API-functions.md

SRT API Functions

• Library Initialization
  • srt_startup
  • srt_cleanup
• Creating and configuring sockets
  • srt_socket
  • srt_create_socket
  • srt_bind
  • srt_bind_acquire
  • srt_getsockstate
  • srt_getsndbuffer
  • srt_close
• Connecting
  • srt_listen
  • srt_accept
  • srt_accept_bond
  • srt_listen_callback
  • srt_connect
  • srt_connect_bind
  • srt_connect_debug
  • srt_rendezvous
  • srt_connect_callback
• Socket group management
  • SRT_GROUP_TYPE
  • SRT_SOCKGROUPCONFIG
  • SRT_SOCKGROUPDATA
  • SRT_MEMBERSTATUS
  • srt_create_group
  • srt_include
  • srt_exclude
  • srt_groupof
  • srt_group_data
  • srt_connect_group
  • srt_prepare_endpoint
  • srt_create_config
  • srt_delete_config
  • srt_config_add
• Options and properties
  • srt_getpeername
  • srt_getsockname
  • srt_getsockopt, srt_getsockflag
  • srt_setsockopt, srt_setsockflag
  • srt_getversion
• Helper data types for transmission
  • SRT_MSGCTRL
• Transmission
  • srt_send, srt_sendmsg, srt_sendmsg2
  • srt_recv, srt_recvmsg, srt_recvmsg2
  • srt_sendfile, srt_recvfile
• Diagnostics
  • srt_getlasterror_str
  • srt_getlasterror
  • srt_strerror
  • srt_clearlasterror
  • srt_getrejectreason
  • srt_rejectreason_str
  • srt_setrejectreason
  • Error Codes
• Performance tracking
  • srt_bstats, srt_bistats
• Asynchronous operations (epoll)
  • srt_epoll_create
  • srt_epoll_add_usock, srt_epoll_add_ssock, srt_epoll_update_usock, srt_epoll_update_ssock
  • srt_epoll_remove_usock, srt_epoll_remove_ssock
  • srt_epoll_wait
  • srt_epoll_uwait
  • srt_epoll_clear_usocks
  • srt_epoll_set
  • srt_epoll_release
• Logging control
  • srt_setloglevel
  • srt_addlogfa, srt_dellogfa, srt_resetlogfa
  • srt_setloghandler
  • srt_setlogflags
• Time Access
  • srt_time_now
  • srt_connection_time

Library initialization

srt_startup

int srt_startup(void);

This function shall be called at the start of an application that uses the SRT library. It provides all necessary platform-specific initializations, sets up global data, and starts the SRT GC thread. If this function isn't explicitly called, it will be called automatically when creating the first socket. However, relying on this behavior is strongly discouraged.

• Returns:

  • 0 = successfully run, or already started
  • 1 = this is the first startup, but the GC thread is already running
  • -1 = failed

• Errors:

  • SRT_ECONNSETUP (with error code set): Reported when required system resource(s) failed to initialize. This is currently used only on Windows to report a failure from WSAStartup.

srt_cleanup

int srt_cleanup(void);

This function cleans up all global SRT resources and shall be called just before exiting the application that uses the SRT library. This cleanup function will still be called from the C++ global destructor, if not called by the application, although relying on this behavior is stronly discouraged.

• Returns:

  • 0 (A possibility to return other values is reserved for future use)

IMPORTANT: Note that the startup/cleanup calls have an instance counter. This means that if you call srt_startup multiple times, you need to call the srt_cleanup function exactly the same number of times.

Creating and configuring sockets

srt_socket

SRTSOCKET srt_socket(int af, int type, int protocol);

Old and deprecated version of srt_create_socket. All arguments are ignored.

NOTE changes with respect to UDT version:

• In UDT (and SRT versions before 1.4.2) the af parameter was specifying the socket family (AF_INET or AF_INET6). This is now not required; this parameter is decided at the call of srt_conenct or srt_bind.

• In UDT the type parameter was used to specify the file or message mode using SOCK_STREAM or SOCK_DGRAM symbols (with the latter being misleading, as the message mode has nothing to do with UDP datagrams and it's rather similar to the SCTP protocol). In SRT these two modes are available by setting SRTO_TRANSTYPE. The default is SRTT_LIVE. If, however, you set SRTO_TRANSTYPE to SRTT_FILE for file mode, you can then leave the SRTO_MESSAGEAPI option as false (default), which corresponds to "stream" mode (TCP-like), or set it to true, which corresponds to "message" mode (SCTP-like).

srt_create_socket

SRTSOCKET srt_create_socket();

Creates an SRT socket.

Note that socket IDs always have the SRTGROUP_MASK bit clear.

• Returns:

  • a valid socket ID on success
  • INVALID_SOCKET (-1) on error

• Errors:

  • SRT_ENOTBUF: not enough memory to allocate required resources

NOTE: This is probably a design flaw (BUG?). Usually underlying system errors are reported by SRT_ECONNSETUP.

srt_bind

int srt_bind(SRTSOCKET u, const struct sockaddr* name, int namelen);

Binds a socket to a local address and port. Binding specifies the local network interface and the UDP port number to be used for the socket. When the local address is a form of INADDR_ANY, then it's bound to all interfaces. When the port number is 0, then the port number will be system-allocated if necessary.

This call is obligatory for a listening socket before calling srt_listen and for rendezvous mode before calling srt_connect, otherwise it's optional. For a listening socket it defines the network interface and the port where the listener should expect a call request. In case of rendezvous mode (when the socket has set SRTO_RENDEZVOUS to true, in this mode both parties connect to one another) it defines the network interface and port from which packets will be sent to the peer and to which the peer is expected to send packets.

For a connecting socket this call can set up the outgoing port to be used in the communication. It is allowed that multiple SRT sockets share one local outgoing port, as long as SRTO_REUSEADDR is set to true (default). Without this call the port will be automatically selected by the system.

NOTE: This function cannot be called on socket group. If you need to have the group-member socket bound to the specified source address before connecting, use srt_connect_bind for that purpose.

• Returns:

  • SRT_ERROR (-1) on error, otherwise 0

• Errors:

  • SRT_EINVSOCK: Socket passed as u designates no valid socket
  • SRT_EINVOP: Socket already bound
  • SRT_ECONNSETUP: Internal creation of a UDP socket failed
  • SRT_ESOCKFAIL: Internal configuration of a UDP socket (bind, setsockopt) failed

srt_bind_acquire

int srt_bind_acquire(SRTSOCKET u, UDPSOCKET udpsock);

A version of srt_bind that acquires a given UDP socket instead of creating one.

srt_getsockstate

SRT_SOCKSTATUS srt_getsockstate(SRTSOCKET u);

Gets the current status of the socket. Possible states are:

• SRTS_INIT: Created, but not bound
• SRTS_OPENED: Created and bound, but not in use yet.
• SRTS_LISTENING: Socket is in listening state
• SRTS_CONNECTING: The connect operation was initiated, but not yet finished. This may also mean that it has timed out; you can only know that after getting a socket error report from srt_epoll_wait. In blocking mode it's not possible because srt_connect does not return until the socket is connected or failed due to timeout or interrupted call.
• SRTS_CONNECTED: The socket is connected and ready for transmission.
• SRTS_BROKEN: The socket was connected, but the connection was broken
• SRTS_CLOSING: The socket may still be open and active, but closing is requested, so no further operations will be accepted (active operations will be completed before closing)
• SRTS_CLOSED: The socket has been closed, but not yet removed by the GC thread
• SRTS_NONEXIST: The specified number does not correspond to a valid socket.

srt_getsndbuffer

int srt_getsndbuffer(SRTSOCKET sock, size_t* blocks, size_t* bytes);

Retrieves information about the sender buffer.

• sock: Socket to test
• blocks: Written information about buffer blocks in use
• bytes: Written information about bytes in use

This function can be used for diagnostics. It is especially useful when the socket needs to be closed asynchronously.

srt_close

int srt_close(SRTSOCKET u);

Closes the socket or group and frees all used resources. Note that underlying UDP sockets may be shared between sockets, so these are freed only with the last user closed.

• Returns:

  • SRT_ERROR (-1) in case of error, otherwise 0

• Errors:

  • SRT_EINVSOCK: Socket u indicates no valid socket ID

Connecting

srt_listen

int srt_listen(SRTSOCKET u, int backlog);

This sets up the listening state on a socket with a backlog setting that defines how many sockets may be allowed to wait until they are accepted (excessive connection requests are rejected in advance).

The following important options may change the behavior of the listener socket and the srt_accept function:

• srt_listen_callback installs a user function that will be called before srt_accept can happen
• SRTO_GROUPCONNECT option allows the listener socket to accept group connections

• Returns:

  • SRT_ERROR (-1) in case of error, otherwise 0.

• Errors:

  • SRT_EINVPARAM: Value of backlog is 0 or negative.
  • SRT_EINVSOCK: Socket u indicates no valid SRT socket.
  • SRT_EUNBOUNDSOCK: srt_bind has not yet been called on that socket.
  • SRT_ERDVNOSERV: SRTO_RENDEZVOUS flag is set to true on specified socket.
  • SRT_EINVOP: Internal error (should not happen when SRT_EUNBOUNDSOCK is reported).
  • SRT_ECONNSOCK: The socket is already connected.
  • SRT_EDUPLISTEN: The address used in srt_bind by this socket is already occupied by another listening socket. Binding multiple sockets to one IP address and port is allowed, as long as SRTO_REUSEADDR is set to true, but only one of these sockets can be set up as a listener.

srt_accept

SRTSOCKET srt_accept(SRTSOCKET lsn, struct sockaddr* addr, int* addrlen);

Accepts a pending connection, then creates and returns a new socket or group ID that handles this connection. The group and socket can be distinguished by checking the SRTGROUP_MASK bit on the returned ID.

• lsn: the listener socket previously configured by srt_listen
• addr: the IP address and port specification for the remote party
• addrlen: INPUT: size of addr pointed object. OUTPUT: real size of the returned object

NOTE: addr is allowed to be NULL, in which case it's understood that the application is not interested in the address from which the connection originated. Otherwise addr should specify an object into which the address will be written, and addrlen must also specify a variable to contain the object size. Note also that in the case of group connection only the initial connection that establishes the group connection is returned, together with its address. As member connections are added or broken within the group, you can obtain this information through srt_group_data or the data filled by srt_sendmsg2 and srt_recvmsg2.

If the pending connection is a group connection (initiated on the peer side by calling the connection function using a group ID, and permitted on the listener socket by SRTO_GROUPCONNECT flag), then the value returned is a group ID. This function then creates a new group, as well as a new socket for this very connection, that will be added to the group. Once the group is created this way, further connections within the same group, as well as sockets for them, will be created in the background. The SRT_EPOLL_UPDATE event is raised on the lsn socket when a new background connection is attached to the group, although it's usually for internal use only.

