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 * @(#)mount.h 8.21 (Berkeley) 5/20/95
32 */
33
34 #ifndef _SYS_MOUNT_H_
35 #define _SYS_MOUNT_H_
36
37 #include <sys/ucred.h>
38 #include <sys/queue.h>
39 #ifdef _KERNEL
40 #include <sys/lock.h>
41 #include <sys/lockmgr.h>
42 #include <sys/tslog.h>
43 #include <sys/_mutex.h>
44 #include <sys/_sx.h>
45 #endif
46
47 /*
48 * NOTE: When changing statfs structure, mount structure, MNT_* flags or
49 * MNTK_* flags also update DDB show mount command in vfs_subr.c.
50 */
51
52 typedef struct fsid { int32_t val[2]; } fsid_t; /* filesystem id type */
53
54 /* Returns non-zero if fsids are different. */
55 static __inline int
fsidcmp(const fsid_t * a,const fsid_t * b)56 fsidcmp(const fsid_t *a, const fsid_t *b)
57 {
58 return (a->val[0] != b->val[0] || a->val[1] != b->val[1]);
59 }
60
61 /*
62 * File identifier.
63 * These are unique per filesystem on a single machine.
64 *
65 * Note that the offset of fid_data is 4 bytes, so care must be taken to avoid
66 * undefined behavior accessing unaligned fields within an embedded struct.
67 */
68 #define MAXFIDSZ 16
69
70 struct fid {
71 u_short fid_len; /* length of data in bytes */
72 u_short fid_data0; /* force longword alignment */
73 char fid_data[MAXFIDSZ]; /* data (variable length) */
74 };
75
76 /*
77 * filesystem statistics
78 */
79 #define MFSNAMELEN 16 /* length of type name including null */
80 #define MNAMELEN 1024 /* size of on/from name bufs */
81 #define STATFS_VERSION 0x20140518 /* current version number */
82 struct statfs {
83 uint32_t f_version; /* structure version number */
84 uint32_t f_type; /* type of filesystem */
85 uint64_t f_flags; /* copy of mount exported flags */
86 uint64_t f_bsize; /* filesystem fragment size */
87 uint64_t f_iosize; /* optimal transfer block size */
88 uint64_t f_blocks; /* total data blocks in filesystem */
89 uint64_t f_bfree; /* free blocks in filesystem */
90 int64_t f_bavail; /* free blocks avail to non-superuser */
91 uint64_t f_files; /* total file nodes in filesystem */
92 int64_t f_ffree; /* free nodes avail to non-superuser */
93 uint64_t f_syncwrites; /* count of sync writes since mount */
94 uint64_t f_asyncwrites; /* count of async writes since mount */
95 uint64_t f_syncreads; /* count of sync reads since mount */
96 uint64_t f_asyncreads; /* count of async reads since mount */
97 uint64_t f_spare[10]; /* unused spare */
98 uint32_t f_namemax; /* maximum filename length */
99 uid_t f_owner; /* user that mounted the filesystem */
100 fsid_t f_fsid; /* filesystem id */
101 char f_charspare[80]; /* spare string space */
102 char f_fstypename[MFSNAMELEN]; /* filesystem type name */
103 char f_mntfromname[MNAMELEN]; /* mounted filesystem */
104 char f_mntonname[MNAMELEN]; /* directory on which mounted */
105 };
106
107 #if defined(_WANT_FREEBSD11_STATFS) || defined(_KERNEL)
108 #define FREEBSD11_STATFS_VERSION 0x20030518 /* current version number */
109 struct freebsd11_statfs {
110 uint32_t f_version; /* structure version number */
111 uint32_t f_type; /* type of filesystem */
112 uint64_t f_flags; /* copy of mount exported flags */
113 uint64_t f_bsize; /* filesystem fragment size */
114 uint64_t f_iosize; /* optimal transfer block size */
115 uint64_t f_blocks; /* total data blocks in filesystem */
116 uint64_t f_bfree; /* free blocks in filesystem */
117 int64_t f_bavail; /* free blocks avail to non-superuser */
118 uint64_t f_files; /* total file nodes in filesystem */
119 int64_t f_ffree; /* free nodes avail to non-superuser */
120 uint64_t f_syncwrites; /* count of sync writes since mount */
121 uint64_t f_asyncwrites; /* count of async writes since mount */
122 uint64_t f_syncreads; /* count of sync reads since mount */
123 uint64_t f_asyncreads; /* count of async reads since mount */
124 uint64_t f_spare[10]; /* unused spare */
125 uint32_t f_namemax; /* maximum filename length */
126 uid_t f_owner; /* user that mounted the filesystem */
127 fsid_t f_fsid; /* filesystem id */
128 char f_charspare[80]; /* spare string space */
129 char f_fstypename[16]; /* filesystem type name */
130 char f_mntfromname[88]; /* mounted filesystem */
131 char f_mntonname[88]; /* directory on which mounted */
132 };
133 #endif /* _WANT_FREEBSD11_STATFS || _KERNEL */
134
135 #ifdef _KERNEL
136 #define OMFSNAMELEN 16 /* length of fs type name, including null */
137 #define OMNAMELEN (88 - 2 * sizeof(long)) /* size of on/from name bufs */
138
139 /* XXX getfsstat.2 is out of date with write and read counter changes here. */
140 /* XXX statfs.2 is out of date with read counter changes here. */
141 struct ostatfs {
142 long f_spare2; /* placeholder */
143 long f_bsize; /* fundamental filesystem block size */
144 long f_iosize; /* optimal transfer block size */
145 long f_blocks; /* total data blocks in filesystem */
146 long f_bfree; /* free blocks in fs */
147 long f_bavail; /* free blocks avail to non-superuser */
148 long f_files; /* total file nodes in filesystem */
149 long f_ffree; /* free file nodes in fs */
150 fsid_t f_fsid; /* filesystem id */
151 uid_t f_owner; /* user that mounted the filesystem */
152 int f_type; /* type of filesystem */
153 int f_flags; /* copy of mount exported flags */
154 long f_syncwrites; /* count of sync writes since mount */
155 long f_asyncwrites; /* count of async writes since mount */
156 char f_fstypename[OMFSNAMELEN]; /* fs type name */
157 char f_mntonname[OMNAMELEN]; /* directory on which mounted */
158 long f_syncreads; /* count of sync reads since mount */
159 long f_asyncreads; /* count of async reads since mount */
160 short f_spares1; /* unused spare */
161 char f_mntfromname[OMNAMELEN];/* mounted filesystem */
162 short f_spares2; /* unused spare */
163 /*
164 * XXX on machines where longs are aligned to 8-byte boundaries, there
165 * is an unnamed int32_t here. This spare was after the apparent end
166 * of the struct until we bit off the read counters from f_mntonname.
167 */
168 long f_spare[2]; /* unused spare */
169 };
170
171 TAILQ_HEAD(vnodelst, vnode);
172
173 /* Mount options list */
174 TAILQ_HEAD(vfsoptlist, vfsopt);
175 struct vfsopt {
176 TAILQ_ENTRY(vfsopt) link;
177 char *name;
178 void *value;
179 int len;
180 int pos;
181 int seen;
182 };
183
184 struct mount_pcpu {
185 int mntp_thread_in_ops;
186 int mntp_ref;
187 int mntp_lockref;
188 int mntp_writeopcount;
189 };
190
191 _Static_assert(sizeof(struct mount_pcpu) == 16,
192 "the struct is allocated from pcpu 16 zone");
193
194 /*
195 * Structure per mounted filesystem. Each mounted filesystem has an
196 * array of operations and an instance record. The filesystems are
197 * put on a doubly linked list.
