| 1 | /* |
| 2 | * Copyright (c) 2012-2020 Apple Inc. All rights reserved. |
| 3 | * |
| 4 | * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ |
| 5 | * |
| 6 | * This file contains Original Code and/or Modifications of Original Code |
| 7 | * as defined in and that are subject to the Apple Public Source License |
| 8 | * Version 2.0 (the 'License'). You may not use this file except in |
| 9 | * compliance with the License. The rights granted to you under the License |
| 10 | * may not be used to create, or enable the creation or redistribution of, |
| 11 | * unlawful or unlicensed copies of an Apple operating system, or to |
| 12 | * circumvent, violate, or enable the circumvention or violation of, any |
| 13 | * terms of an Apple operating system software license agreement. |
| 14 | * |
| 15 | * Please obtain a copy of the License at |
| 16 | * http://www.opensource.apple.com/apsl/ and read it before using this file. |
| 17 | * |
| 18 | * The Original Code and all software distributed under the License are |
| 19 | * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER |
| 20 | * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, |
| 21 | * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, |
| 22 | * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. |
| 23 | * Please see the License for the specific language governing rights and |
| 24 | * limitations under the License. |
| 25 | * |
| 26 | * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ |
| 27 | */ |
| 28 | |
| 29 | #ifndef _NETINET_MPTCP_VAR_H_ |
| 30 | #define _NETINET_MPTCP_VAR_H_ |
| 31 | |
| 32 | #ifdef PRIVATE |
| 33 | #include <netinet/in.h> |
| 34 | #include <netinet/tcp.h> |
| 35 | #endif |
| 36 | |
| 37 | #ifdef BSD_KERNEL_PRIVATE |
| 38 | #include <sys/queue.h> |
| 39 | #include <sys/protosw.h> |
| 40 | #include <kern/locks.h> |
| 41 | #include <mach/boolean.h> |
| 42 | #include <netinet/mp_pcb.h> |
| 43 | #include <netinet/tcp_var.h> |
| 44 | #include <os/log.h> |
| 45 | |
| 46 | struct mpt_itf_info { |
| 47 | uint32_t ifindex; |
| 48 | uint32_t has_v4_conn:1, |
| 49 | has_v6_conn:1, |
| 50 | has_nat64_conn:1, |
| 51 | no_mptcp_support:1; |
| 52 | }; |
| 53 | |
| 54 | /* |
| 55 | * MPTCP Session |
| 56 | * |
| 57 | * This is an extension to the multipath PCB specific for MPTCP, protected by |
| 58 | * the per-PCB mpp_lock (also the socket's lock); |
| 59 | */ |
| 60 | struct mptses { |
| 61 | struct mppcb *mpte_mppcb; /* back ptr to multipath PCB */ |
| 62 | struct mptcb *mpte_mptcb; /* ptr to MPTCP PCB */ |
| 63 | TAILQ_HEAD(, mptopt) mpte_sopts; /* list of socket options */ |
| 64 | TAILQ_HEAD(, mptsub) mpte_subflows; /* list of subflows */ |
| 65 | #define MPTCP_MAX_NUM_SUBFLOWS 256 |
| 66 | uint16_t mpte_numflows; /* # of subflows in list */ |
| 67 | uint16_t mpte_nummpcapflows; /* # of MP_CAP subflows */ |
| 68 | sae_associd_t mpte_associd; /* MPTCP association ID */ |
| 69 | sae_connid_t mpte_connid_last; /* last used connection ID */ |
| 70 | |
| 71 | uint64_t mpte_time_target; |
| 72 | thread_call_t mpte_time_thread; |
| 73 | thread_call_t mpte_stop_urgency; |
| 74 | |
| 75 | uint32_t mpte_last_cellicon_set; |
| 76 | uint32_t mpte_cellicon_increments; |
| 77 | |
| 78 | union { |
| 79 | /* Source address of initial subflow */ |
| 80 | struct sockaddr _mpte_src; |
| 81 | struct sockaddr_in _mpte_src_v4; |
| 82 | struct sockaddr_in6 _mpte_src_v6; |
| 83 | } mpte_u_src; |
| 84 | #define mpte_src mpte_u_src._mpte_src |
| 85 | #define __mpte_src_v4 mpte_u_src._mpte_src_v4 |
| 86 | #define __mpte_src_v6 mpte_u_src._mpte_src_v6 |
| 87 | union { |
| 88 | /* Destination address of initial subflow */ |
| 89 | struct sockaddr _mpte_dst; |
| 90 | struct sockaddr_in _mpte_dst_v4; |
| 91 | struct sockaddr_in6 _mpte_dst_v6; |
| 92 | } mpte_u_dst; |
| 93 | #define mpte_dst mpte_u_dst._mpte_dst |
| 94 | #define __mpte_dst_v4 mpte_u_dst._mpte_dst_v4 |
| 95 | #define __mpte_dst_v6 mpte_u_dst._mpte_dst_v6 |
| 96 | |
| 97 | struct sockaddr_in mpte_sub_dst_v4; |
| 98 | struct sockaddr_in6 mpte_sub_dst_v6; |
| 99 | uint8_t sub_dst_addr_id_v4; |
| 100 | uint8_t sub_dst_addr_id_v6; |
| 101 | |
| 102 | uint16_t mpte_alternate_port; /* Alternate port for subflow establishment (network-byte-order) */ |
| 103 | |
| 104 | int mpte_epid; |
