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 /* 27 * Copyright (c) 1983,1984,1985,1986,1987,1988,1989 AT&T. 28 * All Rights Reserved 29 */ 30 31 #pragma ident "%Z%%M% %I% %E% SMI" 32 33 #include <sys/param.h> 34 #include <sys/types.h> 35 #include <sys/systm.h> 36 #include <sys/cred.h> 37 #include <sys/buf.h> 38 #include <sys/vfs.h> 39 #include <sys/vfs_opreg.h> 40 #include <sys/vnode.h> 41 #include <sys/uio.h> 42 #include <sys/errno.h> 43 #include <sys/sysmacros.h> 44 #include <sys/statvfs.h> 45 #include <sys/kmem.h> 46 #include <sys/dirent.h> 47 #include <sys/cmn_err.h> 48 #include <sys/debug.h> 49 #include <sys/systeminfo.h> 50 #include <sys/flock.h> 51 #include <sys/pathname.h> 52 #include <sys/nbmlock.h> 53 #include <sys/share.h> 54 #include <sys/atomic.h> 55 #include <sys/policy.h> 56 #include <sys/fem.h> 57 #include <sys/sdt.h> 58 #include <sys/ddi.h> 59 60 #include <rpc/types.h> 61 #include <rpc/auth.h> 62 #include <rpc/rpcsec_gss.h> 63 #include <rpc/svc.h> 64 65 #include <nfs/nfs.h> 66 #include <nfs/export.h> 67 #include <nfs/lm.h> 68 #include <nfs/nfs4.h> 69 70 #include <sys/strsubr.h> 71 #include <sys/strsun.h> 72 73 #include <inet/common.h> 74 #include <inet/ip.h> 75 #include <inet/ip6.h> 76 77 #include <sys/tsol/label.h> 78 #include <sys/tsol/tndb.h> 79 80 #define RFS4_MAXLOCK_TRIES 4 /* Try to get the lock this many times */ 81 static int rfs4_maxlock_tries = RFS4_MAXLOCK_TRIES; 82 #define RFS4_LOCK_DELAY 10 /* Milliseconds */ 83 static clock_t rfs4_lock_delay = RFS4_LOCK_DELAY; 84 85 /* End of Tunables */ 86 87 /* 88 * Used to bump the stateid4.seqid value and show changes in the stateid 89 */ 90 #define next_stateid(sp) (++(sp)->bits.chgseq) 91 92 /* 93 * RFS4_MINLEN_ENTRY4: XDR-encoded size of smallest possible dirent. 94 * This is used to return NFS4ERR_TOOSMALL when clients specify 95 * maxcount that isn't large enough to hold the smallest possible 96 * XDR encoded dirent. 97 * 98 * sizeof cookie (8 bytes) + 99 * sizeof name_len (4 bytes) + 100 * sizeof smallest (padded) name (4 bytes) + 101 * sizeof bitmap4_len (12 bytes) + NOTE: we always encode len=2 bm4 102 * sizeof attrlist4_len (4 bytes) + 103 * sizeof next boolean (4 bytes) 104 * 105 * RFS4_MINLEN_RDDIR4: XDR-encoded size of READDIR op reply containing 106 * the smallest possible entry4 (assumes no attrs requested). 107 * sizeof nfsstat4 (4 bytes) + 108 * sizeof verifier4 (8 bytes) + 109 * sizeof entry4list bool (4 bytes) + 110 * sizeof entry4 (36 bytes) + 111 * sizeof eof bool (4 bytes) 112 * 113 * RFS4_MINLEN_RDDIR_BUF: minimum length of buffer server will provide to 114 * VOP_READDIR. Its value is the size of the maximum possible dirent 115 * for solaris. The DIRENT64_RECLEN macro returns the size of dirent 116 * required for a given name length. MAXNAMELEN is the maximum 117 * filename length allowed in Solaris. The first two DIRENT64_RECLEN() 118 * macros are to allow for . and .. entries -- just a minor tweak to try 119 * and guarantee that buffer we give to VOP_READDIR will be large enough 120 * to hold ., .., and the largest possible solaris dirent64. 121 */ 122 #define RFS4_MINLEN_ENTRY4 36 123 #define RFS4_MINLEN_RDDIR4 (4 + NFS4_VERIFIER_SIZE + 4 + RFS4_MINLEN_ENTRY4 + 4) 124 #define RFS4_MINLEN_RDDIR_BUF \ 125 (DIRENT64_RECLEN(1) + DIRENT64_RECLEN(2) + DIRENT64_RECLEN(MAXNAMELEN)) 126 127 /* 128 * It would be better to pad to 4 bytes since that's what XDR would do, 129 * but the dirents UFS gives us are already padded to 8, so just take 130 * what we're given. Dircount is only a hint anyway. Currently the 131 * solaris kernel is ASCII only, so there's no point in calling the 132 * UTF8 functions. 133 * 134 * dirent64: named padded to provide 8 byte struct alignment 135 * d_ino(8) + d_off(8) + d_reclen(2) + d_name(namelen + null(1) + pad) 136 * 137 * cookie: uint64_t + utf8namelen: uint_t + utf8name padded to 8 bytes 138 * 139 */ 140 #define DIRENT64_TO_DIRCOUNT(dp) \ 141 (3 * BYTES_PER_XDR_UNIT + DIRENT64_NAMELEN((dp)->d_reclen)) 142 143 time_t rfs4_start_time; /* Initialized in rfs4_srvrinit */ 144 145 static sysid_t lockt_sysid; /* dummy sysid for all LOCKT calls */ 146 147 u_longlong_t nfs4_srv_caller_id; 148 uint_t nfs4_srv_vkey = 0; 149 150 verifier4 Write4verf; 151 verifier4 Readdir4verf; 152 153 void rfs4_init_compound_state(struct compound_state *); 154 155 static void nullfree(caddr_t); 156 static void rfs4_op_inval(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 157 struct compound_state *); 158 static void rfs4_op_access(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 159 struct compound_state *); 160 static void rfs4_op_close(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 161 struct compound_state *); 162 static void rfs4_op_commit(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 163 struct compound_state *); 164 static void rfs4_op_create(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 165 struct compound_state *); 166 static void rfs4_op_create_free(nfs_resop4 *resop); 167 static void rfs4_op_delegreturn(nfs_argop4 *, nfs_resop4 *, 168 struct svc_req *, struct compound_state *); 169 static void rfs4_op_delegpurge(nfs_argop4 *, nfs_resop4 *, 170 struct svc_req *, struct compound_state *); 171 static void rfs4_op_getattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 172 struct compound_state *); 173 static void rfs4_op_getattr_free(nfs_resop4 *); 174 static void rfs4_op_getfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 175 struct compound_state *); 176 static void rfs4_op_getfh_free(nfs_resop4 *); 177 static void rfs4_op_illegal(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 178 struct compound_state *); 179 static void rfs4_op_link(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 180 struct compound_state *); 181 static void rfs4_op_lock(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 182 struct compound_state *); 183 static void lock_denied_free(nfs_resop4 *); 184 static void rfs4_op_locku(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 185 struct compound_state *); 186 static void rfs4_op_lockt(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 187 struct compound_state *); 188 static void rfs4_op_lookup(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 189 struct compound_state *); 190 static void rfs4_op_lookupp(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 191 struct compound_state *); 192 static void rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, 193 struct svc_req *req, struct compound_state *cs); 194 static void rfs4_op_nverify(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 195 struct compound_state *); 196 static void rfs4_op_open(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 197 struct compound_state *); 198 static void rfs4_op_open_confirm(nfs_argop4 *, nfs_resop4 *, 199 struct svc_req *, struct compound_state *); 200 static void rfs4_op_open_downgrade(nfs_argop4 *, nfs_resop4 *, 201 struct svc_req *, struct compound_state *); 202 static void rfs4_op_putfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 203 struct compound_state *); 204 static void rfs4_op_putpubfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 205 struct compound_state *); 206 static void rfs4_op_putrootfh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 207 struct compound_state *); 208 static void rfs4_op_read(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 209 struct compound_state *); 210 static void rfs4_op_read_free(nfs_resop4 *); 211 static void rfs4_op_readdir_free(nfs_resop4 *resop); 212 static void rfs4_op_readlink(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 213 struct compound_state *); 214 static void rfs4_op_readlink_free(nfs_resop4 *); 215 static void rfs4_op_release_lockowner(nfs_argop4 *, nfs_resop4 *, 216 struct svc_req *, struct compound_state *); 217 static void rfs4_op_remove(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 218 struct compound_state *); 219 static void rfs4_op_rename(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 220 struct compound_state *); 221 static void rfs4_op_renew(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 222 struct compound_state *); 223 static void rfs4_op_restorefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 224 struct compound_state *); 225 static void rfs4_op_savefh(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 226 struct compound_state *); 227 static void rfs4_op_setattr(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 228 struct compound_state *); 229 static void rfs4_op_verify(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 230 struct compound_state *); 231 static void rfs4_op_write(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 232 struct compound_state *); 233 static void rfs4_op_setclientid(nfs_argop4 *, nfs_resop4 *, 234 struct svc_req *, struct compound_state *); 235 static void rfs4_op_setclientid_confirm(nfs_argop4 *, nfs_resop4 *, 236 struct svc_req *req, struct compound_state *); 237 static void rfs4_op_secinfo(nfs_argop4 *, nfs_resop4 *, struct svc_req *, 238 struct compound_state *); 239 static void rfs4_op_secinfo_free(nfs_resop4 *); 240 241 static nfsstat4 check_open_access(uint32_t, 242 struct compound_state *, struct svc_req *); 243 nfsstat4 rfs4_client_sysid(rfs4_client_t *, sysid_t *); 244 static int vop_shrlock(vnode_t *, int, struct shrlock *, int); 245 static int rfs4_shrlock(rfs4_state_t *, int); 246 static int rfs4_share(rfs4_state_t *); 247 void rfs4_ss_clid(rfs4_client_t *, struct svc_req *); 248 249 /* 250 * translation table for attrs 251 */ 252 struct nfs4_ntov_table { 253 union nfs4_attr_u *na; 254 uint8_t amap[NFS4_MAXNUM_ATTRS]; 255 int attrcnt; 256 bool_t vfsstat; 257 }; 258 259 static void nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp); 260 static void nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp, 261 struct nfs4_svgetit_arg *sargp); 262 263 static nfsstat4 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, 264 struct compound_state *cs, struct nfs4_svgetit_arg *sargp, 265 struct nfs4_ntov_table *ntovp, nfs4_attr_cmd_t cmd); 266 267 fem_t *deleg_rdops; 268 fem_t *deleg_wrops; 269 270 rfs4_servinst_t *rfs4_cur_servinst = NULL; /* current server instance */ 271 kmutex_t rfs4_servinst_lock; /* protects linked list */ 272 int rfs4_seen_first_compound; /* set first time we see one */ 273 274 /* 275 * NFS4 op dispatch table 276 */ 277 278 struct rfsv4disp { 279 void (*dis_proc)(); /* proc to call */ 280 void (*dis_resfree)(); /* frees space allocated by proc */ 281 int dis_flags; /* RPC_IDEMPOTENT, etc... */ 282 }; 283 284 static struct rfsv4disp rfsv4disptab[] = { 285 /* 286 * NFS VERSION 4 287 */ 288 289 /* RFS_NULL = 0 */ 290 {rfs4_op_illegal, nullfree, 0}, 291 292 /* UNUSED = 1 */ 293 {rfs4_op_illegal, nullfree, 0}, 294 295 /* UNUSED = 2 */ 296 {rfs4_op_illegal, nullfree, 0}, 297 298 /* OP_ACCESS = 3 */ 299 {rfs4_op_access, nullfree, RPC_IDEMPOTENT}, 300 301 /* OP_CLOSE = 4 */ 302 {rfs4_op_close, nullfree, 0}, 303 304 /* OP_COMMIT = 5 */ 305 {rfs4_op_commit, nullfree, RPC_IDEMPOTENT}, 306 307 /* OP_CREATE = 6 */ 308 {rfs4_op_create, nullfree, 0}, 309 310 /* OP_DELEGPURGE = 7 */ 311 {rfs4_op_delegpurge, nullfree, 0}, 312 313 /* OP_DELEGRETURN = 8 */ 314 {rfs4_op_delegreturn, nullfree, 0}, 315 316 /* OP_GETATTR = 9 */ 317 {rfs4_op_getattr, rfs4_op_getattr_free, RPC_IDEMPOTENT}, 318 319 /* OP_GETFH = 10 */ 320 {rfs4_op_getfh, rfs4_op_getfh_free, RPC_ALL}, 321 322 /* OP_LINK = 11 */ 323 {rfs4_op_link, nullfree, 0}, 324 325 /* OP_LOCK = 12 */ 326 {rfs4_op_lock, lock_denied_free, 0}, 327 328 /* OP_LOCKT = 13 */ 329 {rfs4_op_lockt, lock_denied_free, 0}, 330 331 /* OP_LOCKU = 14 */ 332 {rfs4_op_locku, nullfree, 0}, 333 334 /* OP_LOOKUP = 15 */ 335 {rfs4_op_lookup, nullfree, (RPC_IDEMPOTENT|RPC_PUBLICFH_OK)}, 336 337 /* OP_LOOKUPP = 16 */ 338 {rfs4_op_lookupp, nullfree, (RPC_IDEMPOTENT|RPC_PUBLICFH_OK)}, 339 340 /* OP_NVERIFY = 17 */ 341 {rfs4_op_nverify, nullfree, RPC_IDEMPOTENT}, 342 343 /* OP_OPEN = 18 */ 344 {rfs4_op_open, rfs4_free_reply, 0}, 345 346 /* OP_OPENATTR = 19 */ 347 {rfs4_op_openattr, nullfree, 0}, 348 349 /* OP_OPEN_CONFIRM = 20 */ 350 {rfs4_op_open_confirm, nullfree, 0}, 351 352 /* OP_OPEN_DOWNGRADE = 21 */ 353 {rfs4_op_open_downgrade, nullfree, 0}, 354 355 /* OP_OPEN_PUTFH = 22 */ 356 {rfs4_op_putfh, nullfree, RPC_ALL}, 357 358 /* OP_PUTPUBFH = 23 */ 359 {rfs4_op_putpubfh, nullfree, RPC_ALL}, 360 361 /* OP_PUTROOTFH = 24 */ 362 {rfs4_op_putrootfh, nullfree, RPC_ALL}, 363 364 /* OP_READ = 25 */ 365 {rfs4_op_read, rfs4_op_read_free, RPC_IDEMPOTENT}, 366 367 /* OP_READDIR = 26 */ 368 {rfs4_op_readdir, rfs4_op_readdir_free, RPC_IDEMPOTENT}, 369 370 /* OP_READLINK = 27 */ 371 {rfs4_op_readlink, rfs4_op_readlink_free, RPC_IDEMPOTENT}, 372 373 /* OP_REMOVE = 28 */ 374 {rfs4_op_remove, nullfree, 0}, 375 376 /* OP_RENAME = 29 */ 377 {rfs4_op_rename, nullfree, 0}, 378 379 /* OP_RENEW = 30 */ 380 {rfs4_op_renew, nullfree, 0}, 381 382 /* OP_RESTOREFH = 31 */ 383 {rfs4_op_restorefh, nullfree, RPC_ALL}, 384 385 /* OP_SAVEFH = 32 */ 386 {rfs4_op_savefh, nullfree, RPC_ALL}, 387 388 /* OP_SECINFO = 33 */ 389 {rfs4_op_secinfo, rfs4_op_secinfo_free, 0}, 390 391 /* OP_SETATTR = 34 */ 392 {rfs4_op_setattr, nullfree, 0}, 393 394 /* OP_SETCLIENTID = 35 */ 395 {rfs4_op_setclientid, nullfree, 0}, 396 397 /* OP_SETCLIENTID_CONFIRM = 36 */ 398 {rfs4_op_setclientid_confirm, nullfree, 0}, 399 400 /* OP_VERIFY = 37 */ 401 {rfs4_op_verify, nullfree, RPC_IDEMPOTENT}, 402 403 /* OP_WRITE = 38 */ 404 {rfs4_op_write, nullfree, 0}, 405 406 /* OP_RELEASE_LOCKOWNER = 39 */ 407 {rfs4_op_release_lockowner, nullfree, 0}, 408 }; 409 410 static uint_t rfsv4disp_cnt = sizeof (rfsv4disptab) / sizeof (rfsv4disptab[0]); 411 412 #define OP_ILLEGAL_IDX (rfsv4disp_cnt) 413 414 #ifdef DEBUG 415 416 int rfs4_fillone_debug = 0; 417 int rfs4_shrlock_debug = 0; 418 int rfs4_no_stub_access = 1; 419 int rfs4_rddir_debug = 0; 420 421 static char *rfs4_op_string[] = { 422 "rfs4_op_null", 423 "rfs4_op_1 unused", 424 "rfs4_op_2 unused", 425 "rfs4_op_access", 426 "rfs4_op_close", 427 "rfs4_op_commit", 428 "rfs4_op_create", 429 "rfs4_op_delegpurge", 430 "rfs4_op_delegreturn", 431 "rfs4_op_getattr", 432 "rfs4_op_getfh", 433 "rfs4_op_link", 434 "rfs4_op_lock", 435 "rfs4_op_lockt", 436 "rfs4_op_locku", 437 "rfs4_op_lookup", 438 "rfs4_op_lookupp", 439 "rfs4_op_nverify", 440 "rfs4_op_open", 441 "rfs4_op_openattr", 442 "rfs4_op_open_confirm", 443 "rfs4_op_open_downgrade", 444 "rfs4_op_putfh", 445 "rfs4_op_putpubfh", 446 "rfs4_op_putrootfh", 447 "rfs4_op_read", 448 "rfs4_op_readdir", 449 "rfs4_op_readlink", 450 "rfs4_op_remove", 451 "rfs4_op_rename", 452 "rfs4_op_renew", 453 "rfs4_op_restorefh", 454 "rfs4_op_savefh", 455 "rfs4_op_secinfo", 456 "rfs4_op_setattr", 457 "rfs4_op_setclientid", 458 "rfs4_op_setclient_confirm", 459 "rfs4_op_verify", 460 "rfs4_op_write", 461 "rfs4_op_release_lockowner", 462 "rfs4_op_illegal" 463 }; 464 #endif 465 466 void rfs4_ss_chkclid(rfs4_client_t *); 467 468 extern size_t strlcpy(char *dst, const char *src, size_t dstsize); 469 470 #ifdef nextdp 471 #undef nextdp 472 #endif 473 #define nextdp(dp) ((struct dirent64 *)((char *)(dp) + (dp)->d_reclen)) 474 475 static const fs_operation_def_t nfs4_rd_deleg_tmpl[] = { 476 VOPNAME_OPEN, { .femop_open = deleg_rd_open }, 477 VOPNAME_WRITE, { .femop_write = deleg_rd_write }, 478 VOPNAME_SETATTR, { .femop_setattr = deleg_rd_setattr }, 479 VOPNAME_RWLOCK, { .femop_rwlock = deleg_rd_rwlock }, 480 VOPNAME_SPACE, { .femop_space = deleg_rd_space }, 481 VOPNAME_SETSECATTR, { .femop_setsecattr = deleg_rd_setsecattr }, 482 VOPNAME_VNEVENT, { .femop_vnevent = deleg_rd_vnevent }, 483 NULL, NULL 484 }; 485 static const fs_operation_def_t nfs4_wr_deleg_tmpl[] = { 486 VOPNAME_OPEN, { .femop_open = deleg_wr_open }, 487 VOPNAME_READ, { .femop_read = deleg_wr_read }, 488 VOPNAME_WRITE, { .femop_write = deleg_wr_write }, 489 VOPNAME_SETATTR, { .femop_setattr = deleg_wr_setattr }, 490 VOPNAME_RWLOCK, { .femop_rwlock = deleg_wr_rwlock }, 491 VOPNAME_SPACE, { .femop_space = deleg_wr_space }, 492 VOPNAME_SETSECATTR, { .femop_setsecattr = deleg_wr_setsecattr }, 493 VOPNAME_VNEVENT, { .femop_vnevent = deleg_wr_vnevent }, 494 NULL, NULL 495 }; 496 497 int 498 rfs4_srvrinit(void) 499 { 500 timespec32_t verf; 501 int error; 502 extern void rfs4_attr_init(); 503 extern krwlock_t rfs4_deleg_policy_lock; 504 505 /* 506 * The following algorithm attempts to find a unique verifier 507 * to be used as the write verifier returned from the server 508 * to the client. It is important that this verifier change 509 * whenever the server reboots. Of secondary importance, it 510 * is important for the verifier to be unique between two 511 * different servers. 512 * 513 * Thus, an attempt is made to use the system hostid and the 514 * current time in seconds when the nfssrv kernel module is 515 * loaded. It is assumed that an NFS server will not be able 516 * to boot and then to reboot in less than a second. If the 517 * hostid has not been set, then the current high resolution 518 * time is used. This will ensure different verifiers each 519 * time the server reboots and minimize the chances that two 520 * different servers will have the same verifier. 521 * XXX - this is broken on LP64 kernels. 522 */ 523 verf.tv_sec = (time_t)nfs_atoi(hw_serial); 524 if (verf.tv_sec != 0) { 525 verf.tv_nsec = gethrestime_sec(); 526 } else { 527 timespec_t tverf; 528 529 gethrestime(&tverf); 530 verf.tv_sec = (time_t)tverf.tv_sec; 531 verf.tv_nsec = tverf.tv_nsec; 532 } 533 534 Write4verf = *(uint64_t *)&verf; 535 536 rfs4_attr_init(); 537 mutex_init(&rfs4_deleg_lock, NULL, MUTEX_DEFAULT, NULL); 538 539 /* Used to manage create/destroy of server state */ 540 mutex_init(&rfs4_state_lock, NULL, MUTEX_DEFAULT, NULL); 541 542 /* Used to manage access to server instance linked list */ 543 mutex_init(&rfs4_servinst_lock, NULL, MUTEX_DEFAULT, NULL); 544 545 /* Used to manage access to rfs4_deleg_policy */ 546 rw_init(&rfs4_deleg_policy_lock, NULL, RW_DEFAULT, NULL); 547 548 error = fem_create("deleg_rdops", nfs4_rd_deleg_tmpl, &deleg_rdops); 549 if (error != 0) { 550 rfs4_disable_delegation(); 551 } else { 552 error = fem_create("deleg_wrops", nfs4_wr_deleg_tmpl, 553 &deleg_wrops); 554 if (error != 0) { 555 rfs4_disable_delegation(); 556 fem_free(deleg_rdops); 557 } 558 } 559 560 nfs4_srv_caller_id = fs_new_caller_id(); 561 562 lockt_sysid = lm_alloc_sysidt(); 563 564 vsd_create(&nfs4_srv_vkey, NULL); 565 566 return (0); 567 } 568 569 void 570 rfs4_srvrfini(void) 571 { 572 extern krwlock_t rfs4_deleg_policy_lock; 573 574 if (lockt_sysid != LM_NOSYSID) { 575 lm_free_sysidt(lockt_sysid); 576 lockt_sysid = LM_NOSYSID; 577 } 578 579 mutex_destroy(&rfs4_deleg_lock); 580 mutex_destroy(&rfs4_state_lock); 581 rw_destroy(&rfs4_deleg_policy_lock); 582 583 fem_free(deleg_rdops); 584 fem_free(deleg_wrops); 585 } 586 587 void 588 rfs4_init_compound_state(struct compound_state *cs) 589 { 590 bzero(cs, sizeof (*cs)); 591 cs->cont = TRUE; 592 cs->access = CS_ACCESS_DENIED; 593 cs->deleg = FALSE; 594 cs->mandlock = FALSE; 595 cs->fh.nfs_fh4_val = cs->fhbuf; 596 } 597 598 void 599 rfs4_grace_start(rfs4_servinst_t *sip) 600 { 601 time_t now = gethrestime_sec(); 602 603 rw_enter(&sip->rwlock, RW_WRITER); 604 sip->start_time = now; 605 sip->grace_period = rfs4_grace_period; 606 rw_exit(&sip->rwlock); 607 } 608 609 /* 610 * returns true if the instance's grace period has never been started 611 */ 612 int 613 rfs4_servinst_grace_new(rfs4_servinst_t *sip) 614 { 615 time_t start_time; 616 617 rw_enter(&sip->rwlock, RW_READER); 618 start_time = sip->start_time; 619 rw_exit(&sip->rwlock); 620 621 return (start_time == 0); 622 } 623 624 /* 625 * Indicates if server instance is within the 626 * grace period. 627 */ 628 int 629 rfs4_servinst_in_grace(rfs4_servinst_t *sip) 630 { 631 time_t grace_expiry; 632 633 rw_enter(&sip->rwlock, RW_READER); 634 grace_expiry = sip->start_time + sip->grace_period; 635 rw_exit(&sip->rwlock); 636 637 return (gethrestime_sec() < grace_expiry); 638 } 639 640 int 641 rfs4_clnt_in_grace(rfs4_client_t *cp) 642 { 643 ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0); 644 645 return (rfs4_servinst_in_grace(cp->server_instance)); 646 } 647 648 /* 649 * reset all currently active grace periods 650 */ 651 void 652 rfs4_grace_reset_all(void) 653 { 654 rfs4_servinst_t *sip; 655 656 mutex_enter(&rfs4_servinst_lock); 657 for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev) 658 if (rfs4_servinst_in_grace(sip)) 659 rfs4_grace_start(sip); 660 mutex_exit(&rfs4_servinst_lock); 661 } 662 663 /* 664 * start any new instances' grace periods 665 */ 666 void 667 rfs4_grace_start_new(void) 668 { 669 rfs4_servinst_t *sip; 670 671 mutex_enter(&rfs4_servinst_lock); 672 for (sip = rfs4_cur_servinst; sip != NULL; sip = sip->prev) 673 if (rfs4_servinst_grace_new(sip)) 674 rfs4_grace_start(sip); 675 mutex_exit(&rfs4_servinst_lock); 676 } 677 678 static rfs4_dss_path_t * 679 rfs4_dss_newpath(rfs4_servinst_t *sip, char *path, unsigned index) 680 { 681 size_t len; 682 rfs4_dss_path_t *dss_path; 683 684 dss_path = kmem_alloc(sizeof (rfs4_dss_path_t), KM_SLEEP); 685 686 /* 687 * Take a copy of the string, since the original may be overwritten. 688 * Sadly, no strdup() in the kernel. 689 */ 690 /* allow for NUL */ 691 len = strlen(path) + 1; 692 dss_path->path = kmem_alloc(len, KM_SLEEP); 693 (void) strlcpy(dss_path->path, path, len); 694 695 /* associate with servinst */ 696 dss_path->sip = sip; 697 dss_path->index = index; 698 699 /* 700 * Add to list of served paths. 701 * No locking required, as we're only ever called at startup. 702 */ 703 if (rfs4_dss_pathlist == NULL) { 704 /* this is the first dss_path_t */ 705 706 /* needed for insque/remque */ 707 dss_path->next = dss_path->prev = dss_path; 708 709 rfs4_dss_pathlist = dss_path; 710 } else { 711 insque(dss_path, rfs4_dss_pathlist); 712 } 713 714 return (dss_path); 715 } 716 717 /* 718 * Create a new server instance, and make it the currently active instance. 719 * Note that starting the grace period too early will reduce the clients' 720 * recovery window. 721 */ 722 void 723 rfs4_servinst_create(int start_grace, int dss_npaths, char **dss_paths) 724 { 725 unsigned i; 726 rfs4_servinst_t *sip; 727 rfs4_oldstate_t *oldstate; 728 729 sip = kmem_alloc(sizeof (rfs4_servinst_t), KM_SLEEP); 730 rw_init(&sip->rwlock, NULL, RW_DEFAULT, NULL); 731 732 sip->start_time = (time_t)0; 733 sip->grace_period = (time_t)0; 734 sip->next = NULL; 735 sip->prev = NULL; 736 737 rw_init(&sip->oldstate_lock, NULL, RW_DEFAULT, NULL); 738 /* 739 * This initial dummy entry is required to setup for insque/remque. 740 * It must be skipped over whenever the list is traversed. 741 */ 742 oldstate = kmem_alloc(sizeof (rfs4_oldstate_t), KM_SLEEP); 743 /* insque/remque require initial list entry to be self-terminated */ 744 oldstate->next = oldstate; 745 oldstate->prev = oldstate; 746 sip->oldstate = oldstate; 747 748 749 sip->dss_npaths = dss_npaths; 750 sip->dss_paths = kmem_alloc(dss_npaths * 751 sizeof (rfs4_dss_path_t *), KM_SLEEP); 752 753 for (i = 0; i < dss_npaths; i++) { 754 sip->dss_paths[i] = rfs4_dss_newpath(sip, dss_paths[i], i); 755 } 756 757 mutex_enter(&rfs4_servinst_lock); 758 if (rfs4_cur_servinst != NULL) { 759 /* add to linked list */ 760 sip->prev = rfs4_cur_servinst; 761 rfs4_cur_servinst->next = sip; 762 } 763 if (start_grace) 764 rfs4_grace_start(sip); 765 /* make the new instance "current" */ 766 rfs4_cur_servinst = sip; 767 768 mutex_exit(&rfs4_servinst_lock); 769 } 770 771 /* 772 * In future, we might add a rfs4_servinst_destroy(sip) but, for now, destroy 773 * all instances directly. 774 */ 775 void 776 rfs4_servinst_destroy_all(void) 777 { 778 rfs4_servinst_t *sip, *prev, *current; 779 #ifdef DEBUG 780 int n = 0; 781 #endif 782 783 mutex_enter(&rfs4_servinst_lock); 784 ASSERT(rfs4_cur_servinst != NULL); 785 current = rfs4_cur_servinst; 786 rfs4_cur_servinst = NULL; 787 for (sip = current; sip != NULL; sip = prev) { 788 prev = sip->prev; 789 rw_destroy(&sip->rwlock); 790 if (sip->oldstate) 791 kmem_free(sip->oldstate, sizeof (rfs4_oldstate_t)); 792 if (sip->dss_paths) 793 kmem_free(sip->dss_paths, 794 sip->dss_npaths * sizeof (rfs4_dss_path_t *)); 795 kmem_free(sip, sizeof (rfs4_servinst_t)); 796 #ifdef DEBUG 797 n++; 798 #endif 799 } 800 mutex_exit(&rfs4_servinst_lock); 801 } 802 803 /* 804 * Assign the current server instance to a client_t. 805 * Should be called with cp->dbe held. 806 */ 807 void 808 rfs4_servinst_assign(rfs4_client_t *cp, rfs4_servinst_t *sip) 809 { 810 ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0); 811 812 /* 813 * The lock ensures that if the current instance is in the process 814 * of changing, we will see the new one. 815 */ 816 mutex_enter(&rfs4_servinst_lock); 817 cp->server_instance = sip; 818 mutex_exit(&rfs4_servinst_lock); 819 } 820 821 rfs4_servinst_t * 822 rfs4_servinst(rfs4_client_t *cp) 823 { 824 ASSERT(rfs4_dbe_refcnt(cp->dbe) > 0); 825 826 return (cp->server_instance); 827 } 828 829 /* ARGSUSED */ 830 static void 831 nullfree(caddr_t resop) 832 { 833 } 834 835 /* 836 * This is a fall-through for invalid or not implemented (yet) ops 837 */ 838 /* ARGSUSED */ 839 static void 840 rfs4_op_inval(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 841 struct compound_state *cs) 842 { 843 *cs->statusp = *((nfsstat4 *)&(resop)->nfs_resop4_u) = NFS4ERR_INVAL; 844 } 845 846 /* 847 * Check if the security flavor, nfsnum, is in the flavor_list. 848 */ 849 bool_t 850 in_flavor_list(int nfsnum, int *flavor_list, int count) 851 { 852 int i; 853 854 for (i = 0; i < count; i++) { 855 if (nfsnum == flavor_list[i]) 856 return (TRUE); 857 } 858 return (FALSE); 859 } 860 861 /* 862 * Used by rfs4_op_secinfo to get the security information from the 863 * export structure associated with the component. 864 */ 865 /* ARGSUSED */ 866 static nfsstat4 867 do_rfs4_op_secinfo(struct compound_state *cs, char *nm, SECINFO4res *resp) 868 { 869 int error, different_export = 0; 870 vnode_t *dvp, *vp, *tvp; 871 struct exportinfo *exi = NULL; 872 fid_t fid; 873 uint_t count, i; 874 secinfo4 *resok_val; 875 struct secinfo *secp; 876 seconfig_t *si; 877 bool_t did_traverse; 878 int dotdot, walk; 879 880 dvp = cs->vp; 881 dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0'); 882 883 /* 884 * If dotdotting, then need to check whether it's above the 885 * root of a filesystem, or above an export point. 886 */ 887 if (dotdot) { 888 889 /* 890 * If dotdotting at the root of a filesystem, then 891 * need to traverse back to the mounted-on filesystem 892 * and do the dotdot lookup there. 893 */ 894 if (cs->vp->v_flag & VROOT) { 895 896 /* 897 * If at the system root, then can 898 * go up no further. 899 */ 900 if (VN_CMP(dvp, rootdir)) 901 return (puterrno4(ENOENT)); 902 903 /* 904 * Traverse back to the mounted-on filesystem 905 */ 906 dvp = untraverse(cs->vp); 907 908 /* 909 * Set the different_export flag so we remember 910 * to pick up a new exportinfo entry for 911 * this new filesystem. 912 */ 913 different_export = 1; 914 } else { 915 916 /* 917 * If dotdotting above an export point then set 918 * the different_export to get new export info. 919 */ 920 different_export = nfs_exported(cs->exi, cs->vp); 921 } 922 } 923 924 /* 925 * Get the vnode for the component "nm". 926 */ 927 error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cs->cr, 928 NULL, NULL, NULL); 929 if (error) 930 return (puterrno4(error)); 931 932 /* 933 * If the vnode is in a pseudo filesystem, or if the security flavor 934 * used in the request is valid but not an explicitly shared flavor, 935 * or the access bit indicates that this is a limited access, 936 * check whether this vnode is visible. 937 */ 938 if (!different_export && 939 (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) || 940 cs->access & CS_ACCESS_LIMITED)) { 941 if (! nfs_visible(cs->exi, vp, &different_export)) { 942 VN_RELE(vp); 943 return (puterrno4(ENOENT)); 944 } 945 } 946 947 /* 948 * If it's a mountpoint, then traverse it. 949 */ 950 if (vn_ismntpt(vp)) { 951 tvp = vp; 952 if ((error = traverse(&tvp)) != 0) { 953 VN_RELE(vp); 954 return (puterrno4(error)); 955 } 956 /* remember that we had to traverse mountpoint */ 957 did_traverse = TRUE; 958 vp = tvp; 959 different_export = 1; 960 } else if (vp->v_vfsp != dvp->v_vfsp) { 961 /* 962 * If vp isn't a mountpoint and the vfs ptrs aren't the same, 963 * then vp is probably an LOFS object. We don't need the 964 * realvp, we just need to know that we might have crossed 965 * a server fs boundary and need to call checkexport4. 966 * (LOFS lookup hides server fs mountpoints, and actually calls 967 * traverse) 968 */ 969 different_export = 1; 970 did_traverse = FALSE; 971 } 972 973 /* 974 * Get the export information for it. 975 */ 976 if (different_export) { 977 978 bzero(&fid, sizeof (fid)); 979 fid.fid_len = MAXFIDSZ; 980 error = vop_fid_pseudo(vp, &fid); 981 if (error) { 982 VN_RELE(vp); 983 return (puterrno4(error)); 984 } 985 986 if (dotdot) 987 exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE); 988 else 989 exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp); 990 991 if (exi == NULL) { 992 if (did_traverse == TRUE) { 993 /* 994 * If this vnode is a mounted-on vnode, 995 * but the mounted-on file system is not 996 * exported, send back the secinfo for 997 * the exported node that the mounted-on 998 * vnode lives in. 999 */ 1000 exi = cs->exi; 1001 } else { 1002 VN_RELE(vp); 1003 return (puterrno4(EACCES)); 1004 } 1005 } 1006 } else { 1007 exi = cs->exi; 1008 } 1009 ASSERT(exi != NULL); 1010 1011 1012 /* 1013 * Create the secinfo result based on the security information 1014 * from the exportinfo structure (exi). 1015 * 1016 * Return all flavors for a pseudo node. 1017 * For a real export node, return the flavor that the client 1018 * has access with. 1019 */ 1020 ASSERT(RW_LOCK_HELD(&exported_lock)); 1021 if (PSEUDO(exi)) { 1022 count = exi->exi_export.ex_seccnt; /* total sec count */ 1023 resok_val = kmem_alloc(count * sizeof (secinfo4), KM_SLEEP); 1024 secp = exi->exi_export.ex_secinfo; 1025 1026 for (i = 0; i < count; i++) { 1027 si = &secp[i].s_secinfo; 1028 resok_val[i].flavor = si->sc_rpcnum; 1029 if (resok_val[i].flavor == RPCSEC_GSS) { 1030 rpcsec_gss_info *info; 1031 1032 info = &resok_val[i].flavor_info; 1033 info->qop = si->sc_qop; 1034 info->service = (rpc_gss_svc_t)si->sc_service; 1035 1036 /* get oid opaque data */ 1037 info->oid.sec_oid4_len = 1038 si->sc_gss_mech_type->length; 1039 info->oid.sec_oid4_val = kmem_alloc( 1040 si->sc_gss_mech_type->length, KM_SLEEP); 1041 bcopy( 1042 si->sc_gss_mech_type->elements, 1043 info->oid.sec_oid4_val, 1044 info->oid.sec_oid4_len); 1045 } 1046 } 1047 resp->SECINFO4resok_len = count; 1048 resp->SECINFO4resok_val = resok_val; 1049 } else { 1050 int ret_cnt = 0, k = 0; 1051 int *flavor_list; 1052 1053 count = exi->exi_export.ex_seccnt; /* total sec count */ 1054 secp = exi->exi_export.ex_secinfo; 1055 1056 flavor_list = kmem_alloc(count * sizeof (int), KM_SLEEP); 1057 /* find out which flavors to return */ 1058 for (i = 0; i < count; i ++) { 1059 int access, flavor, perm; 1060 1061 flavor = secp[i].s_secinfo.sc_nfsnum; 1062 perm = secp[i].s_flags; 1063 1064 access = nfsauth4_secinfo_access(exi, cs->req, 1065 flavor, perm); 1066 1067 if (! (access & NFSAUTH_DENIED) && 1068 ! (access & NFSAUTH_WRONGSEC)) { 1069 flavor_list[ret_cnt] = flavor; 1070 ret_cnt++; 1071 } 1072 } 1073 1074 /* Create the returning SECINFO value */ 1075 resok_val = kmem_alloc(ret_cnt * sizeof (secinfo4), KM_SLEEP); 1076 1077 for (i = 0; i < count; i++) { 1078 /* 1079 * If the flavor is in the flavor list, 1080 * fill in resok_val. 1081 */ 1082 si = &secp[i].s_secinfo; 1083 if (in_flavor_list(si->sc_nfsnum, 1084 flavor_list, ret_cnt)) { 1085 resok_val[k].flavor = si->sc_rpcnum; 1086 if (resok_val[k].flavor == RPCSEC_GSS) { 1087 rpcsec_gss_info *info; 1088 1089 info = &resok_val[k].flavor_info; 1090 info->qop = si->sc_qop; 1091 info->service = (rpc_gss_svc_t) 1092 si->sc_service; 1093 1094 /* get oid opaque data */ 1095 info->oid.sec_oid4_len = 1096 si->sc_gss_mech_type->length; 1097 info->oid.sec_oid4_val = kmem_alloc( 1098 si->sc_gss_mech_type->length, 1099 KM_SLEEP); 1100 bcopy(si->sc_gss_mech_type->elements, 1101 info->oid.sec_oid4_val, 1102 info->oid.sec_oid4_len); 1103 } 1104 k++; 1105 } 1106 if (k >= ret_cnt) 1107 break; 1108 } 1109 resp->SECINFO4resok_len = ret_cnt; 1110 resp->SECINFO4resok_val = resok_val; 1111 kmem_free(flavor_list, count * sizeof (int)); 1112 } 1113 1114 VN_RELE(vp); 1115 return (NFS4_OK); 1116 } 1117 1118 /* 1119 * SECINFO (Operation 33): Obtain required security information on 1120 * the component name in the format of (security-mechanism-oid, qop, service) 1121 * triplets. 