xref: /openbmc/linux/fs/fuse/fuse_i.h (revision f39650de)
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 	/**
153 	 * Can't take inode lock in fault path (leads to circular dependency).
154 	 * Introduce another semaphore which can be taken in fault path and
155 	 * then other filesystem paths can take this to block faults.
156 	 */
157 	struct rw_semaphore i_mmap_sem;
158 
159 #ifdef CONFIG_FUSE_DAX
160 	/*
161 	 * Dax specific inode data
162 	 */
163 	struct fuse_inode_dax *dax;
164 #endif
165 };
166 
167 /** FUSE inode state bits */
168 enum {
169 	/** Advise readdirplus  */
170 	FUSE_I_ADVISE_RDPLUS,
171 	/** Initialized with readdirplus */
172 	FUSE_I_INIT_RDPLUS,
173 	/** An operation changing file size is in progress  */
174 	FUSE_I_SIZE_UNSTABLE,
175 	/* Bad inode */
176 	FUSE_I_BAD,
177 };
178 
179 struct fuse_conn;
180 struct fuse_mount;
181 struct fuse_release_args;
182 
183 /** FUSE specific file data */
184 struct fuse_file {
185 	/** Fuse connection for this file */
186 	struct fuse_mount *fm;
187 
188 	/* Argument space reserved for release */
189 	struct fuse_release_args *release_args;
190 
191 	/** Kernel file handle guaranteed to be unique */
192 	u64 kh;
193 
194 	/** File handle used by userspace */
195 	u64 fh;
196 
197 	/** Node id of this file */
198 	u64 nodeid;
199 
200 	/** Refcount */
201 	refcount_t count;
202 
203 	/** FOPEN_* flags returned by open */
204 	u32 open_flags;
205 
206 	/** Entry on inode's write_files list */
207 	struct list_head write_entry;
208 
209 	/* Readdir related */
210 	struct {
211 		/*
212 		 * Protects below fields against (crazy) parallel readdir on
213 		 * same open file.  Uncontended in the normal case.
214 		 */
215 		struct mutex lock;
216 
217 		/* Dir stream position */
218 		loff_t pos;
219 
220 		/* Offset in cache */
221 		loff_t cache_off;
222 
223 		/* Version of cache we are reading */
224 		u64 version;
225 
226 	} readdir;
227 
228 	/** RB node to be linked on fuse_conn->polled_files */
229 	struct rb_node polled_node;
230 
231 	/** Wait queue head for poll */
232 	wait_queue_head_t poll_wait;
233 
234 	/** Has flock been performed on this file? */
235 	bool flock:1;
236 };
237 
238 /** One input argument of a request */
239 struct fuse_in_arg {
240 	unsigned size;
241 	const void *value;
242 };
243 
244 /** One output argument of a request */
245 struct fuse_arg {
246 	unsigned size;
247 	void *value;
248 };
249 
250 /** FUSE page descriptor */
251 struct fuse_page_desc {
252 	unsigned int length;
253 	unsigned int offset;
254 };
255 
256 struct fuse_args {
257 	uint64_t nodeid;
258 	uint32_t opcode;
259 	unsigned short in_numargs;
260 	unsigned short out_numargs;
261 	bool force:1;
262 	bool noreply:1;
263 	bool nocreds:1;
264 	bool in_pages:1;
265 	bool out_pages:1;
266 	bool out_argvar:1;
267 	bool page_zeroing:1;
268 	bool page_replace:1;
269 	bool may_block:1;
270 	struct fuse_in_arg in_args[3];
271 	struct fuse_arg out_args[2];
272 	void (*end)(struct fuse_mount *fm, struct fuse_args *args, int error);
273 };
274 
275 struct fuse_args_pages {
276 	struct fuse_args args;
277 	struct page **pages;
278 	struct fuse_page_desc *descs;
279 	unsigned int num_pages;
280 };
281 
282 #define FUSE_ARGS(args) struct fuse_args args = {}
283 
284 /** The request IO state (for asynchronous processing) */
285 struct fuse_io_priv {
286 	struct kref refcnt;
287 	int async;
288 	spinlock_t lock;
289 	unsigned reqs;
290 	ssize_t bytes;
291 	size_t size;
292 	__u64 offset;
293 	bool write;
294 	bool should_dirty;
295 	int err;
296 	struct kiocb *iocb;
297 	struct completion *done;
298 	bool blocking;
299 };
300 
301 #define FUSE_IO_PRIV_SYNC(i) \
302 {					\
303 	.refcnt = KREF_INIT(1),		\
304 	.async = 0,			\
305 	.iocb = i,			\
306 }
307 
308 /**
309  * Request flags
310  *
311  * FR_ISREPLY:		set if the request has reply
312  * FR_FORCE:		force sending of the request even if interrupted
313  * FR_BACKGROUND:	request is sent in the background
314  * FR_WAITING:		request is counted as "waiting"
315  * FR_ABORTED:		the request was aborted
316  * FR_INTERRUPTED:	the request has been interrupted
317  * FR_LOCKED:		data is being copied to/from the request
318  * FR_PENDING:		request is not yet in userspace
319  * FR_SENT:		request is in userspace, waiting for an answer
320  * FR_FINISHED:		request is finished
321  * FR_PRIVATE:		request is on private list
322  * FR_ASYNC:		request is asynchronous
323  */
324 enum fuse_req_flag {
325 	FR_ISREPLY,
