xref: /openbmc/linux/fs/fuse/fuse_i.h (revision e368cd72)
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 #ifndef pr_fmt
13 # define pr_fmt(fmt) "fuse: " fmt
14 #endif
15 
16 #include <linux/fuse.h>
17 #include <linux/fs.h>
18 #include <linux/mount.h>
19 #include <linux/wait.h>
20 #include <linux/list.h>
21 #include <linux/spinlock.h>
22 #include <linux/mm.h>
23 #include <linux/backing-dev.h>
24 #include <linux/mutex.h>
25 #include <linux/rwsem.h>
26 #include <linux/rbtree.h>
27 #include <linux/poll.h>
28 #include <linux/workqueue.h>
29 #include <linux/kref.h>
30 #include <linux/xattr.h>
31 #include <linux/pid_namespace.h>
32 #include <linux/refcount.h>
33 #include <linux/user_namespace.h>
34 
35 /** Default max number of pages that can be used in a single read request */
36 #define FUSE_DEFAULT_MAX_PAGES_PER_REQ 32
37 
38 /** Maximum of max_pages received in init_out */
39 #define FUSE_MAX_MAX_PAGES 256
40 
41 /** Bias for fi->writectr, meaning new writepages must not be sent */
42 #define FUSE_NOWRITE INT_MIN
43 
44 /** It could be as large as PATH_MAX, but would that have any uses? */
45 #define FUSE_NAME_MAX 1024
46 
47 /** Number of dentries for each connection in the control filesystem */
48 #define FUSE_CTL_NUM_DENTRIES 5
49 
50 /** List of active connections */
51 extern struct list_head fuse_conn_list;
52 
53 /** Global mutex protecting fuse_conn_list and the control filesystem */
54 extern struct mutex fuse_mutex;
55 
56 /** Module parameters */
57 extern unsigned max_user_bgreq;
58 extern unsigned max_user_congthresh;
59 
60 /* One forget request */
61 struct fuse_forget_link {
62 	struct fuse_forget_one forget_one;
63 	struct fuse_forget_link *next;
64 };
65 
66 /** FUSE inode */
67 struct fuse_inode {
68 	/** Inode data */
69 	struct inode inode;
70 
71 	/** Unique ID, which identifies the inode between userspace
72 	 * and kernel */
73 	u64 nodeid;
74 
75 	/** Number of lookups on this inode */
76 	u64 nlookup;
77 
78 	/** The request used for sending the FORGET message */
79 	struct fuse_forget_link *forget;
80 
81 	/** Time in jiffies until the file attributes are valid */
82 	u64 i_time;
83 
84 	/* Which attributes are invalid */
85 	u32 inval_mask;
86 
87 	/** The sticky bit in inode->i_mode may have been removed, so
88 	    preserve the original mode */
89 	umode_t orig_i_mode;
90 
91 	/** 64 bit inode number */
92 	u64 orig_ino;
93 
94 	/** Version of last attribute change */
95 	u64 attr_version;
96 
97 	union {
98 		/* Write related fields (regular file only) */
99 		struct {
100 			/* Files usable in writepage.  Protected by fi->lock */
101 			struct list_head write_files;
102 
103 			/* Writepages pending on truncate or fsync */
104 			struct list_head queued_writes;
105 
106 			/* Number of sent writes, a negative bias
107 			 * (FUSE_NOWRITE) means more writes are blocked */
108 			int writectr;
109 
110 			/* Waitq for writepage completion */
111 			wait_queue_head_t page_waitq;
112 
113 			/* List of writepage requestst (pending or sent) */
114 			struct rb_root writepages;
115 		};
116 
117 		/* readdir cache (directory only) */
118 		struct {
119 			/* true if fully cached */
120 			bool cached;
121 
122 			/* size of cache */
123 			loff_t size;
124 
125 			/* position at end of cache (position of next entry) */
126 			loff_t pos;
127 
128 			/* version of the cache */
129 			u64 version;
130 
131 			/* modification time of directory when cache was
132 			 * started */
133 			struct timespec64 mtime;
134 
135 			/* iversion of directory when cache was started */
136 			u64 iversion;
137 
138 			/* protects above fields */
139 			spinlock_t lock;
140 		} rdc;
141 	};
142 
143 	/** Miscellaneous bits describing inode state */
144 	unsigned long state;
145 
146 	/** Lock for serializing lookup and readdir for back compatibility*/
147 	struct mutex mutex;
148 
149 	/** Lock to protect write related fields */
150 	spinlock_t lock;
151 
152 #ifdef CONFIG_FUSE_DAX
153 	/*
154 	 * Dax specific inode data
155 	 */
156 	struct fuse_inode_dax *dax;
157 #endif
158 };
159 
160 /** FUSE inode state bits */
161 enum {
162 	/** Advise readdirplus  */
163 	FUSE_I_ADVISE_RDPLUS,
164 	/** Initialized with readdirplus */
165 	FUSE_I_INIT_RDPLUS,
166 	/** An operation changing file size is in progress  */
167 	FUSE_I_SIZE_UNSTABLE,
168 	/* Bad inode */
169 	FUSE_I_BAD,
170 };
171 
172 struct fuse_conn;
173 struct fuse_mount;
174 struct fuse_release_args;
175 
176 /** FUSE specific file data */
177 struct fuse_file {
178 	/** Fuse connection for this file */
179 	struct fuse_mount *fm;
180 
181 	/* Argument space reserved for release */
182 	struct fuse_release_args *release_args;
183 
184 	/** Kernel file handle guaranteed to be unique */
185 	u64 kh;
186 
187 	/** File handle used by userspace */
188 	u64 fh;
189 
190 	/** Node id of this file */
191 	u64 nodeid;
192 
193 	/** Refcount */
194 	refcount_t count;
195 
196 	/** FOPEN_* flags returned by open */
197 	u32 open_flags;
198 
199 	/** Entry on inode's write_files list */
200 	struct list_head write_entry;
201 
202 	/* Readdir related */
203 	struct {
204 		/*
205 		 * Protects below fields against (crazy) parallel readdir on
206 		 * same open file.  Uncontended in the normal case.
