X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/c0fea4742e91338fffdcf79f86a7c1d5e2b97eb1..490019cf9519204c5fb36b2fba54ceb983bb6b72:/bsd/net/if_var.h diff --git a/bsd/net/if_var.h b/bsd/net/if_var.h index 6feecc2b9..21066e652 100644 --- a/bsd/net/if_var.h +++ b/bsd/net/if_var.h @@ -1,23 +1,29 @@ /* - * Copyright (c) 2000 Apple Computer, Inc. All rights reserved. + * Copyright (c) 2000-2015 Apple Inc. All rights reserved. * - * @APPLE_LICENSE_HEADER_START@ - * - * The contents of this file constitute Original Code as defined in and - * are subject to the Apple Public Source License Version 1.1 (the - * "License"). You may not use this file except in compliance with the - * License. Please obtain a copy of the License at - * http://www.apple.com/publicsource and read it before using this file. - * - * This Original Code and all software distributed under the License are - * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER + * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ + * + * This file contains Original Code and/or Modifications of Original Code + * as defined in and that are subject to the Apple Public Source License + * Version 2.0 (the 'License'). You may not use this file except in + * compliance with the License. The rights granted to you under the License + * may not be used to create, or enable the creation or redistribution of, + * unlawful or unlicensed copies of an Apple operating system, or to + * circumvent, violate, or enable the circumvention or violation of, any + * terms of an Apple operating system software license agreement. + * + * Please obtain a copy of the License at + * http://www.opensource.apple.com/apsl/ and read it before using this file. + * + * The Original Code and all software distributed under the License are + * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the - * License for the specific language governing rights and limitations - * under the License. - * - * @APPLE_LICENSE_HEADER_END@ + * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. + * Please see the License for the specific language governing rights and + * limitations under the License. + * + * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ */ /* * Copyright (c) 1982, 1986, 1989, 1993 @@ -59,16 +65,20 @@ #define _NET_IF_VAR_H_ #include +#include #include #include #include /* get TAILQ macros */ #ifdef KERNEL_PRIVATE #include #endif /* KERNEL_PRIVATE */ +#ifdef PRIVATE +#include +#endif #ifdef KERNEL #include -#endif KERNEL +#endif /* KERNEL */ #ifdef __APPLE__ #define APPLE_IF_FAM_LOOPBACK 1 @@ -81,7 +91,7 @@ #define APPLE_IF_FAM_DISC 8 #define APPLE_IF_FAM_MDECAP 9 #define APPLE_IF_FAM_GIF 10 -#define APPLE_IF_FAM_FAITH 11 +#define APPLE_IF_FAM_FAITH 11 /* deprecated */ #define APPLE_IF_FAM_STF 12 #define APPLE_IF_FAM_FIREWIRE 13 #define APPLE_IF_FAM_BOND 14 @@ -126,49 +136,56 @@ */ struct net_event_data { - unsigned long if_family; - unsigned long if_unit; - char if_name[IFNAMSIZ]; + u_int32_t if_family; + u_int32_t if_unit; + char if_name[IFNAMSIZ]; }; +#if defined(__LP64__) +#include +#define IF_DATA_TIMEVAL timeval32 +#else +#define IF_DATA_TIMEVAL timeval +#endif + +#pragma pack(4) + /* * Structure describing information about an interface * which may be of interest to management entities. */ struct if_data { /* generic interface information */ - unsigned char ifi_type; /* ethernet, tokenring, etc */ -#ifdef __APPLE__ - unsigned char ifi_typelen; /* Length of frame type id */ -#endif - unsigned char ifi_physical; /* e.g., AUI, Thinnet, 10base-T, etc */ - unsigned char ifi_addrlen; /* media address length */ - unsigned char ifi_hdrlen; /* media header length */ - unsigned char ifi_recvquota; /* polling quota for receive intrs */ - unsigned char ifi_xmitquota; /* polling quota for xmit intrs */ - unsigned char ifi_unused1; /* for future use */ - unsigned long ifi_mtu; /* maximum transmission unit */ - unsigned long ifi_metric; /* routing metric (external only) */ - unsigned long ifi_baudrate; /* linespeed */ + u_char ifi_type; /* ethernet, tokenring, etc */ + u_char ifi_typelen; /* Length of frame type id */ + u_char ifi_physical; /* e.g., AUI, Thinnet, 10base-T, etc */ + u_char ifi_addrlen; /* media address length */ + u_char ifi_hdrlen; /* media header length */ + u_char ifi_recvquota; /* polling quota for receive intrs */ + u_char ifi_xmitquota; /* polling quota for xmit intrs */ + u_char ifi_unused1; /* for future use */ + u_int32_t ifi_mtu; /* maximum transmission unit */ + u_int32_t ifi_metric; /* routing metric (external only) */ + u_int32_t ifi_baudrate; /* linespeed */ /* volatile statistics */ - unsigned long ifi_ipackets; /* packets received on interface */ - unsigned long ifi_ierrors; /* input errors on interface */ - unsigned long ifi_opackets; /* packets sent on interface */ - unsigned long ifi_oerrors; /* output errors on interface */ - unsigned long ifi_collisions; /* collisions on csma interfaces */ - unsigned long ifi_ibytes; /* total number of octets received */ - unsigned long ifi_obytes; /* total number of octets sent */ - unsigned long ifi_imcasts; /* packets received via multicast */ - unsigned long ifi_omcasts; /* packets sent via multicast */ - unsigned long ifi_iqdrops; /* dropped on input, this interface */ - unsigned long ifi_noproto; /* destined for unsupported protocol */ - unsigned long ifi_recvtiming; /* usec spent receiving when timing */ - unsigned long ifi_xmittiming; /* usec spent xmitting when timing */ - struct timeval ifi_lastchange; /* time of last administrative change */ - unsigned long ifi_unused2; /* used to be the default_proto */ - unsigned long ifi_hwassist; /* HW offload capabilities */ - unsigned long ifi_reserved1; /* for future use */ - unsigned long ifi_reserved2; /* for future use */ + u_int32_t ifi_ipackets; /* packets received on interface */ + u_int32_t ifi_ierrors; /* input errors on interface */ + u_int32_t ifi_opackets; /* packets sent on interface */ + u_int32_t ifi_oerrors; /* output errors on interface */ + u_int32_t ifi_collisions; /* collisions on csma interfaces */ + u_int32_t ifi_ibytes; /* total number of octets received */ + u_int32_t ifi_obytes; /* total number of octets sent */ + u_int32_t ifi_imcasts; /* packets received via multicast */ + u_int32_t ifi_omcasts; /* packets sent via multicast */ + u_int32_t ifi_iqdrops; /* dropped on input, this interface */ + u_int32_t ifi_noproto; /* destined for unsupported protocol */ + u_int32_t ifi_recvtiming; /* usec spent receiving when timing */ + u_int32_t ifi_xmittiming; /* usec spent xmitting when timing */ + struct IF_DATA_TIMEVAL ifi_lastchange; /* time of last administrative change */ + u_int32_t ifi_unused2; /* used to be the default_proto */ + u_int32_t ifi_hwassist; /* HW offload capabilities */ + u_int32_t ifi_reserved1; /* for future use */ + u_int32_t ifi_reserved2; /* for future use */ }; /* @@ -177,18 +194,16 @@ struct if_data { */ struct if_data64 { /* generic interface information */ - u_char ifi_type; /* ethernet, tokenring, etc */ -#ifdef __APPLE__ - u_char ifi_typelen; /* Length of frame type id */ -#endif + u_char ifi_type; /* ethernet, tokenring, etc */ + u_char ifi_typelen; /* Length of frame type id */ u_char ifi_physical; /* e.g., AUI, Thinnet, 10base-T, etc */ u_char ifi_addrlen; /* media address length */ u_char ifi_hdrlen; /* media header length */ u_char ifi_recvquota; /* polling quota for receive intrs */ u_char ifi_xmitquota; /* polling quota for xmit intrs */ u_char ifi_unused1; /* for future use */ - u_long ifi_mtu; /* maximum transmission unit */ - u_long ifi_metric; /* routing metric (external only) */ + u_int32_t ifi_mtu; /* maximum transmission unit */ + u_int32_t ifi_metric; /* routing metric (external only) */ u_int64_t ifi_baudrate; /* linespeed */ /* volatile statistics */ u_int64_t ifi_ipackets; /* packets received on interface */ @@ -202,55 +217,443 @@ struct if_data64 { u_int64_t ifi_omcasts; /* packets sent via multicast */ u_int64_t ifi_iqdrops; /* dropped on input, this interface */ u_int64_t ifi_noproto; /* destined for unsupported protocol */ - u_long ifi_recvtiming; /* usec spent receiving when timing */ - u_long ifi_xmittiming; /* usec spent xmitting when timing */ - struct timeval ifi_lastchange; /* time of last administrative change */ + u_int32_t ifi_recvtiming; /* usec