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/refcount.h>
29 /** Max number of pages that can be used in a single read request */
30 #define FUSE_MAX_PAGES_PER_REQ 32
32 /** Bias for fi->writectr, meaning new writepages must not be sent */
33 #define FUSE_NOWRITE INT_MIN
35 /** It could be as large as PATH_MAX, but would that have any uses? */
36 #define FUSE_NAME_MAX 1024
38 /** Number of dentries for each connection in the control filesystem */
39 #define FUSE_CTL_NUM_DENTRIES 5
41 /** Number of page pointers embedded in fuse_req */
42 #define FUSE_REQ_INLINE_PAGES 1
44 /** List of active connections */
45 extern struct list_head fuse_conn_list;
47 /** Global mutex protecting fuse_conn_list and the control filesystem */
48 extern struct mutex fuse_mutex;
50 /** Module parameters */
51 extern unsigned max_user_bgreq;
52 extern unsigned max_user_congthresh;
54 /* One forget request */
55 struct fuse_forget_link {
56 struct fuse_forget_one forget_one;
57 struct fuse_forget_link *next;
65 /** Unique ID, which identifies the inode between userspace
69 /** Number of lookups on this inode */
72 /** The request used for sending the FORGET message */
73 struct fuse_forget_link *forget;
75 /** Time in jiffies until the file attributes are valid */
78 /** The sticky bit in inode->i_mode may have been removed, so
79 preserve the original mode */
82 /** 64 bit inode number */
85 /** Version of last attribute change */
88 /** Files usable in writepage. Protected by fc->lock */
89 struct list_head write_files;
91 /** Writepages pending on truncate or fsync */
92 struct list_head queued_writes;
94 /** Number of sent writes, a negative bias (FUSE_NOWRITE)
95 * means more writes are blocked */
98 /** Waitq for writepage completion */
99 wait_queue_head_t page_waitq;
101 /** List of writepage requestst (pending or sent) */
102 struct list_head writepages;
104 /** Miscellaneous bits describing inode state */
107 /** Lock for serializing lookup and readdir for back compatibility*/
111 /** FUSE inode state bits */
113 /** Advise readdirplus */
114 FUSE_I_ADVISE_RDPLUS,
115 /** Initialized with readdirplus */
117 /** An operation changing file size is in progress */
118 FUSE_I_SIZE_UNSTABLE,
123 /** FUSE specific file data */
125 /** Fuse connection for this file */
126 struct fuse_conn *fc;
128 /** Request reserved for flush and release */
129 struct fuse_req *reserved_req;
131 /** Kernel file handle guaranteed to be unique */
134 /** File handle used by userspace */
137 /** Node id of this file */
143 /** FOPEN_* flags returned by open */
146 /** Entry on inode's write_files list */
147 struct list_head write_entry;
149 /** RB node to be linked on fuse_conn->polled_files */
150 struct rb_node polled_node;
152 /** Wait queue head for poll */
153 wait_queue_head_t poll_wait;
155 /** Has flock been performed on this file? */
159 /** One input argument of a request */
165 /** The request input */
167 /** The request header */
168 struct fuse_in_header h;
170 /** True if the data for the last argument is in req->pages */
173 /** Number of arguments */
176 /** Array of arguments */
177 struct fuse_in_arg args[3];
180 /** One output argument of a request */
186 /** The request output */
188 /** Header returned from userspace */
189 struct fuse_out_header h;
192 * The following bitfields are not changed during the request
196 /** Last argument is variable length (can be shorter than
200 /** Last argument is a list of pages to copy data to */
203 /** Zero partially or not copied pages */
204 unsigned page_zeroing:1;
206 /** Pages may be replaced with new ones */
207 unsigned page_replace:1;
209 /** Number or arguments */
212 /** Array of arguments */
213 struct fuse_arg args[2];
216 /** FUSE page descriptor */
217 struct fuse_page_desc {
229 struct fuse_in_arg args[3];
235 struct fuse_arg args[2];
239 #define FUSE_ARGS(args) struct fuse_args args = {}
241 /** The request IO state (for asynchronous processing) */
242 struct fuse_io_priv {
254 struct completion *done;
258 #define FUSE_IO_PRIV_SYNC(f) \
260 .refcnt = KREF_INIT(1), \
268 * FR_ISREPLY: set if the request has reply
269 * FR_FORCE: force sending of the request even if interrupted
270 * FR_BACKGROUND: request is sent in the background
271 * FR_WAITING: request is counted as "waiting"
272 * FR_ABORTED: the request was aborted
273 * FR_INTERRUPTED: the request has been interrupted
274 * FR_LOCKED: data is being copied to/from the request
275 * FR_PENDING: request is not yet in userspace
