xref: /illumos-gate/usr/src/uts/common/nfs/export.h (revision 66582b606a8194f7f3ba5b3a3a6dca5b0d346361)
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) 1989, 2010, Oracle and/or its affiliates. All rights reserved.
24  * Copyright 2016 Nexenta Systems, Inc.  All rights reserved.
25  * Copyright 2016 Jason King.
26  * Copyright 2018 Nexenta Systems, Inc.  All rights reserved.
27  */
28 
29 /*	Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T	*/
30 /*	  All Rights Reserved	*/
31 
32 #ifndef	_NFS_EXPORT_H
33 #define	_NFS_EXPORT_H
34 
35 #include <nfs/nfs_sec.h>
36 #include <nfs/auth.h>
37 #include <sys/vnode.h>
38 #include <nfs/nfs4.h>
39 #include <sys/kiconv.h>
40 #include <sys/avl.h>
41 #include <sys/zone.h>
42 
43 #ifdef _KERNEL
44 #include <sys/pkp_hash.h> /* for PKP_HASH_SIZE */
45 #endif /* _KERNEL */
46 
47 #ifdef	__cplusplus
48 extern "C" {
49 #endif
50 
51 /*
52  * nfs pseudo flavor number is owned by IANA. Need to make sure the
53  * Solaris specific NFS_FLAVOR_NOMAP number will not overlap with any
54  * new IANA defined pseudo flavor numbers. The chance for the overlap
55  * is very small since the growth of new flavor numbers is expected
56  * to be limited.
57  */
58 #define	NFS_FLAVOR_NOMAP	999999	/* no nfs flavor mapping */
59 
60 /*
61  * As duplicate flavors can be passed into exportfs in the arguments, we
62  * allocate a cleaned up array with non duplicate flavors on the stack.
63  * So we need to know how much to allocate.
64  */
65 #define	MAX_FLAVORS		6	/* none, sys, dh, krb5, krb5i krb5p */
66 
67 /*
68  * Note: exported_lock is currently used to ensure the integrity of
69  * the secinfo fields.
70  */
71 struct secinfo {
72 	seconfig_t	s_secinfo;	/* /etc/nfssec.conf entry */
73 	unsigned int	s_flags;	/* flags (see below) */
74 	int32_t		s_refcnt;	/* reference count for tracking */
75 					/* how many children (self included) */
76 					/* use this flavor. */
77 	int		s_window;	/* window */
78 	uint_t		s_rootid;	/* UID to use for authorized roots */
79 	int		s_rootcnt;	/* count of root names */
80 	caddr_t		*s_rootnames;	/* array of root names */
81 					/* they are strings for AUTH_DES and */
82 					/* rpc_gss_principal_t for RPCSEC_GSS */
83 };
84 
85 #ifdef _SYSCALL32
86 struct secinfo32 {
87 	seconfig32_t	s_secinfo;	/* /etc/nfssec.conf entry */
88 	uint32_t	s_flags;	/* flags (see below) */
89 	int32_t		s_refcnt;	/* reference count for tracking */
90 					/* how many children (self included) */
91 					/* use this flavor. */
92 	int32_t		s_window;	/* window */
93 	uint32_t	s_rootid;	/* UID to use for authorized roots */
94 	int32_t		s_rootcnt;	/* count of root names */
95 	caddr32_t	s_rootnames;	/* array of root names */
96 					/* they are strings for AUTH_DES and */
97 					/* rpc_gss_principal_t for RPCSEC_GSS */
98 };
99 #endif /* _SYSCALL32 */
100 
101 /*
102  * security negotiation related
103  */
104 
105 #define	SEC_QUERY	0x01	/* query sec modes */
106 
107 struct sec_ol {
108 	int		sec_flags;	/* security nego flags */
109 	uint_t		sec_index;	/* index into sec flavor array */
110 };
111 
112 /*
113  * Per-mode flags (secinfo.s_flags)
114  */
115 #define	M_RO		0x01	/* exported ro to all */
116 #define	M_ROL		0x02	/* exported ro to all listed */
117 #define	M_RW		0x04	/* exported rw to all */
118 #define	M_RWL		0x08	/* exported ro to all listed */
119 #define	M_ROOT		0x10	/* root list is defined */
120 #define	M_4SEC_EXPORTED	0x20	/* this is an explicitly shared flavor */
121 #define	M_NONE		0x40	/* none list is defined */
