.OH 'RPC Programming''Page \\\\n(PN'
.EH 'Page \\\\n(PN''RPC Programming'
.OF 'Sun Microsystems''Release 2.0'
.EF 'Release 2.0''Sun Microsystems'
.RP
.rm DY
.TL
.ps 20
Remote Procedure Call
.sp.5
Programming Guide
.
.H 1 "Introduction"
.LP
Programs that communicate over a network
need a paradigm for communication.
A low-level mechanism might 
send a signal on the arrival of incoming packets,
causing a network signal handler to execute.
A high-level mechanism would be the Ada 
.L rendezvous .
The method used at Sun is the
Remote Procedure Call (RPC) paradigm,
in which a client communicates with a server.
In this process,
the client first calls a procedure to send a data packet to the server.
When the packet arrives, the server calls a dispatch routine,  
performs whatever service is requested, sends back the reply,
and the procedure call returns to the client.
.LP
The RPC interface is divided into three layers.
The highest layer is totally transparent to the programmer.
To illustrate,
at this level a program can contain a call to
.L rnusers() ,
which returns the number of users on a remote machine.
You don't have to be aware that RPC is being used,
since you simply make the call in a program,
just as you would call
.L malloc() .
.LP
At the middle layer, the routines
.L registerrpc()
and
.L callrpc()
are used to make RPC calls:
.L registerrpc()
obtains a unique system-wide number, while
.L callrpc()
executes a remote procedure call.
The 
.L rnusers()
call is implemented using these two routines
The middle-layer routines are designed for most common applications,
and shield the user from knowing about sockets.
.LP
The lowest layer is used for more sophisticated applications,
which may want to alter the defaults of the routines.
At this layer, you can explicitly manipulate
sockets used for transmitting RPC messages.
This level should be avoided if possible.
.LP
Section 2 of this manual illustrates use of the highest two layers
while Section 3 presents the low-level interface.
Section 4 of the manual discusses miscellaneous topics.
The final section summarizes 
all the entry points into the RPC system.
.LP
Although this document only discusses the interface to C,
remote procedure calls can be made from any language.
Even though this document discusses RPC
when it is used to communicate
between processes on different machines,
it works just as well for communication
between different processes on the same machine.
.bp
.
.H 1 "Introductory Examples"
.H 2 "Highest Layer"
.LP
Imagine you're writing a program that needs to know
how many users are logged into a remote machine.
You can do this by calling the library routine
.L rnusers() ,
as illustrated below:
.LS
#include <stdio.h>
.sp.5
main(argc, argv)
	int argc;
	char **argv;
{
	unsigned num;
.sp.5
	if (argc < 2) {
		fprintf(stderr, "usage: rnusers hostname\en");
		exit(1);
	}
	if ((num = rnusers(argv[1])) < 0) {
		fprintf(stderr, "error: rnusers\en");
		exit(-1);
	}
	printf("%d users on %s\en", num, argv[1]);
	exit(0);
}
.LE
RPC library routines such as
.L rnusers()
are included in the C library
.L libc.a .
Thus, the program above could be compiled with
.LS
% cc \fIprogram.c\fP
.LE
Some other library routines are
.L rstat()
to gather remote performance statistics, and
.L ypmatch()
to glean information from the yellow pages (YP).
The YP library routines are documented on the manual page
.I ypclnt (3N).
.bp
.
.H 2 "Intermediate Layer"
.LP
The simplest interface, which explicitly makes RPC
calls, uses the functions
.L callrpc()
and
.L registerrpc() .
Using this method, another way to get the number of remote users is:
.LS
#include <stdio.h>
#include <rpcsvc/rusers.h>
.sp.5
main(argc, argv)
	int argc;
	char **argv;
{
	unsigned long nusers;
.sp.5
	if (argc < 2) {
		fprintf(stderr, "usage: nusers hostname\en");
		exit(-1);
	}
	if (callrpc(argv[1], RUSERSPROG, RUSERSVERS, RUSERSPROC_NUM,
	    xdr_void, 0, xdr_u_long, &nusers) != 0) {
		fprintf(stderr, "error: callrpc\en");
		exit(1);
	}
	printf("number of users on %s is %d\en", argv[1], nusers);
	exit(0);
}
.LE
A program number, version number, and procedure number
defines each RPC procedure.
The program number defines a group
of related remote procedures, each of which has a different
procedure number.
Each program also has a version number,
so when a minor change is made to a remote service
(adding a new procedure, for example),
a new program number doesn't have to be assigned.
When you want to call a procedure to
find the number of remote users, you look up the appropriate
program, version and procedure numbers
in a manual, similar to when you look up the name of memory
allocator when you want to allocate memory.
.LP
The simplest routine in the RPC library
used to make remote procedure calls is
.L callrpc() .
It has eight parameters.
The first is the name of the remote machine.
The next three parameters
are the program, version, and procedure numbers.
The following two parameters
define the argument of the RPC call, and the final two parameters
are for the return value of the call. 
If it completes successfully,
.L callrpc()
returns zero, but nonzero otherwise.
The exact meaning of the return codes is found in
.L <rpc/clnt.h> ,
and is in fact an
.L "enum clnt_stat"
cast into an integer.
.LP
Since data types may be represented differently on different machines,
.L callrpc()
needs both the type of the RPC argument, as well as
a pointer to the argument itself (and similarly for the result).
For RUSERSPROC_NUM, the return value is an
.L "unsigned long" ,
so 
.L callrpc()
has
.L xdr_u_long
as its first return parameter, which says 
that the result is of type
.L "unsigned long",
and
.L &nusers 
as its second return parameter,
which is a pointer to where the long result will be placed.
Since RUSERSPROC_NUM takes no argument, the argument parameter of
.L callrpc()
is
.L xdr_void .
.LP
After trying several times to deliver a message, if
.L callrpc()
gets no answer, it returns with an error code.
The delivery mechanism is UDP,
which stands for User Datagram Protocol.
Methods for adjusting the number of retries
or for using a different protocol require you to use the lower
layer of the RPC library, discussed later in this document.
The remote server procedure
corresponding to the above might look like this:
.LS
char *
nuser(indata)
	char *indata;
{
	static int nusers;
.sp.5
	/*
	 * code here to compute the number of users
	 * and place result in variable nusers
	 */
	return ((char *)&nusers);
}
.LE
.LP
It takes one argument, which is a pointer to the input
of the remote procedure call (ignored in our example),
and it returns a pointer to the result.
In the current version of C,
character pointers are the generic pointers,
so both the input argument and the return value are cast to
.L "char *" .
.LP
Normally, a server registers all of the RPC calls it plans
to handle, and then goes into an infinite loop waiting to service requests.