326 	FR_FORCE,
327 	FR_BACKGROUND,
328 	FR_WAITING,
329 	FR_ABORTED,
330 	FR_INTERRUPTED,
331 	FR_LOCKED,
332 	FR_PENDING,
333 	FR_SENT,
334 	FR_FINISHED,
335 	FR_PRIVATE,
336 	FR_ASYNC,
337 };
338 
339 /**
340  * A request to the client
341  *
342  * .waitq.lock protects the following fields:
343  *   - FR_ABORTED
344  *   - FR_LOCKED (may also be modified under fc->lock, tested under both)
345  */
346 struct fuse_req {
347 	/** This can be on either pending processing or io lists in
348 	    fuse_conn */
349 	struct list_head list;
350 
351 	/** Entry on the interrupts list  */
352 	struct list_head intr_entry;
353 
354 	/* Input/output arguments */
355 	struct fuse_args *args;
356 
357 	/** refcount */
358 	refcount_t count;
359 
360 	/* Request flags, updated with test/set/clear_bit() */
361 	unsigned long flags;
362 
363 	/* The request input header */
364 	struct {
365 		struct fuse_in_header h;
366 	} in;
367 
368 	/* The request output header */
369 	struct {
370 		struct fuse_out_header h;
371 	} out;
372 
373 	/** Used to wake up the task waiting for completion of request*/
374 	wait_queue_head_t waitq;
375 
376 #if IS_ENABLED(CONFIG_VIRTIO_FS)
377 	/** virtio-fs's physically contiguous buffer for in and out args */
378 	void *argbuf;
379 #endif
380 
381 	/** fuse_mount this request belongs to */
382 	struct fuse_mount *fm;
383 };
384 
385 struct fuse_iqueue;
386 
387 /**
388  * Input queue callbacks
389  *
390  * Input queue signalling is device-specific.  For example, the /dev/fuse file
391  * uses fiq->waitq and fasync to wake processes that are waiting on queue
392  * readiness.  These callbacks allow other device types to respond to input
393  * queue activity.
394  */
395 struct fuse_iqueue_ops {
396 	/**
397 	 * Signal that a forget has been queued
398 	 */
399 	void (*wake_forget_and_unlock)(struct fuse_iqueue *fiq)
400 		__releases(fiq->lock);
401 
402 	/**
403 	 * Signal that an INTERRUPT request has been queued
404 	 */
405 	void (*wake_interrupt_and_unlock)(struct fuse_iqueue *fiq)
406 		__releases(fiq->lock);
407 
408 	/**
409 	 * Signal that a request has been queued
410 	 */
411 	void (*wake_pending_and_unlock)(struct fuse_iqueue *fiq)
412 		__releases(fiq->lock);
413 
414 	/**
415 	 * Clean up when fuse_iqueue is destroyed
416 	 */
417 	void (*release)(struct fuse_iqueue *fiq);
418 };
419 
420 /** /dev/fuse input queue operations */
421 extern const struct fuse_iqueue_ops fuse_dev_fiq_ops;
422 
423 struct fuse_iqueue {
424 	/** Connection established */
425 	unsigned connected;
426 
427 	/** Lock protecting accesses to members of this structure */
428 	spinlock_t lock;
429 
430 	/** Readers of the connection are waiting on this */
431 	wait_queue_head_t waitq;
432 
433 	/** The next unique request id */
434 	u64 reqctr;
435 
436 	/** The list of pending requests */
437 	struct list_head pending;
438 
439 	/** Pending interrupts */
440 	struct list_head interrupts;
441 
442 	/** Queue of pending forgets */
443 	struct fuse_forget_link forget_list_head;
444 	struct fuse_forget_link *forget_list_tail;
445 
446 	/** Batching of FORGET requests (positive indicates FORGET batch) */
447 	int forget_batch;
448 
449 	/** O_ASYNC requests */
450 	struct fasync_struct *fasync;
451 
452 	/** Device-specific callbacks */
453 	const struct fuse_iqueue_ops *ops;
454 
455 	/** Device-specific state */
456 	void *priv;
457 };
458 
459 #define FUSE_PQ_HASH_BITS 8
460 #define FUSE_PQ_HASH_SIZE (1 << FUSE_PQ_HASH_BITS)
461 
462 struct fuse_pqueue {
463 	/** Connection established */
464 	unsigned connected;
465 
466 	/** Lock protecting accessess to  members of this structure */
467 	spinlock_t lock;
468 
469 	/** Hash table of requests being processed */
470 	struct list_head *processing;
471 
472 	/** The list of requests under I/O */
473 	struct list_head io;
474 };
475 
476 /**
477  * Fuse device instance
478  */
479 struct fuse_dev {
480 	/** Fuse connection for this device */
481 	struct fuse_conn *fc;
482 
483 	/** Processing queue */
484 	struct fuse_pqueue pq;
485 
486 	/** list entry on fc->devices */
487 	struct list_head entry;
488 };
489 
490 struct fuse_fs_context {
491 	int fd;
492 	unsigned int rootmode;
493 	kuid_t user_id;
494 	kgid_t group_id;
495 	bool is_bdev:1;
496 	bool fd_present:1;
497 	bool rootmode_present:1;
498 	bool user_id_present:1;
499 	bool group_id_present:1;
500 	bool default_permissions:1;
501 	bool allow_other:1;
502 	bool destroy:1;
503 	bool no_control:1;
504 	bool no_force_umount:1;
