[plan9front.git] / sys / man / 2 / 9p

.TH 9P 2
.SH NAME
Srv,
chatty9p,
dirread9p,
emalloc9p,
erealloc9p,
estrdup9p,
listensrv,
postmountsrv,
postsharesrv,
postsrv,
readbuf,
readstr,
respond,
responderror,
srv,
srvacquire,
srvforker,
srvrelease,
threadlistensrv,
threadpostmountsrv,
threadpostsharesrv,
threadpostsrv,
threadsrv,
threadsrvforker,
auth9p,
authattach,
authread,
authwrite \- 9P file service
.SH SYNOPSIS
.ft L
.nf
#include <u.h>
#include <libc.h>
#include <fcall.h>
#include <thread.h>
#include <9p.h>
.fi
.PP
.ft L
.nf
.ta \w'\fL1234'u +\w'\fLTree* 'u
typedef struct Srv {
        Tree*   tree;

        void            (*attach)(Req *r);
        void            (*auth)(Req *r);
        void            (*open)(Req *r);
        void            (*create)(Req *r);
        void            (*read)(Req *r);
        void            (*write)(Req *r);
        void            (*remove)(Req *r);
        void            (*flush)(Req *r);
        void            (*stat)(Req *r);
        void            (*wstat)(Req *r);
        void            (*walk)(Req *r);

        char*   (*walk1)(Fid *fid, char *name, Qid *qid);
        char*   (*clone)(Fid *oldfid, Fid *newfid);

        void            (*destroyfid)(Fid *fid);
        void            (*destroyreq)(Req *r);
        void            (*start)(Srv *s);
        void            (*end)(Srv *s);
        void*   aux;

        int             infd;
        int             outfd;

        void            (*forker)(void (*fn)(void*), void *arg, int flags);
} Srv;
.fi
.PP
.nf
.ft L
.ta \w'\fLvoid* 'u
void    srv(Srv *s)
int     postsrv(Srv *s, char *name);
void    postmountsrv(Srv *s, char *name, char *mtpt, int flag)
void    postsharesrv(Srv *s, char *name, char *mtpt, char *desc)
void    listensrv(Srv *s, char *addr)
void    threadsrv(Srv *s)
int     threadpostsrv(Srv *s, char *name);
void    threadpostmountsrv(Srv *s, char *name, char *mtpt, int flag)
void    threadpostsharesrv(Srv *s, char *name, char *mtpt, char *desc)
void    threadlistensrv(Srv *s, char *addr)
void    srvforker(void (*fn)(void*), void *arg, int flags)
void    threadsrvforker(void (*fn)(void*), void *arg, int flags)
void    respond(Req *r, char *error)
void    responderror(Req*)
void    readstr(Req *r, char *src)
void    readbuf(Req *r, void *src, long nsrc)
typedef int Dirgen(int n, Dir *dir, void *aux)
void            dirread9p(Req *r, Dirgen *gen, void *aux)
void    walkandclone(Req *r, char *(*walk1)(Fid *old, char *name, void *v),
          char *(*clone)(Fid *old, Fid *new, void *v), void *v)
.fi
.PP
.nf
.ft L
.ta \w'\fLvoid* 'u
void    srvrelease(Srv *s)
void    srvacquire(Srv *s)
.fi
.PP
.fi
.PP
.nf
.ft L
.ta \w'\fLvoid* 'u
void*   emalloc9p(ulong n)
void*   erealloc9p(void *v, ulong n)
char*   estrdup9p(char *s)
.fi
.PP
.nf
.ft L
extern int chatty9p;
.fi
.PP
.nf
.ft L
.ta \w'\fLvoid* 'u
void    auth9p(Req *r);
int             authattach(Req *r);
void    authread(Req *r);
void    authwrite(Req *r);
.fi
.SH DESCRIPTION
The function
.I srv
and
.I threadsrv
serve a 9P session by reading requests from
.BR s->infd ,
dispatching them to the function pointers kept in 
.BR Srv ,
and
writing the responses to
.BR s->outfd .