In this example, there is only a single procedure
to register, so the main body of the server would look like this:
.LS
#include <stdio.h>
#include <rpcsvc/rusers.h>
.sp.5
char *nuser();
.sp.5
main()
{
	registerrpc(RUSERSPROG, RUSERSVERS, RUSERSPROC_NUM, nuser,
	    xdr_void, xdr_u_long);
	svc_run();		/* never returns */
	fprintf(stderr, "Error: svc_run returned!\en");
	exit(1);
}
.LE
.LP
The
.L registerrpc()
routine establishes what C procedure
corresponds to each RPC procedure number.
The first three parameters,
RUSERPROG, RUSERSVERS, and RUSERSPROC_NUM
are the program, version, and procedure numbers
of the remote procedure to be registered;
.L nuser
is the name of the C procedure implementing it;
and 
.L xdr_void
and
.L xdr_u_long
are the types of the input to and output from the procedure.
.LP
Only the UDP transport mechanism can use
.L registerrpc() ;
thus, it is always safe in conjunction with calls generated by 
.L callrpc() .
.LP
Warning: the UDP transport mechanism can only deal with
arguments and results less than 8K bytes in length.
.
.H 2 "Assigning Program Numbers"
.LP
Program numbers are assigned in groups of 0x20000000 (536870912)
according to the following chart:
.LS
       0 - 1fffffff	defined by sun
20000000 - 3fffffff	defined by user
40000000 - 5fffffff	transient
60000000 - 7fffffff	reserved
80000000 - 9fffffff	reserved
a0000000 - bfffffff	reserved
c0000000 - dfffffff	reserved
e0000000 - ffffffff	reserved
.LE
Sun Microsystems administers the first group of numbers,
which should be identical for all Sun customers.
If a customer develops an application that might be of general
interest, that application should be given an assigned number
in the first range.
The second group of numbers is reserved for specific customer
applications.
This range is intended primarily for debugging new programs.
The third group is reserved for applications that
generate program numbers dynamically.
The final groups
are reserved for future use, and should not be used.
.LP
The exact registration process for Sun defined numbers is yet
to be established.
.
.H 2 "Passing Arbitrary Data Types"
.LP
In the previous example, the RPC call passes a single
.L "unsigned long."
RPC can handle arbitrary data structures, regardless of
different machines' byte orders or structure layout conventions,
by always converting them to a network standard called 
.I "eXternal Data Representation"
(XDR) before
sending them over the wire.
The process of converting from a particular machine representation
to XDR format is called
.I serializing ,
and the reverse process is called
.I deserializing .
The type field parameters of 
.L callrpc()
and
.L registerrpc()
can be a built-in procedure like
.L xdr_u_long()
in the previous example, or a user supplied one.
XDR has these built-in type routines:
.LS
xdr_int()      xdr_u_int()      xdr_enum()
xdr_long()     xdr_u_long()     xdr_bool()  
xdr_short()    xdr_u_short()    xdr_string()
.LE
As an example of a user-defined type routine,
if you wanted to send the structure
.LS
struct simple {
	int a;
	short b;
} simple;
.LE
then you would call
.L callrpc
as
.LS
callrpc(hostname, PROGNUM, VERSNUM, PROCNUM, xdr_simple, &simple ...);
.LE
where
.L xdr_simple()
is written as:
.LS
#include <rpc/rpc.h>
.sp.5
xdr_simple(xdrsp, simplep)
	XDR *xdrsp;
	struct simple *simplep;
{
	if (!xdr_int(xdrsp, &simplep->a))
		return (0);
	if (!xdr_short(xdrsp, &simplep->b))
		return (0);
	return (1);
}
.LE
.LP
An XDR routine returns nonzero (true in the sense of C)
if it completes successfully, and zero otherwise.
A complete description of XDR is in the 
.I "XDR Protocol Specification" ,
so this section only gives a few examples of XDR implementation.
.LP
In addition to the built-in primitives,
there are also the prefabricated building blocks:
.LS
xdr_array()       xdr_bytes()
xdr_reference()   xdr_union()
.LE
To send a variable array of integers,
you might package them up as a structure like this
.LS
struct varintarr {
	int *data;
	int arrlnth;
} arr;
.LE
and make an RPC call such as
.LS
callrpc(hostname, PROGNUM, VERSNUM, PROCNUM, xdr_varintarr, &arr...);
.LE
with
.L xdr_varintarr()
defined as:
.LS
xdr_varintarr(xdrsp, varintarr)
	XDR *xdrsp;
	struct varintarr *arrp;
{
	xdr_array(xdrsp, &arrp->data, &arrp->arrlnth, MAXLEN,
		sizeof(int), xdr_int);
}
.LE
This routine takes as parameters the XDR handle,
a pointer to the array, a pointer to the size of the array,
the maximum allowable array size,
the size of each array element,
and an XDR routine for handling each array element.
.LP
If the size of the array is known in advance, then
the following could also be used to send 
out an array of length SIZE:
.LS
int intarr[SIZE];
.sp.5
xdr_intarr(xdrsp, intarr)
	XDR *xdrsp;
	int intarr[];
{
	int i;
.sp.5
	for (i = 0; i < SIZE; i++) {
		if (!xdr_int(xdrsp, &intarr[i]))
			return (0);
	}
	return (1);
}
.LE
.LP
XDR always converts quantities to 4-byte multiples when deserializing.
Thus, if either of the examples above involved characters
instead of integers, each character would occupy 32 bits.
That is the reason for the XDR routine
.L xdr_bytes() ,
which is like
.L xdr_array()
except that it packs characters.
It has four parameters,
the same as the first four parameters of
.L xdr_array() .
For null-terminated strings, there is also the
.L xdr_string()
routine, which is the same as
.L xdr_bytes()
without the length parameter.
On serializing it gets the string length from 
.L strlen() ,
and on deserializing it creates a null-terminated string.
.LP
Here is a final example that calls the previously written
.L xdr_simple()
as well as the built-in functions
.L xdr_string()
and
.L xdr_reference() ,
which chases pointers:
.LS
struct finalexample {
	char *string;
	struct simple *simplep;
} finalexample;
.LE
.LS
xdr_finalexample(xdrsp, finalp)
	XDR *xdrsp;
	struct finalexample *finalp;
{
	int i;
.sp.5
	if (!xdr_string(xdrsp, &finalp->string, MAXSTRLEN))
		return (0);
	if (!xdr_reference(xdrsp, &finalp->simplep,
	    sizeof(struct simple), xdr_simple);
		return (0);
	return (1);
}
.LE
.bp
.
.H 1 "Lower Layers of RPC"
.LP
In the examples given so far,
RPC takes care of many details automatically for you.
In this section, we'll show you how you can change the defaults
by using lower layers of the RPC library.
It is assumed that you are familiar with sockets
and the system calls for dealing with them.
If not, consult
.I "The IPC Tutorial" .
.H 2 "More on the Server Side"
.LP
There are a number of assumptions built into 
.L registerrpc() .
One is that you are using the UDP datagram protocol.
Another is that
you don't want to do anything unusual while deserializing,
since the deserialization process happens automatically
before the user's server routine is called.
