NAME
Fuse - write filesystems in Perl using FUSE
SYNOPSIS
use Fuse;
my ($mountpoint) = "";
$mountpoint = shift(@ARGV) if @ARGV;
Fuse::main(mountpoint=>$mountpoint, getattr=>"main::my_getattr", getdir=>"main::my_getdir", ...);
DESCRIPTION
This lets you implement filesystems in perl, through the FUSE (Filesystem in USErspace) kernel/lib interface.
FUSE expects you to implement callbacks for the various functions.
In the following definitions, "errno" can be 0 (for a success), -EINVAL, -ENOENT, -EONFIRE, any integer less than 1 really.
You can import standard error constants by saying something like "use POSIX qw(EDOTDOT ENOANO);".
Every constant you need (file types, open() flags, error values, etc) can be imported either from POSIX or from Fcntl, often both. See their respective documentations, for more information.
EXPORTED SYMBOLS
None by default.
You can request all exportable symbols by using the tag ":all".
You can request the extended attribute symbols by using the tag ":xattr". This will export XATTR_CREATE and XATTR_REPLACE.
FUNCTIONS
Fuse::main
Takes arguments in the form of hash key=>value pairs. There are many valid keys. Most of them correspond with names of callback functions, as described in section 'FUNCTIONS YOUR FILESYSTEM MAY IMPLEMENT'. A few special keys also exist:
- debug => boolean
-
This turns FUSE call tracing on and off. Default is 0 (which means off).
- mountpoint => string
-
The point at which to mount this filesystem. There is no default, you must specify this. An example would be '/mnt'.
- mountopts => string
-
This is a comma separated list of mount options to pass to the FUSE kernel module.
At present, it allows the specification of the allow_other argument when mounting the new FUSE filesystem. To use this, you will also need 'user_allow_other' in /etc/fuse.conf as per the FUSE documention
mountopts => "allow_other" or mountopts => ""
- threaded => boolean
-
This turns FUSE multithreading on and off. The default is 0, meaning your FUSE script will run in single-threaded mode. Note that single-threaded mode also means that you will not have to worry about reentrancy, though you will have to worry about recursive lookups. In single-threaded mode, FUSE holds a global lock on your filesystem, and will wait for one callback to return before calling another. This can lead to deadlocks, if your script makes any attempt to access files or directories in the filesystem it is providing. (This includes calling stat() on the mount-point, statfs() calls from the 'df' command, and so on and so forth.) It is worth paying a little attention and being careful about this.
Enabling multithreading will cause FUSE to make multiple simultaneous calls into the various callback functions of your perl script. If you enable threaded mode, you can enjoy all the parallel execution and interactive response benefits of threads, and you get to enjoy all the benefits of race conditions and locking bugs, too. Please also ensure any other perl modules you're using are also thread-safe.
(If enabled, this option will cause a warning if your perl interpreter was not built with USE_ITHREADS, or if you have failed to use threads or threads::shared.)
- nullpath_ok => boolean
-
This flag tells Fuse to not pass paths for functions that operate on file or directory handles. This will yield empty path parameters for functions including read, write, flush, release, fsync, readdir, releasedir, fsyncdir, truncate, fgetattr and lock. If you use this, you must return file/directory handles from open, opendir and create. Default is 0 (off). Only effective on Fuse 2.8 and up; with earlier versions, this does nothing.
- utimens_as_array => boolean
-
This flag causes timestamps passed via the utimens() call to be passed as arrays containing the time in seconds, and a second value containing the number of nanoseconds, instead of a floating point value. This allows for more precise times, as the normal floating point type used by Perl (double) loses accuracy starting at about tenths of a microsecond.
- nopath => boolean
-
Flag indicating that the path need not be calculated for the following operations:
read, write, flush, release, fsync, readdir, releasedir, fsyncdir, ftruncate, fgetattr, lock, ioctl and poll
Closely related to nullpath_ok, but if this flag is set then the path will not be calculated even if the file wasn't unlinked. However the path can still be defined if it needs to be calculated for some other reason.
Only effective on Fuse 2.9 and up.