276 * FR_SENT: request is in userspace, waiting for an answer
277 * FR_FINISHED: request is finished
278 * FR_PRIVATE: request is on private list
295 * A request to the client
297 * .waitq.lock protects the following fields:
299 * - FR_LOCKED (may also be modified under fc->lock, tested under both)
302 /** This can be on either pending processing or io lists in
304 struct list_head list;
306 /** Entry on the interrupts list */
307 struct list_head intr_entry;
312 /** Unique ID for the interrupt request */
315 /* Request flags, updated with test/set/clear_bit() */
318 /** The request input */
321 /** The request output */
324 /** Used to wake up the task waiting for completion of request*/
325 wait_queue_head_t waitq;
327 /** Data for asynchronous requests */
330 struct fuse_release_in in;
333 struct fuse_init_in init_in;
334 struct fuse_init_out init_out;
335 struct cuse_init_in cuse_init_in;
337 struct fuse_read_in in;
341 struct fuse_write_in in;
342 struct fuse_write_out out;
343 struct fuse_req *next;
345 struct fuse_notify_retrieve_in retrieve_in;
351 /** page-descriptor vector */
352 struct fuse_page_desc *page_descs;
354 /** size of the 'pages' array */
357 /** inline page vector */
358 struct page *inline_pages[FUSE_REQ_INLINE_PAGES];
360 /** inline page-descriptor vector */
361 struct fuse_page_desc inline_page_descs[FUSE_REQ_INLINE_PAGES];
363 /** number of pages in vector */
366 /** File used in the request (or NULL) */
367 struct fuse_file *ff;
369 /** Inode used in the request or NULL */
372 /** AIO control block */
373 struct fuse_io_priv *io;
375 /** Link on fi->writepages */
376 struct list_head writepages_entry;
378 /** Request completion callback */
379 void (*end)(struct fuse_conn *, struct fuse_req *);
381 /** Request is stolen from fuse_file->reserved_req */
382 struct file *stolen_file;
386 /** Connection established */
389 /** Readers of the connection are waiting on this */
390 wait_queue_head_t waitq;
392 /** The next unique request id */
395 /** The list of pending requests */
396 struct list_head pending;
398 /** Pending interrupts */
399 struct list_head interrupts;
401 /** Queue of pending forgets */
402 struct fuse_forget_link forget_list_head;
403 struct fuse_forget_link *forget_list_tail;
405 /** Batching of FORGET requests (positive indicates FORGET batch) */
408 /** O_ASYNC requests */
409 struct fasync_struct *fasync;
413 /** Connection established */
416 /** Lock protecting accessess to members of this structure */
419 /** The list of requests being processed */
420 struct list_head processing;
422 /** The list of requests under I/O */
427 * Fuse device instance
430 /** Fuse connection for this device */
431 struct fuse_conn *fc;
433 /** Processing queue */
434 struct fuse_pqueue pq;
436 /** list entry on fc->devices */
437 struct list_head entry;
443 * This structure is created, when the filesystem is mounted, and is
444 * destroyed, when the client device is closed and the filesystem is
448 /** Lock protecting accessess to members of this structure */
454 /** Number of fuse_dev's */
459 /** The user id for this mount */
462 /** The group id for this mount */
465 /** Maximum read size */
468 /** Maximum write size */
472 struct fuse_iqueue iq;
474 /** The next unique kernel file handle */
477 /** rbtree of fuse_files waiting for poll events indexed by ph */
478 struct rb_root polled_files;
480 /** Maximum number of outstanding background requests */
481 unsigned max_background;
483 /** Number of background requests at which congestion starts */
484 unsigned congestion_threshold;
486 /** Number of requests currently in the background */
487 unsigned num_background;
489 /** Number of background requests currently queued for userspace */
490 unsigned active_background;
492 /** The list of background requests set aside for later queuing */
493 struct list_head bg_queue;
495 /** Flag indicating that INIT reply has been received. Allocating
496 * any fuse request will be suspended until the flag is set */
499 /** Flag indicating if connection is blocked. This will be
500 the case before the INIT reply is received, and if there
501 are too many outstading backgrounds requests */
504 /** waitq for blocked connection */
505 wait_queue_head_t blocked_waitq;
507 /** waitq for reserved requests */
508 wait_queue_head_t reserved_req_waitq;
510 /** Connection established, cleared on umount, connection
511 abort and device release */
514 /** Connection failed (version mismatch). Cannot race with
515 setting other bitfields since it is only set once in INIT
