xref: /titanic_52/usr/src/uts/common/nfs/export.h (revision 55553f719b521a0bb4deab6efc944cd30c1a56aa)
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  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 /*	Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T	*/
27 /*	  All Rights Reserved  	*/
28 
29 #ifndef	_NFS_EXPORT_H
30 #define	_NFS_EXPORT_H
31 
32 #pragma ident	"%Z%%M%	%I%	%E% SMI"
33 
34 #include <nfs/nfs_sec.h>
35 #include <nfs/auth.h>
36 #include <sys/vnode.h>
37 #include <nfs/nfs4.h>
38 
39 #ifdef	__cplusplus
40 extern "C" {
41 #endif
42 
43 /*
44  * nfs pseudo flavor number is owned by IANA. Need to make sure the
45  * Solaris specific NFS_FLAVOR_NOMAP number will not overlap with any
46  * new IANA defined pseudo flavor numbers. The chance for the overlap
47  * is very small since the growth of new flavor numbers is expected
48  * to be limited.
49  */
50 #define	NFS_FLAVOR_NOMAP	999999	/* no nfs flavor mapping */
51 
52 /*
53  * As duplicate flavors can be passed into exportfs in the arguments, we
54  * allocate a cleaned up array with non duplicate flavors on the stack.
55  * So we need to know how much to allocate.
56  */
57 #define	MAX_FLAVORS		6	/* none, sys, dh, krb5, krb5i krb5p */
58 
59 /*
60  * Note: exported_lock is currently used to ensure the integrity of
61  * the secinfo fields.
62  */
63 struct secinfo {
64 	seconfig_t	s_secinfo;	/* /etc/nfssec.conf entry */
65 	unsigned int	s_flags;	/* flags (see below) */
66 	int32_t		s_refcnt;	/* reference count for tracking */
67 					/* how many children (self included) */
68 					/* use this flavor. */
69 	int 		s_window;	/* window */
70 	int		s_rootcnt;	/* count of root names */
71 	caddr_t		*s_rootnames;	/* array of root names */
72 					/* they are strings for AUTH_DES and */
73 					/* rpc_gss_principal_t for RPCSEC_GSS */
74 };
75 
76 #ifdef _SYSCALL32
77 struct secinfo32 {
78 	seconfig32_t	s_secinfo;	/* /etc/nfssec.conf entry */
79 	uint32_t	s_flags;	/* flags (see below) */
80 	int32_t		s_refcnt;	/* reference count for tracking */
81 					/* how many children (self included) */
82 					/* use this flavor. */
83 	int32_t 	s_window;	/* window */
84 	int32_t		s_rootcnt;	/* count of root names */
85 	caddr32_t	s_rootnames;	/* array of root names */
86 					/* they are strings for AUTH_DES and */
87 					/* rpc_gss_principal_t for RPCSEC_GSS */
88 };
89 #endif /* _SYSCALL32 */
90 
91 /*
92  * security negotiation related
93  */
94 
95 #define	SEC_QUERY	0x01	/* query sec modes */
96 
97 struct sec_ol {
98 	int		sec_flags;	/* security nego flags */
99 	uint_t		sec_index;	/* index into sec flavor array */
100 };
101 
102 /*
103  * Per-mode flags (secinfo.s_flags)
104  */
105 #define	M_RO		0x01	/* exported ro to all */
106 #define	M_ROL		0x02	/* exported ro to all listed */
107 #define	M_RW		0x04	/* exported rw to all */
108 #define	M_RWL		0x08	/* exported ro to all listed */
109 #define	M_ROOT		0x10	/* root list is defined */
110 #define	M_4SEC_EXPORTED	0x20	/* this is an explicitly shared flavor */
111 
112 /* invalid secinfo reference count */
113 #define	SEC_REF_INVALID(p) ((p)->s_refcnt < 1)
114 