(Typically,
.I postmountsrv
or
.I threadpostmountsrv
initializes the
.B infd
and
.B outfd
structure members.  See the description below.)
.PP
.B Req
and 
.B Fid
structures are allocated one-to-one with uncompleted
requests and active fids, and are described in
.IR 9pfid (2).
.PP
The behavior of
.I srv
depends on whether there is a file tree
(see
.IR 9pfile (2))
associated with the server, that is,
whether the
.B tree
element is nonzero.
The differences are made explicit in the
discussion of the service loop below.
The
.B aux
element is the client's, to do with as it pleases.
.PP
.I Srv
does not return until the 9P conversation is finished.
Since it is usually run in a separate process so that
the caller can exit, the service loop has little chance
to return gracefully on out of memory errors.
It calls 
.IR emalloc9p ,
.IR erealloc9p ,
and
.I estrdup9p
to obtain its memory.
The default implementations of these functions
act as
.IR malloc ,
.IR realloc ,
and
.I strdup 
but abort the program if they run out of memory.
If alternate behavior is desired, clients can link against
alternate implementations of these functions.
.PP
The functions
.I srvforker
and
.I threadsrvforker
handle the creation of new processes on a connection which use
.I rfork
(see
.IR fork (2))
or
.I procrfork
(see
.IR thread (2)).
.PP

.I Postsrv
and
.I threadpostsrv
are wrappers that create a separate process in which to run
.IR srv .
They do the following:
.IP
Initialize
.IB s -> infd
and
.IB s -> outfd
to be one end of a freshly allocated pipe.
.IP
If
.B name
is non-nil, post the other end file descriptor under the name
.BI /srv/ name .
.IP
Initialize
.IB s -> forker
to eigther
.I srvforker
or
.I threadsrvforker
unless already initialized to a non-nil value.
.IP
Fork a child process via
.IB s -> forker
using the
.BR RFPROC ,
.BR RFNOWAIT ,
.BR RFNAMEG ,
.B RFNOTEG
and
.B RFMEM
flags. This isolates the service loop from the callers
namespace and from notes posted to the callers note group
but shares data and bss segments.
.IP
The child process then waits for the parent to copy its
file descriptor table via
.I rfork
using
.B RFFDG
flag. This way, the service loop will share the original
file descriptor table with previously created child
processes of the caller.
.IP
The child process then closes the other end file descriptor
and calls
.IR srv .
.IP
The parent process returns from the function with the
mountable file descriptor
.IR sfd .
On error,
.I postsrv
and
.I threadpostsrv
return a file descriptor of
.BR -1
with error string set.
.PP
.I Postmountsrv
and
.I threadpostmntsrv
call
.I postsrv
with
.I name
and then if
.I mtpt
is non-nil,
call
.BI amount( sfd ,
.IB mtpt ,
.IB flag ,
\fB"")\fR;
otherwise, close the file descriptor.
If any error occurs during
this process, the entire process is terminated by calling
.I sysfatal
(see
.IR perror (2)).
.PP
.I Postsharesrv
is similar to
.I Postmountsrv
but instead of mounting the service on a directory, it is
put in a share (see
.IR shr (3))
where
.IB mtpt
is the name of the share and
.IB desc
is the name of the service channel.
.PP
.I Listensrv
and
.I threadlistensrv
create a separate process to announce as
.IR addr .
The process listens for incoming connections,
creating a new process to serve each.
Using these functions results in 
.I srv
and the service functions
being run in multiple processes simultaneously.
The library locks its own data structures as necessary;
the client may need to lock data it shares between
the multiple connections.
.SS Service functions
The functions in a 
.B Srv
structure named after 9P transactions
are called to satisfy requests as they arrive.
If a function is provided, it
.I must
arrange for
.I respond
to be called when the request is satisfied.
The only parameter of each service function
is a 
.B Req*
parameter (say
.IR r ).