516 reply, before any other request, and never cleared */
517 unsigned conn_error:1;
519 /** Connection successful. Only set in INIT */
520 unsigned conn_init:1;
522 /** Do readpages asynchronously? Only set in INIT */
523 unsigned async_read:1;
525 /** Do not send separate SETATTR request before open(O_TRUNC) */
526 unsigned atomic_o_trunc:1;
528 /** Filesystem supports NFS exporting. Only set in INIT */
529 unsigned export_support:1;
531 /** Set if bdi is valid */
532 unsigned bdi_initialized:1;
534 /** write-back cache policy (default is write-through) */
535 unsigned writeback_cache:1;
537 /** allow parallel lookups and readdir (default is serialized) */
538 unsigned parallel_dirops:1;
540 /** handle fs handles killing suid/sgid/cap on write/chown/trunc */
541 unsigned handle_killpriv:1;
544 * The following bitfields are only for optimization purposes
545 * and hence races in setting them will not cause malfunction
548 /** Is open/release not implemented by fs? */
551 /** Is fsync not implemented by fs? */
554 /** Is fsyncdir not implemented by fs? */
555 unsigned no_fsyncdir:1;
557 /** Is flush not implemented by fs? */
560 /** Is setxattr not implemented by fs? */
561 unsigned no_setxattr:1;
563 /** Is getxattr not implemented by fs? */
564 unsigned no_getxattr:1;
566 /** Is listxattr not implemented by fs? */
567 unsigned no_listxattr:1;
569 /** Is removexattr not implemented by fs? */
570 unsigned no_removexattr:1;
572 /** Are posix file locking primitives not implemented by fs? */
575 /** Is access not implemented by fs? */
576 unsigned no_access:1;
578 /** Is create not implemented by fs? */
579 unsigned no_create:1;
581 /** Is interrupt not implemented by fs? */
582 unsigned no_interrupt:1;
584 /** Is bmap not implemented by fs? */
587 /** Is poll not implemented by fs? */
590 /** Do multi-page cached writes */
591 unsigned big_writes:1;
593 /** Don't apply umask to creation modes */
594 unsigned dont_mask:1;
596 /** Are BSD file locking primitives not implemented by fs? */
599 /** Is fallocate not implemented by fs? */
600 unsigned no_fallocate:1;
602 /** Is rename with flags implemented by fs? */
603 unsigned no_rename2:1;
605 /** Use enhanced/automatic page cache invalidation. */
606 unsigned auto_inval_data:1;
608 /** Does the filesystem support readdirplus? */
609 unsigned do_readdirplus:1;
611 /** Does the filesystem want adaptive readdirplus? */
612 unsigned readdirplus_auto:1;
614 /** Does the filesystem support asynchronous direct-IO submission? */
615 unsigned async_dio:1;
617 /** Is lseek not implemented by fs? */
620 /** Does the filesystem support posix acls? */
621 unsigned posix_acl:1;
623 /** Check permissions based on the file mode or not? */
624 unsigned default_permissions:1;
626 /** Allow other than the mounter user to access the filesystem ? */
627 unsigned allow_other:1;
629 /** The number of requests waiting for completion */
630 atomic_t num_waiting;
632 /** Negotiated minor version */
635 /** Backing dev info */
636 struct backing_dev_info bdi;
638 /** Entry on the fuse_conn_list */
639 struct list_head entry;
641 /** Device ID from super block */
644 /** Dentries in the control filesystem */
645 struct dentry *ctl_dentry[FUSE_CTL_NUM_DENTRIES];
647 /** number of dentries used in the above array */
650 /** Key for lock owner ID scrambling */
653 /** Reserved request for the DESTROY message */
654 struct fuse_req *destroy_req;
656 /** Version counter for attribute changes */
659 /** Called on final put */
660 void (*release)(struct fuse_conn *);
662 /** Super block for this connection. */
663 struct super_block *sb;
665 /** Read/write semaphore to hold when accessing sb. */
666 struct rw_semaphore killsb;
668 /** List of device instances belonging to this connection */
669 struct list_head devices;
672 static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb)
674 return sb->s_fs_info;
677 static inline struct fuse_conn *get_fuse_conn(struct inode *inode)
679 return get_fuse_conn_super(inode->i_sb);
682 static inline struct fuse_inode *get_fuse_inode(struct inode *inode)
684 return container_of(inode, struct fuse_inode, inode);
687 static inline u64 get_node_id(struct inode *inode)
689 return get_fuse_inode(inode)->nodeid;
692 /** Device operations */
693 extern const struct file_operations fuse_dev_operations;
695 extern const struct dentry_operations fuse_dentry_operations;
696 extern const struct dentry_operations fuse_root_dentry_operations;
699 * Inode to nodeid comparison.