122 #define	M_MAP		0x80	/* uidmap and/or gidmap is defined */
123 
124 /* invalid secinfo reference count */
125 #define	SEC_REF_INVALID(p) ((p)->s_refcnt < 1)
126 
127 /* last secinfo reference */
128 #define	SEC_REF_LAST(p) ((p)->s_refcnt == 1)
129 
130 /* sec flavor explicitly shared for the exported node */
131 #define	SEC_REF_EXPORTED(p) ((p)->s_flags & M_4SEC_EXPORTED)
132 
133 /* the only reference count left is for referring itself */
134 #define	SEC_REF_SELF(p) (SEC_REF_LAST(p) && SEC_REF_EXPORTED(p))
135 
136 /*
137  * The export information passed to exportfs() (Version 2)
138  */
139 #define	EX_CURRENT_VERSION 2	/* current version of exportdata struct */
140 
141 struct exportdata {
142 	int		ex_version;	/* structure version */
143 	char		*ex_path;	/* exported path */
144 	size_t		ex_pathlen;	/* path length */
145 	int		ex_flags;	/* flags */
146 	unsigned int	ex_anon;	/* uid for unauthenticated requests */
147 	int		ex_seccnt;	/* count of security modes */
148 	struct secinfo	*ex_secinfo;	/* security mode info */
149 	char		*ex_index;	/* index file for public filesystem */
150 	char		*ex_log_buffer;	/* path to logging buffer file */
151 	size_t		ex_log_bufferlen;	/* buffer file path len */
152 	char		*ex_tag;	/* tag used to identify log config */
153 	size_t		ex_taglen;	/* tag length */
154 };
155 
156 #ifdef _SYSCALL32
157 struct exportdata32 {
158 	int32_t		ex_version;	/* structure version */
159 	caddr32_t	ex_path;	/* exported path */
160 	int32_t		ex_pathlen;	/* path length */
161 	int32_t		ex_flags;	/* flags */
162 	uint32_t	ex_anon;	/* uid for unauthenticated requests */
163 	int32_t		ex_seccnt;	/* count of security modes */
164 	caddr32_t	ex_secinfo;	/* security mode info */
165 	caddr32_t	ex_index;	/* index file for public filesystem */
166 	caddr32_t	ex_log_buffer;	/* path to logging buffer file */
167 	int32_t		ex_log_bufferlen;	/* buffer file path len */
168 	caddr32_t	ex_tag;		/* tag used to identify log config */
169 	int32_t		ex_taglen;	/* tag length */
170 };
171 #endif /* _SYSCALL32 */
172 
173 /*
174  * exported vfs flags.
175  */
176 
177 #define	EX_NOSUID	0x01	/* exported with unsetable set[ug]ids */
178 #define	EX_ACLOK	0x02	/* exported with maximal access if acl exists */
179 #define	EX_PUBLIC	0x04	/* exported with public filehandle */
180 #define	EX_NOSUB	0x08	/* no nfs_getfh or MCL below export point */
181 #define	EX_INDEX	0x10	/* exported with index file specified */
182 #define	EX_LOG		0x20	/* logging enabled */
183 #define	EX_LOG_ALLOPS	0x40	/* logging of all RPC operations enabled */
184 				/* by default only operations which affect */
185 				/* transaction logging are enabled */
186 #define	EX_PSEUDO	0x80	/* pseudo filesystem export */
187 #ifdef VOLATILE_FH_TEST
188 #define	EX_VOLFH	0x100	/* XXX nfsv4 fh may expire anytime */
189 #define	EX_VOLRNM	0x200	/* XXX nfsv4 fh expire at rename */
190 #define	EX_VOLMIG	0x400	/* XXX nfsv4 fh expire at migration */
191 #define	EX_NOEXPOPEN	0x800	/* XXX nfsv4 fh no expire with open */
192 #endif /* VOLATILE_FH_TEST */
193 
194 #define	EX_CHARMAP	0x1000	/* NFS may need a character set conversion */
195 #define	EX_NOACLFAB	0x2000	/* If set, NFSv2 and v3 servers won't */
196 				/* fabricate an aclent_t ACL on file systems */
197 				/* that don't support aclent_t ACLs */
198 #define	EX_NOHIDE	0x4000	/* traversable from exported parent */
199 
200 #ifdef	_KERNEL
201 
202 #define	RPC_IDEMPOTENT	0x1	/* idempotent or not */
203 /*
204  * Be very careful about which NFS procedures get the RPC_ALLOWANON bit.
