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 IO state (for asynchronous processing) */
245 struct fuse_io_priv {
256 struct completion *done;
262 * FR_ISREPLY: set if the request has reply
263 * FR_FORCE: force sending of the request even if interrupted
264 * FR_BACKGROUND: request is sent in the background
265 * FR_WAITING: request is counted as "waiting"
266 * FR_ABORTED: the request was aborted
267 * FR_INTERRUPTED: the request has been interrupted
268 * FR_LOCKED: data is being copied to/from the request
269 * FR_PENDING: request is not yet in userspace
270 * FR_SENT: request is in userspace, waiting for an answer
271 * FR_FINISHED: request is finished
272 * FR_PRIVATE: request is on private list
289 * A request to the client
291 * .waitq.lock protects the following fields:
293 * - FR_LOCKED (may also be modified under fc->lock, tested under both)
296 /** This can be on either pending processing or io lists in
298 struct list_head list;
300 /** Entry on the interrupts list */
301 struct list_head intr_entry;
306 /** Unique ID for the interrupt request */
309 /* Request flags, updated with test/set/clear_bit() */
312 /** The request input */
315 /** The request output */
318 /** Used to wake up the task waiting for completion of request*/
319 wait_queue_head_t waitq;
321 /** Data for asynchronous requests */
324 struct fuse_release_in in;
327 struct fuse_init_in init_in;
328 struct fuse_init_out init_out;
329 struct cuse_init_in cuse_init_in;
331 struct fuse_read_in in;
335 struct fuse_write_in in;
336 struct fuse_write_out out;
337 struct fuse_req *next;
339 struct fuse_notify_retrieve_in retrieve_in;
345 /** page-descriptor vector */
346 struct fuse_page_desc *page_descs;
348 /** size of the 'pages' array */
351 /** inline page vector */
352 struct page *inline_pages[FUSE_REQ_INLINE_PAGES];
354 /** inline page-descriptor vector */
355 struct fuse_page_desc inline_page_descs[FUSE_REQ_INLINE_PAGES];
357 /** number of pages in vector */
360 /** File used in the request (or NULL) */
361 struct fuse_file *ff;
363 /** Inode used in the request or NULL */
366 /** AIO control block */
367 struct fuse_io_priv *io;
369 /** Link on fi->writepages */
370 struct list_head writepages_entry;
372 /** Request completion callback */
373 void (*end)(struct fuse_conn *, struct fuse_req *);
375 /** Request is stolen from fuse_file->reserved_req */
376 struct file *stolen_file;
380 /** Connection established */
383 /** Readers of the connection are waiting on this */
384 wait_queue_head_t waitq;
386 /** The next unique request id */
389 /** The list of pending requests */
390 struct list_head pending;
392 /** Pending interrupts */
393 struct list_head interrupts;
395 /** Queue of pending forgets */
396 struct fuse_forget_link forget_list_head;
397 struct fuse_forget_link *forget_list_tail;
399 /** Batching of FORGET requests (positive indicates FORGET batch) */
402 /** O_ASYNC requests */
403 struct fasync_struct *fasync;
407 /** Connection established */
410 /** Lock protecting accessess to members of this structure */
413 /** The list of requests being processed */
414 struct list_head processing;
416 /** The list of requests under I/O */
423 * This structure is created, when the filesystem is mounted, and is
424 * destroyed, when the client device is closed and the filesystem is
428 /** Lock protecting accessess to members of this structure */
436 /** The user id for this mount */
439 /** The group id for this mount */
442 /** The fuse mount flags for this mount */
445 /** Maximum read size */
448 /** Maximum write size */
452 struct fuse_iqueue iq;
454 /** Processing queue */
455 struct fuse_pqueue pq;
457 /** The next unique kernel file handle */
460 /** rbtree of fuse_files waiting for poll events indexed by ph */
461 struct rb_root polled_files;
463 /** Maximum number of outstanding background requests */
464 unsigned max_background;
466 /** Number of background requests at which congestion starts */
467 unsigned congestion_threshold;
469 /** Number of requests currently in the background */
470 unsigned num_background;
472 /** Number of background requests currently queued for userspace */
473 unsigned active_background;
475 /** The list of background requests set aside for later queuing */
476 struct list_head bg_queue;
478 /** Flag indicating that INIT reply has been received. Allocating
