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>
25 #include <linux/kref.h>
26 #include <linux/xattr.h>
27 #include <linux/pid_namespace.h>
28 #include <linux/refcount.h>
29 #include <linux/user_namespace.h>
31 /** Max number of pages that can be used in a single read request */
32 #define FUSE_MAX_PAGES_PER_REQ 32
34 /** Bias for fi->writectr, meaning new writepages must not be sent */
35 #define FUSE_NOWRITE INT_MIN
37 /** It could be as large as PATH_MAX, but would that have any uses? */
38 #define FUSE_NAME_MAX 1024
40 /** Number of dentries for each connection in the control filesystem */
41 #define FUSE_CTL_NUM_DENTRIES 5
43 /** Number of page pointers embedded in fuse_req */
44 #define FUSE_REQ_INLINE_PAGES 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
;
56 /* One forget request */
57 struct fuse_forget_link
{
58 struct fuse_forget_one forget_one
;
59 struct fuse_forget_link
*next
;
67 /** Unique ID, which identifies the inode between userspace
71 /** Number of lookups on this inode */
74 /** The request used for sending the FORGET message */
75 struct fuse_forget_link
*forget
;
77 /** Time in jiffies until the file attributes are valid */
80 /** The sticky bit in inode->i_mode may have been removed, so
81 preserve the original mode */
84 /** 64 bit inode number */
87 /** Version of last attribute change */
90 /** Files usable in writepage. Protected by fc->lock */
91 struct list_head write_files
;
93 /** Writepages pending on truncate or fsync */
94 struct list_head queued_writes
;
96 /** Number of sent writes, a negative bias (FUSE_NOWRITE)
97 * means more writes are blocked */
100 /** Waitq for writepage completion */
101 wait_queue_head_t page_waitq
;
103 /** List of writepage requestst (pending or sent) */
104 struct list_head writepages
;
106 /** Miscellaneous bits describing inode state */
109 /** Lock for serializing lookup and readdir for back compatibility*/
113 /** FUSE inode state bits */
115 /** Advise readdirplus */
116 FUSE_I_ADVISE_RDPLUS
,
117 /** Initialized with readdirplus */
119 /** An operation changing file size is in progress */
120 FUSE_I_SIZE_UNSTABLE
,
125 /** FUSE specific file data */
127 /** Fuse connection for this file */
128 struct fuse_conn
*fc
;
130 /** Request reserved for flush and release */
131 struct fuse_req
*reserved_req
;
133 /** Kernel file handle guaranteed to be unique */
136 /** File handle used by userspace */
139 /** Node id of this file */
145 /** FOPEN_* flags returned by open */
148 /** Entry on inode's write_files list */
149 struct list_head write_entry
;
151 /** RB node to be linked on fuse_conn->polled_files */
152 struct rb_node polled_node
;
154 /** Wait queue head for poll */
155 wait_queue_head_t poll_wait
;
157 /** Has flock been performed on this file? */
161 /** One input argument of a request */
167 /** The request input */
169 /** The request header */
170 struct fuse_in_header h
;
172 /** True if the data for the last argument is in req->pages */
175 /** Number of arguments */
178 /** Array of arguments */
179 struct fuse_in_arg args
[3];
182 /** One output argument of a request */
188 /** The request output */
190 /** Header returned from userspace */
191 struct fuse_out_header h
;
194 * The following bitfields are not changed during the request
198 /** Last argument is variable length (can be shorter than
202 /** Last argument is a list of pages to copy data to */
205 /** Zero partially or not copied pages */
206 unsigned page_zeroing
:1;
208 /** Pages may be replaced with new ones */
209 unsigned page_replace
:1;
211 /** Number or arguments */
214 /** Array of arguments */
215 struct fuse_arg args
[2];
218 /** FUSE page descriptor */
