xref: /illumos-gate/usr/src/uts/common/sys/vnode.h (revision cd964fce751ca752af4158842063a9579a2d4331)
1 /*
2  * CDDL HEADER START
3  *
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.
7  *
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.
12  *
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]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright (c) 1988, 2010, Oracle and/or its affiliates. All rights reserved.
24  * Copyright (c) 2017, Joyent, Inc.
25  * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
26  */
27 
28 /*	Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T	*/
29 /*	  All Rights Reserved	*/
30 
31 /*
32  * University Copyright- Copyright (c) 1982, 1986, 1988
33  * The Regents of the University of California
34  * All Rights Reserved
35  *
36  * University Acknowledgment- Portions of this document are derived from
37  * software developed by the University of California, Berkeley, and its
38  * contributors.
39  */
40 
41 #ifndef _SYS_VNODE_H
42 #define	_SYS_VNODE_H
43 
44 #include <sys/types.h>
45 #include <sys/t_lock.h>
46 #include <sys/rwstlock.h>
47 #include <sys/time_impl.h>
48 #include <sys/cred.h>
49 #include <sys/uio.h>
50 #include <sys/resource.h>
51 #include <vm/seg_enum.h>
52 #include <sys/kstat.h>
53 #include <sys/kmem.h>
54 #include <sys/list.h>
55 #ifdef	_KERNEL
56 #include <sys/buf.h>
57 #include <sys/sdt.h>
58 #endif	/* _KERNEL */
59 
60 #ifdef	__cplusplus
61 extern "C" {
62 #endif
63 
64 /*
65  * Statistics for all vnode operations.
66  * All operations record number of ops (since boot/mount/zero'ed).
67  * Certain I/O operations (read, write, readdir) also record number
68  * of bytes transferred.
69  * This appears in two places in the system: one is embedded in each
70  * vfs_t.  There is also an array of vopstats_t structures allocated
71  * on a per-fstype basis.
72  */
73 
74 #define	VOPSTATS_STR	"vopstats_"	/* Initial string for vopstat kstats */
75 
76 typedef struct vopstats {
77 	kstat_named_t	nopen;		/* VOP_OPEN */
78 	kstat_named_t	nclose;		/* VOP_CLOSE */
79 	kstat_named_t	nread;		/* VOP_READ */
80 	kstat_named_t	read_bytes;
81 	kstat_named_t	nwrite;		/* VOP_WRITE */
82 	kstat_named_t	write_bytes;
83 	kstat_named_t	nioctl;		/* VOP_IOCTL */
84 	kstat_named_t	nsetfl;		/* VOP_SETFL */
85 	kstat_named_t	ngetattr;	/* VOP_GETATTR */
86 	kstat_named_t	nsetattr;	/* VOP_SETATTR */
87 	kstat_named_t	naccess;	/* VOP_ACCESS */
88 	kstat_named_t	nlookup;	/* VOP_LOOKUP */
89 	kstat_named_t	ncreate;	/* VOP_CREATE */
90 	kstat_named_t	nremove;	/* VOP_REMOVE */
91 	kstat_named_t	nlink;		/* VOP_LINK */
92 	kstat_named_t	nrename;	/* VOP_RENAME */
93 	kstat_named_t	nmkdir;		/* VOP_MKDIR */
94 	kstat_named_t	nrmdir;		/* VOP_RMDIR */
95 	kstat_named_t	nreaddir;	/* VOP_READDIR */
96 	kstat_named_t	readdir_bytes;
97 	kstat_named_t	nsymlink;	/* VOP_SYMLINK */
98 	kstat_named_t	nreadlink;	/* VOP_READLINK */
99 	kstat_named_t	nfsync;		/* VOP_FSYNC */
100 	kstat_named_t	ninactive;	/* VOP_INACTIVE */
101 	kstat_named_t	nfid;		/* VOP_FID */
102 	kstat_named_t	nrwlock;	/* VOP_RWLOCK */
103 	kstat_named_t	nrwunlock;	/* VOP_RWUNLOCK */
104 	kstat_named_t	nseek;		/* VOP_SEEK */
105 	kstat_named_t	ncmp;		/* VOP_CMP */
106 	kstat_named_t	nfrlock;	/* VOP_FRLOCK */
107 	kstat_named_t	nspace;		/* VOP_SPACE */
108 	kstat_named_t	nrealvp;	/* VOP_REALVP */
109 	kstat_named_t	ngetpage;	/* VOP_GETPAGE */
110 	kstat_named_t	nputpage;	/* VOP_PUTPAGE */
111 	kstat_named_t	nmap;		/* VOP_MAP */
112 	kstat_named_t	naddmap;	/* VOP_ADDMAP */
113 	kstat_named_t	ndelmap;	/* VOP_DELMAP */
114 	kstat_named_t	npoll;		/* VOP_POLL */
115 	kstat_named_t	ndump;		/* VOP_DUMP */
116 	kstat_named_t	npathconf;	/* VOP_PATHCONF */
117 	kstat_named_t	npageio;	/* VOP_PAGEIO */
118 	kstat_named_t	ndumpctl;	/* VOP_DUMPCTL */
119 	kstat_named_t	ndispose;	/* VOP_DISPOSE */
120 	kstat_named_t	nsetsecattr;	/* VOP_SETSECATTR */
121 	kstat_named_t	ngetsecattr;	/* VOP_GETSECATTR */
122 	kstat_named_t	nshrlock;	/* VOP_SHRLOCK */
123 	kstat_named_t	nvnevent;	/* VOP_VNEVENT */
124 	kstat_named_t	nreqzcbuf;	/* VOP_REQZCBUF */
125 	kstat_named_t	nretzcbuf;	/* VOP_RETZCBUF */
126 } vopstats_t;
127 
128 /*
129  * The vnode is the focus of all file activity in UNIX.
130  * A vnode is allocated for each active file, each current
131  * directory, each mounted-on file, and the root.
132  *
133  * Each vnode is usually associated with a file-system-specific node (for
134  * UFS, this is the in-memory inode).  Generally, a vnode and an fs-node
135  * should be created and destroyed together as a pair.
136  *
137  * If a vnode is reused for a new file, it should be reinitialized by calling
138  * either vn_reinit() or vn_recycle().
139  *
140  * vn_reinit() resets the entire vnode as if it was returned by vn_alloc().
141  * The caller is responsible for setting up the entire vnode after calling
142  * vn_reinit().  This is important when using kmem caching where the vnode is
143  * allocated by a constructor, for instance.
144  *
145  * vn_recycle() is used when the file system keeps some state around in both
146  * the vnode and the associated FS-node.  In UFS, for example, the inode of
147  * a deleted file can be reused immediately.  The v_data, v_vfsp, v_op, etc.
148  * remains the same but certain fields related to the previous instance need
149  * to be reset.  In particular:
150  *	v_femhead
151  *	v_path
152  *	v_rdcnt, v_wrcnt
153  *	v_mmap_read, v_mmap_write
154  */
155 
156 /*
157  * vnode types.  VNON means no type.  These values are unrelated to
158  * values in on-disk inodes.
159  */
160 typedef enum vtype {
161 	VNON	= 0,
162 	VREG	= 1,
163 	VDIR	= 2,
164 	VBLK	= 3,
165 	VCHR	= 4,
166 	VLNK	= 5,
167 	VFIFO	= 6,
168 	VDOOR	= 7,
169 	VPROC	= 8,
170 	VSOCK	= 9,
171 	VPORT	= 10,
172 	VBAD	= 11
173 } vtype_t;
174 
175 /*
176  * VSD - Vnode Specific Data
177  * Used to associate additional private data with a vnode.
178  */
179 struct vsd_node {
180 	list_node_t vs_nodes;		/* list of all VSD nodes */
181 	uint_t vs_nkeys;		/* entries in value array */
182 	void **vs_value;		/* array of value/key */
183 };
184 
185 /*
186  * Many of the fields in the vnode are read-only once they are initialized
187  * at vnode creation time.  Other fields are protected by locks.
188  *
189  * IMPORTANT: vnodes should be created ONLY by calls to vn_alloc().  They
190  * may not be embedded into the file-system specific node (inode).  The
191  * size of vnodes may change.
192  *
193  * The v_lock protects:
194  *   v_flag
195  *   v_stream
196  *   v_count
197  *   v_shrlocks
198  *   v_path
199  *   v_vsd
200  *   v_xattrdir
201  *
202  * A special lock (implemented by vn_vfswlock in vnode.c) protects:
203  *   v_vfsmountedhere
204  *
205  * The global flock_lock mutex (in flock.c) protects:
206  *   v_filocks
207  *
208  * IMPORTANT NOTE:
209  *
210  *   The following vnode fields are considered public and may safely be
211  *   accessed by file systems or other consumers:
212  *
213  *     v_lock
214  *     v_flag
215  *     v_count
216  *     v_data
217  *     v_vfsp
218  *     v_stream
219  *     v_type
220  *     v_rdev
221  *
222  * ALL OTHER FIELDS SHOULD BE ACCESSED ONLY BY THE OWNER OF THAT FIELD.
223  * In particular, file systems should not access other fields; they may
224  * change or even be removed.  The functionality which was once provided
225  * by these fields is available through vn_* functions.
226  *
227  * VNODE PATH THEORY:
228  * In each vnode, the v_path field holds a cached version of the canonical
229  * filesystem path which that node represents.  Because vnodes lack contextual
230  * information about their own name or position in the VFS hierarchy, this path
231  * must be calculated when the vnode is instantiated by operations such as
232  * fop_create, fop_lookup, or fop_mkdir.  During said operations, both the
233  * parent vnode (and its cached v_path) and future name are known, so the
234  * v_path of the resulting object can easily be set.
235  *
236  * The caching nature of v_path is complicated in the face of directory
237  * renames.  Filesystem drivers are responsible for calling vn_renamepath when
238  * a fop_rename operation succeeds.  While the v_path on the renamed vnode will
239  * be updated, existing children of the directory (direct, or at deeper levels)
240  * will now possess v_path caches which are stale.
241  *
242  * It is expensive (and for non-directories, impossible) to recalculate stale
243  * v_path entries during operations such as vnodetopath.  The best time during
244  * which to correct such wrongs is the same as when v_path is first
245  * initialized: during fop_create/fop_lookup/fop_mkdir/etc, where adequate
246  * context is available to generate the current path.