spent receiving when timing */ + u_int32_t ifi_xmittiming; /* usec spent xmitting when timing */ + struct IF_DATA_TIMEVAL ifi_lastchange; /* time of last administrative change */ }; #ifdef PRIVATE +struct if_traffic_class { + u_int64_t ifi_ibepackets; /* TC_BE packets received on interface */ + u_int64_t ifi_ibebytes; /* TC_BE bytes received on interface */ + u_int64_t ifi_obepackets; /* TC_BE packet sent on interface */ + u_int64_t ifi_obebytes; /* TC_BE bytes sent on interface */ + u_int64_t ifi_ibkpackets; /* TC_BK packets received on interface */ + u_int64_t ifi_ibkbytes; /* TC_BK bytes received on interface */ + u_int64_t ifi_obkpackets; /* TC_BK packet sent on interface */ + u_int64_t ifi_obkbytes; /* TC_BK bytes sent on interface */ + u_int64_t ifi_ivipackets; /* TC_VI packets received on interface */ + u_int64_t ifi_ivibytes; /* TC_VI bytes received on interface */ + u_int64_t ifi_ovipackets; /* TC_VI packets sent on interface */ + u_int64_t ifi_ovibytes; /* TC_VI bytes sent on interface */ + u_int64_t ifi_ivopackets; /* TC_VO packets received on interface */ + u_int64_t ifi_ivobytes; /* TC_VO bytes received on interface */ + u_int64_t ifi_ovopackets; /* TC_VO packets sent on interface */ + u_int64_t ifi_ovobytes; /* TC_VO bytes sent on interface */ + u_int64_t ifi_ipvpackets; /* TC priv packets received on interface */ + u_int64_t ifi_ipvbytes; /* TC priv bytes received on interface */ + u_int64_t ifi_opvpackets; /* TC priv packets sent on interface */ + u_int64_t ifi_opvbytes; /* TC priv bytes sent on interface */ +}; + +struct if_data_extended { + u_int64_t ifi_alignerrs; /* unaligned (32-bit) input pkts */ + u_int64_t ifi_dt_bytes; /* Data threshold counter */ + u_int64_t ifi_fpackets; /* forwarded packets on interface */ + u_int64_t ifi_fbytes; /* forwarded bytes on interface */ + u_int64_t reserved[12]; /* for future */ +}; + +struct if_packet_stats { + /* TCP */ + u_int64_t ifi_tcp_badformat; + u_int64_t ifi_tcp_unspecv6; + u_int64_t ifi_tcp_synfin; + u_int64_t ifi_tcp_badformatipsec; + u_int64_t ifi_tcp_noconnnolist; + u_int64_t ifi_tcp_noconnlist; + u_int64_t ifi_tcp_listbadsyn; + u_int64_t ifi_tcp_icmp6unreach; + u_int64_t ifi_tcp_deprecate6; + u_int64_t ifi_tcp_rstinsynrcv; + u_int64_t ifi_tcp_ooopacket; + u_int64_t ifi_tcp_dospacket; + u_int64_t ifi_tcp_cleanup; + u_int64_t ifi_tcp_synwindow; + u_int64_t reserved[6]; + /* UDP */ + u_int64_t ifi_udp_port_unreach; + u_int64_t ifi_udp_faithprefix; + u_int64_t ifi_udp_port0; + u_int64_t ifi_udp_badlength; + u_int64_t ifi_udp_badchksum; + u_int64_t ifi_udp_badmcast; + u_int64_t ifi_udp_cleanup; + u_int64_t ifi_udp_badipsec; + u_int64_t _reserved[4]; +}; + +struct if_description { + u_int32_t ifd_maxlen; /* must be IF_DESCSIZE */ + u_int32_t ifd_len; /* actual ifd_desc length */ + u_int8_t *ifd_desc; /* ptr to desc buffer */ +}; + +struct if_bandwidths { + u_int64_t eff_bw; /* effective bandwidth */ + u_int64_t max_bw; /* maximum theoretical bandwidth */ +}; + +struct if_latencies { + u_int64_t eff_lt; /* effective latency */ + u_int64_t max_lt; /* maximum theoretical latency */ +}; + +struct if_rxpoll_stats { + u_int32_t ifi_poll_off_req; /* total # of POLL_OFF reqs */ + u_int32_t ifi_poll_off_err; /* total # of POLL_OFF errors */ + u_int32_t ifi_poll_on_req; /* total # of POLL_ON reqs */ + u_int32_t ifi_poll_on_err; /* total # of POLL_ON errors */ + + u_int32_t ifi_poll_wakeups_avg; /* avg # of wakeup reqs */ + u_int32_t ifi_poll_wakeups_lowat; /* wakeups low watermark */ + u_int32_t ifi_poll_wakeups_hiwat; /* wakeups high watermark */ + + u_int64_t ifi_poll_packets; /* total # of polled packets */ + u_int32_t ifi_poll_packets_avg; /* average polled packets */ + u_int32_t ifi_poll_packets_min; /* smallest polled packets */ + u_int32_t ifi_poll_packets_max; /* largest polled packets */ + u_int32_t ifi_poll_packets_lowat; /* packets low watermark */ + u_int32_t ifi_poll_packets_hiwat; /* packets high watermark */ + + u_int64_t ifi_poll_bytes; /* total # of polled bytes */ + u_int32_t ifi_poll_bytes_avg; /* average polled bytes */ + u_int32_t ifi_poll_bytes_min; /* smallest polled bytes */ + u_int32_t ifi_poll_bytes_max; /* largest polled bytes */ + u_int32_t ifi_poll_bytes_lowat; /* bytes low watermark */ + u_int32_t ifi_poll_bytes_hiwat; /* bytes high watermark */ + + u_int32_t ifi_poll_packets_limit; /* max packets per poll call */ + u_int64_t ifi_poll_interval_time; /* poll interval (nsec) */ +}; + +struct if_tcp_ecn_perf_stat { + u_int64_t total_txpkts; + u_int64_t total_rxmitpkts; + u_int64_t total_rxpkts; + u_int64_t total_oopkts; + u_int64_t total_reorderpkts; + u_int64_t rtt_avg; + u_int64_t rtt_var; + u_int64_t sack_episodes; + u_int64_t rxmit_drop; + u_int64_t rst_drop; + u_int64_t oo_percent; + u_int64_t reorder_percent; + u_int64_t rxmit_percent; +}; + +struct if_tcp_ecn_stat { + u_int64_t timestamp; + u_int64_t ecn_client_setup; + u_int64_t ecn_server_setup; + u_int64_t ecn_client_success; + u_int64_t ecn_server_success; + u_int64_t ecn_peer_nosupport; + u_int64_t ecn_syn_lost; + u_int64_t ecn_synack_lost; + u_int64_t ecn_recv_ce; + u_int64_t ecn_recv_ece; + u_int64_t ecn_conn_recv_ce; + u_int64_t ecn_conn_recv_ece; + u_int64_t ecn_conn_plnoce; + u_int64_t ecn_conn_plce; + u_int64_t ecn_conn_noplce; + u_int64_t ecn_fallback_synloss; + u_int64_t ecn_fallback_reorder; + u_int64_t ecn_fallback_ce; + u_int64_t ecn_off_conn; + u_int64_t ecn_total_conn; + struct if_tcp_ecn_perf_stat ecn_on; + struct if_tcp_ecn_perf_stat ecn_off; +}; + +/* + * Interface link status report -- includes statistics related to + * the link layer technology sent by the driver. The driver will monitor + * these statistics over an interval (3-4 secs) and will generate a report + * to the network stack. This will give first-hand information about the + * status of the first hop of the network path. The version and + * length values should be correct for the data to be processed correctly. + * The definitions are different for different kind of interfaces like + * Wifi, Cellular etc,. + */ +#define IF_CELLULAR_STATUS_REPORT_VERSION_1 1 +#define IF_WIFI_STATUS_REPORT_VERSION_1 1 +#define IF_CELLULAR_STATUS_REPORT_CURRENT_VERSION \ + IF_CELLULAR_STATUS_REPORT_VERSION_1 +#define IF_WIFI_STATUS_REPORT_CURRENT_VERSION IF_WIFI_STATUS_REPORT_VERSION_1 +/* + * For cellular interface -- + * There is no way to share common headers between the Baseband and + * the kernel. Any changes to this structure will need to be communicated + * to the Baseband team. It is better to use reserved space instead of + * changing the size or existing fields in the structure. + */ +struct if_cellular_status_v1 { + u_int32_t valid_bitmask; /* indicates which fields are valid */ +#define IF_CELL_LINK_QUALITY_METRIC_VALID 0x1 +#define IF_CELL_UL_EFFECTIVE_BANDWIDTH_VALID 0x2 +#define IF_CELL_UL_MAX_BANDWIDTH_VALID 0x4 +#define IF_CELL_UL_MIN_LATENCY_VALID 0x8 +#define IF_CELL_UL_EFFECTIVE_LATENCY_VALID 0x10 +#define IF_CELL_UL_MAX_LATENCY_VALID 0x20 +#define IF_CELL_UL_RETXT_LEVEL_VALID 0x40 +#define IF_CELL_UL_BYTES_LOST_VALID 0x80 +#define IF_CELL_UL_MIN_QUEUE_SIZE_VALID 0x100 +#define IF_CELL_UL_AVG_QUEUE_SIZE_VALID 0x200 +#define IF_CELL_UL_MAX_QUEUE_SIZE_VALID 0x400 +#define IF_CELL_DL_EFFECTIVE_BANDWIDTH_VALID 0x800 +#define IF_CELL_DL_MAX_BANDWIDTH_VALID 0x1000 +#define IF_CELL_CONFIG_INACTIVITY_TIME_VALID 0x2000 +#define IF_CELL_CONFIG_BACKOFF_TIME_VALID 0x4000 + u_int32_t link_quality_metric; + u_int32_t ul_effective_bandwidth; /* Measured uplink bandwidth based on current activity (bps) */ + u_int32_t ul_max_bandwidth; /* Maximum supported uplink bandwidth (bps) */ + u_int32_t ul_min_latency; /* min expected uplink latency for first hop (ms) */ + u_int32_t ul_effective_latency; /* current expected uplink latency for first hop (ms) */ + u_int32_t ul_max_latency; /* max expected uplink latency first hop (ms) */ + u_int32_t ul_retxt_level; /* Retransmission metric */ +#define IF_CELL_UL_RETXT_LEVEL_NONE 1 +#define IF_CELL_UL_RETXT_LEVEL_LOW 2 +#define IF_CELL_UL_RETXT_LEVEL_MEDIUM 3 +#define IF_CELL_UL_RETXT_LEVEL_HIGH 4 + u_int32_t ul_bytes_lost; /* % of total bytes lost on uplink in Q10 format */ + u_int32_t ul_min_queue_size; /* minimum bytes in queue */ + u_int32_t ul_avg_queue_size; /* average bytes in queue */ + u_int32_t