198 *
199 * Lock reference:
200 * l - mnt_listmtx
201 * m - mountlist_mtx
202 * i - interlock
203 * i* - interlock of uppers' list head
204 * v - vnode freelist mutex
205 *
206 * Unmarked fields are considered stable as long as a ref is held.
207 *
208 */
209 struct mount {
210 int mnt_vfs_ops; /* (i) pending vfs ops */
211 int mnt_kern_flag; /* (i) kernel only flags */
212 uint64_t mnt_flag; /* (i) flags shared with user */
213 struct mount_pcpu *mnt_pcpu; /* per-CPU data */
214 struct vnode *mnt_rootvnode;
215 struct vnode *mnt_vnodecovered; /* vnode we mounted on */
216 struct vfsops *mnt_op; /* operations on fs */
217 struct vfsconf *mnt_vfc; /* configuration info */
218 struct mtx __aligned(CACHE_LINE_SIZE) mnt_mtx; /* mount structure interlock */
219 int mnt_gen; /* struct mount generation */
220 #define mnt_startzero mnt_list
221 TAILQ_ENTRY(mount) mnt_list; /* (m) mount list */
222 struct vnode *mnt_syncer; /* syncer vnode */
223 int mnt_ref; /* (i) Reference count */
224 struct vnodelst mnt_nvnodelist; /* (i) list of vnodes */
225 int mnt_nvnodelistsize; /* (i) # of vnodes */
226 int mnt_writeopcount; /* (i) write syscalls pending */
227 struct vfsoptlist *mnt_opt; /* current mount options */
228 struct vfsoptlist *mnt_optnew; /* new options passed to fs */
229 struct ucred *mnt_exjail; /* (i) jail which did exports */
230 struct statfs mnt_stat; /* cache of filesystem stats */
231 struct ucred *mnt_cred; /* credentials of mounter */
232 void * mnt_data; /* private data */
233 time_t mnt_time; /* last time written*/
234 int mnt_iosize_max; /* max size for clusters, etc */
235 struct netexport *mnt_export; /* export list */
236 struct label *mnt_label; /* MAC label for the fs */
237 u_int mnt_hashseed; /* Random seed for vfs_hash */
238 int mnt_lockref; /* (i) Lock reference count */
239 int mnt_secondary_writes; /* (i) # of secondary writes */
240 int mnt_secondary_accwrites;/* (i) secondary wr. starts */
241 struct thread *mnt_susp_owner; /* (i) thread owning suspension */
242 #define mnt_endzero mnt_gjprovider
243 char *mnt_gjprovider; /* gjournal provider name */
244 struct mtx mnt_listmtx;
245 struct vnodelst mnt_lazyvnodelist; /* (l) list of lazy vnodes */
246 int mnt_lazyvnodelistsize; /* (l) # of lazy vnodes */
247 struct lock mnt_explock; /* vfs_export walkers lock */
248 TAILQ_ENTRY(mount) mnt_upper_link; /* (i*) we in the all uppers */
249 TAILQ_HEAD(, mount) mnt_uppers; /* (i) upper mounts over us */
250 };
251
252 /*
253 * Definitions for MNT_VNODE_FOREACH_ALL.
254 */
255 struct vnode *__mnt_vnode_next_all(struct vnode **mvp, struct mount *mp);
256 struct vnode *__mnt_vnode_first_all(struct vnode **mvp, struct mount *mp);
257 void __mnt_vnode_markerfree_all(struct vnode **mvp, struct mount *mp);
258
259 #define MNT_VNODE_FOREACH_ALL(vp, mp, mvp) \
260 for (vp = __mnt_vnode_first_all(&(mvp), (mp)); \
261 (vp) != NULL; vp = __mnt_vnode_next_all(&(mvp), (mp)))
262
263 #define MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp) \
264 do { \
265 MNT_ILOCK(mp); \
266 __mnt_vnode_markerfree_all(&(mvp), (mp)); \
267 /* MNT_IUNLOCK(mp); -- done in above function */ \
268 mtx_assert(MNT_MTX(mp), MA_NOTOWNED); \
269 } while (0)
270
271 /*
272 * Definitions for MNT_VNODE_FOREACH_LAZY.
273 */
274 typedef int mnt_lazy_cb_t(struct vnode *, void *);
275 struct vnode *__mnt_vnode_next_lazy(struct vnode **mvp, struct mount *mp,
276 mnt_lazy_cb_t *cb, void *cbarg);
277 struct vnode *__mnt_vnode_first_lazy(struct vnode **mvp, struct mount *mp,
278 mnt_lazy_cb_t *cb, void *cbarg);
279 void __mnt_vnode_markerfree_lazy(struct vnode **mvp, struct mount *mp);
280
281 #define MNT_VNODE_FOREACH_LAZY(vp, mp, mvp, cb, cbarg) \
282 for (vp = __mnt_vnode_first_lazy(&(mvp), (mp), (cb), (cbarg)); \
283 (vp) != NULL; \
284 vp = __mnt_vnode_next_lazy(&(mvp), (mp), (cb), (cbarg)))
285
286 #define MNT_VNODE_FOREACH_LAZY_ABORT(mp, mvp) \
287 __mnt_vnode_markerfree_lazy(&(mvp), (mp))
288
289 #define MNT_ILOCK(mp) mtx_lock(&(mp)->mnt_mtx)
290 #define MNT_ITRYLOCK(mp) mtx_trylock(&(mp)->mnt_mtx)
291 #define MNT_IUNLOCK(mp) mtx_unlock(&(mp)->mnt_mtx)
292 #define MNT_MTX(mp) (&(mp)->mnt_mtx)
293
294 #define MNT_REF(mp) do { \
295 mtx_assert(MNT_MTX(mp), MA_OWNED); \
296 mp->mnt_ref++; \
297 } while (0)
298 #define MNT_REL(mp) do { \
299 mtx_assert(MNT_MTX(mp), MA_OWNED); \
300 (mp)->mnt_ref--; \
301 if ((mp)->mnt_vfs_ops && (mp)->mnt_ref < 0) \
302 vfs_dump_mount_counters(mp); \
303 if ((mp)->mnt_ref == 0 && (mp)->mnt_vfs_ops) \
304 wakeup((mp)); \
305 } while (0)
306
307 #endif /* _KERNEL */
308
309 #if defined(_WANT_MNTOPTNAMES) || defined(_KERNEL)
310 struct mntoptnames {
311 uint64_t o_opt;
312 const char *o_name;
313 };
314 #define MNTOPT_NAMES \
315 { MNT_ASYNC, "asynchronous" }, \
316 { MNT_EXPORTED, "NFS exported" }, \
317 { MNT_LOCAL, "local" }, \
318 { MNT_NOATIME, "noatime" }, \
319 { MNT_NOEXEC, "noexec" }, \
320 { MNT_NOSUID, "nosuid" }, \
321 { MNT_NOSYMFOLLOW, "nosymfollow" }, \
322 { MNT_QUOTA, "with quotas" }, \
323 { MNT_RDONLY, "read-only" }, \
324 { MNT_SYNCHRONOUS, "synchronous" }, \
325 { MNT_UNION, "union" }, \
326 { MNT_NOCLUSTERR, "noclusterr" }, \
327 { MNT_NOCLUSTERW, "noclusterw" }, \
328 { MNT_SUIDDIR, "suiddir" }, \
329 { MNT_SOFTDEP, "soft-updates" }, \
330 { MNT_SUJ, "journaled soft-updates" }, \
331 { MNT_MULTILABEL, "multilabel" }, \
332 { MNT_ACLS, "acls" }, \
333 { MNT_NFS4ACLS, "nfsv4acls" }, \
334 { MNT_GJOURNAL, "gjournal" }, \
335 { MNT_AUTOMOUNTED, "automounted" }, \
336 { MNT_VERIFIED, "verified" }, \
337 { MNT_UNTRUSTED, "untrusted" }, \
338 { MNT_NOCOVER, "nocover" }, \
339 { MNT_EMPTYDIR, "emptydir" }, \
340 { MNT_UPDATE, "update" }, \
341 { MNT_DELEXPORT, "delexport" }, \
342 { MNT_RELOAD, "reload" }, \
343 { MNT_FORCE, "force" }, \
344 { MNT_SNAPSHOT, "snapshot" }, \
345 { 0, NULL }
346 #endif
347
348 /*
349 * User specifiable flags, stored in mnt_flag.
