Index: head/sys/netinet/sctp.h =================================================================== --- head/sys/netinet/sctp.h (revision 181053) +++ head/sys/netinet/sctp.h (revision 181054) @@ -1,540 +1,539 @@ /*- * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp.h,v 1.18 2005/03/06 16:04:16 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #ifndef _NETINET_SCTP_H_ #define _NETINET_SCTP_H_ #include #define SCTP_PACKED __attribute__((packed)) /* * SCTP protocol - RFC2960. */ struct sctphdr { uint16_t src_port; /* source port */ uint16_t dest_port; /* destination port */ uint32_t v_tag; /* verification tag of packet */ uint32_t checksum; /* Adler32 C-Sum */ /* chunks follow... */ } SCTP_PACKED; /* * SCTP Chunks */ struct sctp_chunkhdr { uint8_t chunk_type; /* chunk type */ uint8_t chunk_flags; /* chunk flags */ uint16_t chunk_length; /* chunk length */ /* optional params follow */ } SCTP_PACKED; /* * SCTP chunk parameters */ struct sctp_paramhdr { uint16_t param_type; /* parameter type */ uint16_t param_length; /* parameter length */ } SCTP_PACKED; /* * user socket options: socket API defined */ /* * read-write options */ #define SCTP_RTOINFO 0x00000001 #define SCTP_ASSOCINFO 0x00000002 #define SCTP_INITMSG 0x00000003 #define SCTP_NODELAY 0x00000004 #define SCTP_AUTOCLOSE 0x00000005 #define SCTP_SET_PEER_PRIMARY_ADDR 0x00000006 #define SCTP_PRIMARY_ADDR 0x00000007 #define SCTP_ADAPTATION_LAYER 0x00000008 /* same as above */ #define SCTP_ADAPTION_LAYER 0x00000008 #define SCTP_DISABLE_FRAGMENTS 0x00000009 #define SCTP_PEER_ADDR_PARAMS 0x0000000a #define SCTP_DEFAULT_SEND_PARAM 0x0000000b /* ancillary data/notification interest options */ #define SCTP_EVENTS 0x0000000c /* Without this applied we will give V4 and V6 addresses on a V6 socket */ #define SCTP_I_WANT_MAPPED_V4_ADDR 0x0000000d #define SCTP_MAXSEG 0x0000000e #define SCTP_DELAYED_SACK 0x0000000f #define SCTP_FRAGMENT_INTERLEAVE 0x00000010 #define SCTP_PARTIAL_DELIVERY_POINT 0x00000011 /* authentication support */ #define SCTP_AUTH_CHUNK 0x00000012 #define SCTP_AUTH_KEY 0x00000013 #define SCTP_HMAC_IDENT 0x00000014 #define SCTP_AUTH_ACTIVE_KEY 0x00000015 #define SCTP_AUTH_DELETE_KEY 0x00000016 #define SCTP_USE_EXT_RCVINFO 0x00000017 #define SCTP_AUTO_ASCONF 0x00000018 /* rw */ #define SCTP_MAXBURST 0x00000019 /* rw */ #define SCTP_MAX_BURST 0x00000019 /* rw */ /* assoc level context */ #define SCTP_CONTEXT 0x0000001a /* rw */ /* explict EOR signalling */ #define SCTP_EXPLICIT_EOR 0x0000001b +#define SCTP_REUSE_PORT 0x0000001c /* rw */ /* * read-only options */ #define SCTP_STATUS 0x00000100 #define SCTP_GET_PEER_ADDR_INFO 0x00000101 /* authentication support */ #define SCTP_PEER_AUTH_CHUNKS 0x00000102 #define SCTP_LOCAL_AUTH_CHUNKS 0x00000103 #define SCTP_GET_ASSOC_NUMBER 0x00000104 /* ro */ #define SCTP_GET_ASSOC_ID_LIST 0x00000105 /* ro */ /* * user socket options: BSD implementation specific */ /* * Blocking I/O is enabled on any TCP type socket by default. For the UDP * model if this is turned on then the socket buffer is shared for send * resources amongst all associations. The default for the UDP model is that * is SS_NBIO is set. Which means all associations have a seperate send * limit BUT they will NOT ever BLOCK instead you will get an error back * EAGAIN if you try to send to much. If you want the blocking symantics you * set this option at the cost of sharing one socket send buffer size amongst * all associations. Peeled off sockets turn this option off and block. But * since both TCP and peeled off sockets have only one assoc per socket this * is fine. It probably does NOT make sense to set this on SS_NBIO on a TCP * model OR peeled off UDP model, but we do allow you to do so. You just use * the normal syscall to toggle SS_NBIO the way you want. * * Blocking I/O is controled by the SS_NBIO flag on the socket state so_state * field. */ /* these should probably go into sockets API */ #define SCTP_RESET_STREAMS 0x00001004 /* wo */ /* here on down are more implementation specific */ #define SCTP_SET_DEBUG_LEVEL 0x00001005 #define SCTP_CLR_STAT_LOG 0x00001007 /* CMT ON/OFF socket option */ #define SCTP_CMT_ON_OFF 0x00001200 #define SCTP_CMT_USE_DAC 0x00001201 /* JRS - Pluggable Congestion Control Socket option */ #define SCTP_PLUGGABLE_CC 0x00001202 /* read only */ #define SCTP_GET_SNDBUF_USE 0x00001101 #define SCTP_GET_STAT_LOG 0x00001103 #define SCTP_PCB_STATUS 0x00001104 #define SCTP_GET_NONCE_VALUES 0x00001105 /* Special hook for dynamically setting primary for all assoc's, * this is a write only option that requires root privledge. */ #define SCTP_SET_DYNAMIC_PRIMARY 0x00002001 /* VRF (virtual router feature) and multi-VRF support * options. VRF's provide splits within a router * that give the views of multiple routers. A * standard host, without VRF support, is just * a single VRF. If VRF's are supported then * the transport must be VRF aware. This means * that every socket call coming in must be directed * within the endpoint to one of the VRF's it belongs * to. The endpoint, before binding, may select * the "default" VRF it is in by using a set socket * option with SCTP_VRF_ID. This will also * get propegated to the default VRF. Once the * endpoint binds an address then it CANNOT add * additional VRF's to become a Multi-VRF endpoint. * * Before BINDING additional VRF's can be added with * the SCTP_ADD_VRF_ID call or deleted with * SCTP_DEL_VRF_ID. * * Associations are ALWAYS contained inside a single * VRF. They cannot reside in two (or more) VRF's. Incoming * packets, assuming the router is VRF aware, can always * tell us what VRF they arrived on. A host not supporting * any VRF's will find that the packets always arrived on the * single VRF that the host has. * */ #define SCTP_VRF_ID 0x00003001 #define SCTP_ADD_VRF_ID 0x00003002 #define SCTP_GET_VRF_IDS 0x00003003 #define SCTP_GET_ASOC_VRF 0x00003004 #define SCTP_DEL_VRF_ID 0x00003005 /* * If you enable packet logging you can get * a poor mans ethereal output in binary * form. Note this is a compile option to * the kernel, SCTP_PACKET_LOGGING, and * without it in your kernel you * will get a EOPNOTSUPP */ #define SCTP_GET_PACKET_LOG 0x00004001 /* * hidden implementation specific options these are NOT user visible (should * move out of sctp.h) */ /* sctp_bindx() flags as hidden socket options */ #define SCTP_BINDX_ADD_ADDR 0x00008001 #define SCTP_BINDX_REM_ADDR 0x00008002 /* Hidden socket option that gets the addresses */ #define SCTP_GET_PEER_ADDRESSES 0x00008003 #define SCTP_GET_LOCAL_ADDRESSES 0x00008004 /* return the total count in bytes needed to hold all local addresses bound */ #define SCTP_GET_LOCAL_ADDR_SIZE 0x00008005 /* Return the total count in bytes needed to hold the remote address */ #define SCTP_GET_REMOTE_ADDR_SIZE 0x00008006 /* hidden option for connectx */ #define SCTP_CONNECT_X 0x00008007 /* hidden option for connectx_delayed, part of sendx */ #define SCTP_CONNECT_X_DELAYED 0x00008008 #define SCTP_CONNECT_X_COMPLETE 0x00008009 /* hidden socket option based sctp_peeloff */ #define SCTP_PEELOFF 0x0000800a /* the real worker for sctp_getaddrlen() */ #define SCTP_GET_ADDR_LEN 0x0000800b /* temporary workaround for Apple listen() issue, no args used */ #define SCTP_LISTEN_FIX 0x0000800c /* Debug things that need to be purged */ #define SCTP_SET_INITIAL_DBG_SEQ 0x00009f00 /* JRS - Supported congestion control modules for pluggable * congestion control */ /* Standard TCP Congestion Control */ #define SCTP_CC_RFC2581 0x00000000 /* High Speed TCP Congestion Control (Floyd) */ #define SCTP_CC_HSTCP 0x00000001 /* HTCP Congestion Control */ #define SCTP_CC_HTCP 0x00000002 /* fragment interleave constants * setting must be one of these or * EINVAL returned. */ #define SCTP_FRAG_LEVEL_0 0x00000000 #define SCTP_FRAG_LEVEL_1 0x00000001 #define SCTP_FRAG_LEVEL_2 0x00000002 /* * user state values */ #define SCTP_CLOSED 0x0000 #define SCTP_BOUND 0x1000 #define SCTP_LISTEN 0x2000 #define SCTP_COOKIE_WAIT 0x0002 #define SCTP_COOKIE_ECHOED 0x0004 #define SCTP_ESTABLISHED 0x0008 #define SCTP_SHUTDOWN_SENT 0x0010 #define SCTP_SHUTDOWN_RECEIVED 0x0020 #define SCTP_SHUTDOWN_ACK_SENT 0x0040 #define SCTP_SHUTDOWN_PENDING 0x0080 /* * SCTP operational error codes (user visible) */ #define SCTP_CAUSE_NO_ERROR 0x0000 #define SCTP_CAUSE_INVALID_STREAM 0x0001 #define SCTP_CAUSE_MISSING_PARAM 0x0002 #define SCTP_CAUSE_STALE_COOKIE 0x0003 #define SCTP_CAUSE_OUT_OF_RESC 0x0004 #define SCTP_CAUSE_UNRESOLVABLE_ADDR 0x0005 #define SCTP_CAUSE_UNRECOG_CHUNK 0x0006 #define SCTP_CAUSE_INVALID_PARAM 0x0007 #define SCTP_CAUSE_UNRECOG_PARAM 0x0008 #define SCTP_CAUSE_NO_USER_DATA 0x0009 #define SCTP_CAUSE_COOKIE_IN_SHUTDOWN 0x000a #define SCTP_CAUSE_RESTART_W_NEWADDR 0x000b #define SCTP_CAUSE_USER_INITIATED_ABT 0x000c #define SCTP_CAUSE_PROTOCOL_VIOLATION 0x000d /* Error causes from RFC5061 */ #define SCTP_CAUSE_DELETING_LAST_ADDR 0xa0 #define SCTP_CAUSE_RESOURCE_SHORTAGE 0xa1 #define SCTP_CAUSE_DELETING_SRC_ADDR 0xa2 #define SCTP_CAUSE_ILLEGAL_ASCONF_ACK 0xa3 #define SCTP_CAUSE_REQUEST_REFUSED 0xa4 /* Error causes from RFC4895 */ #define SCTP_CAUSE_UNSUPPORTED_HMACID 0x0105 /* * error cause parameters (user visisble) */ struct sctp_error_cause { uint16_t code; uint16_t length; /* optional cause-specific info may follow */ } SCTP_PACKED; struct sctp_error_invalid_stream { struct sctp_error_cause cause; /* code=SCTP_ERROR_INVALID_STREAM */ uint16_t stream_id; /* stream id of the DATA in error */ uint16_t reserved; } SCTP_PACKED; struct sctp_error_missing_param { struct sctp_error_cause cause; /* code=SCTP_ERROR_MISSING_PARAM */ uint32_t num_missing_params; /* number of missing parameters */ /* uint16_t param_type's follow */ } SCTP_PACKED; struct sctp_error_stale_cookie { struct sctp_error_cause cause; /* code=SCTP_ERROR_STALE_COOKIE */ uint32_t stale_time; /* time in usec of staleness */ } SCTP_PACKED; struct sctp_error_out_of_resource { struct sctp_error_cause cause; /* code=SCTP_ERROR_OUT_OF_RESOURCES */ } SCTP_PACKED; struct sctp_error_unresolv_addr { struct sctp_error_cause cause; /* code=SCTP_ERROR_UNRESOLVABLE_ADDR */ } SCTP_PACKED; struct sctp_error_unrecognized_chunk { struct sctp_error_cause cause; /* code=SCTP_ERROR_UNRECOG_CHUNK */ struct sctp_chunkhdr ch;/* header from chunk in error */ } SCTP_PACKED; /* * Main SCTP chunk types we place these here so natd and f/w's in user land * can find them. */ /************0x00 series ***********/ #define SCTP_DATA 0x00 #define SCTP_INITIATION 0x01 #define SCTP_INITIATION_ACK 0x02 #define SCTP_SELECTIVE_ACK 0x03 #define SCTP_HEARTBEAT_REQUEST 0x04 #define SCTP_HEARTBEAT_ACK 0x05 #define SCTP_ABORT_ASSOCIATION 0x06 #define SCTP_SHUTDOWN 0x07 #define SCTP_SHUTDOWN_ACK 0x08 #define SCTP_OPERATION_ERROR 0x09 #define SCTP_COOKIE_ECHO 0x0a #define SCTP_COOKIE_ACK 0x0b #define SCTP_ECN_ECHO 0x0c #define SCTP_ECN_CWR 0x0d #define SCTP_SHUTDOWN_COMPLETE 0x0e /* RFC4895 */ #define SCTP_AUTHENTICATION 0x0f /************0x40 series ***********/ /************0x80 series ***********/ /* RFC5061 */ #define SCTP_ASCONF_ACK 0x80 /* draft-ietf-stewart-pktdrpsctp */ #define SCTP_PACKET_DROPPED 0x81 /* draft-ietf-stewart-strreset-xxx */ #define SCTP_STREAM_RESET 0x82 /* RFC4820 */ #define SCTP_PAD_CHUNK 0x84 /************0xc0 series ***********/ /* RFC3758 */ #define SCTP_FORWARD_CUM_TSN 0xc0 /* RFC5061 */ #define SCTP_ASCONF 0xc1 /* ABORT and SHUTDOWN COMPLETE FLAG */ #define SCTP_HAD_NO_TCB 0x01 /* Packet dropped flags */ #define SCTP_FROM_MIDDLE_BOX SCTP_HAD_NO_TCB #define SCTP_BADCRC 0x02 #define SCTP_PACKET_TRUNCATED 0x04 #define SCTP_SAT_NETWORK_MIN 400 /* min ms for RTT to set satellite * time */ #define SCTP_SAT_NETWORK_BURST_INCR 2 /* how many times to multiply maxburst * in sat */ /* Data Chuck Specific Flags */ #define SCTP_DATA_FRAG_MASK 0x03 #define SCTP_DATA_MIDDLE_FRAG 0x00 #define SCTP_DATA_LAST_FRAG 0x01 #define SCTP_DATA_FIRST_FRAG 0x02 #define SCTP_DATA_NOT_FRAG 0x03 #define SCTP_DATA_UNORDERED 0x04 #define SCTP_DATA_SACK_IMMEDIATELY 0x08 /* ECN Nonce: SACK Chunk Specific Flags */ #define SCTP_SACK_NONCE_SUM 0x01 /* CMT DAC algorithm SACK flag */ #define SCTP_SACK_CMT_DAC 0x80 /* * PCB flags (in sctp_flags bitmask). * Note the features and flags are meant * for use by netstat. */ #define SCTP_PCB_FLAGS_UDPTYPE 0x00000001 #define SCTP_PCB_FLAGS_TCPTYPE 0x00000002 #define SCTP_PCB_FLAGS_BOUNDALL 0x00000004 #define SCTP_PCB_FLAGS_ACCEPTING 0x00000008 #define SCTP_PCB_FLAGS_UNBOUND 0x00000010 -#define SCTP_PCB_FLAGS_LISTENING 0x00000020 #define SCTP_PCB_FLAGS_CLOSE_IP 0x00040000 #define SCTP_PCB_FLAGS_WAS_CONNECTED 0x00080000 #define SCTP_PCB_FLAGS_WAS_ABORTED 0x00100000 /* TCP model support */ #define SCTP_PCB_FLAGS_CONNECTED 0x00200000 #define SCTP_PCB_FLAGS_IN_TCPPOOL 0x00400000 #define SCTP_PCB_FLAGS_DONT_WAKE 0x00800000 #define SCTP_PCB_FLAGS_WAKEOUTPUT 0x01000000 #define SCTP_PCB_FLAGS_WAKEINPUT 0x02000000 #define SCTP_PCB_FLAGS_BOUND_V6 0x04000000 #define SCTP_PCB_FLAGS_BLOCKING_IO 0x08000000 #define SCTP_PCB_FLAGS_SOCKET_GONE 0x10000000 #define SCTP_PCB_FLAGS_SOCKET_ALLGONE 0x20000000 /* flags to copy to new PCB */ #define SCTP_PCB_COPY_FLAGS (SCTP_PCB_FLAGS_BOUNDALL|\ SCTP_PCB_FLAGS_WAKEINPUT|\ SCTP_PCB_FLAGS_BOUND_V6) /* * PCB Features (in sctp_features bitmask) */ #define SCTP_PCB_FLAGS_EXT_RCVINFO 0x00000002 #define SCTP_PCB_FLAGS_DONOT_HEARTBEAT 0x00000004 #define SCTP_PCB_FLAGS_FRAG_INTERLEAVE 0x00000008 #define SCTP_PCB_FLAGS_INTERLEAVE_STRMS 0x00000010 #define SCTP_PCB_FLAGS_DO_ASCONF 0x00000020 #define SCTP_PCB_FLAGS_AUTO_ASCONF 0x00000040 #define SCTP_PCB_FLAGS_ZERO_COPY_ACTIVE 0x00000080 - /* socket options */ #define SCTP_PCB_FLAGS_NODELAY 0x00000100 #define SCTP_PCB_FLAGS_AUTOCLOSE 0x00000200 #define SCTP_PCB_FLAGS_RECVDATAIOEVNT 0x00000400 #define SCTP_PCB_FLAGS_RECVASSOCEVNT 0x00000800 #define SCTP_PCB_FLAGS_RECVPADDREVNT 0x00001000 #define SCTP_PCB_FLAGS_RECVPEERERR 0x00002000 #define SCTP_PCB_FLAGS_RECVSENDFAILEVNT 0x00004000 #define SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT 0x00008000 #define SCTP_PCB_FLAGS_ADAPTATIONEVNT 0x00010000 #define SCTP_PCB_FLAGS_PDAPIEVNT 0x00020000 #define SCTP_PCB_FLAGS_AUTHEVNT 0x00040000 #define SCTP_PCB_FLAGS_STREAM_RESETEVNT 0x00080000 #define SCTP_PCB_FLAGS_NO_FRAGMENT 0x00100000 #define SCTP_PCB_FLAGS_EXPLICIT_EOR 0x00400000 #define SCTP_PCB_FLAGS_NEEDS_MAPPED_V4 0x00800000 #define SCTP_PCB_FLAGS_MULTIPLE_ASCONFS 0x01000000 - +#define SCTP_PCB_FLAGS_PORTREUSE 0x02000000 /*- * mobility_features parameters (by micchie).Note * these features are applied against the * sctp_mobility_features flags.. not the sctp_features * flags. */ #define SCTP_MOBILITY_BASE 0x00000001 #define SCTP_MOBILITY_FASTHANDOFF 0x00000002 #define SCTP_MOBILITY_PRIM_DELETED 0x00000004 #define SCTP_SMALLEST_PMTU 512 /* smallest pmtu allowed when disabling PMTU * discovery */ #include /* This dictates the size of the packet * collection buffer. This only applies * if SCTP_PACKET_LOGGING is enabled in * your config. */ #define SCTP_PACKET_LOG_SIZE 65536 /* Maximum delays and such a user can set for options that * take ms. */ #define SCTP_MAX_SACK_DELAY 500 /* per RFC4960 */ #define SCTP_MAX_HB_INTERVAL 14400000 /* 4 hours in ms */ #define SCTP_MAX_COOKIE_LIFE 3600000 /* 1 hour in ms */ /* Types of logging/KTR tracing that can be enabled via the * sysctl net.inet.sctp.sctp_logging. You must also enable * SUBSYS tracing. * Note that you must have the SCTP option in the kernel * to enable these as well. */ #define SCTP_BLK_LOGGING_ENABLE 0x00000001 #define SCTP_CWND_MONITOR_ENABLE 0x00000002 #define SCTP_CWND_LOGGING_ENABLE 0x00000004 #define SCTP_EARLYFR_LOGGING_ENABLE 0x00000010 #define SCTP_FLIGHT_LOGGING_ENABLE 0x00000020 #define SCTP_FR_LOGGING_ENABLE 0x00000040 #define SCTP_LOCK_LOGGING_ENABLE 0x00000080 #define SCTP_MAP_LOGGING_ENABLE 0x00000100 #define SCTP_MBCNT_LOGGING_ENABLE 0x00000200 #define SCTP_MBUF_LOGGING_ENABLE 0x00000400 #define SCTP_NAGLE_LOGGING_ENABLE 0x00000800 #define SCTP_RECV_RWND_LOGGING_ENABLE 0x00001000 #define SCTP_RTTVAR_LOGGING_ENABLE 0x00002000 #define SCTP_SACK_LOGGING_ENABLE 0x00004000 #define SCTP_SACK_RWND_LOGGING_ENABLE 0x00008000 #define SCTP_SB_LOGGING_ENABLE 0x00010000 #define SCTP_STR_LOGGING_ENABLE 0x00020000 #define SCTP_WAKE_LOGGING_ENABLE 0x00040000 #define SCTP_LOG_MAXBURST_ENABLE 0x00080000 #define SCTP_LOG_RWND_ENABLE 0x00100000 #define SCTP_LOG_SACK_ARRIVALS_ENABLE 0x00200000 #define SCTP_LTRACE_CHUNK_ENABLE 0x00400000 #define SCTP_LTRACE_ERROR_ENABLE 0x00800000 #define SCTP_LAST_PACKET_TRACING 0x01000000 #define SCTP_THRESHOLD_LOGGING 0x02000000 #define SCTP_LOG_AT_SEND_2_SCTP 0x04000000 #define SCTP_LOG_AT_SEND_2_OUTQ 0x08000000 #undef SCTP_PACKED #endif /* !_NETINET_SCTP_H_ */ Index: head/sys/netinet/sctp_output.c =================================================================== --- head/sys/netinet/sctp_output.c (revision 181053) +++ head/sys/netinet/sctp_output.c (revision 181054) @@ -1,13237 +1,13255 @@ /*- * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_output.c,v 1.46 2005/03/06 16:04:17 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define SCTP_MAX_GAPS_INARRAY 4 struct sack_track { uint8_t right_edge; /* mergable on the right edge */ uint8_t left_edge; /* mergable on the left edge */ uint8_t num_entries; uint8_t spare; struct sctp_gap_ack_block gaps[SCTP_MAX_GAPS_INARRAY]; }; struct sack_track sack_array[256] = { {0, 0, 0, 0, /* 0x00 */ {{0, 0}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x01 */ {{0, 0}, {0, 0}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x02 */ {{1, 1}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x03 */ {{0, 1}, {0, 0}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x04 */ {{2, 2}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x05 */ {{0, 0}, {2, 2}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x06 */ {{1, 2}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x07 */ {{0, 2}, {0, 0}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x08 */ {{3, 3}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x09 */ {{0, 0}, {3, 3}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x0a */ {{1, 1}, {3, 3}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x0b */ {{0, 1}, {3, 3}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x0c */ {{2, 3}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x0d */ {{0, 0}, {2, 3}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x0e */ {{1, 3}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x0f */ {{0, 3}, {0, 0}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x10 */ {{4, 4}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x11 */ {{0, 0}, {4, 4}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x12 */ {{1, 1}, {4, 4}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x13 */ {{0, 1}, {4, 4}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x14 */ {{2, 2}, {4, 4}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x15 */ {{0, 0}, {2, 2}, {4, 4}, {0, 0} } }, {0, 0, 2, 0, /* 0x16 */ {{1, 2}, {4, 4}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x17 */ {{0, 2}, {4, 4}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x18 */ {{3, 4}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x19 */ {{0, 0}, {3, 4}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x1a */ {{1, 1}, {3, 4}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x1b */ {{0, 1}, {3, 4}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x1c */ {{2, 4}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x1d */ {{0, 0}, {2, 4}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x1e */ {{1, 4}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x1f */ {{0, 4}, {0, 0}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x20 */ {{5, 5}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x21 */ {{0, 0}, {5, 5}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x22 */ {{1, 1}, {5, 5}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x23 */ {{0, 1}, {5, 5}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x24 */ {{2, 2}, {5, 5}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x25 */ {{0, 0}, {2, 2}, {5, 5}, {0, 0} } }, {0, 0, 2, 0, /* 0x26 */ {{1, 2}, {5, 5}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x27 */ {{0, 2}, {5, 5}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x28 */ {{3, 3}, {5, 5}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x29 */ {{0, 0}, {3, 3}, {5, 5}, {0, 0} } }, {0, 0, 3, 0, /* 0x2a */ {{1, 1}, {3, 3}, {5, 5}, {0, 0} } }, {1, 0, 3, 0, /* 0x2b */ {{0, 1}, {3, 3}, {5, 5}, {0, 0} } }, {0, 0, 2, 0, /* 0x2c */ {{2, 3}, {5, 5}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x2d */ {{0, 0}, {2, 3}, {5, 5}, {0, 0} } }, {0, 0, 2, 0, /* 0x2e */ {{1, 3}, {5, 5}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x2f */ {{0, 3}, {5, 5}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x30 */ {{4, 5}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x31 */ {{0, 0}, {4, 5}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x32 */ {{1, 1}, {4, 5}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x33 */ {{0, 1}, {4, 5}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x34 */ {{2, 2}, {4, 5}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x35 */ {{0, 0}, {2, 2}, {4, 5}, {0, 0} } }, {0, 0, 2, 0, /* 0x36 */ {{1, 2}, {4, 5}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x37 */ {{0, 2}, {4, 5}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x38 */ {{3, 5}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x39 */ {{0, 0}, {3, 5}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x3a */ {{1, 1}, {3, 5}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x3b */ {{0, 1}, {3, 5}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x3c */ {{2, 5}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x3d */ {{0, 0}, {2, 5}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x3e */ {{1, 5}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x3f */ {{0, 5}, {0, 0}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x40 */ {{6, 6}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x41 */ {{0, 0}, {6, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x42 */ {{1, 1}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x43 */ {{0, 1}, {6, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x44 */ {{2, 2}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x45 */ {{0, 0}, {2, 2}, {6, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x46 */ {{1, 2}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x47 */ {{0, 2}, {6, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x48 */ {{3, 3}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x49 */ {{0, 0}, {3, 3}, {6, 6}, {0, 0} } }, {0, 0, 3, 0, /* 0x4a */ {{1, 1}, {3, 3}, {6, 6}, {0, 0} } }, {1, 0, 3, 0, /* 0x4b */ {{0, 1}, {3, 3}, {6, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x4c */ {{2, 3}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x4d */ {{0, 0}, {2, 3}, {6, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x4e */ {{1, 3}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x4f */ {{0, 3}, {6, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x50 */ {{4, 4}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x51 */ {{0, 0}, {4, 4}, {6, 6}, {0, 0} } }, {0, 0, 3, 0, /* 0x52 */ {{1, 1}, {4, 4}, {6, 6}, {0, 0} } }, {1, 0, 3, 0, /* 0x53 */ {{0, 1}, {4, 4}, {6, 6}, {0, 0} } }, {0, 0, 3, 0, /* 0x54 */ {{2, 2}, {4, 4}, {6, 6}, {0, 0} } }, {1, 0, 4, 0, /* 0x55 */ {{0, 0}, {2, 2}, {4, 4}, {6, 6} } }, {0, 0, 3, 0, /* 0x56 */ {{1, 2}, {4, 4}, {6, 6}, {0, 0} } }, {1, 0, 3, 0, /* 0x57 */ {{0, 2}, {4, 4}, {6, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x58 */ {{3, 4}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x59 */ {{0, 0}, {3, 4}, {6, 6}, {0, 0} } }, {0, 0, 3, 0, /* 0x5a */ {{1, 1}, {3, 4}, {6, 6}, {0, 0} } }, {1, 0, 3, 0, /* 0x5b */ {{0, 1}, {3, 4}, {6, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x5c */ {{2, 4}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x5d */ {{0, 0}, {2, 4}, {6, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x5e */ {{1, 4}, {6, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x5f */ {{0, 4}, {6, 6}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x60 */ {{5, 6}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x61 */ {{0, 0}, {5, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x62 */ {{1, 1}, {5, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x63 */ {{0, 1}, {5, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x64 */ {{2, 2}, {5, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x65 */ {{0, 0}, {2, 2}, {5, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x66 */ {{1, 2}, {5, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x67 */ {{0, 2}, {5, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x68 */ {{3, 3}, {5, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x69 */ {{0, 0}, {3, 3}, {5, 6}, {0, 0} } }, {0, 0, 3, 0, /* 0x6a */ {{1, 1}, {3, 3}, {5, 6}, {0, 0} } }, {1, 0, 3, 0, /* 0x6b */ {{0, 1}, {3, 3}, {5, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x6c */ {{2, 3}, {5, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x6d */ {{0, 0}, {2, 3}, {5, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x6e */ {{1, 3}, {5, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x6f */ {{0, 3}, {5, 6}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x70 */ {{4, 6}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x71 */ {{0, 0}, {4, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x72 */ {{1, 1}, {4, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x73 */ {{0, 1}, {4, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x74 */ {{2, 2}, {4, 6}, {0, 0}, {0, 0} } }, {1, 0, 3, 0, /* 0x75 */ {{0, 0}, {2, 2}, {4, 6}, {0, 0} } }, {0, 0, 2, 0, /* 0x76 */ {{1, 2}, {4, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x77 */ {{0, 2}, {4, 6}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x78 */ {{3, 6}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x79 */ {{0, 0}, {3, 6}, {0, 0}, {0, 0} } }, {0, 0, 2, 0, /* 0x7a */ {{1, 1}, {3, 6}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x7b */ {{0, 1}, {3, 6}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x7c */ {{2, 6}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 2, 0, /* 0x7d */ {{0, 0}, {2, 6}, {0, 0}, {0, 0} } }, {0, 0, 1, 0, /* 0x7e */ {{1, 6}, {0, 0}, {0, 0}, {0, 0} } }, {1, 0, 1, 0, /* 0x7f */ {{0, 6}, {0, 0}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0x80 */ {{7, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0x81 */ {{0, 0}, {7, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0x82 */ {{1, 1}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0x83 */ {{0, 1}, {7, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0x84 */ {{2, 2}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0x85 */ {{0, 0}, {2, 2}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0x86 */ {{1, 2}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0x87 */ {{0, 2}, {7, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0x88 */ {{3, 3}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0x89 */ {{0, 0}, {3, 3}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0x8a */ {{1, 1}, {3, 3}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0x8b */ {{0, 1}, {3, 3}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0x8c */ {{2, 3}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0x8d */ {{0, 0}, {2, 3}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0x8e */ {{1, 3}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0x8f */ {{0, 3}, {7, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0x90 */ {{4, 4}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0x91 */ {{0, 0}, {4, 4}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0x92 */ {{1, 1}, {4, 4}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0x93 */ {{0, 1}, {4, 4}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0x94 */ {{2, 2}, {4, 4}, {7, 7}, {0, 0} } }, {1, 1, 4, 0, /* 0x95 */ {{0, 0}, {2, 2}, {4, 4}, {7, 7} } }, {0, 1, 3, 0, /* 0x96 */ {{1, 2}, {4, 4}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0x97 */ {{0, 2}, {4, 4}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0x98 */ {{3, 4}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0x99 */ {{0, 0}, {3, 4}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0x9a */ {{1, 1}, {3, 4}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0x9b */ {{0, 1}, {3, 4}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0x9c */ {{2, 4}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0x9d */ {{0, 0}, {2, 4}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0x9e */ {{1, 4}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0x9f */ {{0, 4}, {7, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xa0 */ {{5, 5}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xa1 */ {{0, 0}, {5, 5}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xa2 */ {{1, 1}, {5, 5}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xa3 */ {{0, 1}, {5, 5}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xa4 */ {{2, 2}, {5, 5}, {7, 7}, {0, 0} } }, {1, 1, 4, 0, /* 0xa5 */ {{0, 0}, {2, 2}, {5, 5}, {7, 7} } }, {0, 1, 3, 0, /* 0xa6 */ {{1, 2}, {5, 5}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xa7 */ {{0, 2}, {5, 5}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xa8 */ {{3, 3}, {5, 5}, {7, 7}, {0, 0} } }, {1, 1, 4, 0, /* 0xa9 */ {{0, 0}, {3, 3}, {5, 5}, {7, 7} } }, {0, 1, 4, 0, /* 0xaa */ {{1, 1}, {3, 3}, {5, 5}, {7, 7} } }, {1, 1, 4, 0, /* 0xab */ {{0, 1}, {3, 3}, {5, 5}, {7, 7} } }, {0, 1, 3, 0, /* 0xac */ {{2, 3}, {5, 5}, {7, 7}, {0, 0} } }, {1, 1, 4, 0, /* 0xad */ {{0, 0}, {2, 3}, {5, 5}, {7, 7} } }, {0, 1, 3, 0, /* 0xae */ {{1, 3}, {5, 5}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xaf */ {{0, 3}, {5, 5}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xb0 */ {{4, 5}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xb1 */ {{0, 0}, {4, 5}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xb2 */ {{1, 1}, {4, 5}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xb3 */ {{0, 1}, {4, 5}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xb4 */ {{2, 2}, {4, 5}, {7, 7}, {0, 0} } }, {1, 1, 4, 0, /* 0xb5 */ {{0, 0}, {2, 2}, {4, 5}, {7, 7} } }, {0, 1, 3, 0, /* 0xb6 */ {{1, 2}, {4, 5}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xb7 */ {{0, 2}, {4, 5}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xb8 */ {{3, 5}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xb9 */ {{0, 0}, {3, 5}, {7, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xba */ {{1, 1}, {3, 5}, {7, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xbb */ {{0, 1}, {3, 5}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xbc */ {{2, 5}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xbd */ {{0, 0}, {2, 5}, {7, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xbe */ {{1, 5}, {7, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xbf */ {{0, 5}, {7, 7}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0xc0 */ {{6, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xc1 */ {{0, 0}, {6, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xc2 */ {{1, 1}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xc3 */ {{0, 1}, {6, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xc4 */ {{2, 2}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xc5 */ {{0, 0}, {2, 2}, {6, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xc6 */ {{1, 2}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xc7 */ {{0, 2}, {6, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xc8 */ {{3, 3}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xc9 */ {{0, 0}, {3, 3}, {6, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xca */ {{1, 1}, {3, 3}, {6, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xcb */ {{0, 1}, {3, 3}, {6, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xcc */ {{2, 3}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xcd */ {{0, 0}, {2, 3}, {6, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xce */ {{1, 3}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xcf */ {{0, 3}, {6, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xd0 */ {{4, 4}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xd1 */ {{0, 0}, {4, 4}, {6, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xd2 */ {{1, 1}, {4, 4}, {6, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xd3 */ {{0, 1}, {4, 4}, {6, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xd4 */ {{2, 2}, {4, 4}, {6, 7}, {0, 0} } }, {1, 1, 4, 0, /* 0xd5 */ {{0, 0}, {2, 2}, {4, 4}, {6, 7} } }, {0, 1, 3, 0, /* 0xd6 */ {{1, 2}, {4, 4}, {6, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xd7 */ {{0, 2}, {4, 4}, {6, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xd8 */ {{3, 4}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xd9 */ {{0, 0}, {3, 4}, {6, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xda */ {{1, 1}, {3, 4}, {6, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xdb */ {{0, 1}, {3, 4}, {6, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xdc */ {{2, 4}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xdd */ {{0, 0}, {2, 4}, {6, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xde */ {{1, 4}, {6, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xdf */ {{0, 4}, {6, 7}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0xe0 */ {{5, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xe1 */ {{0, 0}, {5, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xe2 */ {{1, 1}, {5, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xe3 */ {{0, 1}, {5, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xe4 */ {{2, 2}, {5, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xe5 */ {{0, 0}, {2, 2}, {5, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xe6 */ {{1, 2}, {5, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xe7 */ {{0, 2}, {5, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xe8 */ {{3, 3}, {5, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xe9 */ {{0, 0}, {3, 3}, {5, 7}, {0, 0} } }, {0, 1, 3, 0, /* 0xea */ {{1, 1}, {3, 3}, {5, 7}, {0, 0} } }, {1, 1, 3, 0, /* 0xeb */ {{0, 1}, {3, 3}, {5, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xec */ {{2, 3}, {5, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xed */ {{0, 0}, {2, 3}, {5, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xee */ {{1, 3}, {5, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xef */ {{0, 3}, {5, 7}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0xf0 */ {{4, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xf1 */ {{0, 0}, {4, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xf2 */ {{1, 1}, {4, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xf3 */ {{0, 1}, {4, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xf4 */ {{2, 2}, {4, 7}, {0, 0}, {0, 0} } }, {1, 1, 3, 0, /* 0xf5 */ {{0, 0}, {2, 2}, {4, 7}, {0, 0} } }, {0, 1, 2, 0, /* 0xf6 */ {{1, 2}, {4, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xf7 */ {{0, 2}, {4, 7}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0xf8 */ {{3, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xf9 */ {{0, 0}, {3, 7}, {0, 0}, {0, 0} } }, {0, 1, 2, 0, /* 0xfa */ {{1, 1}, {3, 7}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xfb */ {{0, 1}, {3, 7}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0xfc */ {{2, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 2, 0, /* 0xfd */ {{0, 0}, {2, 7}, {0, 0}, {0, 0} } }, {0, 1, 1, 0, /* 0xfe */ {{1, 7}, {0, 0}, {0, 0}, {0, 0} } }, {1, 1, 1, 0, /* 0xff */ {{0, 7}, {0, 0}, {0, 0}, {0, 0} } } }; int sctp_is_address_in_scope(struct sctp_ifa *ifa, int ipv4_addr_legal, int ipv6_addr_legal, int loopback_scope, int ipv4_local_scope, int local_scope, int site_scope, int do_update) { if ((loopback_scope == 0) && (ifa->ifn_p) && SCTP_IFN_IS_IFT_LOOP(ifa->ifn_p)) { /* * skip loopback if not in scope * */ return (0); } switch (ifa->address.sa.sa_family) { case AF_INET: if (ipv4_addr_legal) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)&ifa->address.sin; if (sin->sin_addr.s_addr == 0) { /* not in scope , unspecified */ return (0); } if ((ipv4_local_scope == 0) && (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { /* private address not in scope */ return (0); } } else { return (0); } break; #ifdef INET6 case AF_INET6: if (ipv6_addr_legal) { struct sockaddr_in6 *sin6; /* * Must update the flags, bummer, which means any * IFA locks must now be applied HERE <-> */ if (do_update) { sctp_gather_internal_ifa_flags(ifa); } if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { return (0); } /* ok to use deprecated addresses? */ sin6 = (struct sockaddr_in6 *)&ifa->address.sin6; if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { /* skip unspecifed addresses */ return (0); } if ( /* (local_scope == 0) && */ (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr))) { return (0); } if ((site_scope == 0) && (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { return (0); } } else { return (0); } break; #endif default: return (0); } return (1); } static struct mbuf * sctp_add_addr_to_mbuf(struct mbuf *m, struct sctp_ifa *ifa) { struct sctp_paramhdr *parmh; struct mbuf *mret; int len; if (ifa->address.sa.sa_family == AF_INET) { len = sizeof(struct sctp_ipv4addr_param); } else if (ifa->address.sa.sa_family == AF_INET6) { len = sizeof(struct sctp_ipv6addr_param); } else { /* unknown type */ return (m); } if (M_TRAILINGSPACE(m) >= len) { /* easy side we just drop it on the end */ parmh = (struct sctp_paramhdr *)(SCTP_BUF_AT(m, SCTP_BUF_LEN(m))); mret = m; } else { /* Need more space */ mret = m; while (SCTP_BUF_NEXT(mret) != NULL) { mret = SCTP_BUF_NEXT(mret); } SCTP_BUF_NEXT(mret) = sctp_get_mbuf_for_msg(len, 0, M_DONTWAIT, 1, MT_DATA); if (SCTP_BUF_NEXT(mret) == NULL) { /* We are hosed, can't add more addresses */ return (m); } mret = SCTP_BUF_NEXT(mret); parmh = mtod(mret, struct sctp_paramhdr *); } /* now add the parameter */ switch (ifa->address.sa.sa_family) { case AF_INET: { struct sctp_ipv4addr_param *ipv4p; struct sockaddr_in *sin; sin = (struct sockaddr_in *)&ifa->address.sin; ipv4p = (struct sctp_ipv4addr_param *)parmh; parmh->param_type = htons(SCTP_IPV4_ADDRESS); parmh->param_length = htons(len); ipv4p->addr = sin->sin_addr.s_addr; SCTP_BUF_LEN(mret) += len; break; } #ifdef INET6 case AF_INET6: { struct sctp_ipv6addr_param *ipv6p; struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&ifa->address.sin6; ipv6p = (struct sctp_ipv6addr_param *)parmh; parmh->param_type = htons(SCTP_IPV6_ADDRESS); parmh->param_length = htons(len); memcpy(ipv6p->addr, &sin6->sin6_addr, sizeof(ipv6p->addr)); /* clear embedded scope in the address */ in6_clearscope((struct in6_addr *)ipv6p->addr); SCTP_BUF_LEN(mret) += len; break; } #endif default: return (m); } return (mret); } struct mbuf * sctp_add_addresses_to_i_ia(struct sctp_inpcb *inp, struct sctp_scoping *scope, struct mbuf *m_at, int cnt_inits_to) { struct sctp_vrf *vrf = NULL; int cnt, limit_out = 0, total_count; uint32_t vrf_id; vrf_id = inp->def_vrf_id; SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { SCTP_IPI_ADDR_RUNLOCK(); return (m_at); } if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { struct sctp_ifa *sctp_ifap; struct sctp_ifn *sctp_ifnp; cnt = cnt_inits_to; if (vrf->total_ifa_count > SCTP_COUNT_LIMIT) { limit_out = 1; cnt = SCTP_ADDRESS_LIMIT; goto skip_count; } LIST_FOREACH(sctp_ifnp, &vrf->ifnlist, next_ifn) { if ((scope->loopback_scope == 0) && SCTP_IFN_IS_IFT_LOOP(sctp_ifnp)) { /* * Skip loopback devices if loopback_scope * not set */ continue; } LIST_FOREACH(sctp_ifap, &sctp_ifnp->ifalist, next_ifa) { if (sctp_is_address_in_scope(sctp_ifap, scope->ipv4_addr_legal, scope->ipv6_addr_legal, scope->loopback_scope, scope->ipv4_local_scope, scope->local_scope, scope->site_scope, 1) == 0) { continue; } cnt++; if (cnt > SCTP_ADDRESS_LIMIT) { break; } } if (cnt > SCTP_ADDRESS_LIMIT) { break; } } skip_count: if (cnt > 1) { total_count = 0; LIST_FOREACH(sctp_ifnp, &vrf->ifnlist, next_ifn) { cnt = 0; if ((scope->loopback_scope == 0) && SCTP_IFN_IS_IFT_LOOP(sctp_ifnp)) { /* * Skip loopback devices if * loopback_scope not set */ continue; } LIST_FOREACH(sctp_ifap, &sctp_ifnp->ifalist, next_ifa) { if (sctp_is_address_in_scope(sctp_ifap, scope->ipv4_addr_legal, scope->ipv6_addr_legal, scope->loopback_scope, scope->ipv4_local_scope, scope->local_scope, scope->site_scope, 0) == 0) { continue; } m_at = sctp_add_addr_to_mbuf(m_at, sctp_ifap); if (limit_out) { cnt++; total_count++; if (cnt >= 2) { /* * two from each * address */ break; } if (total_count > SCTP_ADDRESS_LIMIT) { /* No more addresses */ break; } } } } } } else { struct sctp_laddr *laddr; cnt = cnt_inits_to; /* First, how many ? */ LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { continue; } if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) /* * Address being deleted by the system, dont * list. */ continue; if (laddr->action == SCTP_DEL_IP_ADDRESS) { /* * Address being deleted on this ep don't * list. */ continue; } if (sctp_is_address_in_scope(laddr->ifa, scope->ipv4_addr_legal, scope->ipv6_addr_legal, scope->loopback_scope, scope->ipv4_local_scope, scope->local_scope, scope->site_scope, 1) == 0) { continue; } cnt++; } if (cnt > SCTP_ADDRESS_LIMIT) { limit_out = 1; } /* * To get through a NAT we only list addresses if we have * more than one. That way if you just bind a single address * we let the source of the init dictate our address. */ if (cnt > 1) { LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { cnt = 0; if (laddr->ifa == NULL) { continue; } if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) continue; if (sctp_is_address_in_scope(laddr->ifa, scope->ipv4_addr_legal, scope->ipv6_addr_legal, scope->loopback_scope, scope->ipv4_local_scope, scope->local_scope, scope->site_scope, 0) == 0) { continue; } m_at = sctp_add_addr_to_mbuf(m_at, laddr->ifa); cnt++; if (cnt >= SCTP_ADDRESS_LIMIT) { break; } } } } SCTP_IPI_ADDR_RUNLOCK(); return (m_at); } static struct sctp_ifa * sctp_is_ifa_addr_preferred(struct sctp_ifa *ifa, uint8_t dest_is_loop, uint8_t dest_is_priv, sa_family_t fam) { uint8_t dest_is_global = 0; /* dest_is_priv is true if destination is a private address */ /* dest_is_loop is true if destination is a loopback addresses */ /* * Here we determine if its a preferred address. A preferred address * means it is the same scope or higher scope then the destination. * L = loopback, P = private, G = global * ----------------------------------------- src | dest | result * ---------------------------------------- L | L | yes * ----------------------------------------- P | L | * yes-v4 no-v6 ----------------------------------------- G | * L | yes-v4 no-v6 ----------------------------------------- L * | P | no ----------------------------------------- P | * P | yes ----------------------------------------- G | * P | no ----------------------------------------- L | G * | no ----------------------------------------- P | G | * no ----------------------------------------- G | G | * yes ----------------------------------------- */ if (ifa->address.sa.sa_family != fam) { /* forget mis-matched family */ return (NULL); } if ((dest_is_priv == 0) && (dest_is_loop == 0)) { dest_is_global = 1; } SCTPDBG(SCTP_DEBUG_OUTPUT2, "Is destination preferred:"); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &ifa->address.sa); /* Ok the address may be ok */ if (fam == AF_INET6) { /* ok to use deprecated addresses? no lets not! */ if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:1\n"); return (NULL); } if (ifa->src_is_priv && !ifa->src_is_loop) { if (dest_is_loop) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:2\n"); return (NULL); } } if (ifa->src_is_glob) { if (dest_is_loop) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:3\n"); return (NULL); } } } /* * Now that we know what is what, implement or table this could in * theory be done slicker (it used to be), but this is * straightforward and easier to validate :-) */ SCTPDBG(SCTP_DEBUG_OUTPUT3, "src_loop:%d src_priv:%d src_glob:%d\n", ifa->src_is_loop, ifa->src_is_priv, ifa->src_is_glob); SCTPDBG(SCTP_DEBUG_OUTPUT3, "dest_loop:%d dest_priv:%d dest_glob:%d\n", dest_is_loop, dest_is_priv, dest_is_global); if ((ifa->src_is_loop) && (dest_is_priv)) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:4\n"); return (NULL); } if ((ifa->src_is_glob) && (dest_is_priv)) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:5\n"); return (NULL); } if ((ifa->src_is_loop) && (dest_is_global)) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:6\n"); return (NULL); } if ((ifa->src_is_priv) && (dest_is_global)) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "NO:7\n"); return (NULL); } SCTPDBG(SCTP_DEBUG_OUTPUT3, "YES\n"); /* its a preferred address */ return (ifa); } static struct sctp_ifa * sctp_is_ifa_addr_acceptable(struct sctp_ifa *ifa, uint8_t dest_is_loop, uint8_t dest_is_priv, sa_family_t fam) { uint8_t dest_is_global = 0; /* * Here we determine if its a acceptable address. A acceptable * address means it is the same scope or higher scope but we can * allow for NAT which means its ok to have a global dest and a * private src. * * L = loopback, P = private, G = global * ----------------------------------------- src | dest | result * ----------------------------------------- L | L | yes * ----------------------------------------- P | L | * yes-v4 no-v6 ----------------------------------------- G | * L | yes ----------------------------------------- L | * P | no ----------------------------------------- P | P * | yes ----------------------------------------- G | P * | yes - May not work ----------------------------------------- * L | G | no ----------------------------------------- P * | G | yes - May not work * ----------------------------------------- G | G | yes * ----------------------------------------- */ if (ifa->address.sa.sa_family != fam) { /* forget non matching family */ return (NULL); } /* Ok the address may be ok */ if ((dest_is_loop == 0) && (dest_is_priv == 0)) { dest_is_global = 1; } if (fam == AF_INET6) { /* ok to use deprecated addresses? */ if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { return (NULL); } if (ifa->src_is_priv) { /* Special case, linklocal to loop */ if (dest_is_loop) return (NULL); } } /* * Now that we know what is what, implement our table. This could in * theory be done slicker (it used to be), but this is * straightforward and easier to validate :-) */ if ((ifa->src_is_loop == 1) && (dest_is_priv)) { return (NULL); } if ((ifa->src_is_loop == 1) && (dest_is_global)) { return (NULL); } /* its an acceptable address */ return (ifa); } int sctp_is_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa) { struct sctp_laddr *laddr; if (stcb == NULL) { /* There are no restrictions, no TCB :-) */ return (0); } LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "%s: NULL ifa\n", __FUNCTION__); continue; } if (laddr->ifa == ifa) { /* Yes it is on the list */ return (1); } } return (0); } int sctp_is_addr_in_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa) { struct sctp_laddr *laddr; if (ifa == NULL) return (0); LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "%s: NULL ifa\n", __FUNCTION__); continue; } if ((laddr->ifa == ifa) && laddr->action == 0) /* same pointer */ return (1); } return (0); } static struct sctp_ifa * sctp_choose_boundspecific_inp(struct sctp_inpcb *inp, sctp_route_t * ro, uint32_t vrf_id, int non_asoc_addr_ok, uint8_t dest_is_priv, uint8_t dest_is_loop, sa_family_t fam) { struct sctp_laddr *laddr, *starting_point; void *ifn; int resettotop = 0; struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa, *sifa; struct sctp_vrf *vrf; uint32_t ifn_index; vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) return (NULL); ifn = SCTP_GET_IFN_VOID_FROM_ROUTE(ro); ifn_index = SCTP_GET_IF_INDEX_FROM_ROUTE(ro); sctp_ifn = sctp_find_ifn(ifn, ifn_index); /* * first question, is the ifn we will emit on in our list, if so, we * want such an address. Note that we first looked for a preferred * address. */ if (sctp_ifn) { /* is a preferred one on the interface we route out? */ LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) continue; sifa = sctp_is_ifa_addr_preferred(sctp_ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (sctp_is_addr_in_ep(inp, sifa)) { atomic_add_int(&sifa->refcount, 1); return (sifa); } } } /* * ok, now we now need to find one on the list of the addresses. We * can't get one on the emitting interface so let's find first a * preferred one. If not that an acceptable one otherwise... we * return NULL. */ starting_point = inp->next_addr_touse; once_again: if (inp->next_addr_touse == NULL) { inp->next_addr_touse = LIST_FIRST(&inp->sctp_addr_list); resettotop = 1; } for (laddr = inp->next_addr_touse; laddr; laddr = LIST_NEXT(laddr, sctp_nxt_addr)) { if (laddr->ifa == NULL) { /* address has been removed */ continue; } if (laddr->action == SCTP_DEL_IP_ADDRESS) { /* address is being deleted */ continue; } sifa = sctp_is_ifa_addr_preferred(laddr->ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; atomic_add_int(&sifa->refcount, 1); return (sifa); } if (resettotop == 0) { inp->next_addr_touse = NULL; goto once_again; } inp->next_addr_touse = starting_point; resettotop = 0; once_again_too: if (inp->next_addr_touse == NULL) { inp->next_addr_touse = LIST_FIRST(&inp->sctp_addr_list); resettotop = 1; } /* ok, what about an acceptable address in the inp */ for (laddr = inp->next_addr_touse; laddr; laddr = LIST_NEXT(laddr, sctp_nxt_addr)) { if (laddr->ifa == NULL) { /* address has been removed */ continue; } if (laddr->action == SCTP_DEL_IP_ADDRESS) { /* address is being deleted */ continue; } sifa = sctp_is_ifa_addr_acceptable(laddr->ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; atomic_add_int(&sifa->refcount, 1); return (sifa); } if (resettotop == 0) { inp->next_addr_touse = NULL; goto once_again_too; } /* * no address bound can be a source for the destination we are in * trouble */ return (NULL); } static struct sctp_ifa * sctp_choose_boundspecific_stcb(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, sctp_route_t * ro, uint32_t vrf_id, uint8_t dest_is_priv, uint8_t dest_is_loop, int non_asoc_addr_ok, sa_family_t fam) { struct sctp_laddr *laddr, *starting_point; void *ifn; struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa, *sifa; uint8_t start_at_beginning = 0; struct sctp_vrf *vrf; uint32_t ifn_index; /* * first question, is the ifn we will emit on in our list, if so, we * want that one. */ vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) return (NULL); ifn = SCTP_GET_IFN_VOID_FROM_ROUTE(ro); ifn_index = SCTP_GET_IF_INDEX_FROM_ROUTE(ro); sctp_ifn = sctp_find_ifn(ifn, ifn_index); /* * first question, is the ifn we will emit on in our list? If so, * we want that one. First we look for a preferred. Second, we go * for an acceptable. */ if (sctp_ifn) { /* first try for a preferred address on the ep */ LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) continue; if (sctp_is_addr_in_ep(inp, sctp_ifa)) { sifa = sctp_is_ifa_addr_preferred(sctp_ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* on the no-no list */ continue; } atomic_add_int(&sifa->refcount, 1); return (sifa); } } /* next try for an acceptable address on the ep */ LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) continue; if (sctp_is_addr_in_ep(inp, sctp_ifa)) { sifa = sctp_is_ifa_addr_acceptable(sctp_ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* on the no-no list */ continue; } atomic_add_int(&sifa->refcount, 1); return (sifa); } } } /* * if we can't find one like that then we must look at all addresses * bound to pick one at first preferable then secondly acceptable. */ starting_point = stcb->asoc.last_used_address; sctp_from_the_top: if (stcb->asoc.last_used_address == NULL) { start_at_beginning = 1; stcb->asoc.last_used_address = LIST_FIRST(&inp->sctp_addr_list); } /* search beginning with the last used address */ for (laddr = stcb->asoc.last_used_address; laddr; laddr = LIST_NEXT(laddr, sctp_nxt_addr)) { if (laddr->ifa == NULL) { /* address has been removed */ continue; } if (laddr->action == SCTP_DEL_IP_ADDRESS) { /* address is being deleted */ continue; } sifa = sctp_is_ifa_addr_preferred(laddr->ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* on the no-no list */ continue; } stcb->asoc.last_used_address = laddr; atomic_add_int(&sifa->refcount, 1); return (sifa); } if (start_at_beginning == 0) { stcb->asoc.last_used_address = NULL; goto sctp_from_the_top; } /* now try for any higher scope than the destination */ stcb->asoc.last_used_address = starting_point; start_at_beginning = 0; sctp_from_the_top2: if (stcb->asoc.last_used_address == NULL) { start_at_beginning = 1; stcb->asoc.last_used_address = LIST_FIRST(&inp->sctp_addr_list); } /* search beginning with the last used address */ for (laddr = stcb->asoc.last_used_address; laddr; laddr = LIST_NEXT(laddr, sctp_nxt_addr)) { if (laddr->ifa == NULL) { /* address has been removed */ continue; } if (laddr->action == SCTP_DEL_IP_ADDRESS) { /* address is being deleted */ continue; } sifa = sctp_is_ifa_addr_acceptable(laddr->ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* on the no-no list */ continue; } stcb->asoc.last_used_address = laddr; atomic_add_int(&sifa->refcount, 1); return (sifa); } if (start_at_beginning == 0) { stcb->asoc.last_used_address = NULL; goto sctp_from_the_top2; } return (NULL); } static struct sctp_ifa * sctp_select_nth_preferred_addr_from_ifn_boundall(struct sctp_ifn *ifn, struct sctp_tcb *stcb, int non_asoc_addr_ok, uint8_t dest_is_loop, uint8_t dest_is_priv, int addr_wanted, sa_family_t fam, sctp_route_t * ro ) { struct sctp_ifa *ifa, *sifa; int num_eligible_addr = 0; #ifdef INET6 struct sockaddr_in6 sin6, lsa6; if (fam == AF_INET6) { memcpy(&sin6, &ro->ro_dst, sizeof(struct sockaddr_in6)); (void)sa6_recoverscope(&sin6); } #endif /* INET6 */ LIST_FOREACH(ifa, &ifn->ifalist, next_ifa) { if ((ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) continue; sifa = sctp_is_ifa_addr_preferred(ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; #ifdef INET6 if (fam == AF_INET6 && dest_is_loop && sifa->src_is_loop && sifa->src_is_priv) { /* * don't allow fe80::1 to be a src on loop ::1, we * don't list it to the peer so we will get an * abort. */ continue; } if (fam == AF_INET6 && IN6_IS_ADDR_LINKLOCAL(&sifa->address.sin6.sin6_addr) && IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) { /* * link-local <-> link-local must belong to the same * scope. */ memcpy(&lsa6, &sifa->address.sin6, sizeof(struct sockaddr_in6)); (void)sa6_recoverscope(&lsa6); if (sin6.sin6_scope_id != lsa6.sin6_scope_id) { continue; } } #endif /* INET6 */ /* * Check if the IPv6 address matches to next-hop. In the * mobile case, old IPv6 address may be not deleted from the * interface. Then, the interface has previous and new * addresses. We should use one corresponding to the * next-hop. (by micchie) */ #ifdef INET6 if (stcb && fam == AF_INET6 && sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_BASE)) { if (sctp_v6src_match_nexthop(&sifa->address.sin6, ro) == 0) { continue; } } #endif /* Avoid topologically incorrect IPv4 address */ if (stcb && fam == AF_INET && sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_BASE)) { if (sctp_v4src_match_nexthop(sifa, ro) == 0) { continue; } } if (stcb) { if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* * It is restricted for some reason.. * probably not yet added. */ continue; } } if (num_eligible_addr >= addr_wanted) { return (sifa); } num_eligible_addr++; } return (NULL); } static int sctp_count_num_preferred_boundall(struct sctp_ifn *ifn, struct sctp_tcb *stcb, int non_asoc_addr_ok, uint8_t dest_is_loop, uint8_t dest_is_priv, sa_family_t fam) { struct sctp_ifa *ifa, *sifa; int num_eligible_addr = 0; LIST_FOREACH(ifa, &ifn->ifalist, next_ifa) { if ((ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) { continue; } sifa = sctp_is_ifa_addr_preferred(ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) { continue; } if (stcb) { if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* * It is restricted for some reason.. * probably not yet added. */ continue; } } num_eligible_addr++; } return (num_eligible_addr); } static struct sctp_ifa * sctp_choose_boundall(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, sctp_route_t * ro, uint32_t vrf_id, uint8_t dest_is_priv, uint8_t dest_is_loop, int non_asoc_addr_ok, sa_family_t fam) { int cur_addr_num = 0, num_preferred = 0; void *ifn; struct sctp_ifn *sctp_ifn, *looked_at = NULL, *emit_ifn; struct sctp_ifa *sctp_ifa, *sifa; uint32_t ifn_index; struct sctp_vrf *vrf; /*- * For boundall we can use any address in the association. * If non_asoc_addr_ok is set we can use any address (at least in * theory). So we look for preferred addresses first. If we find one, * we use it. Otherwise we next try to get an address on the * interface, which we should be able to do (unless non_asoc_addr_ok * is false and we are routed out that way). In these cases where we * can't use the address of the interface we go through all the * ifn's looking for an address we can use and fill that in. Punting * means we send back address 0, which will probably cause problems * actually since then IP will fill in the address of the route ifn, * which means we probably already rejected it.. i.e. here comes an * abort :-<. */ vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) return (NULL); ifn = SCTP_GET_IFN_VOID_FROM_ROUTE(ro); ifn_index = SCTP_GET_IF_INDEX_FROM_ROUTE(ro); emit_ifn = looked_at = sctp_ifn = sctp_find_ifn(ifn, ifn_index); if (sctp_ifn == NULL) { /* ?? We don't have this guy ?? */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "No ifn emit interface?\n"); goto bound_all_plan_b; } SCTPDBG(SCTP_DEBUG_OUTPUT2, "ifn_index:%d name:%s is emit interface\n", ifn_index, sctp_ifn->ifn_name); if (net) { cur_addr_num = net->indx_of_eligible_next_to_use; } num_preferred = sctp_count_num_preferred_boundall(sctp_ifn, stcb, non_asoc_addr_ok, dest_is_loop, dest_is_priv, fam); SCTPDBG(SCTP_DEBUG_OUTPUT2, "Found %d preferred source addresses for intf:%s\n", num_preferred, sctp_ifn->ifn_name); if (num_preferred == 0) { /* * no eligible addresses, we must use some other interface * address if we can find one. */ goto bound_all_plan_b; } /* * Ok we have num_eligible_addr set with how many we can use, this * may vary from call to call due to addresses being deprecated * etc.. */ if (cur_addr_num >= num_preferred) { cur_addr_num = 0; } /* * select the nth address from the list (where cur_addr_num is the * nth) and 0 is the first one, 1 is the second one etc... */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "cur_addr_num:%d\n", cur_addr_num); sctp_ifa = sctp_select_nth_preferred_addr_from_ifn_boundall(sctp_ifn, stcb, non_asoc_addr_ok, dest_is_loop, dest_is_priv, cur_addr_num, fam, ro); /* if sctp_ifa is NULL something changed??, fall to plan b. */ if (sctp_ifa) { atomic_add_int(&sctp_ifa->refcount, 1); if (net) { /* save off where the next one we will want */ net->indx_of_eligible_next_to_use = cur_addr_num + 1; } return (sctp_ifa); } /* * plan_b: Look at all interfaces and find a preferred address. If * no preferred fall through to plan_c. */ bound_all_plan_b: SCTPDBG(SCTP_DEBUG_OUTPUT2, "Trying Plan B\n"); LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { SCTPDBG(SCTP_DEBUG_OUTPUT2, "Examine interface %s\n", sctp_ifn->ifn_name); if (dest_is_loop == 0 && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { /* wrong base scope */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "skip\n"); continue; } if ((sctp_ifn == looked_at) && looked_at) { /* already looked at this guy */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "already seen\n"); continue; } num_preferred = sctp_count_num_preferred_boundall(sctp_ifn, stcb, non_asoc_addr_ok, dest_is_loop, dest_is_priv, fam); SCTPDBG(SCTP_DEBUG_OUTPUT2, "Found ifn:%p %d preferred source addresses\n", ifn, num_preferred); if (num_preferred == 0) { /* None on this interface. */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "No prefered -- skipping to next\n"); continue; } SCTPDBG(SCTP_DEBUG_OUTPUT2, "num preferred:%d on interface:%p cur_addr_num:%d\n", num_preferred, sctp_ifn, cur_addr_num); /* * Ok we have num_eligible_addr set with how many we can * use, this may vary from call to call due to addresses * being deprecated etc.. */ if (cur_addr_num >= num_preferred) { cur_addr_num = 0; } sifa = sctp_select_nth_preferred_addr_from_ifn_boundall(sctp_ifn, stcb, non_asoc_addr_ok, dest_is_loop, dest_is_priv, cur_addr_num, fam, ro); if (sifa == NULL) continue; if (net) { net->indx_of_eligible_next_to_use = cur_addr_num + 1; SCTPDBG(SCTP_DEBUG_OUTPUT2, "we selected %d\n", cur_addr_num); SCTPDBG(SCTP_DEBUG_OUTPUT2, "Source:"); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &sifa->address.sa); SCTPDBG(SCTP_DEBUG_OUTPUT2, "Dest:"); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &net->ro._l_addr.sa); } atomic_add_int(&sifa->refcount, 1); return (sifa); } /* plan_c: do we have an acceptable address on the emit interface */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "Trying Plan C: find acceptable on interface\n"); if (emit_ifn == NULL) { goto plan_d; } LIST_FOREACH(sctp_ifa, &emit_ifn->ifalist, next_ifa) { if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) continue; sifa = sctp_is_ifa_addr_acceptable(sctp_ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (stcb) { if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* * It is restricted for some reason.. * probably not yet added. */ continue; } } atomic_add_int(&sifa->refcount, 1); return (sifa); } plan_d: /* * plan_d: We are in trouble. No preferred address on the emit * interface. And not even a preferred address on all interfaces. Go * out and see if we can find an acceptable address somewhere * amongst all interfaces. */ SCTPDBG(SCTP_DEBUG_OUTPUT2, "Trying Plan D\n"); LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { if (dest_is_loop == 0 && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { /* wrong base scope */ continue; } if ((sctp_ifn == looked_at) && looked_at) /* already looked at this guy */ continue; LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if ((sctp_ifa->localifa_flags & SCTP_ADDR_DEFER_USE) && (non_asoc_addr_ok == 0)) continue; sifa = sctp_is_ifa_addr_acceptable(sctp_ifa, dest_is_loop, dest_is_priv, fam); if (sifa == NULL) continue; if (stcb) { if (((non_asoc_addr_ok == 0) && (sctp_is_addr_restricted(stcb, sifa))) || (non_asoc_addr_ok && (sctp_is_addr_restricted(stcb, sifa)) && (!sctp_is_addr_pending(stcb, sifa)))) { /* * It is restricted for some * reason.. probably not yet added. */ continue; } } atomic_add_int(&sifa->refcount, 1); return (sifa); } } /* * Ok we can find NO address to source from that is not on our * restricted list and non_asoc_address is NOT ok, or it is on our * restricted list. We can't source to it :-( */ return (NULL); } /* tcb may be NULL */ struct sctp_ifa * sctp_source_address_selection(struct sctp_inpcb *inp, struct sctp_tcb *stcb, sctp_route_t * ro, struct sctp_nets *net, int non_asoc_addr_ok, uint32_t vrf_id) { struct sockaddr_in *to = (struct sockaddr_in *)&ro->ro_dst; #ifdef INET6 struct sockaddr_in6 *to6 = (struct sockaddr_in6 *)&ro->ro_dst; #endif struct sctp_ifa *answer; uint8_t dest_is_priv, dest_is_loop; sa_family_t fam; /*- * Rules: - Find the route if needed, cache if I can. - Look at * interface address in route, Is it in the bound list. If so we * have the best source. - If not we must rotate amongst the * addresses. * * Cavets and issues * * Do we need to pay attention to scope. We can have a private address * or a global address we are sourcing or sending to. So if we draw * it out * zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz * For V4 *------------------------------------------ * source * dest * result * ----------------------------------------- * Private * Global * NAT * ----------------------------------------- * Private * Private * No problem * ----------------------------------------- * Global * Private * Huh, How will this work? * ----------------------------------------- * Global * Global * No Problem *------------------------------------------ * zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz * For V6 *------------------------------------------ * source * dest * result * ----------------------------------------- * Linklocal * Global * * ----------------------------------------- * Linklocal * Linklocal * No problem * ----------------------------------------- * Global * Linklocal * Huh, How will this work? * ----------------------------------------- * Global * Global * No Problem *------------------------------------------ * zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz * * And then we add to that what happens if there are multiple addresses * assigned to an interface. Remember the ifa on a ifn is a linked * list of addresses. So one interface can have more than one IP * address. What happens if we have both a private and a global * address? Do we then use context of destination to sort out which * one is best? And what about NAT's sending P->G may get you a NAT * translation, or should you select the G thats on the interface in * preference. * * Decisions: * * - count the number of addresses on the interface. * - if it is one, no problem except case . * For we will assume a NAT out there. * - if there are more than one, then we need to worry about scope P * or G. We should prefer G -> G and P -> P if possible. * Then as a secondary fall back to mixed types G->P being a last * ditch one. * - The above all works for bound all, but bound specific we need to * use the same concept but instead only consider the bound * addresses. If the bound set is NOT assigned to the interface then * we must use rotation amongst the bound addresses.. */ if (ro->ro_rt == NULL) { /* * Need a route to cache. */ SCTP_RTALLOC(ro, vrf_id); } if (ro->ro_rt == NULL) { return (NULL); } fam = to->sin_family; dest_is_priv = dest_is_loop = 0; /* Setup our scopes for the destination */ switch (fam) { case AF_INET: /* Scope based on outbound address */ if (IN4_ISLOOPBACK_ADDRESS(&to->sin_addr)) { dest_is_loop = 1; if (net != NULL) { /* mark it as local */ net->addr_is_local = 1; } } else if ((IN4_ISPRIVATE_ADDRESS(&to->sin_addr))) { dest_is_priv = 1; } break; #ifdef INET6 case AF_INET6: /* Scope based on outbound address */ if (IN6_IS_ADDR_LOOPBACK(&to6->sin6_addr) || SCTP_ROUTE_IS_REAL_LOOP(ro)) { /* * If the address is a loopback address, which * consists of "::1" OR "fe80::1%lo0", we are * loopback scope. But we don't use dest_is_priv * (link local addresses). */ dest_is_loop = 1; if (net != NULL) { /* mark it as local */ net->addr_is_local = 1; } } else if (IN6_IS_ADDR_LINKLOCAL(&to6->sin6_addr)) { dest_is_priv = 1; } break; #endif } SCTPDBG(SCTP_DEBUG_OUTPUT2, "Select source addr for:"); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)to); SCTP_IPI_ADDR_RLOCK(); if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* * Bound all case */ answer = sctp_choose_boundall(inp, stcb, net, ro, vrf_id, dest_is_priv, dest_is_loop, non_asoc_addr_ok, fam); SCTP_IPI_ADDR_RUNLOCK(); return (answer); } /* * Subset bound case */ if (stcb) { answer = sctp_choose_boundspecific_stcb(inp, stcb, net, ro, vrf_id, dest_is_priv, dest_is_loop, non_asoc_addr_ok, fam); } else { answer = sctp_choose_boundspecific_inp(inp, ro, vrf_id, non_asoc_addr_ok, dest_is_priv, dest_is_loop, fam); } SCTP_IPI_ADDR_RUNLOCK(); return (answer); } static int sctp_find_cmsg(int c_type, void *data, struct mbuf *control, int cpsize) { struct cmsghdr cmh; int tlen, at; tlen = SCTP_BUF_LEN(control); at = 0; /* * Independent of how many mbufs, find the c_type inside the control * structure and copy out the data. */ while (at < tlen) { if ((tlen - at) < (int)CMSG_ALIGN(sizeof(cmh))) { /* not enough room for one more we are done. */ return (0); } m_copydata(control, at, sizeof(cmh), (caddr_t)&cmh); if (((int)cmh.cmsg_len + at) > tlen) { /* * this is real messed up since there is not enough * data here to cover the cmsg header. We are done. */ return (0); } if ((cmh.cmsg_level == IPPROTO_SCTP) && (c_type == cmh.cmsg_type)) { /* found the one we want, copy it out */ at += CMSG_ALIGN(sizeof(struct cmsghdr)); if ((int)(cmh.cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr))) < cpsize) { /* * space of cmsg_len after header not big * enough */ return (0); } m_copydata(control, at, cpsize, data); return (1); } else { at += CMSG_ALIGN(cmh.cmsg_len); if (cmh.cmsg_len == 0) { break; } } } /* not found */ return (0); } static struct mbuf * sctp_add_cookie(struct sctp_inpcb *inp, struct mbuf *init, int init_offset, struct mbuf *initack, int initack_offset, struct sctp_state_cookie *stc_in, uint8_t ** signature) { struct mbuf *copy_init, *copy_initack, *m_at, *sig, *mret; struct sctp_state_cookie *stc; struct sctp_paramhdr *ph; uint8_t *foo; int sig_offset; uint16_t cookie_sz; mret = NULL; mret = sctp_get_mbuf_for_msg((sizeof(struct sctp_state_cookie) + sizeof(struct sctp_paramhdr)), 0, M_DONTWAIT, 1, MT_DATA); if (mret == NULL) { return (NULL); } copy_init = SCTP_M_COPYM(init, init_offset, M_COPYALL, M_DONTWAIT); if (copy_init == NULL) { sctp_m_freem(mret); return (NULL); } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = copy_init; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif copy_initack = SCTP_M_COPYM(initack, initack_offset, M_COPYALL, M_DONTWAIT); if (copy_initack == NULL) { sctp_m_freem(mret); sctp_m_freem(copy_init); return (NULL); } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = copy_initack; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif /* easy side we just drop it on the end */ ph = mtod(mret, struct sctp_paramhdr *); SCTP_BUF_LEN(mret) = sizeof(struct sctp_state_cookie) + sizeof(struct sctp_paramhdr); stc = (struct sctp_state_cookie *)((caddr_t)ph + sizeof(struct sctp_paramhdr)); ph->param_type = htons(SCTP_STATE_COOKIE); ph->param_length = 0; /* fill in at the end */ /* Fill in the stc cookie data */ memcpy(stc, stc_in, sizeof(struct sctp_state_cookie)); /* tack the INIT and then the INIT-ACK onto the chain */ cookie_sz = 0; m_at = mret; for (m_at = mret; m_at; m_at = SCTP_BUF_NEXT(m_at)) { cookie_sz += SCTP_BUF_LEN(m_at); if (SCTP_BUF_NEXT(m_at) == NULL) { SCTP_BUF_NEXT(m_at) = copy_init; break; } } for (m_at = copy_init; m_at; m_at = SCTP_BUF_NEXT(m_at)) { cookie_sz += SCTP_BUF_LEN(m_at); if (SCTP_BUF_NEXT(m_at) == NULL) { SCTP_BUF_NEXT(m_at) = copy_initack; break; } } for (m_at = copy_initack; m_at; m_at = SCTP_BUF_NEXT(m_at)) { cookie_sz += SCTP_BUF_LEN(m_at); if (SCTP_BUF_NEXT(m_at) == NULL) { break; } } sig = sctp_get_mbuf_for_msg(SCTP_SECRET_SIZE, 0, M_DONTWAIT, 1, MT_DATA); if (sig == NULL) { /* no space, so free the entire chain */ sctp_m_freem(mret); return (NULL); } SCTP_BUF_LEN(sig) = 0; SCTP_BUF_NEXT(m_at) = sig; sig_offset = 0; foo = (uint8_t *) (mtod(sig, caddr_t)+sig_offset); memset(foo, 0, SCTP_SIGNATURE_SIZE); *signature = foo; SCTP_BUF_LEN(sig) += SCTP_SIGNATURE_SIZE; cookie_sz += SCTP_SIGNATURE_SIZE; ph->param_length = htons(cookie_sz); return (mret); } static uint8_t sctp_get_ect(struct sctp_tcb *stcb, struct sctp_tmit_chunk *chk) { uint8_t this_random; /* Huh? */ if (SCTP_BASE_SYSCTL(sctp_ecn_enable) == 0) return (0); if (SCTP_BASE_SYSCTL(sctp_ecn_nonce) == 0) /* no nonce, always return ECT0 */ return (SCTP_ECT0_BIT); if (stcb->asoc.peer_supports_ecn_nonce == 0) { /* Peer does NOT support it, so we send a ECT0 only */ return (SCTP_ECT0_BIT); } if (chk == NULL) return (SCTP_ECT0_BIT); if ((stcb->asoc.hb_random_idx > 3) || ((stcb->asoc.hb_random_idx == 3) && (stcb->asoc.hb_ect_randombit > 7))) { uint32_t rndval; warp_drive_sa: rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep); memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values)); this_random = stcb->asoc.hb_random_values[0]; stcb->asoc.hb_random_idx = 0; stcb->asoc.hb_ect_randombit = 0; } else { if (stcb->asoc.hb_ect_randombit > 7) { stcb->asoc.hb_ect_randombit = 0; stcb->asoc.hb_random_idx++; if (stcb->asoc.hb_random_idx > 3) { goto warp_drive_sa; } } this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx]; } if ((this_random >> stcb->asoc.hb_ect_randombit) & 0x01) { if (chk != NULL) /* ECN Nonce stuff */ chk->rec.data.ect_nonce = SCTP_ECT1_BIT; stcb->asoc.hb_ect_randombit++; return (SCTP_ECT1_BIT); } else { stcb->asoc.hb_ect_randombit++; return (SCTP_ECT0_BIT); } } static int sctp_lowlevel_chunk_output(struct sctp_inpcb *inp, struct sctp_tcb *stcb, /* may be NULL */ struct sctp_nets *net, struct sockaddr *to, struct mbuf *m, uint32_t auth_offset, struct sctp_auth_chunk *auth, int nofragment_flag, int ecn_ok, struct sctp_tmit_chunk *chk, int out_of_asoc_ok, uint16_t port, int so_locked, #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif union sctp_sockstore *over_addr ) /* nofragment_flag to tell if IP_DF should be set (IPv4 only) */ { /* * Given a mbuf chain (via SCTP_BUF_NEXT()) that holds a packet * header WITH an SCTPHDR but no IP header, endpoint inp and sa * structure: - fill in the HMAC digest of any AUTH chunk in the * packet. - calculate and fill in the SCTP checksum. - prepend an * IP address header. - if boundall use INADDR_ANY. - if * boundspecific do source address selection. - set fragmentation * option for ipV4. - On return from IP output, check/adjust mtu * size of output interface and smallest_mtu size as well. */ /* Will need ifdefs around this */ struct mbuf *o_pak; struct mbuf *newm; struct sctphdr *sctphdr; int packet_length; uint32_t csum; int ret; uint32_t vrf_id; sctp_route_t *ro = NULL; struct udphdr *udp = NULL; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so = NULL; #endif if ((net) && (net->dest_state & SCTP_ADDR_OUT_OF_SCOPE)) { SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EFAULT); sctp_m_freem(m); return (EFAULT); } if (stcb) { vrf_id = stcb->asoc.vrf_id; } else { vrf_id = inp->def_vrf_id; } /* fill in the HMAC digest for any AUTH chunk in the packet */ if ((auth != NULL) && (stcb != NULL)) { sctp_fill_hmac_digest_m(m, auth_offset, auth, stcb); } /* Calculate the csum and fill in the length of the packet */ sctphdr = mtod(m, struct sctphdr *); if (SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback) && (stcb) && (to->sa_family == AF_INET) && (stcb->asoc.loopback_scope)) { sctphdr->checksum = 0; /* * This can probably now be taken out since my audit shows * no more bad pktlen's coming in. But we will wait a while * yet. */ packet_length = sctp_calculate_len(m); } else { sctphdr->checksum = 0; csum = sctp_calculate_sum(m, &packet_length, 0); sctphdr->checksum = csum; } if (to->sa_family == AF_INET) { struct ip *ip = NULL; sctp_route_t iproute; uint8_t tos_value; if (port) { newm = sctp_get_mbuf_for_msg(sizeof(struct ip) + sizeof(struct udphdr), 1, M_DONTWAIT, 1, MT_DATA); } else { newm = sctp_get_mbuf_for_msg(sizeof(struct ip), 1, M_DONTWAIT, 1, MT_DATA); } if (newm == NULL) { sctp_m_freem(m); SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } if (port) { SCTP_ALIGN_TO_END(newm, sizeof(struct ip) + sizeof(struct udphdr)); SCTP_BUF_LEN(newm) = sizeof(struct ip) + sizeof(struct udphdr); packet_length += sizeof(struct ip) + sizeof(struct udphdr); } else { SCTP_ALIGN_TO_END(newm, sizeof(struct ip)); SCTP_BUF_LEN(newm) = sizeof(struct ip); packet_length += sizeof(struct ip); } SCTP_BUF_NEXT(newm) = m; m = newm; ip = mtod(m, struct ip *); ip->ip_v = IPVERSION; ip->ip_hl = (sizeof(struct ip) >> 2); if (net) { tos_value = net->tos_flowlabel & 0x000000ff; } else { tos_value = inp->ip_inp.inp.inp_ip_tos; } if ((nofragment_flag) && (port == 0)) { #if defined(WITH_CONVERT_IP_OFF) || defined(__FreeBSD__) || defined(__APPLE__) || defined(__Userspace__) ip->ip_off = IP_DF; #else ip->ip_off = htons(IP_DF); #endif } else ip->ip_off = 0; /* FreeBSD has a function for ip_id's */ ip->ip_id = ip_newid(); ip->ip_ttl = inp->ip_inp.inp.inp_ip_ttl; ip->ip_len = packet_length; if (stcb) { if ((stcb->asoc.ecn_allowed) && ecn_ok) { /* Enable ECN */ ip->ip_tos = ((u_char)(tos_value & 0xfc) | sctp_get_ect(stcb, chk)); } else { /* No ECN */ ip->ip_tos = (u_char)(tos_value & 0xfc); } } else { /* no association at all */ ip->ip_tos = (tos_value & 0xfc); } if (port) { ip->ip_p = IPPROTO_UDP; } else { ip->ip_p = IPPROTO_SCTP; } ip->ip_sum = 0; if (net == NULL) { ro = &iproute; memset(&iproute, 0, sizeof(iproute)); memcpy(&ro->ro_dst, to, to->sa_len); } else { ro = (sctp_route_t *) & net->ro; } /* Now the address selection part */ ip->ip_dst.s_addr = ((struct sockaddr_in *)to)->sin_addr.s_addr; /* call the routine to select the src address */ if (net && out_of_asoc_ok == 0) { if (net->ro._s_addr && (net->ro._s_addr->localifa_flags & (SCTP_BEING_DELETED | SCTP_ADDR_IFA_UNUSEABLE))) { sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; net->src_addr_selected = 0; if (ro->ro_rt) { RTFREE(ro->ro_rt); ro->ro_rt = NULL; } } if (net->src_addr_selected == 0) { /* Cache the source address */ net->ro._s_addr = sctp_source_address_selection(inp, stcb, ro, net, 0, vrf_id); net->src_addr_selected = 1; } if (net->ro._s_addr == NULL) { /* No route to host */ net->src_addr_selected = 0; goto no_route; } ip->ip_src = net->ro._s_addr->address.sin.sin_addr; } else { if (over_addr == NULL) { struct sctp_ifa *_lsrc; _lsrc = sctp_source_address_selection(inp, stcb, ro, net, out_of_asoc_ok, vrf_id); if (_lsrc == NULL) { goto no_route; } ip->ip_src = _lsrc->address.sin.sin_addr; sctp_free_ifa(_lsrc); } else { ip->ip_src = over_addr->sin.sin_addr; SCTP_RTALLOC((&ro->ro_rt), vrf_id); } } if (port) { udp = (struct udphdr *)(ip + 1); udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); udp->uh_dport = port; udp->uh_ulen = htons(packet_length - sizeof(struct ip)); udp->uh_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, udp->uh_ulen + htons(IPPROTO_UDP)); } /* * If source address selection fails and we find no route * then the ip_output should fail as well with a * NO_ROUTE_TO_HOST type error. We probably should catch * that somewhere and abort the association right away * (assuming this is an INIT being sent). */ if ((ro->ro_rt == NULL)) { /* * src addr selection failed to find a route (or * valid source addr), so we can't get there from * here (yet)! */ no_route: SCTPDBG(SCTP_DEBUG_OUTPUT1, "%s: dropped packet - no valid source addr\n", __FUNCTION__); if (net) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Destination was "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT1, &net->ro._l_addr.sa); if (net->dest_state & SCTP_ADDR_CONFIRMED) { if ((net->dest_state & SCTP_ADDR_REACHABLE) && stcb) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "no route takes interface %p down\n", net); sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, stcb, SCTP_FAILED_THRESHOLD, (void *)net, so_locked); net->dest_state &= ~SCTP_ADDR_REACHABLE; net->dest_state |= SCTP_ADDR_NOT_REACHABLE; /* * JRS 5/14/07 - If a * destination is * unreachable, the PF bit * is turned off. This * allows an unambiguous use * of the PF bit for * destinations that are * reachable but potentially * failed. If the * destination is set to the * unreachable state, also * set the destination to * the PF state. */ /* * Add debug message here if * destination is not in PF * state. */ /* * Stop any running T3 * timers here? */ if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf)) { net->dest_state &= ~SCTP_ADDR_PF; SCTPDBG(SCTP_DEBUG_OUTPUT1, "Destination %p moved from PF to unreachable.\n", net); } } } if (stcb) { if (net == stcb->asoc.primary_destination) { /* need a new primary */ struct sctp_nets *alt; alt = sctp_find_alternate_net(stcb, net, 0); if (alt != net) { if (sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, alt) == 0) { net->dest_state |= SCTP_ADDR_WAS_PRIMARY; if (net->ro._s_addr) { sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; } net->src_addr_selected = 0; } } } } } SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EHOSTUNREACH); sctp_m_freem(m); return (EHOSTUNREACH); } if (ro != &iproute) { memcpy(&iproute, ro, sizeof(*ro)); } SCTPDBG(SCTP_DEBUG_OUTPUT3, "Calling ipv4 output routine from low level src addr:%x\n", (uint32_t) (ntohl(ip->ip_src.s_addr))); SCTPDBG(SCTP_DEBUG_OUTPUT3, "Destination is %x\n", (uint32_t) (ntohl(ip->ip_dst.s_addr))); SCTPDBG(SCTP_DEBUG_OUTPUT3, "RTP route is %p through\n", ro->ro_rt); if (SCTP_GET_HEADER_FOR_OUTPUT(o_pak)) { /* failed to prepend data, give up */ SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); sctp_m_freem(m); return (ENOMEM); } #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(m, packet_length); #endif SCTP_ATTACH_CHAIN(o_pak, m, packet_length); if (port) { SCTP_ENABLE_UDP_CSUM(o_pak); } /* send it out. table id is taken from stcb */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if ((SCTP_BASE_SYSCTL(sctp_output_unlocked)) && (so_locked)) { so = SCTP_INP_SO(inp); SCTP_SOCKET_UNLOCK(so, 0); } #endif SCTP_IP_OUTPUT(ret, o_pak, ro, stcb, vrf_id); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if ((SCTP_BASE_SYSCTL(sctp_output_unlocked)) && (so_locked)) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 0); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); } #endif SCTP_STAT_INCR(sctps_sendpackets); SCTP_STAT_INCR_COUNTER64(sctps_outpackets); if (ret) SCTP_STAT_INCR(sctps_senderrors); SCTPDBG(SCTP_DEBUG_OUTPUT3, "IP output returns %d\n", ret); if (net == NULL) { /* free tempy routes */ if (ro->ro_rt) { RTFREE(ro->ro_rt); ro->ro_rt = NULL; } } else { /* PMTU check versus smallest asoc MTU goes here */ if ((ro->ro_rt != NULL) && (net->ro._s_addr)) { uint32_t mtu; mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, ro->ro_rt); if (mtu && (stcb->asoc.smallest_mtu > mtu)) { #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("sctp_mtu_size_reset called after ip_output mtu-change:%d\n", mtu); #endif sctp_mtu_size_reset(inp, &stcb->asoc, mtu); net->mtu = mtu; } } else if (ro->ro_rt == NULL) { /* route was freed */ if (net->ro._s_addr && net->src_addr_selected) { sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; } net->src_addr_selected = 0; } } return (ret); } #ifdef INET6 else if (to->sa_family == AF_INET6) { uint32_t flowlabel; struct ip6_hdr *ip6h; struct route_in6 ip6route; struct ifnet *ifp; u_char flowTop; uint16_t flowBottom; u_char tosBottom, tosTop; struct sockaddr_in6 *sin6, tmp, *lsa6, lsa6_tmp; int prev_scope = 0; struct sockaddr_in6 lsa6_storage; int error; u_short prev_port = 0; if (net != NULL) { flowlabel = net->tos_flowlabel; } else { flowlabel = ((struct in6pcb *)inp)->in6p_flowinfo; } if (port) { newm = sctp_get_mbuf_for_msg(sizeof(struct ip6_hdr) + sizeof(struct udphdr), 1, M_DONTWAIT, 1, MT_DATA); } else { newm = sctp_get_mbuf_for_msg(sizeof(struct ip6_hdr), 1, M_DONTWAIT, 1, MT_DATA); } if (newm == NULL) { sctp_m_freem(m); SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } if (port) { SCTP_ALIGN_TO_END(newm, sizeof(struct ip6_hdr) + sizeof(struct udphdr)); SCTP_BUF_LEN(newm) = sizeof(struct ip6_hdr) + sizeof(struct udphdr); packet_length += sizeof(struct ip6_hdr) + sizeof(struct udphdr); } else { SCTP_ALIGN_TO_END(newm, sizeof(struct ip6_hdr)); SCTP_BUF_LEN(newm) = sizeof(struct ip6_hdr); packet_length += sizeof(struct ip6_hdr); } SCTP_BUF_NEXT(newm) = m; m = newm; ip6h = mtod(m, struct ip6_hdr *); /* * We assume here that inp_flow is in host byte order within * the TCB! */ flowBottom = flowlabel & 0x0000ffff; flowTop = ((flowlabel & 0x000f0000) >> 16); tosTop = (((flowlabel & 0xf0) >> 4) | IPV6_VERSION); /* protect *sin6 from overwrite */ sin6 = (struct sockaddr_in6 *)to; tmp = *sin6; sin6 = &tmp; /* KAME hack: embed scopeid */ if (sa6_embedscope(sin6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone)) != 0) { SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } if (net == NULL) { memset(&ip6route, 0, sizeof(ip6route)); ro = (sctp_route_t *) & ip6route; memcpy(&ro->ro_dst, sin6, sin6->sin6_len); } else { ro = (sctp_route_t *) & net->ro; } if (stcb != NULL) { if ((stcb->asoc.ecn_allowed) && ecn_ok) { /* Enable ECN */ tosBottom = (((((struct in6pcb *)inp)->in6p_flowinfo & 0x0c) | sctp_get_ect(stcb, chk)) << 4); } else { /* No ECN */ tosBottom = ((((struct in6pcb *)inp)->in6p_flowinfo & 0x0c) << 4); } } else { /* we could get no asoc if it is a O-O-T-B packet */ tosBottom = ((((struct in6pcb *)inp)->in6p_flowinfo & 0x0c) << 4); } ip6h->ip6_flow = htonl(((tosTop << 24) | ((tosBottom | flowTop) << 16) | flowBottom)); if (port) { ip6h->ip6_nxt = IPPROTO_UDP; } else { ip6h->ip6_nxt = IPPROTO_SCTP; } ip6h->ip6_plen = (packet_length - sizeof(struct ip6_hdr)); ip6h->ip6_dst = sin6->sin6_addr; /* * Add SRC address selection here: we can only reuse to a * limited degree the kame src-addr-sel, since we can try * their selection but it may not be bound. */ bzero(&lsa6_tmp, sizeof(lsa6_tmp)); lsa6_tmp.sin6_family = AF_INET6; lsa6_tmp.sin6_len = sizeof(lsa6_tmp); lsa6 = &lsa6_tmp; if (net && out_of_asoc_ok == 0) { if (net->ro._s_addr && (net->ro._s_addr->localifa_flags & (SCTP_BEING_DELETED | SCTP_ADDR_IFA_UNUSEABLE))) { sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; net->src_addr_selected = 0; if (ro->ro_rt) { RTFREE(ro->ro_rt); ro->ro_rt = NULL; } } if (net->src_addr_selected == 0) { sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; /* KAME hack: embed scopeid */ if (sa6_embedscope(sin6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone)) != 0) { SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } /* Cache the source address */ net->ro._s_addr = sctp_source_address_selection(inp, stcb, ro, net, 0, vrf_id); (void)sa6_recoverscope(sin6); net->src_addr_selected = 1; } if (net->ro._s_addr == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "V6:No route to host\n"); net->src_addr_selected = 0; goto no_route; } lsa6->sin6_addr = net->ro._s_addr->address.sin6.sin6_addr; } else { sin6 = (struct sockaddr_in6 *)&ro->ro_dst; /* KAME hack: embed scopeid */ if (sa6_embedscope(sin6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone)) != 0) { SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } if (over_addr == NULL) { struct sctp_ifa *_lsrc; _lsrc = sctp_source_address_selection(inp, stcb, ro, net, out_of_asoc_ok, vrf_id); if (_lsrc == NULL) { goto no_route; } lsa6->sin6_addr = _lsrc->address.sin6.sin6_addr; sctp_free_ifa(_lsrc); } else { lsa6->sin6_addr = over_addr->sin6.sin6_addr; SCTP_RTALLOC((&ro->ro_rt), vrf_id); } (void)sa6_recoverscope(sin6); } lsa6->sin6_port = inp->sctp_lport; if (ro->ro_rt == NULL) { /* * src addr selection failed to find a route (or * valid source addr), so we can't get there from * here! */ goto no_route; } /* * XXX: sa6 may not have a valid sin6_scope_id in the * non-SCOPEDROUTING case. */ bzero(&lsa6_storage, sizeof(lsa6_storage)); lsa6_storage.sin6_family = AF_INET6; lsa6_storage.sin6_len = sizeof(lsa6_storage); lsa6_storage.sin6_addr = lsa6->sin6_addr; if ((error = sa6_recoverscope(&lsa6_storage)) != 0) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "recover scope fails error %d\n", error); sctp_m_freem(m); return (error); } /* XXX */ lsa6_storage.sin6_addr = lsa6->sin6_addr; lsa6_storage.sin6_port = inp->sctp_lport; lsa6 = &lsa6_storage; ip6h->ip6_src = lsa6->sin6_addr; if (port) { udp = (struct udphdr *)(ip6h + 1); udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); udp->uh_dport = port; udp->uh_ulen = htons(packet_length - sizeof(struct ip6_hdr)); udp->uh_sum = 0; } /* * We set the hop limit now since there is a good chance * that our ro pointer is now filled */ ip6h->ip6_hlim = SCTP_GET_HLIM(inp, ro); ifp = SCTP_GET_IFN_VOID_FROM_ROUTE(ro); #ifdef SCTP_DEBUG /* Copy to be sure something bad is not happening */ sin6->sin6_addr = ip6h->ip6_dst; lsa6->sin6_addr = ip6h->ip6_src; #endif SCTPDBG(SCTP_DEBUG_OUTPUT3, "Calling ipv6 output routine from low level\n"); SCTPDBG(SCTP_DEBUG_OUTPUT3, "src: "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT3, (struct sockaddr *)lsa6); SCTPDBG(SCTP_DEBUG_OUTPUT3, "dst: "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT3, (struct sockaddr *)sin6); if (net) { sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; /* preserve the port and scope for link local send */ prev_scope = sin6->sin6_scope_id; prev_port = sin6->sin6_port; } if (SCTP_GET_HEADER_FOR_OUTPUT(o_pak)) { /* failed to prepend data, give up */ sctp_m_freem(m); SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(m, packet_length); #endif SCTP_ATTACH_CHAIN(o_pak, m, packet_length); if (port) { if ((udp->uh_sum = in6_cksum(o_pak, IPPROTO_UDP, sizeof(struct ip6_hdr), packet_length - sizeof(struct ip6_hdr))) == 0) { udp->uh_sum = 0xffff; } } /* send it out. table id is taken from stcb */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if ((SCTP_BASE_SYSCTL(sctp_output_unlocked)) && (so_locked)) { so = SCTP_INP_SO(inp); SCTP_SOCKET_UNLOCK(so, 0); } #endif SCTP_IP6_OUTPUT(ret, o_pak, (struct route_in6 *)ro, &ifp, stcb, vrf_id); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if ((SCTP_BASE_SYSCTL(sctp_output_unlocked)) && (so_locked)) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 0); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); } #endif if (net) { /* for link local this must be done */ sin6->sin6_scope_id = prev_scope; sin6->sin6_port = prev_port; } SCTPDBG(SCTP_DEBUG_OUTPUT3, "return from send is %d\n", ret); SCTP_STAT_INCR(sctps_sendpackets); SCTP_STAT_INCR_COUNTER64(sctps_outpackets); if (ret) { SCTP_STAT_INCR(sctps_senderrors); } if (net == NULL) { /* Now if we had a temp route free it */ if (ro->ro_rt) { RTFREE(ro->ro_rt); } } else { /* PMTU check versus smallest asoc MTU goes here */ if (ro->ro_rt == NULL) { /* Route was freed */ if (net->ro._s_addr && net->src_addr_selected) { sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; } net->src_addr_selected = 0; } if ((ro->ro_rt != NULL) && (net->ro._s_addr)) { uint32_t mtu; mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, ro->ro_rt); if (mtu && (stcb->asoc.smallest_mtu > mtu)) { #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("sctp_mtu_size_reset called after ip6_output mtu-change:%d\n", mtu); #endif sctp_mtu_size_reset(inp, &stcb->asoc, mtu); net->mtu = mtu; } } else if (ifp) { if (ND_IFINFO(ifp)->linkmtu && (stcb->asoc.smallest_mtu > ND_IFINFO(ifp)->linkmtu)) { #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("sctp_mtu_size_reset called via ifp ND_IFINFO() linkmtu:%d\n", ND_IFINFO(ifp)->linkmtu); #endif sctp_mtu_size_reset(inp, &stcb->asoc, ND_IFINFO(ifp)->linkmtu); } } } return (ret); } #endif else { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Unknown protocol (TSNH) type %d\n", ((struct sockaddr *)to)->sa_family); sctp_m_freem(m); SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EFAULT); return (EFAULT); } } void sctp_send_initiate(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct mbuf *m, *m_at, *mp_last; struct sctp_nets *net; struct sctp_init_msg *initm; struct sctp_supported_addr_param *sup_addr; struct sctp_ecn_supported_param *ecn; struct sctp_prsctp_supported_param *prsctp; struct sctp_ecn_nonce_supported_param *ecn_nonce; struct sctp_supported_chunk_types_param *pr_supported; int cnt_inits_to = 0; int padval, ret; int num_ext; int p_len; /* INIT's always go to the primary (and usually ONLY address) */ mp_last = NULL; net = stcb->asoc.primary_destination; if (net == NULL) { net = TAILQ_FIRST(&stcb->asoc.nets); if (net == NULL) { /* TSNH */ return; } /* we confirm any address we send an INIT to */ net->dest_state &= ~SCTP_ADDR_UNCONFIRMED; (void)sctp_set_primary_addr(stcb, NULL, net); } else { /* we confirm any address we send an INIT to */ net->dest_state &= ~SCTP_ADDR_UNCONFIRMED; } SCTPDBG(SCTP_DEBUG_OUTPUT4, "Sending INIT\n"); #ifdef INET6 if (((struct sockaddr *)&(net->ro._l_addr))->sa_family == AF_INET6) { /* * special hook, if we are sending to link local it will not * show up in our private address count. */ struct sockaddr_in6 *sin6l; sin6l = &net->ro._l_addr.sin6; if (IN6_IS_ADDR_LINKLOCAL(&sin6l->sin6_addr)) cnt_inits_to = 1; } #endif if (SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { /* This case should not happen */ SCTPDBG(SCTP_DEBUG_OUTPUT4, "Sending INIT - failed timer?\n"); return; } /* start the INIT timer */ sctp_timer_start(SCTP_TIMER_TYPE_INIT, inp, stcb, net); m = sctp_get_mbuf_for_msg(MCLBYTES, 1, M_DONTWAIT, 1, MT_DATA); if (m == NULL) { /* No memory, INIT timer will re-attempt. */ SCTPDBG(SCTP_DEBUG_OUTPUT4, "Sending INIT - mbuf?