The incoming request parameters are stored in 
.IB r -> ifcall \fR;
.IB r -> fid
and
.IB r -> newfid
are pointers to 
.B Fid
structures corresponding to the
numeric fids in
.IB r -> ifcall \fR;
similarly, 
.IB r -> oldreq
is the
.B Req
structure corresponding to
.IB r -> ifcall.oldtag \fR.
The outgoing response data should be stored in 
.IB r -> ofcall \fR.
The one exception to this rule is that 
.I stat
should fill in 
.IB r -> d
rather than
.IB r -> ofcall.stat \fR:
the library will convert the structure into the machine-independent
wire representation.
Similarly, 
.I wstat
may consult
.IB r -> d
rather than decoding
.IB r -> ifcall . stat
itself.
When a request has been handled,
.I respond
should be called with
.I r
and an error string.
If the request was satisfied successfully, the error
string should be a nil pointer.
Note that it is permissible for a function to return
without itself calling 
.IR respond ,
as long as it has arranged for
.I respond
to be called at some point in the future
by another proc sharing its address space,
but see the discussion of
.I flush
below.
Once
.I respond
has been called, the 
.B Req*
as well as any pointers it once contained must
be considered freed and not referenced.
.PP
.I Responderror
calls 
.I respond
with the system error string
(see
.IR errstr (2)).
.PP
If the service loop detects an error in a request
(e.g., an attempt to reuse an extant fid, an open of
an already open fid, a read from a fid opened for write, etc.)
it will reply with an error without consulting
the service functions.
.PP
The service loop provided by
.I srv
(and indirectly by
.I postmountsrv
and
.IR threadpostmountsrv )
is single-threaded.
If it is expected that some requests might
block, arranging for alternate processes
to handle them is suggested (see
.IR 9pqueue (2)).
.PP
.I Srvrelease
temporarily releases the calling process from the server
loop and if neccesary spawns a new process to handle 9p 
requests. When released, the process can do blocking work
that would otherwise halt processing of 9p requests.
.I Srvacquire
rejoins the calling process with the server loop after
a srvrelease.
.PP
The constraints on the service functions are as follows.
These constraints are checked while the server executes.
If a service function fails to do something it ought to have,
.I srv
will call
.I endsrv
and then abort.
.TP 
.I Auth
If authentication is desired,
the
.I auth
function should record that
.IB r -> afid
is the new authentication fid and
set 
.IB r -> afid -> qid
and
.IR ofcall.qid .
.I Auth
may be nil, in which case it will be treated as having
responded with the error
.RI `` "argv0: authentication not required" ,''
where
.I argv0
is the program name variable as set by
.I ARGBEGIN
(see
.IR arg (2)).
.TP
.I Attach
The
.I attach
function should check the authentication state of
.I afid
if desired,
and set
.IB r -> fid -> qid
and
.I ofcall.qid
to the qid of the file system root.
.I Attach
may be nil only if file trees are in use;
in this case, the qid will be filled from the root
of the tree, and no authentication will be done.
.TP
.I Walk
If file trees are in use,
.I walk
is handled internally, and 
.IB srv -> walk
is never called.
.IP
If file trees are not in use,
.I walk
should consult
.IB r -> ifcall . wname
and
.IB r -> ifcall . nwname \fR,
filling in 
.IB ofcall . qid
and
.IB ofcall . nqid \fR,
and also copying any necessary 
.I aux
state from
.IB r -> fid
to
.IB r -> newfid
when the two are different.
As long as
.I walk 
sets
.IB ofcall . nqid
appropriately, it can
.I respond
with a nil error string even when 9P
demands an error 
.RI ( e.g. ,
in the case of a short walk);
the library detects error conditions and handles them appropriately.
.IP
Because implementing the full walk message is intricate and
prone to error, the helper routine
.I walkandclone
will handle the request given pointers to two functions
.I walk1
and (optionally)
.I clone .
.IR Clone ,
if non-nil, is called to signal the creation of
.I newfid
from
.IR oldfid .
Typically a 
.I clone
routine will copy or increment a reference count in
.IR oldfid 's
.I aux
element.