- utime_omit_ok => boolean
-
Flag indicating that the filesystem accepts special UTIME_NOW and UTIME_OMIT values in its
utimens
operation.If you wish to use these constants, make sure to include the ':utime' flag when including the Fuse module, or the ':all' flag.
Only effective on Fuse 2.9 and up.
Fuse::fuse_get_context
use Fuse "fuse_get_context";
my $caller_uid = fuse_get_context()->{"uid"};
my $caller_gid = fuse_get_context()->{"gid"};
my $caller_pid = fuse_get_context()->{"pid"};
Access context information about the current Fuse operation.
Fuse::fuse_version
Indicates the Fuse version in use; more accurately, indicates the version of the Fuse API in use at build time. If called in scalar context, the version will be returned as a decimal value; i.e., for Fuse API v2.6, will return "2.6". If called in array context, an array will be returned, containing the major, minor and micro version numbers of the Fuse API it was built against.
Fuse::fuse_buf_size
Computes the total size of a buffer vector. Applicable for read_buf
and write_buf
operations.
Fuse::fuse_buf_copy
Copies data from one buffer vector to another. Primarily useful if a buffer vector contains multiple, fragmented chunks or if it contains an FD buffer instead of a memory buffer. Applicable for write_buf
.
Fuse::notify_poll
Only available if the Fuse module is built against libfuse 2.8 or later. Use fuse_version() to determine if this is the case. Calling this function with a pollhandle argument (as provided to the poll
operation implementation) will send a notification to the caller poll()ing for I/O operation availability. If more than one pollhandle is provided for the same filehandle, only use the latest; you *can* send notifications to them all, but it is unnecessary and decreases performance.
ONLY supply poll handles fed to you through poll
to this function. Due to thread safety requirements, we can't currently package the pointer up in an object the way we'd like to to prevent this situation, but your filesystem server program may segfault, or worse, if you feed things to this function which it is not supposed to receive. If you do anyway, we take no responsibility for whatever Bad Things(tm) may happen.
Fuse::pollhandle_destroy
Only available if the Fuse module is built against libfuse 2.8 or later. Use fuse_version() to determine if this is the case. This function destroys a poll handle (fed to your program through poll
). When you are done with a poll handle, either because it has been replaced, or because a notification has been sent to it, pass it to this function to dispose of it safely.
ONLY supply poll handles fed to you through poll
to this function. Due to thread safety requirements, we can't currently package the pointer up in an object the way we'd like to to prevent this situation, but your filesystem server program may segfault, or worse, if you feed things to this function which it is not supposed to receive. If you do anyway, we take no responsibility for whatever Bad Things(tm) may happen.
FUNCTIONS YOUR FILESYSTEM MAY IMPLEMENT
getattr
Arguments: filename.
Returns a list, very similar to the 'stat' function (see perlfunc). On error, simply return a single numeric scalar value (e.g. "return -ENOENT();").
FIXME: the "ino" field is currently ignored. I tried setting it to 0 in an example script, which consistently caused segfaults.
Fields (the following was stolen from perlfunc(1) with apologies):
($dev,$ino,$mode,$nlink,$uid,$gid,$rdev,$size, $atime,$mtime,$ctime,$blksize,$blocks) = getattr($filename);
Here are the meaning of the fields:
0 dev device number of filesystem
1 ino inode number
2 mode file mode (type and permissions)
3 nlink number of (hard) links to the file
4 uid numeric user ID of file's owner
5 gid numeric group ID of file's owner
6 rdev the device identifier (special files only)
7 size total size of file, in bytes
8 atime last access time in seconds since the epoch
9 mtime last modify time in seconds since the epoch
10 ctime inode change time (NOT creation time!) in seconds
since the epoch
11 blksize preferred block size for file system I/O
12 blocks actual number of blocks allocated
(The epoch was at 00:00 January 1, 1970 GMT.)
If you wish to provide sub-second precision timestamps, they may be passed either as the fractional part of a floating-point value, or as a two-element array, passed as an array ref, with the first element containing the number of seconds since the epoch, and the second containing the number of nanoseconds. This provides complete time precision, as a floating point number starts losing precision at about a tenth of a microsecond. So if you really care about that sort of thing...
readlink
Arguments: link pathname.
Returns a scalar: either a numeric constant, or a text string.