115 /* last secinfo reference */
116 #define	SEC_REF_LAST(p) ((p)->s_refcnt == 1)
117 
118 /* sec flavor explicitly shared for the exported node */
119 #define	SEC_REF_EXPORTED(p) ((p)->s_flags & M_4SEC_EXPORTED)
120 
121 /* the only reference count left is for referring itself */
122 #define	SEC_REF_SELF(p) (SEC_REF_LAST(p) && SEC_REF_EXPORTED(p))
123 
124 /*
125  * The export information passed to exportfs() (Version 2)
126  */
127 #define	EX_CURRENT_VERSION 2	/* current version of exportdata struct */
128 
129 struct exportdata {
130 	int		ex_version;	/* structure version */
131 	char		*ex_path;	/* exported path */
132 	size_t		ex_pathlen;	/* path length */
133 	int		ex_flags;	/* flags */
134 	unsigned int	ex_anon;	/* uid for unauthenticated requests */
135 	int		ex_seccnt;	/* count of security modes */
136 	struct secinfo	*ex_secinfo;	/* security mode info */
137 	char		*ex_index;	/* index file for public filesystem */
138 	char		*ex_log_buffer;	/* path to logging buffer file */
139 	size_t		ex_log_bufferlen;	/* buffer file path len */
140 	char		*ex_tag;	/* tag used to identify log config */
141 	size_t		ex_taglen;	/* tag length */
142 };
143 
144 #ifdef _SYSCALL32
145 struct exportdata32 {
146 	int32_t		ex_version;	/* structure version */
147 	caddr32_t	ex_path;	/* exported path */
148 	int32_t		ex_pathlen;	/* path length */
149 	int32_t		ex_flags;	/* flags */
150 	uint32_t	ex_anon;	/* uid for unauthenticated requests */
151 	int32_t		ex_seccnt;	/* count of security modes */
152 	caddr32_t	ex_secinfo;	/* security mode info */
153 	caddr32_t	ex_index;	/* index file for public filesystem */
154 	caddr32_t	ex_log_buffer;	/* path to logging buffer file */
155 	int32_t		ex_log_bufferlen;	/* buffer file path len */
156 	caddr32_t	ex_tag;		/* tag used to identify log config */
157 	int32_t		ex_taglen;	/* tag length */
158 };
159 #endif /* _SYSCALL32 */
160 
161 /*
162  * exported vfs flags.
163  */
164 
165 #define	EX_NOSUID	0x01	/* exported with unsetable set[ug]ids */
166 #define	EX_ACLOK	0x02	/* exported with maximal access if acl exists */
167 #define	EX_PUBLIC	0x04	/* exported with public filehandle */
168 #define	EX_NOSUB	0x08	/* no nfs_getfh or MCL below export point */
169 #define	EX_INDEX	0x10	/* exported with index file specified */
170 #define	EX_LOG		0x20	/* logging enabled */
171 #define	EX_LOG_ALLOPS	0x40	/* logging of all RPC operations enabled */
172 				/* by default only operations which affect */
173 				/* transaction logging are enabled */
174 #define	EX_PSEUDO	0x80	/* pseudo filesystem export */
175 #ifdef VOLATILE_FH_TEST
176 #define	EX_VOLFH	0x100	/* XXX nfsv4 fh may expire anytime */
177 #define	EX_VOLRNM	0x200	/* XXX nfsv4 fh expire at rename */
178 #define	EX_VOLMIG	0x400	/* XXX nfsv4 fh expire at migration */
179 #define	EX_NOEXPOPEN	0x800	/* XXX nfsv4 fh no expire with open */
180 #endif /* VOLATILE_FH_TEST */
181 
182 #ifdef	_KERNEL
183 
184 #define	RPC_IDEMPOTENT	0x1	/* idempotent or not */
185 /*
186  * Be very careful about which NFS procedures get the RPC_ALLOWANON bit.
187  * Right now, it this bit is on, we ignore the results of per NFS request
188  * access control.
