From 16c90ceeb35fa6183bb1348c38c30d5f13a5be17 Mon Sep 17 00:00:00 2001 From: Christian Brueffer Date: Mon, 18 Jun 2007 10:20:32 +0000 Subject: [PATCH] Major cleanup: mdoc macros, style, typos etc. --- lib/libc/net/sctp_bindx.3 | 21 +++-- lib/libc/net/sctp_connectx.3 | 16 ++-- lib/libc/net/sctp_freepaddrs.3 | 15 ++-- lib/libc/net/sctp_getaddrlen.3 | 25 ++++-- lib/libc/net/sctp_getassocid.3 | 7 +- lib/libc/net/sctp_getpaddrs.3 | 16 ++-- lib/libc/net/sctp_opt_info.3 | 22 +++-- lib/libc/net/sctp_recvmsg.3 | 118 ++++++++++++++----------- lib/libc/net/sctp_send.3 | 144 ++++++++++++++++++------------- lib/libc/net/sctp_sendmsg.3 | 152 ++++++++++++++++++++------------- 10 files changed, 320 insertions(+), 216 deletions(-) diff --git a/lib/libc/net/sctp_bindx.3 b/lib/libc/net/sctp_bindx.3 index 427b2ab205b0..a2a260008bf3 100644 --- a/lib/libc/net/sctp_bindx.3 +++ b/lib/libc/net/sctp_bindx.3 @@ -50,19 +50,24 @@ The .Fn sctp_bindx call binds or unbinds a address or a list of addresses to an -SCTP endpoint. This allows a user to bind a subset of -addresses. The +SCTP endpoint. +This allows a user to bind a subset of +addresses. +The .Fn sctp_bindx call operates similarly to .Fn bind but allows a list of addresses and also allows a bind or an -unbind. The argument +unbind. +The argument .Fa s -must be a valid SCTP socket descriptor. The argument +must be a valid SCTP socket descriptor. +The argument .Fa addrs is a list of addresses (where the list may be only 1 in length) that the user wishes to bind or unbind to the -socket. The argument +socket. +The argument .Fa type must be one of the following values. .Pp @@ -75,7 +80,8 @@ This value indicates that the listed address(es) need to be removed from the endpoint. .Pp Note that when a user adds or deletes an address to an -association if the dynamic address flag "net.inet.sctp.auto_asconf" +association if the dynamic address flag +.Va net.inet.sctp.auto_asconf is enabled any associations in the endpoint will attempt to have the address(es) added dynamically to the existing association. @@ -104,6 +110,5 @@ The argument is not a socket. .El .Sh SEE ALSO -.Xr sctp 4 , .Xr bind 2 , - +.Xr sctp 4 diff --git a/lib/libc/net/sctp_connectx.3 b/lib/libc/net/sctp_connectx.3 index 9d0faf1ffd0e..c0bbc9059878 100644 --- a/lib/libc/net/sctp_connectx.3 +++ b/lib/libc/net/sctp_connectx.3 @@ -49,17 +49,20 @@ The .Fn sctp_connectx call attempts to initiate an association to a peer SCTP -endpoint. The call operates similarly to +endpoint. +The call operates similarly to .Fn connect but it also provides the ability to specify multiple destination -addresses for the peer. This allows a fault tolerant method -of initiating an association. When one of the peers addresses +addresses for the peer. +This allows a fault tolerant method +of initiating an association. +When one of the peers addresses is unreachable, the subsequent listed addresses will also be used -to setup the association with the peer. +to set up the association with the peer. .Pp The user also needs to consider that any address listed in an .Fn sctp_connectx -call is also consider "confirmed". +call is also considered "confirmed". A confirmed address is one in which the SCTP transport will trust is a part of the association and it will not send a confirmation heartbeat to it with @@ -93,6 +96,5 @@ The argument is not a socket. .El .Sh SEE ALSO -.Xr sctp 4 , .Xr connect 2 , - +.Xr sctp 4 diff --git a/lib/libc/net/sctp_freepaddrs.3 b/lib/libc/net/sctp_freepaddrs.3 index 48f7ae05fbec..a6815f4c461a 100644 --- a/lib/libc/net/sctp_freepaddrs.3 +++ b/lib/libc/net/sctp_freepaddrs.3 @@ -36,9 +36,9 @@ .Dt SCTP_FREEPADDRS 3 .Os .Sh NAME -.Nm sctp_freepaddrs +.Nm sctp_freepaddrs , .Nm sctp_freeladdrs -.Nd release the memory returned from a previous call. +.Nd release the memory returned from a previous call .Sh LIBRARY .Lb libc .Sh SYNOPSIS @@ -49,13 +49,12 @@ .Fn sctp_freepaddrs "struct sockaddr *" .Ft void .Fn sctp_freeladdrs "struct sockaddr *" - .Sh DESCRIPTION -The function +The .Fn sctp_freepaddrs and .Fn sctp_freeladdrs -are used to release the memory allocated by previous +functions are used to release the memory allocated by previous calls to .Fn sctp_getpaddrs or @@ -63,9 +62,7 @@ or respectively. .Sh RETURN VALUES none. -.El .Sh SEE ALSO -.Xr sctp 4 , -.Xr sctp_getpaddrs 3 , .Xr sctp_getladdrs 3 , - +.Xr sctp_getpaddrs 3 , +.Xr sctp 4 diff --git a/lib/libc/net/sctp_getaddrlen.3 b/lib/libc/net/sctp_getaddrlen.3 index cea2140a0876..325e2b0365cd 100644 --- a/lib/libc/net/sctp_getaddrlen.3 +++ b/lib/libc/net/sctp_getaddrlen.3 @@ -45,20 +45,27 @@ .In sys/socket.h .In netinet/sctp.h .Ft int -.Fn