The server for the
.L nusers
program shown below
is written using a lower layer of the RPC package,
which does not make these assumptions.
.LS
#include <stdio.h>
#include <rpc/rpc.h>
#include <rpcsvc/rusers.h>
.sp.5
int nuser();
.sp.5
main()
{
	SVCXPRT *transp;
.sp.5
	transp = svcudp_create(RPC_ANYSOCK);
	if (transp == NULL){
		fprintf(stderr, "couldn't create an RPC server\en");
		exit(1);
	}
	pmap_unset(RUSERSPROG, RUSERSVERS);
	if (!svc_register(transp, RUSERSPROG, RUSERSVERS, nuser,
	    IPPROTO_UDP)) {
		fprintf(stderr, "couldn't register RUSER service\en");
		exit(1);
	}
	svc_run();  /* never returns */
	fprintf(stderr, "should never reach this point\en");
}
.LE
.LS
nuser(rqstp, tranp)
	struct svc_req *rqstp;
	SVCXPRT *transp;
{
	unsigned long nusers;
.sp.5
	switch (rqstp->rq_proc) {
	case NULLPROC:
		if (!svc_sendreply(transp, xdr_void, 0)) {
			fprintf(stderr, "couldn't reply to RPC call\en");
			exit(1);
		}
		return;
	case RUSERSPROC_NUM:
		/*
		 * code here to compute the number of users
		 * and put in variable nusers
		 */
		if (!svc_sendreply(transp, xdr_u_long, &nusers) {
			fprintf(stderr, "couldn't reply to RPC call\en");
			exit(1);
		}
		return;
	default: 
		svcerr_noproc(transp);
		return;
	}
}
.LE
.LP
First, the server gets a transport handle, which is used
for sending out RPC messages.
.L registerrpc()
uses
.L svcudp_create()
to get a UDP handle.
If you require a reliable protocol, call
.L svctcp_create()
instead.
If the argument to
.L svcudp_create()
is RPC_ANYSOCK,
the RPC library creates a socket
on which to send out RPC calls.
Otherwise, 
.L svcudp_create()
expects its argument to be a valid socket number.
If you specify your own socket, it can be bound or unbound.
If it is bound to a port by the user, the port numbers of
.L svcudp_create()
and
.L clntudp_create()
(the low-level client routine) must match.
.LP
When the user specifies RPC_ANYSOCK for a socket 
or gives an unbound socket, the system determines
port numbers in the following way:
when a server starts up,
it advertises to a port mapper demon on its local machine,
which picks a port number for the RPC procedure
if the socket specified to
.L svcudp_create()
isn't already bound.
When the
.L clntudp_create()
call is made with an unbound socket,
the system queries the port mapper on
the machine to which the call is being made,
and gets the appropriate port number.
If the port mapper is not running
or has no port corresponding to the RPC call,
the RPC call fails.
Users can make RPC calls
to the port mapper themselves.
The appropriate procedure
numbers are in the include file
.L <rpc/pmap_prot.h> .
.LP
After creating an SVCXPRT, the next step is to call
.L pmap_unset()
so that if the
.L nusers
server crashed earlier,
any previous trace of it is erased before restarting.
More precisely,
.L pmap_unset()
erases the entry for RUSERS from the port mapper's tables.
.LP
Finally, we associate the program number for
.L nusers
with the procedure 
.L nuser() .
The final argument to
.L svc_register()
is normally the protocol being used,
which, in this case, is IPPROTO_UDP.
Notice that unlike
.L registerrpc() ,
there are no XDR routines involved
in the registration process.
Also, registration is done on the program,
rather than procedure, level.
.LP
The user routine 
.L nuser()
must call and dispatch the appropriate XDR routines
based on the procedure number.
Note that 
two things are handled by 
.L nuser()
that
.L registerrpc()
handles automatically. 
The first is that procedure NULLPROC
(currently zero) returns with no arguments.
This can be used as a simple test
for detecting if a remote program is running.
Second, there is a check for invalid procedure numbers.
If one is detected,
.L svcerr_noproc()
is called to handle the error.
.LP
The user service routine serializes the results and returns
them to the RPC caller via 
.L svc_sendreply() .
Its first parameter is the SVCXPRT handle,
the second is the XDR routine,
and the third is a pointer to the data to be returned.
Not illustrated above is how a server
handles an RPC program that passes data.
As an example, we can add a procedure RUSERSPROC_BOOL,
which has an argument
.L nusers ,
and returns TRUE or FALSE depending on
whether there are nusers logged on.
It would look like this:
.LS
case RUSERSPROC_BOOL: {
	int bool;
	unsigned nuserquery;
.sp.5
	if (!svc_getargs(transp, xdr_u_int, &nuserquery) {
		svcerr_decode(transp);
		return;
	}
	/*
	 * code to set nusers = number of users
	 */
	if (nuserquery == nusers)
		bool = TRUE;
	else
		bool = FALSE;
	if (!svc_sendreply(transp, xdr_bool, &bool){
		 fprintf(stderr, "couldn't reply to RPC call\en");
		 exit(1);
	}
	return;
}
.LE
.LP
The relevant routine is
.L svc_getargs() ,
which takes an SVCXPRT handle, the XDR routine,
and a pointer to where the input is to be placed as arguments.
.H 2 "Memory Allocation with XDR"
.LP
XDR routines not only do input and output,
they also do memory allocation.
This is why the second parameter of
.L xdr_array()
is a pointer to an array, rather than the array itself.
If it is NULL, then
.L xdr_array()
allocates space for the array and returns a pointer to it, 
putting the size of the array in the third argument.
As an example, consider the following XDR routine
.L xdr_chararr1() ,
which deals with a fixed array of bytes with length SIZE:
.LS
xdr_chararr1(xdrsp, chararr)
	XDR *xdrsp;
	char chararr[];
{
	char *p;
	int len;
.sp.5
	p = chararr;
	len = SIZE;
	return (xdr_bytes(xdrsp, &p, &len, SIZE));
}
.LE
It might be called from a server like this,
.LS
char chararr[SIZE];
.sp.5
svc_getargs(transp, xdr_chararr1, chararr);
.LE
where
.L chararr
has already allocated space.
If you want XDR to do the allocation,
you would have to rewrite this routine in the following way:
.LS
xdr_chararr2(xdrsp, chararrp)
	XDR *xdrsp;
	char **chararrp;
{
	int len;
.sp.5
	len = SIZE;
	return (xdr_bytes(xdrsp, charrarrp, &len, SIZE));
}
.LE
Then the RPC call might look like this:
.LS
char *arrptr;
.sp.5
arrptr = NULL;
svc_getargs(transp, xdr_chararr2, &arrptr);
/* 
 * use the result here
 */
svc_freeargs(xdrsp, xdr_chararr2, &arrptr);
.LE
After using the character array, it can be freed with
.L svc_freeargs() .