189  */
190 #define	RPC_ALLOWANON	0x2	/* allow anonymous access */
191 #define	RPC_MAPRESP	0x4	/* use mapped response buffer */
192 #define	RPC_AVOIDWORK	0x8	/* do work avoidance for dups */
193 #define	RPC_PUBLICFH_OK	0x10	/* allow use of public filehandle */
194 
195 /*
196  * RPC_ALL is an or of all above bits to be used with "don't care"
197  * nfsv4 ops. The flags of an nfsv4 request is the bit-AND of the
198  * per-op flags.
199  */
200 #define	RPC_ALL	(RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_AVOIDWORK|RPC_PUBLICFH_OK)
201 
202 
203 #ifdef VOLATILE_FH_TEST
204 struct ex_vol_rename {
205 	nfs_fh4_fmt_t vrn_fh_fmt;
206 	struct ex_vol_rename *vrn_next;
207 };
208 #endif /* VOLATILE_FH_TEST */
209 
210 /*
211  * An authorization cache entry
212  */
213 struct auth_cache {
214 	struct netbuf		auth_addr;
215 	int			auth_flavor;
216 	int			auth_access;
217 	time_t			auth_time;
218 	struct auth_cache	*auth_next;
219 };
220 
221 #define	AUTH_TABLESIZE	32
222 
223 /*
224  * Structure containing log file meta-data.
225  */
226 struct log_file {
227 	unsigned int	lf_flags;	/* flags (see below) */
228 	int		lf_writers;	/* outstanding writers */
229 	int		lf_refcnt;	/* references to this struct */
230 	caddr_t		lf_path;	/* buffer file location */
231 	vnode_t		*lf_vp;		/* vnode for the buffer file */
232 	kmutex_t	lf_lock;
233 	kcondvar_t	lf_cv_waiters;
234 };
235 
236 /*
237  * log_file and log_buffer flags.
238  */
239 #define	L_WAITING	0x01		/* flush of in-core data to stable */
240 					/* storage in progress */
241 #define	L_PRINTED	0x02		/* error message printed to console */
242 #define	L_ERROR		0x04		/* error condition detected */
243 
244 /*
245  * The logging buffer information.
246  * This structure may be shared by multiple exportinfo structures,
247  * if they share the same buffer file.
248  * This structure contains the basic information about the buffer, such
249  * as it's location in the filesystem.
250  *
251  * 'lb_lock' protects all the fields in this structure except for 'lb_path',
252  * and 'lb_next'.
253  * 'lb_path' is a write-once/read-many field which needs no locking, it is
254  * set before the structure is linked to any exportinfo structure.
255  * 'lb_next' is protected by the log_buffer_list_lock.
256  */
257 struct log_buffer {
258 	unsigned int	lb_flags;	/* L_ONLIST set? */
259 	int		lb_refcnt;	/* references to this struct */
260 	unsigned int	lb_rec_id;	/* used to generate unique id */
261 	caddr_t		lb_path;	/* buffer file pathname */
262 	struct log_file	*lb_logfile;	/* points to log_file structure */
263 	kmutex_t	lb_lock;
264 	struct log_buffer	*lb_next;
265 	kcondvar_t	lb_cv_waiters;
266 	caddr_t		lb_records;	/* linked list of records to write */
267 	int		lb_num_recs;	/* # of records to write */
268 	ssize_t		lb_size_queued; /* number of bytes queued for write */
269 };
270 
271 #define	LOG_BUFFER_HOLD(lbp)	{ \
272 	mutex_enter(&(lbp)->lb_lock); \
273 	(lbp)->lb_refcnt++; \
274 	mutex_exit(&(lbp)->lb_lock); \
275 }
276 
277 #define	LOG_BUFFER_RELE(lbp)	{ \
278 	log_buffer_rele(lbp); \
279 }
280 
281 /* Forward declarations */
282 struct exportinfo;
283 struct exp_visible;
284 
285 /*
286  * Treenodes are used to build tree representing every node which is part
287  * of nfs server pseudo namespace.
288  * This tree is interconnected with both exportinfo and exp_visible struct.
289  * When there is a need to walk the namespace (either starting in
290  * exportinfo or in exp_visible) we first make a step aside (to the left),
291  * walk up or down as needed, and then we step back (to the right).
