NAME | SYNOPSIS | DESCRIPTION | RETURN VALUES | FILES | ATTRIBUTES | SEE ALSO
cc [ flag ... ] file ... -lnsl [ library ... ] #include <rpcsvc/ypclnt.h> #include <rpcsvc/yp_prot.h>
This package of functions provides an interface to NIS, Network Information Service Version 2, formerly referred to as YP. In this version of SunOS, NIS version 2 is supported only for compatibility with previous versions. The recommended enterprise level information service is NIS+ or NIS version 3, see nis+(1) . Moreover, this version of SunOS supports only the client interface to NIS version 2. It is expected that this client interface will be served either by an existing ypserv process running on another machine on the network that has an earlier version of SunOS or by an NIS+ server, see rpc.nisd(1M) , running in "YP-compatibility mode". Refer to the NOTES section in ypfiles(4) for implications of being an NIS client of an NIS+ server in "YP-compatibility mode", and to ypbind(1M) , ypwhich(1) , ypmatch(1) , and ypcat(1) for commands to access NIS from a client machine. The package can be loaded from the standard library, /usr/lib/libnsl.so.1 .
All input parameter names begin with in . Output parameters begin with out . Output parameters of type char ** should be addresses of uninitialized character pointers. Memory is allocated by the NIS client package using malloc(3C) , and may be freed by the user code if it has no continuing need for it. For each outkey and outval , two extra bytes of memory are allocated at the end that contain NEWLINE and null, respectively, but these two bytes are not reflected in outkeylen or outvallen . indomain and inmap strings must be non-null and null-terminated. String parameters which are accompanied by a count parameter may not be null, but may point to null strings, with the count parameter indicating this. Counted strings need not be null-terminated.
All functions in this package of type int return 0 if they succeed, and a failure code (YPERR_ xxxx ) otherwise. Failure codes are described in the ERRORS section.
To use the NIS name services, the client process must be "bound" to an NIS server that serves the appropriate domain using yp_bind() . Binding need not be done explicitly by user code; this is done automatically whenever an NIS lookup function is called. yp_bind() can be called directly for processes that make use of a backup strategy (for example, a local file) in cases when NIS services are not available. If a process calls yp_bind() , it should call yp_unbind() when it is done using NIS in order to free up resources.
Each binding allocates (uses up) one client process socket descriptor; each bound domain costs one socket descriptor. However, multiple requests to the same domain use that same descriptor. yp_unbind() is available at the client interface for processes that explicitly manage their socket descriptors while accessing multiple domains. The call to yp_unbind() makes the domain unbound , and frees all per-process and per-node resources used to bind it.
If an RPC failure results upon use of a binding, that domain will be unbound automatically. At that point, the ypclnt() layer will retry a few more times or until the operation succeeds, provided that rpcbind(1M) and ypbind(1M) are running, and either
the client process cannot bind a server for the proper domain, or
RPC requests to the server fail.
If an error is not RPC-related, or if rpcbind is not running, or if ypbind is not running, or if a bound ypserv process returns any answer (success or failure), the ypclnt layer will return control to the user code, either with an error code, or a success code and any results.
The NIS lookup calls require a map name and a domain name, at minimum. It is assumed that the client process knows the name of the map of interest. Client processes should fetch the node's default domain by calling yp_get_default_domain() , and use the returned outdomain as the indomain parameter to successive NIS name service calls. The domain thus returned is the same as that returned using the SI_SRPC_DOMAIN command to the sysinfo(2) system call. The value returned in outdomain should not be freed.
yp_match() returns the value associated with a passed key. This key must be exact; no pattern matching is available. yp_match() requires a full YP map name; for example, hosts.byname instead of the nickname hosts .
yp_first() returns the first key-value pair from the named map in the named domain.
yp_next() returns the next key-value pair in a named map. The inkey parameter must be the outkey returned from an initial call to yp_first() (to get the second key-value pair) or the one returned from the n th call to yp_next() (to get the n th + second key-value pair). Similarly, the inkeylen parameter must be the outkeylen returned from the earlier yp_first() or yp_next() call.
