1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1989, 1991, 1993
5 *	The Regents of the University of California.  All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 *    may be used to endorse or promote products derived from this software
17 *    without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32#ifndef _SYS_MOUNT_H_
33#define _SYS_MOUNT_H_
34
35#include <sys/ucred.h>
36#include <sys/queue.h>
37#ifdef _KERNEL
38#include <sys/types.h>
39#include <sys/lock.h>
40#include <sys/lockmgr.h>
41#include <sys/tslog.h>
42#include <sys/_mutex.h>
43#include <sys/_sx.h>
44#endif
45
46/*
47 * NOTE: When changing statfs structure, mount structure, MNT_* flags or
48 * MNTK_* flags also update DDB show mount command in vfs_subr.c.
49 */
50
51typedef struct fsid { int32_t val[2]; } fsid_t;	/* filesystem id type */
52
53/* Returns non-zero if fsids are different. */
54static __inline int
55fsidcmp(const fsid_t *a, const fsid_t *b)
56{
57	return (a->val[0] != b->val[0] || a->val[1] != b->val[1]);
58}
59
60/*
61 * File identifier.
62 * These are unique per filesystem on a single machine.
63 *
64 * Note that the offset of fid_data is 4 bytes, so care must be taken to avoid
65 * undefined behavior accessing unaligned fields within an embedded struct.
66 */
67#define	MAXFIDSZ	16
68
69struct fid {
70	u_short		fid_len;		/* length of data in bytes */
71	u_short		fid_data0;		/* force longword alignment */
72	char		fid_data[MAXFIDSZ];	/* data (variable length) */
73};
74
75/*
76 * filesystem statistics
77 */
78#define	MFSNAMELEN	16		/* length of type name including null */
79#define	MNAMELEN	1024		/* size of on/from name bufs */
80#define	STATFS_VERSION	0x20140518	/* current version number */
81struct statfs {
82	uint32_t f_version;		/* structure version number */
83	uint32_t f_type;		/* type of filesystem */
84	uint64_t f_flags;		/* copy of mount exported flags */
85	uint64_t f_bsize;		/* filesystem fragment size */
86	uint64_t f_iosize;		/* optimal transfer block size */
87	uint64_t f_blocks;		/* total data blocks in filesystem */
88	uint64_t f_bfree;		/* free blocks in filesystem */
89	int64_t	 f_bavail;		/* free blocks avail to non-superuser */
90	uint64_t f_files;		/* total file nodes in filesystem */
91	int64_t	 f_ffree;		/* free nodes avail to non-superuser */
92	uint64_t f_syncwrites;		/* count of sync writes since mount */
93	uint64_t f_asyncwrites;		/* count of async writes since mount */
94	uint64_t f_syncreads;		/* count of sync reads since mount */
95	uint64_t f_asyncreads;		/* count of async reads since mount */
96	uint32_t f_nvnodelistsize;	/* # of vnodes */
97	uint32_t f_spare0;		/* unused spare */
98	uint64_t f_spare[9];		/* unused spare */
99	uint32_t f_namemax;		/* maximum filename length */
100	uid_t	  f_owner;		/* user that mounted the filesystem */
101	fsid_t	  f_fsid;		/* filesystem id */
102	char	  f_charspare[80];	    /* spare string space */
103	char	  f_fstypename[MFSNAMELEN]; /* filesystem type name */
104	char	  f_mntfromname[MNAMELEN];  /* mounted filesystem */
105	char	  f_mntonname[MNAMELEN];    /* directory on which mounted */
106};
107
108#if defined(_WANT_FREEBSD11_STATFS) || defined(_KERNEL)
109#define	FREEBSD11_STATFS_VERSION	0x20030518 /* current version number */
110struct freebsd11_statfs {
111	uint32_t f_version;		/* structure version number */
112	uint32_t f_type;		/* type of filesystem */
113	uint64_t f_flags;		/* copy of mount exported flags */
114	uint64_t f_bsize;		/* filesystem fragment size */
115	uint64_t f_iosize;		/* optimal transfer block size */
116	uint64_t f_blocks;		/* total data blocks in filesystem */
117	uint64_t f_bfree;		/* free blocks in filesystem */
118	int64_t	 f_bavail;		/* free blocks avail to non-superuser */
119	uint64_t f_files;		/* total file nodes in filesystem */
120	int64_t	 f_ffree;		/* free nodes avail to non-superuser */
121	uint64_t f_syncwrites;		/* count of sync writes since mount */
122	uint64_t f_asyncwrites;		/* count of async writes since mount */
123	uint64_t f_syncreads;		/* count of sync reads since mount */
124	uint64_t f_asyncreads;		/* count of async reads since mount */
125	uint64_t f_spare[10];		/* unused spare */
126	uint32_t f_namemax;		/* maximum filename length */
127	uid_t	  f_owner;		/* user that mounted the filesystem */
128	fsid_t	  f_fsid;		/* filesystem id */
129	char	  f_charspare[80];	/* spare string space */
130	char	  f_fstypename[16];	/* filesystem type name */
131	char	  f_mntfromname[88];	/* mounted filesystem */
132	char	  f_mntonname[88];	/* directory on which mounted */
133};
134#endif /* _WANT_FREEBSD11_STATFS || _KERNEL */
135
136#ifdef _KERNEL
137#define	OMFSNAMELEN	16	/* length of fs type name, including null */
138#define	OMNAMELEN	(88 - 2 * sizeof(long))	/* size of on/from name bufs */
139
140/* XXX getfsstat.2 is out of date with write and read counter changes here. */
141/* XXX statfs.2 is out of date with read counter changes here. */
142struct ostatfs {
143	long	f_spare2;		/* placeholder */
144	long	f_bsize;		/* fundamental filesystem block size */
145	long	f_iosize;		/* optimal transfer block size */
146	long	f_blocks;		/* total data blocks in filesystem */
147	long	f_bfree;		/* free blocks in fs */
148	long	f_bavail;		/* free blocks avail to non-superuser */
149	long	f_files;		/* total file nodes in filesystem */
150	long	f_ffree;		/* free file nodes in fs */
151	fsid_t	f_fsid;			/* filesystem id */
152	uid_t	f_owner;		/* user that mounted the filesystem */
153	int	f_type;			/* type of filesystem */
154	int	f_flags;		/* copy of mount exported flags */
155	long	f_syncwrites;		/* count of sync writes since mount */
156	long	f_asyncwrites;		/* count of async writes since mount */
157	char	f_fstypename[OMFSNAMELEN]; /* fs type name */
158	char	f_mntonname[OMNAMELEN];	/* directory on which mounted */
159	long	f_syncreads;		/* count of sync reads since mount */
160	long	f_asyncreads;		/* count of async reads since mount */
161	short	f_spares1;		/* unused spare */
162	char	f_mntfromname[OMNAMELEN];/* mounted filesystem */
163	short	f_spares2;		/* unused spare */
164	/*
165	 * XXX on machines where longs are aligned to 8-byte boundaries, there
166	 * is an unnamed int32_t here.  This spare was after the apparent end
167	 * of the struct until we bit off the read counters from f_mntonname.
168	 */
169	long	f_spare[2];		/* unused spare */
170};
171#endif	/* _KERNEL */
172
173#if defined(_WANT_MOUNT) || defined(_KERNEL)
174TAILQ_HEAD(vnodelst, vnode);
175
176/* Mount options list */
177TAILQ_HEAD(vfsoptlist, vfsopt);
178struct vfsopt {
179	TAILQ_ENTRY(vfsopt) link;
180	char	*name;
181	void	*value;
182	int	len;
183	int	pos;
184	int	seen;
185};
186
187struct mount_pcpu {
188	int		mntp_thread_in_ops;
189	int		mntp_ref;
190	int		mntp_lockref;
191	int		mntp_writeopcount;
192};
193
194_Static_assert(sizeof(struct mount_pcpu) == 16,
195    "the struct is allocated from pcpu 16 zone");
196
197/*
198 * Structure for tracking a stacked filesystem mounted above another
199 * filesystem.  This is expected to be stored in the upper FS' per-mount data.