207 		 */
208 		struct mutex lock;
209 
210 		/* Dir stream position */
211 		loff_t pos;
212 
213 		/* Offset in cache */
214 		loff_t cache_off;
215 
216 		/* Version of cache we are reading */
217 		u64 version;
218 
219 	} readdir;
220 
221 	/** RB node to be linked on fuse_conn->polled_files */
222 	struct rb_node polled_node;
223 
224 	/** Wait queue head for poll */
225 	wait_queue_head_t poll_wait;
226 
227 	/** Has flock been performed on this file? */
228 	bool flock:1;
229 };
230 
231 /** One input argument of a request */
232 struct fuse_in_arg {
233 	unsigned size;
234 	const void *value;
235 };
236 
237 /** One output argument of a request */
238 struct fuse_arg {
239 	unsigned size;
240 	void *value;
241 };
242 
243 /** FUSE page descriptor */
244 struct fuse_page_desc {
245 	unsigned int length;
246 	unsigned int offset;
247 };
248 
249 struct fuse_args {
250 	uint64_t nodeid;
251 	uint32_t opcode;
252 	unsigned short in_numargs;
253 	unsigned short out_numargs;
254 	bool force:1;
255 	bool noreply:1;
256 	bool nocreds:1;
257 	bool in_pages:1;
258 	bool out_pages:1;
259 	bool out_argvar:1;
260 	bool page_zeroing:1;
261 	bool page_replace:1;
262 	bool may_block:1;
263 	struct fuse_in_arg in_args[3];
264 	struct fuse_arg out_args[2];
265 	void (*end)(struct fuse_mount *fm, struct fuse_args *args, int error);
266 };
267 
268 struct fuse_args_pages {
269 	struct fuse_args args;
270 	struct page **pages;
271 	struct fuse_page_desc *descs;
272 	unsigned int num_pages;
273 };
274 
275 #define FUSE_ARGS(args) struct fuse_args args = {}
276 
277 /** The request IO state (for asynchronous processing) */
278 struct fuse_io_priv {
279 	struct kref refcnt;
280 	int async;
281 	spinlock_t lock;
282 	unsigned reqs;
283 	ssize_t bytes;
284 	size_t size;
285 	__u64 offset;
286 	bool write;
287 	bool should_dirty;
288 	int err;
289 	struct kiocb *iocb;
290 	struct completion *done;
291 	bool blocking;
292 };
293 
294 #define FUSE_IO_PRIV_SYNC(i) \
295 {					\
296 	.refcnt = KREF_INIT(1),		\
297 	.async = 0,			\
298 	.iocb = i,			\
299 }
300 
301 /**
302  * Request flags
303  *
304  * FR_ISREPLY:		set if the request has reply
305  * FR_FORCE:		force sending of the request even if interrupted
306  * FR_BACKGROUND:	request is sent in the background
307  * FR_WAITING:		request is counted as "waiting"
308  * FR_ABORTED:		the request was aborted
309  * FR_INTERRUPTED:	the request has been interrupted
310  * FR_LOCKED:		data is being copied to/from the request
311  * FR_PENDING:		request is not yet in userspace
312  * FR_SENT:		request is in userspace, waiting for an answer
313  * FR_FINISHED:		request is finished
314  * FR_PRIVATE:		request is on private list
315  * FR_ASYNC:		request is asynchronous
316  */
317 enum fuse_req_flag {
318 	FR_ISREPLY,
319 	FR_FORCE,
320 	FR_BACKGROUND,
321 	FR_WAITING,
322 	FR_ABORTED,
323 	FR_INTERRUPTED,
324 	FR_LOCKED,
325 	FR_PENDING,
326 	FR_SENT,
327 	FR_FINISHED,
328 	FR_PRIVATE,
329 	FR_ASYNC,
330 };
331 
332 /**
333  * A request to the client
334  *
335  * .waitq.lock protects the following fields:
336  *   - FR_ABORTED
337  *   - FR_LOCKED (may also be modified under fc->lock, tested under both)
338  */
339 struct fuse_req {
340 	/** This can be on either pending processing or io lists in
341 	    fuse_conn */
342 	struct list_head list;
343 
344 	/** Entry on the interrupts list  */
345 	struct list_head intr_entry;
346 
347 	/* Input/output arguments */
348 	struct fuse_args *args;
349 
350 	/** refcount */
351 	refcount_t count;
352 
353 	/* Request flags, updated with test/set/clear_bit() */
354 	unsigned long flags;
355 
356 	/* The request input header */
357 	struct {
358 		struct fuse_in_header h;
359 	} in;
360 
361 	/* The request output header */
362 	struct {
363 		struct fuse_out_header h;
364 	} out;
365 
366 	/** Used to wake up the task waiting for completion of request*/
367 	wait_queue_head_t waitq;
368 
369 #if IS_ENABLED(CONFIG_VIRTIO_FS)
370 	/** virtio-fs's physically contiguous buffer for in and out args */
371 	void *argbuf;
372 #endif
373 
374 	/** fuse_mount this request belongs to */
375 	struct fuse_mount *fm;
376 };
377 
378 struct fuse_iqueue;
379 
380 /**
381  * Input queue callbacks
382  *
383  * Input queue signalling is device-specific.  For example, the /dev/fuse file
384  * uses fiq->waitq and fasync to wake processes that are waiting on queue
385  * readiness.  These callbacks allow other device types to respond to input
386  * queue activity.