292  *
293  *
294  *
295  *     NEW DATA STRUCT         ORIGINAL DATA STRUCT
296  *
297  * ns_root +---+               +----------+
298  *         | / |               |PSEUDO EXP|-->+---+   +---+   +---+
299  *         +---+---------  ----+----------+   | a |-->| k |-->| b |
300  *          /\                                +---+   +---+   +---+
301  *         /  \                                .       .       .
302  *     +---+...\.........  .....................       .       .
303  *    *| a |    \              +----------+            .       .
304  *     +---+-----\-------  ----|REAL EXP a|            .       .
305  *       /        \            +----------+            .       .
306  *      /        +===+...  .............................       .
307  *     /        *| k |         +----------+                    .
308  *    /          +===+---  ----|REAL EXP k|                    .
309  *   /                         +----------+                    .
310  *  +===+................  .....................................
311  * *| b |                      +----------+
312  *  +===+----------------  ----|REAL EXP b|-->+---+
313  *     \                       +----------+   | d |
314  *     +===+.............  ...................+---+
315  *     | d |                   +----------+
316  *     +===+-------------  ----|PSEUDO EXP|-->+---+   +---+
317  *     /                       +----------+   | e |-->| g |
318  * +---+.................  ...................+---+   +---+
319  * | e |                                              .
320  * +---+                                              .
321  *    \                                               .
322  *    +---+..............  ............................
323  *   *| g |                    +----------+
324  *    +---+--------------  ----|REAL EXP g|
325  *                             +----------+
326  *
327  *
328  *
329  * +===+               +---+                    +---+
330  * | b |..mountpoint   | e |..directory/file   *| a |..node is shared
331  * +===+  (VROOT)      +---+                    +---+
332  *
333  *
334  * Bi-directional interconnect:
335  *
336  * treenode_t::tree_exi ---------  exportinfo_t::exi_tree
337  * treenode_t::tree_vis ......... exp_visible_t::vis_tree
338  */
339 /* Access to treenodei_t is under under protection of exported_lock RW_LOCK */
340 typedef struct treenode {
341 	/* support for generic n-ary trees */
342 	struct treenode *tree_parent;
343 	struct treenode *tree_child_first;
344 	struct treenode *tree_sibling; /* next sibling */
345 	/* private, nfs specific part */
346 	struct exportinfo  *tree_exi;
347 	struct exp_visible *tree_vis;
348 } treenode_t;
349 
350 /*
351  * TREE_ROOT checks if the node corresponds to a filesystem root
352  * TREE_EXPORTED checks if the node is explicitly shared
353  */
354 
355 #define	TREE_ROOT(t) \
356 	((t)->tree_exi && (t)->tree_exi->exi_vp->v_flag & VROOT)
357 
358 #define	TREE_EXPORTED(t) \
359 	((t)->tree_exi && !PSEUDO((t)->tree_exi))
360 
361 /* Root of nfs pseudo namespace */
362 treenode_t *ns_root;
363 
364 #define	EXPTABLESIZE	16
365 
366 /*
367  * A node associated with an export entry on the
368  * list of exported filesystems.
369  *
370  * exi_count+exi_lock protects an individual exportinfo from being freed
371  * when in use.
372  * You must have the writer lock on exported_lock to add/delete an exportinfo
373  * structure to/from the list.
374  *
375  * exi_volatile_dev maps to VSW_VOLATILEDEV.  It means that the
376  * underlying fs devno can change on each mount.  When set, the server
377  * should not use va_fsid for a GETATTR(FATTR4_FSID) reply.  It must
378  * use exi_fsid because it is guaranteed to be persistent.  This isn't
379  * in any way related to NFS4 volatile filehandles.
