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>
24 #include <linux/workqueue.h>
26 /** Max number of pages that can be used in a single read request */
27 #define FUSE_MAX_PAGES_PER_REQ 32
29 /** Bias for fi->writectr, meaning new writepages must not be sent */
30 #define FUSE_NOWRITE INT_MIN
32 /** It could be as large as PATH_MAX, but would that have any uses? */
33 #define FUSE_NAME_MAX 1024
35 /** Number of dentries for each connection in the control filesystem */
36 #define FUSE_CTL_NUM_DENTRIES 5
38 /** If the FUSE_DEFAULT_PERMISSIONS flag is given, the filesystem
39 module will check permissions based on the file mode. Otherwise no
40 permission checking is done in the kernel */
41 #define FUSE_DEFAULT_PERMISSIONS (1 << 0)
43 /** If the FUSE_ALLOW_OTHER flag is given, then not only the user
44 doing the mount will be allowed to access the filesystem */
45 #define FUSE_ALLOW_OTHER (1 << 1)
47 /** Number of page pointers embedded in fuse_req */
48 #define FUSE_REQ_INLINE_PAGES 1
50 /** List of active connections */
51 extern struct list_head fuse_conn_list
;
53 /** Global mutex protecting fuse_conn_list and the control filesystem */
54 extern struct mutex fuse_mutex
;
56 /** Module parameters */
57 extern unsigned max_user_bgreq
;
58 extern unsigned max_user_congthresh
;
60 /* One forget request */
61 struct fuse_forget_link
{
62 struct fuse_forget_one forget_one
;
63 struct fuse_forget_link
*next
;
71 /** Unique ID, which identifies the inode between userspace
75 /** Number of lookups on this inode */
78 /** The request used for sending the FORGET message */
79 struct fuse_forget_link
*forget
;
81 /** Time in jiffies until the file attributes are valid */
84 /** The sticky bit in inode->i_mode may have been removed, so
85 preserve the original mode */
88 /** 64 bit inode number */
91 /** Version of last attribute change */
94 /** Files usable in writepage. Protected by fc->lock */
95 struct list_head write_files
;
97 /** Writepages pending on truncate or fsync */
98 struct list_head queued_writes
;
100 /** Number of sent writes, a negative bias (FUSE_NOWRITE)
101 * means more writes are blocked */
104 /** Waitq for writepage completion */
105 wait_queue_head_t page_waitq
;
107 /** List of writepage requestst (pending or sent) */
108 struct list_head writepages
;
110 /** Miscellaneous bits describing inode state */
114 /** FUSE inode state bits */
116 /** Advise readdirplus */
117 FUSE_I_ADVISE_RDPLUS
,
118 /** Initialized with readdirplus */
120 /** An operation changing file size is in progress */
121 FUSE_I_SIZE_UNSTABLE
,
126 /** FUSE specific file data */
128 /** Fuse connection for this file */
129 struct fuse_conn
*fc
;
131 /** Request reserved for flush and release */
132 struct fuse_req
*reserved_req
;
134 /** Kernel file handle guaranteed to be unique */
137 /** File handle used by userspace */
140 /** Node id of this file */
146 /** FOPEN_* flags returned by open */
149 /** Entry on inode's write_files list */
150 struct list_head write_entry
;
152 /** RB node to be linked on fuse_conn->polled_files */
153 struct rb_node polled_node
;
155 /** Wait queue head for poll */
156 wait_queue_head_t poll_wait
;
158 /** Has flock been performed on this file? */
162 /** One input argument of a request */
168 /** The request input */
170 /** The request header */
171 struct fuse_in_header h
;
173 /** True if the data for the last argument is in req->pages */
176 /** Number of arguments */
179 /** Array of arguments */
180 struct fuse_in_arg args
[3];
183 /** One output argument of a request */
189 /** The request output */
191 /** Header returned from userspace */
192 struct fuse_out_header h
;
195 * The following bitfields are not changed during the request
199 /** Last argument is variable length (can be shorter than
203 /** Last argument is a list of pages to copy data to */
206 /** Zero partially or not copied pages */
207 unsigned page_zeroing
:1;
209 /** Pages may be replaced with new ones */
210 unsigned page_replace
:1;
212 /** Number or arguments */
215 /** Array of arguments */
216 struct fuse_arg args
[2];
219 /** FUSE page descriptor */
220 struct fuse_page_desc
{
232 struct fuse_in_arg args
[3];
238 struct fuse_arg args
[2];
242 #define FUSE_ARGS(args) struct fuse_args args = {}
