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>
25 /** Max number of pages that can be used in a single read request */
26 #define FUSE_MAX_PAGES_PER_REQ 32
28 /** Bias for fi->writectr, meaning new writepages must not be sent */
29 #define FUSE_NOWRITE INT_MIN
31 /** It could be as large as PATH_MAX, but would that have any uses? */
32 #define FUSE_NAME_MAX 1024
34 /** Number of dentries for each connection in the control filesystem */
35 #define FUSE_CTL_NUM_DENTRIES 5
37 /** If the FUSE_DEFAULT_PERMISSIONS flag is given, the filesystem
38 module will check permissions based on the file mode. Otherwise no
39 permission checking is done in the kernel */
40 #define FUSE_DEFAULT_PERMISSIONS (1 << 0)
42 /** If the FUSE_ALLOW_OTHER flag is given, then not only the user
43 doing the mount will be allowed to access the filesystem */
44 #define FUSE_ALLOW_OTHER (1 << 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;
61 /** Unique ID, which identifies the inode between userspace
65 /** Number of lookups on this inode */
68 /** The request used for sending the FORGET message */
69 struct fuse_req *forget_req;
71 /** Time in jiffies until the file attributes are valid */
74 /** The sticky bit in inode->i_mode may have been removed, so
75 preserve the original mode */
78 /** Version of last attribute change */
81 /** Files usable in writepage. Protected by fc->lock */
82 struct list_head write_files;
84 /** Writepages pending on truncate or fsync */
85 struct list_head queued_writes;
87 /** Number of sent writes, a negative bias (FUSE_NOWRITE)
88 * means more writes are blocked */
91 /** Waitq for writepage completion */
92 wait_queue_head_t page_waitq;
94 /** List of writepage requestst (pending or sent) */
95 struct list_head writepages;
100 /** FUSE specific file data */
102 /** Fuse connection for this file */
103 struct fuse_conn *fc;
105 /** Request reserved for flush and release */
106 struct fuse_req *reserved_req;
108 /** Kernel file handle guaranteed to be unique */
111 /** File handle used by userspace */
114 /** Node id of this file */
120 /** FOPEN_* flags returned by open */
123 /** Entry on inode's write_files list */
124 struct list_head write_entry;
126 /** RB node to be linked on fuse_conn->polled_files */
127 struct rb_node polled_node;
129 /** Wait queue head for poll */
130 wait_queue_head_t poll_wait;
133 /** One input argument of a request */
139 /** The request input */
141 /** The request header */
142 struct fuse_in_header h;
144 /** True if the data for the last argument is in req->pages */
147 /** Number of arguments */
150 /** Array of arguments */
151 struct fuse_in_arg args[3];
154 /** One output argument of a request */
160 /** The request output */
162 /** Header returned from userspace */
163 struct fuse_out_header h;
166 * The following bitfields are not changed during the request
170 /** Last argument is variable length (can be shorter than
174 /** Last argument is a list of pages to copy data to */
177 /** Zero partially or not copied pages */
178 unsigned page_zeroing:1;
180 /** Pages may be replaced with new ones */
181 unsigned page_replace:1;
183 /** Number or arguments */
186 /** Array of arguments */
187 struct fuse_arg args[3];
190 /** The request state */
191 enum fuse_req_state {
201 * A request to the client
204 /** This can be on either pending processing or io lists in
206 struct list_head list;
208 /** Entry on the interrupts list */
209 struct list_head intr_entry;
214 /** Unique ID for the interrupt request */
218 * The following bitfields are either set once before the
219 * request is queued or setting/clearing them is protected by
223 /** True if the request has reply */
226 /** Force sending of the request even if interrupted */
229 /** The request was aborted */
232 /** Request is sent in the background */
233 unsigned background:1;
235 /** The request has been interrupted */
236 unsigned interrupted:1;