sctp_getaddrlen "sa_family_t family"" +.Fn sctp_getaddrlen "sa_family_t family" .Sh DESCRIPTION The .Fn sctp_getaddrlen -returns the size of a specific address family. This function +function returns the size of a specific address family. +This function is provided for application binary compatability since it provides the application with the size the operating system -thinks the specific address family is. Note that the function +thinks the specific address family is. +Note that the function will actually create an SCTP socket and then gather the information via a .Fn getsockopt -system calls. If for some reason a SCTP socket cannot -be created or the getsockopt fails, an error will be returned -with errno set as specified in the +system calls. +If for some reason a SCTP socket cannot +be created or the +.Fn getsockopt +call fails, an error will be returned +with +.Va errno +set as specified in the .Fn socket or .Fn getsockopt @@ -69,12 +76,12 @@ system expects for the specific address family or -1. .Sh ERRORS The .Fn sctp_getaddrlen +function can return the following errors. .Bl -tag -width Er .It Bq Er EINVAL The address family specified does NOT exist. .El .Sh SEE ALSO -.Xr sctp 4 , -.Xr socket 2 , .Xr getsockopt 2 , - +.Xr socket 2 , +.Xr sctp 4 diff --git a/lib/libc/net/sctp_getassocid.3 b/lib/libc/net/sctp_getassocid.3 index d54257c0086e..ee0926a8cbaa 100644 --- a/lib/libc/net/sctp_getassocid.3 +++ b/lib/libc/net/sctp_getassocid.3 @@ -48,12 +48,12 @@ .Sh DESCRIPTION The .Fn sctp_getassocid -call attempts to lookup the specified socket address +call attempts to look up the specified socket address .Fa addr and find the respective association identification. .Pp .Sh RETURN VALUES -The call returns the association id upon success and a +The call returns the association id upon success and 0 is returned upon failure. .Sh ERRORS The @@ -72,5 +72,4 @@ The argument is not a socket. .El .Sh SEE ALSO -.Xr sctp 4 , - +.Xr sctp 4 diff --git a/lib/libc/net/sctp_getpaddrs.3 b/lib/libc/net/sctp_getpaddrs.3 index 81171fe5127f..9f6d632cd03d 100644 --- a/lib/libc/net/sctp_getpaddrs.3 +++ b/lib/libc/net/sctp_getpaddrs.3 @@ -36,9 +36,9 @@ .Dt SCTP_GETPADDR 3 .Os .Sh NAME -.Nm sctp_getpaddrs +.Nm sctp_getpaddrs , .Nm sctp_getladdrs -.Nd return a list of addresses to the caller. +.Nd return a list of addresses to the caller .Sh LIBRARY .Lb libc .Sh SYNOPSIS @@ -58,11 +58,12 @@ The function is used to get the list of the local addresses. The association of interest is identified by the .Fa asocid -argument. The addresses are returned in a newly allocated +argument. +The addresses are returned in a newly allocated array of socket addresses returned in the argument .Fa addrs upon success. - .Pp +.Pp After the caller is through the function .Fn sctp_freepaddrs or @@ -93,8 +94,7 @@ The argument is not a socket. .El .Sh SEE ALSO -.Xr sctp 4 , -.Xr sctp_freepaddrs 3 , -.Xr sctp_freeladdrs 3 , .Xr getsockopt 2 , - +.Xr sctp_freeladdrs 3 , +.Xr sctp_freepaddrs 3 , +.Xr sctp 4 diff --git a/lib/libc/net/sctp_opt_info.3 b/lib/libc/net/sctp_opt_info.3 index b51ee858bdec..acfaf1ff9ff9 100644 --- a/lib/libc/net/sctp_opt_info.3 +++ b/lib/libc/net/sctp_opt_info.3 @@ -53,14 +53,21 @@ call provides a multi-os compatible method for getting specific .Fn getsockopt data where an association identification needs to be passed -into the operating system. For FreeBSD a direct +into the operating system. +For +.Fx +a direct .Fn getsockopt -may be used, since FreeBSD has the ability to pass information +may be used, since +.Fx +has the ability to pass information into the operating system on a .Fn getsockopt -call. Other operating systems may not have this ability. For those -who wish to write portable code amongst multiple operating system -this call should be used for the for the following sctp +call. +Other operating systems may not have this ability. +For those +who wish to write portable code amongst multiple operating systems +this call should be used for the following SCTP socket options. .Pp .Dv SCTP_RTOINFO @@ -83,7 +90,7 @@ socket options. .Sh ERRORS The .Fn sctp_opt_info -can return the following errors. +function can return the following errors. .Bl -tag -width Er .It Bq Er EINVAL The argument @@ -104,6 +111,5 @@ The argument is not a socket. .El .Sh SEE ALSO -.Xr sctp 4 , .Xr getsockopt 2 , - +.Xr sctp 4 diff --git a/lib/libc/net/sctp_recvmsg.3 b/lib/libc/net/sctp_recvmsg.3 index 4a085da440e3..b929c9c1a34a 100644 --- a/lib/libc/net/sctp_recvmsg.3 +++ b/lib/libc/net/sctp_recvmsg.3 @@ -44,47 +44,50 @@ .In sys/socket.h .In netinet/sctp.h .Ft ssize_t -.Ft ssize_t -.Fn sctp_recvmsg "int s" "void *msg" "size_t