205  * Right now, if this bit is on, we ignore the results of per NFS request
206  * access control.
207  */
208 #define	RPC_ALLOWANON	0x2	/* allow anonymous access */
209 #define	RPC_MAPRESP	0x4	/* use mapped response buffer */
210 #define	RPC_AVOIDWORK	0x8	/* do work avoidance for dups */
211 #define	RPC_PUBLICFH_OK	0x10	/* allow use of public filehandle */
212 
213 /*
214  * RPC_ALL is an or of all above bits to be used with "don't care"
215  * nfsv4 ops. The flags of an nfsv4 request is the bit-AND of the
216  * per-op flags.
217  */
218 #define	RPC_ALL	(RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_AVOIDWORK|RPC_PUBLICFH_OK)
219 
220 
221 #ifdef VOLATILE_FH_TEST
222 struct ex_vol_rename {
223 	nfs_fh4_fmt_t vrn_fh_fmt;
224 	struct ex_vol_rename *vrn_next;
225 };
226 #endif /* VOLATILE_FH_TEST */
227 
228 /*
229  * An auth cache client entry.  This is the umbrella structure and contains all
230  * related auth_cache entries in the authc_tree AVL tree.
231  */
232 struct auth_cache_clnt {
233 	avl_node_t		authc_link;
234 	struct netbuf		authc_addr;	/* address of the client */
235 	krwlock_t		authc_lock;	/* protects authc_tree */
236 	avl_tree_t		authc_tree;	/* auth_cache entries */
237 };
238 
239 /*
240  * An auth cache entry can exist in 6 states.
241  *
242  * A NEW entry was recently allocated and added to the cache.  It does not
243  * contain the valid auth state yet.
244  *
245  * A WAITING entry is one which is actively engaging the user land mountd code
246  * to authenticate or re-authenticate it.  The auth state might not be valid
247  * yet.  The other threads should wait on auth_cv until the retrieving thread
248  * finishes the retrieval and changes the auth cache entry to FRESH, or NEW (in
249  * a case this entry had no valid auth state yet).
250  *
251  * A REFRESHING entry is one which is actively engaging the user land mountd
252  * code to re-authenticate the cache entry.  There is currently no other thread
253  * waiting for the results of the refresh.
254  *
255  * A FRESH entry is one which is valid (it is either newly retrieved or has
256  * been refreshed at least once).
257  *
258  * A STALE entry is one which has been detected to be too old.  The transition
259  * from FRESH to STALE prevents multiple threads from submitting refresh
260  * requests.
261  *
262  * An INVALID entry is one which was either STALE or REFRESHING and was deleted
263  * out of the encapsulating exi.  Since we can't delete it yet, we mark it as
264  * INVALID, which lets the refresh thread know not to work on it and free it
265  * instead.
266  *
267  * Note that the auth state of the entry is valid, even if the entry is STALE.
268  * Just as you can eat stale bread, you can consume a stale cache entry. The
269  * only time the contents change could be during the transition from REFRESHING
270  * or WAITING to FRESH.
