fuse: make the number of max background requests and congestion threshold tunable
[deliverable/linux.git] / fs / fuse / fuse_i.h
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
25 /** Max number of pages that can be used in a single read request */
26 #define FUSE_MAX_PAGES_PER_REQ 32
27
28 /** Bias for fi->writectr, meaning new writepages must not be sent */
29 #define FUSE_NOWRITE INT_MIN
30
31 /** It could be as large as PATH_MAX, but would that have any uses? */
32 #define FUSE_NAME_MAX 1024
33
34 /** Number of dentries for each connection in the control filesystem */
35 #define FUSE_CTL_NUM_DENTRIES 3
36
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)
41
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)
45
46 /** List of active connections */
47 extern struct list_head fuse_conn_list;
48
49 /** Global mutex protecting fuse_conn_list and the control filesystem */
50 extern struct mutex fuse_mutex;
51
52 /** FUSE inode */
53 struct fuse_inode {
54 /** Inode data */
55 struct inode inode;
56
57 /** Unique ID, which identifies the inode between userspace
58 * and kernel */
59 u64 nodeid;
60
61 /** Number of lookups on this inode */
62 u64 nlookup;
63
64 /** The request used for sending the FORGET message */
65 struct fuse_req *forget_req;
66
67 /** Time in jiffies until the file attributes are valid */
68 u64 i_time;
69
70 /** The sticky bit in inode->i_mode may have been removed, so
71 preserve the original mode */
72 mode_t orig_i_mode;
73
74 /** Version of last attribute change */
75 u64 attr_version;
76
77 /** Files usable in writepage. Protected by fc->lock */
78 struct list_head write_files;
79
80 /** Writepages pending on truncate or fsync */
81 struct list_head queued_writes;
82
83 /** Number of sent writes, a negative bias (FUSE_NOWRITE)
84 * means more writes are blocked */
85 int writectr;
86
87 /** Waitq for writepage completion */
88 wait_queue_head_t page_waitq;
89
90 /** List of writepage requestst (pending or sent) */
91 struct list_head writepages;
92 };
93
94 struct fuse_conn;
95
96 /** FUSE specific file data */
97 struct fuse_file {
98 /** Fuse connection for this file */
99 struct fuse_conn *fc;
100
101 /** Request reserved for flush and release */
102 struct fuse_req *reserved_req;
103
104 /** Kernel file handle guaranteed to be unique */
105 u64 kh;
106
107 /** File handle used by userspace */
108 u64 fh;
109
110 /** Node id of this file */
111 u64 nodeid;
112
113 /** Refcount */
114 atomic_t count;
115
116 /** FOPEN_* flags returned by open */
117 u32 open_flags;
118
119 /** Entry on inode's write_files list */
120 struct list_head write_entry;
121
122 /** RB node to be linked on fuse_conn->polled_files */
123 struct rb_node polled_node;
124
125 /** Wait queue head for poll */
126 wait_queue_head_t poll_wait;
127 };
128
129 /** One input argument of a request */
130 struct fuse_in_arg {
131 unsigned size;
132 const void *value;
133 };
134
135 /** The request input */
136 struct fuse_in {
137 /** The request header */
138 struct fuse_in_header h;
139
140 /** True if the data for the last argument is in req->pages */
141 unsigned argpages:1;
142
143 /** Number of arguments */
144 unsigned numargs;
145
146 /** Array of arguments */
147 struct fuse_in_arg args[3];
148 };
149
150 /** One output argument of a request */
151 struct fuse_arg {
152 unsigned size;
153 void *value;
154 };
155
156 /** The request output */
157 struct fuse_out {