len" "struct sockaddr * restrict from" "socklen_t * restrict fromlen" "struct sctp_sndrcvinfo *sinfo" "int *flags" +.Fo sctp_recvmsg +.Fa "int s" "void *msg" "size_t len" "struct sockaddr * restrict from" +.Fa "socklen_t * restrict fromlen" "struct sctp_sndrcvinfo *sinfo" "int *flags" +.Fc .Sh DESCRIPTION The .Fn sctp_recvmsg -system calls +system call is used to receive a message from another SCTP endpoint. The .Fn sctp_recvmsg call is used by one-to-one (SOCK_STREAM) type sockets after a -sucessful -.Fn connect 2 +successful +.Fn connect call or after the application has performed a .Fn listen followed by a sucessful -.Fn accept 2 -For a one-to-many (SOCK_SEQPACKET)type socket, an endpoint may call +.Fn accept +For a one-to-many (SOCK_SEQPACKET) type socket, an endpoint may call .Fn sctp_recvmsg after having implicitly started an association via one of the send calls including -.Fn sctp_sendmsg 2 -.Fn sendto 2 +.Fn sctp_sendmsg +.Fn sendto and -.Fn sendmsg 2 +.Fn sendmsg . Or, an application may also receive a message after having called -.Fn listen 2 +.Fn listen with a postive backlog to enable the reception of new associations. .Pp -.Pp The address of the sender is held in the .Fa from argument with .Fa fromlen -specifying its size. At the completion of a sucessful +specifying its size. +At the completion of a successful .Fn sctp_recvmsg call .Fa from will hold the address of the peer and .Fa fromlen -will hold the length of that address. Note that +will hold the length of that address. +Note that the address is bounded by the inital value of .Fa fromlen which is used as an in/out variable. @@ -93,36 +96,47 @@ The length of the message .Fa msg to be received is bounded by .Fa len . -If the message is to long to fit in the users +If the message is too long to fit in the users receive buffer, then the .Fa flags -argument will NOT have the MSG_EOF flag -applied. If the message is a complete message then +argument will NOT have the +.Dv MSG_EOF +flag applied. +If the message is a complete message then the .Fa flags -argument will have MSG_EOF set. Locally detected errors are -indicated by a return value of -1 with errno set accordingly. +argument will have +.Dv MSG_EOF +set. +Locally detected errors are +indicated by a return value of -1 with +.Va errno +set accordingly. The .Fa flags -argument may also hold the value MSG_NOTIFICATION. When this +argument may also hold the value +.Dv MSG_NOTIFICATION . +When this occurs this indicates that the message received is NOT from the peer endpoint, but instead is a notification from the SCTP stack (see -.Fn sctp 4 -for more details). Note that no notifications are ever +.Xr sctp 4 +for more details). +Note that no notifications are ever given unless the user subscribes to such notifications using -the SCTP_EVENTS socket option. +the +.Dv SCTP_EVENTS +socket option. .Pp -If no messages is available at the socket then +If no messages are available at the socket then .Fn sctp_recvmsg -normally blocks on the reception of a message or NOTIFICATION, unless the socket has been placed in -non-blocking I/O mode. +normally blocks on the reception of a message or NOTIFICATION, unless the +socket has been placed in non-blocking I/O mode. The .Xr select 2 system call may be used to determine when it is possible to receive a message. .Pp - The .Fa sinfo argument is defined as follows. @@ -139,7 +153,7 @@ struct sctp_sndrcvinfo { sctp_assoc_t sinfo_assoc_id; /* The association id of the peer */ }; .Ed - +.Pp The .Fa sinfo->sinfo_ppid is an opaque 32 bit value that is passed transparently @@ -158,7 +172,8 @@ The .Dv SCTP_UNORDERED flag is used to specify that the message arrived with no specific order and was delivered to the peer application -as soon as possible. When this flag is absent the message +as soon as possible. +When this flag is absent the message was delivered in order within the stream it was received. .Pp .Fa sinfo->sinfo_stream @@ -166,10 +181,12 @@ is the SCTP stream that the message was received on. Streams in SCTP are reliable (or partially reliable) flows of ordered messages. .Pp - The +The .Fa sinfo->sinfo_context -field is used only if the local application set a association level -context with the SCTP_CONTEXT socket option. +field is used only if the local application set an association level +context with the +.Dv SCTP_CONTEXT +socket option. Optionally a user process can use this value to index some application specific data structure for all data coming from a specific association. @@ -183,7 +200,8 @@ For unordered messages this field holds an undefined value. The .Fa sinfo->sinfo_tsn holds a transport sequence