238 /** Data is being copied to/from the request */
241 /** Request is counted as "waiting" */
244 /** State of the request */
245 enum fuse_req_state state;
247 /** The request input */
250 /** The request output */
253 /** Used to wake up the task waiting for completion of request*/
254 wait_queue_head_t waitq;
256 /** Data for asynchronous requests */
258 struct fuse_forget_in forget_in;
260 struct fuse_release_in in;
263 struct fuse_init_in init_in;
264 struct fuse_init_out init_out;
265 struct cuse_init_in cuse_init_in;
266 struct cuse_init_out cuse_init_out;
268 struct fuse_read_in in;
272 struct fuse_write_in in;
273 struct fuse_write_out out;
275 struct fuse_notify_retrieve_in retrieve_in;
276 struct fuse_lk_in lk_in;
280 struct page *pages[FUSE_MAX_PAGES_PER_REQ];
282 /** number of pages in vector */
285 /** offset of data on first page */
286 unsigned page_offset;
288 /** File used in the request (or NULL) */
289 struct fuse_file *ff;
291 /** Inode used in the request or NULL */
294 /** Link on fi->writepages */
295 struct list_head writepages_entry;
297 /** Request completion callback */
298 void (*end)(struct fuse_conn *, struct fuse_req *);
300 /** Request is stolen from fuse_file->reserved_req */
301 struct file *stolen_file;
307 * This structure is created, when the filesystem is mounted, and is
308 * destroyed, when the client device is closed and the filesystem is
312 /** Lock protecting accessess to members of this structure */
315 /** Mutex protecting against directory alias creation */
316 struct mutex inst_mutex;
321 /** The user id for this mount */
324 /** The group id for this mount */
327 /** The fuse mount flags for this mount */
330 /** Maximum read size */
333 /** Maximum write size */
336 /** Readers of the connection are waiting on this */
337 wait_queue_head_t waitq;
339 /** The list of pending requests */
340 struct list_head pending;
342 /** The list of requests being processed */
343 struct list_head processing;
345 /** The list of requests under I/O */
348 /** The next unique kernel file handle */
351 /** rbtree of fuse_files waiting for poll events indexed by ph */
352 struct rb_root polled_files;
354 /** Maximum number of outstanding background requests */
355 unsigned max_background;
357 /** Number of background requests at which congestion starts */
358 unsigned congestion_threshold;
360 /** Number of requests currently in the background */
361 unsigned num_background;
363 /** Number of background requests currently queued for userspace */
364 unsigned active_background;
366 /** The list of background requests set aside for later queuing */
367 struct list_head bg_queue;
369 /** Pending interrupts */
370 struct list_head interrupts;
372 /** Flag indicating if connection is blocked. This will be
373 the case before the INIT reply is received, and if there
374 are too many outstading backgrounds requests */
377 /** waitq for blocked connection */
378 wait_queue_head_t blocked_waitq;
380 /** waitq for reserved requests */
381 wait_queue_head_t reserved_req_waitq;
383 /** The next unique request id */
386 /** Connection established, cleared on umount, connection
387 abort and device release */
390 /** Connection failed (version mismatch). Cannot race with
391 setting other bitfields since it is only set once in INIT
392 reply, before any other request, and never cleared */
393 unsigned conn_error:1;
395 /** Connection successful. Only set in INIT */
396 unsigned conn_init:1;
398 /** Do readpages asynchronously? Only set in INIT */
399 unsigned async_read:1;
401 /** Do not send separate SETATTR request before open(O_TRUNC) */
402 unsigned atomic_o_trunc:1;
404 /** Filesystem supports NFS exporting. Only set in INIT */
405 unsigned export_support:1;
407 /** Set if bdi is valid */
408 unsigned bdi_initialized:1;
411 * The following bitfields are only for optimization purposes
