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 /** List of active connections */
48 extern struct list_head fuse_conn_list;
50 /** Global mutex protecting fuse_conn_list and the control filesystem */
51 extern struct mutex fuse_mutex;
53 /** Module parameters */
54 extern unsigned max_user_bgreq;
55 extern unsigned max_user_congthresh;
57 /* One forget request */
58 struct fuse_forget_link {
59 struct fuse_forget_one forget_one;
60 struct fuse_forget_link *next;
68 /** Unique ID, which identifies the inode between userspace
72 /** Number of lookups on this inode */
75 /** The request used for sending the FORGET message */
76 struct fuse_forget_link *forget;
78 /** Time in jiffies until the file attributes are valid */
81 /** The sticky bit in inode->i_mode may have been removed, so
82 preserve the original mode */
85 /** 64 bit inode number */
88 /** Version of last attribute change */
91 /** Files usable in writepage. Protected by fc->lock */
92 struct list_head write_files;
94 /** Writepages pending on truncate or fsync */
95 struct list_head queued_writes;
97 /** Number of sent writes, a negative bias (FUSE_NOWRITE)
98 * means more writes are blocked */
101 /** Waitq for writepage completion */
102 wait_queue_head_t page_waitq;
104 /** List of writepage requestst (pending or sent) */
105 struct list_head writepages;
110 /** FUSE specific file data */
112 /** Fuse connection for this file */
113 struct fuse_conn *fc;
115 /** Request reserved for flush and release */
116 struct fuse_req *reserved_req;
118 /** Kernel file handle guaranteed to be unique */
121 /** File handle used by userspace */
124 /** Node id of this file */
130 /** FOPEN_* flags returned by open */
133 /** Entry on inode's write_files list */
134 struct list_head write_entry;
136 /** RB node to be linked on fuse_conn->polled_files */
137 struct rb_node polled_node;
139 /** Wait queue head for poll */
140 wait_queue_head_t poll_wait;
142 /** Has flock been performed on this file? */
146 /** One input argument of a request */
152 /** The request input */
154 /** The request header */
155 struct fuse_in_header h;
157 /** True if the data for the last argument is in req->pages */
160 /** Number of arguments */
163 /** Array of arguments */
164 struct fuse_in_arg args[3];
167 /** One output argument of a request */
173 /** The request output */
175 /** Header returned from userspace */
176 struct fuse_out_header h;
179 * The following bitfields are not changed during the request
183 /** Last argument is variable length (can be shorter than
187 /** Last argument is a list of pages to copy data to */
190 /** Zero partially or not copied pages */
191 unsigned page_zeroing:1;
193 /** Pages may be replaced with new ones */
194 unsigned page_replace:1;
196 /** Number or arguments */
199 /** Array of arguments */
200 struct fuse_arg args[3];
203 /** The request state */
204 enum fuse_req_state {
214 * A request to the client
217 /** This can be on either pending processing or io lists in
219 struct list_head list;
221 /** Entry on the interrupts list */
222 struct list_head intr_entry;
227 /** Unique ID for the interrupt request */
231 * The following bitfields are either set once before the
232 * request is queued or setting/clearing them is protected by
236 /** True if the request has reply */
239 /** Force sending of the request even if interrupted */
242 /** The request was aborted */
245 /** Request is sent in the background */
246 unsigned background:1;
248 /** The request has been interrupted */
249 unsigned interrupted:1;
251 /** Data is being copied to/from the request */
254 /** Request is counted as "waiting" */
257 /** State of the request */
258 enum fuse_req_state state;
260 /** The request input */
263 /** The request output */
266 /** Used to wake up the task waiting for completion of request*/
267 wait_queue_head_t waitq;
269 /** Data for asynchronous requests */
273 struct fuse_release_in in;
274 struct work_struct work;
278 struct fuse_init_in init_in;