In the routine
.L xdr_finalexample()
given earlier, if
.L finalp->string
was NULL in the call
.LS
svc_getargs(transp, xdr_finalexample, &finalp);
.LE
then
.LS
svc_freeargs(xdrsp, xdr_finalexample, &finalp);
.LE
frees the array allocated to hold
.L finalp->string ;
otherwise, it frees nothing.
The same is true for
.L finalp->simplep .
.LP
To summarize, each XDR routine is responsible
for serializing, deserializing, and allocating memory.
When an XDR routine is called from
.L callrpc() ,
the serializing part is used.
When called from
.L svc_getargs() ,
the deserializer is used.
And when called from
.L svc_freeargs() ,
the memory deallocator is used.
When building simple examples like those in this section,
a user doesn't have to worry about the three modes.
The XDR reference manual has examples of more
sophisticated XDR routines that 
determine which of the three modes they are in
to function correctly.
.
.H 2 "The Calling Side"
.LP
When you use
.L callrpc,
you have no control over the RPC delivery
mechanism or the socket used to transport the data.
To illustrate the layer of RPC that lets you adjust these
parameters, consider the following code to call the
.L nusers
service:
.LS
#include <stdio.h>
#include <rpc/rpc.h>
#include <rpcsvc/rusers.h>
#include <sys/socket.h>
#include <sys/time.h>
#include <netdb.h>
.sp.5
main(argc, argv)
	int argc;
	char **argv;
{
	struct hostent *hp;
	struct timeval pertry_timeout, total_timeout;
	struct sockaddr_in server_addr;
	int addrlen, sock = RPC_ANYSOCK;
	register CLIENT *client;
	enum clnt_stat clnt_stat;
	unsigned long nusers;
.sp.5
	if (argc < 2) {
		fprintf(stderr, "usage: nusers hostname\en");
		exit(-1);
	}
	if ((hp = gethostbyname(argv[1])) == NULL) {
		fprintf(stderr, "cannot get addr for '%s'\en", argv[1]);
		exit(-1);
	}
	pertry_timeout.tv_sec = 3;
	pertry_timeout.tv_usec = 0;
	addrlen = sizeof(struct sockaddr_in);
	bcopy(hp->h_addr, (caddr_t)&server_addr.sin_addr, hp->h_length);
	server_addr.sin_family = AF_INET;
	server_addr.sin_port =  0;
	if ((client = clntudp_create(&server_addr, RUSERSPROG,
	    RUSERSVERS, pertry_timeout, &sock)) == NULL) {
		perror("clntudp_create");
		exit(-1);
	}
	total_timeout.tv_sec = 20;
	total_timeout.tv_usec = 0;
	clnt_stat = clnt_call(client, RUSERSPROC_NUM, xdr_void, 0,
	    xdr_u_long, &nusers, total_timeout);
	if (clnt_stat != RPC_SUCCESS) {
		clnt_perror(client, "rpc");
		exit(-1);
	}
	clnt_destroy(client);
}
.LE
The low-level version of 
.L callrpc()
is
.L clnt_call() ,
which takes a CLIENT pointer rather than a host name.
The parameters to 
.L clnt_call()
are a CLIENT pointer, the procedure number,
the XDR routine for serializing the argument,
a pointer to the argument,
the XDR routine for deserializing the return value,
a pointer to where the return value will be placed,
and the time in seconds to wait for a reply.
.LP
The CLIENT pointer is encoded with  
the transport mechanism.
.L callrpc()
uses UDP, thus it calls
.L clntudp_create()
to get a CLIENT pointer.
To get TCP (Transport Control Protocol), you would use
.L clnttcp_create() .
.LP
The parameters to
.L clntudp_create()
are the server address, the length of the server address,
the program number, the version number,
a timeout value (between tries), and a pointer to a socket.
The final argument to
.L clnt_call()
is the total time to wait for a response.
Thus, the number of tries is the
.L clnt_call()
timeout divided by the
.L clntudp_create()
timeout.
.LP
There is one thing to note when using the
.L clnt_destroy()
call.
It deallocates any space associated with the CLIENT handle,
but it does not close the socket associated with it,
which was passed as an argument to
.L clntudp_create() .
The reason is that if 
there are multiple client handles using the same socket,
then it is possible to close one handle
without destroying the socket that other handles
are using.
.LP
To make a stream connection, the call to
.L clntudp_create()
is replaced with a call to 
.L clnttcp_create(). 
.LS
clnttcp_create(&server_addr, prognum, versnum, &socket, inputsize,
	outputsize);
.LE
There is no timeout argument; instead, the receive and send buffer
sizes must be specified.  When the
.L clnttcp_create()
call is made, a TCP connection is established.
All RPC calls using that CLIENT handle would use this connection.
The server side of an RPC call using TCP has 
.L svcudp_create()
replaced by
.L svctcp_create() .
.bp
.
.H 1 "Other RPC Features"
.LP
This section discusses some other aspects of RPC
that are occasionally useful.
.
.H 2 "Select on the Server Side"
.LP
Suppose a process is processing RPC requests
while performing some other activity.
If the other activity involves periodically updating a data structure,
the process can set an alarm signal before calling
.L svc_run() .
But if the other activity
involves waiting on a a file descriptor, the
.L svc_run()
call won't work.
The code for
.L svc_run()
is as follows:
.LS
void
svc_run()
{
	int readfds;
.sp.5
	for (;;) {
		readfds = svc_fds;
		switch (select(32, &readfds, NULL, NULL, NULL)) {
.sp.5
		case -1:
			if (errno == EINTR)
				continue;
			perror("rstat: select");
			return;
		case 0:
			break;
		default:
			svc_getreq(readfds);
		}
	}
}
.LE
.LP
You can bypass
.L svc_run() 
and call
.L svc_getreq()
yourself.
All you need to know are the file descriptors
of the socket(s) associated with the programs you are waiting on.
Thus you can have your own
.L select()
that waits on both the RPC socket,
and your own descriptors.
.
.H 2 "Broadcast RPC"
.LP
The
.L pmap
and RPC protocols implement broadcast RPC.
Here are the main differences between broadcast RPC
and normal RPC calls:
.IP 1)
Normal RPC expects one answer, whereas
broadcast RPC expects many answers
(one or more answer from each responding machine).
.IP 2)
Broadcast RPC can only be supported by packet-oriented (connectionless)
transport protocols like UPD/IP.
.IP 3)
The implementation of broadcast RPC
treats all unsuccessful responses as garbage by filtering them out.
Thus, if there is a version mismatch between the
broadcaster and a remote service,
the user of broadcast RPC never knows.
.IP 4)
All broadcast messages are sent to the portmap port.
Thus, only services that register themselves with their portmapper
are accessible via the broadcast RPC mechanism.
.