350 */
351 #define MNT_RDONLY 0x0000000000000001ULL /* read only filesystem */
352 #define MNT_SYNCHRONOUS 0x0000000000000002ULL /* fs written synchronously */
353 #define MNT_NOEXEC 0x0000000000000004ULL /* can't exec from filesystem */
354 #define MNT_NOSUID 0x0000000000000008ULL /* don't honor setuid fs bits */
355 #define MNT_NFS4ACLS 0x0000000000000010ULL /* enable NFS version 4 ACLs */
356 #define MNT_UNION 0x0000000000000020ULL /* union with underlying fs */
357 #define MNT_ASYNC 0x0000000000000040ULL /* fs written asynchronously */
358 #define MNT_SUIDDIR 0x0000000000100000ULL /* special SUID dir handling */
359 #define MNT_SOFTDEP 0x0000000000200000ULL /* using soft updates */
360 #define MNT_NOSYMFOLLOW 0x0000000000400000ULL /* do not follow symlinks */
361 #define MNT_GJOURNAL 0x0000000002000000ULL /* GEOM journal support enabled */
362 #define MNT_MULTILABEL 0x0000000004000000ULL /* MAC support for objects */
363 #define MNT_ACLS 0x0000000008000000ULL /* ACL support enabled */
364 #define MNT_NOATIME 0x0000000010000000ULL /* dont update file access time */
365 #define MNT_NOCLUSTERR 0x0000000040000000ULL /* disable cluster read */
366 #define MNT_NOCLUSTERW 0x0000000080000000ULL /* disable cluster write */
367 #define MNT_SUJ 0x0000000100000000ULL /* using journaled soft updates */
368 #define MNT_AUTOMOUNTED 0x0000000200000000ULL /* mounted by automountd(8) */
369 #define MNT_UNTRUSTED 0x0000000800000000ULL /* filesys metadata untrusted */
370
371 /*
372 * NFS export related mount flags.
373 */
374 #define MNT_EXRDONLY 0x0000000000000080ULL /* exported read only */
375 #define MNT_EXPORTED 0x0000000000000100ULL /* filesystem is exported */
376 #define MNT_DEFEXPORTED 0x0000000000000200ULL /* exported to the world */
377 #define MNT_EXPORTANON 0x0000000000000400ULL /* anon uid mapping for all */
378 #define MNT_EXKERB 0x0000000000000800ULL /* exported with Kerberos */
379 #define MNT_EXPUBLIC 0x0000000020000000ULL /* public export (WebNFS) */
380 #define MNT_EXTLS 0x0000004000000000ULL /* require TLS */
381 #define MNT_EXTLSCERT 0x0000008000000000ULL /* require TLS with client cert */
382 #define MNT_EXTLSCERTUSER 0x0000010000000000ULL /* require TLS with user cert */
383
384 /*
385 * Flags set by internal operations, but visible to the user.
386 */
387 #define MNT_LOCAL 0x0000000000001000ULL /* filesystem is stored locally */
388 #define MNT_QUOTA 0x0000000000002000ULL /* quotas are enabled on fs */
389 #define MNT_ROOTFS 0x0000000000004000ULL /* identifies the root fs */
390 #define MNT_USER 0x0000000000008000ULL /* mounted by a user */
391 #define MNT_IGNORE 0x0000000000800000ULL /* do not show entry in df */
392 #define MNT_VERIFIED 0x0000000400000000ULL /* filesystem is verified */
393
394 /*
395 * Mask of flags that are visible to statfs().
396 * XXX I think that this could now become (~(MNT_CMDFLAGS))
397 * but the 'mount' program may need changing to handle this.
398 */
399 #define MNT_VISFLAGMASK (MNT_RDONLY | MNT_SYNCHRONOUS | MNT_NOEXEC | \
400 MNT_NOSUID | MNT_UNION | MNT_SUJ | \
401 MNT_ASYNC | MNT_EXRDONLY | MNT_EXPORTED | \
402 MNT_DEFEXPORTED | MNT_EXPORTANON| MNT_EXKERB | \
403 MNT_LOCAL | MNT_USER | MNT_QUOTA | \
404 MNT_ROOTFS | MNT_NOATIME | MNT_NOCLUSTERR| \
405 MNT_NOCLUSTERW | MNT_SUIDDIR | MNT_SOFTDEP | \
406 MNT_IGNORE | MNT_EXPUBLIC | MNT_NOSYMFOLLOW | \
407 MNT_GJOURNAL | MNT_MULTILABEL | MNT_ACLS | \
408 MNT_NFS4ACLS | MNT_AUTOMOUNTED | MNT_VERIFIED | \
409 MNT_UNTRUSTED)
410
411 /* Mask of flags that can be updated. */
412 #define MNT_UPDATEMASK (MNT_NOSUID | MNT_NOEXEC | \
413 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | \
414 MNT_NOATIME | \
415 MNT_NOSYMFOLLOW | MNT_IGNORE | \
416 MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR | \
417 MNT_ACLS | MNT_USER | MNT_NFS4ACLS | \
418 MNT_AUTOMOUNTED | MNT_UNTRUSTED)
419
420 /*
421 * External filesystem command modifier flags.
422 * Unmount can use the MNT_FORCE flag.
423 * XXX: These are not STATES and really should be somewhere else.
424 * XXX: MNT_BYFSID and MNT_NONBUSY collide with MNT_ACLS and MNT_MULTILABEL,
425 * but because MNT_ACLS and MNT_MULTILABEL are only used for mount(2),
426 * and MNT_BYFSID and MNT_NONBUSY are only used for unmount(2),
427 * it's harmless.
