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,
122 /** FUSE specific file data */
124 /** Fuse connection for this file */
125 struct fuse_conn *fc;
127 /** Request reserved for flush and release */
128 struct fuse_req *reserved_req;
130 /** Kernel file handle guaranteed to be unique */
133 /** File handle used by userspace */
136 /** Node id of this file */
142 /** FOPEN_* flags returned by open */
145 /** Entry on inode's write_files list */
146 struct list_head write_entry;
148 /** RB node to be linked on fuse_conn->polled_files */
149 struct rb_node polled_node;
151 /** Wait queue head for poll */
152 wait_queue_head_t poll_wait;
154 /** Has flock been performed on this file? */
158 /** One input argument of a request */
164 /** The request input */
166 /** The request header */
167 struct fuse_in_header h;
169 /** True if the data for the last argument is in req->pages */
172 /** Number of arguments */
175 /** Array of arguments */
176 struct fuse_in_arg args[3];
179 /** One output argument of a request */
185 /** The request output */
187 /** Header returned from userspace */
188 struct fuse_out_header h;
191 * The following bitfields are not changed during the request
195 /** Last argument is variable length (can be shorter than
199 /** Last argument is a list of pages to copy data to */
202 /** Zero partially or not copied pages */
203 unsigned page_zeroing:1;
205 /** Pages may be replaced with new ones */
206 unsigned page_replace:1;
208 /** Number or arguments */
211 /** Array of arguments */
212 struct fuse_arg args[3];
215 /** FUSE page descriptor */
216 struct fuse_page_desc {
221 /** The request state */
222 enum fuse_req_state {
232 * A request to the client
235 /** This can be on either pending processing or io lists in
237 struct list_head list;
239 /** Entry on the interrupts list */
240 struct list_head intr_entry;
245 /** Unique ID for the interrupt request */
249 * The following bitfields are either set once before the
250 * request is queued or setting/clearing them is protected by
254 /** True if the request has reply */
257 /** Force sending of the request even if interrupted */
260 /** The request was aborted */
263 /** Request is sent in the background */
264 unsigned background:1;
266 /** The request has been interrupted */
267 unsigned interrupted:1;
269 /** Data is being copied to/from the request */
272 /** Request is counted as "waiting" */
275 /** State of the request */
276 enum fuse_req_state state;
278 /** The request input */
281 /** The request output */
284 /** Used to wake up the task waiting for completion of request*/
285 wait_queue_head_t waitq;
287 /** Data for asynchronous requests */
291 struct fuse_release_in in;
292 struct work_struct work;
296 struct fuse_init_in init_in;
297 struct fuse_init_out init_out;
298 struct cuse_init_in cuse_init_in;
300 struct fuse_read_in in;
304 struct fuse_write_in in;
305 struct fuse_write_out out;
307 struct fuse_notify_retrieve_in retrieve_in;
308 struct fuse_lk_in lk_in;
314 /** page-descriptor vector */
315 struct fuse_page_desc *page_descs;
317 /** size of the 'pages' array */
320 /** inline page vector */
321 struct page *inline_pages[FUSE_REQ_INLINE_PAGES];
323 /** inline page-descriptor vector */
324 struct fuse_page_desc inline_page_descs[FUSE_REQ_INLINE_PAGES];
326 /** number of pages in vector */
329 /** File used in the request (or NULL) */
330 struct fuse_file *ff;
332 /** Inode used in the request or NULL */
335 /** Link on fi->writepages */
336 struct list_head writepages_entry;
338 /** Request completion callback */
339 void (*end)(struct fuse_conn *, struct fuse_req *);
341 /** Request is stolen from fuse_file->reserved_req */
342 struct file *stolen_file;
348 * This structure is created, when the filesystem is mounted, and is
349 * destroyed, when the client device is closed and the filesystem is
353 /** Lock protecting accessess to members of this structure */