| 105 | uuid_t mpte_euuid; |
| 106 | |
| 107 | struct mptsub *mpte_active_sub; /* ptr to last active subf */ |
| 108 | uint16_t mpte_flags; /* per mptcp session flags */ |
| 109 | #define MPTE_SND_REM_ADDR 0x01 /* Send Remove_addr option */ |
| 110 | #define MPTE_SVCTYPE_CHECKED 0x02 /* Did entitlement-check for service-type */ |
| 111 | #define MPTE_FIRSTPARTY 0x04 /* First-party app used multipath_extended entitlement */ |
| 112 | #define MPTE_ACCESS_GRANTED 0x08 /* Access to cellular has been granted for this connection */ |
| 113 | #define MPTE_FORCE_ENABLE 0x10 /* For MPTCP regardless of heuristics to detect middleboxes */ |
| 114 | #define MPTE_IN_WORKLOOP 0x20 /* Are we currently inside the workloop ? */ |
| 115 | #define MPTE_WORKLOOP_RELAUNCH 0x40 /* Another event got queued, we should restart the workloop */ |
| 116 | #define MPTE_UNICAST_IP 0x80 /* New subflows are only being established towards the unicast IP in the ADD_ADDR */ |
| 117 | #define MPTE_CELL_PROHIBITED 0x100 /* Cell access has been prohibited based on signal quality */ |
| 118 | #define MPTE_FORCE_V0 0x200 /* Force MPTCP to use version 0 regradless of tcp cache */ |
| 119 | #define MPTE_FORCE_V1 0x400 /* Force MPTCP to use version 1 regradless of tcp cache */ |
| 120 | #define MPTE_ITFINFO_INIT 0x800 /* Set when the itfinfo has been initialized */ |
| 121 | uint8_t mpte_svctype; /* MPTCP Service type */ |
| 122 | uint8_t mpte_lost_aid; /* storing lost address id */ |
| 123 | uint8_t mpte_addrid_last; /* storing address id parm */ |
| 124 | |
| 125 | #define MPTE_ITFINFO_SIZE 4 |
| 126 | uint32_t mpte_itfinfo_size; |
| 127 | struct mpt_itf_info _mpte_itfinfo[MPTE_ITFINFO_SIZE]; |
| 128 | struct mpt_itf_info *mpte_itfinfo; |
| 129 | |
| 130 | struct mbuf *mpte_reinjectq; |
| 131 | |
| 132 | /* The below is used for stats */ |
| 133 | uint32_t mpte_subflow_switches; /* Number of subflow-switches in sending */ |
| 134 | uint32_t mpte_used_cell:1, |
| 135 | mpte_used_wifi:1, |
| 136 | mpte_initial_cell:1, |
| 137 | mpte_triggered_cell, |
| 138 | mpte_handshake_success:1, |
| 139 | mpte_last_added_addr_is_v4:1; |
| 140 | |
| 141 | struct mptcp_itf_stats mpte_itfstats[MPTCP_ITFSTATS_SIZE]; |
| 142 | uint64_t mpte_init_txbytes __attribute__((aligned(8))); |
| 143 | uint64_t mpte_init_rxbytes __attribute__((aligned(8))); |
| 144 | }; |
| 145 | |
| 146 | static inline struct socket * |
| 147 | mptetoso(struct mptses *mpte) |
| 148 | { |
| 149 | return mpte->mpte_mppcb->mpp_socket; |
| 150 | } |
| 151 | |
| 152 | static inline struct mptses * |
| 153 | mptompte(struct mppcb *mp) |
| 154 | { |
| 155 | return (struct mptses *)mp->mpp_pcbe; |
| 156 | } |
| 157 | |
| 158 | static inline struct mptses * |
| 159 | mpsotompte(struct socket *so) |
| 160 | { |
| 161 | return mptompte(mp: mpsotomppcb(mp_so: so)); |
| 162 | } |
| 163 | |
| 164 | static inline boolean_t |
| 165 | mpp_try_lock(struct mppcb *mp) |
| 166 | { |
| 167 | if (!lck_mtx_try_lock(lck: &mp->mpp_lock)) { |
| 168 | return false; |
| 169 | } |
| 170 | |
| 171 | VERIFY(!(mp->mpp_flags & MPP_INSIDE_OUTPUT)); |
| 172 | VERIFY(!(mp->mpp_flags & MPP_INSIDE_INPUT)); |
| 173 | VERIFY(!(mp->mpp_flags & MPP_INSIDE_SETGETOPT)); |
| 174 | |
| 175 | return true; |
| 176 | } |
| 177 | |
| 178 | static inline void |
| 179 | mpp_lock(struct mppcb *mp) |
| 180 | { |
| 181 | lck_mtx_lock(lck: &mp->mpp_lock); |
| 182 | VERIFY(!(mp->mpp_flags & MPP_INSIDE_OUTPUT)); |
| 183 | VERIFY(!(mp->mpp_flags & MPP_INSIDE_INPUT)); |
| 184 | VERIFY(!(mp->mpp_flags & MPP_INSIDE_SETGETOPT)); |
| 185 | } |
| 186 | |
| 187 | static inline void |
| 188 | mpp_unlock(struct mppcb *mp) |
| 189 | { |
| 190 | VERIFY(!(mp->mpp_flags & MPP_INSIDE_OUTPUT)); |
| 191 | VERIFY(!(mp->mpp_flags & MPP_INSIDE_INPUT)); |
| 192 | VERIFY(!(mp->mpp_flags & MPP_INSIDE_SETGETOPT)); |
| 193 | lck_mtx_unlock(lck: &mp->mpp_lock); |
| 194 | } |
| 195 | |
| 196 | static inline lck_mtx_t * |
| 197 | mpp_getlock(struct mppcb *mp, int flags) |
| 198 | { |
| 199 | if (flags & PR_F_WILLUNLOCK) { |
| 200 | VERIFY(!(mp->mpp_flags & MPP_INSIDE_OUTPUT)); |
| 201 | VERIFY(!(mp->mpp_flags & MPP_INSIDE_INPUT)); |
| 202 | VERIFY(!(mp->mpp_flags & MPP_INSIDE_SETGETOPT)); |
| 203 | } |
| 204 | |
| 205 | return &mp->mpp_lock; |
| 206 | } |
| 207 | |