\n"); return; } SCTP_BUF_LEN(m) = sizeof(struct sctp_init_msg); /* * assume peer supports asconf in order to be able to queue local * address changes while an INIT is in flight and before the assoc * is established. */ stcb->asoc.peer_supports_asconf = 1; /* Now lets put the SCTP header in place */ initm = mtod(m, struct sctp_init_msg *); initm->sh.src_port = inp->sctp_lport; initm->sh.dest_port = stcb->rport; initm->sh.v_tag = 0; initm->sh.checksum = 0; /* calculate later */ /* now the chunk header */ initm->msg.ch.chunk_type = SCTP_INITIATION; initm->msg.ch.chunk_flags = 0; /* fill in later from mbuf we build */ initm->msg.ch.chunk_length = 0; /* place in my tag */ initm->msg.init.initiate_tag = htonl(stcb->asoc.my_vtag); /* set up some of the credits. */ initm->msg.init.a_rwnd = htonl(max(SCTP_SB_LIMIT_RCV(inp->sctp_socket), SCTP_MINIMAL_RWND)); initm->msg.init.num_outbound_streams = htons(stcb->asoc.pre_open_streams); initm->msg.init.num_inbound_streams = htons(stcb->asoc.max_inbound_streams); initm->msg.init.initial_tsn = htonl(stcb->asoc.init_seq_number); /* now the address restriction */ sup_addr = (struct sctp_supported_addr_param *)((caddr_t)initm + sizeof(*initm)); sup_addr->ph.param_type = htons(SCTP_SUPPORTED_ADDRTYPE); #ifdef INET6 /* we support 2 types: IPv6/IPv4 */ sup_addr->ph.param_length = htons(sizeof(*sup_addr) + sizeof(uint16_t)); sup_addr->addr_type[0] = htons(SCTP_IPV4_ADDRESS); sup_addr->addr_type[1] = htons(SCTP_IPV6_ADDRESS); #else /* we support 1 type: IPv4 */ sup_addr->ph.param_length = htons(sizeof(*sup_addr) + sizeof(uint8_t)); sup_addr->addr_type[0] = htons(SCTP_IPV4_ADDRESS); sup_addr->addr_type[1] = htons(0); /* this is the padding */ #endif SCTP_BUF_LEN(m) += sizeof(*sup_addr) + sizeof(uint16_t); if (inp->sctp_ep.adaptation_layer_indicator) { struct sctp_adaptation_layer_indication *ali; ali = (struct sctp_adaptation_layer_indication *)( (caddr_t)sup_addr + sizeof(*sup_addr) + sizeof(uint16_t)); ali->ph.param_type = htons(SCTP_ULP_ADAPTATION); ali->ph.param_length = htons(sizeof(*ali)); ali->indication = ntohl(inp->sctp_ep.adaptation_layer_indicator); SCTP_BUF_LEN(m) += sizeof(*ali); ecn = (struct sctp_ecn_supported_param *)((caddr_t)ali + sizeof(*ali)); } else { ecn = (struct sctp_ecn_supported_param *)((caddr_t)sup_addr + sizeof(*sup_addr) + sizeof(uint16_t)); } /* now any cookie time extensions */ if (stcb->asoc.cookie_preserve_req) { struct sctp_cookie_perserve_param *cookie_preserve; cookie_preserve = (struct sctp_cookie_perserve_param *)(ecn); cookie_preserve->ph.param_type = htons(SCTP_COOKIE_PRESERVE); cookie_preserve->ph.param_length = htons( sizeof(*cookie_preserve)); cookie_preserve->time = htonl(stcb->asoc.cookie_preserve_req); SCTP_BUF_LEN(m) += sizeof(*cookie_preserve); ecn = (struct sctp_ecn_supported_param *)( (caddr_t)cookie_preserve + sizeof(*cookie_preserve)); stcb->asoc.cookie_preserve_req = 0; } /* ECN parameter */ if (SCTP_BASE_SYSCTL(sctp_ecn_enable) == 1) { ecn->ph.param_type = htons(SCTP_ECN_CAPABLE); ecn->ph.param_length = htons(sizeof(*ecn)); SCTP_BUF_LEN(m) += sizeof(*ecn); prsctp = (struct sctp_prsctp_supported_param *)((caddr_t)ecn + sizeof(*ecn)); } else { prsctp = (struct sctp_prsctp_supported_param *)((caddr_t)ecn); } /* And now tell the peer we do pr-sctp */ prsctp->ph.param_type = htons(SCTP_PRSCTP_SUPPORTED); prsctp->ph.param_length = htons(sizeof(*prsctp)); SCTP_BUF_LEN(m) += sizeof(*prsctp); /* And now tell the peer we do all the extensions */ pr_supported = (struct sctp_supported_chunk_types_param *) ((caddr_t)prsctp + sizeof(*prsctp)); pr_supported->ph.param_type = htons(SCTP_SUPPORTED_CHUNK_EXT); num_ext = 0; pr_supported->chunk_types[num_ext++] = SCTP_ASCONF; pr_supported->chunk_types[num_ext++] = SCTP_ASCONF_ACK; pr_supported->chunk_types[num_ext++] = SCTP_FORWARD_CUM_TSN; pr_supported->chunk_types[num_ext++] = SCTP_PACKET_DROPPED; pr_supported->chunk_types[num_ext++] = SCTP_STREAM_RESET; if (!SCTP_BASE_SYSCTL(sctp_auth_disable)) pr_supported->chunk_types[num_ext++] = SCTP_AUTHENTICATION; p_len = sizeof(*pr_supported) + num_ext; pr_supported->ph.param_length = htons(p_len); bzero((caddr_t)pr_supported + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); /* ECN nonce: And now tell the peer we support ECN nonce */ if (SCTP_BASE_SYSCTL(sctp_ecn_nonce)) { ecn_nonce = (struct sctp_ecn_nonce_supported_param *) ((caddr_t)pr_supported + SCTP_SIZE32(p_len)); ecn_nonce->ph.param_type = htons(SCTP_ECN_NONCE_SUPPORTED); ecn_nonce->ph.param_length = htons(sizeof(*ecn_nonce)); SCTP_BUF_LEN(m) += sizeof(*ecn_nonce); } /* add authentication parameters */ if (!SCTP_BASE_SYSCTL(sctp_auth_disable)) { struct sctp_auth_random *randp; struct sctp_auth_hmac_algo *hmacs; struct sctp_auth_chunk_list *chunks; /* attach RANDOM parameter, if available */ if (stcb->asoc.authinfo.random != NULL) { randp = (struct sctp_auth_random *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); p_len = sizeof(*randp) + stcb->asoc.authinfo.random_len; #ifdef SCTP_AUTH_DRAFT_04 randp->ph.param_type = htons(SCTP_RANDOM); randp->ph.param_length = htons(p_len); bcopy(stcb->asoc.authinfo.random->key, randp->random_data, stcb->asoc.authinfo.random_len); #else /* random key already contains the header */ bcopy(stcb->asoc.authinfo.random->key, randp, p_len); #endif /* zero out any padding required */ bzero((caddr_t)randp + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); } /* add HMAC_ALGO parameter */ hmacs = (struct sctp_auth_hmac_algo *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); p_len = sctp_serialize_hmaclist(stcb->asoc.local_hmacs, (uint8_t *) hmacs->hmac_ids); if (p_len > 0) { p_len += sizeof(*hmacs); hmacs->ph.param_type = htons(SCTP_HMAC_LIST); hmacs->ph.param_length = htons(p_len); /* zero out any padding required */ bzero((caddr_t)hmacs + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); } /* add CHUNKS parameter */ chunks = (struct sctp_auth_chunk_list *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); p_len = sctp_serialize_auth_chunks(stcb->asoc.local_auth_chunks, chunks->chunk_types); if (p_len > 0) { p_len += sizeof(*chunks); chunks->ph.param_type = htons(SCTP_CHUNK_LIST); chunks->ph.param_length = htons(p_len); /* zero out any padding required */ bzero((caddr_t)chunks + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); } } m_at = m; /* now the addresses */ { struct sctp_scoping scp; /* * To optimize this we could put the scoping stuff into a * structure and remove the individual uint8's from the * assoc structure. Then we could just sifa in the address * within the stcb.. but for now this is a quick hack to get * the address stuff teased apart. */ scp.ipv4_addr_legal = stcb->asoc.ipv4_addr_legal; scp.ipv6_addr_legal = stcb->asoc.ipv6_addr_legal; scp.loopback_scope = stcb->asoc.loopback_scope; scp.ipv4_local_scope = stcb->asoc.ipv4_local_scope; scp.local_scope = stcb->asoc.local_scope; scp.site_scope = stcb->asoc.site_scope; m_at = sctp_add_addresses_to_i_ia(inp, &scp, m_at, cnt_inits_to); } /* calulate the size and update pkt header and chunk header */ p_len = 0; for (m_at = m; m_at; m_at = SCTP_BUF_NEXT(m_at)) { if (SCTP_BUF_NEXT(m_at) == NULL) mp_last = m_at; p_len += SCTP_BUF_LEN(m_at); } initm->msg.ch.chunk_length = htons((p_len - sizeof(struct sctphdr))); /* * We sifa 0 here to NOT set IP_DF if its IPv4, we ignore the return * here since the timer will drive a retranmission. */ /* I don't expect this to execute but we will be safe here */ padval = p_len % 4; if ((padval) && (mp_last)) { /* * The compiler worries that mp_last may not be set even * though I think it is impossible :-> however we add * mp_last here just in case. */ ret = sctp_add_pad_tombuf(mp_last, (4 - padval)); if (ret) { /* Houston we have a problem, no space */ sctp_m_freem(m); return; } p_len += padval; } SCTPDBG(SCTP_DEBUG_OUTPUT4, "Sending INIT - calls lowlevel_output\n"); ret = sctp_lowlevel_chunk_output(inp, stcb, net, (struct sockaddr *)&net->ro._l_addr, m, 0, NULL, 0, 0, NULL, 0, net->port, so_locked, NULL); SCTPDBG(SCTP_DEBUG_OUTPUT4, "lowlevel_output - %d\n", ret); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); sctp_timer_start(SCTP_TIMER_TYPE_INIT, inp, stcb, net); (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); } struct mbuf * sctp_arethere_unrecognized_parameters(struct mbuf *in_initpkt, int param_offset, int *abort_processing, struct sctp_chunkhdr *cp) { /* * Given a mbuf containing an INIT or INIT-ACK with the param_offset * being equal to the beginning of the params i.e. (iphlen + * sizeof(struct sctp_init_msg) parse through the parameters to the * end of the mbuf verifying that all parameters are known. * * For unknown parameters build and return a mbuf with * UNRECOGNIZED_PARAMETER errors. If the flags indicate to stop * processing this chunk stop, and set *abort_processing to 1. * * By having param_offset be pre-set to where parameters begin it is * hoped that this routine may be reused in the future by new * features. */ struct sctp_paramhdr *phdr, params; struct mbuf *mat, *op_err; char tempbuf[SCTP_PARAM_BUFFER_SIZE]; int at, limit, pad_needed; uint16_t ptype, plen, padded_size; int err_at; *abort_processing = 0; mat = in_initpkt; err_at = 0; limit = ntohs(cp->chunk_length) - sizeof(struct sctp_init_chunk); at = param_offset; op_err = NULL; SCTPDBG(SCTP_DEBUG_OUTPUT1, "Check for unrecognized param's\n"); phdr = sctp_get_next_param(mat, at, ¶ms, sizeof(params)); while ((phdr != NULL) && ((size_t)limit >= sizeof(struct sctp_paramhdr))) { ptype = ntohs(phdr->param_type); plen = ntohs(phdr->param_length); if ((plen > limit) || (plen < sizeof(struct sctp_paramhdr))) { /* wacked parameter */ SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error %d\n", plen); goto invalid_size; } limit -= SCTP_SIZE32(plen); /*- * All parameters for all chunks that we know/understand are * listed here. We process them other places and make * appropriate stop actions per the upper bits. However this * is the generic routine processor's can call to get back * an operr.. to either incorporate (init-ack) or send. */ padded_size = SCTP_SIZE32(plen); switch (ptype) { /* Param's with variable size */ case SCTP_HEARTBEAT_INFO: case SCTP_STATE_COOKIE: case SCTP_UNRECOG_PARAM: case SCTP_ERROR_CAUSE_IND: /* ok skip fwd */ at += padded_size; break; /* Param's with variable size within a range */ case SCTP_CHUNK_LIST: case SCTP_SUPPORTED_CHUNK_EXT: if (padded_size > (sizeof(struct sctp_supported_chunk_types_param) + (sizeof(uint8_t) * SCTP_MAX_SUPPORTED_EXT))) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error chklist %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_SUPPORTED_ADDRTYPE: if (padded_size > SCTP_MAX_ADDR_PARAMS_SIZE) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error supaddrtype %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_RANDOM: if (padded_size > (sizeof(struct sctp_auth_random) + SCTP_RANDOM_MAX_SIZE)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error random %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_SET_PRIM_ADDR: case SCTP_DEL_IP_ADDRESS: case SCTP_ADD_IP_ADDRESS: if ((padded_size != sizeof(struct sctp_asconf_addrv4_param)) && (padded_size != sizeof(struct sctp_asconf_addr_param))) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error setprim %d\n", plen); goto invalid_size; } at += padded_size; break; /* Param's with a fixed size */ case SCTP_IPV4_ADDRESS: if (padded_size != sizeof(struct sctp_ipv4addr_param)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error ipv4 addr %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_IPV6_ADDRESS: if (padded_size != sizeof(struct sctp_ipv6addr_param)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error ipv6 addr %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_COOKIE_PRESERVE: if (padded_size != sizeof(struct sctp_cookie_perserve_param)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error cookie-preserve %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_ECN_NONCE_SUPPORTED: case SCTP_PRSCTP_SUPPORTED: if (padded_size != sizeof(struct sctp_paramhdr)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error ecnnonce/prsctp %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_ECN_CAPABLE: if (padded_size != sizeof(struct sctp_ecn_supported_param)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error ecn %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_ULP_ADAPTATION: if (padded_size != sizeof(struct sctp_adaptation_layer_indication)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error adapatation %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_SUCCESS_REPORT: if (padded_size != sizeof(struct sctp_asconf_paramhdr)) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Invalid size - error success %d\n", plen); goto invalid_size; } at += padded_size; break; case SCTP_HOSTNAME_ADDRESS: { /* We can NOT handle HOST NAME addresses!! */ int l_len; SCTPDBG(SCTP_DEBUG_OUTPUT1, "Can't handle hostname addresses.. abort processing\n"); *abort_processing = 1; if (op_err == NULL) { /* Ok need to try to get a mbuf */ #ifdef INET6 l_len = sizeof(struct ip6_hdr) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); #else l_len = sizeof(struct ip) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); #endif l_len += plen; l_len += sizeof(struct sctp_paramhdr); op_err = sctp_get_mbuf_for_msg(l_len, 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { SCTP_BUF_LEN(op_err) = 0; /* * pre-reserve space for ip * and sctp header and * chunk hdr */ #ifdef INET6 SCTP_BUF_RESV_UF(op_err, sizeof(struct ip6_hdr)); #else SCTP_BUF_RESV_UF(op_err, sizeof(struct ip)); #endif SCTP_BUF_RESV_UF(op_err, sizeof(struct sctphdr)); SCTP_BUF_RESV_UF(op_err, sizeof(struct sctp_chunkhdr)); } } if (op_err) { /* If we have space */ struct sctp_paramhdr s; if (err_at % 4) { uint32_t cpthis = 0; pad_needed = 4 - (err_at % 4); m_copyback(op_err, err_at, pad_needed, (caddr_t)&cpthis); err_at += pad_needed; } s.param_type = htons(SCTP_CAUSE_UNRESOLVABLE_ADDR); s.param_length = htons(sizeof(s) + plen); m_copyback(op_err, err_at, sizeof(s), (caddr_t)&s); err_at += sizeof(s); phdr = sctp_get_next_param(mat, at, (struct sctp_paramhdr *)tempbuf, min(sizeof(tempbuf), plen)); if (phdr == NULL) { sctp_m_freem(op_err); /* * we are out of memory but * we still need to have a * look at what to do (the * system is in trouble * though). */ return (NULL); } m_copyback(op_err, err_at, plen, (caddr_t)phdr); err_at += plen; } return (op_err); break; } default: /* * we do not recognize the parameter figure out what * we do. */ SCTPDBG(SCTP_DEBUG_OUTPUT1, "Hit default param %x\n", ptype); if ((ptype & 0x4000) == 0x4000) { /* Report bit is set?? */ SCTPDBG(SCTP_DEBUG_OUTPUT1, "report op err\n"); if (op_err == NULL) { int l_len; /* Ok need to try to get an mbuf */ #ifdef INET6 l_len = sizeof(struct ip6_hdr) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); #else l_len = sizeof(struct ip) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); #endif l_len += plen; l_len += sizeof(struct sctp_paramhdr); op_err = sctp_get_mbuf_for_msg(l_len, 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { SCTP_BUF_LEN(op_err) = 0; #ifdef INET6 SCTP_BUF_RESV_UF(op_err, sizeof(struct ip6_hdr)); #else SCTP_BUF_RESV_UF(op_err, sizeof(struct ip)); #endif SCTP_BUF_RESV_UF(op_err, sizeof(struct sctphdr)); SCTP_BUF_RESV_UF(op_err, sizeof(struct sctp_chunkhdr)); } } if (op_err) { /* If we have space */ struct sctp_paramhdr s; if (err_at % 4) { uint32_t cpthis = 0; pad_needed = 4 - (err_at % 4); m_copyback(op_err, err_at, pad_needed, (caddr_t)&cpthis); err_at += pad_needed; } s.param_type = htons(SCTP_UNRECOG_PARAM); s.param_length = htons(sizeof(s) + plen); m_copyback(op_err, err_at, sizeof(s), (caddr_t)&s); err_at += sizeof(s); if (plen > sizeof(tempbuf)) { plen = sizeof(tempbuf); } phdr = sctp_get_next_param(mat, at, (struct sctp_paramhdr *)tempbuf, min(sizeof(tempbuf), plen)); if (phdr == NULL) { sctp_m_freem(op_err); /* * we are out of memory but * we still need to have a * look at what to do (the * system is in trouble * though). */ op_err = NULL; goto more_processing; } m_copyback(op_err, err_at, plen, (caddr_t)phdr); err_at += plen; } } more_processing: if ((ptype & 0x8000) == 0x0000) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "stop proc\n"); return (op_err); } else { /* skip this chunk and continue processing */ SCTPDBG(SCTP_DEBUG_OUTPUT1, "move on\n"); at += SCTP_SIZE32(plen); } break; } phdr = sctp_get_next_param(mat, at, ¶ms, sizeof(params)); } return (op_err); invalid_size: SCTPDBG(SCTP_DEBUG_OUTPUT1, "abort flag set\n"); *abort_processing = 1; if ((op_err == NULL) && phdr) { int l_len; #ifdef INET6 l_len = sizeof(struct ip6_hdr) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); #else l_len = sizeof(struct ip) + sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); #endif l_len += (2 * sizeof(struct sctp_paramhdr)); op_err = sctp_get_mbuf_for_msg(l_len, 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { SCTP_BUF_LEN(op_err) = 0; #ifdef INET6 SCTP_BUF_RESV_UF(op_err, sizeof(struct ip6_hdr)); #else SCTP_BUF_RESV_UF(op_err, sizeof(struct ip)); #endif SCTP_BUF_RESV_UF(op_err, sizeof(struct sctphdr)); SCTP_BUF_RESV_UF(op_err, sizeof(struct sctp_chunkhdr)); } } if ((op_err) && phdr) { struct sctp_paramhdr s; if (err_at % 4) { uint32_t cpthis = 0; pad_needed = 4 - (err_at % 4); m_copyback(op_err, err_at, pad_needed, (caddr_t)&cpthis); err_at += pad_needed; } s.param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION); s.param_length = htons(sizeof(s) + sizeof(struct sctp_paramhdr)); m_copyback(op_err, err_at, sizeof(s), (caddr_t)&s); err_at += sizeof(s); /* Only copy back the p-hdr that caused the issue */ m_copyback(op_err, err_at, sizeof(struct sctp_paramhdr), (caddr_t)phdr); } return (op_err); } static int sctp_are_there_new_addresses(struct sctp_association *asoc, struct mbuf *in_initpkt, int iphlen, int offset) { /* * Given a INIT packet, look through the packet to verify that there * are NO new addresses. As we go through the parameters add reports * of any un-understood parameters that require an error. Also we * must return (1) to drop the packet if we see a un-understood * parameter that tells us to drop the chunk. */ struct sockaddr_in sin4, *sa4; #ifdef INET6 struct sockaddr_in6 sin6, *sa6; #endif struct sockaddr *sa_touse; struct sockaddr *sa; struct sctp_paramhdr *phdr, params; struct ip *iph; #ifdef INET6 struct ip6_hdr *ip6h; #endif struct mbuf *mat; uint16_t ptype, plen; int err_at; uint8_t fnd; struct sctp_nets *net; memset(&sin4, 0, sizeof(sin4)); #ifdef INET6 memset(&sin6, 0, sizeof(sin6)); #endif sin4.sin_family = AF_INET; sin4.sin_len = sizeof(sin4); #ifdef INET6 sin6.sin6_family = AF_INET6; sin6.sin6_len = sizeof(sin6); #endif sa_touse = NULL; /* First what about the src address of the pkt ? */ iph = mtod(in_initpkt, struct ip *); switch (iph->ip_v) { case IPVERSION: /* source addr is IPv4 */ sin4.sin_addr = iph->ip_src; sa_touse = (struct sockaddr *)&sin4; break; #ifdef INET6 case IPV6_VERSION >> 4: /* source addr is IPv6 */ ip6h = mtod(in_initpkt, struct ip6_hdr *); sin6.sin6_addr = ip6h->ip6_src; sa_touse = (struct sockaddr *)&sin6; break; #endif default: return (1); } fnd = 0; TAILQ_FOREACH(net, &asoc->nets, sctp_next) { sa = (struct sockaddr *)&net->ro._l_addr; if (sa->sa_family == sa_touse->sa_family) { if (sa->sa_family == AF_INET) { sa4 = (struct sockaddr_in *)sa; if (sa4->sin_addr.s_addr == sin4.sin_addr.s_addr) { fnd = 1; break; } } #ifdef INET6 if (sa->sa_family == AF_INET6) { sa6 = (struct sockaddr_in6 *)sa; if (SCTP6_ARE_ADDR_EQUAL(sa6, &sin6)) { fnd = 1; break; } } #endif } } if (fnd == 0) { /* New address added! no need to look futher. */ return (1); } /* Ok so far lets munge through the rest of the packet */ mat = in_initpkt; err_at = 0; sa_touse = NULL; offset += sizeof(struct sctp_init_chunk); phdr = sctp_get_next_param(mat, offset, ¶ms, sizeof(params)); while (phdr) { ptype = ntohs(phdr->param_type); plen = ntohs(phdr->param_length); if (ptype == SCTP_IPV4_ADDRESS) { struct sctp_ipv4addr_param *p4, p4_buf; phdr = sctp_get_next_param(mat, offset, (struct sctp_paramhdr *)&p4_buf, sizeof(p4_buf)); if (plen != sizeof(struct sctp_ipv4addr_param) || phdr == NULL) { return (1); } p4 = (struct sctp_ipv4addr_param *)phdr; sin4.sin_addr.s_addr = p4->addr; sa_touse = (struct sockaddr *)&sin4; } else if (ptype == SCTP_IPV6_ADDRESS) { struct sctp_ipv6addr_param *p6, p6_buf; phdr = sctp_get_next_param(mat, offset, (struct sctp_paramhdr *)&p6_buf, sizeof(p6_buf)); if (plen != sizeof(struct sctp_ipv6addr_param) || phdr == NULL) { return (1); } p6 = (struct sctp_ipv6addr_param *)phdr; #ifdef INET6 memcpy((caddr_t)&sin6.sin6_addr, p6->addr, sizeof(p6->addr)); #endif sa_touse = (struct sockaddr *)&sin4; } if (sa_touse) { /* ok, sa_touse points to one to check */ fnd = 0; TAILQ_FOREACH(net, &asoc->nets, sctp_next) { sa = (struct sockaddr *)&net->ro._l_addr; if (sa->sa_family != sa_touse->sa_family) { continue; } if (sa->sa_family == AF_INET) { sa4 = (struct sockaddr_in *)sa; if (sa4->sin_addr.s_addr == sin4.sin_addr.s_addr) { fnd = 1; break; } } #ifdef INET6 if (sa->sa_family == AF_INET6) { sa6 = (struct sockaddr_in6 *)sa; if (SCTP6_ARE_ADDR_EQUAL( sa6, &sin6)) { fnd = 1; break; } } #endif } if (!fnd) { /* New addr added! no need to look further */ return (1); } } offset += SCTP_SIZE32(plen); phdr = sctp_get_next_param(mat, offset, ¶ms, sizeof(params)); } return (0); } /* * Given a MBUF chain that was sent into us containing an INIT. Build a * INIT-ACK with COOKIE and send back. We assume that the in_initpkt has done * a pullup to include IPv6/4header, SCTP header and initial part of INIT * message (i.e. the struct sctp_init_msg). */ void sctp_send_initiate_ack(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct mbuf *init_pkt, int iphlen, int offset, struct sctphdr *sh, struct sctp_init_chunk *init_chk, uint32_t vrf_id, uint16_t port, int hold_inp_lock) { struct sctp_association *asoc; struct mbuf *m, *m_at, *m_tmp, *m_cookie, *op_err, *mp_last; struct sctp_init_msg *initackm_out; struct sctp_ecn_supported_param *ecn; struct sctp_prsctp_supported_param *prsctp; struct sctp_ecn_nonce_supported_param *ecn_nonce; struct sctp_supported_chunk_types_param *pr_supported; union sctp_sockstore store, store1, *over_addr; struct sockaddr_in *sin, *to_sin; #ifdef INET6 struct sockaddr_in6 *sin6, *to_sin6; #endif struct ip *iph; #ifdef INET6 struct ip6_hdr *ip6; #endif struct sockaddr *to; struct sctp_state_cookie stc; struct sctp_nets *net = NULL; uint8_t *signature = NULL; int cnt_inits_to = 0; uint16_t his_limit, i_want; int abort_flag, padval; int num_ext; int p_len; struct socket *so; if (stcb) asoc = &stcb->asoc; else asoc = NULL; mp_last = NULL; if ((asoc != NULL) && (SCTP_GET_STATE(asoc) != SCTP_STATE_COOKIE_WAIT) && (sctp_are_there_new_addresses(asoc, init_pkt, iphlen, offset))) { /* new addresses, out of here in non-cookie-wait states */ /* * Send a ABORT, we don't add the new address error clause * though we even set the T bit and copy in the 0 tag.. this * looks no different than if no listener was present. */ sctp_send_abort(init_pkt, iphlen, sh, 0, NULL, vrf_id, port); return; } abort_flag = 0; op_err = sctp_arethere_unrecognized_parameters(init_pkt, (offset + sizeof(struct sctp_init_chunk)), &abort_flag, (struct sctp_chunkhdr *)init_chk); if (abort_flag) { do_a_abort: sctp_send_abort(init_pkt, iphlen, sh, init_chk->init.initiate_tag, op_err, vrf_id, port); return; } m = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_DATA); if (m == NULL) { /* No memory, INIT timer will re-attempt. */ if (op_err) sctp_m_freem(op_err); return; } SCTP_BUF_LEN(m) = sizeof(struct sctp_init_msg); /* the time I built cookie */ (void)SCTP_GETTIME_TIMEVAL(&stc.time_entered); /* populate any tie tags */ if (asoc != NULL) { /* unlock before tag selections */ stc.tie_tag_my_vtag = asoc->my_vtag_nonce; stc.tie_tag_peer_vtag = asoc->peer_vtag_nonce; stc.cookie_life = asoc->cookie_life; net = asoc->primary_destination; } else { stc.tie_tag_my_vtag = 0; stc.tie_tag_peer_vtag = 0; /* life I will award this cookie */ stc.cookie_life = inp->sctp_ep.def_cookie_life; } /* copy in the ports for later check */ stc.myport = sh->dest_port; stc.peerport = sh->src_port; /* * If we wanted to honor cookie life extentions, we would add to * stc.cookie_life. For now we should NOT honor any extension */ stc.site_scope = stc.local_scope = stc.loopback_scope = 0; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { struct inpcb *in_inp; /* Its a V6 socket */ in_inp = (struct inpcb *)inp; stc.ipv6_addr_legal = 1; /* Now look at the binding flag to see if V4 will be legal */ if (SCTP_IPV6_V6ONLY(in_inp) == 0) { stc.ipv4_addr_legal = 1; } else { /* V4 addresses are NOT legal on the association */ stc.ipv4_addr_legal = 0; } } else { /* Its a V4 socket, no - V6 */ stc.ipv4_addr_legal = 1; stc.ipv6_addr_legal = 0; } #ifdef SCTP_DONT_DO_PRIVADDR_SCOPE stc.ipv4_scope = 1; #else stc.ipv4_scope = 0; #endif /* now for scope setup */ memset((caddr_t)&store, 0, sizeof(store)); memset((caddr_t)&store1, 0, sizeof(store1)); sin = &store.sin; to_sin = &store1.sin; #ifdef INET6 sin6 = &store.sin6; to_sin6 = &store1.sin6; #endif iph = mtod(init_pkt, struct ip *); /* establish the to_addr's */ switch (iph->ip_v) { case IPVERSION: to_sin->sin_port = sh->dest_port; to_sin->sin_family = AF_INET; to_sin->sin_len = sizeof(struct sockaddr_in); to_sin->sin_addr = iph->ip_dst; break; #ifdef INET6 case IPV6_VERSION >> 4: ip6 = mtod(init_pkt, struct ip6_hdr *); to_sin6->sin6_addr = ip6->ip6_dst; to_sin6->sin6_scope_id = 0; to_sin6->sin6_port = sh->dest_port; to_sin6->sin6_family = AF_INET6; to_sin6->sin6_len = sizeof(struct sockaddr_in6); break; #endif default: goto do_a_abort; break; }; if (net == NULL) { to = (struct sockaddr *)&store; switch (iph->ip_v) { case IPVERSION: { sin->sin_family = AF_INET; sin->sin_len = sizeof(struct sockaddr_in); sin->sin_port = sh->src_port; sin->sin_addr = iph->ip_src; /* lookup address */ stc.address[0] = sin->sin_addr.s_addr; stc.address[1] = 0; stc.address[2] = 0; stc.address[3] = 0; stc.addr_type = SCTP_IPV4_ADDRESS; /* local from address */ stc.laddress[0] = to_sin->sin_addr.s_addr; stc.laddress[1] = 0; stc.laddress[2] = 0; stc.laddress[3] = 0; stc.laddr_type = SCTP_IPV4_ADDRESS; /* scope_id is only for v6 */ stc.scope_id = 0; #ifndef SCTP_DONT_DO_PRIVADDR_SCOPE if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) { stc.ipv4_scope = 1; } #else stc.ipv4_scope = 1; #endif /* SCTP_DONT_DO_PRIVADDR_SCOPE */ /* Must use the address in this case */ if (sctp_is_address_on_local_host((struct sockaddr *)sin, vrf_id)) { stc.loopback_scope = 1; stc.ipv4_scope = 1; stc.site_scope = 1; stc.local_scope = 0; } break; } #ifdef INET6 case IPV6_VERSION >> 4: { ip6 = mtod(init_pkt, struct ip6_hdr *); sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(struct sockaddr_in6); sin6->sin6_port = sh->src_port; sin6->sin6_addr = ip6->ip6_src; /* lookup address */ memcpy(&stc.address, &sin6->sin6_addr, sizeof(struct in6_addr)); sin6->sin6_scope_id = 0; stc.addr_type = SCTP_IPV6_ADDRESS; stc.scope_id = 0; if (sctp_is_address_on_local_host((struct sockaddr *)sin6, vrf_id)) { /* * FIX ME: does this have scope from * rcvif? */ (void)sa6_recoverscope(sin6); stc.scope_id = sin6->sin6_scope_id; sa6_embedscope(sin6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone)); stc.loopback_scope = 1; stc.local_scope = 0; stc.site_scope = 1; stc.ipv4_scope = 1; } else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { /* * If the new destination is a * LINK_LOCAL we must have common * both site and local scope. Don't * set local scope though since we * must depend on the source to be * added implicitly. We cannot * assure just because we share one * link that all links are common. */ stc.local_scope = 0; stc.site_scope = 1; stc.ipv4_scope = 1; /* * we start counting for the private * address stuff at 1. since the * link local we source from won't * show up in our scoped count. */ cnt_inits_to = 1; /* * pull out the scope_id from * incoming pkt */ /* * FIX ME: does this have scope from * rcvif? */ (void)sa6_recoverscope(sin6); stc.scope_id = sin6->sin6_scope_id; sa6_embedscope(sin6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone)); } else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) { /* * If the new destination is * SITE_LOCAL then we must have site * scope in common. */ stc.site_scope = 1; } memcpy(&stc.laddress, &to_sin6->sin6_addr, sizeof(struct in6_addr)); stc.laddr_type = SCTP_IPV6_ADDRESS; break; } #endif default: /* TSNH */ goto do_a_abort; break; } } else { /* set the scope per the existing tcb */ #ifdef INET6 struct sctp_nets *lnet; #endif stc.loopback_scope = asoc->loopback_scope; stc.ipv4_scope = asoc->ipv4_local_scope; stc.site_scope = asoc->site_scope; stc.local_scope = asoc->local_scope; #ifdef INET6 /* Why do we not consider IPv4 LL addresses? */ TAILQ_FOREACH(lnet, &asoc->nets, sctp_next) { if (lnet->ro._l_addr.sin6.sin6_family == AF_INET6) { if (IN6_IS_ADDR_LINKLOCAL(&lnet->ro._l_addr.sin6.sin6_addr)) { /* * if we have a LL address, start * counting at 1. */ cnt_inits_to = 1; } } } #endif /* use the net pointer */ to = (struct sockaddr *)&net->ro._l_addr; switch (to->sa_family) { case AF_INET: sin = (struct sockaddr_in *)to; stc.address[0] = sin->sin_addr.s_addr; stc.address[1] = 0; stc.address[2] = 0; stc.address[3] = 0; stc.addr_type = SCTP_IPV4_ADDRESS; if (net->src_addr_selected == 0) { /* * strange case here, the INIT should have * did the selection. */ net->ro._s_addr = sctp_source_address_selection(inp, stcb, (sctp_route_t *) & net->ro, net, 0, vrf_id); if (net->ro._s_addr == NULL) return; net->src_addr_selected = 1; } stc.laddress[0] = net->ro._s_addr->address.sin.sin_addr.s_addr; stc.laddress[1] = 0; stc.laddress[2] = 0; stc.laddress[3] = 0; stc.laddr_type = SCTP_IPV4_ADDRESS; break; #ifdef INET6 case AF_INET6: sin6 = (struct sockaddr_in6 *)to; memcpy(&stc.address, &sin6->sin6_addr, sizeof(struct in6_addr)); stc.addr_type = SCTP_IPV6_ADDRESS; if (net->src_addr_selected == 0) { /* * strange case here, the INIT should have * did the selection. */ net->ro._s_addr = sctp_source_address_selection(inp, stcb, (sctp_route_t *) & net->ro, net, 0, vrf_id); if (net->ro._s_addr == NULL) return; net->src_addr_selected = 1; } memcpy(&stc.laddress, &net->ro._s_addr->address.sin6.sin6_addr, sizeof(struct in6_addr)); stc.laddr_type = SCTP_IPV6_ADDRESS; break; #endif } } /* Now lets put the SCTP header in place */ initackm_out = mtod(m, struct sctp_init_msg *); initackm_out->sh.src_port = inp->sctp_lport; initackm_out->sh.dest_port = sh->src_port; initackm_out->sh.v_tag = init_chk->init.initiate_tag; /* Save it off for quick ref */ stc.peers_vtag = init_chk->init.initiate_tag; initackm_out->sh.checksum = 0; /* calculate later */ /* who are we */ memcpy(stc.identification, SCTP_VERSION_STRING, min(strlen(SCTP_VERSION_STRING), sizeof(stc.identification))); /* now the chunk header */ initackm_out->msg.ch.chunk_type = SCTP_INITIATION_ACK; initackm_out->msg.ch.chunk_flags = 0; /* fill in later from mbuf we build */ initackm_out->msg.ch.chunk_length = 0; /* place in my tag */ if ((asoc != NULL) && ((SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_WAIT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_INUSE) || (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED))) { /* re-use the v-tags and init-seq here */ initackm_out->msg.init.initiate_tag = htonl(asoc->my_vtag); initackm_out->msg.init.initial_tsn = htonl(asoc->init_seq_number); } else { uint32_t vtag, itsn; if (hold_inp_lock) { SCTP_INP_INCR_REF(inp); SCTP_INP_RUNLOCK(inp); } if (asoc) { atomic_add_int(&asoc->refcnt, 1); SCTP_TCB_UNLOCK(stcb); vtag = sctp_select_a_tag(inp, 1); initackm_out->msg.init.initiate_tag = htonl(vtag); /* get a TSN to use too */ itsn = sctp_select_initial_TSN(&inp->sctp_ep); initackm_out->msg.init.initial_tsn = htonl(itsn); SCTP_TCB_LOCK(stcb); atomic_add_int(&asoc->refcnt, -1); } else { vtag = sctp_select_a_tag(inp, 1); initackm_out->msg.init.initiate_tag = htonl(vtag); /* get a TSN to use too */ initackm_out->msg.init.initial_tsn = htonl(sctp_select_initial_TSN(&inp->sctp_ep)); } if (hold_inp_lock) { SCTP_INP_RLOCK(inp); SCTP_INP_DECR_REF(inp); } } /* save away my tag to */ stc.my_vtag = initackm_out->msg.init.initiate_tag; /* set up some of the credits. */ so = inp->sctp_socket; if (so == NULL) { /* memory problem */ sctp_m_freem(m); return; } else { initackm_out->msg.init.a_rwnd = htonl(max(SCTP_SB_LIMIT_RCV(so), SCTP_MINIMAL_RWND)); } /* set what I want */ his_limit = ntohs(init_chk->init.num_inbound_streams); /* choose what I want */ if (asoc != NULL) { if (asoc->streamoutcnt > inp->sctp_ep.pre_open_stream_count) { i_want = asoc->streamoutcnt; } else { i_want = inp->sctp_ep.pre_open_stream_count; } } else { i_want = inp->sctp_ep.pre_open_stream_count; } if (his_limit < i_want) { /* I Want more :< */ initackm_out->msg.init.num_outbound_streams = init_chk->init.num_inbound_streams; } else { /* I can have what I want :> */ initackm_out->msg.init.num_outbound_streams = htons(i_want); } /* tell him his limt. */ initackm_out->msg.init.num_inbound_streams = htons(inp->sctp_ep.max_open_streams_intome); /* setup the ECN pointer */ if (inp->sctp_ep.adaptation_layer_indicator) { struct sctp_adaptation_layer_indication *ali; ali = (struct sctp_adaptation_layer_indication *)( (caddr_t)initackm_out + sizeof(*initackm_out)); ali->ph.param_type = htons(SCTP_ULP_ADAPTATION); ali->ph.param_length = htons(sizeof(*ali)); ali->indication = ntohl(inp->sctp_ep.adaptation_layer_indicator); SCTP_BUF_LEN(m) += sizeof(*ali); ecn = (struct sctp_ecn_supported_param *)((caddr_t)ali + sizeof(*ali)); } else { ecn = (struct sctp_ecn_supported_param *)( (caddr_t)initackm_out + sizeof(*initackm_out)); } /* ECN parameter */ if (SCTP_BASE_SYSCTL(sctp_ecn_enable) == 1) { ecn->ph.param_type = htons(SCTP_ECN_CAPABLE); ecn->ph.param_length = htons(sizeof(*ecn)); SCTP_BUF_LEN(m) += sizeof(*ecn); prsctp = (struct sctp_prsctp_supported_param *)((caddr_t)ecn + sizeof(*ecn)); } else { prsctp = (struct sctp_prsctp_supported_param *)((caddr_t)ecn); } /* And now tell the peer we do pr-sctp */ prsctp->ph.param_type = htons(SCTP_PRSCTP_SUPPORTED); prsctp->ph.param_length = htons(sizeof(*prsctp)); SCTP_BUF_LEN(m) += sizeof(*prsctp); /* And now tell the peer we do all the extensions */ pr_supported = (struct sctp_supported_chunk_types_param *) ((caddr_t)prsctp + sizeof(*prsctp)); pr_supported->ph.param_type = htons(SCTP_SUPPORTED_CHUNK_EXT); num_ext = 0; pr_supported->chunk_types[num_ext++] = SCTP_ASCONF; pr_supported->chunk_types[num_ext++] = SCTP_ASCONF_ACK; pr_supported->chunk_types[num_ext++] = SCTP_FORWARD_CUM_TSN; pr_supported->chunk_types[num_ext++] = SCTP_PACKET_DROPPED; pr_supported->chunk_types[num_ext++] = SCTP_STREAM_RESET; if (!SCTP_BASE_SYSCTL(sctp_auth_disable)) pr_supported->chunk_types[num_ext++] = SCTP_AUTHENTICATION; p_len = sizeof(*pr_supported) + num_ext; pr_supported->ph.param_length = htons(p_len); bzero((caddr_t)pr_supported + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); /* ECN nonce: And now tell the peer we support ECN nonce */ if (SCTP_BASE_SYSCTL(sctp_ecn_nonce)) { ecn_nonce = (struct sctp_ecn_nonce_supported_param *) ((caddr_t)pr_supported + SCTP_SIZE32(p_len)); ecn_nonce->ph.param_type = htons(SCTP_ECN_NONCE_SUPPORTED); ecn_nonce->ph.param_length = htons(sizeof(*ecn_nonce)); SCTP_BUF_LEN(m) += sizeof(*ecn_nonce); } /* add authentication parameters */ if (!SCTP_BASE_SYSCTL(sctp_auth_disable)) { struct sctp_auth_random *randp; struct sctp_auth_hmac_algo *hmacs; struct sctp_auth_chunk_list *chunks; uint16_t random_len; /* generate and add RANDOM parameter */ random_len = SCTP_AUTH_RANDOM_SIZE_DEFAULT; randp = (struct sctp_auth_random *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); randp->ph.param_type = htons(SCTP_RANDOM); p_len = sizeof(*randp) + random_len; randp->ph.param_length = htons(p_len); SCTP_READ_RANDOM(randp->random_data, random_len); /* zero out any padding required */ bzero((caddr_t)randp + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); /* add HMAC_ALGO parameter */ hmacs = (struct sctp_auth_hmac_algo *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); p_len = sctp_serialize_hmaclist(inp->sctp_ep.local_hmacs, (uint8_t *) hmacs->hmac_ids); if (p_len > 0) { p_len += sizeof(*hmacs); hmacs->ph.param_type = htons(SCTP_HMAC_LIST); hmacs->ph.param_length = htons(p_len); /* zero out any padding required */ bzero((caddr_t)hmacs + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); } /* add CHUNKS parameter */ chunks = (struct sctp_auth_chunk_list *)(mtod(m, caddr_t)+SCTP_BUF_LEN(m)); p_len = sctp_serialize_auth_chunks(inp->sctp_ep.local_auth_chunks, chunks->chunk_types); if (p_len > 0) { p_len += sizeof(*chunks); chunks->ph.param_type = htons(SCTP_CHUNK_LIST); chunks->ph.param_length = htons(p_len); /* zero out any padding required */ bzero((caddr_t)chunks + p_len, SCTP_SIZE32(p_len) - p_len); SCTP_BUF_LEN(m) += SCTP_SIZE32(p_len); } } m_at = m; /* now the addresses */ { struct sctp_scoping scp; /* * To optimize this we could put the scoping stuff into a * structure and remove the individual uint8's from the stc * structure. Then we could just sifa in the address within * the stc.. but for now this is a quick hack to get the * address stuff teased apart. */ scp.ipv4_addr_legal = stc.ipv4_addr_legal; scp.ipv6_addr_legal = stc.ipv6_addr_legal; scp.loopback_scope = stc.loopback_scope; scp.ipv4_local_scope = stc.ipv4_scope; scp.local_scope = stc.local_scope; scp.site_scope = stc.site_scope; m_at = sctp_add_addresses_to_i_ia(inp, &scp, m_at, cnt_inits_to); } /* tack on the operational error if present */ if (op_err) { struct mbuf *ol; int llen; llen = 0; ol = op_err; while (ol) { llen += SCTP_BUF_LEN(ol); ol = SCTP_BUF_NEXT(ol); } if (llen % 4) { /* must add a pad to the param */ uint32_t cpthis = 0; int padlen; padlen = 4 - (llen % 4); m_copyback(op_err, llen, padlen, (caddr_t)&cpthis); } while (SCTP_BUF_NEXT(m_at) != NULL) { m_at = SCTP_BUF_NEXT(m_at); } SCTP_BUF_NEXT(m_at) = op_err; while (SCTP_BUF_NEXT(m_at) != NULL) { m_at = SCTP_BUF_NEXT(m_at); } } /* pre-calulate the size and update pkt header and chunk header */ p_len = 0; for (m_tmp = m; m_tmp; m_tmp = SCTP_BUF_NEXT(m_tmp)) { p_len += SCTP_BUF_LEN(m_tmp); if (SCTP_BUF_NEXT(m_tmp) == NULL) { /* m_tmp should now point to last one */ break; } } /* Now we must build a cookie */ m_cookie = sctp_add_cookie(inp, init_pkt, offset, m, sizeof(struct sctphdr), &stc, &signature); if (m_cookie == NULL) { /* memory problem */ sctp_m_freem(m); return; } /* Now append the cookie to the end and update the space/size */ SCTP_BUF_NEXT(m_tmp) = m_cookie; for (m_tmp = m_cookie; m_tmp; m_tmp = SCTP_BUF_NEXT(m_tmp)) { p_len += SCTP_BUF_LEN(m_tmp); if (SCTP_BUF_NEXT(m_tmp) == NULL) { /* m_tmp should now point to last one */ mp_last = m_tmp; break; } } /* * Place in the size, but we don't include the last pad (if any) in * the INIT-ACK. */ initackm_out->msg.ch.chunk_length = htons((p_len - sizeof(struct sctphdr))); /* * Time to sign the cookie, we don't sign over the cookie signature * though thus we set trailer. */ (void)sctp_hmac_m(SCTP_HMAC, (uint8_t *) inp->sctp_ep.secret_key[(int)(inp->sctp_ep.current_secret_number)], SCTP_SECRET_SIZE, m_cookie, sizeof(struct sctp_paramhdr), (uint8_t *) signature, SCTP_SIGNATURE_SIZE); /* * We sifa 0 here to NOT set IP_DF if its IPv4, we ignore the return * here since the timer will drive a retranmission. */ padval = p_len % 4; if ((padval) && (mp_last)) { /* see my previous comments on mp_last */ int ret; ret = sctp_add_pad_tombuf(mp_last, (4 - padval)); if (ret) { /* Houston we have a problem, no space */ sctp_m_freem(m); return; } p_len += padval; } if (stc.loopback_scope) { over_addr = &store1; } else { over_addr = NULL; } (void)sctp_lowlevel_chunk_output(inp, NULL, NULL, to, m, 0, NULL, 0, 0, NULL, 0, port, SCTP_SO_NOT_LOCKED, over_addr); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } void sctp_insert_on_wheel(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_stream_out *strq, int holds_lock) { struct sctp_stream_out *stre, *strn; if (holds_lock == 0) { SCTP_TCB_SEND_LOCK(stcb); } if ((strq->next_spoke.tqe_next) || (strq->next_spoke.tqe_prev)) { /* already on wheel */ goto outof_here; } stre = TAILQ_FIRST(&asoc->out_wheel); if (stre == NULL) { /* only one on wheel */ TAILQ_INSERT_HEAD(&asoc->out_wheel, strq, next_spoke); goto outof_here; } for (; stre; stre = strn) { strn = TAILQ_NEXT(stre, next_spoke); if (stre->stream_no > strq->stream_no) { TAILQ_INSERT_BEFORE(stre, strq, next_spoke); goto outof_here; } else if (stre->stream_no == strq->stream_no) { /* huh, should not happen */ goto outof_here; } else if (strn == NULL) { /* next one is null */ TAILQ_INSERT_AFTER(&asoc->out_wheel, stre, strq, next_spoke); } } outof_here: if (holds_lock == 0) { SCTP_TCB_SEND_UNLOCK(stcb); } } static void sctp_remove_from_wheel(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_stream_out *strq) { /* take off and then setup so we know it is not on the wheel */ SCTP_TCB_SEND_LOCK(stcb); if (TAILQ_FIRST(&strq->outqueue)) { /* more was added */ SCTP_TCB_SEND_UNLOCK(stcb); return; } TAILQ_REMOVE(&asoc->out_wheel, strq, next_spoke); strq->next_spoke.tqe_next = NULL; strq->next_spoke.tqe_prev = NULL; SCTP_TCB_SEND_UNLOCK(stcb); } static void sctp_prune_prsctp(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_sndrcvinfo *srcv, int dataout) { int freed_spc = 0; struct sctp_tmit_chunk *chk, *nchk; SCTP_TCB_LOCK_ASSERT(stcb); if ((asoc->peer_supports_prsctp) && (asoc->sent_queue_cnt_removeable > 0)) { TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { /* * Look for chunks marked with the PR_SCTP flag AND * the buffer space flag. If the one being sent is * equal or greater priority then purge the old one * and free some space. */ if (PR_SCTP_BUF_ENABLED(chk->flags)) { /* * This one is PR-SCTP AND buffer space * limited type */ if (chk->rec.data.timetodrop.tv_sec >= (long)srcv->sinfo_timetolive) { /* * Lower numbers equates to higher * priority so if the one we are * looking at has a larger or equal * priority we want to drop the data * and NOT retransmit it. */ if (chk->data) { /* * We release the book_size * if the mbuf is here */ int ret_spc; int cause; if (chk->sent > SCTP_DATAGRAM_UNSENT) cause = SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT; else cause = SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_UNSENT; ret_spc = sctp_release_pr_sctp_chunk(stcb, chk, cause, &asoc->sent_queue, SCTP_SO_LOCKED); freed_spc += ret_spc; if (freed_spc >= dataout) { return; } } /* if chunk was present */ } /* if of sufficent priority */ } /* if chunk has enabled */ } /* tailqforeach */ chk = TAILQ_FIRST(&asoc->send_queue); while (chk) { nchk = TAILQ_NEXT(chk, sctp_next); /* Here we must move to the sent queue and mark */ if (PR_SCTP_TTL_ENABLED(chk->flags)) { if (chk->rec.data.timetodrop.tv_sec >= (long)srcv->sinfo_timetolive) { if (chk->data) { /* * We release the book_size * if the mbuf is here */ int ret_spc; ret_spc = sctp_release_pr_sctp_chunk(stcb, chk, SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_UNSENT, &asoc->send_queue, SCTP_SO_LOCKED); freed_spc += ret_spc; if (freed_spc >= dataout) { return; } } /* end if chk->data */ } /* end if right class */ } /* end if chk pr-sctp */ chk = nchk; } /* end while (chk) */ } /* if enabled in asoc */ } int sctp_get_frag_point(struct sctp_tcb *stcb, struct sctp_association *asoc) { int siz, ovh; /* * For endpoints that have both v6 and v4 addresses we must reserve * room for the ipv6 header, for those that are only dealing with V4 * we use a larger frag point. */ if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ovh = SCTP_MED_OVERHEAD; } else { ovh = SCTP_MED_V4_OVERHEAD; } if (stcb->asoc.sctp_frag_point > asoc->smallest_mtu) siz = asoc->smallest_mtu - ovh; else siz = (stcb->asoc.sctp_frag_point - ovh); /* * if (siz > (MCLBYTES-sizeof(struct sctp_data_chunk))) { */ /* A data chunk MUST fit in a cluster */ /* siz = (MCLBYTES - sizeof(struct sctp_data_chunk)); */ /* } */ /* adjust for an AUTH chunk if DATA requires auth */ if (sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks)) siz -= sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); if (siz % 4) { /* make it an even word boundary please */ siz -= (siz % 4); } return (siz); } static void sctp_set_prsctp_policy(struct sctp_tcb *stcb, struct sctp_stream_queue_pending *sp) { sp->pr_sctp_on = 0; if (stcb->asoc.peer_supports_prsctp) { /* * We assume that the user wants PR_SCTP_TTL if the user * provides a positive lifetime but does not specify any * PR_SCTP policy. This is a BAD assumption and causes * problems at least with the U-Vancovers MPI folks. I will * change this to be no policy means NO PR-SCTP. */ if (PR_SCTP_ENABLED(sp->sinfo_flags)) { sp->act_flags |= PR_SCTP_POLICY(sp->sinfo_flags); sp->pr_sctp_on = 1; } else { return; } switch (PR_SCTP_POLICY(sp->sinfo_flags)) { case CHUNK_FLAGS_PR_SCTP_BUF: /* * Time to live is a priority stored in tv_sec when * doing the buffer drop thing. */ sp->ts.tv_sec = sp->timetolive; sp->ts.tv_usec = 0; break; case CHUNK_FLAGS_PR_SCTP_TTL: { struct timeval tv; (void)SCTP_GETTIME_TIMEVAL(&sp->ts); tv.tv_sec = sp->timetolive / 1000; tv.tv_usec = (sp->timetolive * 1000) % 1000000; /* * TODO sctp_constants.h needs alternative * time macros when _KERNEL is undefined. */ timevaladd(&sp->ts, &tv); } break; case CHUNK_FLAGS_PR_SCTP_RTX: /* * Time to live is a the number or retransmissions * stored in tv_sec. */ sp->ts.tv_sec = sp->timetolive; sp->ts.tv_usec = 0; break; default: SCTPDBG(SCTP_DEBUG_USRREQ1, "Unknown PR_SCTP policy %u.\n", PR_SCTP_POLICY(sp->sinfo_flags)); break; } } } static int sctp_msg_append(struct sctp_tcb *stcb, struct sctp_nets *net, struct mbuf *m, struct sctp_sndrcvinfo *srcv, int hold_stcb_lock) { int error = 0, holds_lock; struct mbuf *at; struct sctp_stream_queue_pending *sp = NULL; struct sctp_stream_out *strm; /* * Given an mbuf chain, put it into the association send queue and * place it on the wheel */ holds_lock = hold_stcb_lock; if (srcv->sinfo_stream >= stcb->asoc.streamoutcnt) { /* Invalid stream number */ SCTP_LTRACE_ERR_RET_PKT(m, NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_now; } if ((stcb->asoc.stream_locked) && (stcb->asoc.stream_locked_on != srcv->sinfo_stream)) { SCTP_LTRACE_ERR_RET_PKT(m, NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_now; } strm = &stcb->asoc.strmout[srcv->sinfo_stream]; /* Now can we send this? */ if ((SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_SENT) || (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_ACK_SENT) || (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED) || (stcb->asoc.state & SCTP_STATE_SHUTDOWN_PENDING)) { /* got data while shutting down */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ECONNRESET); error = ECONNRESET; goto out_now; } sctp_alloc_a_strmoq(stcb, sp); if (sp == NULL) { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); error = ENOMEM; goto out_now; } sp->sinfo_flags = srcv->sinfo_flags; sp->timetolive = srcv->sinfo_timetolive; sp->ppid = srcv->sinfo_ppid; sp->context = srcv->sinfo_context; sp->strseq = 0; if (sp->sinfo_flags & SCTP_ADDR_OVER) { sp->net = net; sp->addr_over = 1; } else { sp->net = stcb->asoc.primary_destination; sp->addr_over = 0; } atomic_add_int(&sp->net->ref_count, 1); (void)SCTP_GETTIME_TIMEVAL(&sp->ts); sp->stream = srcv->sinfo_stream; sp->msg_is_complete = 1; sp->sender_all_done = 1; sp->some_taken = 0; sp->data = m; sp->tail_mbuf = NULL; sp->length = 0; at = m; sctp_set_prsctp_policy(stcb, sp); /* * We could in theory (for sendall) sifa the length in, but we would * still have to hunt through the chain since we need to setup the * tail_mbuf */ while (at) { if (SCTP_BUF_NEXT(at) == NULL) sp->tail_mbuf = at; sp->length += SCTP_BUF_LEN(at); at = SCTP_BUF_NEXT(at); } SCTP_TCB_SEND_LOCK(stcb); sctp_snd_sb_alloc(stcb, sp->length); atomic_add_int(&stcb->asoc.stream_queue_cnt, 1); TAILQ_INSERT_TAIL(&strm->outqueue, sp, next); if ((srcv->sinfo_flags & SCTP_UNORDERED) == 0) { sp->strseq = strm->next_sequence_sent; strm->next_sequence_sent++; } if ((strm->next_spoke.tqe_next == NULL) && (strm->next_spoke.tqe_prev == NULL)) { /* Not on wheel, insert */ sctp_insert_on_wheel(stcb, &stcb->asoc, strm, 1); } m = NULL; SCTP_TCB_SEND_UNLOCK(stcb); out_now: if (m) { sctp_m_freem(m); } return (error); } static struct mbuf * sctp_copy_mbufchain(struct mbuf *clonechain, struct mbuf *outchain, struct mbuf **endofchain, int can_take_mbuf, int sizeofcpy, uint8_t copy_by_ref) { struct mbuf *m; struct mbuf *appendchain; caddr_t cp; int len; if (endofchain == NULL) { /* error */ error_out: if (outchain) sctp_m_freem(outchain); return (NULL); } if (can_take_mbuf) { appendchain = clonechain; } else { if (!copy_by_ref && (sizeofcpy <= (int)((((SCTP_BASE_SYSCTL(sctp_mbuf_threshold_count) - 1) * MLEN) + MHLEN))) ) { /* Its not in a cluster */ if (*endofchain == NULL) { /* lets get a mbuf cluster */ if (outchain == NULL) { /* This is the general case */ new_mbuf: outchain = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_HEADER); if (outchain == NULL) { goto error_out; } SCTP_BUF_LEN(outchain) = 0; *endofchain = outchain; /* get the prepend space */ SCTP_BUF_RESV_UF(outchain, (SCTP_FIRST_MBUF_RESV + 4)); } else { /* * We really should not get a NULL * in endofchain */ /* find end */ m = outchain; while (m) { if (SCTP_BUF_NEXT(m) == NULL) { *endofchain = m; break; } m = SCTP_BUF_NEXT(m); } /* sanity */ if (*endofchain == NULL) { /* * huh, TSNH XXX maybe we * should panic */ sctp_m_freem(outchain); goto new_mbuf; } } /* get the new end of length */ len = M_TRAILINGSPACE(*endofchain); } else { /* how much is left at the end? */ len = M_TRAILINGSPACE(*endofchain); } /* Find the end of the data, for appending */ cp = (mtod((*endofchain), caddr_t)+SCTP_BUF_LEN((*endofchain))); /* Now lets copy it out */ if (len >= sizeofcpy) { /* It all fits, copy it in */ m_copydata(clonechain, 0, sizeofcpy, cp); SCTP_BUF_LEN((*endofchain)) += sizeofcpy; } else { /* fill up the end of the chain */ if (len > 0) { m_copydata(clonechain, 0, len, cp); SCTP_BUF_LEN((*endofchain)) += len; /* now we need another one */ sizeofcpy -= len; } m = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_HEADER); if (m == NULL) { /* We failed */ goto error_out; } SCTP_BUF_NEXT((*endofchain)) = m; *endofchain = m; cp = mtod((*endofchain), caddr_t); m_copydata(clonechain, len, sizeofcpy, cp); SCTP_BUF_LEN((*endofchain)) += sizeofcpy; } return (outchain); } else { /* copy the old fashion way */ appendchain = SCTP_M_COPYM(clonechain, 0, M_COPYALL, M_DONTWAIT); #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = appendchain; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif } } if (appendchain == NULL) { /* error */ if (outchain) sctp_m_freem(outchain); return (NULL); } if (outchain) { /* tack on to the end */ if (*endofchain != NULL) { SCTP_BUF_NEXT(((*endofchain))) = appendchain; } else { m = outchain; while (m) { if (SCTP_BUF_NEXT(m) == NULL) { SCTP_BUF_NEXT(m) = appendchain; break; } m = SCTP_BUF_NEXT(m); } } /* * save off the end and update the end-chain postion */ m = appendchain; while (m) { if (SCTP_BUF_NEXT(m) == NULL) { *endofchain = m; break; } m = SCTP_BUF_NEXT(m); } return (outchain); } else { /* save off the end and update the end-chain postion */ m = appendchain; while (m) { if (SCTP_BUF_NEXT(m) == NULL) { *endofchain = m; break; } m = SCTP_BUF_NEXT(m); } return (appendchain); } } int sctp_med_chunk_output(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_association *asoc, int *num_out, int *reason_code, int control_only, int *cwnd_full, int from_where, struct timeval *now, int *now_filled, int frag_point, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ); static void sctp_sendall_iterator(struct sctp_inpcb *inp, struct sctp_tcb *stcb, void *ptr, uint32_t val) { struct sctp_copy_all *ca; struct mbuf *m; int ret = 0; int added_control = 0; int un_sent, do_chunk_output = 1; struct sctp_association *asoc; ca = (struct sctp_copy_all *)ptr; if (ca->m == NULL) { return; } if (ca->inp != inp) { /* TSNH */ return; } if ((ca->m) && ca->sndlen) { m = SCTP_M_COPYM(ca->m, 0, M_COPYALL, M_DONTWAIT); if (m == NULL) { /* can't copy so we are done */ ca->cnt_failed++; return; } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = m; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif } else { m = NULL; } SCTP_TCB_LOCK_ASSERT(stcb); if (ca->sndrcv.sinfo_flags & SCTP_ABORT) { /* Abort this assoc with m as the user defined reason */ if (m) { struct sctp_paramhdr *ph; SCTP_BUF_PREPEND(m, sizeof(struct sctp_paramhdr), M_DONTWAIT); if (m) { ph = mtod(m, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(ca->sndlen); } /* * We add one here to keep the assoc from * dis-appearing on us. */ atomic_add_int(&stcb->asoc.refcnt, 1); sctp_abort_an_association(inp, stcb, SCTP_RESPONSE_TO_USER_REQ, m, SCTP_SO_NOT_LOCKED); /* * sctp_abort_an_association calls sctp_free_asoc() * free association will NOT free it since we * incremented the refcnt .. we do this to prevent * it being freed and things getting tricky since we * could end up (from free_asoc) calling inpcb_free * which would get a recursive lock call to the * iterator lock.. But as a consequence of that the * stcb will return to us un-locked.. since * free_asoc returns with either no TCB or the TCB * unlocked, we must relock.. to unlock in the * iterator timer :-0 */ SCTP_TCB_LOCK(stcb); atomic_add_int(&stcb->asoc.refcnt, -1); goto no_chunk_output; } } else { if (m) { ret = sctp_msg_append(stcb, stcb->asoc.primary_destination, m, &ca->sndrcv, 1); } asoc = &stcb->asoc; if (ca->sndrcv.sinfo_flags & SCTP_EOF) { /* shutdown this assoc */ int cnt; cnt = sctp_is_there_unsent_data(stcb); if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (cnt == 0)) { if (asoc->locked_on_sending) { goto abort_anyway; } /* * there is nothing queued to send, so I'm * done... */ if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_RECEIVED) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { /* * only send SHUTDOWN the first time * through */ sctp_send_shutdown(stcb, stcb->asoc.primary_destination); if (SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, asoc->primary_destination); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); added_control = 1; do_chunk_output = 0; } } else { /* * we still got (or just got) data to send, * so set SHUTDOWN_PENDING */ /* * XXX sockets draft says that SCTP_EOF * should be sent with no data. currently, * we will allow user data to be sent first * and move to SHUTDOWN-PENDING */ if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_RECEIVED) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { if (asoc->locked_on_sending) { /* * Locked to send out the * data */ struct sctp_stream_queue_pending *sp; sp = TAILQ_LAST(&asoc->locked_on_sending->outqueue, sctp_streamhead); if (sp) { if ((sp->length == 0) && (sp->msg_is_complete == 0)) asoc->state |= SCTP_STATE_PARTIAL_MSG_LEFT; } } asoc->state |= SCTP_STATE_SHUTDOWN_PENDING; if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { abort_anyway: atomic_add_int(&stcb->asoc.refcnt, 1); sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_RESPONSE_TO_USER_REQ, NULL, SCTP_SO_NOT_LOCKED); atomic_add_int(&stcb->asoc.refcnt, -1); goto no_chunk_output; } sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } } } } un_sent = ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) + (stcb->asoc.stream_queue_cnt * sizeof(struct sctp_data_chunk))); if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) && (stcb->asoc.total_flight > 0) && (un_sent < (int)(stcb->asoc.smallest_mtu - SCTP_MIN_OVERHEAD)) ) { do_chunk_output = 0; } if (do_chunk_output) sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_NOT_LOCKED); else if (added_control) { int num_out = 0, reason = 0, cwnd_full = 0, now_filled = 0; struct timeval now; int frag_point; frag_point = sctp_get_frag_point(stcb, &stcb->asoc); (void)sctp_med_chunk_output(inp, stcb, &stcb->asoc, &num_out, &reason, 1, &cwnd_full, 1, &now, &now_filled, frag_point, SCTP_SO_NOT_LOCKED); } no_chunk_output: if (ret) { ca->cnt_failed++; } else { ca->cnt_sent++; } } static void sctp_sendall_completes(void *ptr, uint32_t val) { struct sctp_copy_all *ca; ca = (struct sctp_copy_all *)ptr; /* * Do a notify here? Kacheong suggests that the notify be done at * the send time.. so you would push up a notification if any send * failed. Don't know if this is feasable since the only failures we * have is "memory" related and if you cannot get an mbuf to send * the data you surely can't get an mbuf to send up to notify the * user you can't send the data :-> */ /* now free everything */ sctp_m_freem(ca->m); SCTP_FREE(ca, SCTP_M_COPYAL); } #define MC_ALIGN(m, len) do { \ SCTP_BUF_RESV_UF(m, ((MCLBYTES - (len)) & ~(sizeof(long) - 1)); \ } while (0) static struct mbuf * sctp_copy_out_all(struct uio *uio, int len) { struct mbuf *ret, *at; int left, willcpy, cancpy, error; ret = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_WAIT, 1, MT_DATA); if (ret == NULL) { /* TSNH */ return (NULL); } left = len; SCTP_BUF_LEN(ret) = 0; /* save space for the data chunk header */ cancpy = M_TRAILINGSPACE(ret); willcpy = min(cancpy, left); at = ret; while (left > 0) { /* Align data to the end */ error = uiomove(mtod(at, caddr_t), willcpy, uio); if (error) { err_out_now: sctp_m_freem(at); return (NULL); } SCTP_BUF_LEN(at) = willcpy; SCTP_BUF_NEXT_PKT(at) = SCTP_BUF_NEXT(at) = 0; left -= willcpy; if (left > 0) { SCTP_BUF_NEXT(at) = sctp_get_mbuf_for_msg(left, 0, M_WAIT, 1, MT_DATA); if (SCTP_BUF_NEXT(at) == NULL) { goto err_out_now; } at = SCTP_BUF_NEXT(at); SCTP_BUF_LEN(at) = 0; cancpy = M_TRAILINGSPACE(at); willcpy = min(cancpy, left); } } return (ret); } static int sctp_sendall(struct sctp_inpcb *inp, struct uio *uio, struct mbuf *m, struct sctp_sndrcvinfo *srcv) { int ret; struct sctp_copy_all *ca; SCTP_MALLOC(ca, struct sctp_copy_all *, sizeof(struct sctp_copy_all), SCTP_M_COPYAL); if (ca == NULL) { sctp_m_freem(m); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } memset(ca, 0, sizeof(struct sctp_copy_all)); ca->inp = inp; memcpy(&ca->sndrcv, srcv, sizeof(struct sctp_nonpad_sndrcvinfo)); /* * take off the sendall flag, it would be bad if we failed to do * this :-0 */ ca->sndrcv.sinfo_flags &= ~SCTP_SENDALL; /* get length and mbuf chain */ if (uio) { ca->sndlen = uio->uio_resid; ca->m = sctp_copy_out_all(uio, ca->sndlen); if (ca->m == NULL) { SCTP_FREE(ca, SCTP_M_COPYAL); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } } else { /* Gather the length of the send */ struct mbuf *mat; mat = m; ca->sndlen = 0; while (m) { ca->sndlen += SCTP_BUF_LEN(m); m = SCTP_BUF_NEXT(m); } ca->m = mat; } ret = sctp_initiate_iterator(NULL, sctp_sendall_iterator, NULL, SCTP_PCB_ANY_FLAGS, SCTP_PCB_ANY_FEATURES, SCTP_ASOC_ANY_STATE, (void *)ca, 0, sctp_sendall_completes, inp, 1); if (ret) { SCTP_PRINTF("Failed to initiate iterator for sendall\n"); SCTP_FREE(ca, SCTP_M_COPYAL); SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EFAULT); return (EFAULT); } return (0); } void sctp_toss_old_cookies(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_tmit_chunk *chk, *nchk; chk = TAILQ_FIRST(&asoc->control_send_queue); while (chk) { nchk = TAILQ_NEXT(chk, sctp_next); if (chk->rec.chunk_id.id == SCTP_COOKIE_ECHO) { TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } asoc->ctrl_queue_cnt--; sctp_free_a_chunk(stcb, chk); } chk = nchk; } } void sctp_toss_old_asconf(struct sctp_tcb *stcb) { struct sctp_association *asoc; struct sctp_tmit_chunk *chk, *chk_tmp; struct sctp_asconf_chunk *acp; asoc = &stcb->asoc; for (chk = TAILQ_FIRST(&asoc->asconf_send_queue); chk != NULL; chk = chk_tmp) { /* get next chk */ chk_tmp = TAILQ_NEXT(chk, sctp_next); /* find SCTP_ASCONF chunk in queue */ if (chk->rec.chunk_id.id == SCTP_ASCONF) { if (chk->data) { acp = mtod(chk->data, struct sctp_asconf_chunk *); if (compare_with_wrap(ntohl(acp->serial_number), stcb->asoc.asconf_seq_out_acked, MAX_SEQ)) { /* Not Acked yet */ break; } } TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } asoc->ctrl_queue_cnt--; sctp_free_a_chunk(stcb, chk); } } } static void sctp_clean_up_datalist(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_tmit_chunk **data_list, int bundle_at, struct sctp_nets *net) { int i; struct sctp_tmit_chunk *tp1; for (i = 0; i < bundle_at; i++) { /* off of the send queue */ if (i) { /* * Any chunk NOT 0 you zap the time chunk 0 gets * zapped or set based on if a RTO measurment is * needed. */ data_list[i]->do_rtt = 0; } /* record time */ data_list[i]->sent_rcv_time = net->last_sent_time; data_list[i]->rec.data.fast_retran_tsn = data_list[i]->rec.data.TSN_seq; TAILQ_REMOVE(&asoc->send_queue, data_list[i], sctp_next); /* on to the sent queue */ tp1 = TAILQ_LAST(&asoc->sent_queue, sctpchunk_listhead); if ((tp1) && (compare_with_wrap(tp1->rec.data.TSN_seq, data_list[i]->rec.data.TSN_seq, MAX_TSN))) { struct sctp_tmit_chunk *tpp; /* need to move back */ back_up_more: tpp = TAILQ_PREV(tp1, sctpchunk_listhead, sctp_next); if (tpp == NULL) { TAILQ_INSERT_BEFORE(tp1, data_list[i], sctp_next); goto all_done; } tp1 = tpp; if (compare_with_wrap(tp1->rec.data.TSN_seq, data_list[i]->rec.data.TSN_seq, MAX_TSN)) { goto back_up_more; } TAILQ_INSERT_AFTER(&asoc->sent_queue, tp1, data_list[i], sctp_next); } else { TAILQ_INSERT_TAIL(&asoc->sent_queue, data_list[i], sctp_next); } all_done: /* This does not lower until the cum-ack passes it */ asoc->sent_queue_cnt++; asoc->send_queue_cnt--; if ((asoc->peers_rwnd <= 0) && (asoc->total_flight == 0) && (bundle_at == 1)) { /* Mark the chunk as being a window probe */ SCTP_STAT_INCR(sctps_windowprobed); } #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xC2, 3); #endif data_list[i]->sent = SCTP_DATAGRAM_SENT; data_list[i]->snd_count = 1; data_list[i]->rec.data.chunk_was_revoked = 0; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_UP, data_list[i]->whoTo->flight_size, data_list[i]->book_size, (uintptr_t) data_list[i]->whoTo, data_list[i]->rec.data.TSN_seq); } sctp_flight_size_increase(data_list[i]); sctp_total_flight_increase(stcb, data_list[i]); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { sctp_log_rwnd(SCTP_DECREASE_PEER_RWND, asoc->peers_rwnd, data_list[i]->send_size, SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)); } asoc->peers_rwnd = sctp_sbspace_sub(asoc->peers_rwnd, (uint32_t) (data_list[i]->send_size + SCTP_BASE_SYSCTL(sctp_peer_chunk_oh))); if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { /* SWS sender side engages */ asoc->peers_rwnd = 0; } } } static void sctp_clean_up_ctl(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_tmit_chunk *chk, *nchk; for (chk = TAILQ_FIRST(&asoc->control_send_queue); chk; chk = nchk) { nchk = TAILQ_NEXT(chk, sctp_next); if ((chk->rec.chunk_id.id == SCTP_SELECTIVE_ACK) || (chk->rec.chunk_id.id == SCTP_HEARTBEAT_REQUEST) || (chk->rec.chunk_id.id == SCTP_HEARTBEAT_ACK) || (chk->rec.chunk_id.id == SCTP_SHUTDOWN) || (chk->rec.chunk_id.id == SCTP_SHUTDOWN_ACK) || (chk->rec.chunk_id.id == SCTP_OPERATION_ERROR) || (chk->rec.chunk_id.id == SCTP_PACKET_DROPPED) || (chk->rec.chunk_id.id == SCTP_COOKIE_ACK) || (chk->rec.chunk_id.id == SCTP_ECN_CWR) || (chk->rec.chunk_id.id == SCTP_ASCONF_ACK)) { /* Stray chunks must be cleaned up */ clean_up_anyway: TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } asoc->ctrl_queue_cnt--; sctp_free_a_chunk(stcb, chk); } else if (chk->rec.chunk_id.id == SCTP_STREAM_RESET) { /* special handling, we must look into the param */ if (chk != asoc->str_reset) { goto clean_up_anyway; } } } } static int sctp_can_we_split_this(struct sctp_tcb *stcb, uint32_t length, uint32_t goal_mtu, uint32_t frag_point, int eeor_on) { /* * Make a decision on if I should split a msg into multiple parts. * This is only asked of incomplete messages. */ if (eeor_on) { /* * If we are doing EEOR we need to always send it if its the * entire thing, since it might be all the guy is putting in * the hopper. */ if (goal_mtu >= length) { /*- * If we have data outstanding, * we get another chance when the sack * arrives to transmit - wait for more data */ if (stcb->asoc.total_flight == 0) { /* * If nothing is in flight, we zero the * packet counter. */ return (length); } return (0); } else { /* You can fill the rest */ return (goal_mtu); } } /*- * For those strange folk that make the send buffer * smaller than our fragmentation point, we can't * get a full msg in so we have to allow splitting. */ if (SCTP_SB_LIMIT_SND(stcb->sctp_socket) < frag_point) { return (length); } if ((length <= goal_mtu) || ((length - goal_mtu) < SCTP_BASE_SYSCTL(sctp_min_residual))) { /* Sub-optimial residual don't split in non-eeor mode. */ return (0); } /* * If we reach here length is larger than the goal_mtu. Do we wish * to split it for the sake of packet putting together? */ if (goal_mtu >= min(SCTP_BASE_SYSCTL(sctp_min_split_point), frag_point)) { /* Its ok to split it */ return (min(goal_mtu, frag_point)); } /* Nope, can't split */ return (0); } static uint32_t sctp_move_to_outqueue(struct sctp_tcb *stcb, struct sctp_nets *net, struct sctp_stream_out *strq, uint32_t goal_mtu, uint32_t frag_point, int *locked, int *giveup, int eeor_mode, int *bail) { /* Move from the stream to the send_queue keeping track of the total */ struct sctp_association *asoc; struct sctp_stream_queue_pending *sp; struct sctp_tmit_chunk *chk; struct sctp_data_chunk *dchkh; uint32_t to_move, length; uint8_t rcv_flags = 0; uint8_t some_taken; uint8_t send_lock_up = 0; SCTP_TCB_LOCK_ASSERT(stcb); asoc = &stcb->asoc; one_more_time: /* sa_ignore FREED_MEMORY */ sp = TAILQ_FIRST(&strq->outqueue); if (sp == NULL) { *locked = 0; if (send_lock_up == 0) { SCTP_TCB_SEND_LOCK(stcb); send_lock_up = 1; } sp = TAILQ_FIRST(&strq->outqueue); if (sp) { goto one_more_time; } if (strq->last_msg_incomplete) { SCTP_PRINTF("Huh? Stream:%d lm_in_c=%d but queue is NULL\n", strq->stream_no, strq->last_msg_incomplete); strq->last_msg_incomplete = 0; } to_move = 0; if (send_lock_up) { SCTP_TCB_SEND_UNLOCK(stcb); send_lock_up = 0; } goto out_of; } if ((sp->msg_is_complete) && (sp->length == 0)) { if (sp->sender_all_done) { /* * We are doing differed cleanup. Last time through * when we took all the data the sender_all_done was * not set. */ if (sp->put_last_out == 0) { SCTP_PRINTF("Gak, put out entire msg with NO end!-1\n"); SCTP_PRINTF("sender_done:%d len:%d msg_comp:%d put_last_out:%d send_lock:%d\n", sp->sender_all_done, sp->length, sp->msg_is_complete, sp->put_last_out, send_lock_up); } if ((TAILQ_NEXT(sp, next) == NULL) && (send_lock_up == 0)) { SCTP_TCB_SEND_LOCK(stcb); send_lock_up = 1; } atomic_subtract_int(&asoc->stream_queue_cnt, 1); TAILQ_REMOVE(&strq->outqueue, sp, next); sctp_free_remote_addr(sp->net); if (sp->data) { sctp_m_freem(sp->data); sp->data = NULL; } sctp_free_a_strmoq(stcb, sp); /* we can't be locked to it */ *locked = 0; stcb->asoc.locked_on_sending = NULL; if (send_lock_up) { SCTP_TCB_SEND_UNLOCK(stcb); send_lock_up = 0; } /* back to get the next msg */ goto one_more_time; } else { /* * sender just finished this but still holds a * reference */ *locked = 1; *giveup = 1; to_move = 0; goto out_of; } } else { /* is there some to get */ if (sp->length == 0) { /* no */ *locked = 1; *giveup = 1; to_move = 0; goto out_of; } } some_taken = sp->some_taken; if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { sp->msg_is_complete = 1; } re_look: length = sp->length; if (sp->msg_is_complete) { /* The message is complete */ to_move = min(length, frag_point); if (to_move == length) { /* All of it fits in the MTU */ if (sp->some_taken) { rcv_flags |= SCTP_DATA_LAST_FRAG; sp->put_last_out = 1; } else { rcv_flags |= SCTP_DATA_NOT_FRAG; sp->put_last_out = 1; } } else { /* Not all of it fits, we fragment */ if (sp->some_taken == 0) { rcv_flags |= SCTP_DATA_FIRST_FRAG; } sp->some_taken = 1; } } else { to_move = sctp_can_we_split_this(stcb, length, goal_mtu, frag_point, eeor_mode); if (to_move) { /*- * We use a snapshot of length in case it * is expanding during the compare. */ uint32_t llen; llen = length; if (to_move >= llen) { to_move = llen; if (send_lock_up == 0) { /*- * We are taking all of an incomplete msg * thus we need a send lock. */ SCTP_TCB_SEND_LOCK(stcb); send_lock_up = 1; if (sp->msg_is_complete) { /* * the sender finished the * msg */ goto re_look; } } } if (sp->some_taken == 0) { rcv_flags |= SCTP_DATA_FIRST_FRAG; sp->some_taken = 1; } } else { /* Nothing to take. */ if (sp->some_taken) { *locked = 1; } *giveup = 1; to_move = 0; goto out_of; } } /* If we reach here, we can copy out a chunk */ sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* No chunk memory */ *giveup = 1; to_move = 0; goto out_of; } /* * Setup for unordered if needed by looking at the user sent info * flags. */ if (sp->sinfo_flags & SCTP_UNORDERED) { rcv_flags |= SCTP_DATA_UNORDERED; } /* clear out the chunk before setting up */ memset(chk, 0, sizeof(*chk)); chk->rec.data.rcv_flags = rcv_flags; if (to_move >= length) { /* we think we can steal the whole thing */ if ((sp->sender_all_done == 0) && (send_lock_up == 0)) { SCTP_TCB_SEND_LOCK(stcb); send_lock_up = 1; } if (to_move < sp->length) { /* bail, it changed */ goto dont_do_it; } chk->data = sp->data; chk->last_mbuf = sp->tail_mbuf; /* register the stealing */ sp->data = sp->tail_mbuf = NULL; } else { struct mbuf *m; dont_do_it: chk->data = SCTP_M_COPYM(sp->data, 0, to_move, M_DONTWAIT); chk->last_mbuf = NULL; if (chk->data == NULL) { sp->some_taken = some_taken; sctp_free_a_chunk(stcb, chk); *bail = 1; to_move = 0; goto out_of; } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = chk->data; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif /* Pull off the data */ m_adj(sp->data, to_move); /* Now lets work our way down and compact it */ m = sp->data; while (m && (SCTP_BUF_LEN(m) == 0)) { sp->data = SCTP_BUF_NEXT(m); SCTP_BUF_NEXT(m) = NULL; if (sp->tail_mbuf == m) { /*- * Freeing tail? TSNH since * we supposedly were taking less * than the sp->length. */ #ifdef INVARIANTS panic("Huh, freing tail? - TSNH"); #else SCTP_PRINTF("Huh, freeing tail? - TSNH\n"); sp->tail_mbuf = sp->data = NULL; sp->length = 0; #endif } sctp_m_free(m); m = sp->data; } } if (SCTP_BUF_IS_EXTENDED(chk->data)) { chk->copy_by_ref = 1; } else { chk->copy_by_ref = 0; } /* * get last_mbuf and counts of mb useage This is ugly but hopefully * its only one mbuf. */ if (chk->last_mbuf == NULL) { chk->last_mbuf = chk->data; while (SCTP_BUF_NEXT(chk->last_mbuf) != NULL) { chk->last_mbuf = SCTP_BUF_NEXT(chk->last_mbuf); } } if (to_move > length) { /*- This should not happen either * since we always lower to_move to the size * of sp->length if its larger. */ #ifdef INVARIANTS panic("Huh, how can to_move be larger?"); #else SCTP_PRINTF("Huh, how can to_move be larger?\n"); sp->length = 0; #endif } else { atomic_subtract_int(&sp->length, to_move); } if (M_LEADINGSPACE(chk->data) < (int)sizeof(struct sctp_data_chunk)) { /* Not enough room for a chunk header, get some */ struct mbuf *m; m = sctp_get_mbuf_for_msg(1, 0, M_DONTWAIT, 0, MT_DATA); if (m == NULL) { /* * we're in trouble here. _PREPEND below will free * all the data if there is no leading space, so we * must put the data back and restore. */ if (send_lock_up == 0) { SCTP_TCB_SEND_LOCK(stcb); send_lock_up = 1; } if (chk->data == NULL) { /* unsteal the data */ sp->data = chk->data; sp->tail_mbuf = chk->last_mbuf; } else { struct mbuf *m_tmp; /* reassemble the data */ m_tmp = sp->data; sp->data = chk->data; SCTP_BUF_NEXT(chk->last_mbuf) = m_tmp; } sp->some_taken = some_taken; atomic_add_int(&sp->length, to_move); chk->data = NULL; *bail = 1; sctp_free_a_chunk(stcb, chk); to_move = 0; goto out_of; } else { SCTP_BUF_LEN(m) = 0; SCTP_BUF_NEXT(m) = chk->data; chk->data = m; M_ALIGN(chk->data, 4); } } SCTP_BUF_PREPEND(chk->data, sizeof(struct sctp_data_chunk), M_DONTWAIT); if (chk->data == NULL) { /* HELP, TSNH since we assured it would not above? */ #ifdef INVARIANTS panic("prepend failes HELP?"); #else SCTP_PRINTF("prepend fails HELP?\n"); sctp_free_a_chunk(stcb, chk); #endif *bail = 1; to_move = 0; goto out_of; } sctp_snd_sb_alloc(stcb, sizeof(struct sctp_data_chunk)); chk->book_size = chk->send_size = (to_move + sizeof(struct sctp_data_chunk)); chk->book_size_scale = 0; chk->sent = SCTP_DATAGRAM_UNSENT; chk->flags = 0; chk->asoc = &stcb->asoc; chk->pad_inplace = 0; chk->no_fr_allowed = 0; chk->rec.data.stream_seq = sp->strseq; chk->rec.data.stream_number = sp->stream; chk->rec.data.payloadtype = sp->ppid; chk->rec.data.context = sp->context; chk->rec.data.doing_fast_retransmit = 0; chk->rec.data.ect_nonce = 0; /* ECN Nonce */ chk->rec.data.timetodrop = sp->ts; chk->flags = sp->act_flags; chk->addr_over = sp->addr_over; chk->whoTo = net; atomic_add_int(&chk->whoTo->ref_count, 1); chk->rec.data.TSN_seq = atomic_fetchadd_int(&asoc->sending_seq, 1); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_AT_SEND_2_OUTQ) { sctp_misc_ints(SCTP_STRMOUT_LOG_SEND, (uintptr_t) stcb, sp->length, (uint32_t) ((chk->rec.data.stream_number << 16) | chk->rec.data.stream_seq), chk->rec.data.TSN_seq); } dchkh = mtod(chk->data, struct sctp_data_chunk *); /* * Put the rest of the things in place now. Size was done earlier in * previous loop prior to padding. */ #ifdef SCTP_ASOCLOG_OF_TSNS SCTP_TCB_LOCK_ASSERT(stcb); if (asoc->tsn_out_at >= SCTP_TSN_LOG_SIZE) { asoc->tsn_out_at = 0; asoc->tsn_out_wrapped = 1; } asoc->out_tsnlog[asoc->tsn_out_at].tsn = chk->rec.data.TSN_seq; asoc->out_tsnlog[asoc->tsn_out_at].strm = chk->rec.data.stream_number; asoc->out_tsnlog[asoc->tsn_out_at].seq = chk->rec.data.stream_seq; asoc->out_tsnlog[asoc->tsn_out_at].sz = chk->send_size; asoc->out_tsnlog[asoc->tsn_out_at].flgs = chk->rec.data.rcv_flags; asoc->out_tsnlog[asoc->tsn_out_at].stcb = (void *)stcb; asoc->out_tsnlog[asoc->tsn_out_at].in_pos = asoc->tsn_out_at; asoc->out_tsnlog[asoc->tsn_out_at].in_out = 2; asoc->tsn_out_at++; #endif dchkh->ch.chunk_type = SCTP_DATA; dchkh->ch.chunk_flags = chk->rec.data.rcv_flags; dchkh->dp.tsn = htonl(chk->rec.data.TSN_seq); dchkh->dp.stream_id = htons(strq->stream_no); dchkh->dp.stream_sequence = htons(chk->rec.data.stream_seq); dchkh->dp.protocol_id = chk->rec.data.payloadtype; dchkh->ch.chunk_length = htons(chk->send_size); /* Now advance the chk->send_size by the actual pad needed. */ if (chk->send_size < SCTP_SIZE32(chk->book_size)) { /* need a pad */ struct mbuf *lm; int pads; pads = SCTP_SIZE32(chk->book_size) - chk->send_size; if (sctp_pad_lastmbuf(chk->data, pads, chk->last_mbuf) == 0) { chk->pad_inplace = 1; } if ((lm = SCTP_BUF_NEXT(chk->last_mbuf)) != NULL) { /* pad added an mbuf */ chk->last_mbuf = lm; } chk->send_size += pads; } /* We only re-set the policy if it is on */ if (sp->pr_sctp_on) { sctp_set_prsctp_policy(stcb, sp); asoc->pr_sctp_cnt++; chk->pr_sctp_on = 1; } else { chk->pr_sctp_on = 0; } if (sp->msg_is_complete && (sp->length == 0) && (sp->sender_all_done)) { /* All done pull and kill the message */ atomic_subtract_int(&asoc->stream_queue_cnt, 1); if (sp->put_last_out == 0) { SCTP_PRINTF("Gak, put out entire msg with NO end!-2\n"); SCTP_PRINTF("sender_done:%d len:%d msg_comp:%d put_last_out:%d send_lock:%d\n", sp->sender_all_done, sp->length, sp->msg_is_complete, sp->put_last_out, send_lock_up); } if ((send_lock_up == 0) && (TAILQ_NEXT(sp, next) == NULL)) { SCTP_TCB_SEND_LOCK(stcb); send_lock_up = 1; } TAILQ_REMOVE(&strq->outqueue, sp, next); sctp_free_remote_addr(sp->net); if (sp->data) { sctp_m_freem(sp->data); sp->data = NULL; } sctp_free_a_strmoq(stcb, sp); /* we can't be locked to it */ *locked = 0; stcb->asoc.locked_on_sending = NULL; } else { /* more to go, we are locked */ *locked = 1; } asoc->chunks_on_out_queue++; TAILQ_INSERT_TAIL(&asoc->send_queue, chk, sctp_next); asoc->send_queue_cnt++; out_of: if (send_lock_up) { SCTP_TCB_SEND_UNLOCK(stcb); send_lock_up = 0; } return (to_move); } static struct sctp_stream_out * sctp_select_a_stream(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_stream_out *strq; /* Find the next stream to use */ if (asoc->last_out_stream == NULL) { strq = asoc->last_out_stream = TAILQ_FIRST(&asoc->out_wheel); if (asoc->last_out_stream == NULL) { /* huh nothing on the wheel, TSNH */ return (NULL); } goto done_it; } strq = TAILQ_NEXT(asoc->last_out_stream, next_spoke); done_it: if (strq == NULL) { strq = asoc->last_out_stream = TAILQ_FIRST(&asoc->out_wheel); } /* Save off the last stream */ asoc->last_out_stream = strq; return (strq); } static void sctp_fill_outqueue(struct sctp_tcb *stcb, struct sctp_nets *net, int frag_point, int eeor_mode, int *quit_now) { struct sctp_association *asoc; struct sctp_stream_out *strq, *strqn, *strqt; int goal_mtu, moved_how_much, total_moved = 0, bail = 0; int locked, giveup; struct sctp_stream_queue_pending *sp; SCTP_TCB_LOCK_ASSERT(stcb); asoc = &stcb->asoc; #ifdef INET6 if (net->ro._l_addr.sin6.sin6_family == AF_INET6) { goal_mtu = net->mtu - SCTP_MIN_OVERHEAD; } else { /* ?? not sure what else to do */ goal_mtu = net->mtu - SCTP_MIN_V4_OVERHEAD; } #else goal_mtu = net->mtu - SCTP_MIN_OVERHEAD; #endif /* Need an allowance for the data chunk header too */ goal_mtu -= sizeof(struct sctp_data_chunk); /* must make even word boundary */ goal_mtu &= 0xfffffffc; if (asoc->locked_on_sending) { /* We are stuck on one stream until the message completes. */ strqn = strq = asoc->locked_on_sending; locked = 1; } else { strqn = strq = sctp_select_a_stream(stcb, asoc); locked = 0; } while ((goal_mtu > 0) && strq) { sp = TAILQ_FIRST(&strq->outqueue); /* * If CMT is off, we must validate that the stream in * question has the first item pointed towards are network * destionation requested by the caller. Note that if we * turn out to be locked to a stream (assigning TSN's then * we must stop, since we cannot look for another stream * with data to send to that destination). In CMT's case, by * skipping this check, we will send one data packet towards * the requested net. */ if (sp == NULL) { break; } if ((sp->net != net) && (SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0)) { /* none for this network */ if (locked) { break; } else { strq = sctp_select_a_stream(stcb, asoc); if (strq == NULL) /* none left */ break; if (strqn == strq) { /* I have circled */ break; } continue; } } giveup = 0; bail = 0; moved_how_much = sctp_move_to_outqueue(stcb, net, strq, goal_mtu, frag_point, &locked, &giveup, eeor_mode, &bail); asoc->last_out_stream = strq; if (locked) { asoc->locked_on_sending = strq; if ((moved_how_much == 0) || (giveup) || bail) /* no more to move for now */ break; } else { asoc->locked_on_sending = NULL; strqt = sctp_select_a_stream(stcb, asoc); if (TAILQ_FIRST(&strq->outqueue) == NULL) { if (strq == strqn) { /* Must move start to next one */ strqn = TAILQ_NEXT(asoc->last_out_stream, next_spoke); if (strqn == NULL) { strqn = TAILQ_FIRST(&asoc->out_wheel); if (strqn == NULL) { break; } } } sctp_remove_from_wheel(stcb, asoc, strq); } if ((giveup) || bail) { break; } strq = strqt; if (strq == NULL) { break; } } total_moved += moved_how_much; goal_mtu -= (moved_how_much + sizeof(struct sctp_data_chunk)); goal_mtu &= 0xfffffffc; } if (bail) *quit_now = 1; if (total_moved == 0) { if ((SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0) && (net == stcb->asoc.primary_destination)) { /* ran dry for primary network net */ SCTP_STAT_INCR(sctps_primary_randry); } else if (SCTP_BASE_SYSCTL(sctp_cmt_on_off)) { /* ran dry with CMT on */ SCTP_STAT_INCR(sctps_cmt_randry); } } } void sctp_fix_ecn_echo(struct sctp_association *asoc) { struct sctp_tmit_chunk *chk; TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { if (chk->rec.chunk_id.id == SCTP_ECN_ECHO) { chk->sent = SCTP_DATAGRAM_UNSENT; } } } static void sctp_move_to_an_alt(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_nets *net) { struct sctp_tmit_chunk *chk; struct sctp_nets *a_net; SCTP_TCB_LOCK_ASSERT(stcb); /* * JRS 5/14/07 - If CMT PF is turned on, find an alternate * destination using the PF algorithm for finding alternate * destinations. */ if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf)) { a_net = sctp_find_alternate_net(stcb, net, 2); } else { a_net = sctp_find_alternate_net(stcb, net, 0); } if ((a_net != net) && ((a_net->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE)) { /* * We only proceed if a valid alternate is found that is not * this one and is reachable. Here we must move all chunks * queued in the send queue off of the destination address * to our alternate. */ TAILQ_FOREACH(chk, &asoc->send_queue, sctp_next) { if (chk->whoTo == net) { /* Move the chunk to our alternate */ sctp_free_remote_addr(chk->whoTo); chk->whoTo = a_net; atomic_add_int(&a_net->ref_count, 1); } } } } int sctp_med_chunk_output(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_association *asoc, int *num_out, int *reason_code, int control_only, int *cwnd_full, int from_where, struct timeval *now, int *now_filled, int frag_point, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { /* * Ok this is the generic chunk service queue. we must do the * following: - Service the stream queue that is next, moving any * message (note I must get a complete message i.e. FIRST/MIDDLE and * LAST to the out queue in one pass) and assigning TSN's - Check to * see if the cwnd/rwnd allows any output, if so we go ahead and * fomulate and send the low level chunks. Making sure to combine * any control in the control chunk queue also. */ struct sctp_nets *net; struct mbuf *outchain, *endoutchain; struct sctp_tmit_chunk *chk, *nchk; struct sctphdr *shdr; /* temp arrays for unlinking */ struct sctp_tmit_chunk *data_list[SCTP_MAX_DATA_BUNDLING]; int no_fragmentflg, error; int one_chunk, hbflag, skip_data_for_this_net; int asconf, cookie, no_out_cnt; int bundle_at, ctl_cnt, no_data_chunks, cwnd_full_ind, eeor_mode; unsigned int mtu, r_mtu, omtu, mx_mtu, to_out; struct sctp_nets *start_at, *old_startat = NULL, *send_start_at; int tsns_sent = 0; uint32_t auth_offset = 0; struct sctp_auth_chunk *auth = NULL; /* * JRS 5/14/07 - Add flag for whether a heartbeat is sent to the * destination. */ int pf_hbflag = 0; int quit_now = 0; *num_out = 0; cwnd_full_ind = 0; if ((asoc->state & SCTP_STATE_SHUTDOWN_PENDING) || (asoc->state & SCTP_STATE_SHUTDOWN_RECEIVED) || (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR))) { eeor_mode = 1; } else { eeor_mode = 0; } ctl_cnt = no_out_cnt = asconf = cookie = 0; /* * First lets prime the pump. For each destination, if there is room * in the flight size, attempt to pull an MTU's worth out of the * stream queues into the general send_queue */ #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xC2, 2); #endif SCTP_TCB_LOCK_ASSERT(stcb); hbflag = 0; if ((control_only) || (asoc->stream_reset_outstanding)) no_data_chunks = 1; else no_data_chunks = 0; /* Nothing to possible to send? */ if (TAILQ_EMPTY(&asoc->control_send_queue) && TAILQ_EMPTY(&asoc->asconf_send_queue) && TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->out_wheel)) { *reason_code = 9; return (0); } if (asoc->peers_rwnd == 0) { /* No room in peers rwnd */ *cwnd_full = 1; *reason_code = 1; if (asoc->total_flight > 0) { /* we are allowed one chunk in flight */ no_data_chunks = 1; } } if ((no_data_chunks == 0) && (!TAILQ_EMPTY(&asoc->out_wheel))) { if (SCTP_BASE_SYSCTL(sctp_cmt_on_off)) { /* * for CMT we start at the next one past the one we * last added data to. */ if (TAILQ_FIRST(&asoc->send_queue) != NULL) { goto skip_the_fill_from_streams; } if (asoc->last_net_data_came_from) { net = TAILQ_NEXT(asoc->last_net_data_came_from, sctp_next); if (net == NULL) { net = TAILQ_FIRST(&asoc->nets); } } else { /* back to start */ net = TAILQ_FIRST(&asoc->nets); } /* * JRI-TODO: CMT-MPI. Simply set the first * destination (net) to be optimized for the next * message to be pulled out of the outwheel. 1. peek * at outwheel 2. If large message, set net = * highest_cwnd 3. If small message, set net = * lowest rtt */ } else { net = asoc->primary_destination; if (net == NULL) { /* TSNH */ net = TAILQ_FIRST(&asoc->nets); } } start_at = net; one_more_time: for (; net != NULL; net = TAILQ_NEXT(net, sctp_next)) { net->window_probe = 0; if (old_startat && (old_startat == net)) { break; } /* * JRI: if dest is unreachable or unconfirmed, do * not send data to it */ if ((net->dest_state & SCTP_ADDR_NOT_REACHABLE) || (net->dest_state & SCTP_ADDR_UNCONFIRMED)) { continue; } /* * JRI: if dest is in PF state, do not send data to * it */ if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf) && (net->dest_state & SCTP_ADDR_PF)) { continue; } if ((SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0) && (net->ref_count < 2)) { /* nothing can be in queue for this guy */ continue; } if (net->flight_size >= net->cwnd) { /* skip this network, no room */ cwnd_full_ind++; continue; } /* * JRI : this for loop we are in takes in each net, * if its's got space in cwnd and has data sent to * it (when CMT is off) then it calls * sctp_fill_outqueue for the net. This gets data on * the send queue for that network. * * In sctp_fill_outqueue TSN's are assigned and data is * copied out of the stream buffers. Note mostly * copy by reference (we hope). */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, 0, SCTP_CWND_LOG_FILL_OUTQ_CALLED); } sctp_fill_outqueue(stcb, net, frag_point, eeor_mode, &quit_now); if (quit_now) { /* memory alloc failure */ no_data_chunks = 1; goto skip_the_fill_from_streams; } } if (start_at != TAILQ_FIRST(&asoc->nets)) { /* got to pick up the beginning stuff. */ old_startat = start_at; start_at = net = TAILQ_FIRST(&asoc->nets); if (old_startat) goto one_more_time; } } skip_the_fill_from_streams: *cwnd_full = cwnd_full_ind; /* now service each destination and send out what we can for it */ /* Nothing to send? */ if ((TAILQ_FIRST(&asoc->control_send_queue) == NULL) && (TAILQ_FIRST(&asoc->asconf_send_queue) == NULL) && (TAILQ_FIRST(&asoc->send_queue) == NULL)) { *reason_code = 8; return (0); } if (no_data_chunks) { chk = TAILQ_FIRST(&asoc->asconf_send_queue); if (chk == NULL) chk = TAILQ_FIRST(&asoc->control_send_queue); } else { chk = TAILQ_FIRST(&asoc->send_queue); } if (chk) { send_start_at = chk->whoTo; } else { send_start_at = TAILQ_FIRST(&asoc->nets); } old_startat = NULL; again_one_more_time: for (net = send_start_at; net != NULL; net = TAILQ_NEXT(net, sctp_next)) { /* how much can we send? */ /* SCTPDBG("Examine for sending net:%x\n", (uint32_t)net); */ if (old_startat && (old_startat == net)) { /* through list ocmpletely. */ break; } tsns_sent = 0; if (net->ref_count < 2) { /* * Ref-count of 1 so we cannot have data or control * queued to this address. Skip it. */ continue; } ctl_cnt = bundle_at = 0; endoutchain = outchain = NULL; no_fragmentflg = 1; one_chunk = 0; if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { skip_data_for_this_net = 1; } else { skip_data_for_this_net = 0; } if ((net->ro.ro_rt) && (net->ro.ro_rt->rt_ifp)) { /* * if we have a route and an ifp check to see if we * have room to send to this guy */ struct ifnet *ifp; ifp = net->ro.ro_rt->rt_ifp; if ((ifp->if_snd.ifq_len + 2) >= ifp->if_snd.ifq_maxlen) { SCTP_STAT_INCR(sctps_ifnomemqueued); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_MAXBURST_ENABLE) { sctp_log_maxburst(stcb, net, ifp->if_snd.ifq_len, ifp->if_snd.ifq_maxlen, SCTP_MAX_IFP_APPLIED); } continue; } } switch (((struct sockaddr *)&net->ro._l_addr)->sa_family) { case AF_INET: mtu = net->mtu - (sizeof(struct ip) + sizeof(struct sctphdr)); break; #ifdef INET6 case AF_INET6: mtu = net->mtu - (sizeof(struct ip6_hdr) + sizeof(struct sctphdr)); break; #endif default: /* TSNH */ mtu = net->mtu; break; } mx_mtu = mtu; to_out = 0; if (mtu > asoc->peers_rwnd) { if (asoc->total_flight > 0) { /* We have a packet in flight somewhere */ r_mtu = asoc->peers_rwnd; } else { /* We are always allowed to send one MTU out */ one_chunk = 1; r_mtu = mtu; } } else { r_mtu = mtu; } /************************/ /* ASCONF transmission */ /************************/ /* Now first lets go through the asconf queue */ for (chk = TAILQ_FIRST(&asoc->asconf_send_queue); chk; chk = nchk) { nchk = TAILQ_NEXT(chk, sctp_next); if (chk->rec.chunk_id.id != SCTP_ASCONF) { continue; } if (chk->whoTo != net) { /* * No, not sent to the network we are * looking at */ break; } if (chk->data == NULL) { break; } if (chk->sent != SCTP_DATAGRAM_UNSENT && chk->sent != SCTP_DATAGRAM_RESEND) { break; } /* * if no AUTH is yet included and this chunk * requires it, make sure to account for it. We * don't apply the size until the AUTH chunk is * actually added below in case there is no room for * this chunk. NOTE: we overload the use of "omtu" * here */ if ((auth == NULL) && sctp_auth_is_required_chunk(chk->rec.chunk_id.id, stcb->asoc.peer_auth_chunks)) { omtu = sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); } else omtu = 0; /* Here we do NOT factor the r_mtu */ if ((chk->send_size < (int)(mtu - omtu)) || (chk->flags & CHUNK_FLAGS_FRAGMENT_OK)) { /* * We probably should glom the mbuf chain * from the chk->data for control but the * problem is it becomes yet one more level * of tracking to do if for some reason * output fails. Then I have got to * reconstruct the merged control chain.. el * yucko.. for now we take the easy way and * do the copy */ /* * Add an AUTH chunk, if chunk requires it * save the offset into the chain for AUTH */ if ((auth == NULL) && (sctp_auth_is_required_chunk(chk->rec.chunk_id.id, stcb->asoc.peer_auth_chunks))) { outchain = sctp_add_auth_chunk(outchain, &endoutchain, &auth, &auth_offset, stcb, chk->rec.chunk_id.id); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } outchain = sctp_copy_mbufchain(chk->data, outchain, &endoutchain, (int)chk->rec.chunk_id.can_take_data, chk->send_size, chk->copy_by_ref); if (outchain == NULL) { *reason_code = 8; SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); /* update our MTU size */ if (mtu > (chk->send_size + omtu)) mtu -= (chk->send_size + omtu); else mtu = 0; to_out += (chk->send_size + omtu); /* Do clear IP_DF ? */ if (chk->flags & CHUNK_FLAGS_FRAGMENT_OK) { no_fragmentflg = 0; } if (chk->rec.chunk_id.can_take_data) chk->data = NULL; /* * set hb flag since we can use these for * RTO */ hbflag = 1; asconf = 1; /* * should sysctl this: don't bundle data * with ASCONF since it requires AUTH */ no_data_chunks = 1; chk->sent = SCTP_DATAGRAM_SENT; chk->snd_count++; if (mtu == 0) { /* * Ok we are out of room but we can * output without effecting the * flight size since this little guy * is a control only packet. */ sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, inp, stcb, net); /* * do NOT clear the asconf flag as * it is used to do appropriate * source address selection. */ SCTP_BUF_PREPEND(outchain, sizeof(struct sctphdr), M_DONTWAIT); if (outchain == NULL) { /* no memory */ SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOBUFS); error = ENOBUFS; *reason_code = 7; continue; } shdr = mtod(outchain, struct sctphdr *); shdr->src_port = inp->sctp_lport; shdr->dest_port = stcb->rport; shdr->v_tag = htonl(stcb->asoc.peer_vtag); shdr->checksum = 0; auth_offset += sizeof(struct sctphdr); if ((error = sctp_lowlevel_chunk_output(inp, stcb, net, (struct sockaddr *)&net->ro._l_addr, outchain, auth_offset, auth, no_fragmentflg, 0, NULL, asconf, net->port, so_locked, NULL))) { if (error == ENOBUFS) { asoc->ifp_had_enobuf = 1; SCTP_STAT_INCR(sctps_lowlevelerr); } if (from_where == 0) { SCTP_STAT_INCR(sctps_lowlevelerrusr); } if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); *now_filled = 1; *now = net->last_sent_time; } else { net->last_sent_time = *now; } hbflag = 0; /* error, could not output */ if (error == EHOSTUNREACH) { /* * Destination went * unreachable * during this send */ sctp_move_to_an_alt(stcb, asoc, net); } *reason_code = 7; continue; } else asoc->ifp_had_enobuf = 0; if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); *now_filled = 1; *now = net->last_sent_time; } else { net->last_sent_time = *now; } hbflag = 0; /* * increase the number we sent, if a * cookie is sent we don't tell them * any was sent out. */ outchain = endoutchain = NULL; auth = NULL; auth_offset = 0; if (!no_out_cnt) *num_out += ctl_cnt; /* recalc a clean slate and setup */ if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { mtu = (net->mtu - SCTP_MIN_OVERHEAD); } else { mtu = (net->mtu - SCTP_MIN_V4_OVERHEAD); } to_out = 0; no_fragmentflg = 1; } } } /************************/ /* Control transmission */ /************************/ /* Now first lets go through the control queue */ for (chk = TAILQ_FIRST(&asoc->control_send_queue); chk; chk = nchk) { nchk = TAILQ_NEXT(chk, sctp_next); if (chk->whoTo != net) { /* * No, not sent to the network we are * looking at */ continue; } if (chk->data == NULL) { continue; } if (chk->sent != SCTP_DATAGRAM_UNSENT) { /* * It must be unsent. Cookies and ASCONF's * hang around but there timers will force * when marked for resend. */ continue; } /* * if no AUTH is yet included and this chunk * requires it, make sure to account for it. We * don't apply the size until the AUTH chunk is * actually added below in case there is no room for * this chunk. NOTE: we overload the use of "omtu" * here */ if ((auth == NULL) && sctp_auth_is_required_chunk(chk->rec.chunk_id.id, stcb->asoc.peer_auth_chunks)) { omtu = sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); } else omtu = 0; /* Here we do NOT factor the r_mtu */ if ((chk->send_size < (int)(mtu - omtu)) || (chk->flags & CHUNK_FLAGS_FRAGMENT_OK)) { /* * We probably should glom the mbuf chain * from the chk->data for control but the * problem is it becomes yet one more level * of tracking to do if for some reason * output fails. Then I have got to * reconstruct the merged control chain.. el * yucko.. for now we take the easy way and * do the copy */ /* * Add an AUTH chunk, if chunk requires it * save the offset into the chain for AUTH */ if ((auth == NULL) && (sctp_auth_is_required_chunk(chk->rec.chunk_id.id, stcb->asoc.peer_auth_chunks))) { outchain = sctp_add_auth_chunk(outchain, &endoutchain, &auth, &auth_offset, stcb, chk->rec.chunk_id.id); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } outchain = sctp_copy_mbufchain(chk->data, outchain, &endoutchain, (int)chk->rec.chunk_id.can_take_data, chk->send_size, chk->copy_by_ref); if (outchain == NULL) { *reason_code = 8; SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); /* update our MTU size */ if (mtu > (chk->send_size + omtu)) mtu -= (chk->send_size + omtu); else mtu = 0; to_out += (chk->send_size + omtu); /* Do clear IP_DF ? */ if (chk->flags & CHUNK_FLAGS_FRAGMENT_OK) { no_fragmentflg = 0; } if (chk->rec.chunk_id.can_take_data) chk->data = NULL; /* Mark things to be removed, if needed */ if ((chk->rec.chunk_id.id == SCTP_SELECTIVE_ACK) || (chk->rec.chunk_id.id == SCTP_HEARTBEAT_REQUEST) || (chk->rec.chunk_id.id == SCTP_HEARTBEAT_ACK) || (chk->rec.chunk_id.id == SCTP_SHUTDOWN) || (chk->rec.chunk_id.id == SCTP_SHUTDOWN_ACK) || (chk->rec.chunk_id.id == SCTP_OPERATION_ERROR) || (chk->rec.chunk_id.id == SCTP_COOKIE_ACK) || (chk->rec.chunk_id.id == SCTP_ECN_CWR) || (chk->rec.chunk_id.id == SCTP_PACKET_DROPPED) || (chk->rec.chunk_id.id == SCTP_ASCONF_ACK)) { if (chk->rec.chunk_id.id == SCTP_HEARTBEAT_REQUEST) { hbflag = 1; /* * JRS 5/14/07 - Set the * flag to say a heartbeat * is being sent. */ pf_hbflag = 1; } /* remove these chunks at the end */ if (chk->rec.chunk_id.id == SCTP_SELECTIVE_ACK) { /* turn off the timer */ if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) { sctp_timer_stop(SCTP_TIMER_TYPE_RECV, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_1); } } ctl_cnt++; } else { /* * Other chunks, since they have * timers running (i.e. COOKIE) we * just "trust" that it gets sent or * retransmitted. */ ctl_cnt++; if (chk->rec.chunk_id.id == SCTP_COOKIE_ECHO) { cookie = 1; no_out_cnt = 1; } chk->sent = SCTP_DATAGRAM_SENT; chk->snd_count++; } if (mtu == 0) { /* * Ok we are out of room but we can * output without effecting the * flight size since this little guy * is a control only packet. */ if (asconf) { sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, inp, stcb, net); /* * do NOT clear the asconf * flag as it is used to do * appropriate source * address selection. */ } if (cookie) { sctp_timer_start(SCTP_TIMER_TYPE_COOKIE, inp, stcb, net); cookie = 0; } SCTP_BUF_PREPEND(outchain, sizeof(struct sctphdr), M_DONTWAIT); if (outchain == NULL) { /* no memory */ SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOBUFS); error = ENOBUFS; goto error_out_again; } shdr = mtod(outchain, struct sctphdr *); shdr->src_port = inp->sctp_lport; shdr->dest_port = stcb->rport; shdr->v_tag = htonl(stcb->asoc.peer_vtag); shdr->checksum = 0; auth_offset += sizeof(struct sctphdr); if ((error = sctp_lowlevel_chunk_output(inp, stcb, net, (struct sockaddr *)&net->ro._l_addr, outchain, auth_offset, auth, no_fragmentflg, 0, NULL, asconf, net->port, so_locked, NULL))) { if (error == ENOBUFS) { asoc->ifp_had_enobuf = 1; SCTP_STAT_INCR(sctps_lowlevelerr); } if (from_where == 0) { SCTP_STAT_INCR(sctps_lowlevelerrusr); } error_out_again: /* error, could not output */ if (hbflag) { if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); *now_filled = 1; *now = net->last_sent_time; } else { net->last_sent_time = *now; } hbflag = 0; } if (error == EHOSTUNREACH) { /* * Destination went * unreachable * during this send */ sctp_move_to_an_alt(stcb, asoc, net); } *reason_code = 7; continue; } else asoc->ifp_had_enobuf = 0; /* Only HB or ASCONF advances time */ if (hbflag) { if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); *now_filled = 1; *now = net->last_sent_time; } else { net->last_sent_time = *now; } hbflag = 0; } /* * increase the number we sent, if a * cookie is sent we don't tell them * any was sent out. */ outchain = endoutchain = NULL; auth = NULL; auth_offset = 0; if (!no_out_cnt) *num_out += ctl_cnt; /* recalc a clean slate and setup */ if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { mtu = (net->mtu - SCTP_MIN_OVERHEAD); } else { mtu = (net->mtu - SCTP_MIN_V4_OVERHEAD); } to_out = 0; no_fragmentflg = 1; } } } /*********************/ /* Data transmission */ /*********************/ /* * if AUTH for DATA is required and no AUTH has been added * yet, account for this in the mtu now... if no data can be * bundled, this adjustment won't matter anyways since the * packet will be going out... */ if ((auth == NULL) && sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks)) { mtu -= sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); } /* now lets add any data within the MTU constraints */ switch (((struct sockaddr *)&net->ro._l_addr)->sa_family) { case AF_INET: if (net->mtu > (sizeof(struct ip) + sizeof(struct sctphdr))) omtu = net->mtu - (sizeof(struct ip) + sizeof(struct sctphdr)); else omtu = 0; break; #ifdef INET6 case AF_INET6: if (net->mtu > (sizeof(struct ip6_hdr) + sizeof(struct sctphdr))) omtu = net->mtu - (sizeof(struct ip6_hdr) + sizeof(struct sctphdr)); else omtu = 0; break; #endif default: /* TSNH */ omtu = 0; break; } if ((((asoc->state & SCTP_STATE_OPEN) == SCTP_STATE_OPEN) && (skip_data_for_this_net == 0)) || (cookie)) { for (chk = TAILQ_FIRST(&asoc->send_queue); chk; chk = nchk) { if (no_data_chunks) { /* let only control go out */ *reason_code = 1; break; } if (net->flight_size >= net->cwnd) { /* skip this net, no room for data */ *reason_code = 2; break; } nchk = TAILQ_NEXT(chk, sctp_next); if (chk->whoTo != net) { /* No, not sent to this net */ continue; } if ((chk->send_size > omtu) && ((chk->flags & CHUNK_FLAGS_FRAGMENT_OK) == 0)) { /*- * strange, we have a chunk that is * to big for its destination and * yet no fragment ok flag. * Something went wrong when the * PMTU changed...we did not mark * this chunk for some reason?? I * will fix it here by letting IP * fragment it for now and printing * a warning. This really should not * happen ... */ SCTP_PRINTF("Warning chunk of %d bytes > mtu:%d and yet PMTU disc missed\n", chk->send_size, mtu); chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; } if (((chk->send_size <= mtu) && (chk->send_size <= r_mtu)) || ((chk->flags & CHUNK_FLAGS_FRAGMENT_OK) && (chk->send_size <= asoc->peers_rwnd))) { /* ok we will add this one */ /* * Add an AUTH chunk, if chunk * requires it, save the offset into * the chain for AUTH */ if ((auth == NULL) && (sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks))) { outchain = sctp_add_auth_chunk(outchain, &endoutchain, &auth, &auth_offset, stcb, SCTP_DATA); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } outchain = sctp_copy_mbufchain(chk->data, outchain, &endoutchain, 0, chk->send_size, chk->copy_by_ref); if (outchain == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT3, "No memory?