.I Walk1
should walk
.I fid
to
.IR name ,
initializing
.IB fid -> qid
to the new path's qid.
Both should return nil
on success or an error message on error.
.I Walkandclone
will call
.I respond
after handling the request.
.TP
.I Walk1\fR, \fPClone
If the client provides functions
.IB srv -> walk1
and (optionally)
.IB srv -> clone \fR,
the 9P service loop will call
.I walkandclone
with these functions to handle the request.
Unlike the
.I walk1
above,
.IB srv -> walk1
must fill in both
.IB fid -> qid
and
.BI * qid
with the new qid on a successful walk.
.TP
.I Open
If file trees are in use, the file
metadata will be consulted on open, create, remove, and wstat
to see if the requester has the appropriate permissions.
If not, an error will be sent back without consulting a service function.
.IP
If not using file trees or the user has the appropriate permissions,
.I open
is called with
.IB r -> ofcall . qid
already initialized to the one stored in the 
.B Fid
structure (that is, the one returned in the previous walk).
If the qid changes, both should be updated.
.TP
.I Create
The
.I create
function must fill in
both
.IB r -> fid -> qid
and
.IB r -> ofcall . qid
on success.
When using file trees,
.I create
should allocate a new 
.B File
with
.IR createfile ;
note that
.I createfile
may return nil (because, say, the file already exists).
If the 
.I create
function is nil,
.I srv 
behaves as though it were a function that always responded
with the error ``create prohibited''.
.TP
.I Remove
.I Remove
should mark the file as removed, whether
by calling
.I removefile
when using file trees, or by updating an internal data structure.
In general it is not a good idea to clean up the
.I aux
information associated with the corresponding
.B File
at this time, to avoid memory errors if other
fids have references to that file.
Instead, it is suggested that 
.I remove
simply mark the file as removed (so that further
operations on it know to fail) and wait until the
file tree's destroy function is called to reclaim the
.I aux
pointer.
If not using file trees, it is prudent to take the
analogous measures.
If
.I remove
is not provided, all remove requests will draw
``remove prohibited'' errors.
.TP
.I Read
The
.I read
function must be provided; it fills
.IB r -> ofcall . data
with at most
.IB r -> ifcall . count
bytes of data from offset
.IB r -> ifcall . offset
of the file.
It also sets
.IB r -> ofcall . count
to the number of bytes being returned.
If using file trees, 
.I srv
will handle reads of directories internally, only
calling
.I read
for requests on files.
.I Readstr
and
.I readbuf
are useful for satisfying read requests on a string or buffer.
Consulting the request in
.IB r -> ifcall \fR,
they fill
.IB r -> ofcall . data
and set
.IB r -> ofcall . count \fR;
they do not call
.IB respond .
Similarly,
.I dirread9p
can be used to handle directory reads in servers
not using file trees.
The passed
.I gen
function will be called as necessary to
fill
.I dir
with information for the
.IR n th
entry in the directory.
The string pointers placed in
.I dir
should be fresh copies
made with
.IR estrdup9p ;
they will be freed by
.I dirread9p
after each successful call to 
.IR gen .
.I Gen
should return zero if it successfully filled
.IR dir ,
minus one on end of directory.
.TP
.I Write
The 
.I write 
function is similar but need not be provided.
If it is not, all writes will draw 
``write prohibited'' errors.
Otherwise, 
.I write
should attempt to write the
.IB r -> ifcall . count
bytes of 
.IB r -> ifcall . data
to offset
.IB r -> ifcall . offset
of the file, setting
.IB r -> ofcall . count
to the number of bytes actually written.
Most programs consider it an error to
write less than the requested amount.
.TP
.I Stat
.I Stat
should fill
.IB r -> d
with the stat information for
.IB r -> fid \fR.
If using file trees, 
.IB r -> d
will have been initialized with the stat info from 
the tree, and
.I stat
itself may be nil.
.TP
.I Wstat
The
.I wstat
function consults 
.IB r -> d
in changing the metadata for
.IB r -> fid
as described in
.IR stat (5).