1122 */ 1123 /* ARGSUSED */ 1124 static void 1125 rfs4_op_secinfo(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1126 struct compound_state *cs) 1127 { 1128 SECINFO4args *args = &argop->nfs_argop4_u.opsecinfo; 1129 SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo; 1130 utf8string *utfnm = &args->name; 1131 uint_t len; 1132 char *nm; 1133 1134 DTRACE_NFSV4_2(op__secinfo__start, struct compound_state *, cs, 1135 SECINFO4args *, args); 1136 1137 /* 1138 * Current file handle (cfh) should have been set before getting 1139 * into this function. If not, return error. 1140 */ 1141 if (cs->vp == NULL) { 1142 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1143 goto out; 1144 } 1145 1146 if (cs->vp->v_type != VDIR) { 1147 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 1148 goto out; 1149 } 1150 1151 /* 1152 * Verify the component name. If failed, error out, but 1153 * do not error out if the component name is a "..". 1154 * SECINFO will return its parents secinfo data for SECINFO "..". 1155 */ 1156 if (!utf8_dir_verify(utfnm)) { 1157 if (utfnm->utf8string_len != 2 || 1158 utfnm->utf8string_val[0] != '.' || 1159 utfnm->utf8string_val[1] != '.') { 1160 *cs->statusp = resp->status = NFS4ERR_INVAL; 1161 goto out; 1162 } 1163 } 1164 1165 nm = utf8_to_str(utfnm, &len, NULL); 1166 if (nm == NULL) { 1167 *cs->statusp = resp->status = NFS4ERR_INVAL; 1168 goto out; 1169 } 1170 1171 if (len > MAXNAMELEN) { 1172 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1173 kmem_free(nm, len); 1174 goto out; 1175 } 1176 1177 *cs->statusp = resp->status = do_rfs4_op_secinfo(cs, nm, resp); 1178 1179 kmem_free(nm, len); 1180 1181 out: 1182 DTRACE_NFSV4_2(op__secinfo__done, struct compound_state *, cs, 1183 SECINFO4res *, resp); 1184 } 1185 1186 /* 1187 * Free SECINFO result. 1188 */ 1189 /* ARGSUSED */ 1190 static void 1191 rfs4_op_secinfo_free(nfs_resop4 *resop) 1192 { 1193 SECINFO4res *resp = &resop->nfs_resop4_u.opsecinfo; 1194 int count, i; 1195 secinfo4 *resok_val; 1196 1197 /* If this is not an Ok result, nothing to free. */ 1198 if (resp->status != NFS4_OK) { 1199 return; 1200 } 1201 1202 count = resp->SECINFO4resok_len; 1203 resok_val = resp->SECINFO4resok_val; 1204 1205 for (i = 0; i < count; i++) { 1206 if (resok_val[i].flavor == RPCSEC_GSS) { 1207 rpcsec_gss_info *info; 1208 1209 info = &resok_val[i].flavor_info; 1210 kmem_free(info->oid.sec_oid4_val, 1211 info->oid.sec_oid4_len); 1212 } 1213 } 1214 kmem_free(resok_val, count * sizeof (secinfo4)); 1215 resp->SECINFO4resok_len = 0; 1216 resp->SECINFO4resok_val = NULL; 1217 } 1218 1219 /* ARGSUSED */ 1220 static void 1221 rfs4_op_access(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1222 struct compound_state *cs) 1223 { 1224 ACCESS4args *args = &argop->nfs_argop4_u.opaccess; 1225 ACCESS4res *resp = &resop->nfs_resop4_u.opaccess; 1226 int error; 1227 vnode_t *vp; 1228 struct vattr va; 1229 int checkwriteperm; 1230 cred_t *cr = cs->cr; 1231 bslabel_t *clabel, *slabel; 1232 ts_label_t *tslabel; 1233 boolean_t admin_low_client; 1234 1235 DTRACE_NFSV4_2(op__access__start, struct compound_state *, cs, 1236 ACCESS4args *, args); 1237 1238 #if 0 /* XXX allow access even if !cs->access. Eventually only pseudo fs */ 1239 if (cs->access == CS_ACCESS_DENIED) { 1240 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1241 goto out; 1242 } 1243 #endif 1244 if (cs->vp == NULL) { 1245 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1246 goto out; 1247 } 1248 1249 ASSERT(cr != NULL); 1250 1251 vp = cs->vp; 1252 1253 /* 1254 * If the file system is exported read only, it is not appropriate 1255 * to check write permissions for regular files and directories. 1256 * Special files are interpreted by the client, so the underlying 1257 * permissions are sent back to the client for interpretation. 1258 */ 1259 if (rdonly4(cs->exi, cs->vp, req) && 1260 (vp->v_type == VREG || vp->v_type == VDIR)) 1261 checkwriteperm = 0; 1262 else 1263 checkwriteperm = 1; 1264 1265 /* 1266 * XXX 1267 * We need the mode so that we can correctly determine access 1268 * permissions relative to a mandatory lock file. Access to 1269 * mandatory lock files is denied on the server, so it might 1270 * as well be reflected to the server during the open. 1271 */ 1272 va.va_mask = AT_MODE; 1273 error = VOP_GETATTR(vp, &va, 0, cr, NULL); 1274 if (error) { 1275 *cs->statusp = resp->status = puterrno4(error); 1276 goto out; 1277 } 1278 resp->access = 0; 1279 resp->supported = 0; 1280 1281 if (is_system_labeled()) { 1282 ASSERT(req->rq_label != NULL); 1283 clabel = req->rq_label; 1284 DTRACE_PROBE2(tx__rfs4__log__info__opaccess__clabel, char *, 1285 "got client label from request(1)", 1286 struct svc_req *, req); 1287 if (!blequal(&l_admin_low->tsl_label, clabel)) { 1288 if ((tslabel = nfs_getflabel(vp)) == NULL) { 1289 *cs->statusp = resp->status = puterrno4(EACCES); 1290 goto out; 1291 } 1292 slabel = label2bslabel(tslabel); 1293 DTRACE_PROBE3(tx__rfs4__log__info__opaccess__slabel, 1294 char *, "got server label(1) for vp(2)", 1295 bslabel_t *, slabel, vnode_t *, vp); 1296 1297 admin_low_client = B_FALSE; 1298 } else 1299 admin_low_client = B_TRUE; 1300 } 1301 1302 if (args->access & ACCESS4_READ) { 1303 error = VOP_ACCESS(vp, VREAD, 0, cr, NULL); 1304 if (!error && !MANDLOCK(vp, va.va_mode) && 1305 (!is_system_labeled() || admin_low_client || 1306 bldominates(clabel, slabel))) 1307 resp->access |= ACCESS4_READ; 1308 resp->supported |= ACCESS4_READ; 1309 } 1310 if ((args->access & ACCESS4_LOOKUP) && vp->v_type == VDIR) { 1311 error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL); 1312 if (!error && (!is_system_labeled() || admin_low_client || 1313 bldominates(clabel, slabel))) 1314 resp->access |= ACCESS4_LOOKUP; 1315 resp->supported |= ACCESS4_LOOKUP; 1316 } 1317 if (checkwriteperm && 1318 (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND))) { 1319 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 1320 if (!error && !MANDLOCK(vp, va.va_mode) && 1321 (!is_system_labeled() || admin_low_client || 1322 blequal(clabel, slabel))) 1323 resp->access |= 1324 (args->access & (ACCESS4_MODIFY|ACCESS4_EXTEND)); 1325 resp->supported |= (ACCESS4_MODIFY|ACCESS4_EXTEND); 1326 } 1327 1328 if (checkwriteperm && 1329 (args->access & ACCESS4_DELETE) && vp->v_type == VDIR) { 1330 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 1331 if (!error && (!is_system_labeled() || admin_low_client || 1332 blequal(clabel, slabel))) 1333 resp->access |= ACCESS4_DELETE; 1334 resp->supported |= ACCESS4_DELETE; 1335 } 1336 if (args->access & ACCESS4_EXECUTE && vp->v_type != VDIR) { 1337 error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL); 1338 if (!error && !MANDLOCK(vp, va.va_mode) && 1339 (!is_system_labeled() || admin_low_client || 1340 bldominates(clabel, slabel))) 1341 resp->access |= ACCESS4_EXECUTE; 1342 resp->supported |= ACCESS4_EXECUTE; 1343 } 1344 1345 if (is_system_labeled() && !admin_low_client) 1346 label_rele(tslabel); 1347 1348 *cs->statusp = resp->status = NFS4_OK; 1349 out: 1350 DTRACE_NFSV4_2(op__access__done, struct compound_state *, cs, 1351 ACCESS4res *, resp); 1352 } 1353 1354 /* ARGSUSED */ 1355 static void 1356 rfs4_op_commit(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1357 struct compound_state *cs) 1358 { 1359 COMMIT4args *args = &argop->nfs_argop4_u.opcommit; 1360 COMMIT4res *resp = &resop->nfs_resop4_u.opcommit; 1361 int error; 1362 vnode_t *vp = cs->vp; 1363 cred_t *cr = cs->cr; 1364 vattr_t va; 1365 1366 DTRACE_NFSV4_2(op__commit__start, struct compound_state *, cs, 1367 COMMIT4args *, args); 1368 1369 if (vp == NULL) { 1370 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1371 goto out; 1372 } 1373 if (cs->access == CS_ACCESS_DENIED) { 1374 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1375 goto out; 1376 } 1377 1378 if (args->offset + args->count < args->offset) { 1379 *cs->statusp = resp->status = NFS4ERR_INVAL; 1380 goto out; 1381 } 1382 1383 va.va_mask = AT_UID; 1384 error = VOP_GETATTR(vp, &va, 0, cr, NULL); 1385 1386 /* 1387 * If we can't get the attributes, then we can't do the 1388 * right access checking. So, we'll fail the request. 1389 */ 1390 if (error) { 1391 *cs->statusp = resp->status = puterrno4(error); 1392 goto out; 1393 } 1394 if (rdonly4(cs->exi, cs->vp, req)) { 1395 *cs->statusp = resp->status = NFS4ERR_ROFS; 1396 goto out; 1397 } 1398 1399 if (vp->v_type != VREG) { 1400 if (vp->v_type == VDIR) 1401 resp->status = NFS4ERR_ISDIR; 1402 else 1403 resp->status = NFS4ERR_INVAL; 1404 *cs->statusp = resp->status; 1405 goto out; 1406 } 1407 1408 if (crgetuid(cr) != va.va_uid && 1409 (error = VOP_ACCESS(vp, VWRITE, 0, cs->cr, NULL))) { 1410 *cs->statusp = resp->status = puterrno4(error); 1411 goto out; 1412 } 1413 1414 error = VOP_PUTPAGE(vp, args->offset, args->count, 0, cr, NULL); 1415 if (!error) 1416 error = VOP_FSYNC(vp, FNODSYNC, cr, NULL); 1417 1418 if (error) { 1419 *cs->statusp = resp->status = puterrno4(error); 1420 goto out; 1421 } 1422 1423 *cs->statusp = resp->status = NFS4_OK; 1424 resp->writeverf = Write4verf; 1425 out: 1426 DTRACE_NFSV4_2(op__commit__done, struct compound_state *, cs, 1427 COMMIT4res *, resp); 1428 } 1429 1430 /* 1431 * rfs4_op_mknod is called from rfs4_op_create after all initial verification 1432 * was completed. It does the nfsv4 create for special files. 1433 */ 1434 /* ARGSUSED */ 1435 static vnode_t * 1436 do_rfs4_op_mknod(CREATE4args *args, CREATE4res *resp, struct svc_req *req, 1437 struct compound_state *cs, vattr_t *vap, char *nm) 1438 { 1439 int error; 1440 cred_t *cr = cs->cr; 1441 vnode_t *dvp = cs->vp; 1442 vnode_t *vp = NULL; 1443 int mode; 1444 enum vcexcl excl; 1445 1446 switch (args->type) { 1447 case NF4CHR: 1448 case NF4BLK: 1449 if (secpolicy_sys_devices(cr) != 0) { 1450 *cs->statusp = resp->status = NFS4ERR_PERM; 1451 return (NULL); 1452 } 1453 if (args->type == NF4CHR) 1454 vap->va_type = VCHR; 1455 else 1456 vap->va_type = VBLK; 1457 vap->va_rdev = makedevice(args->ftype4_u.devdata.specdata1, 1458 args->ftype4_u.devdata.specdata2); 1459 vap->va_mask |= AT_RDEV; 1460 break; 1461 case NF4SOCK: 1462 vap->va_type = VSOCK; 1463 break; 1464 case NF4FIFO: 1465 vap->va_type = VFIFO; 1466 break; 1467 default: 1468 *cs->statusp = resp->status = NFS4ERR_BADTYPE; 1469 return (NULL); 1470 } 1471 1472 /* 1473 * Must specify the mode. 1474 */ 1475 if (!(vap->va_mask & AT_MODE)) { 1476 *cs->statusp = resp->status = NFS4ERR_INVAL; 1477 return (NULL); 1478 } 1479 1480 excl = EXCL; 1481 1482 mode = 0; 1483 1484 error = VOP_CREATE(dvp, nm, vap, excl, mode, &vp, cr, 0, NULL, NULL); 1485 if (error) { 1486 *cs->statusp = resp->status = puterrno4(error); 1487 return (NULL); 1488 } 1489 return (vp); 1490 } 1491 1492 /* 1493 * nfsv4 create is used to create non-regular files. For regular files, 1494 * use nfsv4 open. 1495 */ 1496 /* ARGSUSED */ 1497 static void 1498 rfs4_op_create(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1499 struct compound_state *cs) 1500 { 1501 CREATE4args *args = &argop->nfs_argop4_u.opcreate; 1502 CREATE4res *resp = &resop->nfs_resop4_u.opcreate; 1503 int error; 1504 struct vattr bva, iva, iva2, ava, *vap; 1505 cred_t *cr = cs->cr; 1506 vnode_t *dvp = cs->vp; 1507 vnode_t *vp = NULL; 1508 char *nm, *lnm; 1509 uint_t len, llen; 1510 int syncval = 0; 1511 struct nfs4_svgetit_arg sarg; 1512 struct nfs4_ntov_table ntov; 1513 struct statvfs64 sb; 1514 nfsstat4 status; 1515 1516 DTRACE_NFSV4_2(op__create__start, struct compound_state *, cs, 1517 CREATE4args *, args); 1518 1519 resp->attrset = 0; 1520 1521 if (dvp == NULL) { 1522 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 1523 goto out; 1524 } 1525 1526 /* 1527 * If there is an unshared filesystem mounted on this vnode, 1528 * do not allow to create an object in this directory. 1529 */ 1530 if (vn_ismntpt(dvp)) { 1531 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1532 goto out; 1533 } 1534 1535 /* Verify that type is correct */ 1536 switch (args->type) { 1537 case NF4LNK: 1538 case NF4BLK: 1539 case NF4CHR: 1540 case NF4SOCK: 1541 case NF4FIFO: 1542 case NF4DIR: 1543 break; 1544 default: 1545 *cs->statusp = resp->status = NFS4ERR_BADTYPE; 1546 goto out; 1547 }; 1548 1549 if (cs->access == CS_ACCESS_DENIED) { 1550 *cs->statusp = resp->status = NFS4ERR_ACCESS; 1551 goto out; 1552 } 1553 if (dvp->v_type != VDIR) { 1554 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 1555 goto out; 1556 } 1557 if (!utf8_dir_verify(&args->objname)) { 1558 *cs->statusp = resp->status = NFS4ERR_INVAL; 1559 goto out; 1560 } 1561 1562 if (rdonly4(cs->exi, cs->vp, req)) { 1563 *cs->statusp = resp->status = NFS4ERR_ROFS; 1564 goto out; 1565 } 1566 1567 /* 1568 * Name of newly created object 1569 */ 1570 nm = utf8_to_fn(&args->objname, &len, NULL); 1571 if (nm == NULL) { 1572 *cs->statusp = resp->status = NFS4ERR_INVAL; 1573 goto out; 1574 } 1575 1576 if (len > MAXNAMELEN) { 1577 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1578 kmem_free(nm, len); 1579 goto out; 1580 } 1581 1582 resp->attrset = 0; 1583 1584 sarg.sbp = &sb; 1585 nfs4_ntov_table_init(&ntov); 1586 1587 status = do_rfs4_set_attrs(&resp->attrset, 1588 &args->createattrs, cs, &sarg, &ntov, NFS4ATTR_SETIT); 1589 1590 if (sarg.vap->va_mask == 0 && status == NFS4_OK) 1591 status = NFS4ERR_INVAL; 1592 1593 if (status != NFS4_OK) { 1594 *cs->statusp = resp->status = status; 1595 kmem_free(nm, len); 1596 nfs4_ntov_table_free(&ntov, &sarg); 1597 resp->attrset = 0; 1598 goto out; 1599 } 1600 1601 /* Get "before" change value */ 1602 bva.va_mask = AT_CTIME|AT_SEQ; 1603 error = VOP_GETATTR(dvp, &bva, 0, cr, NULL); 1604 if (error) { 1605 *cs->statusp = resp->status = puterrno4(error); 1606 kmem_free(nm, len); 1607 nfs4_ntov_table_free(&ntov, &sarg); 1608 resp->attrset = 0; 1609 goto out; 1610 } 1611 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bva.va_ctime) 1612 1613 vap = sarg.vap; 1614 1615 /* 1616 * Set default initial values for attributes when not specified 1617 * in createattrs. 1618 */ 1619 if ((vap->va_mask & AT_UID) == 0) { 1620 vap->va_uid = crgetuid(cr); 1621 vap->va_mask |= AT_UID; 1622 } 1623 if ((vap->va_mask & AT_GID) == 0) { 1624 vap->va_gid = crgetgid(cr); 1625 vap->va_mask |= AT_GID; 1626 } 1627 1628 vap->va_mask |= AT_TYPE; 1629 switch (args->type) { 1630 case NF4DIR: 1631 vap->va_type = VDIR; 1632 if ((vap->va_mask & AT_MODE) == 0) { 1633 vap->va_mode = 0700; /* default: owner rwx only */ 1634 vap->va_mask |= AT_MODE; 1635 } 1636 error = VOP_MKDIR(dvp, nm, vap, &vp, cr, NULL, 0, NULL); 1637 if (error) 1638 break; 1639 1640 /* 1641 * Get the initial "after" sequence number, if it fails, 1642 * set to zero 1643 */ 1644 iva.va_mask = AT_SEQ; 1645 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1646 iva.va_seq = 0; 1647 break; 1648 case NF4LNK: 1649 vap->va_type = VLNK; 1650 if ((vap->va_mask & AT_MODE) == 0) { 1651 vap->va_mode = 0700; /* default: owner rwx only */ 1652 vap->va_mask |= AT_MODE; 1653 } 1654 1655 /* 1656 * symlink names must be treated as data 1657 */ 1658 lnm = utf8_to_str(&args->ftype4_u.linkdata, &llen, NULL); 1659 1660 if (lnm == NULL) { 1661 *cs->statusp = resp->status = NFS4ERR_INVAL; 1662 kmem_free(nm, len); 1663 nfs4_ntov_table_free(&ntov, &sarg); 1664 resp->attrset = 0; 1665 goto out; 1666 } 1667 1668 if (llen > MAXPATHLEN) { 1669 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 1670 kmem_free(nm, len); 1671 kmem_free(lnm, llen); 1672 nfs4_ntov_table_free(&ntov, &sarg); 1673 resp->attrset = 0; 1674 goto out; 1675 } 1676 1677 error = VOP_SYMLINK(dvp, nm, vap, lnm, cr, NULL, 0); 1678 if (lnm != NULL) 1679 kmem_free(lnm, llen); 1680 if (error) 1681 break; 1682 1683 /* 1684 * Get the initial "after" sequence number, if it fails, 1685 * set to zero 1686 */ 1687 iva.va_mask = AT_SEQ; 1688 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1689 iva.va_seq = 0; 1690 1691 error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, 1692 NULL, NULL, NULL); 1693 if (error) 1694 break; 1695 1696 /* 1697 * va_seq is not safe over VOP calls, check it again 1698 * if it has changed zero out iva to force atomic = FALSE. 1699 */ 1700 iva2.va_mask = AT_SEQ; 1701 if (VOP_GETATTR(dvp, &iva2, 0, cs->cr, NULL) || 1702 iva2.va_seq != iva.va_seq) 1703 iva.va_seq = 0; 1704 break; 1705 default: 1706 /* 1707 * probably a special file. 1708 */ 1709 if ((vap->va_mask & AT_MODE) == 0) { 1710 vap->va_mode = 0600; /* default: owner rw only */ 1711 vap->va_mask |= AT_MODE; 1712 } 1713 syncval = FNODSYNC; 1714 /* 1715 * We know this will only generate one VOP call 1716 */ 1717 vp = do_rfs4_op_mknod(args, resp, req, cs, vap, nm); 1718 1719 if (vp == NULL) { 1720 kmem_free(nm, len); 1721 nfs4_ntov_table_free(&ntov, &sarg); 1722 resp->attrset = 0; 1723 goto out; 1724 } 1725 1726 /* 1727 * Get the initial "after" sequence number, if it fails, 1728 * set to zero 1729 */ 1730 iva.va_mask = AT_SEQ; 1731 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) 1732 iva.va_seq = 0; 1733 1734 break; 1735 } 1736 kmem_free(nm, len); 1737 1738 if (error) { 1739 *cs->statusp = resp->status = puterrno4(error); 1740 } 1741 1742 /* 1743 * Force modified data and metadata out to stable storage. 1744 */ 1745 (void) VOP_FSYNC(dvp, 0, cr, NULL); 1746 1747 if (resp->status != NFS4_OK) { 1748 if (vp != NULL) 1749 VN_RELE(vp); 1750 nfs4_ntov_table_free(&ntov, &sarg); 1751 resp->attrset = 0; 1752 goto out; 1753 } 1754 1755 /* 1756 * Finish setup of cinfo response, "before" value already set. 1757 * Get "after" change value, if it fails, simply return the 1758 * before value. 1759 */ 1760 ava.va_mask = AT_CTIME|AT_SEQ; 1761 if (VOP_GETATTR(dvp, &ava, 0, cr, NULL)) { 1762 ava.va_ctime = bva.va_ctime; 1763 ava.va_seq = 0; 1764 } 1765 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, ava.va_ctime); 1766 1767 /* 1768 * True verification that object was created with correct 1769 * attrs is impossible. The attrs could have been changed 1770 * immediately after object creation. If attributes did 1771 * not verify, the only recourse for the server is to 1772 * destroy the object. Maybe if some attrs (like gid) 1773 * are set incorrectly, the object should be destroyed; 1774 * however, seems bad as a default policy. Do we really 1775 * want to destroy an object over one of the times not 1776 * verifying correctly? For these reasons, the server 1777 * currently sets bits in attrset for createattrs 1778 * that were set; however, no verification is done. 1779 * 1780 * vmask_to_nmask accounts for vattr bits set on create 1781 * [do_rfs4_set_attrs() only sets resp bits for 1782 * non-vattr/vfs bits.] 1783 * Mask off any bits set by default so as not to return 1784 * more attrset bits than were requested in createattrs 1785 */ 1786 nfs4_vmask_to_nmask(sarg.vap->va_mask, &resp->attrset); 1787 resp->attrset &= args->createattrs.attrmask; 1788 nfs4_ntov_table_free(&ntov, &sarg); 1789 1790 error = makefh4(&cs->fh, vp, cs->exi); 1791 if (error) { 1792 *cs->statusp = resp->status = puterrno4(error); 1793 } 1794 1795 /* 1796 * The cinfo.atomic = TRUE only if we got no errors, we have 1797 * non-zero va_seq's, and it has incremented by exactly one 1798 * during the creation and it didn't change during the VOP_LOOKUP 1799 * or VOP_FSYNC. 1800 */ 1801 if (!error && bva.va_seq && iva.va_seq && ava.va_seq && 1802 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 1803 resp->cinfo.atomic = TRUE; 1804 else 1805 resp->cinfo.atomic = FALSE; 1806 1807 (void) VOP_FSYNC(vp, syncval, cr, NULL); 1808 1809 if (resp->status != NFS4_OK) { 1810 VN_RELE(vp); 1811 goto out; 1812 } 1813 if (cs->vp) 1814 VN_RELE(cs->vp); 1815 1816 cs->vp = vp; 1817 *cs->statusp = resp->status = NFS4_OK; 1818 out: 1819 DTRACE_NFSV4_2(op__create__done, struct compound_state *, cs, 1820 CREATE4res *, resp); 1821 } 1822 1823 /*ARGSUSED*/ 1824 static void 1825 rfs4_op_delegpurge(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1826 struct compound_state *cs) 1827 { 1828 DTRACE_NFSV4_2(op__delegpurge__start, struct compound_state *, cs, 1829 DELEGPURGE4args *, &argop->nfs_argop4_u.opdelegpurge); 1830 1831 rfs4_op_inval(argop, resop, req, cs); 1832 1833 DTRACE_NFSV4_2(op__delegpurge__done, struct compound_state *, cs, 1834 DELEGPURGE4res *, &resop->nfs_resop4_u.opdelegpurge); 1835 } 1836 1837 /*ARGSUSED*/ 1838 static void 1839 rfs4_op_delegreturn(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 1840 struct compound_state *cs) 1841 { 1842 DELEGRETURN4args *args = &argop->nfs_argop4_u.opdelegreturn; 1843 DELEGRETURN4res *resp = &resop->nfs_resop4_u.opdelegreturn; 1844 rfs4_deleg_state_t *dsp; 1845 nfsstat4 status; 1846 1847 DTRACE_NFSV4_2(op__delegreturn__start, struct compound_state *, cs, 1848 DELEGRETURN4args *, args); 1849 1850 status = rfs4_get_deleg_state(&args->deleg_stateid, &dsp); 1851 resp->status = *cs->statusp = status; 1852 if (status != NFS4_OK) 1853 goto out; 1854 1855 /* Ensure specified filehandle matches */ 1856 if (cs->vp != dsp->finfo->vp) { 1857 resp->status = *cs->statusp = NFS4ERR_BAD_STATEID; 1858 } else 1859 rfs4_return_deleg(dsp, FALSE); 1860 1861 rfs4_update_lease(dsp->client); 1862 1863 rfs4_deleg_state_rele(dsp); 1864 out: 1865 DTRACE_NFSV4_2(op__delegreturn__done, struct compound_state *, cs, 1866 DELEGRETURN4res *, resp); 1867 } 1868 1869 /* 1870 * Check to see if a given "flavor" is an explicitly shared flavor. 1871 * The assumption of this routine is the "flavor" is already a valid 1872 * flavor in the secinfo list of "exi". 1873 * 1874 * e.g. 1875 * # share -o sec=flavor1 /export 1876 * # share -o sec=flavor2 /export/home 1877 * 1878 * flavor2 is not an explicitly shared flavor for /export, 1879 * however it is in the secinfo list for /export thru the 1880 * server namespace setup. 1881 */ 1882 int 1883 is_exported_sec(int flavor, struct exportinfo *exi) 1884 { 1885 int i; 1886 struct secinfo *sp; 1887 1888 sp = exi->exi_export.ex_secinfo; 1889 for (i = 0; i < exi->exi_export.ex_seccnt; i++) { 1890 if (flavor == sp[i].s_secinfo.sc_nfsnum || 1891 sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) { 1892 return (SEC_REF_EXPORTED(&sp[i])); 1893 } 1894 } 1895 1896 /* Should not reach this point based on the assumption */ 1897 return (0); 1898 } 1899 1900 /* 1901 * Check if the security flavor used in the request matches what is 1902 * required at the export point or at the root pseudo node (exi_root). 1903 * 1904 * returns 1 if there's a match or if exported with AUTH_NONE; 0 otherwise. 1905 * 1906 */ 1907 static int 1908 secinfo_match_or_authnone(struct compound_state *cs) 1909 { 1910 int i; 1911 struct secinfo *sp; 1912 1913 /* 1914 * Check cs->nfsflavor (from the request) against 1915 * the current export data in cs->exi. 1916 */ 1917 sp = cs->exi->exi_export.ex_secinfo; 1918 for (i = 0; i < cs->exi->exi_export.ex_seccnt; i++) { 1919 if (cs->nfsflavor == sp[i].s_secinfo.sc_nfsnum || 1920 sp[i].s_secinfo.sc_nfsnum == AUTH_NONE) 1921 return (1); 1922 } 1923 1924 return (0); 1925 } 1926 1927 /* 1928 * Check the access authority for the client and return the correct error. 1929 */ 1930 nfsstat4 1931 call_checkauth4(struct compound_state *cs, struct svc_req *req) 1932 { 1933 int authres; 1934 1935 /* 1936 * First, check if the security flavor used in the request 1937 * are among the flavors set in the server namespace. 1938 */ 1939 if (!secinfo_match_or_authnone(cs)) { 1940 *cs->statusp = NFS4ERR_WRONGSEC; 1941 return (*cs->statusp); 1942 } 1943 1944 authres = checkauth4(cs, req); 1945 1946 if (authres > 0) { 1947 *cs->statusp = NFS4_OK; 1948 if (! (cs->access & CS_ACCESS_LIMITED)) 1949 cs->access = CS_ACCESS_OK; 1950 } else if (authres == 0) { 1951 *cs->statusp = NFS4ERR_ACCESS; 1952 } else if (authres == -2) { 1953 *cs->statusp = NFS4ERR_WRONGSEC; 1954 } else { 1955 *cs->statusp = NFS4ERR_DELAY; 1956 } 1957 return (*cs->statusp); 1958 } 1959 1960 /* 1961 * bitmap4_to_attrmask is called by getattr and readdir. 1962 * It sets up the vattr mask and determines whether vfsstat call is needed 1963 * based on the input bitmap. 1964 * Returns nfsv4 status. 1965 */ 1966 static nfsstat4 1967 bitmap4_to_attrmask(bitmap4 breq, struct nfs4_svgetit_arg *sargp) 1968 { 1969 int i; 1970 uint_t va_mask; 1971 struct statvfs64 *sbp = sargp->sbp; 1972 1973 sargp->sbp = NULL; 1974 sargp->flag = 0; 1975 sargp->rdattr_error = NFS4_OK; 1976 sargp->mntdfid_set = FALSE; 1977 if (sargp->cs->vp) 1978 sargp->xattr = get_fh4_flag(&sargp->cs->fh, 1979 FH4_ATTRDIR | FH4_NAMEDATTR); 1980 else 1981 sargp->xattr = 0; 1982 1983 /* 1984 * Set rdattr_error_req to true if return error per 1985 * failed entry rather than fail the readdir. 1986 */ 1987 if (breq & FATTR4_RDATTR_ERROR_MASK) 1988 sargp->rdattr_error_req = 1; 1989 else 1990 sargp->rdattr_error_req = 0; 1991 1992 /* 1993 * generate the va_mask 1994 * Handle the easy cases first 1995 */ 1996 switch (breq) { 1997 case NFS4_NTOV_ATTR_MASK: 1998 sargp->vap->va_mask = NFS4_NTOV_ATTR_AT_MASK; 1999 return (NFS4_OK); 2000 2001 case NFS4_FS_ATTR_MASK: 2002 sargp->vap->va_mask = NFS4_FS_ATTR_AT_MASK; 2003 sargp->sbp = sbp; 2004 return (NFS4_OK); 2005 2006 case NFS4_NTOV_ATTR_CACHE_MASK: 2007 sargp->vap->va_mask = NFS4_NTOV_ATTR_CACHE_AT_MASK; 2008 return (NFS4_OK); 2009 2010 case FATTR4_LEASE_TIME_MASK: 2011 sargp->vap->va_mask = 0; 2012 return (NFS4_OK); 2013 2014 default: 2015 va_mask = 0; 2016 for (i = 0; i < nfs4_ntov_map_size; i++) { 2017 if ((breq & nfs4_ntov_map[i].fbit) && 2018 nfs4_ntov_map[i].vbit) 2019 va_mask |= nfs4_ntov_map[i].vbit; 2020 } 2021 2022 /* 2023 * Check is vfsstat is needed 2024 */ 2025 if (breq & NFS4_FS_ATTR_MASK) 2026 sargp->sbp = sbp; 2027 2028 sargp->vap->va_mask = va_mask; 2029 return (NFS4_OK); 2030 } 2031 /* NOTREACHED */ 2032 } 2033 2034 /* 2035 * bitmap4_get_sysattrs is called by getattr and readdir. 2036 * It calls both VOP_GETATTR and VFS_STATVFS calls to get the attrs. 2037 * Returns nfsv4 status. 2038 */ 2039 static nfsstat4 2040 bitmap4_get_sysattrs(struct nfs4_svgetit_arg *sargp) 2041 { 2042 int error; 2043 struct compound_state *cs = sargp->cs; 2044 vnode_t *vp = cs->vp; 2045 2046 if (sargp->sbp != NULL) { 2047 if (error = VFS_STATVFS(vp->v_vfsp, sargp->sbp)) { 2048 sargp->sbp = NULL; /* to identify error */ 2049 return (puterrno4(error)); 2050 } 2051 } 2052 2053 return (rfs4_vop_getattr(vp, sargp->vap, 0, cs->cr)); 2054 } 2055 2056 static void 2057 nfs4_ntov_table_init(struct nfs4_ntov_table *ntovp) 2058 { 2059 ntovp->na = kmem_zalloc(sizeof (union nfs4_attr_u) * nfs4_ntov_map_size, 2060 KM_SLEEP); 2061 ntovp->attrcnt = 0; 2062 ntovp->vfsstat = FALSE; 2063 } 2064 2065 static void 2066 nfs4_ntov_table_free(struct nfs4_ntov_table *ntovp, 2067 struct nfs4_svgetit_arg *sargp) 2068 { 2069 int i; 2070 union nfs4_attr_u *na; 2071 uint8_t *amap; 2072 2073 /* 2074 * XXX Should do the same checks for whether the bit is set 2075 */ 2076 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2077 i < ntovp->attrcnt; i++, na++, amap++) { 2078 (void) (*nfs4_ntov_map[*amap].sv_getit)( 2079 NFS4ATTR_FREEIT, sargp, na); 2080 } 2081 if ((sargp->op == NFS4ATTR_SETIT) || (sargp->op == NFS4ATTR_VERIT)) { 2082 /* 2083 * xdr_free for getattr will be done later 2084 */ 2085 for (i = 0, na = ntovp->na, amap = ntovp->amap; 2086 i < ntovp->attrcnt; i++, na++, amap++) { 2087 xdr_free(nfs4_ntov_map[*amap].xfunc, (caddr_t)na); 2088 } 2089 } 2090 kmem_free(ntovp->na, sizeof (union nfs4_attr_u) * nfs4_ntov_map_size); 2091 } 2092 2093 /* 2094 * do_rfs4_op_getattr gets the system attrs and converts into fattr4. 2095 */ 2096 static nfsstat4 2097 do_rfs4_op_getattr(bitmap4 breq, fattr4 *fattrp, 2098 struct nfs4_svgetit_arg *sargp) 2099 { 2100 int error = 0; 2101 int i, k; 2102 struct nfs4_ntov_table ntov; 2103 XDR xdr; 2104 ulong_t xdr_size; 2105 char *xdr_attrs; 2106 nfsstat4 status = NFS4_OK; 2107 nfsstat4 prev_rdattr_error = sargp->rdattr_error; 2108 union nfs4_attr_u *na; 2109 uint8_t *amap; 2110 2111 sargp->op = NFS4ATTR_GETIT; 2112 sargp->flag = 0; 2113 2114 fattrp->attrmask = 0; 2115 /* if no bits requested, then return empty fattr4 */ 2116 if (breq == 0) { 2117 fattrp->attrlist4_len = 0; 2118 fattrp->attrlist4 = NULL; 2119 return (NFS4_OK); 2120 } 2121 2122 /* 2123 * return NFS4ERR_INVAL when client requests write-only attrs 2124 */ 2125 if (breq & (FATTR4_TIME_ACCESS_SET_MASK | FATTR4_TIME_MODIFY_SET_MASK)) 2126 return (NFS4ERR_INVAL); 2127 2128 nfs4_ntov_table_init(&ntov); 2129 na = ntov.na; 2130 amap = ntov.amap; 2131 2132 /* 2133 * Now loop to get or verify the attrs 2134 */ 2135 for (i = 0; i < nfs4_ntov_map_size; i++) { 2136 if (breq & nfs4_ntov_map[i].fbit) { 2137 if ((*nfs4_ntov_map[i].sv_getit)( 2138 NFS4ATTR_SUPPORTED, sargp, NULL) == 0) { 2139 2140 error = (*nfs4_ntov_map[i].sv_getit)( 2141 NFS4ATTR_GETIT, sargp, na); 2142 2143 /* 2144 * Possible error values: 2145 * >0 if sv_getit failed to 2146 * get the attr; 0 if succeeded; 2147 * <0 if rdattr_error and the 2148 * attribute cannot be returned. 2149 */ 2150 if (error && !(sargp->rdattr_error_req)) 2151 goto done; 2152 /* 2153 * If error then just for entry 2154 */ 2155 if (error == 0) { 2156 fattrp->attrmask |= 2157 nfs4_ntov_map[i].fbit; 2158 *amap++ = 2159 (uint8_t)nfs4_ntov_map[i].nval; 2160 na++; 2161 (ntov.attrcnt)++; 2162 } else if ((error > 0) && 2163 (sargp->rdattr_error == NFS4_OK)) { 2164 sargp->rdattr_error = puterrno4(error); 2165 } 2166 error = 0; 2167 } 2168 } 2169 } 2170 2171 /* 2172 * If rdattr_error was set after the return value for it was assigned, 2173 * update it. 2174 */ 2175 if (prev_rdattr_error != sargp->rdattr_error) { 2176 na = ntov.na; 2177 amap = ntov.amap; 2178 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2179 k = *amap; 2180 if (k < FATTR4_RDATTR_ERROR) { 2181 continue; 2182 } 2183 if ((k == FATTR4_RDATTR_ERROR) && 2184 ((*nfs4_ntov_map[k].sv_getit)( 2185 NFS4ATTR_SUPPORTED, sargp, NULL) == 0)) { 2186 2187 (void) (*nfs4_ntov_map[k].sv_getit)( 2188 NFS4ATTR_GETIT, sargp, na); 2189 } 2190 break; 2191 } 2192 } 2193 2194 xdr_size = 0; 2195 na = ntov.na; 2196 amap = ntov.amap; 2197 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2198 xdr_size += xdr_sizeof(nfs4_ntov_map[*amap].xfunc, na); 2199 } 2200 2201 fattrp->attrlist4_len = xdr_size; 2202 if (xdr_size) { 2203 /* freed by rfs4_op_getattr_free() */ 2204 fattrp->attrlist4 = xdr_attrs = kmem_zalloc(xdr_size, KM_SLEEP); 2205 2206 xdrmem_create(&xdr, xdr_attrs, xdr_size, XDR_ENCODE); 2207 2208 na = ntov.na; 2209 amap = ntov.amap; 2210 for (i = 0; i < ntov.attrcnt; i++, na++, amap++) { 2211 if (!