356 /** Mutex protecting against directory alias creation */
357 struct mutex inst_mutex;
362 /** The user id for this mount */
365 /** The group id for this mount */
368 /** The fuse mount flags for this mount */
371 /** Maximum read size */
374 /** Maximum write size */
377 /** Readers of the connection are waiting on this */
378 wait_queue_head_t waitq;
380 /** The list of pending requests */
381 struct list_head pending;
383 /** The list of requests being processed */
384 struct list_head processing;
386 /** The list of requests under I/O */
389 /** The next unique kernel file handle */
392 /** rbtree of fuse_files waiting for poll events indexed by ph */
393 struct rb_root polled_files;
395 /** Maximum number of outstanding background requests */
396 unsigned max_background;
398 /** Number of background requests at which congestion starts */
399 unsigned congestion_threshold;
401 /** Number of requests currently in the background */
402 unsigned num_background;
404 /** Number of background requests currently queued for userspace */
405 unsigned active_background;
407 /** The list of background requests set aside for later queuing */
408 struct list_head bg_queue;
410 /** Pending interrupts */
411 struct list_head interrupts;
413 /** Queue of pending forgets */
414 struct fuse_forget_link forget_list_head;
415 struct fuse_forget_link *forget_list_tail;
417 /** Batching of FORGET requests (positive indicates FORGET batch) */
420 /** Flag indicating if connection is blocked. This will be
421 the case before the INIT reply is received, and if there
422 are too many outstading backgrounds requests */
425 /** waitq for blocked connection */
426 wait_queue_head_t blocked_waitq;
428 /** waitq for reserved requests */
429 wait_queue_head_t reserved_req_waitq;
431 /** The next unique request id */
434 /** Connection established, cleared on umount, connection
435 abort and device release */
438 /** Connection failed (version mismatch). Cannot race with
439 setting other bitfields since it is only set once in INIT
440 reply, before any other request, and never cleared */
441 unsigned conn_error:1;
443 /** Connection successful. Only set in INIT */
444 unsigned conn_init:1;
446 /** Do readpages asynchronously? Only set in INIT */
447 unsigned async_read:1;
449 /** Do not send separate SETATTR request before open(O_TRUNC) */
450 unsigned atomic_o_trunc:1;
452 /** Filesystem supports NFS exporting. Only set in INIT */
453 unsigned export_support:1;
455 /** Set if bdi is valid */
456 unsigned bdi_initialized:1;
459 * The following bitfields are only for optimization purposes
460 * and hence races in setting them will not cause malfunction
463 /** Is fsync not implemented by fs? */
466 /** Is fsyncdir not implemented by fs? */
467 unsigned no_fsyncdir:1;
469 /** Is flush not implemented by fs? */
472 /** Is setxattr not implemented by fs? */
473 unsigned no_setxattr:1;
475 /** Is getxattr not implemented by fs? */
476 unsigned no_getxattr:1;
478 /** Is listxattr not implemented by fs? */
479 unsigned no_listxattr:1;
481 /** Is removexattr not implemented by fs? */
482 unsigned no_removexattr:1;
484 /** Are posix file locking primitives not implemented by fs? */
487 /** Is access not implemented by fs? */
488 unsigned no_access:1;
490 /** Is create not implemented by fs? */
491 unsigned no_create:1;
493 /** Is interrupt not implemented by fs? */
494 unsigned no_interrupt:1;
496 /** Is bmap not implemented by fs? */
499 /** Is poll not implemented by fs? */
502 /** Do multi-page cached writes */
503 unsigned big_writes:1;
505 /** Don't apply umask to creation modes */
506 unsigned dont_mask:1;
508 /** Are BSD file locking primitives not implemented by fs? */
511 /** Is fallocate not implemented by fs? */
512 unsigned no_fallocate:1;
514 /** Use enhanced/automatic page cache invalidation. */
515 unsigned auto_inval_data:1;
