1 /*
2 FUSE: Filesystem in Userspace
3 Copyright (C) 2001-2008 Miklos Szeredi <miklos@szeredi.hu>
4
5 This program can be distributed under the terms of the GNU GPL.
6 See the file COPYING.
7 */
8
9 #ifndef _FS_FUSE_I_H
10 #define _FS_FUSE_I_H
11
12 #include <linux/fuse.h>
13 #include <linux/fs.h>
14 #include <linux/mount.h>
15 #include <linux/wait.h>
16 #include <linux/list.h>
17 #include <linux/spinlock.h>
18 #include <linux/mm.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
26 /** Max number of pages that can be used in a single read request */
27 #define FUSE_MAX_PAGES_PER_REQ 32
28
29 /** Bias for fi->writectr, meaning new writepages must not be sent */
30 #define FUSE_NOWRITE INT_MIN
31
32 /** It could be as large as PATH_MAX, but would that have any uses? */
33 #define FUSE_NAME_MAX 1024
34
35 /** Number of dentries for each connection in the control filesystem */
36 #define FUSE_CTL_NUM_DENTRIES 5
37
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)
42
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)
46
47 /** List of active connections */
48 extern struct list_head fuse_conn_list;
49
50 /** Global mutex protecting fuse_conn_list and the control filesystem */
51 extern struct mutex fuse_mutex;
52
53 /** Module parameters */
54 extern unsigned max_user_bgreq;
55 extern unsigned max_user_congthresh;
56
57 /* One forget request */
58 struct fuse_forget_link {
59 struct fuse_forget_one forget_one;
60 struct fuse_forget_link *next;
61 };
62
63 /** FUSE inode */
64 struct fuse_inode {
65 /** Inode data */
66 struct inode inode;
67
68 /** Unique ID, which identifies the inode between userspace
69 * and kernel */
70 u64 nodeid;
71
72 /** Number of lookups on this inode */
73 u64 nlookup;
74
75 /** The request used for sending the FORGET message */
76 struct fuse_forget_link *forget;
77
78 /** Time in jiffies until the file attributes are valid */
79 u64 i_time;
80
81 /** The sticky bit in inode->i_mode may have been removed, so
82 preserve the original mode */
83 umode_t orig_i_mode;
84
85 /** 64 bit inode number */
86 u64 orig_ino;
87
88 /** Version of last attribute change */
89 u64 attr_version;
90
91 /** Files usable in writepage. Protected by fc->lock */
92 struct list_head write_files;
93
94 /** Writepages pending on truncate or fsync */
95 struct list_head queued_writes;
96
97 /** Number of sent writes, a negative bias (FUSE_NOWRITE)
98 * means more writes are blocked */
99 int writectr;
100
101 /** Waitq for writepage completion */
102 wait_queue_head_t page_waitq;
103
104 /** List of writepage requestst (pending or sent) */
105 struct list_head writepages;
106
107 /** Miscellaneous bits describing inode state */
108 unsigned long state;
109 };
110
111 /** FUSE inode state bits */
112 enum {
113 /** An operation changing file size is in progress */
114 FUSE_I_SIZE_UNSTABLE,
115 };
116
117 struct fuse_conn;
118
119 /** FUSE specific file data */
120 struct fuse_file {
121 /** Fuse connection for this file */
122 struct fuse_conn *fc;
123
124 /** Request reserved for flush and release */
125 struct fuse_req *reserved_req;
126
127 /** Kernel file handle guaranteed to be unique */
128 u64 kh;
129
130 /** File handle used by userspace */
131 u64 fh;
132
133 /** Node id of this file */
134 u64 nodeid;
135
136 /** Refcount */
137 atomic_t count;
138
139 /** FOPEN_* flags returned by open */
140 u32 open_flags;
141
142 /** Entry on inode's write_files list */
143 struct list_head write_entry;
144
145 /** RB node to be linked on fuse_conn->polled_files */
146 struct rb_node polled_node;
147
148 /** Wait queue head for poll */
149 wait_queue_head_t poll_wait;
150
151 /** Has flock been performed on this file? */
152 bool flock:1;
153 };
154
155 /** One input argument of a request */
156 struct fuse_in_arg {
157 unsigned size;
158 const void *value;
159 };
160
161 /** The request input */
162 struct fuse_in {
163 /** The request header */
164 struct fuse_in_header h;
165