219 struct fuse_page_desc
{
231 struct fuse_in_arg args
[3];
237 struct fuse_arg args
[2];
241 #define FUSE_ARGS(args) struct fuse_args args = {}
243 /** The request IO state (for asynchronous processing) */
244 struct fuse_io_priv
{
256 struct completion
*done
;
260 #define FUSE_IO_PRIV_SYNC(i) \
262 .refcnt = KREF_INIT(1), \
270 * FR_ISREPLY: set if the request has reply
271 * FR_FORCE: force sending of the request even if interrupted
272 * FR_BACKGROUND: request is sent in the background
273 * FR_WAITING: request is counted as "waiting"
274 * FR_ABORTED: the request was aborted
275 * FR_INTERRUPTED: the request has been interrupted
276 * FR_LOCKED: data is being copied to/from the request
277 * FR_PENDING: request is not yet in userspace
278 * FR_SENT: request is in userspace, waiting for an answer
279 * FR_FINISHED: request is finished
280 * FR_PRIVATE: request is on private list
297 * A request to the client
299 * .waitq.lock protects the following fields:
301 * - FR_LOCKED (may also be modified under fc->lock, tested under both)
304 /** This can be on either pending processing or io lists in
306 struct list_head list
;
308 /** Entry on the interrupts list */
309 struct list_head intr_entry
;
314 /** Unique ID for the interrupt request */
317 /* Request flags, updated with test/set/clear_bit() */
320 /** The request input */
323 /** The request output */
326 /** Used to wake up the task waiting for completion of request*/
327 wait_queue_head_t waitq
;
329 /** Data for asynchronous requests */
332 struct fuse_release_in in
;
335 struct fuse_init_in init_in
;
336 struct fuse_init_out init_out
;
337 struct cuse_init_in cuse_init_in
;
339 struct fuse_read_in in
;
343 struct fuse_write_in in
;
344 struct fuse_write_out out
;
345 struct fuse_req
*next
;
347 struct fuse_notify_retrieve_in retrieve_in
;
353 /** page-descriptor vector */
354 struct fuse_page_desc
*page_descs
;
356 /** size of the 'pages' array */
359 /** inline page vector */
360 struct page
*inline_pages
[FUSE_REQ_INLINE_PAGES
];
362 /** inline page-descriptor vector */
363 struct fuse_page_desc inline_page_descs
[FUSE_REQ_INLINE_PAGES
];
365 /** number of pages in vector */
368 /** File used in the request (or NULL) */
369 struct fuse_file
*ff
;
371 /** Inode used in the request or NULL */
374 /** AIO control block */
375 struct fuse_io_priv
*io
;
377 /** Link on fi->writepages */
378 struct list_head writepages_entry
;
380 /** Request completion callback */
381 void (*end
)(struct fuse_conn
*, struct fuse_req
*);
383 /** Request is stolen from fuse_file->reserved_req */
384 struct file
*stolen_file
;
388 /** Connection established */
391 /** Readers of the connection are waiting on this */
392 wait_queue_head_t waitq
;
394 /** The next unique request id */
397 /** The list of pending requests */
398 struct list_head pending
;
400 /** Pending interrupts */
401 struct list_head interrupts
;
403 /** Queue of pending forgets */
404 struct fuse_forget_link forget_list_head
;
405 struct fuse_forget_link
*forget_list_tail
;
407 /** Batching of FORGET requests (positive indicates FORGET batch) */
410 /** O_ASYNC requests */
411 struct fasync_struct
*fasync
;
415 /** Connection established */
418 /** Lock protecting accessess to members of this structure */
421 /** The list of requests being processed */
422 struct list_head processing
;
424 /** The list of requests under I/O */
429 * Fuse device instance
432 /** Fuse connection for this device */
433 struct fuse_conn
*fc
;
435 /** Processing queue */
436 struct fuse_pqueue pq
;
438 /** list entry on fc->devices */
439 struct list_head entry
;
445 * This structure is created, when the filesystem is mounted, and is
