4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright (c) 1988, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright 2016 Toomas Soome <tsoome@me.com>
25 * Copyright 2016 Nexenta Systems, Inc.
28 /* Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T */
29 /* All Rights Reserved */
32 * Portions of this source code were derived from Berkeley 4.3 BSD
33 * under license from the Regents of the University of California.
40 #include <sys/types.h>
41 #include <sys/t_lock.h>
43 #include <sys/vnode.h>
44 #include <sys/statvfs.h>
45 #include <sys/refstr.h>
54 * Data associated with mounted file systems.
58 * Operations vector. This is used internal to the kernel; file systems
59 * supply their list of operations via vfs_setfsops().
62 typedef struct vfsops vfsops_t
;
65 * File system identifier. Should be unique (at least per machine).
68 int val
[2]; /* file system id type */
72 * File identifier. Should be unique per filesystem on a single
73 * machine. This is typically called by a stateless file server
74 * in order to generate "file handles".
76 * Many underlying file systems cast a struct fid into other
77 * file system dependent structures which may require 4 byte alignment.
78 * Because a fid starts with a short it may not be 4 byte aligned, the
79 * fid_pad will force the alignment.
82 #define OLD_MAXFIDSZ 16
88 ushort_t len
; /* length of data in bytes */
89 char data
[MAXFIDSZ
]; /* data (variable len) */
96 * Solaris 64 - use old-style cache format with 32-bit aligned fid for on-disk
97 * struct compatibility.
99 typedef struct fid32
{
103 uint16_t len
; /* length of data in bytes */
104 char data
[MAXFIDSZ
]; /* data (variable len) */
108 #else /* not _SYSCALL32 */
110 typedef fid_t fid32_t
;
111 #endif /* _SYSCALL32 */
113 #define fid_len un._fid.len
114 #define fid_data un._fid.data
117 * Structure defining a mount option for a filesystem.
118 * option names are found in mntent.h
120 typedef struct mntopt
{
121 char *mo_name
; /* option name */
122 char **mo_cancel
; /* list of options cancelled by this one */
123 char *mo_arg
; /* argument string for this option */
124 int mo_flags
; /* flags for this mount option */
125 void *mo_data
; /* filesystem specific data */
129 * Flags that apply to mount options
132 #define MO_SET 0x01 /* option is set */
133 #define MO_NODISPLAY 0x02 /* option not listed in mnttab */
134 #define MO_HASVALUE 0x04 /* option takes a value */
135 #define MO_IGNORE 0x08 /* option ignored by parser */
136 #define MO_DEFAULT MO_SET /* option is on by default */
137 #define MO_TAG 0x10 /* flags a tag set by user program */
138 #define MO_EMPTY 0x20 /* empty space in option table */
140 #define VFS_NOFORCEOPT 0x01 /* honor MO_IGNORE (don't set option) */
141 #define VFS_DISPLAY 0x02 /* Turn off MO_NODISPLAY bit for opt */
142 #define VFS_NODISPLAY 0x04 /* Turn on MO_NODISPLAY bit for opt */
143 #define VFS_CREATEOPT 0x08 /* Create the opt if it's not there */
146 * Structure holding mount option strings for the mounted file system.
148 typedef struct mntopts
{
149 uint_t mo_count
; /* number of entries in table */
150 mntopt_t
*mo_list
; /* list of mount options */
154 * The kstat structures associated with the vopstats are kept in an
155 * AVL tree. This is to avoid the case where a file system does not
156 * use a unique fsid_t for each vfs (e.g., namefs). In order to do
157 * this, we need a structure that the AVL tree can use that also
158 * references the kstat.
159 * Note that the vks_fsid is generated from the value reported by
162 typedef struct vskstat_anchor
{
163 avl_node_t vsk_node
; /* Required for use by AVL routines */
164 kstat_t
*vsk_ksp
; /* kstat structure for vopstats */
165 ulong_t vsk_fsid
; /* fsid associated w/this FS */
168 extern avl_tree_t vskstat_tree
;
169 extern kmutex_t vskstat_tree_lock
;
172 * Structure per mounted file system. Each mounted file system has
173 * an array of operations and an instance record.
