2 FUSE: Filesystem in Userspace
3 Copyright (C) 2001-2008 Miklos Szeredi <miklos@szeredi.hu>
5 This program can be distributed under the terms of the GNU GPL.
12 #include <linux/fuse.h>
14 #include <linux/mount.h>
15 #include <linux/wait.h>
16 #include <linux/list.h>
17 #include <linux/spinlock.h>
19 #include <linux/backing-dev.h>
20 #include <linux/mutex.h>
21 #include <linux/rwsem.h>
22 #include <linux/rbtree.h>
23 #include <linux/poll.h>
25 /** Max number of pages that can be used in a single read request */
26 #define FUSE_MAX_PAGES_PER_REQ 32
28 /** Bias for fi->writectr, meaning new writepages must not be sent */
29 #define FUSE_NOWRITE INT_MIN
31 /** It could be as large as PATH_MAX, but would that have any uses? */
32 #define FUSE_NAME_MAX 1024
34 /** Number of dentries for each connection in the control filesystem */
35 #define FUSE_CTL_NUM_DENTRIES 5
37 /** If the FUSE_DEFAULT_PERMISSIONS flag is given, the filesystem
38 module will check permissions based on the file mode. Otherwise no
39 permission checking is done in the kernel */
40 #define FUSE_DEFAULT_PERMISSIONS (1 << 0)
42 /** If the FUSE_ALLOW_OTHER flag is given, then not only the user
43 doing the mount will be allowed to access the filesystem */
44 #define FUSE_ALLOW_OTHER (1 << 1)
46 /** List of active connections */
47 extern struct list_head fuse_conn_list
;
49 /** Global mutex protecting fuse_conn_list and the control filesystem */
50 extern struct mutex fuse_mutex
;
52 /** Module parameters */
53 extern unsigned max_user_bgreq
;
54 extern unsigned max_user_congthresh
;
61 /** Unique ID, which identifies the inode between userspace
65 /** Number of lookups on this inode */
68 /** The request used for sending the FORGET message */
69 struct fuse_req
*forget_req
;
71 /** Time in jiffies until the file attributes are valid */
74 /** The sticky bit in inode->i_mode may have been removed, so
75 preserve the original mode */
78 /** Version of last attribute change */
81 /** Files usable in writepage. Protected by fc->lock */
82 struct list_head write_files
;
84 /** Writepages pending on truncate or fsync */
85 struct list_head queued_writes
;
87 /** Number of sent writes, a negative bias (FUSE_NOWRITE)
88 * means more writes are blocked */
91 /** Waitq for writepage completion */
92 wait_queue_head_t page_waitq
;
94 /** List of writepage requestst (pending or sent) */
95 struct list_head writepages
;
100 /** FUSE specific file data */
102 /** Fuse connection for this file */
103 struct fuse_conn
*fc
;
105 /** Request reserved for flush and release */
106 struct fuse_req
*reserved_req
;
108 /** Kernel file handle guaranteed to be unique */
111 /** File handle used by userspace */
114 /** Node id of this file */
120 /** FOPEN_* flags returned by open */
123 /** Entry on inode's write_files list */
124 struct list_head write_entry
;
126 /** RB node to be linked on fuse_conn->polled_files */
127 struct rb_node polled_node
;
129 /** Wait queue head for poll */
130 wait_queue_head_t poll_wait
;
133 /** One input argument of a request */
139 /** The request input */
141 /** The request header */
142 struct fuse_in_header h
;
144 /** True if the data for the last argument is in req->pages */
147 /** Number of arguments */
150 /** Array of arguments */
151 struct fuse_in_arg args
[3];
154 /** One output argument of a request */
160 /** The request output */
162 /** Header returned from userspace */
163 struct fuse_out_header h
;
166 * The following bitfields are not changed during the request
170 /** Last argument is variable length (can be shorter than
174 /** Last argument is a list of pages to copy data to */
177 /** Zero partially or not copied pages */
178 unsigned page_zeroing
:1;
180 /** Number or arguments */
183 /** Array of arguments */
184 struct fuse_arg args
[3];
187 /** The request state */
188 enum fuse_req_state
{
198 * A request to the client
201 /** This can be on either pending processing or io lists in
203 struct list_head list
;
205 /** Entry on the interrupts list */
206 struct list_head intr_entry
;
211 /** Unique ID for the interrupt request */