446 * destroyed, when the client device is closed and the filesystem is
450 /** Lock protecting accessess to members of this structure */
456 /** Number of fuse_dev's */
461 /** The user id for this mount */
464 /** The group id for this mount */
467 /** The pid namespace for this mount */
468 struct pid_namespace
*pid_ns
;
470 /** The user namespace for this mount */
471 struct user_namespace
*user_ns
;
473 /** Maximum read size */
476 /** Maximum write size */
480 struct fuse_iqueue iq
;
482 /** The next unique kernel file handle */
485 /** rbtree of fuse_files waiting for poll events indexed by ph */
486 struct rb_root polled_files
;
488 /** Maximum number of outstanding background requests */
489 unsigned max_background
;
491 /** Number of background requests at which congestion starts */
492 unsigned congestion_threshold
;
494 /** Number of requests currently in the background */
495 unsigned num_background
;
497 /** Number of background requests currently queued for userspace */
498 unsigned active_background
;
500 /** The list of background requests set aside for later queuing */
501 struct list_head bg_queue
;
503 /** Flag indicating that INIT reply has been received. Allocating
504 * any fuse request will be suspended until the flag is set */
507 /** Flag indicating if connection is blocked. This will be
508 the case before the INIT reply is received, and if there
509 are too many outstading backgrounds requests */
512 /** waitq for blocked connection */
513 wait_queue_head_t blocked_waitq
;
515 /** waitq for reserved requests */
516 wait_queue_head_t reserved_req_waitq
;
518 /** Connection established, cleared on umount, connection
519 abort and device release */
522 /** Connection aborted via sysfs */
525 /** Connection failed (version mismatch). Cannot race with
526 setting other bitfields since it is only set once in INIT
527 reply, before any other request, and never cleared */
528 unsigned conn_error
:1;
530 /** Connection successful. Only set in INIT */
531 unsigned conn_init
:1;
533 /** Do readpages asynchronously? Only set in INIT */
534 unsigned async_read
:1;
536 /** Return an unique read error after abort. Only set in INIT */
537 unsigned abort_err
:1;
539 /** Do not send separate SETATTR request before open(O_TRUNC) */
540 unsigned atomic_o_trunc
:1;
542 /** Filesystem supports NFS exporting. Only set in INIT */
543 unsigned export_support
:1;
545 /** write-back cache policy (default is write-through) */
546 unsigned writeback_cache
:1;
548 /** allow parallel lookups and readdir (default is serialized) */
549 unsigned parallel_dirops
:1;
551 /** handle fs handles killing suid/sgid/cap on write/chown/trunc */
552 unsigned handle_killpriv
:1;
555 * The following bitfields are only for optimization purposes
556 * and hence races in setting them will not cause malfunction
559 /** Is open/release not implemented by fs? */
562 /** Is fsync not implemented by fs? */
565 /** Is fsyncdir not implemented by fs? */
566 unsigned no_fsyncdir
:1;
568 /** Is flush not implemented by fs? */
571 /** Is setxattr not implemented by fs? */
572 unsigned no_setxattr
:1;
574 /** Is getxattr not implemented by fs? */
575 unsigned no_getxattr
:1;
577 /** Is listxattr not implemented by fs? */
578 unsigned no_listxattr
:1;
580 /** Is removexattr not implemented by fs? */
581 unsigned no_removexattr
:1;
583 /** Are posix file locking primitives not implemented by fs? */
586 /** Is access not implemented by fs? */
587 unsigned no_access
:1;
589 /** Is create not implemented by fs? */
590 unsigned no_create
:1;
592 /** Is interrupt not implemented by fs? */
593 unsigned no_interrupt
:1;
595 /** Is bmap not implemented by fs? */