200 *
201 * Lock reference:
202 *	i - lower mount interlock
203 *	c - constant from node initialization
204 */
205struct mount_upper_node {
206	struct mount 	*mp;	/* (c) mount object for upper FS */
207	TAILQ_ENTRY(mount_upper_node) mnt_upper_link;	/* (i) position in uppers list */
208};
209
210/*
211 * Structure per mounted filesystem.  Each mounted filesystem has an
212 * array of operations and an instance record.  The filesystems are
213 * put on a doubly linked list.
214 *
215 * Lock reference:
216 * 	l - mnt_listmtx
217 *	m - mountlist_mtx
218 *	i - interlock
219 *	v - vnode freelist mutex
220 *	d - deferred unmount list mutex
221 *	e - mnt_explock
222 *
223 * Unmarked fields are considered stable as long as a ref is held.
224 *
225 */
226struct mount {
227	int 		mnt_vfs_ops;		/* (i) pending vfs ops */
228	int		mnt_kern_flag;		/* (i) kernel only flags */
229	uint64_t	mnt_flag;		/* (i) flags shared with user */
230	struct mount_pcpu *mnt_pcpu;		/* per-CPU data */
231	struct vnode	*mnt_rootvnode;
232	struct vnode	*mnt_vnodecovered;	/* vnode we mounted on */
233	struct vfsops	*mnt_op;		/* operations on fs */
234	struct vfsconf	*mnt_vfc;		/* configuration info */
235	struct mtx __aligned(CACHE_LINE_SIZE)	mnt_mtx; /* mount structure interlock */
236	int		mnt_gen;		/* struct mount generation */
237#define	mnt_startzero	mnt_list
238	TAILQ_ENTRY(mount) mnt_list;		/* (m) mount list */
239	struct vnode	*mnt_syncer;		/* syncer vnode */
240	int		mnt_ref;		/* (i) Reference count */
241	struct vnodelst	mnt_nvnodelist;		/* (i) list of vnodes */
242	int		mnt_nvnodelistsize;	/* (i) # of vnodes */
243	int		mnt_writeopcount;	/* (i) write syscalls pending */
244	struct vfsoptlist *mnt_opt;		/* current mount options */
245	struct vfsoptlist *mnt_optnew;		/* new options passed to fs */
246	struct statfs	mnt_stat;		/* cache of filesystem stats */
247	struct ucred	*mnt_cred;		/* credentials of mounter */
248	void *		mnt_data;		/* private data */
249	time_t		mnt_time;		/* last time written*/
250	int		mnt_iosize_max;		/* max size for clusters, etc */
251	struct netexport *mnt_export;		/* (e) export list */
252	struct label	*mnt_label;		/* MAC label for the fs */
253	u_int		mnt_hashseed;		/* Random seed for vfs_hash */
254	int		mnt_lockref;		/* (i) Lock reference count */
255	int		mnt_secondary_writes;   /* (i) # of secondary writes */
256	int		mnt_secondary_accwrites;/* (i) secondary wr. starts */
257	struct thread	*mnt_susp_owner;	/* (i) thread owning suspension */
258	struct ucred	*mnt_exjail;		/* (i) jail which did exports */
259#define	mnt_endzero	mnt_gjprovider
260	char		*mnt_gjprovider;	/* gjournal provider name */
261	struct mtx	mnt_listmtx;
262	struct vnodelst	mnt_lazyvnodelist;	/* (l) list of lazy vnodes */
263	int		mnt_lazyvnodelistsize;	/* (l) # of lazy vnodes */
264	int		mnt_upper_pending;	/* (i) # of pending ops on mnt_uppers */
265	struct lock	mnt_explock;		/* vfs_export walkers lock */
266	TAILQ_HEAD(, mount_upper_node) mnt_uppers; /* (i) upper mounts over us */
267	TAILQ_HEAD(, mount_upper_node) mnt_notify; /* (i) upper mounts for notification */
268	STAILQ_ENTRY(mount) mnt_taskqueue_link;	/* (d) our place in deferred unmount list */
269	uint64_t	mnt_taskqueue_flags;	/* (d) unmount flags passed from taskqueue */
270	unsigned int	mnt_unmount_retries;	/* (d) # of failed deferred unmount attempts */
271};
272#endif	/* _WANT_MOUNT || _KERNEL */
273
274#ifdef _KERNEL
275/*
276 * Definitions for MNT_VNODE_FOREACH_ALL.
277 */
278struct vnode *__mnt_vnode_next_all(struct vnode **mvp, struct mount *mp);
279struct vnode *__mnt_vnode_first_all(struct vnode **mvp, struct mount *mp);
280void          __mnt_vnode_markerfree_all(struct vnode **mvp, struct mount *mp);
281
282#define MNT_VNODE_FOREACH_ALL(vp, mp, mvp)				\
283	for (vp = __mnt_vnode_first_all(&(mvp), (mp));			\
284		(vp) != NULL; vp = __mnt_vnode_next_all(&(mvp), (mp)))
285
286#define MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp)				\
287	do {								\
288		MNT_ILOCK(mp);						\
289		__mnt_vnode_markerfree_all(&(mvp), (mp));		\
290		/* MNT_IUNLOCK(mp); -- done in above function */	\
291		mtx_assert(MNT_MTX(mp), MA_NOTOWNED);			\
292	} while (0)
293
294/*
295 * Definitions for MNT_VNODE_FOREACH_LAZY.