175 * The file systems are kept on a doubly linked circular list headed by
177 * File system implementations should not access this list;
178 * it's intended for use only in the kernel's vfs layer.
180 * Each zone also has its own list of mounts, containing filesystems mounted
181 * somewhere within the filesystem tree rooted at the zone's rootpath. The
182 * list is doubly linked to match the global list.
184 * mnttab locking: the in-kernel mnttab uses the vfs_mntpt, vfs_resource and
185 * vfs_mntopts fields in the vfs_t. mntpt and resource are refstr_ts that
186 * are set at mount time and can only be modified during a remount.
187 * It is safe to read these fields if you can prevent a remount on the vfs,
188 * or through the convenience funcs vfs_getmntpoint() and vfs_getresource().
189 * The mntopts field may only be accessed through the provided convenience
190 * functions, as it is protected by the vfs list lock. Modifying a mount
191 * option requires grabbing the vfs list write lock, which can be a very
194 struct zone
; /* from zone.h */
195 struct fem_head
; /* from fem.h */
198 struct vfs
*vfs_next
; /* next VFS in VFS list */
199 struct vfs
*vfs_prev
; /* prev VFS in VFS list */
201 /* vfs_op should not be used directly. Accessor functions are provided */
202 vfsops_t
*vfs_op
; /* operations on VFS */
204 struct vnode
*vfs_vnodecovered
; /* vnode mounted on */
205 uint_t vfs_flag
; /* flags */
206 uint_t vfs_bsize
; /* native block size */
207 int vfs_fstype
; /* file system type index */
208 fsid_t vfs_fsid
; /* file system id */
209 void *vfs_data
; /* private data */
210 dev_t vfs_dev
; /* device of mounted VFS */
211 ulong_t vfs_bcount
; /* I/O count (accounting) */
212 struct vfs
*vfs_list
; /* sync list pointer */
213 struct vfs
*vfs_hash
; /* hash list pointer */
214 ksema_t vfs_reflock
; /* mount/unmount/sync lock */
215 uint_t vfs_count
; /* vfs reference count */
216 mntopts_t vfs_mntopts
; /* options mounted with */
217 refstr_t
*vfs_resource
; /* mounted resource name */
218 refstr_t
*vfs_mntpt
; /* mount point name */
219 time_t vfs_mtime
; /* time we were mounted */
220 struct vfs_impl
*vfs_implp
; /* impl specific data */
222 * Zones support. Note that the zone that "owns" the mount isn't
223 * necessarily the same as the zone in which the zone is visible.
224 * That is, vfs_zone and (vfs_zone_next|vfs_zone_prev) may refer to
227 struct zone
*vfs_zone
; /* zone that owns the mount */
228 struct vfs
*vfs_zone_next
; /* next VFS visible in zone */
229 struct vfs
*vfs_zone_prev
; /* prev VFS visible in zone */
231 struct fem_head
*vfs_femhead
; /* fs monitoring */
232 uint32_t vfs_lofi_id
; /* ID if lofi mount */
235 #define vfs_featureset vfs_implp->vi_featureset
236 #define vfs_vskap vfs_implp->vi_vskap
237 #define vfs_fstypevsp vfs_implp->vi_fstypevsp
238 #define vfs_vopstats vfs_implp->vi_vopstats
239 #define vfs_hrctime vfs_implp->vi_hrctime
244 #define VFS_RDONLY 0x01 /* read-only vfs */
245 #define VFS_NOMNTTAB 0x02 /* vfs not seen in mnttab */
246 #define VFS_NOSETUID 0x08 /* setuid disallowed */
247 #define VFS_REMOUNT 0x10 /* modify mount options only */
248 #define VFS_NOTRUNC 0x20 /* does not truncate long file names */
249 #define VFS_UNLINKABLE 0x40 /* unlink(2) can be applied to root */
250 #define VFS_PXFS 0x80 /* clustering: global fs proxy vfs */
251 #define VFS_UNMOUNTED 0x100 /* file system has been unmounted */
252 #define VFS_NBMAND 0x200 /* allow non-blocking mandatory locks */
253 #define VFS_XATTR 0x400 /* fs supports extended attributes */
254 #define VFS_NODEVICES 0x800 /* device-special files disallowed */
255 #define VFS_NOEXEC 0x1000 /* executables disallowed */
256 #define VFS_STATS 0x2000 /* file system can collect stats */
257 #define VFS_XID 0x4000 /* file system supports extended ids */
259 #define VFS_NORESOURCE "unspecified_resource"
260 #define VFS_NOMNTPT "unspecified_mountpoint"
263 * VFS features are implemented as bits set in the vfs_t.