.H 3 "Broadcast RPC Synopsis"
.LP
.LS
#include <rpc/pmap_clnt.h>
.sp.5
enum clnt_stat	clnt_stat;
.sp.5
clnt_stat =
clnt_broadcast(prog, vers, proc, xargs, argsp, xresults, resultsp, eachresult)
u_long		prog;		/* program number */
u_long		vers;		/* version number */
u_long		proc;		/* procedure number */
xdrproc_t	xargs;		/* xdr routine for args */
caddr_t		argsp;		/* pointer to args */
xdrproc_t	xresults;	/* xdr routine for results */
caddr_t		resultsp;	/* pointer to results */
bool_t (*eachresult)();		/* call with each result obtained */
.LE
The procedure
.L eachresult()
is called each time a valid result is obtained.
It returns a boolean that indicates
whether or not the client wants more responses.
.LS
bool_t			done;
.sp.5
done =
eachresult(resultsp, raddr)
caddr_t			resultsp;
struct sockaddr_in	*raddr;   /* address of machine that sent response */
.LE
If
.L done
is TRUE, then broadcasting stops and
.L clnt_broadcast()
returns successfully.
Otherwise, the routine waits for another response.
The request is rebroadcast
after a few seconds of waiting.
If no responses come back,
the routine returns with RPC_TIMEDOUT.
To interpret 
.L clnt_stat
errors, feed the error code to
.L clnt_perrno() .
.
.H 2 "Batching"
.LP
The RPC architecture is designed so that clients send a call message,
and wait for servers to reply that the call succeeded.
This implies that clients do not compute
while servers are processing a call.
This is inefficient if the client does not want or need
an acknowledgement for every message sent.
It is possible for clients to continue computing
while waiting for a response,
using RPC batch facilities.
.LP
RPC messages can be placed in a ``pipeline'' of calls
to a desired server; this is called batching.
Batching assumes that:
1) each RPC call in the pipeline requires no response from the server,
and the server does not send a response message; and
2) the pipeline of calls is transported on a reliable
byte stream transport such as TCP/IP.
Since the server does not respond to every call,
the client can generate new calls in parallel
with the server executing previous calls.
Furthermore, the TCP/IP implementation can buffer up
many call messages, and send them to the server in one
.L write
system call.  This overlapped execution
greatly decreases the interprocess communication overhead of
the client and server processes,
and the total elapsed time of a series of calls.
.LP
Since the batched calls are buffered,
the client should eventually do a legitimate call
in order to flush the pipeline.
.LP
A contrived example of batching follows.
Assume a string rendering service (like a window system)
has two similar calls: one renders a string and returns void results,
while the other renders a string and remains silent.
The service (using the TCP/IP transport) may look like:
.LS
#include <stdio.h>
#include <rpc/rpc.h>
#include <rpcsvc/windows.h>
.sp.5
void windowdispatch();
.sp.5
main()
{
	SVCXPRT *transp;
.sp.5
	transp = svctcp_create(RPC_ANYSOCK, 0, 0);
	if (transp == NULL){
		fprintf(stderr, "couldn't create an RPC server\en");
		exit(1);
	}
	pmap_unset(WINDOWPROG, WINDOWVERS);
	if (!svc_register(transp, WINDOWPROG, WINDOWVERS, windowdispatch,
	    IPPROTO_TCP)) {
		fprintf(stderr, "couldn't register WINDOW service\en");
		exit(1);
	}
	svc_run();  /* never returns */
	fprintf(stderr, "should never reach this point\en");
}
.LE
.LS no
void
windowdispatch(rqstp, transp)
	struct svc_req *rqstp;
	SVCXPRT *transp;
{
	char *s = NULL;
.sp.5
	switch (rqstp->rq_proc) {
	case NULLPROC:
		if (!svc_sendreply(transp, xdr_void, 0)) {
			fprintf(stderr, "couldn't reply to RPC call\en");
			exit(1);
		}
		return;
	case RENDERSTRING:
		if (!svc_getargs(transp, xdr_wrapstring, &s)) {
			fprintf(stderr, "couldn't decode arguments\en");
			svcerr_decode(transp); /* tell caller he screwed up */
			break;
		}
		/*
		 * call here to to render the string s
		 */
		if (!svc_sendreply(transp, xdr_void, NULL)) {
			fprintf(stderr, "couldn't reply to RPC call\en");
			exit(1);
		}
		break;
	case RENDERSTRING_BATCHED:
		if (!svc_getargs(transp, xdr_wrapstring, &s)) {
			fprintf(stderr, "couldn't decode arguments\en");
			/*
			 * we are silent in the face of protocol errors
			 */
			break;
		}
		/*
		 * call here to to render the string s,
		 * but sends no reply!
		 */
		break;
	default:
		svcerr_noproc(transp);
		return;
	}
	/*
	 * now free string allocated while decoding arguments
	 */
	svc_freeargs(transp, xdr_wrapstring, &s);
}
.LE
Of course the service could have one procedure
that takes the string and a boolean
to indicate whether or not the procedure should respond.
.LP
In order for a client to take advantage of batching,
the client must perform RPC calls on a TCP-based transport
and the actual calls must have the following attributes:
1) the result's XDR routine must be zero (NULL), and
2) the RPC call's timeout must be zero.
.LP
Here is an example of a client that uses batching
to render a bunch of strings;
the batching is flushed when the client gets a null string:
.LS
#include <stdio.h>
#include <rpc/rpc.h>
#include <rpcsvc/windows.h>
#include <sys/socket.h>
#include <sys/time.h>
#include <netdb.h>
.sp.5
main(argc, argv)
	int argc;
	char **argv;
{
	struct hostent *hp;
	struct timeval pertry_timeout, total_timeout;
	struct sockaddr_in server_addr;
	int addrlen, sock = RPC_ANYSOCK;
	register CLIENT *client;
	enum clnt_stat clnt_stat;
	char buf[1000];
	char *s = buf;
.sp.5 
	/*
	 * initial as in example 3.3
	 */
	if ((client = clnttcp_create(&server_addr, WINDOWPROG,
	    WINDOWVERS, &sock, 0, 0)) == NULL) {
		perror("clnttcp_create");
		exit(-1);
	}
	total_timeout.tv_sec = 0;
	total_timeout.tv_usec = 0;
	while (scanf("%s", s) != EOF) {
		clnt_stat = clnt_call(client, RENDERSTRING_BATCHED,
		    xdr_wrapstring, &s, NULL, NULL, total_timeout);
		if (clnt_stat != RPC_SUCCESS) {
			clnt_perror(client, "batched rpc");
			exit(-1);
		}
	}
	/*
	 * now flush the pipeline
	 */
	total_timeout.tv_sec = 20;
	clnt_stat = clnt_call(client, NULLPROC,
	    xdr_void, NULL, xdr_void, NULL, total_timeout);
	if (clnt_stat != RPC_SUCCESS) {
		clnt_perror(client, "rpc");
		exit(-1);
	}
	
	clnt_destroy(client);
}
.LE
Since the server sends no message,
the clients cannot be notified of any of the failures that may occur.