166 /** True if the data for the last argument is in req->pages */
167 unsigned argpages:1;
168
169 /** Number of arguments */
170 unsigned numargs;
171
172 /** Array of arguments */
173 struct fuse_in_arg args[3];
174 };
175
176 /** One output argument of a request */
177 struct fuse_arg {
178 unsigned size;
179 void *value;
180 };
181
182 /** The request output */
183 struct fuse_out {
184 /** Header returned from userspace */
185 struct fuse_out_header h;
186
187 /*
188 * The following bitfields are not changed during the request
189 * processing
190 */
191
192 /** Last argument is variable length (can be shorter than
193 arg->size) */
194 unsigned argvar:1;
195
196 /** Last argument is a list of pages to copy data to */
197 unsigned argpages:1;
198
199 /** Zero partially or not copied pages */
200 unsigned page_zeroing:1;
201
202 /** Pages may be replaced with new ones */
203 unsigned page_replace:1;
204
205 /** Number or arguments */
206 unsigned numargs;
207
208 /** Array of arguments */
209 struct fuse_arg args[3];
210 };
211
212 /** The request state */
213 enum fuse_req_state {
214 FUSE_REQ_INIT = 0,
215 FUSE_REQ_PENDING,
216 FUSE_REQ_READING,
217 FUSE_REQ_SENT,
218 FUSE_REQ_WRITING,
219 FUSE_REQ_FINISHED
220 };
221
222 /**
223 * A request to the client
224 */
225 struct fuse_req {
226 /** This can be on either pending processing or io lists in
227 fuse_conn */
228 struct list_head list;
229
230 /** Entry on the interrupts list */
231 struct list_head intr_entry;
232
233 /** refcount */
234 atomic_t count;
235
236 /** Unique ID for the interrupt request */
237 u64 intr_unique;
238
239 /*
240 * The following bitfields are either set once before the
241 * request is queued or setting/clearing them is protected by
242 * fuse_conn->lock
243 */
244
245 /** True if the request has reply */
246 unsigned isreply:1;
247
248 /** Force sending of the request even if interrupted */
249 unsigned force:1;
250
251 /** The request was aborted */
252 unsigned aborted:1;
253
254 /** Request is sent in the background */
255 unsigned background:1;
256
257 /** The request has been interrupted */
258 unsigned interrupted:1;
259
260 /** Data is being copied to/from the request */
261 unsigned locked:1;
262
263 /** Request is counted as "waiting" */
264 unsigned waiting:1;
265
266 /** State of the request */
267 enum fuse_req_state state;
268
269 /** The request input */
270 struct fuse_in in;
271
272 /** The request output */
273 struct fuse_out out;
274
275 /** Used to wake up the task waiting for completion of request*/
276 wait_queue_head_t waitq;
277
278 /** Data for asynchronous requests */
279 union {
280 struct {
281 union {
282 struct fuse_release_in in;
283 struct work_struct work;
284 };
285 struct path path;
286 } release;
287 struct fuse_init_in init_in;
288 struct fuse_init_out init_out;
289 struct cuse_init_in cuse_init_in;
290 struct {
291 struct fuse_read_in in;
292 u64 attr_ver;
293 } read;
294 struct {
295 struct fuse_write_in in;
296 struct fuse_write_out out;
297 } write;
298 struct fuse_notify_retrieve_in retrieve_in;
299 struct fuse_lk_in lk_in;
300 } misc;
301
302 /** page vector */
303 struct page *pages[FUSE_MAX_PAGES_PER_REQ];
304
305 /** number of pages in vector */
306 unsigned num_pages;
307
308 /** offset of data on first page */
309 unsigned page_offset;
310
311 /** File used in the request (or NULL) */
312 struct fuse_file *ff;
313
314 /** Inode used in the request or NULL */
315 struct inode *inode;
316
317 /** Link on fi->writepages */
318 struct list_head writepages_entry;
319
320 /** Request completion callback */
321 void (*end)(struct fuse_conn *, struct fuse_req *);
322
323 /** Request is stolen from fuse_file->reserved_req */
324 struct file *stolen_file;
325 };
326
327 /**
328 * A Fuse connection.
329 *
330 * This structure is created, when the filesystem is mounted, and is
331 * destroyed, when the client device is closed and the filesystem is
332 * unmounted.