(*nfs4_ntov_map[*amap].xfunc)(&xdr, na)) { 2212 cmn_err(CE_WARN, "do_rfs4_op_getattr: xdr " 2213 "encode of attribute %d failed\n", *amap); 2214 status = NFS4ERR_SERVERFAULT; 2215 break; 2216 } 2217 } 2218 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 2219 } else { 2220 fattrp->attrlist4 = NULL; 2221 } 2222 done: 2223 2224 nfs4_ntov_table_free(&ntov, sargp); 2225 2226 if (error != 0) 2227 status = puterrno4(error); 2228 2229 return (status); 2230 } 2231 2232 /* ARGSUSED */ 2233 static void 2234 rfs4_op_getattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2235 struct compound_state *cs) 2236 { 2237 GETATTR4args *args = &argop->nfs_argop4_u.opgetattr; 2238 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2239 struct nfs4_svgetit_arg sarg; 2240 struct statvfs64 sb; 2241 nfsstat4 status; 2242 2243 DTRACE_NFSV4_2(op__getattr__start, struct compound_state *, cs, 2244 GETATTR4args *, args); 2245 2246 if (cs->vp == NULL) { 2247 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2248 goto out; 2249 } 2250 2251 if (cs->access == CS_ACCESS_DENIED) { 2252 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2253 goto out; 2254 } 2255 2256 sarg.sbp = &sb; 2257 sarg.cs = cs; 2258 2259 status = bitmap4_to_attrmask(args->attr_request, &sarg); 2260 if (status == NFS4_OK) { 2261 status = bitmap4_get_sysattrs(&sarg); 2262 if (status == NFS4_OK) 2263 status = do_rfs4_op_getattr(args->attr_request, 2264 &resp->obj_attributes, &sarg); 2265 } 2266 *cs->statusp = resp->status = status; 2267 out: 2268 DTRACE_NFSV4_2(op__getattr__done, struct compound_state *, cs, 2269 GETATTR4res *, resp); 2270 } 2271 2272 static void 2273 rfs4_op_getattr_free(nfs_resop4 *resop) 2274 { 2275 GETATTR4res *resp = &resop->nfs_resop4_u.opgetattr; 2276 2277 nfs4_fattr4_free(&resp->obj_attributes); 2278 } 2279 2280 /* ARGSUSED */ 2281 static void 2282 rfs4_op_getfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2283 struct compound_state *cs) 2284 { 2285 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2286 2287 DTRACE_NFSV4_1(op__getfh__start, struct compound_state *, cs); 2288 2289 if (cs->vp == NULL) { 2290 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2291 goto out; 2292 } 2293 if (cs->access == CS_ACCESS_DENIED) { 2294 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2295 goto out; 2296 } 2297 2298 resp->object.nfs_fh4_val = 2299 kmem_alloc(cs->fh.nfs_fh4_len, KM_SLEEP); 2300 nfs_fh4_copy(&cs->fh, &resp->object); 2301 *cs->statusp = resp->status = NFS4_OK; 2302 out: 2303 DTRACE_NFSV4_2(op__getfh__done, struct compound_state *, cs, 2304 GETFH4res *, resp); 2305 } 2306 2307 static void 2308 rfs4_op_getfh_free(nfs_resop4 *resop) 2309 { 2310 GETFH4res *resp = &resop->nfs_resop4_u.opgetfh; 2311 2312 if (resp->status == NFS4_OK && 2313 resp->object.nfs_fh4_val != NULL) { 2314 kmem_free(resp->object.nfs_fh4_val, resp->object.nfs_fh4_len); 2315 resp->object.nfs_fh4_val = NULL; 2316 resp->object.nfs_fh4_len = 0; 2317 } 2318 } 2319 2320 /* 2321 * illegal: args: void 2322 * res : status (NFS4ERR_OP_ILLEGAL) 2323 */ 2324 /* ARGSUSED */ 2325 static void 2326 rfs4_op_illegal(nfs_argop4 *argop, nfs_resop4 *resop, 2327 struct svc_req *req, struct compound_state *cs) 2328 { 2329 ILLEGAL4res *resp = &resop->nfs_resop4_u.opillegal; 2330 2331 resop->resop = OP_ILLEGAL; 2332 *cs->statusp = resp->status = NFS4ERR_OP_ILLEGAL; 2333 } 2334 2335 /* 2336 * link: args: SAVED_FH: file, CURRENT_FH: target directory 2337 * res: status. If success - CURRENT_FH unchanged, return change_info 2338 */ 2339 /* ARGSUSED */ 2340 static void 2341 rfs4_op_link(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2342 struct compound_state *cs) 2343 { 2344 LINK4args *args = &argop->nfs_argop4_u.oplink; 2345 LINK4res *resp = &resop->nfs_resop4_u.oplink; 2346 int error; 2347 vnode_t *vp; 2348 vnode_t *dvp; 2349 struct vattr bdva, idva, adva; 2350 char *nm; 2351 uint_t len; 2352 2353 DTRACE_NFSV4_2(op__link__start, struct compound_state *, cs, 2354 LINK4args *, args); 2355 2356 /* SAVED_FH: source object */ 2357 vp = cs->saved_vp; 2358 if (vp == NULL) { 2359 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2360 goto out; 2361 } 2362 2363 /* CURRENT_FH: target directory */ 2364 dvp = cs->vp; 2365 if (dvp == NULL) { 2366 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2367 goto out; 2368 } 2369 2370 /* 2371 * If there is a non-shared filesystem mounted on this vnode, 2372 * do not allow to link any file in this directory. 2373 */ 2374 if (vn_ismntpt(dvp)) { 2375 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2376 goto out; 2377 } 2378 2379 if (cs->access == CS_ACCESS_DENIED) { 2380 *cs->statusp = resp->status = NFS4ERR_ACCESS; 2381 goto out; 2382 } 2383 2384 /* Check source object's type validity */ 2385 if (vp->v_type == VDIR) { 2386 *cs->statusp = resp->status = NFS4ERR_ISDIR; 2387 goto out; 2388 } 2389 2390 /* Check target directory's type */ 2391 if (dvp->v_type != VDIR) { 2392 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2393 goto out; 2394 } 2395 2396 if (cs->saved_exi != cs->exi) { 2397 *cs->statusp = resp->status = NFS4ERR_XDEV; 2398 goto out; 2399 } 2400 2401 if (!utf8_dir_verify(&args->newname)) { 2402 *cs->statusp = resp->status = NFS4ERR_INVAL; 2403 goto out; 2404 } 2405 2406 nm = utf8_to_fn(&args->newname, &len, NULL); 2407 if (nm == NULL) { 2408 *cs->statusp = resp->status = NFS4ERR_INVAL; 2409 goto out; 2410 } 2411 2412 if (len > MAXNAMELEN) { 2413 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 2414 kmem_free(nm, len); 2415 goto out; 2416 } 2417 2418 if (rdonly4(cs->exi, cs->vp, req)) { 2419 *cs->statusp = resp->status = NFS4ERR_ROFS; 2420 kmem_free(nm, len); 2421 goto out; 2422 } 2423 2424 /* Get "before" change value */ 2425 bdva.va_mask = AT_CTIME|AT_SEQ; 2426 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 2427 if (error) { 2428 *cs->statusp = resp->status = puterrno4(error); 2429 kmem_free(nm, len); 2430 goto out; 2431 } 2432 2433 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 2434 2435 error = VOP_LINK(dvp, vp, nm, cs->cr, NULL, 0); 2436 2437 kmem_free(nm, len); 2438 2439 /* 2440 * Get the initial "after" sequence number, if it fails, set to zero 2441 */ 2442 idva.va_mask = AT_SEQ; 2443 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 2444 idva.va_seq = 0; 2445 2446 /* 2447 * Force modified data and metadata out to stable storage. 2448 */ 2449 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 2450 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 2451 2452 if (error) { 2453 *cs->statusp = resp->status = puterrno4(error); 2454 goto out; 2455 } 2456 2457 /* 2458 * Get "after" change value, if it fails, simply return the 2459 * before value. 2460 */ 2461 adva.va_mask = AT_CTIME|AT_SEQ; 2462 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 2463 adva.va_ctime = bdva.va_ctime; 2464 adva.va_seq = 0; 2465 } 2466 2467 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 2468 2469 /* 2470 * The cinfo.atomic = TRUE only if we have 2471 * non-zero va_seq's, and it has incremented by exactly one 2472 * during the VOP_LINK and it didn't change during the VOP_FSYNC. 2473 */ 2474 if (bdva.va_seq && idva.va_seq && adva.va_seq && 2475 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 2476 resp->cinfo.atomic = TRUE; 2477 else 2478 resp->cinfo.atomic = FALSE; 2479 2480 *cs->statusp = resp->status = NFS4_OK; 2481 out: 2482 DTRACE_NFSV4_2(op__link__done, struct compound_state *, cs, 2483 LINK4res *, resp); 2484 } 2485 2486 /* 2487 * Used by rfs4_op_lookup and rfs4_op_lookupp to do the actual work. 2488 */ 2489 2490 /* ARGSUSED */ 2491 static nfsstat4 2492 do_rfs4_op_lookup(char *nm, uint_t buflen, struct svc_req *req, 2493 struct compound_state *cs) 2494 { 2495 int error; 2496 int different_export = 0; 2497 vnode_t *vp, *tvp, *pre_tvp = NULL, *oldvp = NULL; 2498 struct exportinfo *exi = NULL, *pre_exi = NULL; 2499 nfsstat4 stat; 2500 fid_t fid; 2501 int attrdir, dotdot, walk; 2502 bool_t is_newvp = FALSE; 2503 2504 if (cs->vp->v_flag & V_XATTRDIR) { 2505 attrdir = 1; 2506 ASSERT(get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2507 } else { 2508 attrdir = 0; 2509 ASSERT(! get_fh4_flag(&cs->fh, FH4_ATTRDIR)); 2510 } 2511 2512 dotdot = (nm[0] == '.' && nm[1] == '.' && nm[2] == '\0'); 2513 2514 /* 2515 * If dotdotting, then need to check whether it's 2516 * above the root of a filesystem, or above an 2517 * export point. 2518 */ 2519 if (dotdot) { 2520 2521 /* 2522 * If dotdotting at the root of a filesystem, then 2523 * need to traverse back to the mounted-on filesystem 2524 * and do the dotdot lookup there. 2525 */ 2526 if (cs->vp->v_flag & VROOT) { 2527 2528 /* 2529 * If at the system root, then can 2530 * go up no further. 2531 */ 2532 if (VN_CMP(cs->vp, rootdir)) 2533 return (puterrno4(ENOENT)); 2534 2535 /* 2536 * Traverse back to the mounted-on filesystem 2537 */ 2538 cs->vp = untraverse(cs->vp); 2539 2540 /* 2541 * Set the different_export flag so we remember 2542 * to pick up a new exportinfo entry for 2543 * this new filesystem. 2544 */ 2545 different_export = 1; 2546 } else { 2547 2548 /* 2549 * If dotdotting above an export point then set 2550 * the different_export to get new export info. 2551 */ 2552 different_export = nfs_exported(cs->exi, cs->vp); 2553 } 2554 } 2555 2556 error = VOP_LOOKUP(cs->vp, nm, &vp, NULL, 0, NULL, cs->cr, 2557 NULL, NULL, NULL); 2558 if (error) 2559 return (puterrno4(error)); 2560 2561 /* 2562 * If the vnode is in a pseudo filesystem, check whether it is visible. 2563 * 2564 * XXX if the vnode is a symlink and it is not visible in 2565 * a pseudo filesystem, return ENOENT (not following symlink). 2566 * V4 client can not mount such symlink. This is a regression 2567 * from V2/V3. 2568 * 2569 * In the same exported filesystem, if the security flavor used 2570 * is not an explicitly shared flavor, limit the view to the visible 2571 * list entries only. This is not a WRONGSEC case because it's already 2572 * checked via PUTROOTFH/PUTPUBFH or PUTFH. 2573 */ 2574 if (!different_export && 2575 (PSEUDO(cs->exi) || ! is_exported_sec(cs->nfsflavor, cs->exi) || 2576 cs->access & CS_ACCESS_LIMITED)) { 2577 if (! nfs_visible(cs->exi, vp, &different_export)) { 2578 VN_RELE(vp); 2579 return (puterrno4(ENOENT)); 2580 } 2581 } 2582 2583 /* 2584 * If it's a mountpoint, then traverse it. 2585 */ 2586 if (vn_ismntpt(vp)) { 2587 pre_exi = cs->exi; /* save pre-traversed exportinfo */ 2588 pre_tvp = vp; /* save pre-traversed vnode */ 2589 2590 /* 2591 * hold pre_tvp to counteract rele by traverse. We will 2592 * need pre_tvp below if checkexport4 fails 2593 */ 2594 VN_HOLD(pre_tvp); 2595 tvp = vp; 2596 if ((error = traverse(&tvp)) != 0) { 2597 VN_RELE(vp); 2598 VN_RELE(pre_tvp); 2599 return (puterrno4(error)); 2600 } 2601 vp = tvp; 2602 different_export = 1; 2603 } else if (vp->v_vfsp != cs->vp->v_vfsp) { 2604 /* 2605 * The vfsp comparison is to handle the case where 2606 * a LOFS mount is shared. lo_lookup traverses mount points, 2607 * and NFS is unaware of local fs transistions because 2608 * v_vfsmountedhere isn't set. For this special LOFS case, 2609 * the dir and the obj returned by lookup will have different 2610 * vfs ptrs. 2611 */ 2612 different_export = 1; 2613 } 2614 2615 if (different_export) { 2616 2617 bzero(&fid, sizeof (fid)); 2618 fid.fid_len = MAXFIDSZ; 2619 error = vop_fid_pseudo(vp, &fid); 2620 if (error) { 2621 VN_RELE(vp); 2622 if (pre_tvp) 2623 VN_RELE(pre_tvp); 2624 return (puterrno4(error)); 2625 } 2626 2627 if (dotdot) 2628 exi = nfs_vptoexi(NULL, vp, cs->cr, &walk, NULL, TRUE); 2629 else 2630 exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp); 2631 2632 if (exi == NULL) { 2633 if (pre_tvp) { 2634 /* 2635 * If this vnode is a mounted-on vnode, 2636 * but the mounted-on file system is not 2637 * exported, send back the filehandle for 2638 * the mounted-on vnode, not the root of 2639 * the mounted-on file system. 2640 */ 2641 VN_RELE(vp); 2642 vp = pre_tvp; 2643 exi = pre_exi; 2644 } else { 2645 VN_RELE(vp); 2646 return (puterrno4(EACCES)); 2647 } 2648 } else if (pre_tvp) { 2649 /* we're done with pre_tvp now. release extra hold */ 2650 VN_RELE(pre_tvp); 2651 } 2652 2653 cs->exi = exi; 2654 2655 /* 2656 * Now we do a checkauth4. The reason is that 2657 * this client/user may not have access to the new 2658 * exported file system, and if he does, 2659 * the client/user may be mapped to a different uid. 2660 * 2661 * We start with a new cr, because the checkauth4 done 2662 * in the PUT*FH operation over wrote the cred's uid, 2663 * gid, etc, and we want the real thing before calling 2664 * checkauth4() 2665 */ 2666 crfree(cs->cr); 2667 cs->cr = crdup(cs->basecr); 2668 2669 if (cs->vp) 2670 oldvp = cs->vp; 2671 cs->vp = vp; 2672 is_newvp = TRUE; 2673 2674 stat = call_checkauth4(cs, req); 2675 if (stat != NFS4_OK) { 2676 VN_RELE(cs->vp); 2677 cs->vp = oldvp; 2678 return (stat); 2679 } 2680 } 2681 2682 /* 2683 * After various NFS checks, do a label check on the path 2684 * component. The label on this path should either be the 2685 * global zone's label or a zone's label. We are only 2686 * interested in the zone's label because exported files 2687 * in global zone is accessible (though read-only) to 2688 * clients. The exportability/visibility check is already 2689 * done before reaching this code. 2690 */ 2691 if (is_system_labeled()) { 2692 bslabel_t *clabel; 2693 2694 ASSERT(req->rq_label != NULL); 2695 clabel = req->rq_label; 2696 DTRACE_PROBE2(tx__rfs4__log__info__oplookup__clabel, char *, 2697 "got client label from request(1)", struct svc_req *, req); 2698 2699 if (!blequal(&l_admin_low->tsl_label, clabel)) { 2700 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) { 2701 error = EACCES; 2702 goto err_out; 2703 } 2704 } else { 2705 /* 2706 * We grant access to admin_low label clients 2707 * only if the client is trusted, i.e. also 2708 * running Solaris Trusted Extension. 2709 */ 2710 struct sockaddr *ca; 2711 int addr_type; 2712 void *ipaddr; 2713 tsol_tpc_t *tp; 2714 2715 ca = (struct sockaddr *)svc_getrpccaller( 2716 req->rq_xprt)->buf; 2717 if (ca->sa_family == AF_INET) { 2718 addr_type = IPV4_VERSION; 2719 ipaddr = &((struct sockaddr_in *)ca)->sin_addr; 2720 } else if (ca->sa_family == AF_INET6) { 2721 addr_type = IPV6_VERSION; 2722 ipaddr = &((struct sockaddr_in6 *) 2723 ca)->sin6_addr; 2724 } 2725 tp = find_tpc(ipaddr, addr_type, B_FALSE); 2726 if (tp == NULL || tp->tpc_tp.tp_doi != 2727 l_admin_low->tsl_doi || tp->tpc_tp.host_type != 2728 SUN_CIPSO) { 2729 if (tp != NULL) 2730 TPC_RELE(tp); 2731 error = EACCES; 2732 goto err_out; 2733 } 2734 TPC_RELE(tp); 2735 } 2736 } 2737 2738 error = makefh4(&cs->fh, vp, cs->exi); 2739 2740 err_out: 2741 if (error) { 2742 if (is_newvp) { 2743 VN_RELE(cs->vp); 2744 cs->vp = oldvp; 2745 } else 2746 VN_RELE(vp); 2747 return (puterrno4(error)); 2748 } 2749 2750 if (!is_newvp) { 2751 if (cs->vp) 2752 VN_RELE(cs->vp); 2753 cs->vp = vp; 2754 } else if (oldvp) 2755 VN_RELE(oldvp); 2756 2757 /* 2758 * if did lookup on attrdir and didn't lookup .., set named 2759 * attr fh flag 2760 */ 2761 if (attrdir && ! dotdot) 2762 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 2763 2764 /* Assume false for now, open proc will set this */ 2765 cs->mandlock = FALSE; 2766 2767 return (NFS4_OK); 2768 } 2769 2770 /* ARGSUSED */ 2771 static void 2772 rfs4_op_lookup(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2773 struct compound_state *cs) 2774 { 2775 LOOKUP4args *args = &argop->nfs_argop4_u.oplookup; 2776 LOOKUP4res *resp = &resop->nfs_resop4_u.oplookup; 2777 char *nm; 2778 uint_t len; 2779 2780 DTRACE_NFSV4_2(op__lookup__start, struct compound_state *, cs, 2781 LOOKUP4args *, args); 2782 2783 if (cs->vp == NULL) { 2784 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2785 goto out; 2786 } 2787 2788 if (cs->vp->v_type == VLNK) { 2789 *cs->statusp = resp->status = NFS4ERR_SYMLINK; 2790 goto out; 2791 } 2792 2793 if (cs->vp->v_type != VDIR) { 2794 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2795 goto out; 2796 } 2797 2798 if (!utf8_dir_verify(&args->objname)) { 2799 *cs->statusp = resp->status = NFS4ERR_INVAL; 2800 goto out; 2801 } 2802 2803 nm = utf8_to_str(&args->objname, &len, NULL); 2804 if (nm == NULL) { 2805 *cs->statusp = resp->status = NFS4ERR_INVAL; 2806 goto out; 2807 } 2808 2809 if (len > MAXNAMELEN) { 2810 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 2811 kmem_free(nm, len); 2812 goto out; 2813 } 2814 2815 *cs->statusp = resp->status = do_rfs4_op_lookup(nm, len, req, cs); 2816 2817 kmem_free(nm, len); 2818 2819 out: 2820 DTRACE_NFSV4_2(op__lookup__done, struct compound_state *, cs, 2821 LOOKUP4res *, resp); 2822 } 2823 2824 /* ARGSUSED */ 2825 static void 2826 rfs4_op_lookupp(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 2827 struct compound_state *cs) 2828 { 2829 LOOKUPP4res *resp = &resop->nfs_resop4_u.oplookupp; 2830 2831 DTRACE_NFSV4_1(op__lookupp__start, struct compound_state *, cs); 2832 2833 if (cs->vp == NULL) { 2834 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2835 goto out; 2836 } 2837 2838 if (cs->vp->v_type != VDIR) { 2839 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 2840 goto out; 2841 } 2842 2843 *cs->statusp = resp->status = do_rfs4_op_lookup("..", 3, req, cs); 2844 2845 /* 2846 * From NFSV4 Specification, LOOKUPP should not check for 2847 * NFS4ERR_WRONGSEC. Retrun NFS4_OK instead. 2848 */ 2849 if (resp->status == NFS4ERR_WRONGSEC) { 2850 *cs->statusp = resp->status = NFS4_OK; 2851 } 2852 2853 out: 2854 DTRACE_NFSV4_2(op__lookupp__done, struct compound_state *, cs, 2855 LOOKUPP4res *, resp); 2856 } 2857 2858 2859 /*ARGSUSED2*/ 2860 static void 2861 rfs4_op_openattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 2862 struct compound_state *cs) 2863 { 2864 OPENATTR4args *args = &argop->nfs_argop4_u.opopenattr; 2865 OPENATTR4res *resp = &resop->nfs_resop4_u.opopenattr; 2866 vnode_t *avp = NULL; 2867 int lookup_flags = LOOKUP_XATTR, error; 2868 int exp_ro = 0; 2869 2870 DTRACE_NFSV4_2(op__openattr__start, struct compound_state *, cs, 2871 OPENATTR4args *, args); 2872 2873 if (cs->vp == NULL) { 2874 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 2875 goto out; 2876 } 2877 2878 /* 2879 * Make a couple of checks made by copen() 2880 * 2881 * Check to make sure underlying fs supports xattrs. This 2882 * is required because solaris filesystem implementations 2883 * (UFS/TMPFS) don't enforce the noxattr mount option 2884 * in VOP_LOOKUP(LOOKUP_XATTR). If fs doesn't support this 2885 * pathconf cmd or if fs supports cmd but doesn't claim 2886 * support for xattr, return NOTSUPP. It would be better 2887 * to use VOP_PATHCONF( _PC_XATTR_ENABLED) for this; however, 2888 * that cmd is not available to VOP_PATHCONF interface 2889 * (it's only implemented inside pathconf syscall)... 2890 * 2891 * Verify permission to put attributes on files (access 2892 * checks from copen). 2893 */ 2894 2895 if ((cs->vp->v_vfsp->vfs_flag & VFS_XATTR) == 0 && 2896 !vfs_has_feature(cs->vp->v_vfsp, VFSFT_XVATTR)) { 2897 *cs->statusp = resp->status = puterrno4(ENOTSUP); 2898 goto out; 2899 } 2900 2901 if ((VOP_ACCESS(cs->vp, VREAD, 0, cs->cr, NULL) != 0) && 2902 (VOP_ACCESS(cs->vp, VWRITE, 0, cs->cr, NULL) != 0) && 2903 (VOP_ACCESS(cs->vp, VEXEC, 0, cs->cr, NULL) != 0)) { 2904 *cs->statusp = resp->status = puterrno4(EACCES); 2905 goto out; 2906 } 2907 2908 /* 2909 * The CREATE_XATTR_DIR VOP flag cannot be specified if 2910 * the file system is exported read-only -- regardless of 2911 * createdir flag. Otherwise the attrdir would be created 2912 * (assuming server fs isn't mounted readonly locally). If 2913 * VOP_LOOKUP returns ENOENT in this case, the error will 2914 * be translated into EROFS. ENOSYS is mapped to ENOTSUP 2915 * because specfs has no VOP_LOOKUP op, so the macro would 2916 * return ENOSYS. EINVAL is returned by all (current) 2917 * Solaris file system implementations when any of their 2918 * restrictions are violated (xattr(dir) can't have xattrdir). 2919 * Returning NOTSUPP is more appropriate in this case 2920 * because the object will never be able to have an attrdir. 2921 */ 2922 if (args->createdir && ! (exp_ro = rdonly4(cs->exi, cs->vp, req))) 2923 lookup_flags |= CREATE_XATTR_DIR; 2924 2925 error = VOP_LOOKUP(cs->vp, "", &avp, NULL, lookup_flags, NULL, cs->cr, 2926 NULL, NULL, NULL); 2927 2928 if (error) { 2929 if (error == ENOENT && args->createdir && exp_ro) 2930 *cs->statusp = resp->status = puterrno4(EROFS); 2931 else if (error == EINVAL || error == ENOSYS) 2932 *cs->statusp = resp->status = puterrno4(ENOTSUP); 2933 else 2934 *cs->statusp = resp->status = puterrno4(error); 2935 goto out; 2936 } 2937 2938 ASSERT(avp->v_flag & V_XATTRDIR); 2939 2940 error = makefh4(&cs->fh, avp, cs->exi); 2941 2942 if (error) { 2943 VN_RELE(avp); 2944 *cs->statusp = resp->status = puterrno4(error); 2945 goto out; 2946 } 2947 2948 VN_RELE(cs->vp); 2949 cs->vp = avp; 2950 2951 /* 2952 * There is no requirement for an attrdir fh flag 2953 * because the attrdir has a vnode flag to distinguish 2954 * it from regular (non-xattr) directories. The 2955 * FH4_ATTRDIR flag is set for future sanity checks. 2956 */ 2957 set_fh4_flag(&cs->fh, FH4_ATTRDIR); 2958 *cs->statusp = resp->status = NFS4_OK; 2959 2960 out: 2961 DTRACE_NFSV4_2(op__openattr__done, struct compound_state *, cs, 2962 OPENATTR4res *, resp); 2963 } 2964 2965 static int 2966 do_io(int direction, vnode_t *vp, struct uio *uio, int ioflag, cred_t *cred, 2967 caller_context_t *ct) 2968 { 2969 int error; 2970 int i; 2971 clock_t delaytime; 2972 2973 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 2974 2975 /* 2976 * Don't block on mandatory locks. If this routine returns 2977 * EAGAIN, the caller should return NFS4ERR_LOCKED. 2978 */ 2979 uio->uio_fmode = FNONBLOCK; 2980 2981 for (i = 0; i < rfs4_maxlock_tries; i++) { 2982 2983 2984 if (direction == FREAD) { 2985 (void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, ct); 2986 error = VOP_READ(vp, uio, ioflag, cred, ct); 2987 VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, ct); 2988 } else { 2989 (void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, ct); 2990 error = VOP_WRITE(vp, uio, ioflag, cred, ct); 2991 VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, ct); 2992 } 2993 2994 if (error != EAGAIN) 2995 break; 2996 2997 if (i < rfs4_maxlock_tries - 1) { 2998 delay(delaytime); 2999 delaytime *= 2; 3000 } 3001 } 3002 3003 return (error); 3004 } 3005 3006 /* ARGSUSED */ 3007 static void 3008 rfs4_op_read(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3009 struct compound_state *cs) 3010 { 3011 READ4args *args = &argop->nfs_argop4_u.opread; 3012 READ4res *resp = &resop->nfs_resop4_u.opread; 3013 int error; 3014 int verror; 3015 vnode_t *vp; 3016 struct vattr va; 3017 struct iovec iov; 3018 struct uio uio; 3019 u_offset_t offset; 3020 bool_t *deleg = &cs->deleg; 3021 nfsstat4 stat; 3022 int in_crit = 0; 3023 mblk_t *mp; 3024 int alloc_err = 0; 3025 caller_context_t ct; 3026 3027 DTRACE_NFSV4_2(op__read__start, struct compound_state *, cs, 3028 READ4args, args); 3029 3030 vp = cs->vp; 3031 if (vp == NULL) { 3032 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3033 goto out; 3034 } 3035 if (cs->access == CS_ACCESS_DENIED) { 3036 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3037 goto out; 3038 } 3039 3040 if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE, 3041 deleg, TRUE, &ct)) != NFS4_OK) { 3042 *cs->statusp = resp->status = stat; 3043 goto out; 3044 } 3045 3046 /* 3047 * Enter the critical region before calling VOP_RWLOCK 3048 * to avoid a deadlock with write requests. 3049 */ 3050 if (nbl_need_check(vp)) { 3051 nbl_start_crit(vp, RW_READER); 3052 in_crit = 1; 3053 if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0, 3054 &ct)) { 3055 *cs->statusp = resp->status = NFS4ERR_LOCKED; 3056 goto out; 3057 } 3058 } 3059 3060 if ((stat = rfs4_check_stateid(FREAD, vp, &args->stateid, FALSE, 3061 deleg, TRUE, &ct)) != NFS4_OK) { 3062 *cs->statusp = resp->status = stat; 3063 goto out; 3064 } 3065 3066 va.va_mask = AT_MODE|AT_SIZE|AT_UID; 3067 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3068 3069 /* 3070 * If we can't get the attributes, then we can't do the 3071 * right access checking. So, we'll fail the request. 3072 */ 3073 if (verror) { 3074 *cs->statusp = resp->status = puterrno4(verror); 3075 goto out; 3076 } 3077 3078 if (vp->v_type != VREG) { 3079 *cs->statusp = resp->status = 3080 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 3081 goto out; 3082 } 3083 3084 if (crgetuid(cs->cr) != va.va_uid && 3085 (error = VOP_ACCESS(vp, VREAD, 0, cs->cr, &ct)) && 3086 (error = VOP_ACCESS(vp, VEXEC, 0, cs->cr, &ct))) { 3087 *cs->statusp = resp->status = puterrno4(error); 3088 goto out; 3089 } 3090 3091 if (MANDLOCK(vp, va.va_mode)) { /* XXX - V4 supports mand locking */ 3092 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3093 goto out; 3094 } 3095 3096 offset = args->offset; 3097 if (offset >= va.va_size) { 3098 *cs->statusp = resp->status = NFS4_OK; 3099 resp->eof = TRUE; 3100 resp->data_len = 0; 3101 resp->data_val = NULL; 3102 resp->mblk = NULL; 3103 *cs->statusp = resp->status = NFS4_OK; 3104 goto out; 3105 } 3106 3107 if (args->count == 0) { 3108 *cs->statusp = resp->status = NFS4_OK; 3109 resp->eof = FALSE; 3110 resp->data_len = 0; 3111 resp->data_val = NULL; 3112 resp->mblk = NULL; 3113 goto out; 3114 } 3115 3116 /* 3117 * Do not allocate memory more than maximum allowed 3118 * transfer size 3119 */ 3120 if (args->count > rfs4_tsize(req)) 3121 args->count = rfs4_tsize(req); 3122 3123 /* 3124 * mp will contain the data to be sent out in the read reply. 3125 * It will be freed after the reply has been sent. 3126 * Let's roundup the data to a BYTES_PER_XDR_UNIT multiple, 3127 * so that the call to xdrmblk_putmblk() never fails. 3128 * If the first alloc of the requested size fails, then 3129 * decrease the size to something more reasonable and wait 3130 * for the allocation to occur. 3131 */ 3132 mp = allocb(RNDUP(args->count), BPRI_MED); 3133 if (mp == NULL) { 3134 if (args->count > MAXBSIZE) 3135 args->count = MAXBSIZE; 3136 mp = allocb_wait(RNDUP(args->count), BPRI_MED, 3137 STR_NOSIG, &alloc_err); 3138 } 3139 ASSERT(mp != NULL); 3140 ASSERT(alloc_err == 0); 3141 3142 iov.iov_base = (caddr_t)mp->b_datap->db_base; 3143 iov.iov_len = args->count; 3144 uio.uio_iov = &iov; 3145 uio.uio_iovcnt = 1; 3146 uio.uio_segflg = UIO_SYSSPACE; 3147 uio.uio_extflg = UIO_COPY_CACHED; 3148 uio.uio_loffset = args->offset; 3149 uio.uio_resid = args->count; 3150 3151 error = do_io(FREAD, vp, &uio, 0, cs->cr, &ct); 3152 3153 va.va_mask = AT_SIZE; 3154 verror = VOP_GETATTR(vp, &va, 0, cs->cr, &ct); 3155 3156 if (error) { 3157 freeb(mp); 3158 *cs->statusp = resp->status = puterrno4(error); 3159 goto out; 3160 } 3161 3162 *cs->statusp = resp->status = NFS4_OK; 3163 3164 ASSERT(uio.uio_resid >= 0); 3165 resp->data_len = args->count - uio.uio_resid; 3166 resp->data_val = (char *)mp->b_datap->db_base; 3167 resp->mblk = mp; 3168 3169 if (!verror && offset + resp->data_len == va.va_size) 3170 resp->eof = TRUE; 3171 else 3172 resp->eof = FALSE; 3173 3174 out: 3175 if (in_crit) 3176 nbl_end_crit(vp); 3177 3178 DTRACE_NFSV4_2(op__read__done, struct compound_state *, cs, 3179 READ4res *, resp); 3180 } 3181 3182 static void 3183 rfs4_op_read_free(nfs_resop4 *resop) 3184 { 3185 READ4res *resp = &resop->nfs_resop4_u.opread; 3186 3187 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3188 freeb(resp->mblk); 3189 resp->mblk = NULL; 3190 resp->data_val = NULL; 3191 resp->data_len = 0; 3192 } 3193 } 3194 3195 static void 3196 rfs4_op_readdir_free(nfs_resop4 *resop) 3197 { 3198 READDIR4res *resp = &resop->nfs_resop4_u.opreaddir; 3199 3200 if (resp->status == NFS4_OK && resp->mblk != NULL) { 3201 freeb(resp->mblk); 3202 resp->mblk = NULL; 3203 resp->data_len = 0; 3204 } 3205 } 3206 3207 3208 /* ARGSUSED */ 3209 static void 3210 rfs4_op_putpubfh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 3211 struct compound_state *cs) 3212 { 3213 PUTPUBFH4res *resp = &resop->nfs_resop4_u.opputpubfh; 3214 int error; 3215 vnode_t *vp; 3216 struct exportinfo *exi, *sav_exi; 3217 nfs_fh4_fmt_t *fh_fmtp; 3218 3219 DTRACE_NFSV4_1(op__putpubfh__start, struct compound_state *, cs); 3220 3221 if (cs->vp) { 3222 VN_RELE(cs->vp); 3223 cs->vp = NULL; 3224 } 3225 3226 if (cs->cr) 3227 crfree(cs->cr); 3228 3229 cs->cr = crdup(cs->basecr); 3230 3231 vp = exi_public->exi_vp; 3232 if (vp == NULL) { 3233 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3234 goto out; 3235 } 3236 3237 if (is_system_labeled()) { 3238 bslabel_t *clabel; 3239 3240 ASSERT(req->rq_label != NULL); 3241 clabel = req->rq_label; 3242 DTRACE_PROBE2(tx__rfs4__log__info__opputpubfh__clabel, char *, 3243 "got client label from request(1)", 3244 struct svc_req *, req); 3245 if (!blequal(&l_admin_low->tsl_label, clabel)) { 3246 if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) { 3247 *cs->statusp = resp->status = 3248 NFS4ERR_SERVERFAULT; 3249 return; 3250 } 3251 } 3252 } 3253 3254 error = makefh4(&cs->fh, vp, exi_public); 3255 if (error != 0) { 3256 *cs->statusp = resp->status = puterrno4(error); 3257 goto out; 3258 } 3259 sav_exi = cs->exi; 3260 if (exi_public == exi_root) { 3261 /* 3262 * No filesystem is actually shared public, so we default 3263 * to exi_root. In this case, we must check whether root 3264 * is exported. 3265 */ 3266 fh_fmtp = (nfs_fh4_fmt_t *)cs->fh.nfs_fh4_val; 3267 3268 /* 3269 * if root filesystem is exported, the exportinfo struct that we 3270 * should use is what checkexport4 returns, because root_exi is 3271 * actually a mostly empty struct. 3272 */ 3273 exi = checkexport4(&fh_fmtp->fh4_fsid, 3274 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 3275 cs->exi = ((exi != NULL) ? exi : exi_public); 3276 } else { 3277 /* 3278 * it's a properly shared filesystem 3279 */ 3280 cs->exi = exi_public; 3281 } 3282 3283 VN_HOLD(vp); 3284 cs->vp = vp; 3285 3286 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3287 VN_RELE(cs->vp); 3288 cs->vp = NULL; 3289 cs->exi = sav_exi; 3290 goto out; 3291 } 3292 3293 *cs->statusp = resp->status = NFS4_OK; 3294 out: 3295 DTRACE_NFSV4_2(op__putpubfh__done, struct compound_state *, cs, 3296 PUTPUBFH4res *, resp); 3297 } 3298 3299 /* 3300 * XXX - issue with put*fh operations. Suppose /export/home is exported. 3301 * Suppose an NFS client goes to mount /export/home/joe. If /export, home, 3302 * or joe have restrictive search permissions, then we shouldn't let 3303 * the client get a file handle. This is easy to enforce. However, we 3304 * don't know what security flavor should be used until we resolve the 3305 * path name. Another complication is uid mapping. If root is 3306 * the user, then it will be mapped to the anonymous user by default, 3307 * but we won't know that till we've resolved the path name. And we won't 3308 * know what the anonymous user is. 3309 * Luckily, SECINFO is specified to take a full filename. 3310 * So what we will have to in rfs4_op_lookup is check that flavor of 3311 * the target object matches that of the request, and if root was the 3312 * caller, check for the root= and anon= options, and if necessary, 3313 * repeat the lookup using the right cred_t. But that's not done yet. 3314 */ 3315 /* ARGSUSED */ 3316 static void 3317 rfs4_op_putfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3318 struct compound_state *cs) 3319 { 3320 PUTFH4args *args = &argop->nfs_argop4_u.opputfh; 3321 PUTFH4res *resp = &resop->nfs_resop4_u.opputfh; 3322 nfs_fh4_fmt_t *fh_fmtp; 3323 3324 DTRACE_NFSV4_2(op__putfh__start, struct compound_state *, cs, 3325 PUTFH4args *, args); 3326 3327 if (cs->vp) { 3328 VN_RELE(cs->vp); 3329 cs->vp = NULL; 3330 } 3331 3332 if (cs->cr) { 3333 crfree(cs->cr); 3334 cs->cr = NULL; 3335 } 3336 3337 3338 if (args->object.nfs_fh4_len < NFS_FH4_LEN) { 3339 *cs->statusp = resp->status = NFS4ERR_BADHANDLE; 3340 goto out; 3341 } 3342 3343 fh_fmtp = (nfs_fh4_fmt_t *)args->object.nfs_fh4_val; 3344 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, (fid_t *)&fh_fmtp->fh4_xlen, 3345 NULL); 3346 3347 if (cs->exi == NULL) { 3348 *cs->statusp = resp->status = NFS4ERR_STALE; 3349 goto out; 3350 } 3351 3352 cs->cr = crdup(cs->basecr); 3353 3354 ASSERT(cs->cr != NULL); 3355 3356 if (! (cs->vp = nfs4_fhtovp(&args->object, cs->exi, &resp->status))) { 3357 *cs->statusp = resp->status; 3358 goto out; 3359 } 3360 3361 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3362 VN_RELE(cs->vp); 3363 cs->vp = NULL; 3364 goto out; 3365 } 3366 3367 nfs_fh4_copy(&args->object, &cs->fh); 3368 *cs->statusp = resp->status = NFS4_OK; 3369 cs->deleg = FALSE; 3370 3371 out: 3372 DTRACE_NFSV4_2(op__putfh__done, struct compound_state *, cs, 3373 PUTFH4res *, resp); 3374 } 3375 3376 /* ARGSUSED */ 3377 static void 3378 rfs4_op_putrootfh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3379 struct compound_state *cs) 3380 3381 { 3382 PUTROOTFH4res *resp = &resop->nfs_resop4_u.opputrootfh; 3383 int error; 3384 fid_t fid; 3385 struct exportinfo *exi, *sav_exi; 3386 3387 DTRACE_NFSV4_1(op__putrootfh__start, struct compound_state *, cs); 3388 3389 if (cs->vp) { 3390 VN_RELE(cs->vp); 3391 cs->vp = NULL; 3392 } 3393 3394 if (cs->cr) 3395 crfree(cs->cr); 3396 3397 cs->cr = crdup(cs->basecr); 3398 3399 /* 3400 * Using rootdir, the system root vnode, 3401 * get its fid. 3402 */ 3403 bzero(&fid, sizeof (fid)); 3404 fid.fid_len = MAXFIDSZ; 3405 error = vop_fid_pseudo(rootdir, &fid); 3406 if (error != 0) { 3407 *cs->statusp = resp->status = puterrno4(error); 3408 goto out; 3409 } 3410 3411 /* 3412 * Then use the root fsid & fid it to find out if it's exported 3413 * 3414 * If the server root isn't exported directly, then 3415 * it should at least be a pseudo export based on 3416 * one or more exports further down in the server's 3417 * file tree. 3418 */ 3419 exi = checkexport4(&rootdir->v_vfsp->vfs_fsid, &fid, NULL); 3420 if (exi == NULL || exi->exi_export.ex_flags & EX_PUBLIC) { 3421 NFS4_DEBUG(rfs4_debug, 3422 (CE_WARN, "rfs4_op_putrootfh: export check failure")); 3423 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 3424 goto out; 3425 } 3426 3427 /* 3428 * Now make a filehandle based on the root 3429 * export and root vnode. 