428 */
429 #define MNT_UPDATE 0x0000000000010000ULL /* not real mount, just update */
430 #define MNT_DELEXPORT 0x0000000000020000ULL /* delete export host lists */
431 #define MNT_RELOAD 0x0000000000040000ULL /* reload filesystem data */
432 #define MNT_FORCE 0x0000000000080000ULL /* force unmount or readonly */
433 #define MNT_SNAPSHOT 0x0000000001000000ULL /* snapshot the filesystem */
434 #define MNT_NONBUSY 0x0000000004000000ULL /* check vnode use counts. */
435 #define MNT_BYFSID 0x0000000008000000ULL /* specify filesystem by ID. */
436 #define MNT_NOCOVER 0x0000001000000000ULL /* Do not cover a mount point */
437 #define MNT_EMPTYDIR 0x0000002000000000ULL /* Only mount on empty dir */
438 #define MNT_CMDFLAGS (MNT_UPDATE | MNT_DELEXPORT | MNT_RELOAD | \
439 MNT_FORCE | MNT_SNAPSHOT | MNT_NONBUSY | \
440 MNT_BYFSID | MNT_NOCOVER | MNT_EMPTYDIR)
441 /*
442 * Internal filesystem control flags stored in mnt_kern_flag.
443 *
444 * MNTK_UNMOUNT locks the mount entry so that name lookup cannot
445 * proceed past the mount point. This keeps the subtree stable during
446 * mounts and unmounts. When non-forced unmount flushes all vnodes
447 * from the mp queue, the MNTK_UNMOUNT flag prevents insmntque() from
448 * queueing new vnodes.
449 *
450 * MNTK_UNMOUNTF permits filesystems to detect a forced unmount while
451 * dounmount() is still waiting to lock the mountpoint. This allows
452 * the filesystem to cancel operations that might otherwise deadlock
453 * with the unmount attempt (used by NFS).
454 */
455 #define MNTK_UNMOUNTF 0x00000001 /* forced unmount in progress */
456 #define MNTK_ASYNC 0x00000002 /* filtered async flag */
457 #define MNTK_SOFTDEP 0x00000004 /* async disabled by softdep */
458 #define MNTK_NOMSYNC 0x00000008 /* don't do msync */
459 #define MNTK_DRAINING 0x00000010 /* lock draining is happening */
460 #define MNTK_REFEXPIRE 0x00000020 /* refcount expiring is happening */
461 #define MNTK_EXTENDED_SHARED 0x00000040 /* Allow shared locking for more ops */
462 #define MNTK_SHARED_WRITES 0x00000080 /* Allow shared locking for writes */
463 #define MNTK_NO_IOPF 0x00000100 /* Disallow page faults during reads
464 and writes. Filesystem shall properly
465 handle i/o state on EFAULT. */
466 #define MNTK_VGONE_UPPER 0x00000200
467 #define MNTK_VGONE_WAITER 0x00000400
468 #define MNTK_LOOKUP_EXCL_DOTDOT 0x00000800
469 #define MNTK_MARKER 0x00001000
470 #define MNTK_UNMAPPED_BUFS 0x00002000
471 #define MNTK_USES_BCACHE 0x00004000 /* FS uses the buffer cache. */
472 #define MNTK_TEXT_REFS 0x00008000 /* Keep use ref for text */
473 #define MNTK_VMSETSIZE_BUG 0x00010000
474 #define MNTK_UNIONFS 0x00020000 /* A hack for F_ISUNIONSTACK */
475 #define MNTK_FPLOOKUP 0x00040000 /* fast path lookup is supported */
476 #define MNTK_SUSPEND_ALL 0x00080000 /* Suspended by all-fs suspension */
477 #define MNTK_NOASYNC 0x00800000 /* disable async */
478 #define MNTK_UNMOUNT 0x01000000 /* unmount in progress */
479 #define MNTK_MWAIT 0x02000000 /* waiting for unmount to finish */
480 #define MNTK_SUSPEND 0x08000000 /* request write suspension */
481 #define MNTK_SUSPEND2 0x04000000 /* block secondary writes */
482 #define MNTK_SUSPENDED 0x10000000 /* write operations are suspended */
483 #define MNTK_NULL_NOCACHE 0x20000000 /* auto disable cache for nullfs
484 mounts over this fs */
485 #define MNTK_LOOKUP_SHARED 0x40000000 /* FS supports shared lock lookups */
486 #define MNTK_NOKNOTE 0x80000000 /* Don't send KNOTEs from VOP hooks */
487
488 #ifdef _KERNEL
489 static inline int
MNT_SHARED_WRITES(struct mount * mp)490 MNT_SHARED_WRITES(struct mount *mp)
491 {
492
493 return (mp != NULL && (mp->mnt_kern_flag & MNTK_SHARED_WRITES) != 0);
494 }
495
496 static inline int
MNT_EXTENDED_SHARED(struct mount * mp)497 MNT_EXTENDED_SHARED(struct mount *mp)
498 {
499
500 return (mp != NULL && (mp->mnt_kern_flag & MNTK_EXTENDED_SHARED) != 0);
501 }
502 #endif
503
504 /*
505 * Sysctl CTL_VFS definitions.
506 *
507 * Second level identifier specifies which filesystem. Second level
508 * identifier VFS_VFSCONF returns information about all filesystems.
509 * Second level identifier VFS_GENERIC is non-terminal.
510 */
511 #define VFS_VFSCONF 0 /* get configured filesystems */
512 #define VFS_GENERIC 0 /* generic filesystem information */
513 /*
514 * Third level identifiers for VFS_GENERIC are given below; third
515 * level identifiers for specific filesystems are given in their
516 * mount specific header files.
517 */
518 #define VFS_MAXTYPENUM 1 /* int: highest defined filesystem type */
519 #define VFS_CONF 2 /* struct: vfsconf for filesystem given
520 as next argument */
521
522 /*
523 * Flags for various system call interfaces.
524 *
525 * waitfor flags to vfs_sync() and getfsstat()
526 */
527 #define MNT_WAIT 1 /* synchronously wait for I/O to complete */
528 #define MNT_NOWAIT 2 /* start all I/O, but do not wait for it */
529 #define MNT_LAZY 3 /* push data not written by filesystem syncer */
530 #define MNT_SUSPEND 4 /* Suspend file system after sync */
531
532 /*
533 * Generic file handle
534 */
535 struct fhandle {
536 fsid_t fh_fsid; /* Filesystem id of mount point */
537 struct fid fh_fid; /* Filesys specific id */
538 };
539 typedef struct fhandle fhandle_t;
540
541 /*
542 * Old export arguments without security flavor list
543 */
544 struct oexport_args {
545 int ex_flags; /* export related flags */
546 uid_t ex_root; /* mapping for root uid */
547 struct xucred ex_anon; /* mapping for anonymous user */
548 struct sockaddr *ex_addr; /* net address to which exported */
549 u_char ex_addrlen; /* and the net address length */
550 struct sockaddr *ex_mask; /* mask of valid bits in saddr */
551 u_char ex_masklen; /* and the smask length */
552 char *ex_indexfile; /* index file for WebNFS URLs */
553 };
554
555 /*
556 * Not quite so old export arguments with 32bit ex_flags and xucred ex_anon.