• Returns:

  • On success, a valid SRT socket or group ID to be used for transmission
  • SRT_ERROR (-1) on failure

• Errors:

  • SRT_EINVPARAM: NULL specified as addrlen, when addr is not NULL
  • SRT_EINVSOCK: lsn designates no valid socket ID.
  • SRT_ENOLISTEN: lsn is not set up as a listener (srt_listen not called)
  • SRT_EASYNCRCV: No connection reported so far. This error is reported only when the lsn listener socket was configured as non-blocking for reading (SRTO_RCVSYN set to false); otherwise the call blocks until a connection is reported or an error occurs
  • SRT_ESCLOSED: The lsn socket has been closed while the function was blocking the call (if SRTO_RCVSYN is set to default true). This includes a situation when the socket was closed just at the moment when a connection was made and the socket got closed during processing

srt_accept_bond

SRTSOCKET srt_accept_bond(const SRTSOCKET listeners[], int nlisteners, int msTimeOut);

Accepts a pending connection, like srt_accept, but pending on any of the listener sockets passed in the listeners array of nlisteners size.

• listeners: array of listener sockets (all must be setup by srt_listen)
• nlisteners: size of the listeners array
• msTimeOut: timeout in [ms] or -1 to block forever

This function is for blocking mode only - for non-blocking mode you should simply call srt_accept on the first listener socket that reports readiness, and this function is actually a friendly shortcut that uses waiting on epoll and srt_accept internally. This function supports an important use case for accepting a group connection, for which every member connection is expected to be established over a different listener socket.

Note that there's no special set of settings required or rejected for this function. The group-member connections for the same group can be established over various different listener sockets always when all those listeners are hosted by the same application, as the group management is global for the application, so a connection reporting in for an already connected group gets discovered and the connection will be handled in the background, regardless to which listener socket the call was done - as long as the connection is accepted according to any additional conditions.

This function has still nothing to do with the groups - you can use it in any case when you have one service that accepts connections to multiple endpoints. Note also that the settings as to whether listeners should accept or reject socket or group connections, should be applied to the listener sockets appropriately prior to calling this function.

• Returns:

  • On success, a valid SRT socket or group ID to be used for transmission
  • SRT_ERROR (-1) on failure

• Errors:

  • SRT_EINVPARAM: NULL specified as listeners or nlisteners < 1

  • SRT_EINVSOCK: any socket in listeners designates no valid socket ID. Can also mean Internal Error when an error occurred while creating an accepted socket (BUG?)

  • SRT_ENOLISTEN: any socket in listeners is not set up as a listener (srt_listen not called, or the listener socket has already been closed)

  • SRT_EASYNCRCV: No connection reported on any listener socket as the timeout has been reached. This error is only reported when msTimeOut is not -1

srt_listen_callback

int srt_listen_callback(SRTSOCKET lsn, srt_listen_callback_fn* hook_fn, void* hook_opaque);

This call installs a callback hook, which will be executed on a socket that is automatically created to handle the incoming connection on the listening socket (and is about to be returned by srt_accept), but before the connection has been accepted.

• lsn: Listening socket where you want to install the callback hook
• hook_fn: The callback hook function pointer
• hook_opaque: The pointer value that will be passed to the callback function

• Returns:

  • 0, if successful
  • -1, on error

• Errors:

  • SRT_EINVPARAM reported when hook_fn is a null pointer

The callback function has the signature as per this type definition:

typedef int srt_listen_callback_fn(void* opaque, SRTSOCKET ns, int hs_version
             const struct sockaddr* peeraddr, const char* streamid);

The callback function gets the following parameters passed:

• opaque: The pointer passed as hook_opaque when registering
• ns: The freshly created socket to handle the incoming connection
• hs_version: The handshake version (usually 5, pre-1.3 versions use 4)
• peeraddr: The address of the incoming connection
• streamid: The value set to SRTO_STREAMID option set on the peer side

(Note that versions that use handshake version 4 are incapable of using any extensions, such as streamid, however they do support encryption. Note also that the SRT version isn't yet extracted, however you can prevent too old version connections using SRTO_MINVERSION option).

The callback function is given an opportunity to:

• use the passed information (streamid and peer address) to decide what to do with this connection
• alter any options on the socket, which could not be set properly before on the listening socket to be derived by the accepted socket, and won't be allowed to be altered after the socket is returned by srt_accept

Note that the returned socket has already set all derived options from the listener socket, as it happens normally, and the moment when this callback is called is when the conclusion handshake has been already received from the caller party, but not yet interpreted (the streamid field is extracted from it prematurely). When you, for example, set a passphrase on the socket at this very moment, the Key Material processing will happen against this already set passphrase, after the callback function is finished.

The callback function shall return 0, if the connection is to be accepted. If you return -1, or if the function throws an exception, this will be understood as a request to reject the incoming connection - in which case the about-to-be-accepted socket will be silently deleted and srt_accept will not report it. Note that in case of non-blocking mode the epoll bits for read-ready on the listener socket will not be set if the connection is rejected, including when rejected from this user function.

IMPORTANT: This function is called in the receiver worker thread, which means that it must do its checks and operations as quickly as possible and keep the minimum possible time, as every delay you do in this function will burden the processing of the incoming data on the associated UDP socket, which in case of a listener socket means the listener socket itself and every socket accepted off this listener socket. Avoid any extensive search operations, best cache in memory whatever database you have to check against the data received in streamid or peeraddr.

srt_connect

int srt_connect(SRTSOCKET u, const struct sockaddr* name, int namelen);

Connects a socket or a group to a remote party with a specified address and port.

• u: can be an SRT socket or SRT group, both freshly created and not yet used for any connection, except possibly srt_bind on the socket
• name: specification of the remote address and port
• namelen: size of the object passed by name

NOTES:

1. The socket used here may be bound from upside (or binding and connection can be done in one function, srt_connect_bind) so that it uses a predefined network interface or local outgoing port. If not, it behaves as if it was bound to INADDR_ANY (which binds on all interfaces) and port 0 (which makes the system assign the port automatically).
2. When u is a group, then this call can be done multiple times, each time for another member connection, and a new member SRT socket will be created automatically for every call of this function.
3. If you want to connect a group to multiple links at once and use blocking mode, you might want to use srt_connect_group instead.

• Returns:

  • SRT_ERROR (-1) in case of error
  • 0 in case when used for u socket
  • Socket ID created for connection for u group

• Errors:

  • SRT_EINVSOCK: Socket u indicates no valid socket ID
  • SRT_ERDVUNBOUND: Socket u has set SRTO_RENDEZVOUS to true, but srt_bind hasn't yet been called on it. The srt_connect function is also used to connect a rendezvous socket, but rendezvous sockets must be explicitly bound to a local interface prior to connecting. Non-rendezvous sockets (caller sockets) can be left without binding - the call to srt_connect will bind them automatically.
  • SRT_ECONNSOCK: Socket u is already connected
  • SRT_ECONNREJ: Connection has been rejected
  • SRT_ENOSERVER: Connection has been timed out (see SRTO_CONNTIMEO)
  • SRT_ESCLOSED: The socket u has been closed while the function was blocking the call (if SRTO_RCVSYN is set to default true)

When SRT_ECONNREJ error is reported, you can get the reason for a rejected connection from srt_getrejectreason. In non-blocking mode (when SRTO_RCVSYN is set to false), only SRT_EINVSOCK, SRT_ERDVUNBOUND and SRT_ECONNSOCK can be reported. In all other cases the function returns immediately with a success, and the only way to obtain the connecting status is through the epoll flag with SRT_EPOLL_ERR. In this case you can also call srt_getrejectreason to get the detailed reason for the error, including connection timeout (SRT_REJ_TIMEOUT).

srt_connect_bind

int srt_connect_bind(SRTSOCKET u, const struct sockaddr* source,
                     const struct sockaddr* target, int len);

This function does the same as first srt_bind then srt_connect, if called with u being a socket. If u is a group, then it will execute srt_bind first on the automatically created socket for the connection.

• u: Socket or group to connect
• source: Address to bind u to
• target: Address to connect
• len: size of the original structure of source and target

• Returns:

  • SRT_ERROR (-1) in case of error
  • 0 in case when used for u socket
  • Socket ID created for connection for u group

• Errors:

  • SRT_EINVSOCK: Socket passed as u designates no valid socket
  • SRT_EINVOP: Socket already bound
  • SRT_ECONNSETUP: Internal creation of a UDP socket failed
  • SRT_ESOCKFAIL: Internal configuration of a UDP socket (bind, setsockopt) failed
  • SRT_ERDVUNBOUND: Internal error (srt_connect should not report it after srt_bind was called)
  • SRT_ECONNSOCK: Socket u is already connected
  • SRT_ECONNREJ: Connection has been rejected

IMPORTANT: It's not allowed to bind and connect the same socket to two different families (that is, both source and target must be AF_INET or AF_INET6), although you may mix links over IPv4 and IPv6 in one group.

srt_connect_debug

int srt_connect_debug(SRTSOCKET u, const struct sockaddr* name, int namelen, int forced_isn);

This function is for developers only and can be used for testing. It does the same thing as srt_connect, with the exception that it allows specifying the Initial Sequence Number for data transmission. Normally this value is generated randomly.

srt_rendezvous

int srt_rendezvous(SRTSOCKET u, const struct sockaddr* local_name, int local_namelen,
        const struct sockaddr* remote_name, int remote_namelen);

Performs a rendezvous connection. This is a shortcut for doing bind locally, setting the SRTO_RENDEZVOUS option to true, and doing srt_connect.

• u: socket to connect
• local_name: specifies the local network interface and port to bind
• remote_name: specifies the remote party's IP address and port

• Returns:

  • SRT_ERROR (-1) in case of error, otherwise 0

• Errors:

  • SRT_EINVSOCK: Socket passed as u designates no valid socket
  • SRT_EINVOP: Socket already bound
  • SRT_ECONNSETUP: Internal creation of a UDP socket failed
  • SRT_ESOCKFAIL: Internal configuration of a UDP socket (bind, setsockopt) failed
  • SRT_ERDVUNBOUND: Internal error (srt_connect should not report it after srt_bind was called)
  • SRT_ECONNSOCK: Socket u is already connected
  • SRT_ECONNREJ: Connection has been rejected

IMPORTANT: It's not allowed to perform a rendezvous connection to two different families (that is, both local_name and remote_name must be AF_INET or AF_INET6).

srt_connect_callback

int srt_connect_callback(SRTSOCKET clr, srt_connect_callback_fn* hook_fn, void* hook_opaque);

This call installs a callback hook, which will be executed on a given clr socket just after the pending connection situation in the background has been resolved (that is, when the connection succeeded or failed). Note that this function is not guaranteed to be called if the clr socket is set to blocking mode (SRTO_RCVSYN option set to true). It is guaranteed to be called when a socket is in non-blocking mode, or when you use a group.