158 /** Header returned from userspace */
159 struct fuse_out_header h;
160
161 /*
162 * The following bitfields are not changed during the request
163 * processing
164 */
165
166 /** Last argument is variable length (can be shorter than
167 arg->size) */
168 unsigned argvar:1;
169
170 /** Last argument is a list of pages to copy data to */
171 unsigned argpages:1;
172
173 /** Zero partially or not copied pages */
174 unsigned page_zeroing:1;
175
176 /** Number or arguments */
177 unsigned numargs;
178
179 /** Array of arguments */
180 struct fuse_arg args[3];
181 };
182
183 /** The request state */
184 enum fuse_req_state {
185 FUSE_REQ_INIT = 0,
186 FUSE_REQ_PENDING,
187 FUSE_REQ_READING,
188 FUSE_REQ_SENT,
189 FUSE_REQ_WRITING,
190 FUSE_REQ_FINISHED
191 };
192
193 /**
194 * A request to the client
195 */
196 struct fuse_req {
197 /** This can be on either pending processing or io lists in
198 fuse_conn */
199 struct list_head list;
200
201 /** Entry on the interrupts list */
202 struct list_head intr_entry;
203
204 /** refcount */
205 atomic_t count;
206
207 /** Unique ID for the interrupt request */
208 u64 intr_unique;
209
210 /*
211 * The following bitfields are either set once before the
212 * request is queued or setting/clearing them is protected by
213 * fuse_conn->lock
214 */
215
216 /** True if the request has reply */
217 unsigned isreply:1;
218
219 /** Force sending of the request even if interrupted */
220 unsigned force:1;
221
222 /** The request was aborted */
223 unsigned aborted:1;
224
225 /** Request is sent in the background */
226 unsigned background:1;
227
228 /** The request has been interrupted */
229 unsigned interrupted:1;
230
231 /** Data is being copied to/from the request */
232 unsigned locked:1;
233
234 /** Request is counted as "waiting" */
235 unsigned waiting:1;
236
237 /** State of the request */
238 enum fuse_req_state state;
239
240 /** The request input */
241 struct fuse_in in;
242
243 /** The request output */
244 struct fuse_out out;
245
246 /** Used to wake up the task waiting for completion of request*/
247 wait_queue_head_t waitq;
248
249 /** Data for asynchronous requests */
250 union {
251 struct fuse_forget_in forget_in;
252 struct {
253 struct fuse_release_in in;
254 struct path path;
255 } release;
256 struct fuse_init_in init_in;
257 struct fuse_init_out init_out;
258 struct cuse_init_in cuse_init_in;
259 struct cuse_init_out cuse_init_out;
260 struct {
261 struct fuse_read_in in;
262 u64 attr_ver;
263 } read;
264 struct {
265 struct fuse_write_in in;
266 struct fuse_write_out out;
267 } write;
268 struct fuse_lk_in lk_in;
269 } misc;
270
271 /** page vector */
272 struct page *pages[FUSE_MAX_PAGES_PER_REQ];
273
274 /** number of pages in vector */
275 unsigned num_pages;
276
277 /** offset of data on first page */
278 unsigned page_offset;
279
280 /** File used in the request (or NULL) */
281 struct fuse_file *ff;
282
283 /** Inode used in the request or NULL */
284 struct inode *inode;
285
286 /** Link on fi->writepages */
287 struct list_head writepages_entry;
288
289 /** Request completion callback */
290 void (*end)(struct fuse_conn *, struct fuse_req *);
291
292 /** Request is stolen from fuse_file->reserved_req */
293 struct file *stolen_file;
294 };
295
296 /**
297 * A Fuse connection.
298 *
299 * This structure is created, when the filesystem is mounted, and is
300 * destroyed, when the client device is closed and the filesystem is
301 * unmounted.