264 * The vfs_feature_t typedef is a 64-bit number that will translate
265 * into an element in an array of bitmaps and a bit in that element.
266 * Developers must not depend on the implementation of this and
267 * need to use vfs_has_feature()/vfs_set_feature() routines.
269 typedef uint64_t vfs_feature_t
;
271 #define VFSFT_XVATTR 0x100000001 /* Supports xvattr for attrs */
272 #define VFSFT_CASEINSENSITIVE 0x100000002 /* Supports case-insensitive */
273 #define VFSFT_NOCASESENSITIVE 0x100000004 /* NOT case-sensitive */
274 #define VFSFT_DIRENTFLAGS 0x100000008 /* Supports dirent flags */
275 #define VFSFT_ACLONCREATE 0x100000010 /* Supports ACL on create */
276 #define VFSFT_ACEMASKONACCESS 0x100000020 /* Can use ACEMASK for access */
277 #define VFSFT_SYSATTR_VIEWS 0x100000040 /* Supports sysattr view i/f */
278 #define VFSFT_ACCESS_FILTER 0x100000080 /* dirents filtered by access */
279 #define VFSFT_REPARSE 0x100000100 /* Supports reparse point */
280 #define VFSFT_ZEROCOPY_SUPPORTED 0x100000200
281 /* Support loaning /returning cache buffer */
283 * Argument structure for mount(2).
285 * Flags are defined in <sys/mount.h>.
287 * Note that if the MS_SYSSPACE bit is set in flags, the pointer fields in
288 * this structure are to be interpreted as kernel addresses. File systems
289 * should be prepared for this possibility.
303 * Reasons for calling the vfs_mountroot() operation.
305 enum whymountroot
{ ROOT_INIT
, ROOT_REMOUNT
, ROOT_UNMOUNT
};
306 typedef enum whymountroot whymountroot_t
;
309 * Reasons for calling the VFS_VNSTATE():
317 typedef enum vntrans vntrans_t
;
320 * VFS_OPS defines all the vfs operations. It is used to define
321 * the vfsops structure (below) and the fs_func_p union (vfs_opreg.h).
324 int (*vfs_mount)(vfs_t *, vnode_t *, struct mounta *, cred_t *); \
325 int (*vfs_unmount)(vfs_t *, int, cred_t *); \
326 int (*vfs_root)(vfs_t *, vnode_t **); \
327 int (*vfs_statvfs)(vfs_t *, statvfs64_t *); \
328 int (*vfs_sync)(vfs_t *, short, cred_t *); \
329 int (*vfs_vget)(vfs_t *, vnode_t **, fid_t *); \
330 int (*vfs_mountroot)(vfs_t *, enum whymountroot); \
331 void (*vfs_freevfs)(vfs_t *); \
332 int (*vfs_vnstate)(vfs_t *, vnode_t *, vntrans_t) /* NB: No ";" */
335 * Operations supported on virtual file system.
