plan9fox/sys/man/2/usbfs
2011-03-30 19:35:09 +03:00

341 lines
6.9 KiB
Text

.TH USBFS 2
.SH NAME
usbreadbuf,
usbfsadd,
usbfsdel,
usbdirread,
usbfsinit,
usbdirfs,
usbfs \- USB device driver file system library
.SH SYNOPSIS
.EX
.ta 8n +8n +8n +8n +8n +8n +8n
#include <u.h>
#include <libc.h>
#include <thread.h>
#include "../lib/usb.h"
#include "../lib/usbfs.h"
.sp 0.3v
enum {
Hdrsize = 128, /* plenty of room for headers */
Msgsize = 8 * 1024,
Bufsize = Hdrsize + Msgsize,
Namesz = 40,
Errmax = 128,
ONONE = ~0, /* omode in Fid when not open */
};
.sp 0.3v
struct Fid {
int fid;
Qid qid;
int omode;
Fid* next;
void* aux;
};
.sp 0.3v
struct Usbfs {
char name[Namesz];
uvlong qid;
Dev* dev;
void* aux;
.sp 0.3v
int (*walk)(Usbfs *fs, Fid *f, char *name);
void (*clone)(Usbfs *fs, Fid *of, Fid *nf);
void (*clunk)(Usbfs *fs, Fid *f);
int (*open)(Usbfs *fs, Fid *f, int mode);
long (*read)(Usbfs *fs, Fid *f,
void *data, long count, vlong offset);
long (*write)(Usbfs *fs, Fid*f,
void *data, long count, vlong offset);
int (*stat)(Usbfs *fs, Qid q, Dir *d);
void (*end)(Usbfs *fs);
};
.sp 0.3v
typedef int (*Dirgen)(Usbfs*, Qid, int, Dir*, void*);
.sp 0.3v
long usbreadbuf(void *data, long count,
vlong offset, void *buf, long n);
void usbfsadd(Usbfs *dfs);
void usbfsdel(Usbfs *dfs);
int usbdirread(Usbfs*f, Qid q, char *data, long cnt,
vlong off, Dirgen gen, void *arg);
void usbfsinit(char* srv, char *mnt, Usbfs *f, int flag);
void usbfsdirdump(void);
.sp 0.3v
extern char Enotfound[], Etoosmall[], Eio[], Eperm[], Ebadcall[],
Ebadfid[], Einuse[], Eisopen[], Ebadctl[];
.sp 0.3v
extern Usbfs usbdirfs;
extern int usbfsdebug;
.EE
.SH DESCRIPTION
This library provides an alternative to
.IR 9p (2)
for implementing a file server within a USB driver.
Drivers using this
library may be embedded into
.IR usbd (4).
It may be also desirable to use this library when drivers are
not embedded because it is tailored to work well with the
library for handling USB devices.
.PP
A USB file system is described by a
.I Usbfs
structure.
In most cases, the driver is not responsible for the root of the
file tree.
It is customary that a driver creates a file server
for each device handled and links all of them to a root directory
implemented by the
.I usbdirfs
file system implemented by the library.
This root directory is bound to
.B /dev
in most cases.
.PP
.I Usbdirfs
implements a root directory populated by named file trees,
each one described by a
.B Usbfs
structure.
.PP
The field
.B Usbfs.name
contains the name for the root directory of the file system, usually
a directory seen at
.BI /dev/ name
when the driver is embedded.
.PP
.B Usbfs.qid
maintains a value used to decorate qids for the file tree.
This may be ignored when
.I usbdirfs
is not used.
Otherwise,
.I usbdirfs
assigns a unique value kept at the high 32 bits of
.B Qid.path
for all files on each file tree bound to it.
Each
.I Usbfs
server must bitwise OR
.B Usbfs.qid
to all
.B Qid.path
values returned by its functions.
In the same way,
functions usually clear bits in
.B Usbfs.qid
before processing
.B Qid.path
values supplied as input.
.PP
The USB device handled by a file tree is referenced from
.B Usbfs.dev
(and a reference must be counted for it).