\n"); if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); } *reason_code = 3; SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } /* upate our MTU size */ /* Do clear IP_DF ? */ if (chk->flags & CHUNK_FLAGS_FRAGMENT_OK) { no_fragmentflg = 0; } /* unsigned subtraction of mtu */ if (mtu > chk->send_size) mtu -= chk->send_size; else mtu = 0; /* unsigned subtraction of r_mtu */ if (r_mtu > chk->send_size) r_mtu -= chk->send_size; else r_mtu = 0; to_out += chk->send_size; if ((to_out > mx_mtu) && no_fragmentflg) { #ifdef INVARIANTS panic("Exceeding mtu of %d out size is %d", mx_mtu, to_out); #else SCTP_PRINTF("Exceeding mtu of %d out size is %d\n", mx_mtu, to_out); #endif } chk->window_probe = 0; data_list[bundle_at++] = chk; if (bundle_at >= SCTP_MAX_DATA_BUNDLING) { mtu = 0; break; } if (chk->sent == SCTP_DATAGRAM_UNSENT) { if ((chk->rec.data.rcv_flags & SCTP_DATA_UNORDERED) == 0) { SCTP_STAT_INCR_COUNTER64(sctps_outorderchunks); } else { SCTP_STAT_INCR_COUNTER64(sctps_outunorderchunks); } if (((chk->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) == SCTP_DATA_LAST_FRAG) && ((chk->rec.data.rcv_flags & SCTP_DATA_FIRST_FRAG) == 0)) /* * Count number of * user msg's that * were fragmented * we do this by * counting when we * see a LAST * fragment only. */ SCTP_STAT_INCR_COUNTER64(sctps_fragusrmsgs); } if ((mtu == 0) || (r_mtu == 0) || (one_chunk)) { if (one_chunk) { data_list[0]->window_probe = 1; net->window_probe = 1; } break; } } else { /* * Must be sent in order of the * TSN's (on a network) */ break; } } /* for (chunk gather loop for this net) */ } /* if asoc.state OPEN */ /* Is there something to send for this destination? */ if (outchain) { /* We may need to start a control timer or two */ if (asconf) { sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, inp, stcb, net); /* * do NOT clear the asconf flag as it is * used to do appropriate source address * selection. */ } if (cookie) { sctp_timer_start(SCTP_TIMER_TYPE_COOKIE, inp, stcb, net); cookie = 0; } /* must start a send timer if data is being sent */ if (bundle_at && (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer))) { /* * no timer running on this destination * restart it. */ sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); } else if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf) && pf_hbflag && ((net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF) && (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer))) { /* * JRS 5/14/07 - If a HB has been sent to a * PF destination and no T3 timer is * currently running, start the T3 timer to * track the HBs that were sent. */ sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); } /* Now send it, if there is anything to send :> */ SCTP_BUF_PREPEND(outchain, sizeof(struct sctphdr), M_DONTWAIT); if (outchain == NULL) { /* out of mbufs */ SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOBUFS); error = ENOBUFS; goto errored_send; } shdr = mtod(outchain, struct sctphdr *); shdr->src_port = inp->sctp_lport; shdr->dest_port = stcb->rport; shdr->v_tag = htonl(stcb->asoc.peer_vtag); shdr->checksum = 0; auth_offset += sizeof(struct sctphdr); if ((error = sctp_lowlevel_chunk_output(inp, stcb, net, (struct sockaddr *)&net->ro._l_addr, outchain, auth_offset, auth, no_fragmentflg, bundle_at, data_list[0], asconf, net->port, so_locked, NULL))) { /* error, we could not output */ if (error == ENOBUFS) { SCTP_STAT_INCR(sctps_lowlevelerr); asoc->ifp_had_enobuf = 1; } if (from_where == 0) { SCTP_STAT_INCR(sctps_lowlevelerrusr); } errored_send: SCTPDBG(SCTP_DEBUG_OUTPUT3, "Gak send error %d\n", error); if (hbflag) { if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); *now_filled = 1; *now = net->last_sent_time; } else { net->last_sent_time = *now; } hbflag = 0; } if (error == EHOSTUNREACH) { /* * Destination went unreachable * during this send */ sctp_move_to_an_alt(stcb, asoc, net); } *reason_code = 6; /*- * I add this line to be paranoid. As far as * I can tell the continue, takes us back to * the top of the for, but just to make sure * I will reset these again here. */ ctl_cnt = bundle_at = 0; continue; /* This takes us back to the * for() for the nets. */ } else { asoc->ifp_had_enobuf = 0; } outchain = endoutchain = NULL; auth = NULL; auth_offset = 0; if (bundle_at || hbflag) { /* For data/asconf and hb set time */ if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); *now_filled = 1; *now = net->last_sent_time; } else { net->last_sent_time = *now; } } if (!no_out_cnt) { *num_out += (ctl_cnt + bundle_at); } if (bundle_at) { /* setup for a RTO measurement */ tsns_sent = data_list[0]->rec.data.TSN_seq; /* fill time if not already filled */ if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&asoc->time_last_sent); *now_filled = 1; *now = asoc->time_last_sent; } else { asoc->time_last_sent = *now; } data_list[0]->do_rtt = 1; SCTP_STAT_INCR_BY(sctps_senddata, bundle_at); sctp_clean_up_datalist(stcb, asoc, data_list, bundle_at, net); if (SCTP_BASE_SYSCTL(sctp_early_fr)) { if (net->flight_size < net->cwnd) { /* start or restart it */ if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_2); } SCTP_STAT_INCR(sctps_earlyfrstrout); sctp_timer_start(SCTP_TIMER_TYPE_EARLYFR, inp, stcb, net); } else { /* stop it if its running */ if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) { SCTP_STAT_INCR(sctps_earlyfrstpout); sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_3); } } } } if (one_chunk) { break; } } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, tsns_sent, SCTP_CWND_LOG_FROM_SEND); } } if (old_startat == NULL) { old_startat = send_start_at; send_start_at = TAILQ_FIRST(&asoc->nets); if (old_startat) goto again_one_more_time; } /* * At the end there should be no NON timed chunks hanging on this * queue. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, *num_out, SCTP_CWND_LOG_FROM_SEND); } if ((*num_out == 0) && (*reason_code == 0)) { *reason_code = 4; } else { *reason_code = 5; } sctp_clean_up_ctl(stcb, asoc); return (0); } void sctp_queue_op_err(struct sctp_tcb *stcb, struct mbuf *op_err) { /*- * Prepend a OPERATIONAL_ERROR chunk header and put on the end of * the control chunk queue. */ struct sctp_chunkhdr *hdr; struct sctp_tmit_chunk *chk; struct mbuf *mat; SCTP_TCB_LOCK_ASSERT(stcb); sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(op_err); return; } chk->copy_by_ref = 0; SCTP_BUF_PREPEND(op_err, sizeof(struct sctp_chunkhdr), M_DONTWAIT); if (op_err == NULL) { sctp_free_a_chunk(stcb, chk); return; } chk->send_size = 0; mat = op_err; while (mat != NULL) { chk->send_size += SCTP_BUF_LEN(mat); mat = SCTP_BUF_NEXT(mat); } chk->rec.chunk_id.id = SCTP_OPERATION_ERROR; chk->rec.chunk_id.can_take_data = 1; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = 0; chk->asoc = &stcb->asoc; chk->data = op_err; chk->whoTo = chk->asoc->primary_destination; atomic_add_int(&chk->whoTo->ref_count, 1); hdr = mtod(op_err, struct sctp_chunkhdr *); hdr->chunk_type = SCTP_OPERATION_ERROR; hdr->chunk_flags = 0; hdr->chunk_length = htons(chk->send_size); TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; } int sctp_send_cookie_echo(struct mbuf *m, int offset, struct sctp_tcb *stcb, struct sctp_nets *net) { /*- * pull out the cookie and put it at the front of the control chunk * queue. */ int at; struct mbuf *cookie; struct sctp_paramhdr parm, *phdr; struct sctp_chunkhdr *hdr; struct sctp_tmit_chunk *chk; uint16_t ptype, plen; /* First find the cookie in the param area */ cookie = NULL; at = offset + sizeof(struct sctp_init_chunk); SCTP_TCB_LOCK_ASSERT(stcb); do { phdr = sctp_get_next_param(m, at, &parm, sizeof(parm)); if (phdr == NULL) { return (-3); } ptype = ntohs(phdr->param_type); plen = ntohs(phdr->param_length); if (ptype == SCTP_STATE_COOKIE) { int pad; /* found the cookie */ if ((pad = (plen % 4))) { plen += 4 - pad; } cookie = SCTP_M_COPYM(m, at, plen, M_DONTWAIT); if (cookie == NULL) { /* No memory */ return (-2); } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = cookie; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif break; } at += SCTP_SIZE32(plen); } while (phdr); if (cookie == NULL) { /* Did not find the cookie */ return (-3); } /* ok, we got the cookie lets change it into a cookie echo chunk */ /* first the change from param to cookie */ hdr = mtod(cookie, struct sctp_chunkhdr *); hdr->chunk_type = SCTP_COOKIE_ECHO; hdr->chunk_flags = 0; /* get the chunk stuff now and place it in the FRONT of the queue */ sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(cookie); return (-5); } chk->copy_by_ref = 0; chk->send_size = plen; chk->rec.chunk_id.id = SCTP_COOKIE_ECHO; chk->rec.chunk_id.can_take_data = 0; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = CHUNK_FLAGS_FRAGMENT_OK; chk->asoc = &stcb->asoc; chk->data = cookie; chk->whoTo = chk->asoc->primary_destination; atomic_add_int(&chk->whoTo->ref_count, 1); TAILQ_INSERT_HEAD(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; return (0); } void sctp_send_heartbeat_ack(struct sctp_tcb *stcb, struct mbuf *m, int offset, int chk_length, struct sctp_nets *net) { /* * take a HB request and make it into a HB ack and send it. */ struct mbuf *outchain; struct sctp_chunkhdr *chdr; struct sctp_tmit_chunk *chk; if (net == NULL) /* must have a net pointer */ return; outchain = SCTP_M_COPYM(m, offset, chk_length, M_DONTWAIT); if (outchain == NULL) { /* gak out of memory */ return; } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = outchain; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif chdr = mtod(outchain, struct sctp_chunkhdr *); chdr->chunk_type = SCTP_HEARTBEAT_ACK; chdr->chunk_flags = 0; if (chk_length % 4) { /* need pad */ uint32_t cpthis = 0; int padlen; padlen = 4 - (chk_length % 4); m_copyback(outchain, chk_length, padlen, (caddr_t)&cpthis); } sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(outchain); return; } chk->copy_by_ref = 0; chk->send_size = chk_length; chk->rec.chunk_id.id = SCTP_HEARTBEAT_ACK; chk->rec.chunk_id.can_take_data = 1; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = 0; chk->asoc = &stcb->asoc; chk->data = outchain; chk->whoTo = net; atomic_add_int(&chk->whoTo->ref_count, 1); TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; } void sctp_send_cookie_ack(struct sctp_tcb *stcb) { /* formulate and queue a cookie-ack back to sender */ struct mbuf *cookie_ack; struct sctp_chunkhdr *hdr; struct sctp_tmit_chunk *chk; cookie_ack = NULL; SCTP_TCB_LOCK_ASSERT(stcb); cookie_ack = sctp_get_mbuf_for_msg(sizeof(struct sctp_chunkhdr), 0, M_DONTWAIT, 1, MT_HEADER); if (cookie_ack == NULL) { /* no mbuf's */ return; } SCTP_BUF_RESV_UF(cookie_ack, SCTP_MIN_OVERHEAD); sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(cookie_ack); return; } chk->copy_by_ref = 0; chk->send_size = sizeof(struct sctp_chunkhdr); chk->rec.chunk_id.id = SCTP_COOKIE_ACK; chk->rec.chunk_id.can_take_data = 1; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = 0; chk->asoc = &stcb->asoc; chk->data = cookie_ack; if (chk->asoc->last_control_chunk_from != NULL) { chk->whoTo = chk->asoc->last_control_chunk_from; } else { chk->whoTo = chk->asoc->primary_destination; } atomic_add_int(&chk->whoTo->ref_count, 1); hdr = mtod(cookie_ack, struct sctp_chunkhdr *); hdr->chunk_type = SCTP_COOKIE_ACK; hdr->chunk_flags = 0; hdr->chunk_length = htons(chk->send_size); SCTP_BUF_LEN(cookie_ack) = chk->send_size; TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; return; } void sctp_send_shutdown_ack(struct sctp_tcb *stcb, struct sctp_nets *net) { /* formulate and queue a SHUTDOWN-ACK back to the sender */ struct mbuf *m_shutdown_ack; struct sctp_shutdown_ack_chunk *ack_cp; struct sctp_tmit_chunk *chk; m_shutdown_ack = sctp_get_mbuf_for_msg(sizeof(struct sctp_shutdown_ack_chunk), 0, M_DONTWAIT, 1, MT_HEADER); if (m_shutdown_ack == NULL) { /* no mbuf's */ return; } SCTP_BUF_RESV_UF(m_shutdown_ack, SCTP_MIN_OVERHEAD); sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(m_shutdown_ack); return; } chk->copy_by_ref = 0; chk->send_size = sizeof(struct sctp_chunkhdr); chk->rec.chunk_id.id = SCTP_SHUTDOWN_ACK; chk->rec.chunk_id.can_take_data = 1; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = 0; chk->asoc = &stcb->asoc; chk->data = m_shutdown_ack; chk->whoTo = net; atomic_add_int(&net->ref_count, 1); ack_cp = mtod(m_shutdown_ack, struct sctp_shutdown_ack_chunk *); ack_cp->ch.chunk_type = SCTP_SHUTDOWN_ACK; ack_cp->ch.chunk_flags = 0; ack_cp->ch.chunk_length = htons(chk->send_size); SCTP_BUF_LEN(m_shutdown_ack) = chk->send_size; TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; return; } void sctp_send_shutdown(struct sctp_tcb *stcb, struct sctp_nets *net) { /* formulate and queue a SHUTDOWN to the sender */ struct mbuf *m_shutdown; struct sctp_shutdown_chunk *shutdown_cp; struct sctp_tmit_chunk *chk; m_shutdown = sctp_get_mbuf_for_msg(sizeof(struct sctp_shutdown_chunk), 0, M_DONTWAIT, 1, MT_HEADER); if (m_shutdown == NULL) { /* no mbuf's */ return; } SCTP_BUF_RESV_UF(m_shutdown, SCTP_MIN_OVERHEAD); sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(m_shutdown); return; } chk->copy_by_ref = 0; chk->send_size = sizeof(struct sctp_shutdown_chunk); chk->rec.chunk_id.id = SCTP_SHUTDOWN; chk->rec.chunk_id.can_take_data = 1; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = 0; chk->asoc = &stcb->asoc; chk->data = m_shutdown; chk->whoTo = net; atomic_add_int(&net->ref_count, 1); shutdown_cp = mtod(m_shutdown, struct sctp_shutdown_chunk *); shutdown_cp->ch.chunk_type = SCTP_SHUTDOWN; shutdown_cp->ch.chunk_flags = 0; shutdown_cp->ch.chunk_length = htons(chk->send_size); shutdown_cp->cumulative_tsn_ack = htonl(stcb->asoc.cumulative_tsn); SCTP_BUF_LEN(m_shutdown) = chk->send_size; TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; return; } void sctp_send_asconf(struct sctp_tcb *stcb, struct sctp_nets *net, int addr_locked) { /* * formulate and queue an ASCONF to the peer. ASCONF parameters * should be queued on the assoc queue. */ struct sctp_tmit_chunk *chk; struct mbuf *m_asconf; int len; SCTP_TCB_LOCK_ASSERT(stcb); if ((!TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) && (!sctp_is_feature_on(stcb->sctp_ep, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS))) { /* can't send a new one if there is one in flight already */ return; } /* compose an ASCONF chunk, maximum length is PMTU */ m_asconf = sctp_compose_asconf(stcb, &len, addr_locked); if (m_asconf == NULL) { return; } sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ sctp_m_freem(m_asconf); return; } chk->copy_by_ref = 0; chk->data = m_asconf; chk->send_size = len; chk->rec.chunk_id.id = SCTP_ASCONF; chk->rec.chunk_id.can_take_data = 0; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags = CHUNK_FLAGS_FRAGMENT_OK; chk->asoc = &stcb->asoc; chk->whoTo = net; atomic_add_int(&chk->whoTo->ref_count, 1); TAILQ_INSERT_TAIL(&chk->asoc->asconf_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; return; } void sctp_send_asconf_ack(struct sctp_tcb *stcb) { /* * formulate and queue a asconf-ack back to sender. the asconf-ack * must be stored in the tcb. */ struct sctp_tmit_chunk *chk; struct sctp_asconf_ack *ack, *latest_ack; struct mbuf *m_ack, *m; struct sctp_nets *net = NULL; SCTP_TCB_LOCK_ASSERT(stcb); /* Get the latest ASCONF-ACK */ latest_ack = TAILQ_LAST(&stcb->asoc.asconf_ack_sent, sctp_asconf_ackhead); if (latest_ack == NULL) { return; } if (latest_ack->last_sent_to != NULL && latest_ack->last_sent_to == stcb->asoc.last_control_chunk_from) { /* we're doing a retransmission */ net = sctp_find_alternate_net(stcb, stcb->asoc.last_control_chunk_from, 0); if (net == NULL) { /* no alternate */ if (stcb->asoc.last_control_chunk_from == NULL) net = stcb->asoc.primary_destination; else net = stcb->asoc.last_control_chunk_from; } } else { /* normal case */ if (stcb->asoc.last_control_chunk_from == NULL) net = stcb->asoc.primary_destination; else net = stcb->asoc.last_control_chunk_from; } latest_ack->last_sent_to = net; TAILQ_FOREACH(ack, &stcb->asoc.asconf_ack_sent, next) { if (ack->data == NULL) { continue; } /* copy the asconf_ack */ m_ack = SCTP_M_COPYM(ack->data, 0, M_COPYALL, M_DONTWAIT); if (m_ack == NULL) { /* couldn't copy it */ return; } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { struct mbuf *mat; mat = m_ack; while (mat) { if (SCTP_BUF_IS_EXTENDED(mat)) { sctp_log_mb(mat, SCTP_MBUF_ICOPY); } mat = SCTP_BUF_NEXT(mat); } } #endif sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { /* no memory */ if (m_ack) sctp_m_freem(m_ack); return; } chk->copy_by_ref = 0; chk->whoTo = net; chk->data = m_ack; chk->send_size = 0; /* Get size */ m = m_ack; chk->send_size = ack->len; chk->rec.chunk_id.id = SCTP_ASCONF_ACK; chk->rec.chunk_id.can_take_data = 1; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; /* XXX */ chk->asoc = &stcb->asoc; atomic_add_int(&chk->whoTo->ref_count, 1); TAILQ_INSERT_TAIL(&chk->asoc->control_send_queue, chk, sctp_next); chk->asoc->ctrl_queue_cnt++; } return; } static int sctp_chunk_retransmission(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_association *asoc, int *cnt_out, struct timeval *now, int *now_filled, int *fr_done, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { /*- * send out one MTU of retransmission. If fast_retransmit is * happening we ignore the cwnd. Otherwise we obey the cwnd and * rwnd. For a Cookie or Asconf in the control chunk queue we * retransmit them by themselves. * * For data chunks we will pick out the lowest TSN's in the sent_queue * marked for resend and bundle them all together (up to a MTU of * destination). The address to send to should have been * selected/changed where the retransmission was marked (i.e. in FR * or t3-timeout routines). */ struct sctp_tmit_chunk *data_list[SCTP_MAX_DATA_BUNDLING]; struct sctp_tmit_chunk *chk, *fwd; struct mbuf *m, *endofchain; struct sctphdr *shdr; struct sctp_nets *net = NULL; uint32_t tsns_sent = 0; int no_fragmentflg, bundle_at, cnt_thru; unsigned int mtu; int error, i, one_chunk, fwd_tsn, ctl_cnt, tmr_started; struct sctp_auth_chunk *auth = NULL; uint32_t auth_offset = 0; uint32_t dmtu = 0; SCTP_TCB_LOCK_ASSERT(stcb); tmr_started = ctl_cnt = bundle_at = error = 0; no_fragmentflg = 1; fwd_tsn = 0; *cnt_out = 0; fwd = NULL; endofchain = m = NULL; #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xC3, 1); #endif if ((TAILQ_EMPTY(&asoc->sent_queue)) && (TAILQ_EMPTY(&asoc->control_send_queue))) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "SCTP hits empty queue with cnt set to %d?\n", asoc->sent_queue_retran_cnt); asoc->sent_queue_cnt = 0; asoc->sent_queue_cnt_removeable = 0; /* send back 0/0 so we enter normal transmission */ *cnt_out = 0; return (0); } TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { if ((chk->rec.chunk_id.id == SCTP_COOKIE_ECHO) || (chk->rec.chunk_id.id == SCTP_STREAM_RESET) || (chk->rec.chunk_id.id == SCTP_FORWARD_CUM_TSN)) { if (chk->rec.chunk_id.id == SCTP_STREAM_RESET) { if (chk != asoc->str_reset) { /* * not eligible for retran if its * not ours */ continue; } } ctl_cnt++; if (chk->rec.chunk_id.id == SCTP_FORWARD_CUM_TSN) { fwd_tsn = 1; fwd = chk; } /* * Add an AUTH chunk, if chunk requires it save the * offset into the chain for AUTH */ if ((auth == NULL) && (sctp_auth_is_required_chunk(chk->rec.chunk_id.id, stcb->asoc.peer_auth_chunks))) { m = sctp_add_auth_chunk(m, &endofchain, &auth, &auth_offset, stcb, chk->rec.chunk_id.id); } m = sctp_copy_mbufchain(chk->data, m, &endofchain, 0, chk->send_size, chk->copy_by_ref); break; } } one_chunk = 0; cnt_thru = 0; /* do we have control chunks to retransmit? */ if (m != NULL) { /* Start a timer no matter if we suceed or fail */ if (chk->rec.chunk_id.id == SCTP_COOKIE_ECHO) { sctp_timer_start(SCTP_TIMER_TYPE_COOKIE, inp, stcb, chk->whoTo); } else if (chk->rec.chunk_id.id == SCTP_ASCONF) sctp_timer_start(SCTP_TIMER_TYPE_ASCONF, inp, stcb, chk->whoTo); SCTP_BUF_PREPEND(m, sizeof(struct sctphdr), M_DONTWAIT); if (m == NULL) { SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOBUFS); return (ENOBUFS); } shdr = mtod(m, struct sctphdr *); shdr->src_port = inp->sctp_lport; shdr->dest_port = stcb->rport; shdr->v_tag = htonl(stcb->asoc.peer_vtag); shdr->checksum = 0; auth_offset += sizeof(struct sctphdr); chk->snd_count++; /* update our count */ if ((error = sctp_lowlevel_chunk_output(inp, stcb, chk->whoTo, (struct sockaddr *)&chk->whoTo->ro._l_addr, m, auth_offset, auth, no_fragmentflg, 0, NULL, 0, chk->whoTo->port, so_locked, NULL))) { SCTP_STAT_INCR(sctps_lowlevelerr); return (error); } m = endofchain = NULL; auth = NULL; auth_offset = 0; /* * We don't want to mark the net->sent time here since this * we use this for HB and retrans cannot measure RTT */ /* (void)SCTP_GETTIME_TIMEVAL(&chk->whoTo->last_sent_time); */ *cnt_out += 1; chk->sent = SCTP_DATAGRAM_SENT; sctp_ucount_decr(asoc->sent_queue_retran_cnt); if (fwd_tsn == 0) { return (0); } else { /* Clean up the fwd-tsn list */ sctp_clean_up_ctl(stcb, asoc); return (0); } } /* * Ok, it is just data retransmission we need to do or that and a * fwd-tsn with it all. */ if (TAILQ_EMPTY(&asoc->sent_queue)) { return (SCTP_RETRAN_DONE); } if ((SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED) || (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_WAIT)) { /* not yet open, resend the cookie and that is it */ return (1); } #ifdef SCTP_AUDITING_ENABLED sctp_auditing(20, inp, stcb, NULL); #endif TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { if (chk->sent != SCTP_DATAGRAM_RESEND) { /* No, not sent to this net or not ready for rtx */ continue; } if ((SCTP_BASE_SYSCTL(sctp_max_retran_chunk)) && (chk->snd_count >= SCTP_BASE_SYSCTL(sctp_max_retran_chunk))) { /* Gak, we have exceeded max unlucky retran, abort! */ SCTP_PRINTF("Gak, chk->snd_count:%d >= max:%d - send abort\n", chk->snd_count, SCTP_BASE_SYSCTL(sctp_max_retran_chunk)); atomic_add_int(&stcb->asoc.refcnt, 1); sctp_abort_an_association(stcb->sctp_ep, stcb, 0, NULL, so_locked); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); return (SCTP_RETRAN_EXIT); } /* pick up the net */ net = chk->whoTo; if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { mtu = (net->mtu - SCTP_MIN_OVERHEAD); } else { mtu = net->mtu - SCTP_MIN_V4_OVERHEAD; } if ((asoc->peers_rwnd < mtu) && (asoc->total_flight > 0)) { /* No room in peers rwnd */ uint32_t tsn; tsn = asoc->last_acked_seq + 1; if (tsn == chk->rec.data.TSN_seq) { /* * we make a special exception for this * case. The peer has no rwnd but is missing * the lowest chunk.. which is probably what * is holding up the rwnd. */ goto one_chunk_around; } return (1); } one_chunk_around: if (asoc->peers_rwnd < mtu) { one_chunk = 1; if ((asoc->peers_rwnd == 0) && (asoc->total_flight == 0)) { chk->window_probe = 1; chk->whoTo->window_probe = 1; } } #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xC3, 2); #endif bundle_at = 0; m = NULL; net->fast_retran_ip = 0; if (chk->rec.data.doing_fast_retransmit == 0) { /* * if no FR in progress skip destination that have * flight_size > cwnd. */ if (net->flight_size >= net->cwnd) { continue; } } else { /* * Mark the destination net to have FR recovery * limits put on it. */ *fr_done = 1; net->fast_retran_ip = 1; } /* * if no AUTH is yet included and this chunk requires it, * make sure to account for it. We don't apply the size * until the AUTH chunk is actually added below in case * there is no room for this chunk. */ if ((auth == NULL) && sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks)) { dmtu = sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); } else dmtu = 0; if ((chk->send_size <= (mtu - dmtu)) || (chk->flags & CHUNK_FLAGS_FRAGMENT_OK)) { /* ok we will add this one */ if ((auth == NULL) && (sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks))) { m = sctp_add_auth_chunk(m, &endofchain, &auth, &auth_offset, stcb, SCTP_DATA); } m = sctp_copy_mbufchain(chk->data, m, &endofchain, 0, chk->send_size, chk->copy_by_ref); if (m == NULL) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } /* Do clear IP_DF ? */ if (chk->flags & CHUNK_FLAGS_FRAGMENT_OK) { no_fragmentflg = 0; } /* upate our MTU size */ if (mtu > (chk->send_size + dmtu)) mtu -= (chk->send_size + dmtu); else mtu = 0; data_list[bundle_at++] = chk; if (one_chunk && (asoc->total_flight <= 0)) { SCTP_STAT_INCR(sctps_windowprobed); } } if (one_chunk == 0) { /* * now are there anymore forward from chk to pick * up? */ fwd = TAILQ_NEXT(chk, sctp_next); while (fwd) { if (fwd->sent != SCTP_DATAGRAM_RESEND) { /* Nope, not for retran */ fwd = TAILQ_NEXT(fwd, sctp_next); continue; } if (fwd->whoTo != net) { /* Nope, not the net in question */ fwd = TAILQ_NEXT(fwd, sctp_next); continue; } if ((auth == NULL) && sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks)) { dmtu = sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); } else dmtu = 0; if (fwd->send_size <= (mtu - dmtu)) { if ((auth == NULL) && (sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks))) { m = sctp_add_auth_chunk(m, &endofchain, &auth, &auth_offset, stcb, SCTP_DATA); } m = sctp_copy_mbufchain(fwd->data, m, &endofchain, 0, fwd->send_size, fwd->copy_by_ref); if (m == NULL) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } /* Do clear IP_DF ? */ if (fwd->flags & CHUNK_FLAGS_FRAGMENT_OK) { no_fragmentflg = 0; } /* upate our MTU size */ if (mtu > (fwd->send_size + dmtu)) mtu -= (fwd->send_size + dmtu); else mtu = 0; data_list[bundle_at++] = fwd; if (bundle_at >= SCTP_MAX_DATA_BUNDLING) { break; } fwd = TAILQ_NEXT(fwd, sctp_next); } else { /* can't fit so we are done */ break; } } } /* Is there something to send for this destination? */ if (m) { /* * No matter if we fail/or suceed we should start a * timer. A failure is like a lost IP packet :-) */ if (!SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { /* * no timer running on this destination * restart it. */ sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); tmr_started = 1; } SCTP_BUF_PREPEND(m, sizeof(struct sctphdr), M_DONTWAIT); if (m == NULL) { SCTP_LTRACE_ERR_RET_PKT(m, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOBUFS); return (ENOBUFS); } shdr = mtod(m, struct sctphdr *); shdr->src_port = inp->sctp_lport; shdr->dest_port = stcb->rport; shdr->v_tag = htonl(stcb->asoc.peer_vtag); shdr->checksum = 0; auth_offset += sizeof(struct sctphdr); /* Now lets send it, if there is anything to send :> */ if ((error = sctp_lowlevel_chunk_output(inp, stcb, net, (struct sockaddr *)&net->ro._l_addr, m, auth_offset, auth, no_fragmentflg, 0, NULL, 0, net->port, so_locked, NULL))) { /* error, we could not output */ SCTP_STAT_INCR(sctps_lowlevelerr); return (error); } m = endofchain = NULL; auth = NULL; auth_offset = 0; /* For HB's */ /* * We don't want to mark the net->sent time here * since this we use this for HB and retrans cannot * measure RTT */ /* (void)SCTP_GETTIME_TIMEVAL(&net->last_sent_time); */ /* For auto-close */ cnt_thru++; if (*now_filled == 0) { (void)SCTP_GETTIME_TIMEVAL(&asoc->time_last_sent); *now = asoc->time_last_sent; *now_filled = 1; } else { asoc->time_last_sent = *now; } *cnt_out += bundle_at; #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xC4, bundle_at); #endif if (bundle_at) { tsns_sent = data_list[0]->rec.data.TSN_seq; } for (i = 0; i < bundle_at; i++) { SCTP_STAT_INCR(sctps_sendretransdata); data_list[i]->sent = SCTP_DATAGRAM_SENT; /* * When we have a revoked data, and we * retransmit it, then we clear the revoked * flag since this flag dictates if we * subtracted from the fs */ if (data_list[i]->rec.data.chunk_was_revoked) { /* Deflate the cwnd */ data_list[i]->whoTo->cwnd -= data_list[i]->book_size; data_list[i]->rec.data.chunk_was_revoked = 0; } data_list[i]->snd_count++; sctp_ucount_decr(asoc->sent_queue_retran_cnt); /* record the time */ data_list[i]->sent_rcv_time = asoc->time_last_sent; if (data_list[i]->book_size_scale) { /* * need to double the book size on * this one */ data_list[i]->book_size_scale = 0; /* * Since we double the booksize, we * must also double the output queue * size, since this get shrunk when * we free by this amount. */ atomic_add_int(&((asoc)->total_output_queue_size), data_list[i]->book_size); data_list[i]->book_size *= 2; } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_RWND_ENABLE) { sctp_log_rwnd(SCTP_DECREASE_PEER_RWND, asoc->peers_rwnd, data_list[i]->send_size, SCTP_BASE_SYSCTL(sctp_peer_chunk_oh)); } asoc->peers_rwnd = sctp_sbspace_sub(asoc->peers_rwnd, (uint32_t) (data_list[i]->send_size + SCTP_BASE_SYSCTL(sctp_peer_chunk_oh))); } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_UP_RSND, data_list[i]->whoTo->flight_size, data_list[i]->book_size, (uintptr_t) data_list[i]->whoTo, data_list[i]->rec.data.TSN_seq); } sctp_flight_size_increase(data_list[i]); sctp_total_flight_increase(stcb, data_list[i]); if (asoc->peers_rwnd < stcb->sctp_ep->sctp_ep.sctp_sws_sender) { /* SWS sender side engages */ asoc->peers_rwnd = 0; } if ((i == 0) && (data_list[i]->rec.data.doing_fast_retransmit)) { SCTP_STAT_INCR(sctps_sendfastretrans); if ((data_list[i] == TAILQ_FIRST(&asoc->sent_queue)) && (tmr_started == 0)) { /*- * ok we just fast-retrans'd * the lowest TSN, i.e the * first on the list. In * this case we want to give * some more time to get a * SACK back without a * t3-expiring. */ sctp_timer_stop(SCTP_TIMER_TYPE_SEND, inp, stcb, net, SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_4); sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net); } } } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, net, tsns_sent, SCTP_CWND_LOG_FROM_RESEND); } #ifdef SCTP_AUDITING_ENABLED sctp_auditing(21, inp, stcb, NULL); #endif } else { /* None will fit */ return (1); } if (asoc->sent_queue_retran_cnt <= 0) { /* all done we have no more to retran */ asoc->sent_queue_retran_cnt = 0; break; } if (one_chunk) { /* No more room in rwnd */ return (1); } /* stop the for loop here. we sent out a packet */ break; } return (0); } static int sctp_timer_validation(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_association *asoc, int ret) { struct sctp_nets *net; /* Validate that a timer is running somewhere */ TAILQ_FOREACH(net, &asoc->nets, sctp_next) { if (SCTP_OS_TIMER_PENDING(&net->rxt_timer.timer)) { /* Here is a timer */ return (ret); } } SCTP_TCB_LOCK_ASSERT(stcb); /* Gak, we did not have a timer somewhere */ SCTPDBG(SCTP_DEBUG_OUTPUT3, "Deadlock avoided starting timer on a dest at retran\n"); sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, asoc->primary_destination); return (ret); } void sctp_chunk_output(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_where, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { /*- * Ok this is the generic chunk service queue. we must do the * following: * - See if there are retransmits pending, if so we must * do these first. * - Service the stream queue that is next, moving any * message (note I must get a complete message i.e. * FIRST/MIDDLE and LAST to the out queue in one pass) and assigning * TSN's * - Check to see if the cwnd/rwnd allows any output, if so we * go ahead and fomulate and send the low level chunks. Making sure * to combine any control in the control chunk queue also. */ struct sctp_association *asoc; struct sctp_nets *net; int error = 0, num_out = 0, tot_out = 0, ret = 0, reason_code = 0, burst_cnt = 0, burst_limit = 0; struct timeval now; int now_filled = 0; int cwnd_full = 0; int nagle_on = 0; int frag_point = sctp_get_frag_point(stcb, &stcb->asoc); int un_sent = 0; int fr_done, tot_frs = 0; asoc = &stcb->asoc; if (from_where == SCTP_OUTPUT_FROM_USR_SEND) { if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NODELAY)) { nagle_on = 0; } else { nagle_on = 1; } } SCTP_TCB_LOCK_ASSERT(stcb); un_sent = (stcb->asoc.total_output_queue_size - stcb->asoc.total_flight); if ((un_sent <= 0) && (TAILQ_EMPTY(&asoc->control_send_queue)) && (asoc->sent_queue_retran_cnt == 0)) { /* Nothing to do unless there is something to be sent left */ return; } /* * Do we have something to send, data or control AND a sack timer * running, if so piggy-back the sack. */ if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) { sctp_send_sack(stcb); (void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer); } while (asoc->sent_queue_retran_cnt) { /*- * Ok, it is retransmission time only, we send out only ONE * packet with a single call off to the retran code. */ if (from_where == SCTP_OUTPUT_FROM_COOKIE_ACK) { /*- * Special hook for handling cookiess discarded * by peer that carried data. Send cookie-ack only * and then the next call with get the retran's. */ (void)sctp_med_chunk_output(inp, stcb, asoc, &num_out, &reason_code, 1, &cwnd_full, from_where, &now, &now_filled, frag_point, so_locked); return; } else if (from_where != SCTP_OUTPUT_FROM_HB_TMR) { /* if its not from a HB then do it */ fr_done = 0; ret = sctp_chunk_retransmission(inp, stcb, asoc, &num_out, &now, &now_filled, &fr_done, so_locked); if (fr_done) { tot_frs++; } } else { /* * its from any other place, we don't allow retran * output (only control) */ ret = 1; } if (ret > 0) { /* Can't send anymore */ /*- * now lets push out control by calling med-level * output once. this assures that we WILL send HB's * if queued too. */ (void)sctp_med_chunk_output(inp, stcb, asoc, &num_out, &reason_code, 1, &cwnd_full, from_where, &now, &now_filled, frag_point, so_locked); #ifdef SCTP_AUDITING_ENABLED sctp_auditing(8, inp, stcb, NULL); #endif (void)sctp_timer_validation(inp, stcb, asoc, ret); return; } if (ret < 0) { /*- * The count was off.. retran is not happening so do * the normal retransmission. */ #ifdef SCTP_AUDITING_ENABLED sctp_auditing(9, inp, stcb, NULL); #endif if (ret == SCTP_RETRAN_EXIT) { return; } break; } if (from_where == SCTP_OUTPUT_FROM_T3) { /* Only one transmission allowed out of a timeout */ #ifdef SCTP_AUDITING_ENABLED sctp_auditing(10, inp, stcb, NULL); #endif /* Push out any control */ (void)sctp_med_chunk_output(inp, stcb, asoc, &num_out, &reason_code, 1, &cwnd_full, from_where, &now, &now_filled, frag_point, so_locked); return; } if (tot_frs > asoc->max_burst) { /* Hit FR burst limit */ return; } if ((num_out == 0) && (ret == 0)) { /* No more retrans to send */ break; } } #ifdef SCTP_AUDITING_ENABLED sctp_auditing(12, inp, stcb, NULL); #endif /* Check for bad destinations, if they exist move chunks around. */ burst_limit = asoc->max_burst; TAILQ_FOREACH(net, &asoc->nets, sctp_next) { if ((net->dest_state & SCTP_ADDR_NOT_REACHABLE) == SCTP_ADDR_NOT_REACHABLE) { /*- * if possible move things off of this address we * still may send below due to the dormant state but * we try to find an alternate address to send to * and if we have one we move all queued data on the * out wheel to this alternate address. */ if (net->ref_count > 1) sctp_move_to_an_alt(stcb, asoc, net); } else if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf) && ((net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF)) { /* * JRS 5/14/07 - If CMT PF is on and the current * destination is in PF state, move all queued data * to an alternate desination. */ if (net->ref_count > 1) sctp_move_to_an_alt(stcb, asoc, net); } else { /*- * if ((asoc->sat_network) || (net->addr_is_local)) * { burst_limit = asoc->max_burst * * SCTP_SAT_NETWORK_BURST_INCR; } */ if (SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst)) { if ((net->flight_size + (burst_limit * net->mtu)) < net->cwnd) { /* * JRS - Use the congestion control * given in the congestion control * module */ asoc->cc_functions.sctp_cwnd_update_after_output(stcb, net, burst_limit); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_MAXBURST_ENABLE) { sctp_log_maxburst(stcb, net, 0, burst_limit, SCTP_MAX_BURST_APPLIED); } SCTP_STAT_INCR(sctps_maxburstqueued); } net->fast_retran_ip = 0; } else { if (net->flight_size == 0) { /* Should be decaying the cwnd here */ ; } } } } burst_cnt = 0; cwnd_full = 0; do { error = sctp_med_chunk_output(inp, stcb, asoc, &num_out, &reason_code, 0, &cwnd_full, from_where, &now, &now_filled, frag_point, so_locked); if (error) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "Error %d was returned from med-c-op\n", error); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_MAXBURST_ENABLE) { sctp_log_maxburst(stcb, asoc->primary_destination, error, burst_cnt, SCTP_MAX_BURST_ERROR_STOP); } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, NULL, error, SCTP_SEND_NOW_COMPLETES); sctp_log_cwnd(stcb, NULL, 0xdeadbeef, SCTP_SEND_NOW_COMPLETES); } break; } SCTPDBG(SCTP_DEBUG_OUTPUT3, "m-c-o put out %d\n", num_out); tot_out += num_out; burst_cnt++; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, NULL, num_out, SCTP_SEND_NOW_COMPLETES); if (num_out == 0) { sctp_log_cwnd(stcb, NULL, reason_code, SCTP_SEND_NOW_COMPLETES); } } if (nagle_on) { /*- * When nagle is on, we look at how much is un_sent, then * if its smaller than an MTU and we have data in * flight we stop. */ un_sent = ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) + (stcb->asoc.stream_queue_cnt * sizeof(struct sctp_data_chunk))); if ((un_sent < (int)(stcb->asoc.smallest_mtu - SCTP_MIN_OVERHEAD)) && (stcb->asoc.total_flight > 0)) { break; } } if (TAILQ_EMPTY(&asoc->control_send_queue) && TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->out_wheel)) { /* Nothing left to send */ break; } if ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) <= 0) { /* Nothing left to send */ break; } } while (num_out && (SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst) || (burst_cnt < burst_limit))); if (SCTP_BASE_SYSCTL(sctp_use_cwnd_based_maxburst) == 0) { if (burst_cnt >= burst_limit) { SCTP_STAT_INCR(sctps_maxburstqueued); asoc->burst_limit_applied = 1; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_MAXBURST_ENABLE) { sctp_log_maxburst(stcb, asoc->primary_destination, 0, burst_cnt, SCTP_MAX_BURST_APPLIED); } } else { asoc->burst_limit_applied = 0; } } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) { sctp_log_cwnd(stcb, NULL, tot_out, SCTP_SEND_NOW_COMPLETES); } SCTPDBG(SCTP_DEBUG_OUTPUT1, "Ok, we have put out %d chunks\n", tot_out); /*- * Now we need to clean up the control chunk chain if a ECNE is on * it. It must be marked as UNSENT again so next call will continue * to send it until such time that we get a CWR, to remove it. */ if (stcb->asoc.ecn_echo_cnt_onq) sctp_fix_ecn_echo(asoc); return; } int sctp_output(inp, m, addr, control, p, flags) struct sctp_inpcb *inp; struct mbuf *m; struct sockaddr *addr; struct mbuf *control; struct thread *p; int flags; { if (inp == NULL) { SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } if (inp->sctp_socket == NULL) { SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } return (sctp_sosend(inp->sctp_socket, addr, (struct uio *)NULL, m, control, flags, p )); } void send_forward_tsn(struct sctp_tcb *stcb, struct sctp_association *asoc) { struct sctp_tmit_chunk *chk; struct sctp_forward_tsn_chunk *fwdtsn; SCTP_TCB_LOCK_ASSERT(stcb); TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { if (chk->rec.chunk_id.id == SCTP_FORWARD_CUM_TSN) { /* mark it to unsent */ chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; /* Do we correct its output location? */ if (chk->whoTo != asoc->primary_destination) { sctp_free_remote_addr(chk->whoTo); chk->whoTo = asoc->primary_destination; atomic_add_int(&chk->whoTo->ref_count, 1); } goto sctp_fill_in_rest; } } /* Ok if we reach here we must build one */ sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { return; } chk->copy_by_ref = 0; chk->rec.chunk_id.id = SCTP_FORWARD_CUM_TSN; chk->rec.chunk_id.can_take_data = 0; chk->asoc = asoc; chk->whoTo = NULL; chk->data = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_DATA); if (chk->data == NULL) { sctp_free_a_chunk(stcb, chk); return; } SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->whoTo = asoc->primary_destination; atomic_add_int(&chk->whoTo->ref_count, 1); TAILQ_INSERT_TAIL(&asoc->control_send_queue, chk, sctp_next); asoc->ctrl_queue_cnt++; sctp_fill_in_rest: /*- * Here we go through and fill out the part that deals with * stream/seq of the ones we skip. */ SCTP_BUF_LEN(chk->data) = 0; { struct sctp_tmit_chunk *at, *tp1, *last; struct sctp_strseq *strseq; unsigned int cnt_of_space, i, ovh; unsigned int space_needed; unsigned int cnt_of_skipped = 0; TAILQ_FOREACH(at, &asoc->sent_queue, sctp_next) { if (at->sent != SCTP_FORWARD_TSN_SKIP) { /* no more to look at */ break; } if (at->rec.data.rcv_flags & SCTP_DATA_UNORDERED) { /* We don't report these */ continue; } cnt_of_skipped++; } space_needed = (sizeof(struct sctp_forward_tsn_chunk) + (cnt_of_skipped * sizeof(struct sctp_strseq))); cnt_of_space = M_TRAILINGSPACE(chk->data); if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ovh = SCTP_MIN_OVERHEAD; } else { ovh = SCTP_MIN_V4_OVERHEAD; } if (cnt_of_space > (asoc->smallest_mtu - ovh)) { /* trim to a mtu size */ cnt_of_space = asoc->smallest_mtu - ovh; } if (cnt_of_space < space_needed) { /*- * ok we must trim down the chunk by lowering the * advance peer ack point. */ cnt_of_skipped = (cnt_of_space - ((sizeof(struct sctp_forward_tsn_chunk)) / sizeof(struct sctp_strseq))); /*- * Go through and find the TSN that will be the one * we report. */ at = TAILQ_FIRST(&asoc->sent_queue); for (i = 0; i < cnt_of_skipped; i++) { tp1 = TAILQ_NEXT(at, sctp_next); at = tp1; } last = at; /*- * last now points to last one I can report, update * peer ack point */ asoc->advanced_peer_ack_point = last->rec.data.TSN_seq; space_needed -= (cnt_of_skipped * sizeof(struct sctp_strseq)); } chk->send_size = space_needed; /* Setup the chunk */ fwdtsn = mtod(chk->data, struct sctp_forward_tsn_chunk *); fwdtsn->ch.chunk_length = htons(chk->send_size); fwdtsn->ch.chunk_flags = 0; fwdtsn->ch.chunk_type = SCTP_FORWARD_CUM_TSN; fwdtsn->new_cumulative_tsn = htonl(asoc->advanced_peer_ack_point); chk->send_size = (sizeof(struct sctp_forward_tsn_chunk) + (cnt_of_skipped * sizeof(struct sctp_strseq))); SCTP_BUF_LEN(chk->data) = chk->send_size; fwdtsn++; /*- * Move pointer to after the fwdtsn and transfer to the * strseq pointer. */ strseq = (struct sctp_strseq *)fwdtsn; /*- * Now populate the strseq list. This is done blindly * without pulling out duplicate stream info. This is * inefficent but won't harm the process since the peer will * look at these in sequence and will thus release anything. * It could mean we exceed the PMTU and chop off some that * we could have included.. but this is unlikely (aka 1432/4 * would mean 300+ stream seq's would have to be reported in * one FWD-TSN. With a bit of work we can later FIX this to * optimize and pull out duplcates.. but it does add more * overhead. So for now... not! */ at = TAILQ_FIRST(&asoc->sent_queue); for (i = 0; i < cnt_of_skipped; i++) { tp1 = TAILQ_NEXT(at, sctp_next); if (at->rec.data.rcv_flags & SCTP_DATA_UNORDERED) { /* We don't report these */ i--; at = tp1; continue; } strseq->stream = ntohs(at->rec.data.stream_number); strseq->sequence = ntohs(at->rec.data.stream_seq); strseq++; at = tp1; } } return; } void sctp_send_sack(struct sctp_tcb *stcb) { /*- * Queue up a SACK in the control queue. We must first check to see * if a SACK is somehow on the control queue. If so, we will take * and and remove the old one. */ struct sctp_association *asoc; struct sctp_tmit_chunk *chk, *a_chk; struct sctp_sack_chunk *sack; struct sctp_gap_ack_block *gap_descriptor; struct sack_track *selector; int mergeable = 0; int offset; caddr_t limit; uint32_t *dup; int limit_reached = 0; unsigned int i, jstart, siz, j; unsigned int num_gap_blocks = 0, space; int num_dups = 0; int space_req; a_chk = NULL; asoc = &stcb->asoc; SCTP_TCB_LOCK_ASSERT(stcb); if (asoc->last_data_chunk_from == NULL) { /* Hmm we never received anything */ return; } sctp_set_rwnd(stcb, asoc); TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { if (chk->rec.chunk_id.id == SCTP_SELECTIVE_ACK) { /* Hmm, found a sack already on queue, remove it */ TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); asoc->ctrl_queue_cnt++; a_chk = chk; if (a_chk->data) { sctp_m_freem(a_chk->data); a_chk->data = NULL; } sctp_free_remote_addr(a_chk->whoTo); a_chk->whoTo = NULL; break; } } if (a_chk == NULL) { sctp_alloc_a_chunk(stcb, a_chk); if (a_chk == NULL) { /* No memory so we drop the idea, and set a timer */ if (stcb->asoc.delayed_ack) { sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL, SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_5); sctp_timer_start(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL); } else { stcb->asoc.send_sack = 1; } return; } a_chk->copy_by_ref = 0; /* a_chk->rec.chunk_id.id = SCTP_SELECTIVE_ACK; */ a_chk->rec.chunk_id.id = SCTP_SELECTIVE_ACK; a_chk->rec.chunk_id.can_take_data = 1; } /* Clear our pkt counts */ asoc->data_pkts_seen = 0; a_chk->asoc = asoc; a_chk->snd_count = 0; a_chk->send_size = 0; /* fill in later */ a_chk->sent = SCTP_DATAGRAM_UNSENT; a_chk->whoTo = NULL; if ((asoc->numduptsns) || (asoc->last_data_chunk_from->dest_state & SCTP_ADDR_NOT_REACHABLE) ) { /*- * Ok, we have some duplicates or the destination for the * sack is unreachable, lets see if we can select an * alternate than asoc->last_data_chunk_from */ if ((!(asoc->last_data_chunk_from->dest_state & SCTP_ADDR_NOT_REACHABLE)) && (asoc->used_alt_onsack > asoc->numnets)) { /* We used an alt last time, don't this time */ a_chk->whoTo = NULL; } else { asoc->used_alt_onsack++; a_chk->whoTo = sctp_find_alternate_net(stcb, asoc->last_data_chunk_from, 0); } if (a_chk->whoTo == NULL) { /* Nope, no alternate */ a_chk->whoTo = asoc->last_data_chunk_from; asoc->used_alt_onsack = 0; } } else { /* * No duplicates so we use the last place we received data * from. */ asoc->used_alt_onsack = 0; a_chk->whoTo = asoc->last_data_chunk_from; } if (a_chk->whoTo) { atomic_add_int(&a_chk->whoTo->ref_count, 1); } if (asoc->highest_tsn_inside_map == asoc->cumulative_tsn) { /* no gaps */ space_req = sizeof(struct sctp_sack_chunk); } else { /* gaps get a cluster */ space_req = MCLBYTES; } /* Ok now lets formulate a MBUF with our sack */ a_chk->data = sctp_get_mbuf_for_msg(space_req, 0, M_DONTWAIT, 1, MT_DATA); if ((a_chk->data == NULL) || (a_chk->whoTo == NULL)) { /* rats, no mbuf memory */ if (a_chk->data) { /* was a problem with the destination */ sctp_m_freem(a_chk->data); a_chk->data = NULL; } sctp_free_a_chunk(stcb, a_chk); /* sa_ignore NO_NULL_CHK */ if (stcb->asoc.delayed_ack) { sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL, SCTP_FROM_SCTP_OUTPUT + SCTP_LOC_6); sctp_timer_start(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL); } else { stcb->asoc.send_sack = 1; } return; } /* ok, lets go through and fill it in */ SCTP_BUF_RESV_UF(a_chk->data, SCTP_MIN_OVERHEAD); space = M_TRAILINGSPACE(a_chk->data); if (space > (a_chk->whoTo->mtu - SCTP_MIN_OVERHEAD)) { space = (a_chk->whoTo->mtu - SCTP_MIN_OVERHEAD); } limit = mtod(a_chk->data, caddr_t); limit += space; sack = mtod(a_chk->data, struct sctp_sack_chunk *); sack->ch.chunk_type = SCTP_SELECTIVE_ACK; /* 0x01 is used by nonce for ecn */ if ((SCTP_BASE_SYSCTL(sctp_ecn_enable)) && (SCTP_BASE_SYSCTL(sctp_ecn_nonce)) && (asoc->peer_supports_ecn_nonce)) sack->ch.chunk_flags = (asoc->receiver_nonce_sum & SCTP_SACK_NONCE_SUM); else sack->ch.chunk_flags = 0; if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_use_dac)) { /*- * CMT DAC algorithm: If 2 (i.e., 0x10) packets have been * received, then set high bit to 1, else 0. Reset * pkts_rcvd. */ sack->ch.chunk_flags |= (asoc->cmt_dac_pkts_rcvd << 6); asoc->cmt_dac_pkts_rcvd = 0; } #ifdef SCTP_ASOCLOG_OF_TSNS stcb->asoc.cumack_logsnt[stcb->asoc.cumack_log_atsnt] = asoc->cumulative_tsn; stcb->asoc.cumack_log_atsnt++; if (stcb->asoc.cumack_log_atsnt >= SCTP_TSN_LOG_SIZE) { stcb->asoc.cumack_log_atsnt = 0; } #endif sack->sack.cum_tsn_ack = htonl(asoc->cumulative_tsn); sack->sack.a_rwnd = htonl(asoc->my_rwnd); asoc->my_last_reported_rwnd = asoc->my_rwnd; /* reset the readers interpretation */ stcb->freed_by_sorcv_sincelast = 0; gap_descriptor = (struct sctp_gap_ack_block *)((caddr_t)sack + sizeof(struct sctp_sack_chunk)); siz = (((asoc->highest_tsn_inside_map - asoc->mapping_array_base_tsn) + 1) + 7) / 8; if (compare_with_wrap(asoc->mapping_array_base_tsn, asoc->cumulative_tsn, MAX_TSN)) { offset = 1; /*- * cum-ack behind the mapping array, so we start and use all * entries. */ jstart = 0; } else { offset = asoc->mapping_array_base_tsn - asoc->cumulative_tsn; /*- * we skip the first one when the cum-ack is at or above the * mapping array base. Note this only works if */ jstart = 1; } if (compare_with_wrap(asoc->highest_tsn_inside_map, asoc->cumulative_tsn, MAX_TSN)) { /* we have a gap .. maybe */ for (i = 0; i < siz; i++) { selector = &sack_array[asoc->mapping_array[i]]; if (mergeable && selector->right_edge) { /* * Backup, left and right edges were ok to * merge. */ num_gap_blocks--; gap_descriptor--; } if (selector->num_entries == 0) mergeable = 0; else { for (j = jstart; j < selector->num_entries; j++) { if (mergeable && selector->right_edge) { /* * do a merge by NOT setting * the left side */ mergeable = 0; } else { /* * no merge, set the left * side */ mergeable = 0; gap_descriptor->start = htons((selector->gaps[j].start + offset)); } gap_descriptor->end = htons((selector->gaps[j].end + offset)); num_gap_blocks++; gap_descriptor++; if (((caddr_t)gap_descriptor + sizeof(struct sctp_gap_ack_block)) > limit) { /* no more room */ limit_reached = 1; break; } } if (selector->left_edge) { mergeable = 1; } } if (limit_reached) { /* Reached the limit stop */ break; } jstart = 0; offset += 8; } if (num_gap_blocks == 0) { /* * slide not yet happened, and somehow we got called * to send a sack. Cumack needs to move up. */ int abort_flag = 0; asoc->cumulative_tsn = asoc->highest_tsn_inside_map; sack->sack.cum_tsn_ack = htonl(asoc->cumulative_tsn); sctp_sack_check(stcb, 0, 0, &abort_flag); } } /* now we must add any dups we are going to report. */ if ((limit_reached == 0) && (asoc->numduptsns)) { dup = (uint32_t *) gap_descriptor; for (i = 0; i < asoc->numduptsns; i++) { *dup = htonl(asoc->dup_tsns[i]); dup++; num_dups++; if (((caddr_t)dup + sizeof(uint32_t)) > limit) { /* no more room */ break; } } asoc->numduptsns = 0; } /* * now that the chunk is prepared queue it to the control chunk * queue. */ a_chk->send_size = (sizeof(struct sctp_sack_chunk) + (num_gap_blocks * sizeof(struct sctp_gap_ack_block)) + (num_dups * sizeof(int32_t))); SCTP_BUF_LEN(a_chk->data) = a_chk->send_size; sack->sack.num_gap_ack_blks = htons(num_gap_blocks); sack->sack.num_dup_tsns = htons(num_dups); sack->ch.chunk_length = htons(a_chk->send_size); TAILQ_INSERT_TAIL(&asoc->control_send_queue, a_chk, sctp_next); asoc->ctrl_queue_cnt++; asoc->send_sack = 0; SCTP_STAT_INCR(sctps_sendsacks); return; } void sctp_send_abort_tcb(struct sctp_tcb *stcb, struct mbuf *operr, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct mbuf *m_abort; struct mbuf *m_out = NULL, *m_end = NULL; struct sctp_abort_chunk *abort = NULL; int sz; uint32_t auth_offset = 0; struct sctp_auth_chunk *auth = NULL; struct sctphdr *shdr; /*- * Add an AUTH chunk, if chunk requires it and save the offset into * the chain for AUTH */ if (sctp_auth_is_required_chunk(SCTP_ABORT_ASSOCIATION, stcb->asoc.peer_auth_chunks)) { m_out = sctp_add_auth_chunk(m_out, &m_end, &auth, &auth_offset, stcb, SCTP_ABORT_ASSOCIATION); } SCTP_TCB_LOCK_ASSERT(stcb); m_abort = sctp_get_mbuf_for_msg(sizeof(struct sctp_abort_chunk), 0, M_DONTWAIT, 1, MT_HEADER); if (m_abort == NULL) { /* no mbuf's */ if (m_out) sctp_m_freem(m_out); return; } /* link in any error */ SCTP_BUF_NEXT(m_abort) = operr; sz = 0; if (operr) { struct mbuf *n; n = operr; while (n) { sz += SCTP_BUF_LEN(n); n = SCTP_BUF_NEXT(n); } } SCTP_BUF_LEN(m_abort) = sizeof(*abort); if (m_out == NULL) { /* NO Auth chunk prepended, so reserve space in front */ SCTP_BUF_RESV_UF(m_abort, SCTP_MIN_OVERHEAD); m_out = m_abort; } else { /* Put AUTH chunk at the front of the chain */ SCTP_BUF_NEXT(m_end) = m_abort; } /* fill in the ABORT chunk */ abort = mtod(m_abort, struct sctp_abort_chunk *); abort->ch.chunk_type = SCTP_ABORT_ASSOCIATION; abort->ch.chunk_flags = 0; abort->ch.chunk_length = htons(sizeof(*abort) + sz); /* prepend and fill in the SCTP header */ SCTP_BUF_PREPEND(m_out, sizeof(struct sctphdr), M_DONTWAIT); if (m_out == NULL) { /* TSNH: no memory */ return; } shdr = mtod(m_out, struct sctphdr *); shdr->src_port = stcb->sctp_ep->sctp_lport; shdr->dest_port = stcb->rport; shdr->v_tag = htonl(stcb->asoc.peer_vtag); shdr->checksum = 0; auth_offset += sizeof(struct sctphdr); (void)sctp_lowlevel_chunk_output(stcb->sctp_ep, stcb, stcb->asoc.primary_destination, (struct sockaddr *)&stcb->asoc.primary_destination->ro._l_addr, m_out, auth_offset, auth, 1, 0, NULL, 0, stcb->asoc.primary_destination->port, so_locked, NULL); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); } void sctp_send_shutdown_complete(struct sctp_tcb *stcb, struct sctp_nets *net) { /* formulate and SEND a SHUTDOWN-COMPLETE */ struct mbuf *m_shutdown_comp; struct sctp_shutdown_complete_msg *comp_cp; m_shutdown_comp = sctp_get_mbuf_for_msg(sizeof(struct sctp_shutdown_complete_msg), 0, M_DONTWAIT, 1, MT_HEADER); if (m_shutdown_comp == NULL) { /* no mbuf's */ return; } comp_cp = mtod(m_shutdown_comp, struct sctp_shutdown_complete_msg *); comp_cp->shut_cmp.ch.chunk_type = SCTP_SHUTDOWN_COMPLETE; comp_cp->shut_cmp.ch.chunk_flags = 0; comp_cp->shut_cmp.ch.chunk_length = htons(sizeof(struct sctp_shutdown_complete_chunk)); comp_cp->sh.src_port = stcb->sctp_ep->sctp_lport; comp_cp->sh.dest_port = stcb->rport; comp_cp->sh.v_tag = htonl(stcb->asoc.peer_vtag); comp_cp->sh.checksum = 0; SCTP_BUF_LEN(m_shutdown_comp) = sizeof(struct sctp_shutdown_complete_msg); (void)sctp_lowlevel_chunk_output(stcb->sctp_ep, stcb, net, (struct sockaddr *)&net->ro._l_addr, m_shutdown_comp, 0, NULL, 1, 0, NULL, 0, net->port, SCTP_SO_NOT_LOCKED, NULL); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); return; } void sctp_send_shutdown_complete2(struct mbuf *m, int iphlen, struct sctphdr *sh, uint32_t vrf_id, uint16_t port) { /* formulate and SEND a SHUTDOWN-COMPLETE */ struct mbuf *o_pak; struct mbuf *mout; struct ip *iph, *iph_out; struct udphdr *udp = NULL; #ifdef INET6 struct ip6_hdr *ip6, *ip6_out; #endif int offset_out, len, mlen; struct sctp_shutdown_complete_msg *comp_cp; iph = mtod(m, struct ip *); switch (iph->ip_v) { case IPVERSION: len = (sizeof(struct ip) + sizeof(struct sctp_shutdown_complete_msg)); break; #ifdef INET6 case IPV6_VERSION >> 4: len = (sizeof(struct ip6_hdr) + sizeof(struct sctp_shutdown_complete_msg)); break; #endif default: return; } if (port) { len += sizeof(struct udphdr); } mout = sctp_get_mbuf_for_msg(len, 1, M_DONTWAIT, 1, MT_DATA); if (mout == NULL) { return; } SCTP_BUF_LEN(mout) = len; SCTP_BUF_NEXT(mout) = NULL; iph_out = NULL; #ifdef INET6 ip6_out = NULL; #endif offset_out = 0; switch (iph->ip_v) { case IPVERSION: iph_out = mtod(mout, struct ip *); /* Fill in the IP header for the ABORT */ iph_out->ip_v = IPVERSION; iph_out->ip_hl = (sizeof(struct ip) / 4); iph_out->ip_tos = (u_char)0; iph_out->ip_id = 0; iph_out->ip_off = 0; iph_out->ip_ttl = MAXTTL; if (port) { iph_out->ip_p = IPPROTO_UDP; } else { iph_out->ip_p = IPPROTO_SCTP; } iph_out->ip_src.s_addr = iph->ip_dst.s_addr; iph_out->ip_dst.s_addr = iph->ip_src.s_addr; /* let IP layer calculate this */ iph_out->ip_sum = 0; offset_out += sizeof(*iph_out); comp_cp = (struct sctp_shutdown_complete_msg *)( (caddr_t)iph_out + offset_out); break; #ifdef INET6 case IPV6_VERSION >> 4: ip6 = (struct ip6_hdr *)iph; ip6_out = mtod(mout, struct ip6_hdr *); /* Fill in the IPv6 header for the ABORT */ ip6_out->ip6_flow = ip6->ip6_flow; ip6_out->ip6_hlim = MODULE_GLOBAL(MOD_INET6, ip6_defhlim); if (port) { ip6_out->ip6_nxt = IPPROTO_UDP; } else { ip6_out->ip6_nxt = IPPROTO_SCTP; } ip6_out->ip6_src = ip6->ip6_dst; ip6_out->ip6_dst = ip6->ip6_src; /* * ?? The old code had both the iph len + payload, I think * this is wrong and would never have worked */ ip6_out->ip6_plen = sizeof(struct sctp_shutdown_complete_msg); offset_out += sizeof(*ip6_out); comp_cp = (struct sctp_shutdown_complete_msg *)( (caddr_t)ip6_out + offset_out); break; #endif /* INET6 */ default: /* Currently not supported. */ return; } if (port) { udp = (struct udphdr *)comp_cp; udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); udp->uh_dport = port; udp->uh_ulen = htons(sizeof(struct sctp_shutdown_complete_msg) + sizeof(struct udphdr)); udp->uh_sum = in_pseudo(iph_out->ip_src.s_addr, iph_out->ip_dst.s_addr, udp->uh_ulen + htons(IPPROTO_UDP)); offset_out += sizeof(struct udphdr); comp_cp = (struct sctp_shutdown_complete_msg *)((caddr_t)comp_cp + sizeof(struct udphdr)); } if (SCTP_GET_HEADER_FOR_OUTPUT(o_pak)) { /* no mbuf's */ sctp_m_freem(mout); return; } /* Now copy in and fill in the ABORT tags etc. */ comp_cp->sh.src_port = sh->dest_port; comp_cp->sh.dest_port = sh->src_port; comp_cp->sh.checksum = 0; comp_cp->sh.v_tag = sh->v_tag; comp_cp->shut_cmp.ch.chunk_flags = SCTP_HAD_NO_TCB; comp_cp->shut_cmp.ch.chunk_type = SCTP_SHUTDOWN_COMPLETE; comp_cp->shut_cmp.ch.chunk_length = htons(sizeof(struct sctp_shutdown_complete_chunk)); /* add checksum */ comp_cp->sh.checksum = sctp_calculate_sum(mout, NULL, offset_out); if (iph_out != NULL) { sctp_route_t ro; int ret; struct sctp_tcb *stcb = NULL; mlen = SCTP_BUF_LEN(mout); bzero(&ro, sizeof ro); /* set IPv4 length */ iph_out->ip_len = mlen; #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(mout, mlen); #endif SCTP_ATTACH_CHAIN(o_pak, mout, mlen); if (port) { SCTP_ENABLE_UDP_CSUM(o_pak); } /* out it goes */ SCTP_IP_OUTPUT(ret, o_pak, &ro, stcb, vrf_id); /* Free the route if we got one back */ if (ro.ro_rt) RTFREE(ro.ro_rt); } #ifdef INET6 if (ip6_out != NULL) { struct route_in6 ro; int ret; struct sctp_tcb *stcb = NULL; struct ifnet *ifp = NULL; bzero(&ro, sizeof(ro)); mlen = SCTP_BUF_LEN(mout); #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(mout, mlen); #endif SCTP_ATTACH_CHAIN(o_pak, mout, mlen); if (port) { if ((udp->uh_sum = in6_cksum(o_pak, IPPROTO_UDP, sizeof(struct ip6_hdr), sizeof(struct sctp_shutdown_complete_msg) + sizeof(struct udphdr))) == 0) { udp->uh_sum = 0xffff; } } SCTP_IP6_OUTPUT(ret, o_pak, &ro, &ifp, stcb, vrf_id); /* Free the route if we got one back */ if (ro.ro_rt) RTFREE(ro.ro_rt); } #endif SCTP_STAT_INCR(sctps_sendpackets); SCTP_STAT_INCR_COUNTER64(sctps_outpackets); SCTP_STAT_INCR_COUNTER64(sctps_outcontrolchunks); return; } static struct sctp_nets * sctp_select_hb_destination(struct sctp_tcb *stcb, struct timeval *now) { struct sctp_nets *net, *hnet; int ms_goneby, highest_ms, state_overide = 0; (void)SCTP_GETTIME_TIMEVAL(now); highest_ms = 0; hnet = NULL; SCTP_TCB_LOCK_ASSERT(stcb); TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if ( ((net->dest_state & SCTP_ADDR_NOHB) && ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) || (net->dest_state & SCTP_ADDR_OUT_OF_SCOPE) ) { /* * Skip this guy from consideration if HB is off AND * its confirmed */ continue; } if (sctp_destination_is_reachable(stcb, (struct sockaddr *)&net->ro._l_addr) == 0) { /* skip this dest net from consideration */ continue; } if (net->last_sent_time.tv_sec) { /* Sent to so we subtract */ ms_goneby = (now->tv_sec - net->last_sent_time.tv_sec) * 1000; } else /* Never been sent to */ ms_goneby = 0x7fffffff; /*- * When the address state is unconfirmed but still * considered reachable, we HB at a higher rate. Once it * goes confirmed OR reaches the "unreachable" state, thenw * we cut it back to HB at a more normal pace. */ if ((net->dest_state & (SCTP_ADDR_UNCONFIRMED | SCTP_ADDR_NOT_REACHABLE)) == SCTP_ADDR_UNCONFIRMED) { state_overide = 1; } else { state_overide = 0; } if ((((unsigned int)ms_goneby >= net->RTO) || (state_overide)) && (ms_goneby > highest_ms)) { highest_ms = ms_goneby; hnet = net; } } if (hnet && ((hnet->dest_state & (SCTP_ADDR_UNCONFIRMED | SCTP_ADDR_NOT_REACHABLE)) == SCTP_ADDR_UNCONFIRMED)) { state_overide = 1; } else { state_overide = 0; } if (hnet && highest_ms && (((unsigned int)highest_ms >= hnet->RTO) || state_overide)) { /*- * Found the one with longest delay bounds OR it is * unconfirmed and still not marked unreachable. */ SCTPDBG(SCTP_DEBUG_OUTPUT4, "net:%p is the hb winner -", hnet); #ifdef SCTP_DEBUG if (hnet) { SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT4, (struct sockaddr *)&hnet->ro._l_addr); } else { SCTPDBG(SCTP_DEBUG_OUTPUT4, " none\n"); } #endif /* update the timer now */ hnet->last_sent_time = *now; return (hnet); } /* Nothing to HB */ return (NULL); } int sctp_send_hb(struct sctp_tcb *stcb, int user_req, struct sctp_nets *u_net) { struct sctp_tmit_chunk *chk; struct sctp_nets *net; struct sctp_heartbeat_chunk *hb; struct timeval now; struct sockaddr_in *sin; struct sockaddr_in6 *sin6; SCTP_TCB_LOCK_ASSERT(stcb); if (user_req == 0) { net = sctp_select_hb_destination(stcb, &now); if (net == NULL) { /*- * All our busy none to send to, just start the * timer again. */ if (stcb->asoc.state == 0) { return (0); } sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); return (0); } } else { net = u_net; if (net == NULL) { return (0); } (void)SCTP_GETTIME_TIMEVAL(&now); } sin = (struct sockaddr_in *)&net->ro._l_addr; if (sin->sin_family != AF_INET) { if (sin->sin_family != AF_INET6) { /* huh */ return (0); } } sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT4, "Gak, can't get a chunk for hb\n"); return (0); } chk->copy_by_ref = 0; chk->rec.chunk_id.id = SCTP_HEARTBEAT_REQUEST; chk->rec.chunk_id.can_take_data = 1; chk->asoc = &stcb->asoc; chk->send_size = sizeof(struct sctp_heartbeat_chunk); chk->data = sctp_get_mbuf_for_msg(chk->send_size, 0, M_DONTWAIT, 1, MT_HEADER); if (chk->data == NULL) { sctp_free_a_chunk(stcb, chk); return (0); } SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); SCTP_BUF_LEN(chk->data) = chk->send_size; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->whoTo = net; atomic_add_int(&chk->whoTo->ref_count, 1); /* Now we have a mbuf that we can fill in with the details */ hb = mtod(chk->data, struct sctp_heartbeat_chunk *); memset(hb, 0, sizeof(struct sctp_heartbeat_chunk)); /* fill out chunk header */ hb->ch.chunk_type = SCTP_HEARTBEAT_REQUEST; hb->ch.chunk_flags = 0; hb->ch.chunk_length = htons(chk->send_size); /* Fill out hb parameter */ hb->heartbeat.hb_info.ph.param_type = htons(SCTP_HEARTBEAT_INFO); hb->heartbeat.hb_info.ph.param_length = htons(sizeof(struct sctp_heartbeat_info_param)); hb->heartbeat.hb_info.time_value_1 = now.tv_sec; hb->heartbeat.hb_info.time_value_2 = now.tv_usec; /* Did our user request this one, put it in */ hb->heartbeat.hb_info.user_req = user_req; hb->heartbeat.hb_info.addr_family = sin->sin_family; hb->heartbeat.hb_info.addr_len = sin->sin_len; if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { /* * we only take from the entropy pool if the address is not * confirmed. */ net->heartbeat_random1 = hb->heartbeat.hb_info.random_value1 = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep); net->heartbeat_random2 = hb->heartbeat.hb_info.random_value2 = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep); } else { net->heartbeat_random1 = hb->heartbeat.hb_info.random_value1 = 0; net->heartbeat_random2 = hb->heartbeat.hb_info.random_value2 = 0; } if (sin->sin_family == AF_INET) { memcpy(hb->heartbeat.hb_info.address, &sin->sin_addr, sizeof(sin->sin_addr)); } else if (sin->sin_family == AF_INET6) { /* We leave the scope the way it is in our lookup table. */ sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; memcpy(hb->heartbeat.hb_info.address, &sin6->sin6_addr, sizeof(sin6->sin6_addr)); } else { /* huh compiler bug */ return (0); } /* * JRS 5/14/07 - In CMT PF, the T3 timer is used to track * PF-heartbeats. Because of this, threshold management is done by * the t3 timer handler, and does not need to be done upon the send * of a PF-heartbeat. If CMT PF is on and the destination to which a * heartbeat is being sent is in PF state, do NOT do threshold * management. */ if ((SCTP_BASE_SYSCTL(sctp_cmt_pf) == 0) || ((net->dest_state & SCTP_ADDR_PF) != SCTP_ADDR_PF)) { /* ok we have a destination that needs a beat */ /* lets do the theshold management Qiaobing style */ if (sctp_threshold_management(stcb->sctp_ep, stcb, net, stcb->asoc.max_send_times)) { /*- * we have lost the association, in a way this is * quite bad since we really are one less time since * we really did not send yet. This is the down side * to the Q's style as defined in the RFC and not my * alternate style defined in the RFC. */ if (chk->data != NULL) { sctp_m_freem(chk->data); chk->data = NULL; } /* * Here we do NOT use the macro since the * association is now gone. */ if (chk->whoTo) { sctp_free_remote_addr(chk->whoTo); chk->whoTo = NULL; } sctp_free_a_chunk((struct sctp_tcb *)NULL, chk); return (-1); } } net->hb_responded = 0; TAILQ_INSERT_TAIL(&stcb->asoc.control_send_queue, chk, sctp_next); stcb->asoc.ctrl_queue_cnt++; SCTP_STAT_INCR(sctps_sendheartbeat); /*- * Call directly med level routine to put out the chunk. It will * always tumble out control chunks aka HB but it may even tumble * out data too. */ return (1); } void sctp_send_ecn_echo(struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t high_tsn) { struct sctp_association *asoc; struct sctp_ecne_chunk *ecne; struct sctp_tmit_chunk *chk; asoc = &stcb->asoc; SCTP_TCB_LOCK_ASSERT(stcb); TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { if (chk->rec.chunk_id.id == SCTP_ECN_ECHO) { /* found a previous ECN_ECHO update it if needed */ ecne = mtod(chk->data, struct sctp_ecne_chunk *); ecne->tsn = htonl(high_tsn); return; } } /* nope could not find one to update so we must build one */ sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { return; } chk->copy_by_ref = 0; SCTP_STAT_INCR(sctps_sendecne); chk->rec.chunk_id.id = SCTP_ECN_ECHO; chk->rec.chunk_id.can_take_data = 0; chk->asoc = &stcb->asoc; chk->send_size = sizeof(struct sctp_ecne_chunk); chk->data = sctp_get_mbuf_for_msg(chk->send_size, 0, M_DONTWAIT, 1, MT_HEADER); if (chk->data == NULL) { sctp_free_a_chunk(stcb, chk); return; } SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); SCTP_BUF_LEN(chk->data) = chk->send_size; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->whoTo = net; atomic_add_int(&chk->whoTo->ref_count, 1); stcb->asoc.ecn_echo_cnt_onq++; ecne = mtod(chk->data, struct sctp_ecne_chunk *); ecne->ch.chunk_type = SCTP_ECN_ECHO; ecne->ch.chunk_flags = 0; ecne->ch.chunk_length = htons(sizeof(struct sctp_ecne_chunk)); ecne->tsn = htonl(high_tsn); TAILQ_INSERT_TAIL(&stcb->asoc.control_send_queue, chk, sctp_next); asoc->ctrl_queue_cnt++; } void sctp_send_packet_dropped(struct sctp_tcb *stcb, struct sctp_nets *net, struct mbuf *m, int iphlen, int bad_crc) { struct sctp_association *asoc; struct sctp_pktdrop_chunk *drp; struct sctp_tmit_chunk *chk; uint8_t *datap; int len; int was_trunc = 0; struct ip *iph; #ifdef INET6 struct ip6_hdr *ip6h; #endif int fullsz = 0, extra = 0; long spc; int offset; struct sctp_chunkhdr *ch, chunk_buf; unsigned int chk_length; if (!stcb) { return; } asoc = &stcb->asoc; SCTP_TCB_LOCK_ASSERT(stcb); if (asoc->peer_supports_pktdrop == 0) { /*- * peer must declare support before I send one. */ return; } if (stcb->sctp_socket == NULL) { return; } sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { return; } chk->copy_by_ref = 0; iph = mtod(m, struct ip *); if (iph == NULL) { sctp_free_a_chunk(stcb, chk); return; } switch (iph->ip_v) { case IPVERSION: /* IPv4 */ len = chk->send_size = iph->ip_len; break; #ifdef INET6 case IPV6_VERSION >> 4: /* IPv6 */ ip6h = mtod(m, struct ip6_hdr *); len = chk->send_size = htons(ip6h->ip6_plen); break; #endif default: return; } /* Validate that we do not have an ABORT in here. */ offset = iphlen + sizeof(struct sctphdr); ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, sizeof(*ch), (uint8_t *) & chunk_buf); while (ch != NULL) { chk_length = ntohs(ch->chunk_length); if (chk_length < sizeof(*ch)) { /* break to abort land */ break; } switch (ch->chunk_type) { case SCTP_PACKET_DROPPED: case SCTP_ABORT_ASSOCIATION: /*- * we don't respond with an PKT-DROP to an ABORT * or PKT-DROP */ sctp_free_a_chunk(stcb, chk); return; default: break; } offset += SCTP_SIZE32(chk_length); ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, sizeof(*ch), (uint8_t *) & chunk_buf); } if ((len + SCTP_MAX_OVERHEAD + sizeof(struct sctp_pktdrop_chunk)) > min(stcb->asoc.smallest_mtu, MCLBYTES)) { /* * only send 1 mtu worth, trim off the excess on the end. */ fullsz = len - extra; len = min(stcb->asoc.smallest_mtu, MCLBYTES) - SCTP_MAX_OVERHEAD; was_trunc = 1; } chk->asoc = &stcb->asoc; chk->data = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_DATA); if (chk->data == NULL) { jump_out: sctp_free_a_chunk(stcb, chk); return; } SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); drp = mtod(chk->data, struct sctp_pktdrop_chunk *); if (drp == NULL) { sctp_m_freem(chk->data); chk->data = NULL; goto jump_out; } chk->book_size = SCTP_SIZE32((chk->send_size + sizeof(struct sctp_pktdrop_chunk) + sizeof(struct sctphdr) + SCTP_MED_OVERHEAD)); chk->book_size_scale = 0; if (was_trunc) { drp->ch.chunk_flags = SCTP_PACKET_TRUNCATED; drp->trunc_len = htons(fullsz); /* * Len is already adjusted to size minus overhead above take * out the pkt_drop chunk itself from it. */ chk->send_size = len - sizeof(struct sctp_pktdrop_chunk); len = chk->send_size; } else { /* no truncation needed */ drp->ch.chunk_flags = 0; drp->trunc_len = htons(0); } if (bad_crc) { drp->ch.chunk_flags |= SCTP_BADCRC; } chk->send_size += sizeof(struct sctp_pktdrop_chunk); SCTP_BUF_LEN(chk->data) = chk->send_size; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; if (net) { /* we should hit here */ chk->whoTo = net; } else { chk->whoTo = asoc->primary_destination; } atomic_add_int(&chk->whoTo->ref_count, 1); chk->rec.chunk_id.id = SCTP_PACKET_DROPPED; chk->rec.chunk_id.can_take_data = 1; drp->ch.chunk_type = SCTP_PACKET_DROPPED; drp->ch.chunk_length = htons(chk->send_size); spc = SCTP_SB_LIMIT_RCV(stcb->sctp_socket); if (spc < 0) { spc = 0; } drp->bottle_bw = htonl(spc); if (asoc->my_rwnd) { drp->current_onq = htonl(asoc->size_on_reasm_queue + asoc->size_on_all_streams + asoc->my_rwnd_control_len + stcb->sctp_socket->so_rcv.sb_cc); } else { /*- * If my rwnd is 0, possibly from mbuf depletion as well as * space used, tell the peer there is NO space aka onq == bw */ drp->current_onq = htonl(spc); } drp->reserved = 0; datap = drp->data; m_copydata(m, iphlen, len, (caddr_t)datap); TAILQ_INSERT_TAIL(&stcb->asoc.control_send_queue, chk, sctp_next); asoc->ctrl_queue_cnt++; } void sctp_send_cwr(struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t high_tsn) { struct sctp_association *asoc; struct sctp_cwr_chunk *cwr; struct sctp_tmit_chunk *chk; asoc = &stcb->asoc; SCTP_TCB_LOCK_ASSERT(stcb); TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) { if (chk->rec.chunk_id.id == SCTP_ECN_CWR) { /* found a previous ECN_CWR update it if needed */ cwr = mtod(chk->data, struct sctp_cwr_chunk *); if (compare_with_wrap(high_tsn, ntohl(cwr->tsn), MAX_TSN)) { cwr->tsn = htonl(high_tsn); } return; } } /* nope could not find one to update so we must build one */ sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { return; } chk->copy_by_ref = 0; chk->rec.chunk_id.id = SCTP_ECN_CWR; chk->rec.chunk_id.can_take_data = 1; chk->asoc = &stcb->asoc; chk->send_size = sizeof(struct sctp_cwr_chunk); chk->data = sctp_get_mbuf_for_msg(chk->send_size, 0, M_DONTWAIT, 1, MT_HEADER); if (chk->data == NULL) { sctp_free_a_chunk(stcb, chk); return; } SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); SCTP_BUF_LEN(chk->data) = chk->send_size; chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->whoTo = net; atomic_add_int(&chk->whoTo->ref_count, 1); cwr = mtod(chk->data, struct sctp_cwr_chunk *); cwr->ch.chunk_type = SCTP_ECN_CWR; cwr->ch.chunk_flags = 0; cwr->ch.chunk_length = htons(sizeof(struct sctp_cwr_chunk)); cwr->tsn = htonl(high_tsn); TAILQ_INSERT_TAIL(&stcb->asoc.control_send_queue, chk, sctp_next); asoc->ctrl_queue_cnt++; } void sctp_add_stream_reset_out(struct sctp_tmit_chunk *chk, int number_entries, uint16_t * list, uint32_t seq, uint32_t resp_seq, uint32_t last_sent) { int len, old_len, i; struct sctp_stream_reset_out_request *req_out; struct sctp_chunkhdr *ch; ch = mtod(chk->data, struct sctp_chunkhdr *); old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); /* get to new offset for the param. */ req_out = (struct sctp_stream_reset_out_request *)((caddr_t)ch + len); /* now how long will this param be? */ len = (sizeof(struct sctp_stream_reset_out_request) + (sizeof(uint16_t) * number_entries)); req_out->ph.param_type = htons(SCTP_STR_RESET_OUT_REQUEST); req_out->ph.param_length = htons(len); req_out->request_seq = htonl(seq); req_out->response_seq = htonl(resp_seq); req_out->send_reset_at_tsn = htonl(last_sent); if (number_entries) { for (i = 0; i < number_entries; i++) { req_out->list_of_streams[i] = htons(list[i]); } } if (SCTP_SIZE32(len) > len) { /*- * Need to worry about the pad we may end up adding to the * end. This is easy since the struct is either aligned to 4 * bytes or 2 bytes off. */ req_out->list_of_streams[number_entries] = 0; } /* now fix the chunk length */ ch->chunk_length = htons(len + old_len); chk->book_size = len + old_len; chk->book_size_scale = 0; chk->send_size = SCTP_SIZE32(chk->book_size); SCTP_BUF_LEN(chk->data) = chk->send_size; return; } void sctp_add_stream_reset_in(struct sctp_tmit_chunk *chk, int number_entries, uint16_t * list, uint32_t seq) { int len, old_len, i; struct sctp_stream_reset_in_request *req_in; struct sctp_chunkhdr *ch; ch = mtod(chk->data, struct sctp_chunkhdr *); old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); /* get to new offset for the param. */ req_in = (struct sctp_stream_reset_in_request *)((caddr_t)ch + len); /* now how long will this param be? */ len = (sizeof(struct sctp_stream_reset_in_request) + (sizeof(uint16_t) * number_entries)); req_in->ph.param_type = htons(SCTP_STR_RESET_IN_REQUEST); req_in->ph.param_length = htons(len); req_in->request_seq = htonl(seq); if (number_entries) { for (i = 0; i < number_entries; i++) { req_in->list_of_streams[i] = htons(list[i]); } } if (SCTP_SIZE32(len) > len) { /*- * Need to worry about the pad we may end up adding to the * end. This is easy since the struct is either aligned to 4 * bytes or 2 bytes off. */ req_in->list_of_streams[number_entries] = 0; } /* now fix the chunk length */ ch->chunk_length = htons(len + old_len); chk->book_size = len + old_len; chk->book_size_scale = 0; chk->send_size = SCTP_SIZE32(chk->book_size); SCTP_BUF_LEN(chk->data) = chk->send_size; return; } void sctp_add_stream_reset_tsn(struct sctp_tmit_chunk *chk, uint32_t seq) { int len, old_len; struct sctp_stream_reset_tsn_request *req_tsn; struct sctp_chunkhdr *ch; ch = mtod(chk->data, struct sctp_chunkhdr *); old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); /* get to new offset for the param. */ req_tsn = (struct sctp_stream_reset_tsn_request *)((caddr_t)ch + len); /* now how long will this param be? */ len = sizeof(struct sctp_stream_reset_tsn_request); req_tsn->ph.param_type = htons(SCTP_STR_RESET_TSN_REQUEST); req_tsn->ph.param_length = htons(len); req_tsn->request_seq = htonl(seq); /* now fix the chunk length */ ch->chunk_length = htons(len + old_len); chk->send_size = len + old_len; chk->book_size = SCTP_SIZE32(chk->send_size); chk->book_size_scale = 0; SCTP_BUF_LEN(chk->data) = SCTP_SIZE32(chk->send_size); return; } void sctp_add_stream_reset_result(struct sctp_tmit_chunk *chk, uint32_t resp_seq, uint32_t result) { int len, old_len; struct sctp_stream_reset_response *resp; struct sctp_chunkhdr *ch; ch = mtod(chk->data, struct sctp_chunkhdr *); old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); /* get to new offset for the param. */ resp = (struct sctp_stream_reset_response *)((caddr_t)ch + len); /* now how long will this param be? */ len = sizeof(struct sctp_stream_reset_response); resp->ph.param_type = htons(SCTP_STR_RESET_RESPONSE); resp->ph.param_length = htons(len); resp->response_seq = htonl(resp_seq); resp->result = ntohl(result); /* now fix the chunk length */ ch->chunk_length = htons(len + old_len); chk->book_size = len + old_len; chk->book_size_scale = 0; chk->send_size = SCTP_SIZE32(chk->book_size); SCTP_BUF_LEN(chk->data) = chk->send_size; return; } void sctp_add_stream_reset_result_tsn(struct sctp_tmit_chunk *chk, uint32_t resp_seq, uint32_t result, uint32_t send_una, uint32_t recv_next) { int len, old_len; struct sctp_stream_reset_response_tsn *resp; struct sctp_chunkhdr *ch; ch = mtod(chk->data, struct sctp_chunkhdr *); old_len = len = SCTP_SIZE32(ntohs(ch->chunk_length)); /* get to new offset for the param. */ resp = (struct sctp_stream_reset_response_tsn *)((caddr_t)ch + len); /* now how long will this param be? */ len = sizeof(struct sctp_stream_reset_response_tsn); resp->ph.param_type = htons(SCTP_STR_RESET_RESPONSE); resp->ph.param_length = htons(len); resp->response_seq = htonl(resp_seq); resp->result = htonl(result); resp->senders_next_tsn = htonl(send_una); resp->receivers_next_tsn = htonl(recv_next); /* now fix the chunk length */ ch->chunk_length = htons(len + old_len); chk->book_size = len + old_len; chk->send_size = SCTP_SIZE32(chk->book_size); chk->book_size_scale = 0; SCTP_BUF_LEN(chk->data) = chk->send_size; return; } int sctp_send_str_reset_req(struct sctp_tcb *stcb, int number_entries, uint16_t * list, uint8_t send_out_req, uint32_t resp_seq, uint8_t send_in_req, uint8_t send_tsn_req) { struct sctp_association *asoc; struct sctp_tmit_chunk *chk; struct sctp_chunkhdr *ch; uint32_t seq; asoc = &stcb->asoc; if (asoc->stream_reset_outstanding) { /*- * Already one pending, must get ACK back to clear the flag. */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EBUSY); return (EBUSY); } if ((send_out_req == 0) && (send_in_req == 0) && (send_tsn_req == 0)) { /* nothing to do */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } if (send_tsn_req && (send_out_req || send_in_req)) { /* error, can't do that */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } sctp_alloc_a_chunk(stcb, chk); if (chk == NULL) { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } chk->copy_by_ref = 0; chk->rec.chunk_id.id = SCTP_STREAM_RESET; chk->rec.chunk_id.can_take_data = 0; chk->asoc = &stcb->asoc; chk->book_size = sizeof(struct sctp_chunkhdr); chk->send_size = SCTP_SIZE32(chk->book_size); chk->book_size_scale = 0; chk->data = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_DATA); if (chk->data == NULL) { sctp_free_a_chunk(stcb, chk); SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } SCTP_BUF_RESV_UF(chk->data, SCTP_MIN_OVERHEAD); /* setup chunk parameters */ chk->sent = SCTP_DATAGRAM_UNSENT; chk->snd_count = 0; chk->whoTo = asoc->primary_destination; atomic_add_int(&chk->whoTo->ref_count, 1); ch = mtod(chk->data, struct sctp_chunkhdr *); ch->chunk_type = SCTP_STREAM_RESET; ch->chunk_flags = 0; ch->chunk_length = htons(chk->book_size); SCTP_BUF_LEN(chk->data) = chk->send_size; seq = stcb->asoc.str_reset_seq_out; if (send_out_req) { sctp_add_stream_reset_out(chk, number_entries, list, seq, resp_seq, (stcb->asoc.sending_seq - 1)); asoc->stream_reset_out_is_outstanding = 1; seq++; asoc->stream_reset_outstanding++; } if (send_in_req) { sctp_add_stream_reset_in(chk, number_entries, list, seq); asoc->stream_reset_outstanding++; } if (send_tsn_req) { sctp_add_stream_reset_tsn(chk, seq); asoc->stream_reset_outstanding++; } asoc->str_reset = chk; /* insert the chunk for sending */ TAILQ_INSERT_TAIL(&asoc->control_send_queue, chk, sctp_next); asoc->ctrl_queue_cnt++; sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, stcb->sctp_ep, stcb, chk->whoTo); return (0); } void sctp_send_abort(struct mbuf *m, int iphlen, struct sctphdr *sh, uint32_t vtag, struct mbuf *err_cause, uint32_t vrf_id, uint16_t port) { /*- * Formulate the abort message, and send it back down. */ struct mbuf *o_pak; struct mbuf *mout; struct sctp_abort_msg *abm; struct ip *iph, *iph_out; struct udphdr *udp; #ifdef INET6 struct ip6_hdr *ip6, *ip6_out; #endif int iphlen_out, len; /* don't respond to ABORT with ABORT */ if (sctp_is_there_an_abort_here(m, iphlen, &vtag)) { if (err_cause) sctp_m_freem(err_cause); return; } iph = mtod(m, struct ip *); switch (iph->ip_v) { case IPVERSION: len = (sizeof(struct ip) + sizeof(struct sctp_abort_msg)); break; #ifdef INET6 case IPV6_VERSION >> 4: len = (sizeof(struct ip6_hdr) + sizeof(struct sctp_abort_msg)); break; #endif default: return; } if (port) { len += sizeof(struct udphdr); } mout = sctp_get_mbuf_for_msg(len, 1, M_DONTWAIT, 1, MT_DATA); if (mout == NULL) { if (err_cause) sctp_m_freem(err_cause); return; } SCTP_BUF_LEN(mout) = len; SCTP_BUF_NEXT(mout) = err_cause; iph_out = NULL; #ifdef INET6 ip6_out = NULL; #endif switch (iph->ip_v) { case IPVERSION: iph_out = mtod(mout, struct ip *); /* Fill in the IP header for the ABORT */ iph_out->ip_v = IPVERSION; iph_out->ip_hl = (sizeof(struct ip) / 4); iph_out->ip_tos = (u_char)0; iph_out->ip_id = 0; iph_out->ip_off = 0; iph_out->ip_ttl = MAXTTL; if (port) { iph_out->ip_p = IPPROTO_UDP; } else { iph_out->ip_p = IPPROTO_SCTP; } iph_out->ip_src.s_addr = iph->ip_dst.s_addr; iph_out->ip_dst.s_addr = iph->ip_src.s_addr; /* let IP layer calculate this */ iph_out->ip_sum = 0; iphlen_out = sizeof(*iph_out); abm = (struct sctp_abort_msg *)((caddr_t)iph_out + iphlen_out); break; #ifdef INET6 case IPV6_VERSION >> 4: ip6 = (struct ip6_hdr *)iph; ip6_out = mtod(mout, struct ip6_hdr *); /* Fill in the IP6 header for the ABORT */ ip6_out->ip6_flow = ip6->ip6_flow; ip6_out->ip6_hlim = MODULE_GLOBAL(MOD_INET6, ip6_defhlim); if (port) { ip6_out->ip6_nxt = IPPROTO_UDP; } else { ip6_out->ip6_nxt = IPPROTO_SCTP; } ip6_out->ip6_src = ip6->ip6_dst; ip6_out->ip6_dst = ip6->ip6_src; iphlen_out = sizeof(*ip6_out); abm = (struct sctp_abort_msg *)((caddr_t)ip6_out + iphlen_out); break; #endif /* INET6 */ default: /* Currently not supported */ if (err_cause) sctp_m_freem(err_cause); sctp_m_freem(mout); return; } udp = (struct udphdr *)abm; if (port) { udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); udp->uh_dport = port; /* set udp->uh_ulen later */ udp->uh_sum = 0; iphlen_out += sizeof(struct udphdr); abm = (struct sctp_abort_msg *)((caddr_t)abm + sizeof(struct udphdr)); } abm->sh.src_port = sh->dest_port; abm->sh.dest_port = sh->src_port; abm->sh.checksum = 0; if (vtag == 0) { abm->sh.v_tag = sh->v_tag; abm->msg.ch.chunk_flags = SCTP_HAD_NO_TCB; } else { abm->sh.v_tag = htonl(vtag); abm->msg.ch.chunk_flags = 0; } abm->msg.ch.chunk_type = SCTP_ABORT_ASSOCIATION; if (err_cause) { struct mbuf *m_tmp = err_cause; int err_len = 0; /* get length of the err_cause chain */ while (m_tmp != NULL) { err_len += SCTP_BUF_LEN(m_tmp); m_tmp = SCTP_BUF_NEXT(m_tmp); } len = SCTP_BUF_LEN(mout) + err_len; if (err_len % 4) { /* need pad at end of chunk */ uint32_t cpthis = 0; int padlen; padlen = 4 - (len % 4); m_copyback(mout, len, padlen, (caddr_t)&cpthis); len += padlen; } abm->msg.ch.chunk_length = htons(sizeof(abm->msg.ch) + err_len); } else { len = SCTP_BUF_LEN(mout); abm->msg.ch.chunk_length = htons(sizeof(abm->msg.ch)); } /* add checksum */ abm->sh.checksum = sctp_calculate_sum(mout, NULL, iphlen_out); if (SCTP_GET_HEADER_FOR_OUTPUT(o_pak)) { /* no mbuf's */ sctp_m_freem(mout); return; } if (iph_out != NULL) { sctp_route_t ro; struct sctp_tcb *stcb = NULL; int ret; /* zap the stack pointer to the route */ bzero(&ro, sizeof ro); if (port) { udp->uh_ulen = htons(len - sizeof(struct ip)); udp->uh_sum = in_pseudo(iph_out->ip_src.s_addr, iph_out->ip_dst.s_addr, udp->uh_ulen + htons(IPPROTO_UDP)); } SCTPDBG(SCTP_DEBUG_OUTPUT2, "sctp_send_abort calling ip_output:\n"); SCTPDBG_PKT(SCTP_DEBUG_OUTPUT2, iph_out, &abm->sh); /* set IPv4 length */ iph_out->ip_len = len; /* out it goes */ #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(mout, len); #endif SCTP_ATTACH_CHAIN(o_pak, mout, len); if (port) { SCTP_ENABLE_UDP_CSUM(o_pak); } SCTP_IP_OUTPUT(ret, o_pak, &ro, stcb, vrf_id); /* Free the route if we got one back */ if (ro.ro_rt) RTFREE(ro.ro_rt); } #ifdef INET6 if (ip6_out != NULL) { struct route_in6 ro; int ret; struct sctp_tcb *stcb = NULL; struct ifnet *ifp = NULL; /* zap the stack pointer to the route */ bzero(&ro, sizeof(ro)); if (port) { udp->uh_ulen = htons(len - sizeof(struct ip6_hdr)); } SCTPDBG(SCTP_DEBUG_OUTPUT2, "sctp_send_abort calling ip6_output:\n"); SCTPDBG_PKT(SCTP_DEBUG_OUTPUT2, (struct ip *)ip6_out, &abm->sh); ip6_out->ip6_plen = len - sizeof(*ip6_out); #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(mout, len); #endif SCTP_ATTACH_CHAIN(o_pak, mout, len); if (port) { if ((udp->uh_sum = in6_cksum(o_pak, IPPROTO_UDP, sizeof(struct ip6_hdr), len - sizeof(struct ip6_hdr))) == 0) { udp->uh_sum = 0xffff; } } SCTP_IP6_OUTPUT(ret, o_pak, &ro, &ifp, stcb, vrf_id); /* Free the route if we got one back */ if (ro.ro_rt) RTFREE(ro.ro_rt); } #endif SCTP_STAT_INCR(sctps_sendpackets); SCTP_STAT_INCR_COUNTER64(sctps_outpackets); } void sctp_send_operr_to(struct mbuf *m, int iphlen, struct mbuf *scm, uint32_t vtag, uint32_t vrf_id, uint16_t port) { struct mbuf *o_pak; struct sctphdr *ihdr; int retcode; struct sctphdr *ohdr; struct sctp_chunkhdr *ophdr; struct ip *iph; struct udphdr *udp = NULL; struct mbuf *mout; #ifdef SCTP_DEBUG struct sockaddr_in6 lsa6, fsa6; #endif uint32_t val; struct mbuf *at; int len; iph = mtod(m, struct ip *); ihdr = (struct sctphdr *)((caddr_t)iph + iphlen); SCTP_BUF_PREPEND(scm, (sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr)), M_DONTWAIT); if (scm == NULL) { /* can't send because we can't add a mbuf */ return; } ohdr = mtod(scm, struct sctphdr *); ohdr->src_port = ihdr->dest_port; ohdr->dest_port = ihdr->src_port; ohdr->v_tag = vtag; ohdr->checksum = 0; ophdr = (struct sctp_chunkhdr *)(ohdr + 1); ophdr->chunk_type = SCTP_OPERATION_ERROR; ophdr->chunk_flags = 0; len = 0; at = scm; while (at) { len += SCTP_BUF_LEN(at); at = SCTP_BUF_NEXT(at); } ophdr->chunk_length = htons(len - sizeof(struct sctphdr)); if (len % 4) { /* need padding */ uint32_t cpthis = 0; int padlen; padlen = 4 - (len % 4); m_copyback(scm, len, padlen, (caddr_t)&cpthis); len += padlen; } val = sctp_calculate_sum(scm, NULL, 0); #ifdef INET6 if (port) { mout = sctp_get_mbuf_for_msg(sizeof(struct ip6_hdr) + sizeof(struct udphdr), 1, M_DONTWAIT, 1, MT_DATA); } else { mout = sctp_get_mbuf_for_msg(sizeof(struct ip6_hdr), 1, M_DONTWAIT, 1, MT_DATA); } #else if (port) { mout = sctp_get_mbuf_for_msg(sizeof(struct ip) + sizeof(struct udphdr), 1, M_DONTWAIT, 1, MT_DATA); } else { mout = sctp_get_mbuf_for_msg(sizeof(struct ip), 1, M_DONTWAIT, 1, MT_DATA); } #endif if (mout == NULL) { sctp_m_freem(scm); return; } SCTP_BUF_NEXT(mout) = scm; if (SCTP_GET_HEADER_FOR_OUTPUT(o_pak)) { sctp_m_freem(mout); return; } ohdr->checksum = val; switch (iph->ip_v) { case IPVERSION: { /* V4 */ struct ip *out; sctp_route_t ro; struct sctp_tcb *stcb = NULL; SCTP_BUF_LEN(mout) = sizeof(struct ip); len += sizeof(struct ip); if (port) { SCTP_BUF_LEN(mout) += sizeof(struct udphdr); len += sizeof(struct udphdr); } bzero(&ro, sizeof ro); out = mtod(mout, struct ip *); out->ip_v = iph->ip_v; out->ip_hl = (sizeof(struct ip) / 4); out->ip_tos = iph->ip_tos; out->ip_id = iph->ip_id; out->ip_off = 0; out->ip_ttl = MAXTTL; if (port) { out->ip_p = IPPROTO_UDP; } else { out->ip_p = IPPROTO_SCTP; } out->ip_sum = 0; out->ip_src = iph->ip_dst; out->ip_dst = iph->ip_src; out->ip_len = len; if (port) { udp = (struct udphdr *)(out + 1); udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); udp->uh_dport = port; udp->uh_ulen = htons(len - sizeof(struct ip)); udp->uh_sum = in_pseudo(out->ip_src.s_addr, out->ip_dst.s_addr, udp->uh_ulen + htons(IPPROTO_UDP)); } #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(mout, len); #endif SCTP_ATTACH_CHAIN(o_pak, mout, len); if (port) { SCTP_ENABLE_UDP_CSUM(o_pak); } SCTP_IP_OUTPUT(retcode, o_pak, &ro, stcb, vrf_id); SCTP_STAT_INCR(sctps_sendpackets); SCTP_STAT_INCR_COUNTER64(sctps_outpackets); /* Free the route if we got one back */ if (ro.ro_rt) RTFREE(ro.ro_rt); break; } #ifdef INET6 case IPV6_VERSION >> 4: { /* V6 */ struct route_in6 ro; int ret; struct sctp_tcb *stcb = NULL; struct ifnet *ifp = NULL; struct ip6_hdr *out6, *in6; SCTP_BUF_LEN(mout) = sizeof(struct ip6_hdr); len += sizeof(struct ip6_hdr); bzero(&ro, sizeof ro); if (port) { SCTP_BUF_LEN(mout) += sizeof(struct udphdr); len += sizeof(struct udphdr); } in6 = mtod(m, struct ip6_hdr *); out6 = mtod(mout, struct ip6_hdr *); out6->ip6_flow = in6->ip6_flow; out6->ip6_hlim = MODULE_GLOBAL(MOD_INET6, ip6_defhlim); if (port) { out6->ip6_nxt = IPPROTO_UDP; } else { out6->ip6_nxt = IPPROTO_SCTP; } out6->ip6_src = in6->ip6_dst; out6->ip6_dst = in6->ip6_src; out6->ip6_plen = len - sizeof(struct ip6_hdr); if (port) { udp = (struct udphdr *)(out6 + 1); udp->uh_sport = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); udp->uh_dport = port; udp->uh_ulen = htons(len - sizeof(struct ip6_hdr)); udp->uh_sum = 0; } #ifdef SCTP_DEBUG bzero(&lsa6, sizeof(lsa6)); lsa6.sin6_len = sizeof(lsa6); lsa6.sin6_family = AF_INET6; lsa6.sin6_addr = out6->ip6_src; bzero(&fsa6, sizeof(fsa6)); fsa6.sin6_len = sizeof(fsa6); fsa6.sin6_family = AF_INET6; fsa6.sin6_addr = out6->ip6_dst; #endif SCTPDBG(SCTP_DEBUG_OUTPUT2, "sctp_operr_to calling ipv6 output:\n"); SCTPDBG(SCTP_DEBUG_OUTPUT2, "src: "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)&lsa6); SCTPDBG(SCTP_DEBUG_OUTPUT2, "dst "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)&fsa6); #ifdef SCTP_PACKET_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LAST_PACKET_TRACING) sctp_packet_log(mout, len); #endif SCTP_ATTACH_CHAIN(o_pak, mout, len); if (port) { if ((udp->uh_sum = in6_cksum(o_pak, IPPROTO_UDP, sizeof(struct ip6_hdr), len - sizeof(struct ip6_hdr))) == 0) { udp->uh_sum = 0xffff; } } SCTP_IP6_OUTPUT(ret, o_pak, &ro, &ifp, stcb, vrf_id); SCTP_STAT_INCR(sctps_sendpackets); SCTP_STAT_INCR_COUNTER64(sctps_outpackets); /* Free the route if we got one back */ if (ro.ro_rt) RTFREE(ro.ro_rt); break; } #endif /* INET6 */ default: /* TSNH */ break; } } static struct mbuf * sctp_copy_resume(struct sctp_stream_queue_pending *sp, struct uio *uio, struct sctp_sndrcvinfo *srcv, int max_send_len, int user_marks_eor, int *error, uint32_t * sndout, struct mbuf **new_tail) { struct mbuf *m; m = m_uiotombuf(uio, M_WAITOK, max_send_len, 0, (M_PKTHDR | (user_marks_eor ? M_EOR : 0))); if (m == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); *error = ENOMEM; } else { *sndout = m_length(m, NULL); *new_tail = m_last(m); } return (m); } static int sctp_copy_one(struct sctp_stream_queue_pending *sp, struct uio *uio, int resv_upfront) { int left; left = sp->length; sp->data = m_uiotombuf(uio, M_WAITOK, sp->length, resv_upfront, 0); if (sp->data == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, ENOMEM); return (ENOMEM); } sp->tail_mbuf = m_last(sp->data); return (0); } static struct sctp_stream_queue_pending * sctp_copy_it_in(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_sndrcvinfo *srcv, struct uio *uio, struct sctp_nets *net, int max_send_len, int user_marks_eor, int *error, int non_blocking) { /*- * This routine must be very careful in its work. Protocol * processing is up and running so care must be taken to spl...() * when you need to do something that may effect the stcb/asoc. The * sb is locked however. When data is copied the protocol processing * should be enabled since this is a slower operation... */ struct sctp_stream_queue_pending *sp = NULL; int resv_in_first; *error = 0; /* Now can we send this? */ if ((SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_SENT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_ACK_SENT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED) || (asoc->state & SCTP_STATE_SHUTDOWN_PENDING)) { /* got data while shutting down */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ECONNRESET); *error = ECONNRESET; goto out_now; } sctp_alloc_a_strmoq(stcb, sp); if (sp == NULL) { SCTP_LTRACE_ERR_RET(NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOMEM); *error = ENOMEM; goto out_now; } sp->act_flags = 0; sp->sender_all_done = 0; sp->sinfo_flags = srcv->sinfo_flags; sp->timetolive = srcv->sinfo_timetolive; sp->ppid = srcv->sinfo_ppid; sp->context = srcv->sinfo_context; sp->strseq = 0; (void)SCTP_GETTIME_TIMEVAL(&sp->ts); sp->stream = srcv->sinfo_stream; sp->length = min(uio->uio_resid, max_send_len); if ((sp->length == (uint32_t) uio->uio_resid) && ((user_marks_eor == 0) || (srcv->sinfo_flags & SCTP_EOF) || (user_marks_eor && (srcv->sinfo_flags & SCTP_EOR)))) { sp->msg_is_complete = 1; } else { sp->msg_is_complete = 0; } sp->sender_all_done = 0; sp->some_taken = 0; sp->put_last_out = 0; resv_in_first = sizeof(struct sctp_data_chunk); sp->data = sp->tail_mbuf = NULL; *error = sctp_copy_one(sp, uio, resv_in_first); if (*error) { sctp_free_a_strmoq(stcb, sp); sp = NULL; } else { if (sp->sinfo_flags & SCTP_ADDR_OVER) { sp->net = net; sp->addr_over = 1; } else { sp->net = asoc->primary_destination; sp->addr_over = 0; } atomic_add_int(&sp->net->ref_count, 1); sctp_set_prsctp_policy(stcb, sp); } out_now: return (sp); } int sctp_sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, struct mbuf *top, struct mbuf *control, int flags, struct thread *p ) { struct sctp_inpcb *inp; int error, use_rcvinfo = 0; struct sctp_sndrcvinfo srcv; + struct sockaddr *addr_to_use; +#ifdef INET6 + struct sockaddr_in sin; + +#endif + inp = (struct sctp_inpcb *)so->so_pcb; if (control) { /* process cmsg snd/rcv info (maybe a assoc-id) */ if (sctp_find_cmsg(SCTP_SNDRCV, (void *)&srcv, control, sizeof(srcv))) { /* got one */ use_rcvinfo = 1; } } - error = sctp_lower_sosend(so, addr, uio, top, + addr_to_use = addr; +#ifdef INET6 + if ((addr) && (addr->sa_family == AF_INET6)) { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)addr; + if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { + in6_sin6_2_sin(&sin, sin6); + addr_to_use = (struct sockaddr *)&sin; + } + } +#endif + error = sctp_lower_sosend(so, addr_to_use, uio, top, control, flags, use_rcvinfo, &srcv ,p ); return (error); } int sctp_lower_sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, struct mbuf *i_pak, struct mbuf *control, int flags, int use_rcvinfo, struct sctp_sndrcvinfo *srcv , struct thread *p ) { unsigned int sndlen = 0, max_len; int error, len; struct mbuf *top = NULL; int queue_only = 0, queue_only_for_init = 0; int free_cnt_applied = 0; int un_sent = 0; int now_filled = 0; unsigned int inqueue_bytes = 0; struct sctp_block_entry be; struct sctp_inpcb *inp; struct sctp_tcb *stcb = NULL; struct timeval now; struct sctp_nets *net; struct sctp_association *asoc; struct sctp_inpcb *t_inp; int user_marks_eor; int create_lock_applied = 0; int nagle_applies = 0; int some_on_control = 0; int got_all_of_the_send = 0; int hold_tcblock = 0; int non_blocking = 0; int temp_flags = 0; uint32_t local_add_more, local_soresv = 0; error = 0; net = NULL; stcb = NULL; asoc = NULL; t_inp = inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EFAULT); error = EFAULT; if (i_pak) { SCTP_RELEASE_PKT(i_pak); } return (error); } if ((uio == NULL) && (i_pak == NULL)) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } user_marks_eor = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR); atomic_add_int(&inp->total_sends, 1); if (uio) { if (uio->uio_resid < 0) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); return (EINVAL); } sndlen = uio->uio_resid; } else { top = SCTP_HEADER_TO_CHAIN(i_pak); sndlen = SCTP_HEADER_LEN(i_pak); } SCTPDBG(SCTP_DEBUG_OUTPUT1, "Send called addr:%p send length %d\n", addr, sndlen); /*- * Pre-screen address, if one is given the sin-len * must be set correctly! */ if (addr) { if ((addr->sa_family == AF_INET) && (addr->sa_len != sizeof(struct sockaddr_in))) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } else if ((addr->sa_family == AF_INET6) && (addr->sa_len != sizeof(struct sockaddr_in6))) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } } hold_tcblock = 0; if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && (inp->sctp_socket->so_qlimit)) { /* The listener can NOT send */ SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EFAULT); error = EFAULT; goto out_unlocked; } if ((use_rcvinfo) && srcv) { if (INVALID_SINFO_FLAG(srcv->sinfo_flags) || PR_SCTP_INVALID_POLICY(srcv->sinfo_flags)) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } if (srcv->sinfo_flags) SCTP_STAT_INCR(sctps_sends_with_flags); if (srcv->sinfo_flags & SCTP_SENDALL) { /* its a sendall */ error = sctp_sendall(inp, uio, top, srcv); top = NULL; goto out_unlocked; } } /* now we must find the assoc */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { SCTP_INP_RLOCK(inp); stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOTCONN); error = ENOTCONN; goto out_unlocked; } hold_tcblock = 0; SCTP_INP_RUNLOCK(inp); if (addr) { /* Must locate the net structure if addr given */ net = sctp_findnet(stcb, addr); if (net) { /* validate port was 0 or correct */ struct sockaddr_in *sin; sin = (struct sockaddr_in *)addr; if ((sin->sin_port != 0) && (sin->sin_port != stcb->rport)) { net = NULL; } } temp_flags |= SCTP_ADDR_OVER; } else net = stcb->asoc.primary_destination; if (addr && (net == NULL)) { /* Could not find address, was it legal */ if (addr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)addr; if (sin->sin_addr.s_addr == 0) { if ((sin->sin_port == 0) || (sin->sin_port == stcb->rport)) { net = stcb->asoc.primary_destination; } } } else { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)addr; if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { if ((sin6->sin6_port == 0) || (sin6->sin6_port == stcb->rport)) { net = stcb->asoc.primary_destination; } } } } if (net == NULL) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } } else if (use_rcvinfo && srcv && srcv->sinfo_assoc_id) { stcb = sctp_findassociation_ep_asocid(inp, srcv->sinfo_assoc_id, 0); if (stcb) { if (addr) /* * Must locate the net structure if addr * given */ net = sctp_findnet(stcb, addr); else net = stcb->asoc.primary_destination; if ((srcv->sinfo_flags & SCTP_ADDR_OVER) && ((net == NULL) || (addr == NULL))) { struct sockaddr_in *sin; if (addr == NULL) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } sin = (struct sockaddr_in *)addr; /* Validate port is 0 or correct */ if ((sin->sin_port != 0) && (sin->sin_port != stcb->rport)) { net = NULL; } } } hold_tcblock = 0; } else if (addr) { /*- * Since we did not use findep we must * increment it, and if we don't find a tcb * decrement it. */ SCTP_INP_WLOCK(inp); SCTP_INP_INCR_REF(inp); SCTP_INP_WUNLOCK(inp); stcb = sctp_findassociation_ep_addr(&t_inp, addr, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_WLOCK(inp); SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); } else { hold_tcblock = 1; } } if ((stcb == NULL) && (addr)) { /* Possible implicit send? */ SCTP_ASOC_CREATE_LOCK(inp); create_lock_applied = 1; if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { /* Should I really unlock ? */ SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_OUTPUT, EFAULT); error = EFAULT; goto out_unlocked; } if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) && (addr->sa_family == AF_INET6)) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } SCTP_INP_WLOCK(inp); SCTP_INP_INCR_REF(inp); SCTP_INP_WUNLOCK(inp); /* With the lock applied look again */ stcb = sctp_findassociation_ep_addr(&t_inp, addr, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_WLOCK(inp); SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); } else { hold_tcblock = 1; } if (t_inp != inp) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOTCONN); error = ENOTCONN; goto out_unlocked; } } if (stcb == NULL) { if (addr == NULL) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOENT); error = ENOENT; goto out_unlocked; } else { /* * UDP style, we must go ahead and start the INIT * process */ uint32_t vrf_id; if ((use_rcvinfo) && (srcv) && ((srcv->sinfo_flags & SCTP_ABORT) || ((srcv->sinfo_flags & SCTP_EOF) && (sndlen == 0)))) { /*- * User asks to abort a non-existant assoc, * or EOF a non-existant assoc with no data */ SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOENT); error = ENOENT; goto out_unlocked; } /* get an asoc/stcb struct */ vrf_id = inp->def_vrf_id; #ifdef INVARIANTS if (create_lock_applied == 0) { panic("Error, should hold create lock and I don't?"); } #endif stcb = sctp_aloc_assoc(inp, addr, 1, &error, 0, vrf_id, p ); if (stcb == NULL) { /* Error is setup for us in the call */ goto out_unlocked; } if (create_lock_applied) { SCTP_ASOC_CREATE_UNLOCK(inp); create_lock_applied = 0; } else { SCTP_PRINTF("Huh-3? create lock should have been on??