333 */
334 struct fuse_conn {
335 /** Lock protecting accessess to members of this structure */
336 spinlock_t lock;
337
338 /** Mutex protecting against directory alias creation */
339 struct mutex inst_mutex;
340
341 /** Refcount */
342 atomic_t count;
343
344 /** The user id for this mount */
345 uid_t user_id;
346
347 /** The group id for this mount */
348 gid_t group_id;
349
350 /** The fuse mount flags for this mount */
351 unsigned flags;
352
353 /** Maximum read size */
354 unsigned max_read;
355
356 /** Maximum write size */
357 unsigned max_write;
358
359 /** Readers of the connection are waiting on this */
360 wait_queue_head_t waitq;
361
362 /** The list of pending requests */
363 struct list_head pending;
364
365 /** The list of requests being processed */
366 struct list_head processing;
367
368 /** The list of requests under I/O */
369 struct list_head io;
370
371 /** The next unique kernel file handle */
372 u64 khctr;
373
374 /** rbtree of fuse_files waiting for poll events indexed by ph */
375 struct rb_root polled_files;
376
377 /** Maximum number of outstanding background requests */
378 unsigned max_background;
379
380 /** Number of background requests at which congestion starts */
381 unsigned congestion_threshold;
382
383 /** Number of requests currently in the background */
384 unsigned num_background;
385
386 /** Number of background requests currently queued for userspace */
387 unsigned active_background;
388
389 /** The list of background requests set aside for later queuing */
390 struct list_head bg_queue;
391
392 /** Pending interrupts */
393 struct list_head interrupts;
394
395 /** Queue of pending forgets */
396 struct fuse_forget_link forget_list_head;
397 struct fuse_forget_link *forget_list_tail;
398
399 /** Batching of FORGET requests (positive indicates FORGET batch) */
400 int forget_batch;
401
402 /** Flag indicating if connection is blocked. This will be
403 the case before the INIT reply is received, and if there
404 are too many outstading backgrounds requests */
405 int blocked;
406
407 /** waitq for blocked connection */
408 wait_queue_head_t blocked_waitq;
409
410 /** waitq for reserved requests */
411 wait_queue_head_t reserved_req_waitq;
412
413 /** The next unique request id */
414 u64 reqctr;
415
416 /** Connection established, cleared on umount, connection
417 abort and device release */
418 unsigned connected;
419
420 /** Connection failed (version mismatch). Cannot race with
421 setting other bitfields since it is only set once in INIT
422 reply, before any other request, and never cleared */
423 unsigned conn_error:1;
424
425 /** Connection successful. Only set in INIT */
426 unsigned conn_init:1;
427
428 /** Do readpages asynchronously? Only set in INIT */
429 unsigned async_read:1;
430
431 /** Do not send separate SETATTR request before open(O_TRUNC) */
432 unsigned atomic_o_trunc:1;
433
434 /** Filesystem supports NFS exporting. Only set in INIT */
435 unsigned export_support:1;
436
437 /** Set if bdi is valid */
438 unsigned bdi_initialized:1;
439
440 /*
441 * The following bitfields are only for optimization purposes
442 * and hence races in setting them will not cause malfunction
443 */
444
445 /** Is fsync not implemented by fs? */
446 unsigned no_fsync:1;
447
448 /** Is fsyncdir not implemented by fs? */
449 unsigned no_fsyncdir:1;
450
451 /** Is flush not implemented by fs? */
452 unsigned no_flush:1;
453
454 /** Is setxattr not implemented by fs? */
455 unsigned no_setxattr:1;
456
457 /** Is getxattr not implemented by fs? */
458 unsigned no_getxattr:1;
459
460 /** Is listxattr not implemented by fs? */
461 unsigned no_listxattr:1;
462
463 /** Is removexattr not implemented by fs? */
464 unsigned no_removexattr:1;
465
466 /** Are posix file locking primitives not implemented by fs? */
467 unsigned no_lock:1;
468
469 /** Is access not implemented by fs? */
470 unsigned no_access:1;
471
472 /** Is create not implemented by fs? */
473 unsigned no_create:1;
474
475 /** Is interrupt not implemented by fs? */
476 unsigned no_interrupt:1;
477
478 /** Is bmap not implemented by fs? */
479 unsigned no_bmap:1;
480
481 /** Is poll not implemented by fs? */
482 unsigned no_poll:1;
483
484 /** Do multi-page cached writes */
485 unsigned big_writes:1;
486
487 /** Don't apply umask to creation modes */
488 unsigned dont_mask:1;
489
490 /** Are BSD file locking primitives not implemented by fs? */
491 unsigned no_flock:1;
492
493 /** The number of requests waiting for completion */
494 atomic_t num_waiting;
495
496 /** Negotiated minor version */
497 unsigned minor;
498
499 /** Backing dev info */
500 struct backing_dev_info bdi;
501
502 /** Entry on the fuse_conn_list */
503 struct list_head entry;
504
505 /** Device ID from super block */
506 dev_t dev;
507
508 /** Dentries in the control filesystem */
