; - goto update; - } - rxbuf->buf = mp; - bus_dmamap_sync(rxr->ptag, rxbuf->pmap, - BUS_DMASYNC_PREREAD); - rxbuf->addr = rxr->rx_base[i].read.pkt_addr = - htole64(seg[0].ds_addr); - } else { - rxr->rx_base[i].read.pkt_addr = rxbuf->addr; - rxbuf->flags &= ~IXGBE_RX_COPY; - } - - refreshed = TRUE; - /* Next is precalculated */ - i = j; - rxr->next_to_refresh = i; - if (++j == rxr->num_desc) - j = 0; + for (i = 0, next_pidx = pidx; i < count; i++) { + rxr->rx_base[next_pidx].read.pkt_addr = htole64(paddrs[i]); + if (++next_pidx == ixgbe_sctx->isc_nrxd) + next_pidx = 0; } -update: - if (refreshed) /* Update hardware tail index */ - IXGBE_WRITE_REG(&adapter->hw, - rxr->tail, rxr->next_to_refresh); - return; } -/********************************************************************* - * - * Allocate memory for rx_buffer structures. Since we use one - * rx_buffer per received packet, the maximum number of rx_buffer's - * that we'll need is equal to the number of receive descriptors - * that we've allocated. - * - **********************************************************************/ -int -ixgbe_allocate_receive_buffers(struct rx_ring *rxr) +static void ixgbe_isc_rxd_flush(void *arg, uint16_t rxqid, uint8_t flid __unused, uint32_t pidx) { - struct adapter *adapter = rxr->adapter; - device_t dev = adapter->dev; - struct ixgbe_rx_buf *rxbuf; - int bsize, error; - - bsize = sizeof(struct ixgbe_rx_buf) * rxr->num_desc; - if (!(rxr->rx_buffers = - (struct ixgbe_rx_buf *) malloc(bsize, - M_DEVBUF, M_NOWAIT | M_ZERO))) { - device_printf(dev, "Unable to allocate rx_buffer memory\n"); - error = ENOMEM; - goto fail; - } - - if ((error = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */ - 1, 0, /* alignment, bounds */ - BUS_SPACE_MAXADDR, /* lowaddr */ - BUS_SPACE_MAXADDR, /* highaddr */ - NULL, NULL, /* filter, filterarg */ - MJUM16BYTES, /* maxsize */ - 1, /* nsegments */ - MJUM16BYTES, /* maxsegsize */ - 0, /* flags */ - NULL, /* lockfunc */ - NULL, /* lockfuncarg */ - &rxr->ptag))) { - device_printf(dev, "Unable to create RX DMA tag\n"); - goto fail; - } - - for (int i = 0; i < rxr->num_desc; i++, rxbuf++) { - rxbuf = &rxr->rx_buffers[i]; - error = bus_dmamap_create(rxr->ptag, 0, &rxbuf->pmap); - if (error) { - device_printf(dev, "Unable to create RX dma map\n"); - goto fail; - } - } - - return (0); - -fail: - /* Frees all, but can handle partial completion */ - ixgbe_free_receive_structures(adapter); - return (error); -} - -static void -ixgbe_free_receive_ring(struct rx_ring *rxr) -{ - struct ixgbe_rx_buf *rxbuf; - - for (int i = 0; i < rxr->num_desc; i++) { - rxbuf = &rxr->rx_buffers[i]; - if (rxbuf->buf != NULL) { - bus_dmamap_sync(rxr->ptag, rxbuf->pmap, - BUS_DMASYNC_POSTREAD); - bus_dmamap_unload(rxr->ptag, rxbuf->pmap); - rxbuf->buf->m_flags |= M_PKTHDR; - m_freem(rxbuf->buf); - rxbuf->buf = NULL; - rxbuf->flags = 0; - } - } -} - -/********************************************************************* - * - * Initialize a receive ring and its buffers. - * - **********************************************************************/ -static int -ixgbe_setup_receive_ring(struct rx_ring *rxr) -{ - struct adapter *adapter; - struct ifnet *ifp; - device_t dev; - struct ixgbe_rx_buf *rxbuf; - bus_dma_segment_t seg[1]; - struct lro_ctrl *lro = &rxr->lro; - int rsize, nsegs, error = 0; -#ifdef DEV_NETMAP - struct netmap_adapter *na = NA(rxr->adapter->ifp); - struct netmap_slot *slot; -#endif /* DEV_NETMAP */ - - adapter = rxr->adapter; - ifp = adapter->ifp; - dev = adapter->dev; - - /* Clear the ring contents */ - IXGBE_RX_LOCK(rxr); -#ifdef DEV_NETMAP - /* same as in ixgbe_setup_transmit_ring() */ - slot = netmap_reset(na, NR_RX, rxr->me, 0); -#endif /* DEV_NETMAP */ - rsize = roundup2(adapter->num_rx_desc * - sizeof(union ixgbe_adv_rx_desc), DBA_ALIGN); - bzero((void *)rxr->rx_base, rsize); - /* Cache the size */ - rxr->mbuf_sz = adapter->rx_mbuf_sz; - - /* Free current RX buffer structs and their mbufs */ - ixgbe_free_receive_ring(rxr); - - /* Now replenish the mbufs */ - for (int j = 0; j != rxr->num_desc; ++j) { - struct mbuf *mp; - - rxbuf = &rxr->rx_buffers[j]; -#ifdef DEV_NETMAP - /* - * In netmap mode, fill the map and set the buffer - * address in the NIC ring, considering the offset - * between the netmap and NIC rings (see comment in - * ixgbe_setup_transmit_ring() ). No need to allocate - * an mbuf, so end the block with a continue; - */ - if (slot) { - int sj = netmap_idx_n2k(&na->rx_rings[rxr->me], j); - uint64_t paddr; - void *addr; - - addr = PNMB(na, slot + sj, &paddr); - netmap_load_map(na, rxr->ptag, rxbuf->pmap, addr); - /* Update descriptor and the cached value */ - rxr->rx_base[j].read.pkt_addr = htole64(paddr); - rxbuf->addr = htole64(paddr); - continue; - } -#endif /* DEV_NETMAP */ - rxbuf->flags = 0; - rxbuf->buf = m_getjcl(M_NOWAIT, MT_DATA, - M_PKTHDR, adapter->rx_mbuf_sz); - if (rxbuf->buf == NULL) { - error = ENOBUFS; - goto fail; - } - mp = rxbuf->buf; - mp->m_pkthdr.len = mp->m_len = rxr->mbuf_sz; - /* Get the memory mapping */ - error = bus_dmamap_load_mbuf_sg(rxr->ptag, - rxbuf->pmap, mp, seg, - &nsegs, BUS_DMA_NOWAIT); - if (error != 0) - goto fail; - bus_dmamap_sync(rxr->ptag, - rxbuf->pmap, BUS_DMASYNC_PREREAD); - /* Update the descriptor and the cached value */ - rxr->rx_base[j].read.pkt_addr = htole64(seg[0].ds_addr); - rxbuf->addr = htole64(seg[0].ds_addr); - } - - - /* Setup our descriptor indices */ - rxr->next_to_check = 0; - rxr->next_to_refresh = 0; - rxr->lro_enabled = FALSE; - rxr->rx_copies = 0; - rxr->rx_bytes = 0; - rxr->vtag_strip = FALSE; - - bus_dmamap_sync(rxr->rxdma.dma_tag, rxr->rxdma.dma_map, - BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); - - /* - ** Now set up the LRO interface: - */ - if (ixgbe_rsc_enable) - ixgbe_setup_hw_rsc(rxr); - else if (ifp->if_capenable & IFCAP_LRO) { - int err = tcp_lro_init(lro); - if (err) { - device_printf(dev, "LRO Initialization failed!\n"); - goto fail; - } - INIT_DEBUGOUT("RX Soft LRO Initialized\n"); - rxr->lro_enabled = TRUE; - lro->ifp = adapter->ifp; - } - - IXGBE_RX_UNLOCK(rxr); - return (0); - -fail: - ixgbe_free_receive_ring(rxr); - IXGBE_RX_UNLOCK(rxr); - return (error); -} - -/********************************************************************* - * - * Initialize all receive rings. - * - **********************************************************************/ -int -ixgbe_setup_receive_structures(struct adapter *adapter) -{ - struct rx_ring *rxr = adapter->rx_rings; - int j; - - for (j = 0; j < adapter->num_queues; j++, rxr++) - if (ixgbe_setup_receive_ring(rxr)) - goto fail; - - return (0); -fail: - /* - * Free RX buffers allocated so far, we will only handle - * the rings that completed, the failing case will have - * cleaned up for itself. 