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>
28 /** Max number of pages that can be used in a single read request */
29 #define FUSE_MAX_PAGES_PER_REQ 32
31 /** Bias for fi->writectr, meaning new writepages must not be sent */
32 #define FUSE_NOWRITE INT_MIN
34 /** It could be as large as PATH_MAX, but would that have any uses? */
35 #define FUSE_NAME_MAX 1024
37 /** Number of dentries for each connection in the control filesystem */
38 #define FUSE_CTL_NUM_DENTRIES 5
40 /** Number of page pointers embedded in fuse_req */
41 #define FUSE_REQ_INLINE_PAGES 1
43 /** List of active connections */
44 extern struct list_head fuse_conn_list;
46 /** Global mutex protecting fuse_conn_list and the control filesystem */
47 extern struct mutex fuse_mutex;
49 /** Module parameters */
50 extern unsigned max_user_bgreq;
51 extern unsigned max_user_congthresh;
53 /* One forget request */
54 struct fuse_forget_link {
55 struct fuse_forget_one forget_one;
56 struct fuse_forget_link *next;
64 /** Unique ID, which identifies the inode between userspace
68 /** Number of lookups on this inode */
71 /** The request used for sending the FORGET message */
72 struct fuse_forget_link *forget;
74 /** Time in jiffies until the file attributes are valid */
77 /** The sticky bit in inode->i_mode may have been removed, so
78 preserve the original mode */
81 /** 64 bit inode number */
84 /** Version of last attribute change */
87 /** Files usable in writepage. Protected by fc->lock */
88 struct list_head write_files;
90 /** Writepages pending on truncate or fsync */
91 struct list_head queued_writes;
93 /** Number of sent writes, a negative bias (FUSE_NOWRITE)
94 * means more writes are blocked */
97 /** Waitq for writepage completion */
98 wait_queue_head_t page_waitq;
100 /** List of writepage requestst (pending or sent) */
101 struct list_head writepages;
103 /** Miscellaneous bits describing inode state */
106 /** Lock for serializing lookup and readdir for back compatibility*/
110 /** FUSE inode state bits */
112 /** Advise readdirplus */
113 FUSE_I_ADVISE_RDPLUS,
114 /** Initialized with readdirplus */
116 /** An operation changing file size is in progress */
117 FUSE_I_SIZE_UNSTABLE,
124 /** FUSE specific file data */
126 /** Fuse connection for this file */
127 struct fuse_conn *fc;
129 /** Request reserved for flush and release */
130 struct fuse_req *reserved_req;
132 /** Kernel file handle guaranteed to be unique */
135 /** File handle used by userspace */
138 /** Node id of this file */
144 /** FOPEN_* flags returned by open */
147 /** Entry on inode's write_files list */
148 struct list_head write_entry;
150 /** RB node to be linked on fuse_conn->polled_files */
151 struct rb_node polled_node;
153 /** Wait queue head for poll */
154 wait_queue_head_t poll_wait;
156 /** Has flock been performed on this file? */
160 /** One input argument of a request */
166 /** The request input */
168 /** The request header */
169 struct fuse_in_header h;
171 /** True if the data for the last argument is in req->pages */
174 /** Number of arguments */
177 /** Array of arguments */
178 struct fuse_in_arg args[3];
181 /** One output argument of a request */
187 /** The request output */
189 /** Header returned from userspace */
190 struct fuse_out_header h;
193 * The following bitfields are not changed during the request
197 /** Last argument is variable length (can be shorter than
201 /** Last argument is a list of pages to copy data to */
204 /** Zero partially or not copied pages */
205 unsigned page_zeroing:1;
207 /** Pages may be replaced with new ones */
208 unsigned page_replace:1;
210 /** Number or arguments */
213 /** Array of arguments */
214 struct fuse_arg args[2];
217 /** FUSE page descriptor */
218 struct fuse_page_desc {
230 struct fuse_in_arg args[3];
236 struct fuse_arg args[2];
240 #define FUSE_ARGS(args) struct fuse_args args = {}
242 /** The request IO state (for asynchronous processing) */
243 struct fuse_io_priv {
255 struct completion *done;
259 #define FUSE_IO_PRIV_SYNC(f) \
261 .refcnt = { ATOMIC_INIT(1) }, \
269 * FR_ISREPLY: set if the request has reply
270 * FR_FORCE: force sending of the request even if interrupted
271 * FR_BACKGROUND: request is sent in the background
272 * FR_WAITING: request is counted as "waiting"
273 * FR_ABORTED: the request was aborted
274 * FR_INTERRUPTED: the request has been interrupted
275 * FR_LOCKED: data is being copied to/from the request
276 * FR_PENDING: request is not yet in userspace
277 * FR_SENT: request is in userspace, waiting for an answer
278 * FR_FINISHED: request is finished