\n"); } /* * Turn on queue only flag to prevent data from * being sent */ queue_only = 1; asoc = &stcb->asoc; SCTP_SET_STATE(asoc, SCTP_STATE_COOKIE_WAIT); (void)SCTP_GETTIME_TIMEVAL(&asoc->time_entered); /* initialize authentication params for the assoc */ sctp_initialize_auth_params(inp, stcb); if (control) { /* * see if a init structure exists in cmsg * headers */ struct sctp_initmsg initm; int i; if (sctp_find_cmsg(SCTP_INIT, (void *)&initm, control, sizeof(initm))) { /* * we have an INIT override of the * default */ if (initm.sinit_max_attempts) asoc->max_init_times = initm.sinit_max_attempts; if (initm.sinit_num_ostreams) asoc->pre_open_streams = initm.sinit_num_ostreams; if (initm.sinit_max_instreams) asoc->max_inbound_streams = initm.sinit_max_instreams; if (initm.sinit_max_init_timeo) asoc->initial_init_rto_max = initm.sinit_max_init_timeo; if (asoc->streamoutcnt < asoc->pre_open_streams) { /* Default is NOT correct */ SCTPDBG(SCTP_DEBUG_OUTPUT1, "Ok, defout:%d pre_open:%d\n", asoc->streamoutcnt, asoc->pre_open_streams); /* * What happens if this * fails? we panic ... */ { struct sctp_stream_out *tmp_str; int had_lock = 0; if (hold_tcblock) { had_lock = 1; SCTP_TCB_UNLOCK(stcb); } SCTP_MALLOC(tmp_str, struct sctp_stream_out *, (asoc->pre_open_streams * sizeof(struct sctp_stream_out)), SCTP_M_STRMO); if (had_lock) { SCTP_TCB_LOCK(stcb); } if (tmp_str != NULL) { SCTP_FREE(asoc->strmout, SCTP_M_STRMO); asoc->strmout = tmp_str; asoc->streamoutcnt = asoc->pre_open_streams; } else { asoc->pre_open_streams = asoc->streamoutcnt; } } for (i = 0; i < asoc->streamoutcnt; i++) { /*- * inbound side must be set * to 0xffff, also NOTE when * we get the INIT-ACK back * (for INIT sender) we MUST * reduce the count * (streamoutcnt) but first * check if we sent to any * of the upper streams that * were dropped (if some * were). Those that were * dropped must be notified * to the upper layer as * failed to send. */ asoc->strmout[i].next_sequence_sent = 0x0; TAILQ_INIT(&asoc->strmout[i].outqueue); asoc->strmout[i].stream_no = i; asoc->strmout[i].last_msg_incomplete = 0; asoc->strmout[i].next_spoke.tqe_next = 0; asoc->strmout[i].next_spoke.tqe_prev = 0; } } } } hold_tcblock = 1; /* out with the INIT */ queue_only_for_init = 1; /*- * we may want to dig in after this call and adjust the MTU * value. It defaulted to 1500 (constant) but the ro * structure may now have an update and thus we may need to * change it BEFORE we append the message. */ net = stcb->asoc.primary_destination; asoc = &stcb->asoc; } } if ((SCTP_SO_IS_NBIO(so) || (flags & MSG_NBIO) )) { non_blocking = 1; } asoc = &stcb->asoc; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NO_FRAGMENT)) { if (sndlen > asoc->smallest_mtu) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EMSGSIZE); error = EMSGSIZE; goto out_unlocked; } } /* would we block? */ if (non_blocking) { if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); if ((SCTP_SB_LIMIT_SND(so) < (sndlen + inqueue_bytes + stcb->asoc.sb_send_resv)) || (stcb->asoc.chunks_on_out_queue >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue))) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EWOULDBLOCK); if (sndlen > SCTP_SB_LIMIT_SND(so)) error = EMSGSIZE; else error = EWOULDBLOCK; goto out_unlocked; } stcb->asoc.sb_send_resv += sndlen; SCTP_TCB_UNLOCK(stcb); hold_tcblock = 0; } else { atomic_add_int(&stcb->asoc.sb_send_resv, sndlen); } local_soresv = sndlen; /* Keep the stcb from being freed under our feet */ if (free_cnt_applied) { #ifdef INVARIANTS panic("refcnt already incremented"); #else printf("refcnt:1 already incremented?\n"); #endif } else { atomic_add_int(&stcb->asoc.refcnt, 1); free_cnt_applied = 1; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ECONNRESET); error = ECONNRESET; goto out_unlocked; } if (create_lock_applied) { SCTP_ASOC_CREATE_UNLOCK(inp); create_lock_applied = 0; } if (asoc->stream_reset_outstanding) { /* * Can't queue any data while stream reset is underway. */ SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EAGAIN); error = EAGAIN; goto out_unlocked; } if ((SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_WAIT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED)) { queue_only = 1; } if ((use_rcvinfo == 0) || (srcv == NULL)) { /* Grab the default stuff from the asoc */ srcv = (struct sctp_sndrcvinfo *)&stcb->asoc.def_send; } /* we are now done with all control */ if (control) { sctp_m_freem(control); control = NULL; } if ((SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_SENT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_ACK_SENT) || (asoc->state & SCTP_STATE_SHUTDOWN_PENDING)) { if ((use_rcvinfo) && (srcv->sinfo_flags & SCTP_ABORT)) { ; } else { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ECONNRESET); error = ECONNRESET; goto out_unlocked; } } /* Ok, we will attempt a msgsnd :> */ if (p) { p->td_ru.ru_msgsnd++; } if (stcb) { if (((srcv->sinfo_flags | temp_flags) & SCTP_ADDR_OVER) == 0) { net = stcb->asoc.primary_destination; } } if (net == NULL) { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } if ((net->flight_size > net->cwnd) && (SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0)) { /*- * CMT: Added check for CMT above. net above is the primary * dest. If CMT is ON, sender should always attempt to send * with the output routine sctp_fill_outqueue() that loops * through all destination addresses. Therefore, if CMT is * ON, queue_only is NOT set to 1 here, so that * sctp_chunk_output() can be called below. */ queue_only = 1; } else if (asoc->ifp_had_enobuf) { SCTP_STAT_INCR(sctps_ifnomemqueued); if (net->flight_size > (net->mtu * 2)) queue_only = 1; asoc->ifp_had_enobuf = 0; } else { un_sent = ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) + (stcb->asoc.stream_queue_cnt * sizeof(struct sctp_data_chunk))); } /* Are we aborting? */ if (srcv->sinfo_flags & SCTP_ABORT) { struct mbuf *mm; int tot_demand, tot_out = 0, max_out; SCTP_STAT_INCR(sctps_sends_with_abort); if ((SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_WAIT) || (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED)) { /* It has to be up before we abort */ /* how big is the user initiated abort? */ SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out; } if (hold_tcblock) { SCTP_TCB_UNLOCK(stcb); hold_tcblock = 0; } if (top) { struct mbuf *cntm = NULL; mm = sctp_get_mbuf_for_msg(1, 0, M_WAIT, 1, MT_DATA); if (sndlen != 0) { cntm = top; while (cntm) { tot_out += SCTP_BUF_LEN(cntm); cntm = SCTP_BUF_NEXT(cntm); } } tot_demand = (tot_out + sizeof(struct sctp_paramhdr)); } else { /* Must fit in a MTU */ tot_out = sndlen; tot_demand = (tot_out + sizeof(struct sctp_paramhdr)); if (tot_demand > SCTP_DEFAULT_ADD_MORE) { /* To big */ SCTP_LTRACE_ERR_RET(NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, EMSGSIZE); error = EMSGSIZE; goto out; } mm = sctp_get_mbuf_for_msg(tot_demand, 0, M_WAIT, 1, MT_DATA); } if (mm == NULL) { SCTP_LTRACE_ERR_RET(NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, ENOMEM); error = ENOMEM; goto out; } max_out = asoc->smallest_mtu - sizeof(struct sctp_paramhdr); max_out -= sizeof(struct sctp_abort_msg); if (tot_out > max_out) { tot_out = max_out; } if (mm) { struct sctp_paramhdr *ph; /* now move forward the data pointer */ ph = mtod(mm, struct sctp_paramhdr *); ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons((sizeof(struct sctp_paramhdr) + tot_out)); ph++; SCTP_BUF_LEN(mm) = tot_out + sizeof(struct sctp_paramhdr); if (top == NULL) { error = uiomove((caddr_t)ph, (int)tot_out, uio); if (error) { /*- * Here if we can't get his data we * still abort we just don't get to * send the users note :-0 */ sctp_m_freem(mm); mm = NULL; } } else { if (sndlen != 0) { SCTP_BUF_NEXT(mm) = top; } } } if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } atomic_add_int(&stcb->asoc.refcnt, -1); free_cnt_applied = 0; /* release this lock, otherwise we hang on ourselves */ sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_RESPONSE_TO_USER_REQ, mm, SCTP_SO_LOCKED); /* now relock the stcb so everything is sane */ hold_tcblock = 0; stcb = NULL; /* * In this case top is already chained to mm avoid double * free, since we free it below if top != NULL and driver * would free it after sending the packet out */ if (sndlen != 0) { top = NULL; } goto out_unlocked; } /* Calculate the maximum we can send */ inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); if (SCTP_SB_LIMIT_SND(so) > inqueue_bytes) { if (non_blocking) { /* we already checked for non-blocking above. */ max_len = sndlen; } else { max_len = SCTP_SB_LIMIT_SND(so) - inqueue_bytes; } } else { max_len = 0; } if (hold_tcblock) { SCTP_TCB_UNLOCK(stcb); hold_tcblock = 0; } /* Is the stream no. valid? */ if (srcv->sinfo_stream >= asoc->streamoutcnt) { /* Invalid stream number */ SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } if (asoc->strmout == NULL) { /* huh? software error */ SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EFAULT); error = EFAULT; goto out_unlocked; } /* Unless E_EOR mode is on, we must make a send FIT in one call. */ if ((user_marks_eor == 0) && (sndlen > SCTP_SB_LIMIT_SND(stcb->sctp_socket))) { /* It will NEVER fit */ SCTP_LTRACE_ERR_RET(NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, EMSGSIZE); error = EMSGSIZE; goto out_unlocked; } if ((uio == NULL) && user_marks_eor) { /*- * We do not support eeor mode for * sending with mbuf chains (like sendfile). */ SCTP_LTRACE_ERR_RET(NULL, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out_unlocked; } if (user_marks_eor) { local_add_more = SCTP_BASE_SYSCTL(sctp_add_more_threshold); } else { /*- * For non-eeor the whole message must fit in * the socket send buffer. */ local_add_more = sndlen; } len = 0; if (non_blocking) { goto skip_preblock; } if (((max_len <= local_add_more) && (SCTP_SB_LIMIT_SND(so) > local_add_more)) || ((stcb->asoc.chunks_on_out_queue + stcb->asoc.stream_queue_cnt) >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue))) { /* if */ /* No room right now ! */ SOCKBUF_LOCK(&so->so_snd); inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); while ((SCTP_SB_LIMIT_SND(so) < (inqueue_bytes + SCTP_BASE_SYSCTL(sctp_add_more_threshold))) || ((stcb->asoc.stream_queue_cnt + stcb->asoc.chunks_on_out_queue) >= SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue) /* while */ )) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { sctp_log_block(SCTP_BLOCK_LOG_INTO_BLKA, so, asoc, sndlen); } be.error = 0; stcb->block_entry = &be; error = sbwait(&so->so_snd); stcb->block_entry = NULL; if (error || so->so_error || be.error) { if (error == 0) { if (so->so_error) error = so->so_error; if (be.error) { error = be.error; } } SOCKBUF_UNLOCK(&so->so_snd); goto out_unlocked; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { sctp_log_block(SCTP_BLOCK_LOG_OUTOF_BLK, so, asoc, stcb->asoc.total_output_queue_size); } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { goto out_unlocked; } inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); } inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); if (SCTP_SB_LIMIT_SND(so) > inqueue_bytes) { max_len = SCTP_SB_LIMIT_SND(so) - inqueue_bytes; } else { max_len = 0; } SOCKBUF_UNLOCK(&so->so_snd); } skip_preblock: if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { goto out_unlocked; } atomic_add_int(&stcb->total_sends, 1); /* * sndlen covers for mbuf case uio_resid covers for the non-mbuf * case NOTE: uio will be null when top/mbuf is passed */ if (sndlen == 0) { if (srcv->sinfo_flags & SCTP_EOF) { got_all_of_the_send = 1; goto dataless_eof; } else { SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out; } } if (top == NULL) { struct sctp_stream_queue_pending *sp; struct sctp_stream_out *strm; uint32_t sndout, initial_out; initial_out = uio->uio_resid; SCTP_TCB_SEND_LOCK(stcb); if ((asoc->stream_locked) && (asoc->stream_locked_on != srcv->sinfo_stream)) { SCTP_TCB_SEND_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, stcb, net, SCTP_FROM_SCTP_OUTPUT, EINVAL); error = EINVAL; goto out; } SCTP_TCB_SEND_UNLOCK(stcb); strm = &stcb->asoc.strmout[srcv->sinfo_stream]; if (strm->last_msg_incomplete == 0) { do_a_copy_in: sp = sctp_copy_it_in(stcb, asoc, srcv, uio, net, max_len, user_marks_eor, &error, non_blocking); if ((sp == NULL) || (error)) { goto out; } SCTP_TCB_SEND_LOCK(stcb); if (sp->msg_is_complete) { strm->last_msg_incomplete = 0; asoc->stream_locked = 0; } else { /* * Just got locked to this guy in case of an * interrupt. */ strm->last_msg_incomplete = 1; asoc->stream_locked = 1; asoc->stream_locked_on = srcv->sinfo_stream; sp->sender_all_done = 0; } sctp_snd_sb_alloc(stcb, sp->length); atomic_add_int(&asoc->stream_queue_cnt, 1); if ((srcv->sinfo_flags & SCTP_UNORDERED) == 0) { sp->strseq = strm->next_sequence_sent; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOG_AT_SEND_2_SCTP) { sctp_misc_ints(SCTP_STRMOUT_LOG_ASSIGN, (uintptr_t) stcb, sp->length, (uint32_t) ((srcv->sinfo_stream << 16) | sp->strseq), 0); } strm->next_sequence_sent++; } else { SCTP_STAT_INCR(sctps_sends_with_unord); } TAILQ_INSERT_TAIL(&strm->outqueue, sp, next); if ((strm->next_spoke.tqe_next == NULL) && (strm->next_spoke.tqe_prev == NULL)) { /* Not on wheel, insert */ sctp_insert_on_wheel(stcb, asoc, strm, 1); } SCTP_TCB_SEND_UNLOCK(stcb); } else { SCTP_TCB_SEND_LOCK(stcb); sp = TAILQ_LAST(&strm->outqueue, sctp_streamhead); SCTP_TCB_SEND_UNLOCK(stcb); if (sp == NULL) { /* ???? Huh ??? last msg is gone */ #ifdef INVARIANTS panic("Warning: Last msg marked incomplete, yet nothing left?"); #else SCTP_PRINTF("Warning: Last msg marked incomplete, yet nothing left?\n"); strm->last_msg_incomplete = 0; #endif goto do_a_copy_in; } } while (uio->uio_resid > 0) { /* How much room do we have? */ struct mbuf *new_tail, *mm; if (SCTP_SB_LIMIT_SND(so) > stcb->asoc.total_output_queue_size) max_len = SCTP_SB_LIMIT_SND(so) - stcb->asoc.total_output_queue_size; else max_len = 0; if ((max_len > SCTP_BASE_SYSCTL(sctp_add_more_threshold)) || (max_len && (SCTP_SB_LIMIT_SND(so) < SCTP_BASE_SYSCTL(sctp_add_more_threshold))) || (uio->uio_resid && (uio->uio_resid <= (int)max_len))) { sndout = 0; new_tail = NULL; if (hold_tcblock) { SCTP_TCB_UNLOCK(stcb); hold_tcblock = 0; } mm = sctp_copy_resume(sp, uio, srcv, max_len, user_marks_eor, &error, &sndout, &new_tail); if ((mm == NULL) || error) { if (mm) { sctp_m_freem(mm); } goto out; } /* Update the mbuf and count */ SCTP_TCB_SEND_LOCK(stcb); if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { /* * we need to get out. Peer probably * aborted. */ sctp_m_freem(mm); if (stcb->asoc.state & SCTP_PCB_FLAGS_WAS_ABORTED) { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTP_OUTPUT, ECONNRESET); error = ECONNRESET; } SCTP_TCB_SEND_UNLOCK(stcb); goto out; } if (sp->tail_mbuf) { /* tack it to the end */ SCTP_BUF_NEXT(sp->tail_mbuf) = mm; sp->tail_mbuf = new_tail; } else { /* A stolen mbuf */ sp->data = mm; sp->tail_mbuf = new_tail; } sctp_snd_sb_alloc(stcb, sndout); atomic_add_int(&sp->length, sndout); len += sndout; /* Did we reach EOR? */ if ((uio->uio_resid == 0) && ((user_marks_eor == 0) || (srcv->sinfo_flags & SCTP_EOF) || (user_marks_eor && (srcv->sinfo_flags & SCTP_EOR))) ) { sp->msg_is_complete = 1; } else { sp->msg_is_complete = 0; } SCTP_TCB_SEND_UNLOCK(stcb); } if (uio->uio_resid == 0) { /* got it all? */ continue; } /* PR-SCTP? */ if ((asoc->peer_supports_prsctp) && (asoc->sent_queue_cnt_removeable > 0)) { /* * This is ugly but we must assure locking * order */ if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } sctp_prune_prsctp(stcb, asoc, srcv, sndlen); inqueue_bytes = stcb->asoc.total_output_queue_size - (stcb->asoc.chunks_on_out_queue * sizeof(struct sctp_data_chunk)); if (SCTP_SB_LIMIT_SND(so) > stcb->asoc.total_output_queue_size) max_len = SCTP_SB_LIMIT_SND(so) - inqueue_bytes; else max_len = 0; if (max_len > 0) { continue; } SCTP_TCB_UNLOCK(stcb); hold_tcblock = 0; } /* wait for space now */ if (non_blocking) { /* Non-blocking io in place out */ goto skip_out_eof; } if ((net->flight_size > net->cwnd) && (SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0)) { queue_only = 1; } else if (asoc->ifp_had_enobuf) { SCTP_STAT_INCR(sctps_ifnomemqueued); if (net->flight_size > (net->mtu * 2)) { queue_only = 1; } else { queue_only = 0; } asoc->ifp_had_enobuf = 0; un_sent = ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) + (stcb->asoc.stream_queue_cnt * sizeof(struct sctp_data_chunk))); } else { un_sent = ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) + (stcb->asoc.stream_queue_cnt * sizeof(struct sctp_data_chunk))); if (net->flight_size > (net->mtu * stcb->asoc.max_burst)) { queue_only = 1; SCTP_STAT_INCR(sctps_send_burst_avoid); } else if (net->flight_size > net->cwnd) { queue_only = 1; SCTP_STAT_INCR(sctps_send_cwnd_avoid); } else { queue_only = 0; } } if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) && (stcb->asoc.total_flight > 0) && (stcb->asoc.stream_queue_cnt < SCTP_MAX_DATA_BUNDLING) && (un_sent < (int)(stcb->asoc.smallest_mtu - SCTP_MIN_OVERHEAD)) ) { /*- * Ok, Nagle is set on and we have data outstanding. * Don't send anything and let SACKs drive out the * data unless wen have a "full" segment to send. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_NAGLE_LOGGING_ENABLE) { sctp_log_nagle_event(stcb, SCTP_NAGLE_APPLIED); } SCTP_STAT_INCR(sctps_naglequeued); nagle_applies = 1; } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_NAGLE_LOGGING_ENABLE) { if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) sctp_log_nagle_event(stcb, SCTP_NAGLE_SKIPPED); } SCTP_STAT_INCR(sctps_naglesent); nagle_applies = 0; } /* What about the INIT, send it maybe */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { sctp_misc_ints(SCTP_CWNDLOG_PRESEND, queue_only_for_init, queue_only, nagle_applies, un_sent); sctp_misc_ints(SCTP_CWNDLOG_PRESEND, stcb->asoc.total_output_queue_size, stcb->asoc.total_flight, stcb->asoc.chunks_on_out_queue, stcb->asoc.total_flight_count); } if (queue_only_for_init) { if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) { /* a collision took us forward? */ queue_only_for_init = 0; queue_only = 0; } else { sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); SCTP_SET_STATE(asoc, SCTP_STATE_COOKIE_WAIT); queue_only_for_init = 0; queue_only = 1; } } if ((queue_only == 0) && (nagle_applies == 0) ) { /*- * need to start chunk output * before blocking.. note that if * a lock is already applied, then * the input via the net is happening * and I don't need to start output :-D */ if (hold_tcblock == 0) { if (SCTP_TCB_TRYLOCK(stcb)) { hold_tcblock = 1; sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_LOCKED); } } else { sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_LOCKED); } if (hold_tcblock == 1) { SCTP_TCB_UNLOCK(stcb); hold_tcblock = 0; } } SOCKBUF_LOCK(&so->so_snd); /*- * This is a bit strange, but I think it will * work. The total_output_queue_size is locked and * protected by the TCB_LOCK, which we just released. * There is a race that can occur between releasing it * above, and me getting the socket lock, where sacks * come in but we have not put the SB_WAIT on the * so_snd buffer to get the wakeup. After the LOCK * is applied the sack_processing will also need to * LOCK the so->so_snd to do the actual sowwakeup(). So * once we have the socket buffer lock if we recheck the * size we KNOW we will get to sleep safely with the * wakeup flag in place. */ if (SCTP_SB_LIMIT_SND(so) <= (stcb->asoc.total_output_queue_size + min(SCTP_BASE_SYSCTL(sctp_add_more_threshold), SCTP_SB_LIMIT_SND(so))) ) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { sctp_log_block(SCTP_BLOCK_LOG_INTO_BLK, so, asoc, uio->uio_resid); } be.error = 0; stcb->block_entry = &be; error = sbwait(&so->so_snd); stcb->block_entry = NULL; if (error || so->so_error || be.error) { if (error == 0) { if (so->so_error) error = so->so_error; if (be.error) { error = be.error; } } SOCKBUF_UNLOCK(&so->so_snd); goto out_unlocked; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_BLK_LOGGING_ENABLE) { sctp_log_block(SCTP_BLOCK_LOG_OUTOF_BLK, so, asoc, stcb->asoc.total_output_queue_size); } } SOCKBUF_UNLOCK(&so->so_snd); if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { goto out_unlocked; } } SCTP_TCB_SEND_LOCK(stcb); if (sp) { if (sp->msg_is_complete == 0) { strm->last_msg_incomplete = 1; asoc->stream_locked = 1; asoc->stream_locked_on = srcv->sinfo_stream; } else { sp->sender_all_done = 1; strm->last_msg_incomplete = 0; asoc->stream_locked = 0; } } else { SCTP_PRINTF("Huh no sp TSNH?\n"); strm->last_msg_incomplete = 0; asoc->stream_locked = 0; } SCTP_TCB_SEND_UNLOCK(stcb); if (uio->uio_resid == 0) { got_all_of_the_send = 1; } } else if (top) { /* We send in a 0, since we do NOT have any locks */ error = sctp_msg_append(stcb, net, top, srcv, 0); top = NULL; if (srcv->sinfo_flags & SCTP_EOF) { /* * This should only happen for Panda for the mbuf * send case, which does NOT yet support EEOR mode. * Thus, we can just set this flag to do the proper * EOF handling. */ got_all_of_the_send = 1; } } if (error) { goto out; } dataless_eof: /* EOF thing ? */ if ((srcv->sinfo_flags & SCTP_EOF) && (got_all_of_the_send == 1) && (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) ) { int cnt; SCTP_STAT_INCR(sctps_sends_with_eof); error = 0; if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } cnt = sctp_is_there_unsent_data(stcb); if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (cnt == 0)) { if (asoc->locked_on_sending) { goto abort_anyway; } /* there is nothing queued to send, so I'm done... */ if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_RECEIVED) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { /* only send SHUTDOWN the first time through */ sctp_send_shutdown(stcb, stcb->asoc.primary_destination); if (SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, asoc->primary_destination); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } } else { /*- * we still got (or just got) data to send, so set * SHUTDOWN_PENDING */ /*- * XXX sockets draft says that SCTP_EOF should be * sent with no data. currently, we will allow user * data to be sent first and move to * SHUTDOWN-PENDING */ if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_RECEIVED) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } if (asoc->locked_on_sending) { /* Locked to send out the data */ struct sctp_stream_queue_pending *sp; sp = TAILQ_LAST(&asoc->locked_on_sending->outqueue, sctp_streamhead); if (sp) { if ((sp->length == 0) && (sp->msg_is_complete == 0)) asoc->state |= SCTP_STATE_PARTIAL_MSG_LEFT; } } asoc->state |= SCTP_STATE_SHUTDOWN_PENDING; if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { abort_anyway: if (free_cnt_applied) { atomic_add_int(&stcb->asoc.refcnt, -1); free_cnt_applied = 0; } sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_RESPONSE_TO_USER_REQ, NULL, SCTP_SO_LOCKED); /* * now relock the stcb so everything * is sane */ hold_tcblock = 0; stcb = NULL; goto out; } sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); sctp_feature_off(inp, SCTP_PCB_FLAGS_NODELAY); } } } skip_out_eof: if (!TAILQ_EMPTY(&stcb->asoc.control_send_queue)) { some_on_control = 1; } if ((net->flight_size > net->cwnd) && (SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0)) { queue_only = 1; } else if (asoc->ifp_had_enobuf) { SCTP_STAT_INCR(sctps_ifnomemqueued); if (net->flight_size > (net->mtu * 2)) { queue_only = 1; } else { queue_only = 0; } asoc->ifp_had_enobuf = 0; un_sent = ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) + (stcb->asoc.stream_queue_cnt * sizeof(struct sctp_data_chunk))); } else { un_sent = ((stcb->asoc.total_output_queue_size - stcb->asoc.total_flight) + (stcb->asoc.stream_queue_cnt * sizeof(struct sctp_data_chunk))); if (net->flight_size > (net->mtu * stcb->asoc.max_burst)) { queue_only = 1; SCTP_STAT_INCR(sctps_send_burst_avoid); } else if (net->flight_size > net->cwnd) { queue_only = 1; SCTP_STAT_INCR(sctps_send_cwnd_avoid); } else { queue_only = 0; } } if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) && (stcb->asoc.total_flight > 0) && (stcb->asoc.stream_queue_cnt < SCTP_MAX_DATA_BUNDLING) && (un_sent < (int)(stcb->asoc.smallest_mtu - SCTP_MIN_OVERHEAD)) ) { /*- * Ok, Nagle is set on and we have data outstanding. * Don't send anything and let SACKs drive out the * data unless wen have a "full" segment to send. */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_NAGLE_LOGGING_ENABLE) { sctp_log_nagle_event(stcb, SCTP_NAGLE_APPLIED); } SCTP_STAT_INCR(sctps_naglequeued); nagle_applies = 1; } else { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_NAGLE_LOGGING_ENABLE) { if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_NODELAY)) sctp_log_nagle_event(stcb, SCTP_NAGLE_SKIPPED); } SCTP_STAT_INCR(sctps_naglesent); nagle_applies = 0; } if (queue_only_for_init) { if (hold_tcblock == 0) { SCTP_TCB_LOCK(stcb); hold_tcblock = 1; } if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) { /* a collision took us forward? */ queue_only_for_init = 0; queue_only = 0; } else { sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); SCTP_SET_STATE(&stcb->asoc, SCTP_STATE_COOKIE_WAIT); queue_only_for_init = 0; queue_only = 1; } } if ((queue_only == 0) && (nagle_applies == 0) && (stcb->asoc.peers_rwnd && un_sent)) { /* we can attempt to send too. */ if (hold_tcblock == 0) { /* * If there is activity recv'ing sacks no need to * send */ if (SCTP_TCB_TRYLOCK(stcb)) { sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_LOCKED); hold_tcblock = 1; } } else { sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_LOCKED); } } else if ((queue_only == 0) && (stcb->asoc.peers_rwnd == 0) && (stcb->asoc.total_flight == 0)) { /* We get to have a probe outstanding */ if (hold_tcblock == 0) { hold_tcblock = 1; SCTP_TCB_LOCK(stcb); } sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_USR_SEND, SCTP_SO_LOCKED); } else if (some_on_control) { int num_out, reason, cwnd_full, frag_point; /* Here we do control only */ if (hold_tcblock == 0) { hold_tcblock = 1; SCTP_TCB_LOCK(stcb); } frag_point = sctp_get_frag_point(stcb, &stcb->asoc); (void)sctp_med_chunk_output(inp, stcb, &stcb->asoc, &num_out, &reason, 1, &cwnd_full, 1, &now, &now_filled, frag_point, SCTP_SO_LOCKED); } SCTPDBG(SCTP_DEBUG_OUTPUT1, "USR Send complete qo:%d prw:%d unsent:%d tf:%d cooq:%d toqs:%d err:%d", queue_only, stcb->asoc.peers_rwnd, un_sent, stcb->asoc.total_flight, stcb->asoc.chunks_on_out_queue, stcb->asoc.total_output_queue_size, error); out: out_unlocked: if (local_soresv && stcb) { atomic_subtract_int(&stcb->asoc.sb_send_resv, sndlen); local_soresv = 0; } if (create_lock_applied) { SCTP_ASOC_CREATE_UNLOCK(inp); create_lock_applied = 0; } if ((stcb) && hold_tcblock) { SCTP_TCB_UNLOCK(stcb); } if (stcb && free_cnt_applied) { atomic_add_int(&stcb->asoc.refcnt, -1); } #ifdef INVARIANTS if (stcb) { if (mtx_owned(&stcb->tcb_mtx)) { panic("Leaving with tcb mtx owned?"); } if (mtx_owned(&stcb->tcb_send_mtx)) { panic("Leaving with tcb send mtx owned?"); } } #endif if (top) { sctp_m_freem(top); } if (control) { sctp_m_freem(control); } return (error); } /* * generate an AUTHentication chunk, if required */ struct mbuf * sctp_add_auth_chunk(struct mbuf *m, struct mbuf **m_end, struct sctp_auth_chunk **auth_ret, uint32_t * offset, struct sctp_tcb *stcb, uint8_t chunk) { struct mbuf *m_auth; struct sctp_auth_chunk *auth; int chunk_len; if ((m_end == NULL) || (auth_ret == NULL) || (offset == NULL) || (stcb == NULL)) return (m); /* sysctl disabled auth? */ if (SCTP_BASE_SYSCTL(sctp_auth_disable)) return (m); /* peer doesn't do auth... */ if (!stcb->asoc.peer_supports_auth) { return (m); } /* does the requested chunk require auth? */ if (!sctp_auth_is_required_chunk(chunk, stcb->asoc.peer_auth_chunks)) { return (m); } m_auth = sctp_get_mbuf_for_msg(sizeof(*auth), 0, M_DONTWAIT, 1, MT_HEADER); if (m_auth == NULL) { /* no mbuf's */ return (m); } /* reserve some space if this will be the first mbuf */ if (m == NULL) SCTP_BUF_RESV_UF(m_auth, SCTP_MIN_OVERHEAD); /* fill in the AUTH chunk details */ auth = mtod(m_auth, struct sctp_auth_chunk *); bzero(auth, sizeof(*auth)); auth->ch.chunk_type = SCTP_AUTHENTICATION; auth->ch.chunk_flags = 0; chunk_len = sizeof(*auth) + sctp_get_hmac_digest_len(stcb->asoc.peer_hmac_id); auth->ch.chunk_length = htons(chunk_len); auth->hmac_id = htons(stcb->asoc.peer_hmac_id); /* key id and hmac digest will be computed and filled in upon send */ /* save the offset where the auth was inserted into the chain */ if (m != NULL) { struct mbuf *cn; *offset = 0; cn = m; while (cn) { *offset += SCTP_BUF_LEN(cn); cn = SCTP_BUF_NEXT(cn); } } else *offset = 0; /* update length and return pointer to the auth chunk */ SCTP_BUF_LEN(m_auth) = chunk_len; m = sctp_copy_mbufchain(m_auth, m, m_end, 1, chunk_len, 0); if (auth_ret != NULL) *auth_ret = auth; return (m); } #ifdef INET6 int sctp_v6src_match_nexthop(struct sockaddr_in6 *src6, sctp_route_t * ro) { struct nd_prefix *pfx = NULL; struct nd_pfxrouter *pfxrtr = NULL; struct sockaddr_in6 gw6; if (ro == NULL || ro->ro_rt == NULL || src6->sin6_family != AF_INET6) return (0); /* get prefix entry of address */ LIST_FOREACH(pfx, &MODULE_GLOBAL(MOD_INET6, nd_prefix), ndpr_entry) { if (pfx->ndpr_stateflags & NDPRF_DETACHED) continue; if (IN6_ARE_MASKED_ADDR_EQUAL(&pfx->ndpr_prefix.sin6_addr, &src6->sin6_addr, &pfx->ndpr_mask)) break; } /* no prefix entry in the prefix list */ if (pfx == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT2, "No prefix entry for "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)src6); return (0); } SCTPDBG(SCTP_DEBUG_OUTPUT2, "v6src_match_nexthop(), Prefix entry is "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)src6); /* search installed gateway from prefix entry */ for (pfxrtr = pfx->ndpr_advrtrs.lh_first; pfxrtr; pfxrtr = pfxrtr->pfr_next) { memset(&gw6, 0, sizeof(struct sockaddr_in6)); gw6.sin6_family = AF_INET6; gw6.sin6_len = sizeof(struct sockaddr_in6); memcpy(&gw6.sin6_addr, &pfxrtr->router->rtaddr, sizeof(struct in6_addr)); SCTPDBG(SCTP_DEBUG_OUTPUT2, "prefix router is "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, (struct sockaddr *)&gw6); SCTPDBG(SCTP_DEBUG_OUTPUT2, "installed router is "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, ro->ro_rt->rt_gateway); if (sctp_cmpaddr((struct sockaddr *)&gw6, ro->ro_rt->rt_gateway)) { SCTPDBG(SCTP_DEBUG_OUTPUT2, "pfxrouter is installed\n"); return (1); } } SCTPDBG(SCTP_DEBUG_OUTPUT2, "pfxrouter is not installed\n"); return (0); } #endif int sctp_v4src_match_nexthop(struct sctp_ifa *sifa, sctp_route_t * ro) { struct sockaddr_in *sin, *mask; struct ifaddr *ifa; struct in_addr srcnetaddr, gwnetaddr; if (ro == NULL || ro->ro_rt == NULL || sifa->address.sa.sa_family != AF_INET) { return (0); } ifa = (struct ifaddr *)sifa->ifa; mask = (struct sockaddr_in *)(ifa->ifa_netmask); sin = (struct sockaddr_in *)&sifa->address.sin; srcnetaddr.s_addr = (sin->sin_addr.s_addr & mask->sin_addr.s_addr); SCTPDBG(SCTP_DEBUG_OUTPUT1, "match_nexthop4: src address is "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, &sifa->address.sa); SCTPDBG(SCTP_DEBUG_OUTPUT1, "network address is %x\n", srcnetaddr.s_addr); sin = (struct sockaddr_in *)ro->ro_rt->rt_gateway; gwnetaddr.s_addr = (sin->sin_addr.s_addr & mask->sin_addr.s_addr); SCTPDBG(SCTP_DEBUG_OUTPUT1, "match_nexthop4: nexthop is "); SCTPDBG_ADDR(SCTP_DEBUG_OUTPUT2, ro->ro_rt->rt_gateway); SCTPDBG(SCTP_DEBUG_OUTPUT1, "network address is %x\n", gwnetaddr.s_addr); if (srcnetaddr.s_addr == gwnetaddr.s_addr) { return (1); } return (0); } Index: head/sys/netinet/sctp_pcb.c =================================================================== --- head/sys/netinet/sctp_pcb.c (revision 181053) +++ head/sys/netinet/sctp_pcb.c (revision 181054) @@ -1,6470 +1,6562 @@ /*- * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_pcb.c,v 1.38 2005/03/06 16:04:18 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include struct sctp_base_info system_base_info; /* FIX: we don't handle multiple link local scopes */ /* "scopeless" replacement IN6_ARE_ADDR_EQUAL */ #ifdef INET6 int SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b) { struct sockaddr_in6 tmp_a, tmp_b; memcpy(&tmp_a, a, sizeof(struct sockaddr_in6)); if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(MOD_INET6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone))) != 0) { return 0; } memcpy(&tmp_b, b, sizeof(struct sockaddr_in6)); if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(MOD_INET6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone))) != 0) { return 0; } return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr)); } #endif void sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb) { /* * We really don't need to lock this, but I will just because it * does not hurt. */ SCTP_INP_INFO_RLOCK(); spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep); spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc); spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr); spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr); spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk); spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq); spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq); spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks); SCTP_INP_INFO_RUNLOCK(); } /* * Addresses are added to VRF's (Virtual Router's). For BSD we * have only the default VRF 0. We maintain a hash list of * VRF's. Each VRF has its own list of sctp_ifn's. Each of * these has a list of addresses. When we add a new address * to a VRF we lookup the ifn/ifn_index, if the ifn does * not exist we create it and add it to the list of IFN's * within the VRF. Once we have the sctp_ifn, we add the * address to the list. So we look something like: * * hash-vrf-table * vrf-> ifn-> ifn -> ifn * vrf | * ... +--ifa-> ifa -> ifa * vrf * * We keep these seperate lists since the SCTP subsystem will * point to these from its source address selection nets structure. * When an address is deleted it does not happen right away on * the SCTP side, it gets scheduled. What we do when a * delete happens is immediately remove the address from * the master list and decrement the refcount. As our * addip iterator works through and frees the src address * selection pointing to the sctp_ifa, eventually the refcount * will reach 0 and we will delete it. Note that it is assumed * that any locking on system level ifn/ifa is done at the * caller of these functions and these routines will only * lock the SCTP structures as they add or delete things. * * Other notes on VRF concepts. * - An endpoint can be in multiple VRF's * - An association lives within a VRF and only one VRF. * - Any incoming packet we can deduce the VRF for by * looking at the mbuf/pak inbound (for BSD its VRF=0 :D) * - Any downward send call or connect call must supply the * VRF via ancillary data or via some sort of set default * VRF socket option call (again for BSD no brainer since * the VRF is always 0). * - An endpoint may add multiple VRF's to it. * - Listening sockets can accept associations in any * of the VRF's they are in but the assoc will end up * in only one VRF (gotten from the packet or connect/send). * */ struct sctp_vrf * sctp_allocate_vrf(int vrf_id) { struct sctp_vrf *vrf = NULL; struct sctp_vrflist *bucket; /* First allocate the VRF structure */ vrf = sctp_find_vrf(vrf_id); if (vrf) { /* Already allocated */ return (vrf); } SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf), SCTP_M_VRF); if (vrf == NULL) { /* No memory */ #ifdef INVARIANTS panic("No memory for VRF:%d", vrf_id); #endif return (NULL); } /* setup the VRF */ memset(vrf, 0, sizeof(struct sctp_vrf)); vrf->vrf_id = vrf_id; LIST_INIT(&vrf->ifnlist); vrf->total_ifa_count = 0; vrf->refcount = 0; /* now also setup table ids */ SCTP_INIT_VRF_TABLEID(vrf); /* Init the HASH of addresses */ vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE, &vrf->vrf_addr_hashmark); if (vrf->vrf_addr_hash == NULL) { /* No memory */ #ifdef INVARIANTS panic("No memory for VRF:%d", vrf_id); #endif SCTP_FREE(vrf, SCTP_M_VRF); return (NULL); } /* Add it to the hash table */ bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))]; LIST_INSERT_HEAD(bucket, vrf, next_vrf); atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1); return (vrf); } struct sctp_ifn * sctp_find_ifn(void *ifn, uint32_t ifn_index) { struct sctp_ifn *sctp_ifnp; struct sctp_ifnlist *hash_ifn_head; /* * We assume the lock is held for the addresses if thats wrong * problems could occur :-) */ hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))]; LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) { if (sctp_ifnp->ifn_index == ifn_index) { return (sctp_ifnp); } if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) { return (sctp_ifnp); } } return (NULL); } struct sctp_vrf * sctp_find_vrf(uint32_t vrf_id) { struct sctp_vrflist *bucket; struct sctp_vrf *liste; bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))]; LIST_FOREACH(liste, bucket, next_vrf) { if (vrf_id == liste->vrf_id) { return (liste); } } return (NULL); } void sctp_free_vrf(struct sctp_vrf *vrf) { int ret; ret = atomic_fetchadd_int(&vrf->refcount, -1); if (ret == 1) { if (vrf->vrf_addr_hash) { SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark); vrf->vrf_addr_hash = NULL; } /* We zero'd the count */ LIST_REMOVE(vrf, next_vrf); SCTP_FREE(vrf, SCTP_M_VRF); atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1); } } void sctp_free_ifn(struct sctp_ifn *sctp_ifnp) { int ret; ret = atomic_fetchadd_int(&sctp_ifnp->refcount, -1); if (ret == 1) { /* We zero'd the count */ if (sctp_ifnp->vrf) { sctp_free_vrf(sctp_ifnp->vrf); } SCTP_FREE(sctp_ifnp, SCTP_M_IFN); atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1); } } void sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu) { struct sctp_ifn *sctp_ifnp; sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index); if (sctp_ifnp != NULL) { sctp_ifnp->ifn_mtu = mtu; } } void sctp_free_ifa(struct sctp_ifa *sctp_ifap) { int ret; ret = atomic_fetchadd_int(&sctp_ifap->refcount, -1); if (ret == 1) { /* We zero'd the count */ if (sctp_ifap->ifn_p) { sctp_free_ifn(sctp_ifap->ifn_p); } SCTP_FREE(sctp_ifap, SCTP_M_IFA); atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1); } } static void sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock) { struct sctp_ifn *found; found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index); if (found == NULL) { /* Not in the list.. sorry */ return; } if (hold_addr_lock == 0) SCTP_IPI_ADDR_WLOCK(); LIST_REMOVE(sctp_ifnp, next_bucket); LIST_REMOVE(sctp_ifnp, next_ifn); SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index, sctp_ifnp->registered_af); if (hold_addr_lock == 0) SCTP_IPI_ADDR_WUNLOCK(); /* Take away the reference, and possibly free it */ sctp_free_ifn(sctp_ifnp); } void sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr, const char *if_name, uint32_t ifn_index) { struct sctp_vrf *vrf; struct sctp_ifa *sctp_ifap = NULL; SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); goto out; } sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); if (sctp_ifap == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n"); goto out; } if (sctp_ifap->ifn_p == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n"); goto out; } if (if_name) { int len1, len2; len1 = strlen(if_name); len2 = strlen(sctp_ifap->ifn_p->ifn_name); if (len1 != len2) { SCTPDBG(SCTP_DEBUG_PCB4, "IFN of ifa names different lenght %d vs %d - ignored\n", len1, len2); goto out; } if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, len1) != 0) { SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n", sctp_ifap->ifn_p->ifn_name, if_name); goto out; } } else { if (sctp_ifap->ifn_p->ifn_index != ifn_index) { SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n", sctp_ifap->ifn_p->ifn_index, ifn_index); goto out; } } sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID); sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE; out: SCTP_IPI_ADDR_RUNLOCK(); } void sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr, const char *if_name, uint32_t ifn_index) { struct sctp_vrf *vrf; struct sctp_ifa *sctp_ifap = NULL; SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); goto out; } sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); if (sctp_ifap == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n"); goto out; } if (sctp_ifap->ifn_p == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n"); goto out; } if (if_name) { int len1, len2; len1 = strlen(if_name); len2 = strlen(sctp_ifap->ifn_p->ifn_name); if (len1 != len2) { SCTPDBG(SCTP_DEBUG_PCB4, "IFN of ifa names different lenght %d vs %d - ignored\n", len1, len2); goto out; } if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, len1) != 0) { SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n", sctp_ifap->ifn_p->ifn_name, if_name); goto out; } } else { if (sctp_ifap->ifn_p->ifn_index != ifn_index) { SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n", sctp_ifap->ifn_p->ifn_index, ifn_index); goto out; } } sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE); sctp_ifap->localifa_flags |= SCTP_ADDR_VALID; out: SCTP_IPI_ADDR_RUNLOCK(); } /*- * Add an ifa to an ifn. * Register the interface as necessary. * NOTE: ADDR write lock MUST be held. */ static void sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap) { int ifa_af; LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa); sctp_ifap->ifn_p = sctp_ifnp; atomic_add_int(&sctp_ifap->ifn_p->refcount, 1); /* update address counts */ sctp_ifnp->ifa_count++; ifa_af = sctp_ifap->address.sa.sa_family; if (ifa_af == AF_INET) sctp_ifnp->num_v4++; else sctp_ifnp->num_v6++; if (sctp_ifnp->ifa_count == 1) { /* register the new interface */ SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af); sctp_ifnp->registered_af = ifa_af; } } /*- * Remove an ifa from its ifn. * If no more addresses exist, remove the ifn too. Otherwise, re-register * the interface based on the remaining address families left. * NOTE: ADDR write lock MUST be held. */ static void sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap) { uint32_t ifn_index; LIST_REMOVE(sctp_ifap, next_ifa); if (sctp_ifap->ifn_p) { /* update address counts */ sctp_ifap->ifn_p->ifa_count--; if (sctp_ifap->address.sa.sa_family == AF_INET6) sctp_ifap->ifn_p->num_v6--; else if (sctp_ifap->address.sa.sa_family == AF_INET) sctp_ifap->ifn_p->num_v4--; ifn_index = sctp_ifap->ifn_p->ifn_index; if (SCTP_LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) { /* remove the ifn, possibly freeing it */ sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED); } else { /* re-register address family type, if needed */ if ((sctp_ifap->ifn_p->num_v6 == 0) && (sctp_ifap->ifn_p->registered_af == AF_INET6)) { SCTP_DEREGISTER_INTERFACE(ifn_index, AF_INET6); SCTP_REGISTER_INTERFACE(ifn_index, AF_INET); sctp_ifap->ifn_p->registered_af = AF_INET; } else if ((sctp_ifap->ifn_p->num_v4 == 0) && (sctp_ifap->ifn_p->registered_af == AF_INET)) { SCTP_DEREGISTER_INTERFACE(ifn_index, AF_INET); SCTP_REGISTER_INTERFACE(ifn_index, AF_INET6); sctp_ifap->ifn_p->registered_af = AF_INET6; } /* free the ifn refcount */ sctp_free_ifn(sctp_ifap->ifn_p); } sctp_ifap->ifn_p = NULL; } } struct sctp_ifa * sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index, uint32_t ifn_type, const char *if_name, void *ifa, struct sockaddr *addr, uint32_t ifa_flags, int dynamic_add) { struct sctp_vrf *vrf; struct sctp_ifn *sctp_ifnp = NULL; struct sctp_ifa *sctp_ifap = NULL; struct sctp_ifalist *hash_addr_head; struct sctp_ifnlist *hash_ifn_head; uint32_t hash_of_addr; int new_ifn_af = 0; #ifdef SCTP_DEBUG SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id); SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr); #endif SCTP_IPI_ADDR_WLOCK(); sctp_ifnp = sctp_find_ifn(ifn, ifn_index); if (sctp_ifnp) { vrf = sctp_ifnp->vrf; } else { vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { vrf = sctp_allocate_vrf(vrf_id); if (vrf == NULL) { SCTP_IPI_ADDR_WUNLOCK(); return (NULL); } } } if (sctp_ifnp == NULL) { /* * build one and add it, can't hold lock until after malloc * done though. */ SCTP_IPI_ADDR_WUNLOCK(); SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *, sizeof(struct sctp_ifn), SCTP_M_IFN); if (sctp_ifnp == NULL) { #ifdef INVARIANTS panic("No memory for IFN:%u", sctp_ifnp->ifn_index); #endif return (NULL); } memset(sctp_ifnp, 0, sizeof(struct sctp_ifn)); sctp_ifnp->ifn_index = ifn_index; sctp_ifnp->ifn_p = ifn; sctp_ifnp->ifn_type = ifn_type; sctp_ifnp->refcount = 0; sctp_ifnp->vrf = vrf; atomic_add_int(&vrf->refcount, 1); sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family); if (if_name != NULL) { memcpy(sctp_ifnp->ifn_name, if_name, SCTP_IFNAMSIZ); } else { memcpy(sctp_ifnp->ifn_name, "unknown", min(7, SCTP_IFNAMSIZ)); } hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))]; LIST_INIT(&sctp_ifnp->ifalist); SCTP_IPI_ADDR_WLOCK(); LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket); LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn); atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1); new_ifn_af = 1; } sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); if (sctp_ifap) { /* Hmm, it already exists? */ if ((sctp_ifap->ifn_p) && (sctp_ifap->ifn_p->ifn_index == ifn_index)) { SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n", sctp_ifap->ifn_p->ifn_name, ifn_index, sctp_ifap); if (new_ifn_af) { /* Remove the created one that we don't want */ sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED); } if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) { /* easy to solve, just switch back to active */ SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n"); sctp_ifap->localifa_flags = SCTP_ADDR_VALID; sctp_ifap->ifn_p = sctp_ifnp; atomic_add_int(&sctp_ifap->ifn_p->refcount, 1); } exit_stage_left: SCTP_IPI_ADDR_WUNLOCK(); return (sctp_ifap); } else { if (sctp_ifap->ifn_p) { /* * The last IFN gets the address, removee * the old one */ SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n", sctp_ifap, sctp_ifap->ifn_p->ifn_name, sctp_ifap->ifn_p->ifn_index, if_name, ifn_index); /* remove the address from the old ifn */ sctp_remove_ifa_from_ifn(sctp_ifap); /* move the address over to the new ifn */ sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap); goto exit_stage_left; } else { /* repair ifnp which was NULL ? */ sctp_ifap->localifa_flags = SCTP_ADDR_VALID; SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n", sctp_ifnp, sctp_ifap); sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap); } goto exit_stage_left; } } SCTP_IPI_ADDR_WUNLOCK(); SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA); if (sctp_ifap == NULL) { #ifdef INVARIANTS panic("No memory for IFA"); #endif return (NULL); } memset(sctp_ifap, 0, sizeof(struct sctp_ifa)); sctp_ifap->ifn_p = sctp_ifnp; atomic_add_int(&sctp_ifnp->refcount, 1); sctp_ifap->vrf_id = vrf_id; sctp_ifap->ifa = ifa; memcpy(&sctp_ifap->address, addr, addr->sa_len); sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE; sctp_ifap->flags = ifa_flags; /* Set scope */ switch (sctp_ifap->address.sa.sa_family) { case AF_INET: { struct sockaddr_in *sin; sin = (struct sockaddr_in *)&sctp_ifap->address.sin; if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) || (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) { sctp_ifap->src_is_loop = 1; } if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { sctp_ifap->src_is_priv = 1; } sctp_ifnp->num_v4++; if (new_ifn_af) new_ifn_af = AF_INET; break; } #ifdef INET6 case AF_INET6: { /* ok to use deprecated addresses? */ struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6; if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) || (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) { sctp_ifap->src_is_loop = 1; } if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { sctp_ifap->src_is_priv = 1; } sctp_ifnp->num_v6++; if (new_ifn_af) new_ifn_af = AF_INET6; break; } #endif default: new_ifn_af = 0; break; } hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa); if ((sctp_ifap->src_is_priv == 0) && (sctp_ifap->src_is_loop == 0)) { sctp_ifap->src_is_glob = 1; } SCTP_IPI_ADDR_WLOCK(); hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)]; LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket); sctp_ifap->refcount = 1; LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa); sctp_ifnp->ifa_count++; vrf->total_ifa_count++; atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1); if (new_ifn_af) { SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af); sctp_ifnp->registered_af = new_ifn_af; } SCTP_IPI_ADDR_WUNLOCK(); if (dynamic_add) { /* * Bump up the refcount so that when the timer completes it * will drop back down. */ struct sctp_laddr *wi; atomic_add_int(&sctp_ifap->refcount, 1); wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (wi == NULL) { /* * Gak, what can we do? We have lost an address * change can you say HOSED? */ SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n"); /* Opps, must decrement the count */ sctp_del_addr_from_vrf(vrf_id, addr, ifn_index, if_name); return (NULL); } SCTP_INCR_LADDR_COUNT(); bzero(wi, sizeof(*wi)); (void)SCTP_GETTIME_TIMEVAL(&wi->start_time); wi->ifa = sctp_ifap; wi->action = SCTP_ADD_IP_ADDRESS; SCTP_IPI_ITERATOR_WQ_LOCK(); /* * Should this really be a tailq? As it is we will process * the newest first :-0 */ LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); SCTP_IPI_ITERATOR_WQ_UNLOCK(); sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, (struct sctp_inpcb *)NULL, (struct sctp_tcb *)NULL, (struct sctp_nets *)NULL); } else { /* it's ready for use */ sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE; } return (sctp_ifap); } void sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr, uint32_t ifn_index, const char *if_name) { struct sctp_vrf *vrf; struct sctp_ifa *sctp_ifap = NULL; SCTP_IPI_ADDR_WLOCK(); vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); goto out_now; } #ifdef SCTP_DEBUG SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id); SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr); #endif sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); if (sctp_ifap) { /* Validate the delete */ if (sctp_ifap->ifn_p) { int valid = 0; /*- * The name has priority over the ifn_index * if its given. We do this especially for * panda who might recycle indexes fast. */ if (if_name) { int len1, len2; len1 = min(SCTP_IFNAMSIZ, strlen(if_name)); len2 = min(SCTP_IFNAMSIZ, strlen(sctp_ifap->ifn_p->ifn_name)); if (len1 && len2 && (len1 == len2)) { /* we can compare them */ if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, len1) == 0) { /* * They match its a correct * delete */ valid = 1; } } } if (!valid) { /* last ditch check ifn_index */ if (ifn_index == sctp_ifap->ifn_p->ifn_index) { valid = 1; } } if (!valid) { SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n", ifn_index, ((if_name == NULL) ? "NULL" : if_name)); SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n", sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name); SCTP_IPI_ADDR_WUNLOCK(); return; } } SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", sctp_ifap); sctp_ifap->localifa_flags &= SCTP_ADDR_VALID; sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; vrf->total_ifa_count--; LIST_REMOVE(sctp_ifap, next_bucket); sctp_remove_ifa_from_ifn(sctp_ifap); } #ifdef SCTP_DEBUG else { SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:", ifn_index); SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr); } #endif out_now: SCTP_IPI_ADDR_WUNLOCK(); if (sctp_ifap) { struct sctp_laddr *wi; wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (wi == NULL) { /* * Gak, what can we do? We have lost an address * change can you say HOSED? */ SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n"); /* Oops, must decrement the count */ sctp_free_ifa(sctp_ifap); return; } SCTP_INCR_LADDR_COUNT(); bzero(wi, sizeof(*wi)); (void)SCTP_GETTIME_TIMEVAL(&wi->start_time); wi->ifa = sctp_ifap; wi->action = SCTP_DEL_IP_ADDRESS; SCTP_IPI_ITERATOR_WQ_LOCK(); /* * Should this really be a tailq? As it is we will process * the newest first :-0 */ LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); SCTP_IPI_ITERATOR_WQ_UNLOCK(); sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, (struct sctp_inpcb *)NULL, (struct sctp_tcb *)NULL, (struct sctp_nets *)NULL); } return; } static struct sctp_tcb * sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from, struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id) { /**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */ /* * If we support the TCP model, then we must now dig through to see * if we can find our endpoint in the list of tcp ep's. */ uint16_t lport, rport; struct sctppcbhead *ephead; struct sctp_inpcb *inp; struct sctp_laddr *laddr; struct sctp_tcb *stcb; struct sctp_nets *net; if ((to == NULL) || (from == NULL)) { return (NULL); } if (to->sa_family == AF_INET && from->sa_family == AF_INET) { lport = ((struct sockaddr_in *)to)->sin_port; rport = ((struct sockaddr_in *)from)->sin_port; } else if (to->sa_family == AF_INET6 && from->sa_family == AF_INET6) { lport = ((struct sockaddr_in6 *)to)->sin6_port; rport = ((struct sockaddr_in6 *)from)->sin6_port; } else { return NULL; } - ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR( - (lport + rport), SCTP_BASE_INFO(hashtcpmark))]; + ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport), SCTP_BASE_INFO(hashtcpmark))]; /* * Ok now for each of the guys in this bucket we must look and see: * - Does the remote port match. - Does there single association's * addresses match this address (to). If so we update p_ep to point * to this ep and return the tcb from it. */ LIST_FOREACH(inp, ephead, sctp_hash) { SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(inp); continue; } if (lport != inp->sctp_lport) { SCTP_INP_RUNLOCK(inp); continue; } if (inp->def_vrf_id != vrf_id) { SCTP_INP_RUNLOCK(inp); continue; } /* check to see if the ep has one of the addresses */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { /* We are NOT bound all, so look further */ int match = 0; LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__); continue; } if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n"); continue; } if (laddr->ifa->address.sa.sa_family == to->sa_family) { /* see if it matches */ struct sockaddr_in *intf_addr, *sin; intf_addr = &laddr->ifa->address.sin; sin = (struct sockaddr_in *)to; if (from->sa_family == AF_INET) { if (sin->sin_addr.s_addr == intf_addr->sin_addr.s_addr) { match = 1; break; } } #ifdef INET6 if (from->sa_family == AF_INET6) { struct sockaddr_in6 *intf_addr6; struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *) to; intf_addr6 = &laddr->ifa->address.sin6; if (SCTP6_ARE_ADDR_EQUAL(sin6, intf_addr6)) { match = 1; break; } } #endif } } if (match == 0) { /* This endpoint does not have this address */ SCTP_INP_RUNLOCK(inp); continue; } } /* * Ok if we hit here the ep has the address, does it hold * the tcb? */ stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { SCTP_INP_RUNLOCK(inp); continue; } SCTP_TCB_LOCK(stcb); if (stcb->rport != rport) { /* remote port does not match. */ SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); continue; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); continue; } /* Does this TCB have a matching address? */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (net->ro._l_addr.sa.sa_family != from->sa_family) { /* not the same family, can't be a match */ continue; } switch (from->sa_family) { case AF_INET: { struct sockaddr_in *sin, *rsin; sin = (struct sockaddr_in *)&net->ro._l_addr; rsin = (struct sockaddr_in *)from; if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { /* found it */ if (netp != NULL) { *netp = net; } /* * Update the endpoint * pointer */ *inp_p = inp; SCTP_INP_RUNLOCK(inp); return (stcb); } break; } #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6, *rsin6; sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; rsin6 = (struct sockaddr_in6 *)from; if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { /* found it */ if (netp != NULL) { *netp = net; } /* * Update the endpoint * pointer */ *inp_p = inp; SCTP_INP_RUNLOCK(inp); return (stcb); } break; } #endif default: /* TSNH */ break; } } SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); } return (NULL); } /* * rules for use * * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an * stcb, both will be locked (locked_tcb and stcb) but decrement will be done * (if locked == NULL). 3) Decrement happens on return ONLY if locked == * NULL. */ struct sctp_tcb * sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote, struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb) { struct sctpasochead *head; struct sctp_inpcb *inp; struct sctp_tcb *stcb = NULL; struct sctp_nets *net; uint16_t rport; inp = *inp_p; if (remote->sa_family == AF_INET) { rport = (((struct sockaddr_in *)remote)->sin_port); } else if (remote->sa_family == AF_INET6) { rport = (((struct sockaddr_in6 *)remote)->sin6_port); } else { return (NULL); } if (locked_tcb) { /* * UN-lock so we can do proper locking here this occurs when * called from load_addresses_from_init. */ atomic_add_int(&locked_tcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(locked_tcb); } SCTP_INP_INFO_RLOCK(); if (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) { /*- * Now either this guy is our listener or it's the * connector. If it is the one that issued the connect, then * it's only chance is to be the first TCB in the list. If * it is the acceptor, then do the special_lookup to hash * and find the real inp. */ if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) { /* to is peer addr, from is my addr */ stcb = sctp_tcb_special_locate(inp_p, remote, local, netp, inp->def_vrf_id); if ((stcb != NULL) && (locked_tcb == NULL)) { /* we have a locked tcb, lower refcount */ SCTP_INP_DECR_REF(inp); } if ((locked_tcb != NULL) && (locked_tcb != stcb)) { SCTP_INP_RLOCK(locked_tcb->sctp_ep); SCTP_TCB_LOCK(locked_tcb); atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); SCTP_INP_RUNLOCK(locked_tcb->sctp_ep); } SCTP_INP_INFO_RUNLOCK(); return (stcb); } else { SCTP_INP_WLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { goto null_return; } stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { goto null_return; } SCTP_TCB_LOCK(stcb); if (stcb->rport != rport) { /* remote port does not match. */ SCTP_TCB_UNLOCK(stcb); goto null_return; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_UNLOCK(stcb); goto null_return; } /* now look at the list of remote addresses */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { #ifdef INVARIANTS if (net == (TAILQ_NEXT(net, sctp_next))) { panic("Corrupt net list"); } #endif if (net->ro._l_addr.sa.sa_family != remote->sa_family) { /* not the same family */ continue; } switch (remote->sa_family) { case AF_INET: { struct sockaddr_in *sin, *rsin; sin = (struct sockaddr_in *) &net->ro._l_addr; rsin = (struct sockaddr_in *)remote; if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { /* found it */ if (netp != NULL) { *netp = net; } if (locked_tcb == NULL) { SCTP_INP_DECR_REF(inp); } else if (locked_tcb != stcb) { SCTP_TCB_LOCK(locked_tcb); } if (locked_tcb) { atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); return (stcb); } break; } #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6, *rsin6; sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; rsin6 = (struct sockaddr_in6 *)remote; if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { /* found it */ if (netp != NULL) { *netp = net; } if (locked_tcb == NULL) { SCTP_INP_DECR_REF(inp); } else if (locked_tcb != stcb) { SCTP_TCB_LOCK(locked_tcb); } if (locked_tcb) { atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); return (stcb); } break; } #endif default: /* TSNH */ break; } } SCTP_TCB_UNLOCK(stcb); } } else { SCTP_INP_WLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { goto null_return; } head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport, inp->sctp_hashmark)]; if (head == NULL) { goto null_return; } LIST_FOREACH(stcb, head, sctp_tcbhash) { if (stcb->rport != rport) { /* remote port does not match */ continue; } SCTP_TCB_LOCK(stcb); if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_UNLOCK(stcb); continue; } /* now look at the list of remote addresses */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { #ifdef INVARIANTS if (net == (TAILQ_NEXT(net, sctp_next))) { panic("Corrupt net list"); } #endif if (net->ro._l_addr.sa.sa_family != remote->sa_family) { /* not the same family */ continue; } switch (remote->sa_family) { case AF_INET: { struct sockaddr_in *sin, *rsin; sin = (struct sockaddr_in *) &net->ro._l_addr; rsin = (struct sockaddr_in *)remote; if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { /* found it */ if (netp != NULL) { *netp = net; } if (locked_tcb == NULL) { SCTP_INP_DECR_REF(inp); } else if (locked_tcb != stcb) { SCTP_TCB_LOCK(locked_tcb); } if (locked_tcb) { atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); return (stcb); } break; } #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6, *rsin6; sin6 = (struct sockaddr_in6 *) &net->ro._l_addr; rsin6 = (struct sockaddr_in6 *)remote; if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { /* found it */ if (netp != NULL) { *netp = net; } if (locked_tcb == NULL) { SCTP_INP_DECR_REF(inp); } else if (locked_tcb != stcb) { SCTP_TCB_LOCK(locked_tcb); } if (locked_tcb) { atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); return (stcb); } break; } #endif default: /* TSNH */ break; } } SCTP_TCB_UNLOCK(stcb); } } null_return: /* clean up for returning null */ if (locked_tcb) { SCTP_TCB_LOCK(locked_tcb); atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); /* not found */ return (NULL); } /* * Find an association for a specific endpoint using the association id given * out in the COMM_UP notification */ struct sctp_tcb * sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock) { /* * Use my the assoc_id to find a endpoint */ struct sctpasochead *head; struct sctp_tcb *stcb; uint32_t id; if (asoc_id == 0 || inp == NULL) { return (NULL); } SCTP_INP_INFO_RLOCK(); id = (uint32_t) asoc_id; head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(id, SCTP_BASE_INFO(hashasocmark))]; if (head == NULL) { /* invalid id TSNH */ SCTP_INP_INFO_RUNLOCK(); return (NULL); } LIST_FOREACH(stcb, head, sctp_asocs) { SCTP_INP_RLOCK(stcb->sctp_ep); if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(stcb->sctp_ep); SCTP_INP_INFO_RUNLOCK(); return (NULL); } if (stcb->asoc.assoc_id == id) { /* candidate */ if (inp != stcb->sctp_ep) { /* * some other guy has the same id active (id * collision ??). */ SCTP_INP_RUNLOCK(stcb->sctp_ep); continue; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { continue; } if (want_lock) { SCTP_TCB_LOCK(stcb); } SCTP_INP_RUNLOCK(stcb->sctp_ep); SCTP_INP_INFO_RUNLOCK(); return (stcb); } SCTP_INP_RUNLOCK(stcb->sctp_ep); } /* Ok if we missed here, lets try the restart hash */ head = &SCTP_BASE_INFO(sctp_restarthash)[SCTP_PCBHASH_ASOC(id, SCTP_BASE_INFO(hashrestartmark))]; if (head == NULL) { /* invalid id TSNH */ SCTP_INP_INFO_RUNLOCK(); return (NULL); } LIST_FOREACH(stcb, head, sctp_tcbrestarhash) { SCTP_INP_RLOCK(stcb->sctp_ep); if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(stcb->sctp_ep); continue; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { continue; } if (want_lock) { SCTP_TCB_LOCK(stcb); } if (stcb->asoc.assoc_id == id) { /* candidate */ SCTP_INP_RUNLOCK(stcb->sctp_ep); if (inp != stcb->sctp_ep) { /* * some other guy has the same id active (id * collision ??). */ if (want_lock) { SCTP_TCB_UNLOCK(stcb); } continue; } SCTP_INP_INFO_RUNLOCK(); return (stcb); } else { SCTP_INP_RUNLOCK(stcb->sctp_ep); } if (want_lock) { SCTP_TCB_UNLOCK(stcb); } } SCTP_INP_INFO_RUNLOCK(); return (NULL); } static struct sctp_inpcb * sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head, uint16_t lport, uint32_t vrf_id) { struct sctp_inpcb *inp; struct sockaddr_in *sin; #ifdef INET6 struct sockaddr_in6 *sin6; #endif struct sctp_laddr *laddr; #ifdef INET6 struct sockaddr_in6 *intf_addr6; #endif int fnd; /* * Endpoing probe expects that the INP_INFO is locked. */ sin = NULL; #ifdef INET6 sin6 = NULL; #endif switch (nam->sa_family) { case AF_INET: sin = (struct sockaddr_in *)nam; break; #ifdef INET6 case AF_INET6: sin6 = (struct sockaddr_in6 *)nam; break; #endif default: /* unsupported family */ return (NULL); } + if (head == NULL) return (NULL); LIST_FOREACH(inp, head, sctp_hash) { SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(inp); continue; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) && (inp->sctp_lport == lport)) { /* got it */ if ((nam->sa_family == AF_INET) && (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(inp)) { /* IPv4 on a IPv6 socket with ONLY IPv6 set */ SCTP_INP_RUNLOCK(inp); continue; } /* A V6 address and the endpoint is NOT bound V6 */ if (nam->sa_family == AF_INET6 && (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { SCTP_INP_RUNLOCK(inp); continue; } /* does a VRF id match? */ fnd = 0; if (inp->def_vrf_id == vrf_id) fnd = 1; SCTP_INP_RUNLOCK(inp); if (!fnd) continue; return (inp); } SCTP_INP_RUNLOCK(inp); } - if ((nam->sa_family == AF_INET) && (sin->sin_addr.s_addr == INADDR_ANY)) { /* Can't hunt for one that has no address specified */ return (NULL); } #ifdef INET6 if ((nam->sa_family == AF_INET6) && (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) { /* Can't hunt for one that has no address specified */ return (NULL); } #endif /* * ok, not bound to all so see if we can find a EP bound to this * address. */ LIST_FOREACH(inp, head, sctp_hash) { SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(inp); continue; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) { SCTP_INP_RUNLOCK(inp); continue; } /* * Ok this could be a likely candidate, look at all of its * addresses */ if (inp->sctp_lport != lport) { SCTP_INP_RUNLOCK(inp); continue; } /* does a VRF id match? */ fnd = 0; if (inp->def_vrf_id == vrf_id) fnd = 1; if (!fnd) { SCTP_INP_RUNLOCK(inp); continue; } LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__); continue; } SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ", laddr->ifa); if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n"); continue; } if (laddr->ifa->address.sa.sa_family == nam->sa_family) { /* possible, see if it matches */ struct sockaddr_in *intf_addr; intf_addr = &laddr->ifa->address.sin; switch (nam->sa_family) { case AF_INET: if (sin->sin_addr.s_addr == intf_addr->sin_addr.s_addr) { SCTP_INP_RUNLOCK(inp); return (inp); } break; #ifdef INET6 case AF_INET6: intf_addr6 = &laddr->ifa->address.sin6; if (SCTP6_ARE_ADDR_EQUAL(sin6, intf_addr6)) { SCTP_INP_RUNLOCK(inp); return (inp); } break; #endif } } } SCTP_INP_RUNLOCK(inp); } return (NULL); } + +static struct sctp_inpcb * +sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id) +{ + struct sctppcbhead *head; + struct sctp_inpcb *t_inp; + int fnd; + + head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, + SCTP_BASE_INFO(hashmark))]; + LIST_FOREACH(t_inp, head, sctp_hash) { + if (t_inp->sctp_lport != lport) { + continue; + } + /* is it in the VRF in question */ + fnd = 0; + if (t_inp->def_vrf_id == vrf_id) + fnd = 1; + if (!fnd) + continue; + + /* This one is in use. */ + /* check the v6/v4 binding issue */ + if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + SCTP_IPV6_V6ONLY(t_inp)) { + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + /* collision in V6 space */ + return (t_inp); + } else { + /* inp is BOUND_V4 no conflict */ + continue; + } + } else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + /* t_inp is bound v4 and v6, conflict always */ + return (t_inp); + } else { + /* t_inp is bound only V4 */ + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && + SCTP_IPV6_V6ONLY(inp)) { + /* no conflict */ + continue; + } + /* else fall through to conflict */ + } + return (t_inp); + } + return (NULL); +} + + +int +sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp) +{ + /* For 1-2-1 with port reuse */ + struct sctppcbhead *head; + struct sctp_inpcb *tinp; + + if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) { + /* only works with port reuse on */ + return (-1); + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) { + return (0); + } + SCTP_INP_RUNLOCK(inp); + head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, + SCTP_BASE_INFO(hashmark))]; + /* Kick out all non-listeners to the TCP hash */ + LIST_FOREACH(tinp, head, sctp_hash) { + if (tinp->sctp_lport != inp->sctp_lport) { + continue; + } + if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { + continue; + } + if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { + continue; + } + if (tinp->sctp_socket->so_qlimit) { + continue; + } + SCTP_INP_WLOCK(tinp); + LIST_REMOVE(tinp, sctp_hash); + head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))]; + tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL; + LIST_INSERT_HEAD(head, tinp, sctp_hash); + SCTP_INP_WUNLOCK(tinp); + } + SCTP_INP_WLOCK(inp); + /* Pull from where he was */ + LIST_REMOVE(inp, sctp_hash); + inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL; + head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))]; + LIST_INSERT_HEAD(head, inp, sctp_hash); + SCTP_INP_WUNLOCK(inp); + SCTP_INP_RLOCK(inp); + return (0); +} + + struct sctp_inpcb * sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock, uint32_t vrf_id) { /* * First we check the hash table to see if someone has this port * bound with just the port. */ struct sctp_inpcb *inp; struct sctppcbhead *head; struct sockaddr_in *sin; struct sockaddr_in6 *sin6; int lport; if (nam->sa_family == AF_INET) { sin = (struct sockaddr_in *)nam; lport = ((struct sockaddr_in *)nam)->sin_port; } else if (nam->sa_family == AF_INET6) { sin6 = (struct sockaddr_in6 *)nam; lport = ((struct sockaddr_in6 *)nam)->sin6_port; } else { /* unsupported family */ return (NULL); } /* * I could cheat here and just cast to one of the types but we will * do it right. It also provides the check against an Unsupported * type too. */ /* Find the head of the ALLADDR chain */ if (have_lock == 0) { SCTP_INP_INFO_RLOCK(); } head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))]; inp = sctp_endpoint_probe(nam, head, lport, vrf_id); /* * If the TCP model exists it could be that the main listening * endpoint is gone but there exists a connected socket for this guy * yet. If so we can return the first one that we find. This may NOT * be the correct one but the sctp_findassociation_ep_addr has * further code to look at all TCP models. */ if (inp == NULL && find_tcp_pool) { - unsigned int i; - - for (i = 0; i < SCTP_BASE_INFO(hashtblsize); i++) { - /* - * This is real gross, but we do NOT have a remote - * port at this point depending on who is calling. - * We must therefore look for ANY one that matches - * our local port :/ - */ - head = &SCTP_BASE_INFO(sctp_tcpephash)[i]; - if (LIST_FIRST(head)) { - inp = sctp_endpoint_probe(nam, head, lport, vrf_id); - if (inp) { - /* Found one */ - break; - } - } - } + head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))]; + inp = sctp_endpoint_probe(nam, head, lport, vrf_id); } if (inp) { SCTP_INP_INCR_REF(inp); } if (have_lock == 0) { SCTP_INP_INFO_RUNLOCK(); } return (inp); } /* * Find an association for an endpoint with the pointer to whom you want to * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may * need to change the *to to some other struct like a mbuf... */ struct sctp_tcb * sctp_findassociation_addr_sa(struct sockaddr *to, struct sockaddr *from, struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool, uint32_t vrf_id) { struct sctp_inpcb *inp = NULL; struct sctp_tcb *retval; SCTP_INP_INFO_RLOCK(); if (find_tcp_pool) { if (inp_p != NULL) { retval = sctp_tcb_special_locate(inp_p, from, to, netp, vrf_id); } else { retval = sctp_tcb_special_locate(&inp, from, to, netp, vrf_id); } if (retval != NULL) { SCTP_INP_INFO_RUNLOCK(); return (retval); } } inp = sctp_pcb_findep(to, 0, 1, vrf_id); if (inp_p != NULL) { *inp_p = inp; } SCTP_INP_INFO_RUNLOCK(); if (inp == NULL) { return (NULL); } /* * ok, we have an endpoint, now lets find the assoc for it (if any) * we now place the source address or from in the to of the find * endpoint call. Since in reality this chain is used from the * inbound packet side. */ if (inp_p != NULL) { retval = sctp_findassociation_ep_addr(inp_p, from, netp, to, NULL); } else { retval = sctp_findassociation_ep_addr(&inp, from, netp, to, NULL); } return retval; } /* * This routine will grub through the mbuf that is a INIT or INIT-ACK and * find all addresses that the sender has specified in any address list. Each * address will be used to lookup the TCB and see if one exits. */ static struct sctp_tcb * sctp_findassociation_special_addr(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp, struct sockaddr *dest) { struct sockaddr_in sin4; struct sockaddr_in6 sin6; struct sctp_paramhdr *phdr, parm_buf; struct sctp_tcb *retval; uint32_t ptype, plen; memset(&sin4, 0, sizeof(sin4)); memset(&sin6, 0, sizeof(sin6)); sin4.sin_len = sizeof(sin4); sin4.sin_family = AF_INET; sin4.sin_port = sh->src_port; sin6.sin6_len = sizeof(sin6); sin6.sin6_family = AF_INET6; sin6.sin6_port = sh->src_port; retval = NULL; offset += sizeof(struct sctp_init_chunk); phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf)); while (phdr != NULL) { /* now we must see if we want the parameter */ ptype = ntohs(phdr->param_type); plen = ntohs(phdr->param_length); if (plen == 0) { break; } if (ptype == SCTP_IPV4_ADDRESS && plen == sizeof(struct sctp_ipv4addr_param)) { /* Get the rest of the address */ struct sctp_ipv4addr_param ip4_parm, *p4; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm))); if (phdr == NULL) { return (NULL); } p4 = (struct sctp_ipv4addr_param *)phdr; memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr)); /* look it up */ retval = sctp_findassociation_ep_addr(inp_p, (struct sockaddr *)&sin4, netp, dest, NULL); if (retval != NULL) { return (retval); } } else if (ptype == SCTP_IPV6_ADDRESS && plen == sizeof(struct sctp_ipv6addr_param)) { /* Get the rest of the address */ struct sctp_ipv6addr_param ip6_parm, *p6; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm))); if (phdr == NULL) { return (NULL); } p6 = (struct sctp_ipv6addr_param *)phdr; memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr)); /* look it up */ retval = sctp_findassociation_ep_addr(inp_p, (struct sockaddr *)&sin6, netp, dest, NULL); if (retval != NULL) { return (retval); } } offset += SCTP_SIZE32(plen); phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf)); } return (NULL); } static struct sctp_tcb * sctp_findassoc_by_vtag(struct sockaddr *from, uint32_t vtag, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport, uint16_t lport, int skip_src_check) { /* * Use my vtag to hash. If we find it we then verify the source addr * is in the assoc. If all goes well we save a bit on rec of a * packet. */ struct sctpasochead *head; struct sctp_nets *net; struct sctp_tcb *stcb; *netp = NULL; *inp_p = NULL; SCTP_INP_INFO_RLOCK(); head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag, SCTP_BASE_INFO(hashasocmark))]; if (head == NULL) { /* invalid vtag */ SCTP_INP_INFO_RUNLOCK(); return (NULL); } LIST_FOREACH(stcb, head, sctp_asocs) { SCTP_INP_RLOCK(stcb->sctp_ep); if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(stcb->sctp_ep); continue; } SCTP_TCB_LOCK(stcb); SCTP_INP_RUNLOCK(stcb->sctp_ep); if (stcb->asoc.my_vtag == vtag) { /* candidate */ if (stcb->rport != rport) { /* * we could remove this if vtags are unique * across the system. */ SCTP_TCB_UNLOCK(stcb); continue; } if (stcb->sctp_ep->sctp_lport != lport) { /* * we could remove this if vtags are unique * across the system. */ SCTP_TCB_UNLOCK(stcb); continue; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_UNLOCK(stcb); continue; } if (skip_src_check) { *netp = NULL; /* unknown */ if (inp_p) *inp_p = stcb->sctp_ep; SCTP_INP_INFO_RUNLOCK(); return (stcb); } net = sctp_findnet(stcb, from); if (net) { /* yep its him. */ *netp = net; SCTP_STAT_INCR(sctps_vtagexpress); *inp_p = stcb->sctp_ep; SCTP_INP_INFO_RUNLOCK(); return (stcb); } else { /* * not him, this should only happen in rare * cases so I peg it. */ SCTP_STAT_INCR(sctps_vtagbogus); } } SCTP_TCB_UNLOCK(stcb); } SCTP_INP_INFO_RUNLOCK(); return (NULL); } /* * Find an association with the pointer to the inbound IP packet. This can be * a IPv4 or IPv6 packet. */ struct sctp_tcb * sctp_findassociation_addr(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_chunkhdr *ch, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id) { int find_tcp_pool; struct ip *iph; struct sctp_tcb *retval; struct sockaddr_storage to_store, from_store; struct sockaddr *to = (struct sockaddr *)&to_store; struct sockaddr *from = (struct sockaddr *)&from_store; struct sctp_inpcb *inp; iph = mtod(m, struct ip *); switch (iph->ip_v) { case IPVERSION: { /* its IPv4 */ struct sockaddr_in *from4; from4 = (struct sockaddr_in *)&from_store; bzero(from4, sizeof(*from4)); from4->sin_family = AF_INET; from4->sin_len = sizeof(struct sockaddr_in); from4->sin_addr.s_addr = iph->ip_src.s_addr; from4->sin_port = sh->src_port; break; } #ifdef INET6 case IPV6_VERSION >> 4: { /* its IPv6 */ struct ip6_hdr *ip6; struct sockaddr_in6 *from6; ip6 = mtod(m, struct ip6_hdr *); from6 = (struct sockaddr_in6 *)&from_store; bzero(from6, sizeof(*from6)); from6->sin6_family = AF_INET6; from6->sin6_len = sizeof(struct sockaddr_in6); from6->sin6_addr = ip6->ip6_src; from6->sin6_port = sh->src_port; /* Get the scopes in properly to the sin6 addr's */ /* we probably don't need these operations */ (void)sa6_recoverscope(from6); sa6_embedscope(from6, MODULE_GLOBAL(MOD_INET6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone))); break; } #endif default: /* Currently not supported. */ return (NULL); } if (sh->v_tag) { /* we only go down this path if vtag is non-zero */ retval = sctp_findassoc_by_vtag(from, ntohl(sh->v_tag), inp_p, netp, sh->src_port, sh->dest_port, 0); if (retval) { return (retval); } } switch (iph->ip_v) { case IPVERSION: { /* its IPv4 */ struct sockaddr_in *to4; to4 = (struct sockaddr_in *)&to_store; bzero(to4, sizeof(*to4)); to4->sin_family = AF_INET; to4->sin_len = sizeof(struct sockaddr_in); to4->sin_addr.s_addr = iph->ip_dst.s_addr; to4->sin_port = sh->dest_port; break; } #ifdef INET6 case IPV6_VERSION >> 4: { /* its IPv6 */ struct ip6_hdr *ip6; struct sockaddr_in6 *to6; ip6 = mtod(m, struct ip6_hdr *); to6 = (struct sockaddr_in6 *)&to_store; bzero(to6, sizeof(*to6)); to6->sin6_family = AF_INET6; to6->sin6_len = sizeof(struct sockaddr_in6); to6->sin6_addr = ip6->ip6_dst; to6->sin6_port = sh->dest_port; /* Get the scopes in properly to the sin6 addr's */ /* we probably don't need these operations */ (void)sa6_recoverscope(to6); sa6_embedscope(to6, MODULE_GLOBAL(MOD_INET6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone))); break; } #endif default: /* TSNH */ break; } find_tcp_pool = 0; if ((ch->chunk_type != SCTP_INITIATION) && (ch->chunk_type != SCTP_INITIATION_ACK) && (ch->chunk_type != SCTP_COOKIE_ACK) && (ch->chunk_type != SCTP_COOKIE_ECHO)) { /* Other chunk types go to the tcp pool. */ find_tcp_pool = 1; } if (inp_p) { retval = sctp_findassociation_addr_sa(to, from, inp_p, netp, find_tcp_pool, vrf_id); inp = *inp_p; } else { retval = sctp_findassociation_addr_sa(to, from, &inp, netp, find_tcp_pool, vrf_id); } SCTPDBG(SCTP_DEBUG_PCB1, "retval:%p inp:%p\n", retval, inp); if (retval == NULL && inp) { /* Found a EP but not this address */ if ((ch->chunk_type == SCTP_INITIATION) || (ch->chunk_type == SCTP_INITIATION_ACK)) { /*- * special hook, we do NOT return linp or an * association that is linked to an existing * association that is under the TCP pool (i.e. no * listener exists). The endpoint finding routine * will always find a listner before examining the * TCP pool. */ if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) { if (inp_p) { *inp_p = NULL; } return (NULL); } retval = sctp_findassociation_special_addr(m, iphlen, offset, sh, &inp, netp, to); if (inp_p != NULL) { *inp_p = inp; } } } SCTPDBG(SCTP_DEBUG_PCB1, "retval is %p\n", retval); return (retval); } /* * lookup an association by an ASCONF lookup address. * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup */ struct sctp_tcb * sctp_findassociation_ep_asconf(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp) { struct sctp_tcb *stcb; struct sockaddr_in *sin; #ifdef INET6 struct sockaddr_in6 *sin6; #endif struct sockaddr_storage local_store, remote_store; struct ip *iph; #ifdef INET6 struct ip6_hdr *ip6; #endif struct sctp_paramhdr parm_buf, *phdr; int ptype; int zero_address = 0; memset(&local_store, 0, sizeof(local_store)); memset(&remote_store, 0, sizeof(remote_store)); /* First get the destination address setup too. */ iph = mtod(m, struct ip *); switch (iph->ip_v) { case IPVERSION: /* its IPv4 */ sin = (struct sockaddr_in *)&local_store; sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_port = sh->dest_port; sin->sin_addr.s_addr = iph->ip_dst.s_addr; break; #ifdef INET6 case IPV6_VERSION >> 4: /* its IPv6 */ ip6 = mtod(m, struct ip6_hdr *); sin6 = (struct sockaddr_in6 *)&local_store; sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(*sin6); sin6->sin6_port = sh->dest_port; sin6->sin6_addr = ip6->ip6_dst; break; #endif default: return NULL; } phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk), &parm_buf, sizeof(struct sctp_paramhdr)); if (phdr == NULL) { SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n", __FUNCTION__); return NULL; } ptype = (int)((uint32_t) ntohs(phdr->param_type)); /* get the correlation address */ switch (ptype) { #ifdef INET6 case SCTP_IPV6_ADDRESS: { /* ipv6 address param */ struct sctp_ipv6addr_param *p6, p6_buf; if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) { return NULL; } p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk), &p6_buf.ph, sizeof(*p6)); if (p6 == NULL) { SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n", __FUNCTION__); return (NULL); } sin6 = (struct sockaddr_in6 *)&remote_store; sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(*sin6); sin6->sin6_port = sh->src_port; memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr)); if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) zero_address = 1; break; } #endif case SCTP_IPV4_ADDRESS: { /* ipv4 address param */ struct sctp_ipv4addr_param *p4, p4_buf; if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) { return NULL; } p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk), &p4_buf.ph, sizeof(*p4)); if (p4 == NULL) { SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n", __FUNCTION__); return (NULL); } sin = (struct sockaddr_in *)&remote_store; sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_port = sh->src_port; memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr)); if (sin->sin_addr.s_addr == INADDR_ANY) zero_address = 1; break; } default: /* invalid address param type */ return NULL; } if (zero_address) { stcb = sctp_findassoc_by_vtag(NULL, ntohl(sh->v_tag), inp_p, netp, sh->src_port, sh->dest_port, 1); /* * printf("findassociation_ep_asconf: zero lookup address * finds stcb 0x%x\n", (uint32_t)stcb); */ } else { stcb = sctp_findassociation_ep_addr(inp_p, (struct sockaddr *)&remote_store, netp, (struct sockaddr *)&local_store, NULL); } return (stcb); } /* * allocate a sctp_inpcb and setup a temporary binding to a port/all * addresses. This way if we don't get a bind we by default pick a ephemeral * port with all addresses bound. */ int sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id) { /* * we get called when a new endpoint starts up. We need to allocate * the sctp_inpcb structure from the zone and init it. Mark it as * unbound and find a port that we can use as an ephemeral with * INADDR_ANY. If the user binds later no problem we can then add in * the specific addresses. And setup the default parameters for the * EP. */ int i, error; struct sctp_inpcb *inp; struct sctp_pcb *m; struct timeval time; sctp_sharedkey_t *null_key; error = 0; SCTP_INP_INFO_WLOCK(); inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb); if (inp == NULL) { SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n"); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); return (ENOBUFS); } /* zap it */ bzero(inp, sizeof(*inp)); /* bump generations */ /* setup socket pointers */ inp->sctp_socket = so; inp->ip_inp.inp.inp_socket = so; inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT; inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT; #ifdef IPSEC { struct inpcbpolicy *pcb_sp = NULL; error = ipsec_init_policy(so, &pcb_sp); /* Arrange to share the policy */ inp->ip_inp.inp.inp_sp = pcb_sp; ((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp; } if (error != 0) { SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); SCTP_INP_INFO_WUNLOCK(); return error; } #endif /* IPSEC */ SCTP_INCR_EP_COUNT(); inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(MOD_INET, ip_defttl); SCTP_INP_INFO_WUNLOCK(); so->so_pcb = (caddr_t)inp; if ((SCTP_SO_TYPE(so) == SOCK_DGRAM) || (SCTP_SO_TYPE(so) == SOCK_SEQPACKET)) { /* UDP style socket */ inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE | SCTP_PCB_FLAGS_UNBOUND); /* Be sure it is NON-BLOCKING IO for UDP */ /* SCTP_SET_SO_NBIO(so); */ } else if (SCTP_SO_TYPE(so) == SOCK_STREAM) { /* TCP style socket */ inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE | SCTP_PCB_FLAGS_UNBOUND); /* Be sure we have blocking IO by default */ SCTP_CLEAR_SO_NBIO(so); } else { /* * unsupported socket type (RAW, etc)- in case we missed it * in protosw */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); return (EOPNOTSUPP); } if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) { sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) { sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) { sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize), &inp->sctp_hashmark); if (inp->sctp_tcbhash == NULL) { SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n"); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); return (ENOBUFS); } inp->def_vrf_id = vrf_id; SCTP_INP_INFO_WLOCK(); SCTP_INP_LOCK_INIT(inp); INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp"); SCTP_INP_READ_INIT(inp); SCTP_ASOC_CREATE_LOCK_INIT(inp); /* lock the new ep */ SCTP_INP_WLOCK(inp); /* add it to the info area */ LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list); SCTP_INP_INFO_WUNLOCK(); TAILQ_INIT(&inp->read_queue); LIST_INIT(&inp->sctp_addr_list); LIST_INIT(&inp->sctp_asoc_list); #ifdef SCTP_TRACK_FREED_ASOCS /* TEMP CODE */ LIST_INIT(&inp->sctp_asoc_free_list); #endif /* Init the timer structure for signature change */ SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer); inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE; /* now init the actual endpoint default data */ m = &inp->sctp_ep; /* setup the base timeout information */ m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC); /* needed ? */ m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC); /* needed ? */ m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default)); m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default)); m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default)); m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default)); m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default)); /* all max/min max are in ms */ m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default); m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default); m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default); m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default); m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default); m->max_open_streams_intome = MAX_SCTP_STREAMS; m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default); m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default); m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default); m->sctp_sws_sender = SCTP_SWS_SENDER_DEF; m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF; m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default); if ((SCTP_BASE_SYSCTL(sctp_default_cc_module) >= SCTP_CC_RFC2581) && (SCTP_BASE_SYSCTL(sctp_default_cc_module) <= SCTP_CC_HTCP)) { m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module); } else { /* sysctl done with invalid value, set to 2581 */ m->sctp_default_cc_module = SCTP_CC_RFC2581; } /* number of streams to pre-open on a association */ m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default); /* Add adaptation cookie */ m->adaptation_layer_indicator = 0x504C5253; /* seed random number generator */ m->random_counter = 1; m->store_at = SCTP_SIGNATURE_SIZE; SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers)); sctp_fill_random_store(m); /* Minimum cookie size */ m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) + sizeof(struct sctp_state_cookie); m->size_of_a_cookie += SCTP_SIGNATURE_SIZE; /* Setup the initial secret */ (void)SCTP_GETTIME_TIMEVAL(&time); m->time_of_secret_change = time.tv_sec; for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) { m->secret_key[0][i] = sctp_select_initial_TSN(m); } sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL); /* How long is a cookie good for ? */ m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default)); /* * Initialize authentication parameters */ m->local_hmacs = sctp_default_supported_hmaclist(); m->local_auth_chunks = sctp_alloc_chunklist(); sctp_auth_set_default_chunks(m->local_auth_chunks); LIST_INIT(&m->shared_keys); /* add default NULL key as key id 0 */ null_key = sctp_alloc_sharedkey(); sctp_insert_sharedkey(&m->shared_keys, null_key); SCTP_INP_WUNLOCK(inp); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 12); #endif return (error); } void sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp, struct sctp_tcb *stcb) { struct sctp_nets *net; uint16_t lport, rport; struct sctppcbhead *head; struct sctp_laddr *laddr, *oladdr; atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(old_inp); SCTP_INP_WLOCK(new_inp); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); new_inp->sctp_ep.time_of_secret_change = old_inp->sctp_ep.time_of_secret_change; memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key, sizeof(old_inp->sctp_ep.secret_key)); new_inp->sctp_ep.current_secret_number = old_inp->sctp_ep.current_secret_number; new_inp->sctp_ep.last_secret_number = old_inp->sctp_ep.last_secret_number; new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie; /* make it so new data pours into the new socket */ stcb->sctp_socket = new_inp->sctp_socket; stcb->sctp_ep = new_inp; /* Copy the port across */ lport = new_inp->sctp_lport = old_inp->sctp_lport; rport = stcb->rport; /* Pull the tcb from the old association */ LIST_REMOVE(stcb, sctp_tcbhash); LIST_REMOVE(stcb, sctp_tcblist); /* Now insert the new_inp into the TCP connected hash */ - head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport + rport), + head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport), SCTP_BASE_INFO(hashtcpmark))]; LIST_INSERT_HEAD(head, new_inp, sctp_hash); /* Its safe to access */ new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND; /* Now move the tcb into the endpoint list */ LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist); /* * Question, do we even need to worry about the ep-hash since we * only have one connection? Probably not :> so lets get rid of it * and not suck up any kernel memory in that. */ /* Ok. Let's restart timer. */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp, stcb, net); } SCTP_INP_INFO_WUNLOCK(); if (new_inp->sctp_tcbhash != NULL) { SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark); new_inp->sctp_tcbhash = NULL; } if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { /* Subset bound, so copy in the laddr list from the old_inp */ LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) { laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (laddr == NULL) { /* * Gak, what can we do? This assoc is really * HOSED. We probably should send an abort * here. */ SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n"); continue; } SCTP_INCR_LADDR_COUNT(); bzero(laddr, sizeof(*laddr)); (void)SCTP_GETTIME_TIMEVAL(&laddr->start_time); laddr->ifa = oladdr->ifa; atomic_add_int(&laddr->ifa->refcount, 1); LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr, sctp_nxt_addr); new_inp->laddr_count++; } } /* * Now any running timers need to be adjusted since we really don't * care if they are running or not just blast in the new_inp into * all of them. */ stcb->asoc.hb_timer.ep = (void *)new_inp; stcb->asoc.dack_timer.ep = (void *)new_inp; stcb->asoc.asconf_timer.ep = (void *)new_inp; stcb->asoc.strreset_timer.ep = (void *)new_inp; stcb->asoc.shut_guard_timer.ep = (void *)new_inp; stcb->asoc.autoclose_timer.ep = (void *)new_inp; stcb->asoc.delayed_event_timer.ep = (void *)new_inp; stcb->asoc.delete_prim_timer.ep = (void *)new_inp; /* now what about the nets? */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { net->pmtu_timer.ep = (void *)new_inp; net->rxt_timer.ep = (void *)new_inp; net->fr_timer.ep = (void *)new_inp; } SCTP_INP_WUNLOCK(new_inp); SCTP_INP_WUNLOCK(old_inp); } -static int -sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id) -{ - struct sctppcbhead *head; - struct sctp_inpcb *t_inp; - int fnd; - head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, - SCTP_BASE_INFO(hashmark))]; - LIST_FOREACH(t_inp, head, sctp_hash) { - if (t_inp->sctp_lport != lport) { - continue; - } - /* is it in the VRF in question */ - fnd = 0; - if (t_inp->def_vrf_id == vrf_id) - fnd = 1; - if (!fnd) - continue; - /* This one is in use. */ - /* check the v6/v4 binding issue */ - if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && - SCTP_IPV6_V6ONLY(t_inp)) { - if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { - /* collision in V6 space */ - return (1); - } else { - /* inp is BOUND_V4 no conflict */ - continue; - } - } else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { - /* t_inp is bound v4 and v6, conflict always */ - return (1); - } else { - /* t_inp is bound only V4 */ - if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && - SCTP_IPV6_V6ONLY(inp)) { - /* no conflict */ - continue; - } - /* else fall through to conflict */ - } - return (1); - } - return (0); -} - - /* sctp_ifap is used to bypass normal local address validation checks */ int sctp_inpcb_bind(struct socket *so, struct sockaddr *addr, struct sctp_ifa *sctp_ifap, struct thread *p) { /* bind a ep to a socket address */ struct sctppcbhead *head; struct sctp_inpcb *inp, *inp_tmp; struct inpcb *ip_inp; + int port_reuse_active = 0; int bindall; int prison = 0; uint16_t lport; int error; uint32_t vrf_id; lport = 0; error = 0; bindall = 1; inp = (struct sctp_inpcb *)so->so_pcb; ip_inp = (struct inpcb *)so->so_pcb; #ifdef SCTP_DEBUG if (addr) { SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port:%d\n", ntohs(((struct sockaddr_in *)addr)->sin_port)); SCTPDBG(SCTP_DEBUG_PCB1, "Addr :"); SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr); } #endif if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) { /* already did a bind, subsequent binds NOT allowed ! */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } #ifdef INVARIANTS if (p == NULL) panic("null proc/thread"); #endif if (p && jailed(p->td_ucred)) { prison = 1; } if (addr != NULL) { switch (addr->sa_family) { case AF_INET: { struct sockaddr_in *sin; /* IPV6_V6ONLY socket? */ if (SCTP_IPV6_V6ONLY(ip_inp)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } if (addr->sa_len != sizeof(*sin)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } sin = (struct sockaddr_in *)addr; lport = sin->sin_port; if (prison) { /* * For INADDR_ANY and LOOPBACK the * prison_ip() call will transmute * the ip address to the proper * value (i.e. the IP address owned * by the jail). */ if (prison_ip(p->td_ucred, 0, &sin->sin_addr.s_addr)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } } if (sin->sin_addr.s_addr != INADDR_ANY) { bindall = 0; } break; } #ifdef INET6 case AF_INET6: { /* * Only for pure IPv6 Address. (No IPv4 * Mapped!) */ struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)addr; if (addr->sa_len != sizeof(*sin6)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } lport = sin6->sin6_port; /* * Jail checks for IPv6 should go HERE! i.e. * add the prison_ip() equivilant in this * postion to transmute the addresses to the * proper one jailed. */ if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { bindall = 0; /* KAME hack: embed scopeid */ if (sa6_embedscope(sin6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone)) != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } } /* this must be cleared for ifa_ifwithaddr() */ sin6->sin6_scope_id = 0; break; } #endif default: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT); return (EAFNOSUPPORT); } } SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(inp); /* Setup a vrf_id to be the default for the non-bind-all case. */ vrf_id = inp->def_vrf_id; /* increase our count due to the unlock we do */ SCTP_INP_INCR_REF(inp); if (lport) { /* * Did the caller specify a port? if so we must see if a ep * already has this one bound. */ /* got to be root to get at low ports */ if (ntohs(lport) < IPPORT_RESERVED) { if (p && (error = priv_check(p, PRIV_NETINET_RESERVEDPORT) )) { SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return (error); } } if (p == NULL) { SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); return (error); } SCTP_INP_WUNLOCK(inp); if (bindall) { vrf_id = inp->def_vrf_id; inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id); if (inp_tmp != NULL) { /* * lock guy returned and lower count note * that we are not bound so inp_tmp should * NEVER be inp. And it is this inp * (inp_tmp) that gets the reference bump, * so we must lower it. */ SCTP_INP_DECR_REF(inp_tmp); - SCTP_INP_DECR_REF(inp); /* unlock info */ + if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && + (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { + /* + * Ok, must be one-2-one and + * allowing port re-use + */ + port_reuse_active = 1; + goto continue_anyway; + } + SCTP_INP_DECR_REF(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); return (EADDRINUSE); } } else { inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id); if (inp_tmp != NULL) { /* * lock guy returned and lower count note * that we are not bound so inp_tmp should * NEVER be inp. And it is this inp * (inp_tmp) that gets the reference bump, * so we must lower it. */ SCTP_INP_DECR_REF(inp_tmp); - SCTP_INP_DECR_REF(inp); /* unlock info */ + if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && + (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { + /* + * Ok, must be one-2-one and + * allowing port re-use + */ + port_reuse_active = 1; + goto continue_anyway; + } + SCTP_INP_DECR_REF(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); return (EADDRINUSE); } } +continue_anyway: SCTP_INP_WLOCK(inp); if (bindall) { /* verify that no lport is not used by a singleton */ - if (sctp_isport_inuse(inp, lport, vrf_id)) { + if ((port_reuse_active == 0) && + (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id)) + ) { /* Sorry someone already has this one bound */ - SCTP_INP_DECR_REF(inp); - SCTP_INP_WUNLOCK(inp); - SCTP_INP_INFO_WUNLOCK(); - SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); - return (EADDRINUSE); + if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && + (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { + port_reuse_active = 1; + } else { + SCTP_INP_DECR_REF(inp); + SCTP_INP_WUNLOCK(inp); + SCTP_INP_INFO_WUNLOCK(); + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); + return (EADDRINUSE); + } } } } else { uint16_t first, last, candidate; uint16_t count; int done; if (ip_inp->inp_flags & INP_HIGHPORT) { first = MODULE_GLOBAL(MOD_INET, ipport_hifirstauto); last = MODULE_GLOBAL(MOD_INET, ipport_hilastauto); } else if (ip_inp->inp_flags & INP_LOWPORT) { if (p && (error = priv_check(p, PRIV_NETINET_RESERVEDPORT) )) { SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); return (error); } first = MODULE_GLOBAL(MOD_INET, ipport_lowfirstauto); last = MODULE_GLOBAL(MOD_INET, ipport_lowlastauto); } else { first = ipport_firstauto; last = ipport_lastauto; } if (first > last) { uint16_t temp; temp = first; first = last; last = temp; } count = last - first + 1; /* number of candidates */ candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count); done = 0; while (!done) { - if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == 0) { + if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) { done = 1; } if (!done) { if (--count == 0) { SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); return (EADDRINUSE); } if (candidate == last) candidate = first; else candidate = candidate + 1; } } lport = htons(candidate); } SCTP_INP_DECR_REF(inp); if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { /* * this really should not happen. The guy did a non-blocking * bind and then did a close at the same time. */ SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } /* ok we look clear to give out this port, so lets setup the binding */ if (bindall) { /* binding to all addresses, so just set in the proper flags */ inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL; /* set the automatic addr changes from kernel flag */ if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) { sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF); sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); } else { sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF); sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); } if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) { sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS); } else { sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS); } /* * set the automatic mobility_base from kernel flag (by * micchie) */ if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) { sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE); sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); } else { sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE); sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); } /* * set the automatic mobility_fasthandoff from kernel flag * (by micchie) */ if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) { sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF); sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); } else { sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF); sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); } } else { /* * bind specific, make sure flags is off and add a new * address structure to the sctp_addr_list inside the ep * structure. * * We will need to allocate one and insert it at the head. The * socketopt call can just insert new addresses in there as * well. It will also have to do the embed scope kame hack * too (before adding). */ struct sctp_ifa *ifa; struct sockaddr_storage store_sa; memset(&store_sa, 0, sizeof(store_sa)); if (addr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)&store_sa; memcpy(sin, addr, sizeof(struct sockaddr_in)); sin->sin_port = 0; } else if (addr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&store_sa; memcpy(sin6, addr, sizeof(struct sockaddr_in6)); sin6->sin6_port = 0; } /* * first find the interface with the bound address need to * zero out the port to find the address! yuck! can't do * this earlier since need port for sctp_pcb_findep() */ if (sctp_ifap != NULL) ifa = sctp_ifap; else { /* * Note for BSD we hit here always other O/S's will * pass things in via the sctp_ifap argument * (Panda). */ ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa, vrf_id, SCTP_ADDR_NOT_LOCKED); } if (ifa == NULL) { /* Can't find an interface with that address */ SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL); return (EADDRNOTAVAIL); } if (addr->sa_family == AF_INET6) { /* GAK, more FIXME IFA lock? */ if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { /* Can't bind a non-existent addr. */ SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } } /* we're not bound all */ inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL; /* allow bindx() to send ASCONF's for binding changes */ sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF); /* clear automatic addr changes from kernel flag */ sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); /* add this address to the endpoint list */ error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0); if (error != 0) { SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return (error); } inp->laddr_count++; } /* find the bucket */ - head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, - SCTP_BASE_INFO(hashmark))]; + if (port_reuse_active) { + /* Put it into tcp 1-2-1 hash */ + head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport), + SCTP_BASE_INFO(hashtcpmark))]; + inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL; + } else { + head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, + SCTP_BASE_INFO(hashmark))]; + } /* put it in the bucket */ LIST_INSERT_HEAD(head, inp, sctp_hash); - SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d\n", - head, ntohs(lport)); + SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n", + head, ntohs(lport), port_reuse_active); /* set in the port */ inp->sctp_lport = lport; /* turn off just the unbound flag */ inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND; SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return (0); } static void sctp_iterator_inp_being_freed(struct sctp_inpcb *inp, struct sctp_inpcb *inp_next) { struct sctp_iterator *it; /* * We enter with the only the ITERATOR_LOCK in place and a write * lock on the inp_info stuff. */ /* * Go through all iterators, we must do this since it is possible * that some iterator does NOT have the lock, but is waiting for it. * And the one that had the lock has either moved in the last * iteration or we just cleared it above. We need to find all of * those guys. The list of iterators should never be very big * though. */ TAILQ_FOREACH(it, &SCTP_BASE_INFO(iteratorhead), sctp_nxt_itr) { if (it == inp->inp_starting_point_for_iterator) /* skip this guy, he's special */ continue; if (it->inp == inp) { /* * This is tricky and we DON'T lock the iterator. * Reason is he's running but waiting for me since * inp->inp_starting_point_for_iterator has the lock * on me (the guy above we skipped). This tells us * its is not running but waiting for * inp->inp_starting_point_for_iterator to be * released by the guy that does have our INP in a * lock. */ if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { it->inp = NULL; it->stcb = NULL; } else { /* set him up to do the next guy not me */ it->inp = inp_next; it->stcb = NULL; } } } it = inp->inp_starting_point_for_iterator; if (it) { if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { it->inp = NULL; } else { it->inp = inp_next; } it->stcb = NULL; } } /* release sctp_inpcb unbind the port */ void sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from) { /* * Here we free a endpoint. We must find it (if it is in the Hash * table) and remove it from there. Then we must also find it in the * overall list and remove it from there. After all removals are * complete then any timer has to be stopped. Then start the actual * freeing. a) Any local lists. b) Any associations. c) The hash of * all associations. d) finally the ep itself. */ struct sctp_pcb *m; struct sctp_inpcb *inp_save; struct sctp_tcb *asoc, *nasoc; struct sctp_laddr *laddr, *nladdr; struct inpcb *ip_pcb; struct socket *so; struct sctp_queued_to_read *sq; int cnt; sctp_sharedkey_t *shared_key; #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 0); #endif SCTP_ITERATOR_LOCK(); so = inp->sctp_socket; if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { /* been here before.. eeks.. get out of here */ SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate); SCTP_ITERATOR_UNLOCK(); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 1); #endif return; } SCTP_ASOC_CREATE_LOCK(inp); SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(inp); /* First time through we have the socket lock, after that no more. */ if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) { /* * Once we are in we can remove the flag from = 1 is only * passed from the actual closing routines that are called * via the sockets layer. */ inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP; /* socket is gone, so no more wakeups allowed */ inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE; inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT; inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT; } sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL, SCTP_FROM_SCTP_PCB + SCTP_LOC_1); if (inp->control) { sctp_m_freem(inp->control); inp->control = NULL; } if (inp->pkt) { sctp_m_freem(inp->pkt); inp->pkt = NULL; } m = &inp->sctp_ep; ip_pcb = &inp->ip_inp.inp; /* we could just cast the main pointer * here but I will be nice :> (i.e. * ip_pcb = ep;) */ if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) { int cnt_in_sd; cnt_in_sd = 0; for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL; asoc = nasoc) { SCTP_TCB_LOCK(asoc); nasoc = LIST_NEXT(asoc, sctp_tcblist); if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { /* Skip guys being freed */ /* asoc->sctp_socket = NULL; FIXME MT */ cnt_in_sd++; SCTP_TCB_UNLOCK(asoc); continue; } if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) { /* * If we have data in queue, we don't want * to just free since the app may have done, * send()/close or connect/send/close. And * it wants the data to get across first. */ /* Just abandon things in the front states */ if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) { cnt_in_sd++; } continue; } /* Disconnect the socket please */ asoc->sctp_socket = NULL; asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET; if ((asoc->asoc.size_on_reasm_queue > 0) || (asoc->asoc.control_pdapi) || (asoc->asoc.size_on_all_streams > 0) || (so && (so->so_rcv.sb_cc > 0)) ) { /* Left with Data unread */ struct mbuf *op_err; op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { /* Fill in the user initiated abort */ struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(op_err) = sizeof(struct sctp_paramhdr) + sizeof(uint32_t); ph = mtod(op_err, struct sctp_paramhdr *); ph->param_type = htons( SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(SCTP_BUF_LEN(op_err)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_3); } asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3; #if defined(SCTP_PANIC_ON_ABORT) panic("inpcb_free does an abort"); #endif sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED); SCTP_STAT_INCR_COUNTER32(sctps_aborted); if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) { cnt_in_sd++; } continue; } else if (TAILQ_EMPTY(&asoc->asoc.send_queue) && TAILQ_EMPTY(&asoc->asoc.sent_queue) && (asoc->asoc.stream_queue_cnt == 0) ) { if (asoc->asoc.locked_on_sending) { goto abort_anyway; } if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { /* * there is nothing queued to send, * so I send shutdown */ sctp_send_shutdown(asoc, asoc->asoc.primary_destination); if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc, asoc->asoc.primary_destination); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc, asoc->asoc.primary_destination); sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED); } } else { /* mark into shutdown pending */ struct sctp_stream_queue_pending *sp; asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING; sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc, asoc->asoc.primary_destination); if (asoc->asoc.locked_on_sending) { sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue), sctp_streamhead); if (sp == NULL) { SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n", asoc->asoc.locked_on_sending, asoc->asoc.locked_on_sending->stream_no); } else { if ((sp->length == 0) && (sp->msg_is_complete == 0)) asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT; } } if (TAILQ_EMPTY(&asoc->asoc.send_queue) && TAILQ_EMPTY(&asoc->asoc.sent_queue) && (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) { struct mbuf *op_err; abort_anyway: op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { /* * Fill in the user * initiated abort */ struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) + sizeof(uint32_t)); ph = mtod(op_err, struct sctp_paramhdr *); ph->param_type = htons( SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(SCTP_BUF_LEN(op_err)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_5); } asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5; #if defined(SCTP_PANIC_ON_ABORT) panic("inpcb_free does an abort"); #endif sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED); SCTP_STAT_INCR_COUNTER32(sctps_aborted); if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) { cnt_in_sd++; } continue; } else { sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED); } } cnt_in_sd++; SCTP_TCB_UNLOCK(asoc); } /* now is there some left in our SHUTDOWN state? */ if (cnt_in_sd) { SCTP_INP_WUNLOCK(inp); SCTP_ASOC_CREATE_UNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_ITERATOR_UNLOCK(); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 2); #endif return; } } inp->sctp_socket = NULL; if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) != SCTP_PCB_FLAGS_UNBOUND) { /* * ok, this guy has been bound. It's port is somewhere in * the SCTP_BASE_INFO(hash table). Remove it! */ LIST_REMOVE(inp, sctp_hash); inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND; } /* * If there is a timer running to kill us, forget it, since it may * have a contest on the INP lock.. which would cause us to die ... */ cnt = 0; for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL; asoc = nasoc) { nasoc = LIST_NEXT(asoc, sctp_tcblist); if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { cnt++; continue; } /* Free associations that are NOT killing us */ SCTP_TCB_LOCK(asoc); if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) && ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) { struct mbuf *op_err; uint32_t *ippp; op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { /* Fill in the user initiated abort */ struct sctp_paramhdr *ph; SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) + sizeof(uint32_t)); ph = mtod(op_err, struct sctp_paramhdr *); ph->param_type = htons( SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(SCTP_BUF_LEN(op_err)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_7); } asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7; #if defined(SCTP_PANIC_ON_ABORT) panic("inpcb_free does an abort"); #endif sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED); SCTP_STAT_INCR_COUNTER32(sctps_aborted); } else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { cnt++; SCTP_TCB_UNLOCK(asoc); continue; } if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) { cnt++; } } if (cnt) { /* Ok we have someone out there that will kill us */ (void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer); SCTP_INP_WUNLOCK(inp); SCTP_ASOC_CREATE_UNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_ITERATOR_UNLOCK(); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 3); #endif return; } if ((inp->refcount) || (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) { (void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer); sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL); SCTP_INP_WUNLOCK(inp); SCTP_ASOC_CREATE_UNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_ITERATOR_UNLOCK(); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 4); #endif return; } (void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer); inp->sctp_ep.signature_change.type = 0; inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE; #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 5); #endif (void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer); inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NONE; /* Clear the read queue */ /* sa_ignore FREED_MEMORY */ while ((sq = TAILQ_FIRST(&inp->read_queue)) != NULL) { /* Its only abandoned if it had data left */ if (sq->length) SCTP_STAT_INCR(sctps_left_abandon); TAILQ_REMOVE(&inp->read_queue, sq, next); sctp_free_remote_addr(sq->whoFrom); if (so) so->so_rcv.sb_cc -= sq->length; if (sq->data) { sctp_m_freem(sq->data); sq->data = NULL; } /* * no need to free the net count, since at this point all * assoc's are gone. */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq); SCTP_DECR_READQ_COUNT(); } /* Now the sctp_pcb things */ /* * free each asoc if it is not already closed/free. we can't use the * macro here since le_next will get freed as part of the * sctp_free_assoc() call. */ cnt = 0; if (so) { #ifdef IPSEC ipsec4_delete_pcbpolicy(ip_pcb); #endif /* IPSEC */ /* Unlocks not needed since the socket is gone now */ } if (ip_pcb->inp_options) { (void)sctp_m_free(ip_pcb->inp_options); ip_pcb->inp_options = 0; } if (ip_pcb->inp_moptions) { inp_freemoptions(ip_pcb->inp_moptions); ip_pcb->inp_moptions = 0; } #ifdef INET6 if (ip_pcb->inp_vflag & INP_IPV6) { struct in6pcb *in6p; in6p = (struct in6pcb *)inp; ip6_freepcbopts(in6p->in6p_outputopts); } #endif /* INET6 */ ip_pcb->inp_vflag = 0; /* free up authentication fields */ if (inp->sctp_ep.local_auth_chunks != NULL) sctp_free_chunklist(inp->sctp_ep.local_auth_chunks); if (inp->sctp_ep.local_hmacs != NULL) sctp_free_hmaclist(inp->sctp_ep.local_hmacs); shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys); while (shared_key) { LIST_REMOVE(shared_key, next); sctp_free_sharedkey(shared_key); /* sa_ignore FREED_MEMORY */ shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys); } inp_save = LIST_NEXT(inp, sctp_list); LIST_REMOVE(inp, sctp_list); /* fix any iterators only after out of the list */ sctp_iterator_inp_being_freed(inp, inp_save); /* * if we have an address list the following will free the list of * ifaddr's that are set into this ep. Again macro limitations here, * since the LIST_FOREACH could be a bad idea. */ for ((laddr = LIST_FIRST(&inp->sctp_addr_list)); laddr != NULL; laddr = nladdr) { nladdr = LIST_NEXT(laddr, sctp_nxt_addr); sctp_remove_laddr(laddr); } #ifdef SCTP_TRACK_FREED_ASOCS /* TEMP CODE */ for ((asoc = LIST_FIRST(&inp->sctp_asoc_free_list)); asoc != NULL; asoc = nasoc) { nasoc = LIST_NEXT(asoc, sctp_tcblist); LIST_REMOVE(asoc, sctp_tcblist); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc); SCTP_DECR_ASOC_COUNT(); } /* *** END TEMP CODE *** */ #endif /* Now lets see about freeing the EP hash table. */ if (inp->sctp_tcbhash != NULL) { SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark); inp->sctp_tcbhash = NULL; } /* Now we must put the ep memory back into the zone pool */ INP_LOCK_DESTROY(&inp->ip_inp.inp); SCTP_INP_LOCK_DESTROY(inp); SCTP_INP_READ_DESTROY(inp); SCTP_ASOC_CREATE_LOCK_DESTROY(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_ITERATOR_UNLOCK(); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); SCTP_DECR_EP_COUNT(); } struct sctp_nets * sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr) { struct sctp_nets *net; /* locate the address */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr)) return (net); } return (NULL); } int sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id) { struct sctp_ifa *sctp_ifa; sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED); if (sctp_ifa) { return (1); } else { return (0); } } /* * add's a remote endpoint address, done with the INIT/INIT-ACK as well as * when a ASCONF arrives that adds it. It will also initialize all the cwnd * stats of stuff. */ int sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr, int set_scope, int from) { /* * The following is redundant to the same lines in the * sctp_aloc_assoc() but is needed since other's call the add * address function */ struct sctp_nets *net, *netfirst; int addr_inscope; SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ", from); SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr); netfirst = sctp_findnet(stcb, newaddr); if (netfirst) { /* * Lie and return ok, we don't want to make the association * go away for this behavior. It will happen in the TCP * model in a connected socket. It does not reach the hash * table until after the association is built so it can't be * found. Mark as reachable, since the initial creation will * have been cleared and the NOT_IN_ASSOC flag will have * been added... and we don't want to end up removing it * back out. */ if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) { netfirst->dest_state = (SCTP_ADDR_REACHABLE | SCTP_ADDR_UNCONFIRMED); } else { netfirst->dest_state = SCTP_ADDR_REACHABLE; } return (0); } addr_inscope = 1; if (newaddr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)newaddr; if (sin->sin_addr.s_addr == 0) { /* Invalid address */ return (-1); } /* zero out the bzero area */ memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); /* assure len is set */ sin->sin_len = sizeof(struct sockaddr_in); if (set_scope) { #ifdef SCTP_DONT_DO_PRIVADDR_SCOPE stcb->ipv4_local_scope = 1; #else if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) { stcb->asoc.ipv4_local_scope = 1; } #endif /* SCTP_DONT_DO_PRIVADDR_SCOPE */ } else { /* Validate the address is in scope */ if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) && (stcb->asoc.ipv4_local_scope == 0)) { addr_inscope = 0; } } #ifdef INET6 } else if (newaddr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)newaddr; if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { /* Invalid address */ return (-1); } /* assure len is set */ sin6->sin6_len = sizeof(struct sockaddr_in6); if (set_scope) { if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) { stcb->asoc.loopback_scope = 1; stcb->asoc.local_scope = 0; stcb->asoc.ipv4_local_scope = 1; stcb->asoc.site_scope = 1; } else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { /* * If the new destination is a LINK_LOCAL we * must have common site scope. Don't set * the local scope since we may not share * all links, only loopback can do this. * Links on the local network would also be * on our private network for v4 too. */ stcb->asoc.ipv4_local_scope = 1; stcb->asoc.site_scope = 1; } else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) { /* * If the new destination is SITE_LOCAL then * we must have site scope in common. */ stcb->asoc.site_scope = 1; } } else { /* Validate the address is in scope */ if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) && (stcb->asoc.loopback_scope == 0)) { addr_inscope = 0; } else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) && (stcb->asoc.local_scope == 0)) { addr_inscope = 0; } else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) && (stcb->asoc.site_scope == 0)) { addr_inscope = 0; } } #endif } else { /* not supported family type */ return (-1); } net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets); if (net == NULL) { return (-1); } SCTP_INCR_RADDR_COUNT(); bzero(net, sizeof(*net)); (void)SCTP_GETTIME_TIMEVAL(&net->start_time); memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len); if (newaddr->sa_family == AF_INET) { ((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport; } else if (newaddr->sa_family == AF_INET6) { ((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport; } net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id); if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) { stcb->asoc.loopback_scope = 1; stcb->asoc.ipv4_local_scope = 1; stcb->asoc.local_scope = 0; stcb->asoc.site_scope = 1; addr_inscope = 1; } net->failure_threshold = stcb->asoc.def_net_failure; if (addr_inscope == 0) { net->dest_state = (SCTP_ADDR_REACHABLE | SCTP_ADDR_OUT_OF_SCOPE); } else { if (from == SCTP_ADDR_IS_CONFIRMED) /* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */ net->dest_state = SCTP_ADDR_REACHABLE; else net->dest_state = SCTP_ADDR_REACHABLE | SCTP_ADDR_UNCONFIRMED; } /* * We set this to 0, the timer code knows that this means its an * initial value */ net->RTO = 0; net->RTO_measured = 0; stcb->asoc.numnets++; *(&net->ref_count) = 1; net->tos_flowlabel = 0; if (SCTP_BASE_SYSCTL(sctp_udp_tunneling_for_client_enable)) { net->port = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port)); } else { net->port = 0; } #ifdef INET if (newaddr->sa_family == AF_INET) net->tos_flowlabel = stcb->asoc.default_tos; #endif #ifdef INET6 if (newaddr->sa_family == AF_INET6) net->tos_flowlabel = stcb->asoc.default_flowlabel; #endif /* Init the timer structure */ SCTP_OS_TIMER_INIT(&net->rxt_timer.timer); SCTP_OS_TIMER_INIT(&net->fr_timer.timer); SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer); /* Now generate a route for this guy */ #ifdef INET6 /* KAME hack: embed scopeid */ if (newaddr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; (void)sa6_embedscope(sin6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone)); sin6->sin6_scope_id = 0; } #endif SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id); if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) { /* Get source address */ net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep, stcb, (sctp_route_t *) & net->ro, net, 0, stcb->asoc.vrf_id); /* Now get the interface MTU */ if (net->ro._s_addr && net->ro._s_addr->ifn_p) { net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p); } else { net->mtu = 0; } #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("We have found an interface mtu of %d\n", net->mtu); #endif if (net->mtu == 0) { /* Huh ?? */ net->mtu = SCTP_DEFAULT_MTU; } else { uint32_t rmtu; rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt); #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("The route mtu is %d\n", rmtu); #endif if (rmtu == 0) { /* * Start things off to match mtu of * interface please. */ SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa, net->ro.ro_rt, net->mtu); } else { /* * we take the route mtu over the interface, * since the route may be leading out the * loopback, or a different interface. */ net->mtu = rmtu; } } if (from == SCTP_ALLOC_ASOC) { #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("New assoc sets mtu to :%d\n", net->mtu); #endif stcb->asoc.smallest_mtu = net->mtu; } } else { net->mtu = stcb->asoc.smallest_mtu; } #ifdef INET6 if (newaddr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; (void)sa6_recoverscope(sin6); } #endif if (net->port) { net->mtu -= sizeof(struct udphdr); } if (stcb->asoc.smallest_mtu > net->mtu) { #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("new address mtu:%d smaller than smallest:%d\n", net->mtu, stcb->asoc.smallest_mtu); #endif stcb->asoc.smallest_mtu = net->mtu; } /* JRS - Use the congestion control given in the CC module */ stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net); /* * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning * of assoc (2005/06/27, iyengar@cis.udel.edu) */ net->find_pseudo_cumack = 1; net->find_rtx_pseudo_cumack = 1; net->src_addr_selected = 0; netfirst = TAILQ_FIRST(&stcb->asoc.nets); if (net->ro.ro_rt == NULL) { /* Since we have no route put it at the back */ TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next); } else if (netfirst == NULL) { /* We are the first one in the pool. */ TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); } else if (netfirst->ro.ro_rt == NULL) { /* * First one has NO route. Place this one ahead of the first * one. */ TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); } else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) { /* * This one has a different interface than the one at the * top of the list. Place it ahead. */ TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); } else { /* * Ok we have the same interface as the first one. Move * forward until we find either a) one with a NULL route... * insert ahead of that b) one with a different ifp.. insert * after that. c) end of the list.. insert at the tail. */ struct sctp_nets *netlook; do { netlook = TAILQ_NEXT(netfirst, sctp_next); if (netlook == NULL) { /* End of the list */ TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next); break; } else if (netlook->ro.ro_rt == NULL) { /* next one has NO route */ TAILQ_INSERT_BEFORE(netfirst, net, sctp_next); break; } else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) { TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook, net, sctp_next); break; } /* Shift forward */ netfirst = netlook; } while (netlook != NULL); } /* got to have a primary set */ if (stcb->asoc.primary_destination == 0) { stcb->asoc.primary_destination = net; } else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) && (net->ro.ro_rt) && ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) { /* No route to current primary adopt new primary */ stcb->asoc.primary_destination = net; } sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, stcb->sctp_ep, stcb, net); /* Validate primary is first */ net = TAILQ_FIRST(&stcb->asoc.nets); if ((net != stcb->asoc.primary_destination) && (stcb->asoc.primary_destination)) { /* * first one on the list is NOT the primary sctp_cmpaddr() * is much more efficent if the primary is the first on the * list, make it so. */ TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); } return (0); } /* * allocate an association and add it to the endpoint. The caller must be * careful to add all additional addresses once they are know right away or * else the assoc will be may experience a blackout scenario. */ struct sctp_tcb * sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr, int for_a_init, int *error, uint32_t override_tag, uint32_t vrf_id, struct thread *p ) { /* note the p argument is only valid in unbound sockets */ struct sctp_tcb *stcb; struct sctp_association *asoc; struct sctpasochead *head; uint16_t rport; int err; /* * Assumption made here: Caller has done a * sctp_findassociation_ep_addr(ep, addr's); to make sure the * address does not exist already. */ if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) { /* Hit max assoc, sorry no more */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); *error = ENOBUFS; return (NULL); } if (firstaddr == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } SCTP_INP_RLOCK(inp); - if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) && + ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) || + (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) { /* * If its in the TCP pool, its NOT allowed to create an * association. The parent listener needs to call * sctp_aloc_assoc.. or the one-2-many socket. If a peeled * off, or connected one does this.. its an error. */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:"); #ifdef SCTP_DEBUG if (firstaddr) { SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr); SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n", ntohs(((struct sockaddr_in *)firstaddr)->sin_port)); } else { SCTPDBG(SCTP_DEBUG_PCB3, "None\n"); } #endif /* SCTP_DEBUG */ if (firstaddr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)firstaddr; if ((sin->sin_port == 0) || (sin->sin_addr.s_addr == 0)) { /* Invalid address */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } rport = sin->sin_port; } else if (firstaddr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)firstaddr; if ((sin6->sin6_port == 0) || (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) { /* Invalid address */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } rport = sin6->sin6_port; } else { /* not supported family type */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } SCTP_INP_RUNLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { /* * If you have not performed a bind, then we need to do the * ephemerial bind for you. */ if ((err = sctp_inpcb_bind(inp->sctp_socket, (struct sockaddr *)NULL, (struct sctp_ifa *)NULL, p ))) { /* bind error, probably perm */ *error = err; return (NULL); } } stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb); if (stcb == NULL) { /* out of memory? */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM); *error = ENOMEM; return (NULL); } SCTP_INCR_ASOC_COUNT(); bzero(stcb, sizeof(*stcb)); asoc = &stcb->asoc; SCTP_TCB_LOCK_INIT(stcb); SCTP_TCB_SEND_LOCK_INIT(stcb); /* setup back pointer's */ stcb->sctp_ep = inp; stcb->sctp_socket = inp->sctp_socket; if ((err = sctp_init_asoc(inp, stcb, for_a_init, override_tag, vrf_id))) { /* failed */ SCTP_TCB_LOCK_DESTROY(stcb); SCTP_TCB_SEND_LOCK_DESTROY(stcb); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_DECR_ASOC_COUNT(); *error = err; return (NULL); } /* and the port */ stcb->rport = rport; SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(inp); if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { /* inpcb freed while alloc going on */ SCTP_TCB_LOCK_DESTROY(stcb); SCTP_TCB_SEND_LOCK_DESTROY(stcb); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_DECR_ASOC_COUNT(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } SCTP_TCB_LOCK(stcb); /* now that my_vtag is set, add it to the hash */ head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))]; /* put it in the bucket in the vtag hash of assoc's for the system */ LIST_INSERT_HEAD(head, stcb, sctp_asocs); sctp_delete_from_timewait(stcb->asoc.my_vtag); SCTP_INP_INFO_WUNLOCK(); if ((err = sctp_add_remote_addr(stcb, firstaddr, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) { /* failure.. memory error? */ if (asoc->strmout) { SCTP_FREE(asoc->strmout, SCTP_M_STRMO); asoc->strmout = NULL; } if (asoc->mapping_array) { SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); asoc->mapping_array = NULL; } SCTP_DECR_ASOC_COUNT(); SCTP_TCB_LOCK_DESTROY(stcb); SCTP_TCB_SEND_LOCK_DESTROY(stcb); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_INP_WUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); *error = ENOBUFS; return (NULL); } /* Init all the timers */ SCTP_OS_TIMER_INIT(&asoc->hb_timer.timer); SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer); SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer); SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer); SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer); SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer); SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer); SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer); LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist); /* now file the port under the hash as well */ if (inp->sctp_tcbhash != NULL) { head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport, inp->sctp_hashmark)]; LIST_INSERT_HEAD(head, stcb, sctp_tcbhash); } SCTP_INP_WUNLOCK(inp); SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", stcb); return (stcb); } void sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_association *asoc; asoc = &stcb->asoc; asoc->numnets--; TAILQ_REMOVE(&asoc->nets, net, sctp_next); if (net == asoc->primary_destination) { /* Reset primary */ struct sctp_nets *lnet; lnet = TAILQ_FIRST(&asoc->nets); /* * Mobility adaptation Ideally, if deleted destination is * the primary, it becomes a fast retransmission trigger by * the subsequent SET PRIMARY. (by micchie) */ if (sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_BASE) || sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_FASTHANDOFF)) { SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n"); if (asoc->deleted_primary != NULL) { SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n"); goto out; } asoc->deleted_primary = net; atomic_add_int(&net->ref_count, 1); memset(&net->lastsa, 0, sizeof(net->lastsa)); memset(&net->lastsv, 0, sizeof(net->lastsv)); sctp_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_PRIM_DELETED); sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED, stcb->sctp_ep, stcb, NULL); } out: /* Try to find a confirmed primary */ asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0); } if (net == asoc->last_data_chunk_from) { /* Reset primary */ asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets); } if (net == asoc->last_control_chunk_from) { /* Clear net */ asoc->last_control_chunk_from = NULL; } sctp_free_remote_addr(net); } /* * remove a remote endpoint address from an association, it will fail if the * address does not exist. */ int sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr) { /* * Here we need to remove a remote address. This is quite simple, we * first find it in the list of address for the association * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE * on that item. Note we do not allow it to be removed if there are * no other addresses. */ struct sctp_association *asoc; struct sctp_nets *net, *net_tmp; asoc = &stcb->asoc; /* locate the address */ for (net = TAILQ_FIRST(&asoc->nets); net != NULL; net = net_tmp) { net_tmp = TAILQ_NEXT(net, sctp_next); if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) { continue; } if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr, remaddr)) { /* we found the guy */ if (asoc->numnets < 2) { /* Must have at LEAST two remote addresses */ return (-1); } else { sctp_remove_net(stcb, net); return (0); } } } /* not found. */ return (-2); } void sctp_delete_from_timewait(uint32_t tag) { struct sctpvtaghead *chain; struct sctp_tagblock *twait_block; int found = 0; int i; chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; if (!SCTP_LIST_EMPTY(chain)) { LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { if (twait_block->vtag_block[i].v_tag == tag) { twait_block->vtag_block[i].tv_sec_at_expire = 0; twait_block->vtag_block[i].v_tag = 0; found = 1; break; } } if (found) break; } } } int sctp_is_in_timewait(uint32_t tag) { struct sctpvtaghead *chain; struct sctp_tagblock *twait_block; int found = 0; int i; chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; if (!SCTP_LIST_EMPTY(chain)) { LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { if (twait_block->vtag_block[i].v_tag == tag) { found = 1; break; } } if (found) break; } } return (found); } void sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time) { struct sctpvtaghead *chain; struct sctp_tagblock *twait_block; struct timeval now; int set, i; (void)SCTP_GETTIME_TIMEVAL(&now); chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; set = 0; if (!SCTP_LIST_EMPTY(chain)) { /* Block(s) present, lets find space, and expire on the fly */ LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { if ((twait_block->vtag_block[i].v_tag == 0) && !set) { twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time; twait_block->vtag_block[i].v_tag = tag; set = 1; } else if ((twait_block->vtag_block[i].v_tag) && ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) { /* Audit expires this guy */ twait_block->vtag_block[i].tv_sec_at_expire = 0; twait_block->vtag_block[i].v_tag = 0; if (set == 0) { /* Reuse it for my new tag */ twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time; twait_block->vtag_block[0].v_tag = tag; set = 1; } } } if (set) { /* * We only do up to the block where we can * place our tag for audits */ break; } } } /* Need to add a new block to chain */ if (!set) { SCTP_MALLOC(twait_block, struct sctp_tagblock *, sizeof(struct sctp_tagblock), SCTP_M_TIMW); if (twait_block == NULL) { return; } memset(twait_block, 0, sizeof(struct sctp_tagblock)); LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock); twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time; twait_block->vtag_block[0].v_tag = tag; } } static void sctp_iterator_asoc_being_freed(struct sctp_inpcb *inp, struct sctp_tcb *stcb) { struct sctp_iterator *it; /* * Unlock the tcb lock we do this so we avoid a dead lock scenario * where the iterator is waiting on the TCB lock and the TCB lock is * waiting on the iterator lock. */ it = stcb->asoc.stcb_starting_point_for_iterator; if (it == NULL) { return; } if (it->inp != stcb->sctp_ep) { /* hmm, focused on the wrong one? */ return; } if (it->stcb != stcb) { return; } it->stcb = LIST_NEXT(stcb, sctp_tcblist); if (it->stcb == NULL) { /* done with all asoc's in this assoc */ if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { it->inp = NULL; } else { it->inp = LIST_NEXT(inp, sctp_list); } } } /*- * Free the association after un-hashing the remote port. This * function ALWAYS returns holding NO LOCK on the stcb. It DOES * expect that the input to this function IS a locked TCB. * It will return 0, if it did NOT destroy the association (instead * it unlocks it. It will return NON-zero if it either destroyed the * association OR the association is already destroyed. */ int sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location) { int i; struct sctp_association *asoc; struct sctp_nets *net, *prev; struct sctp_laddr *laddr; struct sctp_tmit_chunk *chk; struct sctp_asconf_addr *aparam; struct sctp_asconf_ack *aack; struct sctp_stream_reset_list *liste; struct sctp_queued_to_read *sq; struct sctp_stream_queue_pending *sp; sctp_sharedkey_t *shared_key; struct socket *so; int ccnt = 0; int cnt = 0; /* first, lets purge the entry from the hash table. */ #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, stcb, 6); #endif if (stcb->asoc.state == 0) { #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 7); #endif /* there is no asoc, really TSNH :-0 */ return (1); } /* TEMP CODE */ if (stcb->freed_from_where == 0) { /* Only record the first place free happened from */ stcb->freed_from_where = from_location; } /* TEMP CODE */ asoc = &stcb->asoc; if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) /* nothing around */ so = NULL; else so = inp->sctp_socket; /* * We used timer based freeing if a reader or writer is in the way. * So we first check if we are actually being called from a timer, * if so we abort early if a reader or writer is still in the way. */ if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) && (from_inpcbfree == SCTP_NORMAL_PROC)) { /* * is it the timer driving us? if so are the reader/writers * gone? */ if (stcb->asoc.refcnt) { /* nope, reader or writer in the way */ sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); /* no asoc destroyed */ SCTP_TCB_UNLOCK(stcb); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, stcb, 8); #endif return (0); } } /* now clean up any other timers */ (void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer); asoc->hb_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer); asoc->dack_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer); /*- * For stream reset we don't blast this unless * it is a str-reset timer, it might be the * free-asoc timer which we DON'T want to * disturb. */ if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET) asoc->strreset_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer); asoc->asconf_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer); asoc->autoclose_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer); asoc->shut_guard_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer); asoc->delayed_event_timer.self = NULL; /* Mobility adaptation */ (void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer); asoc->delete_prim_timer.self = NULL; TAILQ_FOREACH(net, &asoc->nets, sctp_next) { (void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer); net->fr_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer); net->rxt_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer); net->pmtu_timer.self = NULL; } /* Now the read queue needs to be cleaned up (only once) */ cnt = 0; if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) { stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED; SCTP_INP_READ_LOCK(inp); TAILQ_FOREACH(sq, &inp->read_queue, next) { if (sq->stcb == stcb) { sq->do_not_ref_stcb = 1; sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn; /* * If there is no end, there never will be * now. */ if (sq->end_added == 0) { /* Held for PD-API clear that. */ sq->pdapi_aborted = 1; sq->held_length = 0; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) { /* * Need to add a PD-API * aborted indication. * Setting the control_pdapi * assures that it will be * added right after this * msg. */ uint32_t strseq; stcb->asoc.control_pdapi = sq; strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn; sctp_notify_partial_delivery_indication(stcb, SCTP_PARTIAL_DELIVERY_ABORTED, 1, strseq); stcb->asoc.control_pdapi = NULL; } } /* Add an end to wake them */ sq->end_added = 1; cnt++; } } SCTP_INP_READ_UNLOCK(inp); if (stcb->block_entry) { cnt++; SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET); stcb->block_entry->error = ECONNRESET; stcb->block_entry = NULL; } } if (stcb->asoc.refcnt) { /* * reader or writer in the way, we have hopefully given him * something to chew on above. */ sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); SCTP_TCB_UNLOCK(stcb); if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) /* nothing around */ so = NULL; if (so) { /* Wake any reader/writers */ sctp_sorwakeup(inp, so); sctp_sowwakeup(inp, so); } #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, stcb, 9); #endif /* no asoc destroyed */ return (0); } #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, stcb, 10); #endif /* * When I reach here, no others want to kill the assoc yet.. and I * own the lock. Now its possible an abort comes in when I do the * lock exchange below to grab all the locks to do the final take * out. to prevent this we increment the count, which will start a * timer and blow out above thus assuring us that we hold exclusive * killing of the asoc. Note that after getting back the TCB lock we * will go ahead and increment the counter back up and stop any * timer a passing stranger may have started :-S */ if (from_inpcbfree == SCTP_NORMAL_PROC) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_ITERATOR_LOCK(); SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(inp); SCTP_TCB_LOCK(stcb); } /* Double check the GONE flag */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) /* nothing around */ so = NULL; if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { /* * For TCP type we need special handling when we are * connected. We also include the peel'ed off ones to. */ if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED; inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED; if (so) { SOCK_LOCK(so); if (so->so_rcv.sb_cc == 0) { so->so_state &= ~(SS_ISCONNECTING | SS_ISDISCONNECTING | SS_ISCONFIRMING | SS_ISCONNECTED); } SOCK_UNLOCK(so); socantrcvmore(so); sctp_sowwakeup(inp, so); sctp_sorwakeup(inp, so); SCTP_SOWAKEUP(so); } } } /* * Make it invalid too, that way if its about to run it will abort * and return. */ sctp_iterator_asoc_being_freed(inp, stcb); /* re-increment the lock */ if (from_inpcbfree == SCTP_NORMAL_PROC) { atomic_add_int(&stcb->asoc.refcnt, -1); } asoc->state = 0; if (inp->sctp_tcbhash) { LIST_REMOVE(stcb, sctp_tcbhash); } if (stcb->asoc.in_restart_hash) { LIST_REMOVE(stcb, sctp_tcbrestarhash); } /* Now lets remove it from the list of ALL associations in the EP */ LIST_REMOVE(stcb, sctp_tcblist); if (from_inpcbfree == SCTP_NORMAL_PROC) { SCTP_INP_INCR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_ITERATOR_UNLOCK(); } /* pull from vtag hash */ LIST_REMOVE(stcb, sctp_asocs); sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_TIME_WAIT); /* * Now restop the timers to be sure - this is paranoia at is finest! */ (void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer); TAILQ_FOREACH(net, &asoc->nets, sctp_next) { (void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer); (void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer); (void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer); } asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE; prev = NULL; /* * The chunk lists and such SHOULD be empty but we check them just * in case. */ /* anything on the wheel needs to be removed */ for (i = 0; i < asoc->streamoutcnt; i++) { struct sctp_stream_out *outs; outs = &asoc->strmout[i]; /* now clean up any chunks here */ sp = TAILQ_FIRST(&outs->outqueue); while (sp) { TAILQ_REMOVE(&outs->outqueue, sp, next); if (sp->data) { sctp_m_freem(sp->data); sp->data = NULL; sp->tail_mbuf = NULL; } sctp_free_remote_addr(sp->net); sctp_free_spbufspace(stcb, asoc, sp); /* Free the zone stuff */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_strmoq), sp); SCTP_DECR_STRMOQ_COUNT(); /* sa_ignore FREED_MEMORY */ sp = TAILQ_FIRST(&outs->outqueue); } } /* sa_ignore FREED_MEMORY */ while ((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) { TAILQ_REMOVE(&asoc->resetHead, liste, next_resp); SCTP_FREE(liste, SCTP_M_STRESET); } sq = TAILQ_FIRST(&asoc->pending_reply_queue); while (sq) { TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next); if (sq->data) { sctp_m_freem(sq->data); sq->data = NULL; } sctp_free_remote_addr(sq->whoFrom); sq->whoFrom = NULL; sq->stcb = NULL; /* Free the ctl entry */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq); SCTP_DECR_READQ_COUNT(); /* sa_ignore FREED_MEMORY */ sq = TAILQ_FIRST(&asoc->pending_reply_queue); } chk = TAILQ_FIRST(&asoc->free_chunks); while (chk) { TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } ccnt++; SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1); asoc->free_chunk_cnt--; /* sa_ignore FREED_MEMORY */ chk = TAILQ_FIRST(&asoc->free_chunks); } /* pending send queue SHOULD be empty */ if (!TAILQ_EMPTY(&asoc->send_queue)) { chk = TAILQ_FIRST(&asoc->send_queue); while (chk) { TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } ccnt++; sctp_free_remote_addr(chk->whoTo); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ chk = TAILQ_FIRST(&asoc->send_queue); } } /* if(ccnt) { printf("Freed %d from send_queue\n", ccnt); ccnt = 0; } */ /* sent queue SHOULD be empty */ if (!TAILQ_EMPTY(&asoc->sent_queue)) { chk = TAILQ_FIRST(&asoc->sent_queue); while (chk) { TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } ccnt++; sctp_free_remote_addr(chk->whoTo); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ chk = TAILQ_FIRST(&asoc->sent_queue); } } /* if(ccnt) { printf("Freed %d from sent_queue\n", ccnt); ccnt = 0; } */ /* control queue MAY not be empty */ if (!TAILQ_EMPTY(&asoc->control_send_queue)) { chk = TAILQ_FIRST(&asoc->control_send_queue); while (chk) { TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } ccnt++; sctp_free_remote_addr(chk->whoTo); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ chk = TAILQ_FIRST(&asoc->control_send_queue); } } /* if(ccnt) { printf("Freed %d from ctrl_queue\n", ccnt); ccnt = 0; } */ /* ASCONF queue MAY not be empty */ if (!TAILQ_EMPTY(&asoc->asconf_send_queue)) { chk = TAILQ_FIRST(&asoc->asconf_send_queue); while (chk) { TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } ccnt++; sctp_free_remote_addr(chk->whoTo); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ chk = TAILQ_FIRST(&asoc->asconf_send_queue); } } /* if(ccnt) { printf("Freed %d from asconf_queue\n", ccnt); ccnt = 0; } */ if (!TAILQ_EMPTY(&asoc->reasmqueue)) { chk = TAILQ_FIRST(&asoc->reasmqueue); while (chk) { TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } sctp_free_remote_addr(chk->whoTo); ccnt++; SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ chk = TAILQ_FIRST(&asoc->reasmqueue); } } /* if(ccnt) { printf("Freed %d from reasm_queue\n", ccnt); ccnt = 0; } */ if (asoc->mapping_array) { SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); asoc->mapping_array = NULL; } /* the stream outs */ if (asoc->strmout) { SCTP_FREE(asoc->strmout, SCTP_M_STRMO); asoc->strmout = NULL; } asoc->streamoutcnt = 0; if (asoc->strmin) { struct sctp_queued_to_read *ctl; for (i = 0; i < asoc->streamincnt; i++) { if (!TAILQ_EMPTY(&asoc->strmin[i].inqueue)) { /* We have somethings on the streamin queue */ ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue); while (ctl) { TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next); sctp_free_remote_addr(ctl->whoFrom); if (ctl->data) { sctp_m_freem(ctl->data); ctl->data = NULL; } /* * We don't free the address here * since all the net's were freed * above. */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl); SCTP_DECR_READQ_COUNT(); ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue); } } } SCTP_FREE(asoc->strmin, SCTP_M_STRMI); asoc->strmin = NULL; } asoc->streamincnt = 0; while (!TAILQ_EMPTY(&asoc->nets)) { /* sa_ignore FREED_MEMORY */ net = TAILQ_FIRST(&asoc->nets); /* pull from list */ if ((SCTP_BASE_INFO(ipi_count_raddr) == 0) || (prev == net)) { #ifdef INVARIANTS panic("no net's left alloc'ed, or list points to itself"); #endif break; } prev = net; TAILQ_REMOVE(&asoc->nets, net, sctp_next); sctp_free_remote_addr(net); } while (!SCTP_LIST_EMPTY(&asoc->sctp_restricted_addrs)) { /* sa_ignore FREED_MEMORY */ laddr = LIST_FIRST(&asoc->sctp_restricted_addrs); sctp_remove_laddr(laddr); } /* pending asconf (address) parameters */ while (!TAILQ_EMPTY(&asoc->asconf_queue)) { /* sa_ignore FREED_MEMORY */ aparam = TAILQ_FIRST(&asoc->asconf_queue); TAILQ_REMOVE(&asoc->asconf_queue, aparam, next); SCTP_FREE(aparam, SCTP_M_ASC_ADDR); } while (!TAILQ_EMPTY(&asoc->asconf_ack_sent)) { /* sa_ignore FREED_MEMORY */ aack = TAILQ_FIRST(&asoc->asconf_ack_sent); TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next); if (aack->data != NULL) { sctp_m_freem(aack->data); } SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack); } /* clean up auth stuff */ if (asoc->local_hmacs) sctp_free_hmaclist(asoc->local_hmacs); if (asoc->peer_hmacs) sctp_free_hmaclist(asoc->peer_hmacs); if (asoc->local_auth_chunks) sctp_free_chunklist(asoc->local_auth_chunks); if (asoc->peer_auth_chunks) sctp_free_chunklist(asoc->peer_auth_chunks); sctp_free_authinfo(&asoc->authinfo); shared_key = LIST_FIRST(&asoc->shared_keys); while (shared_key) { LIST_REMOVE(shared_key, next); sctp_free_sharedkey(shared_key); /* sa_ignore FREED_MEMORY */ shared_key = LIST_FIRST(&asoc->shared_keys); } /* Insert new items here :> */ /* Get rid of LOCK */ SCTP_TCB_LOCK_DESTROY(stcb); SCTP_TCB_SEND_LOCK_DESTROY(stcb); if (from_inpcbfree == SCTP_NORMAL_PROC) { SCTP_INP_INFO_WUNLOCK(); SCTP_INP_RLOCK(inp); } #ifdef SCTP_TRACK_FREED_ASOCS if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { /* now clean up the tasoc itself */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_DECR_ASOC_COUNT(); } else { LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist); } #else SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_DECR_ASOC_COUNT(); #endif if (from_inpcbfree == SCTP_NORMAL_PROC) { if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { /* * If its NOT the inp_free calling us AND sctp_close * as been called, we call back... */ SCTP_INP_RUNLOCK(inp); /* * This will start the kill timer (if we are the * lastone) since we hold an increment yet. But this * is the only safe way to do this since otherwise * if the socket closes at the same time we are here * we might collide in the cleanup. */ sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE, SCTP_CALLED_DIRECTLY_NOCMPSET); SCTP_INP_DECR_REF(inp); goto out_of; } else { /* The socket is still open. */ SCTP_INP_DECR_REF(inp); } } if (from_inpcbfree == SCTP_NORMAL_PROC) { SCTP_INP_RUNLOCK(inp); } out_of: /* destroyed the asoc */ #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 11); #endif return (1); } /* * determine if a destination is "reachable" based upon the addresses bound * to the current endpoint (e.g. only v4 or v6 currently bound) */ /* * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use * assoc level v4/v6 flags, as the assoc *may* not have the same address * types bound as its endpoint */ int sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr) { struct sctp_inpcb *inp; int answer; /* * No locks here, the TCB, in all cases is already locked and an * assoc is up. There is either a INP lock by the caller applied (in * asconf case when deleting an address) or NOT in the HB case, * however if HB then the INP increment is up and the INP will not * be removed (on top of the fact that we have a TCB lock). So we * only want to read the sctp_flags, which is either bound-all or * not.. no protection needed since once an assoc is up you can't be * changing your binding. */ inp = stcb->sctp_ep; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* if bound all, destination is not restricted */ /* * RRS: Question during lock work: Is this correct? If you * are bound-all you still might need to obey the V4--V6 * flags??? IMO this bound-all stuff needs to be removed! */ return (1); } /* NOTE: all "scope" checks are done when local addresses are added */ if (destaddr->sa_family == AF_INET6) { answer = inp->ip_inp.inp.inp_vflag & INP_IPV6; } else if (destaddr->sa_family == AF_INET) { answer = inp->ip_inp.inp.inp_vflag & INP_IPV4; } else { /* invalid family, so it's unreachable */ answer = 0; } return (answer); } /* * update the inp_vflags on an endpoint */ static void sctp_update_ep_vflag(struct sctp_inpcb *inp) { struct sctp_laddr *laddr; /* first clear the flag */ inp->ip_inp.inp.inp_vflag = 0; /* set the flag based on addresses on the ep list */ LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__); continue; } if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { continue; } if (laddr->ifa->address.sa.sa_family == AF_INET6) { inp->ip_inp.inp.inp_vflag |= INP_IPV6; } else if (laddr->ifa->address.sa.sa_family == AF_INET) { inp->ip_inp.inp.inp_vflag |= INP_IPV4; } } } /* * Add the address to the endpoint local address list There is nothing to be * done if we are bound to all addresses */ void sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action) { struct sctp_laddr *laddr; int fnd, error = 0; fnd = 0; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* You are already bound to all. You have it already */ return; } if (ifa->address.sa.sa_family == AF_INET6) { if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { /* Can't bind a non-useable addr. */ return; } } /* first, is it already present? */ LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == ifa) { fnd = 1; break; } } if (fnd == 0) { /* Not in the ep list */ error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action); if (error != 0) return; inp->laddr_count++; /* update inp_vflag flags */ if (ifa->address.sa.sa_family == AF_INET6) { inp->ip_inp.inp.inp_vflag |= INP_IPV6; } else if (ifa->address.sa.sa_family == AF_INET) { inp->ip_inp.inp.inp_vflag |= INP_IPV4; } } return; } /* * select a new (hopefully reachable) destination net (should only be used * when we deleted an ep addr that is the only usable source address to reach * the destination net) */ static void sctp_select_primary_destination(struct sctp_tcb *stcb) { struct sctp_nets *net; TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { /* for now, we'll just pick the first reachable one we find */ if (net->dest_state & SCTP_ADDR_UNCONFIRMED) continue; if (sctp_destination_is_reachable(stcb, (struct sockaddr *)&net->ro._l_addr)) { /* found a reachable destination */ stcb->asoc.primary_destination = net; } } /* I can't there from here! ...we're gonna die shortly... */ } /* * Delete the address from the endpoint local address list There is nothing * to be done if we are bound to all addresses */ void sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa) { struct sctp_laddr *laddr; int fnd; fnd = 0; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* You are already bound to all. You have it already */ return; } LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == ifa) { fnd = 1; break; } } if (fnd && (inp->laddr_count < 2)) { /* can't delete unless there are at LEAST 2 addresses */ return; } if (fnd) { /* * clean up any use of this address go through our * associations and clear any last_used_address that match * this one for each assoc, see if a new primary_destination * is needed */ struct sctp_tcb *stcb; /* clean up "next_addr_touse" */ if (inp->next_addr_touse == laddr) /* delete this address */ inp->next_addr_touse = NULL; /* clean up "last_used_address" */ LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { struct sctp_nets *net; SCTP_TCB_LOCK(stcb); if (stcb->asoc.last_used_address == laddr) /* delete this address */ stcb->asoc.last_used_address = NULL; /* * Now spin through all the nets and purge any ref * to laddr */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (net->ro._s_addr && (net->ro._s_addr->ifa == laddr->ifa)) { /* Yep, purge src address selected */ sctp_rtentry_t *rt; /* delete this address if cached */ rt = net->ro.ro_rt; if (rt != NULL) { RTFREE(rt); net->ro.ro_rt = NULL; } sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; net->src_addr_selected = 0; } } SCTP_TCB_UNLOCK(stcb); } /* for each tcb */ /* remove it from the ep list */ sctp_remove_laddr(laddr); inp->laddr_count--; /* update inp_vflag flags */ sctp_update_ep_vflag(inp); } return; } /* * Add the address to the TCB local address restricted list. * This is a "pending" address list (eg. addresses waiting for an * ASCONF-ACK response) and cannot be used as a valid source address. */ void sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa) { struct sctp_inpcb *inp; struct sctp_laddr *laddr; struct sctpladdr *list; /* * Assumes TCB is locked.. and possibly the INP. May need to * confirm/fix that if we need it and is not the case. */ list = &stcb->asoc.sctp_restricted_addrs; inp = stcb->sctp_ep; if (ifa->address.sa.sa_family == AF_INET6) { if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { /* Can't bind a non-existent addr. */ return; } } /* does the address already exist? */ LIST_FOREACH(laddr, list, sctp_nxt_addr) { if (laddr->ifa == ifa) { return; } } /* add to the list */ (void)sctp_insert_laddr(list, ifa, 0); return; } /* * insert an laddr entry with the given ifa for the desired list */ int sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act) { struct sctp_laddr *laddr; laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (laddr == NULL) { /* out of memory? */ SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } SCTP_INCR_LADDR_COUNT(); bzero(laddr, sizeof(*laddr)); (void)SCTP_GETTIME_TIMEVAL(&laddr->start_time); laddr->ifa = ifa; laddr->action = act; atomic_add_int(&ifa->refcount, 1); /* insert it */ LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr); return (0); } /* * Remove an laddr entry from the local address list (on an assoc) */ void sctp_remove_laddr(struct sctp_laddr *laddr) { /* remove from the list */ LIST_REMOVE(laddr, sctp_nxt_addr); sctp_free_ifa(laddr->ifa); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr); SCTP_DECR_LADDR_COUNT(); } /* * Remove a local address from the TCB local address restricted list */ void sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa) { struct sctp_inpcb *inp; struct sctp_laddr *laddr; /* * This is called by asconf work. It is assumed that a) The TCB is * locked and b) The INP is locked. This is true in as much as I can * trace through the entry asconf code where I did these locks. * Again, the ASCONF code is a bit different in that it does lock * the INP during its work often times. This must be since we don't * want other proc's looking up things while what they are looking * up is changing :-D */ inp = stcb->sctp_ep; /* if subset bound and don't allow ASCONF's, can't delete last */ if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) && sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) { if (stcb->sctp_ep->laddr_count < 2) { /* can't delete last address */ return; } } LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) { /* remove the address if it exists */ if (laddr->ifa == NULL) continue; if (laddr->ifa == ifa) { sctp_remove_laddr(laddr); return; } } /* address not found! */ return; } /* * Temporarily remove for __APPLE__ until we use the Tiger equivalents */ /* sysctl */ static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC; static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR; void sctp_pcb_init() { /* * SCTP initialization for the PCB structures should be called by * the sctp_init() funciton. */ int i; struct timeval tv; if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) { /* error I was called twice */ return; } SCTP_BASE_VAR(sctp_pcb_initialized) = 1; #if defined(SCTP_LOCAL_TRACE_BUF) bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log)); #endif (void)SCTP_GETTIME_TIMEVAL(&tv); SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec; SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec; /* init the empty list of (All) Endpoints */ LIST_INIT(&SCTP_BASE_INFO(listhead)); /* init the iterator head */ TAILQ_INIT(&SCTP_BASE_INFO(iteratorhead)); /* init the hash table of endpoints */ TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize)); TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize)); TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale)); SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31), &SCTP_BASE_INFO(hashasocmark)); SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize), &SCTP_BASE_INFO(hashmark)); SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize), &SCTP_BASE_INFO(hashtcpmark)); SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize); /* init the small hash table we use to track restarted asoc's */ SCTP_BASE_INFO(sctp_restarthash) = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &SCTP_BASE_INFO(hashrestartmark)); SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH, &SCTP_BASE_INFO(hashvrfmark)); SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE, &SCTP_BASE_INFO(vrf_ifn_hashmark)); /* init the zones */ /* * FIX ME: Should check for NULL returns, but if it does fail we are * doomed to panic anyways... add later maybe. */ SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep", sizeof(struct sctp_inpcb), maxsockets); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc", sizeof(struct sctp_tcb), sctp_max_number_of_assoc); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr", sizeof(struct sctp_laddr), (sctp_max_number_of_assoc * sctp_scale_up_for_address)); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr", sizeof(struct sctp_nets), (sctp_max_number_of_assoc * sctp_scale_up_for_address)); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk", sizeof(struct sctp_tmit_chunk), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq", sizeof(struct sctp_queued_to_read), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out", sizeof(struct sctp_stream_queue_pending), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf", sizeof(struct sctp_asconf), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack", sizeof(struct sctp_asconf_ack), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); /* Master Lock INIT for info structure */ SCTP_INP_INFO_LOCK_INIT(); SCTP_STATLOG_INIT_LOCK(); SCTP_ITERATOR_LOCK_INIT(); SCTP_IPI_COUNT_INIT(); SCTP_IPI_ADDR_INIT(); SCTP_IPI_ITERATOR_WQ_INIT(); #ifdef SCTP_PACKET_LOGGING SCTP_IP_PKTLOG_INIT(); #endif LIST_INIT(&SCTP_BASE_INFO(addr_wq)); /* not sure if we need all the counts */ SCTP_BASE_INFO(ipi_count_ep) = 0; /* assoc/tcb zone info */ SCTP_BASE_INFO(ipi_count_asoc) = 0; /* local addrlist zone info */ SCTP_BASE_INFO(ipi_count_laddr) = 0; /* remote addrlist zone info */ SCTP_BASE_INFO(ipi_count_raddr) = 0; /* chunk info */ SCTP_BASE_INFO(ipi_count_chunk) = 0; /* socket queue zone info */ SCTP_BASE_INFO(ipi_count_readq) = 0; /* stream out queue cont */ SCTP_BASE_INFO(ipi_count_strmoq) = 0; SCTP_BASE_INFO(ipi_free_strmoq) = 0; SCTP_BASE_INFO(ipi_free_chunks) = 0; SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer)); /* Init the TIMEWAIT list */ for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE_A; i++) { LIST_INIT(&SCTP_BASE_INFO(vtag_timewait[i])); } #if defined(SCTP_USE_THREAD_BASED_ITERATOR) SCTP_BASE_INFO(iterator_running) = 0; SCTP_BASE_INFO(threads_must_exit) = 0; sctp_startup_iterator(); #endif /* * INIT the default VRF which for BSD is the only one, other O/S's * may have more. But initially they must start with one and then * add the VRF's as addresses are added. */ sctp_init_vrf_list(SCTP_DEFAULT_VRF); } /* * Assumes that the SCTP_BASE_INFO() lock is NOT held. */ void sctp_pcb_finish(void) { struct sctp_vrflist *vrf_bucket; struct sctp_vrf *vrf; struct sctp_ifn *ifn; struct sctp_ifa *ifa; struct sctpvtaghead *chain; struct sctp_tagblock *twait_block, *prev_twait_block; int i; /* * free the vrf/ifn/ifa lists and hashes (be sure address monitor is * destroyed first). */ vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))]; vrf = LIST_FIRST(vrf_bucket); while (vrf) { ifn = LIST_FIRST(&vrf->ifnlist); while (ifn) { ifa = LIST_FIRST(&ifn->ifalist); while (ifa) { /* free the ifa */ LIST_REMOVE(ifa, next_bucket); LIST_REMOVE(ifa, next_ifa); SCTP_FREE(ifa, SCTP_M_IFA); ifa = LIST_FIRST(&ifn->ifalist); } /* free the ifn */ LIST_REMOVE(ifn, next_bucket); LIST_REMOVE(ifn, next_ifn); SCTP_FREE(ifn, SCTP_M_IFN); ifn = LIST_FIRST(&vrf->ifnlist); } SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark); /* free the vrf */ LIST_REMOVE(vrf, next_vrf); SCTP_FREE(vrf, SCTP_M_VRF); vrf = LIST_FIRST(vrf_bucket); } /* free the vrf hashes */ SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark)); SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark)); /* * free the TIMEWAIT list elements malloc'd in the function * sctp_add_vtag_to_timewait()... */ for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE_A; i++) { chain = &SCTP_BASE_INFO(vtag_timewait)[i]; if (!SCTP_LIST_EMPTY(chain)) { prev_twait_block = NULL; LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { if (prev_twait_block) { SCTP_FREE(prev_twait_block, SCTP_M_TIMW); } prev_twait_block = twait_block; } SCTP_FREE(prev_twait_block, SCTP_M_TIMW); } } /* free the locks and mutexes */ #ifdef SCTP_PACKET_LOGGING SCTP_IP_PKTLOG_DESTROY(); #endif SCTP_IPI_ITERATOR_WQ_DESTROY(); SCTP_IPI_ADDR_DESTROY(); SCTP_ITERATOR_LOCK_DESTROY(); SCTP_STATLOG_DESTROY(); SCTP_INP_INFO_LOCK_DESTROY(); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack)); /* Get rid of other stuff to */ if (SCTP_BASE_INFO(sctp_asochash) != NULL) SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark)); if (SCTP_BASE_INFO(sctp_ephash) != NULL) SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark)); if (SCTP_BASE_INFO(sctp_tcpephash) != NULL) SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark)); if (SCTP_BASE_INFO(sctp_restarthash) != NULL) SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_restarthash), SCTP_BASE_INFO(hashrestartmark)); } int sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m, int iphlen, int offset, int limit, struct sctphdr *sh, struct sockaddr *altsa) { /* * grub through the INIT pulling addresses and loading them to the * nets structure in the asoc. The from address in the mbuf should * also be loaded (if it is not already). This routine can be called * with either INIT or INIT-ACK's as long as the m points to the IP * packet and the offset points to the beginning of the parameters. */ struct sctp_inpcb *inp, *l_inp; struct sctp_nets *net, *net_tmp; struct ip *iph; struct sctp_paramhdr *phdr, parm_buf; struct sctp_tcb *stcb_tmp; uint16_t ptype, plen; struct sockaddr *sa; struct sockaddr_storage dest_store; struct sockaddr *local_sa = (struct sockaddr *)&dest_store; struct sockaddr_in sin; struct sockaddr_in6 sin6; uint8_t random_store[SCTP_PARAM_BUFFER_SIZE]; struct sctp_auth_random *p_random = NULL; uint16_t random_len = 0; uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE]; struct sctp_auth_hmac_algo *hmacs = NULL; uint16_t hmacs_len = 0; uint8_t saw_asconf = 0; uint8_t saw_asconf_ack = 0; uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE]; struct sctp_auth_chunk_list *chunks = NULL; uint16_t num_chunks = 0; sctp_key_t *new_key; uint32_t keylen; int got_random = 0, got_hmacs = 0, got_chklist = 0; /* First get the destination address setup too. */ memset(&sin, 0, sizeof(sin)); memset(&sin6, 0, sizeof(sin6)); sin.sin_family = AF_INET; sin.sin_len = sizeof(sin); sin.sin_port = stcb->rport; sin6.sin6_family = AF_INET6; sin6.sin6_len = sizeof(struct sockaddr_in6); sin6.sin6_port = stcb->rport; if (altsa == NULL) { iph = mtod(m, struct ip *); switch (iph->ip_v) { case IPVERSION: { /* its IPv4 */ struct sockaddr_in *sin_2; sin_2 = (struct sockaddr_in *)(local_sa); memset(sin_2, 0, sizeof(sin)); sin_2->sin_family = AF_INET; sin_2->sin_len = sizeof(sin); sin_2->sin_port = sh->dest_port; sin_2->sin_addr.s_addr = iph->ip_dst.s_addr; sin.sin_addr = iph->ip_src; sa = (struct sockaddr *)&sin; break; } #ifdef INET6 case IPV6_VERSION >> 4: { /* its IPv6 */ struct ip6_hdr *ip6; struct sockaddr_in6 *sin6_2; ip6 = mtod(m, struct ip6_hdr *); sin6_2 = (struct sockaddr_in6 *)(local_sa); memset(sin6_2, 0, sizeof(sin6)); sin6_2->sin6_family = AF_INET6; sin6_2->sin6_len = sizeof(struct sockaddr_in6); sin6_2->sin6_port = sh->dest_port; sin6.sin6_addr = ip6->ip6_src; sa = (struct sockaddr *)&sin6; break; } #endif default: sa = NULL; break; } } else { /* * For cookies we use the src address NOT from the packet * but from the original INIT */ sa = altsa; } /* Turn off ECN until we get through all params */ stcb->asoc.ecn_allowed = 0; TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { /* mark all addresses that we have currently on the list */ net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC; } /* does the source address already exist? if so skip it */ l_inp = inp = stcb->sctp_ep; atomic_add_int(&stcb->asoc.refcnt, 1); stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, local_sa, stcb); atomic_add_int(&stcb->asoc.refcnt, -1); if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) { /* we must add the source address */ /* no scope set here since we have a tcb already. */ if ((sa->sa_family == AF_INET) && (stcb->asoc.ipv4_addr_legal)) { if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) { return (-1); } } else if ((sa->sa_family == AF_INET6) && (stcb->asoc.ipv6_addr_legal)) { if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) { return (-2); } } } else { if (net_tmp != NULL && stcb_tmp == stcb) { net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC; } else if (stcb_tmp != stcb) { /* It belongs to another association? */ if (stcb_tmp) SCTP_TCB_UNLOCK(stcb_tmp); return (-3); } } if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-4); } /* * peer must explicitly turn this on. This may have been initialized * to be "on" in order to allow local addr changes while INIT's are * in flight. */ stcb->asoc.peer_supports_asconf = 0; /* now we must go through each of the params. */ phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf)); while (phdr) { ptype = ntohs(phdr->param_type); plen = ntohs(phdr->param_length); /* * printf("ptype => %0x, plen => %d\n", (uint32_t)ptype, * (int)plen); */ if (offset + plen > limit) { break; } if (plen == 0) { break; } if (ptype == SCTP_IPV4_ADDRESS) { if (stcb->asoc.ipv4_addr_legal) { struct sctp_ipv4addr_param *p4, p4_buf; /* ok get the v4 address and check/add */ phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&p4_buf, sizeof(p4_buf)); if (plen != sizeof(struct sctp_ipv4addr_param) || phdr == NULL) { return (-5); } p4 = (struct sctp_ipv4addr_param *)phdr; sin.sin_addr.s_addr = p4->addr; if (IN_MULTICAST(sin.sin_addr.s_addr)) { /* Skip multi-cast addresses */ goto next_param; } if ((sin.sin_addr.s_addr == INADDR_BROADCAST) || (sin.sin_addr.s_addr == INADDR_ANY)) { goto next_param; } sa = (struct sockaddr *)&sin; inp = stcb->sctp_ep; atomic_add_int(&stcb->asoc.refcnt, 1); stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net, local_sa, stcb); atomic_add_int(&stcb->asoc.refcnt, -1); if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) { /* we must add the source address */ /* * no scope set since we have a tcb * already */ /* * we must validate the state again * here */ + add_it_now: if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-7); } if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) { return (-8); } } else if (stcb_tmp == stcb) { if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-10); } if (net != NULL) { /* clear flag */ net->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC; } } else { /* * strange, address is in another * assoc? straighten out locks. */ - if (stcb_tmp) + if (stcb_tmp) { + if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) { + /* + * in setup state we + * abort this guy + */ + sctp_abort_an_association(stcb_tmp->sctp_ep, + stcb_tmp, 1, NULL, 0); + goto add_it_now; + } SCTP_TCB_UNLOCK(stcb_tmp); - + } if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-12); } return (-13); } } } else if (ptype == SCTP_IPV6_ADDRESS) { if (stcb->asoc.ipv6_addr_legal) { /* ok get the v6 address and check/add */ struct sctp_ipv6addr_param *p6, p6_buf; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&p6_buf, sizeof(p6_buf)); if (plen != sizeof(struct sctp_ipv6addr_param) || phdr == NULL) { return (-14); } p6 = (struct sctp_ipv6addr_param *)phdr; memcpy((caddr_t)&sin6.sin6_addr, p6->addr, sizeof(p6->addr)); if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) { /* Skip multi-cast addresses */ goto next_param; } if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) { /* * Link local make no sense without * scope */ goto next_param; } sa = (struct sockaddr *)&sin6; inp = stcb->sctp_ep; atomic_add_int(&stcb->asoc.refcnt, 1); stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net, local_sa, stcb); atomic_add_int(&stcb->asoc.refcnt, -1); if (stcb_tmp == NULL && (inp == stcb->sctp_ep || inp == NULL)) { /* * we must validate the state again * here */ + add_it_now6: if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-16); } /* * we must add the address, no scope * set */ if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) { return (-17); } } else if (stcb_tmp == stcb) { /* * we must validate the state again * here */ if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-19); } if (net != NULL) { /* clear flag */ net->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC; } } else { /* * strange, address is in another * assoc? straighten out locks. */ if (stcb_tmp) - SCTP_TCB_UNLOCK(stcb_tmp); + if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) { + /* + * in setup state we + * abort this guy + */ + sctp_abort_an_association(stcb_tmp->sctp_ep, + stcb_tmp, 1, NULL, 0); + goto add_it_now6; + } + SCTP_TCB_UNLOCK(stcb_tmp); if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-21); } return (-22); } } } else if (ptype == SCTP_ECN_CAPABLE) { stcb->asoc.ecn_allowed = 1; } else if (ptype == SCTP_ULP_ADAPTATION) { if (stcb->asoc.state != SCTP_STATE_OPEN) { struct sctp_adaptation_layer_indication ai, *aip; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&ai, sizeof(ai)); aip = (struct sctp_adaptation_layer_indication *)phdr; if (aip) { stcb->asoc.peers_adaptation = ntohl(aip->indication); stcb->asoc.adaptation_needed = 1; } } } else if (ptype == SCTP_SET_PRIM_ADDR) { struct sctp_asconf_addr_param lstore, *fee; struct sctp_asconf_addrv4_param *fii; int lptype; struct sockaddr *lsa = NULL; stcb->asoc.peer_supports_asconf = 1; if (plen > sizeof(lstore)) { return (-23); } phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&lstore, min(plen, sizeof(lstore))); if (phdr == NULL) { return (-24); } fee = (struct sctp_asconf_addr_param *)phdr; lptype = ntohs(fee->addrp.ph.param_type); if (lptype == SCTP_IPV4_ADDRESS) { if (plen != sizeof(struct sctp_asconf_addrv4_param)) { SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n", (int)sizeof(struct sctp_asconf_addrv4_param), plen); } else { fii = (struct sctp_asconf_addrv4_param *)fee; sin.sin_addr.s_addr = fii->addrp.addr; lsa = (struct sockaddr *)&sin; } } else if (lptype == SCTP_IPV6_ADDRESS) { if (plen != sizeof(struct sctp_asconf_addr_param)) { SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n", (int)sizeof(struct sctp_asconf_addr_param), plen); } else { memcpy(sin6.sin6_addr.s6_addr, fee->addrp.addr, sizeof(fee->addrp.addr)); lsa = (struct sockaddr *)&sin6; } } if (lsa) { (void)sctp_set_primary_addr(stcb, sa, NULL); } } else if (ptype == SCTP_PRSCTP_SUPPORTED) { /* Peer supports pr-sctp */ stcb->asoc.peer_supports_prsctp = 1; } else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) { /* A supported extension chunk */ struct sctp_supported_chunk_types_param *pr_supported; uint8_t local_store[SCTP_PARAM_BUFFER_SIZE]; int num_ent, i; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen)); if (phdr == NULL) { return (-25); } stcb->asoc.peer_supports_asconf = 0; stcb->asoc.peer_supports_prsctp = 0; stcb->asoc.peer_supports_pktdrop = 0; stcb->asoc.peer_supports_strreset = 0; stcb->asoc.peer_supports_auth = 0; pr_supported = (struct sctp_supported_chunk_types_param *)phdr; num_ent = plen - sizeof(struct sctp_paramhdr); for (i = 0; i < num_ent; i++) { switch (pr_supported->chunk_types[i]) { case SCTP_ASCONF: case SCTP_ASCONF_ACK: stcb->asoc.peer_supports_asconf = 1; break; case SCTP_FORWARD_CUM_TSN: stcb->asoc.peer_supports_prsctp = 1; break; case SCTP_PACKET_DROPPED: stcb->asoc.peer_supports_pktdrop = 1; break; case SCTP_STREAM_RESET: stcb->asoc.peer_supports_strreset = 1; break; case SCTP_AUTHENTICATION: stcb->asoc.peer_supports_auth = 1; break; default: /* one I have not learned yet */ break; } } } else if (ptype == SCTP_ECN_NONCE_SUPPORTED) { /* Peer supports ECN-nonce */ stcb->asoc.peer_supports_ecn_nonce = 1; stcb->asoc.ecn_nonce_allowed = 1; } else if (ptype == SCTP_RANDOM) { if (plen > sizeof(random_store)) break; if (got_random) { /* already processed a RANDOM */ goto next_param; } phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)random_store, min(sizeof(random_store), plen)); if (phdr == NULL) return (-26); p_random = (struct sctp_auth_random *)phdr; random_len = plen - sizeof(*p_random); /* enforce the random length */ if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) { SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n"); return (-27); } got_random = 1; } else if (ptype == SCTP_HMAC_LIST) { int num_hmacs; int i; if (plen > sizeof(hmacs_store)) break; if (got_hmacs) { /* already processed a HMAC list */ goto next_param; } phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)hmacs_store, min(plen, sizeof(hmacs_store))); if (phdr == NULL) return (-28); hmacs = (struct sctp_auth_hmac_algo *)phdr; hmacs_len = plen - sizeof(*hmacs); num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]); /* validate the hmac list */ if (sctp_verify_hmac_param(hmacs, num_hmacs)) { return (-29); } if (stcb->asoc.peer_hmacs != NULL) sctp_free_hmaclist(stcb->asoc.peer_hmacs); stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs); if (stcb->asoc.peer_hmacs != NULL) { for (i = 0; i < num_hmacs; i++) { (void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs, ntohs(hmacs->hmac_ids[i])); } } got_hmacs = 1; } else if (ptype == SCTP_CHUNK_LIST) { int i; if (plen > sizeof(chunks_store)) break; if (got_chklist) { /* already processed a Chunks list */ goto next_param; } phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)chunks_store, min(plen, sizeof(chunks_store))); if (phdr == NULL) return (-30); chunks = (struct sctp_auth_chunk_list *)phdr; num_chunks = plen - sizeof(*chunks); if (stcb->asoc.peer_auth_chunks != NULL) sctp_clear_chunklist(stcb->asoc.peer_auth_chunks); else stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist(); for (i = 0; i < num_chunks; i++) { (void)sctp_auth_add_chunk(chunks->chunk_types[i], stcb->asoc.peer_auth_chunks); /* record asconf/asconf-ack if listed */ if (chunks->chunk_types[i] == SCTP_ASCONF) saw_asconf = 1; if (chunks->chunk_types[i] == SCTP_ASCONF_ACK) saw_asconf_ack = 1; } got_chklist = 1; } else if ((ptype == SCTP_HEARTBEAT_INFO) || (ptype == SCTP_STATE_COOKIE) || (ptype == SCTP_UNRECOG_PARAM) || (ptype == SCTP_COOKIE_PRESERVE) || (ptype == SCTP_SUPPORTED_ADDRTYPE) || (ptype == SCTP_ADD_IP_ADDRESS) || (ptype == SCTP_DEL_IP_ADDRESS) || (ptype == SCTP_ERROR_CAUSE_IND) || (ptype == SCTP_SUCCESS_REPORT)) { /* don't care */ ; } else { if ((ptype & 0x8000) == 0x0000) { /* * must stop processing the rest of the * param's. Any report bits were handled * with the call to * sctp_arethere_unrecognized_parameters() * when the INIT or INIT-ACK was first seen. */ break; } } next_param: offset += SCTP_SIZE32(plen); if (offset >= limit) { break; } phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf)); } /* Now check to see if we need to purge any addresses */ for (net = TAILQ_FIRST(&stcb->asoc.nets); net != NULL; net = net_tmp) { net_tmp = TAILQ_NEXT(net, sctp_next); if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) == SCTP_ADDR_NOT_IN_ASSOC) { /* This address has been removed from the asoc */ /* remove and free it */ stcb->asoc.numnets--; TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next); sctp_free_remote_addr(net); if (net == stcb->asoc.primary_destination) { stcb->asoc.primary_destination = NULL; sctp_select_primary_destination(stcb); } } } /* validate authentication required parameters */ if (got_random && got_hmacs) { stcb->asoc.peer_supports_auth = 1; } else { stcb->asoc.peer_supports_auth = 0; } if (!stcb->asoc.peer_supports_auth && got_chklist) { /* peer does not support auth but sent a chunks list? */ return (-31); } if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf && !stcb->asoc.peer_supports_auth) { /* peer supports asconf but not auth? */ return (-32); } else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) && ((saw_asconf == 0) || (saw_asconf_ack == 0))) { return (-33); } /* concatenate the full random key */ #ifdef SCTP_AUTH_DRAFT_04 keylen = random_len; new_key = sctp_alloc_key(keylen); if (new_key != NULL) { /* copy in the RANDOM */ if (p_random != NULL) bcopy(p_random->random_data, new_key->key, random_len); } #else keylen = sizeof(*p_random) + random_len + sizeof(*chunks) + num_chunks + sizeof(*hmacs) + hmacs_len; new_key = sctp_alloc_key(keylen); if (new_key != NULL) { /* copy in the RANDOM */ if (p_random != NULL) { keylen = sizeof(*p_random) + random_len; bcopy(p_random, new_key->key, keylen); } /* append in the AUTH chunks */ if (chunks != NULL) { bcopy(chunks, new_key->key + keylen, sizeof(*chunks) + num_chunks); keylen += sizeof(*chunks) + num_chunks; } /* append in the HMACs */ if (hmacs != NULL) { bcopy(hmacs, new_key->key + keylen, sizeof(*hmacs) + hmacs_len); } } #endif else { /* failed to get memory for the key */ return (-34); } if (stcb->asoc.authinfo.peer_random != NULL) sctp_free_key(stcb->asoc.authinfo.peer_random); stcb->asoc.authinfo.peer_random = new_key; #ifdef SCTP_AUTH_DRAFT_04 /* don't include the chunks and hmacs for draft -04 */ stcb->asoc.authinfo.peer_random->keylen = random_len; #endif sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid); sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid); return (0); } int sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa, struct sctp_nets *net) { /* make sure the requested primary address exists in the assoc */ if (net == NULL && sa) net = sctp_findnet(stcb, sa); if (net == NULL) { /* didn't find the requested primary address! */ return (-1); } else { /* set the primary address */ if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { /* Must be confirmed, so queue to set */ net->dest_state |= SCTP_ADDR_REQ_PRIMARY; return (0); } stcb->asoc.primary_destination = net; net->dest_state &= ~SCTP_ADDR_WAS_PRIMARY; net = TAILQ_FIRST(&stcb->asoc.nets); if (net != stcb->asoc.primary_destination) { /* * first one on the list is NOT the primary * sctp_cmpaddr() is much more efficent if the * primary is the first on the list, make it so. */ TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); } return (0); } } int sctp_is_vtag_good(struct sctp_inpcb *inp, uint32_t tag, struct timeval *now, int save_in_twait) { /* * This function serves two purposes. It will see if a TAG can be * re-used and return 1 for yes it is ok and 0 for don't use that * tag. A secondary function it will do is purge out old tags that * can be removed. */ struct sctpasochead *head; struct sctpvtaghead *chain; struct sctp_tagblock *twait_block; struct sctp_tcb *stcb; int i; SCTP_INP_INFO_WLOCK(); chain = &SCTP_BASE_INFO(vtag_timewait[(tag % SCTP_STACK_VTAG_HASH_SIZE))]; /* First is the vtag in use ? */ head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag, SCTP_BASE_INFO(hashasocmark))]; if (head == NULL) { goto check_restart; } LIST_FOREACH(stcb, head, sctp_asocs) { if (stcb->asoc.my_vtag == tag) { /* * We should remove this if and return 0 always if * we want vtags unique across all endpoints. For * now within a endpoint is ok. */ if (inp == stcb->sctp_ep) { /* bad tag, in use */ SCTP_INP_INFO_WUNLOCK(); return (0); } } } check_restart: /* Now lets check the restart hash */ head = &SCTP_BASE_INFO(sctp_restarthash)[SCTP_PCBHASH_ASOC(tag, SCTP_BASE_INFO(hashrestartmark))]; if (head == NULL) { goto check_time_wait; } LIST_FOREACH(stcb, head, sctp_tcbrestarhash) { if (stcb->asoc.assoc_id == tag) { /* candidate */ if (inp == stcb->sctp_ep) { /* bad tag, in use */ SCTP_INP_INFO_WUNLOCK(); return (0); } } } check_time_wait: /* Now what about timed wait ? */ if (!SCTP_LIST_EMPTY(chain)) { /* * Block(s) are present, lets see if we have this tag in the * list */ LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { if (twait_block->vtag_block[i].v_tag == 0) { /* not used */ continue; } else if ((long)twait_block->vtag_block[i].tv_sec_at_expire < now->tv_sec) { /* Audit expires this guy */ twait_block->vtag_block[i].tv_sec_at_expire = 0; twait_block->vtag_block[i].v_tag = 0; } else if (twait_block->vtag_block[i].v_tag == tag) { /* Bad tag, sorry :< */ SCTP_INP_INFO_WUNLOCK(); return (0); } } } } /*- * Not found, ok to use the tag, add it to the time wait hash * as well this will prevent two sucessive cookies from getting * the same tag or two inits sent quickly on multi-processors. * We only keep the tag for the life of a cookie and when we * add this tag to the assoc hash we need to purge it from * the t-wait hash. */ if (save_in_twait) sctp_add_vtag_to_timewait(tag, TICKS_TO_SEC(inp->sctp_ep.def_cookie_life)); SCTP_INP_INFO_WUNLOCK(); return (1); } static sctp_assoc_t reneged_asoc_ids[256]; static uint8_t reneged_at = 0; static void sctp_drain_mbufs(struct sctp_inpcb *inp, struct sctp_tcb *stcb) { /* * We must hunt this association for MBUF's past the cumack (i.e. * out of order data that we can renege on). */ struct sctp_association *asoc; struct sctp_tmit_chunk *chk, *nchk; uint32_t cumulative_tsn_p1, tsn; struct sctp_queued_to_read *ctl, *nctl; int cnt, strmat, gap; /* We look for anything larger than the cum-ack + 1 */ SCTP_STAT_INCR(sctps_protocol_drain_calls); if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) { return; } asoc = &stcb->asoc; if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) { /* none we can reneg on. */ return; } SCTP_STAT_INCR(sctps_protocol_drains_done); cumulative_tsn_p1 = asoc->cumulative_tsn + 1; cnt = 0; /* First look in the re-assembly queue */ chk = TAILQ_FIRST(&asoc->reasmqueue); while (chk) { /* Get the next one */ nchk = TAILQ_NEXT(chk, sctp_next); if (compare_with_wrap(chk->rec.data.TSN_seq, cumulative_tsn_p1, MAX_TSN)) { /* Yep it is above cum-ack */ cnt++; tsn = chk->rec.data.TSN_seq; if (tsn >= asoc->mapping_array_base_tsn) { gap = tsn - asoc->mapping_array_base_tsn; } else { gap = (MAX_TSN - asoc->mapping_array_base_tsn) + tsn + 1; } asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size); sctp_ucount_decr(asoc->cnt_on_reasm_queue); SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } sctp_free_a_chunk(stcb, chk); } chk = nchk; } /* Ok that was fun, now we will drain all the inbound streams? */ for (strmat = 0; strmat < asoc->streamincnt; strmat++) { ctl = TAILQ_FIRST(&asoc->strmin[strmat].inqueue); while (ctl) { nctl = TAILQ_NEXT(ctl, next); if (compare_with_wrap(ctl->sinfo_tsn, cumulative_tsn_p1, MAX_TSN)) { /* Yep it is above cum-ack */ cnt++; tsn = ctl->sinfo_tsn; if (tsn >= asoc->mapping_array_base_tsn) { gap = tsn - asoc->mapping_array_base_tsn; } else { gap = (MAX_TSN - asoc->mapping_array_base_tsn) + tsn + 1; } asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length); sctp_ucount_decr(asoc->cnt_on_all_streams); SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next); if (ctl->data) { sctp_m_freem(ctl->data); ctl->data = NULL; } sctp_free_remote_addr(ctl->whoFrom); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl); SCTP_DECR_READQ_COUNT(); } ctl = nctl; } } /* * Question, should we go through the delivery queue? The only * reason things are on here is the app not reading OR a p-d-api up. * An attacker COULD send enough in to initiate the PD-API and then * send a bunch of stuff to other streams... these would wind up on * the delivery queue.. and then we would not get to them. But in * order to do this I then have to back-track and un-deliver * sequence numbers in streams.. el-yucko. I think for now we will * NOT look at the delivery queue and leave it to be something to * consider later. An alternative would be to abort the P-D-API with * a notification and then deliver the data.... Or another method * might be to keep track of how many times the situation occurs and * if we see a possible attack underway just abort the association. */ #ifdef SCTP_DEBUG if (cnt) { SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt); } #endif if (cnt) { /* * Now do we need to find a new * asoc->highest_tsn_inside_map? */ if (asoc->highest_tsn_inside_map >= asoc->mapping_array_base_tsn) { gap = asoc->highest_tsn_inside_map - asoc->mapping_array_base_tsn; } else { gap = (MAX_TSN - asoc->mapping_array_base_tsn) + asoc->highest_tsn_inside_map + 1; } if (gap >= (asoc->mapping_array_size << 3)) { /* * Something bad happened or cum-ack and high were * behind the base, but if so earlier checks should * have found NO data... wierd... we will start at * end of mapping array. */ SCTP_PRINTF("Gap was larger than array?? %d set to max:%d maparraymax:%x\n", (int)gap, (int)(asoc->mapping_array_size << 3), (int)asoc->highest_tsn_inside_map); gap = asoc->mapping_array_size << 3; } while (gap > 0) { if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) { /* found the new highest */ asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn + gap; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { sctp_log_map(0, 8, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT); } break; } gap--; } if (gap == 0) { /* Nothing left in map */ memset(asoc->mapping_array, 0, asoc->mapping_array_size); asoc->mapping_array_base_tsn = asoc->cumulative_tsn + 1; asoc->highest_tsn_inside_map = asoc->cumulative_tsn; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) { sctp_log_map(0, 9, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT); } } asoc->last_revoke_count = cnt; (void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer); /* sa_ignore NO_NULL_CHK */ sctp_send_sack(stcb); sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED); reneged_asoc_ids[reneged_at] = sctp_get_associd(stcb); reneged_at++; } /* * Another issue, in un-setting the TSN's in the mapping array we * DID NOT adjust the higest_tsn marker. This will cause one of two * things to occur. It may cause us to do extra work in checking for * our mapping array movement. More importantly it may cause us to * SACK every datagram. This may not be a bad thing though since we * will recover once we get our cum-ack above and all this stuff we * dumped recovered. */ } void sctp_drain() { /* * We must walk the PCB lists