'j' failed, so its the terminus. - */ - for (int i = 0; i < j; ++i) { - rxr = &adapter->rx_rings[i]; - ixgbe_free_receive_ring(rxr); - } + struct adapter *sc = arg; + struct ix_queue *que = &sc->queues[rxqid]; + struct rx_ring *rxr = &que->rxr; - return (ENOBUFS); + IXGBE_WRITE_REG(&sc->hw, rxr->tail, pidx); } - -/********************************************************************* - * - * Free all receive rings. - * - **********************************************************************/ -void -ixgbe_free_receive_structures(struct adapter *adapter) +static int ixgbe_isc_rxd_available(void *arg, uint16_t rxqid, uint32_t idx) { - struct rx_ring *rxr = adapter->rx_rings; - - INIT_DEBUGOUT("ixgbe_free_receive_structures: begin"); - - for (int i = 0; i < adapter->num_queues; i++, rxr++) { - struct lro_ctrl *lro = &rxr->lro; - ixgbe_free_receive_buffers(rxr); - /* Free LRO memory */ - tcp_lro_free(lro); - /* Free the ring memory as well */ - ixgbe_dma_free(adapter, &rxr->rxdma); - } - - free(adapter->rx_rings, M_DEVBUF); -} - + struct adapter *sc = arg; + struct ix_queue *que = &sc->queues[rxqid]; + struct rx_ring *rxr = &que->rxr; + union ixgbe_adv_rx_desc *rxd; + u16 pkt_info; + u32 staterr = 0; + int cnt, i; + + for (cnt = 0, i = idx; cnt < ixgbe_sctx->isc_nrxd;) { + rxd = &rxr->rx_base[i]; + staterr = le32toh(rxd->wb.upper.status_error); + pkt_info = le16toh(rxd->wb.lower.lo_dword.hs_rss.pkt_info); -/********************************************************************* - * - * Free receive ring data structures - * - **********************************************************************/ -void -ixgbe_free_receive_buffers(struct rx_ring *rxr) -{ - struct adapter *adapter = rxr->adapter; - struct ixgbe_rx_buf *rxbuf; - - INIT_DEBUGOUT("ixgbe_free_receive_buffers: begin"); - - /* Cleanup any existing buffers */ - if (rxr->rx_buffers != NULL) { - for (int i = 0; i < adapter->num_rx_desc; i++) { - rxbuf = &rxr->rx_buffers[i]; - if (rxbuf->buf != NULL) { - bus_dmamap_sync(rxr->ptag, rxbuf->pmap, - BUS_DMASYNC_POSTREAD); - bus_dmamap_unload(rxr->ptag, rxbuf->pmap); - rxbuf->buf->m_flags |= M_PKTHDR; - m_freem(rxbuf->buf); - } - rxbuf->buf = NULL; - if (rxbuf->pmap != NULL) { - bus_dmamap_destroy(rxr->ptag, rxbuf->pmap); - rxbuf->pmap = NULL; - } + if ((staterr & IXGBE_RXD_STAT_DD) == 0) { + break; } - if (rxr->rx_buffers != NULL) { - free(rxr->rx_buffers, M_DEVBUF); - rxr->rx_buffers = NULL; + cnt++; + if (++i == ixgbe_sctx->isc_nrxd) { + i = 0; } } - if (rxr->ptag != NULL) { - bus_dma_tag_destroy(rxr->ptag); - rxr->ptag = NULL; - } - - return; -} - -static __inline void -ixgbe_rx_input(struct rx_ring *rxr, struct ifnet *ifp, struct mbuf *m, u32 ptype) -{ - - /* - * ATM LRO is only for IP/TCP packets and TCP checksum of the packet - * should be computed by hardware. Also it should not have VLAN tag in - * ethernet header. In case of IPv6 we do not yet support ext. hdrs. - */ - if (rxr->lro_enabled && - (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0 && - (ptype & IXGBE_RXDADV_PKTTYPE_ETQF) == 0 && - ((ptype & (IXGBE_RXDADV_PKTTYPE_IPV4 | IXGBE_RXDADV_PKTTYPE_TCP)) == - (IXGBE_RXDADV_PKTTYPE_IPV4 | IXGBE_RXDADV_PKTTYPE_TCP) || - (ptype & (IXGBE_RXDADV_PKTTYPE_IPV6 | IXGBE_RXDADV_PKTTYPE_TCP)) == - (IXGBE_RXDADV_PKTTYPE_IPV6 | IXGBE_RXDADV_PKTTYPE_TCP)) && - (m->m_pkthdr.csum_flags & (CSUM_DATA_VALID | CSUM_PSEUDO_HDR)) == - (CSUM_DATA_VALID | CSUM_PSEUDO_HDR)) { - /* - * Send to the stack if: - ** - LRO not enabled, or - ** - no LRO resources, or - ** - lro enqueue fails - */ - if (rxr->lro.lro_cnt != 0) - if (tcp_lro_rx(&rxr->lro, m, 0) == 0) - return; - } - IXGBE_RX_UNLOCK(rxr); - (*ifp->if_input)(ifp, m); - IXGBE_RX_LOCK(rxr); -} - -static __inline void -ixgbe_rx_discard(struct rx_ring *rxr, int i) -{ - struct ixgbe_rx_buf *rbuf; - - rbuf = &rxr->rx_buffers[i]; - - - /* - ** With advanced descriptors the writeback - ** clobbers the buffer addrs, so its easier - ** to just free the existing mbufs and take - ** the normal refresh path to get new buffers - ** and mapping. - */ - - if (rbuf->fmp != NULL) {/* Partial chain ? */ - rbuf->fmp->m_flags |= M_PKTHDR; - m_freem(rbuf->fmp); - rbuf->fmp = NULL; - rbuf->buf = NULL; /* rbuf->buf is part of fmp's chain */ - } else if (rbuf->buf) { - m_free(rbuf->buf); - rbuf->buf = NULL; - } - bus_dmamap_unload(rxr->ptag, rbuf->pmap); - - rbuf->flags = 0; - - return; + return (cnt); } - -/********************************************************************* +/**************************************************************** + * Routine sends data which has been dma'ed into host memory + * to upper layer. Initialize ri structure. * - * This routine executes in interrupt context. It replenishes - * the mbufs in the descriptor and sends data which has been - * dma'ed into host memory to upper layer. - * - * Return TRUE for more work, FALSE for all clean. - *********************************************************************/ -bool -ixgbe_rxeof(struct ix_queue *que) -{ - struct adapter *adapter = que->adapter; - struct rx_ring *rxr = que->rxr; - struct ifnet *ifp = adapter->ifp; - struct lro_ctrl *lro = &rxr->lro; - struct lro_entry *queued; - int i, nextp, processed = 0; - u32 staterr = 0; - u32 count = adapter->rx_process_limit; - union ixgbe_adv_rx_desc *cur; - struct ixgbe_rx_buf *rbuf, *nbuf; - u16 pkt_info; - - IXGBE_RX_LOCK(rxr); - -#ifdef DEV_NETMAP - /* Same as the txeof routine: wakeup clients on intr. */ - if (netmap_rx_irq(ifp, rxr->me, &processed)) { - IXGBE_RX_UNLOCK(rxr); - return (FALSE); + * Returns 0 upon success, errno on failure + ***************************************************************/ + +static int +ixgbe_isc_rxd_pkt_get(void *arg, if_rxd_info_t ri) +{ + struct adapter *adapter = arg; + struct ix_queue *que = &adapter->queues[ri->iri_qsidx]; + struct rx_ring *rxr = &que->rxr; + struct ifnet *ifp = iflib_get_ifp(adapter->ctx); + union ixgbe_adv_rx_desc *rxd; + + u16 pkt_info, len; + u16 vtag = 0; + u32 ptype; + u32 staterr = 0; + bool eop; + + rxd = &rxr->rx_base[ri->iri_cidx]; + staterr = le32toh(rxd->wb.upper.status_error); + pkt_info = le16toh(rxd->wb.lower.lo_dword.hs_rss.pkt_info); + + /* Error Checking then decrement count */ + MPASS ((staterr & IXGBE_RXD_STAT_DD) != 0); + + len = le16toh(rxd->wb.upper.length); + ptype = le32toh(rxd->wb.lower.lo_dword.data) & + IXGBE_RXDADV_PKTTYPE_MASK; + + ri->iri_len = len; + rxr->bytes += len; + + rxd->wb.upper.status_error = 0; + eop = ((staterr & IXGBE_RXD_STAT_EOP) != 0); + if (staterr & IXGBE_RXD_STAT_VP) { + vtag = le16toh(rxd->wb.upper.vlan); + } else { + vtag = 0; } -#endif /* DEV_NETMAP */ - - for (i = rxr->next_to_check; count != 0;) { - struct mbuf *sendmp, *mp; - u32 rsc, ptype; - u16 len; - u16 vtag = 0; - bool eop; - - /* Sync the ring. */ - bus_dmamap_sync(rxr->rxdma.dma_tag, rxr->rxdma.dma_map, - BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); - - cur = &rxr->rx_base[i]; - staterr = le32toh(cur->wb.upper.status_error); - pkt_info = le16toh(cur->wb.lower.lo_dword.hs_rss.pkt_info); - - if ((staterr & IXGBE_RXD_STAT_DD) == 0) - break; - if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) - break; + + /* Make sure bad packets are discarded */ + if (eop && (staterr & IXGBE_RXDADV_ERR_FRAME_ERR_MASK) != 0) { - count--; - sendmp = NULL; - nbuf = NULL; - rsc = 0; - cur->wb.upper.status_error = 0; - rbuf = &rxr->rx_buffers[i]; - mp = rbuf->buf; - - len = le16toh(cur->wb.upper.length); - ptype = le32toh(cur->wb.lower.lo_dword.data) & - IXGBE_RXDADV_PKTTYPE_MASK; - eop = ((staterr & IXGBE_RXD_STAT_EOP) != 0); - - /* Make sure bad packets are discarded */ - if (eop && (staterr & IXGBE_RXDADV_ERR_FRAME_ERR_MASK) != 0) { #if __FreeBSD_version >= 1100036 - if (IXGBE_IS_VF(adapter)) - if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); -#endif - rxr->rx_discarded++; - ixgbe_rx_discard(rxr, i); - goto next_desc; - } - - /* - ** On 82599 which supports a hardware - ** LRO (called HW RSC), packets need - ** not be fragmented across sequential - ** descriptors, rather the next descriptor - ** is indicated in bits of the descriptor. - ** This also means that we might proceses - ** more than one packet at a time, something - ** that has never been true before, it - ** required eliminating