3430 */ 3431 error = makefh4(&cs->fh, rootdir, exi); 3432 if (error != 0) { 3433 *cs->statusp = resp->status = puterrno4(error); 3434 goto out; 3435 } 3436 3437 sav_exi = cs->exi; 3438 cs->exi = exi; 3439 3440 VN_HOLD(rootdir); 3441 cs->vp = rootdir; 3442 3443 if ((resp->status = call_checkauth4(cs, req)) != NFS4_OK) { 3444 VN_RELE(rootdir); 3445 cs->vp = NULL; 3446 cs->exi = sav_exi; 3447 goto out; 3448 } 3449 3450 *cs->statusp = resp->status = NFS4_OK; 3451 cs->deleg = FALSE; 3452 out: 3453 DTRACE_NFSV4_2(op__putrootfh__done, struct compound_state *, cs, 3454 PUTROOTFH4res *, resp); 3455 } 3456 3457 /* 3458 * A directory entry is a valid nfsv4 entry if 3459 * - it has a non-zero ino 3460 * - it is not a dot or dotdot name 3461 * - it is visible in a pseudo export or in a real export that can 3462 * only have a limited view. 3463 */ 3464 static bool_t 3465 valid_nfs4_entry(struct exportinfo *exi, struct dirent64 *dp, 3466 int *expseudo, int check_visible) 3467 { 3468 if (dp->d_ino == 0 || NFS_IS_DOTNAME(dp->d_name)) { 3469 *expseudo = 0; 3470 return (FALSE); 3471 } 3472 3473 if (! check_visible) { 3474 *expseudo = 0; 3475 return (TRUE); 3476 } 3477 3478 return (nfs_visible_inode(exi, dp->d_ino, expseudo)); 3479 } 3480 3481 /* 3482 * set_rdattr_params sets up the variables used to manage what information 3483 * to get for each directory entry. 3484 */ 3485 static nfsstat4 3486 set_rdattr_params(struct nfs4_svgetit_arg *sargp, 3487 bitmap4 attrs, bool_t *need_to_lookup) 3488 { 3489 uint_t va_mask; 3490 nfsstat4 status; 3491 bitmap4 objbits; 3492 3493 status = bitmap4_to_attrmask(attrs, sargp); 3494 if (status != NFS4_OK) { 3495 /* 3496 * could not even figure attr mask 3497 */ 3498 return (status); 3499 } 3500 va_mask = sargp->vap->va_mask; 3501 3502 /* 3503 * dirent's d_ino is always correct value for mounted_on_fileid. 3504 * mntdfid_set is set once here, but mounted_on_fileid is 3505 * set in main dirent processing loop for each dirent. 3506 * The mntdfid_set is a simple optimization that lets the 3507 * server attr code avoid work when caller is readdir. 3508 */ 3509 sargp->mntdfid_set = TRUE; 3510 3511 /* 3512 * Lookup entry only if client asked for any of the following: 3513 * a) vattr attrs 3514 * b) vfs attrs 3515 * c) attrs w/per-object scope requested (change, filehandle, etc) 3516 * other than mounted_on_fileid (which we can take from dirent) 3517 */ 3518 objbits = attrs ? attrs & NFS4_VP_ATTR_MASK : 0; 3519 3520 if (va_mask || sargp->sbp || (objbits & ~FATTR4_MOUNTED_ON_FILEID_MASK)) 3521 *need_to_lookup = TRUE; 3522 else 3523 *need_to_lookup = FALSE; 3524 3525 if (sargp->sbp == NULL) 3526 return (NFS4_OK); 3527 3528 /* 3529 * If filesystem attrs are requested, get them now from the 3530 * directory vp, as most entries will have same filesystem. The only 3531 * exception are mounted over entries but we handle 3532 * those as we go (XXX mounted over detection not yet implemented). 3533 */ 3534 sargp->vap->va_mask = 0; /* to avoid VOP_GETATTR */ 3535 status = bitmap4_get_sysattrs(sargp); 3536 sargp->vap->va_mask = va_mask; 3537 3538 if ((status != NFS4_OK) && sargp->rdattr_error_req) { 3539 /* 3540 * Failed to get filesystem attributes. 3541 * Return a rdattr_error for each entry, but don't fail. 3542 * However, don't get any obj-dependent attrs. 3543 */ 3544 sargp->rdattr_error = status; /* for rdattr_error */ 3545 *need_to_lookup = FALSE; 3546 /* 3547 * At least get fileid for regular readdir output 3548 */ 3549 sargp->vap->va_mask &= AT_NODEID; 3550 status = NFS4_OK; 3551 } 3552 3553 return (status); 3554 } 3555 3556 /* 3557 * readlink: args: CURRENT_FH. 3558 * res: status. If success - CURRENT_FH unchanged, return linktext. 3559 */ 3560 3561 /* ARGSUSED */ 3562 static void 3563 rfs4_op_readlink(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3564 struct compound_state *cs) 3565 { 3566 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3567 int error; 3568 vnode_t *vp; 3569 struct iovec iov; 3570 struct vattr va; 3571 struct uio uio; 3572 char *data; 3573 3574 DTRACE_NFSV4_1(op__readlink__start, struct compound_state *, cs); 3575 3576 /* CURRENT_FH: directory */ 3577 vp = cs->vp; 3578 if (vp == NULL) { 3579 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3580 goto out; 3581 } 3582 3583 if (cs->access == CS_ACCESS_DENIED) { 3584 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3585 goto out; 3586 } 3587 3588 if (vp->v_type == VDIR) { 3589 *cs->statusp = resp->status = NFS4ERR_ISDIR; 3590 goto out; 3591 } 3592 3593 if (vp->v_type != VLNK) { 3594 *cs->statusp = resp->status = NFS4ERR_INVAL; 3595 goto out; 3596 } 3597 3598 va.va_mask = AT_MODE; 3599 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 3600 if (error) { 3601 *cs->statusp = resp->status = puterrno4(error); 3602 goto out; 3603 } 3604 3605 if (MANDLOCK(vp, va.va_mode)) { 3606 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3607 goto out; 3608 } 3609 3610 data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP); 3611 3612 iov.iov_base = data; 3613 iov.iov_len = MAXPATHLEN; 3614 uio.uio_iov = &iov; 3615 uio.uio_iovcnt = 1; 3616 uio.uio_segflg = UIO_SYSSPACE; 3617 uio.uio_extflg = UIO_COPY_CACHED; 3618 uio.uio_loffset = 0; 3619 uio.uio_resid = MAXPATHLEN; 3620 3621 error = VOP_READLINK(vp, &uio, cs->cr, NULL); 3622 3623 if (error) { 3624 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3625 *cs->statusp = resp->status = puterrno4(error); 3626 goto out; 3627 } 3628 3629 *(data + MAXPATHLEN - uio.uio_resid) = '\0'; 3630 3631 /* 3632 * treat link name as data 3633 */ 3634 (void) str_to_utf8(data, &resp->link); 3635 3636 kmem_free((caddr_t)data, (uint_t)MAXPATHLEN + 1); 3637 *cs->statusp = resp->status = NFS4_OK; 3638 3639 out: 3640 DTRACE_NFSV4_2(op__readlink__done, struct compound_state *, cs, 3641 READLINK4res *, resp); 3642 } 3643 3644 static void 3645 rfs4_op_readlink_free(nfs_resop4 *resop) 3646 { 3647 READLINK4res *resp = &resop->nfs_resop4_u.opreadlink; 3648 utf8string *symlink = &resp->link; 3649 3650 if (symlink->utf8string_val) { 3651 UTF8STRING_FREE(*symlink) 3652 } 3653 } 3654 3655 /* 3656 * release_lockowner: 3657 * Release any state associated with the supplied 3658 * lockowner. Note if any lo_state is holding locks we will not 3659 * rele that lo_state and thus the lockowner will not be destroyed. 3660 * A client using lock after the lock owner stateid has been released 3661 * will suffer the consequence of NFS4ERR_BAD_STATEID and would have 3662 * to reissue the lock with new_lock_owner set to TRUE. 3663 * args: lock_owner 3664 * res: status 3665 */ 3666 /* ARGSUSED */ 3667 static void 3668 rfs4_op_release_lockowner(nfs_argop4 *argop, nfs_resop4 *resop, 3669 struct svc_req *req, struct compound_state *cs) 3670 { 3671 RELEASE_LOCKOWNER4args *ap = &argop->nfs_argop4_u.oprelease_lockowner; 3672 RELEASE_LOCKOWNER4res *resp = &resop->nfs_resop4_u.oprelease_lockowner; 3673 rfs4_lockowner_t *lo; 3674 rfs4_openowner_t *oop; 3675 rfs4_state_t *sp; 3676 rfs4_lo_state_t *lsp; 3677 rfs4_client_t *cp; 3678 bool_t create = FALSE; 3679 locklist_t *llist; 3680 sysid_t sysid; 3681 3682 DTRACE_NFSV4_2(op__release__lockowner__start, struct compound_state *, 3683 cs, RELEASE_LOCKOWNER4args *, ap); 3684 3685 /* Make sure there is a clientid around for this request */ 3686 cp = rfs4_findclient_by_id(ap->lock_owner.clientid, FALSE); 3687 3688 if (cp == NULL) { 3689 *cs->statusp = resp->status = 3690 rfs4_check_clientid(&ap->lock_owner.clientid, 0); 3691 goto out; 3692 } 3693 rfs4_client_rele(cp); 3694 3695 lo = rfs4_findlockowner(&ap->lock_owner, &create); 3696 if (lo == NULL) { 3697 *cs->statusp = resp->status = NFS4_OK; 3698 goto out; 3699 } 3700 ASSERT(lo->client != NULL); 3701 3702 /* 3703 * Check for EXPIRED client. If so will reap state with in a lease 3704 * period or on next set_clientid_confirm step 3705 */ 3706 if (rfs4_lease_expired(lo->client)) { 3707 rfs4_lockowner_rele(lo); 3708 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 3709 goto out; 3710 } 3711 3712 /* 3713 * If no sysid has been assigned, then no locks exist; just return. 3714 */ 3715 rfs4_dbe_lock(lo->client->dbe); 3716 if (lo->client->sysidt == LM_NOSYSID) { 3717 rfs4_lockowner_rele(lo); 3718 rfs4_dbe_unlock(lo->client->dbe); 3719 goto out; 3720 } 3721 3722 sysid = lo->client->sysidt; 3723 rfs4_dbe_unlock(lo->client->dbe); 3724 3725 /* 3726 * Mark the lockowner invalid. 3727 */ 3728 rfs4_dbe_hide(lo->dbe); 3729 3730 /* 3731 * sysid-pid pair should now not be used since the lockowner is 3732 * invalid. If the client were to instantiate the lockowner again 3733 * it would be assigned a new pid. Thus we can get the list of 3734 * current locks. 3735 */ 3736 3737 llist = flk_get_active_locks(sysid, lo->pid); 3738 /* If we are still holding locks fail */ 3739 if (llist != NULL) { 3740 3741 *cs->statusp = resp->status = NFS4ERR_LOCKS_HELD; 3742 3743 flk_free_locklist(llist); 3744 /* 3745 * We need to unhide the lockowner so the client can 3746 * try it again. The bad thing here is if the client 3747 * has a logic error that took it here in the first place 3748 * he probably has lost accounting of the locks that it 3749 * is holding. So we may have dangling state until the 3750 * open owner state is reaped via close. One scenario 3751 * that could possibly occur is that the client has 3752 * sent the unlock request(s) in separate threads 3753 * and has not waited for the replies before sending the 3754 * RELEASE_LOCKOWNER request. Presumably, it would expect 3755 * and deal appropriately with NFS4ERR_LOCKS_HELD, by 3756 * reissuing the request. 3757 */ 3758 rfs4_dbe_unhide(lo->dbe); 3759 rfs4_lockowner_rele(lo); 3760 goto out; 3761 } 3762 3763 /* 3764 * For the corresponding client we need to check each open 3765 * owner for any opens that have lockowner state associated 3766 * with this lockowner. 3767 */ 3768 3769 rfs4_dbe_lock(lo->client->dbe); 3770 for (oop = lo->client->openownerlist.next->oop; oop != NULL; 3771 oop = oop->openownerlist.next->oop) { 3772 3773 rfs4_dbe_lock(oop->dbe); 3774 for (sp = oop->ownerstateids.next->sp; sp != NULL; 3775 sp = sp->ownerstateids.next->sp) { 3776 3777 rfs4_dbe_lock(sp->dbe); 3778 for (lsp = sp->lockownerlist.next->lsp; 3779 lsp != NULL; lsp = lsp->lockownerlist.next->lsp) { 3780 if (lsp->locker == lo) { 3781 rfs4_dbe_lock(lsp->dbe); 3782 rfs4_dbe_invalidate(lsp->dbe); 3783 rfs4_dbe_unlock(lsp->dbe); 3784 } 3785 } 3786 rfs4_dbe_unlock(sp->dbe); 3787 } 3788 rfs4_dbe_unlock(oop->dbe); 3789 } 3790 rfs4_dbe_unlock(lo->client->dbe); 3791 3792 rfs4_lockowner_rele(lo); 3793 3794 *cs->statusp = resp->status = NFS4_OK; 3795 3796 out: 3797 DTRACE_NFSV4_2(op__release__lockowner__done, struct compound_state *, 3798 cs, RELEASE_LOCKOWNER4res *, resp); 3799 } 3800 3801 /* 3802 * short utility function to lookup a file and recall the delegation 3803 */ 3804 static rfs4_file_t * 3805 rfs4_lookup_and_findfile(vnode_t *dvp, char *nm, vnode_t **vpp, 3806 int *lkup_error, cred_t *cr) 3807 { 3808 vnode_t *vp; 3809 rfs4_file_t *fp = NULL; 3810 bool_t fcreate = FALSE; 3811 int error; 3812 3813 if (vpp) 3814 *vpp = NULL; 3815 3816 if ((error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cr, NULL, NULL, 3817 NULL)) == 0) { 3818 if (vp->v_type == VREG) 3819 fp = rfs4_findfile(vp, NULL, &fcreate); 3820 if (vpp) 3821 *vpp = vp; 3822 else 3823 VN_RELE(vp); 3824 } 3825 3826 if (lkup_error) 3827 *lkup_error = error; 3828 3829 return (fp); 3830 } 3831 3832 /* 3833 * remove: args: CURRENT_FH: directory; name. 3834 * res: status. If success - CURRENT_FH unchanged, return change_info 3835 * for directory. 3836 */ 3837 /* ARGSUSED */ 3838 static void 3839 rfs4_op_remove(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 3840 struct compound_state *cs) 3841 { 3842 REMOVE4args *args = &argop->nfs_argop4_u.opremove; 3843 REMOVE4res *resp = &resop->nfs_resop4_u.opremove; 3844 int error; 3845 vnode_t *dvp, *vp; 3846 struct vattr bdva, idva, adva; 3847 char *nm; 3848 uint_t len; 3849 rfs4_file_t *fp; 3850 int in_crit = 0; 3851 bslabel_t *clabel; 3852 3853 DTRACE_NFSV4_2(op__remove__start, struct compound_state *, cs, 3854 REMOVE4args *, args); 3855 3856 /* CURRENT_FH: directory */ 3857 dvp = cs->vp; 3858 if (dvp == NULL) { 3859 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 3860 goto out; 3861 } 3862 3863 if (cs->access == CS_ACCESS_DENIED) { 3864 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3865 goto out; 3866 } 3867 3868 /* 3869 * If there is an unshared filesystem mounted on this vnode, 3870 * Do not allow to remove anything in this directory. 3871 */ 3872 if (vn_ismntpt(dvp)) { 3873 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3874 goto out; 3875 } 3876 3877 if (dvp->v_type != VDIR) { 3878 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 3879 goto out; 3880 } 3881 3882 if (!utf8_dir_verify(&args->target)) { 3883 *cs->statusp = resp->status = NFS4ERR_INVAL; 3884 goto out; 3885 } 3886 3887 /* 3888 * Lookup the file so that we can check if it's a directory 3889 */ 3890 nm = utf8_to_fn(&args->target, &len, NULL); 3891 if (nm == NULL) { 3892 *cs->statusp = resp->status = NFS4ERR_INVAL; 3893 goto out; 3894 } 3895 3896 if (len > MAXNAMELEN) { 3897 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 3898 kmem_free(nm, len); 3899 goto out; 3900 } 3901 3902 if (rdonly4(cs->exi, cs->vp, req)) { 3903 *cs->statusp = resp->status = NFS4ERR_ROFS; 3904 kmem_free(nm, len); 3905 goto out; 3906 } 3907 3908 /* 3909 * Lookup the file to determine type and while we are see if 3910 * there is a file struct around and check for delegation. 3911 * We don't need to acquire va_seq before this lookup, if 3912 * it causes an update, cinfo.before will not match, which will 3913 * trigger a cache flush even if atomic is TRUE. 3914 */ 3915 if (fp = rfs4_lookup_and_findfile(dvp, nm, &vp, &error, cs->cr)) { 3916 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 3917 NULL)) { 3918 VN_RELE(vp); 3919 rfs4_file_rele(fp); 3920 *cs->statusp = resp->status = NFS4ERR_DELAY; 3921 kmem_free(nm, len); 3922 goto out; 3923 } 3924 } 3925 3926 /* Didn't find anything to remove */ 3927 if (vp == NULL) { 3928 *cs->statusp = resp->status = error; 3929 kmem_free(nm, len); 3930 goto out; 3931 } 3932 3933 if (nbl_need_check(vp)) { 3934 nbl_start_crit(vp, RW_READER); 3935 in_crit = 1; 3936 if (nbl_conflict(vp, NBL_REMOVE, 0, 0, 0, NULL)) { 3937 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 3938 kmem_free(nm, len); 3939 nbl_end_crit(vp); 3940 VN_RELE(vp); 3941 if (fp) { 3942 rfs4_clear_dont_grant(fp); 3943 rfs4_file_rele(fp); 3944 } 3945 goto out; 3946 } 3947 } 3948 3949 /* check label before allowing removal */ 3950 if (is_system_labeled()) { 3951 ASSERT(req->rq_label != NULL); 3952 clabel = req->rq_label; 3953 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 3954 "got client label from request(1)", 3955 struct svc_req *, req); 3956 if (!blequal(&l_admin_low->tsl_label, clabel)) { 3957 if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK)) { 3958 *cs->statusp = resp->status = NFS4ERR_ACCESS; 3959 kmem_free(nm, len); 3960 if (in_crit) 3961 nbl_end_crit(vp); 3962 VN_RELE(vp); 3963 if (fp) { 3964 rfs4_clear_dont_grant(fp); 3965 rfs4_file_rele(fp); 3966 } 3967 goto out; 3968 } 3969 } 3970 } 3971 3972 /* Get dir "before" change value */ 3973 bdva.va_mask = AT_CTIME|AT_SEQ; 3974 error = VOP_GETATTR(dvp, &bdva, 0, cs->cr, NULL); 3975 if (error) { 3976 *cs->statusp = resp->status = puterrno4(error); 3977 kmem_free(nm, len); 3978 goto out; 3979 } 3980 NFS4_SET_FATTR4_CHANGE(resp->cinfo.before, bdva.va_ctime) 3981 3982 /* Actually do the REMOVE operation */ 3983 if (vp->v_type == VDIR) { 3984 /* 3985 * Can't remove a directory that has a mounted-on filesystem. 3986 */ 3987 if (vn_ismntpt(vp)) { 3988 error = EACCES; 3989 } else { 3990 /* 3991 * System V defines rmdir to return EEXIST, 3992 * not * ENOTEMPTY, if the directory is not 3993 * empty. A System V NFS server needs to map 3994 * NFS4ERR_EXIST to NFS4ERR_NOTEMPTY to 3995 * transmit over the wire. 3996 */ 3997 if ((error = VOP_RMDIR(dvp, nm, rootdir, cs->cr, 3998 NULL, 0)) == EEXIST) 3999 error = ENOTEMPTY; 4000 } 4001 } else { 4002 if ((error = VOP_REMOVE(dvp, nm, cs->cr, NULL, 0)) == 0 && 4003 fp != NULL) { 4004 struct vattr va; 4005 vnode_t *tvp; 4006 4007 rfs4_dbe_lock(fp->dbe); 4008 tvp = fp->vp; 4009 if (tvp) 4010 VN_HOLD(tvp); 4011 rfs4_dbe_unlock(fp->dbe); 4012 4013 if (tvp) { 4014 /* 4015 * This is va_seq safe because we are not 4016 * manipulating dvp. 4017 */ 4018 va.va_mask = AT_NLINK; 4019 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4020 va.va_nlink == 0) { 4021 /* Remove state on file remove */ 4022 if (in_crit) { 4023 nbl_end_crit(vp); 4024 in_crit = 0; 4025 } 4026 rfs4_close_all_state(fp); 4027 } 4028 VN_RELE(tvp); 4029 } 4030 } 4031 } 4032 4033 if (in_crit) 4034 nbl_end_crit(vp); 4035 VN_RELE(vp); 4036 4037 if (fp) { 4038 rfs4_clear_dont_grant(fp); 4039 rfs4_file_rele(fp); 4040 } 4041 kmem_free(nm, len); 4042 4043 if (error) { 4044 *cs->statusp = resp->status = puterrno4(error); 4045 goto out; 4046 } 4047 4048 /* 4049 * Get the initial "after" sequence number, if it fails, set to zero 4050 */ 4051 idva.va_mask = AT_SEQ; 4052 if (VOP_GETATTR(dvp, &idva, 0, cs->cr, NULL)) 4053 idva.va_seq = 0; 4054 4055 /* 4056 * Force modified data and metadata out to stable storage. 4057 */ 4058 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 4059 4060 /* 4061 * Get "after" change value, if it fails, simply return the 4062 * before value. 4063 */ 4064 adva.va_mask = AT_CTIME|AT_SEQ; 4065 if (VOP_GETATTR(dvp, &adva, 0, cs->cr, NULL)) { 4066 adva.va_ctime = bdva.va_ctime; 4067 adva.va_seq = 0; 4068 } 4069 4070 NFS4_SET_FATTR4_CHANGE(resp->cinfo.after, adva.va_ctime) 4071 4072 /* 4073 * The cinfo.atomic = TRUE only if we have 4074 * non-zero va_seq's, and it has incremented by exactly one 4075 * during the VOP_REMOVE/RMDIR and it didn't change during 4076 * the VOP_FSYNC. 4077 */ 4078 if (bdva.va_seq && idva.va_seq && adva.va_seq && 4079 idva.va_seq == (bdva.va_seq + 1) && idva.va_seq == adva.va_seq) 4080 resp->cinfo.atomic = TRUE; 4081 else 4082 resp->cinfo.atomic = FALSE; 4083 4084 *cs->statusp = resp->status = NFS4_OK; 4085 4086 out: 4087 DTRACE_NFSV4_2(op__remove__done, struct compound_state *, cs, 4088 REMOVE4res *, resp); 4089 } 4090 4091 /* 4092 * rename: args: SAVED_FH: from directory, CURRENT_FH: target directory, 4093 * oldname and newname. 4094 * res: status. If success - CURRENT_FH unchanged, return change_info 4095 * for both from and target directories. 4096 */ 4097 /* ARGSUSED */ 4098 static void 4099 rfs4_op_rename(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4100 struct compound_state *cs) 4101 { 4102 RENAME4args *args = &argop->nfs_argop4_u.oprename; 4103 RENAME4res *resp = &resop->nfs_resop4_u.oprename; 4104 int error; 4105 vnode_t *odvp; 4106 vnode_t *ndvp; 4107 vnode_t *srcvp, *targvp; 4108 struct vattr obdva, oidva, oadva; 4109 struct vattr nbdva, nidva, nadva; 4110 char *onm, *nnm; 4111 uint_t olen, nlen; 4112 rfs4_file_t *fp, *sfp; 4113 int in_crit_src, in_crit_targ; 4114 int fp_rele_grant_hold, sfp_rele_grant_hold; 4115 bslabel_t *clabel; 4116 4117 DTRACE_NFSV4_2(op__rename__start, struct compound_state *, cs, 4118 RENAME4args *, args); 4119 4120 fp = sfp = NULL; 4121 srcvp = targvp = NULL; 4122 in_crit_src = in_crit_targ = 0; 4123 fp_rele_grant_hold = sfp_rele_grant_hold = 0; 4124 4125 /* CURRENT_FH: target directory */ 4126 ndvp = cs->vp; 4127 if (ndvp == NULL) { 4128 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4129 goto out; 4130 } 4131 4132 /* SAVED_FH: from directory */ 4133 odvp = cs->saved_vp; 4134 if (odvp == NULL) { 4135 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4136 goto out; 4137 } 4138 4139 if (cs->access == CS_ACCESS_DENIED) { 4140 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4141 goto out; 4142 } 4143 4144 /* 4145 * If there is an unshared filesystem mounted on this vnode, 4146 * do not allow to rename objects in this directory. 4147 */ 4148 if (vn_ismntpt(odvp)) { 4149 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4150 goto out; 4151 } 4152 4153 /* 4154 * If there is an unshared filesystem mounted on this vnode, 4155 * do not allow to rename to this directory. 4156 */ 4157 if (vn_ismntpt(ndvp)) { 4158 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4159 goto out; 4160 } 4161 4162 if (odvp->v_type != VDIR || ndvp->v_type != VDIR) { 4163 *cs->statusp = resp->status = NFS4ERR_NOTDIR; 4164 goto out; 4165 } 4166 4167 if (cs->saved_exi != cs->exi) { 4168 *cs->statusp = resp->status = NFS4ERR_XDEV; 4169 goto out; 4170 } 4171 4172 if (!utf8_dir_verify(&args->oldname)) { 4173 *cs->statusp = resp->status = NFS4ERR_INVAL; 4174 goto out; 4175 } 4176 4177 if (!utf8_dir_verify(&args->newname)) { 4178 *cs->statusp = resp->status = NFS4ERR_INVAL; 4179 goto out; 4180 } 4181 4182 onm = utf8_to_fn(&args->oldname, &olen, NULL); 4183 if (onm == NULL) { 4184 *cs->statusp = resp->status = NFS4ERR_INVAL; 4185 goto out; 4186 } 4187 4188 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4189 if (nnm == NULL) { 4190 *cs->statusp = resp->status = NFS4ERR_INVAL; 4191 kmem_free(onm, olen); 4192 goto out; 4193 } 4194 4195 if (olen > MAXNAMELEN || nlen > MAXNAMELEN) { 4196 *cs->statusp = resp->status = NFS4ERR_NAMETOOLONG; 4197 kmem_free(onm, olen); 4198 kmem_free(nnm, nlen); 4199 goto out; 4200 } 4201 4202 4203 if (rdonly4(cs->exi, cs->vp, req)) { 4204 *cs->statusp = resp->status = NFS4ERR_ROFS; 4205 kmem_free(onm, olen); 4206 kmem_free(nnm, nlen); 4207 goto out; 4208 } 4209 4210 /* check label of the target dir */ 4211 if (is_system_labeled()) { 4212 ASSERT(req->rq_label != NULL); 4213 clabel = req->rq_label; 4214 DTRACE_PROBE2(tx__rfs4__log__info__oprename__clabel, char *, 4215 "got client label from request(1)", 4216 struct svc_req *, req); 4217 if (!blequal(&l_admin_low->tsl_label, clabel)) { 4218 if (!do_rfs_label_check(clabel, ndvp, 4219 EQUALITY_CHECK)) { 4220 *cs->statusp = resp->status = NFS4ERR_ACCESS; 4221 goto out; 4222 } 4223 } 4224 } 4225 4226 /* 4227 * Is the source a file and have a delegation? 4228 * We don't need to acquire va_seq before these lookups, if 4229 * it causes an update, cinfo.before will not match, which will 4230 * trigger a cache flush even if atomic is TRUE. 4231 */ 4232 if (sfp = rfs4_lookup_and_findfile(odvp, onm, &srcvp, &error, cs->cr)) { 4233 if (rfs4_check_delegated_byfp(FWRITE, sfp, TRUE, TRUE, TRUE, 4234 NULL)) { 4235 *cs->statusp = resp->status = NFS4ERR_DELAY; 4236 goto err_out; 4237 } 4238 } 4239 4240 if (srcvp == NULL) { 4241 *cs->statusp = resp->status = puterrno4(error); 4242 kmem_free(onm, olen); 4243 kmem_free(nnm, nlen); 4244 goto out; 4245 } 4246 4247 sfp_rele_grant_hold = 1; 4248 4249 /* Does the destination exist and a file and have a delegation? */ 4250 if (fp = rfs4_lookup_and_findfile(ndvp, nnm, &targvp, NULL, cs->cr)) { 4251 if (rfs4_check_delegated_byfp(FWRITE, fp, TRUE, TRUE, TRUE, 4252 NULL)) { 4253 *cs->statusp = resp->status = NFS4ERR_DELAY; 4254 goto err_out; 4255 } 4256 } 4257 fp_rele_grant_hold = 1; 4258 4259 4260 /* Check for NBMAND lock on both source and target */ 4261 if (nbl_need_check(srcvp)) { 4262 nbl_start_crit(srcvp, RW_READER); 4263 in_crit_src = 1; 4264 if (nbl_conflict(srcvp, NBL_RENAME, 0, 0, 0, NULL)) { 4265 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4266 goto err_out; 4267 } 4268 } 4269 4270 if (targvp && nbl_need_check(targvp)) { 4271 nbl_start_crit(targvp, RW_READER); 4272 in_crit_targ = 1; 4273 if (nbl_conflict(targvp, NBL_REMOVE, 0, 0, 0, NULL)) { 4274 *cs->statusp = resp->status = NFS4ERR_FILE_OPEN; 4275 goto err_out; 4276 } 4277 } 4278 4279 /* Get source "before" change value */ 4280 obdva.va_mask = AT_CTIME|AT_SEQ; 4281 error = VOP_GETATTR(odvp, &obdva, 0, cs->cr, NULL); 4282 if (!error) { 4283 nbdva.va_mask = AT_CTIME|AT_SEQ; 4284 error = VOP_GETATTR(ndvp, &nbdva, 0, cs->cr, NULL); 4285 } 4286 if (error) { 4287 *cs->statusp = resp->status = puterrno4(error); 4288 goto err_out; 4289 } 4290 4291 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.before, obdva.va_ctime) 4292 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.before, nbdva.va_ctime) 4293 4294 if ((error = VOP_RENAME(odvp, onm, ndvp, nnm, cs->cr, NULL, 0)) == 0 && 4295 fp != NULL) { 4296 struct vattr va; 4297 vnode_t *tvp; 4298 4299 rfs4_dbe_lock(fp->dbe); 4300 tvp = fp->vp; 4301 if (tvp) 4302 VN_HOLD(tvp); 4303 rfs4_dbe_unlock(fp->dbe); 4304 4305 if (tvp) { 4306 va.va_mask = AT_NLINK; 4307 if (!VOP_GETATTR(tvp, &va, 0, cs->cr, NULL) && 4308 va.va_nlink == 0) { 4309 /* The file is gone and so should the state */ 4310 if (in_crit_targ) { 4311 nbl_end_crit(targvp); 4312 in_crit_targ = 0; 4313 } 4314 rfs4_close_all_state(fp); 4315 } 4316 VN_RELE(tvp); 4317 } 4318 } 4319 if (error == 0) { 4320 char *tmp; 4321 4322 /* fix the path name for the renamed file */ 4323 mutex_enter(&srcvp->v_lock); 4324 tmp = srcvp->v_path; 4325 srcvp->v_path = NULL; 4326 mutex_exit(&srcvp->v_lock); 4327 vn_setpath(rootdir, ndvp, srcvp, nnm, nlen - 1); 4328 if (tmp != NULL) 4329 kmem_free(tmp, strlen(tmp) + 1); 4330 } 4331 4332 if (in_crit_src) 4333 nbl_end_crit(srcvp); 4334 if (srcvp) 4335 VN_RELE(srcvp); 4336 if (in_crit_targ) 4337 nbl_end_crit(targvp); 4338 if (targvp) 4339 VN_RELE(targvp); 4340 4341 if (sfp) { 4342 rfs4_clear_dont_grant(sfp); 4343 rfs4_file_rele(sfp); 4344 } 4345 if (fp) { 4346 rfs4_clear_dont_grant(fp); 4347 rfs4_file_rele(fp); 4348 } 4349 4350 kmem_free(onm, olen); 4351 kmem_free(nnm, nlen); 4352 4353 /* 4354 * Get the initial "after" sequence number, if it fails, set to zero 4355 */ 4356 oidva.va_mask = AT_SEQ; 4357 if (VOP_GETATTR(odvp, &oidva, 0, cs->cr, NULL)) 4358 oidva.va_seq = 0; 4359 4360 nidva.va_mask = AT_SEQ; 4361 if (VOP_GETATTR(ndvp, &nidva, 0, cs->cr, NULL)) 4362 nidva.va_seq = 0; 4363 4364 /* 4365 * Force modified data and metadata out to stable storage. 4366 */ 4367 (void) VOP_FSYNC(odvp, 0, cs->cr, NULL); 4368 (void) VOP_FSYNC(ndvp, 0, cs->cr, NULL); 4369 4370 if (error) { 4371 *cs->statusp = resp->status = puterrno4(error); 4372 goto out; 4373 } 4374 4375 /* 4376 * Get "after" change values, if it fails, simply return the 4377 * before value. 4378 */ 4379 oadva.va_mask = AT_CTIME|AT_SEQ; 4380 if (VOP_GETATTR(odvp, &oadva, 0, cs->cr, NULL)) { 4381 oadva.va_ctime = obdva.va_ctime; 4382 oadva.va_seq = 0; 4383 } 4384 4385 nadva.va_mask = AT_CTIME|AT_SEQ; 4386 if (VOP_GETATTR(odvp, &nadva, 0, cs->cr, NULL)) { 4387 nadva.va_ctime = nbdva.va_ctime; 4388 nadva.va_seq = 0; 4389 } 4390 4391 NFS4_SET_FATTR4_CHANGE(resp->source_cinfo.after, oadva.va_ctime) 4392 NFS4_SET_FATTR4_CHANGE(resp->target_cinfo.after, nadva.va_ctime) 4393 4394 /* 4395 * The cinfo.atomic = TRUE only if we have 4396 * non-zero va_seq's, and it has incremented by exactly one 4397 * during the VOP_RENAME and it didn't change during the VOP_FSYNC. 4398 */ 4399 if (obdva.va_seq && oidva.va_seq && oadva.va_seq && 4400 oidva.va_seq == (obdva.va_seq + 1) && oidva.va_seq == oadva.va_seq) 4401 resp->source_cinfo.atomic = TRUE; 4402 else 4403 resp->source_cinfo.atomic = FALSE; 4404 4405 if (nbdva.va_seq && nidva.va_seq && nadva.va_seq && 4406 nidva.va_seq == (nbdva.va_seq + 1) && nidva.va_seq == nadva.va_seq) 4407 resp->target_cinfo.atomic = TRUE; 4408 else 4409 resp->target_cinfo.atomic = FALSE; 4410 4411 #ifdef VOLATILE_FH_TEST 4412 { 4413 extern void add_volrnm_fh(struct exportinfo *, vnode_t *); 4414 4415 /* 4416 * Add the renamed file handle to the volatile rename list 4417 */ 4418 if (cs->exi->exi_export.ex_flags & EX_VOLRNM) { 4419 /* file handles may expire on rename */ 4420 vnode_t *vp; 4421 4422 nnm = utf8_to_fn(&args->newname, &nlen, NULL); 4423 /* 4424 * Already know that nnm will be a valid string 4425 */ 4426 error = VOP_LOOKUP(ndvp, nnm, &vp, NULL, 0, NULL, cs->cr, 4427 NULL, NULL, NULL); 4428 kmem_free(nnm, nlen); 4429 if (!error) { 4430 add_volrnm_fh(cs->exi, vp); 4431 VN_RELE(vp); 4432 } 4433 } 4434 } 4435 #endif /* VOLATILE_FH_TEST */ 4436 4437 *cs->statusp = resp->status = NFS4_OK; 4438 out: 4439 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4440 RENAME4res *, resp); 4441 return; 4442 4443 err_out: 4444 kmem_free(onm, olen); 4445 kmem_free(nnm, nlen); 4446 4447 if (in_crit_src) nbl_end_crit(srcvp); 4448 if (in_crit_targ) nbl_end_crit(targvp); 4449 if (targvp) VN_RELE(targvp); 4450 if (srcvp) VN_RELE(srcvp); 4451 if (sfp) { 4452 if (sfp_rele_grant_hold) rfs4_clear_dont_grant(sfp); 4453 rfs4_file_rele(sfp); 4454 } 4455 if (fp) { 4456 if (fp_rele_grant_hold) rfs4_clear_dont_grant(fp); 4457 rfs4_file_rele(fp); 4458 } 4459 4460 DTRACE_NFSV4_2(op__rename__done, struct compound_state *, cs, 4461 RENAME4res *, resp); 4462 } 4463 4464 /* ARGSUSED */ 4465 static void 4466 rfs4_op_renew(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4467 struct compound_state *cs) 4468 { 4469 RENEW4args *args = &argop->nfs_argop4_u.oprenew; 4470 RENEW4res *resp = &resop->nfs_resop4_u.oprenew; 4471 rfs4_client_t *cp; 4472 4473 DTRACE_NFSV4_2(op__renew__start, struct compound_state *, cs, 4474 RENEW4args *, args); 4475 4476 if ((cp = rfs4_findclient_by_id(args->clientid, FALSE)) == NULL) { 4477 *cs->statusp = resp->status = 4478 rfs4_check_clientid(&args->clientid, 0); 4479 goto out; 4480 } 4481 4482 if (rfs4_lease_expired(cp)) { 4483 rfs4_client_rele(cp); 4484 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 4485 goto out; 4486 } 4487 4488 rfs4_update_lease(cp); 4489 4490 mutex_enter(cp->cbinfo.cb_lock); 4491 if (cp->cbinfo.cb_notified_of_cb_path_down == FALSE) { 4492 cp->cbinfo.cb_notified_of_cb_path_down = TRUE; 4493 *cs->statusp = resp->status = NFS4ERR_CB_PATH_DOWN; 4494 } else { 4495 *cs->statusp = resp->status = NFS4_OK; 4496 } 4497 mutex_exit(cp->cbinfo.cb_lock); 4498 4499 rfs4_client_rele(cp); 4500 4501 out: 4502 DTRACE_NFSV4_2(op__renew__done, struct compound_state *, cs, 4503 RENEW4res *, resp); 4504 } 4505 4506 /* ARGSUSED */ 4507 static void 4508 rfs4_op_restorefh(nfs_argop4 *args, nfs_resop4 *resop, struct svc_req *req, 4509 struct compound_state *cs) 4510 { 4511 RESTOREFH4res *resp = &resop->nfs_resop4_u.oprestorefh; 4512 4513 DTRACE_NFSV4_1(op__restorefh__start, struct compound_state *, cs); 4514 4515 /* No need to check cs->access - we are not accessing any object */ 4516 if ((cs->saved_vp == NULL) || (cs->saved_fh.nfs_fh4_val == NULL)) { 4517 *cs->statusp = resp->status = NFS4ERR_RESTOREFH; 4518 goto out; 4519 } 4520 if (cs->vp != NULL) { 4521 VN_RELE(cs->vp); 4522 } 4523 cs->vp = cs->saved_vp; 4524 cs->saved_vp = NULL; 4525 cs->exi = cs->saved_exi; 4526 nfs_fh4_copy(&cs->saved_fh, &cs->fh); 4527 *cs->statusp = resp->status = NFS4_OK; 4528 cs->deleg = FALSE; 4529 4530 out: 4531 DTRACE_NFSV4_2(op__restorefh__done, struct compound_state *, cs, 4532 RESTOREFH4res *, resp); 4533 } 4534 4535 /* ARGSUSED */ 4536 static void 4537 rfs4_op_savefh(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 4538 struct compound_state *cs) 4539 { 4540 SAVEFH4res *resp = &resop->nfs_resop4_u.opsavefh; 4541 4542 DTRACE_NFSV4_1(op__savefh__start, struct compound_state *, cs); 4543 4544 /* No need to check cs->access - we are not accessing any object */ 4545 if (cs->vp == NULL) { 4546 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 4547 goto out; 4548 } 4549 if (cs->saved_vp != NULL) { 4550 VN_RELE(cs->saved_vp); 4551 } 4552 cs->saved_vp = cs->vp; 4553 VN_HOLD(cs->saved_vp); 4554 cs->saved_exi = cs->exi; 4555 /* 4556 * since SAVEFH is fairly rare, don't alloc space for its fh 4557 * unless necessary. 