247  *
248  * In order to quickly detect stale v_path entries (without full lookup
249  * verification) to trigger a v_path update, the v_path_stamp field has been
250  * added to vnode_t.  As part of successful fop_create/fop_lookup/fop_mkdir
251  * operations, where the name and parent vnode are available, the following
252  * rules are used to determine updates to the child:
253  *
254  * 1. If the parent lacks a v_path, clear any existing v_path and v_path_stamp
255  *    on the child.  Until the parent v_path is refreshed to a valid state, the
256  *    child v_path must be considered invalid too.
257  *
258  * 2. If the child lacks a v_path (implying v_path_stamp == 0), it inherits the
259  *    v_path_stamp value from its parent and its v_path is updated.
260  *
261  * 3. If the child v_path_stamp is less than v_path_stamp in the parent, it is
262  *    an indication that the child v_path is stale.  The v_path is updated and
263  *    v_path_stamp in the child is set to the current hrtime().
264  *
265  *    It does _not_ inherit the parent v_path_stamp in order to propagate the
266  *    the time of v_path invalidation through the directory structure.  This
267  *    prevents concurrent invalidations (operating with a now-incorrect v_path)
268  *    at deeper levels in the tree from persisting.
269  *
270  * 4. If the child v_path_stamp is greater or equal to the parent, no action
271  *    needs to be taken.
272  *
273  * Note that fop_rename operations do not follow this ruleset.  They perform an
274  * explicit update of v_path and v_path_stamp (setting it to the current time)
275  *
276  * With these constraints in place, v_path invalidations and updates should
277  * proceed in a timely manner as vnodes are accessed.  While there still are
278  * limited cases where vnodetopath operations will fail, the risk is minimized.
279  */
280 
281 struct fem_head;	/* from fem.h */
282 
283 typedef struct vnode {
284 	kmutex_t	v_lock;		/* protects vnode fields */
285 	uint_t		v_flag;		/* vnode flags (see below) */
286 	uint_t		v_count;	/* reference count */
287 	void		*v_data;	/* private data for fs */
288 	struct vfs	*v_vfsp;	/* ptr to containing VFS */
289 	struct stdata	*v_stream;	/* associated stream */
290 	enum vtype	v_type;		/* vnode type */
291 	dev_t		v_rdev;		/* device (VCHR, VBLK) */
292 
293 	/* PRIVATE FIELDS BELOW - DO NOT USE */
294 
295 	struct vfs	*v_vfsmountedhere; /* ptr to vfs mounted here */
296 	struct vnodeops	*v_op;		/* vnode operations */
297 	struct page	*v_pages;	/* vnode pages list */
298 	struct filock	*v_filocks;	/* ptr to filock list */
299 	struct shrlocklist *v_shrlocks;	/* ptr to shrlock list */
300 	krwlock_t	v_nbllock;	/* sync for NBMAND locks */
301 	kcondvar_t	v_cv;		/* synchronize locking */
302 	void		*v_locality;	/* hook for locality info */
303 	struct fem_head	*v_femhead;	/* fs monitoring */
304 	char		*v_path;	/* cached path */
305 	hrtime_t	v_path_stamp;	/* timestamp for cached path */
306 	uint_t		v_rdcnt;	/* open for read count  (VREG only) */
307 	uint_t		v_wrcnt;	/* open for write count (VREG only) */
308 	u_longlong_t	v_mmap_read;	/* mmap read count */
309 	u_longlong_t	v_mmap_write;	/* mmap write count */
310 	void		*v_mpssdata;	/* info for large page mappings */
311 	void		*v_fopdata;	/* list of file ops event watches */
312 	kmutex_t	v_vsd_lock;	/* protects v_vsd field */
313 	struct vsd_node *v_vsd;		/* vnode specific data */
314 	struct vnode	*v_xattrdir;	/* unnamed extended attr dir (GFS) */
315 	uint_t		v_count_dnlc;	/* dnlc reference count */
316 } vnode_t;
317 
318 #define	IS_DEVVP(vp)	\
319 	((vp)->v_type == VCHR || (vp)->v_type == VBLK || (vp)->v_type == VFIFO)
320 
321 #define	VNODE_ALIGN	64
322 /* Count of low-order 0 bits in a vnode *, based on size and alignment. */
323 #if defined(_LP64)
324 #define	VNODE_ALIGN_LOG2	8
325 #else
326 #define	VNODE_ALIGN_LOG2	7
327 #endif
328 
329 /*
330  * vnode flags.
331  */
332 #define	VROOT		0x01	/* root of its file system */
333 #define	VNOCACHE	0x02	/* don't keep cache pages on vnode */
334 #define	VNOMAP		0x04	/* file cannot be mapped/faulted */
335 #define	VDUP		0x08	/* file should be dup'ed rather then opened */
336 #define	VNOSWAP		0x10	/* file cannot be used as virtual swap device */
337 #define	VNOMOUNT	0x20	/* file cannot be covered by mount */
338 #define	VISSWAP		0x40	/* vnode is being used for swap */
339 #define	VSWAPLIKE	0x80	/* vnode acts like swap (but may not be) */
340 
341 #define	IS_SWAPVP(vp)	(((vp)->v_flag & (VISSWAP | VSWAPLIKE)) != 0)
342 
343 typedef struct vn_vfslocks_entry {
344 	rwstlock_t ve_lock;
345 	void *ve_vpvfs;
346 	struct vn_vfslocks_entry *ve_next;
347 	uint32_t ve_refcnt;
348 	char pad[64 - sizeof (rwstlock_t) - 2 * sizeof (void *) - \
349 	    sizeof (uint32_t)];
350 } vn_vfslocks_entry_t;
351 
352 /*
353  * The following two flags are used to lock the v_vfsmountedhere field
354  */
355 #define	VVFSLOCK	0x100
356 #define	VVFSWAIT	0x200
357 
358 /*
359  * Used to serialize VM operations on a vnode
360  */
361 #define	VVMLOCK		0x400
362 
363 /*
364  * Tell vn_open() not to fail a directory open for writing but
365  * to go ahead and call VOP_OPEN() to let the filesystem check.
366  */
367 #define	VDIROPEN	0x800
368 
369 /*
370  * Flag to let the VM system know that this file is most likely a binary
371  * or shared library since it has been mmap()ed EXEC at some time.
372  */
373 #define	VVMEXEC		0x1000
374 
375 #define	VPXFS		0x2000  /* clustering: global fs proxy vnode */
376 
377 #define	IS_PXFSVP(vp)	((vp)->v_flag & VPXFS)
378 
379 #define	V_XATTRDIR	0x4000	/* attribute unnamed directory */
380 
381 #define	IS_XATTRDIR(vp)	((vp)->v_flag & V_XATTRDIR)
382 
383 #define	V_LOCALITY	0x8000	/* whether locality aware */
384 
385 /*
386  * Flag that indicates the VM should maintain the v_pages list with all modified
387  * pages on one end and unmodified pages at the other. This makes finding dirty
388  * pages to write back to disk much faster at the expense of taking a minor
389  * fault on the first store instruction which touches a writable page.
390  */
391 #define	VMODSORT	(0x10000)
392 #define	IS_VMODSORT(vp) \
393 	(pvn_vmodsort_supported != 0 && ((vp)->v_flag  & VMODSORT) != 0)
394 
395 #define	VISSWAPFS	0x20000	/* vnode is being used for swapfs */
396 
397 /*
398  * The mdb memstat command assumes that IS_SWAPFSVP only uses the
399  * vnode's v_flag field.  If this changes, cache the additional
400  * fields in mdb; see vn_get in mdb/common/modules/genunix/memory.c
401  */
402 #define	IS_SWAPFSVP(vp)	(((vp)->v_flag & VISSWAPFS) != 0)
403 
404 #define	V_SYSATTR	0x40000	/* vnode is a GFS system attribute */
405 
406 /*
407  * Indication that VOP_LOOKUP operations on this vnode may yield results from a
408  * different VFS instance.  The main use of this is to suppress v_path
409  * calculation logic when filesystems such as procfs emit results which defy
410  * expectations about normal VFS behavior.
411  */
412 #define	VTRAVERSE	0x80000
413 
414 /*
415  * Vnode attributes.  A bit-mask is supplied as part of the
416  * structure to indicate the attributes the caller wants to
417  * set (setattr) or extract (getattr).
418  */
419 
420 /*
421  * Note that va_nodeid and va_nblocks are 64bit data type.
422  * We support large files over NFSV3. With Solaris client and
423  * Server that generates 64bit ino's and sizes these fields
424  * will overflow if they are 32 bit sizes.
425  */
426 
427 typedef struct vattr {
428 	uint_t		va_mask;	/* bit-mask of attributes */
429 	vtype_t		va_type;	/* vnode type (for create) */
430 	mode_t		va_mode;	/* file access mode */
431 	uid_t		va_uid;		/* owner user id */
432 	gid_t		va_gid;		/* owner group id */
433 	dev_t		va_fsid;	/* file system id (dev for now) */
434 	u_longlong_t	va_nodeid;	/* node id */
435 	nlink_t		va_nlink;	/* number of references to file */
436 	u_offset_t	va_size;	/* file size in bytes */
437 	timestruc_t	va_atime;	/* time of last access */
438 	timestruc_t	va_mtime;	/* time of last modification */
439 	timestruc_t	va_ctime;	/* time of last status change */
440 	dev_t		va_rdev;	/* device the file represents */
441 	uint_t		va_blksize;	/* fundamental block size */
442 	u_longlong_t	va_nblocks;	/* # of blocks allocated */
443 	uint_t		va_seq;		/* sequence number */
444 } vattr_t;
445 
446 #define	AV_SCANSTAMP_SZ	32		/* length of anti-virus scanstamp */
447 
448 /*
449  * Structure of all optional attributes.
450  */
451 typedef struct xoptattr {
452 	timestruc_t	xoa_createtime;	/* Create time of file */
453 	uint8_t		xoa_archive;
454 	uint8_t		xoa_system;
455 	uint8_t		xoa_readonly;
456 	uint8_t		xoa_hidden;
457 	uint8_t		xoa_nounlink;
458 	uint8_t		xoa_immutable;
459 	uint8_t		xoa_appendonly;
460 	uint8_t		xoa_nodump;
461 	uint8_t		xoa_opaque;
462 	uint8_t		xoa_av_quarantined;
463 	uint8_t		xoa_av_modified;
464 	uint8_t		xoa_av_scanstamp[AV_SCANSTAMP_SZ];
465 	uint8_t		xoa_reparse;
466 	uint64_t	xoa_generation;
467 	uint8_t		xoa_offline;
468 	uint8_t		xoa_sparse;
469 } xoptattr_t;
470 
471 /*
472  * The xvattr structure is really a variable length structure that
473  * is made up of:
474  * - The classic vattr_t (xva_vattr)
475  * - a 32 bit quantity (xva_mapsize) that specifies the size of the
476  *   attribute bitmaps in 32 bit words.