387  */
388 struct fuse_iqueue_ops {
389 	/**
390 	 * Signal that a forget has been queued
391 	 */
392 	void (*wake_forget_and_unlock)(struct fuse_iqueue *fiq)
393 		__releases(fiq->lock);
394 
395 	/**
396 	 * Signal that an INTERRUPT request has been queued
397 	 */
398 	void (*wake_interrupt_and_unlock)(struct fuse_iqueue *fiq)
399 		__releases(fiq->lock);
400 
401 	/**
402 	 * Signal that a request has been queued
403 	 */
404 	void (*wake_pending_and_unlock)(struct fuse_iqueue *fiq)
405 		__releases(fiq->lock);
406 
407 	/**
408 	 * Clean up when fuse_iqueue is destroyed
409 	 */
410 	void (*release)(struct fuse_iqueue *fiq);
411 };
412 
413 /** /dev/fuse input queue operations */
414 extern const struct fuse_iqueue_ops fuse_dev_fiq_ops;
415 
416 struct fuse_iqueue {
417 	/** Connection established */
418 	unsigned connected;
419 
420 	/** Lock protecting accesses to members of this structure */
421 	spinlock_t lock;
422 
423 	/** Readers of the connection are waiting on this */
424 	wait_queue_head_t waitq;
425 
426 	/** The next unique request id */
427 	u64 reqctr;
428 
429 	/** The list of pending requests */
430 	struct list_head pending;
431 
432 	/** Pending interrupts */
433 	struct list_head interrupts;
434 
435 	/** Queue of pending forgets */
436 	struct fuse_forget_link forget_list_head;
437 	struct fuse_forget_link *forget_list_tail;
438 
439 	/** Batching of FORGET requests (positive indicates FORGET batch) */
440 	int forget_batch;
441 
442 	/** O_ASYNC requests */
443 	struct fasync_struct *fasync;
444 
445 	/** Device-specific callbacks */
446 	const struct fuse_iqueue_ops *ops;
447 
448 	/** Device-specific state */
449 	void *priv;
450 };
451 
452 #define FUSE_PQ_HASH_BITS 8
453 #define FUSE_PQ_HASH_SIZE (1 << FUSE_PQ_HASH_BITS)
454 
455 struct fuse_pqueue {
456 	/** Connection established */
457 	unsigned connected;
458 
459 	/** Lock protecting accessess to  members of this structure */
460 	spinlock_t lock;
461 
462 	/** Hash table of requests being processed */
463 	struct list_head *processing;
464 
465 	/** The list of requests under I/O */
466 	struct list_head io;
467 };
468 
469 /**
470  * Fuse device instance
471  */
472 struct fuse_dev {
473 	/** Fuse connection for this device */
474 	struct fuse_conn *fc;
475 
476 	/** Processing queue */
477 	struct fuse_pqueue pq;
478 
479 	/** list entry on fc->devices */
480 	struct list_head entry;
481 };
482 
483 struct fuse_fs_context {
484 	int fd;
485 	unsigned int rootmode;
486 	kuid_t user_id;
487 	kgid_t group_id;
488 	bool is_bdev:1;
489 	bool fd_present:1;
490 	bool rootmode_present:1;
491 	bool user_id_present:1;
492 	bool group_id_present:1;
493 	bool default_permissions:1;
494 	bool allow_other:1;
495 	bool destroy:1;
496 	bool no_control:1;
497 	bool no_force_umount:1;
498 	bool legacy_opts_show:1;
499 	bool dax:1;
500 	unsigned int max_read;
501 	unsigned int blksize;
502 	const char *subtype;
503 
504 	/* DAX device, may be NULL */
505 	struct dax_device *dax_dev;
506 
507 	/* fuse_dev pointer to fill in, should contain NULL on entry */
508 	void **fudptr;
509 };
510 
511 /**
512  * A Fuse connection.
513  *
514  * This structure is created, when the root filesystem is mounted, and
515  * is destroyed, when the client device is closed and the last
516  * fuse_mount is destroyed.