505 	bool legacy_opts_show:1;
506 	bool dax:1;
507 	unsigned int max_read;
508 	unsigned int blksize;
509 	const char *subtype;
510 
511 	/* DAX device, may be NULL */
512 	struct dax_device *dax_dev;
513 
514 	/* fuse_dev pointer to fill in, should contain NULL on entry */
515 	void **fudptr;
516 };
517 
518 /**
519  * A Fuse connection.
520  *
521  * This structure is created, when the root filesystem is mounted, and
522  * is destroyed, when the client device is closed and the last
523  * fuse_mount is destroyed.
524  */
525 struct fuse_conn {
526 	/** Lock protecting accessess to  members of this structure */
527 	spinlock_t lock;
528 
529 	/** Refcount */
530 	refcount_t count;
531 
532 	/** Number of fuse_dev's */
533 	atomic_t dev_count;
534 
535 	struct rcu_head rcu;
536 
537 	/** The user id for this mount */
538 	kuid_t user_id;
539 
540 	/** The group id for this mount */
541 	kgid_t group_id;
542 
543 	/** The pid namespace for this mount */
544 	struct pid_namespace *pid_ns;
545 
546 	/** The user namespace for this mount */
547 	struct user_namespace *user_ns;
548 
549 	/** Maximum read size */
550 	unsigned max_read;
551 
552 	/** Maximum write size */
553 	unsigned max_write;
554 
555 	/** Maximum number of pages that can be used in a single request */
556 	unsigned int max_pages;
557 
558 	/** Constrain ->max_pages to this value during feature negotiation */
559 	unsigned int max_pages_limit;
560 
561 	/** Input queue */
562 	struct fuse_iqueue iq;
563 
564 	/** The next unique kernel file handle */
565 	atomic64_t khctr;
566 
567 	/** rbtree of fuse_files waiting for poll events indexed by ph */
568 	struct rb_root polled_files;
569 
570 	/** Maximum number of outstanding background requests */
571 	unsigned max_background;
572 
573 	/** Number of background requests at which congestion starts */
574 	unsigned congestion_threshold;
575 
576 	/** Number of requests currently in the background */
577 	unsigned num_background;
578 
579 	/** Number of background requests currently queued for userspace */
580 	unsigned active_background;
581 
582 	/** The list of background requests set aside for later queuing */
583 	struct list_head bg_queue;
584 
585 	/** Protects: max_background, congestion_threshold, num_background,
586 	 * active_background, bg_queue, blocked */
587 	spinlock_t bg_lock;
588 
589 	/** Flag indicating that INIT reply has been received. Allocating
590 	 * any fuse request will be suspended until the flag is set */
591 	int initialized;
592 
593 	/** Flag indicating if connection is blocked.  This will be
594 	    the case before the INIT reply is received, and if there
595 	    are too many outstading backgrounds requests */
596 	int blocked;
597 
598 	/** waitq for blocked connection */
599 	wait_queue_head_t blocked_waitq;
600 
601 	/** Connection established, cleared on umount, connection
602 	    abort and device release */
603 	unsigned connected;
604 
605 	/** Connection aborted via sysfs */
606 	bool aborted;
607 
608 	/** Connection failed (version mismatch).  Cannot race with
609 	    setting other bitfields since it is only set once in INIT
610 	    reply, before any other request, and never cleared */
611 	unsigned conn_error:1;
612 
613 	/** Connection successful.  Only set in INIT */
614 	unsigned conn_init:1;
615 
616 	/** Do readpages asynchronously?  Only set in INIT */
617 	unsigned async_read:1;
618 
619 	/** Return an unique read error after abort.  Only set in INIT */
620 	unsigned abort_err:1;
621 
622 	/** Do not send separate SETATTR request before open(O_TRUNC)  */
623 	unsigned atomic_o_trunc:1;
624 
625 	/** Filesystem supports NFS exporting.  Only set in INIT */
626 	unsigned export_support:1;
627 
628 	/** write-back cache policy (default is write-through) */
629 	unsigned writeback_cache:1;
630 
631 	/** allow parallel lookups and readdir (default is serialized) */
632 	unsigned parallel_dirops:1;
633 
634 	/** handle fs handles killing suid/sgid/cap on write/chown/trunc */
635 	unsigned handle_killpriv:1;
636 
637 	/** cache READLINK responses in page cache */
638 	unsigned cache_symlinks:1;
639 
640 	/* show legacy mount options */
641 	unsigned int legacy_opts_show:1;
642 
643 	/*
644 	 * fs kills suid/sgid/cap on write/chown/trunc. suid is killed on
645 	 * write/trunc only if caller did not have CAP_FSETID.  sgid is killed
646 	 * on write/truncate only if caller did not have CAP_FSETID as well as
647 	 * file has group execute permission.