244 /** The request state */
245 enum fuse_req_state
{
254 /** The request IO state (for asynchronous processing) */
255 struct fuse_io_priv
{
266 struct completion
*done
;
272 * FR_ISREPLY: set if the request has reply
273 * FR_FORCE: force sending of the request even if interrupted
274 * FR_BACKGROUND: request is sent in the background
275 * FR_WAITING: request is counted as "waiting"
276 * FR_ABORTED: the request was aborted
277 * FR_INTERRUPTED: the request has been interrupted
278 * FR_LOCKED: data is being copied to/from the request
291 * A request to the client
293 * .waitq.lock protects the following fields:
295 * - FR_LOCKED (may also be modified under fc->lock, tested under both)
298 /** This can be on either pending processing or io lists in
300 struct list_head list
;
302 /** Entry on the interrupts list */
303 struct list_head intr_entry
;
308 /** Unique ID for the interrupt request */
311 /* Request flags, updated with test/set/clear_bit() */
314 /** State of the request */
315 enum fuse_req_state state
;
317 /** The request input */
320 /** The request output */
323 /** Used to wake up the task waiting for completion of request*/
324 wait_queue_head_t waitq
;
326 /** Data for asynchronous requests */
329 struct fuse_release_in in
;
332 struct fuse_init_in init_in
;
333 struct fuse_init_out init_out
;
334 struct cuse_init_in cuse_init_in
;
336 struct fuse_read_in in
;
340 struct fuse_write_in in
;
341 struct fuse_write_out out
;
342 struct fuse_req
*next
;
344 struct fuse_notify_retrieve_in retrieve_in
;
350 /** page-descriptor vector */
351 struct fuse_page_desc
*page_descs
;
353 /** size of the 'pages' array */
356 /** inline page vector */
357 struct page
*inline_pages
[FUSE_REQ_INLINE_PAGES
];
359 /** inline page-descriptor vector */
360 struct fuse_page_desc inline_page_descs
[FUSE_REQ_INLINE_PAGES
];
362 /** number of pages in vector */
365 /** File used in the request (or NULL) */
366 struct fuse_file
*ff
;
368 /** Inode used in the request or NULL */
371 /** AIO control block */
372 struct fuse_io_priv
*io
;
374 /** Link on fi->writepages */
375 struct list_head writepages_entry
;
377 /** Request completion callback */
378 void (*end
)(struct fuse_conn
*, struct fuse_req
*);
380 /** Request is stolen from fuse_file->reserved_req */
381 struct file
*stolen_file
;
387 * This structure is created, when the filesystem is mounted, and is
388 * destroyed, when the client device is closed and the filesystem is
392 /** Lock protecting accessess to members of this structure */
400 /** The user id for this mount */
403 /** The group id for this mount */
406 /** The fuse mount flags for this mount */
409 /** Maximum read size */
412 /** Maximum write size */
415 /** Readers of the connection are waiting on this */
416 wait_queue_head_t waitq
;
418 /** The list of pending requests */
419 struct list_head pending
;
421 /** The list of requests being processed */
422 struct list_head processing
;
424 /** The list of requests under I/O */
427 /** The next unique kernel file handle */
430 /** rbtree of fuse_files waiting for poll events indexed by ph */
431 struct rb_root polled_files
;
433 /** Maximum number of outstanding background requests */
434 unsigned max_background
;
436 /** Number of background requests at which congestion starts */
437 unsigned congestion_threshold
;
439 /** Number of requests currently in the background */
440 unsigned num_background
;
442 /** Number of background requests currently queued for userspace */
443 unsigned active_background
;
445 /** The list of background requests set aside for later queuing */
446 struct list_head bg_queue
;
448 /** Pending interrupts */
449 struct list_head interrupts
;
451 /** Queue of pending forgets */
452 struct fuse_forget_link forget_list_head
;
453 struct fuse_forget_link
*forget_list_tail
;
455 /** Batching of FORGET requests (positive indicates FORGET batch) */
458 /** Flag indicating that INIT reply has been received. Allocating
459 * any fuse request will be suspended until the flag is set */
462 /** Flag indicating if connection is blocked. This will be
463 the case before the INIT reply is received, and if there
464 are too many outstading backgrounds requests */
467 /** waitq for blocked connection */
468 wait_queue_head_t blocked_waitq