477  * - A pointer to the returned attribute bitmap (needed because the
478  *   previous element, the requested attribute bitmap) is variable lenth.
479  * - The requested attribute bitmap, which is an array of 32 bit words.
480  *   Callers use the XVA_SET_REQ() macro to set the bits corresponding to
481  *   the attributes that are being requested.
482  * - The returned attribute bitmap, which is an array of 32 bit words.
483  *   File systems that support optional attributes use the XVA_SET_RTN()
484  *   macro to set the bits corresponding to the attributes that are being
485  *   returned.
486  * - The xoptattr_t structure which contains the attribute values
487  *
488  * xva_mapsize determines how many words in the attribute bitmaps.
489  * Immediately following the attribute bitmaps is the xoptattr_t.
490  * xva_getxoptattr() is used to get the pointer to the xoptattr_t
491  * section.
492  */
493 
494 #define	XVA_MAPSIZE	3		/* Size of attr bitmaps */
495 #define	XVA_MAGIC	0x78766174	/* Magic # for verification */
496 
497 /*
498  * The xvattr structure is an extensible structure which permits optional
499  * attributes to be requested/returned.  File systems may or may not support
500  * optional attributes.  They do so at their own discretion but if they do
501  * support optional attributes, they must register the VFSFT_XVATTR feature
502  * so that the optional attributes can be set/retrived.
503  *
504  * The fields of the xvattr structure are:
505  *
506  * xva_vattr - The first element of an xvattr is a legacy vattr structure
507  * which includes the common attributes.  If AT_XVATTR is set in the va_mask
508  * then the entire structure is treated as an xvattr.  If AT_XVATTR is not
509  * set, then only the xva_vattr structure can be used.
510  *
511  * xva_magic - 0x78766174 (hex for "xvat"). Magic number for verification.
512  *
513  * xva_mapsize - Size of requested and returned attribute bitmaps.
514  *
515  * xva_rtnattrmapp - Pointer to xva_rtnattrmap[].  We need this since the
516  * size of the array before it, xva_reqattrmap[], could change which means
517  * the location of xva_rtnattrmap[] could change.  This will allow unbundled
518  * file systems to find the location of xva_rtnattrmap[] when the sizes change.
519  *
520  * xva_reqattrmap[] - Array of requested attributes.  Attributes are
521  * represented by a specific bit in a specific element of the attribute
522  * map array.  Callers set the bits corresponding to the attributes
523  * that the caller wants to get/set.
524  *
525  * xva_rtnattrmap[] - Array of attributes that the file system was able to
526  * process.  Not all file systems support all optional attributes.  This map
527  * informs the caller which attributes the underlying file system was able
528  * to set/get.  (Same structure as the requested attributes array in terms
529  * of each attribute  corresponding to specific bits and array elements.)
530  *
531  * xva_xoptattrs - Structure containing values of optional attributes.
532  * These values are only valid if the corresponding bits in xva_reqattrmap
533  * are set and the underlying file system supports those attributes.
534  */
535 typedef struct xvattr {
536 	vattr_t		xva_vattr;	/* Embedded vattr structure */
537 	uint32_t	xva_magic;	/* Magic Number */
538 	uint32_t	xva_mapsize;	/* Size of attr bitmap (32-bit words) */
539 	uint32_t	*xva_rtnattrmapp;	/* Ptr to xva_rtnattrmap[] */
540 	uint32_t	xva_reqattrmap[XVA_MAPSIZE];	/* Requested attrs */
541 	uint32_t	xva_rtnattrmap[XVA_MAPSIZE];	/* Returned attrs */
542 	xoptattr_t	xva_xoptattrs;	/* Optional attributes */
543 } xvattr_t;
544 
545 #ifdef _SYSCALL32
546 /*
547  * For bigtypes time_t changed to 64 bit on the 64-bit kernel.
548  * Define an old version for user/kernel interface
549  */
550 
551 #if _LONG_LONG_ALIGNMENT == 8 && _LONG_LONG_ALIGNMENT_32 == 4
552 #pragma pack(4)
553 #endif
554 
555 typedef struct vattr32 {
556 	uint32_t	va_mask;	/* bit-mask of attributes */
557 	vtype_t		va_type;	/* vnode type (for create) */
558 	mode32_t	va_mode;	/* file access mode */
559 	uid32_t		va_uid;		/* owner user id */
560 	gid32_t		va_gid;		/* owner group id */
561 	dev32_t		va_fsid;	/* file system id (dev for now) */
562 	u_longlong_t	va_nodeid;	/* node id */
563 	nlink_t		va_nlink;	/* number of references to file */
564 	u_offset_t	va_size;	/* file size in bytes */
565 	timestruc32_t	va_atime;	/* time of last access */
566 	timestruc32_t	va_mtime;	/* time of last modification */
567 	timestruc32_t	va_ctime;	/* time of last status change */
568 	dev32_t		va_rdev;	/* device the file represents */
569 	uint32_t	va_blksize;	/* fundamental block size */
570 	u_longlong_t	va_nblocks;	/* # of blocks allocated */
571 	uint32_t	va_seq;		/* sequence number */
572 } vattr32_t;
573 
574 #if _LONG_LONG_ALIGNMENT == 8 && _LONG_LONG_ALIGNMENT_32 == 4
575 #pragma pack()
576 #endif
577 
578 #else  /* not _SYSCALL32 */
579 #define	vattr32		vattr
580 typedef vattr_t		vattr32_t;
581 #endif /* _SYSCALL32 */
582 
583 /*
584  * Attributes of interest to the caller of setattr or getattr.
585  */
586 #define	AT_TYPE		0x00001
587 #define	AT_MODE		0x00002
588 #define	AT_UID		0x00004
589 #define	AT_GID		0x00008
590 #define	AT_FSID		0x00010
591 #define	AT_NODEID	0x00020
592 #define	AT_NLINK	0x00040
593 #define	AT_SIZE		0x00080
594 #define	AT_ATIME	0x00100
595 #define	AT_MTIME	0x00200
596 #define	AT_CTIME	0x00400
597 #define	AT_RDEV		0x00800
598 #define	AT_BLKSIZE	0x01000
599 #define	AT_NBLOCKS	0x02000
600 /*			0x04000 */	/* unused */
601 #define	AT_SEQ		0x08000
602 /*
603  * If AT_XVATTR is set then there are additional bits to process in
604  * the xvattr_t's attribute bitmap.  If this is not set then the bitmap
605  * MUST be ignored.  Note that this bit must be set/cleared explicitly.
606  * That is, setting AT_ALL will NOT set AT_XVATTR.
607  */
608 #define	AT_XVATTR	0x10000
609 
610 #define	AT_ALL		(AT_TYPE|AT_MODE|AT_UID|AT_GID|AT_FSID|AT_NODEID|\
611 			AT_NLINK|AT_SIZE|AT_ATIME|AT_MTIME|AT_CTIME|\
612 			AT_RDEV|AT_BLKSIZE|AT_NBLOCKS|AT_SEQ)
613 
614 #define	AT_STAT		(AT_MODE|AT_UID|AT_GID|AT_FSID|AT_NODEID|AT_NLINK|\
615 			AT_SIZE|AT_ATIME|AT_MTIME|AT_CTIME|AT_RDEV|AT_TYPE)
616 
617 #define	AT_TIMES	(AT_ATIME|AT_MTIME|AT_CTIME)
618 
619 #define	AT_NOSET	(AT_NLINK|AT_RDEV|AT_FSID|AT_NODEID|AT_TYPE|\
620 			AT_BLKSIZE|AT_NBLOCKS|AT_SEQ)
621 
622 /*
623  * Attribute bits used in the extensible attribute's (xva's) attribute
624  * bitmaps.  Note that the bitmaps are made up of a variable length number
625  * of 32-bit words.  The convention is to use XAT{n}_{attrname} where "n"
626  * is the element in the bitmap (starting at 1).  This convention is for
627  * the convenience of the maintainer to keep track of which element each
628  * attribute belongs to.
629  *
630  * NOTE THAT CONSUMERS MUST *NOT* USE THE XATn_* DEFINES DIRECTLY.  CONSUMERS
631  * MUST USE THE XAT_* DEFINES.
632  */
633 #define	XAT0_INDEX	0LL		/* Index into bitmap for XAT0 attrs */
634 #define	XAT0_CREATETIME	0x00000001	/* Create time of file */
635 #define	XAT0_ARCHIVE	0x00000002	/* Archive */
636 #define	XAT0_SYSTEM	0x00000004	/* System */
637 #define	XAT0_READONLY	0x00000008	/* Readonly */
638 #define	XAT0_HIDDEN	0x00000010	/* Hidden */
639 #define	XAT0_NOUNLINK	0x00000020	/* Nounlink */
640 #define	XAT0_IMMUTABLE	0x00000040	/* immutable */
641 #define	XAT0_APPENDONLY	0x00000080	/* appendonly */
642 #define	XAT0_NODUMP	0x00000100	/* nodump */
643 #define	XAT0_OPAQUE	0x00000200	/* opaque */
644 #define	XAT0_AV_QUARANTINED	0x00000400	/* anti-virus quarantine */
645 #define	XAT0_AV_MODIFIED	0x00000800	/* anti-virus modified */
646 #define	XAT0_AV_SCANSTAMP	0x00001000	/* anti-virus scanstamp */
647 #define	XAT0_REPARSE	0x00002000	/* FS reparse point */
648 #define	XAT0_GEN	0x00004000	/* object generation number */
649 #define	XAT0_OFFLINE	0x00008000	/* offline */
650 #define	XAT0_SPARSE	0x00010000	/* sparse */
651 
652 #define	XAT0_ALL_ATTRS	(XAT0_CREATETIME|XAT0_ARCHIVE|XAT0_SYSTEM| \
653     XAT0_READONLY|XAT0_HIDDEN|XAT0_NOUNLINK|XAT0_IMMUTABLE|XAT0_APPENDONLY| \
654     XAT0_NODUMP|XAT0_OPAQUE|XAT0_AV_QUARANTINED|  XAT0_AV_MODIFIED| \
655     XAT0_AV_SCANSTAMP|XAT0_REPARSE|XATO_GEN|XAT0_OFFLINE|XAT0_SPARSE)
656 
657 /* Support for XAT_* optional attributes */
658 #define	XVA_MASK		0xffffffff	/* Used to mask off 32 bits */
659 #define	XVA_SHFT		32		/* Used to shift index */
660 
661 /*
662  * Used to pry out the index and attribute bits from the XAT_* attributes
663  * defined below.  Note that we're masking things down to 32 bits then
664  * casting to uint32_t.