557 */
558 #define MAXSECFLAVORS 5
559 struct o2export_args {
560 int ex_flags; /* export related flags */
561 uid_t ex_root; /* mapping for root uid */
562 struct xucred ex_anon; /* mapping for anonymous user */
563 struct sockaddr *ex_addr; /* net address to which exported */
564 u_char ex_addrlen; /* and the net address length */
565 struct sockaddr *ex_mask; /* mask of valid bits in saddr */
566 u_char ex_masklen; /* and the smask length */
567 char *ex_indexfile; /* index file for WebNFS URLs */
568 int ex_numsecflavors; /* security flavor count */
569 int ex_secflavors[MAXSECFLAVORS]; /* list of security flavors */
570 };
571
572 /*
573 * Export arguments for local filesystem mount calls.
574 */
575 struct export_args {
576 uint64_t ex_flags; /* export related flags */
577 uid_t ex_root; /* mapping for root uid */
578 uid_t ex_uid; /* mapping for anonymous user */
579 int ex_ngroups;
580 gid_t *ex_groups;
581 struct sockaddr *ex_addr; /* net address to which exported */
582 u_char ex_addrlen; /* and the net address length */
583 struct sockaddr *ex_mask; /* mask of valid bits in saddr */
584 u_char ex_masklen; /* and the smask length */
585 char *ex_indexfile; /* index file for WebNFS URLs */
586 int ex_numsecflavors; /* security flavor count */
587 int ex_secflavors[MAXSECFLAVORS]; /* list of security flavors */
588 };
589
590 /*
591 * Structure holding information for a publicly exported filesystem
592 * (WebNFS). Currently the specs allow just for one such filesystem.
593 */
594 struct nfs_public {
595 int np_valid; /* Do we hold valid information */
596 fhandle_t np_handle; /* Filehandle for pub fs (internal) */
597 struct mount *np_mount; /* Mountpoint of exported fs */
598 char *np_index; /* Index file */
599 };
600
601 /*
602 * Filesystem configuration information. One of these exists for each
603 * type of filesystem supported by the kernel. These are searched at
604 * mount time to identify the requested filesystem.
605 *
606 * XXX: Never change the first two arguments!
607 */
608 struct vfsconf {
609 u_int vfc_version; /* ABI version number */
610 char vfc_name[MFSNAMELEN]; /* filesystem type name */
611 struct vfsops *vfc_vfsops; /* filesystem operations vector */
612 struct vfsops *vfc_vfsops_sd; /* ... signal-deferred */
613 int vfc_typenum; /* historic filesystem type number */
614 int vfc_refcount; /* number mounted of this type */
615 int vfc_flags; /* permanent flags */
616 int vfc_prison_flag; /* prison allow.mount.* flag */
617 struct vfsoptdecl *vfc_opts; /* mount options */
618 TAILQ_ENTRY(vfsconf) vfc_list; /* list of vfscons */
619 };
620
621 /* Userland version of the struct vfsconf. */
622 struct xvfsconf {
623 struct vfsops *vfc_vfsops; /* filesystem operations vector */
624 char vfc_name[MFSNAMELEN]; /* filesystem type name */
625 int vfc_typenum; /* historic filesystem type number */
626 int vfc_refcount; /* number mounted of this type */
627 int vfc_flags; /* permanent flags */
628 struct vfsconf *vfc_next; /* next in list */
629 };
630
631 #ifndef BURN_BRIDGES
632 struct ovfsconf {
633 void *vfc_vfsops;
634 char vfc_name[32];
635 int vfc_index;
636 int vfc_refcount;
637 int vfc_flags;
638 };
639 #endif
640
641 /*
642 * NB: these flags refer to IMPLEMENTATION properties, not properties of
643 * any actual mounts; i.e., it does not make sense to change the flags.
644 */
645 #define VFCF_STATIC 0x00010000 /* statically compiled into kernel */
646 #define VFCF_NETWORK 0x00020000 /* may get data over the network */
647 #define VFCF_READONLY 0x00040000 /* writes are not implemented */
648 #define VFCF_SYNTHETIC 0x00080000 /* data does not represent real files */
649 #define VFCF_LOOPBACK 0x00100000 /* aliases some other mounted FS */
650 #define VFCF_UNICODE 0x00200000 /* stores file names as Unicode */
651 #define VFCF_JAIL 0x00400000 /* can be mounted from within a jail */
652 #define VFCF_DELEGADMIN 0x00800000 /* supports delegated administration */
653 #define VFCF_SBDRY 0x01000000 /* Stop at Boundary: defer stop requests
654 to kernel->user (AST) transition */
655 #define VFCF_FILEMOUNT 0x02000000 /* allow mounting files */
656 #define VFCF_FILEREVINC 0x04000000 /* va_filerev is incr. by one */
657 #define VFCF_FILEREVCT 0x08000000 /* va_filerev is set to ctime */
658
659 typedef uint32_t fsctlop_t;
660
661 struct vfsidctl {
662 int vc_vers; /* should be VFSIDCTL_VERS1 (below) */
663 fsid_t vc_fsid; /* fsid to operate on */
664 char vc_fstypename[MFSNAMELEN];
665 /* type of fs 'nfs' or '*' */
666 fsctlop_t vc_op; /* operation VFS_CTL_* (below) */
667 void *vc_ptr; /* pointer to data structure */
668 size_t vc_len; /* sizeof said structure */
669 u_int32_t vc_spare[12]; /* spare (must be zero) */
670 };
671
672 /* vfsidctl API version. */
673 #define VFS_CTL_VERS1 0x01
674
675 /*
676 * New style VFS sysctls, do not reuse/conflict with the namespace for
677 * private sysctls.
678 * All "global" sysctl ops have the 33rd bit set:
679 * 0x...1....
680 * Private sysctl ops should have the 33rd bit unset.
681 */
682 #define VFS_CTL_QUERY 0x00010001 /* anything wrong? (vfsquery) */
683 #define VFS_CTL_TIMEO 0x00010002 /* set timeout for vfs notification */
684 #define VFS_CTL_NOLOCKS 0x00010003 /* disable file locking */
685
686 struct vfsquery {
687 u_int32_t vq_flags;
688 u_int32_t vq_spare[31];
689 };
690
691 /* vfsquery flags */
692 #define VQ_NOTRESP 0x0001 /* server down */
693 #define VQ_NEEDAUTH 0x0002 /* server bad auth */
694 #define VQ_LOWDISK 0x0004 /* we're low on space */
695 #define VQ_MOUNT 0x0008 /* new filesystem arrived */
696 #define VQ_UNMOUNT 0x0010 /* filesystem has left */
697 #define VQ_DEAD 0x0020 /* filesystem is dead, needs force unmount */
698 #define VQ_ASSIST 0x0040 /* filesystem needs assistance from external
699 program */
700 #define VQ_NOTRESPLOCK 0x0080 /* server lockd down */
701 #define VQ_FLAG0100 0x0100 /* placeholder */
702 #define VQ_FLAG0200 0x0200 /* placeholder */
703 #define VQ_FLAG0400 0x0400 /* placeholder */
704 #define VQ_FLAG0800 0x0800 /* placeholder */
705 #define VQ_FLAG1000 0x1000 /* placeholder */
706 #define VQ_FLAG2000 0x2000 /* placeholder */
707 #define VQ_FLAG4000 0x4000 /* placeholder */
708 #define VQ_FLAG8000 0x8000 /* placeholder */
709
710 #ifdef _KERNEL
711 /* Point a sysctl request at a vfsidctl's data. */
712 #define VCTLTOREQ(vc, req) \
713 do { \
714 (req)->newptr = (vc)->vc_ptr; \
715 (req)->newlen = (vc)->vc_len; \
716 (req)->newidx = 0; \
717 } while (0)
718 #endif
719
720 struct iovec;
721 struct uio;
722
723 #ifdef _KERNEL
724
725 /*
726 * vfs_busy specific flags and mask.