479 * any fuse request will be suspended until the flag is set */
482 /** Flag indicating if connection is blocked. This will be
483 the case before the INIT reply is received, and if there
484 are too many outstading backgrounds requests */
487 /** waitq for blocked connection */
488 wait_queue_head_t blocked_waitq;
490 /** waitq for reserved requests */
491 wait_queue_head_t reserved_req_waitq;
493 /** Connection established, cleared on umount, connection
494 abort and device release */
497 /** Connection failed (version mismatch). Cannot race with
498 setting other bitfields since it is only set once in INIT
499 reply, before any other request, and never cleared */
500 unsigned conn_error:1;
502 /** Connection successful. Only set in INIT */
503 unsigned conn_init:1;
505 /** Do readpages asynchronously? Only set in INIT */
506 unsigned async_read:1;
508 /** Do not send separate SETATTR request before open(O_TRUNC) */
509 unsigned atomic_o_trunc:1;
511 /** Filesystem supports NFS exporting. Only set in INIT */
512 unsigned export_support:1;
514 /** Set if bdi is valid */
515 unsigned bdi_initialized:1;
517 /** write-back cache policy (default is write-through) */
518 unsigned writeback_cache:1;
521 * The following bitfields are only for optimization purposes
522 * and hence races in setting them will not cause malfunction
525 /** Is open/release not implemented by fs? */
528 /** Is fsync not implemented by fs? */
531 /** Is fsyncdir not implemented by fs? */
532 unsigned no_fsyncdir:1;
534 /** Is flush not implemented by fs? */
537 /** Is setxattr not implemented by fs? */
538 unsigned no_setxattr:1;
540 /** Is getxattr not implemented by fs? */
541 unsigned no_getxattr:1;
543 /** Is listxattr not implemented by fs? */
544 unsigned no_listxattr:1;
546 /** Is removexattr not implemented by fs? */
547 unsigned no_removexattr:1;
549 /** Are posix file locking primitives not implemented by fs? */
552 /** Is access not implemented by fs? */
553 unsigned no_access:1;
555 /** Is create not implemented by fs? */
556 unsigned no_create:1;
558 /** Is interrupt not implemented by fs? */
559 unsigned no_interrupt:1;
561 /** Is bmap not implemented by fs? */
564 /** Is poll not implemented by fs? */
567 /** Do multi-page cached writes */
568 unsigned big_writes:1;
570 /** Don't apply umask to creation modes */
571 unsigned dont_mask:1;
573 /** Are BSD file locking primitives not implemented by fs? */
576 /** Is fallocate not implemented by fs? */
577 unsigned no_fallocate:1;
579 /** Is rename with flags implemented by fs? */
580 unsigned no_rename2:1;
582 /** Use enhanced/automatic page cache invalidation. */
583 unsigned auto_inval_data:1;
585 /** Does the filesystem support readdirplus? */
586 unsigned do_readdirplus:1;
588 /** Does the filesystem want adaptive readdirplus? */
589 unsigned readdirplus_auto:1;
591 /** Does the filesystem support asynchronous direct-IO submission? */
592 unsigned async_dio:1;
594 /** The number of requests waiting for completion */
595 atomic_t num_waiting;
597 /** Negotiated minor version */
600 /** Backing dev info */
601 struct backing_dev_info bdi;
603 /** Entry on the fuse_conn_list */
604 struct list_head entry;
606 /** Device ID from super block */
609 /** Dentries in the control filesystem */
610 struct dentry *ctl_dentry[FUSE_CTL_NUM_DENTRIES];
612 /** number of dentries used in the above array */
615 /** Key for lock owner ID scrambling */
618 /** Reserved request for the DESTROY message */
619 struct fuse_req *destroy_req;
621 /** Version counter for attribute changes */
624 /** Called on final put */
625 void (*release)(struct fuse_conn *);
627 /** Super block for this connection. */
628 struct super_block *sb;
630 /** Read/write semaphore to hold when accessing sb. */
631 struct rw_semaphore killsb;
634 static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb)
636 return sb->s_fs_info;
639 static inline struct fuse_conn *get_fuse_conn(struct inode *inode)
641 return get_fuse_conn_super(inode->i_sb);
644 static inline struct fuse_inode *get_fuse_inode(struct inode *inode)
646 return container_of(inode, struct fuse_inode, inode);
649 static inline u64 get_node_id(struct inode *inode)
651 return get_fuse_inode(inode)->nodeid;
654 /** Device operations */
655 extern const struct file_operations fuse_dev_operations;
657 extern const struct dentry_operations fuse_dentry_operations;
660 * Inode to nodeid comparison.