665  */
666 #define	XVA_INDEX(attr)		((uint32_t)(((attr) >> XVA_SHFT) & XVA_MASK))
667 #define	XVA_ATTRBIT(attr)	((uint32_t)((attr) & XVA_MASK))
668 
669 /*
670  * The following defines present a "flat namespace" so that consumers don't
671  * need to keep track of which element belongs to which bitmap entry.
672  *
673  * NOTE THAT THESE MUST NEVER BE OR-ed TOGETHER
674  */
675 #define	XAT_CREATETIME		((XAT0_INDEX << XVA_SHFT) | XAT0_CREATETIME)
676 #define	XAT_ARCHIVE		((XAT0_INDEX << XVA_SHFT) | XAT0_ARCHIVE)
677 #define	XAT_SYSTEM		((XAT0_INDEX << XVA_SHFT) | XAT0_SYSTEM)
678 #define	XAT_READONLY		((XAT0_INDEX << XVA_SHFT) | XAT0_READONLY)
679 #define	XAT_HIDDEN		((XAT0_INDEX << XVA_SHFT) | XAT0_HIDDEN)
680 #define	XAT_NOUNLINK		((XAT0_INDEX << XVA_SHFT) | XAT0_NOUNLINK)
681 #define	XAT_IMMUTABLE		((XAT0_INDEX << XVA_SHFT) | XAT0_IMMUTABLE)
682 #define	XAT_APPENDONLY		((XAT0_INDEX << XVA_SHFT) | XAT0_APPENDONLY)
683 #define	XAT_NODUMP		((XAT0_INDEX << XVA_SHFT) | XAT0_NODUMP)
684 #define	XAT_OPAQUE		((XAT0_INDEX << XVA_SHFT) | XAT0_OPAQUE)
685 #define	XAT_AV_QUARANTINED	((XAT0_INDEX << XVA_SHFT) | XAT0_AV_QUARANTINED)
686 #define	XAT_AV_MODIFIED		((XAT0_INDEX << XVA_SHFT) | XAT0_AV_MODIFIED)
687 #define	XAT_AV_SCANSTAMP	((XAT0_INDEX << XVA_SHFT) | XAT0_AV_SCANSTAMP)
688 #define	XAT_REPARSE		((XAT0_INDEX << XVA_SHFT) | XAT0_REPARSE)
689 #define	XAT_GEN			((XAT0_INDEX << XVA_SHFT) | XAT0_GEN)
690 #define	XAT_OFFLINE		((XAT0_INDEX << XVA_SHFT) | XAT0_OFFLINE)
691 #define	XAT_SPARSE		((XAT0_INDEX << XVA_SHFT) | XAT0_SPARSE)
692 
693 /*
694  * The returned attribute map array (xva_rtnattrmap[]) is located past the
695  * requested attribute map array (xva_reqattrmap[]).  Its location changes
696  * when the array sizes change.  We use a separate pointer in a known location
697  * (xva_rtnattrmapp) to hold the location of xva_rtnattrmap[].  This is
698  * set in xva_init()
699  */
700 #define	XVA_RTNATTRMAP(xvap)	((xvap)->xva_rtnattrmapp)
701 
702 /*
703  * XVA_SET_REQ() sets an attribute bit in the proper element in the bitmap
704  * of requested attributes (xva_reqattrmap[]).
705  */
706 #define	XVA_SET_REQ(xvap, attr)					\
707 	ASSERT((xvap)->xva_vattr.va_mask | AT_XVATTR);		\
708 	ASSERT((xvap)->xva_magic == XVA_MAGIC);			\
709 	(xvap)->xva_reqattrmap[XVA_INDEX(attr)] |= XVA_ATTRBIT(attr)
710 /*
711  * XVA_CLR_REQ() clears an attribute bit in the proper element in the bitmap
712  * of requested attributes (xva_reqattrmap[]).
713  */
714 #define	XVA_CLR_REQ(xvap, attr)					\
715 	ASSERT((xvap)->xva_vattr.va_mask | AT_XVATTR);		\
716 	ASSERT((xvap)->xva_magic == XVA_MAGIC);			\
717 	(xvap)->xva_reqattrmap[XVA_INDEX(attr)] &= ~XVA_ATTRBIT(attr)
718 
719 /*
720  * XVA_SET_RTN() sets an attribute bit in the proper element in the bitmap
721  * of returned attributes (xva_rtnattrmap[]).
722  */
723 #define	XVA_SET_RTN(xvap, attr)					\
724 	ASSERT((xvap)->xva_vattr.va_mask | AT_XVATTR);		\
725 	ASSERT((xvap)->xva_magic == XVA_MAGIC);			\
726 	(XVA_RTNATTRMAP(xvap))[XVA_INDEX(attr)] |= XVA_ATTRBIT(attr)
727 
728 /*
729  * XVA_ISSET_REQ() checks the requested attribute bitmap (xva_reqattrmap[])
730  * to see of the corresponding attribute bit is set.  If so, returns non-zero.
731  */
732 #define	XVA_ISSET_REQ(xvap, attr)					\
733 	((((xvap)->xva_vattr.va_mask | AT_XVATTR) &&			\
734 		((xvap)->xva_magic == XVA_MAGIC) &&			\
735 		((xvap)->xva_mapsize > XVA_INDEX(attr))) ?		\
736 	((xvap)->xva_reqattrmap[XVA_INDEX(attr)] & XVA_ATTRBIT(attr)) :	0)
737 
738 /*
739  * XVA_ISSET_RTN() checks the returned attribute bitmap (xva_rtnattrmap[])
740  * to see of the corresponding attribute bit is set.  If so, returns non-zero.
741  */
742 #define	XVA_ISSET_RTN(xvap, attr)					\
743 	((((xvap)->xva_vattr.va_mask | AT_XVATTR) &&			\
744 		((xvap)->xva_magic == XVA_MAGIC) &&			\
745 		((xvap)->xva_mapsize > XVA_INDEX(attr))) ?		\
746 	((XVA_RTNATTRMAP(xvap))[XVA_INDEX(attr)] & XVA_ATTRBIT(attr)) : 0)
747 
748 /*
749  *  Modes.  Some values same as S_xxx entries from stat.h for convenience.
750  */
751 #define	VSUID		04000		/* set user id on execution */
752 #define	VSGID		02000		/* set group id on execution */
753 #define	VSVTX		01000		/* save swapped text even after use */
754 
755 /*
756  * Permissions.
757  */
758 #define	VREAD		00400
759 #define	VWRITE		00200
760 #define	VEXEC		00100
761 
762 #define	MODEMASK	07777		/* mode bits plus permission bits */
763 #define	PERMMASK	00777		/* permission bits */
764 
765 /*
766  * VOP_ACCESS flags
767  */
768 #define	V_ACE_MASK	0x1	/* mask represents  NFSv4 ACE permissions */
769 #define	V_APPEND	0x2	/* want to do append only check */
770 
771 /*
772  * Check whether mandatory file locking is enabled.
773  */
774 
775 #define	MANDMODE(mode)		(((mode) & (VSGID|(VEXEC>>3))) == VSGID)
776 #define	MANDLOCK(vp, mode)	((vp)->v_type == VREG && MANDMODE(mode))
777 
778 /*
779  * Flags for vnode operations.
780  */
781 enum rm		{ RMFILE, RMDIRECTORY };	/* rm or rmdir (remove) */
782 enum symfollow	{ NO_FOLLOW, FOLLOW };		/* follow symlinks (or not) */
783 enum vcexcl	{ NONEXCL, EXCL };		/* (non)excl create */
784 enum create	{ CRCREAT, CRMKNOD, CRMKDIR };	/* reason for create */
785 
786 typedef enum rm		rm_t;
787 typedef enum symfollow	symfollow_t;
788 typedef enum vcexcl	vcexcl_t;
789 typedef enum create	create_t;
790 
791 /*
792  * Vnode Events - Used by VOP_VNEVENT
793  * The VE_PRE_RENAME_* events fire before the rename operation and are
794  * primarily used for specialized applications, such as NFSv4 delegation, which
795  * need to know about rename before it occurs.
796  */
797 typedef enum vnevent	{
798 	VE_SUPPORT	= 0,	/* Query */
799 	VE_RENAME_SRC	= 1,	/* Rename, with vnode as source */
800 	VE_RENAME_DEST	= 2,	/* Rename, with vnode as target/destination */
801 	VE_REMOVE	= 3,	/* Remove of vnode's name */
802 	VE_RMDIR	= 4,	/* Remove of directory vnode's name */
803 	VE_CREATE	= 5,	/* Create with vnode's name which exists */
804 	VE_LINK		= 6, 	/* Link with vnode's name as source */
805 	VE_RENAME_DEST_DIR	= 7, 	/* Rename with vnode as target dir */
806 	VE_MOUNTEDOVER	= 8, 	/* File or Filesystem got mounted over vnode */
807 	VE_TRUNCATE = 9,	/* Truncate */
808 	VE_PRE_RENAME_SRC = 10,	/* Pre-rename, with vnode as source */
809 	VE_PRE_RENAME_DEST = 11, /* Pre-rename, with vnode as target/dest. */
810 	VE_PRE_RENAME_DEST_DIR = 12 /* Pre-rename with vnode as target dir */
811 } vnevent_t;
812 
813 /*
814  * Values for checking vnode open and map counts
815  */
816 enum v_mode { V_READ, V_WRITE, V_RDORWR, V_RDANDWR };
817 
818 typedef enum v_mode v_mode_t;
819 
820 #define	V_TRUE	1
821 #define	V_FALSE	0
822 
823 /*
824  * Structure used on VOP_GETSECATTR and VOP_SETSECATTR operations
825  */
826 
827 typedef struct vsecattr {
828 	uint_t		vsa_mask;	/* See below */
829 	int		vsa_aclcnt;	/* ACL entry count */
830 	void		*vsa_aclentp;	/* pointer to ACL entries */
831 	int		vsa_dfaclcnt;	/* default ACL entry count */
832 	void		*vsa_dfaclentp;	/* pointer to default ACL entries */
833 	size_t		vsa_aclentsz;	/* ACE size in bytes of vsa_aclentp */
834 	uint_t		vsa_aclflags;	/* ACE ACL flags */
835 } vsecattr_t;
836 
837 /* vsa_mask values */
838 #define	VSA_ACL			0x0001
839 #define	VSA_ACLCNT		0x0002
840 #define	VSA_DFACL		0x0004
841 #define	VSA_DFACLCNT		0x0008
842 #define	VSA_ACE			0x0010
843 #define	VSA_ACECNT		0x0020
844 #define	VSA_ACE_ALLTYPES	0x0040
845 #define	VSA_ACE_ACLFLAGS	0x0080	/* get/set ACE ACL flags */
846 
847 /*
848  * Structure used by various vnode operations to determine
849  * the context (pid, host, identity) of a caller.