412 * and hence races in setting them will not cause malfunction
415 /** Is fsync not implemented by fs? */
418 /** Is fsyncdir not implemented by fs? */
419 unsigned no_fsyncdir:1;
421 /** Is flush not implemented by fs? */
424 /** Is setxattr not implemented by fs? */
425 unsigned no_setxattr:1;
427 /** Is getxattr not implemented by fs? */
428 unsigned no_getxattr:1;
430 /** Is listxattr not implemented by fs? */
431 unsigned no_listxattr:1;
433 /** Is removexattr not implemented by fs? */
434 unsigned no_removexattr:1;
436 /** Are file locking primitives not implemented by fs? */
439 /** Is access not implemented by fs? */
440 unsigned no_access:1;
442 /** Is create not implemented by fs? */
443 unsigned no_create:1;
445 /** Is interrupt not implemented by fs? */
446 unsigned no_interrupt:1;
448 /** Is bmap not implemented by fs? */
451 /** Is poll not implemented by fs? */
454 /** Do multi-page cached writes */
455 unsigned big_writes:1;
457 /** Don't apply umask to creation modes */
458 unsigned dont_mask:1;
460 /** The number of requests waiting for completion */
461 atomic_t num_waiting;
463 /** Negotiated minor version */
466 /** Backing dev info */
467 struct backing_dev_info bdi;
469 /** Entry on the fuse_conn_list */
470 struct list_head entry;
472 /** Device ID from super block */
475 /** Dentries in the control filesystem */
476 struct dentry *ctl_dentry[FUSE_CTL_NUM_DENTRIES];
478 /** number of dentries used in the above array */
481 /** O_ASYNC requests */
482 struct fasync_struct *fasync;
484 /** Key for lock owner ID scrambling */
487 /** Reserved request for the DESTROY message */
488 struct fuse_req *destroy_req;
490 /** Version counter for attribute changes */
493 /** Called on final put */
494 void (*release)(struct fuse_conn *);
496 /** Super block for this connection. */
497 struct super_block *sb;
499 /** Read/write semaphore to hold when accessing sb. */
500 struct rw_semaphore killsb;
503 static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb)
505 return sb->s_fs_info;
508 static inline struct fuse_conn *get_fuse_conn(struct inode *inode)
510 return get_fuse_conn_super(inode->i_sb);
513 static inline struct fuse_inode *get_fuse_inode(struct inode *inode)
515 return container_of(inode, struct fuse_inode, inode);
518 static inline u64 get_node_id(struct inode *inode)
520 return get_fuse_inode(inode)->nodeid;
523 /** Device operations */
524 extern const struct file_operations fuse_dev_operations;
526 extern const struct dentry_operations fuse_dentry_operations;
529 * Inode to nodeid comparison.
531 int fuse_inode_eq(struct inode *inode, void *_nodeidp);
534 * Get a filled in inode
536 struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
537 int generation, struct fuse_attr *attr,
538 u64 attr_valid, u64 attr_version);
540 int fuse_lookup_name(struct super_block *sb, u64 nodeid, struct qstr *name,
541 struct fuse_entry_out *outarg, struct inode **inode);
544 * Send FORGET command
546 void fuse_send_forget(struct fuse_conn *fc, struct fuse_req *req,
547 u64 nodeid, u64 nlookup);
550 * Initialize READ or READDIR request
552 void fuse_read_fill(struct fuse_req *req, struct file *file,
553 loff_t pos, size_t count, int opcode);
556 * Send OPEN or OPENDIR request
558 int fuse_open_common(struct inode *inode, struct file *file, bool isdir);
560 struct fuse_file *fuse_file_alloc(struct fuse_conn *fc);
561 struct fuse_file *fuse_file_get(struct fuse_file *ff);
562 void fuse_file_free(struct fuse_file *ff);
563 void fuse_finish_open(struct inode *inode, struct file *file);
565 void fuse_sync_release(struct fuse_file *ff, int flags);
568 * Send RELEASE or RELEASEDIR request
570 void fuse_release_common(struct file *file, int opcode);
573 * Send FSYNC or FSYNCDIR request
575 int fuse_fsync_common(struct file *file, int datasync, int isdir);
580 int fuse_notify_poll_wakeup(struct fuse_conn *fc,
581 struct fuse_notify_poll_wakeup_out *outarg);