This function is mainly intended to be used with group connections. Note that even if you use a group connection in blocking mode, after the group is considered connected the member connections still continue in background. Also, when some connections are still pending and others have failed, the blocking call for srt_connect_group will not exit until at least one of them succeeds or all fail - in such a case those failures also happen only in the background, while the connecting function blocks until all connections are resolved. When all links fail, you will only get a general error code for the group. This mechanism allows you to get individual errors for particular member connections.

You can also use this mechanism as an alternative method for a single-socket connection in non-blocking mode to trigger an action when the connection process is finished.

• clr: Socket that will be used for connecting and for which the hook is installed
• hook_fn: The callback hook function pointer
• hook_opaque: The pointer value that will be passed to the callback function

• Returns:

  • 0, if successful
  • -1, on error

• Errors:

  • SRT_EINVPARAM reported when hook_fn is a null pointer

The callback function signature has the following type definition:

typedef void srt_connect_callback_fn(void* opaq, SRTSOCKET ns, int errorcode, const struct sockaddr* peeraddr, int token);

where:

• opaq: The pointer passed as hook_opaque when registering
• ns: The socket for which the connection process was resolved
• errorcode: The error code, same as for srt_connect for blocking mode
• peeraddr: The target address passed to srt_connect call
• token: The token value, if it was used for group connection, otherwise -1

Socket group management

SRT_GROUP_TYPE

The following group types are collected in an SRT_GROUP_TYPE enum:

• SRT_GTYPE_BROADCAST: broadcast type, all links are actively used at once
• SRT_GTYPE_BACKUP: backup type, idle links take over connection on disturbance
• SRT_GTYPE_BALANCING: balancing type, share bandwidth usage between links

SRT_SOCKGROUPCONFIG

This structure is used to define entry points for connections for the srt_connect_group function:

typedef struct SRT_GroupMemberConfig_
{
    SRTSOCKET id;
    struct sockaddr_storage srcaddr;
    struct sockaddr_storage peeraddr;
    int weight;
    SRT_SOCKOPT_CONFIG* config;
    int errorcode;
    int token;
} SRT_SOCKGROUPCONFIG;

where:

• id: member socket ID (filled back as output)
• srcaddr: address to which id should be bound
• peeraddr: address to which id should be connected
• weight: the weight parameter for the link (group-type dependent)
• config: the configuration object, if used (see srt_create_config())
• errorcode: status of the connecting operation
• token: An integer value unique for every connection, or -1 if unused

The srt_perpare_endpoint sets these fields to default values. After that you can change the value of weight and config and token fields. The weight parameter's meaning is dependent on the group type:

• BROADCAST: not used
• BACKUP: positive value of link priority (the greater, the more preferred)
• BALANCING: relative expected load on this link for fixed algorithm

In any case, the allowed value ranges for weight is between 0 and 32767.

The config parameter is used to provide options to be set separately on a socket for a particular connection (see srt_create_config()).

The token value is intended to allow the application to more easily identify a particular connection. If you don't use it and leave the default value of -1, the library will set a unique value for the next connection (a 32-bit unsigned number that will overflow by itself; the default value will be skipped). The application can also set a unique value by itself and keep the same value for the same connection.

SRT_SOCKGROUPDATA

The most important structure for the group member status is SRT_SOCKGROUPDATA:

typedef struct SRT_SocketGroupData_
{
    SRTSOCKET id;
    struct sockaddr_storage peeraddr;
    SRT_SOCKSTATUS sockstate;
    uint16_t weight;
    SRT_MEMBERSTATUS memberstate;
    int result;
    int token;
} SRT_SOCKGROUPDATA;

where:

• id: member socket ID
• peeraddr: address to which id should be connected
• sockstate: current connection status (see srt_getsockstate)
• weight: current weight value set on the link
• memberstate: current state of the member (see below)
• result: result of the operation (if this operation recently updated this structure)
• token: A token value set for that connection (see SRT_SOCKGROUPCONFIG)

SRT_MEMBERSTATUS

The enumeration type that defines the state of the member connection in the group:

• SRT_GST_PENDING: The connection is in progress, so the socket is not currently being used for transmission, even potentially, and still has a chance to fail and transit into SRT_GST_BROKEN without turning into SRT_GST_IDLE

• SRT_GST_IDLE: The connection is established and ready to take over transmission, but it's not used for transmission at the moment. This state may last for a short moment in case of broadcast or balancing groups. In backup groups this state defines a backup link that is ready to take over when the currently active (running) link gets unstable.

• SRT_GST_RUNNING: The connection is established and at least one packet has already been sent or received over it.

• SRT_GST_BROKEN: The connection was broken. Broken connections are not to be revived. Note also that it is only possible to see this state if it is read by srt_sendmsg2 or srt_recvmsg2 just after the link failure has been detected. Otherwise, the broken link simply disappears from the member list.

Note that internally the member state is separate for sending and receiving. If the memberstate field of SRT_SOCKGROUPDATA is SRT_GST_RUNNING, it means that this is the state in at least one direction, while in the other direction it may be SRT_GST_IDLE. In all other cases the states should be the same in both directions.

States should normally start with SRT_GST_PENDING and then turn into SRT_GST_IDLE. Once a new link is used for sending data, the state becomes SRT_GST_RUNNING. In case of SRT_GTYPE_BACKUP type group, if a link is in SRT_GST_RUNNING state, but another link is chosen to remain as the only active one, this link will be "silenced" (its state will become SRT_GST_IDLE).

Functions to be used on groups:

srt_create_group

SRTSOCKET srt_create_group(SRT_GROUP_TYPE type);

Creates a new group of type type. This is typically called on the caller side to be next used for connecting to the listener peer side. The group ID is of the same domain as socket ID, with the exception that the SRTGROUP_MASK bit is set on it, unlike for socket ID.

srt_include

int srt_include(SRTSOCKET socket, SRTSOCKET group);

This function adds a socket to a group. This is only allowed for unmanaged groups. No such group type is currently implemented.

srt_exclude

int srt_exclude(SRTSOCKET socket);

This function removes a socket from a group to which it currently belongs. This is only allowed for unmanaged groups. No such group type is currently implemented.

srt_groupof

SRTSOCKET srt_groupof(SRTSOCKET socket);

Returns the group ID of the socket, or SRT_INVALID_SOCK if the socket doesn't exist or it's not a member of any group.

srt_group_data

int srt_group_data(SRTSOCKET socketgroup, SRT_SOCKGROUPDATA output[], size_t* inoutlen);

• socketgroup an existing socket group ID
• output points to an output array
• inoutlen points to a variable that stores the size of the output array, and is set to the filled array's size

This function obtains the current member state of the group specified in socketgroup. The output should point to an array large enough to hold all the elements. The inoutlen should point to a variable initially set to the size of the output array. The current number of members will be written back to inoutlen.

If the size of the output array is enough for the current number of members, the output array will be filled with group data and the function will return the number of elements filled. Otherwise the array will not be filled and SRT_ERROR will be returned.

This function can be used to get the group size by setting output to NULL, and providing socketgroup and inoutlen.

• Returns:

  • the number of data elements filled, on success
  • -1, on error

• Errors:

  • SRT_EINVPARAM reported if socketgroup is not an existing group ID
  • SRT_ELARGEMSG reported if inoutlen if less than the size of the group

in:output	in:inoutlen	returns	out:output	out:inoutlen	Error
NULL	NULL	-1	NULL	NULL	SRT_EINVPARAM
NULL	ptr	0	NULL	group.size()	✖️
ptr	NULL	-1	✖️	NULL	SRT_EINVPARAM
ptr	≥ group.size	group.size()	group.data	group.size	✖️
ptr	< group.size	-1	✖️	group.size	SRT_ELARGEMSG

srt_connect_group

int srt_connect_group(SRTSOCKET group,
                      SRT_SOCKGROUPCONFIG name [], int arraysize);

This function does almost the same as calling srt_connect or srt_connect_bind (when the source was specified for srt_prepare_endpoint) in a loop for every item specified in name array. However if you did this in blocking mode, the first call to srt_connect would block until the connection is established, whereas this function blocks until any of the specified connections is established.

If you set the group nonblocking mode (SRTO_RCVSYN option), there's no difference, except that the SRT_SOCKGROUPCONFIG structure allows you to add extra configuration data used by groups. Note also that this function accepts only groups, not sockets.

The elements of the name array need to be prepared with the use of the srt_prepare_endpoint function. Note that it is NOT required that every target address you specify for it is of the same family.

Return value and errors in this function are the same as in srt_connect, although this function reports success when at least one connection has succeeded. If none has succeeded, this function reports SRT_ECONNLOST error. Particular connection states can be obtained from the name array upon return from the errorcode field.

The fields of SRT_SOCKGROUPCONFIG structure have the following meaning:

Input:

• id: unused, should be -1 (default when created by srt_prepare_endpoint)
• srcaddr: address to bind before connecting, if specified (see below for details)
• peeraddr: target address to connect
• weight: weight value to be set on the link
• config: socket options to be set on the socket before connecting
• errorcode: unused, should be SRT_SUCCESS (default)
• token: An integer value unique for every connection, or -1 if unused

Output:

• id: The socket created for that connection (-1 if failed to create)
• srcaddr: unchanged
• peeraddr: unchanged
• weight: unchanged
• config: unchanged (the object should be manually deleted upon return)
• errorcode: status of connection for that link (SRT_SUCCESS if succeeded)
• token: same as in input, or a newly created token value if input was -1

The procedure of connecting for every connection definition specified in the name array is performed the following way:

1. The socket for this connection is first created

2. Socket options derived from the group are set on that socket.

3. If config is not NULL, configuration options stored there are set on that socket.

4. If source address is specified (that is srcaddr value is not default empty, as described in SRT_SOCKGROUPCONFIG), then the binding operation is being done on the socket (see srt_bind).

5. The socket is added to the group as a member.

6. The socket is being connected to the target address, as specified in the peeraddr field.

During this process there can be errors at any stage. There are two possibilities as to what may happen in this case:

1. If creation of a new socket has failed, which may only happen due to problems with system resources, then the whole loop is interrupted and no further items in the array are processed. All sockets that got created until then, and for which the connection attempt has at least successfully started, remain group members, although the function will return immediately with an error status (that is, without waiting for the first successful connection). If your application wants to do any partial recovery from this situation, it can only use epoll mechanism to wait for readiness.