850  *
851  * The cc_caller_id is used to identify one or more callers who invoke
852  * operations, possibly on behalf of others.  For example, the NFS
853  * server could have it's own cc_caller_id which can be detected by
854  * vnode/vfs operations or (FEM) monitors on those operations.  New
855  * caller IDs are generated by fs_new_caller_id().
856  */
857 typedef struct caller_context {
858 	pid_t		cc_pid;		/* Process ID of the caller */
859 	int		cc_sysid;	/* System ID, used for remote calls */
860 	u_longlong_t	cc_caller_id;	/* Identifier for (set of) caller(s) */
861 	ulong_t		cc_flags;
862 } caller_context_t;
863 
864 /*
865  * Flags for caller context.  The caller sets CC_DONTBLOCK if it does not
866  * want to block inside of a FEM monitor.  The monitor will set CC_WOULDBLOCK
867  * and return EAGAIN if the operation would have blocked.
868  */
869 #define	CC_WOULDBLOCK	0x01
870 #define	CC_DONTBLOCK	0x02
871 
872 /*
873  * Structure tags for function prototypes, defined elsewhere.
874  */
875 struct pathname;
876 struct fid;
877 struct flock64;
878 struct flk_callback;
879 struct shrlock;
880 struct page;
881 struct seg;
882 struct as;
883 struct pollhead;
884 struct taskq;
885 
886 #ifdef	_KERNEL
887 
888 /*
889  * VNODE_OPS defines all the vnode operations.  It is used to define
890  * the vnodeops structure (below) and the fs_func_p union (vfs_opreg.h).
891  */
892 #define	VNODE_OPS							\
893 	int	(*vop_open)(vnode_t **, int, cred_t *,			\
894 				caller_context_t *);			\
895 	int	(*vop_close)(vnode_t *, int, int, offset_t, cred_t *,	\
896 				caller_context_t *);			\
897 	int	(*vop_read)(vnode_t *, uio_t *, int, cred_t *,		\
898 				caller_context_t *);			\
899 	int	(*vop_write)(vnode_t *, uio_t *, int, cred_t *,		\
900 				caller_context_t *);			\
901 	int	(*vop_ioctl)(vnode_t *, int, intptr_t, int, cred_t *,	\
902 				int *, caller_context_t *);		\
903 	int	(*vop_setfl)(vnode_t *, int, int, cred_t *,		\
904 				caller_context_t *);			\
905 	int	(*vop_getattr)(vnode_t *, vattr_t *, int, cred_t *,	\
906 				caller_context_t *);			\
907 	int	(*vop_setattr)(vnode_t *, vattr_t *, int, cred_t *,	\
908 				caller_context_t *);			\
909 	int	(*vop_access)(vnode_t *, int, int, cred_t *,		\
910 				caller_context_t *);			\
911 	int	(*vop_lookup)(vnode_t *, char *, vnode_t **,		\
912 				struct pathname *,			\
913 				int, vnode_t *, cred_t *,		\
914 				caller_context_t *, int *,		\
915 				struct pathname *);			\
916 	int	(*vop_create)(vnode_t *, char *, vattr_t *, vcexcl_t,	\
917 				int, vnode_t **, cred_t *, int,		\
918 				caller_context_t *, vsecattr_t *);	\
919 	int	(*vop_remove)(vnode_t *, char *, cred_t *,		\
920 				caller_context_t *, int);		\
921 	int	(*vop_link)(vnode_t *, vnode_t *, char *, cred_t *,	\
922 				caller_context_t *, int);		\
923 	int	(*vop_rename)(vnode_t *, char *, vnode_t *, char *,	\
924 				cred_t *, caller_context_t *, int);	\
925 	int	(*vop_mkdir)(vnode_t *, char *, vattr_t *, vnode_t **,	\
926 				cred_t *, caller_context_t *, int,	\
927 				vsecattr_t *);				\
928 	int	(*vop_rmdir)(vnode_t *, char *, vnode_t *, cred_t *,	\
929 				caller_context_t *, int);		\
930 	int	(*vop_readdir)(vnode_t *, uio_t *, cred_t *, int *,	\
931 				caller_context_t *, int);		\
932 	int	(*vop_symlink)(vnode_t *, char *, vattr_t *, char *,	\
933 				cred_t *, caller_context_t *, int);	\
934 	int	(*vop_readlink)(vnode_t *, uio_t *, cred_t *,		\
935 				caller_context_t *);			\
936 	int	(*vop_fsync)(vnode_t *, int, cred_t *,			\
937 				caller_context_t *);			\
938 	void	(*vop_inactive)(vnode_t *, cred_t *,			\
939 				caller_context_t *);			\
940 	int	(*vop_fid)(vnode_t *, struct fid *,			\
941 				caller_context_t *);			\
942 	int	(*vop_rwlock)(vnode_t *, int, caller_context_t *);	\
943 	void	(*vop_rwunlock)(vnode_t *, int, caller_context_t *);	\
944 	int	(*vop_seek)(vnode_t *, offset_t, offset_t *,		\
945 				caller_context_t *);			\
946 	int	(*vop_cmp)(vnode_t *, vnode_t *, caller_context_t *);	\
947 	int	(*vop_frlock)(vnode_t *, int, struct flock64 *,		\
948 				int, offset_t,				\
949 				struct flk_callback *, cred_t *,	\
950 				caller_context_t *);			\
951 	int	(*vop_space)(vnode_t *, int, struct flock64 *,		\
952 				int, offset_t,				\
953 				cred_t *, caller_context_t *);		\
954 	int	(*vop_realvp)(vnode_t *, vnode_t **,			\
955 				caller_context_t *);			\
956 	int	(*vop_getpage)(vnode_t *, offset_t, size_t, uint_t *,	\
957 				struct page **, size_t, struct seg *,	\
958 				caddr_t, enum seg_rw, cred_t *,		\
959 				caller_context_t *);			\
960 	int	(*vop_putpage)(vnode_t *, offset_t, size_t,		\
961 				int, cred_t *, caller_context_t *);	\
962 	int	(*vop_map)(vnode_t *, offset_t, struct as *,		\
963 				caddr_t *, size_t,			\
964 				uchar_t, uchar_t, uint_t, cred_t *,	\
965 				caller_context_t *);			\
966 	int	(*vop_addmap)(vnode_t *, offset_t, struct as *,		\
967 				caddr_t, size_t,			\
968 				uchar_t, uchar_t, uint_t, cred_t *,	\
969 				caller_context_t *);			\
970 	int	(*vop_delmap)(vnode_t *, offset_t, struct as *,		\
971 				caddr_t, size_t,			\
972 				uint_t, uint_t, uint_t, cred_t *,	\
973 				caller_context_t *);			\
974 	int	(*vop_poll)(vnode_t *, short, int, short *,		\
975 				struct pollhead **,			\
976 				caller_context_t *);			\
977 	int	(*vop_dump)(vnode_t *, caddr_t, offset_t, offset_t,	\
978 				caller_context_t *);			\
979 	int	(*vop_pathconf)(vnode_t *, int, ulong_t *, cred_t *,	\
980 				caller_context_t *);			\
981 	int	(*vop_pageio)(vnode_t *, struct page *,			\
982 				u_offset_t, size_t, int, cred_t *,	\
983 				caller_context_t *);			\
984 	int	(*vop_dumpctl)(vnode_t *, int, offset_t *,		\
985 				caller_context_t *);			\
986 	void	(*vop_dispose)(vnode_t *, struct page *,		\
987 				int, int, cred_t *,			\
988 				caller_context_t *);			\
989 	int	(*vop_setsecattr)(vnode_t *, vsecattr_t *,		\
990 				int, cred_t *, caller_context_t *);	\
991 	int	(*vop_getsecattr)(vnode_t *, vsecattr_t *,		\
992 				int, cred_t *, caller_context_t *);	\
993 	int	(*vop_shrlock)(vnode_t *, int, struct shrlock *,	\
994 				int, cred_t *, caller_context_t *);	\
995 	int	(*vop_vnevent)(vnode_t *, vnevent_t, vnode_t *,		\
996 				char *, caller_context_t *);		\
997 	int	(*vop_reqzcbuf)(vnode_t *, enum uio_rw, xuio_t *,	\
998 				cred_t *, caller_context_t *);		\
999 	int	(*vop_retzcbuf)(vnode_t *, xuio_t *, cred_t *,		\
1000 				caller_context_t *)
1001 	/* NB: No ";" */
1002 
1003 /*
1004  * Operations on vnodes.  Note: File systems must never operate directly
1005  * on a 'vnodeops' structure -- it WILL change in future releases!  They
1006  * must use vn_make_ops() to create the structure.