| 208 | static inline int |
| 209 | mptcp_subflow_cwnd_space(struct socket *so) |
| 210 | { |
| 211 | struct tcpcb *tp = sototcpcb(so); |
| 212 | int cwnd = (int)(MIN(tp->snd_wnd, tp->snd_cwnd) - (so->so_snd.sb_cc)); |
| 213 | |
| 214 | return MIN(cwnd, sbspace(&so->so_snd)); |
| 215 | } |
| 216 | |
| 217 | static inline bool |
| 218 | mptcp_subflows_need_backup_flag(struct mptses *mpte) |
| 219 | { |
| 220 | return mpte->mpte_svctype < MPTCP_SVCTYPE_AGGREGATE || |
| 221 | mpte->mpte_svctype == MPTCP_SVCTYPE_PURE_HANDOVER; |
| 222 | } |
| 223 | |
| 224 | /* |
| 225 | * MPTCP socket options |
| 226 | */ |
| 227 | struct mptopt { |
| 228 | TAILQ_ENTRY(mptopt) mpo_entry; /* glue to other options */ |
| 229 | uint32_t mpo_flags; /* see flags below */ |
| 230 | int mpo_level; /* sopt_level */ |
| 231 | int mpo_name; /* sopt_name */ |
| 232 | int mpo_intval; /* sopt_val */ |
| 233 | }; |
| 234 | |
| 235 | #define MPOF_ATTACHED 0x1 /* attached to MP socket */ |
| 236 | #define MPOF_SUBFLOW_OK 0x2 /* can be issued on subflow socket */ |
| 237 | #define MPOF_INTERIM 0x4 /* has not been issued on any subflow */ |
| 238 | |
| 239 | /* |
| 240 | * MPTCP subflow |
| 241 | */ |
| 242 | struct mptsub { |
| 243 | TAILQ_ENTRY(mptsub) mpts_entry; /* glue to peer subflows */ |
| 244 | uint32_t mpts_refcnt; /* reference count */ |
| 245 | uint32_t mpts_flags; /* see flags below */ |
| 246 | uint32_t mpts_evctl; /* subflow control events */ |
| 247 | sae_connid_t mpts_connid; /* subflow connection ID */ |
| 248 | int mpts_oldintval; /* sopt_val before sosetopt */ |
| 249 | struct mptses *mpts_mpte; /* back ptr to MPTCP session */ |
| 250 | struct socket *mpts_socket; /* subflow socket */ |
| 251 | struct sockaddr *mpts_src; /* source address */ |
| 252 | |
| 253 | union { |
| 254 | /* destination address */ |
| 255 | struct sockaddr _mpts_dst; |
| 256 | struct sockaddr_in _mpts_dst_v4; |
| 257 | struct sockaddr_in6 _mpts_dst_v6; |
| 258 | } mpts_u_dst; |
| 259 | #define mpts_dst mpts_u_dst._mpts_dst |
| 260 | #define __mpts_dst_v4 mpts_u_dst._mpts_dst_v4 |
| 261 | #define __mpts_dst_v6 mpts_u_dst._mpts_dst_v6 |
| 262 | u_int32_t mpts_rel_seq; /* running count of subflow # */ |
| 263 | u_int32_t mpts_iss; /* Initial sequence number, taking TFO into account */ |
| 264 | u_int32_t mpts_ifscope; /* scoped to the interface */ |
| 265 | uint32_t mpts_probesoon; /* send probe after probeto */ |
| 266 | uint32_t mpts_probecnt; /* number of probes sent */ |
| 267 | uint32_t mpts_maxseg; /* cached value of t_maxseg */ |
| 268 | }; |
| 269 | |
| 270 | /* |
| 271 | * Valid values for mpts_flags. In particular: |
| 272 | * |
| 273 | * - MP_CAPABLE means that the connection is successfully established as |
| 274 | * MPTCP and data transfer may occur, but is not yet ready for multipath- |
| 275 | * related semantics until MP_READY. I.e. if this is on the first subflow, |
| 276 | * it causes the MPTCP socket to transition to a connected state, except |
| 277 | * that additional subflows will not be established; they will be marked |
| 278 | * with PENDING and will be processed when the first subflow is marked |
| 279 | * with MP_READY. |
| 280 | * |
| 281 | * - MP_READY implies that an MP_CAPABLE connection has been confirmed as |
| 282 | * an MPTCP connection. See notes above. |
| 283 | * |
| 284 | * - MP_DEGRADED implies that the connection has lost its MPTCP capabilities |
| 285 | * but data transfer on the MPTCP socket is unaffected. Any existing |
| 286 | * PENDING subflows will be disconnected, and further attempts to connect |
| 287 | * additional subflows will be rejected. |
| 288 | * |
| 289 | * Note that these are per-subflow flags. The setting and clearing of MP_READY |
| 290 | * reflects the state of the MPTCP connection with regards to its multipath |
| 291 | * semantics, via the MPTCPF_JOIN_READY flag. Until that flag is set (meaning |
| 292 | * until at least a subflow is marked with MP_READY), further connectx(2) |
| 293 | * attempts to join will be queued. When the flag is cleared (after it has |
| 294 | * been set), further connectx(2) will fail (and existing queued ones will be |
| 295 | * aborted) and the MPTCP connection loses all of its multipath semantics. |
| 296 | * |
| 297 | * Keep in sync with bsd/dev/dtrace/scripts/mptcp.d. |