for ALL associations here. The system * is LOW on MBUF's and needs help. This is where reneging will * occur. We really hope this does NOT happen! */ struct sctp_inpcb *inp; struct sctp_tcb *stcb; SCTP_INP_INFO_RLOCK(); LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) { /* For each endpoint */ SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { /* For each association */ SCTP_TCB_LOCK(stcb); sctp_drain_mbufs(inp, stcb); SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } SCTP_INP_INFO_RUNLOCK(); } /* * start a new iterator * iterates through all endpoints and associations based on the pcb_state * flags and asoc_state. "af" (mandatory) is executed for all matching * assocs and "ef" (optional) is executed when the iterator completes. * "inpf" (optional) is executed for each new endpoint as it is being * iterated through. inpe (optional) is called when the inp completes * its way through all the stcbs. */ int sctp_initiate_iterator(inp_func inpf, asoc_func af, inp_func inpe, uint32_t pcb_state, uint32_t pcb_features, uint32_t asoc_state, void *argp, uint32_t argi, end_func ef, struct sctp_inpcb *s_inp, uint8_t chunk_output_off) { struct sctp_iterator *it = NULL; if (af == NULL) { return (-1); } SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator), SCTP_M_ITER); if (it == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM); return (ENOMEM); } memset(it, 0, sizeof(*it)); it->function_assoc = af; it->function_inp = inpf; if (inpf) it->done_current_ep = 0; else it->done_current_ep = 1; it->function_atend = ef; it->pointer = argp; it->val = argi; it->pcb_flags = pcb_state; it->pcb_features = pcb_features; it->asoc_state = asoc_state; it->function_inp_end = inpe; it->no_chunk_output = chunk_output_off; if (s_inp) { it->inp = s_inp; it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP; } else { SCTP_INP_INFO_RLOCK(); it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead)); SCTP_INP_INFO_RUNLOCK(); it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP; } SCTP_IPI_ITERATOR_WQ_LOCK(); if (it->inp) { SCTP_INP_INCR_REF(it->inp); } TAILQ_INSERT_TAIL(&SCTP_BASE_INFO(iteratorhead), it, sctp_nxt_itr); #if defined(SCTP_USE_THREAD_BASED_ITERATOR) if (SCTP_BASE_INFO(iterator_running) == 0) { sctp_wakeup_iterator(); } SCTP_IPI_ITERATOR_WQ_UNLOCK(); #else if (it->inp) SCTP_INP_DECR_REF(it->inp); SCTP_IPI_ITERATOR_WQ_UNLOCK(); /* Init the timer */ SCTP_OS_TIMER_INIT(&it->tmr.timer); /* add to the list of all iterators */ sctp_timer_start(SCTP_TIMER_TYPE_ITERATOR, (struct sctp_inpcb *)it, NULL, NULL); #endif /* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */ return (0); } Index: head/sys/netinet/sctp_pcb.h =================================================================== --- head/sys/netinet/sctp_pcb.h (revision 181053) +++ head/sys/netinet/sctp_pcb.h (revision 181054) @@ -1,618 +1,620 @@ /*- * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_pcb.h,v 1.21 2005/07/16 01:18:47 suz Exp $ */ #include __FBSDID("$FreeBSD$"); #ifndef __sctp_pcb_h__ #define __sctp_pcb_h__ #include #include #include #include LIST_HEAD(sctppcbhead, sctp_inpcb); LIST_HEAD(sctpasochead, sctp_tcb); LIST_HEAD(sctpladdr, sctp_laddr); LIST_HEAD(sctpvtaghead, sctp_tagblock); LIST_HEAD(sctp_vrflist, sctp_vrf); LIST_HEAD(sctp_ifnlist, sctp_ifn); LIST_HEAD(sctp_ifalist, sctp_ifa); TAILQ_HEAD(sctp_readhead, sctp_queued_to_read); TAILQ_HEAD(sctp_streamhead, sctp_stream_queue_pending); #include #include #define SCTP_PCBHASH_ALLADDR(port, mask) (port & mask) #define SCTP_PCBHASH_ASOC(tag, mask) (tag & mask) struct sctp_vrf { LIST_ENTRY(sctp_vrf) next_vrf; struct sctp_ifalist *vrf_addr_hash; struct sctp_ifnlist ifnlist; uint32_t vrf_id; uint32_t tbl_id_v4; /* default v4 table id */ uint32_t tbl_id_v6; /* default v6 table id */ uint32_t total_ifa_count; u_long vrf_addr_hashmark; uint32_t refcount; }; struct sctp_ifn { struct sctp_ifalist ifalist; struct sctp_vrf *vrf; LIST_ENTRY(sctp_ifn) next_ifn; LIST_ENTRY(sctp_ifn) next_bucket; void *ifn_p; /* never access without appropriate lock */ uint32_t ifn_mtu; uint32_t ifn_type; uint32_t ifn_index; /* shorthand way to look at ifn for reference */ uint32_t refcount; /* number of reference held should be >= * ifa_count */ uint32_t ifa_count; /* IFA's we hold (in our list - ifalist) */ uint32_t num_v6; /* number of v6 addresses */ uint32_t num_v4; /* number of v4 addresses */ uint32_t registered_af; /* registered address family for i/f events */ char ifn_name[SCTP_IFNAMSIZ]; }; /* SCTP local IFA flags */ #define SCTP_ADDR_VALID 0x00000001 /* its up and active */ #define SCTP_BEING_DELETED 0x00000002 /* being deleted, when * refcount = 0. Note that it * is pulled from the ifn list * and ifa_p is nulled right * away but it cannot be freed * until the last *net * pointing to it is deleted. */ #define SCTP_ADDR_DEFER_USE 0x00000004 /* Hold off using this one */ #define SCTP_ADDR_IFA_UNUSEABLE 0x00000008 struct sctp_ifa { LIST_ENTRY(sctp_ifa) next_ifa; LIST_ENTRY(sctp_ifa) next_bucket; struct sctp_ifn *ifn_p; /* back pointer to parent ifn */ void *ifa; /* pointer to ifa, needed for flag update for * that we MUST lock appropriate locks. This * is for V6. */ union sctp_sockstore address; uint32_t refcount; /* number of folks refering to this */ uint32_t flags; uint32_t localifa_flags; uint32_t vrf_id; /* vrf_id of this addr (for deleting) */ uint8_t src_is_loop; uint8_t src_is_priv; uint8_t src_is_glob; uint8_t resv; }; struct sctp_laddr { LIST_ENTRY(sctp_laddr) sctp_nxt_addr; /* next in list */ struct sctp_ifa *ifa; uint32_t action; /* Used during asconf and adding if no-zero * src-addr selection will not consider this * address. */ struct timeval start_time; /* time when this address was created */ }; struct sctp_block_entry { int error; }; struct sctp_timewait { uint32_t tv_sec_at_expire; /* the seconds from boot to expire */ uint32_t v_tag; /* the vtag that can not be reused */ }; struct sctp_tagblock { LIST_ENTRY(sctp_tagblock) sctp_nxt_tagblock; struct sctp_timewait vtag_block[SCTP_NUMBER_IN_VTAG_BLOCK]; }; struct sctp_epinfo { struct sctpasochead *sctp_asochash; u_long hashasocmark; struct sctppcbhead *sctp_ephash; u_long hashmark; struct sctpasochead *sctp_restarthash; u_long hashrestartmark; /*- * The TCP model represents a substantial overhead in that we get an * additional hash table to keep explicit connections in. The * listening TCP endpoint will exist in the usual ephash above and * accept only INIT's. It will be incapable of sending off an INIT. * When a dg arrives we must look in the normal ephash. If we find a * TCP endpoint that will tell us to go to the specific endpoint * hash and re-hash to find the right assoc/socket. If we find a UDP * model socket we then must complete the lookup. If this fails, * i.e. no association can be found then we must continue to see if * a sctp_peeloff()'d socket is in the tcpephash (a spun off socket * acts like a TCP model connected socket). */ struct sctppcbhead *sctp_tcpephash; u_long hashtcpmark; uint32_t hashtblsize; struct sctp_vrflist *sctp_vrfhash; u_long hashvrfmark; struct sctp_ifnlist *vrf_ifn_hash; u_long vrf_ifn_hashmark; struct sctppcbhead listhead; struct sctpladdr addr_wq; struct sctpiterators iteratorhead; int threads_must_exit; /* ep zone info */ sctp_zone_t ipi_zone_ep; sctp_zone_t ipi_zone_asoc; sctp_zone_t ipi_zone_laddr; sctp_zone_t ipi_zone_net; sctp_zone_t ipi_zone_chunk; sctp_zone_t ipi_zone_readq; sctp_zone_t ipi_zone_strmoq; sctp_zone_t ipi_zone_asconf; sctp_zone_t ipi_zone_asconf_ack; struct rwlock ipi_ep_mtx; struct mtx it_mtx; struct mtx ipi_iterator_wq_mtx; struct rwlock ipi_addr_mtx; struct mtx ipi_pktlog_mtx; uint32_t ipi_count_ep; /* assoc/tcb zone info */ uint32_t ipi_count_asoc; /* local addrlist zone info */ uint32_t ipi_count_laddr; /* remote addrlist zone info */ uint32_t ipi_count_raddr; /* chunk structure list for output */ uint32_t ipi_count_chunk; /* socket queue zone info */ uint32_t ipi_count_readq; /* socket queue zone info */ uint32_t ipi_count_strmoq; /* Number of vrfs */ uint32_t ipi_count_vrfs; /* Number of ifns */ uint32_t ipi_count_ifns; /* Number of ifas */ uint32_t ipi_count_ifas; /* system wide number of free chunks hanging around */ uint32_t ipi_free_chunks; uint32_t ipi_free_strmoq; struct sctpvtaghead vtag_timewait[SCTP_STACK_VTAG_HASH_SIZE_A]; /* address work queue handling */ #if defined(SCTP_USE_THREAD_BASED_ITERATOR) uint32_t iterator_running; SCTP_PROCESS_STRUCT thread_proc; #endif struct sctp_timer addr_wq_timer; }; struct sctp_base_info { /* * All static structures that anchor the system must be here. */ struct sctp_epinfo sctppcbinfo; struct sctpstat sctpstat; struct sctp_sysctl sctpsysctl; uint8_t first_time; char sctp_pcb_initialized; #if defined(SCTP_PACKET_LOGGING) int packet_log_writers; int packet_log_end; uint8_t packet_log_buffer[SCTP_PACKET_LOG_SIZE]; #endif }; /*- * Here we have all the relevant information for each SCTP entity created. We * will need to modify this as approprate. We also need to figure out how to * access /dev/random. */ struct sctp_pcb { unsigned int time_of_secret_change; /* number of seconds from * timeval.tv_sec */ uint32_t secret_key[SCTP_HOW_MANY_SECRETS][SCTP_NUMBER_OF_SECRETS]; unsigned int size_of_a_cookie; unsigned int sctp_timeoutticks[SCTP_NUM_TMRS]; unsigned int sctp_minrto; unsigned int sctp_maxrto; unsigned int initial_rto; int initial_init_rto_max; unsigned int sctp_sack_freq; uint32_t sctp_sws_sender; uint32_t sctp_sws_receiver; uint32_t sctp_default_cc_module; /* authentication related fields */ struct sctp_keyhead shared_keys; sctp_auth_chklist_t *local_auth_chunks; sctp_hmaclist_t *local_hmacs; uint16_t default_keyid; /* various thresholds */ /* Max times I will init at a guy */ uint16_t max_init_times; /* Max times I will send before we consider someone dead */ uint16_t max_send_times; uint16_t def_net_failure; /* number of streams to pre-open on a association */ uint16_t pre_open_stream_count; uint16_t max_open_streams_intome; /* random number generator */ uint32_t random_counter; uint8_t random_numbers[SCTP_SIGNATURE_ALOC_SIZE]; uint8_t random_store[SCTP_SIGNATURE_ALOC_SIZE]; /* * This timer is kept running per endpoint. When it fires it will * change the secret key. The default is once a hour */ struct sctp_timer signature_change; /* Zero copy full buffer timer */ struct sctp_timer zero_copy_timer; /* Zero copy app to transport (sendq) read repulse timer */ struct sctp_timer zero_copy_sendq_timer; uint32_t def_cookie_life; /* defaults to 0 */ int auto_close_time; uint32_t initial_sequence_debug; uint32_t adaptation_layer_indicator; uint32_t store_at; uint8_t max_burst; char current_secret_number; char last_secret_number; }; #ifndef SCTP_ALIGNMENT #define SCTP_ALIGNMENT 32 #endif #ifndef SCTP_ALIGNM1 #define SCTP_ALIGNM1 (SCTP_ALIGNMENT-1) #endif #define sctp_lport ip_inp.inp.inp_lport struct sctp_pcbtsn_rlog { uint32_t vtag; uint16_t strm; uint16_t seq; uint16_t sz; uint16_t flgs; }; #define SCTP_READ_LOG_SIZE 135 /* we choose the number to make a pcb a page */ struct sctp_inpcb { /*- * put an inpcb in front of it all, kind of a waste but we need to * for compatability with all the other stuff. */ union { struct inpcb inp; char align[(sizeof(struct in6pcb) + SCTP_ALIGNM1) & ~SCTP_ALIGNM1]; } ip_inp; /* Socket buffer lock protects read_queue and of course sb_cc */ struct sctp_readhead read_queue; LIST_ENTRY(sctp_inpcb) sctp_list; /* lists all endpoints */ /* hash of all endpoints for model */ LIST_ENTRY(sctp_inpcb) sctp_hash; /* count of local addresses bound, 0 if bound all */ int laddr_count; /* list of addrs in use by the EP, NULL if bound-all */ struct sctpladdr sctp_addr_list; /* * used for source address selection rotation when we are subset * bound */ struct sctp_laddr *next_addr_touse; /* back pointer to our socket */ struct socket *sctp_socket; uint32_t sctp_flags; /* INP state flag set */ uint32_t sctp_features; /* Feature flags */ uint32_t sctp_mobility_features; /* Mobility Feature flags */ struct sctp_pcb sctp_ep;/* SCTP ep data */ /* head of the hash of all associations */ struct sctpasochead *sctp_tcbhash; u_long sctp_hashmark; /* head of the list of all associations */ struct sctpasochead sctp_asoc_list; #ifdef SCTP_TRACK_FREED_ASOCS struct sctpasochead sctp_asoc_free_list; #endif struct sctp_iterator *inp_starting_point_for_iterator; uint32_t sctp_frag_point; uint32_t partial_delivery_point; uint32_t sctp_context; struct sctp_nonpad_sndrcvinfo def_send; /*- * These three are here for the sosend_dgram * (pkt, pkt_last and control). * routine. However, I don't think anyone in * the current FreeBSD kernel calls this. So * they are candidates with sctp_sendm for * de-supporting. */ struct mbuf *pkt, *pkt_last; struct mbuf *control; struct mtx inp_mtx; struct mtx inp_create_mtx; struct mtx inp_rdata_mtx; int32_t refcount; uint32_t def_vrf_id; uint32_t total_sends; uint32_t total_recvs; uint32_t last_abort_code; uint32_t total_nospaces; #ifdef SCTP_ASOCLOG_OF_TSNS struct sctp_pcbtsn_rlog readlog[SCTP_READ_LOG_SIZE]; uint32_t readlog_index; #endif }; struct sctp_tcb { struct socket *sctp_socket; /* back pointer to socket */ struct sctp_inpcb *sctp_ep; /* back pointer to ep */ LIST_ENTRY(sctp_tcb) sctp_tcbhash; /* next link in hash * table */ LIST_ENTRY(sctp_tcb) sctp_tcblist; /* list of all of the * TCB's */ LIST_ENTRY(sctp_tcb) sctp_tcbrestarhash; /* next link in restart * hash table */ LIST_ENTRY(sctp_tcb) sctp_asocs; /* vtag hash list */ struct sctp_block_entry *block_entry; /* pointer locked by socket * send buffer */ struct sctp_association asoc; /*- * freed_by_sorcv_sincelast is protected by the sockbuf_lock NOT the * tcb_lock. Its special in this way to help avoid extra mutex calls * in the reading of data. */ uint32_t freed_by_sorcv_sincelast; uint32_t total_sends; uint32_t total_recvs; int freed_from_where; uint16_t rport; /* remote port in network format */ uint16_t resv; struct mtx tcb_mtx; struct mtx tcb_send_mtx; }; #include /* TODO where to put non-_KERNEL things for __Userspace__? */ #if defined(_KERNEL) || defined(__Userspace__) /* Attention Julian, this is the extern that * goes with the base info. sctp_pcb.c has * the real definition. */ extern struct sctp_base_info system_base_info; #ifdef INET6 int SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b); #endif void sctp_fill_pcbinfo(struct sctp_pcbinfo *); struct sctp_ifn * sctp_find_ifn(void *ifn, uint32_t ifn_index); struct sctp_vrf *sctp_allocate_vrf(int vrfid); struct sctp_vrf *sctp_find_vrf(uint32_t vrfid); void sctp_free_vrf(struct sctp_vrf *vrf); /*- * Change address state, can be used if * O/S supports telling transports about * changes to IFA/IFN's (link layer triggers). * If a ifn goes down, we will do src-addr-selection * and NOT use that, as a source address. This does * not stop the routing system from routing out * that interface, but we won't put it as a source. */ void sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr, const char *if_name, uint32_t ifn_index); void sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr, const char *if_name, uint32_t ifn_index); struct sctp_ifa * sctp_add_addr_to_vrf(uint32_t vrfid, void *ifn, uint32_t ifn_index, uint32_t ifn_type, const char *if_name, void *ifa, struct sockaddr *addr, uint32_t ifa_flags, int dynamic_add); void sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu); void sctp_free_ifn(struct sctp_ifn *sctp_ifnp); void sctp_free_ifa(struct sctp_ifa *sctp_ifap); void sctp_del_addr_from_vrf(uint32_t vrfid, struct sockaddr *addr, uint32_t ifn_index, const char *if_name); struct sctp_nets *sctp_findnet(struct sctp_tcb *, struct sockaddr *); struct sctp_inpcb *sctp_pcb_findep(struct sockaddr *, int, int, uint32_t); int sctp_inpcb_bind(struct socket *, struct sockaddr *, struct sctp_ifa *, struct thread *); struct sctp_tcb * sctp_findassociation_addr(struct mbuf *, int, int, struct sctphdr *, struct sctp_chunkhdr *, struct sctp_inpcb **, struct sctp_nets **, uint32_t vrf_id); struct sctp_tcb * sctp_findassociation_addr_sa(struct sockaddr *, struct sockaddr *, struct sctp_inpcb **, struct sctp_nets **, int, uint32_t); void sctp_move_pcb_and_assoc(struct sctp_inpcb *, struct sctp_inpcb *, struct sctp_tcb *); /*- * For this call ep_addr, the to is the destination endpoint address of the * peer (relative to outbound). The from field is only used if the TCP model * is enabled and helps distingush amongst the subset bound (non-boundall). * The TCP model MAY change the actual ep field, this is why it is passed. */ struct sctp_tcb * sctp_findassociation_ep_addr(struct sctp_inpcb **, struct sockaddr *, struct sctp_nets **, struct sockaddr *, struct sctp_tcb *); struct sctp_tcb * sctp_findassociation_ep_asocid(struct sctp_inpcb *, sctp_assoc_t, int); struct sctp_tcb * sctp_findassociation_ep_asconf(struct mbuf *, int, int, struct sctphdr *, struct sctp_inpcb **, struct sctp_nets **); int sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id); int sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id); void sctp_inpcb_free(struct sctp_inpcb *, int, int); struct sctp_tcb * sctp_aloc_assoc(struct sctp_inpcb *, struct sockaddr *, int, int *, uint32_t, uint32_t, struct thread *); int sctp_free_assoc(struct sctp_inpcb *, struct sctp_tcb *, int, int); void sctp_delete_from_timewait(uint32_t); int sctp_is_in_timewait(uint32_t tag); void sctp_add_vtag_to_timewait(uint32_t, uint32_t); void sctp_add_local_addr_ep(struct sctp_inpcb *, struct sctp_ifa *, uint32_t); int sctp_insert_laddr(struct sctpladdr *, struct sctp_ifa *, uint32_t); void sctp_remove_laddr(struct sctp_laddr *); void sctp_del_local_addr_ep(struct sctp_inpcb *, struct sctp_ifa *); int sctp_add_remote_addr(struct sctp_tcb *, struct sockaddr *, int, int); void sctp_remove_net(struct sctp_tcb *, struct sctp_nets *); int sctp_del_remote_addr(struct sctp_tcb *, struct sockaddr *); void sctp_pcb_init(void); void sctp_pcb_finish(void); void sctp_add_local_addr_restricted(struct sctp_tcb *, struct sctp_ifa *); void sctp_del_local_addr_restricted(struct sctp_tcb *, struct sctp_ifa *); int sctp_load_addresses_from_init(struct sctp_tcb *, struct mbuf *, int, int, int, struct sctphdr *, struct sockaddr *); int sctp_set_primary_addr(struct sctp_tcb *, struct sockaddr *, struct sctp_nets *); int sctp_is_vtag_good(struct sctp_inpcb *, uint32_t, struct timeval *, int); /* void sctp_drain(void); */ int sctp_destination_is_reachable(struct sctp_tcb *, struct sockaddr *); +int sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp); + /*- * Null in last arg inpcb indicate run on ALL ep's. Specific inp in last arg * indicates run on ONLY assoc's of the specified endpoint. */ int sctp_initiate_iterator(inp_func inpf, asoc_func af, inp_func inpe, uint32_t, uint32_t, uint32_t, void *, uint32_t, end_func ef, struct sctp_inpcb *, uint8_t co_off); #endif /* _KERNEL */ #endif /* !__sctp_pcb_h__ */ Index: head/sys/netinet/sctp_usrreq.c =================================================================== --- head/sys/netinet/sctp_usrreq.c (revision 181053) +++ head/sys/netinet/sctp_usrreq.c (revision 181054) @@ -1,4521 +1,4629 @@ /*- * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_usrreq.c,v 1.48 2005/03/07 23:26:08 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #if defined(INET6) #endif #include #include #include #include #include #include #include #include #include #include void sctp_init(void) { u_long sb_max_adj; bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat)); /* Initialize and modify the sysctled variables */ sctp_init_sysctls(); if ((nmbclusters / 8) > SCTP_ASOC_MAX_CHUNKS_ON_QUEUE) SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue) = (nmbclusters / 8); /* * Allow a user to take no more than 1/2 the number of clusters or * the SB_MAX whichever is smaller for the send window. */ sb_max_adj = (u_long)((u_quad_t) (SB_MAX) * MCLBYTES / (MSIZE + MCLBYTES)); SCTP_BASE_SYSCTL(sctp_sendspace) = min(sb_max_adj, (((uint32_t) nmbclusters / 2) * SCTP_DEFAULT_MAXSEGMENT)); /* * Now for the recv window, should we take the same amount? or * should I do 1/2 the SB_MAX instead in the SB_MAX min above. For * now I will just copy. */ SCTP_BASE_SYSCTL(sctp_recvspace) = SCTP_BASE_SYSCTL(sctp_sendspace); SCTP_BASE_VAR(first_time) = 0; SCTP_BASE_VAR(sctp_pcb_initialized) = 0; sctp_pcb_init(); #if defined(SCTP_PACKET_LOGGING) SCTP_BASE_VAR(packet_log_writers) = 0; SCTP_BASE_VAR(packet_log_end) = 0; bzero(&SCTP_BASE_VAR(packet_log_buffer), SCTP_PACKET_LOG_SIZE); #endif } void sctp_finish(void) { sctp_pcb_finish(); } void sctp_pathmtu_adjustment(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, uint16_t nxtsz) { struct sctp_tmit_chunk *chk; /* Adjust that too */ stcb->asoc.smallest_mtu = nxtsz; /* now off to subtract IP_DF flag if needed */ #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("sctp_pathmtu_adjust called inp:%p stcb:%p net:%p nxtsz:%d\n", inp, stcb, net, nxtsz); #endif TAILQ_FOREACH(chk, &stcb->asoc.send_queue, sctp_next) { if ((chk->send_size + IP_HDR_SIZE) > nxtsz) { chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; } } TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { if ((chk->send_size + IP_HDR_SIZE) > nxtsz) { /* * For this guy we also mark for immediate resend * since we sent to big of chunk */ chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; if (chk->sent != SCTP_DATAGRAM_RESEND) { sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); } chk->sent = SCTP_DATAGRAM_RESEND; chk->rec.data.doing_fast_retransmit = 0; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_PMTU, chk->whoTo->flight_size, chk->book_size, (uintptr_t) chk->whoTo, chk->rec.data.TSN_seq); } /* Clear any time so NO RTT is being done */ chk->do_rtt = 0; sctp_flight_size_decrease(chk); sctp_total_flight_decrease(stcb, chk); } } } static void sctp_notify_mbuf(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, struct ip *ip, struct sctphdr *sh) { struct icmp *icmph; int totsz, tmr_stopped = 0; uint16_t nxtsz; /* protection */ if ((inp == NULL) || (stcb == NULL) || (net == NULL) || (ip == NULL) || (sh == NULL)) { if (stcb != NULL) { SCTP_TCB_UNLOCK(stcb); } return; } /* First job is to verify the vtag matches what I would send */ if (ntohl(sh->v_tag) != (stcb->asoc.peer_vtag)) { SCTP_TCB_UNLOCK(stcb); return; } icmph = (struct icmp *)((caddr_t)ip - (sizeof(struct icmp) - sizeof(struct ip))); if (icmph->icmp_type != ICMP_UNREACH) { /* We only care about unreachable */ SCTP_TCB_UNLOCK(stcb); return; } if (icmph->icmp_code != ICMP_UNREACH_NEEDFRAG) { /* not a unreachable message due to frag. */ SCTP_TCB_UNLOCK(stcb); return; } totsz = ip->ip_len; nxtsz = ntohs(icmph->icmp_nextmtu); if (nxtsz == 0) { /* * old type router that does not tell us what the next size * mtu is. Rats we will have to guess (in a educated fashion * of course) */ nxtsz = find_next_best_mtu(totsz); } /* Stop any PMTU timer */ if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { tmr_stopped = 1; sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_1); } /* Adjust destination size limit */ if (net->mtu > nxtsz) { net->mtu = nxtsz; } /* now what about the ep? */ if (stcb->asoc.smallest_mtu > nxtsz) { #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("notify_mbuf (ICMP) calls sctp_pathmtu_adjust mtu:%d\n", nxtsz); #endif sctp_pathmtu_adjustment(inp, stcb, net, nxtsz); } if (tmr_stopped) sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net); SCTP_TCB_UNLOCK(stcb); } void sctp_notify(struct sctp_inpcb *inp, struct ip *ip, struct sctphdr *sh, struct sockaddr *to, struct sctp_tcb *stcb, struct sctp_nets *net) { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif /* protection */ int reason; struct icmp *icmph; if ((inp == NULL) || (stcb == NULL) || (net == NULL) || (sh == NULL) || (to == NULL)) { if (stcb) SCTP_TCB_UNLOCK(stcb); return; } /* First job is to verify the vtag matches what I would send */ if (ntohl(sh->v_tag) != (stcb->asoc.peer_vtag)) { SCTP_TCB_UNLOCK(stcb); return; } icmph = (struct icmp *)((caddr_t)ip - (sizeof(struct icmp) - sizeof(struct ip))); if (icmph->icmp_type != ICMP_UNREACH) { /* We only care about unreachable */ SCTP_TCB_UNLOCK(stcb); return; } if ((icmph->icmp_code == ICMP_UNREACH_NET) || (icmph->icmp_code == ICMP_UNREACH_HOST) || (icmph->icmp_code == ICMP_UNREACH_NET_UNKNOWN) || (icmph->icmp_code == ICMP_UNREACH_HOST_UNKNOWN) || (icmph->icmp_code == ICMP_UNREACH_ISOLATED) || (icmph->icmp_code == ICMP_UNREACH_NET_PROHIB) || (icmph->icmp_code == ICMP_UNREACH_HOST_PROHIB) || (icmph->icmp_code == ICMP_UNREACH_FILTER_PROHIB)) { /* * Hmm reachablity problems we must examine closely. If its * not reachable, we may have lost a network. Or if there is * NO protocol at the other end named SCTP. well we consider * it a OOTB abort. */ if (net->dest_state & SCTP_ADDR_REACHABLE) { /* Ok that destination is NOT reachable */ SCTP_PRINTF("ICMP (thresh %d/%d) takes interface %p down\n", net->error_count, net->failure_threshold, net); net->dest_state &= ~SCTP_ADDR_REACHABLE; net->dest_state |= SCTP_ADDR_NOT_REACHABLE; /* * JRS 5/14/07 - If a destination is unreachable, * the PF bit is turned off. This allows an * unambiguous use of the PF bit for destinations * that are reachable but potentially failed. If the * destination is set to the unreachable state, also * set the destination to the PF state. */ /* * Add debug message here if destination is not in * PF state. */ /* Stop any running T3 timers here? */ if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf)) { net->dest_state &= ~SCTP_ADDR_PF; SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from PF to unreachable.\n", net); } net->error_count = net->failure_threshold + 1; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, stcb, SCTP_FAILED_THRESHOLD, (void *)net, SCTP_SO_NOT_LOCKED); } SCTP_TCB_UNLOCK(stcb); } else if ((icmph->icmp_code == ICMP_UNREACH_PROTOCOL) || (icmph->icmp_code == ICMP_UNREACH_PORT)) { /* * Here the peer is either playing tricks on us, including * an address that belongs to someone who does not support * SCTP OR was a userland implementation that shutdown and * now is dead. In either case treat it like a OOTB abort * with no TCB */ reason = SCTP_PEER_FAULTY; sctp_abort_notification(stcb, reason, SCTP_SO_NOT_LOCKED); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(inp); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_2); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); /* SCTP_TCB_UNLOCK(stcb); MT: I think this is not needed. */ #endif /* no need to unlock here, since the TCB is gone */ } else { SCTP_TCB_UNLOCK(stcb); } } void sctp_ctlinput(cmd, sa, vip) int cmd; struct sockaddr *sa; void *vip; { struct ip *ip = vip; struct sctphdr *sh; uint32_t vrf_id; /* FIX, for non-bsd is this right? */ vrf_id = SCTP_DEFAULT_VRFID; if (sa->sa_family != AF_INET || ((struct sockaddr_in *)sa)->sin_addr.s_addr == INADDR_ANY) { return; } if (PRC_IS_REDIRECT(cmd)) { ip = 0; } else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) { return; } if (ip) { struct sctp_inpcb *inp = NULL; struct sctp_tcb *stcb = NULL; struct sctp_nets *net = NULL; struct sockaddr_in to, from; sh = (struct sctphdr *)((caddr_t)ip + (ip->ip_hl << 2)); bzero(&to, sizeof(to)); bzero(&from, sizeof(from)); from.sin_family = to.sin_family = AF_INET; from.sin_len = to.sin_len = sizeof(to); from.sin_port = sh->src_port; from.sin_addr = ip->ip_src; to.sin_port = sh->dest_port; to.sin_addr = ip->ip_dst; /* * 'to' holds the dest of the packet that failed to be sent. * 'from' holds our local endpoint address. Thus we reverse * the to and the from in the lookup. */ stcb = sctp_findassociation_addr_sa((struct sockaddr *)&from, (struct sockaddr *)&to, &inp, &net, 1, vrf_id); if (stcb != NULL && inp && (inp->sctp_socket != NULL)) { if (cmd != PRC_MSGSIZE) { sctp_notify(inp, ip, sh, (struct sockaddr *)&to, stcb, net); } else { /* handle possible ICMP size messages */ sctp_notify_mbuf(inp, stcb, net, ip, sh); } } else { if ((stcb == NULL) && (inp != NULL)) { /* reduce ref-count */ SCTP_INP_WLOCK(inp); SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); } } } return; } static int sctp_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in addrs[2]; struct sctp_inpcb *inp; struct sctp_nets *net; struct sctp_tcb *stcb; int error; uint32_t vrf_id; /* FIX, for non-bsd is this right? */ vrf_id = SCTP_DEFAULT_VRFID; error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); stcb = sctp_findassociation_addr_sa(sintosa(&addrs[0]), sintosa(&addrs[1]), &inp, &net, 1, vrf_id); if (stcb == NULL || inp == NULL || inp->sctp_socket == NULL) { if ((inp != NULL) && (stcb == NULL)) { /* reduce ref-count */ SCTP_INP_WLOCK(inp); SCTP_INP_DECR_REF(inp); goto cred_can_cont; } SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; goto out; } SCTP_TCB_UNLOCK(stcb); /* * We use the write lock here, only since in the error leg we need * it. If we used RLOCK, then we would have to * wlock/decr/unlock/rlock. Which in theory could create a hole. * Better to use higher wlock. */ SCTP_INP_WLOCK(inp); cred_can_cont: error = cr_canseesocket(req->td->td_ucred, inp->sctp_socket); if (error) { SCTP_INP_WUNLOCK(inp); goto out; } cru2x(inp->sctp_socket->so_cred, &xuc); SCTP_INP_WUNLOCK(inp); error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); out: return (error); } SYSCTL_PROC(_net_inet_sctp, OID_AUTO, getcred, CTLTYPE_OPAQUE | CTLFLAG_RW, 0, 0, sctp_getcred, "S,ucred", "Get the ucred of a SCTP connection"); static void sctp_abort(struct socket *so) { struct sctp_inpcb *inp; uint32_t flags; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == 0) { return; } sctp_must_try_again: flags = inp->sctp_flags; #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 17); #endif if (((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && (atomic_cmpset_int(&inp->sctp_flags, flags, (flags | SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_CLOSE_IP)))) { #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 16); #endif sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, SCTP_CALLED_AFTER_CMPSET_OFCLOSE); SOCK_LOCK(so); SCTP_SB_CLEAR(so->so_snd); /* * same for the rcv ones, they are only here for the * accounting/select. */ SCTP_SB_CLEAR(so->so_rcv); /* Now null out the reference, we are completely detached. */ so->so_pcb = NULL; SOCK_UNLOCK(so); } else { flags = inp->sctp_flags; if ((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) { goto sctp_must_try_again; } } return; } static int sctp_attach(struct socket *so, int proto, struct thread *p) { struct sctp_inpcb *inp; struct inpcb *ip_inp; int error; uint32_t vrf_id = SCTP_DEFAULT_VRFID; #ifdef IPSEC uint32_t flags; #endif inp = (struct sctp_inpcb *)so->so_pcb; if (inp != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return EINVAL; } error = SCTP_SORESERVE(so, SCTP_BASE_SYSCTL(sctp_sendspace), SCTP_BASE_SYSCTL(sctp_recvspace)); if (error) { return error; } error = sctp_inpcb_alloc(so, vrf_id); if (error) { return error; } inp = (struct sctp_inpcb *)so->so_pcb; SCTP_INP_WLOCK(inp); inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUND_V6; /* I'm not v6! */ ip_inp = &inp->ip_inp.inp; ip_inp->inp_vflag |= INP_IPV4; ip_inp->inp_ip_ttl = MODULE_GLOBAL(MOD_INET, ip_defttl); #ifdef IPSEC error = ipsec_init_policy(so, &ip_inp->inp_sp); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 17); #endif if (error != 0) { flags = inp->sctp_flags; if (((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && (atomic_cmpset_int(&inp->sctp_flags, flags, (flags | SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_CLOSE_IP)))) { #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 15); #endif SCTP_INP_WUNLOCK(inp); sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, SCTP_CALLED_AFTER_CMPSET_OFCLOSE); } else { SCTP_INP_WUNLOCK(inp); } return error; } #endif /* IPSEC */ SCTP_INP_WUNLOCK(inp); return 0; } static int sctp_bind(struct socket *so, struct sockaddr *addr, struct thread *p) { struct sctp_inpcb *inp = NULL; int error; #ifdef INET6 if (addr && addr->sa_family != AF_INET) { /* must be a v4 address! */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return EINVAL; } #endif /* INET6 */ if (addr && (addr->sa_len != sizeof(struct sockaddr_in))) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return EINVAL; } inp = (struct sctp_inpcb *)so->so_pcb; if (inp == 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return EINVAL; } error = sctp_inpcb_bind(so, addr, NULL, p); return error; } void sctp_close(struct socket *so) { struct sctp_inpcb *inp; uint32_t flags; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == 0) return; /* * Inform all the lower layer assoc that we are done. */ sctp_must_try_again: flags = inp->sctp_flags; #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 17); #endif if (((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && (atomic_cmpset_int(&inp->sctp_flags, flags, (flags | SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_CLOSE_IP)))) { if (((so->so_options & SO_LINGER) && (so->so_linger == 0)) || (so->so_rcv.sb_cc > 0)) { #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 13); #endif sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, SCTP_CALLED_AFTER_CMPSET_OFCLOSE); } else { #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 14); #endif sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE, SCTP_CALLED_AFTER_CMPSET_OFCLOSE); } /* * The socket is now detached, no matter what the state of * the SCTP association. */ SOCK_LOCK(so); SCTP_SB_CLEAR(so->so_snd); /* * same for the rcv ones, they are only here for the * accounting/select. */ SCTP_SB_CLEAR(so->so_rcv); /* Now null out the reference, we are completely detached. */ so->so_pcb = NULL; SOCK_UNLOCK(so); } else { flags = inp->sctp_flags; if ((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) { goto sctp_must_try_again; } } return; } int sctp_sendm(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *p); int sctp_sendm(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *p) { struct sctp_inpcb *inp; int error; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == 0) { if (control) { sctp_m_freem(control); control = NULL; } SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); sctp_m_freem(m); return EINVAL; } /* Got to have an to address if we are NOT a connected socket */ if ((addr == NULL) && ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) || (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) ) { goto connected_type; } else if (addr == NULL) { SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EDESTADDRREQ); error = EDESTADDRREQ; sctp_m_freem(m); if (control) { sctp_m_freem(control); control = NULL; } return (error); } #ifdef INET6 if (addr->sa_family != AF_INET) { /* must be a v4 address! */ SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EDESTADDRREQ); sctp_m_freem(m); if (control) { sctp_m_freem(control); control = NULL; } error = EDESTADDRREQ; return EDESTADDRREQ; } #endif /* INET6 */ connected_type: /* now what about control */ if (control) { if (inp->control) { SCTP_PRINTF("huh? control set?\n"); sctp_m_freem(inp->control); inp->control = NULL; } inp->control = control; } /* Place the data */ if (inp->pkt) { SCTP_BUF_NEXT(inp->pkt_last) = m; inp->pkt_last = m; } else { inp->pkt_last = inp->pkt = m; } if ( /* FreeBSD uses a flag passed */ ((flags & PRUS_MORETOCOME) == 0) ) { /* * note with the current version this code will only be used * by OpenBSD-- NetBSD, FreeBSD, and MacOS have methods for * re-defining sosend to use the sctp_sosend. One can * optionally switch back to this code (by changing back the * definitions) but this is not advisable. This code is used * by FreeBSD when sending a file with sendfile() though. */ int ret; ret = sctp_output(inp, inp->pkt, addr, inp->control, p, flags); inp->pkt = NULL; inp->control = NULL; return (ret); } else { return (0); } } int sctp_disconnect(struct socket *so) { struct sctp_inpcb *inp; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); return (ENOTCONN); } SCTP_INP_RLOCK(inp); if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { if (SCTP_LIST_EMPTY(&inp->sctp_asoc_list)) { /* No connection */ SCTP_INP_RUNLOCK(inp); return (0); } else { struct sctp_association *asoc; struct sctp_tcb *stcb; stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } SCTP_TCB_LOCK(stcb); asoc = &stcb->asoc; if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { /* We are about to be freed, out of here */ SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); return (0); } if (((so->so_options & SO_LINGER) && (so->so_linger == 0)) || (so->so_rcv.sb_cc > 0)) { if (SCTP_GET_STATE(asoc) != SCTP_STATE_COOKIE_WAIT) { /* Left with Data unread */ struct mbuf *err; err = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr), 0, M_DONTWAIT, 1, MT_DATA); if (err) { /* * Fill in the user * initiated abort */ struct sctp_paramhdr *ph; ph = mtod(err, struct sctp_paramhdr *); SCTP_BUF_LEN(err) = sizeof(struct sctp_paramhdr); ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(SCTP_BUF_LEN(err)); } #if defined(SCTP_PANIC_ON_ABORT) panic("disconnect does an abort"); #endif sctp_send_abort_tcb(stcb, err, SCTP_SO_LOCKED); SCTP_STAT_INCR_COUNTER32(sctps_aborted); } SCTP_INP_RUNLOCK(inp); if ((SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_3); /* No unlock tcb assoc is gone */ return (0); } if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (asoc->stream_queue_cnt == 0)) { /* there is nothing queued to send, so done */ if (asoc->locked_on_sending) { goto abort_anyway; } if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { /* only send SHUTDOWN 1st time thru */ sctp_stop_timers_for_shutdown(stcb); sctp_send_shutdown(stcb, stcb->asoc.primary_destination); sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_LOCKED); if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, asoc->primary_destination); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } } else { /* * we still got (or just got) data to send, * so set SHUTDOWN_PENDING */ /* * XXX sockets draft says that SCTP_EOF * should be sent with no data. currently, * we will allow user data to be sent first * and move to SHUTDOWN-PENDING */ asoc->state |= SCTP_STATE_SHUTDOWN_PENDING; sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); if (asoc->locked_on_sending) { /* Locked to send out the data */ struct sctp_stream_queue_pending *sp; sp = TAILQ_LAST(&asoc->locked_on_sending->outqueue, sctp_streamhead); if (sp == NULL) { SCTP_PRINTF("Error, sp is NULL, locked on sending is non-null strm:%d\n", asoc->locked_on_sending->stream_no); } else { if ((sp->length == 0) && (sp->msg_is_complete == 0)) asoc->state |= SCTP_STATE_PARTIAL_MSG_LEFT; } } if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { struct mbuf *op_err; abort_anyway: op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { /* * Fill in the user * initiated abort */ struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) + sizeof(uint32_t)); ph = mtod(op_err, struct sctp_paramhdr *); ph->param_type = htons( SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(SCTP_BUF_LEN(op_err)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_USRREQ + SCTP_LOC_4); } #if defined(SCTP_PANIC_ON_ABORT) panic("disconnect does an abort"); #endif stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_USRREQ + SCTP_LOC_4; sctp_send_abort_tcb(stcb, op_err, SCTP_SO_LOCKED); SCTP_STAT_INCR_COUNTER32(sctps_aborted); if ((SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_INP_RUNLOCK(inp); (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_5); return (0); } else { sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED); } } SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); return (0); } /* not reached */ } else { /* UDP model does not support this */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); return EOPNOTSUPP; } } int sctp_flush(struct socket *so, int how) { /* * We will just clear out the values and let subsequent close clear * out the data, if any. Note if the user did a shutdown(SHUT_RD) * they will not be able to read the data, the socket will block * that from happening. */ if ((how == PRU_FLUSH_RD) || (how == PRU_FLUSH_RDWR)) { /* * First make sure the sb will be happy, we don't use these * except maybe the count */ so->so_rcv.sb_cc = 0; so->so_rcv.sb_mbcnt = 0; so->so_rcv.sb_mb = NULL; } if ((how == PRU_FLUSH_WR) || (how == PRU_FLUSH_RDWR)) { /* * First make sure the sb will be happy, we don't use these * except maybe the count */ so->so_snd.sb_cc = 0; so->so_snd.sb_mbcnt = 0; so->so_snd.sb_mb = NULL; } return (0); } int sctp_shutdown(struct socket *so) { struct sctp_inpcb *inp; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return EINVAL; } SCTP_INP_RLOCK(inp); /* For UDP model this is a invalid call */ if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) { /* Restore the flags that the soshutdown took away. */ so->so_rcv.sb_state &= ~SBS_CANTRCVMORE; /* This proc will wakeup for read and do nothing (I hope) */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); return (EOPNOTSUPP); } /* * Ok if we reach here its the TCP model and it is either a SHUT_WR * or SHUT_RDWR. This means we put the shutdown flag against it. */ { struct sctp_tcb *stcb; struct sctp_association *asoc; socantsendmore(so); stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { /* * Ok we hit the case that the shutdown call was * made after an abort or something. Nothing to do * now. */ SCTP_INP_RUNLOCK(inp); return (0); } SCTP_TCB_LOCK(stcb); asoc = &stcb->asoc; if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (asoc->stream_queue_cnt == 0)) { if (asoc->locked_on_sending) { goto abort_anyway; } /* there is nothing queued to send, so I'm done... */ if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) { /* only send SHUTDOWN the first time through */ sctp_stop_timers_for_shutdown(stcb); sctp_send_shutdown(stcb, stcb->asoc.primary_destination); sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_LOCKED); if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, asoc->primary_destination); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); } } else { /* * we still got (or just got) data to send, so set * SHUTDOWN_PENDING */ asoc->state |= SCTP_STATE_SHUTDOWN_PENDING; sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, asoc->primary_destination); if (asoc->locked_on_sending) { /* Locked to send out the data */ struct sctp_stream_queue_pending *sp; sp = TAILQ_LAST(&asoc->locked_on_sending->outqueue, sctp_streamhead); if (sp == NULL) { SCTP_PRINTF("Error, sp is NULL, locked on sending is non-null strm:%d\n", asoc->locked_on_sending->stream_no); } else { if ((sp->length == 0) && (sp->msg_is_complete == 0)) { asoc->state |= SCTP_STATE_PARTIAL_MSG_LEFT; } } } if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { struct mbuf *op_err; abort_anyway: op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)), 0, M_DONTWAIT, 1, MT_DATA); if (op_err) { /* Fill in the user initiated abort */ struct sctp_paramhdr *ph; uint32_t *ippp; SCTP_BUF_LEN(op_err) = sizeof(struct sctp_paramhdr) + sizeof(uint32_t); ph = mtod(op_err, struct sctp_paramhdr *); ph->param_type = htons( SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(SCTP_BUF_LEN(op_err)); ippp = (uint32_t *) (ph + 1); *ippp = htonl(SCTP_FROM_SCTP_USRREQ + SCTP_LOC_6); } #if defined(SCTP_PANIC_ON_ABORT) panic("shutdown does an abort"); #endif stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_USRREQ + SCTP_LOC_6; sctp_abort_an_association(stcb->sctp_ep, stcb, SCTP_RESPONSE_TO_USER_REQ, op_err, SCTP_SO_LOCKED); goto skip_unlock; } else { sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED); } } SCTP_TCB_UNLOCK(stcb); } skip_unlock: SCTP_INP_RUNLOCK(inp); return 0; } /* * copies a "user" presentable address and removes embedded scope, etc. * returns 0 on success, 1 on error */ static uint32_t sctp_fill_user_address(struct sockaddr_storage *ss, struct sockaddr *sa) { #ifdef INET6 struct sockaddr_in6 lsa6; sa = (struct sockaddr *)sctp_recover_scope((struct sockaddr_in6 *)sa, &lsa6); #endif memcpy(ss, sa, sa->sa_len); return (0); } /* * NOTE: assumes addr lock is held */ static size_t sctp_fill_up_addresses_vrf(struct sctp_inpcb *inp, struct sctp_tcb *stcb, size_t limit, struct sockaddr_storage *sas, uint32_t vrf_id) { struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa; int loopback_scope, ipv4_local_scope, local_scope, site_scope; size_t actual; int ipv4_addr_legal, ipv6_addr_legal; struct sctp_vrf *vrf; actual = 0; if (limit <= 0) return (actual); if (stcb) { /* Turn on all the appropriate scope */ loopback_scope = stcb->asoc.loopback_scope; ipv4_local_scope = stcb->asoc.ipv4_local_scope; local_scope = stcb->asoc.local_scope; site_scope = stcb->asoc.site_scope; } else { /* Turn on ALL scope, since we look at the EP */ loopback_scope = ipv4_local_scope = local_scope = site_scope = 1; } ipv4_addr_legal = ipv6_addr_legal = 0; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ipv6_addr_legal = 1; if (SCTP_IPV6_V6ONLY(inp) == 0) { ipv4_addr_legal = 1; } } else { ipv4_addr_legal = 1; } vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { return (0); } if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { if ((loopback_scope == 0) && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { /* Skip loopback if loopback_scope not set */ continue; } LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if (stcb) { /* * For the BOUND-ALL case, the list * associated with a TCB is Always * considered a reverse list.. i.e. * it lists addresses that are NOT * part of the association. If this * is one of those we must skip it. */ if (sctp_is_addr_restricted(stcb, sctp_ifa)) { continue; } } switch (sctp_ifa->address.sa.sa_family) { case AF_INET: if (ipv4_addr_legal) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)&sctp_ifa->address.sa; if (sin->sin_addr.s_addr == 0) { /* * we skip * unspecifed * addresses */ continue; } if ((ipv4_local_scope == 0) && (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { continue; } #ifdef INET6 if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) { in6_sin_2_v4mapsin6(sin, (struct sockaddr_in6 *)sas); ((struct sockaddr_in6 *)sas)->sin6_port = inp->sctp_lport; sas = (struct sockaddr_storage *)((caddr_t)sas + sizeof(struct sockaddr_in6)); actual += sizeof(struct sockaddr_in6); } else { #endif memcpy(sas, sin, sizeof(*sin)); ((struct sockaddr_in *)sas)->sin_port = inp->sctp_lport; sas = (struct sockaddr_storage *)((caddr_t)sas + sizeof(*sin)); actual += sizeof(*sin); #ifdef INET6 } #endif if (actual >= limit) { return (actual); } } else { continue; } break; #ifdef INET6 case AF_INET6: if (ipv6_addr_legal) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&sctp_ifa->address.sa; if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { /* * we skip * unspecifed * addresses */ continue; } if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { if (local_scope == 0) continue; if (sin6->sin6_scope_id == 0) { if (sa6_recoverscope(sin6) != 0) /* * * bad * * li * nk * * loc * al * * add * re * ss * */ continue; } } if ((site_scope == 0) && (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { continue; } memcpy(sas, sin6, sizeof(*sin6)); ((struct sockaddr_in6 *)sas)->sin6_port = inp->sctp_lport; sas = (struct sockaddr_storage *)((caddr_t)sas + sizeof(*sin6)); actual += sizeof(*sin6); if (actual >= limit) { return (actual); } } else { continue; } break; #endif default: /* TSNH */ break; } } } } else { struct sctp_laddr *laddr; LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (stcb) { if (sctp_is_addr_restricted(stcb, laddr->ifa)) { continue; } } if (sctp_fill_user_address(sas, &laddr->ifa->address.sa)) continue; ((struct sockaddr_in6 *)sas)->sin6_port = inp->sctp_lport; sas = (struct sockaddr_storage *)((caddr_t)sas + laddr->ifa->address.sa.sa_len); actual += laddr->ifa->address.sa.sa_len; if (actual >= limit) { return (actual); } } } return (actual); } static size_t sctp_fill_up_addresses(struct sctp_inpcb *inp, struct sctp_tcb *stcb, size_t limit, struct sockaddr_storage *sas) { size_t size = 0; SCTP_IPI_ADDR_RLOCK(); /* fill up addresses for the endpoint's default vrf */ size = sctp_fill_up_addresses_vrf(inp, stcb, limit, sas, inp->def_vrf_id); SCTP_IPI_ADDR_RUNLOCK(); return (size); } /* * NOTE: assumes addr lock is held */ static int sctp_count_max_addresses_vrf(struct sctp_inpcb *inp, uint32_t vrf_id) { int cnt = 0; struct sctp_vrf *vrf = NULL; /* * In both sub-set bound an bound_all cases we return the MAXIMUM * number of addresses that you COULD get. In reality the sub-set * bound may have an exclusion list for a given TCB OR in the * bound-all case a TCB may NOT include the loopback or other * addresses as well. */ vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { return (0); } if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa; LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { /* Count them if they are the right type */ if (sctp_ifa->address.sa.sa_family == AF_INET) { if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) cnt += sizeof(struct sockaddr_in6); else cnt += sizeof(struct sockaddr_in); } else if (sctp_ifa->address.sa.sa_family == AF_INET6) cnt += sizeof(struct sockaddr_in6); } } } else { struct sctp_laddr *laddr; LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa->address.sa.sa_family == AF_INET) { if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) cnt += sizeof(struct sockaddr_in6); else cnt += sizeof(struct sockaddr_in); } else if (laddr->ifa->address.sa.sa_family == AF_INET6) cnt += sizeof(struct sockaddr_in6); } } return (cnt); } static int sctp_count_max_addresses(struct sctp_inpcb *inp) { int cnt = 0; SCTP_IPI_ADDR_RLOCK(); /* count addresses for the endpoint's default VRF */ cnt = sctp_count_max_addresses_vrf(inp, inp->def_vrf_id); SCTP_IPI_ADDR_RUNLOCK(); return (cnt); } static int sctp_do_connect_x(struct socket *so, struct sctp_inpcb *inp, void *optval, size_t optsize, void *p, int delay) { int error = 0; int creat_lock_on = 0; struct sctp_tcb *stcb = NULL; struct sockaddr *sa; int num_v6 = 0, num_v4 = 0, *totaddrp, totaddr; int added = 0; uint32_t vrf_id; int bad_addresses = 0; sctp_assoc_t *a_id; SCTPDBG(SCTP_DEBUG_PCB1, "Connectx called\n"); if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) { /* We are already connected AND the TCP model */ SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EADDRINUSE); return (EADDRINUSE); } - if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) && + (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE))) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { SCTP_INP_RLOCK(inp); stcb = LIST_FIRST(&inp->sctp_asoc_list); SCTP_INP_RUNLOCK(inp); } if (stcb) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); return (EALREADY); } SCTP_INP_INCR_REF(inp); SCTP_ASOC_CREATE_LOCK(inp); creat_lock_on = 1; if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EFAULT); error = EFAULT; goto out_now; } totaddrp = (int *)optval; totaddr = *totaddrp; sa = (struct sockaddr *)(totaddrp + 1); stcb = sctp_connectx_helper_find(inp, sa, &totaddr, &num_v4, &num_v6, &error, (optsize - sizeof(int)), &bad_addresses); if ((stcb != NULL) || bad_addresses) { /* Already have or am bring up an association */ SCTP_ASOC_CREATE_UNLOCK(inp); creat_lock_on = 0; if (stcb) SCTP_TCB_UNLOCK(stcb); if (bad_addresses == 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); error = EALREADY; } goto out_now; } #ifdef INET6 if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) && (num_v6 > 0)) { error = EINVAL; goto out_now; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && (num_v4 > 0)) { struct in6pcb *inp6; inp6 = (struct in6pcb *)inp; if (SCTP_IPV6_V6ONLY(inp6)) { /* * if IPV6_V6ONLY flag, ignore connections destined * to a v4 addr or v4-mapped addr */ SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto out_now; } } #endif /* INET6 */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == SCTP_PCB_FLAGS_UNBOUND) { /* Bind a ephemeral port */ error = sctp_inpcb_bind(so, NULL, NULL, p); if (error) { goto out_now; } } /* FIX ME: do we want to pass in a vrf on the connect call? */ vrf_id = inp->def_vrf_id; + /* We are GOOD to go */ stcb = sctp_aloc_assoc(inp, sa, 1, &error, 0, vrf_id, (struct thread *)p ); if (stcb == NULL) { /* Gak! no memory */ goto out_now; } SCTP_SET_STATE(&stcb->asoc, SCTP_STATE_COOKIE_WAIT); /* move to second address */ if (sa->sa_family == AF_INET) sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in)); else sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in6)); error = 0; added = sctp_connectx_helper_add(stcb, sa, (totaddr - 1), &error); /* Fill in the return id */ if (error) { (void)sctp_free_assoc(inp, stcb, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_12); goto out_now; } a_id = (sctp_assoc_t *) optval; *a_id = sctp_get_associd(stcb); /* initialize authentication parameters for the assoc */ sctp_initialize_auth_params(inp, stcb); if (delay) { /* doing delayed connection */ stcb->asoc.delayed_connection = 1; sctp_timer_start(SCTP_TIMER_TYPE_INIT, inp, stcb, stcb->asoc.primary_destination); } else { (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); } SCTP_TCB_UNLOCK(stcb); if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) { stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED; /* Set the connected flag so we can queue data */ soisconnecting(so); } out_now: if (creat_lock_on) { SCTP_ASOC_CREATE_UNLOCK(inp); } SCTP_INP_DECR_REF(inp); return error; } #define SCTP_FIND_STCB(inp, stcb, assoc_id) { \ if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||\ (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { \ SCTP_INP_RLOCK(inp); \ stcb = LIST_FIRST(&inp->sctp_asoc_list); \ if (stcb) { \ SCTP_TCB_LOCK(stcb); \ } \ SCTP_INP_RUNLOCK(inp); \ } else if (assoc_id != 0) { \ stcb = sctp_findassociation_ep_asocid(inp, assoc_id, 1); \ if (stcb == NULL) { \ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); \ error = ENOENT; \ break; \ } \ } else { \ stcb = NULL; \ } \ } #define SCTP_CHECK_AND_CAST(destp, srcp, type, size) {\ if (size < sizeof(type)) { \ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); \ error = EINVAL; \ break; \ } else { \ destp = (type *)srcp; \ } \ } static int sctp_getopt(struct socket *so, int optname, void *optval, size_t *optsize, void *p) { struct sctp_inpcb *inp = NULL; int error, val = 0; struct sctp_tcb *stcb = NULL; if (optval == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } inp = (struct sctp_inpcb *)so->so_pcb; if (inp == 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return EINVAL; } error = 0; switch (optname) { case SCTP_NODELAY: case SCTP_AUTOCLOSE: case SCTP_EXPLICIT_EOR: case SCTP_AUTO_ASCONF: case SCTP_DISABLE_FRAGMENTS: case SCTP_I_WANT_MAPPED_V4_ADDR: case SCTP_USE_EXT_RCVINFO: SCTP_INP_RLOCK(inp); switch (optname) { case SCTP_DISABLE_FRAGMENTS: val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NO_FRAGMENT); break; case SCTP_I_WANT_MAPPED_V4_ADDR: val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4); break; case SCTP_AUTO_ASCONF: if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* only valid for bound all sockets */ val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto flags_out; } break; case SCTP_EXPLICIT_EOR: val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR); break; case SCTP_NODELAY: val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NODELAY); break; case SCTP_USE_EXT_RCVINFO: val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO); break; case SCTP_AUTOCLOSE: if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) val = TICKS_TO_SEC(inp->sctp_ep.auto_close_time); else val = 0; break; default: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT); error = ENOPROTOOPT; } /* end switch (sopt->sopt_name) */ if (optname != SCTP_AUTOCLOSE) { /* make it an "on/off" value */ val = (val != 0); } if (*optsize < sizeof(val)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } flags_out: SCTP_INP_RUNLOCK(inp); if (error == 0) { /* return the option value */ *(int *)optval = val; *optsize = sizeof(val); } break; case SCTP_GET_PACKET_LOG: { #ifdef SCTP_PACKET_LOGGING uint8_t *target; int ret; SCTP_CHECK_AND_CAST(target, optval, uint8_t, *optsize); ret = sctp_copy_out_packet_log(target, (int)*optsize); *optsize = ret; #else SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); error = EOPNOTSUPP; #endif break; } + case SCTP_REUSE_PORT: + { + uint32_t *value; + + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE)) { + /* Can't do this for a 1-m socket */ + error = EINVAL; + break; + } + SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); + *value = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE); + *optsize = sizeof(uint32_t); + } + break; case SCTP_PARTIAL_DELIVERY_POINT: { uint32_t *value; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); *value = inp->partial_delivery_point; *optsize = sizeof(uint32_t); } break; case SCTP_FRAGMENT_INTERLEAVE: { uint32_t *value; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE)) { if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS)) { *value = SCTP_FRAG_LEVEL_2; } else { *value = SCTP_FRAG_LEVEL_1; } } else { *value = SCTP_FRAG_LEVEL_0; } *optsize = sizeof(uint32_t); } break; case SCTP_CMT_ON_OFF: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); if (SCTP_BASE_SYSCTL(sctp_cmt_on_off)) { SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.sctp_cmt_on_off; SCTP_TCB_UNLOCK(stcb); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); error = ENOTCONN; } } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT); error = ENOPROTOOPT; } *optsize = sizeof(*av); } break; /* JRS - Get socket option for pluggable congestion control */ case SCTP_PLUGGABLE_CC: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.congestion_control_module; SCTP_TCB_UNLOCK(stcb); } else { av->assoc_value = inp->sctp_ep.sctp_default_cc_module; } *optsize = sizeof(*av); } break; case SCTP_GET_ADDR_LEN: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); error = EINVAL; #ifdef INET if (av->assoc_value == AF_INET) { av->assoc_value = sizeof(struct sockaddr_in); error = 0; } #endif #ifdef INET6 if (av->assoc_value == AF_INET6) { av->assoc_value = sizeof(struct sockaddr_in6); error = 0; } #endif if (error) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); } *optsize = sizeof(*av); } break; case SCTP_GET_ASSOC_NUMBER: { uint32_t *value, cnt; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); cnt = 0; SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { cnt++; } SCTP_INP_RUNLOCK(inp); *value = cnt; *optsize = sizeof(uint32_t); } break; case SCTP_GET_ASSOC_ID_LIST: { struct sctp_assoc_ids *ids; unsigned int at, limit; SCTP_CHECK_AND_CAST(ids, optval, struct sctp_assoc_ids, *optsize); at = 0; limit = *optsize / sizeof(sctp_assoc_t); SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { if (at < limit) { ids->gaids_assoc_id[at++] = sctp_get_associd(stcb); } else { error = EINVAL; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } SCTP_INP_RUNLOCK(inp); *optsize = at * sizeof(sctp_assoc_t); } break; case SCTP_CONTEXT: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.context; SCTP_TCB_UNLOCK(stcb); } else { SCTP_INP_RLOCK(inp); av->assoc_value = inp->sctp_context; SCTP_INP_RUNLOCK(inp); } *optsize = sizeof(*av); } break; case SCTP_VRF_ID: { uint32_t *default_vrfid; SCTP_CHECK_AND_CAST(default_vrfid, optval, uint32_t, *optsize); *default_vrfid = inp->def_vrf_id; break; } case SCTP_GET_ASOC_VRF: { struct sctp_assoc_value *id; SCTP_CHECK_AND_CAST(id, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, id->assoc_id); if (stcb == NULL) { error = EINVAL; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } id->assoc_value = stcb->asoc.vrf_id; break; } case SCTP_GET_VRF_IDS: { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); error = EOPNOTSUPP; break; } case SCTP_GET_NONCE_VALUES: { struct sctp_get_nonce_values *gnv; SCTP_CHECK_AND_CAST(gnv, optval, struct sctp_get_nonce_values, *optsize); SCTP_FIND_STCB(inp, stcb, gnv->gn_assoc_id); if (stcb) { gnv->gn_peers_tag = stcb->asoc.peer_vtag; gnv->gn_local_tag = stcb->asoc.my_vtag; SCTP_TCB_UNLOCK(stcb); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); error = ENOTCONN; } *optsize = sizeof(*gnv); } break; case SCTP_DELAYED_SACK: { struct sctp_sack_info *sack; SCTP_CHECK_AND_CAST(sack, optval, struct sctp_sack_info, *optsize); SCTP_FIND_STCB(inp, stcb, sack->sack_assoc_id); if (stcb) { sack->sack_delay = stcb->asoc.delayed_ack; sack->sack_freq = stcb->asoc.sack_freq; SCTP_TCB_UNLOCK(stcb); } else { SCTP_INP_RLOCK(inp); sack->sack_delay = TICKS_TO_MSEC(inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_RECV]); sack->sack_freq = inp->sctp_ep.sctp_sack_freq; SCTP_INP_RUNLOCK(inp); } *optsize = sizeof(*sack); } break; case SCTP_GET_SNDBUF_USE: { struct sctp_sockstat *ss; SCTP_CHECK_AND_CAST(ss, optval, struct sctp_sockstat, *optsize); SCTP_FIND_STCB(inp, stcb, ss->ss_assoc_id); if (stcb) { ss->ss_total_sndbuf = stcb->asoc.total_output_queue_size; ss->ss_total_recv_buf = (stcb->asoc.size_on_reasm_queue + stcb->asoc.size_on_all_streams); SCTP_TCB_UNLOCK(stcb); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); error = ENOTCONN; } *optsize = sizeof(struct sctp_sockstat); } break; case SCTP_MAX_BURST: { uint8_t *value; SCTP_CHECK_AND_CAST(value, optval, uint8_t, *optsize); SCTP_INP_RLOCK(inp); *value = inp->sctp_ep.max_burst; SCTP_INP_RUNLOCK(inp); *optsize = sizeof(uint8_t); } break; case SCTP_MAXSEG: { struct sctp_assoc_value *av; int ovh; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = sctp_get_frag_point(stcb, &stcb->asoc); SCTP_TCB_UNLOCK(stcb); } else { SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ovh = SCTP_MED_OVERHEAD; } else { ovh = SCTP_MED_V4_OVERHEAD; } if (inp->sctp_frag_point >= SCTP_DEFAULT_MAXSEGMENT) av->assoc_value = 0; else av->assoc_value = inp->sctp_frag_point - ovh; SCTP_INP_RUNLOCK(inp); } *optsize = sizeof(struct sctp_assoc_value); } break; case SCTP_GET_STAT_LOG: error = sctp_fill_stat_log(optval, optsize); break; case SCTP_EVENTS: { struct sctp_event_subscribe *events; SCTP_CHECK_AND_CAST(events, optval, struct sctp_event_subscribe, *optsize); memset(events, 0, sizeof(*events)); SCTP_INP_RLOCK(inp); if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT)) events->sctp_data_io_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVASSOCEVNT)) events->sctp_association_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVPADDREVNT)) events->sctp_address_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVSENDFAILEVNT)) events->sctp_send_failure_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVPEERERR)) events->sctp_peer_error_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT)) events->sctp_shutdown_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT)) events->sctp_partial_delivery_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_ADAPTATIONEVNT)) events->sctp_adaptation_layer_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTHEVNT)) events->sctp_authentication_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_STREAM_RESETEVNT)) events->sctp_stream_reset_events = 1; SCTP_INP_RUNLOCK(inp); *optsize = sizeof(struct sctp_event_subscribe); } break; case SCTP_ADAPTATION_LAYER: { uint32_t *value; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); SCTP_INP_RLOCK(inp); *value = inp->sctp_ep.adaptation_layer_indicator; SCTP_INP_RUNLOCK(inp); *optsize = sizeof(uint32_t); } break; case SCTP_SET_INITIAL_DBG_SEQ: { uint32_t *value; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); SCTP_INP_RLOCK(inp); *value = inp->sctp_ep.initial_sequence_debug; SCTP_INP_RUNLOCK(inp); *optsize = sizeof(uint32_t); } break; case SCTP_GET_LOCAL_ADDR_SIZE: { uint32_t *value; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); SCTP_INP_RLOCK(inp); *value = sctp_count_max_addresses(inp); SCTP_INP_RUNLOCK(inp); *optsize = sizeof(uint32_t); } break; case SCTP_GET_REMOTE_ADDR_SIZE: { uint32_t *value; size_t size; struct sctp_nets *net; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); /* FIXME MT: change to sctp_assoc_value? */ SCTP_FIND_STCB(inp, stcb, (sctp_assoc_t) * value); if (stcb) { size = 0; /* Count the sizes */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) || (((struct sockaddr *)&net->ro._l_addr)->sa_family == AF_INET6)) { size += sizeof(struct sockaddr_in6); } else if (((struct sockaddr *)&net->ro._l_addr)->sa_family == AF_INET) { size += sizeof(struct sockaddr_in); } else { /* huh */ break; } } SCTP_TCB_UNLOCK(stcb); *value = (uint32_t) size; } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); error = ENOTCONN; } *optsize = sizeof(uint32_t); } break; case SCTP_GET_PEER_ADDRESSES: /* * Get the address information, an array is passed in to * fill up we pack it. */ { size_t cpsz, left; struct sockaddr_storage *sas; struct sctp_nets *net; struct sctp_getaddresses *saddr; SCTP_CHECK_AND_CAST(saddr, optval, struct sctp_getaddresses, *optsize); SCTP_FIND_STCB(inp, stcb, saddr->sget_assoc_id); if (stcb) { left = (*optsize) - sizeof(struct sctp_getaddresses); *optsize = sizeof(struct sctp_getaddresses); sas = (struct sockaddr_storage *)&saddr->addr[0]; TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) || (((struct sockaddr *)&net->ro._l_addr)->sa_family == AF_INET6)) { cpsz = sizeof(struct sockaddr_in6); } else if (((struct sockaddr *)&net->ro._l_addr)->sa_family == AF_INET) { cpsz = sizeof(struct sockaddr_in); } else { /* huh */ break; } if (left < cpsz) { /* not enough room. */ break; } #ifdef INET6 if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) && (((struct sockaddr *)&net->ro._l_addr)->sa_family == AF_INET)) { /* Must map the address */ in6_sin_2_v4mapsin6((struct sockaddr_in *)&net->ro._l_addr, (struct sockaddr_in6 *)sas); } else { #endif memcpy(sas, &net->ro._l_addr, cpsz); #ifdef INET6 } #endif ((struct sockaddr_in *)sas)->sin_port = stcb->rport; sas = (struct sockaddr_storage *)((caddr_t)sas + cpsz); left -= cpsz; *optsize += cpsz; } SCTP_TCB_UNLOCK(stcb); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; } } break; case SCTP_GET_LOCAL_ADDRESSES: { size_t limit, actual; struct sockaddr_storage *sas; struct sctp_getaddresses *saddr; SCTP_CHECK_AND_CAST(saddr, optval, struct sctp_getaddresses, *optsize); SCTP_FIND_STCB(inp, stcb, saddr->sget_assoc_id); sas = (struct sockaddr_storage *)&saddr->addr[0]; limit = *optsize - sizeof(sctp_assoc_t); actual = sctp_fill_up_addresses(inp, stcb, limit, sas); if (stcb) { SCTP_TCB_UNLOCK(stcb); } *optsize = sizeof(struct sockaddr_storage) + actual; } break; case SCTP_PEER_ADDR_PARAMS: { struct sctp_paddrparams *paddrp; struct sctp_nets *net; SCTP_CHECK_AND_CAST(paddrp, optval, struct sctp_paddrparams, *optsize); SCTP_FIND_STCB(inp, stcb, paddrp->spp_assoc_id); net = NULL; if (stcb) { net = sctp_findnet(stcb, (struct sockaddr *)&paddrp->spp_address); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, (struct sockaddr *)&paddrp->spp_address, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if (stcb && (net == NULL)) { struct sockaddr *sa; sa = (struct sockaddr *)&paddrp->spp_address; if (sa->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)sa; if (sin->sin_addr.s_addr) { error = EINVAL; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } else if (sa->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)sa; if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { error = EINVAL; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } else { error = EAFNOSUPPORT; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } if (stcb) { /* Applys to the specific association */ paddrp->spp_flags = 0; if (net) { int ovh; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ovh = SCTP_MED_OVERHEAD; } else { ovh = SCTP_MED_V4_OVERHEAD; } paddrp->spp_pathmaxrxt = net->failure_threshold; paddrp->spp_pathmtu = net->mtu - ovh; /* get flags for HB */ if (net->dest_state & SCTP_ADDR_NOHB) paddrp->spp_flags |= SPP_HB_DISABLE; else paddrp->spp_flags |= SPP_HB_ENABLE; /* get flags for PMTU */ if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { paddrp->spp_flags |= SPP_PMTUD_ENABLE; } else { paddrp->spp_flags |= SPP_PMTUD_DISABLE; } #ifdef INET if (net->ro._l_addr.sin.sin_family == AF_INET) { paddrp->spp_ipv4_tos = net->tos_flowlabel & 0x000000fc; paddrp->spp_flags |= SPP_IPV4_TOS; } #endif #ifdef INET6 if (net->ro._l_addr.sin6.sin6_family == AF_INET6) { paddrp->spp_ipv6_flowlabel = net->tos_flowlabel; paddrp->spp_flags |= SPP_IPV6_FLOWLABEL; } #endif } else { /* * No destination so return default * value */ int cnt = 0; paddrp->spp_pathmaxrxt = stcb->asoc.def_net_failure; paddrp->spp_pathmtu = sctp_get_frag_point(stcb, &stcb->asoc); #ifdef INET paddrp->spp_ipv4_tos = stcb->asoc.default_tos & 0x000000fc; paddrp->spp_flags |= SPP_IPV4_TOS; #endif #ifdef INET6 paddrp->spp_ipv6_flowlabel = stcb->asoc.default_flowlabel; paddrp->spp_flags |= SPP_IPV6_FLOWLABEL; #endif /* default settings should be these */ if (stcb->asoc.hb_is_disabled == 0) { paddrp->spp_flags |= SPP_HB_ENABLE; } else { paddrp->spp_flags |= SPP_HB_DISABLE; } TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { cnt++; } } if (cnt) { paddrp->spp_flags |= SPP_PMTUD_ENABLE; } } paddrp->spp_hbinterval = stcb->asoc.heart_beat_delay; paddrp->spp_assoc_id = sctp_get_associd(stcb); SCTP_TCB_UNLOCK(stcb); } else { /* Use endpoint defaults */ SCTP_INP_RLOCK(inp); paddrp->spp_pathmaxrxt = inp->sctp_ep.def_net_failure; paddrp->spp_hbinterval = TICKS_TO_MSEC(inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT]); paddrp->spp_assoc_id = (sctp_assoc_t) 0; /* get inp's default */ #ifdef INET paddrp->spp_ipv4_tos = inp->ip_inp.inp.inp_ip_tos; paddrp->spp_flags |= SPP_IPV4_TOS; #endif #ifdef INET6 if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { paddrp->spp_ipv6_flowlabel = ((struct in6pcb *)inp)->in6p_flowinfo; paddrp->spp_flags |= SPP_IPV6_FLOWLABEL; } #endif /* can't return this */ paddrp->spp_pathmtu = 0; /* default behavior, no stcb */ paddrp->spp_flags = SPP_PMTUD_ENABLE; if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) { paddrp->spp_flags |= SPP_HB_ENABLE; } else { paddrp->spp_flags |= SPP_HB_DISABLE; } SCTP_INP_RUNLOCK(inp); } *optsize = sizeof(struct sctp_paddrparams); } break; case SCTP_GET_PEER_ADDR_INFO: { struct sctp_paddrinfo *paddri; struct sctp_nets *net; SCTP_CHECK_AND_CAST(paddri, optval, struct sctp_paddrinfo, *optsize); SCTP_FIND_STCB(inp, stcb, paddri->spinfo_assoc_id); net = NULL; if (stcb) { net = sctp_findnet(stcb, (struct sockaddr *)&paddri->spinfo_address); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, (struct sockaddr *)&paddri->spinfo_address, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if ((stcb) && (net)) { paddri->spinfo_state = net->dest_state & (SCTP_REACHABLE_MASK | SCTP_ADDR_NOHB); paddri->spinfo_cwnd = net->cwnd; paddri->spinfo_srtt = ((net->lastsa >> 2) + net->lastsv) >> 1; paddri->spinfo_rto = net->RTO; paddri->spinfo_assoc_id = sctp_get_associd(stcb); SCTP_TCB_UNLOCK(stcb); } else { if (stcb) { SCTP_TCB_UNLOCK(stcb); } SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; } *optsize = sizeof(struct sctp_paddrinfo); } break; case SCTP_PCB_STATUS: { struct sctp_pcbinfo *spcb; SCTP_CHECK_AND_CAST(spcb, optval, struct sctp_pcbinfo, *optsize); sctp_fill_pcbinfo(spcb); *optsize = sizeof(struct sctp_pcbinfo); } break; case SCTP_STATUS: { struct sctp_nets *net; struct sctp_status *sstat; SCTP_CHECK_AND_CAST(sstat, optval, struct sctp_status, *optsize); SCTP_FIND_STCB(inp, stcb, sstat->sstat_assoc_id); if (stcb == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); error = EINVAL; break; } /* * I think passing the state is fine since * sctp_constants.h will be available to the user * land. */ sstat->sstat_state = stcb->asoc.state; sstat->sstat_assoc_id = sctp_get_associd(stcb); sstat->sstat_rwnd = stcb->asoc.peers_rwnd; sstat->sstat_unackdata = stcb->asoc.sent_queue_cnt; /* * We can't include chunks that have been passed to * the socket layer. Only things in queue. */ sstat->sstat_penddata = (stcb->asoc.cnt_on_reasm_queue + stcb->asoc.cnt_on_all_streams); sstat->sstat_instrms = stcb->asoc.streamincnt; sstat->sstat_outstrms = stcb->asoc.streamoutcnt; sstat->sstat_fragmentation_point = sctp_get_frag_point(stcb, &stcb->asoc); memcpy(&sstat->sstat_primary.spinfo_address, &stcb->asoc.primary_destination->ro._l_addr, ((struct sockaddr *)(&stcb->asoc.primary_destination->ro._l_addr))->sa_len); net = stcb->asoc.primary_destination; ((struct sockaddr_in *)&sstat->sstat_primary.spinfo_address)->sin_port = stcb->rport; /* * Again the user can get info from sctp_constants.h * for what the state of the network is. */ sstat->sstat_primary.spinfo_state = net->dest_state & SCTP_REACHABLE_MASK; sstat->sstat_primary.spinfo_cwnd = net->cwnd; sstat->sstat_primary.spinfo_srtt = net->lastsa; sstat->sstat_primary.spinfo_rto = net->RTO; sstat->sstat_primary.spinfo_mtu = net->mtu; sstat->sstat_primary.spinfo_assoc_id = sctp_get_associd(stcb); SCTP_TCB_UNLOCK(stcb); *optsize = sizeof(*sstat); } break; case SCTP_RTOINFO: { struct sctp_rtoinfo *srto; SCTP_CHECK_AND_CAST(srto, optval, struct sctp_rtoinfo, *optsize); SCTP_FIND_STCB(inp, stcb, srto->srto_assoc_id); if (stcb) { srto->srto_initial = stcb->asoc.initial_rto; srto->srto_max = stcb->asoc.maxrto; srto->srto_min = stcb->asoc.minrto; SCTP_TCB_UNLOCK(stcb); } else { SCTP_INP_RLOCK(inp); srto->srto_initial = inp->sctp_ep.initial_rto; srto->srto_max = inp->sctp_ep.sctp_maxrto; srto->srto_min = inp->sctp_ep.sctp_minrto; SCTP_INP_RUNLOCK(inp); } *optsize = sizeof(*srto); } break; case SCTP_ASSOCINFO: { struct sctp_assocparams *sasoc; uint32_t oldval; SCTP_CHECK_AND_CAST(sasoc, optval, struct sctp_assocparams, *optsize); SCTP_FIND_STCB(inp, stcb, sasoc->sasoc_assoc_id); if (stcb) { oldval = sasoc->sasoc_cookie_life; sasoc->sasoc_cookie_life = TICKS_TO_MSEC(stcb->asoc.cookie_life); sasoc->sasoc_asocmaxrxt = stcb->asoc.max_send_times; sasoc->sasoc_number_peer_destinations = stcb->asoc.numnets; sasoc->sasoc_peer_rwnd = stcb->asoc.peers_rwnd; sasoc->sasoc_local_rwnd = stcb->asoc.my_rwnd; SCTP_TCB_UNLOCK(stcb); } else { SCTP_INP_RLOCK(inp); sasoc->sasoc_cookie_life = TICKS_TO_MSEC(inp->sctp_ep.def_cookie_life); sasoc->sasoc_asocmaxrxt = inp->sctp_ep.max_send_times; sasoc->sasoc_number_peer_destinations = 0; sasoc->sasoc_peer_rwnd = 0; sasoc->sasoc_local_rwnd = sbspace(&inp->sctp_socket->so_rcv); SCTP_INP_RUNLOCK(inp); } *optsize = sizeof(*sasoc); } break; case SCTP_DEFAULT_SEND_PARAM: { struct sctp_sndrcvinfo *s_info; SCTP_CHECK_AND_CAST(s_info, optval, struct sctp_sndrcvinfo, *optsize); SCTP_FIND_STCB(inp, stcb, s_info->sinfo_assoc_id); if (stcb) { memcpy(s_info, &stcb->asoc.def_send, sizeof(stcb->asoc.def_send)); SCTP_TCB_UNLOCK(stcb); } else { SCTP_INP_RLOCK(inp); memcpy(s_info, &inp->def_send, sizeof(inp->def_send)); SCTP_INP_RUNLOCK(inp); } *optsize = sizeof(*s_info); } break; case SCTP_INITMSG: { struct sctp_initmsg *sinit; SCTP_CHECK_AND_CAST(sinit, optval, struct sctp_initmsg, *optsize); SCTP_INP_RLOCK(inp); sinit->sinit_num_ostreams = inp->sctp_ep.pre_open_stream_count; sinit->sinit_max_instreams = inp->sctp_ep.max_open_streams_intome; sinit->sinit_max_attempts = inp->sctp_ep.max_init_times; sinit->sinit_max_init_timeo = inp->sctp_ep.initial_init_rto_max; SCTP_INP_RUNLOCK(inp); *optsize = sizeof(*sinit); } break; case SCTP_PRIMARY_ADDR: /* we allow a "get" operation on this */ { struct sctp_setprim *ssp; SCTP_CHECK_AND_CAST(ssp, optval, struct sctp_setprim, *optsize); SCTP_FIND_STCB(inp, stcb, ssp->ssp_assoc_id); if (stcb) { /* simply copy out the sockaddr_storage... */ int len; len = *optsize; if (len > stcb->asoc.primary_destination->ro._l_addr.sa.sa_len) len = stcb->asoc.primary_destination->ro._l_addr.sa.sa_len; memcpy(&ssp->ssp_addr, &stcb->asoc.primary_destination->ro._l_addr, len); SCTP_TCB_UNLOCK(stcb); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); error = EINVAL; } *optsize = sizeof(*ssp); } break; case SCTP_HMAC_IDENT: { struct sctp_hmacalgo *shmac; sctp_hmaclist_t *hmaclist; uint32_t size; int i; SCTP_CHECK_AND_CAST(shmac, optval, struct sctp_hmacalgo, *optsize); SCTP_INP_RLOCK(inp); hmaclist = inp->sctp_ep.local_hmacs; if (hmaclist == NULL) { /* no HMACs to return */ *optsize = sizeof(*shmac); SCTP_INP_RUNLOCK(inp); break; } /* is there room for all of the hmac ids? */ size = sizeof(*shmac) + (hmaclist->num_algo * sizeof(shmac->shmac_idents[0])); if ((size_t)(*optsize) < size) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); error = EINVAL; SCTP_INP_RUNLOCK(inp); break; } /* copy in the list */ - for (i = 0; i < hmaclist->num_algo; i++) + shmac->shmac_number_of_idents = hmaclist->num_algo; + for (i = 0; i < hmaclist->num_algo; i++) { shmac->shmac_idents[i] = hmaclist->hmac[i]; + } SCTP_INP_RUNLOCK(inp); *optsize = size; break; } case SCTP_AUTH_ACTIVE_KEY: { struct sctp_authkeyid *scact; SCTP_CHECK_AND_CAST(scact, optval, struct sctp_authkeyid, *optsize); SCTP_FIND_STCB(inp, stcb, scact->scact_assoc_id); if (stcb) { /* get the active key on the assoc */ scact->scact_keynumber = stcb->asoc.authinfo.assoc_keyid; SCTP_TCB_UNLOCK(stcb); } else { /* get the endpoint active key */ SCTP_INP_RLOCK(inp); scact->scact_keynumber = inp->sctp_ep.default_keyid; SCTP_INP_RUNLOCK(inp); } *optsize = sizeof(*scact); break; } case SCTP_LOCAL_AUTH_CHUNKS: { struct sctp_authchunks *sac; sctp_auth_chklist_t *chklist = NULL; size_t size = 0; SCTP_CHECK_AND_CAST(sac, optval, struct sctp_authchunks, *optsize); SCTP_FIND_STCB(inp, stcb, sac->gauth_assoc_id); if (stcb) { /* get off the assoc */ chklist = stcb->asoc.local_auth_chunks; /* is there enough space? */ size = sctp_auth_get_chklist_size(chklist); if (*optsize < (sizeof(struct sctp_authchunks) + size)) { error = EINVAL; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); } else { /* copy in the chunks */ (void)sctp_serialize_auth_chunks(chklist, sac->gauth_chunks); } SCTP_TCB_UNLOCK(stcb); } else { /* get off the endpoint */ SCTP_INP_RLOCK(inp); chklist = inp->sctp_ep.local_auth_chunks; /* is there enough space? */ size = sctp_auth_get_chklist_size(chklist); if (*optsize < (sizeof(struct sctp_authchunks) + size)) { error = EINVAL; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); } else { /* copy in the chunks */ (void)sctp_serialize_auth_chunks(chklist, sac->gauth_chunks); } SCTP_INP_RUNLOCK(inp); } *optsize = sizeof(struct sctp_authchunks) + size; break; } case SCTP_PEER_AUTH_CHUNKS: { struct sctp_authchunks *sac; sctp_auth_chklist_t *chklist = NULL; size_t size = 0; SCTP_CHECK_AND_CAST(sac, optval, struct sctp_authchunks, *optsize); SCTP_FIND_STCB(inp, stcb, sac->gauth_assoc_id); if (stcb) { /* get off the assoc */ chklist = stcb->asoc.peer_auth_chunks; /* is there enough space? */ size = sctp_auth_get_chklist_size(chklist); if (*optsize < (sizeof(struct sctp_authchunks) + size)) { error = EINVAL; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); } else { /* copy in the chunks */ (void)sctp_serialize_auth_chunks(chklist, sac->gauth_chunks); } SCTP_TCB_UNLOCK(stcb); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; } *optsize = sizeof(struct sctp_authchunks) + size; break; } default: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT); error = ENOPROTOOPT; *optsize = 0; break; } /* end switch (sopt->sopt_name) */ return (error); } static int sctp_setopt(struct socket *so, int optname, void *optval, size_t optsize, void *p) { int error, set_opt; uint32_t *mopt; struct sctp_tcb *stcb = NULL; struct sctp_inpcb *inp = NULL; uint32_t vrf_id; if (optval == NULL) { SCTP_PRINTF("optval is NULL\n"); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } inp = (struct sctp_inpcb *)so->so_pcb; if (inp == 0) { SCTP_PRINTF("inp is NULL?\n"); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return EINVAL; } vrf_id = inp->def_vrf_id; error = 0; switch (optname) { case SCTP_NODELAY: case SCTP_AUTOCLOSE: case SCTP_AUTO_ASCONF: case SCTP_EXPLICIT_EOR: case SCTP_DISABLE_FRAGMENTS: case SCTP_USE_EXT_RCVINFO: case SCTP_I_WANT_MAPPED_V4_ADDR: /* copy in the option value */ SCTP_CHECK_AND_CAST(mopt, optval, uint32_t, optsize); set_opt = 0; if (error) break; switch (optname) { case SCTP_DISABLE_FRAGMENTS: set_opt = SCTP_PCB_FLAGS_NO_FRAGMENT; break; case SCTP_AUTO_ASCONF: /* * NOTE: we don't really support this flag */ if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* only valid for bound all sockets */ set_opt = SCTP_PCB_FLAGS_AUTO_ASCONF; } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } break; case SCTP_EXPLICIT_EOR: set_opt = SCTP_PCB_FLAGS_EXPLICIT_EOR; break; case SCTP_USE_EXT_RCVINFO: set_opt = SCTP_PCB_FLAGS_EXT_RCVINFO; break; case SCTP_I_WANT_MAPPED_V4_ADDR: if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { set_opt = SCTP_PCB_FLAGS_NEEDS_MAPPED_V4; } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } break; case SCTP_NODELAY: set_opt = SCTP_PCB_FLAGS_NODELAY; break; case SCTP_AUTOCLOSE: if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } set_opt = SCTP_PCB_FLAGS_AUTOCLOSE; /* * The value is in ticks. Note this does not effect * old associations, only new ones. */ inp->sctp_ep.auto_close_time = SEC_TO_TICKS(*mopt); break; } SCTP_INP_WLOCK(inp); if (*mopt != 0) { sctp_feature_on(inp, set_opt); } else { sctp_feature_off(inp, set_opt); } SCTP_INP_WUNLOCK(inp); break; + case SCTP_REUSE_PORT: + { + SCTP_CHECK_AND_CAST(mopt, optval, uint32_t, optsize); + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) { + /* Can't set it after we are bound */ + error = EINVAL; + break; + } + if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE)) { + /* Can't do this for a 1-m socket */ + error = EINVAL; + break; + } + if (optval) + sctp_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE); + else + sctp_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE); + } + break; case SCTP_PARTIAL_DELIVERY_POINT: { uint32_t *value; SCTP_CHECK_AND_CAST(value, optval, uint32_t, optsize); if (*value > SCTP_SB_LIMIT_RCV(so)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } inp->partial_delivery_point = *value; } break; case SCTP_FRAGMENT_INTERLEAVE: /* not yet until we re-write sctp_recvmsg() */ { uint32_t *level; SCTP_CHECK_AND_CAST(level, optval, uint32_t, optsize); if (*level == SCTP_FRAG_LEVEL_2) { sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } else if (*level == SCTP_FRAG_LEVEL_1) { sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } else if (*level == SCTP_FRAG_LEVEL_0) { sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; case SCTP_CMT_ON_OFF: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); if (SCTP_BASE_SYSCTL(sctp_cmt_on_off)) { SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { stcb->asoc.sctp_cmt_on_off = (uint8_t) av->assoc_value; SCTP_TCB_UNLOCK(stcb); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); error = ENOTCONN; } } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT); error = ENOPROTOOPT; } } break; /* JRS - Set socket option for pluggable congestion control */ case SCTP_PLUGGABLE_CC: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { switch (av->assoc_value) { /* * JRS - Standard TCP congestion * control */ case SCTP_CC_RFC2581: { stcb->asoc.congestion_control_module = SCTP_CC_RFC2581; stcb->asoc.cc_functions.sctp_set_initial_cc_param = &sctp_set_initial_cc_param; stcb->asoc.cc_functions.sctp_cwnd_update_after_sack = &sctp_cwnd_update_after_sack; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr = &sctp_cwnd_update_after_fr; stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout = &sctp_cwnd_update_after_timeout; stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo = &sctp_cwnd_update_after_ecn_echo; stcb->asoc.cc_functions.sctp_cwnd_update_after_packet_dropped = &sctp_cwnd_update_after_packet_dropped; stcb->asoc.cc_functions.sctp_cwnd_update_after_output = &sctp_cwnd_update_after_output; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer = &sctp_cwnd_update_after_fr_timer; SCTP_TCB_UNLOCK(stcb); break; } /* * JRS - High Speed TCP congestion * control (Floyd) */ case SCTP_CC_HSTCP: { stcb->asoc.congestion_control_module = SCTP_CC_HSTCP; stcb->asoc.cc_functions.sctp_set_initial_cc_param = &sctp_set_initial_cc_param; stcb->asoc.cc_functions.sctp_cwnd_update_after_sack = &sctp_hs_cwnd_update_after_sack; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr = &sctp_hs_cwnd_update_after_fr; stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout = &sctp_cwnd_update_after_timeout; stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo = &sctp_cwnd_update_after_ecn_echo; stcb->asoc.cc_functions.sctp_cwnd_update_after_packet_dropped = &sctp_cwnd_update_after_packet_dropped; stcb->asoc.cc_functions.sctp_cwnd_update_after_output = &sctp_cwnd_update_after_output; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer = &sctp_cwnd_update_after_fr_timer; SCTP_TCB_UNLOCK(stcb); break; } /* JRS - HTCP congestion control */ case SCTP_CC_HTCP: { stcb->asoc.congestion_control_module = SCTP_CC_HTCP; stcb->asoc.cc_functions.sctp_set_initial_cc_param = &sctp_htcp_set_initial_cc_param; stcb->asoc.cc_functions.sctp_cwnd_update_after_sack = &sctp_htcp_cwnd_update_after_sack; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr = &sctp_htcp_cwnd_update_after_fr; stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout = &sctp_htcp_cwnd_update_after_timeout; stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo = &sctp_htcp_cwnd_update_after_ecn_echo; stcb->asoc.cc_functions.sctp_cwnd_update_after_packet_dropped = &sctp_cwnd_update_after_packet_dropped; stcb->asoc.cc_functions.sctp_cwnd_update_after_output = &sctp_cwnd_update_after_output; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer = &sctp_htcp_cwnd_update_after_fr_timer; SCTP_TCB_UNLOCK(stcb); break; } /* * JRS - All other values are * invalid */ default: { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; SCTP_TCB_UNLOCK(stcb); break; } } } else { switch (av->assoc_value) { case SCTP_CC_RFC2581: case SCTP_CC_HSTCP: case SCTP_CC_HTCP: inp->sctp_ep.sctp_default_cc_module = av->assoc_value; break; default: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; }; } } break; case SCTP_CLR_STAT_LOG: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); error = EOPNOTSUPP; break; case SCTP_CONTEXT: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { stcb->asoc.context = av->assoc_value; SCTP_TCB_UNLOCK(stcb); } else { SCTP_INP_WLOCK(inp); inp->sctp_context = av->assoc_value; SCTP_INP_WUNLOCK(inp); } } break; case SCTP_VRF_ID: { uint32_t *default_vrfid; SCTP_CHECK_AND_CAST(default_vrfid, optval, uint32_t, optsize); if (*default_vrfid > SCTP_MAX_VRF_ID) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } inp->def_vrf_id = *default_vrfid; break; } case SCTP_DEL_VRF_ID: { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); error = EOPNOTSUPP; break; } case SCTP_ADD_VRF_ID: { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); error = EOPNOTSUPP; break; } case SCTP_DELAYED_SACK: { struct sctp_sack_info *sack; SCTP_CHECK_AND_CAST(sack, optval, struct sctp_sack_info, optsize); SCTP_FIND_STCB(inp, stcb, sack->sack_assoc_id); if (sack->sack_delay) { if (sack->sack_delay > SCTP_MAX_SACK_DELAY) sack->sack_delay = SCTP_MAX_SACK_DELAY; } if (stcb) { if (sack->sack_delay) { if (MSEC_TO_TICKS(sack->sack_delay) < 1) { sack->sack_delay = TICKS_TO_MSEC(1); } stcb->asoc.delayed_ack = sack->sack_delay; } if (sack->sack_freq) { stcb->asoc.sack_freq = sack->sack_freq; } SCTP_TCB_UNLOCK(stcb); } else { SCTP_INP_WLOCK(inp); if (sack->sack_delay) { if (MSEC_TO_TICKS(sack->sack_delay) < 1) { sack->sack_delay = TICKS_TO_MSEC(1); } inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(sack->sack_delay); } if (sack->sack_freq) { inp->sctp_ep.sctp_sack_freq = sack->sack_freq; } SCTP_INP_WUNLOCK(inp); } break; } case SCTP_AUTH_CHUNK: { struct sctp_authchunk *sauth; SCTP_CHECK_AND_CAST(sauth, optval, struct sctp_authchunk, optsize); SCTP_INP_WLOCK(inp); if (sctp_auth_add_chunk(sauth->sauth_chunk, inp->sctp_ep.local_auth_chunks)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_INP_WUNLOCK(inp); break; } case SCTP_AUTH_KEY: { struct sctp_authkey *sca; struct sctp_keyhead *shared_keys; sctp_sharedkey_t *shared_key; sctp_key_t *key = NULL; size_t size; SCTP_CHECK_AND_CAST(sca, optval, struct sctp_authkey, optsize); SCTP_FIND_STCB(inp, stcb, sca->sca_assoc_id); size = optsize - sizeof(*sca); if (stcb) { /* set it on the assoc */ shared_keys = &stcb->asoc.shared_keys; /* clear the cached keys for this key id */ sctp_clear_cachedkeys(stcb, sca->sca_keynumber); /* * create the new shared key and * insert/replace it */ if (size > 0) { key = sctp_set_key(sca->sca_key, (uint32_t) size); if (key == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); error = ENOMEM; SCTP_TCB_UNLOCK(stcb); break; } } shared_key = sctp_alloc_sharedkey(); if (shared_key == NULL) { sctp_free_key(key); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); error = ENOMEM; SCTP_TCB_UNLOCK(stcb); break; } shared_key->key = key; shared_key->keyid = sca->sca_keynumber; sctp_insert_sharedkey(shared_keys, shared_key); SCTP_TCB_UNLOCK(stcb); } else { /* set it on the endpoint */ SCTP_INP_WLOCK(inp); shared_keys = &inp->sctp_ep.shared_keys; /* * clear the cached keys on all assocs for * this key id */ sctp_clear_cachedkeys_ep(inp, sca->sca_keynumber); /* * create the new shared key and * insert/replace it */ if (size > 0) { key = sctp_set_key(sca->sca_key, (uint32_t) size); if (key == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); error = ENOMEM; SCTP_INP_WUNLOCK(inp); break; } } shared_key = sctp_alloc_sharedkey(); if (shared_key == NULL) { sctp_free_key(key); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); error = ENOMEM; SCTP_INP_WUNLOCK(inp); break; } shared_key->key = key; shared_key->keyid = sca->sca_keynumber; sctp_insert_sharedkey(shared_keys, shared_key); SCTP_INP_WUNLOCK(inp); } break; } case SCTP_HMAC_IDENT: { struct sctp_hmacalgo *shmac; sctp_hmaclist_t *hmaclist; - uint32_t hmacid; - size_t size, i, found; + uint16_t hmacid; + uint32_t i; + size_t found; + SCTP_CHECK_AND_CAST(shmac, optval, struct sctp_hmacalgo, optsize); - size = (optsize - sizeof(*shmac)) / sizeof(shmac->shmac_idents[0]); - hmaclist = sctp_alloc_hmaclist(size); + if (optsize < sizeof(struct sctp_hmacalgo) + shmac->shmac_number_of_idents * sizeof(uint16_t)) { + SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); + error = EINVAL; + break; + } + hmaclist = sctp_alloc_hmaclist(shmac->shmac_number_of_idents); if (hmaclist == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); error = ENOMEM; break; } - for (i = 0; i < size; i++) { + for (i = 0; i < shmac->shmac_number_of_idents; i++) { hmacid = shmac->shmac_idents[i]; - if (sctp_auth_add_hmacid(hmaclist, (uint16_t) hmacid)) { + if (sctp_auth_add_hmacid(hmaclist, hmacid)) { /* invalid HMACs were found */ ; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; sctp_free_hmaclist(hmaclist); goto sctp_set_hmac_done; } } found = 0; for (i = 0; i < hmaclist->num_algo; i++) { if (hmaclist->hmac[i] == SCTP_AUTH_HMAC_ID_SHA1) { /* already in list */ found = 1; } } if (!found) { sctp_free_hmaclist(hmaclist); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } /* set it on the endpoint */ SCTP_INP_WLOCK(inp); if (inp->sctp_ep.local_hmacs) sctp_free_hmaclist(inp->sctp_ep.local_hmacs); inp->sctp_ep.local_hmacs = hmaclist; SCTP_INP_WUNLOCK(inp); sctp_set_hmac_done: break; } case SCTP_AUTH_ACTIVE_KEY: { struct sctp_authkeyid *scact; SCTP_CHECK_AND_CAST(scact, optval, struct sctp_authkeyid, optsize); SCTP_FIND_STCB(inp, stcb, scact->scact_assoc_id); /* set the active key on the right place */ if (stcb) { /* set the active key on the assoc */ if (sctp_auth_setactivekey(stcb, scact->scact_keynumber)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_TCB_UNLOCK(stcb); } else { /* set the active key on the endpoint */ SCTP_INP_WLOCK(inp); if (sctp_auth_setactivekey_ep(inp, scact->scact_keynumber)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_INP_WUNLOCK(inp); } break; } case SCTP_AUTH_DELETE_KEY: { struct sctp_authkeyid *scdel; SCTP_CHECK_AND_CAST(scdel, optval, struct sctp_authkeyid, optsize); SCTP_FIND_STCB(inp, stcb, scdel->scact_assoc_id); /* delete the key from the right place */ if (stcb) { if (sctp_delete_sharedkey(stcb, scdel->scact_keynumber)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_TCB_UNLOCK(stcb); } else { SCTP_INP_WLOCK(inp); if (sctp_delete_sharedkey_ep(inp, scdel->scact_keynumber)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_INP_WUNLOCK(inp); } break; } case SCTP_RESET_STREAMS: { struct sctp_stream_reset *strrst; uint8_t send_in = 0, send_tsn = 0, send_out = 0; int i; SCTP_CHECK_AND_CAST(strrst, optval, struct sctp_stream_reset, optsize); SCTP_FIND_STCB(inp, stcb, strrst->strrst_assoc_id); if (stcb == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; break; } if (stcb->asoc.peer_supports_strreset == 0) { /* * Peer does not support it, we return * protocol not supported since this is true * for this feature and this peer, not the * socket request in general. */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EPROTONOSUPPORT); error = EPROTONOSUPPORT; SCTP_TCB_UNLOCK(stcb); break; } if (stcb->asoc.stream_reset_outstanding) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); error = EALREADY; SCTP_TCB_UNLOCK(stcb); break; } if (strrst->strrst_flags == SCTP_RESET_LOCAL_RECV) { send_in = 1; } else if (strrst->strrst_flags == SCTP_RESET_LOCAL_SEND) { send_out = 1; } else if (strrst->strrst_flags == SCTP_RESET_BOTH) { send_in = 1; send_out = 1; } else if (strrst->strrst_flags == SCTP_RESET_TSN) { send_tsn = 1; } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; SCTP_TCB_UNLOCK(stcb); break; } for (i = 0; i < strrst->strrst_num_streams; i++) { if ((send_in) && (strrst->strrst_list[i] > stcb->asoc.streamincnt)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto get_out; } if ((send_out) && (strrst->strrst_list[i] > stcb->asoc.streamoutcnt)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto get_out; } } if (error) { get_out: SCTP_TCB_UNLOCK(stcb); break; } error = sctp_send_str_reset_req(stcb, strrst->strrst_num_streams, strrst->strrst_list, send_out, (stcb->asoc.str_reset_seq_in - 3), send_in, send_tsn); sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_STRRST_REQ, SCTP_SO_LOCKED); SCTP_TCB_UNLOCK(stcb); } break; case SCTP_CONNECT_X: if (optsize < (sizeof(int) + sizeof(struct sockaddr_in))) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } error = sctp_do_connect_x(so, inp, optval, optsize, p, 0); break; case SCTP_CONNECT_X_DELAYED: if (optsize < (sizeof(int) + sizeof(struct sockaddr_in))) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } error = sctp_do_connect_x(so, inp, optval, optsize, p, 1); break; case SCTP_CONNECT_X_COMPLETE: { struct sockaddr *sa; struct sctp_nets *net; /* FIXME MT: check correct? */ SCTP_CHECK_AND_CAST(sa, optval, struct sockaddr, optsize); /* find tcb */ if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { SCTP_INP_RLOCK(inp); stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb) { SCTP_TCB_LOCK(stcb); net = sctp_findnet(stcb, sa); } SCTP_INP_RUNLOCK(inp); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, sa, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if (stcb == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; break; } if (stcb->asoc.delayed_connection == 1) { stcb->asoc.delayed_connection = 0; (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); sctp_timer_stop(SCTP_TIMER_TYPE_INIT, inp, stcb, stcb->asoc.primary_destination, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_9); sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); } else { /* * already expired or did not use delayed * connectx */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); error = EALREADY; } SCTP_TCB_UNLOCK(stcb); } break; case SCTP_MAX_BURST: { uint8_t *burst; SCTP_CHECK_AND_CAST(burst, optval, uint8_t, optsize); SCTP_INP_WLOCK(inp); if (*burst) { inp->sctp_ep.max_burst = *burst; } SCTP_INP_WUNLOCK(inp); } break; case SCTP_MAXSEG: { struct sctp_assoc_value *av; int ovh; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ovh = SCTP_MED_OVERHEAD; } else { ovh = SCTP_MED_V4_OVERHEAD; } if (stcb) { if (av->assoc_value) { stcb->asoc.sctp_frag_point = (av->assoc_value + ovh); } else { stcb->asoc.sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT; } SCTP_TCB_UNLOCK(stcb); } else { SCTP_INP_WLOCK(inp); /* * FIXME MT: I think this is not in tune * with the API ID */ if (av->assoc_value) { inp->sctp_frag_point = (av->assoc_value + ovh); } else { inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT; } SCTP_INP_WUNLOCK(inp); } } break; case SCTP_EVENTS: { struct sctp_event_subscribe *events; SCTP_CHECK_AND_CAST(events, optval, struct sctp_event_subscribe, optsize); SCTP_INP_WLOCK(inp); if (events->sctp_data_io_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT); } if (events->sctp_association_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVASSOCEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVASSOCEVNT); } if (events->sctp_address_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVPADDREVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVPADDREVNT); } if (events->sctp_send_failure_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVSENDFAILEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVSENDFAILEVNT); } if (events->sctp_peer_error_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVPEERERR); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVPEERERR); } if (events->sctp_shutdown_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT); } if (events->sctp_partial_delivery_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_PDAPIEVNT); } if (events->sctp_adaptation_layer_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_ADAPTATIONEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_ADAPTATIONEVNT); } if (events->sctp_authentication_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTHEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTHEVNT); } if (events->sctp_stream_reset_events) { sctp_feature_on(inp, SCTP_PCB_FLAGS_STREAM_RESETEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_STREAM_RESETEVNT); } SCTP_INP_WUNLOCK(inp); } break; case SCTP_ADAPTATION_LAYER: { struct sctp_setadaptation *adap_bits; SCTP_CHECK_AND_CAST(adap_bits, optval, struct sctp_setadaptation, optsize); SCTP_INP_WLOCK(inp); inp->sctp_ep.adaptation_layer_indicator = adap_bits->ssb_adaptation_ind; SCTP_INP_WUNLOCK(inp); } break; #ifdef SCTP_DEBUG case SCTP_SET_INITIAL_DBG_SEQ: { uint32_t *vvv; SCTP_CHECK_AND_CAST(vvv, optval, uint32_t, optsize); SCTP_INP_WLOCK(inp); inp->sctp_ep.initial_sequence_debug = *vvv; SCTP_INP_WUNLOCK(inp); } break; #endif case SCTP_DEFAULT_SEND_PARAM: { struct sctp_sndrcvinfo *s_info; SCTP_CHECK_AND_CAST(s_info, optval, struct sctp_sndrcvinfo, optsize); SCTP_FIND_STCB(inp, stcb, s_info->sinfo_assoc_id); if (stcb) { if (s_info->sinfo_stream <= stcb->asoc.streamoutcnt) { memcpy(&stcb->asoc.def_send, s_info, min(optsize, sizeof(stcb->asoc.def_send))); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_TCB_UNLOCK(stcb); } else { SCTP_INP_WLOCK(inp); memcpy(&inp->def_send, s_info, min(optsize, sizeof(inp->def_send))); SCTP_INP_WUNLOCK(inp); } } break; case SCTP_PEER_ADDR_PARAMS: /* Applys to the specific association */ { struct sctp_paddrparams *paddrp; struct sctp_nets *net; SCTP_CHECK_AND_CAST(paddrp, optval, struct sctp_paddrparams, optsize); SCTP_FIND_STCB(inp, stcb, paddrp->spp_assoc_id); net = NULL; if (stcb) { net = sctp_findnet(stcb, (struct sockaddr *)&paddrp->spp_address); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, (struct sockaddr *)&paddrp->spp_address, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if (stcb && (net == NULL)) { struct sockaddr *sa; sa = (struct sockaddr *)&paddrp->spp_address; if (sa->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)sa; if (sin->sin_addr.s_addr) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); SCTP_TCB_UNLOCK(stcb); error = EINVAL; break; } } else if (sa->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)sa; if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); SCTP_TCB_UNLOCK(stcb); error = EINVAL; break; } } else { error = EAFNOSUPPORT; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } /* sanity checks */ if ((paddrp->spp_flags & SPP_HB_ENABLE) && (paddrp->spp_flags & SPP_HB_DISABLE)) { if (stcb) SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } if ((paddrp->spp_flags & SPP_PMTUD_ENABLE) && (paddrp->spp_flags & SPP_PMTUD_DISABLE)) { if (stcb) SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } if (stcb) { /************************TCB SPECIFIC SET ******************/ /* * do we change the timer for HB, we run * only one? */ int ovh = 0; if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ovh = SCTP_MED_OVERHEAD; } else { ovh = SCTP_MED_V4_OVERHEAD; } if (paddrp->spp_hbinterval) stcb->asoc.heart_beat_delay = paddrp->spp_hbinterval; else if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO) stcb->asoc.heart_beat_delay = 0; /* network sets ? */ if (net) { /************************NET SPECIFIC SET ******************/ if (paddrp->spp_flags & SPP_HB_DEMAND) { /* on demand HB */ if (sctp_send_hb(stcb, 1, net) < 0) { /* asoc destroyed */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } } if (paddrp->spp_flags & SPP_HB_DISABLE) { net->dest_state |= SCTP_ADDR_NOHB; } if (paddrp->spp_flags & SPP_HB_ENABLE) { net->dest_state &= ~SCTP_ADDR_NOHB; } if ((paddrp->spp_flags & SPP_PMTUD_DISABLE) && (paddrp->spp_pathmtu >= SCTP_SMALLEST_PMTU)) { if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_10); } if (paddrp->spp_pathmtu > SCTP_DEFAULT_MINSEGMENT) { net->mtu = paddrp->spp_pathmtu + ovh; if (net->mtu < stcb->asoc.smallest_mtu) { #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("SCTP_PMTU_DISABLE calls sctp_pathmtu_adjustment:%d\n", net->mtu); #endif sctp_pathmtu_adjustment(inp, stcb, net, net->mtu); } } } if (paddrp->spp_flags & SPP_PMTUD_ENABLE) { if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net); } } if (paddrp->spp_pathmaxrxt) net->failure_threshold = paddrp->spp_pathmaxrxt; #ifdef INET if (paddrp->spp_flags & SPP_IPV4_TOS) { if (net->ro._l_addr.sin.sin_family == AF_INET) { net->tos_flowlabel = paddrp->spp_ipv4_tos & 0x000000fc; } } #endif #ifdef INET6 if (paddrp->spp_flags & SPP_IPV6_FLOWLABEL) { if (net->ro._l_addr.sin6.sin6_family == AF_INET6) { net->tos_flowlabel = paddrp->spp_ipv6_flowlabel; } } #endif } else { /************************ASSOC ONLY -- NO NET SPECIFIC SET ******************/ if (paddrp->spp_pathmaxrxt) stcb->asoc.def_net_failure = paddrp->spp_pathmaxrxt; if (paddrp->spp_flags & SPP_HB_ENABLE) { /* Turn back on the timer */ stcb->asoc.hb_is_disabled = 0; sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net); } if ((paddrp->spp_flags & SPP_PMTUD_DISABLE) && (paddrp->spp_pathmtu >= SCTP_SMALLEST_PMTU)) { TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_10); } if (paddrp->spp_pathmtu > SCTP_DEFAULT_MINSEGMENT) { net->mtu = paddrp->spp_pathmtu + ovh; if (net->mtu < stcb->asoc.smallest_mtu) { #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("SCTP_PMTU_DISABLE calls sctp_pathmtu_adjustment:%d\n", net->mtu); #endif sctp_pathmtu_adjustment(inp, stcb, net, net->mtu); } } } } if (paddrp->spp_flags & SPP_PMTUD_ENABLE) { TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net); } } } if (paddrp->spp_flags & SPP_HB_DISABLE) { int cnt_of_unconf = 0; struct sctp_nets *lnet; stcb->asoc.hb_is_disabled = 1; TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { if (lnet->dest_state & SCTP_ADDR_UNCONFIRMED) { cnt_of_unconf++; } } /* * stop the timer ONLY if we * have no unconfirmed * addresses */ if (cnt_of_unconf == 0) { TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_11); } } } if (paddrp->spp_flags & SPP_HB_ENABLE) { /* start up the timer. */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net); } } #ifdef INET if (paddrp->spp_flags & SPP_IPV4_TOS) stcb->asoc.default_tos = paddrp->spp_ipv4_tos & 0x000000fc; #endif #ifdef INET6 if (paddrp->spp_flags & SPP_IPV6_FLOWLABEL) stcb->asoc.default_flowlabel = paddrp->spp_ipv6_flowlabel; #endif } SCTP_TCB_UNLOCK(stcb); } else { /************************NO TCB, SET TO default stuff ******************/ SCTP_INP_WLOCK(inp); /* * For the TOS/FLOWLABEL stuff you set it * with the options on the socket */ if (paddrp->spp_pathmaxrxt) { inp->sctp_ep.def_net_failure = paddrp->spp_pathmaxrxt; } if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO) inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = 0; else if (paddrp->spp_hbinterval) { if (paddrp->spp_hbinterval > SCTP_MAX_HB_INTERVAL) paddrp->spp_hbinterval = SCTP_MAX_HB_INTERVAL; inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(paddrp->spp_hbinterval); } if (paddrp->spp_flags & SPP_HB_ENABLE) { sctp_feature_off(inp, SCTP_PCB_FLAGS_DONOT_HEARTBEAT); } else if (paddrp->spp_flags & SPP_HB_DISABLE) { sctp_feature_on(inp, SCTP_PCB_FLAGS_DONOT_HEARTBEAT); } SCTP_INP_WUNLOCK(inp); } } break; case SCTP_RTOINFO: { struct sctp_rtoinfo *srto; uint32_t new_init, new_min, new_max; SCTP_CHECK_AND_CAST(srto, optval, struct sctp_rtoinfo, optsize); SCTP_FIND_STCB(inp, stcb, srto->srto_assoc_id); if (stcb) { if (srto->srto_initial) new_init = srto->srto_initial; else new_init = stcb->asoc.initial_rto; if (srto->srto_max) new_max = srto->srto_max; else new_max = stcb->asoc.maxrto; if (srto->srto_min) new_min = srto->srto_min; else new_min = stcb->asoc.minrto; if ((new_min <= new_init) && (new_init <= new_max)) { stcb->asoc.initial_rto = new_init; stcb->asoc.maxrto = new_max; stcb->asoc.minrto = new_min; } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_TCB_UNLOCK(stcb); } else { SCTP_INP_WLOCK(inp); if (srto->srto_initial) new_init = srto->srto_initial; else new_init = inp->sctp_ep.initial_rto; if (srto->srto_max) new_max = srto->srto_max; else new_max = inp->sctp_ep.sctp_maxrto; if (srto->srto_min) new_min = srto->srto_min; else new_min = inp->sctp_ep.sctp_minrto; if ((new_min <= new_init) && (new_init <= new_max)) { inp->sctp_ep.initial_rto = new_init; inp->sctp_ep.sctp_maxrto = new_max; inp->sctp_ep.sctp_minrto = new_min; } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_INP_WUNLOCK(inp); } } break; case SCTP_ASSOCINFO: { struct sctp_assocparams *sasoc; SCTP_CHECK_AND_CAST(sasoc, optval, struct sctp_assocparams, optsize); SCTP_FIND_STCB(inp, stcb, sasoc->sasoc_assoc_id); if (sasoc->sasoc_cookie_life) { /* boundary check the cookie life */ if (sasoc->sasoc_cookie_life < 1000) sasoc->sasoc_cookie_life = 1000; if (sasoc->sasoc_cookie_life > SCTP_MAX_COOKIE_LIFE) { sasoc->sasoc_cookie_life = SCTP_MAX_COOKIE_LIFE; } } if (stcb) { if (sasoc->sasoc_asocmaxrxt) stcb->asoc.max_send_times = sasoc->sasoc_asocmaxrxt; sasoc->sasoc_number_peer_destinations = stcb->asoc.numnets; sasoc->sasoc_peer_rwnd = 0; sasoc->sasoc_local_rwnd = 0; if (sasoc->sasoc_cookie_life) { stcb->asoc.cookie_life = MSEC_TO_TICKS(sasoc->sasoc_cookie_life); } SCTP_TCB_UNLOCK(stcb); } else { SCTP_INP_WLOCK(inp); if (sasoc->sasoc_asocmaxrxt) inp->sctp_ep.max_send_times = sasoc->sasoc_asocmaxrxt; sasoc->sasoc_number_peer_destinations = 0; sasoc->sasoc_peer_rwnd = 0; sasoc->sasoc_local_rwnd = 0; if (sasoc->sasoc_cookie_life) { inp->sctp_ep.def_cookie_life = MSEC_TO_TICKS(sasoc->sasoc_cookie_life); } SCTP_INP_WUNLOCK(inp); } } break; case SCTP_INITMSG: { struct sctp_initmsg *sinit; SCTP_CHECK_AND_CAST(sinit, optval, struct sctp_initmsg, optsize); SCTP_INP_WLOCK(inp); if (sinit->sinit_num_ostreams) inp->sctp_ep.pre_open_stream_count = sinit->sinit_num_ostreams; if (sinit->sinit_max_instreams) inp->sctp_ep.max_open_streams_intome = sinit->sinit_max_instreams; if (sinit->sinit_max_attempts) inp->sctp_ep.max_init_times = sinit->sinit_max_attempts; if (sinit->sinit_max_init_timeo) inp->sctp_ep.initial_init_rto_max = sinit->sinit_max_init_timeo; SCTP_INP_WUNLOCK(inp); } break; case SCTP_PRIMARY_ADDR: { struct sctp_setprim *spa; struct sctp_nets *net, *lnet; SCTP_CHECK_AND_CAST(spa, optval, struct sctp_setprim, optsize); SCTP_FIND_STCB(inp, stcb, spa->ssp_assoc_id); net = NULL; if (stcb) { net = sctp_findnet(stcb, (struct sockaddr *)&spa->ssp_addr); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, (struct sockaddr *)&spa->ssp_addr, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if ((stcb) && (net)) { if ((net != stcb->asoc.primary_destination) && (!