517  */
518 struct fuse_conn {
519 	/** Lock protecting accessess to  members of this structure */
520 	spinlock_t lock;
521 
522 	/** Refcount */
523 	refcount_t count;
524 
525 	/** Number of fuse_dev's */
526 	atomic_t dev_count;
527 
528 	struct rcu_head rcu;
529 
530 	/** The user id for this mount */
531 	kuid_t user_id;
532 
533 	/** The group id for this mount */
534 	kgid_t group_id;
535 
536 	/** The pid namespace for this mount */
537 	struct pid_namespace *pid_ns;
538 
539 	/** The user namespace for this mount */
540 	struct user_namespace *user_ns;
541 
542 	/** Maximum read size */
543 	unsigned max_read;
544 
545 	/** Maximum write size */
546 	unsigned max_write;
547 
548 	/** Maximum number of pages that can be used in a single request */
549 	unsigned int max_pages;
550 
551 	/** Constrain ->max_pages to this value during feature negotiation */
552 	unsigned int max_pages_limit;
553 
554 	/** Input queue */
555 	struct fuse_iqueue iq;
556 
557 	/** The next unique kernel file handle */
558 	atomic64_t khctr;
559 
560 	/** rbtree of fuse_files waiting for poll events indexed by ph */
561 	struct rb_root polled_files;
562 
563 	/** Maximum number of outstanding background requests */
564 	unsigned max_background;
565 
566 	/** Number of background requests at which congestion starts */
567 	unsigned congestion_threshold;
568 
569 	/** Number of requests currently in the background */
570 	unsigned num_background;
571 
572 	/** Number of background requests currently queued for userspace */
573 	unsigned active_background;
574 
575 	/** The list of background requests set aside for later queuing */
576 	struct list_head bg_queue;
577 
578 	/** Protects: max_background, congestion_threshold, num_background,
579 	 * active_background, bg_queue, blocked */
580 	spinlock_t bg_lock;
581 
582 	/** Flag indicating that INIT reply has been received. Allocating
583 	 * any fuse request will be suspended until the flag is set */
584 	int initialized;
585 
586 	/** Flag indicating if connection is blocked.  This will be
587 	    the case before the INIT reply is received, and if there
588 	    are too many outstading backgrounds requests */
589 	int blocked;
590 
591 	/** waitq for blocked connection */
592 	wait_queue_head_t blocked_waitq;
593 
594 	/** Connection established, cleared on umount, connection
595 	    abort and device release */
596 	unsigned connected;
597 
598 	/** Connection aborted via sysfs */
599 	bool aborted;
600 
601 	/** Connection failed (version mismatch).  Cannot race with
602 	    setting other bitfields since it is only set once in INIT
603 	    reply, before any other request, and never cleared */
604 	unsigned conn_error:1;
605 
606 	/** Connection successful.  Only set in INIT */
607 	unsigned conn_init:1;
608 
609 	/** Do readpages asynchronously?  Only set in INIT */
610 	unsigned async_read:1;
611 
612 	/** Return an unique read error after abort.  Only set in INIT */
613 	unsigned abort_err:1;
614 
615 	/** Do not send separate SETATTR request before open(O_TRUNC)  */
616 	unsigned atomic_o_trunc:1;
617 
618 	/** Filesystem supports NFS exporting.  Only set in INIT */
619 	unsigned export_support:1;
620 
621 	/** write-back cache policy (default is write-through) */
622 	unsigned writeback_cache:1;
623 
624 	/** allow parallel lookups and readdir (default is serialized) */
625 	unsigned parallel_dirops:1;
626 
627 	/** handle fs handles killing suid/sgid/cap on write/chown/trunc */
628 	unsigned handle_killpriv:1;
629 
630 	/** cache READLINK responses in page cache */
631 	unsigned cache_symlinks:1;
632 
633 	/* show legacy mount options */
634 	unsigned int legacy_opts_show:1;
635 
636 	/*
637 	 * fs kills suid/sgid/cap on write/chown/trunc. suid is killed on
638 	 * write/trunc only if caller did not have CAP_FSETID.  sgid is killed
639 	 * on write/truncate only if caller did not have CAP_FSETID as well as
640 	 * file has group execute permission.