271  *
272  * Valid state transitions:
273  *
274  *          alloc
275  *            |
276  *            v
277  *         +-----+
278  *    +--->| NEW |------>free
279  *    |    +-----+
280  *    |       |
281  *    |       v
282  *    |  +---------+
283  *    +<-| WAITING |
284  *    ^  +---------+
285  *    |       |
286  *    |       v
287  *    |       +<--------------------------+<---------------+
288  *    |       |                           ^                |
289  *    |       v                           |                |
290  *    |   +-------+    +-------+    +------------+    +---------+
291  *    +---| FRESH |--->| STALE |--->| REFRESHING |--->| WAITING |
292  *        +-------+    +-------+    +------------+    +---------+
293  *            |            |              |
294  *            |            v              |
295  *            v       +---------+         |
296  *          free<-----| INVALID |<--------+
297  *                    +---------+
298  */
299 typedef enum auth_state {
300 	NFS_AUTH_FRESH,
301 	NFS_AUTH_STALE,
302 	NFS_AUTH_REFRESHING,
303 	NFS_AUTH_INVALID,
304 	NFS_AUTH_NEW,
305 	NFS_AUTH_WAITING
306 } auth_state_t;
307 
308 /*
309  * An authorization cache entry
310  *
311  * Either the state in auth_state will protect the
312  * contents or auth_lock must be held.
313  */
314 struct auth_cache {
315 	avl_node_t		auth_link;
316 	struct auth_cache_clnt	*auth_clnt;
317 	int			auth_flavor;
318 	cred_t			*auth_clnt_cred;
319 	uid_t			auth_srv_uid;
320 	gid_t			auth_srv_gid;
321 	uint_t			auth_srv_ngids;
322 	gid_t			*auth_srv_gids;
323 	int			auth_access;
324 	time_t			auth_time;
325 	time_t			auth_freshness;
326 	auth_state_t		auth_state;
327 	kmutex_t		auth_lock;
328 	kcondvar_t		auth_cv;
329 };
330 
331 #define	AUTH_TABLESIZE	32
332 
333 /*
334  * Structure containing log file meta-data.
335  */
336 struct log_file {
337 	unsigned int	lf_flags;	/* flags (see below) */
338 	int		lf_writers;	/* outstanding writers */
339 	int		lf_refcnt;	/* references to this struct */
340 	caddr_t		lf_path;	/* buffer file location */
341 	vnode_t		*lf_vp;		/* vnode for the buffer file */
342 	kmutex_t	lf_lock;
343 	kcondvar_t	lf_cv_waiters;
344 };
345 
346 /*
347  * log_file and log_buffer flags.
348  */
349 #define	L_WAITING	0x01		/* flush of in-core data to stable */
350 					/* storage in progress */
351 #define	L_PRINTED	0x02		/* error message printed to console */
352 #define	L_ERROR		0x04		/* error condition detected */
353 
354 /*
355  * The logging buffer information.
356  * This structure may be shared by multiple exportinfo structures,
357  * if they share the same buffer file.
358  * This structure contains the basic information about the buffer, such
359  * as it's location in the filesystem.
360  *
361  * 'lb_lock' protects all the fields in this structure except for 'lb_path',
362  * and 'lb_next'.
363  * 'lb_path' is a write-once/read-many field which needs no locking, it is
364  * set before the structure is linked to any exportinfo structure.
365  * 'lb_next' is protected by the log_buffer_list_lock.
366  */
367 struct log_buffer {
368 	unsigned int	lb_flags;	/* L_ONLIST set? */
369 	int		lb_refcnt;	/* references to this struct */
370 	unsigned int	lb_rec_id;	/* used to generate unique id */
371 	caddr_t		lb_path;	/* buffer file pathname */
372 	struct log_file	*lb_logfile;	/* points to log_file structure */
373 	kmutex_t	lb_lock;
374 	struct log_buffer	*lb_next;
375 	kcondvar_t	lb_cv_waiters;
376 	caddr_t		lb_records;	/* linked list of records to write */
377 	int		lb_num_recs;	/* # of records to write */
378 	ssize_t		lb_size_queued; /* number of bytes queued for write */
379 };
380 
381 #define	LOG_BUFFER_HOLD(lbp)	{ \
382 	mutex_enter(&(lbp)->lb_lock); \
383 	(lbp)->lb_refcnt++; \
384 	mutex_exit(&(lbp)->lb_lock); \
385 }
386 
387 #define	LOG_BUFFER_RELE(lbp)	{ \
388 	log_buffer_rele(lbp); \
389 }
390 
391 /*
392  * Structure for character set conversion mapping based on client address.