This permits the following functions to quickly locate the device of
interest, and also permits releasing the device when
no request is outstanding.
.PP
The field
.B Usbfs.aux
is for the device to use.
The rest of the fields implement the 9P protocol for the device.
Not all the operations need be implemented.
Only
.IR walk ,
.IR open ,
.IR read ,
.IR write ,
and
.IR stat ,
must be implemented (and their corresponding fields in
.B Usbfs
may never be
.BR nil ).
These functions must return -1 upon failure
and set the error string to reflect the cause of a failure.
.PP
In all the functions, a 9P fid is represented by a
.B Fid
structure.
It contains the 9P
.IR fid ,
the corresponding
.IR qid ,
and an auxiliary pointer for the driver to use.
Open
.IR fid s
have a valid open mode in
.I omode
while others have
.B ONONE
to indicate that the
.I fid
is not open.
The library takes care of which
fids
exist and which ones do not.
.PP
.I Walk
must walk
.I f
to
.I name
(a single name, not a file path)
in the supplied
.IR fs .
Its implementation should update the qid in
.I f
to reflect the walk.
This function must bitwise OR any returned Qid with
.B Usbfs.qid ,
if
.I usbdirfs
is used.
.PP
.I Clone
must clone fid
.I of
onto
.I nf
so that,
upon successful completion,
.I nf
also refers to the file that
.I f
refers to.
An implementation must update the Qid of the cloned
fid.
If this function is not supplied, the library copies the
.I aux
field to the cloned fid.
.PP
.I Clunk
clunks
.IR f .
It usually releases data kept in the
.I aux
field, but may be
set to
.B nil
otherwise.
.PP
.I Open
prepares the fid
.I f
for I/O according to
.IR mode .
The open mode in the fid is updated by the library upon return.
The library checks trivial cases like opening already-open fids.
The implementation performs most permission checking.
.PP
.I Read
reads up to
.I count
bytes into
.I data
starting at
.I offset
in the file referenced by
.IR f .
.I Write
is the counterpart.
To read from directories,
the function
.I usbdirread
may be called.
It returns the return value of
.I read
or -1.
.I usbdirread
calls
.I gen
to iterate through files as needed.
The
.B Dirgen
function will be called with index values of 0
and up to ask for the first file and following files.
To read from data already in buffers, the function
.I usbreadbuf
may help.
It must be given the arguments supplied
by the user, plus the buffer and buffer size.
.PP
.I Stat
must fill
.I d
with the directory entry for the file identified by
.IR q.
As an aid,
.I d
is initialized to fake access and modification times,
and user and group ids.
Also, the field
.B name
in
.I d
is initialized to point to a 40-byte buffer.
If the file name fits,
it may be copied directly into
.B d->name
without allocating memory for that purpose.
Otherwise
.B d->name
must be initialized to point to static memory.
.PP
The function
.I end
is called upon termination of the file tree to
release resources.
.PP
Calling
.I usbfsinit
starts a file server for
.I f
that mounts itself at
.I mnt
and posts
.I srv
at
.IR srv (3).
In most cases, the file system supplied is
.IR usbdirfs .
The
.I flag
is used for
.IR mount
(see
.IR bind (2)).
Once
.I usbdirfs
is started, calls to
.IR usbfsadd
add a file tree implemented by
.I dfs
to the root directory of
.I usbdirfs
and
calls to
.I usbfsdel
remove that binding (and release resources including
the reference to the USB device).
.PP
Various error strings are declared as an aid.
The global
.B usbfsdebug
may be set to trigger diagnostics and protocol tracing.
.SH EXAMPLE
See
.B /sys/src/cmd/usb/disk
for an example driver that uses this library.
Looking at an example is strongly suggested
to see how reference counts for the USB device
and the file system are handled.
.SH SOURCE
.B /sys/src/cmd/usb/lib
.SH "SEE ALSO"
.IR usb (2),
.IR usb (3),
.IR usb (4),
.IR usbd (4)