641 	 */
642 	unsigned handle_killpriv_v2:1;
643 
644 	/*
645 	 * The following bitfields are only for optimization purposes
646 	 * and hence races in setting them will not cause malfunction
647 	 */
648 
649 	/** Is open/release not implemented by fs? */
650 	unsigned no_open:1;
651 
652 	/** Is opendir/releasedir not implemented by fs? */
653 	unsigned no_opendir:1;
654 
655 	/** Is fsync not implemented by fs? */
656 	unsigned no_fsync:1;
657 
658 	/** Is fsyncdir not implemented by fs? */
659 	unsigned no_fsyncdir:1;
660 
661 	/** Is flush not implemented by fs? */
662 	unsigned no_flush:1;
663 
664 	/** Is setxattr not implemented by fs? */
665 	unsigned no_setxattr:1;
666 
667 	/** Does file server support extended setxattr */
668 	unsigned setxattr_ext:1;
669 
670 	/** Is getxattr not implemented by fs? */
671 	unsigned no_getxattr:1;
672 
673 	/** Is listxattr not implemented by fs? */
674 	unsigned no_listxattr:1;
675 
676 	/** Is removexattr not implemented by fs? */
677 	unsigned no_removexattr:1;
678 
679 	/** Are posix file locking primitives not implemented by fs? */
680 	unsigned no_lock:1;
681 
682 	/** Is access not implemented by fs? */
683 	unsigned no_access:1;
684 
685 	/** Is create not implemented by fs? */
686 	unsigned no_create:1;
687 
688 	/** Is interrupt not implemented by fs? */
689 	unsigned no_interrupt:1;
690 
691 	/** Is bmap not implemented by fs? */
692 	unsigned no_bmap:1;
693 
694 	/** Is poll not implemented by fs? */
695 	unsigned no_poll:1;
696 
697 	/** Do multi-page cached writes */
698 	unsigned big_writes:1;
699 
700 	/** Don't apply umask to creation modes */
701 	unsigned dont_mask:1;
702 
703 	/** Are BSD file locking primitives not implemented by fs? */
704 	unsigned no_flock:1;
705 
706 	/** Is fallocate not implemented by fs? */
707 	unsigned no_fallocate:1;
708 
709 	/** Is rename with flags implemented by fs? */
710 	unsigned no_rename2:1;
711 
712 	/** Use enhanced/automatic page cache invalidation. */
713 	unsigned auto_inval_data:1;
714 
715 	/** Filesystem is fully responsible for page cache invalidation. */
716 	unsigned explicit_inval_data:1;
717 
718 	/** Does the filesystem support readdirplus? */
719 	unsigned do_readdirplus:1;
720 
721 	/** Does the filesystem want adaptive readdirplus? */
722 	unsigned readdirplus_auto:1;
723 
724 	/** Does the filesystem support asynchronous direct-IO submission? */
725 	unsigned async_dio:1;
726 
727 	/** Is lseek not implemented by fs? */
728 	unsigned no_lseek:1;
729 
730 	/** Does the filesystem support posix acls? */
731 	unsigned posix_acl:1;
732 
733 	/** Check permissions based on the file mode or not? */
734 	unsigned default_permissions:1;
735 
736 	/** Allow other than the mounter user to access the filesystem ? */
737 	unsigned allow_other:1;
738 
739 	/** Does the filesystem support copy_file_range? */
740 	unsigned no_copy_file_range:1;
741 
742 	/* Send DESTROY request */
743 	unsigned int destroy:1;
744 
745 	/* Delete dentries that have gone stale */
746 	unsigned int delete_stale:1;
747 
748 	/** Do not create entry in fusectl fs */
749 	unsigned int no_control:1;
750 
751 	/** Do not allow MNT_FORCE umount */
752 	unsigned int no_force_umount:1;
753 
754 	/* Auto-mount submounts announced by the server */
755 	unsigned int auto_submounts:1;
756 
757 	/* Propagate syncfs() to server */
758 	unsigned int sync_fs:1;
759 
760 	/** The number of requests waiting for completion */
761 	atomic_t num_waiting;
762 
763 	/** Negotiated minor version */
764 	unsigned minor;
765 
766 	/** Entry on the fuse_mount_list */
767 	struct list_head entry;
768 
769 	/** Device ID from the root super block */
770 	dev_t dev;
771 
772 	/** Dentries in the control filesystem */
773 	struct dentry *ctl_dentry[FUSE_CTL_NUM_DENTRIES];
774 
775 	/** number of dentries used in the above array */
776 	int ctl_ndents;
777 
778 	/** Key for lock owner ID scrambling */
779 	u32 scramble_key[4];
780 
781 	/** Version counter for attribute changes */
782 	atomic64_t attr_version;
783 
784 	/** Called on final put */
785 	void (*release)(struct fuse_conn *);
786 
787 	/**
788 	 * Read/write semaphore to hold when accessing the sb of any
789 	 * fuse_mount belonging to this connection
790 	 */
791 	struct rw_semaphore killsb;
792 
793 	/** List of device instances belonging to this connection */
794 	struct list_head devices;
795 
796 #ifdef CONFIG_FUSE_DAX
797 	/* Dax specific conn data, non-NULL if DAX is enabled */
798 	struct fuse_conn_dax *dax;
799 #endif
800 
801 	/** List of filesystems using this connection */
802 	struct list_head mounts;
803 };
804 
805 /*
806  * Represents a mounted filesystem, potentially a submount.
807  *
808  * This object allows sharing a fuse_conn between separate mounts to
809  * allow submounts with dedicated superblocks and thus separate device
810  * IDs.
811  */
812 struct fuse_mount {
813 	/* Underlying (potentially shared) connection to the FUSE server */
814 	struct fuse_conn *fc;
815 
816 	/*
817 	 * Super block for this connection (fc->killsb must be held when
818 	 * accessing this).