279 struct fuse_init_out init_out;
280 struct cuse_init_in cuse_init_in;
282 struct fuse_read_in in;
286 struct fuse_write_in in;
287 struct fuse_write_out out;
289 struct fuse_notify_retrieve_in retrieve_in;
290 struct fuse_lk_in lk_in;
294 struct page *pages[FUSE_MAX_PAGES_PER_REQ];
296 /** number of pages in vector */
299 /** offset of data on first page */
300 unsigned page_offset;
302 /** File used in the request (or NULL) */
303 struct fuse_file *ff;
305 /** Inode used in the request or NULL */
308 /** Link on fi->writepages */
309 struct list_head writepages_entry;
311 /** Request completion callback */
312 void (*end)(struct fuse_conn *, struct fuse_req *);
314 /** Request is stolen from fuse_file->reserved_req */
315 struct file *stolen_file;
321 * This structure is created, when the filesystem is mounted, and is
322 * destroyed, when the client device is closed and the filesystem is
326 /** Lock protecting accessess to members of this structure */
329 /** Mutex protecting against directory alias creation */
330 struct mutex inst_mutex;
335 /** The user id for this mount */
338 /** The group id for this mount */
341 /** The fuse mount flags for this mount */
344 /** Maximum read size */
347 /** Maximum write size */
350 /** Readers of the connection are waiting on this */
351 wait_queue_head_t waitq;
353 /** The list of pending requests */
354 struct list_head pending;
356 /** The list of requests being processed */
357 struct list_head processing;
359 /** The list of requests under I/O */
362 /** The next unique kernel file handle */
365 /** rbtree of fuse_files waiting for poll events indexed by ph */
366 struct rb_root polled_files;
368 /** Maximum number of outstanding background requests */
369 unsigned max_background;
371 /** Number of background requests at which congestion starts */
372 unsigned congestion_threshold;
374 /** Number of requests currently in the background */
375 unsigned num_background;
377 /** Number of background requests currently queued for userspace */
378 unsigned active_background;
380 /** The list of background requests set aside for later queuing */
381 struct list_head bg_queue;
383 /** Pending interrupts */
384 struct list_head interrupts;
386 /** Queue of pending forgets */
387 struct fuse_forget_link forget_list_head;
388 struct fuse_forget_link *forget_list_tail;
390 /** Batching of FORGET requests (positive indicates FORGET batch) */
393 /** Flag indicating if connection is blocked. This will be
394 the case before the INIT reply is received, and if there
395 are too many outstading backgrounds requests */
398 /** waitq for blocked connection */
399 wait_queue_head_t blocked_waitq;
401 /** waitq for reserved requests */
402 wait_queue_head_t reserved_req_waitq;
404 /** The next unique request id */
407 /** Connection established, cleared on umount, connection
408 abort and device release */
411 /** Connection failed (version mismatch). Cannot race with
412 setting other bitfields since it is only set once in INIT
413 reply, before any other request, and never cleared */
414 unsigned conn_error:1;
416 /** Connection successful. Only set in INIT */
417 unsigned conn_init:1;
419 /** Do readpages asynchronously? Only set in INIT */
420 unsigned async_read:1;
422 /** Do not send separate SETATTR request before open(O_TRUNC) */
423 unsigned atomic_o_trunc:1;
425 /** Filesystem supports NFS exporting. Only set in INIT */
426 unsigned export_support:1;
428 /** Set if bdi is valid */
429 unsigned bdi_initialized:1;
432 * The following bitfields are only for optimization purposes
433 * and hence races in setting them will not cause malfunction
436 /** Is fsync not implemented by fs? */
439 /** Is fsyncdir not implemented by fs? */
440 unsigned no_fsyncdir:1;
442 /** Is flush not implemented by fs? */
445 /** Is setxattr not implemented by fs? */
446 unsigned no_setxattr:1;
448 /** Is getxattr not implemented by fs? */
449 unsigned no_getxattr:1;
451 /** Is listxattr not implemented by fs? */
452 unsigned no_listxattr:1;