| 298 | */ |
| 299 | #define MPTSF_CONNECTING 0x00000002 /* connection was attempted */ |
| 300 | #define MPTSF_CONNECT_PENDING 0x00000004 /* will connect when MPTCP is ready */ |
| 301 | #define MPTSF_CONNECTED 0x00000008 /* connection is established */ |
| 302 | #define MPTSF_DISCONNECTING 0x00000010 /* disconnection was attempted */ |
| 303 | #define MPTSF_DISCONNECTED 0x00000020 /* has been disconnected */ |
| 304 | #define MPTSF_MP_CAPABLE 0x00000040 /* connected as a MPTCP subflow */ |
| 305 | #define MPTSF_MP_READY 0x00000080 /* MPTCP has been confirmed */ |
| 306 | #define MPTSF_MP_DEGRADED 0x00000100 /* has lost its MPTCP capabilities */ |
| 307 | #define MPTSF_PREFERRED 0x00000200 /* primary/preferred subflow */ |
| 308 | #define MPTSF_SOPT_OLDVAL 0x00000400 /* old option value is valid */ |
| 309 | #define MPTSF_SOPT_INPROG 0x00000800 /* sosetopt in progress */ |
| 310 | #define MPTSF_FAILINGOVER 0x00001000 /* subflow not used for output */ |
| 311 | #define MPTSF_ACTIVE 0x00002000 /* subflow currently in use */ |
| 312 | #define MPTSF_MPCAP_CTRSET 0x00004000 /* mpcap counter */ |
| 313 | #define MPTSF_CLOSED 0x00008000 /* soclose_locked has been called on this subflow */ |
| 314 | #define MPTSF_TFO_REQD 0x00010000 /* TFO requested */ |
| 315 | #define MPTSF_CLOSE_REQD 0x00020000 /* A close has been requested from NECP */ |
| 316 | #define MPTSF_INITIAL_SUB 0x00040000 /* This is the initial subflow */ |
| 317 | #define MPTSF_READ_STALL 0x00080000 /* A read-stall has been detected */ |
| 318 | #define MPTSF_WRITE_STALL 0x00100000 /* A write-stall has been detected */ |
| 319 | #define MPTSF_FULLY_ESTABLISHED 0x00200000 /* Subflow is fully established and it has been confirmed |
| 320 | * whether or not it supports MPTCP. |
| 321 | * No need for further middlebox-detection. |
| 322 | */ |
| 323 | #define MPTSF_CELLICON_SET 0x00400000 /* This subflow set the cellicon */ |
| 324 | |
| 325 | /* |
| 326 | * MPTCP states |
| 327 | * Keep in sync with bsd/dev/dtrace/scripts/mptcp.d |
| 328 | */ |
| 329 | typedef enum mptcp_state { |
| 330 | MPTCPS_CLOSED = 0, /* closed */ |
| 331 | MPTCPS_LISTEN = 1, /* not yet implemented */ |
| 332 | MPTCPS_ESTABLISHED = 2, /* MPTCP connection established */ |
| 333 | MPTCPS_CLOSE_WAIT = 3, /* rcvd DFIN, waiting for close */ |
| 334 | MPTCPS_FIN_WAIT_1 = 4, /* have closed, sent DFIN */ |
| 335 | MPTCPS_CLOSING = 5, /* closed xchd DFIN, waiting DFIN ACK */ |
| 336 | MPTCPS_LAST_ACK = 6, /* had DFIN and close; await DFIN ACK */ |
| 337 | MPTCPS_FIN_WAIT_2 = 7, /* have closed, DFIN is acked */ |
| 338 | MPTCPS_TIME_WAIT = 8, /* in 2*MSL quiet wait after close */ |
| 339 | MPTCPS_TERMINATE = 9, /* terminal state */ |
| 340 | } mptcp_state_t; |
| 341 | |
| 342 | /* |
| 343 | * WiFi Quality states from MPTCP's perspective |
| 344 | */ |
| 345 | typedef enum mptcp_wifi_quality { |
| 346 | MPTCP_WIFI_QUALITY_GOOD, |
| 347 | MPTCP_WIFI_QUALITY_BAD, |
| 348 | MPTCP_WIFI_QUALITY_UNSURE, |
| 349 | } mptcp_wifi_quality_t; |
| 350 | |
| 351 | typedef u_int64_t mptcp_key_t; |
| 352 | typedef u_int32_t mptcp_token_t; |
| 353 | typedef u_int8_t mptcp_addr_id; |
| 354 | |
| 355 | |
| 356 | /* Address ID list */ |
| 357 | struct mptcp_subf_auth_entry { |
| 358 | LIST_ENTRY(mptcp_subf_auth_entry) msae_next; |
| 359 | u_int32_t msae_laddr_rand; /* Local nonce */ |
| 360 | u_int32_t msae_raddr_rand; /* Remote nonce */ |
| 361 | mptcp_addr_id msae_laddr_id; /* Local addr ID */ |
| 362 | mptcp_addr_id msae_raddr_id; /* Remote addr ID */ |
| 363 | }; |
| 364 | |
| 365 | /* |
| 366 | * MPTCP Protocol Control Block |
| 367 | * |
| 368 | * Protected by per-MPTCP mpt_lock. |
| 369 | * Keep in sync with bsd/dev/dtrace/scripts/mptcp.d. |
| 370 | */ |
| 371 | struct mptcb { |
| 372 | struct mptses *mpt_mpte; /* back ptr to MPTCP session */ |
| 373 | mptcp_state_t mpt_state; /* MPTCP state */ |
| 374 | uint32_t mpt_flags; /* see flags below */ |
| 375 | uint8_t mpt_version; /* MPTCP proto version */ |
| 376 | u_short mpt_softerror; /* error not yet reported */ |
| 377 | /* |
| 378 | * Authentication and metadata invariants |
| 379 | */ |
| 380 | mptcp_key_t mpt_localkey; /* in network byte order */ |
| 381 | mptcp_key_t mpt_remotekey; /* in network byte order */ |
