+=head3 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.
+
+=head3 release
+
+Arguments: Pathname, numeric flags passed to open, file handle
+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.
+
+=head3 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.
+
+=head3 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';
+
+=head3 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.
+
+=head3 listxattr
+
+Arguments: Pathname
+Returns a list: 0 or more text strings (the extended attribute names), followed by a numeric errno (usually 0).
+
+=head3 removexattr
+
+Arguments: Pathname, extended attribute's name
+Returns an errno or 0 on success.
+
+=head3 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.
+
+=head3 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.
+
+=head3 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.
+
+=head3 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.
+
+=head3 init
+
+Arguments: None.
+Returns (optionally) an SV to be passed as private_data via fuse_get_context().
+
+=head3 destroy
+
+Arguments: (optional) private data SV returned from init(), if any.
+Returns nothing.
+
+=head3 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.
+
+=head3 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.
+
+=head3 ftruncate
+
+Arguments: Pathname, numeric offset, (optional) file handle
+Returns errno or 0 on success
+
+Like truncate(), but on an opened file.
+
+=head3 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.
+
+=head3 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.
+
+=head3 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. Currently
+times are passed as "numeric" (internally I believe these are represented
+typically as "long double"), so the sub-second portion is represented as
+fractions of a second.
+
+Note that if this call is implemented, it overrides utime() ALWAYS.
+
+=head3 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.
+