302 */
303 struct fuse_conn {
304 /** Lock protecting accessess to members of this structure */
305 spinlock_t lock;
306
307 /** Mutex protecting against directory alias creation */
308 struct mutex inst_mutex;
309
310 /** Refcount */
311 atomic_t count;
312
313 /** The user id for this mount */
314 uid_t user_id;
315
316 /** The group id for this mount */
317 gid_t group_id;
318
319 /** The fuse mount flags for this mount */
320 unsigned flags;
321
322 /** Maximum read size */
323 unsigned max_read;
324
325 /** Maximum write size */
326 unsigned max_write;
327
328 /** Readers of the connection are waiting on this */
329 wait_queue_head_t waitq;
330
331 /** The list of pending requests */
332 struct list_head pending;
333
334 /** The list of requests being processed */
335 struct list_head processing;
336
337 /** The list of requests under I/O */
338 struct list_head io;
339
340 /** The next unique kernel file handle */
341 u64 khctr;
342
343 /** rbtree of fuse_files waiting for poll events indexed by ph */
344 struct rb_root polled_files;
345
346 /** Maximum number of outstanding background requests */
347 unsigned max_background;
348
349 /** Number of background requests at which congestion starts */
350 unsigned congestion_threshold;
351
352 /** Number of requests currently in the background */
353 unsigned num_background;
354
355 /** Number of background requests currently queued for userspace */
356 unsigned active_background;
357
358 /** The list of background requests set aside for later queuing */
359 struct list_head bg_queue;
360
361 /** Pending interrupts */
362 struct list_head interrupts;
363
364 /** Flag indicating if connection is blocked. This will be
365 the case before the INIT reply is received, and if there
366 are too many outstading backgrounds requests */
367 int blocked;
368
369 /** waitq for blocked connection */
370 wait_queue_head_t blocked_waitq;
371
372 /** waitq for reserved requests */
373 wait_queue_head_t reserved_req_waitq;
374
375 /** The next unique request id */
376 u64 reqctr;
377
378 /** Connection established, cleared on umount, connection
379 abort and device release */
380 unsigned connected;
381
382 /** Connection failed (version mismatch). Cannot race with
383 setting other bitfields since it is only set once in INIT
384 reply, before any other request, and never cleared */
385 unsigned conn_error:1;
386
387 /** Connection successful. Only set in INIT */
388 unsigned conn_init:1;
389
390 /** Do readpages asynchronously? Only set in INIT */
391 unsigned async_read:1;
392
393 /** Do not send separate SETATTR request before open(O_TRUNC) */
394 unsigned atomic_o_trunc:1;
395
396 /** Filesystem supports NFS exporting. Only set in INIT */
397 unsigned export_support:1;
398
399 /** Set if bdi is valid */
400 unsigned bdi_initialized:1;
401
402 /*
403 * The following bitfields are only for optimization purposes
404 * and hence races in setting them will not cause malfunction
405 */
406
407 /** Is fsync not implemented by fs? */
408 unsigned no_fsync:1;
409
410 /** Is fsyncdir not implemented by fs? */
411 unsigned no_fsyncdir:1;
412
413 /** Is flush not implemented by fs? */
414 unsigned no_flush:1;
415
416 /** Is setxattr not implemented by fs? */
417 unsigned no_setxattr:1;
418
419 /** Is getxattr not implemented by fs? */
420 unsigned no_getxattr:1;
421
422 /** Is listxattr not implemented by fs? */
423 unsigned no_listxattr:1;
424
425 /** Is removexattr not implemented by fs? */
426 unsigned no_removexattr:1;
427
428 /** Are file locking primitives not implemented by fs? */
429 unsigned no_lock:1;
430
431 /** Is access not implemented by fs? */
432 unsigned no_access:1;
433
434 /** Is create not implemented by fs? */
435 unsigned no_create:1;
436
437 /** Is interrupt not implemented by fs? */
438 unsigned no_interrupt:1;
439
440 /** Is bmap not implemented by fs? */
441 unsigned no_bmap:1;
442
443 /** Is poll not implemented by fs? */
444 unsigned no_poll:1;
445
446 /** Do multi-page cached writes */
447 unsigned big_writes:1;
448
449 /** Don't apply umask to creation modes */
450 unsigned dont_mask:1;
451
452 /** The number of requests waiting for completion */
453 atomic_t num_waiting;
454
455 /** Negotiated minor version */
456 unsigned minor;
457
458 /** Backing dev info */
459 struct backing_dev_info bdi;
460
461 /** Entry on the fuse_conn_list */
462 struct list_head entry;
463
464 /** Device ID from super block */
465 dev_t dev;
466
467 /** Dentries in the control filesystem */
468 struct dentry *ctl_dentry[FUSE_CTL_NUM_DENTRIES];
469
470 /** number of dentries used in the above array */
471 int ctl_ndents;
472
473 /** O_ASYNC requests */
474 struct fasync_struct *fasync;
475
476 /** Key for lock owner ID scrambling */
477 u32 scramble_key[4];
478
479 /** Reserved request for the DESTROY message */
480 struct fuse_req *destroy_req;
481
482 /** Version counter for attribute changes */
483 u64 attr_version;
484
485 /** Called on final put */
486 void (*release)(struct fuse_conn *);
487
488 /** Super block for this connection. */
489 struct super_block *sb;
490
491 /** Read/write semaphore to hold when accessing sb. */
492 struct rw_semaphore killsb;
493 };
494
495 static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb)
496 {
497 return sb->s_fs_info;
498 }
499
500 static inline struct fuse_conn *get_fuse_conn(struct inode *inode)
501 {
502 return get_fuse_conn_super(inode->i_sb);
503 }
504
505 static inline struct fuse_inode *get_fuse_inode(struct inode *inode)
506 {
507 return container_of(inode, struct fuse_inode, inode);
508 }
509
510 static inline u64 get_node_id(struct inode *inode)
511 {
512 return get_fuse_inode(inode)->nodeid;
513 }
514
515 /** Device operations */
516 extern const struct file_operations fuse_dev_operations;
517
518 extern const struct dentry_operations fuse_dentry_operations;
519
520 /**
521 * Inode to nodeid comparison.