This is called when dereferencing symbolic links, to learn the target.
example rv: return "/proc/self/fd/stdin";
getdir
Arguments: Containing directory name.
Returns a list: 0 or more text strings (the filenames), followed by a numeric errno (usually 0).
This is used to obtain directory listings. It's opendir(), readdir(), filldir() and closedir() all in one call.
example rv: return ('.', 'a', 'b', 0);
mknod
Arguments: Filename, numeric modes, numeric device
Returns an errno (0 upon success, as usual).
This function is called for all non-directory, non-symlink nodes, not just devices.
mkdir
Arguments: New directory pathname, numeric modes.
Returns an errno.
Called to create a directory.
unlink
Arguments: Filename.
Returns an errno.
Called to remove a file, device, or symlink.
rmdir
Arguments: Pathname.
Returns an errno.
Called to remove a directory.
symlink
Arguments: Existing filename, symlink name.
Returns an errno.
Called to create a symbolic link.
rename
Arguments: old filename, new filename.
Returns an errno.
Called to rename a file, and/or move a file from one directory to another.
link
Arguments: Existing filename, hardlink name.
Returns an errno.
Called to create hard links.
chmod
Arguments: Pathname, numeric modes.
Returns an errno.
Called to change permissions on a file/directory/device/symlink.
chown
Arguments: Pathname, numeric uid, numeric gid.
Returns an errno.
Called to change ownership of a file/directory/device/symlink.
truncate
Arguments: Pathname, numeric offset.
Returns an errno.
Called to truncate a file, at the given offset.
utime
Arguments: Pathname, numeric actime, numeric modtime.
Returns an errno.
Called to change access/modification times for a file/directory/device/symlink.
open
Arguments: Pathname, numeric flags (which is an OR-ing of stuff like O_RDONLY and O_SYNC, constants you can import from POSIX), fileinfo hash reference.
Returns an errno, a file handle (optional).
No creation, or truncation flags (O_CREAT, O_EXCL, O_TRUNC) will be passed to open(). The fileinfo hash reference contains flags from the Fuse open call which may be modified by the module. The only fields presently supported are: direct_io (version 2.4 onwards) keep_cache (version 2.4 onwards) nonseekable (version 2.8 onwards) Your open() method needs only check if the operation is permitted for the given flags, and return 0 for success. Optionally a file handle may be returned, which will be passed to subsequent read, write, flush, fsync and release calls.
read
Arguments: Pathname, numeric requested size, numeric offset, file handle
Returns a numeric errno, or a string scalar with up to $requestedsize bytes of data.
Called in an attempt to fetch a portion of the file.
write
Arguments: Pathname, scalar buffer, numeric offset, file handle. You can use length($buffer) to find the buffersize. Returns length($buffer) if successful (number of bytes written).
Called in an attempt to write (or overwrite) a portion of the file. Be prepared because $buffer could contain random binary data with NULs and all sorts of other wonderful stuff.
statfs
Arguments: none
Returns any of the following:
-ENOANO()
or
$namelen, $files, $files_free, $blocks, $blocks_avail, $blocksize
or
-ENOANO(), $namelen, $files, $files_free, $blocks, $blocks_avail, $blocksize
flush
Arguments: Pathname, file handle
Returns an errno or 0 on success.
Called to synchronise any cached data. This is called before the file is closed. It may be called multiple times before a file is closed.
release
Arguments: Pathname, numeric flags passed to open, file handle, flock_release flag (when built against FUSE 2.9 or later), lock owner ID (when built against FUSE 2.9 or later)
Returns an errno or 0 on success.
Called to indicate that there are no more references to the file. Called once for every file with the same pathname and flags as were passed to open.
fsync
Arguments: Pathname, numeric flags
Returns an errno or 0 on success.
Called to synchronise the file's contents. If flags is non-zero, only synchronise the user data. Otherwise synchronise the user and meta data.
setxattr
Arguments: Pathname, extended attribute's name, extended attribute's value, numeric flags (which is an OR-ing of XATTR_CREATE and XATTR_REPLACE
Returns an errno or 0 on success.
Called to set the value of the named extended attribute.