279 * FR_PRIVATE: request is on private list
296 * A request to the client
298 * .waitq.lock protects the following fields:
300 * - FR_LOCKED (may also be modified under fc->lock, tested under both)
303 /** This can be on either pending processing or io lists in
305 struct list_head list;
307 /** Entry on the interrupts list */
308 struct list_head intr_entry;
313 /** Unique ID for the interrupt request */
316 /* Request flags, updated with test/set/clear_bit() */
319 /** The request input */
322 /** The request output */
325 /** Used to wake up the task waiting for completion of request*/
326 wait_queue_head_t waitq;
328 /** Data for asynchronous requests */
331 struct fuse_release_in in;
334 struct fuse_init_in init_in;
335 struct fuse_init_out init_out;
336 struct cuse_init_in cuse_init_in;
338 struct fuse_read_in in;
342 struct fuse_write_in in;
343 struct fuse_write_out out;
344 struct fuse_req *next;
346 struct fuse_notify_retrieve_in retrieve_in;
352 /** page-descriptor vector */
353 struct fuse_page_desc *page_descs;
355 /** size of the 'pages' array */
358 /** inline page vector */
359 struct page *inline_pages[FUSE_REQ_INLINE_PAGES];
361 /** inline page-descriptor vector */
362 struct fuse_page_desc inline_page_descs[FUSE_REQ_INLINE_PAGES];
364 /** number of pages in vector */
367 /** File used in the request (or NULL) */
368 struct fuse_file *ff;
370 /** Inode used in the request or NULL */
373 /** AIO control block */
374 struct fuse_io_priv *io;
376 /** Link on fi->writepages */
377 struct list_head writepages_entry;
379 /** Request completion callback */
380 void (*end)(struct fuse_conn *, struct fuse_req *);
382 /** Request is stolen from fuse_file->reserved_req */
383 struct file *stolen_file;
387 /** Connection established */
390 /** Readers of the connection are waiting on this */
391 wait_queue_head_t waitq;
393 /** The next unique request id */
396 /** The list of pending requests */
397 struct list_head pending;
399 /** Pending interrupts */
400 struct list_head interrupts;
402 /** Queue of pending forgets */
403 struct fuse_forget_link forget_list_head;
404 struct fuse_forget_link *forget_list_tail;
406 /** Batching of FORGET requests (positive indicates FORGET batch) */
409 /** O_ASYNC requests */
410 struct fasync_struct *fasync;
414 /** Connection established */
417 /** Lock protecting accessess to members of this structure */
420 /** The list of requests being processed */
421 struct list_head processing;
423 /** The list of requests under I/O */
428 * Fuse device instance
431 /** Fuse connection for this device */
432 struct fuse_conn *fc;
434 /** Processing queue */
435 struct fuse_pqueue pq;
437 /** list entry on fc->devices */
438 struct list_head entry;
444 * This structure is created, when the filesystem is mounted, and is
445 * destroyed, when the client device is closed and the filesystem is
449 /** Lock protecting accessess to members of this structure */
455 /** Number of fuse_dev's */
460 /** The user id for this mount */
463 /** The group id for this mount */
466 /** Maximum read size */
469 /** Maximum write size */
473 struct fuse_iqueue iq;
475 /** The next unique kernel file handle */
478 /** rbtree of fuse_files waiting for poll events indexed by ph */
479 struct rb_root polled_files;
481 /** Maximum number of outstanding background requests */
482 unsigned max_background;
484 /** Number of background requests at which congestion starts */
485 unsigned congestion_threshold;
487 /** Number of requests currently in the background */
488 unsigned num_background;
490 /** Number of background requests currently queued for userspace */
491 unsigned active_background;
493 /** The list of background requests set aside for later queuing */
494 struct list_head bg_queue;
496 /** Flag indicating that INIT reply has been received. Allocating
497 * any fuse request will be suspended until the flag is set */
500 /** Flag indicating if connection is blocked. This will be
501 the case before the INIT reply is received, and if there
502 are too many outstading backgrounds requests */
505 /** waitq for blocked connection */
506 wait_queue_head_t blocked_waitq;
508 /** waitq for reserved requests */
509 wait_queue_head_t reserved_req_waitq;
511 /** Connection established, cleared on umount, connection
512 abort and device release */
515 /** Connection failed (version mismatch). Cannot race with
516 setting other bitfields since it is only set once in INIT
517 reply, before any other request, and never cleared */
518 unsigned conn_error:1;
520 /** Connection successful. Only set in INIT */
521 unsigned conn_init:1;