819 	 */
820 	struct super_block *sb;
821 
822 	/* Entry on fc->mounts */
823 	struct list_head fc_entry;
824 };
825 
826 static inline struct fuse_mount *get_fuse_mount_super(struct super_block *sb)
827 {
828 	return sb->s_fs_info;
829 }
830 
831 static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb)
832 {
833 	return get_fuse_mount_super(sb)->fc;
834 }
835 
836 static inline struct fuse_mount *get_fuse_mount(struct inode *inode)
837 {
838 	return get_fuse_mount_super(inode->i_sb);
839 }
840 
841 static inline struct fuse_conn *get_fuse_conn(struct inode *inode)
842 {
843 	return get_fuse_mount_super(inode->i_sb)->fc;
844 }
845 
846 static inline struct fuse_inode *get_fuse_inode(struct inode *inode)
847 {
848 	return container_of(inode, struct fuse_inode, inode);
849 }
850 
851 static inline u64 get_node_id(struct inode *inode)
852 {
853 	return get_fuse_inode(inode)->nodeid;
854 }
855 
856 static inline int invalid_nodeid(u64 nodeid)
857 {
858 	return !nodeid || nodeid == FUSE_ROOT_ID;
859 }
860 
861 static inline u64 fuse_get_attr_version(struct fuse_conn *fc)
862 {
863 	return atomic64_read(&fc->attr_version);
864 }
865 
866 static inline bool fuse_stale_inode(const struct inode *inode, int generation,
867 				    struct fuse_attr *attr)
868 {
869 	return inode->i_generation != generation ||
870 		inode_wrong_type(inode, attr->mode);
871 }
872 
873 static inline void fuse_make_bad(struct inode *inode)
874 {
875 	remove_inode_hash(inode);
876 	set_bit(FUSE_I_BAD, &get_fuse_inode(inode)->state);
877 }
878 
879 static inline bool fuse_is_bad(struct inode *inode)
880 {
881 	return unlikely(test_bit(FUSE_I_BAD, &get_fuse_inode(inode)->state));
882 }
883 
884 static inline struct page **fuse_pages_alloc(unsigned int npages, gfp_t flags,
885 					     struct fuse_page_desc **desc)
886 {
887 	struct page **pages;
888 
889 	pages = kzalloc(npages * (sizeof(struct page *) +
890 				  sizeof(struct fuse_page_desc)), flags);
891 	*desc = (void *) (pages + npages);
892 
893 	return pages;
894 }
895 
896 static inline void fuse_page_descs_length_init(struct fuse_page_desc *descs,
897 					       unsigned int index,
898 					       unsigned int nr_pages)
899 {
900 	int i;
901 
902 	for (i = index; i < index + nr_pages; i++)
903 		descs[i].length = PAGE_SIZE - descs[i].offset;
904 }
905 
906 /** Device operations */
907 extern const struct file_operations fuse_dev_operations;
908 
909 extern const struct dentry_operations fuse_dentry_operations;
910 extern const struct dentry_operations fuse_root_dentry_operations;
911 
912 /**
913  * Get a filled in inode
914  */
915 struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
916 			int generation, struct fuse_attr *attr,
917 			u64 attr_valid, u64 attr_version);
918 
919 int fuse_lookup_name(struct super_block *sb, u64 nodeid, const struct qstr *name,
920 		     struct fuse_entry_out *outarg, struct inode **inode);
921 
922 /**
923  * Send FORGET command
924  */
925 void fuse_queue_forget(struct fuse_conn *fc, struct fuse_forget_link *forget,
926 		       u64 nodeid, u64 nlookup);
927 
928 struct fuse_forget_link *fuse_alloc_forget(void);
929 
930 struct fuse_forget_link *fuse_dequeue_forget(struct fuse_iqueue *fiq,
931 					     unsigned int max,
932 					     unsigned int *countp);
933 
934 /*
935  * Initialize READ or READDIR request
936  */
937 struct fuse_io_args {
938 	union {
939 		struct {
940 			struct fuse_read_in in;
941 			u64 attr_ver;
942 		} read;
943 		struct {
944 			struct fuse_write_in in;
945 			struct fuse_write_out out;
946 			bool page_locked;
947 		} write;
948 	};
949 	struct fuse_args_pages ap;
950 	struct fuse_io_priv *io;
951 	struct fuse_file *ff;
952 };
953 
954 void fuse_read_args_fill(struct fuse_io_args *ia, struct file *file, loff_t pos,
955 			 size_t count, int opcode);
956 
957 
958 /**
959  * Send OPEN or OPENDIR request
960  */
961 int fuse_open_common(struct inode *inode, struct file *file, bool isdir);
962 
963 struct fuse_file *fuse_file_alloc(struct fuse_mount *fm);
964 void fuse_file_free(struct fuse_file *ff);
965 void fuse_finish_open(struct inode *inode, struct file *file);
966 
967 void fuse_sync_release(struct fuse_inode *fi, struct fuse_file *ff,
968 		       unsigned int flags);
969 
970 /**
971  * Send RELEASE or RELEASEDIR request
972  */
973 void fuse_release_common(struct file *file, bool isdir);
974 
975 /**
976  * Send FSYNC or FSYNCDIR request
977  */
978 int fuse_fsync_common(struct file *file, loff_t start, loff_t end,
979 		      int datasync, int opcode);
980 
981 /**
982  * Notify poll wakeup
983  */
984 int fuse_notify_poll_wakeup(struct fuse_conn *fc,
985 			    struct fuse_notify_poll_wakeup_out *outarg);
986 
987 /**
988  * Initialize file operations on a regular file
989  */
990 void fuse_init_file_inode(struct inode *inode);
991 
992 /**
993  * Initialize inode operations on regular files and special files
994  */
995 void fuse_init_common(struct inode *inode);
996 
997 /**