Therefore, clients are on their own when it comes to handling errors.
.LP
The above example was completed to render
all of the (2000) lines in the file
.I /etc/termcap .
The rendering service did nothing but to throw the lines away.
The example was run in the following four configurations:
1) machine to itself, regular RPC;
2) machine to itself, batched RPC;
3) machine to another, regular RPC; and
4) machine to another, batched RPC.
The results are as follows:
1) 50 seconds;
2) 16 seconds;
3) 52 seconds;
4) 10 seconds.
Running
.L fscanf()
on
.I /etc/termcap
only requires six seconds.
These timings show the advantage of protocols
that allow for overlapped execution,
though these protocols are often hard to design.
.
.H 2 "Authentication"
.LP
In the examples presented so far,
the caller never identified itself to the server,
and the server never required an ID from the caller.
Clearly, some network services, such as a network filesystem,
require stronger security than what has been presented so far.
.LP
In reality, every RPC call is authenticated by
the RPC package on the server, and similarly,
the RPC client package generates and sends authentication parameters.
Just as different transports (TCP/IP or UDP/IP)
can be used when creating RPC clients and servers,
different forms of authentication can be associated with RPC clients;
the default authentication type used as a default is type
.I none .
.LP
The authentication subsystem of the RPC package is open ended.
That is, numerous types of authentication are easy to support.
However, this section deals only with
.I unix
type authentication, which besides
.I none
is the only supported type.
.
.H 3 "The Client Side"
.LP
When a caller creates a new RPC client handle as in:
.LS
clnt = clntudp_create(address, prognum, versnum, wait, sockp)
.LE
the appropriate transport instance defaults
the associate authentication handle to be
.LS
clnt->cl_auth = authnone_create();
.LE
The RPC client can choose to use
.I unix
style authentication by setting
.L clnt->cl_auth
after creating the RPC client handle:
.LS
clnt->cl_auth = authunix_create_default();
.LE
This causes each RPC call associated with
.L clnt
to carry with it the following authentication credentials structure:
.LS
/*
 * Unix style credentials.
 */
struct authunix_parms {
	u_long 	aup_time;	/* credentials creation time */
	char	*aup_machname;	/* host name of where the client is calling */
	int	aup_uid;	/* client's UNIX effective uid */
	int	aup_gid;	/* client's current UNIX group id */
	u_int	aup_len;	/* the element length of aup_gids array */
	int	*aup_gids;	/* array of 4.2 groups to which user belongs */
};
.LE
These fields are set by
.L authunix_create_default()
by invoking the appropriate system calls.
.LP
Since the RPC user created this new style of authentication,
he is responsible for destroying it with:
.LS
auth_destroy(clnt->cl_auth);
.LE
.
.H 3 "The Server Side"
.LP
Service implementors have a harder time dealing with authentication issues
since the RPC package passes the service dispatch routine a request
that has an arbitrary authentication style associated with it.
Consider the fields of a request handle passed to a service dispatch routine:
.LS
/*
 * An RPC Service request
 */
struct svc_req {
	u_long		rq_prog;	/* service program number */
	u_long		rq_vers;	/* service protocol version number*/
	u_long		rq_proc;	/* the desired procedure number*/
	struct opaque_auth rq_cred;	/* raw credentials from the ``wire'' */
	caddr_t		rq_clntcred;	/* read only, cooked credentials */
};
.LE
The
.L rq_cred
is mostly opaque, except for one field of interest:
the style of authentication credentials:
.LS
/*
 * Authentication info.  Mostly opaque to the programmer.
 */
struct opaque_auth {
	enum_t	oa_flavor;	/* style of credentials */
	caddr_t	oa_base;	/* address of more auth stuff */
	u_int	oa_length;	/* not to exceed MAX_AUTH_BYTES */
};
.LE
The RPC package guarantees the following
to the service dispatch routine:
.IP 1)
That the request's
.L rq_cred
is well formed.  Thus the service implementor may inspect the request's
.L rq_cred.oa_flavor
to determine which style of authentication the caller used.
The service implementor may also wish to inspect the other fields of
.L rq_cred
if the style is not one of the styles supported by the RPC package.
.IP 2)
That the request's
.L rq_clntcred
field is either NULL or points to a well formed structure
that corresponds to a supported style of authentication credentials.
Remember that only
.I unix
style is currently supported, so (currently)
.L rq_clntcred
could be cast to a pointer to an
.L authunix_parms
structure.  If
.L rq_clntcred
is NULL, the service implementor may wish to inspect
the other (opaque) fileds of
.L rq_cred
in case the service knows about a new type of authentication
that the RPC package does not know about.
.LP
Our remote users service example can be extended so that
it computes results for all users except UID 16:
.LS
nuser(rqstp, tranp)
	struct svc_req *rqstp;
	SVCXPRT *transp;
{
	struct authunix_parms *unix_cred;
	int uid;
	unsigned long nusers;
.sp.5
	/*
	 * we don't care about authentication for the null procedure
	 */
	if (rqstp->rq_proc == NULLPROC) {
		if (!svc_sendreply(transp, xdr_void, 0)) {
			fprintf(stderr, "couldn't reply to RPC call\en");
			exit(1);
		 }
		 return;
	}
	/*
	 * now get the uid
	 */
	switch (rqstp->rq_cred.oa_flavor) {
	case AUTH_UNIX:
		unix_cred = (struct authunix_parms *) rqstp->rq_clntcred;
		uid = unix_cred->aup_uid;
		break;
	case AUTH_NULL:
	default:
		svcerr_weakauth(transp);
		return;
	}
	switch (rqstp->rq_proc) {
	case RUSERSPROC_NUM:
		/*
		 * make sure the caller is allow to call this procedure.
		 */
		if (uid == 16) {
			svcerr_systemerr(transp);
			return;
		}
		/*
		 * code here to compute the number of users
		 * and put in variable nusers
		 */
		if (!svc_sendreply(transp, xdr_u_long, &nusers) {
			fprintf(stderr, "couldn't reply to RPC call\en");
			exit(1);
		}
		return;
	default:
		svcerr_noproc(transp);
		return;
	}
}
.LE
A few things should be noted here.
First, it is customary not to check the authentication parameters
associated with the NULLPROC (procedure number zero).
Second, if the authentication parameter's type is not suitable
for your service, you should call
.L svcerr_weakauth() .
And finally, the service protocol itself should return status
for access denied; in the case of our example, the protocol
does not have such a status, so we call the service primitive
.L svcerr_systemerr()
instead.
.LP
The last point underscores the relation between
the RPC authentication package and the services;
RPC deals only with authentication and not with
individual services' access control.
The services themselves must implement their own access control policies
and reflect these policies as return statuses in their protocols.
.
.H 2 "Using Inetd"
.LP
An RPC server can be started from
.L inetd .