(net->dest_state & SCTP_ADDR_UNCONFIRMED))) { /* Ok we need to set it */ lnet = stcb->asoc.primary_destination; if (sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, net) == 0) { if (net->dest_state & SCTP_ADDR_SWITCH_PRIMARY) { net->dest_state |= SCTP_ADDR_DOUBLE_SWITCH; } net->dest_state |= SCTP_ADDR_SWITCH_PRIMARY; } } } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } if (stcb) { SCTP_TCB_UNLOCK(stcb); } } break; case SCTP_SET_DYNAMIC_PRIMARY: { union sctp_sockstore *ss; error = priv_check(curthread, PRIV_NETINET_RESERVEDPORT); if (error) break; SCTP_CHECK_AND_CAST(ss, optval, union sctp_sockstore, optsize); /* SUPER USER CHECK? */ error = sctp_dynamic_set_primary(&ss->sa, vrf_id); } break; case SCTP_SET_PEER_PRIMARY_ADDR: { struct sctp_setpeerprim *sspp; SCTP_CHECK_AND_CAST(sspp, optval, struct sctp_setpeerprim, optsize); SCTP_FIND_STCB(inp, stcb, sspp->sspp_assoc_id); if (stcb != NULL) { struct sctp_ifa *ifa; ifa = sctp_find_ifa_by_addr((struct sockaddr *)&sspp->sspp_addr, stcb->asoc.vrf_id, SCTP_ADDR_NOT_LOCKED); if (ifa == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto out_of_it; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { /* * Must validate the ifa found is in * our ep */ struct sctp_laddr *laddr; int found = 0; LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "%s: NULL ifa\n", __FUNCTION__); continue; } if (laddr->ifa == ifa) { found = 1; break; } } if (!found) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto out_of_it; } } if (sctp_set_primary_ip_address_sa(stcb, (struct sockaddr *)&sspp->sspp_addr) != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } out_of_it: SCTP_TCB_UNLOCK(stcb); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; case SCTP_BINDX_ADD_ADDR: { struct sctp_getaddresses *addrs; size_t sz; struct thread *td; int prison = 0; td = (struct thread *)p; if (jailed(td->td_ucred)) { prison = 1; } SCTP_CHECK_AND_CAST(addrs, optval, struct sctp_getaddresses, optsize); if (addrs->addr->sa_family == AF_INET) { sz = sizeof(struct sctp_getaddresses) - sizeof(struct sockaddr) + sizeof(struct sockaddr_in); if (optsize < sz) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } if (prison && prison_ip(td->td_ucred, 0, &(((struct sockaddr_in *)(addrs->addr))->sin_addr.s_addr))) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EADDRNOTAVAIL); error = EADDRNOTAVAIL; } } else if (addrs->addr->sa_family == AF_INET6) { sz = sizeof(struct sctp_getaddresses) - sizeof(struct sockaddr) + sizeof(struct sockaddr_in6); if (optsize < sz) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } /* JAIL XXXX Add else here for V6 */ } sctp_bindx_add_address(so, inp, addrs->addr, addrs->sget_assoc_id, vrf_id, &error, p); } break; case SCTP_BINDX_REM_ADDR: { struct sctp_getaddresses *addrs; size_t sz; struct thread *td; int prison = 0; td = (struct thread *)p; if (jailed(td->td_ucred)) { prison = 1; } SCTP_CHECK_AND_CAST(addrs, optval, struct sctp_getaddresses, optsize); if (addrs->addr->sa_family == AF_INET) { sz = sizeof(struct sctp_getaddresses) - sizeof(struct sockaddr) + sizeof(struct sockaddr_in); if (optsize < sz) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } if (prison && prison_ip(td->td_ucred, 0, &(((struct sockaddr_in *)(addrs->addr))->sin_addr.s_addr))) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EADDRNOTAVAIL); error = EADDRNOTAVAIL; } } else if (addrs->addr->sa_family == AF_INET6) { sz = sizeof(struct sctp_getaddresses) - sizeof(struct sockaddr) + sizeof(struct sockaddr_in6); if (optsize < sz) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } /* JAIL XXXX Add else here for V6 */ } sctp_bindx_delete_address(so, inp, addrs->addr, addrs->sget_assoc_id, vrf_id, &error); } break; default: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT); error = ENOPROTOOPT; break; } /* end switch (opt) */ return (error); } int sctp_ctloutput(struct socket *so, struct sockopt *sopt) { void *optval = NULL; size_t optsize = 0; struct sctp_inpcb *inp; void *p; int error = 0; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == 0) { /* I made the same as TCP since we are not setup? */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (ECONNRESET); } if (sopt->sopt_level != IPPROTO_SCTP) { /* wrong proto level... send back up to IP */ #ifdef INET6 if (INP_CHECK_SOCKAF(so, AF_INET6)) error = ip6_ctloutput(so, sopt); else #endif /* INET6 */ error = ip_ctloutput(so, sopt); return (error); } optsize = sopt->sopt_valsize; if (optsize) { SCTP_MALLOC(optval, void *, optsize, SCTP_M_SOCKOPT); if (optval == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOBUFS); return (ENOBUFS); } error = sooptcopyin(sopt, optval, optsize, optsize); if (error) { SCTP_FREE(optval, SCTP_M_SOCKOPT); goto out; } } p = (void *)sopt->sopt_td; if (sopt->sopt_dir == SOPT_SET) { error = sctp_setopt(so, sopt->sopt_name, optval, optsize, p); } else if (sopt->sopt_dir == SOPT_GET) { error = sctp_getopt(so, sopt->sopt_name, optval, &optsize, p); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } if ((error == 0) && (optval != NULL)) { error = sooptcopyout(sopt, optval, optsize); SCTP_FREE(optval, SCTP_M_SOCKOPT); } else if (optval != NULL) { SCTP_FREE(optval, SCTP_M_SOCKOPT); } out: return (error); } static int sctp_connect(struct socket *so, struct sockaddr *addr, struct thread *p) { int error = 0; int create_lock_on = 0; uint32_t vrf_id; struct sctp_inpcb *inp; struct sctp_tcb *stcb = NULL; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == 0) { /* I made the same as TCP since we are not setup? */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (ECONNRESET); } if (addr == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return EINVAL; } if ((addr->sa_family == AF_INET6) && (addr->sa_len != sizeof(struct sockaddr_in6))) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } if ((addr->sa_family == AF_INET) && (addr->sa_len != sizeof(struct sockaddr_in))) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } SCTP_INP_INCR_REF(inp); SCTP_ASOC_CREATE_LOCK(inp); create_lock_on = 1; if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { /* Should I really unlock ? */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EFAULT); error = EFAULT; goto out_now; } #ifdef INET6 if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) && (addr->sa_family == AF_INET6)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto out_now; } #endif /* INET6 */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == SCTP_PCB_FLAGS_UNBOUND) { /* Bind a ephemeral port */ error = sctp_inpcb_bind(so, NULL, NULL, p); if (error) { goto out_now; } } /* Now do we connect? */ - if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) { + if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) && + (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE))) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto out_now; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) { /* We are already connected AND the TCP model */ SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EADDRINUSE); error = EADDRINUSE; goto out_now; } if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { SCTP_INP_RLOCK(inp); stcb = LIST_FIRST(&inp->sctp_asoc_list); SCTP_INP_RUNLOCK(inp); } else { /* * We increment here since sctp_findassociation_ep_addr() - * wil do a decrement if it finds the stcb as long as the + * will do a decrement if it finds the stcb as long as the * locked tcb (last argument) is NOT a TCB.. aka NULL. */ SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, addr, NULL, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } else { SCTP_TCB_UNLOCK(stcb); } } if (stcb != NULL) { /* Already have or am bring up an association */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); error = EALREADY; goto out_now; } vrf_id = inp->def_vrf_id; /* We are GOOD to go */ stcb = sctp_aloc_assoc(inp, addr, 1, &error, 0, vrf_id, p); if (stcb == NULL) { /* Gak! no memory */ goto out_now; } if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) { stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED; /* Set the connected flag so we can queue data */ soisconnecting(so); } SCTP_SET_STATE(&stcb->asoc, SCTP_STATE_COOKIE_WAIT); (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); /* initialize authentication parameters for the assoc */ sctp_initialize_auth_params(inp, stcb); sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); SCTP_TCB_UNLOCK(stcb); out_now: if (create_lock_on) { SCTP_ASOC_CREATE_UNLOCK(inp); } SCTP_INP_DECR_REF(inp); return error; } int sctp_listen(struct socket *so, int backlog, struct thread *p) { /* * Note this module depends on the protocol processing being called * AFTER any socket level flags and backlog are applied to the * socket. The traditional way that the socket flags are applied is * AFTER protocol processing. We have made a change to the * sys/kern/uipc_socket.c module to reverse this but this MUST be in * place if the socket API for SCTP is to work properly. */ int error = 0; struct sctp_inpcb *inp; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == 0) { /* I made the same as TCP since we are not setup? */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (ECONNRESET); } + if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) { + /* See if we have a listener */ + struct sctp_inpcb *tinp; + union sctp_sockstore store, *sp; + + sp = &store; + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { + /* not bound all */ + struct sctp_laddr *laddr; + + LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { + memcpy(&store, &laddr->ifa->address, sizeof(store)); + sp->sin.sin_port = inp->sctp_lport; + tinp = sctp_pcb_findep(&sp->sa, 0, 0, inp->def_vrf_id); + if (tinp && (tinp != inp) && + ((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) == 0) && + ((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && + (tinp->sctp_socket->so_qlimit)) { + /* + * we have a listener already and + * its not this inp. + */ + SCTP_INP_DECR_REF(tinp); + return (EADDRINUSE); + } else if (tinp) { + SCTP_INP_DECR_REF(tinp); + } + } + } else { + /* Setup a local addr bound all */ + memset(&store, 0, sizeof(store)); + store.sin.sin_port = inp->sctp_lport; +#ifdef INET6 + if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { + store.sa.sa_family = AF_INET6; + store.sa.sa_len = sizeof(struct sockaddr_in6); + } +#endif + if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { + store.sa.sa_family = AF_INET; + store.sa.sa_len = sizeof(struct sockaddr_in); + } + tinp = sctp_pcb_findep(&sp->sa, 0, 0, inp->def_vrf_id); + if (tinp && (tinp != inp) && + ((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) == 0) && + ((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && + (tinp->sctp_socket->so_qlimit)) { + /* + * we have a listener already and its not + * this inp. + */ + SCTP_INP_DECR_REF(tinp); + return (EADDRINUSE); + } else if (tinp) { + SCTP_INP_DECR_REF(inp); + } + } + } SCTP_INP_RLOCK(inp); #ifdef SCTP_LOCK_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) { sctp_log_lock(inp, (struct sctp_tcb *)NULL, SCTP_LOG_LOCK_SOCK); } #endif SOCK_LOCK(so); error = solisten_proto_check(so); if (error) { SOCK_UNLOCK(so); SCTP_INP_RUNLOCK(inp); return (error); } + if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && + (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { + /* + * The unlucky case - We are in the tcp pool with this guy. + * - Someone else is in the main inp slot. - We must move + * this guy (the listener) to the main slot - We must then + * move the guy that was listener to the TCP Pool. + */ + if (sctp_swap_inpcb_for_listen(inp)) { + goto in_use; + } + } if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) { /* We are already connected AND the TCP model */ +in_use: SCTP_INP_RUNLOCK(inp); SOCK_UNLOCK(so); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EADDRINUSE); return (EADDRINUSE); } + SCTP_INP_RUNLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { /* We must do a bind. */ SOCK_UNLOCK(so); - SCTP_INP_RUNLOCK(inp); if ((error = sctp_inpcb_bind(so, NULL, NULL, p))) { /* bind error, probably perm */ return (error); } SOCK_LOCK(so); - } else { - if (backlog != 0) { - inp->sctp_flags |= SCTP_PCB_FLAGS_LISTENING; - } else { - inp->sctp_flags &= ~SCTP_PCB_FLAGS_LISTENING; - } - SCTP_INP_RUNLOCK(inp); } /* It appears for 7.0 and on, we must always call this. */ solisten_proto(so, backlog); if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) { /* remove the ACCEPTCONN flag for one-to-many sockets */ so->so_options &= ~SO_ACCEPTCONN; } if (backlog == 0) { /* turning off listen */ so->so_options &= ~SO_ACCEPTCONN; } SOCK_UNLOCK(so); return (error); } static int sctp_defered_wakeup_cnt = 0; int sctp_accept(struct socket *so, struct sockaddr **addr) { struct sctp_tcb *stcb; struct sctp_inpcb *inp; union sctp_sockstore store; #ifdef INET6 int error; #endif inp = (struct sctp_inpcb *)so->so_pcb; if (inp == 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (ECONNRESET); } SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) { SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); return (EOPNOTSUPP); } if (so->so_state & SS_ISDISCONNECTED) { SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ECONNABORTED); return (ECONNABORTED); } stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (ECONNRESET); } SCTP_TCB_LOCK(stcb); SCTP_INP_RUNLOCK(inp); store = stcb->asoc.primary_destination->ro._l_addr; SCTP_TCB_UNLOCK(stcb); switch (store.sa.sa_family) { case AF_INET: { struct sockaddr_in *sin; SCTP_MALLOC_SONAME(sin, struct sockaddr_in *, sizeof *sin); sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_port = ((struct sockaddr_in *)&store)->sin_port; sin->sin_addr = ((struct sockaddr_in *)&store)->sin_addr; *addr = (struct sockaddr *)sin; break; } #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6; SCTP_MALLOC_SONAME(sin6, struct sockaddr_in6 *, sizeof *sin6); sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(*sin6); sin6->sin6_port = ((struct sockaddr_in6 *)&store)->sin6_port; sin6->sin6_addr = ((struct sockaddr_in6 *)&store)->sin6_addr; if ((error = sa6_recoverscope(sin6)) != 0) { SCTP_FREE_SONAME(sin6); return (error); } *addr = (struct sockaddr *)sin6; break; } #endif default: /* TSNH */ break; } /* Wake any delayed sleep action */ if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) { SCTP_INP_WLOCK(inp); inp->sctp_flags &= ~SCTP_PCB_FLAGS_DONT_WAKE; if (inp->sctp_flags & SCTP_PCB_FLAGS_WAKEOUTPUT) { inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT; SCTP_INP_WUNLOCK(inp); SOCKBUF_LOCK(&inp->sctp_socket->so_snd); if (sowriteable(inp->sctp_socket)) { sowwakeup_locked(inp->sctp_socket); } else { SOCKBUF_UNLOCK(&inp->sctp_socket->so_snd); } SCTP_INP_WLOCK(inp); } if (inp->sctp_flags & SCTP_PCB_FLAGS_WAKEINPUT) { inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT; SCTP_INP_WUNLOCK(inp); SOCKBUF_LOCK(&inp->sctp_socket->so_rcv); if (soreadable(inp->sctp_socket)) { sctp_defered_wakeup_cnt++; sorwakeup_locked(inp->sctp_socket); } else { SOCKBUF_UNLOCK(&inp->sctp_socket->so_rcv); } SCTP_INP_WLOCK(inp); } SCTP_INP_WUNLOCK(inp); } return (0); } int sctp_ingetaddr(struct socket *so, struct sockaddr **addr) { struct sockaddr_in *sin; uint32_t vrf_id; struct sctp_inpcb *inp; struct sctp_ifa *sctp_ifa; /* * Do the malloc first in case it blocks. */ SCTP_MALLOC_SONAME(sin, struct sockaddr_in *, sizeof *sin); sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); inp = (struct sctp_inpcb *)so->so_pcb; if (!inp) { SCTP_FREE_SONAME(sin); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return ECONNRESET; } SCTP_INP_RLOCK(inp); sin->sin_port = inp->sctp_lport; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { struct sctp_tcb *stcb; struct sockaddr_in *sin_a; struct sctp_nets *net; int fnd; stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { goto notConn; } fnd = 0; sin_a = NULL; SCTP_TCB_LOCK(stcb); TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { sin_a = (struct sockaddr_in *)&net->ro._l_addr; if (sin_a == NULL) /* this will make coverity happy */ continue; if (sin_a->sin_family == AF_INET) { fnd = 1; break; } } if ((!fnd) || (sin_a == NULL)) { /* punt */ SCTP_TCB_UNLOCK(stcb); goto notConn; } vrf_id = inp->def_vrf_id; sctp_ifa = sctp_source_address_selection(inp, stcb, (sctp_route_t *) & net->ro, net, 0, vrf_id); if (sctp_ifa) { sin->sin_addr = sctp_ifa->address.sin.sin_addr; sctp_free_ifa(sctp_ifa); } SCTP_TCB_UNLOCK(stcb); } else { /* For the bound all case you get back 0 */ notConn: sin->sin_addr.s_addr = 0; } } else { /* Take the first IPv4 address in the list */ struct sctp_laddr *laddr; int fnd = 0; LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa->address.sa.sa_family == AF_INET) { struct sockaddr_in *sin_a; sin_a = (struct sockaddr_in *)&laddr->ifa->address.sa; sin->sin_addr = sin_a->sin_addr; fnd = 1; break; } } if (!fnd) { SCTP_FREE_SONAME(sin); SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); return ENOENT; } } SCTP_INP_RUNLOCK(inp); (*addr) = (struct sockaddr *)sin; return (0); } int sctp_peeraddr(struct socket *so, struct sockaddr **addr) { struct sockaddr_in *sin = (struct sockaddr_in *)*addr; int fnd; struct sockaddr_in *sin_a; struct sctp_inpcb *inp; struct sctp_tcb *stcb; struct sctp_nets *net; /* Do the malloc first in case it blocks. */ inp = (struct sctp_inpcb *)so->so_pcb; if ((inp == NULL) || ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) == 0)) { /* UDP type and listeners will drop out here */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); return (ENOTCONN); } SCTP_MALLOC_SONAME(sin, struct sockaddr_in *, sizeof *sin); sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); /* We must recapture incase we blocked */ inp = (struct sctp_inpcb *)so->so_pcb; if (!inp) { SCTP_FREE_SONAME(sin); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return ECONNRESET; } SCTP_INP_RLOCK(inp); stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb) { SCTP_TCB_LOCK(stcb); } SCTP_INP_RUNLOCK(inp); if (stcb == NULL) { SCTP_FREE_SONAME(sin); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return ECONNRESET; } fnd = 0; TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { sin_a = (struct sockaddr_in *)&net->ro._l_addr; if (sin_a->sin_family == AF_INET) { fnd = 1; sin->sin_port = stcb->rport; sin->sin_addr = sin_a->sin_addr; break; } } SCTP_TCB_UNLOCK(stcb); if (!fnd) { /* No IPv4 address */ SCTP_FREE_SONAME(sin); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); return ENOENT; } (*addr) = (struct sockaddr *)sin; return (0); } struct pr_usrreqs sctp_usrreqs = { .pru_abort = sctp_abort, .pru_accept = sctp_accept, .pru_attach = sctp_attach, .pru_bind = sctp_bind, .pru_connect = sctp_connect, .pru_control = in_control, .pru_close = sctp_close, .pru_detach = sctp_close, .pru_sopoll = sopoll_generic, .pru_flush = sctp_flush, .pru_disconnect = sctp_disconnect, .pru_listen = sctp_listen, .pru_peeraddr = sctp_peeraddr, .pru_send = sctp_sendm, .pru_shutdown = sctp_shutdown, .pru_sockaddr = sctp_ingetaddr, .pru_sosend = sctp_sosend, .pru_soreceive = sctp_soreceive }; Index: head/sys/netinet/sctp_var.h =================================================================== --- head/sys/netinet/sctp_var.h (revision 181053) +++ head/sys/netinet/sctp_var.h (revision 181054) @@ -1,338 +1,339 @@ /*- * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctp_var.h,v 1.24 2005/03/06 16:04:19 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #ifndef _NETINET_SCTP_VAR_H_ #define _NETINET_SCTP_VAR_H_ #include #if defined(_KERNEL) || defined(__Userspace__) extern struct pr_usrreqs sctp_usrreqs; #define sctp_feature_on(inp, feature) (inp->sctp_features |= feature) #define sctp_feature_off(inp, feature) (inp->sctp_features &= ~feature) -#define sctp_is_feature_on(inp, feature) (inp->sctp_features & feature) +#define sctp_is_feature_on(inp, feature) ((inp->sctp_features & feature) == feature) #define sctp_is_feature_off(inp, feature) ((inp->sctp_features & feature) == 0) + /* managing mobility_feature in inpcb (by micchie) */ #define sctp_mobility_feature_on(inp, feature) (inp->sctp_mobility_features |= feature) #define sctp_mobility_feature_off(inp, feature) (inp->sctp_mobility_features &= ~feature) #define sctp_is_mobility_feature_on(inp, feature) (inp->sctp_mobility_features & feature) #define sctp_is_mobility_feature_off(inp, feature) ((inp->sctp_mobility_features & feature) == 0) #define sctp_maxspace(sb) (max((sb)->sb_hiwat,SCTP_MINIMAL_RWND)) #define sctp_sbspace(asoc, sb) ((long) ((sctp_maxspace(sb) > (asoc)->sb_cc) ? (sctp_maxspace(sb) - (asoc)->sb_cc) : 0)) #define sctp_sbspace_failedmsgs(sb) ((long) ((sctp_maxspace(sb) > (sb)->sb_cc) ? (sctp_maxspace(sb) - (sb)->sb_cc) : 0)) #define sctp_sbspace_sub(a,b) ((a > b) ? (a - b) : 0) /* * I tried to cache the readq entries at one point. But the reality * is that it did not add any performance since this meant we had to * lock the STCB on read. And at that point once you have to do an * extra lock, it really does not matter if the lock is in the ZONE * stuff or in our code. Note that this same problem would occur with * an mbuf cache as well so it is not really worth doing, at least * right now :-D */ #define sctp_free_a_readq(_stcb, _readq) { \ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), (_readq)); \ SCTP_DECR_READQ_COUNT(); \ } #define sctp_alloc_a_readq(_stcb, _readq) { \ (_readq) = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_readq), struct sctp_queued_to_read); \ if ((_readq)) { \ SCTP_INCR_READQ_COUNT(); \ } \ } #define sctp_free_a_strmoq(_stcb, _strmoq) { \ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_strmoq), (_strmoq)); \ SCTP_DECR_STRMOQ_COUNT(); \ } #define sctp_alloc_a_strmoq(_stcb, _strmoq) { \ (_strmoq) = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_strmoq), struct sctp_stream_queue_pending); \ if ((_strmoq)) { \ SCTP_INCR_STRMOQ_COUNT(); \ } \ } #define sctp_free_a_chunk(_stcb, _chk) { \ if(_stcb) { \ SCTP_TCB_LOCK_ASSERT((_stcb)); \ if ((_chk)->whoTo) { \ sctp_free_remote_addr((_chk)->whoTo); \ (_chk)->whoTo = NULL; \ } \ if (((_stcb)->asoc.free_chunk_cnt > SCTP_BASE_SYSCTL(sctp_asoc_free_resc_limit)) || \ (SCTP_BASE_INFO(ipi_free_chunks) > SCTP_BASE_SYSCTL(sctp_system_free_resc_limit))) { \ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), (_chk)); \ SCTP_DECR_CHK_COUNT(); \ } else { \ TAILQ_INSERT_TAIL(&(_stcb)->asoc.free_chunks, (_chk), sctp_next); \ (_stcb)->asoc.free_chunk_cnt++; \ atomic_add_int(&SCTP_BASE_INFO(ipi_free_chunks), 1); \ } \ } else { \ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), (_chk)); \ SCTP_DECR_CHK_COUNT(); \ } \ } #define sctp_alloc_a_chunk(_stcb, _chk) { \ if (TAILQ_EMPTY(&(_stcb)->asoc.free_chunks)) { \ (_chk) = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_chunk), struct sctp_tmit_chunk); \ if ((_chk)) { \ SCTP_INCR_CHK_COUNT(); \ (_chk)->whoTo = NULL; \ } \ } else { \ (_chk) = TAILQ_FIRST(&(_stcb)->asoc.free_chunks); \ TAILQ_REMOVE(&(_stcb)->asoc.free_chunks, (_chk), sctp_next); \ atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1); \ SCTP_STAT_INCR(sctps_cached_chk); \ (_stcb)->asoc.free_chunk_cnt--; \ } \ } #define sctp_free_remote_addr(__net) { \ if ((__net)) { \ if (atomic_fetchadd_int(&(__net)->ref_count, -1) == 1) { \ (void)SCTP_OS_TIMER_STOP(&(__net)->rxt_timer.timer); \ (void)SCTP_OS_TIMER_STOP(&(__net)->pmtu_timer.timer); \ (void)SCTP_OS_TIMER_STOP(&(__net)->fr_timer.timer); \ if ((__net)->ro.ro_rt) { \ RTFREE((__net)->ro.ro_rt); \ (__net)->ro.ro_rt = NULL; \ } \ if ((__net)->src_addr_selected) { \ sctp_free_ifa((__net)->ro._s_addr); \ (__net)->ro._s_addr = NULL; \ } \ (__net)->src_addr_selected = 0; \ (__net)->dest_state = SCTP_ADDR_NOT_REACHABLE; \ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_net), (__net)); \ SCTP_DECR_RADDR_COUNT(); \ } \ } \ } #define sctp_sbfree(ctl, stcb, sb, m) { \ uint32_t val; \ val = atomic_fetchadd_int(&(sb)->sb_cc,-(SCTP_BUF_LEN((m)))); \ if (val < SCTP_BUF_LEN((m))) { \ panic("sb_cc goes negative"); \ } \ val = atomic_fetchadd_int(&(sb)->sb_mbcnt,-(MSIZE)); \ if (val < MSIZE) { \ panic("sb_mbcnt goes negative"); \ } \ if (((ctl)->do_not_ref_stcb == 0) && stcb) {\ val = atomic_fetchadd_int(&(stcb)->asoc.sb_cc,-(SCTP_BUF_LEN((m)))); \ if (val < SCTP_BUF_LEN((m))) {\ panic("stcb->sb_cc goes negative"); \ } \ val = atomic_fetchadd_int(&(stcb)->asoc.my_rwnd_control_len,-(MSIZE)); \ if (val < MSIZE) { \ panic("asoc->mbcnt goes negative"); \ } \ } \ if (SCTP_BUF_TYPE(m) != MT_DATA && SCTP_BUF_TYPE(m) != MT_HEADER && \ SCTP_BUF_TYPE(m) != MT_OOBDATA) \ atomic_subtract_int(&(sb)->sb_ctl,SCTP_BUF_LEN((m))); \ } #define sctp_sballoc(stcb, sb, m) { \ atomic_add_int(&(sb)->sb_cc,SCTP_BUF_LEN((m))); \ atomic_add_int(&(sb)->sb_mbcnt, MSIZE); \ if (stcb) { \ atomic_add_int(&(stcb)->asoc.sb_cc,SCTP_BUF_LEN((m))); \ atomic_add_int(&(stcb)->asoc.my_rwnd_control_len, MSIZE); \ } \ if (SCTP_BUF_TYPE(m) != MT_DATA && SCTP_BUF_TYPE(m) != MT_HEADER && \ SCTP_BUF_TYPE(m) != MT_OOBDATA) \ atomic_add_int(&(sb)->sb_ctl,SCTP_BUF_LEN((m))); \ } #define sctp_ucount_incr(val) { \ val++; \ } #define sctp_ucount_decr(val) { \ if (val > 0) { \ val--; \ } else { \ val = 0; \ } \ } #define sctp_mbuf_crush(data) do { \ struct mbuf *_m; \ _m = (data); \ while(_m && (SCTP_BUF_LEN(_m) == 0)) { \ (data) = SCTP_BUF_NEXT(_m); \ SCTP_BUF_NEXT(_m) = NULL; \ sctp_m_free(_m); \ _m = (data); \ } \ } while (0) #define sctp_flight_size_decrease(tp1) do { \ if (tp1->whoTo->flight_size >= tp1->book_size) \ tp1->whoTo->flight_size -= tp1->book_size; \ else \ tp1->whoTo->flight_size = 0; \ } while (0) #define sctp_flight_size_increase(tp1) do { \ (tp1)->whoTo->flight_size += (tp1)->book_size; \ } while (0) #ifdef SCTP_FS_SPEC_LOG #define sctp_total_flight_decrease(stcb, tp1) do { \ if(stcb->asoc.fs_index > SCTP_FS_SPEC_LOG_SIZE) \ stcb->asoc.fs_index = 0;\ stcb->asoc.fslog[stcb->asoc.fs_index].total_flight = stcb->asoc.total_flight; \ stcb->asoc.fslog[stcb->asoc.fs_index].tsn = tp1->rec.data.TSN_seq; \ stcb->asoc.fslog[stcb->asoc.fs_index].book = tp1->book_size; \ stcb->asoc.fslog[stcb->asoc.fs_index].sent = tp1->sent; \ stcb->asoc.fslog[stcb->asoc.fs_index].incr = 0; \ stcb->asoc.fslog[stcb->asoc.fs_index].decr = 1; \ stcb->asoc.fs_index++; \ tp1->window_probe = 0; \ if (stcb->asoc.total_flight >= tp1->book_size) { \ stcb->asoc.total_flight -= tp1->book_size; \ if (stcb->asoc.total_flight_count > 0) \ stcb->asoc.total_flight_count--; \ } else { \ stcb->asoc.total_flight = 0; \ stcb->asoc.total_flight_count = 0; \ } \ } while (0) #define sctp_total_flight_increase(stcb, tp1) do { \ if(stcb->asoc.fs_index > SCTP_FS_SPEC_LOG_SIZE) \ stcb->asoc.fs_index = 0;\ stcb->asoc.fslog[stcb->asoc.fs_index].total_flight = stcb->asoc.total_flight; \ stcb->asoc.fslog[stcb->asoc.fs_index].tsn = tp1->rec.data.TSN_seq; \ stcb->asoc.fslog[stcb->asoc.fs_index].book = tp1->book_size; \ stcb->asoc.fslog[stcb->asoc.fs_index].sent = tp1->sent; \ stcb->asoc.fslog[stcb->asoc.fs_index].incr = 1; \ stcb->asoc.fslog[stcb->asoc.fs_index].decr = 0; \ stcb->asoc.fs_index++; \ (stcb)->asoc.total_flight_count++; \ (stcb)->asoc.total_flight += (tp1)->book_size; \ } while (0) #else #define sctp_total_flight_decrease(stcb, tp1) do { \ if (stcb->asoc.total_flight >= tp1->book_size) { \ stcb->asoc.total_flight -= tp1->book_size; \ if (stcb->asoc.total_flight_count > 0) \ stcb->asoc.total_flight_count--; \ } else { \ stcb->asoc.total_flight = 0; \ stcb->asoc.total_flight_count = 0; \ } \ } while (0) #define sctp_total_flight_increase(stcb, tp1) do { \ (stcb)->asoc.total_flight_count++; \ (stcb)->asoc.total_flight += (tp1)->book_size; \ } while (0) #endif struct sctp_nets; struct sctp_inpcb; struct sctp_tcb; struct sctphdr; void sctp_close(struct socket *so); int sctp_disconnect(struct socket *so); void sctp_ctlinput __P((int, struct sockaddr *, void *)); int sctp_ctloutput __P((struct socket *, struct sockopt *)); void sctp_input_with_port __P((struct mbuf *, int, uint16_t)); void sctp_input __P((struct mbuf *, int)); void sctp_pathmtu_adjustment __P((struct sctp_inpcb *, struct sctp_tcb *, struct sctp_nets *, uint16_t)); void sctp_drain __P((void)); void sctp_init __P((void)); void sctp_finish(void); int sctp_flush(struct socket *, int); int sctp_shutdown __P((struct socket *)); void sctp_notify __P((struct sctp_inpcb *, struct ip *ip, struct sctphdr *, struct sockaddr *, struct sctp_tcb *, struct sctp_nets *)); int sctp_bindx(struct socket *, int, struct sockaddr_storage *, int, int, struct proc *); /* can't use sctp_assoc_t here */ int sctp_peeloff(struct socket *, struct socket *, int, caddr_t, int *); int sctp_ingetaddr(struct socket *, struct sockaddr ** ); int sctp_peeraddr(struct socket *, struct sockaddr ** ); int sctp_listen(struct socket *, int, struct thread *); int sctp_accept(struct socket *, struct sockaddr **); #endif /* _KERNEL */ #endif /* !_NETINET_SCTP_VAR_H_ */ Index: head/sys/netinet/sctputil.c =================================================================== --- head/sys/netinet/sctputil.c (revision 181053) +++ head/sys/netinet/sctputil.c (revision 181054) @@ -1,6639 +1,6639 @@ /*- * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /* $KAME: sctputil.c,v 1.37 2005/03/07 23:26:09 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #ifdef INET6 #endif #include #include #include #include #include #include /* for sctp_deliver_data() */ #include #include #include #define NUMBER_OF_MTU_SIZES 18 #ifndef KTR_SCTP #define KTR_SCTP KTR_SUBSYS #endif void sctp_sblog(struct sockbuf *sb, struct sctp_tcb *stcb, int from, int incr) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.sb.stcb = stcb; sctp_clog.x.sb.so_sbcc = sb->sb_cc; if (stcb) sctp_clog.x.sb.stcb_sbcc = stcb->asoc.sb_cc; else sctp_clog.x.sb.stcb_sbcc = 0; sctp_clog.x.sb.incr = incr; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_SB, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_closing(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int16_t loc) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.close.inp = (void *)inp; sctp_clog.x.close.sctp_flags = inp->sctp_flags; if (stcb) { sctp_clog.x.close.stcb = (void *)stcb; sctp_clog.x.close.state = (uint16_t) stcb->asoc.state; } else { sctp_clog.x.close.stcb = 0; sctp_clog.x.close.state = 0; } sctp_clog.x.close.loc = loc; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_CLOSE, 0, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void rto_logging(struct sctp_nets *net, int from) { struct sctp_cwnd_log sctp_clog; memset(&sctp_clog, 0, sizeof(sctp_clog)); sctp_clog.x.rto.net = (void *)net; sctp_clog.x.rto.rtt = net->prev_rtt; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_RTT, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_strm_del_alt(struct sctp_tcb *stcb, uint32_t tsn, uint16_t sseq, uint16_t stream, int from) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.strlog.stcb = stcb; sctp_clog.x.strlog.n_tsn = tsn; sctp_clog.x.strlog.n_sseq = sseq; sctp_clog.x.strlog.e_tsn = 0; sctp_clog.x.strlog.e_sseq = 0; sctp_clog.x.strlog.strm = stream; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_STRM, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_nagle_event(struct sctp_tcb *stcb, int action) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.nagle.stcb = (void *)stcb; sctp_clog.x.nagle.total_flight = stcb->asoc.total_flight; sctp_clog.x.nagle.total_in_queue = stcb->asoc.total_output_queue_size; sctp_clog.x.nagle.count_in_queue = stcb->asoc.chunks_on_out_queue; sctp_clog.x.nagle.count_in_flight = stcb->asoc.total_flight_count; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_NAGLE, action, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_sack(uint32_t old_cumack, uint32_t cumack, uint32_t tsn, uint16_t gaps, uint16_t dups, int from) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.sack.cumack = cumack; sctp_clog.x.sack.oldcumack = old_cumack; sctp_clog.x.sack.tsn = tsn; sctp_clog.x.sack.numGaps = gaps; sctp_clog.x.sack.numDups = dups; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_SACK, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_map(uint32_t map, uint32_t cum, uint32_t high, int from) { struct sctp_cwnd_log sctp_clog; memset(&sctp_clog, 0, sizeof(sctp_clog)); sctp_clog.x.map.base = map; sctp_clog.x.map.cum = cum; sctp_clog.x.map.high = high; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_MAP, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_fr(uint32_t biggest_tsn, uint32_t biggest_new_tsn, uint32_t tsn, int from) { struct sctp_cwnd_log sctp_clog; memset(&sctp_clog, 0, sizeof(sctp_clog)); sctp_clog.x.fr.largest_tsn = biggest_tsn; sctp_clog.x.fr.largest_new_tsn = biggest_new_tsn; sctp_clog.x.fr.tsn = tsn; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_FR, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_mb(struct mbuf *m, int from) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.mb.mp = m; sctp_clog.x.mb.mbuf_flags = (uint8_t) (SCTP_BUF_GET_FLAGS(m)); sctp_clog.x.mb.size = (uint16_t) (SCTP_BUF_LEN(m)); sctp_clog.x.mb.data = SCTP_BUF_AT(m, 0); if (SCTP_BUF_IS_EXTENDED(m)) { sctp_clog.x.mb.ext = SCTP_BUF_EXTEND_BASE(m); sctp_clog.x.mb.refcnt = (uint8_t) (SCTP_BUF_EXTEND_REFCNT(m)); } else { sctp_clog.x.mb.ext = 0; sctp_clog.x.mb.refcnt = 0; } SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_MBUF, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_strm_del(struct sctp_queued_to_read *control, struct sctp_queued_to_read *poschk, int from) { struct sctp_cwnd_log sctp_clog; if (control == NULL) { SCTP_PRINTF("Gak log of NULL?\n"); return; } sctp_clog.x.strlog.stcb = control->stcb; sctp_clog.x.strlog.n_tsn = control->sinfo_tsn; sctp_clog.x.strlog.n_sseq = control->sinfo_ssn; sctp_clog.x.strlog.strm = control->sinfo_stream; if (poschk != NULL) { sctp_clog.x.strlog.e_tsn = poschk->sinfo_tsn; sctp_clog.x.strlog.e_sseq = poschk->sinfo_ssn; } else { sctp_clog.x.strlog.e_tsn = 0; sctp_clog.x.strlog.e_sseq = 0; } SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_STRM, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_cwnd(struct sctp_tcb *stcb, struct sctp_nets *net, int augment, uint8_t from) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.cwnd.net = net; if (stcb->asoc.send_queue_cnt > 255) sctp_clog.x.cwnd.cnt_in_send = 255; else sctp_clog.x.cwnd.cnt_in_send = stcb->asoc.send_queue_cnt; if (stcb->asoc.stream_queue_cnt > 255) sctp_clog.x.cwnd.cnt_in_str = 255; else sctp_clog.x.cwnd.cnt_in_str = stcb->asoc.stream_queue_cnt; if (net) { sctp_clog.x.cwnd.cwnd_new_value = net->cwnd; sctp_clog.x.cwnd.inflight = net->flight_size; sctp_clog.x.cwnd.pseudo_cumack = net->pseudo_cumack; sctp_clog.x.cwnd.meets_pseudo_cumack = net->new_pseudo_cumack; sctp_clog.x.cwnd.need_new_pseudo_cumack = net->find_pseudo_cumack; } if (SCTP_CWNDLOG_PRESEND == from) { sctp_clog.x.cwnd.meets_pseudo_cumack = stcb->asoc.peers_rwnd; } sctp_clog.x.cwnd.cwnd_augment = augment; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_CWND, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_lock(struct sctp_inpcb *inp, struct sctp_tcb *stcb, uint8_t from) { struct sctp_cwnd_log sctp_clog; memset(&sctp_clog, 0, sizeof(sctp_clog)); if (inp) { sctp_clog.x.lock.sock = (void *)inp->sctp_socket; } else { sctp_clog.x.lock.sock = (void *)NULL; } sctp_clog.x.lock.inp = (void *)inp; if (stcb) { sctp_clog.x.lock.tcb_lock = mtx_owned(&stcb->tcb_mtx); } else { sctp_clog.x.lock.tcb_lock = SCTP_LOCK_UNKNOWN; } if (inp) { sctp_clog.x.lock.inp_lock = mtx_owned(&inp->inp_mtx); sctp_clog.x.lock.create_lock = mtx_owned(&inp->inp_create_mtx); } else { sctp_clog.x.lock.inp_lock = SCTP_LOCK_UNKNOWN; sctp_clog.x.lock.create_lock = SCTP_LOCK_UNKNOWN; } sctp_clog.x.lock.info_lock = rw_wowned(&SCTP_BASE_INFO(ipi_ep_mtx)); if (inp->sctp_socket) { sctp_clog.x.lock.sock_lock = mtx_owned(&(inp->sctp_socket->so_rcv.sb_mtx)); sctp_clog.x.lock.sockrcvbuf_lock = mtx_owned(&(inp->sctp_socket->so_rcv.sb_mtx)); sctp_clog.x.lock.socksndbuf_lock = mtx_owned(&(inp->sctp_socket->so_snd.sb_mtx)); } else { sctp_clog.x.lock.sock_lock = SCTP_LOCK_UNKNOWN; sctp_clog.x.lock.sockrcvbuf_lock = SCTP_LOCK_UNKNOWN; sctp_clog.x.lock.socksndbuf_lock = SCTP_LOCK_UNKNOWN; } SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_LOCK_EVENT, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_maxburst(struct sctp_tcb *stcb, struct sctp_nets *net, int error, int burst, uint8_t from) { struct sctp_cwnd_log sctp_clog; memset(&sctp_clog, 0, sizeof(sctp_clog)); sctp_clog.x.cwnd.net = net; sctp_clog.x.cwnd.cwnd_new_value = error; sctp_clog.x.cwnd.inflight = net->flight_size; sctp_clog.x.cwnd.cwnd_augment = burst; if (stcb->asoc.send_queue_cnt > 255) sctp_clog.x.cwnd.cnt_in_send = 255; else sctp_clog.x.cwnd.cnt_in_send = stcb->asoc.send_queue_cnt; if (stcb->asoc.stream_queue_cnt > 255) sctp_clog.x.cwnd.cnt_in_str = 255; else sctp_clog.x.cwnd.cnt_in_str = stcb->asoc.stream_queue_cnt; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_MAXBURST, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_rwnd(uint8_t from, uint32_t peers_rwnd, uint32_t snd_size, uint32_t overhead) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.rwnd.rwnd = peers_rwnd; sctp_clog.x.rwnd.send_size = snd_size; sctp_clog.x.rwnd.overhead = overhead; sctp_clog.x.rwnd.new_rwnd = 0; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_RWND, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_rwnd_set(uint8_t from, uint32_t peers_rwnd, uint32_t flight_size, uint32_t overhead, uint32_t a_rwndval) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.rwnd.rwnd = peers_rwnd; sctp_clog.x.rwnd.send_size = flight_size; sctp_clog.x.rwnd.overhead = overhead; sctp_clog.x.rwnd.new_rwnd = a_rwndval; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_RWND, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_mbcnt(uint8_t from, uint32_t total_oq, uint32_t book, uint32_t total_mbcnt_q, uint32_t mbcnt) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.mbcnt.total_queue_size = total_oq; sctp_clog.x.mbcnt.size_change = book; sctp_clog.x.mbcnt.total_queue_mb_size = total_mbcnt_q; sctp_clog.x.mbcnt.mbcnt_change = mbcnt; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_MBCNT, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_misc_ints(uint8_t from, uint32_t a, uint32_t b, uint32_t c, uint32_t d) { SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_MISC_EVENT, from, a, b, c, d); } void sctp_wakeup_log(struct sctp_tcb *stcb, uint32_t cumtsn, uint32_t wake_cnt, int from) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.wake.stcb = (void *)stcb; sctp_clog.x.wake.wake_cnt = wake_cnt; sctp_clog.x.wake.flight = stcb->asoc.total_flight_count; sctp_clog.x.wake.send_q = stcb->asoc.send_queue_cnt; sctp_clog.x.wake.sent_q = stcb->asoc.sent_queue_cnt; if (stcb->asoc.stream_queue_cnt < 0xff) sctp_clog.x.wake.stream_qcnt = (uint8_t) stcb->asoc.stream_queue_cnt; else sctp_clog.x.wake.stream_qcnt = 0xff; if (stcb->asoc.chunks_on_out_queue < 0xff) sctp_clog.x.wake.chunks_on_oque = (uint8_t) stcb->asoc.chunks_on_out_queue; else sctp_clog.x.wake.chunks_on_oque = 0xff; sctp_clog.x.wake.sctpflags = 0; /* set in the defered mode stuff */ if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) sctp_clog.x.wake.sctpflags |= 1; if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_WAKEOUTPUT) sctp_clog.x.wake.sctpflags |= 2; if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_WAKEINPUT) sctp_clog.x.wake.sctpflags |= 4; /* what about the sb */ if (stcb->sctp_socket) { struct socket *so = stcb->sctp_socket; sctp_clog.x.wake.sbflags = (uint8_t) ((so->so_snd.sb_flags & 0x00ff)); } else { sctp_clog.x.wake.sbflags = 0xff; } SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_WAKE, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } void sctp_log_block(uint8_t from, struct socket *so, struct sctp_association *asoc, int sendlen) { struct sctp_cwnd_log sctp_clog; sctp_clog.x.blk.onsb = asoc->total_output_queue_size; sctp_clog.x.blk.send_sent_qcnt = (uint16_t) (asoc->send_queue_cnt + asoc->sent_queue_cnt); sctp_clog.x.blk.peer_rwnd = asoc->peers_rwnd; sctp_clog.x.blk.stream_qcnt = (uint16_t) asoc->stream_queue_cnt; sctp_clog.x.blk.chunks_on_oque = (uint16_t) asoc->chunks_on_out_queue; sctp_clog.x.blk.flight_size = (uint16_t) (asoc->total_flight / 1024); sctp_clog.x.blk.sndlen = sendlen; SCTP_CTR6(KTR_SCTP, "SCTP:%d[%d]:%x-%x-%x-%x", SCTP_LOG_EVENT_BLOCK, from, sctp_clog.x.misc.log1, sctp_clog.x.misc.log2, sctp_clog.x.misc.log3, sctp_clog.x.misc.log4); } int sctp_fill_stat_log(void *optval, size_t *optsize) { /* May need to fix this if ktrdump does not work */ return (0); } #ifdef SCTP_AUDITING_ENABLED uint8_t sctp_audit_data[SCTP_AUDIT_SIZE][2]; static int sctp_audit_indx = 0; static void sctp_print_audit_report(void) { int i; int cnt; cnt = 0; for (i = sctp_audit_indx; i < SCTP_AUDIT_SIZE; i++) { if ((sctp_audit_data[i][0] == 0xe0) && (sctp_audit_data[i][1] == 0x01)) { cnt = 0; SCTP_PRINTF("\n"); } else if (sctp_audit_data[i][0] == 0xf0) { cnt = 0; SCTP_PRINTF("\n"); } else if ((sctp_audit_data[i][0] == 0xc0) && (sctp_audit_data[i][1] == 0x01)) { SCTP_PRINTF("\n"); cnt = 0; } SCTP_PRINTF("%2.2x%2.2x ", (uint32_t) sctp_audit_data[i][0], (uint32_t) sctp_audit_data[i][1]); cnt++; if ((cnt % 14) == 0) SCTP_PRINTF("\n"); } for (i = 0; i < sctp_audit_indx; i++) { if ((sctp_audit_data[i][0] == 0xe0) && (sctp_audit_data[i][1] == 0x01)) { cnt = 0; SCTP_PRINTF("\n"); } else if (sctp_audit_data[i][0] == 0xf0) { cnt = 0; SCTP_PRINTF("\n"); } else if ((sctp_audit_data[i][0] == 0xc0) && (sctp_audit_data[i][1] == 0x01)) { SCTP_PRINTF("\n"); cnt = 0; } SCTP_PRINTF("%2.2x%2.2x ", (uint32_t) sctp_audit_data[i][0], (uint32_t) sctp_audit_data[i][1]); cnt++; if ((cnt % 14) == 0) SCTP_PRINTF("\n"); } SCTP_PRINTF("\n"); } void sctp_auditing(int from, struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { int resend_cnt, tot_out, rep, tot_book_cnt; struct sctp_nets *lnet; struct sctp_tmit_chunk *chk; sctp_audit_data[sctp_audit_indx][0] = 0xAA; sctp_audit_data[sctp_audit_indx][1] = 0x000000ff & from; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } if (inp == NULL) { sctp_audit_data[sctp_audit_indx][0] = 0xAF; sctp_audit_data[sctp_audit_indx][1] = 0x01; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } return; } if (stcb == NULL) { sctp_audit_data[sctp_audit_indx][0] = 0xAF; sctp_audit_data[sctp_audit_indx][1] = 0x02; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } return; } sctp_audit_data[sctp_audit_indx][0] = 0xA1; sctp_audit_data[sctp_audit_indx][1] = (0x000000ff & stcb->asoc.sent_queue_retran_cnt); sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } rep = 0; tot_book_cnt = 0; resend_cnt = tot_out = 0; TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { if (chk->sent == SCTP_DATAGRAM_RESEND) { resend_cnt++; } else if (chk->sent < SCTP_DATAGRAM_RESEND) { tot_out += chk->book_size; tot_book_cnt++; } } if (resend_cnt != stcb->asoc.sent_queue_retran_cnt) { sctp_audit_data[sctp_audit_indx][0] = 0xAF; sctp_audit_data[sctp_audit_indx][1] = 0xA1; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } SCTP_PRINTF("resend_cnt:%d asoc-tot:%d\n", resend_cnt, stcb->asoc.sent_queue_retran_cnt); rep = 1; stcb->asoc.sent_queue_retran_cnt = resend_cnt; sctp_audit_data[sctp_audit_indx][0] = 0xA2; sctp_audit_data[sctp_audit_indx][1] = (0x000000ff & stcb->asoc.sent_queue_retran_cnt); sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } } if (tot_out != stcb->asoc.total_flight) { sctp_audit_data[sctp_audit_indx][0] = 0xAF; sctp_audit_data[sctp_audit_indx][1] = 0xA2; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } rep = 1; SCTP_PRINTF("tot_flt:%d asoc_tot:%d\n", tot_out, (int)stcb->asoc.total_flight); stcb->asoc.total_flight = tot_out; } if (tot_book_cnt != stcb->asoc.total_flight_count) { sctp_audit_data[sctp_audit_indx][0] = 0xAF; sctp_audit_data[sctp_audit_indx][1] = 0xA5; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } rep = 1; SCTP_PRINTF("tot_flt_book:%d\n", tot_book); stcb->asoc.total_flight_count = tot_book_cnt; } tot_out = 0; TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { tot_out += lnet->flight_size; } if (tot_out != stcb->asoc.total_flight) { sctp_audit_data[sctp_audit_indx][0] = 0xAF; sctp_audit_data[sctp_audit_indx][1] = 0xA3; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } rep = 1; SCTP_PRINTF("real flight:%d net total was %d\n", stcb->asoc.total_flight, tot_out); /* now corrective action */ TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { tot_out = 0; TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { if ((chk->whoTo == lnet) && (chk->sent < SCTP_DATAGRAM_RESEND)) { tot_out += chk->book_size; } } if (lnet->flight_size != tot_out) { SCTP_PRINTF("net:%x flight was %d corrected to %d\n", (uint32_t) lnet, lnet->flight_size, tot_out); lnet->flight_size = tot_out; } } } if (rep) { sctp_print_audit_report(); } } void sctp_audit_log(uint8_t ev, uint8_t fd) { sctp_audit_data[sctp_audit_indx][0] = ev; sctp_audit_data[sctp_audit_indx][1] = fd; sctp_audit_indx++; if (sctp_audit_indx >= SCTP_AUDIT_SIZE) { sctp_audit_indx = 0; } } #endif /* * a list of sizes based on typical mtu's, used only if next hop size not * returned. */ static int sctp_mtu_sizes[] = { 68, 296, 508, 512, 544, 576, 1006, 1492, 1500, 1536, 2002, 2048, 4352, 4464, 8166, 17914, 32000, 65535 }; void sctp_stop_timers_for_shutdown(struct sctp_tcb *stcb) { struct sctp_association *asoc; struct sctp_nets *net; asoc = &stcb->asoc; (void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer); TAILQ_FOREACH(net, &asoc->nets, sctp_next) { (void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer); (void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer); } } int find_next_best_mtu(int totsz) { int i, perfer; /* * if we are in here we must find the next best fit based on the * size of the dg that failed to be sent. */ perfer = 0; for (i = 0; i < NUMBER_OF_MTU_SIZES; i++) { if (totsz < sctp_mtu_sizes[i]) { perfer = i - 1; if (perfer < 0) perfer = 0; break; } } return (sctp_mtu_sizes[perfer]); } void sctp_fill_random_store(struct sctp_pcb *m) { /* * Here we use the MD5/SHA-1 to hash with our good randomNumbers and * our counter. The result becomes our good random numbers and we * then setup to give these out. Note that we do no locking to * protect this. This is ok, since if competing folks call this we * will get more gobbled gook in the random store which is what we * want. There is a danger that two guys will use the same random * numbers, but thats ok too since that is random as well :-> */ m->store_at = 0; (void)sctp_hmac(SCTP_HMAC, (uint8_t *) m->random_numbers, sizeof(m->random_numbers), (uint8_t *) & m->random_counter, sizeof(m->random_counter), (uint8_t *) m->random_store); m->random_counter++; } uint32_t sctp_select_initial_TSN(struct sctp_pcb *inp) { /* * A true implementation should use random selection process to get * the initial stream sequence number, using RFC1750 as a good * guideline */ uint32_t x, *xp; uint8_t *p; int store_at, new_store; if (inp->initial_sequence_debug != 0) { uint32_t ret; ret = inp->initial_sequence_debug; inp->initial_sequence_debug++; return (ret); } retry: store_at = inp->store_at; new_store = store_at + sizeof(uint32_t); if (new_store >= (SCTP_SIGNATURE_SIZE - 3)) { new_store = 0; } if (!atomic_cmpset_int(&inp->store_at, store_at, new_store)) { goto retry; } if (new_store == 0) { /* Refill the random store */ sctp_fill_random_store(inp); } p = &inp->random_store[store_at]; xp = (uint32_t *) p; x = *xp; return (x); } uint32_t sctp_select_a_tag(struct sctp_inpcb *inp, int save_in_twait) { u_long x, not_done; struct timeval now; (void)SCTP_GETTIME_TIMEVAL(&now); not_done = 1; while (not_done) { x = sctp_select_initial_TSN(&inp->sctp_ep); if (x == 0) { /* we never use 0 */ continue; } if (sctp_is_vtag_good(inp, x, &now, save_in_twait)) { not_done = 0; } } return (x); } int sctp_init_asoc(struct sctp_inpcb *m, struct sctp_tcb *stcb, int for_a_init, uint32_t override_tag, uint32_t vrf_id) { struct sctp_association *asoc; /* * Anything set to zero is taken care of by the allocation routine's * bzero */ /* * Up front select what scoping to apply on addresses I tell my peer * Not sure what to do with these right now, we will need to come up * with a way to set them. We may need to pass them through from the * caller in the sctp_aloc_assoc() function. */ int i; asoc = &stcb->asoc; /* init all variables to a known value. */ SCTP_SET_STATE(&stcb->asoc, SCTP_STATE_INUSE); asoc->max_burst = m->sctp_ep.max_burst; asoc->heart_beat_delay = TICKS_TO_MSEC(m->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT]); asoc->cookie_life = m->sctp_ep.def_cookie_life; asoc->sctp_cmt_on_off = (uint8_t) SCTP_BASE_SYSCTL(sctp_cmt_on_off); /* JRS 5/21/07 - Init CMT PF variables */ asoc->sctp_cmt_pf = (uint8_t) SCTP_BASE_SYSCTL(sctp_cmt_pf); asoc->sctp_frag_point = m->sctp_frag_point; #ifdef INET asoc->default_tos = m->ip_inp.inp.inp_ip_tos; #else asoc->default_tos = 0; #endif #ifdef INET6 asoc->default_flowlabel = ((struct in6pcb *)m)->in6p_flowinfo; #else asoc->default_flowlabel = 0; #endif asoc->sb_send_resv = 0; if (override_tag) { struct timeval now; (void)SCTP_GETTIME_TIMEVAL(&now); if (sctp_is_in_timewait(override_tag)) { /* * It must be in the time-wait hash, we put it there * when we aloc one. If not the peer is playing * games. */ asoc->my_vtag = override_tag; } else { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOMEM); panic("Huh is_in_timewait fails"); return (ENOMEM); } } else { asoc->my_vtag = sctp_select_a_tag(m, 1); } /* Get the nonce tags */ asoc->my_vtag_nonce = sctp_select_a_tag(m, 0); asoc->peer_vtag_nonce = sctp_select_a_tag(m, 0); asoc->vrf_id = vrf_id; if (sctp_is_feature_on(m, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) asoc->hb_is_disabled = 1; else asoc->hb_is_disabled = 0; #ifdef SCTP_ASOCLOG_OF_TSNS asoc->tsn_in_at = 0; asoc->tsn_out_at = 0; asoc->tsn_in_wrapped = 0; asoc->tsn_out_wrapped = 0; asoc->cumack_log_at = 0; asoc->cumack_log_atsnt = 0; #endif #ifdef SCTP_FS_SPEC_LOG asoc->fs_index = 0; #endif asoc->refcnt = 0; asoc->assoc_up_sent = 0; asoc->assoc_id = asoc->my_vtag; asoc->asconf_seq_out = asoc->str_reset_seq_out = asoc->init_seq_number = asoc->sending_seq = sctp_select_initial_TSN(&m->sctp_ep); asoc->asconf_seq_out_acked = asoc->asconf_seq_out - 1; /* we are optimisitic here */ asoc->peer_supports_pktdrop = 1; asoc->sent_queue_retran_cnt = 0; /* for CMT */ asoc->last_net_data_came_from = NULL; /* This will need to be adjusted */ asoc->last_cwr_tsn = asoc->init_seq_number - 1; asoc->last_acked_seq = asoc->init_seq_number - 1; asoc->advanced_peer_ack_point = asoc->last_acked_seq; asoc->asconf_seq_in = asoc->last_acked_seq; /* here we are different, we hold the next one we expect */ asoc->str_reset_seq_in = asoc->last_acked_seq + 1; asoc->initial_init_rto_max = m->sctp_ep.initial_init_rto_max; asoc->initial_rto = m->sctp_ep.initial_rto; asoc->max_init_times = m->sctp_ep.max_init_times; asoc->max_send_times = m->sctp_ep.max_send_times; asoc->def_net_failure = m->sctp_ep.def_net_failure; asoc->free_chunk_cnt = 0; asoc->iam_blocking = 0; /* ECN Nonce initialization */ asoc->context = m->sctp_context; asoc->def_send = m->def_send; asoc->ecn_nonce_allowed = 0; asoc->receiver_nonce_sum = 1; asoc->nonce_sum_expect_base = 1; asoc->nonce_sum_check = 1; asoc->nonce_resync_tsn = 0; asoc->nonce_wait_for_ecne = 0; asoc->nonce_wait_tsn = 0; asoc->delayed_ack = TICKS_TO_MSEC(m->sctp_ep.sctp_timeoutticks[SCTP_TIMER_RECV]); asoc->sack_freq = m->sctp_ep.sctp_sack_freq; asoc->pr_sctp_cnt = 0; asoc->total_output_queue_size = 0; if (m->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { struct in6pcb *inp6; /* Its a V6 socket */ inp6 = (struct in6pcb *)m; asoc->ipv6_addr_legal = 1; /* Now look at the binding flag to see if V4 will be legal */ if (SCTP_IPV6_V6ONLY(inp6) == 0) { asoc->ipv4_addr_legal = 1; } else { /* V4 addresses are NOT legal on the association */ asoc->ipv4_addr_legal = 0; } } else { /* Its a V4 socket, no - V6 */ asoc->ipv4_addr_legal = 1; asoc->ipv6_addr_legal = 0; } asoc->my_rwnd = max(SCTP_SB_LIMIT_RCV(m->sctp_socket), SCTP_MINIMAL_RWND); asoc->peers_rwnd = SCTP_SB_LIMIT_RCV(m->sctp_socket); asoc->smallest_mtu = m->sctp_frag_point; #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("smallest_mtu init'd with asoc to :%d\n", asoc->smallest_mtu); #endif asoc->minrto = m->sctp_ep.sctp_minrto; asoc->maxrto = m->sctp_ep.sctp_maxrto; asoc->locked_on_sending = NULL; asoc->stream_locked_on = 0; asoc->ecn_echo_cnt_onq = 0; asoc->stream_locked = 0; asoc->send_sack = 1; LIST_INIT(&asoc->sctp_restricted_addrs); TAILQ_INIT(&asoc->nets); TAILQ_INIT(&asoc->pending_reply_queue); TAILQ_INIT(&asoc->asconf_ack_sent); /* Setup to fill the hb random cache at first HB */ asoc->hb_random_idx = 4; asoc->sctp_autoclose_ticks = m->sctp_ep.auto_close_time; /* * JRS - Pick the default congestion control module based on the * sysctl. */ switch (m->sctp_ep.sctp_default_cc_module) { /* JRS - Standard TCP congestion control */ case SCTP_CC_RFC2581: { stcb->asoc.congestion_control_module = SCTP_CC_RFC2581; stcb->asoc.cc_functions.sctp_set_initial_cc_param = &sctp_set_initial_cc_param; stcb->asoc.cc_functions.sctp_cwnd_update_after_sack = &sctp_cwnd_update_after_sack; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr = &sctp_cwnd_update_after_fr; stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout = &sctp_cwnd_update_after_timeout; stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo = &sctp_cwnd_update_after_ecn_echo; stcb->asoc.cc_functions.sctp_cwnd_update_after_packet_dropped = &sctp_cwnd_update_after_packet_dropped; stcb->asoc.cc_functions.sctp_cwnd_update_after_output = &sctp_cwnd_update_after_output; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer = &sctp_cwnd_update_after_fr_timer; break; } /* JRS - High Speed TCP congestion control (Floyd) */ case SCTP_CC_HSTCP: { stcb->asoc.congestion_control_module = SCTP_CC_HSTCP; stcb->asoc.cc_functions.sctp_set_initial_cc_param = &sctp_set_initial_cc_param; stcb->asoc.cc_functions.sctp_cwnd_update_after_sack = &sctp_hs_cwnd_update_after_sack; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr = &sctp_hs_cwnd_update_after_fr; stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout = &sctp_cwnd_update_after_timeout; stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo = &sctp_cwnd_update_after_ecn_echo; stcb->asoc.cc_functions.sctp_cwnd_update_after_packet_dropped = &sctp_cwnd_update_after_packet_dropped; stcb->asoc.cc_functions.sctp_cwnd_update_after_output = &sctp_cwnd_update_after_output; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer = &sctp_cwnd_update_after_fr_timer; break; } /* JRS - HTCP congestion control */ case SCTP_CC_HTCP: { stcb->asoc.congestion_control_module = SCTP_CC_HTCP; stcb->asoc.cc_functions.sctp_set_initial_cc_param = &sctp_htcp_set_initial_cc_param; stcb->asoc.cc_functions.sctp_cwnd_update_after_sack = &sctp_htcp_cwnd_update_after_sack; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr = &sctp_htcp_cwnd_update_after_fr; stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout = &sctp_htcp_cwnd_update_after_timeout; stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo = &sctp_htcp_cwnd_update_after_ecn_echo; stcb->asoc.cc_functions.sctp_cwnd_update_after_packet_dropped = &sctp_cwnd_update_after_packet_dropped; stcb->asoc.cc_functions.sctp_cwnd_update_after_output = &sctp_cwnd_update_after_output; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer = &sctp_htcp_cwnd_update_after_fr_timer; break; } /* JRS - By default, use RFC2581 */ default: { stcb->asoc.congestion_control_module = SCTP_CC_RFC2581; stcb->asoc.cc_functions.sctp_set_initial_cc_param = &sctp_set_initial_cc_param; stcb->asoc.cc_functions.sctp_cwnd_update_after_sack = &sctp_cwnd_update_after_sack; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr = &sctp_cwnd_update_after_fr; stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout = &sctp_cwnd_update_after_timeout; stcb->asoc.cc_functions.sctp_cwnd_update_after_ecn_echo = &sctp_cwnd_update_after_ecn_echo; stcb->asoc.cc_functions.sctp_cwnd_update_after_packet_dropped = &sctp_cwnd_update_after_packet_dropped; stcb->asoc.cc_functions.sctp_cwnd_update_after_output = &sctp_cwnd_update_after_output; stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer = &sctp_cwnd_update_after_fr_timer; break; } } /* * Now the stream parameters, here we allocate space for all streams * that we request by default. */ asoc->streamoutcnt = asoc->pre_open_streams = m->sctp_ep.pre_open_stream_count; SCTP_MALLOC(asoc->strmout, struct sctp_stream_out *, asoc->streamoutcnt * sizeof(struct sctp_stream_out), SCTP_M_STRMO); if (asoc->strmout == NULL) { /* big trouble no memory */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOMEM); return (ENOMEM); } for (i = 0; i < asoc->streamoutcnt; i++) { /* * inbound side must be set to 0xffff, also NOTE when we get * the INIT-ACK back (for INIT sender) we MUST reduce the * count (streamoutcnt) but first check if we sent to any of * the upper streams that were dropped (if some were). Those * that were dropped must be notified to the upper layer as * failed to send. */ asoc->strmout[i].next_sequence_sent = 0x0; TAILQ_INIT(&asoc->strmout[i].outqueue); asoc->strmout[i].stream_no = i; asoc->strmout[i].last_msg_incomplete = 0; asoc->strmout[i].next_spoke.tqe_next = 0; asoc->strmout[i].next_spoke.tqe_prev = 0; } /* Now the mapping array */ asoc->mapping_array_size = SCTP_INITIAL_MAPPING_ARRAY; SCTP_MALLOC(asoc->mapping_array, uint8_t *, asoc->mapping_array_size, SCTP_M_MAP); if (asoc->mapping_array == NULL) { SCTP_FREE(asoc->strmout, SCTP_M_STRMO); SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOMEM); return (ENOMEM); } memset(asoc->mapping_array, 0, asoc->mapping_array_size); /* Now the init of the other outqueues */ TAILQ_INIT(&asoc->free_chunks); TAILQ_INIT(&asoc->out_wheel); TAILQ_INIT(&asoc->control_send_queue); TAILQ_INIT(&asoc->asconf_send_queue); TAILQ_INIT(&asoc->send_queue); TAILQ_INIT(&asoc->sent_queue); TAILQ_INIT(&asoc->reasmqueue); TAILQ_INIT(&asoc->resetHead); asoc->max_inbound_streams = m->sctp_ep.max_open_streams_intome; TAILQ_INIT(&asoc->asconf_queue); /* authentication fields */ asoc->authinfo.random = NULL; asoc->authinfo.assoc_key = NULL; asoc->authinfo.assoc_keyid = 0; asoc->authinfo.recv_key = NULL; asoc->authinfo.recv_keyid = 0; LIST_INIT(&asoc->shared_keys); asoc->marked_retrans = 0; asoc->timoinit = 0; asoc->timodata = 0; asoc->timosack = 0; asoc->timoshutdown = 0; asoc->timoheartbeat = 0; asoc->timocookie = 0; asoc->timoshutdownack = 0; (void)SCTP_GETTIME_TIMEVAL(&asoc->start_time); asoc->discontinuity_time = asoc->start_time; /* * sa_ignore MEMLEAK {memory is put in the assoc mapping array and * freed later whe the association is freed. */ return (0); } int sctp_expand_mapping_array(struct sctp_association *asoc, uint32_t needed) { /* mapping array needs to grow */ uint8_t *new_array; uint32_t new_size; new_size = asoc->mapping_array_size + ((needed + 7) / 8 + SCTP_MAPPING_ARRAY_INCR); SCTP_MALLOC(new_array, uint8_t *, new_size, SCTP_M_MAP); if (new_array == NULL) { /* can't get more, forget it */ SCTP_PRINTF("No memory for expansion of SCTP mapping array %d\n", new_size); return (-1); } memset(new_array, 0, new_size); memcpy(new_array, asoc->mapping_array, asoc->mapping_array_size); SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); asoc->mapping_array = new_array; asoc->mapping_array_size = new_size; return (0); } #if defined(SCTP_USE_THREAD_BASED_ITERATOR) static void sctp_iterator_work(struct sctp_iterator *it) { int iteration_count = 0; int inp_skip = 0; SCTP_ITERATOR_LOCK(); if (it->inp) { SCTP_INP_DECR_REF(it->inp); } if (it->inp == NULL) { /* iterator is complete */ done_with_iterator: SCTP_ITERATOR_UNLOCK(); if (it->function_atend != NULL) { (*it->function_atend) (it->pointer, it->val); } SCTP_FREE(it, SCTP_M_ITER); return; } select_a_new_ep: SCTP_INP_WLOCK(it->inp); while (((it->pcb_flags) && ((it->inp->sctp_flags & it->pcb_flags) != it->pcb_flags)) || ((it->pcb_features) && ((it->inp->sctp_features & it->pcb_features) != it->pcb_features))) { /* endpoint flags or features don't match, so keep looking */ if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { SCTP_INP_WUNLOCK(it->inp); goto done_with_iterator; } SCTP_INP_WUNLOCK(it->inp); it->inp = LIST_NEXT(it->inp, sctp_list); if (it->inp == NULL) { goto done_with_iterator; } SCTP_INP_WLOCK(it->inp); } SCTP_INP_WUNLOCK(it->inp); SCTP_INP_RLOCK(it->inp); /* now go through each assoc which is in the desired state */ if (it->done_current_ep == 0) { if (it->function_inp != NULL) inp_skip = (*it->function_inp) (it->inp, it->pointer, it->val); it->done_current_ep = 1; } if (it->stcb == NULL) { /* run the per instance function */ it->stcb = LIST_FIRST(&it->inp->sctp_asoc_list); } if ((inp_skip) || it->stcb == NULL) { if (it->function_inp_end != NULL) { inp_skip = (*it->function_inp_end) (it->inp, it->pointer, it->val); } SCTP_INP_RUNLOCK(it->inp); goto no_stcb; } while (it->stcb) { SCTP_TCB_LOCK(it->stcb); if (it->asoc_state && ((it->stcb->asoc.state & it->asoc_state) != it->asoc_state)) { /* not in the right state... keep looking */ SCTP_TCB_UNLOCK(it->stcb); goto next_assoc; } /* see if we have limited out the iterator loop */ iteration_count++; if (iteration_count > SCTP_ITERATOR_MAX_AT_ONCE) { /* Pause to let others grab the lock */ atomic_add_int(&it->stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(it->stcb); SCTP_INP_INCR_REF(it->inp); SCTP_INP_RUNLOCK(it->inp); SCTP_ITERATOR_UNLOCK(); SCTP_ITERATOR_LOCK(); SCTP_INP_RLOCK(it->inp); SCTP_INP_DECR_REF(it->inp); SCTP_TCB_LOCK(it->stcb); atomic_add_int(&it->stcb->asoc.refcnt, -1); iteration_count = 0; } /* run function on this one */ (*it->function_assoc) (it->inp, it->stcb, it->pointer, it->val); /* * we lie here, it really needs to have its own type but * first I must verify that this won't effect things :-0 */ if (it->no_chunk_output == 0) sctp_chunk_output(it->inp, it->stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED); SCTP_TCB_UNLOCK(it->stcb); next_assoc: it->stcb = LIST_NEXT(it->stcb, sctp_tcblist); if (it->stcb == NULL) { /* Run last function */ if (it->function_inp_end != NULL) { inp_skip = (*it->function_inp_end) (it->inp, it->pointer, it->val); } } } SCTP_INP_RUNLOCK(it->inp); no_stcb: /* done with all assocs on this endpoint, move on to next endpoint */ it->done_current_ep = 0; SCTP_INP_WLOCK(it->inp); SCTP_INP_WUNLOCK(it->inp); if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { it->inp = NULL; } else { SCTP_INP_INFO_RLOCK(); it->inp = LIST_NEXT(it->inp, sctp_list); SCTP_INP_INFO_RUNLOCK(); } if (it->inp == NULL) { goto done_with_iterator; } goto select_a_new_ep; } void sctp_iterator_worker(void) { struct sctp_iterator *it = NULL; /* This function is called with the WQ lock in place */ SCTP_BASE_INFO(iterator_running) = 1; again: it = TAILQ_FIRST(&SCTP_BASE_INFO(iteratorhead)); while (it) { /* now lets work on this one */ TAILQ_REMOVE(&SCTP_BASE_INFO(iteratorhead), it, sctp_nxt_itr); SCTP_IPI_ITERATOR_WQ_UNLOCK(); sctp_iterator_work(it); SCTP_IPI_ITERATOR_WQ_LOCK(); /* sa_ignore FREED_MEMORY */ it = TAILQ_FIRST(&SCTP_BASE_INFO(iteratorhead)); } if (TAILQ_FIRST(&SCTP_BASE_INFO(iteratorhead))) { goto again; } SCTP_BASE_INFO(iterator_running) = 0; return; } #endif static void sctp_handle_addr_wq(void) { /* deal with the ADDR wq from the rtsock calls */ struct sctp_laddr *wi; struct sctp_asconf_iterator *asc; SCTP_MALLOC(asc, struct sctp_asconf_iterator *, sizeof(struct sctp_asconf_iterator), SCTP_M_ASC_IT); if (asc == NULL) { /* Try later, no memory */ sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, (struct sctp_inpcb *)NULL, (struct sctp_tcb *)NULL, (struct sctp_nets *)NULL); return; } LIST_INIT(&asc->list_of_work); asc->cnt = 0; SCTP_IPI_ITERATOR_WQ_LOCK(); wi = LIST_FIRST(&SCTP_BASE_INFO(addr_wq)); while (wi != NULL) { LIST_REMOVE(wi, sctp_nxt_addr); LIST_INSERT_HEAD(&asc->list_of_work, wi, sctp_nxt_addr); asc->cnt++; wi = LIST_FIRST(&SCTP_BASE_INFO(addr_wq)); } SCTP_IPI_ITERATOR_WQ_UNLOCK(); if (asc->cnt == 0) { SCTP_FREE(asc, SCTP_M_ASC_IT); } else { (void)sctp_initiate_iterator(sctp_asconf_iterator_ep, sctp_asconf_iterator_stcb, NULL, /* No ep end for boundall */ SCTP_PCB_FLAGS_BOUNDALL, SCTP_PCB_ANY_FEATURES, SCTP_ASOC_ANY_STATE, (void *)asc, 0, sctp_asconf_iterator_end, NULL, 0); } } int retcode = 0; int cur_oerr = 0; void sctp_timeout_handler(void *t) { struct sctp_inpcb *inp; struct sctp_tcb *stcb; struct sctp_nets *net; struct sctp_timer *tmr; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif int did_output, type; struct sctp_iterator *it = NULL; tmr = (struct sctp_timer *)t; inp = (struct sctp_inpcb *)tmr->ep; stcb = (struct sctp_tcb *)tmr->tcb; net = (struct sctp_nets *)tmr->net; did_output = 1; #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xF0, (uint8_t) tmr->type); sctp_auditing(3, inp, stcb, net); #endif /* sanity checks... */ if (tmr->self != (void *)tmr) { /* * SCTP_PRINTF("Stale SCTP timer fired (%p), ignoring...\n", * tmr); */ return; } tmr->stopped_from = 0xa001; if (!SCTP_IS_TIMER_TYPE_VALID(tmr->type)) { /* * SCTP_PRINTF("SCTP timer fired with invalid type: 0x%x\n", * tmr->type); */ return; } tmr->stopped_from = 0xa002; if ((tmr->type != SCTP_TIMER_TYPE_ADDR_WQ) && (inp == NULL)) { return; } /* if this is an iterator timeout, get the struct and clear inp */ tmr->stopped_from = 0xa003; if (tmr->type == SCTP_TIMER_TYPE_ITERATOR) { it = (struct sctp_iterator *)inp; inp = NULL; } type = tmr->type; if (inp) { SCTP_INP_INCR_REF(inp); if ((inp->sctp_socket == 0) && ((tmr->type != SCTP_TIMER_TYPE_INPKILL) && (tmr->type != SCTP_TIMER_TYPE_SHUTDOWN) && (tmr->type != SCTP_TIMER_TYPE_SHUTDOWNACK) && (tmr->type != SCTP_TIMER_TYPE_SHUTDOWNGUARD) && (tmr->type != SCTP_TIMER_TYPE_ASOCKILL)) ) { SCTP_INP_DECR_REF(inp); return; } } tmr->stopped_from = 0xa004; if (stcb) { atomic_add_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state == 0) { atomic_add_int(&stcb->asoc.refcnt, -1); if (inp) { SCTP_INP_DECR_REF(inp); } return; } } tmr->stopped_from = 0xa005; SCTPDBG(SCTP_DEBUG_TIMER1, "Timer type %d goes off\n", tmr->type); if (!SCTP_OS_TIMER_ACTIVE(&tmr->timer)) { if (inp) { SCTP_INP_DECR_REF(inp); } if (stcb) { atomic_add_int(&stcb->asoc.refcnt, -1); } return; } tmr->stopped_from = 0xa006; if (stcb) { SCTP_TCB_LOCK(stcb); atomic_add_int(&stcb->asoc.refcnt, -1); if ((tmr->type != SCTP_TIMER_TYPE_ASOCKILL) && ((stcb->asoc.state == 0) || (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED))) { SCTP_TCB_UNLOCK(stcb); if (inp) { SCTP_INP_DECR_REF(inp); } return; } } /* record in stopped what t-o occured */ tmr->stopped_from = tmr->type; /* mark as being serviced now */ if (SCTP_OS_TIMER_PENDING(&tmr->timer)) { /* * Callout has been rescheduled. */ goto get_out; } if (!SCTP_OS_TIMER_ACTIVE(&tmr->timer)) { /* * Not active, so no action. */ goto get_out; } SCTP_OS_TIMER_DEACTIVATE(&tmr->timer); /* call the handler for the appropriate timer type */ switch (tmr->type) { case SCTP_TIMER_TYPE_ZERO_COPY: if (inp == NULL) { break; } if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_ZERO_COPY_ACTIVE)) { SCTP_ZERO_COPY_EVENT(inp, inp->sctp_socket); } break; case SCTP_TIMER_TYPE_ZCOPY_SENDQ: if (inp == NULL) { break; } if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_ZERO_COPY_ACTIVE)) { SCTP_ZERO_COPY_SENDQ_EVENT(inp, inp->sctp_socket); } break; case SCTP_TIMER_TYPE_ADDR_WQ: sctp_handle_addr_wq(); break; case SCTP_TIMER_TYPE_ITERATOR: SCTP_STAT_INCR(sctps_timoiterator); sctp_iterator_timer(it); break; case SCTP_TIMER_TYPE_SEND: if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timodata); stcb->asoc.timodata++; stcb->asoc.num_send_timers_up--; if (stcb->asoc.num_send_timers_up < 0) { stcb->asoc.num_send_timers_up = 0; } SCTP_TCB_LOCK_ASSERT(stcb); cur_oerr = stcb->asoc.overall_error_count; retcode = sctp_t3rxt_timer(inp, stcb, net); if (retcode) { /* no need to unlock on tcb its gone */ goto out_decr; } SCTP_TCB_LOCK_ASSERT(stcb); #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, net); #endif sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED); if ((stcb->asoc.num_send_timers_up == 0) && (stcb->asoc.sent_queue_cnt > 0) ) { struct sctp_tmit_chunk *chk; /* * safeguard. If there on some on the sent queue * somewhere but no timers running something is * wrong... so we start a timer on the first chunk * on the send queue on whatever net it is sent to. */ chk = TAILQ_FIRST(&stcb->asoc.sent_queue); sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, chk->whoTo); } break; case SCTP_TIMER_TYPE_INIT: if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timoinit); stcb->asoc.timoinit++; if (sctp_t1init_timer(inp, stcb, net)) { /* no need to unlock on tcb its gone */ goto out_decr; } /* We do output but not here */ did_output = 0; break; case SCTP_TIMER_TYPE_RECV: if ((stcb == NULL) || (inp == NULL)) { break; } { int abort_flag; SCTP_STAT_INCR(sctps_timosack); stcb->asoc.timosack++; if (stcb->asoc.cumulative_tsn != stcb->asoc.highest_tsn_inside_map) sctp_sack_check(stcb, 0, 0, &abort_flag); sctp_send_sack(stcb); } #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, net); #endif sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SACK_TMR, SCTP_SO_NOT_LOCKED); break; case SCTP_TIMER_TYPE_SHUTDOWN: if ((stcb == NULL) || (inp == NULL)) { break; } if (sctp_shutdown_timer(inp, stcb, net)) { /* no need to unlock on tcb its gone */ goto out_decr; } SCTP_STAT_INCR(sctps_timoshutdown); stcb->asoc.timoshutdown++; #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, net); #endif sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_NOT_LOCKED); break; case SCTP_TIMER_TYPE_HEARTBEAT: { struct sctp_nets *lnet; int cnt_of_unconf = 0; if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timoheartbeat); stcb->asoc.timoheartbeat++; TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { if ((lnet->dest_state & SCTP_ADDR_UNCONFIRMED) && (lnet->dest_state & SCTP_ADDR_REACHABLE)) { cnt_of_unconf++; } } if (cnt_of_unconf == 0) { if (sctp_heartbeat_timer(inp, stcb, lnet, cnt_of_unconf)) { /* no need to unlock on tcb its gone */ goto out_decr; } } #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, lnet); #endif sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, lnet); sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_HB_TMR, SCTP_SO_NOT_LOCKED); } break; case SCTP_TIMER_TYPE_COOKIE: if ((stcb == NULL) || (inp == NULL)) { break; } if (sctp_cookie_timer(inp, stcb, net)) { /* no need to unlock on tcb its gone */ goto out_decr; } SCTP_STAT_INCR(sctps_timocookie); stcb->asoc.timocookie++; #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, net); #endif /* * We consider T3 and Cookie timer pretty much the same with * respect to where from in chunk_output. */ sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED); break; case SCTP_TIMER_TYPE_NEWCOOKIE: { struct timeval tv; int i, secret; if (inp == NULL) { break; } SCTP_STAT_INCR(sctps_timosecret); (void)SCTP_GETTIME_TIMEVAL(&tv); SCTP_INP_WLOCK(inp); inp->sctp_ep.time_of_secret_change = tv.tv_sec; inp->sctp_ep.last_secret_number = inp->sctp_ep.current_secret_number; inp->sctp_ep.current_secret_number++; if (inp->sctp_ep.current_secret_number >= SCTP_HOW_MANY_SECRETS) { inp->sctp_ep.current_secret_number = 0; } secret = (int)inp->sctp_ep.current_secret_number; for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) { inp->sctp_ep.secret_key[secret][i] = sctp_select_initial_TSN(&inp->sctp_ep); } SCTP_INP_WUNLOCK(inp); sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, stcb, net); } did_output = 0; break; case SCTP_TIMER_TYPE_PATHMTURAISE: if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timopathmtu); sctp_pathmtu_timer(inp, stcb, net); did_output = 0; break; case SCTP_TIMER_TYPE_SHUTDOWNACK: if ((stcb == NULL) || (inp == NULL)) { break; } if (sctp_shutdownack_timer(inp, stcb, net)) { /* no need to unlock on tcb its gone */ goto out_decr; } SCTP_STAT_INCR(sctps_timoshutdownack); stcb->asoc.timoshutdownack++; #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, net); #endif sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SHUT_ACK_TMR, SCTP_SO_NOT_LOCKED); break; case SCTP_TIMER_TYPE_SHUTDOWNGUARD: if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timoshutdownguard); sctp_abort_an_association(inp, stcb, SCTP_SHUTDOWN_GUARD_EXPIRES, NULL, SCTP_SO_NOT_LOCKED); /* no need to unlock on tcb its gone */ goto out_decr; case SCTP_TIMER_TYPE_STRRESET: if ((stcb == NULL) || (inp == NULL)) { break; } if (sctp_strreset_timer(inp, stcb, net)) { /* no need to unlock on tcb its gone */ goto out_decr; } SCTP_STAT_INCR(sctps_timostrmrst); sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_STRRST_TMR, SCTP_SO_NOT_LOCKED); break; case SCTP_TIMER_TYPE_EARLYFR: /* Need to do FR of things for net */ if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timoearlyfr); sctp_early_fr_timer(inp, stcb, net); break; case SCTP_TIMER_TYPE_ASCONF: if ((stcb == NULL) || (inp == NULL)) { break; } if (sctp_asconf_timer(inp, stcb, net)) { /* no need to unlock on tcb its gone */ goto out_decr; } SCTP_STAT_INCR(sctps_timoasconf); #ifdef SCTP_AUDITING_ENABLED sctp_auditing(4, inp, stcb, net); #endif sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_ASCONF_TMR, SCTP_SO_NOT_LOCKED); break; case SCTP_TIMER_TYPE_PRIM_DELETED: if ((stcb == NULL) || (inp == NULL)) { break; } sctp_delete_prim_timer(inp, stcb, net); SCTP_STAT_INCR(sctps_timodelprim); break; case SCTP_TIMER_TYPE_AUTOCLOSE: if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timoautoclose); sctp_autoclose_timer(inp, stcb, net); sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED); did_output = 0; break; case SCTP_TIMER_TYPE_ASOCKILL: if ((stcb == NULL) || (inp == NULL)) { break; } SCTP_STAT_INCR(sctps_timoassockill); /* Can we free it yet? */ SCTP_INP_DECR_REF(inp); sctp_timer_stop(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL, SCTP_FROM_SCTPUTIL + SCTP_LOC_1); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(inp); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTPUTIL + SCTP_LOC_2); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif /* * free asoc, always unlocks (or destroy's) so prevent * duplicate unlock or unlock of a free mtx :-0 */ stcb = NULL; goto out_no_decr; case SCTP_TIMER_TYPE_INPKILL: SCTP_STAT_INCR(sctps_timoinpkill); if (inp == NULL) { break; } /* * special case, take away our increment since WE are the * killer */ SCTP_INP_DECR_REF(inp); sctp_timer_stop(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL, SCTP_FROM_SCTPUTIL + SCTP_LOC_3); sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, SCTP_CALLED_DIRECTLY_NOCMPSET); inp = NULL; goto out_no_decr; default: SCTPDBG(SCTP_DEBUG_TIMER1, "sctp_timeout_handler:unknown timer %d\n", tmr->type); break; }; #ifdef SCTP_AUDITING_ENABLED sctp_audit_log(0xF1, (uint8_t) tmr->type); if (inp) sctp_auditing(5, inp, stcb, net); #endif if ((did_output) && stcb) { /* * Now we need to clean up the control chunk chain if an * ECNE is on it. It must be marked as UNSENT again so next * call will continue to send it until such time that we get * a CWR, to remove it. It is, however, less likely that we * will find a ecn echo on the chain though. */ sctp_fix_ecn_echo(&stcb->asoc); } get_out: if (stcb) { SCTP_TCB_UNLOCK(stcb); } out_decr: if (inp) { SCTP_INP_DECR_REF(inp); } out_no_decr: SCTPDBG(SCTP_DEBUG_TIMER1, "Timer now complete (type %d)\n", type); } void sctp_timer_start(int t_type, struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net) { int to_ticks; struct sctp_timer *tmr; if ((t_type != SCTP_TIMER_TYPE_ADDR_WQ) && (inp == NULL)) return; to_ticks = 0; tmr = NULL; if (stcb) { SCTP_TCB_LOCK_ASSERT(stcb); } switch (t_type) { case SCTP_TIMER_TYPE_ZERO_COPY: tmr = &inp->sctp_ep.zero_copy_timer; to_ticks = SCTP_ZERO_COPY_TICK_DELAY; break; case SCTP_TIMER_TYPE_ZCOPY_SENDQ: tmr = &inp->sctp_ep.zero_copy_sendq_timer; to_ticks = SCTP_ZERO_COPY_SENDQ_TICK_DELAY; break; case SCTP_TIMER_TYPE_ADDR_WQ: /* Only 1 tick away :-) */ tmr = &SCTP_BASE_INFO(addr_wq_timer); to_ticks = SCTP_ADDRESS_TICK_DELAY; break; case SCTP_TIMER_TYPE_ITERATOR: { struct sctp_iterator *it; it = (struct sctp_iterator *)inp; tmr = &it->tmr; to_ticks = SCTP_ITERATOR_TICKS; } break; case SCTP_TIMER_TYPE_SEND: /* Here we use the RTO timer */ { int rto_val; if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; if (net->RTO == 0) { rto_val = stcb->asoc.initial_rto; } else { rto_val = net->RTO; } to_ticks = MSEC_TO_TICKS(rto_val); } break; case SCTP_TIMER_TYPE_INIT: /* * Here we use the INIT timer default usually about 1 * minute. */ if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; if (net->RTO == 0) { to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); } else { to_ticks = MSEC_TO_TICKS(net->RTO); } break; case SCTP_TIMER_TYPE_RECV: /* * Here we use the Delayed-Ack timer value from the inp * ususually about 200ms. */ if (stcb == NULL) { return; } tmr = &stcb->asoc.dack_timer; to_ticks = MSEC_TO_TICKS(stcb->asoc.delayed_ack); break; case SCTP_TIMER_TYPE_SHUTDOWN: /* Here we use the RTO of the destination. */ if ((stcb == NULL) || (net == NULL)) { return; } if (net->RTO == 0) { to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); } else { to_ticks = MSEC_TO_TICKS(net->RTO); } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_HEARTBEAT: /* * the net is used here so that we can add in the RTO. Even * though we use a different timer. We also add the HB timer * PLUS a random jitter. */ if ((inp == NULL) || (stcb == NULL)) { return; } else { uint32_t rndval; uint8_t this_random; int cnt_of_unconf = 0; struct sctp_nets *lnet; TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { if ((lnet->dest_state & SCTP_ADDR_UNCONFIRMED) && (lnet->dest_state & SCTP_ADDR_REACHABLE)) { cnt_of_unconf++; } } if (cnt_of_unconf) { net = lnet = NULL; (void)sctp_heartbeat_timer(inp, stcb, lnet, cnt_of_unconf); } if (stcb->asoc.hb_random_idx > 3) { rndval = sctp_select_initial_TSN(&inp->sctp_ep); memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values)); stcb->asoc.hb_random_idx = 0; } this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx]; stcb->asoc.hb_random_idx++; stcb->asoc.hb_ect_randombit = 0; /* * this_random will be 0 - 256 ms RTO is in ms. */ if ((stcb->asoc.hb_is_disabled) && (cnt_of_unconf == 0)) { return; } if (net) { int delay; delay = stcb->asoc.heart_beat_delay; TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) { if ((lnet->dest_state & SCTP_ADDR_UNCONFIRMED) && ((lnet->dest_state & SCTP_ADDR_OUT_OF_SCOPE) == 0) && (lnet->dest_state & SCTP_ADDR_REACHABLE)) { delay = 0; } } if (net->RTO == 0) { /* Never been checked */ to_ticks = this_random + stcb->asoc.initial_rto + delay; } else { /* set rto_val to the ms */ to_ticks = delay + net->RTO + this_random; } } else { if (cnt_of_unconf) { to_ticks = this_random + stcb->asoc.initial_rto; } else { to_ticks = stcb->asoc.heart_beat_delay + this_random + stcb->asoc.initial_rto; } } /* * Now we must convert the to_ticks that are now in * ms to ticks. */ to_ticks = MSEC_TO_TICKS(to_ticks); tmr = &stcb->asoc.hb_timer; } break; case SCTP_TIMER_TYPE_COOKIE: /* * Here we can use the RTO timer from the network since one * RTT was compelete. If a retran happened then we will be * using the RTO initial value. */ if ((stcb == NULL) || (net == NULL)) { return; } if (net->RTO == 0) { to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); } else { to_ticks = MSEC_TO_TICKS(net->RTO); } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_NEWCOOKIE: /* * nothing needed but the endpoint here ususually about 60 * minutes. */ if (inp == NULL) { return; } tmr = &inp->sctp_ep.signature_change; to_ticks = inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_SIGNATURE]; break; case SCTP_TIMER_TYPE_ASOCKILL: if (stcb == NULL) { return; } tmr = &stcb->asoc.strreset_timer; to_ticks = MSEC_TO_TICKS(SCTP_ASOC_KILL_TIMEOUT); break; case SCTP_TIMER_TYPE_INPKILL: /* * The inp is setup to die. We re-use the signature_chage * timer since that has stopped and we are in the GONE * state. */ if (inp == NULL) { return; } tmr = &inp->sctp_ep.signature_change; to_ticks = MSEC_TO_TICKS(SCTP_INP_KILL_TIMEOUT); break; case SCTP_TIMER_TYPE_PATHMTURAISE: /* * Here we use the value found in the EP for PMTU ususually * about 10 minutes. */ if ((stcb == NULL) || (inp == NULL)) { return; } if (net == NULL) { return; } to_ticks = inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_PMTU]; tmr = &net->pmtu_timer; break; case SCTP_TIMER_TYPE_SHUTDOWNACK: /* Here we use the RTO of the destination */ if ((stcb == NULL) || (net == NULL)) { return; } if (net->RTO == 0) { to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); } else { to_ticks = MSEC_TO_TICKS(net->RTO); } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_SHUTDOWNGUARD: /* * Here we use the endpoints shutdown guard timer usually * about 3 minutes. */ if ((inp == NULL) || (stcb == NULL)) { return; } to_ticks = inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN]; tmr = &stcb->asoc.shut_guard_timer; break; case SCTP_TIMER_TYPE_STRRESET: /* * Here the timer comes from the stcb but its value is from * the net's RTO. */ if ((stcb == NULL) || (net == NULL)) { return; } if (net->RTO == 0) { to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); } else { to_ticks = MSEC_TO_TICKS(net->RTO); } tmr = &stcb->asoc.strreset_timer; break; case SCTP_TIMER_TYPE_EARLYFR: { unsigned int msec; if ((stcb == NULL) || (net == NULL)) { return; } if (net->flight_size > net->cwnd) { /* no need to start */ return; } SCTP_STAT_INCR(sctps_earlyfrstart); if (net->lastsa == 0) { /* Hmm no rtt estimate yet? */ msec = stcb->asoc.initial_rto >> 2; } else { msec = ((net->lastsa >> 2) + net->lastsv) >> 1; } if (msec < SCTP_BASE_SYSCTL(sctp_early_fr_msec)) { msec = SCTP_BASE_SYSCTL(sctp_early_fr_msec); if (msec < SCTP_MINFR_MSEC_FLOOR) { msec = SCTP_MINFR_MSEC_FLOOR; } } to_ticks = MSEC_TO_TICKS(msec); tmr = &net->fr_timer; } break; case SCTP_TIMER_TYPE_ASCONF: /* * Here the timer comes from the stcb but its value is from * the net's RTO. */ if ((stcb == NULL) || (net == NULL)) { return; } if (net->RTO == 0) { to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); } else { to_ticks = MSEC_TO_TICKS(net->RTO); } tmr = &stcb->asoc.asconf_timer; break; case SCTP_TIMER_TYPE_PRIM_DELETED: if ((stcb == NULL) || (net != NULL)) { return; } to_ticks = MSEC_TO_TICKS(stcb->asoc.initial_rto); tmr = &stcb->asoc.delete_prim_timer; break; case SCTP_TIMER_TYPE_AUTOCLOSE: if (stcb == NULL) { return; } if (stcb->asoc.sctp_autoclose_ticks == 0) { /* * Really an error since stcb is NOT set to * autoclose */ return; } to_ticks = stcb->asoc.sctp_autoclose_ticks; tmr = &stcb->asoc.autoclose_timer; break; default: SCTPDBG(SCTP_DEBUG_TIMER1, "%s: Unknown timer type %d\n", __FUNCTION__, t_type); return; break; }; if ((to_ticks <= 0) || (tmr == NULL)) { SCTPDBG(SCTP_DEBUG_TIMER1, "%s: %d:software error to_ticks:%d tmr:%p not set ??