rpc_clnt_create(3N) rpc_clnt_create(3N)
NAME
rpc_clnt_create: clnt_control(), clnt_create(), clnt_create_timed(), clnt_create_vers(), clnt_create_vers_timed(), clnt_destroy(),
clnt_dg_create(), clnt_pcreateerror(), clnt_raw_create(), clnt_spcreateerror(), clnt_tli_create(), clnt_tp_create(), clnt_tp_cre-
ate_timed(), clnt_vc_create(), rpc_createerr - library routines for dealing with creation and manipulation of CLIENT handles
SYNOPSIS
DESCRIPTION
RPC library routines allow C language programs to make procedure calls on other machines across the network. First a handle is created and
then the client calls a procedure to send a request to the server. On receipt of the request, the server calls a dispatch routine to per-
form the requested service, and then sends a reply.
These routines are thread-safe. In the case of multithreaded applications, the flag must be defined on the command line at compilation
time When the flag is defined, becomes a macro which enables each thread to have its own
The HP-UX implementation of RPC only supports the X/Open Transport Interface (XTI). Applications that are written using the Transport
Layer Interface (TLI) and wish to use RPC, must convert their application to XTI.
Routines
See rpc(3N) for the definition of the data structure.
A function macro to change or retrieve various information
about a client object. req indicates the type of operation, and info is a pointer to the information. For both connectionless and
connection-oriented transports, the supported values of req and their argument types and what they do are:
CLSET_TIMEOUT struct timeval * set total timeout
CLGET_TIMEOUT struct timeval * get total timeout
If the timeout is set using the timeout argument passed by is ignored in all subsequent calls.
If the timeout value is set to immediately returns Set the timeout parameter to for batching calls.
CLGET_SERVER_ADDR struct netbuf * get server's address
CLGET_SVC_ADDR struct netbuf * get server's address
CLGET_FD int * get the associated file descriptor
CLSET_FD_CLOSE void * close the file descriptor
when destroying the client handle
(see clnt_destroy())
CLSET_FD_NCLOSE void do not close the file descriptor
when destroying the client handle
CLGET_VERS rpcvers_t get the RPC program's version
number associated with the
client handle
CLSET_VERS rpcvers_t set the RPC program's version
number associated with the
client handle. This assumes that
the RPC server for this new
version is still listening at the
address of the previous version.
CLGET_XID unsigned long * get the XID of the previous
remote procedure call
CLSET_XID unsigned long * set the XID of the next
remote procedure call
CLGET_PROG rpcprog_t get program number
CLSET_PROG rpcprog_t set program number
The following operations are valid for connection-oriented transports only:
CLSET_IO_MODE rpciomode_t * set the IO mode used
to send one-way requests.
The argument for this operation can be either:
All sending operations block until the underlying transport protocol
has accepted requests. If you specify this argument you cannot use flush and getting and setting buffer size is meaningless.
Sending operations do not block and return as soon as requests enter
the buffer. You can now use non-blocking I/O. The requests in the buffer are pending. The requests are sent to the server
as soon as a two-way request is sent or a flush is done. You are responsible for flushing the buffer. When you choose the
argument you have a choice of flush modes as specified by
CLGET_IO_MODE rpciomode_t * get the current IO mode
CLSET_FLUSH_MODE rpcflushmode_t * set the flush mode
The flush mode can only be used in non-blocking I/O mode. The argument for this operation can be either:
All flushes send requests in the buffer until the transport end-point
blocks. If the transport connection is congested, the call returns directly.
Flush blocks until the underlying transport protocol accepts all
pending requests into the queue.
CLGET_FLUSH_MODE rpcflushmode_t * get the current flush mode
CLFLUSH rpcflushmode_t flush the pending requests
This command can only be used in non-blocking I/O mode. The flush policy depends on which of the following parameters is specified:
The flush is done according to the current flush mode policy
(see the option).
The flush tries to send pending requests without blocking; the call
returns directly. If the transport connection is congested, this call could return without the request being sent.
The flush sends all pending requests.
This call will block until all the requests have been accepted by the transport layer.
CLSET_CONNMAXREC_SIZE int * set the buffer size
CLGET_CONNMAXREC_SIZE int * get the current size of the
buffer
CLGET_CURRENT_REC_SIZE int * get the size of the pending
requests stored in the
buffer. Use of this command
is only recommended when you
are in non-blocking I/O
mode. The current size of
the buffer is always zero
when the handle is in
blocking mode as the buffer
is not used in this mode.
The following operations are valid for connectionless transports only:
CLSET_RETRY_TIMEOUT struct timeval * set the retry timeout
CLGET_RETRY_TIMEOUT struct timeval * get the retry timeout
The retry timeout is the time that RPC waits for the server to reply before retransmitting the request.
returns on success and on failure.
Generic client creation routine for program
prognum and version versnum. host identifies the name of the remote host where the server is located. nettype indicates the class
of transport protocol to use. The transports are tried in left to right order in variable or in top to bottom order in the netcon-
fig database.
tries all the transports of the nettype class available from the environment variable and the netconfig database, and chooses the
first successful one. A default timeout is set and can be modified using This routine returns NULL if it fails. The routine can be
used to print the reason for failure.
