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 #include <nfs/nfs_sec.h> 33 #include <nfs/auth.h> 34 #include <sys/vnode.h> 35 #include <nfs/nfs4.h> 36 #include <sys/kiconv.h> 37 38 #ifdef __cplusplus 39 extern "C" { 40 #endif 41 42 /* 43 * nfs pseudo flavor number is owned by IANA. Need to make sure the 44 * Solaris specific NFS_FLAVOR_NOMAP number will not overlap with any 45 * new IANA defined pseudo flavor numbers. The chance for the overlap 46 * is very small since the growth of new flavor numbers is expected 47 * to be limited. 48 */ 49 #define NFS_FLAVOR_NOMAP 999999 /* no nfs flavor mapping */ 50 51 /* 52 * As duplicate flavors can be passed into exportfs in the arguments, we 53 * allocate a cleaned up array with non duplicate flavors on the stack. 54 * So we need to know how much to allocate. 55 */ 56 #define MAX_FLAVORS 6 /* none, sys, dh, krb5, krb5i krb5p */ 57 58 /* 59 * Note: exported_lock is currently used to ensure the integrity of 60 * the secinfo fields. 61 */ 62 struct secinfo { 63 seconfig_t s_secinfo; /* /etc/nfssec.conf entry */ 64 unsigned int s_flags; /* flags (see below) */ 65 int32_t s_refcnt; /* reference count for tracking */ 66 /* how many children (self included) */ 67 /* use this flavor. */ 68 int s_window; /* window */ 69 uint_t s_rootid; /* UID to use for authorized roots */ 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 uint32_t s_rootid; /* UID to use for authorized roots */ 85 int32_t s_rootcnt; /* count of root names */ 86 caddr32_t s_rootnames; /* array of root names */ 87 /* they are strings for AUTH_DES and */ 88 /* rpc_gss_principal_t for RPCSEC_GSS */ 89 }; 90 #endif /* _SYSCALL32 */ 91 92 /* 93 * security negotiation related 94 */ 95 96 #define SEC_QUERY 0x01 /* query sec modes */ 97 98 struct sec_ol { 99 int sec_flags; /* security nego flags */ 100 uint_t sec_index; /* index into sec flavor array */ 101 }; 102 103 /* 104 * Per-mode flags (secinfo.s_flags) 105 */ 106 #define M_RO 0x01 /* exported ro to all */ 107 #define M_ROL 0x02 /* exported ro to all listed */ 108 #define M_RW 0x04 /* exported rw to all */ 109 #define M_RWL 0x08 /* exported ro to all listed */ 110 #define M_ROOT 0x10 /* root list is defined */ 111 #define M_4SEC_EXPORTED 0x20 /* this is an explicitly shared flavor */ 112 #define M_NONE 0x40 /* none list is defined */ 113 114 /* invalid secinfo reference count */ 115 #define SEC_REF_INVALID(p) ((p)->s_refcnt < 1) 116 117 /* last secinfo reference */ 118 #define SEC_REF_LAST(p) ((p)->s_refcnt == 1) 119 120 /* sec flavor explicitly shared for the exported node */ 121 #define SEC_REF_EXPORTED(p) ((p)->s_flags & M_4SEC_EXPORTED) 122 123 /* the only reference count left is for referring itself */ 124 #define SEC_REF_SELF(p) (SEC_REF_LAST(p) && SEC_REF_EXPORTED(p)) 125 126 /* 127 * The export information passed to exportfs() (Version 2) 128 */ 129 #define EX_CURRENT_VERSION 2 /* current version of exportdata struct */ 130 131 struct exportdata { 132 int ex_version; /* structure version */ 133 char *ex_path; /* exported path */ 134 size_t ex_pathlen; /* path length */ 135 int ex_flags; /* flags */ 136 unsigned int ex_anon; /* uid for unauthenticated requests */ 137 int ex_seccnt; /* count of security modes */ 138 struct secinfo *ex_secinfo; /* security mode info */ 139 char *ex_index; /* index file for public filesystem */ 140 char *ex_log_buffer; /* path to logging buffer file */ 141 size_t ex_log_bufferlen; /* buffer file path len */ 142 char *ex_tag; /* tag used to identify log config */ 143 size_t ex_taglen; /* tag length */ 144 }; 145 146 #ifdef _SYSCALL32 147 struct exportdata32 { 148 int32_t ex_version; /* structure version */ 149 caddr32_t ex_path; /* exported path */ 150 int32_t ex_pathlen; /* path