4558 */ 4559 if (cs->saved_fh.nfs_fh4_val == NULL) { 4560 cs->saved_fh.nfs_fh4_val = kmem_alloc(NFS4_FHSIZE, KM_SLEEP); 4561 } 4562 nfs_fh4_copy(&cs->fh, &cs->saved_fh); 4563 *cs->statusp = resp->status = NFS4_OK; 4564 4565 out: 4566 DTRACE_NFSV4_2(op__savefh__done, struct compound_state *, cs, 4567 SAVEFH4res *, resp); 4568 } 4569 4570 /* 4571 * rfs4_verify_attr is called when nfsv4 Setattr failed, but we wish to 4572 * return the bitmap of attrs that were set successfully. It is also 4573 * called by Verify/Nverify to test the vattr/vfsstat attrs. It should 4574 * always be called only after rfs4_do_set_attrs(). 4575 * 4576 * Verify that the attributes are same as the expected ones. sargp->vap 4577 * and sargp->sbp contain the input attributes as translated from fattr4. 4578 * 4579 * This function verifies only the attrs that correspond to a vattr or 4580 * vfsstat struct. That is because of the extra step needed to get the 4581 * corresponding system structs. Other attributes have already been set or 4582 * verified by do_rfs4_set_attrs. 4583 * 4584 * Return 0 if all attrs match, -1 if some don't, error if error processing. 4585 */ 4586 static int 4587 rfs4_verify_attr(struct nfs4_svgetit_arg *sargp, 4588 bitmap4 *resp, struct nfs4_ntov_table *ntovp) 4589 { 4590 int error, ret_error = 0; 4591 int i, k; 4592 uint_t sva_mask = sargp->vap->va_mask; 4593 uint_t vbit; 4594 union nfs4_attr_u *na; 4595 uint8_t *amap; 4596 bool_t getsb = ntovp->vfsstat; 4597 4598 if (sva_mask != 0) { 4599 /* 4600 * Okay to overwrite sargp->vap because we verify based 4601 * on the incoming values. 4602 */ 4603 ret_error = VOP_GETATTR(sargp->cs->vp, sargp->vap, 0, 4604 sargp->cs->cr, NULL); 4605 if (ret_error) { 4606 if (resp == NULL) 4607 return (ret_error); 4608 /* 4609 * Must return bitmap of successful attrs 4610 */ 4611 sva_mask = 0; /* to prevent checking vap later */ 4612 } else { 4613 /* 4614 * Some file systems clobber va_mask. it is probably 4615 * wrong of them to do so, nonethless we practice 4616 * defensive coding. 4617 * See bug id 4276830. 4618 */ 4619 sargp->vap->va_mask = sva_mask; 4620 } 4621 } 4622 4623 if (getsb) { 4624 /* 4625 * Now get the superblock and loop on the bitmap, as there is 4626 * no simple way of translating from superblock to bitmap4. 4627 */ 4628 ret_error = VFS_STATVFS(sargp->cs->vp->v_vfsp, sargp->sbp); 4629 if (ret_error) { 4630 if (resp == NULL) 4631 goto errout; 4632 getsb = FALSE; 4633 } 4634 } 4635 4636 /* 4637 * Now loop and verify each attribute which getattr returned 4638 * whether it's the same as the input. 4639 */ 4640 if (resp == NULL && !getsb && (sva_mask == 0)) 4641 goto errout; 4642 4643 na = ntovp->na; 4644 amap = ntovp->amap; 4645 k = 0; 4646 for (i = 0; i < ntovp->attrcnt; i++, na++, amap++) { 4647 k = *amap; 4648 ASSERT(nfs4_ntov_map[k].nval == k); 4649 vbit = nfs4_ntov_map[k].vbit; 4650 4651 /* 4652 * If vattr attribute but VOP_GETATTR failed, or it's 4653 * superblock attribute but VFS_STATVFS failed, skip 4654 */ 4655 if (vbit) { 4656 if ((vbit & sva_mask) == 0) 4657 continue; 4658 } else if (!(getsb && nfs4_ntov_map[k].vfsstat)) { 4659 continue; 4660 } 4661 error = (*nfs4_ntov_map[k].sv_getit)(NFS4ATTR_VERIT, sargp, na); 4662 if (resp != NULL) { 4663 if (error) 4664 ret_error = -1; /* not all match */ 4665 else /* update response bitmap */ 4666 *resp |= nfs4_ntov_map[k].fbit; 4667 continue; 4668 } 4669 if (error) { 4670 ret_error = -1; /* not all match */ 4671 break; 4672 } 4673 } 4674 errout: 4675 return (ret_error); 4676 } 4677 4678 /* 4679 * Decode the attribute to be set/verified. If the attr requires a sys op 4680 * (VOP_GETATTR, VFS_VFSSTAT), and the request is to verify, then don't 4681 * call the sv_getit function for it, because the sys op hasn't yet been done. 4682 * Return 0 for success, error code if failed. 4683 * 4684 * Note: the decoded arg is not freed here but in nfs4_ntov_table_free. 4685 */ 4686 static int 4687 decode_fattr4_attr(nfs4_attr_cmd_t cmd, struct nfs4_svgetit_arg *sargp, 4688 int k, XDR *xdrp, bitmap4 *resp_bval, union nfs4_attr_u *nap) 4689 { 4690 int error = 0; 4691 bool_t set_later; 4692 4693 sargp->vap->va_mask |= nfs4_ntov_map[k].vbit; 4694 4695 if ((*nfs4_ntov_map[k].xfunc)(xdrp, nap)) { 4696 set_later = nfs4_ntov_map[k].vbit || nfs4_ntov_map[k].vfsstat; 4697 /* 4698 * don't verify yet if a vattr or sb dependent attr, 4699 * because we don't have their sys values yet. 4700 * Will be done later. 4701 */ 4702 if (! (set_later && (cmd == NFS4ATTR_VERIT))) { 4703 /* 4704 * ACLs are a special case, since setting the MODE 4705 * conflicts with setting the ACL. We delay setting 4706 * the ACL until all other attributes have been set. 4707 * The ACL gets set in do_rfs4_op_setattr(). 4708 */ 4709 if (nfs4_ntov_map[k].fbit != FATTR4_ACL_MASK) { 4710 error = (*nfs4_ntov_map[k].sv_getit)(cmd, 4711 sargp, nap); 4712 if (error) { 4713 xdr_free(nfs4_ntov_map[k].xfunc, 4714 (caddr_t)nap); 4715 } 4716 } 4717 } 4718 } else { 4719 #ifdef DEBUG 4720 cmn_err(CE_NOTE, "decode_fattr4_attr: error " 4721 "decoding attribute %d\n", k); 4722 #endif 4723 error = EINVAL; 4724 } 4725 if (!error && resp_bval && !set_later) { 4726 *resp_bval |= nfs4_ntov_map[k].fbit; 4727 } 4728 4729 return (error); 4730 } 4731 4732 /* 4733 * Set vattr based on incoming fattr4 attrs - used by setattr. 4734 * Set response mask. Ignore any values that are not writable vattr attrs. 4735 */ 4736 static nfsstat4 4737 do_rfs4_set_attrs(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 4738 struct nfs4_svgetit_arg *sargp, struct nfs4_ntov_table *ntovp, 4739 nfs4_attr_cmd_t cmd) 4740 { 4741 int error = 0; 4742 int i; 4743 char *attrs = fattrp->attrlist4; 4744 uint32_t attrslen = fattrp->attrlist4_len; 4745 XDR xdr; 4746 nfsstat4 status = NFS4_OK; 4747 vnode_t *vp = cs->vp; 4748 union nfs4_attr_u *na; 4749 uint8_t *amap; 4750 4751 #ifndef lint 4752 /* 4753 * Make sure that maximum attribute number can be expressed as an 4754 * 8 bit quantity. 4755 */ 4756 ASSERT(NFS4_MAXNUM_ATTRS <= (UINT8_MAX + 1)); 4757 #endif 4758 4759 if (vp == NULL) { 4760 if (resp) 4761 *resp = 0; 4762 return (NFS4ERR_NOFILEHANDLE); 4763 } 4764 if (cs->access == CS_ACCESS_DENIED) { 4765 if (resp) 4766 *resp = 0; 4767 return (NFS4ERR_ACCESS); 4768 } 4769 4770 sargp->op = cmd; 4771 sargp->cs = cs; 4772 sargp->flag = 0; /* may be set later */ 4773 sargp->vap->va_mask = 0; 4774 sargp->rdattr_error = NFS4_OK; 4775 sargp->rdattr_error_req = FALSE; 4776 /* sargp->sbp is set by the caller */ 4777 4778 xdrmem_create(&xdr, attrs, attrslen, XDR_DECODE); 4779 4780 na = ntovp->na; 4781 amap = ntovp->amap; 4782 4783 /* 4784 * The following loop iterates on the nfs4_ntov_map checking 4785 * if the fbit is set in the requested bitmap. 4786 * If set then we process the arguments using the 4787 * rfs4_fattr4 conversion functions to populate the setattr 4788 * vattr and va_mask. Any settable attrs that are not using vattr 4789 * will be set in this loop. 4790 */ 4791 for (i = 0; i < nfs4_ntov_map_size; i++) { 4792 if (!(fattrp->attrmask & nfs4_ntov_map[i].fbit)) { 4793 continue; 4794 } 4795 /* 4796 * If setattr, must be a writable attr. 4797 * If verify/nverify, must be a readable attr. 4798 */ 4799 if ((error = (*nfs4_ntov_map[i].sv_getit)( 4800 NFS4ATTR_SUPPORTED, sargp, NULL)) != 0) { 4801 /* 4802 * Client tries to set/verify an 4803 * unsupported attribute, tries to set 4804 * a read only attr or verify a write 4805 * only one - error! 4806 */ 4807 break; 4808 } 4809 /* 4810 * Decode the attribute to set/verify 4811 */ 4812 error = decode_fattr4_attr(cmd, sargp, nfs4_ntov_map[i].nval, 4813 &xdr, resp ? resp : NULL, na); 4814 if (error) 4815 break; 4816 *amap++ = (uint8_t)nfs4_ntov_map[i].nval; 4817 na++; 4818 (ntovp->attrcnt)++; 4819 if (nfs4_ntov_map[i].vfsstat) 4820 ntovp->vfsstat = TRUE; 4821 } 4822 4823 if (error != 0) 4824 status = (error == ENOTSUP ? NFS4ERR_ATTRNOTSUPP : 4825 puterrno4(error)); 4826 /* xdrmem_destroy(&xdrs); */ /* NO-OP */ 4827 return (status); 4828 } 4829 4830 static nfsstat4 4831 do_rfs4_op_setattr(bitmap4 *resp, fattr4 *fattrp, struct compound_state *cs, 4832 stateid4 *stateid) 4833 { 4834 int error = 0; 4835 struct nfs4_svgetit_arg sarg; 4836 bool_t trunc; 4837 4838 nfsstat4 status = NFS4_OK; 4839 cred_t *cr = cs->cr; 4840 vnode_t *vp = cs->vp; 4841 struct nfs4_ntov_table ntov; 4842 struct statvfs64 sb; 4843 struct vattr bva; 4844 struct flock64 bf; 4845 int in_crit = 0; 4846 uint_t saved_mask = 0; 4847 caller_context_t ct; 4848 4849 *resp = 0; 4850 sarg.sbp = &sb; 4851 nfs4_ntov_table_init(&ntov); 4852 status = do_rfs4_set_attrs(resp, fattrp, cs, &sarg, &ntov, 4853 NFS4ATTR_SETIT); 4854 if (status != NFS4_OK) { 4855 /* 4856 * failed set attrs 4857 */ 4858 goto done; 4859 } 4860 if ((sarg.vap->va_mask == 0) && 4861 (! (fattrp->attrmask & FATTR4_ACL_MASK))) { 4862 /* 4863 * no further work to be done 4864 */ 4865 goto done; 4866 } 4867 4868 /* 4869 * If we got a request to set the ACL and the MODE, only 4870 * allow changing VSUID, VSGID, and VSVTX. Attempting 4871 * to change any other bits, along with setting an ACL, 4872 * gives NFS4ERR_INVAL. 4873 */ 4874 if ((fattrp->attrmask & FATTR4_ACL_MASK) && 4875 (fattrp->attrmask & FATTR4_MODE_MASK)) { 4876 vattr_t va; 4877 4878 va.va_mask = AT_MODE; 4879 error = VOP_GETATTR(vp, &va, 0, cs->cr, NULL); 4880 if (error) { 4881 status = puterrno4(error); 4882 goto done; 4883 } 4884 if ((sarg.vap->va_mode ^ va.va_mode) & 4885 ~(VSUID | VSGID | VSVTX)) { 4886 status = NFS4ERR_INVAL; 4887 goto done; 4888 } 4889 } 4890 4891 /* Check stateid only if size has been set */ 4892 if (sarg.vap->va_mask & AT_SIZE) { 4893 trunc = (sarg.vap->va_size == 0); 4894 status = rfs4_check_stateid(FWRITE, cs->vp, stateid, 4895 trunc, &cs->deleg, sarg.vap->va_mask & AT_SIZE, &ct); 4896 if (status != NFS4_OK) 4897 goto done; 4898 } else { 4899 ct.cc_sysid = 0; 4900 ct.cc_pid = 0; 4901 ct.cc_caller_id = nfs4_srv_caller_id; 4902 ct.cc_flags = CC_DONTBLOCK; 4903 } 4904 4905 /* XXX start of possible race with delegations */ 4906 4907 /* 4908 * We need to specially handle size changes because it is 4909 * possible for the client to create a file with read-only 4910 * modes, but with the file opened for writing. If the client 4911 * then tries to set the file size, e.g. ftruncate(3C), 4912 * fcntl(F_FREESP), the normal access checking done in 4913 * VOP_SETATTR would prevent the client from doing it even though 4914 * it should be allowed to do so. To get around this, we do the 4915 * access checking for ourselves and use VOP_SPACE which doesn't 4916 * do the access checking. 4917 * Also the client should not be allowed to change the file 4918 * size if there is a conflicting non-blocking mandatory lock in 4919 * the region of the change. 4920 */ 4921 if (vp->v_type == VREG && (sarg.vap->va_mask & AT_SIZE)) { 4922 u_offset_t offset; 4923 ssize_t length; 4924 4925 /* 4926 * ufs_setattr clears AT_SIZE from vap->va_mask, but 4927 * before returning, sarg.vap->va_mask is used to 4928 * generate the setattr reply bitmap. We also clear 4929 * AT_SIZE below before calling VOP_SPACE. For both 4930 * of these cases, the va_mask needs to be saved here 4931 * and restored after calling VOP_SETATTR. 4932 */ 4933 saved_mask = sarg.vap->va_mask; 4934 4935 /* 4936 * Check any possible conflict due to NBMAND locks. 4937 * Get into critical region before VOP_GETATTR, so the 4938 * size attribute is valid when checking conflicts. 4939 */ 4940 if (nbl_need_check(vp)) { 4941 nbl_start_crit(vp, RW_READER); 4942 in_crit = 1; 4943 } 4944 4945 bva.va_mask = AT_UID|AT_SIZE; 4946 if (error = VOP_GETATTR(vp, &bva, 0, cr, &ct)) { 4947 status = puterrno4(error); 4948 goto done; 4949 } 4950 4951 if (in_crit) { 4952 if (sarg.vap->va_size < bva.va_size) { 4953 offset = sarg.vap->va_size; 4954 length = bva.va_size - sarg.vap->va_size; 4955 } else { 4956 offset = bva.va_size; 4957 length = sarg.vap->va_size - bva.va_size; 4958 } 4959 if (nbl_conflict(vp, NBL_WRITE, offset, length, 0, 4960 &ct)) { 4961 status = NFS4ERR_LOCKED; 4962 goto done; 4963 } 4964 } 4965 4966 if (crgetuid(cr) == bva.va_uid) { 4967 sarg.vap->va_mask &= ~AT_SIZE; 4968 bf.l_type = F_WRLCK; 4969 bf.l_whence = 0; 4970 bf.l_start = (off64_t)sarg.vap->va_size; 4971 bf.l_len = 0; 4972 bf.l_sysid = 0; 4973 bf.l_pid = 0; 4974 error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE, 4975 (offset_t)sarg.vap->va_size, cr, &ct); 4976 } 4977 } 4978 4979 if (!error && sarg.vap->va_mask != 0) 4980 error = VOP_SETATTR(vp, sarg.vap, sarg.flag, cr, &ct); 4981 4982 /* restore va_mask -- ufs_setattr clears AT_SIZE */ 4983 if (saved_mask & AT_SIZE) 4984 sarg.vap->va_mask |= AT_SIZE; 4985 4986 /* 4987 * If an ACL was being set, it has been delayed until now, 4988 * in order to set the mode (via the VOP_SETATTR() above) first. 4989 */ 4990 if ((! error) && (fattrp->attrmask & FATTR4_ACL_MASK)) { 4991 int i; 4992 4993 for (i = 0; i < NFS4_MAXNUM_ATTRS; i++) 4994 if (ntov.amap[i] == FATTR4_ACL) 4995 break; 4996 if (i < NFS4_MAXNUM_ATTRS) { 4997 error = (*nfs4_ntov_map[FATTR4_ACL].sv_getit)( 4998 NFS4ATTR_SETIT, &sarg, &ntov.na[i]); 4999 if (error == 0) { 5000 *resp |= FATTR4_ACL_MASK; 5001 } else if (error == ENOTSUP) { 5002 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5003 status = NFS4ERR_ATTRNOTSUPP; 5004 goto done; 5005 } 5006 } else { 5007 NFS4_DEBUG(rfs4_debug, 5008 (CE_NOTE, "do_rfs4_op_setattr: " 5009 "unable to find ACL in fattr4")); 5010 error = EINVAL; 5011 } 5012 } 5013 5014 if (error) { 5015 /* check if a monitor detected a delegation conflict */ 5016 if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) 5017 status = NFS4ERR_DELAY; 5018 else 5019 status = puterrno4(error); 5020 5021 /* 5022 * Set the response bitmap when setattr failed. 5023 * If VOP_SETATTR partially succeeded, test by doing a 5024 * VOP_GETATTR on the object and comparing the data 5025 * to the setattr arguments. 5026 */ 5027 (void) rfs4_verify_attr(&sarg, resp, &ntov); 5028 } else { 5029 /* 5030 * Force modified metadata out to stable storage. 5031 */ 5032 (void) VOP_FSYNC(vp, FNODSYNC, cr, &ct); 5033 /* 5034 * Set response bitmap 5035 */ 5036 nfs4_vmask_to_nmask_set(sarg.vap->va_mask, resp); 5037 } 5038 5039 /* Return early and already have a NFSv4 error */ 5040 done: 5041 /* 5042 * Except for nfs4_vmask_to_nmask_set(), vattr --> fattr 5043 * conversion sets both readable and writeable NFS4 attrs 5044 * for AT_MTIME and AT_ATIME. The line below masks out 5045 * unrequested attrs from the setattr result bitmap. This 5046 * is placed after the done: label to catch the ATTRNOTSUP 5047 * case. 5048 */ 5049 *resp &= fattrp->attrmask; 5050 5051 if (in_crit) 5052 nbl_end_crit(vp); 5053 5054 nfs4_ntov_table_free(&ntov, &sarg); 5055 5056 return (status); 5057 } 5058 5059 /* ARGSUSED */ 5060 static void 5061 rfs4_op_setattr(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5062 struct compound_state *cs) 5063 { 5064 SETATTR4args *args = &argop->nfs_argop4_u.opsetattr; 5065 SETATTR4res *resp = &resop->nfs_resop4_u.opsetattr; 5066 bslabel_t *clabel; 5067 5068 DTRACE_NFSV4_2(op__setattr__start, struct compound_state *, cs, 5069 SETATTR4args *, args); 5070 5071 if (cs->vp == NULL) { 5072 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5073 goto out; 5074 } 5075 5076 /* 5077 * If there is an unshared filesystem mounted on this vnode, 5078 * do not allow to setattr on this vnode. 5079 */ 5080 if (vn_ismntpt(cs->vp)) { 5081 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5082 goto out; 5083 } 5084 5085 resp->attrsset = 0; 5086 5087 if (rdonly4(cs->exi, cs->vp, req)) { 5088 *cs->statusp = resp->status = NFS4ERR_ROFS; 5089 goto out; 5090 } 5091 5092 /* check label before setting attributes */ 5093 if (is_system_labeled()) { 5094 ASSERT(req->rq_label != NULL); 5095 clabel = req->rq_label; 5096 DTRACE_PROBE2(tx__rfs4__log__info__opsetattr__clabel, char *, 5097 "got client label from request(1)", 5098 struct svc_req *, req); 5099 if (!blequal(&l_admin_low->tsl_label, clabel)) { 5100 if (!do_rfs_label_check(clabel, cs->vp, 5101 EQUALITY_CHECK)) { 5102 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5103 goto out; 5104 } 5105 } 5106 } 5107 5108 *cs->statusp = resp->status = 5109 do_rfs4_op_setattr(&resp->attrsset, &args->obj_attributes, cs, 5110 &args->stateid); 5111 5112 out: 5113 DTRACE_NFSV4_2(op__setattr__done, struct compound_state *, cs, 5114 SETATTR4res *, resp); 5115 } 5116 5117 /* ARGSUSED */ 5118 static void 5119 rfs4_op_verify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5120 struct compound_state *cs) 5121 { 5122 /* 5123 * verify and nverify are exactly the same, except that nverify 5124 * succeeds when some argument changed, and verify succeeds when 5125 * when none changed. 5126 */ 5127 5128 VERIFY4args *args = &argop->nfs_argop4_u.opverify; 5129 VERIFY4res *resp = &resop->nfs_resop4_u.opverify; 5130 5131 int error; 5132 struct nfs4_svgetit_arg sarg; 5133 struct statvfs64 sb; 5134 struct nfs4_ntov_table ntov; 5135 5136 DTRACE_NFSV4_2(op__verify__start, struct compound_state *, cs, 5137 VERIFY4args *, args); 5138 5139 if (cs->vp == NULL) { 5140 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5141 goto out; 5142 } 5143 5144 sarg.sbp = &sb; 5145 nfs4_ntov_table_init(&ntov); 5146 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5147 &sarg, &ntov, NFS4ATTR_VERIT); 5148 if (resp->status != NFS4_OK) { 5149 /* 5150 * do_rfs4_set_attrs will try to verify systemwide attrs, 5151 * so could return -1 for "no match". 5152 */ 5153 if (resp->status == -1) 5154 resp->status = NFS4ERR_NOT_SAME; 5155 goto done; 5156 } 5157 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5158 switch (error) { 5159 case 0: 5160 resp->status = NFS4_OK; 5161 break; 5162 case -1: 5163 resp->status = NFS4ERR_NOT_SAME; 5164 break; 5165 default: 5166 resp->status = puterrno4(error); 5167 break; 5168 } 5169 done: 5170 *cs->statusp = resp->status; 5171 nfs4_ntov_table_free(&ntov, &sarg); 5172 out: 5173 DTRACE_NFSV4_2(op__verify__done, struct compound_state *, cs, 5174 VERIFY4res *, resp); 5175 } 5176 5177 /* ARGSUSED */ 5178 static void 5179 rfs4_op_nverify(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5180 struct compound_state *cs) 5181 { 5182 /* 5183 * verify and nverify are exactly the same, except that nverify 5184 * succeeds when some argument changed, and verify succeeds when 5185 * when none changed. 5186 */ 5187 5188 NVERIFY4args *args = &argop->nfs_argop4_u.opnverify; 5189 NVERIFY4res *resp = &resop->nfs_resop4_u.opnverify; 5190 5191 int error; 5192 struct nfs4_svgetit_arg sarg; 5193 struct statvfs64 sb; 5194 struct nfs4_ntov_table ntov; 5195 5196 DTRACE_NFSV4_2(op__nverify__start, struct compound_state *, cs, 5197 NVERIFY4args *, args); 5198 5199 if (cs->vp == NULL) { 5200 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5201 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5202 NVERIFY4res *, resp); 5203 return; 5204 } 5205 sarg.sbp = &sb; 5206 nfs4_ntov_table_init(&ntov); 5207 resp->status = do_rfs4_set_attrs(NULL, &args->obj_attributes, cs, 5208 &sarg, &ntov, NFS4ATTR_VERIT); 5209 if (resp->status != NFS4_OK) { 5210 /* 5211 * do_rfs4_set_attrs will try to verify systemwide attrs, 5212 * so could return -1 for "no match". 5213 */ 5214 if (resp->status == -1) 5215 resp->status = NFS4_OK; 5216 goto done; 5217 } 5218 error = rfs4_verify_attr(&sarg, NULL, &ntov); 5219 switch (error) { 5220 case 0: 5221 resp->status = NFS4ERR_SAME; 5222 break; 5223 case -1: 5224 resp->status = NFS4_OK; 5225 break; 5226 default: 5227 resp->status = puterrno4(error); 5228 break; 5229 } 5230 done: 5231 *cs->statusp = resp->status; 5232 nfs4_ntov_table_free(&ntov, &sarg); 5233 5234 DTRACE_NFSV4_2(op__nverify__done, struct compound_state *, cs, 5235 NVERIFY4res *, resp); 5236 } 5237 5238 /* 5239 * XXX - This should live in an NFS header file. 5240 */ 5241 #define MAX_IOVECS 12 5242 5243 /* ARGSUSED */ 5244 static void 5245 rfs4_op_write(nfs_argop4 *argop, nfs_resop4 *resop, struct svc_req *req, 5246 struct compound_state *cs) 5247 { 5248 WRITE4args *args = &argop->nfs_argop4_u.opwrite; 5249 WRITE4res *resp = &resop->nfs_resop4_u.opwrite; 5250 int error; 5251 vnode_t *vp; 5252 struct vattr bva; 5253 u_offset_t rlimit; 5254 struct uio uio; 5255 struct iovec iov[MAX_IOVECS]; 5256 struct iovec *iovp; 5257 int iovcnt; 5258 int ioflag; 5259 cred_t *savecred, *cr; 5260 bool_t *deleg = &cs->deleg; 5261 nfsstat4 stat; 5262 int in_crit = 0; 5263 caller_context_t ct; 5264 5265 DTRACE_NFSV4_2(op__write__start, struct compound_state *, cs, 5266 WRITE4args *, args); 5267 5268 vp = cs->vp; 5269 if (vp == NULL) { 5270 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 5271 goto out; 5272 } 5273 if (cs->access == CS_ACCESS_DENIED) { 5274 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5275 goto out; 5276 } 5277 5278 cr = cs->cr; 5279 5280 if ((stat = rfs4_check_stateid(FWRITE, vp, &args->stateid, FALSE, 5281 deleg, TRUE, &ct)) != NFS4_OK) { 5282 *cs->statusp = resp->status = stat; 5283 goto out; 5284 } 5285 5286 /* 5287 * We have to enter the critical region before calling VOP_RWLOCK 5288 * to avoid a deadlock with ufs. 5289 */ 5290 if (nbl_need_check(vp)) { 5291 nbl_start_crit(vp, RW_READER); 5292 in_crit = 1; 5293 if (nbl_conflict(vp, NBL_WRITE, 5294 args->offset, args->data_len, 0, &ct)) { 5295 *cs->statusp = resp->status = NFS4ERR_LOCKED; 5296 goto out; 5297 } 5298 } 5299 5300 bva.va_mask = AT_MODE | AT_UID; 5301 error = VOP_GETATTR(vp, &bva, 0, cr, &ct); 5302 5303 /* 5304 * If we can't get the attributes, then we can't do the 5305 * right access checking. So, we'll fail the request. 5306 */ 5307 if (error) { 5308 *cs->statusp = resp->status = puterrno4(error); 5309 goto out; 5310 } 5311 5312 if (rdonly4(cs->exi, cs->vp, req)) { 5313 *cs->statusp = resp->status = NFS4ERR_ROFS; 5314 goto out; 5315 } 5316 5317 if (vp->v_type != VREG) { 5318 *cs->statusp = resp->status = 5319 ((vp->v_type == VDIR) ? NFS4ERR_ISDIR : NFS4ERR_INVAL); 5320 goto out; 5321 } 5322 5323 if (crgetuid(cr) != bva.va_uid && 5324 (error = VOP_ACCESS(vp, VWRITE, 0, cr, &ct))) { 5325 *cs->statusp = resp->status = puterrno4(error); 5326 goto out; 5327 } 5328 5329 if (MANDLOCK(vp, bva.va_mode)) { 5330 *cs->statusp = resp->status = NFS4ERR_ACCESS; 5331 goto out; 5332 } 5333 5334 if (args->data_len == 0) { 5335 *cs->statusp = resp->status = NFS4_OK; 5336 resp->count = 0; 5337 resp->committed = args->stable; 5338 resp->writeverf = Write4verf; 5339 goto out; 5340 } 5341 5342 if (args->mblk != NULL) { 5343 mblk_t *m; 5344 uint_t bytes, round_len; 5345 5346 iovcnt = 0; 5347 bytes = 0; 5348 round_len = roundup(args->data_len, BYTES_PER_XDR_UNIT); 5349 for (m = args->mblk; 5350 m != NULL && bytes < round_len; 5351 m = m->b_cont) { 5352 iovcnt++; 5353 bytes += MBLKL(m); 5354 } 5355 #ifdef DEBUG 5356 /* should have ended on an mblk boundary */ 5357 if (bytes != round_len) { 5358 printf("bytes=0x%x, round_len=0x%x, req len=0x%x\n", 5359 bytes, round_len, args->data_len); 5360 printf("args=%p, args->mblk=%p, m=%p", (void *)args, 5361 (void *)args->mblk, (void *)m); 5362 ASSERT(bytes == round_len); 5363 } 5364 #endif 5365 if (iovcnt <= MAX_IOVECS) { 5366 iovp = iov; 5367 } else { 5368 iovp = kmem_alloc(sizeof (*iovp) * iovcnt, KM_SLEEP); 5369 } 5370 mblk_to_iov(args->mblk, iovcnt, iovp); 5371 } else { 5372 iovcnt = 1; 5373 iovp = iov; 5374 iovp->iov_base = args->data_val; 5375 iovp->iov_len = args->data_len; 5376 } 5377 5378 uio.uio_iov = iovp; 5379 uio.uio_iovcnt = iovcnt; 5380 5381 uio.uio_segflg = UIO_SYSSPACE; 5382 uio.uio_extflg = UIO_COPY_DEFAULT; 5383 uio.uio_loffset = args->offset; 5384 uio.uio_resid = args->data_len; 5385 uio.uio_llimit = curproc->p_fsz_ctl; 5386 rlimit = uio.uio_llimit - args->offset; 5387 if (rlimit < (u_offset_t)uio.uio_resid) 5388 uio.uio_resid = (int)rlimit; 5389 5390 if (args->stable == UNSTABLE4) 5391 ioflag = 0; 5392 else if (args->stable == FILE_SYNC4) 5393 ioflag = FSYNC; 5394 else if (args->stable == DATA_SYNC4) 5395 ioflag = FDSYNC; 5396 else { 5397 if (iovp != iov) 5398 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5399 *cs->statusp = resp->status = NFS4ERR_INVAL; 5400 goto out; 5401 } 5402 5403 /* 5404 * We're changing creds because VM may fault and we need 5405 * the cred of the current thread to be used if quota 5406 * checking is enabled. 5407 */ 5408 savecred = curthread->t_cred; 5409 curthread->t_cred = cr; 5410 error = do_io(FWRITE, vp, &uio, ioflag, cr, &ct); 5411 curthread->t_cred = savecred; 5412 5413 if (iovp != iov) 5414 kmem_free(iovp, sizeof (*iovp) * iovcnt); 5415 5416 if (error) { 5417 *cs->statusp = resp->status = puterrno4(error); 5418 goto out; 5419 } 5420 5421 *cs->statusp = resp->status = NFS4_OK; 5422 resp->count = args->data_len - uio.uio_resid; 5423 5424 if (ioflag == 0) 5425 resp->committed = UNSTABLE4; 5426 else 5427 resp->committed = FILE_SYNC4; 5428 5429 resp->writeverf = Write4verf; 5430 5431 out: 5432 if (in_crit) 5433 nbl_end_crit(vp); 5434 5435 DTRACE_NFSV4_2(op__write__done, struct compound_state *, cs, 5436 WRITE4res *, resp); 5437 } 5438 5439 5440 /* XXX put in a header file */ 5441 extern int sec_svc_getcred(struct svc_req *, cred_t *, caddr_t *, int *); 5442 5443 void 5444 rfs4_compound(COMPOUND4args *args, COMPOUND4res *resp, struct exportinfo *exi, 5445 struct svc_req *req, cred_t *cr) 5446 { 5447 uint_t i; 5448 struct compound_state cs; 5449 5450 rfs4_init_compound_state(&cs); 5451 /* 5452 * Form a reply tag by copying over the reqeuest tag. 5453 */ 5454 resp->tag.utf8string_val = 5455 kmem_alloc(args->tag.utf8string_len, KM_SLEEP); 5456 resp->tag.utf8string_len = args->tag.utf8string_len; 5457 bcopy(args->tag.utf8string_val, resp->tag.utf8string_val, 5458 resp->tag.utf8string_len); 5459 5460 cs.statusp = &resp->status; 5461 cs.req = req; 5462 5463 /* 5464 * XXX for now, minorversion should be zero 5465 */ 5466 if (args->minorversion != NFS4_MINORVERSION) { 5467 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5468 &cs, COMPOUND4args *, args); 5469 resp->array_len = 0; 5470 resp->array = NULL; 5471 resp->status = NFS4ERR_MINOR_VERS_MISMATCH; 5472 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5473 &cs, COMPOUND4res *, resp); 5474 resp->array_len = 0; 5475 return; 5476 } 5477 5478 resp->array_len = args->array_len; 5479 resp->array = kmem_zalloc(args->array_len * sizeof (nfs_resop4), 5480 KM_SLEEP); 5481 5482 ASSERT(exi == NULL); 5483 ASSERT(cr == NULL); 5484 5485 cr = crget(); 5486 ASSERT(cr != NULL); 5487 5488 if (sec_svc_getcred(req, cr, &cs.principal, &cs.nfsflavor) == 0) { 5489 DTRACE_NFSV4_2(compound__start, struct compound_state *, 5490 &cs, COMPOUND4args *, args); 5491 crfree(cr); 5492 DTRACE_NFSV4_2(compound__done, struct compound_state *, 5493 &cs, COMPOUND4res *, resp); 5494 return; 5495 } 5496 5497 cs.basecr = cr; 5498 5499 DTRACE_NFSV4_2(compound__start, struct compound_state *, &cs, 5500 COMPOUND4args *, args); 5501 5502 /* 5503 * For now, NFS4 compound processing must be protected by 5504 * exported_lock because it can access more than one exportinfo 5505 * per compound and share/unshare can now change multiple 5506 * exinfo structs. The NFS2/3 code only refs 1 exportinfo 5507 * per proc (excluding public exinfo), and exi_count design 5508 * is sufficient to protect concurrent execution of NFS2/3 5509 * ops along with unexport. This lock will be removed as 5510 * part of the NFSv4 phase 2 namespace redesign work. 5511 */ 5512 rw_enter(&exported_lock, RW_READER); 5513 5514 /* 5515 * If this is the first compound we've seen, we need to start all 5516 * new instances' grace periods. 5517 */ 5518 if (rfs4_seen_first_compound == 0) { 5519 rfs4_grace_start_new(); 5520 /* 5521 * This must be set after rfs4_grace_start_new(), otherwise 5522 * another thread could proceed past here before the former 5523 * is finished. 5524 */ 5525 rfs4_seen_first_compound = 1; 5526 } 5527 5528 for (i = 0; i < args->array_len && cs.cont; i++) { 5529 nfs_argop4 *argop; 5530 nfs_resop4 *resop; 5531 uint_t op; 5532 5533 argop = &args->array[i]; 5534 resop = &resp->array[i]; 5535 resop->resop = argop->argop; 5536 op = (uint_t)resop->resop; 5537 5538 if (op < rfsv4disp_cnt) { 5539 /* 5540 * Count the individual ops here; NULL and COMPOUND 5541 * are counted in common_dispatch() 5542 */ 5543 rfsproccnt_v4_ptr[op].value.ui64++; 5544 5545 NFS4_DEBUG(rfs4_debug > 1, 5546 (CE_NOTE, "Executing %s", rfs4_op_string[op])); 5547 (*rfsv4disptab[op].dis_proc)(argop, resop, req, &cs); 5548 NFS4_DEBUG(rfs4_debug > 1, (CE_NOTE, "%s returned %d", 5549 rfs4_op_string[op], *cs.statusp)); 5550 if (*cs.statusp != NFS4_OK) 5551 cs.cont = FALSE; 5552 } else { 5553 /* 5554 * This is effectively dead code since XDR code 5555 * will have already returned BADXDR if op doesn't 5556 * decode to legal value. This only done for a 5557 * day when XDR code doesn't verify v4 opcodes. 5558 */ 5559 op = OP_ILLEGAL; 5560 rfsproccnt_v4_ptr[OP_ILLEGAL_IDX].value.ui64++; 5561 5562 rfs4_op_illegal(argop, resop, req, &cs); 5563 cs.cont = FALSE; 5564 } 5565 5566 /* 5567 * If not at last op, and if we are to stop, then 5568 * compact the results array. 5569 */ 5570 if ((i + 1) < args->array_len && !cs.cont) { 5571 nfs_resop4 *new_res = kmem_alloc( 5572 (i+1) * sizeof (nfs_resop4), KM_SLEEP); 5573 bcopy(resp->array, 5574 new_res, (i+1) * sizeof (nfs_resop4)); 5575 kmem_free(resp->array, 5576 args->array_len * sizeof (nfs_resop4)); 5577 5578 resp->array_len = i + 1; 5579 resp->array = new_res; 5580 } 5581 } 5582 5583 rw_exit(&exported_lock); 5584 5585 DTRACE_NFSV4_2(compound__done, struct compound_state *, &cs, 5586 COMPOUND4res *, resp); 5587 5588 if (cs.vp) 5589 VN_RELE(cs.vp); 5590 if (cs.saved_vp) 5591 VN_RELE(cs.saved_vp); 5592 if (cs.saved_fh.nfs_fh4_val) 5593 kmem_free(cs.saved_fh.nfs_fh4_val, NFS4_FHSIZE); 5594 5595 if (cs.basecr) 5596 crfree(cs.basecr); 5597 if (cs.cr) 5598 crfree(cs.cr); 5599 /* 5600 * done with this compound request, free the label 5601 */ 5602 5603 if (req->rq_label != NULL) { 5604 kmem_free(req->rq_label, sizeof (bslabel_t)); 5605 req->rq_label = NULL; 5606 } 5607 } 5608 5609 /* 5610 * XXX because of what appears to be duplicate calls to rfs4_compound_free 5611 * XXX zero out the tag and array values. Need to investigate why the 5612 * XXX calls occur, but at least prevent the panic for now. 5613 */ 5614 void 5615 rfs4_compound_free(COMPOUND4res *resp) 5616 { 5617 uint_t i; 5618 5619 if (resp->tag.utf8string_val) { 5620 UTF8STRING_FREE(resp->tag) 5621 } 5622 5623 for (i = 0; i < resp->array_len; i++) { 5624 nfs_resop4 *resop; 5625 uint_t op; 5626 5627 resop = &resp->array[i]; 5628 op = (uint_t)resop->resop; 5629 if (op < rfsv4disp_cnt) { 5630 (*rfsv4disptab[op].dis_resfree)(resop); 5631 } 5632 } 5633 if (resp->array != NULL) { 5634 kmem_free(resp->array, resp->array_len * sizeof (nfs_resop4)); 5635 } 5636 } 5637 5638 /* 5639 * Process the value of the compound request rpc flags, as a bit-AND 5640 * of the individual per-op flags (idempotent, allowork, publicfh_ok) 5641 */ 5642 void 5643 rfs4_compound_flagproc(COMPOUND4args *args, int *flagp) 5644 { 5645 int i; 5646 int flag = RPC_ALL; 5647 5648 for (i = 0; flag && i < args->array_len; i++) { 5649 uint_t op; 5650 5651 op = (uint_t)args->array[i].argop; 5652 5653 if (op < rfsv4disp_cnt) 5654 flag &= rfsv4disptab[op].dis_flags; 5655 else 5656 flag = 0; 5657 } 5658 *flagp = flag; 5659 } 5660 5661 nfsstat4 5662 rfs4_client_sysid(rfs4_client_t *cp, sysid_t *sp) 5663 { 5664 nfsstat4 e; 5665 5666 rfs4_dbe_lock(cp->dbe); 5667 5668 if (cp->sysidt != LM_NOSYSID) { 5669 *sp = cp->sysidt; 5670 e = NFS4_OK; 5671 5672 } else if ((cp->sysidt = lm_alloc_sysidt()) != LM_NOSYSID) { 5673 *sp = cp->sysidt; 5674 e = NFS4_OK; 5675 5676 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 5677 "rfs4_client_sysid: allocated 0x%x\n", *sp)); 5678 } else 5679 e = NFS4ERR_DELAY; 5680 5681 rfs4_dbe_unlock(cp->dbe); 5682 return (e); 5683 } 5684 5685 #if defined(DEBUG) && ! defined(lint) 5686 static void lock_print(char *str, int operation, struct flock64 *flk) 5687 { 5688 char *op, *type; 5689 5690 switch (operation) { 5691 case F_GETLK: op = "F_GETLK"; 5692 break; 5693 case F_SETLK: op = "F_SETLK"; 5694 break; 5695 case F_SETLK_NBMAND: op = "F_SETLK_NBMAND"; 5696 break; 5697 default: op = "F_UNKNOWN"; 5698 break; 5699 } 5700 switch (flk->l_type) { 5701 case F_UNLCK: type = "F_UNLCK"; 5702 break; 5703 case F_RDLCK: type = "F_RDLCK"; 5704 break; 5705 case F_WRLCK: type = "F_WRLCK"; 5706 break; 5707 default: type = "F_UNKNOWN"; 5708 break; 5709 } 5710 5711 ASSERT(flk->l_whence == 0); 5712 cmn_err(CE_NOTE, "%s: %s, type = %s, off = %llx len = %llx pid = %d", 5713 str, op, type, (longlong_t)flk->l_start, 5714 flk->l_len ? (longlong_t)flk->l_len : ~0LL, flk->l_pid); 5715 } 5716 5717 #define LOCK_PRINT(d, s, t, f) if (d) lock_print(s, t, f) 5718 #else 5719 #define LOCK_PRINT(d, s, t, f) 5720 #endif 5721 5722 /*ARGSUSED*/ 5723 static bool_t 5724 creds_ok(cred_set_t cr_set, struct svc_req *req, struct compound_state *cs) 5725 { 5726 return (TRUE); 5727 } 5728 5729 /* 5730 * Look up the pathname using the vp in cs as the directory vnode. 