global chain pointers - ** in favor of what we are doing here. -jfv - */ - if (!eop) { - /* - ** Figure out the next descriptor - ** of this frame. - */ - if (rxr->hw_rsc == TRUE) { - rsc = ixgbe_rsc_count(cur); - rxr->rsc_num += (rsc - 1); - } - if (rsc) { /* Get hardware index */ - nextp = ((staterr & - IXGBE_RXDADV_NEXTP_MASK) >> - IXGBE_RXDADV_NEXTP_SHIFT); - } else { /* Just sequential */ - nextp = i + 1; - if (nextp == adapter->num_rx_desc) - nextp = 0; - } - nbuf = &rxr->rx_buffers[nextp]; - prefetch(nbuf); - } - /* - ** Rather than using the fmp/lmp global pointers - ** we now keep the head of a packet chain in the - ** buffer struct and pass this along from one - ** descriptor to the next, until we get EOP. - */ - mp->m_len = len; - /* - ** See if there is a stored head - ** that determines what we are - */ - sendmp = rbuf->fmp; - if (sendmp != NULL) { /* secondary frag */ - rbuf->buf = rbuf->fmp = NULL; - mp->m_flags &= ~M_PKTHDR; - sendmp->m_pkthdr.len += mp->m_len; - } else { - /* - * Optimize. This might be a small packet, - * maybe just a TCP ACK. Do a fast copy that - * is cache aligned into a new mbuf, and - * leave the old mbuf+cluster for re-use. - */ - if (eop && len <= IXGBE_RX_COPY_LEN) { - sendmp = m_gethdr(M_NOWAIT, MT_DATA); - if (sendmp != NULL) { - sendmp->m_data += - IXGBE_RX_COPY_ALIGN; - ixgbe_bcopy(mp->m_data, - sendmp->m_data, len); - sendmp->m_len = len; - rxr->rx_copies++; - rbuf->flags |= IXGBE_RX_COPY; - } - } - if (sendmp == NULL) { - rbuf->buf = rbuf->fmp = NULL; - sendmp = mp; - } - - /* first desc of a non-ps chain */ - sendmp->m_flags |= M_PKTHDR; - sendmp->m_pkthdr.len = mp->m_len; - } - ++processed; - - /* Pass the head pointer on */ - if (eop == 0) { - nbuf->fmp = sendmp; - sendmp = NULL; - mp->m_next = nbuf->buf; - } else { /* Sending this frame */ - sendmp->m_pkthdr.rcvif = ifp; - rxr->rx_packets++; - /* capture data for AIM */ - rxr->bytes += sendmp->m_pkthdr.len; - rxr->rx_bytes += sendmp->m_pkthdr.len; - /* Process vlan info */ - if ((rxr->vtag_strip) && - (staterr & IXGBE_RXD_STAT_VP)) - vtag = le16toh(cur->wb.upper.vlan); - if (vtag) { - sendmp->m_pkthdr.ether_vtag = vtag; - sendmp->m_flags |= M_VLANTAG; - } - if ((ifp->if_capenable & IFCAP_RXCSUM) != 0) - ixgbe_rx_checksum(staterr, sendmp, ptype); - - /* - * In case of multiqueue, we have RXCSUM.PCSD bit set - * and never cleared. This means we have RSS hash - * available to be used. - */ - if (adapter->num_queues > 1) { - sendmp->m_pkthdr.flowid = - le32toh(cur->wb.lower.hi_dword.rss); - switch (pkt_info & IXGBE_RXDADV_RSSTYPE_MASK) { - case IXGBE_RXDADV_RSSTYPE_IPV4: - M_HASHTYPE_SET(sendmp, - M_HASHTYPE_RSS_IPV4); - break; - case IXGBE_RXDADV_RSSTYPE_IPV4_TCP: - M_HASHTYPE_SET(sendmp, - M_HASHTYPE_RSS_TCP_IPV4); - break; - case IXGBE_RXDADV_RSSTYPE_IPV6: - M_HASHTYPE_SET(sendmp, - M_HASHTYPE_RSS_IPV6); - break; - case IXGBE_RXDADV_RSSTYPE_IPV6_TCP: - M_HASHTYPE_SET(sendmp, - M_HASHTYPE_RSS_TCP_IPV6); - break; - case IXGBE_RXDADV_RSSTYPE_IPV6_EX: - M_HASHTYPE_SET(sendmp, - M_HASHTYPE_RSS_IPV6_EX); - break; - case IXGBE_RXDADV_RSSTYPE_IPV6_TCP_EX: - M_HASHTYPE_SET(sendmp, - M_HASHTYPE_RSS_TCP_IPV6_EX); - break; -#if __FreeBSD_version > 1100000 - case IXGBE_RXDADV_RSSTYPE_IPV4_UDP: - M_HASHTYPE_SET(sendmp, - M_HASHTYPE_RSS_UDP_IPV4); - break; - case IXGBE_RXDADV_RSSTYPE_IPV6_UDP: - M_HASHTYPE_SET(sendmp, - M_HASHTYPE_RSS_UDP_IPV6); - break; - case IXGBE_RXDADV_RSSTYPE_IPV6_UDP_EX: - M_HASHTYPE_SET(sendmp, - M_HASHTYPE_RSS_UDP_IPV6_EX); - break; + if (IXGBE_IS_VF(adapter)) + if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); #endif - default: - M_HASHTYPE_SET(sendmp, - M_HASHTYPE_OPAQUE); - } - } else { - sendmp->m_pkthdr.flowid = que->msix; - M_HASHTYPE_SET(sendmp, M_HASHTYPE_OPAQUE); - } - } -next_desc: - bus_dmamap_sync(rxr->rxdma.dma_tag, rxr->rxdma.dma_map, - BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); - - /* Advance our pointers to the next descriptor. */ - if (++i == rxr->num_desc) - i = 0; - /* Now send to the stack or do LRO */ - if (sendmp != NULL) { - rxr->next_to_check = i; - ixgbe_rx_input(rxr, ifp, sendmp, ptype); - i = rxr->next_to_check; - } - - /* Every 8 descriptors we go to refresh mbufs */ - if (processed == 8) { - ixgbe_refresh_mbufs(rxr, i); - processed = 0; - } + rxr->rx_discarded++; + return (EBADMSG); } - /* Refresh any remaining buf structs */ - if (ixgbe_rx_unrefreshed(rxr)) - ixgbe_refresh_mbufs(rxr, i); + /* Prefetch the next buffer */ + if (!eop) { + ri->iri_next_offset = 1; + } else { + rxr->rx_packets++; + rxr->packets++; - rxr->next_to_check = i; + if ((ifp->if_capenable & IFCAP_RXCSUM) != 0) + ixgbe_rx_checksum(staterr, ri, ptype); - /* - * Flush any outstanding LRO work - */ - while ((queued = SLIST_FIRST(&lro->lro_active)) != NULL) { - SLIST_REMOVE_HEAD(&lro->lro_active, next); - tcp_lro_flush(lro, queued); +#ifdef RSS + ri->iri_flowid = + le32toh(rxd->wb.lower.hi_dword.rss); + ri->iri_rsstype = ixgbe_determine_rsstype(pkt_info); +#else + ri->iri_flowid = que->msix; + ri->iri_rsstype = M_HASHTYPE_OPAQUE; +#endif + + if (vtag) { + ri->iri_vtag = vtag; + ri->iri_flags |= M_VLANTAG; + } + ri->iri_next_offset = 0; } - - IXGBE_RX_UNLOCK(rxr); - - /* - ** Still have cleaning to do? - */ - if ((staterr & IXGBE_RXD_STAT_DD) != 0) - return (TRUE); - else - return (FALSE); + return (0); } - /********************************************************************* * * Verify that the hardware indicated that the checksum is valid. @@ -2028,280 +476,67 @@ * *********************************************************************/ static void -ixgbe_rx_checksum(u32 staterr, struct mbuf * mp, u32 ptype) +ixgbe_rx_checksum(u32 staterr, if_rxd_info_t ri, u32 ptype) { u16 status = (u16) staterr; u8 errors = (u8) (staterr >> 24); - bool sctp = false; + bool sctp = FALSE; if ((ptype & IXGBE_RXDADV_PKTTYPE_ETQF) == 0 && (ptype & IXGBE_RXDADV_PKTTYPE_SCTP) != 0) - sctp = true; + sctp = TRUE; - /* IPv4 checksum */ if (status & IXGBE_RXD_STAT_IPCS) { - mp->m_pkthdr.csum_flags |= CSUM_L3_CALC; - /* IP Checksum Good */ - if (!(errors & IXGBE_RXD_ERR_IPE)) - mp->m_pkthdr.csum_flags |= CSUM_L3_VALID; + if (!(errors & IXGBE_RXD_ERR_IPE)) { + /* IP Checksum Good */ + ri->iri_csum_flags = CSUM_IP_CHECKED | CSUM_IP_VALID; + } else + ri->iri_csum_flags = 0; } - /* TCP/UDP/SCTP checksum */ if (status & IXGBE_RXD_STAT_L4CS) { - mp->m_pkthdr.csum_flags |= CSUM_L4_CALC; + u64 type = (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); +#if __FreeBSD_version >= 800000 + if (sctp) + type = CSUM_SCTP_VALID; +#endif if (!