;
470 /** waitq for reserved requests */
471 wait_queue_head_t reserved_req_waitq
;
473 /** The next unique request id */
476 /** Connection established, cleared on umount, connection
477 abort and device release */
480 /** Connection failed (version mismatch). Cannot race with
481 setting other bitfields since it is only set once in INIT
482 reply, before any other request, and never cleared */
483 unsigned conn_error
:1;
485 /** Connection successful. Only set in INIT */
486 unsigned conn_init
:1;
488 /** Do readpages asynchronously? Only set in INIT */
489 unsigned async_read
:1;
491 /** Do not send separate SETATTR request before open(O_TRUNC) */
492 unsigned atomic_o_trunc
:1;
494 /** Filesystem supports NFS exporting. Only set in INIT */
495 unsigned export_support
:1;
497 /** Set if bdi is valid */
498 unsigned bdi_initialized
:1;
500 /** write-back cache policy (default is write-through) */
501 unsigned writeback_cache
:1;
504 * The following bitfields are only for optimization purposes
505 * and hence races in setting them will not cause malfunction
508 /** Is open/release not implemented by fs? */
511 /** Is fsync not implemented by fs? */
514 /** Is fsyncdir not implemented by fs? */
515 unsigned no_fsyncdir
:1;
517 /** Is flush not implemented by fs? */
520 /** Is setxattr not implemented by fs? */
521 unsigned no_setxattr
:1;
523 /** Is getxattr not implemented by fs? */
524 unsigned no_getxattr
:1;
526 /** Is listxattr not implemented by fs? */
527 unsigned no_listxattr
:1;
529 /** Is removexattr not implemented by fs? */
530 unsigned no_removexattr
:1;
532 /** Are posix file locking primitives not implemented by fs? */
535 /** Is access not implemented by fs? */
536 unsigned no_access
:1;
538 /** Is create not implemented by fs? */
539 unsigned no_create
:1;
541 /** Is interrupt not implemented by fs? */
542 unsigned no_interrupt
:1;
544 /** Is bmap not implemented by fs? */
547 /** Is poll not implemented by fs? */
550 /** Do multi-page cached writes */
551 unsigned big_writes
:1;
553 /** Don't apply umask to creation modes */
554 unsigned dont_mask
:1;
556 /** Are BSD file locking primitives not implemented by fs? */
559 /** Is fallocate not implemented by fs? */
560 unsigned no_fallocate
:1;
562 /** Is rename with flags implemented by fs? */
563 unsigned no_rename2
:1;
565 /** Use enhanced/automatic page cache invalidation. */
566 unsigned auto_inval_data
:1;
568 /** Does the filesystem support readdirplus? */
569 unsigned do_readdirplus
:1;
571 /** Does the filesystem want adaptive readdirplus? */
572 unsigned readdirplus_auto
:1;
574 /** Does the filesystem support asynchronous direct-IO submission? */
575 unsigned async_dio
:1;
577 /** The number of requests waiting for completion */
578 atomic_t num_waiting
;
580 /** Negotiated minor version */
583 /** Backing dev info */
584 struct backing_dev_info bdi
;
586 /** Entry on the fuse_conn_list */
587 struct list_head entry
;
589 /** Device ID from super block */
592 /** Dentries in the control filesystem */
593 struct dentry
*ctl_dentry
[FUSE_CTL_NUM_DENTRIES
];
595 /** number of dentries used in the above array */
598 /** O_ASYNC requests */
599 struct fasync_struct
*fasync
;
601 /** Key for lock owner ID scrambling */
604 /** Reserved request for the DESTROY message */
605 struct fuse_req
*destroy_req
;
607 /** Version counter for attribute changes */
610 /** Called on final put */
611 void (*release
)(struct fuse_conn
*);
613 /** Super block for this connection. */
614 struct super_block
*sb
;
616 /** Read/write semaphore to hold when accessing sb. */
617 struct rw_semaphore killsb
;
620 static inline struct fuse_conn
*get_fuse_conn_super(struct super_block
*sb
)
622 return sb
->s_fs_info
;
625 static inline struct fuse_conn
*get_fuse_conn(struct inode
*inode
)
627 return get_fuse_conn_super(inode
->i_sb
);
630 static inline struct fuse_inode
*get_fuse_inode(struct inode
*inode
)
632 return container_of(inode
, struct fuse_inode
, inode
);
635 static inline u64
get_node_id(struct inode
*inode
)
637 return get_fuse_inode(inode
)->nodeid
;
640 /** Device operations */
641 extern const struct file_operations fuse_dev_operations
;
643 extern const struct dentry_operations fuse_dentry_operations
;
646 * Inode to nodeid comparison.