The only difference
from the usual code is that
.L svcudp_create()
should be called as
.LS
transp = svcudp_create(0);
.LE
since
.L inet
passes a socket as file descriptor 0.
Also,
.L svc_register()
should be called as
.LS
svc_register(PROGNUM, VERSNUM, service, transp, 0);
.LE
with the final flag as 0,
since the program would already be registered by
.L inetd .
Remember that if you want to exit
from the server process and return control to
.L inet ,
you need to explicitly exit, since
.L svc_run()
never returns.
.LP
The format of entries in /etc/servers for RPC services is
.LS
rpc udp \fIserver \0program \0version\fP
.LE
where 
.I server
is the C code implementing the server,
and 
.I program
and
.I version
are the program and version numbers of the service.
The key word 
.L udp
can be replaced by 
.L tcp
for TCP-based RPC services.
.LP
If the same program handles multiple versions,
then the version number can be a range,
as in this example:
.LS
rpc udp /usr/etc/rstatd 100001 1-2
.LE
.bp
.
.H 1 "More Examples"
.H 2 "Versions"
.LP
By convention, the first version number of program FOO is 
FOOVERS_ORIG and the most recent version is FOOVERS.
Suppose there is a new version of the 
.L user
program that returns an
.L "unsigned short"
rather than a
.L long .
If we name this version
RUSERSVERS_SHORT, then
a server that wants to support both versions
would do a double register.
.LS
.sp.5
if (!svc_register(transp, RUSERSPROG, RUSERSVERS_ORIG, nuser,
    IPPROTO_TCP)) {
	fprintf(stderr, "couldn't register RUSER service\en");
	exit(1);
}
if (!svc_register(transp, RUSERSPROG, RUSERSVERS_SHORT, nuser,
    IPPROTO_TCP)) {
	fprintf(stderr, "couldn't register RUSER service\en");
	exit(1);
}
.LE
.bp
Both versions can be handled by the same C procedure:
.LS 0
nuser(rqstp, tranp)
	struct svc_req *rqstp;
	SVCXPRT *transp;
{
	unsigned long nusers;
	unsigned short nusers2
.sp.5
	switch (rqstp->rq_proc) {
	case NULLPROC:
		if (!svc_sendreply(transp, xdr_void, 0)) {
			fprintf(stderr, "couldn't reply to RPC call\en");
			exit(1);
		}
		return;
	case RUSERSPROC_NUM:
		/*
		 * code here to compute the number of users
		 * and put in variable nusers
		 */
		nusers2 = nusers;
		if (rqstp->rq_vers == RUSERSVERS_ORIG)
			if (!svc_sendreply(transp, xdr_u_long, &nusers) {
				fprintf(stderr, "couldn't reply to RPC call\en");
				exit(1);
			}
		else
			if (!svc_sendreply(transp, xdr_u_short, &nusers2) {
				fprintf(stderr, "couldn't reply to RPC call\en");
				exit(1);
		return;
	default: 
		svcerr_noproc(transp);
		return;
	}
}
.LE
.H 2 "TCP"
.LP
Here is an example that is essentially  
.L rcp .
The initiator of the RPC 
.L snd()
call takes its standard input and sends it to the server 
.L rcv() ,
which prints it on standard output.
The RPC call uses TCP.
This also illustrates an XDR procedure that behaves differently
on serialization than on deserialization.
.LS 0
/* 
 * The xdr routine:
 *
 * on decode, read from wire, write onto fp
 * on encode, read from fp, write onto wire
 */
#include <stdio.h>
#include <rpc/rpc.h>
.sp.5
xdr_rcp(xdrs, fp)
	XDR *xdrs;
	FILE *fp;
{
	unsigned long size;
	char buf[MAXCHUNK], *p;
.sp.5
	if (xdrs->x_op == XDR_FREE)/* nothing to free */
		return 1;
	while (1) {
		if (xdrs->x_op == XDR_ENCODE) {
			if ((size = fread (buf, sizeof(char), MAXCHUNK, fp))
			    == 0 && ferror(fp)) {
				fprintf(stderr, "couldn't fread\en");
				exit(1);
			}
		}
		p = buf;
		if (!xdr_bytes(xdrs, &p, &size, MAXCHUNK))
			return 0;
		if (size == 0)
			return 1;
		if (xdrs->x_op == XDR_DECODE) {
			if (fwrite(buf, sizeof(char), size, fp) != size) {
				fprintf(stderr, "couldn't fwrite\en");
				exit(1);
			}
		}
	}
}
.LE
.LS 0
/*
 * The sender routines
 */
#include <stdio.h>
#include <netdb.h>
#include <rpc/rpc.h>
#include <sys/socket.h>
#include <sys/time.h>
.sp.5
main(argc, argv)
	int argc;
	char **argv;
{
	int err;
.sp.5
	if (argc < 2) {
		fprintf(stderr, "usage: %s server-name\en", argv[0]);
		exit(-1);
	}
	if ((err = callrpctcp(argv[1], RCPPROG, RCPPROC_FP, RCPVERS,
	    xdr_rcp, stdin, xdr_void, 0) != 0)) {
		clnt_perrno(err);
		fprintf(stderr, " couldn't make RPC call\en");
		exit(1);
	}
}
.sp.5
callrpctcp(host, prognum, procnum, versnum, inproc, in, outproc, out)
	char *host, *in, *out;
	xdrproc_t inproc, outproc;
{
	struct sockaddr_in server_addr;
	int socket = RPC_ANYSOCK;
	enum clnt_stat clnt_stat;
	struct hostent *hp;
	register CLIENT *client;
	struct timeval total_timeout;
.sp.5
	if ((hp = gethostbyname(host)) == NULL) {
		fprintf(stderr, "cannot get addr for '%s'\en", host);
		exit(-1);
	}
	bcopy(hp->h_addr, (caddr_t)&server_addr.sin_addr, hp->h_length);
	server_addr.sin_family = AF_INET;
	server_addr.sin_port =  0;
	if ((client = clnttcp_create(&server_addr, prognum,
	    versnum, &socket, BUFSIZ, BUFSIZ)) == NULL) {
		perror("rpctcp_create");
		exit(-1);
	}
	total_timeout.tv_sec = 20;
	total_timeout.tv_usec = 0;
	clnt_stat = clnt_call(client, procnum, inproc, in, outproc, out, total_timeout);
	clnt_destroy(client)
	return (int)clnt_stat;
}
.LE
.LS 0
/*
 * The receiving routines
 */
#include <stdio.h>
#include <rpc/rpc.h>
.sp.5
main()
{
	register SVCXPRT *transp;
.sp.5
	if ((transp = svctcp_create(RPC_ANYSOCK, 1024, 1024)) == NULL) {
		fprintf("svctcp_create: error\en");
		exit(1);
	}
	pmap_unset(RCPPROG, RCPVERS);
	if (!svc_register(transp, RCPPROG, RCPVERS, rcp_service, IPPROTO_TCP)) {
		fprintf(stderr, "svc_register: error\en");
		exit(1);
	}
	svc_run();  /* never returns */
	fprintf(stderr, "svc_run should never return\en");
}
.sp.5
rcp_service(rqstp, transp)
	register struct svc_req *rqstp;
	register SVCXPRT *transp;
{
	switch (rqstp->rq_proc) {
	case NULLPROC:
		if (svc_sendreply(transp, xdr_void, 0) == 0) {
			fprintf(stderr, "err: rcp_service");
			exit(1);
		}
		return;
	case RCPPROC_FP:
		if (!svc_getargs(transp, xdr_rcp, stdout)) {
			svcerr_decode(transp);
			return;
		}
		if (!svc_sendreply(transp, xdr_void, 0)) {
			fprintf(stderr, "can't reply\en");
			return;
		}
		exit(0);
	default: 
		svcerr_noproc(transp);
		return;
	}
}
.LE
.H 2 "Callback Procedures"
.LP
Occasionally, it is useful to have a server become a client,
and make an RPC call back the process which is its client.