522 */
523 int fuse_inode_eq(struct inode *inode, void *_nodeidp);
524
525 /**
526 * Get a filled in inode
527 */
528 struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
529 int generation, struct fuse_attr *attr,
530 u64 attr_valid, u64 attr_version);
531
532 int fuse_lookup_name(struct super_block *sb, u64 nodeid, struct qstr *name,
533 struct fuse_entry_out *outarg, struct inode **inode);
534
535 /**
536 * Send FORGET command
537 */
538 void fuse_send_forget(struct fuse_conn *fc, struct fuse_req *req,
539 u64 nodeid, u64 nlookup);
540
541 /**
542 * Initialize READ or READDIR request
543 */
544 void fuse_read_fill(struct fuse_req *req, struct file *file,
545 loff_t pos, size_t count, int opcode);
546
547 /**
548 * Send OPEN or OPENDIR request
549 */
550 int fuse_open_common(struct inode *inode, struct file *file, bool isdir);
551
552 struct fuse_file *fuse_file_alloc(struct fuse_conn *fc);
553 struct fuse_file *fuse_file_get(struct fuse_file *ff);
554 void fuse_file_free(struct fuse_file *ff);
555 void fuse_finish_open(struct inode *inode, struct file *file);
556
557 void fuse_sync_release(struct fuse_file *ff, int flags);
558
559 /**
560 * Send RELEASE or RELEASEDIR request
561 */
562 void fuse_release_common(struct file *file, int opcode);
563
564 /**
565 * Send FSYNC or FSYNCDIR request
566 */
567 int fuse_fsync_common(struct file *file, struct dentry *de, int datasync,
568 int isdir);
569
570 /**
571 * Notify poll wakeup
572 */
573 int fuse_notify_poll_wakeup(struct fuse_conn *fc,
574 struct fuse_notify_poll_wakeup_out *outarg);
575
576 /**
577 * Initialize file operations on a regular file
578 */
579 void fuse_init_file_inode(struct inode *inode);
580
581 /**
582 * Initialize inode operations on regular files and special files
583 */
584 void fuse_init_common(struct inode *inode);
585
586 /**
587 * Initialize inode and file operations on a directory
588 */
589 void fuse_init_dir(struct inode *inode);
590
591 /**
592 * Initialize inode operations on a symlink
593 */
594 void fuse_init_symlink(struct inode *inode);
595
596 /**
597 * Change attributes of an inode
598 */
599 void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
600 u64 attr_valid, u64 attr_version);
601
602 void fuse_change_attributes_common(struct inode *inode, struct fuse_attr *attr,
603 u64 attr_valid);
604
605 void fuse_truncate(struct address_space *mapping, loff_t offset);
606
607 /**
608 * Initialize the client device
609 */
610 int fuse_dev_init(void);
611
612 /**
613 * Cleanup the client device
614 */
615 void fuse_dev_cleanup(void);
616
617 int fuse_ctl_init(void);
618 void fuse_ctl_cleanup(void);
619
620 /**
621 * Allocate a request
622 */
623 struct fuse_req *fuse_request_alloc(void);
624
625 struct fuse_req *fuse_request_alloc_nofs(void);
626
627 /**
628 * Free a request
629 */
630 void fuse_request_free(struct fuse_req *req);
631
632 /**
633 * Get a request, may fail with -ENOMEM
634 */
635 struct fuse_req *fuse_get_req(struct fuse_conn *fc);
636
637 /**
638 * Gets a requests for a file operation, always succeeds
639 */
640 struct fuse_req *fuse_get_req_nofail(struct fuse_conn *fc, struct file *file);
641
642 /**
643 * Decrement reference count of a request. If count goes to zero free
644 * the request.