701 int fuse_inode_eq(struct inode *inode, void *_nodeidp);
704 * Get a filled in inode
706 struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
707 int generation, struct fuse_attr *attr,
708 u64 attr_valid, u64 attr_version);
710 int fuse_lookup_name(struct super_block *sb, u64 nodeid, const struct qstr *name,
711 struct fuse_entry_out *outarg, struct inode **inode);
714 * Send FORGET command
716 void fuse_queue_forget(struct fuse_conn *fc, struct fuse_forget_link *forget,
717 u64 nodeid, u64 nlookup);
719 struct fuse_forget_link *fuse_alloc_forget(void);
721 /* Used by READDIRPLUS */
722 void fuse_force_forget(struct file *file, u64 nodeid);
725 * Initialize READ or READDIR request
727 void fuse_read_fill(struct fuse_req *req, struct file *file,
728 loff_t pos, size_t count, int opcode);
731 * Send OPEN or OPENDIR request
733 int fuse_open_common(struct inode *inode, struct file *file, bool isdir);
735 struct fuse_file *fuse_file_alloc(struct fuse_conn *fc);
736 void fuse_file_free(struct fuse_file *ff);
737 void fuse_finish_open(struct inode *inode, struct file *file);
739 void fuse_sync_release(struct fuse_file *ff, int flags);
742 * Send RELEASE or RELEASEDIR request
744 void fuse_release_common(struct file *file, int opcode);
747 * Send FSYNC or FSYNCDIR request
749 int fuse_fsync_common(struct file *file, loff_t start, loff_t end,
750 int datasync, int isdir);
755 int fuse_notify_poll_wakeup(struct fuse_conn *fc,
756 struct fuse_notify_poll_wakeup_out *outarg);
759 * Initialize file operations on a regular file
761 void fuse_init_file_inode(struct inode *inode);
764 * Initialize inode operations on regular files and special files
766 void fuse_init_common(struct inode *inode);
769 * Initialize inode and file operations on a directory
771 void fuse_init_dir(struct inode *inode);
774 * Initialize inode operations on a symlink
776 void fuse_init_symlink(struct inode *inode);
779 * Change attributes of an inode
781 void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
782 u64 attr_valid, u64 attr_version);
784 void fuse_change_attributes_common(struct inode *inode, struct fuse_attr *attr,
788 * Initialize the client device
790 int fuse_dev_init(void);
793 * Cleanup the client device
795 void fuse_dev_cleanup(void);
797 int fuse_ctl_init(void);
798 void __exit fuse_ctl_cleanup(void);
803 struct fuse_req *fuse_request_alloc(unsigned npages);
805 struct fuse_req *fuse_request_alloc_nofs(unsigned npages);
810 void fuse_request_free(struct fuse_req *req);
813 * Get a request, may fail with -ENOMEM,
814 * caller should specify # elements in req->pages[] explicitly
816 struct fuse_req *fuse_get_req(struct fuse_conn *fc, unsigned npages);
817 struct fuse_req *fuse_get_req_for_background(struct fuse_conn *fc,
821 * Increment reference count on request
823 void __fuse_get_request(struct fuse_req *req);
826 * Gets a requests for a file operation, always succeeds
828 struct fuse_req *fuse_get_req_nofail_nopages(struct fuse_conn *fc,
832 * Decrement reference count of a request. If count goes to zero free
835 void fuse_put_request(struct fuse_conn *fc, struct fuse_req *req);
838 * Send a request (synchronous)
840 void fuse_request_send(struct fuse_conn *fc, struct fuse_req *req);
843 * Simple request sending that does request allocation and freeing
845 ssize_t fuse_simple_request(struct fuse_conn *fc, struct fuse_args *args);
848 * Send a request in the background
850 void fuse_request_send_background(struct fuse_conn *fc, struct fuse_req *req);
852 void fuse_request_send_background_locked(struct fuse_conn *fc,
853 struct fuse_req *req);
855 /* Abort all requests */
856 void fuse_abort_conn(struct fuse_conn *fc);
859 * Invalidate inode attributes
861 void fuse_invalidate_attr(struct inode *inode);
863 void fuse_invalidate_entry_cache(struct dentry *entry);
865 void fuse_invalidate_atime(struct inode *inode);
868 * Acquire reference to fuse_conn
870 struct fuse_conn *fuse_conn_get(struct fuse_conn *fc);
873 * Initialize fuse_conn
875 void fuse_conn_init(struct fuse_conn *fc);
878 * Release reference to fuse_conn
880 void fuse_conn_put(struct fuse_conn *fc);
882 struct fuse_dev *fuse_dev_alloc(struct fuse_conn *fc);
883 void fuse_dev_free(struct fuse_dev *fud);
886 * Add connection to control filesystem
888 int fuse_ctl_add_conn(struct fuse_conn *fc);
891 * Remove connection from control filesystem
893 void fuse_ctl_remove_conn(struct fuse_conn *fc);
896 * Is file type valid?