length */ 151 int32_t ex_flags; /* flags */ 152 uint32_t ex_anon; /* uid for unauthenticated requests */ 153 int32_t ex_seccnt; /* count of security modes */ 154 caddr32_t ex_secinfo; /* security mode info */ 155 caddr32_t ex_index; /* index file for public filesystem */ 156 caddr32_t ex_log_buffer; /* path to logging buffer file */ 157 int32_t ex_log_bufferlen; /* buffer file path len */ 158 caddr32_t ex_tag; /* tag used to identify log config */ 159 int32_t ex_taglen; /* tag length */ 160 }; 161 #endif /* _SYSCALL32 */ 162 163 /* 164 * exported vfs flags. 165 */ 166 167 #define EX_NOSUID 0x01 /* exported with unsetable set[ug]ids */ 168 #define EX_ACLOK 0x02 /* exported with maximal access if acl exists */ 169 #define EX_PUBLIC 0x04 /* exported with public filehandle */ 170 #define EX_NOSUB 0x08 /* no nfs_getfh or MCL below export point */ 171 #define EX_INDEX 0x10 /* exported with index file specified */ 172 #define EX_LOG 0x20 /* logging enabled */ 173 #define EX_LOG_ALLOPS 0x40 /* logging of all RPC operations enabled */ 174 /* by default only operations which affect */ 175 /* transaction logging are enabled */ 176 #define EX_PSEUDO 0x80 /* pseudo filesystem export */ 177 #ifdef VOLATILE_FH_TEST 178 #define EX_VOLFH 0x100 /* XXX nfsv4 fh may expire anytime */ 179 #define EX_VOLRNM 0x200 /* XXX nfsv4 fh expire at rename */ 180 #define EX_VOLMIG 0x400 /* XXX nfsv4 fh expire at migration */ 181 #define EX_NOEXPOPEN 0x800 /* XXX nfsv4 fh no expire with open */ 182 #endif /* VOLATILE_FH_TEST */ 183 184 #define EX_CHARMAP 0x1000 /* NFS may need a character set conversion */ 185 186 #ifdef _KERNEL 187 188 #define RPC_IDEMPOTENT 0x1 /* idempotent or not */ 189 /* 190 * Be very careful about which NFS procedures get the RPC_ALLOWANON bit. 191 * Right now, it this bit is on, we ignore the results of per NFS request 192 * access control. 193 */ 194 #define RPC_ALLOWANON 0x2 /* allow anonymous access */ 195 #define RPC_MAPRESP 0x4 /* use mapped response buffer */ 196 #define RPC_AVOIDWORK 0x8 /* do work avoidance for dups */ 197 #define RPC_PUBLICFH_OK 0x10 /* allow use of public filehandle */ 198 199 /* 200 * RPC_ALL is an or of all above bits to be used with "don't care" 201 * nfsv4 ops. The flags of an nfsv4 request is the bit-AND of the 202 * per-op flags. 203 */ 204 #define RPC_ALL (RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_AVOIDWORK|RPC_PUBLICFH_OK) 205 206 207 #ifdef VOLATILE_FH_TEST 208 struct ex_vol_rename { 209 nfs_fh4_fmt_t vrn_fh_fmt; 210 struct ex_vol_rename *vrn_next; 211 }; 212 #endif /* VOLATILE_FH_TEST */ 213 214 /* 215 * An authorization cache entry 216 */ 217 struct auth_cache { 218 struct netbuf auth_addr; 219 int auth_flavor; 220 int auth_access; 221 time_t auth_time; 222 struct auth_cache *auth_next; 223 }; 224 225 #define AUTH_TABLESIZE 32 226 227 /* 228 * Structure containing log file meta-data. 229 */ 230 struct log_file { 231 unsigned int lf_flags; /* flags (see below) */ 232 int lf_writers; /* outstanding writers */ 233 int lf_refcnt; /* references to this struct */ 234 caddr_t lf_path; /* buffer file location */ 235 vnode_t *lf_vp; /* vnode for the buffer file */ 236 kmutex_t lf_lock; 237 kcondvar_t lf_cv_waiters; 238 }; 239 240 /* 241 * log_file and log_buffer flags. 242 */ 243 #define L_WAITING 0x01 /* flush of in-core data to stable */ 244 /* storage in progress */ 245 #define L_PRINTED 0x02 /* error message printed to console */ 246 #define L_ERROR 0x04 /* error condition detected */ 247 248 /* 249 * The logging buffer information. 250 * This structure may be shared by multiple exportinfo structures, 251 * if they share the same buffer file. 252 * This structure contains the basic information about the buffer, such 253 * as it's location in the filesystem. 