517 /** Does the filesystem support readdirplus? */
518 unsigned do_readdirplus:1;
520 /** Does the filesystem want adaptive readdirplus? */
521 unsigned readdirplus_auto:1;
523 /** The number of requests waiting for completion */
524 atomic_t num_waiting;
526 /** Negotiated minor version */
529 /** Backing dev info */
530 struct backing_dev_info bdi;
532 /** Entry on the fuse_conn_list */
533 struct list_head entry;
535 /** Device ID from super block */
538 /** Dentries in the control filesystem */
539 struct dentry *ctl_dentry[FUSE_CTL_NUM_DENTRIES];
541 /** number of dentries used in the above array */
544 /** O_ASYNC requests */
545 struct fasync_struct *fasync;
547 /** Key for lock owner ID scrambling */
550 /** Reserved request for the DESTROY message */
551 struct fuse_req *destroy_req;
553 /** Version counter for attribute changes */
556 /** Called on final put */
557 void (*release)(struct fuse_conn *);
559 /** Super block for this connection. */
560 struct super_block *sb;
562 /** Read/write semaphore to hold when accessing sb. */
563 struct rw_semaphore killsb;
566 static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb)
568 return sb->s_fs_info;
571 static inline struct fuse_conn *get_fuse_conn(struct inode *inode)
573 return get_fuse_conn_super(inode->i_sb);
576 static inline struct fuse_inode *get_fuse_inode(struct inode *inode)
578 return container_of(inode, struct fuse_inode, inode);
581 static inline u64 get_node_id(struct inode *inode)
583 return get_fuse_inode(inode)->nodeid;
586 /** Device operations */
587 extern const struct file_operations fuse_dev_operations;
589 extern const struct dentry_operations fuse_dentry_operations;
592 * Inode to nodeid comparison.
594 int fuse_inode_eq(struct inode *inode, void *_nodeidp);
597 * Get a filled in inode
599 struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
600 int generation, struct fuse_attr *attr,
601 u64 attr_valid, u64 attr_version);
603 int fuse_lookup_name(struct super_block *sb, u64 nodeid, struct qstr *name,
604 struct fuse_entry_out *outarg, struct inode **inode);
607 * Send FORGET command
609 void fuse_queue_forget(struct fuse_conn *fc, struct fuse_forget_link *forget,
610 u64 nodeid, u64 nlookup);
612 struct fuse_forget_link *fuse_alloc_forget(void);
614 /* Used by READDIRPLUS */
615 void fuse_force_forget(struct file *file, u64 nodeid);
618 * Initialize READ or READDIR request
620 void fuse_read_fill(struct fuse_req *req, struct file *file,
621 loff_t pos, size_t count, int opcode);
624 * Send OPEN or OPENDIR request
626 int fuse_open_common(struct inode *inode, struct file *file, bool isdir);
628 struct fuse_file *fuse_file_alloc(struct fuse_conn *fc);
629 struct fuse_file *fuse_file_get(struct fuse_file *ff);
630 void fuse_file_free(struct fuse_file *ff);
631 void fuse_finish_open(struct inode *inode, struct file *file);
633 void fuse_sync_release(struct fuse_file *ff, int flags);
636 * Send RELEASE or RELEASEDIR request
638 void fuse_release_common(struct file *file, int opcode);
641 * Send FSYNC or FSYNCDIR request
643 int fuse_fsync_common(struct file *file, loff_t start, loff_t end,
644 int datasync, int isdir);
649 int fuse_notify_poll_wakeup(struct fuse_conn *fc,
650 struct fuse_notify_poll_wakeup_out *outarg);
653 * Initialize file operations on a regular file
655 void fuse_init_file_inode(struct inode *inode);
658 * Initialize inode operations on regular files and special files
660 void fuse_init_common(struct inode *inode);
663 * Initialize inode and file operations on a directory
665 void fuse_init_dir(struct inode *inode);
668 * Initialize inode operations on a symlink
670 void fuse_init_symlink(struct inode *inode);
673 * Change attributes of an inode
675 void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
676 u64 attr_valid, u64 attr_version);
678 void fuse_change_attributes_common(struct inode *inode, struct fuse_attr *attr,
682 * Initialize the client device