| 382 | mptcp_token_t mpt_localtoken; /* HMAC SHA1 of local key */ |
| 383 | mptcp_token_t mpt_remotetoken; /* HMAC SHA1 of remote key */ |
| 384 | |
| 385 | /* |
| 386 | * Timer vars for scenarios where subflow level acks arrive, but |
| 387 | * Data ACKs do not. |
| 388 | */ |
| 389 | int mpt_rxtshift; /* num of consecutive retrans */ |
| 390 | uint64_t mpt_rxtstart; /* time at which rxt started */ |
| 391 | uint64_t mpt_rtseq; /* seq # being tracked */ |
| 392 | uint64_t mpt_timewait; /* timewait */ |
| 393 | uint32_t mpt_timer_vals; /* timer related values */ |
| 394 | /* |
| 395 | * Sending side |
| 396 | */ |
| 397 | uint64_t mpt_snduna; /* DSN of last unacked byte */ |
| 398 | uint64_t mpt_sndnxt; /* DSN of next byte to send */ |
| 399 | uint64_t mpt_sndmax; /* DSN of max byte sent */ |
| 400 | uint64_t mpt_local_idsn; /* First byte's DSN */ |
| 401 | uint32_t mpt_sndwnd; |
| 402 | uint64_t mpt_sndwl1; |
| 403 | uint64_t mpt_sndwl2; |
| 404 | /* |
| 405 | * Receiving side |
| 406 | */ |
| 407 | uint64_t mpt_rcvnxt; /* Next expected DSN */ |
| 408 | uint64_t mpt_remote_idsn; /* Peer's IDSN */ |
| 409 | uint64_t mpt_rcvadv; |
| 410 | uint32_t mpt_rcvwnd; |
| 411 | LIST_HEAD(, mptcp_subf_auth_entry) mpt_subauth_list; /* address IDs */ |
| 412 | /* |
| 413 | * Fastclose |
| 414 | */ |
| 415 | uint64_t mpt_dsn_at_csum_fail; /* MPFail Opt DSN */ |
| 416 | uint32_t mpt_ssn_at_csum_fail; /* MPFail Subflow Seq */ |
| 417 | /* |
| 418 | * Zombie handling |
| 419 | */ |
| 420 | #define MPT_GC_TICKS (30) |
| 421 | #define MPT_GC_TICKS_FAST (10) |
| 422 | int32_t mpt_gc_ticks; /* Used for zombie deletion */ |
| 423 | |
| 424 | uint32_t mpt_notsent_lowat; /* TCP_NOTSENT_LOWAT support */ |
| 425 | |
| 426 | struct tsegqe_head mpt_segq; |
| 427 | uint32_t mpt_reassqlen; /* length of reassembly queue */ |
| 428 | }; |
| 429 | |
| 430 | /* valid values for mpt_flags (see also notes on mpts_flags above) */ |
| 431 | #define MPTCPF_CHECKSUM 0x001 /* checksum DSS option */ |
| 432 | #define MPTCPF_FALLBACK_TO_TCP 0x002 /* Fallback to TCP */ |
| 433 | #define MPTCPF_JOIN_READY 0x004 /* Ready to start 2 or more subflows */ |
| 434 | #define MPTCPF_RECVD_MPFAIL 0x008 /* Received MP_FAIL option */ |
| 435 | #define MPTCPF_SND_64BITDSN 0x010 /* Send full 64-bit DSN */ |
| 436 | #define MPTCPF_SND_64BITACK 0x020 /* Send 64-bit ACK response */ |
| 437 | #define MPTCPF_RCVD_64BITACK 0x040 /* Received 64-bit Data ACK */ |
| 438 | #define MPTCPF_POST_FALLBACK_SYNC 0x080 /* Post fallback resend data */ |
| 439 | #define MPTCPF_FALLBACK_HEURISTIC 0x100 /* Send SYN without MP_CAPABLE due to heuristic */ |
| 440 | #define MPTCPF_HEURISTIC_TRAC 0x200 /* Tracked this connection in the heuristics as a failure */ |
| 441 | #define MPTCPF_REASS_INPROG 0x400 /* Reassembly is in progress */ |
| 442 | |
| 443 | /* valid values for mpt_timer_vals */ |
| 444 | #define MPTT_REXMT 0x01 /* Starting Retransmit Timer */ |
| 445 | #define MPTT_TW 0x02 /* Starting Timewait Timer */ |
| 446 | #define MPTT_FASTCLOSE 0x04 /* Starting Fastclose wait timer */ |
| 447 | |
| 448 | /* events for close FSM */ |
| 449 | #define MPCE_CLOSE 0x1 |
| 450 | #define MPCE_RECV_DATA_ACK 0x2 |
| 451 | #define MPCE_RECV_DATA_FIN 0x4 |
| 452 | |
| 453 | /* mptcb manipulation */ |
| 454 | static inline struct mptcb * |
| 455 | tptomptp(struct tcpcb *tp) |
| 456 | { |
| 457 | return tp->t_mptcb; |
| 458 | } |
| 459 | |
| 460 | /* |
| 461 | * MPTCP control block and state structures are allocated along with |
| 462 | * the MP protocol control block; the folllowing represents the layout. |
| 463 | */ |
| 464 | struct mpp_mtp { |
| 465 | struct mppcb mpp; /* Multipath PCB */ |
| 466 | struct mptses mpp_ses; /* MPTCP session */ |
| 467 | struct mptcb mtcb; /* MPTCP PCB */ |
| 468 | }; |
| 469 | |
| 470 | #ifdef SYSCTL_DECL |
| 471 | SYSCTL_DECL(_net_inet_mptcp); |
| 472 | #endif /* SYSCTL_DECL */ |
| 473 | |
| 474 | extern struct mppcbinfo mtcbinfo; |
| 475 | extern struct pr_usrreqs mptcp_usrreqs; |
| 476 | extern os_log_t mptcp_log_handle; |
| 477 | |
| 478 | /* Encryption algorithm related definitions */ |
| 479 | #define HMAC_TRUNCATED_SYNACK 8 |
| 480 | #define HMAC_TRUNCATED_ACK 20 |
| 481 | #define HMAC_TRUNCATED_ADD_ADDR 8 |
| 482 | |
| 483 | /* Mask to obtain 32-bit portion of data sequence number */ |