If you wish to reject setting of a particular form of extended attribute name (e.g.: regexps matching user\..* or security\..*), then return - EOPNOTSUPP.
If flags is set to XATTR_CREATE and the extended attribute already exists, this should fail with - EEXIST. If flags is set to XATTR_REPLACE and the extended attribute doesn't exist, this should fail with - ENOATTR.
XATTR_CREATE and XATTR_REPLACE are provided by this module, but not exported by default. To import them:
use Fuse ':xattr';
or:
use Fuse ':all';
getxattr
Arguments: Pathname, extended attribute's name
Returns an errno, 0 if there was no value, or the extended attribute's value.
Called to get the value of the named extended attribute.
listxattr
Arguments: Pathname
Returns a list: 0 or more text strings (the extended attribute names), followed by a numeric errno (usually 0).
removexattr
Arguments: Pathname, extended attribute's name
Returns an errno or 0 on success.
Removes the named extended attribute (if present) from a file.
opendir
Arguments: Pathname of a directory Returns an errno, and a directory handle (optional)
Called when opening a directory for reading. If special handling is required to open a directory, this operation can be implemented to handle that.
readdir
Arguments: Pathname of a directory, numeric offset, (optional) directory handle
Returns a list of 0 or more entries, followed by a numeric errno (usually 0). The entries can be simple strings (filenames), or arrays containing an offset number, the filename, and optionally an array ref containing the stat values (as would be returned from getattr()).
This is used to read entries from a directory. It can be used to return just entry names like getdir(), or can get a segment of the available entries, which requires using array refs and the 2- or 3-item form, with offset values starting from 1. If you wish to return the parameters to fill each entry's struct stat, but do not wish to do partial entry lists/entry counting, set the first element of each array to 0 always.
Note that if this call is implemented, it overrides getdir() ALWAYS.
releasedir
Arguments: Pathname of a directory, (optional) directory handle
Returns an errno or 0 on success
Called when there are no more references to an opened directory. Called once for each pathname or handle passed to opendir(). Similar to release(), but for directories. Accepts a return value, but like release(), the response code will not propagate to any corresponding closedir() calls.
fsyncdir
Arguments: Pathname of a directory, numeric flags, (optional) directory handle
Returns an errno or 0 on success.
Called to synchronize any changes to a directory's contents. If flag is non-zero, only synchronize user data, otherwise synchronize user data and metadata.
init
Arguments: None.
Returns (optionally) an SV to be passed as private_data via fuse_get_context().
destroy
Arguments: (optional) private data SV returned from init(), if any.
Returns nothing.
access
Arguments: Pathname, access mode flags
Returns an errno or 0 on success.
Determine if the user attempting to access the indicated file has access to perform the requested actions. The user ID can be determined by calling fuse_get_context(). See access(2) for more information.
create
Arguments: Pathname, create mask, open mode flags
Returns errno or 0 on success, and (optional) file handle.
Create a file with the path indicated, then open a handle for reading and/or writing with the supplied mode flags. Can also return a file handle like open() as part of the call.
ftruncate
Arguments: Pathname, numeric offset, (optional) file handle
Returns errno or 0 on success
Like truncate(), but on an opened file.
fgetattr
Arguments: Pathname, (optional) file handle
Returns a list, very similar to the 'stat' function (see perlfunc). On error, simply return a single numeric scalar value (e.g. "return -ENOENT();").
Like getattr(), but on an opened file.
lock
Arguments: Pathname, numeric command code, hashref containing lock parameters, (optional) file handle
Returns errno or 0 on success
Used to lock or unlock regions of a file. Locking is handled locally, but this allows (especially for networked file systems) for protocol-level locking semantics to also be employed, if any are available.
See the Fuse documentation for more explanation of lock(). The needed symbols for the lock constants can be obtained by importing Fcntl.
utimens
Arguments: Pathname, last accessed time, last modified time
Returns errno or 0 on success
Like utime(), but allows time resolution down to the nanosecond. By default, times are passed as "numeric" (internally these are typically represented as "double"), so the sub-second portion is represented as fractions of a second. If you want times passed as arrays instead of floating point values, for higher precision, you should pass the utimens_as_array
option to Fuse::main
.