number (TSN) that was assigned -to this message by the peer endpoint. For messages that fit in or less +to this message by the peer endpoint. +For messages that fit in or less than the path MTU this will be the only TSN assigned. Note that for messages that span multiple TSN's this value will be one of the TSN's that was used on the @@ -192,22 +210,25 @@ message. The .Fa sinfo->sinfo_cumtsn holds the current cumulative acknowledgment point of -the transport association. Note that this may be larger +the transport association. +Note that this may be larger or smaller than the TSN assigned to the message itself. .Pp The -sinfo->sinfo_assoc_id +.Fa sinfo->sinfo_assoc_id is the unique association identification that was assigned -to the association. For one-to-many (SOCK_SEQPACKET) type +to the association. +For one-to-many (SOCK_SEQPACKET) type sockets this value can be used to send data to the peer without -the use of an address field. It is also quite useful in +the use of an address field. +It is also quite useful in setting various socket options on the specific association (see -.Fn sctp 4 +.Xr sctp 4 ). .Pp The -sinfo->info_timetolive +.Fa sinfo->info_timetolive field is not used by .Fa sctp_recvmsg . .Sh RETURN VALUES @@ -217,7 +238,7 @@ if an error occurred. The .Fn sctp_recvmsg system call -fail if: +fails if: .Bl -tag -width Er .It Bq Er EBADF An invalid descriptor was specified. @@ -243,7 +264,7 @@ but may be caused by transient congestion. .It Bq Er EHOSTUNREACH The remote host was unreachable. .It Bq Er ENOTCON -On a one to one style socket no association exists. +On a one-to-one style socket no association exists. .It Bq Er ECONNRESET An abort was received by the stack while the user was attempting to send data to the peer. @@ -259,14 +280,13 @@ This typically means that the socket is not connected and is a one-to-one style socket. .El .Sh SEE ALSO -.Xr sctp 4 , -.Xr sendmsg 3 , -.Xr sctp_sendmsg 3 , -.Xr sctp_send 3 , -.Xr getsockopt 2 , -.Xr setsockopt 2 , .Xr recv 2 , .Xr select 2 , .Xr socket 2 , -.Xr write 2 - +.Xr write 2 , +.Xr getsockopt 2 , +.Xr setsockopt 2 , +.Xr sctp_send 3 , +.Xr sctp_sendmsg 3 , +.Xr sendmsg 3 , +.Xr sctp 4 diff --git a/lib/libc/net/sctp_send.3 b/lib/libc/net/sctp_send.3 index db31cb8ecfd4..c44e992fd99f 100644 --- a/lib/libc/net/sctp_send.3 +++ b/lib/libc/net/sctp_send.3 @@ -35,7 +35,7 @@ .Dt SCTP_SEND 3 .Os .Sh NAME -.Nm sctp_send +.Nm sctp_send , .Nm sctp_sendx .Nd send a message from an SCTP socket .Sh LIBRARY @@ -45,15 +45,20 @@ .In sys/socket.h .In netinet/sctp.h .Ft ssize_t -.Fn sctp_send "int sd" "const void *msg" "size_t len" "const struct sctp_sndrcvinfo *sinfo" "int flags" +.Fo sctp_send +.Fa "int sd" "const void *msg" "size_t len" +.Fa "const struct sctp_sndrcvinfo *sinfo" "int flags" +.Fc .Ft ssize_t -.Fn sctp_sendx "int sd" "const void *msg" "size_t len" "struct sockaddr *addrs" "int addrcnt" "const struct sctp_sndrcvinfo *sinfo" "int flags" +.Fo sctp_sendx +.Fa "int sd" "const void *msg" "size_t len" "struct sockaddr *addrs" +.Fa "int addrcnt" "const struct sctp_sndrcvinfo *sinfo" "int flags" +.Fc .Sh DESCRIPTION The .Fn sctp_send -system calls +system call is used to transmit a message to another SCTP endpoint. -The .Fn sctp_send may be used to send data to an existing association for both one-to-many (SOCK_SEQPACKET) and one-to-one (SOCK_STREAM) socket types. @@ -62,9 +67,11 @@ The length of the message is given by .Fa len . If the message is too long to pass atomically through the -underlying protocol, the errno is set to -.Er EMSGSIZE -a -1 is returned, and +underlying protocol, +.Va errno +is set to +.Er EMSGSIZE , +-1 is returned, and the message is not transmitted. .Pp No indication of failure to deliver is implicit in a @@ -77,7 +84,7 @@ the message to be transmitted, then normally blocks, unless the socket has been placed in non-blocking I/O mode. The -.Fn select 2 +.Xr select 2 system call may be used to determine when it is possible to send more data on one-to-one type (SOCK_STREAM) sockets. .Pp @@ -98,13 +105,15 @@ struct sctp_sndrcvinfo { sctp_assoc_t sinfo_assoc_id; /* The association id */ }; .Ed +.Pp The .Fa sinfo->sinfo_ppid argument is an opaque 32 bit value that is passed transparently through the stack to the peer endpoint. It will be available on reception of a message (see -.Fn sctp_recvmsg 2 -). Note that the stack passes this value without regard to byte +.Xr sctp_recvmsg 2 +). +Note that the stack passes this value without regard to byte order. .Pp The @@ -126,54 +135,65 @@ argument may include one or more of the following: The flag .Dv SCTP_EOF is used to instruct the SCTP stack to queue this