523 /** Do readpages asynchronously? Only set in INIT */
524 unsigned async_read:1;
526 /** Do not send separate SETATTR request before open(O_TRUNC) */
527 unsigned atomic_o_trunc:1;
529 /** Filesystem supports NFS exporting. Only set in INIT */
530 unsigned export_support:1;
532 /** Set if bdi is valid */
533 unsigned bdi_initialized:1;
535 /** write-back cache policy (default is write-through) */
536 unsigned writeback_cache:1;
538 /** allow parallel lookups and readdir (default is serialized) */
539 unsigned parallel_dirops:1;
541 /** handle fs handles killing suid/sgid/cap on write/chown/trunc */
542 unsigned handle_killpriv:1;
545 * The following bitfields are only for optimization purposes
546 * and hence races in setting them will not cause malfunction
549 /** Is open/release not implemented by fs? */
552 /** Is fsync not implemented by fs? */
555 /** Is fsyncdir not implemented by fs? */
556 unsigned no_fsyncdir:1;
558 /** Is flush not implemented by fs? */
561 /** Is setxattr not implemented by fs? */
562 unsigned no_setxattr:1;
564 /** Is getxattr not implemented by fs? */
565 unsigned no_getxattr:1;
567 /** Is listxattr not implemented by fs? */
568 unsigned no_listxattr:1;
570 /** Is removexattr not implemented by fs? */
571 unsigned no_removexattr:1;
573 /** Are posix file locking primitives not implemented by fs? */
576 /** Is access not implemented by fs? */
577 unsigned no_access:1;
579 /** Is create not implemented by fs? */
580 unsigned no_create:1;
582 /** Is interrupt not implemented by fs? */
583 unsigned no_interrupt:1;
585 /** Is bmap not implemented by fs? */
588 /** Is poll not implemented by fs? */
591 /** Do multi-page cached writes */
592 unsigned big_writes:1;
594 /** Don't apply umask to creation modes */
595 unsigned dont_mask:1;
597 /** Are BSD file locking primitives not implemented by fs? */
600 /** Is fallocate not implemented by fs? */
601 unsigned no_fallocate:1;
603 /** Is rename with flags implemented by fs? */
604 unsigned no_rename2:1;
606 /** Use enhanced/automatic page cache invalidation. */
607 unsigned auto_inval_data:1;
609 /** Does the filesystem support readdirplus? */
610 unsigned do_readdirplus:1;
612 /** Does the filesystem want adaptive readdirplus? */
613 unsigned readdirplus_auto:1;
615 /** Does the filesystem support asynchronous direct-IO submission? */
616 unsigned async_dio:1;
618 /** Is lseek not implemented by fs? */
621 /** Does the filesystem support posix acls? */
622 unsigned posix_acl:1;
624 /** Check permissions based on the file mode or not? */
625 unsigned default_permissions:1;
627 /** Allow other than the mounter user to access the filesystem ? */
628 unsigned allow_other:1;
630 /** The number of requests waiting for completion */
631 atomic_t num_waiting;
633 /** Negotiated minor version */
636 /** Backing dev info */
637 struct backing_dev_info bdi;
639 /** Entry on the fuse_conn_list */
640 struct list_head entry;
642 /** Device ID from super block */
645 /** Dentries in the control filesystem */
646 struct dentry *ctl_dentry[FUSE_CTL_NUM_DENTRIES];
648 /** number of dentries used in the above array */
651 /** Key for lock owner ID scrambling */
654 /** Reserved request for the DESTROY message */
655 struct fuse_req *destroy_req;
657 /** Version counter for attribute changes */
660 /** Called on final put */
661 void (*release)(struct fuse_conn *);
663 /** Super block for this connection. */
664 struct super_block *sb;
666 /** Read/write semaphore to hold when accessing sb. */
667 struct rw_semaphore killsb;
669 /** List of device instances belonging to this connection */
670 struct list_head devices;
673 static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb)
675 return sb->s_fs_info;
678 static inline struct fuse_conn *get_fuse_conn(struct inode *inode)
680 return get_fuse_conn_super(inode->i_sb);
683 static inline struct fuse_inode *get_fuse_inode(struct inode *inode)
685 return container_of(inode, struct fuse_inode, inode);
688 static inline u64 get_node_id(struct inode *inode)
690 return get_fuse_inode(inode)->nodeid;
693 static inline void fuse_make_bad(struct inode *inode)
695 remove_inode_hash(inode);
696 set_bit(FUSE_I_BAD, &get_fuse_inode(inode)->state);
699 static inline bool fuse_is_bad(struct inode *inode)
701 return unlikely(test_bit(FUSE_I_BAD, &get_fuse_inode(inode)->state));
704 /** Device operations */
705 extern const struct file_operations fuse_dev_operations;
707 extern const struct dentry_operations fuse_dentry_operations;
708 extern const struct dentry_operations fuse_root_dentry_operations;
711 * Inode to nodeid comparison.