648 int fuse_inode_eq(struct inode
*inode
, void *_nodeidp
);
651 * Get a filled in inode
653 struct inode
*fuse_iget(struct super_block
*sb
, u64 nodeid
,
654 int generation
, struct fuse_attr
*attr
,
655 u64 attr_valid
, u64 attr_version
);
657 int fuse_lookup_name(struct super_block
*sb
, u64 nodeid
, struct qstr
*name
,
658 struct fuse_entry_out
*outarg
, struct inode
**inode
);
661 * Send FORGET command
663 void fuse_queue_forget(struct fuse_conn
*fc
, struct fuse_forget_link
*forget
,
664 u64 nodeid
, u64 nlookup
);
666 struct fuse_forget_link
*fuse_alloc_forget(void);
668 /* Used by READDIRPLUS */
669 void fuse_force_forget(struct file
*file
, u64 nodeid
);
672 * Initialize READ or READDIR request
674 void fuse_read_fill(struct fuse_req
*req
, struct file
*file
,
675 loff_t pos
, size_t count
, int opcode
);
678 * Send OPEN or OPENDIR request
680 int fuse_open_common(struct inode
*inode
, struct file
*file
, bool isdir
);
682 struct fuse_file
*fuse_file_alloc(struct fuse_conn
*fc
);
683 struct fuse_file
*fuse_file_get(struct fuse_file
*ff
);
684 void fuse_file_free(struct fuse_file
*ff
);
685 void fuse_finish_open(struct inode
*inode
, struct file
*file
);
687 void fuse_sync_release(struct fuse_file
*ff
, int flags
);
690 * Send RELEASE or RELEASEDIR request
692 void fuse_release_common(struct file
*file
, int opcode
);
695 * Send FSYNC or FSYNCDIR request
697 int fuse_fsync_common(struct file
*file
, loff_t start
, loff_t end
,
698 int datasync
, int isdir
);
703 int fuse_notify_poll_wakeup(struct fuse_conn
*fc
,
704 struct fuse_notify_poll_wakeup_out
*outarg
);
707 * Initialize file operations on a regular file
709 void fuse_init_file_inode(struct inode
*inode
);
712 * Initialize inode operations on regular files and special files
714 void fuse_init_common(struct inode
*inode
);
717 * Initialize inode and file operations on a directory
719 void fuse_init_dir(struct inode
*inode
);
722 * Initialize inode operations on a symlink
724 void fuse_init_symlink(struct inode
*inode
);
727 * Change attributes of an inode
729 void fuse_change_attributes(struct inode
*inode
, struct fuse_attr
*attr
,
730 u64 attr_valid
, u64 attr_version
);
732 void fuse_change_attributes_common(struct inode
*inode
, struct fuse_attr
*attr
,
736 * Initialize the client device
738 int fuse_dev_init(void);
741 * Cleanup the client device
743 void fuse_dev_cleanup(void);
745 int fuse_ctl_init(void);
746 void __exit
fuse_ctl_cleanup(void);
751 struct fuse_req
*fuse_request_alloc(unsigned npages
);
753 struct fuse_req
*fuse_request_alloc_nofs(unsigned npages
);
758 void fuse_request_free(struct fuse_req
*req
);
761 * Get a request, may fail with -ENOMEM,
762 * caller should specify # elements in req->pages[] explicitly
764 struct fuse_req
*fuse_get_req(struct fuse_conn
*fc
, unsigned npages
);
765 struct fuse_req
*fuse_get_req_for_background(struct fuse_conn
*fc
,
769 * Increment reference count on request
771 void __fuse_get_request(struct fuse_req
*req
);
774 * Gets a requests for a file operation, always succeeds
776 struct fuse_req
*fuse_get_req_nofail_nopages(struct fuse_conn
*fc
,
780 * Decrement reference count of a request. If count goes to zero free
783 void fuse_put_request(struct fuse_conn
*fc
, struct fuse_req
*req
);
786 * Send a request (synchronous)
788 void fuse_request_send(struct fuse_conn
*fc
, struct fuse_req
*req
);