215 * The following bitfields are either set once before the
216 * request is queued or setting/clearing them is protected by
220 /** True if the request has reply */
223 /** Force sending of the request even if interrupted */
226 /** The request was aborted */
229 /** Request is sent in the background */
230 unsigned background
:1;
232 /** The request has been interrupted */
233 unsigned interrupted
:1;
235 /** Data is being copied to/from the request */
238 /** Request is counted as "waiting" */
241 /** State of the request */
242 enum fuse_req_state state
;
244 /** The request input */
247 /** The request output */
250 /** Used to wake up the task waiting for completion of request*/
251 wait_queue_head_t waitq
;
253 /** Data for asynchronous requests */
255 struct fuse_forget_in forget_in
;
257 struct fuse_release_in in
;
260 struct fuse_init_in init_in
;
261 struct fuse_init_out init_out
;
262 struct cuse_init_in cuse_init_in
;
263 struct cuse_init_out cuse_init_out
;
265 struct fuse_read_in in
;
269 struct fuse_write_in in
;
270 struct fuse_write_out out
;
272 struct fuse_lk_in lk_in
;
276 struct page
*pages
[FUSE_MAX_PAGES_PER_REQ
];
278 /** number of pages in vector */
281 /** offset of data on first page */
282 unsigned page_offset
;
284 /** File used in the request (or NULL) */
285 struct fuse_file
*ff
;
287 /** Inode used in the request or NULL */
290 /** Link on fi->writepages */
291 struct list_head writepages_entry
;
293 /** Request completion callback */
294 void (*end
)(struct fuse_conn
*, struct fuse_req
*);
296 /** Request is stolen from fuse_file->reserved_req */
297 struct file
*stolen_file
;
303 * This structure is created, when the filesystem is mounted, and is
304 * destroyed, when the client device is closed and the filesystem is
308 /** Lock protecting accessess to members of this structure */
311 /** Mutex protecting against directory alias creation */
312 struct mutex inst_mutex
;
317 /** The user id for this mount */
320 /** The group id for this mount */
323 /** The fuse mount flags for this mount */
326 /** Maximum read size */
329 /** Maximum write size */
332 /** Readers of the connection are waiting on this */
333 wait_queue_head_t waitq
;
335 /** The list of pending requests */
336 struct list_head pending
;
338 /** The list of requests being processed */
339 struct list_head processing
;
341 /** The list of requests under I/O */
344 /** The next unique kernel file handle */
347 /** rbtree of fuse_files waiting for poll events indexed by ph */
348 struct rb_root polled_files
;
350 /** Maximum number of outstanding background requests */
351 unsigned max_background
;
353 /** Number of background requests at which congestion starts */
354 unsigned congestion_threshold
;
356 /** Number of requests currently in the background */
357 unsigned num_background
;
359 /** Number of background requests currently queued for userspace */
360 unsigned active_background
;
362 /** The list of background requests set aside for later queuing */
363 struct list_head bg_queue
;
365 /** Pending interrupts */
366 struct list_head interrupts
;
368 /** Flag indicating if connection is blocked. This will be
369 the case before the INIT reply is received, and if there
370 are too many outstading backgrounds requests */
373 /** waitq for blocked connection */
374 wait_queue_head_t blocked_waitq
;
376 /** waitq for reserved requests */
377 wait_queue_head_t reserved_req_waitq
;
379 /** The next unique request id */
382 /** Connection established, cleared on umount, connection
383 abort and device release */
386 /** Connection failed (version mismatch). Cannot race with
387 setting other bitfields since it is only set once in INIT
388 reply, before any other request, and never cleared */
389 unsigned conn_error
:1;
391 /** Connection successful. Only set in INIT */
392 unsigned conn_init
:1;
394 /** Do readpages asynchronously? Only set in INIT */
395 unsigned async_read
:1;
397 /** Do not send separate SETATTR request before open(O_TRUNC) */
398 unsigned atomic_o_trunc
:1;