296 */
297typedef int mnt_lazy_cb_t(struct vnode *, void *);
298struct vnode *__mnt_vnode_next_lazy(struct vnode **mvp, struct mount *mp,
299    mnt_lazy_cb_t *cb, void *cbarg);
300struct vnode *__mnt_vnode_first_lazy(struct vnode **mvp, struct mount *mp,
301    mnt_lazy_cb_t *cb, void *cbarg);
302void          __mnt_vnode_markerfree_lazy(struct vnode **mvp, struct mount *mp);
303
304#define MNT_VNODE_FOREACH_LAZY(vp, mp, mvp, cb, cbarg)			\
305	for (vp = __mnt_vnode_first_lazy(&(mvp), (mp), (cb), (cbarg));	\
306		(vp) != NULL; 						\
307		vp = __mnt_vnode_next_lazy(&(mvp), (mp), (cb), (cbarg)))
308
309#define MNT_VNODE_FOREACH_LAZY_ABORT(mp, mvp)				\
310	__mnt_vnode_markerfree_lazy(&(mvp), (mp))
311
312#define	MNT_ILOCK(mp)	mtx_lock(&(mp)->mnt_mtx)
313#define	MNT_ITRYLOCK(mp) mtx_trylock(&(mp)->mnt_mtx)
314#define	MNT_IUNLOCK(mp)	mtx_unlock(&(mp)->mnt_mtx)
315#define	MNT_MTX(mp)	(&(mp)->mnt_mtx)
316
317#define	MNT_REF(mp)	do {						\
318	mtx_assert(MNT_MTX(mp), MA_OWNED);				\
319	mp->mnt_ref++;							\
320} while (0)
321#define	MNT_REL(mp)	do {						\
322	mtx_assert(MNT_MTX(mp), MA_OWNED);				\
323	(mp)->mnt_ref--;						\
324	if ((mp)->mnt_vfs_ops && (mp)->mnt_ref < 0)		\
325		vfs_dump_mount_counters(mp);				\
326	if ((mp)->mnt_ref == 0 && (mp)->mnt_vfs_ops)		\
327		wakeup((mp));						\
328} while (0)
329
330#endif /* _KERNEL */
331
332#if defined(_WANT_MNTOPTNAMES) || defined(_KERNEL)
333struct mntoptnames {
334	uint64_t o_opt;
335	const char *o_name;
336};
337#define MNTOPT_NAMES							\
338	{ MNT_ASYNC,		"asynchronous" },			\
339	{ MNT_EXPORTED,		"NFS exported" },			\
340	{ MNT_LOCAL,		"local" },				\
341	{ MNT_NOATIME,		"noatime" },				\
342	{ MNT_NOEXEC,		"noexec" },				\
343	{ MNT_NOSUID,		"nosuid" },				\
344	{ MNT_NOSYMFOLLOW,	"nosymfollow" },			\
345	{ MNT_QUOTA,		"with quotas" },			\
346	{ MNT_RDONLY,		"read-only" },				\
347	{ MNT_SYNCHRONOUS,	"synchronous" },			\
348	{ MNT_UNION,		"union" },				\
349	{ MNT_NOCLUSTERR,	"noclusterr" },				\
350	{ MNT_NOCLUSTERW,	"noclusterw" },				\
351	{ MNT_SUIDDIR,		"suiddir" },				\
352	{ MNT_SOFTDEP,		"soft-updates" },			\
353	{ MNT_SUJ,		"journaled soft-updates" },		\
354	{ MNT_MULTILABEL,	"multilabel" },				\
355	{ MNT_ACLS,		"acls" },				\
356	{ MNT_NFS4ACLS,		"nfsv4acls" },				\
357	{ MNT_GJOURNAL,		"gjournal" },				\
358	{ MNT_AUTOMOUNTED,	"automounted" },			\
359	{ MNT_VERIFIED,		"verified" },				\
360	{ MNT_UNTRUSTED,	"untrusted" },				\
361	{ MNT_NOCOVER,		"nocover" },				\
362	{ MNT_EMPTYDIR,		"emptydir" },				\
363	{ MNT_UPDATE,		"update" },				\
364	{ MNT_DELEXPORT,	"delexport" },				\
365	{ MNT_RELOAD,		"reload" },				\
366	{ MNT_FORCE,		"force" },				\
367	{ MNT_SNAPSHOT,		"snapshot" },				\
368	{ 0, NULL }
369#endif
370
371/*
372 * User specifiable flags, stored in mnt_flag.
373 */
374#define	MNT_RDONLY	0x0000000000000001ULL /* read only filesystem */
375#define	MNT_SYNCHRONOUS	0x0000000000000002ULL /* fs written synchronously */
376#define	MNT_NOEXEC	0x0000000000000004ULL /* can't exec from filesystem */
377#define	MNT_NOSUID	0x0000000000000008ULL /* don't honor setuid fs bits */
378#define	MNT_NFS4ACLS	0x0000000000000010ULL /* enable NFS version 4 ACLs */
379#define	MNT_UNION	0x0000000000000020ULL /* union with underlying fs */
380#define	MNT_ASYNC	0x0000000000000040ULL /* fs written asynchronously */
381#define	MNT_SUIDDIR	0x0000000000100000ULL /* special SUID dir handling */
382#define	MNT_SOFTDEP	0x0000000000200000ULL /* using soft updates */
383#define	MNT_NOSYMFOLLOW	0x0000000000400000ULL /* do not follow symlinks */
384#define	MNT_GJOURNAL	0x0000000002000000ULL /* GEOM journal support enabled */
385#define	MNT_MULTILABEL	0x0000000004000000ULL /* MAC support for objects */
386#define	MNT_ACLS	0x0000000008000000ULL /* ACL support enabled */
387#define	MNT_NOATIME	0x0000000010000000ULL /* dont update file access time */
388#define	MNT_NOCLUSTERR	0x0000000040000000ULL /* disable cluster read */
389#define	MNT_NOCLUSTERW	0x0000000080000000ULL /* disable cluster write */
390#define	MNT_SUJ		0x0000000100000000ULL /* using journaled soft updates */
391#define	MNT_AUTOMOUNTED	0x0000000200000000ULL /* mounted by automountd(8) */
392#define	MNT_UNTRUSTED	0x0000000800000000ULL /* filesys metadata untrusted */
393
394/*
395 * NFS export related mount flags.
396 */
397#define	MNT_EXRDONLY	0x0000000000000080ULL	/* exported read only */
398#define	MNT_EXPORTED	0x0000000000000100ULL	/* filesystem is exported */
399#define	MNT_DEFEXPORTED	0x0000000000000200ULL	/* exported to the world */
400#define	MNT_EXPORTANON	0x0000000000000400ULL	/* anon uid mapping for all */
401#define	MNT_EXKERB	0x0000000000000800ULL	/* exported with Kerberos */
402#define	MNT_EXPUBLIC	0x0000000020000000ULL	/* public export (WebNFS) */
403#define	MNT_EXTLS	0x0000004000000000ULL /* require TLS */
404#define	MNT_EXTLSCERT	0x0000008000000000ULL /* require TLS with client cert */
405#define	MNT_EXTLSCERTUSER 0x0000010000000000ULL /* require TLS with user cert */
406
407/*
408 * Flags set by internal operations, but visible to the user.
409 */
410#define	MNT_LOCAL	0x0000000000001000ULL /* filesystem is stored locally */
411#define	MNT_QUOTA	0x0000000000002000ULL /* quotas are enabled on fs */
412#define	MNT_ROOTFS	0x0000000000004000ULL /* identifies the root fs */
413#define	MNT_USER	0x0000000000008000ULL /* mounted by a user */
414#define	MNT_IGNORE	0x0000000000800000ULL /* do not show entry in df */
415#define	MNT_VERIFIED	0x0000000400000000ULL /* filesystem is verified */
416
417/*
418 * Mask of flags that are visible to statfs().
419 * XXX I think that this could now become (~(MNT_CMDFLAGS))
420 * but the 'mount' program may need changing to handle this.
421 */
422#define	MNT_VISFLAGMASK	(MNT_RDONLY	| MNT_SYNCHRONOUS | MNT_NOEXEC	| \
423			MNT_NOSUID	| MNT_UNION	| MNT_SUJ	| \
424			MNT_ASYNC	| MNT_EXRDONLY	| MNT_EXPORTED	| \
425			MNT_DEFEXPORTED	| MNT_EXPORTANON| MNT_EXKERB	| \
426			MNT_LOCAL	| MNT_USER	| MNT_QUOTA	| \
427			MNT_ROOTFS	| MNT_NOATIME	| MNT_NOCLUSTERR| \
428			MNT_NOCLUSTERW	| MNT_SUIDDIR	| MNT_SOFTDEP	| \
429			MNT_IGNORE	| MNT_EXPUBLIC	| MNT_NOSYMFOLLOW | \
430			MNT_GJOURNAL	| MNT_MULTILABEL | MNT_ACLS	| \
431			MNT_NFS4ACLS	| MNT_AUTOMOUNTED | MNT_VERIFIED | \
432			MNT_UNTRUSTED)
433
434/* Mask of flags that can be updated. */
435#define	MNT_UPDATEMASK (MNT_NOSUID	| MNT_NOEXEC	| \
436			MNT_SYNCHRONOUS	| MNT_UNION	| MNT_ASYNC	| \
437			MNT_NOATIME | \
438			MNT_NOSYMFOLLOW	| MNT_IGNORE	| \
439			MNT_NOCLUSTERR	| MNT_NOCLUSTERW | MNT_SUIDDIR	| \
440			MNT_ACLS	| MNT_USER	| MNT_NFS4ACLS	| \
441			MNT_AUTOMOUNTED | MNT_UNTRUSTED)
442
443/*
444 * External filesystem command modifier flags.