662 int fuse_inode_eq(struct inode *inode, void *_nodeidp);
665 * Get a filled in inode
667 struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
668 int generation, struct fuse_attr *attr,
669 u64 attr_valid, u64 attr_version);
671 int fuse_lookup_name(struct super_block *sb, u64 nodeid, struct qstr *name,
672 struct fuse_entry_out *outarg, struct inode **inode);
675 * Send FORGET command
677 void fuse_queue_forget(struct fuse_conn *fc, struct fuse_forget_link *forget,
678 u64 nodeid, u64 nlookup);
680 struct fuse_forget_link *fuse_alloc_forget(void);
682 /* Used by READDIRPLUS */
683 void fuse_force_forget(struct file *file, u64 nodeid);
686 * Initialize READ or READDIR request
688 void fuse_read_fill(struct fuse_req *req, struct file *file,
689 loff_t pos, size_t count, int opcode);
692 * Send OPEN or OPENDIR request
694 int fuse_open_common(struct inode *inode, struct file *file, bool isdir);
696 struct fuse_file *fuse_file_alloc(struct fuse_conn *fc);
697 struct fuse_file *fuse_file_get(struct fuse_file *ff);
698 void fuse_file_free(struct fuse_file *ff);
699 void fuse_finish_open(struct inode *inode, struct file *file);
701 void fuse_sync_release(struct fuse_file *ff, int flags);
704 * Send RELEASE or RELEASEDIR request
706 void fuse_release_common(struct file *file, int opcode);
709 * Send FSYNC or FSYNCDIR request
711 int fuse_fsync_common(struct file *file, loff_t start, loff_t end,
712 int datasync, int isdir);
717 int fuse_notify_poll_wakeup(struct fuse_conn *fc,
718 struct fuse_notify_poll_wakeup_out *outarg);
721 * Initialize file operations on a regular file
723 void fuse_init_file_inode(struct inode *inode);
726 * Initialize inode operations on regular files and special files
728 void fuse_init_common(struct inode *inode);
731 * Initialize inode and file operations on a directory
733 void fuse_init_dir(struct inode *inode);
736 * Initialize inode operations on a symlink
738 void fuse_init_symlink(struct inode *inode);
741 * Change attributes of an inode
743 void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
744 u64 attr_valid, u64 attr_version);
746 void fuse_change_attributes_common(struct inode *inode, struct fuse_attr *attr,
750 * Initialize the client device
752 int fuse_dev_init(void);
755 * Cleanup the client device
757 void fuse_dev_cleanup(void);
759 int fuse_ctl_init(void);
760 void __exit fuse_ctl_cleanup(void);
765 struct fuse_req *fuse_request_alloc(unsigned npages);
767 struct fuse_req *fuse_request_alloc_nofs(unsigned npages);
772 void fuse_request_free(struct fuse_req *req);
775 * Get a request, may fail with -ENOMEM,
776 * caller should specify # elements in req->pages[] explicitly
778 struct fuse_req *fuse_get_req(struct fuse_conn *fc, unsigned npages);
779 struct fuse_req *fuse_get_req_for_background(struct fuse_conn *fc,
783 * Increment reference count on request
785 void __fuse_get_request(struct fuse_req *req);
788 * Gets a requests for a file operation, always succeeds
790 struct fuse_req *fuse_get_req_nofail_nopages(struct fuse_conn *fc,
794 * Decrement reference count of a request. If count goes to zero free
797 void fuse_put_request(struct fuse_conn *fc, struct fuse_req *req);
800 * Send a request (synchronous)
802 void fuse_request_send(struct fuse_conn *fc, struct fuse_req *req);