791 * Simple request sending that does request allocation and freeing
793 ssize_t
fuse_simple_request(struct fuse_conn
*fc
, struct fuse_args
*args
);
796 * Send a request in the background
798 void fuse_request_send_background(struct fuse_conn
*fc
, struct fuse_req
*req
);
800 void fuse_request_send_background_locked(struct fuse_conn
*fc
,
801 struct fuse_req
*req
);
803 /* Abort all requests */
804 void fuse_abort_conn(struct fuse_conn
*fc
);
807 * Invalidate inode attributes
809 void fuse_invalidate_attr(struct inode
*inode
);
811 void fuse_invalidate_entry_cache(struct dentry
*entry
);
813 void fuse_invalidate_atime(struct inode
*inode
);
816 * Acquire reference to fuse_conn
818 struct fuse_conn
*fuse_conn_get(struct fuse_conn
*fc
);
821 * Initialize fuse_conn
823 void fuse_conn_init(struct fuse_conn
*fc
);
826 * Release reference to fuse_conn
828 void fuse_conn_put(struct fuse_conn
*fc
);
831 * Add connection to control filesystem
833 int fuse_ctl_add_conn(struct fuse_conn
*fc
);
836 * Remove connection from control filesystem
838 void fuse_ctl_remove_conn(struct fuse_conn
*fc
);
841 * Is file type valid?
843 int fuse_valid_type(int m
);
846 * Is current process allowed to perform filesystem operation?
848 int fuse_allow_current_process(struct fuse_conn
*fc
);
850 u64
fuse_lock_owner_id(struct fuse_conn
*fc
, fl_owner_t id
);
852 int fuse_update_attributes(struct inode
*inode
, struct kstat
*stat
,
853 struct file
*file
, bool *refreshed
);
855 void fuse_flush_writepages(struct inode
*inode
);
857 void fuse_set_nowrite(struct inode
*inode
);
858 void fuse_release_nowrite(struct inode
*inode
);
860 u64
fuse_get_attr_version(struct fuse_conn
*fc
);
863 * File-system tells the kernel to invalidate cache for the given node id.
865 int fuse_reverse_inval_inode(struct super_block
*sb
, u64 nodeid
,
866 loff_t offset
, loff_t len
);
869 * File-system tells the kernel to invalidate parent attributes and
870 * the dentry matching parent/name.
872 * If the child_nodeid is non-zero and:
873 * - matches the inode number for the dentry matching parent/name,
874 * - is not a mount point
875 * - is a file or oan empty directory
876 * then the dentry is unhashed (d_delete()).
878 int fuse_reverse_inval_entry(struct super_block
*sb
, u64 parent_nodeid
,
879 u64 child_nodeid
, struct qstr
*name
);
881 int fuse_do_open(struct fuse_conn
*fc
, u64 nodeid
, struct file
*file
,
885 * fuse_direct_io() flags
888 /** If set, it is WRITE; otherwise - READ */
889 #define FUSE_DIO_WRITE (1 << 0)
891 /** CUSE pass fuse_direct_io() a file which f_mapping->host is not from FUSE */
892 #define FUSE_DIO_CUSE (1 << 1)
894 ssize_t
fuse_direct_io(struct fuse_io_priv
*io
, struct iov_iter
*iter
,
895 loff_t
*ppos
, int flags
);
896 long fuse_do_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
,
898 long fuse_ioctl_common(struct file
*file
, unsigned int cmd
,
899 unsigned long arg
, unsigned int flags
);
900 unsigned fuse_file_poll(struct file
*file
, poll_table
*wait
);
901 int fuse_dev_release(struct inode
*inode
, struct file
*file
);
903 bool fuse_write_update_size(struct inode
*inode
, loff_t pos
);
905 int fuse_flush_times(struct inode
*inode
, struct fuse_file
*ff
);
906 int fuse_write_inode(struct inode
*inode
, struct writeback_control
*wbc
);
908 int fuse_do_setattr(struct inode
*inode
, struct iattr
*attr
,
911 void fuse_set_initialized(struct fuse_conn
*fc
);
913 #endif /* _FS_FUSE_I_H */