445 * Unmount can use the MNT_FORCE flag.
446 * XXX: These are not STATES and really should be somewhere else.
447 * XXX: MNT_BYFSID and MNT_NONBUSY collide with MNT_ACLS and MNT_MULTILABEL,
448 *      but because MNT_ACLS and MNT_MULTILABEL are only used for mount(2),
449 *      and MNT_BYFSID and MNT_NONBUSY are only used for unmount(2),
450 *      it's harmless.
451 */
452#define	MNT_UPDATE	0x0000000000010000ULL /* not real mount, just update */
453#define	MNT_DELEXPORT	0x0000000000020000ULL /* delete export host lists */
454#define	MNT_RELOAD	0x0000000000040000ULL /* reload filesystem data */
455#define	MNT_FORCE	0x0000000000080000ULL /* force unmount or readonly */
456#define	MNT_SNAPSHOT	0x0000000001000000ULL /* snapshot the filesystem */
457#define	MNT_NONBUSY	0x0000000004000000ULL /* check vnode use counts. */
458#define	MNT_BYFSID	0x0000000008000000ULL /* specify filesystem by ID. */
459#define	MNT_NOCOVER	0x0000001000000000ULL /* Do not cover a mount point */
460#define	MNT_EMPTYDIR	0x0000002000000000ULL /* Only mount on empty dir */
461#define	MNT_RECURSE	0x0000100000000000ULL /* recursively unmount uppers */
462#define	MNT_DEFERRED    0x0000200000000000ULL /* unmount in async context */
463#define	MNT_CMDFLAGS   (MNT_UPDATE	| MNT_DELEXPORT	| MNT_RELOAD	| \
464			MNT_FORCE	| MNT_SNAPSHOT	| MNT_NONBUSY	| \
465			MNT_BYFSID	| MNT_NOCOVER	| MNT_EMPTYDIR	| \
466			MNT_RECURSE	| MNT_DEFERRED)
467
468/*
469 * Internal filesystem control flags stored in mnt_kern_flag.
470 *
471 * MNTK_UNMOUNT locks the mount entry so that name lookup cannot
472 * proceed past the mount point.  This keeps the subtree stable during
473 * mounts and unmounts.  When non-forced unmount flushes all vnodes
474 * from the mp queue, the MNTK_UNMOUNT flag prevents insmntque() from
475 * queueing new vnodes.
476 *
477 * MNTK_UNMOUNTF permits filesystems to detect a forced unmount while
478 * dounmount() is still waiting to lock the mountpoint. This allows
479 * the filesystem to cancel operations that might otherwise deadlock
480 * with the unmount attempt (used by NFS).
481 */
482#define MNTK_UNMOUNTF		0x00000001 /* forced unmount in progress */
483#define MNTK_ASYNC		0x00000002 /* filtered async flag */
484#define MNTK_SOFTDEP		0x00000004 /* async disabled by softdep */
485#define MNTK_NOMSYNC		0x00000008 /* don't do msync */
486#define	MNTK_DRAINING		0x00000010 /* lock draining is happening */
487#define	MNTK_REFEXPIRE		0x00000020 /* refcount expiring is happening */
488#define MNTK_EXTENDED_SHARED	0x00000040 /* Allow shared locking for more ops */
489#define	MNTK_SHARED_WRITES	0x00000080 /* Allow shared locking for writes */
490#define	MNTK_NO_IOPF		0x00000100 /* Disallow page faults during reads
491					      and writes. Filesystem shall
492					      properly handle i/o state on
493					      EFAULT. */
494#define	MNTK_RECURSE		0x00000200 /* pending recursive unmount */
495#define	MNTK_UPPER_WAITER	0x00000400 /* waiting to drain MNTK_UPPER_PENDING */
496/* UNUSED 			0x00000800 */
497#define	MNTK_UNLOCKED_INSMNTQUE	0x00001000 /* fs does not lock the vnode for
498					      insmntque */
499#define	MNTK_UNMAPPED_BUFS	0x00002000
500#define	MNTK_USES_BCACHE	0x00004000 /* FS uses the buffer cache. */
501/* UNUSED			0x00008000 */
502#define	MNTK_VMSETSIZE_BUG	0x00010000
503#define	MNTK_UNIONFS		0x00020000 /* A hack for F_ISUNIONSTACK */
504#define	MNTK_FPLOOKUP		0x00040000 /* fast path lookup is supported */
505#define	MNTK_SUSPEND_ALL	0x00080000 /* Suspended by all-fs suspension */
506#define	MNTK_TASKQUEUE_WAITER	0x00100000 /* Waiting on unmount taskqueue */
507/* UNUSED			0x00200000 */
508/* UNUSED			0x00400000 */
509#define	MNTK_NOASYNC		0x00800000 /* disable async */
510#define	MNTK_UNMOUNT		0x01000000 /* unmount in progress */
511#define	MNTK_MWAIT		0x02000000 /* waiting for unmount to finish */
512#define	MNTK_SUSPEND		0x08000000 /* request write suspension */
513#define	MNTK_SUSPEND2		0x04000000 /* block secondary writes */
514#define	MNTK_SUSPENDED		0x10000000 /* write operations are suspended */
515#define	MNTK_NULL_NOCACHE	0x20000000 /* auto disable cache for nullfs
516					      mounts over this fs */
517#define MNTK_LOOKUP_SHARED	0x40000000 /* FS supports shared lock lookups */
518/* UNUSED			0x80000000 */
519
520#ifdef _KERNEL
521static inline int
522MNT_SHARED_WRITES(struct mount *mp)
523{
524
525	return (mp != NULL && (mp->mnt_kern_flag & MNTK_SHARED_WRITES) != 0);
526}
527
528static inline int
529MNT_EXTENDED_SHARED(struct mount *mp)
530{
531
532	return (mp != NULL && (mp->mnt_kern_flag & MNTK_EXTENDED_SHARED) != 0);
533}
534#endif
535
536/*
537 * Sysctl CTL_VFS definitions.
538 *
539 * Second level identifier specifies which filesystem. Second level
540 * identifier VFS_VFSCONF returns information about all filesystems.
541 * Second level identifier VFS_GENERIC is non-terminal.
542 */
543#define	VFS_VFSCONF		0	/* get configured filesystems */
544#define	VFS_GENERIC		0	/* generic filesystem information */
545/*
546 * Third level identifiers for VFS_GENERIC are given below; third
547 * level identifiers for specific filesystems are given in their
548 * mount specific header files.
549 */
550#define VFS_MAXTYPENUM	1	/* int: highest defined filesystem type */
551#define VFS_CONF	2	/* struct: vfsconf for filesystem given
552				   as next argument */
553
554/*
555 * Flags for various system call interfaces.
556 *
557 * waitfor flags to vfs_sync() and getfsstat()
558 */
559#define MNT_WAIT	1	/* synchronously wait for I/O to complete */
560#define MNT_NOWAIT	2	/* start all I/O, but do not wait for it */
561#define MNT_LAZY	3	/* push data not written by filesystem syncer */
562#define MNT_SUSPEND	4	/* Suspend file system after sync */
563
564/*
565 * Generic file handle
566 */
567struct fhandle {
568	fsid_t	fh_fsid;	/* Filesystem id of mount point */
569	struct	fid fh_fid;	/* Filesys specific id */
570};
571typedef struct fhandle	fhandle_t;
572
573/*
574 * Old export arguments without security flavor list
575 */
576struct oexport_args {
577	int	ex_flags;		/* export related flags */
578	uid_t	ex_root;		/* mapping for root uid */
579	struct	xucred ex_anon;		/* mapping for anonymous user */
580	struct	sockaddr *ex_addr;	/* net address to which exported */
581	u_char	ex_addrlen;		/* and the net address length */
582	struct	sockaddr *ex_mask;	/* mask of valid bits in saddr */
583	u_char	ex_masklen;		/* and the smask length */
584	char	*ex_indexfile;		/* index file for WebNFS URLs */
585};
586
587/*
588 * Not quite so old export arguments with 32bit ex_flags and xucred ex_anon.