5731 * cs->vp will be the vnode for the file on success 5732 */ 5733 5734 static nfsstat4 5735 rfs4_lookup(component4 *component, struct svc_req *req, 5736 struct compound_state *cs) 5737 { 5738 char *nm; 5739 uint32_t len; 5740 nfsstat4 status; 5741 5742 if (cs->vp == NULL) { 5743 return (NFS4ERR_NOFILEHANDLE); 5744 } 5745 if (cs->vp->v_type != VDIR) { 5746 return (NFS4ERR_NOTDIR); 5747 } 5748 5749 if (!utf8_dir_verify(component)) 5750 return (NFS4ERR_INVAL); 5751 5752 nm = utf8_to_fn(component, &len, NULL); 5753 if (nm == NULL) { 5754 return (NFS4ERR_INVAL); 5755 } 5756 5757 if (len > MAXNAMELEN) { 5758 kmem_free(nm, len); 5759 return (NFS4ERR_NAMETOOLONG); 5760 } 5761 5762 status = do_rfs4_op_lookup(nm, len, req, cs); 5763 5764 kmem_free(nm, len); 5765 5766 return (status); 5767 } 5768 5769 static nfsstat4 5770 rfs4_lookupfile(component4 *component, struct svc_req *req, 5771 struct compound_state *cs, uint32_t access, 5772 change_info4 *cinfo) 5773 { 5774 nfsstat4 status; 5775 vnode_t *dvp = cs->vp; 5776 vattr_t bva, ava, fva; 5777 int error; 5778 5779 /* Get "before" change value */ 5780 bva.va_mask = AT_CTIME|AT_SEQ; 5781 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 5782 if (error) 5783 return (puterrno4(error)); 5784 5785 /* rfs4_lookup may VN_RELE directory */ 5786 VN_HOLD(dvp); 5787 5788 status = rfs4_lookup(component, req, cs); 5789 if (status != NFS4_OK) { 5790 VN_RELE(dvp); 5791 return (status); 5792 } 5793 5794 /* 5795 * Get "after" change value, if it fails, simply return the 5796 * before value. 5797 */ 5798 ava.va_mask = AT_CTIME|AT_SEQ; 5799 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 5800 ava.va_ctime = bva.va_ctime; 5801 ava.va_seq = 0; 5802 } 5803 VN_RELE(dvp); 5804 5805 /* 5806 * Validate the file is a file 5807 */ 5808 fva.va_mask = AT_TYPE|AT_MODE; 5809 error = VOP_GETATTR(cs->vp, &fva, 0, cs->cr, NULL); 5810 if (error) 5811 return (puterrno4(error)); 5812 5813 if (fva.va_type != VREG) { 5814 if (fva.va_type == VDIR) 5815 return (NFS4ERR_ISDIR); 5816 if (fva.va_type == VLNK) 5817 return (NFS4ERR_SYMLINK); 5818 return (NFS4ERR_INVAL); 5819 } 5820 5821 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime); 5822 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 5823 5824 /* 5825 * It is undefined if VOP_LOOKUP will change va_seq, so 5826 * cinfo.atomic = TRUE only if we have 5827 * non-zero va_seq's, and they have not changed. 5828 */ 5829 if (bva.va_seq && ava.va_seq && ava.va_seq == bva.va_seq) 5830 cinfo->atomic = TRUE; 5831 else 5832 cinfo->atomic = FALSE; 5833 5834 /* Check for mandatory locking */ 5835 cs->mandlock = MANDLOCK(cs->vp, fva.va_mode); 5836 return (check_open_access(access, cs, req)); 5837 } 5838 5839 static nfsstat4 5840 create_vnode(vnode_t *dvp, char *nm, vattr_t *vap, createmode4 mode, 5841 timespec32_t *mtime, cred_t *cr, vnode_t **vpp, bool_t *created) 5842 { 5843 int error; 5844 nfsstat4 status = NFS4_OK; 5845 vattr_t va; 5846 5847 tryagain: 5848 5849 /* 5850 * The file open mode used is VWRITE. If the client needs 5851 * some other semantic, then it should do the access checking 5852 * itself. It would have been nice to have the file open mode 5853 * passed as part of the arguments. 5854 */ 5855 5856 *created = TRUE; 5857 error = VOP_CREATE(dvp, nm, vap, EXCL, VWRITE, vpp, cr, 0, NULL, NULL); 5858 5859 if (error) { 5860 *created = FALSE; 5861 5862 /* 5863 * If we got something other than file already exists 5864 * then just return this error. Otherwise, we got 5865 * EEXIST. If we were doing a GUARDED create, then 5866 * just return this error. Otherwise, we need to 5867 * make sure that this wasn't a duplicate of an 5868 * exclusive create request. 5869 * 5870 * The assumption is made that a non-exclusive create 5871 * request will never return EEXIST. 5872 */ 5873 5874 if (error != EEXIST || mode == GUARDED4) { 5875 status = puterrno4(error); 5876 return (status); 5877 } 5878 error = VOP_LOOKUP(dvp, nm, vpp, NULL, 0, NULL, cr, 5879 NULL, NULL, NULL); 5880 5881 if (error) { 5882 /* 5883 * We couldn't find the file that we thought that 5884 * we just created. So, we'll just try creating 5885 * it again. 5886 */ 5887 if (error == ENOENT) 5888 goto tryagain; 5889 5890 status = puterrno4(error); 5891 return (status); 5892 } 5893 5894 if (mode == UNCHECKED4) { 5895 /* existing object must be regular file */ 5896 if ((*vpp)->v_type != VREG) { 5897 if ((*vpp)->v_type == VDIR) 5898 status = NFS4ERR_ISDIR; 5899 else if ((*vpp)->v_type == VLNK) 5900 status = NFS4ERR_SYMLINK; 5901 else 5902 status = NFS4ERR_INVAL; 5903 VN_RELE(*vpp); 5904 return (status); 5905 } 5906 5907 return (NFS4_OK); 5908 } 5909 5910 /* Check for duplicate request */ 5911 ASSERT(mtime != 0); 5912 va.va_mask = AT_MTIME; 5913 error = VOP_GETATTR(*vpp, &va, 0, cr, NULL); 5914 if (!error) { 5915 /* We found the file */ 5916 if (va.va_mtime.tv_sec != mtime->tv_sec || 5917 va.va_mtime.tv_nsec != mtime->tv_nsec) { 5918 /* but its not our creation */ 5919 VN_RELE(*vpp); 5920 return (NFS4ERR_EXIST); 5921 } 5922 *created = TRUE; /* retrans of create == created */ 5923 return (NFS4_OK); 5924 } 5925 VN_RELE(*vpp); 5926 return (NFS4ERR_EXIST); 5927 } 5928 5929 return (NFS4_OK); 5930 } 5931 5932 static nfsstat4 5933 check_open_access(uint32_t access, 5934 struct compound_state *cs, struct svc_req *req) 5935 { 5936 int error; 5937 vnode_t *vp; 5938 bool_t readonly; 5939 cred_t *cr = cs->cr; 5940 5941 /* For now we don't allow mandatory locking as per V2/V3 */ 5942 if (cs->access == CS_ACCESS_DENIED || cs->mandlock) { 5943 return (NFS4ERR_ACCESS); 5944 } 5945 5946 vp = cs->vp; 5947 ASSERT(cr != NULL && vp->v_type == VREG); 5948 5949 /* 5950 * If the file system is exported read only and we are trying 5951 * to open for write, then return NFS4ERR_ROFS 5952 */ 5953 5954 readonly = rdonly4(cs->exi, cs->vp, req); 5955 5956 if ((access & OPEN4_SHARE_ACCESS_WRITE) && readonly) 5957 return (NFS4ERR_ROFS); 5958 5959 if (access & OPEN4_SHARE_ACCESS_READ) { 5960 if ((VOP_ACCESS(vp, VREAD, 0, cr, NULL) != 0) && 5961 (VOP_ACCESS(vp, VEXEC, 0, cr, NULL) != 0)) { 5962 return (NFS4ERR_ACCESS); 5963 } 5964 } 5965 5966 if (access & OPEN4_SHARE_ACCESS_WRITE) { 5967 error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL); 5968 if (error) 5969 return (NFS4ERR_ACCESS); 5970 } 5971 5972 return (NFS4_OK); 5973 } 5974 5975 static nfsstat4 5976 rfs4_createfile(OPEN4args *args, struct svc_req *req, struct compound_state *cs, 5977 change_info4 *cinfo, bitmap4 *attrset, clientid4 clientid) 5978 { 5979 struct nfs4_svgetit_arg sarg; 5980 struct nfs4_ntov_table ntov; 5981 5982 bool_t ntov_table_init = FALSE; 5983 struct statvfs64 sb; 5984 nfsstat4 status; 5985 vnode_t *vp; 5986 vattr_t bva, ava, iva, cva, *vap; 5987 vnode_t *dvp; 5988 timespec32_t *mtime; 5989 char *nm = NULL; 5990 uint_t buflen; 5991 bool_t created; 5992 bool_t setsize = FALSE; 5993 len_t reqsize; 5994 int error; 5995 bool_t trunc; 5996 caller_context_t ct; 5997 component4 *component; 5998 bslabel_t *clabel; 5999 6000 sarg.sbp = &sb; 6001 6002 dvp = cs->vp; 6003 6004 /* Check if the file system is read only */ 6005 if (rdonly4(cs->exi, dvp, req)) 6006 return (NFS4ERR_ROFS); 6007 6008 /* check the label of including directory */ 6009 if (is_system_labeled()) { 6010 ASSERT(req->rq_label != NULL); 6011 clabel = req->rq_label; 6012 DTRACE_PROBE2(tx__rfs4__log__info__opremove__clabel, char *, 6013 "got client label from request(1)", 6014 struct svc_req *, req); 6015 if (!blequal(&l_admin_low->tsl_label, clabel)) { 6016 if (!do_rfs_label_check(clabel, dvp, EQUALITY_CHECK)) { 6017 return (NFS4ERR_ACCESS); 6018 } 6019 } 6020 } 6021 6022 /* 6023 * Get the last component of path name in nm. cs will reference 6024 * the including directory on success. 6025 */ 6026 component = &args->open_claim4_u.file; 6027 if (!utf8_dir_verify(component)) 6028 return (NFS4ERR_INVAL); 6029 6030 nm = utf8_to_fn(component, &buflen, NULL); 6031 6032 if (nm == NULL) 6033 return (NFS4ERR_RESOURCE); 6034 6035 if (buflen > MAXNAMELEN) { 6036 kmem_free(nm, buflen); 6037 return (NFS4ERR_NAMETOOLONG); 6038 } 6039 6040 bva.va_mask = AT_TYPE|AT_CTIME|AT_SEQ; 6041 error = VOP_GETATTR(dvp, &bva, 0, cs->cr, NULL); 6042 if (error) { 6043 kmem_free(nm, buflen); 6044 return (puterrno4(error)); 6045 } 6046 6047 if (bva.va_type != VDIR) { 6048 kmem_free(nm, buflen); 6049 return (NFS4ERR_NOTDIR); 6050 } 6051 6052 NFS4_SET_FATTR4_CHANGE(cinfo->before, bva.va_ctime) 6053 6054 switch (args->mode) { 6055 case GUARDED4: 6056 /*FALLTHROUGH*/ 6057 case UNCHECKED4: 6058 nfs4_ntov_table_init(&ntov); 6059 ntov_table_init = TRUE; 6060 6061 *attrset = 0; 6062 status = do_rfs4_set_attrs(attrset, 6063 &args->createhow4_u.createattrs, 6064 cs, &sarg, &ntov, NFS4ATTR_SETIT); 6065 6066 if (status == NFS4_OK && (sarg.vap->va_mask & AT_TYPE) && 6067 sarg.vap->va_type != VREG) { 6068 if (sarg.vap->va_type == VDIR) 6069 status = NFS4ERR_ISDIR; 6070 else if (sarg.vap->va_type == VLNK) 6071 status = NFS4ERR_SYMLINK; 6072 else 6073 status = NFS4ERR_INVAL; 6074 } 6075 6076 if (status != NFS4_OK) { 6077 kmem_free(nm, buflen); 6078 nfs4_ntov_table_free(&ntov, &sarg); 6079 *attrset = 0; 6080 return (status); 6081 } 6082 6083 vap = sarg.vap; 6084 vap->va_type = VREG; 6085 vap->va_mask |= AT_TYPE; 6086 6087 if ((vap->va_mask & AT_MODE) == 0) { 6088 vap->va_mask |= AT_MODE; 6089 vap->va_mode = (mode_t)0600; 6090 } 6091 6092 if (vap->va_mask & AT_SIZE) { 6093 6094 /* Disallow create with a non-zero size */ 6095 6096 if ((reqsize = sarg.vap->va_size) != 0) { 6097 kmem_free(nm, buflen); 6098 nfs4_ntov_table_free(&ntov, &sarg); 6099 *attrset = 0; 6100 return (NFS4ERR_INVAL); 6101 } 6102 setsize = TRUE; 6103 } 6104 break; 6105 6106 case EXCLUSIVE4: 6107 /* prohibit EXCL create of named attributes */ 6108 if (dvp->v_flag & V_XATTRDIR) { 6109 kmem_free(nm, buflen); 6110 *attrset = 0; 6111 return (NFS4ERR_INVAL); 6112 } 6113 6114 cva.va_mask = AT_TYPE | AT_MTIME | AT_MODE; 6115 cva.va_type = VREG; 6116 /* 6117 * Ensure no time overflows. Assumes underlying 6118 * filesystem supports at least 32 bits. 6119 * Truncate nsec to usec resolution to allow valid 6120 * compares even if the underlying filesystem truncates. 6121 */ 6122 mtime = (timespec32_t *)&args->createhow4_u.createverf; 6123 cva.va_mtime.tv_sec = mtime->tv_sec % TIME32_MAX; 6124 cva.va_mtime.tv_nsec = (mtime->tv_nsec / 1000) * 1000; 6125 cva.va_mode = (mode_t)0; 6126 vap = &cva; 6127 break; 6128 } 6129 6130 status = create_vnode(dvp, nm, vap, args->mode, mtime, 6131 cs->cr, &vp, &created); 6132 kmem_free(nm, buflen); 6133 6134 if (status != NFS4_OK) { 6135 if (ntov_table_init) 6136 nfs4_ntov_table_free(&ntov, &sarg); 6137 *attrset = 0; 6138 return (status); 6139 } 6140 6141 trunc = (setsize && !created); 6142 6143 if (args->mode != EXCLUSIVE4) { 6144 bitmap4 createmask = args->createhow4_u.createattrs.attrmask; 6145 6146 /* 6147 * True verification that object was created with correct 6148 * attrs is impossible. The attrs could have been changed 6149 * immediately after object creation. If attributes did 6150 * not verify, the only recourse for the server is to 6151 * destroy the object. Maybe if some attrs (like gid) 6152 * are set incorrectly, the object should be destroyed; 6153 * however, seems bad as a default policy. Do we really 6154 * want to destroy an object over one of the times not 6155 * verifying correctly? For these reasons, the server 6156 * currently sets bits in attrset for createattrs 6157 * that were set; however, no verification is done. 6158 * 6159 * vmask_to_nmask accounts for vattr bits set on create 6160 * [do_rfs4_set_attrs() only sets resp bits for 6161 * non-vattr/vfs bits.] 6162 * Mask off any bits we set by default so as not to return 6163 * more attrset bits than were requested in createattrs 6164 */ 6165 if (created) { 6166 nfs4_vmask_to_nmask(sarg.vap->va_mask, attrset); 6167 *attrset &= createmask; 6168 } else { 6169 /* 6170 * We did not create the vnode (we tried but it 6171 * already existed). In this case, the only createattr 6172 * that the spec allows the server to set is size, 6173 * and even then, it can only be set if it is 0. 6174 */ 6175 *attrset = 0; 6176 if (trunc) 6177 *attrset = FATTR4_SIZE_MASK; 6178 } 6179 } 6180 if (ntov_table_init) 6181 nfs4_ntov_table_free(&ntov, &sarg); 6182 6183 /* 6184 * Get the initial "after" sequence number, if it fails, 6185 * set to zero, time to before. 6186 */ 6187 iva.va_mask = AT_CTIME|AT_SEQ; 6188 if (VOP_GETATTR(dvp, &iva, 0, cs->cr, NULL)) { 6189 iva.va_seq = 0; 6190 iva.va_ctime = bva.va_ctime; 6191 } 6192 6193 /* 6194 * create_vnode attempts to create the file exclusive, 6195 * if it already exists the VOP_CREATE will fail and 6196 * may not increase va_seq. It is atomic if 6197 * we haven't changed the directory, but if it has changed 6198 * we don't know what changed it. 6199 */ 6200 if (!created) { 6201 if (bva.va_seq && iva.va_seq && 6202 bva.va_seq == iva.va_seq) 6203 cinfo->atomic = TRUE; 6204 else 6205 cinfo->atomic = FALSE; 6206 NFS4_SET_FATTR4_CHANGE(cinfo->after, iva.va_ctime); 6207 } else { 6208 /* 6209 * The entry was created, we need to sync the 6210 * directory metadata. 6211 */ 6212 (void) VOP_FSYNC(dvp, 0, cs->cr, NULL); 6213 6214 /* 6215 * Get "after" change value, if it fails, simply return the 6216 * before value. 6217 */ 6218 ava.va_mask = AT_CTIME|AT_SEQ; 6219 if (VOP_GETATTR(dvp, &ava, 0, cs->cr, NULL)) { 6220 ava.va_ctime = bva.va_ctime; 6221 ava.va_seq = 0; 6222 } 6223 6224 NFS4_SET_FATTR4_CHANGE(cinfo->after, ava.va_ctime); 6225 6226 /* 6227 * The cinfo->atomic = TRUE only if we have 6228 * non-zero va_seq's, and it has incremented by exactly one 6229 * during the create_vnode and it didn't 6230 * change during the VOP_FSYNC. 6231 */ 6232 if (bva.va_seq && iva.va_seq && ava.va_seq && 6233 iva.va_seq == (bva.va_seq + 1) && iva.va_seq == ava.va_seq) 6234 cinfo->atomic = TRUE; 6235 else 6236 cinfo->atomic = FALSE; 6237 } 6238 6239 /* Check for mandatory locking and that the size gets set. */ 6240 cva.va_mask = AT_MODE; 6241 if (setsize) 6242 cva.va_mask |= AT_SIZE; 6243 6244 /* Assume the worst */ 6245 cs->mandlock = TRUE; 6246 6247 if (VOP_GETATTR(vp, &cva, 0, cs->cr, NULL) == 0) { 6248 cs->mandlock = MANDLOCK(cs->vp, cva.va_mode); 6249 6250 /* 6251 * Truncate the file if necessary; this would be 6252 * the case for create over an existing file. 6253 */ 6254 6255 if (trunc) { 6256 int in_crit = 0; 6257 rfs4_file_t *fp; 6258 bool_t create = FALSE; 6259 6260 /* 6261 * We are writing over an existing file. 6262 * Check to see if we need to recall a delegation. 6263 */ 6264 rfs4_hold_deleg_policy(); 6265 if ((fp = rfs4_findfile(vp, NULL, &create)) != NULL) { 6266 if (rfs4_check_delegated_byfp(FWRITE, fp, 6267 (reqsize == 0), FALSE, FALSE, &clientid)) { 6268 rfs4_file_rele(fp); 6269 rfs4_rele_deleg_policy(); 6270 VN_RELE(vp); 6271 *attrset = 0; 6272 return (NFS4ERR_DELAY); 6273 } 6274 rfs4_file_rele(fp); 6275 } 6276 rfs4_rele_deleg_policy(); 6277 6278 if (nbl_need_check(vp)) { 6279 in_crit = 1; 6280 6281 ASSERT(reqsize == 0); 6282 6283 nbl_start_crit(vp, RW_READER); 6284 if (nbl_conflict(vp, NBL_WRITE, 0, 6285 cva.va_size, 0, NULL)) { 6286 in_crit = 0; 6287 nbl_end_crit(vp); 6288 VN_RELE(vp); 6289 *attrset = 0; 6290 return (NFS4ERR_ACCESS); 6291 } 6292 } 6293 ct.cc_sysid = 0; 6294 ct.cc_pid = 0; 6295 ct.cc_caller_id = nfs4_srv_caller_id; 6296 ct.cc_flags = CC_DONTBLOCK; 6297 6298 cva.va_mask = AT_SIZE; 6299 cva.va_size = reqsize; 6300 (void) VOP_SETATTR(vp, &cva, 0, cs->cr, &ct); 6301 if (in_crit) 6302 nbl_end_crit(vp); 6303 } 6304 } 6305 6306 error = makefh4(&cs->fh, vp, cs->exi); 6307 6308 /* 6309 * Force modified data and metadata out to stable storage. 6310 */ 6311 (void) VOP_FSYNC(vp, FNODSYNC, cs->cr, NULL); 6312 6313 if (error) { 6314 VN_RELE(vp); 6315 *attrset = 0; 6316 return (puterrno4(error)); 6317 } 6318 6319 /* if parent dir is attrdir, set namedattr fh flag */ 6320 if (dvp->v_flag & V_XATTRDIR) 6321 set_fh4_flag(&cs->fh, FH4_NAMEDATTR); 6322 6323 if (cs->vp) 6324 VN_RELE(cs->vp); 6325 6326 cs->vp = vp; 6327 6328 /* 6329 * if we did not create the file, we will need to check 6330 * the access bits on the file 6331 */ 6332 6333 if (!created) { 6334 if (setsize) 6335 args->share_access |= OPEN4_SHARE_ACCESS_WRITE; 6336 status = check_open_access(args->share_access, cs, req); 6337 if (status != NFS4_OK) 6338 *attrset = 0; 6339 } 6340 return (status); 6341 } 6342 6343 /*ARGSUSED*/ 6344 static void 6345 rfs4_do_open(struct compound_state *cs, struct svc_req *req, 6346 rfs4_openowner_t *oo, delegreq_t deleg, 6347 uint32_t access, uint32_t deny, 6348 OPEN4res *resp, int deleg_cur) 6349 { 6350 /* XXX Currently not using req */ 6351 rfs4_state_t *state; 6352 rfs4_file_t *file; 6353 bool_t screate = TRUE; 6354 bool_t fcreate = TRUE; 6355 uint32_t amodes; 6356 uint32_t dmodes; 6357 rfs4_deleg_state_t *dsp; 6358 struct shrlock shr; 6359 struct shr_locowner shr_loco; 6360 sysid_t sysid; 6361 nfsstat4 status; 6362 caller_context_t ct; 6363 int fflags = 0; 6364 int recall = 0; 6365 int err; 6366 int cmd; 6367 6368 /* get the file struct and hold a lock on it during initial open */ 6369 file = rfs4_findfile_withlock(cs->vp, &cs->fh, &fcreate); 6370 if (file == NULL) { 6371 NFS4_DEBUG(rfs4_debug, 6372 (CE_NOTE, "rfs4_do_open: can't find file")); 6373 resp->status = NFS4ERR_SERVERFAULT; 6374 return; 6375 } 6376 6377 state = rfs4_findstate_by_owner_file(oo, file, &screate); 6378 if (state == NULL) { 6379 NFS4_DEBUG(rfs4_debug, 6380 (CE_NOTE, "rfs4_do_open: can't find state")); 6381 resp->status = NFS4ERR_RESOURCE; 6382 /* No need to keep any reference */ 6383 rfs4_file_rele_withunlock(file); 6384 return; 6385 } 6386 6387 /* try to get the sysid before continuing */ 6388 if ((status = rfs4_client_sysid(oo->client, &sysid)) != NFS4_OK) { 6389 resp->status = status; 6390 rfs4_file_rele(file); 6391 /* Not a fully formed open; "close" it */ 6392 if (screate == TRUE) 6393 rfs4_state_close(state, FALSE, FALSE, cs->cr); 6394 rfs4_state_rele(state); 6395 return; 6396 } 6397 6398 /* Calculate the fflags for this OPEN. */ 6399 if (access & OPEN4_SHARE_ACCESS_READ) 6400 fflags |= FREAD; 6401 if (access & OPEN4_SHARE_ACCESS_WRITE) 6402 fflags |= FWRITE; 6403 6404 /* 6405 * Calculate the new deny and access mode that this open is adding to 6406 * the file for this open owner; 6407 */ 6408 dmodes = (deny & ~state->share_deny); 6409 amodes = (access & ~state->share_access); 6410 6411 /* 6412 * Check to see the client has already sent an open for this 6413 * open owner on this file with the same share/deny modes. 6414 * If so, we don't need to check for a conflict and we don't 6415 * need to add another shrlock. If not, then we need to 6416 * check for conflicts in deny and access before checking for 6417 * conflicts in delegation. We don't want to recall a 6418 * delegation based on an open that will eventually fail based 6419 * on shares modes. 6420 */ 6421 6422 if (dmodes || amodes) { 6423 shr.s_access = (short)access; 6424 shr.s_deny = (short)deny; 6425 shr.s_pid = rfs4_dbe_getid(oo->dbe); 6426 shr.s_sysid = sysid; 6427 shr_loco.sl_pid = shr.s_pid; 6428 shr_loco.sl_id = shr.s_sysid; 6429 shr.s_owner = (caddr_t)&shr_loco; 6430 shr.s_own_len = sizeof (shr_loco); 6431 6432 cmd = nbl_need_check(cs->vp) ? F_SHARE_NBMAND : F_SHARE; 6433 if ((err = vop_shrlock(cs->vp, cmd, &shr, fflags)) != 0) { 6434 6435 resp->status = err == EAGAIN ? 6436 NFS4ERR_SHARE_DENIED : puterrno4(err); 6437 6438 rfs4_file_rele(file); 6439 /* Not a fully formed open; "close" it */ 6440 if (screate == TRUE) 6441 rfs4_state_close(state, FALSE, FALSE, cs->cr); 6442 rfs4_state_rele(state); 6443 return; 6444 } 6445 } 6446 6447 rfs4_dbe_lock(state->dbe); 6448 rfs4_dbe_lock(file->dbe); 6449 6450 /* 6451 * Check to see if this file is delegated and if so, if a 6452 * recall needs to be done. 6453 */ 6454 if (rfs4_check_recall(state, access)) { 6455 rfs4_dbe_unlock(file->dbe); 6456 rfs4_dbe_unlock(state->dbe); 6457 rfs4_recall_deleg(file, FALSE, state->owner->client); 6458 delay(NFS4_DELEGATION_CONFLICT_DELAY); 6459 rfs4_dbe_lock(state->dbe); 6460 rfs4_dbe_lock(file->dbe); 6461 /* Let's see if the delegation was returned */ 6462 if (rfs4_check_recall(state, access)) { 6463 rfs4_dbe_unlock(file->dbe); 6464 rfs4_dbe_unlock(state->dbe); 6465 rfs4_file_rele(file); 6466 rfs4_update_lease(state->owner->client); 6467 (void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags); 6468 /* Not a fully formed open; "close" it */ 6469 if (screate == TRUE) 6470 rfs4_state_close(state, FALSE, FALSE, cs->cr); 6471 rfs4_state_rele(state); 6472 resp->status = NFS4ERR_DELAY; 6473 return; 6474 } 6475 } 6476 /* 6477 * the share check passed and any delegation conflict has been 6478 * taken care of, now call vop_open. 6479 * if this is the first open then call vop_open with fflags. 6480 * if not, call vn_open_upgrade with just the upgrade flags. 6481 * 6482 * if the file has been opened already, it will have the current 6483 * access mode in the state struct. if it has no share access, then 6484 * this is a new open. 6485 * 6486 * However, if this is open with CLAIM_DLEGATE_CUR, then don't 6487 * call VOP_OPEN(), just do the open upgrade. 6488 */ 6489 if (((state->share_access & OPEN4_SHARE_ACCESS_BOTH) == 0) && 6490 !deleg_cur) { 6491 ct.cc_sysid = sysid; 6492 ct.cc_pid = shr.s_pid; 6493 ct.cc_caller_id = nfs4_srv_caller_id; 6494 ct.cc_flags = CC_DONTBLOCK; 6495 err = VOP_OPEN(&cs->vp, fflags, cs->cr, &ct); 6496 if (err) { 6497 rfs4_dbe_unlock(file->dbe); 6498 rfs4_dbe_unlock(state->dbe); 6499 rfs4_file_rele(file); 6500 (void) vop_shrlock(cs->vp, F_UNSHARE, &shr, fflags); 6501 /* Not a fully formed open; "close" it */ 6502 if (screate == TRUE) 6503 rfs4_state_close(state, FALSE, FALSE, cs->cr); 6504 rfs4_state_rele(state); 6505 resp->status = NFS4ERR_SERVERFAULT; 6506 return; 6507 } 6508 } else { /* open upgrade */ 6509 /* 6510 * calculate the fflags for the new mode that is being added 6511 * by this upgrade. 6512 */ 6513 fflags = 0; 6514 if (amodes & OPEN4_SHARE_ACCESS_READ) 6515 fflags |= FREAD; 6516 if (amodes & OPEN4_SHARE_ACCESS_WRITE) 6517 fflags |= FWRITE; 6518 vn_open_upgrade(cs->vp, fflags); 6519 } 6520 6521 if (dmodes & OPEN4_SHARE_DENY_READ) 6522 file->deny_read++; 6523 if (dmodes & OPEN4_SHARE_DENY_WRITE) 6524 file->deny_write++; 6525 file->share_deny |= deny; 6526 state->share_deny |= deny; 6527 6528 if (amodes & OPEN4_SHARE_ACCESS_READ) 6529 file->access_read++; 6530 if (amodes & OPEN4_SHARE_ACCESS_WRITE) 6531 file->access_write++; 6532 file->share_access |= access; 6533 state->share_access |= access; 6534 6535 /* 6536 * Check for delegation here. if the deleg argument is not 6537 * DELEG_ANY, then this is a reclaim from a client and 6538 * we must honor the delegation requested. If necessary we can 6539 * set the recall flag. 6540 */ 6541 6542 dsp = rfs4_grant_delegation(deleg, state, &recall); 6543 6544 cs->deleg = (file->dinfo->dtype == OPEN_DELEGATE_WRITE); 6545 6546 next_stateid(&state->stateid); 6547 6548 resp->stateid = state->stateid.stateid; 6549 6550 rfs4_dbe_unlock(file->dbe); 6551 rfs4_dbe_unlock(state->dbe); 6552 6553 if (dsp) { 6554 rfs4_set_deleg_response(dsp, &resp->delegation, NULL, recall); 6555 rfs4_deleg_state_rele(dsp); 6556 } 6557 6558 rfs4_file_rele(file); 6559 rfs4_state_rele(state); 6560 6561 resp->status = NFS4_OK; 6562 } 6563 6564 /*ARGSUSED*/ 6565 static void 6566 rfs4_do_opennull(struct compound_state *cs, struct svc_req *req, 6567 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6568 { 6569 change_info4 *cinfo = &resp->cinfo; 6570 bitmap4 *attrset = &resp->attrset; 6571 6572 if (args->opentype == OPEN4_NOCREATE) 6573 resp->status = rfs4_lookupfile(&args->open_claim4_u.file, 6574 req, cs, args->share_access, cinfo); 6575 else { 6576 /* inhibit delegation grants during exclusive create */ 6577 6578 if (args->mode == EXCLUSIVE4) 6579 rfs4_disable_delegation(); 6580 6581 resp->status = rfs4_createfile(args, req, cs, cinfo, attrset, 6582 oo->client->clientid); 6583 } 6584 6585 if (resp->status == NFS4_OK) { 6586 6587 /* cs->vp cs->fh now reference the desired file */ 6588 6589 rfs4_do_open(cs, req, oo, DELEG_ANY, args->share_access, 6590 args->share_deny, resp, 0); 6591 6592 /* 6593 * If rfs4_createfile set attrset, we must 6594 * clear this attrset before the response is copied. 6595 */ 6596 if (resp->status != NFS4_OK && resp->attrset) { 6597 resp->attrset = 0; 6598 } 6599 } 6600 else 6601 *cs->statusp = resp->status; 6602 6603 if (args->mode == EXCLUSIVE4) 6604 rfs4_enable_delegation(); 6605 } 6606 6607 /*ARGSUSED*/ 6608 static void 6609 rfs4_do_openprev(struct compound_state *cs, struct svc_req *req, 6610 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6611 { 6612 change_info4 *cinfo = &resp->cinfo; 6613 vattr_t va; 6614 vtype_t v_type = cs->vp->v_type; 6615 int error = 0; 6616 6617 /* Verify that we have a regular file */ 6618 if (v_type != VREG) { 6619 if (v_type == VDIR) 6620 resp->status = NFS4ERR_ISDIR; 6621 else if (v_type == VLNK) 6622 resp->status = NFS4ERR_SYMLINK; 6623 else 6624 resp->status = NFS4ERR_INVAL; 6625 return; 6626 } 6627 6628 va.va_mask = AT_MODE|AT_UID; 6629 error = VOP_GETATTR(cs->vp, &va, 0, cs->cr, NULL); 6630 if (error) { 6631 resp->status = puterrno4(error); 6632 return; 6633 } 6634 6635 cs->mandlock = MANDLOCK(cs->vp, va.va_mode); 6636 6637 /* 6638 * Check if we have access to the file, Note the the file 6639 * could have originally been open UNCHECKED or GUARDED 6640 * with mode bits that will now fail, but there is nothing 6641 * we can really do about that except in the case that the 6642 * owner of the file is the one requesting the open. 6643 */ 6644 if (crgetuid(cs->cr) != va.va_uid) { 6645 resp->status = check_open_access(args->share_access, cs, req); 6646 if (resp->status != NFS4_OK) { 6647 return; 6648 } 6649 } 6650 6651 /* 6652 * cinfo on a CLAIM_PREVIOUS is undefined, initialize to zero 6653 */ 6654 cinfo->before = 0; 6655 cinfo->after = 0; 6656 cinfo->atomic = FALSE; 6657 6658 rfs4_do_open(cs, req, oo, 6659 NFS4_DELEG4TYPE2REQTYPE(args->open_claim4_u.delegate_type), 6660 args->share_access, args->share_deny, resp, 0); 6661 } 6662 6663 static void 6664 rfs4_do_opendelcur(struct compound_state *cs, struct svc_req *req, 6665 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6666 { 6667 int error; 6668 nfsstat4 status; 6669 stateid4 stateid = 6670 args->open_claim4_u.delegate_cur_info.delegate_stateid; 6671 rfs4_deleg_state_t *dsp; 6672 6673 /* 6674 * Find the state info from the stateid and confirm that the 6675 * file is delegated. If the state openowner is the same as 6676 * the supplied openowner we're done. If not, get the file 6677 * info from the found state info. Use that file info to 6678 * create the state for this lock owner. Note solaris doen't 6679 * really need the pathname to find the file. We may want to 6680 * lookup the pathname and make sure that the vp exist and 6681 * matches the vp in the file structure. However it is 6682 * possible that the pathname nolonger exists (local process 6683 * unlinks the file), so this may not be that useful. 6684 */ 6685 6686 status = rfs4_get_deleg_state(&stateid, &dsp); 6687 if (status != NFS4_OK) { 6688 resp->status = status; 6689 return; 6690 } 6691 6692 ASSERT(dsp->finfo->dinfo->dtype != OPEN_DELEGATE_NONE); 6693 6694 /* 6695 * New lock owner, create state. Since this was probably called 6696 * in response to a CB_RECALL we set deleg to DELEG_NONE 6697 */ 6698 6699 ASSERT(cs->vp != NULL); 6700 VN_RELE(cs->vp); 6701 VN_HOLD(dsp->finfo->vp); 6702 cs->vp = dsp->finfo->vp; 6703 6704 if (error = makefh4(&cs->fh, cs->vp, cs->exi)) { 6705 rfs4_deleg_state_rele(dsp); 6706 *cs->statusp = resp->status = puterrno4(error); 6707 return; 6708 } 6709 6710 /* Mark progress for delegation returns */ 6711 dsp->finfo->dinfo->time_lastwrite = gethrestime_sec(); 6712 rfs4_deleg_state_rele(dsp); 6713 rfs4_do_open(cs, req, oo, DELEG_NONE, 6714 args->share_access, args->share_deny, resp, 1); 6715 } 6716 6717 /*ARGSUSED*/ 6718 static void 6719 rfs4_do_opendelprev(struct compound_state *cs, struct svc_req *req, 6720 OPEN4args *args, rfs4_openowner_t *oo, OPEN4res *resp) 6721 { 6722 /* 6723 * Lookup the pathname, it must already exist since this file 6724 * was delegated. 6725 * 6726 * Find the file and state info for this vp and open owner pair. 6727 * check that they are in fact delegated. 6728 * check that the state access and deny modes are the same. 6729 * 6730 * Return the delgation possibly seting the recall flag. 6731 */ 6732 rfs4_file_t *file; 6733 rfs4_state_t *state; 6734 bool_t create = FALSE; 6735 bool_t dcreate = FALSE; 6736 rfs4_deleg_state_t *dsp; 6737 nfsace4 *ace; 6738 6739 6740 /* Note we ignore oflags */ 6741 resp->status = rfs4_lookupfile(&args->open_claim4_u.file_delegate_prev, 6742 req, cs, args->share_access, &resp->cinfo); 6743 6744 if (resp->status != NFS4_OK) { 6745 return; 6746 } 6747 6748 /* get the file struct and hold a lock on it during initial open */ 6749 file = rfs4_findfile_withlock(cs->vp, NULL, &create); 6750 if (file == NULL) { 6751 NFS4_DEBUG(rfs4_debug, 6752 (CE_NOTE, "rfs4_do_opendelprev: can't find file")); 6753 resp->status = NFS4ERR_SERVERFAULT; 6754 return; 6755 } 6756 6757 state = rfs4_findstate_by_owner_file(oo, file, &create); 6758 if (state == NULL) { 6759 NFS4_DEBUG(rfs4_debug, 6760 (CE_NOTE, "rfs4_do_opendelprev: can't find state")); 6761 resp->status = NFS4ERR_SERVERFAULT; 6762 rfs4_file_rele_withunlock(file); 6763 return; 6764 } 6765 6766 rfs4_dbe_lock(state->dbe); 6767 rfs4_dbe_lock(file->dbe); 6768 if (args->share_access != state->share_access || 6769 args->share_deny != state->share_deny || 6770 state->finfo->dinfo->dtype == OPEN_DELEGATE_NONE) { 6771 NFS4_DEBUG(rfs4_debug, 6772 (CE_NOTE, "rfs4_do_opendelprev: state mixup")); 6773 rfs4_dbe_unlock(file->dbe); 6774 rfs4_dbe_unlock(state->dbe); 6775 rfs4_file_rele(file); 6776 rfs4_state_rele(state); 6777 resp->status = NFS4ERR_SERVERFAULT; 6778 return; 6779 } 6780 rfs4_dbe_unlock(file->dbe); 6781 rfs4_dbe_unlock(state->dbe); 6782 6783 dsp = rfs4_finddeleg(state, &dcreate); 6784 if (dsp == NULL) { 6785 rfs4_state_rele(state); 6786 rfs4_file_rele(file); 6787 resp->status = NFS4ERR_SERVERFAULT; 6788 return; 6789 } 6790 6791 next_stateid(&state->stateid); 6792 6793 resp->stateid = state->stateid.stateid; 6794 6795 resp->delegation.delegation_type = dsp->dtype; 6796 6797 if (dsp->dtype == OPEN_DELEGATE_READ) { 6798 open_read_delegation4 *rv = 6799 &resp->delegation.open_delegation4_u.read; 6800 6801 rv->stateid = dsp->delegid.stateid; 6802 rv->recall = FALSE; /* no policy in place to set to TRUE */ 6803 ace = &rv->permissions; 6804 } else { 6805 open_write_delegation4 *rv = 6806 &resp->delegation.open_delegation4_u.write; 6807 6808 rv->stateid = dsp->delegid.stateid; 6809 rv->recall = FALSE; /* no policy in place to set to TRUE */ 6810 ace = &rv->permissions; 6811 rv->space_limit.limitby = NFS_LIMIT_SIZE; 6812 rv->space_limit.nfs_space_limit4_u.filesize = UINT64_MAX; 6813 } 6814 6815 /* XXX For now */ 6816 ace->type = ACE4_ACCESS_ALLOWED_ACE_TYPE; 6817 ace->flag = 0; 6818 ace->access_mask = 0; 6819 ace->who.utf8string_len = 0; 6820 ace->who.utf8string_val = 0; 6821 6822 rfs4_deleg_state_rele(dsp); 6823 rfs4_state_rele(state); 6824 rfs4_file_rele(file); 6825 } 6826 6827 typedef enum { 6828 NFS4_CHKSEQ_OKAY = 0, 6829 NFS4_CHKSEQ_REPLAY = 1, 6830 NFS4_CHKSEQ_BAD = 2 6831 } rfs4_chkseq_t; 6832 6833 /* 6834 * Generic function for sequence number checks. 