| 484 | #define MPTCP_DATASEQ_LOW32_MASK (0xffffffff) |
| 485 | #define MPTCP_DATASEQ_LOW32(seq) (seq & MPTCP_DATASEQ_LOW32_MASK) |
| 486 | |
| 487 | /* Mask to obtain upper 32-bit portion of data sequence number */ |
| 488 | #define MPTCP_DATASEQ_HIGH32_MASK (0xffffffff00000000) |
| 489 | #define MPTCP_DATASEQ_HIGH32(seq) (seq & MPTCP_DATASEQ_HIGH32_MASK) |
| 490 | |
| 491 | /* Mask to obtain 32-bit portion of data ack */ |
| 492 | #define MPTCP_DATAACK_LOW32_MASK (0xffffffff) |
| 493 | #define MPTCP_DATAACK_LOW32(ack) (ack & MPTCP_DATAACK_LOW32_MASK) |
| 494 | |
| 495 | /* |
| 496 | * x is the 64-bit data sequence number, y the 32-bit data seq number to be |
| 497 | * extended. z is y extended to the appropriate 64-bit value. |
| 498 | * This algorithm is based on the fact that subflow level window sizes are |
| 499 | * at the maximum 2**30 (in reality, they are a lot lesser). A high throughput |
| 500 | * application sending on a large number of subflows can in theory have very |
| 501 | * large MPTCP level send and receive windows. In which case, 64 bit DSNs |
| 502 | * must be sent in place of 32 bit DSNs on wire. For us, with 2 subflows at |
| 503 | * 512K each, sequence wraparound detection can be done by checking whether |
| 504 | * the 32-bit value obtained on wire is 2**31 bytes apart from the stored |
| 505 | * lower 32-bits of the Data Sequence Number. Bogus DSNs are dropped by |
| 506 | * comparing against rwnd. Bogus DSNs within rwnd cannot be protected against |
| 507 | * and are as weak as bogus TCP sequence numbers. |
| 508 | */ |
| 509 | #define MPTCP_EXTEND_DSN(x, y, z) { \ |
| 510 | if ((MPTCP_DATASEQ_LOW32(x) > y) && \ |
| 511 | ((((u_int32_t)MPTCP_DATASEQ_LOW32(x)) - (u_int32_t)y) >= \ |
| 512 | (u_int32_t)(1U << 31))) { \ |
| 513 | /* \ |
| 514 | * y wrapped around and x and y are 2**31 bytes apart \ |
| 515 | */ \ |
| 516 | z = MPTCP_DATASEQ_HIGH32(x) + 0x100000000; \ |
| 517 | z |= y; \ |
| 518 | } else if ((MPTCP_DATASEQ_LOW32(x) < y) && \ |
| 519 | (((u_int32_t)y - \ |
| 520 | ((u_int32_t)MPTCP_DATASEQ_LOW32(x))) >= \ |
| 521 | (u_int32_t)(1U << 31))) { \ |
| 522 | /* \ |
| 523 | * x wrapped around and x and y are 2**31 apart \ |
| 524 | */ \ |
| 525 | z = MPTCP_DATASEQ_HIGH32(x) - 0x100000000; \ |
| 526 | z |= y; \ |
| 527 | } else { \ |
| 528 | z = MPTCP_DATASEQ_HIGH32(x) | y; \ |
| 529 | } \ |
| 530 | } |
| 531 | |
| 532 | extern int mptcp_enable; /* Multipath TCP */ |
| 533 | extern int mptcp_mpcap_retries; /* Multipath TCP retries */ |
| 534 | extern int mptcp_join_retries; /* Multipath TCP Join retries */ |
| 535 | extern int mptcp_dss_csum; /* Multipath DSS Option checksum */ |
| 536 | extern int mptcp_fail_thresh; /* Multipath failover thresh of retransmits */ |
| 537 | extern int mptcp_subflow_keeptime; /* Multipath subflow TCP_KEEPALIVE opt */ |
| 538 | extern int mptcp_developer_mode; /* Allow aggregation mode */ |
| 539 | extern uint32_t mptcp_cellicon_refcount; |
| 540 | extern uint32_t mptcp_enable_v1; |
| 541 | |
| 542 | #define MPTCP_CELLICON_TOGGLE_RATE (5 * TCP_RETRANSHZ) /* Only toggle every 5 seconds */ |
| 543 | |
| 544 | extern int tcp_jack_rxmt; /* Join ACK retransmission value in msecs */ |
| 545 | |
| 546 | extern int mptcp_reass_total_qlen; |
| 547 | |
| 548 | __BEGIN_DECLS |
| 549 | extern void mptcp_init(struct protosw *, struct domain *); |
| 550 | extern int mptcp_ctloutput(struct socket *, struct sockopt *); |
| 551 | extern int mptcp_session_create(struct mppcb *); |
| 552 | extern boolean_t mptcp_ok_to_create_subflows(struct mptcb *mp_tp); |
| 553 | extern void mptcp_check_subflows_and_add(struct mptses *mpte); |
| 554 | extern void mptcp_check_subflows_and_remove(struct mptses *mpte); |
| 555 | extern void mptcpstats_inc_switch(struct mptses *mpte, const struct mptsub *mpts); |
| 556 | extern void mptcpstats_update(struct mptcp_itf_stats *stats, const struct mptsub *mpts); |
| 557 | extern int mptcpstats_get_index_by_ifindex(struct mptcp_itf_stats *stats, u_short ifindex, boolean_t create); |
| 558 | extern struct mptses *mptcp_drop(struct mptses *mpte, struct mptcb *mp_tp, u_short errno); |
| 559 | extern struct mptses *mptcp_close(struct mptses *, struct mptcb *); |
| 560 | extern int mptcp_lock(struct socket *, int, void *); |