The concept of first (and, for that matter, of next) is particular to the structure of the NIS map being processing; there is no relation in retrieval order to either the lexical order within any original (non-NIS name service) data base, or to any obvious numerical sorting order on the keys, values, or key-value pairs. The only ordering guarantee made is that if the yp_first() function is called on a particular map, and then the yp_next() function is repeatedly called on the same map at the same server until the call fails with a reason of YPERR_NOMORE , every entry in the data base will be seen exactly once. Further, if the same sequence of operations is performed on the same map at the same server, the entries will be seen in the same order.
Under conditions of heavy server load or server failure, it is possible for the domain to become unbound, then bound once again (perhaps to a different server) while a client is running. This can cause a break in one of the enumeration rules; specific entries may be seen twice by the client, or not at all. This approach protects the client from error messages that would otherwise be returned in the midst of the enumeration. The next paragraph describes a better solution to enumerating all entries in a map.
The function yp_all() provides a way to transfer an entire map from server to client in a single request using TCP (rather than UDP as with other functions in this package). The entire transaction take place as a single RPC request and response. yp_all() can be used just like any other NIS name service procedure, identify the map in the normal manner, and supply the name of a function which will be called to process each key-value pair within the map. The call to yp_all() returns only when the transaction is completed (successfully or unsuccessfully), or the foreach() function decides that it does not want to see any more key-value pairs.
The third parameter to yp_all() is
struct ypall_callback *incallback { int (*foreach)(); char *data; };
The function foreach() is called
foreach(int instatus, char *inkey, int inkeylen, char *inval, int invallen, char *indata);
The instatus parameter will hold one of the return status values defined in <rpcsvc/yp_prot.h -- either YP_TRUE or an error code. (See ypprot_err() , below, for a function which converts an NIS name service protocol error code to a ypclnt layer error code.)
The key and value parameters are somewhat different than defined in the synopsis section above. First, the memory pointed to by the inkey and inval parameters is private to the yp_all() function, and is overwritten with the arrival of each new key-value pair. It is the responsibility of the foreach() function to do something useful with the contents of that memory, but it does not own the memory itself. Key and value objects presented to the foreach() function look exactly as they do in the server's map -- if they were not NEWLINE-terminated or null-terminated in the map, they will not be here either.
The indata parameter is the contents of the incallback => data element passed to yp_all() . The data element of the callback structure may be used to share state information between the foreach() function and the mainline code. Its use is optional, and no part of the NIS client package inspects its contents -- cast it to something useful, or ignore it.
The foreach() function is a Boolean. It should return 0 to indicate that it wants to be called again for further received key-value pairs, or non-zero to stop the flow of key-value pairs. If foreach() returns a non-zero value, it is not called again; the functional value of yp_all() is then 0 .
yp_order() returns the order number for a map. This function is not supported if the ypbind process on the client's system is bound to an NIS+ server running in "YP-compatibility mode".
yp_master() returns the machine name of the master NIS server for a map.
yperr_string() returns a pointer to an error message string that is null-terminated but contains no period or NEWLINE .
ypprot_err() takes an NIS name service protocol error code as input, and returns a ypclnt layer error code, which may be used in turn as an input to yperr_string() .
All integer functions return 0 if the requested operation is successful, or one of the following errors if the operation fails.
Access violation.
The arguments to the function are bad.
The YP database is bad.
The database is busy.
Cannot bind to server on this domain.
No such key in map.
No such map in server's domain.
Local domain name not set.
No more records in map database.
Cannot communicate with rpcbind .
Resource allocation failure.
RPC failure; domain has been unbound.
Cannot communicate with ypbind .
Internal YP server or client error.
Cannot communicate with ypserv .
YP version mismatch.
See attributes(5) for descriptions of the following attributes:
ATTRIBUTE TYPE | ATTRIBUTE VALUE |
---|---|
MT-Level | Safe |
nis+(1) , ypcat(1) , ypmatch(1) , ypwhich(1) , rpc.nisd(1M) , rpcbind(1M) , ypbind(1M) , ypserv(1M) , sysinfo(2) , malloc(3C) , ypfiles(4) , attributes(5)
NAME | SYNOPSIS | DESCRIPTION | RETURN VALUES | FILES | ATTRIBUTES | SEE ALSO