998  * Initialize inode and file operations on a directory
999  */
1000 void fuse_init_dir(struct inode *inode);
1001 
1002 /**
1003  * Initialize inode operations on a symlink
1004  */
1005 void fuse_init_symlink(struct inode *inode);
1006 
1007 /**
1008  * Change attributes of an inode
1009  */
1010 void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
1011 			    u64 attr_valid, u64 attr_version);
1012 
1013 void fuse_change_attributes_common(struct inode *inode, struct fuse_attr *attr,
1014 				   u64 attr_valid);
1015 
1016 /**
1017  * Initialize the client device
1018  */
1019 int fuse_dev_init(void);
1020 
1021 /**
1022  * Cleanup the client device
1023  */
1024 void fuse_dev_cleanup(void);
1025 
1026 int fuse_ctl_init(void);
1027 void __exit fuse_ctl_cleanup(void);
1028 
1029 /**
1030  * Simple request sending that does request allocation and freeing
1031  */
1032 ssize_t fuse_simple_request(struct fuse_mount *fm, struct fuse_args *args);
1033 int fuse_simple_background(struct fuse_mount *fm, struct fuse_args *args,
1034 			   gfp_t gfp_flags);
1035 
1036 /**
1037  * End a finished request
1038  */
1039 void fuse_request_end(struct fuse_req *req);
1040 
1041 /* Abort all requests */
1042 void fuse_abort_conn(struct fuse_conn *fc);
1043 void fuse_wait_aborted(struct fuse_conn *fc);
1044 
1045 /**
1046  * Invalidate inode attributes
1047  */
1048 void fuse_invalidate_attr(struct inode *inode);
1049 
1050 void fuse_invalidate_entry_cache(struct dentry *entry);
1051 
1052 void fuse_invalidate_atime(struct inode *inode);
1053 
1054 u64 entry_attr_timeout(struct fuse_entry_out *o);
1055 void fuse_change_entry_timeout(struct dentry *entry, struct fuse_entry_out *o);
1056 
1057 /**
1058  * Acquire reference to fuse_conn
1059  */
1060 struct fuse_conn *fuse_conn_get(struct fuse_conn *fc);
1061 
1062 /**
1063  * Initialize fuse_conn
1064  */
1065 void fuse_conn_init(struct fuse_conn *fc, struct fuse_mount *fm,
1066 		    struct user_namespace *user_ns,
1067 		    const struct fuse_iqueue_ops *fiq_ops, void *fiq_priv);
1068 
1069 /**
1070  * Release reference to fuse_conn
1071  */
1072 void fuse_conn_put(struct fuse_conn *fc);
1073 
1074 struct fuse_dev *fuse_dev_alloc_install(struct fuse_conn *fc);
1075 struct fuse_dev *fuse_dev_alloc(void);
1076 void fuse_dev_install(struct fuse_dev *fud, struct fuse_conn *fc);
1077 void fuse_dev_free(struct fuse_dev *fud);
1078 void fuse_send_init(struct fuse_mount *fm);
1079 
1080 /**
1081  * Fill in superblock and initialize fuse connection
1082  * @sb: partially-initialized superblock to fill in
1083  * @ctx: mount context
1084  */
1085 int fuse_fill_super_common(struct super_block *sb, struct fuse_fs_context *ctx);
1086 
1087 /*
1088  * Remove the mount from the connection
1089  *
1090  * Returns whether this was the last mount
1091  */
1092 bool fuse_mount_remove(struct fuse_mount *fm);
1093 
1094 /*
1095  * Setup context ops for submounts
1096  */
1097 int fuse_init_fs_context_submount(struct fs_context *fsc);
1098 
1099 /*
1100  * Shut down the connection (possibly sending DESTROY request).
1101  */
1102 void fuse_conn_destroy(struct fuse_mount *fm);
1103 
1104 /**
1105  * Add connection to control filesystem
1106  */
1107 int fuse_ctl_add_conn(struct fuse_conn *fc);
1108 
1109 /**
1110  * Remove connection from control filesystem
1111  */
1112 void fuse_ctl_remove_conn(struct fuse_conn *fc);
1113 
1114 /**
1115  * Is file type valid?
1116  */
1117 int fuse_valid_type(int m);
1118 
1119 bool fuse_invalid_attr(struct fuse_attr *attr);
1120 
1121 /**
1122  * Is current process allowed to perform filesystem operation?
1123  */
1124 int fuse_allow_current_process(struct fuse_conn *fc);
1125 
1126 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id);
1127 
1128 void fuse_update_ctime(struct inode *inode);
1129 
1130 int fuse_update_attributes(struct inode *inode, struct file *file);
1131 
1132 void fuse_flush_writepages(struct inode *inode);
1133 
1134 void fuse_set_nowrite(struct inode *inode);
1135 void fuse_release_nowrite(struct inode *inode);
1136 
1137 /**
1138  * Scan all fuse_mounts belonging to fc to find the first where
1139  * ilookup5() returns a result.  Return that result and the
1140  * respective fuse_mount in *fm (unless fm is NULL).
1141  *
1142  * The caller must hold fc->killsb.
1143  */
1144 struct inode *fuse_ilookup(struct fuse_conn *fc, u64 nodeid,
1145 			   struct fuse_mount **fm);
1146 
1147 /**
1148  * File-system tells the kernel to invalidate cache for the given node id.
1149  */
1150 int fuse_reverse_inval_inode(struct fuse_conn *fc, u64 nodeid,
1151 			     loff_t offset, loff_t len);
1152 
1153 /**
1154  * File-system tells the kernel to invalidate parent attributes and
1155  * the dentry matching parent/name.
1156  *
1157  * If the child_nodeid is non-zero and:
1158  *    - matches the inode number for the dentry matching parent/name,
1159  *    - is not a mount point
1160  *    - is a file or oan empty directory
1161  * then the dentry is unhashed (d_delete()).