1007  */
1008 typedef struct vnodeops {
1009 	const char *vnop_name;
1010 	VNODE_OPS;	/* Signatures of all vnode operations (vops) */
1011 } vnodeops_t;
1012 
1013 typedef int (*fs_generic_func_p) ();	/* Generic vop/vfsop/femop/fsemop ptr */
1014 
1015 extern int	fop_open(vnode_t **, int, cred_t *, caller_context_t *);
1016 extern int	fop_close(vnode_t *, int, int, offset_t, cred_t *,
1017 				caller_context_t *);
1018 extern int	fop_read(vnode_t *, uio_t *, int, cred_t *, caller_context_t *);
1019 extern int	fop_write(vnode_t *, uio_t *, int, cred_t *,
1020 				caller_context_t *);
1021 extern int	fop_ioctl(vnode_t *, int, intptr_t, int, cred_t *, int *,
1022 				caller_context_t *);
1023 extern int	fop_setfl(vnode_t *, int, int, cred_t *, caller_context_t *);
1024 extern int	fop_getattr(vnode_t *, vattr_t *, int, cred_t *,
1025 				caller_context_t *);
1026 extern int	fop_setattr(vnode_t *, vattr_t *, int, cred_t *,
1027 				caller_context_t *);
1028 extern int	fop_access(vnode_t *, int, int, cred_t *, caller_context_t *);
1029 extern int	fop_lookup(vnode_t *, char *, vnode_t **, struct pathname *,
1030 				int, vnode_t *, cred_t *, caller_context_t *,
1031 				int *, struct pathname *);
1032 extern int	fop_create(vnode_t *, char *, vattr_t *, vcexcl_t, int,
1033 				vnode_t **, cred_t *, int, caller_context_t *,
1034 				vsecattr_t *);
1035 extern int	fop_remove(vnode_t *vp, char *, cred_t *, caller_context_t *,
1036 				int);
1037 extern int	fop_link(vnode_t *, vnode_t *, char *, cred_t *,
1038 				caller_context_t *, int);
1039 extern int	fop_rename(vnode_t *, char *, vnode_t *, char *, cred_t *,
1040 				caller_context_t *, int);
1041 extern int	fop_mkdir(vnode_t *, char *, vattr_t *, vnode_t **, cred_t *,
1042 				caller_context_t *, int, vsecattr_t *);
1043 extern int	fop_rmdir(vnode_t *, char *, vnode_t *, cred_t *,
1044 				caller_context_t *, int);
1045 extern int	fop_readdir(vnode_t *, uio_t *, cred_t *, int *,
1046 				caller_context_t *, int);
1047 extern int	fop_symlink(vnode_t *, char *, vattr_t *, char *, cred_t *,
1048 				caller_context_t *, int);
1049 extern int	fop_readlink(vnode_t *, uio_t *, cred_t *, caller_context_t *);
1050 extern int	fop_fsync(vnode_t *, int, cred_t *, caller_context_t *);
1051 extern void	fop_inactive(vnode_t *, cred_t *, caller_context_t *);
1052 extern int	fop_fid(vnode_t *, struct fid *, caller_context_t *);
1053 extern int	fop_rwlock(vnode_t *, int, caller_context_t *);
1054 extern void	fop_rwunlock(vnode_t *, int, caller_context_t *);
1055 extern int	fop_seek(vnode_t *, offset_t, offset_t *, caller_context_t *);
1056 extern int	fop_cmp(vnode_t *, vnode_t *, caller_context_t *);
1057 extern int	fop_frlock(vnode_t *, int, struct flock64 *, int, offset_t,
1058 				struct flk_callback *, cred_t *,
1059 				caller_context_t *);
1060 extern int	fop_space(vnode_t *, int, struct flock64 *, int, offset_t,
1061 				cred_t *, caller_context_t *);
1062 extern int	fop_realvp(vnode_t *, vnode_t **, caller_context_t *);
1063 extern int	fop_getpage(vnode_t *, offset_t, size_t, uint_t *,
1064 				struct page **, size_t, struct seg *,
1065 				caddr_t, enum seg_rw, cred_t *,
1066 				caller_context_t *);
1067 extern int	fop_putpage(vnode_t *, offset_t, size_t, int, cred_t *,
1068 				caller_context_t *);
1069 extern int	fop_map(vnode_t *, offset_t, struct as *, caddr_t *, size_t,
1070 				uchar_t, uchar_t, uint_t, cred_t *cr,
1071 				caller_context_t *);
1072 extern int	fop_addmap(vnode_t *, offset_t, struct as *, caddr_t, size_t,
1073 				uchar_t, uchar_t, uint_t, cred_t *,
1074 				caller_context_t *);
1075 extern int	fop_delmap(vnode_t *, offset_t, struct as *, caddr_t, size_t,
1076 				uint_t, uint_t, uint_t, cred_t *,
1077 				caller_context_t *);
1078 extern int	fop_poll(vnode_t *, short, int, short *, struct pollhead **,
1079 				caller_context_t *);
1080 extern int	fop_dump(vnode_t *, caddr_t, offset_t, offset_t,
1081     caller_context_t *);
1082 extern int	fop_pathconf(vnode_t *, int, ulong_t *, cred_t *,
1083 				caller_context_t *);
1084 extern int	fop_pageio(vnode_t *, struct page *, u_offset_t, size_t, int,
1085 				cred_t *, caller_context_t *);
1086 extern int	fop_dumpctl(vnode_t *, int, offset_t *, caller_context_t *);
1087 extern void	fop_dispose(vnode_t *, struct page *, int, int, cred_t *,
1088 				caller_context_t *);
1089 extern int	fop_setsecattr(vnode_t *, vsecattr_t *, int, cred_t *,
1090 				caller_context_t *);
1091 extern int	fop_getsecattr(vnode_t *, vsecattr_t *, int, cred_t *,
1092 				caller_context_t *);
1093 extern int	fop_shrlock(vnode_t *, int, struct shrlock *, int, cred_t *,
1094 				caller_context_t *);
1095 extern int	fop_vnevent(vnode_t *, vnevent_t, vnode_t *, char *,
1096 				caller_context_t *);
1097 extern int	fop_reqzcbuf(vnode_t *, enum uio_rw, xuio_t *, cred_t *,
1098 				caller_context_t *);
1099 extern int	fop_retzcbuf(vnode_t *, xuio_t *, cred_t *, caller_context_t *);
1100 
1101 #endif	/* _KERNEL */
1102 
1103 #define	VOP_OPEN(vpp, mode, cr, ct) \
1104 	fop_open(vpp, mode, cr, ct)
1105 #define	VOP_CLOSE(vp, f, c, o, cr, ct) \
1106 	fop_close(vp, f, c, o, cr, ct)
1107 #define	VOP_READ(vp, uiop, iof, cr, ct) \
1108 	fop_read(vp, uiop, iof, cr, ct)
1109 #define	VOP_WRITE(vp, uiop, iof, cr, ct) \
1110 	fop_write(vp, uiop, iof, cr, ct)
1111 #define	VOP_IOCTL(vp, cmd, a, f, cr, rvp, ct) \
1112 	fop_ioctl(vp, cmd, a, f, cr, rvp, ct)
1113 #define	VOP_SETFL(vp, f, a, cr, ct) \
1114 	fop_setfl(vp, f, a, cr, ct)
1115 #define	VOP_GETATTR(vp, vap, f, cr, ct) \
1116 	fop_getattr(vp, vap, f, cr, ct)
1117 #define	VOP_SETATTR(vp, vap, f, cr, ct) \
1118 	fop_setattr(vp, vap, f, cr, ct)
1119 #define	VOP_ACCESS(vp, mode, f, cr, ct) \
1120 	fop_access(vp, mode, f, cr, ct)
1121 #define	VOP_LOOKUP(vp, cp, vpp, pnp, f, rdir, cr, ct, defp, rpnp) \
1122 	fop_lookup(vp, cp, vpp, pnp, f, rdir, cr, ct, defp, rpnp)
1123 #define	VOP_CREATE(dvp, p, vap, ex, mode, vpp, cr, flag, ct, vsap) \
1124 	fop_create(dvp, p, vap, ex, mode, vpp, cr, flag, ct, vsap)
1125 #define	VOP_REMOVE(dvp, p, cr, ct, f) \
1126 	fop_remove(dvp, p, cr, ct, f)
1127 #define	VOP_LINK(tdvp, fvp, p, cr, ct, f) \
1128 	fop_link(tdvp, fvp, p, cr, ct, f)
1129 #define	VOP_RENAME(fvp, fnm, tdvp, tnm, cr, ct, f) \
1130 	fop_rename(fvp, fnm, tdvp, tnm, cr, ct, f)
1131 #define	VOP_MKDIR(dp, p, vap, vpp, cr, ct, f, vsap) \
1132 	fop_mkdir(dp, p, vap, vpp, cr, ct, f, vsap)
1133 #define	VOP_RMDIR(dp, p, cdir, cr, ct, f) \
1134 	fop_rmdir(dp, p, cdir, cr, ct, f)
1135 #define	VOP_READDIR(vp, uiop, cr, eofp, ct, f) \
1136 	fop_readdir(vp, uiop, cr, eofp, ct, f)
1137 #define	VOP_SYMLINK(dvp, lnm, vap, tnm, cr, ct, f) \
1138 	fop_symlink(dvp, lnm, vap, tnm, cr, ct, f)
1139 #define	VOP_READLINK(vp, uiop, cr, ct) \
1140 	fop_readlink(vp, uiop, cr, ct)
1141 #define	VOP_FSYNC(vp, syncflag, cr, ct) \
1142 	fop_fsync(vp, syncflag, cr, ct)
1143 #define	VOP_INACTIVE(vp, cr, ct) \
1144 	fop_inactive(vp, cr, ct)
1145 #define	VOP_FID(vp, fidp, ct) \
1146 	fop_fid(vp, fidp, ct)
1147 #define	VOP_RWLOCK(vp, w, ct) \
1148 	fop_rwlock(vp, w, ct)
1149 #define	VOP_RWUNLOCK(vp, w, ct) \
1150 	fop_rwunlock(vp, w, ct)
1151 #define	VOP_SEEK(vp, ooff, noffp, ct) \
1152 	fop_seek(vp, ooff, noffp, ct)
1153 #define	VOP_CMP(vp1, vp2, ct) \
1154 	fop_cmp(vp1, vp2, ct)
1155 #define	VOP_FRLOCK(vp, cmd, a, f, o, cb, cr, ct) \
1156 	fop_frlock(vp, cmd, a, f, o, cb, cr, ct)
1157 #define	VOP_SPACE(vp, cmd, a, f, o, cr, ct) \
1158 	fop_space(vp, cmd, a, f, o, cr, ct)