589 */
590#define	MAXSECFLAVORS	5
591struct o2export_args {
592	int	ex_flags;		/* export related flags */
593	uid_t	ex_root;		/* mapping for root uid */
594	struct	xucred ex_anon;		/* mapping for anonymous user */
595	struct	sockaddr *ex_addr;	/* net address to which exported */
596	u_char	ex_addrlen;		/* and the net address length */
597	struct	sockaddr *ex_mask;	/* mask of valid bits in saddr */
598	u_char	ex_masklen;		/* and the smask length */
599	char	*ex_indexfile;		/* index file for WebNFS URLs */
600	int	ex_numsecflavors;	/* security flavor count */
601	int	ex_secflavors[MAXSECFLAVORS]; /* list of security flavors */
602};
603
604/*
605 * Export arguments for local filesystem mount calls.
606 */
607struct export_args {
608	uint64_t ex_flags;		/* export related flags */
609	uid_t	ex_root;		/* mapping for root uid */
610	uid_t	ex_uid;			/* mapping for anonymous user */
611	int	ex_ngroups;
612	gid_t	*ex_groups;
613	struct	sockaddr *ex_addr;	/* net address to which exported */
614	u_char	ex_addrlen;		/* and the net address length */
615	struct	sockaddr *ex_mask;	/* mask of valid bits in saddr */
616	u_char	ex_masklen;		/* and the smask length */
617	char	*ex_indexfile;		/* index file for WebNFS URLs */
618	int	ex_numsecflavors;	/* security flavor count */
619	int	ex_secflavors[MAXSECFLAVORS]; /* list of security flavors */
620};
621
622/*
623 * Structure holding information for a publicly exported filesystem
624 * (WebNFS). Currently the specs allow just for one such filesystem.
625 */
626struct nfs_public {
627	int		np_valid;	/* Do we hold valid information */
628	fhandle_t	np_handle;	/* Filehandle for pub fs (internal) */
629	struct mount	*np_mount;	/* Mountpoint of exported fs */
630	char		*np_index;	/* Index file */
631};
632
633/*
634 * Filesystem configuration information. One of these exists for each
635 * type of filesystem supported by the kernel. These are searched at
636 * mount time to identify the requested filesystem.
637 *
638 * XXX: Never change the first two arguments!
639 */
640struct vfsconf {
641	u_int	vfc_version;		/* ABI version number */
642	char	vfc_name[MFSNAMELEN];	/* filesystem type name */
643	struct	vfsops *vfc_vfsops;	/* filesystem operations vector */
644	struct	vfsops *vfc_vfsops_sd;	/* ... signal-deferred */
645	int	vfc_typenum;		/* historic filesystem type number */
646	int	vfc_refcount;		/* number mounted of this type */
647	int	vfc_flags;		/* permanent flags */
648	int	vfc_prison_flag;	/* prison allow.mount.* flag */
649	struct	vfsoptdecl *vfc_opts;	/* mount options */
650	TAILQ_ENTRY(vfsconf) vfc_list;	/* list of vfscons */
651};
652
653/* Userland version of the struct vfsconf. */
654struct xvfsconf {
655	struct	vfsops *vfc_vfsops;	/* filesystem operations vector */
656	char	vfc_name[MFSNAMELEN];	/* filesystem type name */
657	int	vfc_typenum;		/* historic filesystem type number */
658	int	vfc_refcount;		/* number mounted of this type */
659	int	vfc_flags;		/* permanent flags */
660	struct	vfsconf *vfc_next;	/* next in list */
661};
662
663#ifndef BURN_BRIDGES
664struct ovfsconf {
665	void	*vfc_vfsops;
666	char	vfc_name[32];
667	int	vfc_index;
668	int	vfc_refcount;
669	int	vfc_flags;
670};
671#endif
672
673/*
674 * NB: these flags refer to IMPLEMENTATION properties, not properties of
675 * any actual mounts; i.e., it does not make sense to change the flags.
676 */
677#define	VFCF_STATIC	0x00010000	/* statically compiled into kernel */
678#define	VFCF_NETWORK	0x00020000	/* may get data over the network */
679#define	VFCF_READONLY	0x00040000	/* writes are not implemented */
680#define	VFCF_SYNTHETIC	0x00080000	/* data does not represent real files */
681#define	VFCF_LOOPBACK	0x00100000	/* aliases some other mounted FS */
682#define	VFCF_UNICODE	0x00200000	/* stores file names as Unicode */
683#define	VFCF_JAIL	0x00400000	/* can be mounted from within a jail */
684#define	VFCF_DELEGADMIN	0x00800000	/* supports delegated administration */
685#define	VFCF_SBDRY	0x01000000	/* Stop at Boundary: defer stop requests
686					   to kernel->user (AST) transition */
687#define	VFCF_FILEMOUNT	0x02000000	/* allow mounting files */
688
689typedef uint32_t fsctlop_t;
690
691struct vfsidctl {
692	int		vc_vers;	/* should be VFSIDCTL_VERS1 (below) */
693	fsid_t		vc_fsid;	/* fsid to operate on */
694	char		vc_fstypename[MFSNAMELEN];
695					/* type of fs 'nfs' or '*' */
696	fsctlop_t	vc_op;		/* operation VFS_CTL_* (below) */
697	void		*vc_ptr;	/* pointer to data structure */
698	size_t		vc_len;		/* sizeof said structure */
699	u_int32_t	vc_spare[12];	/* spare (must be zero) */
700};
701
702/* vfsidctl API version. */
703#define VFS_CTL_VERS1	0x01
704
705/*
706 * New style VFS sysctls, do not reuse/conflict with the namespace for
707 * private sysctls.
708 * All "global" sysctl ops have the 33rd bit set:
709 * 0x...1....
710 * Private sysctl ops should have the 33rd bit unset.
711 */
712#define VFS_CTL_QUERY	0x00010001	/* anything wrong? (vfsquery) */
713#define VFS_CTL_TIMEO	0x00010002	/* set timeout for vfs notification */
714#define VFS_CTL_NOLOCKS	0x00010003	/* disable file locking */
715
716struct vfsquery {
717	u_int32_t	vq_flags;
718	u_int32_t	vq_spare[31];
719};
720
721/* vfsquery flags */
722#define VQ_NOTRESP	0x0001	/* server down */
723#define VQ_NEEDAUTH	0x0002	/* server bad auth */
724#define VQ_LOWDISK	0x0004	/* we're low on space */
725#define VQ_MOUNT	0x0008	/* new filesystem arrived */
726#define VQ_UNMOUNT	0x0010	/* filesystem has left */
727#define VQ_DEAD		0x0020	/* filesystem is dead, needs force unmount */
728#define VQ_ASSIST	0x0040	/* filesystem needs assistance from external
729				   program */
730#define VQ_NOTRESPLOCK	0x0080	/* server lockd down */
731#define VQ_FLAG0100	0x0100	/* placeholder */
732#define VQ_FLAG0200	0x0200	/* placeholder */
733#define VQ_FLAG0400	0x0400	/* placeholder */
734#define VQ_FLAG0800	0x0800	/* placeholder */
735#define VQ_FLAG1000	0x1000	/* placeholder */
736#define VQ_FLAG2000	0x2000	/* placeholder */
737#define VQ_FLAG4000	0x4000	/* placeholder */
738#define VQ_FLAG8000	0x8000	/* placeholder */
739
740#ifdef _KERNEL
741/* Point a sysctl request at a vfsidctl's data. */
742#define VCTLTOREQ(vc, req)						\
743	do {								\
744		(req)->newptr = (vc)->vc_ptr;				\
745		(req)->newlen = (vc)->vc_len;				\
746		(req)->newidx = 0;					\
747	} while (0)
748#endif
749
750struct iovec;
751struct uio;
752
753#ifdef _KERNEL
754
755/*
756 * vfs_busy specific flags and mask.