2. In any other case, if an error occurs at any stage of the above process, the processing is interrupted for this very array item only, the socket used for it is immediately closed, and the processing of the next elements continues. In case of connection process, it also passes two stages - parameter check and the process itself. Failure at the parameter check breaks this process, while if this check passed, this item is considered correctly processed, even if the connection attempt is going to fail later. If this function is called in the blocking mode, it then blocks until at least one connection reports success or if all of them fail. Connections that continue in the background after this function exits can be then checked status by srt_group_data.

As member socket connections are running in the background, for determining if a particular connection has succeeded or failed it is recommended to use srt_connect_callback. In this case the token callback function parameter will be the same as the token value used for the particular item in the name connection table.

The token value doesn't have any limitations except that the -1 value is a "trap representation", that is, when set on input it will make the internals define a unique value for the token. Your application can also set unique values, in which case the token value will be preserved.

srt_prepare_endpoint

SRT_SOCKGROUPCONFIG srt_prepare_endpoint(const struct sockaddr* src /*nullable*/,
                                       const struct sockaddr* adr, int namelen);

This function prepares a default SRT_SOCKGROUPCONFIG object as an element of the array you can prepare for srt_connect_group function, filled with additional data:

• src: address to which the newly created socket should be bound
• adr: address to which the newly created socket should connect
• namelen: size of both src and adr

The following fields are set by this function:

• id: -1 (unused for input)
• srcaddr: default empty (see below) or copied from src
• peeraddr: copied from adr
• weight: 0
• config: NULL
• errorcode: SRT_SUCCESS

The default empty srcaddr is set the following way:

• ss_family set to the same value as adr->sa_family
• empty address (INADDR_ANY for IPv4 and in6addr_any for IPv6)
• port number 0

If src is not NULL, then srcaddr is copied from src. Otherwise it will remain as default empty.

The adr parameter is obligatory. If src parameter is not NULL, then both adr and src must have the same value of sa_family.

Note though that this function has no possibility of reporting errors - these would be reported only by srt_connect_group, separately for every individual connection, and the status can be obtained from errorcode field.

srt_create_config

SRT_SOCKOPT_CONFIG* srt_create_config();

Creates a dynamic object for specifying the socket options. You can add options to be set on the socket by srt_config_add and then mount this object into the config field in SRT_SOCKGROUPCONFIG object for that particular connection. After the object is no longer needed, you should delete it using srt_delete_config.

Returns:

• The pointer to the created object (memory allocation errors apply)

srt_delete_config

void srt_delete_config(SRT_SOCKOPT_CONFIG* c);

Deletes the configurartion object.

srt_config_add

int srt_config_add(SRT_SOCKOPT_CONFIG* c, SRT_SOCKOPT opt, void* val, int len);

Adds a configuration option to the configuration object. Parameters have meanings similar to srt_setsockflag. Note that not every option is allowed to be set this way. However, the option (if allowed) isn't checked if it doesn't violate other preconditions. This will be checked when the option is being set on the socket, which may fail as a part of the connection process done by srt_connect_group.

This function should be used when this option must be set individually on a socket and differently for particular link. If you need to set some option the same way on every socket, you should rather set this option on the whole group.

The following options are allowed to be set on the member socket:

• SRTO_SNDBUF: Allows for larger sender buffer for slower links
• SRTO_RCVBUF: Allows for larger receiver buffer for longer recovery
• SRTO_UDP_RCVBUF: UDP receiver buffer, if this link has a big flight window
• SRTO_UDP_SNDBUF: UDP sender buffer, if this link has a big flight window
• SRTO_SNDDROPDELAY: When particular link tends to drop too eagerly
• SRTO_NAKREPORT: If you don't want NAKREPORT to work for this link
• SRTO_CONNTIMEO: If you want to give more time to connect on this link
• SRTO_LOSSMAXTTL: If this link tends to suffer from UDP reordering
• SRTO_PEERIDLETIMEO: If you want to be more tolerant for temporary outages
• SRTO_GROUPSTABTIMEO: To set ACK jitter tolerance per individual link

Returns: 0 if succeeded, -1 when failed

Errors:

• SRT_EINVPARAM: this option is not allowed to be set on a socket being a group member

Options and properties

srt_getpeername

int srt_getpeername(SRTSOCKET u, struct sockaddr* name, int* namelen);

Retrieves the remote address to which the socket is connected.

• Returns:

  • SRT_ERROR (-1) in case of error, otherwise 0

• Errors:

  • SRT_EINVSOCK: Socket u indicates no valid socket ID
  • SRT_ENOCONN: Socket u isn't connected, so there's no remote address to return

srt_getsockname

int srt_getsockname(SRTSOCKET u, struct sockaddr* name, int* namelen);

Extracts the address to which the socket was bound. Although you should know the address(es) that you have used for binding yourself, this function can be useful for extracting the local outgoing port number when it was specified as 0 with binding for system autoselection. With this function you can extract the port number after it has been autoselected.

• Returns:

  • SRT_ERROR (-1) in case of error, otherwise 0

• Errors:

  • SRT_EINVSOCK: Socket u indicates no valid socket ID
  • SRT_ENOCONN: Socket u isn't bound, so there's no local address to return (BUG? It should rather be SRT_EUNBOUNDSOCK)

srt_getsockopt, srt_getsockflag

int srt_getsockopt(SRTSOCKET u, int level /*ignored*/, SRT_SOCKOPT opt, void* optval, int* optlen);
int srt_getsockflag(SRTSOCKET u, SRT_SOCKOPT opt, void* optval, int* optlen);

Gets the value of the given socket option (from a socket or a group). The first version (srt_getsockopt) respects the BSD socket API convention, although the "level" parameter is ignored. The second version (srt_getsockflag) omits the "level" parameter completely.

Options correspond to various data types, so you need to know what data type is assigned to a particular option, and to pass a variable of the appropriate data type. Specifications are provided in the apps/socketoptions.hpp file at the srt_options object declaration.

• Returns:

  • SRT_ERROR (-1) in case of error, otherwise 0

• Errors:

  • SRT_EINVSOCK: Socket u indicates no valid socket ID
  • SRT_EINVOP: Option opt indicates no valid option

srt_setsockopt, srt_setsockflag

int srt_setsockopt(SRTSOCKET u, int level /*ignored*/, SRT_SOCKOPT opt, const void* optval, int optlen);
int srt_setsockflag(SRTSOCKET u, SRT_SOCKOPT opt, const void* optval, int optlen);

Sets a value for a socket option in the socket or group. The first version (srt_setsockopt) respects the BSD socket API convention, although the "level" parameter is ignored. The second version (srt_setsockflag) omits the "level" parameter completely.

Options correspond to various data types, so you need to know what data type is assigned to a particular option, and to pass a variable of the appropriate data type with the option value to be set.

Please note that some of the options can only be set on sockets or only on groups, although most of the options can be set on the groups so that they are then derived by the member sockets.

• Returns:

  • SRT_ERROR (-1) in case of error, otherwise 0

-Errors:

• SRT_EINVSOCK: Socket u indicates no valid socket ID
• SRT_EINVOP: Option opt indicates no valid option
• Various other errors that may result from problems when setting a specific option (see option description for details).

srt_getversion

uint32_t srt_getversion();

Get SRT version value. The version format in hex is 0xXXYYZZ for x.y.z in human readable form, where x = ("%d", (version>>16) & 0xff), etc.

• Returns:

  • srt version as an unsigned 32-bit integer

Helper data types for transmission

SRT_MSGCTRL

The SRT_MSGCTRL structure:

typedef struct SRT_MsgCtrl_
{
   int flags;            // Left for future
   int msgttl;           // TTL for a message, default -1 (no TTL limitation)
   int inorder;          // Whether a message is allowed to supersede partially lost one. Unused in stream and live mode.
   int boundary;         // 0:mid pkt, 1(01b):end of frame, 2(11b):complete frame, 3(10b): start of frame
   int64_t srctime;      // source time (microseconds since SRT internal clock epoch)
   int32_t pktseq;       // sequence number of the first packet in received message (unused for sending)
   int32_t msgno;        // message number (output value for both sending and receiving)
} SRT_MSGCTRL;

The SRT_MSGCTRL structure is used in srt_sendmsg2 and srt_recvmsg2 calls and specifies some special extra parameters:

• flags: [IN, OUT]. RESERVED FOR FUTURE USE (should be 0). This is intended to specify some special options controlling the details of how the called function should work.

• msgttl: [IN]. In message and live mode only, specifies the TTL for sending messages (in [ms]). Not used for receiving messages. If this value is not negative, it defines the maximum time up to which this message should stay scheduled for sending for the sake of later retransmission. A message is always sent for the first time, but the UDP packet carrying it may be (also partially) lost, and if so, lacking packets will be retransmitted. If the message is not successfully resent before TTL expires, further retransmission is given up and the message is discarded.

• inorder: [IN]. In message mode only, specifies that sent messages should be extracted by the receiver in the order of sending. This can be meaningful if a packet loss has happened, and a particular message must wait for retransmission so that it can be reassembled and then delivered. When this flag is false, the message can be delivered even if there are any previous messages still waiting for completion.

• boundary: RESERVED FOR FUTURE USE. Intended to be used in a special mode when you are allowed to send or retrieve a part of the message.

• srctime:

  • [OUT] Receiver only. Specifies the time when the packet was intended to be delivered to the receiving application (in microseconds since SRT clock epoch).
  • [IN] Sender only. Specifies the application-provided timestamp to be asociated with the packet. If not provided (specified as 0), the current time of SRT internal clock is used.
  • For details on how to use srctime please refer to (Time Access)[#time-access] section.

• pktseq: Receiver only. Reports the sequence number for the packet carrying out the payload being returned. If the payload is carried out by more than one UDP packet, only the sequence of the first one is reported. Note that in live mode there's always one UDP packet per message.

• msgno: Message number that can be sent by both sender and receiver, although it is required that this value remain monotonic in subsequent send calls. Normally message numbers start with 1 and increase with every message sent.

Helpers for SRT_MSGCTRL:

void srt_msgctrl_init(SRT_MSGCTRL* mctrl);
const SRT_MSGCTRL srt_msgctrl_default;

Helpers for getting an object of SRT_MSGCTRL type ready to use. The first is a function that fills the object with default values. The second is a constant object and can be used as a source for assignment. Note that you cannot pass this constant object into any of the API functions because they require it to be mutable, as they use some fields to output values.