648 	 */
649 	unsigned handle_killpriv_v2:1;
650 
651 	/*
652 	 * The following bitfields are only for optimization purposes
653 	 * and hence races in setting them will not cause malfunction
654 	 */
655 
656 	/** Is open/release not implemented by fs? */
657 	unsigned no_open:1;
658 
659 	/** Is opendir/releasedir not implemented by fs? */
660 	unsigned no_opendir:1;
661 
662 	/** Is fsync not implemented by fs? */
663 	unsigned no_fsync:1;
664 
665 	/** Is fsyncdir not implemented by fs? */
666 	unsigned no_fsyncdir:1;
667 
668 	/** Is flush not implemented by fs? */
669 	unsigned no_flush:1;
670 
671 	/** Is setxattr not implemented by fs? */
672 	unsigned no_setxattr:1;
673 
674 	/** Does file server support extended setxattr */
675 	unsigned setxattr_ext:1;
676 
677 	/** Is getxattr not implemented by fs? */
678 	unsigned no_getxattr:1;
679 
680 	/** Is listxattr not implemented by fs? */
681 	unsigned no_listxattr:1;
682 
683 	/** Is removexattr not implemented by fs? */
684 	unsigned no_removexattr:1;
685 
686 	/** Are posix file locking primitives not implemented by fs? */
687 	unsigned no_lock:1;
688 
689 	/** Is access not implemented by fs? */
690 	unsigned no_access:1;
691 
692 	/** Is create not implemented by fs? */
693 	unsigned no_create:1;
694 
695 	/** Is interrupt not implemented by fs? */
696 	unsigned no_interrupt:1;
697 
698 	/** Is bmap not implemented by fs? */
699 	unsigned no_bmap:1;
700 
701 	/** Is poll not implemented by fs? */
702 	unsigned no_poll:1;
703 
704 	/** Do multi-page cached writes */
705 	unsigned big_writes:1;
706 
707 	/** Don't apply umask to creation modes */
708 	unsigned dont_mask:1;
709 
710 	/** Are BSD file locking primitives not implemented by fs? */
711 	unsigned no_flock:1;
712 
713 	/** Is fallocate not implemented by fs? */
714 	unsigned no_fallocate:1;
715 
716 	/** Is rename with flags implemented by fs? */
717 	unsigned no_rename2:1;
718 
719 	/** Use enhanced/automatic page cache invalidation. */
720 	unsigned auto_inval_data:1;
721 
722 	/** Filesystem is fully responsible for page cache invalidation. */
723 	unsigned explicit_inval_data:1;
724 
725 	/** Does the filesystem support readdirplus? */
726 	unsigned do_readdirplus:1;
727 
728 	/** Does the filesystem want adaptive readdirplus? */
729 	unsigned readdirplus_auto:1;
730 
731 	/** Does the filesystem support asynchronous direct-IO submission? */
732 	unsigned async_dio:1;
733 
734 	/** Is lseek not implemented by fs? */
735 	unsigned no_lseek:1;
736 
737 	/** Does the filesystem support posix acls? */
738 	unsigned posix_acl:1;
739 
740 	/** Check permissions based on the file mode or not? */
741 	unsigned default_permissions:1;
742 
743 	/** Allow other than the mounter user to access the filesystem ? */
744 	unsigned allow_other:1;
745 
746 	/** Does the filesystem support copy_file_range? */
747 	unsigned no_copy_file_range:1;
748 
749 	/* Send DESTROY request */
750 	unsigned int destroy:1;
751 
752 	/* Delete dentries that have gone stale */
753 	unsigned int delete_stale:1;
754 
755 	/** Do not create entry in fusectl fs */
756 	unsigned int no_control:1;
757 
758 	/** Do not allow MNT_FORCE umount */
759 	unsigned int no_force_umount:1;
760 
761 	/* Auto-mount submounts announced by the server */
762 	unsigned int auto_submounts:1;
763 
764 	/** The number of requests waiting for completion */
765 	atomic_t num_waiting;
766 
767 	/** Negotiated minor version */
768 	unsigned minor;
769 
770 	/** Entry on the fuse_mount_list */
771 	struct list_head entry;
772 
773 	/** Device ID from the root super block */
774 	dev_t dev;
775 
776 	/** Dentries in the control filesystem */
777 	struct dentry *ctl_dentry[FUSE_CTL_NUM_DENTRIES];
778 
779 	/** number of dentries used in the above array */