898 int fuse_valid_type(int m);
901 * Is current process allowed to perform filesystem operation?
903 int fuse_allow_current_process(struct fuse_conn *fc);
905 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id);
907 void fuse_update_ctime(struct inode *inode);
909 int fuse_update_attributes(struct inode *inode, struct kstat *stat,
910 struct file *file, bool *refreshed);
912 void fuse_flush_writepages(struct inode *inode);
914 void fuse_set_nowrite(struct inode *inode);
915 void fuse_release_nowrite(struct inode *inode);
917 u64 fuse_get_attr_version(struct fuse_conn *fc);
920 * File-system tells the kernel to invalidate cache for the given node id.
922 int fuse_reverse_inval_inode(struct super_block *sb, u64 nodeid,
923 loff_t offset, loff_t len);
926 * File-system tells the kernel to invalidate parent attributes and
927 * the dentry matching parent/name.
929 * If the child_nodeid is non-zero and:
930 * - matches the inode number for the dentry matching parent/name,
931 * - is not a mount point
932 * - is a file or oan empty directory
933 * then the dentry is unhashed (d_delete()).
935 int fuse_reverse_inval_entry(struct super_block *sb, u64 parent_nodeid,
936 u64 child_nodeid, struct qstr *name);
938 int fuse_do_open(struct fuse_conn *fc, u64 nodeid, struct file *file,
942 * fuse_direct_io() flags
945 /** If set, it is WRITE; otherwise - READ */
946 #define FUSE_DIO_WRITE (1 << 0)
948 /** CUSE pass fuse_direct_io() a file which f_mapping->host is not from FUSE */
949 #define FUSE_DIO_CUSE (1 << 1)
951 ssize_t fuse_direct_io(struct fuse_io_priv *io, struct iov_iter *iter,
952 loff_t *ppos, int flags);
953 long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
955 long fuse_ioctl_common(struct file *file, unsigned int cmd,
956 unsigned long arg, unsigned int flags);
957 unsigned fuse_file_poll(struct file *file, poll_table *wait);
958 int fuse_dev_release(struct inode *inode, struct file *file);
960 bool fuse_write_update_size(struct inode *inode, loff_t pos);
962 int fuse_flush_times(struct inode *inode, struct fuse_file *ff);
963 int fuse_write_inode(struct inode *inode, struct writeback_control *wbc);
965 int fuse_do_setattr(struct dentry *dentry, struct iattr *attr,
968 void fuse_set_initialized(struct fuse_conn *fc);
970 void fuse_unlock_inode(struct inode *inode);
971 void fuse_lock_inode(struct inode *inode);
973 int fuse_setxattr(struct inode *inode, const char *name, const void *value,
974 size_t size, int flags);
975 ssize_t fuse_getxattr(struct inode *inode, const char *name, void *value,
977 ssize_t fuse_listxattr(struct dentry *entry, char *list, size_t size);
978 int fuse_removexattr(struct inode *inode, const char *name);
979 extern const struct xattr_handler *fuse_xattr_handlers[];
980 extern const struct xattr_handler *fuse_acl_xattr_handlers[];
983 struct posix_acl *fuse_get_acl(struct inode *inode, int type);
984 int fuse_set_acl(struct inode *inode, struct posix_acl *acl, int type);
986 #endif /* _FS_FUSE_I_H */