400 /** Filesystem supports NFS exporting. Only set in INIT */
401 unsigned export_support
:1;
403 /** Set if bdi is valid */
404 unsigned bdi_initialized
:1;
407 * The following bitfields are only for optimization purposes
408 * and hence races in setting them will not cause malfunction
411 /** Is fsync not implemented by fs? */
414 /** Is fsyncdir not implemented by fs? */
415 unsigned no_fsyncdir
:1;
417 /** Is flush not implemented by fs? */
420 /** Is setxattr not implemented by fs? */
421 unsigned no_setxattr
:1;
423 /** Is getxattr not implemented by fs? */
424 unsigned no_getxattr
:1;
426 /** Is listxattr not implemented by fs? */
427 unsigned no_listxattr
:1;
429 /** Is removexattr not implemented by fs? */
430 unsigned no_removexattr
:1;
432 /** Are file locking primitives not implemented by fs? */
435 /** Is access not implemented by fs? */
436 unsigned no_access
:1;
438 /** Is create not implemented by fs? */
439 unsigned no_create
:1;
441 /** Is interrupt not implemented by fs? */
442 unsigned no_interrupt
:1;
444 /** Is bmap not implemented by fs? */
447 /** Is poll not implemented by fs? */
450 /** Do multi-page cached writes */
451 unsigned big_writes
:1;
453 /** Don't apply umask to creation modes */
454 unsigned dont_mask
:1;
456 /** The number of requests waiting for completion */
457 atomic_t num_waiting
;
459 /** Negotiated minor version */
462 /** Backing dev info */
463 struct backing_dev_info bdi
;
465 /** Entry on the fuse_conn_list */
466 struct list_head entry
;
468 /** Device ID from super block */
471 /** Dentries in the control filesystem */
472 struct dentry
*ctl_dentry
[FUSE_CTL_NUM_DENTRIES
];
474 /** number of dentries used in the above array */
477 /** O_ASYNC requests */
478 struct fasync_struct
*fasync
;
480 /** Key for lock owner ID scrambling */
483 /** Reserved request for the DESTROY message */
484 struct fuse_req
*destroy_req
;
486 /** Version counter for attribute changes */
489 /** Called on final put */
490 void (*release
)(struct fuse_conn
*);
492 /** Super block for this connection. */
493 struct super_block
*sb
;
495 /** Read/write semaphore to hold when accessing sb. */
496 struct rw_semaphore killsb
;
499 static inline struct fuse_conn
*get_fuse_conn_super(struct super_block
*sb
)
501 return sb
->s_fs_info
;
504 static inline struct fuse_conn
*get_fuse_conn(struct inode
*inode
)
506 return get_fuse_conn_super(inode
->i_sb
);
509 static inline struct fuse_inode
*get_fuse_inode(struct inode
*inode
)
511 return container_of(inode
, struct fuse_inode
, inode
);
514 static inline u64
get_node_id(struct inode
*inode
)
516 return get_fuse_inode(inode
)->nodeid
;
519 /** Device operations */
520 extern const struct file_operations fuse_dev_operations
;
522 extern const struct dentry_operations fuse_dentry_operations
;
525 * Inode to nodeid comparison.
527 int fuse_inode_eq(struct inode
*inode
, void *_nodeidp
);
530 * Get a filled in inode
532 struct inode
*fuse_iget(struct super_block
*sb
, u64 nodeid
,
533 int generation
, struct fuse_attr
*attr
,
534 u64 attr_valid
, u64 attr_version
);
536 int fuse_lookup_name(struct super_block
*sb
, u64 nodeid
, struct qstr
*name
,
537 struct fuse_entry_out
*outarg
, struct inode
**inode
);
540 * Send FORGET command
542 void fuse_send_forget(struct fuse_conn
*fc
, struct fuse_req
*req
,
543 u64 nodeid
, u64 nlookup
);
546 * Initialize READ or READDIR request
548 void fuse_read_fill(struct fuse_req
*req
, struct file
*file
,
549 loff_t pos
, size_t count
, int opcode
);
552 * Send OPEN or OPENDIR request
554 int fuse_open_common(struct inode
*inode
, struct file
*file
, bool isdir
);
556 struct fuse_file
*fuse_file_alloc(struct fuse_conn
*fc
);
557 struct fuse_file
*fuse_file_get(struct fuse_file
*ff
);
558 void fuse_file_free(struct fuse_file
*ff
);
559 void fuse_finish_open(struct inode
*inode
, struct file
*file
);
561 void fuse_sync_release(struct fuse_file
*ff
, int flags
);
564 * Send RELEASE or RELEASEDIR request
566 void fuse_release_common(struct file
*file
, int opcode
);
569 * Send FSYNC or FSYNCDIR request
571 int fuse_fsync_common(struct file
*file