393  */
394 struct charset_cache {
395 	struct charset_cache *next;
396 	kiconv_t	inbound;
397 	kiconv_t	outbound;
398 	struct sockaddr	client_addr;
399 };
400 
401 /* Forward declarations */
402 struct exportinfo;
403 struct exp_visible;
404 struct svc_req;
405 
406 /*
407  * Treenodes are used to build tree representing every node which is part
408  * of nfs server pseudo namespace. They are connected with both exportinfo
409  * and exp_visible struct. They were introduced to avoid lookup of ".."
410  * in the underlying file system during unshare, which was failing if the
411  * file system was forcibly unmounted or if the directory was removed.
412  * One exp_visible_t can be shared via several treenode_t, i.e.
413  * different tree_vis can point to the same exp_visible_t.
414  * This will happen if some directory is on two different shared paths:
415  * E.g. after share /tmp/a/b1 and share /tmp/a/b2 there will be two treenodes
416  * corresponding to /tmp/a and both will have same value in tree_vis.
417  *
418  *
419  *
420  *     NEW DATA STRUCT         ORIGINAL DATA STRUCT
421  *
422  * ns_root +---+               +----------+
423  *         | / |               |PSEUDO EXP|-->+---+   +---+   +---+
424  *         +---+---------  ----+----------+   | a |-->| k |-->| b |
425  *          /\                                +---+   +---+   +---+
426  *         /  \                                .       .       .
427  *     +---+...\.........  .....................       .       .
428  *    *| a |    \              +----------+            .       .
429  *     +---+-----\-------  ----|REAL EXP a|            .       .
430  *       /        \            +----------+            .       .
431  *      /        +===+...  .............................       .
432  *     /        *| k |         +----------+                    .
433  *    /          +===+---  ----|REAL EXP k|                    .
434  *   /                         +----------+                    .
435  *  +===+................  .....................................
436  * *| b |                      +----------+
437  *  +===+----------------  ----|REAL EXP b|-->+---+
438  *     \                       +----------+   | d |
439  *     +===+.............  ...................+---+
440  *     | d |                   +----------+
441  *     +===+-------------  ----|PSEUDO EXP|-->+---+   +---+
442  *     /                       +----------+   | e |-->| g |
443  * +---+.................  ...................+---+   +---+
444  * | e |                                              .
445  * +---+                                              .
446  *    \                                               .
447  *    +---+..............  ............................
448  *   *| g |                    +----------+
449  *    +---+--------------  ----|REAL EXP g|
450  *                             +----------+
451  *
452  *
453  *
454  * +===+               +---+                    +---+
455  * | b |..mountpoint   | e |..directory/file   *| a |..node is shared
456  * +===+  (VROOT)      +---+                    +---+
457  *
458  *
459  * Bi-directional interconnect:
460  * treenode_t::tree_exi ---------  exportinfo_t::exi_tree
461  * One-way direction connection:
462  * treenode_t::tree_vis .........> exp_visible_t
463  */
464 /* Access to treenode_t is under protection of exported_lock RW_LOCK */
465 typedef struct treenode {
466 	/* support for generic n-ary trees */
467 	struct treenode *tree_parent;
468 	struct treenode *tree_child_first;
469 	struct treenode *tree_sibling; /* next sibling */
470 	/* private, nfs specific part */
471 	struct exportinfo  *tree_exi;
472 	struct exp_visible *tree_vis;
473 } treenode_t;
474 
475 /*
476  * Now that we have links to chase, we can get the zone rootvp just from
477  * an export.  No current-zone-context needed.
478  */
479 #define	EXI_TO_ZONEROOTVP(exi) ((exi)->exi_ne->exi_root->exi_vp)
480 
481 /*
482  * TREE_ROOT checks if the node corresponds to a filesystem root or
483  * the zone's root directory.
484  * TREE_EXPORTED checks if the node is explicitly shared
485  */
486 
487 #define	TREE_ROOT(t) \
488 	((t)->tree_exi != NULL && \
489 	(((t)->tree_exi->exi_vp->v_flag & VROOT) || \
490 	VN_CMP(EXI_TO_ZONEROOTVP((t)->tree_exi), (t)->tree_exi->exi_vp)))
491 
492 #define	TREE_EXPORTED(t) \
493 	((t)->tree_exi && !PSEUDO((t)->tree_exi))
494 
495 #define	EXPTABLESIZE   256
496 
497 struct exp_hash {
498 	struct exportinfo	*prev;  /* ptr to the previous exportinfo */
499 	struct exportinfo	*next;  /* ptr to the next exportinfo */
500 	struct exportinfo	**bckt; /* backpointer to the hash bucket */
501 };
502 
503 /*
504  * A node associated with an export entry on the
505  * list of exported filesystems.