ul_max_queue_size; /* maximum bytes in queue */ + u_int32_t dl_effective_bandwidth; /* Measured downlink bandwidth based on current activity (bps) */ + u_int32_t dl_max_bandwidth; /* Maximum supported downlink bandwidth (bps) */ + u_int32_t config_inactivity_time; /* ms */ + u_int32_t config_backoff_time; /* new connections backoff time in ms */ + u_int64_t reserved_1; + u_int64_t reserved_2; + u_int64_t reserved_3; + u_int64_t reserved_4; + u_int64_t reserved_5; +} __attribute__((packed)); + +struct if_cellular_status { + union { + struct if_cellular_status_v1 if_status_v1; + } if_cell_u; +}; + +/* + * These statistics will be provided by the Wifi driver periodically. + * After sending each report, the driver should start computing again + * for the next report duration so that the values represent the link + * status for one report duration. + */ + +struct if_wifi_status_v1 { + u_int32_t valid_bitmask; +#define IF_WIFI_LINK_QUALITY_METRIC_VALID 0x1 +#define IF_WIFI_UL_EFFECTIVE_BANDWIDTH_VALID 0x2 +#define IF_WIFI_UL_MAX_BANDWIDTH_VALID 0x4 +#define IF_WIFI_UL_MIN_LATENCY_VALID 0x8 +#define IF_WIFI_UL_EFFECTIVE_LATENCY_VALID 0x10 +#define IF_WIFI_UL_MAX_LATENCY_VALID 0x20 +#define IF_WIFI_UL_RETXT_LEVEL_VALID 0x40 +#define IF_WIFI_UL_ERROR_RATE_VALID 0x80 +#define IF_WIFI_UL_BYTES_LOST_VALID 0x100 +#define IF_WIFI_DL_EFFECTIVE_BANDWIDTH_VALID 0x200 +#define IF_WIFI_DL_MAX_BANDWIDTH_VALID 0x400 +#define IF_WIFI_DL_MIN_LATENCY_VALID 0x800 +#define IF_WIFI_DL_EFFECTIVE_LATENCY_VALID 0x1000 +#define IF_WIFI_DL_MAX_LATENCY_VALID 0x2000 +#define IF_WIFI_DL_ERROR_RATE_VALID 0x4000 +#define IF_WIFI_CONFIG_FREQUENCY_VALID 0x8000 +#define IF_WIFI_CONFIG_MULTICAST_RATE_VALID 0x10000 +#define IF_WIFI_CONFIG_SCAN_COUNT_VALID 0x20000 +#define IF_WIFI_CONFIG_SCAN_DURATION_VALID 0x40000 + u_int32_t link_quality_metric; /* link quality metric */ + u_int32_t ul_effective_bandwidth; /* Measured uplink bandwidth based on current activity (bps) */ + u_int32_t ul_max_bandwidth; /* Maximum supported uplink bandwidth (bps) */ + u_int32_t ul_min_latency; /* min expected uplink latency for first hop (ms) */ + u_int32_t ul_effective_latency; /* current expected uplink latency for first hop (ms) */ + u_int32_t ul_max_latency; /* max expected uplink latency for first hop (ms) */ + u_int32_t ul_retxt_level; /* Retransmission metric */ +#define IF_WIFI_UL_RETXT_LEVEL_NONE 1 +#define IF_WIFI_UL_RETXT_LEVEL_LOW 2 +#define IF_WIFI_UL_RETXT_LEVEL_MEDIUM 3 +#define IF_WIFI_UL_RETXT_LEVEL_HIGH 4 + u_int32_t ul_bytes_lost; /* % of total bytes lost on uplink in Q10 format */ + u_int32_t ul_error_rate; /* % of bytes dropped on uplink after many retransmissions in Q10 format */ + u_int32_t dl_effective_bandwidth; /* Measured downlink bandwidth based on current activity (bps) */ + u_int32_t dl_max_bandwidth; /* Maximum supported downlink bandwidth (bps) */ + /* + * The download latency values indicate the time AP may have to wait for the + * driver to receive the packet. These values give the range of expected latency + * mainly due to co-existence events and channel hopping where the interface + * becomes unavailable. + */ + u_int32_t dl_min_latency; /* min expected latency for first hop in ms */ + u_int32_t dl_effective_latency; /* current expected latency for first hop in ms */ + u_int32_t dl_max_latency; /* max expected latency for first hop in ms */ + u_int32_t dl_error_rate; /* % of CRC or other errors in Q10 format */ + u_int32_t config_frequency; /* 2.4 or 5 GHz */ +#define IF_WIFI_CONFIG_FREQUENCY_2_4_GHZ 1 +#define IF_WIFI_CONFIG_FREQUENCY_5_0_GHZ 2 + u_int32_t config_multicast_rate; /* bps */ + u_int32_t scan_count; /* scan count during the previous period */ + u_int32_t scan_duration; /* scan duration in ms */ + u_int64_t reserved_1; + u_int64_t reserved_2; + u_int64_t reserved_3; + u_int64_t reserved_4; +} __attribute__((packed)); + +struct if_wifi_status { + union { + struct if_wifi_status_v1 if_status_v1; + } if_wifi_u; +}; + +struct if_link_status { + u_int32_t ifsr_version; /* version of this report */ + u_int32_t ifsr_len; /* length of the following struct */ + union { + struct if_cellular_status ifsr_cell; + struct if_wifi_status ifsr_wifi; + } ifsr_u; +}; + +struct if_interface_state { + /* + * The bitmask tells which of the fields + * to consider: + * - When setting, to control which fields + * are being modified; + * - When getting, it tells which fields are set. + */ + u_int8_t valid_bitmask; +#define IF_INTERFACE_STATE_RRC_STATE_VALID 0x1 +#define IF_INTERFACE_STATE_LQM_STATE_VALID 0x2 +#define IF_INTERFACE_STATE_INTERFACE_AVAILABILITY_VALID 0x4 + + /* + * Valid only for cellular interface + */ + u_int8_t rrc_state; +#define IF_INTERFACE_STATE_RRC_STATE_IDLE 0x0 +#define IF_INTERFACE_STATE_RRC_STATE_CONNECTED 0x1 + + /* + * Values normalized to the edge of the following values + * that are defined on : + * IFNET_LQM_THRESH_BAD + * IFNET_LQM_THRESH_POOR + * IFNET_LQM_THRESH_GOOD + */ + int8_t lqm_state; + + /* + * Indicate if the underlying link is currently + * available + */ + u_int8_t interface_availability; +#define IF_INTERFACE_STATE_INTERFACE_AVAILABLE 0x0 +#define IF_INTERFACE_STATE_INTERFACE_UNAVAILABLE 0x1 +}; + +struct chain_len_stats { + uint64_t cls_one; + uint64_t cls_two; + uint64_t cls_three; + uint64_t cls_four; + uint64_t cls_five_or_more; +}; + +#endif /* PRIVATE */ + +#pragma pack() + +/* + * Structure defining a queue for a network interface. + */ +struct ifqueue { + void *ifq_head; + void *ifq_tail; + int ifq_len; + int ifq_maxlen; + int ifq_drops; +}; + +#ifdef BSD_KERNEL_PRIVATE /* * Internal storage of if_data. This is bound to change. Various places in the * stack will translate this data structure in to the externally visible - * if_data structure above. + * if_data structure above. Note that during interface attach time, the + * embedded if_data structure in ifnet is cleared, with the exception of + * some non-statistics related fields. */ struct if_data_internal { /* generic interface information */ - u_char ifi_type; /* ethernet, tokenring, etc */ - u_char ifi_typelen; /* Length of frame type id */ - u_char ifi_physical; /* e.g., AUI, Thinnet, 10base-T, etc */ - u_char ifi_addrlen; /* media address length */ - u_char ifi_hdrlen; /* media header length */ - u_char ifi_recvquota; /* polling quota for receive intrs */ - u_char ifi_xmitquota; /* polling quota for xmit intrs */ - u_char ifi_unused1; /* for future use */ - u_long ifi_mtu; /* maximum transmission unit */ - u_long ifi_metric; /* routing metric (external only) */ - u_long ifi_baudrate; /* linespeed */ + u_char ifi_type; /* ethernet, tokenring, etc */ + u_char ifi_typelen; /* Length of frame type id */ + u_char ifi_physical; /* e.g., AUI, Thinnet, 10base-T, etc */ + u_char ifi_addrlen; /* media address length */ + u_char ifi_hdrlen; /* media header length */ + u_char ifi_recvquota; /* polling quota for receive intrs */ + u_char ifi_xmitquota; /* polling quota for xmit intrs */ + u_char ifi_unused1; /* for future use */ + u_int32_t ifi_mtu; /* maximum transmission unit */ + u_int32_t ifi_metric; /* routing metric (external only) */ + u_int32_t ifi_baudrate; /* linespeed */ + u_int32_t ifi_preamblelen;/* length of the packet preamble */ /* volatile statistics */ u_int64_t ifi_ipackets; /* packets received on interface */ u_int64_t ifi_ierrors; /* input errors on interface */ u_int64_t ifi_opackets; /* packets sent on interface */ u_int64_t ifi_oerrors; /* output errors on interface */ u_int64_t ifi_collisions; /* collisions on csma interfaces */ - u_int64_t ifi_ibytes; /* total number of octets received */ - u_int64_t ifi_obytes; /* total number of octets sent */ + u_int64_t ifi_ibytes; /* total number of octets received */ + u_int64_t ifi_obytes; /* total number of octets sent */ u_int64_t ifi_imcasts; /* packets received via multicast */ u_int64_t ifi_omcasts; /* packets sent via multicast */ u_int64_t ifi_iqdrops; /* dropped on input, this interface */ u_int64_t ifi_noproto; /* destined for unsupported protocol */ - u_long ifi_recvtiming; /* usec spent receiving when timing */ - u_long ifi_xmittiming; /* usec spent xmitting when timing */ -#define IF_LASTCHANGEUPTIME 1 /* lastchange: 1-uptime 0-calendar time */ + u_int32_t ifi_recvtiming; /* usec spent receiving when timing */ + u_int32_t ifi_xmittiming; /* usec