645 */
646 void fuse_put_request(struct fuse_conn *fc, struct fuse_req *req);
647
648 /**
649 * Send a request (synchronous)
650 */
651 void fuse_request_send(struct fuse_conn *fc, struct fuse_req *req);
652
653 /**
654 * Send a request with no reply
655 */
656 void fuse_request_send_noreply(struct fuse_conn *fc, struct fuse_req *req);
657
658 /**
659 * Send a request in the background
660 */
661 void fuse_request_send_background(struct fuse_conn *fc, struct fuse_req *req);
662
663 void fuse_request_send_background_locked(struct fuse_conn *fc,
664 struct fuse_req *req);
665
666 /* Abort all requests */
667 void fuse_abort_conn(struct fuse_conn *fc);
668
669 /**
670 * Invalidate inode attributes
671 */
672 void fuse_invalidate_attr(struct inode *inode);
673
674 void fuse_invalidate_entry_cache(struct dentry *entry);
675
676 /**
677 * Acquire reference to fuse_conn
678 */
679 struct fuse_conn *fuse_conn_get(struct fuse_conn *fc);
680
681 void fuse_conn_kill(struct fuse_conn *fc);
682
683 /**
684 * Initialize fuse_conn
685 */
686 void fuse_conn_init(struct fuse_conn *fc);
687
688 /**
689 * Release reference to fuse_conn
690 */
691 void fuse_conn_put(struct fuse_conn *fc);
692
693 /**
694 * Add connection to control filesystem
695 */
696 int fuse_ctl_add_conn(struct fuse_conn *fc);
697
698 /**
699 * Remove connection from control filesystem
700 */
701 void fuse_ctl_remove_conn(struct fuse_conn *fc);
702
703 /**
704 * Is file type valid?
705 */
706 int fuse_valid_type(int m);
707
708 /**
709 * Is task allowed to perform filesystem operation?
710 */
711 int fuse_allow_task(struct fuse_conn *fc, struct task_struct *task);
712
713 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id);
714
715 int fuse_update_attributes(struct inode *inode, struct kstat *stat,
716 struct file *file, bool *refreshed);
717
718 void fuse_flush_writepages(struct inode *inode);
719
720 void fuse_set_nowrite(struct inode *inode);
721 void fuse_release_nowrite(struct inode *inode);
722
723 u64 fuse_get_attr_version(struct fuse_conn *fc);
724
725 /**
726 * File-system tells the kernel to invalidate cache for the given node id.
727 */
728 int fuse_reverse_inval_inode(struct super_block *sb, u64 nodeid,
729 loff_t offset, loff_t len);
730
731 /**
732 * File-system tells the kernel to invalidate parent attributes and
733 * the dentry matching parent/name.
734 */
735 int fuse_reverse_inval_entry(struct super_block *sb, u64 parent_nodeid,
736 struct qstr *name);
737
738 int fuse_do_open(struct fuse_conn *fc, u64 nodeid, struct file *file,
739 bool isdir);
740 ssize_t fuse_direct_io(struct file *file, const char __user *buf,
741 size_t count, loff_t *ppos, int write);
742 long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
743 unsigned int flags);
744 unsigned fuse_file_poll(struct file *file, poll_table *wait);
745 int fuse_dev_release(struct inode *inode, struct file *file);
746
747 #endif /* _FS_FUSE_I_H */
This page took 0.076641 seconds and 5 git commands to generate.