(errors & IXGBE_RXD_ERR_TCPE)) { - mp->m_pkthdr.csum_flags |= CSUM_L4_VALID; + ri->iri_csum_flags |= type; if (!sctp) - mp->m_pkthdr.csum_data = htons(0xffff); - } + ri->iri_csum_data = htons(0xffff); + } } } /******************************************************************** - * Manage DMA'able memory. - *******************************************************************/ -static void -ixgbe_dmamap_cb(void *arg, bus_dma_segment_t * segs, int nseg, int error) -{ - if (error) - return; - *(bus_addr_t *) arg = segs->ds_addr; - return; -} - -int -ixgbe_dma_malloc(struct adapter *adapter, bus_size_t size, - struct ixgbe_dma_alloc *dma, int mapflags) -{ - device_t dev = adapter->dev; - int r; - - r = bus_dma_tag_create(bus_get_dma_tag(adapter->dev), /* parent */ - DBA_ALIGN, 0, /* alignment, bounds */ - BUS_SPACE_MAXADDR, /* lowaddr */ - BUS_SPACE_MAXADDR, /* highaddr */ - NULL, NULL, /* filter, filterarg */ - size, /* maxsize */ - 1, /* nsegments */ - size, /* maxsegsize */ - BUS_DMA_ALLOCNOW, /* flags */ - NULL, /* lockfunc */ - NULL, /* lockfuncarg */ - &dma->dma_tag); - if (r != 0) { - device_printf(dev,"ixgbe_dma_malloc: bus_dma_tag_create failed; " - "error %u\n", r); - goto fail_0; - } - r = bus_dmamem_alloc(dma->dma_tag, (void **)&dma->dma_vaddr, - BUS_DMA_NOWAIT, &dma->dma_map); - if (r != 0) { - device_printf(dev,"ixgbe_dma_malloc: bus_dmamem_alloc failed; " - "error %u\n", r); - goto fail_1; - } - r = bus_dmamap_load(dma->dma_tag, dma->dma_map, dma->dma_vaddr, - size, - ixgbe_dmamap_cb, - &dma->dma_paddr, - mapflags | BUS_DMA_NOWAIT); - if (r != 0) { - device_printf(dev,"ixgbe_dma_malloc: bus_dmamap_load failed; " - "error %u\n", r); - goto fail_2; - } - dma->dma_size = size; - return (0); -fail_2: - bus_dmamem_free(dma->dma_tag, dma->dma_vaddr, dma->dma_map); -fail_1: - bus_dma_tag_destroy(dma->dma_tag); -fail_0: - dma->dma_tag = NULL; - return (r); -} - -void -ixgbe_dma_free(struct adapter *adapter, struct ixgbe_dma_alloc *dma) -{ - bus_dmamap_sync(dma->dma_tag, dma->dma_map, - BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); - bus_dmamap_unload(dma->dma_tag, dma->dma_map); - bus_dmamem_free(dma->dma_tag, dma->dma_vaddr, dma->dma_map); - bus_dma_tag_destroy(dma->dma_tag); -} - - -/********************************************************************* * - * Allocate memory for the transmit and receive rings, and then - * the descriptors associated with each, called only once at attach. - * - **********************************************************************/ -int -ixgbe_allocate_queues(struct adapter *adapter) -{ - device_t dev = adapter->dev; - struct ix_queue *que; - struct tx_ring *txr; - struct rx_ring *rxr; - int rsize, tsize, error = IXGBE_SUCCESS; - int txconf = 0, rxconf = 0; -#ifdef PCI_IOV - enum ixgbe_iov_mode iov_mode; -#endif - - /* First allocate the top level queue structs */ - if (!(adapter->queues = - (struct ix_queue *) malloc(sizeof(struct ix_queue) * - adapter->num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) { - device_printf(dev, "Unable to allocate queue memory\n"); - error = ENOMEM; - goto fail; - } - - /* First allocate the TX ring struct memory */ - if (!(adapter->tx_rings = - (struct tx_ring *) malloc(sizeof(struct tx_ring) * - adapter->num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) { - device_printf(dev, "Unable to allocate TX ring memory\n"); - error = ENOMEM; - goto tx_fail; - } - - /* Next allocate the RX */ - if (!(adapter->rx_rings = - (struct rx_ring *) malloc(sizeof(struct rx_ring) * - adapter->num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) { - device_printf(dev, "Unable to allocate RX ring memory\n"); - error = ENOMEM; - goto rx_fail; - } - - /* For the ring itself */ - tsize = roundup2(adapter->num_tx_desc * - sizeof(union ixgbe_adv_tx_desc), DBA_ALIGN); - -#ifdef PCI_IOV - iov_mode = ixgbe_get_iov_mode(adapter); - adapter->pool = ixgbe_max_vfs(iov_mode); -#else - adapter->pool = 0; -#endif - /* - * Now set up the TX queues, txconf is needed to handle the - * possibility that things fail midcourse and we need to - * undo memory gracefully - */ - for (int i = 0; i < adapter->num_queues; i++, txconf++) { - /* Set up some basics */ - txr = &adapter->tx_rings[i]; - txr->adapter = adapter; -#ifdef PCI_IOV - txr->me = ixgbe_pf_que_index(iov_mode, i); -#else - txr->me = i; -#endif - txr->num_desc = adapter->num_tx_desc; - - /* Initialize the TX side lock */ - snprintf(txr->mtx_name, sizeof(txr->mtx_name), "%s:tx(%d)", - device_get_nameunit(dev), txr->me); - mtx_init(&txr->tx_mtx, txr->mtx_name, NULL, MTX_DEF); - - if (ixgbe_dma_malloc(adapter, tsize, - &txr->txdma, BUS_DMA_NOWAIT)) { - device_printf(dev, - "Unable to allocate TX Descriptor memory\n"); - error = ENOMEM; - goto err_tx_desc; - } - txr->tx_base = (union ixgbe_adv_tx_desc *)txr->txdma.dma_vaddr; - bzero((void *)txr->tx_base, tsize); - - /* Now allocate transmit buffers for the ring */ - if (ixgbe_allocate_transmit_buffers(txr)) { - device_printf(dev, - "Critical Failure setting up transmit buffers\n"); - error = ENOMEM; - goto err_tx_desc; - } -#ifndef IXGBE_LEGACY_TX - /* Allocate a buf ring */ - txr->br = buf_ring_alloc(IXGBE_BR_SIZE, M_DEVBUF, - M_WAITOK, &txr->tx_mtx); - if (txr->br == NULL) { - device_printf(dev, - "Critical Failure setting up buf ring\n"); - error = ENOMEM; - goto err_tx_desc; - } -#endif - } - - /* - * Next the RX queues... - */ - rsize = roundup2(adapter->num_rx_desc * - sizeof(union ixgbe_adv_rx_desc), DBA_ALIGN); - for (int i = 0; i < adapter->num_queues; i++, rxconf++) { - rxr = &adapter->rx_rings[i]; - /* Set up some basics */ - rxr->adapter = adapter; -#ifdef PCI_IOV - rxr->me = ixgbe_pf_que_index(iov_mode, i); -#else - rxr->me = i; -#endif - rxr->num_desc = adapter->num_rx_desc; - - /* Initialize the RX side lock */ - snprintf(rxr->mtx_name, sizeof(rxr->mtx_name), "%s:rx(%d)", - device_get_nameunit(dev), rxr->me); - mtx_init(&rxr->rx_mtx, rxr->mtx_name, NULL, MTX_DEF); - - if (ixgbe_dma_malloc(adapter, rsize, - &rxr->rxdma, BUS_DMA_NOWAIT)) { - device_printf(dev, - "Unable to allocate RxDescriptor memory\n"); - error = ENOMEM; - goto err_rx_desc; - } - rxr->rx_base = (union ixgbe_adv_rx_desc *)rxr->rxdma.dma_vaddr; - bzero((void *)rxr->rx_base, rsize); - - /* Allocate receive buffers for the ring*/ - if (ixgbe_allocate_receive_buffers(rxr)) { - device_printf(dev, - "Critical Failure setting up receive buffers\n"); - error = ENOMEM; - goto err_rx_desc; - } - } - - /* - ** Finally set up the queue holding structs - */ - for (int i = 0; i < adapter->num_queues; i++) { - que = &adapter->queues[i]; - que->adapter = adapter; - que->me = i; - que->txr = &adapter->tx_rings[i]; - que->rxr = &adapter->rx_rings[i]; - } - - return (0); - -err_rx_desc: - for (rxr = adapter->rx_rings; rxconf > 0; rxr++, rxconf--) - ixgbe_dma_free(adapter, &rxr->rxdma); -err_tx_desc: - for (txr = adapter->tx_rings; txconf > 0; txr++, txconf--) - ixgbe_dma_free(adapter, &txr->txdma); - free(adapter->rx_rings, M_DEVBUF); -rx_fail: - free(adapter->tx_rings, M_DEVBUF); -tx_fail: - free(adapter->queues, M_DEVBUF); -fail: - return (error); + * Parse the packet type to determine the appropriate hash + * + ******************************************************************/ +#ifdef RSS +static int +ixgbe_determine_rsstype(u16 pkt_info) +{ + switch (pkt_info & IXGBE_RXDADV_RSSTYPE_MASK) { + case IXGBE_RXDADV_RSSTYPE_IPV4_TCP: + return M_HASHTYPE_RSS_TCP_IPV4; + case IXGBE_RXDADV_RSSTYPE_IPV4: + return M_HASHTYPE_RSS_IPV4; + case IXGBE_RXDADV_RSSTYPE_IPV6_TCP: + return M_HASHTYPE_RSS_TCP_IPV6; + case IXGBE_RXDADV_RSSTYPE_IPV6_EX: + return M_HASHTYPE_RSS_IPV6_EX; + case IXGBE_RXDADV_RSSTYPE_IPV6: + return M_HASHTYPE_RSS_IPV6; + case IXGBE_RXDADV_RSSTYPE_IPV6_TCP_EX: + return M_HASHTYPE_RSS_TCP_IPV6_EX; + case IXGBE_RXDADV_RSSTYPE_IPV4_UDP: + return M_HASHTYPE_RSS_UDP_IPV4; + case IXGBE_RXDADV_RSSTYPE_IPV6_UDP: + return M_HASHTYPE_RSS_UDP_IPV6; + case IXGBE_RXDADV_RSSTYPE_IPV6_UDP_EX: + return M_HASHTYPE_RSS_UDP_IPV6_EX; + default: + return M_HASHTYPE_OPAQUE; + } } +#endif Index: sys/dev/ixgbe/ixgbe.h =================================================================== --- sys/dev/ixgbe/ixgbe.h +++ sys/dev/ixgbe/ixgbe.h @@ -59,9 +59,9 @@ #include #include -#include #include #include +#include #include #include @@ -337,42 +337,22 @@ }; /* -** Driver queue struct: this is the interrupt container -** for the associated tx and rx ring. -*/ -struct ix_queue { - struct adapter *adapter; - u32 