727 */
728 #define MBF_NOWAIT 0x01
729 #define MBF_MNTLSTLOCK 0x02
730 #define MBF_MASK (MBF_NOWAIT | MBF_MNTLSTLOCK)
731
732 #ifdef MALLOC_DECLARE
733 MALLOC_DECLARE(M_MOUNT);
734 MALLOC_DECLARE(M_STATFS);
735 #endif
736 extern int maxvfsconf; /* highest defined filesystem type */
737
738 TAILQ_HEAD(vfsconfhead, vfsconf);
739 extern struct vfsconfhead vfsconf;
740
741 /*
742 * Operations supported on mounted filesystem.
743 */
744 struct mount_args;
745 struct nameidata;
746 struct sysctl_req;
747 struct mntarg;
748
749 /*
750 * N.B., vfs_cmount is the ancient vfsop invoked by the old mount(2) syscall.
751 * The new way is vfs_mount.
752 *
753 * vfs_cmount implementations typically translate arguments from their
754 * respective old per-FS structures into the key-value list supported by
755 * nmount(2), then use kernel_mount(9) to mimic nmount(2) from kernelspace.
756 *
757 * Filesystems with mounters that use nmount(2) do not need to and should not
758 * implement vfs_cmount. Hopefully a future cleanup can remove vfs_cmount and
759 * mount(2) entirely.
760 */
761 typedef int vfs_cmount_t(struct mntarg *ma, void *data, uint64_t flags);
762 typedef int vfs_unmount_t(struct mount *mp, int mntflags);
763 typedef int vfs_root_t(struct mount *mp, int flags, struct vnode **vpp);
764 typedef int vfs_quotactl_t(struct mount *mp, int cmds, uid_t uid, void *arg);
765 typedef int vfs_statfs_t(struct mount *mp, struct statfs *sbp);
766 typedef int vfs_sync_t(struct mount *mp, int waitfor);
767 typedef int vfs_vget_t(struct mount *mp, ino_t ino, int flags,
768 struct vnode **vpp);
769 typedef int vfs_fhtovp_t(struct mount *mp, struct fid *fhp,
770 int flags, struct vnode **vpp);
771 typedef int vfs_checkexp_t(struct mount *mp, struct sockaddr *nam,
772 uint64_t *extflagsp, struct ucred **credanonp,
773 int *numsecflavors, int *secflavors);
774 typedef int vfs_init_t(struct vfsconf *);
775 typedef int vfs_uninit_t(struct vfsconf *);
776 typedef int vfs_extattrctl_t(struct mount *mp, int cmd,
777 struct vnode *filename_vp, int attrnamespace,
778 const char *attrname);
779 typedef int vfs_mount_t(struct mount *mp);
780 typedef int vfs_sysctl_t(struct mount *mp, fsctlop_t op,
781 struct sysctl_req *req);
782 typedef void vfs_susp_clean_t(struct mount *mp);
783 typedef void vfs_notify_lowervp_t(struct mount *mp, struct vnode *lowervp);
784 typedef void vfs_purge_t(struct mount *mp);
785 struct sbuf;
786 typedef int vfs_report_lockf_t(struct mount *mp, struct sbuf *sb);
787
788 struct vfsops {
789 vfs_mount_t *vfs_mount;
790 vfs_cmount_t *vfs_cmount;
791 vfs_unmount_t *vfs_unmount;
792 vfs_root_t *vfs_root;
793 vfs_root_t *vfs_cachedroot;
794 vfs_quotactl_t *vfs_quotactl;
795 vfs_statfs_t *vfs_statfs;
796 vfs_sync_t *vfs_sync;
797 vfs_vget_t *vfs_vget;
798 vfs_fhtovp_t *vfs_fhtovp;
799 vfs_checkexp_t *vfs_checkexp;
800 vfs_init_t *vfs_init;
801 vfs_uninit_t *vfs_uninit;
802 vfs_extattrctl_t *vfs_extattrctl;
803 vfs_sysctl_t *vfs_sysctl;
804 vfs_susp_clean_t *vfs_susp_clean;
805 vfs_notify_lowervp_t *vfs_reclaim_lowervp;
806 vfs_notify_lowervp_t *vfs_unlink_lowervp;
807 vfs_purge_t *vfs_purge;
808 vfs_report_lockf_t *vfs_report_lockf;
809 vfs_mount_t *vfs_spare[5]; /* spares for ABI compat */
810 };
811
812 vfs_statfs_t __vfs_statfs;
813
814 #define VFS_MOUNT(MP) ({ \
815 int _rc; \
816 \
817 TSRAW(curthread, TS_ENTER, "VFS_MOUNT", (MP)->mnt_vfc->vfc_name);\
818 _rc = (*(MP)->mnt_op->vfs_mount)(MP); \
819 TSRAW(curthread, TS_EXIT, "VFS_MOUNT", (MP)->mnt_vfc->vfc_name);\
820 _rc; })
821
822 #define VFS_UNMOUNT(MP, FORCE) ({ \
823 int _rc; \
824 \
825 _rc = (*(MP)->mnt_op->vfs_unmount)(MP, FORCE); \
826 _rc; })
827
828 #define VFS_ROOT(MP, FLAGS, VPP) ({ \
829 int _rc; \
830 \
831 _rc = (*(MP)->mnt_op->vfs_root)(MP, FLAGS, VPP); \
832 _rc; })
833
834 #define VFS_CACHEDROOT(MP, FLAGS, VPP) ({ \
835 int _rc; \
836 \
837 _rc = (*(MP)->mnt_op->vfs_cachedroot)(MP, FLAGS, VPP); \
838 _rc; })
839
840 #define VFS_QUOTACTL(MP, C, U, A) ({ \
841 int _rc; \
842 \
843 _rc = (*(MP)->mnt_op->vfs_quotactl)(MP, C, U, A); \
844 _rc; })
845
846 #define VFS_STATFS(MP, SBP) ({ \
847 int _rc; \
848 \
849 _rc = __vfs_statfs((MP), (SBP)); \
850 _rc; })
851
852 #define VFS_SYNC(MP, WAIT) ({ \
853 int _rc; \
854 \
855 _rc = (*(MP)->mnt_op->vfs_sync)(MP, WAIT); \
856 _rc; })
857
858 #define VFS_VGET(MP, INO, FLAGS, VPP) ({ \
859 int _rc; \
860 \
861 _rc = (*(MP)->mnt_op->vfs_vget)(MP, INO, FLAGS, VPP); \
862 _rc; })
863
864 #define VFS_FHTOVP(MP, FIDP, FLAGS, VPP) ({ \
865 int _rc; \
866 \
867 _rc = (*(MP)->mnt_op->vfs_fhtovp)(MP, FIDP, FLAGS, VPP); \
868 _rc; })
869
870 #define VFS_CHECKEXP(MP, NAM, EXFLG, CRED, NUMSEC, SEC) ({ \
871 int _rc; \
872 \
873 _rc = (*(MP)->mnt_op->vfs_checkexp)(MP, NAM, EXFLG, CRED, NUMSEC,\
874 SEC); \
875 _rc; })
876
877 #define VFS_EXTATTRCTL(MP, C, FN, NS, N) ({ \
878 int _rc; \
879 \
880 _rc = (*(MP)->mnt_op->vfs_extattrctl)(MP, C, FN, NS, N); \
881 _rc; })
882
883 #define VFS_SYSCTL(MP, OP, REQ) ({ \
884 int _rc; \
885 \
886 _rc = (*(MP)->mnt_op->vfs_sysctl)(MP, OP, REQ); \
887 _rc; })
888
889 #define VFS_SUSP_CLEAN(MP) do { \
890 if (*(MP)->mnt_op->vfs_susp_clean != NULL) { \
891 (*(MP)->mnt_op->vfs_susp_clean)(MP); \
892 } \
893 } while (0)
894
895 #define VFS_RECLAIM_LOWERVP(MP, VP) do { \
896 if (*(MP)->mnt_op->vfs_reclaim_lowervp != NULL) { \
897 (*(MP)->mnt_op->vfs_reclaim_lowervp)((MP), (VP)); \
898 } \
899 } while (0)
900
901 #define VFS_UNLINK_LOWERVP(MP, VP) do { \
902 if (*(MP)->mnt_op->vfs_unlink_lowervp != NULL) { \
903 (*(MP)->mnt_op->vfs_unlink_lowervp)((MP), (VP)); \
904 } \