506  *
507  * exi_count+exi_lock protects an individual exportinfo from being freed
508  * when in use.
509  *
510  * You must have the writer lock on exported_lock to add/delete an exportinfo
511  * structure to/from the list.
512  *
513  * exi_volatile_dev maps to VSW_VOLATILEDEV.  It means that the
514  * underlying fs devno can change on each mount.  When set, the server
515  * should not use va_fsid for a GETATTR(FATTR4_FSID) reply.  It must
516  * use exi_fsid because it is guaranteed to be persistent.  This isn't
517  * in any way related to NFS4 volatile filehandles.
518  *
519  * The exi_cache_lock protects the exi_cache AVL trees.
520  */
521 struct exportinfo {
522 	struct exportdata	exi_export;
523 	fsid_t			exi_fsid;
524 	struct fid		exi_fid;
525 	struct exp_hash		fid_hash;
526 	struct exp_hash		path_hash;
527 	struct treenode		*exi_tree;
528 	fhandle_t		exi_fh;
529 	krwlock_t		exi_cache_lock;
530 	kmutex_t		exi_lock;
531 	uint_t			exi_count;
532 	zoneid_t		exi_zoneid;
533 	vnode_t			*exi_vp;
534 	vnode_t			*exi_dvp;
535 	avl_tree_t		*exi_cache[AUTH_TABLESIZE];
536 	struct log_buffer	*exi_logbuffer;
537 	struct exp_visible	*exi_visible;
538 	struct charset_cache	*exi_charset;
539 	unsigned		exi_volatile_dev:1;
540 	unsigned		exi_moved:1;
541 	int			exi_id;
542 	avl_node_t		exi_id_link;
543 	/*
544 	 * Soft-reference/backpointer to zone's nfs_export_t.
545 	 * This allows us access to the zone's rootvp (stored in
546 	 * exi_ne->exi_root->exi_vp) even if the current thread isn't in
547 	 * same-zone context.
548 	 */
549 	struct nfs_export	*exi_ne;
550 #ifdef VOLATILE_FH_TEST
551 	uint32_t		exi_volatile_id;
552 	struct ex_vol_rename	*exi_vol_rename;
553 	kmutex_t		exi_vol_rename_lock;
554 #endif /* VOLATILE_FH_TEST -- keep last! */
555 };
556 
557 typedef struct exportinfo exportinfo_t;
558 typedef struct exportdata exportdata_t;
559 typedef struct secinfo secinfo_t;
560 
561 /*
562  * exp_visible is a visible list per filesystem. It is for filesystems
563  * that may need a limited view of its contents. A pseudo export and
564  * a real export at the mount point (VROOT) which has a subtree shared
565  * has a visible list.
566  *
567  * The exi_visible field is NULL for normal, non-pseudo filesystems
568  * which do not have any subtree exported. If the field is non-null,
569  * it points to a list of visible entries, identified by vis_fid and/or
570  * vis_ino. The presence of a "visible" list means that if this export
571  * can only have a limited view, it can only view the entries in the
572  * exp_visible list. The directories in the fid list comprise paths that
573  * lead to exported directories.
574  *
575  * The vis_count field records the number of paths in this filesystem
576  * that use this directory. The vis_exported field is non-zero if the
577  * entry is an exported directory (leaf node).
578  *
579  * exp_visible itself is not reference counted. Each exp_visible is
580  * referenced twice:
581  * 1) from treenode::tree_vis
582  * 2) linked from exportinfo::exi_visible
583  * The 'owner' of exp_visible is the exportinfo structure. exp_visible should
584  * be always freed only from exportinfo_t, never from treenode::tree_vis.