6835 */ 6836 static rfs4_chkseq_t 6837 rfs4_check_seqid(seqid4 seqid, nfs_resop4 *lastop, 6838 seqid4 rqst_seq, nfs_resop4 *resop, bool_t copyres) 6839 { 6840 /* Same sequence ids and matching operations? */ 6841 if (seqid == rqst_seq && resop->resop == lastop->resop) { 6842 if (copyres == TRUE) { 6843 rfs4_free_reply(resop); 6844 rfs4_copy_reply(resop, lastop); 6845 } 6846 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 6847 "Replayed SEQID %d\n", seqid)); 6848 return (NFS4_CHKSEQ_REPLAY); 6849 } 6850 6851 /* If the incoming sequence is not the next expected then it is bad */ 6852 if (rqst_seq != seqid + 1) { 6853 if (rqst_seq == seqid) { 6854 NFS4_DEBUG(rfs4_debug, 6855 (CE_NOTE, "BAD SEQID: Replayed sequence id " 6856 "but last op was %d current op is %d\n", 6857 lastop->resop, resop->resop)); 6858 return (NFS4_CHKSEQ_BAD); 6859 } 6860 NFS4_DEBUG(rfs4_debug, 6861 (CE_NOTE, "BAD SEQID: got %u expecting %u\n", 6862 rqst_seq, seqid)); 6863 return (NFS4_CHKSEQ_BAD); 6864 } 6865 6866 /* Everything okay -- next expected */ 6867 return (NFS4_CHKSEQ_OKAY); 6868 } 6869 6870 6871 static rfs4_chkseq_t 6872 rfs4_check_open_seqid(seqid4 seqid, rfs4_openowner_t *op, nfs_resop4 *resop) 6873 { 6874 rfs4_chkseq_t rc; 6875 6876 rfs4_dbe_lock(op->dbe); 6877 rc = rfs4_check_seqid(op->open_seqid, op->reply, seqid, resop, TRUE); 6878 rfs4_dbe_unlock(op->dbe); 6879 6880 if (rc == NFS4_CHKSEQ_OKAY) 6881 rfs4_update_lease(op->client); 6882 6883 return (rc); 6884 } 6885 6886 static rfs4_chkseq_t 6887 rfs4_check_olo_seqid(seqid4 olo_seqid, rfs4_openowner_t *op, 6888 nfs_resop4 *resop) 6889 { 6890 rfs4_chkseq_t rc; 6891 6892 rfs4_dbe_lock(op->dbe); 6893 rc = rfs4_check_seqid(op->open_seqid, op->reply, 6894 olo_seqid, resop, FALSE); 6895 rfs4_dbe_unlock(op->dbe); 6896 6897 return (rc); 6898 } 6899 6900 static rfs4_chkseq_t 6901 rfs4_check_lock_seqid(seqid4 seqid, rfs4_lo_state_t *lp, nfs_resop4 *resop) 6902 { 6903 rfs4_chkseq_t rc = NFS4_CHKSEQ_OKAY; 6904 6905 rfs4_dbe_lock(lp->dbe); 6906 if (!lp->skip_seqid_check) 6907 rc = rfs4_check_seqid(lp->seqid, lp->reply, seqid, resop, TRUE); 6908 rfs4_dbe_unlock(lp->dbe); 6909 6910 return (rc); 6911 } 6912 6913 static void 6914 rfs4_op_open(nfs_argop4 *argop, nfs_resop4 *resop, 6915 struct svc_req *req, struct compound_state *cs) 6916 { 6917 OPEN4args *args = &argop->nfs_argop4_u.opopen; 6918 OPEN4res *resp = &resop->nfs_resop4_u.opopen; 6919 open_owner4 *owner = &args->owner; 6920 open_claim_type4 claim = args->claim; 6921 rfs4_client_t *cp; 6922 rfs4_openowner_t *oo; 6923 bool_t create; 6924 bool_t replay = FALSE; 6925 int can_reclaim; 6926 6927 DTRACE_NFSV4_2(op__open__start, struct compound_state *, cs, 6928 OPEN4args *, args); 6929 6930 if (cs->vp == NULL) { 6931 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 6932 goto end; 6933 } 6934 6935 /* 6936 * Need to check clientid and lease expiration first based on 6937 * error ordering and incrementing sequence id. 6938 */ 6939 cp = rfs4_findclient_by_id(owner->clientid, FALSE); 6940 if (cp == NULL) { 6941 *cs->statusp = resp->status = 6942 rfs4_check_clientid(&owner->clientid, 0); 6943 goto end; 6944 } 6945 6946 if (rfs4_lease_expired(cp)) { 6947 rfs4_client_close(cp); 6948 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 6949 goto end; 6950 } 6951 can_reclaim = cp->can_reclaim; 6952 6953 /* 6954 * Find the open_owner for use from this point forward. Take 6955 * care in updating the sequence id based on the type of error 6956 * being returned. 6957 */ 6958 retry: 6959 create = TRUE; 6960 oo = rfs4_findopenowner(owner, &create, args->seqid); 6961 if (oo == NULL) { 6962 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 6963 rfs4_client_rele(cp); 6964 goto end; 6965 } 6966 6967 /* Hold off access to the sequence space while the open is done */ 6968 rfs4_sw_enter(&oo->oo_sw); 6969 6970 /* 6971 * If the open_owner existed before at the server, then check 6972 * the sequence id. 6973 */ 6974 if (!create && !oo->postpone_confirm) { 6975 switch (rfs4_check_open_seqid(args->seqid, oo, resop)) { 6976 case NFS4_CHKSEQ_BAD: 6977 if ((args->seqid > oo->open_seqid) && 6978 oo->need_confirm) { 6979 rfs4_free_opens(oo, TRUE, FALSE); 6980 rfs4_sw_exit(&oo->oo_sw); 6981 rfs4_openowner_rele(oo); 6982 goto retry; 6983 } 6984 resp->status = NFS4ERR_BAD_SEQID; 6985 goto out; 6986 case NFS4_CHKSEQ_REPLAY: /* replay of previous request */ 6987 replay = TRUE; 6988 goto out; 6989 default: 6990 break; 6991 } 6992 6993 /* 6994 * Sequence was ok and open owner exists 6995 * check to see if we have yet to see an 6996 * open_confirm. 6997 */ 6998 if (oo->need_confirm) { 6999 rfs4_free_opens(oo, TRUE, FALSE); 7000 rfs4_sw_exit(&oo->oo_sw); 7001 rfs4_openowner_rele(oo); 7002 goto retry; 7003 } 7004 } 7005 /* Grace only applies to regular-type OPENs */ 7006 if (rfs4_clnt_in_grace(cp) && 7007 (claim == CLAIM_NULL || claim == CLAIM_DELEGATE_CUR)) { 7008 *cs->statusp = resp->status = NFS4ERR_GRACE; 7009 goto out; 7010 } 7011 7012 /* 7013 * If previous state at the server existed then can_reclaim 7014 * will be set. If not reply NFS4ERR_NO_GRACE to the 7015 * client. 7016 */ 7017 if (rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS && !can_reclaim) { 7018 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7019 goto out; 7020 } 7021 7022 7023 /* 7024 * Reject the open if the client has missed the grace period 7025 */ 7026 if (!rfs4_clnt_in_grace(cp) && claim == CLAIM_PREVIOUS) { 7027 *cs->statusp = resp->status = NFS4ERR_NO_GRACE; 7028 goto out; 7029 } 7030 7031 /* Couple of up-front bookkeeping items */ 7032 if (oo->need_confirm) { 7033 /* 7034 * If this is a reclaim OPEN then we should not ask 7035 * for a confirmation of the open_owner per the 7036 * protocol specification. 7037 */ 7038 if (claim == CLAIM_PREVIOUS) 7039 oo->need_confirm = FALSE; 7040 else 7041 resp->rflags |= OPEN4_RESULT_CONFIRM; 7042 } 7043 resp->rflags |= OPEN4_RESULT_LOCKTYPE_POSIX; 7044 7045 /* 7046 * If there is an unshared filesystem mounted on this vnode, 7047 * do not allow to open/create in this directory. 7048 */ 7049 if (vn_ismntpt(cs->vp)) { 7050 *cs->statusp = resp->status = NFS4ERR_ACCESS; 7051 goto out; 7052 } 7053 7054 /* 7055 * access must READ, WRITE, or BOTH. No access is invalid. 7056 * deny can be READ, WRITE, BOTH, or NONE. 7057 * bits not defined for access/deny are invalid. 7058 */ 7059 if (! (args->share_access & OPEN4_SHARE_ACCESS_BOTH) || 7060 (args->share_access & ~OPEN4_SHARE_ACCESS_BOTH) || 7061 (args->share_deny & ~OPEN4_SHARE_DENY_BOTH)) { 7062 *cs->statusp = resp->status = NFS4ERR_INVAL; 7063 goto out; 7064 } 7065 7066 7067 /* 7068 * make sure attrset is zero before response is built. 7069 */ 7070 resp->attrset = 0; 7071 7072 switch (claim) { 7073 case CLAIM_NULL: 7074 rfs4_do_opennull(cs, req, args, oo, resp); 7075 break; 7076 case CLAIM_PREVIOUS: 7077 rfs4_do_openprev(cs, req, args, oo, resp); 7078 break; 7079 case CLAIM_DELEGATE_CUR: 7080 rfs4_do_opendelcur(cs, req, args, oo, resp); 7081 break; 7082 case CLAIM_DELEGATE_PREV: 7083 rfs4_do_opendelprev(cs, req, args, oo, resp); 7084 break; 7085 default: 7086 resp->status = NFS4ERR_INVAL; 7087 break; 7088 } 7089 7090 out: 7091 rfs4_client_rele(cp); 7092 7093 /* Catch sequence id handling here to make it a little easier */ 7094 switch (resp->status) { 7095 case NFS4ERR_BADXDR: 7096 case NFS4ERR_BAD_SEQID: 7097 case NFS4ERR_BAD_STATEID: 7098 case NFS4ERR_NOFILEHANDLE: 7099 case NFS4ERR_RESOURCE: 7100 case NFS4ERR_STALE_CLIENTID: 7101 case NFS4ERR_STALE_STATEID: 7102 /* 7103 * The protocol states that if any of these errors are 7104 * being returned, the sequence id should not be 7105 * incremented. Any other return requires an 7106 * increment. 7107 */ 7108 break; 7109 default: 7110 /* Always update the lease in this case */ 7111 rfs4_update_lease(oo->client); 7112 7113 /* Regular response - copy the result */ 7114 if (!replay) 7115 rfs4_update_open_resp(oo, resop, &cs->fh); 7116 7117 /* 7118 * REPLAY case: Only if the previous response was OK 7119 * do we copy the filehandle. If not OK, no 7120 * filehandle to copy. 7121 */ 7122 if (replay == TRUE && 7123 resp->status == NFS4_OK && 7124 oo->reply_fh.nfs_fh4_val) { 7125 /* 7126 * If this is a replay, we must restore the 7127 * current filehandle/vp to that of what was 7128 * returned originally. Try our best to do 7129 * it. 7130 */ 7131 nfs_fh4_fmt_t *fh_fmtp = 7132 (nfs_fh4_fmt_t *)oo->reply_fh.nfs_fh4_val; 7133 7134 cs->exi = checkexport4(&fh_fmtp->fh4_fsid, 7135 (fid_t *)&fh_fmtp->fh4_xlen, NULL); 7136 7137 if (cs->exi == NULL) { 7138 resp->status = NFS4ERR_STALE; 7139 goto finish; 7140 } 7141 7142 VN_RELE(cs->vp); 7143 7144 cs->vp = nfs4_fhtovp(&oo->reply_fh, cs->exi, 7145 &resp->status); 7146 7147 if (cs->vp == NULL) 7148 goto finish; 7149 7150 nfs_fh4_copy(&oo->reply_fh, &cs->fh); 7151 } 7152 7153 /* 7154 * If this was a replay, no need to update the 7155 * sequence id. If the open_owner was not created on 7156 * this pass, then update. The first use of an 7157 * open_owner will not bump the sequence id. 7158 */ 7159 if (replay == FALSE && !create) 7160 rfs4_update_open_sequence(oo); 7161 /* 7162 * If the client is receiving an error and the 7163 * open_owner needs to be confirmed, there is no way 7164 * to notify the client of this fact ignoring the fact 7165 * that the server has no method of returning a 7166 * stateid to confirm. Therefore, the server needs to 7167 * mark this open_owner in a way as to avoid the 7168 * sequence id checking the next time the client uses 7169 * this open_owner. 7170 */ 7171 if (resp->status != NFS4_OK && oo->need_confirm) 7172 oo->postpone_confirm = TRUE; 7173 /* 7174 * If OK response then clear the postpone flag and 7175 * reset the sequence id to keep in sync with the 7176 * client. 7177 */ 7178 if (resp->status == NFS4_OK && oo->postpone_confirm) { 7179 oo->postpone_confirm = FALSE; 7180 oo->open_seqid = args->seqid; 7181 } 7182 break; 7183 } 7184 7185 finish: 7186 *cs->statusp = resp->status; 7187 7188 rfs4_sw_exit(&oo->oo_sw); 7189 rfs4_openowner_rele(oo); 7190 7191 end: 7192 DTRACE_NFSV4_2(op__open__done, struct compound_state *, cs, 7193 OPEN4res *, resp); 7194 } 7195 7196 /*ARGSUSED*/ 7197 void 7198 rfs4_op_open_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 7199 struct svc_req *req, struct compound_state *cs) 7200 { 7201 OPEN_CONFIRM4args *args = &argop->nfs_argop4_u.opopen_confirm; 7202 OPEN_CONFIRM4res *resp = &resop->nfs_resop4_u.opopen_confirm; 7203 rfs4_state_t *sp; 7204 nfsstat4 status; 7205 7206 DTRACE_NFSV4_2(op__open__confirm__start, struct compound_state *, cs, 7207 OPEN_CONFIRM4args *, args); 7208 7209 if (cs->vp == NULL) { 7210 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7211 goto out; 7212 } 7213 7214 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7215 if (status != NFS4_OK) { 7216 *cs->statusp = resp->status = status; 7217 goto out; 7218 } 7219 7220 /* Ensure specified filehandle matches */ 7221 if (cs->vp != sp->finfo->vp) { 7222 rfs4_state_rele(sp); 7223 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7224 goto out; 7225 } 7226 7227 /* hold off other access to open_owner while we tinker */ 7228 rfs4_sw_enter(&sp->owner->oo_sw); 7229 7230 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7231 case NFS4_CHECK_STATEID_OKAY: 7232 if (rfs4_check_open_seqid(args->seqid, sp->owner, 7233 resop) != 0) { 7234 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7235 break; 7236 } 7237 /* 7238 * If it is the appropriate stateid and determined to 7239 * be "OKAY" then this means that the stateid does not 7240 * need to be confirmed and the client is in error for 7241 * sending an OPEN_CONFIRM. 7242 */ 7243 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7244 break; 7245 case NFS4_CHECK_STATEID_OLD: 7246 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7247 break; 7248 case NFS4_CHECK_STATEID_BAD: 7249 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7250 break; 7251 case NFS4_CHECK_STATEID_EXPIRED: 7252 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7253 break; 7254 case NFS4_CHECK_STATEID_CLOSED: 7255 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7256 break; 7257 case NFS4_CHECK_STATEID_REPLAY: 7258 switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) { 7259 case NFS4_CHKSEQ_OKAY: 7260 /* 7261 * This is replayed stateid; if seqid matches 7262 * next expected, then client is using wrong seqid. 7263 */ 7264 /* fall through */ 7265 case NFS4_CHKSEQ_BAD: 7266 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7267 break; 7268 case NFS4_CHKSEQ_REPLAY: 7269 /* 7270 * Note this case is the duplicate case so 7271 * resp->status is already set. 7272 */ 7273 *cs->statusp = resp->status; 7274 rfs4_update_lease(sp->owner->client); 7275 break; 7276 } 7277 break; 7278 case NFS4_CHECK_STATEID_UNCONFIRMED: 7279 if (rfs4_check_open_seqid(args->seqid, sp->owner, 7280 resop) != NFS4_CHKSEQ_OKAY) { 7281 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7282 break; 7283 } 7284 *cs->statusp = resp->status = NFS4_OK; 7285 7286 next_stateid(&sp->stateid); 7287 resp->open_stateid = sp->stateid.stateid; 7288 sp->owner->need_confirm = FALSE; 7289 rfs4_update_lease(sp->owner->client); 7290 rfs4_update_open_sequence(sp->owner); 7291 rfs4_update_open_resp(sp->owner, resop, NULL); 7292 break; 7293 default: 7294 ASSERT(FALSE); 7295 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7296 break; 7297 } 7298 rfs4_sw_exit(&sp->owner->oo_sw); 7299 rfs4_state_rele(sp); 7300 7301 out: 7302 DTRACE_NFSV4_2(op__open__confirm__done, struct compound_state *, cs, 7303 OPEN_CONFIRM4res *, resp); 7304 } 7305 7306 /*ARGSUSED*/ 7307 void 7308 rfs4_op_open_downgrade(nfs_argop4 *argop, nfs_resop4 *resop, 7309 struct svc_req *req, struct compound_state *cs) 7310 { 7311 OPEN_DOWNGRADE4args *args = &argop->nfs_argop4_u.opopen_downgrade; 7312 OPEN_DOWNGRADE4res *resp = &resop->nfs_resop4_u.opopen_downgrade; 7313 uint32_t access = args->share_access; 7314 uint32_t deny = args->share_deny; 7315 nfsstat4 status; 7316 rfs4_state_t *sp; 7317 rfs4_file_t *fp; 7318 int fflags = 0; 7319 7320 DTRACE_NFSV4_2(op__open__downgrade__start, struct compound_state *, cs, 7321 OPEN_DOWNGRADE4args *, args); 7322 7323 if (cs->vp == NULL) { 7324 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7325 goto out; 7326 } 7327 7328 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_VALID); 7329 if (status != NFS4_OK) { 7330 *cs->statusp = resp->status = status; 7331 goto out; 7332 } 7333 7334 /* Ensure specified filehandle matches */ 7335 if (cs->vp != sp->finfo->vp) { 7336 rfs4_state_rele(sp); 7337 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7338 goto out; 7339 } 7340 7341 /* hold off other access to open_owner while we tinker */ 7342 rfs4_sw_enter(&sp->owner->oo_sw); 7343 7344 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7345 case NFS4_CHECK_STATEID_OKAY: 7346 if (rfs4_check_open_seqid(args->seqid, sp->owner, 7347 resop) != NFS4_CHKSEQ_OKAY) { 7348 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7349 goto end; 7350 } 7351 break; 7352 case NFS4_CHECK_STATEID_OLD: 7353 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7354 goto end; 7355 case NFS4_CHECK_STATEID_BAD: 7356 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7357 goto end; 7358 case NFS4_CHECK_STATEID_EXPIRED: 7359 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7360 goto end; 7361 case NFS4_CHECK_STATEID_CLOSED: 7362 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7363 goto end; 7364 case NFS4_CHECK_STATEID_UNCONFIRMED: 7365 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7366 goto end; 7367 case NFS4_CHECK_STATEID_REPLAY: 7368 /* Check the sequence id for the open owner */ 7369 switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) { 7370 case NFS4_CHKSEQ_OKAY: 7371 /* 7372 * This is replayed stateid; if seqid matches 7373 * next expected, then client is using wrong seqid. 7374 */ 7375 /* fall through */ 7376 case NFS4_CHKSEQ_BAD: 7377 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7378 goto end; 7379 case NFS4_CHKSEQ_REPLAY: 7380 /* 7381 * Note this case is the duplicate case so 7382 * resp->status is already set. 7383 */ 7384 *cs->statusp = resp->status; 7385 rfs4_update_lease(sp->owner->client); 7386 goto end; 7387 } 7388 break; 7389 default: 7390 ASSERT(FALSE); 7391 break; 7392 } 7393 7394 rfs4_dbe_lock(sp->dbe); 7395 /* 7396 * Check that the new access modes and deny modes are valid. 7397 * Check that no invalid bits are set. 7398 */ 7399 if ((access & ~(OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) || 7400 (deny & ~(OPEN4_SHARE_DENY_READ | OPEN4_SHARE_DENY_WRITE))) { 7401 *cs->statusp = resp->status = NFS4ERR_INVAL; 7402 rfs4_update_open_sequence(sp->owner); 7403 rfs4_dbe_unlock(sp->dbe); 7404 goto end; 7405 } 7406 7407 /* 7408 * The new modes must be a subset of the current modes and 7409 * the access must specify at least one mode. To test that 7410 * the new mode is a subset of the current modes we bitwise 7411 * AND them together and check that the result equals the new 7412 * mode. For example: 7413 * New mode, access == R and current mode, sp->share_access == RW 7414 * access & sp->share_access == R == access, so the new access mode 7415 * is valid. Consider access == RW, sp->share_access = R 7416 * access & sp->share_access == R != access, so the new access mode 7417 * is invalid. 7418 */ 7419 if ((access & sp->share_access) != access || 7420 (deny & sp->share_deny) != deny || 7421 (access & 7422 (OPEN4_SHARE_ACCESS_READ | OPEN4_SHARE_ACCESS_WRITE)) == 0) { 7423 *cs->statusp = resp->status = NFS4ERR_INVAL; 7424 rfs4_update_open_sequence(sp->owner); 7425 rfs4_dbe_unlock(sp->dbe); 7426 goto end; 7427 } 7428 7429 /* 7430 * Release any share locks associated with this stateID. 7431 * Strictly speaking, this violates the spec because the 7432 * spec effectively requires that open downgrade be atomic. 7433 * At present, fs_shrlock does not have this capability. 7434 */ 7435 rfs4_dbe_unlock(sp->dbe); 7436 rfs4_unshare(sp); 7437 rfs4_dbe_lock(sp->dbe); 7438 7439 fp = sp->finfo; 7440 rfs4_dbe_lock(fp->dbe); 7441 7442 /* 7443 * If the current mode has deny read and the new mode 7444 * does not, decrement the number of deny read mode bits 7445 * and if it goes to zero turn off the deny read bit 7446 * on the file. 7447 */ 7448 if ((sp->share_deny & OPEN4_SHARE_DENY_READ) && 7449 (deny & OPEN4_SHARE_DENY_READ) == 0) { 7450 fp->deny_read--; 7451 if (fp->deny_read == 0) 7452 fp->share_deny &= ~OPEN4_SHARE_DENY_READ; 7453 } 7454 7455 /* 7456 * If the current mode has deny write and the new mode 7457 * does not, decrement the number of deny write mode bits 7458 * and if it goes to zero turn off the deny write bit 7459 * on the file. 7460 */ 7461 if ((sp->share_deny & OPEN4_SHARE_DENY_WRITE) && 7462 (deny & OPEN4_SHARE_DENY_WRITE) == 0) { 7463 fp->deny_write--; 7464 if (fp->deny_write == 0) 7465 fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE; 7466 } 7467 7468 /* 7469 * If the current mode has access read and the new mode 7470 * does not, decrement the number of access read mode bits 7471 * and if it goes to zero turn off the access read bit 7472 * on the file. set fflags to FREAD for the call to 7473 * vn_open_downgrade(). 7474 */ 7475 if ((sp->share_access & OPEN4_SHARE_ACCESS_READ) && 7476 (access & OPEN4_SHARE_ACCESS_READ) == 0) { 7477 fp->access_read--; 7478 if (fp->access_read == 0) 7479 fp->share_access &= ~OPEN4_SHARE_ACCESS_READ; 7480 fflags |= FREAD; 7481 } 7482 7483 /* 7484 * If the current mode has access write and the new mode 7485 * does not, decrement the number of access write mode bits 7486 * and if it goes to zero turn off the access write bit 7487 * on the file. set fflags to FWRITE for the call to 7488 * vn_open_downgrade(). 7489 */ 7490 if ((sp->share_access & OPEN4_SHARE_ACCESS_WRITE) && 7491 (access & OPEN4_SHARE_ACCESS_WRITE) == 0) { 7492 fp->access_write--; 7493 if (fp->access_write == 0) 7494 fp->share_deny &= ~OPEN4_SHARE_ACCESS_WRITE; 7495 fflags |= FWRITE; 7496 } 7497 7498 /* Set the new access and deny modes */ 7499 sp->share_access = access; 7500 sp->share_deny = deny; 7501 /* Check that the file is still accessible */ 7502 ASSERT(fp->share_access); 7503 7504 rfs4_dbe_unlock(fp->dbe); 7505 7506 rfs4_dbe_unlock(sp->dbe); 7507 7508 if ((status = rfs4_share(sp)) != NFS4_OK) { 7509 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 7510 rfs4_update_open_sequence(sp->owner); 7511 goto end; 7512 } 7513 7514 /* 7515 * we successfully downgraded the share lock, now we need to downgrade 7516 * the open. it is possible that the downgrade was only for a deny 7517 * mode and we have nothing else to do. 7518 */ 7519 if ((fflags & (FREAD|FWRITE)) != 0) 7520 vn_open_downgrade(cs->vp, fflags); 7521 7522 rfs4_dbe_lock(sp->dbe); 7523 7524 /* Update the stateid */ 7525 next_stateid(&sp->stateid); 7526 resp->open_stateid = sp->stateid.stateid; 7527 7528 rfs4_dbe_unlock(sp->dbe); 7529 7530 *cs->statusp = resp->status = NFS4_OK; 7531 /* Update the lease */ 7532 rfs4_update_lease(sp->owner->client); 7533 /* And the sequence */ 7534 rfs4_update_open_sequence(sp->owner); 7535 rfs4_update_open_resp(sp->owner, resop, NULL); 7536 7537 end: 7538 rfs4_sw_exit(&sp->owner->oo_sw); 7539 rfs4_state_rele(sp); 7540 out: 7541 DTRACE_NFSV4_2(op__open__downgrade__done, struct compound_state *, cs, 7542 OPEN_DOWNGRADE4res *, resp); 7543 } 7544 7545 /* 7546 * The logic behind this function is detailed in the NFSv4 RFC in the 7547 * SETCLIENTID operation description under IMPLEMENTATION. Refer to 7548 * that section for explicit guidance to server behavior for 7549 * SETCLIENTID. 7550 */ 7551 void 7552 rfs4_op_setclientid(nfs_argop4 *argop, nfs_resop4 *resop, 7553 struct svc_req *req, struct compound_state *cs) 7554 { 7555 SETCLIENTID4args *args = &argop->nfs_argop4_u.opsetclientid; 7556 SETCLIENTID4res *res = &resop->nfs_resop4_u.opsetclientid; 7557 rfs4_client_t *cp, *newcp, *cp_confirmed, *cp_unconfirmed; 7558 bool_t create = TRUE; 7559 char *addr, *netid; 7560 int len; 7561 7562 DTRACE_NFSV4_2(op__setclientid__start, struct compound_state *, cs, 7563 SETCLIENTID4args *, args); 7564 retry: 7565 newcp = cp_confirmed = cp_unconfirmed = NULL; 7566 7567 /* 7568 * In search of an EXISTING client matching the incoming 7569 * request to establish a new client identifier at the server 7570 */ 7571 create = TRUE; 7572 cp = rfs4_findclient(&args->client, &create, NULL); 7573 7574 /* Should never happen */ 7575 ASSERT(cp != NULL); 7576 7577 if (cp == NULL) { 7578 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 7579 goto out; 7580 } 7581 7582 /* 7583 * Easiest case. Client identifier is newly created and is 7584 * unconfirmed. Also note that for this case, no other 7585 * entries exist for the client identifier. Nothing else to 7586 * check. Just setup the response and respond. 7587 */ 7588 if (create) { 7589 *cs->statusp = res->status = NFS4_OK; 7590 res->SETCLIENTID4res_u.resok4.clientid = cp->clientid; 7591 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 7592 cp->confirm_verf; 7593 /* Setup callback information; CB_NULL confirmation later */ 7594 rfs4_client_setcb(cp, &args->callback, args->callback_ident); 7595 7596 rfs4_client_rele(cp); 7597 goto out; 7598 } 7599 7600 /* 7601 * An existing, confirmed client may exist but it may not have 7602 * been active for at least one lease period. If so, then 7603 * "close" the client and create a new client identifier 7604 */ 7605 if (rfs4_lease_expired(cp)) { 7606 rfs4_client_close(cp); 7607 goto retry; 7608 } 7609 7610 if (cp->need_confirm == TRUE) 7611 cp_unconfirmed = cp; 7612 else 7613 cp_confirmed = cp; 7614 7615 cp = NULL; 7616 7617 /* 7618 * We have a confirmed client, now check for an 7619 * unconfimred entry 7620 */ 7621 if (cp_confirmed) { 7622 /* If creds don't match then client identifier is inuse */ 7623 if (!creds_ok(cp_confirmed->cr_set, req, cs)) { 7624 rfs4_cbinfo_t *cbp; 7625 /* 7626 * Some one else has established this client 7627 * id. Try and say * who they are. We will use 7628 * the call back address supplied by * the 7629 * first client. 7630 */ 7631 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 7632 7633 addr = netid = NULL; 7634 7635 cbp = &cp_confirmed->cbinfo; 7636 if (cbp->cb_callback.cb_location.r_addr && 7637 cbp->cb_callback.cb_location.r_netid) { 7638 cb_client4 *cbcp = &cbp->cb_callback; 7639 7640 len = strlen(cbcp->cb_location.r_addr)+1; 7641 addr = kmem_alloc(len, KM_SLEEP); 7642 bcopy(cbcp->cb_location.r_addr, addr, len); 7643 len = strlen(cbcp->cb_location.r_netid)+1; 7644 netid = kmem_alloc(len, KM_SLEEP); 7645 bcopy(cbcp->cb_location.r_netid, netid, len); 7646 } 7647 7648 res->SETCLIENTID4res_u.client_using.r_addr = addr; 7649 res->SETCLIENTID4res_u.client_using.r_netid = netid; 7650 7651 rfs4_client_rele(cp_confirmed); 7652 } 7653 7654 /* 7655 * Confirmed, creds match, and verifier matches; must 7656 * be an update of the callback info 7657 */ 7658 if (cp_confirmed->nfs_client.verifier == 7659 args->client.verifier) { 7660 /* Setup callback information */ 7661 rfs4_client_setcb(cp_confirmed, &args->callback, 7662 args->callback_ident); 7663 7664 /* everything okay -- move ahead */ 7665 *cs->statusp = res->status = NFS4_OK; 7666 res->SETCLIENTID4res_u.resok4.clientid = 7667 cp_confirmed->clientid; 7668 7669 /* update the confirm_verifier and return it */ 7670 rfs4_client_scv_next(cp_confirmed); 7671 res->SETCLIENTID4res_u.resok4.setclientid_confirm = 7672 cp_confirmed->confirm_verf; 7673 7674 rfs4_client_rele(cp_confirmed); 7675 goto out; 7676 } 7677 7678 /* 7679 * Creds match but the verifier doesn't. Must search 7680 * for an unconfirmed client that would be replaced by 7681 * this request. 7682 */ 7683 create = FALSE; 7684 cp_unconfirmed = rfs4_findclient(&args->client, &create, 7685 cp_confirmed); 7686 } 7687 7688 /* 7689 * At this point, we have taken care of the brand new client 7690 * struct, INUSE case, update of an existing, and confirmed 7691 * client struct. 7692 */ 7693 7694 /* 7695 * check to see if things have changed while we originally 7696 * picked up the client struct. If they have, then return and 7697 * retry the processing of this SETCLIENTID request. 7698 */ 7699 if (cp_unconfirmed) { 7700 rfs4_dbe_lock(cp_unconfirmed->dbe); 7701 if (!cp_unconfirmed->need_confirm) { 7702 rfs4_dbe_unlock(cp_unconfirmed->dbe); 7703 rfs4_client_rele(cp_unconfirmed); 7704 if (cp_confirmed) 7705 rfs4_client_rele(cp_confirmed); 7706 goto retry; 7707 } 7708 /* do away with the old unconfirmed one */ 7709 rfs4_dbe_invalidate(cp_unconfirmed->dbe); 7710 rfs4_dbe_unlock(cp_unconfirmed->dbe); 7711 rfs4_client_rele(cp_unconfirmed); 7712 cp_unconfirmed = NULL; 7713 } 7714 7715 /* 7716 * This search will temporarily hide the confirmed client 7717 * struct while a new client struct is created as the 7718 * unconfirmed one. 7719 */ 7720 create = TRUE; 7721 newcp = rfs4_findclient(&args->client, &create, cp_confirmed); 7722 7723 ASSERT(newcp != NULL); 7724 7725 if (newcp == NULL) { 7726 *cs->statusp = res->status = NFS4ERR_SERVERFAULT; 7727 rfs4_client_rele(cp_confirmed); 7728 goto out; 7729 } 7730 7731 /* 7732 * If one was not created, then a similar request must be in 7733 * process so release and start over with this one 7734 */ 7735 if (create != TRUE) { 7736 rfs4_client_rele(newcp); 7737 if (cp_confirmed) 7738 rfs4_client_rele(cp_confirmed); 7739 goto retry; 7740 } 7741 7742 *cs->statusp = res->status = NFS4_OK; 7743 res->SETCLIENTID4res_u.resok4.clientid = newcp->clientid; 7744 res->SETCLIENTID4res_u.resok4.setclientid_confirm = newcp->confirm_verf; 7745 /* Setup callback information; CB_NULL confirmation later */ 7746 rfs4_client_setcb(newcp, &args->callback, args->callback_ident); 7747 7748 newcp->cp_confirmed = cp_confirmed; 7749 7750 rfs4_client_rele(newcp); 7751 7752 out: 7753 DTRACE_NFSV4_2(op__setclientid__done, struct compound_state *, cs, 7754 SETCLIENTID4res *, res); 7755 } 7756 7757 /*ARGSUSED*/ 7758 void 7759 rfs4_op_setclientid_confirm(nfs_argop4 *argop, nfs_resop4 *resop, 7760 struct svc_req *req, struct compound_state *cs) 7761 { 7762 SETCLIENTID_CONFIRM4args *args = 7763 &argop->nfs_argop4_u.opsetclientid_confirm; 7764 SETCLIENTID_CONFIRM4res *res = 7765 &resop->nfs_resop4_u.opsetclientid_confirm; 7766 rfs4_client_t *cp, *cptoclose = NULL; 7767 7768 DTRACE_NFSV4_2(op__setclientid__confirm__start, 7769 struct compound_state *, cs, 7770 SETCLIENTID_CONFIRM4args *, args); 7771 7772 *cs->statusp = res->status = NFS4_OK; 7773 7774 cp = rfs4_findclient_by_id(args->clientid, TRUE); 7775 7776 if (cp == NULL) { 7777 *cs->statusp = res->status = 7778 rfs4_check_clientid(&args->clientid, 1); 7779 goto out; 7780 } 7781 7782 if (!creds_ok(cp, req, cs)) { 7783 *cs->statusp = res->status = NFS4ERR_CLID_INUSE; 7784 rfs4_client_rele(cp); 7785 goto out; 7786 } 7787 7788 /* If the verifier doesn't match, the record doesn't match */ 7789 if (cp->confirm_verf != args->setclientid_confirm) { 7790 *cs->statusp = res->status = NFS4ERR_STALE_CLIENTID; 7791 rfs4_client_rele(cp); 7792 goto out; 7793 } 7794 7795 rfs4_dbe_lock(cp->dbe); 7796 cp->need_confirm = FALSE; 7797 if (cp->cp_confirmed) { 7798 cptoclose = cp->cp_confirmed; 7799 cptoclose->ss_remove = 1; 7800 cp->cp_confirmed = NULL; 7801 } 7802 7803 /* 7804 * Update the client's associated server instance, if it's changed 7805 * since the client was created. 7806 */ 7807 if (rfs4_servinst(cp) != rfs4_cur_servinst) 7808 rfs4_servinst_assign(cp, rfs4_cur_servinst); 7809 7810 /* 7811 * Record clientid in stable storage. 7812 * Must be done after server instance has been assigned. 7813 */ 7814 rfs4_ss_clid(cp, req); 7815 7816 rfs4_dbe_unlock(cp->dbe); 7817 7818 if (cptoclose) 7819 /* don't need to rele, client_close does it */ 7820 rfs4_client_close(cptoclose); 7821 7822 /* If needed, initiate CB_NULL call for callback path */ 7823 rfs4_deleg_cb_check(cp); 7824 rfs4_update_lease(cp); 7825 7826 /* 7827 * Check to see if client can perform reclaims 7828 */ 7829 rfs4_ss_chkclid(cp); 7830 7831 rfs4_client_rele(cp); 7832 7833 out: 7834 DTRACE_NFSV4_2(op__setclientid__confirm__done, 7835 struct compound_state *, cs, 7836 SETCLIENTID_CONFIRM4 *, res); 7837 } 7838 7839 7840 /*ARGSUSED*/ 7841 void 7842 rfs4_op_close(nfs_argop4 *argop, nfs_resop4 *resop, 7843 struct svc_req *req, struct compound_state *cs) 7844 { 7845 CLOSE4args *args = &argop->nfs_argop4_u.opclose; 7846 CLOSE4res *resp = &resop->nfs_resop4_u.opclose; 7847 rfs4_state_t *sp; 7848 nfsstat4 status; 7849 7850 DTRACE_NFSV4_2(op__close__start, struct compound_state *, cs, 7851 CLOSE4args *, args); 7852 7853 if (cs->vp == NULL) { 7854 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 7855 goto out; 7856 } 7857 7858 status = rfs4_get_state(&args->open_stateid, &sp, RFS4_DBS_INVALID); 7859 if (status != NFS4_OK) { 7860 *cs->statusp = resp->status = status; 7861 goto out; 7862 } 7863 7864 /* Ensure specified filehandle matches */ 7865 if (cs->vp != sp->finfo->vp) { 7866 rfs4_state_rele(sp); 7867 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7868 goto out; 7869 } 7870 7871 /* hold off other access to open_owner while we tinker */ 7872 rfs4_sw_enter(&sp->owner->oo_sw); 7873 7874 switch (rfs4_check_stateid_seqid(sp, &args->open_stateid)) { 7875 case NFS4_CHECK_STATEID_OKAY: 7876 if (rfs4_check_open_seqid(args->seqid, sp->owner, 7877 resop) != NFS4_CHKSEQ_OKAY) { 7878 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7879 goto end; 7880 } 7881 break; 7882 case NFS4_CHECK_STATEID_OLD: 7883 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7884 goto end; 7885 case NFS4_CHECK_STATEID_BAD: 7886 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7887 goto end; 7888 case NFS4_CHECK_STATEID_EXPIRED: 7889 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 7890 goto end; 7891 case NFS4_CHECK_STATEID_CLOSED: 7892 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 7893 goto end; 7894 case NFS4_CHECK_STATEID_UNCONFIRMED: 7895 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 7896 goto end; 7897 case NFS4_CHECK_STATEID_REPLAY: 7898 /* Check the sequence id for the open owner */ 7899 switch (rfs4_check_open_seqid(args->seqid, sp->owner, resop)) { 7900 case NFS4_CHKSEQ_OKAY: 7901 /* 7902 * This is replayed stateid; if seqid matches 7903 * next expected, then client is using wrong seqid. 