message -and then start a graceful shutdown of the association. All +and then start a graceful shutdown of the association. +All remaining data in queue will be sent after which the association -will be shutdown. +will be shut down. .Pp .Dv SCTP_ABORT -is used to immediately terminate an association. An abort +is used to immediately terminate an association. +An abort is sent to the peer and the local TCB is destroyed. .Pp .Dv SCTP_UNORDERED is used to specify that the message being sent has no specific order and should be delivered to the peer application -as soon as possible. When this flag is absent messages +as soon as possible. +When this flag is absent messages are delivered in order within the stream they are sent, but without respect to order to peer streams. .Pp The flag .Dv SCTP_ADDR_OVER -is used to specify that an specific address should be used. Normally -SCTP will use only one of a multi-homed peers address as the primary -address to send to. By default, no matter what the +is used to specify that a specific address should be used. +Normally +SCTP will use only one of a multi-homed peers addresses as the primary +address to send to. +By default, no matter what the .Fa to -argument is, this primary address is used to send data. By specifying +argument is, this primary address is used to send data. +By specifying this flag, the user is asking the stack to ignore the primary address -and instead use the specified address not only has a lookup mechanism -to find the association but also has the actual address to send to. +and instead use the specified address not only as a lookup mechanism +to find the association but also as the actual address to send to. .Pp For a one-to-many type (SOCK_SEQPACKET) socket the flag .Dv SCTP_SENDALL -can be used as a convient way to make one send call and have +can be used as a convenient way to make one send call and have all associations that are under the socket get a copy of the message. Note that this mechanism is quite efficent and makes only one actual copy of the data which is shared by all the associations for sending. .Pp The remaining flags are used for the partial reliabilty extension (RFC3758) and will only be effective if the peer endpoint supports this extension. -This option specify's what local policy the local endpoint should use -in skipping data. If none of these options are set, then data is +This option specifies what local policy the local endpoint should use +in skipping data. +If none of these options are set, then data is never skipped over. .Pp .Dv SCTP_PR_SCTP_TTL Is used to indicate that a time based lifetime is being applied -to the data. The +to the data. +The .Fa sinfo->sinfo_timetolive argument is then a number of milliseconds for which the data is -attempted to be transmitted. If that many milliseconds ellapses -and the peer has not acknowledge the data, the data will be -skipped and no longer transmitted. Note that this policy does -not even assure that the data will ever be sent. In times of a congestion +attempted to be transmitted. +If that many milliseconds ellapse +and the peer has not acknowledged the data, the data will be +skipped and no longer transmitted. +Note that this policy does +not even assure that the data will ever be sent. +In times of a congestion with large amounts of data being queued, the .Fa sinfo->sinfo_timetolive may expire before the first transmission is ever made. @@ -183,8 +203,9 @@ The based policy transforms the .Fa sinfo->sinfo_timetolive field into a total number of bytes allowed on the outbound -send queue. If that number or more bytes are in queue, then -other buffer based sends are looked to be removed and +send queue. +If that number or more bytes are in queue, then +other buffer-based sends are looked to be removed and skipped. Note that this policy may also result in the data never being sent if no buffer based sends are in queue and the maximum specified by @@ -195,35 +216,40 @@ The .Dv SCTP_PR_SCTP_RTX policy transforms the .Fa sinfo->sinfo_timetolive -into a number of retransmissions to allow. This policy +into a number of retransmissions to allow. +This policy always assures that at a minimum one send attempt is -made of the data. After which no more than +made of the data. +After which no more than .Fa sinfo->sinfo_timetolive retransmissions will be made before the data is skipped. .Pp .Fa sinfo->sinfo_stream is the SCTP stream that you wish to send the -message on. Streams in SCTP are reliable (or partially reliable) flows of ordered +message on. +Streams in SCTP are reliable (or partially reliable) flows of ordered messages. .Pp The .Fa