spent xmitting when timing */ + u_int64_t ifi_alignerrs; /* unaligned (32-bit) input pkts */ + u_int64_t ifi_dt_bytes; /* Data threshold counter */ + u_int64_t ifi_fpackets; /* forwarded packets on interface */ + u_int64_t ifi_fbytes; /* forwarded bytes on interface */ struct timeval ifi_lastchange; /* time of last administrative change */ - u_long ifi_hwassist; /* HW offload capabilities */ + u_int32_t ifi_hwassist; /* HW offload capabilities */ + u_int32_t ifi_tso_v4_mtu; /* TCP Segment Offload IPv4 maximum segment size */ + u_int32_t ifi_tso_v6_mtu; /* TCP Segment Offload IPv6 maximum segment size */ +}; + +#if MEASURE_BW +/* + * Fields per interface to measure perceived bandwidth. + */ +struct if_measured_bw { + u_int64_t bw; /* measured bandwidth in bytes per ms */ + u_int64_t bytes; /* XXX not needed */ + u_int64_t ts; /* XXX not needed */ + u_int64_t cur_seq __attribute((aligned(8))); /* current sequence for marking a packet */ + u_int64_t start_ts; /* time at which a measurement started */ + u_int64_t start_seq; /* sequence at which a measurement should start */ + u_int64_t last_seq; /* last recorded seq */ + u_int64_t last_ts; /* last recorded ts */ + u_int32_t flags __attribute__((aligned(4))); /* flags */ +#define IF_MEASURED_BW_INPROGRESS 0x1 +#define IF_MEASURED_BW_CALCULATION 0x2 }; +#endif /* MEASURE_BW */ +#endif /* BSD_KERNEL_PRIVATE */ +#ifdef PRIVATE #define if_mtu if_data.ifi_mtu #define if_type if_data.ifi_type #define if_typelen if_data.ifi_typelen #define if_physical if_data.ifi_physical #define if_addrlen if_data.ifi_addrlen #define if_hdrlen if_data.ifi_hdrlen +#define if_preamblelen if_data.ifi_preamblelen #define if_metric if_data.ifi_metric #define if_baudrate if_data.ifi_baudrate #define if_hwassist if_data.ifi_hwassist @@ -268,273 +671,401 @@ struct if_data_internal { #define if_lastchange if_data.ifi_lastchange #define if_recvquota if_data.ifi_recvquota #define if_xmitquota if_data.ifi_xmitquota -#define if_iflags if_data.ifi_iflags - -struct mbuf; -struct ifaddr; -TAILQ_HEAD(ifnethead, ifnet); /* we use TAILQs so that the order of */ -TAILQ_HEAD(ifaddrhead, ifaddr); /* instantiation is preserved in the list */ -TAILQ_HEAD(ifprefixhead, ifprefix); -LIST_HEAD(ifmultihead, ifmultiaddr); -struct tqdummy; -TAILQ_HEAD(tailq_head, tqdummy); +#endif /* PRIVATE */ +#ifdef BSD_KERNEL_PRIVATE +#define if_tso_v4_mtu if_data.ifi_tso_v4_mtu +#define if_tso_v6_mtu if_data.ifi_tso_v6_mtu +#define if_alignerrs if_data.ifi_alignerrs +#define if_dt_bytes if_data.ifi_dt_bytes +#define if_fpackets if_data.ifi_fpackets +#define if_fbytes if_data.ifi_fbytes +#endif /* BSD_KERNEL_PRIVATE */ +#ifdef BSD_KERNEL_PRIVATE /* * Forward structure declarations for function prototypes [sic]. */ -struct proc; -struct rtentry; -struct socket; -struct ether_header; -struct sockaddr_dl; +struct proc; +struct rtentry; +struct socket; struct ifnet_filter; +struct mbuf; +struct ifaddr; +struct tqdummy; +struct proto_hash_entry; +struct dlil_threading_info; +struct tcpstat_local; +struct udpstat_local; +#if PF +struct pfi_kif; +#endif /* PF */ +/* we use TAILQs so that the order of instantiation is preserved in the list */ +TAILQ_HEAD(ifnethead, ifnet); +TAILQ_HEAD(ifaddrhead, ifaddr); +TAILQ_HEAD(ifprefixhead, ifprefix); +LIST_HEAD(ifmultihead, ifmultiaddr); +TAILQ_HEAD(tailq_head, tqdummy); TAILQ_HEAD(ifnet_filter_head, ifnet_filter); TAILQ_HEAD(ddesc_head_name, dlil_demux_desc); +#endif /* BSD_KERNEL_PRIVATE */ -/* bottom 16 bits reserved for hardware checksum */ -#define IF_HWASSIST_CSUM_IP 0x0001 /* will csum IP */ -#define IF_HWASSIST_CSUM_TCP 0x0002 /* will csum TCP */ -#define IF_HWASSIST_CSUM_UDP 0x0004 /* will csum UDP */ -#define IF_HWASSIST_CSUM_IP_FRAGS 0x0008 /* will csum IP fragments */ -#define IF_HWASSIST_CSUM_FRAGMENT 0x0010 /* will do IP fragmentation */ -#define IF_HWASSIST_CSUM_TCP_SUM16 0x1000 /* simple TCP Sum16 computation */ +#ifdef PRIVATE +/* + * All of the following IF_HWASSIST_* flags are defined in kpi_inteface.h as + * IFNET_* flags. These are redefined here as constants to avoid failures to + * build user level programs that can not include kpi_interface.h. It is + * important to keep this in sync with the definitions in kpi_interface.h. + * The corresponding constant for each definition is mentioned in the comment. + * + * Bottom 16 bits reserved for hardware checksum + */ +#define IF_HWASSIST_CSUM_IP 0x0001 /* will csum IP, IFNET_CSUM_IP */ +#define IF_HWASSIST_CSUM_TCP 0x0002 /* will csum TCP, IFNET_CSUM_TCP */ +#define IF_HWASSIST_CSUM_UDP 0x0004 /* will csum UDP, IFNET_CSUM_UDP */ +#define IF_HWASSIST_CSUM_IP_FRAGS 0x0008 /* will csum IP fragments, IFNET_CSUM_FRAGMENT */ +#define IF_HWASSIST_CSUM_FRAGMENT 0x0010 /* will do IP fragmentation, IFNET_IP_FRAGMENT */ +#define IF_HWASSIST_CSUM_TCPIPV6 0x0020 /* will csum TCPv6, IFNET_CSUM_TCPIPV6 */ +#define IF_HWASSIST_CSUM_UDPIPV6 0x0040 /* will csum UDPv6, IFNET_CSUM_UDP */ +#define IF_HWASSIST_CSUM_FRAGMENT_IPV6 0x0080 /* will do IPv6 fragmentation, IFNET_IPV6_FRAGMENT */ +#define IF_HWASSIST_CSUM_PARTIAL 0x1000 /* simple Sum16 computation, IFNET_CSUM_PARTIAL */ #define IF_HWASSIST_CSUM_MASK 0xffff #define IF_HWASSIST_CSUM_FLAGS(hwassist) ((hwassist) & IF_HWASSIST_CSUM_MASK) /* VLAN support */ -#define IF_HWASSIST_VLAN_TAGGING 0x10000 /* supports VLAN tagging */ -#define IF_HWASSIST_VLAN_MTU 0x20000 /* supports VLAN MTU-sized packet (for software VLAN) */ +#define IF_HWASSIST_VLAN_TAGGING 0x00010000 /* supports VLAN tagging, IFNET_VLAN_TAGGING */ +#define IF_HWASSIST_VLAN_MTU 0x00020000 /* supports VLAN MTU-sized packet (for software VLAN), IFNET_VLAN_MTU */ + +/* TCP Segment Offloading support */ -#define IFNET_RW_LOCK 1 +#define IF_HWASSIST_TSO_V4 0x00200000 /* will do TCP Segment offload for IPv4, IFNET_TSO_IPV4 */ +#define IF_HWASSIST_TSO_V6 0x00400000 /* will do TCP Segment offload for IPv6, IFNET_TSO_IPV6 */ +#endif /* PRIVATE */ +#ifdef PRIVATE +#define IFXNAMSIZ (IFNAMSIZ + 8) /* external name (name + unit) */ +#endif + +#ifdef BSD_KERNEL_PRIVATE /* - * Structure defining a queue for a network interface. + * ifnet is private to BSD portion of kernel */ -struct ifqueue { - void *ifq_head; - void *ifq_tail; - int ifq_len; - int ifq_maxlen; - int ifq_drops; -}; +#include +#include +#include +#include +#include +#include -struct ddesc_head_str; -struct proto_hash_entry; -struct kev_msg; +RB_HEAD(ll_reach_tree, if_llreach); /* define struct ll_reach_tree */ +#define if_name(ifp) ifp->if_xname /* * Structure defining a network interface. * * (Would like to call this struct ``if'', but C isn't PL/1.) */ struct ifnet { - void *if_softc; /* pointer to driver state */ - const char *if_name; /* name, e.g. ``en'' or ``lo'' */ - TAILQ_ENTRY(ifnet) if_link; /* all struct ifnets are chained */ - struct ifaddrhead if_addrhead; /* linked list of addresses per if */ - u_long if_refcnt; -#ifdef __KPI_INTERFACE__ - ifnet_check_multi if_check_multi; -#else - void* if_check_multi; -#endif __KPI_INTERFACE__ - int if_pcount; /* number of promiscuous listeners */ - struct bpf_if *if_bpf; /* packet filter structure */ - u_short if_index; /* numeric abbreviation for this if */ - short if_unit; /* sub-unit for lower level driver */ - short if_timer; /* time 'til if_watchdog called */ - short if_flags; /* up/down, broadcast, etc. */ - int if_ipending; /* interrupts pending */ - void *if_linkmib; /* link-type-specific MIB data */ - size_t if_linkmiblen; /* length of above data */ - struct if_data_internal if_data; - -/* New with DLIL */ -#ifdef BSD_KERNEL_PRIVATE - int if_usecnt; -#else - int refcnt; -#endif - int offercnt; -#ifdef __KPI_INTERFACE__ + /* + * Lock (RW or mutex) to protect this data structure (static storage.) + */ + decl_lck_rw_data(, if_lock); + void *if_softc; /* pointer to driver state */ + const char *if_name; /* name, e.g. ``en'' or ``lo'' */ + const char *if_xname; /* external name (name + unit) */ + struct if_description if_desc; /* extended description */ + TAILQ_ENTRY(ifnet) if_link; /* all struct ifnets are chained */ + TAILQ_ENTRY(ifnet) if_detaching_link; /* list of detaching ifnets */ + + decl_lck_mtx_data(, if_ref_lock) + u_int32_t if_refflags; /* see IFRF flags below */ + u_int32_t if_refio; /* number of io ops to the underlying driver */ + +#define if_list if_link + struct ifaddrhead if_addrhead; /* linked list of addresses per if */ +#define if_addrlist if_addrhead + struct ifaddr *if_lladdr; /* link address (first/permanent) */ + + int if_pcount; /* number of promiscuous listeners */ + struct bpf_if *if_bpf; /* packet filter structure */ + u_short if_index; /* numeric abbreviation for this if */ + short if_unit; /* sub-unit for lower level driver */ + short if_timer; /* time 'til if_watchdog called */ + short if_flags; /* up/down, broadcast, etc. */ + u_int32_t if_eflags; /* see */ + + int if_capabilities; /* interface features & capabilities */ + int if_capenable; /* enabled features & capabilities */ + + void *if_linkmib; /* link-type-specific MIB data */ + size_t if_linkmiblen; /* length of above data */ + + struct if_data_internal if_data __attribute__((aligned(8))); + + ifnet_family_t if_family; /* value assigned by Apple */ + ifnet_subfamily_t if_subfamily; /* value assigned by Apple */ + uintptr_t if_family_cookie; ifnet_output_func if_output; + ifnet_pre_enqueue_func if_pre_enqueue; + ifnet_start_func if_start; + ifnet_ctl_func if_output_ctl; + ifnet_input_poll_func if_input_poll; + ifnet_ctl_func if_input_ctl; ifnet_ioctl_func if_ioctl; ifnet_set_bpf_tap if_set_bpf_tap; ifnet_detached_func if_free; ifnet_demux_func if_demux; ifnet_event_func if_event; - ifnet_framer_func if_framer; - ifnet_family_t if_family; /* ulong assigned by Apple */ -#else - void* if_output; - void* if_ioctl; - void* if_set_bpf_tap; - void* if_free; - void* if_demux; - void* if_event; - void* if_framer; - u_long if_family; /* ulong assigned by Apple */ -#endif + ifnet_framer_func if_framer_legacy; + ifnet_framer_extended_func if_framer; + ifnet_add_proto_func if_add_proto; + ifnet_del_proto_func if_del_proto; + ifnet_check_multi if_check_multi; + struct proto_hash_entry *if_proto_hash; + void *if_kpi_storage; - struct ifnet_filter_head if_flt_head; + u_int32_t if_flowhash; /* interface flow control ID */ -/* End DLIL specific */ + decl_lck_mtx_data(, if_start_lock); + u_int32_t if_start_flags; /* see IFSF flags below */ + u_int32_t if_start_req; + u_int16_t if_start_active; /* output is active */ + u_int16_t if_start_delayed; + u_int16_t if_start_delay_qlen; + u_int16_t if_start_delay_idle; + u_int64_t if_start_delay_swin; + u_int32_t if_start_delay_cnt; + u_int32_t if_start_delay_timeout; /* nanoseconds */ + struct timespec if_start_cycle; /* restart interval */ + struct thread *if_start_thread; - u_long if_delayed_detach; /* need to perform delayed detach */ - void *if_private; /* private to interface */ - long if_eflags; /* autoaddr, autoaddr done, etc. */ + struct ifclassq if_snd; /* transmit queue */ + u_int32_t if_output_sched_model; /* tx sched model */ - struct ifmultihead if_multiaddrs; /* multicast addresses configured */ - int if_amcount; /* number of all-multicast requests */ -/* procedure handles */ -#ifdef __KPI_INTERFACE__ - union { - int (*original)(struct ifnet *ifp, u_long protocol_family, - struct ddesc_head_str *demux_desc_head); - ifnet_add_proto_func kpi; - } if_add_proto_u; - ifnet_del_proto_func if_del_proto; -#else __KPI_INTERFACE__ - void* if_add_proto; - void* if_del_proto; -#endif __KPI_INTERFACE__ - struct proto_hash_entry *if_proto_hash; - void *if_kpi_storage; - - void *unused_was_init; - void *unused_was_resolvemulti; - - struct ifqueue if_snd; - u_long unused_2[1]; -#ifdef __APPLE__ - u_long family_cookie; - struct ifprefixhead if_prefixhead; /* list of prefixes per if */ + struct if_bandwidths if_output_bw; + struct if_bandwidths if_input_bw; -#ifdef _KERN_LOCKS_H_ -#if IFNET_RW_LOCK - lck_rw_t *if_lock; /* Lock to protect this interface */ -#else - lck_mtx_t *if_lock; /* Lock to protect this interface */ -#endif -#else - void *if_lock; -#endif + struct if_latencies if_output_lt; + struct if_latencies if_input_lt; -#else - struct ifprefixhead if_prefixhead; /* list of prefixes per if */ -#endif /* __APPLE__ */ + decl_lck_mtx_data(, if_flt_lock) + u_int32_t if_flt_busy; + u_int32_t if_flt_waiters; + struct ifnet_filter_head if_flt_head; + + struct ifmultihead if_multiaddrs; /* multicast addresses */ + u_int32_t if_updatemcasts; /* mcast addrs need updating */ + int if_amcount; /* # of all-multicast reqs */ + decl_lck_mtx_data(, if_addrconfig_lock); /* for serializing addr config */ + struct in_multi *if_allhostsinm; /* store all-hosts inm for this ifp */ + + decl_lck_mtx_data(, if_poll_lock); + u_int16_t if_poll_req; + u_int16_t if_poll_update; /* link update */ + u_int32_t if_poll_active; /* polling is active */ + struct timespec if_poll_cycle; /* poll interval */ + struct thread *if_poll_thread; + + struct dlil_threading_info *if_inp; + + struct ifprefixhead if_prefixhead; /* list of prefixes per if */ struct { - u_long length; + u_int32_t length; union { u_char buffer[8]; u_char *ptr; } u; } if_broadcast; +#if CONFIG_MACF_NET + struct label *if_label; /* interface MAC label */ +#endif + + u_int32_t if_wake_properties; +#if PF + struct pfi_kif *if_pf_kif; +#endif /* PF */ + + decl_lck_mtx_data(, if_cached_route_lock); + u_int32_t if_fwd_cacheok; + struct route if_fwd_route; /* cached forwarding route */ + struct route if_src_route; /* cached ipv4 source route */ + struct route_in6 if_src_route6; /* cached ipv6 source route */ + + decl_lck_rw_data(, if_llreach_lock); + struct ll_reach_tree if_ll_srcs; /* source link-layer tree */ + + void *if_bridge; /* bridge glue */ + + u_int32_t if_want_aggressive_drain; + u_int32_t if_idle_flags; /* idle flags */ + u_int32_t if_idle_new_flags; /* temporary idle flags */ + u_int32_t if_idle_new_flags_mask; /* temporary mask */ + u_int32_t if_route_refcnt; /* idle: route ref count */ + + struct if_traffic_class if_tc __attribute__((aligned(8))); +#if INET + struct igmp_ifinfo *if_igi; /* for IGMPv3 */ +#endif /* INET */ +#if INET6 + struct mld_ifinfo *if_mli; /* for MLDv2 */ +#endif /* INET6 */ + +#if MEASURE_BW + struct if_measured_bw if_bw; +#endif /* MEASURE_BW */ + struct tcpstat_local *if_tcp_stat; /* TCP specific stats */ + struct udpstat_local *if_udp_stat; /* UDP specific stats */ + + struct { + int32_t level; /* cached logging level */ + u_int32_t flags; /* cached logging flags */ + int32_t category; /* cached category */ + int32_t subcategory; /* cached subcategory */ + } if_log; + + struct { + struct ifnet *ifp; /* delegated ifp */ + u_int32_t type; /* delegated i/f type */ + u_int32_t family; /* delegated i/f family */ + u_int32_t subfamily; /* delegated i/f sub-family */ + uint32_t expensive:1; /* delegated i/f expensive? */ + } if_delegated; + +#define IF_MAXAGENTS 8 + uuid_t if_agentids[IF_MAXAGENTS]; + + u_int64_t if_data_threshold; + u_int32_t if_fg_sendts; /* last send on a fg socket in seconds */ + u_int32_t if_rt_sendts; /* last of a real time packet */ + +#if INET + decl_lck_rw_data(, if_inetdata_lock); + void *if_inetdata; +#endif /* INET */ +#if INET6 + decl_lck_rw_data(, if_inet6data_lock); + void *if_inet6data; +#endif + decl_lck_rw_data(, if_link_status_lock); + struct if_link_status *if_link_status; + struct if_interface_state if_interface_state; + struct if_tcp_ecn_stat *if_ipv4_stat; + struct if_tcp_ecn_stat *if_ipv6_stat; }; -#define if_add_proto if_add_proto_u.original +#define IF_TCP_STATINC(_ifp, _s) do { \ + if ((_ifp)->if_tcp_stat != NULL) \ + atomic_add_64(&(_ifp)->if_tcp_stat->_s, 1); \ +} while (0); -#ifndef __APPLE__ -/* for compatibility with other BSDs */ -#define if_addrlist if_addrhead -#define if_list if_link -#endif !