Transmission

srt_send, srt_sendmsg, srt_sendmsg2

int srt_send(SRTSOCKET u, const char* buf, int len);
int srt_sendmsg(SRTSOCKET u, const char* buf, int len, int ttl/* = -1*/, int inorder/* = false*/);
int srt_sendmsg2(SRTSOCKET u, const char* buf, int len, SRT_MSGCTRL *mctrl);

Sends a payload to a remote party over a given socket.

• u: Socket used to send. The socket must be connected for this operation.
• buf: Points to the buffer containing the payload to send.
• len: Size of the payload specified in buf.
• ttl: Time (in [ms]) to wait for a successful delivery. See description of the SRT_MSGCTRL::msgttl field.
• inorder: Required to be received in the order of sending. See SRT_MSGCTRL::inorder.
• mctrl: An object of SRT_MSGCTRL type that contains extra parameters, including ttl and inorder.

The way this function works is determined by the mode set in options, and it has specific requirements:

1. In file/stream mode, the payload is byte-based. You are not required to know the size of the data, although they are only guaranteed to be received in the same byte order.

2. In file/message mode, the payload that you send using this function is a single message that you intend to be received as a whole. In other words, a single call to this function determines a message's boundaries.

3. In live mode, you are only allowed to send up to the length of SRTO_PAYLOADSIZE, which can't be larger than 1456 bytes (1316 default).

• Returns:

  • Size of the data sent, if successful. Note that in file/stream mode the returned size may be less than len, which means that it didn't send the whole contents of the buffer. You would need to call this function again with the rest of the buffer next time to send it completely. In both file/message and live mode the successful return is always equal to len
  • In case of error, SRT_ERROR (-1)

• Errors:

  • SRT_ENOCONN: Socket u used when the operation is not connected.
  • SRT_ECONNLOST: Socket u used for the operation has lost its connection.
  • SRT_EINVALMSGAPI: Incorrect API usage in message mode:
    • live mode: trying to send more bytes at once than SRTO_PAYLOADSIZE or wrong source time was provided.
  • SRT_EINVALBUFFERAPI: Incorrect API usage in stream mode:
    • Reserved for future use. The congestion controller object used for this mode doesn't use any restrictions on this call for now, but this may change in future.
  • SRT_ELARGEMSG: Message to be sent can't fit in the sending buffer (that is, it exceeds the current total space in the sending buffer in bytes). This means that the sender buffer is too small, or the application is trying to send a larger message than initially predicted.
  • SRT_EASYNCSND: There's no free space currently in the buffer to schedule the payload. This is only reported in non-blocking mode (SRTO_SNDSYN set to false); in blocking mode the call is blocked until enough free space in the sending buffer becomes available.
  • SRT_ETIMEOUT: The condition described above still persists and the timeout has passed. This is only reported in blocking mode when SRTO_SNDTIMEO is set to a value other than -1.
  • SRT_EPEERERR: This is reported only in the case where, as a stream is being received by a peer, the srt_recvfile function encounters an error during a write operation on a file. This is reported by a UMSG_PEERERROR message from the peer, and the agent sets the appropriate flag internally. This flag persists up to the moment when the connection is broken or closed.

srt_recv, srt_recvmsg, srt_recvmsg2

int srt_recv(SRTSOCKET u, char* buf, int len);
int srt_recvmsg(SRTSOCKET u, char* buf, int len);
int srt_recvmsg2(SRTSOCKET u, char *buf, int len, SRT_MSGCTRL *mctrl);

Extracts the payload waiting to be received. Note that srt_recv and srt_recvmsg are identical functions, two different names being kept for historical reasons. In the UDT predecessor the application was required to use either the UDT::recv version for stream mode and UDT::recvmsg for message mode. In SRT this distinction is resolved internally by the SRTO_MESSAGEAPI flag.

• u: Socket used to send. The socket must be connected for this operation.
• buf: Points to the buffer to which the payload is copied
• len: Size of the payload specified in buf
• mctrl: An object of SRT_MSGCTRL type that contains extra parameters

The way this function works is determined by the mode set in options, and it has specific requirements:

1. In file/stream mode, as many bytes as possible are retrieved, that is, only so many bytes that fit in the buffer and are currently available. Any data that is available but not extracted this time will be available next time.

2. In file/message mode, exactly one message is retrieved, with the boundaries defined at the moment of sending. If some parts of the messages are already retrieved, but not the whole message, nothing will be received (the function blocks or returns SRT_EASYNCRCV). If the message to be returned does not fit in the buffer, nothing will be received and the error is reported.

3. In live mode, the function behaves as in file/message mode, although the number of bytes retrieved will be at most the size of SRTO_PAYLOADSIZE. In this mode, however, with default settings of SRTO_TSBPDMODE and SRTO_TLPKTDROP, the message will be received only when its time to play has come, and until then it will be kept in the receiver buffer; also, when the time to play has come for a message that is next to the currently lost one, it will be delivered and the lost one dropped.

• Returns:

  • Size (>0) of the data received, if successful.
  • 0, if the connection has been closed
  • SRT_ERROR (-1) when an error occurs

• Errors:

  • SRT_ENOCONN: Socket u used for the operation is not connected.
  • SRT_ECONNLOST: Socket u used for the operation has lost connection (this is reported only if the connection was unexpectedly broken, not when it was closed by the foreign host).
  • SRT_EINVALMSGAPI: Incorrect API usage in message mode:
    • live mode: size of the buffer is less than SRTO_PAYLOADSIZE
  • SRT_EINVALBUFFERAPI: Incorrect API usage in stream mode:
    • Currently not in use. File congestion control used for stream mode does not restrict the parameters. ???
  • SRT_ELARGEMSG: Message to be sent can't fit in the sending buffer (that is, it exceeds the current total space in the sending buffer in bytes). This means that the sender buffer is too small, or the application is trying to send a larger message than initially intended.
  • SRT_EASYNCRCV: There are no data currently waiting for delivery. This happens only in non-blocking mode (when SRTO_RCVSYN is set to false). In blocking mode the call is blocked until the data are ready. How this is defined, depends on the mode:
  • In live mode (with SRTO_TSBPDMODE on), at least one packet must be present in the receiver buffer and its time to play be in the past
  • In file/message mode, one full message must be available,
    • the next one waiting if there are no messages with inorder = false, or possibly the first message ready with inorder = false
  • In file/stream mode, it is expected to have at least one byte of data still not extracted
  • SRT_ETIMEOUT: The readiness condition described above is still not achieved and the timeout has passed. This is only reported in blocking mode when SRTO_RCVTIMEO is set to a value other than -1.

srt_sendfile, srt_recvfile

int64_t srt_sendfile(SRTSOCKET u, const char* path, int64_t* offset, int64_t size, int block);
int64_t srt_recvfile(SRTSOCKET u, const char* path, int64_t* offset, int64_t size, int block);

These are functions dedicated to sending and receiving a file. You need to call this function just once for the whole file, although you need to know the size of the file prior to sending and also define the size of a single block that should be internally retrieved and written into a file in a single step. This influences only the performance of the internal operations; from the application perspective you just have one call that exits only when the transmission is complete.

• u: Socket used for transmission. The socket must be connected.
• path: Path to the file that should be read or written.
• offset: Needed to pass or retrieve the offset used to read or write to a file
• size: Size of transfer (file size, if offset is at 0)
• block: Size of the single block to read at once before writing it to a file

The following values are recommended for the block parameter:

#define SRT_DEFAULT_SENDFILE_BLOCK 364000
#define SRT_DEFAULT_RECVFILE_BLOCK 7280000

You need to pass them to the srt_sendfile or srt_recvfile function if you don't know what value to chose.

• Returns:

  • Size (>0) of the transmitted data of a file. It may be less than size, if the size was greater than the free space in the buffer, in which case you have to send rest of the file next time.
  • -1 in case of error.

• Errors:

  • SRT_ENOCONN: Socket u used for the operation is not connected.
  • SRT_ECONNLOST: Socket u used for the operation has lost its connection.
  • SRT_EINVALBUFFERAPI: When socket has SRTO_MESSAGEAPI = true or SRTO_TSBPDMODE = true. (BUG?: Looxlike MESSAGEAPI isn't checked)
  • SRT_EINVRDOFF: There is a mistake in offset or size parameters, which should match the index availability and size of the bytes available since offset index. This is actually reported for srt_sendfile when the seekg or tellg operations resulted in error.
  • SRT_EINVWROFF: Like above, reported for srt_recvfile and seekp/tellp.
  • SRT_ERDPERM: The read from file operation has failed (srt_sendfile).
  • SRT_EWRPERM: The write to file operation has failed (srt_recvfile).

Diagnostics

General notes concerning the "getlasterror" diagnostic functions: when an API function ends up with error, this error information is stored in a thread-local storage. This means that you'll get the error of the operation that was last performed as long as you call this diagnostic function just after the failed function has returned. In any other situation the information provided by the diagnostic function is undefined.

srt_getlasterror

int srt_getlasterror(int* errno_loc);

Get the numeric code of the last error. Additionally, in the variable passed as errno_loc the system error value is returned, or 0 if there was no system error associated with the last error. The system error is:

• On POSIX systems, the value from errno
• On Windows, the result from GetLastError() call

srt_strerror

const char* srt_strerror(int code, int errnoval);

Returns a string message that represents a given SRT error code and possibly the errno value, if not 0.

NOTE: This function isn't thread safe. It uses a static variable to hold the error description. There's no problem with using it in a multithreaded environment, as long as only one thread in the whole application calls this function at the moment

srt_getlasterror_str

const char* srt_getlasterror_str(void);

Get the text message for the last error. It's a shortcut to calling first srt_getlasterror and then passing the returned value into srt_strerror. Note that, in contradiction to srt_strerror, this function is thread safe.

srt_clearlasterror

void srt_clearlasterror(void);

This function clears the last error. After this call, the srt_getlasterror will report a "successful" code.

srt_getrejectreason

int srt_getrejectreason(SRTSOCKET sock);

This function provides a more detailed reason for the failed connection attempt. It shall be called after a connecting function (such as srt_connect) has returned an error, the code for which is SRT_ECONNREJ. If SRTO_RCVSYN has been set on the socket used for the connection, the function should also be called when the SRT_EPOLL_ERR event is set for this socket. It returns a numeric code, which can be translated into a message by srt_rejectreason_str. The following codes are currently reported:

SRT_REJ_UNKNOWN

A fallback value for cases when there was no connection rejected.

SRT_REJ_SYSTEM

One of system function reported a failure. Usually this means some system error or lack of system resources to complete the task.

SRT_REJ_PEER

The connection has been rejected by peer, but no further details are available. This usually means that the peer doesn't support rejection reason reporting.