Note that returns a valid client handle even if the particular version number supplied to is not registered with the service. This
mismatch will be discovered by a later (see rpc_clnt_calls(3N)).
Generic client creation routine which is similar to
but with the additional parameter timeout, which specifies the maximum amount of time allowed for each transport class tried. In
all other respects, the call behaves exactly like the call.
Generic client creation routine which is similar to
but which also checks for the version availability. host identifies the name of the remote host where the server is located. net-
type indicates the class transport protocols to be used. If the routine is successful it returns a client handle created for the
highest version between vers_low and vers_high that is supported by the server. vers_outp is set to this value. That is, after a
successful return vers_low <= *vers_outp <= vers_high. If no version between vers_low and vers_high is supported by the server then
the routine fails and returns NULL. A default timeout is set and can be modified using This routine returns NULL if it fails. The
routine can be used to print the reason for failure.
Note: returns a valid client handle even if the particular version number supplied to is not registered with the service. This mis-
match will be discovered by a later (see rpc_clnt_calls(3N)). However, does this for you and returns a valid handle only if a ver-
sion within the range supplied is supported by the server.
Generic client creation routine which is similar to
but with the additional parameter timeout, which specifies the maximum amount of time allowed for each transport class tried. In
all other respects, the call behaves exactly like the call.
A function macro that destroys the client's RPC handle.
Destruction usually involves deallocation of private data structures, including clnt itself. Use of clnt is undefined after calling
If the RPC library opened the associated file descriptor, or was set using the file descriptor will be closed.
The caller should call (before calling to destroy the associated AUTH structure (see rpc_clnt_auth(3N)).
This routine creates an RPC client for the remote program
prognum and version versnum; the client uses a connectionless transport. The remote program is located at address svcaddr. The
parameter fildes is an open and bound file descriptor. This routine will resend the call message in intervals of 15 seconds until a
response is received or until the call times out. The total time for the call to time out is specified by (see in
rpc_clnt_calls(3N)). The retry timeout and the total timeout periods can be changed using The user may set the size of the send and
receive buffers with the parameters sendsz and recvsz; values of choose suitable defaults. This routine returns NULL if it fails.
Print a message to standard error indicating why a client
RPC handle could not be created. The message is prepended with the string s and a colon, and appended with a newline.
This routine creates an RPC client handle for the remote program
prognum and version versnum. The transport used to pass messages to the service is a buffer within the process' address space, so
the corresponding RPC server should live in the same address space (see in rpc_svc_create(3N)). This allows simulation of RPC and
measurement of RPC overheads, such as round trip times, without any kernel or networking interference. This routine returns NULL if
it fails. should be called after
Like except that it returns a string instead of printing to the standard error. A newline is not appended to the message in this case.
Warning: returns a pointer to a buffer that is overwritten on each call. In multithreaded applications, this buffer is implemented
as thread-specific data.
This routine creates an RPC client handle for the remote program
prognum and version versnum. The remote program is located at address svcaddr. If svcaddr is NULL and it is connection-oriented,
it is assumed that the file descriptor is connected. For connectionless transports, if svcaddr is NULL, error is set. fildes is a
file descriptor which may be open, bound and connected. If it is it opens a file descriptor on the transport specified by netconf.
If fildes is and netconf is NULL, a error is set. If fildes is unbound, then it will attempt to bind the descriptor. The user may
specify the size of the buffers with the parameters sendsz and recvsz; values of choose suitable defaults. Depending upon the type
of the transport (connection-oriented or connectionless), calls appropriate client creation routines. This routine returns NULL if
it fails. The routine can be used to print the reason for failure. The remote service (see rpcbind(1M)) is not consulted for the
address of the remote service.
Like except tries only one transport specified through netconf.
creates a client handle for the program prognum, the version versnum, and for the transport specified by netconf. Default options
are set, which can be changed using calls. The remote service on the host host is consulted for the address of the remote service.
This routine returns NULL if it fails. The routine can be used to print the reason for failure.
Like except has the extra parameter timeout which specifies the maximum time allowed for the creation attempt to succeed. In all other
respects, the call behaves exactly like the call.
This routine creates an RPC client for the remote program
prognum and version versnum; the client uses a connection-oriented transport. The remote program is located at address svcaddr.
The parameter fildes is an open and bound file descriptor. The user may specify the size of the send and receive buffers with the
parameters sendsz and recvsz; values of choose suitable defaults. This routine returns NULL if it fails.
The address svcaddr should not be NULL and should point to the actual address of the remote program. does not consult the remote
service for this information.
A global variable whose value is set by any RPC
client handle creation routine that fails. It is used by the routine to print the reason for the failure.
In multithreaded applications, becomes a macro which enables each thread to have its own
MULTITHREAD USAGE
Thread Safe: Yes
Cancel Safe: Yes
Fork Safe: No
Async-cancel Safe: No
Async-signal Safe: No
These functions can be called safely in a multithreaded environment. They may be cancellation points in that they call functions that are
cancel points.
In a multithreaded environment, these functions are not safe to be called by a child process after and before These functions should not be
called by a multithreaded application that supports asynchronous cancellation or asynchronous signals.
SEE ALSO
rpc(3N), rpc_clnt_auth(3N), rpc_clnt_calls(3N), rpcbind(1M).
rpc_clnt_create(3N)