9PCLIENT(3) 9PCLIENT(3)
NAME
CFid, CFsys, fsinit, fsmount, fsroot, fssetroot, fsunmount,
nsinit, nsmount, fsversion, fsauth, fsattach, fsclose,
fscreate, fsfcreate, fsremove, fsfremove, fsaccess,
fsdirread, fsdirreadall, fsdirstat, fsdirfstat, fsdirwstat,
fsdirfwstat, fsopen, fsfopen, nsopen, fsopenfd, fspread,
fspwrite, fsread, fsreadn, fsseek, fswrite, fsprint,
fsvprint - 9P client library
SYNOPSIS
#include <u.h>
#include <libc.h>
#include <fcall.h>
#include <thread.h>
#include <9pclient.h>
CFsys* fsmount(int fd, char *aname)
CFsys* nsmount(char *name, char *aname)
CFid* fsroot(CFsys *fsys)
void fsunmount(CFsys *fsys)
CFsys* fsinit(int fd)
CFsys* nsinit(char *name)
int fsversion(CFsys *fsys, int msize, char *version, int
nversion)
CFid* fsauth(CFsys *fsys, char *uname, char *aname)
CFid* fsattach(CFsys *fsys, CFid *afid, char *uname, char
*aname)
void fssetroot(CFsys *fsys, CFid *fid)
void fsclose(CFid *fid)
CFid* fscreate(CFsys *fs, char *path, int mode, ulong perm)
int fsfcreate(CFid *fid, char *path, int mode, ulong
perm)
int fsremove(CFSys *fs, char *path)
Page 1 Plan 9 (printed 11/18/25)
9PCLIENT(3) 9PCLIENT(3)
int fsfremove(CFid *fid)
int fsaccess(CFsys *fs, char *path, int amode)
CFid* fsopen(CFsys *fs, char *path, int mode)
int fsfopen(CFid *fid, char *path, int mode)
long fspread(CFid *fid, void *buf, long n, vlong offset)
long fspwrite(CFid *fid, void *buf, long n, vlong offset)
long fsread(CFid *fid, void *buf, long n)
long fsreadn(CFid *fid, void *buf, long n)
long fswrite(CFid *fid, void *buf, long n)
int fsprint(CFid *fid, char *fmt, ...)
int fsvprint(CFid *fid, char *fmt, ...)
vlong fsseek(CFid *Fid, vlong n, int type)
Qid fsqid(CFid *fid)
long fsdirread(CFid *fid, Dir **d)
long fsdirreadall(CFid *fid, Dir **d)
Dir* fsdirstat(CFsys *fs, char *path)
Dir* fsdirfstat(CFid *fid)
int fsdirwstat(CFsys *fs, char *path, Dir *d)
int fsdirfwstat(CFid *fid, Dir *d)
int fsopenfd(CFsys *fs, char *path, int mode)
CFsys* nsopen(char *name, char *aname, char *path, int mode)
extern int chatty9pclient;
extern int eofkill9pclient;
DESCRIPTION
The 9pclient library helps client programs interact with 9P
servers.
A CFsys* represents a connection to a 9P server. A CFid*
represents an active fid on some connection; see intro(9p).
Page 2 Plan 9 (printed 11/18/25)
9PCLIENT(3) 9PCLIENT(3)
A new connection to a 9P server is typically established by
fsmount or nsmount. Fsmount initializes a new 9P conversa-
tion on the open file descriptor fd; nsmount connects to a
service named name in the current name space directory (see
intro(4)). Both attach to the root of the file system using
the attach name aname. Fsroot returns the CFid* correspond-
ing to this root.
Fsinit, nsinit, fsversion, fsauth, fsattach, and fssetroot
provide more detailed control over the file system connec-
tion than fsmount and nsmount. Fsinit allocates a new CFsys*
corresponding to a 9P conversation on the file descriptor fd
and then calls fsversion to initialize the connection.
Nsinit does the same for name space services. Fsversion
executes a version(9p) transaction to establish maximum mes-
sage size and 9P version. Fsauth executes an auth(9p)
transaction, returning the new auth fid. (Fsread and
fswrite can then be used to run the authentication protocol
over the fid.) Fsattach executes an attach(9p) transaction
to connect to the root of a file tree served by the server.
It presents afid (which may be nil) to establish identity.