Note that if this call is implemented, it overrides utime() ALWAYS.
bmap
Arguments: Pathname, numeric blocksize, numeric block number
Returns errno or 0 on success, and physical block number if successful
Used to map a block number offset in a file to the physical block offset on the block device backing the file system. This is intended for filesystems that are stored on an actual block device, with the 'blkdev' option passed.
ioctl
Arguments: Pathname, ioctl command code, flags, data if ioctl op is a write, (optional) file handle
Returns errno or 0 on success, and data if ioctl op is a read
Used to handle ioctl() operations on files. See ioctl(2) for more information on the fine details of ioctl operation numbers. May need to h2ph system headers to get the necessary macros; keep in mind the macros are highly OS-dependent.
Keep in mind that read and write are from the client perspective, so read from our end means data is going *out*, and write means data is coming *in*. It can be slightly confusing.
poll
Arguments: Pathname, poll handle ID (or undef if none), event mask, (optional) file handle
Returns errno or 0 on success, and updated event mask on success
Used to handle poll() operations on files. See poll(2) to learn more about event polling. Use IO::Poll to get the POLLIN, POLLOUT, and other symbols to describe the events which can happen on the filehandle. Save the poll handle ID to be passed to notify_poll
and pollhandle_destroy
functions, if it is not undef. Threading will likely be necessary for this operation to work.
There is not an "out of band" data transfer channel provided as part of FUSE, so POLLPRI/POLLRDBAND/POLLWRBAND won't work.
Poll handle is currently a read-only scalar; we are investigating a way to make this an object instead.
write_buf
Arguments: Pathname, offset, buffer vector, (optional) file handle.
Write contents of buffer to an open file.
Similar to the write
method, but data is supplied in a generic buffer. Use fuse_buf_copy() to transfer data to the destination if necessary.
read_buf
Arguments: Pathname, size, offset, buffer vector, (optional) file handle.
Store data from an open file in a buffer.
Similar to the read
method, but data is stored and returned in a generic buffer.
No actual copying of data has to take place, the source file descriptor may simply be placed in the 'fd' member of the buffer access hash (and the 'flags' member OR'd with FUSE_BUF_IS_FD) for later retrieval.
Also, if the FUSE_BUF_FD_SEEK constant is OR'd with 'flags', the 'pos' member should contain the offset (in bytes) to seek to in the file descriptor.
If data is to be read, the read data should be placed in the 'mem' member of the buffer access hash, and the 'size' member should be updated if less data was read than requested.
flock
Arguments: pathname, (optional) file handle, unique lock owner ID, operation ID
Perform BSD-style file locking operations.
Operation ID will be one of LOCK_SH, LOCK_EX or LOCK_UN. Non-blocking lock requests will be indicated by having LOCK_NB OR'd into the value.
For more information, see the flock(2) manpage. For the lock symbols, do:
use Fcntl qw(flock);
Locking is handled locally, but this allows (especially for networked file systems) for protocol-level locking semantics to also be employed, if any are available.
fallocate
Arguments: pathname, (optional) file handle, mode, offset, length
Allocates space for an open file.
This function ensures that required space is allocated for specified file. If this function returns success then any subsequent write request to specified range is guaranteed not to fail because of lack of space on the file system media.
EXAMPLES
There are a few example scripts in the examples/ subdirectory. These are:
example.pl
A simple "Hello world" type of script
loopback.pl
A filesystem loopback-device. like fusexmp from the main FUSE dist,
it simply recurses file operations into the real filesystem. Unlike
fusexmp, it only re-shares files under the /tmp/test directory.
rmount.pl
An NFS-workalike which tunnels through SSH. It requires an account
on some ssh server (obviously), with public-key authentication enabled.
(if you have to type in a password, you don't have this. man ssh_keygen.).
Copy rmount_remote.pl to your home directory on the remote machine
and make it executable. Then create a mountpoint subdir somewhere local,
and run the example script: ./rmount.pl host /remote/dir /local/dir
rmount_remote.pl
A ripoff of loopback.pl meant to be used as a backend for rmount.pl.
AUTHOR
Mark Glines, <mark@glines.org>
SEE ALSO
perl, the FUSE documentation.