msix; /* This queue's MSIX vector */ - u32 eims; /* This queue's EIMS bit */ - u32 eitr_setting; - u32 me; - struct resource *res; - void *tag; - int busy; - struct tx_ring *txr; - struct rx_ring *rxr; - struct task que_task; - struct taskqueue *tq; - u64 irqs; -}; - -/* * The transmit ring, one per queue */ struct tx_ring { - struct adapter *adapter; + struct ix_queue *que; + struct adapter *adapter; struct mtx tx_mtx; u32 me; u32 tail; int busy; union ixgbe_adv_tx_desc *tx_base; struct ixgbe_tx_buf *tx_buffers; - struct ixgbe_dma_alloc txdma; + uint64_t tx_paddr; volatile u16 tx_avail; u16 next_avail_desc; u16 next_to_clean; u16 num_desc; - u32 txd_cmd; bus_dma_tag_t txtag; char mtx_name[16]; #ifndef IXGBE_LEGACY_TX @@ -390,7 +370,7 @@ unsigned long no_tx_map_avail; unsigned long no_tx_dma_setup; u64 no_desc_avail; - u64 total_packets; + u64 total_packets; }; @@ -398,22 +378,18 @@ * The Receive ring, one per rx queue */ struct rx_ring { - struct adapter *adapter; + struct ix_queue *que; + struct adapter *adapter; struct mtx rx_mtx; + struct lro_ctrl lro; u32 me; u32 tail; union ixgbe_adv_rx_desc *rx_base; - struct ixgbe_dma_alloc rxdma; - struct lro_ctrl lro; - bool lro_enabled; + uint64_t rx_paddr; bool hw_rsc; bool vtag_strip; - u16 next_to_refresh; - u16 next_to_check; - u16 num_desc; u16 mbuf_sz; char mtx_name[16]; - struct ixgbe_rx_buf *rx_buffers; bus_dma_tag_t ptag; u32 bytes; /* Used for AIM calc */ @@ -431,6 +407,27 @@ #endif }; +/* +** Driver queue struct: this is the interrupt container +** for the associated tx and rx ring. +*/ +struct ix_queue { + struct adapter *adapter; + u32 msix; /* This queue's MSIX vector */ + u32 eims; /* This queue's EIMS bit */ + u32 eitr_setting; + u32 me; + struct resource *res; + void *tag; + int busy; + struct tx_ring txr; + struct rx_ring rxr; + + struct if_irq que_irq; + + u64 irqs; +}; + #ifdef PCI_IOV #define IXGBE_VF_CTS (1 << 0) /* VF is clear to send. */ #define IXGBE_VF_CAP_MAC (1 << 1) /* VF is permitted to change MAC. */ @@ -455,35 +452,33 @@ /* Our adapter structure */ struct adapter { + if_ctx_t ctx; + if_softc_ctx_t shared; +#define num_queues shared->isc_nqsets +#define max_frame_size shared->isc_max_frame_size +#define intr_type shared->isc_intr + struct ifnet *ifp; struct ixgbe_hw hw; struct ixgbe_osdep osdep; struct device *dev; - struct ifnet *ifp; struct resource *pci_mem; - struct resource *msix_mem; /* * Interrupt resources: this set is * either used for legacy, or for Link * when doing MSIX */ + struct if_irq irq; void *tag; struct resource *res; - struct ifmedia media; - struct callout timer; + struct ifmedia *media; int msix; int if_flags; - struct mtx core_mtx; - - eventhandler_tag vlan_attach; - eventhandler_tag vlan_detach; - u16 num_vlans; - u16 num_queues; /* ** Shadow VFTA table, this is needed because @@ -499,7 +494,6 @@ int advertise; /* link speeds */ bool enable_aim; /* adaptive interrupt moderation */ bool link_active; - u16 max_frame_size; u16 num_segs; u32 link_speed; bool link_up; @@ -512,23 +506,21 @@ bool wol_support; u32 wufc; - /* Mbuf cluster size */ - u32 rx_mbuf_sz; - /* Support for pluggable optics */ bool sfp_probe; - struct task link_task; /* Link tasklet */ - struct task mod_task; /* SFP tasklet */ - struct task msf_task; /* Multispeed Fiber */ + + struct grouptask mod_task; /* SFP tasklet */ + struct grouptask msf_task; /* Multispeed Fiber */ + #ifdef PCI_IOV - struct task mbx_task; /* VF -> PF mailbox interrupt */ + struct grouptask mbx_task; /* VF -> PF mailbox interrupt */ #endif /* PCI_IOV */ #ifdef IXGBE_FDIR int fdir_reinit; - struct task fdir_task; + struct grouptask fdir_task; #endif - struct task phy_task; /* PHY intr tasklet */ - struct taskqueue *tq; + + struct grouptask phy_task; /* PHY intr tasklet */ /* ** Queues: @@ -537,22 +529,9 @@ ** with it. */ struct ix_queue *queues; - - /* - * Transmit rings: - * Allocated at run time, an array of rings. - */ - struct tx_ring *tx_rings; - u32 num_tx_desc; - u32 tx_process_limit; - - /* - * Receive rings: - * Allocated at run time, an array of rings. - */ - struct rx_ring *rx_rings; u64 active_queues; - u32 num_rx_desc; + + u32 tx_process_limit; u32 rx_process_limit; /* Multicast array memory */ @@ -568,6 +547,7 @@ #endif /* Misc stats maintained by the driver */ + unsigned long rx_mbuf_sz; unsigned long dropped_pkts; unsigned long mbuf_defrag_failed; unsigned long mbuf_header_failed; @@ -593,7 +573,6 @@ #endif }; - /* Precision Time Sync (IEEE 1588) defines */ #define ETHERTYPE_IEEE1588 0x88F7 #define PICOSECS_PER_TICK 20833 @@ -616,6 +595,8 @@ #define IXGBE_CORE_LOCK_ASSERT(_sc) mtx_assert(&(_sc)->core_mtx, MA_OWNED) #define IXGBE_TX_LOCK_ASSERT(_sc) mtx_assert(&(_sc)->tx_mtx, MA_OWNED) + + /* For backward compatibility */ #if !defined(PCIER_LINK_STA) #define PCIER_LINK_STA PCIR_EXPRESS_LINK_STA @@ -699,19 +680,6 @@ #endif /* -** Find the number of unrefreshed RX descriptors -*/ -static inline u16 -ixgbe_rx_unrefreshed(struct rx_ring *rxr) -{ - if (rxr->next_to_check > rxr->next_to_refresh) - return (rxr->next_to_check - rxr->next_to_refresh - 1); - else - return ((rxr->num_desc + rxr->next_to_check) - - rxr->next_to_refresh - 1); -} - -/* ** This checks for a zero mac addr, something that will be likely ** unless the Admin on the Host has created one. */ @@ -751,6 +719,7 @@ int ixgbe_dma_malloc(struct adapter *, bus_size_t, struct ixgbe_dma_alloc *, int); void ixgbe_dma_free(struct adapter *, struct ixgbe_dma_alloc *); +int ixgbe_get_regs(SYSCTL_HANDLER_ARGS); #ifdef PCI_IOV Index: sys/dev/ixgbe/ixgbe_sysctl.c =================================================================== --- /dev/null +++ sys/dev/ixgbe/ixgbe_sysctl.c @@ -0,0 +1,519 @@ +#include "ixgbe.h" +/******************************************************************** + * + * Dump Registers + * + ********************************************************************/ +#define IXGBE_REGS_LEN 1139 + +int ixgbe_get_regs(SYSCTL_HANDLER_ARGS) +{ + struct adapter *adapter = (struct adapter *) arg1; + struct ixgbe_hw *hw = &adapter->hw; + struct ix_queue *que = &adapter->queues[0]; + struct rx_ring *rxr = &que->rxr; + struct tx_ring *txr = &que->txr; + + struct sbuf *sb; + u32 *regs_buff = (u32 *)malloc(sizeof(u32) * IXGBE_REGS_LEN, M_DEVBUF, M_NOWAIT); + u8 i, j; + int rc; + + memset(regs_buff, 0, IXGBE_REGS_LEN * sizeof(u32)); + + rc = sysctl_wire_old_buffer(req, 0); + MPASS(rc == 0); + if (rc != 0) + return (rc); + + sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req); + MPASS(sb != NULL); + if (sb == NULL) + return (ENOMEM); + + /* General Registers */ + sbuf_printf(sb, "General Registers\n"); + regs_buff[0] = IXGBE_READ_REG(hw, IXGBE_CTRL); + sbuf_printf(sb, "\tIXGBE_CTRL\t %08x\n", regs_buff[0]); + regs_buff[1] = IXGBE_READ_REG(hw, IXGBE_STATUS); + sbuf_printf(sb, "\tIXGBE_STATUS\t %08x\n", regs_buff[1]); + regs_buff[2] = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT); + sbuf_printf(sb, "\tIXGBE_CTRL_EXT\t %08x\n", regs_buff[2]); + regs_buff[3] = IXGBE_READ_REG(hw, IXGBE_ESDP); + sbuf_printf(sb, "\tIXGBE_ESDP\t %08x\n", regs_buff[3]); + regs_buff[4] = IXGBE_READ_REG(hw, IXGBE_EODSDP); + sbuf_printf(sb, "\tIXGBE_EODSDP\t %08x\n", regs_buff[4]); + regs_buff[5] = IXGBE_READ_REG(hw, IXGBE_LEDCTL); + sbuf_printf(sb, "\tIXGBE_LEDCTL\t %08x\n", regs_buff[5]); + regs_buff[6] = IXGBE_READ_REG(hw, IXGBE_FRTIMER); + sbuf_printf(sb, "\tIXGBE_FRTIMER\t %08x\n", regs_buff[6]); + regs_buff[7] = IXGBE_READ_REG(hw, IXGBE_TCPTIMER); + sbuf_printf(sb, "\tIXGBE_TCPTIMER\t %08x\n\n", regs_buff[7]); + + /* Interrupt */ + /* don't read EICR because it can clear interrupt causes, instead + * read EICS which is a shadow but doesn't clear EICR */ + sbuf_printf(sb, "Interrupt\n"); + regs_buff[18] = IXGBE_READ_REG(hw, IXGBE_EICS); + sbuf_printf(sb, "\tIXGBE_EICS\t %08x\n", regs_buff[18]); + regs_buff[19] = IXGBE_READ_REG(hw, IXGBE_EICS); + sbuf_printf(sb, "\tIXGBE_EICS\t %08x\n", regs_buff[19]); + regs_buff[20] = IXGBE_READ_REG(hw, IXGBE_EIMS); + sbuf_printf(sb, "\tIXGBE_EIMS\t %08x\n", regs_buff[20]); + regs_buff[21] = IXGBE_READ_REG(hw, IXGBE_EIMC); + sbuf_printf(sb, "\tIXGBE_EIMC\t %08x\n", regs_buff[21]); + regs_buff[22] = IXGBE_READ_REG(hw, IXGBE_EIAC); + sbuf_printf(sb, "\tIXGBE_EIAC\t %08x\n", regs_buff[22]); + regs_buff[23] = IXGBE_READ_REG(hw, IXGBE_EIAM); + sbuf_printf(sb, "\tIXGBE_EIAM\t %08x\n", regs_buff[23]); + regs_buff[24] = IXGBE_READ_REG(hw, IXGBE_EITR(0)); + sbuf_printf(sb, "\tIXGBE_EITR(0)\t %08x\n", regs_buff[24]); + regs_buff[25] = IXGBE_READ_REG(hw, IXGBE_IVAR(0)); + sbuf_printf(sb, "\tIXGBE_IVAR(0)\t %08x\n", regs_buff[25]); + regs_buff[26] = IXGBE_READ_REG(hw, IXGBE_MSIXT); + sbuf_printf(sb, "\tIXGBE_MSIXT\t %08x\n", regs_buff[26]); + regs_buff[27] = IXGBE_READ_REG(hw, IXGBE_MSIXPBA); + sbuf_printf(sb, "\tIXGBE_MSIXPBA\t %08x\n", regs_buff[27]); + regs_buff[28] = IXGBE_READ_REG(hw, IXGBE_PBACL(0)); + sbuf_printf(sb, "\tIXGBE_PBACL(0)\t %08x\n", regs_buff[28]); + regs_buff[29] = IXGBE_READ_REG(hw, IXGBE_GPIE); + sbuf_printf(sb, "\tIXGBE_GPIE\t %08x\n", regs_buff[29]); + + /* Flow Control */ + sbuf_printf(sb, "\nFlow Control\n"); + regs_buff[30] = IXGBE_READ_REG(hw, IXGBE_PFCTOP); + sbuf_printf(sb, "\tIXGBE_PFCTOP\t %08x\n", regs_buff[30]); + regs_buff[31] = IXGBE_READ_REG(hw, IXGBE_FCTTV(0)); + sbuf_printf(sb, "\tIXGBE_FCTTV(0)\t %08x\n", regs_buff[31]); + regs_buff[32] = IXGBE_READ_REG(hw, IXGBE_FCTTV(1)); + sbuf_printf(sb, "\tIXGBE_FCTTV(1)\t %08x\n", regs_buff[32]); + regs_buff[33] = IXGBE_READ_REG(hw, IXGBE_FCTTV(2)); + sbuf_printf(sb, "\tIXGBE_FCTTV(2)\t %08x\n", regs_buff[33]); + regs_buff[34] = IXGBE_READ_REG(hw, IXGBE_FCTTV(3)); + sbuf_printf(sb, "\tIXGBE_FCTTV(3)\t %08x\n", regs_buff[34]); + + for (i = 0; i < adapter->num_queues; i++) { + switch (hw->mac.type) { + case ixgbe_mac_82598EB: + regs_buff[35 + i] = IXGBE_READ_REG(hw, IXGBE_FCRTL(i)); + sbuf_printf(sb, "\tIXGBE_FCRTL(%2d)\t %08x\n", i, regs_buff[35+i]); + regs_buff[43 + i] = IXGBE_READ_REG(hw, IXGBE_FCRTH(i)); + sbuf_printf(sb, "\tIXGBE_FCRTH(%2d)\t %08x\n", i, regs_buff[43+i]); + break; + case ixgbe_mac_82599EB: + case ixgbe_mac_X540: + case ixgbe_mac_X550: + case ixgbe_mac_X550EM_x: + regs_buff[35 + i] = IXGBE_READ_REG(hw, IXGBE_FCRTL_82599(i)); + regs_buff[43 + i] = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i)); + break; + default: + break; + } + } + regs_buff[51] = IXGBE_READ_REG(hw, IXGBE_FCRTV); + sbuf_printf(sb, "\tIXGBE_FCRTV\t %08x\n", regs_buff[51]); + regs_buff[52] = IXGBE_READ_REG(hw, IXGBE_TFCS); + sbuf_printf(sb, "\tIXGBE_TFCS\t %08x\n", regs_buff[52]); + + /* Receive DMA */ + sbuf_printf(sb, "\nReceive DMA\n"); + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[53 + i] = IXGBE_READ_REG(hw, IXGBE_RDBAL(i)); + sbuf_printf(sb, "\tIXGBE_RDBAL(%2d)\t %08x\n", i, regs_buff[53+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[117 + i] = IXGBE_READ_REG(hw, IXGBE_RDBAH(i)); + sbuf_printf(sb, "\tIXGBE_RDBAH(%2d)\t %08x\n", i, regs_buff[117+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[181 + i] = IXGBE_READ_REG(hw, IXGBE_RDLEN(i)); + sbuf_printf(sb, "\tIXGBE_RDLEN(%2d)\t %08x\n", i, regs_buff[181+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[245 + i] = IXGBE_READ_REG(hw, IXGBE_RDH(i)); + sbuf_printf(sb, "\tIXGBE_RDH(%2d)\t %08x\n", i, regs_buff[245+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[309 + i] = IXGBE_READ_REG(hw, IXGBE_RDT(i)); + sbuf_printf(sb, "\tIXGBE_RDT(%2d)\t %08x\n", i, regs_buff[309+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[373 + i] = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i)); + sbuf_printf(sb, "\tIXGBE_RXDCTL(%2d)\t %08x\n", i, regs_buff[373+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[437 + i] = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i)); + sbuf_printf(sb, "\tIXGBE_SRRCTL(%2d)\t %08x\n", i, regs_buff[437+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[453 + i] = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i)); + sbuf_printf(sb, "\tIXGBE_DCA_RXCTRL(%2d)\t %08x\n", i, regs_buff[453+i]); + } + regs_buff[469] = IXGBE_READ_REG(hw, IXGBE_RDRXCTL); + sbuf_printf(sb, "\tIXGBE_RDRXCTL\t %08x\n", regs_buff[469]); + + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[470 + i] = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(i)); + sbuf_printf(sb, "\tIXGBE_RXPBSIZE(%2d)\t %08x\n", i, regs_buff[470+i]); + } + regs_buff[478] = IXGBE_READ_REG(hw, IXGBE_RXCTRL); + sbuf_printf(sb, "\tIXGBE_RXCTRL\t %08x\n", regs_buff[478]); + regs_buff[479] = IXGBE_READ_REG(hw, IXGBE_DROPEN); + sbuf_printf(sb, "\tIXGBE_DROPEN\t %08x\n", regs_buff[479]); + + /* Receive */ + sbuf_printf(sb, "\nReceive\n"); + regs_buff[480] = IXGBE_READ_REG(hw, IXGBE_RXCSUM); + sbuf_printf(sb, "\tIXGBE_RXCSUM\t %08x\n", regs_buff[480]); + regs_buff[481] = IXGBE_READ_REG(hw, IXGBE_RFCTL); + sbuf_printf(sb, "\tIXGBE_RFCTL\t %08x\n", regs_buff[481]); + + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[482 + i] = IXGBE_READ_REG(hw, IXGBE_RAL(i)); + sbuf_printf(sb, "\tIXGBE_RAL(%2d)\t %08x\n", i, regs_buff[482+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[498 + i] = IXGBE_READ_REG(hw, IXGBE_RAH(i)); + sbuf_printf(sb, "\tIXGBE_RAH(%2d)\t %08x\n", i, regs_buff[498+i]); + } + regs_buff[514] = IXGBE_READ_REG(hw, IXGBE_PSRTYPE(0)); + sbuf_printf(sb, "\tIXGBE_PSRTYPE\t %08x\n", regs_buff[514]); + regs_buff[515] = IXGBE_READ_REG(hw, IXGBE_FCTRL); + sbuf_printf(sb, "\tIXGBE_FCTRL\t %08x\n", regs_buff[515]); + regs_buff[516] = IXGBE_READ_REG(hw, IXGBE_VLNCTRL); + sbuf_printf(sb, "\tIXGBE_VLNCTRL\t %08x\n", regs_buff[516]); + regs_buff[517] = IXGBE_READ_REG(hw, IXGBE_MCSTCTRL); + sbuf_printf(sb, "\tIXGBE_MCSTCTRL\t %08x\n", regs_buff[517]); + regs_buff[518] = IXGBE_READ_REG(hw, IXGBE_MRQC); + sbuf_printf(sb, "\tIXGBE_MRQC\t %08x\n", regs_buff[518]); + regs_buff[519] = IXGBE_READ_REG(hw, IXGBE_VMD_CTL); + sbuf_printf(sb, "\tIXGBE_VMD_CTL\t %08x\n", regs_buff[519]); + + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[520 + i] = IXGBE_READ_REG(hw, IXGBE_IMIR(i)); + sbuf_printf(sb, "\tIXGBE_IMIR(%2d)\t %08x\n", i, regs_buff[520+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[528 + i] = IXGBE_READ_REG(hw, IXGBE_IMIREXT(i)); + sbuf_printf(sb, "\tIXGBE_IMIREXT(%2d)\t %08x\n", i, regs_buff[528+i]); + } + regs_buff[536] = IXGBE_READ_REG(hw, IXGBE_IMIRVP); + sbuf_printf(sb, "\tIXGBE_IMIRVP\t %08x\n", regs_buff[536]); + + /* Transmit */ + sbuf_printf(sb, "\nTransmit\n"); + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[537 + i] = IXGBE_READ_REG(hw, IXGBE_TDBAL(i)); + sbuf_printf(sb, "\tIXGBE_TDBAL(%2d)\t %08x\n", i, regs_buff[537+i]); + } + + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[569 + i] = IXGBE_READ_REG(hw, IXGBE_TDBAH(i)); + sbuf_printf(sb, "\tIXGBE_TDBAH(%2d)\t %08x\n", i, regs_buff[569+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[601 + i] = IXGBE_READ_REG(hw, IXGBE_TDLEN(i)); + sbuf_printf(sb, "\tIXGBE_TDLEN(%2d)\t %08x\n", i, regs_buff[601+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[633 + i] = IXGBE_READ_REG(hw, IXGBE_TDH(i)); + sbuf_printf(sb, "\tIXGBE_TDH(%2d)\t %08x\n", i, regs_buff[633+i]); + } + for (i = 0; i < adapter->num_queues; i++){ + regs_buff[665 + i] = IXGBE_READ_REG(hw, IXGBE_TDT(i)); + sbuf_printf(sb, "\tIXGBE_TDT(%2d)\t %08x\n", i, regs_buff[665+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[697 + i] = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i)); + sbuf_printf(sb, "\tIXGBE_TXDCTL(%2d)\t %08x\n", i, regs_buff[697+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[729 + i] = IXGBE_READ_REG(hw, IXGBE_TDWBAL(i)); + sbuf_printf(sb, "\tIXGBE_TDWBAL(%2d)\t %08x\n", i, regs_buff[729+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[761 + i] = IXGBE_READ_REG(hw, IXGBE_TDWBAH(i)); + sbuf_printf(sb, "\tIXGBE_TDWBAH(%2d)\t %08x\n", i, regs_buff[761+i]); + } + regs_buff[793] = IXGBE_READ_REG(hw, IXGBE_DTXCTL); + sbuf_printf(sb, "\tIXGBE_DTXCTL\t %08x\n", regs_buff[793]); + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[794 + i] = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL(i)); + sbuf_printf(sb, "\tIXGBE_DCA_TXCTRL(%2d)\t %08x\n", i, regs_buff[794+i]); + } + regs_buff[810] = IXGBE_READ_REG(hw, IXGBE_TIPG); + sbuf_printf(sb, "\tIXGBE_TIPG\t %08x\n", regs_buff[810]); + + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[811 + i] = IXGBE_READ_REG(hw, IXGBE_TXPBSIZE(i)); + sbuf_printf(sb, "\tIXGBE_TXPBSIZE(%2d)\t %08x\n", i, regs_buff[811+i]); + } + regs_buff[819] = IXGBE_READ_REG(hw, IXGBE_MNGTXMAP); + sbuf_printf(sb, "\tIXGBE_MNGTXMAP\t %08x\n", regs_buff[819]); + + /* Wake Up */ + sbuf_printf(sb, "\nWake Up\n"); + regs_buff[820] = IXGBE_READ_REG(hw, IXGBE_WUC); + sbuf_printf(sb, "\tIXGBE_WUC\t %08x\n", regs_buff[820]); + regs_buff[821] = IXGBE_READ_REG(hw, IXGBE_WUFC); + sbuf_printf(sb, "\tIXGBE_WUFC\t %08x\n", regs_buff[821]); + regs_buff[822] = IXGBE_READ_REG(hw, IXGBE_WUS); + sbuf_printf(sb, "\tIXGBE_WUS\t %08x\n", regs_buff[822]); + regs_buff[823] = IXGBE_READ_REG(hw, IXGBE_IPAV); + sbuf_printf(sb, "\tIXGBE_IPAV\t %08x\n", regs_buff[823]); + regs_buff[824] = IXGBE_READ_REG(hw, IXGBE_IP4AT); + sbuf_printf(sb, "\tIXGBE_IP4AT\t %08x\n", regs_buff[824]); + regs_buff[825] = IXGBE_READ_REG(hw, IXGBE_IP6AT); + sbuf_printf(sb, "\tIXGBE_IP6AT\t %08x\n", regs_buff[825]); + regs_buff[826] = IXGBE_READ_REG(hw, IXGBE_WUPL); + sbuf_printf(sb, "\tIXGBE_WUPL\t %08x\n", regs_buff[826]); + regs_buff[827] = IXGBE_READ_REG(hw, IXGBE_WUPM); + sbuf_printf(sb, "\tIXGBE_WUPM\t %08x\n", regs_buff[827]); + regs_buff[828] = IXGBE_READ_REG(hw, IXGBE_FHFT(0)); + sbuf_printf(sb, "\tIXGBE_FHFT\t %08x\n", regs_buff[828]); + + /* DCB */ + sbuf_printf(sb, "\nDCB\n"); + regs_buff[829] = IXGBE_READ_REG(hw, IXGBE_RMCS); /* same as FCCFG */ + sbuf_printf(sb, "\tIXGBE_RMCS\t %08x\n", regs_buff[829]); + regs_buff[831] = IXGBE_READ_REG(hw, IXGBE_PDPMCS); /* same as RTTPCS */ + sbuf_printf(sb, "\tIXGBE_PDPMCS\t %08x\n", regs_buff[831]); + + switch (hw->mac.type) { + case ixgbe_mac_82598EB: + regs_buff[830] = IXGBE_READ_REG(hw, IXGBE_DPMCS); + sbuf_printf(sb, "\tIXGBE_DPMCS\t %08x\n", regs_buff[830]); + regs_buff[832] = IXGBE_READ_REG(hw, IXGBE_RUPPBMR); + sbuf_printf(sb, "\tIXGBE_RUPPBMR\t %08x\n", regs_buff[832]); + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[833 + i] = IXGBE_READ_REG(hw, IXGBE_RT2CR(i)); + sbuf_printf(sb, "\tIXGBE_RT2CR(%2d)\t %08x\n", i, regs_buff[833+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[841 + i] = IXGBE_READ_REG(hw, IXGBE_RT2SR(i)); + sbuf_printf(sb, "\tIXGBE_RT2SR(%2d)\t %08x\n", i, regs_buff[841+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[849 + i] = IXGBE_READ_REG(hw, IXGBE_TDTQ2TCCR(i)); + sbuf_printf(sb, "\tIXGBE_TDTQ2TCCR(%2d)\t %08x\n", i, regs_buff[849+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[857 + i] = IXGBE_READ_REG(hw, IXGBE_TDTQ2TCSR(i)); + sbuf_printf(sb, "\tIXGBE_TDTQ2TCSR(%2d)\t %08x\n", i, regs_buff[857+i]); + } + break; + case ixgbe_mac_82599EB: + case ixgbe_mac_X540: + case ixgbe_mac_X550: + case ixgbe_mac_X550EM_x: + regs_buff[830] = IXGBE_READ_REG(hw, IXGBE_RTTDCS); + regs_buff[832] = IXGBE_READ_REG(hw, IXGBE_RTRPCS); + for (i = 0; i < adapter->num_queues; i++) + regs_buff[833 + i] = + IXGBE_READ_REG(hw, IXGBE_RTRPT4C(i)); + for (i = 0; i < adapter->num_queues; i++) + regs_buff[841 + i] = + IXGBE_READ_REG(hw, IXGBE_RTRPT4S(i)); + for (i = 0; i < adapter->num_queues; i++) + regs_buff[849 + i] = + IXGBE_READ_REG(hw, IXGBE_RTTDT2C(i)); + for (i = 0; i < adapter->num_queues; i++) + regs_buff[857 + i] = + IXGBE_READ_REG(hw, IXGBE_RTTDT2S(i)); + break; + default: + break; + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[865 + i] = IXGBE_READ_REG(hw, IXGBE_TDPT2TCCR(i)); /* same as RTTPT2C */ + sbuf_printf(sb, "\tIXGBE_TDPT2TCCR(%2d)\t %08x\n", i, regs_buff[865+i]); + } + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[873 + i] = IXGBE_READ_REG(hw, IXGBE_TDPT2TCSR(i)); /* same as RTTPT2S */ + sbuf_printf(sb, "\tIXGBE_TDPT2TCSR(%2d)\t %08x\n", i, regs_buff[873+i]); + } + /* MAC */ + sbuf_printf(sb, "\nMAC\n"); + regs_buff[1038] = IXGBE_READ_REG(hw, IXGBE_PCS1GCFIG); + sbuf_printf(sb, "\tIXGBE_PCS1GCFIG\t %08x\n", regs_buff[1038]); + regs_buff[1039] = IXGBE_READ_REG(hw, IXGBE_PCS1GLCTL); + sbuf_printf(sb, "\tIXGBE_PCSIG1CTL\t %08x\n", regs_buff[1039]); + regs_buff[1040] = IXGBE_READ_REG(hw, IXGBE_PCS1GLSTA); + sbuf_printf(sb, "\tIXGBE_PCSIGLSTA\t %08x\n", regs_buff[1040]); + regs_buff[1041] = IXGBE_READ_REG(hw, IXGBE_PCS1GDBG0); + sbuf_printf(sb, "\tIXGBE_PCS1GDBG0\t %08x\n", regs_buff[1041]); + regs_buff[1042] = IXGBE_READ_REG(hw, IXGBE_PCS1GDBG1); + sbuf_printf(sb, "\tIXGBE_PCS1GDBG1\t %08x\n", regs_buff[1042]); + regs_buff[1043] = IXGBE_READ_REG(hw, IXGBE_PCS1GANA); + sbuf_printf(sb, "\tIXGBE_PCS1GANA\t %08x\n", regs_buff[1043]); + regs_buff[1044] = IXGBE_READ_REG(hw, IXGBE_PCS1GANLP); + sbuf_printf(sb, "\tIXGBE_PCS1GANLP\t %08x\n", regs_buff[1044]); + regs_buff[1045] = IXGBE_READ_REG(hw, IXGBE_PCS1GANNP); + sbuf_printf(sb, "\tIXGBE_PCS1GANNP\t %08x\n", regs_buff[1045]); + regs_buff[1046] = IXGBE_READ_REG(hw, IXGBE_PCS1GANLPNP); + sbuf_printf(sb, "\tIXGBE_PCS1GANLPNP\t %08x\n", regs_buff[1046]); + regs_buff[1047] = IXGBE_READ_REG(hw, IXGBE_HLREG0); + sbuf_printf(sb, "\tIXGBE_HILREG0\t %08x\n", regs_buff[1047]); + regs_buff[1048] = IXGBE_READ_REG(hw, IXGBE_HLREG1); + sbuf_printf(sb, "\tIXGBE_HILREG1\t %08x\n", regs_buff[1048]); + regs_buff[1049] = IXGBE_READ_REG(hw, IXGBE_PAP); + sbuf_printf(sb, "\tIXGBE_PAP\t %08x\n", regs_buff[1049]); + regs_buff[1050] = IXGBE_READ_REG(hw, IXGBE_MACA); + sbuf_printf(sb, "\tIXGBE_MACA\t %08x\n", regs_buff[1050]); + regs_buff[1051] = IXGBE_READ_REG(hw, IXGBE_APAE); + sbuf_printf(sb, "\tIXGBE_APAE\t %08x\n", regs_buff[1051]); + regs_buff[1052] = IXGBE_READ_REG(hw, IXGBE_ARD); + sbuf_printf(sb, "\tIXGBE_ARD\t %08x\n", regs_buff[1052]); + regs_buff[1053] = IXGBE_READ_REG(hw, IXGBE_AIS); + sbuf_printf(sb, "\tIXGBE_AIS\t %08x\n", regs_buff[1053]); + regs_buff[1054] = IXGBE_READ_REG(hw, IXGBE_MSCA); + sbuf_printf(sb, "\tIXGBE_MSCA\t %08x\n", regs_buff[1054]); + regs_buff[1055] = IXGBE_READ_REG(hw, IXGBE_MSRWD); + sbuf_printf(sb, "\tIXGBE_MSRWD\t %08x\n", regs_buff[1055]); + regs_buff[1056] = IXGBE_READ_REG(hw, IXGBE_MLADD); + sbuf_printf(sb, "\tIXGBE_MLADD\t %08x\n", regs_buff[1056]); + regs_buff[1057] = IXGBE_READ_REG(hw, IXGBE_MHADD); + sbuf_printf(sb, "\tIXGBE_MHADD\t %08x\n", regs_buff[1057]); + regs_buff[1058] = IXGBE_READ_REG(hw, IXGBE_TREG); + sbuf_printf(sb, "\tIXGBE_TREG\t %08x\n", regs_buff[1058]); + regs_buff[1059] = IXGBE_READ_REG(hw, IXGBE_PCSS1); + sbuf_printf(sb, "\tIXGBE_PCSS1\t %08x\n", regs_buff[1059]); + regs_buff[1060] = IXGBE_READ_REG(hw, IXGBE_PCSS2); + sbuf_printf(sb, "\tIXGBE_PCSS2\t %08x\n", regs_buff[1060]); + regs_buff[1061] = IXGBE_READ_REG(hw, IXGBE_XPCSS); + sbuf_printf(sb, "\tIXGBE_XPCSS\t %08x\n", regs_buff[1061]); + regs_buff[1062] = IXGBE_READ_REG(hw, IXGBE_SERDESC); + sbuf_printf(sb, "\tIXGBE_SERDESC\t %08x\n", regs_buff[1062]); + regs_buff[1063] = IXGBE_READ_REG(hw, IXGBE_MACS); + sbuf_printf(sb, "\tIXGBE_MACS\t %08x\n", regs_buff[1063]); + regs_buff[1064] = IXGBE_READ_REG(hw, IXGBE_AUTOC); + sbuf_printf(sb, "\tIXGBE_AUTOC\t %08x\n", regs_buff[1064]); + regs_buff[1065] = IXGBE_READ_REG(hw, IXGBE_LINKS); + sbuf_printf(sb, "\tIXGBE_LINKS\t %08x\n", regs_buff[1065]); + regs_buff[1066] = IXGBE_READ_REG(hw, IXGBE_AUTOC2); + sbuf_printf(sb, "\tIXGBE_AUTOC2\t %08x\n", regs_buff[1066]); + regs_buff[1067] = IXGBE_READ_REG(hw, IXGBE_AUTOC3); + sbuf_printf(sb, "\tIXGBE_AUTOC3\t %08x\n", regs_buff[1067]); + regs_buff[1068] = IXGBE_READ_REG(hw, IXGBE_ANLP1); + sbuf_printf(sb, "\tIXGBE_ANLP1\t %08x\n", regs_buff[1068]); + regs_buff[1069] = IXGBE_READ_REG(hw, IXGBE_ANLP2); + sbuf_printf(sb, "\tIXGBE_ANLP2\t %08x\n", regs_buff[1069]); + regs_buff[1070] = IXGBE_READ_REG(hw, IXGBE_ATLASCTL); + sbuf_printf(sb, "\tIXGBE_ATLASCTL\t %08x\n", regs_buff[1070]); + + /* Diagnostic */ + sbuf_printf(sb, "\nDiagnostic\n"); + regs_buff[1071] = IXGBE_READ_REG(hw, IXGBE_RDSTATCTL); + sbuf_printf(sb, "\tIXGBE_RDSTATCTL\t %08x\n", regs_buff[1071]); + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[1072 + i] = IXGBE_READ_REG(hw, IXGBE_RDSTAT(i)); + sbuf_printf(sb, "\tIXGBE_RDSTAT(%2d)\t %08x\n", i, regs_buff[1072+i]); + } + regs_buff[1080] = IXGBE_READ_REG(hw, IXGBE_RDHMPN); + sbuf_printf(sb, "\tIXGBE_RDHMPN\t %08x\n", regs_buff[1080]); + for (i = 0; i < 4; i++) { + regs_buff[1081 + i] = IXGBE_READ_REG(hw, IXGBE_RIC_DW(i)); + sbuf_printf(sb, "\tIXGBE_RIC_DW(%2d)\t %08x\n", i, regs_buff[1081+i]); + } + regs_buff[1085] = IXGBE_READ_REG(hw, IXGBE_RDPROBE); + sbuf_printf(sb, "\tIXGBE_RDPROBE\t %08x\n", regs_buff[1085]); + /* regs_buff[1086] = IXGBE_READ_REG(hw, IXGBE_TDSTATCTL); + sbuf_printf(sb, "\tIXGBE_TDSTATCTL\t %08x\n", regs_buff[1086]); */ + /* for (i = 0; i < adapter->num_queues; i++) { + regs_buff[1087 + i] = IXGBE_READ_REG(hw, IXGBE_TDSTAT(i)); + sbuf_printf(sb, "\n\tIXGBE_TDSTAT(%2d)\t %08x\n", i, regs_buff[1087+i]); + } */ + regs_buff[1095] = IXGBE_READ_REG(hw, IXGBE_TDHMPN); + sbuf_printf(sb, "\tIXGBE_TDHMPN\t %08x\n", regs_buff[1095]); + + for (i = 0; i < 4; i++) { + regs_buff[1096 + i] = IXGBE_READ_REG(hw, IXGBE_TIC_DW(i)); + sbuf_printf(sb, "\tIXGBE_TIC_DW(%2d)\t %08x\n", i, regs_buff[1096+i]); + } + regs_buff[1100] = IXGBE_READ_REG(hw, IXGBE_TDPROBE); + sbuf_printf(sb, "\tIXGBE_TDPROBE\t %08x\n", regs_buff[1100]); + regs_buff[1101] = IXGBE_READ_REG(hw, IXGBE_TXBUFCTRL); + sbuf_printf(sb, "\tIXGBE_TXBUFCTRL\t %08x\n", regs_buff[1101]); + regs_buff[1102] = IXGBE_READ_REG(hw, IXGBE_TXBUFDATA0); + sbuf_printf(sb, "\tIXGBE_TXBUFDATA0\t %08x\n", regs_buff[1102]); + regs_buff[1103] = IXGBE_READ_REG(hw, IXGBE_TXBUFDATA1); + sbuf_printf(sb, "\tIXGBE_TXBUFDATA1\t %08x\n", regs_buff[1103]); + regs_buff[1104] = IXGBE_READ_REG(hw, IXGBE_TXBUFDATA2); + sbuf_printf(sb, "\tIXGBE_TXBUFDATA2\t %08x\n", regs_buff[1104]); + regs_buff[1105] = IXGBE_READ_REG(hw, IXGBE_TXBUFDATA3); + sbuf_printf(sb, "\tIXGBE_TXBUFDATA3\t %08x\n", regs_buff[1105]); + regs_buff[1106] = IXGBE_READ_REG(hw, IXGBE_RXBUFCTRL); + sbuf_printf(sb, "\tIXGBE_RXBUFCTRL\t %08x\n", regs_buff[1106]); + regs_buff[1107] = IXGBE_READ_REG(hw, IXGBE_RXBUFDATA0); + sbuf_printf(sb, "\tIXGBE_RXBIFDATA0\t %08x\n", regs_buff[1107]); + regs_buff[1108] = IXGBE_READ_REG(hw, IXGBE_RXBUFDATA1); + sbuf_printf(sb, "\tIXGBE_RXBUFDATA1\t %08x\n", regs_buff[1108]); + regs_buff[1109] = IXGBE_READ_REG(hw, IXGBE_RXBUFDATA2); + sbuf_printf(sb, "\tIXGBE_RXBUFDATA2\t %08x\n", regs_buff[1109]); + regs_buff[1110] = IXGBE_READ_REG(hw, IXGBE_RXBUFDATA3); + sbuf_printf(sb, "\tIXGBE_RXBUFDATA3\t %08x\n", regs_buff[1110]); + for (i = 0; i < adapter->num_queues; i++) { + regs_buff[1111 + i] = IXGBE_READ_REG(hw, IXGBE_PCIE_DIAG(i)); + sbuf_printf(sb, "\tIXGBE_PCIE_DIAG(%2d)\t %08x\n", i, regs_buff[1111+i]); + } + regs_buff[1119] = IXGBE_READ_REG(hw, IXGBE_RFVAL); + sbuf_printf(sb, "\tIXGBE_RFVAL\t %08x\n", regs_buff[1119]); + regs_buff[1120] = IXGBE_READ_REG(hw, IXGBE_MDFTC1); + sbuf_printf(sb, "\tIXGBE_MDFTC1\t %08x\n", regs_buff[1120]); + regs_buff[1121] = IXGBE_READ_REG(hw, IXGBE_MDFTC2); + sbuf_printf(sb, "\tIXGBE_MDFTC2\t %08x\n", regs_buff[1121]); + regs_buff[1122] = IXGBE_READ_REG(hw, IXGBE_MDFTFIFO1); + sbuf_printf(sb, "\tIXGBE_MDFTFIF01\t %08x\n", regs_buff[1122]); + regs_buff[1123] = IXGBE_READ_REG(hw, IXGBE_MDFTFIFO2); + sbuf_printf(sb, "\tIXGBE_MDFTFIF02\t %08x\n", regs_buff[1123]); + regs_buff[1124] = IXGBE_READ_REG(hw, IXGBE_MDFTS); + sbuf_printf(sb, "\tIXGBE_MDFTS\t %08x\n", regs_buff[1124]); + regs_buff[1125] = IXGBE_READ_REG(hw, IXGBE_PCIEECCCTL); + sbuf_printf(sb, "\tIXGBE_PCIEECCCTL\t %08x\n", regs_buff[1125]); + regs_buff[1126] = IXGBE_READ_REG(hw, IXGBE_PBTXECC); + sbuf_printf(sb, "\tIXGBE_PBTXECC\t %08x\n", regs_buff[1126]); + regs_buff[1127] = IXGBE_READ_REG(hw, IXGBE_PBRXECC); + sbuf_printf(sb, "\tIXGBE_PBRXECC\t %08x\n", regs_buff[1127]); + + /* 82599 X540 specific registers */ + regs_buff[1128] = IXGBE_READ_REG(hw, IXGBE_MFLCN); + + /* 82599 X540 specific DCB registers */ + regs_buff[1129] = IXGBE_READ_REG(hw, IXGBE_RTRUP2TC); + regs_buff[1130] = IXGBE_READ_REG(hw, IXGBE_RTTUP2TC); + for (i = 0; i < 4; i++) { + regs_buff[1131 + i] = IXGBE_READ_REG(hw, IXGBE_TXLLQ(i)); + sbuf_printf(sb, "\tIXGBE_TXLLQ(%2d)\t %08x\n", i, regs_buff[1131+i]); + } + regs_buff[1135] = IXGBE_READ_REG(hw, IXGBE_RTTBCNRM); + /* same as RTTQCNRM */ + sbuf_printf(sb, "\tIXGBE_RTTBCNRM\t %08x\n", regs_buff[1135]); + regs_buff[1136] = IXGBE_READ_REG(hw, IXGBE_RTTBCNRD); + /* same as RTTQCNRR */ + sbuf_printf(sb, "\tIXGBE_RTTBCNRD\t %08x\n", regs_buff[1136]); + + for (j = 0; j < 64; j++) { + u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error); + u32 length = le32toh(rxr->rx_base[j].wb.upper.length); + sbuf_printf(sb, "\tReceive Descriptor Address %d: %08lx Error:%d Length:%d\n", j, rxr->rx_base[j].read.pkt_addr, staterr, length); + } + + for (j = 0; j < 64; j++) { + struct ixgbe_tx_buf *buf = &txr->tx_buffers[j]; + u64 buffer_addr = txr->tx_base[j].read.buffer_addr; + u32 cmd_type_len = txr->tx_base[j].read.cmd_type_len; + u32 olinfo_status = txr->tx_base[j].read.olinfo_status; + sbuf_printf(sb, "\tTXD[%d] addr: %08lx Cmd Len:%d olinfo_status:%d eop: %p DD=%d\n", + j, buffer_addr, cmd_type_len, olinfo_status, buf->eop, + buf->eop != NULL ? buf->eop->wb.status & IXGBE_TXD_STAT_DD : 0); + } + /* X540 specific DCB registers + regs_buff[1137] = IXGBE_READ_REG(hw, IXGBE_RTTQCNCR); + regs_buff[1138] = IXGBE_READ_REG(hw, IXGBE_RTTQCNTG); */ + + rc = sbuf_finish(sb); + sbuf_delete(sb); + return(rc); +} + Index: sys/modules/ix/Makefile =================================================================== --- sys/modules/ix/Makefile +++ sys/modules/ix/Makefile @@ -5,7 +5,7 @@ KMOD = if_ix SRCS = device_if.h bus_if.h pci_if.h pci_iov_if.h SRCS += opt_inet.h opt_inet6.h opt_rss.h -SRCS += if_ix.c ix_txrx.c ixgbe_osdep.c +SRCS += if_ix.c ix_txrx.c ixgbe_sysctl.c ixgbe_osdep.c ifdi_if.h # Shared source SRCS += ixgbe_common.c ixgbe_api.c ixgbe_phy.c ixgbe_mbx.c ixgbe_vf.c SRCS += ixgbe_dcb.c ixgbe_dcb_82598.c ixgbe_dcb_82599.c