805 * Simple request sending that does request allocation and freeing
807 ssize_t fuse_simple_request(struct fuse_conn *fc, struct fuse_args *args);
810 * Send a request in the background
812 void fuse_request_send_background(struct fuse_conn *fc, struct fuse_req *req);
814 void fuse_request_send_background_locked(struct fuse_conn *fc,
815 struct fuse_req *req);
817 /* Abort all requests */
818 void fuse_abort_conn(struct fuse_conn *fc);
821 * Invalidate inode attributes
823 void fuse_invalidate_attr(struct inode *inode);
825 void fuse_invalidate_entry_cache(struct dentry *entry);
827 void fuse_invalidate_atime(struct inode *inode);
830 * Acquire reference to fuse_conn
832 struct fuse_conn *fuse_conn_get(struct fuse_conn *fc);
835 * Initialize fuse_conn
837 void fuse_conn_init(struct fuse_conn *fc);
840 * Release reference to fuse_conn
842 void fuse_conn_put(struct fuse_conn *fc);
845 * Add connection to control filesystem
847 int fuse_ctl_add_conn(struct fuse_conn *fc);
850 * Remove connection from control filesystem
852 void fuse_ctl_remove_conn(struct fuse_conn *fc);
855 * Is file type valid?
857 int fuse_valid_type(int m);
860 * Is current process allowed to perform filesystem operation?
862 int fuse_allow_current_process(struct fuse_conn *fc);
864 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id);
866 int fuse_update_attributes(struct inode *inode, struct kstat *stat,
867 struct file *file, bool *refreshed);
869 void fuse_flush_writepages(struct inode *inode);
871 void fuse_set_nowrite(struct inode *inode);
872 void fuse_release_nowrite(struct inode *inode);
874 u64 fuse_get_attr_version(struct fuse_conn *fc);
877 * File-system tells the kernel to invalidate cache for the given node id.
879 int fuse_reverse_inval_inode(struct super_block *sb, u64 nodeid,
880 loff_t offset, loff_t len);
883 * File-system tells the kernel to invalidate parent attributes and
884 * the dentry matching parent/name.
886 * If the child_nodeid is non-zero and:
887 * - matches the inode number for the dentry matching parent/name,
888 * - is not a mount point
889 * - is a file or oan empty directory
890 * then the dentry is unhashed (d_delete()).
892 int fuse_reverse_inval_entry(struct super_block *sb, u64 parent_nodeid,
893 u64 child_nodeid, struct qstr *name);
895 int fuse_do_open(struct fuse_conn *fc, u64 nodeid, struct file *file,
899 * fuse_direct_io() flags
902 /** If set, it is WRITE; otherwise - READ */
903 #define FUSE_DIO_WRITE (1 << 0)
905 /** CUSE pass fuse_direct_io() a file which f_mapping->host is not from FUSE */
906 #define FUSE_DIO_CUSE (1 << 1)
908 ssize_t fuse_direct_io(struct fuse_io_priv *io, struct iov_iter *iter,
909 loff_t *ppos, int flags);
910 long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
912 long fuse_ioctl_common(struct file *file, unsigned int cmd,
913 unsigned long arg, unsigned int flags);
914 unsigned fuse_file_poll(struct file *file, poll_table *wait);
915 int fuse_dev_release(struct inode *inode, struct file *file);
917 bool fuse_write_update_size(struct inode *inode, loff_t pos);
919 int fuse_flush_times(struct inode *inode, struct fuse_file *ff);
920 int fuse_write_inode(struct inode *inode, struct writeback_control *wbc);
922 int fuse_do_setattr(struct inode *inode, struct iattr *attr,
925 void fuse_set_initialized(struct fuse_conn *fc);
927 #endif /* _FS_FUSE_I_H */