757 */
758#define	MBF_NOWAIT	0x01
759#define	MBF_MNTLSTLOCK	0x02
760#define	MBF_MASK	(MBF_NOWAIT | MBF_MNTLSTLOCK)
761
762#ifdef MALLOC_DECLARE
763MALLOC_DECLARE(M_MOUNT);
764MALLOC_DECLARE(M_STATFS);
765#endif
766extern int maxvfsconf;		/* highest defined filesystem type */
767
768TAILQ_HEAD(vfsconfhead, vfsconf);
769extern struct vfsconfhead vfsconf;
770
771/*
772 * Operations supported on mounted filesystem.
773 */
774struct mount_args;
775struct nameidata;
776struct sysctl_req;
777struct mntarg;
778
779/*
780 * N.B., vfs_cmount is the ancient vfsop invoked by the old mount(2) syscall.
781 * The new way is vfs_mount.
782 *
783 * vfs_cmount implementations typically translate arguments from their
784 * respective old per-FS structures into the key-value list supported by
785 * nmount(2), then use kernel_mount(9) to mimic nmount(2) from kernelspace.
786 *
787 * Filesystems with mounters that use nmount(2) do not need to and should not
788 * implement vfs_cmount.  Hopefully a future cleanup can remove vfs_cmount and
789 * mount(2) entirely.
790 */
791typedef int vfs_cmount_t(struct mntarg *ma, void *data, uint64_t flags);
792typedef int vfs_unmount_t(struct mount *mp, int mntflags);
793typedef int vfs_root_t(struct mount *mp, int flags, struct vnode **vpp);
794typedef	int vfs_quotactl_t(struct mount *mp, int cmds, uid_t uid, void *arg,
795		    bool *mp_busy);
796typedef	int vfs_statfs_t(struct mount *mp, struct statfs *sbp);
797typedef	int vfs_sync_t(struct mount *mp, int waitfor);
798typedef	int vfs_vget_t(struct mount *mp, ino_t ino, int flags,
799		    struct vnode **vpp);
800typedef	int vfs_fhtovp_t(struct mount *mp, struct fid *fhp,
801		    int flags, struct vnode **vpp);
802typedef	int vfs_checkexp_t(struct mount *mp, struct sockaddr *nam,
803		    uint64_t *extflagsp, struct ucred **credanonp,
804		    int *numsecflavors, int *secflavors);
805typedef	int vfs_init_t(struct vfsconf *);
806typedef	int vfs_uninit_t(struct vfsconf *);
807typedef	int vfs_extattrctl_t(struct mount *mp, int cmd,
808		    struct vnode *filename_vp, int attrnamespace,
809		    const char *attrname);
810typedef	int vfs_mount_t(struct mount *mp);
811typedef int vfs_sysctl_t(struct mount *mp, fsctlop_t op,
812		    struct sysctl_req *req);
813typedef void vfs_susp_clean_t(struct mount *mp);
814typedef void vfs_notify_lowervp_t(struct mount *mp, struct vnode *lowervp);
815typedef void vfs_purge_t(struct mount *mp);
816struct sbuf;
817typedef int vfs_report_lockf_t(struct mount *mp, struct sbuf *sb);
818
819struct vfsops {
820	vfs_mount_t		*vfs_mount;
821	vfs_cmount_t		*vfs_cmount;
822	vfs_unmount_t		*vfs_unmount;
823	vfs_root_t		*vfs_root;
824	vfs_root_t		*vfs_cachedroot;
825	vfs_quotactl_t		*vfs_quotactl;
826	vfs_statfs_t		*vfs_statfs;
827	vfs_sync_t		*vfs_sync;
828	vfs_vget_t		*vfs_vget;
829	vfs_fhtovp_t		*vfs_fhtovp;
830	vfs_checkexp_t		*vfs_checkexp;
831	vfs_init_t		*vfs_init;
832	vfs_uninit_t		*vfs_uninit;
833	vfs_extattrctl_t	*vfs_extattrctl;
834	vfs_sysctl_t		*vfs_sysctl;
835	vfs_susp_clean_t	*vfs_susp_clean;
836	vfs_notify_lowervp_t	*vfs_reclaim_lowervp;
837	vfs_notify_lowervp_t	*vfs_unlink_lowervp;
838	vfs_purge_t		*vfs_purge;
839	vfs_report_lockf_t	*vfs_report_lockf;
840	vfs_mount_t		*vfs_spare[6];	/* spares for ABI compat */
841};
842
843vfs_statfs_t	__vfs_statfs;
844
845#define	VFS_MOUNT(MP) ({						\
846	int _rc;							\
847									\
848	TSRAW(curthread, TS_ENTER, "VFS_MOUNT", (MP)->mnt_vfc->vfc_name);\
849	_rc = (*(MP)->mnt_op->vfs_mount)(MP);				\
850	TSRAW(curthread, TS_EXIT, "VFS_MOUNT", (MP)->mnt_vfc->vfc_name);\
851	_rc; })
852
853#define	VFS_UNMOUNT(MP, FORCE) ({					\
854	int _rc;							\
855									\
856	_rc = (*(MP)->mnt_op->vfs_unmount)(MP, FORCE);			\
857	_rc; })
858
859#define	VFS_ROOT(MP, FLAGS, VPP) ({					\
860	int _rc;							\
861									\
862	_rc = (*(MP)->mnt_op->vfs_root)(MP, FLAGS, VPP);		\
863	_rc; })
864
865#define	VFS_CACHEDROOT(MP, FLAGS, VPP) ({				\
866	int _rc;							\
867									\
868	_rc = (*(MP)->mnt_op->vfs_cachedroot)(MP, FLAGS, VPP);		\
869	_rc; })
870
871#define	VFS_QUOTACTL(MP, C, U, A, MP_BUSY) ({				\
872	int _rc;							\
873									\
874	_rc = (*(MP)->mnt_op->vfs_quotactl)(MP, C, U, A, MP_BUSY);	\
875	_rc; })
876
877#define	VFS_STATFS(MP, SBP) ({						\
878	int _rc;							\
879									\
880	_rc = __vfs_statfs((MP), (SBP));				\
881	_rc; })
882
883#define	VFS_SYNC(MP, WAIT) ({						\
884	int _rc;							\
885									\
886	_rc = (*(MP)->mnt_op->vfs_sync)(MP, WAIT);			\
887	_rc; })
888
889#define	VFS_VGET(MP, INO, FLAGS, VPP) ({				\
890	int _rc;							\
891									\
892	_rc = (*(MP)->mnt_op->vfs_vget)(MP, INO, FLAGS, VPP);		\
893	_rc; })
894
895#define	VFS_FHTOVP(MP, FIDP, FLAGS, VPP) ({				\
896	int _rc;							\
897									\
898	_rc = (*(MP)->mnt_op->vfs_fhtovp)(MP, FIDP, FLAGS, VPP);	\
899	_rc; })
900
901#define	VFS_CHECKEXP(MP, NAM, EXFLG, CRED, NUMSEC, SEC) ({		\
902	int _rc;							\
903									\
904	_rc = (*(MP)->mnt_op->vfs_checkexp)(MP, NAM, EXFLG, CRED, NUMSEC,\
905	    SEC);							\
906	_rc; })
907
908#define	VFS_EXTATTRCTL(MP, C, FN, NS, N) ({				\
909	int _rc;							\
910									\
911	_rc = (*(MP)->mnt_op->vfs_extattrctl)(MP, C, FN, NS, N);	\
912	_rc; })
913
914#define	VFS_SYSCTL(MP, OP, REQ) ({					\
915	int _rc;							\
916									\
917	_rc = (*(MP)->mnt_op->vfs_sysctl)(MP, OP, REQ);			\
918	_rc; })
919
920#define	VFS_SUSP_CLEAN(MP) do {						\
921	if (*(MP)->mnt_op->vfs_susp_clean != NULL) {			\
922		(*(MP)->mnt_op->vfs_susp_clean)(MP);			\
923	}								\
924} while (0)
925
926#define	VFS_RECLAIM_LOWERVP(MP, VP) do {				\
927	if (*(MP)->mnt_op->vfs_reclaim_lowervp != NULL) {		\
928		(*(MP)->mnt_op->vfs_reclaim_lowervp)((MP), (VP));	\
929	}								\
930} while (0)
931
932#define	VFS_UNLINK_LOWERVP(MP, VP) do {					\
933	if (*(MP)->mnt_op->vfs_unlink_lowervp != NULL) {		\
934		(*(MP)->mnt_op->vfs_unlink_lowervp)((MP), (VP));	\
935	}								\
936} while (0)
937
938#define	VFS_PURGE(MP) do {						\