585  */
586 
587 struct exp_visible {
588 	vnode_t			*vis_vp;
589 	fid_t			vis_fid;
590 	u_longlong_t		vis_ino;
591 	int			vis_count;
592 	int			vis_exported;
593 	struct exp_visible	*vis_next;
594 	struct secinfo		*vis_secinfo;
595 	int			vis_seccnt;
596 	timespec_t		vis_change;
597 };
598 typedef struct exp_visible exp_visible_t;
599 
600 #define	PSEUDO(exi)	((exi)->exi_export.ex_flags & EX_PSEUDO)
601 #define	EXP_LINKED(exi)	((exi)->fid_hash.bckt != NULL)
602 
603 #define	EQFSID(fsidp1, fsidp2)	\
604 	(((fsidp1)->val[0] == (fsidp2)->val[0]) && \
605 	    ((fsidp1)->val[1] == (fsidp2)->val[1]))
606 
607 #define	EQFID(fidp1, fidp2)	\
608 	((fidp1)->fid_len == (fidp2)->fid_len && \
609 	    bcmp((char *)(fidp1)->fid_data, (char *)(fidp2)->fid_data, \
610 	    (uint_t)(fidp1)->fid_len) == 0)
611 
612 #define	exportmatch(exi, fsid, fid)	\
613 	(EQFSID(&(exi)->exi_fsid, (fsid)) && EQFID(&(exi)->exi_fid, (fid)))
614 
615 /*
616  * Returns true iff exported filesystem is read-only to the given host.
617  *
618  * Note:  this macro should be as fast as possible since it's called
619  * on each NFS modification request.
620  */
621 #define	rdonly(ro, vp)	((ro) || vn_is_readonly(vp))
622 #define	rdonly4(req, cs)  \
623 	(vn_is_readonly((cs)->vp) || \
624 	    (nfsauth4_access((cs)->exi, (cs)->vp, (req), (cs)->basecr, NULL, \
625 	    NULL, NULL, NULL) & (NFSAUTH_RO | NFSAUTH_LIMITED)))
626 
627 extern int	nfsauth4_access(struct exportinfo *, vnode_t *,
628     struct svc_req *, cred_t *, uid_t *, gid_t *, uint_t *, gid_t **);
629 extern int	nfsauth4_secinfo_access(struct exportinfo *,
630     struct svc_req *, int, int, cred_t *);
631 extern int	nfsauth_cache_clnt_compar(const void *, const void *);
632 extern int	nfs_fhbcmp(char *, char *, int);
633 extern void	nfs_exportinit(void);
634 extern void	nfs_exportfini(void);
635 extern void	nfs_export_zone_init(nfs_globals_t *);
636 extern void	nfs_export_zone_fini(nfs_globals_t *);
637 extern void	nfs_export_zone_shutdown(nfs_globals_t *);
638 extern int	nfs_export_get_rootfh(nfs_globals_t *);
639 extern int	chk_clnt_sec(struct exportinfo *, struct svc_req *);
640 extern int	makefh(fhandle_t *, struct vnode *, struct exportinfo *);
641 extern int	makefh_ol(fhandle_t *, struct exportinfo *, uint_t);
642 extern int	makefh3(nfs_fh3 *, struct vnode *, struct exportinfo *);
643 extern int	makefh3_ol(nfs_fh3 *, struct exportinfo *, uint_t);
644 extern vnode_t *nfs_fhtovp(fhandle_t *, struct exportinfo *);
645 extern vnode_t *nfs3_fhtovp(nfs_fh3 *, struct exportinfo *);
646 extern struct	exportinfo *checkexport(fsid_t *, struct fid *);
647 extern struct	exportinfo *checkexport4(fsid_t *, struct fid *, vnode_t *);
648 extern void	exi_hold(struct exportinfo *);
649 extern void	exi_rele(struct exportinfo *);
650 extern struct exportinfo *nfs_vptoexi(vnode_t *, vnode_t *, cred_t *, int *,
651     int *, bool_t);
652 extern int	nfs_check_vpexi(vnode_t *, vnode_t *, cred_t *,
653 			struct exportinfo **);
654 extern vnode_t *untraverse(vnode_t *, vnode_t *);
655 extern int	vn_is_nfs_reparse(vnode_t *, cred_t *);
656 extern int	client_is_downrev(struct svc_req *);
657 extern char    *build_symlink(vnode_t *, cred_t *, size_t *);
658 
659 extern fhandle_t nullfh2;	/* for comparing V2 filehandles */
660 
661 typedef struct nfs_export {
662 	/* Root of nfs pseudo namespace */
663 	treenode_t *ns_root;
664 
665 	nfs_globals_t		*ne_globals;	/* "up" pointer */
666 
667 	struct exportinfo *exptable_path_hash[PKP_HASH_SIZE];
668 	struct exportinfo *exptable[EXPTABLESIZE];
669 
670 	/*
671 	 * Read/Write lock that protects the exportinfo list.  This lock
672 	 * must be held when searching or modifiying the exportinfo list.