SRT_REJ_RESOURCE

A problem with resource allocation (usually memory).

SRT_REJ_ROGUE

The data sent by one party to another cannot be properly interpreted. This should not happen during normal usage, unless it's a bug, or some weird events are happening on the network.

SRT_REJ_BACKLOG

The listener's backlog has exceeded (there are many other callers waiting for the opportunity of being connected and wait in the queue, which has reached its limit).

SRT_REJ_IPE

Internal Program Error. This should not happen during normal usage and it usually means a bug in the software (although this can be reported by both local and foreign host).

SRT_REJ_CLOSE

The listener socket was able to receive your request, but at this moment it is being closed. It's likely that your next attempt will result with timeout.

SRT_REJ_VERSION

Any party of the connection has set up minimum version that is required for that connection, and the other party didn't satisfy this requirement.

SRT_REJ_RDVCOOKIE

Rendezvous cookie collision. This normally should never happen, or the probability that this will really happen is negligible. However this can be also a result of a misconfiguration that you are trying to make a rendezvous connection where both parties try to bind to the same IP address, or both are local addresses of the same host - in which case the sent handshake packets are returning to the same host as if they were sent by the peer, who is this party itself. When this happens, this reject reason will be reported by every attempt.

SRT_REJ_BADSECRET

Both parties have defined a passprhase for connection and they differ.

SRT_REJ_UNSECURE

Only one connection party has set up a password. See also SRTO_ENFORCEDENCRYPTION flag in API.md.

SRT_REJ_MESSAGEAPI

The value for SRTO_MESSAGEAPI flag is different on both connection parties.

SRT_REJ_CONGESTION

The SRTO_CONGESTION option has been set up differently on both connection parties.

SRT_REJ_FILTER

The SRTO_PACKETFILTER option has been set differently on both connection parties.

SRT_REJ_GROUP

The group type or some group settings are incompatible for both connection parties. While every connection within a bonding group may have different target addresses, they should all designate the same endpoint and the same SRT application. If this condition isn't satisfied, then the peer will respond with a different peer group ID for the connection that is trying to contact a machine/application that is completely different from the existing connections in the bonding group.

SRT_REJ_TIMEOUT

The connection wasn't rejected, but it timed out. This code is always set on connection timeout, but this is the only way to get this state in non-blocking mode (see SRTO_RCVSYN).

There may also be server and user rejection codes, as defined by the SRT_REJC_INTERNAL, SRT_REJC_PREDEFINED and SRT_REJC_USERDEFINED constants. Note that the number space from the value of SRT_REJC_PREDEFINED and above is reserved for "predefined codes" (SRT_REJC_PREDEFINED value plus adopted HTTP codes). Values above SRT_REJC_USERDEFINED are freely defined by the application.

srt_rejectreason_str

const char* srt_rejectreason_str(enum SRT_REJECT_REASON id);

Returns a constant string for the reason of the connection rejected, as per given code ID. It provides a system-defined message for values below SRT_REJ_E_SIZE. For other values below SRT_REJC_PREDEFINED it returns the string for SRT_REJ_UNKNOWN. For values since SRT_REJC_PREDEFINED on, returns "Application-defined rejection reason".

The actual messages assigned to the internal rejection codes, that is, less than SRT_REJ_E_SIZE, can be also obtained from srt_rejectreason_msg array.

srt_setrejectreason

int srt_setrejectreason(SRTSOCKET sock, int value);

Sets the rejection code on the socket. This call is only useful in the listener callback. The code from value set this way will be set as a rejection reason for the socket. After the callback rejects the connection, the code will be passed back to the caller peer with the handshake response.

Note that allowed values for this function begin with SRT_REJC_PREDEFINED (that is, you cannot set a system rejection code). For example, your application can inform the calling side that the resource specified under the r key in the StreamID string (see SRTO_STREAMID) is not availble - it then sets the value to SRT_REJC_PREDEFINED + 404.

• Returns:

  • 0 in case of success.
  • -1 in case of error.

• Errors:

  • SRT_EINVSOCK: Socket sock is not an ID of a valid socket
  • SRT_EINVPARAM: value is less than SRT_REJC_PREDEFINED

Error codes

All functions that return the status via int value return -1 (designated as SRT_ERROR) always when the call has failed (in case of resource creation functions an appropriate symbol is defined, like SRT_INVALID_SOCK for SRTSOCKET). When this happens, the error code can be obtained from the srt_getlasterror function. The values for the error are collected in an SRT_ERRNO enum:

SRT_EUNKNOWN

Internal error when setting the right error code.

SRT_SUCCESS

The value set when the last error was cleared and no error has occurred since then.

SRT_ECONNSETUP

General setup error resulting from internal system state.

SRT_ENOSERVER

Connection timed out while attempting to connect to the remote address. Note that when this happens, srt_getrejectreason also reports the timeout reason.

SRT_ECONNREJ

Connection has been rejected. Additional reject reason can be obtained through srt_getrejectreason (see above).

SRT_ESOCKFAIL

An error occurred when trying to call a system function on an internally used UDP socket. Note that the detailed system error is available in the extra variable passed by pointer to srt_getlasterror.

SRT_ESECFAIL

A possible tampering with the handshake packets was detected, or encryption request wasn't properly fulfilled.

SRT_ESCLOSED

A socket that was vital for an operation called in blocking mode has been closed during the operation. Please note that this situation is handled differently than the system calls for connect and accept functions for TCP, which simply block indefinitely (or until the standard timeout) when the key socket was closed during an operation. When this error is reported, it usually means that the socket passed as the first parameter to srt_connect* or srt_accept is no longer usable.

SRT_ECONNFAIL

General connection failure of unknown details.

SRT_ECONNLOST

The socket was properly connected, but the connection has been broken. This specialzation is reported from the transmission functions.

SRT_ENOCONN

The socket is not connected. This can be reported also when the connection was broken for a function that checks some characteristic socket data.

SRT_ERESOURCE

System or standard library error reported unexpectedly for unknown purpose. Usually it means some internal error.

SRT_ETHREAD

System was unable to spawn a new thread when requried.

SRT_ENOBUF

System was unable to allocate memory for buffers.

SRT_ESYSOBJ

System was unable to allocate system specific objects (such as sockets, mutexes or condition variables).

SRT_EFILE

General filesystem error (for functions operating with file transmission).

SRT_EINVRDOFF

Failure when trying to read from a given position in the file (file could be modified while it was read from).

SRT_ERDPERM

Read permission was denied when trying to read from file.

SRT_EINVWROFF

Failed to set position in the written file.

SRT_EWRPERM

Write permission was denied when trying to write to a file.

SRT_EINVOP

Invalid operation performed for the current state of a socket. This mainly concerns performing srt_bind* operations on a socket that is already bound. Once a socket has been been bound, it cannot be bound again.

SRT_EBOUNDSOCK

The socket is currently bound and the required operation cannot be performed in this state. Usually it's about an option that can only be set on the socket before binding (srt_bind*). Note that a socket that is currently connected is also considered bound.

SRT_ECONNSOCK

The socket is currently connected and therefore performing the required operation is not possible. Usually concerns setting an option that must be set before connecting (although it is allowed to be altered after binding), or when trying to start a connecting operation (srt_connect*) while the socket isn't in a state that allows it (only SRTS_INIT or SRTS_OPENED are allowed).

SRT_EINVPARAM

This error is reported in a variety of situations when call parameters for API functions have some requirements defined and these were not satisfied. This error should be reported after an initial check of the parameters of the call before even performing any operation. This error can be easily avoided if you set the values correctly.

SRT_EINVSOCK

The API function required an ID of an entity (socket or group) and it was invalid. Note that some API functions work only with socket or only with group, so they would also return this error if inappropriate type of entity was passed, even if it was valid.

SRT_EUNBOUNDSOCK

The operation to be performed on a socket requires that it first be explicitly bound (using srt_bind* functions). Currently it applies when calling srt_listen, which cannot work with an implicitly bound socket.

SRT_ENOLISTEN

The socket passed for the operation is required to be in the listen state (srt_listen must be called first).

SRT_ERDVNOSERV

The required operation cannot be performed when the socket is set to rendezvous mode (SRTO_RENDEZVOUS set to true). Usually applies when trying to call srt_listen on such a socket.

SRT_ERDVUNBOUND

An attempt was made to connect to a socket set to rendezvous mode (SRTO_RENDEZVOUS set to true) that was not first bound. A rendezvous connection requires setting up two addresses and ports on both sides of the connection, then setting the local one with srt_bind and using the remote one with srt_connect (or you can simply use srt_rendezvous). Calling srt_connect* on an unbound socket (in SRTS_INIT state) that is to be bound implicitly is only allowed for regular caller sockets (not rendezvous).

SRT_EINVALMSGAPI

The function was used incorrectly in the message API. This can happen if:

• The parameters specific for the message API in SRT_MSGCTRL type parameter were incorrectly specified

• The extra parameter check performed by the congestion controller has failed

• The socket is a member of a self-managing group, therefore you should perform the operation on the group, not on this socket

SRT_EINVALBUFFERAPI

The function was used incorrectly in the stream (buffer) API, that is, either the stream-only functions were used with set message API (srt_sendfile/srt_recvfile) or TSBPD mode was used with buffer API (SRTO_TSBPDMODE set to true) or the congestion controller has failed to check call parameters.

SRT_EDUPLISTEN

The port tried to be bound for listening is already busy. Note that binding to the same port is allowed in general (when SRTO_REUSEADDR is true on every socket that bound it), but only one such socket can be a listener.

SRT_ELARGEMSG

Size exceeded. This is reported in the following situations:

• Trying to receive a message, but the read-ready message is larger than the buffer passed to the receiving function

• Trying to send a message, but the size of this message exceeds the size of the preset sender buffer, so it cannot be stored in the sender buffer.

• With getting group data, the array to be filled is too small.

SRT_EINVPOLLID

The epoll ID passed to an epoll function is invalid

SRT_EPOLLEMPTY

The epoll container currently has no subscribed sockets. This is reported by an epoll waiting function that would in this case block forever. This problem might be reported both in a situation where you have created a new epoll container and didn't subscribe any sockets to it, or you did, but these sockets have been closed (including when closed in a separate thread while the waiting function was blocking). Note that this situation can be prevented by setting the SRT_EPOLL_ENABLE_EMPTY flag, which may be useful when you use multiple threads and start waiting without subscribed sockets, so that you can subscribe them later from another thread.

SRT_EASYNCFAIL

General asynchronous failure (not in use currently).

SRT_EASYNCSND

Sending operation is not ready to perform. This error is reported when trying to perform a sending operation on a socket that is not ready for sending, but SRTO_SNDSYN was set to false (when true, the function would block the call otherwise).