338 VFS_OPS
; /* Signature of all vfs operations (vfsops) */
341 extern int fsop_mount(vfs_t
*, vnode_t
*, struct mounta
*, cred_t
*);
342 extern int fsop_unmount(vfs_t
*, int, cred_t
*);
343 extern int fsop_root(vfs_t
*, vnode_t
**);
344 extern int fsop_statfs(vfs_t
*, statvfs64_t
*);
345 extern int fsop_sync(vfs_t
*, short, cred_t
*);
346 extern int fsop_vget(vfs_t
*, vnode_t
**, fid_t
*);
347 extern int fsop_mountroot(vfs_t
*, enum whymountroot
);
348 extern void fsop_freefs(vfs_t
*);
349 extern int fsop_sync_by_kind(int, short, cred_t
*);
350 extern int fsop_vnstate(vfs_t
*, vnode_t
*, vntrans_t
);
352 #define VFS_MOUNT(vfsp, mvp, uap, cr) fsop_mount(vfsp, mvp, uap, cr)
353 #define VFS_UNMOUNT(vfsp, flag, cr) fsop_unmount(vfsp, flag, cr)
354 #define VFS_ROOT(vfsp, vpp) fsop_root(vfsp, vpp)
355 #define VFS_STATVFS(vfsp, sp) fsop_statfs(vfsp, sp)
356 #define VFS_SYNC(vfsp, flag, cr) fsop_sync(vfsp, flag, cr)
357 #define VFS_VGET(vfsp, vpp, fidp) fsop_vget(vfsp, vpp, fidp)
358 #define VFS_MOUNTROOT(vfsp, init) fsop_mountroot(vfsp, init)
359 #define VFS_FREEVFS(vfsp) fsop_freefs(vfsp)
360 #define VFS_VNSTATE(vfsp, vn, ns) fsop_vnstate(vfsp, vn, ns)
362 #define VFSNAME_MOUNT "mount"
363 #define VFSNAME_UNMOUNT "unmount"
364 #define VFSNAME_ROOT "root"
365 #define VFSNAME_STATVFS "statvfs"
366 #define VFSNAME_SYNC "sync"
367 #define VFSNAME_VGET "vget"
368 #define VFSNAME_MOUNTROOT "mountroot"
369 #define VFSNAME_FREEVFS "freevfs"
370 #define VFSNAME_VNSTATE "vnstate"
372 * Filesystem type switch table.
375 typedef struct vfssw
{
376 char *vsw_name
; /* type name -- max len _ST_FSTYPSZ */
377 int (*vsw_init
) (int, char *);
378 /* init routine (for non-loadable fs only) */
379 int vsw_flag
; /* flags */
380 mntopts_t vsw_optproto
; /* mount options table prototype */
381 uint_t vsw_count
; /* count of references */
382 kmutex_t vsw_lock
; /* lock to protect vsw_count */
383 vfsops_t vsw_vfsops
; /* filesystem operations vector */
387 * Filesystem type definition record. All file systems must export a record
388 * of this type through their modlfs structure. N.B., changing the version
389 * number requires a change in sys/modctl.h.
392 typedef struct vfsdef_v5
{
393 int def_version
; /* structure version, must be first */
394 char *name
; /* filesystem type name */
395 int (*init
) (int, char *); /* init routine */
396 int flags
; /* filesystem flags */
397 mntopts_t
*optproto
; /* mount options table prototype */
400 typedef struct vfsdef_v5 vfsdef_t
;
407 * flags for vfssw and vfsdef
409 #define VSW_HASPROTO 0x01 /* struct has a mount options prototype */
410 #define VSW_CANRWRO 0x02 /* file system can transition from rw to ro */
411 #define VSW_CANREMOUNT 0x04 /* file system supports remounts */
412 #define VSW_NOTZONESAFE 0x08 /* zone_enter(2) should fail for these files */
413 #define VSW_VOLATILEDEV 0x10 /* vfs_dev can change each time fs is mounted */
414 #define VSW_STATS 0x20 /* file system can collect stats */
415 #define VSW_XID 0x40 /* file system supports extended ids */
416 #define VSW_CANLOFI 0x80 /* file system supports lofi mounts */
417 #define VSW_ZMOUNT 0x100 /* file system always allowed in a zone */
419 #define VSW_INSTALLED 0x8000 /* this vsw is associated with a file system */
422 * A flag for vfs_setpath().
424 #define VFSSP_VERBATIM 0x1 /* do not prefix the supplied path */
426 #if defined(_KERNEL) || defined(_FAKE_KERNEL)
429 * Private vfs data, NOT to be used by a file system implementation.