__APPLE__ +#define IF_UDP_STATINC(_ifp, _s) do { \ + if ((_ifp)->if_udp_stat != NULL) \ + atomic_add_64(&(_ifp)->if_udp_stat->_s, 1); \ +} while (0); +/* + * Valid values for if_refflags + */ +#define IFRF_ATTACHED 0x1 /* ifnet attach is completely done */ +#define IFRF_DETACHING 0x2 /* detach has been requested */ -#endif /* PRIVATE */ +/* + * Valid values for if_start_flags + */ +#define IFSF_FLOW_CONTROLLED 0x1 /* flow controlled */ -#ifdef KERNEL_PRIVATE /* * Structure describing a `cloning' interface. */ struct if_clone { LIST_ENTRY(if_clone) ifc_list; /* on list of cloners */ - const char *ifc_name; /* name of device, e.g. `vlan' */ - size_t ifc_namelen; /* length of name */ - int ifc_minifs; /* minimum number of interfaces */ - int ifc_maxunit; /* maximum unit number */ - unsigned char *ifc_units; /* bitmap to handle units */ - int ifc_bmlen; /* bitmap length */ - - int (*ifc_create)(struct if_clone *, int); - void (*ifc_destroy)(struct ifnet *); + const char *ifc_name; /* name of device, e.g. `vlan' */ + size_t ifc_namelen; /* length of name */ + u_int32_t ifc_minifs; /* minimum number of interfaces */ + u_int32_t ifc_maxunit; /* maximum unit number */ + unsigned char *ifc_units; /* bitmap to handle units */ + u_int32_t ifc_bmlen; /* bitmap length */ + + int (*ifc_create)(struct if_clone *, u_int32_t, void *); + int (*ifc_destroy)(struct ifnet *); }; -#define IF_CLONE_INITIALIZER(name, create, destroy, minifs, maxunit) \ - { { 0, 0 }, name, sizeof(name) - 1, minifs, maxunit, NULL, 0, create, destroy } +#define IF_CLONE_INITIALIZER(name, create, destroy, minifs, maxunit) { \ + { NULL, NULL }, name, (sizeof (name) - 1), minifs, maxunit, NULL, 0, \ + create, destroy \ +} -/* - * Bit values in if_ipending - */ -#define IFI_RECV 1 /* I want to receive */ -#define IFI_XMIT 2 /* I want to transmit */ +#define M_CLONE M_IFADDR /* - * Output queues (ifp->if_snd) and slow device input queues (*ifp->if_slowq) - * are queues of messages stored on ifqueue structures - * (defined above). Entries are added to and deleted from these structures - * by these macros, which should be called with ipl raised to splimp(). + * Macros to manipulate ifqueue. Users of these macros are responsible + * for serialization, by holding whatever lock is appropriate for the + * corresponding structure that is referring the ifqueue. */ #define IF_QFULL(ifq) ((ifq)->ifq_len >= (ifq)->ifq_maxlen) #define IF_DROP(ifq) ((ifq)->ifq_drops++) -#define IF_ENQUEUE(ifq, m) { \ - (m)->m_nextpkt = 0; \ - if ((ifq)->ifq_tail == 0) \ - (ifq)->ifq_head = m; \ - else \ - ((struct mbuf*)(ifq)->ifq_tail)->m_nextpkt = m; \ - (ifq)->ifq_tail = m; \ - (ifq)->ifq_len++; \ -} -#define IF_PREPEND(ifq, m) { \ - (m)->m_nextpkt = (ifq)->ifq_head; \ - if ((ifq)->ifq_tail == 0) \ - (ifq)->ifq_tail = (m); \ - (ifq)->ifq_head = (m); \ - (ifq)->ifq_len++; \ -} -#define IF_DEQUEUE(ifq, m) { \ - (m) = (ifq)->ifq_head; \ - if (m) { \ - if (((ifq)->ifq_head = (m)->m_nextpkt) == 0) \ - (ifq)->ifq_tail = 0; \ - (m)->m_nextpkt = 0; \ - (ifq)->ifq_len--; \ - } \ -} - -#define IF_ENQ_DROP(ifq, m) if_enq_drop(ifq, m) -#if defined(__GNUC__) && defined(MT_HEADER) -static __inline int -if_queue_drop(struct ifqueue *ifq, __unused struct mbuf *m) -{ - IF_DROP(ifq); - return 0; -} +#define IF_ENQUEUE(ifq, m) do { \ + (m)->m_nextpkt = NULL; \ + if ((ifq)->ifq_tail == NULL) \ + (ifq)->ifq_head = m; \ + else \ + ((struct mbuf*)(ifq)->ifq_tail)->m_nextpkt = m; \ + (ifq)->ifq_tail = m; \ + (ifq)->ifq_len++; \ +} while (0) -static __inline int -if_enq_drop(struct ifqueue *ifq, struct mbuf *m) -{ - if (IF_QFULL(ifq) && - !if_queue_drop(ifq, m)) - return 0; - IF_ENQUEUE(ifq, m); - return 1; -} -#else +#define IF_PREPEND(ifq, m) do { \ + (m)->m_nextpkt = (ifq)->ifq_head; \ + if ((ifq)->ifq_tail == NULL) \ + (ifq)->ifq_tail = (m); \ + (ifq)->ifq_head = (m); \ + (ifq)->ifq_len++; \ +} while (0) -#ifdef MT_HEADER -int if_enq_drop(struct ifqueue *, struct mbuf *); -#endif MT_HEADER +#define IF_DEQUEUE(ifq, m) do { \ + (m) = (ifq)->ifq_head; \ + if (m != NULL) { \ + if (((ifq)->ifq_head = (m)->m_nextpkt) == NULL) \ + (ifq)->ifq_tail = NULL; \ + (m)->m_nextpkt = NULL; \ + (ifq)->ifq_len--; \ + } \ +} while (0) -#endif defined(__GNUC__) && defined(MT_HEADER) - -#endif /* KERNEL_PRIVATE */ +#define IF_REMQUEUE(ifq, m) do { \ + struct mbuf *_p = (ifq)->ifq_head; \ + struct mbuf *_n = (m)->m_nextpkt; \ + if ((m) == _p) \ + _p = NULL; \ + while (_p != NULL) { \ + if (_p->m_nextpkt == (m)) \ + break; \ + _p = _p->m_nextpkt; \ + } \ + VERIFY(_p != NULL || ((m) == (ifq)->ifq_head)); \ + if ((m) == (ifq)->ifq_head) \ + (ifq)->ifq_head = _n; \ + if ((m) == (ifq)->ifq_tail) \ + (ifq)->ifq_tail = _p; \ + VERIFY((ifq)->ifq_tail != NULL || (ifq)->ifq_head == NULL); \ + VERIFY((ifq)->ifq_len != 0); \ + --(ifq)->ifq_len; \ + if (_p != NULL) \ + _p->m_nextpkt = _n; \ + (m)->m_nextpkt = NULL; \ +} while (0) +#define IF_DRAIN(ifq) do { \ + struct mbuf *_m; \ + for (;;) { \ + IF_DEQUEUE(ifq, _m); \ + if (_m == NULL) \ + break; \ + m_freem(_m); \ + } \ +} while (0) -#ifdef PRIVATE /* * The ifaddr structure contains information about one address * of an interface. They are maintained by the different address families, @@ -542,31 +1073,77 @@ int if_enq_drop(struct ifqueue *, struct mbuf *); * together so all addresses for an interface can be located. */ struct ifaddr { - struct sockaddr *ifa_addr; /* address of interface */ - struct sockaddr *ifa_dstaddr; /* other end of p-to-p link */ + decl_lck_mtx_data(, ifa_lock); /* lock for ifaddr */ + uint32_t ifa_refcnt; /* ref count, use IFA_{ADD,REM}REF */ + uint32_t ifa_debug; /* debug flags */ + struct sockaddr *ifa_addr; /* address of interface */ + struct sockaddr *ifa_dstaddr; /* other end of p-to-p link */ #define ifa_broadaddr ifa_dstaddr /* broadcast address interface */ - struct sockaddr *ifa_netmask; /* used to determine subnet */ - struct ifnet *ifa_ifp; /* back-pointer to interface */ + struct sockaddr *ifa_netmask; /* used to determine subnet */ + struct ifnet *ifa_ifp; /* back-pointer to interface */ TAILQ_ENTRY(ifaddr) ifa_link; /* queue macro glue */ - void (*ifa_rtrequest) /* check or clean routes (+ or -)'d */ - (int, struct rtentry *, struct sockaddr *); - u_short ifa_flags; /* mostly rt_flags for cloning */ - int ifa_refcnt;/* 32bit ref count, use ifaref, ifafree */ - int ifa_metric; /* cost of going out this interface */ -#ifdef notdef - struct rtentry *ifa_rt; /* XXXX for ROUTETOIF ????? */ -#endif - int (*ifa_claim_addr) /* check if an addr goes to this if */ - (struct ifaddr *, const struct sockaddr *); - u_long ifa_debug; /* debug flags */ + void (*ifa_rtrequest) /* check or clean routes (+ or -)'d */ + (int, struct rtentry *, struct sockaddr *); + uint32_t ifa_flags; /* mostly rt_flags for cloning */ + int32_t ifa_metric; /* cost of going out this interface */ + void (*ifa_free)(struct ifaddr *); /* callback fn for freeing */ + void (*ifa_trace) /* callback fn for tracing refs */ + (struct ifaddr *, int); + void (*ifa_attached)(struct ifaddr *); /* callback fn for attaching */ + void (*ifa_detached)(struct ifaddr *); /* callback fn for detaching */ }; + + +/* + * Valid values for ifa_flags + */ #define IFA_ROUTE RTF_UP /* route installed (0x1) */ #define IFA_CLONING RTF_CLONING /* (0x100) */ -#define IFA_ATTACHED 0x1 /* ifa_debug: IFA is attached to an interface */ -#endif /* PRIVATE */ +/* + * Valid values for ifa_debug + */ +#define IFD_ATTACHED 0x1 /* attached to list */ +#define IFD_ALLOC 0x2 /* dynamically allocated */ +#define IFD_DEBUG 0x4 /* has debugging info */ +#define IFD_LINK 0x8 /* link address */ +#define IFD_TRASHED 0x10 /* in trash list */ +#define IFD_SKIP 0x20 /* skip this entry */ +#define IFD_NOTREADY 0x40 /* embryonic; not yet ready */ + +#define IFA_LOCK_ASSERT_HELD(_ifa) \ + lck_mtx_assert(&(_ifa)->ifa_lock, LCK_MTX_ASSERT_OWNED) + +#define IFA_LOCK_ASSERT_NOTHELD(_ifa) \ + lck_mtx_assert(&(_ifa)->ifa_lock, LCK_MTX_ASSERT_NOTOWNED) + +#define IFA_LOCK(_ifa) \ + lck_mtx_lock(&(_ifa)->ifa_lock) + +#define IFA_LOCK_SPIN(_ifa) \ + lck_mtx_lock_spin(&(_ifa)->ifa_lock) + +#define IFA_CONVERT_LOCK(_ifa) do { \ + IFA_LOCK_ASSERT_HELD(_ifa); \ + lck_mtx_convert_spin(&(_ifa)->ifa_lock); \ +} while (0) + +#define IFA_UNLOCK(_ifa) \ + lck_mtx_unlock(&(_ifa)->ifa_lock) + +#define IFA_ADDREF(_ifa) \ + ifa_addref(_ifa, 0) + +#define IFA_ADDREF_LOCKED(_ifa) \ + ifa_addref(_ifa, 1) + +#define IFA_REMREF(_ifa) do { \ + (void) ifa_remref(_ifa, 0); \ +} while (0) + +#define IFA_REMREF_LOCKED(_ifa) \ + ifa_remref(_ifa, 1) -#ifdef KERNEL_PRIVATE /* * The prefix structure contains information about one prefix * of an interface. They are maintained by the different address families, @@ -580,95 +1157,337 @@ struct ifprefix { u_char ifpr_plen; /* prefix length in bits */ u_char ifpr_type; /* protocol dependent prefix type */ }; -#endif /* KERNEL_PRIVATE */ - -#ifdef PRIVATE -typedef void (*ifma_protospec_free_func)(void* ifma_protospec); /* * Multicast address structure. This is analogous to the ifaddr * structure except that it keeps track of multicast addresses. - * Also, the reference count here is a count of requests