254 * 255 * 'lb_lock' protects all the fields in this structure except for 'lb_path', 256 * and 'lb_next'. 257 * 'lb_path' is a write-once/read-many field which needs no locking, it is 258 * set before the structure is linked to any exportinfo structure. 259 * 'lb_next' is protected by the log_buffer_list_lock. 260 */ 261 struct log_buffer { 262 unsigned int lb_flags; /* L_ONLIST set? */ 263 int lb_refcnt; /* references to this struct */ 264 unsigned int lb_rec_id; /* used to generate unique id */ 265 caddr_t lb_path; /* buffer file pathname */ 266 struct log_file *lb_logfile; /* points to log_file structure */ 267 kmutex_t lb_lock; 268 struct log_buffer *lb_next; 269 kcondvar_t lb_cv_waiters; 270 caddr_t lb_records; /* linked list of records to write */ 271 int lb_num_recs; /* # of records to write */ 272 ssize_t lb_size_queued; /* number of bytes queued for write */ 273 }; 274 275 #define LOG_BUFFER_HOLD(lbp) { \ 276 mutex_enter(&(lbp)->lb_lock); \ 277 (lbp)->lb_refcnt++; \ 278 mutex_exit(&(lbp)->lb_lock); \ 279 } 280 281 #define LOG_BUFFER_RELE(lbp) { \ 282 log_buffer_rele(lbp); \ 283 } 284 285 /* 286 * Structure for character set conversion mapping based on client address. 287 */ 288 struct charset_cache { 289 struct charset_cache *next; 290 kiconv_t inbound; 291 kiconv_t outbound; 292 struct sockaddr client_addr; 293 }; 294 295 /* Forward declarations */ 296 struct exportinfo; 297 struct exp_visible; 298 299 /* 300 * Treenodes are used to build tree representing every node which is part 301 * of nfs server pseudo namespace. 302 * This tree is interconnected with both exportinfo and exp_visible struct. 303 * When there is a need to walk the namespace (either starting in 304 * exportinfo or in exp_visible) we first make a step aside (to the left), 305 * walk up or down as needed, and then we step back (to the right). 306 * 307 * 308 * 309 * NEW DATA STRUCT ORIGINAL DATA STRUCT 310 * 311 * ns_root +---+ +----------+ 312 * | / | |PSEUDO EXP|-->+---+ +---+ +---+ 313 * +---+--------- ----+----------+ | a |-->| k |-->| b | 314 * /\ +---+ +---+ +---+ 315 * / \ . . . 316 * +---+...\......... ..................... . . 317 * *| a | \ +----------+ . . 318 * +---+-----\------- ----|REAL EXP a| . . 319 * / \ +----------+ . . 320 * / +===+... ............................. . 321 * / *| k | +----------+ . 322 * / +===+--- ----|REAL EXP k| . 323 * / +----------+ . 324 * +===+................ ..................................... 325 * *| b | +----------+ 326 * +===+---------------- ----|REAL EXP b|-->+---+ 327 * \ +----------+ | d | 328 * +===+............. ...................+---+ 329 * | d | +----------+ 330 * +===+------------- ----|PSEUDO EXP|-->+---+ +---+ 331 * / +----------+ | e |-->| g | 332 * +---+................. ...................+---+ +---+ 333 * | e | . 334 * +---+ . 335 * \ . 336 * +---+.............. ............................ 337 * *| g | +----------+ 338 * +---+-------------- ----|REAL EXP g| 339 * +----------+ 340 * 341 * 342 * 343 * +===+ +---+ +---+ 344 * | b |..mountpoint | e |..directory/file *| a |..node is shared 345 * +===+ (VROOT) +---+ +---+ 346 * 347 * 348 * Bi-directional interconnect: 349 * 350 * treenode_t::tree_exi --------- exportinfo_t::exi_tree 351 * treenode_t::tree_vis ......... exp_visible_t::vis_tree 352 */ 353 /* Access to treenodei_t is under under protection of exported_lock RW_LOCK */ 354 typedef struct treenode { 355 /* support for generic n-ary trees */ 356 struct treenode *tree_parent; 357 struct treenode *tree_child_first; 358 struct treenode *tree_sibling; /* next sibling */ 359 /* private, nfs specific part */ 360 struct exportinfo *tree_exi; 361 struct exp_visible *tree_vis; 362 } treenode_t; 363 364 /* 365 * TREE_ROOT checks if the node corresponds to a filesystem root 366 * TREE_EXPORTED checks if the node is explicitly shared 367 */ 368 369 #define TREE_ROOT(t) \ 370 ((t)->tree_exi && (t)->tree_exi->exi_vp->v_flag & VROOT) 371 372 #define TREE_EXPORTED(t) \ 373 ((t)->tree_exi && !PSEUDO((t)->tree_exi)) 374 375 /* Root of nfs pseudo namespace */ 376 treenode_t *ns_root; 377 378 #define EXPTABLESIZE 16 379 380 /* 381 * A node associated with an export entry on the 382 * list of exported filesystems. 