1162  */
1163 int fuse_reverse_inval_entry(struct fuse_conn *fc, u64 parent_nodeid,
1164 			     u64 child_nodeid, struct qstr *name);
1165 
1166 int fuse_do_open(struct fuse_mount *fm, u64 nodeid, struct file *file,
1167 		 bool isdir);
1168 
1169 /**
1170  * fuse_direct_io() flags
1171  */
1172 
1173 /** If set, it is WRITE; otherwise - READ */
1174 #define FUSE_DIO_WRITE (1 << 0)
1175 
1176 /** CUSE pass fuse_direct_io() a file which f_mapping->host is not from FUSE */
1177 #define FUSE_DIO_CUSE  (1 << 1)
1178 
1179 ssize_t fuse_direct_io(struct fuse_io_priv *io, struct iov_iter *iter,
1180 		       loff_t *ppos, int flags);
1181 long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
1182 		   unsigned int flags);
1183 long fuse_ioctl_common(struct file *file, unsigned int cmd,
1184 		       unsigned long arg, unsigned int flags);
1185 __poll_t fuse_file_poll(struct file *file, poll_table *wait);
1186 int fuse_dev_release(struct inode *inode, struct file *file);
1187 
1188 bool fuse_write_update_size(struct inode *inode, loff_t pos);
1189 
1190 int fuse_flush_times(struct inode *inode, struct fuse_file *ff);
1191 int fuse_write_inode(struct inode *inode, struct writeback_control *wbc);
1192 
1193 int fuse_do_setattr(struct dentry *dentry, struct iattr *attr,
1194 		    struct file *file);
1195 
1196 void fuse_set_initialized(struct fuse_conn *fc);
1197 
1198 void fuse_unlock_inode(struct inode *inode, bool locked);
1199 bool fuse_lock_inode(struct inode *inode);
1200 
1201 int fuse_setxattr(struct inode *inode, const char *name, const void *value,
1202 		  size_t size, int flags, unsigned int extra_flags);
1203 ssize_t fuse_getxattr(struct inode *inode, const char *name, void *value,
1204 		      size_t size);
1205 ssize_t fuse_listxattr(struct dentry *entry, char *list, size_t size);
1206 int fuse_removexattr(struct inode *inode, const char *name);
1207 extern const struct xattr_handler *fuse_xattr_handlers[];
1208 extern const struct xattr_handler *fuse_acl_xattr_handlers[];
1209 extern const struct xattr_handler *fuse_no_acl_xattr_handlers[];
1210 
1211 struct posix_acl;
1212 struct posix_acl *fuse_get_acl(struct inode *inode, int type, bool rcu);
1213 int fuse_set_acl(struct user_namespace *mnt_userns, struct inode *inode,
1214 		 struct posix_acl *acl, int type);
1215 
1216 /* readdir.c */
1217 int fuse_readdir(struct file *file, struct dir_context *ctx);
1218 
1219 /**
1220  * Return the number of bytes in an arguments list
1221  */
1222 unsigned int fuse_len_args(unsigned int numargs, struct fuse_arg *args);
1223 
1224 /**
1225  * Get the next unique ID for a request
1226  */
1227 u64 fuse_get_unique(struct fuse_iqueue *fiq);
1228 void fuse_free_conn(struct fuse_conn *fc);
1229 
1230 /* dax.c */
1231 
1232 #define FUSE_IS_DAX(inode) (IS_ENABLED(CONFIG_FUSE_DAX) && IS_DAX(inode))
1233 
1234 ssize_t fuse_dax_read_iter(struct kiocb *iocb, struct iov_iter *to);
1235 ssize_t fuse_dax_write_iter(struct kiocb *iocb, struct iov_iter *from);
1236 int fuse_dax_mmap(struct file *file, struct vm_area_struct *vma);
1237 int fuse_dax_break_layouts(struct inode *inode, u64 dmap_start, u64 dmap_end);
1238 int fuse_dax_conn_alloc(struct fuse_conn *fc, struct dax_device *dax_dev);
1239 void fuse_dax_conn_free(struct fuse_conn *fc);
1240 bool fuse_dax_inode_alloc(struct super_block *sb, struct fuse_inode *fi);
1241 void fuse_dax_inode_init(struct inode *inode);
1242 void fuse_dax_inode_cleanup(struct inode *inode);
1243 bool fuse_dax_check_alignment(struct fuse_conn *fc, unsigned int map_alignment);
1244 void fuse_dax_cancel_work(struct fuse_conn *fc);
1245 
1246 /* ioctl.c */
1247 long fuse_file_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1248 long fuse_file_compat_ioctl(struct file *file, unsigned int cmd,
1249 			    unsigned long arg);
1250 int fuse_fileattr_get(struct dentry *dentry, struct fileattr *fa);
1251 int fuse_fileattr_set(struct user_namespace *mnt_userns,
1252 		      struct dentry *dentry, struct fileattr *fa);
1253 
1254 /* file.c */
1255 
1256 struct fuse_file *fuse_file_open(struct fuse_mount *fm, u64 nodeid,
1257 				 unsigned int open_flags, bool isdir);
1258 void fuse_file_release(struct inode *inode, struct fuse_file *ff,
1259 		       unsigned int open_flags, fl_owner_t id, bool isdir);
1260 
1261 #endif /* _FS_FUSE_I_H */
1262