for this - * address, not a count of pointers to this structure. + * Also, the request count here is a count of requests for this + * address, not a count of pointers to this structure; anonymous + * membership(s) holds one outstanding request count. */ struct ifmultiaddr { + decl_lck_mtx_data(, ifma_lock); + u_int32_t ifma_refcount; /* reference count */ + u_int32_t ifma_anoncnt; /* # of anonymous requests */ + u_int32_t ifma_reqcnt; /* total requests for this address */ + u_int32_t ifma_debug; /* see ifa_debug flags */ + u_int32_t ifma_flags; /* see below */ LIST_ENTRY(ifmultiaddr) ifma_link; /* queue macro glue */ - struct sockaddr *ifma_addr; /* address this membership is for */ + struct sockaddr *ifma_addr; /* address this membership is for */ struct ifmultiaddr *ifma_ll; /* link-layer translation, if any */ - struct ifnet *ifma_ifp; /* back-pointer to interface */ - u_int ifma_usecount; /* use count, protected by ifp's lock */ - void *ifma_protospec; /* protocol-specific state, if any */ - int32_t ifma_refcount; /* reference count, atomically protected */ - ifma_protospec_free_func ifma_free; /* function called to free ifma_protospec */ + struct ifnet *ifma_ifp; /* back-pointer to interface */ + void *ifma_protospec; /* protocol-specific state, if any */ + void (*ifma_trace) /* callback fn for tracing refs */ + (struct ifmultiaddr *, int); }; -#endif /* PRIVATE */ -#ifdef KERNEL_PRIVATE -#define IFAREF(ifa) ifaref(ifa) -#define IFAFREE(ifa) ifafree(ifa) - -/* - * To preserve kmem compatibility, we define - * ifnet_head to ifnet. This should be temp. - */ -#define ifnet_head ifnet -extern struct ifnethead ifnet_head; -extern struct ifnet **ifindex2ifnet; -extern int ifqmaxlen; -extern struct ifnet loif[]; -extern int if_index; -extern struct ifaddr **ifnet_addrs; -extern struct ifnet *lo_ifp; - -int if_addmulti(struct ifnet *, const struct sockaddr *, struct ifmultiaddr **); -int if_allmulti(struct ifnet *, int); -void if_attach(struct ifnet *); -int if_delmultiaddr(struct ifmultiaddr *ifma, int locked); -int if_delmulti(struct ifnet *, const struct sockaddr *); -void if_down(struct ifnet *); -void if_route(struct ifnet *, int flag, int fam); -void if_unroute(struct ifnet *, int flag, int fam); -void if_up(struct ifnet *); -void if_updown(struct ifnet *ifp, int up); -/*void ifinit(void));*/ /* declared in systm.h for main( */ -int ifioctl(struct socket *, u_long, caddr_t, struct proc *); -int ifioctllocked(struct socket *, u_long, caddr_t, struct proc *); -struct ifnet *ifunit(const char *); -struct ifnet *if_withname(struct sockaddr *); - -void if_clone_attach(struct if_clone *); -void if_clone_detach(struct if_clone *); - -void ifnet_lock_assert(struct ifnet *ifp, int what); -void ifnet_lock_shared(struct ifnet *ifp); -void ifnet_lock_exclusive(struct ifnet *ifp); -void ifnet_lock_done(struct ifnet *ifp); - -void ifnet_head_lock_shared(void); -void ifnet_head_lock_exclusive(void); -void ifnet_head_done(void); - -void if_attach_ifa(struct ifnet * ifp, struct ifaddr *ifa); -void if_detach_ifa(struct ifnet * ifp, struct ifaddr *ifa); - -void ifma_reference(struct ifmultiaddr *ifma); -void ifma_release(struct ifmultiaddr *ifma); - -struct ifaddr *ifa_ifwithaddr(const struct sockaddr *); -struct ifaddr *ifa_ifwithdstaddr(const struct sockaddr *); -struct ifaddr *ifa_ifwithnet(const struct sockaddr *); -struct ifaddr *ifa_ifwithroute(int, const struct sockaddr *, const struct sockaddr *); -struct ifaddr *ifaof_ifpforaddr(const struct sockaddr *, struct ifnet *); -void ifafree(struct ifaddr *); -void ifaref(struct ifaddr *); - -struct ifmultiaddr *ifmaof_ifpforaddr(const struct sockaddr *, struct ifnet *); +/* + * Values for ifma_flags + */ +#define IFMAF_ANONYMOUS 0x1 /* has anonymous request ref(s) held */ -#ifdef BSD_KERNEL_PRIVATE -void if_data_internal_to_if_data(const struct if_data_internal *if_data_int, - struct if_data *if_data); -void if_data_internal_to_if_data64(const struct if_data_internal *if_data_int, - struct if_data64 *if_data64); +#define IFMA_LOCK_ASSERT_HELD(_ifma) \ + lck_mtx_assert(&(_ifma)->ifma_lock, LCK_MTX_ASSERT_OWNED) + +#define IFMA_LOCK_ASSERT_NOTHELD(_ifma) \ + lck_mtx_assert(&(_ifma)->ifma_lock, LCK_MTX_ASSERT_NOTOWNED) + +#define IFMA_LOCK(_ifma) \ + lck_mtx_lock(&(_ifma)->ifma_lock) + +#define IFMA_LOCK_SPIN(_ifma) \ + lck_mtx_lock_spin(&(_ifma)->ifma_lock) + +#define IFMA_CONVERT_LOCK(_ifma) do { \ + IFMA_LOCK_ASSERT_HELD(_ifma); \ + lck_mtx_convert_spin(&(_ifma)->ifma_lock); \ +} while (0) + +#define IFMA_UNLOCK(_ifma) \ + lck_mtx_unlock(&(_ifma)->ifma_lock) + +#define IFMA_ADDREF(_ifma) \ + ifma_addref(_ifma, 0) + +#define IFMA_ADDREF_LOCKED(_ifma) \ + ifma_addref(_ifma, 1) + +#define IFMA_REMREF(_ifma) \ + ifma_remref(_ifma) + +/* + * Indicate whether or not the immediate interface, or the interface delegated + * by it, is a cellular interface (IFT_CELLULAR). Delegated interface type is + * set/cleared along with the delegated ifp; we cache the type for performance + * to avoid dereferencing delegated ifp each time. + * + * Note that this is meant to be used only for accounting and policy purposes; + * certain places need to explicitly know the immediate interface type, and + * this macro should not be used there. + * + * The test is done against IFT_CELLULAR instead of IFNET_FAMILY_CELLULAR to + * handle certain cases where the family isn't set to the latter. + */ +#define IFNET_IS_CELLULAR(_ifp) \ + ((_ifp)->if_type == IFT_CELLULAR || \ + (_ifp)->if_delegated.type == IFT_CELLULAR) + +/* + * Indicate whether or not the immediate interface, or the interface delegated + * by it, is an ETHERNET interface. + */ +#define IFNET_IS_ETHERNET(_ifp) \ + ((_ifp)->if_family == IFNET_FAMILY_ETHERNET || \ + (_ifp)->if_delegated.family == IFNET_FAMILY_ETHERNET) +/* + * Indicate whether or not the immediate interface, or the interface delegated + * by it, is a Wi-Fi interface (IFNET_SUBFAMILY_WIFI). Delegated interface + * subfamily is set/cleared along with the delegated ifp; we cache the subfamily + * for performance to avoid dereferencing delegated ifp each time. + * + * Note that this is meant to be used only for accounting and policy purposes; + * certain places need to explicitly know the immediate interface type, and + * this macro should not be used there. + * + * The test is done against IFNET_SUBFAMILY_WIFI as the family may be set to + * IFNET_FAMILY_ETHERNET (as well as type to IFT_ETHER) which is too generic. + */ +#define IFNET_IS_WIFI(_ifp) \ + ((_ifp)->if_subfamily == IFNET_SUBFAMILY_WIFI || \ + (_ifp)->if_delegated.subfamily == IFNET_SUBFAMILY_WIFI) + +/* + * Indicate whether or not the immediate interface, or the interface delegated + * by it, is a Wired interface (several families). Delegated interface + * family is set/cleared along with the delegated ifp; we cache the family + * for performance to avoid dereferencing delegated ifp each time. + * + * Note that this is meant to be used only for accounting and policy purposes; + * certain places need to explicitly know the immediate interface type, and + * this macro should not be used there. + */ +#define IFNET_IS_WIRED(_ifp) \ + ((_ifp)->if_family == IFNET_FAMILY_ETHERNET || \ + (_ifp)->if_delegated.family == IFNET_FAMILY_ETHERNET || \ + (_ifp)->if_family == IFNET_FAMILY_FIREWIRE || \ + (_ifp)->if_delegated.family == IFNET_FAMILY_FIREWIRE) + +/* + * Indicate whether or not the immediate interface, or the interface delegated + * by it, is marked as expensive. The delegated interface is set/cleared + * along with the delegated ifp; we cache the flag for performance to avoid + * dereferencing delegated ifp each time. + * + * Note that this is meant to be used only for policy purposes. + */ +#define IFNET_IS_EXPENSIVE(_ifp) \ + ((_ifp)->if_eflags & IFEF_EXPENSIVE || \ + (_ifp)->if_delegated.expensive) + +/* + * We don't support AWDL interface delegation. + */ +#define IFNET_IS_AWDL_RESTRICTED(_ifp) \ + (((_ifp)->if_eflags & (IFEF_AWDL|IFEF_AWDL_RESTRICTED)) == \ + (IFEF_AWDL|IFEF_AWDL_RESTRICTED)) + + +extern struct ifnethead ifnet_head; +extern struct ifnet **ifindex2ifnet; +extern u_int32_t if_sndq_maxlen; +extern u_int32_t if_rcvq_maxlen; +extern int if_index; +extern struct ifaddr **ifnet_addrs; +extern lck_attr_t *ifa_mtx_attr; +extern lck_grp_t *ifa_mtx_grp; +extern