584 * Initialize file operations on a regular file
586 void fuse_init_file_inode(struct inode *inode);
589 * Initialize inode operations on regular files and special files
591 void fuse_init_common(struct inode *inode);
594 * Initialize inode and file operations on a directory
596 void fuse_init_dir(struct inode *inode);
599 * Initialize inode operations on a symlink
601 void fuse_init_symlink(struct inode *inode);
604 * Change attributes of an inode
606 void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
607 u64 attr_valid, u64 attr_version);
609 void fuse_change_attributes_common(struct inode *inode, struct fuse_attr *attr,
613 * Initialize the client device
615 int fuse_dev_init(void);
618 * Cleanup the client device
620 void fuse_dev_cleanup(void);
622 int fuse_ctl_init(void);
623 void fuse_ctl_cleanup(void);
628 struct fuse_req *fuse_request_alloc(void);
630 struct fuse_req *fuse_request_alloc_nofs(void);
635 void fuse_request_free(struct fuse_req *req);
638 * Get a request, may fail with -ENOMEM
640 struct fuse_req *fuse_get_req(struct fuse_conn *fc);
643 * Gets a requests for a file operation, always succeeds
645 struct fuse_req *fuse_get_req_nofail(struct fuse_conn *fc, struct file *file);
648 * Decrement reference count of a request. If count goes to zero free
651 void fuse_put_request(struct fuse_conn *fc, struct fuse_req *req);
654 * Send a request (synchronous)
656 void fuse_request_send(struct fuse_conn *fc, struct fuse_req *req);
659 * Send a request with no reply
661 void fuse_request_send_noreply(struct fuse_conn *fc, struct fuse_req *req);
664 * Send a request in the background
666 void fuse_request_send_background(struct fuse_conn *fc, struct fuse_req *req);
668 void fuse_request_send_background_locked(struct fuse_conn *fc,
669 struct fuse_req *req);
671 /* Abort all requests */
672 void fuse_abort_conn(struct fuse_conn *fc);
675 * Invalidate inode attributes
677 void fuse_invalidate_attr(struct inode *inode);
679 void fuse_invalidate_entry_cache(struct dentry *entry);
682 * Acquire reference to fuse_conn
684 struct fuse_conn *fuse_conn_get(struct fuse_conn *fc);
686 void fuse_conn_kill(struct fuse_conn *fc);
689 * Initialize fuse_conn
691 void fuse_conn_init(struct fuse_conn *fc);
694 * Release reference to fuse_conn
696 void fuse_conn_put(struct fuse_conn *fc);
699 * Add connection to control filesystem
701 int fuse_ctl_add_conn(struct fuse_conn *fc);
704 * Remove connection from control filesystem
706 void fuse_ctl_remove_conn(struct fuse_conn *fc);
709 * Is file type valid?
711 int fuse_valid_type(int m);
714 * Is task allowed to perform filesystem operation?
716 int fuse_allow_task(struct fuse_conn *fc, struct task_struct *task);
718 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id);
720 int fuse_update_attributes(struct inode *inode, struct kstat *stat,
721 struct file *file, bool *refreshed);
723 void fuse_flush_writepages(struct inode *inode);
725 void fuse_set_nowrite(struct inode *inode);
726 void fuse_release_nowrite(struct inode *inode);
728 u64 fuse_get_attr_version(struct fuse_conn *fc);
731 * File-system tells the kernel to invalidate cache for the given node id.
733 int fuse_reverse_inval_inode(struct super_block *sb, u64 nodeid,
734 loff_t offset, loff_t len);
737 * File-system tells the kernel to invalidate parent attributes and
738 * the dentry matching parent/name.
740 int fuse_reverse_inval_entry(struct super_block *sb, u64 parent_nodeid,
743 int fuse_do_open(struct fuse_conn *fc, u64 nodeid, struct file *file,
745 ssize_t fuse_direct_io(struct file *file, const char __user *buf,
746 size_t count, loff_t *ppos, int write);
747 long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
749 unsigned fuse_file_poll(struct file *file, poll_table *wait);
750 int fuse_dev_release(struct inode *inode, struct file *file);
752 void fuse_write_update_size(struct inode *inode, loff_t pos);
754 #endif /* _FS_FUSE_I_H */