713 int fuse_inode_eq(struct inode *inode, void *_nodeidp);
716 * Get a filled in inode
718 struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
719 int generation, struct fuse_attr *attr,
720 u64 attr_valid, u64 attr_version);
722 int fuse_lookup_name(struct super_block *sb, u64 nodeid, const struct qstr *name,
723 struct fuse_entry_out *outarg, struct inode **inode);
726 * Send FORGET command
728 void fuse_queue_forget(struct fuse_conn *fc, struct fuse_forget_link *forget,
729 u64 nodeid, u64 nlookup);
731 struct fuse_forget_link *fuse_alloc_forget(void);
733 /* Used by READDIRPLUS */
734 void fuse_force_forget(struct file *file, u64 nodeid);
737 * Initialize READ or READDIR request
739 void fuse_read_fill(struct fuse_req *req, struct file *file,
740 loff_t pos, size_t count, int opcode);
743 * Send OPEN or OPENDIR request
745 int fuse_open_common(struct inode *inode, struct file *file, bool isdir);
747 struct fuse_file *fuse_file_alloc(struct fuse_conn *fc);
748 struct fuse_file *fuse_file_get(struct fuse_file *ff);
749 void fuse_file_free(struct fuse_file *ff);
750 void fuse_finish_open(struct inode *inode, struct file *file);
752 void fuse_sync_release(struct fuse_file *ff, int flags);
755 * Send RELEASE or RELEASEDIR request
757 void fuse_release_common(struct file *file, int opcode);
760 * Send FSYNC or FSYNCDIR request
762 int fuse_fsync_common(struct file *file, loff_t start, loff_t end,
763 int datasync, int isdir);
768 int fuse_notify_poll_wakeup(struct fuse_conn *fc,
769 struct fuse_notify_poll_wakeup_out *outarg);
772 * Initialize file operations on a regular file
774 void fuse_init_file_inode(struct inode *inode);
777 * Initialize inode operations on regular files and special files
779 void fuse_init_common(struct inode *inode);
782 * Initialize inode and file operations on a directory
784 void fuse_init_dir(struct inode *inode);
787 * Initialize inode operations on a symlink
789 void fuse_init_symlink(struct inode *inode);
792 * Change attributes of an inode
794 void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
795 u64 attr_valid, u64 attr_version);
797 void fuse_change_attributes_common(struct inode *inode, struct fuse_attr *attr,
801 * Initialize the client device
803 int fuse_dev_init(void);
806 * Cleanup the client device
808 void fuse_dev_cleanup(void);
810 int fuse_ctl_init(void);
811 void __exit fuse_ctl_cleanup(void);
816 struct fuse_req *fuse_request_alloc(unsigned npages);
818 struct fuse_req *fuse_request_alloc_nofs(unsigned npages);
823 void fuse_request_free(struct fuse_req *req);
826 * Get a request, may fail with -ENOMEM,
827 * caller should specify # elements in req->pages[] explicitly
829 struct fuse_req *fuse_get_req(struct fuse_conn *fc, unsigned npages);
830 struct fuse_req *fuse_get_req_for_background(struct fuse_conn *fc,
834 * Increment reference count on request
836 void __fuse_get_request(struct fuse_req *req);
839 * Gets a requests for a file operation, always succeeds
841 struct fuse_req *fuse_get_req_nofail_nopages(struct fuse_conn *fc,
845 * Decrement reference count of a request. If count goes to zero free
848 void fuse_put_request(struct fuse_conn *fc, struct fuse_req *req);
851 * Send a request (synchronous)
853 void fuse_request_send(struct fuse_conn *fc, struct fuse_req *req);
856 * Simple request sending that does request allocation and freeing
858 ssize_t fuse_simple_request(struct fuse_conn *fc, struct fuse_args *args);
861 * Send a request in the background
863 void fuse_request_send_background(struct fuse_conn *fc, struct fuse_req *req);
865 void fuse_request_send_background_locked(struct fuse_conn *fc,
866 struct fuse_req *req);
868 /* Abort all requests */
869 void fuse_abort_conn(struct fuse_conn *fc);
870 void fuse_wait_aborted(struct fuse_conn *fc);
873 * Invalidate inode attributes
875 void fuse_invalidate_attr(struct inode *inode);
877 void fuse_invalidate_entry_cache(struct dentry *entry);
879 void fuse_invalidate_atime(struct inode *inode);
882 * Acquire reference to fuse_conn
884 struct fuse_conn *fuse_conn_get(struct fuse_conn *fc);
887 * Initialize fuse_conn
889 void fuse_conn_init(struct fuse_conn *fc);
892 * Release reference to fuse_conn
894 void fuse_conn_put(struct fuse_conn *fc);
896 struct fuse_dev *fuse_dev_alloc(struct fuse_conn *fc);
897 void fuse_dev_free(struct fuse_dev *fud);
900 * Add connection to control filesystem
902 int fuse_ctl_add_conn(struct fuse_conn *fc);
905 * Remove connection from control filesystem
907 void fuse_ctl_remove_conn(struct fuse_conn *fc);
910 * Is file type valid?