1159 #define	VOP_REALVP(vp1, vp2, ct) \
1160 	fop_realvp(vp1, vp2, ct)
1161 #define	VOP_GETPAGE(vp, of, sz, pr, pl, ps, sg, a, rw, cr, ct) \
1162 	fop_getpage(vp, of, sz, pr, pl, ps, sg, a, rw, cr, ct)
1163 #define	VOP_PUTPAGE(vp, of, sz, fl, cr, ct) \
1164 	fop_putpage(vp, of, sz, fl, cr, ct)
1165 #define	VOP_MAP(vp, of, as, a, sz, p, mp, fl, cr, ct) \
1166 	fop_map(vp, of, as, a, sz, p, mp, fl, cr, ct)
1167 #define	VOP_ADDMAP(vp, of, as, a, sz, p, mp, fl, cr, ct) \
1168 	fop_addmap(vp, of, as, a, sz, p, mp, fl, cr, ct)
1169 #define	VOP_DELMAP(vp, of, as, a, sz, p, mp, fl, cr, ct) \
1170 	fop_delmap(vp, of, as, a, sz, p, mp, fl, cr, ct)
1171 #define	VOP_POLL(vp, events, anyyet, reventsp, phpp, ct) \
1172 	fop_poll(vp, events, anyyet, reventsp, phpp, ct)
1173 #define	VOP_DUMP(vp, addr, bn, count, ct) \
1174 	fop_dump(vp, addr, bn, count, ct)
1175 #define	VOP_PATHCONF(vp, cmd, valp, cr, ct) \
1176 	fop_pathconf(vp, cmd, valp, cr, ct)
1177 #define	VOP_PAGEIO(vp, pp, io_off, io_len, flags, cr, ct) \
1178 	fop_pageio(vp, pp, io_off, io_len, flags, cr, ct)
1179 #define	VOP_DUMPCTL(vp, action, blkp, ct) \
1180 	fop_dumpctl(vp, action, blkp, ct)
1181 #define	VOP_DISPOSE(vp, pp, flag, dn, cr, ct) \
1182 	fop_dispose(vp, pp, flag, dn, cr, ct)
1183 #define	VOP_GETSECATTR(vp, vsap, f, cr, ct) \
1184 	fop_getsecattr(vp, vsap, f, cr, ct)
1185 #define	VOP_SETSECATTR(vp, vsap, f, cr, ct) \
1186 	fop_setsecattr(vp, vsap, f, cr, ct)
1187 #define	VOP_SHRLOCK(vp, cmd, shr, f, cr, ct) \
1188 	fop_shrlock(vp, cmd, shr, f, cr, ct)
1189 #define	VOP_VNEVENT(vp, vnevent, dvp, fnm, ct) \
1190 	fop_vnevent(vp, vnevent, dvp, fnm, ct)
1191 #define	VOP_REQZCBUF(vp, rwflag, xuiop, cr, ct) \
1192 	fop_reqzcbuf(vp, rwflag, xuiop, cr, ct)
1193 #define	VOP_RETZCBUF(vp, xuiop, cr, ct) \
1194 	fop_retzcbuf(vp, xuiop, cr, ct)
1195 
1196 #define	VOPNAME_OPEN		"open"
1197 #define	VOPNAME_CLOSE		"close"
1198 #define	VOPNAME_READ		"read"
1199 #define	VOPNAME_WRITE		"write"
1200 #define	VOPNAME_IOCTL		"ioctl"
1201 #define	VOPNAME_SETFL		"setfl"
1202 #define	VOPNAME_GETATTR		"getattr"
1203 #define	VOPNAME_SETATTR		"setattr"
1204 #define	VOPNAME_ACCESS		"access"
1205 #define	VOPNAME_LOOKUP		"lookup"
1206 #define	VOPNAME_CREATE		"create"
1207 #define	VOPNAME_REMOVE		"remove"
1208 #define	VOPNAME_LINK		"link"
1209 #define	VOPNAME_RENAME		"rename"
1210 #define	VOPNAME_MKDIR		"mkdir"
1211 #define	VOPNAME_RMDIR		"rmdir"
1212 #define	VOPNAME_READDIR		"readdir"
1213 #define	VOPNAME_SYMLINK		"symlink"
1214 #define	VOPNAME_READLINK	"readlink"
1215 #define	VOPNAME_FSYNC		"fsync"
1216 #define	VOPNAME_INACTIVE	"inactive"
1217 #define	VOPNAME_FID		"fid"
1218 #define	VOPNAME_RWLOCK		"rwlock"
1219 #define	VOPNAME_RWUNLOCK	"rwunlock"
1220 #define	VOPNAME_SEEK		"seek"
1221 #define	VOPNAME_CMP		"cmp"
1222 #define	VOPNAME_FRLOCK		"frlock"
1223 #define	VOPNAME_SPACE		"space"
1224 #define	VOPNAME_REALVP		"realvp"
1225 #define	VOPNAME_GETPAGE		"getpage"
1226 #define	VOPNAME_PUTPAGE		"putpage"
1227 #define	VOPNAME_MAP		"map"
1228 #define	VOPNAME_ADDMAP		"addmap"
1229 #define	VOPNAME_DELMAP		"delmap"
1230 #define	VOPNAME_POLL		"poll"
1231 #define	VOPNAME_DUMP		"dump"
1232 #define	VOPNAME_PATHCONF	"pathconf"
1233 #define	VOPNAME_PAGEIO		"pageio"
1234 #define	VOPNAME_DUMPCTL		"dumpctl"
1235 #define	VOPNAME_DISPOSE		"dispose"
1236 #define	VOPNAME_GETSECATTR	"getsecattr"
1237 #define	VOPNAME_SETSECATTR	"setsecattr"
1238 #define	VOPNAME_SHRLOCK		"shrlock"
1239 #define	VOPNAME_VNEVENT		"vnevent"
1240 #define	VOPNAME_REQZCBUF	"reqzcbuf"
1241 #define	VOPNAME_RETZCBUF	"retzcbuf"
1242 
1243 /*
1244  * Flags for VOP_LOOKUP
1245  *
1246  * Defined in file.h, but also possible, FIGNORECASE and FSEARCH
1247  *
1248  */
1249 #define	LOOKUP_DIR		0x01	/* want parent dir vp */
1250 #define	LOOKUP_XATTR		0x02	/* lookup up extended attr dir */
1251 #define	CREATE_XATTR_DIR	0x04	/* Create extended attr dir */
1252 #define	LOOKUP_HAVE_SYSATTR_DIR	0x08	/* Already created virtual GFS dir */
1253 
1254 /*
1255  * Flags for VOP_READDIR
1256  */
1257 #define	V_RDDIR_ENTFLAGS	0x01	/* request dirent flags */
1258 #define	V_RDDIR_ACCFILTER	0x02	/* filter out inaccessible dirents */
1259 
1260 /*
1261  * Flags for VOP_RWLOCK/VOP_RWUNLOCK
1262  * VOP_RWLOCK will return the flag that was actually set, or -1 if none.
1263  */
1264 #define	V_WRITELOCK_TRUE	(1)	/* Request write-lock on the vnode */
1265 #define	V_WRITELOCK_FALSE	(0)	/* Request read-lock on the vnode */
1266 
1267 /*
1268  * Flags for VOP_DUMPCTL
1269  */
1270 #define	DUMP_ALLOC	0
1271 #define	DUMP_FREE	1
1272 #define	DUMP_SCAN	2
1273 
1274 /*
1275  * Public vnode manipulation functions.
1276  */
1277 #ifdef	_KERNEL
1278 
1279 vnode_t *vn_alloc(int);
1280 void	vn_reinit(vnode_t *);
1281 void	vn_recycle(vnode_t *);
1282 void	vn_free(vnode_t *);
1283 
1284 int	vn_is_readonly(vnode_t *);
1285 int   	vn_is_opened(vnode_t *, v_mode_t);
1286 int   	vn_is_mapped(vnode_t *, v_mode_t);
1287 int   	vn_has_other_opens(vnode_t *, v_mode_t);
1288 void	vn_open_upgrade(vnode_t *, int);
1289 void	vn_open_downgrade(vnode_t *, int);
1290 
1291 int	vn_can_change_zones(vnode_t *vp);
1292 
1293 int	vn_has_flocks(vnode_t *);
1294 int	vn_has_mandatory_locks(vnode_t *, int);
1295 int	vn_has_cached_data(vnode_t *);
1296 
1297 void	vn_setops(vnode_t *, vnodeops_t *);
1298 vnodeops_t *vn_getops(vnode_t *);
1299 int	vn_matchops(vnode_t *, vnodeops_t *);
1300 int	vn_matchopval(vnode_t *, char *, fs_generic_func_p);
1301 int	vn_ismntpt(vnode_t *);
1302 
1303 struct vfs *vn_mountedvfs(vnode_t *);
1304 
1305 int	vn_in_dnlc(vnode_t *);
1306 
1307 void	vn_create_cache(void);
1308 void	vn_destroy_cache(void);
1309 
1310 void	vn_freevnodeops(vnodeops_t *);
1311 
1312 int	vn_open(char *pnamep, enum uio_seg seg, int filemode, int createmode,
1313 		struct vnode **vpp, enum create crwhy, mode_t umask);
1314 int	vn_openat(char *pnamep, enum uio_seg seg, int filemode, int createmode,
1315 		struct vnode **vpp, enum create crwhy,
1316 		mode_t umask, struct vnode *startvp, int fd);
1317 int	vn_create(char *pnamep, enum uio_seg seg, struct vattr *vap,
1318 		enum vcexcl excl, int mode, struct vnode **vpp,
1319 		enum create why, int flag, mode_t umask);
1320 int	vn_createat(char *pnamep, enum uio_seg seg, struct vattr *vap,
1321 		enum vcexcl excl, int mode, struct vnode **vpp,
1322 		enum create why, int flag, mode_t umask, struct vnode *startvp);
1323 int	vn_rdwr(enum uio_rw rw, struct vnode *vp, caddr_t base, ssize_t len,
1324 		offset_t offset, enum uio_seg seg, int ioflag, rlim64_t ulimit,
1325 		cred_t *cr, ssize_t *residp);
1326 void	vn_rele(struct vnode *vp);
1327 void	vn_rele_async(struct vnode *vp, struct taskq *taskq);
1328 void	vn_rele_dnlc(struct vnode *vp);
1329 void	vn_rele_stream(struct vnode *vp);
1330 int	vn_link(char *from, char *to, enum uio_seg seg);
1331 int	vn_linkat(vnode_t *fstartvp, char *from, enum symfollow follow,
1332 		vnode_t *tstartvp, char *to, enum uio_seg seg);
1333 int	vn_rename(char *from, char *to, enum uio_seg seg);
1334 int	vn_renameat(vnode_t *fdvp, char *fname, vnode_t *tdvp, char *tname,
1335 		enum uio_seg seg);
1336 int	vn_remove(char *fnamep, enum uio_seg seg, enum rm dirflag);
1337 int	vn_removeat(vnode_t *startvp, char *fnamep, enum uio_seg seg,
1338 		enum rm dirflag);
1339 int	vn_compare(vnode_t *vp1, vnode_t *vp2);
1340 int	vn_vfswlock(struct vnode *vp);
1341 int	vn_vfswlock_wait(struct vnode *vp);
1342 int	vn_vfsrlock(struct vnode *vp);
1343 int	vn_vfsrlock_wait(struct vnode *vp);