sinfo->sinfo_assoc_id field is used to -select the association to send to on an one-to-many socket. For a -one-to-one socket, this field is ignored. +select the association to send to on an one-to-many socket. +For a one-to-one socket, this field is ignored. .Pp .Fa sinfo->sinfo_context -field is used only in the event the message cannot be sent. This is an opaque +field is used only in the event the message cannot be sent. +This is an opaque value that the stack retains and will give to the user when a failed send is given if that notification is enabled (see -.Tn sctp -). Normally a user process can use this value to index some application +.Xr sctp 4 +). +Normally a user process can use this value to index some application specific data structure when a send cannot be fulfilled. .Pp The .Fa flags -argument holds the same meaning and values has those found in -.Fn sendmsg 2 +argument holds the same meaning and values as those found in +.Xr sendmsg 2 but is generally ignored by SCTP. .Pp The fields @@ -232,30 +258,35 @@ The fields and .Fa sinfo->sinfo_cumtsn are used only when receiving messages and are thus ignored by -.Fn sctp_send. +.Fn sctp_send . The function .Fn sctp_sendx has the same properties as .Fn sctp_send with the additional arguments of an array of sockaddr structures -passed in. With the +passed in. +With the .Fa addrs argument being given as an array of addresses to be sent to and the .Fa addrcnt argument indicating how many socket addresses are in the passed -in array. Note that all of the addresses will only be used -when an implicit association is being setup. This allows the +in array. +Note that all of the addresses will only be used +when an implicit association is being set up. +This allows the user the equivilant behavior as doing a .Fn sctp_connectx followed by a .Fn sctp_send -to the association. Note that if the association id. +to the association. +Note that if the .Fa sinfo->sinfo_assoc_id field is 0, then the first address will be used to look up -the association in place of the association id. If both -an address and and association id are specified, the association -id has priority. +the association in place of the association id. +If both +an address and an association id are specified, the association +id has priority. .Sh RETURN VALUES The call returns the number of characters sent, or -1 if an error occurred. @@ -289,12 +320,12 @@ but may be caused by transient congestion. .It Bq Er EHOSTUNREACH The remote host was unreachable. .It Bq Er ENOTCON -On a one to one style socket no association exists. +On a one-to-one style socket no association exists. .It Bq Er ECONNRESET An abort was received by the stack while the user was attempting to send data to the peer. .It Bq Er ENOENT -On a one to many style socket no address is specified +On a one-to-many style socket no address is specified so that the association cannot be located or the SCTP_ABORT flag was specified on a non-existing association. .It Bq Er EPIPE @@ -305,20 +336,19 @@ This typically means that the socket is not connected and is a one-to-one style socket. .El .Sh SEE ALSO -.Xr sctp 4 , -.Xr sendmsg 2 , -.Xr sctp_sendmsg 3 , -.Xr sctp_recvmsg 3 , -.Xr sctp_connectx 3 , .Xr getsockopt 2 , .Xr recv 2 , .Xr select 2 , +.Xr sendmsg 2 , .Xr socket 2 , .Xr write 2 +.Xr sctp_connectx 3 , +.Xr sctp_recvmsg 3 , +.Xr sctp_sendmsg 3 , +.Xr sctp 4 .Sh BUGS Because .Fn sctp_send may have multiple associations under one endpoint, a select on write will only work for a one-to-one style socket. - diff --git a/lib/libc/net/sctp_sendmsg.3 b/lib/libc/net/sctp_sendmsg.3 index 8c3f6edbc828..bc644b672939 100644 --- a/lib/libc/net/sctp_sendmsg.3 +++ b/lib/libc/net/sctp_sendmsg.3 @@ -36,7 +36,7 @@ .Dt SCTP_SENDMSG 3 .Os .Sh NAME -.Nm sctp_sendmsg +.Nm sctp_sendmsg , .Nm sctp_sendmsgx .Nd send a message from an SCTP socket .Sh LIBRARY @@ -46,31 +46,44 @@ .In sys/socket.h .In netinet/sctp.h .Ft ssize_t -.Fn sctp_sendmsg "int s" "const void *msg" "size_t len" "const struct sockaddr *to" "socklen_t tolen" "uint32_t ppid" "uint32_t flags" "uint16_t stream_no" "uint32_t timetolive" "uint32_t context" +.Fo sctp_sendmsg +.Fa "int s" "const void *msg" "size_t len" "const struct sockaddr *to" +.Fa "socklen_t tolen" "uint32_t ppid" "uint32_t flags" "uint16_t stream_no" +.Fa "uint32_t timetolive" "uint32_t context" +.Fc .Ft ssize_t -.Fn sctp_sendmsgx "int s" "const void *msg" "size_t len" "const struct sockaddr *to" "int addrcnt" "uint32_t ppid" "uint32_t flags" "uint16_t stream_no" "uint32_t timetolive" "uint32_t context" - +.Fo sctp_sendmsgx +.Fa "int s" "const void *msg" "size_t len" "const struct sockaddr *to" +.Fa "int addrcnt" "uint32_t ppid" "uint32_t flags" "uint16_t stream_no" +.Fa "uint32_t timetolive" "uint32_t context" +.Fc .Sh DESCRIPTION The .Fn sctp_sendmsg -system calls +system call is used to transmit a message to another SCTP endpoint. The .Fn sctp_sendmsg -may be used at any time. If the socket is a one-to-many type (SOCK_SEQPACKET) +may be used at any time. +If the socket is a one-to-many type (SOCK_SEQPACKET) socket then an attempt to send to an address that no association exists to will -implicitly create a new association. Data sent in such an instance will result in -the data being sent on the third leg of the SCTP four-way handshake. Note that if +implicitly create a new association. +Data sent in such an instance will result in +the data being sent on the third leg of the SCTP four-way handshake. +Note that if the socket is a one-to-one type (SOCK_STREAM) socket then an association must be in existance (by use of the -.Fn connect 2 -system call). Calling +.Xr connect 2 +system call). +Calling .Fn sctp_sendmsg or .Fn sctp_sendmsgx -on a non-connected one-to-one socket will result in the errno being set to -.Er ENOTCONN -a -1 being returned, and the message is not transmitted. +on a non-connected one-to-one socket will result in +.Va errno +being set to +.Er ENOTCONN , +-1 being returned, and the message not being transmitted. .Pp The address of the target is given by .Fa to @@ -82,32 +95,37 @@ The length of the message is given by .Fa len . If the message is too long to pass atomically through the -underlying protocol, the errno is set to -.Er EMSGSIZE -a -1 is returned, and +underlying protocol, +.Va errno +is set to +.Er EMSGSIZE , +-1 is returned, and the message is not transmitted. .Pp No indication of failure to deliver is implicit in a -.Fn sctp_sendmsg 2 +.Xr sctp_sendmsg 2 +call. Locally detected errors are indicated by a return value of -1. .Pp -If no messages space is available at the socket to hold +If no space is available at the socket to hold the message to be transmitted, then -.Fn sctp_sendmsg 2 +.Xr sctp_sendmsg 2 normally blocks, unless the socket has been placed in non-blocking I/O mode. The -.Fn select 2 +.Xr select 2 system call may be used to determine when it is possible to send more data on one-to-one type (SOCK_STREAM) sockets. .Pp The .Fa ppid argument is an opaque 32 bit value that is passed transparently -through the stack to the peer endpoint. It will be available on +through the stack to the peer endpoint. +It will be available on reception of a message (see -.Fn sctp_recvmsg 2 -). Note that the stack passes this value without regard to byte +.Xr sctp_recvmsg 2 +). +Note that the stack passes this value without regard to byte order. .Pp The @@ -129,31 +147,37 @@ argument may include one or more of the following: The flag .Dv SCTP_EOF is used to instruct the SCTP stack to queue this message -and then start a graceful shutdown of the association. All +and then start a graceful shutdown of the association. +All remaining data in queue will be sent after which the association -will be shutdown. +will be shut down. .Pp .Dv SCTP_ABORT -is used to immediately terminate an association. An abort +is used to immediately terminate an association. +An abort is sent to the peer and the local TCB is destroyed. .Pp .Dv SCTP_UNORDERED is used to specify that the message being sent has no specific order and should be delivered to the peer application -as soon as possible. When this flag is absent messages +as soon as possible. +When this flag is absent messages are delivered in order within the stream they are sent, but without respect to order to peer streams. .Pp The flag .Dv SCTP_ADDR_OVER -is used to specify that an specific address should be used. Normally -SCTP will use only one of a multi-homed peers address as the primary -address to send to. By default, no matter what the +is used to specify that an specific address should be used. +Normally +SCTP will use only one of a multi-homed peers addresses as the primary +address to send to. +By default, no matter what the .Fa to -argument is, this primary address is used to send data. By specifying +argument is, this primary address is used to send data. +By specifying this flag, the user is asking the stack to ignore the primary address -and instead use the specified address not only has a lookup mechanism -to find the association but also has the actual address to send to. +and instead use the specified address not only as a lookup mechanism +to find the association but also as the actual address to send to. .Pp For a one-to-many type (SOCK_SEQPACKET) socket the flag .Dv SCTP_SENDALL @@ -165,18 +189,23 @@ copy of the data which is shared by all the associations for sending. The remaining flags are used for the partial reliabilty extension (RFC3758) and will only be effective if the peer endpoint supports this extension. This option specify's what local policy the local endpoint should use -in skipping data. If none of these options are set, then data is +in skipping data. +If none of these options are set, then data is never skipped over. .Pp .Dv SCTP_PR_SCTP_TTL Is used to indicate that a time based lifetime is being applied -to the data. The +to the data. +The .Fa timetolive argument is then a number of milliseconds for which the data is -attempted to be transmitted. If that many milliseconds ellapses -and the peer has not acknowledge the data, the data will be -skipped and no longer transmitted. Note that this policy does -not even assure that the data will ever be sent. In times of a congestion +attempted to be transmitted. +If that many milliseconds ellapse +and the peer has not acknowledged the data, the data will be +skipped and no longer transmitted. +Note that this policy does +not even assure that the data will ever be sent. +In times of a congestion with large amounts of data being queued, the .Fa timetolive may expire before the first transmission is ever made. @@ -186,9 +215,11 @@ The based policy transforms the .Fa timetolive field into a total number of bytes allowed on the outbound -send queue. If that number or more bytes are in queue, then +send queue. +If that number or more bytes are in queue, then other buffer based sends are looked to be removed and -skipped. Note that this policy may also result in the data +skipped. +Note that this policy may also result in the data never being sent if no buffer based sends are in queue and the maximum specified by .Fa timetolive @@ -198,22 +229,28 @@ The .Dv SCTP_PR_SCTP_RTX policy transforms the .Fa timetolive -into a number of retransmissions to allow. This policy +into a number of retransmissions to allow. +This policy always assures that at a minimum one send attempt is -made of the data. After which no more than +made of the data. +After which no more than .Fa timetolive retransmissions will be made before the data is skipped. .Pp .Fa stream_no is the SCTP stream that you wish to send the -message on. Streams in SCTP are reliable (or partially reliable) flows of ordered -messages. The +message on. +Streams in SCTP are reliable (or partially reliable) flows of ordered +messages. +The .Fa context -field is used only in the event the message cannot be sent. This is an opaque +field is used only in the event the message cannot be sent. +This is an opaque value that the stack retains and will give to the user when a failed send is given if that notification is enabled (see -.Tn sctp -). Normally a user process can use this value to index some application +.Xr sctp 4 +). +Normally a user process can use this value to index some application specific data structure when a send cannot be fulfilled. .Fn sctp_sendmsgx is identical to @@ -223,7 +260,8 @@ argument .Fa to and adds the additional argument .Fa addrcnt -which specifies how many addresses are in the array. This allows a +which specifies how many addresses are in the array. +This allows a caller to implictly setup an association passing multiple addresses as if an .Fn sctp_connectx @@ -261,14 +299,15 @@ but may be caused by transient congestion. .It Bq Er EHOSTUNREACH The remote host was unreachable. .It Bq Er ENOTCON -On a one to one style socket no association exists. +On a one-to-one style socket no association exists. .It Bq Er ECONNRESET An abort was received by the stack while the user was attempting to send data to the peer. .It Bq Er ENOENT -On a one to many style socket no address is specified +On a one-to-many style socket no address is specified so that the association cannot be located or the -SCTP_ABORT flag was specified on a non-existing association. +.Dv SCTP_ABORT +flag was specified on a non-existing association. .It Bq Er EPIPE The socket is unable to send anymore data .Dv ( SBS_CANTSENDMORE @@ -277,15 +316,15 @@ This typically means that the socket is not connected and is a one-to-one style socket. .El .Sh SEE ALSO -.Xr sctp 4 , -.Xr sendmsg 3 , .Xr connect 2 , -.Xr sctp_connectx 3 , .Xr getsockopt 2 , .Xr recv 2 , .Xr select 2 , .Xr socket 2 , -.Xr write 2 +.Xr write 2 , +.Xr sendmsg 3 , +.Xr sctp_connectx 3 , +.Xr sctp 4 .Sh BUGS Because in the one-to-many style socket the .Fn sctp_sendmsg @@ -294,4 +333,3 @@ or may have multiple associations under one endpoint, a select on write will only work for a one-to-one style socket. -