598 /** Is poll not implemented by fs? */
601 /** Do multi-page cached writes */
602 unsigned big_writes
:1;
604 /** Don't apply umask to creation modes */
605 unsigned dont_mask
:1;
607 /** Are BSD file locking primitives not implemented by fs? */
610 /** Is fallocate not implemented by fs? */
611 unsigned no_fallocate
:1;
613 /** Is rename with flags implemented by fs? */
614 unsigned no_rename2
:1;
616 /** Use enhanced/automatic page cache invalidation. */
617 unsigned auto_inval_data
:1;
619 /** Does the filesystem support readdirplus? */
620 unsigned do_readdirplus
:1;
622 /** Does the filesystem want adaptive readdirplus? */
623 unsigned readdirplus_auto
:1;
625 /** Does the filesystem support asynchronous direct-IO submission? */
626 unsigned async_dio
:1;
628 /** Is lseek not implemented by fs? */
631 /** Does the filesystem support posix acls? */
632 unsigned posix_acl
:1;
634 /** Check permissions based on the file mode or not? */
635 unsigned default_permissions
:1;
637 /** Allow other than the mounter user to access the filesystem ? */
638 unsigned allow_other
:1;
640 /** The number of requests waiting for completion */
641 atomic_t num_waiting
;
643 /** Negotiated minor version */
646 /** Entry on the fuse_conn_list */
647 struct list_head entry
;
649 /** Device ID from super block */
652 /** Dentries in the control filesystem */
653 struct dentry
*ctl_dentry
[FUSE_CTL_NUM_DENTRIES
];
655 /** number of dentries used in the above array */
658 /** Key for lock owner ID scrambling */
661 /** Reserved request for the DESTROY message */
662 struct fuse_req
*destroy_req
;
664 /** Version counter for attribute changes */
667 /** Called on final put */
668 void (*release
)(struct fuse_conn
*);
670 /** Super block for this connection. */
671 struct super_block
*sb
;
673 /** Read/write semaphore to hold when accessing sb. */
674 struct rw_semaphore killsb
;
676 /** List of device instances belonging to this connection */
677 struct list_head devices
;
680 static inline struct fuse_conn
*get_fuse_conn_super(struct super_block
*sb
)
682 return sb
->s_fs_info
;
685 static inline struct fuse_conn
*get_fuse_conn(struct inode
*inode
)
687 return get_fuse_conn_super(inode
->i_sb
);
690 static inline struct fuse_inode
*get_fuse_inode(struct inode
*inode
)
692 return container_of(inode
, struct fuse_inode
, inode
);
695 static inline u64
get_node_id(struct inode
*inode
)
697 return get_fuse_inode(inode
)->nodeid
;
700 /** Device operations */
701 extern const struct file_operations fuse_dev_operations
;
703 extern const struct dentry_operations fuse_dentry_operations
;
704 extern const struct dentry_operations fuse_root_dentry_operations
;
707 * Inode to nodeid comparison.
709 int fuse_inode_eq(struct inode
*inode
, void *_nodeidp
);
712 * Get a filled in inode
714 struct inode
*fuse_iget(struct super_block
*sb
, u64 nodeid
,
715 int generation
, struct fuse_attr
*attr
,
716 u64 attr_valid
, u64 attr_version
);
718 int fuse_lookup_name(struct super_block
*sb
, u64 nodeid
, const struct qstr
*name
,
719 struct fuse_entry_out
*outarg
, struct inode
**inode
);
722 * Send FORGET command
724 void fuse_queue_forget(struct fuse_conn
*fc
, struct fuse_forget_link
*forget
,
725 u64 nodeid
, u64 nlookup
);
727 struct fuse_forget_link
*fuse_alloc_forget(void);
729 /* Used by READDIRPLUS */
730 void fuse_force_forget(struct file
*file
, u64 nodeid
);
733 * Initialize READ or READDIR request
735 void fuse_read_fill(struct fuse_req
*req
, struct file
*file
,
736 loff_t pos
, size_t count
, int opcode
);
739 * Send OPEN or OPENDIR request
741 int fuse_open_common(struct inode
*inode
, struct file
*file
, bool isdir
);
743 struct fuse_file