| 561 | extern int mptcp_unlock(struct socket *, int, void *); |
| 562 | extern lck_mtx_t *mptcp_getlock(struct socket *, int); |
| 563 | extern void mptcp_subflow_workloop(struct mptses *); |
| 564 | |
| 565 | extern void mptcp_sched_create_subflows(struct mptses *); |
| 566 | |
| 567 | extern void mptcp_finish_usrclosed(struct mptses *mpte); |
| 568 | extern struct mptopt *mptcp_sopt_alloc(zalloc_flags_t); |
| 569 | extern const char *mptcp_sopt2str(int, int); |
| 570 | extern void mptcp_sopt_free(struct mptopt *); |
| 571 | extern void mptcp_sopt_insert(struct mptses *, struct mptopt *); |
| 572 | extern void mptcp_sopt_remove(struct mptses *, struct mptopt *); |
| 573 | extern struct mptopt *mptcp_sopt_find(struct mptses *, struct sockopt *); |
| 574 | |
| 575 | extern int mptcp_subflow_add(struct mptses *, struct sockaddr *, |
| 576 | struct sockaddr *, uint32_t, sae_connid_t *); |
| 577 | extern void mptcp_subflow_del(struct mptses *, struct mptsub *); |
| 578 | |
| 579 | extern void mptcp_handle_input(struct socket *so); |
| 580 | #define MPTCP_SUBOUT_PROBING 0x01 |
| 581 | extern int mptcp_subflow_output(struct mptses *mpte, struct mptsub *mpts, int flags); |
| 582 | extern void mptcp_clean_reinjectq(struct mptses *mpte); |
| 583 | extern void mptcp_subflow_shutdown(struct mptses *, struct mptsub *); |
| 584 | extern void mptcp_subflow_disconnect(struct mptses *, struct mptsub *); |
| 585 | extern int mptcp_subflow_sosetopt(struct mptses *, struct mptsub *, |
| 586 | struct mptopt *); |
| 587 | extern int mptcp_subflow_sogetopt(struct mptses *, struct socket *, |
| 588 | struct mptopt *); |
| 589 | |
| 590 | extern void mptcp_input(struct mptses *, struct mbuf *); |
| 591 | extern boolean_t mptcp_can_send_more(struct mptcb *mp_tp, boolean_t ignore_reinject); |
| 592 | extern int mptcp_output(struct mptses *); |
| 593 | extern void mptcp_close_fsm(struct mptcb *, uint32_t); |
| 594 | |
| 595 | extern void mptcp_hmac_sha1(mptcp_key_t, mptcp_key_t, u_int32_t, u_int32_t, |
| 596 | u_char*); |
| 597 | extern void mptcp_hmac_sha256(mptcp_key_t, mptcp_key_t, u_char*, uint16_t, |
| 598 | u_char*); |
| 599 | extern void mptcp_get_mpjoin_hmac(mptcp_addr_id, struct mptcb *, u_char *, uint8_t); |
| 600 | extern void mptcp_get_rands(mptcp_addr_id, struct mptcb *, u_int32_t *, |
| 601 | u_int32_t *); |
| 602 | extern void mptcp_set_raddr_rand(mptcp_addr_id, struct mptcb *, mptcp_addr_id, |
| 603 | u_int32_t); |
| 604 | extern int mptcp_init_remote_parms(struct mptcb *); |
| 605 | extern boolean_t mptcp_ok_to_keepalive(struct mptcb *); |
| 606 | extern void mptcp_insert_dsn(struct mppcb *, struct mbuf *); |
| 607 | extern void mptcp_output_getm_dsnmap32(struct socket *so, int off, |
| 608 | uint32_t *dsn, uint32_t *relseq, |
| 609 | uint16_t *data_len, uint16_t *dss_csum); |
| 610 | extern void mptcp_output_getm_dsnmap64(struct socket *so, int off, |
| 611 | uint64_t *dsn, uint32_t *relseq, |
| 612 | uint16_t *data_len, uint16_t *dss_csum); |
| 613 | extern void mptcp_output_getm_data_level_details(struct socket *so, int off, |
| 614 | uint16_t *data_len, uint16_t *dss_csum); |
| 615 | extern void mptcp_act_on_txfail(struct socket *); |
| 616 | extern struct mptsub *mptcp_get_subflow(struct mptses *mpte, struct mptsub **preferred); |
| 617 | extern int mptcp_get_map_for_dsn(struct socket *so, uint64_t dsn_fail, uint32_t *tcp_seq); |
| 618 | extern int32_t mptcp_adj_sendlen(struct socket *so, int32_t off); |
| 619 | extern void mptcp_sbrcv_grow(struct mptcb *mp_tp); |
| 620 | extern int32_t mptcp_sbspace(struct mptcb *); |
| 621 | extern void mptcp_notify_mpready(struct socket *); |
| 622 | extern void mptcp_notify_mpfail(struct socket *); |
| 623 | extern void mptcp_notify_close(struct socket *); |
| 624 | extern boolean_t mptcp_no_rto_spike(struct socket*); |
| 625 | extern int mptcp_set_notsent_lowat(struct mptses *mpte, int optval); |
| 626 | extern u_int32_t mptcp_get_notsent_lowat(struct mptses *mpte); |
| 627 | extern int mptcp_notsent_lowat_check(struct socket *so); |
| 628 | extern void mptcp_ask_symptoms(struct mptses *mpte); |
| 629 | extern void mptcp_control_register(void); |
| 630 | extern mptcp_wifi_quality_t mptcp_wifi_quality_for_session(struct mptses *mpte); |