780 	int ctl_ndents;
781 
782 	/** Key for lock owner ID scrambling */
783 	u32 scramble_key[4];
784 
785 	/** Version counter for attribute changes */
786 	atomic64_t attr_version;
787 
788 	/** Called on final put */
789 	void (*release)(struct fuse_conn *);
790 
791 	/**
792 	 * Read/write semaphore to hold when accessing the sb of any
793 	 * fuse_mount belonging to this connection
794 	 */
795 	struct rw_semaphore killsb;
796 
797 	/** List of device instances belonging to this connection */
798 	struct list_head devices;
799 
800 #ifdef CONFIG_FUSE_DAX
801 	/* Dax specific conn data, non-NULL if DAX is enabled */
802 	struct fuse_conn_dax *dax;
803 #endif
804 
805 	/** List of filesystems using this connection */
806 	struct list_head mounts;
807 };
808 
809 /*
810  * Represents a mounted filesystem, potentially a submount.
811  *
812  * This object allows sharing a fuse_conn between separate mounts to
813  * allow submounts with dedicated superblocks and thus separate device
814  * IDs.
815  */
816 struct fuse_mount {
817 	/* Underlying (potentially shared) connection to the FUSE server */
818 	struct fuse_conn *fc;
819 
820 	/*
821 	 * Super block for this connection (fc->killsb must be held when
822 	 * accessing this).
823 	 */
824 	struct super_block *sb;
825 
826 	/* Entry on fc->mounts */
827 	struct list_head fc_entry;
828 };
829 
830 static inline struct fuse_mount *get_fuse_mount_super(struct super_block *sb)
831 {
832 	return sb->s_fs_info;
833 }
834 
835 static inline struct fuse_conn *get_fuse_conn_super(struct super_block *sb)
836 {
837 	return get_fuse_mount_super(sb)->fc;
838 }
839 
840 static inline struct fuse_mount *get_fuse_mount(struct inode *inode)
841 {
842 	return get_fuse_mount_super(inode->i_sb);
843 }
844 
845 static inline struct fuse_conn *get_fuse_conn(struct inode *inode)
846 {
847 	return get_fuse_mount_super(inode->i_sb)->fc;
848 }
849 
850 static inline struct fuse_inode *get_fuse_inode(struct inode *inode)
851 {
852 	return container_of(inode, struct fuse_inode, inode);
853 }
854 
855 static inline u64 get_node_id(struct inode *inode)
856 {
857 	return get_fuse_inode(inode)->nodeid;
858 }
859 
860 static inline int invalid_nodeid(u64 nodeid)
861 {
862 	return !nodeid || nodeid == FUSE_ROOT_ID;
863 }
864 
865 static inline u64 fuse_get_attr_version(struct fuse_conn *fc)
866 {
867 	return atomic64_read(&fc->attr_version);
868 }
869 
870 static inline void fuse_make_bad(struct inode *inode)
871 {
872 	remove_inode_hash(inode);
873 	set_bit(FUSE_I_BAD, &get_fuse_inode(inode)->state);
874 }
875 
876 static inline bool fuse_is_bad(struct inode *inode)
877 {
878 	return unlikely(test_bit(FUSE_I_BAD, &get_fuse_inode(inode)->state));
879 }
880 
881 static inline struct page **fuse_pages_alloc(unsigned int npages, gfp_t flags,
882 					     struct fuse_page_desc **desc)
883 {
884 	struct page **pages;
885 
886 	pages = kzalloc(npages * (sizeof(struct page *) +
887 				  sizeof(struct fuse_page_desc)), flags);
888 	*desc = (void *) (pages + npages);
889 
890 	return pages;
891 }
892 
893 static inline void fuse_page_descs_length_init(struct fuse_page_desc *descs,
894 					       unsigned int index,
895 					       unsigned int nr_pages)
896 {
897 	int i;
898 
899 	for (i = index; i < index + nr_pages; i++)
900 		descs[i].length = PAGE_SIZE - descs[i].offset;
901 }
902 
903 /** Device operations */
904 extern const struct file_operations fuse_dev_operations;
905 
906 extern const struct dentry_operations fuse_dentry_operations;
907 extern const struct dentry_operations fuse_root_dentry_operations;
908 
909 /**
910  * Get a filled in inode
911  */
912 struct inode *fuse_iget(struct super_block *sb, u64 nodeid,
913 			int generation, struct fuse_attr *attr,