\n", __FUNCTION__, t_type, to_ticks, tmr); return; } if (SCTP_OS_TIMER_PENDING(&tmr->timer)) { /* * we do NOT allow you to have it already running. if it is * we leave the current one up unchanged */ return; } /* At this point we can proceed */ if (t_type == SCTP_TIMER_TYPE_SEND) { stcb->asoc.num_send_timers_up++; } tmr->stopped_from = 0; tmr->type = t_type; tmr->ep = (void *)inp; tmr->tcb = (void *)stcb; tmr->net = (void *)net; tmr->self = (void *)tmr; tmr->ticks = sctp_get_tick_count(); (void)SCTP_OS_TIMER_START(&tmr->timer, to_ticks, sctp_timeout_handler, tmr); return; } void sctp_timer_stop(int t_type, struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, uint32_t from) { struct sctp_timer *tmr; if ((t_type != SCTP_TIMER_TYPE_ADDR_WQ) && (inp == NULL)) return; tmr = NULL; if (stcb) { SCTP_TCB_LOCK_ASSERT(stcb); } switch (t_type) { case SCTP_TIMER_TYPE_ZERO_COPY: tmr = &inp->sctp_ep.zero_copy_timer; break; case SCTP_TIMER_TYPE_ZCOPY_SENDQ: tmr = &inp->sctp_ep.zero_copy_sendq_timer; break; case SCTP_TIMER_TYPE_ADDR_WQ: tmr = &SCTP_BASE_INFO(addr_wq_timer); break; case SCTP_TIMER_TYPE_EARLYFR: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->fr_timer; SCTP_STAT_INCR(sctps_earlyfrstop); break; case SCTP_TIMER_TYPE_ITERATOR: { struct sctp_iterator *it; it = (struct sctp_iterator *)inp; tmr = &it->tmr; } break; case SCTP_TIMER_TYPE_SEND: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_INIT: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_RECV: if (stcb == NULL) { return; } tmr = &stcb->asoc.dack_timer; break; case SCTP_TIMER_TYPE_SHUTDOWN: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_HEARTBEAT: if (stcb == NULL) { return; } tmr = &stcb->asoc.hb_timer; break; case SCTP_TIMER_TYPE_COOKIE: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_NEWCOOKIE: /* nothing needed but the endpoint here */ tmr = &inp->sctp_ep.signature_change; /* * We re-use the newcookie timer for the INP kill timer. We * must assure that we do not kill it by accident. */ break; case SCTP_TIMER_TYPE_ASOCKILL: /* * Stop the asoc kill timer. */ if (stcb == NULL) { return; } tmr = &stcb->asoc.strreset_timer; break; case SCTP_TIMER_TYPE_INPKILL: /* * The inp is setup to die. We re-use the signature_chage * timer since that has stopped and we are in the GONE * state. */ tmr = &inp->sctp_ep.signature_change; break; case SCTP_TIMER_TYPE_PATHMTURAISE: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->pmtu_timer; break; case SCTP_TIMER_TYPE_SHUTDOWNACK: if ((stcb == NULL) || (net == NULL)) { return; } tmr = &net->rxt_timer; break; case SCTP_TIMER_TYPE_SHUTDOWNGUARD: if (stcb == NULL) { return; } tmr = &stcb->asoc.shut_guard_timer; break; case SCTP_TIMER_TYPE_STRRESET: if (stcb == NULL) { return; } tmr = &stcb->asoc.strreset_timer; break; case SCTP_TIMER_TYPE_ASCONF: if (stcb == NULL) { return; } tmr = &stcb->asoc.asconf_timer; break; case SCTP_TIMER_TYPE_PRIM_DELETED: if (stcb == NULL) { return; } tmr = &stcb->asoc.delete_prim_timer; break; case SCTP_TIMER_TYPE_AUTOCLOSE: if (stcb == NULL) { return; } tmr = &stcb->asoc.autoclose_timer; break; default: SCTPDBG(SCTP_DEBUG_TIMER1, "%s: Unknown timer type %d\n", __FUNCTION__, t_type); break; }; if (tmr == NULL) { return; } if ((tmr->type != t_type) && tmr->type) { /* * Ok we have a timer that is under joint use. Cookie timer * per chance with the SEND timer. We therefore are NOT * running the timer that the caller wants stopped. So just * return. */ return; } if ((t_type == SCTP_TIMER_TYPE_SEND) && (stcb != NULL)) { stcb->asoc.num_send_timers_up--; if (stcb->asoc.num_send_timers_up < 0) { stcb->asoc.num_send_timers_up = 0; } } tmr->self = NULL; tmr->stopped_from = from; (void)SCTP_OS_TIMER_STOP(&tmr->timer); return; } #ifdef SCTP_USE_ADLER32 static uint32_t update_adler32(uint32_t adler, uint8_t * buf, int32_t len) { uint32_t s1 = adler & 0xffff; uint32_t s2 = (adler >> 16) & 0xffff; int n; for (n = 0; n < len; n++, buf++) { /* s1 = (s1 + buf[n]) % BASE */ /* first we add */ s1 = (s1 + *buf); /* * now if we need to, we do a mod by subtracting. It seems a * bit faster since I really will only ever do one subtract * at the MOST, since buf[n] is a max of 255. */ if (s1 >= SCTP_ADLER32_BASE) { s1 -= SCTP_ADLER32_BASE; } /* s2 = (s2 + s1) % BASE */ /* first we add */ s2 = (s2 + s1); /* * again, it is more efficent (it seems) to subtract since * the most s2 will ever be is (BASE-1 + BASE-1) in the * worse case. This would then be (2 * BASE) - 2, which will * still only do one subtract. On Intel this is much better * to do this way and avoid the divide. Have not -pg'd on * sparc. */ if (s2 >= SCTP_ADLER32_BASE) { s2 -= SCTP_ADLER32_BASE; } } /* Return the adler32 of the bytes buf[0..len-1] */ return ((s2 << 16) + s1); } #endif uint32_t sctp_calculate_len(struct mbuf *m) { uint32_t tlen = 0; struct mbuf *at; at = m; while (at) { tlen += SCTP_BUF_LEN(at); at = SCTP_BUF_NEXT(at); } return (tlen); } #if defined(SCTP_WITH_NO_CSUM) uint32_t sctp_calculate_sum(struct mbuf *m, int32_t * pktlen, uint32_t offset) { /* * given a mbuf chain with a packetheader offset by 'offset' * pointing at a sctphdr (with csum set to 0) go through the chain * of SCTP_BUF_NEXT()'s and calculate the SCTP checksum. This also * has a side bonus as it will calculate the total length of the * mbuf chain. Note: if offset is greater than the total mbuf * length, checksum=1, pktlen=0 is returned (ie. no real error code) */ if (pktlen == NULL) return (0); *pktlen = sctp_calculate_len(m); return (0); } #elif defined(SCTP_USE_INCHKSUM) #include uint32_t sctp_calculate_sum(struct mbuf *m, int32_t * pktlen, uint32_t offset) { /* * given a mbuf chain with a packetheader offset by 'offset' * pointing at a sctphdr (with csum set to 0) go through the chain * of SCTP_BUF_NEXT()'s and calculate the SCTP checksum. This also * has a side bonus as it will calculate the total length of the * mbuf chain. Note: if offset is greater than the total mbuf * length, checksum=1, pktlen=0 is returned (ie. no real error code) */ int32_t tlen = 0; struct mbuf *at; uint32_t the_sum, retsum; at = m; while (at) { tlen += SCTP_BUF_LEN(at); at = SCTP_BUF_NEXT(at); } the_sum = (uint32_t) (in_cksum_skip(m, tlen, offset)); if (pktlen != NULL) *pktlen = (tlen - offset); retsum = htons(the_sum); return (the_sum); } #else uint32_t sctp_calculate_sum(struct mbuf *m, int32_t * pktlen, uint32_t offset) { /* * given a mbuf chain with a packetheader offset by 'offset' * pointing at a sctphdr (with csum set to 0) go through the chain * of SCTP_BUF_NEXT()'s and calculate the SCTP checksum. This also * has a side bonus as it will calculate the total length of the * mbuf chain. Note: if offset is greater than the total mbuf * length, checksum=1, pktlen=0 is returned (ie. no real error code) */ int32_t tlen = 0; #ifdef SCTP_USE_ADLER32 uint32_t base = 1L; #else uint32_t base = 0xffffffff; #endif struct mbuf *at; at = m; /* find the correct mbuf and offset into mbuf */ while ((at != NULL) && (offset > (uint32_t) SCTP_BUF_LEN(at))) { offset -= SCTP_BUF_LEN(at); /* update remaining offset * left */ at = SCTP_BUF_NEXT(at); } while (at != NULL) { if ((SCTP_BUF_LEN(at) - offset) > 0) { #ifdef SCTP_USE_ADLER32 base = update_adler32(base, (unsigned char *)(SCTP_BUF_AT(at, offset)), (unsigned int)(SCTP_BUF_LEN(at) - offset)); #else if ((SCTP_BUF_LEN(at) - offset) < 4) { /* Use old method if less than 4 bytes */ base = old_update_crc32(base, (unsigned char *)(SCTP_BUF_AT(at, offset)), (unsigned int)(SCTP_BUF_LEN(at) - offset)); } else { base = update_crc32(base, (unsigned char *)(SCTP_BUF_AT(at, offset)), (unsigned int)(SCTP_BUF_LEN(at) - offset)); } #endif tlen += SCTP_BUF_LEN(at) - offset; /* we only offset once into the first mbuf */ } if (offset) { if (offset < (uint32_t) SCTP_BUF_LEN(at)) offset = 0; else offset -= SCTP_BUF_LEN(at); } at = SCTP_BUF_NEXT(at); } if (pktlen != NULL) { *pktlen = tlen; } #ifdef SCTP_USE_ADLER32 /* Adler32 */ base = htonl(base); #else /* CRC-32c */ base = sctp_csum_finalize(base); #endif return (base); } #endif void sctp_mtu_size_reset(struct sctp_inpcb *inp, struct sctp_association *asoc, uint32_t mtu) { /* * Reset the P-MTU size on this association, this involves changing * the asoc MTU, going through ANY chunk+overhead larger than mtu to * allow the DF flag to be cleared. */ struct sctp_tmit_chunk *chk; unsigned int eff_mtu, ovh; #ifdef SCTP_PRINT_FOR_B_AND_M SCTP_PRINTF("sctp_mtu_size_reset(%p, asoc:%p mtu:%d\n", inp, asoc, mtu); #endif asoc->smallest_mtu = mtu; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ovh = SCTP_MIN_OVERHEAD; } else { ovh = SCTP_MIN_V4_OVERHEAD; } eff_mtu = mtu - ovh; TAILQ_FOREACH(chk, &asoc->send_queue, sctp_next) { if (chk->send_size > eff_mtu) { chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; } } TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) { if (chk->send_size > eff_mtu) { chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; } } } /* * given an association and starting time of the current RTT period return * RTO in number of msecs net should point to the current network */ uint32_t sctp_calculate_rto(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_nets *net, struct timeval *told, int safe) { /*- * given an association and the starting time of the current RTT * period (in value1/value2) return RTO in number of msecs. */ int calc_time = 0; int o_calctime; uint32_t new_rto = 0; int first_measure = 0; struct timeval now, then, *old; /* Copy it out for sparc64 */ if (safe == sctp_align_unsafe_makecopy) { old = &then; memcpy(&then, told, sizeof(struct timeval)); } else if (safe == sctp_align_safe_nocopy) { old = told; } else { /* error */ SCTP_PRINTF("Huh, bad rto calc call\n"); return (0); } /************************/ /* 1. calculate new RTT */ /************************/ /* get the current time */ (void)SCTP_GETTIME_TIMEVAL(&now); /* compute the RTT value */ if ((u_long)now.tv_sec > (u_long)old->tv_sec) { calc_time = ((u_long)now.tv_sec - (u_long)old->tv_sec) * 1000; if ((u_long)now.tv_usec > (u_long)old->tv_usec) { calc_time += (((u_long)now.tv_usec - (u_long)old->tv_usec) / 1000); } else if ((u_long)now.tv_usec < (u_long)old->tv_usec) { /* Borrow 1,000ms from current calculation */ calc_time -= 1000; /* Add in the slop over */ calc_time += ((int)now.tv_usec / 1000); /* Add in the pre-second ms's */ calc_time += (((int)1000000 - (int)old->tv_usec) / 1000); } } else if ((u_long)now.tv_sec == (u_long)old->tv_sec) { if ((u_long)now.tv_usec > (u_long)old->tv_usec) { calc_time = ((u_long)now.tv_usec - (u_long)old->tv_usec) / 1000; } else if ((u_long)now.tv_usec < (u_long)old->tv_usec) { /* impossible .. garbage in nothing out */ goto calc_rto; } else if ((u_long)now.tv_usec == (u_long)old->tv_usec) { /* * We have to have 1 usec :-D this must be the * loopback. */ calc_time = 1; } else { /* impossible .. garbage in nothing out */ goto calc_rto; } } else { /* Clock wrapped? */ goto calc_rto; } /***************************/ /* 2. update RTTVAR & SRTT */ /***************************/ o_calctime = calc_time; /* this is Van Jacobson's integer version */ if (net->RTO_measured) { calc_time -= (net->lastsa >> SCTP_RTT_SHIFT); /* take away 1/8th when * shift=3 */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_RTTVAR_LOGGING_ENABLE) { rto_logging(net, SCTP_LOG_RTTVAR); } net->prev_rtt = o_calctime; net->lastsa += calc_time; /* add 7/8th into sa when * shift=3 */ if (calc_time < 0) { calc_time = -calc_time; } calc_time -= (net->lastsv >> SCTP_RTT_VAR_SHIFT); /* take away 1/4 when * VAR shift=2 */ net->lastsv += calc_time; if (net->lastsv == 0) { net->lastsv = SCTP_CLOCK_GRANULARITY; } } else { /* First RTO measurment */ net->RTO_measured = 1; net->lastsa = calc_time << SCTP_RTT_SHIFT; /* Multiply by 8 when * shift=3 */ net->lastsv = calc_time; if (net->lastsv == 0) { net->lastsv = SCTP_CLOCK_GRANULARITY; } first_measure = 1; net->prev_rtt = o_calctime; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_RTTVAR_LOGGING_ENABLE) { rto_logging(net, SCTP_LOG_INITIAL_RTT); } } calc_rto: new_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv; if ((new_rto > SCTP_SAT_NETWORK_MIN) && (stcb->asoc.sat_network_lockout == 0)) { stcb->asoc.sat_network = 1; } else if ((!first_measure) && stcb->asoc.sat_network) { stcb->asoc.sat_network = 0; stcb->asoc.sat_network_lockout = 1; } /* bound it, per C6/C7 in Section 5.3.1 */ if (new_rto < stcb->asoc.minrto) { new_rto = stcb->asoc.minrto; } if (new_rto > stcb->asoc.maxrto) { new_rto = stcb->asoc.maxrto; } /* we are now returning the RTO */ return (new_rto); } /* * return a pointer to a contiguous piece of data from the given mbuf chain * starting at 'off' for 'len' bytes. If the desired piece spans more than * one mbuf, a copy is made at 'ptr'. caller must ensure that the buffer size * is >= 'len' returns NULL if there there isn't 'len' bytes in the chain. */ caddr_t sctp_m_getptr(struct mbuf *m, int off, int len, uint8_t * in_ptr) { uint32_t count; uint8_t *ptr; ptr = in_ptr; if ((off < 0) || (len <= 0)) return (NULL); /* find the desired start location */ while ((m != NULL) && (off > 0)) { if (off < SCTP_BUF_LEN(m)) break; off -= SCTP_BUF_LEN(m); m = SCTP_BUF_NEXT(m); } if (m == NULL) return (NULL); /* is the current mbuf large enough (eg. contiguous)? */ if ((SCTP_BUF_LEN(m) - off) >= len) { return (mtod(m, caddr_t)+off); } else { /* else, it spans more than one mbuf, so save a temp copy... */ while ((m != NULL) && (len > 0)) { count = min(SCTP_BUF_LEN(m) - off, len); bcopy(mtod(m, caddr_t)+off, ptr, count); len -= count; ptr += count; off = 0; m = SCTP_BUF_NEXT(m); } if ((m == NULL) && (len > 0)) return (NULL); else return ((caddr_t)in_ptr); } } struct sctp_paramhdr * sctp_get_next_param(struct mbuf *m, int offset, struct sctp_paramhdr *pull, int pull_limit) { /* This just provides a typed signature to Peter's Pull routine */ return ((struct sctp_paramhdr *)sctp_m_getptr(m, offset, pull_limit, (uint8_t *) pull)); } int sctp_add_pad_tombuf(struct mbuf *m, int padlen) { /* * add padlen bytes of 0 filled padding to the end of the mbuf. If * padlen is > 3 this routine will fail. */ uint8_t *dp; int i; if (padlen > 3) { SCTP_LTRACE_ERR_RET_PKT(m, NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, ENOBUFS); return (ENOBUFS); } if (padlen <= M_TRAILINGSPACE(m)) { /* * The easy way. We hope the majority of the time we hit * here :) */ dp = (uint8_t *) (mtod(m, caddr_t)+SCTP_BUF_LEN(m)); SCTP_BUF_LEN(m) += padlen; } else { /* Hard way we must grow the mbuf */ struct mbuf *tmp; tmp = sctp_get_mbuf_for_msg(padlen, 0, M_DONTWAIT, 1, MT_DATA); if (tmp == NULL) { /* Out of space GAK! we are in big trouble. */ SCTP_LTRACE_ERR_RET_PKT(m, NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); return (ENOSPC); } /* setup and insert in middle */ SCTP_BUF_LEN(tmp) = padlen; SCTP_BUF_NEXT(tmp) = NULL; SCTP_BUF_NEXT(m) = tmp; dp = mtod(tmp, uint8_t *); } /* zero out the pad */ for (i = 0; i < padlen; i++) { *dp = 0; dp++; } return (0); } int sctp_pad_lastmbuf(struct mbuf *m, int padval, struct mbuf *last_mbuf) { /* find the last mbuf in chain and pad it */ struct mbuf *m_at; m_at = m; if (last_mbuf) { return (sctp_add_pad_tombuf(last_mbuf, padval)); } else { while (m_at) { if (SCTP_BUF_NEXT(m_at) == NULL) { return (sctp_add_pad_tombuf(m_at, padval)); } m_at = SCTP_BUF_NEXT(m_at); } } SCTP_LTRACE_ERR_RET_PKT(m, NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, EFAULT); return (EFAULT); } int sctp_asoc_change_wake = 0; static void sctp_notify_assoc_change(uint32_t event, struct sctp_tcb *stcb, uint32_t error, void *data, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct mbuf *m_notify; struct sctp_assoc_change *sac; struct sctp_queued_to_read *control; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif /* * First if we are are going down dump everything we can to the * socket rcv queue. */ if ((stcb == NULL) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) ) { /* If the socket is gone we are out of here */ return; } /* * For TCP model AND UDP connected sockets we will send an error up * when an ABORT comes in. */ if (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) && ((event == SCTP_COMM_LOST) || (event == SCTP_CANT_STR_ASSOC))) { if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_WAIT) { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ECONNREFUSED); stcb->sctp_socket->so_error = ECONNREFUSED; } else { SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ECONNRESET); stcb->sctp_socket->so_error = ECONNRESET; } /* Wake ANY sleepers */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); if (!so_locked) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { SCTP_SOCKET_UNLOCK(so, 1); return; } } #endif sorwakeup(stcb->sctp_socket); sowwakeup(stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if (!so_locked) { SCTP_SOCKET_UNLOCK(so, 1); } #endif sctp_asoc_change_wake++; } if (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_RECVASSOCEVNT)) { /* event not enabled */ return; } m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_assoc_change), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; SCTP_BUF_LEN(m_notify) = 0; sac = mtod(m_notify, struct sctp_assoc_change *); sac->sac_type = SCTP_ASSOC_CHANGE; sac->sac_flags = 0; sac->sac_length = sizeof(struct sctp_assoc_change); sac->sac_state = event; sac->sac_error = error; /* XXX verify these stream counts */ sac->sac_outbound_streams = stcb->asoc.streamoutcnt; sac->sac_inbound_streams = stcb->asoc.streamincnt; sac->sac_assoc_id = sctp_get_associd(stcb); SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_assoc_change); SCTP_BUF_NEXT(m_notify) = NULL; control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->length = SCTP_BUF_LEN(m_notify); /* not that we need this */ control->tail_mbuf = m_notify; control->spec_flags = M_NOTIFICATION; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, so_locked); if (event == SCTP_COMM_LOST) { /* Wake up any sleeper */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); if (!so_locked) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { SCTP_SOCKET_UNLOCK(so, 1); return; } } #endif sctp_sowwakeup(stcb->sctp_ep, stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if (!so_locked) { SCTP_SOCKET_UNLOCK(so, 1); } #endif } } static void sctp_notify_peer_addr_change(struct sctp_tcb *stcb, uint32_t state, struct sockaddr *sa, uint32_t error) { struct mbuf *m_notify; struct sctp_paddr_change *spc; struct sctp_queued_to_read *control; if ((stcb == NULL) || (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_RECVPADDREVNT))) /* event not enabled */ return; m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_paddr_change), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) return; SCTP_BUF_LEN(m_notify) = 0; spc = mtod(m_notify, struct sctp_paddr_change *); spc->spc_type = SCTP_PEER_ADDR_CHANGE; spc->spc_flags = 0; spc->spc_length = sizeof(struct sctp_paddr_change); switch (sa->sa_family) { case AF_INET: memcpy(&spc->spc_aaddr, sa, sizeof(struct sockaddr_in)); break; #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6; memcpy(&spc->spc_aaddr, sa, sizeof(struct sockaddr_in6)); sin6 = (struct sockaddr_in6 *)&spc->spc_aaddr; if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) { if (sin6->sin6_scope_id == 0) { /* recover scope_id for user */ (void)sa6_recoverscope(sin6); } else { /* clear embedded scope_id for user */ in6_clearscope(&sin6->sin6_addr); } } break; } #endif default: /* TSNH */ break; } spc->spc_state = state; spc->spc_error = error; spc->spc_assoc_id = sctp_get_associd(stcb); SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_paddr_change); SCTP_BUF_NEXT(m_notify) = NULL; /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->length = SCTP_BUF_LEN(m_notify); control->spec_flags = M_NOTIFICATION; /* not that we need this */ control->tail_mbuf = m_notify; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_SO_NOT_LOCKED); } static void sctp_notify_send_failed(struct sctp_tcb *stcb, uint32_t error, struct sctp_tmit_chunk *chk, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct mbuf *m_notify, *tt; struct sctp_send_failed *ssf; struct sctp_queued_to_read *control; int length; if ((stcb == NULL) || (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_RECVSENDFAILEVNT))) /* event not enabled */ return; m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_send_failed), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; length = sizeof(struct sctp_send_failed) + chk->send_size; length -= sizeof(struct sctp_data_chunk); SCTP_BUF_LEN(m_notify) = 0; ssf = mtod(m_notify, struct sctp_send_failed *); ssf->ssf_type = SCTP_SEND_FAILED; if (error == SCTP_NOTIFY_DATAGRAM_UNSENT) ssf->ssf_flags = SCTP_DATA_UNSENT; else ssf->ssf_flags = SCTP_DATA_SENT; ssf->ssf_length = length; ssf->ssf_error = error; /* not exactly what the user sent in, but should be close :) */ bzero(&ssf->ssf_info, sizeof(ssf->ssf_info)); ssf->ssf_info.sinfo_stream = chk->rec.data.stream_number; ssf->ssf_info.sinfo_ssn = chk->rec.data.stream_seq; ssf->ssf_info.sinfo_flags = chk->rec.data.rcv_flags; ssf->ssf_info.sinfo_ppid = chk->rec.data.payloadtype; ssf->ssf_info.sinfo_context = chk->rec.data.context; ssf->ssf_info.sinfo_assoc_id = sctp_get_associd(stcb); ssf->ssf_assoc_id = sctp_get_associd(stcb); /* Take off the chunk header */ m_adj(chk->data, sizeof(struct sctp_data_chunk)); /* trim out any 0 len mbufs */ while (SCTP_BUF_LEN(chk->data) == 0) { tt = chk->data; chk->data = SCTP_BUF_NEXT(tt); SCTP_BUF_NEXT(tt) = NULL; sctp_m_freem(tt); } SCTP_BUF_NEXT(m_notify) = chk->data; SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_send_failed); /* Steal off the mbuf */ chk->data = NULL; /* * For this case, we check the actual socket buffer, since the assoc * is going away we don't want to overfill the socket buffer for a * non-reader */ if (sctp_sbspace_failedmsgs(&stcb->sctp_socket->so_rcv) < SCTP_BUF_LEN(m_notify)) { sctp_m_freem(m_notify); return; } /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->spec_flags = M_NOTIFICATION; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, so_locked); } static void sctp_notify_send_failed2(struct sctp_tcb *stcb, uint32_t error, struct sctp_stream_queue_pending *sp, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct mbuf *m_notify; struct sctp_send_failed *ssf; struct sctp_queued_to_read *control; int length; if ((stcb == NULL) || (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_RECVSENDFAILEVNT))) /* event not enabled */ return; length = sizeof(struct sctp_send_failed) + sp->length; m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_send_failed), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; SCTP_BUF_LEN(m_notify) = 0; ssf = mtod(m_notify, struct sctp_send_failed *); ssf->ssf_type = SCTP_SEND_FAILED; if (error == SCTP_NOTIFY_DATAGRAM_UNSENT) ssf->ssf_flags = SCTP_DATA_UNSENT; else ssf->ssf_flags = SCTP_DATA_SENT; ssf->ssf_length = length; ssf->ssf_error = error; /* not exactly what the user sent in, but should be close :) */ bzero(&ssf->ssf_info, sizeof(ssf->ssf_info)); ssf->ssf_info.sinfo_stream = sp->stream; ssf->ssf_info.sinfo_ssn = sp->strseq; if (sp->some_taken) { ssf->ssf_info.sinfo_flags = SCTP_DATA_LAST_FRAG; } else { ssf->ssf_info.sinfo_flags = SCTP_DATA_NOT_FRAG; } ssf->ssf_info.sinfo_ppid = sp->ppid; ssf->ssf_info.sinfo_context = sp->context; ssf->ssf_info.sinfo_assoc_id = sctp_get_associd(stcb); ssf->ssf_assoc_id = sctp_get_associd(stcb); SCTP_BUF_NEXT(m_notify) = sp->data; SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_send_failed); /* Steal off the mbuf */ sp->data = NULL; /* * For this case, we check the actual socket buffer, since the assoc * is going away we don't want to overfill the socket buffer for a * non-reader */ if (sctp_sbspace_failedmsgs(&stcb->sctp_socket->so_rcv) < SCTP_BUF_LEN(m_notify)) { sctp_m_freem(m_notify); return; } /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->spec_flags = M_NOTIFICATION; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, so_locked); } static void sctp_notify_adaptation_layer(struct sctp_tcb *stcb, uint32_t error) { struct mbuf *m_notify; struct sctp_adaptation_event *sai; struct sctp_queued_to_read *control; if ((stcb == NULL) || (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_ADAPTATIONEVNT))) /* event not enabled */ return; m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_adaption_event), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; SCTP_BUF_LEN(m_notify) = 0; sai = mtod(m_notify, struct sctp_adaptation_event *); sai->sai_type = SCTP_ADAPTATION_INDICATION; sai->sai_flags = 0; sai->sai_length = sizeof(struct sctp_adaptation_event); sai->sai_adaptation_ind = stcb->asoc.peers_adaptation; sai->sai_assoc_id = sctp_get_associd(stcb); SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_adaptation_event); SCTP_BUF_NEXT(m_notify) = NULL; /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->length = SCTP_BUF_LEN(m_notify); control->spec_flags = M_NOTIFICATION; /* not that we need this */ control->tail_mbuf = m_notify; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_SO_NOT_LOCKED); } /* This always must be called with the read-queue LOCKED in the INP */ void sctp_notify_partial_delivery_indication(struct sctp_tcb *stcb, uint32_t error, int nolock, uint32_t val) { struct mbuf *m_notify; struct sctp_pdapi_event *pdapi; struct sctp_queued_to_read *control; struct sockbuf *sb; if ((stcb == NULL) || (stcb->sctp_socket == NULL) || sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_PDAPIEVNT)) /* event not enabled */ return; m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_pdapi_event), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; SCTP_BUF_LEN(m_notify) = 0; pdapi = mtod(m_notify, struct sctp_pdapi_event *); pdapi->pdapi_type = SCTP_PARTIAL_DELIVERY_EVENT; pdapi->pdapi_flags = 0; pdapi->pdapi_length = sizeof(struct sctp_pdapi_event); pdapi->pdapi_indication = error; pdapi->pdapi_stream = (val >> 16); pdapi->pdapi_seq = (val & 0x0000ffff); pdapi->pdapi_assoc_id = sctp_get_associd(stcb); SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_pdapi_event); SCTP_BUF_NEXT(m_notify) = NULL; control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->spec_flags = M_NOTIFICATION; control->length = SCTP_BUF_LEN(m_notify); /* not that we need this */ control->tail_mbuf = m_notify; control->held_length = 0; control->length = 0; if (nolock == 0) { SCTP_INP_READ_LOCK(stcb->sctp_ep); } sb = &stcb->sctp_socket->so_rcv; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(sb, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBALLOC, SCTP_BUF_LEN(m_notify)); } sctp_sballoc(stcb, sb, m_notify); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(sb, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } atomic_add_int(&control->length, SCTP_BUF_LEN(m_notify)); control->end_added = 1; if (stcb->asoc.control_pdapi) TAILQ_INSERT_AFTER(&stcb->sctp_ep->read_queue, stcb->asoc.control_pdapi, control, next); else { /* we really should not see this case */ TAILQ_INSERT_TAIL(&stcb->sctp_ep->read_queue, control, next); } if (nolock == 0) { SCTP_INP_READ_UNLOCK(stcb->sctp_ep); } if (stcb->sctp_ep && stcb->sctp_socket) { /* This should always be the case */ sctp_sorwakeup(stcb->sctp_ep, stcb->sctp_socket); } } static void sctp_notify_shutdown_event(struct sctp_tcb *stcb) { struct mbuf *m_notify; struct sctp_shutdown_event *sse; struct sctp_queued_to_read *control; /* * For TCP model AND UDP connected sockets we will send an error up * when an SHUTDOWN completes */ if (stcb == NULL) { return; } if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { /* mark socket closed for read/write and wakeup! */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; so = SCTP_INP_SO(stcb->sctp_ep); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { SCTP_SOCKET_UNLOCK(so, 1); return; } #endif socantsendmore(stcb->sctp_socket); socantrcvmore(stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } if (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT)) /* event not enabled */ return; m_notify = sctp_get_mbuf_for_msg(sizeof(struct sctp_shutdown_event), 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; sse = mtod(m_notify, struct sctp_shutdown_event *); sse->sse_type = SCTP_SHUTDOWN_EVENT; sse->sse_flags = 0; sse->sse_length = sizeof(struct sctp_shutdown_event); sse->sse_assoc_id = sctp_get_associd(stcb); SCTP_BUF_LEN(m_notify) = sizeof(struct sctp_shutdown_event); SCTP_BUF_NEXT(m_notify) = NULL; /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->spec_flags = M_NOTIFICATION; control->length = SCTP_BUF_LEN(m_notify); /* not that we need this */ control->tail_mbuf = m_notify; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_SO_NOT_LOCKED); } static void sctp_notify_stream_reset(struct sctp_tcb *stcb, int number_entries, uint16_t * list, int flag) { struct mbuf *m_notify; struct sctp_queued_to_read *control; struct sctp_stream_reset_event *strreset; int len; if (stcb == NULL) { return; } if (sctp_is_feature_off(stcb->sctp_ep, SCTP_PCB_FLAGS_STREAM_RESETEVNT)) /* event not enabled */ return; m_notify = sctp_get_mbuf_for_msg(MCLBYTES, 0, M_DONTWAIT, 1, MT_DATA); if (m_notify == NULL) /* no space left */ return; SCTP_BUF_LEN(m_notify) = 0; len = sizeof(struct sctp_stream_reset_event) + (number_entries * sizeof(uint16_t)); if (len > M_TRAILINGSPACE(m_notify)) { /* never enough room */ sctp_m_freem(m_notify); return; } strreset = mtod(m_notify, struct sctp_stream_reset_event *); strreset->strreset_type = SCTP_STREAM_RESET_EVENT; if (number_entries == 0) { strreset->strreset_flags = flag | SCTP_STRRESET_ALL_STREAMS; } else { strreset->strreset_flags = flag | SCTP_STRRESET_STREAM_LIST; } strreset->strreset_length = len; strreset->strreset_assoc_id = sctp_get_associd(stcb); if (number_entries) { int i; for (i = 0; i < number_entries; i++) { strreset->strreset_list[i] = ntohs(list[i]); } } SCTP_BUF_LEN(m_notify) = len; SCTP_BUF_NEXT(m_notify) = NULL; if (sctp_sbspace(&stcb->asoc, &stcb->sctp_socket->so_rcv) < SCTP_BUF_LEN(m_notify)) { /* no space */ sctp_m_freem(m_notify); return; } /* append to socket */ control = sctp_build_readq_entry(stcb, stcb->asoc.primary_destination, 0, 0, 0, 0, 0, 0, m_notify); if (control == NULL) { /* no memory */ sctp_m_freem(m_notify); return; } control->spec_flags = M_NOTIFICATION; control->length = SCTP_BUF_LEN(m_notify); /* not that we need this */ control->tail_mbuf = m_notify; sctp_add_to_readq(stcb->sctp_ep, stcb, control, &stcb->sctp_socket->so_rcv, 1, SCTP_SO_NOT_LOCKED); } void sctp_ulp_notify(uint32_t notification, struct sctp_tcb *stcb, uint32_t error, void *data, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { if (stcb == NULL) { /* unlikely but */ return; } if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) ) { /* No notifications up when we are in a no socket state */ return; } if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { /* Can't send up to a closed socket any notifications */ return; } if (stcb && ((stcb->asoc.state & SCTP_STATE_COOKIE_WAIT) || (stcb->asoc.state & SCTP_STATE_COOKIE_ECHOED))) { if ((notification == SCTP_NOTIFY_INTERFACE_DOWN) || (notification == SCTP_NOTIFY_INTERFACE_UP) || (notification == SCTP_NOTIFY_INTERFACE_CONFIRMED)) { /* Don't report these in front states */ return; } } switch (notification) { case SCTP_NOTIFY_ASSOC_UP: if (stcb->asoc.assoc_up_sent == 0) { sctp_notify_assoc_change(SCTP_COMM_UP, stcb, error, NULL, so_locked); stcb->asoc.assoc_up_sent = 1; } if (stcb->asoc.adaptation_needed && (stcb->asoc.adaptation_sent == 0)) { sctp_notify_adaptation_layer(stcb, error); } break; case SCTP_NOTIFY_ASSOC_DOWN: sctp_notify_assoc_change(SCTP_SHUTDOWN_COMP, stcb, error, NULL, so_locked); break; case SCTP_NOTIFY_INTERFACE_DOWN: { struct sctp_nets *net; net = (struct sctp_nets *)data; sctp_notify_peer_addr_change(stcb, SCTP_ADDR_UNREACHABLE, (struct sockaddr *)&net->ro._l_addr, error); break; } case SCTP_NOTIFY_INTERFACE_UP: { struct sctp_nets *net; net = (struct sctp_nets *)data; sctp_notify_peer_addr_change(stcb, SCTP_ADDR_AVAILABLE, (struct sockaddr *)&net->ro._l_addr, error); break; } case SCTP_NOTIFY_INTERFACE_CONFIRMED: { struct sctp_nets *net; net = (struct sctp_nets *)data; sctp_notify_peer_addr_change(stcb, SCTP_ADDR_CONFIRMED, (struct sockaddr *)&net->ro._l_addr, error); break; } case SCTP_NOTIFY_SPECIAL_SP_FAIL: sctp_notify_send_failed2(stcb, error, (struct sctp_stream_queue_pending *)data, so_locked); break; case SCTP_NOTIFY_DG_FAIL: sctp_notify_send_failed(stcb, error, (struct sctp_tmit_chunk *)data, so_locked); break; case SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION: { uint32_t val; val = *((uint32_t *) data); sctp_notify_partial_delivery_indication(stcb, error, 0, val); } break; case SCTP_NOTIFY_STRDATA_ERR: break; case SCTP_NOTIFY_ASSOC_ABORTED: if ((stcb) && (((stcb->asoc.state & SCTP_STATE_MASK) == SCTP_STATE_COOKIE_WAIT) || ((stcb->asoc.state & SCTP_STATE_MASK) == SCTP_STATE_COOKIE_ECHOED))) { sctp_notify_assoc_change(SCTP_CANT_STR_ASSOC, stcb, error, NULL, so_locked); } else { sctp_notify_assoc_change(SCTP_COMM_LOST, stcb, error, NULL, so_locked); } break; case SCTP_NOTIFY_PEER_OPENED_STREAM: break; case SCTP_NOTIFY_STREAM_OPENED_OK: break; case SCTP_NOTIFY_ASSOC_RESTART: sctp_notify_assoc_change(SCTP_RESTART, stcb, error, data, so_locked); break; case SCTP_NOTIFY_HB_RESP: break; case SCTP_NOTIFY_STR_RESET_SEND: sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), SCTP_STRRESET_OUTBOUND_STR); break; case SCTP_NOTIFY_STR_RESET_RECV: sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), SCTP_STRRESET_INBOUND_STR); break; case SCTP_NOTIFY_STR_RESET_FAILED_OUT: sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), (SCTP_STRRESET_OUTBOUND_STR | SCTP_STRRESET_FAILED)); break; case SCTP_NOTIFY_STR_RESET_FAILED_IN: sctp_notify_stream_reset(stcb, error, ((uint16_t *) data), (SCTP_STRRESET_INBOUND_STR | SCTP_STRRESET_FAILED)); break; case SCTP_NOTIFY_ASCONF_ADD_IP: sctp_notify_peer_addr_change(stcb, SCTP_ADDR_ADDED, data, error); break; case SCTP_NOTIFY_ASCONF_DELETE_IP: sctp_notify_peer_addr_change(stcb, SCTP_ADDR_REMOVED, data, error); break; case SCTP_NOTIFY_ASCONF_SET_PRIMARY: sctp_notify_peer_addr_change(stcb, SCTP_ADDR_MADE_PRIM, data, error); break; case SCTP_NOTIFY_ASCONF_SUCCESS: break; case SCTP_NOTIFY_ASCONF_FAILED: break; case SCTP_NOTIFY_PEER_SHUTDOWN: sctp_notify_shutdown_event(stcb); break; case SCTP_NOTIFY_AUTH_NEW_KEY: sctp_notify_authentication(stcb, SCTP_AUTH_NEWKEY, error, (uint16_t) (uintptr_t) data); break; #if 0 case SCTP_NOTIFY_AUTH_KEY_CONFLICT: sctp_notify_authentication(stcb, SCTP_AUTH_KEY_CONFLICT, error, (uint16_t) (uintptr_t) data); break; #endif /* not yet? remove? */ default: SCTPDBG(SCTP_DEBUG_UTIL1, "%s: unknown notification %xh (%u)\n", __FUNCTION__, notification, notification); break; } /* end switch */ } void sctp_report_all_outbound(struct sctp_tcb *stcb, int holds_lock, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { struct sctp_association *asoc; struct sctp_stream_out *outs; struct sctp_tmit_chunk *chk; struct sctp_stream_queue_pending *sp; int i; asoc = &stcb->asoc; if (stcb == NULL) { return; } if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET)) { return; } /* now through all the gunk freeing chunks */ if (holds_lock == 0) { SCTP_TCB_SEND_LOCK(stcb); } /* sent queue SHOULD be empty */ if (!TAILQ_EMPTY(&asoc->sent_queue)) { chk = TAILQ_FIRST(&asoc->sent_queue); while (chk) { TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next); asoc->sent_queue_cnt--; if (chk->data) { /* * trim off the sctp chunk header(it should * be there) */ if (chk->send_size >= sizeof(struct sctp_data_chunk)) { m_adj(chk->data, sizeof(struct sctp_data_chunk)); sctp_mbuf_crush(chk->data); chk->send_size -= sizeof(struct sctp_data_chunk); } } sctp_free_bufspace(stcb, asoc, chk, 1); sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb, SCTP_NOTIFY_DATAGRAM_SENT, chk, so_locked); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } sctp_free_a_chunk(stcb, chk); /* sa_ignore FREED_MEMORY */ chk = TAILQ_FIRST(&asoc->sent_queue); } } /* pending send queue SHOULD be empty */ if (!TAILQ_EMPTY(&asoc->send_queue)) { chk = TAILQ_FIRST(&asoc->send_queue); while (chk) { TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next); asoc->send_queue_cnt--; if (chk->data) { /* * trim off the sctp chunk header(it should * be there) */ if (chk->send_size >= sizeof(struct sctp_data_chunk)) { m_adj(chk->data, sizeof(struct sctp_data_chunk)); sctp_mbuf_crush(chk->data); chk->send_size -= sizeof(struct sctp_data_chunk); } } sctp_free_bufspace(stcb, asoc, chk, 1); sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb, SCTP_NOTIFY_DATAGRAM_UNSENT, chk, so_locked); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } sctp_free_a_chunk(stcb, chk); /* sa_ignore FREED_MEMORY */ chk = TAILQ_FIRST(&asoc->send_queue); } } for (i = 0; i < stcb->asoc.streamoutcnt; i++) { /* For each stream */ outs = &stcb->asoc.strmout[i]; /* clean up any sends there */ stcb->asoc.locked_on_sending = NULL; sp = TAILQ_FIRST(&outs->outqueue); while (sp) { stcb->asoc.stream_queue_cnt--; TAILQ_REMOVE(&outs->outqueue, sp, next); sctp_free_spbufspace(stcb, asoc, sp); sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb, SCTP_NOTIFY_DATAGRAM_UNSENT, (void *)sp, so_locked); if (sp->data) { sctp_m_freem(sp->data); sp->data = NULL; } if (sp->net) sctp_free_remote_addr(sp->net); sp->net = NULL; /* Free the chunk */ sctp_free_a_strmoq(stcb, sp); /* sa_ignore FREED_MEMORY */ sp = TAILQ_FIRST(&outs->outqueue); } } if (holds_lock == 0) { SCTP_TCB_SEND_UNLOCK(stcb); } } void sctp_abort_notification(struct sctp_tcb *stcb, int error, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { if (stcb == NULL) { return; } if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET)) { return; } /* Tell them we lost the asoc */ sctp_report_all_outbound(stcb, 1, so_locked); if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) { stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_WAS_ABORTED; } sctp_ulp_notify(SCTP_NOTIFY_ASSOC_ABORTED, stcb, error, NULL, so_locked); } void sctp_abort_association(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct mbuf *m, int iphlen, struct sctphdr *sh, struct mbuf *op_err, uint32_t vrf_id, uint16_t port) { uint32_t vtag; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif vtag = 0; if (stcb != NULL) { /* We have a TCB to abort, send notification too */ vtag = stcb->asoc.peer_vtag; sctp_abort_notification(stcb, 0, SCTP_SO_NOT_LOCKED); /* get the assoc vrf id and table id */ vrf_id = stcb->asoc.vrf_id; stcb->asoc.state |= SCTP_STATE_WAS_ABORTED; } sctp_send_abort(m, iphlen, sh, vtag, op_err, vrf_id, port); if (stcb != NULL) { /* Ok, now lets free it */ #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(inp); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTPUTIL + SCTP_LOC_4); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } else { if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { if (LIST_FIRST(&inp->sctp_asoc_list) == NULL) { sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, SCTP_CALLED_DIRECTLY_NOCMPSET); } } } } #ifdef SCTP_ASOCLOG_OF_TSNS void sctp_print_out_track_log(struct sctp_tcb *stcb) { #ifdef NOSIY_PRINTS int i; SCTP_PRINTF("Last ep reason:%x\n", stcb->sctp_ep->last_abort_code); SCTP_PRINTF("IN bound TSN log-aaa\n"); if ((stcb->asoc.tsn_in_at == 0) && (stcb->asoc.tsn_in_wrapped == 0)) { SCTP_PRINTF("None rcvd\n"); goto none_in; } if (stcb->asoc.tsn_in_wrapped) { for (i = stcb->asoc.tsn_in_at; i < SCTP_TSN_LOG_SIZE; i++) { SCTP_PRINTF("TSN:%x strm:%d seq:%d flags:%x sz:%d\n", stcb->asoc.in_tsnlog[i].tsn, stcb->asoc.in_tsnlog[i].strm, stcb->asoc.in_tsnlog[i].seq, stcb->asoc.in_tsnlog[i].flgs, stcb->asoc.in_tsnlog[i].sz); } } if (stcb->asoc.tsn_in_at) { for (i = 0; i < stcb->asoc.tsn_in_at; i++) { SCTP_PRINTF("TSN:%x strm:%d seq:%d flags:%x sz:%d\n", stcb->asoc.in_tsnlog[i].tsn, stcb->asoc.in_tsnlog[i].strm, stcb->asoc.in_tsnlog[i].seq, stcb->asoc.in_tsnlog[i].flgs, stcb->asoc.in_tsnlog[i].sz); } } none_in: SCTP_PRINTF("OUT bound TSN log-aaa\n"); if ((stcb->asoc.tsn_out_at == 0) && (stcb->asoc.tsn_out_wrapped == 0)) { SCTP_PRINTF("None sent\n"); } if (stcb->asoc.tsn_out_wrapped) { for (i = stcb->asoc.tsn_out_at; i < SCTP_TSN_LOG_SIZE; i++) { SCTP_PRINTF("TSN:%x strm:%d seq:%d flags:%x sz:%d\n", stcb->asoc.out_tsnlog[i].tsn, stcb->asoc.out_tsnlog[i].strm, stcb->asoc.out_tsnlog[i].seq, stcb->asoc.out_tsnlog[i].flgs, stcb->asoc.out_tsnlog[i].sz); } } if (stcb->asoc.tsn_out_at) { for (i = 0; i < stcb->asoc.tsn_out_at; i++) { SCTP_PRINTF("TSN:%x strm:%d seq:%d flags:%x sz:%d\n", stcb->asoc.out_tsnlog[i].tsn, stcb->asoc.out_tsnlog[i].strm, stcb->asoc.out_tsnlog[i].seq, stcb->asoc.out_tsnlog[i].flgs, stcb->asoc.out_tsnlog[i].sz); } } #endif } #endif void sctp_abort_an_association(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int error, struct mbuf *op_err, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { uint32_t vtag; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(inp); #endif if (stcb == NULL) { /* Got to have a TCB */ if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { if (LIST_FIRST(&inp->sctp_asoc_list) == NULL) { sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, SCTP_CALLED_DIRECTLY_NOCMPSET); } } return; } else { stcb->asoc.state |= SCTP_STATE_WAS_ABORTED; } vtag = stcb->asoc.peer_vtag; /* notify the ulp */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) sctp_abort_notification(stcb, error, so_locked); /* notify the peer */ #if defined(SCTP_PANIC_ON_ABORT) panic("aborting an association"); #endif sctp_send_abort_tcb(stcb, op_err, so_locked); SCTP_STAT_INCR_COUNTER32(sctps_aborted); if ((SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } /* now free the asoc */ #ifdef SCTP_ASOCLOG_OF_TSNS sctp_print_out_track_log(stcb); #endif #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if (!so_locked) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); } #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTPUTIL + SCTP_LOC_5); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if (!so_locked) { SCTP_SOCKET_UNLOCK(so, 1); } #endif } void sctp_handle_ootb(struct mbuf *m, int iphlen, int offset, struct sctphdr *sh, struct sctp_inpcb *inp, struct mbuf *op_err, uint32_t vrf_id, uint16_t port) { struct sctp_chunkhdr *ch, chunk_buf; unsigned int chk_length; SCTP_STAT_INCR_COUNTER32(sctps_outoftheblue); /* Generate a TO address for future reference */ if (inp && (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { if (LIST_FIRST(&inp->sctp_asoc_list) == NULL) { sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, SCTP_CALLED_DIRECTLY_NOCMPSET); } } ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, sizeof(*ch), (uint8_t *) & chunk_buf); while (ch != NULL) { chk_length = ntohs(ch->chunk_length); if (chk_length < sizeof(*ch)) { /* break to abort land */ break; } switch (ch->chunk_type) { case SCTP_COOKIE_ECHO: /* We hit here only if the assoc is being freed */ return; case SCTP_PACKET_DROPPED: /* we don't respond to pkt-dropped */ return; case SCTP_ABORT_ASSOCIATION: /* we don't respond with an ABORT to an ABORT */ return; case SCTP_SHUTDOWN_COMPLETE: /* * we ignore it since we are not waiting for it and * peer is gone */ return; case SCTP_SHUTDOWN_ACK: sctp_send_shutdown_complete2(m, iphlen, sh, vrf_id, port); return; default: break; } offset += SCTP_SIZE32(chk_length); ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, sizeof(*ch), (uint8_t *) & chunk_buf); } sctp_send_abort(m, iphlen, sh, 0, op_err, vrf_id, port); } /* * check the inbound datagram to make sure there is not an abort inside it, * if there is return 1, else return 0. */ int sctp_is_there_an_abort_here(struct mbuf *m, int iphlen, uint32_t * vtagfill) { struct sctp_chunkhdr *ch; struct sctp_init_chunk *init_chk, chunk_buf; int offset; unsigned int chk_length; offset = iphlen + sizeof(struct sctphdr); ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, sizeof(*ch), (uint8_t *) & chunk_buf); while (ch != NULL) { chk_length = ntohs(ch->chunk_length); if (chk_length < sizeof(*ch)) { /* packet is probably corrupt */ break; } /* we seem to be ok, is it an abort? */ if (ch->chunk_type == SCTP_ABORT_ASSOCIATION) { /* yep, tell them */ return (1); } if (ch->chunk_type == SCTP_INITIATION) { /* need to update the Vtag */ init_chk = (struct sctp_init_chunk *)sctp_m_getptr(m, offset, sizeof(*init_chk), (uint8_t *) & chunk_buf); if (init_chk != NULL) { *vtagfill = ntohl(init_chk->init.initiate_tag); } } /* Nope, move to the next chunk */ offset += SCTP_SIZE32(chk_length); ch = (struct sctp_chunkhdr *)sctp_m_getptr(m, offset, sizeof(*ch), (uint8_t *) & chunk_buf); } return (0); } /* * currently (2/02), ifa_addr embeds scope_id's and don't have sin6_scope_id * set (i.e. it's 0) so, create this function to compare link local scopes */ #ifdef INET6 uint32_t sctp_is_same_scope(struct sockaddr_in6 *addr1, struct sockaddr_in6 *addr2) { struct sockaddr_in6 a, b; /* save copies */ a = *addr1; b = *addr2; if (a.sin6_scope_id == 0) if (sa6_recoverscope(&a)) { /* can't get scope, so can't match */ return (0); } if (b.sin6_scope_id == 0) if (sa6_recoverscope(&b)) { /* can't get scope, so can't match */ return (0); } if (a.sin6_scope_id != b.sin6_scope_id) return (0); return (1); } /* * returns a sockaddr_in6 with embedded scope recovered and removed */ struct sockaddr_in6 * sctp_recover_scope(struct sockaddr_in6 *addr, struct sockaddr_in6 *store) { /* check and strip embedded scope junk */ if (addr->sin6_family == AF_INET6) { if (IN6_IS_SCOPE_LINKLOCAL(&addr->sin6_addr)) { if (addr->sin6_scope_id == 0) { *store = *addr; if (!sa6_recoverscope(store)) { /* use the recovered scope */ addr = store; } } else { /* else, return the original "to" addr */ in6_clearscope(&addr->sin6_addr); } } } return (addr); } #endif /* * are the two addresses the same? currently a "scopeless" check returns: 1 * if same, 0 if not */ int sctp_cmpaddr(struct sockaddr *sa1, struct sockaddr *sa2) { /* must be valid */ if (sa1 == NULL || sa2 == NULL) return (0); /* must be the same family */ if (sa1->sa_family != sa2->sa_family) return (0); switch (sa1->sa_family) { #ifdef INET6 case AF_INET6: { /* IPv6 addresses */ struct sockaddr_in6 *sin6_1, *sin6_2; sin6_1 = (struct sockaddr_in6 *)sa1; sin6_2 = (struct sockaddr_in6 *)sa2; return (SCTP6_ARE_ADDR_EQUAL(sin6_1, sin6_2)); } #endif case AF_INET: { /* IPv4 addresses */ struct sockaddr_in *sin_1, *sin_2; sin_1 = (struct sockaddr_in *)sa1; sin_2 = (struct sockaddr_in *)sa2; return (sin_1->sin_addr.s_addr == sin_2->sin_addr.s_addr); } default: /* we don't do these... */ return (0); } } void sctp_print_address(struct sockaddr *sa) { #ifdef INET6 char ip6buf[INET6_ADDRSTRLEN]; ip6buf[0] = 0; #endif switch (sa->sa_family) { #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)sa; SCTP_PRINTF("IPv6 address: %s:port:%d scope:%u\n", ip6_sprintf(ip6buf, &sin6->sin6_addr), ntohs(sin6->sin6_port), sin6->sin6_scope_id); break; } #endif case AF_INET: { struct sockaddr_in *sin; unsigned char *p; sin = (struct sockaddr_in *)sa; p = (unsigned char *)&sin->sin_addr; SCTP_PRINTF("IPv4 address: %u.%u.%u.%u:%d\n", p[0], p[1], p[2], p[3], ntohs(sin->sin_port)); break; } default: SCTP_PRINTF("?\n"); break; } } void sctp_print_address_pkt(struct ip *iph, struct sctphdr *sh) { switch (iph->ip_v) { case IPVERSION: { struct sockaddr_in lsa, fsa; bzero(&lsa, sizeof(lsa)); lsa.sin_len = sizeof(lsa); lsa.sin_family = AF_INET; lsa.sin_addr = iph->ip_src; lsa.sin_port = sh->src_port; bzero(&fsa, sizeof(fsa)); fsa.sin_len = sizeof(fsa); fsa.sin_family = AF_INET; fsa.sin_addr = iph->ip_dst; fsa.sin_port = sh->dest_port; SCTP_PRINTF("src: "); sctp_print_address((struct sockaddr *)&lsa); SCTP_PRINTF("dest: "); sctp_print_address((struct sockaddr *)&fsa); break; } #ifdef INET6 case IPV6_VERSION >> 4: { struct ip6_hdr *ip6; struct sockaddr_in6 lsa6, fsa6; ip6 = (struct ip6_hdr *)iph; bzero(&lsa6, sizeof(lsa6)); lsa6.sin6_len = sizeof(lsa6); lsa6.sin6_family = AF_INET6; lsa6.sin6_addr = ip6->ip6_src; lsa6.sin6_port = sh->src_port; bzero(&fsa6, sizeof(fsa6)); fsa6.sin6_len = sizeof(fsa6); fsa6.sin6_family = AF_INET6; fsa6.sin6_addr = ip6->ip6_dst; fsa6.sin6_port = sh->dest_port; SCTP_PRINTF("src: "); sctp_print_address((struct sockaddr *)&lsa6); SCTP_PRINTF("dest: "); sctp_print_address((struct sockaddr *)&fsa6); break; } #endif default: /* TSNH */ break; } } void sctp_pull_off_control_to_new_inp(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp, struct sctp_tcb *stcb, int waitflags) { /* * go through our old INP and pull off any control structures that * belong to stcb and move then to the new inp. */ struct socket *old_so, *new_so; struct sctp_queued_to_read *control, *nctl; struct sctp_readhead tmp_queue; struct mbuf *m; int error = 0; old_so = old_inp->sctp_socket; new_so = new_inp->sctp_socket; TAILQ_INIT(&tmp_queue); error = sblock(&old_so->so_rcv, waitflags); if (error) { /* * Gak, can't get sblock, we have a problem. data will be * left stranded.. and we don't dare look at it since the * other thread may be reading something. Oh well, its a * screwed up app that does a peeloff OR a accept while * reading from the main socket... actually its only the * peeloff() case, since I think read will fail on a * listening socket.. */ return; } /* lock the socket buffers */ SCTP_INP_READ_LOCK(old_inp); control = TAILQ_FIRST(&old_inp->read_queue); /* Pull off all for out target stcb */ while (control) { nctl = TAILQ_NEXT(control, next); if (control->stcb == stcb) { /* remove it we want it */ TAILQ_REMOVE(&old_inp->read_queue, control, next); TAILQ_INSERT_TAIL(&tmp_queue, control, next); m = control->data; while (m) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&old_so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBFREE, SCTP_BUF_LEN(m)); } sctp_sbfree(control, stcb, &old_so->so_rcv, m); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&old_so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } m = SCTP_BUF_NEXT(m); } } control = nctl; } SCTP_INP_READ_UNLOCK(old_inp); /* Remove the sb-lock on the old socket */ sbunlock(&old_so->so_rcv); /* Now we move them over to the new socket buffer */ control = TAILQ_FIRST(&tmp_queue); SCTP_INP_READ_LOCK(new_inp); while (control) { nctl = TAILQ_NEXT(control, next); TAILQ_INSERT_TAIL(&new_inp->read_queue, control, next); m = control->data; while (m) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&new_so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBALLOC, SCTP_BUF_LEN(m)); } sctp_sballoc(stcb, &new_so->so_rcv, m); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&new_so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } m = SCTP_BUF_NEXT(m); } control = nctl; } SCTP_INP_READ_UNLOCK(new_inp); } void sctp_add_to_readq(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_queued_to_read *control, struct sockbuf *sb, int end, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { /* * Here we must place the control on the end of the socket read * queue AND increment sb_cc so that select will work properly on * read. */ struct mbuf *m, *prev = NULL; if (inp == NULL) { /* Gak, TSNH!! */ #ifdef INVARIANTS panic("Gak, inp NULL on add_to_readq"); #endif return; } SCTP_INP_READ_LOCK(inp); if (!(control->spec_flags & M_NOTIFICATION)) { atomic_add_int(&inp->total_recvs, 1); if (!control->do_not_ref_stcb) { atomic_add_int(&stcb->total_recvs, 1); } } m = control->data; control->held_length = 0; control->length = 0; while (m) { if (SCTP_BUF_LEN(m) == 0) { /* Skip mbufs with NO length */ if (prev == NULL) { /* First one */ control->data = sctp_m_free(m); m = control->data; } else { SCTP_BUF_NEXT(prev) = sctp_m_free(m); m = SCTP_BUF_NEXT(prev); } if (m == NULL) { control->tail_mbuf = prev;; } continue; } prev = m; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(sb, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBALLOC, SCTP_BUF_LEN(m)); } sctp_sballoc(stcb, sb, m); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(sb, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } atomic_add_int(&control->length, SCTP_BUF_LEN(m)); m = SCTP_BUF_NEXT(m); } if (prev != NULL) { control->tail_mbuf = prev; } else { /* Everything got collapsed out?? */ return; } if (end) { control->end_added = 1; } TAILQ_INSERT_TAIL(&inp->read_queue, control, next); SCTP_INP_READ_UNLOCK(inp); if (inp && inp->sctp_socket) { if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_ZERO_COPY_ACTIVE)) { SCTP_ZERO_COPY_EVENT(inp, inp->sctp_socket); } else { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; so = SCTP_INP_SO(inp); if (!so_locked) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { SCTP_SOCKET_UNLOCK(so, 1); return; } } #endif sctp_sorwakeup(inp, inp->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if (!so_locked) { SCTP_SOCKET_UNLOCK(so, 1); } #endif } } } int sctp_append_to_readq(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_queued_to_read *control, struct mbuf *m, int end, int ctls_cumack, struct sockbuf *sb) { /* * A partial delivery API event is underway. OR we are appending on * the reassembly queue. * * If PDAPI this means we need to add m to the end of the data. * Increase the length in the control AND increment the sb_cc. * Otherwise sb is NULL and all we need to do is put it at the end * of the mbuf chain. */ int len = 0; struct mbuf *mm, *tail = NULL, *prev = NULL; if (inp) { SCTP_INP_READ_LOCK(inp); } if (control == NULL) { get_out: if (inp) { SCTP_INP_READ_UNLOCK(inp); } return (-1); } if (control->end_added) { /* huh this one is complete? */ goto get_out; } mm = m; if (mm == NULL) { goto get_out; } while (mm) { if (SCTP_BUF_LEN(mm) == 0) { /* Skip mbufs with NO lenght */ if (prev == NULL) { /* First one */ m = sctp_m_free(mm); mm = m; } else { SCTP_BUF_NEXT(prev) = sctp_m_free(mm); mm = SCTP_BUF_NEXT(prev); } continue; } prev = mm; len += SCTP_BUF_LEN(mm); if (sb) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(sb, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBALLOC, SCTP_BUF_LEN(mm)); } sctp_sballoc(stcb, sb, mm); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(sb, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } } mm = SCTP_BUF_NEXT(mm); } if (prev) { tail = prev; } else { /* Really there should always be a prev */ if (m == NULL) { /* Huh nothing left? */ #ifdef INVARIANTS panic("Nothing left to add?"); #else goto get_out; #endif } tail = m; } if (control->tail_mbuf) { /* append */ SCTP_BUF_NEXT(control->tail_mbuf) = m; control->tail_mbuf = tail; } else { /* nothing there */ #ifdef INVARIANTS if (control->data != NULL) { panic("This should NOT happen"); } #endif control->data = m; control->tail_mbuf = tail; } atomic_add_int(&control->length, len); if (end) { /* message is complete */ if (stcb && (control == stcb->asoc.control_pdapi)) { stcb->asoc.control_pdapi = NULL; } control->held_length = 0; control->end_added = 1; } if (stcb == NULL) { control->do_not_ref_stcb = 1; } /* * When we are appending in partial delivery, the cum-ack is used * for the actual pd-api highest tsn on this mbuf. The true cum-ack * is populated in the outbound sinfo structure from the true cumack * if the association exists... */ control->sinfo_tsn = control->sinfo_cumtsn = ctls_cumack; if (inp) { SCTP_INP_READ_UNLOCK(inp); } if (inp && inp->sctp_socket) { if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_ZERO_COPY_ACTIVE)) { SCTP_ZERO_COPY_EVENT(inp, inp->sctp_socket); } else { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; so = SCTP_INP_SO(inp); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { SCTP_SOCKET_UNLOCK(so, 1); return (0); } #endif sctp_sorwakeup(inp, inp->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); #endif } } return (0); } /*************HOLD THIS COMMENT FOR PATCH FILE OF *************ALTERNATE ROUTING CODE */ /*************HOLD THIS COMMENT FOR END OF PATCH FILE OF *************ALTERNATE ROUTING CODE */ struct mbuf * sctp_generate_invmanparam(int err) { /* Return a MBUF with a invalid mandatory parameter */ struct mbuf *m; m = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr), 0, M_DONTWAIT, 1, MT_DATA); if (m) { struct sctp_paramhdr *ph; SCTP_BUF_LEN(m) = sizeof(struct sctp_paramhdr); ph = mtod(m, struct sctp_paramhdr *); ph->param_length = htons(sizeof(struct sctp_paramhdr)); ph->param_type = htons(err); } return (m); } #ifdef SCTP_MBCNT_LOGGING void sctp_free_bufspace(struct sctp_tcb *stcb, struct sctp_association *asoc, struct sctp_tmit_chunk *tp1, int chk_cnt) { if (tp1->data == NULL) { return; } asoc->chunks_on_out_queue -= chk_cnt; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBCNT_LOGGING_ENABLE) { sctp_log_mbcnt(SCTP_LOG_MBCNT_DECREASE, asoc->total_output_queue_size, tp1->book_size, 0, tp1->mbcnt); } if (asoc->total_output_queue_size >= tp1->book_size) { atomic_add_int(&asoc->total_output_queue_size, -tp1->book_size); } else { asoc->total_output_queue_size = 0; } if (stcb->sctp_socket && (((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) || ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)))) { if (stcb->sctp_socket->so_snd.sb_cc >= tp1->book_size) { stcb->sctp_socket->so_snd.sb_cc -= tp1->book_size; } else { stcb->sctp_socket->so_snd.sb_cc = 0; } } } #endif int sctp_release_pr_sctp_chunk(struct sctp_tcb *stcb, struct sctp_tmit_chunk *tp1, int reason, struct sctpchunk_listhead *queue, int so_locked #if !defined(__APPLE__) && !defined(SCTP_SO_LOCK_TESTING) SCTP_UNUSED #endif ) { int ret_sz = 0; int notdone; uint8_t foundeom = 0; do { ret_sz += tp1->book_size; tp1->sent = SCTP_FORWARD_TSN_SKIP; if (tp1->data) { #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif sctp_free_bufspace(stcb, &stcb->asoc, tp1, 1); sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb, reason, tp1, SCTP_SO_NOT_LOCKED); sctp_m_freem(tp1->data); tp1->data = NULL; #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(stcb->sctp_ep); if (!so_locked) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & SCTP_STATE_CLOSED_SOCKET) { /* * assoc was freed while we were * unlocked */ SCTP_SOCKET_UNLOCK(so, 1); return (ret_sz); } } #endif sctp_sowwakeup(stcb->sctp_ep, stcb->sctp_socket); #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING) if (!so_locked) { SCTP_SOCKET_UNLOCK(so, 1); } #endif } if (PR_SCTP_BUF_ENABLED(tp1->flags)) { stcb->asoc.sent_queue_cnt_removeable--; } if (queue == &stcb->asoc.send_queue) { TAILQ_REMOVE(&stcb->asoc.send_queue, tp1, sctp_next); /* on to the sent queue */ TAILQ_INSERT_TAIL(&stcb->asoc.sent_queue, tp1, sctp_next); stcb->asoc.sent_queue_cnt++; } if ((tp1->rec.data.rcv_flags & SCTP_DATA_NOT_FRAG) == SCTP_DATA_NOT_FRAG) { /* not frag'ed we ae done */ notdone = 0; foundeom = 1; } else if (tp1->rec.data.rcv_flags & SCTP_DATA_LAST_FRAG) { /* end of frag, we are done */ notdone = 0; foundeom = 1; } else { /* * Its a begin or middle piece, we must mark all of * it */ notdone = 1; tp1 = TAILQ_NEXT(tp1, sctp_next); } } while (tp1 && notdone); if ((foundeom == 0) && (queue == &stcb->asoc.sent_queue)) { /* * The multi-part message was scattered across the send and * sent queue. */ tp1 = TAILQ_FIRST(&stcb->asoc.send_queue); /* * recurse throught the send_queue too, starting at the * beginning. */ if (tp1) { ret_sz += sctp_release_pr_sctp_chunk(stcb, tp1, reason, &stcb->asoc.send_queue, so_locked); } else { SCTP_PRINTF("hmm, nothing on the send queue and no EOM?\n"); } } return (ret_sz); } /* * checks to see if the given address, sa, is one that is currently known by * the kernel note: can't distinguish the same address on multiple interfaces * and doesn't handle multiple addresses with different zone/scope id's note: * ifa_ifwithaddr() compares the entire sockaddr struct */ struct sctp_ifa * sctp_find_ifa_in_ep(struct sctp_inpcb *inp, struct sockaddr *addr, int holds_lock) { struct sctp_laddr *laddr; if (holds_lock == 0) { SCTP_INP_RLOCK(inp); } LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) continue; if (addr->sa_family != laddr->ifa->address.sa.sa_family) continue; if (addr->sa_family == AF_INET) { if (((struct sockaddr_in *)addr)->sin_addr.s_addr == laddr->ifa->address.sin.sin_addr.s_addr) { /* found him. */ if (holds_lock == 0) { SCTP_INP_RUNLOCK(inp); } return (laddr->ifa); break; } } #ifdef INET6 if (addr->sa_family == AF_INET6) { if (SCTP6_ARE_ADDR_EQUAL((struct sockaddr_in6 *)addr, &laddr->ifa->address.sin6)) { /* found him. */ if (holds_lock == 0) { SCTP_INP_RUNLOCK(inp); } return (laddr->ifa); break; } } #endif } if (holds_lock == 0) { SCTP_INP_RUNLOCK(inp); } return (NULL); } uint32_t sctp_get_ifa_hash_val(struct sockaddr *addr) { if (addr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)addr; return (sin->sin_addr.s_addr ^ (sin->sin_addr.s_addr >> 16)); } else if (addr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; uint32_t hash_of_addr; sin6 = (struct sockaddr_in6 *)addr; hash_of_addr = (sin6->sin6_addr.s6_addr32[0] + sin6->sin6_addr.s6_addr32[1] + sin6->sin6_addr.s6_addr32[2] + sin6->sin6_addr.s6_addr32[3]); hash_of_addr = (hash_of_addr ^ (hash_of_addr >> 16)); return (hash_of_addr); } return (0); } struct sctp_ifa * sctp_find_ifa_by_addr(struct sockaddr *addr, uint32_t vrf_id, int holds_lock) { struct sctp_ifa *sctp_ifap; struct sctp_vrf *vrf; struct sctp_ifalist *hash_head; uint32_t hash_of_addr; if (holds_lock == 0) SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { if (holds_lock == 0) SCTP_IPI_ADDR_RUNLOCK(); return (NULL); } hash_of_addr = sctp_get_ifa_hash_val(addr); hash_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)]; if (hash_head == NULL) { SCTP_PRINTF("hash_of_addr:%x mask:%x table:%x - ", hash_of_addr, (uint32_t) vrf->vrf_addr_hashmark, (uint32_t) (hash_of_addr & vrf->vrf_addr_hashmark)); sctp_print_address(addr); SCTP_PRINTF("No such bucket for address\n"); if (holds_lock == 0) SCTP_IPI_ADDR_RUNLOCK(); return (NULL); } LIST_FOREACH(sctp_ifap, hash_head, next_bucket) { if (sctp_ifap == NULL) { panic("Huh LIST_FOREACH corrupt"); } if (addr->sa_family != sctp_ifap->address.sa.sa_family) continue; if (addr->sa_family == AF_INET) { if (((struct sockaddr_in *)addr)->sin_addr.s_addr == sctp_ifap->address.sin.sin_addr.s_addr) { /* found him. */ if (holds_lock == 0) SCTP_IPI_ADDR_RUNLOCK(); return (sctp_ifap); break; } } #ifdef INET6 if (addr->sa_family == AF_INET6) { if (SCTP6_ARE_ADDR_EQUAL((struct sockaddr_in6 *)addr, &sctp_ifap->address.sin6)) { /* found him. */ if (holds_lock == 0) SCTP_IPI_ADDR_RUNLOCK(); return (sctp_ifap); break; } } #endif } if (holds_lock == 0) SCTP_IPI_ADDR_RUNLOCK(); return (NULL); } static void sctp_user_rcvd(struct sctp_tcb *stcb, uint32_t * freed_so_far, int hold_rlock, uint32_t rwnd_req) { /* User pulled some data, do we need a rwnd update? */ int r_unlocked = 0; uint32_t dif, rwnd; struct socket *so = NULL; if (stcb == NULL) return; atomic_add_int(&stcb->asoc.refcnt, 1); if (stcb->asoc.state & (SCTP_STATE_ABOUT_TO_BE_FREED | SCTP_STATE_SHUTDOWN_RECEIVED | SCTP_STATE_SHUTDOWN_ACK_SENT)) { /* Pre-check If we are freeing no update */ goto no_lock; } SCTP_INP_INCR_REF(stcb->sctp_ep); if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { goto out; } so = stcb->sctp_socket; if (so == NULL) { goto out; } atomic_add_int(&stcb->freed_by_sorcv_sincelast, *freed_so_far); /* Have you have freed enough to look */ *freed_so_far = 0; /* Yep, its worth a look and the lock overhead */ /* Figure out what the rwnd would be */ rwnd = sctp_calc_rwnd(stcb, &stcb->asoc); if (rwnd >= stcb->asoc.my_last_reported_rwnd) { dif = rwnd - stcb->asoc.my_last_reported_rwnd; } else { dif = 0; } if (dif >= rwnd_req) { if (hold_rlock) { SCTP_INP_READ_UNLOCK(stcb->sctp_ep); r_unlocked = 1; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { /* * One last check before we allow the guy possibly * to get in. There is a race, where the guy has not * reached the gate. In that case */ goto out; } SCTP_TCB_LOCK(stcb); if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { /* No reports here */ SCTP_TCB_UNLOCK(stcb); goto out; } SCTP_STAT_INCR(sctps_wu_sacks_sent); sctp_send_sack(stcb); sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_USR_RCVD, SCTP_SO_LOCKED); /* make sure no timer is running */ sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL, SCTP_FROM_SCTPUTIL + SCTP_LOC_6); SCTP_TCB_UNLOCK(stcb); } else { /* Update how much we have pending */ stcb->freed_by_sorcv_sincelast = dif; } out: if (so && r_unlocked && hold_rlock) { SCTP_INP_READ_LOCK(stcb->sctp_ep); } SCTP_INP_DECR_REF(stcb->sctp_ep); no_lock: atomic_add_int(&stcb->asoc.refcnt, -1); return; } int sctp_sorecvmsg(struct socket *so, struct uio *uio, struct mbuf **mp, struct sockaddr *from, int fromlen, int *msg_flags, struct sctp_sndrcvinfo *sinfo, int filling_sinfo) { /* * MSG flags we will look at MSG_DONTWAIT - non-blocking IO. * MSG_PEEK - Look don't touch :-D (only valid with OUT mbuf copy * mp=NULL thus uio is the copy method to userland) MSG_WAITALL - ?? * On the way out we may send out any combination of: * MSG_NOTIFICATION MSG_EOR * */ struct sctp_inpcb *inp = NULL; int my_len = 0; int cp_len = 0, error = 0; struct sctp_queued_to_read *control = NULL, *ctl = NULL, *nxt = NULL; struct mbuf *m = NULL, *embuf = NULL; struct sctp_tcb *stcb = NULL; int wakeup_read_socket = 0; int freecnt_applied = 0; int out_flags = 0, in_flags = 0; int block_allowed = 1; uint32_t freed_so_far = 0; uint32_t copied_so_far = 0; int in_eeor_mode = 0; int no_rcv_needed = 0; uint32_t rwnd_req = 0; int hold_sblock = 0; int hold_rlock = 0; int slen = 0; uint32_t held_length = 0; int sockbuf_lock = 0; if (uio == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); return (EINVAL); } if (msg_flags) { in_flags = *msg_flags; if (in_flags & MSG_PEEK) SCTP_STAT_INCR(sctps_read_peeks); } else { in_flags = 0; } slen = uio->uio_resid; /* Pull in and set up our int flags */ if (in_flags & MSG_OOB) { /* Out of band's NOT supported */ return (EOPNOTSUPP); } if ((in_flags & MSG_PEEK) && (mp != NULL)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); return (EINVAL); } if ((in_flags & (MSG_DONTWAIT | MSG_NBIO )) || SCTP_SO_IS_NBIO(so)) { block_allowed = 0; } /* setup the endpoint */ inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, EFAULT); return (EFAULT); } rwnd_req = (SCTP_SB_LIMIT_RCV(so) >> SCTP_RWND_HIWAT_SHIFT); /* Must be at least a MTU's worth */ if (rwnd_req < SCTP_MIN_RWND) rwnd_req = SCTP_MIN_RWND; in_eeor_mode = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_RECV_RWND_LOGGING_ENABLE) { sctp_misc_ints(SCTP_SORECV_ENTER, rwnd_req, in_eeor_mode, so->so_rcv.sb_cc, uio->uio_resid); } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_RECV_RWND_LOGGING_ENABLE) { sctp_misc_ints(SCTP_SORECV_ENTERPL, rwnd_req, block_allowed, so->so_rcv.sb_cc, uio->uio_resid); } error = sblock(&so->so_rcv, (block_allowed ? SBL_WAIT : 0)); sockbuf_lock = 1; if (error) { goto release_unlocked; } restart: restart_nosblocks: if (hold_sblock == 0) { SOCKBUF_LOCK(&so->so_rcv); hold_sblock = 1; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { goto out; } if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { if (so->so_error) { error = so->so_error; if ((in_flags & MSG_PEEK) == 0) so->so_error = 0; goto out; } else { if (so->so_rcv.sb_cc == 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, ENOTCONN); /* indicate EOF */ error = 0; goto out; } } } if ((so->so_rcv.sb_cc <= held_length) && block_allowed) { /* we need to wait for data */ if ((so->so_rcv.sb_cc == 0) && ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL))) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) == 0) { /* * For active open side clear flags for * re-use passive open is blocked by * connect. */ if (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED) { /* * You were aborted, passive side * always hits here */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, ECONNRESET); error = ECONNRESET; /* * You get this once if you are * active open side */ if (!(inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { /* * Remove flag if on the * active open side */ inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAS_ABORTED; } } so->so_state &= ~(SS_ISCONNECTING | SS_ISDISCONNECTING | SS_ISCONFIRMING | SS_ISCONNECTED); if (error == 0) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) == 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, ENOTCONN); error = ENOTCONN; } else { inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAS_CONNECTED; } } goto out; } } error = sbwait(&so->so_rcv); if (error) { goto out; } held_length = 0; goto restart_nosblocks; } else if (so->so_rcv.sb_cc == 0) { if (so->so_error) { error = so->so_error; if ((in_flags & MSG_PEEK) == 0) so->so_error = 0; } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) == 0) { /* * For active open side clear flags * for re-use passive open is * blocked by connect. */ if (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED) { /* * You were aborted, passive * side always hits here */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, ECONNRESET); error = ECONNRESET; /* * You get this once if you * are active open side */ if (!(inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { /* * Remove flag if on * the active open * side */ inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAS_ABORTED; } } so->so_state &= ~(SS_ISCONNECTING | SS_ISDISCONNECTING | SS_ISCONFIRMING | SS_ISCONNECTED); if (error == 0) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) == 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, ENOTCONN); error = ENOTCONN; } else { inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAS_CONNECTED; } } goto out; } } SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EWOULDBLOCK); error = EWOULDBLOCK; } goto out; } if (hold_sblock == 1) { SOCKBUF_UNLOCK(&so->so_rcv); hold_sblock = 0; } /* we possibly have data we can read */ /* sa_ignore FREED_MEMORY */ control = TAILQ_FIRST(&inp->read_queue); if (control == NULL) { /* * This could be happening since the appender did the * increment but as not yet did the tailq insert onto the * read_queue */ if (hold_rlock == 0) { SCTP_INP_READ_LOCK(inp); hold_rlock = 1; } control = TAILQ_FIRST(&inp->read_queue); if ((control == NULL) && (so->so_rcv.sb_cc != 0)) { #ifdef INVARIANTS panic("Huh, its non zero and nothing on control?"); #endif so->so_rcv.sb_cc = 0; } SCTP_INP_READ_UNLOCK(inp); hold_rlock = 0; goto restart; } if ((control->length == 0) && (control->do_not_ref_stcb)) { /* * Clean up code for freeing assoc that left behind a * pdapi.. maybe a peer in EEOR that just closed after * sending and never indicated a EOR. */ if (hold_rlock == 0) { hold_rlock = 1; SCTP_INP_READ_LOCK(inp); } control->held_length = 0; if (control->data) { /* Hmm there is data here .. fix */ struct mbuf *m_tmp; int cnt = 0; m_tmp = control->data; while (m_tmp) { cnt += SCTP_BUF_LEN(m_tmp); if (SCTP_BUF_NEXT(m_tmp) == NULL) { control->tail_mbuf = m_tmp; control->end_added = 1; } m_tmp = SCTP_BUF_NEXT(m_tmp); } control->length = cnt; } else { /* remove it */ TAILQ_REMOVE(&inp->read_queue, control, next); /* Add back any hiddend data */ sctp_free_remote_addr(control->whoFrom); sctp_free_a_readq(stcb, control); } if (hold_rlock) { hold_rlock = 0; SCTP_INP_READ_UNLOCK(inp); } goto restart; } if (control->length == 0) { if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE)) && (filling_sinfo)) { /* find a more suitable one then this */ ctl = TAILQ_NEXT(control, next); while (ctl) { if ((ctl->stcb != control->stcb) && (ctl->length) && (ctl->some_taken || (ctl->spec_flags & M_NOTIFICATION) || ((ctl->do_not_ref_stcb == 0) && (ctl->stcb->asoc.strmin[ctl->sinfo_stream].delivery_started == 0))) ) { /*- * If we have a different TCB next, and there is data * present. If we have already taken some (pdapi), OR we can * ref the tcb and no delivery as started on this stream, we * take it. Note we allow a notification on a different * assoc to be delivered.. */ control = ctl; goto found_one; } else if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS)) && (ctl->length) && ((ctl->some_taken) || ((ctl->do_not_ref_stcb == 0) && ((ctl->spec_flags & M_NOTIFICATION) == 0) && (ctl->stcb->asoc.strmin[ctl->sinfo_stream].delivery_started == 0))) ) { /*- * If we have the same tcb, and there is data present, and we * have the strm interleave feature present. Then if we have * taken some (pdapi) or we can refer to tht tcb AND we have * not started a delivery for this stream, we can take it. * Note we do NOT allow a notificaiton on the same assoc to * be delivered. */ control = ctl; goto found_one; } ctl = TAILQ_NEXT(ctl, next); } } /* * if we reach here, not suitable replacement is available * fragment interleave is NOT on. So stuff the sb_cc * into the our held count, and its time to sleep again. */ held_length = so->so_rcv.sb_cc; control->held_length = so->so_rcv.sb_cc; goto restart; } /* Clear the held length since there is something to read */ control->held_length = 0; if (hold_rlock) { SCTP_INP_READ_UNLOCK(inp); hold_rlock = 0; } found_one: /* * If we reach here, control has a some data for us to read off. * Note that stcb COULD be NULL. */ control->some_taken++; if (hold_sblock) { SOCKBUF_UNLOCK(&so->so_rcv); hold_sblock = 0; } stcb = control->stcb; if (stcb) { if ((control->do_not_ref_stcb == 0) && (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED)) { if (freecnt_applied == 0) stcb = NULL; } else if (control->do_not_ref_stcb == 0) { /* you can't free it on me please */ /* * The lock on the socket buffer protects us so the * free code will stop. But since we used the * socketbuf lock and the sender uses the tcb_lock * to increment, we need to use the atomic add to * the refcnt */ if (freecnt_applied) { #ifdef INVARIANTS panic("refcnt already incremented"); #else printf("refcnt already incremented?\n"); #endif } else { atomic_add_int(&stcb->asoc.refcnt, 1); freecnt_applied = 1; } /* * Setup to remember how much we have not yet told * the peer our rwnd has opened up. Note we grab the * value from the tcb from last time. Note too that * sack sending clears this when a sack is sent, * which is fine. Once we hit the rwnd_req, we then * will go to the sctp_user_rcvd() that will not * lock until it KNOWs it MUST send a WUP-SACK. */ freed_so_far = stcb->freed_by_sorcv_sincelast; stcb->freed_by_sorcv_sincelast = 0; } } if (stcb && ((control->spec_flags & M_NOTIFICATION) == 0) && control->do_not_ref_stcb == 0) { stcb->asoc.strmin[control->sinfo_stream].delivery_started = 1; } /* First lets get off the sinfo and sockaddr info */ if ((sinfo) && filling_sinfo) { memcpy(sinfo, control, sizeof(struct sctp_nonpad_sndrcvinfo)); nxt = TAILQ_NEXT(control, next); if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO)) { struct sctp_extrcvinfo *s_extra; s_extra = (struct sctp_extrcvinfo *)sinfo; if ((nxt) && (nxt->length)) { s_extra->sreinfo_next_flags = SCTP_NEXT_MSG_AVAIL; if (nxt->sinfo_flags & SCTP_UNORDERED) { s_extra->sreinfo_next_flags |= SCTP_NEXT_MSG_IS_UNORDERED; } if (nxt->spec_flags & M_NOTIFICATION) { s_extra->sreinfo_next_flags |= SCTP_NEXT_MSG_IS_NOTIFICATION; } s_extra->sreinfo_next_aid = nxt->sinfo_assoc_id; s_extra->sreinfo_next_length = nxt->length; s_extra->sreinfo_next_ppid = nxt->sinfo_ppid; s_extra->sreinfo_next_stream = nxt->sinfo_stream; if (nxt->tail_mbuf != NULL) { if (nxt->end_added) { s_extra->sreinfo_next_flags |= SCTP_NEXT_MSG_ISCOMPLETE; } } } else { /* * we explicitly 0 this, since the memcpy * got some other things beyond the older * sinfo_ that is on the control's structure * :-D */ nxt = NULL; s_extra->sreinfo_next_flags = SCTP_NO_NEXT_MSG; s_extra->sreinfo_next_aid = 0; s_extra->sreinfo_next_length = 0; s_extra->sreinfo_next_ppid = 0; s_extra->sreinfo_next_stream = 0; } } /* * update off the real current cum-ack, if we have an stcb. */ if ((control->do_not_ref_stcb == 0) && stcb) sinfo->sinfo_cumtsn = stcb->asoc.cumulative_tsn; /* * mask off the high bits, we keep the actual chunk bits in * there. */ sinfo->sinfo_flags &= 0x00ff; if ((control->sinfo_flags >> 8) & SCTP_DATA_UNORDERED) { sinfo->sinfo_flags |= SCTP_UNORDERED; } } #ifdef SCTP_ASOCLOG_OF_TSNS { int index, newindex; struct sctp_pcbtsn_rlog *entry; do { index = inp->readlog_index; newindex = index + 1; if (newindex >= SCTP_READ_LOG_SIZE) { newindex = 0; } } while (atomic_cmpset_int(&inp->readlog_index, index, newindex) == 0); entry = &inp->readlog[index]; entry->vtag = control->sinfo_assoc_id; entry->strm = control->sinfo_stream; entry->seq = control->sinfo_ssn; entry->sz = control->length; entry->flgs = control->sinfo_flags; } #endif if (fromlen && from) { struct sockaddr *to; #ifdef INET cp_len = min((size_t)fromlen, (size_t)control->whoFrom->ro._l_addr.sin.sin_len); memcpy(from, &control->whoFrom->ro._l_addr, cp_len); ((struct sockaddr_in *)from)->sin_port = control->port_from; #else /* No AF_INET use AF_INET6 */ cp_len = min((size_t)fromlen, (size_t)control->whoFrom->ro._l_addr.sin6.sin6_len); memcpy(from, &control->whoFrom->ro._l_addr, cp_len); ((struct sockaddr_in6 *)from)->sin6_port = control->port_from; #endif to = from; #if defined(INET) && defined(INET6) if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) && (to->sa_family == AF_INET) && ((size_t)fromlen >= sizeof(struct sockaddr_in6))) { struct sockaddr_in *sin; struct sockaddr_in6 sin6; sin = (struct sockaddr_in *)to; bzero(&sin6, sizeof(sin6)); sin6.sin6_family = AF_INET6; sin6.sin6_len = sizeof(struct sockaddr_in6); sin6.sin6_addr.s6_addr32[2] = htonl(0xffff); bcopy(&sin->sin_addr, &sin6.sin6_addr.s6_addr32[3], sizeof(sin6.sin6_addr.s6_addr32[3])); sin6.sin6_port = sin->sin_port; memcpy(from, (caddr_t)&sin6, sizeof(sin6)); } #endif #if defined(INET6) { struct sockaddr_in6 lsa6, *to6; to6 = (struct sockaddr_in6 *)to; sctp_recover_scope_mac(to6, (&lsa6)); } #endif } /* now copy out what data we can */ if (mp == NULL) { /* copy out each mbuf in the chain up to length */ get_more_data: m = control->data; while (m) { /* Move out all we can */ cp_len = (int)uio->uio_resid; my_len = (int)SCTP_BUF_LEN(m); if (cp_len > my_len) { /* not enough in this buf */ cp_len = my_len; } if (hold_rlock) { SCTP_INP_READ_UNLOCK(inp); hold_rlock = 0; } if (cp_len > 0) error = uiomove(mtod(m, char *), cp_len, uio); /* re-read */ if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { goto release; } if ((control->do_not_ref_stcb == 0) && stcb && stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { no_rcv_needed = 1; } if (error) { /* error we are out of here */ goto release; } if ((SCTP_BUF_NEXT(m) == NULL) && (cp_len >= SCTP_BUF_LEN(m)) && ((control->end_added == 0) || (control->end_added && (TAILQ_NEXT(control, next) == NULL))) ) { SCTP_INP_READ_LOCK(inp); hold_rlock = 1; } if (cp_len == SCTP_BUF_LEN(m)) { if ((SCTP_BUF_NEXT(m) == NULL) && (control->end_added)) { out_flags |= MSG_EOR; if ((control->do_not_ref_stcb == 0) && ((control->spec_flags & M_NOTIFICATION) == 0)) control->stcb->asoc.strmin[control->sinfo_stream].delivery_started = 0; } if (control->spec_flags & M_NOTIFICATION) { out_flags |= MSG_NOTIFICATION; } /* we ate up the mbuf */ if (in_flags & MSG_PEEK) { /* just looking */ m = SCTP_BUF_NEXT(m); copied_so_far += cp_len; } else { /* dispose of the mbuf */ if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBFREE, SCTP_BUF_LEN(m)); } sctp_sbfree(control, stcb, &so->so_rcv, m); if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } embuf = m; copied_so_far += cp_len; freed_so_far += cp_len; freed_so_far += MSIZE; atomic_subtract_int(&control->length, cp_len); control->data = sctp_m_free(m); m = control->data; /* * been through it all, must hold sb * lock ok to null tail */ if (control->data == NULL) { #ifdef INVARIANTS if ((control->end_added == 0) || (TAILQ_NEXT(control, next) == NULL)) { /* * If the end is not * added, OR the * next is NOT null * we MUST have the * lock. */ if (mtx_owned(&inp->inp_rdata_mtx) == 0) { panic("Hmm we don't own the lock?"); } } #endif control->tail_mbuf = NULL; #ifdef INVARIANTS if ((control->end_added) && ((out_flags & MSG_EOR) == 0)) { panic("end_added, nothing left and no MSG_EOR"); } #endif } } } else { /* Do we need to trim the mbuf? */ if (control->spec_flags & M_NOTIFICATION) { out_flags |= MSG_NOTIFICATION; } if ((in_flags & MSG_PEEK) == 0) { SCTP_BUF_RESV_UF(m, cp_len); SCTP_BUF_LEN(m) -= cp_len; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBFREE, cp_len); } atomic_subtract_int(&so->so_rcv.sb_cc, cp_len); if ((control->do_not_ref_stcb == 0) && stcb) { atomic_subtract_int(&stcb->asoc.sb_cc, cp_len); } copied_so_far += cp_len; embuf = m; freed_so_far += cp_len; freed_so_far += MSIZE; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } atomic_subtract_int(&control->length, cp_len); } else { copied_so_far += cp_len; } } if ((out_flags & MSG_EOR) || (uio->uio_resid == 0)) { break; } if (((stcb) && (in_flags & MSG_PEEK) == 0) && (control->do_not_ref_stcb == 0) && (freed_so_far >= rwnd_req)) { sctp_user_rcvd(stcb, &freed_so_far, hold_rlock, rwnd_req); } } /* end while(m) */ /* * At this point we have looked at it all and we either have * a MSG_EOR/or read all the user wants... * control->length == 0. */ if ((out_flags & MSG_EOR) && ((in_flags & MSG_PEEK) == 0)) { /* we are done with this control */ if (control->length == 0) { if (control->data) { #ifdef INVARIANTS panic("control->data not null at read eor?"); #else SCTP_PRINTF("Strange, data left in the control buffer .. invarients would panic?\n"); sctp_m_freem(control->data); control->data = NULL; #endif } done_with_control: if (TAILQ_NEXT(control, next) == NULL) { /* * If we don't have a next we need a * lock, if there is a next * interrupt is filling ahead of us * and we don't need a lock to * remove this guy (which is the * head of the queue). */ if (hold_rlock == 0) { SCTP_INP_READ_LOCK(inp); hold_rlock = 1; } } TAILQ_REMOVE(&inp->read_queue, control, next); /* Add back any hiddend data */ if (control->held_length) { held_length = 0; control->held_length = 0; wakeup_read_socket = 1; } if (control->aux_data) { sctp_m_free(control->aux_data); control->aux_data = NULL; } no_rcv_needed = control->do_not_ref_stcb; sctp_free_remote_addr(control->whoFrom); control->data = NULL; sctp_free_a_readq(stcb, control); control = NULL; if ((freed_so_far >= rwnd_req) && (no_rcv_needed == 0)) sctp_user_rcvd(stcb, &freed_so_far, hold_rlock, rwnd_req); } else { /* * The user did not read all of this * message, turn off the returned MSG_EOR * since we are leaving more behind on the * control to read. */ #ifdef INVARIANTS if (control->end_added && (control->data == NULL) && (control->tail_mbuf == NULL)) { panic("Gak, control->length is corrupt?"); } #endif no_rcv_needed = control->do_not_ref_stcb; out_flags &= ~MSG_EOR; } } if (out_flags & MSG_EOR) { goto release; } if ((uio->uio_resid == 0) || ((in_eeor_mode) && (copied_so_far >= max(so->so_rcv.sb_lowat, 1))) ) { goto release; } /* * If I hit here the receiver wants more and this message is * NOT done (pd-api). So two questions. Can we block? if not * we are done. Did the user NOT set MSG_WAITALL? */ if (block_allowed == 0) { goto release; } /* * We need to wait for more data a few things: - We don't * sbunlock() so we don't get someone else reading. - We * must be sure to account for the case where what is added * is NOT to our control when we wakeup. */ /* * Do we need to tell the transport a rwnd update might be * needed before we go to sleep? */ if (((stcb) && (in_flags & MSG_PEEK) == 0) && ((freed_so_far >= rwnd_req) && (control->do_not_ref_stcb == 0) && (no_rcv_needed == 0))) { sctp_user_rcvd(stcb, &freed_so_far, hold_rlock, rwnd_req); } wait_some_more: if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { goto release; } if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) goto release; if (hold_rlock == 1) { SCTP_INP_READ_UNLOCK(inp); hold_rlock = 0; } if (hold_sblock == 0) { SOCKBUF_LOCK(&so->so_rcv); hold_sblock = 1; } if ((copied_so_far) && (control->length == 0) && (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE)) ) { goto release; } if (so->so_rcv.sb_cc <= control->held_length) { error = sbwait(&so->so_rcv); if (error) { goto release; } control->held_length = 0; } if (hold_sblock) { SOCKBUF_UNLOCK(&so->so_rcv); hold_sblock = 0; } if (control->length == 0) { /* still nothing here */ if (control->end_added == 1) { /* he aborted, or is done i.e.did a shutdown */ out_flags |= MSG_EOR; if (control->pdapi_aborted) { if ((control->do_not_ref_stcb == 0) && ((control->spec_flags & M_NOTIFICATION) == 0)) control->stcb->asoc.strmin[control->sinfo_stream].delivery_started = 0; out_flags |= MSG_TRUNC; } else { if ((control->do_not_ref_stcb == 0) && ((control->spec_flags & M_NOTIFICATION) == 0)) control->stcb->asoc.strmin[control->sinfo_stream].delivery_started = 0; } goto done_with_control; } if (so->so_rcv.sb_cc > held_length) { control->held_length = so->so_rcv.sb_cc; held_length = 0; } goto wait_some_more; } else if (control->data == NULL) { /* * we must re-sync since data is probably being * added */ SCTP_INP_READ_LOCK(inp); if ((control->length > 0) && (control->data == NULL)) { /* * big trouble.. we have the lock and its * corrupt? */ #ifdef INVARIANTS panic("Impossible data==NULL length !=0"); #endif out_flags |= MSG_EOR; out_flags |= MSG_TRUNC; control->length = 0; SCTP_INP_READ_UNLOCK(inp); goto done_with_control; } SCTP_INP_READ_UNLOCK(inp); /* We will fall around to get more data */ } goto get_more_data; } else { /*- * Give caller back the mbuf chain, * store in uio_resid the length */ wakeup_read_socket = 0; if ((control->end_added == 0) || (TAILQ_NEXT(control, next) == NULL)) { /* Need to get rlock */ if (hold_rlock == 0) { SCTP_INP_READ_LOCK(inp); hold_rlock = 1; } } if (control->end_added) { out_flags |= MSG_EOR; if ((control->do_not_ref_stcb == 0) && ((control->spec_flags & M_NOTIFICATION) == 0)) control->stcb->asoc.strmin[control->sinfo_stream].delivery_started = 0; } if (control->spec_flags & M_NOTIFICATION) { out_flags |= MSG_NOTIFICATION; } uio->uio_resid = control->length; *mp = control->data; m = control->data; while (m) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBFREE, SCTP_BUF_LEN(m)); } sctp_sbfree(control, stcb, &so->so_rcv, m); freed_so_far += SCTP_BUF_LEN(m); freed_so_far += MSIZE; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_SB_LOGGING_ENABLE) { sctp_sblog(&so->so_rcv, control->do_not_ref_stcb ? NULL : stcb, SCTP_LOG_SBRESULT, 0); } m = SCTP_BUF_NEXT(m); } control->data = control->tail_mbuf = NULL; control->length = 0; if (out_flags & MSG_EOR) { /* Done with this control */ goto done_with_control; } } release: if (hold_rlock == 1) { SCTP_INP_READ_UNLOCK(inp); hold_rlock = 0; } if (hold_sblock == 1) { SOCKBUF_UNLOCK(&so->so_rcv); hold_sblock = 0; } sbunlock(&so->so_rcv); sockbuf_lock = 0; release_unlocked: if (hold_sblock) { SOCKBUF_UNLOCK(&so->so_rcv); hold_sblock = 0; } if ((stcb) && (in_flags & MSG_PEEK) == 0) { if ((freed_so_far >= rwnd_req) && (control && (control->do_not_ref_stcb == 0)) && (no_rcv_needed == 0)) sctp_user_rcvd(stcb, &freed_so_far, hold_rlock, rwnd_req); } if (msg_flags) *msg_flags = out_flags; out: if (((out_flags & MSG_EOR) == 0) && ((in_flags & MSG_PEEK) == 0) && (sinfo) && (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO))) { struct sctp_extrcvinfo *s_extra; s_extra = (struct sctp_extrcvinfo *)sinfo; s_extra->sreinfo_next_flags = SCTP_NO_NEXT_MSG; } if (hold_rlock == 1) { SCTP_INP_READ_UNLOCK(inp); hold_rlock = 0; } if (hold_sblock) { SOCKBUF_UNLOCK(&so->so_rcv); hold_sblock = 0; } if (sockbuf_lock) { sbunlock(&so->so_rcv); } if (freecnt_applied) { /* * The lock on the socket buffer protects us so the free * code will stop. But since we used the socketbuf lock and * the sender uses the tcb_lock to increment, we need to use * the atomic add to the refcnt. */ if (stcb == NULL) { panic("stcb for refcnt has gone NULL?"); } atomic_add_int(&stcb->asoc.refcnt, -1); freecnt_applied = 0; /* Save the value back for next time */ stcb->freed_by_sorcv_sincelast = freed_so_far; } if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_RECV_RWND_LOGGING_ENABLE) { if (stcb) { sctp_misc_ints(SCTP_SORECV_DONE, freed_so_far, ((uio) ? (slen - uio->uio_resid) : slen), stcb->asoc.my_rwnd, so->so_rcv.sb_cc); } else { sctp_misc_ints(SCTP_SORECV_DONE, freed_so_far, ((uio) ? (slen - uio->uio_resid) : slen), 0, so->so_rcv.sb_cc); } } if (wakeup_read_socket) { sctp_sorwakeup(inp, so); } return (error); } #ifdef SCTP_MBUF_LOGGING struct mbuf * sctp_m_free(struct mbuf *m) { if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { if (SCTP_BUF_IS_EXTENDED(m)) { sctp_log_mb(m, SCTP_MBUF_IFREE); } } return (m_free(m)); } void sctp_m_freem(struct mbuf *mb) { while (mb != NULL) mb = sctp_m_free(mb); } #endif int sctp_dynamic_set_primary(struct sockaddr *sa, uint32_t vrf_id) { /* * Given a local address. For all associations that holds the * address, request a peer-set-primary. */ struct sctp_ifa *ifa; struct sctp_laddr *wi; ifa = sctp_find_ifa_by_addr(sa, vrf_id, 0); if (ifa == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, EADDRNOTAVAIL); return (EADDRNOTAVAIL); } /* * Now that we have the ifa we must awaken the iterator with this * message. */ wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (wi == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTPUTIL, ENOMEM); return (ENOMEM); } /* Now incr the count and int wi structure */ SCTP_INCR_LADDR_COUNT(); bzero(wi, sizeof(*wi)); (void)SCTP_GETTIME_TIMEVAL(&wi->start_time); wi->ifa = ifa; wi->action = SCTP_SET_PRIM_ADDR; atomic_add_int(&ifa->refcount, 1); /* Now add it to the work queue */ SCTP_IPI_ITERATOR_WQ_LOCK(); /* * Should this really be a tailq? As it is we will process the * newest first :-0 */ LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, (struct sctp_inpcb *)NULL, (struct sctp_tcb *)NULL, (struct sctp_nets *)NULL); SCTP_IPI_ITERATOR_WQ_UNLOCK(); return (0); } int sctp_soreceive(struct socket *so, struct sockaddr **psa, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp) { int error, fromlen; uint8_t sockbuf[256]; struct sockaddr *from; struct sctp_extrcvinfo sinfo; int filling_sinfo = 1; struct sctp_inpcb *inp; inp = (struct sctp_inpcb *)so->so_pcb; /* pickup the assoc we are reading from */ if (inp == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); return (EINVAL); } if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT)) || (controlp == NULL)) { /* user does not want the sndrcv ctl */ filling_sinfo = 0; } if (psa) { from = (struct sockaddr *)sockbuf; fromlen = sizeof(sockbuf); from->sa_len = 0; } else { from = NULL; fromlen = 0; } error = sctp_sorecvmsg(so, uio, mp0, from, fromlen, flagsp, (struct sctp_sndrcvinfo *)&sinfo, filling_sinfo); if ((controlp) && (filling_sinfo)) { /* copy back the sinfo in a CMSG format */ if (filling_sinfo) *controlp = sctp_build_ctl_nchunk(inp, (struct sctp_sndrcvinfo *)&sinfo); else *controlp = NULL; } if (psa) { /* copy back the address info */ if (from && from->sa_len) { *psa = sodupsockaddr(from, M_NOWAIT); } else { *psa = NULL; } } return (error); } int sctp_l_soreceive(struct socket *so, struct sockaddr **name, struct uio *uio, char **controlp, int *controllen, int *flag) { int error, fromlen; uint8_t sockbuf[256]; struct sockaddr *from; struct sctp_extrcvinfo sinfo; int filling_sinfo = 1; struct sctp_inpcb *inp; inp = (struct sctp_inpcb *)so->so_pcb; /* pickup the assoc we are reading from */ if (inp == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); return (EINVAL); } if ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT)) || (controlp == NULL)) { /* user does not want the sndrcv ctl */ filling_sinfo = 0; } if (name) { from = (struct sockaddr *)sockbuf; fromlen = sizeof(sockbuf); from->sa_len = 0; } else { from = NULL; fromlen = 0; } error = sctp_sorecvmsg(so, uio, (struct mbuf **)NULL, from, fromlen, flag, (struct sctp_sndrcvinfo *)&sinfo, filling_sinfo); if ((controlp) && (filling_sinfo)) { /* * copy back the sinfo in a CMSG format note that the caller * has reponsibility for freeing the memory. */ if (filling_sinfo) *controlp = sctp_build_ctl_cchunk(inp, controllen, (struct sctp_sndrcvinfo *)&sinfo); } if (name) { /* copy back the address info */ if (from && from->sa_len) { *name = sodupsockaddr(from, M_WAIT); } else { *name = NULL; } } return (error); } int sctp_connectx_helper_add(struct sctp_tcb *stcb, struct sockaddr *addr, int totaddr, int *error) { int added = 0; int i; struct sctp_inpcb *inp; struct sockaddr *sa; size_t incr = 0; sa = addr; inp = stcb->sctp_ep; *error = 0; for (i = 0; i < totaddr; i++) { if (sa->sa_family == AF_INET) { incr = sizeof(struct sockaddr_in); if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_ADDR_IS_CONFIRMED)) { /* assoc gone no un-lock */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOBUFS); (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_7); *error = ENOBUFS; goto out_now; } added++; } else if (sa->sa_family == AF_INET6) { incr = sizeof(struct sockaddr_in6); if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_ADDR_IS_CONFIRMED)) { /* assoc gone no un-lock */ SCTP_LTRACE_ERR_RET(NULL, stcb, NULL, SCTP_FROM_SCTPUTIL, ENOBUFS); (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_8); *error = ENOBUFS; goto out_now; } added++; } sa = (struct sockaddr *)((caddr_t)sa + incr); } out_now: return (added); } struct sctp_tcb * sctp_connectx_helper_find(struct sctp_inpcb *inp, struct sockaddr *addr, int *totaddr, int *num_v4, int *num_v6, int *error, int limit, int *bad_addr) { struct sockaddr *sa; struct sctp_tcb *stcb = NULL; size_t incr, at, i; at = incr = 0; sa = addr; *error = *num_v6 = *num_v4 = 0; /* account and validate addresses */ for (i = 0; i < (size_t)*totaddr; i++) { if (sa->sa_family == AF_INET) { (*num_v4) += 1; incr = sizeof(struct sockaddr_in); if (sa->sa_len != incr) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; *bad_addr = 1; return (NULL); } } else if (sa->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)sa; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { /* Must be non-mapped for connectx */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; *bad_addr = 1; return (NULL); } (*num_v6) += 1; incr = sizeof(struct sockaddr_in6); if (sa->sa_len != incr) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; *bad_addr = 1; return (NULL); } } else { *totaddr = i; /* we are done */ break; } SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, sa, NULL, NULL, NULL); if (stcb != NULL) { /* Already have or am bring up an association */ return (stcb); } else { SCTP_INP_DECR_REF(inp); } if ((at + incr) > (size_t)limit) { *totaddr = i; break; } sa = (struct sockaddr *)((caddr_t)sa + incr); } return ((struct sctp_tcb *)NULL); } /* * sctp_bindx(ADD) for one address. * assumes all arguments are valid/checked by caller. */ void sctp_bindx_add_address(struct socket *so, struct sctp_inpcb *inp, struct sockaddr *sa, sctp_assoc_t assoc_id, uint32_t vrf_id, int *error, void *p) { struct sockaddr *addr_touse; #ifdef INET6 struct sockaddr_in sin; #endif /* see if we're bound all already! */ if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } addr_touse = sa; #if defined(INET6) && !defined(__Userspace__) /* TODO port in6_sin6_2_sin */ if (sa->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; if (sa->sa_len != sizeof(struct sockaddr_in6)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { /* can only bind v6 on PF_INET6 sockets */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } sin6 = (struct sockaddr_in6 *)addr_touse; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(inp)) { /* can't bind v4-mapped on PF_INET sockets */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } in6_sin6_2_sin(&sin, sin6); addr_touse = (struct sockaddr *)&sin; } } #endif if (sa->sa_family == AF_INET) { if (sa->sa_len != sizeof(struct sockaddr_in)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(inp)) { /* can't bind v4 on PF_INET sockets */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } } if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { if (p == NULL) { /* Can't get proc for Net/Open BSD */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } *error = sctp_inpcb_bind(so, addr_touse, NULL, p); return; } /* * No locks required here since bind and mgmt_ep_sa all do their own * locking. If we do something for the FIX: below we may need to * lock in that case. */ if (assoc_id == 0) { /* add the address */ struct sctp_inpcb *lep; struct sockaddr_in *lsin = (struct sockaddr_in *)addr_touse; /* validate the incoming port */ if ((lsin->sin_port != 0) && (lsin->sin_port != inp->sctp_lport)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } else { /* user specified 0 port, set it to existing port */ lsin->sin_port = inp->sctp_lport; } lep = sctp_pcb_findep(addr_touse, 1, 0, vrf_id); if (lep != NULL) { /* * We must decrement the refcount since we have the * ep already and are binding. No remove going on * here. */ - SCTP_INP_DECR_REF(inp); + SCTP_INP_DECR_REF(lep); } if (lep == inp) { /* already bound to it.. ok */ return; } else if (lep == NULL) { ((struct sockaddr_in *)addr_touse)->sin_port = 0; *error = sctp_addr_mgmt_ep_sa(inp, addr_touse, SCTP_ADD_IP_ADDRESS, vrf_id, NULL); } else { *error = EADDRINUSE; } if (*error) return; } else { /* * FIX: decide whether we allow assoc based bindx */ } } /* * sctp_bindx(DELETE) for one address. * assumes all arguments are valid/checked by caller. */ void sctp_bindx_delete_address(struct socket *so, struct sctp_inpcb *inp, struct sockaddr *sa, sctp_assoc_t assoc_id, uint32_t vrf_id, int *error) { struct sockaddr *addr_touse; #ifdef INET6 struct sockaddr_in sin; #endif /* see if we're bound all already! */ if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } addr_touse = sa; #if defined(INET6) && !defined(__Userspace__) /* TODO port in6_sin6_2_sin */ if (sa->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; if (sa->sa_len != sizeof(struct sockaddr_in6)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { /* can only bind v6 on PF_INET6 sockets */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } sin6 = (struct sockaddr_in6 *)addr_touse; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(inp)) { /* can't bind mapped-v4 on PF_INET sockets */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } in6_sin6_2_sin(&sin, sin6); addr_touse = (struct sockaddr *)&sin; } } #endif if (sa->sa_family == AF_INET) { if (sa->sa_len != sizeof(struct sockaddr_in)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(inp)) { /* can't bind v4 on PF_INET sockets */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTPUTIL, EINVAL); *error = EINVAL; return; } } /* * No lock required mgmt_ep_sa does its own locking. If the FIX: * below is ever changed we may need to lock before calling * association level binding. */ if (assoc_id == 0) { /* delete the address */ *error = sctp_addr_mgmt_ep_sa(inp, addr_touse, SCTP_DEL_IP_ADDRESS, vrf_id, NULL); } else { /* * FIX: decide whether we allow assoc based bindx */ } } /* * returns the valid local address count for an assoc, taking into account * all scoping rules */ int sctp_local_addr_count(struct sctp_tcb *stcb) { int loopback_scope, ipv4_local_scope, local_scope, site_scope; int ipv4_addr_legal, ipv6_addr_legal; struct sctp_vrf *vrf; struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa; int count = 0; /* Turn on all the appropriate scopes */ loopback_scope = stcb->asoc.loopback_scope; ipv4_local_scope = stcb->asoc.ipv4_local_scope; local_scope = stcb->asoc.local_scope; site_scope = stcb->asoc.site_scope; ipv4_addr_legal = ipv6_addr_legal = 0; if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ipv6_addr_legal = 1; if (SCTP_IPV6_V6ONLY(stcb->sctp_ep) == 0) { ipv4_addr_legal = 1; } } else { ipv4_addr_legal = 1; } SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(stcb->asoc.vrf_id); if (vrf == NULL) { /* no vrf, no addresses */ SCTP_IPI_ADDR_RUNLOCK(); return (0); } if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* * bound all case: go through all ifns on the vrf */ LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { if ((loopback_scope == 0) && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { continue; } LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if (sctp_is_addr_restricted(stcb, sctp_ifa)) continue; switch (sctp_ifa->address.sa.sa_family) { case AF_INET: if (ipv4_addr_legal) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)&sctp_ifa->address.sa; if (sin->sin_addr.s_addr == 0) { /* * skip unspecified * addrs */ continue; } if ((ipv4_local_scope == 0) && (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { continue; } /* count this one */ count++; } else { continue; } break; #ifdef INET6 case AF_INET6: if (ipv6_addr_legal) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&sctp_ifa->address.sa; if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { continue; } if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { if (local_scope == 0) continue; if (sin6->sin6_scope_id == 0) { if (sa6_recoverscope(sin6) != 0) /* * * bad * * li * nk * * loc * al * * add * re * ss * */ continue; } } if ((site_scope == 0) && (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { continue; } /* count this one */ count++; } break; #endif default: /* TSNH */ break; } } } } else { /* * subset bound case */ struct sctp_laddr *laddr; LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) { if (sctp_is_addr_restricted(stcb, laddr->ifa)) { continue; } /* count this one */ count++; } } SCTP_IPI_ADDR_RUNLOCK(); return (count); } #if defined(SCTP_LOCAL_TRACE_BUF) void sctp_log_trace(uint32_t subsys, const char *str SCTP_UNUSED, uint32_t a, uint32_t b, uint32_t c, uint32_t d, uint32_t e, uint32_t f) { uint32_t saveindex, newindex; do { saveindex = SCTP_BASE_SYSCTL(sctp_log).index; if (saveindex >= SCTP_MAX_LOGGING_SIZE) { newindex = 1; } else { newindex = saveindex + 1; } } while (atomic_cmpset_int(&SCTP_BASE_SYSCTL(sctp_log).index, saveindex, newindex) == 0); if (saveindex >= SCTP_MAX_LOGGING_SIZE) { saveindex = 0; } SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].timestamp = SCTP_GET_CYCLECOUNT; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].subsys = subsys; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[0] = a; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[1] = b; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[2] = c; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[3] = d; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[4] = e; SCTP_BASE_SYSCTL(sctp_log).entry[saveindex].params[5] = f; } #endif /* We will need to add support * to bind the ports and such here * so we can do UDP tunneling. In * the mean-time, we return error */ void sctp_over_udp_stop(void) { return; } int sctp_over_udp_start(void) { return (-1); }