454 /** Is removexattr not implemented by fs? */
455 unsigned no_removexattr:1;
457 /** Are posix file locking primitives not implemented by fs? */
460 /** Is access not implemented by fs? */
461 unsigned no_access:1;
463 /** Is create not implemented by fs? */
464 unsigned no_create:1;
466 /** Is interrupt not implemented by fs? */
467 unsigned no_interrupt:1;
469 /** Is bmap not implemented by fs? */
472 /** Is poll not implemented by fs? */
475 /** Do multi-page cached writes */
476 unsigned big_writes:1;
478 /** Don't apply umask to creation modes */
479 unsigned dont_mask:1;
481 /** Are BSD file locking primitives not implemented by fs? */
484 /** Is fallocate not implemented by fs? */
485 unsigned no_fallocate:1;
487 /** Use enhanced/automatic page cache invalidation. */
488 unsigned auto_inval_data:1;
490 /** The number of requests waiting for completion */
491 atomic_t num_waiting;
493 /** Negotiated minor version */
496 /** Backing dev info */
497 struct backing_dev_info bdi;
499 /** Entry on the fuse_conn_list */
500 struct list_head entry;
502 /** Device ID from super block */
505 /** Dentries in the control filesystem */
506 struct dentry *ctl_dentry[FUSE_CTL_NUM_DENTRIES];
508 /** number of dentries used in the above array */
511 /** O_ASYNC requests */
512 struct fasync_struct *fasync;
514 /** Key for lock owner ID scrambling */
517 /** Reserved request for the DESTROY message */
518 struct fuse_req *destroy_req;
520 /** Version counter for attribute changes */
523 /** Called on final put */
524 void (*release)(struct fuse_conn *);
526 /** Super block for this connection. */
527 struct super_block *sb;
529 /** Read/write semaphore to hold when accessing sb. */
530 struct rw_semaphore killsb;
533 static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb)
535 return sb->s_fs_info;
538 static inline struct fuse_conn *get_fuse_conn(struct inode *inode)
540 return get_fuse_conn_super(inode->i_sb);
543 static inline struct fuse_inode *get_fuse_inode(struct inode *inode)
545 return container_of(inode, struct fuse_inode, inode);
548 static inline u64 get_node_id(struct inode *inode)
550 return get_fuse_inode(inode)->nodeid;
553 /** Device operations */
554 extern const struct file_operations fuse_dev_operations;
556 extern const struct dentry_operations fuse_dentry_operations;
559 * Inode to nodeid comparison.
561 int fuse_inode_eq(struct inode *inode, void *_nodeidp);
564 * Get a filled in inode
566 struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
567 int generation, struct fuse_attr *attr,
568 u64 attr_valid, u64 attr_version);
570 int fuse_lookup_name(struct super_block *sb, u64 nodeid, struct qstr *name,
571 struct fuse_entry_out *outarg, struct inode **inode);
574 * Send FORGET command
576 void fuse_queue_forget(struct fuse_conn *fc, struct fuse_forget_link *forget,
577 u64 nodeid, u64 nlookup);
579 struct fuse_forget_link *fuse_alloc_forget(void);
582 * Initialize READ or READDIR request
584 void fuse_read_fill(struct fuse_req *req, struct file *file,
585 loff_t pos, size_t count, int opcode);
588 * Send OPEN or OPENDIR request
590 int fuse_open_common(struct inode *inode, struct file *file, bool isdir);
592 struct fuse_file *fuse_file_alloc(struct fuse_conn *fc);
593 struct fuse_file *fuse_file_get(struct fuse_file *ff);
594 void fuse_file_free(struct fuse_file *ff);
595 void fuse_finish_open(struct inode *inode, struct file *file);
597 void fuse_sync_release(struct fuse_file *ff, int flags);
600 * Send RELEASE or RELEASEDIR request
602 void fuse_release_common(struct file *file, int opcode);
605 * Send FSYNC or FSYNCDIR request
607 int fuse_fsync_common(struct file *file, loff_t start, loff_t end,
608 int datasync, int isdir);
613 int fuse_notify_poll_wakeup(struct fuse_conn *fc,
614 struct fuse_notify_poll_wakeup_out *outarg);
617 * Initialize file operations on a regular file
619 void fuse_init_file_inode(struct inode *inode);
622 * Initialize inode operations on regular files and special files