914 			u64 attr_valid, u64 attr_version);
915 
916 int fuse_lookup_name(struct super_block *sb, u64 nodeid, const struct qstr *name,
917 		     struct fuse_entry_out *outarg, struct inode **inode);
918 
919 /**
920  * Send FORGET command
921  */
922 void fuse_queue_forget(struct fuse_conn *fc, struct fuse_forget_link *forget,
923 		       u64 nodeid, u64 nlookup);
924 
925 struct fuse_forget_link *fuse_alloc_forget(void);
926 
927 struct fuse_forget_link *fuse_dequeue_forget(struct fuse_iqueue *fiq,
928 					     unsigned int max,
929 					     unsigned int *countp);
930 
931 /*
932  * Initialize READ or READDIR request
933  */
934 struct fuse_io_args {
935 	union {
936 		struct {
937 			struct fuse_read_in in;
938 			u64 attr_ver;
939 		} read;
940 		struct {
941 			struct fuse_write_in in;
942 			struct fuse_write_out out;
943 			bool page_locked;
944 		} write;
945 	};
946 	struct fuse_args_pages ap;
947 	struct fuse_io_priv *io;
948 	struct fuse_file *ff;
949 };
950 
951 void fuse_read_args_fill(struct fuse_io_args *ia, struct file *file, loff_t pos,
952 			 size_t count, int opcode);
953 
954 
955 /**
956  * Send OPEN or OPENDIR request
957  */
958 int fuse_open_common(struct inode *inode, struct file *file, bool isdir);
959 
960 struct fuse_file *fuse_file_alloc(struct fuse_mount *fm);
961 void fuse_file_free(struct fuse_file *ff);
962 void fuse_finish_open(struct inode *inode, struct file *file);
963 
964 void fuse_sync_release(struct fuse_inode *fi, struct fuse_file *ff,
965 		       unsigned int flags);
966 
967 /**
968  * Send RELEASE or RELEASEDIR request
969  */
970 void fuse_release_common(struct file *file, bool isdir);
971 
972 /**
973  * Send FSYNC or FSYNCDIR request
974  */
975 int fuse_fsync_common(struct file *file, loff_t start, loff_t end,
976 		      int datasync, int opcode);
977 
978 /**
979  * Notify poll wakeup
980  */
981 int fuse_notify_poll_wakeup(struct fuse_conn *fc,
982 			    struct fuse_notify_poll_wakeup_out *outarg);
983 
984 /**
985  * Initialize file operations on a regular file
986  */
987 void fuse_init_file_inode(struct inode *inode);
988 
989 /**
990  * Initialize inode operations on regular files and special files
991  */
992 void fuse_init_common(struct inode *inode);
993 
994 /**
995  * Initialize inode and file operations on a directory
996  */
997 void fuse_init_dir(struct inode *inode);
998 
999 /**
1000  * Initialize inode operations on a symlink
1001  */
1002 void fuse_init_symlink(struct inode *inode);
1003 
1004 /**
1005  * Change attributes of an inode
1006  */
1007 void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr,
1008 			    u64 attr_valid, u64 attr_version);
1009 
1010 void fuse_change_attributes_common(struct inode *inode, struct fuse_attr *attr,
1011 				   u64 attr_valid);
1012 
1013 /**
1014  * Initialize the client device
1015  */
1016 int fuse_dev_init(void);
1017 
1018 /**
1019  * Cleanup the client device
1020  */
1021 void fuse_dev_cleanup(void);
1022 
1023 int fuse_ctl_init(void);
1024 void __exit fuse_ctl_cleanup(void);
1025 
1026 /**
1027  * Simple request sending that does request allocation and freeing
1028  */
1029 ssize_t fuse_simple_request(struct fuse_mount *fm, struct fuse_args *args);
1030 int fuse_simple_background(struct fuse_mount *fm, struct fuse_args *args,
1031 			   gfp_t gfp_flags);
1032 
1033 /**
1034  * End a finished request
1035  */
1036 void fuse_request_end(struct fuse_req *req);
1037 
1038 /* Abort all requests */
1039 void fuse_abort_conn(struct fuse_conn *fc);
1040 void fuse_wait_aborted(struct fuse_conn *fc);
1041 
1042 /**
1043  * Invalidate inode attributes
1044  */
1045 void fuse_invalidate_attr(struct inode *inode);
1046 
1047 void fuse_invalidate_entry_cache(struct dentry *entry);