*fuse_file_alloc(struct fuse_conn
*fc
);
744 void fuse_file_free(struct fuse_file
*ff
);
745 void fuse_finish_open(struct inode
*inode
, struct file
*file
);
747 void fuse_sync_release(struct fuse_file
*ff
, int flags
);
750 * Send RELEASE or RELEASEDIR request
752 void fuse_release_common(struct file
*file
, int opcode
);
755 * Send FSYNC or FSYNCDIR request
757 int fuse_fsync_common(struct file
*file
, loff_t start
, loff_t end
,
758 int datasync
, int isdir
);
763 int fuse_notify_poll_wakeup(struct fuse_conn
*fc
,
764 struct fuse_notify_poll_wakeup_out
*outarg
);
767 * Initialize file operations on a regular file
769 void fuse_init_file_inode(struct inode
*inode
);
772 * Initialize inode operations on regular files and special files
774 void fuse_init_common(struct inode
*inode
);
777 * Initialize inode and file operations on a directory
779 void fuse_init_dir(struct inode
*inode
);
782 * Initialize inode operations on a symlink
784 void fuse_init_symlink(struct inode
*inode
);
787 * Change attributes of an inode
789 void fuse_change_attributes(struct inode
*inode
, struct fuse_attr
*attr
,
790 u64 attr_valid
, u64 attr_version
);
792 void fuse_change_attributes_common(struct inode
*inode
, struct fuse_attr
*attr
,
796 * Initialize the client device
798 int fuse_dev_init(void);
801 * Cleanup the client device
803 void fuse_dev_cleanup(void);
805 int fuse_ctl_init(void);
806 void __exit
fuse_ctl_cleanup(void);
811 struct fuse_req
*fuse_request_alloc(unsigned npages
);
813 struct fuse_req
*fuse_request_alloc_nofs(unsigned npages
);
818 void fuse_request_free(struct fuse_req
*req
);
821 * Get a request, may fail with -ENOMEM,
822 * caller should specify # elements in req->pages[] explicitly
824 struct fuse_req
*fuse_get_req(struct fuse_conn
*fc
, unsigned npages
);
825 struct fuse_req
*fuse_get_req_for_background(struct fuse_conn
*fc
,
829 * Increment reference count on request
831 void __fuse_get_request(struct fuse_req
*req
);
834 * Gets a requests for a file operation, always succeeds
836 struct fuse_req
*fuse_get_req_nofail_nopages(struct fuse_conn
*fc
,
840 * Decrement reference count of a request. If count goes to zero free
843 void fuse_put_request(struct fuse_conn
*fc
, struct fuse_req
*req
);
846 * Send a request (synchronous)
848 void fuse_request_send(struct fuse_conn
*fc
, struct fuse_req
*req
);
851 * Simple request sending that does request allocation and freeing
853 ssize_t
fuse_simple_request(struct fuse_conn
*fc
, struct fuse_args
*args
);
856 * Send a request in the background
858 void fuse_request_send_background(struct fuse_conn
*fc
, struct fuse_req
*req
);
860 void fuse_request_send_background_locked(struct fuse_conn
*fc
,
861 struct fuse_req
*req
);
863 /* Abort all requests */
864 void fuse_abort_conn(struct fuse_conn
*fc
, bool is_abort
);
865 void fuse_wait_aborted(struct fuse_conn
*fc
);
868 * Invalidate inode attributes
870 void fuse_invalidate_attr(struct inode
*inode
);
872 void fuse_invalidate_entry_cache(struct dentry
*entry
);
874 void fuse_invalidate_atime(struct inode
*inode
);
877 * Acquire reference to fuse_conn
879 struct fuse_conn
*fuse_conn_get(struct fuse_conn
*fc
);
882 * Initialize fuse_conn
884 void fuse_conn_init(struct fuse_conn
*fc
, struct user_namespace
*user_ns
);
887 * Release reference to fuse_conn
889 void fuse_conn_put(struct fuse_conn
*fc
);
891 struct fuse_dev
*fuse_dev_alloc(struct fuse_conn
*fc
);
892 void fuse_dev_free(struct fuse_dev
*fud
);
895 * Add connection to control filesystem
897 int fuse_ctl_add_conn(struct fuse_conn
*fc
);
900 * Remove connection from control filesystem
902 void fuse_ctl_remove_conn(struct fuse_conn
*fc
);
905 * Is file type valid?