SRT_EASYNCRCV

Receiving operation is not ready to perform. This error is reported when trying to perform a receiving operation or accept a new socket from the listener socket, when the socket is not ready for that operation, but SRTO_RCVSYN was set to false (when true, the function would block the call otherwise).

SRT_ETIMEOUT

The operation timed out. This can happen if you have a timeout set by an option (SRTO_RCVTIMEO or SRTO_SNDTIMEO), or passed as an extra argument (srt_epoll_wait or srt_accept_bond) and the function call was blocking, but the required timeout time has passed.

SRT_ECONGEST

NOTE: This error is used only in an experimental version that requires setting the SRT_ENABLE_ECN macro at compile time. Otherwise the situation described below results in the usual successful report.

This error should be reported by the sending function when, with SRTO_TSBPDMODE and SRTO_TLPKTDROP set to true, some packets were dropped at the sender side (see the description of SRTO_TLPKTDROP for details). This doesn't concern the data that were passed for sending by the sending function (these data are placed at the back of the sender buffer, while the dropped packets are at the front). In other words, the operation done by the sending function is successful, but the application might want to slow down the sending rate to avoid congestion.

SRT_EPEERERR

This error is reported in a situation when the receiver peer is writing to a file that the agent is sending. When the peer encounters an error when writing the received data to a file, it sends the UMSG_PEERERROR message back to the sender, and the sender reports this error from the API sending function.

Performance tracking

General note concerning sequence numbers used in SRT: they are 32-bit "circular numbers" with the most significant bit not included. For example 0x7FFFFFFF shifted forward by 3 becomes 2. As far as any comparison is concerned, it can be thought of as a "distance" which is an integer value expressing an offset to be added to one sequence in order to get the second one. This distance is only valid as long as the threshold value isn't exceeded, so it's stated that all sequence numbers that are anywhere taken into account were systematically updated and they are kept in the range between 0 and half of the maximum 0x7FFFFFFF. Hence the distance counting procedure always assumes that the sequence number are in the required range already, so for a numbers like 0x7FFFFFF0 and 0x10, for which the "numeric difference" would be 0x7FFFFFE0, the "distance" is 0x20.

srt_bstats, srt_bistats

// Performance monitor with Byte counters for better bitrate estimation.
int srt_bstats(SRTSOCKET u, SRT_TRACEBSTATS * perf, int clear);

// Performance monitor with Byte counters and instantaneous stats instead of moving averages for Snd/Rcvbuffer sizes.
int srt_bistats(SRTSOCKET u, SRT_TRACEBSTATS * perf, int clear, int instantaneous);

Reports the current statistics

• u: Socket from which to get statistics
• perf: Pointer to an object to be written with the statistics
• clear: 1 if the statistics should be cleared after retrieval
• instantaneous: 1 if the statistics should use instant data, not moving averages

SRT_TRACEBSTATS is an alias to struct CBytePerfMon. For a complete description of the fields please refer to the document statistics.md.

Asynchronous operations (epoll)

The epoll system is currently the only method for using multiple sockets in one thread with having the blocking operation moved to epoll waiting so that it can block on multiple sockets at once. That is, instead of blocking a single reading or writing operation, as it's in blocking mode, it blocks until at least one of the sockets subscribed for a single waiting call in given operation mode is ready to do this operation without blocking. It's usually combined with setting the nonblocking mode on a socket, which in SRT is set separately for reading and writing (SRTO_RCVSYN and SRTO_SNDSYN respectively) in order to ensure that in case of some internal error in the application (or even possibly a bug in SRT that has reported a spurious readiness report) the operation will end up with error rather than cause blocking, which would be more dangerous for the application in this case (SRT_EASYNCRCV and SRT_EASYNCRCV respectively).

The epoll system, similar to the one on Linux, relies on eid objects managed internally in SRT, which can be subscribed to particular sockets and the readiness status of particular operations. The srt_epoll_wait function can then be used to block until any readiness status in the whole eid is set.

srt_epoll_create

int srt_epoll_create(void);

Creates a new epoll container.

• Returns:

  • valid EID on success
  • -1 on failure

• Errors:

  • SRT_ECONNSETUP: System operation failed or not enough space to create a new epoll. System error might happen on systems that use a special method for the system part of epoll (epoll_create(), kqueue()), and therefore associated resources, like epoll on Linux.

srt_epoll_add_usock, srt_epoll_add_ssock, srt_epoll_update_usock, srt_epoll_update_ssock

int srt_epoll_add_usock(int eid, SRTSOCKET u, const int* events);
int srt_epoll_add_ssock(int eid, SYSSOCKET s, const int* events);
int srt_epoll_update_usock(int eid, SRTSOCKET u, const int* events);
int srt_epoll_update_ssock(int eid, SYSSOCKET s, const int* events);

Adds a socket to a container, or updates an existing socket subscription.

The _usock suffix refers to a user socket (SRT socket). The _ssock suffix refers to a system socket.

The _add_ functions add new sockets. The _update_ functions act on a socket that is in the container already and just allow changes in the subscription details. For example, if you have already subscribed a socket with SRT_EPOLL_OUT to wait until it's connected, to change it into poll for read-readiness, you use this function on that same socket with a variable set to SRT_EPOLL_IN. This will not only change the event type which is polled on the socket, but also remove any readiness status for flags that are no longer set. It is discouraged to perform socket removal and adding back (instead of using _update_) because this way you may miss an event that could happen in a short moment between these two calls.

• eid: epoll container id
• u: SRT socket
• s: system socket
• events: points to
  • a variable set to epoll flags (see below) to use only selected events
  • NULL if you want to subscribe a socket for all events in level-triggered mode

Possible epoll flags are the following:

• SRT_EPOLL_IN: report readiness for reading or incoming connection on a listener socket
• SRT_EPOLL_OUT: report readiness for writing or a successful connection
• SRT_EPOLL_ERR: report errors on the socket
• SRT_EPOLL_UPDATE: an important event has happened that requires attention
• SRT_EPOLL_ET: the event will be edge-triggered

All flags except SRT_EPOLL_ET are event type flags (important for functions that expect only event types and not other flags).

The SRT_EPOLL_IN, SRT_EPOLL_OUT and SRT_EPOLL_ERR events are by default level-triggered. With SRT_EPOLL_ET flag they become edge-triggered.

The SRT_EPOLL_UPDATE flag is always edge-triggered. It designates a special event that happens on a group, or on a listener socket that has the SRTO_GROUPCONNECT flag set to allow group connections. This flag is triggered in the following situations:

• for group connections, when a new link has been established for a group that is already connected (that is, has at least one connection established), SRT_EPOLL_UPDATE is reported for the listener socket accepting the connection. This is intended for internal use only. An initial connection results in reporting the group connection on that listener. But when the group is already connected, SRT_EPOLL_UPDATE is reported instead.

• when one of group member connection has been broken

Note that at this time the edge-triggered mode is supported only for SRT sockets, not for system sockets.

In the edge-triggered mode the function will only return socket states that have changed since the last call of the waiting function. All events reported in particular call of the waiting function will be cleared in the internal flags and will not be reported until the internal signaling logic clears this state and raises it again.

In the level-triggered mode the function will always return the readiness state as long as it lasts, until the internal signaling logic clear it.

Note that when you use SRT_EPOLL_ET flag in one subscription call, it defines edge-triggered mode for all events passed together with it. However, if you want to have some events reported as edge-triggered and others as level-triggered, you can do two separate subscriptions for the same socket.

IMPORTANT: The srt_epoll_wait function does not report SRT_EPOLL_UPDATE events. If you need the ability to get any possible flag, you must use srt_epoll_uwait. Note that this function doesn't work with system file descriptors.

• Returns:

  • 0 if successful, otherwise -1

• Errors:

  • SRT_EINVPOLLID: eid parameter doesn't refer to a valid epoll container

BUG?: for add_ssock the system error results in an empty CUDTException() call which actually results in SRT_SUCCESS. For cases like that the SRT_ECONNSETUP code is predicted.

srt_epoll_remove_usock, srt_epoll_remove_ssock

int srt_epoll_remove_usock(int eid, SRTSOCKET u);
int srt_epoll_remove_ssock(int eid, SYSSOCKET s);

Removes a specified socket from an epoll container and clears all readiness states recorded for that socket.

The _usock suffix refers to a user socket (SRT socket). The _ssock suffix refers to a system socket.

• Returns:

  • 0 if successful, otherwise -1

• Errors:

  • SRT_EINVPOLLID: eid parameter doesn't refer to a valid epoll container

srt_epoll_wait

int srt_epoll_wait(int eid, SRTSOCKET* readfds, int* rnum, SRTSOCKET* writefds, int* wnum, int64_t msTimeOut,
                        SYSSOCKET* lrfds, int* lrnum, SYSSOCKET* lwfds, int* lwnum);

Blocks the call until any readiness state occurs in the epoll container.

Readiness can be on a socket in the container for the event type as per subscription. Note that in case when particular event was subscribed with SRT_EPOLL_ET flag, this event, when once reported in this function, will be cleared internally.

The first readiness state causes this function to exit, but all ready sockets are reported. This function blocks until the timeout specified in msTimeOut parameter. If timeout is 0, it exits immediately after checking. If timeout is -1, it blocks indefinitely until a readiness state occurs.

• eid: epoll container
• readfds and rnum: A pointer and length of an array to write SRT sockets that are read-ready
• writefds and wnum: A pointer and length of an array to write SRT sockets that are write-ready
• msTimeOut: Timeout specified in milliseconds, or special values (0 or -1)
• lwfds and lwnum:A pointer and length of an array to write system sockets that are read-ready
• lwfds and lwnum:A pointer and length of an array to write system sockets that are write-ready

Note that the following flags are reported:

• SRT_EPOLL_IN as read-ready (also a listener socket ready to accept)
• SRT_EPOLL_OUT as write-ready (also a connected socket)
• SRT_EPOLL_ERR as both read-ready and write-ready
• SRT_EPOLL_UPDATE is not reported

There is no space here to report sockets for which it's already known that the operation will end up with error (athough such a state is known internally). If an error occurred on a socket then that socket is reported in both read-ready and write-ready arrays, regardless of what event types it was subscribed for. Usually then you subscribe given socket for only read readiness, for example (SRT_EPOLL_IN), but pass both arrays for read and write readiness. This socket will not be reported in the write readiness array even if it's write ready (because this isn't what it was subscribed for), but it will be reported there, if the next operation on this socket is about to be erroneous. On such sockets you can still perform an operation, just you should expect that it will always report and error. On the other hand that's the only way to know what kind of error has occurred on the socket.