1048 
1049 void fuse_invalidate_atime(struct inode *inode);
1050 
1051 u64 entry_attr_timeout(struct fuse_entry_out *o);
1052 void fuse_change_entry_timeout(struct dentry *entry, struct fuse_entry_out *o);
1053 
1054 /**
1055  * Acquire reference to fuse_conn
1056  */
1057 struct fuse_conn *fuse_conn_get(struct fuse_conn *fc);
1058 
1059 /**
1060  * Initialize fuse_conn
1061  */
1062 void fuse_conn_init(struct fuse_conn *fc, struct fuse_mount *fm,
1063 		    struct user_namespace *user_ns,
1064 		    const struct fuse_iqueue_ops *fiq_ops, void *fiq_priv);
1065 
1066 /**
1067  * Release reference to fuse_conn
1068  */
1069 void fuse_conn_put(struct fuse_conn *fc);
1070 
1071 struct fuse_dev *fuse_dev_alloc_install(struct fuse_conn *fc);
1072 struct fuse_dev *fuse_dev_alloc(void);
1073 void fuse_dev_install(struct fuse_dev *fud, struct fuse_conn *fc);
1074 void fuse_dev_free(struct fuse_dev *fud);
1075 void fuse_send_init(struct fuse_mount *fm);
1076 
1077 /**
1078  * Fill in superblock and initialize fuse connection
1079  * @sb: partially-initialized superblock to fill in
1080  * @ctx: mount context
1081  */
1082 int fuse_fill_super_common(struct super_block *sb, struct fuse_fs_context *ctx);
1083 
1084 /*
1085  * Fill in superblock for submounts
1086  * @sb: partially-initialized superblock to fill in
1087  * @parent_fi: The fuse_inode of the parent filesystem where this submount is
1088  * 	       mounted
1089  */
1090 int fuse_fill_super_submount(struct super_block *sb,
1091 			     struct fuse_inode *parent_fi);
1092 
1093 /*
1094  * Remove the mount from the connection
1095  *
1096  * Returns whether this was the last mount
1097  */
1098 bool fuse_mount_remove(struct fuse_mount *fm);
1099 
1100 /*
1101  * Shut down the connection (possibly sending DESTROY request).
1102  */
1103 void fuse_conn_destroy(struct fuse_mount *fm);
1104 
1105 /**
1106  * Add connection to control filesystem
1107  */
1108 int fuse_ctl_add_conn(struct fuse_conn *fc);
1109 
1110 /**
1111  * Remove connection from control filesystem
1112  */
1113 void fuse_ctl_remove_conn(struct fuse_conn *fc);
1114 
1115 /**
1116  * Is file type valid?
1117  */
1118 int fuse_valid_type(int m);
1119 
1120 bool fuse_invalid_attr(struct fuse_attr *attr);
1121 
1122 /**
1123  * Is current process allowed to perform filesystem operation?
1124  */
1125 int fuse_allow_current_process(struct fuse_conn *fc);
1126 
1127 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id);
1128 
1129 void fuse_update_ctime(struct inode *inode);
1130 
1131 int fuse_update_attributes(struct inode *inode, struct file *file);
1132 
1133 void fuse_flush_writepages(struct inode *inode);
1134 
1135 void fuse_set_nowrite(struct inode *inode);
1136 void fuse_release_nowrite(struct inode *inode);
1137 
1138 /**
1139  * Scan all fuse_mounts belonging to fc to find the first where
1140  * ilookup5() returns a result.  Return that result and the
1141  * respective fuse_mount in *fm (unless fm is NULL).
1142  *
1143  * The caller must hold fc->killsb.
1144  */
1145 struct inode *fuse_ilookup(struct fuse_conn *fc, u64 nodeid,
1146 			   struct fuse_mount **fm);
1147 
1148 /**
1149  * File-system tells the kernel to invalidate cache for the given node id.
1150  */
1151 int fuse_reverse_inval_inode(struct fuse_conn *fc, u64 nodeid,
1152 			     loff_t offset, loff_t len);
1153 
1154 /**
1155  * File-system tells the kernel to invalidate parent attributes and
1156  * the dentry matching parent/name.
1157  *
1158  * If the child_nodeid is non-zero and:
1159  *    - matches the inode number for the dentry matching parent/name,
1160  *    - is not a mount point
1161  *    - is a file or oan empty directory
1162  * then the dentry is unhashed (d_delete()).