432 #define VFS_FEATURE_MAXSZ 4
434 typedef struct vfs_impl
{
435 /* Counted array - Bitmap of vfs features */
436 uint32_t vi_featureset
[VFS_FEATURE_MAXSZ
];
438 * Support for statistics on the vnode operations
440 vsk_anchor_t
*vi_vskap
; /* anchor for vopstats' kstat */
441 vopstats_t
*vi_fstypevsp
; /* ptr to per-fstype vopstats */
442 vopstats_t vi_vopstats
; /* per-mount vnode op stats */
444 timespec_t vi_hrctime
; /* High-res creation time */
446 zone_ref_t vi_zone_ref
; /* reference to zone */
456 void vfs_freevfsops(vfsops_t
*);
457 int vfs_freevfsops_by_type(int);
458 void vfs_setops(vfs_t
*, vfsops_t
*);
459 vfsops_t
*vfs_getops(vfs_t
*vfsp
);
460 int vfs_matchops(vfs_t
*, vfsops_t
*);
461 int vfs_can_sync(vfs_t
*vfsp
);
462 vfs_t
*vfs_alloc(int);
463 void vfs_free(vfs_t
*);
464 void vfs_init(vfs_t
*vfsp
, vfsops_t
*, void *);
465 void vfsimpl_setup(vfs_t
*vfsp
);
466 void vfsimpl_teardown(vfs_t
*vfsp
);
467 void vn_exists(vnode_t
*);
468 void vn_idle(vnode_t
*);
469 void vn_reclaim(vnode_t
*);
470 void vn_invalid(vnode_t
*);
473 int domount(char *, struct mounta
*, vnode_t
*, struct cred
*,
475 int dounmount(struct vfs
*, int, cred_t
*);
476 int vfs_lock(struct vfs
*);
477 int vfs_rlock(struct vfs
*);
478 void vfs_lock_wait(struct vfs
*);
479 void vfs_rlock_wait(struct vfs
*);
480 void vfs_unlock(struct vfs
*);
481 int vfs_lock_held(struct vfs
*);
482 struct _kthread
*vfs_lock_owner(struct vfs
*);
485 void vfs_mountroot(void);
486 void vfs_add(vnode_t
*, struct vfs
*, int);
487 void vfs_remove(struct vfs
*);
489 /* VFS feature routines */
490 void vfs_set_feature(vfs_t
*, vfs_feature_t
);
491 void vfs_clear_feature(vfs_t
*, vfs_feature_t
);
492 int vfs_has_feature(vfs_t
*, vfs_feature_t
);
493 void vfs_propagate_features(vfs_t
*, vfs_t
*);
495 /* The following functions are not for general use by filesystems */
497 void vfs_createopttbl(mntopts_t
*, const char *);
498 void vfs_copyopttbl(const mntopts_t
*, mntopts_t
*);
499 void vfs_mergeopttbl(const mntopts_t
*, const mntopts_t
*, mntopts_t
*);
500 void vfs_freeopttbl(mntopts_t
*);
501 void vfs_parsemntopts(mntopts_t
*, char *, int);
502 int vfs_buildoptionstr(const mntopts_t
*, char *, int);
503 struct mntopt
*vfs_hasopt(const mntopts_t
*, const char *);
504 void vfs_mnttab_modtimeupd(void);
506 void vfs_clearmntopt(struct vfs
*, const char *);
507 void vfs_setmntopt(struct vfs
*, const char *, const char *, int);
508 void vfs_setresource(struct vfs
*, const char *, uint32_t);
509 void vfs_setmntpoint(struct vfs
*, const char *, uint32_t);
510 refstr_t
*vfs_getresource(const struct vfs
*);
511 refstr_t
*vfs_getmntpoint(const struct vfs
*);
512 int vfs_optionisset(const struct vfs
*, const char *, char **);
513 int vfs_settag(uint_t
, uint_t
, const char *, const char *, cred_t
*);
514 int vfs_clrtag(uint_t
, uint_t
, const char *, const char *, cred_t
*);
515 void vfs_syncall(void);
516 void vfs_syncprogress(void);
518 void vfs_unmountall(void);
519 void vfs_make_fsid(fsid_t