lck_grp_t *ifnet_lock_group; +extern lck_attr_t *ifnet_lock_attr; +extern ifnet_t lo_ifp; +extern uint32_t if_bw_measure_size; +extern u_int32_t if_bw_smoothing_val; + +extern int if_addmulti(struct ifnet *, const struct sockaddr *, + struct ifmultiaddr **); +extern int if_addmulti_anon(struct ifnet *, const struct sockaddr *, + struct ifmultiaddr **); +extern int if_allmulti(struct ifnet *, int); +extern int if_delmulti(struct ifnet *, const struct sockaddr *); +extern int if_delmulti_ifma(struct ifmultiaddr *); +extern int if_delmulti_anon(struct ifnet *, const struct sockaddr *); +extern void if_down(struct ifnet *); +extern int if_down_all(void); +extern void if_up(struct ifnet *); +__private_extern__ void if_updown(struct ifnet *ifp, int up); +extern int ifioctl(struct socket *, u_long, caddr_t, struct proc *); +extern int ifioctllocked(struct socket *, u_long, caddr_t, struct proc *); +extern struct ifnet *ifunit(const char *); +extern struct ifnet *if_withname(struct sockaddr *); +extern void if_qflush(struct ifnet *, int); +extern void if_qflush_sc(struct ifnet *, mbuf_svc_class_t, u_int32_t, + u_int32_t *, u_int32_t *, int); + +extern struct if_clone *if_clone_lookup(const char *, u_int32_t *); +extern int if_clone_attach(struct if_clone *); +extern void if_clone_detach(struct if_clone *); + +extern u_int32_t if_functional_type(struct ifnet *); + +extern errno_t if_mcasts_update(struct ifnet *); +extern int32_t total_snd_byte_count; + +typedef enum { + IFNET_LCK_ASSERT_EXCLUSIVE, /* RW: held as writer */ + IFNET_LCK_ASSERT_SHARED, /* RW: held as reader */ + IFNET_LCK_ASSERT_OWNED, /* RW: writer/reader, MTX: held */ + IFNET_LCK_ASSERT_NOTOWNED /* not held */ +} ifnet_lock_assert_t; + +#define IF_LLADDR(_ifp) \ + (LLADDR(SDL(((_ifp)->if_lladdr)->ifa_addr))) + +__private_extern__ void ifnet_lock_assert(struct ifnet *, ifnet_lock_assert_t); +__private_extern__ void ifnet_lock_shared(struct ifnet *ifp); +__private_extern__ void ifnet_lock_exclusive(struct ifnet *ifp); +__private_extern__ void ifnet_lock_done(struct ifnet *ifp); + +#if INET +__private_extern__ void if_inetdata_lock_shared(struct ifnet *ifp); +__private_extern__ void if_inetdata_lock_exclusive(struct ifnet *ifp); +__private_extern__ void if_inetdata_lock_done(struct ifnet *ifp); +#endif + +#if INET6 +__private_extern__ void if_inet6data_lock_shared(struct ifnet *ifp); +__private_extern__ void if_inet6data_lock_exclusive(struct ifnet *ifp); +__private_extern__ void if_inet6data_lock_done(struct ifnet *ifp); +#endif + +__private_extern__ void ifnet_head_lock_shared(void); +__private_extern__ void ifnet_head_lock_exclusive(void); +__private_extern__ void ifnet_head_done(void); + +__private_extern__ errno_t ifnet_set_idle_flags_locked(ifnet_t, u_int32_t, + u_int32_t); +__private_extern__ int ifnet_is_attached(struct ifnet *, int refio); +__private_extern__ void ifnet_decr_iorefcnt(struct ifnet *); +__private_extern__ void ifnet_set_start_cycle(struct ifnet *, + struct timespec *); +__private_extern__ void ifnet_set_poll_cycle(struct ifnet *, + struct timespec *); + +__private_extern__ void if_attach_ifa(struct ifnet *, struct ifaddr *); +__private_extern__ void if_attach_link_ifa(struct ifnet *, struct ifaddr *); +__private_extern__ void if_detach_ifa(struct ifnet *, struct ifaddr *); +__private_extern__ void if_detach_link_ifa(struct ifnet *, struct ifaddr *); + +__private_extern__ void dlil_if_lock(void); +__private_extern__ void dlil_if_unlock(void); +__private_extern__ void dlil_if_lock_assert(void); + +extern struct ifaddr *ifa_ifwithaddr(const struct sockaddr *); +extern struct ifaddr *ifa_ifwithaddr_scoped(const struct sockaddr *, + unsigned int); +extern struct ifaddr *ifa_ifwithdstaddr(const struct sockaddr *); +extern struct ifaddr *ifa_ifwithnet(const struct sockaddr *); +extern struct ifaddr *ifa_ifwithnet_scoped(const struct sockaddr *, + unsigned int); +extern struct ifaddr *ifa_ifwithroute(int, const struct sockaddr *, + const struct sockaddr *); +extern struct ifaddr *ifa_ifwithroute_locked(int, const struct sockaddr *, + const struct sockaddr *); +extern struct ifaddr *ifa_ifwithroute_scoped_locked(int, + const struct sockaddr *, const struct sockaddr *, unsigned int); +extern struct ifaddr *ifaof_ifpforaddr(const struct sockaddr *, struct ifnet *); +__private_extern__ struct ifaddr *ifa_ifpgetprimary(struct ifnet *, int); +extern void ifa_addref(struct ifaddr *, int); +extern struct ifaddr *ifa_remref(struct ifaddr *, int); +extern void ifa_lock_init(struct ifaddr *); +extern void ifa_lock_destroy(struct ifaddr *); +extern void ifma_addref(struct ifmultiaddr *, int); +extern void ifma_remref(struct ifmultiaddr *); + +extern void ifa_init(void); + +__private_extern__ struct in_ifaddr *ifa_foraddr(unsigned int); +__private_extern__ struct in_ifaddr *ifa_foraddr_scoped(unsigned int, + unsigned int); + +struct ifreq; +extern errno_t ifnet_getset_opportunistic(struct ifnet *, u_long, + struct ifreq *, struct proc *); +extern int ifnet_get_throttle(struct ifnet *, u_int32_t *); +extern int ifnet_set_throttle(struct ifnet *, u_int32_t); +extern errno_t ifnet_getset_log(struct ifnet *, u_long, + struct ifreq *, struct proc *); +extern int ifnet_set_log(struct ifnet *, int32_t, uint32_t, int32_t, int32_t); +extern int ifnet_get_log(struct ifnet *, int32_t *, uint32_t *, int32_t *, + int32_t *); +extern int ifnet_notify_address(struct ifnet *, int); + +#if INET6 +struct in6_addr; +__private_extern__ struct in6_ifaddr *ifa_foraddr6(struct in6_addr *); +__private_extern__ struct in6_ifaddr *ifa_foraddr6_scoped(struct in6_addr *, + unsigned int); +#endif /* INET6 */ + +__private_extern__ void if_data_internal_to_if_data(struct ifnet *ifp, + const struct if_data_internal *if_data_int, struct if_data *if_data); +__private_extern__ void if_data_internal_to_if_data64(struct ifnet *ifp, + const struct if_data_internal *if_data_int, struct if_data64 *if_data64); +__private_extern__ void if_copy_traffic_class(struct ifnet *ifp, + struct if_traffic_class *if_tc); +__private_extern__ void if_copy_data_extended(struct ifnet *ifp, + struct if_data_extended *if_de); +__private_extern__ void if_copy_packet_stats(struct ifnet *ifp, + struct if_packet_stats *if_ps); +__private_extern__ void if_copy_rxpoll_stats(struct ifnet *ifp, + struct if_rxpoll_stats *if_rs); + +__private_extern__ struct rtentry *ifnet_cached_rtlookup_inet(struct ifnet *, + struct in_addr); +#if INET6 +__private_extern__ struct rtentry *ifnet_cached_rtlookup_inet6(struct ifnet *, + struct in6_addr *); +#endif /* INET6 */ + +__private_extern__ errno_t if_state_update(struct ifnet *, + struct if_interface_state *); +__private_extern__ void if_get_state(struct ifnet *, + struct if_interface_state *); +__private_extern__ errno_t if_probe_connectivity(struct ifnet *ifp, + u_int32_t conn_probe); +__private_extern__ void if_lqm_update(struct ifnet *, int32_t, int); +__private_extern__ void ifnet_update_sndq(struct ifclassq *, cqev_t); +__private_extern__ void ifnet_update_rcv(struct ifnet *, cqev_t); + +__private_extern__ void ifnet_flowadv(uint32_t); + +__private_extern__ errno_t ifnet_set_input_bandwidths(struct ifnet *, + struct if_bandwidths *); +__private_extern__ errno_t ifnet_set_output_bandwidths(struct ifnet *, + struct if_bandwidths *, boolean_t); +__private_extern__ u_int64_t ifnet_output_linkrate(struct ifnet *); +__private_extern__ u_int64_t ifnet_input_linkrate(struct ifnet *); + +__private_extern__ errno_t ifnet_set_input_latencies(struct ifnet *, + struct if_latencies *); +__private_extern__ errno_t ifnet_set_output_latencies(struct ifnet *, + struct if_latencies *, boolean_t); + +__private_extern__ void ifnet_clear_netagent(uuid_t); + +__private_extern__ int ifnet_set_netsignature(struct ifnet *, uint8_t, + uint8_t, uint16_t, uint8_t *); +__private_extern__ int ifnet_get_netsignature(struct ifnet *, uint8_t, + uint8_t *, uint16_t *, uint8_t *); + +__private_extern__ errno_t ifnet_framer_stub(struct ifnet *, struct mbuf **, + const struct sockaddr *, const char *, const char *, u_int32_t *, + u_int32_t *); #endif /* BSD_KERNEL_PRIVATE */ -#endif /* KERNEL_PRIVATE */ +#ifdef XNU_KERNEL_PRIVATE +/* for uuid.c */ +__private_extern__ int uuid_get_ethernet(u_int8_t *); +#endif /* XNU_KERNEL_PRIVATE */ #endif /* !_NET_IF_VAR_H_ */