509 struct dentry *ctl_dentry[FUSE_CTL_NUM_DENTRIES];
510
511 /** number of dentries used in the above array */
512 int ctl_ndents;
513
514 /** O_ASYNC requests */
515 struct fasync_struct *fasync;
516
517 /** Key for lock owner ID scrambling */
518 u32 scramble_key[4];
519
520 /** Reserved request for the DESTROY message */
521 struct fuse_req *destroy_req;
522
523 /** Version counter for attribute changes */
524 u64 attr_version;
525
526 /** Called on final put */
527 void (*release)(struct fuse_conn *);
528
529 /** Super block for this connection. */
530 struct super_block *sb;
531
532 /** Read/write semaphore to hold when accessing sb. */
533 struct rw_semaphore killsb;
534 };
535
get_fuse_conn_super(struct super_block * sb)536 static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb)
537 {
538 return sb->s_fs_info;
539 }
540
get_fuse_conn(struct inode * inode)541 static inline struct fuse_conn *get_fuse_conn(struct inode *inode)
542 {
543 return get_fuse_conn_super(inode->i_sb);
544 }
545
get_fuse_inode(struct inode * inode)546 static inline struct fuse_inode *get_fuse_inode(struct inode *inode)
547 {
548 return container_of(inode, struct fuse_inode, inode);
549 }
550
get_node_id(struct inode * inode)551 static inline u64 get_node_id(struct inode *inode)
552 {
553 return get_fuse_inode(inode)->nodeid;
554 }
555
556 /** Device operations */
557 extern const struct file_operations fuse_dev_operations;
558
559 extern const struct dentry_operations fuse_dentry_operations;
560
561 /**
562 * Inode to nodeid comparison.
563 */
564 int fuse_inode_eq(struct inode *inode, void *_nodeidp);
565
566 /**
567 * Get a filled in inode
568 */
569 struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
570 int generation, struct fuse_attr *attr,
571 u64 attr_valid, u64 attr_version);
572
573 int fuse_lookup_name(struct super_block *sb, u64 nodeid, struct qstr *name,
574 struct fuse_entry_out *outarg, struct inode **inode);
575
576 /**
577 * Send FORGET command
578 */
579 void fuse_queue_forget(struct fuse_conn *fc, struct fuse_forget_link *forget,
580 u64 nodeid, u64 nlookup);
581
582 struct fuse_forget_link *fuse_alloc_forget(void);
583
584 /**
585 * Initialize READ or READDIR request
586 */
587 void fuse_read_fill(struct fuse_req *req, struct file *file,
588 loff_t pos, size_t count, int opcode);
589
590 /**
591 * Send OPEN or OPENDIR request
592 */
593 int fuse_open_common(struct inode *inode, struct file *file, bool isdir);
594
595 struct fuse_file *fuse_file_alloc(struct fuse_conn *fc);
596 struct fuse_file *fuse_file_get(struct fuse_file *ff);
597 void fuse_file_free(struct fuse_file *ff);
598 void fuse_finish_open(struct inode *inode, struct file *file);
599
600 void fuse_sync_release(struct fuse_file *ff, int flags);
601
602 /**
603 * Send RELEASE or RELEASEDIR request
604 */
605 void fuse_release_common(struct file *file, int opcode);
606
607 /**
608 * Send FSYNC or FSYNCDIR request
609 */
610 int fuse_fsync_common(struct file *file, loff_t start, loff_t end,
611 int datasync, int isdir);
612
613 /**
614 * Notify poll wakeup
615 */
616 int fuse_notify_poll_wakeup(struct fuse_conn *fc,
617 struct fuse_notify_poll_wakeup_out *outarg);
618
619 /**
620 * Initialize file operations on a regular file
621 */
622 void fuse_init_file_inode(struct inode *inode);
623
624 /**
625 * Initialize inode operations on regular files and special files
626 */
627 void fuse_init_common(struct inode *inode);
628
629 /**
630 * Initialize inode and file operations on a directory
631 */
632 void fuse_init_dir(struct inode *inode);
633
634 /**
635 * Initialize inode operations on a symlink
636 */
637 void fuse_init_symlink(struct inode *inode);
638
639 /**
640 * Change attributes of an inode
641 */
642 void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
643 u64 attr_valid, u64 attr_version);
644
645 void fuse_change_attributes_common(struct inode *inode, struct fuse_attr *attr,
646 u64 attr_valid);
647
648 /**
649 * Initialize the client device
650 */
651 int fuse_dev_init(void);
652
653 /**
654 * Cleanup the client device
655 */
656 void fuse_dev_cleanup(void);
657
658 int fuse_ctl_init(void);
659 void fuse_ctl_cleanup(void);
660
661 /**
662 * Allocate a request
663 */
664 struct fuse_req *fuse_request_alloc(void);
665
666 struct fuse_req *fuse_request_alloc_nofs(void);
667
668 /**
669 * Free a request
670 */
671 void fuse_request_free(struct fuse_req *req);
672
673 /**
674 * Get a request, may fail with -ENOMEM
675 */
676 struct fuse_req *fuse_get_req(struct fuse_conn *fc);
677
678 /**
679 * Gets a requests for a file operation, always succeeds
680 */
681 struct fuse_req *fuse_get_req_nofail(struct fuse_conn *fc, struct file *file);
682
683 /**
684 * Decrement reference count of a request. If count goes to zero free
685 * the request.