Fssetroot sets the root fid used by fsopen, fsopenfd,
fsdirstat, and fsdirwstat, which evaluate rooted path names.
When a fid is no longer needed, it should be clunked by
calling fsclose and then considered freed. Similarly, when
the connection to the server is no longer needed, it should
be closed by calling fsunmount, which will take care of
calling fsclose on the current root fid. Once all fids have
been clunked and the connection has been closed (the order
is not important), the allocated structures will be freed
and the file descriptor corresponding to the connection will
be closed (see close(2)). Fids are not reference counted:
when fsclose is called, the clunk transaction and freeing of
storage happen immediately. Despite its name, fsclose can
be used to clunk fids that are not open for I/O.
Fscreate and fsopen establish new fids using the walk,
create and open transactions (see walk(9p) and open(9p)).
The path argument is evaluated relative to the CFsys root
(see fsroot and fssetroot above). The path is parsed as a
slash-separated sequence of path elements, as on Unix and
Plan 9. Elements that are empty or dot (.) are ignored.
Alternately, fswalk walks from a fid to a given name to cre-
ate a new fid. The name may be nil, corresponding to a walk
with no names. Otherwise the name is taken as a slash-
separated sequence of path elements. Fsfcreate and fsfopen
issue create and open transactions using the passed fid
argument, which should have been obtained by calling fswalk.
Once opened, fids can be read and written using fspread and
Page 3 Plan 9 (printed 11/18/25)
9PCLIENT(3) 9PCLIENT(3)
fspwrite, which execute read and write transactions (see
read(9p)). The library maintains an offset for each fid,
analagous to the offset maintained by the kernel for each
open file descriptor. Fsread and fswrite read and write
from this offset, and update it after successful calls.
Fsseek sets the offset; the n and type arguments are used as
in seek(3). Calling fspread or fspwrite with an offset of -1
is identical to calling fsread or fswrite. Fsreadn calls
fsread repeatedly to obtain exactly n bytes of data, unless
it encounters end-of-file or an error.
Attach, walk, create, and open transactions include in their
replies an updated qid for the fid being manipulated. Fsqid
returns the most recent qid returned by one of these trans-
actions for the given fid.
Fsaccess behaves like Unix's access(2). Fsremove removes the
named path. Fsfremove removes the path corresponding to an
open CFid*.
Reading an open a directory returns directory entries
encoded as described in stat(9p).
Fsprint and fsvprint are like fprint and vfprint (see
print(3)) but write to CFid*s.
Fsdirread calls fsread and then parses the encoded entries
into an array of Dir* data structures, storing a pointer to
the array in *d and returning the number of entries.
Fsdirreadall is similar but reads the entire directory. The
returned pointer should be freed with free (see malloc(3))
when no longer needed.
Fsdirfstat and fsdirfwstat execute stat and wstat (see
stat(9p)) transactions. The Dir structure returned by
fsdirfstat should be freed with free (see malloc(3)) when no
longer needed.
Fsdirstat and fsdirwstat are similar to fsdirfstat and
fsdirfwstat but operate on paths relative to the file system
root (see fsopen and fscreate above).
Fsopenfd opens a file on the 9P server for reading or writ-
ing but returns a Unix file descriptor instead of a fid
structure. The file descriptor is actually one end of a
pipe(2). A proxy process on the other end is ferrying data
between the pipe and the 9P fid. Because of the implementa-
tion as a pipe, the only signal of a read or write error is
the closing of the pipe. The file descriptor remains valid
even after the CFsys is unmounted.
Nsopen opens a single file on a name space server: it runs
Page 4 Plan 9 (printed 11/18/25)
9PCLIENT(3) 9PCLIENT(3)
nsmount, fsopen, and then fsunmount.
If the chatty9pclient flag is set, the library prints all 9P
messages to standard error. If the eofkill9pclient flag is
set, the library calls threadexitsall (see thread(3)) when
it detects EOF on a 9P connection.
SOURCE
/usr/local/plan9/src/lib9pclient
SEE ALSO
intro(4), intro(9p), fsaopen and nsaopen in auth(3)
BUGS
The implementation should use a special version string to
distinguish between servers that support openfd(9p) and
servers that do not.
The interface does not provide access to the walk(9p) trans-
action, or to open and create on already-established fids.
Page 5 Plan 9 (printed 11/18/25)