1344 void	vn_vfsunlock(struct vnode *vp);
1345 int	vn_vfswlock_held(struct vnode *vp);
1346 vnode_t *specvp(struct vnode *vp, dev_t dev, vtype_t type, struct cred *cr);
1347 vnode_t *makespecvp(dev_t dev, vtype_t type);
1348 vn_vfslocks_entry_t *vn_vfslocks_getlock(void *);
1349 void	vn_vfslocks_rele(vn_vfslocks_entry_t *);
1350 boolean_t vn_is_reparse(vnode_t *, cred_t *, caller_context_t *);
1351 
1352 void vn_copypath(struct vnode *src, struct vnode *dst);
1353 void vn_setpath_str(struct vnode *vp, const char *str, size_t len);
1354 void vn_setpath(vnode_t *rootvp, struct vnode *startvp, struct vnode *vp,
1355     const char *path, size_t plen);
1356 void vn_renamepath(vnode_t *dvp, vnode_t *vp, const char *nm, size_t len);
1357 
1358 /* Private vnode manipulation functions */
1359 void vn_clearpath(vnode_t *, hrtime_t);
1360 void vn_updatepath(vnode_t *, vnode_t *, const char *);
1361 
1362 
1363 /* Vnode event notification */
1364 void	vnevent_rename_src(vnode_t *, vnode_t *, char *, caller_context_t *);
1365 void	vnevent_rename_dest(vnode_t *, vnode_t *, char *, caller_context_t *);
1366 void	vnevent_remove(vnode_t *, vnode_t *, char *, caller_context_t *);
1367 void	vnevent_rmdir(vnode_t *, vnode_t *, char *, caller_context_t *);
1368 void	vnevent_create(vnode_t *, caller_context_t *);
1369 void	vnevent_link(vnode_t *, caller_context_t *);
1370 void	vnevent_rename_dest_dir(vnode_t *, caller_context_t *ct);
1371 void	vnevent_mountedover(vnode_t *, caller_context_t *);
1372 void	vnevent_truncate(vnode_t *, caller_context_t *);
1373 int	vnevent_support(vnode_t *, caller_context_t *);
1374 void	vnevent_pre_rename_src(vnode_t *, vnode_t *, char *,
1375 	    caller_context_t *);
1376 void	vnevent_pre_rename_dest(vnode_t *, vnode_t *, char *,
1377 	    caller_context_t *);
1378 void	vnevent_pre_rename_dest_dir(vnode_t *, vnode_t *, char *,
1379 	    caller_context_t *);
1380 
1381 /* Vnode specific data */
1382 void vsd_create(uint_t *, void (*)(void *));
1383 void vsd_destroy(uint_t *);
1384 void *vsd_get(vnode_t *, uint_t);
1385 int vsd_set(vnode_t *, uint_t, void *);
1386 void vsd_free(vnode_t *);
1387 
1388 /*
1389  * Extensible vnode attribute (xva) routines:
1390  * xva_init() initializes an xvattr_t (zero struct, init mapsize, set AT_XATTR)
1391  * xva_getxoptattr() returns a ponter to the xoptattr_t section of xvattr_t
1392  */
1393 void		xva_init(xvattr_t *);
1394 xoptattr_t	*xva_getxoptattr(xvattr_t *);	/* Get ptr to xoptattr_t */
1395 
1396 void xattr_init(void);		/* Initialize vnodeops for xattrs */
1397 
1398 /* GFS tunnel for xattrs */
1399 int xattr_dir_lookup(vnode_t *, vnode_t **, int, cred_t *);
1400 
1401 /* Reparse Point */
1402 void reparse_point_init(void);
1403 
1404 /* Context identification */
1405 u_longlong_t	fs_new_caller_id();
1406 
1407 int	vn_vmpss_usepageio(vnode_t *);
1408 
1409 /* Empty v_path placeholder */
1410 extern char *vn_vpath_empty;
1411 
1412 /*
1413  * Needed for use of IS_VMODSORT() in kernel.
1414  */
1415 extern uint_t pvn_vmodsort_supported;
1416 
1417 /*
1418  * All changes to v_count should be done through VN_HOLD() or VN_RELE(), or
1419  * one of their variants. This makes it possible to ensure proper locking,
1420  * and to guarantee that all modifications are accompanied by a firing of
1421  * the vn-hold or vn-rele SDT DTrace probe.
1422  *
1423  * Example DTrace command for tracing vnode references using these probes:
1424  *
1425  * dtrace -q -n 'sdt:::vn-hold,sdt:::vn-rele
1426  * {
1427  *	this->vp = (vnode_t *)arg0;
1428  *	printf("%s %s(%p[%s]) %d\n", execname, probename, this->vp,
1429  *	    this->vp->v_path == NULL ? "NULL" : stringof(this->vp->v_path),
1430  *	    this->vp->v_count)
1431  * }'
1432  */
1433 #define	VN_HOLD_LOCKED(vp) {			\
1434 	ASSERT(mutex_owned(&(vp)->v_lock));	\
1435 	(vp)->v_count++;			\
1436 	DTRACE_PROBE1(vn__hold, vnode_t *, vp);	\
1437 }
1438 
1439 #define	VN_HOLD(vp)	{		\
1440 	mutex_enter(&(vp)->v_lock);	\
1441 	VN_HOLD_LOCKED(vp);		\
1442 	mutex_exit(&(vp)->v_lock);	\
1443 }
1444 
1445 #define	VN_RELE(vp)	{ \
1446 	vn_rele(vp); \
1447 }
1448 
1449 #define	VN_RELE_ASYNC(vp, taskq)	{ \
1450 	vn_rele_async(vp, taskq); \
1451 }
1452 
1453 #define	VN_RELE_LOCKED(vp) {			\
1454 	ASSERT(mutex_owned(&(vp)->v_lock));	\
1455 	ASSERT((vp)->v_count >= 1);		\
1456 	(vp)->v_count--;			\
1457 	DTRACE_PROBE1(vn__rele, vnode_t *, vp);	\
1458 }
1459 
1460 #define	VN_SET_VFS_TYPE_DEV(vp, vfsp, type, dev)	{ \
1461 	(vp)->v_vfsp = (vfsp); \
1462 	(vp)->v_type = (type); \
1463 	(vp)->v_rdev = (dev); \
1464 }
1465 
1466 /*
1467  * Compare two vnodes for equality.  In general this macro should be used
1468  * in preference to calling VOP_CMP directly.
1469  */
1470 #define	VN_CMP(VP1, VP2)	((VP1) == (VP2) ? 1 : 	\
1471 	((VP1) && (VP2) && (vn_getops(VP1) == vn_getops(VP2)) ? \
1472 	VOP_CMP(VP1, VP2, NULL) : 0))
1473 
1474 /*
1475  * Some well-known global vnodes used by the VM system to name pages.
1476  */
1477 extern struct vnode kvps[];
1478 
1479 typedef enum {
1480 	KV_KVP,		/* vnode for all segkmem pages */
1481 	KV_ZVP,		/* vnode for all ZFS pages */
1482 #if defined(__sparc)
1483 	KV_MPVP,	/* vnode for all page_t meta-pages */
1484 	KV_PROMVP,	/* vnode for all PROM pages */
1485 #endif	/* __sparc */
1486 	KV_MAX		/* total number of vnodes in kvps[] */
1487 } kvps_index_t;
1488 
1489 #define	VN_ISKAS(vp)	((vp) >= &kvps[0] && (vp) < &kvps[KV_MAX])
1490 
1491 #endif	/* _KERNEL */
1492 
1493 /*
1494  * Flags to VOP_SETATTR/VOP_GETATTR.
1495  */
1496 #define	ATTR_UTIME	0x01	/* non-default utime(2) request */
1497 #define	ATTR_EXEC	0x02	/* invocation from exec(2) */
1498 #define	ATTR_COMM	0x04	/* yield common vp attributes */
1499 #define	ATTR_HINT	0x08	/* information returned will be `hint' */
1500 #define	ATTR_REAL	0x10	/* yield attributes of the real vp */
1501 #define	ATTR_NOACLCHECK	0x20	/* Don't check ACL when checking permissions */
1502 #define	ATTR_TRIGGER	0x40	/* Mount first if vnode is a trigger mount */
1503 /*
1504  * Generally useful macros.
1505  */
1506 #define	VBSIZE(vp)	((vp)->v_vfsp->vfs_bsize)
1507 
1508 #define	VTOZONE(vp)	((vp)->v_vfsp->vfs_zone)
1509 
1510 #define	NULLVP		((struct vnode *)0)
1511 #define	NULLVPP		((struct vnode **)0)
1512 
1513 #ifdef	_KERNEL
1514 
1515 /*
1516  * Structure used while handling asynchronous VOP_PUTPAGE operations.
1517  */
1518 struct async_reqs {
1519 	struct async_reqs *a_next;	/* pointer to next arg struct */
1520 	struct vnode *a_vp;		/* vnode pointer */
1521 	u_offset_t a_off;			/* offset in file */
1522 	uint_t a_len;			/* size of i/o request */
1523 	int a_flags;			/* flags to indicate operation type */
1524 	struct cred *a_cred;		/* cred pointer	*/
1525 	ushort_t a_prealloced;		/* set if struct is pre-allocated */
1526 };
1527 
1528 /*
1529  * VN_DISPOSE() -- given a page pointer, safely invoke VOP_DISPOSE().
1530  * Note that there is no guarantee that the page passed in will be
1531  * freed.  If that is required, then a check after calling VN_DISPOSE would
1532  * be necessary to ensure the page was freed.
1533  */
1534 #define	VN_DISPOSE(pp, flag, dn, cr)	{ \
1535 	if ((pp)->p_vnode != NULL && !VN_ISKAS((pp)->p_vnode)) \
1536 		VOP_DISPOSE((pp)->p_vnode, (pp), (flag), (dn), (cr), NULL); \
1537 	else if ((flag) == B_FREE) \
1538 		page_free((pp), (dn)); \
1539 	else \
1540 		page_destroy((pp), (dn)); \
1541 	}
1542 
1543 #endif	/* _KERNEL */
1544 
1545 #ifdef	__cplusplus
1546 }
1547 #endif
1548 
1549 #endif	/* _SYS_VNODE_H */
1550