907 int fuse_valid_type(int m
);
910 * Is current process allowed to perform filesystem operation?
912 int fuse_allow_current_process(struct fuse_conn
*fc
);
914 u64
fuse_lock_owner_id(struct fuse_conn
*fc
, fl_owner_t id
);
916 void fuse_update_ctime(struct inode
*inode
);
918 int fuse_update_attributes(struct inode
*inode
, struct file
*file
);
920 void fuse_flush_writepages(struct inode
*inode
);
922 void fuse_set_nowrite(struct inode
*inode
);
923 void fuse_release_nowrite(struct inode
*inode
);
925 u64
fuse_get_attr_version(struct fuse_conn
*fc
);
928 * File-system tells the kernel to invalidate cache for the given node id.
930 int fuse_reverse_inval_inode(struct super_block
*sb
, u64 nodeid
,
931 loff_t offset
, loff_t len
);
934 * File-system tells the kernel to invalidate parent attributes and
935 * the dentry matching parent/name.
937 * If the child_nodeid is non-zero and:
938 * - matches the inode number for the dentry matching parent/name,
939 * - is not a mount point
940 * - is a file or oan empty directory
941 * then the dentry is unhashed (d_delete()).
943 int fuse_reverse_inval_entry(struct super_block
*sb
, u64 parent_nodeid
,
944 u64 child_nodeid
, struct qstr
*name
);
946 int fuse_do_open(struct fuse_conn
*fc
, u64 nodeid
, struct file
*file
,
950 * fuse_direct_io() flags
953 /** If set, it is WRITE; otherwise - READ */
954 #define FUSE_DIO_WRITE (1 << 0)
956 /** CUSE pass fuse_direct_io() a file which f_mapping->host is not from FUSE */
957 #define FUSE_DIO_CUSE (1 << 1)
959 ssize_t
fuse_direct_io(struct fuse_io_priv
*io
, struct iov_iter
*iter
,
960 loff_t
*ppos
, int flags
);
961 long fuse_do_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
,
963 long fuse_ioctl_common(struct file
*file
, unsigned int cmd
,
964 unsigned long arg
, unsigned int flags
);
965 __poll_t
fuse_file_poll(struct file
*file
, poll_table
*wait
);
966 int fuse_dev_release(struct inode
*inode
, struct file
*file
);
968 bool fuse_write_update_size(struct inode
*inode
, loff_t pos
);
970 int fuse_flush_times(struct inode
*inode
, struct fuse_file
*ff
);
971 int fuse_write_inode(struct inode
*inode
, struct writeback_control
*wbc
);
973 int fuse_do_setattr(struct dentry
*dentry
, struct iattr
*attr
,
976 void fuse_set_initialized(struct fuse_conn
*fc
);
978 void fuse_unlock_inode(struct inode
*inode
, bool locked
);
979 bool fuse_lock_inode(struct inode
*inode
);
981 int fuse_setxattr(struct inode
*inode
, const char *name
, const void *value
,
982 size_t size
, int flags
);
983 ssize_t
fuse_getxattr(struct inode
*inode
, const char *name
, void *value
,
985 ssize_t
fuse_listxattr(struct dentry
*entry
, char *list
, size_t size
);
986 int fuse_removexattr(struct inode
*inode
, const char *name
);
987 extern const struct xattr_handler
*fuse_xattr_handlers
[];
988 extern const struct xattr_handler
*fuse_acl_xattr_handlers
[];
989 extern const struct xattr_handler
*fuse_no_acl_xattr_handlers
[];
992 struct posix_acl
*fuse_get_acl(struct inode
*inode
, int type
);
993 int fuse_set_acl(struct inode
*inode
, struct posix_acl
*acl
, int type
);
995 #endif /* _FS_FUSE_I_H */