380  */
381 struct exportinfo {
382 	struct exportdata	exi_export;
383 	fsid_t			exi_fsid;
384 	struct fid		exi_fid;
385 	struct exportinfo	*exi_hash;
386 	struct treenode		*exi_tree;
387 	fhandle_t		exi_fh;
388 	krwlock_t		exi_cache_lock;
389 	kmutex_t		exi_lock;
390 	uint_t			exi_count;
391 	vnode_t			*exi_vp;
392 	vnode_t			*exi_dvp;
393 	struct auth_cache	*exi_cache[AUTH_TABLESIZE];
394 	struct log_buffer	*exi_logbuffer;
395 	struct exp_visible	*exi_visible;
396 	unsigned		exi_volatile_dev:1;
397 #ifdef VOLATILE_FH_TEST
398 	uint32_t		exi_volatile_id;
399 	struct ex_vol_rename	*exi_vol_rename;
400 	kmutex_t		exi_vol_rename_lock;
401 #endif /* VOLATILE_FH_TEST */
402 };
403 
404 typedef struct exportinfo exportinfo_t;
405 typedef struct exportdata exportdata_t;
406 typedef struct secinfo secinfo_t;
407 
408 /*
409  * exp_visible is a visible list per filesystem. It is for filesystems
410  * that may need a limited view of its contents. A pseudo export and
411  * a real export at the mount point (VROOT) which has a subtree shared
412  * has a visible list.
413  *
414  * The exi_visible field is NULL for normal, non=pseudo filesystems
415  * which do not have any subtree exported. If the field is non-null,
416  * it points to a list of visible entries, identified by vis_fid and/or
417  * vis_ino. The presence of a "visible" list means that if this export
418  * can only have a limited view, it can only view the entries in the
419  * exp_visible list. The directories in the fid list comprise paths that
420  * lead to exported directories.
421  *
422  * The vis_count field records the number of paths in this filesystem
423  * that use this directory. The vis_exported field is non-zero if the
424  * entry is an exported directory (leaf node).
425  */
426 
427 struct exp_visible {
428 	vnode_t			*vis_vp;
429 	fid_t			vis_fid;
430 	u_longlong_t		vis_ino;
431 	int			vis_count;
432 	int			vis_exported;
433 	struct exp_visible	*vis_next;
434 	struct treenode		*vis_tree;
435 	struct secinfo		*vis_secinfo;
436 	int			vis_seccnt;
437 };
438 typedef struct exp_visible exp_visible_t;
439 
440 #define	PSEUDO(exi)	((exi)->exi_export.ex_flags & EX_PSEUDO)
441 
442 #define	EQFSID(fsidp1, fsidp2)	\
443 	(((fsidp1)->val[0] == (fsidp2)->val[0]) && \
444 	    ((fsidp1)->val[1] == (fsidp2)->val[1]))
445 
446 #define	EQFID(fidp1, fidp2)	\
447 	((fidp1)->fid_len == (fidp2)->fid_len && \
448 	    bcmp((char *)(fidp1)->fid_data, (char *)(fidp2)->fid_data, \
449 	    (uint_t)(fidp1)->fid_len) == 0)
450 
451 #define	exportmatch(exi, fsid, fid)	\
452 	(EQFSID(&(exi)->exi_fsid, (fsid)) && EQFID(&(exi)->exi_fid, (fid)))
453 
454 /*
455  * Returns true iff exported filesystem is read-only to the given host.
456  *
457  * Note:  this macro should be as fast as possible since it's called
458  * on each NFS modification request.