905 } while (0)
906
907 #define VFS_PURGE(MP) do { \
908 if (*(MP)->mnt_op->vfs_purge != NULL) { \
909 (*(MP)->mnt_op->vfs_purge)(MP); \
910 } \
911 } while (0)
912
913 #define VFS_KNOTE_LOCKED(vp, hint) do \
914 { \
915 if (((vp)->v_vflag & VV_NOKNOTE) == 0) \
916 VN_KNOTE((vp), (hint), KNF_LISTLOCKED); \
917 } while (0)
918
919 #define VFS_KNOTE_UNLOCKED(vp, hint) do \
920 { \
921 if (((vp)->v_vflag & VV_NOKNOTE) == 0) \
922 VN_KNOTE((vp), (hint), 0); \
923 } while (0)
924
925 #define VFS_NOTIFY_UPPER_RECLAIM 1
926 #define VFS_NOTIFY_UPPER_UNLINK 2
927
928 #include <sys/module.h>
929
930 /*
931 * Version numbers.
932 */
933 #define VFS_VERSION_00 0x19660120
934 #define VFS_VERSION_01 0x20121030
935 #define VFS_VERSION_02 0x20180504
936 #define VFS_VERSION VFS_VERSION_02
937
938 #define VFS_SET(vfsops, fsname, flags) \
939 static struct vfsconf fsname ## _vfsconf = { \
940 .vfc_version = VFS_VERSION, \
941 .vfc_name = #fsname, \
942 .vfc_vfsops = &vfsops, \
943 .vfc_typenum = -1, \
944 .vfc_flags = flags, \
945 }; \
946 static moduledata_t fsname ## _mod = { \
947 #fsname, \
948 vfs_modevent, \
949 & fsname ## _vfsconf \
950 }; \
951 DECLARE_MODULE(fsname, fsname ## _mod, SI_SUB_VFS, SI_ORDER_MIDDLE)
952
953 /*
954 * exported vnode operations
955 */
956
957 /* Define this to indicate that vfs_exjail_clone() exists for ZFS to use. */
958 #define VFS_SUPPORTS_EXJAIL_CLONE 1
959
960 int dounmount(struct mount *, int, struct thread *);
961
962 int kernel_mount(struct mntarg *ma, uint64_t flags);
963 int kernel_vmount(int flags, ...);
964 struct mntarg *mount_arg(struct mntarg *ma, const char *name, const void *val, int len);
965 struct mntarg *mount_argb(struct mntarg *ma, int flag, const char *name);
966 struct mntarg *mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...);
967 struct mntarg *mount_argsu(struct mntarg *ma, const char *name, const void *val, int len);
968 void statfs_scale_blocks(struct statfs *sf, long max_size);
969 struct vfsconf *vfs_byname(const char *);
970 struct vfsconf *vfs_byname_kld(const char *, struct thread *td, int *);
971 void vfs_mount_destroy(struct mount *);
972 void vfs_event_signal(fsid_t *, u_int32_t, intptr_t);
973 void vfs_freeopts(struct vfsoptlist *opts);
974 void vfs_deleteopt(struct vfsoptlist *opts, const char *name);
975 int vfs_buildopts(struct uio *auio, struct vfsoptlist **options);
976 int vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
977 uint64_t val);
978 int vfs_getopt(struct vfsoptlist *, const char *, void **, int *);
979 int vfs_getopt_pos(struct vfsoptlist *opts, const char *name);
980 int vfs_getopt_size(struct vfsoptlist *opts, const char *name,
981 off_t *value);
982 char *vfs_getopts(struct vfsoptlist *, const char *, int *error);
983 int vfs_copyopt(struct vfsoptlist *, const char *, void *, int);
984 int vfs_filteropt(struct vfsoptlist *, const char **legal);
985 void vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...);
986 int vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...);
987 int vfs_setopt(struct vfsoptlist *opts, const char *name, void *value,
988 int len);
989 int vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value,
990 int len);
991 int vfs_setopts(struct vfsoptlist *opts, const char *name,
992 const char *value);
993 int vfs_setpublicfs /* set publicly exported fs */
994 (struct mount *, struct netexport *, struct export_args *);
995 void vfs_periodic(struct mount *, int);
996 int vfs_busy(struct mount *, int);
997 void vfs_exjail_clone(struct mount *, struct mount *);
998 int vfs_export /* process mount export info */
999 (struct mount *, struct export_args *, int);
1000 void vfs_free_addrlist(struct netexport *);
1001 void vfs_allocate_syncvnode(struct mount *);
1002 void vfs_deallocate_syncvnode(struct mount *);
1003 int vfs_donmount(struct thread *td, uint64_t fsflags,
1004 struct uio *fsoptions);
1005 void vfs_getnewfsid(struct mount *);
1006 struct cdev *vfs_getrootfsid(struct mount *);
1007 struct mount *vfs_getvfs(fsid_t *); /* return vfs given fsid */
1008 struct mount *vfs_busyfs(fsid_t *);
1009 int vfs_modevent(module_t, int, void *);
1010 void vfs_mount_error(struct mount *, const char *, ...);
1011 void vfs_mountroot(void); /* mount our root filesystem */
1012 void vfs_mountedfrom(struct mount *, const char *from);
1013 void vfs_notify_upper(struct vnode *, int);
1014 struct mount *vfs_ref_from_vp(struct vnode *);
1015 void vfs_ref(struct mount *);
1016 void vfs_rel(struct mount *);
1017 struct mount *vfs_mount_alloc(struct vnode *, struct vfsconf *, const char *,
1018 struct ucred *);
1019 int vfs_suser(struct mount *, struct thread *);
1020 void vfs_unbusy(struct mount *);
1021 void vfs_unmountall(void);
1022 int vfs_remount_ro(struct mount *mp);
1023 int vfs_report_lockf(struct mount *mp, struct sbuf *sb);
1024
1025 extern TAILQ_HEAD(mntlist, mount) mountlist; /* mounted filesystem list */
1026 extern struct mtx_padalign mountlist_mtx;
1027 extern struct nfs_public nfs_pub;
1028 extern struct sx vfsconf_sx;
1029 #define vfsconf_lock() sx_xlock(&vfsconf_sx)
1030 #define vfsconf_unlock() sx_xunlock(&vfsconf_sx)
1031 #define vfsconf_slock() sx_slock(&vfsconf_sx)
1032 #define vfsconf_sunlock() sx_sunlock(&vfsconf_sx)
1033 struct vnode *mntfs_allocvp(struct mount *, struct vnode *);
1034 void mntfs_freevp(struct vnode *);
1035
1036 /*
1037 * Declarations for these vfs default operations are located in
1038 * kern/vfs_default.c. They will be automatically used to replace
1039 * null entries in VFS ops tables when registering a new filesystem
1040 * type in the global table.