• Returns:

  • The number (>0) of ready sockets, of whatever kind (if any)
  • -1 in case of error

• Errors:

  • SRT_EINVPOLLID: eid parameter doesn't refer to a valid epoll container
  • SRT_ETIMEOUT: Up to msTimeOut no sockets subscribed in eid were ready. This is reported only if msTimeOut was >=0, otherwise the function waits indefinitely.

srt_epoll_uwait

int srt_epoll_uwait(int eid, SRT_EPOLL_EVENT* fdsSet, int fdsSize, int64_t msTimeOut);

This function blocks a call until any readiness state occurs in the epoll container. Unlike srt_epoll_wait, it can only be used with eid subscribed to user sockets (SRT sockets), not system sockets.

This function blocks until the timeout specified in msTimeOut parameter. If timeout is 0, it exits immediately after checking. If timeout is -1, it blocks indefinitely until a readiness state occurs.

• eid: epoll container
• fdsSet : A pointer to an array of SRT_EPOLL_EVENT
• fdsSize : The size of the fdsSet array
• msTimeOut : Timeout specified in milliseconds, or special values (0 or -1):
  • 0: Don't wait, return immediately (report any sockets currently ready)
  • -1: Wait indefinitely.

• Returns:

  • The number of user socket (SRT socket) state changes that have been reported in fdsSet, if this number isn't greater than fdsSize

  • Otherwise the return value is fdsSize + 1. This means that there was not enough space in the output array to report all events. For events subscribed with SRT_EPOLL_ET flag only those will be cleared that were reported. Others will wait for the next call.

  • If no readiness state was found on any socket and the timeout has passed, 0 is returned (this is not possible when waiting indefinitely)

  • -1 in case of error

• Errors:

  • SRT_EINVPOLLID: eid parameter doesn't refer to a valid epoll container
  • SRT_EINVPARAM: One of possible usage errors:
    • fdsSize is < 0
    • fdsSize is > 0 and fdsSet is a null pointer
    • eid was subscribed to any system socket

(IMPORTANT: this function reports timeout by returning 0, not by SRT_ETIMEOUT error.)

The SRT_EPOLL_EVENT structure:

typedef struct SRT_EPOLL_EVENT_
{
    SRTSOCKET fd;
    int       events;
} SRT_EPOLL_EVENT;

• fd : the user socket (SRT socket)
• events : event flags that report readiness of this socket - a combination of SRT_EPOLL_IN, SRT_EPOLL_OUT and SRT_EPOLL_ERR - see srt_epoll_add_usock for details

Note that when the SRT_EPOLL_ERR is set, the underlying socket error can't be retrieved with srt_getlasterror(). The socket will be automatically closed and its state can be verified with a call to srt_getsockstate.

srt_epoll_clear_usocks

int srt_epoll_clear_usocks(int eid);

This function removes all SRT ("user") socket subscriptions from the epoll container identified by eid.

• Returns:

  • 0 on success
  • -1 in case of error

• Errors:

  • SRT_EINVPOLLID: eid parameter doesn't refer to a valid epoll container

srt_epoll_set

int32_t srt_epoll_set(int eid, int32_t flags);

This function allows to set or retrieve flags that change the default behavior of the epoll functions. All default values for these flags are 0. The following flags are available:

• SRT_EPOLL_ENABLE_EMPTY: allows the srt_epoll_wait and srt_epoll_uwait functions to be called with the EID not subscribed to any socket. The default behavior of these function is to report error in this case.

• SRT_EPOLL_ENABLE_OUTPUTCHECK: Forces the srt_epoll_wait and srt_epoll_uwait functions to check if the output array is not empty. For srt_epoll_wait it is still allowed that either system or user array is empty, as long as EID isn't subscribed to this type of socket/fd. srt_epoll_uwait only checks if the general output array is not empty.

• Parameters:

  • eid: the epoll container id
  • flags: a nonzero set of the above flags, or special values:
    • 0: clear all flags (set all defaults)
    • -1: do not modify any flags

• Returns:

This function returns the state of the flags at the time before the call, or a special value -1 in case when an error occurred.

• Errors:

  • SRT_EINVPOLLID: eid parameter doesn't refer to a valid epoll container

srt_epoll_release

int srt_epoll_release(int eid);

Deletes the epoll container.

• Returns:

  • The number (>0) of ready sockets, of whatever kind (if any)
  • -1 in case of error

• Errors:

  • SRT_EINVPOLLID: eid parameter doesn't refer to a valid epoll container

Logging control

SRT has a widely used system of logs, as this is usually the only way to determine how the internals are working, without changing the rules by the act of tracing. Logs are split into levels (5 levels out of those defined by syslog are in use) and additional filtering is possible on FA (functional area). By default only up to the Note log level are displayed and from all FAs.

Logging can only be manipulated globally, with no regard to a specific socket. This is because lots of operations in SRT are not dedicated to any particular socket, and some are shared between sockets.

srt_setloglevel

void srt_setloglevel(int ll);

Sets the minimum severity for logging. A particular log entry is displayed only if it has a severity greater than or equal to the minimum. Setting this value to LOG_DEBUG turns on all levels.

The constants for this value are those from <sys/syslog.h> (for Windows, refer to common/win/syslog_defs.h). The only meaningful are:

• LOG_DEBUG: Highly detailed and very frequent messages
• LOG_NOTICE: Occasionally displayed information
• LOG_WARNING: Unusual behavior
• LOG_ERR: Abnormal behavior
• LOG_CRIT: Error that makes the current socket unusable

srt_addlogfa, srt_dellogfa, srt_resetlogfa

void srt_addlogfa(int fa);
void srt_dellogfa(int fa);
void srt_resetlogfa(const int* fara, size_t fara_size);

A functional area (FA) is an additional filtering mechanism for logging. You can set up logging to display logs only from selected FAs. The list of FAs is collected in srt.h file, as identified by the SRT_LOGFA_ prefix. They are not enumerated here because they may be changed very often.

All FAs are turned on by default, except potentially dangerous ones (such as SRT_LOGFA_HAICRYPT). The reaons is that they may display either some security information that shall remain in the memory only (so, only if strictly required for the development), or some duplicated information (so you may want to turn this FA on, while turning off the others).

srt_setloghandler

void srt_setloghandler(void* opaque, SRT_LOG_HANDLER_FN* handler);
typedef void SRT_LOG_HANDLER_FN(void* opaque, int level, const char* file, int line, const char* area, const char* message);

By default logs are printed to standard error stream. This function replaces the sending to a stream with a handler function that will receive them.

srt_setlogflags

void srt_setlogflags(int flags);

When you set a log handler with srt_setloghandler, you may also want to configure which parts of the log information you do not wish to be passed in the log line (the message parameter). A user's logging facility may, for example, not wish to get the current time or log level marker, as it will provide this information on its own.

The following flags are available, as collected in logging_api.h public header:

• SRT_LOGF_DISABLE_TIME: Do not provide the time in the header
• SRT_LOGF_DISABLE_THREADNAME: Do not provide the thread name in the header
• SRT_LOGF_DISABLE_SEVERITY: Do not provide severity information in the header
• SRT_LOGF_DISABLE_EOL: Do not add the end-of-line character to the log line

Time Access

The following set of functions is intended to retrieve timestamps from the clock used by SRT. The sender can pass the timestamp in MSGCTRL::srctime of the srt_sendmsg2(..) function together with the packet being submitted to SRT. If the srctime value is not provided (the default value of 0 is set), SRT will use internal clock and assign the packet submission time as the packet timestamp. If the sender wants to explicitly assign a timestamp to a certain packet. this timestamp MUST be taken from SRT Time Access functions. The time value provided MUST equal or exceed the connection start time (srt_connection_time(..)) of the SRT socket passed to srt_sendmsg2(..).

The current time value as of the SRT internal clock can be retrieved using the srt_time_now() function.

There are two known cases where you might want to use srctime:

1. SRT passthrough (for stream gateways). You may wish to simply retrieve packets from an SRT source and pass them transparently to an SRT output (possibly re-encrypting). In that case, every packet you read should preserve the original value of srctime as obtained from srt_recvmsg2, and the original srctime for each packet should be then passed to srt_sendmsg2. This mechanism could be used to avoid jitter resulting from varying differences between the time of receiving and sending the same packet.

2. Stable timing source. In the case of a live streaming procedure, when spreading packets evenly into the stream, you might want to predefine times for every single packet to keep time intervals perfectly equal. Or, if you believe that your input signal delivers packets at the exact times that should be assigned to them, you might want to preserve these times at the SRT receiving side to avoid jitter that may result from varying time differences between the packet arrival and the moment when sending it over SRT. In such cases you might do the following:

   • At the packet arrival time, grab the current time at that moment using srt_time_now().

   • When you want a precalculated packet time, use a private relative time counter set at the moment when the connection was made. From the moment when your first packet is ready, start precalculating packet times relative to the connection start time obtained from srt_connection_time(). Although you still have to synchronize sending times with these predefined times, by explicitly specifying the source time you avoid the jitter resulting from a lost accuracy due to waiting time and unfortunate thread scheduling.

Note that srctime uses an internally defined clock that is intended to be monotonic (the definition depends on the build flags, see below). Because of that the application should not define this time basing on values obtained from the system functions for getting the current system time (such as time, ftime or gettimeofday). To avoid problems and misunderstanding you should rely exclusively on time values provided by srt_time_now() and srt_connection_time() functions.

The clock used by SRT internal clock, is determined by the following build flags:

• ENABLE_MONOTONIC makes use of CLOCK_MONOTONIC with clock_gettime function.
• ENABLE_STDXXX_SYNC makes use of std::chrono::steady_clock.

The default is currently to use the system clock as internal SRT clock, although it's highly recommended to use one of the above monotonic clocks, as system clock is vulnerable to time modifications during transmission.

srt_time_now

int64_t srt_time_now();

Get time in microseconds elapsed since epoch using SRT internal clock (steady or monotonic clock).

• Returns:
  • Current time in microseconds elapsed since epoch of SRT internal clock.

srt_connection_time

int64_t srt_connection_time(SRTSOCKET sock);

Get connection time in microseconds elapsed since epoch using SRT internal clock (steady or monotonic clock). The connection time represents the time when SRT socket was open to establish a connection. Milliseconds elapsed since connection start time can be determined using Performance tracking functions and msTimeStamp value of the SRT_TRACEBSTATS (see statistics.md).

• Returns:

  • Connection time in microseconds elapsed since epoch of SRT internal clock.
  • -1 in case of error

• Errors:

  • SRT_EINVSOCK: Socket sock is not an ID of a valid SRT socket

RETURN TO TOP OF PAGE