684 int fuse_dev_init(void);
687 * Cleanup the client device
689 void fuse_dev_cleanup(void);
691 int fuse_ctl_init(void);
692 void fuse_ctl_cleanup(void);
697 struct fuse_req *fuse_request_alloc(unsigned npages);
699 struct fuse_req *fuse_request_alloc_nofs(unsigned npages);
704 void fuse_request_free(struct fuse_req *req);
707 * Get a request, may fail with -ENOMEM,
708 * caller should specify # elements in req->pages[] explicitly
710 struct fuse_req *fuse_get_req(struct fuse_conn *fc, unsigned npages);
711 struct fuse_req *fuse_get_req_for_background(struct fuse_conn *fc,
715 * Get a request, may fail with -ENOMEM,
716 * useful for callers who doesn't use req->pages[]
718 static inline struct fuse_req *fuse_get_req_nopages(struct fuse_conn *fc)
720 return fuse_get_req(fc, 0);
724 * Gets a requests for a file operation, always succeeds
726 struct fuse_req *fuse_get_req_nofail_nopages(struct fuse_conn *fc,
730 * Decrement reference count of a request. If count goes to zero free
733 void fuse_put_request(struct fuse_conn *fc, struct fuse_req *req);
736 * Send a request (synchronous)
738 void fuse_request_send(struct fuse_conn *fc, struct fuse_req *req);
741 * Send a request in the background
743 void fuse_request_send_background(struct fuse_conn *fc, struct fuse_req *req);
745 void fuse_request_send_background_locked(struct fuse_conn *fc,
746 struct fuse_req *req);
748 /* Abort all requests */
749 void fuse_abort_conn(struct fuse_conn *fc);
752 * Invalidate inode attributes
754 void fuse_invalidate_attr(struct inode *inode);
756 void fuse_invalidate_entry_cache(struct dentry *entry);
759 * Acquire reference to fuse_conn
761 struct fuse_conn *fuse_conn_get(struct fuse_conn *fc);
763 void fuse_conn_kill(struct fuse_conn *fc);
766 * Initialize fuse_conn
768 void fuse_conn_init(struct fuse_conn *fc);
771 * Release reference to fuse_conn
773 void fuse_conn_put(struct fuse_conn *fc);
776 * Add connection to control filesystem
778 int fuse_ctl_add_conn(struct fuse_conn *fc);
781 * Remove connection from control filesystem
783 void fuse_ctl_remove_conn(struct fuse_conn *fc);
786 * Is file type valid?
788 int fuse_valid_type(int m);
791 * Is current process allowed to perform filesystem operation?
793 int fuse_allow_current_process(struct fuse_conn *fc);
795 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id);
797 int fuse_update_attributes(struct inode *inode, struct kstat *stat,
798 struct file *file, bool *refreshed);
800 void fuse_flush_writepages(struct inode *inode);
802 void fuse_set_nowrite(struct inode *inode);
803 void fuse_release_nowrite(struct inode *inode);
805 u64 fuse_get_attr_version(struct fuse_conn *fc);
808 * File-system tells the kernel to invalidate cache for the given node id.
810 int fuse_reverse_inval_inode(struct super_block *sb, u64 nodeid,
811 loff_t offset, loff_t len);
814 * File-system tells the kernel to invalidate parent attributes and
815 * the dentry matching parent/name.
817 * If the child_nodeid is non-zero and:
818 * - matches the inode number for the dentry matching parent/name,
819 * - is not a mount point
820 * - is a file or oan empty directory
821 * then the dentry is unhashed (d_delete()).
823 int fuse_reverse_inval_entry(struct super_block *sb, u64 parent_nodeid,
824 u64 child_nodeid, struct qstr *name);
826 int fuse_do_open(struct fuse_conn *fc, u64 nodeid, struct file *file,
828 ssize_t fuse_direct_io(struct file *file, const struct iovec *iov,
829 unsigned long nr_segs, size_t count, loff_t *ppos,
831 long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
833 long fuse_ioctl_common(struct file *file, unsigned int cmd,
834 unsigned long arg, unsigned int flags);
835 unsigned fuse_file_poll(struct file *file, poll_table *wait);
836 int fuse_dev_release(struct inode *inode, struct file *file);
838 void fuse_write_update_size(struct inode *inode, loff_t pos);
840 #endif /* _FS_FUSE_I_H */