459  */
460 #define	rdonly(exi, req)  (nfsauth_access(exi, req) & NFSAUTH_RO)
461 #define	rdonly4(exi, vp, req)  \
462 	(vn_is_readonly(vp) || \
463 	    (nfsauth4_access(exi, vp, req) & (NFSAUTH_RO | NFSAUTH_LIMITED)))
464 
465 extern int	nfsauth4_access(struct exportinfo *, vnode_t *,
466 				struct svc_req *);
467 extern int	nfsauth4_secinfo_access(struct exportinfo *,
468 				struct svc_req *, int, int);
469 extern int	nfs_fhhash(fsid_t *, fid_t *);
470 extern int	nfs_fhbcmp(char *, char *, int);
471 extern int	nfs_exportinit(void);
472 extern void	nfs_exportfini(void);
473 extern int	chk_clnt_sec(struct exportinfo *, struct svc_req *req);
474 extern int	makefh(fhandle_t *, struct vnode *, struct exportinfo *);
475 extern int	makefh_ol(fhandle_t *, struct exportinfo *, uint_t);
476 extern int	makefh3(nfs_fh3 *, struct vnode *, struct exportinfo *);
477 extern int	makefh3_ol(nfs_fh3 *, struct exportinfo *, uint_t);
478 extern vnode_t *nfs_fhtovp(fhandle_t *, struct exportinfo *);
479 extern vnode_t *nfs3_fhtovp(nfs_fh3 *, struct exportinfo *);
480 extern vnode_t *lm_fhtovp(fhandle_t *fh);
481 extern vnode_t *lm_nfs3_fhtovp(nfs_fh3 *fh);
482 extern struct	exportinfo *checkexport(fsid_t *, struct fid *);
483 extern struct	exportinfo *checkexport4(fsid_t *, struct fid *, vnode_t *vp);
484 extern void	exi_rele(struct exportinfo *);
485 extern struct exportinfo *nfs_vptoexi(vnode_t *, vnode_t *, cred_t *, int *,
486     int *, bool_t);
487 extern int	nfs_check_vpexi(vnode_t *, vnode_t *, cred_t *,
488 			struct exportinfo **);
489 extern void	export_link(struct exportinfo *);
490 extern int	export_unlink(fsid_t *, fid_t *, vnode_t *,
491 			struct exportinfo **);
492 extern vnode_t *untraverse(vnode_t *);
493 
494 /*
495  * Functions that handle the NFSv4 server namespace
496  */
497 extern exportinfo_t *vis2exi(struct exp_visible *);
498 extern int	treeclimb_export(struct exportinfo *);
499 extern void	treeclimb_unexport(struct exportinfo *);
500 extern int	nfs_visible(struct exportinfo *, vnode_t *, int *);
501 extern int	nfs_visible_inode(struct exportinfo *, ino64_t, int *);
502 extern int	has_visible(struct exportinfo *, vnode_t *);
503 extern void	free_visible(struct exp_visible *);
504 extern int	nfs_exported(struct exportinfo *, vnode_t *);
505 extern int	pseudo_exportfs(vnode_t *, struct exp_visible *,
506     struct exportdata *, struct exportinfo **);
507 extern int	vop_fid_pseudo(vnode_t *, fid_t *fidp);
508 extern int	nfs4_vget_pseudo(struct exportinfo *, vnode_t **, fid_t *);
509 /*
510  * Functions that handle the NFSv4 server namespace security flavors
511  * information.
512  */
513 extern void	srv_secinfo_exp2pseu(struct exportdata *, struct exportdata *);
514 extern void	srv_secinfo_list_free(struct secinfo *, int);
515 
516 /*
517  * "public" and default (root) location for public filehandle
518  */
519 extern struct exportinfo *exi_public, *exi_root;
520 extern fhandle_t nullfh2;	/* for comparing V2 filehandles */
521 extern krwlock_t exported_lock;
522 extern struct exportinfo *exptable[];
523 
524 /*
525  * Two macros for identifying public filehandles.
526  * A v2 public filehandle is 32 zero bytes.
527  * A v3 public filehandle is zero length.
528  */
529 #define	PUBLIC_FH2(fh) \
530 	((fh)->fh_fsid.val[1] == 0 && \
531 	bcmp((fh), &nullfh2, sizeof (fhandle_t)) == 0)
532 
533 #define	PUBLIC_FH3(fh) \
534 	((fh)->fh3_length == 0)
535 
536 extern int	makefh4(nfs_fh4 *, struct vnode *, struct exportinfo *);
537 extern vnode_t *nfs4_fhtovp(nfs_fh4 *, struct exportinfo *, nfsstat4 *);
538 
539 #endif /* _KERNEL */
540 
541 #ifdef	__cplusplus
542 }
543 #endif
544 
545 #endif	/* _NFS_EXPORT_H */
546