7904 */ 7905 /* FALL THROUGH */ 7906 case NFS4_CHKSEQ_BAD: 7907 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 7908 goto end; 7909 case NFS4_CHKSEQ_REPLAY: 7910 /* 7911 * Note this case is the duplicate case so 7912 * resp->status is already set. 7913 */ 7914 *cs->statusp = resp->status; 7915 rfs4_update_lease(sp->owner->client); 7916 goto end; 7917 } 7918 break; 7919 default: 7920 ASSERT(FALSE); 7921 break; 7922 } 7923 7924 rfs4_dbe_lock(sp->dbe); 7925 7926 /* Update the stateid. */ 7927 next_stateid(&sp->stateid); 7928 resp->open_stateid = sp->stateid.stateid; 7929 7930 rfs4_dbe_unlock(sp->dbe); 7931 7932 rfs4_update_lease(sp->owner->client); 7933 rfs4_update_open_sequence(sp->owner); 7934 rfs4_update_open_resp(sp->owner, resop, NULL); 7935 7936 rfs4_state_close(sp, FALSE, FALSE, cs->cr); 7937 7938 *cs->statusp = resp->status = status; 7939 7940 end: 7941 rfs4_sw_exit(&sp->owner->oo_sw); 7942 rfs4_state_rele(sp); 7943 out: 7944 DTRACE_NFSV4_2(op__close__done, struct compound_state *, cs, 7945 CLOSE4res *, resp); 7946 } 7947 7948 /* 7949 * Manage the counts on the file struct and close all file locks 7950 */ 7951 /*ARGSUSED*/ 7952 void 7953 rfs4_release_share_lock_state(rfs4_state_t *sp, cred_t *cr, 7954 bool_t close_of_client) 7955 { 7956 rfs4_file_t *fp = sp->finfo; 7957 rfs4_lo_state_t *lsp; 7958 struct shrlock shr; 7959 struct shr_locowner shr_loco; 7960 int fflags, s_access, s_deny; 7961 7962 fflags = s_access = s_deny = 0; 7963 /* 7964 * Decrement the count for each access and deny bit that this 7965 * state has contributed to the file. If the file counts go to zero 7966 * clear the appropriate bit in the appropriate mask. 7967 */ 7968 7969 if (sp->share_access & OPEN4_SHARE_ACCESS_READ) { 7970 fp->access_read--; 7971 fflags |= FREAD; 7972 s_access |= F_RDACC; 7973 if (fp->access_read == 0) 7974 fp->share_access &= ~OPEN4_SHARE_ACCESS_READ; 7975 } 7976 if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) { 7977 fp->access_write--; 7978 fflags |= FWRITE; 7979 s_access |= F_WRACC; 7980 if (fp->access_write == 0) 7981 fp->share_access &= ~OPEN4_SHARE_ACCESS_WRITE; 7982 } 7983 if (sp->share_deny & OPEN4_SHARE_DENY_READ) { 7984 fp->deny_read--; 7985 s_deny |= F_RDDNY; 7986 if (fp->deny_read == 0) 7987 fp->share_deny &= ~OPEN4_SHARE_DENY_READ; 7988 } 7989 if (sp->share_deny & OPEN4_SHARE_DENY_WRITE) { 7990 fp->deny_write--; 7991 s_deny |= F_WRDNY; 7992 if (fp->deny_write == 0) 7993 fp->share_deny &= ~OPEN4_SHARE_DENY_WRITE; 7994 } 7995 7996 /* 7997 * If this call is part of the larger closing down of client 7998 * state then it is just easier to release all locks 7999 * associated with this client instead of going through each 8000 * individual file and cleaning locks there. 8001 */ 8002 if (close_of_client) { 8003 if (sp->owner->client->unlksys_completed == FALSE && 8004 sp->lockownerlist.next->lsp != NULL && 8005 sp->owner->client->sysidt != LM_NOSYSID) { 8006 /* Is the PxFS kernel module loaded? */ 8007 if (lm_remove_file_locks != NULL) { 8008 int new_sysid; 8009 8010 /* Encode the cluster nodeid in new sysid */ 8011 new_sysid = sp->owner->client->sysidt; 8012 lm_set_nlmid_flk(&new_sysid); 8013 8014 /* 8015 * This PxFS routine removes file locks for a 8016 * client over all nodes of a cluster. 8017 */ 8018 NFS4_DEBUG(rfs4_debug, (CE_NOTE, 8019 "lm_remove_file_locks(sysid=0x%x)\n", 8020 new_sysid)); 8021 (*lm_remove_file_locks)(new_sysid); 8022 } else { 8023 struct flock64 flk; 8024 8025 /* Release all locks for this client */ 8026 flk.l_type = F_UNLKSYS; 8027 flk.l_whence = 0; 8028 flk.l_start = 0; 8029 flk.l_len = 0; 8030 flk.l_sysid = sp->owner->client->sysidt; 8031 flk.l_pid = 0; 8032 (void) VOP_FRLOCK(sp->finfo->vp, F_SETLK, &flk, 8033 F_REMOTELOCK | FREAD | FWRITE, 8034 (u_offset_t)0, NULL, CRED(), NULL); 8035 } 8036 8037 sp->owner->client->unlksys_completed = TRUE; 8038 } 8039 } 8040 8041 /* 8042 * Release all locks on this file by this lock owner or at 8043 * least mark the locks as having been released 8044 */ 8045 for (lsp = sp->lockownerlist.next->lsp; lsp != NULL; 8046 lsp = lsp->lockownerlist.next->lsp) { 8047 8048 lsp->locks_cleaned = TRUE; 8049 8050 /* Was this already taken care of above? */ 8051 if (!close_of_client && 8052 sp->owner->client->sysidt != LM_NOSYSID) 8053 (void) cleanlocks(sp->finfo->vp, lsp->locker->pid, 8054 lsp->locker->client->sysidt); 8055 } 8056 8057 /* 8058 * Release any shrlocks associated with this open state ID. 8059 * This must be done before the rfs4_state gets marked closed. 8060 */ 8061 if (sp->owner->client->sysidt != LM_NOSYSID) { 8062 shr.s_access = s_access; 8063 shr.s_deny = s_deny; 8064 shr.s_pid = rfs4_dbe_getid(sp->owner->dbe); 8065 shr.s_sysid = sp->owner->client->sysidt; 8066 shr_loco.sl_pid = shr.s_pid; 8067 shr_loco.sl_id = shr.s_sysid; 8068 shr.s_owner = (caddr_t)&shr_loco; 8069 shr.s_own_len = sizeof (shr_loco); 8070 (void) vop_shrlock(sp->finfo->vp, F_UNSHARE, &shr, fflags); 8071 } 8072 8073 (void) VOP_CLOSE(fp->vp, fflags, 1, (offset_t)0, cr, NULL); 8074 } 8075 8076 /* 8077 * lock_denied: Fill in a LOCK4deneid structure given an flock64 structure. 8078 */ 8079 static nfsstat4 8080 lock_denied(LOCK4denied *dp, struct flock64 *flk) 8081 { 8082 rfs4_lockowner_t *lo; 8083 rfs4_client_t *cp; 8084 uint32_t len; 8085 8086 lo = rfs4_findlockowner_by_pid(flk->l_pid); 8087 if (lo != NULL) { 8088 cp = lo->client; 8089 if (rfs4_lease_expired(cp)) { 8090 rfs4_lockowner_rele(lo); 8091 rfs4_dbe_hold(cp->dbe); 8092 rfs4_client_close(cp); 8093 return (NFS4ERR_EXPIRED); 8094 } 8095 dp->owner.clientid = lo->owner.clientid; 8096 len = lo->owner.owner_len; 8097 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8098 bcopy(lo->owner.owner_val, dp->owner.owner_val, len); 8099 dp->owner.owner_len = len; 8100 rfs4_lockowner_rele(lo); 8101 goto finish; 8102 } 8103 8104 /* 8105 * Its not a NFS4 lock. We take advantage that the upper 32 bits 8106 * of the client id contain the boot time for a NFS4 lock. So we 8107 * fabricate and identity by setting clientid to the sysid, and 8108 * the lock owner to the pid. 8109 */ 8110 dp->owner.clientid = flk->l_sysid; 8111 len = sizeof (pid_t); 8112 dp->owner.owner_len = len; 8113 dp->owner.owner_val = kmem_alloc(len, KM_SLEEP); 8114 bcopy(&flk->l_pid, dp->owner.owner_val, len); 8115 finish: 8116 dp->offset = flk->l_start; 8117 dp->length = flk->l_len; 8118 8119 if (flk->l_type == F_RDLCK) 8120 dp->locktype = READ_LT; 8121 else if (flk->l_type == F_WRLCK) 8122 dp->locktype = WRITE_LT; 8123 else 8124 return (NFS4ERR_INVAL); /* no mapping from POSIX ltype to v4 */ 8125 8126 return (NFS4_OK); 8127 } 8128 8129 static int 8130 setlock(vnode_t *vp, struct flock64 *flock, int flag, cred_t *cred) 8131 { 8132 int error; 8133 struct flock64 flk; 8134 int i; 8135 clock_t delaytime; 8136 int cmd; 8137 8138 cmd = nbl_need_check(vp) ? F_SETLK_NBMAND : F_SETLK; 8139 retry: 8140 delaytime = MSEC_TO_TICK_ROUNDUP(rfs4_lock_delay); 8141 8142 for (i = 0; i < rfs4_maxlock_tries; i++) { 8143 LOCK_PRINT(rfs4_debug, "setlock", cmd, flock); 8144 error = VOP_FRLOCK(vp, cmd, 8145 flock, flag, (u_offset_t)0, NULL, cred, NULL); 8146 8147 if (error != EAGAIN && error != EACCES) 8148 break; 8149 8150 if (i < rfs4_maxlock_tries - 1) { 8151 delay(delaytime); 8152 delaytime *= 2; 8153 } 8154 } 8155 8156 if (error == EAGAIN || error == EACCES) { 8157 /* Get the owner of the lock */ 8158 flk = *flock; 8159 LOCK_PRINT(rfs4_debug, "setlock", F_GETLK, &flk); 8160 if (VOP_FRLOCK(vp, F_GETLK, &flk, flag, 8161 (u_offset_t)0, NULL, cred, NULL) == 0) { 8162 if (flk.l_type == F_UNLCK) { 8163 /* No longer locked, retry */ 8164 goto retry; 8165 } 8166 *flock = flk; 8167 LOCK_PRINT(rfs4_debug, "setlock(blocking lock)", 8168 F_GETLK, &flk); 8169 } 8170 } 8171 8172 return (error); 8173 } 8174 8175 /*ARGSUSED*/ 8176 static nfsstat4 8177 rfs4_do_lock(rfs4_lo_state_t *lp, nfs_lock_type4 locktype, 8178 seqid4 seqid, offset4 offset, 8179 length4 length, cred_t *cred, nfs_resop4 *resop) 8180 { 8181 nfsstat4 status; 8182 rfs4_lockowner_t *lo = lp->locker; 8183 rfs4_state_t *sp = lp->state; 8184 struct flock64 flock; 8185 int16_t ltype; 8186 int flag; 8187 int error; 8188 sysid_t sysid; 8189 LOCK4res *lres; 8190 8191 if (rfs4_lease_expired(lo->client)) { 8192 return (NFS4ERR_EXPIRED); 8193 } 8194 8195 if ((status = rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK) 8196 return (status); 8197 8198 /* Check for zero length. To lock to end of file use all ones for V4 */ 8199 if (length == 0) 8200 return (NFS4ERR_INVAL); 8201 else if (length == (length4)(~0)) 8202 length = 0; /* Posix to end of file */ 8203 8204 retry: 8205 rfs4_dbe_lock(sp->dbe); 8206 8207 8208 if (resop->resop != OP_LOCKU) { 8209 switch (locktype) { 8210 case READ_LT: 8211 case READW_LT: 8212 if ((sp->share_access 8213 & OPEN4_SHARE_ACCESS_READ) == 0) { 8214 rfs4_dbe_unlock(sp->dbe); 8215 8216 return (NFS4ERR_OPENMODE); 8217 } 8218 ltype = F_RDLCK; 8219 break; 8220 case WRITE_LT: 8221 case WRITEW_LT: 8222 if ((sp->share_access 8223 & OPEN4_SHARE_ACCESS_WRITE) == 0) { 8224 rfs4_dbe_unlock(sp->dbe); 8225 8226 return (NFS4ERR_OPENMODE); 8227 } 8228 ltype = F_WRLCK; 8229 break; 8230 } 8231 } else 8232 ltype = F_UNLCK; 8233 8234 flock.l_type = ltype; 8235 flock.l_whence = 0; /* SEEK_SET */ 8236 flock.l_start = offset; 8237 flock.l_len = length; 8238 flock.l_sysid = sysid; 8239 flock.l_pid = lp->locker->pid; 8240 8241 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 8242 if (flock.l_len < 0 || flock.l_start < 0) { 8243 rfs4_dbe_unlock(sp->dbe); 8244 return (NFS4ERR_INVAL); 8245 } 8246 8247 /* 8248 * N.B. FREAD has the same value as OPEN4_SHARE_ACCESS_READ and 8249 * FWRITE has the same value as OPEN4_SHARE_ACCESS_WRITE. 8250 */ 8251 flag = (int)sp->share_access | F_REMOTELOCK; 8252 8253 error = setlock(sp->finfo->vp, &flock, flag, cred); 8254 if (error == 0) { 8255 rfs4_dbe_lock(lp->dbe); 8256 next_stateid(&lp->lockid); 8257 rfs4_dbe_unlock(lp->dbe); 8258 } 8259 8260 rfs4_dbe_unlock(sp->dbe); 8261 8262 /* 8263 * N.B. We map error values to nfsv4 errors. This is differrent 8264 * than puterrno4 routine. 8265 */ 8266 switch (error) { 8267 case 0: 8268 status = NFS4_OK; 8269 break; 8270 case EAGAIN: 8271 case EACCES: /* Old value */ 8272 /* Can only get here if op is OP_LOCK */ 8273 ASSERT(resop->resop == OP_LOCK); 8274 lres = &resop->nfs_resop4_u.oplock; 8275 status = NFS4ERR_DENIED; 8276 if (lock_denied(&lres->LOCK4res_u.denied, &flock) 8277 == NFS4ERR_EXPIRED) 8278 goto retry; 8279 break; 8280 case ENOLCK: 8281 status = NFS4ERR_DELAY; 8282 break; 8283 case EOVERFLOW: 8284 status = NFS4ERR_INVAL; 8285 break; 8286 case EINVAL: 8287 status = NFS4ERR_NOTSUPP; 8288 break; 8289 default: 8290 cmn_err(CE_WARN, "rfs4_do_lock: unexpected errno (%d)", error); 8291 status = NFS4ERR_SERVERFAULT; 8292 break; 8293 } 8294 8295 return (status); 8296 } 8297 8298 /*ARGSUSED*/ 8299 void 8300 rfs4_op_lock(nfs_argop4 *argop, nfs_resop4 *resop, 8301 struct svc_req *req, struct compound_state *cs) 8302 { 8303 LOCK4args *args = &argop->nfs_argop4_u.oplock; 8304 LOCK4res *resp = &resop->nfs_resop4_u.oplock; 8305 nfsstat4 status; 8306 stateid4 *stateid; 8307 rfs4_lockowner_t *lo; 8308 rfs4_client_t *cp; 8309 rfs4_state_t *sp = NULL; 8310 rfs4_lo_state_t *lsp = NULL; 8311 bool_t ls_sw_held = FALSE; 8312 bool_t create = TRUE; 8313 bool_t lcreate = TRUE; 8314 bool_t dup_lock = FALSE; 8315 int rc; 8316 8317 DTRACE_NFSV4_2(op__lock__start, struct compound_state *, cs, 8318 LOCK4args *, args); 8319 8320 if (cs->vp == NULL) { 8321 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8322 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8323 cs, LOCK4res *, resp); 8324 return; 8325 } 8326 8327 if (args->locker.new_lock_owner) { 8328 /* Create a new lockowner for this instance */ 8329 open_to_lock_owner4 *olo = &args->locker.locker4_u.open_owner; 8330 8331 NFS4_DEBUG(rfs4_debug, (CE_NOTE, "Creating new lock owner")); 8332 8333 stateid = &olo->open_stateid; 8334 status = rfs4_get_state(stateid, &sp, RFS4_DBS_VALID); 8335 if (status != NFS4_OK) { 8336 NFS4_DEBUG(rfs4_debug, 8337 (CE_NOTE, "Get state failed in lock %d", status)); 8338 *cs->statusp = resp->status = status; 8339 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8340 cs, LOCK4res *, resp); 8341 return; 8342 } 8343 8344 /* Ensure specified filehandle matches */ 8345 if (cs->vp != sp->finfo->vp) { 8346 rfs4_state_rele(sp); 8347 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8348 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8349 cs, LOCK4res *, resp); 8350 return; 8351 } 8352 8353 /* hold off other access to open_owner while we tinker */ 8354 rfs4_sw_enter(&sp->owner->oo_sw); 8355 8356 switch (rc = rfs4_check_stateid_seqid(sp, stateid)) { 8357 case NFS4_CHECK_STATEID_OLD: 8358 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8359 goto end; 8360 case NFS4_CHECK_STATEID_BAD: 8361 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8362 goto end; 8363 case NFS4_CHECK_STATEID_EXPIRED: 8364 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8365 goto end; 8366 case NFS4_CHECK_STATEID_UNCONFIRMED: 8367 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8368 goto end; 8369 case NFS4_CHECK_STATEID_CLOSED: 8370 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8371 goto end; 8372 case NFS4_CHECK_STATEID_OKAY: 8373 case NFS4_CHECK_STATEID_REPLAY: 8374 switch (rfs4_check_olo_seqid(olo->open_seqid, 8375 sp->owner, resop)) { 8376 case NFS4_CHKSEQ_OKAY: 8377 if (rc == NFS4_CHECK_STATEID_OKAY) 8378 break; 8379 /* 8380 * This is replayed stateid; if seqid 8381 * matches next expected, then client 8382 * is using wrong seqid. 8383 */ 8384 /* FALLTHROUGH */ 8385 case NFS4_CHKSEQ_BAD: 8386 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8387 goto end; 8388 case NFS4_CHKSEQ_REPLAY: 8389 /* This is a duplicate LOCK request */ 8390 dup_lock = TRUE; 8391 8392 /* 8393 * For a duplicate we do not want to 8394 * create a new lockowner as it should 8395 * already exist. 8396 * Turn off the lockowner create flag. 8397 */ 8398 lcreate = FALSE; 8399 } 8400 break; 8401 } 8402 8403 lo = rfs4_findlockowner(&olo->lock_owner, &lcreate); 8404 if (lo == NULL) { 8405 NFS4_DEBUG(rfs4_debug, 8406 (CE_NOTE, "rfs4_op_lock: no lock owner")); 8407 *cs->statusp = resp->status = NFS4ERR_RESOURCE; 8408 goto end; 8409 } 8410 8411 lsp = rfs4_findlo_state_by_owner(lo, sp, &create); 8412 if (lsp == NULL) { 8413 rfs4_update_lease(sp->owner->client); 8414 /* 8415 * Only update theh open_seqid if this is not 8416 * a duplicate request 8417 */ 8418 if (dup_lock == FALSE) { 8419 rfs4_update_open_sequence(sp->owner); 8420 } 8421 8422 NFS4_DEBUG(rfs4_debug, 8423 (CE_NOTE, "rfs4_op_lock: no state")); 8424 *cs->statusp = resp->status = NFS4ERR_SERVERFAULT; 8425 rfs4_update_open_resp(sp->owner, resop, NULL); 8426 rfs4_lockowner_rele(lo); 8427 goto end; 8428 } 8429 8430 /* 8431 * This is the new_lock_owner branch and the client is 8432 * supposed to be associating a new lock_owner with 8433 * the open file at this point. If we find that a 8434 * lock_owner/state association already exists and a 8435 * successful LOCK request was returned to the client, 8436 * an error is returned to the client since this is 8437 * not appropriate. The client should be using the 8438 * existing lock_owner branch. 8439 */ 8440 if (dup_lock == FALSE && create == FALSE) { 8441 if (lsp->lock_completed == TRUE) { 8442 *cs->statusp = 8443 resp->status = NFS4ERR_BAD_SEQID; 8444 rfs4_lockowner_rele(lo); 8445 goto end; 8446 } 8447 } 8448 8449 rfs4_update_lease(sp->owner->client); 8450 8451 /* 8452 * Only update theh open_seqid if this is not 8453 * a duplicate request 8454 */ 8455 if (dup_lock == FALSE) { 8456 rfs4_update_open_sequence(sp->owner); 8457 } 8458 8459 /* 8460 * If this is a duplicate lock request, just copy the 8461 * previously saved reply and return. 8462 */ 8463 if (dup_lock == TRUE) { 8464 /* verify that lock_seqid's match */ 8465 if (lsp->seqid != olo->lock_seqid) { 8466 NFS4_DEBUG(rfs4_debug, 8467 (CE_NOTE, "rfs4_op_lock: Dup-Lock seqid bad" 8468 "lsp->seqid=%d old->seqid=%d", 8469 lsp->seqid, olo->lock_seqid)); 8470 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8471 } else { 8472 rfs4_copy_reply(resop, lsp->reply); 8473 /* 8474 * Make sure to copy the just 8475 * retrieved reply status into the 8476 * overall compound status 8477 */ 8478 *cs->statusp = resp->status; 8479 } 8480 rfs4_lockowner_rele(lo); 8481 goto end; 8482 } 8483 8484 rfs4_dbe_lock(lsp->dbe); 8485 8486 /* Make sure to update the lock sequence id */ 8487 lsp->seqid = olo->lock_seqid; 8488 8489 NFS4_DEBUG(rfs4_debug, 8490 (CE_NOTE, "Lock seqid established as %d", lsp->seqid)); 8491 8492 /* 8493 * This is used to signify the newly created lockowner 8494 * stateid and its sequence number. The checks for 8495 * sequence number and increment don't occur on the 8496 * very first lock request for a lockowner. 8497 */ 8498 lsp->skip_seqid_check = TRUE; 8499 8500 /* hold off other access to lsp while we tinker */ 8501 rfs4_sw_enter(&lsp->ls_sw); 8502 ls_sw_held = TRUE; 8503 8504 rfs4_dbe_unlock(lsp->dbe); 8505 8506 rfs4_lockowner_rele(lo); 8507 } else { 8508 stateid = &args->locker.locker4_u.lock_owner.lock_stateid; 8509 /* get lsp and hold the lock on the underlying file struct */ 8510 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) 8511 != NFS4_OK) { 8512 *cs->statusp = resp->status = status; 8513 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8514 cs, LOCK4res *, resp); 8515 return; 8516 } 8517 create = FALSE; /* We didn't create lsp */ 8518 8519 /* Ensure specified filehandle matches */ 8520 if (cs->vp != lsp->state->finfo->vp) { 8521 rfs4_lo_state_rele(lsp, TRUE); 8522 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8523 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, 8524 cs, LOCK4res *, resp); 8525 return; 8526 } 8527 8528 /* hold off other access to lsp while we tinker */ 8529 rfs4_sw_enter(&lsp->ls_sw); 8530 ls_sw_held = TRUE; 8531 8532 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 8533 /* 8534 * The stateid looks like it was okay (expected to be 8535 * the next one) 8536 */ 8537 case NFS4_CHECK_STATEID_OKAY: 8538 /* 8539 * The sequence id is now checked. Determine 8540 * if this is a replay or if it is in the 8541 * expected (next) sequence. In the case of a 8542 * replay, there are two replay conditions 8543 * that may occur. The first is the normal 8544 * condition where a LOCK is done with a 8545 * NFS4_OK response and the stateid is 8546 * updated. That case is handled below when 8547 * the stateid is identified as a REPLAY. The 8548 * second is the case where an error is 8549 * returned, like NFS4ERR_DENIED, and the 8550 * sequence number is updated but the stateid 8551 * is not updated. This second case is dealt 8552 * with here. So it may seem odd that the 8553 * stateid is okay but the sequence id is a 8554 * replay but it is okay. 8555 */ 8556 switch (rfs4_check_lock_seqid( 8557 args->locker.locker4_u.lock_owner.lock_seqid, 8558 lsp, resop)) { 8559 case NFS4_CHKSEQ_REPLAY: 8560 if (resp->status != NFS4_OK) { 8561 /* 8562 * Here is our replay and need 8563 * to verify that the last 8564 * response was an error. 8565 */ 8566 *cs->statusp = resp->status; 8567 goto end; 8568 } 8569 /* 8570 * This is done since the sequence id 8571 * looked like a replay but it didn't 8572 * pass our check so a BAD_SEQID is 8573 * returned as a result. 8574 */ 8575 /*FALLTHROUGH*/ 8576 case NFS4_CHKSEQ_BAD: 8577 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8578 goto end; 8579 case NFS4_CHKSEQ_OKAY: 8580 /* Everything looks okay move ahead */ 8581 break; 8582 } 8583 break; 8584 case NFS4_CHECK_STATEID_OLD: 8585 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8586 goto end; 8587 case NFS4_CHECK_STATEID_BAD: 8588 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8589 goto end; 8590 case NFS4_CHECK_STATEID_EXPIRED: 8591 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8592 goto end; 8593 case NFS4_CHECK_STATEID_CLOSED: 8594 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8595 goto end; 8596 case NFS4_CHECK_STATEID_REPLAY: 8597 switch (rfs4_check_lock_seqid( 8598 args->locker.locker4_u.lock_owner.lock_seqid, 8599 lsp, resop)) { 8600 case NFS4_CHKSEQ_OKAY: 8601 /* 8602 * This is a replayed stateid; if 8603 * seqid matches the next expected, 8604 * then client is using wrong seqid. 8605 */ 8606 case NFS4_CHKSEQ_BAD: 8607 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8608 goto end; 8609 case NFS4_CHKSEQ_REPLAY: 8610 rfs4_update_lease(lsp->locker->client); 8611 *cs->statusp = status = resp->status; 8612 goto end; 8613 } 8614 break; 8615 default: 8616 ASSERT(FALSE); 8617 break; 8618 } 8619 8620 rfs4_update_lock_sequence(lsp); 8621 rfs4_update_lease(lsp->locker->client); 8622 } 8623 8624 /* 8625 * NFS4 only allows locking on regular files, so 8626 * verify type of object. 8627 */ 8628 if (cs->vp->v_type != VREG) { 8629 if (cs->vp->v_type == VDIR) 8630 status = NFS4ERR_ISDIR; 8631 else 8632 status = NFS4ERR_INVAL; 8633 goto out; 8634 } 8635 8636 cp = lsp->state->owner->client; 8637 8638 if (rfs4_clnt_in_grace(cp) && !args->reclaim) { 8639 status = NFS4ERR_GRACE; 8640 goto out; 8641 } 8642 8643 if (rfs4_clnt_in_grace(cp) && args->reclaim && !cp->can_reclaim) { 8644 status = NFS4ERR_NO_GRACE; 8645 goto out; 8646 } 8647 8648 if (!rfs4_clnt_in_grace(cp) && args->reclaim) { 8649 status = NFS4ERR_NO_GRACE; 8650 goto out; 8651 } 8652 8653 if (lsp->state->finfo->dinfo->dtype == OPEN_DELEGATE_WRITE) 8654 cs->deleg = TRUE; 8655 8656 status = rfs4_do_lock(lsp, args->locktype, 8657 args->locker.locker4_u.lock_owner.lock_seqid, 8658 args->offset, args->length, cs->cr, resop); 8659 8660 out: 8661 lsp->skip_seqid_check = FALSE; 8662 8663 *cs->statusp = resp->status = status; 8664 8665 if (status == NFS4_OK) { 8666 resp->LOCK4res_u.lock_stateid = lsp->lockid.stateid; 8667 lsp->lock_completed = TRUE; 8668 } 8669 /* 8670 * Only update the "OPEN" response here if this was a new 8671 * lock_owner 8672 */ 8673 if (sp) 8674 rfs4_update_open_resp(sp->owner, resop, NULL); 8675 8676 rfs4_update_lock_resp(lsp, resop); 8677 8678 end: 8679 if (lsp) { 8680 if (ls_sw_held) 8681 rfs4_sw_exit(&lsp->ls_sw); 8682 /* 8683 * If an sp obtained, then the lsp does not represent 8684 * a lock on the file struct. 8685 */ 8686 if (sp != NULL) 8687 rfs4_lo_state_rele(lsp, FALSE); 8688 else 8689 rfs4_lo_state_rele(lsp, TRUE); 8690 } 8691 if (sp) { 8692 rfs4_sw_exit(&sp->owner->oo_sw); 8693 rfs4_state_rele(sp); 8694 } 8695 8696 DTRACE_NFSV4_2(op__lock__done, struct compound_state *, cs, 8697 LOCK4res *, resp); 8698 } 8699 8700 /* free function for LOCK/LOCKT */ 8701 static void 8702 lock_denied_free(nfs_resop4 *resop) 8703 { 8704 LOCK4denied *dp = NULL; 8705 8706 switch (resop->resop) { 8707 case OP_LOCK: 8708 if (resop->nfs_resop4_u.oplock.status == NFS4ERR_DENIED) 8709 dp = &resop->nfs_resop4_u.oplock.LOCK4res_u.denied; 8710 break; 8711 case OP_LOCKT: 8712 if (resop->nfs_resop4_u.oplockt.status == NFS4ERR_DENIED) 8713 dp = &resop->nfs_resop4_u.oplockt.denied; 8714 break; 8715 default: 8716 break; 8717 } 8718 8719 if (dp) 8720 kmem_free(dp->owner.owner_val, dp->owner.owner_len); 8721 } 8722 8723 /*ARGSUSED*/ 8724 void 8725 rfs4_op_locku(nfs_argop4 *argop, nfs_resop4 *resop, 8726 struct svc_req *req, struct compound_state *cs) 8727 { 8728 LOCKU4args *args = &argop->nfs_argop4_u.oplocku; 8729 LOCKU4res *resp = &resop->nfs_resop4_u.oplocku; 8730 nfsstat4 status; 8731 stateid4 *stateid = &args->lock_stateid; 8732 rfs4_lo_state_t *lsp; 8733 8734 DTRACE_NFSV4_2(op__locku__start, struct compound_state *, cs, 8735 LOCKU4args *, args); 8736 8737 if (cs->vp == NULL) { 8738 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8739 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 8740 LOCKU4res *, resp); 8741 return; 8742 } 8743 8744 if ((status = rfs4_get_lo_state(stateid, &lsp, TRUE)) != NFS4_OK) { 8745 *cs->statusp = resp->status = status; 8746 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 8747 LOCKU4res *, resp); 8748 return; 8749 } 8750 8751 /* Ensure specified filehandle matches */ 8752 if (cs->vp != lsp->state->finfo->vp) { 8753 rfs4_lo_state_rele(lsp, TRUE); 8754 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8755 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 8756 LOCKU4res *, resp); 8757 return; 8758 } 8759 8760 /* hold off other access to lsp while we tinker */ 8761 rfs4_sw_enter(&lsp->ls_sw); 8762 8763 switch (rfs4_check_lo_stateid_seqid(lsp, stateid)) { 8764 case NFS4_CHECK_STATEID_OKAY: 8765 if (rfs4_check_lock_seqid(args->seqid, lsp, resop) 8766 != NFS4_CHKSEQ_OKAY) { 8767 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8768 goto end; 8769 } 8770 break; 8771 case NFS4_CHECK_STATEID_OLD: 8772 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8773 goto end; 8774 case NFS4_CHECK_STATEID_BAD: 8775 *cs->statusp = resp->status = NFS4ERR_BAD_STATEID; 8776 goto end; 8777 case NFS4_CHECK_STATEID_EXPIRED: 8778 *cs->statusp = resp->status = NFS4ERR_EXPIRED; 8779 goto end; 8780 case NFS4_CHECK_STATEID_CLOSED: 8781 *cs->statusp = resp->status = NFS4ERR_OLD_STATEID; 8782 goto end; 8783 case NFS4_CHECK_STATEID_REPLAY: 8784 switch (rfs4_check_lock_seqid(args->seqid, lsp, resop)) { 8785 case NFS4_CHKSEQ_OKAY: 8786 /* 8787 * This is a replayed stateid; if 8788 * seqid matches the next expected, 8789 * then client is using wrong seqid. 8790 */ 8791 case NFS4_CHKSEQ_BAD: 8792 *cs->statusp = resp->status = NFS4ERR_BAD_SEQID; 8793 goto end; 8794 case NFS4_CHKSEQ_REPLAY: 8795 rfs4_update_lease(lsp->locker->client); 8796 *cs->statusp = status = resp->status; 8797 goto end; 8798 } 8799 break; 8800 default: 8801 ASSERT(FALSE); 8802 break; 8803 } 8804 8805 rfs4_update_lock_sequence(lsp); 8806 rfs4_update_lease(lsp->locker->client); 8807 8808 /* 8809 * NFS4 only allows locking on regular files, so 8810 * verify type of object. 8811 */ 8812 if (cs->vp->v_type != VREG) { 8813 if (cs->vp->v_type == VDIR) 8814 status = NFS4ERR_ISDIR; 8815 else 8816 status = NFS4ERR_INVAL; 8817 goto out; 8818 } 8819 8820 if (rfs4_clnt_in_grace(lsp->state->owner->client)) { 8821 status = NFS4ERR_GRACE; 8822 goto out; 8823 } 8824 8825 status = rfs4_do_lock(lsp, args->locktype, 8826 args->seqid, args->offset, args->length, cs->cr, resop); 8827 8828 out: 8829 *cs->statusp = resp->status = status; 8830 8831 if (status == NFS4_OK) 8832 resp->lock_stateid = lsp->lockid.stateid; 8833 8834 rfs4_update_lock_resp(lsp, resop); 8835 8836 end: 8837 rfs4_sw_exit(&lsp->ls_sw); 8838 rfs4_lo_state_rele(lsp, TRUE); 8839 8840 DTRACE_NFSV4_2(op__locku__done, struct compound_state *, cs, 8841 LOCKU4res *, resp); 8842 } 8843 8844 /* 8845 * LOCKT is a best effort routine, the client can not be guaranteed that 8846 * the status return is still in effect by the time the reply is received. 8847 * They are numerous race conditions in this routine, but we are not required 8848 * and can not be accurate. 8849 */ 8850 /*ARGSUSED*/ 8851 void 8852 rfs4_op_lockt(nfs_argop4 *argop, nfs_resop4 *resop, 8853 struct svc_req *req, struct compound_state *cs) 8854 { 8855 LOCKT4args *args = &argop->nfs_argop4_u.oplockt; 8856 LOCKT4res *resp = &resop->nfs_resop4_u.oplockt; 8857 rfs4_lockowner_t *lo; 8858 rfs4_client_t *cp; 8859 bool_t create = FALSE; 8860 struct flock64 flk; 8861 int error; 8862 int flag = FREAD | FWRITE; 8863 int ltype; 8864 length4 posix_length; 8865 sysid_t sysid; 8866 pid_t pid; 8867 8868 DTRACE_NFSV4_2(op__lockt__start, struct compound_state *, cs, 8869 LOCKT4args *, args); 8870 8871 if (cs->vp == NULL) { 8872 *cs->statusp = resp->status = NFS4ERR_NOFILEHANDLE; 8873 goto out; 8874 } 8875 8876 /* 8877 * NFS4 only allows locking on regular files, so 8878 * verify type of object. 8879 */ 8880 if (cs->vp->v_type != VREG) { 8881 if (cs->vp->v_type == VDIR) 8882 *cs->statusp = resp->status = NFS4ERR_ISDIR; 8883 else 8884 *cs->statusp = resp->status = NFS4ERR_INVAL; 8885 goto out; 8886 } 8887 8888 /* 8889 * Check out the clientid to ensure the server knows about it 8890 * so that we correctly inform the client of a server reboot. 8891 */ 8892 if ((cp = rfs4_findclient_by_id(args->owner.clientid, FALSE)) 8893 == NULL) { 8894 *cs->statusp = resp->status = 8895 rfs4_check_clientid(&args->owner.clientid, 0); 8896 goto out; 8897 } 8898 if (rfs4_lease_expired(cp)) { 8899 rfs4_client_close(cp); 8900 /* 8901 * Protocol doesn't allow returning NFS4ERR_STALE as 8902 * other operations do on this check so STALE_CLIENTID 8903 * is returned instead 8904 */ 8905 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID; 8906 goto out; 8907 } 8908 8909 if (rfs4_clnt_in_grace(cp) && !(cp->can_reclaim)) { 8910 *cs->statusp = resp->status = NFS4ERR_GRACE; 8911 rfs4_client_rele(cp); 8912 goto out; 8913 } 8914 rfs4_client_rele(cp); 8915 8916 resp->status = NFS4_OK; 8917 8918 switch (args->locktype) { 8919 case READ_LT: 8920 case READW_LT: 8921 ltype = F_RDLCK; 8922 break; 8923 case WRITE_LT: 8924 case WRITEW_LT: 8925 ltype = F_WRLCK; 8926 break; 8927 } 8928 8929 posix_length = args->length; 8930 /* Check for zero length. To lock to end of file use all ones for V4 */ 8931 if (posix_length == 0) { 8932 *cs->statusp = resp->status = NFS4ERR_INVAL; 8933 goto out; 8934 } else if (posix_length == (length4)(~0)) { 8935 posix_length = 0; /* Posix to end of file */ 8936 } 8937 8938 /* Find or create a lockowner */ 8939 lo = rfs4_findlockowner(&args->owner, &create); 8940 8941 if (lo) { 8942 pid = lo->pid; 8943 if ((resp->status = 8944 rfs4_client_sysid(lo->client, &sysid)) != NFS4_OK) 8945 goto err; 8946 } else { 8947 pid = 0; 8948 sysid = lockt_sysid; 8949 } 8950 retry: 8951 flk.l_type = ltype; 8952 flk.l_whence = 0; /* SEEK_SET */ 8953 flk.l_start = args->offset; 8954 flk.l_len = posix_length; 8955 flk.l_sysid = sysid; 8956 flk.l_pid = pid; 8957 flag |= F_REMOTELOCK; 8958 8959 LOCK_PRINT(rfs4_debug, "rfs4_op_lockt", F_GETLK, &flk); 8960 8961 /* Note that length4 is uint64_t but l_len and l_start are off64_t */ 8962 if (flk.l_len < 0 || flk.l_start < 0) { 8963 resp->status = NFS4ERR_INVAL; 8964 goto err; 8965 } 8966 error = VOP_FRLOCK(cs->vp, F_GETLK, &flk, flag, (u_offset_t)0, 8967 NULL, cs->cr, NULL); 8968 8969 /* 8970 * N.B. We map error values to nfsv4 errors. This is differrent 8971 * than puterrno4 routine. 8972 */ 8973 switch (error) { 8974 case 0: 8975 if (flk.l_type == F_UNLCK) 8976 resp->status = NFS4_OK; 8977 else { 8978 if (lock_denied(&resp->denied, &flk) == NFS4ERR_EXPIRED) 8979 goto retry; 8980 resp->status = NFS4ERR_DENIED; 8981 } 8982 break; 8983 case EOVERFLOW: 8984 resp->status = NFS4ERR_INVAL; 8985 break; 8986 case EINVAL: 8987 resp->status = NFS4ERR_NOTSUPP; 8988 break; 8989 default: 8990 cmn_err(CE_WARN, "rfs4_op_lockt: unexpected errno (%d)", 8991 error); 8992 resp->status = NFS4ERR_SERVERFAULT; 8993 break; 8994 } 8995 8996 err: 8997 if (lo) 8998 rfs4_lockowner_rele(lo); 8999 *cs->statusp = resp->status; 9000 out: 9001 DTRACE_NFSV4_2(op__lockt__done, struct compound_state *, cs, 9002 LOCKT4res *, resp); 9003 } 9004 9005 static int 9006 vop_shrlock(vnode_t *vp, int cmd, struct shrlock *sp, int fflags) 9007 { 9008 int err; 9009 9010 if (cmd == F_UNSHARE && sp->s_deny == 0 && sp->s_access == 0) 9011 return (0); 9012 9013 err = VOP_SHRLOCK(vp, cmd, sp, fflags, CRED(), NULL); 9014 9015 NFS4_DEBUG(rfs4_shrlock_debug, 9016 (CE_NOTE, "rfs4_shrlock %s vp=%p acc=%d dny=%d sysid=%d " 9017 "pid=%d err=%d\n", cmd == F_UNSHARE ? "UNSHR" : "SHARE", 9018 (void *) vp, sp->s_access, sp->s_deny, sp->s_sysid, sp->s_pid, 9019 err)); 9020 9021 return (err); 9022 } 9023 9024 static int 9025 rfs4_shrlock(rfs4_state_t *sp, int cmd) 9026 { 9027 struct shrlock shr; 9028 struct shr_locowner shr_loco; 9029 int fflags; 9030 9031 fflags = shr.s_access = shr.s_deny = 0; 9032 9033 if (sp->share_access & OPEN4_SHARE_ACCESS_READ) { 9034 fflags |= FREAD; 9035 shr.s_access |= F_RDACC; 9036 } 9037 if (sp->share_access & OPEN4_SHARE_ACCESS_WRITE) { 9038 fflags |= FWRITE; 9039 shr.s_access |= F_WRACC; 9040 } 9041 if (sp->share_deny & OPEN4_SHARE_DENY_READ) 9042 shr.s_deny |= F_RDDNY; 9043 if (sp->share_deny & OPEN4_SHARE_DENY_WRITE) 9044 shr.s_deny |= F_WRDNY; 9045 9046 shr.s_pid = rfs4_dbe_getid(sp->owner->dbe); 9047 shr.s_sysid = sp->owner->client->sysidt; 9048 shr_loco.sl_pid = shr.s_pid; 9049 shr_loco.sl_id = shr.s_sysid; 9050 shr.s_owner = (caddr_t)&shr_loco; 9051 shr.s_own_len = sizeof (shr_loco); 9052 return (vop_shrlock(sp->finfo->vp, cmd, &shr, fflags)); 9053 } 9054 9055 static int 9056 rfs4_share(rfs4_state_t *sp) 9057 { 9058 int cmd; 9059 cmd = nbl_need_check(sp->finfo->vp) ? F_SHARE_NBMAND : F_SHARE; 9060 return (rfs4_shrlock(sp, cmd)); 9061 } 9062 9063 void 9064 rfs4_unshare(rfs4_state_t *sp) 9065 { 9066 (void) rfs4_shrlock(sp, F_UNSHARE); 9067 } 9068