*, dev_t
, int);
520 void vfs_addmip(dev_t
, struct vfs
*);
521 void vfs_delmip(struct vfs
*);
522 int vfs_devismounted(dev_t
);
523 int vfs_devmounting(dev_t
, struct vfs
*);
524 int vfs_opsinuse(vfsops_t
*);
525 struct vfs
*getvfs(fsid_t
*);
526 struct vfs
*vfs_dev2vfsp(dev_t
);
527 struct vfs
*vfs_mntpoint2vfsp(const char *);
528 struct vfssw
*allocate_vfssw(const char *);
529 struct vfssw
*vfs_getvfssw(const char *);
530 struct vfssw
*vfs_getvfsswbyname(const char *);
531 struct vfssw
*vfs_getvfsswbyvfsops(vfsops_t
*);
532 void vfs_refvfssw(struct vfssw
*);
533 void vfs_unrefvfssw(struct vfssw
*);
534 uint_t
vf_to_stf(uint_t
);
535 void vfs_mnttab_modtime(timespec_t
*);
536 void vfs_mnttab_poll(timespec_t
*, struct pollhead
**);
538 void vfs_list_lock(void);
539 void vfs_list_read_lock(void);
540 void vfs_list_unlock(void);
541 void vfs_list_add(struct vfs
*);
542 void vfs_list_remove(struct vfs
*);
543 void vfs_hold(vfs_t
*vfsp
);
544 void vfs_rele(vfs_t
*vfsp
);
545 void fs_freevfs(vfs_t
*);
546 void vfs_root_redev(vfs_t
*vfsp
, dev_t ndev
, int fstype
);
548 int vfs_zone_change_safe(vfs_t
*);
550 int vfs_get_lofi(vfs_t
*, vnode_t
**);
552 #define VFSHASH(maj, min) (((int)((maj)+(min))) & (vfshsz - 1))
553 #define VFS_ON_LIST(vfsp) \
554 ((vfsp)->vfs_next != (vfsp) && (vfsp)->vfs_next != NULL)
560 extern struct vfssw vfssw
[]; /* table of filesystem types */
561 extern krwlock_t vfssw_lock
;
562 extern char rootfstype
[]; /* name of root fstype */
563 extern const int nfstype
; /* # of elements in vfssw array */
564 extern vfsops_t
*EIO_vfsops
; /* operations for vfs being torn-down */
567 * The following variables are private to the the kernel's vfs layer. File
568 * system implementations should not access them.
570 extern struct vfs
*rootvfs
; /* ptr to root vfs structure */
572 struct vfs
*rvfs_head
; /* head vfs in chain */
573 kmutex_t rvfs_lock
; /* mutex protecting this chain */
574 uint32_t rvfs_len
; /* length of this chain */
576 extern rvfs_t
*rvfs_list
;
577 extern int vfshsz
; /* # of elements in rvfs_head array */
578 extern const mntopts_t vfs_mntopts
; /* globally recognized options */
580 #endif /* defined(_KERNEL) */
582 #define VFS_HOLD(vfsp) { \
586 #define VFS_RELE(vfsp) { \
590 #define VFS_INIT(vfsp, op, data) { \
591 vfs_init((vfsp), (op), (data)); \
595 #define VFS_INSTALLED(vfsswp) (((vfsswp)->vsw_flag & VSW_INSTALLED) != 0)
596 #define ALLOCATED_VFSSW(vswp) ((vswp)->vsw_name[0] != '\0')
597 #define RLOCK_VFSSW() (rw_enter(&vfssw_lock, RW_READER))
598 #define RUNLOCK_VFSSW() (rw_exit(&vfssw_lock))
599 #define WLOCK_VFSSW() (rw_enter(&vfssw_lock, RW_WRITER))
600 #define WUNLOCK_VFSSW() (rw_exit(&vfssw_lock))
601 #define VFSSW_LOCKED() (RW_LOCK_HELD(&vfssw_lock))
602 #define VFSSW_WRITE_LOCKED() (RW_WRITE_HELD(&vfssw_lock))
606 #define SYNC_ATTR 0x01 /* sync attributes only */
607 #define SYNC_CLOSE 0x02 /* close open file */
608 #define SYNC_ALL 0x04 /* force to sync all fs */
614 #endif /* _SYS_VFS_H */