1163  */
1164 int fuse_reverse_inval_entry(struct fuse_conn *fc, u64 parent_nodeid,
1165 			     u64 child_nodeid, struct qstr *name);
1166 
1167 int fuse_do_open(struct fuse_mount *fm, u64 nodeid, struct file *file,
1168 		 bool isdir);
1169 
1170 /**
1171  * fuse_direct_io() flags
1172  */
1173 
1174 /** If set, it is WRITE; otherwise - READ */
1175 #define FUSE_DIO_WRITE (1 << 0)
1176 
1177 /** CUSE pass fuse_direct_io() a file which f_mapping->host is not from FUSE */
1178 #define FUSE_DIO_CUSE  (1 << 1)
1179 
1180 ssize_t fuse_direct_io(struct fuse_io_priv *io, struct iov_iter *iter,
1181 		       loff_t *ppos, int flags);
1182 long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
1183 		   unsigned int flags);
1184 long fuse_ioctl_common(struct file *file, unsigned int cmd,
1185 		       unsigned long arg, unsigned int flags);
1186 __poll_t fuse_file_poll(struct file *file, poll_table *wait);
1187 int fuse_dev_release(struct inode *inode, struct file *file);
1188 
1189 bool fuse_write_update_size(struct inode *inode, loff_t pos);
1190 
1191 int fuse_flush_times(struct inode *inode, struct fuse_file *ff);
1192 int fuse_write_inode(struct inode *inode, struct writeback_control *wbc);
1193 
1194 int fuse_do_setattr(struct dentry *dentry, struct iattr *attr,
1195 		    struct file *file);
1196 
1197 void fuse_set_initialized(struct fuse_conn *fc);
1198 
1199 void fuse_unlock_inode(struct inode *inode, bool locked);
1200 bool fuse_lock_inode(struct inode *inode);
1201 
1202 int fuse_setxattr(struct inode *inode, const char *name, const void *value,
1203 		  size_t size, int flags, unsigned int extra_flags);
1204 ssize_t fuse_getxattr(struct inode *inode, const char *name, void *value,
1205 		      size_t size);
1206 ssize_t fuse_listxattr(struct dentry *entry, char *list, size_t size);
1207 int fuse_removexattr(struct inode *inode, const char *name);
1208 extern const struct xattr_handler *fuse_xattr_handlers[];
1209 extern const struct xattr_handler *fuse_acl_xattr_handlers[];
1210 extern const struct xattr_handler *fuse_no_acl_xattr_handlers[];
1211 
1212 struct posix_acl;
1213 struct posix_acl *fuse_get_acl(struct inode *inode, int type);
1214 int fuse_set_acl(struct user_namespace *mnt_userns, struct inode *inode,
1215 		 struct posix_acl *acl, int type);
1216 
1217 /* readdir.c */
1218 int fuse_readdir(struct file *file, struct dir_context *ctx);
1219 
1220 /**
1221  * Return the number of bytes in an arguments list
1222  */
1223 unsigned int fuse_len_args(unsigned int numargs, struct fuse_arg *args);
1224 
1225 /**
1226  * Get the next unique ID for a request
1227  */
1228 u64 fuse_get_unique(struct fuse_iqueue *fiq);
1229 void fuse_free_conn(struct fuse_conn *fc);
1230 
1231 /* dax.c */
1232 
1233 #define FUSE_IS_DAX(inode) (IS_ENABLED(CONFIG_FUSE_DAX) && IS_DAX(inode))
1234 
1235 ssize_t fuse_dax_read_iter(struct kiocb *iocb, struct iov_iter *to);
1236 ssize_t fuse_dax_write_iter(struct kiocb *iocb, struct iov_iter *from);
1237 int fuse_dax_mmap(struct file *file, struct vm_area_struct *vma);
1238 int fuse_dax_break_layouts(struct inode *inode, u64 dmap_start, u64 dmap_end);
1239 int fuse_dax_conn_alloc(struct fuse_conn *fc, struct dax_device *dax_dev);
1240 void fuse_dax_conn_free(struct fuse_conn *fc);
1241 bool fuse_dax_inode_alloc(struct super_block *sb, struct fuse_inode *fi);
1242 void fuse_dax_inode_init(struct inode *inode);
1243 void fuse_dax_inode_cleanup(struct inode *inode);
1244 bool fuse_dax_check_alignment(struct fuse_conn *fc, unsigned int map_alignment);
1245 void fuse_dax_cancel_work(struct fuse_conn *fc);
1246 
1247 /* ioctl.c */
1248 long fuse_file_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1249 long fuse_file_compat_ioctl(struct file *file, unsigned int cmd,
1250 			    unsigned long arg);
1251 int fuse_fileattr_get(struct dentry *dentry, struct fileattr *fa);
1252 int fuse_fileattr_set(struct user_namespace *mnt_userns,
1253 		      struct dentry *dentry, struct fileattr *fa);
1254 
1255 /* file.c */
1256 
1257 struct fuse_file *fuse_file_open(struct fuse_mount *fm, u64 nodeid,
1258 				 unsigned int open_flags, bool isdir);
1259 void fuse_file_release(struct inode *inode, struct fuse_file *ff,
1260 		       unsigned int open_flags, fl_owner_t id, bool isdir);
1261 
1262 #endif /* _FS_FUSE_I_H */
1263