673 	 */
674 	krwlock_t exported_lock;
675 
676 	/* "public" and default (root) location for public filehandle */
677 	struct exportinfo *exi_public;
678 	struct exportinfo *exi_root;
679 	/* For checking default public file handle */
680 	fid_t exi_rootfid;
681 	/* For comparing V2 filehandles */
682 	fhandle_t nullfh2;
683 
684 	/* The change attribute value of the root of nfs pseudo namespace */
685 	timespec_t ns_root_change;
686 } nfs_export_t;
687 
688 /*
689  * Functions that handle the NFSv4 server namespace
690  */
691 extern exportinfo_t *vis2exi(treenode_t *);
692 extern int	treeclimb_export(struct exportinfo *);
693 extern void	treeclimb_unexport(nfs_export_t *, struct exportinfo *);
694 extern int	nfs_visible(struct exportinfo *, vnode_t *, int *);
695 extern int	nfs_visible_inode(struct exportinfo *, ino64_t,
696 		    struct exp_visible **);
697 extern int	has_visible(struct exportinfo *, vnode_t *);
698 extern void	free_visible(struct exp_visible *);
699 extern int	nfs_exported(struct exportinfo *, vnode_t *);
700 extern struct exportinfo *pseudo_exportfs(nfs_export_t *, vnode_t *, fid_t *,
701 		    struct exp_visible *, struct exportdata *);
702 extern int	vop_fid_pseudo(vnode_t *, fid_t *);
703 extern int	nfs4_vget_pseudo(struct exportinfo *, vnode_t **, fid_t *);
704 extern bool_t	nfs_visible_change(struct exportinfo *, vnode_t *,
705 		    timespec_t *);
706 extern void	tree_update_change(nfs_export_t *, treenode_t *, timespec_t *);
707 extern void	rfs4_clean_state_exi(nfs_export_t *, struct exportinfo *);
708 
709 /*
710  * Functions that handle the NFSv4 server namespace security flavors
711  * information.
712  */
713 extern void	srv_secinfo_exp2pseu(struct exportdata *, struct exportdata *);
714 extern void	srv_secinfo_list_free(struct secinfo *, int);
715 
716 extern nfs_export_t *nfs_get_export();
717 extern void	export_link(nfs_export_t *, struct exportinfo *);
718 extern void	export_unlink(nfs_export_t *, struct exportinfo *);
719 
720 /*
721  * exi_id support
722  */
723 extern kmutex_t  nfs_exi_id_lock;
724 extern avl_tree_t exi_id_tree;
725 extern int exi_id_get_next(void);
726 
727 /*
728  * Two macros for identifying public filehandles.
729  * A v2 public filehandle is 32 zero bytes.
730  * A v3 public filehandle is zero length.
731  */
732 #define	PUBLIC_FH2(fh) \
733 	((fh)->fh_fsid.val[1] == 0 && \
734 	bcmp((fh), &nullfh2, sizeof (fhandle_t)) == 0)
735 
736 #define	PUBLIC_FH3(fh) \
737 	((fh)->fh3_length == 0)
738 
739 extern int	makefh4(nfs_fh4 *, struct vnode *, struct exportinfo *);
740 extern vnode_t *nfs4_fhtovp(nfs_fh4 *, struct exportinfo *, nfsstat4 *);
741 
742 #endif /* _KERNEL */
743 
744 #ifdef	__cplusplus
745 }
746 #endif
747 
748 #endif	/* _NFS_EXPORT_H */
749