An example is remote debugging,
where the client is a window system program,
and the server is a debugger running on the remote machine.
Most of the time,
the user clicks a mouse button at the debugging window,
which converts this to a debugger command,
and then makes an RPC call to the server
(where the debugger is actually running),
telling it to execute that command.
However, when the debugger hits a breakpoint, the roles are reversed,
and the debugger wants to make an rpc call to the window program,
so that it can inform the user that a breakpoint has been reached.
.LP
In order to do an RPC callback,
you need a program number to make the RPC call on.
Since this will be a dynamically generated program number,
it should be in the transient range, 0x40000000 - 0x5fffffff.
The routine
.L gettransient()
returns a valid program number in the transient range,
and registers it with the portmapper.
It only talks to the portmapper running on the same machine as the
.L gettransient()
routine itself.
The call to
.L pmap_set()
is a test and set operation,
in that it indivisibly tests whether a program number
has already been registered,
and if it has not, then reserves it.
On return, the
.L sockp
argument will contain a socket that can be used
as the argument to an
.L svcudp_create()
or
.L svctcp_create()
call.
.LS
#include <stdio.h>
#include <rpc/rpc.h>
#include <sys/socket.h>
.sp.5
gettransient(proto, vers, sockp)
	int *sockp;
{
	static int prognum = 0x40000000;
	int s, len, socktype;
	struct sockaddr_in addr;
.sp.5
	switch(proto) {
		case IPPROTO_UDP:
			socktype = SOCK_DGRAM;
			break;
		case IPPROTO_TCP:
			socktype = SOCK_STREAM;
			break;
		default:
			fprintf(stderr, "unknown protocol type\en");
			return 0;
	}
	if (*sockp == RPC_ANYSOCK) {
		if ((s = socket(AF_INET, socktype, 0)) < 0) {
			perror("socket");
			return (0);
		}
		*sockp = s;
	}
	else
		s = *sockp;
	addr.sin_addr.s_addr = 0;
	addr.sin_family = AF_INET;
	addr.sin_port = 0;
	len = sizeof(addr);
	/*
	 * may be already bound, so don't check for err
	 */
	bind(s, &addr, len);
	if (getsockname(s, &addr, &len)< 0) {
		perror("getsockname");
		return (0);
	}
	while (pmap_set(prognum++, vers, proto, addr.sin_port) == 0)
		continue;
	return (prognum-1);
}
.LE
The following pair of programs illustrate how to use the
.L gettransient()
routine.
The client makes an RPC call to the server,
passing it a transient program number.
Then the client waits around to receive a callback
from the server at that program number.
The server registers the program EXAMPELPROG,
so that it can receive the RPC call
informing it of the callback program number.
Then at some random time (on receiving an ALRM signal in this example),
it sends a callback RPC call,
using the program number it received earlier.
.LS
/*
 * client
 */
#include <stdio.h>
#include <rpc/rpc.h>
.sp.5
int callback();
char hostname[256];
.sp.5
main(argc, argv)
	char **argv;
{
	int x, ans, s;
	SVCXPRT *xprt;
.sp.5
	gethostname(hostname, sizeof(hostname));
	s = RPC_ANYSOCK;
	x = gettransient(IPPROTO_UDP, 1, &s);
	fprintf(stderr, "client gets prognum %d\en", x);

	if ((xprt = svcudp_create(s)) == NULL) {
	    fprintf(stderr, "rpc_server: svcudp_create\en");
	    exit(1);
	}
	(void)svc_register(xprt, x, 1, callback, 0);

	ans = callrpc(hostname, EXAMPLEPROG, EXAMPLEPROC_CALLBACK,
	    EXAMPLEVERS, xdr_int, &x, xdr_void, 0);
	if (ans != 0) {
		fprintf(stderr, "call: ");
		clnt_perrno(ans);
		fprintf(stderr, "\en");
	}
	svc_run();
	fprintf(stderr, "Error: svc_run shouldn't have returned\en");
}
.LE
.LS
callback(rqstp, transp)
	register struct svc_req *rqstp;
	register SVCXPRT *transp;
{
	switch (rqstp->rq_proc) {
		case 0:
			if (svc_sendreply(transp, xdr_void, 0)  == FALSE) {
				fprintf(stderr, "err: rusersd\en");
				exit(1);
			    }
			exit(0);
		case 1:
			if (!svc_getargs(transp, xdr_void, 0)) {
			    	svcerr_decode(transp);
				exit(1);
			}
			fprintf(stderr, "client got callback\en");
			if (svc_sendreply(transp, xdr_void, 0)  == FALSE) {
				fprintf(stderr, "err: rusersd");
				exit(1);
			}
	}
}
.LE
.LS
/*
 * server
 */
#include <stdio.h>
#include <rpc/rpc.h>
#include <sys/signal.h>
.sp.5
char *getnewprog();
char hostname[256];
int docallback();
int pnum;		/*program number for callback routine */ 
.sp.5
main(argc, argv)
	char **argv;
{
	gethostname(hostname, sizeof(hostname));
	registerrpc(EXAMPLEPROG, EXAMPLEPROC_CALLBACK, EXAMPLEVERS,
	    getnewprog, xdr_int, xdr_void);
	fprintf(stderr, "server going into svc_run\en");
	alarm(10);
	signal(SIGALRM, docallback);
	svc_run();
	fprintf(stderr, "Error: svc_run shouldn't have returned\en");
}
.sp.5
char *
getnewprog(pnump)
	char *pnump;
{
	pnum = *(int *)pnump;
	return NULL;
}
.sp.5
docallback()
{
	int ans;
.sp.5
	ans = callrpc(hostname, pnum, 1, 1, xdr_void, 0, xdr_void, 0);
	if (ans != 0) {
		fprintf(stderr, "server: ");
		clnt_perrno(ans);
		fprintf(stderr, "\en");
	}
}
.LE