939	if (*(MP)->mnt_op->vfs_purge != NULL) {				\
940		(*(MP)->mnt_op->vfs_purge)(MP);				\
941	}								\
942} while (0)
943
944#define VFS_KNOTE_LOCKED(vp, hint) do					\
945{									\
946	VN_KNOTE((vp), (hint), KNF_LISTLOCKED);				\
947} while (0)
948
949#define VFS_KNOTE_UNLOCKED(vp, hint) do					\
950{									\
951	VN_KNOTE((vp), (hint), 0);					\
952} while (0)
953
954#include <sys/module.h>
955
956/*
957 * Version numbers.
958 */
959#define VFS_VERSION_00	0x19660120
960#define VFS_VERSION_01	0x20121030
961#define VFS_VERSION_02	0x20180504
962#define VFS_VERSION	VFS_VERSION_02
963
964#define VFS_SET(vfsops, fsname, flags) \
965	static struct vfsconf fsname ## _vfsconf = {		\
966		.vfc_version = VFS_VERSION,			\
967		.vfc_name = #fsname,				\
968		.vfc_vfsops = &vfsops,				\
969		.vfc_typenum = -1,				\
970		.vfc_flags = flags,				\
971	};							\
972	static moduledata_t fsname ## _mod = {			\
973		#fsname,					\
974		vfs_modevent,					\
975		& fsname ## _vfsconf				\
976	};							\
977	DECLARE_MODULE(fsname, fsname ## _mod, SI_SUB_VFS, SI_ORDER_MIDDLE)
978
979enum vfs_notify_upper_type {
980	VFS_NOTIFY_UPPER_RECLAIM,
981	VFS_NOTIFY_UPPER_UNLINK,
982};
983
984/*
985 * exported vnode operations
986 */
987
988/* Define this to indicate that vfs_exjail_clone() exists for ZFS to use. */
989#define	VFS_SUPPORTS_EXJAIL_CLONE	1
990
991int	dounmount(struct mount *, uint64_t, struct thread *);
992
993int	kernel_mount(struct mntarg *ma, uint64_t flags);
994struct mntarg *mount_arg(struct mntarg *ma, const char *name, const void *val, int len);
995struct mntarg *mount_argb(struct mntarg *ma, int flag, const char *name);
996struct mntarg *mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...);
997struct mntarg *mount_argsu(struct mntarg *ma, const char *name, const void *val, int len);
998void	statfs_scale_blocks(struct statfs *sf, long max_size);
999struct vfsconf *vfs_byname(const char *);
1000struct vfsconf *vfs_byname_kld(const char *, struct thread *td, int *);
1001void	vfs_mount_destroy(struct mount *);
1002void	vfs_event_signal(fsid_t *, u_int32_t, intptr_t);
1003void	vfs_freeopts(struct vfsoptlist *opts);
1004void	vfs_deleteopt(struct vfsoptlist *opts, const char *name);
1005int	vfs_buildopts(struct uio *auio, struct vfsoptlist **options);
1006int	vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
1007	    uint64_t val);
1008int	vfs_getopt(struct vfsoptlist *, const char *, void **, int *);
1009int	vfs_getopt_pos(struct vfsoptlist *opts, const char *name);
1010int	vfs_getopt_size(struct vfsoptlist *opts, const char *name,
1011	    off_t *value);
1012char	*vfs_getopts(struct vfsoptlist *, const char *, int *error);
1013int	vfs_copyopt(struct vfsoptlist *, const char *, void *, int);
1014int	vfs_filteropt(struct vfsoptlist *, const char **legal);
1015void	vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...);
1016int	vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...);
1017int	vfs_setopt(struct vfsoptlist *opts, const char *name, void *value,
1018	    int len);
1019int	vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value,
1020	    int len);
1021int	vfs_setopts(struct vfsoptlist *opts, const char *name,
1022	    const char *value);
1023int	vfs_setpublicfs			    /* set publicly exported fs */
1024	    (struct mount *, struct netexport *, struct export_args *);
1025void	vfs_periodic(struct mount *, int);
1026int	vfs_busy(struct mount *, int);
1027void	vfs_exjail_clone(struct mount *, struct mount *);
1028void	vfs_exjail_delete(struct prison *);
1029int	vfs_export			 /* process mount export info */
1030	    (struct mount *, struct export_args *, bool);
1031void	vfs_free_addrlist(struct netexport *);
1032void	vfs_allocate_syncvnode(struct mount *);
1033void	vfs_deallocate_syncvnode(struct mount *);
1034int	vfs_donmount(struct thread *td, uint64_t fsflags,
1035	    struct uio *fsoptions);
1036void	vfs_getnewfsid(struct mount *);
1037struct	mount *vfs_getvfs(fsid_t *);      /* return vfs given fsid */
1038struct	mount *vfs_busyfs(fsid_t *);
1039int	vfs_modevent(module_t, int, void *);
1040void	vfs_mount_error(struct mount *, const char *, ...);
1041void	vfs_mountroot(void);			/* mount our root filesystem */
1042void	vfs_mountedfrom(struct mount *, const char *from);
1043void	vfs_notify_upper(struct vnode *, enum vfs_notify_upper_type);
1044struct mount *vfs_ref_from_vp(struct vnode *);
1045void	vfs_ref(struct mount *);
1046void	vfs_rel(struct mount *);
1047struct mount *vfs_mount_alloc(struct vnode *, struct vfsconf *, const char *,
1048	    struct ucred *);
1049int	vfs_suser(struct mount *, struct thread *);
1050void	vfs_unbusy(struct mount *);
1051void	vfs_unmountall(void);
1052struct mount *vfs_register_upper_from_vp(struct vnode *,
1053	    struct mount *ump, struct mount_upper_node *);
1054void	vfs_register_for_notification(struct mount *, struct mount *,
1055	    struct mount_upper_node *);
1056void	vfs_unregister_for_notification(struct mount *,
1057	    struct mount_upper_node *);
1058void	vfs_unregister_upper(struct mount *, struct mount_upper_node *);
1059int	vfs_remount_ro(struct mount *mp);
1060int	vfs_report_lockf(struct mount *mp, struct sbuf *sb);
1061
1062extern	TAILQ_HEAD(mntlist, mount) mountlist;	/* mounted filesystem list */
1063extern	struct mtx_padalign mountlist_mtx;
1064extern	struct nfs_public nfs_pub;
1065extern	struct sx vfsconf_sx;
1066#define	vfsconf_lock()		sx_xlock(&vfsconf_sx)
1067#define	vfsconf_unlock()	sx_xunlock(&vfsconf_sx)
1068#define	vfsconf_slock()		sx_slock(&vfsconf_sx)
1069#define	vfsconf_sunlock()	sx_sunlock(&vfsconf_sx)
1070struct vnode *mntfs_allocvp(struct mount *, struct vnode *);
1071void   mntfs_freevp(struct vnode *);
1072
1073/*
1074 * Declarations for these vfs default operations are located in
1075 * kern/vfs_default.c.  They will be automatically used to replace
1076 * null entries in VFS ops tables when registering a new filesystem
1077 * type in the global table.