686 */
687 void fuse_put_request(struct fuse_conn *fc, struct fuse_req *req);
688
689 /**
690 * Send a request (synchronous)
691 */
692 void fuse_request_send(struct fuse_conn *fc, struct fuse_req *req);
693
694 /**
695 * Send a request in the background
696 */
697 void fuse_request_send_background(struct fuse_conn *fc, struct fuse_req *req);
698
699 void fuse_request_send_background_locked(struct fuse_conn *fc,
700 struct fuse_req *req);
701
702 /* Abort all requests */
703 void fuse_abort_conn(struct fuse_conn *fc);
704
705 /**
706 * Invalidate inode attributes
707 */
708 void fuse_invalidate_attr(struct inode *inode);
709
710 void fuse_invalidate_entry_cache(struct dentry *entry);
711
712 /**
713 * Acquire reference to fuse_conn
714 */
715 struct fuse_conn *fuse_conn_get(struct fuse_conn *fc);
716
717 void fuse_conn_kill(struct fuse_conn *fc);
718
719 /**
720 * Initialize fuse_conn
721 */
722 void fuse_conn_init(struct fuse_conn *fc);
723
724 /**
725 * Release reference to fuse_conn
726 */
727 void fuse_conn_put(struct fuse_conn *fc);
728
729 /**
730 * Add connection to control filesystem
731 */
732 int fuse_ctl_add_conn(struct fuse_conn *fc);
733
734 /**
735 * Remove connection from control filesystem
736 */
737 void fuse_ctl_remove_conn(struct fuse_conn *fc);
738
739 /**
740 * Is file type valid?
741 */
742 int fuse_valid_type(int m);
743
744 /**
745 * Is task allowed to perform filesystem operation?
746 */
747 int fuse_allow_task(struct fuse_conn *fc, struct task_struct *task);
748
749 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id);
750
751 int fuse_update_attributes(struct inode *inode, struct kstat *stat,
752 struct file *file, bool *refreshed);
753
754 void fuse_flush_writepages(struct inode *inode);
755
756 void fuse_set_nowrite(struct inode *inode);
757 void fuse_release_nowrite(struct inode *inode);
758
759 u64 fuse_get_attr_version(struct fuse_conn *fc);
760
761 /**
762 * File-system tells the kernel to invalidate cache for the given node id.
763 */
764 int fuse_reverse_inval_inode(struct super_block *sb, u64 nodeid,
765 loff_t offset, loff_t len);
766
767 /**
768 * File-system tells the kernel to invalidate parent attributes and
769 * the dentry matching parent/name.
770 *
771 * If the child_nodeid is non-zero and:
772 * - matches the inode number for the dentry matching parent/name,
773 * - is not a mount point
774 * - is a file or oan empty directory
775 * then the dentry is unhashed (d_delete()).
776 */
777 int fuse_reverse_inval_entry(struct super_block *sb, u64 parent_nodeid,
778 u64 child_nodeid, struct qstr *name);
779
780 int fuse_do_open(struct fuse_conn *fc, u64 nodeid, struct file *file,
781 bool isdir);
782 ssize_t fuse_direct_io(struct file *file, const char __user *buf,
783 size_t count, loff_t *ppos, int write);
784 long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
785 unsigned int flags);
786 long fuse_ioctl_common(struct file *file, unsigned int cmd,
787 unsigned long arg, unsigned int flags);
788 unsigned fuse_file_poll(struct file *file, poll_table *wait);
789 int fuse_dev_release(struct inode *inode, struct file *file);
790
791 void fuse_write_update_size(struct inode *inode, loff_t pos);
792
793 #endif /* _FS_FUSE_I_H */
794