, int datasync
, int isdir
);
576 int fuse_notify_poll_wakeup(struct fuse_conn
*fc
,
577 struct fuse_notify_poll_wakeup_out
*outarg
);
580 * Initialize file operations on a regular file
582 void fuse_init_file_inode(struct inode
*inode
);
585 * Initialize inode operations on regular files and special files
587 void fuse_init_common(struct inode
*inode
);
590 * Initialize inode and file operations on a directory
592 void fuse_init_dir(struct inode
*inode
);
595 * Initialize inode operations on a symlink
597 void fuse_init_symlink(struct inode
*inode
);
600 * Change attributes of an inode
602 void fuse_change_attributes(struct inode
*inode
, struct fuse_attr
*attr
,
603 u64 attr_valid
, u64 attr_version
);
605 void fuse_change_attributes_common(struct inode
*inode
, struct fuse_attr
*attr
,
609 * Initialize the client device
611 int fuse_dev_init(void);
614 * Cleanup the client device
616 void fuse_dev_cleanup(void);
618 int fuse_ctl_init(void);
619 void fuse_ctl_cleanup(void);
624 struct fuse_req
*fuse_request_alloc(void);
626 struct fuse_req
*fuse_request_alloc_nofs(void);
631 void fuse_request_free(struct fuse_req
*req
);
634 * Get a request, may fail with -ENOMEM
636 struct fuse_req
*fuse_get_req(struct fuse_conn
*fc
);
639 * Gets a requests for a file operation, always succeeds
641 struct fuse_req
*fuse_get_req_nofail(struct fuse_conn
*fc
, struct file
*file
);
644 * Decrement reference count of a request. If count goes to zero free
647 void fuse_put_request(struct fuse_conn
*fc
, struct fuse_req
*req
);
650 * Send a request (synchronous)
652 void fuse_request_send(struct fuse_conn
*fc
, struct fuse_req
*req
);
655 * Send a request with no reply
657 void fuse_request_send_noreply(struct fuse_conn
*fc
, struct fuse_req
*req
);
660 * Send a request in the background
662 void fuse_request_send_background(struct fuse_conn
*fc
, struct fuse_req
*req
);
664 void fuse_request_send_background_locked(struct fuse_conn
*fc
,
665 struct fuse_req
*req
);
667 /* Abort all requests */
668 void fuse_abort_conn(struct fuse_conn
*fc
);
671 * Invalidate inode attributes
673 void fuse_invalidate_attr(struct inode
*inode
);
675 void fuse_invalidate_entry_cache(struct dentry
*entry
);
678 * Acquire reference to fuse_conn
680 struct fuse_conn
*fuse_conn_get(struct fuse_conn
*fc
);
682 void fuse_conn_kill(struct fuse_conn
*fc
);
685 * Initialize fuse_conn
687 void fuse_conn_init(struct fuse_conn
*fc
);
690 * Release reference to fuse_conn
692 void fuse_conn_put(struct fuse_conn
*fc
);
695 * Add connection to control filesystem
697 int fuse_ctl_add_conn(struct fuse_conn
*fc
);
700 * Remove connection from control filesystem
702 void fuse_ctl_remove_conn(struct fuse_conn
*fc
);
705 * Is file type valid?
707 int fuse_valid_type(int m
);
710 * Is task allowed to perform filesystem operation?
712 int fuse_allow_task(struct fuse_conn
*fc
, struct task_struct
*task
);
714 u64
fuse_lock_owner_id(struct fuse_conn
*fc
, fl_owner_t id
);
716 int fuse_update_attributes(struct inode
*inode
, struct kstat
*stat
,
717 struct file
*file
, bool *refreshed
);
719 void fuse_flush_writepages(struct inode
*inode
);
721 void fuse_set_nowrite(struct inode
*inode
);
722 void fuse_release_nowrite(struct inode
*inode
);
724 u64
fuse_get_attr_version(struct fuse_conn
*fc
);
727 * File-system tells the kernel to invalidate cache for the given node id.
729 int fuse_reverse_inval_inode(struct super_block
*sb
, u64 nodeid
,
730 loff_t offset
, loff_t len
);
733 * File-system tells the kernel to invalidate parent attributes and
734 * the dentry matching parent/name.
736 int fuse_reverse_inval_entry(struct super_block
*sb
, u64 parent_nodeid
,
739 int fuse_do_open(struct fuse_conn
*fc
, u64 nodeid
, struct file
*file
,
741 ssize_t
fuse_direct_io(struct file
*file
, const char __user
*buf
,
742 size_t count
, loff_t
*ppos
, int write
);
743 long fuse_do_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
,
745 unsigned fuse_file_poll(struct file
*file
, poll_table
*wait
);
746 int fuse_dev_release(struct inode
*inode
, struct file
*file
);
748 #endif /* _FS_FUSE_I_H */