383 * 384 * exi_count+exi_lock protects an individual exportinfo from being freed 385 * when in use. 386 * You must have the writer lock on exported_lock to add/delete an exportinfo 387 * structure to/from the list. 388 * 389 * exi_volatile_dev maps to VSW_VOLATILEDEV. It means that the 390 * underlying fs devno can change on each mount. When set, the server 391 * should not use va_fsid for a GETATTR(FATTR4_FSID) reply. It must 392 * use exi_fsid because it is guaranteed to be persistent. This isn't 393 * in any way related to NFS4 volatile filehandles. 394 */ 395 struct exportinfo { 396 struct exportdata exi_export; 397 fsid_t exi_fsid; 398 struct fid exi_fid; 399 struct exportinfo *exi_hash; 400 struct treenode *exi_tree; 401 fhandle_t exi_fh; 402 krwlock_t exi_cache_lock; 403 kmutex_t exi_lock; 404 uint_t exi_count; 405 vnode_t *exi_vp; 406 vnode_t *exi_dvp; 407 struct auth_cache *exi_cache[AUTH_TABLESIZE]; 408 struct log_buffer *exi_logbuffer; 409 struct exp_visible *exi_visible; 410 struct charset_cache *exi_charset; 411 unsigned exi_volatile_dev:1; 412 #ifdef VOLATILE_FH_TEST 413 uint32_t exi_volatile_id; 414 struct ex_vol_rename *exi_vol_rename; 415 kmutex_t exi_vol_rename_lock; 416 #endif /* VOLATILE_FH_TEST */ 417 }; 418 419 typedef struct exportinfo exportinfo_t; 420 typedef struct exportdata exportdata_t; 421 typedef struct secinfo secinfo_t; 422 423 /* 424 * exp_visible is a visible list per filesystem. It is for filesystems 425 * that may need a limited view of its contents. A pseudo export and 426 * a real export at the mount point (VROOT) which has a subtree shared 427 * has a visible list. 428 * 429 * The exi_visible field is NULL for normal, non=pseudo filesystems 430 * which do not have any subtree exported. If the field is non-null, 431 * it points to a list of visible entries, identified by vis_fid and/or 432 * vis_ino. The presence of a "visible" list means that if this export 433 * can only have a limited view, it can only view the entries in the 434 * exp_visible list. The directories in the fid list comprise paths that 435 * lead to exported directories. 436 * 437 * The vis_count field records the number of paths in this filesystem 438 * that use this directory. The vis_exported field is non-zero if the 439 * entry is an exported directory (leaf node). 440 */ 441 442 struct exp_visible { 443 vnode_t *vis_vp; 444 fid_t vis_fid; 445 u_longlong_t vis_ino; 446 int vis_count; 447 int vis_exported; 448 struct exp_visible *vis_next; 449 struct treenode *vis_tree; 450 struct secinfo *vis_secinfo; 451 int vis_seccnt; 452 }; 453 typedef struct exp_visible exp_visible_t; 454 455 #define PSEUDO(exi) ((exi)->exi_export.ex_flags & EX_PSEUDO) 456 457 #define EQFSID(fsidp1, fsidp2) \ 458 (((fsidp1)->val[0] == (fsidp2)->val[0]) && \ 459 ((fsidp1)->val[1] == (fsidp2)->val[1])) 460 461 #define EQFID(fidp1, fidp2) \ 462 ((fidp1)->fid_len == (fidp2)->fid_len && \ 463 bcmp((char *)(fidp1)->fid_data, (char *)(fidp2)->fid_data, \ 464 (uint_t)(fidp1)->fid_len) == 0) 465 466 #define exportmatch(exi, fsid, fid) \ 467 (EQFSID(&(exi)->exi_fsid, (fsid)) && EQFID(&(exi)->exi_fid, (fid))) 468 469 /* 470 * Returns true iff exported filesystem is read-only to the given host. 471 * 472 * Note: this macro should be as fast as possible since it's called 473 * on each NFS modification request. 