| 631 | extern boolean_t symptoms_is_wifi_lossy(void); |
| 632 | extern void mptcp_session_necp_cb(void *, int, uint32_t, uint32_t, bool *); |
| 633 | extern struct sockaddr *mptcp_get_session_dst(struct mptses *mpte, |
| 634 | boolean_t has_v6, boolean_t has_v4); |
| 635 | extern void mptcp_set_restrictions(struct socket *mp_so); |
| 636 | extern void mptcp_clear_cellicon(void); |
| 637 | extern void mptcp_unset_cellicon(struct mptses *mpte, struct mptsub *mpts, uint32_t val); |
| 638 | extern void mptcp_reset_rexmit_state(struct tcpcb *tp); |
| 639 | extern void mptcp_reset_keepalive(struct tcpcb *tp); |
| 640 | extern int mptcp_validate_csum(struct tcpcb *tp, struct mbuf *m, uint64_t dsn, |
| 641 | uint32_t sseq, uint16_t dlen, uint16_t csum, int dfin); |
| 642 | __END_DECLS |
| 643 | |
| 644 | #endif /* BSD_KERNEL_PRIVATE */ |
| 645 | #ifdef PRIVATE |
| 646 | |
| 647 | typedef struct mptcp_flow { |
| 648 | uint64_t flow_len; |
| 649 | uint64_t flow_tcpci_offset; |
| 650 | uint32_t flow_flags; |
| 651 | sae_connid_t flow_cid; |
| 652 | struct sockaddr_storage flow_src; |
| 653 | struct sockaddr_storage flow_dst; |
| 654 | uint32_t flow_relseq; /* last subflow rel seq# */ |
| 655 | int32_t flow_soerror; /* subflow level error */ |
| 656 | uint32_t flow_probecnt; /* number of probes sent */ |
| 657 | conninfo_tcp_t flow_ci; /* must be the last field */ |
| 658 | } mptcp_flow_t; |
| 659 | |
| 660 | typedef struct conninfo_mptcp { |
| 661 | uint64_t mptcpci_len; |
| 662 | uint64_t mptcpci_flow_offset; /* offsetof first flow */ |
| 663 | uint64_t mptcpci_nflows; /* number of subflows */ |
| 664 | uint32_t mptcpci_state; /* MPTCP level state */ |
| 665 | uint32_t mptcpci_mpte_flags; /* Session flags */ |
| 666 | uint32_t mptcpci_flags; /* MPTCB flags */ |
| 667 | uint32_t mptcpci_ltoken; /* local token */ |
| 668 | uint32_t mptcpci_rtoken; /* remote token */ |
| 669 | uint32_t mptcpci_notsent_lowat; /* NOTSENT_LOWAT */ |
| 670 | |
| 671 | /* Send side */ |
| 672 | uint64_t mptcpci_snduna; /* DSN of last unacked byte */ |
| 673 | uint64_t mptcpci_sndnxt; /* DSN of next byte to send */ |
| 674 | uint64_t mptcpci_sndmax; /* DSN of max byte sent */ |
| 675 | uint64_t mptcpci_lidsn; /* Local IDSN */ |
| 676 | uint32_t mptcpci_sndwnd; /* Send window snapshot */ |
| 677 | |
| 678 | /* Receive side */ |
| 679 | uint64_t mptcpci_rcvnxt; /* Next expected DSN */ |
| 680 | uint64_t mptcpci_rcvatmark; /* Session level rcvnxt */ |
| 681 | uint64_t mptcpci_ridsn; /* Peer's IDSN */ |
| 682 | uint32_t mptcpci_rcvwnd; /* Receive window */ |
| 683 | |
| 684 | uint8_t mptcpci_mpte_addrid; /* last addr id */ |
| 685 | |
| 686 | mptcp_flow_t mptcpci_flows[1]; |
| 687 | } conninfo_mptcp_t; |
| 688 | |
| 689 | /* Use SymptomsD notifications of wifi and cell status in subflow selection */ |
| 690 | #define MPTCP_KERN_CTL_NAME "com.apple.network.advisory" |
| 691 | typedef struct symptoms_advisory { |
| 692 | union { |
| 693 | uint32_t sa_nwk_status_int; |
| 694 | struct { |
| 695 | union { |
| 696 | #define 0x0000 |
| 697 | #define SYMPTOMS_ADVISORY_USEAPP 0xFFFF /* Very ugly workaround to avoid breaking backwards compatibility - ToDo: Fix it in +1 */ |
| 698 | uint16_t sa_nwk_status; |
| 699 | struct { |
| 700 | #define SYMPTOMS_ADVISORY_WIFI_BAD 0x01 |
| 701 | #define SYMPTOMS_ADVISORY_WIFI_OK 0x02 |
| 702 | uint8_t sa_wifi_status; |
| 703 | #define SYMPTOMS_ADVISORY_CELL_BAD 0x01 |
| 704 | #define SYMPTOMS_ADVISORY_CELL_OK 0x02 |
| 705 | uint8_t sa_cell_status; |
| 706 | }; |
| 707 | }; |
| 708 | uint16_t sa_unused; |
| 709 | }; |
| 710 | }; |
| 711 | } symptoms_advisory_t; |
| 712 | |
| 713 | #define -75 |
| 714 | struct mptcp_symptoms_answer { |
| 715 | struct symptoms_advisory advisory; |
| 716 | uuid_t uuid; |
| 717 | int32_t ; |
| 718 | }; |
| 719 | |
| 720 | struct mptcp_symptoms_ask_uuid { |
| 721 | uint32_t cmd; |
| 722 | #define MPTCP_SYMPTOMS_ASK_UUID 1 |
| 723 | uuid_t uuid; |
| 724 | uint32_t priority; |
| 725 | #define MPTCP_SYMPTOMS_UNKNOWN 0 |
| 726 | #define MPTCP_SYMPTOMS_BACKGROUND 1 |
| 727 | #define MPTCP_SYMPTOMS_FOREGROUND 2 |
| 728 | }; |
| 729 | |
| 730 | struct kev_mptcp_data { |
| 731 | int value; |
| 732 | }; |
| 733 | |
| 734 | #endif /* PRIVATE */ |
| 735 | #endif /* _NETINET_MPTCP_VAR_H_ */ |
| 736 | |