912 int fuse_valid_type(int m);
914 bool fuse_invalid_attr(struct fuse_attr *attr);
917 * Is current process allowed to perform filesystem operation?
919 int fuse_allow_current_process(struct fuse_conn *fc);
921 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id);
923 void fuse_update_ctime(struct inode *inode);
925 int fuse_update_attributes(struct inode *inode, struct kstat *stat,
926 struct file *file, bool *refreshed);
928 void fuse_flush_writepages(struct inode *inode);
930 void fuse_set_nowrite(struct inode *inode);
931 void fuse_release_nowrite(struct inode *inode);
933 u64 fuse_get_attr_version(struct fuse_conn *fc);
936 * File-system tells the kernel to invalidate cache for the given node id.
938 int fuse_reverse_inval_inode(struct super_block *sb, u64 nodeid,
939 loff_t offset, loff_t len);
942 * File-system tells the kernel to invalidate parent attributes and
943 * the dentry matching parent/name.
945 * If the child_nodeid is non-zero and:
946 * - matches the inode number for the dentry matching parent/name,
947 * - is not a mount point
948 * - is a file or oan empty directory
949 * then the dentry is unhashed (d_delete()).
951 int fuse_reverse_inval_entry(struct super_block *sb, u64 parent_nodeid,
952 u64 child_nodeid, struct qstr *name);
954 int fuse_do_open(struct fuse_conn *fc, u64 nodeid, struct file *file,
958 * fuse_direct_io() flags
961 /** If set, it is WRITE; otherwise - READ */
962 #define FUSE_DIO_WRITE (1 << 0)
964 /** CUSE pass fuse_direct_io() a file which f_mapping->host is not from FUSE */
965 #define FUSE_DIO_CUSE (1 << 1)
967 ssize_t fuse_direct_io(struct fuse_io_priv *io, struct iov_iter *iter,
968 loff_t *ppos, int flags);
969 long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
971 long fuse_ioctl_common(struct file *file, unsigned int cmd,
972 unsigned long arg, unsigned int flags);
973 unsigned fuse_file_poll(struct file *file, poll_table *wait);
974 int fuse_dev_release(struct inode *inode, struct file *file);
976 bool fuse_write_update_size(struct inode *inode, loff_t pos);
978 int fuse_flush_times(struct inode *inode, struct fuse_file *ff);
979 int fuse_write_inode(struct inode *inode, struct writeback_control *wbc);
981 int fuse_do_setattr(struct dentry *dentry, struct iattr *attr,
984 void fuse_set_initialized(struct fuse_conn *fc);
986 void fuse_unlock_inode(struct inode *inode, bool locked);
987 bool fuse_lock_inode(struct inode *inode);
989 int fuse_setxattr(struct inode *inode, const char *name, const void *value,
990 size_t size, int flags);
991 ssize_t fuse_getxattr(struct inode *inode, const char *name, void *value,
993 ssize_t fuse_listxattr(struct dentry *entry, char *list, size_t size);
994 int fuse_removexattr(struct inode *inode, const char *name);
995 extern const struct xattr_handler *fuse_xattr_handlers[];
996 extern const struct xattr_handler *fuse_acl_xattr_handlers[];
999 struct posix_acl *fuse_get_acl(struct inode *inode, int type);
1000 int fuse_set_acl(struct inode *inode, struct posix_acl *acl, int type);
1002 #endif /* _FS_FUSE_I_H */