1078 */
1079vfs_root_t		vfs_stdroot;
1080vfs_quotactl_t		vfs_stdquotactl;
1081vfs_statfs_t		vfs_stdstatfs;
1082vfs_sync_t		vfs_stdsync;
1083vfs_sync_t		vfs_stdnosync;
1084vfs_vget_t		vfs_stdvget;
1085vfs_fhtovp_t		vfs_stdfhtovp;
1086vfs_checkexp_t		vfs_stdcheckexp;
1087vfs_init_t		vfs_stdinit;
1088vfs_uninit_t		vfs_stduninit;
1089vfs_extattrctl_t	vfs_stdextattrctl;
1090vfs_sysctl_t		vfs_stdsysctl;
1091
1092void	syncer_suspend(void);
1093void	syncer_resume(void);
1094
1095struct vnode *vfs_cache_root_clear(struct mount *);
1096void	vfs_cache_root_set(struct mount *, struct vnode *);
1097
1098void	vfs_op_barrier_wait(struct mount *);
1099void	vfs_op_enter(struct mount *);
1100void	vfs_op_exit_locked(struct mount *);
1101void	vfs_op_exit(struct mount *);
1102
1103#ifdef DIAGNOSTIC
1104void	vfs_assert_mount_counters(struct mount *);
1105void	vfs_dump_mount_counters(struct mount *);
1106#else
1107#define vfs_assert_mount_counters(mp) do { } while (0)
1108#define vfs_dump_mount_counters(mp) do { } while (0)
1109#endif
1110
1111enum mount_counter { MNT_COUNT_REF, MNT_COUNT_LOCKREF, MNT_COUNT_WRITEOPCOUNT };
1112int	vfs_mount_fetch_counter(struct mount *, enum mount_counter);
1113
1114void suspend_all_fs(void);
1115void resume_all_fs(void);
1116
1117/*
1118 * Code transitioning mnt_vfs_ops to > 0 issues IPIs until it observes
1119 * all CPUs not executing code enclosed by thread_in_ops_pcpu variable.
1120 *
1121 * This provides an invariant that by the time the last CPU is observed not
1122 * executing, everyone else entering will see the counter > 0 and exit.
1123 *
1124 * Note there is no barrier between vfs_ops and the rest of the code in the
1125 * section. It is not necessary as the writer has to wait for everyone to drain
1126 * before making any changes or only make changes safe while the section is
1127 * executed.
1128 */
1129#define	vfs_mount_pcpu(mp)		zpcpu_get(mp->mnt_pcpu)
1130#define	vfs_mount_pcpu_remote(mp, cpu)	zpcpu_get_cpu(mp->mnt_pcpu, cpu)
1131
1132#define vfs_op_thread_entered(mp) ({				\
1133	MPASS(curthread->td_critnest > 0);			\
1134	struct mount_pcpu *_mpcpu = vfs_mount_pcpu(mp);		\
1135	_mpcpu->mntp_thread_in_ops == 1;			\
1136})
1137
1138#define vfs_op_thread_enter_crit(mp, _mpcpu) ({			\
1139	bool _retval_crit = true;				\
1140	MPASS(curthread->td_critnest > 0);			\
1141	_mpcpu = vfs_mount_pcpu(mp);				\
1142	MPASS(mpcpu->mntp_thread_in_ops == 0);			\
1143	_mpcpu->mntp_thread_in_ops = 1;				\
1144	atomic_interrupt_fence();					\
1145	if (__predict_false(mp->mnt_vfs_ops > 0)) {		\
1146		vfs_op_thread_exit_crit(mp, _mpcpu);		\
1147		_retval_crit = false;				\
1148	}							\
1149	_retval_crit;						\
1150})
1151
1152#define vfs_op_thread_enter(mp, _mpcpu) ({			\
1153	bool _retval;						\
1154	critical_enter();					\
1155	_retval = vfs_op_thread_enter_crit(mp, _mpcpu);		\
1156	if (__predict_false(!_retval))				\
1157		critical_exit();				\
1158	_retval;						\
1159})
1160
1161#define vfs_op_thread_exit_crit(mp, _mpcpu) do {		\
1162	MPASS(_mpcpu == vfs_mount_pcpu(mp));			\
1163	MPASS(_mpcpu->mntp_thread_in_ops == 1);			\
1164	atomic_interrupt_fence();					\
1165	_mpcpu->mntp_thread_in_ops = 0;				\
1166} while (0)
1167
1168#define vfs_op_thread_exit(mp, _mpcpu) do {			\
1169	vfs_op_thread_exit_crit(mp, _mpcpu);			\
1170	critical_exit();					\
1171} while (0)
1172
1173#define vfs_mp_count_add_pcpu(_mpcpu, count, val) do {		\
1174	MPASS(_mpcpu->mntp_thread_in_ops == 1);			\
1175	_mpcpu->mntp_##count += val;				\
1176} while (0)
1177
1178#define vfs_mp_count_sub_pcpu(_mpcpu, count, val) do {		\
1179	MPASS(_mpcpu->mntp_thread_in_ops == 1);			\
1180	_mpcpu->mntp_##count -= val;				\
1181} while (0)
1182
1183#else /* !_KERNEL */
1184
1185#include <sys/cdefs.h>
1186
1187struct stat;
1188
1189__BEGIN_DECLS
1190int	fhlink(struct fhandle *, const char *);
1191int	fhlinkat(struct fhandle *, int, const char *);
1192int	fhopen(const struct fhandle *, int);
1193int	fhreadlink(struct fhandle *, char *, size_t);
1194int	fhstat(const struct fhandle *, struct stat *);
1195int	fhstatfs(const struct fhandle *, struct statfs *);
1196int	fstatfs(int, struct statfs *);
1197int	getfh(const char *, fhandle_t *);
1198int	getfhat(int, char *, struct fhandle *, int);
1199int	getfsstat(struct statfs *, long, int);
1200int	getmntinfo(struct statfs **, int);
1201int	lgetfh(const char *, fhandle_t *);
1202int	mount(const char *, const char *, int, void *);
1203int	nmount(struct iovec *, unsigned int, int);
1204int	statfs(const char *, struct statfs *);
1205int	unmount(const char *, int);
1206
1207/* C library stuff */
1208int	getvfsbyname(const char *, struct xvfsconf *);
1209__END_DECLS
1210
1211#endif /* _KERNEL */
1212
1213#endif /* !_SYS_MOUNT_H_ */
1214