474 */ 475 #define rdonly(exi, req) (nfsauth_access(exi, req) & NFSAUTH_RO) 476 #define rdonly4(exi, vp, req) \ 477 (vn_is_readonly(vp) || \ 478 (nfsauth4_access(exi, vp, req) & (NFSAUTH_RO | NFSAUTH_LIMITED))) 479 480 extern int nfsauth4_access(struct exportinfo *, vnode_t *, 481 struct svc_req *); 482 extern int nfsauth4_secinfo_access(struct exportinfo *, 483 struct svc_req *, int, int); 484 extern int nfs_fhhash(fsid_t *, fid_t *); 485 extern int nfs_fhbcmp(char *, char *, int); 486 extern int nfs_exportinit(void); 487 extern void nfs_exportfini(void); 488 extern int chk_clnt_sec(struct exportinfo *, struct svc_req *req); 489 extern int makefh(fhandle_t *, struct vnode *, struct exportinfo *); 490 extern int makefh_ol(fhandle_t *, struct exportinfo *, uint_t); 491 extern int makefh3(nfs_fh3 *, struct vnode *, struct exportinfo *); 492 extern int makefh3_ol(nfs_fh3 *, struct exportinfo *, uint_t); 493 extern vnode_t *nfs_fhtovp(fhandle_t *, struct exportinfo *); 494 extern vnode_t *nfs3_fhtovp(nfs_fh3 *, struct exportinfo *); 495 extern vnode_t *lm_fhtovp(fhandle_t *fh); 496 extern vnode_t *lm_nfs3_fhtovp(nfs_fh3 *fh); 497 extern struct exportinfo *checkexport(fsid_t *, struct fid *); 498 extern struct exportinfo *checkexport4(fsid_t *, struct fid *, vnode_t *vp); 499 extern void exi_rele(struct exportinfo *); 500 extern struct exportinfo *nfs_vptoexi(vnode_t *, vnode_t *, cred_t *, int *, 501 int *, bool_t); 502 extern int nfs_check_vpexi(vnode_t *, vnode_t *, cred_t *, 503 struct exportinfo **); 504 extern void export_link(struct exportinfo *); 505 extern int export_unlink(fsid_t *, fid_t *, vnode_t *, 506 struct exportinfo **); 507 extern vnode_t *untraverse(vnode_t *); 508 509 /* 510 * Functions that handle the NFSv4 server namespace 511 */ 512 extern exportinfo_t *vis2exi(struct exp_visible *); 513 extern int treeclimb_export(struct exportinfo *); 514 extern void treeclimb_unexport(struct exportinfo *); 515 extern int nfs_visible(struct exportinfo *, vnode_t *, int *); 516 extern int nfs_visible_inode(struct exportinfo *, ino64_t, int *); 517 extern int has_visible(struct exportinfo *, vnode_t *); 518 extern void free_visible(struct exp_visible *); 519 extern int nfs_exported(struct exportinfo *, vnode_t *); 520 extern int pseudo_exportfs(vnode_t *, struct exp_visible *, 521 struct exportdata *, struct exportinfo **); 522 extern int vop_fid_pseudo(vnode_t *, fid_t *fidp); 523 extern int nfs4_vget_pseudo(struct exportinfo *, vnode_t **, fid_t *); 524 /* 525 * Functions that handle the NFSv4 server namespace security flavors 526 * information. 527 */ 528 extern void srv_secinfo_exp2pseu(struct exportdata *, struct exportdata *); 529 extern void srv_secinfo_list_free(struct secinfo *, int); 530 531 /* 532 * "public" and default (root) location for public filehandle 533 */ 534 extern struct exportinfo *exi_public, *exi_root; 535 extern fhandle_t nullfh2; /* for comparing V2 filehandles */ 536 extern krwlock_t exported_lock; 537 extern struct exportinfo *exptable[]; 538 539 /* 540 * Two macros for identifying public filehandles. 541 * A v2 public filehandle is 32 zero bytes. 542 * A v3 public filehandle is zero length. 543 */ 544 #define PUBLIC_FH2(fh) \ 545 ((fh)->fh_fsid.val[1] == 0 && \ 546 bcmp((fh), &nullfh2, sizeof (fhandle_t)) == 0) 547 548 #define PUBLIC_FH3(fh) \ 549 ((fh)->fh3_length == 0) 550 551 extern int makefh4(nfs_fh4 *, struct vnode *, struct exportinfo *); 552 extern vnode_t *nfs4_fhtovp(nfs_fh4 *, struct exportinfo *, nfsstat4 *); 553 554 #endif /* _KERNEL */ 555 556 #ifdef __cplusplus 557 } 558 #endif 559 560 #endif /* _NFS_EXPORT_H */ 561