1041 */
1042 vfs_root_t vfs_stdroot;
1043 vfs_quotactl_t vfs_stdquotactl;
1044 vfs_statfs_t vfs_stdstatfs;
1045 vfs_sync_t vfs_stdsync;
1046 vfs_sync_t vfs_stdnosync;
1047 vfs_vget_t vfs_stdvget;
1048 vfs_fhtovp_t vfs_stdfhtovp;
1049 vfs_checkexp_t vfs_stdcheckexp;
1050 vfs_init_t vfs_stdinit;
1051 vfs_uninit_t vfs_stduninit;
1052 vfs_extattrctl_t vfs_stdextattrctl;
1053 vfs_sysctl_t vfs_stdsysctl;
1054
1055 void syncer_suspend(void);
1056 void syncer_resume(void);
1057
1058 struct vnode *vfs_cache_root_clear(struct mount *);
1059 void vfs_cache_root_set(struct mount *, struct vnode *);
1060
1061 void vfs_op_barrier_wait(struct mount *);
1062 void vfs_op_enter(struct mount *);
1063 void vfs_op_exit_locked(struct mount *);
1064 void vfs_op_exit(struct mount *);
1065
1066 #ifdef DIAGNOSTIC
1067 void vfs_assert_mount_counters(struct mount *);
1068 void vfs_dump_mount_counters(struct mount *);
1069 #else
1070 #define vfs_assert_mount_counters(mp) do { } while (0)
1071 #define vfs_dump_mount_counters(mp) do { } while (0)
1072 #endif
1073
1074 enum mount_counter { MNT_COUNT_REF, MNT_COUNT_LOCKREF, MNT_COUNT_WRITEOPCOUNT };
1075 int vfs_mount_fetch_counter(struct mount *, enum mount_counter);
1076
1077 void suspend_all_fs(void);
1078 void resume_all_fs(void);
1079
1080 /*
1081 * Code transitioning mnt_vfs_ops to > 0 issues IPIs until it observes
1082 * all CPUs not executing code enclosed by thread_in_ops_pcpu variable.
1083 *
1084 * This provides an invariant that by the time the last CPU is observed not
1085 * executing, everyone else entering will see the counter > 0 and exit.
1086 *
1087 * Note there is no barrier between vfs_ops and the rest of the code in the
1088 * section. It is not necessary as the writer has to wait for everyone to drain
1089 * before making any changes or only make changes safe while the section is
1090 * executed.
1091 */
1092 #define vfs_mount_pcpu(mp) zpcpu_get(mp->mnt_pcpu)
1093 #define vfs_mount_pcpu_remote(mp, cpu) zpcpu_get_cpu(mp->mnt_pcpu, cpu)
1094
1095 #define vfs_op_thread_entered(mp) ({ \
1096 MPASS(curthread->td_critnest > 0); \
1097 struct mount_pcpu *_mpcpu = vfs_mount_pcpu(mp); \
1098 _mpcpu->mntp_thread_in_ops == 1; \
1099 })
1100
1101 #define vfs_op_thread_enter_crit(mp, _mpcpu) ({ \
1102 bool _retval_crit = true; \
1103 MPASS(curthread->td_critnest > 0); \
1104 _mpcpu = vfs_mount_pcpu(mp); \
1105 MPASS(mpcpu->mntp_thread_in_ops == 0); \
1106 _mpcpu->mntp_thread_in_ops = 1; \
1107 atomic_interrupt_fence(); \
1108 if (__predict_false(mp->mnt_vfs_ops > 0)) { \
1109 vfs_op_thread_exit_crit(mp, _mpcpu); \
1110 _retval_crit = false; \
1111 } \
1112 _retval_crit; \
1113 })
1114
1115 #define vfs_op_thread_enter(mp, _mpcpu) ({ \
1116 bool _retval; \
1117 critical_enter(); \
1118 _retval = vfs_op_thread_enter_crit(mp, _mpcpu); \
1119 if (__predict_false(!_retval)) \
1120 critical_exit(); \
1121 _retval; \
1122 })
1123
1124 #define vfs_op_thread_exit_crit(mp, _mpcpu) do { \
1125 MPASS(_mpcpu == vfs_mount_pcpu(mp)); \
1126 MPASS(_mpcpu->mntp_thread_in_ops == 1); \
1127 atomic_interrupt_fence(); \
1128 _mpcpu->mntp_thread_in_ops = 0; \
1129 } while (0)
1130
1131 #define vfs_op_thread_exit(mp, _mpcpu) do { \
1132 vfs_op_thread_exit_crit(mp, _mpcpu); \
1133 critical_exit(); \
1134 } while (0)
1135
1136 #define vfs_mp_count_add_pcpu(_mpcpu, count, val) do { \
1137 MPASS(_mpcpu->mntp_thread_in_ops == 1); \
1138 _mpcpu->mntp_##count += val; \
1139 } while (0)
1140
1141 #define vfs_mp_count_sub_pcpu(_mpcpu, count, val) do { \
1142 MPASS(_mpcpu->mntp_thread_in_ops == 1); \
1143 _mpcpu->mntp_##count -= val; \
1144 } while (0)
1145
1146 #else /* !_KERNEL */
1147
1148 #include <sys/cdefs.h>
1149
1150 struct stat;
1151
1152 __BEGIN_DECLS
1153 int fhlink(struct fhandle *, const char *);
1154 int fhlinkat(struct fhandle *, int, const char *);
1155 int fhopen(const struct fhandle *, int);
1156 int fhreadlink(struct fhandle *, char *, size_t);
1157 int fhstat(const struct fhandle *, struct stat *);
1158 int fhstatfs(const struct fhandle *, struct statfs *);
1159 int fstatfs(int, struct statfs *);
1160 int getfh(const char *, fhandle_t *);
1161 int getfhat(int, char *, struct fhandle *, int);
1162 int getfsstat(struct statfs *, long, int);
1163 int getmntinfo(struct statfs **, int);
1164 int lgetfh(const char *, fhandle_t *);
1165 int mount(const char *, const char *, int, void *);
1166 int nmount(struct iovec *, unsigned int, int);
1167 int statfs(const char *, struct statfs *);
1168 int unmount(const char *, int);
1169
1170 /* C library stuff */
1171 int getvfsbyname(const char *, struct xvfsconf *);
1172 __END_DECLS
1173
1174 #endif /* _KERNEL */
1175
1176 #endif /* !_SYS_MOUNT_H_ */
1177