When using file trees,
.I srv
will take care to check that the request satisfies
the permissions outlined in
.IR stat (5).
Otherwise 
.I wstat
should take care to enforce permissions
where appropriate.
.TP
.I Flush
Servers that always call 
.I respond
before returning from the service functions
need not provide a 
.I flush
implementation:
.I flush
is only necessary in programs
that arrange for 
.I respond
to be called asynchronously.
.I Flush
should cause the request
.IB r -> oldreq
to be cancelled or hurried along.
If
.I oldreq
is cancelled, this should be signalled by calling
.I respond
on
.I oldreq
with error string
.RB ` interrupted '.
.I Flush
must respond to
.I r
with a nil error string.
.I Flush
may respond to
.I r
before forcing a response to
.IB r -> oldreq \fR.
In this case, the library will delay sending
the
.I Rflush
message until the response to 
.IB r -> oldreq
has been sent.
.PD
.PP
.IR Destroyfid ,
.IR destroyreq ,
.IR start ,
and
.I end
are auxiliary functions, not called in direct response to 9P requests.
.TP
.I Destroyfid
When a 
.BR Fid 's
reference count drops to zero
.RI ( i.e., 
it has been clunked and there are no outstanding
requests referring to it),
.I destroyfid
is called to allow the program to dispose
of the
.IB fid -> aux
pointer.
.TP
.I Destroyreq
Similarly, when a
.BR Req 's
reference count drops to zero
.RI ( i.e. ,
it has been handled via
.I respond
and other outstanding pointers to it have been closed),
.I destroyreq
is called to allow the program to dispose of the
.IB r -> aux
pointer.
.TP
.I Start
This gets called (from the forked service process) 
prior entering the 9P service loop.
.TP
.I End
Once the 9P service loop has finished
(end of file been reached on the service pipe
or a bad message has been read),
.I end
is called (if provided) to allow any final cleanup.
For example, it was used by the Palm Pilot synchronization
file system (never finished) to gracefully terminate the serial conversation once
the file system had been unmounted.
After calling
.IR end ,
the service loop (which runs in a separate process
from its caller) terminates using 
.I _exits
(see
.IR exits (2)).
.PD
.PP
.I Auth9p,
.I authattach,
.I authwrite,
and
.I authread
are functions to assist in the creation of authenticated
file servers.
.I Auth9p
sets
.I afid
to point to
.B /mnt/factotum/rpc
then calls
.I respond.
It can be used as the
.I auth
service function address.
.I Authattach
will return 0 if the session has properly
authenticated or will return -1 and
call
.I respond
with an apropiate error.
.I Authread
and
.I authwrite
can be called from the
.I read
and
.I write
service functions respectfully to handle
calls on the
.I afid.
.PP
If the 
.B chatty9p
flag is at least one,
a transcript of the 9P session is printed
on standard error.
If the
.B chatty9p
flag is greater than one,
additional unspecified debugging output is generated.
By convention, servers written using this library
accept the
.B -D
option to increment
.BR chatty9p .
.SH EXAMPLES
.IR Archfs (4),
.IR cdfs (4),
.IR nntpfs (4),
.IR snap (4),
and
.B /sys/src/lib9p/ramfs.c
are good examples of simple single-threaded file servers.
.PP
In general, the
.B File
interface is appropriate for maintaining arbitrary file trees (as in
.IR ramfs ).
The 
.B File
interface is best avoided when the 
tree structure is easily generated as necessary;
this is true when the tree is highly structured (as in
.I cdfs
and
.IR nntpfs )
or is maintained elsewhere.
.SH SOURCE
.B /sys/src/lib9p
.SH SEE ALSO
.IR 9pfid (2),
.IR 9pfile (2),
.IR 9pqueue (2),
.IR srv (3),
.IR shr (3),
.IR intro (5)
.SH BUGS
The switch to 9P2000 was taken as an opportunity to tidy
much of the interface; we promise to avoid such gratuitous change
in the future.