624 void fuse_init_common(struct inode *inode);
627 * Initialize inode and file operations on a directory
629 void fuse_init_dir(struct inode *inode);
632 * Initialize inode operations on a symlink
634 void fuse_init_symlink(struct inode *inode);
637 * Change attributes of an inode
639 void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
640 u64 attr_valid, u64 attr_version);
642 void fuse_change_attributes_common(struct inode *inode, struct fuse_attr *attr,
646 * Initialize the client device
648 int fuse_dev_init(void);
651 * Cleanup the client device
653 void fuse_dev_cleanup(void);
655 int fuse_ctl_init(void);
656 void fuse_ctl_cleanup(void);
661 struct fuse_req *fuse_request_alloc(void);
663 struct fuse_req *fuse_request_alloc_nofs(void);
668 void fuse_request_free(struct fuse_req *req);
671 * Get a request, may fail with -ENOMEM
673 struct fuse_req *fuse_get_req(struct fuse_conn *fc);
676 * Gets a requests for a file operation, always succeeds
678 struct fuse_req *fuse_get_req_nofail(struct fuse_conn *fc, struct file *file);
681 * Decrement reference count of a request. If count goes to zero free
684 void fuse_put_request(struct fuse_conn *fc, struct fuse_req *req);
687 * Send a request (synchronous)
689 void fuse_request_send(struct fuse_conn *fc, struct fuse_req *req);
692 * Send a request in the background
694 void fuse_request_send_background(struct fuse_conn *fc, struct fuse_req *req);
696 void fuse_request_send_background_locked(struct fuse_conn *fc,
697 struct fuse_req *req);
699 /* Abort all requests */
700 void fuse_abort_conn(struct fuse_conn *fc);
703 * Invalidate inode attributes
705 void fuse_invalidate_attr(struct inode *inode);
707 void fuse_invalidate_entry_cache(struct dentry *entry);
710 * Acquire reference to fuse_conn
712 struct fuse_conn *fuse_conn_get(struct fuse_conn *fc);
714 void fuse_conn_kill(struct fuse_conn *fc);
717 * Initialize fuse_conn
719 void fuse_conn_init(struct fuse_conn *fc);
722 * Release reference to fuse_conn
724 void fuse_conn_put(struct fuse_conn *fc);
727 * Add connection to control filesystem
729 int fuse_ctl_add_conn(struct fuse_conn *fc);
732 * Remove connection from control filesystem
734 void fuse_ctl_remove_conn(struct fuse_conn *fc);
737 * Is file type valid?
739 int fuse_valid_type(int m);
742 * Is task allowed to perform filesystem operation?
744 int fuse_allow_task(struct fuse_conn *fc, struct task_struct *task);
746 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id);
748 int fuse_update_attributes(struct inode *inode, struct kstat *stat,
749 struct file *file, bool *refreshed);
751 void fuse_flush_writepages(struct inode *inode);
753 void fuse_set_nowrite(struct inode *inode);
754 void fuse_release_nowrite(struct inode *inode);
756 u64 fuse_get_attr_version(struct fuse_conn *fc);
759 * File-system tells the kernel to invalidate cache for the given node id.
761 int fuse_reverse_inval_inode(struct super_block *sb, u64 nodeid,
762 loff_t offset, loff_t len);
765 * File-system tells the kernel to invalidate parent attributes and
766 * the dentry matching parent/name.
768 * If the child_nodeid is non-zero and:
769 * - matches the inode number for the dentry matching parent/name,
770 * - is not a mount point
771 * - is a file or oan empty directory
772 * then the dentry is unhashed (d_delete()).
774 int fuse_reverse_inval_entry(struct super_block *sb, u64 parent_nodeid,
775 u64 child_nodeid, struct qstr *name);
777 int fuse_do_open(struct fuse_conn *fc, u64 nodeid, struct file *file,
779 ssize_t fuse_direct_io(struct file *file, const char __user *buf,
780 size_t count, loff_t *ppos, int write);
781 long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
783 long fuse_ioctl_common(struct file *file, unsigned int cmd,
784 unsigned long arg, unsigned int flags);
785 unsigned fuse_file_poll(struct file *file, poll_table *wait);
786 int fuse_dev_release(struct inode *inode, struct file *file);
788 void fuse_write_update_size(struct inode *inode, loff_t pos);
790 #endif /* _FS_FUSE_I_H */